End of training

all_results.json

@@ -0,0 +1,15 @@
+{
+    "epoch": 31.18,
+    "eval_accuracy": 0.1376987935196624,
+    "eval_loss": 6.486697196960449,
+    "eval_runtime": 169.2274,
+    "eval_samples": 24055,
+    "eval_samples_per_second": 142.146,
+    "eval_steps_per_second": 2.222,
+    "perplexity": 656.351978052217,
+    "train_loss": 6.89371246590102,
+    "train_runtime": 29548.4223,
+    "train_samples": 24910,
+    "train_samples_per_second": 33.721,
+    "train_steps_per_second": 0.263
+}

eval_results.json

@@ -0,0 +1,10 @@
+{
+    "epoch": 31.18,
+    "eval_accuracy": 0.1376987935196624,
+    "eval_loss": 6.486697196960449,
+    "eval_runtime": 169.2274,
+    "eval_samples": 24055,
+    "eval_samples_per_second": 142.146,
+    "eval_steps_per_second": 2.222,
+    "perplexity": 656.351978052217
+}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 31.18,
+    "train_loss": 6.89371246590102,
+    "train_runtime": 29548.4223,
+    "train_samples": 24910,
+    "train_samples_per_second": 33.721,
+    "train_steps_per_second": 0.263
+}

trainer_state.json

@@ -0,0 +1,4744 @@
+{
+  "best_metric": 6.480716228485107,
+  "best_model_checkpoint": "/data1/attanasiog/babylm/roberta-tiny-10M/checkpoint-5900",
+  "epoch": 31.184852374839537,
+  "global_step": 6050,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 10.8666,
+      "step": 10
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 10.8334,
+      "step": 20
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.2e-05,
+      "loss": 10.7554,
+      "step": 30
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 10.6494,
+      "step": 40
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2e-05,
+      "loss": 10.5287,
+      "step": 50
+    },
+    {
+      "epoch": 0.26,
+      "eval_accuracy": 0.048115526656560687,
+      "eval_loss": 10.44776725769043,
+      "eval_runtime": 168.8841,
+      "eval_samples_per_second": 142.435,
+      "eval_steps_per_second": 2.226,
+      "step": 50
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2e-05,
+      "loss": 10.4004,
+      "step": 60
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2e-05,
+      "loss": 10.2909,
+      "step": 70
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2e-05,
+      "loss": 10.1876,
+      "step": 80
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2e-05,
+      "loss": 10.092,
+      "step": 90
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2e-05,
+      "loss": 10.0073,
+      "step": 100
+    },
+    {
+      "epoch": 0.51,
+      "eval_accuracy": 0.04883696521337363,
+      "eval_loss": 9.954952239990234,
+      "eval_runtime": 168.8432,
+      "eval_samples_per_second": 142.469,
+      "eval_steps_per_second": 2.227,
+      "step": 100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2e-05,
+      "loss": 9.9271,
+      "step": 110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2e-05,
+      "loss": 9.8418,
+      "step": 120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2e-05,
+      "loss": 9.7731,
+      "step": 130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2e-05,
+      "loss": 9.6976,
+      "step": 140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2e-05,
+      "loss": 9.6268,
+      "step": 150
+    },
+    {
+      "epoch": 0.77,
+      "eval_accuracy": 0.04877340177584069,
+      "eval_loss": 9.586450576782227,
+      "eval_runtime": 168.9917,
+      "eval_samples_per_second": 142.344,
+      "eval_steps_per_second": 2.225,
+      "step": 150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2e-05,
+      "loss": 9.5519,
+      "step": 160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2e-05,
+      "loss": 9.4822,
+      "step": 170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2e-05,
+      "loss": 9.4239,
+      "step": 180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2e-05,
+      "loss": 9.3531,
+      "step": 190
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2e-05,
+      "loss": 9.9837,
+      "step": 200
+    },
+    {
+      "epoch": 1.03,
+      "eval_accuracy": 0.04709288795319612,
+      "eval_loss": 9.250182151794434,
+      "eval_runtime": 169.1628,
+      "eval_samples_per_second": 142.2,
+      "eval_steps_per_second": 2.223,
+      "step": 200
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2e-05,
+      "loss": 9.2265,
+      "step": 210
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2e-05,
+      "loss": 9.1555,
+      "step": 220
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2e-05,
+      "loss": 9.0872,
+      "step": 230
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2e-05,
+      "loss": 9.0276,
+      "step": 240
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2e-05,
+      "loss": 8.9701,
+      "step": 250
+    },
+    {
+      "epoch": 1.29,
+      "eval_accuracy": 0.04664432564609462,
+      "eval_loss": 8.936966896057129,
+      "eval_runtime": 168.9806,
+      "eval_samples_per_second": 142.354,
+      "eval_steps_per_second": 2.225,
+      "step": 250
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2e-05,
+      "loss": 8.913,
+      "step": 260
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2e-05,
+      "loss": 8.8506,
+      "step": 270
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2e-05,
+      "loss": 8.7763,
+      "step": 280
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2e-05,
+      "loss": 8.7262,
+      "step": 290
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2e-05,
+      "loss": 8.6689,
+      "step": 300
+    },
+    {
+      "epoch": 1.54,
+      "eval_accuracy": 0.04727021760311789,
+      "eval_loss": 8.644742965698242,
+      "eval_runtime": 168.948,
+      "eval_samples_per_second": 142.381,
+      "eval_steps_per_second": 2.226,
+      "step": 300
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2e-05,
+      "loss": 8.6119,
+      "step": 310
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2e-05,
+      "loss": 8.5511,
+      "step": 320
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2e-05,
+      "loss": 8.503,
+      "step": 330
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2e-05,
+      "loss": 8.4466,
+      "step": 340
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2e-05,
+      "loss": 8.3893,
+      "step": 350
+    },
+    {
+      "epoch": 1.8,
+      "eval_accuracy": 0.04733514138538159,
+      "eval_loss": 8.379368782043457,
+      "eval_runtime": 169.22,
+      "eval_samples_per_second": 142.152,
+      "eval_steps_per_second": 2.222,
+      "step": 350
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2e-05,
+      "loss": 8.3375,
+      "step": 360
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2e-05,
+      "loss": 8.2824,
+      "step": 370
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2e-05,
+      "loss": 8.2317,
+      "step": 380
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2e-05,
+      "loss": 8.8178,
+      "step": 390
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2e-05,
+      "loss": 8.1697,
+      "step": 400
+    },
+    {
+      "epoch": 2.06,
+      "eval_accuracy": 0.050607646781225514,
+      "eval_loss": 8.134186744689941,
+      "eval_runtime": 169.087,
+      "eval_samples_per_second": 142.264,
+      "eval_steps_per_second": 2.224,
+      "step": 400
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2e-05,
+      "loss": 8.0758,
+      "step": 410
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2e-05,
+      "loss": 8.048,
+      "step": 420
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2e-05,
+      "loss": 7.9864,
+      "step": 430
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2e-05,
+      "loss": 7.9734,
+      "step": 440
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2e-05,
+      "loss": 7.926,
+      "step": 450
+    },
+    {
+      "epoch": 2.32,
+      "eval_accuracy": 0.06168861888884143,
+      "eval_loss": 7.922093868255615,
+      "eval_runtime": 168.9407,
+      "eval_samples_per_second": 142.387,
+      "eval_steps_per_second": 2.226,
+      "step": 450
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2e-05,
+      "loss": 7.8712,
+      "step": 460
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2e-05,
+      "loss": 7.8416,
+      "step": 470
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2e-05,
+      "loss": 7.8118,
+      "step": 480
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2e-05,
+      "loss": 7.7706,
+      "step": 490
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2e-05,
+      "loss": 7.7329,
+      "step": 500
+    },
+    {
+      "epoch": 2.58,
+      "eval_accuracy": 0.06273108274751771,
+      "eval_loss": 7.739767074584961,
+      "eval_runtime": 169.0756,
+      "eval_samples_per_second": 142.274,
+      "eval_steps_per_second": 2.224,
+      "step": 500
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2e-05,
+      "loss": 7.7048,
+      "step": 510
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2e-05,
+      "loss": 7.6478,
+      "step": 520
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2e-05,
+      "loss": 7.6647,
+      "step": 530
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2e-05,
+      "loss": 7.6017,
+      "step": 540
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2e-05,
+      "loss": 7.582,
+      "step": 550
+    },
+    {
+      "epoch": 2.83,
+      "eval_accuracy": 0.06907213036340595,
+      "eval_loss": 7.5844244956970215,
+      "eval_runtime": 169.0372,
+      "eval_samples_per_second": 142.306,
+      "eval_steps_per_second": 2.224,
+      "step": 550
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2e-05,
+      "loss": 7.5512,
+      "step": 560
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2e-05,
+      "loss": 7.505,
+      "step": 570
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2e-05,
+      "loss": 7.523,
+      "step": 580
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2e-05,
+      "loss": 8.0083,
+      "step": 590
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2e-05,
+      "loss": 7.4419,
+      "step": 600
+    },
+    {
+      "epoch": 3.09,
+      "eval_accuracy": 0.07293786059545417,
+      "eval_loss": 7.46204137802124,
+      "eval_runtime": 169.0193,
+      "eval_samples_per_second": 142.321,
+      "eval_steps_per_second": 2.225,
+      "step": 600
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2e-05,
+      "loss": 7.3961,
+      "step": 610
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2e-05,
+      "loss": 7.3728,
+      "step": 620
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2e-05,
+      "loss": 7.3709,
+      "step": 630
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2e-05,
+      "loss": 7.3295,
+      "step": 640
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2e-05,
+      "loss": 7.3658,
+      "step": 650
+    },
+    {
+      "epoch": 3.35,
+      "eval_accuracy": 0.07808497071718955,
+      "eval_loss": 7.373513698577881,
+      "eval_runtime": 169.1773,
+      "eval_samples_per_second": 142.188,
+      "eval_steps_per_second": 2.223,
+      "step": 650
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2e-05,
+      "loss": 7.342,
+      "step": 660
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2e-05,
+      "loss": 7.3369,
+      "step": 670
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2e-05,
+      "loss": 7.292,
+      "step": 680
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2e-05,
+      "loss": 7.2974,
+      "step": 690
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 2e-05,
+      "loss": 7.2857,
+      "step": 700
+    },
+    {
+      "epoch": 3.61,
+      "eval_accuracy": 0.08012986753100795,
+      "eval_loss": 7.304928302764893,
+      "eval_runtime": 169.0533,
+      "eval_samples_per_second": 142.292,
+      "eval_steps_per_second": 2.224,
