diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,7474 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.0,
+  "eval_steps": 500,
+  "global_step": 12055,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99998302127482e-06,
+      "loss": 0.6695,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99993208521459e-06,
+      "loss": 0.6754,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999847192165244e-06,
+      "loss": 0.6642,
+      "step": 30
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.999728342703329e-06,
+      "loss": 0.6416,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.999575537636011e-06,
+      "loss": 0.6391,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.999388778001065e-06,
+      "loss": 0.6745,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.999168065066867e-06,
+      "loss": 0.6693,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.998913400332386e-06,
+      "loss": 0.6759,
+      "step": 80
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.998624785527176e-06,
+      "loss": 0.6505,
+      "step": 90
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.99830222261136e-06,
+      "loss": 0.658,
+      "step": 100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.997945713775624e-06,
+      "loss": 0.6856,
+      "step": 110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.99755526144119e-06,
+      "loss": 0.663,
+      "step": 120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.997130868259815e-06,
+      "loss": 0.6695,
+      "step": 130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.996672537113758e-06,
+      "loss": 0.6249,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.996180271115773e-06,
+      "loss": 0.6652,
+      "step": 150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.995654073609078e-06,
+      "loss": 0.6687,
+      "step": 160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.995093948167339e-06,
+      "loss": 0.6757,
+      "step": 170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.994499898594641e-06,
+      "loss": 0.646,
+      "step": 180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.993871928925466e-06,
+      "loss": 0.6731,
+      "step": 190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.993210043424665e-06,
+      "loss": 0.6662,
+      "step": 200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.992514246587428e-06,
+      "loss": 0.6689,
+      "step": 210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.99178454313925e-06,
+      "loss": 0.6584,
+      "step": 220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.991020938035905e-06,
+      "loss": 0.6227,
+      "step": 230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.990223436463409e-06,
+      "loss": 0.6527,
+      "step": 240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.989392043837988e-06,
+      "loss": 0.6694,
+      "step": 250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.988526765806035e-06,
+      "loss": 0.6258,
+      "step": 260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.987627608244078e-06,
+      "loss": 0.662,
+      "step": 270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.986694577258736e-06,
+      "loss": 0.6307,
+      "step": 280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.985727679186681e-06,
+      "loss": 0.6429,
+      "step": 290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.98472692059459e-06,
+      "loss": 0.6463,
+      "step": 300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.983692308279105e-06,
+      "loss": 0.6542,
+      "step": 310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.982623849266788e-06,
+      "loss": 0.6509,
+      "step": 320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.981521550814064e-06,
+      "loss": 0.6545,
+      "step": 330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.980385420407184e-06,
+      "loss": 0.6437,
+      "step": 340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.979215465762168e-06,
+      "loss": 0.6401,
+      "step": 350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.978011694824747e-06,
+      "loss": 0.6446,
+      "step": 360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.976774115770325e-06,
+      "loss": 0.6369,
+      "step": 370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.975502737003903e-06,
+      "loss": 0.6756,
+      "step": 380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.97419756716004e-06,
+      "loss": 0.6389,
+      "step": 390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.97285861510278e-06,
+      "loss": 0.6676,
+      "step": 400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.97148588992561e-06,
+      "loss": 0.648,
+      "step": 410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.970079400951372e-06,
+      "loss": 0.5976,
+      "step": 420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.968639157732225e-06,
+      "loss": 0.6726,
+      "step": 430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.967165170049568e-06,
+      "loss": 0.6316,
+      "step": 440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.965657447913973e-06,
+      "loss": 0.6561,
+      "step": 450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.964116001565119e-06,
+      "loss": 0.6254,
+      "step": 460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.962540841471721e-06,
+      "loss": 0.6139,
+      "step": 470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.96093197833147e-06,
+      "loss": 0.6481,
+      "step": 480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.959289423070937e-06,
+      "loss": 0.6078,
+      "step": 490
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.957613186845522e-06,
+      "loss": 0.6448,
+      "step": 500
+    },
+    {
+      "epoch": 0.21,
+      "eval_accuracy": 0.6346886912325286,
+      "eval_loss": 0.6498483419418335,
+      "eval_runtime": 51.4302,
+      "eval_samples_per_second": 30.605,
+      "eval_steps_per_second": 3.83,
+      "step": 500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.95590328103937e-06,
+      "loss": 0.648,
+      "step": 510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.954159717265285e-06,
+      "loss": 0.6377,
+      "step": 520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.952382507364665e-06,
+      "loss": 0.6272,
+      "step": 530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.950571663407414e-06,
+      "loss": 0.6343,
+      "step": 540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.94872719769186e-06,
+      "loss": 0.6495,
+      "step": 550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.946849122744671e-06,
+      "loss": 0.6296,
+      "step": 560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.944937451320779e-06,
+      "loss": 0.6717,
+      "step": 570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.942992196403279e-06,
+      "loss": 0.6254,
+      "step": 580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.94101337120335e-06,
+      "loss": 0.6469,
+      "step": 590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.939000989160166e-06,
+      "loss": 0.6624,
+      "step": 600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.936955063940796e-06,
+      "loss": 0.6625,
+      "step": 610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.934875609440125e-06,
+      "loss": 0.6321,
+      "step": 620
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.932762639780745e-06,
+      "loss": 0.657,
+      "step": 630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.930616169312869e-06,
+      "loss": 0.6154,
+      "step": 640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.92843621261423e-06,
+      "loss": 0.6238,
+      "step": 650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.926222784489981e-06,
+      "loss": 0.6132,
+      "step": 660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.924202093528908e-06,
+      "loss": 0.6121,
+      "step": 670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.92192511129827e-06,
+      "loss": 0.6568,
+      "step": 680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.919614701862104e-06,
+      "loss": 0.6188,
+      "step": 690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.917270880911531e-06,
+      "loss": 0.6597,
+      "step": 700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.914893664364591e-06,
+      "loss": 0.627,
+      "step": 710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.912483068366122e-06,
+      "loss": 0.6286,
+      "step": 720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.910039109287665e-06,
+      "loss": 0.6645,
+      "step": 730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.907561803727344e-06,
+      "loss": 0.6654,
+      "step": 740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.905051168509756e-06,
+      "loss": 0.609,
+      "step": 750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.902507220685854e-06,
+      "loss": 0.6233,
+      "step": 760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.899929977532833e-06,
+      "loss": 0.6424,
+      "step": 770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.897319456554016e-06,
+      "loss": 0.6119,
+      "step": 780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.89467567547873e-06,
+      "loss": 0.6272,
+      "step": 790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.891998652262187e-06,
+      "loss": 0.6372,
+      "step": 800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.889288405085366e-06,
+      "loss": 0.6072,
+      "step": 810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.886544952354884e-06,
+      "loss": 0.6256,
+      "step": 820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.883768312702867e-06,
+      "loss": 0.593,
+      "step": 830
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.880958504986844e-06,
+      "loss": 0.5883,
+      "step": 840
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.87811554828959e-06,
+      "loss": 0.6027,
+      "step": 850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.875239461919021e-06,
+      "loss": 0.586,
+      "step": 860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.872330265408047e-06,
+      "loss": 0.6206,
+      "step": 870
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.869683695703284e-06,
+      "loss": 0.603,
+      "step": 880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.866711644543966e-06,
+      "loss": 0.6394,
+      "step": 890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.863706541160834e-06,
+      "loss": 0.6399,
+      "step": 900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.860668405963011e-06,
+      "loss": 0.6058,
+      "step": 910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.857597259583967e-06,
+      "loss": 0.6303,
+      "step": 920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.85449312288136e-06,
+      "loss": 0.5934,
+      "step": 930
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.851356016936901e-06,
+      "loss": 0.6505,
+      "step": 940
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.848185963056221e-06,
+      "loss": 0.6406,
+      "step": 950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.844982982768702e-06,
+      "loss": 0.6187,
+      "step": 960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.841747097827358e-06,
+      "loss": 0.6269,
+      "step": 970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.83847833020867e-06,
+      "loss": 0.6239,
+      "step": 980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.835176702112435e-06,
+      "loss": 0.6498,
+      "step": 990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.831842235961632e-06,
+      "loss": 0.6401,
+      "step": 1000
+    },
+    {
+      "epoch": 0.41,
+      "eval_accuracy": 0.6442185514612452,
+      "eval_loss": 0.6312425136566162,
+      "eval_runtime": 51.5654,
+      "eval_samples_per_second": 30.524,
+      "eval_steps_per_second": 3.82,
+      "step": 1000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.828474954402252e-06,
+      "loss": 0.6169,
+      "step": 1010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.825074880303159e-06,
+      "loss": 0.6195,
+      "step": 1020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.821642036755916e-06,
+      "loss": 0.6259,
+      "step": 1030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.818176447074652e-06,
+      "loss": 0.5888,
+      "step": 1040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.814678134795882e-06,
+      "loss": 0.6526,
+      "step": 1050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.811147123678359e-06,
+      "loss": 0.6121,
+      "step": 1060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.80758343770291e-06,
+      "loss": 0.6318,
+      "step": 1070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.803987101072272e-06,
+      "loss": 0.6432,
+      "step": 1080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.800358138210932e-06,
+      "loss": 0.6623,
+      "step": 1090
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.796696573764955e-06,
+      "loss": 0.6156,
+      "step": 1100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.793002432601817e-06,
+      "loss": 0.6342,
+      "step": 1110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.789275739810243e-06,
+      "loss": 0.6234,
+      "step": 1120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.78551652070003e-06,
+      "loss": 0.6459,
+      "step": 1130
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.781724800801874e-06,
+      "loss": 0.609,
+      "step": 1140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.77790060586721e-06,
+      "loss": 0.628,
+      "step": 1150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.774043961868014e-06,
+      "loss": 0.6102,
+      "step": 1160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.770154894996648e-06,
+      "loss": 0.5822,
+      "step": 1170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.76623343166567e-06,
+      "loss": 0.6268,
+      "step": 1180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.762279598507659e-06,
+      "loss": 0.617,
+      "step": 1190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.758293422375034e-06,
+      "loss": 0.6248,
+      "step": 1200
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.754274930339874e-06,
+      "loss": 0.6171,
+      "step": 1210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.750224149693721e-06,
+      "loss": 0.6359,
+      "step": 1220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.746141107947417e-06,
+      "loss": 0.641,
+      "step": 1230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.742025832830898e-06,
+      "loss": 0.627,
+      "step": 1240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.737878352293014e-06,
+      "loss": 0.6222,
+      "step": 1250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.733698694501337e-06,
+      "loss": 0.5874,
+      "step": 1260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.729486887841972e-06,
+      "loss": 0.6209,
+      "step": 1270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.72524296091936e-06,
+      "loss": 0.6425,
+      "step": 1280
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.720966942556094e-06,
+      "loss": 0.6502,
+      "step": 1290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.716658861792708e-06,
+      "loss": 0.6416,
+      "step": 1300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.712318747887486e-06,
+      "loss": 0.6209,
+      "step": 1310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.707946630316271e-06,
+      "loss": 0.6372,
+      "step": 1320
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.70354253877226e-06,
+      "loss": 0.6564,
+      "step": 1330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.69910650316579e-06,
+      "loss": 0.63,
+      "step": 1340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.694638553624157e-06,
+      "loss": 0.6399,
+      "step": 1350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.690138720491395e-06,
+      "loss": 0.6043,
+      "step": 1360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.685607034328077e-06,
+      "loss": 0.5991,
+      "step": 1370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.681043525911102e-06,
+      "loss": 0.6398,
+      "step": 1380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.676448226233495e-06,
+      "loss": 0.6058,
+      "step": 1390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.671821166504187e-06,
+      "loss": 0.6727,
+      "step": 1400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.667162378147809e-06,
+      "loss": 0.6395,
+      "step": 1410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.662471892804475e-06,
+      "loss": 0.6209,
+      "step": 1420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.657749742329568e-06,
+      "loss": 0.5832,
+      "step": 1430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.652995958793529e-06,
+      "loss": 0.6323,
+      "step": 1440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.648210574481629e-06,
+      "loss": 0.6029,
+      "step": 1450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.643393621893762e-06,
+      "loss": 0.6325,
+      "step": 1460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.638545133744208e-06,
+      "loss": 0.6226,
+      "step": 1470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.633665142961433e-06,
+      "loss": 0.6154,
+      "step": 1480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.628753682687839e-06,
+      "loss": 0.608,
+      "step": 1490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.623810786279563e-06,
+      "loss": 0.6557,
+      "step": 1500
+    },
+    {
+      "epoch": 0.62,
+      "eval_accuracy": 0.6581956797966964,
+      "eval_loss": 0.6313868165016174,
+      "eval_runtime": 51.0498,
+      "eval_samples_per_second": 30.833,
+      "eval_steps_per_second": 3.859,
+      "step": 1500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.61883648730624e-06,
+      "loss": 0.6109,
+      "step": 1510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.613830819550767e-06,
+      "loss": 0.6391,
+      "step": 1520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.60879381700909e-06,
+      "loss": 0.6,
+      "step": 1530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.603725513889959e-06,
+      "loss": 0.6411,
