diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,8 +1,8 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 30.0,
-  "global_step": 10364160,
+  "epoch": 40.0,
+  "global_step": 13818880,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -124704,11 +124704,41575 @@
       "eval_samples_per_second": 1316.115,
       "eval_steps_per_second": 54.839,
       "step": 10364160
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 3.500492804047796e-05,
+      "loss": 2.1213,
+      "step": 10364500
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 3.500420439283068e-05,
+      "loss": 2.1103,
+      "step": 10365000
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 3.50034821924787e-05,
+      "loss": 2.1067,
+      "step": 10365500
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 3.500276143942201e-05,
+      "loss": 2.0826,
+      "step": 10366000
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 3.5002037791774736e-05,
+      "loss": 2.0969,
+      "step": 10366500
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 3.500131414412746e-05,
+      "loss": 2.0776,
+      "step": 10367000
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 3.500059049648018e-05,
+      "loss": 2.0948,
+      "step": 10367500
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 3.49998668488329e-05,
+      "loss": 2.0853,
+      "step": 10368000
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 3.4999143201185625e-05,
+      "loss": 2.0972,
+      "step": 10368500
+    },
+    {
+      "epoch": 30.01,
+      "learning_rate": 3.499841955353835e-05,
+      "loss": 2.0717,
+      "step": 10369000
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 3.499769590589107e-05,
+      "loss": 2.0894,
+      "step": 10369500
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 3.499697225824379e-05,
+      "loss": 2.0816,
+      "step": 10370000
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 3.499624861059652e-05,
+      "loss": 2.0836,
+      "step": 10370500
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 3.499552496294924e-05,
+      "loss": 2.0927,
+      "step": 10371000
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 3.4994801315301965e-05,
+      "loss": 2.1031,
+      "step": 10371500
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 3.499407766765469e-05,
+      "loss": 2.1029,
+      "step": 10372000
+    },
+    {
+      "epoch": 30.02,
+      "learning_rate": 3.499335546730271e-05,
+      "loss": 2.1119,
+      "step": 10372500
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 3.499263181965543e-05,
+      "loss": 2.0929,
+      "step": 10373000
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 3.4991908172008154e-05,
+      "loss": 2.0962,
+      "step": 10373500
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 3.4991184524360877e-05,
+      "loss": 2.0846,
+      "step": 10374000
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 3.499046232400889e-05,
+      "loss": 2.1082,
+      "step": 10374500
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 3.4989738676361614e-05,
+      "loss": 2.1004,
+      "step": 10375000
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 3.4989015028714337e-05,
+      "loss": 2.0722,
+      "step": 10375500
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 3.498829138106706e-05,
+      "loss": 2.0787,
+      "step": 10376000
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 3.498756773341979e-05,
+      "loss": 2.1036,
+      "step": 10376500
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 3.49868455330678e-05,
+      "loss": 2.1034,
+      "step": 10377000
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 3.498612333271582e-05,
+      "loss": 2.0852,
+      "step": 10377500
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 3.498539968506854e-05,
+      "loss": 2.1011,
+      "step": 10378000
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 3.498467603742127e-05,
+      "loss": 2.0806,
+      "step": 10378500
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 3.498395238977399e-05,
+      "loss": 2.0907,
+      "step": 10379000
+    },
+    {
+      "epoch": 30.04,
+      "learning_rate": 3.4983228742126715e-05,
+      "loss": 2.0767,
+      "step": 10379500
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 3.498250509447944e-05,
+      "loss": 2.0895,
+      "step": 10380000
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 3.4981781446832166e-05,
+      "loss": 2.1007,
+      "step": 10380500
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 3.498105779918489e-05,
+      "loss": 2.089,
+      "step": 10381000
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 3.498033415153761e-05,
+      "loss": 2.0967,
+      "step": 10381500
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 3.4979611951185626e-05,
+      "loss": 2.065,
+      "step": 10382000
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 3.497888830353835e-05,
+      "loss": 2.0805,
+      "step": 10382500
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 3.497816465589107e-05,
+      "loss": 2.0907,
+      "step": 10383000
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 3.497744100824379e-05,
+      "loss": 2.0715,
+      "step": 10383500
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 3.4976717360596515e-05,
+      "loss": 2.0941,
+      "step": 10384000
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 3.497599516024454e-05,
+      "loss": 2.0916,
+      "step": 10384500
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 3.497527151259726e-05,
+      "loss": 2.1088,
+      "step": 10385000
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 3.497454786494999e-05,
+      "loss": 2.097,
+      "step": 10385500
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 3.497382421730271e-05,
+      "loss": 2.1012,
+      "step": 10386000
+    },
+    {
+      "epoch": 30.06,
+      "learning_rate": 3.497310056965543e-05,
+      "loss": 2.0997,
+      "step": 10386500
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 3.497237836930345e-05,
+      "loss": 2.1141,
+      "step": 10387000
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 3.497165472165617e-05,
+      "loss": 2.0816,
+      "step": 10387500
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 3.497093107400889e-05,
+      "loss": 2.1172,
+      "step": 10388000
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 3.4970207426361615e-05,
+      "loss": 2.1078,
+      "step": 10388500
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 3.496948377871434e-05,
+      "loss": 2.0878,
+      "step": 10389000
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 3.4968760131067066e-05,
+      "loss": 2.0848,
+      "step": 10389500
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 3.496803793071508e-05,
+      "loss": 2.1086,
+      "step": 10390000
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 3.4967314283067804e-05,
+      "loss": 2.1157,
+      "step": 10390500
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 3.4966590635420526e-05,
+      "loss": 2.0851,
+      "step": 10391000
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 3.496586698777325e-05,
+      "loss": 2.0881,
+      "step": 10391500
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 3.496514334012597e-05,
+      "loss": 2.0887,
+      "step": 10392000
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 3.4964421139773986e-05,
+      "loss": 2.0919,
+      "step": 10392500
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 3.4963697492126715e-05,
+      "loss": 2.097,
+      "step": 10393000
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 3.496297384447944e-05,
+      "loss": 2.0882,
+      "step": 10393500
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 3.496225019683217e-05,
+      "loss": 2.1106,
+      "step": 10394000
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 3.4961529443775476e-05,
+      "loss": 2.0996,
+      "step": 10394500
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 3.49608057961282e-05,
+      "loss": 2.0901,
+      "step": 10395000
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 3.496008214848092e-05,
+      "loss": 2.0748,
+      "step": 10395500
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 3.495935850083364e-05,
+      "loss": 2.1166,
+      "step": 10396000
+    },
+    {
+      "epoch": 30.09,
+      "learning_rate": 3.4958634853186364e-05,
+      "loss": 2.0924,
+      "step": 10396500
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 3.495791120553909e-05,
+      "loss": 2.0888,
+      "step": 10397000
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 3.4957187557891816e-05,
+      "loss": 2.1078,
+      "step": 10397500
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 3.495646391024454e-05,
+      "loss": 2.0979,
+      "step": 10398000
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 3.495574026259726e-05,
+      "loss": 2.1075,
+      "step": 10398500
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 3.495501661494998e-05,
+      "loss": 2.0849,
+      "step": 10399000
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 3.4954292967302705e-05,
+      "loss": 2.0983,
+      "step": 10399500
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 3.495356931965543e-05,
+      "loss": 2.1227,
+      "step": 10400000
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 3.4952845672008156e-05,
+      "loss": 2.102,
+      "step": 10400500
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 3.495212202436088e-05,
+      "loss": 2.1033,
+      "step": 10401000
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 3.49513983767136e-05,
+      "loss": 2.0828,
+      "step": 10401500
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 3.495067472906632e-05,
+      "loss": 2.0795,
+      "step": 10402000
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 3.494995252871434e-05,
+      "loss": 2.0932,
+      "step": 10402500
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 3.494922888106707e-05,
+      "loss": 2.0807,
+      "step": 10403000
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 3.494850523341979e-05,
+      "loss": 2.1094,
+      "step": 10403500
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 3.494778158577251e-05,
+      "loss": 2.0948,
+      "step": 10404000
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 3.494705938542053e-05,
+      "loss": 2.0983,
+      "step": 10404500
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 3.494633573777325e-05,
+      "loss": 2.1104,
+      "step": 10405000
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 3.494561209012597e-05,
+      "loss": 2.077,
+      "step": 10405500
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 3.4944888442478694e-05,
+      "loss": 2.0896,
+      "step": 10406000
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 3.4944164794831416e-05,
+      "loss": 2.1105,
+      "step": 10406500
+    },
+    {
+      "epoch": 30.12,
+      "learning_rate": 3.494344114718414e-05,
+      "loss": 2.0885,
+      "step": 10407000
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 3.494271749953687e-05,
+      "loss": 2.101,
+      "step": 10407500
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 3.494199385188959e-05,
+      "loss": 2.0973,
+      "step": 10408000
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 3.494127165153761e-05,
+      "loss": 2.0951,
+      "step": 10408500
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 3.4940548003890334e-05,
+      "loss": 2.0993,
+      "step": 10409000
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 3.4939824356243056e-05,
+      "loss": 2.1071,
+      "step": 10409500
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 3.493910070859578e-05,
+      "loss": 2.1001,
+      "step": 10410000
+    },
+    {
+      "epoch": 30.13,
+      "learning_rate": 3.49383770609485e-05,
+      "loss": 2.1157,
+      "step": 10410500
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 3.493765341330122e-05,
+      "loss": 2.0842,
+      "step": 10411000
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 3.4936929765653945e-05,
+      "loss": 2.1065,
+      "step": 10411500
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 3.493620611800667e-05,
+      "loss": 2.0963,
+      "step": 10412000
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 3.493548247035939e-05,
+      "loss": 2.1174,
+      "step": 10412500
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 3.493476027000741e-05,
+      "loss": 2.0945,
+      "step": 10413000
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 3.4934036622360134e-05,
+      "loss": 2.0761,
+      "step": 10413500
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 3.4933312974712857e-05,
+      "loss": 2.0952,
+      "step": 10414000
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 3.493258932706558e-05,
+      "loss": 2.096,
+      "step": 10414500
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 3.493186567941831e-05,
+      "loss": 2.0889,
+      "step": 10415000
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 3.4931143479066323e-05,
+      "loss": 2.0994,
+      "step": 10415500
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 3.4930421278714346e-05,
+      "loss": 2.0818,
+      "step": 10416000
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 3.492969763106707e-05,
+      "loss": 2.0852,
+      "step": 10416500
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 3.492897398341979e-05,
+      "loss": 2.1043,
+      "step": 10417000
+    },
+    {
+      "epoch": 30.15,
+      "learning_rate": 3.4928251783067806e-05,
+      "loss": 2.1028,
+      "step": 10417500
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 3.492752813542053e-05,
+      "loss": 2.0783,
+      "step": 10418000
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 3.492680448777325e-05,
+      "loss": 2.0817,
+      "step": 10418500
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 3.492608084012597e-05,
+      "loss": 2.1086,
+      "step": 10419000
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 3.492536008706929e-05,
+      "loss": 2.0912,
+      "step": 10419500
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 3.492463643942201e-05,
+      "loss": 2.0897,
+      "step": 10420000
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 3.492391279177473e-05,
+      "loss": 2.1039,
+      "step": 10420500
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 3.4923189144127455e-05,
+      "loss": 2.1048,
+      "step": 10421000
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 3.492246549648018e-05,
+      "loss": 2.0901,
+      "step": 10421500
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 3.49217418488329e-05,
+      "loss": 2.0923,
+      "step": 10422000
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 3.492101820118562e-05,
+      "loss": 2.0534,
+      "step": 10422500
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 3.492029455353835e-05,
+      "loss": 2.098,
+      "step": 10423000
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 3.4919572353186366e-05,
+      "loss": 2.0857,
+      "step": 10423500
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 3.4918848705539095e-05,
+      "loss": 2.0862,
+      "step": 10424000
+    },
+    {
+      "epoch": 30.17,
+      "learning_rate": 3.491812505789182e-05,
+      "loss": 2.0934,
+      "step": 10424500
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 3.491740141024454e-05,
+      "loss": 2.112,
+      "step": 10425000
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 3.491667776259726e-05,
+      "loss": 2.0991,
+      "step": 10425500
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 3.4915954114949984e-05,
+      "loss": 2.099,
+      "step": 10426000
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 3.4915230467302706e-05,
+      "loss": 2.0994,
+      "step": 10426500
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 3.491450826695072e-05,
+      "loss": 2.1091,
+      "step": 10427000
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 3.4913784619303444e-05,
+      "loss": 2.1217,
+      "step": 10427500
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 3.4913060971656166e-05,
+      "loss": 2.102,
+      "step": 10428000
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 3.4912337324008895e-05,
+      "loss": 2.1201,
+      "step": 10428500
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 3.491161367636162e-05,
+      "loss": 2.097,
+      "step": 10429000
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 3.491089147600963e-05,
+      "loss": 2.1024,
+      "step": 10429500
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 3.4910167828362355e-05,
+      "loss": 2.0938,
+      "step": 10430000
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 3.4909444180715084e-05,
+      "loss": 2.094,
+      "step": 10430500
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 3.4908720533067807e-05,
+      "loss": 2.1125,
+      "step": 10431000
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 3.490799688542053e-05,
+      "loss": 2.1181,
+      "step": 10431500
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 3.490727323777325e-05,
+      "loss": 2.0849,
+      "step": 10432000
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 3.490655103742127e-05,
+      "loss": 2.0887,
+      "step": 10432500
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 3.4905827389773996e-05,
+      "loss": 2.084,
+      "step": 10433000
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 3.490510374212672e-05,
+      "loss": 2.0454,
+      "step": 10433500
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 3.490438154177473e-05,
+      "loss": 2.1311,
+      "step": 10434000
+    },
+    {
+      "epoch": 30.2,
+      "learning_rate": 3.4903657894127456e-05,
+      "loss": 2.118,
+      "step": 10434500
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 3.490293424648018e-05,
+      "loss": 2.0895,
+      "step": 10435000
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 3.49022105988329e-05,
+      "loss": 2.0945,
+      "step": 10435500
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 3.490148695118562e-05,
+      "loss": 2.0768,
+      "step": 10436000
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 3.490076619812894e-05,
+      "loss": 2.1068,
+      "step": 10436500
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 3.490004255048166e-05,
+      "loss": 2.1096,
+      "step": 10437000
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 3.489931890283438e-05,
+      "loss": 2.1142,
+      "step": 10437500
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 3.4898595255187105e-05,
+      "loss": 2.1015,
+      "step": 10438000
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 3.489787305483513e-05,
+      "loss": 2.0993,
+      "step": 10438500
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 3.489714940718785e-05,
+      "loss": 2.0991,
+      "step": 10439000
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 3.489642575954057e-05,
+      "loss": 2.0732,
+      "step": 10439500
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 3.4895702111893294e-05,
+      "loss": 2.091,
+      "step": 10440000
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 3.489497846424602e-05,
+      "loss": 2.0875,
+      "step": 10440500
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 3.4894254816598745e-05,
+      "loss": 2.0889,
+      "step": 10441000
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 3.489353116895147e-05,
+      "loss": 2.1032,
+      "step": 10441500
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 3.489280752130419e-05,
+      "loss": 2.0874,
+      "step": 10442000
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 3.489208387365691e-05,
+      "loss": 2.0921,
+      "step": 10442500
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 3.4891360226009634e-05,
+      "loss": 2.0944,
+      "step": 10443000
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 3.4890636578362356e-05,
+      "loss": 2.0889,
+      "step": 10443500
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 3.488991293071508e-05,
+      "loss": 2.1125,
+      "step": 10444000
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 3.48891892830678e-05,
+      "loss": 2.1107,
+      "step": 10444500
+    },
+    {
+      "epoch": 30.23,
+      "learning_rate": 3.4888468530011116e-05,
+      "loss": 2.135,
+      "step": 10445000
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 3.488774488236384e-05,
+      "loss": 2.1006,
+      "step": 10445500
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 3.488702268201186e-05,
+      "loss": 2.0979,
+      "step": 10446000
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 3.4886300481659876e-05,
+      "loss": 2.1041,
+      "step": 10446500
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 3.48855768340126e-05,
+      "loss": 2.1155,
+      "step": 10447000
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 3.488485318636532e-05,
+      "loss": 2.1172,
+      "step": 10447500
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 3.488412953871805e-05,
+      "loss": 2.0934,
+      "step": 10448000
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 3.488340589107077e-05,
+      "loss": 2.1162,
+      "step": 10448500
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 3.4882682243423494e-05,
+      "loss": 2.1004,
+      "step": 10449000
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 3.4881958595776217e-05,
+      "loss": 2.1148,
+      "step": 10449500
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 3.488123494812894e-05,
+      "loss": 2.101,
+      "step": 10450000
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 3.488051130048166e-05,
+      "loss": 2.0839,
+      "step": 10450500
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 3.487978765283438e-05,
+      "loss": 2.1271,
+      "step": 10451000
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 3.4879064005187105e-05,
+      "loss": 2.1042,
+      "step": 10451500
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 3.487834035753983e-05,
+      "loss": 2.0907,
+      "step": 10452000
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 3.487761815718785e-05,
+      "loss": 2.0969,
+      "step": 10452500
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 3.487689450954057e-05,
+      "loss": 2.0737,
+      "step": 10453000
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 3.48761708618933e-05,
+      "loss": 2.0969,
+      "step": 10453500
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 3.4875447214246023e-05,
+      "loss": 2.0856,
+      "step": 10454000
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 3.4874723566598746e-05,
+      "loss": 2.1022,
+      "step": 10454500
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 3.487399991895147e-05,
+      "loss": 2.0988,
+      "step": 10455000
+    },
+    {
+      "epoch": 30.26,
+      "learning_rate": 3.487327627130419e-05,
+      "loss": 2.1063,
+      "step": 10455500
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 3.487255262365691e-05,
+      "loss": 2.0833,
+      "step": 10456000
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 3.4871828976009635e-05,
+      "loss": 2.1004,
+      "step": 10456500
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 3.487110532836236e-05,
+      "loss": 2.1153,
+      "step": 10457000
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 3.487038168071508e-05,
+      "loss": 2.0634,
+      "step": 10457500
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 3.48696580330678e-05,
+      "loss": 2.0819,
+      "step": 10458000
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 3.4868934385420524e-05,
+      "loss": 2.0988,
+      "step": 10458500
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 3.4868210737773246e-05,
+      "loss": 2.0891,
+      "step": 10459000
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 3.4867487090125975e-05,
+      "loss": 2.1027,
+      "step": 10459500
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 3.48667634424787e-05,
+      "loss": 2.089,
+      "step": 10460000
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 3.4866039794831426e-05,
+      "loss": 2.0912,
+      "step": 10460500
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 3.486531759447944e-05,
+      "loss": 2.0854,
+      "step": 10461000
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 3.4864593946832164e-05,
+      "loss": 2.0956,
+      "step": 10461500
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 3.486387174648018e-05,
+      "loss": 2.106,
+      "step": 10462000
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 3.48631480988329e-05,
+      "loss": 2.0862,
+      "step": 10462500
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 3.4862424451185624e-05,
+      "loss": 2.1113,
+      "step": 10463000
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 3.486170080353835e-05,
+      "loss": 2.1022,
+      "step": 10463500
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 3.4860977155891075e-05,
+      "loss": 2.0718,
+      "step": 10464000
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 3.48602535082438e-05,
+      "loss": 2.1125,
+      "step": 10464500
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 3.485952986059652e-05,
+      "loss": 2.117,
+      "step": 10465000
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 3.485880621294924e-05,
+      "loss": 2.0922,
+      "step": 10465500
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 3.4858082565301964e-05,
+      "loss": 2.1053,
+      "step": 10466000
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 3.485736036494998e-05,
+      "loss": 2.0858,
+      "step": 10466500
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 3.48566367173027e-05,
+      "loss": 2.098,
+      "step": 10467000
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 3.4855913069655424e-05,
+      "loss": 2.0941,
+      "step": 10467500
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 3.485518942200815e-05,
+      "loss": 2.1012,
+      "step": 10468000
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 3.4854465774360875e-05,
+      "loss": 2.1077,
+      "step": 10468500
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 3.4853742126713604e-05,
+      "loss": 2.0932,
+      "step": 10469000
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 3.4853018479066327e-05,
+      "loss": 2.0795,
+      "step": 10469500
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 3.485229483141905e-05,
+      "loss": 2.099,
+      "step": 10470000
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 3.485157118377177e-05,
+      "loss": 2.1142,
+      "step": 10470500
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 3.485084753612449e-05,
+      "loss": 2.089,
+      "step": 10471000
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 3.4850123888477216e-05,
+      "loss": 2.1337,
+      "step": 10471500
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 3.484940168812523e-05,
+      "loss": 2.1182,
+      "step": 10472000
+    },
+    {
+      "epoch": 30.31,
+      "learning_rate": 3.484867804047795e-05,
+      "loss": 2.0872,
+      "step": 10472500
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 3.4847955840125976e-05,
+      "loss": 2.0921,
+      "step": 10473000
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 3.484723363977399e-05,
+      "loss": 2.103,
+      "step": 10473500
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 3.4846509992126713e-05,
+      "loss": 2.1102,
+      "step": 10474000
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 3.4845786344479436e-05,
+      "loss": 2.1045,
+      "step": 10474500
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 3.484506269683216e-05,
+      "loss": 2.1252,
+      "step": 10475000
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 3.484433904918489e-05,
+      "loss": 2.1201,
+      "step": 10475500
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 3.484361540153761e-05,
+      "loss": 2.0907,
+      "step": 10476000
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 3.484289175389033e-05,
+      "loss": 2.0748,
+      "step": 10476500
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 3.4842169553538354e-05,
+      "loss": 2.126,
+      "step": 10477000
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 3.4841445905891076e-05,
+      "loss": 2.0855,
+      "step": 10477500
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 3.48407222582438e-05,
+      "loss": 2.0942,
+      "step": 10478000
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 3.483999861059652e-05,
+      "loss": 2.1122,
+      "step": 10478500
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 3.483927496294924e-05,
+      "loss": 2.0802,
+      "step": 10479000
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 3.4838551315301965e-05,
+      "loss": 2.1032,
+      "step": 10479500
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 3.483782766765469e-05,
+      "loss": 2.0793,
+      "step": 10480000
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 3.483710402000741e-05,
+      "loss": 2.0851,
+      "step": 10480500
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 3.483638037236013e-05,
+      "loss": 2.108,
+      "step": 10481000
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 3.483565961930345e-05,
+      "loss": 2.1147,
+      "step": 10481500
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 3.483493741895146e-05,
+      "loss": 2.1067,
+      "step": 10482000
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 3.4834213771304185e-05,
+      "loss": 2.1116,
+      "step": 10482500
+    },
+    {
+      "epoch": 30.34,
+      "learning_rate": 3.48334915709522e-05,
+      "loss": 2.0866,
+      "step": 10483000
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 3.483276792330493e-05,
+      "loss": 2.1022,
+      "step": 10483500
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 3.483204427565765e-05,
+      "loss": 2.1074,
+      "step": 10484000
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 3.483132062801038e-05,
+      "loss": 2.0932,
+      "step": 10484500
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 3.48305969803631e-05,
+      "loss": 2.1031,
+      "step": 10485000
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 3.4829873332715825e-05,
+      "loss": 2.0863,
+      "step": 10485500
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 3.482914968506855e-05,
+      "loss": 2.0826,
+      "step": 10486000
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 3.482842603742127e-05,
+      "loss": 2.1108,
+      "step": 10486500
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 3.482770238977399e-05,
+      "loss": 2.1278,
+      "step": 10487000
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 3.482698018942201e-05,
+      "loss": 2.1054,
+      "step": 10487500
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 3.482625654177473e-05,
+      "loss": 2.1043,
+      "step": 10488000
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 3.482553289412745e-05,
+      "loss": 2.0743,
+      "step": 10488500
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 3.482480924648018e-05,
+      "loss": 2.1235,
+      "step": 10489000
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 3.4824087046128197e-05,
+      "loss": 2.0896,
+      "step": 10489500
+    },
+    {
+      "epoch": 30.36,
+      "learning_rate": 3.482336339848092e-05,
+      "loss": 2.0856,
+      "step": 10490000
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 3.482263975083364e-05,
+      "loss": 2.1063,
+      "step": 10490500
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 3.482191610318637e-05,
+      "loss": 2.1214,
+      "step": 10491000
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 3.482119245553909e-05,
+      "loss": 2.1202,
+      "step": 10491500
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 3.4820468807891815e-05,
+      "loss": 2.098,
+      "step": 10492000
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 3.481974516024454e-05,
+      "loss": 2.0809,
+      "step": 10492500
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 3.481902295989255e-05,
+      "loss": 2.0923,
+      "step": 10493000
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 3.4818300759540575e-05,
+      "loss": 2.082,
+      "step": 10493500
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 3.48175771118933e-05,
+      "loss": 2.096,
+      "step": 10494000
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 3.481685346424602e-05,
+      "loss": 2.1176,
+      "step": 10494500
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 3.481612981659874e-05,
+      "loss": 2.1086,
+      "step": 10495000
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 3.4815406168951464e-05,
+      "loss": 2.1244,
+      "step": 10495500
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 3.4814682521304186e-05,
+      "loss": 2.0735,
+      "step": 10496000
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 3.481395887365691e-05,
+      "loss": 2.1281,
+      "step": 10496500
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 3.481323522600963e-05,
+      "loss": 2.0921,
+      "step": 10497000
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 3.481251302565765e-05,
+      "loss": 2.1079,
+      "step": 10497500
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 3.4811789378010375e-05,
+      "loss": 2.1079,
+      "step": 10498000
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 3.4811065730363104e-05,
+      "loss": 2.1152,
+      "step": 10498500
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 3.4810342082715826e-05,
+      "loss": 2.1016,
+      "step": 10499000
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 3.480961843506855e-05,
+      "loss": 2.1077,
+      "step": 10499500
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 3.480889478742127e-05,
+      "loss": 2.1101,
+      "step": 10500000
+    },
+    {
+      "epoch": 30.39,
+      "learning_rate": 3.480817113977399e-05,
+      "loss": 2.0933,
+      "step": 10500500
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 3.4807447492126715e-05,
+      "loss": 2.1145,
+      "step": 10501000
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 3.480672384447944e-05,
+      "loss": 2.0921,
+      "step": 10501500
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 3.480600019683216e-05,
+      "loss": 2.1182,
+      "step": 10502000
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 3.480527654918488e-05,
+      "loss": 2.1088,
+      "step": 10502500
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 3.4804552901537604e-05,
+      "loss": 2.1015,
+      "step": 10503000
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 3.480382925389033e-05,
+      "loss": 2.1247,
+      "step": 10503500
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 3.480310705353835e-05,
+      "loss": 2.0996,
+      "step": 10504000
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 3.480238340589107e-05,
+      "loss": 2.1378,
+      "step": 10504500
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 3.480165975824379e-05,
+      "loss": 2.1126,
+      "step": 10505000
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 3.4800937557891815e-05,
+      "loss": 2.0946,
+      "step": 10505500
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 3.480021391024454e-05,
+      "loss": 2.1013,
+      "step": 10506000
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 3.479949026259726e-05,
+      "loss": 2.1005,
+      "step": 10506500
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 3.479876661494998e-05,
+      "loss": 2.0826,
+      "step": 10507000
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 3.4798042967302704e-05,
+      "loss": 2.1098,
+      "step": 10507500
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 3.479731931965543e-05,
+      "loss": 2.121,
+      "step": 10508000
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 3.4796595672008155e-05,
+      "loss": 2.1286,
+      "step": 10508500
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 3.479587202436088e-05,
+      "loss": 2.0955,
+      "step": 10509000
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 3.47951483767136e-05,
+      "loss": 2.1016,
+      "step": 10509500
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 3.4794426176361616e-05,
+      "loss": 2.1107,
+      "step": 10510000
+    },
+    {
+      "epoch": 30.42,
+      "learning_rate": 3.479370252871434e-05,
+      "loss": 2.0876,
+      "step": 10510500
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 3.479297888106706e-05,
+      "loss": 2.0942,
+      "step": 10511000
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 3.479225523341978e-05,
+      "loss": 2.1061,
+      "step": 10511500
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 3.4791531585772504e-05,
+      "loss": 2.1005,
+      "step": 10512000
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 3.4790807938125233e-05,
+      "loss": 2.1172,
+      "step": 10512500
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 3.4790084290477956e-05,
+      "loss": 2.0962,
+      "step": 10513000
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 3.4789360642830685e-05,
+      "loss": 2.0949,
+      "step": 10513500
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 3.478863699518341e-05,
+      "loss": 2.0859,
+      "step": 10514000
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 3.478791334753613e-05,
+      "loss": 2.0992,
+      "step": 10514500
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 3.478718969988885e-05,
+      "loss": 2.0895,
+      "step": 10515000
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 3.478646749953687e-05,
+      "loss": 2.1089,
+      "step": 10515500
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 3.478574385188959e-05,
+      "loss": 2.1144,
+      "step": 10516000
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 3.478502020424231e-05,
+      "loss": 2.101,
+      "step": 10516500
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 3.4784296556595034e-05,
+      "loss": 2.1223,
+      "step": 10517000
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 3.4783572908947756e-05,
+      "loss": 2.1121,
+      "step": 10517500
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 3.4782849261300485e-05,
+      "loss": 2.1003,
+      "step": 10518000
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 3.478212561365321e-05,
+      "loss": 2.0867,
+      "step": 10518500
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 3.478140196600593e-05,
+      "loss": 2.1225,
+      "step": 10519000
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 3.4780679765653945e-05,
+      "loss": 2.1012,
+      "step": 10519500
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 3.4779956118006674e-05,
+      "loss": 2.085,
+      "step": 10520000
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 3.4779232470359396e-05,
+      "loss": 2.1006,
+      "step": 10520500
+    },
+    {
+      "epoch": 30.45,
+      "learning_rate": 3.477850882271212e-05,
+      "loss": 2.1184,
+      "step": 10521000
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 3.4777786622360134e-05,
+      "loss": 2.0995,
+      "step": 10521500
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 3.4777062974712856e-05,
+      "loss": 2.0993,
+      "step": 10522000
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 3.4776339327065585e-05,
+      "loss": 2.1256,
+      "step": 10522500
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 3.477561567941831e-05,
+      "loss": 2.1224,
+      "step": 10523000
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 3.477489347906632e-05,
+      "loss": 2.1225,
+      "step": 10523500
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 3.4774169831419045e-05,
+      "loss": 2.1086,
+      "step": 10524000
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 3.477344618377177e-05,
+      "loss": 2.0947,
+      "step": 10524500
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 3.477272253612449e-05,
+      "loss": 2.0866,
+      "step": 10525000
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 3.477200033577251e-05,
+      "loss": 2.1321,
+      "step": 10525500
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 3.4771276688125234e-05,
+      "loss": 2.1213,
+      "step": 10526000
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 3.4770553040477956e-05,
+      "loss": 2.0891,
+      "step": 10526500
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 3.476982939283068e-05,
+      "loss": 2.1095,
+      "step": 10527000
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 3.476910574518341e-05,
+      "loss": 2.0887,
+      "step": 10527500
+    },
+    {
+      "epoch": 30.47,
+      "learning_rate": 3.476838209753613e-05,
+      "loss": 2.1086,
+      "step": 10528000
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 3.476765844988885e-05,
+      "loss": 2.1144,
+      "step": 10528500
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 3.4766934802241574e-05,
+      "loss": 2.0794,
+      "step": 10529000
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 3.47662111545943e-05,
+      "loss": 2.1258,
+      "step": 10529500
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 3.476548895424231e-05,
+      "loss": 2.1027,
+      "step": 10530000
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 3.4764765306595034e-05,
+      "loss": 2.0945,
+      "step": 10530500
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 3.476404310624306e-05,
+      "loss": 2.1142,
+      "step": 10531000
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 3.476331945859578e-05,
+      "loss": 2.0891,
+      "step": 10531500
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 3.47625958109485e-05,
+      "loss": 2.0964,
+      "step": 10532000
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 3.4761872163301223e-05,
+      "loss": 2.1022,
+      "step": 10532500
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 3.476114996294924e-05,
+      "loss": 2.0993,
+      "step": 10533000
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 3.476042631530196e-05,
+      "loss": 2.0903,
+      "step": 10533500
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 3.4759702667654684e-05,
+      "loss": 2.1145,
+      "step": 10534000
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 3.475897902000741e-05,
+      "loss": 2.1025,
+      "step": 10534500
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 3.4758255372360135e-05,
+      "loss": 2.0829,
+      "step": 10535000
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 3.4757531724712864e-05,
+      "loss": 2.0857,
+      "step": 10535500
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 3.4756808077065586e-05,
+      "loss": 2.0852,
+      "step": 10536000
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 3.475608442941831e-05,
+      "loss": 2.1172,
+      "step": 10536500
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 3.475536078177103e-05,
+      "loss": 2.088,
+      "step": 10537000
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 3.475463713412375e-05,
+      "loss": 2.1164,
+      "step": 10537500
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 3.4753913486476475e-05,
+      "loss": 2.108,
+      "step": 10538000
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 3.47531898388292e-05,
+      "loss": 2.1157,
+      "step": 10538500
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 3.475246619118192e-05,
+      "loss": 2.0807,
+      "step": 10539000
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 3.475174254353464e-05,
+      "loss": 2.0747,
+      "step": 10539500
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 3.4751018895887364e-05,
+      "loss": 2.1153,
+      "step": 10540000
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 3.4750295248240086e-05,
+      "loss": 2.1244,
+      "step": 10540500
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 3.474957304788811e-05,
+      "loss": 2.0827,
+      "step": 10541000
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 3.474884940024083e-05,
+      "loss": 2.0851,
+      "step": 10541500
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 3.474812575259355e-05,
+      "loss": 2.0935,
+      "step": 10542000
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 3.474740210494628e-05,
+      "loss": 2.1113,
+      "step": 10542500
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 3.4746678457299004e-05,
+      "loss": 2.1156,
+      "step": 10543000
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 3.474595770424231e-05,
+      "loss": 2.1006,
+      "step": 10543500
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 3.4745234056595035e-05,
+      "loss": 2.0893,
+      "step": 10544000
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 3.474451185624306e-05,
+      "loss": 2.108,
+      "step": 10544500
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 3.474378820859578e-05,
+      "loss": 2.1079,
+      "step": 10545000
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 3.47430645609485e-05,
+      "loss": 2.1155,
+      "step": 10545500
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 3.4742340913301224e-05,
+      "loss": 2.1003,
+      "step": 10546000
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 3.4741617265653947e-05,
+      "loss": 2.1159,
+      "step": 10546500
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 3.474089506530196e-05,
+      "loss": 2.1369,
+      "step": 10547000
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 3.474017141765469e-05,
+      "loss": 2.0783,
+      "step": 10547500
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 3.473944777000741e-05,
+      "loss": 2.1329,
+      "step": 10548000
+    },
+    {
+      "epoch": 30.53,
+      "learning_rate": 3.4738724122360136e-05,
+      "loss": 2.1144,
+      "step": 10548500
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 3.473800047471286e-05,
+      "loss": 2.1095,
+      "step": 10549000
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 3.473727682706558e-05,
+      "loss": 2.1005,
+      "step": 10549500
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 3.473655317941831e-05,
+      "loss": 2.1084,
+      "step": 10550000
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 3.4735830979066325e-05,
+      "loss": 2.0824,
+      "step": 10550500
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 3.473511022600964e-05,
+      "loss": 2.1026,
+      "step": 10551000
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 3.473438657836236e-05,
+      "loss": 2.0889,
+      "step": 10551500
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 3.4733662930715085e-05,
+      "loss": 2.1154,
+      "step": 10552000
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 3.473293928306781e-05,
+      "loss": 2.0945,
+      "step": 10552500
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 3.473221563542053e-05,
+      "loss": 2.1108,
+      "step": 10553000
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 3.473149198777325e-05,
+      "loss": 2.0928,
+      "step": 10553500
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 3.4730768340125974e-05,
+      "loss": 2.1428,
+      "step": 10554000
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 3.4730044692478696e-05,
+      "loss": 2.103,
+      "step": 10554500
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 3.472932104483142e-05,
+      "loss": 2.1319,
+      "step": 10555000
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 3.472859739718414e-05,
+      "loss": 2.1045,
+      "step": 10555500
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 3.472787374953686e-05,
+      "loss": 2.1106,
+      "step": 10556000
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 3.4727151549184885e-05,
+      "loss": 2.1103,
+      "step": 10556500
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 3.472642790153761e-05,
+      "loss": 2.1271,
+      "step": 10557000
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 3.472570425389033e-05,
+      "loss": 2.1034,
+      "step": 10557500
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 3.472498060624306e-05,
+      "loss": 2.088,
+      "step": 10558000
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 3.472425695859578e-05,
+      "loss": 2.1319,
+      "step": 10558500
+    },
+    {
+      "epoch": 30.56,
+      "learning_rate": 3.47235333109485e-05,
+      "loss": 2.1103,
+      "step": 10559000
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 3.472281111059652e-05,
+      "loss": 2.1127,
+      "step": 10559500
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 3.472208746294924e-05,
+      "loss": 2.1161,
+      "step": 10560000
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 3.472136381530196e-05,
+      "loss": 2.1091,
+      "step": 10560500
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 3.472064016765469e-05,
+      "loss": 2.0902,
+      "step": 10561000
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 3.4719916520007414e-05,
+      "loss": 2.104,
+      "step": 10561500
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 3.4719192872360136e-05,
+      "loss": 2.1171,
+      "step": 10562000
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 3.471846922471286e-05,
+      "loss": 2.0987,
+      "step": 10562500
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 3.4717747024360874e-05,
+      "loss": 2.097,
+      "step": 10563000
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 3.4717023376713596e-05,
+      "loss": 2.0914,
+      "step": 10563500
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 3.471629972906632e-05,
+      "loss": 2.113,
+      "step": 10564000
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 3.471557608141904e-05,
+      "loss": 2.1005,
+      "step": 10564500
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 3.471485243377176e-05,
+      "loss": 2.1241,
+      "step": 10565000
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 3.471412878612449e-05,
+      "loss": 2.0977,
+      "step": 10565500
+    },
+    {
+      "epoch": 30.58,
+      "learning_rate": 3.4713405138477214e-05,
+      "loss": 2.0874,
+      "step": 10566000
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 3.471268149082994e-05,
+      "loss": 2.0743,
+      "step": 10566500
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 3.4711957843182666e-05,
+      "loss": 2.079,
+      "step": 10567000
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 3.471123564283068e-05,
+      "loss": 2.1048,
+      "step": 10567500
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 3.47105119951834e-05,
+      "loss": 2.0906,
+      "step": 10568000
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 3.470978979483142e-05,
+      "loss": 2.0914,
+      "step": 10568500
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 3.470906614718414e-05,
+      "loss": 2.1006,
+      "step": 10569000
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 3.470834249953686e-05,
+      "loss": 2.1233,
+      "step": 10569500
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 3.470761885188959e-05,
+      "loss": 2.093,
+      "step": 10570000
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 3.4706895204242315e-05,
+      "loss": 2.1286,
+      "step": 10570500
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 3.470617155659504e-05,
+      "loss": 2.1026,
+      "step": 10571000
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 3.470544790894776e-05,
+      "loss": 2.0967,
+      "step": 10571500
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 3.470472426130048e-05,
+      "loss": 2.0947,
+      "step": 10572000
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 3.470400061365321e-05,
+      "loss": 2.107,
+      "step": 10572500
+    },
+    {
+      "epoch": 30.6,
+      "learning_rate": 3.470327696600593e-05,
+      "loss": 2.0974,
+      "step": 10573000
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 3.4702553318358655e-05,
+      "loss": 2.0987,
+      "step": 10573500
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 3.470183111800667e-05,
+      "loss": 2.1058,
+      "step": 10574000
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 3.470110747035939e-05,
+      "loss": 2.0925,
+      "step": 10574500
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 3.4700383822712115e-05,
+      "loss": 2.1235,
+      "step": 10575000
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 3.4699660175064844e-05,
+      "loss": 2.0879,
+      "step": 10575500
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 3.4698936527417566e-05,
+      "loss": 2.1142,
+      "step": 10576000
+    },
+    {
+      "epoch": 30.61,
+      "learning_rate": 3.469821287977029e-05,
+      "loss": 2.0795,
+      "step": 10576500
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 3.469748923212301e-05,
+      "loss": 2.1155,
+      "step": 10577000
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 3.469676558447573e-05,
+      "loss": 2.1153,
+      "step": 10577500
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 3.469604338412375e-05,
+      "loss": 2.1202,
+      "step": 10578000
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 3.469532118377177e-05,
+      "loss": 2.1204,
+      "step": 10578500
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 3.469459753612449e-05,
+      "loss": 2.1243,
+      "step": 10579000
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 3.4693873888477215e-05,
+      "loss": 2.112,
+      "step": 10579500
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 3.4693150240829944e-05,
+      "loss": 2.1113,
+      "step": 10580000
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 3.4692426593182666e-05,
+      "loss": 2.0985,
+      "step": 10580500
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 3.469170294553539e-05,
+      "loss": 2.1281,
+      "step": 10581000
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 3.469097929788811e-05,
+      "loss": 2.0971,
+      "step": 10581500
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 3.469025565024083e-05,
+      "loss": 2.1048,
+      "step": 10582000
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 3.4689532002593555e-05,
+      "loss": 2.1022,
+      "step": 10582500
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 3.468880835494628e-05,
+      "loss": 2.1156,
+      "step": 10583000
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 3.4688084707299e-05,
+      "loss": 2.0974,
+      "step": 10583500
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 3.468736105965172e-05,
+      "loss": 2.1066,
+      "step": 10584000
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 3.4686637412004444e-05,
+      "loss": 2.0973,
+      "step": 10584500
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 3.4685915211652467e-05,
+      "loss": 2.1149,
+      "step": 10585000
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 3.468519301130048e-05,
+      "loss": 2.0976,
+      "step": 10585500
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 3.4684469363653204e-05,
+      "loss": 2.1073,
+      "step": 10586000
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 3.4683745716005927e-05,
+      "loss": 2.1206,
+      "step": 10586500
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 3.468302206835865e-05,
+      "loss": 2.1014,
+      "step": 10587000
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 3.468229842071138e-05,
+      "loss": 2.1026,
+      "step": 10587500
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 3.46815747730641e-05,
+      "loss": 2.0665,
+      "step": 10588000
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 3.468085112541682e-05,
+      "loss": 2.0805,
+      "step": 10588500
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 3.4680127477769545e-05,
+      "loss": 2.1147,
+      "step": 10589000
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 3.467940383012227e-05,
+      "loss": 2.0928,
+      "step": 10589500
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 3.467868162977029e-05,
+      "loss": 2.1106,
+      "step": 10590000
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 3.467795798212301e-05,
+      "loss": 2.099,
+      "step": 10590500
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 3.4677234334475734e-05,
+      "loss": 2.1048,
+      "step": 10591000
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 3.4676510686828456e-05,
+      "loss": 2.0968,
+      "step": 10591500
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 3.467578848647647e-05,
+      "loss": 2.1245,
+      "step": 10592000
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 3.4675066286124494e-05,
+      "loss": 2.107,
+      "step": 10592500
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 3.4674342638477216e-05,
+      "loss": 2.0908,
+      "step": 10593000
+    },
+    {
+      "epoch": 30.66,
+      "learning_rate": 3.467362043812523e-05,
+      "loss": 2.0927,
+      "step": 10593500
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 3.4672896790477954e-05,
+      "loss": 2.1311,
+      "step": 10594000
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 3.4672173142830676e-05,
+      "loss": 2.1037,
+      "step": 10594500
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 3.4671449495183405e-05,
+      "loss": 2.123,
+      "step": 10595000
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 3.467072729483142e-05,
+      "loss": 2.1299,
+      "step": 10595500
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 3.467000364718414e-05,
+      "loss": 2.0969,
+      "step": 10596000
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 3.466927999953687e-05,
+      "loss": 2.1337,
+      "step": 10596500
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 3.4668556351889594e-05,
+      "loss": 2.1021,
+      "step": 10597000
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 3.4667832704242316e-05,
+      "loss": 2.1204,
+      "step": 10597500
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 3.466710905659504e-05,
+      "loss": 2.1056,
+      "step": 10598000
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 3.466638540894776e-05,
+      "loss": 2.1403,
+      "step": 10598500
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 3.466566176130048e-05,
+      "loss": 2.1092,
+      "step": 10599000
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 3.46649395609485e-05,
+      "loss": 2.1259,
+      "step": 10599500
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 3.466421591330122e-05,
+      "loss": 2.1083,
+      "step": 10600000
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 3.466349226565394e-05,
+      "loss": 2.1092,
+      "step": 10600500
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 3.466276861800667e-05,
+      "loss": 2.1048,
+      "step": 10601000
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 3.4662044970359394e-05,
+      "loss": 2.1048,
+      "step": 10601500
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 3.466132277000741e-05,
+      "loss": 2.1221,
+      "step": 10602000
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 3.466059912236013e-05,
+      "loss": 2.081,
+      "step": 10602500
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 3.465987547471286e-05,
+      "loss": 2.1052,
+      "step": 10603000
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 3.465915182706558e-05,
+      "loss": 2.107,
+      "step": 10603500
+    },
+    {
+      "epoch": 30.69,
+      "learning_rate": 3.4658428179418305e-05,
+      "loss": 2.0856,
+      "step": 10604000
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 3.465770453177103e-05,
+      "loss": 2.115,
+      "step": 10604500
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 3.465698088412375e-05,
+      "loss": 2.1242,
+      "step": 10605000
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 3.465625723647647e-05,
+      "loss": 2.1285,
+      "step": 10605500
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 3.4655535036124494e-05,
+      "loss": 2.1191,
+      "step": 10606000
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 3.465481138847722e-05,
+      "loss": 2.1342,
+      "step": 10606500
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 3.465408774082994e-05,
+      "loss": 2.0741,
+      "step": 10607000
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 3.465336409318266e-05,
+      "loss": 2.1249,
+      "step": 10607500
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 3.4652640445535383e-05,
+      "loss": 2.1406,
+      "step": 10608000
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 3.4651916797888106e-05,
+      "loss": 2.1068,
+      "step": 10608500
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 3.465119315024083e-05,
+      "loss": 2.112,
+      "step": 10609000
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 3.465047094988885e-05,
+      "loss": 2.0934,
+      "step": 10609500
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 3.464974730224157e-05,
+      "loss": 2.1061,
+      "step": 10610000
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 3.46490236545943e-05,
+      "loss": 2.0968,
+      "step": 10610500
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 3.464830145424232e-05,
+      "loss": 2.1406,
+      "step": 10611000
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 3.464757780659504e-05,
+      "loss": 2.1136,
+      "step": 10611500
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 3.464685415894776e-05,
+      "loss": 2.1141,
+      "step": 10612000
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 3.4646130511300484e-05,
+      "loss": 2.125,
+      "step": 10612500
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 3.46454083109485e-05,
+      "loss": 2.094,
+      "step": 10613000
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 3.464468466330122e-05,
+      "loss": 2.1236,
+      "step": 10613500
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 3.464396101565395e-05,
+      "loss": 2.1084,
+      "step": 10614000
+    },
+    {
+      "epoch": 30.72,
+      "learning_rate": 3.464323736800667e-05,
+      "loss": 2.108,
+      "step": 10614500
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 3.4642513720359395e-05,
+      "loss": 2.1089,
+      "step": 10615000
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 3.464179007271212e-05,
+      "loss": 2.126,
+      "step": 10615500
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 3.464106642506484e-05,
+      "loss": 2.0948,
+      "step": 10616000
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 3.464034277741756e-05,
+      "loss": 2.1059,
+      "step": 10616500
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 3.4639619129770284e-05,
+      "loss": 2.1235,
+      "step": 10617000
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 3.4638896929418306e-05,
+      "loss": 2.0988,
+      "step": 10617500
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 3.463817472906632e-05,
+      "loss": 2.1331,
+      "step": 10618000
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 3.463745108141905e-05,
+      "loss": 2.1103,
+      "step": 10618500
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 3.4636728881067066e-05,
+      "loss": 2.1302,
+      "step": 10619000
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 3.463600523341979e-05,
+      "loss": 2.1352,
+      "step": 10619500
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 3.463528158577251e-05,
+      "loss": 2.134,
+      "step": 10620000
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 3.463455793812523e-05,
+      "loss": 2.0998,
+      "step": 10620500
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 3.4633834290477955e-05,
+      "loss": 2.0873,
+      "step": 10621000
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 3.463311064283068e-05,
+      "loss": 2.1095,
+      "step": 10621500
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 3.46323869951834e-05,
+      "loss": 2.0785,
+      "step": 10622000
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 3.463166334753612e-05,
+      "loss": 2.1043,
+      "step": 10622500
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 3.4630941147184144e-05,
+      "loss": 2.1015,
+      "step": 10623000
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 3.4630217499536867e-05,
+      "loss": 2.1199,
+      "step": 10623500
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 3.462949385188959e-05,
+      "loss": 2.0892,
+      "step": 10624000
+    },
+    {
+      "epoch": 30.75,
+      "learning_rate": 3.462877020424231e-05,
+      "loss": 2.1227,
+      "step": 10624500
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 3.462804655659504e-05,
+      "loss": 2.1063,
+      "step": 10625000
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 3.462732290894776e-05,
+      "loss": 2.1008,
+      "step": 10625500
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 3.4626599261300484e-05,
+      "loss": 2.1084,
+      "step": 10626000
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 3.462587561365321e-05,
+      "loss": 2.082,
+      "step": 10626500
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 3.462515341330122e-05,
+      "loss": 2.1205,
+      "step": 10627000
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 3.4624431212949245e-05,
+      "loss": 2.0917,
+      "step": 10627500
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 3.462370756530197e-05,
+      "loss": 2.0926,
+      "step": 10628000
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 3.462298391765469e-05,
+      "loss": 2.1074,
+      "step": 10628500
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 3.462226027000741e-05,
+      "loss": 2.1121,
+      "step": 10629000
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 3.4621536622360134e-05,
+      "loss": 2.0819,
+      "step": 10629500
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 3.4620812974712856e-05,
+      "loss": 2.1125,
+      "step": 10630000
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 3.462008932706558e-05,
+      "loss": 2.1056,
+      "step": 10630500
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 3.46193656794183e-05,
+      "loss": 2.1028,
+      "step": 10631000
+    },
+    {
+      "epoch": 30.77,
+      "learning_rate": 3.461864203177102e-05,
+      "loss": 2.1043,
+      "step": 10631500
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 3.461791838412375e-05,
+      "loss": 2.0865,
+      "step": 10632000
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 3.461719618377177e-05,
+      "loss": 2.1143,
+      "step": 10632500
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 3.4616472536124496e-05,
+      "loss": 2.0973,
+      "step": 10633000
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 3.461574888847722e-05,
+      "loss": 2.1167,
+      "step": 10633500
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 3.461502524082994e-05,
+      "loss": 2.0848,
+      "step": 10634000
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 3.461430159318266e-05,
+      "loss": 2.1045,
+      "step": 10634500
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 3.4613577945535385e-05,
+      "loss": 2.1214,
+      "step": 10635000
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 3.461285429788811e-05,
+      "loss": 2.1033,
+      "step": 10635500
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 3.461213065024083e-05,
+      "loss": 2.1035,
+      "step": 10636000
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 3.461140844988885e-05,
+      "loss": 2.0896,
+      "step": 10636500
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 3.4610684802241574e-05,
+      "loss": 2.1127,
+      "step": 10637000
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 3.4609961154594296e-05,
+      "loss": 2.1077,
+      "step": 10637500
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 3.460923750694702e-05,
+      "loss": 2.1448,
+      "step": 10638000
+    },
+    {
+      "epoch": 30.79,
+      "learning_rate": 3.4608515306595034e-05,
+      "loss": 2.1095,
+      "step": 10638500
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 3.4607791658947756e-05,
+      "loss": 2.0712,
+      "step": 10639000
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 3.460706801130048e-05,
+      "loss": 2.1032,
+      "step": 10639500
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 3.460634436365321e-05,
+      "loss": 2.1296,
+      "step": 10640000
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 3.460562071600593e-05,
+      "loss": 2.1002,
+      "step": 10640500
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 3.460489851565395e-05,
+      "loss": 2.1017,
+      "step": 10641000
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 3.4604174868006674e-05,
+      "loss": 2.1172,
+      "step": 10641500
+    },
+    {
+      "epoch": 30.8,
+      "learning_rate": 3.4603451220359397e-05,
+      "loss": 2.1023,
+      "step": 10642000
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 3.460272902000741e-05,
+      "loss": 2.1137,
+      "step": 10642500
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 3.4602005372360134e-05,
+      "loss": 2.1161,
+      "step": 10643000
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 3.460128317200815e-05,
+      "loss": 2.1348,
+      "step": 10643500
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 3.460055952436088e-05,
+      "loss": 2.1076,
+      "step": 10644000
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 3.45998358767136e-05,
+      "loss": 2.1262,
+      "step": 10644500
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 3.459911222906632e-05,
+      "loss": 2.0831,
+      "step": 10645000
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 3.4598388581419046e-05,
+      "loss": 2.1251,
+      "step": 10645500
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 3.459766493377177e-05,
+      "loss": 2.1163,
+      "step": 10646000
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 3.459694128612449e-05,
+      "loss": 2.1153,
+      "step": 10646500
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 3.459621763847721e-05,
+      "loss": 2.0995,
+      "step": 10647000
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 3.459549399082994e-05,
+      "loss": 2.1295,
+      "step": 10647500
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 3.4594770343182664e-05,
+      "loss": 2.0819,
+      "step": 10648000
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 3.4594046695535386e-05,
+      "loss": 2.1078,
+      "step": 10648500
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 3.459332304788811e-05,
+      "loss": 2.1318,
+      "step": 10649000
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 3.459259940024083e-05,
+      "loss": 2.1269,
+      "step": 10649500
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 3.459187719988885e-05,
+      "loss": 2.1018,
+      "step": 10650000
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 3.4591153552241575e-05,
+      "loss": 2.1007,
+      "step": 10650500
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 3.45904299045943e-05,
+      "loss": 2.1133,
+      "step": 10651000
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 3.458970625694702e-05,
+      "loss": 2.1134,
+      "step": 10651500
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 3.4588984056595035e-05,
+      "loss": 2.1079,
+      "step": 10652000
+    },
+    {
+      "epoch": 30.83,
+      "learning_rate": 3.458826185624306e-05,
+      "loss": 2.0908,
+      "step": 10652500
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 3.458753820859578e-05,
+      "loss": 2.0961,
+      "step": 10653000
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 3.45868145609485e-05,
+      "loss": 2.1317,
+      "step": 10653500
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 3.4586090913301224e-05,
+      "loss": 2.0974,
+      "step": 10654000
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 3.4585367265653946e-05,
+      "loss": 2.1093,
+      "step": 10654500
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 3.4584643618006675e-05,
+      "loss": 2.0993,
+      "step": 10655000
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 3.45839199703594e-05,
+      "loss": 2.1435,
+      "step": 10655500
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 3.458319777000741e-05,
+      "loss": 2.101,
+      "step": 10656000
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 3.458247556965543e-05,
+      "loss": 2.0932,
+      "step": 10656500
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 3.458175192200816e-05,
+      "loss": 2.1,
+      "step": 10657000
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 3.458102827436088e-05,
+      "loss": 2.0907,
+      "step": 10657500
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 3.45803046267136e-05,
+      "loss": 2.115,
+      "step": 10658000
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 3.4579580979066324e-05,
+      "loss": 2.1094,
+      "step": 10658500
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 3.4578857331419046e-05,
+      "loss": 2.1002,
+      "step": 10659000
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 3.457813368377177e-05,
+      "loss": 2.0962,
+      "step": 10659500
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 3.457741003612449e-05,
+      "loss": 2.095,
+      "step": 10660000
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 3.457668638847721e-05,
+      "loss": 2.1143,
+      "step": 10660500
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 3.4575962740829935e-05,
+      "loss": 2.1077,
+      "step": 10661000
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 3.457523909318266e-05,
+      "loss": 2.1018,
+      "step": 10661500
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 3.457451689283068e-05,
+      "loss": 2.087,
+      "step": 10662000
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 3.4573794692478695e-05,
+      "loss": 2.1108,
+      "step": 10662500
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 3.4573071044831424e-05,
+      "loss": 2.1193,
+      "step": 10663000
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 3.457234739718415e-05,
+      "loss": 2.1091,
+      "step": 10663500
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 3.457162374953687e-05,
+      "loss": 2.1125,
+      "step": 10664000
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 3.457090010188959e-05,
+      "loss": 2.0936,
+      "step": 10664500
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 3.4570176454242313e-05,
+      "loss": 2.1075,
+      "step": 10665000
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 3.4569452806595036e-05,
+      "loss": 2.1165,
+      "step": 10665500
+    },
+    {
+      "epoch": 30.87,
+      "learning_rate": 3.456872915894776e-05,
+      "loss": 2.1046,
+      "step": 10666000
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 3.456800551130048e-05,
+      "loss": 2.1345,
+      "step": 10666500
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 3.456728186365321e-05,
+      "loss": 2.132,
+      "step": 10667000
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 3.456655821600593e-05,
+      "loss": 2.1084,
+      "step": 10667500
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 3.456583601565395e-05,
+      "loss": 2.0939,
+      "step": 10668000
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 3.456511236800667e-05,
+      "loss": 2.099,
+      "step": 10668500
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 3.456438872035939e-05,
+      "loss": 2.103,
+      "step": 10669000
+    },
+    {
+      "epoch": 30.88,
+      "learning_rate": 3.4563665072712114e-05,
+      "loss": 2.1018,
+      "step": 10669500
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 3.456294142506484e-05,
+      "loss": 2.1097,
+      "step": 10670000
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 3.456221922471286e-05,
+      "loss": 2.1088,
+      "step": 10670500
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 3.456149557706558e-05,
+      "loss": 2.1186,
+      "step": 10671000
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 3.456077192941831e-05,
+      "loss": 2.1183,
+      "step": 10671500
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 3.456004828177103e-05,
+      "loss": 2.0985,
+      "step": 10672000
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 3.4559324634123754e-05,
+      "loss": 2.0993,
+      "step": 10672500
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 3.4558600986476476e-05,
+      "loss": 2.1145,
+      "step": 10673000
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 3.45578773388292e-05,
+      "loss": 2.1013,
+      "step": 10673500
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 3.455715369118192e-05,
+      "loss": 2.1034,
+      "step": 10674000
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 3.4556431490829936e-05,
+      "loss": 2.0987,
+      "step": 10674500
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 3.455570784318266e-05,
+      "loss": 2.1138,
+      "step": 10675000
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 3.455498419553538e-05,
+      "loss": 2.0997,
+      "step": 10675500
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 3.455426054788811e-05,
+      "loss": 2.1045,
+      "step": 10676000
+    },
+    {
+      "epoch": 30.9,
+      "learning_rate": 3.455353690024083e-05,
+      "loss": 2.1221,
+      "step": 10676500
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 3.4552813252593554e-05,
+      "loss": 2.1028,
+      "step": 10677000
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 3.455208960494628e-05,
+      "loss": 2.0933,
+      "step": 10677500
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 3.4551365957299005e-05,
+      "loss": 2.0702,
+      "step": 10678000
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 3.455064230965173e-05,
+      "loss": 2.123,
+      "step": 10678500
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 3.454991866200445e-05,
+      "loss": 2.1145,
+      "step": 10679000
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 3.454919501435717e-05,
+      "loss": 2.1137,
+      "step": 10679500
+    },
+    {
+      "epoch": 30.91,
+      "learning_rate": 3.4548471366709894e-05,
+      "loss": 2.1112,
+      "step": 10680000
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 3.454775061365321e-05,
+      "loss": 2.0965,
+      "step": 10680500
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 3.454702696600593e-05,
+      "loss": 2.1289,
+      "step": 10681000
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 3.4546303318358654e-05,
+      "loss": 2.0981,
+      "step": 10681500
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 3.4545579670711377e-05,
+      "loss": 2.1094,
+      "step": 10682000
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 3.45448560230641e-05,
+      "loss": 2.1064,
+      "step": 10682500
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 3.4544133822712114e-05,
+      "loss": 2.0846,
+      "step": 10683000
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 3.454341017506484e-05,
+      "loss": 2.145,
+      "step": 10683500
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 3.454268652741756e-05,
+      "loss": 2.1017,
+      "step": 10684000
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 3.454196287977028e-05,
+      "loss": 2.1485,
+      "step": 10684500
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 3.454123923212301e-05,
+      "loss": 2.0905,
+      "step": 10685000
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 3.454051703177103e-05,
+      "loss": 2.1336,
+      "step": 10685500
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 3.4539793384123755e-05,
+      "loss": 2.0857,
+      "step": 10686000
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 3.453907118377177e-05,
+      "loss": 2.1163,
+      "step": 10686500
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 3.453834753612449e-05,
+      "loss": 2.1486,
+      "step": 10687000
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 3.453762533577251e-05,
+      "loss": 2.108,
+      "step": 10687500
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 3.453690168812524e-05,
+      "loss": 2.1128,
+      "step": 10688000
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 3.453617804047796e-05,
+      "loss": 2.0781,
+      "step": 10688500
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 3.453545439283068e-05,
+      "loss": 2.0972,
+      "step": 10689000
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 3.4534730745183404e-05,
+      "loss": 2.1034,
+      "step": 10689500
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 3.4534007097536126e-05,
+      "loss": 2.1139,
+      "step": 10690000
+    },
+    {
+      "epoch": 30.94,
+      "learning_rate": 3.453328344988885e-05,
+      "loss": 2.1235,
+      "step": 10690500
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 3.453255980224157e-05,
+      "loss": 2.113,
+      "step": 10691000
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 3.453183615459429e-05,
+      "loss": 2.109,
+      "step": 10691500
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 3.4531112506947015e-05,
+      "loss": 2.1155,
+      "step": 10692000
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 3.4530388859299744e-05,
+      "loss": 2.1178,
+      "step": 10692500
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 3.4529665211652466e-05,
+      "loss": 2.0883,
+      "step": 10693000
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 3.452894301130049e-05,
+      "loss": 2.0839,
+      "step": 10693500
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 3.452821936365321e-05,
+      "loss": 2.109,
+      "step": 10694000
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 3.452749571600593e-05,
+      "loss": 2.0958,
+      "step": 10694500
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 3.452677351565395e-05,
+      "loss": 2.0982,
+      "step": 10695000
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 3.452604986800667e-05,
+      "loss": 2.1025,
+      "step": 10695500
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 3.4525327667654686e-05,
+      "loss": 2.0891,
+      "step": 10696000
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 3.452460402000741e-05,
+      "loss": 2.0991,
+      "step": 10696500
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 3.452388181965543e-05,
+      "loss": 2.1349,
+      "step": 10697000
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 3.452315817200815e-05,
+      "loss": 2.1267,
+      "step": 10697500
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 3.4522434524360875e-05,
+      "loss": 2.1275,
+      "step": 10698000
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 3.45217108767136e-05,
+      "loss": 2.1063,
+      "step": 10698500
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 3.452098722906632e-05,
+      "loss": 2.1058,
+      "step": 10699000
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 3.452026358141904e-05,
+      "loss": 2.1145,
+      "step": 10699500
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 3.4519539933771764e-05,
+      "loss": 2.1079,
+      "step": 10700000
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 3.451881628612449e-05,
+      "loss": 2.1016,
+      "step": 10700500
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 3.4518092638477215e-05,
+      "loss": 2.0703,
+      "step": 10701000
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 3.451736899082994e-05,
+      "loss": 2.1236,
+      "step": 10701500
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 3.451664534318266e-05,
+      "loss": 2.1111,
+      "step": 10702000
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 3.451592314283068e-05,
+      "loss": 2.1126,
+      "step": 10702500
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 3.45152009424787e-05,
+      "loss": 2.1096,
+      "step": 10703000
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 3.451447729483142e-05,
+      "loss": 2.1062,
+      "step": 10703500
+    },
+    {
+      "epoch": 30.98,
+      "learning_rate": 3.451375364718414e-05,
+      "loss": 2.0982,
+      "step": 10704000
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 3.4513029999536865e-05,
+      "loss": 2.0822,
+      "step": 10704500
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 3.451230635188959e-05,
+      "loss": 2.1035,
+      "step": 10705000
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 3.451158415153761e-05,
+      "loss": 2.1218,
+      "step": 10705500
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 3.451086050389033e-05,
+      "loss": 2.1137,
+      "step": 10706000
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 3.4510136856243054e-05,
+      "loss": 2.1128,
+      "step": 10706500
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 3.4509413208595776e-05,
+      "loss": 2.1017,
+      "step": 10707000
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 3.45086895609485e-05,
+      "loss": 2.0891,
+      "step": 10707500
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 3.450796591330123e-05,
+      "loss": 2.1031,
+      "step": 10708000
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 3.450724226565395e-05,
+      "loss": 2.0928,
+      "step": 10708500
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 3.4506520065301965e-05,
+      "loss": 2.1086,
+      "step": 10709000
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 3.450579641765469e-05,
+      "loss": 2.0887,
+      "step": 10709500
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy": 0.66679647998167,
+      "eval_accuracy_mlm": 0.6318043767726271,
+      "eval_accuracy_nsp": 0.8543835045120605,
+      "eval_loss": 2.1734020709991455,
+      "eval_runtime": 331.1869,
+      "eval_samples_per_second": 1317.643,
+      "eval_steps_per_second": 54.903,
