diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,9433 @@
+{
+  "best_metric": 0.10603297501802444,
+  "best_model_checkpoint": "./modeversion2_m7_e8/checkpoint-13600",
+  "epoch": 8.0,
+  "global_step": 13648,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019985345838218055,
+      "loss": 4.9362,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00019970691676436108,
+      "loss": 4.7228,
+      "step": 20
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019956037514654162,
+      "loss": 4.6088,
+      "step": 30
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00019941383352872216,
+      "loss": 4.5012,
+      "step": 40
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001992672919109027,
+      "loss": 4.4055,
+      "step": 50
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019912075029308323,
+      "loss": 4.3348,
+      "step": 60
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019897420867526376,
+      "loss": 4.1101,
+      "step": 70
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019882766705744433,
+      "loss": 4.0032,
+      "step": 80
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019868112543962486,
+      "loss": 4.0011,
+      "step": 90
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001985345838218054,
+      "loss": 4.0231,
+      "step": 100
+    },
+    {
+      "epoch": 0.06,
+      "eval_accuracy": 0.18829589354700244,
+      "eval_loss": 3.8568217754364014,
+      "eval_runtime": 154.561,
+      "eval_samples_per_second": 45.218,
+      "eval_steps_per_second": 5.655,
+      "step": 100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019838804220398594,
+      "loss": 3.8027,
+      "step": 110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019824150058616647,
+      "loss": 3.6998,
+      "step": 120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000198094958968347,
+      "loss": 3.6945,
+      "step": 130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019794841735052755,
+      "loss": 3.6753,
+      "step": 140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001978018757327081,
+      "loss": 3.5269,
+      "step": 150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019765533411488865,
+      "loss": 3.5352,
+      "step": 160
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019750879249706918,
+      "loss": 3.4251,
+      "step": 170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019736225087924972,
+      "loss": 3.3935,
+      "step": 180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019721570926143026,
+      "loss": 3.1822,
+      "step": 190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001970691676436108,
+      "loss": 3.3863,
+      "step": 200
+    },
+    {
+      "epoch": 0.12,
+      "eval_accuracy": 0.25955072256402917,
+      "eval_loss": 3.250999927520752,
+      "eval_runtime": 146.755,
+      "eval_samples_per_second": 47.624,
+      "eval_steps_per_second": 5.956,
+      "step": 200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019692262602579133,
+      "loss": 3.186,
+      "step": 210
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001967760844079719,
+      "loss": 3.1788,
+      "step": 220
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019662954279015243,
+      "loss": 3.0544,
+      "step": 230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019648300117233296,
+      "loss": 2.9295,
+      "step": 240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001963364595545135,
+      "loss": 2.9505,
+      "step": 250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019618991793669404,
+      "loss": 3.0351,
+      "step": 260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019604337631887457,
+      "loss": 2.8201,
+      "step": 270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001958968347010551,
+      "loss": 2.7123,
+      "step": 280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019575029308323565,
+      "loss": 2.6649,
+      "step": 290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019560375146541618,
+      "loss": 2.6187,
+      "step": 300
+    },
+    {
+      "epoch": 0.18,
+      "eval_accuracy": 0.3881814279582201,
+      "eval_loss": 2.6243486404418945,
+      "eval_runtime": 147.2875,
+      "eval_samples_per_second": 47.451,
+      "eval_steps_per_second": 5.934,
+      "step": 300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00019545720984759672,
+      "loss": 2.7383,
+      "step": 310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00019531066822977726,
+      "loss": 2.4618,
+      "step": 320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001951641266119578,
+      "loss": 2.6167,
+      "step": 330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00019501758499413833,
+      "loss": 2.5502,
+      "step": 340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019487104337631887,
+      "loss": 2.3974,
+      "step": 350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00019472450175849943,
+      "loss": 2.3332,
+      "step": 360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00019457796014067997,
+      "loss": 2.2212,
+      "step": 370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001944314185228605,
+      "loss": 2.3352,
+      "step": 380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019428487690504104,
+      "loss": 2.1789,
+      "step": 390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00019413833528722157,
+      "loss": 2.3097,
+      "step": 400
+    },
+    {
+      "epoch": 0.23,
+      "eval_accuracy": 0.4527114036342824,
+      "eval_loss": 2.218879222869873,
+      "eval_runtime": 146.5058,
+      "eval_samples_per_second": 47.705,
+      "eval_steps_per_second": 5.966,
+      "step": 400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001939917936694021,
+      "loss": 2.2105,
+      "step": 410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00019384525205158265,
+      "loss": 2.1158,
+      "step": 420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001936987104337632,
+      "loss": 2.3474,
+      "step": 430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019355216881594375,
+      "loss": 2.0149,
+      "step": 440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00019340562719812428,
+      "loss": 2.0887,
+      "step": 450
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00019325908558030482,
+      "loss": 2.0321,
+      "step": 460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00019311254396248536,
+      "loss": 2.0894,
+      "step": 470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001929660023446659,
+      "loss": 1.9558,
+      "step": 480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00019281946072684643,
+      "loss": 1.9417,
+      "step": 490
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000192672919109027,
+      "loss": 1.9016,
+      "step": 500
+    },
+    {
+      "epoch": 0.29,
+      "eval_accuracy": 0.5243954786092431,
+      "eval_loss": 1.949493408203125,
+      "eval_runtime": 146.4678,
+      "eval_samples_per_second": 47.717,
+      "eval_steps_per_second": 5.967,
+      "step": 500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019252637749120753,
+      "loss": 2.0172,
+      "step": 510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00019237983587338807,
+      "loss": 1.8567,
+      "step": 520
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001922332942555686,
+      "loss": 1.867,
+      "step": 530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019208675263774914,
+      "loss": 1.8106,
+      "step": 540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00019194021101992967,
+      "loss": 1.8043,
+      "step": 550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001917936694021102,
+      "loss": 1.6615,
+      "step": 560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019164712778429075,
+      "loss": 1.6955,
+      "step": 570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019150058616647128,
+      "loss": 1.5323,
+      "step": 580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019135404454865182,
+      "loss": 1.8823,
+      "step": 590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00019120750293083236,
+      "loss": 1.7478,
+      "step": 600
+    },
+    {
+      "epoch": 0.35,
+      "eval_accuracy": 0.6091000143081986,
+      "eval_loss": 1.6609317064285278,
+      "eval_runtime": 145.5033,
+      "eval_samples_per_second": 48.033,
+      "eval_steps_per_second": 6.007,
+      "step": 600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001910609613130129,
+      "loss": 1.5716,
+      "step": 610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00019091441969519343,
+      "loss": 1.6264,
+      "step": 620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00019076787807737397,
+      "loss": 1.4168,
+      "step": 630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019062133645955453,
+      "loss": 1.5568,
+      "step": 640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00019047479484173507,
+      "loss": 1.4259,
+      "step": 650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001903282532239156,
+      "loss": 1.6078,
+      "step": 660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019018171160609614,
+      "loss": 1.6134,
+      "step": 670
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019003516998827668,
+      "loss": 1.5145,
+      "step": 680
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001898886283704572,
+      "loss": 1.4064,
+      "step": 690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00018974208675263775,
+      "loss": 1.2345,
+      "step": 700
+    },
+    {
+      "epoch": 0.41,
+      "eval_accuracy": 0.6425811990270425,
+      "eval_loss": 1.4334971904754639,
+      "eval_runtime": 155.0534,
+      "eval_samples_per_second": 45.075,
+      "eval_steps_per_second": 5.637,
+      "step": 700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018959554513481828,
+      "loss": 1.4409,
+      "step": 710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018944900351699885,
+      "loss": 1.5001,
+      "step": 720
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018930246189917938,
+      "loss": 1.4612,
+      "step": 730
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018915592028135992,
+      "loss": 1.435,
+      "step": 740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018900937866354046,
+      "loss": 1.5151,
+      "step": 750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000188862837045721,
+      "loss": 1.4164,
+      "step": 760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018871629542790153,
+      "loss": 1.4199,
+      "step": 770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018856975381008207,
+      "loss": 1.3292,
+      "step": 780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00018842321219226263,
+      "loss": 1.1989,
+      "step": 790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00018827667057444317,
+      "loss": 1.4129,
+      "step": 800
+    },
+    {
+      "epoch": 0.47,
+      "eval_accuracy": 0.6752038918300186,
+      "eval_loss": 1.30009925365448,
+      "eval_runtime": 145.5894,
+      "eval_samples_per_second": 48.005,
+      "eval_steps_per_second": 6.003,
+      "step": 800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001881301289566237,
+      "loss": 1.2392,
+      "step": 810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00018798358733880424,
+      "loss": 1.2219,
+      "step": 820
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00018783704572098478,
+      "loss": 1.1573,
+      "step": 830
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001876905041031653,
+      "loss": 1.1639,
+      "step": 840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018754396248534585,
+      "loss": 1.3702,
+      "step": 850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00018739742086752638,
+      "loss": 1.2852,
+      "step": 860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00018725087924970692,
+      "loss": 1.3457,
+      "step": 870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00018710433763188746,
+      "loss": 1.0771,
+      "step": 880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000186957796014068,
+      "loss": 1.234,
+      "step": 890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018681125439624853,
+      "loss": 1.1722,
+      "step": 900
+    },
+    {
+      "epoch": 0.53,
+      "eval_accuracy": 0.6784947775075119,
+      "eval_loss": 1.2030422687530518,
+      "eval_runtime": 145.3194,
+      "eval_samples_per_second": 48.094,
+      "eval_steps_per_second": 6.014,
+      "step": 900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00018666471277842907,
+      "loss": 1.2407,
+      "step": 910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001865181711606096,
+      "loss": 1.1659,
+      "step": 920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00018637162954279017,
+      "loss": 1.1053,
+      "step": 930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001862250879249707,
+      "loss": 1.039,
+      "step": 940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00018607854630715124,
+      "loss": 1.036,
+      "step": 950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00018593200468933178,
+      "loss": 1.0771,
+      "step": 960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001857854630715123,
+      "loss": 0.9866,
+      "step": 970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00018563892145369285,
+      "loss": 1.1297,
+      "step": 980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00018549237983587339,
+      "loss": 1.0704,
+      "step": 990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018534583821805395,
+      "loss": 1.0808,
+      "step": 1000
+    },
+    {
+      "epoch": 0.59,
+      "eval_accuracy": 0.727285734725998,
+      "eval_loss": 1.0051182508468628,
+      "eval_runtime": 154.513,
+      "eval_samples_per_second": 45.232,
+      "eval_steps_per_second": 5.656,
+      "step": 1000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00018519929660023449,
+      "loss": 1.1615,
+      "step": 1010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018505275498241502,
+      "loss": 1.0797,
+      "step": 1020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00018490621336459556,
+      "loss": 1.0826,
+      "step": 1030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001847596717467761,
+      "loss": 1.1367,
+      "step": 1040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018461313012895663,
+      "loss": 1.0246,
+      "step": 1050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00018446658851113717,
+      "loss": 1.1359,
+      "step": 1060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018432004689331773,
+      "loss": 1.0534,
+      "step": 1070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00018417350527549827,
+      "loss": 1.0266,
+      "step": 1080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0001840269636576788,
+      "loss": 1.1158,
+      "step": 1090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00018388042203985934,
+      "loss": 0.8814,
+      "step": 1100
+    },
+    {
+      "epoch": 0.64,
+      "eval_accuracy": 0.7062526827872371,
+      "eval_loss": 1.0715495347976685,
+      "eval_runtime": 151.9784,
+      "eval_samples_per_second": 45.987,
+      "eval_steps_per_second": 5.751,
+      "step": 1100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00018373388042203988,
+      "loss": 1.0724,
+      "step": 1110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0001835873388042204,
+      "loss": 0.9641,
+      "step": 1120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00018344079718640095,
+      "loss": 0.9985,
+      "step": 1130
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018329425556858149,
+      "loss": 0.7201,
+      "step": 1140
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00018314771395076202,
+      "loss": 0.9213,
+      "step": 1150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00018300117233294256,
+      "loss": 0.8745,
+      "step": 1160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0001828546307151231,
+      "loss": 0.9735,
+      "step": 1170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00018270808909730363,
+      "loss": 0.8539,
+      "step": 1180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00018256154747948417,
+      "loss": 0.8508,
+      "step": 1190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001824150058616647,
+      "loss": 0.9831,
+      "step": 1200
+    },
+    {
+      "epoch": 0.7,
+      "eval_accuracy": 0.7334382601230505,
+      "eval_loss": 0.9282865524291992,
+      "eval_runtime": 152.1233,
+      "eval_samples_per_second": 45.943,
+      "eval_steps_per_second": 5.745,
+      "step": 1200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00018226846424384527,
+      "loss": 0.8615,
+      "step": 1210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0001821219226260258,
+      "loss": 0.9357,
+      "step": 1220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00018197538100820634,
+      "loss": 0.9575,
+      "step": 1230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00018182883939038688,
+      "loss": 1.0384,
+      "step": 1240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0001816822977725674,
+      "loss": 0.9829,
+      "step": 1250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018153575615474795,
+      "loss": 1.0311,
+      "step": 1260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00018138921453692849,
+      "loss": 0.9393,
+      "step": 1270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00018124267291910902,
+      "loss": 0.8901,
+      "step": 1280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018109613130128959,
+      "loss": 0.9026,
+      "step": 1290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00018094958968347012,
+      "loss": 0.8118,
+      "step": 1300
+    },
+    {
+      "epoch": 0.76,
+      "eval_accuracy": 0.7630562312204894,
+      "eval_loss": 0.8524749875068665,
+      "eval_runtime": 143.9267,
+      "eval_samples_per_second": 48.559,
+      "eval_steps_per_second": 6.073,
+      "step": 1300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00018080304806565066,
+      "loss": 0.8655,
+      "step": 1310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0001806565064478312,
+      "loss": 0.8091,
+      "step": 1320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00018050996483001173,
+      "loss": 0.7504,
+      "step": 1330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00018036342321219227,
+      "loss": 0.8198,
+      "step": 1340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0001802168815943728,
+      "loss": 0.6709,
+      "step": 1350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00018007033997655337,
+      "loss": 0.5819,
+      "step": 1360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0001799237983587339,
+      "loss": 0.8105,
+      "step": 1370
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017977725674091444,
+      "loss": 0.8451,
+      "step": 1380
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00017963071512309498,
+      "loss": 0.759,
+      "step": 1390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00017948417350527551,
+      "loss": 0.7203,
+      "step": 1400
+    },
+    {
+      "epoch": 0.82,
+      "eval_accuracy": 0.7756474459865503,
+      "eval_loss": 0.7848576307296753,
+      "eval_runtime": 140.891,
+      "eval_samples_per_second": 49.606,
+      "eval_steps_per_second": 6.203,
+      "step": 1400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017933763188745605,
+      "loss": 0.6735,
+      "step": 1410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00017919109026963659,
+      "loss": 0.8848,
+      "step": 1420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017904454865181712,
+      "loss": 0.9537,
+      "step": 1430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00017889800703399766,
+      "loss": 0.8524,
+      "step": 1440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0001787514654161782,
+      "loss": 0.7838,
+      "step": 1450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017860492379835873,
+      "loss": 0.7812,
+      "step": 1460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00017845838218053927,
+      "loss": 0.8121,
+      "step": 1470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0001783118405627198,
+      "loss": 0.7117,
+      "step": 1480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00017816529894490034,
+      "loss": 0.8253,
+      "step": 1490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0001780187573270809,
+      "loss": 0.8881,
+      "step": 1500
+    },
+    {
+      "epoch": 0.88,
+      "eval_accuracy": 0.7487480326226928,
+      "eval_loss": 0.8785643577575684,
+      "eval_runtime": 142.9471,
+      "eval_samples_per_second": 48.892,
+      "eval_steps_per_second": 6.114,
+      "step": 1500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017787221570926144,
+      "loss": 0.835,
+      "step": 1510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00017772567409144198,
+      "loss": 0.772,
+      "step": 1520
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017757913247362251,
+      "loss": 0.8038,
+      "step": 1530
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00017743259085580305,
+      "loss": 0.6412,
+      "step": 1540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0001772860492379836,
+      "loss": 0.7474,
+      "step": 1550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00017713950762016412,
+      "loss": 0.6952,
+      "step": 1560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0001769929660023447,
+      "loss": 0.7074,
+      "step": 1570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017684642438452522,
+      "loss": 0.8314,
+      "step": 1580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00017669988276670576,
+      "loss": 0.642,
+      "step": 1590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0001765533411488863,
+      "loss": 0.6407,
+      "step": 1600
+    },
+    {
+      "epoch": 0.94,
+      "eval_accuracy": 0.7999713836028044,
+      "eval_loss": 0.6896187663078308,
+      "eval_runtime": 132.1077,
+      "eval_samples_per_second": 52.904,
+      "eval_steps_per_second": 6.616,
+      "step": 1600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00017640679953106683,
+      "loss": 0.8928,
+      "step": 1610
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00017626025791324737,
+      "loss": 0.6333,
+      "step": 1620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0001761137162954279,
+      "loss": 0.6867,
+      "step": 1630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00017596717467760847,
+      "loss": 0.6426,
+      "step": 1640
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.000175820633059789,
+      "loss": 0.7396,
+      "step": 1650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00017567409144196954,
+      "loss": 0.7627,
+      "step": 1660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017552754982415008,
+      "loss": 0.6725,
+      "step": 1670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00017538100820633061,
+      "loss": 0.7504,
+      "step": 1680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00017523446658851115,
+      "loss": 0.8719,
+      "step": 1690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0001750879249706917,
+      "loss": 0.7574,
+      "step": 1700
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7753612820145943,
+      "eval_loss": 0.731360137462616,
+      "eval_runtime": 125.1906,
+      "eval_samples_per_second": 55.827,
+      "eval_steps_per_second": 6.981,
+      "step": 1700
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00017494138335287222,
+      "loss": 0.5538,
+      "step": 1710
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00017479484173505276,
+      "loss": 0.4408,
+      "step": 1720
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0001746483001172333,
+      "loss": 0.5641,
+      "step": 1730
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00017450175849941383,
+      "loss": 0.555,
+      "step": 1740
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00017435521688159437,
