End of training

all_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 59.00721153846154,
-    "eval_accuracy": 0.7319819819819819,
-    "eval_loss": 1.084416389465332,
-    "eval_runtime": 75.3126,
-    "eval_samples_per_second": 5.895,
-    "eval_steps_per_second": 0.093
+    "epoch": 119.00833333333334,
+    "eval_accuracy": 0.5539772727272727,
+    "eval_loss": 1.9018956422805786,
+    "eval_runtime": 133.7632,
+    "eval_samples_per_second": 10.526,
+    "eval_steps_per_second": 0.164
 }

runs/Jan16_11-24-13_GAN-SVR/events.out.tfevents.1737080111.GAN-SVR.864097.1

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bba6adbc35ef809071241cec5b961f758e5a25c8e06a4b0f62104a6bf8356c0e
-size 411
+oid sha256:1b76810f99e49e3c5531f5cf094dc0380534fa9fdda0f2e9bfdf313f9262f6d4
+size 734

test_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 59.00721153846154,
-    "eval_accuracy": 0.7319819819819819,
-    "eval_loss": 1.084416389465332,
-    "eval_runtime": 75.3126,
-    "eval_samples_per_second": 5.895,
-    "eval_steps_per_second": 0.093
+    "epoch": 119.00833333333334,
+    "eval_accuracy": 0.5539772727272727,
+    "eval_loss": 1.9018956422805786,
+    "eval_runtime": 133.7632,
+    "eval_samples_per_second": 10.526,
+    "eval_steps_per_second": 0.164
 }

trainer_state.json

@@ -1,1164 +1,2467 @@
 {
-  "best_metric": 0.7319819819819819,
-  "best_model_checkpoint": "videomae-base-finetuned-ucf101-subset/checkpoint-812",
-  "epoch": 59.00721153846154,
+  "best_metric": 0.5539772727272727,
+  "best_model_checkpoint": "videomae-base-finetuned-ucf101-subset/checkpoint-1600",
+  "epoch": 119.00833333333334,
   "eval_steps": 500,
-  "global_step": 832,
+  "global_step": 1920,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.01201923076923077,
-      "grad_norm": 4.912038326263428,
-      "learning_rate": 5.9523809523809525e-06,
-      "loss": 4.238,
+      "epoch": 0.005208333333333333,
+      "grad_norm": 3.7305831909179688,
+      "learning_rate": 2.604166666666667e-06,
+      "loss": 4.2451,
       "step": 10
     },
     {
-      "epoch": 0.016826923076923076,
-      "eval_accuracy": 0.018018018018018018,
-      "eval_loss": 4.223487377166748,
-      "eval_runtime": 70.4378,
-      "eval_samples_per_second": 6.303,
-      "eval_steps_per_second": 0.099,
-      "step": 14
+      "epoch": 0.008333333333333333,
+      "eval_accuracy": 0.01065340909090909,
+      "eval_loss": 4.228378772735596,
+      "eval_runtime": 133.9626,
+      "eval_samples_per_second": 10.51,
+      "eval_steps_per_second": 0.164,
+      "step": 16
     },
     {
-      "epoch": 1.0072115384615385,
-      "grad_norm": 5.0958333015441895,
-      "learning_rate": 1.1904761904761905e-05,
-      "loss": 4.2658,
+      "epoch": 1.0020833333333334,
+      "grad_norm": 3.0969808101654053,
+      "learning_rate": 5.208333333333334e-06,
+      "loss": 4.231,
       "step": 20
     },
     {
-      "epoch": 1.0168269230769231,
-      "eval_accuracy": 0.02027027027027027,
-      "eval_loss": 4.207489490509033,
-      "eval_runtime": 72.9674,
-      "eval_samples_per_second": 6.085,
-      "eval_steps_per_second": 0.096,
-      "step": 28
+      "epoch": 1.0072916666666667,
+      "grad_norm": 4.606812000274658,
+      "learning_rate": 7.8125e-06,
+      "loss": 4.2251,
+      "step": 30
     },
     {
-      "epoch": 2.0024038461538463,
-      "grad_norm": 3.8229589462280273,
-      "learning_rate": 1.785714285714286e-05,
-      "loss": 4.2236,
-      "step": 30
+      "epoch": 1.0083333333333333,
+      "eval_accuracy": 0.01065340909090909,
+      "eval_loss": 4.215222358703613,
+      "eval_runtime": 129.5768,
+      "eval_samples_per_second": 10.866,
+      "eval_steps_per_second": 0.17,
+      "step": 32
     },
     {
-      "epoch": 2.014423076923077,
-      "grad_norm": 3.7527875900268555,
-      "learning_rate": 2.380952380952381e-05,
-      "loss": 4.219,
+      "epoch": 2.004166666666667,
+      "grad_norm": 3.574765682220459,
+      "learning_rate": 1.0416666666666668e-05,
+      "loss": 4.2276,
       "step": 40
     },
     {
-      "epoch": 2.016826923076923,
-      "eval_accuracy": 0.015765765765765764,
-      "eval_loss": 4.205322265625,
-      "eval_runtime": 70.576,
-      "eval_samples_per_second": 6.291,
-      "eval_steps_per_second": 0.099,
-      "step": 42
+      "epoch": 2.0083333333333333,
+      "eval_accuracy": 0.012073863636363636,
+      "eval_loss": 4.209568500518799,
+      "eval_runtime": 130.2735,
+      "eval_samples_per_second": 10.808,
+      "eval_steps_per_second": 0.169,
+      "step": 48
     },
     {
-      "epoch": 3.0096153846153846,
-      "grad_norm": 2.5568084716796875,
-      "learning_rate": 2.9761904761904762e-05,
-      "loss": 4.2146,
+      "epoch": 3.0010416666666666,
+      "grad_norm": 4.630516052246094,
+      "learning_rate": 1.3020833333333334e-05,
+      "loss": 4.2126,
       "step": 50
     },
     {
-      "epoch": 3.016826923076923,
-      "eval_accuracy": 0.02252252252252252,
-      "eval_loss": 4.209786415100098,
-      "eval_runtime": 73.3716,
-      "eval_samples_per_second": 6.051,
-      "eval_steps_per_second": 0.095,
-      "step": 56
-    },
-    {
-      "epoch": 4.0048076923076925,
-      "grad_norm": 2.623307943344116,
-      "learning_rate": 3.571428571428572e-05,
-      "loss": 4.2083,
+      "epoch": 3.00625,
+      "grad_norm": 3.0715420246124268,
+      "learning_rate": 1.5625e-05,
+      "loss": 4.2146,
       "step": 60
     },
     {
-      "epoch": 4.016826923076923,
-      "grad_norm": 11.633626937866211,
-      "learning_rate": 4.166666666666667e-05,
-      "loss": 4.1925,
-      "step": 70
+      "epoch": 3.0083333333333333,
+      "eval_accuracy": 0.014914772727272728,
+      "eval_loss": 4.212440490722656,
+      "eval_runtime": 128.256,
+      "eval_samples_per_second": 10.978,
+      "eval_steps_per_second": 0.172,
+      "step": 64
     },
     {
-      "epoch": 4.016826923076923,
-      "eval_accuracy": 0.02252252252252252,
-      "eval_loss": 4.182406902313232,
-      "eval_runtime": 73.5871,
-      "eval_samples_per_second": 6.034,
-      "eval_steps_per_second": 0.095,
+      "epoch": 4.003125,
+      "grad_norm": 2.934682846069336,
+      "learning_rate": 1.8229166666666668e-05,
+      "loss": 4.2235,
       "step": 70
     },
     {
-      "epoch": 5.012019230769231,
-      "grad_norm": 4.2888689041137695,
-      "learning_rate": 4.761904761904762e-05,
-      "loss": 4.1192,
+      "epoch": 4.008333333333334,
+      "grad_norm": 8.70511245727539,
+      "learning_rate": 2.0833333333333336e-05,
+      "loss": 4.2217,
       "step": 80
     },
     {
-      "epoch": 5.016826923076923,
-      "eval_accuracy": 0.038288288288288286,
-      "eval_loss": 4.081945896148682,
-      "eval_runtime": 77.2884,
-      "eval_samples_per_second": 5.745,
-      "eval_steps_per_second": 0.091,
-      "step": 84
+      "epoch": 4.008333333333334,
+      "eval_accuracy": 0.01775568181818182,
+      "eval_loss": 4.2041802406311035,
+      "eval_runtime": 126.2921,
+      "eval_samples_per_second": 11.149,
+      "eval_steps_per_second": 0.174,
+      "step": 80
     },
     {
-      "epoch": 6.007211538461538,
-      "grad_norm": 3.5548884868621826,
-      "learning_rate": 4.959893048128342e-05,
-      "loss": 4.0297,
+      "epoch": 5.005208333333333,
+      "grad_norm": 2.5297322273254395,
+      "learning_rate": 2.34375e-05,
+      "loss": 4.2091,
       "step": 90
     },
     {
-      "epoch": 6.016826923076923,
-      "eval_accuracy": 0.06981981981981981,
-      "eval_loss": 3.981917142868042,
-      "eval_runtime": 72.62,
-      "eval_samples_per_second": 6.114,
-      "eval_steps_per_second": 0.096,
-      "step": 98
+      "epoch": 5.008333333333334,
+      "eval_accuracy": 0.0234375,
+      "eval_loss": 4.205115795135498,
+      "eval_runtime": 127.2102,
+      "eval_samples_per_second": 11.068,
+      "eval_steps_per_second": 0.173,
+      "step": 96
     },
     {
-      "epoch": 7.002403846153846,
-      "grad_norm": 6.507851600646973,
-      "learning_rate": 4.8930481283422465e-05,
-      "loss": 3.9499,
+      "epoch": 6.002083333333333,
+      "grad_norm": 2.4549720287323,
+      "learning_rate": 2.604166666666667e-05,
+      "loss": 4.2256,
       "step": 100
     },
     {
-      "epoch": 7.014423076923077,
-      "grad_norm": 5.1739959716796875,
-      "learning_rate": 4.8262032085561496e-05,
-      "loss": 3.7134,
+      "epoch": 6.007291666666666,
+      "grad_norm": 2.3255746364593506,
+      "learning_rate": 2.8645833333333333e-05,
+      "loss": 4.2085,
       "step": 110
     },
     {
-      "epoch": 7.016826923076923,
-      "eval_accuracy": 0.1036036036036036,
-      "eval_loss": 3.7339625358581543,
-      "eval_runtime": 73.0624,
-      "eval_samples_per_second": 6.077,
-      "eval_steps_per_second": 0.096,
+      "epoch": 6.008333333333334,
+      "eval_accuracy": 0.018465909090909092,
+      "eval_loss": 4.193885803222656,
+      "eval_runtime": 130.9427,
+      "eval_samples_per_second": 10.753,
+      "eval_steps_per_second": 0.168,
       "step": 112
     },
     {
-      "epoch": 8.009615384615385,
-      "grad_norm": 4.7554779052734375,
-      "learning_rate": 4.759358288770054e-05,
-      "loss": 3.5289,
+      "epoch": 7.004166666666666,
+      "grad_norm": 2.676177978515625,
+      "learning_rate": 3.125e-05,
+      "loss": 4.2044,
       "step": 120
     },
     {
-      "epoch": 8.016826923076923,
-      "eval_accuracy": 0.18018018018018017,
-      "eval_loss": 3.420488119125366,
-      "eval_runtime": 73.5384,
-      "eval_samples_per_second": 6.038,
-      "eval_steps_per_second": 0.095,
-      "step": 126
+      "epoch": 7.008333333333334,
+      "eval_accuracy": 0.0390625,
+      "eval_loss": 4.1791582107543945,
+      "eval_runtime": 125.2796,
+      "eval_samples_per_second": 11.239,
+      "eval_steps_per_second": 0.176,
+      "step": 128
     },
     {
-      "epoch": 9.004807692307692,
-      "grad_norm": 6.714512348175049,
-      "learning_rate": 4.6925133689839576e-05,
-      "loss": 3.3709,
+      "epoch": 8.001041666666667,
+      "grad_norm": 2.616762161254883,
+      "learning_rate": 3.385416666666667e-05,
+      "loss": 4.1911,
       "step": 130
     },
     {
-      "epoch": 9.016826923076923,
-      "grad_norm": 18.599552154541016,
-      "learning_rate": 4.625668449197861e-05,
-      "loss": 3.0625,
+      "epoch": 8.00625,
+      "grad_norm": 2.983841896057129,
+      "learning_rate": 3.6458333333333336e-05,
+      "loss": 4.1624,
       "step": 140
     },
     {
-      "epoch": 9.016826923076923,
-      "eval_accuracy": 0.26126126126126126,
-      "eval_loss": 3.196960687637329,
-      "eval_runtime": 74.7592,
-      "eval_samples_per_second": 5.939,
-      "eval_steps_per_second": 0.094,
-      "step": 140
+      "epoch": 8.008333333333333,
+      "eval_accuracy": 0.022017045454545456,
+      "eval_loss": 4.201480388641357,
+      "eval_runtime": 136.9987,
+      "eval_samples_per_second": 10.277,
+      "eval_steps_per_second": 0.161,
+      "step": 144
     },
     {
-      "epoch": 10.01201923076923,
-      "grad_norm": 7.006440162658691,
-      "learning_rate": 4.558823529411765e-05,
-      "loss": 2.8776,
+      "epoch": 9.003125,
+      "grad_norm": 2.874126434326172,
+      "learning_rate": 3.90625e-05,
+      "loss": 4.1349,
       "step": 150
     },
     {
-      "epoch": 10.016826923076923,
-      "eval_accuracy": 0.30180180180180183,
-      "eval_loss": 2.9350106716156006,
-      "eval_runtime": 75.4948,
-      "eval_samples_per_second": 5.881,
-      "eval_steps_per_second": 0.093,
-      "step": 154
-    },
-    {
-      "epoch": 11.007211538461538,
-      "grad_norm": 6.219061374664307,
-      "learning_rate": 4.491978609625669e-05,
-      "loss": 2.6375,
+      "epoch": 9.008333333333333,
+      "grad_norm": 10.345636367797852,
+      "learning_rate": 4.166666666666667e-05,
+      "loss": 4.1253,
       "step": 160
     },
     {
-      "epoch": 11.016826923076923,
-      "eval_accuracy": 0.36036036036036034,