+      "step": 10709632
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 3.4505072770007416e-05,
+      "loss": 2.0638,
+      "step": 10710000
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 3.450434912236014e-05,
+      "loss": 2.075,
+      "step": 10710500
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 3.450362547471286e-05,
+      "loss": 2.0787,
+      "step": 10711000
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 3.450290182706558e-05,
+      "loss": 2.0977,
+      "step": 10711500
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 3.4502178179418305e-05,
+      "loss": 2.0766,
+      "step": 10712000
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 3.450145453177103e-05,
+      "loss": 2.071,
+      "step": 10712500
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 3.450073088412375e-05,
+      "loss": 2.0904,
+      "step": 10713000
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 3.450000723647647e-05,
+      "loss": 2.0882,
+      "step": 10713500
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 3.449928648341979e-05,
+      "loss": 2.0788,
+      "step": 10714000
+    },
+    {
+      "epoch": 31.01,
+      "learning_rate": 3.449856283577251e-05,
+      "loss": 2.1167,
+      "step": 10714500
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 3.449783918812523e-05,
+      "loss": 2.0884,
+      "step": 10715000
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 3.449711554047796e-05,
+      "loss": 2.0807,
+      "step": 10715500
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 3.449639189283068e-05,
+      "loss": 2.0777,
+      "step": 10716000
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 3.4495668245183405e-05,
+      "loss": 2.0795,
+      "step": 10716500
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 3.449494459753613e-05,
+      "loss": 2.0772,
+      "step": 10717000
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 3.449422094988885e-05,
+      "loss": 2.0969,
+      "step": 10717500
+    },
+    {
+      "epoch": 31.02,
+      "learning_rate": 3.449349730224157e-05,
+      "loss": 2.1089,
+      "step": 10718000
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 3.4492773654594294e-05,
+      "loss": 2.0837,
+      "step": 10718500
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 3.4492050006947016e-05,
+      "loss": 2.0876,
+      "step": 10719000
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 3.449132635929974e-05,
+      "loss": 2.0689,
+      "step": 10719500
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 3.449060415894776e-05,
+      "loss": 2.0878,
+      "step": 10720000
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 3.448988051130048e-05,
+      "loss": 2.0833,
+      "step": 10720500
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 3.4489156863653206e-05,
+      "loss": 2.12,
+      "step": 10721000
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 3.448843321600593e-05,
+      "loss": 2.0591,
+      "step": 10721500
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 3.448770956835865e-05,
+      "loss": 2.0979,
+      "step": 10722000
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 3.448698736800667e-05,
+      "loss": 2.0614,
+      "step": 10722500
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 3.4486263720359395e-05,
+      "loss": 2.0765,
+      "step": 10723000
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 3.448554152000742e-05,
+      "loss": 2.0922,
+      "step": 10723500
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 3.448481787236014e-05,
+      "loss": 2.0879,
+      "step": 10724000
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 3.448409422471286e-05,
+      "loss": 2.0897,
+      "step": 10724500
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 3.4483370577065584e-05,
+      "loss": 2.1274,
+      "step": 10725000
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 3.4482646929418306e-05,
+      "loss": 2.068,
+      "step": 10725500
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 3.448192472906632e-05,
+      "loss": 2.0966,
+      "step": 10726000
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 3.448120252871434e-05,
+      "loss": 2.1057,
+      "step": 10726500
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 3.4480478881067066e-05,
+      "loss": 2.0975,
+      "step": 10727000
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 3.447975523341979e-05,
+      "loss": 2.1057,
+      "step": 10727500
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 3.447903158577251e-05,
+      "loss": 2.0921,
+      "step": 10728000
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 3.447830793812523e-05,
+      "loss": 2.071,
+      "step": 10728500
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 3.4477584290477955e-05,
+      "loss": 2.086,
+      "step": 10729000
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 3.447686064283068e-05,
+      "loss": 2.0777,
+      "step": 10729500
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 3.4476136995183406e-05,
+      "loss": 2.1007,
+      "step": 10730000
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 3.447541334753613e-05,
+      "loss": 2.0912,
+      "step": 10730500
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 3.447468969988885e-05,
+      "loss": 2.0892,
+      "step": 10731000
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 3.447396605224157e-05,
+      "loss": 2.0664,
+      "step": 10731500
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 3.447324385188959e-05,
+      "loss": 2.1078,
+      "step": 10732000
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 3.447252020424232e-05,
+      "loss": 2.0718,
+      "step": 10732500
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 3.447179655659504e-05,
+      "loss": 2.1038,
+      "step": 10733000
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 3.447107290894776e-05,
+      "loss": 2.0838,
+      "step": 10733500
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 3.447035070859578e-05,
+      "loss": 2.1088,
+      "step": 10734000
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 3.44696270609485e-05,
+      "loss": 2.0947,
+      "step": 10734500
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 3.446890341330122e-05,
+      "loss": 2.0899,
+      "step": 10735000
+    },
+    {
+      "epoch": 31.07,
+      "learning_rate": 3.4468179765653944e-05,
+      "loss": 2.0798,
+      "step": 10735500
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 3.4467456118006666e-05,
+      "loss": 2.1049,
+      "step": 10736000
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 3.446673391765469e-05,
+      "loss": 2.073,
+      "step": 10736500
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 3.446601027000741e-05,
+      "loss": 2.0686,
+      "step": 10737000
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 3.446528662236013e-05,
+      "loss": 2.0981,
+      "step": 10737500
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 3.446456297471286e-05,
+      "loss": 2.0835,
+      "step": 10738000
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 3.4463839327065584e-05,
+      "loss": 2.0774,
+      "step": 10738500
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 3.4463115679418307e-05,
+      "loss": 2.0779,
+      "step": 10739000
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 3.446239203177103e-05,
+      "loss": 2.0736,
+      "step": 10739500
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 3.446166838412375e-05,
+      "loss": 2.0892,
+      "step": 10740000
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 3.446094618377177e-05,
+      "loss": 2.0819,
+      "step": 10740500
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 3.4460222536124496e-05,
+      "loss": 2.0993,
+      "step": 10741000
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 3.445950033577251e-05,
+      "loss": 2.0968,
+      "step": 10741500
+    },
+    {
+      "epoch": 31.09,
+      "learning_rate": 3.4458776688125233e-05,
+      "loss": 2.0609,
+      "step": 10742000
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 3.4458053040477956e-05,
+      "loss": 2.0779,
+      "step": 10742500
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 3.445732939283068e-05,
+      "loss": 2.1065,
+      "step": 10743000
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 3.44566057451834e-05,
+      "loss": 2.1202,
+      "step": 10743500
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 3.445588209753612e-05,
+      "loss": 2.0803,
+      "step": 10744000
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 3.4455158449888845e-05,
+      "loss": 2.092,
+      "step": 10744500
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 3.4454434802241574e-05,
+      "loss": 2.1197,
+      "step": 10745000
+    },
+    {
+      "epoch": 31.1,
+      "learning_rate": 3.4453711154594296e-05,
+      "loss": 2.1096,
+      "step": 10745500
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 3.445298750694702e-05,
+      "loss": 2.0993,
+      "step": 10746000
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 3.445226385929975e-05,
+      "loss": 2.0928,
+      "step": 10746500
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 3.445154021165247e-05,
+      "loss": 2.1068,
+      "step": 10747000
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 3.445081656400519e-05,
+      "loss": 2.1097,
+      "step": 10747500
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 3.4450092916357914e-05,
+      "loss": 2.098,
+      "step": 10748000
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 3.4449369268710636e-05,
+      "loss": 2.0802,
+      "step": 10748500
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 3.444864706835865e-05,
+      "loss": 2.0878,
+      "step": 10749000
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 3.444792486800667e-05,
+      "loss": 2.0926,
+      "step": 10749500
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 3.4447201220359396e-05,
+      "loss": 2.0935,
+      "step": 10750000
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 3.444647757271212e-05,
+      "loss": 2.095,
+      "step": 10750500
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 3.444575392506484e-05,
+      "loss": 2.1155,
+      "step": 10751000
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 3.444503027741756e-05,
+      "loss": 2.1019,
+      "step": 10751500
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 3.444430807706558e-05,
+      "loss": 2.0898,
+      "step": 10752000
+    },
+    {
+      "epoch": 31.12,
+      "learning_rate": 3.444358442941831e-05,
+      "loss": 2.0855,
+      "step": 10752500
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 3.444286078177103e-05,
+      "loss": 2.1153,
+      "step": 10753000
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 3.444213713412375e-05,
+      "loss": 2.1018,
+      "step": 10753500
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 3.444141493377177e-05,
+      "loss": 2.1045,
+      "step": 10754000
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 3.4440691286124496e-05,
+      "loss": 2.0989,
+      "step": 10754500
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 3.443996763847722e-05,
+      "loss": 2.0887,
+      "step": 10755000
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 3.443924688542053e-05,
+      "loss": 2.0815,
+      "step": 10755500
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 3.443852323777325e-05,
+      "loss": 2.0994,
+      "step": 10756000
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 3.443779959012597e-05,
+      "loss": 2.0949,
+      "step": 10756500
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 3.4437075942478694e-05,
+      "loss": 2.1008,
+      "step": 10757000
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 3.443635229483142e-05,
+      "loss": 2.0858,
+      "step": 10757500
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 3.4435628647184145e-05,
+      "loss": 2.0785,
+      "step": 10758000
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 3.443490499953687e-05,
+      "loss": 2.0872,
+      "step": 10758500
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 3.443418135188959e-05,
+      "loss": 2.102,
+      "step": 10759000
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 3.443345770424231e-05,
+      "loss": 2.0882,
+      "step": 10759500
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 3.443273405659504e-05,
+      "loss": 2.0813,
+      "step": 10760000
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 3.4432010408947763e-05,
+      "loss": 2.1039,
+      "step": 10760500
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 3.4431286761300486e-05,
+      "loss": 2.0822,
+      "step": 10761000
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 3.443056311365321e-05,
+      "loss": 2.1055,
+      "step": 10761500
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 3.4429840913301223e-05,
+      "loss": 2.0929,
+      "step": 10762000
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 3.4429117265653946e-05,
+      "loss": 2.0828,
+      "step": 10762500
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 3.442839361800667e-05,
+      "loss": 2.0714,
+      "step": 10763000
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 3.44276699703594e-05,
+      "loss": 2.1033,
+      "step": 10763500
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 3.442694632271212e-05,
+      "loss": 2.0889,
+      "step": 10764000
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 3.442622267506484e-05,
+      "loss": 2.096,
+      "step": 10764500
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 3.4425499027417564e-05,
+      "loss": 2.0881,
+      "step": 10765000
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 3.4424775379770286e-05,
+      "loss": 2.0817,
+      "step": 10765500
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 3.442405173212301e-05,
+      "loss": 2.0926,
+      "step": 10766000
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 3.4423329531771024e-05,
+      "loss": 2.0763,
+      "step": 10766500
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 3.4422605884123746e-05,
+      "loss": 2.0771,
+      "step": 10767000
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 3.4421882236476475e-05,
+      "loss": 2.1136,
+      "step": 10767500
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 3.44211585888292e-05,
+      "loss": 2.0859,
+      "step": 10768000
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 3.442043494118192e-05,
+      "loss": 2.1017,
+      "step": 10768500
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 3.441971274082994e-05,
+      "loss": 2.0705,
+      "step": 10769000
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 3.4418989093182664e-05,
+      "loss": 2.083,
+      "step": 10769500
+    },
+    {
+      "epoch": 31.17,
+      "learning_rate": 3.441826689283068e-05,
+      "loss": 2.0818,
+      "step": 10770000
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 3.44175432451834e-05,
+      "loss": 2.088,
+      "step": 10770500
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 3.4416819597536124e-05,
+      "loss": 2.0957,
+      "step": 10771000
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 3.4416095949888846e-05,
+      "loss": 2.0825,
+      "step": 10771500
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 3.4415372302241575e-05,
+      "loss": 2.095,
+      "step": 10772000
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 3.44146486545943e-05,
+      "loss": 2.0682,
+      "step": 10772500
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 3.441392645424231e-05,
+      "loss": 2.086,
+      "step": 10773000
+    },
+    {
+      "epoch": 31.18,
+      "learning_rate": 3.4413202806595035e-05,
+      "loss": 2.1061,
+      "step": 10773500
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 3.441247915894776e-05,
+      "loss": 2.1071,
+      "step": 10774000
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 3.441175695859577e-05,
+      "loss": 2.1062,
+      "step": 10774500
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 3.4411033310948495e-05,
+      "loss": 2.1015,
+      "step": 10775000
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 3.4410309663301224e-05,
+      "loss": 2.1257,
+      "step": 10775500
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 3.4409586015653946e-05,
+      "loss": 2.1254,
+      "step": 10776000
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 3.4408862368006675e-05,
+      "loss": 2.0726,
+      "step": 10776500
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 3.44081387203594e-05,
+      "loss": 2.0901,
+      "step": 10777000
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 3.440741652000741e-05,
+      "loss": 2.1135,
+      "step": 10777500
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 3.4406692872360136e-05,
+      "loss": 2.0998,
+      "step": 10778000
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 3.440596922471286e-05,
+      "loss": 2.0861,
+      "step": 10778500
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 3.440524557706558e-05,
+      "loss": 2.0851,
+      "step": 10779000
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 3.44045219294183e-05,
+      "loss": 2.1217,
+      "step": 10779500
+    },
+    {
+      "epoch": 31.2,
+      "learning_rate": 3.4403798281771024e-05,
+      "loss": 2.0857,
+      "step": 10780000
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 3.440307463412375e-05,
+      "loss": 2.098,
+      "step": 10780500
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 3.4402350986476476e-05,
+      "loss": 2.0865,
+      "step": 10781000
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 3.44016273388292e-05,
+      "loss": 2.1236,
+      "step": 10781500
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 3.440090369118192e-05,
+      "loss": 2.0979,
+      "step": 10782000
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 3.440018004353465e-05,
+      "loss": 2.0919,
+      "step": 10782500
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 3.439945639588737e-05,
+      "loss": 2.0959,
+      "step": 10783000
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 3.439873419553539e-05,
+      "loss": 2.1093,
+      "step": 10783500
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 3.439801054788811e-05,
+      "loss": 2.1078,
+      "step": 10784000
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 3.439728690024083e-05,
+      "loss": 2.0734,
+      "step": 10784500
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 3.4396563252593554e-05,
+      "loss": 2.1036,
+      "step": 10785000
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 3.4395839604946276e-05,
+      "loss": 2.1082,
+      "step": 10785500
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 3.4395115957299e-05,
+      "loss": 2.1076,
+      "step": 10786000
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 3.439439230965173e-05,
+      "loss": 2.0776,
+      "step": 10786500
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 3.439366866200445e-05,
+      "loss": 2.1019,
+      "step": 10787000
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 3.439294501435717e-05,
+      "loss": 2.0676,
+      "step": 10787500
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 3.439222281400519e-05,
+      "loss": 2.103,
+      "step": 10788000
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 3.439149916635791e-05,
+      "loss": 2.0785,
+      "step": 10788500
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 3.439077551871063e-05,
+      "loss": 2.0874,
+      "step": 10789000
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 3.4390051871063354e-05,
+      "loss": 2.0858,
+      "step": 10789500
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 3.438932822341608e-05,
+      "loss": 2.0758,
+      "step": 10790000
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 3.43886060230641e-05,
+      "loss": 2.0788,
+      "step": 10790500
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 3.438788237541683e-05,
+      "loss": 2.0967,
+      "step": 10791000
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 3.438715872776955e-05,
+      "loss": 2.1101,
+      "step": 10791500
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 3.438643508012227e-05,
+      "loss": 2.1064,
+      "step": 10792000
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 3.438571432706558e-05,
+      "loss": 2.0906,
+      "step": 10792500
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 3.43849906794183e-05,
+      "loss": 2.1126,
+      "step": 10793000
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 3.4384267031771025e-05,
+      "loss": 2.1057,
+      "step": 10793500
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 3.438354338412375e-05,
+      "loss": 2.1015,
+      "step": 10794000
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 3.4382819736476476e-05,
+      "loss": 2.0912,
+      "step": 10794500
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 3.43820960888292e-05,
+      "loss": 2.0883,
+      "step": 10795000
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 3.438137244118192e-05,
+      "loss": 2.0857,
+      "step": 10795500
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 3.438064879353464e-05,
+      "loss": 2.0828,
+      "step": 10796000
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 3.4379925145887365e-05,
+      "loss": 2.0939,
+      "step": 10796500
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 3.437920149824009e-05,
+      "loss": 2.1193,
+      "step": 10797000
+    },
+    {
+      "epoch": 31.25,
+      "learning_rate": 3.437847785059282e-05,
+      "loss": 2.0771,
+      "step": 10797500
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 3.437775420294554e-05,
+      "loss": 2.1036,
+      "step": 10798000
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 3.437703055529826e-05,
+      "loss": 2.119,
+      "step": 10798500
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 3.437630835494628e-05,
+      "loss": 2.0783,
+      "step": 10799000
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 3.43755861545943e-05,
+      "loss": 2.1315,
+      "step": 10799500
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 3.4374863954242315e-05,
+      "loss": 2.0895,
+      "step": 10800000
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 3.437414030659504e-05,
+      "loss": 2.095,
+      "step": 10800500
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 3.437341665894776e-05,
+      "loss": 2.0913,
+      "step": 10801000
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 3.437269301130048e-05,
+      "loss": 2.0927,
+      "step": 10801500
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 3.4371969363653204e-05,
+      "loss": 2.0789,
+      "step": 10802000
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 3.4371245716005926e-05,
+      "loss": 2.1138,
+      "step": 10802500
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 3.4370522068358655e-05,
+      "loss": 2.0896,
+      "step": 10803000
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 3.436979842071138e-05,
+      "loss": 2.1014,
+      "step": 10803500
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 3.436907622035939e-05,
+      "loss": 2.1004,
+      "step": 10804000
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 3.4368352572712115e-05,
+      "loss": 2.0893,
+      "step": 10804500
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 3.4367628925064844e-05,
+      "loss": 2.1175,
+      "step": 10805000
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 3.436690672471286e-05,
+      "loss": 2.1006,
+      "step": 10805500
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 3.436618307706558e-05,
+      "loss": 2.0857,
+      "step": 10806000
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 3.4365459429418304e-05,
+      "loss": 2.1031,
+      "step": 10806500
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 3.4364735781771026e-05,
+      "loss": 2.0897,
+      "step": 10807000
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 3.4364012134123755e-05,
+      "loss": 2.0926,
+      "step": 10807500
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 3.436328848647648e-05,
+      "loss": 2.0875,
+      "step": 10808000
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 3.43625648388292e-05,
+      "loss": 2.0746,
+      "step": 10808500
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 3.4361842638477215e-05,
+      "loss": 2.0896,
+      "step": 10809000
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 3.436111899082994e-05,
+      "loss": 2.0824,
+      "step": 10809500
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 3.436039534318266e-05,
+      "loss": 2.1134,
+      "step": 10810000
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 3.435967169553538e-05,
+      "loss": 2.1017,
+      "step": 10810500
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 3.4358948047888104e-05,
+      "loss": 2.0751,
+      "step": 10811000
+    },
+    {
+      "epoch": 31.29,
+      "learning_rate": 3.4358224400240826e-05,
+      "loss": 2.0961,
+      "step": 10811500
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 3.4357500752593555e-05,
+      "loss": 2.1102,
+      "step": 10812000
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 3.435677710494628e-05,
+      "loss": 2.0795,
+      "step": 10812500
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 3.4356053457299007e-05,
+      "loss": 2.0851,
+      "step": 10813000
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 3.435532980965173e-05,
+      "loss": 2.0845,
+      "step": 10813500
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 3.435460616200445e-05,
+      "loss": 2.0845,
+      "step": 10814000
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 3.435388251435717e-05,
+      "loss": 2.0991,
+      "step": 10814500
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 3.435316031400519e-05,
+      "loss": 2.1168,
+      "step": 10815000
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 3.435243666635791e-05,
+      "loss": 2.098,
+      "step": 10815500
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 3.4351714466005927e-05,
+      "loss": 2.1273,
+      "step": 10816000
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 3.435099226565395e-05,
+      "loss": 2.1069,
+      "step": 10816500
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 3.435026861800667e-05,
+      "loss": 2.09,
+      "step": 10817000
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 3.434954497035939e-05,
+      "loss": 2.1015,
+      "step": 10817500
+    },
+    {
+      "epoch": 31.31,
+      "learning_rate": 3.4348821322712116e-05,
+      "loss": 2.0699,
+      "step": 10818000
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 3.434809767506484e-05,
+      "loss": 2.0949,
+      "step": 10818500
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 3.434737547471285e-05,
+      "loss": 2.1007,
+      "step": 10819000
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 3.434665182706558e-05,
+      "loss": 2.0969,
+      "step": 10819500
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 3.4345928179418305e-05,
+      "loss": 2.0821,
+      "step": 10820000
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 3.4345204531771034e-05,
+      "loss": 2.108,
+      "step": 10820500
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 3.4344480884123756e-05,
+      "loss": 2.1132,
+      "step": 10821000
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 3.434375723647648e-05,
+      "loss": 2.1031,
+      "step": 10821500
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 3.43430335888292e-05,
+      "loss": 2.0709,
+      "step": 10822000
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 3.434230994118192e-05,
+      "loss": 2.0799,
+      "step": 10822500
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 3.4341586293534645e-05,
+      "loss": 2.1011,
+      "step": 10823000
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 3.434086409318266e-05,
+      "loss": 2.1128,
+      "step": 10823500
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 3.434014044553538e-05,
+      "loss": 2.0982,
+      "step": 10824000
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 3.4339416797888105e-05,
+      "loss": 2.1034,
+      "step": 10824500
+    },
+    {
+      "epoch": 31.33,
+      "learning_rate": 3.4338693150240834e-05,
+      "loss": 2.0936,
+      "step": 10825000
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 3.4337969502593556e-05,
+      "loss": 2.1077,
+      "step": 10825500
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 3.433724730224157e-05,
+      "loss": 2.1107,
+      "step": 10826000
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 3.4336523654594294e-05,
+      "loss": 2.1106,
+      "step": 10826500
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 3.4335800006947016e-05,
+      "loss": 2.1106,
+      "step": 10827000
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 3.4335076359299745e-05,
+      "loss": 2.0784,
+      "step": 10827500
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 3.433435415894776e-05,
+      "loss": 2.1091,
+      "step": 10828000
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 3.433363051130048e-05,
+      "loss": 2.1297,
+      "step": 10828500
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 3.4332908310948505e-05,
+      "loss": 2.1143,
+      "step": 10829000
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 3.433218466330123e-05,
+      "loss": 2.0899,
+      "step": 10829500
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 3.433146101565395e-05,
+      "loss": 2.1223,
+      "step": 10830000
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 3.433073736800667e-05,
+      "loss": 2.0839,
+      "step": 10830500
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 3.4330013720359394e-05,
+      "loss": 2.0801,
+      "step": 10831000
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 3.432929152000741e-05,
+      "loss": 2.0864,
+      "step": 10831500
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 3.432856787236013e-05,
+      "loss": 2.0825,
+      "step": 10832000
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 3.4327845672008154e-05,
+      "loss": 2.1311,
+      "step": 10832500
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 3.4327122024360876e-05,
+      "loss": 2.0945,
+      "step": 10833000
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 3.432639982400889e-05,
+      "loss": 2.0941,
+      "step": 10833500
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 3.4325676176361614e-05,
+      "loss": 2.1079,
+      "step": 10834000
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 3.4324952528714337e-05,
+      "loss": 2.055,
+      "step": 10834500
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 3.432422888106706e-05,
+      "loss": 2.1063,
+      "step": 10835000
+    },
+    {
+      "epoch": 31.36,
+      "learning_rate": 3.432350523341979e-05,
+      "loss": 2.0908,
+      "step": 10835500
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 3.432278158577251e-05,
+      "loss": 2.1023,
+      "step": 10836000
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 3.432205793812523e-05,
+      "loss": 2.0847,
+      "step": 10836500
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 3.4321334290477954e-05,
+      "loss": 2.0946,
+      "step": 10837000
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 3.4320610642830683e-05,
+      "loss": 2.0747,
+      "step": 10837500
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 3.4319886995183406e-05,
+      "loss": 2.1078,
+      "step": 10838000
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 3.431916334753613e-05,
+      "loss": 2.0835,
+      "step": 10838500
+    },
+    {
+      "epoch": 31.37,
+      "learning_rate": 3.431843969988885e-05,
+      "loss": 2.1133,
+      "step": 10839000
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 3.431771605224157e-05,
+      "loss": 2.1159,
+      "step": 10839500
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 3.4316992404594295e-05,
+      "loss": 2.0959,
+      "step": 10840000
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 3.431627020424231e-05,
+      "loss": 2.1028,
+      "step": 10840500
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 3.431554655659503e-05,
+      "loss": 2.0914,
+      "step": 10841000
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 3.4314822908947755e-05,
+      "loss": 2.0882,
+      "step": 10841500
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 3.4314099261300484e-05,
+      "loss": 2.1068,
+      "step": 10842000
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 3.4313375613653206e-05,
+      "loss": 2.1187,
+      "step": 10842500
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 3.4312651966005935e-05,
+      "loss": 2.0559,
+      "step": 10843000
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 3.431192831835866e-05,
+      "loss": 2.0938,
+      "step": 10843500
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 3.431120611800667e-05,
+      "loss": 2.1121,
+      "step": 10844000
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 3.4310482470359395e-05,
+      "loss": 2.1055,
+      "step": 10844500
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 3.430975882271212e-05,
+      "loss": 2.0949,
+      "step": 10845000
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 3.430903517506484e-05,
+      "loss": 2.1139,
+      "step": 10845500
+    },
+    {
+      "epoch": 31.39,
+      "learning_rate": 3.430831152741756e-05,
+      "loss": 2.1157,
+      "step": 10846000
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 3.4307587879770284e-05,
+      "loss": 2.1047,
+      "step": 10846500
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 3.4306865679418306e-05,
+      "loss": 2.0747,
+      "step": 10847000
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 3.430614347906632e-05,
+      "loss": 2.1091,
+      "step": 10847500
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 3.4305419831419044e-05,
+      "loss": 2.0779,
+      "step": 10848000
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 3.4304696183771766e-05,
+      "loss": 2.0986,
+      "step": 10848500
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 3.430397253612449e-05,
+      "loss": 2.0735,
+      "step": 10849000
+    },
+    {
+      "epoch": 31.4,
+      "learning_rate": 3.430325033577251e-05,
+      "loss": 2.1085,
+      "step": 10849500
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 3.430252668812523e-05,
+      "loss": 2.0823,
+      "step": 10850000
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 3.430180304047796e-05,
+      "loss": 2.0913,
+      "step": 10850500
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 3.4301079392830684e-05,
+      "loss": 2.0983,
+      "step": 10851000
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 3.4300355745183406e-05,
+      "loss": 2.0837,
+      "step": 10851500
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 3.429963209753613e-05,
+      "loss": 2.0772,
+      "step": 10852000
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 3.429890844988885e-05,
+      "loss": 2.1056,
+      "step": 10852500
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 3.429818480224157e-05,
+      "loss": 2.0794,
+      "step": 10853000
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 3.429746260188959e-05,
+      "loss": 2.0961,
+      "step": 10853500
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 3.429673895424231e-05,
+      "loss": 2.0784,
+      "step": 10854000
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 3.429601530659503e-05,
+      "loss": 2.1127,
+      "step": 10854500
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 3.4295293106243056e-05,
+      "loss": 2.0786,
+      "step": 10855000
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 3.429456945859578e-05,
+      "loss": 2.0987,
+      "step": 10855500
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 3.42938458109485e-05,
+      "loss": 2.0833,
+      "step": 10856000
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 3.429312216330122e-05,
+      "loss": 2.1303,
+      "step": 10856500
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 3.4292398515653944e-05,
+      "loss": 2.1016,
+      "step": 10857000
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 3.4291674868006673e-05,
+      "loss": 2.0897,
+      "step": 10857500
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 3.4290951220359396e-05,
+      "loss": 2.0975,
+      "step": 10858000
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 3.429022902000741e-05,
+      "loss": 2.0983,
+      "step": 10858500
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 3.4289505372360134e-05,
+      "loss": 2.1078,
+      "step": 10859000
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 3.428878172471286e-05,
+      "loss": 2.0993,
+      "step": 10859500
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 3.4288058077065585e-05,
+      "loss": 2.1001,
+      "step": 10860000
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 3.428733442941831e-05,
+      "loss": 2.0968,
+      "step": 10860500
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 3.428661078177103e-05,
+      "loss": 2.1032,
+      "step": 10861000
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 3.428588713412375e-05,
+      "loss": 2.0934,
+      "step": 10861500
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 3.4285163486476474e-05,
+      "loss": 2.096,
+      "step": 10862000
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 3.4284439838829196e-05,
+      "loss": 2.0808,
+      "step": 10862500
+    },
+    {
+      "epoch": 31.44,
+      "learning_rate": 3.428371619118192e-05,
+      "loss": 2.0996,
+      "step": 10863000
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 3.428299254353464e-05,
+      "loss": 2.0976,
+      "step": 10863500
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 3.428226889588736e-05,
+      "loss": 2.0983,
+      "step": 10864000
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 3.4281545248240085e-05,
+      "loss": 2.1039,
+      "step": 10864500
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 3.4280821600592814e-05,
+      "loss": 2.0984,
+      "step": 10865000
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 3.4280097952945536e-05,
+      "loss": 2.0732,
+      "step": 10865500
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 3.427937575259356e-05,
+      "loss": 2.0955,
+      "step": 10866000
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 3.427865210494628e-05,
+      "loss": 2.0859,
+      "step": 10866500
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 3.4277928457299e-05,
+      "loss": 2.1031,
+      "step": 10867000
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 3.4277204809651725e-05,
+      "loss": 2.1112,
+      "step": 10867500
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 3.427648260929974e-05,
+      "loss": 2.1046,
+      "step": 10868000
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 3.427575896165246e-05,
+      "loss": 2.1204,
+      "step": 10868500
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 3.4275035314005185e-05,
+      "loss": 2.0923,
+      "step": 10869000
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 3.427431311365321e-05,
+      "loss": 2.1067,
+      "step": 10869500
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 3.427358946600593e-05,
+      "loss": 2.0891,
+      "step": 10870000
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 3.427286581835865e-05,
+      "loss": 2.1113,
+      "step": 10870500
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 3.4272142170711374e-05,
+      "loss": 2.0872,
+      "step": 10871000
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 3.4271418523064096e-05,
+      "loss": 2.0944,
+      "step": 10871500
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 3.427069632271211e-05,
+      "loss": 2.0745,
+      "step": 10872000
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 3.426997267506484e-05,
+      "loss": 2.0872,
+      "step": 10872500
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 3.426924902741756e-05,
+      "loss": 2.0983,
+      "step": 10873000
+    },
+    {
+      "epoch": 31.47,
+      "learning_rate": 3.4268525379770285e-05,
+      "loss": 2.1221,
+      "step": 10873500
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 3.4267801732123014e-05,
+      "loss": 2.1001,
+      "step": 10874000
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 3.426707808447574e-05,
+      "loss": 2.0905,
+      "step": 10874500
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 3.426635588412375e-05,
+      "loss": 2.087,
+      "step": 10875000
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 3.4265632236476474e-05,
+      "loss": 2.1269,
+      "step": 10875500
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 3.42649085888292e-05,
+      "loss": 2.1054,
+      "step": 10876000
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 3.426418494118192e-05,
+      "loss": 2.0945,
+      "step": 10876500
+    },
+    {
+      "epoch": 31.48,
+      "learning_rate": 3.426346274082994e-05,
+      "loss": 2.0911,
+      "step": 10877000
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 3.4262739093182664e-05,
+      "loss": 2.0731,
+      "step": 10877500
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 3.4262015445535386e-05,
+      "loss": 2.096,
+      "step": 10878000
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 3.426129179788811e-05,
+      "loss": 2.094,
+      "step": 10878500
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 3.426056815024083e-05,
+      "loss": 2.0833,
+      "step": 10879000
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 3.425984450259355e-05,
+      "loss": 2.1098,
+      "step": 10879500
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 3.425912085494628e-05,
+      "loss": 2.1133,
+      "step": 10880000
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 3.4258397207299004e-05,
+      "loss": 2.1041,
+      "step": 10880500
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 3.4257673559651726e-05,
+      "loss": 2.1054,
+      "step": 10881000
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 3.425695135929974e-05,
+      "loss": 2.0711,
+      "step": 10881500
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 3.4256227711652464e-05,
+      "loss": 2.0978,
+      "step": 10882000
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 3.4255505511300486e-05,
+      "loss": 2.0898,
+      "step": 10882500
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 3.425478186365321e-05,
+      "loss": 2.1039,
+      "step": 10883000
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 3.425405821600593e-05,
+      "loss": 2.068,
+      "step": 10883500
+    },
+    {
+      "epoch": 31.5,
+      "learning_rate": 3.425333456835865e-05,
+      "loss": 2.0865,
+      "step": 10884000
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 3.4252610920711375e-05,
+      "loss": 2.0796,
+      "step": 10884500
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 3.42518872730641e-05,
+      "loss": 2.084,
+      "step": 10885000
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 3.425116362541682e-05,
+      "loss": 2.0894,
+      "step": 10885500
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 3.425043997776954e-05,
+      "loss": 2.0892,
+      "step": 10886000
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 3.4249716330122264e-05,
+      "loss": 2.1096,
+      "step": 10886500
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 3.424899268247499e-05,
+      "loss": 2.0915,
+      "step": 10887000
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 3.4248270482123015e-05,
+      "loss": 2.0995,
+      "step": 10887500
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 3.424754683447574e-05,
+      "loss": 2.11,
+      "step": 10888000
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 3.424682318682846e-05,
+      "loss": 2.0688,
+      "step": 10888500
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 3.424609953918118e-05,
+      "loss": 2.093,
+      "step": 10889000
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 3.4245375891533904e-05,
+      "loss": 2.1091,
+      "step": 10889500
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 3.4244652243886626e-05,
+      "loss": 2.1233,
+      "step": 10890000
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 3.424392859623935e-05,
+      "loss": 2.0905,
+      "step": 10890500
+    },
+    {
+      "epoch": 31.52,
+      "learning_rate": 3.424320494859207e-05,
+      "loss": 2.0978,
+      "step": 10891000
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 3.424248130094479e-05,
+      "loss": 2.0914,
+      "step": 10891500
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 3.4241757653297515e-05,
+      "loss": 2.0883,
+      "step": 10892000
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 3.4241034005650244e-05,
+      "loss": 2.0893,
+      "step": 10892500
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 3.424031325259355e-05,
+      "loss": 2.1158,
+      "step": 10893000
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 3.4239589604946275e-05,
+      "loss": 2.09,
+      "step": 10893500
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 3.4238865957299e-05,
+      "loss": 2.1035,
+      "step": 10894000
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 3.423814375694701e-05,
+      "loss": 2.0978,
+      "step": 10894500
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 3.423742010929974e-05,
+      "loss": 2.1041,
+      "step": 10895000
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 3.4236696461652465e-05,
+      "loss": 2.1077,
+      "step": 10895500
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 3.4235972814005194e-05,
+      "loss": 2.0975,
+      "step": 10896000
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 3.4235249166357916e-05,
+      "loss": 2.0728,
+      "step": 10896500
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 3.423452551871064e-05,
+      "loss": 2.0961,
+      "step": 10897000
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 3.423380187106336e-05,
+      "loss": 2.0974,
+      "step": 10897500
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 3.423307822341608e-05,
+      "loss": 2.0854,
+      "step": 10898000
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 3.4232354575768805e-05,
+      "loss": 2.0985,
+      "step": 10898500
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 3.423163092812153e-05,
+      "loss": 2.0783,
+      "step": 10899000
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 3.423090728047425e-05,
+      "loss": 2.1012,
+      "step": 10899500
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 3.423018363282697e-05,
+      "loss": 2.0991,
+      "step": 10900000
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 3.4229459985179694e-05,
+      "loss": 2.0948,
+      "step": 10900500
+    },
+    {
+      "epoch": 31.55,
+      "learning_rate": 3.4228736337532416e-05,
+      "loss": 2.0835,
+      "step": 10901000
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 3.4228012689885145e-05,
+      "loss": 2.1164,
+      "step": 10901500
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 3.422728904223787e-05,
+      "loss": 2.0969,
+      "step": 10902000
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 3.4226565394590596e-05,
+      "loss": 2.0923,
+      "step": 10902500
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 3.422584319423861e-05,
+      "loss": 2.0715,
+      "step": 10903000
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 3.4225119546591334e-05,
+      "loss": 2.1102,
+      "step": 10903500
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 3.4224395898944056e-05,
+      "loss": 2.1148,
+      "step": 10904000
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 3.422367369859207e-05,
+      "loss": 2.1002,
+      "step": 10904500
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 3.4222950050944794e-05,
+      "loss": 2.1048,
+      "step": 10905000
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 3.4222226403297516e-05,
+      "loss": 2.1031,
+      "step": 10905500
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 3.422150420294554e-05,
+      "loss": 2.0951,
+      "step": 10906000
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 3.422078055529826e-05,
+      "loss": 2.0905,
+      "step": 10906500
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 3.4220058354946276e-05,
+      "loss": 2.1179,
+      "step": 10907000
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 3.4219334707299e-05,
+      "loss": 2.1117,
+      "step": 10907500
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 3.421861105965172e-05,
+      "loss": 2.0828,
+      "step": 10908000
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 3.421788741200444e-05,
+      "loss": 2.1007,
+      "step": 10908500
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 3.4217163764357165e-05,
+      "loss": 2.1162,
+      "step": 10909000
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 3.4216440116709894e-05,
+      "loss": 2.1099,
+      "step": 10909500
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 3.4215716469062616e-05,
+      "loss": 2.1086,
+      "step": 10910000
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 3.4214992821415345e-05,
+      "loss": 2.0836,
+      "step": 10910500
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 3.421426917376807e-05,
+      "loss": 2.0936,
+      "step": 10911000
+    },
+    {
+      "epoch": 31.58,
+      "learning_rate": 3.421354552612079e-05,
+      "loss": 2.0824,
+      "step": 10911500
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 3.421282187847351e-05,
+      "loss": 2.0822,
+      "step": 10912000
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 3.4212098230826234e-05,
+      "loss": 2.1023,
+      "step": 10912500
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 3.421137458317896e-05,
+      "loss": 2.0839,
+      "step": 10913000
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 3.421065093553168e-05,
+      "loss": 2.1395,
+      "step": 10913500
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 3.42099272878844e-05,
+      "loss": 2.1137,
+      "step": 10914000
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 3.420920364023712e-05,
+      "loss": 2.1097,
+      "step": 10914500
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 3.4208479992589846e-05,
+      "loss": 2.1132,
+      "step": 10915000
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 3.420775634494257e-05,
+      "loss": 2.1244,
+      "step": 10915500
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 3.42070326972953e-05,
+      "loss": 2.0953,
+      "step": 10916000
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 3.420630904964802e-05,
+      "loss": 2.0949,
+      "step": 10916500
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 3.420558540200075e-05,
+      "loss": 2.1194,
+      "step": 10917000
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 3.4204863201648764e-05,
+      "loss": 2.0949,
+      "step": 10917500
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 3.4204139554001486e-05,
+      "loss": 2.1254,
+      "step": 10918000
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 3.42034173536495e-05,
+      "loss": 2.0932,
+      "step": 10918500
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 3.4202693706002224e-05,
+      "loss": 2.0915,
+      "step": 10919000
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 3.4201970058354946e-05,
+      "loss": 2.0977,
+      "step": 10919500
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 3.420124785800297e-05,
+      "loss": 2.1169,
+      "step": 10920000
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 3.4200525657650984e-05,
+      "loss": 2.1015,
+      "step": 10920500
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 3.4199802010003706e-05,
+      "loss": 2.1039,
+      "step": 10921000
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 3.419907836235643e-05,
+      "loss": 2.0963,
+      "step": 10921500
+    },
+    {
+      "epoch": 31.61,
+      "learning_rate": 3.419835471470915e-05,
+      "loss": 2.0729,
+      "step": 10922000
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 3.419763106706187e-05,
+      "loss": 2.1114,
+      "step": 10922500
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 3.4196907419414595e-05,
+      "loss": 2.1,
+      "step": 10923000
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 3.4196183771767324e-05,
+      "loss": 2.0952,
+      "step": 10923500
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 3.4195460124120046e-05,
+      "loss": 2.1091,
+      "step": 10924000
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 3.419473792376807e-05,
+      "loss": 2.1133,
+      "step": 10924500
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 3.4194015723416084e-05,
+      "loss": 2.0939,
+      "step": 10925000
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 3.4193292075768806e-05,
+      "loss": 2.098,
+      "step": 10925500
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 3.419256842812153e-05,
+      "loss": 2.0894,
+      "step": 10926000
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 3.419184478047425e-05,
+      "loss": 2.0987,
+      "step": 10926500
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 3.419112113282697e-05,
+      "loss": 2.0984,
+      "step": 10927000
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 3.4190397485179695e-05,
+      "loss": 2.0829,
+      "step": 10927500
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 3.4189673837532424e-05,
+      "loss": 2.0904,
+      "step": 10928000
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 3.4188950189885146e-05,
+      "loss": 2.1089,
+      "step": 10928500
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 3.418822654223787e-05,
+      "loss": 2.0907,
+      "step": 10929000
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 3.418750289459059e-05,
+      "loss": 2.0969,
+      "step": 10929500
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 3.418677924694331e-05,
+      "loss": 2.0721,
+      "step": 10930000
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 3.4186055599296035e-05,
+      "loss": 2.0954,
+      "step": 10930500
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 3.418533195164876e-05,
+      "loss": 2.1008,
+      "step": 10931000
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 3.418460830400149e-05,
+      "loss": 2.0946,
+      "step": 10931500
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 3.41838861036495e-05,
+      "loss": 2.1146,
+      "step": 10932000
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 3.4183162456002224e-05,
+      "loss": 2.0996,
+      "step": 10932500
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 3.418243880835495e-05,
+      "loss": 2.1003,
+      "step": 10933000
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 3.4181715160707676e-05,
+      "loss": 2.0839,
+      "step": 10933500
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 3.41809915130604e-05,
+      "loss": 2.0881,
+      "step": 10934000
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 3.418026786541312e-05,
+      "loss": 2.1211,
+      "step": 10934500
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 3.417954421776584e-05,
+      "loss": 2.082,
+      "step": 10935000
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 3.417882201741386e-05,
+      "loss": 2.1041,
+      "step": 10935500
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 3.4178099817061873e-05,
+      "loss": 2.0947,
+      "step": 10936000
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 3.4177376169414596e-05,
+      "loss": 2.0974,
+      "step": 10936500
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 3.4176652521767325e-05,
+      "loss": 2.112,
+      "step": 10937000
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 3.417592887412005e-05,
+      "loss": 2.1184,
+      "step": 10937500
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 3.417520522647277e-05,
+      "loss": 2.1018,
+      "step": 10938000
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 3.4174483026120785e-05,
+      "loss": 2.1195,
+      "step": 10938500
+    },
+    {
+      "epoch": 31.66,
+      "learning_rate": 3.417375937847351e-05,
+      "loss": 2.1054,
+      "step": 10939000
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 3.4173035730826236e-05,
+      "loss": 2.0946,
+      "step": 10939500
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 3.417231353047425e-05,
+      "loss": 2.1097,
+      "step": 10940000
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 3.4171589882826974e-05,
+      "loss": 2.096,
+      "step": 10940500
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 3.4170866235179696e-05,
+      "loss": 2.1072,
+      "step": 10941000
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 3.4170142587532425e-05,
+      "loss": 2.1039,
+      "step": 10941500
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 3.416941893988515e-05,
+      "loss": 2.089,
+      "step": 10942000
+    },
+    {
+      "epoch": 31.67,
+      "learning_rate": 3.416869529223787e-05,
+      "loss": 2.0889,
+      "step": 10942500
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 3.416797164459059e-05,
+      "loss": 2.0931,
+      "step": 10943000
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 3.4167247996943314e-05,
+      "loss": 2.1067,
+      "step": 10943500
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 3.4166524349296036e-05,
+      "loss": 2.1005,
+      "step": 10944000
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 3.416580070164876e-05,
+      "loss": 2.0979,
+      "step": 10944500
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 3.416507705400148e-05,
+      "loss": 2.1032,
+      "step": 10945000
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 3.4164356300944796e-05,
+      "loss": 2.0937,
+      "step": 10945500
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 3.416363265329752e-05,
+      "loss": 2.1034,
+      "step": 10946000
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 3.416290900565024e-05,
+      "loss": 2.134,
+      "step": 10946500
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 3.416218535800297e-05,
+      "loss": 2.0973,
+      "step": 10947000
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 3.416146171035569e-05,
+      "loss": 2.0949,
+      "step": 10947500
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 3.4160738062708414e-05,
+      "loss": 2.0857,
+      "step": 10948000
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 3.4160014415061136e-05,
+      "loss": 2.1004,
+      "step": 10948500
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 3.415929076741386e-05,
+      "loss": 2.1149,
+      "step": 10949000
+    },
+    {
+      "epoch": 31.69,
+      "learning_rate": 3.415856711976658e-05,
+      "loss": 2.0953,
+      "step": 10949500
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 3.41578434721193e-05,
+      "loss": 2.1168,
+      "step": 10950000
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 3.4157121271767326e-05,
+      "loss": 2.1018,
+      "step": 10950500
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 3.415639762412005e-05,
+      "loss": 2.1135,
+      "step": 10951000
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 3.415567397647277e-05,
+      "loss": 2.1014,
+      "step": 10951500
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 3.415495032882549e-05,
+      "loss": 2.0938,
+      "step": 10952000
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 3.4154226681178214e-05,
+      "loss": 2.1251,
+      "step": 10952500
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 3.415350303353094e-05,
+      "loss": 2.0813,
+      "step": 10953000
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 3.415277938588366e-05,
+      "loss": 2.1092,
+      "step": 10953500
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 3.415205573823639e-05,
+      "loss": 2.0988,
+      "step": 10954000
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 3.415133209058911e-05,
+      "loss": 2.0968,
+      "step": 10954500
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 3.415060844294183e-05,
+      "loss": 2.0729,
+      "step": 10955000
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 3.4149886242589855e-05,
+      "loss": 2.1287,
+      "step": 10955500
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 3.414916404223787e-05,
+      "loss": 2.0745,
+      "step": 10956000
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 3.4148441841885886e-05,
+      "loss": 2.1094,
+      "step": 10956500
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 3.414771819423861e-05,
+      "loss": 2.124,
+      "step": 10957000
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 3.414699454659133e-05,
+      "loss": 2.1024,
+      "step": 10957500
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 3.414627089894405e-05,
+      "loss": 2.0991,
+      "step": 10958000
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 3.4145547251296775e-05,
+      "loss": 2.1159,
+      "step": 10958500
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 3.4144823603649504e-05,
+      "loss": 2.1087,
+      "step": 10959000
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 3.4144099956002226e-05,
+      "loss": 2.0965,
+      "step": 10959500
+    },
+    {
+      "epoch": 31.72,
+      "learning_rate": 3.414337775565024e-05,
+      "loss": 2.1053,
+      "step": 10960000
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 3.4142654108002964e-05,
+      "loss": 2.0802,
+      "step": 10960500
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 3.4141930460355686e-05,
+      "loss": 2.1218,
+      "step": 10961000
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 3.414120681270841e-05,
+      "loss": 2.0974,
+      "step": 10961500
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 3.414048316506114e-05,
+      "loss": 2.1067,
+      "step": 10962000
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 3.413975951741386e-05,
+      "loss": 2.0764,
+      "step": 10962500
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 3.4139037317061875e-05,
+      "loss": 2.1022,
+      "step": 10963000
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 3.4138313669414604e-05,
+      "loss": 2.1155,
+      "step": 10963500
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 3.4137590021767326e-05,
+      "loss": 2.104,
+      "step": 10964000
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 3.413686637412005e-05,
+      "loss": 2.1135,
+      "step": 10964500
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 3.413614272647277e-05,
+      "loss": 2.0734,
+      "step": 10965000
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 3.413541907882549e-05,
+      "loss": 2.0836,
+      "step": 10965500
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 3.4134695431178215e-05,
+      "loss": 2.085,
+      "step": 10966000
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 3.413397178353094e-05,
+      "loss": 2.0995,
+      "step": 10966500
+    },
+    {
+      "epoch": 31.74,
+      "learning_rate": 3.413324813588366e-05,
+      "loss": 2.0803,
+      "step": 10967000
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 3.4132525935531675e-05,
+      "loss": 2.1066,
+      "step": 10967500
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 3.4131802287884404e-05,
+      "loss": 2.1115,
+      "step": 10968000
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 3.4131078640237127e-05,
+      "loss": 2.0802,
+      "step": 10968500
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 3.4130354992589856e-05,
+      "loss": 2.0897,
+      "step": 10969000
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 3.412963279223787e-05,
+      "loss": 2.0987,
+      "step": 10969500
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 3.412890914459059e-05,
+      "loss": 2.0865,
+      "step": 10970000
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 3.4128185496943316e-05,
+      "loss": 2.073,
+      "step": 10970500
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 3.412746184929604e-05,
+      "loss": 2.0925,
+      "step": 10971000
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 3.412673820164876e-05,
+      "loss": 2.0883,
+      "step": 10971500
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 3.4126016001296776e-05,
+      "loss": 2.1024,
+      "step": 10972000
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 3.4125292353649505e-05,
+      "loss": 2.1092,
+      "step": 10972500
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 3.412456870600223e-05,
+      "loss": 2.0778,
+      "step": 10973000
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 3.412384505835495e-05,
+      "loss": 2.0979,
+      "step": 10973500
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 3.412312141070767e-05,
+      "loss": 2.0916,
+      "step": 10974000
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 3.4122397763060394e-05,
+      "loss": 2.1182,
+      "step": 10974500
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 3.4121674115413116e-05,
+      "loss": 2.1062,
+      "step": 10975000
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 3.412095191506113e-05,
+      "loss": 2.0791,
+      "step": 10975500
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 3.4120228267413854e-05,
+      "loss": 2.1213,
+      "step": 10976000
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 3.4119504619766576e-05,
+      "loss": 2.1095,
+      "step": 10976500
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 3.4118780972119305e-05,
+      "loss": 2.0942,
+      "step": 10977000
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 3.411805732447203e-05,
+      "loss": 2.0898,
+      "step": 10977500
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 3.4117333676824756e-05,
+      "loss": 2.1023,
+      "step": 10978000
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 3.411661002917748e-05,
+      "loss": 2.0885,
+      "step": 10978500
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 3.41158863815302e-05,
+      "loss": 2.1062,
+      "step": 10979000
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 3.4115164181178216e-05,
+      "loss": 2.0876,
+      "step": 10979500
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 3.411444053353094e-05,
+      "loss": 2.1045,
+      "step": 10980000
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 3.411371688588366e-05,
+      "loss": 2.1202,
+      "step": 10980500
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 3.411299323823638e-05,
+      "loss": 2.0842,
+      "step": 10981000
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 3.4112269590589105e-05,
+      "loss": 2.1279,
+      "step": 10981500
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 3.411154739023713e-05,
+      "loss": 2.1003,
+      "step": 10982000
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 3.411082374258985e-05,
+      "loss": 2.112,
+      "step": 10982500
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 3.411010009494257e-05,
+      "loss": 2.0932,
+      "step": 10983000
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 3.410937789459059e-05,
+      "loss": 2.0937,
+      "step": 10983500
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 3.410865424694331e-05,
+      "loss": 2.0832,
+      "step": 10984000
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 3.410793059929604e-05,
+      "loss": 2.0915,
+      "step": 10984500
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 3.410720695164876e-05,
+      "loss": 2.1044,
+      "step": 10985000
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 3.410648330400148e-05,
+      "loss": 2.1193,
+      "step": 10985500
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 3.4105761103649505e-05,
+      "loss": 2.0999,
+      "step": 10986000
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 3.410503745600223e-05,
+      "loss": 2.1316,
+      "step": 10986500
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 3.410431380835495e-05,
+      "loss": 2.0909,
+      "step": 10987000
+    },
+    {
+      "epoch": 31.8,
+      "learning_rate": 3.410359016070767e-05,
+      "loss": 2.103,
+      "step": 10987500
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 3.4102866513060394e-05,
+      "loss": 2.0997,
+      "step": 10988000
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 3.410214431270841e-05,
+      "loss": 2.123,
+      "step": 10988500
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 3.410142066506113e-05,
+      "loss": 2.1264,
+      "step": 10989000
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 3.4100697017413854e-05,
+      "loss": 2.0962,
+      "step": 10989500
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 3.409997336976658e-05,
+      "loss": 2.1111,
+      "step": 10990000
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 3.4099249722119306e-05,
+      "loss": 2.1121,
+      "step": 10990500
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 3.409852752176732e-05,
+      "loss": 2.0882,
+      "step": 10991000
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 3.409780387412004e-05,
+      "loss": 2.1222,
+      "step": 10991500
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 3.409708022647277e-05,
+      "loss": 2.1022,
+      "step": 10992000
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 3.4096356578825495e-05,
+      "loss": 2.1035,
+      "step": 10992500
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 3.409563293117822e-05,
+      "loss": 2.1012,
+      "step": 10993000
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 3.409490928353094e-05,
+      "loss": 2.0714,
+      "step": 10993500
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 3.409418563588366e-05,
+      "loss": 2.0741,
+      "step": 10994000
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 3.4093461988236384e-05,
+      "loss": 2.0775,
+      "step": 10994500
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 3.4092739787884406e-05,
+      "loss": 2.1194,
+      "step": 10995000
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 3.409201758753242e-05,
+      "loss": 2.0808,
+      "step": 10995500
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 3.4091293939885144e-05,
+      "loss": 2.1009,
+      "step": 10996000
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 3.4090570292237866e-05,
+      "loss": 2.1154,
+      "step": 10996500
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 3.408984664459059e-05,
+      "loss": 2.1056,
+      "step": 10997000
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 3.408912299694331e-05,
+      "loss": 2.106,
+      "step": 10997500
+    },
+    {
+      "epoch": 31.83,
+      "learning_rate": 3.408840079659133e-05,
+      "loss": 2.0857,
+      "step": 10998000
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 3.408767859623935e-05,
+      "loss": 2.1097,
+      "step": 10998500
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 3.408695494859207e-05,
+      "loss": 2.0926,
+      "step": 10999000
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 3.40862313009448e-05,
+      "loss": 2.0938,
+      "step": 10999500
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 3.408550765329752e-05,
+      "loss": 2.0924,
+      "step": 11000000
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 3.408478545294554e-05,
+      "loss": 2.0892,
+      "step": 11000500
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 3.408406180529826e-05,
+      "loss": 2.1199,
+      "step": 11001000
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 3.408333815765098e-05,
+      "loss": 2.0798,
+      "step": 11001500
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 3.408261451000371e-05,
+      "loss": 2.1098,
+      "step": 11002000
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 3.408189086235643e-05,
+      "loss": 2.0651,
+      "step": 11002500
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 3.4081167214709155e-05,
+      "loss": 2.0822,
+      "step": 11003000
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 3.408044356706188e-05,
+      "loss": 2.0953,
+      "step": 11003500
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 3.40797199194146e-05,
+      "loss": 2.0927,
+      "step": 11004000
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 3.407899627176732e-05,
+      "loss": 2.1134,
+      "step": 11004500
+    },
+    {
+      "epoch": 31.85,
+      "learning_rate": 3.4078272624120044e-05,
+      "loss": 2.0864,
+      "step": 11005000
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 3.4077548976472766e-05,
+      "loss": 2.0868,
+      "step": 11005500
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 3.407682532882549e-05,
+      "loss": 2.1155,
+      "step": 11006000
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 3.407610168117822e-05,
+      "loss": 2.1002,
+      "step": 11006500
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 3.407537803353094e-05,
+      "loss": 2.1142,
+      "step": 11007000
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 3.407465583317896e-05,
+      "loss": 2.0898,
+      "step": 11007500
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 3.4073932185531684e-05,
+      "loss": 2.1062,
+      "step": 11008000
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 3.407320853788441e-05,
+      "loss": 2.0961,
+      "step": 11008500
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 3.407248489023713e-05,
+      "loss": 2.1142,
+      "step": 11009000
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 3.407176124258985e-05,
+      "loss": 2.0889,
+      "step": 11009500
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 3.407103759494257e-05,
+      "loss": 2.0843,
+      "step": 11010000
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 3.4070313947295296e-05,
+      "loss": 2.0904,
+      "step": 11010500
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 3.406959319423861e-05,
+      "loss": 2.0881,
+      "step": 11011000
+    },
+    {
+      "epoch": 31.87,
+      "learning_rate": 3.4068869546591333e-05,
+      "loss": 2.0938,
+      "step": 11011500
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 3.4068145898944056e-05,
+      "loss": 2.0809,
+      "step": 11012000
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 3.406742225129678e-05,
+      "loss": 2.0756,
+      "step": 11012500
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 3.4066700050944794e-05,
+      "loss": 2.1003,
+      "step": 11013000
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 3.4065976403297516e-05,
+      "loss": 2.1023,
+      "step": 11013500
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 3.406525275565024e-05,
+      "loss": 2.1022,
+      "step": 11014000
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 3.406452910800297e-05,
+      "loss": 2.1127,
+      "step": 11014500
+    },
+    {
+      "epoch": 31.88,
+      "learning_rate": 3.406380546035569e-05,
+      "loss": 2.1,
+      "step": 11015000
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 3.406308326000371e-05,
+      "loss": 2.0899,
+      "step": 11015500
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 3.4062359612356434e-05,
+      "loss": 2.1049,
+      "step": 11016000
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 3.4061635964709156e-05,
+      "loss": 2.123,
+      "step": 11016500
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 3.406091231706188e-05,
+      "loss": 2.1166,
+      "step": 11017000
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 3.40601886694146e-05,
+      "loss": 2.1091,
+      "step": 11017500
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 3.405946502176732e-05,
+      "loss": 2.1262,
+      "step": 11018000
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 3.4058741374120045e-05,
+      "loss": 2.1127,
+      "step": 11018500
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 3.405801772647277e-05,
+      "loss": 2.1186,
+      "step": 11019000
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 3.405729552612078e-05,
+      "loss": 2.1037,
+      "step": 11019500
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 3.405657187847351e-05,
+      "loss": 2.0914,
+      "step": 11020000
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 3.4055848230826234e-05,
+      "loss": 2.1114,
+      "step": 11020500
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 3.4055124583178956e-05,
+      "loss": 2.1115,
+      "step": 11021000
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 3.405440093553168e-05,
+      "loss": 2.1175,
+      "step": 11021500
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 3.405367728788441e-05,
+      "loss": 2.0948,
+      "step": 11022000
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 3.405295508753242e-05,
+      "loss": 2.0919,
+      "step": 11022500
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 3.405223288718044e-05,
+      "loss": 2.1249,
+      "step": 11023000
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 3.405150923953316e-05,
+      "loss": 2.0901,
+      "step": 11023500
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 3.405078559188589e-05,
+      "loss": 2.083,
+      "step": 11024000
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 3.405006194423861e-05,
+      "loss": 2.0982,
+      "step": 11024500
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 3.404933974388663e-05,
+      "loss": 2.1035,
+      "step": 11025000
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 3.404861754353464e-05,
+      "loss": 2.0904,
+      "step": 11025500
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 3.4047893895887365e-05,
+      "loss": 2.0962,
+      "step": 11026000
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 3.404717024824009e-05,
+      "loss": 2.1016,
+      "step": 11026500
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 3.404644660059281e-05,
+      "loss": 2.1133,
+      "step": 11027000
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 3.404572295294554e-05,
+      "loss": 2.1082,
+      "step": 11027500
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 3.404499930529826e-05,
+      "loss": 2.1004,
+      "step": 11028000
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 3.404427565765098e-05,
+      "loss": 2.0961,
+      "step": 11028500
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 3.4043552010003706e-05,
+      "loss": 2.1086,
+      "step": 11029000
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 3.4042828362356435e-05,
+      "loss": 2.1021,
+      "step": 11029500
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 3.404210471470916e-05,
+      "loss": 2.1187,
+      "step": 11030000
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 3.404138106706188e-05,
+      "loss": 2.0988,
+      "step": 11030500
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 3.40406574194146e-05,
+      "loss": 2.1115,
+      "step": 11031000
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 3.4039933771767324e-05,
+      "loss": 2.0857,
+      "step": 11031500
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 3.4039210124120046e-05,
+      "loss": 2.1237,
+      "step": 11032000
+    },
+    {
+      "epoch": 31.93,
+      "learning_rate": 3.403848792376806e-05,
+      "loss": 2.1112,
+      "step": 11032500
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 3.403776427612079e-05,
+      "loss": 2.1057,
+      "step": 11033000
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 3.403704062847351e-05,
+      "loss": 2.0866,
+      "step": 11033500
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 3.4036316980826235e-05,
+      "loss": 2.1,
+      "step": 11034000
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 3.403559333317896e-05,
+      "loss": 2.0885,
+      "step": 11034500
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 3.403486968553168e-05,
+      "loss": 2.0845,
+      "step": 11035000
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 3.40341460378844e-05,
+      "loss": 2.0994,
+      "step": 11035500
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 3.4033422390237124e-05,
+      "loss": 2.0806,
+      "step": 11036000
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 3.403270018988514e-05,
+      "loss": 2.1238,
+      "step": 11036500
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 3.403197654223787e-05,
+      "loss": 2.1197,
+      "step": 11037000
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 3.403125289459059e-05,
+      "loss": 2.1138,
+      "step": 11037500
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 3.403052924694331e-05,
+      "loss": 2.1471,
+      "step": 11038000
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 3.4029807046591335e-05,
+      "loss": 2.0917,
+      "step": 11038500
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 3.402908339894406e-05,
+      "loss": 2.0972,
+      "step": 11039000
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 3.402836119859207e-05,
+      "loss": 2.095,
+      "step": 11039500
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 3.4027637550944795e-05,
+      "loss": 2.1102,
+      "step": 11040000
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 3.402691390329752e-05,
+      "loss": 2.1143,
+      "step": 11040500
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 3.402619025565024e-05,
+      "loss": 2.0817,
+      "step": 11041000
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 3.402546660800296e-05,
+      "loss": 2.0853,
+      "step": 11041500
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 3.402474296035569e-05,
+      "loss": 2.1062,
+      "step": 11042000
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 3.402401931270841e-05,
+      "loss": 2.092,
+      "step": 11042500
+    },
+    {
+      "epoch": 31.96,
+      "learning_rate": 3.4023295665061135e-05,
+      "loss": 2.1142,
+      "step": 11043000
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 3.402257201741386e-05,
+      "loss": 2.0706,
+      "step": 11043500
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 3.402184836976658e-05,
+      "loss": 2.1074,
+      "step": 11044000
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 3.402112472211931e-05,
+      "loss": 2.0848,
+      "step": 11044500
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 3.402040107447203e-05,
+      "loss": 2.0974,
+      "step": 11045000
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 3.401967742682475e-05,
+      "loss": 2.0988,
+      "step": 11045500
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 3.4018953779177475e-05,
+      "loss": 2.0857,
+      "step": 11046000
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 3.40182301315302e-05,
+      "loss": 2.1264,
+      "step": 11046500
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 3.401750937847351e-05,
+      "loss": 2.0891,
+      "step": 11047000
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 3.401678717812153e-05,
+      "loss": 2.1024,
+      "step": 11047500
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 3.401606353047425e-05,
+      "loss": 2.0933,
+      "step": 11048000
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 3.401533988282697e-05,
+      "loss": 2.1329,
+      "step": 11048500
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 3.4014616235179696e-05,
+      "loss": 2.0897,
+      "step": 11049000
+    },
+    {
+      "epoch": 31.98,
+      "learning_rate": 3.401389258753242e-05,
+      "loss": 2.1008,
+      "step": 11049500
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 3.401317038718044e-05,
+      "loss": 2.1168,
+      "step": 11050000
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 3.401244673953316e-05,
+      "loss": 2.0799,
+      "step": 11050500
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 3.4011723091885885e-05,
+      "loss": 2.1242,
+      "step": 11051000
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 3.401099944423861e-05,
+      "loss": 2.1073,
+      "step": 11051500
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 3.4010275796591336e-05,
+      "loss": 2.108,
+      "step": 11052000
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 3.400955214894406e-05,
+      "loss": 2.0969,
+      "step": 11052500
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 3.400882850129678e-05,
+      "loss": 2.1011,
+      "step": 11053000
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 3.40081048536495e-05,
+      "loss": 2.1123,
+      "step": 11053500
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 3.4007381206002225e-05,
+      "loss": 2.1004,
+      "step": 11054000
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 3.400665755835495e-05,
+      "loss": 2.1224,
+      "step": 11054500
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 3.400593391070767e-05,
+      "loss": 2.0872,
+      "step": 11055000
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.6675262658987756,
+      "eval_accuracy_mlm": 0.6324042771749461,
+      "eval_accuracy_nsp": 0.8559600903787015,
+      "eval_loss": 2.1762564182281494,
+      "eval_runtime": 331.7013,
+      "eval_samples_per_second": 1315.599,
+      "eval_steps_per_second": 54.817,
+      "step": 11055104
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 3.400521171035569e-05,
+      "loss": 2.0846,
+      "step": 11055500
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 3.4004488062708414e-05,
+      "loss": 2.06,
+      "step": 11056000