+      "loss": 0.4495,
+      "step": 1750
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0001742086752637749,
+      "loss": 0.4559,
+      "step": 1760
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00017406213364595544,
+      "loss": 0.6162,
+      "step": 1770
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.000173915592028136,
+      "loss": 0.4602,
+      "step": 1780
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00017376905041031654,
+      "loss": 0.3142,
+      "step": 1790
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00017362250879249708,
+      "loss": 0.6063,
+      "step": 1800
+    },
+    {
+      "epoch": 1.06,
+      "eval_accuracy": 0.8068393189297467,
+      "eval_loss": 0.6312341690063477,
+      "eval_runtime": 132.7037,
+      "eval_samples_per_second": 52.666,
+      "eval_steps_per_second": 6.586,
+      "step": 1800
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00017347596717467761,
+      "loss": 0.4283,
+      "step": 1810
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00017332942555685815,
+      "loss": 0.333,
+      "step": 1820
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0001731828839390387,
+      "loss": 0.4981,
+      "step": 1830
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00017303634232121922,
+      "loss": 0.4265,
+      "step": 1840
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00017288980070339976,
+      "loss": 0.5487,
+      "step": 1850
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00017274325908558032,
+      "loss": 0.48,
+      "step": 1860
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00017259671746776086,
+      "loss": 0.5629,
+      "step": 1870
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0001724501758499414,
+      "loss": 0.7469,
+      "step": 1880
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00017230363423212193,
+      "loss": 0.4352,
+      "step": 1890
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00017215709261430247,
+      "loss": 0.4797,
+      "step": 1900
+    },
+    {
+      "epoch": 1.11,
+      "eval_accuracy": 0.8295893547002432,
+      "eval_loss": 0.5791635513305664,
+      "eval_runtime": 125.6253,
+      "eval_samples_per_second": 55.634,
+      "eval_steps_per_second": 6.957,
+      "step": 1900
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.000172010550996483,
+      "loss": 0.556,
+      "step": 1910
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00017186400937866357,
+      "loss": 0.3858,
+      "step": 1920
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0001717174677608441,
+      "loss": 0.3804,
+      "step": 1930
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00017157092614302464,
+      "loss": 0.4595,
+      "step": 1940
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00017142438452520518,
+      "loss": 0.3698,
+      "step": 1950
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00017127784290738572,
+      "loss": 0.3749,
+      "step": 1960
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00017113130128956625,
+      "loss": 0.478,
+      "step": 1970
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0001709847596717468,
+      "loss": 0.4529,
+      "step": 1980
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00017083821805392732,
+      "loss": 0.3144,
+      "step": 1990
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00017069167643610786,
+      "loss": 0.4973,
+      "step": 2000
+    },
+    {
+      "epoch": 1.17,
+      "eval_accuracy": 0.8221490914293891,
+      "eval_loss": 0.5846105217933655,
+      "eval_runtime": 124.9156,
+      "eval_samples_per_second": 55.95,
+      "eval_steps_per_second": 6.997,
+      "step": 2000
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0001705451348182884,
+      "loss": 0.4261,
+      "step": 2010
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00017039859320046893,
+      "loss": 0.4794,
+      "step": 2020
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00017025205158264947,
+      "loss": 0.6335,
+      "step": 2030
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00017010550996483,
+      "loss": 0.4208,
+      "step": 2040
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00016995896834701054,
+      "loss": 0.4563,
+      "step": 2050
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00016981242672919108,
+      "loss": 0.434,
+      "step": 2060
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00016966588511137164,
+      "loss": 0.4537,
+      "step": 2070
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00016951934349355218,
+      "loss": 0.6688,
+      "step": 2080
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00016937280187573272,
+      "loss": 0.6074,
+      "step": 2090
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00016922626025791325,
+      "loss": 0.4432,
+      "step": 2100
+    },
+    {
+      "epoch": 1.23,
+      "eval_accuracy": 0.7905279725282587,
+      "eval_loss": 0.7057243585586548,
+      "eval_runtime": 124.8847,
+      "eval_samples_per_second": 55.964,
+      "eval_steps_per_second": 6.998,
+      "step": 2100
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0001690797186400938,
+      "loss": 0.5777,
+      "step": 2110
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00016893317702227432,
+      "loss": 0.492,
+      "step": 2120
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00016878663540445486,
+      "loss": 0.4781,
+      "step": 2130
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00016864009378663542,
+      "loss": 0.459,
+      "step": 2140
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00016849355216881596,
+      "loss": 0.4623,
+      "step": 2150
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0001683470105509965,
+      "loss": 0.4226,
+      "step": 2160
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00016821512309495896,
+      "loss": 0.4729,
+      "step": 2170
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0001680685814771395,
+      "loss": 0.4572,
+      "step": 2180
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00016792203985932004,
+      "loss": 0.5003,
+      "step": 2190
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00016777549824150057,
+      "loss": 0.5518,
+      "step": 2200
+    },
+    {
+      "epoch": 1.29,
+      "eval_accuracy": 0.830447846616111,
+      "eval_loss": 0.5621276497840881,
+      "eval_runtime": 131.9063,
+      "eval_samples_per_second": 52.985,
+      "eval_steps_per_second": 6.626,
+      "step": 2200
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00016762895662368114,
+      "loss": 0.4762,
+      "step": 2210
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00016748241500586167,
+      "loss": 0.4631,
+      "step": 2220
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0001673358733880422,
+      "loss": 0.3945,
+      "step": 2230
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00016718933177022275,
+      "loss": 0.5351,
+      "step": 2240
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00016704279015240328,
+      "loss": 0.3174,
+      "step": 2250
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00016689624853458382,
+      "loss": 0.4819,
+      "step": 2260
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00016674970691676438,
+      "loss": 0.4126,
+      "step": 2270
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00016660316529894492,
+      "loss": 0.3867,
+      "step": 2280
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00016645662368112546,
+      "loss": 0.3899,
+      "step": 2290
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.000166310082063306,
+      "loss": 0.3256,
+      "step": 2300
+    },
+    {
+      "epoch": 1.35,
+      "eval_accuracy": 0.8142795822006009,
+      "eval_loss": 0.5890085101127625,
+      "eval_runtime": 132.1089,
+      "eval_samples_per_second": 52.903,
+      "eval_steps_per_second": 6.616,
+      "step": 2300
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00016616354044548653,
+      "loss": 0.3759,
+      "step": 2310
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00016601699882766706,
+      "loss": 0.3657,
+      "step": 2320
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0001658704572098476,
+      "loss": 0.4711,
+      "step": 2330
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00016572391559202814,
+      "loss": 0.5025,
+      "step": 2340
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0001655773739742087,
+      "loss": 0.447,
+      "step": 2350
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00016543083235638924,
+      "loss": 0.4121,
+      "step": 2360
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00016528429073856977,
+      "loss": 0.3962,
+      "step": 2370
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0001651377491207503,
+      "loss": 0.417,
+      "step": 2380
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00016499120750293085,
+      "loss": 0.3845,
+      "step": 2390
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00016484466588511138,
+      "loss": 0.4284,
+      "step": 2400
+    },
+    {
+      "epoch": 1.41,
+      "eval_accuracy": 0.8484761768493346,
+      "eval_loss": 0.5204383730888367,
+      "eval_runtime": 132.2063,
+      "eval_samples_per_second": 52.864,
+      "eval_steps_per_second": 6.611,
+      "step": 2400
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00016469812426729192,
+      "loss": 0.4315,
+      "step": 2410
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00016455158264947246,
+      "loss": 0.4059,
+      "step": 2420
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.000164405041031653,
+      "loss": 0.5097,
+      "step": 2430
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00016425849941383353,
+      "loss": 0.4221,
+      "step": 2440
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00016411195779601406,
+      "loss": 0.3041,
+      "step": 2450
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0001639654161781946,
+      "loss": 0.3355,
+      "step": 2460
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00016381887456037514,
+      "loss": 0.4833,
+      "step": 2470
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00016367233294255567,
+      "loss": 0.3784,
+      "step": 2480
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00016352579132473624,
+      "loss": 0.4241,
+      "step": 2490
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00016337924970691677,
+      "loss": 0.3702,
+      "step": 2500
+    },
+    {
+      "epoch": 1.47,
+      "eval_accuracy": 0.8255830590928602,
+      "eval_loss": 0.5698599815368652,
+      "eval_runtime": 132.0309,
+      "eval_samples_per_second": 52.935,
+      "eval_steps_per_second": 6.62,
+      "step": 2500
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0001632327080890973,
+      "loss": 0.5582,
+      "step": 2510
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00016308616647127785,
+      "loss": 0.4042,
+      "step": 2520
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00016293962485345838,
+      "loss": 0.521,
+      "step": 2530
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00016279308323563892,
+      "loss": 0.4506,
+      "step": 2540
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00016264654161781948,
+      "loss": 0.4792,
+      "step": 2550
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00016250000000000002,
+      "loss": 0.3871,
+      "step": 2560
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00016235345838218056,
+      "loss": 0.4158,
+      "step": 2570
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0001622069167643611,
+      "loss": 0.344,
+      "step": 2580
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00016206037514654163,
+      "loss": 0.443,
+      "step": 2590
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00016191383352872217,
+      "loss": 0.2858,
+      "step": 2600
+    },
+    {
+      "epoch": 1.52,
+      "eval_accuracy": 0.8287308627843755,
+      "eval_loss": 0.5814529061317444,
+      "eval_runtime": 131.7574,
+      "eval_samples_per_second": 53.044,
+      "eval_steps_per_second": 6.633,
+      "step": 2600
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0001617672919109027,
+      "loss": 0.4386,
+      "step": 2610
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00016162075029308324,
+      "loss": 0.4039,
+      "step": 2620
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00016147420867526377,
+      "loss": 0.3744,
+      "step": 2630
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00016132766705744434,
+      "loss": 0.3651,
+      "step": 2640
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00016118112543962487,
+      "loss": 0.3443,
+      "step": 2650
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0001610345838218054,
+      "loss": 0.2685,
+      "step": 2660
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00016088804220398595,
+      "loss": 0.3829,
+      "step": 2670
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00016075615474794844,
+      "loss": 0.6062,
+      "step": 2680
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00016060961313012898,
+      "loss": 0.4935,
+      "step": 2690
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00016046307151230951,
+      "loss": 0.3706,
+      "step": 2700
+    },
+    {
+      "epoch": 1.58,
+      "eval_accuracy": 0.8570610960080126,
+      "eval_loss": 0.4615406095981598,
+      "eval_runtime": 132.121,
+      "eval_samples_per_second": 52.898,
+      "eval_steps_per_second": 6.615,
+      "step": 2700
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00016031652989449005,
+      "loss": 0.3985,
+      "step": 2710
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0001601699882766706,
+      "loss": 0.455,
+      "step": 2720
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00016002344665885112,
+      "loss": 0.4864,
+      "step": 2730
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00015987690504103166,
+      "loss": 0.4129,
+      "step": 2740
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0001597303634232122,
+      "loss": 0.4379,
+      "step": 2750
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00015958382180539273,
+      "loss": 0.368,
+      "step": 2760
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00015943728018757327,
+      "loss": 0.4663,
+      "step": 2770
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00015929073856975383,
+      "loss": 0.4286,
+      "step": 2780
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00015914419695193437,
+      "loss": 0.4098,
+      "step": 2790
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0001589976553341149,
+      "loss": 0.3484,
+      "step": 2800
+    },
+    {
+      "epoch": 1.64,
+      "eval_accuracy": 0.8517670625268279,
+      "eval_loss": 0.4811908006668091,
+      "eval_runtime": 132.1551,
+      "eval_samples_per_second": 52.885,
+      "eval_steps_per_second": 6.613,
+      "step": 2800
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00015885111371629544,
+      "loss": 0.4884,
+      "step": 2810
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00015870457209847598,
+      "loss": 0.4281,
+      "step": 2820
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00015855803048065651,
+      "loss": 0.332,
+      "step": 2830
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00015841148886283705,
+      "loss": 0.4424,
+      "step": 2840
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0001582649472450176,
+      "loss": 0.3052,
+      "step": 2850
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00015811840562719812,
+      "loss": 0.3357,
+      "step": 2860
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00015797186400937866,
+      "loss": 0.3333,
+      "step": 2870
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0001578253223915592,
+      "loss": 0.3892,
+      "step": 2880
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00015767878077373973,
+      "loss": 0.291,
+      "step": 2890
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00015753223915592027,
+      "loss": 0.2865,
+      "step": 2900
+    },
+    {
+      "epoch": 1.7,
+      "eval_accuracy": 0.863785949348977,
+      "eval_loss": 0.42850247025489807,
+      "eval_runtime": 132.124,
+      "eval_samples_per_second": 52.897,
+      "eval_steps_per_second": 6.615,
+      "step": 2900
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0001573856975381008,
+      "loss": 0.5337,
+      "step": 2910
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00015723915592028137,
+      "loss": 0.3917,
+      "step": 2920
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0001570926143024619,
+      "loss": 0.3865,
+      "step": 2930
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00015694607268464244,
+      "loss": 0.2454,
+      "step": 2940
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00015679953106682298,
+      "loss": 0.3168,
+      "step": 2950
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00015665298944900354,
+      "loss": 0.5034,
+      "step": 2960
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00015650644783118408,
+      "loss": 0.4298,
+      "step": 2970
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00015635990621336461,
+      "loss": 0.3871,
+      "step": 2980
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00015621336459554515,
+      "loss": 0.3953,
+      "step": 2990
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0001560668229777257,
+      "loss": 0.4474,
+      "step": 3000
+    },
+    {
+      "epoch": 1.76,
+      "eval_accuracy": 0.8377450279009873,
+      "eval_loss": 0.5217348337173462,
+      "eval_runtime": 131.5378,
+      "eval_samples_per_second": 53.133,
+      "eval_steps_per_second": 6.644,
+      "step": 3000
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00015592028135990622,
+      "loss": 0.4237,
+      "step": 3010
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00015577373974208676,
+      "loss": 0.5681,
+      "step": 3020
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0001556271981242673,
+      "loss": 0.4514,
+      "step": 3030
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00015548065650644783,
+      "loss": 0.4116,
+      "step": 3040
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00015533411488862837,
+      "loss": 0.2819,
+      "step": 3050
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00015518757327080893,
+      "loss": 0.3867,
+      "step": 3060
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00015504103165298947,
+      "loss": 0.3764,
+      "step": 3070
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00015489449003517,
+      "loss": 0.2372,
+      "step": 3080
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00015474794841735054,
+      "loss": 0.4577,
+      "step": 3090
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00015460140679953108,
+      "loss": 0.2101,
+      "step": 3100
+    },
+    {
+      "epoch": 1.82,
+      "eval_accuracy": 0.8589211618257261,
+      "eval_loss": 0.447763592004776,
+      "eval_runtime": 125.3105,
+      "eval_samples_per_second": 55.773,
+      "eval_steps_per_second": 6.975,
+      "step": 3100
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00015445486518171161,
+      "loss": 0.2983,
+      "step": 3110
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00015430832356389215,
+      "loss": 0.3625,
+      "step": 3120
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0001541617819460727,
+      "loss": 0.3615,
+      "step": 3130
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00015401524032825322,
+      "loss": 0.3006,
+      "step": 3140
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00015386869871043376,
+      "loss": 0.3519,
+      "step": 3150
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0001537221570926143,
+      "loss": 0.437,
+      "step": 3160
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00015357561547479483,
+      "loss": 0.2969,
+      "step": 3170
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00015342907385697537,
+      "loss": 0.31,
+      "step": 3180
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0001532825322391559,
+      "loss": 0.3726,
+      "step": 3190
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00015313599062133644,
+      "loss": 0.3545,
+      "step": 3200
+    },
+    {
+      "epoch": 1.88,
+      "eval_accuracy": 0.8612104736013736,
+      "eval_loss": 0.4443872272968292,
+      "eval_runtime": 125.1071,
+      "eval_samples_per_second": 55.864,
+      "eval_steps_per_second": 6.986,
+      "step": 3200
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.000152989449003517,
+      "loss": 0.3303,
+      "step": 3210
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00015284290738569754,
+      "loss": 0.3757,
+      "step": 3220
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00015269636576787808,
+      "loss": 0.245,
+      "step": 3230
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00015254982415005864,
+      "loss": 0.3251,
+      "step": 3240
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00015240328253223918,
+      "loss": 0.2172,
+      "step": 3250
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00015225674091441972,
+      "loss": 0.3141,
+      "step": 3260
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00015211019929660025,
+      "loss": 0.3574,
+      "step": 3270
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0001519636576787808,
+      "loss": 0.4054,
+      "step": 3280
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00015181711606096132,
+      "loss": 0.4211,
+      "step": 3290
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00015167057444314186,
+      "loss": 0.2728,
+      "step": 3300
+    },
+    {
+      "epoch": 1.93,
+      "eval_accuracy": 0.8645013592788668,
+      "eval_loss": 0.42130500078201294,
+      "eval_runtime": 123.9846,
+      "eval_samples_per_second": 56.37,
+      "eval_steps_per_second": 7.049,
+      "step": 3300
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0001515240328253224,
+      "loss": 0.2755,
+      "step": 3310
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00015137749120750293,
+      "loss": 0.3201,
+      "step": 3320
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00015123094958968347,
+      "loss": 0.2831,
+      "step": 3330
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.000151084407971864,
+      "loss": 0.2945,
+      "step": 3340
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00015093786635404457,
+      "loss": 0.3272,
+      "step": 3350
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0001507913247362251,
+      "loss": 0.4223,
+      "step": 3360
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00015064478311840564,
+      "loss": 0.3458,
+      "step": 3370
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00015049824150058618,
+      "loss": 0.3664,
+      "step": 3380
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00015035169988276672,
+      "loss": 0.408,
+      "step": 3390
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00015020515826494725,
+      "loss": 0.3525,
+      "step": 3400
+    },
+    {
+      "epoch": 1.99,
+      "eval_accuracy": 0.8848190012877378,
+      "eval_loss": 0.3551097512245178,
+      "eval_runtime": 125.3559,
+      "eval_samples_per_second": 55.753,
+      "eval_steps_per_second": 6.972,
+      "step": 3400
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0001500586166471278,
+      "loss": 0.2747,
+      "step": 3410
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00014991207502930832,
+      "loss": 0.2152,
+      "step": 3420
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.00014976553341148886,
+      "loss": 0.1515,
+      "step": 3430
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.0001496189917936694,
+      "loss": 0.1195,
+      "step": 3440
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.00014947245017584993,
+      "loss": 0.1625,
+      "step": 3450
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00014932590855803047,
+      "loss": 0.2271,
+      "step": 3460
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.000149179366940211,
+      "loss": 0.1916,
+      "step": 3470
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.00014903282532239154,