-      "eval_loss": 2.762986660003662,
-      "eval_runtime": 73.4162,
-      "eval_samples_per_second": 6.048,
-      "eval_steps_per_second": 0.095,
-      "step": 168
+      "epoch": 9.008333333333333,
+      "eval_accuracy": 0.029829545454545456,
+      "eval_loss": 4.1215105056762695,
+      "eval_runtime": 132.9763,
+      "eval_samples_per_second": 10.588,
+      "eval_steps_per_second": 0.165,
+      "step": 160
     },
     {
-      "epoch": 12.002403846153847,
-      "grad_norm": 6.574148178100586,
-      "learning_rate": 4.4251336898395724e-05,
-      "loss": 2.4071,
+      "epoch": 10.005208333333334,
+      "grad_norm": 3.5383424758911133,
+      "learning_rate": 4.4270833333333337e-05,
+      "loss": 4.0308,
       "step": 170
     },
     {
-      "epoch": 12.014423076923077,
-      "grad_norm": 8.191123008728027,
-      "learning_rate": 4.358288770053476e-05,
-      "loss": 2.2954,
-      "step": 180
+      "epoch": 10.008333333333333,
+      "eval_accuracy": 0.07102272727272728,
+      "eval_loss": 4.002456188201904,
+      "eval_runtime": 127.9472,
+      "eval_samples_per_second": 11.005,
+      "eval_steps_per_second": 0.172,
+      "step": 176
     },
     {
-      "epoch": 12.016826923076923,
-      "eval_accuracy": 0.42792792792792794,
-      "eval_loss": 2.4990594387054443,
-      "eval_runtime": 80.1492,
-      "eval_samples_per_second": 5.54,
-      "eval_steps_per_second": 0.087,
-      "step": 182
+      "epoch": 11.002083333333333,
+      "grad_norm": 4.1187543869018555,
+      "learning_rate": 4.6875e-05,
+      "loss": 3.9438,
+      "step": 180
     },
     {
-      "epoch": 13.009615384615385,
-      "grad_norm": 6.4414825439453125,
-      "learning_rate": 4.29144385026738e-05,
-      "loss": 2.1337,
+      "epoch": 11.007291666666667,
+      "grad_norm": 4.370882034301758,
+      "learning_rate": 4.947916666666667e-05,
+      "loss": 3.8065,
       "step": 190
     },
     {
-      "epoch": 13.016826923076923,
-      "eval_accuracy": 0.43243243243243246,
-      "eval_loss": 2.3827102184295654,
-      "eval_runtime": 73.7035,
-      "eval_samples_per_second": 6.024,
-      "eval_steps_per_second": 0.095,
-      "step": 196
+      "epoch": 11.008333333333333,
+      "eval_accuracy": 0.08380681818181818,
+      "eval_loss": 3.872328758239746,
+      "eval_runtime": 142.2129,
+      "eval_samples_per_second": 9.901,
+      "eval_steps_per_second": 0.155,
+      "step": 192
     },
     {
-      "epoch": 14.004807692307692,
-      "grad_norm": 8.87990951538086,
-      "learning_rate": 4.224598930481284e-05,
-      "loss": 1.9319,
+      "epoch": 12.004166666666666,
+      "grad_norm": 3.4101152420043945,
+      "learning_rate": 4.976851851851852e-05,
+      "loss": 3.7614,
       "step": 200
     },
     {
-      "epoch": 14.016826923076923,
-      "grad_norm": 22.130002975463867,
-      "learning_rate": 4.157754010695187e-05,
-      "loss": 1.8195,
-      "step": 210
+      "epoch": 12.008333333333333,
+      "eval_accuracy": 0.09943181818181818,
+      "eval_loss": 3.7138783931732178,
+      "eval_runtime": 125.568,
+      "eval_samples_per_second": 11.213,
+      "eval_steps_per_second": 0.175,
+      "step": 208
     },
     {
-      "epoch": 14.016826923076923,
-      "eval_accuracy": 0.481981981981982,
-      "eval_loss": 2.2311551570892334,
-      "eval_runtime": 75.5729,
-      "eval_samples_per_second": 5.875,
-      "eval_steps_per_second": 0.093,
+      "epoch": 13.001041666666667,
+      "grad_norm": 4.585361480712891,
+      "learning_rate": 4.947916666666667e-05,
+      "loss": 3.6302,
       "step": 210
     },
     {
-      "epoch": 15.01201923076923,
-      "grad_norm": 7.134392261505127,
-      "learning_rate": 4.0909090909090915e-05,
-      "loss": 1.6436,
+      "epoch": 13.00625,
+      "grad_norm": 4.209935665130615,
+      "learning_rate": 4.9189814814814815e-05,
+      "loss": 3.4761,
       "step": 220
     },
     {
-      "epoch": 15.016826923076923,
-      "eval_accuracy": 0.44594594594594594,
-      "eval_loss": 2.3628787994384766,
-      "eval_runtime": 76.2152,
-      "eval_samples_per_second": 5.826,
-      "eval_steps_per_second": 0.092,
+      "epoch": 13.008333333333333,
+      "eval_accuracy": 0.1434659090909091,
+      "eval_loss": 3.6160459518432617,
+      "eval_runtime": 133.4296,
+      "eval_samples_per_second": 10.552,
+      "eval_steps_per_second": 0.165,
       "step": 224
     },
     {
-      "epoch": 16.00721153846154,
-      "grad_norm": 11.737651824951172,
-      "learning_rate": 4.024064171122995e-05,
-      "loss": 1.6289,
+      "epoch": 14.003125,
+      "grad_norm": 4.6234049797058105,
+      "learning_rate": 4.8900462962962965e-05,
+      "loss": 3.3764,
       "step": 230
     },
     {
-      "epoch": 16.016826923076923,
-      "eval_accuracy": 0.4752252252252252,
-      "eval_loss": 2.151165723800659,
-      "eval_runtime": 72.7861,
-      "eval_samples_per_second": 6.1,
-      "eval_steps_per_second": 0.096,
-      "step": 238
+      "epoch": 14.008333333333333,
+      "grad_norm": 13.464088439941406,
+      "learning_rate": 4.8611111111111115e-05,
+      "loss": 3.278,
+      "step": 240
     },
     {
-      "epoch": 17.002403846153847,
-      "grad_norm": 12.988323211669922,
-      "learning_rate": 3.957219251336899e-05,
-      "loss": 1.423,
+      "epoch": 14.008333333333333,
+      "eval_accuracy": 0.1924715909090909,
+      "eval_loss": 3.3939576148986816,
+      "eval_runtime": 129.3255,
+      "eval_samples_per_second": 10.887,
+      "eval_steps_per_second": 0.17,
       "step": 240
     },
     {
-      "epoch": 17.014423076923077,
-      "grad_norm": 5.394784450531006,
-      "learning_rate": 3.8903743315508025e-05,
-      "loss": 1.2957,
+      "epoch": 15.005208333333334,
+      "grad_norm": 4.547211647033691,
+      "learning_rate": 4.8321759259259265e-05,
+      "loss": 3.0999,
       "step": 250
     },
     {
-      "epoch": 17.016826923076923,
-      "eval_accuracy": 0.5022522522522522,
-      "eval_loss": 2.0142312049865723,
-      "eval_runtime": 74.6687,
-      "eval_samples_per_second": 5.946,
-      "eval_steps_per_second": 0.094,
-      "step": 252
+      "epoch": 15.008333333333333,
+      "eval_accuracy": 0.2080965909090909,
+      "eval_loss": 3.3182637691497803,
+      "eval_runtime": 123.805,
+      "eval_samples_per_second": 11.373,
+      "eval_steps_per_second": 0.178,
+      "step": 256
     },
     {
-      "epoch": 18.009615384615383,
-      "grad_norm": 6.641128063201904,
-      "learning_rate": 3.8235294117647055e-05,
-      "loss": 1.2761,
+      "epoch": 16.002083333333335,
+      "grad_norm": 4.825346946716309,
+      "learning_rate": 4.803240740740741e-05,
+      "loss": 3.0213,
       "step": 260
     },
     {
-      "epoch": 18.016826923076923,
-      "eval_accuracy": 0.527027027027027,
-      "eval_loss": 1.906096339225769,
-      "eval_runtime": 75.2853,
-      "eval_samples_per_second": 5.898,
-      "eval_steps_per_second": 0.093,
-      "step": 266
-    },
-    {
-      "epoch": 19.004807692307693,
-      "grad_norm": 7.859009742736816,
-      "learning_rate": 3.75668449197861e-05,
-      "loss": 1.217,
+      "epoch": 16.007291666666667,
+      "grad_norm": 5.93734884262085,
+      "learning_rate": 4.774305555555556e-05,
+      "loss": 2.9721,
       "step": 270
     },
     {
-      "epoch": 19.016826923076923,
-      "grad_norm": 29.06666374206543,
-      "learning_rate": 3.6898395721925136e-05,
-      "loss": 1.1118,
-      "step": 280
+      "epoch": 16.008333333333333,
+      "eval_accuracy": 0.24857954545454544,
+      "eval_loss": 3.159555673599243,
+      "eval_runtime": 124.115,
+      "eval_samples_per_second": 11.344,
+      "eval_steps_per_second": 0.177,
+      "step": 272
     },
     {
-      "epoch": 19.016826923076923,
-      "eval_accuracy": 0.5495495495495496,
-      "eval_loss": 1.811281681060791,
-      "eval_runtime": 81.5227,
-      "eval_samples_per_second": 5.446,
-      "eval_steps_per_second": 0.086,
+      "epoch": 17.004166666666666,
+      "grad_norm": 4.84201717376709,
+      "learning_rate": 4.745370370370371e-05,
+      "loss": 2.8064,
       "step": 280
     },
     {
-      "epoch": 20.01201923076923,
-      "grad_norm": 8.949642181396484,
-      "learning_rate": 3.622994652406417e-05,
-      "loss": 0.9642,
-      "step": 290
+      "epoch": 17.008333333333333,
+      "eval_accuracy": 0.26704545454545453,
+      "eval_loss": 3.023179054260254,
+      "eval_runtime": 131.2545,
+      "eval_samples_per_second": 10.727,
+      "eval_steps_per_second": 0.168,
+      "step": 288
     },
     {
-      "epoch": 20.016826923076923,
-      "eval_accuracy": 0.6036036036036037,
-      "eval_loss": 1.727989673614502,
-      "eval_runtime": 72.1469,
-      "eval_samples_per_second": 6.154,
-      "eval_steps_per_second": 0.097,
-      "step": 294
+      "epoch": 18.001041666666666,
+      "grad_norm": 5.273012161254883,
+      "learning_rate": 4.716435185185186e-05,
+      "loss": 2.813,
+      "step": 290
     },
     {
-      "epoch": 21.00721153846154,
-      "grad_norm": 6.73243522644043,
-      "learning_rate": 3.556149732620321e-05,
-      "loss": 0.894,
+      "epoch": 18.00625,
+      "grad_norm": 5.6201276779174805,
+      "learning_rate": 4.6875e-05,
+      "loss": 2.6554,
       "step": 300
     },
     {
-      "epoch": 21.016826923076923,
-      "eval_accuracy": 0.5382882882882883,
-      "eval_loss": 1.8722944259643555,
-      "eval_runtime": 76.1286,
-      "eval_samples_per_second": 5.832,
-      "eval_steps_per_second": 0.092,
-      "step": 308
+      "epoch": 18.008333333333333,
+      "eval_accuracy": 0.2911931818181818,
+      "eval_loss": 2.9448060989379883,
+      "eval_runtime": 123.4251,
+      "eval_samples_per_second": 11.408,
+      "eval_steps_per_second": 0.178,
+      "step": 304
     },
     {
-      "epoch": 22.002403846153847,
-      "grad_norm": 8.856271743774414,
-      "learning_rate": 3.489304812834225e-05,
-      "loss": 0.8454,
+      "epoch": 19.003125,
+      "grad_norm": 7.461442470550537,
+      "learning_rate": 4.658564814814815e-05,
+      "loss": 2.6668,
       "step": 310
     },
     {
-      "epoch": 22.014423076923077,
-      "grad_norm": 6.268283843994141,
-      "learning_rate": 3.4224598930481284e-05,
-      "loss": 0.7974,
+      "epoch": 19.008333333333333,
+      "grad_norm": 14.555140495300293,
+      "learning_rate": 4.62962962962963e-05,
+      "loss": 2.5052,
       "step": 320
     },
     {
-      "epoch": 22.016826923076923,
-      "eval_accuracy": 0.6058558558558559,
-      "eval_loss": 1.658478021621704,
-      "eval_runtime": 69.8498,
-      "eval_samples_per_second": 6.356,
-      "eval_steps_per_second": 0.1,
-      "step": 322
+      "epoch": 19.008333333333333,
+      "eval_accuracy": 0.3309659090909091,
+      "eval_loss": 2.828484535217285,
+      "eval_runtime": 127.0838,
+      "eval_samples_per_second": 11.079,
+      "eval_steps_per_second": 0.173,
+      "step": 320
     },
     {
-      "epoch": 23.009615384615383,
-      "grad_norm": 6.027498722076416,
-      "learning_rate": 3.355614973262032e-05,
-      "loss": 0.833,
+      "epoch": 20.005208333333332,
+      "grad_norm": 6.753479957580566,
+      "learning_rate": 4.6006944444444444e-05,
+      "loss": 2.4322,
       "step": 330
     },
     {
-      "epoch": 23.016826923076923,
-      "eval_accuracy": 0.6148648648648649,
-      "eval_loss": 1.599608302116394,
-      "eval_runtime": 76.7983,
-      "eval_samples_per_second": 5.781,
-      "eval_steps_per_second": 0.091,
+      "epoch": 20.008333333333333,
+      "eval_accuracy": 0.34019886363636365,
+      "eval_loss": 2.7478535175323486,
+      "eval_runtime": 125.7771,
+      "eval_samples_per_second": 11.194,
+      "eval_steps_per_second": 0.175,
       "step": 336
     },
     {
-      "epoch": 24.004807692307693,
-      "grad_norm": 5.669070243835449,
-      "learning_rate": 3.288770053475936e-05,
-      "loss": 0.6668,
+      "epoch": 21.002083333333335,
+      "grad_norm": 5.665611743927002,
+      "learning_rate": 4.5717592592592594e-05,