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 3.400376586235643e-05,
+      "loss": 2.0646,
+      "step": 11056500
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 3.400304221470915e-05,
+      "loss": 2.0927,
+      "step": 11057000
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 3.4002318567061874e-05,
+      "loss": 2.0445,
+      "step": 11057500
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 3.4001594919414596e-05,
+      "loss": 2.0489,
+      "step": 11058000
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 3.400087271906262e-05,
+      "loss": 2.0669,
+      "step": 11058500
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 3.400014907141534e-05,
+      "loss": 2.0876,
+      "step": 11059000
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 3.399942542376807e-05,
+      "loss": 2.099,
+      "step": 11059500
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 3.399870177612079e-05,
+      "loss": 2.0681,
+      "step": 11060000
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 3.3997978128473514e-05,
+      "loss": 2.076,
+      "step": 11060500
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 3.3997254480826236e-05,
+      "loss": 2.0616,
+      "step": 11061000
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 3.399653083317896e-05,
+      "loss": 2.0765,
+      "step": 11061500
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 3.399580718553168e-05,
+      "loss": 2.0781,
+      "step": 11062000
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 3.39950835378844e-05,
+      "loss": 2.0885,
+      "step": 11062500
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 3.3994359890237125e-05,
+      "loss": 2.0804,
+      "step": 11063000
+    },
+    {
+      "epoch": 32.02,
+      "learning_rate": 3.399363624258985e-05,
+      "loss": 2.0707,
+      "step": 11063500
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 3.399291259494257e-05,
+      "loss": 2.0815,
+      "step": 11064000
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 3.399218894729529e-05,
+      "loss": 2.0873,
+      "step": 11064500
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 3.3991465299648014e-05,
+      "loss": 2.0757,
+      "step": 11065000
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 3.399074165200074e-05,
+      "loss": 2.0726,
+      "step": 11065500
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 3.399001945164876e-05,
+      "loss": 2.0832,
+      "step": 11066000
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 3.398929580400149e-05,
+      "loss": 2.087,
+      "step": 11066500
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 3.398857215635421e-05,
+      "loss": 2.0782,
+      "step": 11067000
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 3.398784850870693e-05,
+      "loss": 2.0785,
+      "step": 11067500
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 3.3987124861059655e-05,
+      "loss": 2.0597,
+      "step": 11068000
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 3.398640121341238e-05,
+      "loss": 2.0627,
+      "step": 11068500
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 3.398567901306039e-05,
+      "loss": 2.0933,
+      "step": 11069000
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 3.398495536541312e-05,
+      "loss": 2.0995,
+      "step": 11069500
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 3.3984231717765844e-05,
+      "loss": 2.0712,
+      "step": 11070000
+    },
+    {
+      "epoch": 32.04,
+      "learning_rate": 3.398350951741386e-05,
+      "loss": 2.0735,
+      "step": 11070500
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 3.398278586976658e-05,
+      "loss": 2.0865,
+      "step": 11071000
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 3.3982062222119304e-05,
+      "loss": 2.0783,
+      "step": 11071500
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 3.3981338574472026e-05,
+      "loss": 2.0867,
+      "step": 11072000
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 3.398061492682475e-05,
+      "loss": 2.0735,
+      "step": 11072500
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 3.397989272647277e-05,
+      "loss": 2.107,
+      "step": 11073000
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 3.397916907882549e-05,
+      "loss": 2.0847,
+      "step": 11073500
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 3.397844543117822e-05,
+      "loss": 2.0888,
+      "step": 11074000
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 3.3977721783530944e-05,
+      "loss": 2.0778,
+      "step": 11074500
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 3.3976998135883666e-05,
+      "loss": 2.0654,
+      "step": 11075000
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 3.397627448823639e-05,
+      "loss": 2.0775,
+      "step": 11075500
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 3.397555084058911e-05,
+      "loss": 2.0729,
+      "step": 11076000
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 3.397482719294183e-05,
+      "loss": 2.0838,
+      "step": 11076500
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 3.3974103545294555e-05,
+      "loss": 2.1062,
+      "step": 11077000
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 3.397337989764728e-05,
+      "loss": 2.0761,
+      "step": 11077500
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 3.397265625e-05,
+      "loss": 2.0998,
+      "step": 11078000
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 3.397193260235272e-05,
+      "loss": 2.0944,
+      "step": 11078500
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 3.3971210402000744e-05,
+      "loss": 2.1006,
+      "step": 11079000
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 3.3970486754353466e-05,
+      "loss": 2.1022,
+      "step": 11079500
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 3.396976310670619e-05,
+      "loss": 2.088,
+      "step": 11080000
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 3.396903945905891e-05,
+      "loss": 2.0724,
+      "step": 11080500
+    },
+    {
+      "epoch": 32.07,
+      "learning_rate": 3.396831581141163e-05,
+      "loss": 2.1133,
+      "step": 11081000
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 3.3967593611059655e-05,
+      "loss": 2.0864,
+      "step": 11081500
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 3.396686996341238e-05,
+      "loss": 2.0808,
+      "step": 11082000
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 3.39661463157651e-05,
+      "loss": 2.0805,
+      "step": 11082500
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 3.396542266811782e-05,
+      "loss": 2.0874,
+      "step": 11083000
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 3.3964699020470544e-05,
+      "loss": 2.077,
+      "step": 11083500
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 3.396397537282327e-05,
+      "loss": 2.0846,
+      "step": 11084000
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 3.3963251725175995e-05,
+      "loss": 2.1005,
+      "step": 11084500
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 3.396252807752872e-05,
+      "loss": 2.0891,
+      "step": 11085000
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 3.396180442988144e-05,
+      "loss": 2.0949,
+      "step": 11085500
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 3.396108078223416e-05,
+      "loss": 2.0748,
+      "step": 11086000
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 3.396035858188218e-05,
+      "loss": 2.0744,
+      "step": 11086500
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 3.39596349342349e-05,
+      "loss": 2.0722,
+      "step": 11087000
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 3.395891273388292e-05,
+      "loss": 2.0697,
+      "step": 11087500
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 3.3958189086235645e-05,
+      "loss": 2.0847,
+      "step": 11088000
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 3.395746688588366e-05,
+      "loss": 2.0969,
+      "step": 11088500
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 3.395674323823639e-05,
+      "loss": 2.0733,
+      "step": 11089000
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 3.395601959058911e-05,
+      "loss": 2.0939,
+      "step": 11089500
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 3.3955295942941834e-05,
+      "loss": 2.1077,
+      "step": 11090000
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 3.3954572295294556e-05,
+      "loss": 2.1059,
+      "step": 11090500
+    },
+    {
+      "epoch": 32.1,
+      "learning_rate": 3.395384864764728e-05,
+      "loss": 2.0811,
+      "step": 11091000
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 3.3953125e-05,
+      "loss": 2.0783,
+      "step": 11091500
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 3.395240135235272e-05,
+      "loss": 2.085,
+      "step": 11092000
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 3.3951677704705445e-05,
+      "loss": 2.0935,
+      "step": 11092500
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 3.3950954057058174e-05,
+      "loss": 2.1206,
+      "step": 11093000
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 3.3950230409410896e-05,
+      "loss": 2.0831,
+      "step": 11093500
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 3.394950820905891e-05,
+      "loss": 2.0824,
+      "step": 11094000
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 3.3948784561411634e-05,
+      "loss": 2.0653,
+      "step": 11094500
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 3.3948060913764356e-05,
+      "loss": 2.0849,
+      "step": 11095000
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 3.394733871341237e-05,
+      "loss": 2.0787,
+      "step": 11095500
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 3.39466150657651e-05,
+      "loss": 2.0752,
+      "step": 11096000
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 3.394589141811782e-05,
+      "loss": 2.0753,
+      "step": 11096500
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 3.3945167770470545e-05,
+      "loss": 2.0889,
+      "step": 11097000
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 3.3944444122823274e-05,
+      "loss": 2.0537,
+      "step": 11097500
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 3.3943720475175996e-05,
+      "loss": 2.1074,
+      "step": 11098000
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 3.394299682752872e-05,
+      "loss": 2.0719,
+      "step": 11098500
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 3.394227317988144e-05,
+      "loss": 2.089,
+      "step": 11099000
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 3.394154953223416e-05,
+      "loss": 2.0875,
+      "step": 11099500
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 3.394082733188218e-05,
+      "loss": 2.0743,
+      "step": 11100000
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 3.39401036842349e-05,
+      "loss": 2.0679,
+      "step": 11100500
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 3.393938003658762e-05,
+      "loss": 2.0899,
+      "step": 11101000
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 3.3938656388940345e-05,
+      "loss": 2.0655,
+      "step": 11101500
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 3.3937932741293074e-05,
+      "loss": 2.0916,
+      "step": 11102000
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 3.3937209093645796e-05,
+      "loss": 2.0972,
+      "step": 11102500
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 3.393648544599852e-05,
+      "loss": 2.1027,
+      "step": 11103000
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 3.393576179835125e-05,
+      "loss": 2.0755,
+      "step": 11103500
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 3.393503959799926e-05,
+      "loss": 2.0763,
+      "step": 11104000
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 3.3934315950351986e-05,
+      "loss": 2.0968,
+      "step": 11104500
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 3.393359230270471e-05,
+      "loss": 2.092,
+      "step": 11105000
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 3.393286865505743e-05,
+      "loss": 2.1001,
+      "step": 11105500
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 3.393214645470545e-05,
+      "loss": 2.0733,
+      "step": 11106000
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 3.3931422807058175e-05,
+      "loss": 2.1094,
+      "step": 11106500
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 3.39306991594109e-05,
+      "loss": 2.083,
+      "step": 11107000
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 3.392997551176362e-05,
+      "loss": 2.0913,
+      "step": 11107500
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 3.392925186411634e-05,
+      "loss": 2.0907,
+      "step": 11108000
+    },
+    {
+      "epoch": 32.15,
+      "learning_rate": 3.392852966376436e-05,
+      "loss": 2.07,
+      "step": 11108500
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 3.392780601611708e-05,
+      "loss": 2.0948,
+      "step": 11109000
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 3.39270823684698e-05,
+      "loss": 2.0947,
+      "step": 11109500
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 3.3926358720822524e-05,
+      "loss": 2.1131,
+      "step": 11110000
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 3.392563507317525e-05,
+      "loss": 2.0729,
+      "step": 11110500
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 3.3924911425527975e-05,
+      "loss": 2.0791,
+      "step": 11111000
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 3.3924187777880704e-05,
+      "loss": 2.1045,
+      "step": 11111500
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 3.392346557752872e-05,
+      "loss": 2.1057,
+      "step": 11112000
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 3.392274771906262e-05,
+      "loss": 2.0983,
+      "step": 11112500
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 3.3922024071415344e-05,
+      "loss": 2.0947,
+      "step": 11113000
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 3.3921300423768066e-05,
+      "loss": 2.0808,
+      "step": 11113500
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 3.392057677612079e-05,
+      "loss": 2.1142,
+      "step": 11114000
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 3.391985312847351e-05,
+      "loss": 2.081,
+      "step": 11114500
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 3.391912948082623e-05,
+      "loss": 2.091,
+      "step": 11115000
+    },
+    {
+      "epoch": 32.17,
+      "learning_rate": 3.3918405833178955e-05,
+      "loss": 2.0748,
+      "step": 11115500
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 3.391768218553168e-05,
+      "loss": 2.0833,
+      "step": 11116000
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 3.39169585378844e-05,
+      "loss": 2.0951,
+      "step": 11116500
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 3.391623489023713e-05,
+      "loss": 2.0947,
+      "step": 11117000
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 3.391551124258985e-05,
+      "loss": 2.0941,
+      "step": 11117500
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 3.391478759494257e-05,
+      "loss": 2.0664,
+      "step": 11118000
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 3.3914063947295295e-05,
+      "loss": 2.084,
+      "step": 11118500
+    },
+    {
+      "epoch": 32.18,
+      "learning_rate": 3.3913340299648024e-05,
+      "loss": 2.0732,
+      "step": 11119000
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 3.3912616652000746e-05,
+      "loss": 2.0726,
+      "step": 11119500
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 3.391189300435347e-05,
+      "loss": 2.0911,
+      "step": 11120000
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 3.391116935670619e-05,
+      "loss": 2.1006,
+      "step": 11120500
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 3.391044570905891e-05,
+      "loss": 2.0632,
+      "step": 11121000
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 3.3909722061411635e-05,
+      "loss": 2.1067,
+      "step": 11121500
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 3.390899841376436e-05,
+      "loss": 2.0869,
+      "step": 11122000
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 3.390827476611708e-05,
+      "loss": 2.1037,
+      "step": 11122500
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 3.39075511184698e-05,
+      "loss": 2.0864,
+      "step": 11123000
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 3.3906827470822524e-05,
+      "loss": 2.078,
+      "step": 11123500
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 3.390610382317525e-05,
+      "loss": 2.0904,
+      "step": 11124000
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 3.3905380175527976e-05,
+      "loss": 2.0825,
+      "step": 11124500
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 3.39046565278807e-05,
+      "loss": 2.0744,
+      "step": 11125000
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 3.390393432752871e-05,
+      "loss": 2.0716,
+      "step": 11125500
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 3.390321067988144e-05,
+      "loss": 2.0882,
+      "step": 11126000
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 3.3902487032234165e-05,
+      "loss": 2.0557,
+      "step": 11126500
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 3.390176338458689e-05,
+      "loss": 2.0811,
+      "step": 11127000
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 3.390103973693961e-05,
+      "loss": 2.0752,
+      "step": 11127500
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 3.390031608929233e-05,
+      "loss": 2.069,
+      "step": 11128000
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 3.3899592441645054e-05,
+      "loss": 2.1071,
+      "step": 11128500
+    },
+    {
+      "epoch": 32.21,
+      "learning_rate": 3.3898868793997776e-05,
+      "loss": 2.0685,
+      "step": 11129000
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 3.3898145146350505e-05,
+      "loss": 2.0597,
+      "step": 11129500
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 3.389742294599852e-05,
+      "loss": 2.0599,
+      "step": 11130000
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 3.389669929835124e-05,
+      "loss": 2.0682,
+      "step": 11130500
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 3.3895975650703965e-05,
+      "loss": 2.076,
+      "step": 11131000
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 3.389525200305669e-05,
+      "loss": 2.0684,
+      "step": 11131500
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 3.389452835540941e-05,
+      "loss": 2.052,
+      "step": 11132000
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 3.3893806155057425e-05,
+      "loss": 2.0908,
+      "step": 11132500
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 3.3893082507410154e-05,
+      "loss": 2.0819,
+      "step": 11133000
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 3.3892358859762876e-05,
+      "loss": 2.1044,
+      "step": 11133500
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 3.38916366594109e-05,
+      "loss": 2.0985,
+      "step": 11134000
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 3.389091301176362e-05,
+      "loss": 2.047,
+      "step": 11134500
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 3.389018936411634e-05,
+      "loss": 2.0924,
+      "step": 11135000
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 3.3889465716469065e-05,
+      "loss": 2.0566,
+      "step": 11135500
+    },
+    {
+      "epoch": 32.23,
+      "learning_rate": 3.388874206882179e-05,
+      "loss": 2.0814,
+      "step": 11136000
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 3.388801842117451e-05,
+      "loss": 2.0559,
+      "step": 11136500
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 3.388729477352723e-05,
+      "loss": 2.0783,
+      "step": 11137000
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 3.3886572573175254e-05,
+      "loss": 2.0934,
+      "step": 11137500
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 3.388585037282327e-05,
+      "loss": 2.0781,
+      "step": 11138000
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 3.388512672517599e-05,
+      "loss": 2.0807,
+      "step": 11138500
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 3.3884403077528714e-05,
+      "loss": 2.1042,
+      "step": 11139000
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 3.3883679429881436e-05,
+      "loss": 2.0761,
+      "step": 11139500
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 3.388295578223416e-05,
+      "loss": 2.0859,
+      "step": 11140000
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 3.388223213458688e-05,
+      "loss": 2.1075,
+      "step": 11140500
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 3.388150848693961e-05,
+      "loss": 2.1082,
+      "step": 11141000
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 3.388078628658763e-05,
+      "loss": 2.0654,
+      "step": 11141500
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 3.3880062638940354e-05,
+      "loss": 2.0954,
+      "step": 11142000
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 3.387933899129308e-05,
+      "loss": 2.0707,
+      "step": 11142500
+    },
+    {
+      "epoch": 32.25,
+      "learning_rate": 3.38786153436458e-05,
+      "loss": 2.0878,
+      "step": 11143000
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 3.387789169599852e-05,
+      "loss": 2.0615,
+      "step": 11143500
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 3.387716804835124e-05,
+      "loss": 2.1077,
+      "step": 11144000
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 3.3876444400703966e-05,
+      "loss": 2.0768,
+      "step": 11144500
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 3.387572075305669e-05,
+      "loss": 2.0653,
+      "step": 11145000
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 3.387499710540941e-05,
+      "loss": 2.0721,
+      "step": 11145500
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 3.387427345776213e-05,
+      "loss": 2.0938,
+      "step": 11146000
+    },
+    {
+      "epoch": 32.26,
+      "learning_rate": 3.3873551257410155e-05,
+      "loss": 2.0851,
+      "step": 11146500
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 3.387282760976288e-05,
+      "loss": 2.0681,
+      "step": 11147000
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 3.38721039621156e-05,
+      "loss": 2.0779,
+      "step": 11147500
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 3.387138031446832e-05,
+      "loss": 2.0749,
+      "step": 11148000
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 3.387065666682105e-05,
+      "loss": 2.0871,
+      "step": 11148500
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 3.386993301917377e-05,
+      "loss": 2.0828,
+      "step": 11149000
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 3.3869209371526495e-05,
+      "loss": 2.0798,
+      "step": 11149500
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 3.386848717117451e-05,
+      "loss": 2.0705,
+      "step": 11150000
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 3.386776352352723e-05,
+      "loss": 2.095,
+      "step": 11150500
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 3.3867039875879955e-05,
+      "loss": 2.0804,
+      "step": 11151000
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 3.3866316228232684e-05,
+      "loss": 2.0664,
+      "step": 11151500
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 3.386559547517599e-05,
+      "loss": 2.0525,
+      "step": 11152000
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 3.3864871827528715e-05,
+      "loss": 2.0952,
+      "step": 11152500
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 3.386414817988144e-05,
+      "loss": 2.081,
+      "step": 11153000
+    },
+    {
+      "epoch": 32.28,
+      "learning_rate": 3.386342453223416e-05,
+      "loss": 2.0729,
+      "step": 11153500
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 3.386270088458688e-05,
+      "loss": 2.0992,
+      "step": 11154000
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 3.3861977236939604e-05,
+      "loss": 2.0812,
+      "step": 11154500
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 3.386125648388292e-05,
+      "loss": 2.0853,
+      "step": 11155000
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 3.386053283623564e-05,
+      "loss": 2.0781,
+      "step": 11155500
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 3.3859809188588364e-05,
+      "loss": 2.0671,
+      "step": 11156000
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 3.385908554094109e-05,
+      "loss": 2.0945,
+      "step": 11156500
+    },
+    {
+      "epoch": 32.29,
+      "learning_rate": 3.3858361893293815e-05,
+      "loss": 2.0844,
+      "step": 11157000
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 3.385763969294183e-05,
+      "loss": 2.0966,
+      "step": 11157500
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 3.385691604529456e-05,
+      "loss": 2.0792,
+      "step": 11158000
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 3.385619239764728e-05,
+      "loss": 2.0907,
+      "step": 11158500
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 3.3855468750000004e-05,
+      "loss": 2.0776,
+      "step": 11159000
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 3.3854745102352726e-05,
+      "loss": 2.0798,
+      "step": 11159500
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 3.385402145470545e-05,
+      "loss": 2.0812,
+      "step": 11160000
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 3.385329780705817e-05,
+      "loss": 2.094,
+      "step": 11160500
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 3.385257415941089e-05,
+      "loss": 2.0936,
+      "step": 11161000
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 3.3851850511763615e-05,
+      "loss": 2.09,
+      "step": 11161500
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 3.385112686411634e-05,
+      "loss": 2.0672,
+      "step": 11162000
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 3.385040466376436e-05,
+      "loss": 2.0852,
+      "step": 11162500
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 3.384968101611708e-05,
+      "loss": 2.0829,
+      "step": 11163000
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 3.38489588157651e-05,
+      "loss": 2.0884,
+      "step": 11163500
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 3.384823516811783e-05,
+      "loss": 2.0782,
+      "step": 11164000
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 3.384751152047055e-05,
+      "loss": 2.0956,
+      "step": 11164500
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 3.384678787282327e-05,
+      "loss": 2.084,
+      "step": 11165000
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 3.3846064225175993e-05,
+      "loss": 2.08,
+      "step": 11165500
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 3.384534347211931e-05,
+      "loss": 2.0759,
+      "step": 11166000
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 3.384461982447203e-05,
+      "loss": 2.1057,
+      "step": 11166500
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 3.384389762412005e-05,
+      "loss": 2.1112,
+      "step": 11167000
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 3.384317397647277e-05,
+      "loss": 2.0893,
+      "step": 11167500
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 3.384245032882549e-05,
+      "loss": 2.1149,
+      "step": 11168000
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 3.3841726681178214e-05,
+      "loss": 2.0912,
+      "step": 11168500
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 3.3841003033530936e-05,
+      "loss": 2.0752,
+      "step": 11169000
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 3.384027938588366e-05,
+      "loss": 2.0858,
+      "step": 11169500
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 3.383955573823638e-05,
+      "loss": 2.089,
+      "step": 11170000
+    },
+    {
+      "epoch": 32.33,
+      "learning_rate": 3.383883209058911e-05,
+      "loss": 2.0805,
+      "step": 11170500
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 3.383810844294183e-05,
+      "loss": 2.0929,
+      "step": 11171000
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 3.383738479529456e-05,
+      "loss": 2.0887,
+      "step": 11171500
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 3.383666114764728e-05,
+      "loss": 2.1026,
+      "step": 11172000
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 3.3835937500000005e-05,
+      "loss": 2.0894,
+      "step": 11172500
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 3.383521385235273e-05,
+      "loss": 2.0812,
+      "step": 11173000
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 3.383449165200074e-05,
+      "loss": 2.0684,
+      "step": 11173500
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 3.3833768004353465e-05,
+      "loss": 2.0899,
+      "step": 11174000
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 3.383304435670619e-05,
+      "loss": 2.064,
+      "step": 11174500
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 3.383232070905891e-05,
+      "loss": 2.079,
+      "step": 11175000
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 3.383159706141163e-05,
+      "loss": 2.0886,
+      "step": 11175500
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 3.383087341376436e-05,
+      "loss": 2.1082,
+      "step": 11176000
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 3.383014976611708e-05,
+      "loss": 2.0656,
+      "step": 11176500
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 3.3829426118469805e-05,
+      "loss": 2.0977,
+      "step": 11177000
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 3.382870247082253e-05,
+      "loss": 2.0813,
+      "step": 11177500
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 3.382797882317525e-05,
+      "loss": 2.1031,
+      "step": 11178000
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 3.3827256622823265e-05,
+      "loss": 2.0898,
+      "step": 11178500
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 3.3826532975175994e-05,
+      "loss": 2.0997,
+      "step": 11179000
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 3.3825809327528717e-05,
+      "loss": 2.0862,
+      "step": 11179500
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 3.382508567988144e-05,
+      "loss": 2.1061,
+      "step": 11180000
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 3.382436203223416e-05,
+      "loss": 2.0902,
+      "step": 11180500
+    },
+    {
+      "epoch": 32.36,
+      "learning_rate": 3.382363838458688e-05,
+      "loss": 2.0885,
+      "step": 11181000
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 3.3822916184234906e-05,
+      "loss": 2.0977,
+      "step": 11181500
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 3.382219253658763e-05,
+      "loss": 2.0862,
+      "step": 11182000
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 3.382146888894035e-05,
+      "loss": 2.0839,
+      "step": 11182500
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 3.382074524129307e-05,
+      "loss": 2.0732,
+      "step": 11183000
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 3.3820021593645794e-05,
+      "loss": 2.079,
+      "step": 11183500
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 3.381929794599852e-05,
+      "loss": 2.1104,
+      "step": 11184000
+    },
+    {
+      "epoch": 32.37,
+      "learning_rate": 3.381857429835124e-05,
+      "loss": 2.0937,
+      "step": 11184500
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 3.381785065070396e-05,
+      "loss": 2.0813,
+      "step": 11185000
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 3.3817127003056683e-05,
+      "loss": 2.0905,
+      "step": 11185500
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 3.381640335540941e-05,
+      "loss": 2.1026,
+      "step": 11186000
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 3.3815679707762135e-05,
+      "loss": 2.0749,
+      "step": 11186500
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 3.381495750741016e-05,
+      "loss": 2.1167,
+      "step": 11187000
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 3.381423385976288e-05,
+      "loss": 2.0801,
+      "step": 11187500
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 3.38135102121156e-05,
+      "loss": 2.0857,
+      "step": 11188000
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 3.3812786564468324e-05,
+      "loss": 2.1024,
+      "step": 11188500
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 3.3812062916821046e-05,
+      "loss": 2.0947,
+      "step": 11189000
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 3.381133926917377e-05,
+      "loss": 2.0816,
+      "step": 11189500
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 3.381061562152649e-05,
+      "loss": 2.0785,
+      "step": 11190000
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 3.380989197387921e-05,
+      "loss": 2.0753,
+      "step": 11190500
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 3.3809168326231935e-05,
+      "loss": 2.0976,
+      "step": 11191000
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 3.3808444678584664e-05,
+      "loss": 2.0756,
+      "step": 11191500
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 3.380772247823268e-05,
+      "loss": 2.0869,
+      "step": 11192000
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 3.38069988305854e-05,
+      "loss": 2.0933,
+      "step": 11192500
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 3.3806275182938124e-05,
+      "loss": 2.09,
+      "step": 11193000
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 3.380555442988144e-05,
+      "loss": 2.0881,
+      "step": 11193500
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 3.380483078223416e-05,
+      "loss": 2.1068,
+      "step": 11194000
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 3.380410713458689e-05,
+      "loss": 2.0821,
+      "step": 11194500
+    },
+    {
+      "epoch": 32.4,
+      "learning_rate": 3.380338348693961e-05,
+      "loss": 2.1017,
+      "step": 11195000
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 3.3802659839292335e-05,
+      "loss": 2.0759,
+      "step": 11195500
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 3.380193619164506e-05,
+      "loss": 2.072,
+      "step": 11196000
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 3.380121254399778e-05,
+      "loss": 2.1036,
+      "step": 11196500
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 3.38004888963505e-05,
+      "loss": 2.1077,
+      "step": 11197000
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 3.3799765248703224e-05,
+      "loss": 2.0965,
+      "step": 11197500
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 3.3799041601055946e-05,
+      "loss": 2.0795,
+      "step": 11198000
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 3.379831795340867e-05,
+      "loss": 2.0776,
+      "step": 11198500
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 3.379759430576139e-05,
+      "loss": 2.0629,
+      "step": 11199000
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 3.379687065811411e-05,
+      "loss": 2.0885,
+      "step": 11199500
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 3.379614990505743e-05,
+      "loss": 2.0822,
+      "step": 11200000
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 3.379542625741015e-05,
+      "loss": 2.0746,
+      "step": 11200500
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 3.379470260976288e-05,
+      "loss": 2.0891,
+      "step": 11201000
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 3.37939789621156e-05,
+      "loss": 2.0929,
+      "step": 11201500
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 3.3793255314468324e-05,
+      "loss": 2.0688,
+      "step": 11202000
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 3.379253166682105e-05,
+      "loss": 2.0744,
+      "step": 11202500
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 3.379180801917377e-05,
+      "loss": 2.1042,
+      "step": 11203000
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 3.379108437152649e-05,
+      "loss": 2.0733,
+      "step": 11203500
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 3.3790362171174514e-05,
+      "loss": 2.0711,
+      "step": 11204000
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 3.378963997082253e-05,
+      "loss": 2.0879,
+      "step": 11204500
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 3.378891632317525e-05,
+      "loss": 2.0799,
+      "step": 11205000
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 3.3788192675527974e-05,
+      "loss": 2.0805,
+      "step": 11205500
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 3.3787469027880696e-05,
+      "loss": 2.0987,
+      "step": 11206000
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 3.378674682752871e-05,
+      "loss": 2.1066,
+      "step": 11206500
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 3.378602317988144e-05,
+      "loss": 2.0684,
+      "step": 11207000
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 3.3785300979529456e-05,
+      "loss": 2.0823,
+      "step": 11207500
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 3.378457733188218e-05,
+      "loss": 2.1032,
+      "step": 11208000
+    },
+    {
+      "epoch": 32.44,
+      "learning_rate": 3.37838536842349e-05,
+      "loss": 2.0986,
+      "step": 11208500
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 3.378313003658763e-05,
+      "loss": 2.0898,
+      "step": 11209000
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 3.378240638894035e-05,
+      "loss": 2.0918,
+      "step": 11209500
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 3.3781682741293074e-05,
+      "loss": 2.0964,
+      "step": 11210000
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 3.3780959093645796e-05,
+      "loss": 2.0627,
+      "step": 11210500
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 3.378023544599852e-05,
+      "loss": 2.1028,
+      "step": 11211000
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 3.377951179835124e-05,
+      "loss": 2.0859,
+      "step": 11211500
+    },
+    {
+      "epoch": 32.45,
+      "learning_rate": 3.377878815070396e-05,
+      "loss": 2.0824,
+      "step": 11212000
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 3.377806450305669e-05,
+      "loss": 2.0965,
+      "step": 11212500
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 3.3777340855409414e-05,
+      "loss": 2.0746,
+      "step": 11213000
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 3.3776617207762136e-05,
+      "loss": 2.0992,
+      "step": 11213500
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 3.377589500741015e-05,
+      "loss": 2.1114,
+      "step": 11214000
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 3.3775171359762874e-05,
+      "loss": 2.1001,
+      "step": 11214500
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 3.377444915941089e-05,
+      "loss": 2.0893,
+      "step": 11215000
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 3.377372551176362e-05,
+      "loss": 2.0945,
+      "step": 11215500
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 3.377300186411634e-05,
+      "loss": 2.0785,
+      "step": 11216000
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 3.377227821646907e-05,
+      "loss": 2.0726,
+      "step": 11216500
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 3.377155456882179e-05,
+      "loss": 2.0993,
+      "step": 11217000
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 3.377083236846981e-05,
+      "loss": 2.1135,
+      "step": 11217500
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 3.377010872082253e-05,
+      "loss": 2.0802,
+      "step": 11218000
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 3.376938507317525e-05,
+      "loss": 2.1042,
+      "step": 11218500
+    },
+    {
+      "epoch": 32.47,
+      "learning_rate": 3.3768661425527974e-05,
+      "loss": 2.0902,
+      "step": 11219000
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 3.3767937777880697e-05,
+      "loss": 2.0698,
+      "step": 11219500
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 3.376721557752872e-05,
+      "loss": 2.0865,
+      "step": 11220000
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 3.376649192988144e-05,
+      "loss": 2.0949,
+      "step": 11220500
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 3.376576828223416e-05,
+      "loss": 2.094,
+      "step": 11221000
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 3.3765044634586886e-05,
+      "loss": 2.1094,
+      "step": 11221500
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 3.37643224342349e-05,
+      "loss": 2.0853,
+      "step": 11222000
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 3.3763598786587623e-05,
+      "loss": 2.1028,
+      "step": 11222500
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 3.3762875138940346e-05,
+      "loss": 2.0865,
+      "step": 11223000
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 3.3762151491293075e-05,
+      "loss": 2.0944,
+      "step": 11223500
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 3.37614278436458e-05,
+      "loss": 2.0765,
+      "step": 11224000
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 3.376070419599852e-05,
+      "loss": 2.0894,
+      "step": 11224500
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 3.375998054835124e-05,
+      "loss": 2.0821,
+      "step": 11225000
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 3.375925690070397e-05,
+      "loss": 2.093,
+      "step": 11225500
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 3.375853325305669e-05,
+      "loss": 2.084,
+      "step": 11226000
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 3.375781105270471e-05,
+      "loss": 2.0619,
+      "step": 11226500
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 3.375708740505743e-05,
+      "loss": 2.101,
+      "step": 11227000
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 3.375636375741015e-05,
+      "loss": 2.1095,
+      "step": 11227500
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 3.375564155705817e-05,
+      "loss": 2.0583,
+      "step": 11228000
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 3.375491790941089e-05,
+      "loss": 2.0993,
+      "step": 11228500
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 3.375419426176362e-05,
+      "loss": 2.1048,
+      "step": 11229000
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 3.375347061411634e-05,
+      "loss": 2.1003,
+      "step": 11229500
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 3.3752746966469064e-05,
+      "loss": 2.0843,
+      "step": 11230000
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 3.3752023318821786e-05,
+      "loss": 2.0868,
+      "step": 11230500
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 3.375130111846981e-05,
+      "loss": 2.0789,
+      "step": 11231000
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 3.375057747082253e-05,
+      "loss": 2.084,
+      "step": 11231500
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 3.374985382317525e-05,
+      "loss": 2.0965,
+      "step": 11232000
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 3.3749130175527975e-05,
+      "loss": 2.1112,
+      "step": 11232500
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 3.37484065278807e-05,
+      "loss": 2.0792,
+      "step": 11233000
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 3.374768288023342e-05,
+      "loss": 2.0668,
+      "step": 11233500
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 3.374695923258614e-05,
+      "loss": 2.083,
+      "step": 11234000
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 3.374623558493887e-05,
+      "loss": 2.0988,
+      "step": 11234500
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 3.374551193729159e-05,
+      "loss": 2.1102,
+      "step": 11235000
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 3.3744788289644315e-05,
+      "loss": 2.09,
+      "step": 11235500
+    },
+    {
+      "epoch": 32.52,
+      "learning_rate": 3.374406464199704e-05,
+      "loss": 2.1049,
+      "step": 11236000
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 3.374334244164505e-05,
+      "loss": 2.097,
+      "step": 11236500
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 3.374262024129307e-05,
+      "loss": 2.0907,
+      "step": 11237000
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 3.374189659364579e-05,
+      "loss": 2.102,
+      "step": 11237500
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 3.374117294599852e-05,
+      "loss": 2.0797,
+      "step": 11238000
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 3.374044929835124e-05,
+      "loss": 2.093,
+      "step": 11238500
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 3.373972565070397e-05,
+      "loss": 2.0878,
+      "step": 11239000
+    },
+    {
+      "epoch": 32.53,
+      "learning_rate": 3.373900200305669e-05,
+      "loss": 2.092,
+      "step": 11239500
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 3.373827980270471e-05,
+      "loss": 2.0813,
+      "step": 11240000
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 3.373755615505743e-05,
+      "loss": 2.0877,
+      "step": 11240500
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 3.373683395470545e-05,
+      "loss": 2.106,
+      "step": 11241000
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 3.373611030705817e-05,
+      "loss": 2.0805,
+      "step": 11241500
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 3.37353866594109e-05,
+      "loss": 2.0982,
+      "step": 11242000
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 3.373466301176362e-05,
+      "loss": 2.088,
+      "step": 11242500
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 3.3733940811411636e-05,
+      "loss": 2.1118,
+      "step": 11243000
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 3.373321716376436e-05,
+      "loss": 2.0854,
+      "step": 11243500
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 3.373249351611708e-05,
+      "loss": 2.12,
+      "step": 11244000
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 3.37317698684698e-05,
+      "loss": 2.102,
+      "step": 11244500
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 3.3731046220822525e-05,
+      "loss": 2.085,
+      "step": 11245000
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 3.373032257317525e-05,
+      "loss": 2.0741,
+      "step": 11245500
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 3.3729598925527976e-05,
+      "loss": 2.1125,
+      "step": 11246000
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 3.37288752778807e-05,
+      "loss": 2.0879,
+      "step": 11246500
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 3.372815307752872e-05,
+      "loss": 2.0923,
+      "step": 11247000
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 3.372742942988144e-05,
+      "loss": 2.0812,
+      "step": 11247500
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 3.3726705782234165e-05,
+      "loss": 2.103,
+      "step": 11248000
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 3.372598213458689e-05,
+      "loss": 2.0896,
+      "step": 11248500
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 3.372525848693961e-05,
+      "loss": 2.112,
+      "step": 11249000
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 3.372453483929233e-05,
+      "loss": 2.1155,
+      "step": 11249500
+    },
+    {
+      "epoch": 32.56,
+      "learning_rate": 3.3723811191645054e-05,
+      "loss": 2.1146,
+      "step": 11250000
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 3.3723087543997776e-05,
+      "loss": 2.1001,
+      "step": 11250500
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 3.37223638963505e-05,
+      "loss": 2.1008,
+      "step": 11251000
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 3.372164024870322e-05,
+      "loss": 2.1154,
+      "step": 11251500
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 3.372091804835124e-05,
+      "loss": 2.073,
+      "step": 11252000
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 3.3720194400703965e-05,
+      "loss": 2.073,
+      "step": 11252500
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 3.371947075305669e-05,
+      "loss": 2.0811,
+      "step": 11253000
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 3.3718747105409416e-05,
+      "loss": 2.1081,
+      "step": 11253500
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 3.371802345776214e-05,
+      "loss": 2.1001,
+      "step": 11254000
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 3.371729981011486e-05,
+      "loss": 2.0837,
+      "step": 11254500
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 3.371657616246758e-05,
+      "loss": 2.0763,
+      "step": 11255000
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 3.37158539621156e-05,
+      "loss": 2.0844,
+      "step": 11255500
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 3.371513031446832e-05,
+      "loss": 2.1067,
+      "step": 11256000
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 3.371440666682105e-05,
+      "loss": 2.0847,
+      "step": 11256500
+    },
+    {
+      "epoch": 32.58,
+      "learning_rate": 3.371368301917377e-05,
+      "loss": 2.1134,
+      "step": 11257000
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 3.371296081882179e-05,
+      "loss": 2.0919,
+      "step": 11257500
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 3.371223717117451e-05,
+      "loss": 2.0988,
+      "step": 11258000
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 3.371151352352723e-05,
+      "loss": 2.0723,
+      "step": 11258500
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 3.3710789875879954e-05,
+      "loss": 2.1005,
+      "step": 11259000
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 3.371006622823268e-05,
+      "loss": 2.0693,
+      "step": 11259500
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 3.37093425805854e-05,
+      "loss": 2.0951,
+      "step": 11260000
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 3.370862038023342e-05,
+      "loss": 2.0944,
+      "step": 11260500
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 3.3707898179881444e-05,
+      "loss": 2.115,
+      "step": 11261000
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 3.3707174532234166e-05,
+      "loss": 2.071,
+      "step": 11261500
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 3.370645088458689e-05,
+      "loss": 2.0731,
+      "step": 11262000
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 3.370572723693961e-05,
+      "loss": 2.0894,
+      "step": 11262500
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 3.370500358929233e-05,
+      "loss": 2.1161,
+      "step": 11263000
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 3.3704279941645055e-05,
+      "loss": 2.0792,
+      "step": 11263500
+    },
+    {
+      "epoch": 32.6,
+      "learning_rate": 3.370355629399778e-05,
+      "loss": 2.1103,
+      "step": 11264000
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 3.37028326463505e-05,
+      "loss": 2.0601,
+      "step": 11264500
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 3.370210899870322e-05,
+      "loss": 2.0947,
+      "step": 11265000
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 3.370138535105595e-05,
+      "loss": 2.0628,
+      "step": 11265500
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 3.370066170340867e-05,
+      "loss": 2.0947,
+      "step": 11266000
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 3.3699938055761395e-05,
+      "loss": 2.0868,
+      "step": 11266500
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 3.3699217302704704e-05,
+      "loss": 2.0957,
+      "step": 11267000
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 3.3698493655057426e-05,
+      "loss": 2.1232,
+      "step": 11267500
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 3.369777145470545e-05,
+      "loss": 2.0901,
+      "step": 11268000
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 3.369704780705817e-05,
+      "loss": 2.1072,
+      "step": 11268500
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 3.36963241594109e-05,
+      "loss": 2.084,
+      "step": 11269000
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 3.369560051176362e-05,
+      "loss": 2.0901,
+      "step": 11269500
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 3.3694876864116344e-05,
+      "loss": 2.1101,
+      "step": 11270000
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 3.369415466376436e-05,
+      "loss": 2.0799,
+      "step": 11270500
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 3.369343101611708e-05,
+      "loss": 2.1083,
+      "step": 11271000
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 3.3692707368469804e-05,
+      "loss": 2.0771,
+      "step": 11271500
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 3.3691985168117826e-05,
+      "loss": 2.1,
+      "step": 11272000
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 3.369126152047055e-05,
+      "loss": 2.0937,
+      "step": 11272500
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 3.369053787282327e-05,
+      "loss": 2.0851,
+      "step": 11273000
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 3.368981422517599e-05,
+      "loss": 2.0988,
+      "step": 11273500
+    },
+    {
+      "epoch": 32.63,
+      "learning_rate": 3.368909202482401e-05,
+      "loss": 2.1007,
+      "step": 11274000
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 3.368836837717673e-05,
+      "loss": 2.1077,
+      "step": 11274500
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 3.368764472952945e-05,
+      "loss": 2.0934,
+      "step": 11275000
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 3.3686921081882175e-05,
+      "loss": 2.0773,
+      "step": 11275500
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 3.36861974342349e-05,
+      "loss": 2.0915,
+      "step": 11276000
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 3.3685473786587627e-05,
+      "loss": 2.108,
+      "step": 11276500
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 3.368475013894035e-05,
+      "loss": 2.0966,
+      "step": 11277000
+    },
+    {
+      "epoch": 32.64,
+      "learning_rate": 3.368402649129308e-05,
+      "loss": 2.0984,
+      "step": 11277500
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 3.36833028436458e-05,
+      "loss": 2.0917,
+      "step": 11278000
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 3.368257919599852e-05,
+      "loss": 2.0909,
+      "step": 11278500
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 3.3681855548351245e-05,
+      "loss": 2.1118,
+      "step": 11279000
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 3.368113190070397e-05,
+      "loss": 2.1053,
+      "step": 11279500
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 3.368040825305669e-05,
+      "loss": 2.1065,
+      "step": 11280000
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 3.367968460540941e-05,
+      "loss": 2.1134,
+      "step": 11280500
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 3.3678960957762133e-05,
+      "loss": 2.0757,
+      "step": 11281000
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 3.367823875741015e-05,
+      "loss": 2.0862,
+      "step": 11281500
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 3.367751510976288e-05,
+      "loss": 2.113,
+      "step": 11282000
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 3.36767914621156e-05,
+      "loss": 2.1142,
+      "step": 11282500
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 3.367606781446832e-05,
+      "loss": 2.105,
+      "step": 11283000
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 3.367534416682105e-05,
+      "loss": 2.1001,
+      "step": 11283500
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 3.3674620519173774e-05,
+      "loss": 2.084,
+      "step": 11284000
+    },
+    {
+      "epoch": 32.66,
+      "learning_rate": 3.3673896871526496e-05,
+      "loss": 2.0848,
+      "step": 11284500
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 3.367317467117451e-05,
+      "loss": 2.0942,
+      "step": 11285000
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 3.3672451023527234e-05,
+      "loss": 2.0953,
+      "step": 11285500
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 3.3671727375879956e-05,
+      "loss": 2.0793,
+      "step": 11286000
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 3.367100372823268e-05,
+      "loss": 2.0688,
+      "step": 11286500
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 3.36702800805854e-05,
+      "loss": 2.0824,
+      "step": 11287000
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 3.366955643293813e-05,
+      "loss": 2.0989,
+      "step": 11287500
+    },
+    {
+      "epoch": 32.67,
+      "learning_rate": 3.366883278529085e-05,
+      "loss": 2.085,
+      "step": 11288000
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 3.3668109137643574e-05,
+      "loss": 2.1048,
+      "step": 11288500
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 3.3667385489996296e-05,
+      "loss": 2.0837,
+      "step": 11289000
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 3.366666328964431e-05,
+      "loss": 2.1116,
+      "step": 11289500
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 3.3665939641997034e-05,
+      "loss": 2.0819,
+      "step": 11290000
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 3.3665215994349756e-05,
+      "loss": 2.0911,
+      "step": 11290500
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 3.3664492346702485e-05,
+      "loss": 2.0735,
+      "step": 11291000
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 3.366376869905521e-05,
+      "loss": 2.1221,
+      "step": 11291500
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 3.366304649870323e-05,
+      "loss": 2.0917,
+      "step": 11292000
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 3.3662324298351245e-05,
+      "loss": 2.0987,
+      "step": 11292500
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 3.366160065070397e-05,
+      "loss": 2.0973,
+      "step": 11293000
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 3.366087700305669e-05,
+      "loss": 2.1181,
+      "step": 11293500
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 3.366015335540941e-05,
+      "loss": 2.1046,
+      "step": 11294000
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 3.365943115505743e-05,
+      "loss": 2.1022,
+      "step": 11294500
+    },
+    {
+      "epoch": 32.69,
+      "learning_rate": 3.3658707507410157e-05,
+      "loss": 2.0805,
+      "step": 11295000
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 3.365798385976288e-05,
+      "loss": 2.1143,
+      "step": 11295500
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 3.36572602121156e-05,
+      "loss": 2.1168,
+      "step": 11296000
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 3.365653656446832e-05,
+      "loss": 2.0981,
+      "step": 11296500
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 3.365581436411634e-05,
+      "loss": 2.1093,
+      "step": 11297000
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 3.3655092163764354e-05,
+      "loss": 2.0904,
+      "step": 11297500
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 3.365436851611708e-05,
+      "loss": 2.1018,
+      "step": 11298000
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 3.3653644868469806e-05,
+      "loss": 2.1027,
+      "step": 11298500
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 3.365292122082253e-05,
+      "loss": 2.0853,
+      "step": 11299000
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 3.365219757317526e-05,
+      "loss": 2.1083,
+      "step": 11299500
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 3.365147392552798e-05,
+      "loss": 2.0685,
+      "step": 11300000
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 3.36507502778807e-05,
+      "loss": 2.1188,
+      "step": 11300500
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 3.3650026630233424e-05,
+      "loss": 2.084,
+      "step": 11301000
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 3.3649302982586146e-05,
+      "loss": 2.0781,
+      "step": 11301500
+    },
+    {
+      "epoch": 32.71,
+      "learning_rate": 3.364857933493887e-05,
+      "loss": 2.1031,
+      "step": 11302000
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 3.364785568729159e-05,
+      "loss": 2.1133,
+      "step": 11302500
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 3.364713203964431e-05,
+      "loss": 2.0922,
+      "step": 11303000
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 3.3646408391997035e-05,
+      "loss": 2.0739,
+      "step": 11303500
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 3.364568474434976e-05,
+      "loss": 2.1121,
+      "step": 11304000
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 3.364496109670248e-05,
+      "loss": 2.0926,
+      "step": 11304500
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 3.36442374490552e-05,
+      "loss": 2.1013,
+      "step": 11305000
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 3.364351380140793e-05,
+      "loss": 2.1293,
+      "step": 11305500
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 3.3642793048351246e-05,
+      "loss": 2.12,
+      "step": 11306000
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 3.364207084799926e-05,
+      "loss": 2.081,
+      "step": 11306500
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 3.3641347200351984e-05,
+      "loss": 2.0862,
+      "step": 11307000
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 3.3640623552704706e-05,
+      "loss": 2.1169,
+      "step": 11307500
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 3.363990135235273e-05,
+      "loss": 2.0968,
+      "step": 11308000
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 3.363917770470545e-05,
+      "loss": 2.1026,
+      "step": 11308500
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 3.363845405705817e-05,
+      "loss": 2.0812,
+      "step": 11309000
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 3.3637730409410895e-05,
+      "loss": 2.1379,
+      "step": 11309500
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 3.363700676176362e-05,
+      "loss": 2.0791,
+      "step": 11310000
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 3.363628311411634e-05,
+      "loss": 2.0931,
+      "step": 11310500
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 3.363555946646906e-05,
+      "loss": 2.098,
+      "step": 11311000
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 3.363483726611708e-05,
+      "loss": 2.084,
+      "step": 11311500
+    },
+    {
+      "epoch": 32.74,
+      "learning_rate": 3.3634113618469806e-05,
+      "loss": 2.0872,
+      "step": 11312000
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 3.363338997082253e-05,
+      "loss": 2.0726,
+      "step": 11312500
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 3.363266632317525e-05,
+      "loss": 2.1144,
+      "step": 11313000
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 3.363194267552798e-05,
+      "loss": 2.0936,
+      "step": 11313500
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 3.3631220475175995e-05,
+      "loss": 2.089,
+      "step": 11314000
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 3.363049682752872e-05,
+      "loss": 2.0971,
+      "step": 11314500
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 3.362977317988144e-05,
+      "loss": 2.0992,
+      "step": 11315000
+    },
+    {
+      "epoch": 32.75,
+      "learning_rate": 3.3629052426824756e-05,
+      "loss": 2.1014,
+      "step": 11315500
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 3.362832877917748e-05,
+      "loss": 2.0929,
+      "step": 11316000
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 3.36276051315302e-05,
+      "loss": 2.1029,
+      "step": 11316500
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 3.362688148388292e-05,
+      "loss": 2.0849,
+      "step": 11317000
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 3.3626157836235645e-05,
+      "loss": 2.102,
+      "step": 11317500
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 3.362543418858837e-05,
+      "loss": 2.0979,
+      "step": 11318000
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 3.362471054094109e-05,
+      "loss": 2.089,
+      "step": 11318500
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 3.362398689329381e-05,
+      "loss": 2.0959,
+      "step": 11319000
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 3.3623263245646533e-05,
+      "loss": 2.0978,
+      "step": 11319500
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 3.3622539597999256e-05,
+      "loss": 2.0868,
+      "step": 11320000
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 3.3621815950351985e-05,
+      "loss": 2.1081,
+      "step": 11320500
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 3.362109230270471e-05,
+      "loss": 2.0985,
+      "step": 11321000
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 3.3620368655057436e-05,
+      "loss": 2.095,
+      "step": 11321500
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 3.361964500741016e-05,
+      "loss": 2.1272,
+      "step": 11322000
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 3.361892135976288e-05,
+      "loss": 2.0827,
+      "step": 11322500
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 3.36181977121156e-05,
+      "loss": 2.079,
+      "step": 11323000
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 3.361747551176362e-05,
+      "loss": 2.1054,
+      "step": 11323500
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 3.361675186411634e-05,
+      "loss": 2.0962,
+      "step": 11324000
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 3.361602821646906e-05,
+      "loss": 2.1041,
+      "step": 11324500
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 3.3615304568821785e-05,
+      "loss": 2.0852,
+      "step": 11325000
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 3.361458092117451e-05,
+      "loss": 2.106,
+      "step": 11325500
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 3.3613857273527236e-05,
+      "loss": 2.0935,
+      "step": 11326000
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 3.361313362587996e-05,
+      "loss": 2.076,
+      "step": 11326500
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 3.361240997823268e-05,
+      "loss": 2.0947,
+      "step": 11327000
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 3.36116863305854e-05,
+      "loss": 2.0925,
+      "step": 11327500
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 3.3610962682938125e-05,
+      "loss": 2.0888,
+      "step": 11328000
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 3.3610239035290854e-05,
+      "loss": 2.0735,
+      "step": 11328500
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 3.360951683493887e-05,
+      "loss": 2.0968,
+      "step": 11329000
+    },
+    {
+      "epoch": 32.79,
+      "learning_rate": 3.360879318729159e-05,
+      "loss": 2.0888,
+      "step": 11329500
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 3.3608069539644314e-05,
+      "loss": 2.0904,
+      "step": 11330000
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 3.3607345891997036e-05,
+      "loss": 2.1117,
+      "step": 11330500
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 3.360662369164506e-05,
+      "loss": 2.0867,
+      "step": 11331000
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 3.360590004399778e-05,
+      "loss": 2.1024,
+      "step": 11331500
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 3.36051763963505e-05,
+      "loss": 2.0573,
+      "step": 11332000
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 3.3604452748703225e-05,
+      "loss": 2.0718,
+      "step": 11332500
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 3.360373054835124e-05,
+      "loss": 2.0827,
+      "step": 11333000
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 3.360300690070396e-05,
+      "loss": 2.062,
+      "step": 11333500
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 3.3602283253056685e-05,
+      "loss": 2.1088,
+      "step": 11334000
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 3.360155960540941e-05,
+      "loss": 2.0983,
+      "step": 11334500
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 3.360083595776214e-05,
+      "loss": 2.1146,
+      "step": 11335000
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 3.360011231011486e-05,
+      "loss": 2.0939,
+      "step": 11335500
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 3.359938866246759e-05,
+      "loss": 2.0852,
+      "step": 11336000
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 3.359866501482031e-05,
+      "loss": 2.1146,
+      "step": 11336500
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 3.359794136717303e-05,
+      "loss": 2.0949,
+      "step": 11337000
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 3.3597217719525755e-05,
+      "loss": 2.0887,
+      "step": 11337500
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 3.359649407187848e-05,
+      "loss": 2.0961,
+      "step": 11338000
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 3.359577187152649e-05,
+      "loss": 2.1032,
+      "step": 11338500
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 3.3595048223879215e-05,
+      "loss": 2.1102,
+      "step": 11339000
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 3.359432457623194e-05,
+      "loss": 2.1022,
+      "step": 11339500
+    },
+    {
+      "epoch": 32.82,
+      "learning_rate": 3.359360092858466e-05,
+      "loss": 2.1093,
+      "step": 11340000
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 3.359287728093739e-05,
+      "loss": 2.0867,
+      "step": 11340500
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 3.359215363329011e-05,
+      "loss": 2.0863,
+      "step": 11341000
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 3.3591431432938126e-05,
+      "loss": 2.1054,
+      "step": 11341500
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 3.359070778529085e-05,
+      "loss": 2.0876,
+      "step": 11342000
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 3.358998413764357e-05,
+      "loss": 2.0917,
+      "step": 11342500
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 3.35892604899963e-05,
+      "loss": 2.0925,
+      "step": 11343000
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 3.358853684234902e-05,
+      "loss": 2.109,
+      "step": 11343500
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 3.3587813194701744e-05,
+      "loss": 2.1015,
+      "step": 11344000
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 3.3587089547054466e-05,
+      "loss": 2.1005,
+      "step": 11344500
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 3.358636589940719e-05,
+      "loss": 2.0668,
+      "step": 11345000
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 3.358564369905521e-05,
+      "loss": 2.1128,
+      "step": 11345500
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 3.358492005140793e-05,
+      "loss": 2.1104,
+      "step": 11346000
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 3.3584196403760655e-05,
+      "loss": 2.0704,
+      "step": 11346500
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 3.358347275611338e-05,
+      "loss": 2.1089,
+      "step": 11347000
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 3.35827491084661e-05,
+      "loss": 2.1091,
+      "step": 11347500
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 3.358202546081882e-05,
+      "loss": 2.0904,
+      "step": 11348000
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 3.3581301813171544e-05,
+      "loss": 2.1002,
+      "step": 11348500
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 3.3580578165524266e-05,
+      "loss": 2.0919,
+      "step": 11349000
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 3.357985451787699e-05,
+      "loss": 2.1037,
+      "step": 11349500
+    },
+    {
+      "epoch": 32.85,
+      "learning_rate": 3.357913087022971e-05,
+      "loss": 2.0985,
+      "step": 11350000
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 3.357840866987774e-05,
+      "loss": 2.097,
+      "step": 11350500
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 3.3577686469525755e-05,
+      "loss": 2.1152,
+      "step": 11351000
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 3.357696282187848e-05,
+      "loss": 2.1039,
+      "step": 11351500
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 3.35762391742312e-05,
+      "loss": 2.0926,
+      "step": 11352000
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 3.357551552658392e-05,
+      "loss": 2.0684,
+      "step": 11352500
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 3.3574791878936644e-05,
+      "loss": 2.0872,
+      "step": 11353000
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 3.3574068231289367e-05,
+      "loss": 2.0898,
+      "step": 11353500
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 3.357334458364209e-05,
+      "loss": 2.0997,
+      "step": 11354000
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 3.357262093599481e-05,
+      "loss": 2.1043,
+      "step": 11354500
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 3.357189728834754e-05,
+      "loss": 2.1159,
+      "step": 11355000
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 3.357117364070026e-05,
+      "loss": 2.0984,
+      "step": 11355500
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 3.3570449993052984e-05,
+      "loss": 2.1179,
+      "step": 11356000
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 3.356972634540571e-05,
+      "loss": 2.1214,
+      "step": 11356500
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 3.356900414505372e-05,
+      "loss": 2.0721,
+      "step": 11357000
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 3.3568280497406445e-05,
+      "loss": 2.1199,
+      "step": 11357500
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 3.3567556849759174e-05,
+      "loss": 2.0843,
+      "step": 11358000
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 3.3566833202111896e-05,
+      "loss": 2.0991,
+      "step": 11358500
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 3.356610955446462e-05,
+      "loss": 2.1026,
+      "step": 11359000
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 3.356538590681734e-05,
+      "loss": 2.0788,
+      "step": 11359500
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 3.356466370646536e-05,
+      "loss": 2.1061,
+      "step": 11360000
+    },
+    {
+      "epoch": 32.88,
+      "learning_rate": 3.3563940058818085e-05,
+      "loss": 2.0739,
+      "step": 11360500
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 3.356321641117081e-05,
+      "loss": 2.0938,
+      "step": 11361000
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 3.356249276352353e-05,
+      "loss": 2.1041,
+      "step": 11361500
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 3.356177201046684e-05,
+      "loss": 2.096,
+      "step": 11362000
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 3.356104836281957e-05,
+      "loss": 2.0736,
+      "step": 11362500
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 3.356032471517229e-05,
+      "loss": 2.0807,
+      "step": 11363000
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 3.355960106752501e-05,
+      "loss": 2.087,
+      "step": 11363500
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 3.355887886717303e-05,
+      "loss": 2.1138,
+      "step": 11364000
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 3.355815521952575e-05,
+      "loss": 2.1079,
+      "step": 11364500
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 3.355743157187847e-05,
+      "loss": 2.083,
+      "step": 11365000
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 3.35567079242312e-05,
+      "loss": 2.1071,
+      "step": 11365500
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 3.3555985723879216e-05,
+      "loss": 2.1076,
+      "step": 11366000
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 3.355526207623194e-05,
+      "loss": 2.084,
+      "step": 11366500
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 3.355453842858467e-05,
+      "loss": 2.0993,
+      "step": 11367000
+    },
+    {
+      "epoch": 32.9,
+      "learning_rate": 3.355381478093739e-05,
+      "loss": 2.1086,
+      "step": 11367500
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 3.355309113329011e-05,
+      "loss": 2.0906,
+      "step": 11368000
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 3.3552367485642834e-05,
+      "loss": 2.1024,
+      "step": 11368500
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 3.3551643837995556e-05,
+      "loss": 2.0882,
+      "step": 11369000
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 3.355092019034828e-05,
+      "loss": 2.0854,
+      "step": 11369500
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 3.3550197989996294e-05,
+      "loss": 2.0926,
+      "step": 11370000
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 3.3549474342349016e-05,
+      "loss": 2.1039,
+      "step": 11370500
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 3.354875069470174e-05,
+      "loss": 2.1137,
+      "step": 11371000
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 3.354802849434976e-05,
+      "loss": 2.0968,
+      "step": 11371500
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 3.354730484670248e-05,
+      "loss": 2.0875,
+      "step": 11372000
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 3.3546581199055205e-05,
+      "loss": 2.0789,
+      "step": 11372500
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 3.354585899870322e-05,
+      "loss": 2.1142,
+      "step": 11373000
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 3.354513535105595e-05,
+      "loss": 2.0667,
+      "step": 11373500
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 3.354441170340867e-05,
+      "loss": 2.1118,
+      "step": 11374000
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 3.3543688055761394e-05,
+      "loss": 2.0949,
+      "step": 11374500
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 3.354296440811412e-05,
+      "loss": 2.1049,
+      "step": 11375000
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 3.354224076046684e-05,
+      "loss": 2.1126,
+      "step": 11375500
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 3.354151711281957e-05,
+      "loss": 2.1137,
+      "step": 11376000
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 3.354079346517229e-05,
+      "loss": 2.0933,
+      "step": 11376500
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 3.354006981752501e-05,
+      "loss": 2.0885,
+      "step": 11377000
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 3.3539346169877735e-05,
+      "loss": 2.0898,
+      "step": 11377500
+    },
+    {
+      "epoch": 32.93,
+      "learning_rate": 3.353862252223046e-05,
+      "loss": 2.1123,
+      "step": 11378000
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 3.353789887458318e-05,
+      "loss": 2.0843,
+      "step": 11378500
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 3.35371752269359e-05,
+      "loss": 2.0734,
+      "step": 11379000
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 3.3536451579288624e-05,
+      "loss": 2.1219,
+      "step": 11379500
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 3.3535727931641346e-05,
+      "loss": 2.1021,
+      "step": 11380000
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 3.3535004283994075e-05,
+      "loss": 2.1185,
+      "step": 11380500
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 3.353428208364209e-05,
+      "loss": 2.0792,
+      "step": 11381000
+    },
+    {
+      "epoch": 32.94,
+      "learning_rate": 3.353355988329011e-05,
+      "loss": 2.0908,
+      "step": 11381500
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 3.3532836235642835e-05,
+      "loss": 2.0868,
+      "step": 11382000
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 3.353211403529085e-05,
+      "loss": 2.0985,
+      "step": 11382500
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 3.353139038764357e-05,
+      "loss": 2.0991,
+      "step": 11383000
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 3.3530666739996295e-05,
+      "loss": 2.1102,
+      "step": 11383500
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 3.352994309234902e-05,
+      "loss": 2.0819,
+      "step": 11384000
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 3.352921944470174e-05,
+      "loss": 2.1002,
+      "step": 11384500
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 3.352849579705447e-05,
+      "loss": 2.1269,
+      "step": 11385000
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 3.3527773596702484e-05,
+      "loss": 2.0805,
+      "step": 11385500
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 3.3527049949055206e-05,
+      "loss": 2.0919,
+      "step": 11386000
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 3.352632630140793e-05,
+      "loss": 2.0928,
+      "step": 11386500
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 3.352560265376065e-05,
+      "loss": 2.0823,
+      "step": 11387000
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 3.352487900611337e-05,
+      "loss": 2.0988,
+      "step": 11387500
+    },
+    {
+      "epoch": 32.96,
+      "learning_rate": 3.35241553584661e-05,
+      "loss": 2.0692,
+      "step": 11388000
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 3.3523431710818824e-05,
+      "loss": 2.0975,
+      "step": 11388500
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 3.3522709510466847e-05,
+      "loss": 2.0655,
+      "step": 11389000
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 3.352198731011486e-05,
+      "loss": 2.0752,
+      "step": 11389500
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 3.3521263662467584e-05,
+      "loss": 2.1022,
+      "step": 11390000
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 3.3520540014820307e-05,
+      "loss": 2.0926,
+      "step": 11390500
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 3.351981636717303e-05,
+      "loss": 2.0916,
+      "step": 11391000
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 3.3519094166821044e-05,
+      "loss": 2.0954,
+      "step": 11391500
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 3.3518370519173767e-05,
+      "loss": 2.0929,
+      "step": 11392000
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 3.3517646871526496e-05,
+      "loss": 2.1118,
+      "step": 11392500
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 3.351692322387922e-05,
+      "loss": 2.0894,
+      "step": 11393000
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 3.351619957623194e-05,
+      "loss": 2.0921,
+      "step": 11393500
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 3.351547592858466e-05,
+      "loss": 2.0946,
+      "step": 11394000
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 3.351475372823268e-05,
+      "loss": 2.1136,
+      "step": 11394500
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 3.351403152788069e-05,
+      "loss": 2.1199,
+      "step": 11395000
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 3.3513307880233416e-05,
+      "loss": 2.0861,
+      "step": 11395500
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 3.3512584232586145e-05,
+      "loss": 2.1246,
+      "step": 11396000
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 3.351186203223417e-05,
+      "loss": 2.0875,
+      "step": 11396500
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 3.351113838458689e-05,
+      "loss": 2.1086,
+      "step": 11397000
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 3.351041473693961e-05,
+      "loss": 2.0919,
+      "step": 11397500
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 3.3509691089292334e-05,
+      "loss": 2.0804,
+      "step": 11398000
+    },
+    {
+      "epoch": 32.99,
+      "learning_rate": 3.3508967441645056e-05,
+      "loss": 2.1205,
+      "step": 11398500
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 3.350824379399778e-05,
+      "loss": 2.0986,
+      "step": 11399000
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 3.35075201463505e-05,
+      "loss": 2.093,
+      "step": 11399500
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 3.350679649870322e-05,
+      "loss": 2.0906,
+      "step": 11400000
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 3.3506072851055945e-05,
+      "loss": 2.1081,
+      "step": 11400500
+    },
+    {
+      "epoch": 33.0,
+      "eval_accuracy": 0.6677652966372922,
+      "eval_accuracy_mlm": 0.6326864891368785,
+      "eval_accuracy_nsp": 0.8559715481248252,
+      "eval_loss": 2.179738759994507,
+      "eval_runtime": 331.543,
+      "eval_samples_per_second": 1316.227,
+      "eval_steps_per_second": 54.844,
+      "step": 11400576
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 3.350534920340867e-05,
+      "loss": 2.0683,
+      "step": 11401000
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 3.3504625555761396e-05,
+      "loss": 2.0833,
+      "step": 11401500
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 3.350390190811412e-05,
+      "loss": 2.0711,
+      "step": 11402000
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 3.350317826046684e-05,
+      "loss": 2.1027,
+      "step": 11402500
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 3.350245461281957e-05,
+      "loss": 2.0772,
+      "step": 11403000
