diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,9745 @@
+{
+  "best_metric": 0.2633333333333333,
+  "best_model_checkpoint": "swinv2-small-patch4-window16-256-mineral\\checkpoint-6693",
+  "epoch": 480.0,
+  "global_step": 9000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.53,
+      "learning_rate": 5.555555555555556e-07,
+      "loss": 5.6941,
+      "step": 10
+    },
+    {
+      "epoch": 0.96,
+      "eval_accuracy": 0.005,
+      "eval_loss": 5.692106246948242,
+      "eval_runtime": 9.9033,
+      "eval_samples_per_second": 60.586,
+      "eval_steps_per_second": 1.919,
+      "step": 18
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.1111111111111112e-06,
+      "loss": 5.6939,
+      "step": 20
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 5.6886,
+      "step": 30
+    },
+    {
+      "epoch": 1.97,
+      "eval_accuracy": 0.005,
+      "eval_loss": 5.682541370391846,
+      "eval_runtime": 4.1429,
+      "eval_samples_per_second": 144.826,
+      "eval_steps_per_second": 4.586,
+      "step": 37
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.2222222222222225e-06,
+      "loss": 5.6844,
+      "step": 40
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.777777777777778e-06,
+      "loss": 5.6735,
+      "step": 50
+    },
+    {
+      "epoch": 2.99,
+      "eval_accuracy": 0.005,
+      "eval_loss": 5.669071674346924,
+      "eval_runtime": 4.142,
+      "eval_samples_per_second": 144.858,
+      "eval_steps_per_second": 4.587,
+      "step": 56
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 5.6534,
+      "step": 60
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 3.888888888888889e-06,
+      "loss": 5.6521,
+      "step": 70
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.0033333333333333335,
+      "eval_loss": 5.654940605163574,
+      "eval_runtime": 4.1572,
+      "eval_samples_per_second": 144.326,
+      "eval_steps_per_second": 4.57,
+      "step": 75
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 4.444444444444445e-06,
+      "loss": 5.6431,
+      "step": 80
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5e-06,
+      "loss": 5.6394,
+      "step": 90
+    },
+    {
+      "epoch": 4.96,
+      "eval_accuracy": 0.0033333333333333335,
+      "eval_loss": 5.641611099243164,
+      "eval_runtime": 4.1535,
+      "eval_samples_per_second": 144.457,
+      "eval_steps_per_second": 4.574,
+      "step": 93
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 5.555555555555556e-06,
+      "loss": 5.5941,
+      "step": 100
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 6.111111111111111e-06,
+      "loss": 5.6078,
+      "step": 110
+    },
+    {
+      "epoch": 5.97,
+      "eval_accuracy": 0.0033333333333333335,
+      "eval_loss": 5.627758979797363,
+      "eval_runtime": 4.1893,
+      "eval_samples_per_second": 143.22,
+      "eval_steps_per_second": 4.535,
+      "step": 112
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 5.5762,
+      "step": 120
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 7.222222222222222e-06,
+      "loss": 5.5743,
+      "step": 130
+    },
+    {
+      "epoch": 6.99,
+      "eval_accuracy": 0.0016666666666666668,
+      "eval_loss": 5.612813472747803,
+      "eval_runtime": 4.224,
+      "eval_samples_per_second": 142.046,
+      "eval_steps_per_second": 4.498,
+      "step": 131
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 7.777777777777777e-06,
+      "loss": 5.5413,
+      "step": 140
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 5.5509,
+      "step": 150
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.0016666666666666668,
+      "eval_loss": 5.591813087463379,
+      "eval_runtime": 4.1747,
+      "eval_samples_per_second": 143.723,
+      "eval_steps_per_second": 4.551,
+      "step": 150
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 8.88888888888889e-06,
+      "loss": 5.5115,
+      "step": 160
+    },
+    {
+      "epoch": 8.96,
+      "eval_accuracy": 0.006666666666666667,
+      "eval_loss": 5.569559097290039,
+      "eval_runtime": 4.1739,
+      "eval_samples_per_second": 143.75,
+      "eval_steps_per_second": 4.552,
+      "step": 168
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 9.444444444444445e-06,
+      "loss": 5.4912,
+      "step": 170
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 1e-05,
+      "loss": 5.4411,
+      "step": 180
+    },
+    {
+      "epoch": 9.97,
+      "eval_accuracy": 0.01,
+      "eval_loss": 5.5439839363098145,
+      "eval_runtime": 4.1644,
+      "eval_samples_per_second": 144.078,
+      "eval_steps_per_second": 4.562,
+      "step": 187
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 1.0555555555555555e-05,
+      "loss": 5.3942,
+      "step": 190
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 1.1111111111111112e-05,
+      "loss": 5.3335,
+      "step": 200
+    },
+    {
+      "epoch": 10.99,
+      "eval_accuracy": 0.016666666666666666,
+      "eval_loss": 5.513484001159668,
+      "eval_runtime": 4.1932,
+      "eval_samples_per_second": 143.088,
+      "eval_steps_per_second": 4.531,
+      "step": 206
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 1.1666666666666668e-05,
+      "loss": 5.2998,
+      "step": 210
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 1.2222222222222222e-05,
+      "loss": 5.2413,
+      "step": 220
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.021666666666666667,
+      "eval_loss": 5.464037895202637,
+      "eval_runtime": 4.2106,
+      "eval_samples_per_second": 142.496,
+      "eval_steps_per_second": 4.512,
+      "step": 225
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.2777777777777777e-05,
+      "loss": 5.2175,
+      "step": 230
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 5.1738,
+      "step": 240
+    },
+    {
+      "epoch": 12.96,
+      "eval_accuracy": 0.03333333333333333,
+      "eval_loss": 5.408351421356201,
+      "eval_runtime": 4.1793,
+      "eval_samples_per_second": 143.566,
+      "eval_steps_per_second": 4.546,
+      "step": 243
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.388888888888889e-05,
+      "loss": 5.0966,
+      "step": 250
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.4444444444444444e-05,
+      "loss": 5.0222,
+      "step": 260
+    },
+    {
+      "epoch": 13.97,
+      "eval_accuracy": 0.045,
+      "eval_loss": 5.3320746421813965,
+      "eval_runtime": 4.1797,
+      "eval_samples_per_second": 143.55,
+      "eval_steps_per_second": 4.546,
+      "step": 262
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.5e-05,
+      "loss": 4.913,
+      "step": 270
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.5555555555555555e-05,
+      "loss": 4.8594,
+      "step": 280
+    },
+    {
+      "epoch": 14.99,
+      "eval_accuracy": 0.05333333333333334,
+      "eval_loss": 5.248490333557129,
+      "eval_runtime": 4.1704,
+      "eval_samples_per_second": 143.871,
+      "eval_steps_per_second": 4.556,
+      "step": 281
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.6111111111111115e-05,
+      "loss": 4.7591,
+      "step": 290
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 4.7441,
+      "step": 300
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.065,
+      "eval_loss": 5.150908946990967,
+      "eval_runtime": 4.1485,
+      "eval_samples_per_second": 144.629,
+      "eval_steps_per_second": 4.58,
+      "step": 300
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 1.7222222222222224e-05,
+      "loss": 4.5946,
+      "step": 310
+    },
+    {
+      "epoch": 16.96,
+      "eval_accuracy": 0.07166666666666667,
+      "eval_loss": 5.070082187652588,
+      "eval_runtime": 4.1543,
+      "eval_samples_per_second": 144.43,
+      "eval_steps_per_second": 4.574,
+      "step": 318
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 1.777777777777778e-05,
+      "loss": 4.4995,
+      "step": 320
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 1.8333333333333333e-05,
+      "loss": 4.3382,
+      "step": 330
+    },
+    {
+      "epoch": 17.97,
+      "eval_accuracy": 0.08666666666666667,
+      "eval_loss": 4.976734161376953,
+      "eval_runtime": 4.158,
+      "eval_samples_per_second": 144.301,
+      "eval_steps_per_second": 4.57,
+      "step": 337
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.888888888888889e-05,
+      "loss": 4.3477,
+      "step": 340
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.9444444444444445e-05,
+      "loss": 4.2008,
+      "step": 350
+    },
+    {
+      "epoch": 18.99,
+      "eval_accuracy": 0.105,
+      "eval_loss": 4.862234115600586,
+      "eval_runtime": 4.1637,
+      "eval_samples_per_second": 144.103,
+      "eval_steps_per_second": 4.563,
+      "step": 356
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2e-05,
+      "loss": 4.0491,
+      "step": 360
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 2.0555555555555555e-05,
+      "loss": 4.0563,
+      "step": 370
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.10333333333333333,
+      "eval_loss": 4.772627830505371,
+      "eval_runtime": 4.1762,
+      "eval_samples_per_second": 143.672,
+      "eval_steps_per_second": 4.55,
+      "step": 375
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 2.111111111111111e-05,
+      "loss": 3.8101,
+      "step": 380
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 2.1666666666666667e-05,
+      "loss": 3.8064,
+      "step": 390
+    },
+    {
+      "epoch": 20.96,
+      "eval_accuracy": 0.115,
+      "eval_loss": 4.689815044403076,
+      "eval_runtime": 4.151,
+      "eval_samples_per_second": 144.542,
+      "eval_steps_per_second": 4.577,
+      "step": 393
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 2.2222222222222223e-05,
+      "loss": 3.6083,
+      "step": 400
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 2.277777777777778e-05,
+      "loss": 3.5584,
+      "step": 410
+    },
+    {
+      "epoch": 21.97,
+      "eval_accuracy": 0.125,
+      "eval_loss": 4.599685192108154,
+      "eval_runtime": 4.1649,
+      "eval_samples_per_second": 144.062,
+      "eval_steps_per_second": 4.562,
+      "step": 412
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 2.3333333333333336e-05,
+      "loss": 3.4029,
+      "step": 420
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 2.3888888888888892e-05,
+      "loss": 3.3377,
+      "step": 430
+    },
+    {
+      "epoch": 22.99,
+      "eval_accuracy": 0.13666666666666666,
+      "eval_loss": 4.484786510467529,
+      "eval_runtime": 4.165,
+      "eval_samples_per_second": 144.057,
+      "eval_steps_per_second": 4.562,
+      "step": 431
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 2.4444444444444445e-05,
+      "loss": 3.1633,
+      "step": 440
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 2.5e-05,
+      "loss": 3.1119,
+      "step": 450
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.15333333333333332,
+      "eval_loss": 4.405167102813721,
+      "eval_runtime": 4.1901,
+      "eval_samples_per_second": 143.196,
+      "eval_steps_per_second": 4.535,
+      "step": 450
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 2.5555555555555554e-05,
+      "loss": 2.8686,
+      "step": 460
+    },
+    {
+      "epoch": 24.96,
+      "eval_accuracy": 0.15,
+      "eval_loss": 4.37052583694458,
+      "eval_runtime": 4.1543,
+      "eval_samples_per_second": 144.427,
+      "eval_steps_per_second": 4.574,
+      "step": 468
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 2.6111111111111114e-05,
+      "loss": 3.0138,
+      "step": 470
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 2.7649,
+      "step": 480
+    },
+    {
+      "epoch": 25.97,
+      "eval_accuracy": 0.165,
+      "eval_loss": 4.297973155975342,
+      "eval_runtime": 4.1559,
+      "eval_samples_per_second": 144.373,
+      "eval_steps_per_second": 4.572,
+      "step": 487
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 2.7222222222222223e-05,
+      "loss": 2.6887,
+      "step": 490
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 2.777777777777778e-05,
+      "loss": 2.5698,
+      "step": 500
+    },
+    {
+      "epoch": 26.99,
+      "eval_accuracy": 0.17666666666666667,
+      "eval_loss": 4.236337184906006,
+      "eval_runtime": 4.1581,
+      "eval_samples_per_second": 144.296,
+      "eval_steps_per_second": 4.569,
+      "step": 506
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 2.8333333333333335e-05,
+      "loss": 2.5301,
+      "step": 510
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 2.8888888888888888e-05,
+      "loss": 2.4344,
+      "step": 520
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.17666666666666667,
+      "eval_loss": 4.17328405380249,
+      "eval_runtime": 4.1589,
+      "eval_samples_per_second": 144.269,
+      "eval_steps_per_second": 4.569,
+      "step": 525
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 2.9444444444444448e-05,
+      "loss": 2.2843,
+      "step": 530
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 3e-05,
+      "loss": 2.2186,
+      "step": 540
+    },
+    {
+      "epoch": 28.96,
+      "eval_accuracy": 0.17333333333333334,
+      "eval_loss": 4.178333282470703,
+      "eval_runtime": 4.1547,
+      "eval_samples_per_second": 144.413,
+      "eval_steps_per_second": 4.573,
+      "step": 543
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 3.055555555555556e-05,
+      "loss": 2.1025,
+      "step": 550
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 3.111111111111111e-05,
+      "loss": 2.0227,
+      "step": 560
+    },
+    {
+      "epoch": 29.97,
+      "eval_accuracy": 0.18,
+      "eval_loss": 4.13058614730835,
+      "eval_runtime": 4.1618,
+      "eval_samples_per_second": 144.168,
+      "eval_steps_per_second": 4.565,
+      "step": 562
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 3.1666666666666666e-05,
+      "loss": 1.8851,
+      "step": 570
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 3.222222222222223e-05,
+      "loss": 1.9153,
+      "step": 580
+    },
+    {
+      "epoch": 30.99,
+      "eval_accuracy": 0.175,
+      "eval_loss": 4.094816207885742,
+      "eval_runtime": 4.1706,
+      "eval_samples_per_second": 143.864,
+      "eval_steps_per_second": 4.556,
+      "step": 581
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 3.277777777777778e-05,
+      "loss": 1.7959,
+      "step": 590
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.7363,
+      "step": 600
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.17833333333333334,
+      "eval_loss": 4.061172008514404,
+      "eval_runtime": 4.1527,
+      "eval_samples_per_second": 144.483,
+      "eval_steps_per_second": 4.575,
+      "step": 600
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 3.388888888888889e-05,
+      "loss": 1.6171,
+      "step": 610
+    },
+    {
+      "epoch": 32.96,
+      "eval_accuracy": 0.185,
+      "eval_loss": 4.020925521850586,
+      "eval_runtime": 4.1794,
+      "eval_samples_per_second": 143.562,
+      "eval_steps_per_second": 4.546,
+      "step": 618
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 3.444444444444445e-05,
+      "loss": 1.5284,
+      "step": 620
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 3.5e-05,
+      "loss": 1.4865,
+      "step": 630
+    },
+    {
+      "epoch": 33.97,
+      "eval_accuracy": 0.185,
+      "eval_loss": 4.019384860992432,
+      "eval_runtime": 4.1555,
+      "eval_samples_per_second": 144.386,
+      "eval_steps_per_second": 4.572,
+      "step": 637
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 3.555555555555556e-05,
+      "loss": 1.4216,
+      "step": 640
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 3.611111111111111e-05,
+      "loss": 1.3194,
+      "step": 650
+    },
+    {
+      "epoch": 34.99,
+      "eval_accuracy": 0.205,
+      "eval_loss": 3.988067626953125,
+      "eval_runtime": 4.1907,
+      "eval_samples_per_second": 143.174,
+      "eval_steps_per_second": 4.534,
+      "step": 656
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 3.6666666666666666e-05,
+      "loss": 1.3247,
+      "step": 660
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 3.722222222222222e-05,
+      "loss": 1.2811,
+      "step": 670
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.215,
+      "eval_loss": 3.9861950874328613,
+      "eval_runtime": 4.2139,
+      "eval_samples_per_second": 142.387,
+      "eval_steps_per_second": 4.509,
+      "step": 675
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 3.777777777777778e-05,
+      "loss": 1.1968,
+      "step": 680
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 3.8333333333333334e-05,
+      "loss": 1.1703,
+      "step": 690
+    },
+    {
+      "epoch": 36.96,
+      "eval_accuracy": 0.20333333333333334,
+      "eval_loss": 3.9904768466949463,
+      "eval_runtime": 4.165,
+      "eval_samples_per_second": 144.057,
+      "eval_steps_per_second": 4.562,
+      "step": 693
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 3.888888888888889e-05,
+      "loss": 1.1136,
+      "step": 700
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 3.944444444444445e-05,
+      "loss": 1.114,
+      "step": 710
+    },
+    {
+      "epoch": 37.97,
+      "eval_accuracy": 0.21333333333333335,
+      "eval_loss": 3.951385021209717,
+      "eval_runtime": 4.1675,
+      "eval_samples_per_second": 143.97,
+      "eval_steps_per_second": 4.559,
+      "step": 712
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 4e-05,
+      "loss": 1.0194,
+      "step": 720
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 4.055555555555556e-05,
+      "loss": 0.9645,
+      "step": 730
+    },
+    {
+      "epoch": 38.99,
+      "eval_accuracy": 0.20666666666666667,
+      "eval_loss": 3.9677815437316895,
+      "eval_runtime": 4.1678,
+      "eval_samples_per_second": 143.96,
+      "eval_steps_per_second": 4.559,
+      "step": 731
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 4.111111111111111e-05,
+      "loss": 0.9037,
+      "step": 740
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 4.166666666666667e-05,
+      "loss": 0.8976,
+      "step": 750
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.21666666666666667,
+      "eval_loss": 3.987384080886841,
+      "eval_runtime": 4.1794,
+      "eval_samples_per_second": 143.56,
+      "eval_steps_per_second": 4.546,
+      "step": 750
+    },
+    {
+      "epoch": 40.53,
+      "learning_rate": 4.222222222222222e-05,
+      "loss": 0.8147,
+      "step": 760
+    },
+    {
+      "epoch": 40.96,
+      "eval_accuracy": 0.20833333333333334,
+      "eval_loss": 3.925668954849243,
+      "eval_runtime": 4.2286,
+      "eval_samples_per_second": 141.892,
+      "eval_steps_per_second": 4.493,
+      "step": 768
+    },
+    {
+      "epoch": 41.07,
+      "learning_rate": 4.277777777777778e-05,
+      "loss": 0.8449,
+      "step": 770
+    },
+    {
+      "epoch": 41.6,
+      "learning_rate": 4.3333333333333334e-05,
+      "loss": 0.7239,
+      "step": 780
+    },
+    {
+      "epoch": 41.97,
+      "eval_accuracy": 0.22166666666666668,
+      "eval_loss": 3.9394490718841553,
+      "eval_runtime": 4.2239,
+      "eval_samples_per_second": 142.05,
+      "eval_steps_per_second": 4.498,
+      "step": 787
+    },
+    {
+      "epoch": 42.13,
+      "learning_rate": 4.388888888888889e-05,
+      "loss": 0.7847,
+      "step": 790
+    },
+    {
+      "epoch": 42.67,
+      "learning_rate": 4.4444444444444447e-05,
+      "loss": 0.7732,
+      "step": 800
+    },
+    {
+      "epoch": 42.99,
+      "eval_accuracy": 0.215,
+      "eval_loss": 3.9472830295562744,
+      "eval_runtime": 4.1769,
+      "eval_samples_per_second": 143.646,
+      "eval_steps_per_second": 4.549,
+      "step": 806
+    },
+    {
+      "epoch": 43.2,
+      "learning_rate": 4.5e-05,
+      "loss": 0.7159,
+      "step": 810
+    },
+    {
+      "epoch": 43.73,
+      "learning_rate": 4.555555555555556e-05,
+      "loss": 0.7009,
+      "step": 820
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy": 0.215,
+      "eval_loss": 3.946096897125244,
+      "eval_runtime": 4.1968,
+      "eval_samples_per_second": 142.965,
+      "eval_steps_per_second": 4.527,
+      "step": 825
+    },
+    {
+      "epoch": 44.27,
+      "learning_rate": 4.6111111111111115e-05,
+      "loss": 0.663,
+      "step": 830
+    },
+    {
+      "epoch": 44.8,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 0.5945,
+      "step": 840
+    },
+    {
+      "epoch": 44.96,
+      "eval_accuracy": 0.21333333333333335,
+      "eval_loss": 4.0206685066223145,
+      "eval_runtime": 4.207,
+      "eval_samples_per_second": 142.62,
+      "eval_steps_per_second": 4.516,
+      "step": 843
+    },
+    {
+      "epoch": 45.33,
+      "learning_rate": 4.722222222222222e-05,
+      "loss": 0.5903,
+      "step": 850
+    },
+    {
+      "epoch": 45.87,
+      "learning_rate": 4.7777777777777784e-05,
+      "loss": 0.555,
+      "step": 860
+    },
+    {
+      "epoch": 45.97,
+      "eval_accuracy": 0.20833333333333334,
+      "eval_loss": 4.035262107849121,
+      "eval_runtime": 4.1806,
+      "eval_samples_per_second": 143.522,
+      "eval_steps_per_second": 4.545,
+      "step": 862
+    },
+    {
+      "epoch": 46.4,
+      "learning_rate": 4.8333333333333334e-05,
+      "loss": 0.5768,
+      "step": 870
+    },
+    {
+      "epoch": 46.93,
+      "learning_rate": 4.888888888888889e-05,
+      "loss": 0.5241,
+      "step": 880
+    },
+    {
+      "epoch": 46.99,
+      "eval_accuracy": 0.21666666666666667,
+      "eval_loss": 4.023153305053711,
+      "eval_runtime": 4.1743,
+      "eval_samples_per_second": 143.736,
+      "eval_steps_per_second": 4.552,
+      "step": 881
+    },
+    {
+      "epoch": 47.47,
+      "learning_rate": 4.9444444444444446e-05,
+      "loss": 0.5324,
+      "step": 890
+    },
+    {
+      "epoch": 48.0,
+      "learning_rate": 5e-05,
+      "loss": 0.4789,
+      "step": 900
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy": 0.22,
+      "eval_loss": 4.002644062042236,
+      "eval_runtime": 4.1605,
+      "eval_samples_per_second": 144.215,
+      "eval_steps_per_second": 4.567,
+      "step": 900
+    },
+    {
+      "epoch": 48.53,
+      "learning_rate": 4.9938271604938276e-05,
+      "loss": 0.4284,
+      "step": 910
+    },
+    {
+      "epoch": 48.96,
+      "eval_accuracy": 0.22,
+      "eval_loss": 4.0031256675720215,
+      "eval_runtime": 4.1788,
+      "eval_samples_per_second": 143.582,
+      "eval_steps_per_second": 4.547,
+      "step": 918
+    },
+    {
+      "epoch": 49.07,
+      "learning_rate": 4.987654320987655e-05,
+      "loss": 0.5321,
+      "step": 920
+    },
+    {
+      "epoch": 49.6,
+      "learning_rate": 4.981481481481482e-05,
+      "loss": 0.4701,
+      "step": 930
+    },
+    {
+      "epoch": 49.97,
+      "eval_accuracy": 0.215,
+      "eval_loss": 4.057220935821533,
+      "eval_runtime": 4.1731,
+      "eval_samples_per_second": 143.779,
+      "eval_steps_per_second": 4.553,
+      "step": 937
+    },
+    {
+      "epoch": 50.13,
+      "learning_rate": 4.9753086419753084e-05,
+      "loss": 0.491,
+      "step": 940
+    },
+    {
+      "epoch": 50.67,
+      "learning_rate": 4.969135802469136e-05,
+      "loss": 0.4501,
+      "step": 950
+    },
+    {
+      "epoch": 50.99,
+      "eval_accuracy": 0.215,
+      "eval_loss": 4.087738513946533,
+      "eval_runtime": 4.1876,
+      "eval_samples_per_second": 143.279,
+      "eval_steps_per_second": 4.537,
+      "step": 956
+    },
+    {
+      "epoch": 51.2,
+      "learning_rate": 4.962962962962963e-05,
+      "loss": 0.4453,
+      "step": 960
+    },
+    {
+      "epoch": 51.73,
+      "learning_rate": 4.9567901234567905e-05,
+      "loss": 0.3966,
+      "step": 970
+    },
+    {
+      "epoch": 52.0,
+      "eval_accuracy": 0.21666666666666667,
+      "eval_loss": 4.020733833312988,
+      "eval_runtime": 4.1529,
+      "eval_samples_per_second": 144.478,
+      "eval_steps_per_second": 4.575,
+      "step": 975
+    },
+    {
+      "epoch": 52.27,
+      "learning_rate": 4.950617283950618e-05,
+      "loss": 0.4295,
+      "step": 980
+    },
+    {
+      "epoch": 52.8,
+      "learning_rate": 4.9444444444444446e-05,
+      "loss": 0.3564,
+      "step": 990
+    },
+    {
+      "epoch": 52.96,
+      "eval_accuracy": 0.215,
+      "eval_loss": 4.082664966583252,
+      "eval_runtime": 4.1603,
+      "eval_samples_per_second": 144.219,
+      "eval_steps_per_second": 4.567,
+      "step": 993
+    },
+    {
+      "epoch": 53.33,
+      "learning_rate": 4.938271604938271e-05,
+      "loss": 0.4013,
+      "step": 1000
+    },
+    {
+      "epoch": 53.87,
+      "learning_rate": 4.932098765432099e-05,
+      "loss": 0.3472,
+      "step": 1010
+    },
+    {
+      "epoch": 53.97,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.090172290802002,
+      "eval_runtime": 4.1532,
+      "eval_samples_per_second": 144.465,
+      "eval_steps_per_second": 4.575,
+      "step": 1012
+    },
+    {
+      "epoch": 54.4,
+      "learning_rate": 4.925925925925926e-05,
+      "loss": 0.383,
+      "step": 1020
+    },
+    {
+      "epoch": 54.93,
+      "learning_rate": 4.9197530864197535e-05,
+      "loss": 0.3731,
+      "step": 1030
+    },
+    {
+      "epoch": 54.99,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.095271587371826,
+      "eval_runtime": 4.2172,
+      "eval_samples_per_second": 142.273,
+      "eval_steps_per_second": 4.505,
+      "step": 1031
+    },
+    {
+      "epoch": 55.47,
+      "learning_rate": 4.913580246913581e-05,
+      "loss": 0.3237,
+      "step": 1040
+    },
+    {
+      "epoch": 56.0,
+      "learning_rate": 4.9074074074074075e-05,
+      "loss": 0.3161,
+      "step": 1050
+    },
+    {
+      "epoch": 56.0,
+      "eval_accuracy": 0.20333333333333334,
+      "eval_loss": 4.165963172912598,
+      "eval_runtime": 4.2197,
+      "eval_samples_per_second": 142.19,
+      "eval_steps_per_second": 4.503,
+      "step": 1050
+    },
+    {
+      "epoch": 56.53,
+      "learning_rate": 4.901234567901235e-05,
+      "loss": 0.3352,
+      "step": 1060
+    },
+    {
+      "epoch": 56.96,
+      "eval_accuracy": 0.22166666666666668,
+      "eval_loss": 4.115335464477539,
+      "eval_runtime": 4.1813,
+      "eval_samples_per_second": 143.495,
+      "eval_steps_per_second": 4.544,
+      "step": 1068
+    },
+    {
+      "epoch": 57.07,
+      "learning_rate": 4.8950617283950616e-05,
+      "loss": 0.3455,
+      "step": 1070
+    },
+    {
+      "epoch": 57.6,
+      "learning_rate": 4.888888888888889e-05,
+      "loss": 0.3317,
+      "step": 1080
+    },
+    {
+      "epoch": 57.97,
+      "eval_accuracy": 0.21666666666666667,
+      "eval_loss": 4.1095662117004395,
+      "eval_runtime": 4.2026,
+      "eval_samples_per_second": 142.768,
+      "eval_steps_per_second": 4.521,
+      "step": 1087
+    },
+    {
+      "epoch": 58.13,
+      "learning_rate": 4.8827160493827164e-05,
+      "loss": 0.3594,
+      "step": 1090
+    },
+    {
+      "epoch": 58.67,
+      "learning_rate": 4.876543209876544e-05,
+      "loss": 0.294,
+      "step": 1100
+    },
+    {
+      "epoch": 58.99,
+      "eval_accuracy": 0.215,
+      "eval_loss": 4.185626029968262,
+      "eval_runtime": 4.176,
+      "eval_samples_per_second": 143.679,
+      "eval_steps_per_second": 4.55,
+      "step": 1106
+    },
+    {
+      "epoch": 59.2,
+      "learning_rate": 4.8703703703703704e-05,
+      "loss": 0.3059,
+      "step": 1110
+    },
+    {
+      "epoch": 59.73,
+      "learning_rate": 4.864197530864198e-05,
+      "loss": 0.3299,
+      "step": 1120
+    },
+    {
+      "epoch": 60.0,
+      "eval_accuracy": 0.22333333333333333,
+      "eval_loss": 4.1475830078125,
+      "eval_runtime": 4.1797,
+      "eval_samples_per_second": 143.552,
+      "eval_steps_per_second": 4.546,
+      "step": 1125
+    },
+    {
+      "epoch": 60.27,
+      "learning_rate": 4.858024691358025e-05,
+      "loss": 0.2965,
+      "step": 1130
+    },
+    {
+      "epoch": 60.8,
+      "learning_rate": 4.851851851851852e-05,
+      "loss": 0.2847,
+      "step": 1140
+    },
+    {
+      "epoch": 60.96,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.204588890075684,
+      "eval_runtime": 4.1808,
+      "eval_samples_per_second": 143.515,
+      "eval_steps_per_second": 4.545,
+      "step": 1143
+    },
+    {
+      "epoch": 61.33,
+      "learning_rate": 4.845679012345679e-05,
+      "loss": 0.3176,
+      "step": 1150
+    },
+    {
+      "epoch": 61.87,
+      "learning_rate": 4.8395061728395067e-05,
+      "loss": 0.2924,
+      "step": 1160
+    },
+    {
+      "epoch": 61.97,