+      "loss": 2.4376,
       "step": 340
     },
     {
-      "epoch": 24.016826923076923,
-      "grad_norm": 12.669683456420898,
-      "learning_rate": 3.22192513368984e-05,
-      "loss": 0.6431,
+      "epoch": 21.007291666666667,
+      "grad_norm": 5.6526570320129395,
+      "learning_rate": 4.5428240740740744e-05,
+      "loss": 2.3193,
       "step": 350
     },
     {
-      "epoch": 24.016826923076923,
-      "eval_accuracy": 0.6148648648648649,
-      "eval_loss": 1.560927152633667,
-      "eval_runtime": 69.5019,
-      "eval_samples_per_second": 6.388,
-      "eval_steps_per_second": 0.101,
-      "step": 350
+      "epoch": 21.008333333333333,
+      "eval_accuracy": 0.3309659090909091,
+      "eval_loss": 2.794116735458374,
+      "eval_runtime": 126.5702,
+      "eval_samples_per_second": 11.124,
+      "eval_steps_per_second": 0.174,
+      "step": 352
     },
     {
-      "epoch": 25.01201923076923,
-      "grad_norm": 5.4236626625061035,
-      "learning_rate": 3.155080213903743e-05,
-      "loss": 0.5873,
+      "epoch": 22.004166666666666,
+      "grad_norm": 6.312890529632568,
+      "learning_rate": 4.5138888888888894e-05,
+      "loss": 2.2565,
       "step": 360
     },
     {
-      "epoch": 25.016826923076923,
-      "eval_accuracy": 0.6171171171171171,
-      "eval_loss": 1.6108297109603882,
-      "eval_runtime": 73.8756,
-      "eval_samples_per_second": 6.01,
-      "eval_steps_per_second": 0.095,
-      "step": 364
+      "epoch": 22.008333333333333,
+      "eval_accuracy": 0.3671875,
+      "eval_loss": 2.6383402347564697,
+      "eval_runtime": 123.0345,
+      "eval_samples_per_second": 11.444,
+      "eval_steps_per_second": 0.179,
+      "step": 368
     },
     {
-      "epoch": 26.00721153846154,
-      "grad_norm": 5.605001449584961,
-      "learning_rate": 3.0882352941176475e-05,
-      "loss": 0.5554,
+      "epoch": 23.001041666666666,
+      "grad_norm": 5.865190505981445,
+      "learning_rate": 4.484953703703704e-05,
+      "loss": 2.1257,
       "step": 370
     },
     {
-      "epoch": 26.016826923076923,
-      "eval_accuracy": 0.6531531531531531,
-      "eval_loss": 1.4013562202453613,
-      "eval_runtime": 79.7998,
-      "eval_samples_per_second": 5.564,
-      "eval_steps_per_second": 0.088,
-      "step": 378
-    },
-    {
-      "epoch": 27.002403846153847,
-      "grad_norm": 6.0180277824401855,
-      "learning_rate": 3.0213903743315508e-05,
-      "loss": 0.5428,
+      "epoch": 23.00625,
+      "grad_norm": 5.813269138336182,
+      "learning_rate": 4.456018518518519e-05,
+      "loss": 2.1405,
       "step": 380
     },
     {
-      "epoch": 27.014423076923077,
-      "grad_norm": 7.198999404907227,
-      "learning_rate": 2.954545454545455e-05,
-      "loss": 0.4786,
-      "step": 390
+      "epoch": 23.008333333333333,
+      "eval_accuracy": 0.36079545454545453,
+      "eval_loss": 2.5905861854553223,
+      "eval_runtime": 126.2698,
+      "eval_samples_per_second": 11.151,
+      "eval_steps_per_second": 0.174,
+      "step": 384
     },
     {
-      "epoch": 27.016826923076923,
-      "eval_accuracy": 0.6621621621621622,
-      "eval_loss": 1.433493733406067,
-      "eval_runtime": 71.2505,
-      "eval_samples_per_second": 6.232,
-      "eval_steps_per_second": 0.098,
-      "step": 392
+      "epoch": 24.003125,
+      "grad_norm": 6.6013569831848145,
+      "learning_rate": 4.4270833333333337e-05,
+      "loss": 1.9715,
+      "step": 390
     },
     {
-      "epoch": 28.009615384615383,
-      "grad_norm": 4.281993389129639,
-      "learning_rate": 2.8877005347593582e-05,
-      "loss": 0.4252,
+      "epoch": 24.008333333333333,
+      "grad_norm": 14.786352157592773,
+      "learning_rate": 4.3981481481481486e-05,
+      "loss": 2.1049,
       "step": 400
     },
     {
-      "epoch": 28.016826923076923,
-      "eval_accuracy": 0.6509009009009009,
-      "eval_loss": 1.444474220275879,
-      "eval_runtime": 71.5031,
-      "eval_samples_per_second": 6.21,
-      "eval_steps_per_second": 0.098,
-      "step": 406
+      "epoch": 24.008333333333333,
+      "eval_accuracy": 0.37855113636363635,
+      "eval_loss": 2.5515265464782715,
+      "eval_runtime": 134.242,
+      "eval_samples_per_second": 10.489,
+      "eval_steps_per_second": 0.164,
+      "step": 400
     },
     {
-      "epoch": 29.004807692307693,
-      "grad_norm": 6.2630534172058105,
-      "learning_rate": 2.8208556149732622e-05,
-      "loss": 0.4527,
+      "epoch": 25.005208333333332,
+      "grad_norm": 6.399519920349121,
+      "learning_rate": 4.369212962962963e-05,
+      "loss": 1.8424,
       "step": 410
     },
     {
-      "epoch": 29.016826923076923,
-      "grad_norm": 4.163498878479004,
-      "learning_rate": 2.754010695187166e-05,
-      "loss": 0.382,
-      "step": 420
+      "epoch": 25.008333333333333,
+      "eval_accuracy": 0.39204545454545453,
+      "eval_loss": 2.4692277908325195,
+      "eval_runtime": 134.5245,
+      "eval_samples_per_second": 10.466,
+      "eval_steps_per_second": 0.164,
+      "step": 416
     },
     {
-      "epoch": 29.016826923076923,
-      "eval_accuracy": 0.6621621621621622,
-      "eval_loss": 1.3914759159088135,
-      "eval_runtime": 72.0609,
-      "eval_samples_per_second": 6.161,
-      "eval_steps_per_second": 0.097,
+      "epoch": 26.002083333333335,
+      "grad_norm": 5.429991245269775,
+      "learning_rate": 4.340277777777778e-05,
+      "loss": 2.0616,
       "step": 420
     },
     {
-      "epoch": 30.01201923076923,
-      "grad_norm": 5.012676239013672,
-      "learning_rate": 2.68716577540107e-05,
-      "loss": 0.365,
+      "epoch": 26.007291666666667,
+      "grad_norm": 6.501383304595947,
+      "learning_rate": 4.311342592592593e-05,
+      "loss": 1.8685,
       "step": 430
     },
     {
-      "epoch": 30.016826923076923,
-      "eval_accuracy": 0.6846846846846847,
-      "eval_loss": 1.297812819480896,
-      "eval_runtime": 73.674,
-      "eval_samples_per_second": 6.027,
-      "eval_steps_per_second": 0.095,
-      "step": 434
+      "epoch": 26.008333333333333,
+      "eval_accuracy": 0.4275568181818182,
+      "eval_loss": 2.4325406551361084,
+      "eval_runtime": 138.3084,
+      "eval_samples_per_second": 10.18,
+      "eval_steps_per_second": 0.159,
+      "step": 432
     },
     {
-      "epoch": 31.00721153846154,
-      "grad_norm": 4.618253707885742,
-      "learning_rate": 2.6203208556149733e-05,
-      "loss": 0.319,
+      "epoch": 27.004166666666666,
+      "grad_norm": 6.909026622772217,
+      "learning_rate": 4.282407407407408e-05,
+      "loss": 1.7478,
       "step": 440
     },
     {
-      "epoch": 31.016826923076923,
-      "eval_accuracy": 0.6824324324324325,
-      "eval_loss": 1.3218427896499634,
-      "eval_runtime": 74.4745,
-      "eval_samples_per_second": 5.962,
-      "eval_steps_per_second": 0.094,
+      "epoch": 27.008333333333333,
+      "eval_accuracy": 0.4147727272727273,
+      "eval_loss": 2.416501522064209,
+      "eval_runtime": 128.9504,
+      "eval_samples_per_second": 10.919,
+      "eval_steps_per_second": 0.171,
       "step": 448
     },
     {
-      "epoch": 32.00240384615385,
-      "grad_norm": 6.823300361633301,
-      "learning_rate": 2.5534759358288773e-05,
-      "loss": 0.3829,
+      "epoch": 28.001041666666666,
+      "grad_norm": 6.359575271606445,
+      "learning_rate": 4.253472222222222e-05,
+      "loss": 2.0103,
       "step": 450
     },
     {
-      "epoch": 32.01442307692308,
-      "grad_norm": 5.096348285675049,
-      "learning_rate": 2.4866310160427807e-05,
-      "loss": 0.3167,
+      "epoch": 28.00625,
+      "grad_norm": 7.018951416015625,
+      "learning_rate": 4.224537037037037e-05,
+      "loss": 1.7072,
       "step": 460
     },
     {
-      "epoch": 32.01682692307692,
-      "eval_accuracy": 0.6644144144144144,
-      "eval_loss": 1.3495796918869019,
-      "eval_runtime": 73.0271,
-      "eval_samples_per_second": 6.08,
-      "eval_steps_per_second": 0.096,
-      "step": 462
+      "epoch": 28.008333333333333,
+      "eval_accuracy": 0.4268465909090909,
+      "eval_loss": 2.3617048263549805,
+      "eval_runtime": 126.6826,
+      "eval_samples_per_second": 11.114,
+      "eval_steps_per_second": 0.174,
+      "step": 464
     },
     {
-      "epoch": 33.00961538461539,
-      "grad_norm": 6.042089462280273,
-      "learning_rate": 2.4197860962566847e-05,
-      "loss": 0.2797,
+      "epoch": 29.003125,
+      "grad_norm": 5.399332523345947,
+      "learning_rate": 4.195601851851852e-05,
+      "loss": 1.8236,
       "step": 470
     },
     {
-      "epoch": 33.01682692307692,
-      "eval_accuracy": 0.6801801801801802,
-      "eval_loss": 1.2806001901626587,
-      "eval_runtime": 72.9489,
-      "eval_samples_per_second": 6.086,
-      "eval_steps_per_second": 0.096,
-      "step": 476
+      "epoch": 29.008333333333333,
+      "grad_norm": 14.481021881103516,
+      "learning_rate": 4.166666666666667e-05,
+      "loss": 1.7206,
+      "step": 480
     },
     {
-      "epoch": 34.00480769230769,
-      "grad_norm": 4.125849723815918,
-      "learning_rate": 2.3529411764705884e-05,
-      "loss": 0.281,
+      "epoch": 29.008333333333333,
+      "eval_accuracy": 0.4303977272727273,
+      "eval_loss": 2.372326135635376,
+      "eval_runtime": 128.5457,
+      "eval_samples_per_second": 10.953,
+      "eval_steps_per_second": 0.171,
       "step": 480
     },
     {
-      "epoch": 34.01682692307692,
-      "grad_norm": 7.1154704093933105,
-      "learning_rate": 2.286096256684492e-05,
-      "loss": 0.2864,
+      "epoch": 30.005208333333332,
+      "grad_norm": 7.11575984954834,
+      "learning_rate": 4.1377314814814815e-05,
+      "loss": 1.693,
       "step": 490
     },
     {
-      "epoch": 34.01682692307692,
-      "eval_accuracy": 0.7072072072072072,
-      "eval_loss": 1.219117283821106,
-      "eval_runtime": 74.6962,
-      "eval_samples_per_second": 5.944,
-      "eval_steps_per_second": 0.094,
-      "step": 490
+      "epoch": 30.008333333333333,
+      "eval_accuracy": 0.4424715909090909,
+      "eval_loss": 2.2890071868896484,
+      "eval_runtime": 128.9784,
+      "eval_samples_per_second": 10.917,
+      "eval_steps_per_second": 0.171,
+      "step": 496
     },
     {
-      "epoch": 35.01201923076923,
-      "grad_norm": 6.536057949066162,
-      "learning_rate": 2.2192513368983957e-05,
-      "loss": 0.2927,
+      "epoch": 31.002083333333335,
+      "grad_norm": 7.223865985870361,
+      "learning_rate": 4.1087962962962965e-05,
+      "loss": 1.6722,
       "step": 500
     },
     {
-      "epoch": 35.01682692307692,
-      "eval_accuracy": 0.7207207207207207,
-      "eval_loss": 1.2134795188903809,
-      "eval_runtime": 72.1404,
-      "eval_samples_per_second": 6.155,
-      "eval_steps_per_second": 0.097,
-      "step": 504
-    },
-    {
-      "epoch": 36.00721153846154,
-      "grad_norm": 4.703156471252441,
-      "learning_rate": 2.1524064171122994e-05,
-      "loss": 0.2698,
+      "epoch": 31.007291666666667,
+      "grad_norm": 7.608266353607178,
+      "learning_rate": 4.0798611111111115e-05,
+      "loss": 1.6347,
       "step": 510
     },
     {
-      "epoch": 36.01682692307692,
-      "eval_accuracy": 0.6914414414414415,
-      "eval_loss": 1.250654697418213,
-      "eval_runtime": 73.2059,
-      "eval_samples_per_second": 6.065,
-      "eval_steps_per_second": 0.096,
-      "step": 518
+      "epoch": 31.008333333333333,
+      "eval_accuracy": 0.44105113636363635,
+      "eval_loss": 2.244246482849121,
+      "eval_runtime": 129.1752,
+      "eval_samples_per_second": 10.9,
+      "eval_steps_per_second": 0.17,
+      "step": 512
     },
     {
-      "epoch": 37.00240384615385,
-      "grad_norm": 4.395312786102295,
-      "learning_rate": 2.0855614973262035e-05,
-      "loss": 0.256,
+      "epoch": 32.00416666666667,
+      "grad_norm": 7.774643898010254,
+      "learning_rate": 4.0509259259259265e-05,
+      "loss": 1.5276,
       "step": 520
     },
