diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,12163 @@
+{
+  "best_metric": 0.5004011988639832,
+  "best_model_checkpoint": "./hf_models/pedestrian_age_recognition_local/checkpoint-20080",
+  "epoch": 10.0,
+  "global_step": 20080,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9990039840637452e-05,
+      "loss": 1.4689,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9980079681274903e-05,
+      "loss": 1.1129,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9970119521912353e-05,
+      "loss": 1.1167,
+      "step": 30
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9960159362549804e-05,
+      "loss": 1.1319,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9950199203187254e-05,
+      "loss": 1.1954,
+      "step": 50
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99402390438247e-05,
+      "loss": 1.0682,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9930278884462152e-05,
+      "loss": 0.9902,
+      "step": 70
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9920318725099602e-05,
+      "loss": 1.0565,
+      "step": 80
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9910358565737053e-05,
+      "loss": 1.2222,
+      "step": 90
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9900398406374503e-05,
+      "loss": 1.0621,
+      "step": 100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9890438247011954e-05,
+      "loss": 1.1944,
+      "step": 110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9880478087649404e-05,
+      "loss": 1.0264,
+      "step": 120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9870517928286855e-05,
+      "loss": 1.1163,
+      "step": 130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9860557768924305e-05,
+      "loss": 1.0442,
+      "step": 140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9850597609561756e-05,
+      "loss": 1.0925,
+      "step": 150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9840637450199206e-05,
+      "loss": 1.0375,
+      "step": 160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9830677290836654e-05,
+      "loss": 0.913,
+      "step": 170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820717131474104e-05,
+      "loss": 1.0721,
+      "step": 180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810756972111555e-05,
+      "loss": 0.9865,
+      "step": 190
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9800796812749005e-05,
+      "loss": 1.1469,
+      "step": 200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9790836653386456e-05,
+      "loss": 1.1865,
+      "step": 210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9780876494023906e-05,
+      "loss": 1.1224,
+      "step": 220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9770916334661357e-05,
+      "loss": 1.0292,
+      "step": 230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9760956175298807e-05,
+      "loss": 1.1153,
+      "step": 240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9750996015936258e-05,
+      "loss": 1.0058,
+      "step": 250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9741035856573708e-05,
+      "loss": 1.1484,
+      "step": 260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.973107569721116e-05,
+      "loss": 1.2046,
+      "step": 270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9721115537848606e-05,
+      "loss": 1.0889,
+      "step": 280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9711155378486056e-05,
+      "loss": 1.17,
+      "step": 290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9701195219123507e-05,
+      "loss": 1.1336,
+      "step": 300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9691235059760957e-05,
+      "loss": 1.2032,
+      "step": 310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9681274900398408e-05,
+      "loss": 1.0043,
+      "step": 320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.967131474103586e-05,
+      "loss": 0.9185,
+      "step": 330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.966135458167331e-05,
+      "loss": 1.1718,
+      "step": 340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.965139442231076e-05,
+      "loss": 0.9823,
+      "step": 350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.964143426294821e-05,
+      "loss": 1.107,
+      "step": 360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.963147410358566e-05,
+      "loss": 0.9359,
+      "step": 370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.962151394422311e-05,
+      "loss": 1.0351,
+      "step": 380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.961155378486056e-05,
+      "loss": 1.1267,
+      "step": 390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.960159362549801e-05,
+      "loss": 0.9142,
+      "step": 400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.959163346613546e-05,
+      "loss": 0.9873,
+      "step": 410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.958167330677291e-05,
+      "loss": 1.3306,
+      "step": 420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.957171314741036e-05,
+      "loss": 1.1228,
+      "step": 430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.956175298804781e-05,
+      "loss": 1.0582,
+      "step": 440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.955179282868526e-05,
+      "loss": 1.055,
+      "step": 450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.954183266932271e-05,
+      "loss": 1.1559,
+      "step": 460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.9531872509960162e-05,
+      "loss": 0.9687,
+      "step": 470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9521912350597613e-05,
+      "loss": 0.8925,
+      "step": 480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.9511952191235063e-05,
+      "loss": 1.0781,
+      "step": 490
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.9501992031872514e-05,
+      "loss": 0.9877,
+      "step": 500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.949203187250996e-05,
+      "loss": 1.1891,
+      "step": 510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.948207171314741e-05,
+      "loss": 1.1934,
+      "step": 520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.9472111553784862e-05,
+      "loss": 1.0067,
+      "step": 530
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9462151394422312e-05,
+      "loss": 1.0404,
+      "step": 540
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.9452191235059763e-05,
+      "loss": 1.1674,
+      "step": 550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9442231075697213e-05,
+      "loss": 0.9697,
+      "step": 560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.9432270916334664e-05,
+      "loss": 1.0253,
+      "step": 570
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9422310756972114e-05,
+      "loss": 1.0359,
+      "step": 580
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.9412350597609565e-05,
+      "loss": 0.8933,
+      "step": 590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9402390438247015e-05,
+      "loss": 0.8644,
+      "step": 600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.9392430278884466e-05,
+      "loss": 0.9537,
+      "step": 610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9382470119521913e-05,
+      "loss": 1.0248,
+      "step": 620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.9372509960159363e-05,
+      "loss": 1.094,
+      "step": 630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9362549800796814e-05,
+      "loss": 1.1684,
+      "step": 640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.9352589641434264e-05,
+      "loss": 1.0236,
+      "step": 650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9342629482071715e-05,
+      "loss": 1.1779,
+      "step": 660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.9332669322709166e-05,
+      "loss": 0.8955,
+      "step": 670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9322709163346616e-05,
+      "loss": 1.1052,
+      "step": 680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.9312749003984067e-05,
+      "loss": 1.1084,
+      "step": 690
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9302788844621517e-05,
+      "loss": 1.0582,
+      "step": 700
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.9292828685258968e-05,
+      "loss": 1.1393,
+      "step": 710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9282868525896418e-05,
+      "loss": 0.8981,
+      "step": 720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.9272908366533865e-05,
+      "loss": 0.9853,
+      "step": 730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9262948207171316e-05,
+      "loss": 0.8977,
+      "step": 740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.9252988047808766e-05,
+      "loss": 0.8489,
+      "step": 750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9243027888446217e-05,
+      "loss": 0.9956,
+      "step": 760
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.9233067729083667e-05,
+      "loss": 1.0202,
+      "step": 770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9223107569721118e-05,
+      "loss": 0.9069,
+      "step": 780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.9213147410358568e-05,
+      "loss": 0.9806,
+      "step": 790
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.920318725099602e-05,
+      "loss": 1.0364,
+      "step": 800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.919322709163347e-05,
+      "loss": 1.0004,
+      "step": 810
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.918326693227092e-05,
+      "loss": 1.0176,
+      "step": 820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.917330677290837e-05,
+      "loss": 1.1888,
+      "step": 830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9163346613545817e-05,
+      "loss": 1.0092,
+      "step": 840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.9153386454183268e-05,
+      "loss": 1.1509,
+      "step": 850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.914342629482072e-05,
+      "loss": 1.0061,
+      "step": 860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.913346613545817e-05,
+      "loss": 0.9612,
+      "step": 870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.912350597609562e-05,
+      "loss": 0.9162,
+      "step": 880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.911354581673307e-05,
+      "loss": 0.8402,
+      "step": 890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.910358565737052e-05,
+      "loss": 0.972,
+      "step": 900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.909362549800797e-05,
+      "loss": 1.001,
+      "step": 910
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.908366533864542e-05,
+      "loss": 0.8827,
+      "step": 920
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.9073705179282872e-05,
+      "loss": 1.0681,
+      "step": 930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.9063745019920323e-05,
+      "loss": 0.848,
+      "step": 940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.905378486055777e-05,
+      "loss": 1.0642,
+      "step": 950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.904382470119522e-05,
+      "loss": 0.9796,
+      "step": 960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.903386454183267e-05,
+      "loss": 1.0757,
+      "step": 970
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.902390438247012e-05,
+      "loss": 0.8905,
+      "step": 980
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.901394422310757e-05,
+      "loss": 0.9239,
+      "step": 990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.9003984063745022e-05,
+      "loss": 1.0321,
+      "step": 1000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.8994023904382473e-05,
+      "loss": 0.9084,
+      "step": 1010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8984063745019923e-05,
+      "loss": 0.9873,
+      "step": 1020
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.8974103585657374e-05,
+      "loss": 0.9286,
+      "step": 1030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8964143426294824e-05,
+      "loss": 0.8777,
+      "step": 1040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.8954183266932275e-05,
+      "loss": 1.0323,
+      "step": 1050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8944223107569725e-05,
+      "loss": 1.0043,
+      "step": 1060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.8934262948207172e-05,
+      "loss": 1.1198,
+      "step": 1070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8924302788844623e-05,
+      "loss": 0.9407,
+      "step": 1080
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8914342629482073e-05,
+      "loss": 0.9462,
+      "step": 1090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8904382470119524e-05,
+      "loss": 0.9172,
+      "step": 1100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.8894422310756974e-05,
+      "loss": 1.0548,
+      "step": 1110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8884462151394425e-05,
+      "loss": 0.8905,
+      "step": 1120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8874501992031875e-05,
+      "loss": 0.9873,
+      "step": 1130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8864541832669326e-05,
+      "loss": 0.8582,
+      "step": 1140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.8854581673306776e-05,
+      "loss": 0.9684,
+      "step": 1150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8844621513944227e-05,
+      "loss": 0.9349,
+      "step": 1160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.8834661354581677e-05,
+      "loss": 1.0601,
+      "step": 1170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8824701195219125e-05,
+      "loss": 1.0681,
+      "step": 1180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.8814741035856575e-05,
+      "loss": 0.8343,
+      "step": 1190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8804780876494026e-05,
+      "loss": 1.0553,
+      "step": 1200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.8794820717131476e-05,
+      "loss": 0.8823,
+      "step": 1210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8784860557768927e-05,
+      "loss": 0.9141,
+      "step": 1220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.8774900398406377e-05,
+      "loss": 0.7898,
+      "step": 1230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8764940239043828e-05,
+      "loss": 1.2327,
+      "step": 1240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.8754980079681278e-05,
+      "loss": 0.9651,
+      "step": 1250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8745019920318725e-05,
+      "loss": 0.9444,
+      "step": 1260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.8735059760956176e-05,
+      "loss": 0.9382,
+      "step": 1270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8725099601593626e-05,
+      "loss": 0.9849,
+      "step": 1280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.8715139442231077e-05,
+      "loss": 0.766,
+      "step": 1290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.8705179282868527e-05,
+      "loss": 1.0273,
+      "step": 1300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.8695219123505978e-05,
+      "loss": 0.7574,
+      "step": 1310
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.868525896414343e-05,
+      "loss": 0.9437,
+      "step": 1320
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.8675298804780875e-05,
+      "loss": 0.9649,
+      "step": 1330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.8665338645418326e-05,
+      "loss": 0.8873,
+      "step": 1340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.8655378486055777e-05,
+      "loss": 0.9987,
+      "step": 1350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.8645418326693227e-05,
+      "loss": 1.0478,
+      "step": 1360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.8635458167330678e-05,
+      "loss": 1.059,
+      "step": 1370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.8625498007968128e-05,
+      "loss": 0.8411,
+      "step": 1380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.861553784860558e-05,
+      "loss": 0.9811,
+      "step": 1390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.860557768924303e-05,
+      "loss": 0.9445,
+      "step": 1400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.859561752988048e-05,
+      "loss": 0.7851,
+      "step": 1410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.858565737051793e-05,
+      "loss": 1.119,
+      "step": 1420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.857569721115538e-05,
+      "loss": 0.7936,
+      "step": 1430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8565737051792828e-05,
+      "loss": 0.9803,
+      "step": 1440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8555776892430278e-05,
+      "loss": 0.891,
+      "step": 1450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.854581673306773e-05,
+      "loss": 1.0097,
+      "step": 1460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.853585657370518e-05,
+      "loss": 0.81,
+      "step": 1470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.852589641434263e-05,
+      "loss": 0.8764,
+      "step": 1480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.851593625498008e-05,
+      "loss": 0.7936,
+      "step": 1490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.850597609561753e-05,
+      "loss": 1.1809,
+      "step": 1500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.849601593625498e-05,
+      "loss": 1.081,
+      "step": 1510
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.8486055776892432e-05,
+      "loss": 0.9378,
+      "step": 1520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.8476095617529882e-05,
+      "loss": 0.901,
+      "step": 1530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.8466135458167333e-05,
+      "loss": 0.8377,
+      "step": 1540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.845617529880478e-05,
+      "loss": 0.9604,
+      "step": 1550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.844621513944223e-05,
+      "loss": 0.8409,
+      "step": 1560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.843625498007968e-05,
+      "loss": 0.8396,
+      "step": 1570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.842629482071713e-05,
+      "loss": 0.9565,
+      "step": 1580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.8416334661354582e-05,
+      "loss": 0.9798,
+      "step": 1590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8406374501992033e-05,
+      "loss": 0.9729,
+      "step": 1600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.8396414342629483e-05,
+      "loss": 0.8596,
+      "step": 1610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8386454183266934e-05,
+      "loss": 0.8209,
+      "step": 1620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8376494023904384e-05,
+      "loss": 1.0466,
+      "step": 1630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8366533864541835e-05,
+      "loss": 1.0604,
+      "step": 1640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.8356573705179285e-05,
+      "loss": 1.0126,
+      "step": 1650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.8346613545816732e-05,
+      "loss": 0.9113,
+      "step": 1660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.8336653386454183e-05,
+      "loss": 1.0855,
+      "step": 1670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8326693227091633e-05,
+      "loss": 0.9407,
+      "step": 1680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8316733067729084e-05,
+      "loss": 0.8055,
+      "step": 1690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.8306772908366534e-05,
+      "loss": 1.0912,
+      "step": 1700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.8296812749003985e-05,
+      "loss": 1.061,
+      "step": 1710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8286852589641435e-05,
+      "loss": 0.9365,
+      "step": 1720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.8276892430278886e-05,
+      "loss": 0.8512,
+      "step": 1730
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.8266932270916336e-05,
+      "loss": 0.8024,
+      "step": 1740
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.8256972111553787e-05,
+      "loss": 0.8694,
+      "step": 1750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8247011952191237e-05,
+      "loss": 0.9038,
+      "step": 1760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8237051792828684e-05,
+      "loss": 0.8766,
+      "step": 1770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.8227091633466135e-05,
+      "loss": 1.1225,
+      "step": 1780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.8217131474103585e-05,
+      "loss": 0.8447,
+      "step": 1790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.8207171314741036e-05,
+      "loss": 0.8833,
+      "step": 1800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.8197211155378486e-05,
+      "loss": 1.0964,
+      "step": 1810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8187250996015937e-05,
+      "loss": 0.8619,
+      "step": 1820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.8177290836653387e-05,
+      "loss": 0.8513,
+      "step": 1830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.8167330677290838e-05,
+      "loss": 0.9351,
+      "step": 1840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.815737051792829e-05,
+      "loss": 0.9942,
+      "step": 1850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.814741035856574e-05,
+      "loss": 0.9342,
+      "step": 1860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.813745019920319e-05,
+      "loss": 1.0993,
+      "step": 1870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.812749003984064e-05,
+      "loss": 0.7938,
+      "step": 1880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8117529880478087e-05,
+      "loss": 0.6681,
+      "step": 1890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8107569721115538e-05,
+      "loss": 0.9543,
+      "step": 1900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8097609561752988e-05,
+      "loss": 1.0372,
+      "step": 1910
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.808764940239044e-05,
+      "loss": 0.8814,
+      "step": 1920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.807768924302789e-05,
+      "loss": 0.8848,
+      "step": 1930
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.806772908366534e-05,
+      "loss": 0.7841,
+      "step": 1940
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.805776892430279e-05,
+      "loss": 0.876,
+      "step": 1950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.804780876494024e-05,
+      "loss": 0.8038,
+      "step": 1960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.803784860557769e-05,
+      "loss": 0.8161,
+      "step": 1970
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8027888446215142e-05,
+      "loss": 1.0651,
+      "step": 1980
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8017928286852592e-05,
+      "loss": 0.8864,
+      "step": 1990
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.800796812749004e-05,
+      "loss": 0.8849,
+      "step": 2000
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.6806633733239238,
+      "eval_loss": 0.79388827085495,
+      "eval_runtime": 12.645,
+      "eval_samples_per_second": 224.121,
+      "eval_steps_per_second": 28.074,
+      "step": 2008
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.799800796812749e-05,
+      "loss": 0.7629,
+      "step": 2010
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.798804780876494e-05,
+      "loss": 0.8553,
+      "step": 2020
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.797808764940239e-05,
+      "loss": 0.8713,
+      "step": 2030
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.796812749003984e-05,
+      "loss": 0.9278,
+      "step": 2040
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.7958167330677292e-05,
+      "loss": 0.8988,
+      "step": 2050
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.7948207171314742e-05,
+      "loss": 0.7795,
+      "step": 2060
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.7938247011952193e-05,
+      "loss": 0.9166,
+      "step": 2070
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.7928286852589643e-05,
+      "loss": 0.86,
+      "step": 2080
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.7918326693227094e-05,
+      "loss": 0.8851,
+      "step": 2090
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.7908366533864544e-05,
+      "loss": 0.8794,
+      "step": 2100
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.789840637450199e-05,
+      "loss": 0.9633,
+      "step": 2110
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.7888446215139442e-05,
+      "loss": 0.8549,
+      "step": 2120
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.7878486055776893e-05,
+      "loss": 0.7932,
+      "step": 2130
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.7868525896414343e-05,
+      "loss": 0.6964,
+      "step": 2140
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.7858565737051794e-05,
+      "loss": 0.7684,
+      "step": 2150
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.7848605577689244e-05,
+      "loss": 0.8694,
+      "step": 2160
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.7838645418326695e-05,
+      "loss": 0.9167,
+      "step": 2170
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.7828685258964145e-05,
+      "loss": 0.8935,
+      "step": 2180
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.7818725099601596e-05,
+      "loss": 0.7446,
+      "step": 2190
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.7808764940239046e-05,
+      "loss": 0.7525,
+      "step": 2200
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.7798804780876497e-05,
+      "loss": 0.8569,
+      "step": 2210
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.7788844621513944e-05,
+      "loss": 0.8718,
+      "step": 2220
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.7778884462151394e-05,
+      "loss": 0.8311,
+      "step": 2230
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.7768924302788845e-05,
+      "loss": 0.9541,
+      "step": 2240
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.7758964143426295e-05,
+      "loss": 0.8528,
+      "step": 2250
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.7749003984063746e-05,
+      "loss": 0.8423,
+      "step": 2260
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.7739043824701196e-05,
+      "loss": 0.7685,
+      "step": 2270
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.7729083665338647e-05,
+      "loss": 0.867,
+      "step": 2280
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.7719123505976097e-05,
+      "loss": 0.684,
+      "step": 2290
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.7709163346613548e-05,
+      "loss": 0.9472,
+      "step": 2300
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.7699203187251e-05,
+      "loss": 1.153,
+      "step": 2310
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.768924302788845e-05,
+      "loss": 0.8524,
+      "step": 2320
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.7679282868525896e-05,
+      "loss": 0.8807,
+      "step": 2330
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.7669322709163347e-05,
+      "loss": 0.7363,
+      "step": 2340
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.7659362549800797e-05,
+      "loss": 0.8823,
+      "step": 2350
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.7649402390438248e-05,
+      "loss": 0.722,
+      "step": 2360
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.7639442231075698e-05,
+      "loss": 0.9221,
+      "step": 2370
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.762948207171315e-05,
+      "loss": 0.747,
+      "step": 2380
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.76195219123506e-05,
+      "loss": 0.8488,
+      "step": 2390
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.760956175298805e-05,
+      "loss": 0.7938,
+      "step": 2400
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.75996015936255e-05,
+      "loss": 0.8894,
+      "step": 2410
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.758964143426295e-05,
+      "loss": 0.9678,
+      "step": 2420
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.75796812749004e-05,
+      "loss": 0.8414,
+      "step": 2430
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.756972111553785e-05,
+      "loss": 0.8822,
+      "step": 2440
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.75597609561753e-05,
+      "loss": 0.7541,
+      "step": 2450
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.754980079681275e-05,
+      "loss": 0.8758,
+      "step": 2460
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.75398406374502e-05,
+      "loss": 0.8398,
+      "step": 2470
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.752988047808765e-05,
+      "loss": 0.8705,
+      "step": 2480
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.75199203187251e-05,
+      "loss": 0.6453,
+      "step": 2490
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.750996015936255e-05,
+      "loss": 0.7195,
+      "step": 2500
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 0.9367,
+      "step": 2510
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.7490039840637452e-05,
+      "loss": 0.7887,
+      "step": 2520
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.7480079681274903e-05,
+      "loss": 1.087,
+      "step": 2530
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.7470119521912353e-05,
+      "loss": 0.721,
+      "step": 2540
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.7460159362549804e-05,
+      "loss": 0.9043,
+      "step": 2550
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.745019920318725e-05,
+      "loss": 0.9658,
+      "step": 2560
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.74402390438247e-05,
