diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,338 +1,4774 @@
 {
-  "best_metric": 1.5699154138565063,
-  "best_model_checkpoint": "output/headie-one/checkpoint-235",
-  "epoch": 5.0,
-  "global_step": 235,
+  "best_metric": 5.471560001373291,
+  "best_model_checkpoint": "output/headie-one/checkpoint-76",
+  "epoch": 48.0,
+  "global_step": 3648,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.11,
-      "learning_rate": 0.00013340429968430325,
-      "loss": 1.271,
+      "epoch": 0.07,
+      "learning_rate": 0.0001357399755894565,
+      "loss": 0.4785,
       "step": 5
     },
     {
-      "epoch": 0.21,
-      "learning_rate": 0.00012243723783011623,
-      "loss": 1.1922,
+      "epoch": 0.13,
+      "learning_rate": 0.00013142205020853694,
+      "loss": 0.55,
       "step": 10
     },
     {
-      "epoch": 0.32,
-      "learning_rate": 0.00010551244942700729,
-      "loss": 1.0699,
+      "epoch": 0.2,
+      "learning_rate": 0.00012443002200473538,
+      "loss": 0.5739,
       "step": 15
     },
     {
-      "epoch": 0.43,
-      "learning_rate": 8.450286230835261e-05,
-      "loss": 1.0247,
+      "epoch": 0.26,
+      "learning_rate": 0.00011506151581352585,
+      "loss": 0.5448,
       "step": 20
     },
     {
-      "epoch": 0.53,
-      "learning_rate": 6.17334355545868e-05,
-      "loss": 0.9594,
+      "epoch": 0.33,
+      "learning_rate": 0.00010371531436754662,
+      "loss": 0.5863,
       "step": 25
     },
     {
-      "epoch": 0.64,
-      "learning_rate": 3.9723875297416415e-05,
-      "loss": 0.9693,
+      "epoch": 0.39,
+      "learning_rate": 9.08743835874413e-05,
+      "loss": 0.5127,
       "step": 30
     },
     {
-      "epoch": 0.74,
-      "learning_rate": 2.090979946151953e-05,
-      "loss": 1.106,
+      "epoch": 0.46,
+      "learning_rate": 7.708531450507727e-05,
+      "loss": 0.5548,
       "step": 35
     },
     {
-      "epoch": 0.85,
-      "learning_rate": 7.373207860012262e-06,
-      "loss": 1.0108,
+      "epoch": 0.53,
+      "learning_rate": 6.293505690059801e-05,
+      "loss": 0.5604,
       "step": 40
     },
     {
-      "epoch": 0.96,
-      "learning_rate": 6.1208431258209e-07,
-      "loss": 1.0614,
+      "epoch": 0.59,
+      "learning_rate": 4.902593501651139e-05,
+      "loss": 0.5144,
       "step": 45
     },
     {
-      "epoch": 1.0,
-      "eval_loss": 1.6016333103179932,
-      "eval_runtime": 1.5387,
-      "eval_samples_per_second": 43.543,
-      "eval_steps_per_second": 5.849,
-      "step": 47
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 1.3746270344901794e-06,
-      "loss": 1.1841,
+      "epoch": 0.66,
+      "learning_rate": 3.5950008837656755e-05,
+      "loss": 0.6611,
       "step": 50
     },
     {
-      "epoch": 1.17,
-      "learning_rate": 9.5764516627544e-06,
-      "loss": 1.0057,
+      "epoch": 0.72,
+      "learning_rate": 2.4263872282739445e-05,
+      "loss": 0.5217,
       "step": 55
     },
     {
-      "epoch": 1.28,
-      "learning_rate": 2.4309929383065895e-05,
-      "loss": 0.8867,
+      "epoch": 0.79,
+      "learning_rate": 1.4464961055407408e-05,
+      "loss": 0.544,
       "step": 60
     },
     {
-      "epoch": 1.38,
-      "learning_rate": 4.394462678334651e-05,
-      "loss": 1.0264,
+      "epoch": 0.86,
+      "learning_rate": 6.97037864221025e-06,
+      "loss": 0.5656,
       "step": 65
     },
     {
-      "epoch": 1.49,
-      "learning_rate": 6.630773257727362e-05,
-      "loss": 0.9567,
+      "epoch": 0.92,
+      "learning_rate": 2.0991417565619363e-06,
+      "loss": 0.4698,
       "step": 70
     },
     {
-      "epoch": 1.6,
-      "learning_rate": 8.892450484875433e-05,
-      "loss": 0.8565,
+      "epoch": 0.99,
+      "learning_rate": 5.860097463115006e-08,
+      "loss": 0.5957,
       "step": 75
     },
     {
-      "epoch": 1.7,
-      "learning_rate": 0.00010929213048843382,
-      "loss": 1.0029,
+      "epoch": 1.0,
+      "eval_loss": 5.471560001373291,
+      "eval_runtime": 2.3923,
+      "eval_samples_per_second": 42.637,
+      "eval_steps_per_second": 5.434,
+      "step": 76
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.35614586573241e-07,
+      "loss": 0.5587,
       "step": 80
     },
     {
-      "epoch": 1.81,
-      "learning_rate": 0.00012515669103944463,
-      "loss": 1.0274,
+      "epoch": 1.12,
+      "learning_rate": 4.692851359682261e-06,
+      "loss": 0.5065,
       "step": 85
     },
     {
-      "epoch": 1.91,
-      "learning_rate": 0.00013476258540873022,
-      "loss": 0.9839,
+      "epoch": 1.18,
+      "learning_rate": 1.1170379591190527e-05,
+      "loss": 0.4748,
       "step": 90
     },
     {
-      "epoch": 2.0,
-      "eval_loss": 1.6458494663238525,
-      "eval_runtime": 1.5583,
-      "eval_samples_per_second": 42.996,
-      "eval_steps_per_second": 5.776,
-      "step": 94
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 0.00013704680787354832,
-      "loss": 0.9767,
+      "epoch": 1.25,
+      "learning_rate": 2.009247481060283e-05,
+      "loss": 0.408,
       "step": 95
     },
     {
-      "epoch": 2.13,
-      "learning_rate": 0.00013175658222600302,
-      "loss": 1.1312,
+      "epoch": 1.32,
+      "learning_rate": 3.1079356352801514e-05,
+      "loss": 0.3892,
       "step": 100
     },
     {
-      "epoch": 2.23,
-      "learning_rate": 0.00011947733444745008,
-      "loss": 0.7833,
+      "epoch": 1.38,
+      "learning_rate": 4.3663353219321276e-05,
+      "loss": 0.4958,
       "step": 105
     },
     {
-      "epoch": 2.34,
-      "learning_rate": 0.00010156790840580658,
-      "loss": 0.8376,
+      "epoch": 1.45,
+      "learning_rate": 5.7308811106741675e-05,
+      "loss": 0.3676,
       "step": 110
     },
     {
-      "epoch": 2.45,
-      "learning_rate": 8.001019372440257e-05,
-      "loss": 0.9393,
+      "epoch": 1.51,
+      "learning_rate": 7.143489323346855e-05,
+      "loss": 0.4046,
       "step": 115
     },
     {
-      "epoch": 2.55,
-      "learning_rate": 5.71898062755974e-05,
-      "loss": 0.8203,
+      "epoch": 1.58,
+      "learning_rate": 8.54403044178588e-05,
+      "loss": 0.4215,
       "step": 120
     },
     {
-      "epoch": 2.66,
-      "learning_rate": 3.5632091594193825e-05,
-      "loss": 0.9363,
+      "epoch": 1.64,
+      "learning_rate": 9.87288859949296e-05,
+      "loss": 0.552,
       "step": 125
     },
     {
-      "epoch": 2.77,
-      "learning_rate": 1.7722665552550217e-05,
-      "loss": 0.9023,
+      "epoch": 1.71,
+      "learning_rate": 0.00011073499209051121,
+      "loss": 0.4987,
       "step": 130
     },
     {
-      "epoch": 2.87,
-      "learning_rate": 5.443417773996963e-06,
-      "loss": 0.7999,
+      "epoch": 1.78,
+      "learning_rate": 0.00012094756707850673,
+      "loss": 0.5333,
       "step": 135
     },
     {
-      "epoch": 2.98,
-      "learning_rate": 1.5319212645170819e-07,
-      "loss": 0.8643,
+      "epoch": 1.84,
+      "learning_rate": 0.00012893189933276512,
+      "loss": 0.5114,
       "step": 140
     },
     {
-      "epoch": 3.0,
-      "eval_loss": 1.5993356704711914,
-      "eval_runtime": 1.5942,
-      "eval_samples_per_second": 42.027,
-      "eval_steps_per_second": 5.645,
-      "step": 141
-    },
-    {
-      "epoch": 3.09,
-      "learning_rate": 2.4374145912697747e-06,
-      "loss": 0.6501,
+      "epoch": 1.91,
+      "learning_rate": 0.00013434812529663614,
+      "loss": 0.4435,
       "step": 145
     },
     {
-      "epoch": 3.19,
-      "learning_rate": 1.204330896055538e-05,
-      "loss": 0.7052,
+      "epoch": 1.97,
+      "learning_rate": 0.00013696569622025754,
+      "loss": 0.5415,
       "step": 150
     },
     {
-      "epoch": 3.3,
-      "learning_rate": 2.7907869511566212e-05,
-      "loss": 0.9473,
+      "epoch": 2.0,
+      "eval_loss": 5.518349647521973,
+      "eval_runtime": 2.4007,
+      "eval_samples_per_second": 42.487,
+      "eval_steps_per_second": 5.415,
+      "step": 152
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.00013667319176996132,
+      "loss": 0.4488,
       "step": 155
     },
     {
-      "epoch": 3.4,
-      "learning_rate": 4.8275495151245236e-05,
-      "loss": 0.7318,
+      "epoch": 2.11,
+      "learning_rate": 0.00013348306278066356,
+      "loss": 0.4539,
       "step": 160
     },
     {
-      "epoch": 3.51,
-      "learning_rate": 7.08922674227264e-05,
-      "loss": 0.7595,
+      "epoch": 2.17,
+      "learning_rate": 0.00012753110126977412,
+      "loss": 0.3929,
       "step": 165
     },
     {
-      "epoch": 3.62,
-      "learning_rate": 9.325537321665307e-05,
-      "loss": 0.8141,
+      "epoch": 2.24,
+      "learning_rate": 0.00011907066027217684,
+      "loss": 0.4031,
       "step": 170
     },
     {
-      "epoch": 3.72,
-      "learning_rate": 0.00011289007061693377,
-      "loss": 0.7721,
+      "epoch": 2.3,
+      "learning_rate": 0.00010846186953657296,
+      "loss": 0.4346,
       "step": 175
     },
     {
-      "epoch": 3.83,
-      "learning_rate": 0.0001276235483372456,
-      "loss": 0.8703,
+      "epoch": 2.37,
+      "learning_rate": 9.615630613119375e-05,
+      "loss": 0.5194,
       "step": 180
     },
     {
-      "epoch": 3.94,
-      "learning_rate": 0.00013582537296550983,
-      "loss": 0.8111,
+      "epoch": 2.43,
+      "learning_rate": 8.267777247462053e-05,
+      "loss": 0.4893,
       "step": 185
     },
     {
-      "epoch": 4.0,
-      "eval_loss": 1.582382082939148,
-      "eval_runtime": 1.6284,
-      "eval_samples_per_second": 41.144,
-      "eval_steps_per_second": 5.527,
-      "step": 188
-    },
-    {
-      "epoch": 4.04,
-      "learning_rate": 0.00013658791568741795,
-      "loss": 0.8595,
+      "epoch": 2.5,
+      "learning_rate": 6.860000000000001e-05,
+      "loss": 0.4725,
       "step": 190
     },
     {
-      "epoch": 4.15,
-      "learning_rate": 0.00012982679213998795,
-      "loss": 0.6313,
+      "epoch": 2.57,
+      "learning_rate": 5.45222275253795e-05,
+      "loss": 0.4897,
       "step": 195
     },
     {
-      "epoch": 4.26,
-      "learning_rate": 0.00011629020053848043,
-      "loss": 0.8127,
+      "epoch": 2.63,
+      "learning_rate": 4.1043693868806304e-05,
+      "loss": 0.4427,
       "step": 200
     },
     {
-      "epoch": 4.36,
-      "learning_rate": 9.747612470258399e-05,
-      "loss": 0.8002,
+      "epoch": 2.7,
+      "learning_rate": 2.873813046342713e-05,
+      "loss": 0.4814,
       "step": 205
     },
     {
-      "epoch": 4.47,
-      "learning_rate": 7.546656444541318e-05,
-      "loss": 0.7541,
+      "epoch": 2.76,
+      "learning_rate": 1.812933972782316e-05,
+      "loss": 0.4612,
       "step": 210
     },
     {
-      "epoch": 4.57,
-      "learning_rate": 5.269713769164736e-05,
-      "loss": 0.7412,
+      "epoch": 2.83,
+      "learning_rate": 9.668898730225861e-06,
+      "loss": 0.4338,
       "step": 215
     },
     {
-      "epoch": 4.68,
-      "learning_rate": 3.1687550572992684e-05,
-      "loss": 0.6833,
+      "epoch": 2.89,
+      "learning_rate": 3.7169372193364486e-06,
+      "loss": 0.4686,
       "step": 220
     },
     {
-      "epoch": 4.79,
-      "learning_rate": 1.4762762169884062e-05,
-      "loss": 0.7674,
+      "epoch": 2.96,
+      "learning_rate": 5.268082300387043e-07,
+      "loss": 0.4222,
       "step": 225
     },
     {
-      "epoch": 4.89,
-      "learning_rate": 3.795700315696802e-06,
-      "loss": 0.6441,
+      "epoch": 3.0,
+      "eval_loss": 5.570201873779297,
+      "eval_runtime": 2.378,
+      "eval_samples_per_second": 42.893,
+      "eval_steps_per_second": 5.467,
+      "step": 228
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.3430377974244087e-07,
+      "loss": 0.3902,
       "step": 230
     },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.851874703363849e-06,
+      "loss": 0.3359,
+      "step": 235
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 8.268100667234827e-06,
+      "loss": 0.3495,
+      "step": 240
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 1.6252432921493247e-05,
+      "loss": 0.4064,
+      "step": 245
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.6465007909488756e-05,
+      "loss": 0.4271,
+      "step": 250
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.847111400507037e-05,
+      "loss": 0.3916,
+      "step": 255
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 5.175969558214116e-05,
+      "loss": 0.356,
+      "step": 260
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.576510676653141e-05,
+      "loss": 0.3824,
+      "step": 265
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 7.989118889325835e-05,
+      "loss": 0.3627,
+      "step": 270
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 9.353664678067862e-05,
+      "loss": 0.4006,
+      "step": 275
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0001061206436471984,
+      "loss": 0.382,
+      "step": 280