     {
-      "epoch": 37.01442307692308,
-      "grad_norm": 3.43831729888916,
-      "learning_rate": 2.018716577540107e-05,
-      "loss": 0.2333,
-      "step": 530
+      "epoch": 32.00833333333333,
+      "eval_accuracy": 0.46732954545454547,
+      "eval_loss": 2.210369110107422,
+      "eval_runtime": 120.0766,
+      "eval_samples_per_second": 11.726,
+      "eval_steps_per_second": 0.183,
+      "step": 528
     },
     {
-      "epoch": 37.01682692307692,
-      "eval_accuracy": 0.7094594594594594,
-      "eval_loss": 1.2037817239761353,
-      "eval_runtime": 74.757,
-      "eval_samples_per_second": 5.939,
-      "eval_steps_per_second": 0.094,
-      "step": 532
+      "epoch": 33.001041666666666,
+      "grad_norm": 6.766725063323975,
+      "learning_rate": 4.021990740740741e-05,
+      "loss": 1.5203,
+      "step": 530
     },
     {
-      "epoch": 38.00961538461539,
-      "grad_norm": 4.047479152679443,
-      "learning_rate": 1.951871657754011e-05,
-      "loss": 0.2366,
+      "epoch": 33.00625,
+      "grad_norm": 5.995940208435059,
+      "learning_rate": 3.993055555555556e-05,
+      "loss": 1.4576,
       "step": 540
     },
     {
-      "epoch": 38.01682692307692,
-      "eval_accuracy": 0.7207207207207207,
-      "eval_loss": 1.1517395973205566,
-      "eval_runtime": 73.764,
-      "eval_samples_per_second": 6.019,
-      "eval_steps_per_second": 0.095,
-      "step": 546
+      "epoch": 33.00833333333333,
+      "eval_accuracy": 0.45454545454545453,
+      "eval_loss": 2.227916717529297,
+      "eval_runtime": 124.0445,
+      "eval_samples_per_second": 11.351,
+      "eval_steps_per_second": 0.177,
+      "step": 544
     },
     {
-      "epoch": 39.00480769230769,
-      "grad_norm": 5.707859516143799,
-      "learning_rate": 1.8850267379679145e-05,
-      "loss": 0.1938,
+      "epoch": 34.003125,
+      "grad_norm": 6.610184669494629,
+      "learning_rate": 3.964120370370371e-05,
+      "loss": 1.5347,
       "step": 550
     },
     {
-      "epoch": 39.01682692307692,
-      "grad_norm": 3.7590107917785645,
-      "learning_rate": 1.8181818181818182e-05,
-      "loss": 0.1886,
+      "epoch": 34.00833333333333,
+      "grad_norm": 24.694988250732422,
+      "learning_rate": 3.935185185185186e-05,
+      "loss": 1.5455,
       "step": 560
     },
     {
-      "epoch": 39.01682692307692,
-      "eval_accuracy": 0.7094594594594594,
-      "eval_loss": 1.2073884010314941,
-      "eval_runtime": 72.0607,
-      "eval_samples_per_second": 6.161,
-      "eval_steps_per_second": 0.097,
+      "epoch": 34.00833333333333,
+      "eval_accuracy": 0.4524147727272727,
+      "eval_loss": 2.204953908920288,
+      "eval_runtime": 131.2835,
+      "eval_samples_per_second": 10.725,
+      "eval_steps_per_second": 0.168,
       "step": 560
     },
     {
-      "epoch": 40.01201923076923,
-      "grad_norm": 2.7063074111938477,
-      "learning_rate": 1.7513368983957222e-05,
-      "loss": 0.1804,
+      "epoch": 35.005208333333336,
+      "grad_norm": 6.918396472930908,
+      "learning_rate": 3.90625e-05,
+      "loss": 1.4485,
       "step": 570
     },
     {
-      "epoch": 40.01682692307692,
-      "eval_accuracy": 0.7027027027027027,
-      "eval_loss": 1.1658011674880981,
-      "eval_runtime": 77.6928,
-      "eval_samples_per_second": 5.715,
-      "eval_steps_per_second": 0.09,
-      "step": 574
+      "epoch": 35.00833333333333,
+      "eval_accuracy": 0.4737215909090909,
+      "eval_loss": 2.1584527492523193,
+      "eval_runtime": 134.6721,
+      "eval_samples_per_second": 10.455,
+      "eval_steps_per_second": 0.163,
+      "step": 576
     },
     {
-      "epoch": 41.00721153846154,
-      "grad_norm": 6.454692363739014,
-      "learning_rate": 1.684491978609626e-05,
-      "loss": 0.1778,
+      "epoch": 36.00208333333333,
+      "grad_norm": 6.077505588531494,
+      "learning_rate": 3.877314814814815e-05,
+      "loss": 1.4391,
       "step": 580
     },
     {
-      "epoch": 41.01682692307692,
-      "eval_accuracy": 0.6824324324324325,
-      "eval_loss": 1.2350265979766846,
-      "eval_runtime": 73.7916,
-      "eval_samples_per_second": 6.017,
-      "eval_steps_per_second": 0.095,
-      "step": 588
+      "epoch": 36.00729166666667,
+      "grad_norm": 6.617316722869873,
+      "learning_rate": 3.84837962962963e-05,
+      "loss": 1.3896,
+      "step": 590
     },
     {
-      "epoch": 42.00240384615385,
-      "grad_norm": 3.444183349609375,
-      "learning_rate": 1.6176470588235296e-05,
-      "loss": 0.2187,
-      "step": 590
+      "epoch": 36.00833333333333,
+      "eval_accuracy": 0.4446022727272727,
+      "eval_loss": 2.1850500106811523,
+      "eval_runtime": 129.9914,
+      "eval_samples_per_second": 10.831,
+      "eval_steps_per_second": 0.169,
+      "step": 592
     },
     {
-      "epoch": 42.01442307692308,
-      "grad_norm": 3.9001522064208984,
-      "learning_rate": 1.5508021390374333e-05,
-      "loss": 0.1728,
+      "epoch": 37.00416666666667,
+      "grad_norm": 7.900727272033691,
+      "learning_rate": 3.8194444444444444e-05,
+      "loss": 1.3766,
       "step": 600
     },
     {
-      "epoch": 42.01682692307692,
-      "eval_accuracy": 0.7162162162162162,
-      "eval_loss": 1.1637804508209229,
-      "eval_runtime": 77.911,
-      "eval_samples_per_second": 5.699,
-      "eval_steps_per_second": 0.09,
-      "step": 602
+      "epoch": 37.00833333333333,
+      "eval_accuracy": 0.4872159090909091,
+      "eval_loss": 2.118501663208008,
+      "eval_runtime": 131.0307,
+      "eval_samples_per_second": 10.746,
+      "eval_steps_per_second": 0.168,
+      "step": 608
     },
     {
-      "epoch": 43.00961538461539,
-      "grad_norm": 7.290647029876709,
-      "learning_rate": 1.4839572192513372e-05,
-      "loss": 0.1998,
+      "epoch": 38.001041666666666,
+      "grad_norm": 6.658583641052246,
+      "learning_rate": 3.7905092592592594e-05,
+      "loss": 1.287,
       "step": 610
     },
     {
-      "epoch": 43.01682692307692,
-      "eval_accuracy": 0.6959459459459459,
-      "eval_loss": 1.2359126806259155,
-      "eval_runtime": 72.9535,
-      "eval_samples_per_second": 6.086,
-      "eval_steps_per_second": 0.096,
-      "step": 616
-    },
-    {
-      "epoch": 44.00480769230769,
-      "grad_norm": 5.636298656463623,
-      "learning_rate": 1.4171122994652408e-05,
-      "loss": 0.1639,
+      "epoch": 38.00625,
+      "grad_norm": 5.483121395111084,
+      "learning_rate": 3.7615740740740744e-05,
+      "loss": 1.4035,
       "step": 620
     },
     {
-      "epoch": 44.01682692307692,
-      "grad_norm": 3.6947968006134033,
-      "learning_rate": 1.3502673796791445e-05,
-      "loss": 0.1727,
-      "step": 630
+      "epoch": 38.00833333333333,
+      "eval_accuracy": 0.4794034090909091,
+      "eval_loss": 2.116427183151245,
+      "eval_runtime": 133.3027,
+      "eval_samples_per_second": 10.562,
+      "eval_steps_per_second": 0.165,
+      "step": 624
     },
     {
-      "epoch": 44.01682692307692,
-      "eval_accuracy": 0.6936936936936937,
-      "eval_loss": 1.232095718383789,
-      "eval_runtime": 81.4582,
-      "eval_samples_per_second": 5.451,
-      "eval_steps_per_second": 0.086,
+      "epoch": 39.003125,
+      "grad_norm": 6.027029991149902,
+      "learning_rate": 3.7326388888888893e-05,
+      "loss": 1.416,
       "step": 630
     },
     {
-      "epoch": 45.01201923076923,
-      "grad_norm": 3.2019245624542236,
-      "learning_rate": 1.2834224598930484e-05,
-      "loss": 0.1564,
+      "epoch": 39.00833333333333,
+      "grad_norm": 17.550609588623047,
+      "learning_rate": 3.7037037037037037e-05,
+      "loss": 1.5892,
       "step": 640
     },
     {
-      "epoch": 45.01682692307692,
-      "eval_accuracy": 0.713963963963964,
-      "eval_loss": 1.1604844331741333,
-      "eval_runtime": 72.1149,
-      "eval_samples_per_second": 6.157,
-      "eval_steps_per_second": 0.097,
-      "step": 644
+      "epoch": 39.00833333333333,
+      "eval_accuracy": 0.48011363636363635,
+      "eval_loss": 2.102943181991577,
+      "eval_runtime": 124.9558,
+      "eval_samples_per_second": 11.268,
+      "eval_steps_per_second": 0.176,
+      "step": 640
     },
     {
-      "epoch": 46.00721153846154,
-      "grad_norm": 1.9841790199279785,
-      "learning_rate": 1.2165775401069519e-05,
-      "loss": 0.1888,
+      "epoch": 40.005208333333336,
+      "grad_norm": 8.286770820617676,
+      "learning_rate": 3.6747685185185186e-05,
+      "loss": 1.3647,
       "step": 650
     },
     {
-      "epoch": 46.01682692307692,
-      "eval_accuracy": 0.7094594594594594,
-      "eval_loss": 1.1609092950820923,
-      "eval_runtime": 71.8777,
-      "eval_samples_per_second": 6.177,
-      "eval_steps_per_second": 0.097,
-      "step": 658
+      "epoch": 40.00833333333333,
+      "eval_accuracy": 0.4928977272727273,
+      "eval_loss": 2.0912482738494873,
+      "eval_runtime": 125.7809,
+      "eval_samples_per_second": 11.194,
+      "eval_steps_per_second": 0.175,
+      "step": 656
     },
     {
-      "epoch": 47.00240384615385,
-      "grad_norm": 7.085150241851807,
-      "learning_rate": 1.1497326203208558e-05,
-      "loss": 0.1618,
+      "epoch": 41.00208333333333,
+      "grad_norm": 6.9772138595581055,
+      "learning_rate": 3.6458333333333336e-05,
+      "loss": 1.3769,
       "step": 660
     },
     {
-      "epoch": 47.01442307692308,
-      "grad_norm": 5.41819429397583,
-      "learning_rate": 1.0828877005347594e-05,
-      "loss": 0.1227,
+      "epoch": 41.00729166666667,
+      "grad_norm": 6.215446472167969,
+      "learning_rate": 3.6168981481481486e-05,
+      "loss": 1.388,
       "step": 670
     },
     {
-      "epoch": 47.01682692307692,
-      "eval_accuracy": 0.7117117117117117,
-      "eval_loss": 1.1588457822799683,
-      "eval_runtime": 82.846,
-      "eval_samples_per_second": 5.359,
-      "eval_steps_per_second": 0.084,
+      "epoch": 41.00833333333333,
+      "eval_accuracy": 0.47301136363636365,
+      "eval_loss": 2.1330864429473877,
+      "eval_runtime": 134.2267,
+      "eval_samples_per_second": 10.49,
+      "eval_steps_per_second": 0.164,
       "step": 672
     },
     {
-      "epoch": 48.00961538461539,
-      "grad_norm": 2.3894031047821045,
-      "learning_rate": 1.0160427807486631e-05,
-      "loss": 0.134,
+      "epoch": 42.00416666666667,
+      "grad_norm": 6.233066082000732,
+      "learning_rate": 3.587962962962963e-05,
+      "loss": 1.3425,
       "step": 680
     },
     {
-      "epoch": 48.01682692307692,
-      "eval_accuracy": 0.7072072072072072,
-      "eval_loss": 1.1698147058486938,
-      "eval_runtime": 81.0499,
-      "eval_samples_per_second": 5.478,
-      "eval_steps_per_second": 0.086,
-      "step": 686
+      "epoch": 42.00833333333333,
+      "eval_accuracy": 0.4794034090909091,
+      "eval_loss": 2.143657922744751,
+      "eval_runtime": 126.4352,
+      "eval_samples_per_second": 11.136,
+      "eval_steps_per_second": 0.174,
+      "step": 688
     },
     {
-      "epoch": 49.00480769230769,
-      "grad_norm": 4.01511287689209,
-      "learning_rate": 9.49197860962567e-06,
-      "loss": 0.1234,
+      "epoch": 43.001041666666666,
+      "grad_norm": 7.223750591278076,
+      "learning_rate": 3.559027777777778e-05,
+      "loss": 1.1958,
       "step": 690
     },
     {
-      "epoch": 49.01682692307692,
-      "grad_norm": 7.249420642852783,
-      "learning_rate": 8.823529411764707e-06,
-      "loss": 0.1622,
+      "epoch": 43.00625,
+      "grad_norm": 6.594886779785156,
+      "learning_rate": 3.530092592592593e-05,
+      "loss": 1.2909,
       "step": 700
     },
     {
-      "epoch": 49.01682692307692,
-      "eval_accuracy": 0.6981981981981982,
-      "eval_loss": 1.2014145851135254,
-      "eval_runtime": 73.38,
-      "eval_samples_per_second": 6.051,
-      "eval_steps_per_second": 0.095,
-      "step": 700
+      "epoch": 43.00833333333333,
+      "eval_accuracy": 0.4715909090909091,
+      "eval_loss": 2.109005928039551,