+      "step": 1540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.598625944614705e-06,
+      "loss": 0.6304,
+      "step": 1550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.593495143817004e-06,
+      "loss": 0.6242,
+      "step": 1560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.588333146342639e-06,
+      "loss": 0.6461,
+      "step": 1570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.583139987249262e-06,
+      "loss": 0.5948,
+      "step": 1580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.577915701806161e-06,
+      "loss": 0.5828,
+      "step": 1590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.572660325494022e-06,
+      "loss": 0.5988,
+      "step": 1600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.567373894004682e-06,
+      "loss": 0.5526,
+      "step": 1610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.562056443240882e-06,
+      "loss": 0.6564,
+      "step": 1620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.55670800931604e-06,
+      "loss": 0.6172,
+      "step": 1630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.551328628553993e-06,
+      "loss": 0.5986,
+      "step": 1640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.54591833748875e-06,
+      "loss": 0.6407,
+      "step": 1650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.540477172864249e-06,
+      "loss": 0.6337,
+      "step": 1660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.535005171634106e-06,
+      "loss": 0.5735,
+      "step": 1670
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.529502370961364e-06,
+      "loss": 0.6253,
+      "step": 1680
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.523968808218236e-06,
+      "loss": 0.6198,
+      "step": 1690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.518404520985862e-06,
+      "loss": 0.6293,
+      "step": 1700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.512809547054041e-06,
+      "loss": 0.6366,
+      "step": 1710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.507183924420985e-06,
+      "loss": 0.6298,
+      "step": 1720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.501527691293052e-06,
+      "loss": 0.6193,
+      "step": 1730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.495840886084498e-06,
+      "loss": 0.6307,
+      "step": 1740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.490123547417198e-06,
+      "loss": 0.6284,
+      "step": 1750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.484375714120405e-06,
+      "loss": 0.641,
+      "step": 1760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.478597425230466e-06,
+      "loss": 0.6026,
+      "step": 1770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.47278871999058e-06,
+      "loss": 0.6026,
+      "step": 1780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.466949637850507e-06,
+      "loss": 0.6094,
+      "step": 1790
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.461080218466316e-06,
+      "loss": 0.6217,
+      "step": 1800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.455180501700113e-06,
+      "loss": 0.5985,
+      "step": 1810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.44925052761976e-06,
+      "loss": 0.642,
+      "step": 1820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.44329033649862e-06,
+      "loss": 0.6266,
+      "step": 1830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.437299968815274e-06,
+      "loss": 0.606,
+      "step": 1840
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.431279465253244e-06,
+      "loss": 0.6194,
+      "step": 1850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.42522886670072e-06,
+      "loss": 0.5998,
+      "step": 1860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.419148214250278e-06,
+      "loss": 0.5689,
+      "step": 1870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.413037549198613e-06,
+      "loss": 0.5994,
+      "step": 1880
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.406896913046244e-06,
+      "loss": 0.6436,
+      "step": 1890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.400726347497243e-06,
+      "loss": 0.6326,
+      "step": 1900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.394525894458943e-06,
+      "loss": 0.6285,
+      "step": 1910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.388295596041658e-06,
+      "loss": 0.5859,
+      "step": 1920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.382035494558401e-06,
+      "loss": 0.609,
+      "step": 1930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.375745632524586e-06,
+      "loss": 0.6284,
+      "step": 1940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.369426052657751e-06,
+      "loss": 0.6237,
+      "step": 1950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.363076797877258e-06,
+      "loss": 0.6411,
+      "step": 1960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.356697911304006e-06,
+      "loss": 0.6322,
+      "step": 1970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.350289436260146e-06,
+      "loss": 0.5894,
+      "step": 1980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.343851416268768e-06,
+      "loss": 0.6418,
+      "step": 1990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.337383895053622e-06,
+      "loss": 0.5819,
+      "step": 2000
+    },
+    {
+      "epoch": 0.83,
+      "eval_accuracy": 0.658831003811944,
+      "eval_loss": 0.6320155262947083,
+      "eval_runtime": 51.0571,
+      "eval_samples_per_second": 30.828,
+      "eval_steps_per_second": 3.858,
+      "step": 2000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.330886916538813e-06,
+      "loss": 0.5965,
+      "step": 2010
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.324360524848508e-06,
+      "loss": 0.5914,
+      "step": 2020
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.31780476430663e-06,
+      "loss": 0.6031,
+      "step": 2030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.31121967943656e-06,
+      "loss": 0.6282,
+      "step": 2040
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.304605314960842e-06,
+      "loss": 0.5813,
+      "step": 2050
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.297961715800862e-06,
+      "loss": 0.6214,
+      "step": 2060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.291288927076559e-06,
+      "loss": 0.6366,
+      "step": 2070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.284586994106111e-06,
+      "loss": 0.6155,
+      "step": 2080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.27785596240563e-06,
+      "loss": 0.6308,
+      "step": 2090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.27109587768885e-06,
+      "loss": 0.6273,
+      "step": 2100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.264306785866821e-06,
+      "loss": 0.6027,
+      "step": 2110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.257488733047591e-06,
+      "loss": 0.5968,
+      "step": 2120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.250641765535897e-06,
+      "loss": 0.5679,
+      "step": 2130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.243765929832854e-06,
+      "loss": 0.5994,
+      "step": 2140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.236861272635631e-06,
+      "loss": 0.5943,
+      "step": 2150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.229927840837137e-06,
+      "loss": 0.6378,
+      "step": 2160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.222965681525707e-06,
+      "loss": 0.6334,
+      "step": 2170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.215974841984775e-06,
+      "loss": 0.6191,
+      "step": 2180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.208955369692561e-06,
+      "loss": 0.6456,
+      "step": 2190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.201907312321739e-06,
+      "loss": 0.6001,
+      "step": 2200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.19483071773912e-06,
+      "loss": 0.5984,
+      "step": 2210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.187725634005332e-06,
+      "loss": 0.6105,
+      "step": 2220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.180592109374472e-06,
+      "loss": 0.6204,
+      "step": 2230
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.173430192293807e-06,
+      "loss": 0.6247,
+      "step": 2240
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.166239931403425e-06,
+      "loss": 0.609,
+      "step": 2250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.15902137553591e-06,
+      "loss": 0.6355,
+      "step": 2260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.151774573716015e-06,
+      "loss": 0.6123,
+      "step": 2270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.144499575160319e-06,
+      "loss": 0.6175,
+      "step": 2280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.137196429276906e-06,
+      "loss": 0.5905,
+      "step": 2290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.129865185665019e-06,
+      "loss": 0.6168,
+      "step": 2300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.122505894114723e-06,
+      "loss": 0.6398,
+      "step": 2310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.115118604606574e-06,
+      "loss": 0.5965,
+      "step": 2320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.107703367311278e-06,
+      "loss": 0.6096,
+      "step": 2330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.100260232589342e-06,
+      "loss": 0.62,
+      "step": 2340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.092789250990745e-06,
+      "loss": 0.6141,
+      "step": 2350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.085290473254584e-06,
+      "loss": 0.6006,
+      "step": 2360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.07776395030873e-06,
+      "loss": 0.6166,
+      "step": 2370
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.070209733269493e-06,
+      "loss": 0.6047,
+      "step": 2380
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.062627873441258e-06,
+      "loss": 0.6005,
+      "step": 2390
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.055018422316158e-06,
+      "loss": 0.6192,
+      "step": 2400
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.047381431573699e-06,
+      "loss": 0.59,
+      "step": 2410
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.039716953080428e-06,
+      "loss": 0.6295,
+      "step": 2420
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.032025038889579e-06,
+      "loss": 0.5751,
+      "step": 2430
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.024305741240707e-06,
+      "loss": 0.5837,
+      "step": 2440
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.016559112559346e-06,
+      "loss": 0.5758,
+      "step": 2450
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.008785205456649e-06,
+      "loss": 0.5702,
+      "step": 2460
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 9.000984072729026e-06,
+      "loss": 0.5963,
+      "step": 2470
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.993155767357797e-06,
+      "loss": 0.5829,
+      "step": 2480
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.985300342508818e-06,
+      "loss": 0.6382,
+      "step": 2490
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.977417851532127e-06,
+      "loss": 0.6086,
+      "step": 2500
+    },
+    {
+      "epoch": 1.04,
+      "eval_accuracy": 0.656289707750953,
+      "eval_loss": 0.6343497633934021,
+      "eval_runtime": 50.9481,
+      "eval_samples_per_second": 30.894,
+      "eval_steps_per_second": 3.867,
+      "step": 2500
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.969508347961583e-06,
+      "loss": 0.5933,
+      "step": 2510
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.961571885514506e-06,
+      "loss": 0.5909,
+      "step": 2520
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.953608518091299e-06,
+      "loss": 0.6408,
+      "step": 2530
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.945618299775091e-06,
+      "loss": 0.6553,
+      "step": 2540
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.937601284831373e-06,
+      "loss": 0.6035,
+      "step": 2550
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.92955752770762e-06,
+      "loss": 0.6013,
+      "step": 2560
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.921487083032929e-06,
+      "loss": 0.5957,
+      "step": 2570
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.913390005617649e-06,
+      "loss": 0.6078,
+      "step": 2580
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.905266350452995e-06,
+      "loss": 0.5967,
+      "step": 2590
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.897116172710694e-06,
+      "loss": 0.621,
+      "step": 2600
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.888939527742596e-06,
+      "loss": 0.5993,
+      "step": 2610
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.880736471080304e-06,
+      "loss": 0.58,
+      "step": 2620
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.872507058434798e-06,
+      "loss": 0.5769,
+      "step": 2630
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.86425134569605e-06,
+      "loss": 0.5661,
+      "step": 2640
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.85596938893265e-06,
+      "loss": 0.6389,
+      "step": 2650
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.84766124439143e-06,
+      "loss": 0.5874,
+      "step": 2660
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.839326968497067e-06,
+      "loss": 0.5938,
+      "step": 2670
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.83180382466322e-06,
+      "loss": 0.6148,
+      "step": 2680
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.823420055283185e-06,
+      "loss": 0.5892,
+      "step": 2690
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.815010319183804e-06,
+      "loss": 0.6012,
+      "step": 2700
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.806574673479715e-06,
+      "loss": 0.5919,
+      "step": 2710
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.798113175461522e-06,
+      "loss": 0.5844,
+      "step": 2720
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.789625882595405e-06,
+      "loss": 0.5946,
+      "step": 2730
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.781112852522728e-06,
+      "loss": 0.5648,
+      "step": 2740
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.772574143059655e-06,
+      "loss": 0.6119,
+      "step": 2750
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.764009812196743e-06,
+      "loss": 0.5653,
+      "step": 2760
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.755419918098557e-06,
+      "loss": 0.5999,
+      "step": 2770
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.746804519103284e-06,
+      "loss": 0.6357,
+      "step": 2780
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.738163673722318e-06,
+      "loss": 0.6187,
+      "step": 2790
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.729497440639872e-06,
+      "loss": 0.5753,
+      "step": 2800
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.720805878712584e-06,
+      "loss": 0.6289,
+      "step": 2810
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.712089046969111e-06,
+      "loss": 0.5614,
+      "step": 2820
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.703347004609728e-06,
+      "loss": 0.6035,
+      "step": 2830
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.694579811005933e-06,
+      "loss": 0.5728,
+      "step": 2840
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.685787525700028e-06,
+      "loss": 0.6046,
+      "step": 2850
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.676970208404737e-06,
+      "loss": 0.5726,
+      "step": 2860
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.668127919002777e-06,
+      "loss": 0.5725,
+      "step": 2870
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.659260717546474e-06,
+      "loss": 0.6082,
+      "step": 2880
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.650368664257335e-06,
+      "loss": 0.6188,
+      "step": 2890
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.641451819525655e-06,
+      "loss": 0.6282,
+      "step": 2900
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.632510243910094e-06,
+      "loss": 0.5896,
+      "step": 2910
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.623543998137276e-06,
+      "loss": 0.6138,
+      "step": 2920
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.614553143101368e-06,
+      "loss": 0.6081,
+      "step": 2930
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.605537739863674e-06,
+      "loss": 0.5918,
+      "step": 2940
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.596497849652215e-06,
+      "loss": 0.5911,
+      "step": 2950
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.587433533861316e-06,
+      "loss": 0.579,
+      "step": 2960
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.578344854051187e-06,
+      "loss": 0.5432,
+      "step": 2970
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.569231871947507e-06,
+      "loss": 0.6095,
+      "step": 2980
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.560094649441006e-06,
+      "loss": 0.5996,
+      "step": 2990
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.550933248587036e-06,
+      "loss": 0.6011,
+      "step": 3000
+    },
+    {
+      "epoch": 1.24,
+      "eval_accuracy": 0.6556543837357052,
+      "eval_loss": 0.616517186164856,
+      "eval_runtime": 50.9988,
+      "eval_samples_per_second": 30.863,
+      "eval_steps_per_second": 3.863,
+      "step": 3000
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.54174773160516e-06,
+      "loss": 0.6363,
+      "step": 3010
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.53253816087873e-06,
+      "loss": 0.5931,
+      "step": 3020
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.523304598954449e-06,
+      "loss": 0.5681,
+      "step": 3030
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.514047108541965e-06,
+      "loss": 0.5712,
+      "step": 3040
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.50476575251343e-06,
+      "loss": 0.6139,
+      "step": 3050
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.495460593903083e-06,
+      "loss": 0.6135,
+      "step": 3060
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.486131695906813e-06,