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0001171075251893971,
+      "loss": 0.4378,
+      "step": 285
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.0001260296204088095,
+      "loss": 0.4331,
+      "step": 290
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00013250714864031774,
+      "loss": 0.3896,
+      "step": 295
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.00013626438541342674,
+      "loss": 0.4182,
+      "step": 300
+    },
+    {
+      "epoch": 4.0,
+      "eval_loss": 5.707557678222656,
+      "eval_runtime": 2.3844,
+      "eval_samples_per_second": 42.778,
+      "eval_steps_per_second": 5.452,
+      "step": 304
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.00013714139902536884,
+      "loss": 0.3989,
+      "step": 305
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0001351008582434381,
+      "loss": 0.3894,
+      "step": 310
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.0001302296213577898,
+      "loss": 0.3742,
+      "step": 315
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.00012273503894459262,
+      "loss": 0.3937,
+      "step": 320
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.00011293612771726056,
+      "loss": 0.4215,
+      "step": 325
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00010124999116234326,
+      "loss": 0.3937,
+      "step": 330
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 8.817406498348863e-05,
+      "loss": 0.4438,
+      "step": 335
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 7.426494309940199e-05,
+      "loss": 0.4253,
+      "step": 340
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 6.011468549492274e-05,
+      "loss": 0.3437,
+      "step": 345
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 4.6325616412558686e-05,
+      "loss": 0.3752,
+      "step": 350
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.3484685632453346e-05,
+      "loss": 0.4207,
+      "step": 355
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 2.2138484186474223e-05,
+      "loss": 0.3517,
+      "step": 360
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.2769977995264667e-05,
+      "loss": 0.3324,
+      "step": 365
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.777949791463086e-06,
+      "loss": 0.3575,
+      "step": 370
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.4600244105434905e-06,
+      "loss": 0.3555,
+      "step": 375
+    },
     {
       "epoch": 5.0,
       "learning_rate": 0.0,
-      "loss": 0.761,
-      "step": 235
+      "loss": 0.3544,
+      "step": 380
     },
     {
       "epoch": 5.0,
-      "eval_loss": 1.5699154138565063,
-      "eval_runtime": 1.6286,
-      "eval_samples_per_second": 41.139,
-      "eval_steps_per_second": 5.526,
-      "step": 235
+      "eval_loss": 5.685351371765137,
+      "eval_runtime": 2.3666,
+      "eval_samples_per_second": 43.1,
+      "eval_steps_per_second": 5.493,
+      "step": 380
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 1.4600244105434752e-06,
+      "loss": 0.3192,
+      "step": 385
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 5.777949791463048e-06,
+      "loss": 0.3275,
+      "step": 390
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 1.2769977995264689e-05,
+      "loss": 0.282,
+      "step": 395
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 2.2138484186474162e-05,
+      "loss": 0.2882,
+      "step": 400
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.3484685632453264e-05,
+      "loss": 0.3381,
+      "step": 405
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.632561641255849e-05,
+      "loss": 0.3144,
+      "step": 410
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 6.011468549492265e-05,
+      "loss": 0.2941,
+      "step": 415
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 7.426494309940202e-05,
+      "loss": 0.2993,
+      "step": 420
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 8.817406498348855e-05,
+      "loss": 0.3661,
+      "step": 425
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.00010124999116234329,
+      "loss": 0.3125,
+      "step": 430
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.0001129361277172605,
+      "loss": 0.3826,
+      "step": 435
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00012273503894459265,
+      "loss": 0.3616,
+      "step": 440
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.00013022962135778973,
+      "loss": 0.3224,
+      "step": 445
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.0001351008582434381,
+      "loss": 0.3485,
+      "step": 450
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.00013714139902536884,
+      "loss": 0.3931,
+      "step": 455
+    },
+    {
+      "epoch": 6.0,
+      "eval_loss": 5.795953273773193,
+      "eval_runtime": 2.3912,
+      "eval_samples_per_second": 42.656,
+      "eval_steps_per_second": 5.437,
+      "step": 456
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00013626438541342677,
+      "loss": 0.4352,
+      "step": 460
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00013250714864031774,
+      "loss": 0.3058,
+      "step": 465
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.00012602962040880953,
+      "loss": 0.3144,
+      "step": 470
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00011710752518939715,
+      "loss": 0.2813,
+      "step": 475
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.00010612064364719856,
+      "loss": 0.366,
+      "step": 480
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 9.35366467806787e-05,
+      "loss": 0.357,
+      "step": 485
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 7.989118889325843e-05,
+      "loss": 0.3169,
+      "step": 490
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 6.576510676653137e-05,
+      "loss": 0.304,
+      "step": 495
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 5.1759695582141247e-05,
+      "loss": 0.3576,
+      "step": 500
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.847111400507055e-05,
+      "loss": 0.3524,
+      "step": 505
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 2.6465007909488827e-05,
+      "loss": 0.2983,
+      "step": 510
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 1.6252432921493375e-05,
+      "loss": 0.3261,
+      "step": 515
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 8.268100667234864e-06,
+      "loss": 0.3443,
+      "step": 520
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 2.85187470336391e-06,
+      "loss": 0.3208,
+      "step": 525
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 2.343037797424485e-07,
+      "loss": 0.3161,
+      "step": 530
+    },
+    {
+      "epoch": 7.0,
+      "eval_loss": 5.787503719329834,
+      "eval_runtime": 2.3762,
+      "eval_samples_per_second": 42.925,
+      "eval_steps_per_second": 5.471,
+      "step": 532
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 5.268082300386967e-07,
+      "loss": 0.2751,
+      "step": 535
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.716937219336464e-06,
+      "loss": 0.2778,
+      "step": 540
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 9.668898730225944e-06,
+      "loss": 0.2676,
+      "step": 545
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 1.812933972782302e-05,
+      "loss": 0.2433,
+      "step": 550
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.8738130463426962e-05,
+      "loss": 0.2658,
+      "step": 555
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 4.1043693868806114e-05,
+      "loss": 0.2334,
+      "step": 560
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 5.452222752537942e-05,
+      "loss": 0.2809,
+      "step": 565
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 6.859999999999982e-05,
+      "loss": 0.283,
+      "step": 570
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 8.267777247462046e-05,
+      "loss": 0.2846,
+      "step": 575
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 9.615630613119378e-05,
+      "loss": 0.2926,
+      "step": 580
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00010846186953657294,
+      "loss": 0.2787,
+      "step": 585
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.0001190706602721769,
+      "loss": 0.2776,
+      "step": 590
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.000127531101269774,
+      "loss": 0.2803,
+      "step": 595
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.0001334830627806635,
+      "loss": 0.3028,
+      "step": 600
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.0001366731917699613,
+      "loss": 0.3338,
+      "step": 605
+    },
+    {
+      "epoch": 8.0,
+      "eval_loss": 5.904983043670654,
+      "eval_runtime": 2.3887,
+      "eval_samples_per_second": 42.702,
+      "eval_steps_per_second": 5.442,
+      "step": 608
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00013696569622025754,
+      "loss": 0.3256,
+      "step": 610
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00013434812529663614,
+      "loss": 0.2792,
+      "step": 615
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.0001289318993327653,
+      "loss": 0.2754,
+      "step": 620
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00012094756707850687,
+      "loss": 0.3003,
+      "step": 625
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00011073499209051147,
+      "loss": 0.288,
+      "step": 630
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 9.872888599492977e-05,
+      "loss": 0.2662,
+      "step": 635
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 8.544030441785888e-05,
+      "loss": 0.2787,
+      "step": 640
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 7.143489323346877e-05,
+      "loss": 0.3101,
+      "step": 645
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 5.7308811106741695e-05,
+      "loss": 0.2716,
+      "step": 650
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.366335321932141e-05,
+      "loss": 0.2699,
+      "step": 655
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 3.1079356352801534e-05,
+      "loss": 0.3072,
+      "step": 660
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.0092474810602934e-05,
+      "loss": 0.2632,
+      "step": 665
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 1.1170379591190542e-05,
+      "loss": 0.2761,
+      "step": 670
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 4.692851359682307e-06,
+      "loss": 0.2779,
+      "step": 675
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 9.35614586573241e-07,
+      "loss": 0.2747,
+      "step": 680
+    },
+    {
+      "epoch": 9.0,
+      "eval_loss": 5.883668422698975,
+      "eval_runtime": 2.3812,
+      "eval_samples_per_second": 42.835,
+      "eval_steps_per_second": 5.459,
+      "step": 684
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 5.860097463114244e-08,
+      "loss": 0.2531,
+      "step": 685
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.099141756561944e-06,
+      "loss": 0.2619,
+      "step": 690
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 6.970378642210319e-06,
+      "loss": 0.2333,
+      "step": 695
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 1.4464961055407432e-05,
+      "loss": 0.2043,
+      "step": 700
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.426387228273959e-05,
+      "loss": 0.2374,
+      "step": 705
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 3.5950008837656816e-05,
+      "loss": 0.228,
+      "step": 710
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 4.90259350165111e-05,
+      "loss": 0.2088,
+      "step": 715
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 6.293505690059785e-05,
+      "loss": 0.2676,
+      "step": 720
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 7.708531450507698e-05,
+      "loss": 0.2139,
+      "step": 725
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 9.087438358744116e-05,
+      "loss": 0.2626,
+      "step": 730
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00010371531436754642,
+      "loss": 0.2233,
+      "step": 735
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00011506151581352575,
+      "loss": 0.262,
+      "step": 740
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00012443002200473525,
+      "loss": 0.2799,
+      "step": 745
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00013142205020853692,
+      "loss": 0.224,
+      "step": 750
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00013573997558945648,
+      "loss": 0.2749,
+      "step": 755
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.0001372,
+      "loss": 0.2684,
+      "step": 760
+    },
+    {
+      "epoch": 10.0,
+      "eval_loss": 5.94313907623291,
+      "eval_runtime": 2.3788,
+      "eval_samples_per_second": 42.88,
+      "eval_steps_per_second": 5.465,
+      "step": 760
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.0001357399755894565,