+      "step": 700
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2e-05,
+      "loss": 7.2677,
+      "step": 710
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 2e-05,
+      "loss": 7.2592,
+      "step": 720
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 2e-05,
+      "loss": 7.2564,
+      "step": 730
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 2e-05,
+      "loss": 7.2111,
+      "step": 740
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 2e-05,
+      "loss": 7.224,
+      "step": 750
+    },
+    {
+      "epoch": 3.86,
+      "eval_accuracy": 0.08311436007920397,
+      "eval_loss": 7.2554144859313965,
+      "eval_runtime": 169.0481,
+      "eval_samples_per_second": 142.297,
+      "eval_steps_per_second": 2.224,
+      "step": 750
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 2e-05,
+      "loss": 7.2082,
+      "step": 760
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 2e-05,
+      "loss": 7.2165,
+      "step": 770
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 2e-05,
+      "loss": 7.7143,
+      "step": 780
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 2e-05,
+      "loss": 7.1994,
+      "step": 790
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 2e-05,
+      "loss": 7.1851,
+      "step": 800
+    },
+    {
+      "epoch": 4.12,
+      "eval_accuracy": 0.0853165114983084,
+      "eval_loss": 7.208171844482422,
+      "eval_runtime": 169.0677,
+      "eval_samples_per_second": 142.28,
+      "eval_steps_per_second": 2.224,
+      "step": 800
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 2e-05,
+      "loss": 7.1016,
+      "step": 810
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 2e-05,
+      "loss": 7.1461,
+      "step": 820
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 2e-05,
+      "loss": 7.1433,
+      "step": 830
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 2e-05,
+      "loss": 7.1318,
+      "step": 840
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 2e-05,
+      "loss": 7.1327,
+      "step": 850
+    },
+    {
+      "epoch": 4.38,
+      "eval_accuracy": 0.08777750406730751,
+      "eval_loss": 7.16784143447876,
+      "eval_runtime": 169.0166,
+      "eval_samples_per_second": 142.323,
+      "eval_steps_per_second": 2.225,
+      "step": 850
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 2e-05,
+      "loss": 7.1428,
+      "step": 860
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 2e-05,
+      "loss": 7.1264,
+      "step": 870
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 2e-05,
+      "loss": 7.1309,
+      "step": 880
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 2e-05,
+      "loss": 7.1299,
+      "step": 890
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 2e-05,
+      "loss": 7.0947,
+      "step": 900
+    },
+    {
+      "epoch": 4.64,
+      "eval_accuracy": 0.09090520258816484,
+      "eval_loss": 7.132561683654785,
+      "eval_runtime": 168.9513,
+      "eval_samples_per_second": 142.378,
+      "eval_steps_per_second": 2.225,
+      "step": 900
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 2e-05,
+      "loss": 7.0953,
+      "step": 910
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2e-05,
+      "loss": 7.1143,
+      "step": 920
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 2e-05,
+      "loss": 7.0972,
+      "step": 930
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2e-05,
+      "loss": 7.109,
+      "step": 940
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 2e-05,
+      "loss": 7.0761,
+      "step": 950
+    },
+    {
+      "epoch": 4.89,
+      "eval_accuracy": 0.09188660007613132,
+      "eval_loss": 7.10692834854126,
+      "eval_runtime": 168.925,
+      "eval_samples_per_second": 142.4,
+      "eval_steps_per_second": 2.226,
+      "step": 950
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 2e-05,
+      "loss": 7.0359,
+      "step": 960
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 2e-05,
+      "loss": 7.0479,
+      "step": 970
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 2e-05,
+      "loss": 7.5649,
+      "step": 980
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 2e-05,
+      "loss": 7.0153,
+      "step": 990
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 2e-05,
+      "loss": 7.0551,
+      "step": 1000
+    },
+    {
+      "epoch": 5.15,
+      "eval_accuracy": 0.0943313810914526,
+      "eval_loss": 7.080649375915527,
+      "eval_runtime": 169.1004,
+      "eval_samples_per_second": 142.253,
+      "eval_steps_per_second": 2.224,
+      "step": 1000
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 2e-05,
+      "loss": 7.0119,
+      "step": 1010
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 2e-05,
+      "loss": 7.0506,
+      "step": 1020
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 2e-05,
+      "loss": 7.0266,
+      "step": 1030
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 2e-05,
+      "loss": 7.0121,
+      "step": 1040
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 2e-05,
+      "loss": 7.0389,
+      "step": 1050
+    },
+    {
+      "epoch": 5.41,
+      "eval_accuracy": 0.09524817664480986,
+      "eval_loss": 7.058758735656738,
+      "eval_runtime": 169.0231,
+      "eval_samples_per_second": 142.318,
+      "eval_steps_per_second": 2.225,
+      "step": 1050
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 2e-05,
+      "loss": 7.002,
+      "step": 1060
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 2e-05,
+      "loss": 7.0223,
+      "step": 1070
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 2e-05,
+      "loss": 6.9822,
+      "step": 1080
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 2e-05,
+      "loss": 6.9993,
+      "step": 1090
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 2e-05,
+      "loss": 7.0226,
+      "step": 1100
+    },
+    {
+      "epoch": 5.67,
+      "eval_accuracy": 0.09638965813145792,
+      "eval_loss": 7.037946701049805,
+      "eval_runtime": 169.0695,
+      "eval_samples_per_second": 142.279,
+      "eval_steps_per_second": 2.224,
+      "step": 1100
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 2e-05,
+      "loss": 7.0049,
+      "step": 1110
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 2e-05,
+      "loss": 6.9835,
+      "step": 1120
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 2e-05,
+      "loss": 6.9924,
+      "step": 1130
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 2e-05,
+      "loss": 6.9933,
+      "step": 1140
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 2e-05,
+      "loss": 6.9992,
+      "step": 1150
+    },
+    {
+      "epoch": 5.92,
+      "eval_accuracy": 0.09752125105444163,
+      "eval_loss": 7.01423454284668,
+      "eval_runtime": 169.1114,
+      "eval_samples_per_second": 142.244,
+      "eval_steps_per_second": 2.223,
+      "step": 1150
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 2e-05,
+      "loss": 6.9667,
+      "step": 1160
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 2e-05,
+      "loss": 7.4753,
+      "step": 1170
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 2e-05,
+      "loss": 6.9355,
+      "step": 1180
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 2e-05,
+      "loss": 6.9903,
+      "step": 1190
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 2e-05,
+      "loss": 6.9382,
+      "step": 1200
+    },
+    {
+      "epoch": 6.18,
+      "eval_accuracy": 0.09855111513654274,
+      "eval_loss": 6.997906684875488,
+      "eval_runtime": 169.2157,
+      "eval_samples_per_second": 142.156,
+      "eval_steps_per_second": 2.222,
+      "step": 1200
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 2e-05,
+      "loss": 6.9745,
+      "step": 1210
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 2e-05,
+      "loss": 6.9784,
+      "step": 1220
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 2e-05,
+      "loss": 6.9698,
+      "step": 1230
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 2e-05,
+      "loss": 6.9203,
+      "step": 1240
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 2e-05,
+      "loss": 6.956,
+      "step": 1250
+    },
+    {
+      "epoch": 6.44,
+      "eval_accuracy": 0.09872446648175658,
+      "eval_loss": 6.982813358306885,
+      "eval_runtime": 168.9414,
+      "eval_samples_per_second": 142.387,
+      "eval_steps_per_second": 2.226,
+      "step": 1250
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 2e-05,
+      "loss": 6.9476,
+      "step": 1260
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 2e-05,
+      "loss": 6.9329,
+      "step": 1270
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 2e-05,
+      "loss": 6.9346,
+      "step": 1280
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 2e-05,
+      "loss": 6.9047,
+      "step": 1290
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 2e-05,
+      "loss": 6.9425,
+      "step": 1300
+    },
+    {
+      "epoch": 6.7,
+      "eval_accuracy": 0.1007779473820151,
+      "eval_loss": 6.961859703063965,
+      "eval_runtime": 169.104,
+      "eval_samples_per_second": 142.25,
+      "eval_steps_per_second": 2.223,
+      "step": 1300
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 2e-05,
+      "loss": 6.9086,
+      "step": 1310
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 2e-05,
+      "loss": 6.9056,
+      "step": 1320
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 2e-05,
+      "loss": 6.902,
+      "step": 1330
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 2e-05,
+      "loss": 6.9013,
+      "step": 1340
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 2e-05,
+      "loss": 6.8872,
+      "step": 1350
+    },
+    {
+      "epoch": 6.96,
+      "eval_accuracy": 0.10138427920145159,
+      "eval_loss": 6.946805477142334,
+      "eval_runtime": 169.0,
+      "eval_samples_per_second": 142.337,
+      "eval_steps_per_second": 2.225,
+      "step": 1350
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2e-05,
+      "loss": 7.4327,
+      "step": 1360
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2e-05,
+      "loss": 6.8642,
+      "step": 1370
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2e-05,
+      "loss": 6.9062,
+      "step": 1380
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2e-05,
+      "loss": 6.9209,
+      "step": 1390
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2e-05,
+      "loss": 6.8848,
+      "step": 1400
+    },
+    {
+      "epoch": 7.22,
+      "eval_accuracy": 0.1024037766438749,
+      "eval_loss": 6.931956768035889,
+      "eval_runtime": 169.0148,
+      "eval_samples_per_second": 142.325,
+      "eval_steps_per_second": 2.225,
+      "step": 1400
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2e-05,
+      "loss": 6.9138,
+      "step": 1410
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2e-05,
+      "loss": 6.8736,
+      "step": 1420
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2e-05,
+      "loss": 6.9051,
+      "step": 1430
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2e-05,
+      "loss": 6.8833,
+      "step": 1440
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2e-05,
+      "loss": 6.8578,
+      "step": 1450
+    },
+    {
+      "epoch": 7.47,
+      "eval_accuracy": 0.10391318947463475,
+      "eval_loss": 6.9190216064453125,
+      "eval_runtime": 168.9483,
+      "eval_samples_per_second": 142.381,
+      "eval_steps_per_second": 2.226,
+      "step": 1450
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2e-05,
+      "loss": 6.8757,
+      "step": 1460
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2e-05,
+      "loss": 6.8638,
+      "step": 1470
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2e-05,
+      "loss": 6.8717,