+      "loss": 0.5973,
+      "step": 3070
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.47677912188174e-06,
+      "loss": 0.5557,
+      "step": 3080
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.467402935345778e-06,
+      "loss": 0.6532,
+      "step": 3090
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.458003199977207e-06,
+      "loss": 0.6305,
+      "step": 3100
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.448579979614234e-06,
+      "loss": 0.6029,
+      "step": 3110
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.439133338254566e-06,
+      "loss": 0.594,
+      "step": 3120
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.429663340054974e-06,
+      "loss": 0.5722,
+      "step": 3130
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.420170049330857e-06,
+      "loss": 0.5835,
+      "step": 3140
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.410653530555806e-06,
+      "loss": 0.5997,
+      "step": 3150
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.401113848361161e-06,
+      "loss": 0.5717,
+      "step": 3160
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.391551067535586e-06,
+      "loss": 0.6137,
+      "step": 3170
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.381965253024603e-06,
+      "loss": 0.6025,
+      "step": 3180
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.37235646993018e-06,
+      "loss": 0.5866,
+      "step": 3190
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.362724783510273e-06,
+      "loss": 0.6271,
+      "step": 3200
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.353070259178385e-06,
+      "loss": 0.5826,
+      "step": 3210
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.343392962503118e-06,
+      "loss": 0.5705,
+      "step": 3220
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.333692959207739e-06,
+      "loss": 0.6393,
+      "step": 3230
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.323970315169728e-06,
+      "loss": 0.5504,
+      "step": 3240
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.31422509642032e-06,
+      "loss": 0.584,
+      "step": 3250
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.30445736914407e-06,
+      "loss": 0.5725,
+      "step": 3260
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.294667199678405e-06,
+      "loss": 0.5823,
+      "step": 3270
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.284854654513165e-06,
+      "loss": 0.5999,
+      "step": 3280
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.275019800290149e-06,
+      "loss": 0.6067,
+      "step": 3290
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.265162703802674e-06,
+      "loss": 0.6204,
+      "step": 3300
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.255283431995111e-06,
+      "loss": 0.5953,
+      "step": 3310
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.24538205196244e-06,
+      "loss": 0.6242,
+      "step": 3320
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.235458630949782e-06,
+      "loss": 0.5856,
+      "step": 3330
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.225513236351954e-06,
+      "loss": 0.5746,
+      "step": 3340
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.215545935713003e-06,
+      "loss": 0.6144,
+      "step": 3350
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.205556796725752e-06,
+      "loss": 0.5943,
+      "step": 3360
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.195545887231345e-06,
+      "loss": 0.617,
+      "step": 3370
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.185513275218767e-06,
+      "loss": 0.6239,
+      "step": 3380
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.175459028824407e-06,
+      "loss": 0.5933,
+      "step": 3390
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.165383216331578e-06,
+      "loss": 0.6187,
+      "step": 3400
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.155285906170059e-06,
+      "loss": 0.6052,
+      "step": 3410
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.145167166915636e-06,
+      "loss": 0.599,
+      "step": 3420
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.135027067289622e-06,
+      "loss": 0.6392,
+      "step": 3430
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.124865676158407e-06,
+      "loss": 0.5939,
+      "step": 3440
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.114683062532975e-06,
+      "loss": 0.5753,
+      "step": 3450
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.104479295568446e-06,
+      "loss": 0.5904,
+      "step": 3460
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.094254444563601e-06,
+      "loss": 0.5807,
+      "step": 3470
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.084008578960417e-06,
+      "loss": 0.5588,
+      "step": 3480
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.073741768343586e-06,
+      "loss": 0.6072,
+      "step": 3490
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.063454082440052e-06,
+      "loss": 0.5616,
+      "step": 3500
+    },
+    {
+      "epoch": 1.45,
+      "eval_accuracy": 0.6461245235069886,
+      "eval_loss": 0.6375634074211121,
+      "eval_runtime": 50.9745,
+      "eval_samples_per_second": 30.878,
+      "eval_steps_per_second": 3.865,
+      "step": 3500
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.05314559111853e-06,
+      "loss": 0.5951,
+      "step": 3510
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.042816364389038e-06,
+      "loss": 0.5719,
+      "step": 3520
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.032466472402416e-06,
+      "loss": 0.5994,
+      "step": 3530
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.02209598544985e-06,
+      "loss": 0.567,
+      "step": 3540
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.011704973962406e-06,
+      "loss": 0.629,
+      "step": 3550
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.001293508510531e-06,
+      "loss": 0.5924,
+      "step": 3560
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.990861659803589e-06,
+      "loss": 0.5885,
+      "step": 3570
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.980409498689376e-06,
+      "loss": 0.5501,
+      "step": 3580
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.969937096153643e-06,
+      "loss": 0.6151,
+      "step": 3590
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.959444523319605e-06,
+      "loss": 0.5915,
+      "step": 3600
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.948931851447472e-06,
+      "loss": 0.6035,
+      "step": 3610
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.938399151933944e-06,
+      "loss": 0.6333,
+      "step": 3620
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.927846496311749e-06,
+      "loss": 0.5883,
+      "step": 3630
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.917273956249139e-06,
+      "loss": 0.6129,
+      "step": 3640
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 7.90668160354942e-06,
+      "loss": 0.5985,
+      "step": 3650
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.896069510150447e-06,
+      "loss": 0.6103,
+      "step": 3660
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 7.885437748124148e-06,
+      "loss": 0.5561,
+      "step": 3670
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.87478638967603e-06,
+      "loss": 0.5401,
+      "step": 3680
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.864115507144684e-06,
+      "loss": 0.5606,
+      "step": 3690
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 7.85342517300131e-06,
+      "loss": 0.553,
+      "step": 3700
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 7.842715459849201e-06,
+      "loss": 0.6264,
+      "step": 3710
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 7.831986440423269e-06,
+      "loss": 0.5893,
+      "step": 3720
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 7.82123818758954e-06,
+      "loss": 0.5745,
+      "step": 3730
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 7.810470774344671e-06,
+      "loss": 0.6288,
+      "step": 3740
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 7.79968427381544e-06,
+      "loss": 0.5945,
+      "step": 3750
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 7.788878759258261e-06,
+      "loss": 0.5824,
+      "step": 3760
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 7.778054304058672e-06,
+      "loss": 0.545,
+      "step": 3770
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 7.767210981730859e-06,
+      "loss": 0.6237,
+      "step": 3780
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 7.756348865917135e-06,
+      "loss": 0.5741,
+      "step": 3790
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 7.745468030387456e-06,
+      "loss": 0.6197,
+      "step": 3800
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 7.734568549038904e-06,
+      "loss": 0.5892,
+      "step": 3810
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 7.723650495895199e-06,
+      "loss": 0.5712,
+      "step": 3820
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 7.712713945106192e-06,
+      "loss": 0.5847,
+      "step": 3830
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 7.701758970947357e-06,
+      "loss": 0.6221,
+      "step": 3840
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 7.690785647819297e-06,
+      "loss": 0.6087,
+      "step": 3850
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 7.67979405024722e-06,
+      "loss": 0.5989,
+      "step": 3860
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 7.66878425288046e-06,
+      "loss": 0.6044,
+      "step": 3870
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 7.657756330491937e-06,
+      "loss": 0.6155,
+      "step": 3880
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 7.646710357977685e-06,
+      "loss": 0.6068,
+      "step": 3890
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 7.635646410356312e-06,
+      "loss": 0.5871,
+      "step": 3900
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 7.6245645627685086e-06,
+      "loss": 0.5588,
+      "step": 3910
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 7.613464890476534e-06,
+      "loss": 0.5588,
+      "step": 3920
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 7.602347468863702e-06,
+      "loss": 0.5886,
+      "step": 3930
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 7.5912123734338724e-06,
+      "loss": 0.5818,
+      "step": 3940
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 7.580059679810934e-06,
+      "loss": 0.6092,
+      "step": 3950
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 7.5688894637382935e-06,
+      "loss": 0.6137,
+      "step": 3960
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 7.557701801078365e-06,
+      "loss": 0.5907,
+      "step": 3970
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 7.546496767812046e-06,
+      "loss": 0.5777,
+      "step": 3980
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 7.53527444003821e-06,
+      "loss": 0.6078,
+      "step": 3990
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 7.524034893973185e-06,
+      "loss": 0.5885,
+      "step": 4000
+    },
+    {
+      "epoch": 1.66,
+      "eval_accuracy": 0.6467598475222364,
+      "eval_loss": 0.6304181814193726,
+      "eval_runtime": 50.9704,
+      "eval_samples_per_second": 30.881,
+      "eval_steps_per_second": 3.865,
+      "step": 4000
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 7.512778205950236e-06,
+      "loss": 0.6168,
+      "step": 4010
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 7.501504452419048e-06,
+      "loss": 0.6261,
+      "step": 4020
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 7.490213709945205e-06,
+      "loss": 0.5942,
+      "step": 4030
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 7.478906055209675e-06,
+      "loss": 0.5862,
+      "step": 4040
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 7.467581565008281e-06,
+      "loss": 0.6262,
+      "step": 4050
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 7.456240316251187e-06,
+      "loss": 0.5593,
+      "step": 4060
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.444882385962369e-06,
+      "loss": 0.6106,
+      "step": 4070
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 7.433507851279101e-06,
+      "loss": 0.5816,
+      "step": 4080
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.422116789451419e-06,
+      "loss": 0.5834,
+      "step": 4090
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.41070927784161e-06,
+      "loss": 0.6138,
+      "step": 4100
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 7.399285393923672e-06,
+      "loss": 0.5899,
+      "step": 4110
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.387845215282803e-06,
+      "loss": 0.5979,
+      "step": 4120
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 7.376388819614859e-06,
+      "loss": 0.5719,
+      "step": 4130
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.364916284725839e-06,
+      "loss": 0.5875,
+      "step": 4140
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.353427688531352e-06,
+      "loss": 0.6164,
+      "step": 4150
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.34192310905608e-06,
+      "loss": 0.5647,
+      "step": 4160
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.330402624433266e-06,
+      "loss": 0.5931,
+      "step": 4170
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 7.318866312904163e-06,
+      "loss": 0.6099,
+      "step": 4180
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.307314252817519e-06,
+      "loss": 0.5736,
+      "step": 4190
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 7.295746522629032e-06,
+      "loss": 0.5992,
+      "step": 4200
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.284163200900829e-06,
+      "loss": 0.6159,
+      "step": 4210
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.272564366300924e-06,
+      "loss": 0.557,
+      "step": 4220
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 7.260950097602687e-06,
+      "loss": 0.6099,
+      "step": 4230
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.2493204736843105e-06,
+      "loss": 0.5766,
+      "step": 4240
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 7.237675573528267e-06,
+      "loss": 0.5919,
+      "step": 4250
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.22601547622078e-06,
+      "loss": 0.5843,
+      "step": 4260
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 7.214340260951287e-06,
+      "loss": 0.5901,
+      "step": 4270
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.202650007011896e-06,
+      "loss": 0.5708,
+      "step": 4280
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.19094479379685e-06,
+      "loss": 0.624,
+      "step": 4290
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 7.179224700801987e-06,
+      "loss": 0.5714,
+      "step": 4300
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.167489807624205e-06,
+      "loss": 0.5953,
+      "step": 4310
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.155740193960913e-06,
+      "loss": 0.603,
+      "step": 4320
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.143975939609495e-06,
+      "loss": 0.54,
+      "step": 4330
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.13219712446677e-06,
+      "loss": 0.592,
+      "step": 4340
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.120403828528441e-06,
+      "loss": 0.5528,
+      "step": 4350
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.1085961318885635e-06,
+      "loss": 0.6071,
+      "step": 4360
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.096774114738988e-06,
+      "loss": 0.5681,
+      "step": 4370
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.084937857368831e-06,
+      "loss": 0.5935,
+      "step": 4380
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.073087440163914e-06,
+      "loss": 0.5798,
+      "step": 4390
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.061222943606228e-06,
+      "loss": 0.5674,
+      "step": 4400
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.049344448273384e-06,
+      "loss": 0.595,
+      "step": 4410
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.037452034838067e-06,
+      "loss": 0.6073,
+      "step": 4420
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.025545784067481e-06,
+      "loss": 0.5769,
+      "step": 4430
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.013625776822813e-06,
+      "loss": 0.607,
+      "step": 4440
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.001692094058673e-06,
+      "loss": 0.5659,
+      "step": 4450
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.989744816822548e-06,
+      "loss": 0.5722,
+      "step": 4460
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 6.977784026254255e-06,
+      "loss": 0.5717,
+      "step": 4470
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.965809803585382e-06,
+      "loss": 0.5705,
+      "step": 4480
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.953822230138745e-06,
+      "loss": 0.5551,
+      "step": 4490
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.941821387327829e-06,
+      "loss": 0.6198,
+      "step": 4500
+    },
+    {
+      "epoch": 1.87,
+      "eval_accuracy": 0.6423125794155019,
+      "eval_loss": 0.6448229551315308,
+      "eval_runtime": 51.0009,
+      "eval_samples_per_second": 30.862,
+      "eval_steps_per_second": 3.863,
+      "step": 4500
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.929807356656239e-06,
+      "loss": 0.648,
+      "step": 4510
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 6.917780219717145e-06,
+      "loss": 0.6189,
+      "step": 4520
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.905740058192729e-06,
+      "loss": 0.5731,
+      "step": 4530
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 6.893686953853629e-06,
+      "loss": 0.6248,
+      "step": 4540
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.881620988558379e-06,
+      "loss": 0.6132,
+      "step": 4550
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.869542244252869e-06,
+      "loss": 0.5959,
+      "step": 4560
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 6.857450802969766e-06,
+      "loss": 0.5954,
+      "step": 4570
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 6.845346746827976e-06,