+      "loss": 0.3051,
+      "step": 3480
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.0001488862837045721,
+      "loss": 0.1721,
+      "step": 3490
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00014873974208675264,
+      "loss": 0.0936,
+      "step": 3500
+    },
+    {
+      "epoch": 2.05,
+      "eval_accuracy": 0.8748032622692803,
+      "eval_loss": 0.4074080288410187,
+      "eval_runtime": 124.7781,
+      "eval_samples_per_second": 56.011,
+      "eval_steps_per_second": 7.004,
+      "step": 3500
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.00014859320046893318,
+      "loss": 0.2097,
+      "step": 3510
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.00014844665885111372,
+      "loss": 0.2946,
+      "step": 3520
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.00014830011723329428,
+      "loss": 0.1574,
+      "step": 3530
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.00014815357561547482,
+      "loss": 0.2264,
+      "step": 3540
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.00014800703399765535,
+      "loss": 0.3075,
+      "step": 3550
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.0001478604923798359,
+      "loss": 0.1412,
+      "step": 3560
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.00014771395076201643,
+      "loss": 0.1275,
+      "step": 3570
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.00014756740914419696,
+      "loss": 0.2044,
+      "step": 3580
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0001474208675263775,
+      "loss": 0.1696,
+      "step": 3590
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.00014727432590855803,
+      "loss": 0.2118,
+      "step": 3600
+    },
+    {
+      "epoch": 2.11,
+      "eval_accuracy": 0.8812419516382888,
+      "eval_loss": 0.4089324474334717,
+      "eval_runtime": 131.9917,
+      "eval_samples_per_second": 52.95,
+      "eval_steps_per_second": 6.622,
+      "step": 3600
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.00014712778429073857,
+      "loss": 0.2358,
+      "step": 3610
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.0001469812426729191,
+      "loss": 0.1418,
+      "step": 3620
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.00014683470105509967,
+      "loss": 0.2353,
+      "step": 3630
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0001466881594372802,
+      "loss": 0.1102,
+      "step": 3640
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.00014654161781946074,
+      "loss": 0.1651,
+      "step": 3650
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.00014639507620164128,
+      "loss": 0.2924,
+      "step": 3660
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.00014624853458382182,
+      "loss": 0.2254,
+      "step": 3670
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.00014610199296600235,
+      "loss": 0.2249,
+      "step": 3680
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.0001459554513481829,
+      "loss": 0.1807,
+      "step": 3690
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00014580890973036343,
+      "loss": 0.2744,
+      "step": 3700
+    },
+    {
+      "epoch": 2.17,
+      "eval_accuracy": 0.8893976248390327,
+      "eval_loss": 0.35342270135879517,
+      "eval_runtime": 131.7411,
+      "eval_samples_per_second": 53.051,
+      "eval_steps_per_second": 6.634,
+      "step": 3700
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00014566236811254396,
+      "loss": 0.118,
+      "step": 3710
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.0001455158264947245,
+      "loss": 0.1974,
+      "step": 3720
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00014536928487690504,
+      "loss": 0.3136,
+      "step": 3730
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00014522274325908557,
+      "loss": 0.2877,
+      "step": 3740
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.0001450762016412661,
+      "loss": 0.2308,
+      "step": 3750
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.00014492966002344664,
+      "loss": 0.3069,
+      "step": 3760
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.0001447831184056272,
+      "loss": 0.2199,
+      "step": 3770
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.00014463657678780774,
+      "loss": 0.2017,
+      "step": 3780
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.00014449003516998828,
+      "loss": 0.2587,
+      "step": 3790
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00014434349355216882,
+      "loss": 0.211,
+      "step": 3800
+    },
+    {
+      "epoch": 2.23,
+      "eval_accuracy": 0.8599227357275719,
+      "eval_loss": 0.44224393367767334,
+      "eval_runtime": 125.0237,
+      "eval_samples_per_second": 55.901,
+      "eval_steps_per_second": 6.991,
+      "step": 3800
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00014419695193434935,
+      "loss": 0.2985,
+      "step": 3810
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.00014405041031652992,
+      "loss": 0.2918,
+      "step": 3820
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.00014390386869871045,
+      "loss": 0.1259,
+      "step": 3830
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.000143757327080891,
+      "loss": 0.2545,
+      "step": 3840
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.00014361078546307153,
+      "loss": 0.2291,
+      "step": 3850
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.00014346424384525206,
+      "loss": 0.1138,
+      "step": 3860
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.0001433177022274326,
+      "loss": 0.1128,
+      "step": 3870
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.00014317116060961314,
+      "loss": 0.1546,
+      "step": 3880
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00014302461899179367,
+      "loss": 0.2188,
+      "step": 3890
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0001428780773739742,
+      "loss": 0.1684,
+      "step": 3900
+    },
+    {
+      "epoch": 2.29,
+      "eval_accuracy": 0.8858205751895837,
+      "eval_loss": 0.3704982399940491,
+      "eval_runtime": 131.7337,
+      "eval_samples_per_second": 53.054,
+      "eval_steps_per_second": 6.635,
+      "step": 3900
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.00014273153575615474,
+      "loss": 0.2303,
+      "step": 3910
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0001425849941383353,
+      "loss": 0.1955,
+      "step": 3920
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.00014243845252051584,
+      "loss": 0.3027,
+      "step": 3930
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.00014229191090269638,
+      "loss": 0.2722,
+      "step": 3940
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00014214536928487692,
+      "loss": 0.1931,
+      "step": 3950
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00014199882766705745,
+      "loss": 0.1963,
+      "step": 3960
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.000141852286049238,
+      "loss": 0.1625,
+      "step": 3970
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.00014170574443141853,
+      "loss": 0.1784,
+      "step": 3980
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00014155920281359906,
+      "loss": 0.1873,
+      "step": 3990
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.0001414126611957796,
+      "loss": 0.1885,
+      "step": 4000
+    },
+    {
+      "epoch": 2.34,
+      "eval_accuracy": 0.8862498211475175,
+      "eval_loss": 0.3651321530342102,
+      "eval_runtime": 133.5681,
+      "eval_samples_per_second": 52.325,
+      "eval_steps_per_second": 6.543,
+      "step": 4000
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.00014126611957796014,
+      "loss": 0.2071,
+      "step": 4010
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.00014111957796014067,
+      "loss": 0.2175,
+      "step": 4020
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0001409730363423212,
+      "loss": 0.2051,
+      "step": 4030
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.00014082649472450175,
+      "loss": 0.2552,
+      "step": 4040
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.00014067995310668228,
+      "loss": 0.1143,
+      "step": 4050
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.00014053341148886284,
+      "loss": 0.2751,
+      "step": 4060
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00014038686987104338,
+      "loss": 0.2387,
+      "step": 4070
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00014024032825322392,
+      "loss": 0.2206,
+      "step": 4080
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00014009378663540445,
+      "loss": 0.1251,
+      "step": 4090
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00013994724501758502,
+      "loss": 0.249,
+      "step": 4100
+    },
+    {
+      "epoch": 2.4,
+      "eval_accuracy": 0.8686507368722278,
+      "eval_loss": 0.4234098494052887,
+      "eval_runtime": 131.053,
+      "eval_samples_per_second": 53.33,
+      "eval_steps_per_second": 6.669,
+      "step": 4100
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.00013980070339976555,
+      "loss": 0.3265,
+      "step": 4110
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0001396541617819461,
+      "loss": 0.1465,
+      "step": 4120
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.00013950762016412663,
+      "loss": 0.1933,
+      "step": 4130
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.00013936107854630716,
+      "loss": 0.1933,
+      "step": 4140
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0001392145369284877,
+      "loss": 0.3512,
+      "step": 4150
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.00013906799531066824,
+      "loss": 0.1803,
+      "step": 4160
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.00013892145369284877,
+      "loss": 0.1808,
+      "step": 4170
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.0001387749120750293,
+      "loss": 0.2828,
+      "step": 4180
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.00013862837045720985,
+      "loss": 0.1576,
+      "step": 4190
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0001384818288393904,
+      "loss": 0.1485,
+      "step": 4200
+    },
+    {
+      "epoch": 2.46,
+      "eval_accuracy": 0.8798111317785091,
+      "eval_loss": 0.3784010708332062,
+      "eval_runtime": 131.7299,
+      "eval_samples_per_second": 53.056,
+      "eval_steps_per_second": 6.635,
+      "step": 4200
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00013833528722157095,
+      "loss": 0.2885,
+      "step": 4210
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00013818874560375148,
+      "loss": 0.2119,
+      "step": 4220
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.00013804220398593202,
+      "loss": 0.2142,
+      "step": 4230
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.00013789566236811255,
+      "loss": 0.1219,
+      "step": 4240
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0001377491207502931,
+      "loss": 0.2116,
+      "step": 4250
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.00013760257913247363,
+      "loss": 0.0718,
+      "step": 4260
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.00013745603751465416,
+      "loss": 0.1567,
+      "step": 4270
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0001373094958968347,
+      "loss": 0.1922,
+      "step": 4280
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.00013716295427901524,
+      "loss": 0.1765,
+      "step": 4290
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.00013701641266119577,
+      "loss": 0.1188,
+      "step": 4300
+    },
+    {
+      "epoch": 2.52,
+      "eval_accuracy": 0.8872513950493632,
+      "eval_loss": 0.3589094579219818,
+      "eval_runtime": 132.0311,
+      "eval_samples_per_second": 52.934,
+      "eval_steps_per_second": 6.62,
+      "step": 4300
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.0001368698710433763,
+      "loss": 0.0609,
+      "step": 4310
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.00013672332942555685,
+      "loss": 0.139,
+      "step": 4320
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00013657678780773738,
+      "loss": 0.1079,
+      "step": 4330
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00013643024618991795,
+      "loss": 0.1031,
+      "step": 4340
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.00013628370457209848,
+      "loss": 0.2559,
+      "step": 4350
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00013613716295427902,
+      "loss": 0.2049,
+      "step": 4360
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00013599062133645955,
+      "loss": 0.2282,
+      "step": 4370
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0001358440797186401,
+      "loss": 0.2482,
+      "step": 4380
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.00013569753810082065,
+      "loss": 0.216,
+      "step": 4390
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0001355509964830012,
+      "loss": 0.1274,
+      "step": 4400
+    },
+    {
+      "epoch": 2.58,
+      "eval_accuracy": 0.8916869366146802,
+      "eval_loss": 0.3570205867290497,
+      "eval_runtime": 131.9667,
+      "eval_samples_per_second": 52.96,
+      "eval_steps_per_second": 6.623,
+      "step": 4400
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.00013540445486518173,
+      "loss": 0.0962,
+      "step": 4410
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.00013525791324736226,
+      "loss": 0.1757,
+      "step": 4420
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.0001351113716295428,
+      "loss": 0.21,
+      "step": 4430
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00013496483001172334,
+      "loss": 0.1687,
+      "step": 4440
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.00013481828839390387,
+      "loss": 0.1697,
+      "step": 4450
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0001346717467760844,
+      "loss": 0.1938,
+      "step": 4460
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00013452520515826495,
+      "loss": 0.1768,
+      "step": 4470
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.00013437866354044548,
+      "loss": 0.1346,
+      "step": 4480
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.00013423212192262605,
+      "loss": 0.1008,
+      "step": 4490
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00013408558030480658,
+      "loss": 0.2206,
+      "step": 4500
+    },
+    {
+      "epoch": 2.64,
+      "eval_accuracy": 0.8919731005866361,
+      "eval_loss": 0.33766970038414,
+      "eval_runtime": 125.8837,
+      "eval_samples_per_second": 55.52,
+      "eval_steps_per_second": 6.943,
+      "step": 4500
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00013393903868698712,
+      "loss": 0.0995,
+      "step": 4510
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.00013379249706916766,
+      "loss": 0.1878,
+      "step": 4520
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.0001336459554513482,
+      "loss": 0.2319,
+      "step": 4530
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.00013349941383352873,
+      "loss": 0.2662,
+      "step": 4540
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.00013335287221570926,
+      "loss": 0.1882,
+      "step": 4550
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0001332063305978898,
+      "loss": 0.2197,
+      "step": 4560
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00013305978898007034,
+      "loss": 0.1936,
+      "step": 4570
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00013291324736225087,
+      "loss": 0.1674,
+      "step": 4580
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.0001327667057444314,
+      "loss": 0.1732,
+      "step": 4590
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00013262016412661195,
+      "loss": 0.1287,
+      "step": 4600
+    },
+    {
+      "epoch": 2.7,
+      "eval_accuracy": 0.9022750035770496,
+      "eval_loss": 0.3170166611671448,
+      "eval_runtime": 132.1639,
+      "eval_samples_per_second": 52.881,
+      "eval_steps_per_second": 6.613,
+      "step": 4600
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.00013247362250879248,
+      "loss": 0.3019,
+      "step": 4610
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.00013232708089097302,
+      "loss": 0.1496,
+      "step": 4620
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.00013218053927315358,
+      "loss": 0.0726,
+      "step": 4630
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00013203399765533412,
+      "loss": 0.1167,
+      "step": 4640
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.00013188745603751466,
+      "loss": 0.2261,
+      "step": 4650
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0001317409144196952,
+      "loss": 0.2373,
+      "step": 4660
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.00013159437280187576,
+      "loss": 0.1427,
+      "step": 4670
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.0001314478311840563,
+      "loss": 0.2416,
+      "step": 4680
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00013130128956623683,
+      "loss": 0.15,
+      "step": 4690
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00013115474794841736,
+      "loss": 0.1805,
+      "step": 4700
+    },
+    {
+      "epoch": 2.75,
+      "eval_accuracy": 0.8934039204464158,
+      "eval_loss": 0.34686848521232605,
+      "eval_runtime": 125.0204,
+      "eval_samples_per_second": 55.903,
+      "eval_steps_per_second": 6.991,
+      "step": 4700
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.0001310082063305979,
+      "loss": 0.2412,
+      "step": 4710
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00013086166471277844,
+      "loss": 0.1448,
+      "step": 4720
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00013071512309495897,
+      "loss": 0.2682,
+      "step": 4730
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0001305685814771395,
+      "loss": 0.1454,
+      "step": 4740
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.00013042203985932005,
+      "loss": 0.1455,
+      "step": 4750
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.00013027549824150058,
+      "loss": 0.1965,
+      "step": 4760
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.00013012895662368115,
+      "loss": 0.1012,
+      "step": 4770
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.00012998241500586168,
+      "loss": 0.1611,
+      "step": 4780
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00012983587338804222,
+      "loss": 0.1632,
+      "step": 4790
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00012968933177022276,
+      "loss": 0.1505,
+      "step": 4800
+    },
+    {
+      "epoch": 2.81,
+      "eval_accuracy": 0.8756617541851481,
+      "eval_loss": 0.42582932114601135,
+      "eval_runtime": 131.5209,
+      "eval_samples_per_second": 53.14,
+      "eval_steps_per_second": 6.645,
+      "step": 4800
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.0001295427901524033,
+      "loss": 0.1077,
+      "step": 4810
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00012939624853458383,
+      "loss": 0.2108,
+      "step": 4820
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.00012924970691676437,
+      "loss": 0.0965,
+      "step": 4830
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.0001291031652989449,
+      "loss": 0.2512,
+      "step": 4840
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.00012895662368112544,
+      "loss": 0.1869,
+      "step": 4850
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.00012881008206330597,
+      "loss": 0.1474,
+      "step": 4860
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0001286635404454865,
+      "loss": 0.1372,
+      "step": 4870
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.00012851699882766705,
+      "loss": 0.1315,
+      "step": 4880
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.00012837045720984758,
+      "loss": 0.1833,
+      "step": 4890
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.00012822391559202812,
+      "loss": 0.1592,
+      "step": 4900
+    },
+    {
+      "epoch": 2.87,
+      "eval_accuracy": 0.8948347403061955,
+      "eval_loss": 0.3414626121520996,
+      "eval_runtime": 131.7926,
+      "eval_samples_per_second": 53.03,
+      "eval_steps_per_second": 6.632,
+      "step": 4900
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00012807737397420868,
+      "loss": 0.2438,
+      "step": 4910
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00012793083235638922,
+      "loss": 0.2031,
+      "step": 4920
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.00012778429073856976,
+      "loss": 0.1168,
+      "step": 4930
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.0001276377491207503,
+      "loss": 0.104,
+      "step": 4940
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.00012749120750293083,
+      "loss": 0.1847,
+      "step": 4950
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0001273446658851114,
+      "loss": 0.1623,
+      "step": 4960
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.00012719812426729193,
+      "loss": 0.1452,
+      "step": 4970
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00012705158264947247,
+      "loss": 0.144,
+      "step": 4980
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.000126905041031653,
+      "loss": 0.1342,
+      "step": 4990
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.00012675849941383354,
+      "loss": 0.1297,
+      "step": 5000
+    },
+    {
+      "epoch": 2.93,
+      "eval_accuracy": 0.9028473315209615,
+      "eval_loss": 0.31676042079925537,
+      "eval_runtime": 131.6924,
+      "eval_samples_per_second": 53.071,
+      "eval_steps_per_second": 6.637,
+      "step": 5000
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.00012661195779601407,
+      "loss": 0.2152,
+      "step": 5010
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.0001264654161781946,
+      "loss": 0.1199,
+      "step": 5020
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.00012631887456037515,
+      "loss": 0.1125,
+      "step": 5030
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.00012617233294255568,
+      "loss": 0.21,
+      "step": 5040
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00012602579132473625,
+      "loss": 0.2209,
+      "step": 5050
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00012587924970691678,
+      "loss": 0.1697,
+      "step": 5060
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00012573270808909732,
+      "loss": 0.1927,
+      "step": 5070
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.00012558616647127786,
+      "loss": 0.1699,
+      "step": 5080
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.0001254396248534584,
+      "loss": 0.1926,
+      "step": 5090
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.00012529308323563893,
+      "loss": 0.1284,
+      "step": 5100
+    },
+    {
+      "epoch": 2.99,
+      "eval_accuracy": 0.9088567749320361,
+      "eval_loss": 0.3060346841812134,
+      "eval_runtime": 124.3218,
+      "eval_samples_per_second": 56.217,
+      "eval_steps_per_second": 7.03,
+      "step": 5100
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00012514654161781947,
+      "loss": 0.2588,
+      "step": 5110
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.000125,
+      "loss": 0.2415,
+      "step": 5120
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.00012485345838218054,
+      "loss": 0.1319,
+      "step": 5130
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.00012470691676436108,
+      "loss": 0.0835,
+      "step": 5140
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.0001245603751465416,
+      "loss": 0.1037,
+      "step": 5150
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.00012441383352872215,
+      "loss": 0.0888,
+      "step": 5160
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00012426729191090268,
+      "loss": 0.1312,
+      "step": 5170
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00012412075029308322,
+      "loss": 0.0788,
+      "step": 5180
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.00012397420867526376,
+      "loss": 0.0947,
+      "step": 5190
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00012382766705744432,
+      "loss": 0.0833,