+      "loss": 0.2072,
+      "step": 765
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00013142205020853697,
+      "loss": 0.2447,
+      "step": 770
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.0001244300220047355,
+      "loss": 0.2514,
+      "step": 775
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.00011506151581352587,
+      "loss": 0.2337,
+      "step": 780
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00010371531436754655,
+      "loss": 0.2403,
+      "step": 785
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 9.087438358744108e-05,
+      "loss": 0.2396,
+      "step": 790
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 7.708531450507739e-05,
+      "loss": 0.2401,
+      "step": 795
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 6.293505690059801e-05,
+      "loss": 0.2772,
+      "step": 800
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 4.9025935016511264e-05,
+      "loss": 0.2206,
+      "step": 805
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 3.595000883765696e-05,
+      "loss": 0.2247,
+      "step": 810
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.4263872282739537e-05,
+      "loss": 0.2496,
+      "step": 815
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 1.4464961055407682e-05,
+      "loss": 0.2492,
+      "step": 820
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 6.970378642210395e-06,
+      "loss": 0.2406,
+      "step": 825
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.099141756561982e-06,
+      "loss": 0.2398,
+      "step": 830
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 5.860097463115006e-08,
+      "loss": 0.2381,
+      "step": 835
+    },
+    {
+      "epoch": 11.0,
+      "eval_loss": 5.973106384277344,
+      "eval_runtime": 2.3796,
+      "eval_samples_per_second": 42.864,
+      "eval_steps_per_second": 5.463,
+      "step": 836
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 9.356145865732563e-07,
+      "loss": 0.1709,
+      "step": 840
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 4.6928513596821544e-06,
+      "loss": 0.1884,
+      "step": 845
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 1.117037959119045e-05,
+      "loss": 0.2089,
+      "step": 850
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.0092474810602812e-05,
+      "loss": 0.1931,
+      "step": 855
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 3.1079356352801595e-05,
+      "loss": 0.1894,
+      "step": 860
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 4.366335321932103e-05,
+      "loss": 0.2227,
+      "step": 865
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 5.7308811106741526e-05,
+      "loss": 0.1859,
+      "step": 870
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 7.143489323346859e-05,
+      "loss": 0.2015,
+      "step": 875
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 8.544030441785894e-05,
+      "loss": 0.194,
+      "step": 880
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 9.872888599492985e-05,
+      "loss": 0.2035,
+      "step": 885
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00011073499209051113,
+      "loss": 0.2189,
+      "step": 890
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00012094756707850674,
+      "loss": 0.2264,
+      "step": 895
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012893189933276523,
+      "loss": 0.2029,
+      "step": 900
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00013434812529663609,
+      "loss": 0.2352,
+      "step": 905
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00013696569622025754,
+      "loss": 0.2269,
+      "step": 910
+    },
+    {
+      "epoch": 12.0,
+      "eval_loss": 6.038574695587158,
+      "eval_runtime": 2.3927,
+      "eval_samples_per_second": 42.629,
+      "eval_steps_per_second": 5.433,
+      "step": 912
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00013667319176996134,
+      "loss": 0.2264,
+      "step": 915
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.0001334830627806636,
+      "loss": 0.1967,
+      "step": 920
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0001275311012697742,
+      "loss": 0.2084,
+      "step": 925
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00011907066027217684,
+      "loss": 0.214,
+      "step": 930
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00010846186953657327,
+      "loss": 0.22,
+      "step": 935
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 9.615630613119393e-05,
+      "loss": 0.198,
+      "step": 940
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 8.267777247462063e-05,
+      "loss": 0.23,
+      "step": 945
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 6.859999999999999e-05,
+      "loss": 0.2105,
+      "step": 950
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 5.452222752537934e-05,
+      "loss": 0.2301,
+      "step": 955
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 4.104369386880649e-05,
+      "loss": 0.2116,
+      "step": 960
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.8738130463427098e-05,
+      "loss": 0.2089,
+      "step": 965
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.8129339727823136e-05,
+      "loss": 0.2142,
+      "step": 970
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 9.668898730225785e-06,
+      "loss": 0.2095,
+      "step": 975
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 3.7169372193365172e-06,
+      "loss": 0.2194,
+      "step": 980
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 5.268082300387119e-07,
+      "loss": 0.1903,
+      "step": 985
+    },
+    {
+      "epoch": 13.0,
+      "eval_loss": 6.056723594665527,
+      "eval_runtime": 2.3764,
+      "eval_samples_per_second": 42.923,
+      "eval_steps_per_second": 5.471,
+      "step": 988
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.343037797424561e-07,
+      "loss": 0.1768,
+      "step": 990
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.851874703363925e-06,
+      "loss": 0.1757,
+      "step": 995
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 8.268100667234788e-06,
+      "loss": 0.1981,
+      "step": 1000
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.625243292149296e-05,
+      "loss": 0.1515,
+      "step": 1005
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 2.6465007909488506e-05,
+      "loss": 0.1837,
+      "step": 1010
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 3.847111400507018e-05,
+      "loss": 0.1692,
+      "step": 1015
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 5.1759695582141084e-05,
+      "loss": 0.1552,
+      "step": 1020
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 6.576510676653144e-05,
+      "loss": 0.1596,
+      "step": 1025
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 7.989118889325803e-05,
+      "loss": 0.1919,
+      "step": 1030
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 9.353664678067856e-05,
+      "loss": 0.1759,
+      "step": 1035
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 0.00010612064364719844,
+      "loss": 0.1885,
+      "step": 1040
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 0.00011710752518939722,
+      "loss": 0.204,
+      "step": 1045
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 0.0001260296204088093,
+      "loss": 0.196,
+      "step": 1050
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 0.0001325071486403177,
+      "loss": 0.1722,
+      "step": 1055
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 0.00013626438541342674,
+      "loss": 0.2164,
+      "step": 1060
+    },
+    {
+      "epoch": 14.0,
+      "eval_loss": 6.018917083740234,
+      "eval_runtime": 2.3894,
+      "eval_samples_per_second": 42.688,
+      "eval_steps_per_second": 5.441,
+      "step": 1064
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 0.00013714139902536884,
+      "loss": 0.1849,
+      "step": 1065
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 0.00013510085824343813,
+      "loss": 0.1751,
+      "step": 1070
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 0.0001302296213577898,
+      "loss": 0.185,
+      "step": 1075
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 0.0001227350389445926,
+      "loss": 0.1819,
+      "step": 1080
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 0.00011293612771726044,
+      "loss": 0.1962,
+      "step": 1085
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 0.00010124999116234302,
+      "loss": 0.1865,
+      "step": 1090
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.81740649834887e-05,
+      "loss": 0.188,
+      "step": 1095
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 7.426494309940244e-05,
+      "loss": 0.2,
+      "step": 1100
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 6.011468549492306e-05,
+      "loss": 0.1769,
+      "step": 1105
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 4.632561641255889e-05,
+      "loss": 0.1978,
+      "step": 1110
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 3.3484685632453414e-05,
+      "loss": 0.1777,
+      "step": 1115
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.2138484186474457e-05,
+      "loss": 0.2118,
+      "step": 1120
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.2769977995264789e-05,
+      "loss": 0.182,
+      "step": 1125
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 5.777949791463116e-06,
+      "loss": 0.1877,
+      "step": 1130
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.4600244105434828e-06,
+      "loss": 0.1746,
+      "step": 1135
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 0.0,
+      "loss": 0.1651,
+      "step": 1140
+    },
+    {
+      "epoch": 15.0,
+      "eval_loss": 6.139342784881592,
+      "eval_runtime": 2.3739,
+      "eval_samples_per_second": 42.966,
+      "eval_steps_per_second": 5.476,
+      "step": 1140
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.460024410543437e-06,
+      "loss": 0.1611,
+      "step": 1145
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 5.777949791463017e-06,
+      "loss": 0.1537,
+      "step": 1150
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.2769977995264636e-05,
+      "loss": 0.1527,
+      "step": 1155
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 2.2138484186474274e-05,
+      "loss": 0.1497,
+      "step": 1160
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 3.34846856324532e-05,
+      "loss": 0.1454,
+      "step": 1165
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 4.6325616412558645e-05,
+      "loss": 0.1285,
+      "step": 1170
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 6.0114685494922816e-05,
+      "loss": 0.1492,
+      "step": 1175
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 7.42649430994022e-05,
+      "loss": 0.1309,
+      "step": 1180
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 8.817406498348845e-05,
+      "loss": 0.1646,
+      "step": 1185
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 0.00010124999116234279,
+      "loss": 0.1558,
+      "step": 1190
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 0.00011293612771726026,
+      "loss": 0.1872,
+      "step": 1195
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 0.00012273503894459243,
+      "loss": 0.1446,
+      "step": 1200
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 0.0001302296213577897,
+      "loss": 0.1797,
+      "step": 1205
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 0.00013510085824343807,
+      "loss": 0.1899,
+      "step": 1210
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 0.00013714139902536884,
+      "loss": 0.1733,
+      "step": 1215
+    },
+    {
+      "epoch": 16.0,
+      "eval_loss": 6.194965362548828,
+      "eval_runtime": 2.3865,
+      "eval_samples_per_second": 42.741,
+      "eval_steps_per_second": 5.447,
+      "step": 1216
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 0.00013626438541342671,
+      "loss": 0.1955,
+      "step": 1220
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 0.00013250714864031777,
+      "loss": 0.1686,
+      "step": 1225
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 0.00012602962040880945,
+      "loss": 0.1755,
+      "step": 1230
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 0.00011710752518939739,
+      "loss": 0.1859,
+      "step": 1235
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 0.00010612064364719905,
+      "loss": 0.1609,
+      "step": 1240
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.353664678067879e-05,
+      "loss": 0.1691,
+      "step": 1245
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 7.989118889325875e-05,
+      "loss": 0.1567,
+      "step": 1250
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 6.57651067665317e-05,
+      "loss": 0.1615,
+      "step": 1255
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.17596955821418e-05,
+      "loss": 0.1546,