+      "loss": 0.6321,
+      "step": 4580
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 6.833230158032076e-06,
+      "loss": 0.6009,
+      "step": 4590
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 6.821101118871756e-06,
+      "loss": 0.5741,
+      "step": 4600
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 6.808959711721269e-06,
+      "loss": 0.6012,
+      "step": 4610
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 6.7968060190388574e-06,
+      "loss": 0.5913,
+      "step": 4620
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 6.784640123366208e-06,
+      "loss": 0.6105,
+      "step": 4630
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 6.772462107327876e-06,
+      "loss": 0.5622,
+      "step": 4640
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.760272053630741e-06,
+      "loss": 0.6126,
+      "step": 4650
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 6.7480700450634295e-06,
+      "loss": 0.584,
+      "step": 4660
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 6.735856164495761e-06,
+      "loss": 0.5517,
+      "step": 4670
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 6.723630494878184e-06,
+      "loss": 0.6212,
+      "step": 4680
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.711393119241214e-06,
+      "loss": 0.6105,
+      "step": 4690
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.699144120694866e-06,
+      "loss": 0.58,
+      "step": 4700
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.68688358242809e-06,
+      "loss": 0.5887,
+      "step": 4710
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.674611587708212e-06,
+      "loss": 0.639,
+      "step": 4720
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.662328219880361e-06,
+      "loss": 0.5995,
+      "step": 4730
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.650033562366909e-06,
+      "loss": 0.6064,
+      "step": 4740
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.637727698666898e-06,
+      "loss": 0.5668,
+      "step": 4750
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.625410712355482e-06,
+      "loss": 0.5656,
+      "step": 4760
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.6130826870833496e-06,
+      "loss": 0.6278,
+      "step": 4770
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.600743706576163e-06,
+      "loss": 0.6113,
+      "step": 4780
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.588393854633986e-06,
+      "loss": 0.5991,
+      "step": 4790
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.5760332151307125e-06,
+      "loss": 0.5832,
+      "step": 4800
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.563661872013506e-06,
+      "loss": 0.567,
+      "step": 4810
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.551279909302221e-06,
+      "loss": 0.6161,
+      "step": 4820
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.53888741108883e-06,
+      "loss": 0.5546,
+      "step": 4830
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.526484461536867e-06,
+      "loss": 0.5325,
+      "step": 4840
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.514071144880838e-06,
+      "loss": 0.6121,
+      "step": 4850
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.50164754542566e-06,
+      "loss": 0.5507,
+      "step": 4860
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.489213747546085e-06,
+      "loss": 0.5565,
+      "step": 4870
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.4767698356861275e-06,
+      "loss": 0.5729,
+      "step": 4880
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.464315894358495e-06,
+      "loss": 0.5035,
+      "step": 4890
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.451852008144e-06,
+      "loss": 0.566,
+      "step": 4900
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.43937826169101e-06,
+      "loss": 0.5883,
+      "step": 4910
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4268947397148415e-06,
+      "loss": 0.557,
+      "step": 4920
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.414401526997217e-06,
+      "loss": 0.6021,
+      "step": 4930
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.401898708385662e-06,
+      "loss": 0.5828,
+      "step": 4940
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.389386368792949e-06,
+      "loss": 0.5707,
+      "step": 4950
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.376864593196505e-06,
+      "loss": 0.4898,
+      "step": 4960
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.364333466637846e-06,
+      "loss": 0.6099,
+      "step": 4970
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.351793074221992e-06,
+      "loss": 0.5568,
+      "step": 4980
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.339243501116896e-06,
+      "loss": 0.5708,
+      "step": 4990
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.326684832552859e-06,
+      "loss": 0.5838,
+      "step": 5000
+    },
+    {
+      "epoch": 2.07,
+      "eval_accuracy": 0.6664548919949174,
+      "eval_loss": 0.6319934725761414,
+      "eval_runtime": 50.9976,
+      "eval_samples_per_second": 30.864,
+      "eval_steps_per_second": 3.863,
+      "step": 5000
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.314117153821952e-06,
+      "loss": 0.5559,
+      "step": 5010
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.301540550277442e-06,
+      "loss": 0.5471,
+      "step": 5020
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.288955107333206e-06,
+      "loss": 0.6066,
+      "step": 5030
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.276360910463154e-06,
+      "loss": 0.5683,
+      "step": 5040
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.2637580452006505e-06,
+      "loss": 0.5419,
+      "step": 5050
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.2511465971379305e-06,
+      "loss": 0.5612,
+      "step": 5060
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.238526651925518e-06,
+      "loss": 0.5724,
+      "step": 5070
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 6.225898295271644e-06,
+      "loss": 0.567,
+      "step": 5080
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 6.213261612941668e-06,
+      "loss": 0.5523,
+      "step": 5090
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 6.200616690757493e-06,
+      "loss": 0.4972,
+      "step": 5100
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 6.187963614596982e-06,
+      "loss": 0.5604,
+      "step": 5110
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 6.175302470393378e-06,
+      "loss": 0.5925,
+      "step": 5120
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 6.162633344134712e-06,
+      "loss": 0.5419,
+      "step": 5130
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 6.149956321863235e-06,
+      "loss": 0.5809,
+      "step": 5140
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 6.137271489674813e-06,
+      "loss": 0.5348,
+      "step": 5150
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 6.1245789337183605e-06,
+      "loss": 0.5941,
+      "step": 5160
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 6.111878740195243e-06,
+      "loss": 0.5322,
+      "step": 5170
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 6.0991709953587e-06,
+      "loss": 0.6017,
+      "step": 5180
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 6.086455785513256e-06,
+      "loss": 0.5852,
+      "step": 5190
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 6.0737331970141285e-06,
+      "loss": 0.5577,
+      "step": 5200
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 6.061003316266657e-06,
+      "loss": 0.5299,
+      "step": 5210
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 6.048266229725694e-06,
+      "loss": 0.5185,
+      "step": 5220
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 6.035522023895042e-06,
+      "loss": 0.5758,
+      "step": 5230
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 6.022770785326842e-06,
+      "loss": 0.5524,
+      "step": 5240
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 6.010012600621009e-06,
+      "loss": 0.5431,
+      "step": 5250
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.997247556424626e-06,
+      "loss": 0.5662,
+      "step": 5260
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.984475739431364e-06,
+      "loss": 0.5404,
+      "step": 5270
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.97169723638089e-06,
+      "loss": 0.5755,
+      "step": 5280
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.9589121340582835e-06,
+      "loss": 0.5503,
+      "step": 5290
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.9461205192934394e-06,
+      "loss": 0.5627,
+      "step": 5300
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.93332247896048e-06,
+      "loss": 0.5393,
+      "step": 5310
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.920518099977172e-06,
+      "loss": 0.5923,
+      "step": 5320
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.9077074693043255e-06,
+      "loss": 0.5096,
+      "step": 5330
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.894890673945213e-06,
+      "loss": 0.546,
+      "step": 5340
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.882067800944974e-06,
+      "loss": 0.5621,
+      "step": 5350
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.86923893739002e-06,
+      "loss": 0.5399,
+      "step": 5360
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.856404170407455e-06,
+      "loss": 0.5159,
+      "step": 5370
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.843563587164468e-06,
+      "loss": 0.5437,
+      "step": 5380
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.830717274867753e-06,
+      "loss": 0.5919,
+      "step": 5390
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.8178653207629135e-06,
+      "loss": 0.5447,
+      "step": 5400
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.805007812133867e-06,
+      "loss": 0.5484,
+      "step": 5410
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.792144836302258e-06,
+      "loss": 0.5282,
+      "step": 5420
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.779276480626855e-06,
+      "loss": 0.5522,
+      "step": 5430
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 5.766402832502973e-06,
+      "loss": 0.53,
+      "step": 5440
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 5.7535239793618625e-06,
+      "loss": 0.5726,
+      "step": 5450
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 5.740640008670128e-06,
+      "loss": 0.5401,
+      "step": 5460
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 5.727751007929124e-06,
+      "loss": 0.576,
+      "step": 5470
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 5.714857064674378e-06,
+      "loss": 0.5673,
+      "step": 5480
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 5.701958266474973e-06,
+      "loss": 0.5385,
+      "step": 5490
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 5.68905470093297e-06,
+      "loss": 0.5564,
+      "step": 5500
+    },
+    {
+      "epoch": 2.28,
+      "eval_accuracy": 0.6683608640406608,
+      "eval_loss": 0.6428170800209045,
+      "eval_runtime": 50.9753,
+      "eval_samples_per_second": 30.878,
+      "eval_steps_per_second": 3.865,
+      "step": 5500
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 5.676146455682807e-06,
+      "loss": 0.6073,
+      "step": 5510
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 5.663233618390703e-06,
+      "loss": 0.5685,
+      "step": 5520
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 5.650316276754065e-06,
+      "loss": 0.5168,
+      "step": 5530
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 5.637394518500887e-06,
+      "loss": 0.6063,
+      "step": 5540
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 5.624468431389168e-06,
+      "loss": 0.5695,
+      "step": 5550
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 5.611538103206295e-06,
+      "loss": 0.5592,
+      "step": 5560
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 5.598603621768466e-06,
+      "loss": 0.5302,
+      "step": 5570
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 5.585665074920083e-06,
+      "loss": 0.6123,
+      "step": 5580
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 5.572722550533156e-06,
+      "loss": 0.5531,
+      "step": 5590
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 5.559776136506716e-06,
+      "loss": 0.6083,
+      "step": 5600
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 5.5468259207662e-06,
+      "loss": 0.587,
+      "step": 5610
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 5.5338719912628735e-06,
+      "loss": 0.5528,
+      "step": 5620
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 5.520914435973217e-06,
+      "loss": 0.5858,
+      "step": 5630
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 5.507953342898341e-06,
+      "loss": 0.5747,
+      "step": 5640
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 5.494988800063381e-06,
+      "loss": 0.5299,
+      "step": 5650
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 5.482020895516897e-06,
+      "loss": 0.5543,
+      "step": 5660
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 5.469049717330289e-06,
+      "loss": 0.5014,
+      "step": 5670
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 5.456075353597181e-06,
+      "loss": 0.6067,
+      "step": 5680
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 5.443097892432837e-06,
+      "loss": 0.5718,
+      "step": 5690
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 5.430117421973555e-06,
+      "loss": 0.5778,
+      "step": 5700
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 5.4171340303760735e-06,
+      "loss": 0.5496,
+      "step": 5710
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 5.404147805816966e-06,
+      "loss": 0.6065,
+      "step": 5720
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 5.391158836492045e-06,
+      "loss": 0.5875,
+      "step": 5730
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 5.3781672106157725e-06,
+      "loss": 0.5706,
+      "step": 5740
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 5.365173016420642e-06,
+      "loss": 0.5391,
+      "step": 5750
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 5.352176342156596e-06,
+      "loss": 0.5388,
+      "step": 5760
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 5.339177276090418e-06,
+      "loss": 0.5701,
+      "step": 5770
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 5.326175906505139e-06,
+      "loss": 0.5512,
+      "step": 5780
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 5.3131723216994265e-06,
+      "loss": 0.5423,
+      "step": 5790
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 5.300166609987002e-06,
+      "loss": 0.5806,
+      "step": 5800
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 5.287158859696025e-06,
+      "loss": 0.558,
+      "step": 5810
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 5.274149159168504e-06,
+      "loss": 0.5695,
+      "step": 5820
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 5.26113759675969e-06,
+      "loss": 0.5403,
+      "step": 5830
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 5.248124260837479e-06,
+      "loss": 0.5479,
+      "step": 5840
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 5.235109239781816e-06,
+      "loss": 0.5434,
+      "step": 5850
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 5.222092621984083e-06,
+      "loss": 0.5792,
+      "step": 5860
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 5.209074495846513e-06,
+      "loss": 0.5743,
+      "step": 5870
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 5.196054949781582e-06,
+      "loss": 0.5168,
+      "step": 5880
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 5.183034072211405e-06,
+      "loss": 0.5627,
+      "step": 5890
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 5.170011951567144e-06,
+      "loss": 0.5779,
+      "step": 5900
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 5.156988676288401e-06,
+      "loss": 0.5446,
+      "step": 5910
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 5.143964334822622e-06,
+      "loss": 0.5262,
+      "step": 5920
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 5.130939015624492e-06,
+      "loss": 0.5864,
+      "step": 5930
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 5.1179128071553375e-06,
+      "loss": 0.5432,
+      "step": 5940
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 5.104885797882525e-06,
+      "loss": 0.5907,
+      "step": 5950
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 5.091858076278855e-06,
+      "loss": 0.5548,
+      "step": 5960
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 5.078829730821974e-06,
+      "loss": 0.5233,
+      "step": 5970
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 5.065800849993758e-06,
+      "loss": 0.5669,
+      "step": 5980
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 5.052771522279723e-06,
+      "loss": 0.554,
+      "step": 5990
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 5.039741836168417e-06,
+      "loss": 0.5726,
+      "step": 6000
+    },
+    {
+      "epoch": 2.49,
+      "eval_accuracy": 0.670266836086404,
+      "eval_loss": 0.6400668025016785,
+      "eval_runtime": 51.0392,
+      "eval_samples_per_second": 30.839,
+      "eval_steps_per_second": 3.86,
+      "step": 6000
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 5.0267118801508265e-06,
+      "loss": 0.5433,
+      "step": 6010
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 5.013681742719767e-06,
+      "loss": 0.525,
+      "step": 6020
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 5.0006515123692874e-06,
+      "loss": 0.5502,
+      "step": 6030
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.987621277594069e-06,
+      "loss": 0.5682,
+      "step": 6040
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.974591126888818e-06,
+      "loss": 0.6182,
+      "step": 6050
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.961561148747679e-06,
+      "loss": 0.5702,
+      "step": 6060
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.948531431663616e-06,
+      "loss": 0.5598,
+      "step": 6070
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.935502064127825e-06,
+      "loss": 0.5793,
+      "step": 6080