+      "step": 5200
+    },
+    {
+      "epoch": 3.05,
+      "eval_accuracy": 0.9207325797682072,
+      "eval_loss": 0.26095303893089294,
+      "eval_runtime": 132.1262,
+      "eval_samples_per_second": 52.896,
+      "eval_steps_per_second": 6.615,
+      "step": 5200
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00012368112543962486,
+      "loss": 0.099,
+      "step": 5210
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.00012354923798358735,
+      "loss": 0.1089,
+      "step": 5220
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0001234026963657679,
+      "loss": 0.0636,
+      "step": 5230
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.00012325615474794842,
+      "loss": 0.041,
+      "step": 5240
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.00012310961313012896,
+      "loss": 0.093,
+      "step": 5250
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0001229630715123095,
+      "loss": 0.0884,
+      "step": 5260
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.00012281652989449003,
+      "loss": 0.1423,
+      "step": 5270
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.00012266998827667057,
+      "loss": 0.113,
+      "step": 5280
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0001225234466588511,
+      "loss": 0.097,
+      "step": 5290
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.00012237690504103164,
+      "loss": 0.0334,
+      "step": 5300
+    },
+    {
+      "epoch": 3.11,
+      "eval_accuracy": 0.9197310058663615,
+      "eval_loss": 0.27662691473960876,
+      "eval_runtime": 125.074,
+      "eval_samples_per_second": 55.879,
+      "eval_steps_per_second": 6.988,
+      "step": 5300
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.00012223036342321218,
+      "loss": 0.0971,
+      "step": 5310
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.00012208382180539272,
+      "loss": 0.1892,
+      "step": 5320
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.00012193728018757327,
+      "loss": 0.0872,
+      "step": 5330
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.00012179073856975383,
+      "loss": 0.0763,
+      "step": 5340
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00012164419695193437,
+      "loss": 0.0516,
+      "step": 5350
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0001214976553341149,
+      "loss": 0.0301,
+      "step": 5360
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00012135111371629544,
+      "loss": 0.0616,
+      "step": 5370
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00012120457209847597,
+      "loss": 0.064,
+      "step": 5380
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.00012105803048065651,
+      "loss": 0.0979,
+      "step": 5390
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00012091148886283705,
+      "loss": 0.0847,
+      "step": 5400
+    },
+    {
+      "epoch": 3.17,
+      "eval_accuracy": 0.9015595936471599,
+      "eval_loss": 0.3365921974182129,
+      "eval_runtime": 125.0018,
+      "eval_samples_per_second": 55.911,
+      "eval_steps_per_second": 6.992,
+      "step": 5400
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.00012076494724501758,
+      "loss": 0.0439,
+      "step": 5410
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00012063305978898008,
+      "loss": 0.1307,
+      "step": 5420
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.00012048651817116061,
+      "loss": 0.1127,
+      "step": 5430
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.00012033997655334115,
+      "loss": 0.1524,
+      "step": 5440
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.00012019343493552169,
+      "loss": 0.1865,
+      "step": 5450
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.00012004689331770222,
+      "loss": 0.1067,
+      "step": 5460
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.00011990035169988276,
+      "loss": 0.0644,
+      "step": 5470
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.00011975381008206332,
+      "loss": 0.0786,
+      "step": 5480
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.00011960726846424386,
+      "loss": 0.1206,
+      "step": 5490
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0001194607268464244,
+      "loss": 0.1112,
+      "step": 5500
+    },
+    {
+      "epoch": 3.22,
+      "eval_accuracy": 0.9078552010301904,
+      "eval_loss": 0.3097987473011017,
+      "eval_runtime": 131.4659,
+      "eval_samples_per_second": 53.162,
+      "eval_steps_per_second": 6.648,
+      "step": 5500
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.00011931418522860493,
+      "loss": 0.1539,
+      "step": 5510
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00011916764361078547,
+      "loss": 0.1464,
+      "step": 5520
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.000119021101992966,
+      "loss": 0.0947,
+      "step": 5530
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00011887456037514655,
+      "loss": 0.128,
+      "step": 5540
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00011872801875732709,
+      "loss": 0.0248,
+      "step": 5550
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00011858147713950763,
+      "loss": 0.1434,
+      "step": 5560
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00011843493552168816,
+      "loss": 0.0732,
+      "step": 5570
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0001182883939038687,
+      "loss": 0.1121,
+      "step": 5580
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00011814185228604924,
+      "loss": 0.0738,
+      "step": 5590
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00011799531066822977,
+      "loss": 0.0477,
+      "step": 5600
+    },
+    {
+      "epoch": 3.28,
+      "eval_accuracy": 0.9041350693947632,
+      "eval_loss": 0.3384874761104584,
+      "eval_runtime": 132.3074,
+      "eval_samples_per_second": 52.824,
+      "eval_steps_per_second": 6.606,
+      "step": 5600
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.00011784876905041031,
+      "loss": 0.0989,
+      "step": 5610
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.00011770222743259087,
+      "loss": 0.0424,
+      "step": 5620
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.00011755568581477141,
+      "loss": 0.0857,
+      "step": 5630
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.00011740914419695195,
+      "loss": 0.0629,
+      "step": 5640
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.00011726260257913248,
+      "loss": 0.0466,
+      "step": 5650
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00011711606096131302,
+      "loss": 0.0456,
+      "step": 5660
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00011696951934349356,
+      "loss": 0.0598,
+      "step": 5670
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.0001168229777256741,
+      "loss": 0.0721,
+      "step": 5680
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00011667643610785464,
+      "loss": 0.0838,
+      "step": 5690
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.00011652989449003518,
+      "loss": 0.0419,
+      "step": 5700
+    },
+    {
+      "epoch": 3.34,
+      "eval_accuracy": 0.9138646444412648,
+      "eval_loss": 0.2944459617137909,
+      "eval_runtime": 124.8391,
+      "eval_samples_per_second": 55.984,
+      "eval_steps_per_second": 7.001,
+      "step": 5700
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.00011638335287221571,
+      "loss": 0.1409,
+      "step": 5710
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.00011623681125439625,
+      "loss": 0.0575,
+      "step": 5720
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.00011609026963657679,
+      "loss": 0.0739,
+      "step": 5730
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.00011594372801875732,
+      "loss": 0.0674,
+      "step": 5740
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.00011579718640093786,
+      "loss": 0.0607,
+      "step": 5750
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0001156506447831184,
+      "loss": 0.1805,
+      "step": 5760
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.00011550410316529896,
+      "loss": 0.0809,
+      "step": 5770
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0001153575615474795,
+      "loss": 0.0769,
+      "step": 5780
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00011521101992966003,
+      "loss": 0.0258,
+      "step": 5790
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.00011506447831184057,
+      "loss": 0.0827,
+      "step": 5800
+    },
+    {
+      "epoch": 3.4,
+      "eval_accuracy": 0.9238803834597225,
+      "eval_loss": 0.27152764797210693,
+      "eval_runtime": 132.7583,
+      "eval_samples_per_second": 52.645,
+      "eval_steps_per_second": 6.583,
+      "step": 5800
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0001149179366940211,
+      "loss": 0.0716,
+      "step": 5810
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.00011477139507620166,
+      "loss": 0.0639,
+      "step": 5820
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00011462485345838219,
+      "loss": 0.0269,
+      "step": 5830
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00011447831184056273,
+      "loss": 0.0493,
+      "step": 5840
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.00011433177022274326,
+      "loss": 0.0439,
+      "step": 5850
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0001141852286049238,
+      "loss": 0.1426,
+      "step": 5860
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.00011403868698710434,
+      "loss": 0.0994,
+      "step": 5870
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.00011389214536928487,
+      "loss": 0.0267,
+      "step": 5880
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.00011374560375146541,
+      "loss": 0.073,
+      "step": 5890
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00011359906213364595,
+      "loss": 0.0659,
+      "step": 5900
+    },
+    {
+      "epoch": 3.46,
+      "eval_accuracy": 0.9230218915438546,
+      "eval_loss": 0.26954904198646545,
+      "eval_runtime": 133.8759,
+      "eval_samples_per_second": 52.205,
+      "eval_steps_per_second": 6.528,
+      "step": 5900
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00011345252051582651,
+      "loss": 0.0812,
+      "step": 5910
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.00011330597889800705,
+      "loss": 0.1031,
+      "step": 5920
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.00011315943728018758,
+      "loss": 0.1367,
+      "step": 5930
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.00011301289566236812,
+      "loss": 0.0175,
+      "step": 5940
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00011286635404454866,
+      "loss": 0.0395,
+      "step": 5950
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00011271981242672919,
+      "loss": 0.0192,
+      "step": 5960
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.00011257327080890974,
+      "loss": 0.0198,
+      "step": 5970
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.00011242672919109028,
+      "loss": 0.0626,
+      "step": 5980
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.00011228018757327082,
+      "loss": 0.0431,
+      "step": 5990
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00011213364595545135,
+      "loss": 0.0244,
+      "step": 6000
+    },
+    {
+      "epoch": 3.52,
+      "eval_accuracy": 0.9147231363571326,
+      "eval_loss": 0.30501335859298706,
+      "eval_runtime": 131.6894,
+      "eval_samples_per_second": 53.072,
+      "eval_steps_per_second": 6.637,
+      "step": 6000
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00011198710433763189,
+      "loss": 0.0997,
+      "step": 6010
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00011184056271981242,
+      "loss": 0.0872,
+      "step": 6020
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00011169402110199296,
+      "loss": 0.0791,
+      "step": 6030
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.0001115474794841735,
+      "loss": 0.0161,
+      "step": 6040
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.00011140093786635406,
+      "loss": 0.0247,
+      "step": 6050
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0001112543962485346,
+      "loss": 0.0564,
+      "step": 6060
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.00011110785463071513,
+      "loss": 0.1098,
+      "step": 6070
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.00011096131301289567,
+      "loss": 0.0636,
+      "step": 6080
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0001108147713950762,
+      "loss": 0.0881,
+      "step": 6090
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.00011066822977725674,
+      "loss": 0.0883,
+      "step": 6100
+    },
+    {
+      "epoch": 3.58,
+      "eval_accuracy": 0.9203033338102733,
+      "eval_loss": 0.2862415015697479,
+      "eval_runtime": 131.721,
+      "eval_samples_per_second": 53.059,
+      "eval_steps_per_second": 6.635,
+      "step": 6100
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.00011052168815943729,
+      "loss": 0.0294,
+      "step": 6110
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00011037514654161783,
+      "loss": 0.119,
+      "step": 6120
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00011022860492379837,
+      "loss": 0.1097,
+      "step": 6130
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.0001100820633059789,
+      "loss": 0.1704,
+      "step": 6140
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.00010993552168815944,
+      "loss": 0.1924,
+      "step": 6150
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00010978898007033997,
+      "loss": 0.1146,
+      "step": 6160
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.00010964243845252051,
+      "loss": 0.0637,
+      "step": 6170
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.00010949589683470105,
+      "loss": 0.1162,
+      "step": 6180
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.00010934935521688161,
+      "loss": 0.1158,
+      "step": 6190
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.00010920281359906215,
+      "loss": 0.0527,
+      "step": 6200
+    },
+    {
+      "epoch": 3.63,
+      "eval_accuracy": 0.9318929746744885,
+      "eval_loss": 0.23830993473529816,
+      "eval_runtime": 124.9507,
+      "eval_samples_per_second": 55.934,
+      "eval_steps_per_second": 6.995,
+      "step": 6200
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.00010905627198124268,
+      "loss": 0.0568,
+      "step": 6210
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.00010890973036342322,
+      "loss": 0.1423,
+      "step": 6220
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.00010876318874560376,
+      "loss": 0.0232,
+      "step": 6230
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00010861664712778429,
+      "loss": 0.1261,
+      "step": 6240
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00010847010550996484,
+      "loss": 0.0385,
+      "step": 6250
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.00010832356389214538,
+      "loss": 0.0789,
+      "step": 6260
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.00010817702227432592,
+      "loss": 0.109,
+      "step": 6270
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.00010803048065650645,
+      "loss": 0.0935,
+      "step": 6280
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.00010788393903868699,
+      "loss": 0.1142,
+      "step": 6290
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.00010773739742086753,
+      "loss": 0.0828,
+      "step": 6300
+    },
+    {
+      "epoch": 3.69,
+      "eval_accuracy": 0.9181571040206038,
+      "eval_loss": 0.29844602942466736,
+      "eval_runtime": 125.0359,
+      "eval_samples_per_second": 55.896,
+      "eval_steps_per_second": 6.99,
+      "step": 6300
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.00010759085580304806,
+      "loss": 0.0893,
+      "step": 6310
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.0001074443141852286,
+      "loss": 0.0597,
+      "step": 6320
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.00010729777256740913,
+      "loss": 0.0588,
+      "step": 6330
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0001071512309495897,
+      "loss": 0.1045,
+      "step": 6340
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.00010700468933177023,
+      "loss": 0.0383,
+      "step": 6350
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.00010685814771395077,
+      "loss": 0.1054,
+      "step": 6360
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.00010671160609613131,
+      "loss": 0.0814,
+      "step": 6370
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00010656506447831184,
+      "loss": 0.0156,
+      "step": 6380
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0001064185228604924,
+      "loss": 0.0259,
+      "step": 6390
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.00010627198124267293,
+      "loss": 0.0678,
+      "step": 6400
+    },
+    {
+      "epoch": 3.75,
+      "eval_accuracy": 0.9436256975246816,
+      "eval_loss": 0.21348373591899872,
+      "eval_runtime": 125.0033,
+      "eval_samples_per_second": 55.911,
+      "eval_steps_per_second": 6.992,
+      "step": 6400
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00010612543962485347,
+      "loss": 0.0626,
+      "step": 6410
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.000105978898007034,
+      "loss": 0.0556,
+      "step": 6420
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.00010583235638921454,
+      "loss": 0.1722,
+      "step": 6430
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.00010568581477139508,
+      "loss": 0.102,
+      "step": 6440
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.00010553927315357561,
+      "loss": 0.0304,
+      "step": 6450
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.00010539273153575615,
+      "loss": 0.0167,
+      "step": 6460
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.00010524618991793668,
+      "loss": 0.0763,
+      "step": 6470
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.00010509964830011725,
+      "loss": 0.0199,
+      "step": 6480
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.00010495310668229778,
+      "loss": 0.0433,
+      "step": 6490
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00010480656506447832,
+      "loss": 0.0492,
+      "step": 6500
+    },
+    {
+      "epoch": 3.81,
+      "eval_accuracy": 0.929603662898841,
+      "eval_loss": 0.2604553997516632,
+      "eval_runtime": 124.6857,
+      "eval_samples_per_second": 56.053,
+      "eval_steps_per_second": 7.01,
+      "step": 6500
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.00010466002344665886,
+      "loss": 0.1748,
+      "step": 6510
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.0001045134818288394,
+      "loss": 0.1189,
+      "step": 6520
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.00010436694021101993,
+      "loss": 0.0547,
+      "step": 6530
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.00010422039859320048,
+      "loss": 0.1063,
+      "step": 6540
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.00010407385697538102,
+      "loss": 0.0726,
+      "step": 6550
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00010392731535756155,
+      "loss": 0.0414,
+      "step": 6560
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00010378077373974209,
+      "loss": 0.172,
+      "step": 6570
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.00010363423212192263,
+      "loss": 0.052,
+      "step": 6580
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.00010348769050410316,
+      "loss": 0.1037,
+      "step": 6590
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.0001033411488862837,
+      "loss": 0.0374,
+      "step": 6600
+    },
+    {
+      "epoch": 3.87,
+      "eval_accuracy": 0.938045500071541,
+      "eval_loss": 0.21922510862350464,
+      "eval_runtime": 124.604,
+      "eval_samples_per_second": 56.09,
+      "eval_steps_per_second": 7.014,
+      "step": 6600
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.00010319460726846424,
+      "loss": 0.0364,
+      "step": 6610
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.0001030480656506448,
+      "loss": 0.0965,
+      "step": 6620
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00010290152403282534,
+      "loss": 0.0719,
+      "step": 6630
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00010275498241500587,
+      "loss": 0.0484,
+      "step": 6640
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.00010260844079718641,
+      "loss": 0.0606,
+      "step": 6650
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.00010246189917936694,
+      "loss": 0.0999,
+      "step": 6660
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.00010231535756154748,
+      "loss": 0.1628,
+      "step": 6670
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.00010216881594372803,
+      "loss": 0.0442,
+      "step": 6680
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.00010202227432590857,
+      "loss": 0.1007,
+      "step": 6690
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0001018757327080891,
+      "loss": 0.1846,
+      "step": 6700
+    },
+    {
+      "epoch": 3.93,
+      "eval_accuracy": 0.9187294319645156,
+      "eval_loss": 0.28041157126426697,
+      "eval_runtime": 124.644,
+      "eval_samples_per_second": 56.072,
+      "eval_steps_per_second": 7.012,
+      "step": 6700
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.00010172919109026964,
+      "loss": 0.0226,
+      "step": 6710
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.00010158264947245018,
+      "loss": 0.017,
+      "step": 6720
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.00010143610785463071,
+      "loss": 0.0541,
+      "step": 6730
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.00010128956623681125,
+      "loss": 0.0435,
+      "step": 6740
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00010114302461899179,
+      "loss": 0.0226,
+      "step": 6750
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00010099648300117235,
+      "loss": 0.0893,
+      "step": 6760
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00010084994138335289,
+      "loss": 0.0183,
+      "step": 6770
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.00010070339976553342,
+      "loss": 0.0645,
+      "step": 6780
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.00010055685814771396,
+      "loss": 0.085,
+      "step": 6790
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.0001004103165298945,
+      "loss": 0.0557,
+      "step": 6800
+    },
+    {
+      "epoch": 3.99,
+      "eval_accuracy": 0.9253112033195021,
+      "eval_loss": 0.2598896622657776,
+      "eval_runtime": 131.5952,
+      "eval_samples_per_second": 53.11,
+      "eval_steps_per_second": 6.642,
+      "step": 6800
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.00010026377491207503,
+      "loss": 0.0727,
+      "step": 6810
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.00010011723329425558,
+      "loss": 0.0319,
+      "step": 6820
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 9.997069167643612e-05,
+      "loss": 0.1264,
+      "step": 6830
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 9.982415005861665e-05,
+      "loss": 0.0433,
+      "step": 6840
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 9.967760844079719e-05,
+      "loss": 0.0487,
+      "step": 6850
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 9.953106682297773e-05,
+      "loss": 0.01,
+      "step": 6860
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 9.938452520515828e-05,
+      "loss": 0.0413,
+      "step": 6870
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 9.923798358733881e-05,
+      "loss": 0.0272,
+      "step": 6880
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 9.909144196951935e-05,
+      "loss": 0.0356,
+      "step": 6890
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 9.894490035169989e-05,
+      "loss": 0.0127,
+      "step": 6900
+    },
+    {
+      "epoch": 4.04,
+      "eval_accuracy": 0.9336099585062241,
+      "eval_loss": 0.24118462204933167,
+      "eval_runtime": 125.0963,