+      "step": 1260
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 3.847111400507041e-05,
+      "loss": 0.161,
+      "step": 1265
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 2.6465007909489085e-05,
+      "loss": 0.1455,
+      "step": 1270
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.6252432921493118e-05,
+      "loss": 0.1649,
+      "step": 1275
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 8.268100667234903e-06,
+      "loss": 0.1703,
+      "step": 1280
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 2.851874703363864e-06,
+      "loss": 0.1799,
+      "step": 1285
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 2.3430377974247896e-07,
+      "loss": 0.1598,
+      "step": 1290
+    },
+    {
+      "epoch": 17.0,
+      "eval_loss": 6.190419673919678,
+      "eval_runtime": 2.3796,
+      "eval_samples_per_second": 42.864,
+      "eval_steps_per_second": 5.463,
+      "step": 1292
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 5.268082300387423e-07,
+      "loss": 0.1591,
+      "step": 1295
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 3.7169372193364334e-06,
+      "loss": 0.1468,
+      "step": 1300
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 9.668898730225899e-06,
+      "loss": 0.1398,
+      "step": 1305
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.8129339727822966e-05,
+      "loss": 0.12,
+      "step": 1310
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 2.8738130463427288e-05,
+      "loss": 0.1334,
+      "step": 1315
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.104369386880626e-05,
+      "loss": 0.1346,
+      "step": 1320
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 5.452222752537957e-05,
+      "loss": 0.144,
+      "step": 1325
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.859999999999973e-05,
+      "loss": 0.1241,
+      "step": 1330
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 8.26777724746199e-05,
+      "loss": 0.1374,
+      "step": 1335
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 9.61563061311937e-05,
+      "loss": 0.1446,
+      "step": 1340
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 0.00010846186953657267,
+      "loss": 0.1361,
+      "step": 1345
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 0.00011907066027217668,
+      "loss": 0.1447,
+      "step": 1350
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 0.00012753110126977383,
+      "loss": 0.1544,
+      "step": 1355
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 0.00013348306278066356,
+      "loss": 0.1502,
+      "step": 1360
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 0.00013667319176996126,
+      "loss": 0.1565,
+      "step": 1365
+    },
+    {
+      "epoch": 18.0,
+      "eval_loss": 6.231907844543457,
+      "eval_runtime": 2.391,
+      "eval_samples_per_second": 42.66,
+      "eval_steps_per_second": 5.437,
+      "step": 1368
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 0.00013696569622025757,
+      "loss": 0.1556,
+      "step": 1370
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 0.0001343481252966363,
+      "loss": 0.1436,
+      "step": 1375
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 0.00012893189933276512,
+      "loss": 0.1651,
+      "step": 1380
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 0.00012094756707850692,
+      "loss": 0.1512,
+      "step": 1385
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 0.00011073499209051094,
+      "loss": 0.1571,
+      "step": 1390
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 9.872888599493006e-05,
+      "loss": 0.1541,
+      "step": 1395
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 8.544030441785871e-05,
+      "loss": 0.1489,
+      "step": 1400
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 7.143489323346883e-05,
+      "loss": 0.1501,
+      "step": 1405
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 5.73088111067413e-05,
+      "loss": 0.1598,
+      "step": 1410
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 4.366335321932126e-05,
+      "loss": 0.1461,
+      "step": 1415
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.10793563528014e-05,
+      "loss": 0.1573,
+      "step": 1420
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 2.0092474810602995e-05,
+      "loss": 0.1337,
+      "step": 1425
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.1170379591190853e-05,
+      "loss": 0.1445,
+      "step": 1430
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 4.692851359682246e-06,
+      "loss": 0.1426,
+      "step": 1435
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 9.356145865732944e-07,
+      "loss": 0.1351,
+      "step": 1440
+    },
+    {
+      "epoch": 19.0,
+      "eval_loss": 6.250270843505859,
+      "eval_runtime": 2.3794,
+      "eval_samples_per_second": 42.868,
+      "eval_steps_per_second": 5.464,
+      "step": 1444
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 5.860097463114244e-08,
+      "loss": 0.1946,
+      "step": 1445
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.099141756561799e-06,
+      "loss": 0.1179,
+      "step": 1450
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 6.970378642210281e-06,
+      "loss": 0.1372,
+      "step": 1455
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.4464961055407225e-05,
+      "loss": 0.1174,
+      "step": 1460
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 2.426387228273934e-05,
+      "loss": 0.1142,
+      "step": 1465
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 3.5950008837656314e-05,
+      "loss": 0.1221,
+      "step": 1470
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 4.902593501651148e-05,
+      "loss": 0.11,
+      "step": 1475
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 6.293505690059777e-05,
+      "loss": 0.1134,
+      "step": 1480
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.708531450507714e-05,
+      "loss": 0.1229,
+      "step": 1485
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.087438358744086e-05,
+      "loss": 0.1244,
+      "step": 1490
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 0.00010371531436754677,
+      "loss": 0.1288,
+      "step": 1495
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 0.0001150615158135257,
+      "loss": 0.1384,
+      "step": 1500
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 0.00012443002200473535,
+      "loss": 0.1294,
+      "step": 1505
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 0.00013142205020853678,
+      "loss": 0.1234,
+      "step": 1510
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 0.00013573997558945656,
+      "loss": 0.1569,
+      "step": 1515
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 0.0001372,
+      "loss": 0.1293,
+      "step": 1520
+    },
+    {
+      "epoch": 20.0,
+      "eval_loss": 6.285115718841553,
+      "eval_runtime": 2.3913,
+      "eval_samples_per_second": 42.655,
+      "eval_steps_per_second": 5.436,
+      "step": 1520
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 0.00013573997558945661,
+      "loss": 0.1331,
+      "step": 1525
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 0.00013142205020853692,
+      "loss": 0.1233,
+      "step": 1530
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 0.00012443002200473552,
+      "loss": 0.1277,
+      "step": 1535
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 0.00011506151581352594,
+      "loss": 0.1432,
+      "step": 1540
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 0.00010371531436754704,
+      "loss": 0.1511,
+      "step": 1545
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 9.087438358744162e-05,
+      "loss": 0.1448,
+      "step": 1550
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 7.708531450507796e-05,
+      "loss": 0.1498,
+      "step": 1555
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 6.29350569005981e-05,
+      "loss": 0.1375,
+      "step": 1560
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 4.9025935016511806e-05,
+      "loss": 0.1355,
+      "step": 1565
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 3.5950008837656606e-05,
+      "loss": 0.132,
+      "step": 1570
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 2.4263872282739598e-05,
+      "loss": 0.1243,
+      "step": 1575
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.4464961055407135e-05,
+      "loss": 0.1311,
+      "step": 1580
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 6.97037864221022e-06,
+      "loss": 0.1217,
+      "step": 1585
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 2.099141756562005e-06,
+      "loss": 0.1347,
+      "step": 1590
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 5.860097463117291e-08,
+      "loss": 0.1239,
+      "step": 1595
+    },
+    {
+      "epoch": 21.0,
+      "eval_loss": 6.295395374298096,
+      "eval_runtime": 2.3772,
+      "eval_samples_per_second": 42.908,
+      "eval_steps_per_second": 5.469,
+      "step": 1596
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 9.35614586573241e-07,
+      "loss": 0.1241,
+      "step": 1600
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 4.692851359682124e-06,
+      "loss": 0.1257,
+      "step": 1605
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.117037959119067e-05,
+      "loss": 0.1089,
+      "step": 1610
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 2.009247481060276e-05,
+      "loss": 0.1013,
+      "step": 1615
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 3.107935635280112e-05,
+      "loss": 0.1159,
+      "step": 1620
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 4.3663353219321405e-05,
+      "loss": 0.1031,
+      "step": 1625
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 5.7308811106741445e-05,
+      "loss": 0.1039,
+      "step": 1630
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 7.143489323346802e-05,
+      "loss": 0.1228,
+      "step": 1635
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 8.544030441785793e-05,
+      "loss": 0.1095,
+      "step": 1640
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 9.872888599492933e-05,
+      "loss": 0.1061,
+      "step": 1645
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 0.00011073499209051068,
+      "loss": 0.1078,
+      "step": 1650
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 0.0001209475670785067,
+      "loss": 0.1246,
+      "step": 1655
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 0.00012893189933276496,
+      "loss": 0.1134,
+      "step": 1660
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 0.0001343481252966362,
+      "loss": 0.1251,
+      "step": 1665
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 0.00013696569622025754,
+      "loss": 0.1355,
+      "step": 1670
+    },
+    {
+      "epoch": 22.0,
+      "eval_loss": 6.330929756164551,
+      "eval_runtime": 2.3944,
+      "eval_samples_per_second": 42.599,
+      "eval_steps_per_second": 5.429,
+      "step": 1672
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 0.00013667319176996124,
+      "loss": 0.1321,
+      "step": 1675
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 0.0001334830627806635,
+      "loss": 0.1258,
+      "step": 1680
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 0.00012753110126977426,
+      "loss": 0.1376,
+      "step": 1685
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 0.00011907066027217723,
+      "loss": 0.1364,
+      "step": 1690
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 0.00010846186953657294,
+      "loss": 0.1255,
+      "step": 1695
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 9.615630613119401e-05,
+      "loss": 0.1304,
+      "step": 1700
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 8.267777247462023e-05,
+      "loss": 0.1214,
+      "step": 1705
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 6.860000000000005e-05,
+      "loss": 0.1299,
+      "step": 1710
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 5.452222752537991e-05,
+      "loss": 0.1184,
+      "step": 1715
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 4.104369386880612e-05,
+      "loss": 0.1361,
+      "step": 1720
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 2.8738130463427166e-05,
+      "loss": 0.1135,
+      "step": 1725
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 1.8129339727823525e-05,
+      "loss": 0.1117,
+      "step": 1730
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 9.668898730226326e-06,
+      "loss": 0.1102,
+      "step": 1735
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 3.7169372193365477e-06,
+      "loss": 0.1149,
+      "step": 1740
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 5.26808230038788e-07,
+      "loss": 0.0961,
+      "step": 1745
+    },
+    {
+      "epoch": 23.0,
+      "eval_loss": 6.352334022521973,
+      "eval_runtime": 2.3763,