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.922473134629124e-06,
+      "loss": 0.4937,
+      "step": 6090
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.909444731653359e-06,
+      "loss": 0.5261,
+      "step": 6100
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.896416943682802e-06,
+      "loss": 0.5149,
+      "step": 6110
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.883389859195542e-06,
+      "loss": 0.5502,
+      "step": 6120
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.870363566664895e-06,
+      "loss": 0.5321,
+      "step": 6130
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.857338154558798e-06,
+      "loss": 0.5824,
+      "step": 6140
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.844313711339209e-06,
+      "loss": 0.5606,
+      "step": 6150
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.831290325461504e-06,
+      "loss": 0.541,
+      "step": 6160
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.818268085373878e-06,
+      "loss": 0.5616,
+      "step": 6170
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.805247079516747e-06,
+      "loss": 0.5759,
+      "step": 6180
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.792227396322141e-06,
+      "loss": 0.5141,
+      "step": 6190
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.77920912421311e-06,
+      "loss": 0.5685,
+      "step": 6200
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.766192351603119e-06,
+      "loss": 0.5525,
+      "step": 6210
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.753177166895453e-06,
+      "loss": 0.5415,
+      "step": 6220
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.740163658482606e-06,
+      "loss": 0.5443,
+      "step": 6230
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.727151914745694e-06,
+      "loss": 0.55,
+      "step": 6240
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.7141420240538445e-06,
+      "loss": 0.5038,
+      "step": 6250
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.701134074763599e-06,
+      "loss": 0.6429,
+      "step": 6260
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.68812815521832e-06,
+      "loss": 0.5406,
+      "step": 6270
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.675124353747574e-06,
+      "loss": 0.5808,
+      "step": 6280
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.662122758666559e-06,
+      "loss": 0.5551,
+      "step": 6290
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.6491234582754715e-06,
+      "loss": 0.5557,
+      "step": 6300
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.636126540858934e-06,
+      "loss": 0.5276,
+      "step": 6310
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.623132094685383e-06,
+      "loss": 0.5635,
+      "step": 6320
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.610140208006469e-06,
+      "loss": 0.5572,
+      "step": 6330
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.597150969056462e-06,
+      "loss": 0.6144,
+      "step": 6340
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.584164466051648e-06,
+      "loss": 0.5772,
+      "step": 6350
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.571180787189734e-06,
+      "loss": 0.5744,
+      "step": 6360
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.558200020649248e-06,
+      "loss": 0.5044,
+      "step": 6370
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.545222254588936e-06,
+      "loss": 0.5337,
+      "step": 6380
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.532247577147166e-06,
+      "loss": 0.5405,
+      "step": 6390
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.519276076441331e-06,
+      "loss": 0.5581,
+      "step": 6400
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.506307840567253e-06,
+      "loss": 0.5216,
+      "step": 6410
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.4933429575985725e-06,
+      "loss": 0.5389,
+      "step": 6420
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.480381515586165e-06,
+      "loss": 0.5585,
+      "step": 6430
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.467423602557535e-06,
+      "loss": 0.5817,
+      "step": 6440
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.454469306516224e-06,
+      "loss": 0.5544,
+      "step": 6450
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.4415187154412e-06,
+      "loss": 0.6024,
+      "step": 6460
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.428571917286275e-06,
+      "loss": 0.5219,
+      "step": 6470
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.4156289999795025e-06,
+      "loss": 0.4917,
+      "step": 6480
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.4026900514225735e-06,
+      "loss": 0.5897,
+      "step": 6490
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.389755159490231e-06,
+      "loss": 0.5491,
+      "step": 6500
+    },
+    {
+      "epoch": 2.7,
+      "eval_accuracy": 0.6683608640406608,
+      "eval_loss": 0.6455017924308777,
+      "eval_runtime": 50.9772,
+      "eval_samples_per_second": 30.877,
+      "eval_steps_per_second": 3.864,
+      "step": 6500
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.376824412029662e-06,
+      "loss": 0.5678,
+      "step": 6510
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.3638978968599145e-06,
+      "loss": 0.5464,
+      "step": 6520
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.350975701771284e-06,
+      "loss": 0.5651,
+      "step": 6530
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.3380579145247314e-06,
+      "loss": 0.5724,
+      "step": 6540
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.325144622851282e-06,
+      "loss": 0.5364,
+      "step": 6550
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.312235914451426e-06,
+      "loss": 0.5679,
+      "step": 6560
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.29933187699453e-06,
+      "loss": 0.5317,
+      "step": 6570
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.286432598118231e-06,
+      "loss": 0.5459,
+      "step": 6580
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.273538165427861e-06,
+      "loss": 0.6025,
+      "step": 6590
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.260648666495828e-06,
+      "loss": 0.5508,
+      "step": 6600
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.247764188861037e-06,
+      "loss": 0.5373,
+      "step": 6610
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.234884820028288e-06,
+      "loss": 0.582,
+      "step": 6620
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.222010647467688e-06,
+      "loss": 0.5262,
+      "step": 6630
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.2091417586140524e-06,
+      "loss": 0.5225,
+      "step": 6640
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.196278240866311e-06,
+      "loss": 0.5623,
+      "step": 6650
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.183420181586917e-06,
+      "loss": 0.5469,
+      "step": 6660
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.170567668101252e-06,
+      "loss": 0.5648,
+      "step": 6670
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.157720787697037e-06,
+      "loss": 0.5042,
+      "step": 6680
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.1448796276237295e-06,
+      "loss": 0.5646,
+      "step": 6690
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.1320442750919405e-06,
+      "loss": 0.5273,
+      "step": 6700
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.119214817272842e-06,
+      "loss": 0.4953,
+      "step": 6710
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.106391341297566e-06,
+      "loss": 0.5193,
+      "step": 6720
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.093573934256624e-06,
+      "loss": 0.5785,
+      "step": 6730
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.080762683199308e-06,
+      "loss": 0.5814,
+      "step": 6740
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.067957675133104e-06,
+      "loss": 0.5883,
+      "step": 6750
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.055158997023097e-06,
+      "loss": 0.5629,
+      "step": 6760
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.042366735791379e-06,
+      "loss": 0.5605,
+      "step": 6770
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.02958097831647e-06,
+      "loss": 0.549,
+      "step": 6780
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.016801811432711e-06,
+      "loss": 0.5818,
+      "step": 6790
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.00402932192969e-06,
+      "loss": 0.5889,
+      "step": 6800
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.991263596551638e-06,
+      "loss": 0.5623,
+      "step": 6810
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.9785047219968584e-06,
+      "loss": 0.5674,
+      "step": 6820
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.9657527849171165e-06,
+      "loss": 0.5435,
+      "step": 6830
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.9530078719170686e-06,
+      "loss": 0.5567,
+      "step": 6840
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.940270069553665e-06,
+      "loss": 0.5966,
+      "step": 6850
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.9275394643355615e-06,
+      "loss": 0.5412,
+      "step": 6860
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.914816142722541e-06,
+      "loss": 0.5377,
+      "step": 6870
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.902100191124911e-06,
+      "loss": 0.4923,
+      "step": 6880
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.889391695902933e-06,
+      "loss": 0.5096,
+      "step": 6890
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.876690743366227e-06,
+      "loss": 0.5274,
+      "step": 6900
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.863997419773186e-06,
+      "loss": 0.5767,
+      "step": 6910
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.85131181133039e-06,
+      "loss": 0.5602,
+      "step": 6920
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.8386340041920225e-06,
+      "loss": 0.5397,
+      "step": 6930
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.8259640844592876e-06,
+      "loss": 0.5581,
+      "step": 6940
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.8133021381798157e-06,
+      "loss": 0.5254,
+      "step": 6950
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.80064825134709e-06,
+      "loss": 0.5014,
+      "step": 6960
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.7880025098998595e-06,
+      "loss": 0.5981,
+      "step": 6970
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.7753649997215514e-06,
+      "loss": 0.5681,
+      "step": 6980
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.762735806639688e-06,
+      "loss": 0.5367,
+      "step": 6990
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.7501150164253106e-06,
+      "loss": 0.5303,
+      "step": 7000
+    },
+    {
+      "epoch": 2.9,
+      "eval_accuracy": 0.670266836086404,
+      "eval_loss": 0.6339167952537537,
+      "eval_runtime": 51.0066,
+      "eval_samples_per_second": 30.859,
+      "eval_steps_per_second": 3.862,
+      "step": 7000
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.7375027147923914e-06,
+      "loss": 0.5189,
+      "step": 7010
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.7248989873972495e-06,
+      "loss": 0.5633,
+      "step": 7020
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.7123039198379764e-06,
+      "loss": 0.5544,
+      "step": 7030
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6997175976538456e-06,
+      "loss": 0.5455,
+      "step": 7040
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6871401063247438e-06,
+      "loss": 0.567,
+      "step": 7050
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.674571531270575e-06,
+      "loss": 0.5774,
+      "step": 7060
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.6620119578506925e-06,
+      "loss": 0.5811,
+      "step": 7070
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.6494614713633158e-06,
+      "loss": 0.5332,
+      "step": 7080
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.6369201570449474e-06,
+      "loss": 0.5782,
+      "step": 7090
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.624388100069801e-06,
+      "loss": 0.547,
+      "step": 7100
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.6118653855492126e-06,
+      "loss": 0.5649,
+      "step": 7110
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.5993520985310802e-06,
+      "loss": 0.5446,
+      "step": 7120
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.586848323999264e-06,
+      "loss": 0.5501,
+      "step": 7130
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.5743541468730258e-06,
+      "loss": 0.5705,
+      "step": 7140
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.5618696520064465e-06,
+      "loss": 0.5207,
+      "step": 7150
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.5493949241878466e-06,
+      "loss": 0.5419,
+      "step": 7160
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.536930048139219e-06,
+      "loss": 0.5531,
+      "step": 7170
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.524475108515643e-06,
+      "loss": 0.5803,
+      "step": 7180
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.5120301899047177e-06,
+      "loss": 0.492,
+      "step": 7190
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.499595376825986e-06,
+      "loss": 0.5564,
+      "step": 7200
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.487170753730358e-06,
+      "loss": 0.544,
+      "step": 7210
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.4747564049995353e-06,
+      "loss": 0.553,
+      "step": 7220
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.4623524149454455e-06,
+      "loss": 0.5628,
+      "step": 7230
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.449958867809666e-06,
+      "loss": 0.5304,
+      "step": 7240
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.437575847762845e-06,
+      "loss": 0.5263,
+      "step": 7250
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.4252034389041423e-06,
+      "loss": 0.5057,
+      "step": 7260
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.412841725260649e-06,
+      "loss": 0.5299,
+      "step": 7270
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.4004907907868247e-06,
+      "loss": 0.5101,
+      "step": 7280
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.3881507193639125e-06,
+      "loss": 0.5463,
+      "step": 7290
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 3.375821594799388e-06,
+      "loss": 0.5055,
+      "step": 7300
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 3.3635035008263785e-06,
+      "loss": 0.5391,
+      "step": 7310
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 3.351196521103095e-06,
+      "loss": 0.5352,
+      "step": 7320
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 3.3389007392122687e-06,
+      "loss": 0.539,
+      "step": 7330
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 3.3266162386605805e-06,
+      "loss": 0.5358,
+      "step": 7340
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 3.314343102878096e-06,
+      "loss": 0.5116,
+      "step": 7350
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 3.3020814152176923e-06,
+      "loss": 0.5075,
+      "step": 7360
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 3.2898312589545e-06,
+      "loss": 0.4827,
+      "step": 7370
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 3.277592717285336e-06,
+      "loss": 0.527,
+      "step": 7380
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 3.2653658733281307e-06,
+      "loss": 0.4855,
+      "step": 7390
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 3.253150810121377e-06,
+      "loss": 0.5128,
+      "step": 7400
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 3.2409476106235504e-06,
+      "loss": 0.5041,
+      "step": 7410
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 3.228756357712565e-06,
+      "loss": 0.4662,
+      "step": 7420
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 3.2165771341851916e-06,
+      "loss": 0.5226,
+      "step": 7430
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 3.2044100227565045e-06,
+      "loss": 0.5439,
+      "step": 7440
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 3.192255106059323e-06,
+      "loss": 0.5777,
+      "step": 7450
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 3.1801124666436404e-06,
+      "loss": 0.5248,
+      "step": 7460
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 3.167982186976074e-06,
+      "loss": 0.5777,
+      "step": 7470
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 3.155864349439296e-06,
+      "loss": 0.5385,
+      "step": 7480
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 3.1437590363314786e-06,
+      "loss": 0.4889,
+      "step": 7490
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 3.1316663298657386e-06,
+      "loss": 0.497,
+      "step": 7500
+    },
+    {
+      "epoch": 3.11,
+      "eval_accuracy": 0.6607369758576874,
+      "eval_loss": 0.6541444063186646,
+      "eval_runtime": 51.0133,
+      "eval_samples_per_second": 30.855,
+      "eval_steps_per_second": 3.862,
+      "step": 7500
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 3.1195863121695714e-06,
+      "loss": 0.5098,
+      "step": 7510
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 3.1075190652842955e-06,
+      "loss": 0.4958,
+      "step": 7520
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 3.0954646711645e-06,
+      "loss": 0.5011,
+      "step": 7530
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 3.0834232116774832e-06,
+      "loss": 0.512,
+      "step": 7540
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 3.0713947686026944e-06,
+      "loss": 0.496,
+      "step": 7550
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 3.059379423631189e-06,
+      "loss": 0.5538,
+      "step": 7560
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 3.0473772583650595e-06,
+      "loss": 0.5268,
+      "step": 7570
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 3.035388354316897e-06,
+      "loss": 0.442,
+      "step": 7580
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 3.0234127929092204e-06,
+      "loss": 0.5327,
+      "step": 7590