+      "eval_samples_per_second": 55.869,
+      "eval_steps_per_second": 6.987,
+      "step": 6900
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 9.879835873388042e-05,
+      "loss": 0.0699,
+      "step": 6910
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 9.865181711606096e-05,
+      "loss": 0.0422,
+      "step": 6920
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 9.85052754982415e-05,
+      "loss": 0.0064,
+      "step": 6930
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 9.835873388042205e-05,
+      "loss": 0.0071,
+      "step": 6940
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 9.821219226260258e-05,
+      "loss": 0.0268,
+      "step": 6950
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 9.806565064478312e-05,
+      "loss": 0.0075,
+      "step": 6960
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 9.791910902696367e-05,
+      "loss": 0.1285,
+      "step": 6970
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 9.77725674091442e-05,
+      "loss": 0.0966,
+      "step": 6980
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 9.762602579132474e-05,
+      "loss": 0.0452,
+      "step": 6990
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 9.747948417350528e-05,
+      "loss": 0.0203,
+      "step": 7000
+    },
+    {
+      "epoch": 4.1,
+      "eval_accuracy": 0.9414794677350121,
+      "eval_loss": 0.22137586772441864,
+      "eval_runtime": 124.5679,
+      "eval_samples_per_second": 56.106,
+      "eval_steps_per_second": 7.016,
+      "step": 7000
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 9.733294255568583e-05,
+      "loss": 0.0464,
+      "step": 7010
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 9.718640093786636e-05,
+      "loss": 0.0097,
+      "step": 7020
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 9.70398593200469e-05,
+      "loss": 0.0401,
+      "step": 7030
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 9.689331770222744e-05,
+      "loss": 0.0125,
+      "step": 7040
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 9.674677608440797e-05,
+      "loss": 0.0281,
+      "step": 7050
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 9.660023446658851e-05,
+      "loss": 0.0193,
+      "step": 7060
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 9.645369284876905e-05,
+      "loss": 0.0268,
+      "step": 7070
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 9.63071512309496e-05,
+      "loss": 0.0415,
+      "step": 7080
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 9.616060961313013e-05,
+      "loss": 0.0067,
+      "step": 7090
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 9.601406799531067e-05,
+      "loss": 0.0272,
+      "step": 7100
+    },
+    {
+      "epoch": 4.16,
+      "eval_accuracy": 0.9356131063099156,
+      "eval_loss": 0.23218870162963867,
+      "eval_runtime": 124.8005,
+      "eval_samples_per_second": 56.001,
+      "eval_steps_per_second": 7.003,
+      "step": 7100
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 9.586752637749122e-05,
+      "loss": 0.014,
+      "step": 7110
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 9.572098475967175e-05,
+      "loss": 0.0348,
+      "step": 7120
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 9.557444314185229e-05,
+      "loss": 0.0049,
+      "step": 7130
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 9.542790152403283e-05,
+      "loss": 0.0365,
+      "step": 7140
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 9.528135990621338e-05,
+      "loss": 0.0434,
+      "step": 7150
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 9.513481828839391e-05,
+      "loss": 0.0378,
+      "step": 7160
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 9.498827667057445e-05,
+      "loss": 0.0562,
+      "step": 7170
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 9.484173505275499e-05,
+      "loss": 0.0491,
+      "step": 7180
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 9.469519343493552e-05,
+      "loss": 0.0303,
+      "step": 7190
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 9.454865181711606e-05,
+      "loss": 0.066,
+      "step": 7200
+    },
+    {
+      "epoch": 4.22,
+      "eval_accuracy": 0.9324653026184003,
+      "eval_loss": 0.2642953395843506,
+      "eval_runtime": 125.3377,
+      "eval_samples_per_second": 55.761,
+      "eval_steps_per_second": 6.973,
+      "step": 7200
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 9.44021101992966e-05,
+      "loss": 0.0179,
+      "step": 7210
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 9.425556858147715e-05,
+      "loss": 0.0201,
+      "step": 7220
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 9.410902696365768e-05,
+      "loss": 0.0511,
+      "step": 7230
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 9.396248534583822e-05,
+      "loss": 0.0456,
+      "step": 7240
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 9.381594372801877e-05,
+      "loss": 0.0448,
+      "step": 7250
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 9.36694021101993e-05,
+      "loss": 0.0171,
+      "step": 7260
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 9.352286049237984e-05,
+      "loss": 0.0323,
+      "step": 7270
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 9.337631887456038e-05,
+      "loss": 0.035,
+      "step": 7280
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 9.322977725674091e-05,
+      "loss": 0.035,
+      "step": 7290
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 9.308323563892146e-05,
+      "loss": 0.0628,
+      "step": 7300
+    },
+    {
+      "epoch": 4.28,
+      "eval_accuracy": 0.9406209758191444,
+      "eval_loss": 0.2170339971780777,
+      "eval_runtime": 124.7315,
+      "eval_samples_per_second": 56.032,
+      "eval_steps_per_second": 7.007,
+      "step": 7300
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 9.2936694021102e-05,
+      "loss": 0.0295,
+      "step": 7310
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 9.279015240328254e-05,
+      "loss": 0.0602,
+      "step": 7320
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 9.264361078546307e-05,
+      "loss": 0.0586,
+      "step": 7330
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 9.249706916764361e-05,
+      "loss": 0.0591,
+      "step": 7340
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 9.235052754982415e-05,
+      "loss": 0.0426,
+      "step": 7350
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 9.220398593200468e-05,
+      "loss": 0.0048,
+      "step": 7360
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 9.205744431418523e-05,
+      "loss": 0.0111,
+      "step": 7370
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 9.191090269636577e-05,
+      "loss": 0.0073,
+      "step": 7380
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 9.176436107854632e-05,
+      "loss": 0.0152,
+      "step": 7390
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 9.161781946072686e-05,
+      "loss": 0.0108,
+      "step": 7400
+    },
+    {
+      "epoch": 4.34,
+      "eval_accuracy": 0.9404778938331664,
+      "eval_loss": 0.23883628845214844,
+      "eval_runtime": 131.7955,
+      "eval_samples_per_second": 53.029,
+      "eval_steps_per_second": 6.631,
+      "step": 7400
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 9.147127784290739e-05,
+      "loss": 0.0069,
+      "step": 7410
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 9.132473622508793e-05,
+      "loss": 0.0343,
+      "step": 7420
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 9.117819460726846e-05,
+      "loss": 0.0127,
+      "step": 7430
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 9.103165298944901e-05,
+      "loss": 0.004,
+      "step": 7440
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 9.088511137162955e-05,
+      "loss": 0.0362,
+      "step": 7450
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 9.073856975381009e-05,
+      "loss": 0.0526,
+      "step": 7460
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 9.059202813599062e-05,
+      "loss": 0.0157,
+      "step": 7470
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 9.044548651817116e-05,
+      "loss": 0.0654,
+      "step": 7480
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 9.02989449003517e-05,
+      "loss": 0.002,
+      "step": 7490
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 9.015240328253223e-05,
+      "loss": 0.026,
+      "step": 7500
+    },
+    {
+      "epoch": 4.4,
+      "eval_accuracy": 0.9371870081556732,
+      "eval_loss": 0.2533361315727234,
+      "eval_runtime": 131.9358,
+      "eval_samples_per_second": 52.973,
+      "eval_steps_per_second": 6.624,
+      "step": 7500
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 9.000586166471278e-05,
+      "loss": 0.0584,
+      "step": 7510
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 8.985932004689332e-05,
+      "loss": 0.1142,
+      "step": 7520
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 8.971277842907386e-05,
+      "loss": 0.0509,
+      "step": 7530
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 8.95662368112544e-05,
+      "loss": 0.0774,
+      "step": 7540
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 8.941969519343494e-05,
+      "loss": 0.0116,
+      "step": 7550
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 8.927315357561548e-05,
+      "loss": 0.03,
+      "step": 7560
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 8.912661195779602e-05,
+      "loss": 0.0951,
+      "step": 7570
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 8.898007033997657e-05,
+      "loss": 0.0158,
+      "step": 7580
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 8.88335287221571e-05,
+      "loss": 0.0638,
+      "step": 7590
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 8.868698710433764e-05,
+      "loss": 0.0401,
+      "step": 7600
+    },
+    {
+      "epoch": 4.45,
+      "eval_accuracy": 0.9357561882958936,
+      "eval_loss": 0.24073971807956696,
+      "eval_runtime": 131.5796,
+      "eval_samples_per_second": 53.116,
+      "eval_steps_per_second": 6.642,
+      "step": 7600
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 8.854044548651817e-05,
+      "loss": 0.0078,
+      "step": 7610
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 8.839390386869871e-05,
+      "loss": 0.0673,
+      "step": 7620
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 8.824736225087925e-05,
+      "loss": 0.0463,
+      "step": 7630
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 8.810082063305978e-05,
+      "loss": 0.0175,
+      "step": 7640
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 8.795427901524033e-05,
+      "loss": 0.0964,
+      "step": 7650
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 8.780773739742087e-05,
+      "loss": 0.103,
+      "step": 7660
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 8.76611957796014e-05,
+      "loss": 0.1191,
+      "step": 7670
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 8.751465416178196e-05,
+      "loss": 0.0351,
+      "step": 7680
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 8.738276670574444e-05,
+      "loss": 0.037,
+      "step": 7690
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 8.723622508792497e-05,
+      "loss": 0.0493,
+      "step": 7700
+    },
+    {
+      "epoch": 4.51,
+      "eval_accuracy": 0.9414794677350121,
+      "eval_loss": 0.22126536071300507,
+      "eval_runtime": 132.171,
+      "eval_samples_per_second": 52.878,
+      "eval_steps_per_second": 6.613,
+      "step": 7700
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 8.708968347010551e-05,
+      "loss": 0.085,
+      "step": 7710
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 8.694314185228606e-05,
+      "loss": 0.0244,
+      "step": 7720
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 8.67966002344666e-05,
+      "loss": 0.0389,
+      "step": 7730
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 8.665005861664713e-05,
+      "loss": 0.0152,
+      "step": 7740
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 8.650351699882767e-05,
+      "loss": 0.0857,
+      "step": 7750
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 8.63569753810082e-05,
+      "loss": 0.0231,
+      "step": 7760
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 8.621043376318874e-05,
+      "loss": 0.0076,
+      "step": 7770
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 8.606389214536929e-05,
+      "loss": 0.0312,
+      "step": 7780
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 8.591735052754983e-05,
+      "loss": 0.1142,
+      "step": 7790
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 8.577080890973038e-05,
+      "loss": 0.0951,
+      "step": 7800
+    },
+    {
+      "epoch": 4.57,
+      "eval_accuracy": 0.9237373014737444,
+      "eval_loss": 0.301612913608551,
+      "eval_runtime": 132.0697,
+      "eval_samples_per_second": 52.919,
+      "eval_steps_per_second": 6.618,
+      "step": 7800
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 8.562426729191091e-05,
+      "loss": 0.0508,
+      "step": 7810
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 8.547772567409145e-05,
+      "loss": 0.0277,
+      "step": 7820
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 8.533118405627199e-05,
+      "loss": 0.0208,
+      "step": 7830
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 8.518464243845252e-05,
+      "loss": 0.0537,
+      "step": 7840
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 8.503810082063306e-05,
+      "loss": 0.0094,
+      "step": 7850
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 8.489155920281361e-05,
+      "loss": 0.0098,
+      "step": 7860
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 8.474501758499415e-05,
+      "loss": 0.0098,
+      "step": 7870
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 8.459847596717468e-05,
+      "loss": 0.0337,
+      "step": 7880
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 8.445193434935522e-05,
+      "loss": 0.0563,
+      "step": 7890
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 8.430539273153576e-05,
+      "loss": 0.0017,
+      "step": 7900
+    },
+    {
+      "epoch": 4.63,
+      "eval_accuracy": 0.9447703534125054,
+      "eval_loss": 0.218331441283226,
+      "eval_runtime": 132.27,
+      "eval_samples_per_second": 52.839,
+      "eval_steps_per_second": 6.608,
+      "step": 7900
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 8.415885111371629e-05,
+      "loss": 0.0154,
+      "step": 7910
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 8.401230949589684e-05,
+      "loss": 0.0661,
+      "step": 7920
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 8.386576787807738e-05,
+      "loss": 0.0569,
+      "step": 7930
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 8.371922626025793e-05,
+      "loss": 0.0323,
+      "step": 7940
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 8.357268464243846e-05,
+      "loss": 0.0155,
+      "step": 7950
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 8.3426143024619e-05,
+      "loss": 0.0679,
+      "step": 7960
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 8.327960140679954e-05,
+      "loss": 0.0159,
+      "step": 7970
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 8.313305978898007e-05,
+      "loss": 0.0561,
+      "step": 7980
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 8.298651817116061e-05,
+      "loss": 0.0348,
+      "step": 7990
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 8.283997655334115e-05,
+      "loss": 0.0561,
+      "step": 8000
+    },
+    {
+      "epoch": 4.69,
+      "eval_accuracy": 0.9492058949778223,
+      "eval_loss": 0.19620640575885773,
+      "eval_runtime": 131.9329,
+      "eval_samples_per_second": 52.974,
+      "eval_steps_per_second": 6.625,
+      "step": 8000
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 8.26934349355217e-05,
+      "loss": 0.0143,
+      "step": 8010
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 8.254689331770223e-05,
+      "loss": 0.0049,
+      "step": 8020
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 8.240035169988277e-05,
+      "loss": 0.0269,
+      "step": 8030
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 8.22538100820633e-05,
+      "loss": 0.0109,
+      "step": 8040
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 8.210726846424384e-05,
+      "loss": 0.0425,
+      "step": 8050
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 8.196072684642438e-05,
+      "loss": 0.0051,
+      "step": 8060
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 8.181418522860493e-05,
+      "loss": 0.0235,
+      "step": 8070
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 8.166764361078546e-05,
+      "loss": 0.0147,
+      "step": 8080
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 8.152110199296601e-05,
+      "loss": 0.0175,
+      "step": 8090
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 8.137456037514655e-05,
+      "loss": 0.0063,
+      "step": 8100
+    },
+    {
+      "epoch": 4.75,
+      "eval_accuracy": 0.9522106166833596,
+      "eval_loss": 0.18679927289485931,
+      "eval_runtime": 131.6399,
+      "eval_samples_per_second": 53.092,
+      "eval_steps_per_second": 6.639,
+      "step": 8100
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 8.122801875732709e-05,
+      "loss": 0.0021,
+      "step": 8110
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 8.108147713950762e-05,
+      "loss": 0.0569,
+      "step": 8120
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 8.093493552168816e-05,
+      "loss": 0.0257,
+      "step": 8130
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 8.07883939038687e-05,
+      "loss": 0.0079,
+      "step": 8140
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 8.064185228604925e-05,
+      "loss": 0.0404,
+      "step": 8150
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 8.049531066822978e-05,
+      "loss": 0.0325,
+      "step": 8160
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 8.034876905041032e-05,
+      "loss": 0.0417,
+      "step": 8170
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 8.020222743259086e-05,
+      "loss": 0.0183,
+      "step": 8180
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 8.005568581477139e-05,
+      "loss": 0.0078,
+      "step": 8190
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 7.990914419695193e-05,
+      "loss": 0.0054,
+      "step": 8200
+    },
+    {
+      "epoch": 4.81,
+      "eval_accuracy": 0.9459150093003291,
+      "eval_loss": 0.20684893429279327,
+      "eval_runtime": 125.0363,
+      "eval_samples_per_second": 55.896,
+      "eval_steps_per_second": 6.99,
+      "step": 8200
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 7.976260257913248e-05,
+      "loss": 0.0604,
+      "step": 8210
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 7.961606096131302e-05,
+      "loss": 0.0156,
+      "step": 8220
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 7.946951934349357e-05,
+      "loss": 0.0332,
+      "step": 8230
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 7.93229777256741e-05,
+      "loss": 0.0081,
+      "step": 8240
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 7.917643610785464e-05,
+      "loss": 0.0581,
+      "step": 8250
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 7.902989449003517e-05,
+      "loss": 0.0264,
+      "step": 8260
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 7.888335287221571e-05,
+      "loss": 0.0351,
+      "step": 8270
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 7.873681125439625e-05,
+      "loss": 0.0222,
+      "step": 8280
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 7.85902696365768e-05,
+      "loss": 0.0526,
+      "step": 8290
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 7.844372801875733e-05,
+      "loss": 0.0519,
+      "step": 8300
+    },
+    {
+      "epoch": 4.87,
+      "eval_accuracy": 0.9429102875947918,
+      "eval_loss": 0.21411900222301483,
+      "eval_runtime": 131.8333,
+      "eval_samples_per_second": 53.014,
+      "eval_steps_per_second": 6.63,
+      "step": 8300
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 7.829718640093787e-05,
+      "loss": 0.0085,
+      "step": 8310
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 7.81506447831184e-05,
+      "loss": 0.0045,
+      "step": 8320
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 7.800410316529894e-05,
+      "loss": 0.0111,
+      "step": 8330
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 7.785756154747948e-05,
+      "loss": 0.0776,
+      "step": 8340
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 7.771101992966003e-05,
+      "loss": 0.024,
+      "step": 8350
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 7.756447831184057e-05,
+      "loss": 0.0448,
+      "step": 8360
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 7.741793669402112e-05,
+      "loss": 0.0455,
+      "step": 8370
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 7.727139507620165e-05,
+      "loss": 0.06,
+      "step": 8380
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 7.712485345838219e-05,
+      "loss": 0.0486,
+      "step": 8390
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 7.697831184056272e-05,
+      "loss": 0.027,
+      "step": 8400
+    },
+    {
+      "epoch": 4.92,
+      "eval_accuracy": 0.9437687795106596,
+      "eval_loss": 0.21381603181362152,
+      "eval_runtime": 132.1229,
+      "eval_samples_per_second": 52.898,
+      "eval_steps_per_second": 6.615,
+      "step": 8400
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 7.683177022274326e-05,
+      "loss": 0.0347,
+      "step": 8410
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 7.66852286049238e-05,
+      "loss": 0.0399,
+      "step": 8420
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 7.653868698710435e-05,
+      "loss": 0.0876,
+      "step": 8430
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 7.639214536928488e-05,
+      "loss": 0.0069,
+      "step": 8440
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 7.624560375146542e-05,
+      "loss": 0.0121,
+      "step": 8450
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 7.609906213364596e-05,
+      "loss": 0.0336,
+      "step": 8460
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 7.595252051582649e-05,
+      "loss": 0.0281,
+      "step": 8470
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 7.580597889800703e-05,
+      "loss": 0.0059,
+      "step": 8480
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 7.565943728018758e-05,
+      "loss": 0.0015,
+      "step": 8490
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 7.551289566236812e-05,
+      "loss": 0.0034,
+      "step": 8500
+    },
+    {
+      "epoch": 4.98,
+      "eval_accuracy": 0.9529260266132494,
+      "eval_loss": 0.1774452030658722,
+      "eval_runtime": 131.79,
+      "eval_samples_per_second": 53.031,
+      "eval_steps_per_second": 6.632,
+      "step": 8500
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 7.536635404454867e-05,
+      "loss": 0.0606,
+      "step": 8510
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 7.52198124267292e-05,
+      "loss": 0.0617,
+      "step": 8520
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 7.507327080890974e-05,
+      "loss": 0.0012,
+      "step": 8530
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 7.492672919109028e-05,
+      "loss": 0.0232,
+      "step": 8540
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 7.478018757327081e-05,
+      "loss": 0.0023,
+      "step": 8550
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 7.463364595545135e-05,
+      "loss": 0.011,
+      "step": 8560
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 7.448710433763188e-05,
+      "loss": 0.0554,
+      "step": 8570
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 7.434056271981243e-05,
+      "loss": 0.002,
+      "step": 8580
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 7.419402110199297e-05,
+      "loss": 0.0007,
+      "step": 8590
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 7.404747948417351e-05,
+      "loss": 0.0096,
+      "step": 8600
+    },
+    {
+      "epoch": 5.04,
+      "eval_accuracy": 0.9512090427815139,
+      "eval_loss": 0.17777030169963837,
+      "eval_runtime": 131.4881,