+      "eval_samples_per_second": 42.925,
+      "eval_steps_per_second": 5.471,
+      "step": 1748
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 2.343037797424485e-07,
+      "loss": 0.0998,
+      "step": 1750
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 2.851874703363765e-06,
+      "loss": 0.0979,
+      "step": 1755
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 8.26810066723498e-06,
+      "loss": 0.0952,
+      "step": 1760
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 1.6252432921493216e-05,
+      "loss": 0.103,
+      "step": 1765
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 2.6465007909489207e-05,
+      "loss": 0.0907,
+      "step": 1770
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 3.847111400506967e-05,
+      "loss": 0.0932,
+      "step": 1775
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 5.1759695582141e-05,
+      "loss": 0.0952,
+      "step": 1780
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 6.576510676653087e-05,
+      "loss": 0.0902,
+      "step": 1785
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 7.989118889325843e-05,
+      "loss": 0.0934,
+      "step": 1790
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 9.353664678067846e-05,
+      "loss": 0.1085,
+      "step": 1795
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 0.00010612064364719878,
+      "loss": 0.1076,
+      "step": 1800
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 0.00011710752518939715,
+      "loss": 0.1122,
+      "step": 1805
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 0.00012602962040880926,
+      "loss": 0.1105,
+      "step": 1810
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 0.00013250714864031782,
+      "loss": 0.1049,
+      "step": 1815
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 0.00013626438541342674,
+      "loss": 0.1191,
+      "step": 1820
+    },
+    {
+      "epoch": 24.0,
+      "eval_loss": 6.427064418792725,
+      "eval_runtime": 2.3955,
+      "eval_samples_per_second": 42.579,
+      "eval_steps_per_second": 5.427,
+      "step": 1824
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 0.00013714139902536887,
+      "loss": 0.1256,
+      "step": 1825
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 0.0001351008582434383,
+      "loss": 0.1109,
+      "step": 1830
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 0.00013022962135778984,
+      "loss": 0.1139,
+      "step": 1835
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 0.00012273503894459295,
+      "loss": 0.1074,
+      "step": 1840
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 0.0001129361277172605,
+      "loss": 0.1146,
+      "step": 1845
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 0.00010124999116234352,
+      "loss": 0.1217,
+      "step": 1850
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 8.817406498348832e-05,
+      "loss": 0.1122,
+      "step": 1855
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 7.426494309940203e-05,
+      "loss": 0.1164,
+      "step": 1860
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 6.011468549492314e-05,
+      "loss": 0.1218,
+      "step": 1865
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 4.632561641255942e-05,
+      "loss": 0.1065,
+      "step": 1870
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 3.348468563245349e-05,
+      "loss": 0.1072,
+      "step": 1875
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 2.2138484186474528e-05,
+      "loss": 0.1088,
+      "step": 1880
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 1.2769977995264552e-05,
+      "loss": 0.1169,
+      "step": 1885
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 5.777949791463154e-06,
+      "loss": 0.1083,
+      "step": 1890
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 1.460024410543399e-06,
+      "loss": 0.0946,
+      "step": 1895
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 0.0,
+      "loss": 0.1189,
+      "step": 1900
+    },
+    {
+      "epoch": 25.0,
+      "eval_loss": 6.392660617828369,
+      "eval_runtime": 2.3742,
+      "eval_samples_per_second": 42.962,
+      "eval_steps_per_second": 5.476,
+      "step": 1900
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 1.4600244105434142e-06,
+      "loss": 0.1014,
+      "step": 1905
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 5.777949791463177e-06,
+      "loss": 0.0827,
+      "step": 1910
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 1.276997799526459e-05,
+      "loss": 0.092,
+      "step": 1915
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 2.2138484186473857e-05,
+      "loss": 0.0932,
+      "step": 1920
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 3.34846856324527e-05,
+      "loss": 0.0958,
+      "step": 1925
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 4.6325616412558564e-05,
+      "loss": 0.0902,
+      "step": 1930
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 6.011468549492225e-05,
+      "loss": 0.0864,
+      "step": 1935
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 7.42649430994021e-05,
+      "loss": 0.0765,
+      "step": 1940
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 8.817406498348838e-05,
+      "loss": 0.0921,
+      "step": 1945
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 0.00010124999116234357,
+      "loss": 0.0898,
+      "step": 1950
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 0.00011293612771726056,
+      "loss": 0.0909,
+      "step": 1955
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 0.0001227350389445924,
+      "loss": 0.0994,
+      "step": 1960
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 0.00013022962135778946,
+      "loss": 0.1071,
+      "step": 1965
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 0.00013510085824343805,
+      "loss": 0.1047,
+      "step": 1970
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 0.00013714139902536884,
+      "loss": 0.1126,
+      "step": 1975
+    },
+    {
+      "epoch": 26.0,
+      "eval_loss": 6.413173198699951,
+      "eval_runtime": 2.384,
+      "eval_samples_per_second": 42.785,
+      "eval_steps_per_second": 5.453,
+      "step": 1976
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 0.00013626438541342674,
+      "loss": 0.0971,
+      "step": 1980
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 0.0001325071486403178,
+      "loss": 0.0972,
+      "step": 1985
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 0.00012602962040880923,
+      "loss": 0.1153,
+      "step": 1990
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 0.0001171075251893971,
+      "loss": 0.1152,
+      "step": 1995
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 0.00010612064364719871,
+      "loss": 0.1087,
+      "step": 2000
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 9.353664678067931e-05,
+      "loss": 0.112,
+      "step": 2005
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 7.989118889325931e-05,
+      "loss": 0.1073,
+      "step": 2010
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 6.576510676653178e-05,
+      "loss": 0.103,
+      "step": 2015
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 5.1759695582141884e-05,
+      "loss": 0.0993,
+      "step": 2020
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 3.8471114005070485e-05,
+      "loss": 0.1074,
+      "step": 2025
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 2.6465007909489153e-05,
+      "loss": 0.0895,
+      "step": 2030
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 1.6252432921493172e-05,
+      "loss": 0.1028,
+      "step": 2035
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 8.268100667234949e-06,
+      "loss": 0.0976,
+      "step": 2040
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 2.851874703363742e-06,
+      "loss": 0.0956,
+      "step": 2045
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 2.3430377974244087e-07,
+      "loss": 0.0928,
+      "step": 2050
+    },
+    {
+      "epoch": 27.0,
+      "eval_loss": 6.435521602630615,
+      "eval_runtime": 2.3807,
+      "eval_samples_per_second": 42.844,
+      "eval_steps_per_second": 5.461,
+      "step": 2052
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 5.268082300386738e-07,
+      "loss": 0.0919,
+      "step": 2055
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.716937219336251e-06,
+      "loss": 0.0829,
+      "step": 2060
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 9.668898730225861e-06,
+      "loss": 0.0813,
+      "step": 2065
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 1.8129339727822905e-05,
+      "loss": 0.0878,
+      "step": 2070
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 2.873813046342722e-05,
+      "loss": 0.0754,
+      "step": 2075
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 4.104369386880618e-05,
+      "loss": 0.0738,
+      "step": 2080
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 5.452222752537997e-05,
+      "loss": 0.077,
+      "step": 2085
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 6.860000000000014e-05,
+      "loss": 0.0792,
+      "step": 2090
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 8.267777247462029e-05,
+      "loss": 0.0824,
+      "step": 2095
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 9.615630613119318e-05,
+      "loss": 0.08,
+      "step": 2100
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 0.00010846186953657221,
+      "loss": 0.0833,
+      "step": 2105
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 0.00011907066027217662,
+      "loss": 0.0927,
+      "step": 2110
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 0.0001275311012697738,
+      "loss": 0.0996,
+      "step": 2115
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 0.00013348306278066353,
+      "loss": 0.0852,
+      "step": 2120
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 0.00013667319176996126,
+      "loss": 0.1059,
+      "step": 2125
+    },
+    {
+      "epoch": 28.0,
+      "eval_loss": 6.499926567077637,
+      "eval_runtime": 2.3706,
+      "eval_samples_per_second": 43.027,
+      "eval_steps_per_second": 5.484,
+      "step": 2128
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 0.00013696569622025754,
+      "loss": 0.09,
+      "step": 2130
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 0.00013434812529663617,
+      "loss": 0.0978,
+      "step": 2135
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 0.0001289318993327654,
+      "loss": 0.0895,
+      "step": 2140
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 0.0001209475670785073,
+      "loss": 0.0973,
+      "step": 2145
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 0.0001107349920905114,
+      "loss": 0.1049,
+      "step": 2150
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 9.872888599493015e-05,
+      "loss": 0.0981,
+      "step": 2155
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 8.544030441785881e-05,
+      "loss": 0.0997,
+      "step": 2160
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 7.143489323346893e-05,
+      "loss": 0.0941,
+      "step": 2165
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 5.730881110674138e-05,
+      "loss": 0.0977,
+      "step": 2170
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 4.366335321932134e-05,
+      "loss": 0.0927,
+      "step": 2175
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 3.1079356352801066e-05,
+      "loss": 0.0978,
+      "step": 2180
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 2.0092474810602707e-05,
+      "loss": 0.0931,
+      "step": 2185
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.1170379591190633e-05,
+      "loss": 0.0927,
+      "step": 2190
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 4.692851359682452e-06,
+      "loss": 0.078,
+      "step": 2195
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 9.356145865733934e-07,
+      "loss": 0.0822,
+      "step": 2200
+    },
+    {
+      "epoch": 29.0,
+      "eval_loss": 6.48267936706543,
+      "eval_runtime": 2.3919,
+      "eval_samples_per_second": 42.644,
+      "eval_steps_per_second": 5.435,
+      "step": 2204
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 5.8600974631134825e-08,
+      "loss": 0.0864,
+      "step": 2205
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 2.099141756561784e-06,
+      "loss": 0.0784,
+      "step": 2210
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 6.97037864221025e-06,
+      "loss": 0.0736,
+      "step": 2215
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.446496105540718e-05,
+      "loss": 0.0811,
+      "step": 2220
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 2.4263872282739652e-05,
+      "loss": 0.0735,
+      "step": 2225
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 3.595000883765667e-05,
+      "loss": 0.0773,
+      "step": 2230
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 4.902593501651094e-05,