+      "loss": 0.7609,
+      "step": 2570
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.7430278884462152e-05,
+      "loss": 0.7956,
+      "step": 2580
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.7420318725099603e-05,
+      "loss": 0.7127,
+      "step": 2590
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.7410358565737053e-05,
+      "loss": 0.8127,
+      "step": 2600
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.7400398406374504e-05,
+      "loss": 0.8822,
+      "step": 2610
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.7390438247011954e-05,
+      "loss": 0.9149,
+      "step": 2620
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.7380478087649405e-05,
+      "loss": 0.7677,
+      "step": 2630
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.7370517928286855e-05,
+      "loss": 0.8626,
+      "step": 2640
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.7360557768924306e-05,
+      "loss": 0.7123,
+      "step": 2650
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.7350597609561756e-05,
+      "loss": 0.9579,
+      "step": 2660
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.7340637450199203e-05,
+      "loss": 0.9945,
+      "step": 2670
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.7330677290836654e-05,
+      "loss": 0.9694,
+      "step": 2680
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.7320717131474104e-05,
+      "loss": 0.8601,
+      "step": 2690
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.7310756972111555e-05,
+      "loss": 0.8867,
+      "step": 2700
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.7300796812749005e-05,
+      "loss": 0.8647,
+      "step": 2710
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.7290836653386456e-05,
+      "loss": 0.79,
+      "step": 2720
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.7280876494023906e-05,
+      "loss": 0.9502,
+      "step": 2730
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.7270916334661357e-05,
+      "loss": 0.8664,
+      "step": 2740
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.7260956175298807e-05,
+      "loss": 0.9108,
+      "step": 2750
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.7250996015936258e-05,
+      "loss": 0.749,
+      "step": 2760
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.724103585657371e-05,
+      "loss": 0.7422,
+      "step": 2770
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.7231075697211155e-05,
+      "loss": 0.7023,
+      "step": 2780
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.7221115537848606e-05,
+      "loss": 0.9479,
+      "step": 2790
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.7211155378486056e-05,
+      "loss": 0.7941,
+      "step": 2800
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.7201195219123507e-05,
+      "loss": 0.7438,
+      "step": 2810
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.7191235059760958e-05,
+      "loss": 0.8912,
+      "step": 2820
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.7181274900398408e-05,
+      "loss": 0.7396,
+      "step": 2830
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.717131474103586e-05,
+      "loss": 0.7853,
+      "step": 2840
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.716135458167331e-05,
+      "loss": 0.8629,
+      "step": 2850
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.715139442231076e-05,
+      "loss": 0.7716,
+      "step": 2860
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.714143426294821e-05,
+      "loss": 0.8541,
+      "step": 2870
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.713147410358566e-05,
+      "loss": 0.7971,
+      "step": 2880
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.7121513944223108e-05,
+      "loss": 0.9978,
+      "step": 2890
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.7111553784860558e-05,
+      "loss": 0.717,
+      "step": 2900
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.710159362549801e-05,
+      "loss": 0.7371,
+      "step": 2910
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.709163346613546e-05,
+      "loss": 0.7527,
+      "step": 2920
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.708167330677291e-05,
+      "loss": 0.9245,
+      "step": 2930
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.707171314741036e-05,
+      "loss": 0.9383,
+      "step": 2940
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.706175298804781e-05,
+      "loss": 0.9054,
+      "step": 2950
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.705179282868526e-05,
+      "loss": 0.8007,
+      "step": 2960
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.7041832669322712e-05,
+      "loss": 0.9661,
+      "step": 2970
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.7031872509960162e-05,
+      "loss": 1.1299,
+      "step": 2980
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.7021912350597613e-05,
+      "loss": 0.8637,
+      "step": 2990
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.701195219123506e-05,
+      "loss": 0.8469,
+      "step": 3000
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.700199203187251e-05,
+      "loss": 0.9003,
+      "step": 3010
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.699203187250996e-05,
+      "loss": 0.7667,
+      "step": 3020
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.698207171314741e-05,
+      "loss": 0.7415,
+      "step": 3030
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.6972111553784862e-05,
+      "loss": 0.8631,
+      "step": 3040
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.6962151394422312e-05,
+      "loss": 0.9492,
+      "step": 3050
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.6952191235059763e-05,
+      "loss": 0.8998,
+      "step": 3060
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.6942231075697214e-05,
+      "loss": 0.8409,
+      "step": 3070
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.6932270916334664e-05,
+      "loss": 0.8415,
+      "step": 3080
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.6922310756972115e-05,
+      "loss": 0.8349,
+      "step": 3090
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.6912350597609565e-05,
+      "loss": 0.6596,
+      "step": 3100
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.6902390438247016e-05,
+      "loss": 0.8337,
+      "step": 3110
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.6892430278884463e-05,
+      "loss": 0.8204,
+      "step": 3120
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.6882470119521913e-05,
+      "loss": 0.7054,
+      "step": 3130
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.6872509960159364e-05,
+      "loss": 0.7432,
+      "step": 3140
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.6862549800796814e-05,
+      "loss": 0.8096,
+      "step": 3150
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.6852589641434265e-05,
+      "loss": 0.7895,
+      "step": 3160
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.6842629482071715e-05,
+      "loss": 0.7825,
+      "step": 3170
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.6832669322709166e-05,
+      "loss": 0.7706,
+      "step": 3180
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.6822709163346616e-05,
+      "loss": 0.8894,
+      "step": 3190
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.6812749003984067e-05,
+      "loss": 0.7612,
+      "step": 3200
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.6802788844621517e-05,
+      "loss": 0.767,
+      "step": 3210
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.6792828685258968e-05,
+      "loss": 0.6877,
+      "step": 3220
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.6782868525896415e-05,
+      "loss": 0.8885,
+      "step": 3230
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.6772908366533865e-05,
+      "loss": 1.0056,
+      "step": 3240
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.6762948207171316e-05,
+      "loss": 0.7839,
+      "step": 3250
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 1.6752988047808766e-05,
+      "loss": 0.6917,
+      "step": 3260
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.6743027888446217e-05,
+      "loss": 0.8032,
+      "step": 3270
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.6733067729083667e-05,
+      "loss": 0.6976,
+      "step": 3280
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.6723107569721118e-05,
+      "loss": 0.7373,
+      "step": 3290
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 1.671314741035857e-05,
+      "loss": 0.6648,
+      "step": 3300
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.670318725099602e-05,
+      "loss": 0.8966,
+      "step": 3310
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 1.669322709163347e-05,
+      "loss": 0.9145,
+      "step": 3320
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.668326693227092e-05,
+      "loss": 0.8366,
+      "step": 3330
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.6673306772908367e-05,
+      "loss": 0.7289,
+      "step": 3340
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6663346613545818e-05,
+      "loss": 0.8252,
+      "step": 3350
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 1.6653386454183268e-05,
+      "loss": 0.7782,
+      "step": 3360
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.664342629482072e-05,
+      "loss": 0.9192,
+      "step": 3370
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 1.663346613545817e-05,
+      "loss": 0.8619,
+      "step": 3380
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.662350597609562e-05,
+      "loss": 0.7842,
+      "step": 3390
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.661354581673307e-05,
+      "loss": 0.7818,
+      "step": 3400
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.660358565737052e-05,
+      "loss": 0.7216,
+      "step": 3410
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 1.659362549800797e-05,
+      "loss": 0.7956,
+      "step": 3420
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.6583665338645422e-05,
+      "loss": 0.7667,
+      "step": 3430
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.6573705179282872e-05,
+      "loss": 0.9664,
+      "step": 3440
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.656374501992032e-05,
+      "loss": 0.8326,
+      "step": 3450
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 1.655378486055777e-05,
+      "loss": 0.9226,
+      "step": 3460
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.654382470119522e-05,
+      "loss": 0.9168,
+      "step": 3470
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 1.653386454183267e-05,
+      "loss": 0.8733,
+      "step": 3480
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.652390438247012e-05,
+      "loss": 0.6706,
+      "step": 3490
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.6513944223107572e-05,
+      "loss": 0.9142,
+      "step": 3500
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.6503984063745022e-05,
+      "loss": 0.8239,
+      "step": 3510
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 1.6494023904382473e-05,
+      "loss": 0.663,
+      "step": 3520
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.6484063745019923e-05,
+      "loss": 0.6629,
+      "step": 3530
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 1.6474103585657374e-05,
+      "loss": 0.7573,
+      "step": 3540
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.6464143426294824e-05,
+      "loss": 0.9648,
+      "step": 3550
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.645418326693227e-05,
+      "loss": 0.9071,
+      "step": 3560
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.6444223107569722e-05,
+      "loss": 0.6767,
+      "step": 3570
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 1.6434262948207173e-05,
+      "loss": 0.7979,
+      "step": 3580
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.6424302788844623e-05,
+      "loss": 0.7986,
+      "step": 3590
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 1.6414342629482074e-05,
+      "loss": 0.6969,
+      "step": 3600
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.6404382470119524e-05,
+      "loss": 0.9112,
+      "step": 3610
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.6394422310756975e-05,
+      "loss": 0.9941,
+      "step": 3620
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.6384462151394425e-05,
+      "loss": 0.7969,
+      "step": 3630
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.6374501992031876e-05,
+      "loss": 1.0053,
+      "step": 3640
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.6364541832669326e-05,
+      "loss": 0.6397,
+      "step": 3650
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.6354581673306777e-05,
+      "loss": 0.8681,
+      "step": 3660
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.6344621513944224e-05,
+      "loss": 0.743,
+      "step": 3670
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.6334661354581674e-05,
+      "loss": 0.8639,
+      "step": 3680
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6324701195219125e-05,
+      "loss": 0.7352,
+      "step": 3690
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6314741035856575e-05,
+      "loss": 0.8693,
+      "step": 3700
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.6304780876494026e-05,
+      "loss": 0.9401,
+      "step": 3710
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.6294820717131476e-05,
+      "loss": 0.6446,
+      "step": 3720
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.6284860557768927e-05,
+      "loss": 0.6993,
+      "step": 3730
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.6274900398406377e-05,
+      "loss": 0.8636,
+      "step": 3740
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.6264940239043828e-05,
+      "loss": 0.658,
+      "step": 3750
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.625498007968128e-05,
+      "loss": 0.7295,
+      "step": 3760
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.6245019920318726e-05,
+      "loss": 0.9316,
+      "step": 3770
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.6235059760956176e-05,
+      "loss": 0.834,
+      "step": 3780
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6225099601593627e-05,
+      "loss": 0.7642,
+      "step": 3790
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.6215139442231077e-05,
+      "loss": 0.6506,
+      "step": 3800
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.6205179282868528e-05,
+      "loss": 0.9876,
+      "step": 3810
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.6195219123505975e-05,
+      "loss": 0.8066,
+      "step": 3820
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.6185258964143425e-05,
+      "loss": 1.0183,
+      "step": 3830
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.6175298804780876e-05,
+      "loss": 0.7388,
+      "step": 3840
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.6165338645418326e-05,
+      "loss": 0.7146,
+      "step": 3850
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.6155378486055777e-05,
+      "loss": 0.6693,
+      "step": 3860
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.6145418326693227e-05,
+      "loss": 0.7391,
+      "step": 3870
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.6135458167330678e-05,
+      "loss": 0.7267,
+      "step": 3880
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.6125498007968128e-05,
+      "loss": 1.0671,
+      "step": 3890
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.611553784860558e-05,
+      "loss": 0.7249,
+      "step": 3900
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.610557768924303e-05,
+      "loss": 0.7674,
+      "step": 3910
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.609561752988048e-05,
+      "loss": 0.7563,
+      "step": 3920
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.608565737051793e-05,
+      "loss": 0.7904,
+      "step": 3930
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.6075697211155377e-05,
+      "loss": 0.9466,
+      "step": 3940
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.6065737051792828e-05,
+      "loss": 0.7063,
+      "step": 3950
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 1.605577689243028e-05,
+      "loss": 0.6213,
+      "step": 3960
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.604581673306773e-05,
+      "loss": 0.8229,
+      "step": 3970
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 1.603585657370518e-05,
+      "loss": 0.8078,
+      "step": 3980
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.602589641434263e-05,
+      "loss": 0.8423,
+      "step": 3990
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 1.601593625498008e-05,
+      "loss": 0.6573,
+      "step": 4000
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.600597609561753e-05,
+      "loss": 0.9836,
+      "step": 4010
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.733592095977417,
+      "eval_loss": 0.6694180965423584,
+      "eval_runtime": 12.4658,
+      "eval_samples_per_second": 227.342,
+      "eval_steps_per_second": 28.478,
+      "step": 4016
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 1.599601593625498e-05,
+      "loss": 0.6801,
+      "step": 4020
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 1.5986055776892432e-05,
+      "loss": 0.7124,
+      "step": 4030
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 1.5976095617529883e-05,
+      "loss": 0.7104,
+      "step": 4040
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 1.596613545816733e-05,
+      "loss": 0.6681,
+      "step": 4050
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 1.595617529880478e-05,
+      "loss": 0.8053,
+      "step": 4060
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 1.594621513944223e-05,
+      "loss": 0.8205,
+      "step": 4070
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 1.593625498007968e-05,
+      "loss": 0.7367,
+      "step": 4080
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 1.5926294820717132e-05,
+      "loss": 0.8046,
+      "step": 4090
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 1.5916334661354582e-05,
+      "loss": 0.7983,
+      "step": 4100
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 1.5906374501992033e-05,
+      "loss": 0.6582,
+      "step": 4110
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 1.5896414342629483e-05,
+      "loss": 0.836,
+      "step": 4120
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 1.5886454183266934e-05,
+      "loss": 0.9061,
+      "step": 4130
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 1.5876494023904384e-05,
+      "loss": 0.8273,
+      "step": 4140
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 1.5866533864541835e-05,
+      "loss": 0.8314,
+      "step": 4150
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 1.5856573705179282e-05,
+      "loss": 0.8168,
+      "step": 4160
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 1.5846613545816732e-05,
+      "loss": 0.7776,
+      "step": 4170
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 1.5836653386454183e-05,
+      "loss": 0.6865,
+      "step": 4180
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 1.5826693227091633e-05,
+      "loss": 0.9374,
+      "step": 4190
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 1.5816733067729084e-05,
+      "loss": 0.7698,
+      "step": 4200
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 1.5806772908366534e-05,
+      "loss": 0.5878,
+      "step": 4210
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 1.5796812749003985e-05,
+      "loss": 0.8018,
+      "step": 4220
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 1.5786852589641435e-05,
+      "loss": 0.701,
+      "step": 4230
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 1.5776892430278886e-05,
+      "loss": 0.7707,
+      "step": 4240
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 1.5766932270916336e-05,
+      "loss": 0.8841,
+      "step": 4250
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 1.5756972111553787e-05,
+      "loss": 0.726,
+      "step": 4260
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 1.5747011952191234e-05,
+      "loss": 0.6116,
+      "step": 4270
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 1.5737051792828685e-05,
+      "loss": 0.9335,
+      "step": 4280
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 1.5727091633466135e-05,
+      "loss": 0.8876,
+      "step": 4290
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 1.5717131474103586e-05,
+      "loss": 0.8354,
+      "step": 4300
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 1.5707171314741036e-05,
+      "loss": 0.6105,
+      "step": 4310
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 1.5697211155378487e-05,
+      "loss": 0.7545,
+      "step": 4320
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 1.5687250996015937e-05,
+      "loss": 0.6566,
+      "step": 4330
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 1.5677290836653388e-05,
+      "loss": 0.6197,
+      "step": 4340
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 1.5667330677290838e-05,
+      "loss": 0.8349,
+      "step": 4350
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 1.565737051792829e-05,
+      "loss": 0.6547,
+      "step": 4360
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 1.564741035856574e-05,
+      "loss": 0.6429,
+      "step": 4370
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 1.5637450199203186e-05,
+      "loss": 0.753,
+      "step": 4380
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 1.5627490039840637e-05,
+      "loss": 0.7152,
+      "step": 4390
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 1.5617529880478087e-05,
+      "loss": 0.7747,
+      "step": 4400
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 1.5607569721115538e-05,
+      "loss": 0.6632,
+      "step": 4410
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 1.559760956175299e-05,
+      "loss": 0.7944,
+      "step": 4420
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 1.558764940239044e-05,
+      "loss": 0.7096,
+      "step": 4430
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 1.557768924302789e-05,
+      "loss": 0.7821,
+      "step": 4440
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 1.556772908366534e-05,
+      "loss": 0.7711,
+      "step": 4450
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 1.555776892430279e-05,
+      "loss": 0.707,
+      "step": 4460
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 1.554780876494024e-05,
+      "loss": 0.8105,
+      "step": 4470
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 1.553784860557769e-05,
+      "loss": 0.7196,
+      "step": 4480
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 1.552788844621514e-05,
+      "loss": 0.8178,
+      "step": 4490
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 1.551792828685259e-05,
+      "loss": 0.8944,
+      "step": 4500
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 1.550796812749004e-05,
+      "loss": 0.729,
+      "step": 4510
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 1.549800796812749e-05,
+      "loss": 0.7182,
+      "step": 4520
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 1.548804780876494e-05,
+      "loss": 0.7866,
+      "step": 4530
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 1.547808764940239e-05,
+      "loss": 0.8899,
+      "step": 4540
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 1.546812749003984e-05,
+      "loss": 0.8174,
+      "step": 4550
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 1.5458167330677292e-05,
+      "loss": 0.6495,
+      "step": 4560
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 1.5448207171314743e-05,
+      "loss": 0.7917,
+      "step": 4570
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 1.5438247011952193e-05,
+      "loss": 0.9344,
+      "step": 4580
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 1.5428286852589644e-05,
+      "loss": 0.9611,
+      "step": 4590
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 1.5418326693227094e-05,
+      "loss": 0.8078,
+      "step": 4600
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 1.540836653386454e-05,
+      "loss": 0.6751,
+      "step": 4610
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 1.5398406374501992e-05,
+      "loss": 0.7829,
+      "step": 4620
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 1.5388446215139442e-05,
+      "loss": 0.7538,
+      "step": 4630
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 1.5378486055776893e-05,
+      "loss": 0.9052,
+      "step": 4640
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 1.5368525896414343e-05,
+      "loss": 0.8216,
+      "step": 4650
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 1.5358565737051794e-05,
+      "loss": 0.8,
+      "step": 4660
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 1.5348605577689244e-05,
+      "loss": 0.7506,
+      "step": 4670
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 1.5338645418326695e-05,
+      "loss": 0.7105,
+      "step": 4680
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 1.5328685258964145e-05,
+      "loss": 0.6814,
+      "step": 4690
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 1.5318725099601596e-05,
+      "loss": 0.7265,
+      "step": 4700
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 1.5308764940239046e-05,
+      "loss": 0.6906,
+      "step": 4710
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 1.5298804780876494e-05,
+      "loss": 0.73,
+      "step": 4720
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 1.5288844621513944e-05,
+      "loss": 0.5701,
+      "step": 4730
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 1.5278884462151395e-05,
+      "loss": 0.7003,
+      "step": 4740
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 1.5268924302788845e-05,
+      "loss": 0.7284,
+      "step": 4750
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 1.5258964143426296e-05,
+      "loss": 0.7239,
+      "step": 4760
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 1.5249003984063746e-05,
+      "loss": 0.6051,
+      "step": 4770
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 1.5239043824701197e-05,
+      "loss": 0.6882,
+      "step": 4780
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.5229083665338647e-05,
+      "loss": 1.0491,
+      "step": 4790
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 1.5219123505976096e-05,
+      "loss": 0.6975,
+      "step": 4800
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.5209163346613546e-05,
+      "loss": 0.7106,
+      "step": 4810
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 1.5199203187250997e-05,
+      "loss": 0.6782,
+      "step": 4820
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.5189243027888447e-05,
+      "loss": 0.7955,
+      "step": 4830
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 1.5179282868525898e-05,
+      "loss": 0.6715,
+      "step": 4840
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.5169322709163348e-05,
+      "loss": 0.6772,
+      "step": 4850
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 1.5159362549800797e-05,
+      "loss": 0.6932,
+      "step": 4860
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.5149402390438248e-05,
+      "loss": 0.6818,
+      "step": 4870
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 1.5139442231075698e-05,
+      "loss": 0.7105,
+      "step": 4880
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.5129482071713149e-05,
+      "loss": 0.5646,
+      "step": 4890
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 1.51195219123506e-05,
+      "loss": 0.7916,
+      "step": 4900
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5109561752988048e-05,
+      "loss": 0.7126,
+      "step": 4910
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 1.5099601593625499e-05,
+      "loss": 0.7023,
+      "step": 4920
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.508964143426295e-05,
+      "loss": 0.6397,
+      "step": 4930
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 1.50796812749004e-05,
+      "loss": 0.7289,
+      "step": 4940
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.506972111553785e-05,
+      "loss": 0.8576,
+      "step": 4950
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 1.50597609561753e-05,
+      "loss": 0.5869,
+      "step": 4960
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.504980079681275e-05,
+      "loss": 0.7487,
+      "step": 4970
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 1.50398406374502e-05,
+      "loss": 0.7766,
+      "step": 4980
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.502988047808765e-05,
+      "loss": 0.7278,
+      "step": 4990
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 1.5019920318725101e-05,
+      "loss": 0.6754,
+      "step": 5000
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.5009960159362552e-05,
+      "loss": 0.56,
+      "step": 5010
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.9614,
+      "step": 5020
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 1.4990039840637451e-05,
+      "loss": 0.7505,
+      "step": 5030
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.4980079681274901e-05,
+      "loss": 0.6653,
+      "step": 5040
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 1.4970119521912352e-05,
+      "loss": 0.7601,
+      "step": 5050
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.4960159362549802e-05,
+      "loss": 0.742,
+      "step": 5060
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 1.4950199203187253e-05,
+      "loss": 0.6162,
+      "step": 5070
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.4940239043824702e-05,
+      "loss": 0.6897,
+      "step": 5080
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 1.4930278884462152e-05,
+      "loss": 0.8665,
+      "step": 5090
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.4920318725099603e-05,