+      "step": 1480
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2e-05,
+      "loss": 6.8967,
+      "step": 1490
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2e-05,
+      "loss": 6.8699,
+      "step": 1500
+    },
+    {
+      "epoch": 7.73,
+      "eval_accuracy": 0.10498492735216337,
+      "eval_loss": 6.902231216430664,
+      "eval_runtime": 168.9322,
+      "eval_samples_per_second": 142.394,
+      "eval_steps_per_second": 2.226,
+      "step": 1500
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2e-05,
+      "loss": 6.867,
+      "step": 1510
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2e-05,
+      "loss": 6.8533,
+      "step": 1520
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2e-05,
+      "loss": 6.8669,
+      "step": 1530
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2e-05,
+      "loss": 6.862,
+      "step": 1540
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2e-05,
+      "loss": 6.8402,
+      "step": 1550
+    },
+    {
+      "epoch": 7.99,
+      "eval_accuracy": 0.10569349761970831,
+      "eval_loss": 6.8910064697265625,
+      "eval_runtime": 168.8729,
+      "eval_samples_per_second": 142.444,
+      "eval_steps_per_second": 2.227,
+      "step": 1550
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2e-05,
+      "loss": 7.3336,
+      "step": 1560
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2e-05,
+      "loss": 6.8451,
+      "step": 1570
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2e-05,
+      "loss": 6.8418,
+      "step": 1580
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2e-05,
+      "loss": 6.8648,
+      "step": 1590
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2e-05,
+      "loss": 6.8172,
+      "step": 1600
+    },
+    {
+      "epoch": 8.25,
+      "eval_accuracy": 0.1068832420259414,
+      "eval_loss": 6.8729939460754395,
+      "eval_runtime": 168.964,
+      "eval_samples_per_second": 142.368,
+      "eval_steps_per_second": 2.225,
+      "step": 1600
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2e-05,
+      "loss": 6.832,
+      "step": 1610
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2e-05,
+      "loss": 6.8032,
+      "step": 1620
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2e-05,
+      "loss": 6.8527,
+      "step": 1630
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2e-05,
+      "loss": 6.8175,
+      "step": 1640
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2e-05,
+      "loss": 6.823,
+      "step": 1650
+    },
+    {
+      "epoch": 8.5,
+      "eval_accuracy": 0.10725955083624485,
+      "eval_loss": 6.86623477935791,
+      "eval_runtime": 168.9967,
+      "eval_samples_per_second": 142.34,
+      "eval_steps_per_second": 2.225,
+      "step": 1650
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2e-05,
+      "loss": 6.8428,
+      "step": 1660
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2e-05,
+      "loss": 6.8136,
+      "step": 1670
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2e-05,
+      "loss": 6.828,
+      "step": 1680
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2e-05,
+      "loss": 6.8215,
+      "step": 1690
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2e-05,
+      "loss": 6.8028,
+      "step": 1700
+    },
+    {
+      "epoch": 8.76,
+      "eval_accuracy": 0.1081789668246189,
+      "eval_loss": 6.848670482635498,
+      "eval_runtime": 168.985,
+      "eval_samples_per_second": 142.35,
+      "eval_steps_per_second": 2.225,
+      "step": 1700
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2e-05,
+      "loss": 6.8065,
+      "step": 1710
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2e-05,
+      "loss": 6.8185,
+      "step": 1720
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2e-05,
+      "loss": 6.8206,
+      "step": 1730
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2e-05,
+      "loss": 6.8051,
+      "step": 1740
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2e-05,
+      "loss": 7.3146,
+      "step": 1750
+    },
+    {
+      "epoch": 9.02,
+      "eval_accuracy": 0.10825639090130809,
+      "eval_loss": 6.8400139808654785,
+      "eval_runtime": 168.9057,
+      "eval_samples_per_second": 142.417,
+      "eval_steps_per_second": 2.226,
+      "step": 1750
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2e-05,
+      "loss": 6.8451,
+      "step": 1760
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2e-05,
+      "loss": 6.8243,
+      "step": 1770
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2e-05,
+      "loss": 6.8415,
+      "step": 1780
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2e-05,
+      "loss": 6.7823,
+      "step": 1790
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2e-05,
+      "loss": 6.8014,
+      "step": 1800
+    },
+    {
+      "epoch": 9.28,
+      "eval_accuracy": 0.10921870585675643,
+      "eval_loss": 6.8303141593933105,
+      "eval_runtime": 169.0668,
+      "eval_samples_per_second": 142.281,
+      "eval_steps_per_second": 2.224,
+      "step": 1800
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2e-05,
+      "loss": 6.7885,
+      "step": 1810
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2e-05,
+      "loss": 6.7663,
+      "step": 1820
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2e-05,
+      "loss": 6.7476,
+      "step": 1830
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2e-05,
+      "loss": 6.8016,
+      "step": 1840
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2e-05,
+      "loss": 6.8028,
+      "step": 1850
+    },
+    {
+      "epoch": 9.53,
+      "eval_accuracy": 0.108833053639469,
+      "eval_loss": 6.822628021240234,
+      "eval_runtime": 169.002,
+      "eval_samples_per_second": 142.336,
+      "eval_steps_per_second": 2.225,
+      "step": 1850
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2e-05,
+      "loss": 6.7865,
+      "step": 1860
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2e-05,
+      "loss": 6.7951,
+      "step": 1870
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2e-05,
+      "loss": 6.7739,
+      "step": 1880
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2e-05,
+      "loss": 6.7718,
+      "step": 1890
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2e-05,
+      "loss": 6.7817,
+      "step": 1900
+    },
+    {
+      "epoch": 9.79,
+      "eval_accuracy": 0.11072009541672845,
+      "eval_loss": 6.807923316955566,
+      "eval_runtime": 168.9193,
+      "eval_samples_per_second": 142.405,
+      "eval_steps_per_second": 2.226,
+      "step": 1900
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2e-05,
+      "loss": 6.7725,
+      "step": 1910
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2e-05,
+      "loss": 6.7347,
+      "step": 1920
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2e-05,
+      "loss": 6.7559,
+      "step": 1930
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2e-05,
+      "loss": 6.7409,
+      "step": 1940
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2e-05,
+      "loss": 7.28,
+      "step": 1950
+    },
+    {
+      "epoch": 10.05,
+      "eval_accuracy": 0.11147970049664972,
+      "eval_loss": 6.802090167999268,
+      "eval_runtime": 169.0764,
+      "eval_samples_per_second": 142.273,
+      "eval_steps_per_second": 2.224,
+      "step": 1950
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2e-05,
+      "loss": 6.7606,
+      "step": 1960
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2e-05,
+      "loss": 6.7658,
+      "step": 1970
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2e-05,
+      "loss": 6.7526,
+      "step": 1980
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2e-05,
+      "loss": 6.7531,
+      "step": 1990
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2e-05,
+      "loss": 6.7624,
+      "step": 2000
+    },
+    {
+      "epoch": 10.31,
+      "eval_accuracy": 0.11175787497505224,
+      "eval_loss": 6.793049335479736,
+      "eval_runtime": 169.0207,
+      "eval_samples_per_second": 142.32,
+      "eval_steps_per_second": 2.225,
+      "step": 2000
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2e-05,
+      "loss": 6.7658,
+      "step": 2010
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2e-05,
+      "loss": 6.7513,
+      "step": 2020
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2e-05,
+      "loss": 6.7618,
+      "step": 2030
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2e-05,
+      "loss": 6.7146,
+      "step": 2040
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2e-05,
+      "loss": 6.7416,
+      "step": 2050
+    },
+    {
+      "epoch": 10.56,
+      "eval_accuracy": 0.11243676542273737,
+      "eval_loss": 6.786773681640625,
+      "eval_runtime": 168.9144,
+      "eval_samples_per_second": 142.409,
+      "eval_steps_per_second": 2.226,
+      "step": 2050
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2e-05,
+      "loss": 6.7629,
+      "step": 2060
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2e-05,
+      "loss": 6.7654,
+      "step": 2070
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2e-05,
+      "loss": 6.7179,
+      "step": 2080
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2e-05,
+      "loss": 6.7545,
+      "step": 2090
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2e-05,
+      "loss": 6.7288,
+      "step": 2100
+    },
+    {
+      "epoch": 10.82,
+      "eval_accuracy": 0.11334799297462662,
+      "eval_loss": 6.780516147613525,
+      "eval_runtime": 168.9613,
+      "eval_samples_per_second": 142.37,
+      "eval_steps_per_second": 2.225,
+      "step": 2100
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2e-05,
+      "loss": 6.751,
+      "step": 2110
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2e-05,
+      "loss": 6.7746,
+      "step": 2120
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2e-05,
+      "loss": 6.7309,
+      "step": 2130
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2e-05,
+      "loss": 7.2015,
+      "step": 2140
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2e-05,
+      "loss": 6.7468,
+      "step": 2150
+    },
+    {
+      "epoch": 11.08,
+      "eval_accuracy": 0.11234059549981652,
+      "eval_loss": 6.772030353546143,
+      "eval_runtime": 169.0603,
+      "eval_samples_per_second": 142.287,
+      "eval_steps_per_second": 2.224,
+      "step": 2150
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2e-05,
+      "loss": 6.6947,
+      "step": 2160
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2e-05,
+      "loss": 6.7154,
+      "step": 2170
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2e-05,
+      "loss": 6.6892,
+      "step": 2180
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2e-05,
+      "loss": 6.7256,
+      "step": 2190
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2e-05,
+      "loss": 6.7387,
+      "step": 2200
+    },
+    {
+      "epoch": 11.34,
+      "eval_accuracy": 0.11353738741762727,
+      "eval_loss": 6.7636189460754395,
+      "eval_runtime": 168.99,
+      "eval_samples_per_second": 142.346,
+      "eval_steps_per_second": 2.225,
+      "step": 2200
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2e-05,
+      "loss": 6.6785,
+      "step": 2210
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2e-05,
+      "loss": 6.6991,
+      "step": 2220
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2e-05,
+      "loss": 6.7317,
+      "step": 2230
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2e-05,
+      "loss": 6.7243,
+      "step": 2240
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2e-05,
+      "loss": 6.7242,
+      "step": 2250
+    },
+    {
+      "epoch": 11.6,
+      "eval_accuracy": 0.11342761785068739,