+      "loss": 0.0769,
+      "step": 2235
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 6.29350569005972e-05,
+      "loss": 0.0734,
+      "step": 2240
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 7.708531450507706e-05,
+      "loss": 0.0796,
+      "step": 2245
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 9.087438358744077e-05,
+      "loss": 0.0719,
+      "step": 2250
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 0.00010371531436754669,
+      "loss": 0.072,
+      "step": 2255
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 0.00011506151581352563,
+      "loss": 0.0799,
+      "step": 2260
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 0.00012443002200473557,
+      "loss": 0.0839,
+      "step": 2265
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 0.00013142205020853694,
+      "loss": 0.0816,
+      "step": 2270
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 0.00013573997558945664,
+      "loss": 0.0891,
+      "step": 2275
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 0.0001372,
+      "loss": 0.1024,
+      "step": 2280
+    },
+    {
+      "epoch": 30.0,
+      "eval_loss": 6.566282749176025,
+      "eval_runtime": 2.3662,
+      "eval_samples_per_second": 43.107,
+      "eval_steps_per_second": 5.494,
+      "step": 2280
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 0.00013573997558945675,
+      "loss": 0.0863,
+      "step": 2285
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 0.00013142205020853713,
+      "loss": 0.0869,
+      "step": 2290
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 0.00012443002200473587,
+      "loss": 0.0866,
+      "step": 2295
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 0.000115061515813526,
+      "loss": 0.087,
+      "step": 2300
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 0.00010371531436754714,
+      "loss": 0.08,
+      "step": 2305
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 9.087438358744125e-05,
+      "loss": 0.0955,
+      "step": 2310
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 7.708531450507755e-05,
+      "loss": 0.0862,
+      "step": 2315
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 6.29350569005977e-05,
+      "loss": 0.0974,
+      "step": 2320
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 4.902593501651142e-05,
+      "loss": 0.0996,
+      "step": 2325
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 3.5950008837657114e-05,
+      "loss": 0.082,
+      "step": 2330
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 2.4263872282740035e-05,
+      "loss": 0.0779,
+      "step": 2335
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 1.4464961055407484e-05,
+      "loss": 0.0959,
+      "step": 2340
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 6.970378642210471e-06,
+      "loss": 0.0768,
+      "step": 2345
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 2.0991417565619058e-06,
+      "loss": 0.0897,
+      "step": 2350
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 5.8600974631157675e-08,
+      "loss": 0.0862,
+      "step": 2355
+    },
+    {
+      "epoch": 31.0,
+      "eval_loss": 6.527225494384766,
+      "eval_runtime": 2.3894,
+      "eval_samples_per_second": 42.689,
+      "eval_steps_per_second": 5.441,
+      "step": 2356
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 9.356145865733097e-07,
+      "loss": 0.0767,
+      "step": 2360
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 4.692851359682269e-06,
+      "loss": 0.0717,
+      "step": 2365
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 1.1170379591190359e-05,
+      "loss": 0.0769,
+      "step": 2370
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 2.0092474810602348e-05,
+      "loss": 0.0754,
+      "step": 2375
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 3.107935635280064e-05,
+      "loss": 0.0707,
+      "step": 2380
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 4.3663353219320876e-05,
+      "loss": 0.0739,
+      "step": 2385
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 5.730881110674088e-05,
+      "loss": 0.0702,
+      "step": 2390
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 7.143489323346843e-05,
+      "loss": 0.0686,
+      "step": 2395
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 8.544030441785831e-05,
+      "loss": 0.0702,
+      "step": 2400
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 9.872888599492969e-05,
+      "loss": 0.0638,
+      "step": 2405
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 0.00011073499209051101,
+      "loss": 0.0761,
+      "step": 2410
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 0.00012094756707850696,
+      "loss": 0.0735,
+      "step": 2415
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 0.00012893189933276515,
+      "loss": 0.0779,
+      "step": 2420
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 0.00013434812529663603,
+      "loss": 0.0811,
+      "step": 2425
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 0.00013696569622025749,
+      "loss": 0.0808,
+      "step": 2430
+    },
+    {
+      "epoch": 32.0,
+      "eval_loss": 6.547709941864014,
+      "eval_runtime": 2.3835,
+      "eval_samples_per_second": 42.795,
+      "eval_steps_per_second": 5.454,
+      "step": 2432
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 0.00013667319176996132,
+      "loss": 0.0711,
+      "step": 2435
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 0.00013348306278066337,
+      "loss": 0.0734,
+      "step": 2440
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 0.00012753110126977453,
+      "loss": 0.0796,
+      "step": 2445
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 0.00011907066027217696,
+      "loss": 0.0897,
+      "step": 2450
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 0.0001084618695365726,
+      "loss": 0.0848,
+      "step": 2455
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 9.615630613119364e-05,
+      "loss": 0.0986,
+      "step": 2460
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 8.267777247462079e-05,
+      "loss": 0.0795,
+      "step": 2465
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 6.860000000000064e-05,
+      "loss": 0.0877,
+      "step": 2470
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 5.452222752537951e-05,
+      "loss": 0.0807,
+      "step": 2475
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 4.104369386880754e-05,
+      "loss": 0.0848,
+      "step": 2480
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 2.8738130463427633e-05,
+      "loss": 0.079,
+      "step": 2485
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 1.812933972782325e-05,
+      "loss": 0.0741,
+      "step": 2490
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 9.668898730225618e-06,
+      "loss": 0.084,
+      "step": 2495
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 3.7169372193367307e-06,
+      "loss": 0.0808,
+      "step": 2500
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 5.268082300387347e-07,
+      "loss": 0.0737,
+      "step": 2505
+    },
+    {
+      "epoch": 33.0,
+      "eval_loss": 6.572361469268799,
+      "eval_runtime": 2.376,
+      "eval_samples_per_second": 42.929,
+      "eval_steps_per_second": 5.471,
+      "step": 2508
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 2.343037797424028e-07,
+      "loss": 0.0638,
+      "step": 2510
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 2.8518747033638794e-06,
+      "loss": 0.0645,
+      "step": 2515
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 8.26810066723424e-06,
+      "loss": 0.0724,
+      "step": 2520
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 1.6252432921492843e-05,
+      "loss": 0.0662,
+      "step": 2525
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 2.6465007909488756e-05,
+      "loss": 0.0643,
+      "step": 2530
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 3.847111400507091e-05,
+      "loss": 0.0656,
+      "step": 2535
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 5.175969558214045e-05,
+      "loss": 0.0731,
+      "step": 2540
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 6.576510676653128e-05,
+      "loss": 0.0711,
+      "step": 2545
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 7.989118889325882e-05,
+      "loss": 0.0606,
+      "step": 2550
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 9.353664678067884e-05,
+      "loss": 0.0641,
+      "step": 2555
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 0.00010612064364719829,
+      "loss": 0.0674,
+      "step": 2560
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 0.00011710752518939675,
+      "loss": 0.0743,
+      "step": 2565
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 0.00012602962040880947,
+      "loss": 0.0674,
+      "step": 2570
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 0.00013250714864031726,
+      "loss": 0.074,
+      "step": 2575
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 0.00013626438541342663,
+      "loss": 0.0803,
+      "step": 2580
+    },
+    {
+      "epoch": 34.0,
+      "eval_loss": 6.624580383300781,
+      "eval_runtime": 2.3898,
+      "eval_samples_per_second": 42.681,
+      "eval_steps_per_second": 5.44,
+      "step": 2584
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 0.00013714139902536884,
+      "loss": 0.0696,
+      "step": 2585
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 0.00013510085824343794,
+      "loss": 0.0735,
+      "step": 2590
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 0.00013022962135779008,
+      "loss": 0.0738,
+      "step": 2595
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 0.0001227350389445927,
+      "loss": 0.074,
+      "step": 2600
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 0.00011293612771726094,
+      "loss": 0.079,
+      "step": 2605
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 0.00010124999116234316,
+      "loss": 0.0713,
+      "step": 2610
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 8.81740649834898e-05,
+      "loss": 0.0775,
+      "step": 2615
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 7.42649430994026e-05,
+      "loss": 0.0827,
+      "step": 2620
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 6.011468549492275e-05,
+      "loss": 0.0746,
+      "step": 2625
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 4.632561641255812e-05,
+      "loss": 0.0814,
+      "step": 2630
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 3.3484685632453976e-05,
+      "loss": 0.0829,
+      "step": 2635
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 2.213848418647423e-05,
+      "loss": 0.0812,
+      "step": 2640
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 1.276997799526488e-05,
+      "loss": 0.0637,
+      "step": 2645
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 5.777949791462987e-06,
+      "loss": 0.0834,
+      "step": 2650
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 1.460024410543719e-06,
+      "loss": 0.0753,
+      "step": 2655
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 0.0,
+      "loss": 0.0667,
+      "step": 2660
+    },
+    {
+      "epoch": 35.0,
+      "eval_loss": 6.608242034912109,
+      "eval_runtime": 2.3714,
+      "eval_samples_per_second": 43.012,
+      "eval_steps_per_second": 5.482,
+      "step": 2660
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 1.460024410543498e-06,
+      "loss": 0.0652,
+      "step": 2665
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 5.77794979146256e-06,
+      "loss": 0.0662,
+      "step": 2670
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 1.2769977995264255e-05,
+      "loss": 0.0526,
+      "step": 2675
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 2.2138484186474152e-05,
+      "loss": 0.0655,
+      "step": 2680
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 3.3484685632453895e-05,
+      "loss": 0.0666,
+      "step": 2685
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 4.632561641255803e-05,
+      "loss": 0.0599,
+      "step": 2690
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 6.0114685494922647e-05,
+      "loss": 0.0552,
+      "step": 2695
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 7.426494309940153e-05,
+      "loss": 0.0632,
+      "step": 2700
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 8.817406498348876e-05,
+      "loss": 0.064,
+      "step": 2705
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 0.00010124999116234222,
+      "loss": 0.0597,
+      "step": 2710
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 0.00011293612771726013,
+      "loss": 0.0753,
+      "step": 2715
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 0.00012273503894459265,
+      "loss": 0.0719,
+      "step": 2720
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 0.00013022962135779005,