+      "loss": 0.7284,
+      "step": 5100
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 1.4910358565737053e-05,
+      "loss": 0.5914,
+      "step": 5110
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.4900398406374504e-05,
+      "loss": 0.7063,
+      "step": 5120
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 1.4890438247011954e-05,
+      "loss": 0.7157,
+      "step": 5130
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.4880478087649403e-05,
+      "loss": 0.7803,
+      "step": 5140
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 1.4870517928286854e-05,
+      "loss": 0.6951,
+      "step": 5150
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.4860557768924304e-05,
+      "loss": 0.7263,
+      "step": 5160
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 1.4850597609561755e-05,
+      "loss": 0.7659,
+      "step": 5170
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.4840637450199205e-05,
+      "loss": 0.7443,
+      "step": 5180
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 1.4830677290836654e-05,
+      "loss": 0.7188,
+      "step": 5190
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.4820717131474104e-05,
+      "loss": 0.7931,
+      "step": 5200
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 1.4810756972111555e-05,
+      "loss": 0.7457,
+      "step": 5210
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.4800796812749006e-05,
+      "loss": 0.673,
+      "step": 5220
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 1.4790836653386456e-05,
+      "loss": 0.7712,
+      "step": 5230
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.4780876494023907e-05,
+      "loss": 0.9287,
+      "step": 5240
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 1.4770916334661355e-05,
+      "loss": 0.6749,
+      "step": 5250
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.4760956175298806e-05,
+      "loss": 0.6252,
+      "step": 5260
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 1.4750996015936256e-05,
+      "loss": 0.7767,
+      "step": 5270
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.4741035856573707e-05,
+      "loss": 0.8165,
+      "step": 5280
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 1.4731075697211157e-05,
+      "loss": 0.5819,
+      "step": 5290
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.4721115537848606e-05,
+      "loss": 0.6778,
+      "step": 5300
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 1.4711155378486057e-05,
+      "loss": 0.7556,
+      "step": 5310
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.4701195219123507e-05,
+      "loss": 0.7072,
+      "step": 5320
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 1.4691235059760958e-05,
+      "loss": 0.7781,
+      "step": 5330
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.4681274900398408e-05,
+      "loss": 0.5562,
+      "step": 5340
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 1.4671314741035859e-05,
+      "loss": 0.6716,
+      "step": 5350
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.4661354581673308e-05,
+      "loss": 0.4993,
+      "step": 5360
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 1.4651394422310758e-05,
+      "loss": 0.6522,
+      "step": 5370
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4641434262948209e-05,
+      "loss": 0.7985,
+      "step": 5380
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 1.4631474103585659e-05,
+      "loss": 0.8129,
+      "step": 5390
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.462151394422311e-05,
+      "loss": 0.612,
+      "step": 5400
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.461155378486056e-05,
+      "loss": 0.7853,
+      "step": 5410
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.4601593625498009e-05,
+      "loss": 0.7914,
+      "step": 5420
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.459163346613546e-05,
+      "loss": 0.6586,
+      "step": 5430
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.458167330677291e-05,
+      "loss": 0.7156,
+      "step": 5440
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.457171314741036e-05,
+      "loss": 0.7093,
+      "step": 5450
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.4561752988047811e-05,
+      "loss": 0.749,
+      "step": 5460
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.455179282868526e-05,
+      "loss": 0.9567,
+      "step": 5470
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.454183266932271e-05,
+      "loss": 0.8337,
+      "step": 5480
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.453187250996016e-05,
+      "loss": 0.852,
+      "step": 5490
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.4521912350597611e-05,
+      "loss": 0.6419,
+      "step": 5500
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.4511952191235062e-05,
+      "loss": 0.8001,
+      "step": 5510
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.4501992031872512e-05,
+      "loss": 0.8242,
+      "step": 5520
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.4492031872509961e-05,
+      "loss": 0.7631,
+      "step": 5530
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.4482071713147412e-05,
+      "loss": 0.7759,
+      "step": 5540
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.4472111553784862e-05,
+      "loss": 0.6634,
+      "step": 5550
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.4462151394422313e-05,
+      "loss": 0.7725,
+      "step": 5560
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.4452191235059763e-05,
+      "loss": 0.6164,
+      "step": 5570
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.4442231075697212e-05,
+      "loss": 0.7918,
+      "step": 5580
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 1.4432270916334663e-05,
+      "loss": 0.7932,
+      "step": 5590
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.4422310756972113e-05,
+      "loss": 0.6857,
+      "step": 5600
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.4412350597609564e-05,
+      "loss": 0.8168,
+      "step": 5610
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 1.4402390438247014e-05,
+      "loss": 0.7334,
+      "step": 5620
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 1.4392430278884465e-05,
+      "loss": 0.6208,
+      "step": 5630
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.4382470119521913e-05,
+      "loss": 0.5499,
+      "step": 5640
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.4372509960159364e-05,
+      "loss": 0.8342,
+      "step": 5650
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.4362549800796814e-05,
+      "loss": 0.6332,
+      "step": 5660
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.4352589641434265e-05,
+      "loss": 0.9666,
+      "step": 5670
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.4342629482071715e-05,
+      "loss": 0.8214,
+      "step": 5680
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 1.4332669322709166e-05,
+      "loss": 0.6709,
+      "step": 5690
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4322709163346615e-05,
+      "loss": 0.5711,
+      "step": 5700
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.4312749003984065e-05,
+      "loss": 0.6811,
+      "step": 5710
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.4302788844621516e-05,
+      "loss": 0.5967,
+      "step": 5720
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 1.4292828685258966e-05,
+      "loss": 0.758,
+      "step": 5730
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.4282868525896417e-05,
+      "loss": 0.7305,
+      "step": 5740
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.4272908366533866e-05,
+      "loss": 0.6534,
+      "step": 5750
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 1.4262948207171316e-05,
+      "loss": 0.6304,
+      "step": 5760
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 1.4252988047808767e-05,
+      "loss": 0.7935,
+      "step": 5770
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 1.4243027888446217e-05,
+      "loss": 0.6772,
+      "step": 5780
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 1.4233067729083668e-05,
+      "loss": 0.7283,
+      "step": 5790
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 1.4223107569721118e-05,
+      "loss": 0.6191,
+      "step": 5800
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 1.4213147410358567e-05,
+      "loss": 0.6432,
+      "step": 5810
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 1.4203187250996018e-05,
+      "loss": 0.7632,
+      "step": 5820
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 1.4193227091633468e-05,
+      "loss": 0.9391,
+      "step": 5830
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 1.4183266932270919e-05,
+      "loss": 1.006,
+      "step": 5840
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 1.4173306772908369e-05,
+      "loss": 0.7879,
+      "step": 5850
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 1.4163346613545818e-05,
+      "loss": 0.7906,
+      "step": 5860
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 1.4153386454183268e-05,
+      "loss": 0.7283,
+      "step": 5870
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 1.4143426294820719e-05,
+      "loss": 0.673,
+      "step": 5880
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 1.413346613545817e-05,
+      "loss": 0.6293,
+      "step": 5890
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.412350597609562e-05,
+      "loss": 0.7544,
+      "step": 5900
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 1.411354581673307e-05,
+      "loss": 0.5892,
+      "step": 5910
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.410358565737052e-05,
+      "loss": 0.7518,
+      "step": 5920
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.409362549800797e-05,
+      "loss": 0.6958,
+      "step": 5930
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.408366533864542e-05,
+      "loss": 0.7782,
+      "step": 5940
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.407370517928287e-05,
+      "loss": 0.5405,
+      "step": 5950
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.4063745019920321e-05,
+      "loss": 0.6358,
+      "step": 5960
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.405378486055777e-05,
+      "loss": 0.7265,
+      "step": 5970
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.404382470119522e-05,
+      "loss": 0.7709,
+      "step": 5980
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.4033864541832671e-05,
+      "loss": 0.6181,
+      "step": 5990
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.4023904382470122e-05,
+      "loss": 0.7159,
+      "step": 6000
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 1.4013944223107572e-05,
+      "loss": 0.8848,
+      "step": 6010
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.4003984063745023e-05,
+      "loss": 0.8128,
+      "step": 6020
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.7667607621736062,
+      "eval_loss": 0.5767571330070496,
+      "eval_runtime": 12.4014,
+      "eval_samples_per_second": 228.523,
+      "eval_steps_per_second": 28.626,
+      "step": 6024
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.3994023904382471e-05,
+      "loss": 0.6871,
+      "step": 6030
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 1.3984063745019922e-05,
+      "loss": 0.6732,
+      "step": 6040
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 1.3974103585657372e-05,
+      "loss": 0.8479,
+      "step": 6050
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 1.3964143426294823e-05,
+      "loss": 0.802,
+      "step": 6060
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 1.3954183266932273e-05,
+      "loss": 0.8159,
+      "step": 6070
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 1.3944223107569724e-05,
+      "loss": 0.537,
+      "step": 6080
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 1.3934262948207173e-05,
+      "loss": 0.7607,
+      "step": 6090
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 1.3924302788844623e-05,
+      "loss": 0.6537,
+      "step": 6100
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 1.3914342629482074e-05,
+      "loss": 0.4504,
+      "step": 6110
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 1.3904382470119524e-05,
+      "loss": 0.7238,
+      "step": 6120
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 1.3894422310756975e-05,
+      "loss": 0.6512,
+      "step": 6130
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 1.3884462151394424e-05,
+      "loss": 0.8828,
+      "step": 6140
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 1.3874501992031874e-05,
+      "loss": 0.6161,
+      "step": 6150
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 1.3864541832669325e-05,
+      "loss": 0.6594,
+      "step": 6160
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 1.3854581673306775e-05,
+      "loss": 0.657,
+      "step": 6170
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 1.3844621513944226e-05,
+      "loss": 0.6383,
+      "step": 6180
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 1.3834661354581676e-05,
+      "loss": 0.7496,
+      "step": 6190
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 1.3824701195219125e-05,
+      "loss": 0.7393,
+      "step": 6200
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 1.3814741035856576e-05,
+      "loss": 0.7018,
+      "step": 6210
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 1.3804780876494026e-05,
+      "loss": 0.7096,
+      "step": 6220
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 1.3794820717131477e-05,
+      "loss": 0.5861,
+      "step": 6230
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 1.3784860557768927e-05,
+      "loss": 0.8608,
+      "step": 6240
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 1.3774900398406376e-05,
+      "loss": 0.6554,
+      "step": 6250
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 1.3764940239043826e-05,
+      "loss": 0.5953,
+      "step": 6260
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 1.3754980079681277e-05,
+      "loss": 0.6316,
+      "step": 6270
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 1.3745019920318726e-05,
+      "loss": 0.5364,
+      "step": 6280
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 1.3735059760956175e-05,
+      "loss": 0.8754,
+      "step": 6290
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 1.3725099601593625e-05,
+      "loss": 0.5543,
+      "step": 6300
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 1.3715139442231076e-05,
+      "loss": 0.4965,
+      "step": 6310
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 1.3705179282868526e-05,
+      "loss": 0.8246,
+      "step": 6320
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 1.3695219123505977e-05,
+      "loss": 0.7588,
+      "step": 6330
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 1.3685258964143427e-05,
+      "loss": 0.7063,
+      "step": 6340
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 1.3675298804780876e-05,
+      "loss": 0.6032,
+      "step": 6350
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 1.3665338645418326e-05,
+      "loss": 0.5508,
+      "step": 6360
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 1.3655378486055777e-05,
+      "loss": 0.6647,
+      "step": 6370
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 1.3645418326693227e-05,
+      "loss": 0.8579,
+      "step": 6380
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 1.3635458167330678e-05,
+      "loss": 0.6804,
+      "step": 6390
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 1.3625498007968127e-05,
+      "loss": 0.5512,
+      "step": 6400
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 1.3615537848605577e-05,
+      "loss": 0.5801,
+      "step": 6410
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 1.3605577689243028e-05,
+      "loss": 0.5298,
+      "step": 6420
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 1.3595617529880478e-05,
+      "loss": 0.8435,
+      "step": 6430
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 1.3585657370517929e-05,
+      "loss": 0.8138,
+      "step": 6440
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 1.357569721115538e-05,
+      "loss": 0.614,
+      "step": 6450
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 1.3565737051792828e-05,
+      "loss": 0.7013,
+      "step": 6460
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 1.3555776892430279e-05,
+      "loss": 0.7499,
+      "step": 6470
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 1.354581673306773e-05,
+      "loss": 0.6637,
+      "step": 6480
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 1.353585657370518e-05,
+      "loss": 0.787,
+      "step": 6490
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 1.352589641434263e-05,
+      "loss": 0.8298,
+      "step": 6500
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 1.351593625498008e-05,
+      "loss": 0.7632,
+      "step": 6510
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 1.350597609561753e-05,
+      "loss": 0.5762,
+      "step": 6520
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 1.349601593625498e-05,
+      "loss": 0.7177,
+      "step": 6530
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 1.348605577689243e-05,
+      "loss": 0.7018,
+      "step": 6540
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 1.3476095617529881e-05,
+      "loss": 0.7508,
+      "step": 6550
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 1.3466135458167332e-05,
+      "loss": 0.5311,
+      "step": 6560
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 1.345617529880478e-05,
+      "loss": 0.6881,
+      "step": 6570
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 1.3446215139442231e-05,
+      "loss": 0.5727,
+      "step": 6580
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 1.3436254980079681e-05,
+      "loss": 0.8187,
+      "step": 6590
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 1.3426294820717132e-05,
+      "loss": 0.5865,
+      "step": 6600
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 1.3416334661354582e-05,
+      "loss": 0.6394,
+      "step": 6610
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 1.3406374501992033e-05,
+      "loss": 0.6118,
+      "step": 6620
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 1.3396414342629482e-05,
+      "loss": 0.7971,
+      "step": 6630
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 1.3386454183266932e-05,
+      "loss": 0.6505,
+      "step": 6640
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 1.3376494023904383e-05,
+      "loss": 0.6598,
+      "step": 6650
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 1.3366533864541833e-05,
+      "loss": 0.6268,
+      "step": 6660
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 1.3356573705179284e-05,
+      "loss": 0.6986,
+      "step": 6670
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 1.3346613545816733e-05,
+      "loss": 0.6067,
+      "step": 6680
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 1.3336653386454183e-05,
+      "loss": 0.7246,
+      "step": 6690
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.3326693227091634e-05,
+      "loss": 0.5813,
+      "step": 6700
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.3316733067729084e-05,
+      "loss": 0.8305,
+      "step": 6710
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.3306772908366535e-05,
+      "loss": 0.6774,
+      "step": 6720
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 1.3296812749003985e-05,
+      "loss": 0.6053,
+      "step": 6730
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.3286852589641434e-05,
+      "loss": 0.6385,
+      "step": 6740
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.3276892430278884e-05,
+      "loss": 0.678,
+      "step": 6750
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.3266932270916335e-05,
+      "loss": 0.6764,
+      "step": 6760
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 1.3256972111553786e-05,
+      "loss": 0.8155,
+      "step": 6770
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.3247011952191236e-05,
+      "loss": 0.6272,
+      "step": 6780
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.3237051792828685e-05,
+      "loss": 0.663,
+      "step": 6790
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.3227091633466135e-05,
+      "loss": 0.6421,
+      "step": 6800
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 1.3217131474103586e-05,
+      "loss": 0.6271,
+      "step": 6810
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.3207171314741036e-05,
+      "loss": 0.7884,
+      "step": 6820
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 1.3197211155378487e-05,
+      "loss": 0.7534,
+      "step": 6830
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.3187250996015937e-05,
+      "loss": 0.7807,
+      "step": 6840
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 1.3177290836653386e-05,
+      "loss": 0.6913,
+      "step": 6850
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.3167330677290837e-05,
+      "loss": 0.5774,
+      "step": 6860
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 1.3157370517928287e-05,
+      "loss": 0.5888,
+      "step": 6870
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.3147410358565738e-05,
+      "loss": 0.6487,
+      "step": 6880
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 1.3137450199203188e-05,
+      "loss": 0.887,
+      "step": 6890
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.3127490039840639e-05,
+      "loss": 0.5255,
+      "step": 6900
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 1.3117529880478088e-05,
+      "loss": 0.6181,
+      "step": 6910
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.3107569721115538e-05,
+      "loss": 0.6911,
+      "step": 6920
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 1.3097609561752989e-05,
+      "loss": 0.824,
+      "step": 6930
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.3087649402390439e-05,
+      "loss": 0.5866,
+      "step": 6940
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 1.307768924302789e-05,
+      "loss": 0.8112,
+      "step": 6950
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.3067729083665338e-05,
+      "loss": 0.5409,
+      "step": 6960
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 1.3057768924302789e-05,
+      "loss": 0.6594,
+      "step": 6970
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.304780876494024e-05,
+      "loss": 0.7699,
+      "step": 6980
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 1.303784860557769e-05,
+      "loss": 0.8822,
+      "step": 6990
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.302788844621514e-05,
+      "loss": 0.6163,
+      "step": 7000
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 1.3017928286852591e-05,
+      "loss": 0.6812,
+      "step": 7010
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.300796812749004e-05,
+      "loss": 0.662,
+      "step": 7020
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 1.299800796812749e-05,
+      "loss": 0.5568,
+      "step": 7030
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.298804780876494e-05,
+      "loss": 0.8407,
+      "step": 7040
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 1.2978087649402391e-05,
+      "loss": 0.9296,
+      "step": 7050
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.2968127490039842e-05,
+      "loss": 0.7527,
+      "step": 7060
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 1.295816733067729e-05,
+      "loss": 0.4591,
+      "step": 7070
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.2948207171314741e-05,
+      "loss": 0.5224,
+      "step": 7080
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 1.2938247011952192e-05,
+      "loss": 0.6344,
+      "step": 7090
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.2928286852589642e-05,
+      "loss": 0.7245,
+      "step": 7100
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 1.2918326693227093e-05,
+      "loss": 0.719,
+      "step": 7110
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.2908366533864543e-05,
+      "loss": 0.7353,
+      "step": 7120
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 1.2898406374501992e-05,
+      "loss": 0.7411,
+      "step": 7130
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.2888446215139443e-05,
+      "loss": 0.811,
+      "step": 7140
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 1.2878486055776893e-05,
+      "loss": 0.716,
+      "step": 7150
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.2868525896414344e-05,
+      "loss": 0.6672,
+      "step": 7160
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 1.2858565737051794e-05,
+      "loss": 0.6952,
+      "step": 7170
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.2848605577689245e-05,
+      "loss": 0.6989,
+      "step": 7180
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 1.2838645418326693e-05,
+      "loss": 0.7464,
+      "step": 7190
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.2828685258964144e-05,
+      "loss": 0.6873,
+      "step": 7200
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 1.2818725099601594e-05,
+      "loss": 0.6019,
+      "step": 7210
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.2808764940239045e-05,
+      "loss": 0.6366,
+      "step": 7220
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 1.2798804780876495e-05,
+      "loss": 0.6065,
+      "step": 7230
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.2788844621513944e-05,
+      "loss": 0.7833,
+      "step": 7240
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 1.2778884462151395e-05,
+      "loss": 0.7096,
+      "step": 7250
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.2768924302788845e-05,
+      "loss": 0.5601,
+      "step": 7260
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 1.2758964143426296e-05,
+      "loss": 0.6017,
+      "step": 7270
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.2749003984063746e-05,
+      "loss": 0.8801,
+      "step": 7280
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 1.2739043824701197e-05,
+      "loss": 0.6946,
+      "step": 7290
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.2729083665338646e-05,
+      "loss": 0.6949,
+      "step": 7300
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 1.2719123505976096e-05,
+      "loss": 0.5858,
+      "step": 7310
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.2709163346613547e-05,
+      "loss": 0.5958,
+      "step": 7320
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 1.2699203187250997e-05,
+      "loss": 0.7674,
+      "step": 7330
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.2689243027888448e-05,
+      "loss": 0.6512,
+      "step": 7340
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 1.2679282868525896e-05,
+      "loss": 0.5964,
+      "step": 7350
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.2669322709163347e-05,
+      "loss": 0.686,
+      "step": 7360
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 1.2659362549800798e-05,
+      "loss": 0.7152,
+      "step": 7370
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.2649402390438248e-05,
+      "loss": 0.6055,
+      "step": 7380
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.2639442231075699e-05,
+      "loss": 0.7611,
+      "step": 7390
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.2629482071713149e-05,
+      "loss": 0.6743,
+      "step": 7400
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.2619521912350598e-05,
+      "loss": 0.6299,
+      "step": 7410
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.2609561752988048e-05,
+      "loss": 0.9105,
+      "step": 7420
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.2599601593625499e-05,
+      "loss": 0.7445,
+      "step": 7430
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.258964143426295e-05,
+      "loss": 0.583,
+      "step": 7440
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.25796812749004e-05,
+      "loss": 0.758,
+      "step": 7450
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.256972111553785e-05,
+      "loss": 0.6762,
+      "step": 7460
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.25597609561753e-05,
+      "loss": 0.5847,
+      "step": 7470
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.254980079681275e-05,
+      "loss": 0.5937,
+      "step": 7480
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.25398406374502e-05,
+      "loss": 0.864,
+      "step": 7490
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.252988047808765e-05,
+      "loss": 0.8127,
+      "step": 7500
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.2519920318725101e-05,
+      "loss": 0.751,
+      "step": 7510
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.250996015936255e-05,
+      "loss": 0.735,
+      "step": 7520
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.25e-05,
+      "loss": 0.7009,
+      "step": 7530
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.2490039840637451e-05,
+      "loss": 0.5561,
+      "step": 7540
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.2480079681274902e-05,
+      "loss": 0.714,
+      "step": 7550
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.2470119521912352e-05,
+      "loss": 0.7095,
+      "step": 7560
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.2460159362549803e-05,
+      "loss": 0.7138,
+      "step": 7570
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.2450199203187251e-05,
+      "loss": 0.5978,
+      "step": 7580
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.2440239043824702e-05,
+      "loss": 0.8403,
+      "step": 7590
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.2430278884462152e-05,
+      "loss": 0.6188,
+      "step": 7600
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.2420318725099603e-05,
+      "loss": 0.8506,
+      "step": 7610
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.2410358565737053e-05,
+      "loss": 0.577,