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 3.350173096517229e-05,
+      "loss": 2.0842,
+      "step": 11403500
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 3.350100876482031e-05,
+      "loss": 2.0676,
+      "step": 11404000
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 3.350028511717303e-05,
+      "loss": 2.0813,
+      "step": 11404500
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 3.349956146952575e-05,
+      "loss": 2.0817,
+      "step": 11405000
+    },
+    {
+      "epoch": 33.01,
+      "learning_rate": 3.3498837821878474e-05,
+      "loss": 2.0761,
+      "step": 11405500
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 3.3498114174231196e-05,
+      "loss": 2.0555,
+      "step": 11406000
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 3.349739052658392e-05,
+      "loss": 2.0629,
+      "step": 11406500
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 3.349666687893665e-05,
+      "loss": 2.113,
+      "step": 11407000
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 3.349594323128937e-05,
+      "loss": 2.0806,
+      "step": 11407500
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 3.3495221030937385e-05,
+      "loss": 2.0568,
+      "step": 11408000
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 3.349449738329011e-05,
+      "loss": 2.0908,
+      "step": 11408500
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 3.349377373564283e-05,
+      "loss": 2.0872,
+      "step": 11409000
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 3.349305008799555e-05,
+      "loss": 2.0867,
+      "step": 11409500
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 3.349232788764357e-05,
+      "loss": 2.0412,
+      "step": 11410000
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 3.3491604239996297e-05,
+      "loss": 2.0786,
+      "step": 11410500
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 3.349088059234902e-05,
+      "loss": 2.0591,
+      "step": 11411000
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 3.349015694470175e-05,
+      "loss": 2.1039,
+      "step": 11411500
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 3.348943329705447e-05,
+      "loss": 2.0768,
+      "step": 11412000
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 3.348870964940719e-05,
+      "loss": 2.092,
+      "step": 11412500
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 3.3487986001759915e-05,
+      "loss": 2.0586,
+      "step": 11413000
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 3.348726235411264e-05,
+      "loss": 2.0882,
+      "step": 11413500
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 3.348653870646536e-05,
+      "loss": 2.0505,
+      "step": 11414000
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 3.348581505881808e-05,
+      "loss": 2.054,
+      "step": 11414500
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 3.3485091411170803e-05,
+      "loss": 2.1065,
+      "step": 11415000
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 3.348436921081882e-05,
+      "loss": 2.0608,
+      "step": 11415500
+    },
+    {
+      "epoch": 33.04,
+      "learning_rate": 3.348364556317155e-05,
+      "loss": 2.0652,
+      "step": 11416000
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 3.348292191552427e-05,
+      "loss": 2.0784,
+      "step": 11416500
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 3.348219826787699e-05,
+      "loss": 2.0764,
+      "step": 11417000
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 3.348147606752501e-05,
+      "loss": 2.1013,
+      "step": 11417500
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 3.348075241987774e-05,
+      "loss": 2.0917,
+      "step": 11418000
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 3.348003021952575e-05,
+      "loss": 2.0621,
+      "step": 11418500
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 3.3479306571878475e-05,
+      "loss": 2.1033,
+      "step": 11419000
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 3.34785829242312e-05,
+      "loss": 2.0716,
+      "step": 11419500
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 3.3477859276583926e-05,
+      "loss": 2.0767,
+      "step": 11420000
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 3.347713562893665e-05,
+      "loss": 2.0628,
+      "step": 11420500
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 3.3476413428584664e-05,
+      "loss": 2.1032,
+      "step": 11421000
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 3.3475689780937386e-05,
+      "loss": 2.0615,
+      "step": 11421500
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 3.347496613329011e-05,
+      "loss": 2.0781,
+      "step": 11422000
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 3.347424248564283e-05,
+      "loss": 2.0958,
+      "step": 11422500
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 3.347351883799555e-05,
+      "loss": 2.061,
+      "step": 11423000
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 3.3472795190348275e-05,
+      "loss": 2.0641,
+      "step": 11423500
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 3.34720729899963e-05,
+      "loss": 2.0751,
+      "step": 11424000
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 3.347135078964431e-05,
+      "loss": 2.0642,
+      "step": 11424500
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 3.3470627141997035e-05,
+      "loss": 2.0978,
+      "step": 11425000
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 3.346990349434976e-05,
+      "loss": 2.078,
+      "step": 11425500
+    },
+    {
+      "epoch": 33.07,
+      "learning_rate": 3.3469179846702486e-05,
+      "loss": 2.0577,
+      "step": 11426000
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 3.346845619905521e-05,
+      "loss": 2.0678,
+      "step": 11426500
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 3.346773255140793e-05,
+      "loss": 2.0769,
+      "step": 11427000
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 3.346700890376065e-05,
+      "loss": 2.0774,
+      "step": 11427500
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 3.3466285256113375e-05,
+      "loss": 2.0919,
+      "step": 11428000
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 3.34655616084661e-05,
+      "loss": 2.0641,
+      "step": 11428500
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 3.346483940811412e-05,
+      "loss": 2.0819,
+      "step": 11429000
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 3.346411576046684e-05,
+      "loss": 2.0918,
+      "step": 11429500
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 3.3463392112819564e-05,
+      "loss": 2.0683,
+      "step": 11430000
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 3.3462668465172287e-05,
+      "loss": 2.0706,
+      "step": 11430500
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 3.346194481752501e-05,
+      "loss": 2.0698,
+      "step": 11431000
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 3.346122116987773e-05,
+      "loss": 2.0849,
+      "step": 11431500
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 3.346049752223045e-05,
+      "loss": 2.0735,
+      "step": 11432000
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 3.3459773874583176e-05,
+      "loss": 2.1006,
+      "step": 11432500
+    },
+    {
+      "epoch": 33.09,
+      "learning_rate": 3.3459050226935905e-05,
+      "loss": 2.0963,
+      "step": 11433000
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 3.345832802658393e-05,
+      "loss": 2.072,
+      "step": 11433500
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 3.345760437893665e-05,
+      "loss": 2.0803,
+      "step": 11434000
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 3.345688073128937e-05,
+      "loss": 2.0648,
+      "step": 11434500
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 3.345615853093739e-05,
+      "loss": 2.0695,
+      "step": 11435000
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 3.34554363305854e-05,
+      "loss": 2.0677,
+      "step": 11435500
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 3.3454712682938125e-05,
+      "loss": 2.0911,
+      "step": 11436000
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 3.345398903529085e-05,
+      "loss": 2.0878,
+      "step": 11436500
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 3.3453265387643576e-05,
+      "loss": 2.0665,
+      "step": 11437000
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 3.34525417399963e-05,
+      "loss": 2.0942,
+      "step": 11437500
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 3.345181809234902e-05,
+      "loss": 2.0719,
+      "step": 11438000
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 3.345109444470174e-05,
+      "loss": 2.088,
+      "step": 11438500
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 3.3450370797054465e-05,
+      "loss": 2.0724,
+      "step": 11439000
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 3.344964714940719e-05,
+      "loss": 2.0826,
+      "step": 11439500
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 3.344892350175991e-05,
+      "loss": 2.0807,
+      "step": 11440000
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 3.344819985411264e-05,
+      "loss": 2.0644,
+      "step": 11440500
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 3.344747620646536e-05,
+      "loss": 2.0844,
+      "step": 11441000
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 3.344675255881808e-05,
+      "loss": 2.0972,
+      "step": 11441500
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 3.3446028911170805e-05,
+      "loss": 2.0783,
+      "step": 11442000
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 3.344530526352353e-05,
+      "loss": 2.0699,
+      "step": 11442500
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 3.344458161587625e-05,
+      "loss": 2.0467,
+      "step": 11443000
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 3.344385941552427e-05,
+      "loss": 2.055,
+      "step": 11443500
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 3.344313721517229e-05,
+      "loss": 2.0756,
+      "step": 11444000
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 3.344241356752501e-05,
+      "loss": 2.1077,
+      "step": 11444500
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 3.344168991987773e-05,
+      "loss": 2.0813,
+      "step": 11445000
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 3.3440966272230454e-05,
+      "loss": 2.0583,
+      "step": 11445500
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 3.3440242624583176e-05,
+      "loss": 2.0638,
+      "step": 11446000
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 3.34395189769359e-05,
+      "loss": 2.0927,
+      "step": 11446500
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 3.343879532928863e-05,
+      "loss": 2.06,
+      "step": 11447000
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 3.343807168164135e-05,
+      "loss": 2.07,
+      "step": 11447500
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 3.343734948128937e-05,
+      "loss": 2.0928,
+      "step": 11448000
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 3.3436625833642094e-05,
+      "loss": 2.0777,
+      "step": 11448500
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 3.3435902185994817e-05,
+      "loss": 2.0747,
+      "step": 11449000
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 3.343517998564283e-05,
+      "loss": 2.096,
+      "step": 11449500
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 3.3434456337995554e-05,
+      "loss": 2.0788,
+      "step": 11450000
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 3.343373269034828e-05,
+      "loss": 2.0896,
+      "step": 11450500
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 3.3433009042701006e-05,
+      "loss": 2.0989,
+      "step": 11451000
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 3.343228684234902e-05,
+      "loss": 2.0513,
+      "step": 11451500
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 3.343156464199704e-05,
+      "loss": 2.0647,
+      "step": 11452000
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 3.343084099434976e-05,
+      "loss": 2.0896,
+      "step": 11452500
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 3.343011734670248e-05,
+      "loss": 2.0622,
+      "step": 11453000
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 3.3429393699055203e-05,
+      "loss": 2.0936,
+      "step": 11453500
+    },
+    {
+      "epoch": 33.15,
+      "learning_rate": 3.3428670051407926e-05,
+      "loss": 2.123,
+      "step": 11454000
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 3.3427946403760655e-05,
+      "loss": 2.093,
+      "step": 11454500
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 3.342722275611338e-05,
+      "loss": 2.0935,
+      "step": 11455000
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 3.3426499108466106e-05,
+      "loss": 2.0646,
+      "step": 11455500
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 3.342577690811412e-05,
+      "loss": 2.0842,
+      "step": 11456000
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 3.3425053260466844e-05,
+      "loss": 2.0923,
+      "step": 11456500
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 3.3424329612819566e-05,
+      "loss": 2.0735,
+      "step": 11457000
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 3.342360596517229e-05,
+      "loss": 2.0636,
+      "step": 11457500
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 3.342288231752501e-05,
+      "loss": 2.0603,
+      "step": 11458000
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 3.342215866987773e-05,
+      "loss": 2.0746,
+      "step": 11458500
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 3.3421435022230455e-05,
+      "loss": 2.0655,
+      "step": 11459000
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 3.342071137458318e-05,
+      "loss": 2.0839,
+      "step": 11459500
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 3.34199891742312e-05,
+      "loss": 2.0877,
+      "step": 11460000
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 3.341926552658392e-05,
+      "loss": 2.0788,
+      "step": 11460500
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 3.3418541878936644e-05,
+      "loss": 2.086,
+      "step": 11461000
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 3.3417818231289366e-05,
+      "loss": 2.1039,
+      "step": 11461500
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 3.341709603093738e-05,
+      "loss": 2.0772,
+      "step": 11462000
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 3.3416372383290104e-05,
+      "loss": 2.081,
+      "step": 11462500
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 3.341564873564283e-05,
+      "loss": 2.0633,
+      "step": 11463000
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 3.3414925087995555e-05,
+      "loss": 2.0848,
+      "step": 11463500
+    },
+    {
+      "epoch": 33.18,
+      "learning_rate": 3.341420144034828e-05,
+      "loss": 2.0765,
+      "step": 11464000
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 3.3413477792701006e-05,
+      "loss": 2.079,
+      "step": 11464500
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 3.341275414505373e-05,
+      "loss": 2.0342,
+      "step": 11465000
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 3.341203049740645e-05,
+      "loss": 2.0952,
+      "step": 11465500
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 3.341130684975917e-05,
+      "loss": 2.0717,
+      "step": 11466000
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 3.3410583202111895e-05,
+      "loss": 2.0851,
+      "step": 11466500
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 3.340985955446462e-05,
+      "loss": 2.0834,
+      "step": 11467000
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 3.340913590681734e-05,
+      "loss": 2.0978,
+      "step": 11467500
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 3.340841225917006e-05,
+      "loss": 2.0766,
+      "step": 11468000
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 3.3407688611522784e-05,
+      "loss": 2.077,
+      "step": 11468500
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 3.3406964963875507e-05,
+      "loss": 2.0923,
+      "step": 11469000
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 3.340624131622823e-05,
+      "loss": 2.0699,
+      "step": 11469500
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 3.340551911587626e-05,
+      "loss": 2.0861,
+      "step": 11470000
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 3.340479546822898e-05,
+      "loss": 2.0851,
+      "step": 11470500
+    },
+    {
+      "epoch": 33.2,
+      "learning_rate": 3.34040718205817e-05,
+      "loss": 2.0693,
+      "step": 11471000
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 3.340334962022972e-05,
+      "loss": 2.0764,
+      "step": 11471500
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 3.340262597258244e-05,
+      "loss": 2.0567,
+      "step": 11472000
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 3.340190232493516e-05,
+      "loss": 2.0875,
+      "step": 11472500
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 3.3401178677287885e-05,
+      "loss": 2.0741,
+      "step": 11473000
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 3.340045502964061e-05,
+      "loss": 2.0924,
+      "step": 11473500
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 3.339973282928863e-05,
+      "loss": 2.084,
+      "step": 11474000
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 3.339900918164135e-05,
+      "loss": 2.0798,
+      "step": 11474500
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 3.3398285533994074e-05,
+      "loss": 2.0763,
+      "step": 11475000
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 3.3397561886346796e-05,
+      "loss": 2.1009,
+      "step": 11475500
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 3.339683823869952e-05,
+      "loss": 2.0899,
+      "step": 11476000
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 3.3396116038347534e-05,
+      "loss": 2.0763,
+      "step": 11476500
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 3.3395392390700256e-05,
+      "loss": 2.0953,
+      "step": 11477000
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 3.339466874305298e-05,
+      "loss": 2.0838,
+      "step": 11477500
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 3.339394509540571e-05,
+      "loss": 2.0802,
+      "step": 11478000
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 3.339322144775843e-05,
+      "loss": 2.0884,
+      "step": 11478500
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 3.339249780011116e-05,
+      "loss": 2.081,
+      "step": 11479000
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 3.3391775599759174e-05,
+      "loss": 2.077,
+      "step": 11479500
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 3.3391051952111896e-05,
+      "loss": 2.0723,
+      "step": 11480000
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 3.339032830446462e-05,
+      "loss": 2.0806,
+      "step": 11480500
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 3.338960465681734e-05,
+      "loss": 2.0627,
+      "step": 11481000
+    },
+    {
+      "epoch": 33.23,
+      "learning_rate": 3.338888100917006e-05,
+      "loss": 2.1063,
+      "step": 11481500
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 3.3388157361522785e-05,
+      "loss": 2.0832,
+      "step": 11482000
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 3.338743371387551e-05,
+      "loss": 2.0894,
+      "step": 11482500
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 3.338671006622823e-05,
+      "loss": 2.0948,
+      "step": 11483000
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 3.338598786587625e-05,
+      "loss": 2.0817,
+      "step": 11483500
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 3.3385264218228974e-05,
+      "loss": 2.1093,
+      "step": 11484000
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 3.3384540570581696e-05,
+      "loss": 2.1173,
+      "step": 11484500
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 3.338381837022971e-05,
+      "loss": 2.0922,
+      "step": 11485000
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 3.338309472258244e-05,
+      "loss": 2.0476,
+      "step": 11485500
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 3.338237107493516e-05,
+      "loss": 2.0984,
+      "step": 11486000
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 3.3381647427287885e-05,
+      "loss": 2.1033,
+      "step": 11486500
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 3.338092522693591e-05,
+      "loss": 2.0707,
+      "step": 11487000
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 3.338020157928863e-05,
+      "loss": 2.0766,
+      "step": 11487500
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 3.337947793164135e-05,
+      "loss": 2.0936,
+      "step": 11488000
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 3.3378754283994074e-05,
+      "loss": 2.0636,
+      "step": 11488500
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 3.337803208364209e-05,
+      "loss": 2.0943,
+      "step": 11489000
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 3.337730843599481e-05,
+      "loss": 2.0804,
+      "step": 11489500
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 3.3376586235642835e-05,
+      "loss": 2.0611,
+      "step": 11490000
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 3.337586258799556e-05,
+      "loss": 2.0696,
+      "step": 11490500
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 3.337514038764357e-05,
+      "loss": 2.0771,
+      "step": 11491000
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 3.3374416739996295e-05,
+      "loss": 2.0766,
+      "step": 11491500
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 3.337369309234902e-05,
+      "loss": 2.0777,
+      "step": 11492000
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 3.337296944470174e-05,
+      "loss": 2.095,
+      "step": 11492500
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 3.337224579705447e-05,
+      "loss": 2.0916,
+      "step": 11493000
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 3.337152214940719e-05,
+      "loss": 2.0849,
+      "step": 11493500
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 3.337079850175991e-05,
+      "loss": 2.0765,
+      "step": 11494000
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 3.3370074854112635e-05,
+      "loss": 2.076,
+      "step": 11494500
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 3.336935120646536e-05,
+      "loss": 2.0734,
+      "step": 11495000
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 3.3368627558818086e-05,
+      "loss": 2.0969,
+      "step": 11495500
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 3.336790391117081e-05,
+      "loss": 2.0708,
+      "step": 11496000
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 3.336718026352353e-05,
+      "loss": 2.0842,
+      "step": 11496500
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 3.336645661587625e-05,
+      "loss": 2.0752,
+      "step": 11497000
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 3.336573441552427e-05,
+      "loss": 2.0887,
+      "step": 11497500
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 3.336501076787699e-05,
+      "loss": 2.0941,
+      "step": 11498000
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 3.336428712022971e-05,
+      "loss": 2.0907,
+      "step": 11498500
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 3.3363563472582435e-05,
+      "loss": 2.0685,
+      "step": 11499000
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 3.336283982493516e-05,
+      "loss": 2.0836,
+      "step": 11499500
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 3.3362116177287886e-05,
+      "loss": 2.1096,
+      "step": 11500000
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 3.336139252964061e-05,
+      "loss": 2.0907,
+      "step": 11500500
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 3.336066888199334e-05,
+      "loss": 2.1103,
+      "step": 11501000
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 3.335994523434606e-05,
+      "loss": 2.0673,
+      "step": 11501500
+    },
+    {
+      "epoch": 33.29,
+      "learning_rate": 3.335922158669878e-05,
+      "loss": 2.0736,
+      "step": 11502000
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 3.3358497939051504e-05,
+      "loss": 2.0711,
+      "step": 11502500
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 3.335777573869952e-05,
+      "loss": 2.1109,
+      "step": 11503000
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 3.335705209105224e-05,
+      "loss": 2.0901,
+      "step": 11503500
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 3.3356328443404964e-05,
+      "loss": 2.0703,
+      "step": 11504000
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 3.3355604795757686e-05,
+      "loss": 2.0939,
+      "step": 11504500
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 3.335488114811041e-05,
+      "loss": 2.0705,
+      "step": 11505000
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 3.335415750046314e-05,
+      "loss": 2.1021,
+      "step": 11505500
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 3.335343530011115e-05,
+      "loss": 2.0933,
+      "step": 11506000
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 3.335271309975917e-05,
+      "loss": 2.0714,
+      "step": 11506500
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 3.335198945211189e-05,
+      "loss": 2.0749,
+      "step": 11507000
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 3.335126580446461e-05,
+      "loss": 2.095,
+      "step": 11507500
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 3.335054215681734e-05,
+      "loss": 2.0777,
+      "step": 11508000
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 3.3349818509170064e-05,
+      "loss": 2.0662,
+      "step": 11508500
+    },
+    {
+      "epoch": 33.31,
+      "learning_rate": 3.334909486152279e-05,
+      "loss": 2.0801,
+      "step": 11509000
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 3.334837121387551e-05,
+      "loss": 2.0928,
+      "step": 11509500
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 3.334764756622824e-05,
+      "loss": 2.0737,
+      "step": 11510000
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 3.334692391858096e-05,
+      "loss": 2.0943,
+      "step": 11510500
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 3.334620027093368e-05,
+      "loss": 2.0996,
+      "step": 11511000
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 3.3345476623286405e-05,
+      "loss": 2.0876,
+      "step": 11511500
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 3.334475297563913e-05,
+      "loss": 2.0753,
+      "step": 11512000
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 3.334402932799185e-05,
+      "loss": 2.0833,
+      "step": 11512500
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 3.334330568034457e-05,
+      "loss": 2.0791,
+      "step": 11513000
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 3.3342582032697294e-05,
+      "loss": 2.0813,
+      "step": 11513500
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 3.334185983234531e-05,
+      "loss": 2.0702,
+      "step": 11514000
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 3.334113618469804e-05,
+      "loss": 2.0834,
+      "step": 11514500
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 3.334041253705076e-05,
+      "loss": 2.063,
+      "step": 11515000
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 3.333968888940349e-05,
+      "loss": 2.0778,
+      "step": 11515500
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 3.333896524175621e-05,
+      "loss": 2.0858,
+      "step": 11516000
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 3.3338241594108934e-05,
+      "loss": 2.0836,
+      "step": 11516500
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 3.3337517946461656e-05,
+      "loss": 2.094,
+      "step": 11517000
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 3.333679574610967e-05,
+      "loss": 2.1021,
+      "step": 11517500
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 3.333607354575769e-05,
+      "loss": 2.0933,
+      "step": 11518000
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 3.3335349898110416e-05,
+      "loss": 2.1025,
+      "step": 11518500
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 3.333462769775843e-05,
+      "loss": 2.098,
+      "step": 11519000
+    },
+    {
+      "epoch": 33.34,
+      "learning_rate": 3.3333904050111154e-05,
+      "loss": 2.0757,
+      "step": 11519500
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 3.3333180402463876e-05,
+      "loss": 2.0699,
+      "step": 11520000
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 3.33324567548166e-05,
+      "loss": 2.1122,
+      "step": 11520500
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 3.333173310716932e-05,
+      "loss": 2.0884,
+      "step": 11521000
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 3.333100945952204e-05,
+      "loss": 2.0961,
+      "step": 11521500
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 3.3330285811874765e-05,
+      "loss": 2.075,
+      "step": 11522000
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 3.3329562164227494e-05,
+      "loss": 2.0751,
+      "step": 11522500
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 3.3328838516580216e-05,
+      "loss": 2.0973,
+      "step": 11523000
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 3.332811631622824e-05,
+      "loss": 2.0818,
+      "step": 11523500
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 3.3327394115876254e-05,
+      "loss": 2.074,
+      "step": 11524000
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 3.332667191552427e-05,
+      "loss": 2.0958,
+      "step": 11524500
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 3.332594826787699e-05,
+      "loss": 2.0992,
+      "step": 11525000
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 3.3325224620229714e-05,
+      "loss": 2.0839,
+      "step": 11525500
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 3.3324500972582437e-05,
+      "loss": 2.0944,
+      "step": 11526000
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 3.3323777324935166e-05,
+      "loss": 2.0777,
+      "step": 11526500
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 3.332305367728789e-05,
+      "loss": 2.1044,
+      "step": 11527000
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 3.332233002964061e-05,
+      "loss": 2.0879,
+      "step": 11527500
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 3.332160638199333e-05,
+      "loss": 2.0903,
+      "step": 11528000
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 3.3320882734346054e-05,
+      "loss": 2.095,
+      "step": 11528500
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 3.332016053399407e-05,
+      "loss": 2.0844,
+      "step": 11529000
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 3.331943688634679e-05,
+      "loss": 2.0696,
+      "step": 11529500
+    },
+    {
+      "epoch": 33.37,
+      "learning_rate": 3.331871323869952e-05,
+      "loss": 2.0752,
+      "step": 11530000
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 3.3317989591052244e-05,
+      "loss": 2.0717,
+      "step": 11530500
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 3.3317265943404966e-05,
+      "loss": 2.0931,
+      "step": 11531000
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 3.331654229575769e-05,
+      "loss": 2.079,
+      "step": 11531500
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 3.331581864811042e-05,
+      "loss": 2.083,
+      "step": 11532000
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 3.331509500046314e-05,
+      "loss": 2.0682,
+      "step": 11532500
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 3.331437135281586e-05,
+      "loss": 2.0844,
+      "step": 11533000
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 3.3313647705168584e-05,
+      "loss": 2.0643,
+      "step": 11533500
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 3.33129255048166e-05,
+      "loss": 2.0775,
+      "step": 11534000
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 3.331220185716932e-05,
+      "loss": 2.0757,
+      "step": 11534500
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 3.3311478209522044e-05,
+      "loss": 2.1054,
+      "step": 11535000
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 3.3310754561874766e-05,
+      "loss": 2.0905,
+      "step": 11535500
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 3.331003091422749e-05,
+      "loss": 2.0975,
+      "step": 11536000
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 3.330930726658022e-05,
+      "loss": 2.0661,
+      "step": 11536500
+    },
+    {
+      "epoch": 33.39,
+      "learning_rate": 3.330858361893294e-05,
+      "loss": 2.0998,
+      "step": 11537000
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 3.330785997128567e-05,
+      "loss": 2.0972,
+      "step": 11537500
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 3.330713921822898e-05,
+      "loss": 2.1132,
+      "step": 11538000
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 3.330641846517229e-05,
+      "loss": 2.098,
+      "step": 11538500
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 3.3305694817525015e-05,
+      "loss": 2.0933,
+      "step": 11539000
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 3.330497116987774e-05,
+      "loss": 2.0741,
+      "step": 11539500
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 3.330424752223046e-05,
+      "loss": 2.0794,
+      "step": 11540000
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 3.330352387458318e-05,
+      "loss": 2.0816,
+      "step": 11540500
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 3.3302800226935904e-05,
+      "loss": 2.068,
+      "step": 11541000
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 3.3302076579288626e-05,
+      "loss": 2.0877,
+      "step": 11541500
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 3.330135293164135e-05,
+      "loss": 2.0897,
+      "step": 11542000
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 3.330062928399407e-05,
+      "loss": 2.0861,
+      "step": 11542500
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 3.329990563634679e-05,
+      "loss": 2.0585,
+      "step": 11543000
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 3.3299181988699515e-05,
+      "loss": 2.0959,
+      "step": 11543500
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 3.3298458341052244e-05,
+      "loss": 2.1084,
+      "step": 11544000
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 3.329773614070026e-05,
+      "loss": 2.0686,
+      "step": 11544500
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 3.329701249305298e-05,
+      "loss": 2.0929,
+      "step": 11545000
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 3.329628884540571e-05,
+      "loss": 2.0968,
+      "step": 11545500
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 3.329556519775843e-05,
+      "loss": 2.0719,
+      "step": 11546000
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 3.3294841550111156e-05,
+      "loss": 2.068,
+      "step": 11546500
+    },
+    {
+      "epoch": 33.42,
+      "learning_rate": 3.329411790246388e-05,
+      "loss": 2.1199,
+      "step": 11547000
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 3.32933942548166e-05,
+      "loss": 2.0854,
+      "step": 11547500
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 3.329267060716932e-05,
+      "loss": 2.078,
+      "step": 11548000
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 3.3291946959522044e-05,
+      "loss": 2.0751,
+      "step": 11548500
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 3.329122331187477e-05,
+      "loss": 2.0871,
+      "step": 11549000
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 3.3290499664227496e-05,
+      "loss": 2.0945,
+      "step": 11549500
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 3.328977601658022e-05,
+      "loss": 2.1027,
+      "step": 11550000
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 3.328905236893294e-05,
+      "loss": 2.0926,
+      "step": 11550500
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 3.328832872128566e-05,
+      "loss": 2.1086,
+      "step": 11551000
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 3.328760652093368e-05,
+      "loss": 2.0981,
+      "step": 11551500
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 3.32868828732864e-05,
+      "loss": 2.0928,
+      "step": 11552000
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 3.328616067293442e-05,
+      "loss": 2.0993,
+      "step": 11552500
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 3.3285437025287145e-05,
+      "loss": 2.0687,
+      "step": 11553000
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 3.328471337763987e-05,
+      "loss": 2.0729,
+      "step": 11553500
+    },
+    {
+      "epoch": 33.44,
+      "learning_rate": 3.328399117728789e-05,
+      "loss": 2.0749,
+      "step": 11554000
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 3.328326752964061e-05,
+      "loss": 2.1239,
+      "step": 11554500
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 3.3282543881993334e-05,
+      "loss": 2.0652,
+      "step": 11555000
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 3.3281820234346056e-05,
+      "loss": 2.0885,
+      "step": 11555500
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 3.328109658669878e-05,
+      "loss": 2.0694,
+      "step": 11556000
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 3.32803729390515e-05,
+      "loss": 2.1112,
+      "step": 11556500
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 3.327964929140422e-05,
+      "loss": 2.0858,
+      "step": 11557000
+    },
+    {
+      "epoch": 33.45,
+      "learning_rate": 3.3278925643756945e-05,
+      "loss": 2.0997,
+      "step": 11557500
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 3.327820199610967e-05,
+      "loss": 2.0658,
+      "step": 11558000
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 3.3277478348462396e-05,
+      "loss": 2.0909,
+      "step": 11558500
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 3.327675614811041e-05,
+      "loss": 2.0986,
+      "step": 11559000
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 3.3276032500463134e-05,
+      "loss": 2.0684,
+      "step": 11559500
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 3.3275310300111156e-05,
+      "loss": 2.089,
+      "step": 11560000
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 3.327458665246388e-05,
+      "loss": 2.0624,
+      "step": 11560500
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 3.32738630048166e-05,
+      "loss": 2.0653,
+      "step": 11561000
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 3.327313935716932e-05,
+      "loss": 2.0674,
+      "step": 11561500
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 3.3272415709522045e-05,
+      "loss": 2.0904,
+      "step": 11562000
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 3.327169206187477e-05,
+      "loss": 2.0747,
+      "step": 11562500
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 3.3270968414227497e-05,
+      "loss": 2.1005,
+      "step": 11563000
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 3.327024476658022e-05,
+      "loss": 2.1081,
+      "step": 11563500
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 3.326952111893294e-05,
+      "loss": 2.0867,
+      "step": 11564000
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 3.3268798918580957e-05,
+      "loss": 2.1002,
+      "step": 11564500
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 3.326807527093368e-05,
+      "loss": 2.0904,
+      "step": 11565000
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 3.32673516232864e-05,
+      "loss": 2.0602,
+      "step": 11565500
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 3.326662797563912e-05,
+      "loss": 2.0829,
+      "step": 11566000
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 3.3265904327991845e-05,
+      "loss": 2.1092,
+      "step": 11566500
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 3.326518068034457e-05,
+      "loss": 2.0531,
+      "step": 11567000
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 3.32644584799926e-05,
+      "loss": 2.0869,
+      "step": 11567500
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 3.326373483234532e-05,
+      "loss": 2.0923,
+      "step": 11568000
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 3.3263012631993335e-05,
+      "loss": 2.0881,
+      "step": 11568500
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 3.326228898434606e-05,
+      "loss": 2.0997,
+      "step": 11569000
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 3.326156533669878e-05,
+      "loss": 2.0881,
+      "step": 11569500
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 3.32608416890515e-05,
+      "loss": 2.1037,
+      "step": 11570000
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 3.3260118041404224e-05,
+      "loss": 2.0884,
+      "step": 11570500
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 3.3259394393756946e-05,
+      "loss": 2.0825,
+      "step": 11571000
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 3.325867219340497e-05,
+      "loss": 2.0976,
+      "step": 11571500
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 3.325794854575769e-05,
+      "loss": 2.0872,
+      "step": 11572000
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 3.325722489811041e-05,
+      "loss": 2.0889,
+      "step": 11572500
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 3.3256501250463135e-05,
+      "loss": 2.092,
+      "step": 11573000
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 3.325577760281586e-05,
+      "loss": 2.0732,
+      "step": 11573500
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 3.325505540246387e-05,
+      "loss": 2.1021,
+      "step": 11574000
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 3.3254331754816595e-05,
+      "loss": 2.0922,
+      "step": 11574500
+    },
+    {
+      "epoch": 33.5,
+      "learning_rate": 3.3253608107169324e-05,
+      "loss": 2.1004,
+      "step": 11575000
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 3.3252884459522046e-05,
+      "loss": 2.1022,
+      "step": 11575500
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 3.325216225917007e-05,
+      "loss": 2.0685,
+      "step": 11576000
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 3.325143861152279e-05,
+      "loss": 2.0806,
+      "step": 11576500
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 3.325071496387551e-05,
+      "loss": 2.0952,
+      "step": 11577000
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 3.3249991316228235e-05,
+      "loss": 2.0781,
+      "step": 11577500
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 3.324926766858096e-05,
+      "loss": 2.0673,
+      "step": 11578000
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 3.324854402093368e-05,
+      "loss": 2.1041,
+      "step": 11578500
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 3.32478203732864e-05,
+      "loss": 2.0995,
+      "step": 11579000
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 3.3247096725639124e-05,
+      "loss": 2.098,
+      "step": 11579500
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 3.3246373077991846e-05,
+      "loss": 2.0748,
+      "step": 11580000
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 3.3245649430344575e-05,
+      "loss": 2.0739,
+      "step": 11580500
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 3.324492722999259e-05,
+      "loss": 2.0889,
+      "step": 11581000
+    },
+    {
+      "epoch": 33.52,
+      "learning_rate": 3.324420358234531e-05,
+      "loss": 2.1168,
+      "step": 11581500
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 3.3243479934698035e-05,
+      "loss": 2.0992,
+      "step": 11582000
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 3.3242756287050764e-05,
+      "loss": 2.1035,
+      "step": 11582500
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 3.3242032639403487e-05,
+      "loss": 2.0847,
+      "step": 11583000
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 3.32413104390515e-05,
+      "loss": 2.0796,
+      "step": 11583500
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 3.3240588238699524e-05,
+      "loss": 2.0792,
+      "step": 11584000
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 3.323986459105225e-05,
+      "loss": 2.1101,
+      "step": 11584500
+    },
+    {
+      "epoch": 33.53,
+      "learning_rate": 3.323914094340497e-05,
+      "loss": 2.094,
+      "step": 11585000
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 3.3238418743052984e-05,
+      "loss": 2.0644,
+      "step": 11585500
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 3.323769509540571e-05,
+      "loss": 2.0967,
+      "step": 11586000
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 3.323697144775843e-05,
+      "loss": 2.1022,
+      "step": 11586500
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 3.323624780011115e-05,
+      "loss": 2.0837,
+      "step": 11587000
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 3.3235524152463873e-05,
+      "loss": 2.1079,
+      "step": 11587500
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 3.3234800504816596e-05,
+      "loss": 2.0808,
+      "step": 11588000
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 3.3234076857169325e-05,
+      "loss": 2.0879,
+      "step": 11588500
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 3.323335320952205e-05,
+      "loss": 2.084,
+      "step": 11589000
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 3.323262956187477e-05,
+      "loss": 2.0853,
+      "step": 11589500
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 3.323190736152279e-05,
+      "loss": 2.0783,
+      "step": 11590000
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 3.3231183713875514e-05,
+      "loss": 2.0802,
+      "step": 11590500
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 3.3230460066228236e-05,
+      "loss": 2.09,
+      "step": 11591000
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 3.322973641858096e-05,
+      "loss": 2.0863,
+      "step": 11591500
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 3.322901277093368e-05,
+      "loss": 2.0967,
+      "step": 11592000
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 3.32282905705817e-05,
+      "loss": 2.0886,
+      "step": 11592500
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 3.3227566922934425e-05,
+      "loss": 2.085,
+      "step": 11593000
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 3.322684327528715e-05,
+      "loss": 2.063,
+      "step": 11593500
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 3.322611962763987e-05,
+      "loss": 2.0998,
+      "step": 11594000
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 3.322539597999259e-05,
+      "loss": 2.0888,
+      "step": 11594500
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 3.322467377964061e-05,
+      "loss": 2.0835,
+      "step": 11595000
+    },
+    {
+      "epoch": 33.56,
+      "learning_rate": 3.322395013199333e-05,
+      "loss": 2.0818,
+      "step": 11595500
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 3.322322648434605e-05,
+      "loss": 2.1062,
+      "step": 11596000
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 3.3222504283994074e-05,
+      "loss": 2.0759,
+      "step": 11596500
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 3.3221780636346796e-05,
+      "loss": 2.0884,
+      "step": 11597000
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 3.3221056988699525e-05,
+      "loss": 2.1077,
+      "step": 11597500
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 3.322033334105225e-05,
+      "loss": 2.1144,
+      "step": 11598000
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 3.321960969340497e-05,
+      "loss": 2.0901,
+      "step": 11598500
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 3.321888604575769e-05,
+      "loss": 2.0746,
+      "step": 11599000
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 3.3218162398110414e-05,
+      "loss": 2.0872,
+      "step": 11599500
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 3.3217438750463136e-05,
+      "loss": 2.1001,
+      "step": 11600000
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 3.321671510281586e-05,
+      "loss": 2.1071,
+      "step": 11600500
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 3.321599145516858e-05,
+      "loss": 2.089,
+      "step": 11601000
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 3.32152678075213e-05,
+      "loss": 2.1002,
+      "step": 11601500
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 3.3214544159874025e-05,
+      "loss": 2.1014,
+      "step": 11602000
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 3.321382195952205e-05,
+      "loss": 2.0694,
+      "step": 11602500
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 3.321309831187477e-05,
+      "loss": 2.0773,
+      "step": 11603000
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 3.321237466422749e-05,
+      "loss": 2.0994,
+      "step": 11603500
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 3.321165246387551e-05,
+      "loss": 2.0688,
+      "step": 11604000
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 3.321092881622823e-05,
+      "loss": 2.0688,
+      "step": 11604500
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 3.321020516858096e-05,
+      "loss": 2.086,
+      "step": 11605000
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 3.320948152093368e-05,
+      "loss": 2.1103,
+      "step": 11605500
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 3.3208757873286403e-05,
+      "loss": 2.093,
+      "step": 11606000
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 3.3208034225639126e-05,
+      "loss": 2.0946,
+      "step": 11606500
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 3.3207310577991855e-05,
+      "loss": 2.0923,
+      "step": 11607000
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 3.320658693034458e-05,
+      "loss": 2.0866,
+      "step": 11607500
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 3.32058632826973e-05,
+      "loss": 2.1166,
+      "step": 11608000
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 3.3205141082345315e-05,
+      "loss": 2.1265,
+      "step": 11608500
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 3.320441743469804e-05,
+      "loss": 2.0866,
+      "step": 11609000
+    },
+    {
+      "epoch": 33.6,
+      "learning_rate": 3.320369378705076e-05,
+      "loss": 2.078,
+      "step": 11609500
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 3.320297013940348e-05,
+      "loss": 2.0856,
+      "step": 11610000
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 3.3202246491756204e-05,
+      "loss": 2.0681,
+      "step": 11610500
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 3.3201522844108926e-05,
+      "loss": 2.0941,
+      "step": 11611000
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 3.3200799196461655e-05,
+      "loss": 2.0878,
+      "step": 11611500
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 3.320007554881438e-05,
+      "loss": 2.0854,
+      "step": 11612000
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 3.31993533484624e-05,
+      "loss": 2.0848,
+      "step": 11612500
+    },
+    {
+      "epoch": 33.61,
+      "learning_rate": 3.319862970081512e-05,
+      "loss": 2.0728,
+      "step": 11613000
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 3.3197906053167844e-05,
+      "loss": 2.071,
+      "step": 11613500
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 3.3197182405520566e-05,
+      "loss": 2.0997,
+      "step": 11614000
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 3.319646020516858e-05,
+      "loss": 2.0955,
+      "step": 11614500
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 3.3195738004816604e-05,
+      "loss": 2.0833,
+      "step": 11615000
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 3.3195014357169326e-05,
+      "loss": 2.0796,
+      "step": 11615500
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 3.319429070952205e-05,
+      "loss": 2.0919,
+      "step": 11616000
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 3.319356706187477e-05,
+      "loss": 2.0931,
+      "step": 11616500
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 3.319284341422749e-05,
+      "loss": 2.0923,
+      "step": 11617000
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 3.3192119766580215e-05,
+      "loss": 2.0906,
+      "step": 11617500
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 3.319139611893294e-05,
+      "loss": 2.0889,
+      "step": 11618000
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 3.319067247128566e-05,
+      "loss": 2.0902,
+      "step": 11618500
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 3.318994882363838e-05,
+      "loss": 2.095,
+      "step": 11619000
+    },
+    {
+      "epoch": 33.63,
+      "learning_rate": 3.3189226623286404e-05,
+      "loss": 2.1011,
+      "step": 11619500
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 3.3188502975639126e-05,
+      "loss": 2.0994,
+      "step": 11620000
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 3.318778222258244e-05,
+      "loss": 2.0852,
+      "step": 11620500
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 3.3187058574935164e-05,
+      "loss": 2.0961,
+      "step": 11621000
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 3.318633637458318e-05,
+      "loss": 2.0978,
+      "step": 11621500
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 3.31856127269359e-05,
+      "loss": 2.0926,
+      "step": 11622000
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 3.318488907928863e-05,
+      "loss": 2.1234,
+      "step": 11622500
+    },
+    {
+      "epoch": 33.64,
+      "learning_rate": 3.318416543164135e-05,
+      "loss": 2.0914,
+      "step": 11623000
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 3.3183441783994076e-05,
+      "loss": 2.0934,
+      "step": 11623500
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 3.31827181363468e-05,
+      "loss": 2.0761,
+      "step": 11624000
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 3.318199593599481e-05,
+      "loss": 2.1006,
+      "step": 11624500
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 3.3181272288347536e-05,
+      "loss": 2.0661,
+      "step": 11625000
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 3.318055008799555e-05,
+      "loss": 2.0856,
+      "step": 11625500
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 3.317982644034828e-05,
+      "loss": 2.0716,
+      "step": 11626000
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 3.3179102792701e-05,
+      "loss": 2.0958,
+      "step": 11626500
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 3.3178379145053725e-05,
+      "loss": 2.1009,
+      "step": 11627000
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 3.317765549740645e-05,
+      "loss": 2.093,
+      "step": 11627500
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 3.3176931849759176e-05,
+      "loss": 2.0871,
+      "step": 11628000
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 3.317620964940719e-05,
+      "loss": 2.0731,
+      "step": 11628500
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 3.3175486001759914e-05,
+      "loss": 2.1009,
+      "step": 11629000
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 3.3174762354112636e-05,
+      "loss": 2.0928,
+      "step": 11629500
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 3.317403870646536e-05,
+      "loss": 2.099,
+      "step": 11630000
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 3.317331505881808e-05,
+      "loss": 2.0959,
+      "step": 11630500
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 3.31725914111708e-05,
+      "loss": 2.079,
+      "step": 11631000
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 3.317186776352353e-05,
+      "loss": 2.0774,
+      "step": 11631500
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 3.3171144115876254e-05,
+      "loss": 2.1021,
+      "step": 11632000
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 3.317042191552427e-05,
+      "loss": 2.0902,
+      "step": 11632500
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 3.316969826787699e-05,
+      "loss": 2.0684,
+      "step": 11633000
+    },
+    {
+      "epoch": 33.67,
+      "learning_rate": 3.3168974620229714e-05,
+      "loss": 2.1003,
+      "step": 11633500
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 3.3168250972582436e-05,
+      "loss": 2.092,
+      "step": 11634000
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 3.316752732493516e-05,
+      "loss": 2.0773,
+      "step": 11634500
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 3.316680367728789e-05,
+      "loss": 2.0833,
+      "step": 11635000
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 3.316608002964061e-05,
+      "loss": 2.1025,
+      "step": 11635500
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 3.316535638199333e-05,
+      "loss": 2.0918,
+      "step": 11636000
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 3.3164632734346054e-05,
+      "loss": 2.085,
+      "step": 11636500
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 3.316390908669878e-05,
+      "loss": 2.0753,
+      "step": 11637000
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 3.3163185439051505e-05,
+      "loss": 2.0601,
+      "step": 11637500
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 3.316246179140423e-05,
+      "loss": 2.0716,
+      "step": 11638000
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 3.316173814375695e-05,
+      "loss": 2.0851,
+      "step": 11638500
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 3.316101449610967e-05,
+      "loss": 2.0798,
+      "step": 11639000
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 3.316029229575769e-05,
+      "loss": 2.0774,
+      "step": 11639500
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 3.315956864811041e-05,
+      "loss": 2.0796,
+      "step": 11640000
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 3.315884500046313e-05,
+      "loss": 2.0668,
+      "step": 11640500
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 3.3158121352815854e-05,
+      "loss": 2.0986,
+      "step": 11641000
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 3.3157399152463877e-05,
+      "loss": 2.1004,
+      "step": 11641500
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 3.31566755048166e-05,
+      "loss": 2.0798,
+      "step": 11642000
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 3.315595185716933e-05,
+      "loss": 2.0764,
+      "step": 11642500
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 3.315522820952205e-05,
+      "loss": 2.1072,
+      "step": 11643000
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 3.315450456187477e-05,
+      "loss": 2.0855,
+      "step": 11643500
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 3.3153780914227495e-05,
+      "loss": 2.0839,
+      "step": 11644000
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 3.315305726658022e-05,
+      "loss": 2.0819,
+      "step": 11644500
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 3.315233361893294e-05,
+      "loss": 2.07,
+      "step": 11645000
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 3.315160997128566e-05,
+      "loss": 2.0754,
+      "step": 11645500
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 3.3150886323638383e-05,
+      "loss": 2.0952,
+      "step": 11646000
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 3.3150162675991106e-05,
+      "loss": 2.1035,
+      "step": 11646500
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 3.314944047563913e-05,
+      "loss": 2.0864,
+      "step": 11647000
+    },
+    {
+      "epoch": 33.71,
+      "learning_rate": 3.3148718275287144e-05,
+      "loss": 2.0822,
+      "step": 11647500
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 3.3147994627639866e-05,
+      "loss": 2.0955,
+      "step": 11648000
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 3.314727242728788e-05,
+      "loss": 2.0962,
+      "step": 11648500
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 3.314654877964061e-05,
+      "loss": 2.0907,
+      "step": 11649000
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 3.314582513199333e-05,
+      "loss": 2.0844,
+      "step": 11649500
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 3.314510148434606e-05,
+      "loss": 2.0706,
+      "step": 11650000
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 3.3144377836698784e-05,
+      "loss": 2.1027,
+      "step": 11650500
+    },
+    {
+      "epoch": 33.72,
+      "learning_rate": 3.3143654189051506e-05,
+      "loss": 2.0854,
+      "step": 11651000
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 3.314293054140423e-05,
+      "loss": 2.1021,
+      "step": 11651500
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 3.314220689375695e-05,
+      "loss": 2.1129,
+      "step": 11652000
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 3.3141484693404966e-05,
+      "loss": 2.1121,
+      "step": 11652500
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 3.314076104575769e-05,
+      "loss": 2.0973,
+      "step": 11653000
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 3.314003739811041e-05,
+      "loss": 2.0778,
+      "step": 11653500
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 3.313931375046313e-05,
+      "loss": 2.1001,
+      "step": 11654000
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 3.313859010281586e-05,
+      "loss": 2.1015,
+      "step": 11654500
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 3.3137866455168584e-05,
+      "loss": 2.0898,
+      "step": 11655000
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 3.3137142807521306e-05,
+      "loss": 2.0829,
+      "step": 11655500
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 3.313641915987403e-05,
+      "loss": 2.0735,
+      "step": 11656000
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 3.313569551222675e-05,
+      "loss": 2.0993,
+      "step": 11656500
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 3.313497186457948e-05,
+      "loss": 2.0967,
+      "step": 11657000
+    },
+    {
+      "epoch": 33.74,
+      "learning_rate": 3.31342482169322e-05,
+      "loss": 2.0873,
+      "step": 11657500
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 3.3133524569284924e-05,
+      "loss": 2.0707,
+      "step": 11658000
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 3.3132800921637646e-05,
+      "loss": 2.0912,
+      "step": 11658500
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 3.313207872128566e-05,
+      "loss": 2.0852,
+      "step": 11659000
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 3.3131355073638384e-05,
+      "loss": 2.1049,
+      "step": 11659500
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 3.313063142599111e-05,
+      "loss": 2.0779,
+      "step": 11660000
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 3.3129907778343836e-05,
+      "loss": 2.1034,
+      "step": 11660500
+    },
+    {
+      "epoch": 33.75,
+      "learning_rate": 3.312918413069656e-05,
+      "loss": 2.1041,
+      "step": 11661000
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 3.312846193034457e-05,
+      "loss": 2.0867,
+      "step": 11661500
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 3.3127738282697296e-05,
+      "loss": 2.0956,
+      "step": 11662000
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 3.312701608234531e-05,
+      "loss": 2.1141,
+      "step": 11662500
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 3.312629243469803e-05,
+      "loss": 2.0966,
+      "step": 11663000
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 3.312556878705076e-05,
+      "loss": 2.0547,
+      "step": 11663500
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 3.312484803399407e-05,
+      "loss": 2.0933,
+      "step": 11664000
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 3.3124124386346793e-05,
+      "loss": 2.0734,
+      "step": 11664500
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 3.312340073869952e-05,
+      "loss": 2.0764,
+      "step": 11665000
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 3.3122677091052245e-05,
+      "loss": 2.0909,
+      "step": 11665500
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 3.312195344340497e-05,
+      "loss": 2.0643,
+      "step": 11666000
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 3.312122979575769e-05,
+      "loss": 2.0907,
+      "step": 11666500
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 3.312050614811041e-05,
+      "loss": 2.0981,
+      "step": 11667000
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 3.3119782500463134e-05,
+      "loss": 2.0756,
+      "step": 11667500
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 3.3119060300111156e-05,
+      "loss": 2.0727,
+      "step": 11668000
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 3.311833665246388e-05,
+      "loss": 2.0736,
+      "step": 11668500
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 3.31176130048166e-05,
+      "loss": 2.0759,
+      "step": 11669000
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 3.311688935716932e-05,
+      "loss": 2.0822,
+      "step": 11669500
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 3.3116165709522045e-05,
+      "loss": 2.0938,
+      "step": 11670000
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 3.311544206187477e-05,
+      "loss": 2.0871,
+      "step": 11670500
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 3.311471841422749e-05,
+      "loss": 2.114,
+      "step": 11671000
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 3.311399476658021e-05,
+      "loss": 2.0847,
+      "step": 11671500
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 3.3113271118932934e-05,
+      "loss": 2.0483,
+      "step": 11672000
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 3.311254747128566e-05,
+      "loss": 2.0913,
+      "step": 11672500
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 3.3111823823638385e-05,
+      "loss": 2.0688,
+      "step": 11673000
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 3.311110162328641e-05,
+      "loss": 2.0822,
+      "step": 11673500
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 3.311037797563913e-05,
+      "loss": 2.0927,
+      "step": 11674000
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 3.310965432799185e-05,
+      "loss": 2.0776,
+      "step": 11674500
+    },
+    {
+      "epoch": 33.79,
+      "learning_rate": 3.3108930680344574e-05,
+      "loss": 2.1098,
+      "step": 11675000
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 3.310820847999259e-05,
+      "loss": 2.0925,
+      "step": 11675500
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 3.310748483234531e-05,
+      "loss": 2.0847,
+      "step": 11676000
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 3.3106761184698034e-05,
+      "loss": 2.0652,
+      "step": 11676500
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 3.310603753705076e-05,
+      "loss": 2.1004,
+      "step": 11677000
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 3.3105313889403485e-05,
+      "loss": 2.0999,
+      "step": 11677500
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 3.310459024175621e-05,
+      "loss": 2.1025,
+      "step": 11678000
+    },
+    {
+      "epoch": 33.8,
+      "learning_rate": 3.310386659410893e-05,
+      "loss": 2.0901,
+      "step": 11678500
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 3.310314294646165e-05,
+      "loss": 2.0812,
+      "step": 11679000
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 3.310241929881438e-05,
+      "loss": 2.0678,
+      "step": 11679500
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 3.31016956511671e-05,
+      "loss": 2.0775,
+      "step": 11680000
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 3.3100972003519826e-05,
+      "loss": 2.0689,
+      "step": 11680500
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 3.310024835587255e-05,
+      "loss": 2.0798,
+      "step": 11681000
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 3.309952470822527e-05,
+      "loss": 2.0749,
+      "step": 11681500
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 3.309880106057799e-05,
+      "loss": 2.0686,
+      "step": 11682000
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 3.3098078860226015e-05,
+      "loss": 2.0719,
+      "step": 11682500
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 3.309735521257874e-05,
+      "loss": 2.0871,
+      "step": 11683000
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 3.309663156493146e-05,
+      "loss": 2.0917,
+      "step": 11683500
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 3.309590791728418e-05,
+      "loss": 2.0887,
+      "step": 11684000
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 3.3095184269636904e-05,
+      "loss": 2.0879,
+      "step": 11684500
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 3.3094460621989626e-05,
+      "loss": 2.0852,
+      "step": 11685000
+    },
+    {
+      "epoch": 33.82,
+      "learning_rate": 3.309373842163764e-05,
+      "loss": 2.0985,
+      "step": 11685500
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 3.3093014773990364e-05,
+      "loss": 2.0871,
+      "step": 11686000
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 3.3092292573638386e-05,
+      "loss": 2.1035,
+      "step": 11686500
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 3.3091568925991115e-05,
+      "loss": 2.0803,
+      "step": 11687000
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 3.309084527834384e-05,
+      "loss": 2.0848,
+      "step": 11687500
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 3.309012307799185e-05,
+      "loss": 2.0928,
+      "step": 11688000
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 3.3089399430344575e-05,
+      "loss": 2.1,
+      "step": 11688500
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 3.30886757826973e-05,
+      "loss": 2.0821,
+      "step": 11689000
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 3.308795213505002e-05,
+      "loss": 2.0917,
+      "step": 11689500
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 3.308722848740274e-05,
+      "loss": 2.0952,
+      "step": 11690000
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 3.3086504839755464e-05,
+      "loss": 2.0686,
+      "step": 11690500
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 3.308578119210819e-05,
+      "loss": 2.0799,
+      "step": 11691000
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 3.3085057544460915e-05,
+      "loss": 2.0828,
+      "step": 11691500
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 3.308433389681364e-05,
+      "loss": 2.0852,
+      "step": 11692000
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 3.308361024916636e-05,
+      "loss": 2.1174,
+      "step": 11692500
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 3.308288660151908e-05,
+      "loss": 2.0695,
+      "step": 11693000
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 3.3082162953871804e-05,
+      "loss": 2.0851,
+      "step": 11693500
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 3.3081439306224526e-05,
+      "loss": 2.0796,
+      "step": 11694000
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 3.308071710587255e-05,
+      "loss": 2.1108,
+      "step": 11694500
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 3.307999345822527e-05,
+      "loss": 2.111,
+      "step": 11695000
+    },
+    {
+      "epoch": 33.85,
+      "learning_rate": 3.307926981057799e-05,
+      "loss": 2.0707,
+      "step": 11695500
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 3.3078547610226015e-05,
+      "loss": 2.1109,
+      "step": 11696000
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 3.307782396257874e-05,
+      "loss": 2.0798,
+      "step": 11696500
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 3.307710031493146e-05,
+      "loss": 2.0527,
+      "step": 11697000
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 3.307637666728418e-05,
+      "loss": 2.0901,
+      "step": 11697500
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 3.3075653019636904e-05,
+      "loss": 2.096,
+      "step": 11698000
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 3.3074929371989627e-05,
+      "loss": 2.0759,
+      "step": 11698500
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 3.307420572434235e-05,
+      "loss": 2.1082,
+      "step": 11699000
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 3.307348207669507e-05,
+      "loss": 2.074,
+      "step": 11699500
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 3.307275987634309e-05,
+      "loss": 2.0813,
+      "step": 11700000
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 3.3072036228695816e-05,
+      "loss": 2.0708,
+      "step": 11700500
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 3.307131258104854e-05,
+      "loss": 2.0742,
+      "step": 11701000
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 3.307058893340126e-05,
+      "loss": 2.056,
+      "step": 11701500
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 3.306986528575399e-05,
+      "loss": 2.099,
+      "step": 11702000
+    },
+    {
+      "epoch": 33.87,
+      "learning_rate": 3.306914163810671e-05,
+      "loss": 2.067,
+      "step": 11702500
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 3.306841943775473e-05,
+      "loss": 2.0963,
+      "step": 11703000
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 3.306769579010745e-05,
+      "loss": 2.0731,
+      "step": 11703500
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 3.306697214246017e-05,
+      "loss": 2.0767,
+      "step": 11704000
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 3.3066249942108194e-05,
+      "loss": 2.0569,
+      "step": 11704500
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 3.3065526294460916e-05,
+      "loss": 2.09,
+      "step": 11705000
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 3.306480264681364e-05,
+      "loss": 2.1249,
+      "step": 11705500
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 3.306407899916636e-05,
+      "loss": 2.0881,
+      "step": 11706000
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 3.3063356798814376e-05,
+      "loss": 2.1119,
+      "step": 11706500
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 3.30626331511671e-05,
+      "loss": 2.1004,
+      "step": 11707000
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 3.306190950351982e-05,
+      "loss": 2.0924,
+      "step": 11707500
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 3.306118585587254e-05,
+      "loss": 2.1048,
+      "step": 11708000
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 3.3060462208225265e-05,
+      "loss": 2.0896,
+      "step": 11708500
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 3.3059738560577994e-05,
+      "loss": 2.0899,
+      "step": 11709000
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 3.3059014912930716e-05,
+      "loss": 2.0724,
+      "step": 11709500
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 3.3058291265283445e-05,
+      "loss": 2.0603,
+      "step": 11710000
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 3.305756906493146e-05,
+      "loss": 2.0818,
+      "step": 11710500
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 3.305684541728418e-05,
+      "loss": 2.0761,
+      "step": 11711000
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 3.3056121769636905e-05,
+      "loss": 2.0831,
+      "step": 11711500
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 3.305539812198963e-05,
+      "loss": 2.077,
+      "step": 11712000
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 3.305467447434235e-05,
+      "loss": 2.0693,
+      "step": 11712500
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 3.3053952273990365e-05,
+      "loss": 2.0949,
+      "step": 11713000
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 3.3053228626343094e-05,
+      "loss": 2.0816,
+      "step": 11713500
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 3.305250642599111e-05,
+      "loss": 2.0977,
+      "step": 11714000
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 3.305178277834383e-05,
+      "loss": 2.0706,
+      "step": 11714500
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 3.305106057799185e-05,
+      "loss": 2.0847,
+      "step": 11715000
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 3.305033693034457e-05,
+      "loss": 2.0959,
+      "step": 11715500
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 3.304961328269729e-05,
+      "loss": 2.0836,
+      "step": 11716000
+    },
+    {
+      "epoch": 33.91,
+      "learning_rate": 3.3048891082345314e-05,
+      "loss": 2.0753,
+      "step": 11716500
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 3.3048167434698037e-05,
+      "loss": 2.0853,
+      "step": 11717000
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 3.3047443787050766e-05,
+      "loss": 2.0998,
+      "step": 11717500
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 3.304672013940349e-05,
+      "loss": 2.0762,
+      "step": 11718000
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 3.304599649175621e-05,
+      "loss": 2.1111,
+      "step": 11718500
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 3.304527284410893e-05,
+      "loss": 2.1008,
+      "step": 11719000
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 3.3044549196461654e-05,
+      "loss": 2.1127,
+      "step": 11719500
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 3.304382554881438e-05,
+      "loss": 2.1209,
+      "step": 11720000
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 3.30431019011671e-05,
+      "loss": 2.0972,
+      "step": 11720500
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 3.304237825351982e-05,
+      "loss": 2.0906,
+      "step": 11721000
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 3.304165460587254e-05,
+      "loss": 2.0788,
+      "step": 11721500
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 3.304093095822527e-05,
+      "loss": 2.0565,
+      "step": 11722000
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 3.3040207310577995e-05,
+      "loss": 2.0848,
+      "step": 11722500
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 3.303948366293072e-05,
+      "loss": 2.0761,
+      "step": 11723000
+    },
+    {
+      "epoch": 33.93,
+      "learning_rate": 3.303876001528344e-05,
+      "loss": 2.073,
+      "step": 11723500
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 3.303803636763616e-05,
+      "loss": 2.0886,
+      "step": 11724000
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 3.3037314167284184e-05,
+      "loss": 2.1171,
+      "step": 11724500
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 3.3036590519636906e-05,
+      "loss": 2.0833,
+      "step": 11725000
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 3.303586687198963e-05,
+      "loss": 2.0907,
+      "step": 11725500
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 3.303514322434235e-05,
+      "loss": 2.088,
+      "step": 11726000
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 3.303441957669507e-05,
+      "loss": 2.0699,
+      "step": 11726500
+    },
+    {
+      "epoch": 33.94,
+      "learning_rate": 3.3033695929047795e-05,
+      "loss": 2.0844,
+      "step": 11727000
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 3.3032972281400524e-05,
+      "loss": 2.0925,
+      "step": 11727500
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 3.3032248633753246e-05,
+      "loss": 2.0913,
+      "step": 11728000
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 3.303152498610597e-05,
+      "loss": 2.0799,
+      "step": 11728500
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 3.303080133845869e-05,
+      "loss": 2.0879,
+      "step": 11729000
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 3.303007769081141e-05,
+      "loss": 2.1103,
+      "step": 11729500
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 3.3029354043164135e-05,
+      "loss": 2.0887,
+      "step": 11730000
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 3.302863184281215e-05,
+      "loss": 2.0939,
+      "step": 11730500
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 3.302790819516487e-05,
+      "loss": 2.0801,
+      "step": 11731000
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 3.30271845475176e-05,
+      "loss": 2.0742,
+      "step": 11731500
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 3.3026460899870324e-05,
+      "loss": 2.0848,
+      "step": 11732000
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 3.3025738699518346e-05,
+      "loss": 2.1011,
+      "step": 11732500
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 3.302501505187107e-05,
+      "loss": 2.075,
+      "step": 11733000
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 3.3024292851519084e-05,
+      "loss": 2.0803,
+      "step": 11733500
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 3.3023569203871806e-05,
+      "loss": 2.101,
+      "step": 11734000
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 3.302284700351982e-05,
+      "loss": 2.0901,
+      "step": 11734500
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 3.3022123355872544e-05,
+      "loss": 2.0943,
+      "step": 11735000
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 3.302139970822527e-05,
+      "loss": 2.0945,
+      "step": 11735500
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 3.3020676060577995e-05,
+      "loss": 2.0736,
+      "step": 11736000
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 3.301995386022601e-05,
+      "loss": 2.0864,
+      "step": 11736500
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 3.301923021257873e-05,
+      "loss": 2.0809,
+      "step": 11737000
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 3.3018506564931455e-05,
+      "loss": 2.0885,
+      "step": 11737500
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 3.301778291728418e-05,
+      "loss": 2.1041,
+      "step": 11738000
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 3.301706071693219e-05,
+      "loss": 2.0821,
+      "step": 11738500
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 3.301633706928492e-05,
+      "loss": 2.0999,
+      "step": 11739000
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 3.3015613421637644e-05,
+      "loss": 2.1192,
+      "step": 11739500
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 3.3014889773990373e-05,
+      "loss": 2.067,
+      "step": 11740000
+    },
+    {
+      "epoch": 33.98,
+      "learning_rate": 3.3014166126343096e-05,
+      "loss": 2.0791,
+      "step": 11740500
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 3.301344247869582e-05,
+      "loss": 2.1023,
+      "step": 11741000
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 3.301271883104854e-05,
+      "loss": 2.0997,
+      "step": 11741500
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 3.301199518340126e-05,
+      "loss": 2.0968,
+      "step": 11742000
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 3.3011271535753985e-05,
+      "loss": 2.1097,
+      "step": 11742500
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 3.301054788810671e-05,
+      "loss": 2.0738,
+      "step": 11743000
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 3.300982424045943e-05,
+      "loss": 2.075,
+      "step": 11743500
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 3.300910059281215e-05,
+      "loss": 2.0988,
+      "step": 11744000
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 3.3008376945164874e-05,
+      "loss": 2.104,
+      "step": 11744500
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 3.3007654744812896e-05,
+      "loss": 2.0784,
+      "step": 11745000
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 3.300693109716562e-05,
+      "loss": 2.0744,
+      "step": 11745500
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 3.300620744951834e-05,
+      "loss": 2.1208,
+      "step": 11746000
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.6683752018054113,
+      "eval_accuracy_mlm": 0.6330868903676227,
+      "eval_accuracy_nsp": 0.8576833353957276,
+      "eval_loss": 2.1848316192626953,
+      "eval_runtime": 331.8363,
+      "eval_samples_per_second": 1315.064,
+      "eval_steps_per_second": 54.795,
+      "step": 11746048
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 3.300548380187107e-05,
+      "loss": 2.0541,
+      "step": 11746500
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 3.3004761601519085e-05,
+      "loss": 2.0843,
+      "step": 11747000
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 3.300403795387181e-05,
+      "loss": 2.0417,
+      "step": 11747500
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 3.300331430622453e-05,
+      "loss": 2.0662,
+      "step": 11748000
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 3.300259065857725e-05,
+      "loss": 2.0647,
+      "step": 11748500
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 3.3001867010929974e-05,
+      "loss": 2.052,
+      "step": 11749000
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 3.3001143363282696e-05,
+      "loss": 2.0667,
+      "step": 11749500
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 3.300042116293072e-05,