+      "eval_runtime": 129.7594,
+      "eval_samples_per_second": 10.851,
+      "eval_steps_per_second": 0.17,
+      "step": 704
     },
     {
-      "epoch": 50.01201923076923,
-      "grad_norm": 4.699652194976807,
-      "learning_rate": 8.155080213903744e-06,
-      "loss": 0.1391,
+      "epoch": 44.003125,
+      "grad_norm": 5.431155681610107,
+      "learning_rate": 3.501157407407408e-05,
+      "loss": 1.2277,
       "step": 710
     },
     {
-      "epoch": 50.01682692307692,
-      "eval_accuracy": 0.7162162162162162,
-      "eval_loss": 1.1005959510803223,
-      "eval_runtime": 73.5034,
-      "eval_samples_per_second": 6.041,
-      "eval_steps_per_second": 0.095,
-      "step": 714
-    },
-    {
-      "epoch": 51.00721153846154,
-      "grad_norm": 3.143523931503296,
-      "learning_rate": 7.4866310160427806e-06,
-      "loss": 0.1276,
+      "epoch": 44.00833333333333,
+      "grad_norm": 23.15688133239746,
+      "learning_rate": 3.472222222222222e-05,
+      "loss": 1.2757,
       "step": 720
     },
     {
-      "epoch": 51.01682692307692,
-      "eval_accuracy": 0.6891891891891891,
-      "eval_loss": 1.1485284566879272,
-      "eval_runtime": 77.8096,
-      "eval_samples_per_second": 5.706,
-      "eval_steps_per_second": 0.09,
-      "step": 728
+      "epoch": 44.00833333333333,
+      "eval_accuracy": 0.4900568181818182,
+      "eval_loss": 2.0685956478118896,
+      "eval_runtime": 117.5767,
+      "eval_samples_per_second": 11.975,
+      "eval_steps_per_second": 0.187,
+      "step": 720
     },
     {
-      "epoch": 52.00240384615385,
-      "grad_norm": 1.1826622486114502,
-      "learning_rate": 6.818181818181818e-06,
-      "loss": 0.0979,
+      "epoch": 45.005208333333336,
+      "grad_norm": 6.826360702514648,
+      "learning_rate": 3.443287037037037e-05,
+      "loss": 1.181,
       "step": 730
     },
     {
-      "epoch": 52.01442307692308,
-      "grad_norm": 3.218552589416504,
-      "learning_rate": 6.149732620320856e-06,
-      "loss": 0.1222,
-      "step": 740
+      "epoch": 45.00833333333333,
+      "eval_accuracy": 0.4893465909090909,
+      "eval_loss": 2.0484628677368164,
+      "eval_runtime": 122.183,
+      "eval_samples_per_second": 11.524,
+      "eval_steps_per_second": 0.18,
+      "step": 736
     },
     {
-      "epoch": 52.01682692307692,
-      "eval_accuracy": 0.713963963963964,
-      "eval_loss": 1.0801595449447632,
-      "eval_runtime": 73.9346,
-      "eval_samples_per_second": 6.005,
-      "eval_steps_per_second": 0.095,
-      "step": 742
+      "epoch": 46.00208333333333,
+      "grad_norm": 6.446974277496338,
+      "learning_rate": 3.414351851851852e-05,
+      "loss": 1.2034,
+      "step": 740
     },
     {
-      "epoch": 53.00961538461539,
-      "grad_norm": 4.1740312576293945,
-      "learning_rate": 5.481283422459893e-06,
-      "loss": 0.1024,
+      "epoch": 46.00729166666667,
+      "grad_norm": 6.026095390319824,
+      "learning_rate": 3.385416666666667e-05,
+      "loss": 1.1825,
       "step": 750
     },
     {
-      "epoch": 53.01682692307692,
-      "eval_accuracy": 0.704954954954955,
-      "eval_loss": 1.1242848634719849,
-      "eval_runtime": 72.7853,
-      "eval_samples_per_second": 6.1,
-      "eval_steps_per_second": 0.096,
-      "step": 756
+      "epoch": 46.00833333333333,
+      "eval_accuracy": 0.484375,
+      "eval_loss": 2.0560503005981445,
+      "eval_runtime": 129.8384,
+      "eval_samples_per_second": 10.844,
+      "eval_steps_per_second": 0.169,
+      "step": 752
     },
     {
-      "epoch": 54.00480769230769,
-      "grad_norm": 3.4602904319763184,
-      "learning_rate": 4.812834224598931e-06,
-      "loss": 0.1014,
+      "epoch": 47.00416666666667,
+      "grad_norm": 7.490734577178955,
+      "learning_rate": 3.3564814814814815e-05,
+      "loss": 1.1594,
       "step": 760
     },
     {
-      "epoch": 54.01682692307692,
-      "grad_norm": 24.67173957824707,
-      "learning_rate": 4.144385026737968e-06,
-      "loss": 0.1186,
-      "step": 770
+      "epoch": 47.00833333333333,
+      "eval_accuracy": 0.49644886363636365,
+      "eval_loss": 2.0327041149139404,
+      "eval_runtime": 125.1147,
+      "eval_samples_per_second": 11.254,
+      "eval_steps_per_second": 0.176,
+      "step": 768
     },
     {
-      "epoch": 54.01682692307692,
-      "eval_accuracy": 0.7297297297297297,
-      "eval_loss": 1.0909080505371094,
-      "eval_runtime": 74.1097,
-      "eval_samples_per_second": 5.991,
-      "eval_steps_per_second": 0.094,
+      "epoch": 48.001041666666666,
+      "grad_norm": 6.074591636657715,
+      "learning_rate": 3.3275462962962965e-05,
+      "loss": 1.0989,
       "step": 770
     },
     {
-      "epoch": 55.01201923076923,
-      "grad_norm": 2.6615827083587646,
-      "learning_rate": 3.4759358288770056e-06,
-      "loss": 0.1121,
+      "epoch": 48.00625,
+      "grad_norm": 7.924504280090332,
+      "learning_rate": 3.2986111111111115e-05,
+      "loss": 1.1699,
       "step": 780
     },
     {
-      "epoch": 55.01682692307692,
-      "eval_accuracy": 0.7094594594594594,
-      "eval_loss": 1.1351189613342285,
-      "eval_runtime": 73.5115,
-      "eval_samples_per_second": 6.04,
-      "eval_steps_per_second": 0.095,
+      "epoch": 48.00833333333333,
+      "eval_accuracy": 0.4765625,
+      "eval_loss": 2.095003366470337,
+      "eval_runtime": 131.8398,
+      "eval_samples_per_second": 10.68,
+      "eval_steps_per_second": 0.167,
       "step": 784
     },
     {
-      "epoch": 56.00721153846154,
-      "grad_norm": 3.412928581237793,
-      "learning_rate": 2.807486631016043e-06,
-      "loss": 0.1284,
+      "epoch": 49.003125,
+      "grad_norm": 7.409806728363037,
+      "learning_rate": 3.2696759259259265e-05,
+      "loss": 1.2027,
       "step": 790
     },
     {
-      "epoch": 56.01682692307692,
-      "eval_accuracy": 0.7274774774774775,
-      "eval_loss": 1.1095025539398193,
-      "eval_runtime": 73.055,
-      "eval_samples_per_second": 6.078,
-      "eval_steps_per_second": 0.096,
-      "step": 798
+      "epoch": 49.00833333333333,
+      "grad_norm": 23.586345672607422,
+      "learning_rate": 3.240740740740741e-05,
+      "loss": 1.1908,
+      "step": 800
     },
     {
-      "epoch": 57.00240384615385,
-      "grad_norm": 4.214470386505127,
-      "learning_rate": 2.1390374331550802e-06,
-      "loss": 0.1418,
+      "epoch": 49.00833333333333,
+      "eval_accuracy": 0.4850852272727273,
+      "eval_loss": 2.0465078353881836,
+      "eval_runtime": 129.3358,
+      "eval_samples_per_second": 10.886,
+      "eval_steps_per_second": 0.17,
       "step": 800
     },
     {
-      "epoch": 57.01442307692308,
-      "grad_norm": 2.0941081047058105,
-      "learning_rate": 1.4705882352941177e-06,
-      "loss": 0.0893,
+      "epoch": 50.005208333333336,
+      "grad_norm": 7.079756736755371,
+      "learning_rate": 3.211805555555556e-05,
+      "loss": 1.1149,
       "step": 810
     },
     {
-      "epoch": 57.01682692307692,
-      "eval_accuracy": 0.7319819819819819,
-      "eval_loss": 1.084416389465332,
-      "eval_runtime": 71.5145,
-      "eval_samples_per_second": 6.209,
-      "eval_steps_per_second": 0.098,
-      "step": 812
+      "epoch": 50.00833333333333,
+      "eval_accuracy": 0.48792613636363635,
+      "eval_loss": 2.0569851398468018,
+      "eval_runtime": 123.867,
+      "eval_samples_per_second": 11.367,
+      "eval_steps_per_second": 0.178,
+      "step": 816
     },
     {
-      "epoch": 58.00961538461539,
-      "grad_norm": 4.0398101806640625,
-      "learning_rate": 8.021390374331552e-07,
-      "loss": 0.0878,
+      "epoch": 51.00208333333333,
+      "grad_norm": 5.848146438598633,
+      "learning_rate": 3.182870370370371e-05,
+      "loss": 1.0907,
       "step": 820
     },
     {
-      "epoch": 58.01682692307692,
-      "eval_accuracy": 0.7297297297297297,
-      "eval_loss": 1.0803806781768799,
-      "eval_runtime": 71.6842,
-      "eval_samples_per_second": 6.194,
-      "eval_steps_per_second": 0.098,
-      "step": 826
-    },
-    {
-      "epoch": 59.00480769230769,
-      "grad_norm": 1.0705435276031494,
-      "learning_rate": 1.3368983957219251e-07,
-      "loss": 0.0887,
+      "epoch": 51.00729166666667,
+      "grad_norm": 7.725104808807373,
+      "learning_rate": 3.153935185185186e-05,
+      "loss": 1.1388,
       "step": 830
     },
     {
-      "epoch": 59.00721153846154,
-      "eval_accuracy": 0.7297297297297297,
-      "eval_loss": 1.0809234380722046,
-      "eval_runtime": 72.7582,
-      "eval_samples_per_second": 6.102,
-      "eval_steps_per_second": 0.096,
+      "epoch": 51.00833333333333,
+      "eval_accuracy": 0.4978693181818182,
+      "eval_loss": 2.0232162475585938,
+      "eval_runtime": 124.9726,
+      "eval_samples_per_second": 11.266,
+      "eval_steps_per_second": 0.176,
       "step": 832
     },
     {
-      "epoch": 59.00721153846154,
-      "step": 832,
-      "total_flos": 6.1975283578694664e+19,
-      "train_loss": 1.175738023289551,
-      "train_runtime": 12896.1846,
-      "train_samples_per_second": 4.129,
-      "train_steps_per_second": 0.065
+      "epoch": 52.00416666666667,
+      "grad_norm": 6.411397457122803,
+      "learning_rate": 3.125e-05,
+      "loss": 1.0421,
+      "step": 840
     },
     {
-      "epoch": 59.00721153846154,
-      "eval_accuracy": 0.7319819819819819,
-      "eval_loss": 1.084416389465332,
-      "eval_runtime": 73.94,
-      "eval_samples_per_second": 6.005,
-      "eval_steps_per_second": 0.095,
-      "step": 832
+      "epoch": 52.00833333333333,
+      "eval_accuracy": 0.49857954545454547,
+      "eval_loss": 2.0132648944854736,
+      "eval_runtime": 130.0486,
+      "eval_samples_per_second": 10.827,
+      "eval_steps_per_second": 0.169,
+      "step": 848
     },
     {
-      "epoch": 59.00721153846154,
-      "eval_accuracy": 0.7319819819819819,
-      "eval_loss": 1.084416389465332,
-      "eval_runtime": 75.3126,
-      "eval_samples_per_second": 5.895,
-      "eval_steps_per_second": 0.093,
-      "step": 832
+      "epoch": 53.001041666666666,
+      "grad_norm": 6.417796611785889,
+      "learning_rate": 3.0960648148148144e-05,
+      "loss": 1.128,
+      "step": 850
+    },
+    {
+      "epoch": 53.00625,
+      "grad_norm": 6.483155727386475,
+      "learning_rate": 3.06712962962963e-05,
+      "loss": 1.1243,
+      "step": 860
+    },
+    {
+      "epoch": 53.00833333333333,
+      "eval_accuracy": 0.4900568181818182,
+      "eval_loss": 2.0420944690704346,
+      "eval_runtime": 131.0098,
+      "eval_samples_per_second": 10.747,
+      "eval_steps_per_second": 0.168,
+      "step": 864
+    },
+    {
+      "epoch": 54.003125,
+      "grad_norm": 7.885180950164795,
+      "learning_rate": 3.0381944444444444e-05,
+      "loss": 1.0331,
+      "step": 870
+    },
+    {
+      "epoch": 54.00833333333333,
+      "grad_norm": 28.23676300048828,
+      "learning_rate": 3.0092592592592593e-05,
+      "loss": 1.1064,
+      "step": 880
+    },
+    {
+      "epoch": 54.00833333333333,
+      "eval_accuracy": 0.5042613636363636,
+      "eval_loss": 1.9613640308380127,
+      "eval_runtime": 129.6513,
+      "eval_samples_per_second": 10.86,
+      "eval_steps_per_second": 0.17,
+      "step": 880
+    },
+    {
+      "epoch": 55.005208333333336,
+      "grad_norm": 7.0548319816589355,
+      "learning_rate": 2.980324074074074e-05,
+      "loss": 0.9778,
+      "step": 890
+    },
+    {
+      "epoch": 55.00833333333333,
+      "eval_accuracy": 0.5071022727272727,
+      "eval_loss": 1.9938956499099731,
+      "eval_runtime": 125.6294,
+      "eval_samples_per_second": 11.208,
+      "eval_steps_per_second": 0.175,
+      "step": 896
+    },
+    {
+      "epoch": 56.00208333333333,
+      "grad_norm": 7.993034362792969,
+      "learning_rate": 2.951388888888889e-05,
+      "loss": 1.0192,
+      "step": 900
+    },
+    {
+      "epoch": 56.00729166666667,
+      "grad_norm": 8.833446502685547,
+      "learning_rate": 2.9224537037037036e-05,