+      "eval_accuracy": 0.21833333333333332,
+      "eval_loss": 4.156820297241211,
+      "eval_runtime": 4.1645,
+      "eval_samples_per_second": 144.075,
+      "eval_steps_per_second": 4.562,
+      "step": 1162
+    },
+    {
+      "epoch": 62.4,
+      "learning_rate": 4.8333333333333334e-05,
+      "loss": 0.255,
+      "step": 1170
+    },
+    {
+      "epoch": 62.93,
+      "learning_rate": 4.827160493827161e-05,
+      "loss": 0.2818,
+      "step": 1180
+    },
+    {
+      "epoch": 62.99,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.151925563812256,
+      "eval_runtime": 4.175,
+      "eval_samples_per_second": 143.711,
+      "eval_steps_per_second": 4.551,
+      "step": 1181
+    },
+    {
+      "epoch": 63.47,
+      "learning_rate": 4.820987654320988e-05,
+      "loss": 0.2618,
+      "step": 1190
+    },
+    {
+      "epoch": 64.0,
+      "learning_rate": 4.814814814814815e-05,
+      "loss": 0.2698,
+      "step": 1200
+    },
+    {
+      "epoch": 64.0,
+      "eval_accuracy": 0.215,
+      "eval_loss": 4.227489471435547,
+      "eval_runtime": 4.1793,
+      "eval_samples_per_second": 143.565,
+      "eval_steps_per_second": 4.546,
+      "step": 1200
+    },
+    {
+      "epoch": 64.53,
+      "learning_rate": 4.808641975308642e-05,
+      "loss": 0.2579,
+      "step": 1210
+    },
+    {
+      "epoch": 64.96,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.162615776062012,
+      "eval_runtime": 4.1756,
+      "eval_samples_per_second": 143.693,
+      "eval_steps_per_second": 4.55,
+      "step": 1218
+    },
+    {
+      "epoch": 65.07,
+      "learning_rate": 4.8024691358024696e-05,
+      "loss": 0.2555,
+      "step": 1220
+    },
+    {
+      "epoch": 65.6,
+      "learning_rate": 4.796296296296296e-05,
+      "loss": 0.2597,
+      "step": 1230
+    },
+    {
+      "epoch": 65.97,
+      "eval_accuracy": 0.22166666666666668,
+      "eval_loss": 4.227728843688965,
+      "eval_runtime": 4.1826,
+      "eval_samples_per_second": 143.452,
+      "eval_steps_per_second": 4.543,
+      "step": 1237
+    },
+    {
+      "epoch": 66.13,
+      "learning_rate": 4.7901234567901237e-05,
+      "loss": 0.1989,
+      "step": 1240
+    },
+    {
+      "epoch": 66.67,
+      "learning_rate": 4.783950617283951e-05,
+      "loss": 0.2443,
+      "step": 1250
+    },
+    {
+      "epoch": 66.99,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.192920207977295,
+      "eval_runtime": 4.1844,
+      "eval_samples_per_second": 143.39,
+      "eval_steps_per_second": 4.541,
+      "step": 1256
+    },
+    {
+      "epoch": 67.2,
+      "learning_rate": 4.7777777777777784e-05,
+      "loss": 0.2686,
+      "step": 1260
+    },
+    {
+      "epoch": 67.73,
+      "learning_rate": 4.771604938271605e-05,
+      "loss": 0.2532,
+      "step": 1270
+    },
+    {
+      "epoch": 68.0,
+      "eval_accuracy": 0.22333333333333333,
+      "eval_loss": 4.277872085571289,
+      "eval_runtime": 4.1836,
+      "eval_samples_per_second": 143.417,
+      "eval_steps_per_second": 4.542,
+      "step": 1275
+    },
+    {
+      "epoch": 68.27,
+      "learning_rate": 4.7654320987654325e-05,
+      "loss": 0.2647,
+      "step": 1280
+    },
+    {
+      "epoch": 68.8,
+      "learning_rate": 4.759259259259259e-05,
+      "loss": 0.2305,
+      "step": 1290
+    },
+    {
+      "epoch": 68.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.244070053100586,
+      "eval_runtime": 4.1785,
+      "eval_samples_per_second": 143.591,
+      "eval_steps_per_second": 4.547,
+      "step": 1293
+    },
+    {
+      "epoch": 69.33,
+      "learning_rate": 4.7530864197530866e-05,
+      "loss": 0.2347,
+      "step": 1300
+    },
+    {
+      "epoch": 69.87,
+      "learning_rate": 4.746913580246914e-05,
+      "loss": 0.2423,
+      "step": 1310
+    },
+    {
+      "epoch": 69.97,
+      "eval_accuracy": 0.22166666666666668,
+      "eval_loss": 4.25825309753418,
+      "eval_runtime": 4.1881,
+      "eval_samples_per_second": 143.264,
+      "eval_steps_per_second": 4.537,
+      "step": 1312
+    },
+    {
+      "epoch": 70.4,
+      "learning_rate": 4.740740740740741e-05,
+      "loss": 0.2167,
+      "step": 1320
+    },
+    {
+      "epoch": 70.93,
+      "learning_rate": 4.734567901234569e-05,
+      "loss": 0.222,
+      "step": 1330
+    },
+    {
+      "epoch": 70.99,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.293475151062012,
+      "eval_runtime": 4.178,
+      "eval_samples_per_second": 143.61,
+      "eval_steps_per_second": 4.548,
+      "step": 1331
+    },
+    {
+      "epoch": 71.47,
+      "learning_rate": 4.7283950617283954e-05,
+      "loss": 0.2781,
+      "step": 1340
+    },
+    {
+      "epoch": 72.0,
+      "learning_rate": 4.722222222222222e-05,
+      "loss": 0.2096,
+      "step": 1350
+    },
+    {
+      "epoch": 72.0,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.271422863006592,
+      "eval_runtime": 4.1808,
+      "eval_samples_per_second": 143.513,
+      "eval_steps_per_second": 4.545,
+      "step": 1350
+    },
+    {
+      "epoch": 72.53,
+      "learning_rate": 4.7160493827160495e-05,
+      "loss": 0.1776,
+      "step": 1360
+    },
+    {
+      "epoch": 72.96,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.234805583953857,
+      "eval_runtime": 4.1972,
+      "eval_samples_per_second": 142.953,
+      "eval_steps_per_second": 4.527,
+      "step": 1368
+    },
+    {
+      "epoch": 73.07,
+      "learning_rate": 4.709876543209877e-05,
+      "loss": 0.1977,
+      "step": 1370
+    },
+    {
+      "epoch": 73.6,
+      "learning_rate": 4.703703703703704e-05,
+      "loss": 0.2009,
+      "step": 1380
+    },
+    {
+      "epoch": 73.97,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.292957305908203,
+      "eval_runtime": 4.1773,
+      "eval_samples_per_second": 143.633,
+      "eval_steps_per_second": 4.548,
+      "step": 1387
+    },
+    {
+      "epoch": 74.13,
+      "learning_rate": 4.6975308641975316e-05,
+      "loss": 0.2292,
+      "step": 1390
+    },
+    {
+      "epoch": 74.67,
+      "learning_rate": 4.691358024691358e-05,
+      "loss": 0.2087,
+      "step": 1400
+    },
+    {
+      "epoch": 74.99,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.307061672210693,
+      "eval_runtime": 4.182,
+      "eval_samples_per_second": 143.472,
+      "eval_steps_per_second": 4.543,
+      "step": 1406
+    },
+    {
+      "epoch": 75.2,
+      "learning_rate": 4.685185185185185e-05,
+      "loss": 0.2235,
+      "step": 1410
+    },
+    {
+      "epoch": 75.73,
+      "learning_rate": 4.6790123456790124e-05,
+      "loss": 0.1818,
+      "step": 1420
+    },
+    {
+      "epoch": 76.0,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.296043872833252,
+      "eval_runtime": 4.1959,
+      "eval_samples_per_second": 142.996,
+      "eval_steps_per_second": 4.528,
+      "step": 1425
+    },
+    {
+      "epoch": 76.27,
+      "learning_rate": 4.67283950617284e-05,
+      "loss": 0.2019,
+      "step": 1430
+    },
+    {
+      "epoch": 76.8,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 0.2236,
+      "step": 1440
+    },
+    {
+      "epoch": 76.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.290974140167236,
+      "eval_runtime": 4.1784,
+      "eval_samples_per_second": 143.596,
+      "eval_steps_per_second": 4.547,
+      "step": 1443
+    },
+    {
+      "epoch": 77.33,
+      "learning_rate": 4.6604938271604945e-05,
+      "loss": 0.2008,
+      "step": 1450
+    },
+    {
+      "epoch": 77.87,
+      "learning_rate": 4.654320987654321e-05,
+      "loss": 0.1802,
+      "step": 1460
+    },
+    {
+      "epoch": 77.97,
+      "eval_accuracy": 0.25,
+      "eval_loss": 4.289625644683838,
+      "eval_runtime": 4.224,
+      "eval_samples_per_second": 142.044,
+      "eval_steps_per_second": 4.498,
+      "step": 1462
+    },
+    {
+      "epoch": 78.4,
+      "learning_rate": 4.648148148148148e-05,
+      "loss": 0.2296,
+      "step": 1470
+    },
+    {
+      "epoch": 78.93,
+      "learning_rate": 4.641975308641975e-05,
+      "loss": 0.2037,
+      "step": 1480
+    },
+    {
+      "epoch": 78.99,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.331364154815674,
+      "eval_runtime": 4.2445,
+      "eval_samples_per_second": 141.359,
+      "eval_steps_per_second": 4.476,
+      "step": 1481
+    },
+    {
+      "epoch": 79.47,
+      "learning_rate": 4.635802469135803e-05,
+      "loss": 0.1697,
+      "step": 1490
+    },
+    {
+      "epoch": 80.0,
+      "learning_rate": 4.62962962962963e-05,
+      "loss": 0.1912,
+      "step": 1500
+    },
+    {
+      "epoch": 80.0,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.261221408843994,
+      "eval_runtime": 4.1986,
+      "eval_samples_per_second": 142.905,
+      "eval_steps_per_second": 4.525,
+      "step": 1500
+    },
+    {
+      "epoch": 80.53,
+      "learning_rate": 4.623456790123457e-05,
+      "loss": 0.2305,
+      "step": 1510
+    },
+    {
+      "epoch": 80.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.279019832611084,
+      "eval_runtime": 4.1994,
+      "eval_samples_per_second": 142.879,
+      "eval_steps_per_second": 4.524,
+      "step": 1518
+    },
+    {
+      "epoch": 81.07,
+      "learning_rate": 4.617283950617284e-05,
+      "loss": 0.1808,
+      "step": 1520
+    },
+    {
+      "epoch": 81.6,
+      "learning_rate": 4.6111111111111115e-05,
+      "loss": 0.2188,
+      "step": 1530
+    },
+    {
+      "epoch": 81.97,
+      "eval_accuracy": 0.22166666666666668,
+      "eval_loss": 4.306877613067627,
+      "eval_runtime": 4.1818,
+      "eval_samples_per_second": 143.477,
+      "eval_steps_per_second": 4.543,
+      "step": 1537
+    },
+    {
+      "epoch": 82.13,
+      "learning_rate": 4.604938271604938e-05,
+      "loss": 0.2251,
+      "step": 1540
+    },
+    {
+      "epoch": 82.67,
+      "learning_rate": 4.5987654320987656e-05,
+      "loss": 0.1639,
+      "step": 1550
+    },
+    {
+      "epoch": 82.99,
+      "eval_accuracy": 0.21833333333333332,
+      "eval_loss": 4.353877067565918,
+      "eval_runtime": 4.2313,
+      "eval_samples_per_second": 141.801,
+      "eval_steps_per_second": 4.49,
+      "step": 1556
+    },
+    {
+      "epoch": 83.2,
+      "learning_rate": 4.592592592592593e-05,
+      "loss": 0.1407,
+      "step": 1560
+    },
+    {
+      "epoch": 83.73,
+      "learning_rate": 4.58641975308642e-05,
+      "loss": 0.1741,
+      "step": 1570
+    },
+    {
+      "epoch": 84.0,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.321075916290283,
+      "eval_runtime": 4.2392,
+      "eval_samples_per_second": 141.536,
+      "eval_steps_per_second": 4.482,
+      "step": 1575
+    },
+    {
+      "epoch": 84.27,
+      "learning_rate": 4.580246913580247e-05,
+      "loss": 0.1926,
+      "step": 1580
+    },
+    {
+      "epoch": 84.8,
+      "learning_rate": 4.5740740740740745e-05,
+      "loss": 0.1937,
+      "step": 1590
+    },
+    {
+      "epoch": 84.96,
+      "eval_accuracy": 0.21166666666666667,
+      "eval_loss": 4.357635498046875,
+      "eval_runtime": 4.1967,
+      "eval_samples_per_second": 142.97,
+      "eval_steps_per_second": 4.527,
+      "step": 1593
+    },
+    {
+      "epoch": 85.33,
+      "learning_rate": 4.567901234567901e-05,
+      "loss": 0.1932,
+      "step": 1600
+    },
+    {
+      "epoch": 85.87,
+      "learning_rate": 4.5617283950617285e-05,
+      "loss": 0.1712,
+      "step": 1610
+    },
+    {
+      "epoch": 85.97,
+      "eval_accuracy": 0.22333333333333333,
+      "eval_loss": 4.3434247970581055,
+      "eval_runtime": 4.2166,
+      "eval_samples_per_second": 142.294,
+      "eval_steps_per_second": 4.506,
+      "step": 1612
+    },
+    {
+      "epoch": 86.4,
+      "learning_rate": 4.555555555555556e-05,
+      "loss": 0.1574,
+      "step": 1620
+    },
+    {
+      "epoch": 86.93,
+      "learning_rate": 4.5493827160493826e-05,
+      "loss": 0.1665,
+      "step": 1630
+    },
+    {
+      "epoch": 86.99,
+      "eval_accuracy": 0.21166666666666667,
+      "eval_loss": 4.334897041320801,
+      "eval_runtime": 4.1946,
+      "eval_samples_per_second": 143.042,
+      "eval_steps_per_second": 4.53,
+      "step": 1631
+    },
+    {
+      "epoch": 87.47,
+      "learning_rate": 4.54320987654321e-05,
+      "loss": 0.1565,
+      "step": 1640
+    },
+    {
+      "epoch": 88.0,
+      "learning_rate": 4.5370370370370374e-05,
+      "loss": 0.1846,
+      "step": 1650
+    },
+    {
+      "epoch": 88.0,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.417025566101074,
+      "eval_runtime": 4.202,
+      "eval_samples_per_second": 142.79,
+      "eval_steps_per_second": 4.522,
+      "step": 1650
+    },
+    {
+      "epoch": 88.53,
+      "learning_rate": 4.530864197530865e-05,
+      "loss": 0.1827,
+      "step": 1660
+    },
+    {
+      "epoch": 88.96,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.335045337677002,
+      "eval_runtime": 4.1833,
+      "eval_samples_per_second": 143.428,
+      "eval_steps_per_second": 4.542,
+      "step": 1668
+    },
+    {
+      "epoch": 89.07,
+      "learning_rate": 4.5246913580246914e-05,
+      "loss": 0.1504,
+      "step": 1670
+    },
+    {
+      "epoch": 89.6,
+      "learning_rate": 4.518518518518519e-05,
+      "loss": 0.1591,
+      "step": 1680
+    },
+    {
+      "epoch": 89.97,
+      "eval_accuracy": 0.215,
+      "eval_loss": 4.339655876159668,
+      "eval_runtime": 4.1906,
+      "eval_samples_per_second": 143.178,
+      "eval_steps_per_second": 4.534,
+      "step": 1687
+    },
+    {
+      "epoch": 90.13,
+      "learning_rate": 4.5123456790123455e-05,
+      "loss": 0.1328,
+      "step": 1690
+    },
+    {
+      "epoch": 90.67,
+      "learning_rate": 4.506172839506173e-05,
+      "loss": 0.1508,
+      "step": 1700
+    },
+    {
+      "epoch": 90.99,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.3272809982299805,
+      "eval_runtime": 4.1836,
+      "eval_samples_per_second": 143.418,
+      "eval_steps_per_second": 4.542,
+      "step": 1706
+    },
+    {
+      "epoch": 91.2,
+      "learning_rate": 4.5e-05,
+      "loss": 0.1493,
+      "step": 1710
+    },
+    {
+      "epoch": 91.73,
+      "learning_rate": 4.493827160493828e-05,
+      "loss": 0.1808,
+      "step": 1720
+    },
+    {
+      "epoch": 92.0,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.3314714431762695,
+      "eval_runtime": 4.1951,
+      "eval_samples_per_second": 143.024,
+      "eval_steps_per_second": 4.529,
+      "step": 1725
+    },
+    {
+      "epoch": 92.27,
+      "learning_rate": 4.4876543209876544e-05,
+      "loss": 0.1551,
+      "step": 1730
+    },
+    {
+      "epoch": 92.8,
+      "learning_rate": 4.481481481481482e-05,
+      "loss": 0.17,
+      "step": 1740
+    },
+    {
+      "epoch": 92.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.275998592376709,
+      "eval_runtime": 4.1877,
+      "eval_samples_per_second": 143.278,
+      "eval_steps_per_second": 4.537,
+      "step": 1743
+    },
+    {
+      "epoch": 93.33,
+      "learning_rate": 4.4753086419753084e-05,
+      "loss": 0.1676,
+      "step": 1750
+    },
+    {
+      "epoch": 93.87,
+      "learning_rate": 4.469135802469136e-05,
+      "loss": 0.14,
+      "step": 1760
+    },
+    {
+      "epoch": 93.97,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.314431190490723,
+      "eval_runtime": 4.222,
+      "eval_samples_per_second": 142.112,
+      "eval_steps_per_second": 4.5,
+      "step": 1762
+    },
+    {
+      "epoch": 94.4,
+      "learning_rate": 4.462962962962963e-05,
+      "loss": 0.1526,
+      "step": 1770
+    },
+    {
+      "epoch": 94.93,
+      "learning_rate": 4.4567901234567906e-05,
+      "loss": 0.1734,
+      "step": 1780
+    },
+    {
+      "epoch": 94.99,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.366744041442871,
+      "eval_runtime": 4.2453,
+      "eval_samples_per_second": 141.334,
+      "eval_steps_per_second": 4.476,
+      "step": 1781
+    },
+    {
+      "epoch": 95.47,
+      "learning_rate": 4.450617283950618e-05,
+      "loss": 0.1472,
+      "step": 1790
+    },
+    {
+      "epoch": 96.0,
+      "learning_rate": 4.4444444444444447e-05,
+      "loss": 0.1593,
+      "step": 1800
+    },
+    {
+      "epoch": 96.0,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.390308380126953,
+      "eval_runtime": 4.2118,
+      "eval_samples_per_second": 142.456,
+      "eval_steps_per_second": 4.511,
+      "step": 1800
+    },
+    {
+      "epoch": 96.53,
+      "learning_rate": 4.4382716049382714e-05,
+      "loss": 0.1523,
+      "step": 1810
+    },
+    {
+      "epoch": 96.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.331397533416748,
+      "eval_runtime": 4.1925,
+      "eval_samples_per_second": 143.113,
+      "eval_steps_per_second": 4.532,
+      "step": 1818
+    },
+    {
+      "epoch": 97.07,
+      "learning_rate": 4.432098765432099e-05,
+      "loss": 0.1665,
+      "step": 1820
+    },
+    {
+      "epoch": 97.6,
+      "learning_rate": 4.425925925925926e-05,
+      "loss": 0.1599,
+      "step": 1830
+    },
+    {
+      "epoch": 97.97,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.411539077758789,
+      "eval_runtime": 4.1869,
+      "eval_samples_per_second": 143.306,
+      "eval_steps_per_second": 4.538,
+      "step": 1837
+    },
+    {
+      "epoch": 98.13,
+      "learning_rate": 4.4197530864197535e-05,
+      "loss": 0.1319,
+      "step": 1840
+    },
+    {
+      "epoch": 98.67,
+      "learning_rate": 4.413580246913581e-05,
+      "loss": 0.1352,
+      "step": 1850
+    },
+    {
+      "epoch": 98.99,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.362613201141357,
+      "eval_runtime": 4.181,
+      "eval_samples_per_second": 143.506,
+      "eval_steps_per_second": 4.544,
+      "step": 1856
+    },
+    {
+      "epoch": 99.2,
+      "learning_rate": 4.4074074074074076e-05,
+      "loss": 0.1625,
+      "step": 1860
+    },
+    {
+      "epoch": 99.73,
+      "learning_rate": 4.401234567901234e-05,
+      "loss": 0.1406,
+      "step": 1870
+    },
+    {
+      "epoch": 100.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.355536937713623,
+      "eval_runtime": 4.2021,
+      "eval_samples_per_second": 142.787,
+      "eval_steps_per_second": 4.522,
+      "step": 1875
+    },
+    {
+      "epoch": 100.27,
+      "learning_rate": 4.3950617283950617e-05,
+      "loss": 0.1404,
+      "step": 1880
+    },
+    {
+      "epoch": 100.8,
+      "learning_rate": 4.388888888888889e-05,
+      "loss": 0.1486,
+      "step": 1890
+    },
+    {
+      "epoch": 100.96,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.311557769775391,
+      "eval_runtime": 4.1983,
+      "eval_samples_per_second": 142.914,
+      "eval_steps_per_second": 4.526,
+      "step": 1893
+    },
+    {
+      "epoch": 101.33,
+      "learning_rate": 4.3827160493827164e-05,
+      "loss": 0.1458,
+      "step": 1900
+    },
+    {
+      "epoch": 101.87,
+      "learning_rate": 4.376543209876544e-05,
+      "loss": 0.149,
+      "step": 1910
+    },
+    {
+      "epoch": 101.97,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.389366149902344,
+      "eval_runtime": 4.2353,
+      "eval_samples_per_second": 141.666,
+      "eval_steps_per_second": 4.486,
+      "step": 1912
+    },
+    {
+      "epoch": 102.4,
+      "learning_rate": 4.3703703703703705e-05,
+      "loss": 0.1317,
+      "step": 1920
+    },
+    {
+      "epoch": 102.93,
+      "learning_rate": 4.364197530864197e-05,
+      "loss": 0.115,
+      "step": 1930
+    },
+    {
+      "epoch": 102.99,
+      "eval_accuracy": 0.22333333333333333,
+      "eval_loss": 4.3754754066467285,
+      "eval_runtime": 4.2516,
+      "eval_samples_per_second": 141.122,
+      "eval_steps_per_second": 4.469,
+      "step": 1931
+    },
+    {
+      "epoch": 103.47,
+      "learning_rate": 4.3580246913580246e-05,
+      "loss": 0.1476,
+      "step": 1940
+    },
+    {
+      "epoch": 104.0,
+      "learning_rate": 4.351851851851852e-05,
+      "loss": 0.1301,
+      "step": 1950
+    },
+    {
+      "epoch": 104.0,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.376509189605713,
+      "eval_runtime": 4.2304,
+      "eval_samples_per_second": 141.83,
+      "eval_steps_per_second": 4.491,
+      "step": 1950
+    },
+    {
+      "epoch": 104.53,
+      "learning_rate": 4.345679012345679e-05,
+      "loss": 0.1429,
+      "step": 1960
+    },
+    {
+      "epoch": 104.96,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.402740955352783,
+      "eval_runtime": 4.1986,
+      "eval_samples_per_second": 142.905,
+      "eval_steps_per_second": 4.525,
+      "step": 1968
+    },
+    {
+      "epoch": 105.07,
+      "learning_rate": 4.339506172839507e-05,
+      "loss": 0.123,
+      "step": 1970
+    },
+    {
+      "epoch": 105.6,
+      "learning_rate": 4.3333333333333334e-05,
+      "loss": 0.1209,
+      "step": 1980
+    },
+    {
+      "epoch": 105.97,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.380291938781738,
+      "eval_runtime": 4.2686,
+      "eval_samples_per_second": 140.56,
+      "eval_steps_per_second": 4.451,
+      "step": 1987
+    },
+    {
+      "epoch": 106.13,
+      "learning_rate": 4.327160493827161e-05,
+      "loss": 0.1189,
+      "step": 1990
+    },
+    {
+      "epoch": 106.67,
+      "learning_rate": 4.3209876543209875e-05,
+      "loss": 0.1287,
+      "step": 2000
+    },
+    {
+      "epoch": 106.99,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.323451995849609,
+      "eval_runtime": 5.6461,
+      "eval_samples_per_second": 106.268,
+      "eval_steps_per_second": 3.365,
+      "step": 2006
+    },
+    {
+      "epoch": 107.2,
+      "learning_rate": 4.314814814814815e-05,
+      "loss": 0.1318,
+      "step": 2010
+    },
+    {
+      "epoch": 107.73,
+      "learning_rate": 4.308641975308642e-05,
+      "loss": 0.1318,
+      "step": 2020
+    },
+    {
+      "epoch": 108.0,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.348374843597412,
+      "eval_runtime": 4.2065,
+      "eval_samples_per_second": 142.635,
+      "eval_steps_per_second": 4.517,
+      "step": 2025
+    },
+    {
+      "epoch": 108.27,
+      "learning_rate": 4.3024691358024696e-05,
+      "loss": 0.1328,
+      "step": 2030
+    },
+    {
+      "epoch": 108.8,
+      "learning_rate": 4.296296296296296e-05,
+      "loss": 0.1136,
+      "step": 2040
+    },
+    {
+      "epoch": 108.96,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.39766263961792,
+      "eval_runtime": 4.2019,
+      "eval_samples_per_second": 142.792,
+      "eval_steps_per_second": 4.522,
+      "step": 2043
+    },
+    {
+      "epoch": 109.33,
+      "learning_rate": 4.290123456790124e-05,
+      "loss": 0.1218,
+      "step": 2050
+    },
+    {
+      "epoch": 109.87,
+      "learning_rate": 4.283950617283951e-05,
+      "loss": 0.1326,
+      "step": 2060
+    },
+    {
+      "epoch": 109.97,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.397821426391602,
+      "eval_runtime": 4.1836,
+      "eval_samples_per_second": 143.416,
+      "eval_steps_per_second": 4.542,
+      "step": 2062
+    },
+    {
+      "epoch": 110.4,
+      "learning_rate": 4.277777777777778e-05,
+      "loss": 0.1223,
+      "step": 2070
+    },
+    {
+      "epoch": 110.93,
+      "learning_rate": 4.271604938271605e-05,
+      "loss": 0.1415,
+      "step": 2080
+    },
+    {
+      "epoch": 110.99,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.321408748626709,
+      "eval_runtime": 4.1873,
+      "eval_samples_per_second": 143.29,
+      "eval_steps_per_second": 4.538,
+      "step": 2081
+    },
+    {
+      "epoch": 111.47,
+      "learning_rate": 4.2654320987654325e-05,
+      "loss": 0.1281,
+      "step": 2090
+    },
+    {
+      "epoch": 112.0,
+      "learning_rate": 4.259259259259259e-05,
+      "loss": 0.1229,
+      "step": 2100
+    },
+    {
+      "epoch": 112.0,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.369870662689209,
+      "eval_runtime": 4.2352,
+      "eval_samples_per_second": 141.67,
+      "eval_steps_per_second": 4.486,
+      "step": 2100
+    },
+    {
+      "epoch": 112.53,
+      "learning_rate": 4.2530864197530866e-05,
+      "loss": 0.1004,
+      "step": 2110
+    },
+    {
+      "epoch": 112.96,
+      "eval_accuracy": 0.25833333333333336,
+      "eval_loss": 4.382783889770508,
+      "eval_runtime": 4.2396,
+      "eval_samples_per_second": 141.524,
+      "eval_steps_per_second": 4.482,
+      "step": 2118
+    },
+    {
+      "epoch": 113.07,
+      "learning_rate": 4.246913580246914e-05,
+      "loss": 0.1238,
+      "step": 2120
+    },
+    {
+      "epoch": 113.6,
+      "learning_rate": 4.240740740740741e-05,
+      "loss": 0.0961,
+      "step": 2130
+    },
+    {
+      "epoch": 113.97,
+      "eval_accuracy": 0.25166666666666665,
+      "eval_loss": 4.356354713439941,
+      "eval_runtime": 4.2272,
+      "eval_samples_per_second": 141.937,
+      "eval_steps_per_second": 4.495,
+      "step": 2137
+    },
+    {
+      "epoch": 114.13,
+      "learning_rate": 4.234567901234568e-05,
+      "loss": 0.1181,
+      "step": 2140
+    },
+    {
+      "epoch": 114.67,
+      "learning_rate": 4.2283950617283955e-05,
+      "loss": 0.1132,
+      "step": 2150
+    },
+    {
+      "epoch": 114.99,
+      "eval_accuracy": 0.25333333333333335,
+      "eval_loss": 4.338386058807373,
+      "eval_runtime": 4.2183,
+      "eval_samples_per_second": 142.239,
+      "eval_steps_per_second": 4.504,
+      "step": 2156
+    },
+    {
+      "epoch": 115.2,
+      "learning_rate": 4.222222222222222e-05,
+      "loss": 0.1232,
+      "step": 2160
+    },
+    {
+      "epoch": 115.73,
+      "learning_rate": 4.2160493827160495e-05,
+      "loss": 0.1166,
+      "step": 2170
+    },
+    {
+      "epoch": 116.0,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.415248394012451,
+      "eval_runtime": 4.233,
+      "eval_samples_per_second": 141.743,
+      "eval_steps_per_second": 4.489,
+      "step": 2175
+    },
+    {
+      "epoch": 116.27,
+      "learning_rate": 4.209876543209877e-05,
+      "loss": 0.129,
+      "step": 2180
+    },
+    {
+      "epoch": 116.8,
+      "learning_rate": 4.203703703703704e-05,
+      "loss": 0.1193,
+      "step": 2190
+    },
+    {
+      "epoch": 116.96,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.363390922546387,
+      "eval_runtime": 4.2449,
+      "eval_samples_per_second": 141.345,
+      "eval_steps_per_second": 4.476,
+      "step": 2193
+    },
+    {
+      "epoch": 117.33,
+      "learning_rate": 4.197530864197531e-05,
+      "loss": 0.1348,
+      "step": 2200
+    },
+    {
+      "epoch": 117.87,
+      "learning_rate": 4.1913580246913584e-05,
+      "loss": 0.096,
+      "step": 2210
+    },
+    {
+      "epoch": 117.97,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.382623672485352,
+      "eval_runtime": 4.236,
+      "eval_samples_per_second": 141.643,
+      "eval_steps_per_second": 4.485,
+      "step": 2212
+    },
+    {
+      "epoch": 118.4,
+      "learning_rate": 4.185185185185185e-05,
+      "loss": 0.1375,
+      "step": 2220
+    },
+    {
+      "epoch": 118.93,
+      "learning_rate": 4.1790123456790124e-05,
+      "loss": 0.1158,
+      "step": 2230
+    },
+    {
+      "epoch": 118.99,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.452427387237549,
+      "eval_runtime": 4.1862,
+      "eval_samples_per_second": 143.327,
+      "eval_steps_per_second": 4.539,
+      "step": 2231
+    },
+    {
+      "epoch": 119.47,
+      "learning_rate": 4.17283950617284e-05,
+      "loss": 0.0863,
+      "step": 2240
+    },
+    {
+      "epoch": 120.0,
+      "learning_rate": 4.166666666666667e-05,
+      "loss": 0.099,
+      "step": 2250
+    },
+    {
+      "epoch": 120.0,
+      "eval_accuracy": 0.22333333333333333,
+      "eval_loss": 4.497795581817627,
+      "eval_runtime": 4.2433,
+      "eval_samples_per_second": 141.399,
+      "eval_steps_per_second": 4.478,
+      "step": 2250
+    },
+    {
+      "epoch": 120.53,
+      "learning_rate": 4.1604938271604946e-05,
+      "loss": 0.1065,
+      "step": 2260
+    },
+    {
+      "epoch": 120.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.412367343902588,
+      "eval_runtime": 4.2487,
+      "eval_samples_per_second": 141.22,
+      "eval_steps_per_second": 4.472,
+      "step": 2268
+    },
+    {
+      "epoch": 121.07,
+      "learning_rate": 4.154320987654321e-05,
+      "loss": 0.1202,
+      "step": 2270
+    },
+    {
+      "epoch": 121.6,
+      "learning_rate": 4.148148148148148e-05,
+      "loss": 0.129,
+      "step": 2280
+    },
+    {