+      "loss": 2.0503,
+      "step": 11750000
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 3.299969751528344e-05,
+      "loss": 2.0598,
+      "step": 11750500
+    },
+    {
+      "epoch": 34.01,
+      "learning_rate": 3.2998975314931456e-05,
+      "loss": 2.0656,
+      "step": 11751000
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 3.299825166728418e-05,
+      "loss": 2.0474,
+      "step": 11751500
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 3.29975280196369e-05,
+      "loss": 2.0609,
+      "step": 11752000
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 3.299680581928492e-05,
+      "loss": 2.0618,
+      "step": 11752500
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 3.2996082171637645e-05,
+      "loss": 2.0827,
+      "step": 11753000
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 3.299535852399037e-05,
+      "loss": 2.0515,
+      "step": 11753500
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 3.299463487634309e-05,
+      "loss": 2.0554,
+      "step": 11754000
+    },
+    {
+      "epoch": 34.02,
+      "learning_rate": 3.299391122869582e-05,
+      "loss": 2.0759,
+      "step": 11754500
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 3.299318758104854e-05,
+      "loss": 2.0613,
+      "step": 11755000
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 3.299246393340126e-05,
+      "loss": 2.0762,
+      "step": 11755500
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 3.2991740285753985e-05,
+      "loss": 2.0491,
+      "step": 11756000
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 3.299101663810671e-05,
+      "loss": 2.0709,
+      "step": 11756500
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 3.299029443775472e-05,
+      "loss": 2.0752,
+      "step": 11757000
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 3.298957079010745e-05,
+      "loss": 2.063,
+      "step": 11757500
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 3.2988847142460174e-05,
+      "loss": 2.0705,
+      "step": 11758000
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 3.29881234948129e-05,
+      "loss": 2.0852,
+      "step": 11758500
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 3.298740129446091e-05,
+      "loss": 2.0856,
+      "step": 11759000
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 3.2986677646813635e-05,
+      "loss": 2.073,
+      "step": 11759500
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 3.298595689375695e-05,
+      "loss": 2.0644,
+      "step": 11760000
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 3.298523324610967e-05,
+      "loss": 2.061,
+      "step": 11760500
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 3.2984509598462395e-05,
+      "loss": 2.0538,
+      "step": 11761000
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 3.298378595081512e-05,
+      "loss": 2.0657,
+      "step": 11761500
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 3.2983062303167846e-05,
+      "loss": 2.0939,
+      "step": 11762000
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 3.298233865552057e-05,
+      "loss": 2.0421,
+      "step": 11762500
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 3.298161500787329e-05,
+      "loss": 2.0746,
+      "step": 11763000
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 3.298089136022601e-05,
+      "loss": 2.0625,
+      "step": 11763500
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 3.298016915987403e-05,
+      "loss": 2.0915,
+      "step": 11764000
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 3.297944551222675e-05,
+      "loss": 2.0715,
+      "step": 11764500
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 3.297872186457948e-05,
+      "loss": 2.0651,
+      "step": 11765000
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 3.29779982169322e-05,
+      "loss": 2.1096,
+      "step": 11765500
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 3.2977274569284924e-05,
+      "loss": 2.0695,
+      "step": 11766000
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 3.2976550921637646e-05,
+      "loss": 2.0827,
+      "step": 11766500
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 3.297582727399037e-05,
+      "loss": 2.0595,
+      "step": 11767000
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 3.297510362634309e-05,
+      "loss": 2.0558,
+      "step": 11767500
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 3.297437997869581e-05,
+      "loss": 2.0537,
+      "step": 11768000
+    },
+    {
+      "epoch": 34.06,
+      "learning_rate": 3.297365777834383e-05,
+      "loss": 2.0665,
+      "step": 11768500
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 3.297293413069655e-05,
+      "loss": 2.0891,
+      "step": 11769000
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 3.297221048304928e-05,
+      "loss": 2.0659,
+      "step": 11769500
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 3.2971486835402e-05,
+      "loss": 2.0969,
+      "step": 11770000
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 3.2970763187754724e-05,
+      "loss": 2.0883,
+      "step": 11770500
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 3.297004243469804e-05,
+      "loss": 2.0772,
+      "step": 11771000
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 3.296931878705076e-05,
+      "loss": 2.0877,
+      "step": 11771500
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 3.2968595139403484e-05,
+      "loss": 2.0845,
+      "step": 11772000
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 3.2967871491756206e-05,
+      "loss": 2.0875,
+      "step": 11772500
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 3.296714784410893e-05,
+      "loss": 2.0819,
+      "step": 11773000
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 3.296642419646165e-05,
+      "loss": 2.0609,
+      "step": 11773500
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 3.296570054881438e-05,
+      "loss": 2.0783,
+      "step": 11774000
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 3.29649769011671e-05,
+      "loss": 2.084,
+      "step": 11774500
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 3.2964253253519824e-05,
+      "loss": 2.0782,
+      "step": 11775000
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 3.2963529605872547e-05,
+      "loss": 2.0791,
+      "step": 11775500
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 3.296280740552056e-05,
+      "loss": 2.058,
+      "step": 11776000
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 3.2962083757873284e-05,
+      "loss": 2.0687,
+      "step": 11776500
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 3.296136011022601e-05,
+      "loss": 2.0633,
+      "step": 11777000
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 3.2960636462578736e-05,
+      "loss": 2.0861,
+      "step": 11777500
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 3.295991281493146e-05,
+      "loss": 2.0464,
+      "step": 11778000
+    },
+    {
+      "epoch": 34.09,
+      "learning_rate": 3.295918916728418e-05,
+      "loss": 2.0552,
+      "step": 11778500
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 3.29584655196369e-05,
+      "loss": 2.0559,
+      "step": 11779000
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 3.295774187198963e-05,
+      "loss": 2.0849,
+      "step": 11779500
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 3.2957018224342354e-05,
+      "loss": 2.0898,
+      "step": 11780000
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 3.2956294576695076e-05,
+      "loss": 2.0519,
+      "step": 11780500
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 3.29555709290478e-05,
+      "loss": 2.071,
+      "step": 11781000
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 3.295484728140052e-05,
+      "loss": 2.0886,
+      "step": 11781500
+    },
+    {
+      "epoch": 34.1,
+      "learning_rate": 3.295412363375324e-05,
+      "loss": 2.0841,
+      "step": 11782000
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 3.2953399986105965e-05,
+      "loss": 2.0711,
+      "step": 11782500
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 3.295267633845869e-05,
+      "loss": 2.084,
+      "step": 11783000
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 3.295195269081141e-05,
+      "loss": 2.0786,
+      "step": 11783500
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 3.295122904316414e-05,
+      "loss": 2.0444,
+      "step": 11784000
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 3.2950506842812154e-05,
+      "loss": 2.0636,
+      "step": 11784500
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 3.2949784642460176e-05,
+      "loss": 2.068,
+      "step": 11785000
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 3.294906244210819e-05,
+      "loss": 2.0591,
+      "step": 11785500
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 3.2948338794460914e-05,
+      "loss": 2.0663,
+      "step": 11786000
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 3.2947615146813636e-05,
+      "loss": 2.059,
+      "step": 11786500
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 3.294689294646165e-05,
+      "loss": 2.0956,
+      "step": 11787000
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 3.294616929881438e-05,
+      "loss": 2.078,
+      "step": 11787500
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 3.29454456511671e-05,
+      "loss": 2.0863,
+      "step": 11788000
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 3.2944722003519825e-05,
+      "loss": 2.0534,
+      "step": 11788500
+    },
+    {
+      "epoch": 34.12,
+      "learning_rate": 3.294399835587255e-05,
+      "loss": 2.0767,
+      "step": 11789000
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 3.294327470822527e-05,
+      "loss": 2.0804,
+      "step": 11789500
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 3.294255106057799e-05,
+      "loss": 2.0808,
+      "step": 11790000
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 3.2941827412930714e-05,
+      "loss": 2.0672,
+      "step": 11790500
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 3.2941103765283436e-05,
+      "loss": 2.0873,
+      "step": 11791000
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 3.294038011763616e-05,
+      "loss": 2.0539,
+      "step": 11791500
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 3.293965646998889e-05,
+      "loss": 2.0846,
+      "step": 11792000
+    },
+    {
+      "epoch": 34.13,
+      "learning_rate": 3.293893282234161e-05,
+      "loss": 2.0529,
+      "step": 11792500
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 3.293820917469433e-05,
+      "loss": 2.071,
+      "step": 11793000
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 3.2937486974342354e-05,
+      "loss": 2.0645,
+      "step": 11793500
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 3.2936763326695077e-05,
+      "loss": 2.0782,
+      "step": 11794000
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 3.29360396790478e-05,
+      "loss": 2.0544,
+      "step": 11794500
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 3.293531603140052e-05,
+      "loss": 2.0782,
+      "step": 11795000
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 3.293459238375324e-05,
+      "loss": 2.0772,
+      "step": 11795500
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 3.2933868736105966e-05,
+      "loss": 2.0607,
+      "step": 11796000
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 3.293314508845869e-05,
+      "loss": 2.082,
+      "step": 11796500
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 3.29324228881067e-05,
+      "loss": 2.0909,
+      "step": 11797000
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 3.2931700687754726e-05,
+      "loss": 2.0745,
+      "step": 11797500
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 3.293097704010745e-05,
+      "loss": 2.0617,
+      "step": 11798000
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 3.293025339246017e-05,
+      "loss": 2.0706,
+      "step": 11798500
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 3.292952974481289e-05,
+      "loss": 2.0894,
+      "step": 11799000
+    },
+    {
+      "epoch": 34.15,
+      "learning_rate": 3.292880609716562e-05,
+      "loss": 2.0677,
+      "step": 11799500
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 3.292808389681364e-05,
+      "loss": 2.0815,
+      "step": 11800000
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 3.292736024916636e-05,
+      "loss": 2.09,
+      "step": 11800500
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 3.292663660151908e-05,
+      "loss": 2.0706,
+      "step": 11801000
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 3.2925912953871804e-05,
+      "loss": 2.0848,
+      "step": 11801500
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 3.292518930622453e-05,
+      "loss": 2.0687,
+      "step": 11802000
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 3.2924465658577255e-05,
+      "loss": 2.0518,
+      "step": 11802500
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 3.292374345822527e-05,
+      "loss": 2.0762,
+      "step": 11803000
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 3.292301981057799e-05,
+      "loss": 2.0949,
+      "step": 11803500
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 3.2922296162930715e-05,
+      "loss": 2.0569,
+      "step": 11804000
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 3.292157251528344e-05,
+      "loss": 2.0685,
+      "step": 11804500
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 3.292085031493146e-05,
+      "loss": 2.0883,
+      "step": 11805000
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 3.292012666728418e-05,
+      "loss": 2.0779,
+      "step": 11805500
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 3.2919403019636904e-05,
+      "loss": 2.096,
+      "step": 11806000
+    },
+    {
+      "epoch": 34.17,
+      "learning_rate": 3.2918679371989626e-05,
+      "loss": 2.0712,
+      "step": 11806500
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 3.2917955724342355e-05,
+      "loss": 2.0623,
+      "step": 11807000
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 3.291723207669508e-05,
+      "loss": 2.0802,
+      "step": 11807500
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 3.29165084290478e-05,
+      "loss": 2.0674,
+      "step": 11808000
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 3.291578478140052e-05,
+      "loss": 2.0715,
+      "step": 11808500
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 3.291506258104854e-05,
+      "loss": 2.0689,
+      "step": 11809000
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 3.291433893340126e-05,
+      "loss": 2.0656,
+      "step": 11809500
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 3.291361528575398e-05,
+      "loss": 2.0631,
+      "step": 11810000
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 3.291289163810671e-05,
+      "loss": 2.0952,
+      "step": 11810500
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 3.291216799045943e-05,
+      "loss": 2.0691,
+      "step": 11811000
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 3.2911444342812155e-05,
+      "loss": 2.093,
+      "step": 11811500
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 3.291072069516488e-05,
+      "loss": 2.0667,
+      "step": 11812000
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 3.29099970475176e-05,
+      "loss": 2.0783,
+      "step": 11812500
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 3.290927339987032e-05,
+      "loss": 2.0768,
+      "step": 11813000
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 3.2908549752223044e-05,
+      "loss": 2.0774,
+      "step": 11813500
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 3.290782610457577e-05,
+      "loss": 2.0762,
+      "step": 11814000
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 3.2907102456928496e-05,
+      "loss": 2.0663,
+      "step": 11814500
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 3.290637880928122e-05,
+      "loss": 2.0947,
+      "step": 11815000
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 3.290565516163394e-05,
+      "loss": 2.092,
+      "step": 11815500
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 3.290493151398666e-05,
+      "loss": 2.0701,
+      "step": 11816000
+    },
+    {
+      "epoch": 34.2,
+      "learning_rate": 3.2904209313634685e-05,
+      "loss": 2.0864,
+      "step": 11816500
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 3.29034871132827e-05,
+      "loss": 2.0738,
+      "step": 11817000
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 3.290276346563542e-05,
+      "loss": 2.0635,
+      "step": 11817500
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 3.2902039817988145e-05,
+      "loss": 2.0787,
+      "step": 11818000
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 3.290131617034087e-05,
+      "loss": 2.0751,
+      "step": 11818500
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 3.290059252269359e-05,
+      "loss": 2.0808,
+      "step": 11819000
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 3.289986887504631e-05,
+      "loss": 2.0526,
+      "step": 11819500
+    },
+    {
+      "epoch": 34.21,
+      "learning_rate": 3.2899145227399033e-05,
+      "loss": 2.0808,
+      "step": 11820000
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 3.289842157975176e-05,
+      "loss": 2.0637,
+      "step": 11820500
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 3.2897697932104485e-05,
+      "loss": 2.0806,
+      "step": 11821000
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 3.2896974284457214e-05,
+      "loss": 2.0874,
+      "step": 11821500
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 3.2896250636809936e-05,
+      "loss": 2.079,
+      "step": 11822000
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 3.289552843645795e-05,
+      "loss": 2.0667,
+      "step": 11822500
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 3.2894804788810674e-05,
+      "loss": 2.0665,
+      "step": 11823000
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 3.2894081141163396e-05,
+      "loss": 2.0862,
+      "step": 11823500
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 3.289335749351612e-05,
+      "loss": 2.0858,
+      "step": 11824000
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 3.289263384586884e-05,
+      "loss": 2.0813,
+      "step": 11824500
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 3.289191019822156e-05,
+      "loss": 2.0796,
+      "step": 11825000
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 3.2891186550574285e-05,
+      "loss": 2.0635,
+      "step": 11825500
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 3.2890462902927014e-05,
+      "loss": 2.101,
+      "step": 11826000
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 3.2889739255279736e-05,
+      "loss": 2.0824,
+      "step": 11826500
+    },
+    {
+      "epoch": 34.23,
+      "learning_rate": 3.288901705492775e-05,
+      "loss": 2.0819,
+      "step": 11827000
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 3.2888293407280474e-05,
+      "loss": 2.0891,
+      "step": 11827500
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 3.2887569759633196e-05,
+      "loss": 2.0645,
+      "step": 11828000
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 3.2886846111985925e-05,
+      "loss": 2.0446,
+      "step": 11828500
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 3.288612391163394e-05,
+      "loss": 2.0595,
+      "step": 11829000
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 3.288540026398666e-05,
+      "loss": 2.0683,
+      "step": 11829500
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 3.2884676616339385e-05,
+      "loss": 2.0903,
+      "step": 11830000
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 3.2883952968692114e-05,
+      "loss": 2.0702,
+      "step": 11830500
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 3.2883229321044836e-05,
+      "loss": 2.0579,
+      "step": 11831000
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 3.288250567339756e-05,
+      "loss": 2.0784,
+      "step": 11831500
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 3.288178202575028e-05,
+      "loss": 2.0906,
+      "step": 11832000
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 3.2881058378103e-05,
+      "loss": 2.0933,
+      "step": 11832500
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 3.2880334730455725e-05,
+      "loss": 2.0592,
+      "step": 11833000
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 3.287961108280845e-05,
+      "loss": 2.0663,
+      "step": 11833500
+    },
+    {
+      "epoch": 34.25,
+      "learning_rate": 3.287888743516117e-05,
+      "loss": 2.0624,
+      "step": 11834000
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 3.2878165234809185e-05,
+      "loss": 2.0725,
+      "step": 11834500
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 3.2877441587161914e-05,
+      "loss": 2.071,
+      "step": 11835000
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 3.287671793951464e-05,
+      "loss": 2.0681,
+      "step": 11835500
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 3.2875994291867366e-05,
+      "loss": 2.0825,
+      "step": 11836000
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 3.287527209151538e-05,
+      "loss": 2.0762,
+      "step": 11836500
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 3.28745498911634e-05,
+      "loss": 2.0685,
+      "step": 11837000
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 3.287382624351612e-05,
+      "loss": 2.0911,
+      "step": 11837500
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 3.287310259586884e-05,
+      "loss": 2.0727,
+      "step": 11838000
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 3.2872378948221564e-05,
+      "loss": 2.0694,
+      "step": 11838500
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 3.2871656747869586e-05,
+      "loss": 2.0561,
+      "step": 11839000
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 3.287093310022231e-05,
+      "loss": 2.0549,
+      "step": 11839500
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 3.287020945257503e-05,
+      "loss": 2.0601,
+      "step": 11840000
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 3.286948580492775e-05,
+      "loss": 2.0514,
+      "step": 11840500
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 3.286876360457577e-05,
+      "loss": 2.0579,
+      "step": 11841000
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 3.286803995692849e-05,
+      "loss": 2.0559,
+      "step": 11841500
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 3.286731630928121e-05,
+      "loss": 2.0964,
+      "step": 11842000
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 3.2866592661633935e-05,
+      "loss": 2.0831,
+      "step": 11842500
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 3.2865869013986664e-05,
+      "loss": 2.0662,
+      "step": 11843000
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 3.2865145366339386e-05,
+      "loss": 2.056,
+      "step": 11843500
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 3.2864421718692115e-05,
+      "loss": 2.0835,
+      "step": 11844000
+    },
+    {
+      "epoch": 34.28,
+      "learning_rate": 3.286369807104484e-05,
+      "loss": 2.0764,
+      "step": 11844500
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 3.286297442339756e-05,
+      "loss": 2.0726,
+      "step": 11845000
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 3.286225077575028e-05,
+      "loss": 2.0781,
+      "step": 11845500
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 3.2861527128103004e-05,
+      "loss": 2.0948,
+      "step": 11846000
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 3.2860803480455726e-05,
+      "loss": 2.0848,
+      "step": 11846500
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 3.286007983280845e-05,
+      "loss": 2.0764,
+      "step": 11847000
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 3.285935618516117e-05,
+      "loss": 2.0622,
+      "step": 11847500
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 3.2858633984809186e-05,
+      "loss": 2.0842,
+      "step": 11848000
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 3.2857910337161915e-05,
+      "loss": 2.0715,
+      "step": 11848500
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 3.285718668951464e-05,
+      "loss": 2.078,
+      "step": 11849000
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 3.285646304186736e-05,
+      "loss": 2.0526,
+      "step": 11849500
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 3.285573939422008e-05,
+      "loss": 2.0827,
+      "step": 11850000
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 3.2855015746572804e-05,
+      "loss": 2.069,
+      "step": 11850500
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 3.285429209892553e-05,
+      "loss": 2.0583,
+      "step": 11851000
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 3.2853568451278255e-05,
+      "loss": 2.072,
+      "step": 11851500
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 3.285284625092627e-05,
+      "loss": 2.0703,
+      "step": 11852000
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 3.285212405057429e-05,
+      "loss": 2.0796,
+      "step": 11852500
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 3.2851400402927016e-05,
+      "loss": 2.0784,
+      "step": 11853000
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 3.285067675527974e-05,
+      "loss": 2.0633,
+      "step": 11853500
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 3.284995310763246e-05,
+      "loss": 2.0706,
+      "step": 11854000
+    },
+    {
+      "epoch": 34.31,
+      "learning_rate": 3.284922945998518e-05,
+      "loss": 2.0692,
+      "step": 11854500
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 3.2848505812337904e-05,
+      "loss": 2.0722,
+      "step": 11855000
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 3.284778216469063e-05,
+      "loss": 2.0788,
+      "step": 11855500
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 3.284705996433864e-05,
+      "loss": 2.0685,
+      "step": 11856000
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 3.2846336316691364e-05,
+      "loss": 2.0481,
+      "step": 11856500
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 3.2845612669044094e-05,
+      "loss": 2.0759,
+      "step": 11857000
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 3.284489046869211e-05,
+      "loss": 2.1118,
+      "step": 11857500
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 3.284416682104483e-05,
+      "loss": 2.0922,
+      "step": 11858000
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 3.284344317339756e-05,
+      "loss": 2.1051,
+      "step": 11858500
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 3.284271952575028e-05,
+      "loss": 2.0855,
+      "step": 11859000
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 3.2841995878103005e-05,
+      "loss": 2.0725,
+      "step": 11859500
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 3.284127367775102e-05,
+      "loss": 2.0398,
+      "step": 11860000
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 3.284055003010374e-05,
+      "loss": 2.0646,
+      "step": 11860500
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 3.2839826382456465e-05,
+      "loss": 2.0676,
+      "step": 11861000
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 3.2839102734809194e-05,
+      "loss": 2.0883,
+      "step": 11861500
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 3.2838379087161916e-05,
+      "loss": 2.0984,
+      "step": 11862000
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 3.283765543951464e-05,
+      "loss": 2.088,
+      "step": 11862500
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 3.283693179186736e-05,
+      "loss": 2.0904,
+      "step": 11863000
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 3.283620814422008e-05,
+      "loss": 2.0876,
+      "step": 11863500
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 3.2835484496572805e-05,
+      "loss": 2.0788,
+      "step": 11864000
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 3.283476229622082e-05,
+      "loss": 2.0978,
+      "step": 11864500
+    },
+    {
+      "epoch": 34.34,
+      "learning_rate": 3.283403864857354e-05,
+      "loss": 2.0833,
+      "step": 11865000
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 3.2833315000926265e-05,
+      "loss": 2.0943,
+      "step": 11865500
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 3.2832591353278994e-05,
+      "loss": 2.0586,
+      "step": 11866000
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 3.283187060022231e-05,
+      "loss": 2.0882,
+      "step": 11866500
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 3.283114695257503e-05,
+      "loss": 2.1042,
+      "step": 11867000
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 3.2830423304927754e-05,
+      "loss": 2.0841,
+      "step": 11867500
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 3.2829699657280476e-05,
+      "loss": 2.056,
+      "step": 11868000
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 3.28289760096332e-05,
+      "loss": 2.0748,
+      "step": 11868500
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 3.282825236198592e-05,
+      "loss": 2.065,
+      "step": 11869000
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 3.282753016163394e-05,
+      "loss": 2.0835,
+      "step": 11869500
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 3.2826806513986665e-05,
+      "loss": 2.0824,
+      "step": 11870000
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 3.282608286633939e-05,
+      "loss": 2.0795,
+      "step": 11870500
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 3.282535921869211e-05,
+      "loss": 2.0756,
+      "step": 11871000
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 3.282463557104483e-05,
+      "loss": 2.0781,
+      "step": 11871500
+    },
+    {
+      "epoch": 34.36,
+      "learning_rate": 3.2823911923397554e-05,
+      "loss": 2.0535,
+      "step": 11872000
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 3.2823188275750277e-05,
+      "loss": 2.0538,
+      "step": 11872500
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 3.2822464628103e-05,
+      "loss": 2.0771,
+      "step": 11873000
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 3.282174098045573e-05,
+      "loss": 2.098,
+      "step": 11873500
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 3.282101733280845e-05,
+      "loss": 2.0624,
+      "step": 11874000
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 3.282029368516117e-05,
+      "loss": 2.079,
+      "step": 11874500
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 3.2819570037513895e-05,
+      "loss": 2.0846,
+      "step": 11875000
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 3.281884638986662e-05,
+      "loss": 2.049,
+      "step": 11875500
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 3.2818122742219346e-05,
+      "loss": 2.073,
+      "step": 11876000
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 3.281740054186736e-05,
+      "loss": 2.0808,
+      "step": 11876500
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 3.2816676894220084e-05,
+      "loss": 2.0761,
+      "step": 11877000
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 3.2815953246572806e-05,
+      "loss": 2.0803,
+      "step": 11877500
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 3.281523104622082e-05,
+      "loss": 2.0762,
+      "step": 11878000
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 3.2814507398573544e-05,
+      "loss": 2.0462,
+      "step": 11878500
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 3.2813783750926266e-05,
+      "loss": 2.0695,
+      "step": 11879000
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 3.2813060103278995e-05,
+      "loss": 2.0565,
+      "step": 11879500
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 3.281233645563172e-05,
+      "loss": 2.0948,
+      "step": 11880000
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 3.281161425527973e-05,
+      "loss": 2.0787,
+      "step": 11880500
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 3.281089060763246e-05,
+      "loss": 2.0626,
+      "step": 11881000
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 3.2810166959985184e-05,
+      "loss": 2.0615,
+      "step": 11881500
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 3.2809443312337906e-05,
+      "loss": 2.0861,
+      "step": 11882000
+    },
+    {
+      "epoch": 34.39,
+      "learning_rate": 3.280871966469063e-05,
+      "loss": 2.0907,
+      "step": 11882500
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 3.280799601704335e-05,
+      "loss": 2.0611,
+      "step": 11883000
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 3.280727236939607e-05,
+      "loss": 2.0667,
+      "step": 11883500
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 3.2806548721748795e-05,
+      "loss": 2.0649,
+      "step": 11884000
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 3.280582507410152e-05,
+      "loss": 2.0867,
+      "step": 11884500
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 3.2805101426454246e-05,
+      "loss": 2.0984,
+      "step": 11885000
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 3.280437922610226e-05,
+      "loss": 2.0933,
+      "step": 11885500
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 3.2803655578454984e-05,
+      "loss": 2.0601,
+      "step": 11886000
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 3.2802933378103e-05,
+      "loss": 2.0586,
+      "step": 11886500
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 3.280220973045572e-05,
+      "loss": 2.0949,
+      "step": 11887000
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 3.2801486082808444e-05,
+      "loss": 2.0407,
+      "step": 11887500
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 3.280076243516117e-05,
+      "loss": 2.0776,
+      "step": 11888000
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 3.2800040234809195e-05,
+      "loss": 2.0552,
+      "step": 11888500
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 3.279931658716192e-05,
+      "loss": 2.0498,
+      "step": 11889000
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 3.279859438680993e-05,
+      "loss": 2.086,
+      "step": 11889500
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 3.2797870739162655e-05,
+      "loss": 2.0788,
+      "step": 11890000
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 3.279714709151538e-05,
+      "loss": 2.0848,
+      "step": 11890500
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 3.27964234438681e-05,
+      "loss": 2.0765,
+      "step": 11891000
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 3.279569979622082e-05,
+      "loss": 2.1022,
+      "step": 11891500
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 3.2794976148573544e-05,
+      "loss": 2.0715,
+      "step": 11892000
+    },
+    {
+      "epoch": 34.42,
+      "learning_rate": 3.279425250092627e-05,
+      "loss": 2.074,
+      "step": 11892500
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 3.2793528853278996e-05,
+      "loss": 2.074,
+      "step": 11893000
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 3.279280665292701e-05,
+      "loss": 2.0818,
+      "step": 11893500
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 3.279208300527973e-05,
+      "loss": 2.1109,
+      "step": 11894000
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 3.2791359357632456e-05,
+      "loss": 2.0853,
+      "step": 11894500
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 3.279063570998518e-05,
+      "loss": 2.0809,
+      "step": 11895000
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 3.27899120623379e-05,
+      "loss": 2.0612,
+      "step": 11895500
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 3.2789191309281216e-05,
+      "loss": 2.0567,
+      "step": 11896000
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 3.2788467661633945e-05,
+      "loss": 2.0724,
+      "step": 11896500
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 3.278774401398667e-05,
+      "loss": 2.0594,
+      "step": 11897000
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 3.278702036633939e-05,
+      "loss": 2.0726,
+      "step": 11897500
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 3.278629671869211e-05,
+      "loss": 2.0808,
+      "step": 11898000
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 3.2785573071044834e-05,
+      "loss": 2.0963,
+      "step": 11898500
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 3.2784849423397556e-05,
+      "loss": 2.0972,
+      "step": 11899000
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 3.278412577575028e-05,
+      "loss": 2.07,
+      "step": 11899500
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 3.2783402128103e-05,
+      "loss": 2.0636,
+      "step": 11900000
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 3.278267848045572e-05,
+      "loss": 2.0752,
+      "step": 11900500
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 3.2781954832808445e-05,
+      "loss": 2.0842,
+      "step": 11901000
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 3.2781231185161174e-05,
+      "loss": 2.0751,
+      "step": 11901500
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 3.2780507537513896e-05,
+      "loss": 2.0788,
+      "step": 11902000
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 3.277978388986662e-05,
+      "loss": 2.1004,
+      "step": 11902500
+    },
+    {
+      "epoch": 34.45,
+      "learning_rate": 3.277906024221934e-05,
+      "loss": 2.0559,
+      "step": 11903000
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 3.277833659457207e-05,
+      "loss": 2.0879,
+      "step": 11903500
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 3.2777614394220085e-05,
+      "loss": 2.0761,
+      "step": 11904000
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 3.277689074657281e-05,
+      "loss": 2.0718,
+      "step": 11904500
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 3.277616709892553e-05,
+      "loss": 2.0913,
+      "step": 11905000
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 3.277544345127825e-05,
+      "loss": 2.077,
+      "step": 11905500
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 3.2774719803630974e-05,
+      "loss": 2.0781,
+      "step": 11906000
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 3.2773996155983696e-05,
+      "loss": 2.0837,
+      "step": 11906500
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 3.2773272508336425e-05,
+      "loss": 2.078,
+      "step": 11907000
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 3.277254886068915e-05,
+      "loss": 2.0896,
+      "step": 11907500
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 3.277182521304187e-05,
+      "loss": 2.0808,
+      "step": 11908000
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 3.2771103012689885e-05,
+      "loss": 2.0781,
+      "step": 11908500
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 3.27703808123379e-05,
+      "loss": 2.0693,
+      "step": 11909000
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 3.276965716469062e-05,
+      "loss": 2.0693,
+      "step": 11909500
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 3.2768933517043345e-05,
+      "loss": 2.11,
+      "step": 11910000
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 3.2768209869396074e-05,
+      "loss": 2.0712,
+      "step": 11910500
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 3.2767486221748797e-05,
+      "loss": 2.0934,
+      "step": 11911000
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 3.276676402139682e-05,
+      "loss": 2.0813,
+      "step": 11911500
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 3.276604037374954e-05,
+      "loss": 2.0885,
+      "step": 11912000
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 3.2765316726102263e-05,
+      "loss": 2.0887,
+      "step": 11912500
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 3.2764593078454986e-05,
+      "loss": 2.0613,
+      "step": 11913000
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 3.276386943080771e-05,
+      "loss": 2.0774,
+      "step": 11913500
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 3.2763147230455723e-05,
+      "loss": 2.092,
+      "step": 11914000
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 3.276242358280845e-05,
+      "loss": 2.061,
+      "step": 11914500
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 3.2761699935161175e-05,
+      "loss": 2.0815,
+      "step": 11915000
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 3.27609762875139e-05,
+      "loss": 2.0528,
+      "step": 11915500
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 3.276025408716191e-05,
+      "loss": 2.1093,
+      "step": 11916000
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 3.2759530439514635e-05,
+      "loss": 2.0673,
+      "step": 11916500
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 3.275880679186736e-05,
+      "loss": 2.0648,
+      "step": 11917000
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 3.275808314422008e-05,
+      "loss": 2.0772,
+      "step": 11917500
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 3.27573594965728e-05,
+      "loss": 2.0927,
+      "step": 11918000
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 3.275663584892553e-05,
+      "loss": 2.0706,
+      "step": 11918500
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 3.275591220127825e-05,
+      "loss": 2.0917,
+      "step": 11919000
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 3.2755188553630975e-05,
+      "loss": 2.077,
+      "step": 11919500
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 3.2754464905983704e-05,
+      "loss": 2.0827,
+      "step": 11920000
+    },
+    {
+      "epoch": 34.5,
+      "learning_rate": 3.275374270563172e-05,
+      "loss": 2.068,
+      "step": 11920500
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 3.2753020505279735e-05,
+      "loss": 2.077,
+      "step": 11921000
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 3.275229685763246e-05,
+      "loss": 2.1051,
+      "step": 11921500
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 3.275157320998518e-05,
+      "loss": 2.0502,
+      "step": 11922000
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 3.27508510096332e-05,
+      "loss": 2.0981,
+      "step": 11922500
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 3.2750127361985924e-05,
+      "loss": 2.0931,
+      "step": 11923000
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 3.2749403714338646e-05,
+      "loss": 2.1017,
+      "step": 11923500
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 3.274868006669137e-05,
+      "loss": 2.0654,
+      "step": 11924000
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 3.274795641904409e-05,
+      "loss": 2.0828,
+      "step": 11924500
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 3.274723277139681e-05,
+      "loss": 2.0747,
+      "step": 11925000
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 3.2746509123749535e-05,
+      "loss": 2.1031,
+      "step": 11925500
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 3.2745785476102264e-05,
+      "loss": 2.0889,
+      "step": 11926000
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 3.274506327575028e-05,
+      "loss": 2.0766,
+      "step": 11926500
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 3.2744339628103e-05,
+      "loss": 2.0603,
+      "step": 11927000
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 3.2743615980455724e-05,
+      "loss": 2.0838,
+      "step": 11927500
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 3.274289233280845e-05,
+      "loss": 2.0849,
+      "step": 11928000
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 3.2742168685161175e-05,
+      "loss": 2.0595,
+      "step": 11928500
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 3.27414450375139e-05,
+      "loss": 2.0846,
+      "step": 11929000
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 3.274072138986662e-05,
+      "loss": 2.0898,
+      "step": 11929500
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 3.273999774221934e-05,
+      "loss": 2.0833,
+      "step": 11930000
+    },
+    {
+      "epoch": 34.53,
+      "learning_rate": 3.2739274094572064e-05,
+      "loss": 2.079,
+      "step": 11930500
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 3.273855044692479e-05,
+      "loss": 2.0859,
+      "step": 11931000
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 3.27378282465728e-05,
+      "loss": 2.0545,
+      "step": 11931500
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 3.2737104598925524e-05,
+      "loss": 2.087,
+      "step": 11932000
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 3.2736380951278253e-05,
+      "loss": 2.087,
+      "step": 11932500
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 3.2735657303630976e-05,
+      "loss": 2.0946,
+      "step": 11933000
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 3.2734933655983705e-05,
+      "loss": 2.0676,
+      "step": 11933500
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 3.273421145563172e-05,
+      "loss": 2.1048,
+      "step": 11934000
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 3.273348780798444e-05,
+      "loss": 2.0739,
+      "step": 11934500
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 3.273276705492775e-05,
+      "loss": 2.0723,
+      "step": 11935000
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 3.273204340728048e-05,
+      "loss": 2.0772,
+      "step": 11935500
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 3.2731321206928496e-05,
+      "loss": 2.0949,
+      "step": 11936000
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 3.273059755928122e-05,
+      "loss": 2.0744,
+      "step": 11936500
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 3.272987391163394e-05,
+      "loss": 2.0634,
+      "step": 11937000
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 3.272915026398666e-05,
+      "loss": 2.0971,
+      "step": 11937500
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 3.2728426616339385e-05,
+      "loss": 2.0943,
+      "step": 11938000
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 3.272770296869211e-05,
+      "loss": 2.0614,
+      "step": 11938500
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 3.272697932104483e-05,
+      "loss": 2.087,
+      "step": 11939000
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 3.272625567339755e-05,
+      "loss": 2.1045,
+      "step": 11939500
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 3.272553202575028e-05,
+      "loss": 2.0836,
+      "step": 11940000
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 3.2724808378103e-05,
+      "loss": 2.0468,
+      "step": 11940500
+    },
+    {
+      "epoch": 34.56,
+      "learning_rate": 3.272408473045573e-05,
+      "loss": 2.0786,
+      "step": 11941000
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 3.2723361082808454e-05,
+      "loss": 2.076,
+      "step": 11941500
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 3.2722637435161176e-05,
+      "loss": 2.0663,
+      "step": 11942000
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 3.27219137875139e-05,
+      "loss": 2.0839,
+      "step": 11942500
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 3.2721191587161914e-05,
+      "loss": 2.0755,
+      "step": 11943000
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 3.2720467939514636e-05,
+      "loss": 2.0853,
+      "step": 11943500
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 3.271974429186736e-05,
+      "loss": 2.0611,
+      "step": 11944000
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 3.271902209151538e-05,
+      "loss": 2.0878,
+      "step": 11944500
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 3.27182984438681e-05,
+      "loss": 2.0829,
+      "step": 11945000
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 3.2717574796220825e-05,
+      "loss": 2.0937,
+      "step": 11945500
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 3.271685114857355e-05,
+      "loss": 2.0874,
+      "step": 11946000
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 3.271612750092627e-05,
+      "loss": 2.0916,
+      "step": 11946500
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 3.271540385327899e-05,
+      "loss": 2.0858,
+      "step": 11947000
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 3.2714680205631714e-05,
+      "loss": 2.0838,
+      "step": 11947500
+    },
+    {
+      "epoch": 34.58,
+      "learning_rate": 3.2713956557984436e-05,
+      "loss": 2.0866,
+      "step": 11948000
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 3.2713232910337165e-05,
+      "loss": 2.0602,
+      "step": 11948500
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 3.271250926268989e-05,
+      "loss": 2.0881,
+      "step": 11949000
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 3.271178561504261e-05,
+      "loss": 2.0691,
+      "step": 11949500
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 3.271106196739533e-05,
+      "loss": 2.0895,
+      "step": 11950000
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 3.2710338319748054e-05,
+      "loss": 2.0833,
+      "step": 11950500
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 3.270961611939608e-05,
+      "loss": 2.0772,
+      "step": 11951000
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 3.27088924717488e-05,
+      "loss": 2.1088,
+      "step": 11951500
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 3.270816882410152e-05,
+      "loss": 2.0685,
+      "step": 11952000
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 3.2707445176454243e-05,
+      "loss": 2.0982,
+      "step": 11952500
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 3.2706721528806966e-05,
+      "loss": 2.0671,
+      "step": 11953000
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 3.270599788115969e-05,
+      "loss": 2.0903,
+      "step": 11953500
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 3.270527423351241e-05,
+      "loss": 2.0947,
+      "step": 11954000
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 3.270455058586513e-05,
+      "loss": 2.0933,
+      "step": 11954500
+    },
+    {
+      "epoch": 34.6,
+      "learning_rate": 3.2703826938217855e-05,
+      "loss": 2.1156,
+      "step": 11955000
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 3.2703104737865884e-05,
+      "loss": 2.1087,
+      "step": 11955500
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 3.2702381090218606e-05,
+      "loss": 2.0669,
+      "step": 11956000
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 3.270165888986662e-05,
+      "loss": 2.0991,
+      "step": 11956500
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 3.2700935242219344e-05,
+      "loss": 2.0677,
+      "step": 11957000
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 3.2700211594572066e-05,
+      "loss": 2.0821,
+      "step": 11957500
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 3.269948794692479e-05,
+      "loss": 2.0859,
+      "step": 11958000
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 3.269876429927751e-05,
+      "loss": 2.0726,
+      "step": 11958500
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 3.269804065163023e-05,
+      "loss": 2.0995,
+      "step": 11959000
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 3.2697317003982955e-05,
+      "loss": 2.0438,
+      "step": 11959500
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 3.2696593356335684e-05,
+      "loss": 2.0668,
+      "step": 11960000
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 3.2695869708688406e-05,
+      "loss": 2.082,
+      "step": 11960500
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 3.269514750833642e-05,
+      "loss": 2.0803,
+      "step": 11961000
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 3.269442530798444e-05,
+      "loss": 2.0648,
+      "step": 11961500
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 3.269370166033716e-05,
+      "loss": 2.0954,
+      "step": 11962000
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 3.269297801268988e-05,
+      "loss": 2.0851,
+      "step": 11962500
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 3.2692254365042604e-05,
+      "loss": 2.0841,
+      "step": 11963000
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 3.269153216469063e-05,
+      "loss": 2.09,
+      "step": 11963500
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 3.269080996433865e-05,
+      "loss": 2.0971,
+      "step": 11964000
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 3.269008631669137e-05,
+      "loss": 2.0969,
+      "step": 11964500
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 3.268936266904409e-05,
+      "loss": 2.0825,
+      "step": 11965000
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 3.2688639021396815e-05,
+      "loss": 2.0841,
+      "step": 11965500
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 3.268791537374954e-05,
+      "loss": 2.0821,
+      "step": 11966000
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 3.268719317339756e-05,
+      "loss": 2.0811,
+      "step": 11966500
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 3.268646952575028e-05,
+      "loss": 2.0811,
+      "step": 11967000
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 3.26857473253983e-05,
+      "loss": 2.0651,
+      "step": 11967500
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 3.268502367775102e-05,
+      "loss": 2.0607,
+      "step": 11968000
+    },
+    {
+      "epoch": 34.64,
+      "learning_rate": 3.268430003010374e-05,
+      "loss": 2.0755,
+      "step": 11968500
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 3.2683576382456464e-05,
+      "loss": 2.0628,
+      "step": 11969000
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 3.268285273480919e-05,
+      "loss": 2.1,
+      "step": 11969500
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 3.268212908716191e-05,
+      "loss": 2.0861,
+      "step": 11970000
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 3.268140543951463e-05,
+      "loss": 2.0819,
+      "step": 11970500
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 3.268068179186736e-05,
+      "loss": 2.0896,
+      "step": 11971000
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 3.267995814422008e-05,
+      "loss": 2.1143,
+      "step": 11971500
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 3.2679235943868105e-05,
+      "loss": 2.0821,
+      "step": 11972000
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 3.267851229622083e-05,
+      "loss": 2.0787,
+      "step": 11972500
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 3.267778864857355e-05,
+      "loss": 2.1069,
+      "step": 11973000
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 3.267706500092627e-05,
+      "loss": 2.0839,
+      "step": 11973500
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 3.2676341353278994e-05,
+      "loss": 2.0626,
+      "step": 11974000
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 3.2675617705631716e-05,
+      "loss": 2.0934,
+      "step": 11974500
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 3.267489405798444e-05,
+      "loss": 2.093,
+      "step": 11975000
+    },
+    {
+      "epoch": 34.66,
+      "learning_rate": 3.267417041033716e-05,
+      "loss": 2.0745,
+      "step": 11975500
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 3.267344676268988e-05,
+      "loss": 2.0733,
+      "step": 11976000
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 3.267272311504261e-05,
+      "loss": 2.0769,
+      "step": 11976500
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 3.2671999467395334e-05,
+      "loss": 2.0726,
+      "step": 11977000
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 3.267127726704335e-05,
+      "loss": 2.073,
+      "step": 11977500
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 3.267055361939607e-05,
+      "loss": 2.1022,
+      "step": 11978000
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 3.26698299717488e-05,
+      "loss": 2.102,
+      "step": 11978500
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 3.266910632410152e-05,
+      "loss": 2.0613,
+      "step": 11979000
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 3.2668382676454245e-05,
+      "loss": 2.0785,
+      "step": 11979500
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 3.266765902880697e-05,
+      "loss": 2.0822,
+      "step": 11980000
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 3.266693682845498e-05,
+      "loss": 2.0708,
+      "step": 11980500
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 3.266621318080771e-05,
+      "loss": 2.0768,
+      "step": 11981000
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 3.2665489533160434e-05,
+      "loss": 2.0807,
+      "step": 11981500
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 3.2664765885513156e-05,
+      "loss": 2.0663,
+      "step": 11982000
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 3.266404223786588e-05,
+      "loss": 2.0749,
+      "step": 11982500
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 3.2663320037513894e-05,
+      "loss": 2.0892,
+      "step": 11983000
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 3.2662596389866616e-05,
+      "loss": 2.0901,
+      "step": 11983500
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 3.266187418951463e-05,
+      "loss": 2.0713,
+      "step": 11984000
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 3.266115054186736e-05,
+      "loss": 2.0984,
+      "step": 11984500
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 3.2660428341515376e-05,
+      "loss": 2.0754,
+      "step": 11985000
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 3.26597046938681e-05,
+      "loss": 2.0672,
+      "step": 11985500
+    },
+    {
+      "epoch": 34.69,
+      "learning_rate": 3.265898104622083e-05,
+      "loss": 2.073,
+      "step": 11986000
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 3.265825739857355e-05,
+      "loss": 2.1204,
+      "step": 11986500
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 3.265753375092627e-05,
+      "loss": 2.0937,
+      "step": 11987000
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 3.2656810103278994e-05,
+      "loss": 2.0778,
+      "step": 11987500
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 3.265608645563172e-05,
+      "loss": 2.0797,
+      "step": 11988000
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 3.265536280798444e-05,
+      "loss": 2.0784,
+      "step": 11988500
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 3.265463916033716e-05,
+      "loss": 2.0834,
+      "step": 11989000
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 3.265391551268988e-05,
+      "loss": 2.094,
+      "step": 11989500
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 3.265319186504261e-05,
+      "loss": 2.0934,
+      "step": 11990000
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 3.2652468217395335e-05,
+      "loss": 2.0778,
+      "step": 11990500
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 3.265174456974806e-05,
+      "loss": 2.1027,
+      "step": 11991000
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 3.265102092210078e-05,
+      "loss": 2.0873,
+      "step": 11991500
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 3.26502972744535e-05,
+      "loss": 2.0908,
+      "step": 11992000
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 3.2649573626806224e-05,
+      "loss": 2.0577,
+      "step": 11992500
+    },
+    {
+      "epoch": 34.71,
+      "learning_rate": 3.264885142645424e-05,
+      "loss": 2.08,
+      "step": 11993000
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 3.264812777880697e-05,
+      "loss": 2.0687,
+      "step": 11993500
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 3.264740413115969e-05,
+      "loss": 2.0808,
+      "step": 11994000
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 3.264668048351241e-05,
+      "loss": 2.0877,
+      "step": 11994500
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 3.264595973045573e-05,
+      "loss": 2.0892,
+      "step": 11995000
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 3.264523608280845e-05,
+      "loss": 2.0852,
+      "step": 11995500
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 3.264451243516117e-05,
+      "loss": 2.0792,
+      "step": 11996000
+    },
+    {
+      "epoch": 34.72,
+      "learning_rate": 3.2643788787513895e-05,
+      "loss": 2.1024,
+      "step": 11996500
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 3.264306513986662e-05,
+      "loss": 2.074,
+      "step": 11997000
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 3.264234149221934e-05,
+      "loss": 2.0853,
+      "step": 11997500
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 3.264161929186736e-05,
+      "loss": 2.0843,
+      "step": 11998000
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 3.2640895644220084e-05,
+      "loss": 2.0797,
+      "step": 11998500
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 3.2640171996572806e-05,
+      "loss": 2.0811,
+      "step": 11999000
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 3.263944834892553e-05,
+      "loss": 2.0893,
+      "step": 11999500
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 3.263872470127825e-05,
+      "loss": 2.0723,
+      "step": 12000000
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 3.263800105363097e-05,
+      "loss": 2.0815,
+      "step": 12000500
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 3.26372774059837e-05,
+      "loss": 2.0744,
+      "step": 12001000
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 3.2636553758336424e-05,
+      "loss": 2.1033,
+      "step": 12001500
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 3.2635830110689146e-05,
+      "loss": 2.0763,
+      "step": 12002000
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 3.263510646304187e-05,
+      "loss": 2.087,
+      "step": 12002500
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 3.263438281539459e-05,
+      "loss": 2.1124,
+      "step": 12003000
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 3.2633662062337906e-05,
+      "loss": 2.0633,
+      "step": 12003500
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 3.263293841469063e-05,
+      "loss": 2.1015,
+      "step": 12004000
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 3.263221476704335e-05,
+      "loss": 2.0766,
+      "step": 12004500
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 3.263149111939607e-05,
+      "loss": 2.0924,
+      "step": 12005000
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 3.2630767471748795e-05,
+      "loss": 2.1032,
+      "step": 12005500
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 3.263004527139681e-05,
+      "loss": 2.0861,
+      "step": 12006000
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 3.262932162374954e-05,
+      "loss": 2.1026,
+      "step": 12006500
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 3.262859797610226e-05,
+      "loss": 2.1052,
+      "step": 12007000
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 3.2627874328454984e-05,
+      "loss": 2.0769,
+      "step": 12007500
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 3.2627152128103e-05,
+      "loss": 2.0665,
+      "step": 12008000
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 3.262642848045573e-05,
+      "loss": 2.0785,
+      "step": 12008500
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 3.262570483280845e-05,
+      "loss": 2.0918,
+      "step": 12009000
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 3.2624981185161173e-05,
+      "loss": 2.0614,
+      "step": 12009500
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 3.2624257537513896e-05,
+      "loss": 2.0881,
+      "step": 12010000
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 3.262353388986662e-05,
+      "loss": 2.0819,
+      "step": 12010500
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 3.262281024221934e-05,
+      "loss": 2.0733,
+      "step": 12011000
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 3.262208659457206e-05,
+      "loss": 2.0676,
+      "step": 12011500
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 3.262136294692479e-05,
+      "loss": 2.0768,
+      "step": 12012000
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 3.2620639299277514e-05,
+      "loss": 2.0678,
+      "step": 12012500
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 3.2619915651630236e-05,
+      "loss": 2.082,
+      "step": 12013000
+    },
+    {
+      "epoch": 34.77,
+      "learning_rate": 3.261919345127825e-05,
+      "loss": 2.0705,
+      "step": 12013500
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 3.2618469803630974e-05,
+      "loss": 2.0806,
+      "step": 12014000
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 3.2617746155983696e-05,
+      "loss": 2.0724,
+      "step": 12014500
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 3.261702250833642e-05,
+      "loss": 2.0845,
+      "step": 12015000
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 3.2616301755279734e-05,
+      "loss": 2.0712,
+      "step": 12015500
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 3.261557810763246e-05,
+      "loss": 2.0783,
+      "step": 12016000
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 3.2614854459985185e-05,
+      "loss": 2.0938,
+      "step": 12016500
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 3.261413081233791e-05,
+      "loss": 2.0853,
+      "step": 12017000
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 3.261340716469063e-05,
+      "loss": 2.0955,
+      "step": 12017500
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 3.261268351704335e-05,
+      "loss": 2.0813,
+      "step": 12018000
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 3.261196131669137e-05,
+      "loss": 2.0991,
+      "step": 12018500
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 3.261123766904409e-05,
+      "loss": 2.0716,
+      "step": 12019000
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 3.261051402139682e-05,
+      "loss": 2.0892,
+      "step": 12019500
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 3.260979037374954e-05,
+      "loss": 2.0958,
+      "step": 12020000
+    },
+    {
+      "epoch": 34.79,
+      "learning_rate": 3.260906672610226e-05,
+      "loss": 2.0905,
+      "step": 12020500
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 3.2608343078454985e-05,
+      "loss": 2.0892,
+      "step": 12021000
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 3.260761943080771e-05,
+      "loss": 2.096,
+      "step": 12021500
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 3.260689578316043e-05,
+      "loss": 2.0891,
+      "step": 12022000
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 3.2606173582808445e-05,
+      "loss": 2.0639,
+      "step": 12022500
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 3.260544993516117e-05,
+      "loss": 2.084,
+      "step": 12023000
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 3.2604726287513896e-05,
+      "loss": 2.0998,
+      "step": 12023500
+    },
+    {
+      "epoch": 34.8,
+      "learning_rate": 3.260400408716192e-05,
+      "loss": 2.0922,
+      "step": 12024000
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 3.260328043951464e-05,
+      "loss": 2.0703,
+      "step": 12024500
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 3.260255679186736e-05,
+      "loss": 2.0885,
+      "step": 12025000
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 3.2601833144220086e-05,
+      "loss": 2.0647,
+      "step": 12025500
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 3.260110949657281e-05,
+      "loss": 2.0588,
+      "step": 12026000
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 3.260038584892553e-05,
+      "loss": 2.0893,
+      "step": 12026500
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 3.259966220127825e-05,
+      "loss": 2.1022,
+      "step": 12027000
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 3.2598938553630974e-05,
+      "loss": 2.0795,
+      "step": 12027500
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 3.25982149059837e-05,
+      "loss": 2.0918,
+      "step": 12028000
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 3.259749125833642e-05,
+      "loss": 2.1076,
+      "step": 12028500
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 3.259676761068914e-05,
+      "loss": 2.0761,
+      "step": 12029000
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 3.259604396304187e-05,
+      "loss": 2.101,
+      "step": 12029500
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 3.259532031539459e-05,
+      "loss": 2.0778,
+      "step": 12030000
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 3.259459666774732e-05,
+      "loss": 2.0602,
+      "step": 12030500
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 3.2593873020100044e-05,
+      "loss": 2.0911,
+      "step": 12031000
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 3.2593149372452766e-05,
+      "loss": 2.0719,
+      "step": 12031500
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 3.259242572480549e-05,
+      "loss": 2.073,
+      "step": 12032000
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 3.25917049717488e-05,
+      "loss": 2.1019,
+      "step": 12032500
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 3.259098132410152e-05,
+      "loss": 2.0765,
+      "step": 12033000
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 3.259025767645424e-05,
+      "loss": 2.0886,
+      "step": 12033500
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 3.258953402880697e-05,
+      "loss": 2.1079,
+      "step": 12034000
+    },
+    {
+      "epoch": 34.83,
+      "learning_rate": 3.258881038115969e-05,
+      "loss": 2.0948,
+      "step": 12034500
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 3.2588086733512415e-05,
+      "loss": 2.0956,
+      "step": 12035000
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 3.258736308586514e-05,
+      "loss": 2.081,
+      "step": 12035500
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 3.258663943821786e-05,
+      "loss": 2.0815,
+      "step": 12036000
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 3.258591579057058e-05,
+      "loss": 2.0742,
+      "step": 12036500
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 3.258519503751389e-05,
+      "loss": 2.0836,
+      "step": 12037000
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 3.258447138986662e-05,
+      "loss": 2.062,
+      "step": 12037500
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 3.258374774221934e-05,
+      "loss": 2.0807,
+      "step": 12038000
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 3.258302409457207e-05,
+      "loss": 2.0817,
+      "step": 12038500
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 3.258230044692479e-05,
+      "loss": 2.0669,
+      "step": 12039000
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 3.2581576799277515e-05,
+      "loss": 2.074,
+      "step": 12039500
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 3.258085315163024e-05,
+      "loss": 2.0799,
+      "step": 12040000
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 3.258012950398296e-05,
+      "loss": 2.0888,
+      "step": 12040500
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 3.257940585633568e-05,
+      "loss": 2.0654,
+      "step": 12041000
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 3.2578682208688404e-05,
+      "loss": 2.0959,
+      "step": 12041500
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 3.2577958561041126e-05,
+      "loss": 2.0891,
+      "step": 12042000
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 3.257723491339385e-05,
+      "loss": 2.0797,
+      "step": 12042500
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 3.257651271304187e-05,
+      "loss": 2.0764,
+      "step": 12043000
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 3.257578906539459e-05,
+      "loss": 2.0892,
+      "step": 12043500
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 3.2575065417747315e-05,
+      "loss": 2.0833,
+      "step": 12044000
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 3.2574344664690624e-05,
+      "loss": 2.0869,
+      "step": 12044500
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 3.2573621017043347e-05,
+      "loss": 2.0733,
+      "step": 12045000
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 3.257289736939607e-05,
+      "loss": 2.089,
+      "step": 12045500
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 3.25721737217488e-05,
+      "loss": 2.0676,
+      "step": 12046000
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 3.257145007410152e-05,
+      "loss": 2.0774,
+      "step": 12046500
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 3.257072642645424e-05,
+      "loss": 2.0773,
+      "step": 12047000
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 3.257000277880697e-05,
+      "loss": 2.083,
+      "step": 12047500
+    },
+    {
+      "epoch": 34.87,
+      "learning_rate": 3.2569279131159693e-05,
+      "loss": 2.071,
+      "step": 12048000
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 3.256855693080771e-05,
+      "loss": 2.0971,
+      "step": 12048500
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 3.256783328316043e-05,
+      "loss": 2.0888,
+      "step": 12049000
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 3.2567109635513154e-05,
+      "loss": 2.0955,
+      "step": 12049500
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 3.2566385987865876e-05,
+      "loss": 2.0779,
+      "step": 12050000
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 3.25656637875139e-05,
+      "loss": 2.0785,
+      "step": 12050500
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 3.256494013986662e-05,
+      "loss": 2.074,
+      "step": 12051000
+    },
+    {
+      "epoch": 34.88,
+      "learning_rate": 3.256421649221934e-05,
+      "loss": 2.0935,
+      "step": 12051500
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 3.2563492844572065e-05,
+      "loss": 2.0813,
+      "step": 12052000
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 3.256276919692479e-05,
+      "loss": 2.0955,
+      "step": 12052500
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 3.2562045549277516e-05,
+      "loss": 2.073,
+      "step": 12053000
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 3.256132190163024e-05,
+      "loss": 2.0923,
+      "step": 12053500
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 3.2560599701278254e-05,
+      "loss": 2.0972,
+      "step": 12054000
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 3.2559876053630976e-05,
+      "loss": 2.0733,
+      "step": 12054500
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 3.25591524059837e-05,
+      "loss": 2.0845,
+      "step": 12055000
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 3.255842875833642e-05,
+      "loss": 2.0692,
+      "step": 12055500
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 3.255770511068915e-05,
+      "loss": 2.0942,
+      "step": 12056000
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 3.255698146304187e-05,
+      "loss": 2.0874,
+      "step": 12056500
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 3.2556257815394594e-05,
+      "loss": 2.0938,
+      "step": 12057000
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 3.255553561504261e-05,
+      "loss": 2.067,
+      "step": 12057500
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 3.255481196739533e-05,
+      "loss": 2.0708,
+      "step": 12058000
+    },
+    {
+      "epoch": 34.9,
+      "learning_rate": 3.2554088319748054e-05,
+      "loss": 2.0572,
+      "step": 12058500
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 3.2553364672100776e-05,
+      "loss": 2.0899,
+      "step": 12059000
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 3.25526410244535e-05,
+      "loss": 2.0797,
+      "step": 12059500
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 3.255191737680622e-05,
+      "loss": 2.0753,
+      "step": 12060000
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 3.255119372915895e-05,
+      "loss": 2.0831,
+      "step": 12060500
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 3.255047008151167e-05,
+      "loss": 2.0678,
+      "step": 12061000
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 3.2549747881159694e-05,
+      "loss": 2.0799,
+      "step": 12061500
+    },
+    {
+      "epoch": 34.91,
+      "learning_rate": 3.2549024233512417e-05,
+      "loss": 2.0813,
+      "step": 12062000
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 3.254830058586514e-05,
+      "loss": 2.1055,
+      "step": 12062500
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 3.254757693821786e-05,
+      "loss": 2.0941,
+      "step": 12063000
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 3.254685329057058e-05,
+      "loss": 2.0584,
+      "step": 12063500
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 3.25461310902186e-05,
+      "loss": 2.0878,
+      "step": 12064000
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 3.254540888986662e-05,
+      "loss": 2.1005,
+      "step": 12064500
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 3.254468524221934e-05,
+      "loss": 2.0908,
+      "step": 12065000
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 3.2543961594572066e-05,
+      "loss": 2.0991,
+      "step": 12065500
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 3.254323794692479e-05,
+      "loss": 2.0596,
+      "step": 12066000
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 3.254251429927751e-05,
+      "loss": 2.0855,
+      "step": 12066500
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 3.254179065163023e-05,
+      "loss": 2.078,
+      "step": 12067000
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 3.2541067003982955e-05,
+      "loss": 2.088,
+      "step": 12067500
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 3.2540343356335684e-05,
+      "loss": 2.1078,
+      "step": 12068000
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 3.25396211559837e-05,
+      "loss": 2.0772,
+      "step": 12068500
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 3.253889750833642e-05,
+      "loss": 2.0777,
+      "step": 12069000
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 3.253817386068915e-05,
+      "loss": 2.0688,
+      "step": 12069500
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 3.253745021304187e-05,
+      "loss": 2.0972,
+      "step": 12070000
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 3.2536726565394595e-05,
+      "loss": 2.1017,
+      "step": 12070500
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 3.253600291774732e-05,
+      "loss": 2.0907,
+      "step": 12071000
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 3.253528071739533e-05,
+      "loss": 2.1142,
+      "step": 12071500
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 3.2534557069748055e-05,
+      "loss": 2.0709,
+      "step": 12072000
+    },
+    {
+      "epoch": 34.94,
+      "learning_rate": 3.253383342210078e-05,
+      "loss": 2.0945,
+      "step": 12072500
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 3.25331097744535e-05,
+      "loss": 2.0985,
+      "step": 12073000
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 3.253238612680622e-05,
+      "loss": 2.0889,
+      "step": 12073500
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 3.253166247915895e-05,
+      "loss": 2.0878,
+      "step": 12074000
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 3.253093883151167e-05,
+      "loss": 2.0581,
+      "step": 12074500
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 3.2530215183864395e-05,
+      "loss": 2.0765,
+      "step": 12075000
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 3.2529491536217124e-05,
+      "loss": 2.1122,
+      "step": 12075500
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 3.252876933586514e-05,
+      "loss": 2.0669,
+      "step": 12076000
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 3.252804568821786e-05,
+      "loss": 2.0982,
+      "step": 12076500
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 3.2527322040570584e-05,
+      "loss": 2.0815,
+      "step": 12077000
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 3.2526598392923306e-05,
+      "loss": 2.1089,
+      "step": 12077500
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 3.252587619257132e-05,
+      "loss": 2.0947,
+      "step": 12078000
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 3.252515254492405e-05,
+      "loss": 2.071,
+      "step": 12078500
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 3.252442889727677e-05,
+      "loss": 2.0594,
+      "step": 12079000
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 3.2523705249629495e-05,
+      "loss": 2.081,
+      "step": 12079500
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 3.252298160198222e-05,
+      "loss": 2.0784,
+      "step": 12080000
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 3.252225795433494e-05,
+      "loss": 2.0592,
+      "step": 12080500
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 3.252153430668766e-05,
+      "loss": 2.0773,
+      "step": 12081000
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 3.2520810659040384e-05,
+      "loss": 2.091,
+      "step": 12081500
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 3.2520087011393106e-05,
+      "loss": 2.0784,
+      "step": 12082000
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 3.251936336374583e-05,
+      "loss": 2.0763,
+      "step": 12082500
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 3.251863971609856e-05,
+      "loss": 2.0959,
+      "step": 12083000
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 3.251791751574657e-05,
+      "loss": 2.0677,
+      "step": 12083500
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 3.2517195315394596e-05,
+      "loss": 2.1041,
+      "step": 12084000
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 3.251647166774732e-05,
+      "loss": 2.0712,
+      "step": 12084500
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 3.251574802010004e-05,
+      "loss": 2.0801,
+      "step": 12085000
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 3.251502437245276e-05,
+      "loss": 2.0483,
+      "step": 12085500
+    },
+    {
+      "epoch": 34.98,
+      "learning_rate": 3.2514300724805485e-05,
+      "loss": 2.0741,
+      "step": 12086000
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 3.251357707715821e-05,
+      "loss": 2.0703,
+      "step": 12086500
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 3.251285342951093e-05,
+      "loss": 2.095,
+      "step": 12087000
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 3.251212978186365e-05,
+      "loss": 2.0847,
+      "step": 12087500
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 3.2511406134216373e-05,
+      "loss": 2.0831,
+      "step": 12088000
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 3.25106824865691e-05,
+      "loss": 2.1114,
+      "step": 12088500
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 3.2509958838921825e-05,
+      "loss": 2.0647,
+      "step": 12089000
+    },
+    {
+      "epoch": 34.99,
+      "learning_rate": 3.250923663856984e-05,
+      "loss": 2.0886,
+      "step": 12089500
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 3.250851299092256e-05,
+      "loss": 2.0871,
+      "step": 12090000
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 3.250778934327529e-05,
+      "loss": 2.087,
+      "step": 12090500
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 3.2507065695628014e-05,
+      "loss": 2.0969,
+      "step": 12091000
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 3.250634349527603e-05,
+      "loss": 2.0697,
+      "step": 12091500
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy": 0.6687623409562915,
+      "eval_accuracy_mlm": 0.6335090121750351,
+      "eval_accuracy_nsp": 0.8578345776445624,
+      "eval_loss": 2.168201446533203,
+      "eval_runtime": 331.5723,
+      "eval_samples_per_second": 1316.111,
+      "eval_steps_per_second": 54.839,
+      "step": 12091520
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 3.250561984762875e-05,
+      "loss": 2.0743,
+      "step": 12092000
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 3.250489619998148e-05,
+      "loss": 2.0658,
+      "step": 12092500
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 3.25041725523342e-05,
+      "loss": 2.0524,
+      "step": 12093000
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 3.2503448904686925e-05,
+      "loss": 2.052,
+      "step": 12093500
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 3.250272525703965e-05,
+      "loss": 2.0333,
+      "step": 12094000
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 3.250200160939237e-05,
+      "loss": 2.0897,
+      "step": 12094500
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 3.250127796174509e-05,
+      "loss": 2.0747,
+      "step": 12095000
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 3.2500554314097814e-05,
+      "loss": 2.066,
+      "step": 12095500
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 3.249983211374583e-05,
+      "loss": 2.0713,
+      "step": 12096000
+    },
+    {
+      "epoch": 35.01,
+      "learning_rate": 3.249910991339385e-05,
+      "loss": 2.0707,
+      "step": 12096500
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 3.2498386265746574e-05,