+      "eval_loss": 6.755679607391357,
+      "eval_runtime": 169.0043,
+      "eval_samples_per_second": 142.334,
+      "eval_steps_per_second": 2.225,
+      "step": 2250
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2e-05,
+      "loss": 6.7085,
+      "step": 2260
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2e-05,
+      "loss": 6.7231,
+      "step": 2270
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2e-05,
+      "loss": 6.6973,
+      "step": 2280
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2e-05,
+      "loss": 6.6921,
+      "step": 2290
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2e-05,
+      "loss": 6.702,
+      "step": 2300
+    },
+    {
+      "epoch": 11.85,
+      "eval_accuracy": 0.11411896604340402,
+      "eval_loss": 6.749605178833008,
+      "eval_runtime": 169.2579,
+      "eval_samples_per_second": 142.12,
+      "eval_steps_per_second": 2.221,
+      "step": 2300
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2e-05,
+      "loss": 6.7191,
+      "step": 2310
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2e-05,
+      "loss": 6.6818,
+      "step": 2320
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2e-05,
+      "loss": 7.236,
+      "step": 2330
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2e-05,
+      "loss": 6.6929,
+      "step": 2340
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2e-05,
+      "loss": 6.6662,
+      "step": 2350
+    },
+    {
+      "epoch": 12.11,
+      "eval_accuracy": 0.11496613116598788,
+      "eval_loss": 6.743268013000488,
+      "eval_runtime": 169.0391,
+      "eval_samples_per_second": 142.304,
+      "eval_steps_per_second": 2.224,
+      "step": 2350
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2e-05,
+      "loss": 6.7133,
+      "step": 2360
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2e-05,
+      "loss": 6.7061,
+      "step": 2370
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2e-05,
+      "loss": 6.6505,
+      "step": 2380
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2e-05,
+      "loss": 6.7027,
+      "step": 2390
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2e-05,
+      "loss": 6.6781,
+      "step": 2400
+    },
+    {
+      "epoch": 12.37,
+      "eval_accuracy": 0.11483893825636936,
+      "eval_loss": 6.736194610595703,
+      "eval_runtime": 168.9398,
+      "eval_samples_per_second": 142.388,
+      "eval_steps_per_second": 2.226,
+      "step": 2400
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2e-05,
+      "loss": 6.7312,
+      "step": 2410
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2e-05,
+      "loss": 6.6969,
+      "step": 2420
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2e-05,
+      "loss": 6.6962,
+      "step": 2430
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2e-05,
+      "loss": 6.729,
+      "step": 2440
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2e-05,
+      "loss": 6.6743,
+      "step": 2450
+    },
+    {
+      "epoch": 12.63,
+      "eval_accuracy": 0.11607017718543403,
+      "eval_loss": 6.727517127990723,
+      "eval_runtime": 169.0176,
+      "eval_samples_per_second": 142.322,
+      "eval_steps_per_second": 2.225,
+      "step": 2450
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2e-05,
+      "loss": 6.6952,
+      "step": 2460
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2e-05,
+      "loss": 6.6939,
+      "step": 2470
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2e-05,
+      "loss": 6.6848,
+      "step": 2480
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2e-05,
+      "loss": 6.6694,
+      "step": 2490
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2e-05,
+      "loss": 6.6843,
+      "step": 2500
+    },
+    {
+      "epoch": 12.88,
+      "eval_accuracy": 0.11647291428955145,
+      "eval_loss": 6.724733829498291,
+      "eval_runtime": 169.033,
+      "eval_samples_per_second": 142.31,
+      "eval_steps_per_second": 2.224,
+      "step": 2500
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2e-05,
+      "loss": 6.6913,
+      "step": 2510
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2e-05,
+      "loss": 6.6832,
+      "step": 2520
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2e-05,
+      "loss": 7.1899,
+      "step": 2530
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2e-05,
+      "loss": 6.683,
+      "step": 2540
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2e-05,
+      "loss": 6.6726,
+      "step": 2550
+    },
+    {
+      "epoch": 13.14,
+      "eval_accuracy": 0.11728262763830573,
+      "eval_loss": 6.712704658508301,
+      "eval_runtime": 169.1283,
+      "eval_samples_per_second": 142.229,
+      "eval_steps_per_second": 2.223,
+      "step": 2550
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2e-05,
+      "loss": 6.6882,
+      "step": 2560
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 2e-05,
+      "loss": 6.6761,
+      "step": 2570
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2e-05,
+      "loss": 6.6757,
+      "step": 2580
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 2e-05,
+      "loss": 6.6374,
+      "step": 2590
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2e-05,
+      "loss": 6.6656,
+      "step": 2600
+    },
+    {
+      "epoch": 13.4,
+      "eval_accuracy": 0.11703056957986009,
+      "eval_loss": 6.709805011749268,
+      "eval_runtime": 169.0648,
+      "eval_samples_per_second": 142.283,
+      "eval_steps_per_second": 2.224,
+      "step": 2600
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 2e-05,
+      "loss": 6.6369,
+      "step": 2610
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2e-05,
+      "loss": 6.6619,
+      "step": 2620
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 2e-05,
+      "loss": 6.6576,
+      "step": 2630
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 2e-05,
+      "loss": 6.6737,
+      "step": 2640
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2e-05,
+      "loss": 6.6428,
+      "step": 2650
+    },
+    {
+      "epoch": 13.66,
+      "eval_accuracy": 0.11852539903484209,
+      "eval_loss": 6.701896667480469,
+      "eval_runtime": 169.0251,
+      "eval_samples_per_second": 142.316,
+      "eval_steps_per_second": 2.225,
+      "step": 2650
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2e-05,
+      "loss": 6.6634,
+      "step": 2660
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2e-05,
+      "loss": 6.6575,
+      "step": 2670
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2e-05,
+      "loss": 6.6788,
+      "step": 2680
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2e-05,
+      "loss": 6.659,
+      "step": 2690
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2e-05,
+      "loss": 6.6355,
+      "step": 2700
+    },
+    {
+      "epoch": 13.91,
+      "eval_accuracy": 0.11754058144545079,
+      "eval_loss": 6.697854995727539,
+      "eval_runtime": 169.1175,
+      "eval_samples_per_second": 142.238,
+      "eval_steps_per_second": 2.223,
+      "step": 2700
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2e-05,
+      "loss": 6.6451,
+      "step": 2710
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2e-05,
+      "loss": 7.1351,
+      "step": 2720
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2e-05,
+      "loss": 6.6533,
+      "step": 2730
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2e-05,
+      "loss": 6.6421,
+      "step": 2740
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2e-05,
+      "loss": 6.6521,
+      "step": 2750
+    },
+    {
+      "epoch": 14.17,
+      "eval_accuracy": 0.11875193333953445,
+      "eval_loss": 6.692318439483643,
+      "eval_runtime": 169.1423,
+      "eval_samples_per_second": 142.218,
+      "eval_steps_per_second": 2.223,
+      "step": 2750
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2e-05,
+      "loss": 6.6099,
+      "step": 2760
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2e-05,
+      "loss": 6.6254,
+      "step": 2770
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2e-05,
+      "loss": 6.6545,
+      "step": 2780
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2e-05,
+      "loss": 6.6443,
+      "step": 2790
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2e-05,
+      "loss": 6.6735,
+      "step": 2800
+    },
+    {
+      "epoch": 14.43,
+      "eval_accuracy": 0.1185968174100908,
+      "eval_loss": 6.684227466583252,
+      "eval_runtime": 169.1201,
+      "eval_samples_per_second": 142.236,
+      "eval_steps_per_second": 2.223,
+      "step": 2800
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2e-05,
+      "loss": 6.6032,
+      "step": 2810
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2e-05,
+      "loss": 6.6217,
+      "step": 2820
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2e-05,
+      "loss": 6.6212,
+      "step": 2830
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 2e-05,
+      "loss": 6.636,
+      "step": 2840
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2e-05,
+      "loss": 6.6151,
+      "step": 2850
+    },
+    {
+      "epoch": 14.69,
+      "eval_accuracy": 0.11951815218710449,
+      "eval_loss": 6.679075241088867,
+      "eval_runtime": 168.9877,
+      "eval_samples_per_second": 142.348,
+      "eval_steps_per_second": 2.225,
+      "step": 2850
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2e-05,
+      "loss": 6.6269,
+      "step": 2860
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2e-05,
+      "loss": 6.6356,
+      "step": 2870
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2e-05,
+      "loss": 6.6496,
+      "step": 2880
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2e-05,
+      "loss": 6.6702,
+      "step": 2890
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2e-05,
+      "loss": 6.6248,
+      "step": 2900
+    },
+    {
+      "epoch": 14.94,
+      "eval_accuracy": 0.11975263870094723,
+      "eval_loss": 6.675192832946777,
+      "eval_runtime": 169.15,
+      "eval_samples_per_second": 142.211,
+      "eval_steps_per_second": 2.223,
+      "step": 2900
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 2e-05,
+      "loss": 6.6325,
+      "step": 2910
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2e-05,
+      "loss": 7.124,
+      "step": 2920
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 2e-05,
+      "loss": 6.6072,
+      "step": 2930
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 2e-05,
+      "loss": 6.6165,
+      "step": 2940
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 2e-05,
+      "loss": 6.6427,
+      "step": 2950
+    },
+    {
+      "epoch": 15.21,
+      "eval_accuracy": 0.12074793730646591,
+      "eval_loss": 6.666472434997559,
+      "eval_runtime": 169.0435,
+      "eval_samples_per_second": 142.301,
+      "eval_steps_per_second": 2.224,
+      "step": 2950
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 2e-05,
+      "loss": 6.6282,
+      "step": 2960
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 2e-05,
+      "loss": 6.6127,
+      "step": 2970
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 2e-05,
+      "loss": 6.6067,
+      "step": 2980
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 2e-05,
+      "loss": 6.6559,
+      "step": 2990
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 2e-05,
+      "loss": 6.5947,
+      "step": 3000
+    },
+    {
+      "epoch": 15.46,
+      "eval_accuracy": 0.12065681925283153,
+      "eval_loss": 6.663944721221924,
+      "eval_runtime": 169.0216,
+      "eval_samples_per_second": 142.319,
+      "eval_steps_per_second": 2.225,
+      "step": 3000
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 2e-05,
+      "loss": 6.62,
+      "step": 3010
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 2e-05,