+      "step": 7620
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.2400398406374502e-05,
+      "loss": 0.7674,
+      "step": 7630
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.2390438247011953e-05,
+      "loss": 0.6315,
+      "step": 7640
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.2380478087649403e-05,
+      "loss": 0.5692,
+      "step": 7650
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.2370517928286854e-05,
+      "loss": 0.7386,
+      "step": 7660
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.2360557768924304e-05,
+      "loss": 0.6458,
+      "step": 7670
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.2350597609561755e-05,
+      "loss": 0.6332,
+      "step": 7680
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.2340637450199204e-05,
+      "loss": 0.6986,
+      "step": 7690
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.2330677290836654e-05,
+      "loss": 0.619,
+      "step": 7700
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.2320717131474105e-05,
+      "loss": 0.5602,
+      "step": 7710
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.2310756972111555e-05,
+      "loss": 0.7717,
+      "step": 7720
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.2300796812749006e-05,
+      "loss": 0.6981,
+      "step": 7730
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.2290836653386455e-05,
+      "loss": 0.6402,
+      "step": 7740
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.2280876494023905e-05,
+      "loss": 0.7541,
+      "step": 7750
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.2270916334661356e-05,
+      "loss": 0.6464,
+      "step": 7760
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.2260956175298806e-05,
+      "loss": 0.5339,
+      "step": 7770
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.2250996015936257e-05,
+      "loss": 0.5893,
+      "step": 7780
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.2241035856573707e-05,
+      "loss": 0.8025,
+      "step": 7790
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.2231075697211156e-05,
+      "loss": 0.597,
+      "step": 7800
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.2221115537848606e-05,
+      "loss": 0.6974,
+      "step": 7810
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.2211155378486057e-05,
+      "loss": 0.7521,
+      "step": 7820
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.2201195219123507e-05,
+      "loss": 0.6875,
+      "step": 7830
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.2191235059760958e-05,
+      "loss": 0.6489,
+      "step": 7840
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.2181274900398408e-05,
+      "loss": 0.4919,
+      "step": 7850
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.2171314741035857e-05,
+      "loss": 0.8321,
+      "step": 7860
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.2161354581673308e-05,
+      "loss": 0.7614,
+      "step": 7870
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.2151394422310758e-05,
+      "loss": 0.627,
+      "step": 7880
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.2141434262948209e-05,
+      "loss": 0.6162,
+      "step": 7890
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.213147410358566e-05,
+      "loss": 0.6564,
+      "step": 7900
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.2121513944223108e-05,
+      "loss": 0.5367,
+      "step": 7910
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.2111553784860559e-05,
+      "loss": 0.5283,
+      "step": 7920
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.210159362549801e-05,
+      "loss": 0.7426,
+      "step": 7930
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.209163346613546e-05,
+      "loss": 0.4438,
+      "step": 7940
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.208167330677291e-05,
+      "loss": 0.7928,
+      "step": 7950
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.207171314741036e-05,
+      "loss": 0.5781,
+      "step": 7960
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.206175298804781e-05,
+      "loss": 0.6955,
+      "step": 7970
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.205179282868526e-05,
+      "loss": 0.5313,
+      "step": 7980
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.204183266932271e-05,
+      "loss": 0.573,
+      "step": 7990
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.2031872509960161e-05,
+      "loss": 0.6761,
+      "step": 8000
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.2021912350597612e-05,
+      "loss": 0.7084,
+      "step": 8010
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.201195219123506e-05,
+      "loss": 0.8027,
+      "step": 8020
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.2001992031872511e-05,
+      "loss": 0.7611,
+      "step": 8030
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.7833450952717008,
+      "eval_loss": 0.554111659526825,
+      "eval_runtime": 12.3805,
+      "eval_samples_per_second": 228.908,
+      "eval_steps_per_second": 28.674,
+      "step": 8032
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.1992031872509961e-05,
+      "loss": 0.6747,
+      "step": 8040
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.1982071713147412e-05,
+      "loss": 0.9124,
+      "step": 8050
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.1972111553784862e-05,
+      "loss": 0.5193,
+      "step": 8060
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.1962151394422313e-05,
+      "loss": 0.6755,
+      "step": 8070
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.1952191235059762e-05,
+      "loss": 0.5438,
+      "step": 8080
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.1942231075697212e-05,
+      "loss": 0.5929,
+      "step": 8090
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.1932270916334663e-05,
+      "loss": 0.5668,
+      "step": 8100
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.1922310756972113e-05,
+      "loss": 0.5961,
+      "step": 8110
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.1912350597609564e-05,
+      "loss": 0.6437,
+      "step": 8120
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.1902390438247014e-05,
+      "loss": 0.9184,
+      "step": 8130
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.1892430278884463e-05,
+      "loss": 0.6197,
+      "step": 8140
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.1882470119521914e-05,
+      "loss": 0.7642,
+      "step": 8150
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.1872509960159364e-05,
+      "loss": 0.6014,
+      "step": 8160
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.1862549800796815e-05,
+      "loss": 0.5701,
+      "step": 8170
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.1852589641434265e-05,
+      "loss": 0.7278,
+      "step": 8180
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.1842629482071714e-05,
+      "loss": 0.6398,
+      "step": 8190
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.1832669322709164e-05,
+      "loss": 0.5999,
+      "step": 8200
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.1822709163346615e-05,
+      "loss": 0.6651,
+      "step": 8210
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.1812749003984065e-05,
+      "loss": 0.615,
+      "step": 8220
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.1802788844621516e-05,
+      "loss": 0.5668,
+      "step": 8230
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.1792828685258967e-05,
+      "loss": 0.6382,
+      "step": 8240
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.1782868525896415e-05,
+      "loss": 0.6822,
+      "step": 8250
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.1772908366533866e-05,
+      "loss": 0.7294,
+      "step": 8260
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.1762948207171316e-05,
+      "loss": 0.7187,
+      "step": 8270
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.1752988047808767e-05,
+      "loss": 0.8094,
+      "step": 8280
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.1743027888446217e-05,
+      "loss": 0.706,
+      "step": 8290
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.1733067729083666e-05,
+      "loss": 0.5308,
+      "step": 8300
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.1723107569721117e-05,
+      "loss": 0.536,
+      "step": 8310
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.1713147410358567e-05,
+      "loss": 0.5415,
+      "step": 8320
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.1703187250996018e-05,
+      "loss": 0.7349,
+      "step": 8330
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.1693227091633468e-05,
+      "loss": 0.5349,
+      "step": 8340
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.1683266932270919e-05,
+      "loss": 0.694,
+      "step": 8350
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.1673306772908368e-05,
+      "loss": 0.7136,
+      "step": 8360
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.1663346613545818e-05,
+      "loss": 0.5287,
+      "step": 8370
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.1653386454183269e-05,
+      "loss": 0.8524,
+      "step": 8380
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.1643426294820719e-05,
+      "loss": 0.6588,
+      "step": 8390
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.163346613545817e-05,
+      "loss": 0.6301,
+      "step": 8400
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.1623505976095618e-05,
+      "loss": 0.7839,
+      "step": 8410
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.1613545816733069e-05,
+      "loss": 0.4903,
+      "step": 8420
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.160358565737052e-05,
+      "loss": 0.6657,
+      "step": 8430
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.159362549800797e-05,
+      "loss": 0.5542,
+      "step": 8440
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.158366533864542e-05,
+      "loss": 0.5381,
+      "step": 8450
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.1573705179282871e-05,
+      "loss": 0.4691,
+      "step": 8460
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.156374501992032e-05,
+      "loss": 0.7469,
+      "step": 8470
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.155378486055777e-05,
+      "loss": 0.6524,
+      "step": 8480
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.154382470119522e-05,
+      "loss": 0.6606,
+      "step": 8490
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.1533864541832671e-05,
+      "loss": 0.8643,
+      "step": 8500
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.1523904382470122e-05,
+      "loss": 0.7505,
+      "step": 8510
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.1513944223107572e-05,
+      "loss": 0.8198,
+      "step": 8520
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.1503984063745021e-05,
+      "loss": 0.5468,
+      "step": 8530
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.1494023904382472e-05,
+      "loss": 0.5666,
+      "step": 8540
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.1484063745019922e-05,
+      "loss": 0.6286,
+      "step": 8550
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.1474103585657373e-05,
+      "loss": 0.6504,
+      "step": 8560
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.1464143426294823e-05,
+      "loss": 0.5713,
+      "step": 8570
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.1454183266932272e-05,
+      "loss": 0.501,
+      "step": 8580
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.1444223107569723e-05,
+      "loss": 0.6122,
+      "step": 8590
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.1434262948207173e-05,
+      "loss": 0.625,
+      "step": 8600
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.1424302788844624e-05,
+      "loss": 0.4513,
+      "step": 8610
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.1414342629482074e-05,
+      "loss": 0.7724,
+      "step": 8620
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.1404382470119525e-05,
+      "loss": 0.6723,
+      "step": 8630
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.1394422310756973e-05,
+      "loss": 0.6128,
+      "step": 8640
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.1384462151394424e-05,
+      "loss": 0.5595,
+      "step": 8650
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.1374501992031874e-05,
+      "loss": 0.5553,
+      "step": 8660
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.1364541832669325e-05,
+      "loss": 0.656,
+      "step": 8670
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.1354581673306775e-05,
+      "loss": 0.6754,
+      "step": 8680
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.1344621513944224e-05,
+      "loss": 0.5423,
+      "step": 8690
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.1334661354581675e-05,
+      "loss": 0.5632,
+      "step": 8700
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.1324701195219125e-05,
+      "loss": 0.5655,
+      "step": 8710
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.1314741035856576e-05,
+      "loss": 0.5476,
+      "step": 8720
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.1304780876494026e-05,
+      "loss": 0.6249,
+      "step": 8730
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 1.1294820717131477e-05,
+      "loss": 0.5834,
+      "step": 8740
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.1284860557768926e-05,
+      "loss": 0.5972,
+      "step": 8750
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 1.1274900398406376e-05,
+      "loss": 0.6231,
+      "step": 8760
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.1264940239043827e-05,
+      "loss": 0.5717,
+      "step": 8770
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 1.1254980079681277e-05,
+      "loss": 0.5073,
+      "step": 8780
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.1245019920318724e-05,
+      "loss": 0.6835,
+      "step": 8790
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 1.1235059760956175e-05,
+      "loss": 0.6284,
+      "step": 8800
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.1225099601593625e-05,
+      "loss": 0.6213,
+      "step": 8810
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 1.1215139442231076e-05,
+      "loss": 0.659,
+      "step": 8820
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.1205179282868526e-05,
+      "loss": 0.4998,
+      "step": 8830
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 1.1195219123505975e-05,
+      "loss": 0.7572,
+      "step": 8840
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.1185258964143426e-05,
+      "loss": 0.5937,
+      "step": 8850
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 1.1175298804780876e-05,
+      "loss": 0.4928,
+      "step": 8860
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.1165338645418327e-05,
+      "loss": 0.5478,
+      "step": 8870
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 1.1155378486055777e-05,
+      "loss": 0.737,
+      "step": 8880
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.1145418326693228e-05,
+      "loss": 0.6778,
+      "step": 8890
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 1.1135458167330676e-05,
+      "loss": 0.6994,
+      "step": 8900
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.1125498007968127e-05,
+      "loss": 0.668,
+      "step": 8910
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 1.1115537848605578e-05,
+      "loss": 0.6935,
+      "step": 8920
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.1105577689243028e-05,
+      "loss": 0.5602,
+      "step": 8930
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 1.1095617529880479e-05,
+      "loss": 0.7141,
+      "step": 8940
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.1085657370517929e-05,
+      "loss": 0.5478,
+      "step": 8950
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 1.1075697211155378e-05,
+      "loss": 0.581,
+      "step": 8960
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.1065737051792828e-05,
+      "loss": 0.7616,
+      "step": 8970
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 1.1055776892430279e-05,
+      "loss": 0.5553,
+      "step": 8980
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.104581673306773e-05,
+      "loss": 0.6461,
+      "step": 8990
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 1.103585657370518e-05,
+      "loss": 0.662,
+      "step": 9000
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.1025896414342629e-05,
+      "loss": 0.544,
+      "step": 9010
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 1.101593625498008e-05,
+      "loss": 0.6106,
+      "step": 9020
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.100597609561753e-05,
+      "loss": 0.6738,
+      "step": 9030
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 1.099601593625498e-05,
+      "loss": 0.6346,
+      "step": 9040
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.098605577689243e-05,
+      "loss": 0.5935,
+      "step": 9050
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 1.0976095617529881e-05,
+      "loss": 0.6276,
+      "step": 9060
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.096613545816733e-05,
+      "loss": 0.6324,
+      "step": 9070
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 1.095617529880478e-05,
+      "loss": 0.6854,
+      "step": 9080
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.0946215139442231e-05,
+      "loss": 0.5761,
+      "step": 9090
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 1.0936254980079682e-05,
+      "loss": 0.5624,
+      "step": 9100
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.0926294820717132e-05,
+      "loss": 0.6294,
+      "step": 9110
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 1.0916334661354581e-05,
+      "loss": 0.513,
+      "step": 9120
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.0906374501992031e-05,
+      "loss": 0.484,
+      "step": 9130
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 1.0896414342629482e-05,
+      "loss": 0.8093,
+      "step": 9140
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.0886454183266932e-05,
+      "loss": 0.6066,
+      "step": 9150
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 1.0876494023904383e-05,
+      "loss": 0.572,
+      "step": 9160
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.0866533864541834e-05,
+      "loss": 0.8022,
+      "step": 9170
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 1.0856573705179282e-05,
+      "loss": 0.4559,
+      "step": 9180
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.0846613545816733e-05,
+      "loss": 0.8481,
+      "step": 9190
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 1.0836653386454183e-05,
+      "loss": 0.7148,
+      "step": 9200
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.0826693227091634e-05,
+      "loss": 0.5483,
+      "step": 9210
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 1.0816733067729084e-05,
+      "loss": 0.722,
+      "step": 9220
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.0806772908366535e-05,
+      "loss": 0.4038,
+      "step": 9230
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 1.0796812749003984e-05,
+      "loss": 0.8075,
+      "step": 9240
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.0786852589641434e-05,
+      "loss": 0.7069,
+      "step": 9250
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 1.0776892430278885e-05,
+      "loss": 0.5291,
+      "step": 9260
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.0766932270916335e-05,
+      "loss": 0.549,
+      "step": 9270
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 1.0756972111553786e-05,
+      "loss": 0.5375,
+      "step": 9280
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.0747011952191235e-05,
+      "loss": 0.689,
+      "step": 9290
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 1.0737051792828685e-05,
+      "loss": 0.5831,
+      "step": 9300
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.0727091633466136e-05,
+      "loss": 0.7251,
+      "step": 9310
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 1.0717131474103586e-05,
+      "loss": 0.5846,
+      "step": 9320
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.0707171314741037e-05,
+      "loss": 0.6294,
+      "step": 9330
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 1.0697211155378487e-05,
+      "loss": 0.4381,
+      "step": 9340
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.0687250996015936e-05,
+      "loss": 0.5542,
+      "step": 9350
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 1.0677290836653386e-05,
+      "loss": 0.6181,
+      "step": 9360
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.0667330677290837e-05,
+      "loss": 0.618,
+      "step": 9370
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 1.0657370517928287e-05,
+      "loss": 0.5858,
+      "step": 9380
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.0647410358565738e-05,
+      "loss": 0.5361,
+      "step": 9390
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 1.0637450199203187e-05,
+      "loss": 0.9741,
+      "step": 9400
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.0627490039840637e-05,
+      "loss": 0.6503,
+      "step": 9410
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 1.0617529880478088e-05,
+      "loss": 0.5861,
+      "step": 9420
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.0607569721115538e-05,
+      "loss": 0.6078,
+      "step": 9430
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 1.0597609561752989e-05,
+      "loss": 0.6741,
+      "step": 9440
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.058764940239044e-05,
+      "loss": 0.5918,
+      "step": 9450
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 1.0577689243027888e-05,
+      "loss": 0.8044,
+      "step": 9460
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.0567729083665339e-05,
+      "loss": 0.6268,
+      "step": 9470
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 1.055776892430279e-05,
+      "loss": 0.6959,
+      "step": 9480
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.054780876494024e-05,
+      "loss": 0.6308,
+      "step": 9490
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 1.053784860557769e-05,
+      "loss": 0.5799,
+      "step": 9500
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.0527888446215139e-05,
+      "loss": 0.6327,
+      "step": 9510
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 1.051792828685259e-05,
+      "loss": 0.5992,
+      "step": 9520
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.050796812749004e-05,
+      "loss": 0.6713,
+      "step": 9530
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 1.049800796812749e-05,
+      "loss": 0.612,
+      "step": 9540
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.0488047808764941e-05,
+      "loss": 0.5239,
+      "step": 9550
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 1.0478087649402392e-05,
+      "loss": 0.5259,
+      "step": 9560
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.046812749003984e-05,
+      "loss": 0.5947,
+      "step": 9570
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 1.0458167330677291e-05,
+      "loss": 0.5616,
+      "step": 9580
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.0448207171314741e-05,
+      "loss": 0.6224,
+      "step": 9590
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 1.0438247011952192e-05,
+      "loss": 0.5434,
+      "step": 9600
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.0428286852589642e-05,
+      "loss": 0.7836,
+      "step": 9610
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 1.0418326693227093e-05,
+      "loss": 0.5701,
+      "step": 9620
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.0408366533864542e-05,
+      "loss": 0.4821,
+      "step": 9630
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 1.0398406374501992e-05,
+      "loss": 0.6637,
+      "step": 9640
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.0388446215139443e-05,
+      "loss": 0.6768,
+      "step": 9650
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 1.0378486055776893e-05,
+      "loss": 0.7112,
+      "step": 9660
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.0368525896414344e-05,
+      "loss": 0.6812,
+      "step": 9670
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 1.0358565737051793e-05,
+      "loss": 0.7659,
+      "step": 9680
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.0348605577689243e-05,
+      "loss": 0.7142,
+      "step": 9690
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 1.0338645418326694e-05,
+      "loss": 0.385,
+      "step": 9700
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.0328685258964144e-05,
+      "loss": 0.6218,
+      "step": 9710
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 1.0318725099601595e-05,
+      "loss": 0.6973,
+      "step": 9720
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.0308764940239045e-05,
+      "loss": 0.5339,
+      "step": 9730
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 1.0298804780876494e-05,
+      "loss": 0.5088,
+      "step": 9740
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.0288844621513944e-05,
+      "loss": 0.5024,
+      "step": 9750
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 1.0278884462151395e-05,
+      "loss": 0.497,
+      "step": 9760
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.0268924302788846e-05,
+      "loss": 0.5011,
+      "step": 9770
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.0258964143426296e-05,
+      "loss": 0.5837,
+      "step": 9780
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.0249003984063745e-05,
+      "loss": 0.6643,
+      "step": 9790
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.0239043824701195e-05,
+      "loss": 0.6562,
+      "step": 9800
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.0229083665338646e-05,
+      "loss": 0.844,
+      "step": 9810
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.0219123505976096e-05,
+      "loss": 0.5727,
+      "step": 9820
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.0209163346613547e-05,
+      "loss": 0.4428,
+      "step": 9830
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.0199203187250997e-05,
+      "loss": 0.6486,
+      "step": 9840
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.0189243027888446e-05,
+      "loss": 0.7022,
+      "step": 9850
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.0179282868525897e-05,
+      "loss": 0.6975,
+      "step": 9860
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.0169322709163347e-05,
+      "loss": 0.4616,
+      "step": 9870
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.0159362549800798e-05,
+      "loss": 0.6477,
+      "step": 9880
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.0149402390438248e-05,
+      "loss": 0.5991,
+      "step": 9890
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.0139442231075699e-05,
+      "loss": 0.593,
+      "step": 9900
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.0129482071713148e-05,
+      "loss": 0.5538,
+      "step": 9910
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.0119521912350598e-05,
+      "loss": 0.7266,
+      "step": 9920
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.0109561752988049e-05,
+      "loss": 0.6822,
+      "step": 9930
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 1.0099601593625499e-05,
+      "loss": 0.6699,
+      "step": 9940
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.008964143426295e-05,
+      "loss": 0.4431,
+      "step": 9950
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 1.0079681274900398e-05,
+      "loss": 0.703,
+      "step": 9960
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.0069721115537849e-05,
+      "loss": 0.7132,
+      "step": 9970
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 1.00597609561753e-05,
+      "loss": 0.5608,
+      "step": 9980
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.004980079681275e-05,
+      "loss": 0.6622,
+      "step": 9990
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 1.00398406374502e-05,
+      "loss": 0.7245,
+      "step": 10000
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.0029880478087651e-05,
+      "loss": 0.6612,
+      "step": 10010
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.00199203187251e-05,
+      "loss": 0.5423,
+      "step": 10020
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.000996015936255e-05,
+      "loss": 0.6154,
+      "step": 10030
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1e-05,
+      "loss": 0.6441,
+      "step": 10040
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.7773465067043048,
+      "eval_loss": 0.5472666025161743,
+      "eval_runtime": 12.6285,
+      "eval_samples_per_second": 224.412,
+      "eval_steps_per_second": 28.111,
+      "step": 10040
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 9.990039840637451e-06,
+      "loss": 0.6501,
+      "step": 10050
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 9.980079681274902e-06,
+      "loss": 0.8181,
+      "step": 10060
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 9.97011952191235e-06,
+      "loss": 0.4814,
+      "step": 10070
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 9.960159362549801e-06,
+      "loss": 0.5222,
+      "step": 10080
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 9.950199203187252e-06,
+      "loss": 0.6022,
+      "step": 10090
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 9.940239043824702e-06,
+      "loss": 0.5172,
+      "step": 10100
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 9.930278884462153e-06,
+      "loss": 0.6522,
+      "step": 10110
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 9.920318725099603e-06,
+      "loss": 0.7355,
+      "step": 10120
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 9.910358565737052e-06,
+      "loss": 0.5913,
+      "step": 10130