+      "eval_samples_per_second": 53.153,
+      "eval_steps_per_second": 6.647,
+      "step": 8600
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 7.390093786635404e-05,
+      "loss": 0.0011,
+      "step": 8610
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 7.375439624853458e-05,
+      "loss": 0.0014,
+      "step": 8620
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 7.360785463071512e-05,
+      "loss": 0.0172,
+      "step": 8630
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 7.346131301289567e-05,
+      "loss": 0.0154,
+      "step": 8640
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 7.33147713950762e-05,
+      "loss": 0.0025,
+      "step": 8650
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 7.316822977725675e-05,
+      "loss": 0.0144,
+      "step": 8660
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 7.302168815943729e-05,
+      "loss": 0.0096,
+      "step": 8670
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 7.287514654161783e-05,
+      "loss": 0.0196,
+      "step": 8680
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 7.272860492379836e-05,
+      "loss": 0.0019,
+      "step": 8690
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 7.25820633059789e-05,
+      "loss": 0.0011,
+      "step": 8700
+    },
+    {
+      "epoch": 5.1,
+      "eval_accuracy": 0.9512090427815139,
+      "eval_loss": 0.18543057143688202,
+      "eval_runtime": 125.1978,
+      "eval_samples_per_second": 55.824,
+      "eval_steps_per_second": 6.981,
+      "step": 8700
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 7.243552168815943e-05,
+      "loss": 0.0011,
+      "step": 8710
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 7.228898007033998e-05,
+      "loss": 0.0034,
+      "step": 8720
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 7.214243845252052e-05,
+      "loss": 0.0152,
+      "step": 8730
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 7.199589683470106e-05,
+      "loss": 0.0338,
+      "step": 8740
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 7.18493552168816e-05,
+      "loss": 0.0097,
+      "step": 8750
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 7.170281359906213e-05,
+      "loss": 0.0078,
+      "step": 8760
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 7.155627198124267e-05,
+      "loss": 0.0019,
+      "step": 8770
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 7.140973036342322e-05,
+      "loss": 0.0035,
+      "step": 8780
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 7.126318874560375e-05,
+      "loss": 0.0559,
+      "step": 8790
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 7.11166471277843e-05,
+      "loss": 0.0195,
+      "step": 8800
+    },
+    {
+      "epoch": 5.16,
+      "eval_accuracy": 0.9483474030619545,
+      "eval_loss": 0.19136610627174377,
+      "eval_runtime": 124.2831,
+      "eval_samples_per_second": 56.235,
+      "eval_steps_per_second": 7.032,
+      "step": 8800
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 7.097010550996484e-05,
+      "loss": 0.0041,
+      "step": 8810
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 7.082356389214538e-05,
+      "loss": 0.0427,
+      "step": 8820
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 7.067702227432591e-05,
+      "loss": 0.0028,
+      "step": 8830
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 7.053048065650645e-05,
+      "loss": 0.0746,
+      "step": 8840
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 7.038393903868699e-05,
+      "loss": 0.0023,
+      "step": 8850
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 7.023739742086754e-05,
+      "loss": 0.0486,
+      "step": 8860
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 7.009085580304807e-05,
+      "loss": 0.0009,
+      "step": 8870
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 6.994431418522861e-05,
+      "loss": 0.002,
+      "step": 8880
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 6.979777256740914e-05,
+      "loss": 0.0006,
+      "step": 8890
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 6.965123094958968e-05,
+      "loss": 0.0245,
+      "step": 8900
+    },
+    {
+      "epoch": 5.22,
+      "eval_accuracy": 0.9470596651881528,
+      "eval_loss": 0.21557945013046265,
+      "eval_runtime": 131.081,
+      "eval_samples_per_second": 53.318,
+      "eval_steps_per_second": 6.668,
+      "step": 8900
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 6.950468933177022e-05,
+      "loss": 0.0175,
+      "step": 8910
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 6.935814771395077e-05,
+      "loss": 0.0234,
+      "step": 8920
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 6.92116060961313e-05,
+      "loss": 0.0245,
+      "step": 8930
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 6.906506447831185e-05,
+      "loss": 0.017,
+      "step": 8940
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 6.891852286049239e-05,
+      "loss": 0.0065,
+      "step": 8950
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 6.877198124267293e-05,
+      "loss": 0.0011,
+      "step": 8960
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 6.862543962485346e-05,
+      "loss": 0.0038,
+      "step": 8970
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 6.8478898007034e-05,
+      "loss": 0.0039,
+      "step": 8980
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 6.833235638921454e-05,
+      "loss": 0.0436,
+      "step": 8990
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 6.818581477139509e-05,
+      "loss": 0.0055,
+      "step": 9000
+    },
+    {
+      "epoch": 5.28,
+      "eval_accuracy": 0.9573615681785663,
+      "eval_loss": 0.16400882601737976,
+      "eval_runtime": 131.2197,
+      "eval_samples_per_second": 53.262,
+      "eval_steps_per_second": 6.661,
+      "step": 9000
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 6.803927315357562e-05,
+      "loss": 0.0106,
+      "step": 9010
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 6.789273153575616e-05,
+      "loss": 0.0016,
+      "step": 9020
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 6.77461899179367e-05,
+      "loss": 0.0221,
+      "step": 9030
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 6.759964830011723e-05,
+      "loss": 0.0086,
+      "step": 9040
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 6.745310668229777e-05,
+      "loss": 0.0054,
+      "step": 9050
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 6.73065650644783e-05,
+      "loss": 0.0097,
+      "step": 9060
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 6.716002344665885e-05,
+      "loss": 0.002,
+      "step": 9070
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 6.70134818288394e-05,
+      "loss": 0.0517,
+      "step": 9080
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 6.686694021101994e-05,
+      "loss": 0.0023,
+      "step": 9090
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 6.672039859320048e-05,
+      "loss": 0.0166,
+      "step": 9100
+    },
+    {
+      "epoch": 5.33,
+      "eval_accuracy": 0.9567892402346545,
+      "eval_loss": 0.1769905984401703,
+      "eval_runtime": 131.3469,
+      "eval_samples_per_second": 53.21,
+      "eval_steps_per_second": 6.654,
+      "step": 9100
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 6.657385697538101e-05,
+      "loss": 0.0087,
+      "step": 9110
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 6.642731535756155e-05,
+      "loss": 0.001,
+      "step": 9120
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 6.628077373974209e-05,
+      "loss": 0.0384,
+      "step": 9130
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 6.613423212192264e-05,
+      "loss": 0.039,
+      "step": 9140
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 6.598769050410317e-05,
+      "loss": 0.0101,
+      "step": 9150
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 6.584114888628371e-05,
+      "loss": 0.0297,
+      "step": 9160
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 6.569460726846425e-05,
+      "loss": 0.0052,
+      "step": 9170
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 6.554806565064478e-05,
+      "loss": 0.0547,
+      "step": 9180
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 6.540152403282532e-05,
+      "loss": 0.0216,
+      "step": 9190
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 6.525498241500585e-05,
+      "loss": 0.0217,
+      "step": 9200
+    },
+    {
+      "epoch": 5.39,
+      "eval_accuracy": 0.9479181571040206,
+      "eval_loss": 0.20112772285938263,
+      "eval_runtime": 124.3427,
+      "eval_samples_per_second": 56.208,
+      "eval_steps_per_second": 7.029,
+      "step": 9200
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 6.51084407971864e-05,
+      "loss": 0.0027,
+      "step": 9210
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 6.496189917936694e-05,
+      "loss": 0.0021,
+      "step": 9220
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 6.481535756154749e-05,
+      "loss": 0.0251,
+      "step": 9230
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 6.466881594372803e-05,
+      "loss": 0.0386,
+      "step": 9240
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 6.452227432590856e-05,
+      "loss": 0.0021,
+      "step": 9250
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 6.43757327080891e-05,
+      "loss": 0.0011,
+      "step": 9260
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 6.422919109026964e-05,
+      "loss": 0.0241,
+      "step": 9270
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 6.408264947245017e-05,
+      "loss": 0.0418,
+      "step": 9280
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 6.393610785463072e-05,
+      "loss": 0.0018,
+      "step": 9290
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 6.378956623681126e-05,
+      "loss": 0.0017,
+      "step": 9300
+    },
+    {
+      "epoch": 5.45,
+      "eval_accuracy": 0.946201173272285,
+      "eval_loss": 0.2209555059671402,
+      "eval_runtime": 124.6429,
+      "eval_samples_per_second": 56.072,
+      "eval_steps_per_second": 7.012,
+      "step": 9300
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 6.36430246189918e-05,
+      "loss": 0.0206,
+      "step": 9310
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 6.349648300117233e-05,
+      "loss": 0.0009,
+      "step": 9320
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 6.334994138335287e-05,
+      "loss": 0.0069,
+      "step": 9330
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 6.32033997655334e-05,
+      "loss": 0.0335,
+      "step": 9340
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 6.305685814771395e-05,
+      "loss": 0.0123,
+      "step": 9350
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 6.291031652989449e-05,
+      "loss": 0.0048,
+      "step": 9360
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 6.276377491207504e-05,
+      "loss": 0.0116,
+      "step": 9370
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 6.261723329425558e-05,
+      "loss": 0.001,
+      "step": 9380
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 6.247069167643611e-05,
+      "loss": 0.002,
+      "step": 9390
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 6.232415005861665e-05,
+      "loss": 0.0161,
+      "step": 9400
+    },
+    {
+      "epoch": 5.51,
+      "eval_accuracy": 0.9620832737158391,
+      "eval_loss": 0.15102699398994446,
+      "eval_runtime": 130.7987,
+      "eval_samples_per_second": 53.433,
+      "eval_steps_per_second": 6.682,
+      "step": 9400
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 6.217760844079719e-05,
+      "loss": 0.0012,
+      "step": 9410
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 6.203106682297772e-05,
+      "loss": 0.0024,
+      "step": 9420
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 6.188452520515827e-05,
+      "loss": 0.0037,
+      "step": 9430
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 6.173798358733881e-05,
+      "loss": 0.0047,
+      "step": 9440
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 6.159144196951935e-05,
+      "loss": 0.0417,
+      "step": 9450
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 6.144490035169988e-05,
+      "loss": 0.014,
+      "step": 9460
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 6.129835873388042e-05,
+      "loss": 0.005,
+      "step": 9470
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 6.115181711606096e-05,
+      "loss": 0.0726,
+      "step": 9480
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 6.10052754982415e-05,
+      "loss": 0.0282,
+      "step": 9490
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 6.085873388042205e-05,
+      "loss": 0.0193,
+      "step": 9500
+    },
+    {
+      "epoch": 5.57,
+      "eval_accuracy": 0.958649306052368,
+      "eval_loss": 0.16425824165344238,
+      "eval_runtime": 124.7029,
+      "eval_samples_per_second": 56.045,
+      "eval_steps_per_second": 7.009,
+      "step": 9500
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 6.0712192262602585e-05,
+      "loss": 0.0141,
+      "step": 9510
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 6.056565064478312e-05,
+      "loss": 0.0077,
+      "step": 9520
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 6.041910902696366e-05,
+      "loss": 0.0311,
+      "step": 9530
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 6.02725674091442e-05,
+      "loss": 0.0437,
+      "step": 9540
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 6.012602579132474e-05,
+      "loss": 0.0277,
+      "step": 9550
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 5.9979484173505274e-05,
+      "loss": 0.0027,
+      "step": 9560
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 5.9832942555685824e-05,
+      "loss": 0.0606,
+      "step": 9570
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 5.968640093786636e-05,
+      "loss": 0.0006,
+      "step": 9580
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 5.9539859320046896e-05,
+      "loss": 0.0052,
+      "step": 9590
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 5.939331770222743e-05,
+      "loss": 0.0121,
+      "step": 9600
+    },
+    {
+      "epoch": 5.63,
+      "eval_accuracy": 0.9534983545571613,
+      "eval_loss": 0.17156189680099487,
+      "eval_runtime": 131.2369,
+      "eval_samples_per_second": 53.255,
+      "eval_steps_per_second": 6.66,
+      "step": 9600
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 5.9246776084407976e-05,
+      "loss": 0.0011,
+      "step": 9610
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 5.910023446658851e-05,
+      "loss": 0.0028,
+      "step": 9620
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 5.895369284876905e-05,
+      "loss": 0.0077,
+      "step": 9630
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 5.88071512309496e-05,
+      "loss": 0.0116,
+      "step": 9640
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 5.8660609613130135e-05,
+      "loss": 0.041,
+      "step": 9650
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 5.851406799531067e-05,
+      "loss": 0.027,
+      "step": 9660
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 5.836752637749121e-05,
+      "loss": 0.0005,
+      "step": 9670
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 5.822098475967175e-05,
+      "loss": 0.024,
+      "step": 9680
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 5.807444314185229e-05,
+      "loss": 0.0529,
+      "step": 9690
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 5.7927901524032824e-05,
+      "loss": 0.0146,
+      "step": 9700
+    },
+    {
+      "epoch": 5.69,
+      "eval_accuracy": 0.9553584203748748,
+      "eval_loss": 0.17195869982242584,
+      "eval_runtime": 131.4278,
+      "eval_samples_per_second": 53.177,
+      "eval_steps_per_second": 6.65,
+      "step": 9700
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 5.7781359906213374e-05,
+      "loss": 0.0305,
+      "step": 9710
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 5.763481828839391e-05,
+      "loss": 0.021,
+      "step": 9720
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 5.748827667057445e-05,
+      "loss": 0.0062,
+      "step": 9730
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 5.734173505275498e-05,
+      "loss": 0.0014,
+      "step": 9740
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 5.7195193434935526e-05,
+      "loss": 0.004,
+      "step": 9750
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 5.704865181711606e-05,
+      "loss": 0.062,
+      "step": 9760
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 5.69021101992966e-05,
+      "loss": 0.0171,
+      "step": 9770
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 5.675556858147715e-05,
+      "loss": 0.0009,
+      "step": 9780
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 5.6609026963657686e-05,
+      "loss": 0.0089,
+      "step": 9790
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 5.646248534583822e-05,
+      "loss": 0.0071,
+      "step": 9800
+    },
+    {
+      "epoch": 5.74,
+      "eval_accuracy": 0.9540706825010731,
+      "eval_loss": 0.1830926537513733,
+      "eval_runtime": 130.7438,
+      "eval_samples_per_second": 53.456,
+      "eval_steps_per_second": 6.685,
+      "step": 9800
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 5.631594372801876e-05,
+      "loss": 0.0052,
+      "step": 9810
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 5.61694021101993e-05,
+      "loss": 0.0036,
+      "step": 9820
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 5.602286049237984e-05,
+      "loss": 0.0284,
+      "step": 9830
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 5.5876318874560374e-05,
+      "loss": 0.018,
+      "step": 9840
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 5.572977725674091e-05,
+      "loss": 0.0025,
+      "step": 9850
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 5.558323563892146e-05,
+      "loss": 0.0037,
+      "step": 9860
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 5.5436694021102e-05,
+      "loss": 0.004,
+      "step": 9870
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 5.5290152403282534e-05,
+      "loss": 0.0105,
+      "step": 9880
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 5.514361078546308e-05,
+      "loss": 0.0187,
+      "step": 9890
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 5.499706916764361e-05,
+      "loss": 0.0018,
+      "step": 9900
+    },
+    {
+      "epoch": 5.8,
+      "eval_accuracy": 0.9484904850479324,
+      "eval_loss": 0.20763841271400452,
+      "eval_runtime": 125.1627,
+      "eval_samples_per_second": 55.839,
+      "eval_steps_per_second": 6.983,
+      "step": 9900
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 5.485052754982415e-05,
+      "loss": 0.0016,
+      "step": 9910
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 5.4703985932004686e-05,
+      "loss": 0.0372,
+      "step": 9920
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 5.4557444314185236e-05,
+      "loss": 0.002,
+      "step": 9930
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 5.441090269636577e-05,
+      "loss": 0.0034,
+      "step": 9940
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 5.426436107854631e-05,
+      "loss": 0.0008,
+      "step": 9950
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 5.4117819460726845e-05,
+      "loss": 0.0008,
+      "step": 9960
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 5.397127784290739e-05,
+      "loss": 0.0046,
+      "step": 9970
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 5.3824736225087925e-05,
+      "loss": 0.0035,
+      "step": 9980
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 5.367819460726846e-05,
+      "loss": 0.0019,
+      "step": 9990
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 5.353165298944901e-05,
+      "loss": 0.0007,
+      "step": 10000
+    },
+    {
+      "epoch": 5.86,
+      "eval_accuracy": 0.9599370439261697,
+      "eval_loss": 0.16359391808509827,
+      "eval_runtime": 130.8629,
+      "eval_samples_per_second": 53.407,
+      "eval_steps_per_second": 6.679,
+      "step": 10000
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 5.338511137162955e-05,
+      "loss": 0.0014,
+      "step": 10010
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 5.3238569753810084e-05,
+      "loss": 0.0021,
+      "step": 10020
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 5.309202813599062e-05,
+      "loss": 0.0355,
+      "step": 10030
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 5.2945486518171164e-05,
+      "loss": 0.0013,
+      "step": 10040
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 5.27989449003517e-05,
+      "loss": 0.0061,
+      "step": 10050
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 5.2652403282532236e-05,
+      "loss": 0.0011,
+      "step": 10060
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 5.2505861664712786e-05,
+      "loss": 0.006,
+      "step": 10070
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 5.235932004689332e-05,
+      "loss": 0.0403,
+      "step": 10080
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 5.221277842907386e-05,
+      "loss": 0.0032,
+      "step": 10090
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 5.2066236811254396e-05,
+      "loss": 0.0005,
+      "step": 10100
+    },
+    {
+      "epoch": 5.92,
+      "eval_accuracy": 0.9602232078981257,
+      "eval_loss": 0.16248224675655365,
+      "eval_runtime": 131.0667,
+      "eval_samples_per_second": 53.324,
+      "eval_steps_per_second": 6.668,
+      "step": 10100
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 5.191969519343494e-05,
+      "loss": 0.0019,
+      "step": 10110
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 5.1773153575615475e-05,
+      "loss": 0.0451,
+      "step": 10120
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 5.162661195779601e-05,
+      "loss": 0.0014,
+      "step": 10130
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 5.148007033997656e-05,
+      "loss": 0.0277,
+      "step": 10140
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 5.13335287221571e-05,
+      "loss": 0.0035,
+      "step": 10150
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 5.1186987104337634e-05,
+      "loss": 0.0279,
+      "step": 10160
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 5.104044548651817e-05,
+      "loss": 0.0056,
+      "step": 10170
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 5.0893903868698714e-05,
+      "loss": 0.0394,
+      "step": 10180
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 5.074736225087925e-05,
+      "loss": 0.0004,
+      "step": 10190
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 5.060082063305979e-05,
+      "loss": 0.0277,
+      "step": 10200
+    },
+    {
+      "epoch": 5.98,
+      "eval_accuracy": 0.954643010444985,
+      "eval_loss": 0.1874370574951172,
+      "eval_runtime": 131.5111,
+      "eval_samples_per_second": 53.144,
+      "eval_steps_per_second": 6.646,
+      "step": 10200
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 5.045427901524034e-05,
+      "loss": 0.0299,
+      "step": 10210
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 5.030773739742087e-05,
+      "loss": 0.0515,
+      "step": 10220
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 5.016119577960141e-05,
+      "loss": 0.012,
+      "step": 10230
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 5.0014654161781946e-05,
+      "loss": 0.03,
+      "step": 10240
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.986811254396249e-05,
+      "loss": 0.034,
+      "step": 10250
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.972157092614303e-05,
+      "loss": 0.0016,
+      "step": 10260
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 4.957502930832357e-05,
+      "loss": 0.002,
+      "step": 10270
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 4.9428487690504105e-05,
+      "loss": 0.0127,
+      "step": 10280
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 4.928194607268464e-05,
+      "loss": 0.0005,
+      "step": 10290
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 4.9135404454865185e-05,
+      "loss": 0.0005,
+      "step": 10300
+    },
+    {
+      "epoch": 6.04,
+      "eval_accuracy": 0.9579338961224781,