+      "loss": 2.0554,
+      "step": 12097000
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 3.249766406539459e-05,
+      "loss": 2.0761,
+      "step": 12097500
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 3.249694041774732e-05,
+      "loss": 2.0655,
+      "step": 12098000
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 3.249621966469063e-05,
+      "loss": 2.0674,
+      "step": 12098500
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 3.2495496017043356e-05,
+      "loss": 2.0642,
+      "step": 12099000
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 3.249477236939608e-05,
+      "loss": 2.0858,
+      "step": 12099500
+    },
+    {
+      "epoch": 35.02,
+      "learning_rate": 3.24940487217488e-05,
+      "loss": 2.0681,
+      "step": 12100000
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 3.2493326521396817e-05,
+      "loss": 2.0769,
+      "step": 12100500
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 3.249260287374954e-05,
+      "loss": 2.0707,
+      "step": 12101000
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 3.249187922610226e-05,
+      "loss": 2.0592,
+      "step": 12101500
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 3.249115557845498e-05,
+      "loss": 2.0667,
+      "step": 12102000
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 3.2490431930807705e-05,
+      "loss": 2.0766,
+      "step": 12102500
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 3.248970828316043e-05,
+      "loss": 2.0546,
+      "step": 12103000
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 3.248898463551316e-05,
+      "loss": 2.0661,
+      "step": 12103500
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 3.248826098786588e-05,
+      "loss": 2.0761,
+      "step": 12104000
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 3.24875373402186e-05,
+      "loss": 2.0575,
+      "step": 12104500
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 3.2486813692571323e-05,
+      "loss": 2.0716,
+      "step": 12105000
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 3.248609149221934e-05,
+      "loss": 2.0664,
+      "step": 12105500
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 3.248536784457207e-05,
+      "loss": 2.0647,
+      "step": 12106000
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 3.248464419692479e-05,
+      "loss": 2.0736,
+      "step": 12106500
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 3.248392054927751e-05,
+      "loss": 2.0873,
+      "step": 12107000
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 3.2483196901630235e-05,
+      "loss": 2.056,
+      "step": 12107500
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 3.248247325398296e-05,
+      "loss": 2.0707,
+      "step": 12108000
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 3.248174960633568e-05,
+      "loss": 2.0668,
+      "step": 12108500
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 3.24810274059837e-05,
+      "loss": 2.0785,
+      "step": 12109000
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 3.2480303758336424e-05,
+      "loss": 2.047,
+      "step": 12109500
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 3.2479580110689146e-05,
+      "loss": 2.0746,
+      "step": 12110000
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 3.247885646304187e-05,
+      "loss": 2.0683,
+      "step": 12110500
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 3.247813281539459e-05,
+      "loss": 2.048,
+      "step": 12111000
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 3.247740916774731e-05,
+      "loss": 2.07,
+      "step": 12111500
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 3.2476685520100035e-05,
+      "loss": 2.0647,
+      "step": 12112000
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 3.247596187245276e-05,
+      "loss": 2.0839,
+      "step": 12112500
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 3.2475238224805486e-05,
+      "loss": 2.0798,
+      "step": 12113000
+    },
+    {
+      "epoch": 35.06,
+      "learning_rate": 3.247451457715821e-05,
+      "loss": 2.0349,
+      "step": 12113500
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 3.247379092951093e-05,
+      "loss": 2.0658,
+      "step": 12114000
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 3.247306728186365e-05,
+      "loss": 2.074,
+      "step": 12114500
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 3.247234363421638e-05,
+      "loss": 2.0921,
+      "step": 12115000
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 3.2471619986569104e-05,
+      "loss": 2.0795,
+      "step": 12115500
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 3.2470896338921826e-05,
+      "loss": 2.0553,
+      "step": 12116000
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 3.247017269127455e-05,
+      "loss": 2.101,
+      "step": 12116500
+    },
+    {
+      "epoch": 35.07,
+      "learning_rate": 3.246944904362727e-05,
+      "loss": 2.0659,
+      "step": 12117000
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 3.2468726843275286e-05,
+      "loss": 2.0565,
+      "step": 12117500
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 3.246800319562801e-05,
+      "loss": 2.0969,
+      "step": 12118000
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 3.246728099527603e-05,
+      "loss": 2.0751,
+      "step": 12118500
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 3.246655734762875e-05,
+      "loss": 2.0692,
+      "step": 12119000
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 3.2465833699981475e-05,
+      "loss": 2.0561,
+      "step": 12119500
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 3.24651100523342e-05,
+      "loss": 2.0483,
+      "step": 12120000
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 3.2464386404686927e-05,
+      "loss": 2.0668,
+      "step": 12120500
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 3.2463665651630235e-05,
+      "loss": 2.0676,
+      "step": 12121000
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 3.246294200398296e-05,
+      "loss": 2.0721,
+      "step": 12121500
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 3.246221835633568e-05,
+      "loss": 2.0589,
+      "step": 12122000
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 3.246149470868841e-05,
+      "loss": 2.0807,
+      "step": 12122500
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 3.246077106104113e-05,
+      "loss": 2.0684,
+      "step": 12123000
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 3.2460047413393853e-05,
+      "loss": 2.0988,
+      "step": 12123500
+    },
+    {
+      "epoch": 35.09,
+      "learning_rate": 3.2459323765746576e-05,
+      "loss": 2.0418,
+      "step": 12124000
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 3.24586001180993e-05,
+      "loss": 2.0782,
+      "step": 12124500
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 3.245787647045202e-05,
+      "loss": 2.0636,
+      "step": 12125000
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 3.2457154270100036e-05,
+      "loss": 2.0665,
+      "step": 12125500
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 3.245643062245276e-05,
+      "loss": 2.0548,
+      "step": 12126000
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 3.245570697480548e-05,
+      "loss": 2.0844,
+      "step": 12126500
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 3.245498332715821e-05,
+      "loss": 2.0619,
+      "step": 12127000
+    },
+    {
+      "epoch": 35.1,
+      "learning_rate": 3.2454261126806225e-05,
+      "loss": 2.0653,
+      "step": 12127500
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 3.2453537479158954e-05,
+      "loss": 2.0521,
+      "step": 12128000
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 3.245281672610226e-05,
+      "loss": 2.0764,
+      "step": 12128500
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 3.2452093078454985e-05,
+      "loss": 2.0577,
+      "step": 12129000
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 3.245136943080771e-05,
+      "loss": 2.0606,
+      "step": 12129500
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 3.245064867775102e-05,
+      "loss": 2.0783,
+      "step": 12130000
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 3.2449925030103745e-05,
+      "loss": 2.0718,
+      "step": 12130500
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 3.244920138245647e-05,
+      "loss": 2.0621,
+      "step": 12131000
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 3.244847773480919e-05,
+      "loss": 2.0813,
+      "step": 12131500
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 3.244775408716191e-05,
+      "loss": 2.0704,
+      "step": 12132000
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 3.2447030439514634e-05,
+      "loss": 2.076,
+      "step": 12132500
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 3.2446306791867356e-05,
+      "loss": 2.0878,
+      "step": 12133000
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 3.2445583144220085e-05,
+      "loss": 2.0542,
+      "step": 12133500
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 3.244485949657281e-05,
+      "loss": 2.089,
+      "step": 12134000
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 3.244413729622082e-05,
+      "loss": 2.059,
+      "step": 12134500
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 3.2443413648573545e-05,
+      "loss": 2.0647,
+      "step": 12135000
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 3.244269000092627e-05,
+      "loss": 2.0436,
+      "step": 12135500
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 3.2441966353278996e-05,
+      "loss": 2.0623,
+      "step": 12136000
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 3.244124270563172e-05,
+      "loss": 2.0721,
+      "step": 12136500
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 3.2440520505279734e-05,
+      "loss": 2.0778,
+      "step": 12137000
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 3.2439796857632456e-05,
+      "loss": 2.0532,
+      "step": 12137500
+    },
+    {
+      "epoch": 35.13,
+      "learning_rate": 3.2439073209985185e-05,
+      "loss": 2.0734,
+      "step": 12138000
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 3.243834956233791e-05,
+      "loss": 2.0808,
+      "step": 12138500
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 3.243762591469063e-05,
+      "loss": 2.0473,
+      "step": 12139000
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 3.243690226704335e-05,
+      "loss": 2.0713,
+      "step": 12139500
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 3.2436178619396074e-05,
+      "loss": 2.0456,
+      "step": 12140000
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 3.2435454971748797e-05,
+      "loss": 2.0842,
+      "step": 12140500
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 3.243473132410152e-05,
+      "loss": 2.0818,
+      "step": 12141000
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 3.243400767645424e-05,
+      "loss": 2.0641,
+      "step": 12141500
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 3.243328402880696e-05,
+      "loss": 2.0575,
+      "step": 12142000
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 3.2432560381159686e-05,
+      "loss": 2.0435,
+      "step": 12142500
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 3.2431836733512415e-05,
+      "loss": 2.0739,
+      "step": 12143000
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 3.243111308586514e-05,
+      "loss": 2.0605,
+      "step": 12143500
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 3.243038943821786e-05,
+      "loss": 2.0506,
+      "step": 12144000
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 3.242966579057059e-05,
+      "loss": 2.0518,
+      "step": 12144500
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 3.2428943590218604e-05,
+      "loss": 2.0576,
+      "step": 12145000
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 3.2428219942571326e-05,
+      "loss": 2.0692,
+      "step": 12145500
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 3.242749629492405e-05,
+      "loss": 2.0478,
+      "step": 12146000
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 3.242677264727677e-05,
+      "loss": 2.0601,
+      "step": 12146500
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 3.2426050446924786e-05,
+      "loss": 2.0757,
+      "step": 12147000
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 3.242532679927751e-05,
+      "loss": 2.0788,
+      "step": 12147500
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 3.242460315163024e-05,
+      "loss": 2.0752,
+      "step": 12148000
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 3.242387950398296e-05,
+      "loss": 2.0685,
+      "step": 12148500
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 3.242315585633568e-05,
+      "loss": 2.0691,
+      "step": 12149000
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 3.2422432208688404e-05,
+      "loss": 2.0572,
+      "step": 12149500
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 3.2421708561041126e-05,
+      "loss": 2.0729,
+      "step": 12150000
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 3.2420984913393855e-05,
+      "loss": 2.0654,
+      "step": 12150500
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 3.242026126574658e-05,
+      "loss": 2.0406,
+      "step": 12151000
+    },
+    {
+      "epoch": 35.17,
+      "learning_rate": 3.241953906539459e-05,
+      "loss": 2.0694,
+      "step": 12151500
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 3.2418815417747315e-05,
+      "loss": 2.0702,
+      "step": 12152000
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 3.241809177010004e-05,
+      "loss": 2.0556,
+      "step": 12152500
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 3.241736812245276e-05,
+      "loss": 2.0581,
+      "step": 12153000
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 3.241664447480549e-05,
+      "loss": 2.0782,
+      "step": 12153500
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 3.241592082715821e-05,
+      "loss": 2.0882,
+      "step": 12154000
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 3.241519717951093e-05,
+      "loss": 2.0635,
+      "step": 12154500
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 3.2414473531863655e-05,
+      "loss": 2.0851,
+      "step": 12155000
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 3.241374988421638e-05,
+      "loss": 2.0636,
+      "step": 12155500
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 3.241302768386439e-05,
+      "loss": 2.0619,
+      "step": 12156000
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 3.2412304036217115e-05,
+      "loss": 2.0762,
+      "step": 12156500
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 3.241158038856984e-05,
+      "loss": 2.0818,
+      "step": 12157000
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 3.241085674092256e-05,
+      "loss": 2.0783,
+      "step": 12157500
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 3.241013309327529e-05,
+      "loss": 2.0624,
+      "step": 12158000
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 3.240941089292331e-05,
+      "loss": 2.0692,
+      "step": 12158500
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 3.240868724527603e-05,
+      "loss": 2.0675,
+      "step": 12159000
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 3.240796504492405e-05,
+      "loss": 2.0897,
+      "step": 12159500
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 3.240724139727677e-05,
+      "loss": 2.0824,
+      "step": 12160000
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 3.240651774962949e-05,
+      "loss": 2.0612,
+      "step": 12160500
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 3.2405794101982216e-05,
+      "loss": 2.0581,
+      "step": 12161000
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 3.240507045433494e-05,
+      "loss": 2.0456,
+      "step": 12161500
+    },
+    {
+      "epoch": 35.2,
+      "learning_rate": 3.240434680668766e-05,
+      "loss": 2.0833,
+      "step": 12162000
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 3.240362315904039e-05,
+      "loss": 2.0646,
+      "step": 12162500
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 3.2402900958688405e-05,
+      "loss": 2.0602,
+      "step": 12163000
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 3.240217731104113e-05,
+      "loss": 2.068,
+      "step": 12163500
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 3.240145366339385e-05,
+      "loss": 2.0979,
+      "step": 12164000
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 3.240073001574657e-05,
+      "loss": 2.0633,
+      "step": 12164500
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 3.240000781539459e-05,
+      "loss": 2.076,
+      "step": 12165000
+    },
+    {
+      "epoch": 35.21,
+      "learning_rate": 3.2399284167747316e-05,
+      "loss": 2.0641,
+      "step": 12165500
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 3.239856052010004e-05,
+      "loss": 2.075,
+      "step": 12166000
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 3.239783687245277e-05,
+      "loss": 2.0591,
+      "step": 12166500
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 3.239711322480549e-05,
+      "loss": 2.0698,
+      "step": 12167000
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 3.239638957715821e-05,
+      "loss": 2.073,
+      "step": 12167500
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 3.2395665929510934e-05,
+      "loss": 2.0493,
+      "step": 12168000
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 3.239494372915895e-05,
+      "loss": 2.0493,
+      "step": 12168500
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 3.239422008151167e-05,
+      "loss": 2.0375,
+      "step": 12169000
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 3.2393496433864394e-05,
+      "loss": 2.0633,
+      "step": 12169500
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 3.2392774233512416e-05,
+      "loss": 2.0761,
+      "step": 12170000
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 3.239205203316043e-05,
+      "loss": 2.0731,
+      "step": 12170500
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 3.2391328385513154e-05,
+      "loss": 2.0812,
+      "step": 12171000
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 3.2390604737865876e-05,
+      "loss": 2.0638,
+      "step": 12171500
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 3.23898810902186e-05,
+      "loss": 2.09,
+      "step": 12172000
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 3.238915744257132e-05,
+      "loss": 2.0494,
+      "step": 12172500
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 3.238843379492405e-05,
+      "loss": 2.0893,
+      "step": 12173000
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 3.238771014727677e-05,
+      "loss": 2.0811,
+      "step": 12173500
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 3.2386986499629494e-05,
+      "loss": 2.0592,
+      "step": 12174000
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 3.2386262851982216e-05,
+      "loss": 2.0701,
+      "step": 12174500
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 3.238553920433494e-05,
+      "loss": 2.0845,
+      "step": 12175000
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 3.238481555668767e-05,
+      "loss": 2.0674,
+      "step": 12175500
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 3.238409190904039e-05,
+      "loss": 2.0488,
+      "step": 12176000
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 3.238336826139311e-05,
+      "loss": 2.0668,
+      "step": 12176500
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 3.238264606104113e-05,
+      "loss": 2.078,
+      "step": 12177000
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 3.238192241339385e-05,
+      "loss": 2.0522,
+      "step": 12177500
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 3.2381200213041865e-05,
+      "loss": 2.0394,
+      "step": 12178000
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 3.238047656539459e-05,
+      "loss": 2.0748,
+      "step": 12178500
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 3.2379752917747317e-05,
+      "loss": 2.0557,
+      "step": 12179000
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 3.237902927010004e-05,
+      "loss": 2.081,
+      "step": 12179500
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 3.237830562245276e-05,
+      "loss": 2.0678,
+      "step": 12180000
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 3.237758197480549e-05,
+      "loss": 2.0674,
+      "step": 12180500
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 3.237685832715821e-05,
+      "loss": 2.0892,
+      "step": 12181000
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 3.2376134679510935e-05,
+      "loss": 2.0745,
+      "step": 12181500
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 3.237541103186366e-05,
+      "loss": 2.0829,
+      "step": 12182000
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 3.237468738421638e-05,
+      "loss": 2.0612,
+      "step": 12182500
+    },
+    {
+      "epoch": 35.26,
+      "learning_rate": 3.23739637365691e-05,
+      "loss": 2.0788,
+      "step": 12183000
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 3.2373240088921823e-05,
+      "loss": 2.0533,
+      "step": 12183500
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 3.2372516441274546e-05,
+      "loss": 2.0654,
+      "step": 12184000
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 3.237179279362727e-05,
+      "loss": 2.0455,
+      "step": 12184500
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 3.237106914597999e-05,
+      "loss": 2.0516,
+      "step": 12185000
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 3.237034549833272e-05,
+      "loss": 2.0756,
+      "step": 12185500
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 3.2369623297980735e-05,
+      "loss": 2.0543,
+      "step": 12186000
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 3.236889965033346e-05,
+      "loss": 2.0646,
+      "step": 12186500
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 3.236817600268618e-05,
+      "loss": 2.074,
+      "step": 12187000
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 3.2367453802334195e-05,
+      "loss": 2.0673,
+      "step": 12187500
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 3.2366730154686924e-05,
+      "loss": 2.0652,
+      "step": 12188000
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 3.2366006507039646e-05,
+      "loss": 2.0729,
+      "step": 12188500
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 3.236528430668767e-05,
+      "loss": 2.0495,
+      "step": 12189000
+    },
+    {
+      "epoch": 35.28,
+      "learning_rate": 3.236456065904039e-05,
+      "loss": 2.0616,
+      "step": 12189500
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 3.236383701139311e-05,
+      "loss": 2.0432,
+      "step": 12190000
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 3.2363113363745835e-05,
+      "loss": 2.0543,
+      "step": 12190500
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 3.236238971609856e-05,
+      "loss": 2.0598,
+      "step": 12191000
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 3.236166606845128e-05,
+      "loss": 2.0285,
+      "step": 12191500
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 3.2360942420804e-05,
+      "loss": 2.0781,
+      "step": 12192000
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 3.236022022045202e-05,
+      "loss": 2.0553,
+      "step": 12192500
+    },
+    {
+      "epoch": 35.29,
+      "learning_rate": 3.235949657280474e-05,
+      "loss": 2.1,
+      "step": 12193000
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 3.235877292515747e-05,
+      "loss": 2.0696,
+      "step": 12193500
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 3.235804927751019e-05,
+      "loss": 2.0732,
+      "step": 12194000
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 3.235732562986291e-05,
+      "loss": 2.0585,
+      "step": 12194500
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 3.235660198221564e-05,
+      "loss": 2.078,
+      "step": 12195000
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 3.2355878334568364e-05,
+      "loss": 2.0572,
+      "step": 12195500
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 3.2355154686921086e-05,
+      "loss": 2.0912,
+      "step": 12196000
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 3.235443103927381e-05,
+      "loss": 2.0968,
+      "step": 12196500
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 3.2353708838921824e-05,
+      "loss": 2.0851,
+      "step": 12197000
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 3.2352986638569847e-05,
+      "loss": 2.0791,
+      "step": 12197500
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 3.235226299092257e-05,
+      "loss": 2.0744,
+      "step": 12198000
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 3.235153934327529e-05,
+      "loss": 2.0541,
+      "step": 12198500
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 3.235081569562801e-05,
+      "loss": 2.0565,
+      "step": 12199000
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 3.235009349527603e-05,
+      "loss": 2.0762,
+      "step": 12199500
+    },
+    {
+      "epoch": 35.31,
+      "learning_rate": 3.234936984762875e-05,
+      "loss": 2.0674,
+      "step": 12200000
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 3.234864619998147e-05,
+      "loss": 2.0827,
+      "step": 12200500
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 3.2347922552334196e-05,
+      "loss": 2.0736,
+      "step": 12201000
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 3.234720035198222e-05,
+      "loss": 2.0705,
+      "step": 12201500
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 3.234647670433494e-05,
+      "loss": 2.0769,
+      "step": 12202000
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 3.234575305668766e-05,
+      "loss": 2.0783,
+      "step": 12202500
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 3.234502940904039e-05,
+      "loss": 2.0699,
+      "step": 12203000
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 3.2344305761393114e-05,
+      "loss": 2.0631,
+      "step": 12203500
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 3.234358500833642e-05,
+      "loss": 2.0913,
+      "step": 12204000
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 3.2342861360689145e-05,
+      "loss": 2.0568,
+      "step": 12204500
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 3.234213771304187e-05,
+      "loss": 2.0799,
+      "step": 12205000
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 3.2341414065394596e-05,
+      "loss": 2.0778,
+      "step": 12205500
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 3.234069041774732e-05,
+      "loss": 2.0746,
+      "step": 12206000
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 3.233996677010004e-05,
+      "loss": 2.088,
+      "step": 12206500
+    },
+    {
+      "epoch": 35.33,
+      "learning_rate": 3.2339244569748056e-05,
+      "loss": 2.0757,
+      "step": 12207000
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 3.233852092210078e-05,
+      "loss": 2.0521,
+      "step": 12207500
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 3.23377972744535e-05,
+      "loss": 2.0711,
+      "step": 12208000
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 3.233707362680622e-05,
+      "loss": 2.0797,
+      "step": 12208500
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 3.2336349979158945e-05,
+      "loss": 2.0569,
+      "step": 12209000
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 3.233562633151167e-05,
+      "loss": 2.0608,
+      "step": 12209500
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 3.2334902683864396e-05,
+      "loss": 2.0726,
+      "step": 12210000
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 3.233417903621712e-05,
+      "loss": 2.037,
+      "step": 12210500
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 3.233345538856985e-05,
+      "loss": 2.0704,
+      "step": 12211000
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 3.233273174092257e-05,
+      "loss": 2.0634,
+      "step": 12211500
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 3.233200809327529e-05,
+      "loss": 2.0502,
+      "step": 12212000
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 3.2331284445628014e-05,
+      "loss": 2.1014,
+      "step": 12212500
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 3.2330560797980736e-05,
+      "loss": 2.0574,
+      "step": 12213000
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 3.232983715033346e-05,
+      "loss": 2.0642,
+      "step": 12213500
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 3.232911350268618e-05,
+      "loss": 2.0668,
+      "step": 12214000
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 3.23283898550389e-05,
+      "loss": 2.063,
+      "step": 12214500
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 3.232766910198222e-05,
+      "loss": 2.0976,
+      "step": 12215000
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 3.232694545433494e-05,
+      "loss": 2.0702,
+      "step": 12215500
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 3.2326223253982956e-05,
+      "loss": 2.0849,
+      "step": 12216000
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 3.232549960633568e-05,
+      "loss": 2.0877,
+      "step": 12216500
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 3.23247759586884e-05,
+      "loss": 2.0614,
+      "step": 12217000
+    },
+    {
+      "epoch": 35.36,
+      "learning_rate": 3.232405231104112e-05,
+      "loss": 2.0972,
+      "step": 12217500
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 3.232332866339385e-05,
+      "loss": 2.1018,
+      "step": 12218000
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 3.2322605015746574e-05,
+      "loss": 2.033,
+      "step": 12218500
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 3.23218828153946e-05,
+      "loss": 2.0856,
+      "step": 12219000
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 3.232115916774732e-05,
+      "loss": 2.0785,
+      "step": 12219500
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 3.232043552010004e-05,
+      "loss": 2.0803,
+      "step": 12220000
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 3.2319711872452763e-05,
+      "loss": 2.0695,
+      "step": 12220500
+    },
+    {
+      "epoch": 35.37,
+      "learning_rate": 3.2318988224805486e-05,
+      "loss": 2.0613,
+      "step": 12221000
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 3.231826457715821e-05,
+      "loss": 2.0619,
+      "step": 12221500
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 3.231754092951093e-05,
+      "loss": 2.0543,
+      "step": 12222000
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 3.2316818729158946e-05,
+      "loss": 2.0491,
+      "step": 12222500
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 3.2316095081511675e-05,
+      "loss": 2.072,
+      "step": 12223000
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 3.23153714338644e-05,
+      "loss": 2.0871,
+      "step": 12223500
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 3.231464778621712e-05,
+      "loss": 2.0799,
+      "step": 12224000
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 3.231392413856984e-05,
+      "loss": 2.0892,
+      "step": 12224500
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 3.2313200490922564e-05,
+      "loss": 2.0758,
+      "step": 12225000
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 3.2312478290570586e-05,
+      "loss": 2.0857,
+      "step": 12225500
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 3.231175464292331e-05,
+      "loss": 2.0655,
+      "step": 12226000
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 3.231103099527603e-05,
+      "loss": 2.0632,
+      "step": 12226500
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 3.231030734762875e-05,
+      "loss": 2.0652,
+      "step": 12227000
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 3.2309583699981475e-05,
+      "loss": 2.0707,
+      "step": 12227500
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 3.23088600523342e-05,
+      "loss": 2.0817,
+      "step": 12228000
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 3.2308136404686926e-05,
+      "loss": 2.0837,
+      "step": 12228500
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 3.230741275703965e-05,
+      "loss": 2.0564,
+      "step": 12229000
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 3.230668910939237e-05,
+      "loss": 2.0624,
+      "step": 12229500
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 3.230596546174509e-05,
+      "loss": 2.0808,
+      "step": 12230000
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 3.2305241814097815e-05,
+      "loss": 2.0792,
+      "step": 12230500
+    },
+    {
+      "epoch": 35.4,
+      "learning_rate": 3.230451961374583e-05,
+      "loss": 2.0936,
+      "step": 12231000
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 3.230379596609855e-05,
+      "loss": 2.0718,
+      "step": 12231500
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 3.2303072318451275e-05,
+      "loss": 2.0643,
+      "step": 12232000
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 3.2302348670804004e-05,
+      "loss": 2.1005,
+      "step": 12232500
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 3.2301625023156726e-05,
+      "loss": 2.062,
+      "step": 12233000
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 3.230090282280475e-05,
+      "loss": 2.0647,
+      "step": 12233500
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 3.230017917515747e-05,
+      "loss": 2.0661,
+      "step": 12234000
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 3.229945552751019e-05,
+      "loss": 2.0802,
+      "step": 12234500
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 3.2298731879862915e-05,
+      "loss": 2.0832,
+      "step": 12235000
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 3.229800967951093e-05,
+      "loss": 2.0676,
+      "step": 12235500
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 3.229728603186365e-05,
+      "loss": 2.0479,
+      "step": 12236000
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 3.2296562384216375e-05,
+      "loss": 2.0753,
+      "step": 12236500
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 3.22958387365691e-05,
+      "loss": 2.0754,
+      "step": 12237000
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 3.229511508892183e-05,
+      "loss": 2.0603,
+      "step": 12237500
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 3.229439144127455e-05,
+      "loss": 2.0779,
+      "step": 12238000
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 3.229366779362727e-05,
+      "loss": 2.0482,
+      "step": 12238500
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 3.229294414597999e-05,
+      "loss": 2.0655,
+      "step": 12239000
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 3.229222194562801e-05,
+      "loss": 2.0695,
+      "step": 12239500
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 3.229149829798074e-05,
+      "loss": 2.1075,
+      "step": 12240000
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 3.229077465033346e-05,
+      "loss": 2.0964,
+      "step": 12240500
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 3.229005100268618e-05,
+      "loss": 2.0552,
+      "step": 12241000
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 3.2289327355038905e-05,
+      "loss": 2.0803,
+      "step": 12241500
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 3.228860370739163e-05,
+      "loss": 2.0712,
+      "step": 12242000
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 3.228788005974435e-05,
+      "loss": 2.0827,
+      "step": 12242500
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 3.228715641209708e-05,
+      "loss": 2.0821,
+      "step": 12243000
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 3.22864327644498e-05,
+      "loss": 2.0582,
+      "step": 12243500
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 3.228570911680252e-05,
+      "loss": 2.0728,
+      "step": 12244000
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 3.228498691645054e-05,
+      "loss": 2.0619,
+      "step": 12244500
+    },
+    {
+      "epoch": 35.44,
+      "learning_rate": 3.228426326880326e-05,
+      "loss": 2.0783,
+      "step": 12245000
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 3.228353962115598e-05,
+      "loss": 2.056,
+      "step": 12245500
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 3.2282815973508705e-05,
+      "loss": 2.099,
+      "step": 12246000
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 3.228209232586143e-05,
+      "loss": 2.0765,
+      "step": 12246500
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 3.228136867821415e-05,
+      "loss": 2.0844,
+      "step": 12247000
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 3.228064503056688e-05,
+      "loss": 2.0828,
+      "step": 12247500
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 3.22799213829196e-05,
+      "loss": 2.0511,
+      "step": 12248000
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 3.227919918256762e-05,
+      "loss": 2.0754,
+      "step": 12248500
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 3.2278475534920345e-05,
+      "loss": 2.0722,
+      "step": 12249000
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 3.227775188727307e-05,
+      "loss": 2.0702,
+      "step": 12249500
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 3.227702823962579e-05,
+      "loss": 2.0926,
+      "step": 12250000
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 3.2276306039273805e-05,
+      "loss": 2.0378,
+      "step": 12250500
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 3.227558239162653e-05,
+      "loss": 2.0549,
+      "step": 12251000
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 3.227486019127455e-05,
+      "loss": 2.048,
+      "step": 12251500
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 3.227413654362727e-05,
+      "loss": 2.093,
+      "step": 12252000
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 3.227341434327529e-05,
+      "loss": 2.0532,
+      "step": 12252500
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 3.227269069562801e-05,
+      "loss": 2.064,
+      "step": 12253000
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 3.227196704798073e-05,
+      "loss": 2.0835,
+      "step": 12253500
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 3.2271243400333454e-05,
+      "loss": 2.0598,
+      "step": 12254000
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 3.2270519752686176e-05,
+      "loss": 2.0532,
+      "step": 12254500
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 3.2269796105038905e-05,
+      "loss": 2.0699,
+      "step": 12255000
+    },
+    {
+      "epoch": 35.47,
+      "learning_rate": 3.226907245739163e-05,
+      "loss": 2.0665,
+      "step": 12255500
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 3.226834880974435e-05,
+      "loss": 2.0671,
+      "step": 12256000
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 3.226762660939237e-05,
+      "loss": 2.0706,
+      "step": 12256500
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 3.2266902961745094e-05,
+      "loss": 2.0848,
+      "step": 12257000
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 3.226617931409782e-05,
+      "loss": 2.0727,
+      "step": 12257500
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 3.226545566645054e-05,
+      "loss": 2.0737,
+      "step": 12258000
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 3.226473201880326e-05,
+      "loss": 2.0648,
+      "step": 12258500
+    },
+    {
+      "epoch": 35.48,
+      "learning_rate": 3.226400837115598e-05,
+      "loss": 2.082,
+      "step": 12259000
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 3.2263284723508706e-05,
+      "loss": 2.0811,
+      "step": 12259500
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 3.226256252315673e-05,
+      "loss": 2.0753,
+      "step": 12260000
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 3.226183887550945e-05,
+      "loss": 2.0901,
+      "step": 12260500
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 3.226111522786217e-05,
+      "loss": 2.0706,
+      "step": 12261000
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 3.2260391580214895e-05,
+      "loss": 2.0491,
+      "step": 12261500
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 3.225966793256762e-05,
+      "loss": 2.0731,
+      "step": 12262000
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 3.2258944284920346e-05,
+      "loss": 2.0785,
+      "step": 12262500
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 3.225822063727307e-05,
+      "loss": 2.0675,
+      "step": 12263000
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 3.2257498436921084e-05,
+      "loss": 2.0879,
+      "step": 12263500
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 3.2256774789273806e-05,
+      "loss": 2.0721,
+      "step": 12264000
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 3.225605114162653e-05,
+      "loss": 2.0678,
+      "step": 12264500
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 3.225532749397926e-05,
+      "loss": 2.0605,
+      "step": 12265000
+    },
+    {
+      "epoch": 35.5,
+      "learning_rate": 3.225460384633198e-05,
+      "loss": 2.0596,
+      "step": 12265500
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 3.22538801986847e-05,
+      "loss": 2.076,
+      "step": 12266000
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 3.2253156551037424e-05,
+      "loss": 2.0753,
+      "step": 12266500
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 3.2252432903390146e-05,
+      "loss": 2.048,
+      "step": 12267000
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 3.225170925574287e-05,
+      "loss": 2.0902,
+      "step": 12267500
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 3.225098560809559e-05,
+      "loss": 2.0803,
+      "step": 12268000
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 3.225026196044831e-05,
+      "loss": 2.0709,
+      "step": 12268500
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 3.224953976009633e-05,
+      "loss": 2.0868,
+      "step": 12269000
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 3.224881611244906e-05,
+      "loss": 2.0854,
+      "step": 12269500
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 3.224809246480178e-05,
+      "loss": 2.0749,
+      "step": 12270000
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 3.22473702644498e-05,
+      "loss": 2.0713,
+      "step": 12270500
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 3.2246646616802524e-05,
+      "loss": 2.0581,
+      "step": 12271000
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 3.2245922969155246e-05,
+      "loss": 2.0842,
+      "step": 12271500
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 3.224519932150797e-05,
+      "loss": 2.0859,
+      "step": 12272000
+    },
+    {
+      "epoch": 35.52,
+      "learning_rate": 3.224447567386069e-05,
+      "loss": 2.0401,
+      "step": 12272500
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 3.224375202621341e-05,
+      "loss": 2.0788,
+      "step": 12273000
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 3.2243028378566135e-05,
+      "loss": 2.0872,
+      "step": 12273500
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 3.224230473091886e-05,
+      "loss": 2.0843,
+      "step": 12274000
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 3.224158108327158e-05,
+      "loss": 2.075,
+      "step": 12274500
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 3.22408588829196e-05,
+      "loss": 2.0868,
+      "step": 12275000
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 3.2240135235272324e-05,
+      "loss": 2.0877,
+      "step": 12275500
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 3.2239411587625047e-05,
+      "loss": 2.0911,
+      "step": 12276000
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 3.223868938727306e-05,
+      "loss": 2.085,
+      "step": 12276500
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 3.2237965739625784e-05,
+      "loss": 2.1145,
+      "step": 12277000
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 3.2237242091978513e-05,
+      "loss": 2.0601,
+      "step": 12277500
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 3.2236518444331236e-05,
+      "loss": 2.0697,
+      "step": 12278000
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 3.223579624397926e-05,
+      "loss": 2.0831,
+      "step": 12278500
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 3.223507259633198e-05,
+      "loss": 2.0835,
+      "step": 12279000
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 3.22343489486847e-05,
+      "loss": 2.0755,
+      "step": 12279500
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 3.2233625301037425e-05,
+      "loss": 2.0889,
+      "step": 12280000
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 3.223290165339015e-05,
+      "loss": 2.072,
+      "step": 12280500
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 3.223217800574287e-05,
+      "loss": 2.0646,
+      "step": 12281000
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 3.223145435809559e-05,
+      "loss": 2.0966,
+      "step": 12281500
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 3.2230730710448314e-05,
+      "loss": 2.0992,
+      "step": 12282000
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 3.2230007062801036e-05,
+      "loss": 2.0593,
+      "step": 12282500
+    },
+    {
+      "epoch": 35.55,
+      "learning_rate": 3.222928341515376e-05,
+      "loss": 2.069,
+      "step": 12283000
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 3.222856121480178e-05,
+      "loss": 2.0557,
+      "step": 12283500
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 3.22278375671545e-05,
+      "loss": 2.0895,
+      "step": 12284000
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 3.222711391950723e-05,
+      "loss": 2.0716,
+      "step": 12284500
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 3.2226390271859954e-05,
+      "loss": 2.0772,
+      "step": 12285000
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 3.2225666624212676e-05,
+      "loss": 2.0702,
+      "step": 12285500
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 3.22249429765654e-05,
+      "loss": 2.0564,
+      "step": 12286000
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 3.222421932891812e-05,
+      "loss": 2.0736,
+      "step": 12286500
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 3.2223497128566136e-05,
+      "loss": 2.0926,
+      "step": 12287000
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 3.222277348091886e-05,
+      "loss": 2.0628,
+      "step": 12287500
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 3.222204983327158e-05,
+      "loss": 2.0888,
+      "step": 12288000
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 3.222132618562431e-05,
+      "loss": 2.0615,
+      "step": 12288500
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 3.222060253797703e-05,
+      "loss": 2.0765,
+      "step": 12289000
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 3.2219878890329754e-05,
+      "loss": 2.0754,
+      "step": 12289500
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 3.2219155242682476e-05,
+      "loss": 2.0803,
+      "step": 12290000
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 3.22184315950352e-05,
+      "loss": 2.0889,
+      "step": 12290500
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 3.221770794738792e-05,
+      "loss": 2.069,
+      "step": 12291000
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 3.221698429974064e-05,
+      "loss": 2.0823,
+      "step": 12291500
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 3.221626065209337e-05,
+      "loss": 2.0598,
+      "step": 12292000
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 3.221553845174139e-05,
+      "loss": 2.0813,
+      "step": 12292500
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 3.221481480409411e-05,
+      "loss": 2.0689,
+      "step": 12293000
+    },
+    {
+      "epoch": 35.58,
+      "learning_rate": 3.221409115644683e-05,
+      "loss": 2.0732,
+      "step": 12293500
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 3.221336750879956e-05,
+      "loss": 2.0603,
+      "step": 12294000
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 3.221264386115228e-05,
+      "loss": 2.0729,
+      "step": 12294500
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 3.22119216608003e-05,
+      "loss": 2.072,
+      "step": 12295000
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 3.221119801315302e-05,
+      "loss": 2.0608,
+      "step": 12295500
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 3.221047436550574e-05,
+      "loss": 2.0845,
+      "step": 12296000
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 3.2209750717858466e-05,
+      "loss": 2.061,
+      "step": 12296500
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 3.220902707021119e-05,
+      "loss": 2.0877,
+      "step": 12297000
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 3.220830342256391e-05,
+      "loss": 2.0725,
+      "step": 12297500
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 3.220757977491663e-05,
+      "loss": 2.0828,
+      "step": 12298000
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 3.220685612726936e-05,
+      "loss": 2.058,
+      "step": 12298500
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 3.2206132479622083e-05,
+      "loss": 2.0852,
+      "step": 12299000
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 3.220540883197481e-05,
+      "loss": 2.036,
+      "step": 12299500
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 3.2204685184327535e-05,
+      "loss": 2.0623,
+      "step": 12300000
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 3.220396298397555e-05,
+      "loss": 2.0668,
+      "step": 12300500
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 3.220323933632827e-05,
+      "loss": 2.0614,
+      "step": 12301000
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 3.2202515688680995e-05,
+      "loss": 2.0814,
+      "step": 12301500
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 3.220179204103372e-05,
+      "loss": 2.0684,
+      "step": 12302000
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 3.220106839338644e-05,
+      "loss": 2.078,
+      "step": 12302500
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 3.220034474573916e-05,
+      "loss": 2.0873,
+      "step": 12303000
+    },
+    {
+      "epoch": 35.61,
+      "learning_rate": 3.2199621098091884e-05,
+      "loss": 2.0888,
+      "step": 12303500
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 3.219889745044461e-05,
+      "loss": 2.0935,
+      "step": 12304000
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 3.219817525009263e-05,
+      "loss": 2.0724,
+      "step": 12304500
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 3.2197453049740644e-05,
+      "loss": 2.0775,
+      "step": 12305000
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 3.219673084938866e-05,
+      "loss": 2.0745,
+      "step": 12305500
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 3.219600720174138e-05,
+      "loss": 2.0632,
+      "step": 12306000
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 3.219528355409411e-05,
+      "loss": 2.0734,
+      "step": 12306500
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 3.219455990644683e-05,
+      "loss": 2.0734,
+      "step": 12307000
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 3.219383625879956e-05,
+      "loss": 2.0821,
+      "step": 12307500
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 3.2193112611152284e-05,
+      "loss": 2.0717,
+      "step": 12308000
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 3.2192388963505006e-05,
+      "loss": 2.0798,
+      "step": 12308500
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 3.219166531585773e-05,
+      "loss": 2.0824,
+      "step": 12309000
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 3.219094166821045e-05,
+      "loss": 2.0678,
+      "step": 12309500
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 3.219021802056317e-05,
+      "loss": 2.0835,
+      "step": 12310000
+    },
+    {
+      "epoch": 35.63,
+      "learning_rate": 3.2189494372915895e-05,
+      "loss": 2.0984,
+      "step": 12310500
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 3.218877072526862e-05,
+      "loss": 2.0645,
+      "step": 12311000
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 3.218804852491663e-05,
+      "loss": 2.0843,
+      "step": 12311500
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 3.2187326324564655e-05,
+      "loss": 2.0683,
+      "step": 12312000
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 3.218660267691738e-05,
+      "loss": 2.0697,
+      "step": 12312500
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 3.21858790292701e-05,
+      "loss": 2.0849,
+      "step": 12313000
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 3.218515538162282e-05,
+      "loss": 2.0767,
+      "step": 12313500
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 3.218443173397555e-05,
+      "loss": 2.0676,
+      "step": 12314000
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 3.218370808632827e-05,
+      "loss": 2.0795,
+      "step": 12314500
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 3.218298588597629e-05,
+      "loss": 2.0858,
+      "step": 12315000
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 3.218226223832901e-05,
+      "loss": 2.0429,
+      "step": 12315500
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 3.218153859068174e-05,
+      "loss": 2.0856,
+      "step": 12316000
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 3.218081494303446e-05,
+      "loss": 2.0565,
+      "step": 12316500
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 3.218009274268248e-05,
+      "loss": 2.0792,
+      "step": 12317000
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 3.2179370542330493e-05,
+      "loss": 2.0675,
+      "step": 12317500
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 3.2178646894683216e-05,
+      "loss": 2.0846,
+      "step": 12318000
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 3.217792324703594e-05,
+      "loss": 2.0871,
+      "step": 12318500
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 3.217719959938866e-05,
+      "loss": 2.0927,
+      "step": 12319000
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 3.217647595174139e-05,
+      "loss": 2.0976,
+      "step": 12319500
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 3.217575230409411e-05,
+      "loss": 2.069,
+      "step": 12320000
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 3.2175028656446834e-05,
+      "loss": 2.069,
+      "step": 12320500
+    },
+    {
+      "epoch": 35.66,
+      "learning_rate": 3.2174305008799556e-05,
+      "loss": 2.0692,
+      "step": 12321000
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 3.217358136115228e-05,
+      "loss": 2.0808,
+      "step": 12321500
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 3.217285771350501e-05,
+      "loss": 2.0601,
+      "step": 12322000
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 3.217213406585773e-05,
+      "loss": 2.0905,
+      "step": 12322500
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 3.217141041821045e-05,
+      "loss": 2.0784,
+      "step": 12323000
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 3.2170686770563174e-05,
+      "loss": 2.0639,
+      "step": 12323500
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 3.216996457021119e-05,
+      "loss": 2.0925,
+      "step": 12324000
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 3.216924236985921e-05,
+      "loss": 2.0599,
+      "step": 12324500
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 3.2168518722211934e-05,
+      "loss": 2.0615,
+      "step": 12325000
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 3.2167795074564656e-05,
+      "loss": 2.0804,
+      "step": 12325500
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 3.216707142691738e-05,
+      "loss": 2.0625,
+      "step": 12326000
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 3.21663477792701e-05,
+      "loss": 2.08,
+      "step": 12326500
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 3.216562413162282e-05,
+      "loss": 2.0676,
+      "step": 12327000
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 3.2164900483975545e-05,
+      "loss": 2.0551,
+      "step": 12327500
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 3.216417683632827e-05,
+      "loss": 2.0876,
+      "step": 12328000
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 3.216345318868099e-05,
+      "loss": 2.0757,
+      "step": 12328500
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 3.216272954103372e-05,
+      "loss": 2.0543,
+      "step": 12329000
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 3.216200734068174e-05,
+      "loss": 2.0827,
+      "step": 12329500
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 3.216128369303446e-05,
+      "loss": 2.0829,
+      "step": 12330000
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 3.216056149268248e-05,
+      "loss": 2.0717,
+      "step": 12330500
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 3.21598378450352e-05,
+      "loss": 2.0608,
+      "step": 12331000
+    },
+    {
+      "epoch": 35.69,
+      "learning_rate": 3.215911419738792e-05,
+      "loss": 2.0463,
+      "step": 12331500
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 3.2158390549740645e-05,
+      "loss": 2.0652,
+      "step": 12332000
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 3.215766690209337e-05,
+      "loss": 2.0926,
+      "step": 12332500
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 3.215694325444609e-05,
+      "loss": 2.0827,
+      "step": 12333000
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 3.215621960679881e-05,
+      "loss": 2.0611,
+      "step": 12333500
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 3.215549595915154e-05,
+      "loss": 2.0753,
+      "step": 12334000
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 3.215477375879956e-05,
+      "loss": 2.0613,
+      "step": 12334500
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 3.215405011115228e-05,
+      "loss": 2.0682,
+      "step": 12335000
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 3.2153326463505e-05,
+      "loss": 2.0723,
+      "step": 12335500
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 3.215260281585772e-05,
+      "loss": 2.0738,
+      "step": 12336000
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 3.215188061550574e-05,
+      "loss": 2.0801,
+      "step": 12336500
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 3.215115696785847e-05,
+      "loss": 2.0988,
+      "step": 12337000
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 3.215043332021119e-05,
+      "loss": 2.0863,
+      "step": 12337500
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 3.214970967256391e-05,
+      "loss": 2.0766,
+      "step": 12338000
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 3.2148987472211935e-05,
+      "loss": 2.057,
+      "step": 12338500
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 3.214826382456466e-05,
+      "loss": 2.0669,
+      "step": 12339000
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 3.214754017691738e-05,
+      "loss": 2.0566,
+      "step": 12339500
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 3.2146817976565395e-05,
+      "loss": 2.07,
+      "step": 12340000
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 3.214609432891812e-05,
+      "loss": 2.0699,
+      "step": 12340500
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 3.214537068127084e-05,
+      "loss": 2.0725,
+      "step": 12341000
+    },
+    {
+      "epoch": 35.72,
+      "learning_rate": 3.214464703362357e-05,
+      "loss": 2.0654,
+      "step": 12341500
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 3.214392338597629e-05,
+      "loss": 2.095,
+      "step": 12342000
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 3.214319973832901e-05,
+      "loss": 2.0897,
+      "step": 12342500
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 3.2142476090681735e-05,
+      "loss": 2.0639,
+      "step": 12343000
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 3.214175244303446e-05,
+      "loss": 2.0747,
+      "step": 12343500
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 3.214103024268247e-05,
+      "loss": 2.0985,
+      "step": 12344000
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 3.21403065950352e-05,
+      "loss": 2.0821,
+      "step": 12344500
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 3.2139582947387924e-05,
+      "loss": 2.0772,
+      "step": 12345000
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 3.2138859299740646e-05,
+      "loss": 2.0793,
+      "step": 12345500
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 3.213813565209337e-05,
+      "loss": 2.0757,
+      "step": 12346000
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 3.213741345174139e-05,
+      "loss": 2.0579,
+      "step": 12346500
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 3.2136691251389406e-05,
+      "loss": 2.0839,
+      "step": 12347000
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 3.213596760374213e-05,
+      "loss": 2.0958,
+      "step": 12347500
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 3.213524395609485e-05,
+      "loss": 2.0923,
+      "step": 12348000
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 3.213452030844757e-05,
+      "loss": 2.072,
+      "step": 12348500
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 3.2133796660800295e-05,
+      "loss": 2.0561,
+      "step": 12349000
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 3.213307301315302e-05,
+      "loss": 2.1017,
+      "step": 12349500
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 3.213234936550574e-05,
+      "loss": 2.0697,
+      "step": 12350000
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 3.213162571785847e-05,
+      "loss": 2.0788,
+      "step": 12350500
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 3.213090207021119e-05,
+      "loss": 2.0646,
+      "step": 12351000
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 3.213017842256392e-05,
+      "loss": 2.1037,
+      "step": 12351500
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 3.212945477491664e-05,
+      "loss": 2.0851,
+      "step": 12352000
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 3.2128731127269364e-05,
+      "loss": 2.0882,
+      "step": 12352500
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 3.212800747962209e-05,
+      "loss": 2.1017,
+      "step": 12353000
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 3.212728383197481e-05,
+      "loss": 2.0967,
+      "step": 12353500
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 3.2126561631622824e-05,
+      "loss": 2.0773,
+      "step": 12354000
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 3.212583798397555e-05,
+      "loss": 2.0672,
+      "step": 12354500
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 3.212511433632827e-05,
+      "loss": 2.1051,
+      "step": 12355000
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 3.212439068868099e-05,
+      "loss": 2.0756,
+      "step": 12355500
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 3.212366704103372e-05,
+      "loss": 2.0727,
+      "step": 12356000
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 3.212294339338644e-05,
+      "loss": 2.0646,
+      "step": 12356500
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 3.212222119303446e-05,
+      "loss": 2.0542,
+      "step": 12357000
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 3.212149754538718e-05,
+      "loss": 2.0827,
+      "step": 12357500
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 3.2120775345035196e-05,
+      "loss": 2.0811,
+      "step": 12358000
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 3.212005169738792e-05,
+      "loss": 2.0549,
+      "step": 12358500
+    },
+    {
+      "epoch": 35.77,
+      "learning_rate": 3.211932804974064e-05,
+      "loss": 2.0801,
+      "step": 12359000
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 3.211860440209337e-05,
+      "loss": 2.0765,
+      "step": 12359500
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 3.211788220174139e-05,
+      "loss": 2.0914,
+      "step": 12360000
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 3.2117158554094114e-05,
+      "loss": 2.0775,
+      "step": 12360500
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 3.2116434906446836e-05,
+      "loss": 2.0919,
+      "step": 12361000
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 3.211571125879956e-05,
+      "loss": 2.0749,
+      "step": 12361500
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 3.2114989058447574e-05,
+      "loss": 2.069,
+      "step": 12362000
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 3.2114265410800296e-05,
+      "loss": 2.0814,
+      "step": 12362500
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 3.211354176315302e-05,
+      "loss": 2.0694,
+      "step": 12363000
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 3.211281811550575e-05,
+      "loss": 2.08,
+      "step": 12363500
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 3.211209446785847e-05,
+      "loss": 2.0777,
+      "step": 12364000
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 3.211137082021119e-05,
+      "loss": 2.0801,
+      "step": 12364500
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 3.2110647172563914e-05,
+      "loss": 2.0677,
+      "step": 12365000
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 3.2109923524916636e-05,
+      "loss": 2.0672,
+      "step": 12365500
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 3.210919987726936e-05,
+      "loss": 2.0767,
+      "step": 12366000
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 3.210847622962209e-05,
+      "loss": 2.0748,
+      "step": 12366500
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 3.210775258197481e-05,
+      "loss": 2.0755,
+      "step": 12367000
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 3.210702893432753e-05,
+      "loss": 2.0837,
+      "step": 12367500
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 3.2106305286680254e-05,
+      "loss": 2.0788,
+      "step": 12368000
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 3.2105581639032976e-05,
+      "loss": 2.0958,
+      "step": 12368500
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 3.21048579913857e-05,
+      "loss": 2.0801,
+      "step": 12369000
+    },
+    {
+      "epoch": 35.8,
+      "learning_rate": 3.210413434373842e-05,
+      "loss": 2.074,
+      "step": 12369500
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 3.210341069609114e-05,
+      "loss": 2.068,
+      "step": 12370000
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 3.2102688495739165e-05,
+      "loss": 2.0894,
+      "step": 12370500
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 3.210196484809189e-05,
+      "loss": 2.0845,
+      "step": 12371000
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 3.210124120044461e-05,
+      "loss": 2.0939,
+      "step": 12371500
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 3.210051755279733e-05,
+      "loss": 2.0645,
+      "step": 12372000
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 3.2099793905150054e-05,
+      "loss": 2.0725,
+      "step": 12372500
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 3.2099070257502777e-05,
+      "loss": 2.0645,
+      "step": 12373000
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 3.2098346609855506e-05,
+      "loss": 2.0696,
+      "step": 12373500
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 3.209762440950352e-05,
+      "loss": 2.0794,
+      "step": 12374000
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 3.2096902209151543e-05,
+      "loss": 2.0687,
+      "step": 12374500
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 3.2096178561504266e-05,
+      "loss": 2.0684,
+      "step": 12375000
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 3.209545491385699e-05,
+      "loss": 2.1013,
+      "step": 12375500
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 3.209473126620971e-05,
+      "loss": 2.0805,
+      "step": 12376000
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 3.209400761856243e-05,
+      "loss": 2.0652,
+      "step": 12376500
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 3.2093283970915155e-05,
+      "loss": 2.078,
+      "step": 12377000
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 3.209256032326788e-05,
+      "loss": 2.0767,
+      "step": 12377500
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 3.20918366756206e-05,
+      "loss": 2.0718,
+      "step": 12378000
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 3.209111302797332e-05,
+      "loss": 2.0854,
+      "step": 12378500
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 3.2090390827621344e-05,
+      "loss": 2.0548,
+      "step": 12379000
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 3.2089667179974066e-05,
+      "loss": 2.0751,
+      "step": 12379500
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 3.208894353232679e-05,
+      "loss": 2.0785,
+      "step": 12380000
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 3.208821988467951e-05,
+      "loss": 2.0649,
+      "step": 12380500
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 3.2087497684327526e-05,
+      "loss": 2.06,
+      "step": 12381000
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 3.2086774036680255e-05,
+      "loss": 2.0807,
+      "step": 12381500
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 3.208605038903298e-05,
+      "loss": 2.0539,
+      "step": 12382000
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 3.20853267413857e-05,
+      "loss": 2.0846,
+      "step": 12382500
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 3.2084605988329015e-05,
+      "loss": 2.0738,
+      "step": 12383000
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 3.208388234068174e-05,
+      "loss": 2.0586,
+      "step": 12383500
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 3.208315869303446e-05,
+      "loss": 2.0694,
+      "step": 12384000
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 3.208243504538718e-05,
+      "loss": 2.0706,
+      "step": 12384500
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 3.2081711397739904e-05,
+      "loss": 2.0649,
+      "step": 12385000
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 3.2080987750092626e-05,
+      "loss": 2.0862,
+      "step": 12385500
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 3.208026410244535e-05,
+      "loss": 2.07,
+      "step": 12386000
+    },
+    {
+      "epoch": 35.85,
+      "learning_rate": 3.207954045479807e-05,
+      "loss": 2.0759,
+      "step": 12386500
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 3.20788168071508e-05,
+      "loss": 2.0763,
+      "step": 12387000
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 3.207809605409411e-05,
+      "loss": 2.0803,
+      "step": 12387500
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 3.2077373853742124e-05,
+      "loss": 2.0679,
+      "step": 12388000
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 3.2076650206094846e-05,
+      "loss": 2.0741,
+      "step": 12388500
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 3.2075926558447575e-05,
+      "loss": 2.1108,
+      "step": 12389000
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 3.20752029108003e-05,
+      "loss": 2.0748,
+      "step": 12389500
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 3.207447926315303e-05,
+      "loss": 2.0829,
+      "step": 12390000
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 3.207375706280104e-05,
+      "loss": 2.0733,
+      "step": 12390500
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 3.2073033415153764e-05,
+      "loss": 2.0505,
+      "step": 12391000
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 3.207230976750649e-05,
+      "loss": 2.0956,
+      "step": 12391500
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 3.207158611985921e-05,
+      "loss": 2.0625,
+      "step": 12392000
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 3.207086247221193e-05,
+      "loss": 2.1094,
+      "step": 12392500
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 3.207013882456465e-05,
+      "loss": 2.0948,
+      "step": 12393000
+    },
+    {
+      "epoch": 35.87,
+      "learning_rate": 3.2069415176917376e-05,
+      "loss": 2.0757,
+      "step": 12393500
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 3.20686915292701e-05,
+      "loss": 2.0879,
+      "step": 12394000
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 3.206796932891812e-05,
+      "loss": 2.083,
+      "step": 12394500
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 3.206724568127084e-05,
+      "loss": 2.0923,
+      "step": 12395000
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 3.2066522033623565e-05,
+      "loss": 2.0571,
+      "step": 12395500
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 3.206579838597629e-05,
+      "loss": 2.0994,
+      "step": 12396000
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 3.206507473832901e-05,
+      "loss": 2.0584,
+      "step": 12396500
+    },
+    {
+      "epoch": 35.88,
+      "learning_rate": 3.206435109068174e-05,
+      "loss": 2.0532,
+      "step": 12397000
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 3.206362744303446e-05,
+      "loss": 2.0969,
+      "step": 12397500
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 3.2062905242682476e-05,
+      "loss": 2.0907,
+      "step": 12398000
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 3.20621815950352e-05,
+      "loss": 2.0594,
+      "step": 12398500
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 3.206145794738793e-05,
+      "loss": 2.0663,
+      "step": 12399000
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 3.206073429974065e-05,
+      "loss": 2.076,
+      "step": 12399500
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 3.206001065209337e-05,
+      "loss": 2.0678,
+      "step": 12400000
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 3.2059287004446094e-05,
+      "loss": 2.0838,
+      "step": 12400500
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 3.2058563356798816e-05,
+      "loss": 2.065,
+      "step": 12401000
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 3.205784115644683e-05,
+      "loss": 2.0945,
+      "step": 12401500
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 3.2057117508799554e-05,
+      "loss": 2.0806,
+      "step": 12402000
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 3.2056393861152276e-05,
+      "loss": 2.0867,
+      "step": 12402500
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 3.2055670213505e-05,
+      "loss": 2.1111,
+      "step": 12403000
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 3.205494656585773e-05,
+      "loss": 2.0866,
+      "step": 12403500
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 3.205422291821045e-05,
+      "loss": 2.0932,
+      "step": 12404000
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 3.205349927056318e-05,
+      "loss": 2.0781,
+      "step": 12404500
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 3.20527756229159e-05,
+      "loss": 2.094,
+      "step": 12405000
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 3.205205197526862e-05,
+      "loss": 2.0925,
+      "step": 12405500
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 3.2051328327621345e-05,
+      "loss": 2.0854,
+      "step": 12406000
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 3.205060467997407e-05,
+      "loss": 2.0633,
+      "step": 12406500
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 3.204988247962208e-05,
+      "loss": 2.0532,
+      "step": 12407000
+    },
+    {
+      "epoch": 35.91,
+      "learning_rate": 3.2049158831974805e-05,
+      "loss": 2.0783,
+      "step": 12407500
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 3.204843518432753e-05,
+      "loss": 2.0743,
+      "step": 12408000
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 3.204771298397555e-05,
+      "loss": 2.0825,
+      "step": 12408500
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 3.204698933632827e-05,
+      "loss": 2.0698,
+      "step": 12409000
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 3.2046265688680994e-05,
+      "loss": 2.0744,
+      "step": 12409500
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 3.2045542041033717e-05,
+      "loss": 2.081,
+      "step": 12410000
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 3.204481839338644e-05,
+      "loss": 2.0975,
+      "step": 12410500
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 3.204409474573916e-05,
+      "loss": 2.0849,
+      "step": 12411000
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 3.204337254538718e-05,
+      "loss": 2.085,
+      "step": 12411500
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 3.2042648897739906e-05,
+      "loss": 2.0817,
+      "step": 12412000
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 3.204192525009263e-05,
+      "loss": 2.0916,
+      "step": 12412500
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 3.204120160244535e-05,
+      "loss": 2.0827,
+      "step": 12413000
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 3.204047795479808e-05,
+      "loss": 2.0756,
+      "step": 12413500
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 3.20397543071508e-05,
+      "loss": 2.0696,
+      "step": 12414000
+    },
+    {
+      "epoch": 35.93,
+      "learning_rate": 3.2039030659503524e-05,
+      "loss": 2.0673,
+      "step": 12414500
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 3.203830845915154e-05,
+      "loss": 2.0982,
+      "step": 12415000
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 3.203758481150426e-05,
+      "loss": 2.0743,
+      "step": 12415500
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 3.2036861163856984e-05,
+      "loss": 2.0681,
+      "step": 12416000
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 3.2036137516209706e-05,
+      "loss": 2.084,
+      "step": 12416500
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 3.203541386856243e-05,
+      "loss": 2.0789,
+      "step": 12417000
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 3.203469166821045e-05,
+      "loss": 2.0807,
+      "step": 12417500
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 3.203396802056317e-05,
+      "loss": 2.0525,
+      "step": 12418000
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 3.2033244372915895e-05,
+      "loss": 2.0452,
+      "step": 12418500
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 3.2032520725268624e-05,
+      "loss": 2.0895,
+      "step": 12419000
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 3.2031797077621346e-05,
+      "loss": 2.0621,
+      "step": 12419500
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 3.203107342997407e-05,
+      "loss": 2.0865,
+      "step": 12420000
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 3.203034978232679e-05,
+      "loss": 2.049,
+      "step": 12420500
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 3.202962613467951e-05,
+      "loss": 2.0984,
+      "step": 12421000
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 3.202890393432753e-05,
+      "loss": 2.0844,
+      "step": 12421500
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 3.202818028668025e-05,
+      "loss": 2.0891,
+      "step": 12422000
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 3.202745663903298e-05,
+      "loss": 2.0873,
+      "step": 12422500
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 3.2026734438680995e-05,
+      "loss": 2.0764,
+      "step": 12423000
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 3.202601079103372e-05,
+      "loss": 2.0731,
+      "step": 12423500
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 3.202528714338644e-05,
+      "loss": 2.0592,
+      "step": 12424000
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 3.202456349573916e-05,
+      "loss": 2.0657,
+      "step": 12424500
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 3.2023839848091884e-05,
+      "loss": 2.0917,
+      "step": 12425000
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 3.2023117647739906e-05,
+      "loss": 2.08,
+      "step": 12425500
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 3.202239400009263e-05,
+      "loss": 2.0762,
+      "step": 12426000
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 3.202167035244536e-05,
+      "loss": 2.0708,
+      "step": 12426500
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 3.202094670479808e-05,
+      "loss": 2.0826,
+      "step": 12427000
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 3.20202230571508e-05,
+      "loss": 2.0832,
+      "step": 12427500
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 3.2019499409503524e-05,
+      "loss": 2.0948,
+      "step": 12428000
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 3.201877720915154e-05,
+      "loss": 2.0785,
+      "step": 12428500
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 3.201805356150426e-05,
+      "loss": 2.0773,
+      "step": 12429000
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 3.2017329913856984e-05,
+      "loss": 2.0607,
+      "step": 12429500
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 3.2016606266209707e-05,
+      "loss": 2.1062,
+      "step": 12430000
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 3.201588261856243e-05,
+      "loss": 2.0751,
+      "step": 12430500
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 3.201515897091516e-05,
+      "loss": 2.0714,
+      "step": 12431000
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 3.201443532326788e-05,
+      "loss": 2.0867,
+      "step": 12431500
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 3.20137116756206e-05,
+      "loss": 2.0759,
+      "step": 12432000
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 3.2012988027973325e-05,
+      "loss": 2.0983,
+      "step": 12432500
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 3.201226438032605e-05,
+      "loss": 2.0554,
+      "step": 12433000
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 3.2011540732678776e-05,
+      "loss": 2.0906,
+      "step": 12433500
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 3.201081853232679e-05,
+      "loss": 2.0784,
+      "step": 12434000
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 3.2010094884679514e-05,
+      "loss": 2.068,
+      "step": 12434500
+    },
+    {
+      "epoch": 35.99,
+      "learning_rate": 3.2009371237032236e-05,
+      "loss": 2.06,
+      "step": 12435000
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 3.200864903668026e-05,
+      "loss": 2.0647,
+      "step": 12435500
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 3.200792538903298e-05,
+      "loss": 2.0761,
+      "step": 12436000
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 3.2007203188680996e-05,
+      "loss": 2.0755,
+      "step": 12436500
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.6696317492810138,
+      "eval_accuracy_mlm": 0.6346618176717469,
+      "eval_accuracy_nsp": 0.8570164945713199,
+      "eval_loss": 2.170750141143799,
+      "eval_runtime": 331.4091,
+      "eval_samples_per_second": 1316.759,
+      "eval_steps_per_second": 54.866,
+      "step": 12436992
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 3.200647954103372e-05,
+      "loss": 2.0835,
+      "step": 12437000
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 3.200575589338644e-05,
+      "loss": 2.0695,
+      "step": 12437500
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 3.200503224573916e-05,
+      "loss": 2.0772,
+      "step": 12438000
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 3.2004308598091885e-05,
+      "loss": 2.0507,
+      "step": 12438500
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 3.200358495044461e-05,
+      "loss": 2.0681,
+      "step": 12439000
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 3.200286130279733e-05,
+      "loss": 2.0525,
+      "step": 12439500
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 3.200213765515006e-05,
+      "loss": 2.0395,
+      "step": 12440000
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 3.200141400750278e-05,
+      "loss": 2.0736,
+      "step": 12440500
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 3.200069035985551e-05,
+      "loss": 2.0698,
+      "step": 12441000
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 3.199996671220823e-05,
+      "loss": 2.0789,
+      "step": 12441500
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 3.199924595915154e-05,
+      "loss": 2.0589,
+      "step": 12442000
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 3.199852231150426e-05,
+      "loss": 2.0533,
+      "step": 12442500
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 3.1997798663856985e-05,
+      "loss": 2.0502,
+      "step": 12443000
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 3.199707501620971e-05,
+      "loss": 2.0702,
+      "step": 12443500
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 3.199635136856243e-05,