+      "loss": 0.0701,
+      "step": 2725
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 0.0001351008582434379,
+      "loss": 0.0655,
+      "step": 2730
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 0.00013714139902536884,
+      "loss": 0.086,
+      "step": 2735
+    },
+    {
+      "epoch": 36.0,
+      "eval_loss": 6.644789695739746,
+      "eval_runtime": 2.3906,
+      "eval_samples_per_second": 42.667,
+      "eval_steps_per_second": 5.438,
+      "step": 2736
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 0.00013626438541342682,
+      "loss": 0.0695,
+      "step": 2740
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 0.00013250714864031766,
+      "loss": 0.0606,
+      "step": 2745
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 0.00012602962040881007,
+      "loss": 0.0733,
+      "step": 2750
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 0.00011710752518939751,
+      "loss": 0.0775,
+      "step": 2755
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 0.00010612064364719839,
+      "loss": 0.0782,
+      "step": 2760
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 9.353664678067986e-05,
+      "loss": 0.0732,
+      "step": 2765
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 7.989118889325892e-05,
+      "loss": 0.076,
+      "step": 2770
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 6.576510676653137e-05,
+      "loss": 0.0736,
+      "step": 2775
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 5.175969558214054e-05,
+      "loss": 0.0693,
+      "step": 2780
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 3.847111400507101e-05,
+      "loss": 0.0771,
+      "step": 2785
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 2.6465007909489603e-05,
+      "loss": 0.0622,
+      "step": 2790
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 1.6252432921493545e-05,
+      "loss": 0.0668,
+      "step": 2795
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 8.26810066723475e-06,
+      "loss": 0.0638,
+      "step": 2800
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 2.851874703364184e-06,
+      "loss": 0.0742,
+      "step": 2805
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 2.3430377974248658e-07,
+      "loss": 0.0651,
+      "step": 2810
+    },
+    {
+      "epoch": 37.0,
+      "eval_loss": 6.649299144744873,
+      "eval_runtime": 2.3884,
+      "eval_samples_per_second": 42.707,
+      "eval_steps_per_second": 5.443,
+      "step": 2812
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 5.268082300387194e-07,
+      "loss": 0.067,
+      "step": 2815
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 3.7169372193367e-06,
+      "loss": 0.0551,
+      "step": 2820
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 9.668898730225563e-06,
+      "loss": 0.0525,
+      "step": 2825
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 1.812933972782318e-05,
+      "loss": 0.0596,
+      "step": 2830
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 2.8738130463426756e-05,
+      "loss": 0.063,
+      "step": 2835
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 4.104369386880655e-05,
+      "loss": 0.0566,
+      "step": 2840
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 5.452222752537846e-05,
+      "loss": 0.0547,
+      "step": 2845
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 6.859999999999957e-05,
+      "loss": 0.0582,
+      "step": 2850
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 8.26777724746207e-05,
+      "loss": 0.0583,
+      "step": 2855
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 9.615630613119265e-05,
+      "loss": 0.0561,
+      "step": 2860
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 0.00010846186953657255,
+      "loss": 0.0592,
+      "step": 2865
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 0.0001190706602721769,
+      "loss": 0.0549,
+      "step": 2870
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 0.000127531101269774,
+      "loss": 0.0686,
+      "step": 2875
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 0.00013348306278066334,
+      "loss": 0.0682,
+      "step": 2880
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 0.00013667319176996116,
+      "loss": 0.0665,
+      "step": 2885
+    },
+    {
+      "epoch": 38.0,
+      "eval_loss": 6.660224437713623,
+      "eval_runtime": 2.3808,
+      "eval_samples_per_second": 42.844,
+      "eval_steps_per_second": 5.46,
+      "step": 2888
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 0.0001369656962202576,
+      "loss": 0.0664,
+      "step": 2890
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 0.00013434812529663606,
+      "loss": 0.0631,
+      "step": 2895
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 0.00012893189933276566,
+      "loss": 0.0616,
+      "step": 2900
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 0.00012094756707850702,
+      "loss": 0.0671,
+      "step": 2905
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 0.00011073499209051109,
+      "loss": 0.0692,
+      "step": 2910
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 9.87288859949289e-05,
+      "loss": 0.0676,
+      "step": 2915
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 8.544030441785936e-05,
+      "loss": 0.0729,
+      "step": 2920
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 7.143489323346852e-05,
+      "loss": 0.0826,
+      "step": 2925
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 5.730881110674194e-05,
+      "loss": 0.0716,
+      "step": 2930
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 4.366335321932097e-05,
+      "loss": 0.0631,
+      "step": 2935
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 3.107935635280236e-05,
+      "loss": 0.0674,
+      "step": 2940
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 2.009247481060311e-05,
+      "loss": 0.0674,
+      "step": 2945
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 1.1170379591190413e-05,
+      "loss": 0.0611,
+      "step": 2950
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 4.692851359682665e-06,
+      "loss": 0.068,
+      "step": 2955
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 9.356145865733248e-07,
+      "loss": 0.0651,
+      "step": 2960
+    },
+    {
+      "epoch": 39.0,
+      "eval_loss": 6.686155319213867,
+      "eval_runtime": 2.3897,
+      "eval_samples_per_second": 42.683,
+      "eval_steps_per_second": 5.44,
+      "step": 2964
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 5.8600974631157675e-08,
+      "loss": 0.0541,
+      "step": 2965
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 2.099141756561883e-06,
+      "loss": 0.0574,
+      "step": 2970
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 6.970378642209999e-06,
+      "loss": 0.057,
+      "step": 2975
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 1.446496105540683e-05,
+      "loss": 0.0571,
+      "step": 2980
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 2.4263872282739218e-05,
+      "loss": 0.0495,
+      "step": 2985
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 3.595000883765702e-05,
+      "loss": 0.0575,
+      "step": 2990
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 4.902593501651039e-05,
+      "loss": 0.0554,
+      "step": 2995
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 6.29350569005976e-05,
+      "loss": 0.0578,
+      "step": 3000
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 7.708531450507746e-05,
+      "loss": 0.0509,
+      "step": 3005
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 9.087438358744115e-05,
+      "loss": 0.0458,
+      "step": 3010
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 0.00010371531436754619,
+      "loss": 0.05,
+      "step": 3015
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 0.00011506151581352521,
+      "loss": 0.0553,
+      "step": 3020
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 0.00012443002200473525,
+      "loss": 0.0575,
+      "step": 3025
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 0.00013142205020853708,
+      "loss": 0.0554,
+      "step": 3030
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 0.00013573997558945632,
+      "loss": 0.063,
+      "step": 3035
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 0.0001372,
+      "loss": 0.0733,
+      "step": 3040
+    },
+    {
+      "epoch": 40.0,
+      "eval_loss": 6.705848693847656,
+      "eval_runtime": 2.382,
+      "eval_samples_per_second": 42.821,
+      "eval_steps_per_second": 5.458,
+      "step": 3040
+    },
+    {
+      "epoch": 40.07,
+      "learning_rate": 0.00013573997558945645,
+      "loss": 0.0557,
+      "step": 3045
+    },
+    {
+      "epoch": 40.13,
+      "learning_rate": 0.00013142205020853738,
+      "loss": 0.0663,
+      "step": 3050
+    },
+    {
+      "epoch": 40.2,
+      "learning_rate": 0.00012443002200473563,
+      "loss": 0.0627,
+      "step": 3055
+    },
+    {
+      "epoch": 40.26,
+      "learning_rate": 0.0001150615158135257,
+      "loss": 0.0658,
+      "step": 3060
+    },
+    {
+      "epoch": 40.33,
+      "learning_rate": 0.00010371531436754677,
+      "loss": 0.0668,
+      "step": 3065
+    },
+    {
+      "epoch": 40.39,
+      "learning_rate": 9.087438358744178e-05,
+      "loss": 0.0801,
+      "step": 3070
+    },
+    {
+      "epoch": 40.46,
+      "learning_rate": 7.708531450507813e-05,
+      "loss": 0.0727,
+      "step": 3075
+    },
+    {
+      "epoch": 40.53,
+      "learning_rate": 6.293505690059827e-05,
+      "loss": 0.0678,
+      "step": 3080
+    },
+    {
+      "epoch": 40.59,
+      "learning_rate": 4.902593501651104e-05,
+      "loss": 0.0593,
+      "step": 3085
+    },
+    {
+      "epoch": 40.66,
+      "learning_rate": 3.5950008837657616e-05,
+      "loss": 0.0635,
+      "step": 3090
+    },
+    {
+      "epoch": 40.72,
+      "learning_rate": 2.426387228273973e-05,
+      "loss": 0.0556,
+      "step": 3095
+    },
+    {
+      "epoch": 40.79,
+      "learning_rate": 1.4464961055407835e-05,
+      "loss": 0.0694,
+      "step": 3100
+    },
+    {
+      "epoch": 40.86,
+      "learning_rate": 6.970378642209862e-06,
+      "loss": 0.0597,
+      "step": 3105
+    },
+    {
+      "epoch": 40.92,
+      "learning_rate": 2.0991417565622865e-06,
+      "loss": 0.0636,
+      "step": 3110
+    },
+    {
+      "epoch": 40.99,
+      "learning_rate": 5.860097463114244e-08,
+      "loss": 0.0535,
+      "step": 3115
+    },
+    {
+      "epoch": 41.0,
+      "eval_loss": 6.696343421936035,
+      "eval_runtime": 2.3795,
+      "eval_samples_per_second": 42.866,
+      "eval_steps_per_second": 5.463,
+      "step": 3116
+    },
+    {
+      "epoch": 41.05,
+      "learning_rate": 9.356145865732106e-07,
+      "loss": 0.0552,
+      "step": 3120
+    },
+    {
+      "epoch": 41.12,
+      "learning_rate": 4.692851359682063e-06,
+      "loss": 0.0537,
+      "step": 3125
+    },
+    {
+      "epoch": 41.18,
+      "learning_rate": 1.1170379591190047e-05,
+      "loss": 0.0546,
+      "step": 3130
+    },
+    {
+      "epoch": 41.25,
+      "learning_rate": 2.0092474810601945e-05,
+      "loss": 0.051,
+      "step": 3135
+    },
+    {
+      "epoch": 41.32,
+      "learning_rate": 3.10793563528018e-05,
+      "loss": 0.0501,
+      "step": 3140
+    },
+    {
+      "epoch": 41.38,
+      "learning_rate": 4.366335321931943e-05,
+      "loss": 0.0524,
+      "step": 3145
+    },
+    {
+      "epoch": 41.45,
+      "learning_rate": 5.730881110674128e-05,
+      "loss": 0.0483,
+      "step": 3150
+    },
+    {
+      "epoch": 41.51,
+      "learning_rate": 7.143489323346784e-05,
+      "loss": 0.0532,
+      "step": 3155
+    },
+    {
+      "epoch": 41.58,
+      "learning_rate": 8.544030441785965e-05,
+      "loss": 0.0542,
+      "step": 3160
+    },
+    {
+      "epoch": 41.64,
+      "learning_rate": 9.87288859949283e-05,
+      "loss": 0.0503,
+      "step": 3165
+    },
+    {
+      "epoch": 41.71,
+      "learning_rate": 0.00011073499209051133,
+      "loss": 0.0561,
+      "step": 3170
+    },
+    {
+      "epoch": 41.78,
+      "learning_rate": 0.0001209475670785066,
+      "loss": 0.0526,
+      "step": 3175
+    },
+    {
+      "epoch": 41.84,
+      "learning_rate": 0.00012893189933276488,
+      "loss": 0.0548,
+      "step": 3180
+    },
+    {
+      "epoch": 41.91,
+      "learning_rate": 0.00013434812529663587,
+      "loss": 0.0662,
+      "step": 3185
+    },
+    {
+      "epoch": 41.97,
+      "learning_rate": 0.00013696569622025746,
+      "loss": 0.0633,
+      "step": 3190
+    },
+    {
+      "epoch": 42.0,
+      "eval_loss": 6.702749252319336,
+      "eval_runtime": 2.3956,
+      "eval_samples_per_second": 42.577,
+      "eval_steps_per_second": 5.427,
+      "step": 3192
+    },
+    {
+      "epoch": 42.04,
+      "learning_rate": 0.00013667319176996126,
+      "loss": 0.0521,