+      "eval_loss": 0.17904821038246155,
+      "eval_runtime": 124.074,
+      "eval_samples_per_second": 56.329,
+      "eval_steps_per_second": 7.044,
+      "step": 10300
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 4.898886283704572e-05,
+      "loss": 0.0161,
+      "step": 10310
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 4.8842321219226264e-05,
+      "loss": 0.0216,
+      "step": 10320
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 4.869577960140681e-05,
+      "loss": 0.0004,
+      "step": 10330
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 4.8549237983587344e-05,
+      "loss": 0.0167,
+      "step": 10340
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 4.840269636576788e-05,
+      "loss": 0.0005,
+      "step": 10350
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 4.825615474794842e-05,
+      "loss": 0.0016,
+      "step": 10360
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 4.810961313012896e-05,
+      "loss": 0.0022,
+      "step": 10370
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 4.7963071512309496e-05,
+      "loss": 0.0044,
+      "step": 10380
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 4.781652989449004e-05,
+      "loss": 0.0156,
+      "step": 10390
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 4.7669988276670576e-05,
+      "loss": 0.0012,
+      "step": 10400
+    },
+    {
+      "epoch": 6.1,
+      "eval_accuracy": 0.9543568464730291,
+      "eval_loss": 0.1839716136455536,
+      "eval_runtime": 131.6659,
+      "eval_samples_per_second": 53.081,
+      "eval_steps_per_second": 6.638,
+      "step": 10400
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 4.752344665885112e-05,
+      "loss": 0.0019,
+      "step": 10410
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 4.7376905041031656e-05,
+      "loss": 0.0142,
+      "step": 10420
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 4.723036342321219e-05,
+      "loss": 0.0005,
+      "step": 10430
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 4.7083821805392735e-05,
+      "loss": 0.0003,
+      "step": 10440
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 4.693728018757327e-05,
+      "loss": 0.0005,
+      "step": 10450
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 4.679073856975381e-05,
+      "loss": 0.0018,
+      "step": 10460
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 4.664419695193435e-05,
+      "loss": 0.0004,
+      "step": 10470
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 4.6497655334114894e-05,
+      "loss": 0.0018,
+      "step": 10480
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 4.635111371629543e-05,
+      "loss": 0.0024,
+      "step": 10490
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 4.620457209847597e-05,
+      "loss": 0.0431,
+      "step": 10500
+    },
+    {
+      "epoch": 6.15,
+      "eval_accuracy": 0.962798683645729,
+      "eval_loss": 0.15711738169193268,
+      "eval_runtime": 124.0214,
+      "eval_samples_per_second": 56.353,
+      "eval_steps_per_second": 7.047,
+      "step": 10500
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 4.605803048065651e-05,
+      "loss": 0.001,
+      "step": 10510
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 4.591148886283705e-05,
+      "loss": 0.0012,
+      "step": 10520
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 4.576494724501758e-05,
+      "loss": 0.009,
+      "step": 10530
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 4.5618405627198126e-05,
+      "loss": 0.009,
+      "step": 10540
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 4.547186400937867e-05,
+      "loss": 0.0005,
+      "step": 10550
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 4.5325322391559206e-05,
+      "loss": 0.0025,
+      "step": 10560
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 4.517878077373974e-05,
+      "loss": 0.0023,
+      "step": 10570
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 4.5032239155920286e-05,
+      "loss": 0.001,
+      "step": 10580
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 4.488569753810082e-05,
+      "loss": 0.0008,
+      "step": 10590
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 4.473915592028136e-05,
+      "loss": 0.0332,
+      "step": 10600
+    },
+    {
+      "epoch": 6.21,
+      "eval_accuracy": 0.9590785520103019,
+      "eval_loss": 0.1598556488752365,
+      "eval_runtime": 131.1284,
+      "eval_samples_per_second": 53.299,
+      "eval_steps_per_second": 6.665,
+      "step": 10600
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 4.45926143024619e-05,
+      "loss": 0.0019,
+      "step": 10610
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 4.4446072684642445e-05,
+      "loss": 0.0019,
+      "step": 10620
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 4.429953106682298e-05,
+      "loss": 0.0003,
+      "step": 10630
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 4.415298944900352e-05,
+      "loss": 0.0005,
+      "step": 10640
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 4.400644783118406e-05,
+      "loss": 0.0006,
+      "step": 10650
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 4.38599062133646e-05,
+      "loss": 0.0003,
+      "step": 10660
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 4.3713364595545134e-05,
+      "loss": 0.0005,
+      "step": 10670
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 4.356682297772568e-05,
+      "loss": 0.0004,
+      "step": 10680
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 4.342028135990622e-05,
+      "loss": 0.0023,
+      "step": 10690
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 4.3273739742086756e-05,
+      "loss": 0.0014,
+      "step": 10700
+    },
+    {
+      "epoch": 6.27,
+      "eval_accuracy": 0.9632279296036629,
+      "eval_loss": 0.14932304620742798,
+      "eval_runtime": 124.5288,
+      "eval_samples_per_second": 56.124,
+      "eval_steps_per_second": 7.018,
+      "step": 10700
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 4.312719812426729e-05,
+      "loss": 0.0007,
+      "step": 10710
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 4.298065650644783e-05,
+      "loss": 0.0427,
+      "step": 10720
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 4.283411488862837e-05,
+      "loss": 0.0014,
+      "step": 10730
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 4.268757327080891e-05,
+      "loss": 0.0004,
+      "step": 10740
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 4.254103165298945e-05,
+      "loss": 0.0003,
+      "step": 10750
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 4.2394490035169995e-05,
+      "loss": 0.0003,
+      "step": 10760
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 4.224794841735053e-05,
+      "loss": 0.0003,
+      "step": 10770
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 4.210140679953107e-05,
+      "loss": 0.0006,
+      "step": 10780
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 4.1954865181711604e-05,
+      "loss": 0.0008,
+      "step": 10790
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 4.180832356389215e-05,
+      "loss": 0.0014,
+      "step": 10800
+    },
+    {
+      "epoch": 6.33,
+      "eval_accuracy": 0.9660895693232222,
+      "eval_loss": 0.13655245304107666,
+      "eval_runtime": 124.3673,
+      "eval_samples_per_second": 56.196,
+      "eval_steps_per_second": 7.028,
+      "step": 10800
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 4.1661781946072684e-05,
+      "loss": 0.0002,
+      "step": 10810
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 4.151524032825323e-05,
+      "loss": 0.0086,
+      "step": 10820
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 4.136869871043377e-05,
+      "loss": 0.0008,
+      "step": 10830
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 4.122215709261431e-05,
+      "loss": 0.0003,
+      "step": 10840
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 4.107561547479484e-05,
+      "loss": 0.0005,
+      "step": 10850
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 4.092907385697538e-05,
+      "loss": 0.004,
+      "step": 10860
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 4.078253223915592e-05,
+      "loss": 0.0003,
+      "step": 10870
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 4.063599062133646e-05,
+      "loss": 0.0004,
+      "step": 10880
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 4.0489449003517e-05,
+      "loss": 0.0162,
+      "step": 10890
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 4.0342907385697546e-05,
+      "loss": 0.0006,
+      "step": 10900
+    },
+    {
+      "epoch": 6.39,
+      "eval_accuracy": 0.9609386178280155,
+      "eval_loss": 0.15821775794029236,
+      "eval_runtime": 124.3428,
+      "eval_samples_per_second": 56.208,
+      "eval_steps_per_second": 7.029,
+      "step": 10900
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 4.019636576787808e-05,
+      "loss": 0.0081,
+      "step": 10910
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 4.004982415005862e-05,
+      "loss": 0.0006,
+      "step": 10920
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.9903282532239155e-05,
+      "loss": 0.0112,
+      "step": 10930
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.97567409144197e-05,
+      "loss": 0.0006,
+      "step": 10940
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.9610199296600234e-05,
+      "loss": 0.003,
+      "step": 10950
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.946365767878078e-05,
+      "loss": 0.0004,
+      "step": 10960
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.931711606096132e-05,
+      "loss": 0.0004,
+      "step": 10970
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.917057444314186e-05,
+      "loss": 0.0126,
+      "step": 10980
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.9024032825322394e-05,
+      "loss": 0.0156,
+      "step": 10990
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.887749120750293e-05,
+      "loss": 0.0005,
+      "step": 11000
+    },
+    {
+      "epoch": 6.45,
+      "eval_accuracy": 0.9589354700243239,
+      "eval_loss": 0.17037606239318848,
+      "eval_runtime": 124.3235,
+      "eval_samples_per_second": 56.216,
+      "eval_steps_per_second": 7.03,
+      "step": 11000
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.873094958968347e-05,
+      "loss": 0.0109,
+      "step": 11010
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.858440797186401e-05,
+      "loss": 0.0004,
+      "step": 11020
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.8437866354044546e-05,
+      "loss": 0.0025,
+      "step": 11030
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.829132473622509e-05,
+      "loss": 0.0095,
+      "step": 11040
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.814478311840563e-05,
+      "loss": 0.0018,
+      "step": 11050
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.799824150058617e-05,
+      "loss": 0.0005,
+      "step": 11060
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.7851699882766705e-05,
+      "loss": 0.0004,
+      "step": 11070
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.770515826494725e-05,
+      "loss": 0.0026,
+      "step": 11080
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.7558616647127785e-05,
+      "loss": 0.0004,
+      "step": 11090
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.741207502930832e-05,
+      "loss": 0.0004,
+      "step": 11100
+    },
+    {
+      "epoch": 6.51,
+      "eval_accuracy": 0.967091143225068,
+      "eval_loss": 0.1376296877861023,
+      "eval_runtime": 124.4558,
+      "eval_samples_per_second": 56.156,
+      "eval_steps_per_second": 7.023,
+      "step": 11100
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.7265533411488864e-05,
+      "loss": 0.0029,
+      "step": 11110
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.711899179366941e-05,
+      "loss": 0.0003,
+      "step": 11120
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.6972450175849944e-05,
+      "loss": 0.0004,
+      "step": 11130
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.682590855803048e-05,
+      "loss": 0.0011,
+      "step": 11140
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.6679366940211024e-05,
+      "loss": 0.0066,
+      "step": 11150
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.653282532239156e-05,
+      "loss": 0.0007,
+      "step": 11160
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.6386283704572096e-05,
+      "loss": 0.0004,
+      "step": 11170
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.623974208675264e-05,
+      "loss": 0.0008,
+      "step": 11180
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.609320046893318e-05,
+      "loss": 0.0081,
+      "step": 11190
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.594665885111372e-05,
+      "loss": 0.0755,
+      "step": 11200
+    },
+    {
+      "epoch": 6.57,
+      "eval_accuracy": 0.9653741593933324,
+      "eval_loss": 0.13746266067028046,
+      "eval_runtime": 124.8281,
+      "eval_samples_per_second": 55.989,
+      "eval_steps_per_second": 7.002,
+      "step": 11200
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.5800117233294256e-05,
+      "loss": 0.0026,
+      "step": 11210
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.56535756154748e-05,
+      "loss": 0.0048,
+      "step": 11220
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.5507033997655335e-05,
+      "loss": 0.0016,
+      "step": 11230
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.536049237983587e-05,
+      "loss": 0.0086,
+      "step": 11240
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.5213950762016415e-05,
+      "loss": 0.0003,
+      "step": 11250
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.506740914419696e-05,
+      "loss": 0.0003,
+      "step": 11260
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.4920867526377494e-05,
+      "loss": 0.0004,
+      "step": 11270
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.477432590855803e-05,
+      "loss": 0.0073,
+      "step": 11280
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.4627784290738574e-05,
+      "loss": 0.0012,
+      "step": 11290
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.448124267291911e-05,
+      "loss": 0.0002,
+      "step": 11300
+    },
+    {
+      "epoch": 6.62,
+      "eval_accuracy": 0.9660895693232222,
+      "eval_loss": 0.13609342277050018,
+      "eval_runtime": 124.5272,
+      "eval_samples_per_second": 56.124,
+      "eval_steps_per_second": 7.019,
+      "step": 11300
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.433470105509965e-05,
+      "loss": 0.0062,
+      "step": 11310
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.418815943728019e-05,
+      "loss": 0.0002,
+      "step": 11320
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.404161781946073e-05,
+      "loss": 0.0002,
+      "step": 11330
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.389507620164127e-05,
+      "loss": 0.0034,
+      "step": 11340
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3748534583821806e-05,
+      "loss": 0.0011,
+      "step": 11350
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.360199296600234e-05,
+      "loss": 0.0111,
+      "step": 11360
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3455451348182886e-05,
+      "loss": 0.0009,
+      "step": 11370
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.330890973036342e-05,
+      "loss": 0.0004,
+      "step": 11380
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3162368112543965e-05,
+      "loss": 0.0006,
+      "step": 11390
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.301582649472451e-05,
+      "loss": 0.0006,
+      "step": 11400
+    },
+    {
+      "epoch": 6.68,
+      "eval_accuracy": 0.9675203891830019,
+      "eval_loss": 0.13233859837055206,
+      "eval_runtime": 131.4319,
+      "eval_samples_per_second": 53.176,
+      "eval_steps_per_second": 6.65,
+      "step": 11400
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.2869284876905045e-05,
+      "loss": 0.0004,
+      "step": 11410
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.272274325908558e-05,
+      "loss": 0.0003,
+      "step": 11420
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.257620164126612e-05,
+      "loss": 0.0003,
+      "step": 11430
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.242966002344666e-05,
+      "loss": 0.0009,
+      "step": 11440
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.22831184056272e-05,
+      "loss": 0.0002,
+      "step": 11450
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.213657678780774e-05,
+      "loss": 0.0179,
+      "step": 11460
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.1990035169988284e-05,
+      "loss": 0.0016,
+      "step": 11470
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.184349355216882e-05,
+      "loss": 0.005,
+      "step": 11480
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.1696951934349356e-05,
+      "loss": 0.0003,
+      "step": 11490
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.155041031652989e-05,
+      "loss": 0.0009,
+      "step": 11500
+    },
+    {
+      "epoch": 6.74,
+      "eval_accuracy": 0.9692373730147374,
+      "eval_loss": 0.1238512396812439,
+      "eval_runtime": 130.5049,
+      "eval_samples_per_second": 53.554,
+      "eval_steps_per_second": 6.697,
+      "step": 11500
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.1403868698710436e-05,
+      "loss": 0.001,
+      "step": 11510
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.125732708089097e-05,
+      "loss": 0.0003,
+      "step": 11520
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.1110785463071516e-05,
+      "loss": 0.0002,
+      "step": 11530
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.096424384525206e-05,
+      "loss": 0.0003,
+      "step": 11540
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.0817702227432595e-05,
+      "loss": 0.0003,
+      "step": 11550
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.067116060961313e-05,
+      "loss": 0.0003,
+      "step": 11560
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.052461899179367e-05,
+      "loss": 0.0043,
+      "step": 11570
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.037807737397421e-05,
+      "loss": 0.0011,
+      "step": 11580
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.023153575615475e-05,
+      "loss": 0.0006,
+      "step": 11590
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.0084994138335287e-05,
+      "loss": 0.0004,
+      "step": 11600
+    },
+    {
+      "epoch": 6.8,
+      "eval_accuracy": 0.963084847617685,
+      "eval_loss": 0.15142901241779327,
+      "eval_runtime": 122.8963,
+      "eval_samples_per_second": 56.869,
+      "eval_steps_per_second": 7.112,
+      "step": 11600
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 2.993845252051583e-05,
+      "loss": 0.012,
+      "step": 11610
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 2.9791910902696367e-05,
+      "loss": 0.0006,
+      "step": 11620
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 2.9645369284876907e-05,
+      "loss": 0.0015,
+      "step": 11630
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 2.9498827667057443e-05,
+      "loss": 0.039,
+      "step": 11640
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 2.9352286049237986e-05,
+      "loss": 0.0045,
+      "step": 11650
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 2.9205744431418526e-05,
+      "loss": 0.0043,
+      "step": 11660
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 2.9059202813599063e-05,
+      "loss": 0.0008,
+      "step": 11670
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 2.89126611957796e-05,
+      "loss": 0.0002,
+      "step": 11680
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 2.8766119577960142e-05,
+      "loss": 0.0023,
+      "step": 11690
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 2.8619577960140682e-05,
+      "loss": 0.0002,
+      "step": 11700
+    },
+    {
+      "epoch": 6.86,
+      "eval_accuracy": 0.9663757332951781,
+      "eval_loss": 0.1385500580072403,
+      "eval_runtime": 130.5226,
+      "eval_samples_per_second": 53.546,
+      "eval_steps_per_second": 6.696,
+      "step": 11700
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 2.847303634232122e-05,
+      "loss": 0.0003,
+      "step": 11710
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 2.832649472450176e-05,
+      "loss": 0.0006,
+      "step": 11720
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 2.8179953106682298e-05,
+      "loss": 0.0004,
+      "step": 11730
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 2.8033411488862838e-05,
+      "loss": 0.0003,
+      "step": 11740
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 2.7886869871043374e-05,
+      "loss": 0.0017,
+      "step": 11750
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 2.7740328253223917e-05,
+      "loss": 0.0002,
+      "step": 11760
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 2.7593786635404457e-05,
+      "loss": 0.0034,
+      "step": 11770
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 2.7447245017584994e-05,
+      "loss": 0.0002,
+      "step": 11780
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 2.7300703399765537e-05,
+      "loss": 0.0006,
+      "step": 11790
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 2.7154161781946073e-05,
+      "loss": 0.0004,
+      "step": 11800
+    },
+    {
+      "epoch": 6.92,
+      "eval_accuracy": 0.9659464873372442,
+      "eval_loss": 0.13676127791404724,
+      "eval_runtime": 131.3154,
+      "eval_samples_per_second": 53.223,
+      "eval_steps_per_second": 6.656,
+      "step": 11800
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 2.7007620164126613e-05,
+      "loss": 0.0127,
+      "step": 11810
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 2.686107854630715e-05,
+      "loss": 0.0004,
+      "step": 11820
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 2.6714536928487693e-05,
+      "loss": 0.0002,
+      "step": 11830
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 2.6567995310668232e-05,
+      "loss": 0.0004,
+      "step": 11840
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 2.642145369284877e-05,
+      "loss": 0.0263,
+      "step": 11850
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 2.6274912075029312e-05,
+      "loss": 0.0384,
+      "step": 11860
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 2.612837045720985e-05,
+      "loss": 0.0003,
+      "step": 11870
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 2.5981828839390388e-05,
+      "loss": 0.0005,
+      "step": 11880
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 2.5835287221570925e-05,
+      "loss": 0.0003,
+      "step": 11890
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 2.5688745603751468e-05,
+      "loss": 0.0004,
+      "step": 11900
+    },
+    {
+      "epoch": 6.98,
+      "eval_accuracy": 0.9683788810988696,
+      "eval_loss": 0.12763340771198273,
+      "eval_runtime": 123.1553,
+      "eval_samples_per_second": 56.749,
+      "eval_steps_per_second": 7.097,
+      "step": 11900
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 2.5542203985932008e-05,
+      "loss": 0.0007,
+      "step": 11910
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 2.5395662368112544e-05,
+      "loss": 0.0003,
+      "step": 11920
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 2.5249120750293087e-05,
+      "loss": 0.0005,
+      "step": 11930
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 2.5102579132473624e-05,
+      "loss": 0.0002,
+      "step": 11940
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 2.4956037514654163e-05,
+      "loss": 0.0004,
+      "step": 11950
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.4809495896834703e-05,
+      "loss": 0.0002,
+      "step": 11960
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.466295427901524e-05,
+      "loss": 0.0003,
+      "step": 11970
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.451641266119578e-05,
+      "loss": 0.0004,
+      "step": 11980
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.436987104337632e-05,
+      "loss": 0.0002,
+      "step": 11990