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 9.900398406374503e-06,
+      "loss": 0.6335,
+      "step": 10140
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 9.890438247011953e-06,
+      "loss": 0.5895,
+      "step": 10150
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 9.880478087649404e-06,
+      "loss": 0.5145,
+      "step": 10160
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 9.870517928286854e-06,
+      "loss": 0.5867,
+      "step": 10170
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 9.860557768924303e-06,
+      "loss": 0.509,
+      "step": 10180
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 9.850597609561753e-06,
+      "loss": 0.6523,
+      "step": 10190
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 9.840637450199204e-06,
+      "loss": 0.5949,
+      "step": 10200
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 9.830677290836654e-06,
+      "loss": 0.6623,
+      "step": 10210
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 9.820717131474105e-06,
+      "loss": 0.6212,
+      "step": 10220
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 9.810756972111555e-06,
+      "loss": 0.624,
+      "step": 10230
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 9.800796812749004e-06,
+      "loss": 0.4706,
+      "step": 10240
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 9.790836653386455e-06,
+      "loss": 0.7216,
+      "step": 10250
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 9.780876494023905e-06,
+      "loss": 0.6191,
+      "step": 10260
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 9.770916334661356e-06,
+      "loss": 0.6242,
+      "step": 10270
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 9.760956175298806e-06,
+      "loss": 0.6594,
+      "step": 10280
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 9.750996015936257e-06,
+      "loss": 0.5346,
+      "step": 10290
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 9.741035856573706e-06,
+      "loss": 0.487,
+      "step": 10300
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 9.731075697211156e-06,
+      "loss": 0.688,
+      "step": 10310
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 9.721115537848607e-06,
+      "loss": 0.5676,
+      "step": 10320
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 9.711155378486057e-06,
+      "loss": 0.4808,
+      "step": 10330
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 9.701195219123508e-06,
+      "loss": 0.6092,
+      "step": 10340
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 9.691235059760956e-06,
+      "loss": 0.6759,
+      "step": 10350
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 9.681274900398407e-06,
+      "loss": 0.4499,
+      "step": 10360
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 9.671314741035858e-06,
+      "loss": 0.5645,
+      "step": 10370
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 9.661354581673308e-06,
+      "loss": 0.5436,
+      "step": 10380
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 9.651394422310759e-06,
+      "loss": 0.6921,
+      "step": 10390
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 9.641434262948209e-06,
+      "loss": 0.7377,
+      "step": 10400
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 9.631474103585658e-06,
+      "loss": 0.5902,
+      "step": 10410
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 9.621513944223108e-06,
+      "loss": 0.5209,
+      "step": 10420
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 9.611553784860559e-06,
+      "loss": 0.4325,
+      "step": 10430
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 9.60159362549801e-06,
+      "loss": 0.6181,
+      "step": 10440
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 9.59163346613546e-06,
+      "loss": 0.5895,
+      "step": 10450
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 9.581673306772909e-06,
+      "loss": 0.6064,
+      "step": 10460
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 9.57171314741036e-06,
+      "loss": 0.577,
+      "step": 10470
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 9.56175298804781e-06,
+      "loss": 0.5619,
+      "step": 10480
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 9.55179282868526e-06,
+      "loss": 0.541,
+      "step": 10490
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 9.54183266932271e-06,
+      "loss": 0.4666,
+      "step": 10500
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 9.531872509960161e-06,
+      "loss": 0.6817,
+      "step": 10510
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 9.52191235059761e-06,
+      "loss": 0.5872,
+      "step": 10520
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 9.51195219123506e-06,
+      "loss": 0.5357,
+      "step": 10530
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 9.501992031872511e-06,
+      "loss": 0.5895,
+      "step": 10540
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 9.492031872509962e-06,
+      "loss": 0.5624,
+      "step": 10550
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 9.482071713147412e-06,
+      "loss": 0.5557,
+      "step": 10560
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 9.472111553784863e-06,
+      "loss": 0.6568,
+      "step": 10570
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 9.462151394422311e-06,
+      "loss": 0.612,
+      "step": 10580
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 9.452191235059762e-06,
+      "loss": 0.535,
+      "step": 10590
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 9.442231075697212e-06,
+      "loss": 0.6675,
+      "step": 10600
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 9.432270916334663e-06,
+      "loss": 0.5896,
+      "step": 10610
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 9.422310756972113e-06,
+      "loss": 0.773,
+      "step": 10620
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 9.412350597609562e-06,
+      "loss": 0.6259,
+      "step": 10630
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 9.402390438247013e-06,
+      "loss": 0.6296,
+      "step": 10640
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 9.392430278884463e-06,
+      "loss": 0.6578,
+      "step": 10650
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 9.382470119521914e-06,
+      "loss": 0.6263,
+      "step": 10660
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 9.372509960159363e-06,
+      "loss": 0.6126,
+      "step": 10670
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 9.362549800796813e-06,
+      "loss": 0.5728,
+      "step": 10680
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 9.352589641434264e-06,
+      "loss": 0.5979,
+      "step": 10690
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 9.342629482071714e-06,
+      "loss": 0.7778,
+      "step": 10700
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 9.332669322709163e-06,
+      "loss": 0.7482,
+      "step": 10710
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 9.322709163346614e-06,
+      "loss": 0.7143,
+      "step": 10720
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 9.312749003984064e-06,
+      "loss": 0.5812,
+      "step": 10730
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 9.302788844621515e-06,
+      "loss": 0.6395,
+      "step": 10740
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 9.292828685258965e-06,
+      "loss": 0.6104,
+      "step": 10750
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 9.282868525896414e-06,
+      "loss": 0.6874,
+      "step": 10760
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 9.272908366533864e-06,
+      "loss": 0.4897,
+      "step": 10770
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 9.262948207171315e-06,
+      "loss": 0.6022,
+      "step": 10780
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 9.252988047808765e-06,
+      "loss": 0.5411,
+      "step": 10790
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 9.243027888446216e-06,
+      "loss": 0.6435,
+      "step": 10800
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 9.233067729083666e-06,
+      "loss": 0.5647,
+      "step": 10810
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 9.223107569721115e-06,
+      "loss": 0.5665,
+      "step": 10820
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 9.213147410358566e-06,
+      "loss": 0.4375,
+      "step": 10830
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 9.203187250996016e-06,
+      "loss": 0.4753,
+      "step": 10840
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 9.193227091633467e-06,
+      "loss": 0.3867,
+      "step": 10850
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 9.183266932270917e-06,
+      "loss": 0.5699,
+      "step": 10860
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 9.173306772908366e-06,
+      "loss": 0.4792,
+      "step": 10870
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 9.163346613545817e-06,
+      "loss": 0.5708,
+      "step": 10880
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 9.153386454183267e-06,
+      "loss": 0.6693,
+      "step": 10890
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 9.143426294820718e-06,
+      "loss": 0.4431,
+      "step": 10900
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 9.133466135458168e-06,
+      "loss": 0.4788,
+      "step": 10910
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 9.123505976095619e-06,
+      "loss": 0.5874,
+      "step": 10920
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 9.113545816733067e-06,
+      "loss": 0.4666,
+      "step": 10930
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 9.103585657370518e-06,
+      "loss": 0.4501,
+      "step": 10940
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 9.093625498007968e-06,
+      "loss": 0.6068,
+      "step": 10950
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 9.083665338645419e-06,
+      "loss": 0.7871,
+      "step": 10960
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 9.07370517928287e-06,
+      "loss": 0.6819,
+      "step": 10970
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 9.06374501992032e-06,
+      "loss": 0.6343,
+      "step": 10980
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 9.053784860557769e-06,
+      "loss": 0.5027,
+      "step": 10990
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 9.04382470119522e-06,
+      "loss": 0.3952,
+      "step": 11000
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 9.03386454183267e-06,
+      "loss": 0.5696,
+      "step": 11010
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 9.02390438247012e-06,
+      "loss": 0.5764,
+      "step": 11020
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 9.013944223107571e-06,
+      "loss": 0.6592,
+      "step": 11030
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 9.00398406374502e-06,
+      "loss": 0.6274,
+      "step": 11040
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 8.99402390438247e-06,
+      "loss": 0.6047,
+      "step": 11050
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 8.98406374501992e-06,
+      "loss": 0.5847,
+      "step": 11060
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 8.974103585657371e-06,
+      "loss": 0.5707,
+      "step": 11070
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 8.964143426294822e-06,
+      "loss": 0.529,
+      "step": 11080
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 8.954183266932272e-06,
+      "loss": 0.8672,
+      "step": 11090
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 8.944223107569721e-06,
+      "loss": 0.7242,
+      "step": 11100
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 8.934262948207172e-06,
+      "loss": 0.7125,
+      "step": 11110
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 8.924302788844622e-06,
+      "loss": 0.7131,
+      "step": 11120
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 8.914342629482073e-06,
+      "loss": 0.8014,
+      "step": 11130
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 8.904382470119523e-06,
+      "loss": 0.5761,
+      "step": 11140
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 8.894422310756972e-06,
+      "loss": 0.3705,
+      "step": 11150
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 8.884462151394422e-06,
+      "loss": 0.4697,
+      "step": 11160
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 8.874501992031873e-06,
+      "loss": 0.6275,
+      "step": 11170
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 8.864541832669323e-06,
+      "loss": 0.4263,
+      "step": 11180
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 8.854581673306774e-06,
+      "loss": 0.5357,
+      "step": 11190
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 8.844621513944224e-06,
+      "loss": 0.5909,
+      "step": 11200
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 8.834661354581673e-06,
+      "loss": 0.858,
+      "step": 11210
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 8.824701195219124e-06,
+      "loss": 0.444,
+      "step": 11220
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 8.814741035856574e-06,
+      "loss": 0.6035,
+      "step": 11230
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 8.804780876494025e-06,
+      "loss": 0.5319,
+      "step": 11240
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 8.794820717131475e-06,
+      "loss": 0.3976,
+      "step": 11250
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 8.784860557768926e-06,
+      "loss": 0.7064,
+      "step": 11260
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 8.774900398406375e-06,
+      "loss": 0.5722,
+      "step": 11270
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 8.764940239043825e-06,
+      "loss": 0.5212,
+      "step": 11280
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 8.754980079681276e-06,
+      "loss": 0.5634,
+      "step": 11290
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 8.745019920318726e-06,
+      "loss": 0.609,
+      "step": 11300
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 8.735059760956177e-06,
+      "loss": 0.4342,
+      "step": 11310
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 8.725099601593626e-06,
+      "loss": 0.3864,
+      "step": 11320
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 8.715139442231076e-06,
+      "loss": 0.6731,
+      "step": 11330
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 8.705179282868527e-06,
+      "loss": 0.4634,
+      "step": 11340
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 8.695219123505977e-06,
+      "loss": 0.6497,
+      "step": 11350
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 8.685258964143428e-06,
+      "loss": 0.5574,
+      "step": 11360
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 8.675298804780878e-06,
+      "loss": 0.6428,
+      "step": 11370
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 8.665338645418327e-06,
+      "loss": 0.6133,
+      "step": 11380
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 8.655378486055777e-06,
+      "loss": 0.6774,
+      "step": 11390
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 8.645418326693228e-06,
+      "loss": 0.4935,
+      "step": 11400
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 8.635458167330678e-06,
+      "loss": 0.4892,
+      "step": 11410
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 8.625498007968129e-06,
+      "loss": 0.4676,
+      "step": 11420
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 8.615537848605578e-06,
+      "loss": 0.5762,
+      "step": 11430
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 8.605577689243028e-06,
+      "loss": 0.7039,
+      "step": 11440
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 8.595617529880479e-06,
+      "loss": 0.6463,
+      "step": 11450
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 8.58565737051793e-06,
+      "loss": 0.5726,
+      "step": 11460
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 8.57569721115538e-06,
+      "loss": 0.7153,
+      "step": 11470
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 8.56573705179283e-06,
+      "loss": 0.5372,
+      "step": 11480
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 8.555776892430279e-06,
+      "loss": 0.6465,
+      "step": 11490
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 8.54581673306773e-06,
+      "loss": 0.5487,
+      "step": 11500
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 8.53585657370518e-06,
+      "loss": 0.6862,
+      "step": 11510
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 8.52589641434263e-06,
+      "loss": 0.7063,
+      "step": 11520
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 8.515936254980081e-06,
+      "loss": 0.7095,
+      "step": 11530
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 8.50597609561753e-06,
+      "loss": 0.6512,
+      "step": 11540
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 8.49601593625498e-06,
+      "loss": 0.8048,
+      "step": 11550
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 8.486055776892431e-06,
+      "loss": 0.5851,
+      "step": 11560
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 8.476095617529881e-06,
+      "loss": 0.5727,
+      "step": 11570
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 8.466135458167332e-06,
+      "loss": 0.5966,
+      "step": 11580
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 8.456175298804783e-06,
+      "loss": 0.3418,
+      "step": 11590
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 8.446215139442231e-06,
+      "loss": 0.5834,
+      "step": 11600
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 8.436254980079682e-06,
+      "loss": 0.7495,
+      "step": 11610
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 8.426294820717132e-06,
+      "loss": 0.4903,
+      "step": 11620
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 8.416334661354583e-06,
+      "loss": 0.5303,
+      "step": 11630
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 8.406374501992033e-06,
+      "loss": 0.6463,
+      "step": 11640
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 8.396414342629484e-06,
+      "loss": 0.5901,
+      "step": 11650
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 8.386454183266933e-06,
+      "loss": 0.5551,
+      "step": 11660
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 8.376494023904383e-06,
+      "loss": 0.5811,
+      "step": 11670
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 8.366533864541834e-06,
+      "loss": 0.5387,
+      "step": 11680
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 8.356573705179284e-06,
+      "loss": 0.6098,
+      "step": 11690
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 8.346613545816735e-06,
+      "loss": 0.6618,
+      "step": 11700
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 8.336653386454184e-06,
+      "loss": 0.603,
+      "step": 11710
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 8.326693227091634e-06,
+      "loss": 0.5484,
+      "step": 11720
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 8.316733067729085e-06,
+      "loss": 0.5999,
+      "step": 11730
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 8.306772908366535e-06,
+      "loss": 0.3665,
+      "step": 11740
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 8.296812749003986e-06,
+      "loss": 0.511,
+      "step": 11750
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 8.286852589641436e-06,
+      "loss": 0.7029,
+      "step": 11760
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 8.276892430278885e-06,
+      "loss": 0.6271,
+      "step": 11770
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 8.266932270916335e-06,
+      "loss": 0.7447,
+      "step": 11780
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 8.256972111553786e-06,
+      "loss": 0.6625,
+      "step": 11790
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 8.247011952191236e-06,
+      "loss": 0.5774,
+      "step": 11800
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 8.237051792828687e-06,
+      "loss": 0.5926,
+      "step": 11810
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 8.227091633466136e-06,
+      "loss": 0.6911,
+      "step": 11820
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 8.217131474103586e-06,
+      "loss": 0.6734,
+      "step": 11830
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 8.207171314741037e-06,
+      "loss": 0.5457,
+      "step": 11840
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 8.197211155378487e-06,
+      "loss": 0.6366,
+      "step": 11850
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 8.187250996015938e-06,
+      "loss": 0.616,
+      "step": 11860
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 8.177290836653388e-06,
+      "loss": 0.5629,
+      "step": 11870
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 8.167330677290837e-06,
+      "loss": 0.6825,
+      "step": 11880
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 8.157370517928288e-06,
+      "loss": 0.5056,
+      "step": 11890
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 8.147410358565738e-06,
+      "loss": 0.6219,
+      "step": 11900
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 8.137450199203189e-06,
+      "loss": 0.7134,
+      "step": 11910
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 8.12749003984064e-06,
+      "loss": 0.635,
+      "step": 11920
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 8.117529880478088e-06,
+      "loss": 0.3869,
+      "step": 11930
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 8.107569721115539e-06,
+      "loss": 0.4641,
+      "step": 11940
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 8.097609561752987e-06,
+      "loss": 0.6095,
+      "step": 11950
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 8.087649402390438e-06,
+      "loss": 0.4757,
+      "step": 11960
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 8.077689243027888e-06,
+      "loss": 0.5375,
+      "step": 11970
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 8.067729083665339e-06,
+      "loss": 0.7282,
+      "step": 11980
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 8.05776892430279e-06,
+      "loss": 0.6474,
+      "step": 11990
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 8.04780876494024e-06,
+      "loss": 0.5032,
+      "step": 12000
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 8.037848605577689e-06,
+      "loss": 0.427,
+      "step": 12010
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 8.02788844621514e-06,
+      "loss": 0.6275,
+      "step": 12020
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 8.01792828685259e-06,
+      "loss": 0.5081,
+      "step": 12030
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 8.00796812749004e-06,
+      "loss": 0.5696,
+      "step": 12040
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.797106563161609,
+      "eval_loss": 0.5187363028526306,
+      "eval_runtime": 12.5247,
+      "eval_samples_per_second": 226.273,
+      "eval_steps_per_second": 28.344,
+      "step": 12048
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 7.99800796812749e-06,
+      "loss": 0.6695,
+      "step": 12050
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 7.988047808764941e-06,
+      "loss": 0.7677,
+      "step": 12060
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 7.97808764940239e-06,
+      "loss": 0.7116,
+      "step": 12070
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 7.96812749003984e-06,
+      "loss": 0.4418,
+      "step": 12080
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 7.958167330677291e-06,
+      "loss": 0.6227,
+      "step": 12090
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 7.948207171314742e-06,
+      "loss": 0.7783,
+      "step": 12100
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 7.938247011952192e-06,
+      "loss": 0.601,
+      "step": 12110
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 7.928286852589641e-06,
+      "loss": 0.4237,
+      "step": 12120
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 7.918326693227091e-06,
+      "loss": 0.6167,
+      "step": 12130
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 7.908366533864542e-06,
+      "loss": 0.7637,
+      "step": 12140
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 7.898406374501992e-06,
+      "loss": 0.6366,
+      "step": 12150
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 7.888446215139443e-06,
+      "loss": 0.4891,
+      "step": 12160
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 7.878486055776893e-06,
+      "loss": 0.6284,
+      "step": 12170
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 7.868525896414342e-06,
+      "loss": 0.5783,
+      "step": 12180
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 7.858565737051793e-06,
+      "loss": 0.4514,
+      "step": 12190
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 7.848605577689243e-06,
+      "loss": 0.4417,
+      "step": 12200
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 7.838645418326694e-06,
+      "loss": 0.4861,
+      "step": 12210
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 7.828685258964144e-06,
+      "loss": 0.4081,
+      "step": 12220
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 7.818725099601593e-06,
+      "loss": 0.6688,
+      "step": 12230
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 7.808764940239044e-06,
+      "loss": 0.6554,
+      "step": 12240
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 7.798804780876494e-06,
+      "loss": 0.6114,
+      "step": 12250
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 7.788844621513945e-06,
+      "loss": 0.5268,
+      "step": 12260
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 7.778884462151395e-06,
+      "loss": 0.7053,
+      "step": 12270
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 7.768924302788846e-06,
+      "loss": 0.5817,
+      "step": 12280
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 7.758964143426295e-06,
+      "loss": 0.4887,
+      "step": 12290
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 7.749003984063745e-06,
+      "loss": 0.6909,
+      "step": 12300
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 7.739043824701196e-06,
+      "loss": 0.5208,
+      "step": 12310
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 7.729083665338646e-06,
+      "loss": 0.4697,
+      "step": 12320
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 7.719123505976097e-06,
+      "loss": 0.5739,
+      "step": 12330
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 7.709163346613547e-06,
+      "loss": 0.6711,
+      "step": 12340
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 7.699203187250996e-06,
+      "loss": 0.6656,
+      "step": 12350
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 7.689243027888446e-06,
+      "loss": 0.6536,
+      "step": 12360
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 7.679282868525897e-06,
+      "loss": 0.5429,
+      "step": 12370
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 7.669322709163347e-06,
+      "loss": 0.6721,
+      "step": 12380
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 7.659362549800798e-06,
+      "loss": 0.5073,
+      "step": 12390
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 7.649402390438247e-06,
+      "loss": 0.4724,
+      "step": 12400
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 7.639442231075697e-06,
+      "loss": 0.8172,
+      "step": 12410
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 7.629482071713148e-06,
+      "loss": 0.6466,
+      "step": 12420
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 7.619521912350598e-06,
+      "loss": 0.5664,
+      "step": 12430
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 7.609561752988048e-06,
+      "loss": 0.454,
+      "step": 12440
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 7.5996015936254985e-06,
+      "loss": 0.4968,
+      "step": 12450
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 7.589641434262949e-06,
+      "loss": 0.4288,
+      "step": 12460
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 7.579681274900399e-06,
+      "loss": 0.6098,
+      "step": 12470
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 7.569721115537849e-06,
+      "loss": 0.6802,
+      "step": 12480
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 7.5597609561753e-06,
+      "loss": 0.5829,
+      "step": 12490
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 7.549800796812749e-06,
+      "loss": 0.5111,
+      "step": 12500
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 7.5398406374502e-06,
+      "loss": 0.5891,
+      "step": 12510
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 7.52988047808765e-06,
+      "loss": 0.5702,
+      "step": 12520
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 7.5199203187251e-06,
+      "loss": 0.5772,
+      "step": 12530
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 7.5099601593625505e-06,
+      "loss": 0.4572,
+      "step": 12540
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.4659,
+      "step": 12550
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 7.490039840637451e-06,
+      "loss": 0.4549,
+      "step": 12560
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 7.480079681274901e-06,
+      "loss": 0.5067,
+      "step": 12570
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 7.470119521912351e-06,
+      "loss": 0.5489,
+      "step": 12580
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 7.460159362549801e-06,
+      "loss": 0.5839,
+      "step": 12590
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 7.450199203187252e-06,
+      "loss": 0.4875,
+      "step": 12600
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 7.4402390438247016e-06,
+      "loss": 0.5331,
+      "step": 12610
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 7.430278884462152e-06,
+      "loss": 0.4222,
+      "step": 12620
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 7.420318725099603e-06,
+      "loss": 0.5819,
+      "step": 12630
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 7.410358565737052e-06,