+      "loss": 6.6387,
+      "step": 3020
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 2e-05,
+      "loss": 6.6129,
+      "step": 3030
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 2e-05,
+      "loss": 6.6413,
+      "step": 3040
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 2e-05,
+      "loss": 6.6199,
+      "step": 3050
+    },
+    {
+      "epoch": 15.72,
+      "eval_accuracy": 0.12169938085571898,
+      "eval_loss": 6.659843921661377,
+      "eval_runtime": 169.0987,
+      "eval_samples_per_second": 142.254,
+      "eval_steps_per_second": 2.224,
+      "step": 3050
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 2e-05,
+      "loss": 6.6067,
+      "step": 3060
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 2e-05,
+      "loss": 6.6333,
+      "step": 3070
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 2e-05,
+      "loss": 6.619,
+      "step": 3080
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 2e-05,
+      "loss": 6.6274,
+      "step": 3090
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 2e-05,
+      "loss": 6.6127,
+      "step": 3100
+    },
+    {
+      "epoch": 15.98,
+      "eval_accuracy": 0.12192011294272531,
+      "eval_loss": 6.659284591674805,
+      "eval_runtime": 169.0589,
+      "eval_samples_per_second": 142.288,
+      "eval_steps_per_second": 2.224,
+      "step": 3100
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 2e-05,
+      "loss": 7.1321,
+      "step": 3110
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 2e-05,
+      "loss": 6.5976,
+      "step": 3120
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 2e-05,
+      "loss": 6.6261,
+      "step": 3130
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 2e-05,
+      "loss": 6.6105,
+      "step": 3140
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 2e-05,
+      "loss": 6.6031,
+      "step": 3150
+    },
+    {
+      "epoch": 16.24,
+      "eval_accuracy": 0.12264471521223672,
+      "eval_loss": 6.651196002960205,
+      "eval_runtime": 169.0792,
+      "eval_samples_per_second": 142.271,
+      "eval_steps_per_second": 2.224,
+      "step": 3150
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 2e-05,
+      "loss": 6.5796,
+      "step": 3160
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 2e-05,
+      "loss": 6.5983,
+      "step": 3170
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 2e-05,
+      "loss": 6.5812,
+      "step": 3180
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 2e-05,
+      "loss": 6.6298,
+      "step": 3190
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 2e-05,
+      "loss": 6.5742,
+      "step": 3200
+    },
+    {
+      "epoch": 16.49,
+      "eval_accuracy": 0.12269288211432908,
+      "eval_loss": 6.64845609664917,
+      "eval_runtime": 169.2018,
+      "eval_samples_per_second": 142.168,
+      "eval_steps_per_second": 2.222,
+      "step": 3200
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 2e-05,
+      "loss": 6.6001,
+      "step": 3210
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 2e-05,
+      "loss": 6.6124,
+      "step": 3220
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 2e-05,
+      "loss": 6.6131,
+      "step": 3230
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 2e-05,
+      "loss": 6.6051,
+      "step": 3240
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 2e-05,
+      "loss": 6.621,
+      "step": 3250
+    },
+    {
+      "epoch": 16.75,
+      "eval_accuracy": 0.12212792135914549,
+      "eval_loss": 6.647208213806152,
+      "eval_runtime": 169.0564,
+      "eval_samples_per_second": 142.29,
+      "eval_steps_per_second": 2.224,
+      "step": 3250
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 2e-05,
+      "loss": 6.597,
+      "step": 3260
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 2e-05,
+      "loss": 6.6102,
+      "step": 3270
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 2e-05,
+      "loss": 6.5693,
+      "step": 3280
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 2e-05,
+      "loss": 6.5589,
+      "step": 3290
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 2e-05,
+      "loss": 7.0655,
+      "step": 3300
+    },
+    {
+      "epoch": 17.01,
+      "eval_accuracy": 0.123154659717467,
+      "eval_loss": 6.636902332305908,
+      "eval_runtime": 169.0684,
+      "eval_samples_per_second": 142.28,
+      "eval_steps_per_second": 2.224,
+      "step": 3300
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 2e-05,
+      "loss": 6.5935,
+      "step": 3310
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 2e-05,
+      "loss": 6.5824,
+      "step": 3320
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 2e-05,
+      "loss": 6.5584,
+      "step": 3330
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 2e-05,
+      "loss": 6.5853,
+      "step": 3340
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 2e-05,
+      "loss": 6.5866,
+      "step": 3350
+    },
+    {
+      "epoch": 17.27,
+      "eval_accuracy": 0.12344173360979259,
+      "eval_loss": 6.637628555297852,
+      "eval_runtime": 169.0659,
+      "eval_samples_per_second": 142.282,
+      "eval_steps_per_second": 2.224,
+      "step": 3350
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 2e-05,
+      "loss": 6.5515,
+      "step": 3360
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 2e-05,
+      "loss": 6.623,
+      "step": 3370
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 2e-05,
+      "loss": 6.5592,
+      "step": 3380
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 2e-05,
+      "loss": 6.5367,
+      "step": 3390
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 2e-05,
+      "loss": 6.6098,
+      "step": 3400
+    },
+    {
+      "epoch": 17.52,
+      "eval_accuracy": 0.12519736165745915,
+      "eval_loss": 6.631270885467529,
+      "eval_runtime": 169.1186,
+      "eval_samples_per_second": 142.237,
+      "eval_steps_per_second": 2.223,
+      "step": 3400
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 2e-05,
+      "loss": 6.5829,
+      "step": 3410
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 2e-05,
+      "loss": 6.564,
+      "step": 3420
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 2e-05,
+      "loss": 6.5818,
+      "step": 3430
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 2e-05,
+      "loss": 6.5715,
+      "step": 3440
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 2e-05,
+      "loss": 6.5676,
+      "step": 3450
+    },
+    {
+      "epoch": 17.78,
+      "eval_accuracy": 0.124823313950876,
+      "eval_loss": 6.625356674194336,
+      "eval_runtime": 169.1641,
+      "eval_samples_per_second": 142.199,
+      "eval_steps_per_second": 2.223,
+      "step": 3450
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 2e-05,
+      "loss": 6.5796,
+      "step": 3460
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 2e-05,
+      "loss": 6.595,
+      "step": 3470
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 2e-05,
+      "loss": 6.5946,
+      "step": 3480
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 2e-05,
+      "loss": 6.5585,
+      "step": 3490
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 2e-05,
+      "loss": 7.0636,
+      "step": 3500
+    },
+    {
+      "epoch": 18.04,
+      "eval_accuracy": 0.1255522859600729,
+      "eval_loss": 6.622625350952148,
+      "eval_runtime": 169.0712,
+      "eval_samples_per_second": 142.277,
+      "eval_steps_per_second": 2.224,
+      "step": 3500
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 2e-05,
+      "loss": 6.5962,
+      "step": 3510
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2e-05,
+      "loss": 6.544,
+      "step": 3520
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2e-05,
+      "loss": 6.5729,
+      "step": 3530
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2e-05,
+      "loss": 6.5665,
+      "step": 3540
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2e-05,
+      "loss": 6.5444,
+      "step": 3550
+    },
+    {
+      "epoch": 18.3,
+      "eval_accuracy": 0.12534055374011083,
+      "eval_loss": 6.616397857666016,
+      "eval_runtime": 169.0563,
+      "eval_samples_per_second": 142.29,
+      "eval_steps_per_second": 2.224,
+      "step": 3550
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2e-05,
+      "loss": 6.6036,
+      "step": 3560
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 2e-05,
+      "loss": 6.5666,
+      "step": 3570
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 2e-05,
+      "loss": 6.5536,
+      "step": 3580
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 2e-05,
+      "loss": 6.5631,
+      "step": 3590
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2e-05,
+      "loss": 6.561,
+      "step": 3600
+    },
+    {
+      "epoch": 18.55,
+      "eval_accuracy": 0.12544435584844563,
+      "eval_loss": 6.615684986114502,
+      "eval_runtime": 169.0849,
+      "eval_samples_per_second": 142.266,
+      "eval_steps_per_second": 2.224,
+      "step": 3600
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2e-05,
+      "loss": 6.5385,
+      "step": 3610
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2e-05,
+      "loss": 6.5558,
+      "step": 3620
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 2e-05,
+      "loss": 6.5936,
+      "step": 3630
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 2e-05,
+      "loss": 6.5451,
+      "step": 3640
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2e-05,
+      "loss": 6.5882,
+      "step": 3650
+    },
+    {
+      "epoch": 18.81,
+      "eval_accuracy": 0.12569411736714234,
+      "eval_loss": 6.607241153717041,
+      "eval_runtime": 169.0962,
+      "eval_samples_per_second": 142.256,
+      "eval_steps_per_second": 2.224,
+      "step": 3650
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2e-05,
+      "loss": 6.5568,
+      "step": 3660
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2e-05,
+      "loss": 6.5808,
+      "step": 3670
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2e-05,
+      "loss": 6.5542,
+      "step": 3680
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2e-05,
+      "loss": 7.0483,
+      "step": 3690
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2e-05,
+      "loss": 6.5518,
+      "step": 3700
+    },
+    {
+      "epoch": 19.07,
+      "eval_accuracy": 0.12669918361249846,
+      "eval_loss": 6.6064453125,
+      "eval_runtime": 169.0526,
+      "eval_samples_per_second": 142.293,
+      "eval_steps_per_second": 2.224,
+      "step": 3700
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2e-05,
+      "loss": 6.5397,
+      "step": 3710
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2e-05,
+      "loss": 6.5834,
+      "step": 3720
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 2e-05,
+      "loss": 6.5657,
+      "step": 3730
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 2e-05,
+      "loss": 6.5657,
+      "step": 3740
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 2e-05,
+      "loss": 6.5599,
+      "step": 3750
+    },
+    {
+      "epoch": 19.33,
+      "eval_accuracy": 0.12710447874336364,
+      "eval_loss": 6.60552453994751,
+      "eval_runtime": 169.0746,
+      "eval_samples_per_second": 142.274,
+      "eval_steps_per_second": 2.224,
+      "step": 3750
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 2e-05,
+      "loss": 6.5841,
+      "step": 3760
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 2e-05,
+      "loss": 6.5189,
+      "step": 3770
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 2e-05,
+      "loss": 6.551,
+      "step": 3780
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 2e-05,
+      "loss": 6.5466,
+      "step": 3790