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.422332942555686e-05,
+      "loss": 0.0002,
+      "step": 12000
+    },
+    {
+      "epoch": 7.03,
+      "eval_accuracy": 0.9712405208184289,
+      "eval_loss": 0.117136649787426,
+      "eval_runtime": 130.5938,
+      "eval_samples_per_second": 53.517,
+      "eval_steps_per_second": 6.693,
+      "step": 12000
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.40767878077374e-05,
+      "loss": 0.0015,
+      "step": 12010
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.393024618991794e-05,
+      "loss": 0.0003,
+      "step": 12020
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.378370457209848e-05,
+      "loss": 0.0002,
+      "step": 12030
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.3637162954279015e-05,
+      "loss": 0.0002,
+      "step": 12040
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.3490621336459555e-05,
+      "loss": 0.0004,
+      "step": 12050
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.3344079718640094e-05,
+      "loss": 0.0002,
+      "step": 12060
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.3197538100820634e-05,
+      "loss": 0.0002,
+      "step": 12070
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.305099648300117e-05,
+      "loss": 0.0002,
+      "step": 12080
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.2904454865181714e-05,
+      "loss": 0.0003,
+      "step": 12090
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.2757913247362254e-05,
+      "loss": 0.0002,
+      "step": 12100
+    },
+    {
+      "epoch": 7.09,
+      "eval_accuracy": 0.971097438832451,
+      "eval_loss": 0.11417105048894882,
+      "eval_runtime": 124.1945,
+      "eval_samples_per_second": 56.275,
+      "eval_steps_per_second": 7.037,
+      "step": 12100
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.261137162954279e-05,
+      "loss": 0.0001,
+      "step": 12110
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.246483001172333e-05,
+      "loss": 0.0002,
+      "step": 12120
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.231828839390387e-05,
+      "loss": 0.0002,
+      "step": 12130
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.217174677608441e-05,
+      "loss": 0.0002,
+      "step": 12140
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.2025205158264946e-05,
+      "loss": 0.0023,
+      "step": 12150
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.187866354044549e-05,
+      "loss": 0.0002,
+      "step": 12160
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.173212192262603e-05,
+      "loss": 0.0002,
+      "step": 12170
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.1585580304806565e-05,
+      "loss": 0.0002,
+      "step": 12180
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.1439038686987105e-05,
+      "loss": 0.0001,
+      "step": 12190
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.1292497069167645e-05,
+      "loss": 0.0001,
+      "step": 12200
+    },
+    {
+      "epoch": 7.15,
+      "eval_accuracy": 0.9726713406782086,
+      "eval_loss": 0.11834342777729034,
+      "eval_runtime": 124.7545,
+      "eval_samples_per_second": 56.022,
+      "eval_steps_per_second": 7.006,
+      "step": 12200
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.1145955451348185e-05,
+      "loss": 0.0003,
+      "step": 12210
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.099941383352872e-05,
+      "loss": 0.0002,
+      "step": 12220
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.0852872215709264e-05,
+      "loss": 0.0002,
+      "step": 12230
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.07063305978898e-05,
+      "loss": 0.0002,
+      "step": 12240
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.055978898007034e-05,
+      "loss": 0.0002,
+      "step": 12250
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.041324736225088e-05,
+      "loss": 0.0002,
+      "step": 12260
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.026670574443142e-05,
+      "loss": 0.0002,
+      "step": 12270
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.012016412661196e-05,
+      "loss": 0.0002,
+      "step": 12280
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 1.9973622508792496e-05,
+      "loss": 0.0002,
+      "step": 12290
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 1.9827080890973036e-05,
+      "loss": 0.0002,
+      "step": 12300
+    },
+    {
+      "epoch": 7.21,
+      "eval_accuracy": 0.9732436686221205,
+      "eval_loss": 0.1166636124253273,
+      "eval_runtime": 124.5811,
+      "eval_samples_per_second": 56.1,
+      "eval_steps_per_second": 7.016,
+      "step": 12300
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 1.9680539273153576e-05,
+      "loss": 0.0002,
+      "step": 12310
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 1.9533997655334116e-05,
+      "loss": 0.0002,
+      "step": 12320
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 1.9387456037514655e-05,
+      "loss": 0.0002,
+      "step": 12330
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 1.9240914419695195e-05,
+      "loss": 0.0002,
+      "step": 12340
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 1.9094372801875735e-05,
+      "loss": 0.0002,
+      "step": 12350
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 1.894783118405627e-05,
+      "loss": 0.0002,
+      "step": 12360
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 1.880128956623681e-05,
+      "loss": 0.0002,
+      "step": 12370
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 1.865474794841735e-05,
+      "loss": 0.0002,
+      "step": 12380
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 1.850820633059789e-05,
+      "loss": 0.0002,
+      "step": 12390
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 1.8361664712778427e-05,
+      "loss": 0.0002,
+      "step": 12400
+    },
+    {
+      "epoch": 7.27,
+      "eval_accuracy": 0.9736729145800543,
+      "eval_loss": 0.11427531391382217,
+      "eval_runtime": 123.7156,
+      "eval_samples_per_second": 56.492,
+      "eval_steps_per_second": 7.065,
+      "step": 12400
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 1.821512309495897e-05,
+      "loss": 0.0003,
+      "step": 12410
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 1.806858147713951e-05,
+      "loss": 0.0002,
+      "step": 12420
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 1.7922039859320047e-05,
+      "loss": 0.0002,
+      "step": 12430
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 1.7775498241500586e-05,
+      "loss": 0.0002,
+      "step": 12440
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 1.7628956623681126e-05,
+      "loss": 0.0003,
+      "step": 12450
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 1.7482415005861666e-05,
+      "loss": 0.0002,
+      "step": 12460
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 1.7335873388042202e-05,
+      "loss": 0.0002,
+      "step": 12470
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 1.7189331770222746e-05,
+      "loss": 0.0001,
+      "step": 12480
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 1.7042790152403285e-05,
+      "loss": 0.0002,
+      "step": 12490
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 1.6896248534583822e-05,
+      "loss": 0.0001,
+      "step": 12500
+    },
+    {
+      "epoch": 7.33,
+      "eval_accuracy": 0.9736729145800543,
+      "eval_loss": 0.11291757225990295,
+      "eval_runtime": 131.3202,
+      "eval_samples_per_second": 53.221,
+      "eval_steps_per_second": 6.655,
+      "step": 12500
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 1.674970691676436e-05,
+      "loss": 0.0002,
+      "step": 12510
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 1.66031652989449e-05,
+      "loss": 0.0002,
+      "step": 12520
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 1.645662368112544e-05,
+      "loss": 0.0002,
+      "step": 12530
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 1.6310082063305978e-05,
+      "loss": 0.0002,
+      "step": 12540
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 1.6163540445486517e-05,
+      "loss": 0.0002,
+      "step": 12550
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 1.6016998827667057e-05,
+      "loss": 0.0002,
+      "step": 12560
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 1.5870457209847597e-05,
+      "loss": 0.0002,
+      "step": 12570
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 1.5723915592028137e-05,
+      "loss": 0.0002,
+      "step": 12580
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 1.5577373974208677e-05,
+      "loss": 0.0002,
+      "step": 12590
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 1.5430832356389216e-05,
+      "loss": 0.0002,
+      "step": 12600
+    },
+    {
+      "epoch": 7.39,
+      "eval_accuracy": 0.9742452425239663,
+      "eval_loss": 0.11156066507101059,
+      "eval_runtime": 131.1872,
+      "eval_samples_per_second": 53.275,
+      "eval_steps_per_second": 6.662,
+      "step": 12600
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 1.5284290738569753e-05,
+      "loss": 0.0002,
+      "step": 12610
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 1.5137749120750294e-05,
+      "loss": 0.0016,
+      "step": 12620
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 1.4991207502930832e-05,
+      "loss": 0.0002,
+      "step": 12630
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 1.4844665885111372e-05,
+      "loss": 0.0003,
+      "step": 12640
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 1.4698124267291912e-05,
+      "loss": 0.0002,
+      "step": 12650
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 1.455158264947245e-05,
+      "loss": 0.0001,
+      "step": 12660
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 1.4405041031652992e-05,
+      "loss": 0.0001,
+      "step": 12670
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 1.4258499413833528e-05,
+      "loss": 0.0002,
+      "step": 12680
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 1.411195779601407e-05,
+      "loss": 0.0002,
+      "step": 12690
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 1.3965416178194608e-05,
+      "loss": 0.0002,
+      "step": 12700
+    },
+    {
+      "epoch": 7.44,
+      "eval_accuracy": 0.9745314064959222,
+      "eval_loss": 0.11262710392475128,
+      "eval_runtime": 131.0803,
+      "eval_samples_per_second": 53.318,
+      "eval_steps_per_second": 6.668,
+      "step": 12700
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 1.3818874560375147e-05,
+      "loss": 0.0002,
+      "step": 12710
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 1.3672332942555686e-05,
+      "loss": 0.0002,
+      "step": 12720
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 1.3525791324736225e-05,
+      "loss": 0.0002,
+      "step": 12730
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 1.3379249706916767e-05,
+      "loss": 0.0002,
+      "step": 12740
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 1.3232708089097303e-05,
+      "loss": 0.0002,
+      "step": 12750
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 1.3086166471277845e-05,
+      "loss": 0.0002,
+      "step": 12760
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 1.2939624853458381e-05,
+      "loss": 0.0002,
+      "step": 12770
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 1.2793083235638923e-05,
+      "loss": 0.0002,
+      "step": 12780
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 1.264654161781946e-05,
+      "loss": 0.0002,
+      "step": 12790
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 1.25e-05,
+      "loss": 0.0002,
+      "step": 12800
+    },
+    {
+      "epoch": 7.5,
+      "eval_accuracy": 0.9748175704678781,
+      "eval_loss": 0.11113562434911728,
+      "eval_runtime": 131.633,
+      "eval_samples_per_second": 53.095,
+      "eval_steps_per_second": 6.64,
+      "step": 12800
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 1.235345838218054e-05,
+      "loss": 0.0002,
+      "step": 12810
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 1.2206916764361079e-05,
+      "loss": 0.0001,
+      "step": 12820
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 1.2060375146541618e-05,
+      "loss": 0.0002,
+      "step": 12830
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 1.1913833528722158e-05,
+      "loss": 0.0002,
+      "step": 12840
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 1.1767291910902698e-05,
+      "loss": 0.0002,
+      "step": 12850
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 1.1620750293083236e-05,
+      "loss": 0.0001,
+      "step": 12860
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 1.1474208675263776e-05,
+      "loss": 0.0001,
+      "step": 12870
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 1.1327667057444314e-05,
+      "loss": 0.0002,
+      "step": 12880
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 1.1181125439624854e-05,
+      "loss": 0.0001,
+      "step": 12890
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 1.1034583821805394e-05,
+      "loss": 0.0002,
+      "step": 12900
+    },
+    {
+      "epoch": 7.56,
+      "eval_accuracy": 0.9746744884819001,
+      "eval_loss": 0.11020047217607498,
+      "eval_runtime": 130.4666,
+      "eval_samples_per_second": 53.569,
+      "eval_steps_per_second": 6.699,
+      "step": 12900
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 1.0888042203985932e-05,
+      "loss": 0.0002,
+      "step": 12910
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 1.0741500586166473e-05,
+      "loss": 0.0002,
+      "step": 12920
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 1.0594958968347011e-05,
+      "loss": 0.0002,
+      "step": 12930
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 1.0448417350527551e-05,
+      "loss": 0.0001,
+      "step": 12940
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 1.0301875732708089e-05,
+      "loss": 0.0001,
+      "step": 12950
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 1.0155334114888629e-05,
+      "loss": 0.0002,
+      "step": 12960
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 1.0008792497069169e-05,
+      "loss": 0.0001,
+      "step": 12970
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 9.862250879249707e-06,
+      "loss": 0.0002,
+      "step": 12980
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 9.715709261430247e-06,
+      "loss": 0.0002,
+      "step": 12990
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 9.569167643610786e-06,
+      "loss": 0.0001,
+      "step": 13000
+    },
+    {
+      "epoch": 7.62,
+      "eval_accuracy": 0.9746744884819001,
+      "eval_loss": 0.10937321931123734,
+      "eval_runtime": 130.3281,
+      "eval_samples_per_second": 53.626,
+      "eval_steps_per_second": 6.706,
+      "step": 13000
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 9.422626025791326e-06,
+      "loss": 0.0001,
+      "step": 13010
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 9.276084407971864e-06,
+      "loss": 0.0002,
+      "step": 13020
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 9.129542790152404e-06,
+      "loss": 0.0001,
+      "step": 13030
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 8.983001172332942e-06,
+      "loss": 0.0002,
+      "step": 13040
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 8.836459554513482e-06,
+      "loss": 0.0002,
+      "step": 13050
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 8.689917936694022e-06,
+      "loss": 0.0002,
+      "step": 13060
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 8.54337631887456e-06,
+      "loss": 0.0001,
+      "step": 13070
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 8.396834701055101e-06,
+      "loss": 0.0001,
+      "step": 13080
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 8.25029308323564e-06,
+      "loss": 0.0053,
+      "step": 13090
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 8.10375146541618e-06,
+      "loss": 0.0001,
+      "step": 13100
+    },
+    {
+      "epoch": 7.68,
+      "eval_accuracy": 0.9742452425239663,
+      "eval_loss": 0.10856575518846512,
+      "eval_runtime": 124.665,
+      "eval_samples_per_second": 56.062,
+      "eval_steps_per_second": 7.011,
+      "step": 13100
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 7.957209847596717e-06,
+      "loss": 0.0002,
+      "step": 13110
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 7.810668229777257e-06,
+      "loss": 0.0001,
+      "step": 13120
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 7.664126611957797e-06,
+      "loss": 0.0002,
+      "step": 13130
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 7.517584994138335e-06,
+      "loss": 0.0001,
+      "step": 13140
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 7.371043376318874e-06,
+      "loss": 0.0002,
+      "step": 13150
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 7.224501758499415e-06,
+      "loss": 0.0002,
+      "step": 13160
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 7.077960140679954e-06,
+      "loss": 0.0002,
+      "step": 13170
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.9314185228604935e-06,
+      "loss": 0.0002,
+      "step": 13180
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 6.784876905041032e-06,
+      "loss": 0.0003,
+      "step": 13190
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.638335287221571e-06,
+      "loss": 0.0001,
+      "step": 13200
+    },
+    {
+      "epoch": 7.74,
+      "eval_accuracy": 0.9748175704678781,
+      "eval_loss": 0.10791758447885513,
+      "eval_runtime": 132.0225,
+      "eval_samples_per_second": 52.938,
+      "eval_steps_per_second": 6.62,
+      "step": 13200
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 6.49179366940211e-06,
+      "loss": 0.0002,
+      "step": 13210
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.345252051582649e-06,
+      "loss": 0.0002,
+      "step": 13220
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 6.198710433763189e-06,
+      "loss": 0.0002,
+      "step": 13230
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 6.052168815943728e-06,
+      "loss": 0.0002,
+      "step": 13240
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 5.905627198124267e-06,
+      "loss": 0.0002,
+      "step": 13250
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 5.759085580304807e-06,
+      "loss": 0.0002,
+      "step": 13260
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 5.6125439624853466e-06,
+      "loss": 0.0001,
+      "step": 13270
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 5.4660023446658855e-06,
+      "loss": 0.0001,
+      "step": 13280
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 5.3194607268464245e-06,
+      "loss": 0.0002,
+      "step": 13290
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 5.172919109026964e-06,
+      "loss": 0.0002,
+      "step": 13300
+    },
+    {
+      "epoch": 7.8,
+      "eval_accuracy": 0.9753898984117899,
+      "eval_loss": 0.10622181743383408,
+      "eval_runtime": 131.5367,
+      "eval_samples_per_second": 53.133,
+      "eval_steps_per_second": 6.645,
+      "step": 13300
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 5.026377491207503e-06,
+      "loss": 0.0002,
+      "step": 13310
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 4.879835873388042e-06,
+      "loss": 0.0002,
+      "step": 13320
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 4.733294255568581e-06,
+      "loss": 0.0001,
+      "step": 13330
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 4.586752637749121e-06,
+      "loss": 0.0009,
+      "step": 13340
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 4.440211019929661e-06,
+      "loss": 0.0001,
+      "step": 13350
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 4.2936694021102e-06,
+      "loss": 0.0002,
+      "step": 13360
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 4.147127784290739e-06,
+      "loss": 0.0001,
+      "step": 13370
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 4.000586166471278e-06,
+      "loss": 0.0002,
+      "step": 13380
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.854044548651817e-06,
+      "loss": 0.0002,
+      "step": 13390
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.7075029308323567e-06,
+      "loss": 0.0002,
+      "step": 13400
+    },
+    {
+      "epoch": 7.85,
+      "eval_accuracy": 0.9756760623837459,
+      "eval_loss": 0.1067742183804512,
+      "eval_runtime": 131.3503,
+      "eval_samples_per_second": 53.209,
+      "eval_steps_per_second": 6.654,
+      "step": 13400
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.5609613130128957e-06,
+      "loss": 0.0002,
+      "step": 13410
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.4144196951934355e-06,
+      "loss": 0.0001,
+      "step": 13420
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.2678780773739744e-06,
+      "loss": 0.0002,
+      "step": 13430
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.1213364595545134e-06,
+      "loss": 0.0002,
+      "step": 13440
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.974794841735053e-06,
+      "loss": 0.0002,
+      "step": 13450
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.828253223915592e-06,
+      "loss": 0.0007,
+      "step": 13460
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.6817116060961315e-06,
+      "loss": 0.0008,
+      "step": 13470
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.5351699882766705e-06,
+      "loss": 0.0001,
+      "step": 13480
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.3886283704572103e-06,
+      "loss": 0.0001,
+      "step": 13490
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.2420867526377492e-06,
+      "loss": 0.0001,
+      "step": 13500
+    },
+    {
+      "epoch": 7.91,
+      "eval_accuracy": 0.9762483903276578,
+      "eval_loss": 0.10610503703355789,
+      "eval_runtime": 132.8087,
+      "eval_samples_per_second": 52.625,
+      "eval_steps_per_second": 6.581,
+      "step": 13500
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0955451348182886e-06,
+      "loss": 0.0004,
+      "step": 13510
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 1.9490035169988275e-06,
+      "loss": 0.0002,
+      "step": 13520
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 1.8024618991793671e-06,
+      "loss": 0.0002,
+      "step": 13530
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 1.6559202813599063e-06,
+      "loss": 0.0002,
+      "step": 13540
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 1.5093786635404455e-06,
+      "loss": 0.0001,
+      "step": 13550
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 1.3628370457209848e-06,
+      "loss": 0.0002,
+      "step": 13560
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 1.216295427901524e-06,
+      "loss": 0.0002,
+      "step": 13570
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 1.0697538100820634e-06,
+      "loss": 0.0001,
+      "step": 13580
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 9.232121922626025e-07,
+      "loss": 0.0002,
+      "step": 13590
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 7.766705744431419e-07,
+      "loss": 0.0001,
+      "step": 13600
+    },
+    {
+      "epoch": 7.97,
+      "eval_accuracy": 0.9761053083416797,
+      "eval_loss": 0.10603297501802444,
+      "eval_runtime": 131.2027,
+      "eval_samples_per_second": 53.269,
+      "eval_steps_per_second": 6.661,
+      "step": 13600
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 6.301289566236812e-07,
+      "loss": 0.0001,
+      "step": 13610
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 4.835873388042204e-07,
+      "loss": 0.0003,
+      "step": 13620
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.3704572098475966e-07,
+      "loss": 0.0002,
+      "step": 13630
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 1.9050410316529896e-07,
+      "loss": 0.0001,
+      "step": 13640
+    },
+    {
+      "epoch": 8.0,
+      "step": 13648,
+      "total_flos": 1.6940961703140508e+19,
+      "train_loss": 0.3000723277693496,
+      "train_runtime": 26821.1431,
+      "train_samples_per_second": 8.14,
+      "train_steps_per_second": 0.509
+    }
+  ],
+  "max_steps": 13648,
+  "num_train_epochs": 8,
+  "total_flos": 1.6940961703140508e+19,
+  "trial_name": null,
+  "trial_params": null
+}