+      "loss": 0.706,
+      "step": 12640
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 7.400398406374503e-06,
+      "loss": 0.4294,
+      "step": 12650
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 7.390438247011953e-06,
+      "loss": 0.4794,
+      "step": 12660
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 7.380478087649403e-06,
+      "loss": 0.4752,
+      "step": 12670
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 7.3705179282868534e-06,
+      "loss": 0.4139,
+      "step": 12680
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 7.360557768924303e-06,
+      "loss": 0.5388,
+      "step": 12690
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 7.350597609561754e-06,
+      "loss": 0.3659,
+      "step": 12700
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 7.340637450199204e-06,
+      "loss": 0.5128,
+      "step": 12710
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 7.330677290836654e-06,
+      "loss": 0.6473,
+      "step": 12720
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 7.320717131474104e-06,
+      "loss": 0.5582,
+      "step": 12730
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 7.310756972111555e-06,
+      "loss": 0.4374,
+      "step": 12740
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 7.3007968127490045e-06,
+      "loss": 0.7489,
+      "step": 12750
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 7.290836653386455e-06,
+      "loss": 0.3461,
+      "step": 12760
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 7.2808764940239055e-06,
+      "loss": 0.5546,
+      "step": 12770
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 7.270916334661355e-06,
+      "loss": 0.691,
+      "step": 12780
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 7.260956175298806e-06,
+      "loss": 0.6567,
+      "step": 12790
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 7.250996015936256e-06,
+      "loss": 0.3622,
+      "step": 12800
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 7.241035856573706e-06,
+      "loss": 0.2811,
+      "step": 12810
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 7.231075697211156e-06,
+      "loss": 0.8515,
+      "step": 12820
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 7.221115537848606e-06,
+      "loss": 0.6433,
+      "step": 12830
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 7.2111553784860565e-06,
+      "loss": 0.5357,
+      "step": 12840
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 7.201195219123507e-06,
+      "loss": 0.3968,
+      "step": 12850
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 7.191235059760957e-06,
+      "loss": 0.4755,
+      "step": 12860
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 7.181274900398407e-06,
+      "loss": 0.6498,
+      "step": 12870
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 7.171314741035858e-06,
+      "loss": 0.7729,
+      "step": 12880
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 7.161354581673307e-06,
+      "loss": 0.6945,
+      "step": 12890
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 7.151394422310758e-06,
+      "loss": 0.5372,
+      "step": 12900
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 7.141434262948208e-06,
+      "loss": 0.532,
+      "step": 12910
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 7.131474103585658e-06,
+      "loss": 0.6142,
+      "step": 12920
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 7.121513944223109e-06,
+      "loss": 0.6126,
+      "step": 12930
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 7.111553784860559e-06,
+      "loss": 0.5998,
+      "step": 12940
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 7.101593625498009e-06,
+      "loss": 0.57,
+      "step": 12950
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 7.091633466135459e-06,
+      "loss": 0.4756,
+      "step": 12960
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 7.081673306772909e-06,
+      "loss": 0.6896,
+      "step": 12970
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 7.0717131474103594e-06,
+      "loss": 0.5261,
+      "step": 12980
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 7.06175298804781e-06,
+      "loss": 0.5949,
+      "step": 12990
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 7.05179282868526e-06,
+      "loss": 0.642,
+      "step": 13000
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 7.04183266932271e-06,
+      "loss": 0.7364,
+      "step": 13010
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 7.031872509960161e-06,
+      "loss": 0.4784,
+      "step": 13020
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 7.02191235059761e-06,
+      "loss": 0.492,
+      "step": 13030
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 7.011952191235061e-06,
+      "loss": 0.6409,
+      "step": 13040
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 7.001992031872511e-06,
+      "loss": 0.5648,
+      "step": 13050
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 6.992031872509961e-06,
+      "loss": 0.5776,
+      "step": 13060
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 6.9820717131474115e-06,
+      "loss": 0.5583,
+      "step": 13070
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 6.972111553784862e-06,
+      "loss": 0.5127,
+      "step": 13080
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 6.962151394422312e-06,
+      "loss": 0.4751,
+      "step": 13090
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 6.952191235059762e-06,
+      "loss": 0.4075,
+      "step": 13100
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 6.942231075697212e-06,
+      "loss": 0.5972,
+      "step": 13110
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 6.932270916334662e-06,
+      "loss": 0.5381,
+      "step": 13120
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 6.922310756972113e-06,
+      "loss": 0.6118,
+      "step": 13130
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 6.9123505976095625e-06,
+      "loss": 0.6272,
+      "step": 13140
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 6.902390438247013e-06,
+      "loss": 0.5835,
+      "step": 13150
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 6.8924302788844635e-06,
+      "loss": 0.6003,
+      "step": 13160
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 6.882470119521913e-06,
+      "loss": 0.4997,
+      "step": 13170
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 6.872509960159363e-06,
+      "loss": 0.4771,
+      "step": 13180
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 6.8625498007968125e-06,
+      "loss": 0.6436,
+      "step": 13190
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 6.852589641434263e-06,
+      "loss": 0.7057,
+      "step": 13200
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 6.8426294820717136e-06,
+      "loss": 0.5272,
+      "step": 13210
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 6.832669322709163e-06,
+      "loss": 0.4446,
+      "step": 13220
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 6.822709163346614e-06,
+      "loss": 0.4614,
+      "step": 13230
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 6.812749003984063e-06,
+      "loss": 0.8576,
+      "step": 13240
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 6.802788844621514e-06,
+      "loss": 0.6073,
+      "step": 13250
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 6.792828685258964e-06,
+      "loss": 0.5205,
+      "step": 13260
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 6.782868525896414e-06,
+      "loss": 0.6615,
+      "step": 13270
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 6.772908366533865e-06,
+      "loss": 0.4443,
+      "step": 13280
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 6.762948207171315e-06,
+      "loss": 0.4867,
+      "step": 13290
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 6.752988047808765e-06,
+      "loss": 0.5896,
+      "step": 13300
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 6.743027888446215e-06,
+      "loss": 0.7068,
+      "step": 13310
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 6.733067729083666e-06,
+      "loss": 0.5249,
+      "step": 13320
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 6.7231075697211154e-06,
+      "loss": 0.7351,
+      "step": 13330
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 6.713147410358566e-06,
+      "loss": 0.5428,
+      "step": 13340
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 6.7031872509960165e-06,
+      "loss": 0.4854,
+      "step": 13350
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 6.693227091633466e-06,
+      "loss": 0.5583,
+      "step": 13360
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 6.683266932270917e-06,
+      "loss": 0.507,
+      "step": 13370
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 6.673306772908366e-06,
+      "loss": 0.3868,
+      "step": 13380
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 6.663346613545817e-06,
+      "loss": 0.4286,
+      "step": 13390
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 6.653386454183267e-06,
+      "loss": 0.719,
+      "step": 13400
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 6.643426294820717e-06,
+      "loss": 0.5989,
+      "step": 13410
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 6.6334661354581675e-06,
+      "loss": 0.6752,
+      "step": 13420
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 6.623505976095618e-06,
+      "loss": 0.7239,
+      "step": 13430
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 6.613545816733068e-06,
+      "loss": 0.5251,
+      "step": 13440
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 6.603585657370518e-06,
+      "loss": 0.5813,
+      "step": 13450
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 6.593625498007969e-06,
+      "loss": 0.6049,
+      "step": 13460
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 6.583665338645418e-06,
+      "loss": 0.5703,
+      "step": 13470
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 6.573705179282869e-06,
+      "loss": 0.4525,
+      "step": 13480
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 6.563745019920319e-06,
+      "loss": 0.3916,
+      "step": 13490
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 6.553784860557769e-06,
+      "loss": 0.5247,
+      "step": 13500
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 6.5438247011952196e-06,
+      "loss": 0.4046,
+      "step": 13510
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 6.533864541832669e-06,
+      "loss": 0.576,
+      "step": 13520
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 6.52390438247012e-06,
+      "loss": 0.5616,
+      "step": 13530
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 6.51394422310757e-06,
+      "loss": 0.5975,
+      "step": 13540
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 6.50398406374502e-06,
+      "loss": 0.4555,
+      "step": 13550
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 6.49402390438247e-06,
+      "loss": 0.4115,
+      "step": 13560
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 6.484063745019921e-06,
+      "loss": 0.7309,
+      "step": 13570
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 6.474103585657371e-06,
+      "loss": 0.4828,
+      "step": 13580
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 6.464143426294821e-06,
+      "loss": 0.58,
+      "step": 13590
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 6.454183266932272e-06,
+      "loss": 0.4831,
+      "step": 13600
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 6.444223107569721e-06,
+      "loss": 0.4578,
+      "step": 13610
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 6.434262948207172e-06,
+      "loss": 0.5578,
+      "step": 13620
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 6.424302788844622e-06,
+      "loss": 0.5746,
+      "step": 13630
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 6.414342629482072e-06,
+      "loss": 0.5564,
+      "step": 13640
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 6.4043824701195225e-06,
+      "loss": 0.2511,
+      "step": 13650
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 6.394422310756972e-06,
+      "loss": 0.3815,
+      "step": 13660
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 6.384462151394423e-06,
+      "loss": 0.6406,
+      "step": 13670
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 6.374501992031873e-06,
+      "loss": 0.481,
+      "step": 13680
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 6.364541832669323e-06,
+      "loss": 0.5058,
+      "step": 13690
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 6.354581673306773e-06,
+      "loss": 0.4552,
+      "step": 13700
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 6.344621513944224e-06,
+      "loss": 0.6729,
+      "step": 13710
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 6.3346613545816735e-06,
+      "loss": 0.674,
+      "step": 13720
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 6.324701195219124e-06,
+      "loss": 0.4651,
+      "step": 13730
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 6.3147410358565745e-06,
+      "loss": 0.6089,
+      "step": 13740
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 6.304780876494024e-06,
+      "loss": 0.5601,
+      "step": 13750
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 6.294820717131475e-06,
+      "loss": 0.58,
+      "step": 13760
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 6.284860557768925e-06,
+      "loss": 0.6908,
+      "step": 13770
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 6.274900398406375e-06,
+      "loss": 0.4736,
+      "step": 13780
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 6.264940239043825e-06,
+      "loss": 0.6372,
+      "step": 13790
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 6.254980079681275e-06,
+      "loss": 0.5164,
+      "step": 13800
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 6.2450199203187256e-06,
+      "loss": 0.6392,
+      "step": 13810
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 6.235059760956176e-06,
+      "loss": 0.5604,
+      "step": 13820
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 6.225099601593626e-06,
+      "loss": 0.3918,
+      "step": 13830
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 6.215139442231076e-06,
+      "loss": 0.4223,
+      "step": 13840
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 6.205179282868527e-06,
+      "loss": 0.6227,
+      "step": 13850
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 6.195219123505976e-06,
+      "loss": 0.6022,
+      "step": 13860
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 6.185258964143427e-06,
+      "loss": 0.6236,
+      "step": 13870
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 6.1752988047808774e-06,
+      "loss": 0.4162,
+      "step": 13880
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 6.165338645418327e-06,
+      "loss": 0.5005,
+      "step": 13890
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 6.155378486055778e-06,
+      "loss": 0.5562,
+      "step": 13900
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 6.145418326693227e-06,
+      "loss": 0.5187,
+      "step": 13910
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 6.135458167330678e-06,
+      "loss": 0.569,
+      "step": 13920
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 6.125498007968128e-06,
+      "loss": 0.4239,
+      "step": 13930
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 6.115537848605578e-06,
+      "loss": 0.5256,
+      "step": 13940
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 6.1055776892430285e-06,
+      "loss": 0.7434,
+      "step": 13950
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 6.095617529880479e-06,
+      "loss": 0.6039,
+      "step": 13960
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 6.085657370517929e-06,
+      "loss": 0.6296,
+      "step": 13970
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 6.075697211155379e-06,
+      "loss": 0.4782,
+      "step": 13980
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 6.06573705179283e-06,
+      "loss": 0.5992,
+      "step": 13990
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 6.055776892430279e-06,
+      "loss": 0.6428,
+      "step": 14000
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 6.04581673306773e-06,
+      "loss": 0.5274,
+      "step": 14010
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 6.03585657370518e-06,
+      "loss": 0.5456,
+      "step": 14020
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 6.02589641434263e-06,
+      "loss": 0.4351,
+      "step": 14030
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 6.0159362549800805e-06,
+      "loss": 0.6747,
+      "step": 14040
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 6.00597609561753e-06,
+      "loss": 0.6925,
+      "step": 14050
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.8038108680310515,
+      "eval_loss": 0.5082055330276489,
+      "eval_runtime": 12.4353,
+      "eval_samples_per_second": 227.9,
+      "eval_steps_per_second": 28.548,
+      "step": 14056
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 5.996015936254981e-06,
+      "loss": 0.4451,
+      "step": 14060
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 5.986055776892431e-06,
+      "loss": 0.5667,
+      "step": 14070
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 5.976095617529881e-06,
+      "loss": 0.5743,
+      "step": 14080
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 5.966135458167331e-06,
+      "loss": 0.4747,
+      "step": 14090
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 5.956175298804782e-06,
+      "loss": 0.7133,
+      "step": 14100
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 5.9462151394422316e-06,
+      "loss": 0.5799,
+      "step": 14110
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 5.936254980079682e-06,
+      "loss": 0.546,
+      "step": 14120
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 5.9262948207171326e-06,
+      "loss": 0.4935,
+      "step": 14130
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 5.916334661354582e-06,
+      "loss": 0.5308,
+      "step": 14140
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 5.906374501992033e-06,
+      "loss": 0.5171,
+      "step": 14150
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 5.896414342629483e-06,
+      "loss": 0.3923,
+      "step": 14160
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 5.886454183266933e-06,
+      "loss": 0.5213,
+      "step": 14170
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 5.8764940239043834e-06,
+      "loss": 0.6722,
+      "step": 14180
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 5.866533864541833e-06,
+      "loss": 0.4981,
+      "step": 14190
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 5.856573705179284e-06,
+      "loss": 0.3095,
+      "step": 14200
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 5.846613545816734e-06,
+      "loss": 0.5843,
+      "step": 14210
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 5.836653386454184e-06,
+      "loss": 0.5124,
+      "step": 14220
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 5.826693227091634e-06,
+      "loss": 0.6459,
+      "step": 14230
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 5.816733067729085e-06,
+      "loss": 0.7528,
+      "step": 14240
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 5.8067729083665345e-06,
+      "loss": 0.5661,
+      "step": 14250
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 5.796812749003985e-06,
+      "loss": 0.456,
+      "step": 14260
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 5.7868525896414355e-06,
+      "loss": 0.4424,
+      "step": 14270
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 5.776892430278885e-06,
+      "loss": 0.4202,
+      "step": 14280
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 5.766932270916336e-06,
+      "loss": 0.5372,
+      "step": 14290
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 5.756972111553786e-06,
+      "loss": 0.3845,
+      "step": 14300
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 5.747011952191236e-06,
+      "loss": 0.4565,
+      "step": 14310
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 5.737051792828686e-06,
+      "loss": 0.4562,
+      "step": 14320
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 5.727091633466136e-06,
+      "loss": 0.4594,
+      "step": 14330
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 5.7171314741035865e-06,
+      "loss": 0.542,
+      "step": 14340
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 5.707171314741037e-06,
+      "loss": 0.4657,
+      "step": 14350
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 5.697211155378487e-06,
+      "loss": 0.6771,
+      "step": 14360
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 5.687250996015937e-06,
+      "loss": 0.7343,
+      "step": 14370
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 5.677290836653388e-06,
+      "loss": 0.5098,
+      "step": 14380
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 5.667330677290837e-06,
+      "loss": 0.6568,
+      "step": 14390
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 5.657370517928288e-06,
+      "loss": 0.7888,
+      "step": 14400
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 5.647410358565738e-06,
+      "loss": 0.4103,
+      "step": 14410
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 5.637450199203188e-06,
+      "loss": 0.8035,
+      "step": 14420
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 5.6274900398406386e-06,
+      "loss": 0.6049,
+      "step": 14430
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 5.617529880478087e-06,
+      "loss": 0.5686,
+      "step": 14440
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 5.607569721115538e-06,
+      "loss": 0.7052,
+      "step": 14450
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 5.5976095617529876e-06,
+      "loss": 0.5038,
+      "step": 14460
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 5.587649402390438e-06,
+      "loss": 0.4682,
+      "step": 14470
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 5.577689243027889e-06,
+      "loss": 0.5077,
+      "step": 14480
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 5.567729083665338e-06,
+      "loss": 0.513,
+      "step": 14490
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 5.557768924302789e-06,
+      "loss": 0.5489,
+      "step": 14500
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 5.547808764940239e-06,
+      "loss": 0.4823,
+      "step": 14510
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 5.537848605577689e-06,
+      "loss": 0.5289,
+      "step": 14520
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 5.5278884462151394e-06,
+      "loss": 0.4977,
+      "step": 14530
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 5.51792828685259e-06,
+      "loss": 0.434,
+      "step": 14540
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 5.50796812749004e-06,
+      "loss": 0.452,
+      "step": 14550
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 5.49800796812749e-06,
+      "loss": 0.6038,
+      "step": 14560
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 5.488047808764941e-06,
+      "loss": 0.6288,
+      "step": 14570
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 5.47808764940239e-06,
+      "loss": 0.5716,
+      "step": 14580
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 5.468127490039841e-06,
+      "loss": 0.6424,
+      "step": 14590
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 5.4581673306772905e-06,
+      "loss": 0.5993,
+      "step": 14600
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 5.448207171314741e-06,
+      "loss": 0.5591,
+      "step": 14610
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 5.4382470119521915e-06,
+      "loss": 0.4687,
+      "step": 14620
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 5.428286852589641e-06,
+      "loss": 0.6839,
+      "step": 14630
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 5.418326693227092e-06,
+      "loss": 0.6262,
+      "step": 14640
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 5.408366533864542e-06,
+      "loss": 0.383,
+      "step": 14650
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 5.398406374501992e-06,
+      "loss": 0.5422,
+      "step": 14660
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 5.388446215139442e-06,
+      "loss": 0.5963,
+      "step": 14670
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 5.378486055776893e-06,
+      "loss": 0.5684,
+      "step": 14680
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 5.3685258964143425e-06,
+      "loss": 0.5124,
+      "step": 14690
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 5.358565737051793e-06,
+      "loss": 0.5892,
+      "step": 14700
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 5.3486055776892436e-06,
+      "loss": 0.6804,
+      "step": 14710
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 5.338645418326693e-06,
+      "loss": 0.744,
+      "step": 14720
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 5.328685258964144e-06,
+      "loss": 0.5033,
+      "step": 14730
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 5.318725099601593e-06,
+      "loss": 0.6045,
+      "step": 14740
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 5.308764940239044e-06,
+      "loss": 0.461,
+      "step": 14750
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 5.298804780876494e-06,
+      "loss": 0.5858,
+      "step": 14760
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 5.288844621513944e-06,
+      "loss": 0.6894,
+      "step": 14770
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 5.278884462151395e-06,
+      "loss": 0.4376,
+      "step": 14780
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 5.268924302788845e-06,
+      "loss": 0.3911,
+      "step": 14790
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 5.258964143426295e-06,
+      "loss": 0.481,
+      "step": 14800
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 5.249003984063745e-06,
+      "loss": 0.581,
+      "step": 14810
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 5.239043824701196e-06,
+      "loss": 0.4364,
+      "step": 14820
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 5.2290836653386454e-06,
+      "loss": 0.4107,
+      "step": 14830
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 5.219123505976096e-06,
+      "loss": 0.502,
+      "step": 14840
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 5.2091633466135465e-06,
+      "loss": 0.5012,
+      "step": 14850
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 5.199203187250996e-06,
+      "loss": 0.4538,
+      "step": 14860
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 5.189243027888447e-06,
+      "loss": 0.4881,
+      "step": 14870
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 5.179282868525896e-06,
+      "loss": 0.7348,
+      "step": 14880
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 5.169322709163347e-06,
+      "loss": 0.582,
+      "step": 14890
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 5.159362549800797e-06,
+      "loss": 0.6186,
+      "step": 14900
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 5.149402390438247e-06,
+      "loss": 0.4549,
+      "step": 14910
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 5.1394422310756975e-06,
+      "loss": 0.5866,
+      "step": 14920
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 5.129482071713148e-06,
+      "loss": 0.6193,
+      "step": 14930
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 5.119521912350598e-06,
+      "loss": 0.5156,
+      "step": 14940
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 5.109561752988048e-06,
+      "loss": 0.5604,
+      "step": 14950
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 5.099601593625499e-06,
+      "loss": 0.3008,
+      "step": 14960
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 5.089641434262948e-06,
+      "loss": 0.4341,
+      "step": 14970
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 5.079681274900399e-06,
+      "loss": 0.5978,
+      "step": 14980
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 5.069721115537849e-06,
+      "loss": 0.6953,
+      "step": 14990
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 5.059760956175299e-06,
+      "loss": 0.4379,
+      "step": 15000
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 5.0498007968127496e-06,
+      "loss": 0.6335,
+      "step": 15010
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 5.039840637450199e-06,
+      "loss": 0.5588,
+      "step": 15020
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 5.02988047808765e-06,
+      "loss": 0.5238,
+      "step": 15030
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 5.0199203187251e-06,
+      "loss": 0.4946,
+      "step": 15040
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 5.00996015936255e-06,
+      "loss": 0.537,
+      "step": 15050
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 5e-06,
+      "loss": 0.3886,
+      "step": 15060
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 4.990039840637451e-06,
+      "loss": 0.3871,
+      "step": 15070
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 4.980079681274901e-06,
+      "loss": 0.5236,
+      "step": 15080
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 4.970119521912351e-06,
+      "loss": 0.4614,
+      "step": 15090
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 4.960159362549802e-06,
+      "loss": 0.4073,
+      "step": 15100
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 4.950199203187251e-06,
+      "loss": 0.3829,
+      "step": 15110
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 4.940239043824702e-06,
+      "loss": 0.6448,
+      "step": 15120
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 4.9302788844621514e-06,
+      "loss": 0.6287,
+      "step": 15130