+      "loss": 1.1417,
+      "step": 910
+    },
+    {
+      "epoch": 56.00833333333333,
+      "eval_accuracy": 0.5106534090909091,
+      "eval_loss": 1.977372646331787,
+      "eval_runtime": 141.0715,
+      "eval_samples_per_second": 9.981,
+      "eval_steps_per_second": 0.156,
+      "step": 912
+    },
+    {
+      "epoch": 57.00416666666667,
+      "grad_norm": 6.480681896209717,
+      "learning_rate": 2.8935185185185186e-05,
+      "loss": 1.0578,
+      "step": 920
+    },
+    {
+      "epoch": 57.00833333333333,
+      "eval_accuracy": 0.5305397727272727,
+      "eval_loss": 1.9625455141067505,
+      "eval_runtime": 137.3354,
+      "eval_samples_per_second": 10.252,
+      "eval_steps_per_second": 0.16,
+      "step": 928
+    },
+    {
+      "epoch": 58.001041666666666,
+      "grad_norm": 6.03174352645874,
+      "learning_rate": 2.8645833333333333e-05,
+      "loss": 0.9695,
+      "step": 930
+    },
+    {
+      "epoch": 58.00625,
+      "grad_norm": 5.930421352386475,
+      "learning_rate": 2.8356481481481483e-05,
+      "loss": 1.0904,
+      "step": 940
+    },
+    {
+      "epoch": 58.00833333333333,
+      "eval_accuracy": 0.5056818181818182,
+      "eval_loss": 1.971279501914978,
+      "eval_runtime": 124.5828,
+      "eval_samples_per_second": 11.302,
+      "eval_steps_per_second": 0.177,
+      "step": 944
+    },
+    {
+      "epoch": 59.003125,
+      "grad_norm": 7.177443027496338,
+      "learning_rate": 2.806712962962963e-05,
+      "loss": 1.0493,
+      "step": 950
+    },
+    {
+      "epoch": 59.00833333333333,
+      "grad_norm": 32.896873474121094,
+      "learning_rate": 2.777777777777778e-05,
+      "loss": 1.2569,
+      "step": 960
+    },
+    {
+      "epoch": 59.00833333333333,
+      "eval_accuracy": 0.5255681818181818,
+      "eval_loss": 1.9495558738708496,
+      "eval_runtime": 125.0434,
+      "eval_samples_per_second": 11.26,
+      "eval_steps_per_second": 0.176,
+      "step": 960
+    },
+    {
+      "epoch": 60.005208333333336,
+      "grad_norm": 6.914553642272949,
+      "learning_rate": 2.7488425925925926e-05,
+      "loss": 1.076,
+      "step": 970
+    },
+    {
+      "epoch": 60.00833333333333,
+      "eval_accuracy": 0.5369318181818182,
+      "eval_loss": 1.9237945079803467,
+      "eval_runtime": 130.5511,
+      "eval_samples_per_second": 10.785,
+      "eval_steps_per_second": 0.169,
+      "step": 976
+    },
+    {
+      "epoch": 61.00208333333333,
+      "grad_norm": 6.131436347961426,
+      "learning_rate": 2.7199074074074076e-05,
+      "loss": 0.9972,
+      "step": 980
+    },
+    {
+      "epoch": 61.00729166666667,
+      "grad_norm": 8.375364303588867,
+      "learning_rate": 2.6909722222222222e-05,
+      "loss": 1.018,
+      "step": 990
+    },
+    {
+      "epoch": 61.00833333333333,
+      "eval_accuracy": 0.515625,
+      "eval_loss": 1.9578070640563965,
+      "eval_runtime": 128.8964,
+      "eval_samples_per_second": 10.924,
+      "eval_steps_per_second": 0.171,
+      "step": 992
+    },
+    {
+      "epoch": 62.00416666666667,
+      "grad_norm": 8.409364700317383,
+      "learning_rate": 2.6620370370370372e-05,
+      "loss": 0.8569,
+      "step": 1000
+    },
+    {
+      "epoch": 62.00833333333333,
+      "eval_accuracy": 0.5184659090909091,
+      "eval_loss": 1.9410430192947388,
+      "eval_runtime": 126.0956,
+      "eval_samples_per_second": 11.166,
+      "eval_steps_per_second": 0.174,
+      "step": 1008
+    },
+    {
+      "epoch": 63.001041666666666,
+      "grad_norm": 6.331228256225586,
+      "learning_rate": 2.633101851851852e-05,
+      "loss": 1.1689,
+      "step": 1010
+    },
+    {
+      "epoch": 63.00625,
+      "grad_norm": 5.7633161544799805,
+      "learning_rate": 2.604166666666667e-05,
+      "loss": 0.9847,
+      "step": 1020
+    },
+    {
+      "epoch": 63.00833333333333,
+      "eval_accuracy": 0.5134943181818182,
+      "eval_loss": 1.965501308441162,
+      "eval_runtime": 134.8984,
+      "eval_samples_per_second": 10.437,
+      "eval_steps_per_second": 0.163,
+      "step": 1024
+    },
+    {
+      "epoch": 64.003125,
+      "grad_norm": 6.778160572052002,
+      "learning_rate": 2.5752314814814815e-05,
+      "loss": 1.0541,
+      "step": 1030
+    },
+    {
+      "epoch": 64.00833333333334,
+      "grad_norm": 15.068023681640625,
+      "learning_rate": 2.5462962962962965e-05,
+      "loss": 0.8992,
+      "step": 1040
+    },
+    {
+      "epoch": 64.00833333333334,
+      "eval_accuracy": 0.5184659090909091,
+      "eval_loss": 1.97407865524292,
+      "eval_runtime": 131.1607,
+      "eval_samples_per_second": 10.735,
+      "eval_steps_per_second": 0.168,
+      "step": 1040
+    },
+    {
+      "epoch": 65.00520833333333,
+      "grad_norm": 6.6322126388549805,
+      "learning_rate": 2.517361111111111e-05,
+      "loss": 0.9781,
+      "step": 1050
+    },
+    {
+      "epoch": 65.00833333333334,
+      "eval_accuracy": 0.5248579545454546,
+      "eval_loss": 1.9591399431228638,
+      "eval_runtime": 130.8261,
+      "eval_samples_per_second": 10.762,
+      "eval_steps_per_second": 0.168,
+      "step": 1056
+    },
+    {
+      "epoch": 66.00208333333333,
+      "grad_norm": 6.270596504211426,
+      "learning_rate": 2.488425925925926e-05,
+      "loss": 1.1274,
+      "step": 1060
+    },
+    {
+      "epoch": 66.00729166666666,
+      "grad_norm": 5.216790676116943,
+      "learning_rate": 2.4594907407407408e-05,
+      "loss": 0.9016,
+      "step": 1070
+    },
+    {
+      "epoch": 66.00833333333334,
+      "eval_accuracy": 0.5134943181818182,
+      "eval_loss": 1.9802237749099731,
+      "eval_runtime": 138.409,
+      "eval_samples_per_second": 10.173,
+      "eval_steps_per_second": 0.159,
+      "step": 1072
+    },
+    {
+      "epoch": 67.00416666666666,
+      "grad_norm": 4.889580249786377,
+      "learning_rate": 2.4305555555555558e-05,
+      "loss": 0.9443,
+      "step": 1080
+    },
+    {
+      "epoch": 67.00833333333334,
+      "eval_accuracy": 0.5035511363636364,
+      "eval_loss": 1.9881755113601685,
+      "eval_runtime": 132.577,
+      "eval_samples_per_second": 10.62,
+      "eval_steps_per_second": 0.166,
+      "step": 1088
+    },
+    {
+      "epoch": 68.00104166666667,
+      "grad_norm": 8.338197708129883,
+      "learning_rate": 2.4016203703703704e-05,
+      "loss": 0.9719,
+      "step": 1090
+    },
+    {
+      "epoch": 68.00625,
+      "grad_norm": 7.402223110198975,
+      "learning_rate": 2.3726851851851854e-05,
+      "loss": 0.9359,
+      "step": 1100
+    },
+    {
+      "epoch": 68.00833333333334,
+      "eval_accuracy": 0.5092329545454546,
+      "eval_loss": 2.0045785903930664,
+      "eval_runtime": 122.9735,
+      "eval_samples_per_second": 11.45,
+      "eval_steps_per_second": 0.179,
+      "step": 1104
+    },
+    {
+      "epoch": 69.003125,
+      "grad_norm": 5.299436092376709,
+      "learning_rate": 2.34375e-05,
+      "loss": 0.9554,
+      "step": 1110
+    },
+    {
+      "epoch": 69.00833333333334,
+      "grad_norm": 22.75337028503418,
+      "learning_rate": 2.314814814814815e-05,
+      "loss": 0.7735,
+      "step": 1120
+    },
+    {
+      "epoch": 69.00833333333334,
+      "eval_accuracy": 0.5063920454545454,
+      "eval_loss": 2.017183542251587,
+      "eval_runtime": 132.6432,
+      "eval_samples_per_second": 10.615,
+      "eval_steps_per_second": 0.166,
+      "step": 1120
+    },
+    {
+      "epoch": 70.00520833333333,
+      "grad_norm": 6.869340896606445,
+      "learning_rate": 2.2858796296296297e-05,
+      "loss": 0.9405,
+      "step": 1130
+    },
+    {
+      "epoch": 70.00833333333334,
+      "eval_accuracy": 0.5269886363636364,
+      "eval_loss": 1.955207109451294,
+      "eval_runtime": 135.5316,
+      "eval_samples_per_second": 10.389,
+      "eval_steps_per_second": 0.162,
+      "step": 1136
+    },
+    {
+      "epoch": 71.00208333333333,
+      "grad_norm": 7.737886428833008,
+      "learning_rate": 2.2569444444444447e-05,
+      "loss": 0.9419,
+      "step": 1140
+    },
+    {
+      "epoch": 71.00729166666666,
+      "grad_norm": 6.463024139404297,
+      "learning_rate": 2.2280092592592593e-05,
+      "loss": 0.9709,
+      "step": 1150
+    },
+    {
+      "epoch": 71.00833333333334,
+      "eval_accuracy": 0.5227272727272727,
+      "eval_loss": 1.9572663307189941,
+      "eval_runtime": 134.3176,
+      "eval_samples_per_second": 10.483,
+      "eval_steps_per_second": 0.164,
+      "step": 1152
+    },
+    {
+      "epoch": 72.00416666666666,
+      "grad_norm": 6.5754923820495605,
+      "learning_rate": 2.1990740740740743e-05,
+      "loss": 0.9914,
+      "step": 1160
+    },
+    {
+      "epoch": 72.00833333333334,
+      "eval_accuracy": 0.5248579545454546,
+      "eval_loss": 1.9767735004425049,
+      "eval_runtime": 125.4669,
+      "eval_samples_per_second": 11.222,
+      "eval_steps_per_second": 0.175,
+      "step": 1168
+    },
+    {
+      "epoch": 73.00104166666667,
+      "grad_norm": 6.831247329711914,
+      "learning_rate": 2.170138888888889e-05,
+      "loss": 0.8804,
+      "step": 1170
+    },
+    {
+      "epoch": 73.00625,
+      "grad_norm": 7.8077392578125,
+      "learning_rate": 2.141203703703704e-05,
+      "loss": 0.8487,
+      "step": 1180
+    },
+    {
+      "epoch": 73.00833333333334,
+      "eval_accuracy": 0.5326704545454546,
+      "eval_loss": 1.9569602012634277,
+      "eval_runtime": 134.6123,
+      "eval_samples_per_second": 10.46,
+      "eval_steps_per_second": 0.163,
+      "step": 1184
+    },
+    {
+      "epoch": 74.003125,
+      "grad_norm": 5.786107063293457,
+      "learning_rate": 2.1122685185185186e-05,
+      "loss": 0.9529,
+      "step": 1190
+    },
+    {
+      "epoch": 74.00833333333334,
+      "grad_norm": 21.245386123657227,
+      "learning_rate": 2.0833333333333336e-05,
+      "loss": 0.835,
+      "step": 1200
+    },
+    {
+      "epoch": 74.00833333333334,
+      "eval_accuracy": 0.5241477272727273,
+      "eval_loss": 1.9758590459823608,
+      "eval_runtime": 129.6528,
+      "eval_samples_per_second": 10.86,
+      "eval_steps_per_second": 0.17,
+      "step": 1200
+    },
+    {
+      "epoch": 75.00520833333333,
+      "grad_norm": 5.697115421295166,
+      "learning_rate": 2.0543981481481483e-05,
+      "loss": 0.8914,
+      "step": 1210
+    },
+    {
+      "epoch": 75.00833333333334,
+      "eval_accuracy": 0.5298295454545454,
+      "eval_loss": 1.9309029579162598,
+      "eval_runtime": 130.377,
+      "eval_samples_per_second": 10.799,
+      "eval_steps_per_second": 0.169,
+      "step": 1216
+    },
+    {
+      "epoch": 76.00208333333333,
+      "grad_norm": 4.525904655456543,
+      "learning_rate": 2.0254629629629632e-05,
+      "loss": 0.9268,
+      "step": 1220
+    },
+    {
+      "epoch": 76.00729166666666,
+      "grad_norm": 6.657559871673584,
+      "learning_rate": 1.996527777777778e-05,
+      "loss": 0.9242,
+      "step": 1230
+    },
+    {
+      "epoch": 76.00833333333334,
+      "eval_accuracy": 0.5241477272727273,
+      "eval_loss": 1.9594990015029907,
+      "eval_runtime": 130.7531,
+      "eval_samples_per_second": 10.768,
+      "eval_steps_per_second": 0.168,
+      "step": 1232
+    },
+    {
+      "epoch": 77.00416666666666,
+      "grad_norm": 6.962241172790527,
+      "learning_rate": 1.967592592592593e-05,
+      "loss": 0.8235,
+      "step": 1240
+    },
+    {
+      "epoch": 77.00833333333334,
+      "eval_accuracy": 0.5276988636363636,
+      "eval_loss": 1.9556376934051514,
+      "eval_runtime": 124.6215,
+      "eval_samples_per_second": 11.298,
+      "eval_steps_per_second": 0.177,
+      "step": 1248
+    },
+    {
+      "epoch": 78.00104166666667,
+      "grad_norm": 6.555630207061768,
+      "learning_rate": 1.9386574074074075e-05,
+      "loss": 1.1044,
+      "step": 1250
+    },
+    {
+      "epoch": 78.00625,
+      "grad_norm": 7.0458455085754395,
+      "learning_rate": 1.9097222222222222e-05,
+      "loss": 0.8664,