+      "epoch": 121.97,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.381356239318848,
+      "eval_runtime": 4.2226,
+      "eval_samples_per_second": 142.092,
+      "eval_steps_per_second": 4.5,
+      "step": 2287
+    },
+    {
+      "epoch": 122.13,
+      "learning_rate": 4.1419753086419754e-05,
+      "loss": 0.1152,
+      "step": 2290
+    },
+    {
+      "epoch": 122.67,
+      "learning_rate": 4.135802469135803e-05,
+      "loss": 0.1047,
+      "step": 2300
+    },
+    {
+      "epoch": 122.99,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.366286754608154,
+      "eval_runtime": 4.1923,
+      "eval_samples_per_second": 143.119,
+      "eval_steps_per_second": 4.532,
+      "step": 2306
+    },
+    {
+      "epoch": 123.2,
+      "learning_rate": 4.12962962962963e-05,
+      "loss": 0.1075,
+      "step": 2310
+    },
+    {
+      "epoch": 123.73,
+      "learning_rate": 4.1234567901234575e-05,
+      "loss": 0.101,
+      "step": 2320
+    },
+    {
+      "epoch": 124.0,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.511256694793701,
+      "eval_runtime": 4.2032,
+      "eval_samples_per_second": 142.748,
+      "eval_steps_per_second": 4.52,
+      "step": 2325
+    },
+    {
+      "epoch": 124.27,
+      "learning_rate": 4.117283950617284e-05,
+      "loss": 0.1111,
+      "step": 2330
+    },
+    {
+      "epoch": 124.8,
+      "learning_rate": 4.111111111111111e-05,
+      "loss": 0.1076,
+      "step": 2340
+    },
+    {
+      "epoch": 124.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.455278396606445,
+      "eval_runtime": 4.1874,
+      "eval_samples_per_second": 143.286,
+      "eval_steps_per_second": 4.537,
+      "step": 2343
+    },
+    {
+      "epoch": 125.33,
+      "learning_rate": 4.104938271604938e-05,
+      "loss": 0.112,
+      "step": 2350
+    },
+    {
+      "epoch": 125.87,
+      "learning_rate": 4.0987654320987657e-05,
+      "loss": 0.1135,
+      "step": 2360
+    },
+    {
+      "epoch": 125.97,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.435062885284424,
+      "eval_runtime": 4.2228,
+      "eval_samples_per_second": 142.087,
+      "eval_steps_per_second": 4.499,
+      "step": 2362
+    },
+    {
+      "epoch": 126.4,
+      "learning_rate": 4.092592592592593e-05,
+      "loss": 0.0839,
+      "step": 2370
+    },
+    {
+      "epoch": 126.93,
+      "learning_rate": 4.0864197530864204e-05,
+      "loss": 0.1066,
+      "step": 2380
+    },
+    {
+      "epoch": 126.99,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.4874396324157715,
+      "eval_runtime": 4.1945,
+      "eval_samples_per_second": 143.044,
+      "eval_steps_per_second": 4.53,
+      "step": 2381
+    },
+    {
+      "epoch": 127.47,
+      "learning_rate": 4.080246913580247e-05,
+      "loss": 0.1007,
+      "step": 2390
+    },
+    {
+      "epoch": 128.0,
+      "learning_rate": 4.074074074074074e-05,
+      "loss": 0.1256,
+      "step": 2400
+    },
+    {
+      "epoch": 128.0,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.463526725769043,
+      "eval_runtime": 4.2437,
+      "eval_samples_per_second": 141.387,
+      "eval_steps_per_second": 4.477,
+      "step": 2400
+    },
+    {
+      "epoch": 128.53,
+      "learning_rate": 4.067901234567901e-05,
+      "loss": 0.0932,
+      "step": 2410
+    },
+    {
+      "epoch": 128.96,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.457594394683838,
+      "eval_runtime": 4.2373,
+      "eval_samples_per_second": 141.599,
+      "eval_steps_per_second": 4.484,
+      "step": 2418
+    },
+    {
+      "epoch": 129.07,
+      "learning_rate": 4.0617283950617286e-05,
+      "loss": 0.1158,
+      "step": 2420
+    },
+    {
+      "epoch": 129.6,
+      "learning_rate": 4.055555555555556e-05,
+      "loss": 0.1189,
+      "step": 2430
+    },
+    {
+      "epoch": 129.97,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.577010154724121,
+      "eval_runtime": 4.2164,
+      "eval_samples_per_second": 142.301,
+      "eval_steps_per_second": 4.506,
+      "step": 2437
+    },
+    {
+      "epoch": 130.13,
+      "learning_rate": 4.049382716049383e-05,
+      "loss": 0.116,
+      "step": 2440
+    },
+    {
+      "epoch": 130.67,
+      "learning_rate": 4.04320987654321e-05,
+      "loss": 0.1096,
+      "step": 2450
+    },
+    {
+      "epoch": 130.99,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.492093563079834,
+      "eval_runtime": 4.2174,
+      "eval_samples_per_second": 142.267,
+      "eval_steps_per_second": 4.505,
+      "step": 2456
+    },
+    {
+      "epoch": 131.2,
+      "learning_rate": 4.0370370370370374e-05,
+      "loss": 0.1004,
+      "step": 2460
+    },
+    {
+      "epoch": 131.73,
+      "learning_rate": 4.030864197530864e-05,
+      "loss": 0.0791,
+      "step": 2470
+    },
+    {
+      "epoch": 132.0,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.508973598480225,
+      "eval_runtime": 4.207,
+      "eval_samples_per_second": 142.618,
+      "eval_steps_per_second": 4.516,
+      "step": 2475
+    },
+    {
+      "epoch": 132.27,
+      "learning_rate": 4.0246913580246915e-05,
+      "loss": 0.0935,
+      "step": 2480
+    },
+    {
+      "epoch": 132.8,
+      "learning_rate": 4.018518518518519e-05,
+      "loss": 0.1152,
+      "step": 2490
+    },
+    {
+      "epoch": 132.96,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.457157611846924,
+      "eval_runtime": 4.1941,
+      "eval_samples_per_second": 143.058,
+      "eval_steps_per_second": 4.53,
+      "step": 2493
+    },
+    {
+      "epoch": 133.33,
+      "learning_rate": 4.012345679012346e-05,
+      "loss": 0.091,
+      "step": 2500
+    },
+    {
+      "epoch": 133.87,
+      "learning_rate": 4.006172839506173e-05,
+      "loss": 0.1264,
+      "step": 2510
+    },
+    {
+      "epoch": 133.97,
+      "eval_accuracy": 0.25,
+      "eval_loss": 4.510921001434326,
+      "eval_runtime": 4.2057,
+      "eval_samples_per_second": 142.663,
+      "eval_steps_per_second": 4.518,
+      "step": 2512
+    },
+    {
+      "epoch": 134.4,
+      "learning_rate": 4e-05,
+      "loss": 0.0749,
+      "step": 2520
+    },
+    {
+      "epoch": 134.93,
+      "learning_rate": 3.993827160493827e-05,
+      "loss": 0.1009,
+      "step": 2530
+    },
+    {
+      "epoch": 134.99,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.523574352264404,
+      "eval_runtime": 4.1971,
+      "eval_samples_per_second": 142.957,
+      "eval_steps_per_second": 4.527,
+      "step": 2531
+    },
+    {
+      "epoch": 135.47,
+      "learning_rate": 3.9876543209876544e-05,
+      "loss": 0.0964,
+      "step": 2540
+    },
+    {
+      "epoch": 136.0,
+      "learning_rate": 3.981481481481482e-05,
+      "loss": 0.0956,
+      "step": 2550
+    },
+    {
+      "epoch": 136.0,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.478328227996826,
+      "eval_runtime": 4.1949,
+      "eval_samples_per_second": 143.029,
+      "eval_steps_per_second": 4.529,
+      "step": 2550
+    },
+    {
+      "epoch": 136.53,
+      "learning_rate": 3.975308641975309e-05,
+      "loss": 0.0919,
+      "step": 2560
+    },
+    {
+      "epoch": 136.96,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.5484466552734375,
+      "eval_runtime": 4.2364,
+      "eval_samples_per_second": 141.629,
+      "eval_steps_per_second": 4.485,
+      "step": 2568
+    },
+    {
+      "epoch": 137.07,
+      "learning_rate": 3.969135802469136e-05,
+      "loss": 0.1154,
+      "step": 2570
+    },
+    {
+      "epoch": 137.6,
+      "learning_rate": 3.962962962962963e-05,
+      "loss": 0.1042,
+      "step": 2580
+    },
+    {
+      "epoch": 137.97,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.542301654815674,
+      "eval_runtime": 4.2535,
+      "eval_samples_per_second": 141.061,
+      "eval_steps_per_second": 4.467,
+      "step": 2587
+    },
+    {
+      "epoch": 138.13,
+      "learning_rate": 3.9567901234567906e-05,
+      "loss": 0.0859,
+      "step": 2590
+    },
+    {
+      "epoch": 138.67,
+      "learning_rate": 3.950617283950617e-05,
+      "loss": 0.1039,
+      "step": 2600
+    },
+    {
+      "epoch": 138.99,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.491814136505127,
+      "eval_runtime": 4.1944,
+      "eval_samples_per_second": 143.049,
+      "eval_steps_per_second": 4.53,
+      "step": 2606
+    },
+    {
+      "epoch": 139.2,
+      "learning_rate": 3.944444444444445e-05,
+      "loss": 0.0846,
+      "step": 2610
+    },
+    {
+      "epoch": 139.73,
+      "learning_rate": 3.938271604938272e-05,
+      "loss": 0.094,
+      "step": 2620
+    },
+    {
+      "epoch": 140.0,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.545647144317627,
+      "eval_runtime": 4.1844,
+      "eval_samples_per_second": 143.39,
+      "eval_steps_per_second": 4.541,
+      "step": 2625
+    },
+    {
+      "epoch": 140.27,
+      "learning_rate": 3.932098765432099e-05,
+      "loss": 0.0929,
+      "step": 2630
+    },
+    {
+      "epoch": 140.8,
+      "learning_rate": 3.925925925925926e-05,
+      "loss": 0.1056,
+      "step": 2640
+    },
+    {
+      "epoch": 140.96,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.521935939788818,
+      "eval_runtime": 4.1914,
+      "eval_samples_per_second": 143.151,
+      "eval_steps_per_second": 4.533,
+      "step": 2643
+    },
+    {
+      "epoch": 141.33,
+      "learning_rate": 3.9197530864197535e-05,
+      "loss": 0.1025,
+      "step": 2650
+    },
+    {
+      "epoch": 141.87,
+      "learning_rate": 3.91358024691358e-05,
+      "loss": 0.0918,
+      "step": 2660
+    },
+    {
+      "epoch": 141.97,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.52545166015625,
+      "eval_runtime": 4.2013,
+      "eval_samples_per_second": 142.812,
+      "eval_steps_per_second": 4.522,
+      "step": 2662
+    },
+    {
+      "epoch": 142.4,
+      "learning_rate": 3.9074074074074076e-05,
+      "loss": 0.098,
+      "step": 2670
+    },
+    {
+      "epoch": 142.93,
+      "learning_rate": 3.901234567901234e-05,
+      "loss": 0.0877,
+      "step": 2680
+    },
+    {
+      "epoch": 142.99,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.492305278778076,
+      "eval_runtime": 4.1813,
+      "eval_samples_per_second": 143.497,
+      "eval_steps_per_second": 4.544,
+      "step": 2681
+    },
+    {
+      "epoch": 143.47,
+      "learning_rate": 3.895061728395062e-05,
+      "loss": 0.0936,
+      "step": 2690
+    },
+    {
+      "epoch": 144.0,
+      "learning_rate": 3.888888888888889e-05,
+      "loss": 0.105,
+      "step": 2700
+    },
+    {
+      "epoch": 144.0,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.535154819488525,
+      "eval_runtime": 4.1912,
+      "eval_samples_per_second": 143.158,
+      "eval_steps_per_second": 4.533,
+      "step": 2700
+    },
+    {
+      "epoch": 144.53,
+      "learning_rate": 3.8827160493827165e-05,
+      "loss": 0.0892,
+      "step": 2710
+    },
+    {
+      "epoch": 144.96,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.471460819244385,
+      "eval_runtime": 4.1911,
+      "eval_samples_per_second": 143.161,
+      "eval_steps_per_second": 4.533,
+      "step": 2718
+    },
+    {
+      "epoch": 145.07,
+      "learning_rate": 3.876543209876544e-05,
+      "loss": 0.1041,
+      "step": 2720
+    },
+    {
+      "epoch": 145.6,
+      "learning_rate": 3.8703703703703705e-05,
+      "loss": 0.0963,
+      "step": 2730
+    },
+    {
+      "epoch": 145.97,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.506024360656738,
+      "eval_runtime": 4.1827,
+      "eval_samples_per_second": 143.449,
+      "eval_steps_per_second": 4.543,
+      "step": 2737
+    },
+    {
+      "epoch": 146.13,
+      "learning_rate": 3.864197530864197e-05,
+      "loss": 0.089,
+      "step": 2740
+    },
+    {
+      "epoch": 146.67,
+      "learning_rate": 3.8580246913580246e-05,
+      "loss": 0.095,
+      "step": 2750
+    },
+    {
+      "epoch": 146.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.559337615966797,
+      "eval_runtime": 4.1884,
+      "eval_samples_per_second": 143.254,
+      "eval_steps_per_second": 4.536,
+      "step": 2756
+    },
+    {
+      "epoch": 147.2,
+      "learning_rate": 3.851851851851852e-05,
+      "loss": 0.0978,
+      "step": 2760
+    },
+    {
+      "epoch": 147.73,
+      "learning_rate": 3.8456790123456794e-05,
+      "loss": 0.0997,
+      "step": 2770
+    },
+    {
+      "epoch": 148.0,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.580421447753906,
+      "eval_runtime": 4.2171,
+      "eval_samples_per_second": 142.278,
+      "eval_steps_per_second": 4.505,
+      "step": 2775
+    },
+    {
+      "epoch": 148.27,
+      "learning_rate": 3.839506172839507e-05,
+      "loss": 0.0689,
+      "step": 2780
+    },
+    {
+      "epoch": 148.8,
+      "learning_rate": 3.8333333333333334e-05,
+      "loss": 0.0839,
+      "step": 2790
+    },
+    {
+      "epoch": 148.96,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.59170389175415,
+      "eval_runtime": 4.2419,
+      "eval_samples_per_second": 141.446,
+      "eval_steps_per_second": 4.479,
+      "step": 2793
+    },
+    {
+      "epoch": 149.33,
+      "learning_rate": 3.82716049382716e-05,
+      "loss": 0.1028,
+      "step": 2800
+    },
+    {
+      "epoch": 149.87,
+      "learning_rate": 3.8209876543209875e-05,
+      "loss": 0.0924,
+      "step": 2810
+    },
+    {
+      "epoch": 149.97,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.593113422393799,
+      "eval_runtime": 4.2519,
+      "eval_samples_per_second": 141.113,
+      "eval_steps_per_second": 4.469,
+      "step": 2812
+    },
+    {
+      "epoch": 150.4,
+      "learning_rate": 3.814814814814815e-05,
+      "loss": 0.0694,
+      "step": 2820
+    },
+    {
+      "epoch": 150.93,
+      "learning_rate": 3.808641975308642e-05,
+      "loss": 0.0781,
+      "step": 2830
+    },
+    {
+      "epoch": 150.99,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.578421115875244,
+      "eval_runtime": 4.2293,
+      "eval_samples_per_second": 141.867,
+      "eval_steps_per_second": 4.492,
+      "step": 2831
+    },
+    {
+      "epoch": 151.47,
+      "learning_rate": 3.80246913580247e-05,
+      "loss": 0.1092,
+      "step": 2840
+    },
+    {
+      "epoch": 152.0,
+      "learning_rate": 3.7962962962962964e-05,
+      "loss": 0.0986,
+      "step": 2850
+    },
+    {
+      "epoch": 152.0,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.654634475708008,
+      "eval_runtime": 4.2497,
+      "eval_samples_per_second": 141.185,
+      "eval_steps_per_second": 4.471,
+      "step": 2850
+    },
+    {
+      "epoch": 152.53,
+      "learning_rate": 3.790123456790123e-05,
+      "loss": 0.0823,
+      "step": 2860
+    },
+    {
+      "epoch": 152.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.598492622375488,
+      "eval_runtime": 4.242,
+      "eval_samples_per_second": 141.443,
+      "eval_steps_per_second": 4.479,
+      "step": 2868
+    },
+    {
+      "epoch": 153.07,
+      "learning_rate": 3.7839506172839504e-05,
+      "loss": 0.0913,
+      "step": 2870
+    },
+    {
+      "epoch": 153.6,
+      "learning_rate": 3.777777777777778e-05,
+      "loss": 0.0887,
+      "step": 2880
+    },
+    {
+      "epoch": 153.97,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.614808559417725,
+      "eval_runtime": 4.2174,
+      "eval_samples_per_second": 142.267,
+      "eval_steps_per_second": 4.505,
+      "step": 2887
+    },
+    {
+      "epoch": 154.13,
+      "learning_rate": 3.771604938271605e-05,
+      "loss": 0.0787,
+      "step": 2890
+    },
+    {
+      "epoch": 154.67,
+      "learning_rate": 3.7654320987654326e-05,
+      "loss": 0.0671,
+      "step": 2900
+    },
+    {
+      "epoch": 154.99,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.63968563079834,
+      "eval_runtime": 4.1904,
+      "eval_samples_per_second": 143.186,
+      "eval_steps_per_second": 4.534,
+      "step": 2906
+    },
+    {
+      "epoch": 155.2,
+      "learning_rate": 3.759259259259259e-05,
+      "loss": 0.104,
+      "step": 2910
+    },
+    {
+      "epoch": 155.73,
+      "learning_rate": 3.7530864197530867e-05,
+      "loss": 0.0897,
+      "step": 2920
+    },
+    {
+      "epoch": 156.0,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.583400249481201,
+      "eval_runtime": 4.1921,
+      "eval_samples_per_second": 143.126,
+      "eval_steps_per_second": 4.532,
+      "step": 2925
+    },
+    {
+      "epoch": 156.27,
+      "learning_rate": 3.7469135802469134e-05,
+      "loss": 0.0795,
+      "step": 2930
+    },
+    {
+      "epoch": 156.8,
+      "learning_rate": 3.740740740740741e-05,
+      "loss": 0.093,
+      "step": 2940
+    },
+    {
+      "epoch": 156.96,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.539726257324219,
+      "eval_runtime": 4.203,
+      "eval_samples_per_second": 142.754,
+      "eval_steps_per_second": 4.521,
+      "step": 2943
+    },
+    {
+      "epoch": 157.33,
+      "learning_rate": 3.734567901234568e-05,
+      "loss": 0.0903,
+      "step": 2950
+    },
+    {
+      "epoch": 157.87,
+      "learning_rate": 3.7283950617283955e-05,
+      "loss": 0.0973,
+      "step": 2960
+    },
+    {
+      "epoch": 157.97,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.55323600769043,
+      "eval_runtime": 4.2018,
+      "eval_samples_per_second": 142.796,
+      "eval_steps_per_second": 4.522,
+      "step": 2962
+    },
+    {
+      "epoch": 158.4,
+      "learning_rate": 3.722222222222222e-05,
+      "loss": 0.064,
+      "step": 2970
+    },
+    {
+      "epoch": 158.93,
+      "learning_rate": 3.7160493827160496e-05,
+      "loss": 0.1001,
+      "step": 2980
+    },
+    {
+      "epoch": 158.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.58270263671875,
+      "eval_runtime": 4.2074,
+      "eval_samples_per_second": 142.605,
+      "eval_steps_per_second": 4.516,
+      "step": 2981
+    },
+    {
+      "epoch": 159.47,
+      "learning_rate": 3.709876543209877e-05,
+      "loss": 0.0788,
+      "step": 2990
+    },
+    {
+      "epoch": 160.0,
+      "learning_rate": 3.7037037037037037e-05,
+      "loss": 0.0884,
+      "step": 3000
+    },
+    {
+      "epoch": 160.0,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.572762966156006,
+      "eval_runtime": 4.217,
+      "eval_samples_per_second": 142.282,
+      "eval_steps_per_second": 4.506,
+      "step": 3000
+    },
+    {
+      "epoch": 160.53,
+      "learning_rate": 3.697530864197531e-05,
+      "loss": 0.084,
+      "step": 3010
+    },
+    {
+      "epoch": 160.96,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.654175758361816,
+      "eval_runtime": 4.2104,
+      "eval_samples_per_second": 142.503,
+      "eval_steps_per_second": 4.513,
+      "step": 3018
+    },
+    {
+      "epoch": 161.07,
+      "learning_rate": 3.6913580246913584e-05,
+      "loss": 0.0773,
+      "step": 3020
+    },
+    {
+      "epoch": 161.6,
+      "learning_rate": 3.685185185185185e-05,
+      "loss": 0.0902,
+      "step": 3030
+    },
+    {
+      "epoch": 161.97,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.636648178100586,
+      "eval_runtime": 4.2002,
+      "eval_samples_per_second": 142.849,
+      "eval_steps_per_second": 4.524,
+      "step": 3037
+    },
+    {
+      "epoch": 162.13,
+      "learning_rate": 3.6790123456790125e-05,
+      "loss": 0.0819,
+      "step": 3040
+    },
+    {
+      "epoch": 162.67,
+      "learning_rate": 3.67283950617284e-05,
+      "loss": 0.0944,
+      "step": 3050
+    },
+    {
+      "epoch": 162.99,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.5957183837890625,
+      "eval_runtime": 4.2207,
+      "eval_samples_per_second": 142.157,
+      "eval_steps_per_second": 4.502,
+      "step": 3056
+    },
+    {
+      "epoch": 163.2,
+      "learning_rate": 3.6666666666666666e-05,
+      "loss": 0.0792,
+      "step": 3060
+    },
+    {
+      "epoch": 163.73,
+      "learning_rate": 3.660493827160494e-05,
+      "loss": 0.0828,
+      "step": 3070
+    },
+    {
+      "epoch": 164.0,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.652061462402344,
+      "eval_runtime": 4.2215,
+      "eval_samples_per_second": 142.13,
+      "eval_steps_per_second": 4.501,
+      "step": 3075
+    },
+    {
+      "epoch": 164.27,
+      "learning_rate": 3.654320987654321e-05,
+      "loss": 0.0729,
+      "step": 3080
+    },
+    {
+      "epoch": 164.8,
+      "learning_rate": 3.648148148148148e-05,
+      "loss": 0.0812,
+      "step": 3090
+    },
+    {
+      "epoch": 164.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.676053047180176,
+      "eval_runtime": 4.2058,
+      "eval_samples_per_second": 142.66,
+      "eval_steps_per_second": 4.518,
+      "step": 3093
+    },
+    {
+      "epoch": 165.33,
+      "learning_rate": 3.6419753086419754e-05,
+      "loss": 0.0723,
+      "step": 3100
+    },
+    {
+      "epoch": 165.87,
+      "learning_rate": 3.635802469135803e-05,
+      "loss": 0.0817,
+      "step": 3110
+    },
+    {
+      "epoch": 165.97,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.627193927764893,
+      "eval_runtime": 4.22,
+      "eval_samples_per_second": 142.182,
+      "eval_steps_per_second": 4.502,
+      "step": 3112
+    },
+    {
+      "epoch": 166.4,
+      "learning_rate": 3.62962962962963e-05,
+      "loss": 0.0718,
+      "step": 3120
+    },
+    {
+      "epoch": 166.93,
+      "learning_rate": 3.623456790123457e-05,
+      "loss": 0.07,
+      "step": 3130
+    },
+    {
+      "epoch": 166.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.653589725494385,
+      "eval_runtime": 4.255,
+      "eval_samples_per_second": 141.01,
+      "eval_steps_per_second": 4.465,
+      "step": 3131
+    },
+    {
+      "epoch": 167.47,
+      "learning_rate": 3.617283950617284e-05,
+      "loss": 0.0672,
+      "step": 3140
+    },
+    {
+      "epoch": 168.0,
+      "learning_rate": 3.611111111111111e-05,
+      "loss": 0.0746,
+      "step": 3150
+    },
+    {
+      "epoch": 168.0,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.567090034484863,
+      "eval_runtime": 4.2691,
+      "eval_samples_per_second": 140.544,
+      "eval_steps_per_second": 4.451,
+      "step": 3150
+    },
+    {
+      "epoch": 168.53,
+      "learning_rate": 3.604938271604938e-05,
+      "loss": 0.0782,
+      "step": 3160
+    },
+    {
+      "epoch": 168.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.591490745544434,
+      "eval_runtime": 4.2101,
+      "eval_samples_per_second": 142.516,
+      "eval_steps_per_second": 4.513,
+      "step": 3168
+    },
+    {
+      "epoch": 169.07,
+      "learning_rate": 3.598765432098766e-05,
+      "loss": 0.0667,
+      "step": 3170
+    },
+    {
+      "epoch": 169.6,
+      "learning_rate": 3.592592592592593e-05,
+      "loss": 0.0677,
+      "step": 3180
+    },
+    {
+      "epoch": 169.97,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.637347221374512,
+      "eval_runtime": 4.2173,
+      "eval_samples_per_second": 142.272,
+      "eval_steps_per_second": 4.505,
+      "step": 3187
+    },
+    {
+      "epoch": 170.13,
+      "learning_rate": 3.5864197530864205e-05,
+      "loss": 0.0839,
+      "step": 3190
+    },
+    {
+      "epoch": 170.67,
+      "learning_rate": 3.580246913580247e-05,
+      "loss": 0.0626,
+      "step": 3200
+    },
+    {
+      "epoch": 170.99,
+      "eval_accuracy": 0.25833333333333336,
+      "eval_loss": 4.672304630279541,
+      "eval_runtime": 4.2092,
+      "eval_samples_per_second": 142.546,
+      "eval_steps_per_second": 4.514,
+      "step": 3206
+    },
+    {
+      "epoch": 171.2,
+      "learning_rate": 3.574074074074074e-05,
+      "loss": 0.076,
+      "step": 3210
+    },
+    {
+      "epoch": 171.73,
+      "learning_rate": 3.567901234567901e-05,
+      "loss": 0.0697,
+      "step": 3220
+    },
+    {
+      "epoch": 172.0,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.681668281555176,
+      "eval_runtime": 4.1945,
+      "eval_samples_per_second": 143.044,
+      "eval_steps_per_second": 4.53,
+      "step": 3225
+    },
+    {
+      "epoch": 172.27,
+      "learning_rate": 3.5617283950617286e-05,
+      "loss": 0.0826,
+      "step": 3230
+    },
+    {
+      "epoch": 172.8,
+      "learning_rate": 3.555555555555556e-05,
+      "loss": 0.077,
+      "step": 3240
+    },
+    {
+      "epoch": 172.96,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.679342269897461,
+      "eval_runtime": 4.2134,
+      "eval_samples_per_second": 142.403,
+      "eval_steps_per_second": 4.509,
+      "step": 3243
+    },
+    {
+      "epoch": 173.33,
+      "learning_rate": 3.5493827160493834e-05,
+      "loss": 0.0766,
+      "step": 3250
+    },
+    {
+      "epoch": 173.87,
+      "learning_rate": 3.54320987654321e-05,
+      "loss": 0.068,
+      "step": 3260
+    },
+    {
+      "epoch": 173.97,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.711019039154053,
+      "eval_runtime": 4.211,
+      "eval_samples_per_second": 142.484,
+      "eval_steps_per_second": 4.512,
+      "step": 3262
+    },
+    {
+      "epoch": 174.4,
+      "learning_rate": 3.537037037037037e-05,
+      "loss": 0.0587,
+      "step": 3270
+    },
+    {
+      "epoch": 174.93,
+      "learning_rate": 3.530864197530864e-05,
+      "loss": 0.0875,
+      "step": 3280
+    },
+    {
+      "epoch": 174.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.701217174530029,
+      "eval_runtime": 4.2016,
+      "eval_samples_per_second": 142.802,
+      "eval_steps_per_second": 4.522,
+      "step": 3281
+    },
+    {
+      "epoch": 175.47,
+      "learning_rate": 3.5246913580246915e-05,
+      "loss": 0.0765,
+      "step": 3290
+    },
+    {
+      "epoch": 176.0,
+      "learning_rate": 3.518518518518519e-05,
+      "loss": 0.0787,
+      "step": 3300
+    },
+    {
+      "epoch": 176.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.711310863494873,
+      "eval_runtime": 4.2282,
+      "eval_samples_per_second": 141.904,
+      "eval_steps_per_second": 4.494,
+      "step": 3300
+    },
+    {
+      "epoch": 176.53,
+      "learning_rate": 3.512345679012346e-05,
+      "loss": 0.0779,
+      "step": 3310
+    },
+    {
+      "epoch": 176.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.699758529663086,
+      "eval_runtime": 4.1993,
+      "eval_samples_per_second": 142.882,
+      "eval_steps_per_second": 4.525,
+      "step": 3318
+    },
+    {
+      "epoch": 177.07,
+      "learning_rate": 3.506172839506173e-05,
+      "loss": 0.0632,
+      "step": 3320
+    },
+    {
+      "epoch": 177.6,
+      "learning_rate": 3.5e-05,
+      "loss": 0.0823,
+      "step": 3330
+    },
+    {
+      "epoch": 177.97,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.709224224090576,
+      "eval_runtime": 4.2154,
+      "eval_samples_per_second": 142.335,
+      "eval_steps_per_second": 4.507,
+      "step": 3337
+    },
+    {
+      "epoch": 178.13,
+      "learning_rate": 3.493827160493827e-05,
+      "loss": 0.0819,
+      "step": 3340
+    },
+    {
+      "epoch": 178.67,
+      "learning_rate": 3.4876543209876545e-05,
+      "loss": 0.0685,
+      "step": 3350
+    },
+    {
+      "epoch": 178.99,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.676272869110107,
+      "eval_runtime": 4.2277,
+      "eval_samples_per_second": 141.922,
+      "eval_steps_per_second": 4.494,
+      "step": 3356
+    },
+    {
+      "epoch": 179.2,
+      "learning_rate": 3.481481481481482e-05,
+      "loss": 0.0592,
+      "step": 3360
+    },
+    {
+      "epoch": 179.73,
+      "learning_rate": 3.475308641975309e-05,
+      "loss": 0.0698,
+      "step": 3370
+    },
+    {
+      "epoch": 180.0,
+      "eval_accuracy": 0.25666666666666665,
+      "eval_loss": 4.718149185180664,
+      "eval_runtime": 4.2111,
+      "eval_samples_per_second": 142.479,
+      "eval_steps_per_second": 4.512,
+      "step": 3375
+    },
+    {
+      "epoch": 180.27,
+      "learning_rate": 3.469135802469136e-05,
+      "loss": 0.074,
+      "step": 3380
+    },
+    {
+      "epoch": 180.8,
+      "learning_rate": 3.4629629629629626e-05,
+      "loss": 0.0924,
+      "step": 3390
+    },
+    {
+      "epoch": 180.96,
+      "eval_accuracy": 0.24833333333333332,
+      "eval_loss": 4.715093612670898,
+      "eval_runtime": 4.2612,
+      "eval_samples_per_second": 140.804,
+      "eval_steps_per_second": 4.459,
+      "step": 3393
+    },
+    {
+      "epoch": 181.33,
+      "learning_rate": 3.45679012345679e-05,
+      "loss": 0.0792,
+      "step": 3400
+    },
+    {
+      "epoch": 181.87,
+      "learning_rate": 3.4506172839506174e-05,
+      "loss": 0.084,
+      "step": 3410