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 4.920318725099602e-06,
+      "loss": 0.4138,
+      "step": 15140
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 4.9103585657370525e-06,
+      "loss": 0.7237,
+      "step": 15150
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 4.900398406374502e-06,
+      "loss": 0.4557,
+      "step": 15160
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 4.890438247011953e-06,
+      "loss": 0.5248,
+      "step": 15170
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 4.880478087649403e-06,
+      "loss": 0.6194,
+      "step": 15180
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 4.870517928286853e-06,
+      "loss": 0.5036,
+      "step": 15190
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 4.860557768924303e-06,
+      "loss": 0.5905,
+      "step": 15200
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 4.850597609561754e-06,
+      "loss": 0.4352,
+      "step": 15210
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 4.8406374501992035e-06,
+      "loss": 0.6527,
+      "step": 15220
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 4.830677290836654e-06,
+      "loss": 0.7561,
+      "step": 15230
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 4.8207171314741045e-06,
+      "loss": 0.5117,
+      "step": 15240
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 4.810756972111554e-06,
+      "loss": 0.4597,
+      "step": 15250
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 4.800796812749005e-06,
+      "loss": 0.6125,
+      "step": 15260
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 4.790836653386454e-06,
+      "loss": 0.5426,
+      "step": 15270
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 4.780876494023905e-06,
+      "loss": 0.5405,
+      "step": 15280
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 4.770916334661355e-06,
+      "loss": 0.4973,
+      "step": 15290
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 4.760956175298805e-06,
+      "loss": 0.6566,
+      "step": 15300
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 4.7509960159362555e-06,
+      "loss": 0.5763,
+      "step": 15310
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 4.741035856573706e-06,
+      "loss": 0.597,
+      "step": 15320
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 4.731075697211156e-06,
+      "loss": 0.5112,
+      "step": 15330
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 4.721115537848606e-06,
+      "loss": 0.5034,
+      "step": 15340
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 4.711155378486057e-06,
+      "loss": 0.5071,
+      "step": 15350
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 4.701195219123506e-06,
+      "loss": 0.5127,
+      "step": 15360
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 4.691235059760957e-06,
+      "loss": 0.4724,
+      "step": 15370
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 4.681274900398407e-06,
+      "loss": 0.4786,
+      "step": 15380
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 4.671314741035857e-06,
+      "loss": 0.578,
+      "step": 15390
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 4.661354581673307e-06,
+      "loss": 0.4916,
+      "step": 15400
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 4.651394422310757e-06,
+      "loss": 0.5525,
+      "step": 15410
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 4.641434262948207e-06,
+      "loss": 0.5526,
+      "step": 15420
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 4.6314741035856574e-06,
+      "loss": 0.679,
+      "step": 15430
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 4.621513944223108e-06,
+      "loss": 0.5366,
+      "step": 15440
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 4.611553784860558e-06,
+      "loss": 0.4451,
+      "step": 15450
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 4.601593625498008e-06,
+      "loss": 0.6483,
+      "step": 15460
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 4.591633466135459e-06,
+      "loss": 0.5922,
+      "step": 15470
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 4.581673306772908e-06,
+      "loss": 0.4873,
+      "step": 15480
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 4.571713147410359e-06,
+      "loss": 0.3956,
+      "step": 15490
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 4.561752988047809e-06,
+      "loss": 0.5001,
+      "step": 15500
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 4.551792828685259e-06,
+      "loss": 0.4239,
+      "step": 15510
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 4.5418326693227095e-06,
+      "loss": 0.3169,
+      "step": 15520
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 4.53187250996016e-06,
+      "loss": 0.5022,
+      "step": 15530
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 4.52191235059761e-06,
+      "loss": 0.7488,
+      "step": 15540
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 4.51195219123506e-06,
+      "loss": 0.6319,
+      "step": 15550
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 4.50199203187251e-06,
+      "loss": 0.5754,
+      "step": 15560
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 4.49203187250996e-06,
+      "loss": 0.6268,
+      "step": 15570
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 4.482071713147411e-06,
+      "loss": 0.5245,
+      "step": 15580
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 4.4721115537848605e-06,
+      "loss": 0.575,
+      "step": 15590
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 4.462151394422311e-06,
+      "loss": 0.5156,
+      "step": 15600
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 4.4521912350597615e-06,
+      "loss": 0.5182,
+      "step": 15610
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 4.442231075697211e-06,
+      "loss": 0.5924,
+      "step": 15620
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 4.432270916334662e-06,
+      "loss": 0.4892,
+      "step": 15630
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 4.422310756972112e-06,
+      "loss": 0.5357,
+      "step": 15640
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 4.412350597609562e-06,
+      "loss": 0.5746,
+      "step": 15650
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 4.402390438247012e-06,
+      "loss": 0.4142,
+      "step": 15660
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 4.392430278884463e-06,
+      "loss": 0.4429,
+      "step": 15670
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 4.382470119521913e-06,
+      "loss": 0.4797,
+      "step": 15680
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 4.372509960159363e-06,
+      "loss": 0.4927,
+      "step": 15690
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 4.362549800796813e-06,
+      "loss": 0.5911,
+      "step": 15700
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 4.352589641434263e-06,
+      "loss": 0.3478,
+      "step": 15710
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 4.342629482071714e-06,
+      "loss": 0.4697,
+      "step": 15720
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 4.3326693227091634e-06,
+      "loss": 0.5573,
+      "step": 15730
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 4.322709163346614e-06,
+      "loss": 0.5747,
+      "step": 15740
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 4.3127490039840645e-06,
+      "loss": 0.424,
+      "step": 15750
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 4.302788844621514e-06,
+      "loss": 0.6629,
+      "step": 15760
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 4.292828685258965e-06,
+      "loss": 0.4201,
+      "step": 15770
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 4.282868525896415e-06,
+      "loss": 0.4973,
+      "step": 15780
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 4.272908366533865e-06,
+      "loss": 0.4969,
+      "step": 15790
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 4.262948207171315e-06,
+      "loss": 0.3714,
+      "step": 15800
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 4.252988047808765e-06,
+      "loss": 0.3269,
+      "step": 15810
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 4.2430278884462155e-06,
+      "loss": 0.8514,
+      "step": 15820
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 4.233067729083666e-06,
+      "loss": 0.7288,
+      "step": 15830
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 4.223107569721116e-06,
+      "loss": 0.6349,
+      "step": 15840
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 4.213147410358566e-06,
+      "loss": 0.6079,
+      "step": 15850
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 4.203187250996017e-06,
+      "loss": 0.5376,
+      "step": 15860
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 4.193227091633466e-06,
+      "loss": 0.6511,
+      "step": 15870
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 4.183266932270917e-06,
+      "loss": 0.6402,
+      "step": 15880
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 4.173306772908367e-06,
+      "loss": 0.4694,
+      "step": 15890
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 4.163346613545817e-06,
+      "loss": 0.615,
+      "step": 15900
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 4.1533864541832675e-06,
+      "loss": 0.6239,
+      "step": 15910
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 4.143426294820718e-06,
+      "loss": 0.6424,
+      "step": 15920
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 4.133466135458168e-06,
+      "loss": 0.4866,
+      "step": 15930
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 4.123505976095618e-06,
+      "loss": 0.5413,
+      "step": 15940
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 4.113545816733068e-06,
+      "loss": 0.5445,
+      "step": 15950
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 4.103585657370518e-06,
+      "loss": 0.487,
+      "step": 15960
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 4.093625498007969e-06,
+      "loss": 0.5342,
+      "step": 15970
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 4.083665338645419e-06,
+      "loss": 0.561,
+      "step": 15980
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 4.073705179282869e-06,
+      "loss": 0.4157,
+      "step": 15990
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 4.06374501992032e-06,
+      "loss": 0.2521,
+      "step": 16000
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 4.053784860557769e-06,
+      "loss": 0.511,
+      "step": 16010
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 4.043824701195219e-06,
+      "loss": 0.386,
+      "step": 16020
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 4.0338645418326694e-06,
+      "loss": 0.5799,
+      "step": 16030
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 4.02390438247012e-06,
+      "loss": 0.5091,
+      "step": 16040
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 4.01394422310757e-06,
+      "loss": 0.4357,
+      "step": 16050
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 4.00398406374502e-06,
+      "loss": 0.5711,
+      "step": 16060
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.8098094565984474,
+      "eval_loss": 0.5091592073440552,
+      "eval_runtime": 12.4207,
+      "eval_samples_per_second": 228.168,
+      "eval_steps_per_second": 28.581,
+      "step": 16064
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.994023904382471e-06,
+      "loss": 0.651,
+      "step": 16070
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.98406374501992e-06,
+      "loss": 0.4787,
+      "step": 16080
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.974103585657371e-06,
+      "loss": 0.4972,
+      "step": 16090
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.9641434262948205e-06,
+      "loss": 0.6024,
+      "step": 16100
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.954183266932271e-06,
+      "loss": 0.6903,
+      "step": 16110
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 3.9442231075697215e-06,
+      "loss": 0.5023,
+      "step": 16120
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 3.934262948207171e-06,
+      "loss": 0.5712,
+      "step": 16130
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 3.924302788844622e-06,
+      "loss": 0.5017,
+      "step": 16140
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 3.914342629482072e-06,
+      "loss": 0.5389,
+      "step": 16150
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 3.904382470119522e-06,
+      "loss": 0.4496,
+      "step": 16160
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 3.894422310756972e-06,
+      "loss": 0.5249,
+      "step": 16170
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 3.884462151394423e-06,
+      "loss": 0.4915,
+      "step": 16180
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 3.8745019920318725e-06,
+      "loss": 0.5935,
+      "step": 16190
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 3.864541832669323e-06,
+      "loss": 0.6911,
+      "step": 16200
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 3.8545816733067735e-06,
+      "loss": 0.4751,
+      "step": 16210
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 3.844621513944223e-06,
+      "loss": 0.5733,
+      "step": 16220
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 3.834661354581674e-06,
+      "loss": 0.4281,
+      "step": 16230
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 3.824701195219123e-06,
+      "loss": 0.4155,
+      "step": 16240
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 3.814741035856574e-06,
+      "loss": 0.5646,
+      "step": 16250
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 3.804780876494024e-06,
+      "loss": 0.6027,
+      "step": 16260
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 3.7948207171314745e-06,
+      "loss": 0.4043,
+      "step": 16270
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 3.7848605577689246e-06,
+      "loss": 0.4298,
+      "step": 16280
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 3.7749003984063747e-06,
+      "loss": 0.659,
+      "step": 16290
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 3.764940239043825e-06,
+      "loss": 0.5131,
+      "step": 16300
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 3.7549800796812753e-06,
+      "loss": 0.4614,
+      "step": 16310
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 3.7450199203187254e-06,
+      "loss": 0.4767,
+      "step": 16320
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 3.7350597609561754e-06,
+      "loss": 0.4648,
+      "step": 16330
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 3.725099601593626e-06,
+      "loss": 0.4651,
+      "step": 16340
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 3.715139442231076e-06,
+      "loss": 0.6144,
+      "step": 16350
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 3.705179282868526e-06,
+      "loss": 0.6606,
+      "step": 16360
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 3.6952191235059766e-06,
+      "loss": 0.4969,
+      "step": 16370
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 3.6852589641434267e-06,
+      "loss": 0.4191,
+      "step": 16380
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 3.675298804780877e-06,
+      "loss": 0.5402,
+      "step": 16390
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 3.665338645418327e-06,
+      "loss": 0.5864,
+      "step": 16400
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 3.6553784860557774e-06,
+      "loss": 0.5457,
+      "step": 16410
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 3.6454183266932275e-06,
+      "loss": 0.4721,
+      "step": 16420
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 3.6354581673306776e-06,
+      "loss": 0.4041,
+      "step": 16430
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 3.625498007968128e-06,
+      "loss": 0.4841,
+      "step": 16440
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 3.615537848605578e-06,
+      "loss": 0.4831,
+      "step": 16450
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 3.6055776892430283e-06,
+      "loss": 0.4934,
+      "step": 16460
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 3.5956175298804783e-06,
+      "loss": 0.4875,
+      "step": 16470
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 3.585657370517929e-06,
+      "loss": 0.4133,
+      "step": 16480
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 3.575697211155379e-06,
+      "loss": 0.509,
+      "step": 16490
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 3.565737051792829e-06,
+      "loss": 0.4949,
+      "step": 16500
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 3.5557768924302795e-06,
+      "loss": 0.5502,
+      "step": 16510
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 3.5458167330677296e-06,
+      "loss": 0.5946,
+      "step": 16520
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 3.5358565737051797e-06,
+      "loss": 0.5481,
+      "step": 16530
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 3.52589641434263e-06,
+      "loss": 0.426,
+      "step": 16540
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 3.5159362549800803e-06,
+      "loss": 0.6544,
+      "step": 16550
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 3.5059760956175304e-06,
+      "loss": 0.6521,
+      "step": 16560
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 3.4960159362549805e-06,
+      "loss": 0.3717,
+      "step": 16570
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 3.486055776892431e-06,
+      "loss": 0.252,
+      "step": 16580
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 3.476095617529881e-06,
+      "loss": 0.6715,
+      "step": 16590
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 3.466135458167331e-06,
+      "loss": 0.5278,
+      "step": 16600
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 3.4561752988047813e-06,
+      "loss": 0.3346,
+      "step": 16610
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 3.4462151394422318e-06,
+      "loss": 0.3946,
+      "step": 16620
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 3.4362549800796814e-06,
+      "loss": 0.486,
+      "step": 16630
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 3.4262948207171315e-06,
+      "loss": 0.68,
+      "step": 16640
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 3.4163346613545816e-06,
+      "loss": 0.828,
+      "step": 16650
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 3.4063745019920317e-06,
+      "loss": 0.5069,
+      "step": 16660
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 3.396414342629482e-06,
+      "loss": 0.3648,
+      "step": 16670
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 3.3864541832669323e-06,
+      "loss": 0.3244,
+      "step": 16680
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 3.3764940239043824e-06,
+      "loss": 0.4121,
+      "step": 16690
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 3.366533864541833e-06,
+      "loss": 0.7469,
+      "step": 16700
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 3.356573705179283e-06,
+      "loss": 0.4817,
+      "step": 16710
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 3.346613545816733e-06,
+      "loss": 0.5267,
+      "step": 16720
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 3.336653386454183e-06,
+      "loss": 0.5436,
+      "step": 16730
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 3.3266932270916337e-06,
+      "loss": 0.3844,
+      "step": 16740
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 3.3167330677290838e-06,
+      "loss": 0.3665,
+      "step": 16750
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 3.306772908366534e-06,
+      "loss": 0.5977,
+      "step": 16760
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 3.2968127490039843e-06,
+      "loss": 0.4835,
+      "step": 16770
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 3.2868525896414344e-06,
+      "loss": 0.4295,
+      "step": 16780
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 3.2768924302788845e-06,
+      "loss": 0.384,
+      "step": 16790
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 3.2669322709163346e-06,
+      "loss": 0.6228,
+      "step": 16800
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 3.256972111553785e-06,
+      "loss": 0.5195,
+      "step": 16810
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 3.247011952191235e-06,
+      "loss": 0.3761,
+      "step": 16820
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 3.2370517928286853e-06,
+      "loss": 0.5723,
+      "step": 16830
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 3.227091633466136e-06,
+      "loss": 0.5137,
+      "step": 16840
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 3.217131474103586e-06,
+      "loss": 0.4093,
+      "step": 16850
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 3.207171314741036e-06,
+      "loss": 0.5546,
+      "step": 16860
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 3.197211155378486e-06,
+      "loss": 0.4617,
+      "step": 16870
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 3.1872509960159366e-06,
+      "loss": 0.3662,
+      "step": 16880
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 3.1772908366533867e-06,
+      "loss": 0.503,
+      "step": 16890
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 3.1673306772908368e-06,
+      "loss": 0.6061,
+      "step": 16900
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 3.1573705179282873e-06,
+      "loss": 0.4235,
+      "step": 16910
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 3.1474103585657373e-06,
+      "loss": 0.6107,
+      "step": 16920
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 3.1374501992031874e-06,
+      "loss": 0.4328,
+      "step": 16930
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 3.1274900398406375e-06,
+      "loss": 0.5346,
+      "step": 16940
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 3.117529880478088e-06,
+      "loss": 0.5466,
+      "step": 16950
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 3.107569721115538e-06,
+      "loss": 0.5708,
+      "step": 16960
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 3.097609561752988e-06,
+      "loss": 0.6335,
+      "step": 16970
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 3.0876494023904387e-06,
+      "loss": 0.5643,
+      "step": 16980
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 3.077689243027889e-06,
+      "loss": 0.5899,
+      "step": 16990
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 3.067729083665339e-06,
+      "loss": 0.5567,
+      "step": 17000
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 3.057768924302789e-06,
+      "loss": 0.4018,
+      "step": 17010
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 3.0478087649402395e-06,
+      "loss": 0.598,
+      "step": 17020
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 3.0378486055776896e-06,
+      "loss": 0.6264,
+      "step": 17030
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 3.0278884462151397e-06,
+      "loss": 0.5363,
+      "step": 17040
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 3.01792828685259e-06,
+      "loss": 0.444,
+      "step": 17050
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 3.0079681274900403e-06,
+      "loss": 0.4705,
+      "step": 17060
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.9980079681274903e-06,
+      "loss": 0.7316,
+      "step": 17070
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.9880478087649404e-06,
+      "loss": 0.5141,
+      "step": 17080
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.978087649402391e-06,
+      "loss": 0.4503,
+      "step": 17090
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.968127490039841e-06,
+      "loss": 0.5346,
+      "step": 17100
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.958167330677291e-06,
+      "loss": 0.5459,
+      "step": 17110
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.9482071713147416e-06,
+      "loss": 0.4072,
+      "step": 17120
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.9382470119521917e-06,
+      "loss": 0.4393,
+      "step": 17130
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.928286852589642e-06,
+      "loss": 0.5378,
+      "step": 17140
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.918326693227092e-06,
+      "loss": 0.5621,
+      "step": 17150
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.9083665338645424e-06,
+      "loss": 0.3753,
+      "step": 17160
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.8984063745019925e-06,
+      "loss": 0.6446,
+      "step": 17170
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.8884462151394426e-06,
+      "loss": 0.4302,
+      "step": 17180
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.878486055776893e-06,
+      "loss": 0.5676,
+      "step": 17190
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.868525896414343e-06,
+      "loss": 0.3882,
+      "step": 17200
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.8585657370517933e-06,
+      "loss": 0.5615,
+      "step": 17210
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.8486055776892433e-06,
+      "loss": 0.4734,
+      "step": 17220
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.838645418326694e-06,
+      "loss": 0.4726,
+      "step": 17230
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.828685258964144e-06,
+      "loss": 0.4573,
+      "step": 17240
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.818725099601594e-06,
+      "loss": 0.352,
+      "step": 17250
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.8087649402390437e-06,
+      "loss": 0.5791,
+      "step": 17260
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.7988047808764938e-06,
+      "loss": 0.5656,
+      "step": 17270
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.7888446215139443e-06,
+      "loss": 0.5345,
+      "step": 17280
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.7788844621513944e-06,
+      "loss": 0.4134,
+      "step": 17290
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.7689243027888445e-06,
+      "loss": 0.6257,
+      "step": 17300
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.758964143426295e-06,
+      "loss": 0.4928,
+      "step": 17310
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.749003984063745e-06,
+      "loss": 0.7073,
+      "step": 17320
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.739043824701195e-06,
+      "loss": 0.4397,
+      "step": 17330
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.7290836653386452e-06,
+      "loss": 0.4114,
+      "step": 17340
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.7191235059760957e-06,
+      "loss": 0.4087,
+      "step": 17350
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.709163346613546e-06,
+      "loss": 0.5071,
+      "step": 17360
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.699203187250996e-06,
+      "loss": 0.5657,
+      "step": 17370
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.6892430278884464e-06,
+      "loss": 0.5589,
+      "step": 17380
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.6792828685258965e-06,
+      "loss": 0.6309,
+      "step": 17390
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.6693227091633466e-06,
+      "loss": 0.3281,
+      "step": 17400
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.6593625498007967e-06,
+      "loss": 0.4182,
+      "step": 17410
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.649402390438247e-06,
+      "loss": 0.3686,
+      "step": 17420
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.6394422310756973e-06,
+      "loss": 0.6028,
+      "step": 17430
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.6294820717131474e-06,
+      "loss": 0.4353,
+      "step": 17440
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.619521912350598e-06,
+      "loss": 0.5111,
+      "step": 17450
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.609561752988048e-06,
+      "loss": 0.5888,
+      "step": 17460
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.599601593625498e-06,
+      "loss": 0.4984,
+      "step": 17470
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.589641434262948e-06,
+      "loss": 0.4041,
+      "step": 17480
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.5796812749003987e-06,
+      "loss": 0.4358,
+      "step": 17490
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.5697211155378487e-06,
+      "loss": 0.4403,
+      "step": 17500
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.559760956175299e-06,
+      "loss": 0.6421,
+      "step": 17510
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.5498007968127493e-06,
+      "loss": 0.5042,
+      "step": 17520
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.5398406374501994e-06,
+      "loss": 0.3647,
+      "step": 17530
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.5298804780876495e-06,
+      "loss": 0.4954,
+      "step": 17540
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.5199203187250996e-06,
+      "loss": 0.3949,
+      "step": 17550
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.50996015936255e-06,
+      "loss": 0.6486,
+      "step": 17560
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.5e-06,
+      "loss": 0.5872,
+      "step": 17570
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.4900398406374503e-06,
+      "loss": 0.478,
+      "step": 17580
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.480079681274901e-06,
+      "loss": 0.7052,
+      "step": 17590
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.470119521912351e-06,
+      "loss": 0.7672,
+      "step": 17600
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.460159362549801e-06,
+      "loss": 0.5348,
+      "step": 17610
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.450199203187251e-06,
+      "loss": 0.4957,
+      "step": 17620
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.4402390438247016e-06,
+      "loss": 0.5885,
+      "step": 17630