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 2e-05,
+      "loss": 6.5407,
+      "step": 3800
+    },
+    {
+      "epoch": 19.59,
+      "eval_accuracy": 0.12744788950670777,
+      "eval_loss": 6.598654270172119,
+      "eval_runtime": 169.0329,
+      "eval_samples_per_second": 142.31,
+      "eval_steps_per_second": 2.224,
+      "step": 3800
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 2e-05,
+      "loss": 6.5842,
+      "step": 3810
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 2e-05,
+      "loss": 6.5077,
+      "step": 3820
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 2e-05,
+      "loss": 6.5263,
+      "step": 3830
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 2e-05,
+      "loss": 6.5459,
+      "step": 3840
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 2e-05,
+      "loss": 6.5373,
+      "step": 3850
+    },
+    {
+      "epoch": 19.84,
+      "eval_accuracy": 0.12802956265143414,
+      "eval_loss": 6.595397472381592,
+      "eval_runtime": 169.0397,
+      "eval_samples_per_second": 142.304,
+      "eval_steps_per_second": 2.224,
+      "step": 3850
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2e-05,
+      "loss": 6.5126,
+      "step": 3860
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 2e-05,
+      "loss": 6.5966,
+      "step": 3870
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 2e-05,
+      "loss": 6.5305,
+      "step": 3880
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 2e-05,
+      "loss": 7.0605,
+      "step": 3890
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 2e-05,
+      "loss": 6.5381,
+      "step": 3900
+    },
+    {
+      "epoch": 20.1,
+      "eval_accuracy": 0.1281530363855393,
+      "eval_loss": 6.589879989624023,
+      "eval_runtime": 169.1405,
+      "eval_samples_per_second": 142.219,
+      "eval_steps_per_second": 2.223,
+      "step": 3900
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 2e-05,
+      "loss": 6.5167,
+      "step": 3910
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 2e-05,
+      "loss": 6.5289,
+      "step": 3920
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 2e-05,
+      "loss": 6.5296,
+      "step": 3930
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 2e-05,
+      "loss": 6.4745,
+      "step": 3940
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 2e-05,
+      "loss": 6.5517,
+      "step": 3950
+    },
+    {
+      "epoch": 20.36,
+      "eval_accuracy": 0.12833386885114662,
+      "eval_loss": 6.58883810043335,
+      "eval_runtime": 168.9823,
+      "eval_samples_per_second": 142.352,
+      "eval_steps_per_second": 2.225,
+      "step": 3950
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 2e-05,
+      "loss": 6.5466,
+      "step": 3960
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 2e-05,
+      "loss": 6.5482,
+      "step": 3970
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 2e-05,
+      "loss": 6.5047,
+      "step": 3980
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 2e-05,
+      "loss": 6.499,
+      "step": 3990
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 2e-05,
+      "loss": 6.5371,
+      "step": 4000
+    },
+    {
+      "epoch": 20.62,
+      "eval_accuracy": 0.1295166100847349,
+      "eval_loss": 6.585357189178467,
+      "eval_runtime": 169.1057,
+      "eval_samples_per_second": 142.248,
+      "eval_steps_per_second": 2.223,
+      "step": 4000
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 2e-05,
+      "loss": 6.5112,
+      "step": 4010
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 2e-05,
+      "loss": 6.538,
+      "step": 4020
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 2e-05,
+      "loss": 6.5547,
+      "step": 4030
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 2e-05,
+      "loss": 6.5062,
+      "step": 4040
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 2e-05,
+      "loss": 6.5819,
+      "step": 4050
+    },
+    {
+      "epoch": 20.87,
+      "eval_accuracy": 0.12822338785929896,
+      "eval_loss": 6.582521915435791,
+      "eval_runtime": 169.044,
+      "eval_samples_per_second": 142.3,
+      "eval_steps_per_second": 2.224,
+      "step": 4050
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 2e-05,
+      "loss": 6.4947,
+      "step": 4060
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 2e-05,
+      "loss": 6.5279,
+      "step": 4070
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 2e-05,
+      "loss": 7.0033,
+      "step": 4080
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 2e-05,
+      "loss": 6.5198,
+      "step": 4090
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 2e-05,
+      "loss": 6.5425,
+      "step": 4100
+    },
+    {
+      "epoch": 21.13,
+      "eval_accuracy": 0.12887030827275436,
+      "eval_loss": 6.579444408416748,
+      "eval_runtime": 169.204,
+      "eval_samples_per_second": 142.166,
+      "eval_steps_per_second": 2.222,
+      "step": 4100
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 2e-05,
+      "loss": 6.4964,
+      "step": 4110
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 2e-05,
+      "loss": 6.5506,
+      "step": 4120
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 2e-05,
+      "loss": 6.5544,
+      "step": 4130
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 2e-05,
+      "loss": 6.5089,
+      "step": 4140
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 2e-05,
+      "loss": 6.5372,
+      "step": 4150
+    },
+    {
+      "epoch": 21.39,
+      "eval_accuracy": 0.12997385168937536,
+      "eval_loss": 6.575997829437256,
+      "eval_runtime": 169.1292,
+      "eval_samples_per_second": 142.229,
+      "eval_steps_per_second": 2.223,
+      "step": 4150
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 2e-05,
+      "loss": 6.5305,
+      "step": 4160
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 2e-05,
+      "loss": 6.5443,
+      "step": 4170
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 2e-05,
+      "loss": 6.5214,
+      "step": 4180
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 2e-05,
+      "loss": 6.4939,
+      "step": 4190
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 2e-05,
+      "loss": 6.544,
+      "step": 4200
+    },
+    {
+      "epoch": 21.65,
+      "eval_accuracy": 0.13029984560795255,
+      "eval_loss": 6.571804046630859,
+      "eval_runtime": 168.9481,
+      "eval_samples_per_second": 142.381,
+      "eval_steps_per_second": 2.226,
+      "step": 4200
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 2e-05,
+      "loss": 6.5006,
+      "step": 4210
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 2e-05,
+      "loss": 6.5257,
+      "step": 4220
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 2e-05,
+      "loss": 6.5282,
+      "step": 4230
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 2e-05,
+      "loss": 6.5319,
+      "step": 4240
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 2e-05,
+      "loss": 6.5129,
+      "step": 4250
+    },
+    {
+      "epoch": 21.9,
+      "eval_accuracy": 0.13096158217776768,
+      "eval_loss": 6.565962314605713,
+      "eval_runtime": 168.9477,
+      "eval_samples_per_second": 142.381,
+      "eval_steps_per_second": 2.226,
+      "step": 4250
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 2e-05,
+      "loss": 6.4916,
+      "step": 4260
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 2e-05,
+      "loss": 6.9742,
+      "step": 4270
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 2e-05,
+      "loss": 6.4991,
+      "step": 4280
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 2e-05,
+      "loss": 6.516,
+      "step": 4290
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 2e-05,
+      "loss": 6.4798,
+      "step": 4300
+    },
+    {
+      "epoch": 22.16,
+      "eval_accuracy": 0.1304970998721391,
+      "eval_loss": 6.56817102432251,
+      "eval_runtime": 168.9692,
+      "eval_samples_per_second": 142.363,
+      "eval_steps_per_second": 2.225,
+      "step": 4300
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 2e-05,
+      "loss": 6.5255,
+      "step": 4310
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 2e-05,
+      "loss": 6.4953,
+      "step": 4320
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 2e-05,
+      "loss": 6.5558,
+      "step": 4330
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 2e-05,
+      "loss": 6.5029,
+      "step": 4340
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 2e-05,
+      "loss": 6.5556,
+      "step": 4350
+    },
+    {
+      "epoch": 22.42,
+      "eval_accuracy": 0.13152063993455215,
+      "eval_loss": 6.561850547790527,
+      "eval_runtime": 168.9174,
+      "eval_samples_per_second": 142.407,
+      "eval_steps_per_second": 2.226,
+      "step": 4350
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 2e-05,
+      "loss": 6.4888,
+      "step": 4360
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 2e-05,
+      "loss": 6.4516,
+      "step": 4370
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 2e-05,
+      "loss": 6.5203,
+      "step": 4380
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 2e-05,
+      "loss": 6.502,
+      "step": 4390
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 2e-05,
+      "loss": 6.4946,
+      "step": 4400
+    },
+    {
+      "epoch": 22.68,
+      "eval_accuracy": 0.13143446792084687,
+      "eval_loss": 6.558900833129883,
+      "eval_runtime": 168.9992,
+      "eval_samples_per_second": 142.338,
+      "eval_steps_per_second": 2.225,
+      "step": 4400
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 2e-05,
+      "loss": 6.4919,
+      "step": 4410
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 2e-05,
+      "loss": 6.482,
+      "step": 4420
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 2e-05,
+      "loss": 6.5231,
+      "step": 4430
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 2e-05,
+      "loss": 6.5293,
+      "step": 4440
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 2e-05,
+      "loss": 6.5212,
+      "step": 4450
+    },
+    {
+      "epoch": 22.93,
+      "eval_accuracy": 0.13184165697753022,
+      "eval_loss": 6.559324264526367,
+      "eval_runtime": 168.9331,
+      "eval_samples_per_second": 142.394,
+      "eval_steps_per_second": 2.226,
+      "step": 4450
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 2e-05,
+      "loss": 6.4944,
+      "step": 4460
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 2e-05,
+      "loss": 6.9696,
+      "step": 4470
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 2e-05,
+      "loss": 6.5051,
+      "step": 4480
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 2e-05,
+      "loss": 6.4822,
+      "step": 4490
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 2e-05,
+      "loss": 6.5055,
+      "step": 4500
+    },
+    {
+      "epoch": 23.2,
+      "eval_accuracy": 0.1310761376111694,
+      "eval_loss": 6.5552473068237305,
+      "eval_runtime": 169.0117,
+      "eval_samples_per_second": 142.327,
+      "eval_steps_per_second": 2.225,
+      "step": 4500
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 2e-05,
+      "loss": 6.5152,
+      "step": 4510
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 2e-05,
+      "loss": 6.504,
+      "step": 4520
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 2e-05,
+      "loss": 6.5149,
+      "step": 4530
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 2e-05,
+      "loss": 6.4913,
+      "step": 4540
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 2e-05,
+      "loss": 6.4693,
+      "step": 4550
+    },
+    {
+      "epoch": 23.45,
+      "eval_accuracy": 0.13253956813479448,
+      "eval_loss": 6.548061847686768,
+      "eval_runtime": 169.0655,