+    },
+    {
+      "epoch": 181.97,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.723077774047852,
+      "eval_runtime": 4.2506,
+      "eval_samples_per_second": 141.157,
+      "eval_steps_per_second": 4.47,
+      "step": 3412
+    },
+    {
+      "epoch": 182.4,
+      "learning_rate": 3.444444444444445e-05,
+      "loss": 0.08,
+      "step": 3420
+    },
+    {
+      "epoch": 182.93,
+      "learning_rate": 3.438271604938272e-05,
+      "loss": 0.0508,
+      "step": 3430
+    },
+    {
+      "epoch": 182.99,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.685625076293945,
+      "eval_runtime": 4.2741,
+      "eval_samples_per_second": 140.38,
+      "eval_steps_per_second": 4.445,
+      "step": 3431
+    },
+    {
+      "epoch": 183.47,
+      "learning_rate": 3.432098765432099e-05,
+      "loss": 0.0604,
+      "step": 3440
+    },
+    {
+      "epoch": 184.0,
+      "learning_rate": 3.425925925925926e-05,
+      "loss": 0.0637,
+      "step": 3450
+    },
+    {
+      "epoch": 184.0,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.704137325286865,
+      "eval_runtime": 4.274,
+      "eval_samples_per_second": 140.382,
+      "eval_steps_per_second": 4.445,
+      "step": 3450
+    },
+    {
+      "epoch": 184.53,
+      "learning_rate": 3.419753086419753e-05,
+      "loss": 0.06,
+      "step": 3460
+    },
+    {
+      "epoch": 184.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.720521450042725,
+      "eval_runtime": 4.2381,
+      "eval_samples_per_second": 141.572,
+      "eval_steps_per_second": 4.483,
+      "step": 3468
+    },
+    {
+      "epoch": 185.07,
+      "learning_rate": 3.41358024691358e-05,
+      "loss": 0.062,
+      "step": 3470
+    },
+    {
+      "epoch": 185.6,
+      "learning_rate": 3.4074074074074077e-05,
+      "loss": 0.0659,
+      "step": 3480
+    },
+    {
+      "epoch": 185.97,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.725123405456543,
+      "eval_runtime": 4.2202,
+      "eval_samples_per_second": 142.172,
+      "eval_steps_per_second": 4.502,
+      "step": 3487
+    },
+    {
+      "epoch": 186.13,
+      "learning_rate": 3.401234567901235e-05,
+      "loss": 0.0788,
+      "step": 3490
+    },
+    {
+      "epoch": 186.67,
+      "learning_rate": 3.395061728395062e-05,
+      "loss": 0.0842,
+      "step": 3500
+    },
+    {
+      "epoch": 186.99,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.721489906311035,
+      "eval_runtime": 4.232,
+      "eval_samples_per_second": 141.778,
+      "eval_steps_per_second": 4.49,
+      "step": 3506
+    },
+    {
+      "epoch": 187.2,
+      "learning_rate": 3.388888888888889e-05,
+      "loss": 0.0773,
+      "step": 3510
+    },
+    {
+      "epoch": 187.73,
+      "learning_rate": 3.3827160493827165e-05,
+      "loss": 0.0733,
+      "step": 3520
+    },
+    {
+      "epoch": 188.0,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.706781387329102,
+      "eval_runtime": 4.2434,
+      "eval_samples_per_second": 141.395,
+      "eval_steps_per_second": 4.478,
+      "step": 3525
+    },
+    {
+      "epoch": 188.27,
+      "learning_rate": 3.376543209876543e-05,
+      "loss": 0.0722,
+      "step": 3530
+    },
+    {
+      "epoch": 188.8,
+      "learning_rate": 3.3703703703703706e-05,
+      "loss": 0.0647,
+      "step": 3540
+    },
+    {
+      "epoch": 188.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.759402751922607,
+      "eval_runtime": 4.2349,
+      "eval_samples_per_second": 141.679,
+      "eval_steps_per_second": 4.486,
+      "step": 3543
+    },
+    {
+      "epoch": 189.33,
+      "learning_rate": 3.364197530864198e-05,
+      "loss": 0.0714,
+      "step": 3550
+    },
+    {
+      "epoch": 189.87,
+      "learning_rate": 3.3580246913580247e-05,
+      "loss": 0.0569,
+      "step": 3560
+    },
+    {
+      "epoch": 189.97,
+      "eval_accuracy": 0.22333333333333333,
+      "eval_loss": 4.783107280731201,
+      "eval_runtime": 4.2157,
+      "eval_samples_per_second": 142.324,
+      "eval_steps_per_second": 4.507,
+      "step": 3562
+    },
+    {
+      "epoch": 190.4,
+      "learning_rate": 3.351851851851852e-05,
+      "loss": 0.0673,
+      "step": 3570
+    },
+    {
+      "epoch": 190.93,
+      "learning_rate": 3.3456790123456794e-05,
+      "loss": 0.0883,
+      "step": 3580
+    },
+    {
+      "epoch": 190.99,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.721207618713379,
+      "eval_runtime": 4.2273,
+      "eval_samples_per_second": 141.935,
+      "eval_steps_per_second": 4.495,
+      "step": 3581
+    },
+    {
+      "epoch": 191.47,
+      "learning_rate": 3.339506172839506e-05,
+      "loss": 0.0704,
+      "step": 3590
+    },
+    {
+      "epoch": 192.0,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 0.0622,
+      "step": 3600
+    },
+    {
+      "epoch": 192.0,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.6877546310424805,
+      "eval_runtime": 4.2265,
+      "eval_samples_per_second": 141.962,
+      "eval_steps_per_second": 4.495,
+      "step": 3600
+    },
+    {
+      "epoch": 192.53,
+      "learning_rate": 3.327160493827161e-05,
+      "loss": 0.057,
+      "step": 3610
+    },
+    {
+      "epoch": 192.96,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.665402889251709,
+      "eval_runtime": 4.2317,
+      "eval_samples_per_second": 141.788,
+      "eval_steps_per_second": 4.49,
+      "step": 3618
+    },
+    {
+      "epoch": 193.07,
+      "learning_rate": 3.3209876543209876e-05,
+      "loss": 0.0563,
+      "step": 3620
+    },
+    {
+      "epoch": 193.6,
+      "learning_rate": 3.314814814814815e-05,
+      "loss": 0.0654,
+      "step": 3630
+    },
+    {
+      "epoch": 193.97,
+      "eval_accuracy": 0.25166666666666665,
+      "eval_loss": 4.635808944702148,
+      "eval_runtime": 4.2421,
+      "eval_samples_per_second": 141.438,
+      "eval_steps_per_second": 4.479,
+      "step": 3637
+    },
+    {
+      "epoch": 194.13,
+      "learning_rate": 3.308641975308642e-05,
+      "loss": 0.0771,
+      "step": 3640
+    },
+    {
+      "epoch": 194.67,
+      "learning_rate": 3.30246913580247e-05,
+      "loss": 0.0868,
+      "step": 3650
+    },
+    {
+      "epoch": 194.99,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.662071704864502,
+      "eval_runtime": 4.2363,
+      "eval_samples_per_second": 141.634,
+      "eval_steps_per_second": 4.485,
+      "step": 3656
+    },
+    {
+      "epoch": 195.2,
+      "learning_rate": 3.2962962962962964e-05,
+      "loss": 0.0592,
+      "step": 3660
+    },
+    {
+      "epoch": 195.73,
+      "learning_rate": 3.290123456790124e-05,
+      "loss": 0.0789,
+      "step": 3670
+    },
+    {
+      "epoch": 196.0,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.698493003845215,
+      "eval_runtime": 4.2392,
+      "eval_samples_per_second": 141.537,
+      "eval_steps_per_second": 4.482,
+      "step": 3675
+    },
+    {
+      "epoch": 196.27,
+      "learning_rate": 3.2839506172839505e-05,
+      "loss": 0.0683,
+      "step": 3680
+    },
+    {
+      "epoch": 196.8,
+      "learning_rate": 3.277777777777778e-05,
+      "loss": 0.0657,
+      "step": 3690
+    },
+    {
+      "epoch": 196.96,
+      "eval_accuracy": 0.25666666666666665,
+      "eval_loss": 4.663649559020996,
+      "eval_runtime": 4.2398,
+      "eval_samples_per_second": 141.517,
+      "eval_steps_per_second": 4.481,
+      "step": 3693
+    },
+    {
+      "epoch": 197.33,
+      "learning_rate": 3.271604938271605e-05,
+      "loss": 0.057,
+      "step": 3700
+    },
+    {
+      "epoch": 197.87,
+      "learning_rate": 3.2654320987654326e-05,
+      "loss": 0.0648,
+      "step": 3710
+    },
+    {
+      "epoch": 197.97,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.7698283195495605,
+      "eval_runtime": 4.2294,
+      "eval_samples_per_second": 141.864,
+      "eval_steps_per_second": 4.492,
+      "step": 3712
+    },
+    {
+      "epoch": 198.4,
+      "learning_rate": 3.25925925925926e-05,
+      "loss": 0.0577,
+      "step": 3720
+    },
+    {
+      "epoch": 198.93,
+      "learning_rate": 3.253086419753087e-05,
+      "loss": 0.0635,
+      "step": 3730
+    },
+    {
+      "epoch": 198.99,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.722621440887451,
+      "eval_runtime": 4.2242,
+      "eval_samples_per_second": 142.038,
+      "eval_steps_per_second": 4.498,
+      "step": 3731
+    },
+    {
+      "epoch": 199.47,
+      "learning_rate": 3.2469135802469134e-05,
+      "loss": 0.0563,
+      "step": 3740
+    },
+    {
+      "epoch": 200.0,
+      "learning_rate": 3.240740740740741e-05,
+      "loss": 0.0637,
+      "step": 3750
+    },
+    {
+      "epoch": 200.0,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.748103618621826,
+      "eval_runtime": 4.2435,
+      "eval_samples_per_second": 141.393,
+      "eval_steps_per_second": 4.477,
+      "step": 3750
+    },
+    {
+      "epoch": 200.53,
+      "learning_rate": 3.234567901234568e-05,
+      "loss": 0.0665,
+      "step": 3760
+    },
+    {
+      "epoch": 200.96,
+      "eval_accuracy": 0.24833333333333332,
+      "eval_loss": 4.778894901275635,
+      "eval_runtime": 4.2222,
+      "eval_samples_per_second": 142.106,
+      "eval_steps_per_second": 4.5,
+      "step": 3768
+    },
+    {
+      "epoch": 201.07,
+      "learning_rate": 3.2283950617283955e-05,
+      "loss": 0.0649,
+      "step": 3770
+    },
+    {
+      "epoch": 201.6,
+      "learning_rate": 3.222222222222223e-05,
+      "loss": 0.0799,
+      "step": 3780
+    },
+    {
+      "epoch": 201.97,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.701383590698242,
+      "eval_runtime": 4.2577,
+      "eval_samples_per_second": 140.92,
+      "eval_steps_per_second": 4.462,
+      "step": 3787
+    },
+    {
+      "epoch": 202.13,
+      "learning_rate": 3.216049382716049e-05,
+      "loss": 0.049,
+      "step": 3790
+    },
+    {
+      "epoch": 202.67,
+      "learning_rate": 3.209876543209876e-05,
+      "loss": 0.064,
+      "step": 3800
+    },
+    {
+      "epoch": 202.99,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.752817153930664,
+      "eval_runtime": 4.2702,
+      "eval_samples_per_second": 140.507,
+      "eval_steps_per_second": 4.449,
+      "step": 3806
+    },
+    {
+      "epoch": 203.2,
+      "learning_rate": 3.203703703703704e-05,
+      "loss": 0.0554,
+      "step": 3810
+    },
+    {
+      "epoch": 203.73,
+      "learning_rate": 3.197530864197531e-05,
+      "loss": 0.0772,
+      "step": 3820
+    },
+    {
+      "epoch": 204.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.740113735198975,
+      "eval_runtime": 4.2544,
+      "eval_samples_per_second": 141.031,
+      "eval_steps_per_second": 4.466,
+      "step": 3825
+    },
+    {
+      "epoch": 204.27,
+      "learning_rate": 3.1913580246913585e-05,
+      "loss": 0.0625,
+      "step": 3830
+    },
+    {
+      "epoch": 204.8,
+      "learning_rate": 3.185185185185185e-05,
+      "loss": 0.0438,
+      "step": 3840
+    },
+    {
+      "epoch": 204.96,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.76779317855835,
+      "eval_runtime": 4.2439,
+      "eval_samples_per_second": 141.379,
+      "eval_steps_per_second": 4.477,
+      "step": 3843
+    },
+    {
+      "epoch": 205.33,
+      "learning_rate": 3.1790123456790125e-05,
+      "loss": 0.0716,
+      "step": 3850
+    },
+    {
+      "epoch": 205.87,
+      "learning_rate": 3.172839506172839e-05,
+      "loss": 0.0766,
+      "step": 3860
+    },
+    {
+      "epoch": 205.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.71795654296875,
+      "eval_runtime": 4.2289,
+      "eval_samples_per_second": 141.881,
+      "eval_steps_per_second": 4.493,
+      "step": 3862
+    },
+    {
+      "epoch": 206.4,
+      "learning_rate": 3.1666666666666666e-05,
+      "loss": 0.08,
+      "step": 3870
+    },
+    {
+      "epoch": 206.93,
+      "learning_rate": 3.160493827160494e-05,
+      "loss": 0.0687,
+      "step": 3880
+    },
+    {
+      "epoch": 206.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.7058281898498535,
+      "eval_runtime": 4.2979,
+      "eval_samples_per_second": 139.602,
+      "eval_steps_per_second": 4.421,
+      "step": 3881
+    },
+    {
+      "epoch": 207.47,
+      "learning_rate": 3.1543209876543214e-05,
+      "loss": 0.0686,
+      "step": 3890
+    },
+    {
+      "epoch": 208.0,
+      "learning_rate": 3.148148148148148e-05,
+      "loss": 0.0801,
+      "step": 3900
+    },
+    {
+      "epoch": 208.0,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.7583746910095215,
+      "eval_runtime": 4.3035,
+      "eval_samples_per_second": 139.423,
+      "eval_steps_per_second": 4.415,
+      "step": 3900
+    },
+    {
+      "epoch": 208.53,
+      "learning_rate": 3.1419753086419755e-05,
+      "loss": 0.0772,
+      "step": 3910
+    },
+    {
+      "epoch": 208.96,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.73037576675415,
+      "eval_runtime": 4.265,
+      "eval_samples_per_second": 140.681,
+      "eval_steps_per_second": 4.455,
+      "step": 3918
+    },
+    {
+      "epoch": 209.07,
+      "learning_rate": 3.135802469135803e-05,
+      "loss": 0.0504,
+      "step": 3920
+    },
+    {
+      "epoch": 209.6,
+      "learning_rate": 3.1296296296296295e-05,
+      "loss": 0.0663,
+      "step": 3930
+    },
+    {
+      "epoch": 209.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.693957805633545,
+      "eval_runtime": 4.2721,
+      "eval_samples_per_second": 140.445,
+      "eval_steps_per_second": 4.447,
+      "step": 3937
+    },
+    {
+      "epoch": 210.13,
+      "learning_rate": 3.123456790123457e-05,
+      "loss": 0.0497,
+      "step": 3940
+    },
+    {
+      "epoch": 210.67,
+      "learning_rate": 3.117283950617284e-05,
+      "loss": 0.0529,
+      "step": 3950
+    },
+    {
+      "epoch": 210.99,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.694019317626953,
+      "eval_runtime": 4.2603,
+      "eval_samples_per_second": 140.834,
+      "eval_steps_per_second": 4.46,
+      "step": 3956
+    },
+    {
+      "epoch": 211.2,
+      "learning_rate": 3.111111111111111e-05,
+      "loss": 0.0559,
+      "step": 3960
+    },
+    {
+      "epoch": 211.73,
+      "learning_rate": 3.1049382716049384e-05,
+      "loss": 0.0568,
+      "step": 3970
+    },
+    {
+      "epoch": 212.0,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.733299255371094,
+      "eval_runtime": 4.2654,
+      "eval_samples_per_second": 140.666,
+      "eval_steps_per_second": 4.454,
+      "step": 3975
+    },
+    {
+      "epoch": 212.27,
+      "learning_rate": 3.098765432098766e-05,
+      "loss": 0.057,
+      "step": 3980
+    },
+    {
+      "epoch": 212.8,
+      "learning_rate": 3.0925925925925924e-05,
+      "loss": 0.0697,
+      "step": 3990
+    },
+    {
+      "epoch": 212.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.667250156402588,
+      "eval_runtime": 4.2885,
+      "eval_samples_per_second": 139.909,
+      "eval_steps_per_second": 4.43,
+      "step": 3993
+    },
+    {
+      "epoch": 213.33,
+      "learning_rate": 3.08641975308642e-05,
+      "loss": 0.0402,
+      "step": 4000
+    },
+    {
+      "epoch": 213.87,
+      "learning_rate": 3.080246913580247e-05,
+      "loss": 0.0394,
+      "step": 4010
+    },
+    {
+      "epoch": 213.97,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.673309326171875,
+      "eval_runtime": 4.2825,
+      "eval_samples_per_second": 140.105,
+      "eval_steps_per_second": 4.437,
+      "step": 4012
+    },
+    {
+      "epoch": 214.4,
+      "learning_rate": 3.074074074074074e-05,
+      "loss": 0.0398,
+      "step": 4020
+    },
+    {
+      "epoch": 214.93,
+      "learning_rate": 3.067901234567901e-05,
+      "loss": 0.0625,
+      "step": 4030
+    },
+    {
+      "epoch": 214.99,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.738312244415283,
+      "eval_runtime": 4.3054,
+      "eval_samples_per_second": 139.359,
+      "eval_steps_per_second": 4.413,
+      "step": 4031
+    },
+    {
+      "epoch": 215.47,
+      "learning_rate": 3.061728395061729e-05,
+      "loss": 0.0626,
+      "step": 4040
+    },
+    {
+      "epoch": 216.0,
+      "learning_rate": 3.055555555555556e-05,
+      "loss": 0.0588,
+      "step": 4050
+    },
+    {
+      "epoch": 216.0,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.767359256744385,
+      "eval_runtime": 4.3344,
+      "eval_samples_per_second": 138.426,
+      "eval_steps_per_second": 4.383,
+      "step": 4050
+    },
+    {
+      "epoch": 216.53,
+      "learning_rate": 3.0493827160493827e-05,
+      "loss": 0.0594,
+      "step": 4060
+    },
+    {
+      "epoch": 216.96,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.687304496765137,
+      "eval_runtime": 4.285,
+      "eval_samples_per_second": 140.024,
+      "eval_steps_per_second": 4.434,
+      "step": 4068
+    },
+    {
+      "epoch": 217.07,
+      "learning_rate": 3.0432098765432098e-05,
+      "loss": 0.0721,
+      "step": 4070
+    },
+    {
+      "epoch": 217.6,
+      "learning_rate": 3.037037037037037e-05,
+      "loss": 0.0451,
+      "step": 4080
+    },
+    {
+      "epoch": 217.97,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.671844959259033,
+      "eval_runtime": 4.2925,
+      "eval_samples_per_second": 139.78,
+      "eval_steps_per_second": 4.426,
+      "step": 4087
+    },
+    {
+      "epoch": 218.13,
+      "learning_rate": 3.0308641975308642e-05,
+      "loss": 0.0445,
+      "step": 4090
+    },
+    {
+      "epoch": 218.67,
+      "learning_rate": 3.0246913580246916e-05,
+      "loss": 0.047,
+      "step": 4100
+    },
+    {
+      "epoch": 218.99,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.714609146118164,
+      "eval_runtime": 4.2883,
+      "eval_samples_per_second": 139.914,
+      "eval_steps_per_second": 4.431,
+      "step": 4106
+    },
+    {
+      "epoch": 219.2,
+      "learning_rate": 3.018518518518519e-05,
+      "loss": 0.0648,
+      "step": 4110
+    },
+    {
+      "epoch": 219.73,
+      "learning_rate": 3.012345679012346e-05,
+      "loss": 0.0445,
+      "step": 4120
+    },
+    {
+      "epoch": 220.0,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.717392921447754,
+      "eval_runtime": 4.2776,
+      "eval_samples_per_second": 140.266,
+      "eval_steps_per_second": 4.442,
+      "step": 4125
+    },
+    {
+      "epoch": 220.27,
+      "learning_rate": 3.0061728395061727e-05,
+      "loss": 0.0438,
+      "step": 4130
+    },
+    {
+      "epoch": 220.8,
+      "learning_rate": 3e-05,
+      "loss": 0.0746,
+      "step": 4140
+    },
+    {
+      "epoch": 220.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.670175075531006,
+      "eval_runtime": 4.2998,
+      "eval_samples_per_second": 139.542,
+      "eval_steps_per_second": 4.419,
+      "step": 4143
+    },
+    {
+      "epoch": 221.33,
+      "learning_rate": 2.993827160493827e-05,
+      "loss": 0.0617,
+      "step": 4150
+    },
+    {
+      "epoch": 221.87,
+      "learning_rate": 2.9876543209876545e-05,
+      "loss": 0.0697,
+      "step": 4160
+    },
+    {
+      "epoch": 221.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.646183967590332,
+      "eval_runtime": 4.3381,
+      "eval_samples_per_second": 138.308,
+      "eval_steps_per_second": 4.38,
+      "step": 4162
+    },
+    {
+      "epoch": 222.4,
+      "learning_rate": 2.981481481481482e-05,
+      "loss": 0.0866,
+      "step": 4170
+    },
+    {
+      "epoch": 222.93,
+      "learning_rate": 2.975308641975309e-05,
+      "loss": 0.0562,
+      "step": 4180
+    },
+    {
+      "epoch": 222.99,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.695559024810791,
+      "eval_runtime": 4.3517,
+      "eval_samples_per_second": 137.876,
+      "eval_steps_per_second": 4.366,
+      "step": 4181
+    },
+    {
+      "epoch": 223.47,
+      "learning_rate": 2.9691358024691356e-05,
+      "loss": 0.0986,
+      "step": 4190
+    },
+    {
+      "epoch": 224.0,
+      "learning_rate": 2.962962962962963e-05,
+      "loss": 0.047,
+      "step": 4200
+    },
+    {
+      "epoch": 224.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.727797985076904,
+      "eval_runtime": 4.3288,
+      "eval_samples_per_second": 138.607,
+      "eval_steps_per_second": 4.389,
+      "step": 4200
+    },
+    {
+      "epoch": 224.53,
+      "learning_rate": 2.95679012345679e-05,
+      "loss": 0.0612,
+      "step": 4210
+    },
+    {
+      "epoch": 224.96,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.730659484863281,
+      "eval_runtime": 4.3287,
+      "eval_samples_per_second": 138.611,
+      "eval_steps_per_second": 4.389,
+      "step": 4218
+    },
+    {
+      "epoch": 225.07,
+      "learning_rate": 2.9506172839506174e-05,
+      "loss": 0.048,
+      "step": 4220
+    },
+    {
+      "epoch": 225.6,
+      "learning_rate": 2.9444444444444448e-05,
+      "loss": 0.0625,
+      "step": 4230
+    },
+    {
+      "epoch": 225.97,
+      "eval_accuracy": 0.25666666666666665,
+      "eval_loss": 4.667015075683594,
+      "eval_runtime": 4.3594,
+      "eval_samples_per_second": 137.632,
+      "eval_steps_per_second": 4.358,
+      "step": 4237
+    },
+    {
+      "epoch": 226.13,
+      "learning_rate": 2.9382716049382718e-05,
+      "loss": 0.0558,
+      "step": 4240
+    },
+    {
+      "epoch": 226.67,
+      "learning_rate": 2.9320987654320992e-05,
+      "loss": 0.0739,
+      "step": 4250
+    },
+    {
+      "epoch": 226.99,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.710987091064453,
+      "eval_runtime": 4.3562,
+      "eval_samples_per_second": 137.734,
+      "eval_steps_per_second": 4.362,
+      "step": 4256
+    },
+    {
+      "epoch": 227.2,
+      "learning_rate": 2.925925925925926e-05,
+      "loss": 0.054,
+      "step": 4260
+    },
+    {
+      "epoch": 227.73,
+      "learning_rate": 2.919753086419753e-05,
+      "loss": 0.0637,
+      "step": 4270
+    },
+    {
+      "epoch": 228.0,
+      "eval_accuracy": 0.22,
+      "eval_loss": 4.703871726989746,
+      "eval_runtime": 4.3307,
+      "eval_samples_per_second": 138.545,
+      "eval_steps_per_second": 4.387,
+      "step": 4275
+    },
+    {
+      "epoch": 228.27,
+      "learning_rate": 2.9135802469135803e-05,
+      "loss": 0.0491,
+      "step": 4280
+    },
+    {
+      "epoch": 228.8,
+      "learning_rate": 2.9074074074074077e-05,
+      "loss": 0.0461,
+      "step": 4290
+    },
+    {
+      "epoch": 228.96,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.711916923522949,
+      "eval_runtime": 4.3352,
+      "eval_samples_per_second": 138.402,
+      "eval_steps_per_second": 4.383,
+      "step": 4293
+    },
+    {
+      "epoch": 229.33,
+      "learning_rate": 2.9012345679012347e-05,
+      "loss": 0.0544,
+      "step": 4300
+    },
+    {
+      "epoch": 229.87,
+      "learning_rate": 2.895061728395062e-05,
+      "loss": 0.0506,
+      "step": 4310
+    },
+    {
+      "epoch": 229.97,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.7098541259765625,
+      "eval_runtime": 4.3688,
+      "eval_samples_per_second": 137.339,
+      "eval_steps_per_second": 4.349,
+      "step": 4312
+    },
+    {
+      "epoch": 230.4,
+      "learning_rate": 2.8888888888888888e-05,
+      "loss": 0.0511,
+      "step": 4320
+    },
+    {
+      "epoch": 230.93,
+      "learning_rate": 2.882716049382716e-05,
+      "loss": 0.0412,
+      "step": 4330
+    },
+    {
+      "epoch": 230.99,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.671385288238525,
+      "eval_runtime": 4.3317,
+      "eval_samples_per_second": 138.515,
+      "eval_steps_per_second": 4.386,
+      "step": 4331
+    },
+    {
+      "epoch": 231.47,
+      "learning_rate": 2.8765432098765432e-05,
+      "loss": 0.0681,
+      "step": 4340
+    },
+    {
+      "epoch": 232.0,
+      "learning_rate": 2.8703703703703706e-05,
+      "loss": 0.057,
+      "step": 4350
+    },
+    {
+      "epoch": 232.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.692080497741699,
+      "eval_runtime": 4.3407,
+      "eval_samples_per_second": 138.227,
+      "eval_steps_per_second": 4.377,
+      "step": 4350
+    },
+    {
+      "epoch": 232.53,
+      "learning_rate": 2.8641975308641977e-05,
+      "loss": 0.0402,
+      "step": 4360
+    },
+    {
+      "epoch": 232.96,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.754528045654297,
+      "eval_runtime": 4.3356,
+      "eval_samples_per_second": 138.389,
+      "eval_steps_per_second": 4.382,
+      "step": 4368
+    },
+    {
+      "epoch": 233.07,
+      "learning_rate": 2.858024691358025e-05,
+      "loss": 0.0766,
+      "step": 4370
+    },
+    {
+      "epoch": 233.6,
+      "learning_rate": 2.851851851851852e-05,
+      "loss": 0.058,
+      "step": 4380
+    },
+    {
+      "epoch": 233.97,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.7573161125183105,
+      "eval_runtime": 4.353,
+      "eval_samples_per_second": 137.837,
+      "eval_steps_per_second": 4.365,
+      "step": 4387
+    },
+    {
+      "epoch": 234.13,
+      "learning_rate": 2.8456790123456788e-05,
+      "loss": 0.0749,
+      "step": 4390
+    },
+    {
+      "epoch": 234.67,
+      "learning_rate": 2.839506172839506e-05,
+      "loss": 0.0661,
+      "step": 4400
+    },
+    {
+      "epoch": 234.99,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.6800408363342285,
+      "eval_runtime": 4.3414,
+      "eval_samples_per_second": 138.205,
+      "eval_steps_per_second": 4.377,
+      "step": 4406
+    },
+    {
+      "epoch": 235.2,
+      "learning_rate": 2.8333333333333335e-05,
+      "loss": 0.0613,
+      "step": 4410
+    },
+    {
+      "epoch": 235.73,
+      "learning_rate": 2.8271604938271606e-05,
+      "loss": 0.0613,
+      "step": 4420
+    },
+    {
+      "epoch": 236.0,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.653302192687988,
+      "eval_runtime": 4.3257,
+      "eval_samples_per_second": 138.707,
+      "eval_steps_per_second": 4.392,
+      "step": 4425
+    },
+    {
+      "epoch": 236.27,
+      "learning_rate": 2.820987654320988e-05,
+      "loss": 0.0555,
+      "step": 4430
+    },
+    {
+      "epoch": 236.8,
+      "learning_rate": 2.814814814814815e-05,
+      "loss": 0.0462,
+      "step": 4440
+    },
+    {
+      "epoch": 236.96,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.674839496612549,
+      "eval_runtime": 4.3407,
+      "eval_samples_per_second": 138.228,
+      "eval_steps_per_second": 4.377,
+      "step": 4443
+    },
+    {
+      "epoch": 237.33,
+      "learning_rate": 2.8086419753086424e-05,
+      "loss": 0.0742,
+      "step": 4450
+    },
+    {
+      "epoch": 237.87,
+      "learning_rate": 2.802469135802469e-05,
+      "loss": 0.0494,
+      "step": 4460
+    },
+    {
+      "epoch": 237.97,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.687388896942139,
+      "eval_runtime": 4.3685,
+      "eval_samples_per_second": 137.346,
+      "eval_steps_per_second": 4.349,
+      "step": 4462
+    },
+    {
+      "epoch": 238.4,
+      "learning_rate": 2.7962962962962965e-05,
+      "loss": 0.0617,
+      "step": 4470
+    },
+    {
+      "epoch": 238.93,
+      "learning_rate": 2.7901234567901235e-05,
+      "loss": 0.0643,
+      "step": 4480
+    },
+    {
+      "epoch": 238.99,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.729123592376709,
+      "eval_runtime": 4.3628,
+      "eval_samples_per_second": 137.527,
+      "eval_steps_per_second": 4.355,
+      "step": 4481
+    },
+    {
+      "epoch": 239.47,
+      "learning_rate": 2.783950617283951e-05,
+      "loss": 0.0557,
+      "step": 4490
+    },
+    {
+      "epoch": 240.0,
+      "learning_rate": 2.777777777777778e-05,
+      "loss": 0.0422,
+      "step": 4500
+    },
+    {
+      "epoch": 240.0,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.7088117599487305,
+      "eval_runtime": 4.3986,
+      "eval_samples_per_second": 136.407,
+      "eval_steps_per_second": 4.32,
+      "step": 4500
+    },
+    {
+      "epoch": 240.53,
+      "learning_rate": 2.7716049382716053e-05,
+      "loss": 0.0376,
+      "step": 4510
+    },
+    {
+      "epoch": 240.96,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.74221658706665,
+      "eval_runtime": 4.4146,
+      "eval_samples_per_second": 135.913,
+      "eval_steps_per_second": 4.304,
+      "step": 4518
+    },
+    {
+      "epoch": 241.07,
+      "learning_rate": 2.765432098765432e-05,
+      "loss": 0.0343,
+      "step": 4520
+    },
+    {
+      "epoch": 241.6,
+      "learning_rate": 2.7592592592592594e-05,
+      "loss": 0.0696,
+      "step": 4530
+    },
+    {