+      "step": 1260
+    },
+    {
+      "epoch": 78.00833333333334,
+      "eval_accuracy": 0.5134943181818182,
+      "eval_loss": 1.978991985321045,
+      "eval_runtime": 130.4341,
+      "eval_samples_per_second": 10.795,
+      "eval_steps_per_second": 0.169,
+      "step": 1264
+    },
+    {
+      "epoch": 79.003125,
+      "grad_norm": 4.815821170806885,
+      "learning_rate": 1.8807870370370372e-05,
+      "loss": 0.7884,
+      "step": 1270
+    },
+    {
+      "epoch": 79.00833333333334,
+      "grad_norm": 26.361722946166992,
+      "learning_rate": 1.8518518518518518e-05,
+      "loss": 0.8699,
+      "step": 1280
+    },
+    {
+      "epoch": 79.00833333333334,
+      "eval_accuracy": 0.5227272727272727,
+      "eval_loss": 1.9835097789764404,
+      "eval_runtime": 119.0563,
+      "eval_samples_per_second": 11.826,
+      "eval_steps_per_second": 0.185,
+      "step": 1280
+    },
+    {
+      "epoch": 80.00520833333333,
+      "grad_norm": 4.6080474853515625,
+      "learning_rate": 1.8229166666666668e-05,
+      "loss": 0.9112,
+      "step": 1290
+    },
+    {
+      "epoch": 80.00833333333334,
+      "eval_accuracy": 0.5291193181818182,
+      "eval_loss": 1.9426430463790894,
+      "eval_runtime": 137.1174,
+      "eval_samples_per_second": 10.269,
+      "eval_steps_per_second": 0.16,
+      "step": 1296
+    },
+    {
+      "epoch": 81.00208333333333,
+      "grad_norm": 5.045470237731934,
+      "learning_rate": 1.7939814814814815e-05,
+      "loss": 0.8392,
+      "step": 1300
+    },
+    {
+      "epoch": 81.00729166666666,
+      "grad_norm": 6.203034400939941,
+      "learning_rate": 1.7650462962962965e-05,
+      "loss": 0.7901,
+      "step": 1310
+    },
+    {
+      "epoch": 81.00833333333334,
+      "eval_accuracy": 0.5255681818181818,
+      "eval_loss": 1.959786295890808,
+      "eval_runtime": 120.2871,
+      "eval_samples_per_second": 11.705,
+      "eval_steps_per_second": 0.183,
+      "step": 1312
+    },
+    {
+      "epoch": 82.00416666666666,
+      "grad_norm": 6.214341163635254,
+      "learning_rate": 1.736111111111111e-05,
+      "loss": 0.8186,
+      "step": 1320
+    },
+    {
+      "epoch": 82.00833333333334,
+      "eval_accuracy": 0.5319602272727273,
+      "eval_loss": 1.9397464990615845,
+      "eval_runtime": 130.0941,
+      "eval_samples_per_second": 10.823,
+      "eval_steps_per_second": 0.169,
+      "step": 1328
+    },
+    {
+      "epoch": 83.00104166666667,
+      "grad_norm": 5.813119411468506,
+      "learning_rate": 1.707175925925926e-05,
+      "loss": 0.7438,
+      "step": 1330
+    },
+    {
+      "epoch": 83.00625,
+      "grad_norm": 5.390239715576172,
+      "learning_rate": 1.6782407407407408e-05,
+      "loss": 0.8229,
+      "step": 1340
+    },
+    {
+      "epoch": 83.00833333333334,
+      "eval_accuracy": 0.5326704545454546,
+      "eval_loss": 1.938385009765625,
+      "eval_runtime": 124.8818,
+      "eval_samples_per_second": 11.275,
+      "eval_steps_per_second": 0.176,
+      "step": 1344
+    },
+    {
+      "epoch": 84.003125,
+      "grad_norm": 5.821295738220215,
+      "learning_rate": 1.6493055555555557e-05,
+      "loss": 1.0251,
+      "step": 1350
+    },
+    {
+      "epoch": 84.00833333333334,
+      "grad_norm": 23.789480209350586,
+      "learning_rate": 1.6203703703703704e-05,
+      "loss": 0.9063,
+      "step": 1360
+    },
+    {
+      "epoch": 84.00833333333334,
+      "eval_accuracy": 0.5291193181818182,
+      "eval_loss": 1.9323524236679077,
+      "eval_runtime": 122.6737,
+      "eval_samples_per_second": 11.478,
+      "eval_steps_per_second": 0.179,
+      "step": 1360
+    },
+    {
+      "epoch": 85.00520833333333,
+      "grad_norm": 7.094452381134033,
+      "learning_rate": 1.5914351851851854e-05,
+      "loss": 0.8843,
+      "step": 1370
+    },
+    {
+      "epoch": 85.00833333333334,
+      "eval_accuracy": 0.5369318181818182,
+      "eval_loss": 1.9315829277038574,
+      "eval_runtime": 124.9222,
+      "eval_samples_per_second": 11.271,
+      "eval_steps_per_second": 0.176,
+      "step": 1376
+    },
+    {
+      "epoch": 86.00208333333333,
+      "grad_norm": 5.776772975921631,
+      "learning_rate": 1.5625e-05,
+      "loss": 0.7684,
+      "step": 1380
+    },
+    {
+      "epoch": 86.00729166666666,
+      "grad_norm": 8.940707206726074,
+      "learning_rate": 1.533564814814815e-05,
+      "loss": 0.7904,
+      "step": 1390
+    },
+    {
+      "epoch": 86.00833333333334,
+      "eval_accuracy": 0.5376420454545454,
+      "eval_loss": 1.926943302154541,
+      "eval_runtime": 126.2416,
+      "eval_samples_per_second": 11.153,
+      "eval_steps_per_second": 0.174,
+      "step": 1392
+    },
+    {
+      "epoch": 87.00416666666666,
+      "grad_norm": 6.900567054748535,
+      "learning_rate": 1.5046296296296297e-05,
+      "loss": 0.7942,
+      "step": 1400
+    },
+    {
+      "epoch": 87.00833333333334,
+      "eval_accuracy": 0.5291193181818182,
+      "eval_loss": 1.9505839347839355,
+      "eval_runtime": 129.4495,
+      "eval_samples_per_second": 10.877,
+      "eval_steps_per_second": 0.17,
+      "step": 1408
+    },
+    {
+      "epoch": 88.00104166666667,
+      "grad_norm": 6.561887741088867,
+      "learning_rate": 1.4756944444444445e-05,
+      "loss": 0.7708,
+      "step": 1410
+    },
+    {
+      "epoch": 88.00625,
+      "grad_norm": 6.302940368652344,
+      "learning_rate": 1.4467592592592593e-05,
+      "loss": 0.8798,
+      "step": 1420
+    },
+    {
+      "epoch": 88.00833333333334,
+      "eval_accuracy": 0.5404829545454546,
+      "eval_loss": 1.9184972047805786,
+      "eval_runtime": 136.5919,
+      "eval_samples_per_second": 10.308,
+      "eval_steps_per_second": 0.161,
+      "step": 1424
+    },
+    {
+      "epoch": 89.003125,
+      "grad_norm": 6.015424728393555,
+      "learning_rate": 1.4178240740740741e-05,
+      "loss": 0.7788,
+      "step": 1430
+    },
+    {
+      "epoch": 89.00833333333334,
+      "grad_norm": 9.586054801940918,
+      "learning_rate": 1.388888888888889e-05,
+      "loss": 0.7678,
+      "step": 1440
+    },
+    {
+      "epoch": 89.00833333333334,
+      "eval_accuracy": 0.5326704545454546,
+      "eval_loss": 1.9361698627471924,
+      "eval_runtime": 128.5409,
+      "eval_samples_per_second": 10.954,
+      "eval_steps_per_second": 0.171,
+      "step": 1440
+    },
+    {
+      "epoch": 90.00520833333333,
+      "grad_norm": 6.406522750854492,
+      "learning_rate": 1.3599537037037038e-05,
+      "loss": 0.7589,
+      "step": 1450
+    },
+    {
+      "epoch": 90.00833333333334,
+      "eval_accuracy": 0.5276988636363636,
+      "eval_loss": 1.9496175050735474,
+      "eval_runtime": 128.3627,
+      "eval_samples_per_second": 10.969,
+      "eval_steps_per_second": 0.171,
+      "step": 1456
+    },
+    {
+      "epoch": 91.00208333333333,
+      "grad_norm": 5.885864734649658,
+      "learning_rate": 1.3310185185185186e-05,
+      "loss": 0.7746,
+      "step": 1460
+    },
+    {
+      "epoch": 91.00729166666666,
+      "grad_norm": 6.763460159301758,
+      "learning_rate": 1.3020833333333334e-05,
+      "loss": 0.6679,
+      "step": 1470
+    },
+    {
+      "epoch": 91.00833333333334,
+      "eval_accuracy": 0.5298295454545454,
+      "eval_loss": 1.9507235288619995,
+      "eval_runtime": 133.1729,
+      "eval_samples_per_second": 10.573,
+      "eval_steps_per_second": 0.165,
+      "step": 1472
+    },
+    {
+      "epoch": 92.00416666666666,
+      "grad_norm": 8.202223777770996,
+      "learning_rate": 1.2731481481481482e-05,
+      "loss": 0.8042,
+      "step": 1480
+    },
+    {
+      "epoch": 92.00833333333334,
+      "eval_accuracy": 0.5369318181818182,
+      "eval_loss": 1.950987696647644,
+      "eval_runtime": 134.0073,
+      "eval_samples_per_second": 10.507,
+      "eval_steps_per_second": 0.164,
+      "step": 1488
+    },
+    {
+      "epoch": 93.00104166666667,
+      "grad_norm": 6.163456439971924,
+      "learning_rate": 1.244212962962963e-05,
+      "loss": 0.8758,
+      "step": 1490
+    },
+    {
+      "epoch": 93.00625,
+      "grad_norm": 8.015287399291992,
+      "learning_rate": 1.2152777777777779e-05,
+      "loss": 0.7722,
+      "step": 1500
+    },
+    {
+      "epoch": 93.00833333333334,
+      "eval_accuracy": 0.5333806818181818,
+      "eval_loss": 1.9502681493759155,
+      "eval_runtime": 125.0273,
+      "eval_samples_per_second": 11.262,
+      "eval_steps_per_second": 0.176,
+      "step": 1504
+    },
+    {
+      "epoch": 94.003125,
+      "grad_norm": 5.918884754180908,
+      "learning_rate": 1.1863425925925927e-05,
+      "loss": 0.7579,
+      "step": 1510
+    },
+    {
+      "epoch": 94.00833333333334,
+      "grad_norm": 18.762502670288086,
+      "learning_rate": 1.1574074074074075e-05,
+      "loss": 0.6831,
+      "step": 1520
+    },
+    {
+      "epoch": 94.00833333333334,
+      "eval_accuracy": 0.5348011363636364,
+      "eval_loss": 1.9531010389328003,
+      "eval_runtime": 120.9603,
+      "eval_samples_per_second": 11.64,
+      "eval_steps_per_second": 0.182,
+      "step": 1520
+    },
+    {
+      "epoch": 95.00520833333333,
+      "grad_norm": 7.426961421966553,
+      "learning_rate": 1.1284722222222223e-05,
+      "loss": 0.766,
+      "step": 1530
+    },
+    {
+      "epoch": 95.00833333333334,
+      "eval_accuracy": 0.5383522727272727,
+      "eval_loss": 1.9345489740371704,
+      "eval_runtime": 129.3794,
+      "eval_samples_per_second": 10.883,
+      "eval_steps_per_second": 0.17,
+      "step": 1536
+    },
+    {
+      "epoch": 96.00208333333333,
+      "grad_norm": 5.175162315368652,
+      "learning_rate": 1.0995370370370372e-05,
+      "loss": 0.7168,
+      "step": 1540
+    },
+    {
+      "epoch": 96.00729166666666,
+      "grad_norm": 7.063689708709717,
+      "learning_rate": 1.070601851851852e-05,
+      "loss": 0.8099,
+      "step": 1550
+    },
+    {
+      "epoch": 96.00833333333334,
+      "eval_accuracy": 0.5376420454545454,
+      "eval_loss": 1.9349414110183716,
+      "eval_runtime": 123.5725,
+      "eval_samples_per_second": 11.394,
+      "eval_steps_per_second": 0.178,
+      "step": 1552
+    },
+    {
+      "epoch": 97.00416666666666,
+      "grad_norm": 6.602786064147949,
+      "learning_rate": 1.0416666666666668e-05,
+      "loss": 0.7513,
+      "step": 1560
+    },
+    {
+      "epoch": 97.00833333333334,
+      "eval_accuracy": 0.5461647727272727,
+      "eval_loss": 1.9238044023513794,
+      "eval_runtime": 139.032,
+      "eval_samples_per_second": 10.127,
+      "eval_steps_per_second": 0.158,
+      "step": 1568
+    },
+    {
+      "epoch": 98.00104166666667,
+      "grad_norm": 5.149336338043213,
+      "learning_rate": 1.0127314814814816e-05,
+      "loss": 0.8236,
+      "step": 1570
+    },
+    {
+      "epoch": 98.00625,
+      "grad_norm": 8.144251823425293,
+      "learning_rate": 9.837962962962964e-06,
+      "loss": 0.6561,
+      "step": 1580
+    },
+    {
+      "epoch": 98.00833333333334,
+      "eval_accuracy": 0.5426136363636364,
+      "eval_loss": 1.9337557554244995,
+      "eval_runtime": 133.1537,
+      "eval_samples_per_second": 10.574,
+      "eval_steps_per_second": 0.165,
+      "step": 1584
+    },
+    {
+      "epoch": 99.003125,
+      "grad_norm": 5.338318347930908,
+      "learning_rate": 9.548611111111111e-06,
+      "loss": 0.778,
+      "step": 1590
+    },
+    {
+      "epoch": 99.00833333333334,
+      "grad_norm": 15.469060897827148,
+      "learning_rate": 9.259259259259259e-06,
+      "loss": 0.7423,
+      "step": 1600
+    },
+    {
+      "epoch": 99.00833333333334,
+      "eval_accuracy": 0.5539772727272727,
+      "eval_loss": 1.901895523071289,
+      "eval_runtime": 131.3276,
+      "eval_samples_per_second": 10.721,
+      "eval_steps_per_second": 0.168,
+      "step": 1600
+    },
+    {
+      "epoch": 100.00520833333333,
+      "grad_norm": 4.922194957733154,
+      "learning_rate": 8.969907407407407e-06,