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 3.0114506554739375e-06,
+      "loss": 0.5309,
+      "step": 7600
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.999502023251787e-06,
+      "loss": 0.5044,
+      "step": 7610
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.9875669773917836e-06,
+      "loss": 0.5235,
+      "step": 7620
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.9756455989506737e-06,
+      "loss": 0.5446,
+      "step": 7630
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.96373796889238e-06,
+      "loss": 0.5191,
+      "step": 7640
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.9518441680874565e-06,
+      "loss": 0.4934,
+      "step": 7650
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.9399642773125316e-06,
+      "loss": 0.5224,
+      "step": 7660
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.928098377249764e-06,
+      "loss": 0.5514,
+      "step": 7670
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.9162465484862986e-06,
+      "loss": 0.5096,
+      "step": 7680
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.904408871513712e-06,
+      "loss": 0.5834,
+      "step": 7690
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.89258542672747e-06,
+      "loss": 0.5257,
+      "step": 7700
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.8807762944263784e-06,
+      "loss": 0.4836,
+      "step": 7710
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.8689815548120435e-06,
+      "loss": 0.4929,
+      "step": 7720
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.857201287988325e-06,
+      "loss": 0.5144,
+      "step": 7730
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.845435573960784e-06,
+      "loss": 0.5394,
+      "step": 7740
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.833684492636152e-06,
+      "loss": 0.5519,
+      "step": 7750
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.8219481238217823e-06,
+      "loss": 0.5055,
+      "step": 7760
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.8102265472251063e-06,
+      "loss": 0.5947,
+      "step": 7770
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.7985198424530946e-06,
+      "loss": 0.469,
+      "step": 7780
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.7868280890117126e-06,
+      "loss": 0.4897,
+      "step": 7790
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.7751513663053966e-06,
+      "loss": 0.507,
+      "step": 7800
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.763489753636488e-06,
+      "loss": 0.5186,
+      "step": 7810
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.751843330204711e-06,
+      "loss": 0.5235,
+      "step": 7820
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.74021217510664e-06,
+      "loss": 0.5618,
+      "step": 7830
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.728596367335147e-06,
+      "loss": 0.5595,
+      "step": 7840
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.716995985778874e-06,
+      "loss": 0.4756,
+      "step": 7850
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.7054111092216972e-06,
+      "loss": 0.5412,
+      "step": 7860
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.6938418163421907e-06,
+      "loss": 0.5077,
+      "step": 7870
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.6822881857130954e-06,
+      "loss": 0.4589,
+      "step": 7880
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.670750295800777e-06,
+      "loss": 0.5212,
+      "step": 7890
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.659228224964698e-06,
+      "loss": 0.5068,
+      "step": 7900
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.647722051456892e-06,
+      "loss": 0.5101,
+      "step": 7910
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.6362318534214206e-06,
+      "loss": 0.4968,
+      "step": 7920
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.624757708893847e-06,
+      "loss": 0.5245,
+      "step": 7930
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.6132996958007116e-06,
+      "loss": 0.5218,
+      "step": 7940
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.6018578919590004e-06,
+      "loss": 0.5017,
+      "step": 7950
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.5904323750756066e-06,
+      "loss": 0.5221,
+      "step": 7960
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.5790232227468137e-06,
+      "loss": 0.5148,
+      "step": 7970
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.5676305124577704e-06,
+      "loss": 0.5125,
+      "step": 7980
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.5562543215819533e-06,
+      "loss": 0.5148,
+      "step": 7990
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.5448947273806477e-06,
+      "loss": 0.5041,
+      "step": 8000
+    },
+    {
+      "epoch": 3.32,
+      "eval_accuracy": 0.6759847522236341,
+      "eval_loss": 0.665330171585083,
+      "eval_runtime": 51.0245,
+      "eval_samples_per_second": 30.848,
+      "eval_steps_per_second": 3.861,
+      "step": 8000
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.533551807002429e-06,
+      "loss": 0.5905,
+      "step": 8010
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.522225637482625e-06,
+      "loss": 0.5326,
+      "step": 8020
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.510916295742806e-06,
+      "loss": 0.5496,
+      "step": 8030
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.499623858590251e-06,
+      "loss": 0.5171,
+      "step": 8040
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.4883484027174394e-06,
+      "loss": 0.544,
+      "step": 8050
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.4770900047015145e-06,
+      "loss": 0.5228,
+      "step": 8060
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.4658487410037742e-06,
+      "loss": 0.5229,
+      "step": 8070
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.4546246879691514e-06,
+      "loss": 0.5266,
+      "step": 8080
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.443417921825691e-06,
+      "loss": 0.5429,
+      "step": 8090
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.4322285186840304e-06,
+      "loss": 0.5335,
+      "step": 8100
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.421056554536893e-06,
+      "loss": 0.5003,
+      "step": 8110
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.4099021052585635e-06,
+      "loss": 0.5313,
+      "step": 8120
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.3987652466043726e-06,
+      "loss": 0.508,
+      "step": 8130
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.3876460542101813e-06,
+      "loss": 0.4822,
+      "step": 8140
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.3765446035918805e-06,
+      "loss": 0.4845,
+      "step": 8150
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.365460970144859e-06,
+      "loss": 0.4896,
+      "step": 8160
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.3543952291435007e-06,
+      "loss": 0.5868,
+      "step": 8170
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.343347455740678e-06,
+      "loss": 0.4883,
+      "step": 8180
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.3323177249672375e-06,
+      "loss": 0.4871,
+      "step": 8190
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.3213061117314828e-06,
+      "loss": 0.5097,
+      "step": 8200
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.3103126908186745e-06,
+      "loss": 0.5473,
+      "step": 8210
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.299337536890526e-06,
+      "loss": 0.4838,
+      "step": 8220
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.2883807244846828e-06,
+      "loss": 0.4678,
+      "step": 8230
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.2774423280142266e-06,
+      "loss": 0.5296,
+      "step": 8240
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.266522421767171e-06,
+      "loss": 0.5395,
+      "step": 8250
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.255621079905952e-06,
+      "loss": 0.5025,
+      "step": 8260
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.2447383764669224e-06,
+      "loss": 0.4966,
+      "step": 8270
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.233874385359853e-06,
+      "loss": 0.5259,
+      "step": 8280
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.223029180367436e-06,
+      "loss": 0.491,
+      "step": 8290
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.2122028351447706e-06,
+      "loss": 0.4988,
+      "step": 8300
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.201395423218872e-06,
+      "loss": 0.4868,
+      "step": 8310
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.1906070179881717e-06,
+      "loss": 0.5045,
+      "step": 8320
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.1798376927220187e-06,
+      "loss": 0.4979,
+      "step": 8330
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.169087520560178e-06,
+      "loss": 0.5268,
+      "step": 8340
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.158356574512334e-06,
+      "loss": 0.5226,
+      "step": 8350
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.147644927457605e-06,
+      "loss": 0.5045,
+      "step": 8360
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.1369526521440327e-06,
+      "loss": 0.5452,
+      "step": 8370
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.126279821188099e-06,
+      "loss": 0.5064,
+      "step": 8380
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.1156265070742283e-06,
+      "loss": 0.4937,
+      "step": 8390
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.1049927821542985e-06,
+      "loss": 0.5872,
+      "step": 8400
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.09437871864715e-06,
+      "loss": 0.5033,
+      "step": 8410
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.0837843886380843e-06,
+      "loss": 0.5027,
+      "step": 8420
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.073209864078393e-06,
+      "loss": 0.4867,
+      "step": 8430
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.0626552167848535e-06,
+      "loss": 0.5243,
+      "step": 8440
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.0521205184392475e-06,
+      "loss": 0.4919,
+      "step": 8450
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.0416058405878714e-06,
+      "loss": 0.5175,
+      "step": 8460
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.031111254641059e-06,
+      "loss": 0.4879,
+      "step": 8470
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.020636831872687e-06,
+      "loss": 0.5268,
+      "step": 8480
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.010182643419693e-06,
+      "loss": 0.4955,
+      "step": 8490
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.999748760281592e-06,
+      "loss": 0.4978,
+      "step": 8500
+    },
+    {
+      "epoch": 3.53,
+      "eval_accuracy": 0.6696315120711563,
+      "eval_loss": 0.6626632809638977,
+      "eval_runtime": 51.0233,
+      "eval_samples_per_second": 30.849,
+      "eval_steps_per_second": 3.861,
+      "step": 8500
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9893352533200006e-06,
+      "loss": 0.4821,
+      "step": 8510
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.9789421932581467e-06,
+      "loss": 0.5165,
+      "step": 8520
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9685696506803927e-06,
+      "loss": 0.5464,
+      "step": 8530
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.9582176960317616e-06,
+      "loss": 0.4815,
+      "step": 8540
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.947886399617448e-06,
+      "loss": 0.4829,
+      "step": 8550
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.937575831602352e-06,
+      "loss": 0.589,
+      "step": 8560
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.92728606201059e-06,
+      "loss": 0.527,
+      "step": 8570
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.9170171607250344e-06,
+      "loss": 0.5249,
+      "step": 8580
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.906769197486824e-06,
+      "loss": 0.4985,
+      "step": 8590
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.8965422418948975e-06,
+      "loss": 0.4909,
+      "step": 8600
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.8863363634055265e-06,
+      "loss": 0.4602,
+      "step": 8610
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.8761516313318289e-06,
+      "loss": 0.5304,
+      "step": 8620
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.8659881148433156e-06,
+      "loss": 0.5646,
+      "step": 8630
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.8558458829654046e-06,
+      "loss": 0.4989,
+      "step": 8640
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.8457250045789665e-06,
+      "loss": 0.5796,
+      "step": 8650
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.8356255484198437e-06,
+      "loss": 0.496,
+      "step": 8660
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.825547583078392e-06,
+      "loss": 0.5491,
+      "step": 8670
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.8154911769990152e-06,
+      "loss": 0.5242,
+      "step": 8680
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.8054563984796946e-06,
+      "loss": 0.524,
+      "step": 8690
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.7954433156715274e-06,
+      "loss": 0.5517,
+      "step": 8700
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.7854519965782662e-06,
+      "loss": 0.49,
+      "step": 8710
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.7754825090558586e-06,
+      "loss": 0.5027,
+      "step": 8720
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.7655349208119782e-06,
+      "loss": 0.5032,
+      "step": 8730
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.7556092994055695e-06,
+      "loss": 0.5422,
+      "step": 8740
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.7457057122463956e-06,
+      "loss": 0.532,
+      "step": 8750
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.7358242265945686e-06,
+      "loss": 0.494,
+      "step": 8760
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.7259649095600984e-06,
+      "loss": 0.5087,
+      "step": 8770
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.7161278281024407e-06,
+      "loss": 0.4746,
+      "step": 8780
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.7063130490300384e-06,
+      "loss": 0.4745,
+      "step": 8790
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.6965206389998634e-06,
+      "loss": 0.5424,
+      "step": 8800
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.6867506645169707e-06,
+      "loss": 0.5325,
+      "step": 8810
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.6770031919340478e-06,
+      "loss": 0.5303,
+      "step": 8820
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.6672782874509563e-06,
+      "loss": 0.4971,
+      "step": 8830
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.6575760171142874e-06,
+      "loss": 0.5109,
+      "step": 8840
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.6478964468169145e-06,
+      "loss": 0.5009,
+      "step": 8850
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.638239642297545e-06,
+      "loss": 0.4573,
+      "step": 8860
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.6286056691402702e-06,
+      "loss": 0.5325,
+      "step": 8870
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.61899459277412e-06,
+      "loss": 0.5533,
+      "step": 8880
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.6094064784726287e-06,
+      "loss": 0.5485,
+      "step": 8890
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.5998413913533777e-06,
+      "loss": 0.5363,
+      "step": 8900
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.5902993963775605e-06,
+      "loss": 0.4425,
+      "step": 8910
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.5807805583495395e-06,
+      "loss": 0.5034,
+      "step": 8920
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.571284941916411e-06,
+      "loss": 0.462,
+      "step": 8930
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.5618126115675613e-06,
+      "loss": 0.5476,
+      "step": 8940
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.5523636316342238e-06,
+      "loss": 0.5058,
+      "step": 8950
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.5429380662890564e-06,
+      "loss": 0.5327,
+      "step": 8960
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.53353597954569e-06,
+      "loss": 0.5053,
+      "step": 8970
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.5241574352583038e-06,
+      "loss": 0.5187,
+      "step": 8980
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.5148024971211866e-06,
+      "loss": 0.5093,
+      "step": 8990
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.50547122866831e-06,
+      "loss": 0.5272,
+      "step": 9000
+    },
+    {
+      "epoch": 3.73,
+      "eval_accuracy": 0.667725540025413,
+      "eval_loss": 0.6683589816093445,
+      "eval_runtime": 51.0188,
+      "eval_samples_per_second": 30.851,
+      "eval_steps_per_second": 3.861,
+      "step": 9000
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.4961636932728912e-06,
+      "loss": 0.4706,
+      "step": 9010
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.4868799541469643e-06,
+      "loss": 0.4773,
+      "step": 9020
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.4776200743409486e-06,
+      "loss": 0.508,
+      "step": 9030
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.46838411674323e-06,
+      "loss": 0.5204,
+      "step": 9040
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.4591721440797207e-06,
+      "loss": 0.4865,
+      "step": 9050
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.4499842189134406e-06,
+      "loss": 0.4963,
+      "step": 9060
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.440820403644095e-06,
+      "loss": 0.5922,
+      "step": 9070
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.4316807605076405e-06,
+      "loss": 0.5172,
+      "step": 9080
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.4225653515758775e-06,
+      "loss": 0.5172,
+      "step": 9090
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.4134742387560108e-06,
+      "loss": 0.5234,
+      "step": 9100