+      "epoch": 241.97,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.801132678985596,
+      "eval_runtime": 4.3976,
+      "eval_samples_per_second": 136.44,
+      "eval_steps_per_second": 4.321,
+      "step": 4537
+    },
+    {
+      "epoch": 242.13,
+      "learning_rate": 2.7530864197530864e-05,
+      "loss": 0.0603,
+      "step": 4540
+    },
+    {
+      "epoch": 242.67,
+      "learning_rate": 2.7469135802469138e-05,
+      "loss": 0.0609,
+      "step": 4550
+    },
+    {
+      "epoch": 242.99,
+      "eval_accuracy": 0.22166666666666668,
+      "eval_loss": 4.801338195800781,
+      "eval_runtime": 4.3836,
+      "eval_samples_per_second": 136.873,
+      "eval_steps_per_second": 4.334,
+      "step": 4556
+    },
+    {
+      "epoch": 243.2,
+      "learning_rate": 2.7407407407407408e-05,
+      "loss": 0.0484,
+      "step": 4560
+    },
+    {
+      "epoch": 243.73,
+      "learning_rate": 2.7345679012345682e-05,
+      "loss": 0.0637,
+      "step": 4570
+    },
+    {
+      "epoch": 244.0,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.760260105133057,
+      "eval_runtime": 4.4032,
+      "eval_samples_per_second": 136.265,
+      "eval_steps_per_second": 4.315,
+      "step": 4575
+    },
+    {
+      "epoch": 244.27,
+      "learning_rate": 2.7283950617283956e-05,
+      "loss": 0.0489,
+      "step": 4580
+    },
+    {
+      "epoch": 244.8,
+      "learning_rate": 2.7222222222222223e-05,
+      "loss": 0.0529,
+      "step": 4590
+    },
+    {
+      "epoch": 244.96,
+      "eval_accuracy": 0.22333333333333333,
+      "eval_loss": 4.789524078369141,
+      "eval_runtime": 4.388,
+      "eval_samples_per_second": 136.736,
+      "eval_steps_per_second": 4.33,
+      "step": 4593
+    },
+    {
+      "epoch": 245.33,
+      "learning_rate": 2.7160493827160493e-05,
+      "loss": 0.0617,
+      "step": 4600
+    },
+    {
+      "epoch": 245.87,
+      "learning_rate": 2.7098765432098767e-05,
+      "loss": 0.0603,
+      "step": 4610
+    },
+    {
+      "epoch": 245.97,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.763910293579102,
+      "eval_runtime": 4.3989,
+      "eval_samples_per_second": 136.397,
+      "eval_steps_per_second": 4.319,
+      "step": 4612
+    },
+    {
+      "epoch": 246.4,
+      "learning_rate": 2.7037037037037037e-05,
+      "loss": 0.0558,
+      "step": 4620
+    },
+    {
+      "epoch": 246.93,
+      "learning_rate": 2.697530864197531e-05,
+      "loss": 0.0365,
+      "step": 4630
+    },
+    {
+      "epoch": 246.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.728492259979248,
+      "eval_runtime": 4.3868,
+      "eval_samples_per_second": 136.774,
+      "eval_steps_per_second": 4.331,
+      "step": 4631
+    },
+    {
+      "epoch": 247.47,
+      "learning_rate": 2.6913580246913585e-05,
+      "loss": 0.0862,
+      "step": 4640
+    },
+    {
+      "epoch": 248.0,
+      "learning_rate": 2.6851851851851855e-05,
+      "loss": 0.0732,
+      "step": 4650
+    },
+    {
+      "epoch": 248.0,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.725191116333008,
+      "eval_runtime": 4.4162,
+      "eval_samples_per_second": 135.863,
+      "eval_steps_per_second": 4.302,
+      "step": 4650
+    },
+    {
+      "epoch": 248.53,
+      "learning_rate": 2.6790123456790122e-05,
+      "loss": 0.0709,
+      "step": 4660
+    },
+    {
+      "epoch": 248.96,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.761960506439209,
+      "eval_runtime": 4.3965,
+      "eval_samples_per_second": 136.472,
+      "eval_steps_per_second": 4.322,
+      "step": 4668
+    },
+    {
+      "epoch": 249.07,
+      "learning_rate": 2.6728395061728396e-05,
+      "loss": 0.0463,
+      "step": 4670
+    },
+    {
+      "epoch": 249.6,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 0.0485,
+      "step": 4680
+    },
+    {
+      "epoch": 249.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.752857208251953,
+      "eval_runtime": 4.404,
+      "eval_samples_per_second": 136.24,
+      "eval_steps_per_second": 4.314,
+      "step": 4687
+    },
+    {
+      "epoch": 250.13,
+      "learning_rate": 2.660493827160494e-05,
+      "loss": 0.0453,
+      "step": 4690
+    },
+    {
+      "epoch": 250.67,
+      "learning_rate": 2.654320987654321e-05,
+      "loss": 0.0449,
+      "step": 4700
+    },
+    {
+      "epoch": 250.99,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.800561428070068,
+      "eval_runtime": 4.4195,
+      "eval_samples_per_second": 135.761,
+      "eval_steps_per_second": 4.299,
+      "step": 4706
+    },
+    {
+      "epoch": 251.2,
+      "learning_rate": 2.6481481481481485e-05,
+      "loss": 0.0629,
+      "step": 4710
+    },
+    {
+      "epoch": 251.73,
+      "learning_rate": 2.641975308641975e-05,
+      "loss": 0.0506,
+      "step": 4720
+    },
+    {
+      "epoch": 252.0,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.802790641784668,
+      "eval_runtime": 4.4472,
+      "eval_samples_per_second": 134.916,
+      "eval_steps_per_second": 4.272,
+      "step": 4725
+    },
+    {
+      "epoch": 252.27,
+      "learning_rate": 2.6358024691358025e-05,
+      "loss": 0.038,
+      "step": 4730
+    },
+    {
+      "epoch": 252.8,
+      "learning_rate": 2.6296296296296296e-05,
+      "loss": 0.0455,
+      "step": 4740
+    },
+    {
+      "epoch": 252.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.777773380279541,
+      "eval_runtime": 4.4493,
+      "eval_samples_per_second": 134.853,
+      "eval_steps_per_second": 4.27,
+      "step": 4743
+    },
+    {
+      "epoch": 253.33,
+      "learning_rate": 2.623456790123457e-05,
+      "loss": 0.0418,
+      "step": 4750
+    },
+    {
+      "epoch": 253.87,
+      "learning_rate": 2.617283950617284e-05,
+      "loss": 0.0594,
+      "step": 4760
+    },
+    {
+      "epoch": 253.97,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.743904113769531,
+      "eval_runtime": 4.4293,
+      "eval_samples_per_second": 135.461,
+      "eval_steps_per_second": 4.29,
+      "step": 4762
+    },
+    {
+      "epoch": 254.4,
+      "learning_rate": 2.6111111111111114e-05,
+      "loss": 0.0511,
+      "step": 4770
+    },
+    {
+      "epoch": 254.93,
+      "learning_rate": 2.6049382716049388e-05,
+      "loss": 0.0551,
+      "step": 4780
+    },
+    {
+      "epoch": 254.99,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.806947708129883,
+      "eval_runtime": 4.4446,
+      "eval_samples_per_second": 134.995,
+      "eval_steps_per_second": 4.275,
+      "step": 4781
+    },
+    {
+      "epoch": 255.47,
+      "learning_rate": 2.5987654320987655e-05,
+      "loss": 0.0512,
+      "step": 4790
+    },
+    {
+      "epoch": 256.0,
+      "learning_rate": 2.5925925925925925e-05,
+      "loss": 0.0435,
+      "step": 4800
+    },
+    {
+      "epoch": 256.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.817090034484863,
+      "eval_runtime": 4.463,
+      "eval_samples_per_second": 134.44,
+      "eval_steps_per_second": 4.257,
+      "step": 4800
+    },
+    {
+      "epoch": 256.53,
+      "learning_rate": 2.58641975308642e-05,
+      "loss": 0.042,
+      "step": 4810
+    },
+    {
+      "epoch": 256.96,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.796122074127197,
+      "eval_runtime": 4.4296,
+      "eval_samples_per_second": 135.453,
+      "eval_steps_per_second": 4.289,
+      "step": 4818
+    },
+    {
+      "epoch": 257.07,
+      "learning_rate": 2.580246913580247e-05,
+      "loss": 0.0614,
+      "step": 4820
+    },
+    {
+      "epoch": 257.6,
+      "learning_rate": 2.5740740740740743e-05,
+      "loss": 0.0403,
+      "step": 4830
+    },
+    {
+      "epoch": 257.97,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.817234516143799,
+      "eval_runtime": 4.4455,
+      "eval_samples_per_second": 134.967,
+      "eval_steps_per_second": 4.274,
+      "step": 4837
+    },
+    {
+      "epoch": 258.13,
+      "learning_rate": 2.5679012345679017e-05,
+      "loss": 0.0428,
+      "step": 4840
+    },
+    {
+      "epoch": 258.67,
+      "learning_rate": 2.5617283950617287e-05,
+      "loss": 0.0524,
+      "step": 4850
+    },
+    {
+      "epoch": 258.99,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.8536577224731445,
+      "eval_runtime": 4.4559,
+      "eval_samples_per_second": 134.654,
+      "eval_steps_per_second": 4.264,
+      "step": 4856
+    },
+    {
+      "epoch": 259.2,
+      "learning_rate": 2.5555555555555554e-05,
+      "loss": 0.0495,
+      "step": 4860
+    },
+    {
+      "epoch": 259.73,
+      "learning_rate": 2.5493827160493828e-05,
+      "loss": 0.0461,
+      "step": 4870
+    },
+    {
+      "epoch": 260.0,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.769797325134277,
+      "eval_runtime": 4.4781,
+      "eval_samples_per_second": 133.984,
+      "eval_steps_per_second": 4.243,
+      "step": 4875
+    },
+    {
+      "epoch": 260.27,
+      "learning_rate": 2.5432098765432098e-05,
+      "loss": 0.0547,
+      "step": 4880
+    },
+    {
+      "epoch": 260.8,
+      "learning_rate": 2.5370370370370372e-05,
+      "loss": 0.05,
+      "step": 4890
+    },
+    {
+      "epoch": 260.96,
+      "eval_accuracy": 0.24833333333333332,
+      "eval_loss": 4.805800437927246,
+      "eval_runtime": 4.4587,
+      "eval_samples_per_second": 134.567,
+      "eval_steps_per_second": 4.261,
+      "step": 4893
+    },
+    {
+      "epoch": 261.33,
+      "learning_rate": 2.5308641975308646e-05,
+      "loss": 0.0463,
+      "step": 4900
+    },
+    {
+      "epoch": 261.87,
+      "learning_rate": 2.5246913580246916e-05,
+      "loss": 0.0545,
+      "step": 4910
+    },
+    {
+      "epoch": 261.97,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.839805603027344,
+      "eval_runtime": 5.2865,
+      "eval_samples_per_second": 113.497,
+      "eval_steps_per_second": 3.594,
+      "step": 4912
+    },
+    {
+      "epoch": 262.4,
+      "learning_rate": 2.5185185185185183e-05,
+      "loss": 0.066,
+      "step": 4920
+    },
+    {
+      "epoch": 262.93,
+      "learning_rate": 2.5123456790123457e-05,
+      "loss": 0.0405,
+      "step": 4930
+    },
+    {
+      "epoch": 262.99,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.82278299331665,
+      "eval_runtime": 4.4772,
+      "eval_samples_per_second": 134.011,
+      "eval_steps_per_second": 4.244,
+      "step": 4931
+    },
+    {
+      "epoch": 263.47,
+      "learning_rate": 2.5061728395061727e-05,
+      "loss": 0.0466,
+      "step": 4940
+    },
+    {
+      "epoch": 264.0,
+      "learning_rate": 2.5e-05,
+      "loss": 0.0615,
+      "step": 4950
+    },
+    {
+      "epoch": 264.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.839545726776123,
+      "eval_runtime": 4.4638,
+      "eval_samples_per_second": 134.416,
+      "eval_steps_per_second": 4.257,
+      "step": 4950
+    },
+    {
+      "epoch": 264.53,
+      "learning_rate": 2.4938271604938275e-05,
+      "loss": 0.0381,
+      "step": 4960
+    },
+    {
+      "epoch": 264.96,
+      "eval_accuracy": 0.22333333333333333,
+      "eval_loss": 4.823088645935059,
+      "eval_runtime": 4.4558,
+      "eval_samples_per_second": 134.657,
+      "eval_steps_per_second": 4.264,
+      "step": 4968
+    },
+    {
+      "epoch": 265.07,
+      "learning_rate": 2.4876543209876542e-05,
+      "loss": 0.0609,
+      "step": 4970
+    },
+    {
+      "epoch": 265.6,
+      "learning_rate": 2.4814814814814816e-05,
+      "loss": 0.0464,
+      "step": 4980
+    },
+    {
+      "epoch": 265.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.818026065826416,
+      "eval_runtime": 4.4544,
+      "eval_samples_per_second": 134.698,
+      "eval_steps_per_second": 4.265,
+      "step": 4987
+    },
+    {
+      "epoch": 266.13,
+      "learning_rate": 2.475308641975309e-05,
+      "loss": 0.0454,
+      "step": 4990
+    },
+    {
+      "epoch": 266.67,
+      "learning_rate": 2.4691358024691357e-05,
+      "loss": 0.058,
+      "step": 5000
+    },
+    {
+      "epoch": 266.99,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.87436056137085,
+      "eval_runtime": 4.4992,
+      "eval_samples_per_second": 133.357,
+      "eval_steps_per_second": 4.223,
+      "step": 5006
+    },
+    {
+      "epoch": 267.2,
+      "learning_rate": 2.462962962962963e-05,
+      "loss": 0.044,
+      "step": 5010
+    },
+    {
+      "epoch": 267.73,
+      "learning_rate": 2.4567901234567904e-05,
+      "loss": 0.0553,
+      "step": 5020
+    },
+    {
+      "epoch": 268.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.886570453643799,
+      "eval_runtime": 4.4931,
+      "eval_samples_per_second": 133.538,
+      "eval_steps_per_second": 4.229,
+      "step": 5025
+    },
+    {
+      "epoch": 268.27,
+      "learning_rate": 2.4506172839506175e-05,
+      "loss": 0.0536,
+      "step": 5030
+    },
+    {
+      "epoch": 268.8,
+      "learning_rate": 2.4444444444444445e-05,
+      "loss": 0.0505,
+      "step": 5040
+    },
+    {
+      "epoch": 268.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.853390216827393,
+      "eval_runtime": 4.4705,
+      "eval_samples_per_second": 134.214,
+      "eval_steps_per_second": 4.25,
+      "step": 5043
+    },
+    {
+      "epoch": 269.33,
+      "learning_rate": 2.438271604938272e-05,
+      "loss": 0.0269,
+      "step": 5050
+    },
+    {
+      "epoch": 269.87,
+      "learning_rate": 2.432098765432099e-05,
+      "loss": 0.049,
+      "step": 5060
+    },
+    {
+      "epoch": 269.97,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.870242118835449,
+      "eval_runtime": 4.4681,
+      "eval_samples_per_second": 134.285,
+      "eval_steps_per_second": 4.252,
+      "step": 5062
+    },
+    {
+      "epoch": 270.4,
+      "learning_rate": 2.425925925925926e-05,
+      "loss": 0.0464,
+      "step": 5070
+    },
+    {
+      "epoch": 270.93,
+      "learning_rate": 2.4197530864197533e-05,
+      "loss": 0.0444,
+      "step": 5080
+    },
+    {
+      "epoch": 270.99,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.871464729309082,
+      "eval_runtime": 4.4942,
+      "eval_samples_per_second": 133.505,
+      "eval_steps_per_second": 4.228,
+      "step": 5081
+    },
+    {
+      "epoch": 271.47,
+      "learning_rate": 2.4135802469135804e-05,
+      "loss": 0.0433,
+      "step": 5090
+    },
+    {
+      "epoch": 272.0,
+      "learning_rate": 2.4074074074074074e-05,
+      "loss": 0.0457,
+      "step": 5100
+    },
+    {
+      "epoch": 272.0,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.827383518218994,
+      "eval_runtime": 4.4878,
+      "eval_samples_per_second": 133.697,
+      "eval_steps_per_second": 4.234,
+      "step": 5100
+    },
+    {
+      "epoch": 272.53,
+      "learning_rate": 2.4012345679012348e-05,
+      "loss": 0.0546,
+      "step": 5110
+    },
+    {
+      "epoch": 272.96,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.844120502471924,
+      "eval_runtime": 4.5115,
+      "eval_samples_per_second": 132.993,
+      "eval_steps_per_second": 4.211,
+      "step": 5118
+    },
+    {
+      "epoch": 273.07,
+      "learning_rate": 2.3950617283950618e-05,
+      "loss": 0.0427,
+      "step": 5120
+    },
+    {
+      "epoch": 273.6,
+      "learning_rate": 2.3888888888888892e-05,
+      "loss": 0.0378,
+      "step": 5130
+    },
+    {
+      "epoch": 273.97,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.822915077209473,
+      "eval_runtime": 4.4872,
+      "eval_samples_per_second": 133.713,
+      "eval_steps_per_second": 4.234,
+      "step": 5137
+    },
+    {
+      "epoch": 274.13,
+      "learning_rate": 2.3827160493827162e-05,
+      "loss": 0.0603,
+      "step": 5140
+    },
+    {
+      "epoch": 274.67,
+      "learning_rate": 2.3765432098765433e-05,
+      "loss": 0.0374,
+      "step": 5150
+    },
+    {
+      "epoch": 274.99,
+      "eval_accuracy": 0.22166666666666668,
+      "eval_loss": 4.805349349975586,
+      "eval_runtime": 4.5038,
+      "eval_samples_per_second": 133.22,
+      "eval_steps_per_second": 4.219,
+      "step": 5156
+    },
+    {
+      "epoch": 275.2,
+      "learning_rate": 2.3703703703703707e-05,
+      "loss": 0.0384,
+      "step": 5160
+    },
+    {
+      "epoch": 275.73,
+      "learning_rate": 2.3641975308641977e-05,
+      "loss": 0.047,
+      "step": 5170
+    },
+    {
+      "epoch": 276.0,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.861939907073975,
+      "eval_runtime": 4.5272,
+      "eval_samples_per_second": 132.531,
+      "eval_steps_per_second": 4.197,
+      "step": 5175
+    },
+    {
+      "epoch": 276.27,
+      "learning_rate": 2.3580246913580247e-05,
+      "loss": 0.0352,
+      "step": 5180
+    },
+    {
+      "epoch": 276.8,
+      "learning_rate": 2.351851851851852e-05,
+      "loss": 0.0526,
+      "step": 5190
+    },
+    {
+      "epoch": 276.96,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.87931489944458,
+      "eval_runtime": 4.5397,
+      "eval_samples_per_second": 132.166,
+      "eval_steps_per_second": 4.185,
+      "step": 5193
+    },
+    {
+      "epoch": 277.33,
+      "learning_rate": 2.345679012345679e-05,
+      "loss": 0.0406,
+      "step": 5200
+    },
+    {
+      "epoch": 277.87,
+      "learning_rate": 2.3395061728395062e-05,
+      "loss": 0.0503,
+      "step": 5210
+    },
+    {
+      "epoch": 277.97,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.9059576988220215,
+      "eval_runtime": 4.5301,
+      "eval_samples_per_second": 132.446,
+      "eval_steps_per_second": 4.194,
+      "step": 5212
+    },
+    {
+      "epoch": 278.4,
+      "learning_rate": 2.3333333333333336e-05,
+      "loss": 0.043,
+      "step": 5220
+    },
+    {
+      "epoch": 278.93,
+      "learning_rate": 2.3271604938271606e-05,
+      "loss": 0.0414,
+      "step": 5230
+    },
+    {
+      "epoch": 278.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.86867094039917,
+      "eval_runtime": 4.4935,
+      "eval_samples_per_second": 133.527,
+      "eval_steps_per_second": 4.228,
+      "step": 5231
+    },
+    {
+      "epoch": 279.47,
+      "learning_rate": 2.3209876543209877e-05,
+      "loss": 0.0561,
+      "step": 5240
+    },
+    {
+      "epoch": 280.0,
+      "learning_rate": 2.314814814814815e-05,
+      "loss": 0.0361,
+      "step": 5250
+    },
+    {
+      "epoch": 280.0,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.853731155395508,
+      "eval_runtime": 4.5103,
+      "eval_samples_per_second": 133.029,
+      "eval_steps_per_second": 4.213,
+      "step": 5250
+    },
+    {
+      "epoch": 280.53,
+      "learning_rate": 2.308641975308642e-05,
+      "loss": 0.0449,
+      "step": 5260
+    },
+    {
+      "epoch": 280.96,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.8204240798950195,
+      "eval_runtime": 4.5205,
+      "eval_samples_per_second": 132.729,
+      "eval_steps_per_second": 4.203,
+      "step": 5268
+    },
+    {
+      "epoch": 281.07,
+      "learning_rate": 2.302469135802469e-05,
+      "loss": 0.0527,
+      "step": 5270
+    },
+    {
+      "epoch": 281.6,
+      "learning_rate": 2.2962962962962965e-05,
+      "loss": 0.0596,
+      "step": 5280
+    },
+    {
+      "epoch": 281.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.803044319152832,
+      "eval_runtime": 4.5651,
+      "eval_samples_per_second": 131.433,
+      "eval_steps_per_second": 4.162,
+      "step": 5287
+    },
+    {
+      "epoch": 282.13,
+      "learning_rate": 2.2901234567901235e-05,
+      "loss": 0.056,
+      "step": 5290
+    },
+    {
+      "epoch": 282.67,
+      "learning_rate": 2.2839506172839506e-05,
+      "loss": 0.0494,
+      "step": 5300
+    },
+    {
+      "epoch": 282.99,
+      "eval_accuracy": 0.24833333333333332,
+      "eval_loss": 4.8059892654418945,
+      "eval_runtime": 4.5556,
+      "eval_samples_per_second": 131.706,
+      "eval_steps_per_second": 4.171,
+      "step": 5306
+    },
+    {
+      "epoch": 283.2,
+      "learning_rate": 2.277777777777778e-05,
+      "loss": 0.0311,
+      "step": 5310
+    },
+    {
+      "epoch": 283.73,
+      "learning_rate": 2.271604938271605e-05,
+      "loss": 0.0483,
+      "step": 5320
+    },
+    {
+      "epoch": 284.0,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.7877960205078125,
+      "eval_runtime": 4.5534,
+      "eval_samples_per_second": 131.77,
+      "eval_steps_per_second": 4.173,
+      "step": 5325
+    },
+    {
+      "epoch": 284.27,
+      "learning_rate": 2.2654320987654324e-05,
+      "loss": 0.0606,
+      "step": 5330
+    },
+    {
+      "epoch": 284.8,
+      "learning_rate": 2.2592592592592594e-05,
+      "loss": 0.0338,
+      "step": 5340
+    },
+    {
+      "epoch": 284.96,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.825440406799316,
+      "eval_runtime": 4.5446,
+      "eval_samples_per_second": 132.024,
+      "eval_steps_per_second": 4.181,
+      "step": 5343
+    },
+    {
+      "epoch": 285.33,
+      "learning_rate": 2.2530864197530865e-05,
+      "loss": 0.0454,
+      "step": 5350
+    },
+    {
+      "epoch": 285.87,
+      "learning_rate": 2.246913580246914e-05,
+      "loss": 0.0319,
+      "step": 5360
+    },
+    {
+      "epoch": 285.97,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.826366424560547,
+      "eval_runtime": 4.5558,
+      "eval_samples_per_second": 131.701,
+      "eval_steps_per_second": 4.171,
+      "step": 5362
+    },
+    {
+      "epoch": 286.4,
+      "learning_rate": 2.240740740740741e-05,
+      "loss": 0.0414,
+      "step": 5370
+    },
+    {
+      "epoch": 286.93,
+      "learning_rate": 2.234567901234568e-05,
+      "loss": 0.0454,
+      "step": 5380
+    },
+    {
+      "epoch": 286.99,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.842591762542725,
+      "eval_runtime": 4.55,
+      "eval_samples_per_second": 131.868,
+      "eval_steps_per_second": 4.176,
+      "step": 5381
+    },
+    {
+      "epoch": 287.47,
+      "learning_rate": 2.2283950617283953e-05,
+      "loss": 0.0426,
+      "step": 5390
+    },
+    {
+      "epoch": 288.0,
+      "learning_rate": 2.2222222222222223e-05,
+      "loss": 0.0409,
+      "step": 5400
+    },
+    {
+      "epoch": 288.0,
+      "eval_accuracy": 0.24833333333333332,
+      "eval_loss": 4.819784164428711,
+      "eval_runtime": 4.5681,
+      "eval_samples_per_second": 131.345,
+      "eval_steps_per_second": 4.159,
+      "step": 5400
+    },
+    {
+      "epoch": 288.53,
+      "learning_rate": 2.2160493827160494e-05,
+      "loss": 0.0435,
+      "step": 5410
+    },
+    {
+      "epoch": 288.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.8339433670043945,
+      "eval_runtime": 4.5995,
+      "eval_samples_per_second": 130.45,
+      "eval_steps_per_second": 4.131,
+      "step": 5418
+    },
+    {
+      "epoch": 289.07,
+      "learning_rate": 2.2098765432098767e-05,
+      "loss": 0.0452,
+      "step": 5420
+    },
+    {
+      "epoch": 289.6,
+      "learning_rate": 2.2037037037037038e-05,
+      "loss": 0.0498,
+      "step": 5430
+    },
+    {
+      "epoch": 289.97,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.838677883148193,
+      "eval_runtime": 4.6181,
+      "eval_samples_per_second": 129.923,
+      "eval_steps_per_second": 4.114,
+      "step": 5437
+    },
+    {
+      "epoch": 290.13,
+      "learning_rate": 2.1975308641975308e-05,
+      "loss": 0.0467,
+      "step": 5440
+    },
+    {
+      "epoch": 290.67,
+      "learning_rate": 2.1913580246913582e-05,
+      "loss": 0.0447,
+      "step": 5450
+    },
+    {
+      "epoch": 290.99,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.834191799163818,
+      "eval_runtime": 4.5905,
+      "eval_samples_per_second": 130.706,
+      "eval_steps_per_second": 4.139,
+      "step": 5456
+    },
+    {
+      "epoch": 291.2,
+      "learning_rate": 2.1851851851851852e-05,
+      "loss": 0.0441,
+      "step": 5460
+    },
+    {
+      "epoch": 291.73,
+      "learning_rate": 2.1790123456790123e-05,
+      "loss": 0.0402,
+      "step": 5470
+    },
+    {
+      "epoch": 292.0,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.849569320678711,
+      "eval_runtime": 4.5792,
+      "eval_samples_per_second": 131.028,
+      "eval_steps_per_second": 4.149,
+      "step": 5475
+    },
+    {
+      "epoch": 292.27,
+      "learning_rate": 2.1728395061728397e-05,
+      "loss": 0.0314,
+      "step": 5480
+    },
+    {
+      "epoch": 292.8,
+      "learning_rate": 2.1666666666666667e-05,
+      "loss": 0.0366,
+      "step": 5490
+    },
+    {
+      "epoch": 292.96,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.867130279541016,
+      "eval_runtime": 4.5608,
+      "eval_samples_per_second": 131.555,
+      "eval_steps_per_second": 4.166,
+      "step": 5493
+    },
+    {
+      "epoch": 293.33,
+      "learning_rate": 2.1604938271604937e-05,
+      "loss": 0.0388,
+      "step": 5500
+    },
+    {
+      "epoch": 293.87,
+      "learning_rate": 2.154320987654321e-05,
+      "loss": 0.0369,
+      "step": 5510
+    },
+    {
+      "epoch": 293.97,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.836596488952637,
+      "eval_runtime": 4.583,
+      "eval_samples_per_second": 130.917,
+      "eval_steps_per_second": 4.146,
+      "step": 5512
+    },
+    {
+      "epoch": 294.4,
+      "learning_rate": 2.148148148148148e-05,
+      "loss": 0.0651,
+      "step": 5520
+    },
+    {
+      "epoch": 294.93,
+      "learning_rate": 2.1419753086419755e-05,
+      "loss": 0.0361,
+      "step": 5530
+    },
+    {
+      "epoch": 294.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.799242973327637,
+      "eval_runtime": 4.6324,
+      "eval_samples_per_second": 129.522,
+      "eval_steps_per_second": 4.102,
+      "step": 5531
+    },
+    {
+      "epoch": 295.47,
+      "learning_rate": 2.1358024691358026e-05,
+      "loss": 0.0549,
+      "step": 5540
+    },
+    {
+      "epoch": 296.0,
+      "learning_rate": 2.1296296296296296e-05,
+      "loss": 0.0448,
+      "step": 5550
+    },
+    {
+      "epoch": 296.0,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.848645210266113,
+      "eval_runtime": 4.6425,
+      "eval_samples_per_second": 129.241,
+      "eval_steps_per_second": 4.093,
+      "step": 5550
+    },
+    {
+      "epoch": 296.53,
+      "learning_rate": 2.123456790123457e-05,
+      "loss": 0.055,
+      "step": 5560
+    },
+    {
+      "epoch": 296.96,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.897942066192627,
+      "eval_runtime": 4.5933,
+      "eval_samples_per_second": 130.624,
+      "eval_steps_per_second": 4.136,
+      "step": 5568
+    },
+    {
+      "epoch": 297.07,
+      "learning_rate": 2.117283950617284e-05,
+      "loss": 0.0427,
+      "step": 5570
+    },
+    {
+      "epoch": 297.6,
+      "learning_rate": 2.111111111111111e-05,
+      "loss": 0.0585,
+      "step": 5580
+    },
+    {
+      "epoch": 297.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.866022109985352,
+      "eval_runtime": 4.6056,
+      "eval_samples_per_second": 130.276,
+      "eval_steps_per_second": 4.125,
+      "step": 5587
+    },
+    {
+      "epoch": 298.13,
+      "learning_rate": 2.1049382716049385e-05,
+      "loss": 0.0478,
+      "step": 5590
+    },
+    {
+      "epoch": 298.67,
+      "learning_rate": 2.0987654320987655e-05,
+      "loss": 0.0477,
+      "step": 5600
+    },
+    {
+      "epoch": 298.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.8717451095581055,
+      "eval_runtime": 4.6241,
+      "eval_samples_per_second": 129.756,
+      "eval_steps_per_second": 4.109,
+      "step": 5606
+    },
+    {
+      "epoch": 299.2,
+      "learning_rate": 2.0925925925925925e-05,
+      "loss": 0.0515,
+      "step": 5610
+    },
+    {
+      "epoch": 299.73,
+      "learning_rate": 2.08641975308642e-05,
+      "loss": 0.0247,
+      "step": 5620
+    },
+    {
+      "epoch": 300.0,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.883845806121826,
+      "eval_runtime": 4.6098,
+      "eval_samples_per_second": 130.158,
+      "eval_steps_per_second": 4.122,
+      "step": 5625
+    },
+    {
+      "epoch": 300.27,
+      "learning_rate": 2.0802469135802473e-05,
+      "loss": 0.0438,
+      "step": 5630
+    },
+    {
+      "epoch": 300.8,
+      "learning_rate": 2.074074074074074e-05,
+      "loss": 0.047,
+      "step": 5640
+    },
+    {
+      "epoch": 300.96,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.824845314025879,
+      "eval_runtime": 4.5984,
+      "eval_samples_per_second": 130.482,
+      "eval_steps_per_second": 4.132,
+      "step": 5643
+    },
+    {
+      "epoch": 301.33,
+      "learning_rate": 2.0679012345679014e-05,