+      "loss": 0.7739,
+      "step": 1610
+    },
+    {
+      "epoch": 100.00833333333334,
+      "eval_accuracy": 0.5504261363636364,
+      "eval_loss": 1.916474461555481,
+      "eval_runtime": 124.6426,
+      "eval_samples_per_second": 11.296,
+      "eval_steps_per_second": 0.177,
+      "step": 1616
+    },
+    {
+      "epoch": 101.00208333333333,
+      "grad_norm": 7.351630210876465,
+      "learning_rate": 8.680555555555556e-06,
+      "loss": 0.6743,
+      "step": 1620
+    },
+    {
+      "epoch": 101.00729166666666,
+      "grad_norm": 5.356750965118408,
+      "learning_rate": 8.391203703703704e-06,
+      "loss": 0.6562,
+      "step": 1630
+    },
+    {
+      "epoch": 101.00833333333334,
+      "eval_accuracy": 0.5433238636363636,
+      "eval_loss": 1.9271347522735596,
+      "eval_runtime": 121.1594,
+      "eval_samples_per_second": 11.621,
+      "eval_steps_per_second": 0.182,
+      "step": 1632
+    },
+    {
+      "epoch": 102.00416666666666,
+      "grad_norm": 6.858729839324951,
+      "learning_rate": 8.101851851851852e-06,
+      "loss": 0.7182,
+      "step": 1640
+    },
+    {
+      "epoch": 102.00833333333334,
+      "eval_accuracy": 0.5440340909090909,
+      "eval_loss": 1.9096194505691528,
+      "eval_runtime": 133.4895,
+      "eval_samples_per_second": 10.548,
+      "eval_steps_per_second": 0.165,
+      "step": 1648
+    },
+    {
+      "epoch": 103.00104166666667,
+      "grad_norm": 6.165463924407959,
+      "learning_rate": 7.8125e-06,
+      "loss": 0.75,
+      "step": 1650
+    },
+    {
+      "epoch": 103.00625,
+      "grad_norm": 7.791365623474121,
+      "learning_rate": 7.523148148148148e-06,
+      "loss": 0.6898,
+      "step": 1660
+    },
+    {
+      "epoch": 103.00833333333334,
+      "eval_accuracy": 0.5482954545454546,
+      "eval_loss": 1.9213480949401855,
+      "eval_runtime": 121.4191,
+      "eval_samples_per_second": 11.596,
+      "eval_steps_per_second": 0.181,
+      "step": 1664
+    },
+    {
+      "epoch": 104.003125,
+      "grad_norm": 7.106443405151367,
+      "learning_rate": 7.2337962962962966e-06,
+      "loss": 0.8279,
+      "step": 1670
+    },
+    {
+      "epoch": 104.00833333333334,
+      "grad_norm": 11.49569320678711,
+      "learning_rate": 6.944444444444445e-06,
+      "loss": 0.6541,
+      "step": 1680
+    },
+    {
+      "epoch": 104.00833333333334,
+      "eval_accuracy": 0.5433238636363636,
+      "eval_loss": 1.926275372505188,
+      "eval_runtime": 128.2374,
+      "eval_samples_per_second": 10.98,
+      "eval_steps_per_second": 0.172,
+      "step": 1680
+    },
+    {
+      "epoch": 105.00520833333333,
+      "grad_norm": 5.9457316398620605,
+      "learning_rate": 6.655092592592593e-06,
+      "loss": 0.7131,
+      "step": 1690
+    },
+    {
+      "epoch": 105.00833333333334,
+      "eval_accuracy": 0.546875,
+      "eval_loss": 1.9148098230361938,
+      "eval_runtime": 128.976,
+      "eval_samples_per_second": 10.917,
+      "eval_steps_per_second": 0.171,
+      "step": 1696
+    },
+    {
+      "epoch": 106.00208333333333,
+      "grad_norm": 7.0640716552734375,
+      "learning_rate": 6.365740740740741e-06,
+      "loss": 0.7,
+      "step": 1700
+    },
+    {
+      "epoch": 106.00729166666666,
+      "grad_norm": 5.8036699295043945,
+      "learning_rate": 6.076388888888889e-06,
+      "loss": 0.7076,
+      "step": 1710
+    },
+    {
+      "epoch": 106.00833333333334,
+      "eval_accuracy": 0.5454545454545454,
+      "eval_loss": 1.9192754030227661,
+      "eval_runtime": 128.1384,
+      "eval_samples_per_second": 10.988,
+      "eval_steps_per_second": 0.172,
+      "step": 1712
+    },
+    {
+      "epoch": 107.00416666666666,
+      "grad_norm": 6.425831317901611,
+      "learning_rate": 5.787037037037038e-06,
+      "loss": 0.7822,
+      "step": 1720
+    },
+    {
+      "epoch": 107.00833333333334,
+      "eval_accuracy": 0.5440340909090909,
+      "eval_loss": 1.916624903678894,
+      "eval_runtime": 128.0714,
+      "eval_samples_per_second": 10.994,
+      "eval_steps_per_second": 0.172,
+      "step": 1728
+    },
+    {
+      "epoch": 108.00104166666667,
+      "grad_norm": 5.592070579528809,
+      "learning_rate": 5.497685185185186e-06,
+      "loss": 0.7629,
+      "step": 1730
+    },
+    {
+      "epoch": 108.00625,
+      "grad_norm": 6.38240909576416,
+      "learning_rate": 5.208333333333334e-06,
+      "loss": 0.6955,
+      "step": 1740
+    },
+    {
+      "epoch": 108.00833333333334,
+      "eval_accuracy": 0.5490056818181818,
+      "eval_loss": 1.9166675806045532,
+      "eval_runtime": 133.0053,
+      "eval_samples_per_second": 10.586,
+      "eval_steps_per_second": 0.165,
+      "step": 1744
+    },
+    {
+      "epoch": 109.003125,
+      "grad_norm": 4.68496036529541,
+      "learning_rate": 4.918981481481482e-06,
+      "loss": 0.8267,
+      "step": 1750
+    },
+    {
+      "epoch": 109.00833333333334,
+      "grad_norm": 22.761470794677734,
+      "learning_rate": 4.6296296296296296e-06,
+      "loss": 0.6939,
+      "step": 1760
+    },
+    {
+      "epoch": 109.00833333333334,
+      "eval_accuracy": 0.5426136363636364,
+      "eval_loss": 1.9129440784454346,
+      "eval_runtime": 123.3598,
+      "eval_samples_per_second": 11.414,
+      "eval_steps_per_second": 0.178,
+      "step": 1760
+    },
+    {
+      "epoch": 110.00520833333333,
+      "grad_norm": 6.633167266845703,
+      "learning_rate": 4.340277777777778e-06,
+      "loss": 0.7149,
+      "step": 1770
+    },
+    {
+      "epoch": 110.00833333333334,
+      "eval_accuracy": 0.5355113636363636,
+      "eval_loss": 1.9237289428710938,
+      "eval_runtime": 133.1436,
+      "eval_samples_per_second": 10.575,
+      "eval_steps_per_second": 0.165,
+      "step": 1776
+    },
+    {
+      "epoch": 111.00208333333333,
+      "grad_norm": 6.027252674102783,
+      "learning_rate": 4.050925925925926e-06,
+      "loss": 0.7654,
+      "step": 1780
+    },
+    {
+      "epoch": 111.00729166666666,
+      "grad_norm": 7.40338134765625,
+      "learning_rate": 3.761574074074074e-06,
+      "loss": 0.7341,
+      "step": 1790
+    },
+    {
+      "epoch": 111.00833333333334,
+      "eval_accuracy": 0.5433238636363636,
+      "eval_loss": 1.9047017097473145,
+      "eval_runtime": 133.6961,
+      "eval_samples_per_second": 10.531,
+      "eval_steps_per_second": 0.165,
+      "step": 1792
+    },
+    {
+      "epoch": 112.00416666666666,
+      "grad_norm": 6.874715328216553,
+      "learning_rate": 3.4722222222222224e-06,
+      "loss": 0.7101,
+      "step": 1800
+    },
+    {
+      "epoch": 112.00833333333334,
+      "eval_accuracy": 0.5433238636363636,
+      "eval_loss": 1.9010353088378906,
+      "eval_runtime": 130.9035,
+      "eval_samples_per_second": 10.756,
+      "eval_steps_per_second": 0.168,
+      "step": 1808
+    },
+    {
+      "epoch": 113.00104166666667,
+      "grad_norm": 6.984443664550781,
+      "learning_rate": 3.1828703703703706e-06,
+      "loss": 0.6031,
+      "step": 1810
+    },
+    {
+      "epoch": 113.00625,
+      "grad_norm": 7.135616302490234,
+      "learning_rate": 2.893518518518519e-06,
+      "loss": 0.764,
+      "step": 1820
+    },
+    {
+      "epoch": 113.00833333333334,
+      "eval_accuracy": 0.5454545454545454,
+      "eval_loss": 1.9023563861846924,
+      "eval_runtime": 128.9627,
+      "eval_samples_per_second": 10.918,
+      "eval_steps_per_second": 0.171,
+      "step": 1824
+    },
+    {
+      "epoch": 114.003125,
+      "grad_norm": 5.495314121246338,
+      "learning_rate": 2.604166666666667e-06,
+      "loss": 0.7183,
+      "step": 1830
+    },
+    {
+      "epoch": 114.00833333333334,
+      "grad_norm": 25.88104248046875,
+      "learning_rate": 2.3148148148148148e-06,
+      "loss": 0.667,
+      "step": 1840
+    },
+    {
+      "epoch": 114.00833333333334,
+      "eval_accuracy": 0.5475852272727273,
+      "eval_loss": 1.9040805101394653,
+      "eval_runtime": 132.319,
+      "eval_samples_per_second": 10.641,
+      "eval_steps_per_second": 0.166,
+      "step": 1840
+    },
+    {
+      "epoch": 115.00520833333333,
+      "grad_norm": 7.5589280128479,
+      "learning_rate": 2.025462962962963e-06,
+      "loss": 0.7465,
+      "step": 1850
+    },
+    {
+      "epoch": 115.00833333333334,
+      "eval_accuracy": 0.5482954545454546,
+      "eval_loss": 1.9005664587020874,
+      "eval_runtime": 129.3051,
+      "eval_samples_per_second": 10.889,
+      "eval_steps_per_second": 0.17,
+      "step": 1856
+    },
+    {
+      "epoch": 116.00208333333333,
+      "grad_norm": 8.857230186462402,
+      "learning_rate": 1.7361111111111112e-06,
+      "loss": 0.5929,
+      "step": 1860
+    },
+    {
+      "epoch": 116.00729166666666,
+      "grad_norm": 5.749704837799072,
+      "learning_rate": 1.4467592592592594e-06,
+      "loss": 0.6935,
+      "step": 1870
+    },
+    {
+      "epoch": 116.00833333333334,
+      "eval_accuracy": 0.5461647727272727,
+      "eval_loss": 1.901587724685669,
+      "eval_runtime": 135.6846,
+      "eval_samples_per_second": 10.377,
+      "eval_steps_per_second": 0.162,
+      "step": 1872
+    },
+    {
+      "epoch": 117.00416666666666,
+      "grad_norm": 6.5324883460998535,
+      "learning_rate": 1.1574074074074074e-06,
+      "loss": 0.7306,
+      "step": 1880
+    },
+    {
+      "epoch": 117.00833333333334,
+      "eval_accuracy": 0.5482954545454546,
+      "eval_loss": 1.9008712768554688,
+      "eval_runtime": 123.3184,
+      "eval_samples_per_second": 11.418,
+      "eval_steps_per_second": 0.178,
+      "step": 1888
+    },
+    {
+      "epoch": 118.00104166666667,
+      "grad_norm": 6.570457458496094,
+      "learning_rate": 8.680555555555556e-07,
+      "loss": 0.6663,
+      "step": 1890
+    },
+    {
+      "epoch": 118.00625,
+      "grad_norm": 4.956265449523926,
+      "learning_rate": 5.787037037037037e-07,
+      "loss": 0.6578,
+      "step": 1900
+    },
+    {
+      "epoch": 118.00833333333334,
+      "eval_accuracy": 0.5482954545454546,
+      "eval_loss": 1.9007748365402222,
+      "eval_runtime": 136.6271,
+      "eval_samples_per_second": 10.305,
+      "eval_steps_per_second": 0.161,
+      "step": 1904
+    },
+    {
+      "epoch": 119.003125,
+      "grad_norm": 7.002548694610596,
+      "learning_rate": 2.8935185185185185e-07,
+      "loss": 0.6235,
+      "step": 1910
+    },
+    {
+      "epoch": 119.00833333333334,
+      "grad_norm": 2.7440524101257324,
+      "learning_rate": 0.0,
+      "loss": 0.6427,
+      "step": 1920
+    },
+    {
+      "epoch": 119.00833333333334,
+      "eval_accuracy": 0.5504261363636364,
+      "eval_loss": 1.9013855457305908,
+      "eval_runtime": 128.5584,
+      "eval_samples_per_second": 10.952,
+      "eval_steps_per_second": 0.171,
+      "step": 1920
+    },
+    {
+      "epoch": 119.00833333333334,
+      "step": 1920,
+      "total_flos": 1.4452753827235627e+20,
+      "train_loss": 1.5160956154266994,
+      "train_runtime": 26874.3406,
+      "train_samples_per_second": 4.572,
+      "train_steps_per_second": 0.071
+    },
+    {
+      "epoch": 119.00833333333334,
+      "eval_accuracy": 0.5539772727272727,
+      "eval_loss": 1.901895523071289,
+      "eval_runtime": 137.4484,
+      "eval_samples_per_second": 10.244,
+      "eval_steps_per_second": 0.16,
+      "step": 1920
+    },
+    {
+      "epoch": 119.00833333333334,
+      "eval_accuracy": 0.5539772727272727,
+      "eval_loss": 1.9018956422805786,
+      "eval_runtime": 133.7632,
+      "eval_samples_per_second": 10.526,
+      "eval_steps_per_second": 0.164,
+      "step": 1920
     }
   ],
   "logging_steps": 10,
-  "max_steps": 832,
+  "max_steps": 1920,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 9223372036854775807,
   "save_steps": 500,
@@ -1174,7 +2477,7 @@
       "attributes": {}
     }
   },
-  "total_flos": 6.1975283578694664e+19,
+  "total_flos": 1.4452753827235627e+20,
   "train_batch_size": 64,
   "trial_name": null,
   "trial_params": null