+      "eval_samples_per_second": 142.282,
+      "eval_steps_per_second": 2.224,
+      "step": 4550
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 2e-05,
+      "loss": 6.4741,
+      "step": 4560
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 2e-05,
+      "loss": 6.5033,
+      "step": 4570
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 2e-05,
+      "loss": 6.5313,
+      "step": 4580
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 2e-05,
+      "loss": 6.4789,
+      "step": 4590
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 2e-05,
+      "loss": 6.4706,
+      "step": 4600
+    },
+    {
+      "epoch": 23.71,
+      "eval_accuracy": 0.13172442724332517,
+      "eval_loss": 6.54693078994751,
+      "eval_runtime": 169.0562,
+      "eval_samples_per_second": 142.29,
+      "eval_steps_per_second": 2.224,
+      "step": 4600
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 2e-05,
+      "loss": 6.5044,
+      "step": 4610
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 2e-05,
+      "loss": 6.5038,
+      "step": 4620
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 2e-05,
+      "loss": 6.5173,
+      "step": 4630
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 2e-05,
+      "loss": 6.5048,
+      "step": 4640
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 2e-05,
+      "loss": 6.495,
+      "step": 4650
+    },
+    {
+      "epoch": 23.97,
+      "eval_accuracy": 0.13237315726859447,
+      "eval_loss": 6.546177387237549,
+      "eval_runtime": 168.9825,
+      "eval_samples_per_second": 142.352,
+      "eval_steps_per_second": 2.225,
+      "step": 4650
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 2e-05,
+      "loss": 6.9878,
+      "step": 4660
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 2e-05,
+      "loss": 6.4801,
+      "step": 4670
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 2e-05,
+      "loss": 6.5032,
+      "step": 4680
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 2e-05,
+      "loss": 6.4802,
+      "step": 4690
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 2e-05,
+      "loss": 6.4901,
+      "step": 4700
+    },
+    {
+      "epoch": 24.23,
+      "eval_accuracy": 0.13279457912397785,
+      "eval_loss": 6.541420936584473,
+      "eval_runtime": 168.8748,
+      "eval_samples_per_second": 142.443,
+      "eval_steps_per_second": 2.227,
+      "step": 4700
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 2e-05,
+      "loss": 6.4895,
+      "step": 4710
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 2e-05,
+      "loss": 6.5127,
+      "step": 4720
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 2e-05,
+      "loss": 6.5062,
+      "step": 4730
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 2e-05,
+      "loss": 6.4874,
+      "step": 4740
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 2e-05,
+      "loss": 6.4936,
+      "step": 4750
+    },
+    {
+      "epoch": 24.48,
+      "eval_accuracy": 0.13342969669901336,
+      "eval_loss": 6.538527011871338,
+      "eval_runtime": 168.9618,
+      "eval_samples_per_second": 142.369,
+      "eval_steps_per_second": 2.225,
+      "step": 4750
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 2e-05,
+      "loss": 6.4868,
+      "step": 4760
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 2e-05,
+      "loss": 6.4657,
+      "step": 4770
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 2e-05,
+      "loss": 6.4732,
+      "step": 4780
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 2e-05,
+      "loss": 6.5089,
+      "step": 4790
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 2e-05,
+      "loss": 6.481,
+      "step": 4800
+    },
+    {
+      "epoch": 24.74,
+      "eval_accuracy": 0.13314294552046066,
+      "eval_loss": 6.536244869232178,
+      "eval_runtime": 168.9165,
+      "eval_samples_per_second": 142.408,
+      "eval_steps_per_second": 2.226,
+      "step": 4800
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 2e-05,
+      "loss": 6.4942,
+      "step": 4810
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 2e-05,
+      "loss": 6.4739,
+      "step": 4820
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 2e-05,
+      "loss": 6.4646,
+      "step": 4830
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 2e-05,
+      "loss": 6.4478,
+      "step": 4840
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 2e-05,
+      "loss": 6.5186,
+      "step": 4850
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.13348454019073938,
+      "eval_loss": 6.535652160644531,
+      "eval_runtime": 168.9724,
+      "eval_samples_per_second": 142.361,
+      "eval_steps_per_second": 2.225,
+      "step": 4850
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 2e-05,
+      "loss": 6.9644,
+      "step": 4860
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 2e-05,
+      "loss": 6.4549,
+      "step": 4870
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 2e-05,
+      "loss": 6.4725,
+      "step": 4880
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 2e-05,
+      "loss": 6.467,
+      "step": 4890
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 2e-05,
+      "loss": 6.4711,
+      "step": 4900
+    },
+    {
+      "epoch": 25.26,
+      "eval_accuracy": 0.1339031847051157,
+      "eval_loss": 6.53087854385376,
+      "eval_runtime": 168.9489,
+      "eval_samples_per_second": 142.38,
+      "eval_steps_per_second": 2.226,
+      "step": 4900
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 2e-05,
+      "loss": 6.4475,
+      "step": 4910
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 2e-05,
+      "loss": 6.4674,
+      "step": 4920
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 2e-05,
+      "loss": 6.4737,
+      "step": 4930
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 2e-05,
+      "loss": 6.5061,
+      "step": 4940
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 2e-05,
+      "loss": 6.4513,
+      "step": 4950
+    },
+    {
+      "epoch": 25.51,
+      "eval_accuracy": 0.13372873418009035,
+      "eval_loss": 6.528397083282471,
+      "eval_runtime": 168.9948,
+      "eval_samples_per_second": 142.342,
+      "eval_steps_per_second": 2.225,
+      "step": 4950
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 2e-05,
+      "loss": 6.4663,
+      "step": 4960
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 2e-05,
+      "loss": 6.4488,
+      "step": 4970
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 2e-05,
+      "loss": 6.4882,
+      "step": 4980
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 2e-05,
+      "loss": 6.4644,
+      "step": 4990
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 2e-05,
+      "loss": 6.4652,
+      "step": 5000
+    },
+    {
+      "epoch": 25.77,
+      "eval_accuracy": 0.1343316375042143,
+      "eval_loss": 6.524177074432373,
+      "eval_runtime": 168.9768,
+      "eval_samples_per_second": 142.357,
+      "eval_steps_per_second": 2.225,
+      "step": 5000
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 2e-05,
+      "loss": 6.5203,
+      "step": 5010
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 2e-05,
+      "loss": 6.4518,
+      "step": 5020
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 2e-05,
+      "loss": 6.4791,
+      "step": 5030
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 2e-05,
+      "loss": 6.4843,
+      "step": 5040
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 2e-05,
+      "loss": 6.9335,
+      "step": 5050
+    },
+    {
+      "epoch": 26.03,
+      "eval_accuracy": 0.13449298626675552,
+      "eval_loss": 6.521671772003174,
+      "eval_runtime": 168.9858,
+      "eval_samples_per_second": 142.349,
+      "eval_steps_per_second": 2.225,
+      "step": 5050
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 2e-05,
+      "loss": 6.4677,
+      "step": 5060
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 2e-05,
+      "loss": 6.4347,
+      "step": 5070
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 2e-05,
+      "loss": 6.446,
+      "step": 5080
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 2e-05,
+      "loss": 6.4386,
+      "step": 5090
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 2e-05,
+      "loss": 6.4747,
+      "step": 5100
+    },
+    {
+      "epoch": 26.29,
+      "eval_accuracy": 0.13449609380922367,
+      "eval_loss": 6.520633697509766,
+      "eval_runtime": 168.9534,
+      "eval_samples_per_second": 142.377,
+      "eval_steps_per_second": 2.225,
+      "step": 5100
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 2e-05,
+      "loss": 6.4392,
+      "step": 5110
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 2e-05,
+      "loss": 6.4715,
+      "step": 5120
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 2e-05,
+      "loss": 6.4502,
+      "step": 5130
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 2e-05,
+      "loss": 6.4566,
+      "step": 5140
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 2e-05,
+      "loss": 6.4702,
+      "step": 5150
+    },
+    {
+      "epoch": 26.54,
+      "eval_accuracy": 0.13501667146291277,
+      "eval_loss": 6.52007532119751,
+      "eval_runtime": 168.9331,
+      "eval_samples_per_second": 142.394,
+      "eval_steps_per_second": 2.226,
+      "step": 5150
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 2e-05,
+      "loss": 6.4778,
+      "step": 5160
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 2e-05,
+      "loss": 6.4631,
+      "step": 5170
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 2e-05,
+      "loss": 6.453,
+      "step": 5180
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 2e-05,
+      "loss": 6.4477,
+      "step": 5190
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 2e-05,
+      "loss": 6.4524,
+      "step": 5200
+    },
+    {
+      "epoch": 26.8,
+      "eval_accuracy": 0.13524723051481435,
+      "eval_loss": 6.515605926513672,
+      "eval_runtime": 168.9736,
+      "eval_samples_per_second": 142.359,
+      "eval_steps_per_second": 2.225,
+      "step": 5200
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 2e-05,
+      "loss": 6.4479,
+      "step": 5210
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 2e-05,
+      "loss": 6.4726,
+      "step": 5220
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 2e-05,
+      "loss": 6.4736,
+      "step": 5230
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 2e-05,
+      "loss": 6.9569,
+      "step": 5240
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 2e-05,
+      "loss": 6.4225,
+      "step": 5250
+    },
+    {
+      "epoch": 27.06,
+      "eval_accuracy": 0.13492381162327038,
+      "eval_loss": 6.515013694763184,
+      "eval_runtime": 168.8677,
+      "eval_samples_per_second": 142.449,
+      "eval_steps_per_second": 2.227,
+      "step": 5250
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 2e-05,
+      "loss": 6.4502,
+      "step": 5260
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 2e-05,
+      "loss": 6.4587,
+      "step": 5270
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 2e-05,
+      "loss": 6.4512,
+      "step": 5280
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 2e-05,
+      "loss": 6.4495,
+      "step": 5290
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 2e-05,
+      "loss": 6.4599,
+      "step": 5300
+    },
+    {
+      "epoch": 27.32,
+      "eval_accuracy": 0.13545740402985318,
+      "eval_loss": 6.511618137359619,
+      "eval_runtime": 168.8908,
+      "eval_samples_per_second": 142.429,
+      "eval_steps_per_second": 2.226,
+      "step": 5300
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 2e-05,
+      "loss": 6.4499,
+      "step": 5310
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 2e-05,