+      "loss": 0.0497,
+      "step": 5650
+    },
+    {
+      "epoch": 301.87,
+      "learning_rate": 2.0617283950617287e-05,
+      "loss": 0.0608,
+      "step": 5660
+    },
+    {
+      "epoch": 301.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.832959175109863,
+      "eval_runtime": 4.6004,
+      "eval_samples_per_second": 130.422,
+      "eval_steps_per_second": 4.13,
+      "step": 5662
+    },
+    {
+      "epoch": 302.4,
+      "learning_rate": 2.0555555555555555e-05,
+      "loss": 0.0413,
+      "step": 5670
+    },
+    {
+      "epoch": 302.93,
+      "learning_rate": 2.0493827160493828e-05,
+      "loss": 0.0417,
+      "step": 5680
+    },
+    {
+      "epoch": 302.99,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.823630332946777,
+      "eval_runtime": 4.5991,
+      "eval_samples_per_second": 130.462,
+      "eval_steps_per_second": 4.131,
+      "step": 5681
+    },
+    {
+      "epoch": 303.47,
+      "learning_rate": 2.0432098765432102e-05,
+      "loss": 0.0329,
+      "step": 5690
+    },
+    {
+      "epoch": 304.0,
+      "learning_rate": 2.037037037037037e-05,
+      "loss": 0.0494,
+      "step": 5700
+    },
+    {
+      "epoch": 304.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.807046413421631,
+      "eval_runtime": 4.6151,
+      "eval_samples_per_second": 130.008,
+      "eval_steps_per_second": 4.117,
+      "step": 5700
+    },
+    {
+      "epoch": 304.53,
+      "learning_rate": 2.0308641975308643e-05,
+      "loss": 0.0316,
+      "step": 5710
+    },
+    {
+      "epoch": 304.96,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.821282386779785,
+      "eval_runtime": 4.6056,
+      "eval_samples_per_second": 130.275,
+      "eval_steps_per_second": 4.125,
+      "step": 5718
+    },
+    {
+      "epoch": 305.07,
+      "learning_rate": 2.0246913580246917e-05,
+      "loss": 0.0473,
+      "step": 5720
+    },
+    {
+      "epoch": 305.6,
+      "learning_rate": 2.0185185185185187e-05,
+      "loss": 0.0421,
+      "step": 5730
+    },
+    {
+      "epoch": 305.97,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.863409996032715,
+      "eval_runtime": 4.6418,
+      "eval_samples_per_second": 129.259,
+      "eval_steps_per_second": 4.093,
+      "step": 5737
+    },
+    {
+      "epoch": 306.13,
+      "learning_rate": 2.0123456790123457e-05,
+      "loss": 0.0382,
+      "step": 5740
+    },
+    {
+      "epoch": 306.67,
+      "learning_rate": 2.006172839506173e-05,
+      "loss": 0.0411,
+      "step": 5750
+    },
+    {
+      "epoch": 306.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.877004623413086,
+      "eval_runtime": 4.6583,
+      "eval_samples_per_second": 128.802,
+      "eval_steps_per_second": 4.079,
+      "step": 5756
+    },
+    {
+      "epoch": 307.2,
+      "learning_rate": 2e-05,
+      "loss": 0.042,
+      "step": 5760
+    },
+    {
+      "epoch": 307.73,
+      "learning_rate": 1.9938271604938272e-05,
+      "loss": 0.0404,
+      "step": 5770
+    },
+    {
+      "epoch": 308.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.9029860496521,
+      "eval_runtime": 4.6363,
+      "eval_samples_per_second": 129.412,
+      "eval_steps_per_second": 4.098,
+      "step": 5775
+    },
+    {
+      "epoch": 308.27,
+      "learning_rate": 1.9876543209876546e-05,
+      "loss": 0.0401,
+      "step": 5780
+    },
+    {
+      "epoch": 308.8,
+      "learning_rate": 1.9814814814814816e-05,
+      "loss": 0.0397,
+      "step": 5790
+    },
+    {
+      "epoch": 308.96,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.943274974822998,
+      "eval_runtime": 4.6219,
+      "eval_samples_per_second": 129.817,
+      "eval_steps_per_second": 4.111,
+      "step": 5793
+    },
+    {
+      "epoch": 309.33,
+      "learning_rate": 1.9753086419753087e-05,
+      "loss": 0.0467,
+      "step": 5800
+    },
+    {
+      "epoch": 309.87,
+      "learning_rate": 1.969135802469136e-05,
+      "loss": 0.053,
+      "step": 5810
+    },
+    {
+      "epoch": 309.97,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.930065631866455,
+      "eval_runtime": 4.6916,
+      "eval_samples_per_second": 127.887,
+      "eval_steps_per_second": 4.05,
+      "step": 5812
+    },
+    {
+      "epoch": 310.4,
+      "learning_rate": 1.962962962962963e-05,
+      "loss": 0.0459,
+      "step": 5820
+    },
+    {
+      "epoch": 310.93,
+      "learning_rate": 1.95679012345679e-05,
+      "loss": 0.0303,
+      "step": 5830
+    },
+    {
+      "epoch": 310.99,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.8961286544799805,
+      "eval_runtime": 4.6651,
+      "eval_samples_per_second": 128.614,
+      "eval_steps_per_second": 4.073,
+      "step": 5831
+    },
+    {
+      "epoch": 311.47,
+      "learning_rate": 1.950617283950617e-05,
+      "loss": 0.0314,
+      "step": 5840
+    },
+    {
+      "epoch": 312.0,
+      "learning_rate": 1.9444444444444445e-05,
+      "loss": 0.0369,
+      "step": 5850
+    },
+    {
+      "epoch": 312.0,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.856044292449951,
+      "eval_runtime": 4.6488,
+      "eval_samples_per_second": 129.067,
+      "eval_steps_per_second": 4.087,
+      "step": 5850
+    },
+    {
+      "epoch": 312.53,
+      "learning_rate": 1.938271604938272e-05,
+      "loss": 0.0423,
+      "step": 5860
+    },
+    {
+      "epoch": 312.96,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.917734146118164,
+      "eval_runtime": 4.6571,
+      "eval_samples_per_second": 128.835,
+      "eval_steps_per_second": 4.08,
+      "step": 5868
+    },
+    {
+      "epoch": 313.07,
+      "learning_rate": 1.9320987654320986e-05,
+      "loss": 0.03,
+      "step": 5870
+    },
+    {
+      "epoch": 313.6,
+      "learning_rate": 1.925925925925926e-05,
+      "loss": 0.0343,
+      "step": 5880
+    },
+    {
+      "epoch": 313.97,
+      "eval_accuracy": 0.22333333333333333,
+      "eval_loss": 4.892765522003174,
+      "eval_runtime": 4.6848,
+      "eval_samples_per_second": 128.074,
+      "eval_steps_per_second": 4.056,
+      "step": 5887
+    },
+    {
+      "epoch": 314.13,
+      "learning_rate": 1.9197530864197534e-05,
+      "loss": 0.0299,
+      "step": 5890
+    },
+    {
+      "epoch": 314.67,
+      "learning_rate": 1.91358024691358e-05,
+      "loss": 0.0216,
+      "step": 5900
+    },
+    {
+      "epoch": 314.99,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.895847320556641,
+      "eval_runtime": 4.6722,
+      "eval_samples_per_second": 128.42,
+      "eval_steps_per_second": 4.067,
+      "step": 5906
+    },
+    {
+      "epoch": 315.2,
+      "learning_rate": 1.9074074074074075e-05,
+      "loss": 0.0604,
+      "step": 5910
+    },
+    {
+      "epoch": 315.73,
+      "learning_rate": 1.901234567901235e-05,
+      "loss": 0.0287,
+      "step": 5920
+    },
+    {
+      "epoch": 316.0,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.880258083343506,
+      "eval_runtime": 4.6757,
+      "eval_samples_per_second": 128.323,
+      "eval_steps_per_second": 4.064,
+      "step": 5925
+    },
+    {
+      "epoch": 316.27,
+      "learning_rate": 1.8950617283950615e-05,
+      "loss": 0.0269,
+      "step": 5930
+    },
+    {
+      "epoch": 316.8,
+      "learning_rate": 1.888888888888889e-05,
+      "loss": 0.0286,
+      "step": 5940
+    },
+    {
+      "epoch": 316.96,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.86151123046875,
+      "eval_runtime": 4.6507,
+      "eval_samples_per_second": 129.012,
+      "eval_steps_per_second": 4.085,
+      "step": 5943
+    },
+    {
+      "epoch": 317.33,
+      "learning_rate": 1.8827160493827163e-05,
+      "loss": 0.0478,
+      "step": 5950
+    },
+    {
+      "epoch": 317.87,
+      "learning_rate": 1.8765432098765433e-05,
+      "loss": 0.0304,
+      "step": 5960
+    },
+    {
+      "epoch": 317.97,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.873566150665283,
+      "eval_runtime": 4.6681,
+      "eval_samples_per_second": 128.532,
+      "eval_steps_per_second": 4.07,
+      "step": 5962
+    },
+    {
+      "epoch": 318.4,
+      "learning_rate": 1.8703703703703704e-05,
+      "loss": 0.0346,
+      "step": 5970
+    },
+    {
+      "epoch": 318.93,
+      "learning_rate": 1.8641975308641977e-05,
+      "loss": 0.0486,
+      "step": 5980
+    },
+    {
+      "epoch": 318.99,
+      "eval_accuracy": 0.22333333333333333,
+      "eval_loss": 4.882538318634033,
+      "eval_runtime": 4.6432,
+      "eval_samples_per_second": 129.222,
+      "eval_steps_per_second": 4.092,
+      "step": 5981
+    },
+    {
+      "epoch": 319.47,
+      "learning_rate": 1.8580246913580248e-05,
+      "loss": 0.0391,
+      "step": 5990
+    },
+    {
+      "epoch": 320.0,
+      "learning_rate": 1.8518518518518518e-05,
+      "loss": 0.0404,
+      "step": 6000
+    },
+    {
+      "epoch": 320.0,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.861847400665283,
+      "eval_runtime": 4.6741,
+      "eval_samples_per_second": 128.368,
+      "eval_steps_per_second": 4.065,
+      "step": 6000
+    },
+    {
+      "epoch": 320.53,
+      "learning_rate": 1.8456790123456792e-05,
+      "loss": 0.0439,
+      "step": 6010
+    },
+    {
+      "epoch": 320.96,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.884802341461182,
+      "eval_runtime": 4.66,
+      "eval_samples_per_second": 128.756,
+      "eval_steps_per_second": 4.077,
+      "step": 6018
+    },
+    {
+      "epoch": 321.07,
+      "learning_rate": 1.8395061728395062e-05,
+      "loss": 0.0387,
+      "step": 6020
+    },
+    {
+      "epoch": 321.6,
+      "learning_rate": 1.8333333333333333e-05,
+      "loss": 0.0428,
+      "step": 6030
+    },
+    {
+      "epoch": 321.97,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.897517681121826,
+      "eval_runtime": 4.6751,
+      "eval_samples_per_second": 128.34,
+      "eval_steps_per_second": 4.064,
+      "step": 6037
+    },
+    {
+      "epoch": 322.13,
+      "learning_rate": 1.8271604938271607e-05,
+      "loss": 0.0261,
+      "step": 6040
+    },
+    {
+      "epoch": 322.67,
+      "learning_rate": 1.8209876543209877e-05,
+      "loss": 0.0498,
+      "step": 6050
+    },
+    {
+      "epoch": 322.99,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.8614420890808105,
+      "eval_runtime": 4.6819,
+      "eval_samples_per_second": 128.153,
+      "eval_steps_per_second": 4.058,
+      "step": 6056
+    },
+    {
+      "epoch": 323.2,
+      "learning_rate": 1.814814814814815e-05,
+      "loss": 0.0403,
+      "step": 6060
+    },
+    {
+      "epoch": 323.73,
+      "learning_rate": 1.808641975308642e-05,
+      "loss": 0.0314,
+      "step": 6070
+    },
+    {
+      "epoch": 324.0,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.871830463409424,
+      "eval_runtime": 4.6693,
+      "eval_samples_per_second": 128.5,
+      "eval_steps_per_second": 4.069,
+      "step": 6075
+    },
+    {
+      "epoch": 324.27,
+      "learning_rate": 1.802469135802469e-05,
+      "loss": 0.0365,
+      "step": 6080
+    },
+    {
+      "epoch": 324.8,
+      "learning_rate": 1.7962962962962965e-05,
+      "loss": 0.0334,
+      "step": 6090
+    },
+    {
+      "epoch": 324.96,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.902100563049316,
+      "eval_runtime": 4.7082,
+      "eval_samples_per_second": 127.436,
+      "eval_steps_per_second": 4.035,
+      "step": 6093
+    },
+    {
+      "epoch": 325.33,
+      "learning_rate": 1.7901234567901236e-05,
+      "loss": 0.0446,
+      "step": 6100
+    },
+    {
+      "epoch": 325.87,
+      "learning_rate": 1.7839506172839506e-05,
+      "loss": 0.0431,
+      "step": 6110
+    },
+    {
+      "epoch": 325.97,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.897326946258545,
+      "eval_runtime": 4.7241,
+      "eval_samples_per_second": 127.008,
+      "eval_steps_per_second": 4.022,
+      "step": 6112
+    },
+    {
+      "epoch": 326.4,
+      "learning_rate": 1.777777777777778e-05,
+      "loss": 0.0316,
+      "step": 6120
+    },
+    {
+      "epoch": 326.93,
+      "learning_rate": 1.771604938271605e-05,
+      "loss": 0.0473,
+      "step": 6130
+    },
+    {
+      "epoch": 326.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.867129802703857,
+      "eval_runtime": 4.6999,
+      "eval_samples_per_second": 127.662,
+      "eval_steps_per_second": 4.043,
+      "step": 6131
+    },
+    {
+      "epoch": 327.47,
+      "learning_rate": 1.765432098765432e-05,
+      "loss": 0.0365,
+      "step": 6140
+    },
+    {
+      "epoch": 328.0,
+      "learning_rate": 1.7592592592592595e-05,
+      "loss": 0.0348,
+      "step": 6150
+    },
+    {
+      "epoch": 328.0,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.905031204223633,
+      "eval_runtime": 4.6993,
+      "eval_samples_per_second": 127.679,
+      "eval_steps_per_second": 4.043,
+      "step": 6150
+    },
+    {
+      "epoch": 328.53,
+      "learning_rate": 1.7530864197530865e-05,
+      "loss": 0.0718,
+      "step": 6160
+    },
+    {
+      "epoch": 328.96,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.886887073516846,
+      "eval_runtime": 4.7046,
+      "eval_samples_per_second": 127.535,
+      "eval_steps_per_second": 4.039,
+      "step": 6168
+    },
+    {
+      "epoch": 329.07,
+      "learning_rate": 1.7469135802469135e-05,
+      "loss": 0.0418,
+      "step": 6170
+    },
+    {
+      "epoch": 329.6,
+      "learning_rate": 1.740740740740741e-05,
+      "loss": 0.0387,
+      "step": 6180
+    },
+    {
+      "epoch": 329.97,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.855226039886475,
+      "eval_runtime": 4.7051,
+      "eval_samples_per_second": 127.52,
+      "eval_steps_per_second": 4.038,
+      "step": 6187
+    },
+    {
+      "epoch": 330.13,
+      "learning_rate": 1.734567901234568e-05,
+      "loss": 0.032,
+      "step": 6190
+    },
+    {
+      "epoch": 330.67,
+      "learning_rate": 1.728395061728395e-05,
+      "loss": 0.0335,
+      "step": 6200
+    },
+    {
+      "epoch": 330.99,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.893190860748291,
+      "eval_runtime": 4.7105,
+      "eval_samples_per_second": 127.376,
+      "eval_steps_per_second": 4.034,
+      "step": 6206
+    },
+    {
+      "epoch": 331.2,
+      "learning_rate": 1.7222222222222224e-05,
+      "loss": 0.0271,
+      "step": 6210
+    },
+    {
+      "epoch": 331.73,
+      "learning_rate": 1.7160493827160494e-05,
+      "loss": 0.0355,
+      "step": 6220
+    },
+    {
+      "epoch": 332.0,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.919488906860352,
+      "eval_runtime": 4.7258,
+      "eval_samples_per_second": 126.963,
+      "eval_steps_per_second": 4.02,
+      "step": 6225
+    },
+    {
+      "epoch": 332.27,
+      "learning_rate": 1.7098765432098765e-05,
+      "loss": 0.038,
+      "step": 6230
+    },
+    {
+      "epoch": 332.8,
+      "learning_rate": 1.7037037037037038e-05,
+      "loss": 0.0407,
+      "step": 6240
+    },
+    {
+      "epoch": 332.96,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.9162702560424805,
+      "eval_runtime": 4.7131,
+      "eval_samples_per_second": 127.305,
+      "eval_steps_per_second": 4.031,
+      "step": 6243
+    },
+    {
+      "epoch": 333.33,
+      "learning_rate": 1.697530864197531e-05,
+      "loss": 0.0434,
+      "step": 6250
+    },
+    {
+      "epoch": 333.87,
+      "learning_rate": 1.6913580246913582e-05,
+      "loss": 0.0471,
+      "step": 6260
+    },
+    {
+      "epoch": 333.97,
+      "eval_accuracy": 0.225,
+      "eval_loss": 4.885989189147949,
+      "eval_runtime": 4.7137,
+      "eval_samples_per_second": 127.289,
+      "eval_steps_per_second": 4.031,
+      "step": 6262
+    },
+    {
+      "epoch": 334.4,
+      "learning_rate": 1.6851851851851853e-05,
+      "loss": 0.0336,
+      "step": 6270
+    },
+    {
+      "epoch": 334.93,
+      "learning_rate": 1.6790123456790123e-05,
+      "loss": 0.0334,
+      "step": 6280
+    },
+    {
+      "epoch": 334.99,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.894328594207764,
+      "eval_runtime": 4.6988,
+      "eval_samples_per_second": 127.693,
+      "eval_steps_per_second": 4.044,
+      "step": 6281
+    },
+    {
+      "epoch": 335.47,
+      "learning_rate": 1.6728395061728397e-05,
+      "loss": 0.0441,
+      "step": 6290
+    },
+    {
+      "epoch": 336.0,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 0.0301,
+      "step": 6300
+    },
+    {
+      "epoch": 336.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.922252178192139,
+      "eval_runtime": 4.7453,
+      "eval_samples_per_second": 126.441,
+      "eval_steps_per_second": 4.004,
+      "step": 6300
+    },
+    {
+      "epoch": 336.53,
+      "learning_rate": 1.6604938271604938e-05,
+      "loss": 0.0281,
+      "step": 6310
+    },
+    {
+      "epoch": 336.96,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.910050392150879,
+      "eval_runtime": 4.7284,
+      "eval_samples_per_second": 126.892,
+      "eval_steps_per_second": 4.018,
+      "step": 6318
+    },
+    {
+      "epoch": 337.07,
+      "learning_rate": 1.654320987654321e-05,
+      "loss": 0.0365,
+      "step": 6320
+    },
+    {
+      "epoch": 337.6,
+      "learning_rate": 1.6481481481481482e-05,
+      "loss": 0.0305,
+      "step": 6330
+    },
+    {
+      "epoch": 337.97,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.889711856842041,
+      "eval_runtime": 4.7203,
+      "eval_samples_per_second": 127.111,
+      "eval_steps_per_second": 4.025,
+      "step": 6337
+    },
+    {
+      "epoch": 338.13,
+      "learning_rate": 1.6419753086419752e-05,
+      "loss": 0.0542,
+      "step": 6340
+    },
+    {
+      "epoch": 338.67,
+      "learning_rate": 1.6358024691358026e-05,
+      "loss": 0.0505,
+      "step": 6350
+    },
+    {
+      "epoch": 338.99,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.929032802581787,
+      "eval_runtime": 4.7098,
+      "eval_samples_per_second": 127.395,
+      "eval_steps_per_second": 4.034,
+      "step": 6356
+    },
+    {
+      "epoch": 339.2,
+      "learning_rate": 1.62962962962963e-05,
+      "loss": 0.0414,
+      "step": 6360
+    },
+    {
+      "epoch": 339.73,
+      "learning_rate": 1.6234567901234567e-05,
+      "loss": 0.024,
+      "step": 6370
+    },
+    {
+      "epoch": 340.0,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.944223880767822,
+      "eval_runtime": 4.7276,
+      "eval_samples_per_second": 126.914,
+      "eval_steps_per_second": 4.019,
+      "step": 6375
+    },
+    {
+      "epoch": 340.27,
+      "learning_rate": 1.617283950617284e-05,
+      "loss": 0.0267,
+      "step": 6380
+    },
+    {
+      "epoch": 340.8,
+      "learning_rate": 1.6111111111111115e-05,
+      "loss": 0.0504,
+      "step": 6390
+    },
+    {
+      "epoch": 340.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.918275833129883,
+      "eval_runtime": 4.709,
+      "eval_samples_per_second": 127.416,
+      "eval_steps_per_second": 4.035,
+      "step": 6393
+    },
+    {
+      "epoch": 341.33,
+      "learning_rate": 1.604938271604938e-05,
+      "loss": 0.0413,
+      "step": 6400
+    },
+    {
+      "epoch": 341.87,
+      "learning_rate": 1.5987654320987655e-05,
+      "loss": 0.0259,
+      "step": 6410
+    },
+    {
+      "epoch": 341.97,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.883179187774658,
+      "eval_runtime": 4.734,
+      "eval_samples_per_second": 126.742,
+      "eval_steps_per_second": 4.014,
+      "step": 6412
+    },
+    {
+      "epoch": 342.4,
+      "learning_rate": 1.5925925925925926e-05,
+      "loss": 0.0338,
+      "step": 6420
+    },
+    {
+      "epoch": 342.93,
+      "learning_rate": 1.5864197530864196e-05,
+      "loss": 0.0313,
+      "step": 6430
+    },
+    {
+      "epoch": 342.99,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.895809650421143,
+      "eval_runtime": 4.7766,
+      "eval_samples_per_second": 125.612,
+      "eval_steps_per_second": 3.978,
+      "step": 6431
+    },
+    {
+      "epoch": 343.47,
+      "learning_rate": 1.580246913580247e-05,
+      "loss": 0.0213,
+      "step": 6440
+    },
+    {
+      "epoch": 344.0,
+      "learning_rate": 1.574074074074074e-05,
+      "loss": 0.0293,
+      "step": 6450
+    },
+    {
+      "epoch": 344.0,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.897879123687744,
+      "eval_runtime": 4.8088,
+      "eval_samples_per_second": 124.771,
+      "eval_steps_per_second": 3.951,
+      "step": 6450
+    },
+    {
+      "epoch": 344.53,
+      "learning_rate": 1.5679012345679014e-05,
+      "loss": 0.0427,
+      "step": 6460
+    },
+    {
+      "epoch": 344.96,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.9055495262146,
+      "eval_runtime": 4.8414,
+      "eval_samples_per_second": 123.93,
+      "eval_steps_per_second": 3.924,
+      "step": 6468
+    },
+    {
+      "epoch": 345.07,
+      "learning_rate": 1.5617283950617285e-05,
+      "loss": 0.0344,
+      "step": 6470
+    },
+    {
+      "epoch": 345.6,
+      "learning_rate": 1.5555555555555555e-05,
+      "loss": 0.0399,
+      "step": 6480
+    },
+    {
+      "epoch": 345.97,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.895743370056152,
+      "eval_runtime": 4.7399,
+      "eval_samples_per_second": 126.584,
+      "eval_steps_per_second": 4.009,
+      "step": 6487
+    },
+    {
+      "epoch": 346.13,
+      "learning_rate": 1.549382716049383e-05,
+      "loss": 0.0253,
+      "step": 6490
+    },
+    {
+      "epoch": 346.67,
+      "learning_rate": 1.54320987654321e-05,
+      "loss": 0.0273,
+      "step": 6500
+    },
+    {
+      "epoch": 346.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.8988518714904785,
+      "eval_runtime": 4.7852,
+      "eval_samples_per_second": 125.387,
+      "eval_steps_per_second": 3.971,
+      "step": 6506
+    },
+    {
+      "epoch": 347.2,
+      "learning_rate": 1.537037037037037e-05,
+      "loss": 0.0261,
+      "step": 6510
+    },
+    {
+      "epoch": 347.73,
+      "learning_rate": 1.5308641975308643e-05,
+      "loss": 0.0388,
+      "step": 6520
+    },
+    {
+      "epoch": 348.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.9087018966674805,
+      "eval_runtime": 4.7683,
+      "eval_samples_per_second": 125.83,
+      "eval_steps_per_second": 3.985,
+      "step": 6525
+    },
+    {
+      "epoch": 348.27,
+      "learning_rate": 1.5246913580246914e-05,
+      "loss": 0.0323,
+      "step": 6530
+    },
+    {
+      "epoch": 348.8,
+      "learning_rate": 1.5185185185185186e-05,
+      "loss": 0.0306,
+      "step": 6540
+    },
+    {
+      "epoch": 348.96,
+      "eval_accuracy": 0.22833333333333333,
+      "eval_loss": 4.926441669464111,
+      "eval_runtime": 4.7772,
+      "eval_samples_per_second": 125.598,
+      "eval_steps_per_second": 3.977,
+      "step": 6543
+    },
+    {
+      "epoch": 349.33,
+      "learning_rate": 1.5123456790123458e-05,
+      "loss": 0.0249,
+      "step": 6550
+    },
+    {
+      "epoch": 349.87,
+      "learning_rate": 1.506172839506173e-05,
+      "loss": 0.0411,
+      "step": 6560
+    },
+    {
+      "epoch": 349.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.921908855438232,
+      "eval_runtime": 4.7531,
+      "eval_samples_per_second": 126.234,
+      "eval_steps_per_second": 3.997,
+      "step": 6562
+    },
+    {
+      "epoch": 350.4,
+      "learning_rate": 1.5e-05,
+      "loss": 0.031,
+      "step": 6570
+    },
+    {
+      "epoch": 350.93,
+      "learning_rate": 1.4938271604938272e-05,
+      "loss": 0.0394,
+      "step": 6580
+    },
+    {
+      "epoch": 350.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.89980936050415,
+      "eval_runtime": 4.7836,
+      "eval_samples_per_second": 125.428,
+      "eval_steps_per_second": 3.972,
+      "step": 6581
+    },
+    {
+      "epoch": 351.47,
+      "learning_rate": 1.4876543209876545e-05,
+      "loss": 0.0465,
+      "step": 6590
+    },
+    {
+      "epoch": 352.0,
+      "learning_rate": 1.4814814814814815e-05,
+      "loss": 0.0507,
+      "step": 6600
+    },
+    {
+      "epoch": 352.0,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.930387496948242,
+      "eval_runtime": 4.8303,
+      "eval_samples_per_second": 124.216,
+      "eval_steps_per_second": 3.934,
+      "step": 6600
+    },
+    {
+      "epoch": 352.53,
+      "learning_rate": 1.4753086419753087e-05,
+      "loss": 0.0263,
+      "step": 6610
+    },
+    {
+      "epoch": 352.96,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.923248767852783,
+      "eval_runtime": 4.822,
+      "eval_samples_per_second": 124.429,
+      "eval_steps_per_second": 3.94,
+      "step": 6618
+    },
+    {
+      "epoch": 353.07,
+      "learning_rate": 1.4691358024691359e-05,
+      "loss": 0.0355,
+      "step": 6620
+    },
+    {
+      "epoch": 353.6,
+      "learning_rate": 1.462962962962963e-05,
+      "loss": 0.0395,
+      "step": 6630
+    },
+    {
+      "epoch": 353.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.92411470413208,
+      "eval_runtime": 4.7975,
+      "eval_samples_per_second": 125.066,
+      "eval_steps_per_second": 3.96,
+      "step": 6637
+    },
+    {
+      "epoch": 354.13,
+      "learning_rate": 1.4567901234567902e-05,
+      "loss": 0.0258,
+      "step": 6640
+    },
+    {
+      "epoch": 354.67,
+      "learning_rate": 1.4506172839506174e-05,
+      "loss": 0.0394,
+      "step": 6650
+    },
+    {
+      "epoch": 354.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.926273345947266,
+      "eval_runtime": 4.7995,
+      "eval_samples_per_second": 125.013,
+      "eval_steps_per_second": 3.959,
+      "step": 6656
+    },
+    {
+      "epoch": 355.2,
+      "learning_rate": 1.4444444444444444e-05,
+      "loss": 0.0345,
+      "step": 6660
+    },
+    {
+      "epoch": 355.73,
+      "learning_rate": 1.4382716049382716e-05,
+      "loss": 0.0391,
+      "step": 6670
+    },
+    {
+      "epoch": 356.0,
+      "eval_accuracy": 0.26,
+      "eval_loss": 4.927285671234131,
+      "eval_runtime": 4.819,
+      "eval_samples_per_second": 124.507,
+      "eval_steps_per_second": 3.943,
+      "step": 6675
+    },
+    {
+      "epoch": 356.27,
+      "learning_rate": 1.4320987654320988e-05,
+      "loss": 0.0274,
+      "step": 6680
+    },
+    {
+      "epoch": 356.8,
+      "learning_rate": 1.425925925925926e-05,
+      "loss": 0.0647,
+      "step": 6690
+    },
+    {
+      "epoch": 356.96,
+      "eval_accuracy": 0.2633333333333333,
+      "eval_loss": 4.903398513793945,
+      "eval_runtime": 4.8546,
+      "eval_samples_per_second": 123.593,
+      "eval_steps_per_second": 3.914,
+      "step": 6693
+    },
+    {
+      "epoch": 357.33,
+      "learning_rate": 1.419753086419753e-05,
+      "loss": 0.0427,
+      "step": 6700
+    },
+    {
+      "epoch": 357.87,
+      "learning_rate": 1.4135802469135803e-05,
+      "loss": 0.038,
+      "step": 6710
+    },
+    {
+      "epoch": 357.97,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.891026496887207,
+      "eval_runtime": 4.8918,
+      "eval_samples_per_second": 122.655,
+      "eval_steps_per_second": 3.884,
+      "step": 6712
+    },
+    {
+      "epoch": 358.4,
+      "learning_rate": 1.4074074074074075e-05,
+      "loss": 0.0254,
+      "step": 6720
+    },
+    {
+      "epoch": 358.93,
+      "learning_rate": 1.4012345679012345e-05,
+      "loss": 0.0368,
+      "step": 6730
+    },
+    {
+      "epoch": 358.99,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.883033275604248,
+      "eval_runtime": 4.8388,
+      "eval_samples_per_second": 123.997,
+      "eval_steps_per_second": 3.927,
+      "step": 6731
+    },
+    {
+      "epoch": 359.47,
+      "learning_rate": 1.3950617283950617e-05,
+      "loss": 0.0278,
+      "step": 6740
+    },
+    {
+      "epoch": 360.0,
+      "learning_rate": 1.388888888888889e-05,
+      "loss": 0.0308,
+      "step": 6750
+    },
+    {
+      "epoch": 360.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.886683940887451,
+      "eval_runtime": 4.8656,
+      "eval_samples_per_second": 123.316,
+      "eval_steps_per_second": 3.905,
+      "step": 6750
+    },
+    {
+      "epoch": 360.53,
+      "learning_rate": 1.382716049382716e-05,
+      "loss": 0.0346,
+      "step": 6760
+    },
+    {
+      "epoch": 360.96,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.86568021774292,
+      "eval_runtime": 4.8606,
+      "eval_samples_per_second": 123.442,
+      "eval_steps_per_second": 3.909,
+      "step": 6768
+    },
+    {
+      "epoch": 361.07,
+      "learning_rate": 1.3765432098765432e-05,
+      "loss": 0.0254,
+      "step": 6770