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.4044074837902462e-06,
+      "loss": 0.4496,
+      "step": 9110
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.3953651482553586e-06,
+      "loss": 0.4917,
+      "step": 9120
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.386347293562279e-06,
+      "loss": 0.4721,
+      "step": 9130
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.3773539809556802e-06,
+      "loss": 0.5205,
+      "step": 9140
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.3683852715135533e-06,
+      "loss": 0.5332,
+      "step": 9150
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.3594412261468015e-06,
+      "loss": 0.5106,
+      "step": 9160
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.3505219055988178e-06,
+      "loss": 0.5283,
+      "step": 9170
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.3416273704450821e-06,
+      "loss": 0.5479,
+      "step": 9180
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.33275768109274e-06,
+      "loss": 0.5457,
+      "step": 9190
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.3239128977801968e-06,
+      "loss": 0.4529,
+      "step": 9200
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.3150930805767142e-06,
+      "loss": 0.5522,
+      "step": 9210
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.3062982893819915e-06,
+      "loss": 0.4901,
+      "step": 9220
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.297528583925764e-06,
+      "loss": 0.5215,
+      "step": 9230
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.288784023767401e-06,
+      "loss": 0.5607,
+      "step": 9240
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.280064668295497e-06,
+      "loss": 0.4933,
+      "step": 9250
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.2713705767274675e-06,
+      "loss": 0.4887,
+      "step": 9260
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.262701808109147e-06,
+      "loss": 0.4977,
+      "step": 9270
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.254058421314394e-06,
+      "loss": 0.4802,
+      "step": 9280
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.245440475044684e-06,
+      "loss": 0.554,
+      "step": 9290
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.236848027828711e-06,
+      "loss": 0.4416,
+      "step": 9300
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.2282811380219967e-06,
+      "loss": 0.4841,
+      "step": 9310
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.2197398638064895e-06,
+      "loss": 0.482,
+      "step": 9320
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.2112242631901671e-06,
+      "loss": 0.562,
+      "step": 9330
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.2027343940066456e-06,
+      "loss": 0.551,
+      "step": 9340
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.1942703139147894e-06,
+      "loss": 0.4475,
+      "step": 9350
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.185832080398313e-06,
+      "loss": 0.531,
+      "step": 9360
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.1774197507653944e-06,
+      "loss": 0.515,
+      "step": 9370
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.1690333821482885e-06,
+      "loss": 0.5638,
+      "step": 9380
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.1606730315029352e-06,
+      "loss": 0.5021,
+      "step": 9390
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.1523387556085718e-06,
+      "loss": 0.4604,
+      "step": 9400
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.1440306110673494e-06,
+      "loss": 0.5132,
+      "step": 9410
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.1357486543039515e-06,
+      "loss": 0.517,
+      "step": 9420
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.1274929415652036e-06,
+      "loss": 0.5594,
+      "step": 9430
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.1192635289196962e-06,
+      "loss": 0.5123,
+      "step": 9440
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.1110604722574053e-06,
+      "loss": 0.5055,
+      "step": 9450
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.1028838272893066e-06,
+      "loss": 0.5069,
+      "step": 9460
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.0947336495470063e-06,
+      "loss": 0.495,
+      "step": 9470
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.0866099943823527e-06,
+      "loss": 0.5369,
+      "step": 9480
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.0785129169670717e-06,
+      "loss": 0.547,
+      "step": 9490
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.0704424722923823e-06,
+      "loss": 0.5487,
+      "step": 9500
+    },
+    {
+      "epoch": 3.94,
+      "eval_accuracy": 0.6759847522236341,
+      "eval_loss": 0.6592991352081299,
+      "eval_runtime": 50.9849,
+      "eval_samples_per_second": 30.872,
+      "eval_steps_per_second": 3.864,
+      "step": 9500
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.0623987151686304e-06,
+      "loss": 0.5652,
+      "step": 9510
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.0543817002249107e-06,
+      "loss": 0.4853,
+      "step": 9520
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.0463914819087024e-06,
+      "loss": 0.4757,
+      "step": 9530
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.0384281144854947e-06,
+      "loss": 0.4777,
+      "step": 9540
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.0304916520384156e-06,
+      "loss": 0.4585,
+      "step": 9550
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.0225821484678743e-06,
+      "loss": 0.4736,
+      "step": 9560
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.0146996574911844e-06,
+      "loss": 0.5137,
+      "step": 9570
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.0068442326422045e-06,
+      "loss": 0.5496,
+      "step": 9580
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 9.990159272709738e-07,
+      "loss": 0.4905,
+      "step": 9590
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 9.91214794543353e-07,
+      "loss": 0.5584,
+      "step": 9600
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 9.834408874406549e-07,
+      "loss": 0.5232,
+      "step": 9610
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 9.756942587592943e-07,
+      "loss": 0.4861,
+      "step": 9620
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 9.679749611104218e-07,
+      "loss": 0.5231,
+      "step": 9630
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 9.602830469195718e-07,
+      "loss": 0.5002,
+      "step": 9640
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 9.526185684263028e-07,
+      "loss": 0.5272,
+      "step": 9650
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 9.449815776838434e-07,
+      "loss": 0.5239,
+      "step": 9660
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 9.373721265587421e-07,
+      "loss": 0.5013,
+      "step": 9670
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 9.297902667305087e-07,
+      "loss": 0.4697,
+      "step": 9680
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 9.222360496912714e-07,
+      "loss": 0.5173,
+      "step": 9690
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 9.147095267454176e-07,
+      "loss": 0.4486,
+      "step": 9700
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 9.072107490092557e-07,
+      "loss": 0.4937,
+      "step": 9710
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 8.997397674106589e-07,
+      "loss": 0.4948,
+      "step": 9720
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 8.922966326887233e-07,
+      "loss": 0.4646,
+      "step": 9730
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 8.848813953934271e-07,
+      "loss": 0.5394,
+      "step": 9740
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 8.774941058852798e-07,
+      "loss": 0.4968,
+      "step": 9750
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 8.701348143349825e-07,
+      "loss": 0.4447,
+      "step": 9760
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 8.62803570723093e-07,
+      "loss": 0.4508,
+      "step": 9770
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 8.555004248396809e-07,
+      "loss": 0.4928,
+      "step": 9780
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 8.482254262839867e-07,
+      "loss": 0.4924,
+      "step": 9790
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 8.409786244640899e-07,
+      "loss": 0.5271,
+      "step": 9800
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 8.337600685965764e-07,
+      "loss": 0.4866,
+      "step": 9810
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 8.265698077061946e-07,
+      "loss": 0.469,
+      "step": 9820
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 8.194078906255293e-07,
+      "loss": 0.4378,
+      "step": 9830
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 8.122743659946696e-07,
+      "loss": 0.5039,
+      "step": 9840
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 8.051692822608798e-07,
+      "loss": 0.4981,
+      "step": 9850
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 7.98092687678263e-07,
+      "loss": 0.506,
+      "step": 9860
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 7.910446303074404e-07,
+      "loss": 0.5178,
+      "step": 9870
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 7.840251580152264e-07,
+      "loss": 0.5176,
+      "step": 9880
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 7.770343184742957e-07,
+      "loss": 0.4865,
+      "step": 9890
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 7.700721591628646e-07,
+      "loss": 0.4668,
+      "step": 9900
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 7.631387273643698e-07,
+      "loss": 0.5073,
+      "step": 9910
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 7.562340701671467e-07,
+      "loss": 0.5112,
+      "step": 9920
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 7.493582344641037e-07,
+      "loss": 0.4861,
+      "step": 9930
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 7.425112669524104e-07,
+      "loss": 0.4667,
+      "step": 9940
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 7.356932141331807e-07,
+      "loss": 0.519,
+      "step": 9950
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 7.289041223111515e-07,
+      "loss": 0.5019,
+      "step": 9960
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 7.221440375943711e-07,
+      "loss": 0.4876,
+      "step": 9970
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 7.15413005893889e-07,
+      "loss": 0.47,
+      "step": 9980
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 7.08711072923442e-07,
+      "loss": 0.4966,
+      "step": 9990
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 7.020382841991391e-07,
+      "loss": 0.4998,
+      "step": 10000
+    },
+    {
+      "epoch": 4.15,
+      "eval_accuracy": 0.6747141041931385,
+      "eval_loss": 0.6738409996032715,
+      "eval_runtime": 50.8651,
+      "eval_samples_per_second": 30.945,
+      "eval_steps_per_second": 3.873,
+      "step": 10000
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 6.953946850391585e-07,
+      "loss": 0.4908,
+      "step": 10010
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 6.887803205634397e-07,
+      "loss": 0.438,
+      "step": 10020
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 6.821952356933719e-07,
+      "loss": 0.504,
+      "step": 10030
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 6.756394751514944e-07,
+      "loss": 0.5082,
+      "step": 10040
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 6.691130834611887e-07,
+      "loss": 0.4924,
+      "step": 10050
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 6.626161049463792e-07,
+      "loss": 0.517,
+      "step": 10060
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 6.561485837312332e-07,
+      "loss": 0.4403,
+      "step": 10070
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 6.497105637398543e-07,
+      "loss": 0.4661,
+      "step": 10080
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 6.433020886959934e-07,
+      "loss": 0.5099,
+      "step": 10090
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 6.369232021227445e-07,
+      "loss": 0.459,
+      "step": 10100
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 6.305739473422517e-07,
+      "loss": 0.4583,
+      "step": 10110
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 6.242543674754148e-07,
+      "loss": 0.5359,
+      "step": 10120
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 6.179645054416006e-07,
+      "loss": 0.4623,
+      "step": 10130
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 6.11704403958342e-07,
+      "loss": 0.5055,
+      "step": 10140
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 6.054741055410584e-07,
+      "loss": 0.5315,
+      "step": 10150
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.992736525027576e-07,
+      "loss": 0.5102,
+      "step": 10160
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.931030869537563e-07,
+      "loss": 0.5163,
+      "step": 10170
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.869624508013888e-07,
+      "loss": 0.4868,
+      "step": 10180
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.808517857497237e-07,
+      "loss": 0.5501,
+      "step": 10190
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.747711332992834e-07,
+      "loss": 0.463,
+      "step": 10200
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.687205347467567e-07,
+      "loss": 0.4987,
+      "step": 10210
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.627000311847264e-07,
+      "loss": 0.4603,
+      "step": 10220
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.567096635013797e-07,
+      "loss": 0.5538,
+      "step": 10230
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.507494723802414e-07,
+      "loss": 0.4539,
+      "step": 10240
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.4481949829989e-07,
+      "loss": 0.453,
+      "step": 10250
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.389197815336844e-07,
+      "loss": 0.502,
+      "step": 10260
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.330503621494943e-07,
+      "loss": 0.4637,
+      "step": 10270
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.272112800094215e-07,
+      "loss": 0.478,
+      "step": 10280
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.214025747695345e-07,
+      "loss": 0.5017,
+      "step": 10290
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.156242858795973e-07,
+      "loss": 0.4295,
+      "step": 10300
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.098764525828031e-07,
+      "loss": 0.4873,
+      "step": 10310
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.041591139155039e-07,
+      "loss": 0.5405,
+      "step": 10320
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 4.98472308706947e-07,
+      "loss": 0.5016,
+      "step": 10330
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 4.928160755790157e-07,
+      "loss": 0.4776,
+      "step": 10340
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 4.871904529459598e-07,
+      "loss": 0.4979,
+      "step": 10350
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 4.815954790141392e-07,
+      "loss": 0.4865,
+      "step": 10360
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 4.760311917817645e-07,
+      "loss": 0.4634,
+      "step": 10370
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 4.7049762903863816e-07,
+      "loss": 0.4966,
+      "step": 10380
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 4.649948283658956e-07,
+      "loss": 0.4894,
+      "step": 10390
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 4.5952282713575213e-07,
+      "loss": 0.5108,
+      "step": 10400
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 4.540816625112515e-07,
+      "loss": 0.4657,
+      "step": 10410
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 4.486713714460084e-07,
+      "loss": 0.4639,
+      "step": 10420
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 4.432919906839594e-07,
+      "loss": 0.4609,
+      "step": 10430
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 4.379435567591178e-07,
+      "loss": 0.5085,
+      "step": 10440
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 4.3262610599531907e-07,
+      "loss": 0.443,
+      "step": 10450
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 4.2733967450597714e-07,
+      "loss": 0.4905,
+      "step": 10460
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 4.220842981938378e-07,
+      "loss": 0.4888,
+      "step": 10470
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 4.168600127507394e-07,
+      "loss": 0.5175,
+      "step": 10480
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 4.116668536573626e-07,
+      "loss": 0.5126,
+      "step": 10490
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 4.0650485618299594e-07,
+      "loss": 0.4626,
+      "step": 10500
+    },
+    {
+      "epoch": 4.36,
+      "eval_accuracy": 0.6753494282083863,
+      "eval_loss": 0.6780598163604736,
+      "eval_runtime": 50.8685,
+      "eval_samples_per_second": 30.943,
+      "eval_steps_per_second": 3.873,
+      "step": 10500
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 4.0137405538529516e-07,
+      "loss": 0.4796,
+      "step": 10510
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.962744861100426e-07,
+      "loss": 0.5017,
+      "step": 10520
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.912061829909125e-07,
+      "loss": 0.4647,
+      "step": 10530
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.86169180449234e-07,
+      "loss": 0.4616,
+      "step": 10540
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.8116351269376126e-07,
+      "loss": 0.4694,
+      "step": 10550
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.761892137204365e-07,
+      "loss": 0.4901,
+      "step": 10560
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.7124631731216163e-07,
+      "loss": 0.438,
+      "step": 10570
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.663348570385694e-07,
+      "loss": 0.539,
+      "step": 10580
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.6145486625579206e-07,
+      "loss": 0.4985,
+      "step": 10590
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.566063781062401e-07,
+      "loss": 0.5121,