+      "step": 3195
+    },
+    {
+      "epoch": 42.11,
+      "learning_rate": 0.00013348306278066356,
+      "loss": 0.0563,
+      "step": 3200
+    },
+    {
+      "epoch": 42.17,
+      "learning_rate": 0.00012753110126977434,
+      "loss": 0.0619,
+      "step": 3205
+    },
+    {
+      "epoch": 42.24,
+      "learning_rate": 0.00011907066027217736,
+      "loss": 0.0651,
+      "step": 3210
+    },
+    {
+      "epoch": 42.3,
+      "learning_rate": 0.00010846186953657228,
+      "loss": 0.0628,
+      "step": 3215
+    },
+    {
+      "epoch": 42.37,
+      "learning_rate": 9.615630613119328e-05,
+      "loss": 0.0659,
+      "step": 3220
+    },
+    {
+      "epoch": 42.43,
+      "learning_rate": 8.26777724746204e-05,
+      "loss": 0.0655,
+      "step": 3225
+    },
+    {
+      "epoch": 42.5,
+      "learning_rate": 6.860000000000023e-05,
+      "loss": 0.0623,
+      "step": 3230
+    },
+    {
+      "epoch": 42.57,
+      "learning_rate": 5.4522227525380065e-05,
+      "loss": 0.064,
+      "step": 3235
+    },
+    {
+      "epoch": 42.63,
+      "learning_rate": 4.1043693868807165e-05,
+      "loss": 0.0627,
+      "step": 3240
+    },
+    {
+      "epoch": 42.7,
+      "learning_rate": 2.8738130463428097e-05,
+      "loss": 0.057,
+      "step": 3245
+    },
+    {
+      "epoch": 42.76,
+      "learning_rate": 1.8129339727822976e-05,
+      "loss": 0.056,
+      "step": 3250
+    },
+    {
+      "epoch": 42.83,
+      "learning_rate": 9.668898730225914e-06,
+      "loss": 0.0651,
+      "step": 3255
+    },
+    {
+      "epoch": 42.89,
+      "learning_rate": 3.716937219336601e-06,
+      "loss": 0.0567,
+      "step": 3260
+    },
+    {
+      "epoch": 42.96,
+      "learning_rate": 5.268082300388033e-07,
+      "loss": 0.0803,
+      "step": 3265
+    },
+    {
+      "epoch": 43.0,
+      "eval_loss": 6.75062370300293,
+      "eval_runtime": 2.3741,
+      "eval_samples_per_second": 42.963,
+      "eval_steps_per_second": 5.476,
+      "step": 3268
+    },
+    {
+      "epoch": 43.03,
+      "learning_rate": 2.343037797423571e-07,
+      "loss": 0.0462,
+      "step": 3270
+    },
+    {
+      "epoch": 43.09,
+      "learning_rate": 2.8518747033639933e-06,
+      "loss": 0.0484,
+      "step": 3275
+    },
+    {
+      "epoch": 43.16,
+      "learning_rate": 8.268100667233966e-06,
+      "loss": 0.0574,
+      "step": 3280
+    },
+    {
+      "epoch": 43.22,
+      "learning_rate": 1.625243292149311e-05,
+      "loss": 0.0424,
+      "step": 3285
+    },
+    {
+      "epoch": 43.29,
+      "learning_rate": 2.64650079094883e-05,
+      "loss": 0.0483,
+      "step": 3290
+    },
+    {
+      "epoch": 43.36,
+      "learning_rate": 3.847111400507128e-05,
+      "loss": 0.0512,
+      "step": 3295
+    },
+    {
+      "epoch": 43.42,
+      "learning_rate": 5.17596955821399e-05,
+      "loss": 0.0528,
+      "step": 3300
+    },
+    {
+      "epoch": 43.49,
+      "learning_rate": 6.576510676653168e-05,
+      "loss": 0.0431,
+      "step": 3305
+    },
+    {
+      "epoch": 43.55,
+      "learning_rate": 7.989118889325826e-05,
+      "loss": 0.0531,
+      "step": 3310
+    },
+    {
+      "epoch": 43.62,
+      "learning_rate": 9.353664678067831e-05,
+      "loss": 0.0506,
+      "step": 3315
+    },
+    {
+      "epoch": 43.68,
+      "learning_rate": 0.0001061206436471978,
+      "loss": 0.0495,
+      "step": 3320
+    },
+    {
+      "epoch": 43.75,
+      "learning_rate": 0.00011710752518939633,
+      "loss": 0.059,
+      "step": 3325
+    },
+    {
+      "epoch": 43.82,
+      "learning_rate": 0.0001260296204088097,
+      "loss": 0.0589,
+      "step": 3330
+    },
+    {
+      "epoch": 43.88,
+      "learning_rate": 0.00013250714864031707,
+      "loss": 0.0594,
+      "step": 3335
+    },
+    {
+      "epoch": 43.95,
+      "learning_rate": 0.00013626438541342671,
+      "loss": 0.061,
+      "step": 3340
+    },
+    {
+      "epoch": 44.0,
+      "eval_loss": 6.78729248046875,
+      "eval_runtime": 2.3812,
+      "eval_samples_per_second": 42.835,
+      "eval_steps_per_second": 5.459,
+      "step": 3344
+    },
+    {
+      "epoch": 44.01,
+      "learning_rate": 0.00013714139902536887,
+      "loss": 0.0634,
+      "step": 3345
+    },
+    {
+      "epoch": 44.08,
+      "learning_rate": 0.00013510085824343783,
+      "loss": 0.0513,
+      "step": 3350
+    },
+    {
+      "epoch": 44.14,
+      "learning_rate": 0.00013022962135779035,
+      "loss": 0.0597,
+      "step": 3355
+    },
+    {
+      "epoch": 44.21,
+      "learning_rate": 0.00012273503894459246,
+      "loss": 0.0558,
+      "step": 3360
+    },
+    {
+      "epoch": 44.28,
+      "learning_rate": 0.00011293612771726063,
+      "loss": 0.0677,
+      "step": 3365
+    },
+    {
+      "epoch": 44.34,
+      "learning_rate": 0.00010124999116234365,
+      "loss": 0.0662,
+      "step": 3370
+    },
+    {
+      "epoch": 44.41,
+      "learning_rate": 8.817406498348941e-05,
+      "loss": 0.0628,
+      "step": 3375
+    },
+    {
+      "epoch": 44.47,
+      "learning_rate": 7.426494309940318e-05,
+      "loss": 0.0616,
+      "step": 3380
+    },
+    {
+      "epoch": 44.54,
+      "learning_rate": 6.011468549492234e-05,
+      "loss": 0.0614,
+      "step": 3385
+    },
+    {
+      "epoch": 44.61,
+      "learning_rate": 4.632561641255866e-05,
+      "loss": 0.0577,
+      "step": 3390
+    },
+    {
+      "epoch": 44.67,
+      "learning_rate": 3.348468563245363e-05,
+      "loss": 0.0629,
+      "step": 3395
+    },
+    {
+      "epoch": 44.74,
+      "learning_rate": 2.2138484186474647e-05,
+      "loss": 0.0543,
+      "step": 3400
+    },
+    {
+      "epoch": 44.8,
+      "learning_rate": 1.2769977995265214e-05,
+      "loss": 0.0467,
+      "step": 3405
+    },
+    {
+      "epoch": 44.87,
+      "learning_rate": 5.7779497914628265e-06,
+      "loss": 0.0532,
+      "step": 3410
+    },
+    {
+      "epoch": 44.93,
+      "learning_rate": 1.4600244105438407e-06,
+      "loss": 0.0566,
+      "step": 3415
+    },
+    {
+      "epoch": 45.0,
+      "learning_rate": 0.0,
+      "loss": 0.0512,
+      "step": 3420
+    },
+    {
+      "epoch": 45.0,
+      "eval_loss": 6.75970458984375,
+      "eval_runtime": 2.3797,
+      "eval_samples_per_second": 42.863,
+      "eval_steps_per_second": 5.463,
+      "step": 3420
+    },
+    {
+      "epoch": 45.07,
+      "learning_rate": 1.4600244105433837e-06,
+      "loss": 0.0475,
+      "step": 3425
+    },
+    {
+      "epoch": 45.13,
+      "learning_rate": 5.77794979146272e-06,
+      "loss": 0.0525,
+      "step": 3430
+    },
+    {
+      "epoch": 45.2,
+      "learning_rate": 1.2769977995263928e-05,
+      "loss": 0.0459,
+      "step": 3435
+    },
+    {
+      "epoch": 45.26,
+      "learning_rate": 2.213848418647445e-05,
+      "loss": 0.0464,
+      "step": 3440
+    },
+    {
+      "epoch": 45.33,
+      "learning_rate": 3.348468563245339e-05,
+      "loss": 0.0515,
+      "step": 3445
+    },
+    {
+      "epoch": 45.39,
+      "learning_rate": 4.632561641255841e-05,
+      "loss": 0.0465,
+      "step": 3450
+    },
+    {
+      "epoch": 45.46,
+      "learning_rate": 6.011468549492208e-05,
+      "loss": 0.0493,
+      "step": 3455
+    },
+    {
+      "epoch": 45.53,
+      "learning_rate": 7.426494309940098e-05,
+      "loss": 0.0489,
+      "step": 3460
+    },
+    {
+      "epoch": 45.59,
+      "learning_rate": 8.817406498348916e-05,
+      "loss": 0.0471,
+      "step": 3465
+    },
+    {
+      "epoch": 45.66,
+      "learning_rate": 0.0001012499911623417,
+      "loss": 0.0401,
+      "step": 3470
+    },
+    {
+      "epoch": 45.72,
+      "learning_rate": 0.00011293612771726042,
+      "loss": 0.0492,
+      "step": 3475
+    },
+    {
+      "epoch": 45.79,
+      "learning_rate": 0.0001227350389445923,
+      "loss": 0.0478,
+      "step": 3480
+    },
+    {
+      "epoch": 45.86,
+      "learning_rate": 0.00013022962135779022,
+      "loss": 0.0487,
+      "step": 3485
+    },
+    {
+      "epoch": 45.92,
+      "learning_rate": 0.00013510085824343775,
+      "loss": 0.0522,
+      "step": 3490
+    },
+    {
+      "epoch": 45.99,
+      "learning_rate": 0.00013714139902536887,
+      "loss": 0.0539,
+      "step": 3495
+    },
+    {
+      "epoch": 46.0,
+      "eval_loss": 6.77425479888916,
+      "eval_runtime": 2.3831,
+      "eval_samples_per_second": 42.802,
+      "eval_steps_per_second": 5.455,
+      "step": 3496
+    },
+    {
+      "epoch": 46.05,
+      "learning_rate": 0.00013626438541342674,
+      "loss": 0.0588,
+      "step": 3500
+    },
+    {
+      "epoch": 46.12,
+      "learning_rate": 0.00013250714864031785,
+      "loss": 0.0596,
+      "step": 3505
+    },
+    {
+      "epoch": 46.18,
+      "learning_rate": 0.00012602962040880985,
+      "loss": 0.0537,
+      "step": 3510
+    },
+    {
+      "epoch": 46.25,
+      "learning_rate": 0.00011710752518939791,
+      "loss": 0.0534,
+      "step": 3515
+    },
+    {
+      "epoch": 46.32,
+      "learning_rate": 0.00010612064364719805,
+      "loss": 0.0602,
+      "step": 3520
+    },
+    {
+      "epoch": 46.38,
+      "learning_rate": 9.353664678068038e-05,
+      "loss": 0.0561,
+      "step": 3525
+    },
+    {
+      "epoch": 46.45,
+      "learning_rate": 7.989118889325852e-05,
+      "loss": 0.0507,
+      "step": 3530
+    },
+    {
+      "epoch": 46.51,
+      "learning_rate": 6.576510676653195e-05,
+      "loss": 0.0508,
+      "step": 3535
+    },
+    {
+      "epoch": 46.58,
+      "learning_rate": 5.1759695582140156e-05,
+      "loss": 0.0514,
+      "step": 3540
+    },
+    {
+      "epoch": 46.64,
+      "learning_rate": 3.8471114005071515e-05,
+      "loss": 0.0513,
+      "step": 3545
+    },
+    {
+      "epoch": 46.71,
+      "learning_rate": 2.646500790949005e-05,
+      "loss": 0.0556,
+      "step": 3550
+    },
+    {
+      "epoch": 46.78,
+      "learning_rate": 1.6252432921493277e-05,
+      "loss": 0.0532,
+      "step": 3555
+    },
+    {
+      "epoch": 46.84,
+      "learning_rate": 8.268100667235025e-06,
+      "loss": 0.0507,
+      "step": 3560
+    },
+    {
+      "epoch": 46.91,
+      "learning_rate": 2.8518747033640696e-06,
+      "loss": 0.052,
+      "step": 3565
+    },
+    {
+      "epoch": 46.97,
+      "learning_rate": 2.343037797425399e-07,
+      "loss": 0.0503,
+      "step": 3570
+    },
+    {
+      "epoch": 47.0,
+      "eval_loss": 6.793975353240967,
+      "eval_runtime": 2.3872,
+      "eval_samples_per_second": 42.727,
+      "eval_steps_per_second": 5.446,
+      "step": 3572
+    },
+    {
+      "epoch": 47.04,
+      "learning_rate": 5.268082300387728e-07,
+      "loss": 0.0538,
+      "step": 3575
+    },
+    {
+      "epoch": 47.11,
+      "learning_rate": 3.7169372193365096e-06,
+      "loss": 0.0408,
+      "step": 3580
+    },
+    {
+      "epoch": 47.17,
+      "learning_rate": 9.668898730225777e-06,
+      "loss": 0.053,
+      "step": 3585
+    },
+    {
+      "epoch": 47.24,
+      "learning_rate": 1.8129339727822793e-05,
+      "loss": 0.047,
+      "step": 3590
+    },
+    {
+      "epoch": 47.3,
+      "learning_rate": 2.873813046342629e-05,
+      "loss": 0.0432,
+      "step": 3595
+    },
+    {
+      "epoch": 47.37,
+      "learning_rate": 4.104369386880692e-05,
+      "loss": 0.0432,
+      "step": 3600
+    },
+    {
+      "epoch": 47.43,
+      "learning_rate": 5.4522227525377896e-05,
+      "loss": 0.0438,
+      "step": 3605
+    },
+    {
+      "epoch": 47.5,
+      "learning_rate": 6.859999999999997e-05,
+      "loss": 0.0433,
+      "step": 3610
+    },
+    {
+      "epoch": 47.57,
+      "learning_rate": 8.267777247462013e-05,
+      "loss": 0.0453,
+      "step": 3615
+    },
+    {
+      "epoch": 47.63,
+      "learning_rate": 9.615630613119303e-05,
+      "loss": 0.0466,
+      "step": 3620
+    },
+    {
+      "epoch": 47.7,
+      "learning_rate": 0.00010846186953657208,
+      "loss": 0.0471,
+      "step": 3625
+    },
+    {
+      "epoch": 47.76,
+      "learning_rate": 0.00011907066027217717,
+      "loss": 0.0438,
+      "step": 3630
+    },
+    {
+      "epoch": 47.83,
+      "learning_rate": 0.00012753110126977418,
+      "loss": 0.0497,
+      "step": 3635
+    },
+    {
+      "epoch": 47.89,
+      "learning_rate": 0.00013348306278066348,
+      "loss": 0.049,
+      "step": 3640
+    },
+    {
+      "epoch": 47.96,
+      "learning_rate": 0.0001366731917699612,
+      "loss": 0.0489,
+      "step": 3645
+    },
+    {
+      "epoch": 48.0,
+      "eval_loss": 6.83149528503418,
+      "eval_runtime": 2.3845,
+      "eval_samples_per_second": 42.777,
+      "eval_steps_per_second": 5.452,
+      "step": 3648
     }
   ],
-  "max_steps": 893,
-  "num_train_epochs": 19,
-  "total_flos": 244308049920000.0,
+  "max_steps": 3800,
+  "num_train_epochs": 50,
+  "total_flos": 3768876269568000.0,
   "trial_name": null,
   "trial_params": null
 }