+    },
+    {
+      "epoch": 361.6,
+      "learning_rate": 1.3703703703703704e-05,
+      "loss": 0.0279,
+      "step": 6780
+    },
+    {
+      "epoch": 361.97,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.8677825927734375,
+      "eval_runtime": 4.8214,
+      "eval_samples_per_second": 124.446,
+      "eval_steps_per_second": 3.941,
+      "step": 6787
+    },
+    {
+      "epoch": 362.13,
+      "learning_rate": 1.3641975308641978e-05,
+      "loss": 0.0367,
+      "step": 6790
+    },
+    {
+      "epoch": 362.67,
+      "learning_rate": 1.3580246913580247e-05,
+      "loss": 0.0443,
+      "step": 6800
+    },
+    {
+      "epoch": 362.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.872296333312988,
+      "eval_runtime": 4.8453,
+      "eval_samples_per_second": 123.832,
+      "eval_steps_per_second": 3.921,
+      "step": 6806
+    },
+    {
+      "epoch": 363.2,
+      "learning_rate": 1.3518518518518519e-05,
+      "loss": 0.0224,
+      "step": 6810
+    },
+    {
+      "epoch": 363.73,
+      "learning_rate": 1.3456790123456793e-05,
+      "loss": 0.027,
+      "step": 6820
+    },
+    {
+      "epoch": 364.0,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.875555515289307,
+      "eval_runtime": 4.8571,
+      "eval_samples_per_second": 123.529,
+      "eval_steps_per_second": 3.912,
+      "step": 6825
+    },
+    {
+      "epoch": 364.27,
+      "learning_rate": 1.3395061728395061e-05,
+      "loss": 0.0394,
+      "step": 6830
+    },
+    {
+      "epoch": 364.8,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 0.0447,
+      "step": 6840
+    },
+    {
+      "epoch": 364.96,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.874227046966553,
+      "eval_runtime": 4.8623,
+      "eval_samples_per_second": 123.397,
+      "eval_steps_per_second": 3.908,
+      "step": 6843
+    },
+    {
+      "epoch": 365.33,
+      "learning_rate": 1.3271604938271605e-05,
+      "loss": 0.0402,
+      "step": 6850
+    },
+    {
+      "epoch": 365.87,
+      "learning_rate": 1.3209876543209876e-05,
+      "loss": 0.028,
+      "step": 6860
+    },
+    {
+      "epoch": 365.97,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.904233455657959,
+      "eval_runtime": 4.8676,
+      "eval_samples_per_second": 123.264,
+      "eval_steps_per_second": 3.903,
+      "step": 6862
+    },
+    {
+      "epoch": 366.4,
+      "learning_rate": 1.3148148148148148e-05,
+      "loss": 0.0306,
+      "step": 6870
+    },
+    {
+      "epoch": 366.93,
+      "learning_rate": 1.308641975308642e-05,
+      "loss": 0.0483,
+      "step": 6880
+    },
+    {
+      "epoch": 366.99,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.908579349517822,
+      "eval_runtime": 4.8554,
+      "eval_samples_per_second": 123.573,
+      "eval_steps_per_second": 3.913,
+      "step": 6881
+    },
+    {
+      "epoch": 367.47,
+      "learning_rate": 1.3024691358024694e-05,
+      "loss": 0.0321,
+      "step": 6890
+    },
+    {
+      "epoch": 368.0,
+      "learning_rate": 1.2962962962962962e-05,
+      "loss": 0.034,
+      "step": 6900
+    },
+    {
+      "epoch": 368.0,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.888582229614258,
+      "eval_runtime": 4.9075,
+      "eval_samples_per_second": 122.261,
+      "eval_steps_per_second": 3.872,
+      "step": 6900
+    },
+    {
+      "epoch": 368.53,
+      "learning_rate": 1.2901234567901235e-05,
+      "loss": 0.0363,
+      "step": 6910
+    },
+    {
+      "epoch": 368.96,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.877806186676025,
+      "eval_runtime": 4.9146,
+      "eval_samples_per_second": 122.085,
+      "eval_steps_per_second": 3.866,
+      "step": 6918
+    },
+    {
+      "epoch": 369.07,
+      "learning_rate": 1.2839506172839508e-05,
+      "loss": 0.0401,
+      "step": 6920
+    },
+    {
+      "epoch": 369.6,
+      "learning_rate": 1.2777777777777777e-05,
+      "loss": 0.0417,
+      "step": 6930
+    },
+    {
+      "epoch": 369.97,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.905084609985352,
+      "eval_runtime": 4.8859,
+      "eval_samples_per_second": 122.802,
+      "eval_steps_per_second": 3.889,
+      "step": 6937
+    },
+    {
+      "epoch": 370.13,
+      "learning_rate": 1.2716049382716049e-05,
+      "loss": 0.0338,
+      "step": 6940
+    },
+    {
+      "epoch": 370.67,
+      "learning_rate": 1.2654320987654323e-05,
+      "loss": 0.0326,
+      "step": 6950
+    },
+    {
+      "epoch": 370.99,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.911184310913086,
+      "eval_runtime": 4.8779,
+      "eval_samples_per_second": 123.003,
+      "eval_steps_per_second": 3.895,
+      "step": 6956
+    },
+    {
+      "epoch": 371.2,
+      "learning_rate": 1.2592592592592592e-05,
+      "loss": 0.0261,
+      "step": 6960
+    },
+    {
+      "epoch": 371.73,
+      "learning_rate": 1.2530864197530864e-05,
+      "loss": 0.028,
+      "step": 6970
+    },
+    {
+      "epoch": 372.0,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.911579608917236,
+      "eval_runtime": 4.906,
+      "eval_samples_per_second": 122.299,
+      "eval_steps_per_second": 3.873,
+      "step": 6975
+    },
+    {
+      "epoch": 372.27,
+      "learning_rate": 1.2469135802469137e-05,
+      "loss": 0.0196,
+      "step": 6980
+    },
+    {
+      "epoch": 372.8,
+      "learning_rate": 1.2407407407407408e-05,
+      "loss": 0.0343,
+      "step": 6990
+    },
+    {
+      "epoch": 372.96,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.910400390625,
+      "eval_runtime": 4.8963,
+      "eval_samples_per_second": 122.541,
+      "eval_steps_per_second": 3.88,
+      "step": 6993
+    },
+    {
+      "epoch": 373.33,
+      "learning_rate": 1.2345679012345678e-05,
+      "loss": 0.0246,
+      "step": 7000
+    },
+    {
+      "epoch": 373.87,
+      "learning_rate": 1.2283950617283952e-05,
+      "loss": 0.0229,
+      "step": 7010
+    },
+    {
+      "epoch": 373.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.940114974975586,
+      "eval_runtime": 4.8862,
+      "eval_samples_per_second": 122.794,
+      "eval_steps_per_second": 3.888,
+      "step": 7012
+    },
+    {
+      "epoch": 374.4,
+      "learning_rate": 1.2222222222222222e-05,
+      "loss": 0.0235,
+      "step": 7020
+    },
+    {
+      "epoch": 374.93,
+      "learning_rate": 1.2160493827160495e-05,
+      "loss": 0.0337,
+      "step": 7030
+    },
+    {
+      "epoch": 374.99,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.934114933013916,
+      "eval_runtime": 4.9143,
+      "eval_samples_per_second": 122.092,
+      "eval_steps_per_second": 3.866,
+      "step": 7031
+    },
+    {
+      "epoch": 375.47,
+      "learning_rate": 1.2098765432098767e-05,
+      "loss": 0.0334,
+      "step": 7040
+    },
+    {
+      "epoch": 376.0,
+      "learning_rate": 1.2037037037037037e-05,
+      "loss": 0.0356,
+      "step": 7050
+    },
+    {
+      "epoch": 376.0,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.933629035949707,
+      "eval_runtime": 4.9706,
+      "eval_samples_per_second": 120.709,
+      "eval_steps_per_second": 3.822,
+      "step": 7050
+    },
+    {
+      "epoch": 376.53,
+      "learning_rate": 1.1975308641975309e-05,
+      "loss": 0.029,
+      "step": 7060
+    },
+    {
+      "epoch": 376.96,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.913231372833252,
+      "eval_runtime": 4.888,
+      "eval_samples_per_second": 122.75,
+      "eval_steps_per_second": 3.887,
+      "step": 7068
+    },
+    {
+      "epoch": 377.07,
+      "learning_rate": 1.1913580246913581e-05,
+      "loss": 0.0348,
+      "step": 7070
+    },
+    {
+      "epoch": 377.6,
+      "learning_rate": 1.1851851851851853e-05,
+      "loss": 0.0272,
+      "step": 7080
+    },
+    {
+      "epoch": 377.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.910186290740967,
+      "eval_runtime": 4.9298,
+      "eval_samples_per_second": 121.709,
+      "eval_steps_per_second": 3.854,
+      "step": 7087
+    },
+    {
+      "epoch": 378.13,
+      "learning_rate": 1.1790123456790124e-05,
+      "loss": 0.0269,
+      "step": 7090
+    },
+    {
+      "epoch": 378.67,
+      "learning_rate": 1.1728395061728396e-05,
+      "loss": 0.0256,
+      "step": 7100
+    },
+    {
+      "epoch": 378.99,
+      "eval_accuracy": 0.23166666666666666,
+      "eval_loss": 4.925504207611084,
+      "eval_runtime": 4.948,
+      "eval_samples_per_second": 121.26,
+      "eval_steps_per_second": 3.84,
+      "step": 7106
+    },
+    {
+      "epoch": 379.2,
+      "learning_rate": 1.1666666666666668e-05,
+      "loss": 0.0315,
+      "step": 7110
+    },
+    {
+      "epoch": 379.73,
+      "learning_rate": 1.1604938271604938e-05,
+      "loss": 0.0276,
+      "step": 7120
+    },
+    {
+      "epoch": 380.0,
+      "eval_accuracy": 0.22666666666666666,
+      "eval_loss": 4.928166389465332,
+      "eval_runtime": 4.9278,
+      "eval_samples_per_second": 121.757,
+      "eval_steps_per_second": 3.856,
+      "step": 7125
+    },
+    {
+      "epoch": 380.27,
+      "learning_rate": 1.154320987654321e-05,
+      "loss": 0.0286,
+      "step": 7130
+    },
+    {
+      "epoch": 380.8,
+      "learning_rate": 1.1481481481481482e-05,
+      "loss": 0.026,
+      "step": 7140
+    },
+    {
+      "epoch": 380.96,
+      "eval_accuracy": 0.22,
+      "eval_loss": 4.952660083770752,
+      "eval_runtime": 4.9502,
+      "eval_samples_per_second": 121.208,
+      "eval_steps_per_second": 3.838,
+      "step": 7143
+    },
+    {
+      "epoch": 381.33,
+      "learning_rate": 1.1419753086419753e-05,
+      "loss": 0.031,
+      "step": 7150
+    },
+    {
+      "epoch": 381.87,
+      "learning_rate": 1.1358024691358025e-05,
+      "loss": 0.0385,
+      "step": 7160
+    },
+    {
+      "epoch": 381.97,
+      "eval_accuracy": 0.22166666666666668,
+      "eval_loss": 4.941068649291992,
+      "eval_runtime": 4.9601,
+      "eval_samples_per_second": 120.966,
+      "eval_steps_per_second": 3.831,
+      "step": 7162
+    },
+    {
+      "epoch": 382.4,
+      "learning_rate": 1.1296296296296297e-05,
+      "loss": 0.0269,
+      "step": 7170
+    },
+    {
+      "epoch": 382.93,
+      "learning_rate": 1.123456790123457e-05,
+      "loss": 0.026,
+      "step": 7180
+    },
+    {
+      "epoch": 382.99,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.952951908111572,
+      "eval_runtime": 4.9497,
+      "eval_samples_per_second": 121.22,
+      "eval_steps_per_second": 3.839,
+      "step": 7181
+    },
+    {
+      "epoch": 383.47,
+      "learning_rate": 1.117283950617284e-05,
+      "loss": 0.0262,
+      "step": 7190
+    },
+    {
+      "epoch": 384.0,
+      "learning_rate": 1.1111111111111112e-05,
+      "loss": 0.0444,
+      "step": 7200
+    },
+    {
+      "epoch": 384.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.9387335777282715,
+      "eval_runtime": 4.9515,
+      "eval_samples_per_second": 121.176,
+      "eval_steps_per_second": 3.837,
+      "step": 7200
+    },
+    {
+      "epoch": 384.53,
+      "learning_rate": 1.1049382716049384e-05,
+      "loss": 0.0369,
+      "step": 7210
+    },
+    {
+      "epoch": 384.96,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.904226303100586,
+      "eval_runtime": 4.9498,
+      "eval_samples_per_second": 121.217,
+      "eval_steps_per_second": 3.839,
+      "step": 7218
+    },
+    {
+      "epoch": 385.07,
+      "learning_rate": 1.0987654320987654e-05,
+      "loss": 0.0258,
+      "step": 7220
+    },
+    {
+      "epoch": 385.6,
+      "learning_rate": 1.0925925925925926e-05,
+      "loss": 0.0203,
+      "step": 7230
+    },
+    {
+      "epoch": 385.97,
+      "eval_accuracy": 0.23,
+      "eval_loss": 4.886034965515137,
+      "eval_runtime": 4.9753,
+      "eval_samples_per_second": 120.596,
+      "eval_steps_per_second": 3.819,
+      "step": 7237
+    },
+    {
+      "epoch": 386.13,
+      "learning_rate": 1.0864197530864198e-05,
+      "loss": 0.0332,
+      "step": 7240
+    },
+    {
+      "epoch": 386.67,
+      "learning_rate": 1.0802469135802469e-05,
+      "loss": 0.0238,
+      "step": 7250
+    },
+    {
+      "epoch": 386.99,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.877529144287109,
+      "eval_runtime": 4.9631,
+      "eval_samples_per_second": 120.891,
+      "eval_steps_per_second": 3.828,
+      "step": 7256
+    },
+    {
+      "epoch": 387.2,
+      "learning_rate": 1.074074074074074e-05,
+      "loss": 0.0266,
+      "step": 7260
+    },
+    {
+      "epoch": 387.73,
+      "learning_rate": 1.0679012345679013e-05,
+      "loss": 0.0315,
+      "step": 7270
+    },
+    {
+      "epoch": 388.0,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.864088535308838,
+      "eval_runtime": 4.9534,
+      "eval_samples_per_second": 121.128,
+      "eval_steps_per_second": 3.836,
+      "step": 7275
+    },
+    {
+      "epoch": 388.27,
+      "learning_rate": 1.0617283950617285e-05,
+      "loss": 0.0208,
+      "step": 7280
+    },
+    {
+      "epoch": 388.8,
+      "learning_rate": 1.0555555555555555e-05,
+      "loss": 0.0349,
+      "step": 7290
+    },
+    {
+      "epoch": 388.96,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.86765718460083,
+      "eval_runtime": 4.9642,
+      "eval_samples_per_second": 120.866,
+      "eval_steps_per_second": 3.827,
+      "step": 7293
+    },
+    {
+      "epoch": 389.33,
+      "learning_rate": 1.0493827160493827e-05,
+      "loss": 0.0336,
+      "step": 7300
+    },
+    {
+      "epoch": 389.87,
+      "learning_rate": 1.04320987654321e-05,
+      "loss": 0.038,
+      "step": 7310
+    },
+    {
+      "epoch": 389.97,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.868815898895264,
+      "eval_runtime": 4.9771,
+      "eval_samples_per_second": 120.552,
+      "eval_steps_per_second": 3.817,
+      "step": 7312
+    },
+    {
+      "epoch": 390.4,
+      "learning_rate": 1.037037037037037e-05,
+      "loss": 0.0188,
+      "step": 7320
+    },
+    {
+      "epoch": 390.93,
+      "learning_rate": 1.0308641975308644e-05,
+      "loss": 0.0301,
+      "step": 7330
+    },
+    {
+      "epoch": 390.99,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.8932037353515625,
+      "eval_runtime": 4.9575,
+      "eval_samples_per_second": 121.03,
+      "eval_steps_per_second": 3.833,
+      "step": 7331
+    },
+    {
+      "epoch": 391.47,
+      "learning_rate": 1.0246913580246914e-05,
+      "loss": 0.0418,
+      "step": 7340
+    },
+    {
+      "epoch": 392.0,
+      "learning_rate": 1.0185185185185185e-05,
+      "loss": 0.0363,
+      "step": 7350
+    },
+    {
+      "epoch": 392.0,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.902304172515869,
+      "eval_runtime": 4.9867,
+      "eval_samples_per_second": 120.321,
+      "eval_steps_per_second": 3.81,
+      "step": 7350
+    },
+    {
+      "epoch": 392.53,
+      "learning_rate": 1.0123456790123458e-05,
+      "loss": 0.0329,
+      "step": 7360
+    },
+    {
+      "epoch": 392.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.882464408874512,
+      "eval_runtime": 4.9782,
+      "eval_samples_per_second": 120.525,
+      "eval_steps_per_second": 3.817,
+      "step": 7368
+    },
+    {
+      "epoch": 393.07,
+      "learning_rate": 1.0061728395061729e-05,
+      "loss": 0.0255,
+      "step": 7370
+    },
+    {
+      "epoch": 393.6,
+      "learning_rate": 1e-05,
+      "loss": 0.0174,
+      "step": 7380
+    },
+    {
+      "epoch": 393.97,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.87109375,
+      "eval_runtime": 4.9821,
+      "eval_samples_per_second": 120.431,
+      "eval_steps_per_second": 3.814,
+      "step": 7387
+    },
+    {
+      "epoch": 394.13,
+      "learning_rate": 9.938271604938273e-06,
+      "loss": 0.0256,
+      "step": 7390
+    },
+    {
+      "epoch": 394.67,
+      "learning_rate": 9.876543209876543e-06,
+      "loss": 0.0284,
+      "step": 7400
+    },
+    {
+      "epoch": 394.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.876201152801514,
+      "eval_runtime": 5.0563,
+      "eval_samples_per_second": 118.663,
+      "eval_steps_per_second": 3.758,
+      "step": 7406
+    },
+    {
+      "epoch": 395.2,
+      "learning_rate": 9.814814814814815e-06,
+      "loss": 0.0458,
+      "step": 7410
+    },
+    {
+      "epoch": 395.73,
+      "learning_rate": 9.753086419753086e-06,
+      "loss": 0.0178,
+      "step": 7420
+    },
+    {
+      "epoch": 396.0,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.868426322937012,
+      "eval_runtime": 5.0279,
+      "eval_samples_per_second": 119.333,
+      "eval_steps_per_second": 3.779,
+      "step": 7425
+    },
+    {
+      "epoch": 396.27,
+      "learning_rate": 9.69135802469136e-06,
+      "loss": 0.0154,
+      "step": 7430
+    },
+    {
+      "epoch": 396.8,
+      "learning_rate": 9.62962962962963e-06,
+      "loss": 0.0359,
+      "step": 7440
+    },
+    {
+      "epoch": 396.96,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.865981578826904,
+      "eval_runtime": 5.0157,
+      "eval_samples_per_second": 119.625,
+      "eval_steps_per_second": 3.788,
+      "step": 7443
+    },
+    {
+      "epoch": 397.33,
+      "learning_rate": 9.5679012345679e-06,
+      "loss": 0.0295,
+      "step": 7450
+    },
+    {
+      "epoch": 397.87,
+      "learning_rate": 9.506172839506174e-06,
+      "loss": 0.029,
+      "step": 7460
+    },
+    {
+      "epoch": 397.97,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.879904270172119,
+      "eval_runtime": 5.0461,
+      "eval_samples_per_second": 118.904,
+      "eval_steps_per_second": 3.765,
+      "step": 7462
+    },
+    {
+      "epoch": 398.4,
+      "learning_rate": 9.444444444444445e-06,
+      "loss": 0.0368,
+      "step": 7470
+    },
+    {
+      "epoch": 398.93,
+      "learning_rate": 9.382716049382717e-06,
+      "loss": 0.0227,
+      "step": 7480
+    },
+    {
+      "epoch": 398.99,
+      "eval_accuracy": 0.25,
+      "eval_loss": 4.884500503540039,
+      "eval_runtime": 5.007,
+      "eval_samples_per_second": 119.833,
+      "eval_steps_per_second": 3.795,
+      "step": 7481
+    },
+    {
+      "epoch": 399.47,
+      "learning_rate": 9.320987654320989e-06,
+      "loss": 0.0338,
+      "step": 7490
+    },
+    {
+      "epoch": 400.0,
+      "learning_rate": 9.259259259259259e-06,
+      "loss": 0.0135,
+      "step": 7500
+    },
+    {
+      "epoch": 400.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.889830112457275,
+      "eval_runtime": 5.0086,
+      "eval_samples_per_second": 119.793,
+      "eval_steps_per_second": 3.793,
+      "step": 7500
+    },
+    {
+      "epoch": 400.53,
+      "learning_rate": 9.197530864197531e-06,
+      "loss": 0.0297,
+      "step": 7510
+    },
+    {
+      "epoch": 400.96,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.896719455718994,
+      "eval_runtime": 5.0281,
+      "eval_samples_per_second": 119.329,
+      "eval_steps_per_second": 3.779,
+      "step": 7518
+    },
+    {
+      "epoch": 401.07,
+      "learning_rate": 9.135802469135803e-06,
+      "loss": 0.0364,
+      "step": 7520
+    },
+    {
+      "epoch": 401.6,
+      "learning_rate": 9.074074074074075e-06,
+      "loss": 0.0263,
+      "step": 7530
+    },
+    {
+      "epoch": 401.97,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.888412952423096,
+      "eval_runtime": 5.0155,
+      "eval_samples_per_second": 119.63,
+      "eval_steps_per_second": 3.788,
+      "step": 7537
+    },
+    {
+      "epoch": 402.13,
+      "learning_rate": 9.012345679012346e-06,
+      "loss": 0.035,
+      "step": 7540
+    },
+    {
+      "epoch": 402.67,
+      "learning_rate": 8.950617283950618e-06,
+      "loss": 0.0386,
+      "step": 7550
+    },
+    {
+      "epoch": 402.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.871886730194092,
+      "eval_runtime": 4.9851,
+      "eval_samples_per_second": 120.358,
+      "eval_steps_per_second": 3.811,
+      "step": 7556
+    },
+    {
+      "epoch": 403.2,
+      "learning_rate": 8.88888888888889e-06,
+      "loss": 0.0339,
+      "step": 7560
+    },
+    {
+      "epoch": 403.73,
+      "learning_rate": 8.82716049382716e-06,
+      "loss": 0.0298,
+      "step": 7570
+    },
+    {
+      "epoch": 404.0,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.860945701599121,
+      "eval_runtime": 5.0553,
+      "eval_samples_per_second": 118.688,
+      "eval_steps_per_second": 3.758,
+      "step": 7575
+    },
+    {
+      "epoch": 404.27,
+      "learning_rate": 8.765432098765432e-06,
+      "loss": 0.0318,
+      "step": 7580
+    },
+    {
+      "epoch": 404.8,
+      "learning_rate": 8.703703703703705e-06,
+      "loss": 0.0232,
+      "step": 7590
+    },
+    {
+      "epoch": 404.96,
+      "eval_accuracy": 0.24833333333333332,
+      "eval_loss": 4.860249996185303,
+      "eval_runtime": 5.0445,
+      "eval_samples_per_second": 118.942,
+      "eval_steps_per_second": 3.767,
+      "step": 7593
+    },
+    {
+      "epoch": 405.33,
+      "learning_rate": 8.641975308641975e-06,
+      "loss": 0.0367,
+      "step": 7600
+    },
+    {
+      "epoch": 405.87,
+      "learning_rate": 8.580246913580247e-06,
+      "loss": 0.0232,
+      "step": 7610
+    },
+    {
+      "epoch": 405.97,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.866739749908447,
+      "eval_runtime": 5.0222,
+      "eval_samples_per_second": 119.47,
+      "eval_steps_per_second": 3.783,
+      "step": 7612
+    },
+    {
+      "epoch": 406.4,
+      "learning_rate": 8.518518518518519e-06,
+      "loss": 0.0223,
+      "step": 7620
+    },
+    {
+      "epoch": 406.93,
+      "learning_rate": 8.456790123456791e-06,
+      "loss": 0.032,
+      "step": 7630
+    },
+    {
+      "epoch": 406.99,
+      "eval_accuracy": 0.24833333333333332,
+      "eval_loss": 4.8684258460998535,
+      "eval_runtime": 5.0294,
+      "eval_samples_per_second": 119.298,
+      "eval_steps_per_second": 3.778,
+      "step": 7631
+    },
+    {
+      "epoch": 407.47,
+      "learning_rate": 8.395061728395062e-06,
+      "loss": 0.0272,
+      "step": 7640
+    },
+    {
+      "epoch": 408.0,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.0306,
+      "step": 7650
+    },
+    {
+      "epoch": 408.0,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.875509262084961,
+      "eval_runtime": 5.0495,
+      "eval_samples_per_second": 118.823,
+      "eval_steps_per_second": 3.763,
+      "step": 7650
+    },
+    {
+      "epoch": 408.53,
+      "learning_rate": 8.271604938271606e-06,
+      "loss": 0.0299,
+      "step": 7660
+    },
+    {
+      "epoch": 408.96,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.868679046630859,
+      "eval_runtime": 5.094,
+      "eval_samples_per_second": 117.785,
+      "eval_steps_per_second": 3.73,
+      "step": 7668
+    },
+    {
+      "epoch": 409.07,
+      "learning_rate": 8.209876543209876e-06,
+      "loss": 0.0402,
+      "step": 7670
+    },
+    {
+      "epoch": 409.6,
+      "learning_rate": 8.14814814814815e-06,
+      "loss": 0.0307,
+      "step": 7680
+    },
+    {
+      "epoch": 409.97,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.872376441955566,
+      "eval_runtime": 5.0699,
+      "eval_samples_per_second": 118.345,
+      "eval_steps_per_second": 3.748,
+      "step": 7687
+    },
+    {
+      "epoch": 410.13,
+      "learning_rate": 8.08641975308642e-06,
+      "loss": 0.0315,
+      "step": 7690
+    },
+    {
+      "epoch": 410.67,
+      "learning_rate": 8.02469135802469e-06,
+      "loss": 0.0304,
+      "step": 7700
+    },
+    {
+      "epoch": 410.99,
+      "eval_accuracy": 0.25,
+      "eval_loss": 4.879815101623535,
+      "eval_runtime": 5.0824,
+      "eval_samples_per_second": 118.054,
+      "eval_steps_per_second": 3.738,
+      "step": 7706
+    },
+    {
+      "epoch": 411.2,
+      "learning_rate": 7.962962962962963e-06,
+      "loss": 0.039,
+      "step": 7710
+    },
+    {
+      "epoch": 411.73,
+      "learning_rate": 7.901234567901235e-06,
+      "loss": 0.0293,
+      "step": 7720
+    },
+    {
+      "epoch": 412.0,
+      "eval_accuracy": 0.24833333333333332,
+      "eval_loss": 4.890056133270264,
+      "eval_runtime": 5.1118,
+      "eval_samples_per_second": 117.376,
+      "eval_steps_per_second": 3.717,
+      "step": 7725
+    },
+    {
+      "epoch": 412.27,
+      "learning_rate": 7.839506172839507e-06,
+      "loss": 0.0276,
+      "step": 7730
+    },
+    {
+      "epoch": 412.8,
+      "learning_rate": 7.777777777777777e-06,
+      "loss": 0.0273,
+      "step": 7740
+    },
+    {
+      "epoch": 412.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.902527332305908,
+      "eval_runtime": 5.0871,
+      "eval_samples_per_second": 117.946,
+      "eval_steps_per_second": 3.735,
+      "step": 7743
+    },
+    {
+      "epoch": 413.33,
+      "learning_rate": 7.71604938271605e-06,
+      "loss": 0.0334,
+      "step": 7750
+    },
+    {
+      "epoch": 413.87,
+      "learning_rate": 7.654320987654322e-06,
+      "loss": 0.0184,
+      "step": 7760
+    },
+    {
+      "epoch": 413.97,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.886964321136475,
+      "eval_runtime": 5.0738,
+      "eval_samples_per_second": 118.254,
+      "eval_steps_per_second": 3.745,
+      "step": 7762
+    },
+    {
+      "epoch": 414.4,
+      "learning_rate": 7.592592592592593e-06,
+      "loss": 0.0227,
+      "step": 7770
+    },
+    {
+      "epoch": 414.93,
+      "learning_rate": 7.530864197530865e-06,
+      "loss": 0.0377,
+      "step": 7780
+    },
+    {
+      "epoch": 414.99,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.890088081359863,
+      "eval_runtime": 5.0985,
+      "eval_samples_per_second": 117.682,
+      "eval_steps_per_second": 3.727,
+      "step": 7781
+    },
+    {
+      "epoch": 415.47,
+      "learning_rate": 7.469135802469136e-06,
+      "loss": 0.0284,
+      "step": 7790
+    },
+    {
+      "epoch": 416.0,
+      "learning_rate": 7.4074074074074075e-06,
+      "loss": 0.0278,
+      "step": 7800
+    },
+    {
+      "epoch": 416.0,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.889472484588623,
+      "eval_runtime": 5.0982,
+      "eval_samples_per_second": 117.689,
+      "eval_steps_per_second": 3.727,
+      "step": 7800
+    },
+    {
+      "epoch": 416.53,
+      "learning_rate": 7.3456790123456796e-06,
+      "loss": 0.0345,
+      "step": 7810
+    },
+    {
+      "epoch": 416.96,
+      "eval_accuracy": 0.25333333333333335,
+      "eval_loss": 4.904553413391113,
+      "eval_runtime": 5.0835,
+      "eval_samples_per_second": 118.028,
+      "eval_steps_per_second": 3.738,
+      "step": 7818
+    },
+    {
+      "epoch": 417.07,
+      "learning_rate": 7.283950617283951e-06,
+      "loss": 0.0376,
+      "step": 7820
+    },
+    {
+      "epoch": 417.6,
+      "learning_rate": 7.222222222222222e-06,
+      "loss": 0.0301,
+      "step": 7830
+    },
+    {
+      "epoch": 417.97,
+      "eval_accuracy": 0.24833333333333332,
+      "eval_loss": 4.900204181671143,
+      "eval_runtime": 5.1232,
+      "eval_samples_per_second": 117.114,
+      "eval_steps_per_second": 3.709,
+      "step": 7837
+    },
+    {
+      "epoch": 418.13,
+      "learning_rate": 7.160493827160494e-06,
+      "loss": 0.0204,
+      "step": 7840
+    },
+    {
+      "epoch": 418.67,
+      "learning_rate": 7.098765432098765e-06,
+      "loss": 0.0159,
+      "step": 7850
+    },
+    {
+      "epoch": 418.99,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.89817476272583,
+      "eval_runtime": 5.1276,
+      "eval_samples_per_second": 117.015,
+      "eval_steps_per_second": 3.705,
+      "step": 7856
+    },
+    {
+      "epoch": 419.2,
+      "learning_rate": 7.0370370370370375e-06,
+      "loss": 0.0217,
+      "step": 7860
+    },
+    {
+      "epoch": 419.73,
+      "learning_rate": 6.975308641975309e-06,
+      "loss": 0.0203,
+      "step": 7870
+    },
+    {
+      "epoch": 420.0,
+      "eval_accuracy": 0.24833333333333332,
+      "eval_loss": 4.900780200958252,
+      "eval_runtime": 5.1291,
+      "eval_samples_per_second": 116.98,
+      "eval_steps_per_second": 3.704,
+      "step": 7875
+    },
+    {
+      "epoch": 420.27,
+      "learning_rate": 6.91358024691358e-06,
+      "loss": 0.0295,
+      "step": 7880
+    },
+    {
+      "epoch": 420.8,
+      "learning_rate": 6.851851851851852e-06,
+      "loss": 0.0182,
+      "step": 7890
+    },
+    {
+      "epoch": 420.96,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.9113335609436035,
+      "eval_runtime": 5.1351,
+      "eval_samples_per_second": 116.842,
+      "eval_steps_per_second": 3.7,