+      "step": 10600
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.5178942551837123e-07,
+      "loss": 0.482,
+      "step": 10610
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.4700404120647235e-07,
+      "loss": 0.4905,
+      "step": 10620
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.422502576704334e-07,
+      "loss": 0.4636,
+      "step": 10630
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.3752810719552776e-07,
+      "loss": 0.5135,
+      "step": 10640
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.328376218521923e-07,
+      "loss": 0.5483,
+      "step": 10650
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.2817883349581303e-07,
+      "loss": 0.4481,
+      "step": 10660
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.2355177376650503e-07,
+      "loss": 0.4614,
+      "step": 10670
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.1895647408889787e-07,
+      "loss": 0.482,
+      "step": 10680
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.1439296567192414e-07,
+      "loss": 0.4494,
+      "step": 10690
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.098612795086059e-07,
+      "loss": 0.5117,
+      "step": 10700
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.0536144637584417e-07,
+      "loss": 0.4962,
+      "step": 10710
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.008934968342109e-07,
+      "loss": 0.5207,
+      "step": 10720
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 2.964574612277421e-07,
+      "loss": 0.4722,
+      "step": 10730
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 2.9205336968372856e-07,
+      "loss": 0.5085,
+      "step": 10740
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 2.8768125211251497e-07,
+      "loss": 0.4833,
+      "step": 10750
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 2.833411382072937e-07,
+      "loss": 0.5006,
+      "step": 10760
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 2.79033057443906e-07,
+      "loss": 0.4587,
+      "step": 10770
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 2.7475703908063977e-07,
+      "loss": 0.5463,
+      "step": 10780
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 2.705131121580296e-07,
+      "loss": 0.4722,
+      "step": 10790
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 2.6630130549866515e-07,
+      "loss": 0.4491,
+      "step": 10800
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 2.6212164770698824e-07,
+      "loss": 0.4507,
+      "step": 10810
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 2.579741671691033e-07,
+      "loss": 0.4585,
+      "step": 10820
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 2.538588920525836e-07,
+      "loss": 0.5144,
+      "step": 10830
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 2.497758503062797e-07,
+      "loss": 0.478,
+      "step": 10840
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 2.4572506966012875e-07,
+      "loss": 0.5033,
+      "step": 10850
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 2.4210697306145893e-07,
+      "loss": 0.5843,
+      "step": 10860
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 2.3811756411671494e-07,
+      "loss": 0.488,
+      "step": 10870
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 2.3416049544925878e-07,
+      "loss": 0.4682,
+      "step": 10880
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 2.3023579393348316e-07,
+      "loss": 0.4582,
+      "step": 10890
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 2.263434862239583e-07,
+      "loss": 0.5019,
+      "step": 10900
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 2.2248359875525483e-07,
+      "loss": 0.4367,
+      "step": 10910
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 2.1865615774175952e-07,
+      "loss": 0.5049,
+      "step": 10920
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 2.148611891774993e-07,
+      "loss": 0.4658,
+      "step": 10930
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 2.11098718835967e-07,
+      "loss": 0.5211,
+      "step": 10940
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 2.0736877226994189e-07,
+      "loss": 0.551,
+      "step": 10950
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 2.0367137481131895e-07,
+      "loss": 0.4597,
+      "step": 10960
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 2.000065515709365e-07,
+      "loss": 0.4609,
+      "step": 10970
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.9637432743840501e-07,
+      "loss": 0.4884,
+      "step": 10980
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.9277472708193857e-07,
+      "loss": 0.4719,
+      "step": 10990
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.8920777494818743e-07,
+      "loss": 0.5202,
+      "step": 11000
+    },
+    {
+      "epoch": 4.56,
+      "eval_accuracy": 0.6721728081321474,
+      "eval_loss": 0.6763395071029663,
+      "eval_runtime": 50.9097,
+      "eval_samples_per_second": 30.917,
+      "eval_steps_per_second": 3.87,
+      "step": 11000
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.856734952620709e-07,
+      "loss": 0.4223,
+      "step": 11010
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.8217191202661578e-07,
+      "loss": 0.5036,
+      "step": 11020
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.7870304902278924e-07,
+      "loss": 0.4597,
+      "step": 11030
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.7526692980933957e-07,
+      "loss": 0.488,
+      "step": 11040
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.7186357772263574e-07,
+      "loss": 0.5227,
+      "step": 11050
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.6849301587651134e-07,
+      "loss": 0.498,
+      "step": 11060
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.651552671621026e-07,
+      "loss": 0.4683,
+      "step": 11070
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.6185035424769678e-07,
+      "loss": 0.4731,
+      "step": 11080
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.5857829957857674e-07,
+      "loss": 0.4649,
+      "step": 11090
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.5533912537687102e-07,
+      "loss": 0.4852,
+      "step": 11100
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.521328536413974e-07,
+      "loss": 0.4608,
+      "step": 11110
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.4895950614751853e-07,
+      "loss": 0.4661,
+      "step": 11120
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.458191044469931e-07,
+      "loss": 0.5243,
+      "step": 11130
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.4271166986782825e-07,
+      "loss": 0.5064,
+      "step": 11140
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.396372235141341e-07,
+      "loss": 0.5359,
+      "step": 11150
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.3659578626598336e-07,
+      "loss": 0.4972,
+      "step": 11160
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.3358737877926686e-07,
+      "loss": 0.4889,
+      "step": 11170
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.3061202148555386e-07,
+      "loss": 0.5048,
+      "step": 11180
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.2766973459195365e-07,
+      "loss": 0.5625,
+      "step": 11190
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.2476053808097966e-07,
+      "loss": 0.4774,
+      "step": 11200
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.2188445171041008e-07,
+      "loss": 0.5569,
+      "step": 11210
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.1904149501315631e-07,
+      "loss": 0.4817,
+      "step": 11220
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.1623168729713141e-07,
+      "loss": 0.4588,
+      "step": 11230
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.1345504764511738e-07,
+      "loss": 0.5213,
+      "step": 11240
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.1071159491463368e-07,
+      "loss": 0.4525,
+      "step": 11250
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.0800134773781279e-07,
+      "loss": 0.4998,
+      "step": 11260
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.0532432452127095e-07,
+      "loss": 0.5169,
+      "step": 11270
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.0268054344598544e-07,
+      "loss": 0.4641,
+      "step": 11280
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.0007002246716858e-07,
+      "loss": 0.525,
+      "step": 11290
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 9.749277931414835e-08,
+      "loss": 0.4876,
+      "step": 11300
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 9.494883149024525e-08,
+      "loss": 0.4615,
+      "step": 11310
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 9.243819627265671e-08,
+      "loss": 0.5418,
+      "step": 11320
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 8.99608907123356e-08,
+      "loss": 0.4389,
+      "step": 11330
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 8.75169316338792e-08,
+      "loss": 0.5149,
+      "step": 11340
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 8.510633563541093e-08,
+      "loss": 0.4848,
+      "step": 11350
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 8.272911908846881e-08,
+      "loss": 0.4801,
+      "step": 11360
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 8.038529813789664e-08,
+      "loss": 0.4783,
+      "step": 11370
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 7.80748887017302e-08,
+      "loss": 0.4966,
+      "step": 11380
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 7.579790647109286e-08,
+      "loss": 0.4576,
+      "step": 11390
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 7.355436691008633e-08,
+      "loss": 0.489,
+      "step": 11400
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 7.134428525568782e-08,
+      "loss": 0.4672,
+      "step": 11410
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 6.916767651764467e-08,
+      "loss": 0.5108,
+      "step": 11420
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 6.702455547837327e-08,
+      "loss": 0.5137,
+      "step": 11430
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 6.491493669285976e-08,
+      "loss": 0.5045,
+      "step": 11440
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 6.283883448855832e-08,
+      "loss": 0.4732,
+      "step": 11450
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 6.079626296529584e-08,
+      "loss": 0.5089,
+      "step": 11460
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.878723599517799e-08,
+      "loss": 0.4818,
+      "step": 11470
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.6811767222491045e-08,
+      "loss": 0.4396,
+      "step": 11480
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.486987006361133e-08,
+      "loss": 0.4476,
+      "step": 11490
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.2961557706914245e-08,
+      "loss": 0.4623,
+      "step": 11500
+    },
+    {
+      "epoch": 4.77,
+      "eval_accuracy": 0.6728081321473952,
+      "eval_loss": 0.677772581577301,
+      "eval_runtime": 50.9617,
+      "eval_samples_per_second": 30.886,
+      "eval_steps_per_second": 3.866,
+      "step": 11500
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.108684311268375e-08,
+      "loss": 0.4756,
+      "step": 11510
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 4.924573901302632e-08,
+      "loss": 0.5532,
+      "step": 11520
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 4.743825791178103e-08,
+      "loss": 0.4798,
+      "step": 11530
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 4.56644120844385e-08,
+      "loss": 0.4843,
+      "step": 11540
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 4.392421357805543e-08,
+      "loss": 0.5409,
+      "step": 11550
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 4.22176742111724e-08,
+      "loss": 0.4491,
+      "step": 11560
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 4.054480557373452e-08,
+      "loss": 0.5167,
+      "step": 11570
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.8905619027012065e-08,
+      "loss": 0.4778,
+      "step": 11580
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.7300125703524926e-08,
+      "loss": 0.5106,
+      "step": 11590
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.572833650696439e-08,
+      "loss": 0.4312,
+      "step": 11600
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.419026211212151e-08,
+      "loss": 0.4533,
+      "step": 11610
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.268591296481327e-08,
+      "loss": 0.4726,
+      "step": 11620
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.1215299281812664e-08,
+      "loss": 0.4886,
+      "step": 11630
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.977843105077649e-08,
+      "loss": 0.5077,
+      "step": 11640
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.837531803018212e-08,
+      "loss": 0.5139,
+      "step": 11650
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.700596974925751e-08,
+      "loss": 0.4693,
+      "step": 11660
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.5670395507917944e-08,
+      "loss": 0.4477,
+      "step": 11670
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.4368604376702188e-08,
+      "loss": 0.477,
+      "step": 11680
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.310060519671198e-08,
+      "loss": 0.4652,
+      "step": 11690
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.1866406579551525e-08,
+      "loss": 0.4607,
+      "step": 11700
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.066601690726866e-08,
+      "loss": 0.4795,
+      "step": 11710
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.9499444332297114e-08,
+      "loss": 0.4746,
+      "step": 11720
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.8366696777403213e-08,
+      "loss": 0.493,
+      "step": 11730
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.7267781935630945e-08,
+      "loss": 0.4591,
+      "step": 11740
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.6202707270249196e-08,
+      "loss": 0.4566,
+      "step": 11750
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.5171480014702366e-08,
+      "loss": 0.4885,
+      "step": 11760
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.4174107172560404e-08,
+      "loss": 0.4971,
+      "step": 11770
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.3210595517470503e-08,
+      "loss": 0.467,
+      "step": 11780
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.2280951593112711e-08,
+      "loss": 0.4982,
+      "step": 11790
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.1385181713154393e-08,
+      "loss": 0.4673,
+      "step": 11800
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.0523291961207494e-08,
+      "loss": 0.4683,
+      "step": 11810
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 9.695288190788577e-09,
+      "loss": 0.5238,
+      "step": 11820
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 8.90117602527607e-09,
+      "loss": 0.5104,
+      "step": 11830
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 8.140960857875858e-09,
+      "loss": 0.4234,
+      "step": 11840
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 7.414647851580747e-09,
+      "loss": 0.5093,
+      "step": 11850
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 6.7222419391388406e-09,
+      "loss": 0.5298,
+      "step": 11860
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 6.063747823017996e-09,
+      "loss": 0.4824,
+      "step": 11870
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.4391699753741925e-09,
+      "loss": 0.5416,
+      "step": 11880
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 4.848512638022107e-09,
+      "loss": 0.4853,
+      "step": 11890
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 4.2917798224051396e-09,
+      "loss": 0.5182,
+      "step": 11900
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.7689753095676574e-09,
+      "loss": 0.4784,
+      "step": 11910
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.280102650132788e-09,
+      "loss": 0.4605,
+      "step": 11920
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 2.825165164273558e-09,
+      "loss": 0.493,
+      "step": 11930
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 2.4041659416934593e-09,
+      "loss": 0.5345,
+      "step": 11940
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 2.017107841604804e-09,
+      "loss": 0.5016,
+      "step": 11950
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.6639934927087375e-09,
+      "loss": 0.5209,
+      "step": 11960
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.3448252931774763e-09,
+      "loss": 0.4814,
+      "step": 11970
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.0596054106387642e-09,
+      "loss": 0.4849,
+      "step": 11980
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 8.08335782160885e-10,
+      "loss": 0.421,
+      "step": 11990
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.910181142393389e-10,
+      "loss": 0.4383,
+      "step": 12000
+    },
+    {
+      "epoch": 4.98,
+      "eval_accuracy": 0.6740787801778907,
+      "eval_loss": 0.6774696707725525,
+      "eval_runtime": 50.878,
+      "eval_samples_per_second": 30.937,
+      "eval_steps_per_second": 3.872,
+      "step": 12000
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 4.076538827840759e-10,
+      "loss": 0.4511,
+      "step": 12010
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 2.5824433311227857e-10,
+      "loss": 0.4853,
+      "step": 12020
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.4279047993670525e-10,
+      "loss": 0.4587,
+      "step": 12030
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 6.129310736124883e-11,
+      "loss": 0.4851,
+      "step": 12040
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.3752768874830679e-11,
+      "loss": 0.4848,
+      "step": 12050
+    },
+    {
+      "epoch": 5.0,
+      "step": 12055,
+      "total_flos": 1.5224931143958528e+17,
+      "train_loss": 0.0,
+      "train_runtime": 5.2952,
+      "train_samples_per_second": 72851.348,
+      "train_steps_per_second": 1517.422
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 8035,
+  "num_train_epochs": 5,
+  "save_steps": 500,
+  "total_flos": 1.5224931143958528e+17,
+  "trial_name": null,
+  "trial_params": null
+}