+      "loss": 2.0126,
+      "step": 12444000
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 3.199562772091516e-05,
+      "loss": 2.0865,
+      "step": 12444500
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 3.199490407326788e-05,
+      "loss": 2.0577,
+      "step": 12445000
+    },
+    {
+      "epoch": 36.02,
+      "learning_rate": 3.19941804256206e-05,
+      "loss": 2.0676,
+      "step": 12445500
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 3.1993456777973325e-05,
+      "loss": 2.0435,
+      "step": 12446000
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 3.199273313032605e-05,
+      "loss": 2.0495,
+      "step": 12446500
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 3.199200948267877e-05,
+      "loss": 2.0509,
+      "step": 12447000
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 3.199128583503149e-05,
+      "loss": 2.0273,
+      "step": 12447500
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 3.1990562187384214e-05,
+      "loss": 2.0517,
+      "step": 12448000
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 3.198983853973694e-05,
+      "loss": 2.0624,
+      "step": 12448500
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 3.1989114892089666e-05,
+      "loss": 2.0471,
+      "step": 12449000
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 3.198839269173768e-05,
+      "loss": 2.0762,
+      "step": 12449500
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 3.198766904409041e-05,
+      "loss": 2.0488,
+      "step": 12450000
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 3.1986946843738426e-05,
+      "loss": 2.0747,
+      "step": 12450500
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 3.198622464338644e-05,
+      "loss": 2.059,
+      "step": 12451000
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 3.1985500995739163e-05,
+      "loss": 2.0505,
+      "step": 12451500
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 3.1984777348091886e-05,
+      "loss": 2.0622,
+      "step": 12452000
+    },
+    {
+      "epoch": 36.04,
+      "learning_rate": 3.198405370044461e-05,
+      "loss": 2.0598,
+      "step": 12452500
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 3.198333005279733e-05,
+      "loss": 2.0451,
+      "step": 12453000
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 3.198260640515006e-05,
+      "loss": 2.0541,
+      "step": 12453500
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 3.198188275750278e-05,
+      "loss": 2.0547,
+      "step": 12454000
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 3.1981159109855504e-05,
+      "loss": 2.0726,
+      "step": 12454500
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 3.1980435462208226e-05,
+      "loss": 2.0486,
+      "step": 12455000
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 3.197971181456095e-05,
+      "loss": 2.0548,
+      "step": 12455500
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 3.197898816691368e-05,
+      "loss": 2.0418,
+      "step": 12456000
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 3.19782645192664e-05,
+      "loss": 2.0776,
+      "step": 12456500
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 3.197754087161912e-05,
+      "loss": 2.0645,
+      "step": 12457000
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 3.1976817223971844e-05,
+      "loss": 2.0543,
+      "step": 12457500
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 3.1976093576324566e-05,
+      "loss": 2.0534,
+      "step": 12458000
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 3.197536992867729e-05,
+      "loss": 2.0317,
+      "step": 12458500
+    },
+    {
+      "epoch": 36.06,
+      "learning_rate": 3.197464772832531e-05,
+      "loss": 2.0799,
+      "step": 12459000
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 3.197392408067803e-05,
+      "loss": 2.0623,
+      "step": 12459500
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 3.197320188032605e-05,
+      "loss": 2.0328,
+      "step": 12460000
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 3.197247823267877e-05,
+      "loss": 2.0482,
+      "step": 12460500
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 3.197175458503149e-05,
+      "loss": 2.0596,
+      "step": 12461000
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 3.1971030937384215e-05,
+      "loss": 2.0633,
+      "step": 12461500
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 3.197030728973694e-05,
+      "loss": 2.0215,
+      "step": 12462000
+    },
+    {
+      "epoch": 36.07,
+      "learning_rate": 3.196958364208966e-05,
+      "loss": 2.0716,
+      "step": 12462500
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 3.196885999444238e-05,
+      "loss": 2.0564,
+      "step": 12463000
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 3.196813779409041e-05,
+      "loss": 2.065,
+      "step": 12463500
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 3.196741414644313e-05,
+      "loss": 2.0426,
+      "step": 12464000
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 3.1966690498795855e-05,
+      "loss": 2.0379,
+      "step": 12464500
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 3.196596685114858e-05,
+      "loss": 2.075,
+      "step": 12465000
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 3.19652432035013e-05,
+      "loss": 2.0741,
+      "step": 12465500
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 3.196451955585402e-05,
+      "loss": 2.0446,
+      "step": 12466000
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 3.196379735550204e-05,
+      "loss": 2.0782,
+      "step": 12466500
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 3.196307370785476e-05,
+      "loss": 2.0721,
+      "step": 12467000
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 3.196235006020749e-05,
+      "loss": 2.09,
+      "step": 12467500
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 3.196162641256021e-05,
+      "loss": 2.0738,
+      "step": 12468000
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 3.196090276491293e-05,
+      "loss": 2.051,
+      "step": 12468500
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 3.1960179117265656e-05,
+      "loss": 2.0632,
+      "step": 12469000
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 3.195945691691367e-05,
+      "loss": 2.0712,
+      "step": 12469500
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 3.195873326926639e-05,
+      "loss": 2.0805,
+      "step": 12470000
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 3.1958009621619116e-05,
+      "loss": 2.0623,
+      "step": 12470500
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 3.1957285973971845e-05,
+      "loss": 2.0721,
+      "step": 12471000
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 3.195656377361986e-05,
+      "loss": 2.0635,
+      "step": 12471500
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 3.195584012597259e-05,
+      "loss": 2.0499,
+      "step": 12472000
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 3.195511647832531e-05,
+      "loss": 2.0377,
+      "step": 12472500
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 3.1954392830678034e-05,
+      "loss": 2.0515,
+      "step": 12473000
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 3.1953669183030756e-05,
+      "loss": 2.0584,
+      "step": 12473500
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 3.195294553538348e-05,
+      "loss": 2.099,
+      "step": 12474000
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 3.19522218877362e-05,
+      "loss": 2.0449,
+      "step": 12474500
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 3.195149824008892e-05,
+      "loss": 2.0751,
+      "step": 12475000
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 3.1950774592441645e-05,
+      "loss": 2.0738,
+      "step": 12475500
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 3.195005239208966e-05,
+      "loss": 2.0558,
+      "step": 12476000
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 3.194932874444239e-05,
+      "loss": 2.0682,
+      "step": 12476500
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 3.194860509679511e-05,
+      "loss": 2.0579,
+      "step": 12477000
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 3.1947881449147834e-05,
+      "loss": 2.0891,
+      "step": 12477500
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 3.194715924879585e-05,
+      "loss": 2.0719,
+      "step": 12478000
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 3.194643560114858e-05,
+      "loss": 2.066,
+      "step": 12478500
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 3.1945713400796594e-05,
+      "loss": 2.0694,
+      "step": 12479000
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 3.1944989753149316e-05,
+      "loss": 2.0415,
+      "step": 12479500
+    },
+    {
+      "epoch": 36.12,
+      "learning_rate": 3.194426610550204e-05,
+      "loss": 2.0602,
+      "step": 12480000
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 3.194354245785476e-05,
+      "loss": 2.0712,
+      "step": 12480500
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 3.194281881020749e-05,
+      "loss": 2.0438,
+      "step": 12481000
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 3.194209516256021e-05,
+      "loss": 2.0441,
+      "step": 12481500
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 3.1941371514912934e-05,
+      "loss": 2.0848,
+      "step": 12482000
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 3.1940647867265656e-05,
+      "loss": 2.0767,
+      "step": 12482500
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 3.193992566691367e-05,
+      "loss": 2.0787,
+      "step": 12483000
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 3.1939202019266394e-05,
+      "loss": 2.0593,
+      "step": 12483500
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 3.1938478371619116e-05,
+      "loss": 2.0529,
+      "step": 12484000
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 3.193775472397184e-05,
+      "loss": 2.0689,
+      "step": 12484500
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 3.193703107632456e-05,
+      "loss": 2.068,
+      "step": 12485000
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 3.193630887597258e-05,
+      "loss": 2.0508,
+      "step": 12485500
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 3.193558522832531e-05,
+      "loss": 2.065,
+      "step": 12486000
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 3.1934861580678034e-05,
+      "loss": 2.0539,
+      "step": 12486500
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 3.193413793303076e-05,
+      "loss": 2.0388,
+      "step": 12487000
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 3.193341428538348e-05,
+      "loss": 2.0539,
+      "step": 12487500
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 3.19326906377362e-05,
+      "loss": 2.0467,
+      "step": 12488000
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 3.193196699008892e-05,
+      "loss": 2.0355,
+      "step": 12488500
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 3.1931243342441646e-05,
+      "loss": 2.0612,
+      "step": 12489000
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 3.193051969479437e-05,
+      "loss": 2.0495,
+      "step": 12489500
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 3.192979604714709e-05,
+      "loss": 2.0295,
+      "step": 12490000
+    },
+    {
+      "epoch": 36.15,
+      "learning_rate": 3.192907239949981e-05,
+      "loss": 2.0481,
+      "step": 12490500
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 3.1928350199147835e-05,
+      "loss": 2.0497,
+      "step": 12491000
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 3.192762655150056e-05,
+      "loss": 2.0675,
+      "step": 12491500
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 3.192690290385328e-05,
+      "loss": 2.0984,
+      "step": 12492000
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 3.1926179256206e-05,
+      "loss": 2.0355,
+      "step": 12492500
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 3.192545560855873e-05,
+      "loss": 2.0512,
+      "step": 12493000
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 3.1924733408206746e-05,
+      "loss": 2.0467,
+      "step": 12493500
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 3.192400976055947e-05,
+      "loss": 2.0507,
+      "step": 12494000
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 3.192328611291219e-05,
+      "loss": 2.0522,
+      "step": 12494500
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 3.192256246526491e-05,
+      "loss": 2.0341,
+      "step": 12495000
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 3.1921840264912935e-05,
+      "loss": 2.078,
+      "step": 12495500
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 3.192111661726566e-05,
+      "loss": 2.0741,
+      "step": 12496000
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 3.192039296961838e-05,
+      "loss": 2.06,
+      "step": 12496500
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 3.19196693219711e-05,
+      "loss": 2.0527,
+      "step": 12497000
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 3.1918945674323824e-05,
+      "loss": 2.0608,
+      "step": 12497500
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 3.1918222026676546e-05,
+      "loss": 2.0688,
+      "step": 12498000
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 3.191749837902927e-05,
+      "loss": 2.0853,
+      "step": 12498500
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 3.1916777625972584e-05,
+      "loss": 2.0431,
+      "step": 12499000
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 3.1916053978325306e-05,
+      "loss": 2.0694,
+      "step": 12499500
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 3.191533033067803e-05,
+      "loss": 2.0751,
+      "step": 12500000
+    },
+    {
+      "epoch": 36.18,
+      "learning_rate": 3.191460668303075e-05,
+      "loss": 2.0665,
+      "step": 12500500
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 3.191388303538348e-05,
+      "loss": 2.0632,
+      "step": 12501000
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 3.19131593877362e-05,
+      "loss": 2.0771,
+      "step": 12501500
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 3.1912435740088924e-05,
+      "loss": 2.0455,
+      "step": 12502000
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 3.191171353973694e-05,
+      "loss": 2.0751,
+      "step": 12502500
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 3.191098989208967e-05,
+      "loss": 2.0598,
+      "step": 12503000
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 3.191026624444239e-05,
+      "loss": 2.0535,
+      "step": 12503500
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 3.190954259679511e-05,
+      "loss": 2.0529,
+      "step": 12504000
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 3.1908818949147835e-05,
+      "loss": 2.0678,
+      "step": 12504500
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 3.190809530150056e-05,
+      "loss": 2.0713,
+      "step": 12505000
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 3.190737165385328e-05,
+      "loss": 2.0488,
+      "step": 12505500
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 3.1906648006206e-05,
+      "loss": 2.0585,
+      "step": 12506000
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 3.190592580585402e-05,
+      "loss": 2.0721,
+      "step": 12506500
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 3.190520215820674e-05,
+      "loss": 2.0435,
+      "step": 12507000
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 3.190447851055947e-05,
+      "loss": 2.05,
+      "step": 12507500
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 3.190375486291219e-05,
+      "loss": 2.0783,
+      "step": 12508000
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 3.190303121526492e-05,
+      "loss": 2.0287,
+      "step": 12508500
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 3.190231046220823e-05,
+      "loss": 2.0612,
+      "step": 12509000
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 3.190158681456095e-05,
+      "loss": 2.0493,
+      "step": 12509500
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 3.1900863166913673e-05,
+      "loss": 2.0914,
+      "step": 12510000
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 3.1900139519266396e-05,
+      "loss": 2.0358,
+      "step": 12510500
+    },
+    {
+      "epoch": 36.21,
+      "learning_rate": 3.189941587161912e-05,
+      "loss": 2.0578,
+      "step": 12511000
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 3.189869222397184e-05,
+      "loss": 2.056,
+      "step": 12511500
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 3.189796857632457e-05,
+      "loss": 2.067,
+      "step": 12512000
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 3.189724492867729e-05,
+      "loss": 2.08,
+      "step": 12512500
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 3.189652272832531e-05,
+      "loss": 2.0568,
+      "step": 12513000
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 3.189580052797332e-05,
+      "loss": 2.0492,
+      "step": 12513500
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 3.1895076880326045e-05,
+      "loss": 2.0488,
+      "step": 12514000
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 3.189435323267877e-05,
+      "loss": 2.069,
+      "step": 12514500
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 3.189362958503149e-05,
+      "loss": 2.0782,
+      "step": 12515000
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 3.189290593738422e-05,
+      "loss": 2.0644,
+      "step": 12515500
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 3.1892183737032234e-05,
+      "loss": 2.0795,
+      "step": 12516000
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 3.189146008938496e-05,
+      "loss": 2.0847,
+      "step": 12516500
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 3.189073788903298e-05,
+      "loss": 2.0753,
+      "step": 12517000
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 3.18900142413857e-05,
+      "loss": 2.0615,
+      "step": 12517500
+    },
+    {
+      "epoch": 36.23,
+      "learning_rate": 3.188929059373842e-05,
+      "loss": 2.0735,
+      "step": 12518000
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 3.1888566946091145e-05,
+      "loss": 2.0576,
+      "step": 12518500
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 3.188784329844387e-05,
+      "loss": 2.0499,
+      "step": 12519000
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 3.1887119650796596e-05,
+      "loss": 2.0407,
+      "step": 12519500
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 3.188639600314932e-05,
+      "loss": 2.0644,
+      "step": 12520000
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 3.188567235550204e-05,
+      "loss": 2.0722,
+      "step": 12520500
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 3.188494870785476e-05,
+      "loss": 2.0538,
+      "step": 12521000
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 3.1884225060207485e-05,
+      "loss": 2.0547,
+      "step": 12521500
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 3.188350141256021e-05,
+      "loss": 2.0647,
+      "step": 12522000
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 3.188277776491293e-05,
+      "loss": 2.064,
+      "step": 12522500
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 3.188205411726565e-05,
+      "loss": 2.0364,
+      "step": 12523000
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 3.188133046961838e-05,
+      "loss": 2.0598,
+      "step": 12523500
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 3.18806068219711e-05,
+      "loss": 2.0673,
+      "step": 12524000
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 3.1879883174323825e-05,
+      "loss": 2.0524,
+      "step": 12524500
+    },
+    {
+      "epoch": 36.25,
+      "learning_rate": 3.187915952667655e-05,
+      "loss": 2.0609,
+      "step": 12525000
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 3.187843732632457e-05,
+      "loss": 2.0602,
+      "step": 12525500
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 3.187771367867729e-05,
+      "loss": 2.0545,
+      "step": 12526000
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 3.187699147832531e-05,
+      "loss": 2.0826,
+      "step": 12526500
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 3.187626783067803e-05,
+      "loss": 2.0555,
+      "step": 12527000
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 3.187554418303075e-05,
+      "loss": 2.0643,
+      "step": 12527500
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 3.1874820535383474e-05,
+      "loss": 2.0474,
+      "step": 12528000
+    },
+    {
+      "epoch": 36.26,
+      "learning_rate": 3.18740968877362e-05,
+      "loss": 2.0573,
+      "step": 12528500
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 3.187337324008892e-05,
+      "loss": 2.0887,
+      "step": 12529000
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 3.187264959244165e-05,
+      "loss": 2.0556,
+      "step": 12529500
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 3.187192594479437e-05,
+      "loss": 2.0809,
+      "step": 12530000
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 3.187120229714709e-05,
+      "loss": 2.0732,
+      "step": 12530500
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 3.1870480096795115e-05,
+      "loss": 2.0548,
+      "step": 12531000
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 3.186975789644313e-05,
+      "loss": 2.081,
+      "step": 12531500
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 3.186903424879585e-05,
+      "loss": 2.054,
+      "step": 12532000
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 3.1868310601148575e-05,
+      "loss": 2.0834,
+      "step": 12532500
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 3.18675869535013e-05,
+      "loss": 2.0576,
+      "step": 12533000
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 3.186686330585402e-05,
+      "loss": 2.0484,
+      "step": 12533500
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 3.186614110550204e-05,
+      "loss": 2.0445,
+      "step": 12534000
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 3.1865417457854764e-05,
+      "loss": 2.0685,
+      "step": 12534500
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 3.1864693810207486e-05,
+      "loss": 2.0674,
+      "step": 12535000
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 3.186397016256021e-05,
+      "loss": 2.078,
+      "step": 12535500
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 3.186324651491293e-05,
+      "loss": 2.0461,
+      "step": 12536000
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 3.1862524314560946e-05,
+      "loss": 2.082,
+      "step": 12536500
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 3.186180066691367e-05,
+      "loss": 2.0629,
+      "step": 12537000
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 3.18610770192664e-05,
+      "loss": 2.0693,
+      "step": 12537500
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 3.186035337161912e-05,
+      "loss": 2.0524,
+      "step": 12538000
+    },
+    {
+      "epoch": 36.29,
+      "learning_rate": 3.185962972397185e-05,
+      "loss": 2.0835,
+      "step": 12538500
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 3.1858907523619864e-05,
+      "loss": 2.0698,
+      "step": 12539000
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 3.1858183875972586e-05,
+      "loss": 2.0731,
+      "step": 12539500
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 3.185746022832531e-05,
+      "loss": 2.0697,
+      "step": 12540000
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 3.185673658067803e-05,
+      "loss": 2.0506,
+      "step": 12540500
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 3.185601293303075e-05,
+      "loss": 2.0571,
+      "step": 12541000
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 3.1855289285383475e-05,
+      "loss": 2.0708,
+      "step": 12541500
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 3.18545656377362e-05,
+      "loss": 2.0597,
+      "step": 12542000
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 3.185384343738422e-05,
+      "loss": 2.0473,
+      "step": 12542500
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 3.185311978973694e-05,
+      "loss": 2.0741,
+      "step": 12543000
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 3.1852396142089664e-05,
+      "loss": 2.0739,
+      "step": 12543500
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 3.1851672494442387e-05,
+      "loss": 2.058,
+      "step": 12544000
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 3.185094884679511e-05,
+      "loss": 2.0827,
+      "step": 12544500
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 3.1850226646443124e-05,
+      "loss": 2.0693,
+      "step": 12545000
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 3.1849502998795847e-05,
+      "loss": 2.0824,
+      "step": 12545500
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 3.1848779351148576e-05,
+      "loss": 2.0731,
+      "step": 12546000
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 3.18480557035013e-05,
+      "loss": 2.0742,
+      "step": 12546500
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 3.184733350314932e-05,
+      "loss": 2.0738,
+      "step": 12547000
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 3.184660985550204e-05,
+      "loss": 2.0807,
+      "step": 12547500
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 3.1845886207854765e-05,
+      "loss": 2.0725,
+      "step": 12548000
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 3.184516256020749e-05,
+      "loss": 2.0557,
+      "step": 12548500
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 3.18444403598555e-05,
+      "loss": 2.0655,
+      "step": 12549000
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 3.1843716712208225e-05,
+      "loss": 2.0949,
+      "step": 12549500
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 3.184299306456095e-05,
+      "loss": 2.0792,
+      "step": 12550000
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 3.1842269416913676e-05,
+      "loss": 2.0602,
+      "step": 12550500
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 3.18415457692664e-05,
+      "loss": 2.0678,
+      "step": 12551000
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 3.184082212161912e-05,
+      "loss": 2.0731,
+      "step": 12551500
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 3.184009847397184e-05,
+      "loss": 2.0799,
+      "step": 12552000
+    },
+    {
+      "epoch": 36.33,
+      "learning_rate": 3.1839374826324565e-05,
+      "loss": 2.0524,
+      "step": 12552500
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 3.183865117867729e-05,
+      "loss": 2.0685,
+      "step": 12553000
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 3.1837927531030016e-05,
+      "loss": 2.0711,
+      "step": 12553500
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 3.183720533067803e-05,
+      "loss": 2.0388,
+      "step": 12554000
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 3.1836481683030754e-05,
+      "loss": 2.0613,
+      "step": 12554500
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 3.1835758035383476e-05,
+      "loss": 2.0601,
+      "step": 12555000
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 3.18350343877362e-05,
+      "loss": 2.0549,
+      "step": 12555500
+    },
+    {
+      "epoch": 36.34,
+      "learning_rate": 3.183431218738422e-05,
+      "loss": 2.067,
+      "step": 12556000
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 3.1833589987032236e-05,
+      "loss": 2.0604,
+      "step": 12556500
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 3.183286633938496e-05,
+      "loss": 2.0924,
+      "step": 12557000
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 3.1832144139032974e-05,
+      "loss": 2.0479,
+      "step": 12557500
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 3.1831420491385696e-05,
+      "loss": 2.0861,
+      "step": 12558000
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 3.1830696843738425e-05,
+      "loss": 2.0665,
+      "step": 12558500
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 3.182997319609115e-05,
+      "loss": 2.07,
+      "step": 12559000
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 3.182924954844387e-05,
+      "loss": 2.0512,
+      "step": 12559500
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 3.182852590079659e-05,
+      "loss": 2.0528,
+      "step": 12560000
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 3.1827802253149314e-05,
+      "loss": 2.0628,
+      "step": 12560500
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 3.182707860550204e-05,
+      "loss": 2.0573,
+      "step": 12561000
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 3.1826354957854765e-05,
+      "loss": 2.0932,
+      "step": 12561500
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 3.182563131020749e-05,
+      "loss": 2.067,
+      "step": 12562000
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 3.182490766256021e-05,
+      "loss": 2.0702,
+      "step": 12562500
+    },
+    {
+      "epoch": 36.36,
+      "learning_rate": 3.182418401491293e-05,
+      "loss": 2.0744,
+      "step": 12563000
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 3.1823460367265654e-05,
+      "loss": 2.0666,
+      "step": 12563500
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 3.182273816691368e-05,
+      "loss": 2.0565,
+      "step": 12564000
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 3.18220145192664e-05,
+      "loss": 2.0499,
+      "step": 12564500
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 3.182129087161912e-05,
+      "loss": 2.0701,
+      "step": 12565000
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 3.182056722397184e-05,
+      "loss": 2.0617,
+      "step": 12565500
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 3.1819843576324566e-05,
+      "loss": 2.0708,
+      "step": 12566000
+    },
+    {
+      "epoch": 36.37,
+      "learning_rate": 3.181911992867729e-05,
+      "loss": 2.0718,
+      "step": 12566500
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 3.18183977283253e-05,
+      "loss": 2.0884,
+      "step": 12567000
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 3.1817674080678026e-05,
+      "loss": 2.0645,
+      "step": 12567500
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 3.181695043303075e-05,
+      "loss": 2.0532,
+      "step": 12568000
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 3.181622678538348e-05,
+      "loss": 2.0727,
+      "step": 12568500
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 3.18155031377362e-05,
+      "loss": 2.0658,
+      "step": 12569000
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 3.181477949008893e-05,
+      "loss": 2.053,
+      "step": 12569500
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 3.181405584244165e-05,
+      "loss": 2.0572,
+      "step": 12570000
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 3.181333219479437e-05,
+      "loss": 2.0814,
+      "step": 12570500
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 3.1812608547147095e-05,
+      "loss": 2.035,
+      "step": 12571000
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 3.181188634679511e-05,
+      "loss": 2.0591,
+      "step": 12571500
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 3.1811164146443126e-05,
+      "loss": 2.0613,
+      "step": 12572000
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 3.181044049879585e-05,
+      "loss": 2.058,
+      "step": 12572500
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 3.180971685114858e-05,
+      "loss": 2.0695,
+      "step": 12573000
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 3.18089932035013e-05,
+      "loss": 2.1201,
+      "step": 12573500
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 3.180826955585402e-05,
+      "loss": 2.0596,
+      "step": 12574000
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 3.180754735550204e-05,
+      "loss": 2.0749,
+      "step": 12574500
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 3.180682370785476e-05,
+      "loss": 2.0636,
+      "step": 12575000
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 3.180610006020748e-05,
+      "loss": 2.0635,
+      "step": 12575500
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 3.180537641256021e-05,
+      "loss": 2.0504,
+      "step": 12576000
+    },
+    {
+      "epoch": 36.4,
+      "learning_rate": 3.180465276491293e-05,
+      "loss": 2.0673,
+      "step": 12576500
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 3.1803929117265655e-05,
+      "loss": 2.0859,
+      "step": 12577000
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 3.180320546961838e-05,
+      "loss": 2.0704,
+      "step": 12577500
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 3.18024832692664e-05,
+      "loss": 2.0792,
+      "step": 12578000
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 3.180175962161912e-05,
+      "loss": 2.0506,
+      "step": 12578500
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 3.180103742126714e-05,
+      "loss": 2.0946,
+      "step": 12579000
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 3.180031377361986e-05,
+      "loss": 2.0607,
+      "step": 12579500
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 3.179959012597258e-05,
+      "loss": 2.0874,
+      "step": 12580000
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 3.1798866478325304e-05,
+      "loss": 2.042,
+      "step": 12580500
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 3.1798142830678026e-05,
+      "loss": 2.0639,
+      "step": 12581000
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 3.1797419183030755e-05,
+      "loss": 2.0736,
+      "step": 12581500
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 3.179669553538348e-05,
+      "loss": 2.0681,
+      "step": 12582000
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 3.17959718877362e-05,
+      "loss": 2.0664,
+      "step": 12582500
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 3.179524824008892e-05,
+      "loss": 2.0801,
+      "step": 12583000
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 3.179452459244165e-05,
+      "loss": 2.0522,
+      "step": 12583500
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 3.179380094479437e-05,
+      "loss": 2.0806,
+      "step": 12584000
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 3.1793077297147096e-05,
+      "loss": 2.0561,
+      "step": 12584500
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 3.179235364949982e-05,
+      "loss": 2.0504,
+      "step": 12585000
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 3.1791631449147833e-05,
+      "loss": 2.0542,
+      "step": 12585500
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 3.1790907801500556e-05,
+      "loss": 2.0479,
+      "step": 12586000
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 3.179018415385328e-05,
+      "loss": 2.0858,
+      "step": 12586500
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 3.178946050620601e-05,
+      "loss": 2.0499,
+      "step": 12587000
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 3.178873685855873e-05,
+      "loss": 2.0841,
+      "step": 12587500
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 3.178801321091145e-05,
+      "loss": 2.0682,
+      "step": 12588000
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 3.1787289563264174e-05,
+      "loss": 2.0732,
+      "step": 12588500
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 3.1786565915616896e-05,
+      "loss": 2.0696,
+      "step": 12589000
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 3.178584371526491e-05,
+      "loss": 2.073,
+      "step": 12589500
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 3.1785120067617634e-05,
+      "loss": 2.0417,
+      "step": 12590000
+    },
+    {
+      "epoch": 36.44,
+      "learning_rate": 3.178439641997036e-05,
+      "loss": 2.0772,
+      "step": 12590500
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 3.1783672772323085e-05,
+      "loss": 2.0733,
+      "step": 12591000
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 3.178295057197111e-05,
+      "loss": 2.0526,
+      "step": 12591500
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 3.178222692432383e-05,
+      "loss": 2.0773,
+      "step": 12592000
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 3.178150327667655e-05,
+      "loss": 2.0613,
+      "step": 12592500
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 3.1780779629029274e-05,
+      "loss": 2.0955,
+      "step": 12593000
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 3.178005742867729e-05,
+      "loss": 2.0782,
+      "step": 12593500
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 3.177933378103001e-05,
+      "loss": 2.0618,
+      "step": 12594000
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 3.177861158067803e-05,
+      "loss": 2.051,
+      "step": 12594500
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 3.177788938032605e-05,
+      "loss": 2.051,
+      "step": 12595000
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 3.177716573267877e-05,
+      "loss": 2.0485,
+      "step": 12595500
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 3.1776442085031494e-05,
+      "loss": 2.0578,
+      "step": 12596000
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 3.1775718437384216e-05,
+      "loss": 2.0655,
+      "step": 12596500
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 3.177499478973694e-05,
+      "loss": 2.0603,
+      "step": 12597000
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 3.177427114208966e-05,
+      "loss": 2.0762,
+      "step": 12597500
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 3.177354749444238e-05,
+      "loss": 2.0693,
+      "step": 12598000
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 3.177282384679511e-05,
+      "loss": 2.0849,
+      "step": 12598500
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 3.1772100199147834e-05,
+      "loss": 2.0501,
+      "step": 12599000
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 3.1771376551500556e-05,
+      "loss": 2.0663,
+      "step": 12599500
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 3.177065290385328e-05,
+      "loss": 2.0592,
+      "step": 12600000
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 3.176992925620601e-05,
+      "loss": 2.0939,
+      "step": 12600500
+    },
+    {
+      "epoch": 36.47,
+      "learning_rate": 3.176920560855873e-05,
+      "loss": 2.063,
+      "step": 12601000
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 3.176848196091145e-05,
+      "loss": 2.056,
+      "step": 12601500
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 3.176775976055947e-05,
+      "loss": 2.0547,
+      "step": 12602000
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 3.176703611291219e-05,
+      "loss": 2.0505,
+      "step": 12602500
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 3.176631246526491e-05,
+      "loss": 2.0755,
+      "step": 12603000
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 3.1765588817617634e-05,
+      "loss": 2.0679,
+      "step": 12603500
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 3.176486661726566e-05,
+      "loss": 2.0531,
+      "step": 12604000
+    },
+    {
+      "epoch": 36.48,
+      "learning_rate": 3.176414296961838e-05,
+      "loss": 2.0535,
+      "step": 12604500
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 3.17634193219711e-05,
+      "loss": 2.0584,
+      "step": 12605000
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 3.1762695674323823e-05,
+      "loss": 2.0739,
+      "step": 12605500
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 3.176197202667655e-05,
+      "loss": 2.059,
+      "step": 12606000
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 3.1761248379029275e-05,
+      "loss": 2.0676,
+      "step": 12606500
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 3.1760524731382e-05,
+      "loss": 2.0699,
+      "step": 12607000
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 3.175980253103001e-05,
+      "loss": 2.0665,
+      "step": 12607500
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 3.1759078883382735e-05,
+      "loss": 2.0838,
+      "step": 12608000
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 3.175835523573546e-05,
+      "loss": 2.0707,
+      "step": 12608500
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 3.175763158808818e-05,
+      "loss": 2.0458,
+      "step": 12609000
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 3.175690794044091e-05,
+      "loss": 2.0703,
+      "step": 12609500
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 3.1756185740088924e-05,
+      "loss": 2.0733,
+      "step": 12610000
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 3.1755462092441646e-05,
+      "loss": 2.0484,
+      "step": 12610500
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 3.175473844479437e-05,
+      "loss": 2.0336,
+      "step": 12611000
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 3.175401479714709e-05,
+      "loss": 2.062,
+      "step": 12611500
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 3.175329114949981e-05,
+      "loss": 2.0709,
+      "step": 12612000
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 3.1752567501852535e-05,
+      "loss": 2.0859,
+      "step": 12612500
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 3.1751843854205264e-05,
+      "loss": 2.0608,
+      "step": 12613000
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 3.1751120206557986e-05,
+      "loss": 2.0506,
+      "step": 12613500
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 3.175039655891071e-05,
+      "loss": 2.0451,
+      "step": 12614000
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 3.174967291126343e-05,
+      "loss": 2.0358,
+      "step": 12614500
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 3.174895071091145e-05,
+      "loss": 2.0668,
+      "step": 12615000
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 3.1748227063264175e-05,
+      "loss": 2.0542,
+      "step": 12615500
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 3.174750486291219e-05,
+      "loss": 2.0573,
+      "step": 12616000
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 3.174678121526491e-05,
+      "loss": 2.0668,
+      "step": 12616500
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 3.1746057567617635e-05,
+      "loss": 2.0519,
+      "step": 12617000
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 3.174533391997036e-05,
+      "loss": 2.0695,
+      "step": 12617500
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 3.1744610272323086e-05,
+      "loss": 2.0605,
+      "step": 12618000
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 3.174388662467581e-05,
+      "loss": 2.0551,
+      "step": 12618500
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 3.174316297702853e-05,
+      "loss": 2.0726,
+      "step": 12619000
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 3.174243932938125e-05,
+      "loss": 2.0758,
+      "step": 12619500
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 3.174171712902927e-05,
+      "loss": 2.0716,
+      "step": 12620000
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 3.1740993481382e-05,
+      "loss": 2.0688,
+      "step": 12620500
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 3.174026983373472e-05,
+      "loss": 2.052,
+      "step": 12621000
+    },
+    {
+      "epoch": 36.53,
+      "learning_rate": 3.173954618608744e-05,
+      "loss": 2.05,
+      "step": 12621500
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 3.173882398573546e-05,
+      "loss": 2.0581,
+      "step": 12622000
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 3.173810033808819e-05,
+      "loss": 2.0521,
+      "step": 12622500
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 3.173737669044091e-05,
+      "loss": 2.0562,
+      "step": 12623000
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 3.173665304279363e-05,
+      "loss": 2.0772,
+      "step": 12623500
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 3.1735929395146353e-05,
+      "loss": 2.0666,
+      "step": 12624000
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 3.1735205747499076e-05,
+      "loss": 2.0754,
+      "step": 12624500
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 3.17344820998518e-05,
+      "loss": 2.0739,
+      "step": 12625000
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 3.1733759899499813e-05,
+      "loss": 2.0664,
+      "step": 12625500
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 3.1733036251852536e-05,
+      "loss": 2.0523,
+      "step": 12626000
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 3.173231260420526e-05,
+      "loss": 2.0627,
+      "step": 12626500
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 3.173159040385328e-05,
+      "loss": 2.0682,
+      "step": 12627000
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 3.1730866756206e-05,
+      "loss": 2.0834,
+      "step": 12627500
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 3.173014310855873e-05,
+      "loss": 2.0681,
+      "step": 12628000
+    },
+    {
+      "epoch": 36.55,
+      "learning_rate": 3.1729419460911454e-05,
+      "loss": 2.0664,
+      "step": 12628500
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 3.172869726055947e-05,
+      "loss": 2.0528,
+      "step": 12629000
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 3.172797361291219e-05,
+      "loss": 2.0527,
+      "step": 12629500
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 3.1727249965264914e-05,
+      "loss": 2.0463,
+      "step": 12630000
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 3.1726526317617636e-05,
+      "loss": 2.0894,
+      "step": 12630500
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 3.172580411726566e-05,
+      "loss": 2.0656,
+      "step": 12631000
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 3.172508046961838e-05,
+      "loss": 2.0634,
+      "step": 12631500
+    },
+    {
+      "epoch": 36.56,
+      "learning_rate": 3.17243568219711e-05,
+      "loss": 2.0582,
+      "step": 12632000
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 3.1723633174323825e-05,
+      "loss": 2.0577,
+      "step": 12632500
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 3.172291097397184e-05,
+      "loss": 2.0799,
+      "step": 12633000
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 3.172218732632456e-05,
+      "loss": 2.0579,
+      "step": 12633500
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 3.1721463678677285e-05,
+      "loss": 2.0683,
+      "step": 12634000
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 3.1720740031030014e-05,
+      "loss": 2.0944,
+      "step": 12634500
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 3.1720016383382736e-05,
+      "loss": 2.0434,
+      "step": 12635000
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 3.171929273573546e-05,
+      "loss": 2.056,
+      "step": 12635500
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 3.171856908808819e-05,
+      "loss": 2.0588,
+      "step": 12636000
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 3.17178468877362e-05,
+      "loss": 2.0727,
+      "step": 12636500
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 3.1717123240088925e-05,
+      "loss": 2.0743,
+      "step": 12637000
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 3.171639959244165e-05,
+      "loss": 2.0527,
+      "step": 12637500
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 3.171567594479437e-05,
+      "loss": 2.0651,
+      "step": 12638000
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 3.171495229714709e-05,
+      "loss": 2.0901,
+      "step": 12638500
+    },
+    {
+      "epoch": 36.58,
+      "learning_rate": 3.1714230096795114e-05,
+      "loss": 2.0665,
+      "step": 12639000
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 3.1713506449147837e-05,
+      "loss": 2.0571,
+      "step": 12639500
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 3.171278280150056e-05,
+      "loss": 2.0418,
+      "step": 12640000
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 3.171205915385328e-05,
+      "loss": 2.0524,
+      "step": 12640500
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 3.1711335506206e-05,
+      "loss": 2.0844,
+      "step": 12641000
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 3.1710611858558726e-05,
+      "loss": 2.0399,
+      "step": 12641500
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 3.170988821091145e-05,
+      "loss": 2.0801,
+      "step": 12642000
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 3.170916456326417e-05,
+      "loss": 2.0806,
+      "step": 12642500
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 3.17084409156169e-05,
+      "loss": 2.0713,
+      "step": 12643000
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 3.170771726796962e-05,
+      "loss": 2.0473,
+      "step": 12643500
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 3.170699506761764e-05,
+      "loss": 2.0453,
+      "step": 12644000
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 3.1706271419970366e-05,
+      "loss": 2.063,
+      "step": 12644500
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 3.170554777232309e-05,
+      "loss": 2.0704,
+      "step": 12645000
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 3.170482412467581e-05,
+      "loss": 2.0672,
+      "step": 12645500
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 3.1704101924323826e-05,
+      "loss": 2.0567,
+      "step": 12646000
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 3.170337972397184e-05,
+      "loss": 2.0655,
+      "step": 12646500
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 3.1702656076324564e-05,
+      "loss": 2.0744,
+      "step": 12647000
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 3.1701932428677286e-05,
+      "loss": 2.0839,
+      "step": 12647500
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 3.1701208781030015e-05,
+      "loss": 2.0484,
+      "step": 12648000
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 3.170048658067803e-05,
+      "loss": 2.0819,
+      "step": 12648500
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 3.169976293303075e-05,
+      "loss": 2.0769,
+      "step": 12649000
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 3.1699039285383475e-05,
+      "loss": 2.0761,
+      "step": 12649500
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 3.16983156377362e-05,
+      "loss": 2.0794,
+      "step": 12650000
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 3.169759199008892e-05,
+      "loss": 2.0877,
+      "step": 12650500
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 3.169686834244165e-05,
+      "loss": 2.0623,
+      "step": 12651000
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 3.169614469479437e-05,
+      "loss": 2.0667,
+      "step": 12651500
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 3.169542104714709e-05,
+      "loss": 2.0622,
+      "step": 12652000
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 3.1694697399499815e-05,
+      "loss": 2.0746,
+      "step": 12652500
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 3.169397375185254e-05,
+      "loss": 2.0573,
+      "step": 12653000
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 3.1693250104205266e-05,
+      "loss": 2.0526,
+      "step": 12653500
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 3.169252645655799e-05,
+      "loss": 2.0641,
+      "step": 12654000
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 3.169180280891071e-05,
+      "loss": 2.0551,
+      "step": 12654500
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 3.169107916126343e-05,
+      "loss": 2.0704,
+      "step": 12655000
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 3.1690355513616155e-05,
+      "loss": 2.067,
+      "step": 12655500
+    },
+    {
+      "epoch": 36.63,
+      "learning_rate": 3.168963186596888e-05,
+      "loss": 2.0509,
+      "step": 12656000
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 3.168890966561689e-05,
+      "loss": 2.0597,
+      "step": 12656500
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 3.1688186017969615e-05,
+      "loss": 2.0444,
+      "step": 12657000
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 3.168746237032234e-05,
+      "loss": 2.0595,
+      "step": 12657500
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 3.1686738722675066e-05,
+      "loss": 2.0894,
+      "step": 12658000
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 3.168601507502779e-05,
+      "loss": 2.0656,
+      "step": 12658500
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 3.168529287467581e-05,
+      "loss": 2.0933,
+      "step": 12659000
+    },
+    {
+      "epoch": 36.64,
+      "learning_rate": 3.1684570674323827e-05,
+      "loss": 2.0664,
+      "step": 12659500
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 3.168384702667655e-05,
+      "loss": 2.0482,
+      "step": 12660000
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 3.1683124826324564e-05,
+      "loss": 2.0539,
+      "step": 12660500
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 3.168240262597259e-05,
+      "loss": 2.0585,
+      "step": 12661000
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 3.168167897832531e-05,
+      "loss": 2.0746,
+      "step": 12661500
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 3.168095533067803e-05,
+      "loss": 2.0833,
+      "step": 12662000
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 3.1680231683030753e-05,
+      "loss": 2.0802,
+      "step": 12662500
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 3.1679508035383476e-05,
+      "loss": 2.0608,
+      "step": 12663000
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 3.16787843877362e-05,
+      "loss": 2.0587,
+      "step": 12663500
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 3.167806074008892e-05,
+      "loss": 2.0797,
+      "step": 12664000
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 3.167733709244164e-05,
+      "loss": 2.082,
+      "step": 12664500
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 3.1676613444794365e-05,
+      "loss": 2.0864,
+      "step": 12665000
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 3.167589124444239e-05,
+      "loss": 2.0772,
+      "step": 12665500
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 3.1675167596795116e-05,
+      "loss": 2.067,
+      "step": 12666000
+    },
+    {
+      "epoch": 36.66,
+      "learning_rate": 3.167444394914784e-05,
+      "loss": 2.0584,
+      "step": 12666500
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 3.167372030150056e-05,
+      "loss": 2.0672,
+      "step": 12667000
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 3.167299665385328e-05,
+      "loss": 2.0786,
+      "step": 12667500
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 3.1672273006206005e-05,
+      "loss": 2.0602,
+      "step": 12668000
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 3.167155080585402e-05,
+      "loss": 2.0625,
+      "step": 12668500
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 3.167082860550204e-05,
+      "loss": 2.0763,
+      "step": 12669000
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 3.1670104957854765e-05,
+      "loss": 2.1129,
+      "step": 12669500
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 3.166938131020749e-05,
+      "loss": 2.074,
+      "step": 12670000
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 3.166865766256021e-05,
+      "loss": 2.0682,
+      "step": 12670500
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 3.166793401491293e-05,
+      "loss": 2.0589,
+      "step": 12671000
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 3.1667210367265654e-05,
+      "loss": 2.0605,
+      "step": 12671500
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 3.1666486719618376e-05,
+      "loss": 2.0575,
+      "step": 12672000
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 3.16657630719711e-05,
+      "loss": 2.0631,
+      "step": 12672500
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 3.166503942432382e-05,
+      "loss": 2.067,
+      "step": 12673000
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 3.166431722397184e-05,
+      "loss": 2.0627,
+      "step": 12673500
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 3.1663593576324565e-05,
+      "loss": 2.0724,
+      "step": 12674000
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 3.1662869928677294e-05,
+      "loss": 2.0721,
+      "step": 12674500
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 3.1662146281030016e-05,
+      "loss": 2.0515,
+      "step": 12675000
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 3.166142263338274e-05,
+      "loss": 2.074,
+      "step": 12675500
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 3.166069898573546e-05,
+      "loss": 2.0665,
+      "step": 12676000
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 3.165997533808818e-05,
+      "loss": 2.0734,
+      "step": 12676500
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 3.1659251690440905e-05,
+      "loss": 2.0627,
+      "step": 12677000
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 3.165852804279363e-05,
+      "loss": 2.0897,
+      "step": 12677500
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 3.165780584244164e-05,
+      "loss": 2.0763,
+      "step": 12678000
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 3.1657082194794365e-05,
+      "loss": 2.0686,
+      "step": 12678500
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 3.1656358547147094e-05,
+      "loss": 2.0831,
+      "step": 12679000
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 3.165563634679511e-05,
+      "loss": 2.0902,
+      "step": 12679500
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 3.165491269914783e-05,
+      "loss": 2.0512,
+      "step": 12680000
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 3.1654189051500554e-05,
+      "loss": 2.068,
+      "step": 12680500
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 3.1653465403853283e-05,
+      "loss": 2.0822,
+      "step": 12681000
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 3.1652741756206006e-05,
+      "loss": 2.0591,
+      "step": 12681500
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 3.165201810855873e-05,
+      "loss": 2.0488,
+      "step": 12682000
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 3.165129446091145e-05,
+      "loss": 2.0934,
+      "step": 12682500
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 3.165057081326417e-05,
+      "loss": 2.0733,
+      "step": 12683000
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 3.1649848612912195e-05,
+      "loss": 2.0722,
+      "step": 12683500
+    },
+    {
+      "epoch": 36.71,
+      "learning_rate": 3.164912496526492e-05,
+      "loss": 2.0554,
+      "step": 12684000
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 3.164840131761764e-05,
+      "loss": 2.0537,
+      "step": 12684500
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 3.164767766997036e-05,
+      "loss": 2.0735,
+      "step": 12685000
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 3.1646954022323084e-05,
+      "loss": 2.0683,
+      "step": 12685500
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 3.1646230374675806e-05,
+      "loss": 2.0938,
+      "step": 12686000
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 3.164550817432382e-05,
+      "loss": 2.0483,
+      "step": 12686500
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 3.1644784526676544e-05,
+      "loss": 2.0652,
+      "step": 12687000
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 3.1644060879029266e-05,
+      "loss": 2.0851,
+      "step": 12687500
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 3.1643337231381995e-05,
+      "loss": 2.0602,
+      "step": 12688000
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 3.164261647832531e-05,
+      "loss": 2.0715,
+      "step": 12688500
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 3.164189283067803e-05,
+      "loss": 2.0778,
+      "step": 12689000
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 3.164117063032605e-05,
+      "loss": 2.0604,
+      "step": 12689500
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 3.164044698267877e-05,
+      "loss": 2.055,
+      "step": 12690000
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 3.163972333503149e-05,
+      "loss": 2.0502,
+      "step": 12690500
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 3.163899968738422e-05,
+      "loss": 2.0592,
+      "step": 12691000
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 3.1638276039736944e-05,
+      "loss": 2.0644,
+      "step": 12691500
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 3.1637552392089666e-05,
+      "loss": 2.098,
+      "step": 12692000
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 3.163682874444239e-05,
+      "loss": 2.0938,
+      "step": 12692500
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 3.163610509679511e-05,
+      "loss": 2.0596,
+      "step": 12693000
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 3.1635382896443126e-05,
+      "loss": 2.096,
+      "step": 12693500
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 3.163465924879585e-05,
+      "loss": 2.0509,
+      "step": 12694000
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 3.163393560114857e-05,
+      "loss": 2.0958,
+      "step": 12694500
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 3.163321195350129e-05,
+      "loss": 2.0719,
+      "step": 12695000
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 3.163248830585402e-05,
+      "loss": 2.0543,
+      "step": 12695500
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 3.1631766105502044e-05,
+      "loss": 2.0775,
+      "step": 12696000
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 3.163104390515006e-05,
+      "loss": 2.0577,
+      "step": 12696500
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 3.163032025750278e-05,
+      "loss": 2.0538,
+      "step": 12697000
+    },
+    {
+      "epoch": 36.75,
+      "learning_rate": 3.1629596609855504e-05,
+      "loss": 2.0926,
+      "step": 12697500
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 3.1628872962208227e-05,
+      "loss": 2.0751,
+      "step": 12698000
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 3.162815076185625e-05,
+      "loss": 2.0637,
+      "step": 12698500
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 3.162742711420897e-05,
+      "loss": 2.0587,
+      "step": 12699000
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 3.1626703466561693e-05,
+      "loss": 2.0477,
+      "step": 12699500
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 3.1625979818914416e-05,
+      "loss": 2.1098,
+      "step": 12700000
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 3.162525617126714e-05,
+      "loss": 2.0774,
+      "step": 12700500
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 3.1624533970915153e-05,
+      "loss": 2.0933,
+      "step": 12701000
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 3.1623810323267876e-05,
+      "loss": 2.0576,
+      "step": 12701500
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 3.16230866756206e-05,
+      "loss": 2.0706,
+      "step": 12702000
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 3.162236302797332e-05,
+      "loss": 2.0675,
+      "step": 12702500
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 3.162163938032605e-05,
+      "loss": 2.0766,
+      "step": 12703000
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 3.162091573267877e-05,
+      "loss": 2.052,
+      "step": 12703500
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 3.1620193532326794e-05,
+      "loss": 2.0706,
+      "step": 12704000
+    },
+    {
+      "epoch": 36.77,
+      "learning_rate": 3.1619469884679516e-05,
+      "loss": 2.0435,
+      "step": 12704500
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 3.161874623703224e-05,
+      "loss": 2.088,
+      "step": 12705000
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 3.161802258938496e-05,
+      "loss": 2.0638,
+      "step": 12705500
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 3.161729894173768e-05,
+      "loss": 2.0671,
+      "step": 12706000
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 3.1616575294090405e-05,
+      "loss": 2.0562,
+      "step": 12706500
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 3.161585164644313e-05,
+      "loss": 2.0838,
+      "step": 12707000
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 3.161512799879585e-05,
+      "loss": 2.0816,
+      "step": 12707500
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 3.161440435114857e-05,
+      "loss": 2.0669,
+      "step": 12708000
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 3.1613680703501294e-05,
+      "loss": 2.0566,
+      "step": 12708500
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 3.161295705585402e-05,
+      "loss": 2.0692,
+      "step": 12709000
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 3.1612233408206745e-05,
+      "loss": 2.048,
+      "step": 12709500
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 3.161151120785476e-05,
+      "loss": 2.0985,
+      "step": 12710000
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 3.1610789007502776e-05,
+      "loss": 2.0501,
+      "step": 12710500
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 3.1610065359855505e-05,
+      "loss": 2.0688,
+      "step": 12711000
+    },
+    {
+      "epoch": 36.79,
+      "learning_rate": 3.160934315950352e-05,
+      "loss": 2.0719,
+      "step": 12711500
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 3.160861951185625e-05,
+      "loss": 2.0785,
+      "step": 12712000
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 3.160789586420897e-05,
+      "loss": 2.0934,
+      "step": 12712500
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 3.1607172216561694e-05,
+      "loss": 2.0412,
+      "step": 12713000
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 3.1606448568914416e-05,
+      "loss": 2.0508,
+      "step": 12713500
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 3.160572492126714e-05,
+      "loss": 2.0788,
+      "step": 12714000
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 3.160500127361986e-05,
+      "loss": 2.0706,
+      "step": 12714500
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 3.160427762597258e-05,
+      "loss": 2.0825,
+      "step": 12715000
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 3.1603553978325305e-05,
+      "loss": 2.0819,
+      "step": 12715500
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 3.160283033067803e-05,
+      "loss": 2.0483,
+      "step": 12716000
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 3.160210668303075e-05,
+      "loss": 2.0716,
+      "step": 12716500
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 3.160138303538347e-05,
+      "loss": 2.0915,
+      "step": 12717000
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 3.16006593877362e-05,
+      "loss": 2.0612,
+      "step": 12717500
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 3.159993718738422e-05,
+      "loss": 2.0707,
+      "step": 12718000
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 3.1599213539736946e-05,
+      "loss": 2.0501,
+      "step": 12718500
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 3.159848989208967e-05,
+      "loss": 2.066,
+      "step": 12719000
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 3.159776624444239e-05,
+      "loss": 2.0651,
+      "step": 12719500
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 3.159704259679511e-05,
+      "loss": 2.0551,
+      "step": 12720000
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 3.1596318949147835e-05,
+      "loss": 2.106,
+      "step": 12720500
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 3.159559530150056e-05,
+      "loss": 2.0619,
+      "step": 12721000
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 3.159487310114857e-05,
+      "loss": 2.1086,
+      "step": 12721500
+    },
+    {
+      "epoch": 36.82,
+      "learning_rate": 3.1594150900796595e-05,
+      "loss": 2.0686,
+      "step": 12722000
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 3.159342725314932e-05,
+      "loss": 2.0675,
+      "step": 12722500
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 3.159270360550204e-05,
+      "loss": 2.0759,
+      "step": 12723000
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 3.159197995785476e-05,
+      "loss": 2.0939,
+      "step": 12723500
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 3.1591256310207484e-05,
+      "loss": 2.0779,
+      "step": 12724000
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 3.1590532662560206e-05,
+      "loss": 2.0706,
+      "step": 12724500
+    },
+    {
+      "epoch": 36.83,
+      "learning_rate": 3.158980901491293e-05,
+      "loss": 2.0687,
+      "step": 12725000
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 3.158908536726565e-05,
+      "loss": 2.0627,
+      "step": 12725500
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 3.158836171961838e-05,
+      "loss": 2.0562,
+      "step": 12726000
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 3.15876395192664e-05,
+      "loss": 2.0708,
+      "step": 12726500
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 3.1586915871619124e-05,
+      "loss": 2.08,
+      "step": 12727000
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 3.1586192223971846e-05,
+      "loss": 2.0824,
+      "step": 12727500
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 3.158546857632457e-05,
+      "loss": 2.0634,
+      "step": 12728000
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 3.1584746375972584e-05,
+      "loss": 2.0767,
+      "step": 12728500
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 3.15840241756206e-05,
+      "loss": 2.0821,
+      "step": 12729000
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 3.158330052797333e-05,
+      "loss": 2.0752,
+      "step": 12729500
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 3.1582578327621344e-05,
+      "loss": 2.0675,
+      "step": 12730000
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 3.1581854679974066e-05,
+      "loss": 2.0592,
+      "step": 12730500
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 3.158113103232679e-05,
+      "loss": 2.0781,
+      "step": 12731000
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 3.158040738467951e-05,
+      "loss": 2.055,
+      "step": 12731500
+    },
+    {
+      "epoch": 36.85,
+      "learning_rate": 3.157968373703223e-05,
+      "loss": 2.0775,
+      "step": 12732000
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 3.1578960089384955e-05,
+      "loss": 2.0791,
+      "step": 12732500
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 3.157823644173768e-05,
+      "loss": 2.0744,
+      "step": 12733000
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 3.1577512794090406e-05,
+      "loss": 2.0881,
+      "step": 12733500
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 3.157678914644313e-05,
+      "loss": 2.0961,
+      "step": 12734000
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 3.157606549879585e-05,
+      "loss": 2.0858,
+      "step": 12734500
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 3.157534185114858e-05,
+      "loss": 2.0725,
+      "step": 12735000
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 3.15746182035013e-05,
+      "loss": 2.073,
+      "step": 12735500
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 3.1573894555854024e-05,
+      "loss": 2.0817,
+      "step": 12736000
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 3.157317090820675e-05,
+      "loss": 2.0584,
+      "step": 12736500
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 3.157244726055947e-05,
+      "loss": 2.0604,
+      "step": 12737000
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 3.1571725060207484e-05,
+      "loss": 2.0869,
+      "step": 12737500
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 3.157100141256021e-05,
+      "loss": 2.0552,
+      "step": 12738000
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 3.157027776491293e-05,
+      "loss": 2.0627,
+      "step": 12738500
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 3.156955411726565e-05,
+      "loss": 2.037,
+      "step": 12739000
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 3.156883046961838e-05,
+      "loss": 2.0663,
+      "step": 12739500
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 3.1568108269266396e-05,
+      "loss": 2.0965,
+      "step": 12740000
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 3.156738462161912e-05,
+      "loss": 2.0379,
+      "step": 12740500
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 3.156666097397185e-05,
+      "loss": 2.0607,
+      "step": 12741000
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 3.156593732632457e-05,
+      "loss": 2.0599,
+      "step": 12741500
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 3.156521367867729e-05,
+      "loss": 2.0616,
+      "step": 12742000
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 3.1564490031030014e-05,
+      "loss": 2.0908,
+      "step": 12742500
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 3.1563766383382736e-05,
+      "loss": 2.054,
+      "step": 12743000
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 3.156304273573546e-05,
+      "loss": 2.0677,
+      "step": 12743500
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 3.156231908808818e-05,
+      "loss": 2.0786,
+      "step": 12744000
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 3.15615954404409e-05,
+      "loss": 2.1018,
+      "step": 12744500
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 3.1560871792793625e-05,
+      "loss": 2.0509,
+      "step": 12745000
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 3.1560148145146354e-05,
+      "loss": 2.0691,
+      "step": 12745500
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 3.1559424497499076e-05,
+      "loss": 2.0662,
+      "step": 12746000
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 3.155870229714709e-05,
+      "loss": 2.0704,
+      "step": 12746500
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 3.1557978649499814e-05,
+      "loss": 2.0782,
+      "step": 12747000
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 3.1557255001852536e-05,
+      "loss": 2.069,
+      "step": 12747500
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 3.1556531354205265e-05,
+      "loss": 2.11,
+      "step": 12748000
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 3.155580915385328e-05,
+      "loss": 2.0636,
+      "step": 12748500
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 3.1555085506206e-05,
+      "loss": 2.0516,
+      "step": 12749000
+    },
+    {
+      "epoch": 36.9,
+      "learning_rate": 3.155436185855873e-05,
+      "loss": 2.0641,
+      "step": 12749500
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 3.1553638210911454e-05,
+      "loss": 2.0642,
+      "step": 12750000
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 3.155291601055947e-05,
+      "loss": 2.0694,
+      "step": 12750500
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 3.155219236291219e-05,
+      "loss": 2.049,
+      "step": 12751000
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 3.1551468715264914e-05,
+      "loss": 2.0668,
+      "step": 12751500
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 3.1550745067617636e-05,
+      "loss": 2.0779,
+      "step": 12752000
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 3.155002141997036e-05,
+      "loss": 2.0885,
+      "step": 12752500
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 3.154929921961838e-05,
+      "loss": 2.0349,
+      "step": 12753000
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 3.15485755719711e-05,
+      "loss": 2.0686,
+      "step": 12753500
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 3.154785337161912e-05,
+      "loss": 2.0688,
+      "step": 12754000
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 3.1547131171267134e-05,
+      "loss": 2.0775,
+      "step": 12754500
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 3.1546407523619857e-05,
+      "loss": 2.0653,
+      "step": 12755000
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 3.154568387597258e-05,
+      "loss": 2.0572,
+      "step": 12755500
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 3.154496022832531e-05,
+      "loss": 2.0532,
+      "step": 12756000
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 3.154423658067803e-05,
+      "loss": 2.0693,
+      "step": 12756500
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 3.154351293303075e-05,
+      "loss": 2.0566,
+      "step": 12757000
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 3.154278928538348e-05,
+      "loss": 2.068,
+      "step": 12757500
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 3.1542065637736203e-05,
+      "loss": 2.0328,
+      "step": 12758000
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 3.1541341990088926e-05,
+      "loss": 2.0955,
+      "step": 12758500
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 3.154061834244165e-05,
+      "loss": 2.0796,
+      "step": 12759000
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 3.153989469479437e-05,
+      "loss": 2.0867,
+      "step": 12759500
+    },
+    {
+      "epoch": 36.93,
+      "learning_rate": 3.153917104714709e-05,
+      "loss": 2.0888,
+      "step": 12760000
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 3.1538447399499815e-05,
+      "loss": 2.0613,
+      "step": 12760500
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 3.153772375185254e-05,
+      "loss": 2.0667,
+      "step": 12761000
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 3.153700010420526e-05,
+      "loss": 2.0723,
+      "step": 12761500
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 3.153627645655798e-05,
+      "loss": 2.0533,
+      "step": 12762000
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 3.1535552808910704e-05,
+      "loss": 2.0645,
+      "step": 12762500
+    },
+    {
+      "epoch": 36.94,
+      "learning_rate": 3.153482916126343e-05,
+      "loss": 2.0566,
+      "step": 12763000
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 3.1534106960911455e-05,
+      "loss": 2.0647,
+      "step": 12763500
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 3.1533386207854764e-05,
+      "loss": 2.0874,
+      "step": 12764000
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 3.1532662560207486e-05,
+      "loss": 2.0741,
+      "step": 12764500
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 3.153193891256021e-05,
+      "loss": 2.0715,
+      "step": 12765000
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 3.153121526491293e-05,
+      "loss": 2.0716,
+      "step": 12765500
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 3.153049161726566e-05,
+      "loss": 2.0678,
+      "step": 12766000
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 3.152976796961838e-05,
+      "loss": 2.0927,
+      "step": 12766500
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 3.1529044321971104e-05,
+      "loss": 2.0689,
+      "step": 12767000
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 3.1528320674323826e-05,
+      "loss": 2.0645,
+      "step": 12767500
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 3.1527599921267135e-05,
+      "loss": 2.0771,
+      "step": 12768000
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 3.152687627361986e-05,
+      "loss": 2.032,
+      "step": 12768500
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 3.152615262597258e-05,
+      "loss": 2.0926,
+      "step": 12769000
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 3.152542897832531e-05,
+      "loss": 2.0896,
+      "step": 12769500
+    },
+    {
+      "epoch": 36.96,
+      "learning_rate": 3.152470533067803e-05,
+      "loss": 2.0903,
+      "step": 12770000
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 3.152398168303075e-05,
+      "loss": 2.0673,
+      "step": 12770500
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 3.1523259482678775e-05,
+      "loss": 2.0557,
+      "step": 12771000
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 3.15225358350315e-05,
+      "loss": 2.072,
+      "step": 12771500
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 3.152181363467951e-05,
+      "loss": 2.0714,
+      "step": 12772000
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 3.1521089987032235e-05,
+      "loss": 2.0792,
+      "step": 12772500
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 3.152036633938496e-05,
+      "loss": 2.0703,
+      "step": 12773000
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 3.151964269173768e-05,
+      "loss": 2.0501,
+      "step": 12773500
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 3.151891904409041e-05,
+      "loss": 2.0818,
+      "step": 12774000
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 3.151819539644313e-05,
+      "loss": 2.108,
+      "step": 12774500
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 3.151747319609115e-05,
+      "loss": 2.0914,
+      "step": 12775000
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 3.151674954844387e-05,
+      "loss": 2.0618,
+      "step": 12775500
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 3.151602590079659e-05,
+      "loss": 2.0905,
+      "step": 12776000
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 3.151530225314931e-05,
+      "loss": 2.0843,
+      "step": 12776500
+    },
+    {
+      "epoch": 36.98,
+      "learning_rate": 3.1514578605502036e-05,
+      "loss": 2.0828,
+      "step": 12777000
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 3.151385495785476e-05,
+      "loss": 2.0532,
+      "step": 12777500
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 3.151313131020748e-05,
+      "loss": 2.0684,
+      "step": 12778000
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 3.151240766256021e-05,
+      "loss": 2.0754,
+      "step": 12778500
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 3.151168401491293e-05,
+      "loss": 2.0595,
+      "step": 12779000
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 3.151096036726566e-05,
+      "loss": 2.0687,
+      "step": 12779500
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 3.151023671961838e-05,
+      "loss": 2.0687,
+      "step": 12780000
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 3.15095145192664e-05,
+      "loss": 2.0732,
+      "step": 12780500
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 3.150879087161912e-05,
+      "loss": 2.1001,
+      "step": 12781000
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 3.1508068671267136e-05,
+      "loss": 2.0559,
+      "step": 12781500
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 3.150734502361986e-05,
+      "loss": 2.0639,
+      "step": 12782000
+    },
+    {
+      "epoch": 37.0,
+      "eval_accuracy": 0.6695824526693803,
+      "eval_accuracy_mlm": 0.6348121816080246,
+      "eval_accuracy_nsp": 0.8559096762957565,
+      "eval_loss": 2.1710124015808105,
+      "eval_runtime": 331.6656,
+      "eval_samples_per_second": 1315.741,
+      "eval_steps_per_second": 54.823,
+      "step": 12782464
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 3.150662137597258e-05,
+      "loss": 2.0733,
+      "step": 12782500
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 3.150589772832531e-05,
+      "loss": 2.0443,
+      "step": 12783000
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 3.150517408067803e-05,
+      "loss": 2.0477,
+      "step": 12783500
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 3.1504450433030754e-05,
+      "loss": 2.0526,
+      "step": 12784000
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 3.1503726785383476e-05,
+      "loss": 2.0618,
+      "step": 12784500
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 3.15030031377362e-05,
+      "loss": 2.0671,
+      "step": 12785000
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 3.150227949008892e-05,
+      "loss": 2.0591,
+      "step": 12785500
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 3.150155584244165e-05,
+      "loss": 2.0296,
+      "step": 12786000
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 3.150083219479437e-05,
+      "loss": 2.075,
+      "step": 12786500
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 3.1500108547147094e-05,
+      "loss": 2.0501,
+      "step": 12787000
+    },
+    {
+      "epoch": 37.01,
+      "learning_rate": 3.1499384899499816e-05,
+      "loss": 2.0698,
+      "step": 12787500
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 3.149866125185254e-05,
+      "loss": 2.057,
+      "step": 12788000
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 3.149793905150056e-05,
+      "loss": 2.0676,
+      "step": 12788500
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 3.149721540385328e-05,
+      "loss": 2.0529,
+      "step": 12789000
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 3.1496491756206005e-05,
+      "loss": 2.0617,
+      "step": 12789500
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 3.149576810855873e-05,
+      "loss": 2.0428,
+      "step": 12790000
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 3.149504590820674e-05,
+      "loss": 2.0492,
+      "step": 12790500