+      "step": 7893
+    },
+    {
+      "epoch": 421.33,
+      "learning_rate": 6.790123456790123e-06,
+      "loss": 0.0321,
+      "step": 7900
+    },
+    {
+      "epoch": 421.87,
+      "learning_rate": 6.728395061728396e-06,
+      "loss": 0.0258,
+      "step": 7910
+    },
+    {
+      "epoch": 421.97,
+      "eval_accuracy": 0.25,
+      "eval_loss": 4.918017387390137,
+      "eval_runtime": 5.1337,
+      "eval_samples_per_second": 116.875,
+      "eval_steps_per_second": 3.701,
+      "step": 7912
+    },
+    {
+      "epoch": 422.4,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.0277,
+      "step": 7920
+    },
+    {
+      "epoch": 422.93,
+      "learning_rate": 6.604938271604938e-06,
+      "loss": 0.0266,
+      "step": 7930
+    },
+    {
+      "epoch": 422.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.913443565368652,
+      "eval_runtime": 5.1296,
+      "eval_samples_per_second": 116.969,
+      "eval_steps_per_second": 3.704,
+      "step": 7931
+    },
+    {
+      "epoch": 423.47,
+      "learning_rate": 6.54320987654321e-06,
+      "loss": 0.037,
+      "step": 7940
+    },
+    {
+      "epoch": 424.0,
+      "learning_rate": 6.481481481481481e-06,
+      "loss": 0.0304,
+      "step": 7950
+    },
+    {
+      "epoch": 424.0,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.900519847869873,
+      "eval_runtime": 5.1632,
+      "eval_samples_per_second": 116.207,
+      "eval_steps_per_second": 3.68,
+      "step": 7950
+    },
+    {
+      "epoch": 424.53,
+      "learning_rate": 6.419753086419754e-06,
+      "loss": 0.0247,
+      "step": 7960
+    },
+    {
+      "epoch": 424.96,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.893669605255127,
+      "eval_runtime": 5.2111,
+      "eval_samples_per_second": 115.139,
+      "eval_steps_per_second": 3.646,
+      "step": 7968
+    },
+    {
+      "epoch": 425.07,
+      "learning_rate": 6.3580246913580246e-06,
+      "loss": 0.0335,
+      "step": 7970
+    },
+    {
+      "epoch": 425.6,
+      "learning_rate": 6.296296296296296e-06,
+      "loss": 0.0493,
+      "step": 7980
+    },
+    {
+      "epoch": 425.97,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.883533954620361,
+      "eval_runtime": 5.1746,
+      "eval_samples_per_second": 115.951,
+      "eval_steps_per_second": 3.672,
+      "step": 7987
+    },
+    {
+      "epoch": 426.13,
+      "learning_rate": 6.234567901234569e-06,
+      "loss": 0.0157,
+      "step": 7990
+    },
+    {
+      "epoch": 426.67,
+      "learning_rate": 6.172839506172839e-06,
+      "loss": 0.0286,
+      "step": 8000
+    },
+    {
+      "epoch": 426.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.896754741668701,
+      "eval_runtime": 5.1903,
+      "eval_samples_per_second": 115.6,
+      "eval_steps_per_second": 3.661,
+      "step": 8006
+    },
+    {
+      "epoch": 427.2,
+      "learning_rate": 6.111111111111111e-06,
+      "loss": 0.0249,
+      "step": 8010
+    },
+    {
+      "epoch": 427.73,
+      "learning_rate": 6.049382716049383e-06,
+      "loss": 0.0228,
+      "step": 8020
+    },
+    {
+      "epoch": 428.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.906573295593262,
+      "eval_runtime": 5.2526,
+      "eval_samples_per_second": 114.23,
+      "eval_steps_per_second": 3.617,
+      "step": 8025
+    },
+    {
+      "epoch": 428.27,
+      "learning_rate": 5.9876543209876546e-06,
+      "loss": 0.0324,
+      "step": 8030
+    },
+    {
+      "epoch": 428.8,
+      "learning_rate": 5.925925925925927e-06,
+      "loss": 0.0362,
+      "step": 8040
+    },
+    {
+      "epoch": 428.96,
+      "eval_accuracy": 0.245,
+      "eval_loss": 4.903099536895752,
+      "eval_runtime": 5.2207,
+      "eval_samples_per_second": 114.927,
+      "eval_steps_per_second": 3.639,
+      "step": 8043
+    },
+    {
+      "epoch": 429.33,
+      "learning_rate": 5.864197530864198e-06,
+      "loss": 0.0213,
+      "step": 8050
+    },
+    {
+      "epoch": 429.87,
+      "learning_rate": 5.802469135802469e-06,
+      "loss": 0.0244,
+      "step": 8060
+    },
+    {
+      "epoch": 429.97,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.899692535400391,
+      "eval_runtime": 5.2551,
+      "eval_samples_per_second": 114.174,
+      "eval_steps_per_second": 3.616,
+      "step": 8062
+    },
+    {
+      "epoch": 430.4,
+      "learning_rate": 5.740740740740741e-06,
+      "loss": 0.0405,
+      "step": 8070
+    },
+    {
+      "epoch": 430.93,
+      "learning_rate": 5.6790123456790125e-06,
+      "loss": 0.0204,
+      "step": 8080
+    },
+    {
+      "epoch": 430.99,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.905935287475586,
+      "eval_runtime": 5.2136,
+      "eval_samples_per_second": 115.084,
+      "eval_steps_per_second": 3.644,
+      "step": 8081
+    },
+    {
+      "epoch": 431.47,
+      "learning_rate": 5.617283950617285e-06,
+      "loss": 0.0175,
+      "step": 8090
+    },
+    {
+      "epoch": 432.0,
+      "learning_rate": 5.555555555555556e-06,
+      "loss": 0.0344,
+      "step": 8100
+    },
+    {
+      "epoch": 432.0,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.905205726623535,
+      "eval_runtime": 5.2045,
+      "eval_samples_per_second": 115.286,
+      "eval_steps_per_second": 3.651,
+      "step": 8100
+    },
+    {
+      "epoch": 432.53,
+      "learning_rate": 5.493827160493827e-06,
+      "loss": 0.0252,
+      "step": 8110
+    },
+    {
+      "epoch": 432.96,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.897459983825684,
+      "eval_runtime": 5.1933,
+      "eval_samples_per_second": 115.533,
+      "eval_steps_per_second": 3.659,
+      "step": 8118
+    },
+    {
+      "epoch": 433.07,
+      "learning_rate": 5.432098765432099e-06,
+      "loss": 0.0185,
+      "step": 8120
+    },
+    {
+      "epoch": 433.6,
+      "learning_rate": 5.37037037037037e-06,
+      "loss": 0.0242,
+      "step": 8130
+    },
+    {
+      "epoch": 433.97,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.896070957183838,
+      "eval_runtime": 5.2428,
+      "eval_samples_per_second": 114.443,
+      "eval_steps_per_second": 3.624,
+      "step": 8137
+    },
+    {
+      "epoch": 434.13,
+      "learning_rate": 5.3086419753086425e-06,
+      "loss": 0.0372,
+      "step": 8140
+    },
+    {
+      "epoch": 434.67,
+      "learning_rate": 5.246913580246914e-06,
+      "loss": 0.0135,
+      "step": 8150
+    },
+    {
+      "epoch": 434.99,
+      "eval_accuracy": 0.24666666666666667,
+      "eval_loss": 4.908581733703613,
+      "eval_runtime": 5.248,
+      "eval_samples_per_second": 114.329,
+      "eval_steps_per_second": 3.62,
+      "step": 8156
+    },
+    {
+      "epoch": 435.2,
+      "learning_rate": 5.185185185185185e-06,
+      "loss": 0.0322,
+      "step": 8160
+    },
+    {
+      "epoch": 435.73,
+      "learning_rate": 5.123456790123457e-06,
+      "loss": 0.0296,
+      "step": 8170
+    },
+    {
+      "epoch": 436.0,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.9134602546691895,
+      "eval_runtime": 5.2357,
+      "eval_samples_per_second": 114.598,
+      "eval_steps_per_second": 3.629,
+      "step": 8175
+    },
+    {
+      "epoch": 436.27,
+      "learning_rate": 5.061728395061729e-06,
+      "loss": 0.0164,
+      "step": 8180
+    },
+    {
+      "epoch": 436.8,
+      "learning_rate": 5e-06,
+      "loss": 0.0432,
+      "step": 8190
+    },
+    {
+      "epoch": 436.96,
+      "eval_accuracy": 0.24333333333333335,
+      "eval_loss": 4.907933712005615,
+      "eval_runtime": 5.268,
+      "eval_samples_per_second": 113.896,
+      "eval_steps_per_second": 3.607,
+      "step": 8193
+    },
+    {
+      "epoch": 437.33,
+      "learning_rate": 4.938271604938272e-06,
+      "loss": 0.017,
+      "step": 8200
+    },
+    {
+      "epoch": 437.87,
+      "learning_rate": 4.876543209876543e-06,
+      "loss": 0.0242,
+      "step": 8210
+    },
+    {
+      "epoch": 437.97,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.898138046264648,
+      "eval_runtime": 5.2464,
+      "eval_samples_per_second": 114.365,
+      "eval_steps_per_second": 3.622,
+      "step": 8212
+    },
+    {
+      "epoch": 438.4,
+      "learning_rate": 4.814814814814815e-06,
+      "loss": 0.026,
+      "step": 8220
+    },
+    {
+      "epoch": 438.93,
+      "learning_rate": 4.753086419753087e-06,
+      "loss": 0.0227,
+      "step": 8230
+    },
+    {
+      "epoch": 438.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.8857102394104,
+      "eval_runtime": 5.264,
+      "eval_samples_per_second": 113.981,
+      "eval_steps_per_second": 3.609,
+      "step": 8231
+    },
+    {
+      "epoch": 439.47,
+      "learning_rate": 4.691358024691358e-06,
+      "loss": 0.0309,
+      "step": 8240
+    },
+    {
+      "epoch": 440.0,
+      "learning_rate": 4.6296296296296296e-06,
+      "loss": 0.021,
+      "step": 8250
+    },
+    {
+      "epoch": 440.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.887371063232422,
+      "eval_runtime": 5.2856,
+      "eval_samples_per_second": 113.516,
+      "eval_steps_per_second": 3.595,
+      "step": 8250
+    },
+    {
+      "epoch": 440.53,
+      "learning_rate": 4.567901234567902e-06,
+      "loss": 0.0244,
+      "step": 8260
+    },
+    {
+      "epoch": 440.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.884664058685303,
+      "eval_runtime": 5.2546,
+      "eval_samples_per_second": 114.187,
+      "eval_steps_per_second": 3.616,
+      "step": 8268
+    },
+    {
+      "epoch": 441.07,
+      "learning_rate": 4.506172839506173e-06,
+      "loss": 0.0484,
+      "step": 8270
+    },
+    {
+      "epoch": 441.6,
+      "learning_rate": 4.444444444444445e-06,
+      "loss": 0.0234,
+      "step": 8280
+    },
+    {
+      "epoch": 441.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.896438121795654,
+      "eval_runtime": 5.2769,
+      "eval_samples_per_second": 113.703,
+      "eval_steps_per_second": 3.601,
+      "step": 8287
+    },
+    {
+      "epoch": 442.13,
+      "learning_rate": 4.382716049382716e-06,
+      "loss": 0.0293,
+      "step": 8290
+    },
+    {
+      "epoch": 442.67,
+      "learning_rate": 4.3209876543209875e-06,
+      "loss": 0.0278,
+      "step": 8300
+    },
+    {
+      "epoch": 442.99,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.916093349456787,
+      "eval_runtime": 5.2536,
+      "eval_samples_per_second": 114.207,
+      "eval_steps_per_second": 3.617,
+      "step": 8306
+    },
+    {
+      "epoch": 443.2,
+      "learning_rate": 4.2592592592592596e-06,
+      "loss": 0.0224,
+      "step": 8310
+    },
+    {
+      "epoch": 443.73,
+      "learning_rate": 4.197530864197531e-06,
+      "loss": 0.0322,
+      "step": 8320
+    },
+    {
+      "epoch": 444.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.921158790588379,
+      "eval_runtime": 5.2757,
+      "eval_samples_per_second": 113.729,
+      "eval_steps_per_second": 3.601,
+      "step": 8325
+    },
+    {
+      "epoch": 444.27,
+      "learning_rate": 4.135802469135803e-06,
+      "loss": 0.0201,
+      "step": 8330
+    },
+    {
+      "epoch": 444.8,
+      "learning_rate": 4.074074074074075e-06,
+      "loss": 0.038,
+      "step": 8340
+    },
+    {
+      "epoch": 444.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.925096035003662,
+      "eval_runtime": 5.3078,
+      "eval_samples_per_second": 113.041,
+      "eval_steps_per_second": 3.58,
+      "step": 8343
+    },
+    {
+      "epoch": 445.33,
+      "learning_rate": 4.012345679012345e-06,
+      "loss": 0.035,
+      "step": 8350
+    },
+    {
+      "epoch": 445.87,
+      "learning_rate": 3.9506172839506175e-06,
+      "loss": 0.0327,
+      "step": 8360
+    },
+    {
+      "epoch": 445.97,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.933958530426025,
+      "eval_runtime": 5.2639,
+      "eval_samples_per_second": 113.985,
+      "eval_steps_per_second": 3.61,
+      "step": 8362
+    },
+    {
+      "epoch": 446.4,
+      "learning_rate": 3.888888888888889e-06,
+      "loss": 0.0226,
+      "step": 8370
+    },
+    {
+      "epoch": 446.93,
+      "learning_rate": 3.827160493827161e-06,
+      "loss": 0.0256,
+      "step": 8380
+    },
+    {
+      "epoch": 446.99,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.924610614776611,
+      "eval_runtime": 5.2666,
+      "eval_samples_per_second": 113.926,
+      "eval_steps_per_second": 3.608,
+      "step": 8381
+    },
+    {
+      "epoch": 447.47,
+      "learning_rate": 3.7654320987654325e-06,
+      "loss": 0.027,
+      "step": 8390
+    },
+    {
+      "epoch": 448.0,
+      "learning_rate": 3.7037037037037037e-06,
+      "loss": 0.0327,
+      "step": 8400
+    },
+    {
+      "epoch": 448.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.929351329803467,
+      "eval_runtime": 5.2953,
+      "eval_samples_per_second": 113.308,
+      "eval_steps_per_second": 3.588,
+      "step": 8400
+    },
+    {
+      "epoch": 448.53,
+      "learning_rate": 3.6419753086419754e-06,
+      "loss": 0.0246,
+      "step": 8410
+    },
+    {
+      "epoch": 448.96,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.931128025054932,
+      "eval_runtime": 5.3144,
+      "eval_samples_per_second": 112.901,
+      "eval_steps_per_second": 3.575,
+      "step": 8418
+    },
+    {
+      "epoch": 449.07,
+      "learning_rate": 3.580246913580247e-06,
+      "loss": 0.027,
+      "step": 8420
+    },
+    {
+      "epoch": 449.6,
+      "learning_rate": 3.5185185185185187e-06,
+      "loss": 0.0239,
+      "step": 8430
+    },
+    {
+      "epoch": 449.97,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.92203950881958,
+      "eval_runtime": 5.2899,
+      "eval_samples_per_second": 113.424,
+      "eval_steps_per_second": 3.592,
+      "step": 8437
+    },
+    {
+      "epoch": 450.13,
+      "learning_rate": 3.45679012345679e-06,
+      "loss": 0.0313,
+      "step": 8440
+    },
+    {
+      "epoch": 450.67,
+      "learning_rate": 3.3950617283950617e-06,
+      "loss": 0.0219,
+      "step": 8450
+    },
+    {
+      "epoch": 450.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.920533180236816,
+      "eval_runtime": 5.4385,
+      "eval_samples_per_second": 110.325,
+      "eval_steps_per_second": 3.494,
+      "step": 8456
+    },
+    {
+      "epoch": 451.2,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.0214,
+      "step": 8460
+    },
+    {
+      "epoch": 451.73,
+      "learning_rate": 3.271604938271605e-06,
+      "loss": 0.0287,
+      "step": 8470
+    },
+    {
+      "epoch": 452.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.924895763397217,
+      "eval_runtime": 5.3449,
+      "eval_samples_per_second": 112.256,
+      "eval_steps_per_second": 3.555,
+      "step": 8475
+    },
+    {
+      "epoch": 452.27,
+      "learning_rate": 3.209876543209877e-06,
+      "loss": 0.0226,
+      "step": 8480
+    },
+    {
+      "epoch": 452.8,
+      "learning_rate": 3.148148148148148e-06,
+      "loss": 0.0244,
+      "step": 8490
+    },
+    {
+      "epoch": 452.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.927518844604492,
+      "eval_runtime": 5.3234,
+      "eval_samples_per_second": 112.711,
+      "eval_steps_per_second": 3.569,
+      "step": 8493
+    },
+    {
+      "epoch": 453.33,
+      "learning_rate": 3.0864197530864196e-06,
+      "loss": 0.0215,
+      "step": 8500
+    },
+    {
+      "epoch": 453.87,
+      "learning_rate": 3.0246913580246917e-06,
+      "loss": 0.0222,
+      "step": 8510
+    },
+    {
+      "epoch": 453.97,
+      "eval_accuracy": 0.24166666666666667,
+      "eval_loss": 4.932238578796387,
+      "eval_runtime": 5.3268,
+      "eval_samples_per_second": 112.637,
+      "eval_steps_per_second": 3.567,
+      "step": 8512
+    },
+    {
+      "epoch": 454.4,
+      "learning_rate": 2.9629629629629633e-06,
+      "loss": 0.0269,
+      "step": 8520
+    },
+    {
+      "epoch": 454.93,
+      "learning_rate": 2.9012345679012346e-06,
+      "loss": 0.0277,
+      "step": 8530
+    },
+    {
+      "epoch": 454.99,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.931756019592285,
+      "eval_runtime": 5.3343,
+      "eval_samples_per_second": 112.48,
+      "eval_steps_per_second": 3.562,
+      "step": 8531
+    },
+    {
+      "epoch": 455.47,
+      "learning_rate": 2.8395061728395062e-06,
+      "loss": 0.0355,
+      "step": 8540
+    },
+    {
+      "epoch": 456.0,
+      "learning_rate": 2.777777777777778e-06,
+      "loss": 0.0315,
+      "step": 8550
+    },
+    {
+      "epoch": 456.0,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.929112911224365,
+      "eval_runtime": 5.4018,
+      "eval_samples_per_second": 111.075,
+      "eval_steps_per_second": 3.517,
+      "step": 8550
+    },
+    {
+      "epoch": 456.53,
+      "learning_rate": 2.7160493827160496e-06,
+      "loss": 0.021,
+      "step": 8560
+    },
+    {
+      "epoch": 456.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.929343223571777,
+      "eval_runtime": 5.3767,
+      "eval_samples_per_second": 111.594,
+      "eval_steps_per_second": 3.534,
+      "step": 8568
+    },
+    {
+      "epoch": 457.07,
+      "learning_rate": 2.6543209876543212e-06,
+      "loss": 0.0232,
+      "step": 8570
+    },
+    {
+      "epoch": 457.6,
+      "learning_rate": 2.5925925925925925e-06,
+      "loss": 0.0288,
+      "step": 8580
+    },
+    {
+      "epoch": 457.97,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.923261642456055,
+      "eval_runtime": 5.3734,
+      "eval_samples_per_second": 111.661,
+      "eval_steps_per_second": 3.536,
+      "step": 8587
+    },
+    {
+      "epoch": 458.13,
+      "learning_rate": 2.5308641975308646e-06,
+      "loss": 0.0381,
+      "step": 8590
+    },
+    {
+      "epoch": 458.67,
+      "learning_rate": 2.469135802469136e-06,
+      "loss": 0.0229,
+      "step": 8600
+    },
+    {
+      "epoch": 458.99,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.923562049865723,
+      "eval_runtime": 5.3963,
+      "eval_samples_per_second": 111.186,
+      "eval_steps_per_second": 3.521,
+      "step": 8606
+    },
+    {
+      "epoch": 459.2,
+      "learning_rate": 2.4074074074074075e-06,
+      "loss": 0.0308,
+      "step": 8610
+    },
+    {
+      "epoch": 459.73,
+      "learning_rate": 2.345679012345679e-06,
+      "loss": 0.0257,
+      "step": 8620
+    },
+    {
+      "epoch": 460.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.922508716583252,
+      "eval_runtime": 5.4351,
+      "eval_samples_per_second": 110.394,
+      "eval_steps_per_second": 3.496,
+      "step": 8625
+    },
+    {
+      "epoch": 460.27,
+      "learning_rate": 2.283950617283951e-06,
+      "loss": 0.0193,
+      "step": 8630
+    },
+    {
+      "epoch": 460.8,
+      "learning_rate": 2.2222222222222225e-06,
+      "loss": 0.0291,
+      "step": 8640
+    },
+    {
+      "epoch": 460.96,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.922194004058838,
+      "eval_runtime": 5.3746,
+      "eval_samples_per_second": 111.636,
+      "eval_steps_per_second": 3.535,
+      "step": 8643
+    },
+    {
+      "epoch": 461.33,
+      "learning_rate": 2.1604938271604937e-06,
+      "loss": 0.0292,
+      "step": 8650
+    },
+    {
+      "epoch": 461.87,
+      "learning_rate": 2.0987654320987654e-06,
+      "loss": 0.0325,
+      "step": 8660
+    },
+    {
+      "epoch": 461.97,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.921637535095215,
+      "eval_runtime": 5.38,
+      "eval_samples_per_second": 111.524,
+      "eval_steps_per_second": 3.532,
+      "step": 8662
+    },
+    {
+      "epoch": 462.4,
+      "learning_rate": 2.0370370370370375e-06,
+      "loss": 0.0292,
+      "step": 8670
+    },
+    {
+      "epoch": 462.93,
+      "learning_rate": 1.9753086419753087e-06,
+      "loss": 0.0268,
+      "step": 8680
+    },
+    {
+      "epoch": 462.99,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.92024564743042,
+      "eval_runtime": 5.3601,
+      "eval_samples_per_second": 111.939,
+      "eval_steps_per_second": 3.545,
+      "step": 8681
+    },
+    {
+      "epoch": 463.47,
+      "learning_rate": 1.9135802469135804e-06,
+      "loss": 0.0268,
+      "step": 8690
+    },
+    {
+      "epoch": 464.0,
+      "learning_rate": 1.8518518518518519e-06,
+      "loss": 0.0156,
+      "step": 8700
+    },
+    {
+      "epoch": 464.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.91749906539917,
+      "eval_runtime": 5.3813,
+      "eval_samples_per_second": 111.497,
+      "eval_steps_per_second": 3.531,
+      "step": 8700
+    },
+    {
+      "epoch": 464.53,
+      "learning_rate": 1.7901234567901235e-06,
+      "loss": 0.0196,
+      "step": 8710
+    },
+    {
+      "epoch": 464.96,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.914690017700195,
+      "eval_runtime": 5.3651,
+      "eval_samples_per_second": 111.834,
+      "eval_steps_per_second": 3.541,
+      "step": 8718
+    },
+    {
+      "epoch": 465.07,
+      "learning_rate": 1.728395061728395e-06,
+      "loss": 0.0199,
+      "step": 8720
+    },
+    {
+      "epoch": 465.6,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 0.0448,
+      "step": 8730
+    },
+    {
+      "epoch": 465.97,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.910024642944336,
+      "eval_runtime": 5.359,
+      "eval_samples_per_second": 111.962,
+      "eval_steps_per_second": 3.545,
+      "step": 8737
+    },
+    {
+      "epoch": 466.13,
+      "learning_rate": 1.6049382716049385e-06,
+      "loss": 0.021,
+      "step": 8740
+    },
+    {
+      "epoch": 466.67,
+      "learning_rate": 1.5432098765432098e-06,
+      "loss": 0.0232,
+      "step": 8750
+    },
+    {
+      "epoch": 466.99,
+      "eval_accuracy": 0.23333333333333334,
+      "eval_loss": 4.908828258514404,
+      "eval_runtime": 5.3987,
+      "eval_samples_per_second": 111.138,
+      "eval_steps_per_second": 3.519,
+      "step": 8756
+    },
+    {
+      "epoch": 467.2,
+      "learning_rate": 1.4814814814814817e-06,
+      "loss": 0.0224,
+      "step": 8760
+    },
+    {
+      "epoch": 467.73,
+      "learning_rate": 1.4197530864197531e-06,
+      "loss": 0.0274,
+      "step": 8770
+    },
+    {
+      "epoch": 468.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.909602165222168,
+      "eval_runtime": 5.4071,
+      "eval_samples_per_second": 110.966,
+      "eval_steps_per_second": 3.514,
+      "step": 8775
+    },
+    {
+      "epoch": 468.27,
+      "learning_rate": 1.3580246913580248e-06,
+      "loss": 0.0344,
+      "step": 8780
+    },
+    {
+      "epoch": 468.8,
+      "learning_rate": 1.2962962962962962e-06,
+      "loss": 0.029,
+      "step": 8790
+    },
+    {
+      "epoch": 468.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.910549640655518,
+      "eval_runtime": 5.3868,
+      "eval_samples_per_second": 111.383,
+      "eval_steps_per_second": 3.527,
+      "step": 8793
+    },
+    {
+      "epoch": 469.33,
+      "learning_rate": 1.234567901234568e-06,
+      "loss": 0.0265,
+      "step": 8800
+    },
+    {
+      "epoch": 469.87,
+      "learning_rate": 1.1728395061728396e-06,
+      "loss": 0.0337,
+      "step": 8810
+    },
+    {
+      "epoch": 469.97,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.912467956542969,
+      "eval_runtime": 5.3985,
+      "eval_samples_per_second": 111.143,
+      "eval_steps_per_second": 3.52,
+      "step": 8812
+    },
+    {
+      "epoch": 470.4,
+      "learning_rate": 1.1111111111111112e-06,
+      "loss": 0.0298,
+      "step": 8820
+    },
+    {
+      "epoch": 470.93,
+      "learning_rate": 1.0493827160493827e-06,
+      "loss": 0.0178,
+      "step": 8830
+    },
+    {
+      "epoch": 470.99,
+      "eval_accuracy": 0.235,
+      "eval_loss": 4.91201639175415,
+      "eval_runtime": 5.399,
+      "eval_samples_per_second": 111.132,
+      "eval_steps_per_second": 3.519,
+      "step": 8831
+    },
+    {
+      "epoch": 471.47,
+      "learning_rate": 9.876543209876544e-07,
+      "loss": 0.0179,
+      "step": 8840
+    },
+    {
+      "epoch": 472.0,
+      "learning_rate": 9.259259259259259e-07,
+      "loss": 0.0286,
+      "step": 8850
+    },
+    {
+      "epoch": 472.0,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.912468910217285,
+      "eval_runtime": 5.4913,
+      "eval_samples_per_second": 109.263,
+      "eval_steps_per_second": 3.46,
+      "step": 8850
+    },
+    {
+      "epoch": 472.53,
+      "learning_rate": 8.641975308641975e-07,
+      "loss": 0.0159,
+      "step": 8860
+    },
+    {
+      "epoch": 472.96,
+      "eval_accuracy": 0.23666666666666666,
+      "eval_loss": 4.910217761993408,
+      "eval_runtime": 5.4334,
+      "eval_samples_per_second": 110.428,
+      "eval_steps_per_second": 3.497,
+      "step": 8868
+    },
+    {
+      "epoch": 473.07,
+      "learning_rate": 8.024691358024693e-07,
+      "loss": 0.0247,
+      "step": 8870
+    },
+    {
+      "epoch": 473.6,
+      "learning_rate": 7.407407407407408e-07,
+      "loss": 0.0318,
+      "step": 8880
+    },
+    {
+      "epoch": 473.97,
+      "eval_accuracy": 0.23833333333333334,
+      "eval_loss": 4.9116291999816895,
+      "eval_runtime": 5.4254,
+      "eval_samples_per_second": 110.59,
+      "eval_steps_per_second": 3.502,
+      "step": 8887
+    },
+    {
+      "epoch": 474.13,
+      "learning_rate": 6.790123456790124e-07,
+      "loss": 0.0311,
+      "step": 8890
+    },
+    {
+      "epoch": 474.67,
+      "learning_rate": 6.17283950617284e-07,
+      "loss": 0.0302,
+      "step": 8900
+    },
+    {
+      "epoch": 474.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.911314487457275,
+      "eval_runtime": 5.4263,
+      "eval_samples_per_second": 110.573,
+      "eval_steps_per_second": 3.501,
+      "step": 8906
+    },
+    {
+      "epoch": 475.2,
+      "learning_rate": 5.555555555555556e-07,
+      "loss": 0.0305,
+      "step": 8910
+    },
+    {
+      "epoch": 475.73,
+      "learning_rate": 4.938271604938272e-07,
+      "loss": 0.0184,
+      "step": 8920
+    },
+    {
+      "epoch": 476.0,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.911987781524658,
+      "eval_runtime": 5.4618,
+      "eval_samples_per_second": 109.854,
+      "eval_steps_per_second": 3.479,
+      "step": 8925
+    },
+    {
+      "epoch": 476.27,
+      "learning_rate": 4.3209876543209875e-07,
+      "loss": 0.0415,
+      "step": 8930
+    },
+    {
+      "epoch": 476.8,
+      "learning_rate": 3.703703703703704e-07,
+      "loss": 0.025,
+      "step": 8940
+    },
+    {
+      "epoch": 476.96,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.912769794464111,
+      "eval_runtime": 5.4549,
+      "eval_samples_per_second": 109.993,
+      "eval_steps_per_second": 3.483,
+      "step": 8943
+    },
+    {
+      "epoch": 477.33,
+      "learning_rate": 3.08641975308642e-07,
+      "loss": 0.013,
+      "step": 8950
+    },
+    {
+      "epoch": 477.87,
+      "learning_rate": 2.469135802469136e-07,
+      "loss": 0.027,
+      "step": 8960
+    },
+    {
+      "epoch": 477.97,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.912613868713379,
+      "eval_runtime": 5.4968,
+      "eval_samples_per_second": 109.154,
+      "eval_steps_per_second": 3.457,
+      "step": 8962
+    },
+    {
+      "epoch": 478.4,
+      "learning_rate": 1.851851851851852e-07,
+      "loss": 0.0156,
+      "step": 8970
+    },
+    {
+      "epoch": 478.93,
+      "learning_rate": 1.234567901234568e-07,
+      "loss": 0.0298,
+      "step": 8980
+    },
+    {
+      "epoch": 478.99,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.913006782531738,
+      "eval_runtime": 5.4672,
+      "eval_samples_per_second": 109.745,
+      "eval_steps_per_second": 3.475,
+      "step": 8981
+    },
+    {
+      "epoch": 479.47,
+      "learning_rate": 6.17283950617284e-08,
+      "loss": 0.0213,
+      "step": 8990
+    },
+    {
+      "epoch": 480.0,
+      "learning_rate": 0.0,
+      "loss": 0.0349,
+      "step": 9000
+    },
+    {
+      "epoch": 480.0,
+      "eval_accuracy": 0.24,
+      "eval_loss": 4.9129838943481445,
+      "eval_runtime": 5.4668,
+      "eval_samples_per_second": 109.753,
+      "eval_steps_per_second": 3.476,
+      "step": 9000
+    },
+    {
+      "epoch": 480.0,
+      "step": 9000,
+      "total_flos": 6.6744785965028475e+19,
+      "train_loss": 0.3827814753833744,
+      "train_runtime": 27227.0796,
+      "train_samples_per_second": 44.019,
+      "train_steps_per_second": 0.331
+    }
+  ],
+  "max_steps": 9000,
+  "num_train_epochs": 500,
+  "total_flos": 6.6744785965028475e+19,
+  "trial_name": null,
+  "trial_params": null
+}