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.4302788844621517e-06,
+      "loss": 0.5404,
+      "step": 17640
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.4203187250996017e-06,
+      "loss": 0.2981,
+      "step": 17650
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.4103585657370523e-06,
+      "loss": 0.6734,
+      "step": 17660
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.4003984063745023e-06,
+      "loss": 0.7027,
+      "step": 17670
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.3904382470119524e-06,
+      "loss": 0.3376,
+      "step": 17680
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.3804780876494025e-06,
+      "loss": 0.5217,
+      "step": 17690
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.370517928286853e-06,
+      "loss": 0.4542,
+      "step": 17700
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.360557768924303e-06,
+      "loss": 0.5874,
+      "step": 17710
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.350597609561753e-06,
+      "loss": 0.5207,
+      "step": 17720
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.3406374501992033e-06,
+      "loss": 0.2966,
+      "step": 17730
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.3306772908366534e-06,
+      "loss": 0.5118,
+      "step": 17740
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.3207171314741035e-06,
+      "loss": 0.6041,
+      "step": 17750
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.310756972111554e-06,
+      "loss": 0.3993,
+      "step": 17760
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.300796812749004e-06,
+      "loss": 0.4822,
+      "step": 17770
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.290836653386454e-06,
+      "loss": 0.479,
+      "step": 17780
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.2808764940239047e-06,
+      "loss": 0.3099,
+      "step": 17790
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.2709163346613547e-06,
+      "loss": 0.4412,
+      "step": 17800
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.260956175298805e-06,
+      "loss": 0.5802,
+      "step": 17810
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.250996015936255e-06,
+      "loss": 0.5592,
+      "step": 17820
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.2410358565737054e-06,
+      "loss": 0.5247,
+      "step": 17830
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.2310756972111555e-06,
+      "loss": 0.5257,
+      "step": 17840
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.2211155378486056e-06,
+      "loss": 0.2897,
+      "step": 17850
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.211155378486056e-06,
+      "loss": 0.579,
+      "step": 17860
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.201195219123506e-06,
+      "loss": 0.5301,
+      "step": 17870
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.1912350597609563e-06,
+      "loss": 0.4541,
+      "step": 17880
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.1812749003984064e-06,
+      "loss": 0.5475,
+      "step": 17890
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.171314741035857e-06,
+      "loss": 0.5013,
+      "step": 17900
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.161354581673307e-06,
+      "loss": 0.4796,
+      "step": 17910
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.151394422310757e-06,
+      "loss": 0.5106,
+      "step": 17920
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.1414342629482076e-06,
+      "loss": 0.5796,
+      "step": 17930
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.1314741035856577e-06,
+      "loss": 0.3942,
+      "step": 17940
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.1215139442231077e-06,
+      "loss": 0.6188,
+      "step": 17950
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.111553784860558e-06,
+      "loss": 0.4709,
+      "step": 17960
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.1015936254980083e-06,
+      "loss": 0.5582,
+      "step": 17970
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.0916334661354584e-06,
+      "loss": 0.4182,
+      "step": 17980
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.0816733067729085e-06,
+      "loss": 0.4703,
+      "step": 17990
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.071713147410359e-06,
+      "loss": 0.6225,
+      "step": 18000
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.061752988047809e-06,
+      "loss": 0.6051,
+      "step": 18010
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.051792828685259e-06,
+      "loss": 0.3064,
+      "step": 18020
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.0418326693227093e-06,
+      "loss": 0.5314,
+      "step": 18030
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.03187250996016e-06,
+      "loss": 0.4133,
+      "step": 18040
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.0219123505976095e-06,
+      "loss": 0.3675,
+      "step": 18050
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.01195219123506e-06,
+      "loss": 0.5816,
+      "step": 18060
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.00199203187251e-06,
+      "loss": 0.7741,
+      "step": 18070
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.8020465772759351,
+      "eval_loss": 0.5026426315307617,
+      "eval_runtime": 12.9735,
+      "eval_samples_per_second": 218.445,
+      "eval_steps_per_second": 27.363,
+      "step": 18072
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 1.99203187250996e-06,
+      "loss": 0.4301,
+      "step": 18080
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 1.9820717131474102e-06,
+      "loss": 0.4781,
+      "step": 18090
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 1.9721115537848607e-06,
+      "loss": 0.5096,
+      "step": 18100
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 1.962151394422311e-06,
+      "loss": 0.5254,
+      "step": 18110
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 1.952191235059761e-06,
+      "loss": 0.5067,
+      "step": 18120
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 1.9422310756972114e-06,
+      "loss": 0.467,
+      "step": 18130
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 1.9322709163346615e-06,
+      "loss": 0.4926,
+      "step": 18140
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 1.9223107569721116e-06,
+      "loss": 0.5285,
+      "step": 18150
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 1.9123505976095617e-06,
+      "loss": 0.421,
+      "step": 18160
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 1.902390438247012e-06,
+      "loss": 0.412,
+      "step": 18170
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 1.8924302788844623e-06,
+      "loss": 0.6288,
+      "step": 18180
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 1.8824701195219126e-06,
+      "loss": 0.4522,
+      "step": 18190
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 1.8725099601593627e-06,
+      "loss": 0.4245,
+      "step": 18200
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 1.862549800796813e-06,
+      "loss": 0.6394,
+      "step": 18210
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 1.852589641434263e-06,
+      "loss": 0.5844,
+      "step": 18220
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 1.8426294820717134e-06,
+      "loss": 0.5381,
+      "step": 18230
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 1.8326693227091634e-06,
+      "loss": 0.5711,
+      "step": 18240
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 1.8227091633466137e-06,
+      "loss": 0.3873,
+      "step": 18250
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 1.812749003984064e-06,
+      "loss": 0.5811,
+      "step": 18260
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 1.8027888446215141e-06,
+      "loss": 0.5592,
+      "step": 18270
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 1.7928286852589644e-06,
+      "loss": 0.5231,
+      "step": 18280
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 1.7828685258964145e-06,
+      "loss": 0.4418,
+      "step": 18290
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 1.7729083665338648e-06,
+      "loss": 0.4434,
+      "step": 18300
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 1.762948207171315e-06,
+      "loss": 0.509,
+      "step": 18310
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 1.7529880478087652e-06,
+      "loss": 0.5703,
+      "step": 18320
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 1.7430278884462155e-06,
+      "loss": 0.3696,
+      "step": 18330
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 1.7330677290836656e-06,
+      "loss": 0.5819,
+      "step": 18340
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 1.7231075697211159e-06,
+      "loss": 0.4706,
+      "step": 18350
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 1.7131474103585658e-06,
+      "loss": 0.6293,
+      "step": 18360
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 1.7031872509960158e-06,
+      "loss": 0.5069,
+      "step": 18370
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 1.6932270916334661e-06,
+      "loss": 0.4487,
+      "step": 18380
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 1.6832669322709164e-06,
+      "loss": 0.6765,
+      "step": 18390
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 1.6733067729083665e-06,
+      "loss": 0.4756,
+      "step": 18400
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 1.6633466135458168e-06,
+      "loss": 0.4949,
+      "step": 18410
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 1.653386454183267e-06,
+      "loss": 0.4847,
+      "step": 18420
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 1.6434262948207172e-06,
+      "loss": 0.3738,
+      "step": 18430
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 1.6334661354581673e-06,
+      "loss": 0.4346,
+      "step": 18440
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 1.6235059760956176e-06,
+      "loss": 0.5203,
+      "step": 18450
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 1.613545816733068e-06,
+      "loss": 0.4491,
+      "step": 18460
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 1.603585657370518e-06,
+      "loss": 0.5585,
+      "step": 18470
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 1.5936254980079683e-06,
+      "loss": 0.3489,
+      "step": 18480
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 1.5836653386454184e-06,
+      "loss": 0.4117,
+      "step": 18490
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 1.5737051792828687e-06,
+      "loss": 0.4464,
+      "step": 18500
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 1.5637450199203188e-06,
+      "loss": 0.512,
+      "step": 18510
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 1.553784860557769e-06,
+      "loss": 0.6989,
+      "step": 18520
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 1.5438247011952194e-06,
+      "loss": 0.6097,
+      "step": 18530
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 1.5338645418326694e-06,
+      "loss": 0.6298,
+      "step": 18540
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 1.5239043824701197e-06,
+      "loss": 0.475,
+      "step": 18550
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 1.5139442231075698e-06,
+      "loss": 0.444,
+      "step": 18560
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 1.5039840637450201e-06,
+      "loss": 0.4242,
+      "step": 18570
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 1.4940239043824702e-06,
+      "loss": 0.486,
+      "step": 18580
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 1.4840637450199205e-06,
+      "loss": 0.4435,
+      "step": 18590
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 1.4741035856573708e-06,
+      "loss": 0.3449,
+      "step": 18600
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 1.464143426294821e-06,
+      "loss": 0.4791,
+      "step": 18610
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 1.4541832669322712e-06,
+      "loss": 0.6855,
+      "step": 18620
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 1.4442231075697213e-06,
+      "loss": 0.3786,
+      "step": 18630
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 1.4342629482071716e-06,
+      "loss": 0.2834,
+      "step": 18640
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 1.4243027888446217e-06,
+      "loss": 0.4438,
+      "step": 18650
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 1.414342629482072e-06,
+      "loss": 0.4993,
+      "step": 18660
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 1.4043824701195218e-06,
+      "loss": 0.6736,
+      "step": 18670
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 1.3944223107569721e-06,
+      "loss": 0.5658,
+      "step": 18680
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 1.3844621513944222e-06,
+      "loss": 0.5433,
+      "step": 18690
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 1.3745019920318725e-06,
+      "loss": 0.6247,
+      "step": 18700
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 1.3645418326693226e-06,
+      "loss": 0.4513,
+      "step": 18710
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 1.354581673306773e-06,
+      "loss": 0.553,
+      "step": 18720
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.3446215139442232e-06,
+      "loss": 0.6193,
+      "step": 18730
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 1.3346613545816733e-06,
+      "loss": 0.3928,
+      "step": 18740
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.3247011952191236e-06,
+      "loss": 0.6589,
+      "step": 18750
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 1.3147410358565737e-06,
+      "loss": 0.6309,
+      "step": 18760
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.304780876494024e-06,
+      "loss": 0.4566,
+      "step": 18770
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 1.294820717131474e-06,
+      "loss": 0.3944,
+      "step": 18780
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.2848605577689244e-06,
+      "loss": 0.5422,
+      "step": 18790
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 1.2749003984063747e-06,
+      "loss": 0.6068,
+      "step": 18800
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.2649402390438248e-06,
+      "loss": 0.4014,
+      "step": 18810
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 1.254980079681275e-06,
+      "loss": 0.48,
+      "step": 18820
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.2450199203187251e-06,
+      "loss": 0.4813,
+      "step": 18830
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 1.2350597609561754e-06,
+      "loss": 0.643,
+      "step": 18840
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.2250996015936255e-06,
+      "loss": 0.4656,
+      "step": 18850
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 1.2151394422310758e-06,
+      "loss": 0.4535,
+      "step": 18860
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.2051792828685261e-06,
+      "loss": 0.4754,
+      "step": 18870
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 1.1952191235059762e-06,
+      "loss": 0.5362,
+      "step": 18880
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.1852589641434265e-06,
+      "loss": 0.4409,
+      "step": 18890
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 1.1752988047808766e-06,
+      "loss": 0.4482,
+      "step": 18900
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.1653386454183267e-06,
+      "loss": 0.4116,
+      "step": 18910
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 1.155378486055777e-06,
+      "loss": 0.482,
+      "step": 18920
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.145418326693227e-06,
+      "loss": 0.4568,
+      "step": 18930
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 1.1354581673306774e-06,
+      "loss": 0.4859,
+      "step": 18940
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.1254980079681275e-06,
+      "loss": 0.3848,
+      "step": 18950
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 1.1155378486055778e-06,
+      "loss": 0.3764,
+      "step": 18960
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.105577689243028e-06,
+      "loss": 0.4438,
+      "step": 18970
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 1.0956175298804781e-06,
+      "loss": 0.5759,
+      "step": 18980
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.0856573705179284e-06,
+      "loss": 0.4626,
+      "step": 18990
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 1.0756972111553785e-06,
+      "loss": 0.4384,
+      "step": 19000
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.0657370517928288e-06,
+      "loss": 0.444,
+      "step": 19010
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 1.055776892430279e-06,
+      "loss": 0.5854,
+      "step": 19020
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.0458167330677292e-06,
+      "loss": 0.3492,
+      "step": 19030
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 1.0358565737051795e-06,
+      "loss": 0.4963,
+      "step": 19040
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.0258964143426296e-06,
+      "loss": 0.4611,
+      "step": 19050
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 1.01593625498008e-06,
+      "loss": 0.5395,
+      "step": 19060
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 1.00597609561753e-06,
+      "loss": 0.414,
+      "step": 19070
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 9.9601593625498e-07,
+      "loss": 0.5948,
+      "step": 19080
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 9.860557768924304e-07,
+      "loss": 0.4222,
+      "step": 19090
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 9.760956175298805e-07,
+      "loss": 0.4281,
+      "step": 19100
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 9.661354581673308e-07,
+      "loss": 0.3709,
+      "step": 19110
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 9.561752988047808e-07,
+      "loss": 0.346,
+      "step": 19120
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 9.462151394422311e-07,
+      "loss": 0.3866,
+      "step": 19130
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 9.362549800796813e-07,
+      "loss": 0.5622,
+      "step": 19140
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 9.262948207171315e-07,
+      "loss": 0.5845,
+      "step": 19150
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 9.163346613545817e-07,
+      "loss": 0.4669,
+      "step": 19160
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 9.06374501992032e-07,
+      "loss": 0.4703,
+      "step": 19170
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 8.964143426294822e-07,
+      "loss": 0.2456,
+      "step": 19180
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 8.864541832669324e-07,
+      "loss": 0.5112,
+      "step": 19190
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 8.764940239043826e-07,
+      "loss": 0.5304,
+      "step": 19200
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 8.665338645418328e-07,
+      "loss": 0.4995,
+      "step": 19210
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 8.565737051792829e-07,
+      "loss": 0.4116,
+      "step": 19220
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 8.466135458167331e-07,
+      "loss": 0.4861,
+      "step": 19230
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 8.366533864541833e-07,
+      "loss": 0.475,
+      "step": 19240
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 8.266932270916335e-07,
+      "loss": 0.5003,
+      "step": 19250
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 8.167330677290837e-07,
+      "loss": 0.5105,
+      "step": 19260
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 8.06772908366534e-07,
+      "loss": 0.5235,
+      "step": 19270
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 7.968127490039841e-07,
+      "loss": 0.59,
+      "step": 19280
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 7.868525896414343e-07,
+      "loss": 0.4664,
+      "step": 19290
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 7.768924302788845e-07,
+      "loss": 0.525,
+      "step": 19300
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 7.669322709163347e-07,
+      "loss": 0.4091,
+      "step": 19310
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 7.569721115537849e-07,
+      "loss": 0.4575,
+      "step": 19320
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 7.470119521912351e-07,
+      "loss": 0.6476,
+      "step": 19330
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 7.370517928286854e-07,
+      "loss": 0.5403,
+      "step": 19340
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 7.270916334661356e-07,
+      "loss": 0.4886,
+      "step": 19350
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 7.171314741035858e-07,
+      "loss": 0.4423,
+      "step": 19360
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 7.07171314741036e-07,
+      "loss": 0.3999,
+      "step": 19370
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 6.972111553784861e-07,
+      "loss": 0.5015,
+      "step": 19380
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 6.872509960159363e-07,
+      "loss": 0.5836,
+      "step": 19390
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 6.772908366533865e-07,
+      "loss": 0.3493,
+      "step": 19400
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 6.673306772908367e-07,
+      "loss": 0.4769,
+      "step": 19410
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 6.573705179282868e-07,
+      "loss": 0.5498,
+      "step": 19420
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 6.47410358565737e-07,
+      "loss": 0.4907,
+      "step": 19430
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 6.374501992031873e-07,
+      "loss": 0.5274,
+      "step": 19440
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 6.274900398406375e-07,
+      "loss": 0.5112,
+      "step": 19450
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 6.175298804780877e-07,
+      "loss": 0.454,
+      "step": 19460
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 6.075697211155379e-07,
+      "loss": 0.5685,
+      "step": 19470
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 5.976095617529881e-07,
+      "loss": 0.4543,
+      "step": 19480
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 5.876494023904383e-07,
+      "loss": 0.3657,
+      "step": 19490
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 5.776892430278885e-07,
+      "loss": 0.5083,
+      "step": 19500
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 5.677290836653387e-07,
+      "loss": 0.7822,
+      "step": 19510
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 5.577689243027889e-07,
+      "loss": 0.5013,
+      "step": 19520
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 5.478087649402391e-07,
+      "loss": 0.4994,
+      "step": 19530
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 5.378486055776893e-07,
+      "loss": 0.5511,
+      "step": 19540
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 5.278884462151395e-07,
+      "loss": 0.4256,
+      "step": 19550
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 5.179282868525898e-07,
+      "loss": 0.5524,
+      "step": 19560
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 5.0796812749004e-07,
+      "loss": 0.6717,
+      "step": 19570
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 4.9800796812749e-07,
+      "loss": 0.4841,
+      "step": 19580
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 4.880478087649402e-07,
+      "loss": 0.4773,
+      "step": 19590
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 4.780876494023904e-07,
+      "loss": 0.3983,
+      "step": 19600
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 4.6812749003984067e-07,
+      "loss": 0.6142,
+      "step": 19610
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 4.5816733067729086e-07,
+      "loss": 0.5228,
+      "step": 19620
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 4.482071713147411e-07,
+      "loss": 0.5548,
+      "step": 19630
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 4.382470119521913e-07,
+      "loss": 0.5297,
+      "step": 19640
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 4.2828685258964144e-07,
+      "loss": 0.4601,
+      "step": 19650
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 4.1832669322709163e-07,
+      "loss": 0.5587,
+      "step": 19660
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 4.083665338645418e-07,
+      "loss": 0.6129,
+      "step": 19670
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 3.9840637450199207e-07,
+      "loss": 0.553,
+      "step": 19680
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 3.8844621513944226e-07,
+      "loss": 0.4228,
+      "step": 19690
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 3.7848605577689246e-07,
+      "loss": 0.4654,
+      "step": 19700
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 3.685258964143427e-07,
+      "loss": 0.6356,
+      "step": 19710
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 3.585657370517929e-07,
+      "loss": 0.6593,
+      "step": 19720
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 3.4860557768924304e-07,
+      "loss": 0.4893,
+      "step": 19730
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 3.3864541832669323e-07,
+      "loss": 0.558,
+      "step": 19740
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 3.286852589641434e-07,
+      "loss": 0.5378,
+      "step": 19750
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 3.1872509960159367e-07,
+      "loss": 0.4484,
+      "step": 19760
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 3.0876494023904386e-07,
+      "loss": 0.4393,
+      "step": 19770
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.9880478087649405e-07,
+      "loss": 0.4581,
+      "step": 19780
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.8884462151394425e-07,
+      "loss": 0.4133,
+      "step": 19790
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.7888446215139444e-07,
+      "loss": 0.4295,
+      "step": 19800
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.6892430278884463e-07,
+      "loss": 0.5205,
+      "step": 19810
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.589641434262949e-07,
+      "loss": 0.5441,
+      "step": 19820
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.49003984063745e-07,
+      "loss": 0.5287,
+      "step": 19830
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.390438247011952e-07,
+      "loss": 0.355,
+      "step": 19840
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.2908366533864543e-07,
+      "loss": 0.3812,
+      "step": 19850
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.1912350597609565e-07,
+      "loss": 0.6408,
+      "step": 19860
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.0916334661354582e-07,
+      "loss": 0.3709,
+      "step": 19870
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 1.9920318725099604e-07,
+      "loss": 0.4777,
+      "step": 19880
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 1.8924302788844623e-07,
+      "loss": 0.3695,
+      "step": 19890
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 1.7928286852589645e-07,
+      "loss": 0.5942,
+      "step": 19900
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 1.6932270916334661e-07,
+      "loss": 0.4631,
+      "step": 19910
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 1.5936254980079683e-07,
+      "loss": 0.5435,
+      "step": 19920
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.4940239043824703e-07,
+      "loss": 0.5133,
+      "step": 19930
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 1.3944223107569722e-07,
+      "loss": 0.4818,
+      "step": 19940
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.2948207171314744e-07,
+      "loss": 0.3684,
+      "step": 19950
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 1.195219123505976e-07,
+      "loss": 0.3934,
+      "step": 19960
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 1.0956175298804783e-07,
+      "loss": 0.3967,
+      "step": 19970
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 9.960159362549802e-08,
+      "loss": 0.5681,
+      "step": 19980
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 8.964143426294822e-08,
+      "loss": 0.5905,
+      "step": 19990
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 7.968127490039842e-08,
+      "loss": 0.5304,
+      "step": 20000
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 6.972111553784861e-08,
+      "loss": 0.4843,
+      "step": 20010
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 5.97609561752988e-08,
+      "loss": 0.5102,
+      "step": 20020
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 4.980079681274901e-08,
+      "loss": 0.5088,
+      "step": 20030
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 3.984063745019921e-08,
+      "loss": 0.3727,
+      "step": 20040
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.98804780876494e-08,
+      "loss": 0.754,
+      "step": 20050
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.9920318725099604e-08,
+      "loss": 0.4626,
+      "step": 20060
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 9.960159362549802e-09,
+      "loss": 0.4763,
+      "step": 20070
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.0,
+      "loss": 0.5269,
+      "step": 20080
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.8073394495412844,
+      "eval_loss": 0.5004011988639832,
+      "eval_runtime": 12.511,
+      "eval_samples_per_second": 226.521,
+      "eval_steps_per_second": 28.375,
+      "step": 20080
+    },
+    {
+      "epoch": 10.0,
+      "step": 20080,
+      "total_flos": 1.243945432122882e+19,
+      "train_loss": 0.6558418143674197,
+      "train_runtime": 2165.352,
+      "train_samples_per_second": 74.163,
+      "train_steps_per_second": 9.273
+    }
+  ],
+  "max_steps": 20080,
+  "num_train_epochs": 10,
+  "total_flos": 1.243945432122882e+19,
+  "trial_name": null,
+  "trial_params": null
+}