+      "loss": 6.4468,
+      "step": 5320
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 2e-05,
+      "loss": 6.4448,
+      "step": 5330
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 2e-05,
+      "loss": 6.4436,
+      "step": 5340
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 2e-05,
+      "loss": 6.4591,
+      "step": 5350
+    },
+    {
+      "epoch": 27.58,
+      "eval_accuracy": 0.1358384989998707,
+      "eval_loss": 6.509791374206543,
+      "eval_runtime": 169.0437,
+      "eval_samples_per_second": 142.301,
+      "eval_steps_per_second": 2.224,
+      "step": 5350
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 2e-05,
+      "loss": 6.4593,
+      "step": 5360
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 2e-05,
+      "loss": 6.4223,
+      "step": 5370
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 2e-05,
+      "loss": 6.4278,
+      "step": 5380
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 2e-05,
+      "loss": 6.4607,
+      "step": 5390
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 2e-05,
+      "loss": 6.4184,
+      "step": 5400
+    },
+    {
+      "epoch": 27.83,
+      "eval_accuracy": 0.1352895741192505,
+      "eval_loss": 6.509643077850342,
+      "eval_runtime": 169.086,
+      "eval_samples_per_second": 142.265,
+      "eval_steps_per_second": 2.224,
+      "step": 5400
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 2e-05,
+      "loss": 6.455,
+      "step": 5410
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 2e-05,
+      "loss": 6.4447,
+      "step": 5420
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 2e-05,
+      "loss": 6.3932,
+      "step": 5430
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 2e-05,
+      "loss": 6.953,
+      "step": 5440
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 2e-05,
+      "loss": 6.43,
+      "step": 5450
+    },
+    {
+      "epoch": 28.09,
+      "eval_accuracy": 0.1360815394195743,
+      "eval_loss": 6.507402420043945,
+      "eval_runtime": 169.2594,
+      "eval_samples_per_second": 142.119,
+      "eval_steps_per_second": 2.221,
+      "step": 5450
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 2e-05,
+      "loss": 6.4622,
+      "step": 5460
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 2e-05,
+      "loss": 6.4898,
+      "step": 5470
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 2e-05,
+      "loss": 6.4297,
+      "step": 5480
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 2e-05,
+      "loss": 6.4151,
+      "step": 5490
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 2e-05,
+      "loss": 6.4604,
+      "step": 5500
+    },
+    {
+      "epoch": 28.35,
+      "eval_accuracy": 0.136740573949904,
+      "eval_loss": 6.4999098777771,
+      "eval_runtime": 169.0085,
+      "eval_samples_per_second": 142.33,
+      "eval_steps_per_second": 2.225,
+      "step": 5500
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 2e-05,
+      "loss": 6.4254,
+      "step": 5510
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 2e-05,
+      "loss": 6.4188,
+      "step": 5520
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 2e-05,
+      "loss": 6.4418,
+      "step": 5530
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 2e-05,
+      "loss": 6.4201,
+      "step": 5540
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 2e-05,
+      "loss": 6.4593,
+      "step": 5550
+    },
+    {
+      "epoch": 28.61,
+      "eval_accuracy": 0.13586421209634997,
+      "eval_loss": 6.499355792999268,
+      "eval_runtime": 169.0692,
+      "eval_samples_per_second": 142.279,
+      "eval_steps_per_second": 2.224,
+      "step": 5550
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 2e-05,
+      "loss": 6.4249,
+      "step": 5560
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 2e-05,
+      "loss": 6.4335,
+      "step": 5570
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 2e-05,
+      "loss": 6.4411,
+      "step": 5580
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 2e-05,
+      "loss": 6.4238,
+      "step": 5590
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 2e-05,
+      "loss": 6.4648,
+      "step": 5600
+    },
+    {
+      "epoch": 28.86,
+      "eval_accuracy": 0.13563087896124043,
+      "eval_loss": 6.498055934906006,
+      "eval_runtime": 168.9957,
+      "eval_samples_per_second": 142.341,
+      "eval_steps_per_second": 2.225,
+      "step": 5600
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 2e-05,
+      "loss": 6.4051,
+      "step": 5610
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 2e-05,
+      "loss": 6.4433,
+      "step": 5620
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 2e-05,
+      "loss": 6.9371,
+      "step": 5630
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 2e-05,
+      "loss": 6.4164,
+      "step": 5640
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 2e-05,
+      "loss": 6.4453,
+      "step": 5650
+    },
+    {
+      "epoch": 29.12,
+      "eval_accuracy": 0.1374106846803503,
+      "eval_loss": 6.494868755340576,
+      "eval_runtime": 169.1796,
+      "eval_samples_per_second": 142.186,
+      "eval_steps_per_second": 2.222,
+      "step": 5650
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 2e-05,
+      "loss": 6.4349,
+      "step": 5660
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 2e-05,
+      "loss": 6.468,
+      "step": 5670
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 2e-05,
+      "loss": 6.4532,
+      "step": 5680
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 2e-05,
+      "loss": 6.4349,
+      "step": 5690
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 2e-05,
+      "loss": 6.4275,
+      "step": 5700
+    },
+    {
+      "epoch": 29.38,
+      "eval_accuracy": 0.13618291226339158,
+      "eval_loss": 6.49544095993042,
+      "eval_runtime": 169.054,
+      "eval_samples_per_second": 142.292,
+      "eval_steps_per_second": 2.224,
+      "step": 5700
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 2e-05,
+      "loss": 6.4304,
+      "step": 5710
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 2e-05,
+      "loss": 6.4465,
+      "step": 5720
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 2e-05,
+      "loss": 6.4144,
+      "step": 5730
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 2e-05,
+      "loss": 6.4441,
+      "step": 5740
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 2e-05,
+      "loss": 6.4165,
+      "step": 5750
+    },
+    {
+      "epoch": 29.64,
+      "eval_accuracy": 0.1368880629366925,
+      "eval_loss": 6.493827819824219,
+      "eval_runtime": 169.0908,
+      "eval_samples_per_second": 142.261,
+      "eval_steps_per_second": 2.224,
+      "step": 5750
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 2e-05,
+      "loss": 6.4493,
+      "step": 5760
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 2e-05,
+      "loss": 6.4144,
+      "step": 5770
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 2e-05,
+      "loss": 6.4323,
+      "step": 5780
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 2e-05,
+      "loss": 6.4063,
+      "step": 5790
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 2e-05,
+      "loss": 6.4211,
+      "step": 5800
+    },
+    {
+      "epoch": 29.89,
+      "eval_accuracy": 0.1375700871847227,
+      "eval_loss": 6.4911370277404785,
+      "eval_runtime": 169.0097,
+      "eval_samples_per_second": 142.329,
+      "eval_steps_per_second": 2.225,
+      "step": 5800
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 2e-05,
+      "loss": 6.3907,
+      "step": 5810
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 2e-05,
+      "loss": 6.4137,
+      "step": 5820
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 2e-05,
+      "loss": 6.9177,
+      "step": 5830
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 2e-05,
+      "loss": 6.4397,
+      "step": 5840
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 2e-05,
+      "loss": 6.4188,
+      "step": 5850
+    },
+    {
+      "epoch": 30.15,
+      "eval_accuracy": 0.13739219384676168,
+      "eval_loss": 6.486028671264648,
+      "eval_runtime": 169.0403,
+      "eval_samples_per_second": 142.303,
+      "eval_steps_per_second": 2.224,
+      "step": 5850
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 2e-05,
+      "loss": 6.395,
+      "step": 5860
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 2e-05,
+      "loss": 6.4017,
+      "step": 5870
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 2e-05,
+      "loss": 6.4429,
+      "step": 5880
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 2e-05,
+      "loss": 6.3932,
+      "step": 5890
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 2e-05,
+      "loss": 6.4337,
+      "step": 5900
+    },
+    {
+      "epoch": 30.41,
+      "eval_accuracy": 0.13796805623574424,
+      "eval_loss": 6.480716228485107,
+      "eval_runtime": 169.1216,
+      "eval_samples_per_second": 142.235,
+      "eval_steps_per_second": 2.223,
+      "step": 5900
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 2e-05,
+      "loss": 6.4171,
+      "step": 5910
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 2e-05,
+      "loss": 6.42,
+      "step": 5920
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 2e-05,
+      "loss": 6.4512,
+      "step": 5930
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 2e-05,
+      "loss": 6.4306,
+      "step": 5940
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 2e-05,
+      "loss": 6.4228,
+      "step": 5950
+    },
+    {
+      "epoch": 30.67,
+      "eval_accuracy": 0.1374551952909284,
+      "eval_loss": 6.487619400024414,
+      "eval_runtime": 169.1032,
+      "eval_samples_per_second": 142.25,
+      "eval_steps_per_second": 2.223,
+      "step": 5950
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 2e-05,
+      "loss": 6.4332,
+      "step": 5960
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 2e-05,
+      "loss": 6.413,
+      "step": 5970
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 2e-05,
+      "loss": 6.3904,
+      "step": 5980
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 2e-05,
+      "loss": 6.4198,
+      "step": 5990
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 2e-05,
+      "loss": 6.3841,
+      "step": 6000
+    },
+    {
+      "epoch": 30.92,
+      "eval_accuracy": 0.1375611450497736,
+      "eval_loss": 6.4811110496521,
+      "eval_runtime": 169.0873,
+      "eval_samples_per_second": 142.264,
+      "eval_steps_per_second": 2.224,
+      "step": 6000
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 2e-05,
+      "loss": 6.4452,
+      "step": 6010
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 2e-05,
+      "loss": 6.9049,
+      "step": 6020
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 2e-05,
+      "loss": 6.4161,
+      "step": 6030
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 2e-05,
+      "loss": 6.3858,
+      "step": 6040
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 2e-05,
+      "loss": 6.4383,
+      "step": 6050
+    },
+    {
+      "epoch": 31.18,
+      "eval_accuracy": 0.13791837603204882,
+      "eval_loss": 6.483151912689209,
+      "eval_runtime": 169.0866,
+      "eval_samples_per_second": 142.264,
+      "eval_steps_per_second": 2.224,
+      "step": 6050
+    },
+    {
+      "epoch": 31.18,
+      "step": 6050,
+      "total_flos": 1.125873723276288e+16,
+      "train_loss": 6.89371246590102,
+      "train_runtime": 29548.4223,
+      "train_samples_per_second": 33.721,
+      "train_steps_per_second": 0.263
+    }
+  ],
+  "max_steps": 7760,
+  "num_train_epochs": 40,
+  "total_flos": 1.125873723276288e+16,
+  "trial_name": null,
+  "trial_params": null
+}