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 3.1494322260559465e-05,
+      "loss": 2.0331,
+      "step": 12791000
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 3.149360006020749e-05,
+      "loss": 2.0371,
+      "step": 12791500
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 3.149287641256021e-05,
+      "loss": 2.0314,
+      "step": 12792000
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 3.149215276491293e-05,
+      "loss": 2.0745,
+      "step": 12792500
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 3.1491429117265654e-05,
+      "loss": 2.0652,
+      "step": 12793000
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 3.149070546961838e-05,
+      "loss": 2.0504,
+      "step": 12793500
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 3.14899832692664e-05,
+      "loss": 2.0866,
+      "step": 12794000
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 3.148925962161912e-05,
+      "loss": 2.0443,
+      "step": 12794500
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 3.148853597397184e-05,
+      "loss": 2.0399,
+      "step": 12795000
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 3.1487812326324566e-05,
+      "loss": 2.0583,
+      "step": 12795500
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 3.148708867867729e-05,
+      "loss": 2.0441,
+      "step": 12796000
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 3.148636503103001e-05,
+      "loss": 2.061,
+      "step": 12796500
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 3.148564138338274e-05,
+      "loss": 2.0219,
+      "step": 12797000
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 3.1484919183030755e-05,
+      "loss": 2.0553,
+      "step": 12797500
+    },
+    {
+      "epoch": 37.04,
+      "learning_rate": 3.148419553538348e-05,
+      "loss": 2.0748,
+      "step": 12798000
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 3.14834718877362e-05,
+      "loss": 2.0494,
+      "step": 12798500
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 3.148274824008892e-05,
+      "loss": 2.0528,
+      "step": 12799000
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 3.1482024592441644e-05,
+      "loss": 2.0329,
+      "step": 12799500
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 3.1481300944794366e-05,
+      "loss": 2.0571,
+      "step": 12800000
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 3.148057729714709e-05,
+      "loss": 2.0598,
+      "step": 12800500
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 3.147985364949982e-05,
+      "loss": 2.0591,
+      "step": 12801000
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 3.147913000185254e-05,
+      "loss": 2.0435,
+      "step": 12801500
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 3.147840780150056e-05,
+      "loss": 2.0701,
+      "step": 12802000
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 3.1477684153853284e-05,
+      "loss": 2.0745,
+      "step": 12802500
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 3.1476960506206006e-05,
+      "loss": 2.0636,
+      "step": 12803000
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 3.147623685855873e-05,
+      "loss": 2.0603,
+      "step": 12803500
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 3.147551321091145e-05,
+      "loss": 2.0392,
+      "step": 12804000
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 3.1474791010559466e-05,
+      "loss": 2.0594,
+      "step": 12804500
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 3.147406736291219e-05,
+      "loss": 2.0611,
+      "step": 12805000
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 3.147334371526491e-05,
+      "loss": 2.0382,
+      "step": 12805500
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 3.147262006761764e-05,
+      "loss": 2.027,
+      "step": 12806000
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 3.1471897867265655e-05,
+      "loss": 2.0509,
+      "step": 12806500
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 3.147117421961838e-05,
+      "loss": 2.0347,
+      "step": 12807000
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 3.147045201926639e-05,
+      "loss": 2.05,
+      "step": 12807500
+    },
+    {
+      "epoch": 37.07,
+      "learning_rate": 3.1469728371619115e-05,
+      "loss": 2.0598,
+      "step": 12808000
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 3.1469004723971844e-05,
+      "loss": 2.056,
+      "step": 12808500
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 3.146828252361986e-05,
+      "loss": 2.0344,
+      "step": 12809000
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 3.146755887597259e-05,
+      "loss": 2.0352,
+      "step": 12809500
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 3.146683522832531e-05,
+      "loss": 2.0678,
+      "step": 12810000
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 3.146611158067803e-05,
+      "loss": 2.0444,
+      "step": 12810500
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 3.1465387933030755e-05,
+      "loss": 2.0615,
+      "step": 12811000
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 3.146466428538348e-05,
+      "loss": 2.0288,
+      "step": 12811500
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 3.14639406377362e-05,
+      "loss": 2.0388,
+      "step": 12812000
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 3.146321699008892e-05,
+      "loss": 2.0561,
+      "step": 12812500
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 3.1462493342441644e-05,
+      "loss": 2.0349,
+      "step": 12813000
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 3.1461769694794367e-05,
+      "loss": 2.0255,
+      "step": 12813500
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 3.146104604714709e-05,
+      "loss": 2.0367,
+      "step": 12814000
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 3.146032239949981e-05,
+      "loss": 2.0356,
+      "step": 12814500
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 3.145960019914783e-05,
+      "loss": 2.0546,
+      "step": 12815000
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 3.1458876551500556e-05,
+      "loss": 2.068,
+      "step": 12815500
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 3.1458152903853285e-05,
+      "loss": 2.0404,
+      "step": 12816000
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 3.145742925620601e-05,
+      "loss": 2.0273,
+      "step": 12816500
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 3.145670560855873e-05,
+      "loss": 2.0524,
+      "step": 12817000
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 3.1455983408206745e-05,
+      "loss": 2.0588,
+      "step": 12817500
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 3.145525976055947e-05,
+      "loss": 2.0481,
+      "step": 12818000
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 3.145453611291219e-05,
+      "loss": 2.0685,
+      "step": 12818500
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 3.145381246526491e-05,
+      "loss": 2.0495,
+      "step": 12819000
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 3.145308881761764e-05,
+      "loss": 2.0588,
+      "step": 12819500
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 3.145236516997036e-05,
+      "loss": 2.0462,
+      "step": 12820000
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 3.1451641522323085e-05,
+      "loss": 2.0389,
+      "step": 12820500
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 3.145091787467581e-05,
+      "loss": 2.0441,
+      "step": 12821000
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 3.145019422702853e-05,
+      "loss": 2.036,
+      "step": 12821500
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 3.1449472026676545e-05,
+      "loss": 2.0632,
+      "step": 12822000
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 3.144874837902927e-05,
+      "loss": 2.0653,
+      "step": 12822500
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 3.1448024731381996e-05,
+      "loss": 2.044,
+      "step": 12823000
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 3.144730108373472e-05,
+      "loss": 2.049,
+      "step": 12823500
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 3.144657743608744e-05,
+      "loss": 2.0516,
+      "step": 12824000
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 3.144585378844016e-05,
+      "loss": 2.0379,
+      "step": 12824500
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 3.144513014079289e-05,
+      "loss": 2.0465,
+      "step": 12825000
+    },
+    {
+      "epoch": 37.12,
+      "learning_rate": 3.1444406493145614e-05,
+      "loss": 2.0307,
+      "step": 12825500
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 3.1443682845498336e-05,
+      "loss": 2.0547,
+      "step": 12826000
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 3.1442962092441645e-05,
+      "loss": 2.0525,
+      "step": 12826500
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 3.144223844479437e-05,
+      "loss": 2.0422,
+      "step": 12827000
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 3.144151479714709e-05,
+      "loss": 2.0376,
+      "step": 12827500
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 3.144079259679511e-05,
+      "loss": 2.0835,
+      "step": 12828000
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 3.1440068949147834e-05,
+      "loss": 2.0259,
+      "step": 12828500
+    },
+    {
+      "epoch": 37.13,
+      "learning_rate": 3.143934674879585e-05,
+      "loss": 2.0629,
+      "step": 12829000
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 3.143862310114857e-05,
+      "loss": 2.053,
+      "step": 12829500
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 3.143790090079659e-05,
+      "loss": 2.0567,
+      "step": 12830000
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 3.1437177253149317e-05,
+      "loss": 2.0595,
+      "step": 12830500
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 3.143645360550204e-05,
+      "loss": 2.0538,
+      "step": 12831000
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 3.143572995785477e-05,
+      "loss": 2.0499,
+      "step": 12831500
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 3.143500631020749e-05,
+      "loss": 2.0701,
+      "step": 12832000
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 3.143428266256021e-05,
+      "loss": 2.0489,
+      "step": 12832500
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 3.1433559014912934e-05,
+      "loss": 2.0404,
+      "step": 12833000
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 3.143283536726566e-05,
+      "loss": 2.0546,
+      "step": 12833500
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 3.143211171961838e-05,
+      "loss": 2.0572,
+      "step": 12834000
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 3.14313880719711e-05,
+      "loss": 2.0383,
+      "step": 12834500
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 3.1430664424323823e-05,
+      "loss": 2.0599,
+      "step": 12835000
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 3.1429940776676546e-05,
+      "loss": 2.0561,
+      "step": 12835500
+    },
+    {
+      "epoch": 37.15,
+      "learning_rate": 3.142921712902927e-05,
+      "loss": 2.0625,
+      "step": 12836000
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 3.142849348138199e-05,
+      "loss": 2.0657,
+      "step": 12836500
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 3.142776983373472e-05,
+      "loss": 2.0629,
+      "step": 12837000
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 3.142704618608744e-05,
+      "loss": 2.0451,
+      "step": 12837500
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 3.142632253844017e-05,
+      "loss": 2.0775,
+      "step": 12838000
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 3.1425600338088186e-05,
+      "loss": 2.0539,
+      "step": 12838500
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 3.142487669044091e-05,
+      "loss": 2.0889,
+      "step": 12839000
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 3.142415304279363e-05,
+      "loss": 2.0352,
+      "step": 12839500
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 3.142342939514635e-05,
+      "loss": 2.0453,
+      "step": 12840000
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 3.1422705747499075e-05,
+      "loss": 2.061,
+      "step": 12840500
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 3.14219820998518e-05,
+      "loss": 2.0602,
+      "step": 12841000
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 3.142125845220452e-05,
+      "loss": 2.0656,
+      "step": 12841500
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 3.142053480455724e-05,
+      "loss": 2.0397,
+      "step": 12842000
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 3.1419812604205264e-05,
+      "loss": 2.0388,
+      "step": 12842500
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 3.141909040385328e-05,
+      "loss": 2.0363,
+      "step": 12843000
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 3.1418368203501295e-05,
+      "loss": 2.054,
+      "step": 12843500
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 3.141764600314932e-05,
+      "loss": 2.0775,
+      "step": 12844000
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 3.141692235550204e-05,
+      "loss": 2.0386,
+      "step": 12844500
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 3.141619870785476e-05,
+      "loss": 2.0678,
+      "step": 12845000
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 3.1415475060207484e-05,
+      "loss": 2.041,
+      "step": 12845500
+    },
+    {
+      "epoch": 37.18,
+      "learning_rate": 3.141475141256021e-05,
+      "loss": 2.0508,
+      "step": 12846000
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 3.1414027764912935e-05,
+      "loss": 2.0469,
+      "step": 12846500
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 3.141330411726566e-05,
+      "loss": 2.0586,
+      "step": 12847000
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 3.141258046961838e-05,
+      "loss": 2.0609,
+      "step": 12847500
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 3.14118568219711e-05,
+      "loss": 2.0562,
+      "step": 12848000
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 3.1411133174323824e-05,
+      "loss": 2.0672,
+      "step": 12848500
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 3.1410409526676546e-05,
+      "loss": 2.0222,
+      "step": 12849000
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 3.140968587902927e-05,
+      "loss": 2.0456,
+      "step": 12849500
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 3.1408965125972584e-05,
+      "loss": 2.0156,
+      "step": 12850000
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 3.1408241478325307e-05,
+      "loss": 2.0696,
+      "step": 12850500
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 3.140751783067803e-05,
+      "loss": 2.0446,
+      "step": 12851000
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 3.140679418303075e-05,
+      "loss": 2.0562,
+      "step": 12851500
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 3.140607053538347e-05,
+      "loss": 2.0444,
+      "step": 12852000
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 3.1405346887736195e-05,
+      "loss": 2.0644,
+      "step": 12852500
+    },
+    {
+      "epoch": 37.2,
+      "learning_rate": 3.140462324008892e-05,
+      "loss": 2.0397,
+      "step": 12853000
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 3.140389959244165e-05,
+      "loss": 2.0498,
+      "step": 12853500
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 3.140317594479437e-05,
+      "loss": 2.0677,
+      "step": 12854000
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 3.14024522971471e-05,
+      "loss": 2.0474,
+      "step": 12854500
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 3.1401730096795114e-05,
+      "loss": 2.0337,
+      "step": 12855000
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 3.1401006449147836e-05,
+      "loss": 2.0331,
+      "step": 12855500
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 3.140028424879585e-05,
+      "loss": 2.0735,
+      "step": 12856000
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 3.139956204844387e-05,
+      "loss": 2.0488,
+      "step": 12856500
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 3.1398838400796596e-05,
+      "loss": 2.0594,
+      "step": 12857000
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 3.139811475314932e-05,
+      "loss": 2.0403,
+      "step": 12857500
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 3.139739110550204e-05,
+      "loss": 2.053,
+      "step": 12858000
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 3.139666745785476e-05,
+      "loss": 2.0542,
+      "step": 12858500
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 3.1395943810207485e-05,
+      "loss": 2.0581,
+      "step": 12859000
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 3.13952216098555e-05,
+      "loss": 2.027,
+      "step": 12859500
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 3.139449796220822e-05,
+      "loss": 2.0397,
+      "step": 12860000
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 3.1393775761856245e-05,
+      "loss": 2.0382,
+      "step": 12860500
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 3.139305211420897e-05,
+      "loss": 2.0236,
+      "step": 12861000
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 3.1392328466561696e-05,
+      "loss": 2.0529,
+      "step": 12861500
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 3.139160481891442e-05,
+      "loss": 2.0593,
+      "step": 12862000
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 3.139088117126714e-05,
+      "loss": 2.0597,
+      "step": 12862500
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 3.139015752361986e-05,
+      "loss": 2.0431,
+      "step": 12863000
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 3.1389433875972585e-05,
+      "loss": 2.0572,
+      "step": 12863500
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 3.138871022832531e-05,
+      "loss": 2.0598,
+      "step": 12864000
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 3.138798658067803e-05,
+      "loss": 2.0615,
+      "step": 12864500
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 3.138726293303075e-05,
+      "loss": 2.0538,
+      "step": 12865000
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 3.1386539285383474e-05,
+      "loss": 2.0478,
+      "step": 12865500
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 3.1385815637736196e-05,
+      "loss": 2.0511,
+      "step": 12866000
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 3.138509199008892e-05,
+      "loss": 2.0737,
+      "step": 12866500
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 3.138436978973694e-05,
+      "loss": 2.0632,
+      "step": 12867000
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 3.138364614208966e-05,
+      "loss": 2.0562,
+      "step": 12867500
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 3.1382922494442385e-05,
+      "loss": 2.0515,
+      "step": 12868000
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 3.1382198846795114e-05,
+      "loss": 2.0701,
+      "step": 12868500
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 3.1381475199147837e-05,
+      "loss": 2.0683,
+      "step": 12869000
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 3.138075299879585e-05,
+      "loss": 2.0691,
+      "step": 12869500
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 3.1380029351148574e-05,
+      "loss": 2.0651,
+      "step": 12870000
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 3.13793071507966e-05,
+      "loss": 2.0578,
+      "step": 12870500
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 3.137858350314932e-05,
+      "loss": 2.0752,
+      "step": 12871000
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 3.137785985550204e-05,
+      "loss": 2.0624,
+      "step": 12871500
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 3.137713620785476e-05,
+      "loss": 2.0585,
+      "step": 12872000
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 3.1376412560207486e-05,
+      "loss": 2.0778,
+      "step": 12872500
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 3.137568891256021e-05,
+      "loss": 2.0747,
+      "step": 12873000
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 3.137496526491293e-05,
+      "loss": 2.0462,
+      "step": 12873500
+    },
+    {
+      "epoch": 37.26,
+      "learning_rate": 3.1374243064560946e-05,
+      "loss": 2.0707,
+      "step": 12874000
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 3.1373519416913675e-05,
+      "loss": 2.0582,
+      "step": 12874500
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 3.13727957692664e-05,
+      "loss": 2.0685,
+      "step": 12875000
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 3.137207212161912e-05,
+      "loss": 2.0375,
+      "step": 12875500
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 3.137134847397185e-05,
+      "loss": 2.0823,
+      "step": 12876000
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 3.137062482632457e-05,
+      "loss": 2.0763,
+      "step": 12876500
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 3.136990117867729e-05,
+      "loss": 2.0604,
+      "step": 12877000
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 3.1369177531030015e-05,
+      "loss": 2.061,
+      "step": 12877500
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 3.136845388338274e-05,
+      "loss": 2.0807,
+      "step": 12878000
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 3.136773023573546e-05,
+      "loss": 2.0394,
+      "step": 12878500
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 3.136700658808818e-05,
+      "loss": 2.0419,
+      "step": 12879000
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 3.1366282940440904e-05,
+      "loss": 2.045,
+      "step": 12879500
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 3.1365559292793626e-05,
+      "loss": 2.0743,
+      "step": 12880000
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 3.136483564514635e-05,
+      "loss": 2.0666,
+      "step": 12880500
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 3.136411199749907e-05,
+      "loss": 2.0324,
+      "step": 12881000
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 3.136338979714709e-05,
+      "loss": 2.0815,
+      "step": 12881500
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 3.1362666149499815e-05,
+      "loss": 2.0679,
+      "step": 12882000
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 3.136194394914783e-05,
+      "loss": 2.0357,
+      "step": 12882500
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 3.136122030150055e-05,
+      "loss": 2.0576,
+      "step": 12883000
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 3.136049665385328e-05,
+      "loss": 2.0622,
+      "step": 12883500
+    },
+    {
+      "epoch": 37.29,
+      "learning_rate": 3.1359773006206004e-05,
+      "loss": 2.0647,
+      "step": 12884000
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 3.1359049358558726e-05,
+      "loss": 2.0793,
+      "step": 12884500
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 3.135832715820675e-05,
+      "loss": 2.0534,
+      "step": 12885000
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 3.135760351055947e-05,
+      "loss": 2.0466,
+      "step": 12885500
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 3.135687986291219e-05,
+      "loss": 2.0615,
+      "step": 12886000
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 3.1356156215264915e-05,
+      "loss": 2.0571,
+      "step": 12886500
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 3.135543256761764e-05,
+      "loss": 2.0773,
+      "step": 12887000
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 3.135470891997036e-05,
+      "loss": 2.0507,
+      "step": 12887500
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 3.135398527232308e-05,
+      "loss": 2.0288,
+      "step": 12888000
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 3.1353261624675804e-05,
+      "loss": 2.0455,
+      "step": 12888500
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 3.1352537977028526e-05,
+      "loss": 2.0661,
+      "step": 12889000
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 3.135181432938125e-05,
+      "loss": 2.0501,
+      "step": 12889500
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 3.135109068173398e-05,
+      "loss": 2.0811,
+      "step": 12890000
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 3.13503670340867e-05,
+      "loss": 2.045,
+      "step": 12890500
+    },
+    {
+      "epoch": 37.31,
+      "learning_rate": 3.134964483373472e-05,
+      "loss": 2.0582,
+      "step": 12891000
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 3.1348921186087445e-05,
+      "loss": 2.0642,
+      "step": 12891500
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 3.134819753844017e-05,
+      "loss": 2.061,
+      "step": 12892000
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 3.134747389079289e-05,
+      "loss": 2.0654,
+      "step": 12892500
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 3.134675024314561e-05,
+      "loss": 2.0475,
+      "step": 12893000
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 3.1346026595498333e-05,
+      "loss": 2.0303,
+      "step": 12893500
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 3.134530439514635e-05,
+      "loss": 2.0547,
+      "step": 12894000
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 3.134458074749908e-05,
+      "loss": 2.0492,
+      "step": 12894500
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 3.13438570998518e-05,
+      "loss": 2.0527,
+      "step": 12895000
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 3.1343134899499816e-05,
+      "loss": 2.0729,
+      "step": 12895500
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 3.134241125185254e-05,
+      "loss": 2.0566,
+      "step": 12896000
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 3.1341689051500554e-05,
+      "loss": 2.0602,
+      "step": 12896500
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 3.1340965403853276e-05,
+      "loss": 2.0541,
+      "step": 12897000
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 3.1340241756206e-05,
+      "loss": 2.0517,
+      "step": 12897500
+    },
+    {
+      "epoch": 37.33,
+      "learning_rate": 3.133951810855873e-05,
+      "loss": 2.0539,
+      "step": 12898000
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 3.133879446091145e-05,
+      "loss": 2.0384,
+      "step": 12898500
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 3.133807081326418e-05,
+      "loss": 2.0552,
+      "step": 12899000
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 3.13373471656169e-05,
+      "loss": 2.062,
+      "step": 12899500
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 3.133662351796962e-05,
+      "loss": 2.0334,
+      "step": 12900000
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 3.1335899870322345e-05,
+      "loss": 2.0572,
+      "step": 12900500
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 3.133517622267507e-05,
+      "loss": 2.0613,
+      "step": 12901000
+    },
+    {
+      "epoch": 37.34,
+      "learning_rate": 3.133445402232308e-05,
+      "loss": 2.0381,
+      "step": 12901500
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 3.1333730374675805e-05,
+      "loss": 2.0464,
+      "step": 12902000
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 3.133300672702853e-05,
+      "loss": 2.0224,
+      "step": 12902500
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 3.133228307938125e-05,
+      "loss": 2.029,
+      "step": 12903000
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 3.133155943173398e-05,
+      "loss": 2.05,
+      "step": 12903500
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 3.1330837231381994e-05,
+      "loss": 2.0893,
+      "step": 12904000
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 3.1330113583734716e-05,
+      "loss": 2.0721,
+      "step": 12904500
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 3.132939138338273e-05,
+      "loss": 2.0575,
+      "step": 12905000
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 3.1328667735735454e-05,
+      "loss": 2.0882,
+      "step": 12905500
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 3.132794408808818e-05,
+      "loss": 2.0648,
+      "step": 12906000
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 3.1327220440440905e-05,
+      "loss": 2.0655,
+      "step": 12906500
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 3.132649824008893e-05,
+      "loss": 2.0813,
+      "step": 12907000
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 3.132577459244165e-05,
+      "loss": 2.0641,
+      "step": 12907500
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 3.132505094479437e-05,
+      "loss": 2.0575,
+      "step": 12908000
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 3.1324327297147094e-05,
+      "loss": 2.0594,
+      "step": 12908500
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 3.1323603649499817e-05,
+      "loss": 2.0893,
+      "step": 12909000
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 3.132288000185254e-05,
+      "loss": 2.0487,
+      "step": 12909500
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 3.132215635420526e-05,
+      "loss": 2.0791,
+      "step": 12910000
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 3.132143270655798e-05,
+      "loss": 2.0554,
+      "step": 12910500
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 3.1320709058910706e-05,
+      "loss": 2.0522,
+      "step": 12911000
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 3.131998541126343e-05,
+      "loss": 2.0863,
+      "step": 12911500
+    },
+    {
+      "epoch": 37.37,
+      "learning_rate": 3.131926176361615e-05,
+      "loss": 2.0742,
+      "step": 12912000
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 3.131853811596888e-05,
+      "loss": 2.0438,
+      "step": 12912500
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 3.131781736291219e-05,
+      "loss": 2.0639,
+      "step": 12913000
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 3.131709371526492e-05,
+      "loss": 2.0705,
+      "step": 12913500
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 3.131637006761764e-05,
+      "loss": 2.053,
+      "step": 12914000
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 3.131564641997036e-05,
+      "loss": 2.0696,
+      "step": 12914500
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 3.1314922772323084e-05,
+      "loss": 2.0758,
+      "step": 12915000
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 3.1314199124675806e-05,
+      "loss": 2.0698,
+      "step": 12915500
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 3.131347547702853e-05,
+      "loss": 2.0601,
+      "step": 12916000
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 3.131275182938126e-05,
+      "loss": 2.0536,
+      "step": 12916500
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 3.131202818173398e-05,
+      "loss": 2.0587,
+      "step": 12917000
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 3.13113045340867e-05,
+      "loss": 2.0739,
+      "step": 12917500
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 3.131058233373472e-05,
+      "loss": 2.0908,
+      "step": 12918000
+    },
+    {
+      "epoch": 37.39,
+      "learning_rate": 3.130986013338273e-05,
+      "loss": 2.0574,
+      "step": 12918500
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 3.1309136485735455e-05,
+      "loss": 2.077,
+      "step": 12919000
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 3.130841283808818e-05,
+      "loss": 2.0645,
+      "step": 12919500
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 3.1307689190440906e-05,
+      "loss": 2.0395,
+      "step": 12920000
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 3.130696554279363e-05,
+      "loss": 2.0849,
+      "step": 12920500
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 3.130624189514636e-05,
+      "loss": 2.0505,
+      "step": 12921000
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 3.130551969479437e-05,
+      "loss": 2.0597,
+      "step": 12921500
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 3.1304796047147095e-05,
+      "loss": 2.0682,
+      "step": 12922000
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 3.130407239949982e-05,
+      "loss": 2.0476,
+      "step": 12922500
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 3.130334875185254e-05,
+      "loss": 2.0432,
+      "step": 12923000
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 3.130262510420526e-05,
+      "loss": 2.0715,
+      "step": 12923500
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 3.1301901456557984e-05,
+      "loss": 2.0566,
+      "step": 12924000
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 3.1301179256206006e-05,
+      "loss": 2.068,
+      "step": 12924500
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 3.130045560855873e-05,
+      "loss": 2.0501,
+      "step": 12925000
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 3.129973196091145e-05,
+      "loss": 2.0386,
+      "step": 12925500
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 3.129900831326417e-05,
+      "loss": 2.0582,
+      "step": 12926000
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 3.129828611291219e-05,
+      "loss": 2.0804,
+      "step": 12926500
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 3.129756246526491e-05,
+      "loss": 2.0418,
+      "step": 12927000
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 3.129684026491293e-05,
+      "loss": 2.0449,
+      "step": 12927500
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 3.129611806456095e-05,
+      "loss": 2.04,
+      "step": 12928000
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 3.129539441691368e-05,
+      "loss": 2.0667,
+      "step": 12928500
+    },
+    {
+      "epoch": 37.42,
+      "learning_rate": 3.12946707692664e-05,
+      "loss": 2.0602,
+      "step": 12929000
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 3.129394712161912e-05,
+      "loss": 2.0692,
+      "step": 12929500
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 3.1293223473971845e-05,
+      "loss": 2.0568,
+      "step": 12930000
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 3.129250127361986e-05,
+      "loss": 2.0477,
+      "step": 12930500
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 3.129177762597258e-05,
+      "loss": 2.0329,
+      "step": 12931000
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 3.1291053978325305e-05,
+      "loss": 2.0705,
+      "step": 12931500
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 3.1290330330678034e-05,
+      "loss": 2.0621,
+      "step": 12932000
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 3.1289606683030756e-05,
+      "loss": 2.0655,
+      "step": 12932500
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 3.128888303538348e-05,
+      "loss": 2.0733,
+      "step": 12933000
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 3.12881593877362e-05,
+      "loss": 2.0578,
+      "step": 12933500
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 3.128743574008892e-05,
+      "loss": 2.0701,
+      "step": 12934000
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 3.1286712092441645e-05,
+      "loss": 2.0704,
+      "step": 12934500
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 3.128598844479437e-05,
+      "loss": 2.0413,
+      "step": 12935000
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 3.128526479714709e-05,
+      "loss": 2.0599,
+      "step": 12935500
+    },
+    {
+      "epoch": 37.44,
+      "learning_rate": 3.128454114949982e-05,
+      "loss": 2.0868,
+      "step": 12936000
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 3.1283818949147834e-05,
+      "loss": 2.0592,
+      "step": 12936500
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 3.1283095301500556e-05,
+      "loss": 2.0408,
+      "step": 12937000
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 3.1282371653853285e-05,
+      "loss": 2.0478,
+      "step": 12937500
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 3.128164800620601e-05,
+      "loss": 2.0526,
+      "step": 12938000
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 3.128092435855873e-05,
+      "loss": 2.0834,
+      "step": 12938500
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 3.1280202158206745e-05,
+      "loss": 2.0592,
+      "step": 12939000
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 3.127947851055947e-05,
+      "loss": 2.0545,
+      "step": 12939500
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 3.127875486291219e-05,
+      "loss": 2.0677,
+      "step": 12940000
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 3.127803121526491e-05,
+      "loss": 2.0807,
+      "step": 12940500
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 3.1277309014912934e-05,
+      "loss": 2.0602,
+      "step": 12941000
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 3.1276585367265656e-05,
+      "loss": 2.0448,
+      "step": 12941500
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 3.127586171961838e-05,
+      "loss": 2.0497,
+      "step": 12942000
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 3.1275139519266394e-05,
+      "loss": 2.0583,
+      "step": 12942500
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 3.1274415871619116e-05,
+      "loss": 2.0252,
+      "step": 12943000
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 3.1273692223971845e-05,
+      "loss": 2.0688,
+      "step": 12943500
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 3.127296857632457e-05,
+      "loss": 2.0565,
+      "step": 12944000
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 3.127224492867729e-05,
+      "loss": 2.0713,
+      "step": 12944500
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 3.127152128103001e-05,
+      "loss": 2.066,
+      "step": 12945000
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 3.1270797633382734e-05,
+      "loss": 2.0758,
+      "step": 12945500
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 3.1270073985735456e-05,
+      "loss": 2.0659,
+      "step": 12946000
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 3.1269350338088185e-05,
+      "loss": 2.0594,
+      "step": 12946500
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 3.12686281377362e-05,
+      "loss": 2.0371,
+      "step": 12947000
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 3.126790449008892e-05,
+      "loss": 2.0485,
+      "step": 12947500
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 3.1267180842441646e-05,
+      "loss": 2.0671,
+      "step": 12948000
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 3.126645864208966e-05,
+      "loss": 2.0744,
+      "step": 12948500
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 3.126573499444238e-05,
+      "loss": 2.0475,
+      "step": 12949000
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 3.1265011346795106e-05,
+      "loss": 2.0244,
+      "step": 12949500
+    },
+    {
+      "epoch": 37.48,
+      "learning_rate": 3.126428914644313e-05,
+      "loss": 2.0682,
+      "step": 12950000
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 3.126356549879585e-05,
+      "loss": 2.0616,
+      "step": 12950500
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 3.126284185114858e-05,
+      "loss": 2.0552,
+      "step": 12951000
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 3.1262119650796595e-05,
+      "loss": 2.0635,
+      "step": 12951500
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 3.126139600314932e-05,
+      "loss": 2.0865,
+      "step": 12952000
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 3.126067235550204e-05,
+      "loss": 2.0461,
+      "step": 12952500
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 3.125995015515006e-05,
+      "loss": 2.0444,
+      "step": 12953000
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 3.1259226507502784e-05,
+      "loss": 2.0739,
+      "step": 12953500
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 3.1258502859855506e-05,
+      "loss": 2.025,
+      "step": 12954000
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 3.125777921220823e-05,
+      "loss": 2.0528,
+      "step": 12954500
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 3.1257057011856244e-05,
+      "loss": 2.0711,
+      "step": 12955000
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 3.1256333364208966e-05,
+      "loss": 2.0575,
+      "step": 12955500
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 3.125560971656169e-05,
+      "loss": 2.053,
+      "step": 12956000
+    },
+    {
+      "epoch": 37.5,
+      "learning_rate": 3.125488606891441e-05,
+      "loss": 2.0497,
+      "step": 12956500
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 3.125416242126713e-05,
+      "loss": 2.0575,
+      "step": 12957000
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 3.125343877361986e-05,
+      "loss": 2.0728,
+      "step": 12957500
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 3.1252715125972584e-05,
+      "loss": 2.0736,
+      "step": 12958000
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 3.125199147832531e-05,
+      "loss": 2.0505,
+      "step": 12958500
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 3.1251267830678035e-05,
+      "loss": 2.0499,
+      "step": 12959000
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 3.125054418303076e-05,
+      "loss": 2.0535,
+      "step": 12959500
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 3.124982053538348e-05,
+      "loss": 2.0633,
+      "step": 12960000
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 3.12490968877362e-05,
+      "loss": 2.0805,
+      "step": 12960500
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 3.1248373240088924e-05,
+      "loss": 2.037,
+      "step": 12961000
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 3.1247649592441646e-05,
+      "loss": 2.0465,
+      "step": 12961500
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 3.124692594479437e-05,
+      "loss": 2.039,
+      "step": 12962000
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 3.124620229714709e-05,
+      "loss": 2.0435,
+      "step": 12962500
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 3.124547864949981e-05,
+      "loss": 2.0831,
+      "step": 12963000
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 3.1244756449147835e-05,
+      "loss": 2.096,
+      "step": 12963500
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 3.124403280150056e-05,
+      "loss": 2.0609,
+      "step": 12964000
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 3.124330915385328e-05,
+      "loss": 2.057,
+      "step": 12964500
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 3.1242585506206e-05,
+      "loss": 2.0515,
+      "step": 12965000
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 3.124186185855873e-05,
+      "loss": 2.0721,
+      "step": 12965500
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 3.124113821091145e-05,
+      "loss": 2.0519,
+      "step": 12966000
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 3.1240414563264176e-05,
+      "loss": 2.0615,
+      "step": 12966500
+    },
+    {
+      "epoch": 37.53,
+      "learning_rate": 3.12396909156169e-05,
+      "loss": 2.0499,
+      "step": 12967000
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 3.1238970162560213e-05,
+      "loss": 2.064,
+      "step": 12967500
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 3.1238246514912936e-05,
+      "loss": 2.0364,
+      "step": 12968000
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 3.123752286726566e-05,
+      "loss": 2.0707,
+      "step": 12968500
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 3.123679921961838e-05,
+      "loss": 2.0493,
+      "step": 12969000
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 3.12360755719711e-05,
+      "loss": 2.0695,
+      "step": 12969500
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 3.1235351924323825e-05,
+      "loss": 2.0385,
+      "step": 12970000
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 3.123462827667655e-05,
+      "loss": 2.0549,
+      "step": 12970500
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 3.123390462902927e-05,
+      "loss": 2.0601,
+      "step": 12971000
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 3.123318098138199e-05,
+      "loss": 2.0542,
+      "step": 12971500
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 3.1232458781030014e-05,
+      "loss": 2.0738,
+      "step": 12972000
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 3.1231735133382736e-05,
+      "loss": 2.0707,
+      "step": 12972500
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 3.123101148573546e-05,
+      "loss": 2.069,
+      "step": 12973000
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 3.123028783808819e-05,
+      "loss": 2.0603,
+      "step": 12973500
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 3.12295656377362e-05,
+      "loss": 2.0437,
+      "step": 12974000
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 3.1228841990088925e-05,
+      "loss": 2.0795,
+      "step": 12974500
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 3.122811834244165e-05,
+      "loss": 2.0322,
+      "step": 12975000
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 3.122739469479437e-05,
+      "loss": 2.0521,
+      "step": 12975500
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 3.122667104714709e-05,
+      "loss": 2.0425,
+      "step": 12976000
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 3.1225948846795114e-05,
+      "loss": 2.0516,
+      "step": 12976500
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 3.1225225199147836e-05,
+      "loss": 2.0721,
+      "step": 12977000
+    },
+    {
+      "epoch": 37.56,
+      "learning_rate": 3.122450155150056e-05,
+      "loss": 2.0462,
+      "step": 12977500
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 3.122377790385328e-05,
+      "loss": 2.0781,
+      "step": 12978000
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 3.1223054256206e-05,
+      "loss": 2.0781,
+      "step": 12978500
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 3.1222330608558725e-05,
+      "loss": 2.0595,
+      "step": 12979000
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 3.122160696091145e-05,
+      "loss": 2.0636,
+      "step": 12979500
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 3.122088331326417e-05,
+      "loss": 2.0642,
+      "step": 12980000
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 3.12201596656169e-05,
+      "loss": 2.0513,
+      "step": 12980500
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 3.1219437465264914e-05,
+      "loss": 2.0417,
+      "step": 12981000
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 3.1218713817617636e-05,
+      "loss": 2.0468,
+      "step": 12981500
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 3.1217990169970365e-05,
+      "loss": 2.0397,
+      "step": 12982000
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 3.121726796961838e-05,
+      "loss": 2.0677,
+      "step": 12982500
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 3.12165443219711e-05,
+      "loss": 2.088,
+      "step": 12983000
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 3.1215820674323825e-05,
+      "loss": 2.0451,
+      "step": 12983500
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 3.121509702667655e-05,
+      "loss": 2.0528,
+      "step": 12984000
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 3.121437337902927e-05,
+      "loss": 2.0534,
+      "step": 12984500
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 3.121364973138199e-05,
+      "loss": 2.0491,
+      "step": 12985000
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 3.1212926083734714e-05,
+      "loss": 2.0669,
+      "step": 12985500
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 3.1212202436087437e-05,
+      "loss": 2.0712,
+      "step": 12986000
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 3.121148023573546e-05,
+      "loss": 2.0734,
+      "step": 12986500
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 3.1210758035383474e-05,
+      "loss": 2.0759,
+      "step": 12987000
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 3.12100343877362e-05,
+      "loss": 2.0645,
+      "step": 12987500
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 3.120931074008892e-05,
+      "loss": 2.0643,
+      "step": 12988000
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 3.120858709244165e-05,
+      "loss": 2.0725,
+      "step": 12988500
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 3.120786344479437e-05,
+      "loss": 2.0572,
+      "step": 12989000
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 3.120713979714709e-05,
+      "loss": 2.0993,
+      "step": 12989500
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 3.1206416149499815e-05,
+      "loss": 2.0685,
+      "step": 12990000
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 3.1205692501852544e-05,
+      "loss": 2.0974,
+      "step": 12990500
+    },
+    {
+      "epoch": 37.6,
+      "learning_rate": 3.120497030150056e-05,
+      "loss": 2.0513,
+      "step": 12991000
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 3.1204248101148575e-05,
+      "loss": 2.0729,
+      "step": 12991500
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 3.12035244535013e-05,
+      "loss": 2.0614,
+      "step": 12992000
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 3.120280080585402e-05,
+      "loss": 2.0464,
+      "step": 12992500
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 3.120207715820674e-05,
+      "loss": 2.0649,
+      "step": 12993000
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 3.1201353510559464e-05,
+      "loss": 2.0765,
+      "step": 12993500
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 3.120062986291219e-05,
+      "loss": 2.0598,
+      "step": 12994000
+    },
+    {
+      "epoch": 37.61,
+      "learning_rate": 3.1199906215264915e-05,
+      "loss": 2.0928,
+      "step": 12994500
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 3.119918401491293e-05,
+      "loss": 2.0184,
+      "step": 12995000
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 3.1198461814560946e-05,
+      "loss": 2.0607,
+      "step": 12995500
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 3.119773961420897e-05,
+      "loss": 2.0571,
+      "step": 12996000
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 3.119701596656169e-05,
+      "loss": 2.0721,
+      "step": 12996500
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 3.119629231891442e-05,
+      "loss": 2.0784,
+      "step": 12997000
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 3.119556867126714e-05,
+      "loss": 2.067,
+      "step": 12997500
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 3.119484647091516e-05,
+      "loss": 2.079,
+      "step": 12998000
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 3.119412282326788e-05,
+      "loss": 2.0682,
+      "step": 12998500
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 3.11933991756206e-05,
+      "loss": 2.0483,
+      "step": 12999000
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 3.1192675527973324e-05,
+      "loss": 2.0239,
+      "step": 12999500
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 3.1191951880326046e-05,
+      "loss": 2.072,
+      "step": 13000000
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 3.119122823267877e-05,
+      "loss": 2.0434,
+      "step": 13000500
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 3.119050458503149e-05,
+      "loss": 2.0682,
+      "step": 13001000
+    },
+    {
+      "epoch": 37.63,
+      "learning_rate": 3.118978093738422e-05,
+      "loss": 2.0724,
+      "step": 13001500
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 3.118905728973694e-05,
+      "loss": 2.0547,
+      "step": 13002000
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 3.1188333642089664e-05,
+      "loss": 2.0732,
+      "step": 13002500
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 3.1187609994442386e-05,
+      "loss": 2.0517,
+      "step": 13003000
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 3.1186886346795115e-05,
+      "loss": 2.0773,
+      "step": 13003500
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 3.118616269914784e-05,
+      "loss": 2.0505,
+      "step": 13004000
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 3.118543905150056e-05,
+      "loss": 2.0826,
+      "step": 13004500
+    },
+    {
+      "epoch": 37.64,
+      "learning_rate": 3.1184716851148576e-05,
+      "loss": 2.0816,
+      "step": 13005000
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 3.11839932035013e-05,
+      "loss": 2.0768,
+      "step": 13005500
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 3.118326955585402e-05,
+      "loss": 2.0453,
+      "step": 13006000
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 3.118254590820674e-05,
+      "loss": 2.0612,
+      "step": 13006500
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 3.118182226055947e-05,
+      "loss": 2.0591,
+      "step": 13007000
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 3.1181098612912193e-05,
+      "loss": 2.0704,
+      "step": 13007500
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 3.1180374965264916e-05,
+      "loss": 2.0676,
+      "step": 13008000
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 3.117965131761764e-05,
+      "loss": 2.0484,
+      "step": 13008500
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 3.117892766997036e-05,
+      "loss": 2.0778,
+      "step": 13009000
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 3.117820402232308e-05,
+      "loss": 2.0581,
+      "step": 13009500
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 3.11774818219711e-05,
+      "loss": 2.0618,
+      "step": 13010000
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 3.117675817432382e-05,
+      "loss": 2.0848,
+      "step": 13010500
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 3.117603452667655e-05,
+      "loss": 2.084,
+      "step": 13011000
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 3.117531087902927e-05,
+      "loss": 2.0389,
+      "step": 13011500
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 3.1174587231381994e-05,
+      "loss": 2.0614,
+      "step": 13012000
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 3.1173863583734716e-05,
+      "loss": 2.0797,
+      "step": 13012500
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 3.1173139936087445e-05,
+      "loss": 2.055,
+      "step": 13013000
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 3.117241628844017e-05,
+      "loss": 2.0407,
+      "step": 13013500
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 3.117169264079289e-05,
+      "loss": 2.0679,
+      "step": 13014000
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 3.117096899314561e-05,
+      "loss": 2.0674,
+      "step": 13014500
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 3.1170245345498334e-05,
+      "loss": 2.0371,
+      "step": 13015000
+    },
+    {
+      "epoch": 37.67,
+      "learning_rate": 3.1169521697851056e-05,
+      "loss": 2.0629,
+      "step": 13015500
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 3.116879805020378e-05,
+      "loss": 2.0723,
+      "step": 13016000
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 3.11680744025565e-05,
+      "loss": 2.0338,
+      "step": 13016500
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 3.116735075490922e-05,
+      "loss": 2.0541,
+      "step": 13017000
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 3.116662710726195e-05,
+      "loss": 2.0581,
+      "step": 13017500
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 3.116590635420527e-05,
+      "loss": 2.0729,
+      "step": 13018000
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 3.116518270655799e-05,
+      "loss": 2.0544,
+      "step": 13018500
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 3.116445905891071e-05,
+      "loss": 2.0492,
+      "step": 13019000
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 3.1163735411263434e-05,
+      "loss": 2.0652,
+      "step": 13019500
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 3.1163011763616156e-05,
+      "loss": 2.0614,
+      "step": 13020000
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 3.116228956326417e-05,
+      "loss": 2.0669,
+      "step": 13020500
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 3.1161565915616894e-05,
+      "loss": 2.0642,
+      "step": 13021000
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 3.116084226796962e-05,
+      "loss": 2.0572,
+      "step": 13021500
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 3.1160118620322345e-05,
+      "loss": 2.046,
+      "step": 13022000
+    },
+    {
+      "epoch": 37.69,
+      "learning_rate": 3.115939497267507e-05,
+      "loss": 2.0786,
+      "step": 13022500
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 3.115867132502779e-05,
+      "loss": 2.0744,
+      "step": 13023000
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 3.1157949124675805e-05,
+      "loss": 2.0651,
+      "step": 13023500
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 3.115722547702853e-05,
+      "loss": 2.0604,
+      "step": 13024000
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 3.115650182938125e-05,
+      "loss": 2.068,
+      "step": 13024500
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 3.115577818173397e-05,
+      "loss": 2.0445,
+      "step": 13025000
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 3.11550545340867e-05,
+      "loss": 2.0803,
+      "step": 13025500
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 3.115433088643942e-05,
+      "loss": 2.0443,
+      "step": 13026000
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 3.1153607238792146e-05,
+      "loss": 2.0782,
+      "step": 13026500
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 3.1152883591144875e-05,
+      "loss": 2.0559,
+      "step": 13027000
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 3.11521599434976e-05,
+      "loss": 2.068,
+      "step": 13027500
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 3.115143774314561e-05,
+      "loss": 2.0674,
+      "step": 13028000
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 3.1150714095498335e-05,
+      "loss": 2.0528,
+      "step": 13028500
+    },
+    {
+      "epoch": 37.71,
+      "learning_rate": 3.114999044785106e-05,
+      "loss": 2.0493,
+      "step": 13029000
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 3.114926680020378e-05,
+      "loss": 2.0774,
+      "step": 13029500
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 3.11485431525565e-05,
+      "loss": 2.0528,
+      "step": 13030000
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 3.1147820952204524e-05,
+      "loss": 2.0843,
+      "step": 13030500
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 3.114710019914783e-05,
+      "loss": 2.074,
+      "step": 13031000
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 3.1146376551500555e-05,
+      "loss": 2.0806,
+      "step": 13031500
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 3.114565290385328e-05,
+      "loss": 2.0679,
+      "step": 13032000
+    },
+    {
+      "epoch": 37.72,
+      "learning_rate": 3.1144929256206e-05,
+      "loss": 2.0684,
+      "step": 13032500
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 3.114420705585402e-05,
+      "loss": 2.0601,
+      "step": 13033000
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 3.114348340820675e-05,
+      "loss": 2.0453,
+      "step": 13033500
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 3.1142761207854766e-05,
+      "loss": 2.0596,
+      "step": 13034000
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 3.114203756020749e-05,
+      "loss": 2.0577,
+      "step": 13034500
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 3.1141315359855504e-05,
+      "loss": 2.0553,
+      "step": 13035000
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 3.1140591712208226e-05,
+      "loss": 2.0602,
+      "step": 13035500
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 3.113986806456095e-05,
+      "loss": 2.0892,
+      "step": 13036000
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 3.113914441691367e-05,
+      "loss": 2.0753,
+      "step": 13036500
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 3.11384207692664e-05,
+      "loss": 2.0805,
+      "step": 13037000
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 3.113769712161912e-05,
+      "loss": 2.0836,
+      "step": 13037500
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 3.1136973473971844e-05,
+      "loss": 2.0609,
+      "step": 13038000
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 3.1136249826324566e-05,
+      "loss": 2.0647,
+      "step": 13038500
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 3.113552617867729e-05,
+      "loss": 2.0672,
+      "step": 13039000
+    },
+    {
+      "epoch": 37.74,
+      "learning_rate": 3.113480253103001e-05,
+      "loss": 2.0502,
+      "step": 13039500
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 3.113407888338273e-05,
+      "loss": 2.0691,
+      "step": 13040000
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 3.1133355235735455e-05,
+      "loss": 2.0763,
+      "step": 13040500
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 3.113263303538348e-05,
+      "loss": 2.0577,
+      "step": 13041000
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 3.11319093877362e-05,
+      "loss": 2.0579,
+      "step": 13041500
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 3.113118574008892e-05,
+      "loss": 2.064,
+      "step": 13042000
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 3.113046209244165e-05,
+      "loss": 2.0682,
+      "step": 13042500
+    },
+    {
+      "epoch": 37.75,
+      "learning_rate": 3.112973844479437e-05,
+      "loss": 2.0798,
+      "step": 13043000
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 3.1129014797147096e-05,
+      "loss": 2.0705,
+      "step": 13043500
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 3.112829114949982e-05,
+      "loss": 2.0457,
+      "step": 13044000
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 3.112756750185254e-05,
+      "loss": 2.05,
+      "step": 13044500
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 3.112684385420526e-05,
+      "loss": 2.0612,
+      "step": 13045000
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 3.112612165385328e-05,
+      "loss": 2.0359,
+      "step": 13045500
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 3.1125398006206e-05,
+      "loss": 2.0482,
+      "step": 13046000
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 3.112467435855872e-05,
+      "loss": 2.087,
+      "step": 13046500
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 3.112395071091145e-05,
+      "loss": 2.0763,
+      "step": 13047000
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 3.1123227063264174e-05,
+      "loss": 2.0689,
+      "step": 13047500
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 3.11225034156169e-05,
+      "loss": 2.0445,
+      "step": 13048000
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 3.1121779767969625e-05,
+      "loss": 2.0461,
+      "step": 13048500
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 3.112105612032235e-05,
+      "loss": 2.0613,
+      "step": 13049000
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 3.112033391997036e-05,
+      "loss": 2.0532,
+      "step": 13049500
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 3.1119610272323085e-05,
+      "loss": 2.0528,
+      "step": 13050000
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 3.111888662467581e-05,
+      "loss": 2.0489,
+      "step": 13050500
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 3.111816442432382e-05,
+      "loss": 2.0627,
+      "step": 13051000
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 3.111744077667655e-05,
+      "loss": 2.0741,
+      "step": 13051500
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 3.1116717129029274e-05,
+      "loss": 2.0807,
+      "step": 13052000
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 3.1115993481381996e-05,
+      "loss": 2.0447,
+      "step": 13052500
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 3.111526983373472e-05,
+      "loss": 2.0637,
+      "step": 13053000
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 3.111454618608744e-05,
+      "loss": 2.0359,
+      "step": 13053500
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 3.111382253844016e-05,
+      "loss": 2.0793,
+      "step": 13054000
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 3.1113098890792885e-05,
+      "loss": 2.0547,
+      "step": 13054500
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 3.111237524314561e-05,
+      "loss": 2.0743,
+      "step": 13055000
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 3.1111651595498336e-05,
+      "loss": 2.0598,
+      "step": 13055500
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 3.111092794785106e-05,
+      "loss": 2.0667,
+      "step": 13056000
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 3.111020430020378e-05,
+      "loss": 2.0702,
+      "step": 13056500
+    },
+    {
+      "epoch": 37.79,
+      "learning_rate": 3.11094820998518e-05,
+      "loss": 2.0662,
+      "step": 13057000
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 3.1108758452204525e-05,
+      "loss": 2.0537,
+      "step": 13057500
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 3.110803480455725e-05,
+      "loss": 2.0857,
+      "step": 13058000
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 3.110731260420526e-05,
+      "loss": 2.0733,
+      "step": 13058500
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 3.1106588956557985e-05,
+      "loss": 2.0607,
+      "step": 13059000
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 3.110586530891071e-05,
+      "loss": 2.0756,
+      "step": 13059500
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 3.110514166126343e-05,
+      "loss": 2.039,
+      "step": 13060000
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 3.110441801361615e-05,
+      "loss": 2.0757,
+      "step": 13060500
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 3.1103694365968874e-05,
+      "loss": 2.0881,
+      "step": 13061000
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 3.11029707183216e-05,
+      "loss": 2.0675,
+      "step": 13061500
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 3.1102247070674325e-05,
+      "loss": 2.0592,
+      "step": 13062000
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 3.110152342302705e-05,
+      "loss": 2.0589,
+      "step": 13062500
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 3.110080122267507e-05,
+      "loss": 2.0908,
+      "step": 13063000
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 3.110007757502779e-05,
+      "loss": 2.0524,
+      "step": 13063500
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 3.1099353927380514e-05,
+      "loss": 2.0789,
+      "step": 13064000
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 3.109863027973324e-05,
+      "loss": 2.089,
+      "step": 13064500
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 3.109790807938125e-05,
+      "loss": 2.0471,
+      "step": 13065000
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 3.1097184431733975e-05,
+      "loss": 2.0857,
+      "step": 13065500
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 3.1096460784086704e-05,
+      "loss": 2.0523,
+      "step": 13066000
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 3.1095737136439426e-05,
+      "loss": 2.0631,
+      "step": 13066500
+    },
+    {
+      "epoch": 37.82,
+      "learning_rate": 3.109501348879215e-05,
+      "loss": 2.0495,
+      "step": 13067000
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 3.109428984114487e-05,
+      "loss": 2.0886,
+      "step": 13067500
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 3.109356619349759e-05,
+      "loss": 2.0564,
+      "step": 13068000
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 3.1092842545850315e-05,
+      "loss": 2.0927,
+      "step": 13068500
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 3.109211889820304e-05,
+      "loss": 2.0418,
+      "step": 13069000
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 3.109139525055576e-05,
+      "loss": 2.0646,
+      "step": 13069500
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 3.109067160290849e-05,
+      "loss": 2.0707,
+      "step": 13070000
+    },
+    {
+      "epoch": 37.83,
+      "learning_rate": 3.1089949402556504e-05,
+      "loss": 2.0763,
+      "step": 13070500
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 3.1089227202204526e-05,
+      "loss": 2.0855,
+      "step": 13071000
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 3.108850355455725e-05,
+      "loss": 2.0744,
+      "step": 13071500
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 3.108777990690997e-05,
+      "loss": 2.0608,
+      "step": 13072000
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 3.108705625926269e-05,
+      "loss": 2.0469,
+      "step": 13072500
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 3.108633405891071e-05,
+      "loss": 2.0781,
+      "step": 13073000
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 3.108561041126343e-05,
+      "loss": 2.0749,
+      "step": 13073500
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 3.108488676361615e-05,
+      "loss": 2.0825,
+      "step": 13074000
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 3.108416311596888e-05,
+      "loss": 2.0708,
+      "step": 13074500
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 3.1083439468321604e-05,
+      "loss": 2.0712,
+      "step": 13075000
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 3.108271726796962e-05,
+      "loss": 2.0474,
+      "step": 13075500
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 3.108199362032234e-05,
+      "loss": 2.0654,
+      "step": 13076000
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 3.1081269972675064e-05,
+      "loss": 2.0559,
+      "step": 13076500
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 3.1080546325027786e-05,
+      "loss": 2.0528,
+      "step": 13077000
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 3.107982267738051e-05,
+      "loss": 2.0515,
+      "step": 13077500
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 3.107909902973324e-05,
+      "loss": 2.0816,
+      "step": 13078000
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 3.107837538208596e-05,
+      "loss": 2.0847,
+      "step": 13078500
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 3.107765173443868e-05,
+      "loss": 2.072,
+      "step": 13079000
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 3.1076928086791404e-05,
+      "loss": 2.0768,
+      "step": 13079500
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 3.107620733373472e-05,
+      "loss": 2.0708,
+      "step": 13080000
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 3.107548368608744e-05,
+      "loss": 2.0646,
+      "step": 13080500
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 3.1074760038440164e-05,
+      "loss": 2.0765,
+      "step": 13081000
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 3.1074036390792887e-05,
+      "loss": 2.0481,
+      "step": 13081500
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 3.107331274314561e-05,
+      "loss": 2.0584,
+      "step": 13082000
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 3.107258909549833e-05,
+      "loss": 2.0704,
+      "step": 13082500
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 3.107186544785105e-05,
+      "loss": 2.04,
+      "step": 13083000
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 3.107114180020378e-05,
+      "loss": 2.0568,
+      "step": 13083500
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 3.1070418152556505e-05,
+      "loss": 2.0814,
+      "step": 13084000
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 3.106969450490923e-05,
+      "loss": 2.0688,
+      "step": 13084500
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 3.1068970857261956e-05,
+      "loss": 2.0679,
+      "step": 13085000
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 3.106824720961468e-05,
+      "loss": 2.0804,
+      "step": 13085500
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 3.10675235619674e-05,
+      "loss": 2.074,
+      "step": 13086000
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 3.106679991432012e-05,
+      "loss": 2.073,
+      "step": 13086500
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 3.1066076266672845e-05,
+      "loss": 2.062,
+      "step": 13087000
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 3.106535261902557e-05,
+      "loss": 2.0791,
+      "step": 13087500
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 3.106462897137829e-05,
+      "loss": 2.0748,
+      "step": 13088000
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 3.106390532373101e-05,
+      "loss": 2.068,
+      "step": 13088500
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 3.1063183123379034e-05,
+      "loss": 2.0415,
+      "step": 13089000
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 3.1062459475731756e-05,
+      "loss": 2.073,
+      "step": 13089500
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 3.106173582808448e-05,
+      "loss": 2.0688,
+      "step": 13090000
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 3.1061013627732494e-05,
+      "loss": 2.0263,
+      "step": 13090500
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 3.1060289980085216e-05,
+      "loss": 2.0741,
+      "step": 13091000
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 3.105956633243794e-05,
+      "loss": 2.0441,
+      "step": 13091500
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 3.105884268479066e-05,
+      "loss": 2.0585,
+      "step": 13092000
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 3.105811903714339e-05,
+      "loss": 2.0773,
+      "step": 13092500
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 3.105739538949611e-05,
+      "loss": 2.1005,
+      "step": 13093000
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 3.1056671741848834e-05,
+      "loss": 2.0548,
+      "step": 13093500
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 3.1055948094201556e-05,
+      "loss": 2.0559,
+      "step": 13094000
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 3.1055224446554285e-05,
+      "loss": 2.0522,
+      "step": 13094500
+    },
+    {
+      "epoch": 37.9,
+      "learning_rate": 3.10545022462023e-05,
+      "loss": 2.0576,
+      "step": 13095000
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 3.105377859855502e-05,
+      "loss": 2.0774,
+      "step": 13095500
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 3.105305639820304e-05,
+      "loss": 2.0454,
+      "step": 13096000
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 3.105233275055576e-05,
+      "loss": 2.0919,
+      "step": 13096500
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 3.105160910290848e-05,
+      "loss": 2.0861,
+      "step": 13097000
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 3.1050885455261205e-05,
+      "loss": 2.0446,
+      "step": 13097500
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 3.1050161807613934e-05,
+      "loss": 2.0686,
+      "step": 13098000
+    },
+    {
+      "epoch": 37.91,
+      "learning_rate": 3.104943960726195e-05,
+      "loss": 2.0686,
+      "step": 13098500
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 3.104871595961467e-05,
+      "loss": 2.0757,
+      "step": 13099000
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 3.104799375926269e-05,
+      "loss": 2.063,
+      "step": 13099500
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 3.104727011161541e-05,
+      "loss": 2.0573,
+      "step": 13100000
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 3.104654646396814e-05,
+      "loss": 2.0646,
+      "step": 13100500
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 3.104582281632086e-05,
+      "loss": 2.0649,
+      "step": 13101000
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 3.104509916867358e-05,
+      "loss": 2.0574,
+      "step": 13101500
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 3.1044375521026306e-05,
+      "loss": 2.0667,
+      "step": 13102000
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 3.104365332067433e-05,
+      "loss": 2.031,
+      "step": 13102500
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 3.104292967302705e-05,
+      "loss": 2.0743,
+      "step": 13103000
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 3.104220602537977e-05,
+      "loss": 2.0723,
+      "step": 13103500
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 3.1041482377732495e-05,
+      "loss": 2.0787,
+      "step": 13104000
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 3.104075873008522e-05,
+      "loss": 2.0749,
+      "step": 13104500
+    },
+    {
+      "epoch": 37.93,
+      "learning_rate": 3.104003508243794e-05,
+      "loss": 2.0696,
+      "step": 13105000
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 3.103931143479066e-05,
+      "loss": 2.0562,
+      "step": 13105500
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 3.1038587787143383e-05,
+      "loss": 2.0592,
+      "step": 13106000
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 3.1037865586791406e-05,
+      "loss": 2.0658,
+      "step": 13106500
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 3.103714193914413e-05,
+      "loss": 2.0603,
+      "step": 13107000
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 3.103641829149686e-05,
+      "loss": 2.0739,
+      "step": 13107500
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 3.103569464384958e-05,
+      "loss": 2.0598,
+      "step": 13108000
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 3.10349709962023e-05,
+      "loss": 2.0548,
+      "step": 13108500
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 3.1034247348555024e-05,
+      "loss": 2.0726,
+      "step": 13109000
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 3.103352514820304e-05,
+      "loss": 2.052,
+      "step": 13109500
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 3.103280150055576e-05,
+      "loss": 2.0629,
+      "step": 13110000
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 3.1032077852908484e-05,
+      "loss": 2.0712,
+      "step": 13110500
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 3.1031355652556506e-05,
+      "loss": 2.0754,
+      "step": 13111000
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 3.103063200490923e-05,
+      "loss": 2.0396,
+      "step": 13111500
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 3.1029909804557244e-05,
+      "loss": 2.0422,
+      "step": 13112000
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 3.1029186156909966e-05,
+      "loss": 2.0361,
+      "step": 13112500
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 3.102846250926269e-05,
+      "loss": 2.0856,
+      "step": 13113000
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 3.102773886161541e-05,
+      "loss": 2.0545,
+      "step": 13113500
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 3.102701521396813e-05,
+      "loss": 2.0788,
+      "step": 13114000
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 3.102629156632086e-05,
+      "loss": 2.0677,
+      "step": 13114500
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 3.102556936596888e-05,
+      "loss": 2.0607,
+      "step": 13115000
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 3.1024845718321606e-05,
+      "loss": 2.0728,
+      "step": 13115500
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 3.102412207067433e-05,
+      "loss": 2.0563,
+      "step": 13116000
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 3.102339842302705e-05,
+      "loss": 2.0363,
+      "step": 13116500
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 3.1022676222675066e-05,
+      "loss": 2.0633,
+      "step": 13117000
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 3.102195257502779e-05,
+      "loss": 2.0755,
+      "step": 13117500
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 3.102122892738051e-05,
+      "loss": 2.0742,
+      "step": 13118000
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 3.102050527973323e-05,
+      "loss": 2.0685,
+      "step": 13118500
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 3.101978163208596e-05,
+      "loss": 2.0674,
+      "step": 13119000
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 3.1019057984438684e-05,
+      "loss": 2.0524,
+      "step": 13119500
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 3.10183357840867e-05,
+      "loss": 2.0875,
+      "step": 13120000
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 3.101761213643942e-05,
+      "loss": 2.0385,
+      "step": 13120500
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 3.1016888488792144e-05,
+      "loss": 2.0654,
+      "step": 13121000
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 3.101616484114487e-05,
+      "loss": 2.0788,
+      "step": 13121500
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 3.101544119349759e-05,
+      "loss": 2.063,
+      "step": 13122000
+    },
+    {
+      "epoch": 37.98,
+      "learning_rate": 3.101471754585032e-05,
+      "loss": 2.0616,
+      "step": 13122500
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 3.101399389820304e-05,
+      "loss": 2.0683,
+      "step": 13123000
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 3.101327025055576e-05,
+      "loss": 2.056,
+      "step": 13123500
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 3.1012546602908485e-05,
+      "loss": 2.0578,
+      "step": 13124000
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 3.101182440255651e-05,
+      "loss": 2.0557,
+      "step": 13124500
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 3.101110075490923e-05,
+      "loss": 2.0658,
+      "step": 13125000
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 3.101037710726195e-05,
+      "loss": 2.0562,
+      "step": 13125500
+    },
+    {
+      "epoch": 37.99,
+      "learning_rate": 3.1009653459614674e-05,
+      "loss": 2.0643,
+      "step": 13126000
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 3.1008929811967396e-05,
+      "loss": 2.0631,
+      "step": 13126500
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 3.100820761161541e-05,
+      "loss": 2.0552,
+      "step": 13127000
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 3.1007483963968134e-05,
+      "loss": 2.068,
+      "step": 13127500
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.6700079034069664,
+      "eval_accuracy_mlm": 0.6351199675846458,
+      "eval_accuracy_nsp": 0.857055450908141,
+      "eval_loss": 2.176142454147339,
+      "eval_runtime": 331.6481,
+      "eval_samples_per_second": 1315.81,
+      "eval_steps_per_second": 54.826,
+      "step": 13127936
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 3.100676031632086e-05,
+      "loss": 2.0675,
+      "step": 13128000
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 3.1006036668673585e-05,
+      "loss": 2.0444,
+      "step": 13128500
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 3.100531302102631e-05,
+      "loss": 2.0409,
+      "step": 13129000
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 3.100459082067432e-05,
+      "loss": 2.0686,
+      "step": 13129500
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 3.1003867173027045e-05,
+      "loss": 2.0486,
+      "step": 13130000
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 3.1003143525379774e-05,
+      "loss": 2.0344,
+      "step": 13130500
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 3.1002419877732496e-05,
+      "loss": 2.0368,
+      "step": 13131000
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 3.100169623008522e-05,
+      "loss": 2.0158,
+      "step": 13131500
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 3.100097258243794e-05,
+      "loss": 2.0411,
+      "step": 13132000
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 3.100025038208596e-05,
+      "loss": 2.0327,
+      "step": 13132500
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 3.0999526734438685e-05,
+      "loss": 2.0615,
+      "step": 13133000
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 3.099880308679141e-05,
+      "loss": 2.03,
+      "step": 13133500
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 3.099807943914413e-05,
+      "loss": 2.0321,
+      "step": 13134000
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 3.099735579149685e-05,
+      "loss": 2.0642,
+      "step": 13134500
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 3.0996632143849574e-05,
+      "loss": 2.0562,
+      "step": 13135000
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 3.0995908496202296e-05,
+      "loss": 2.0545,
+      "step": 13135500
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 3.099518629585031e-05,
+      "loss": 2.038,
+      "step": 13136000
+    },
+    {
+      "epoch": 38.02,
+      "learning_rate": 3.099446264820304e-05,
+      "loss": 2.0516,
+      "step": 13136500
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 3.099373900055576e-05,
+      "loss": 2.0788,
+      "step": 13137000
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 3.099301535290849e-05,
+      "loss": 2.0427,
+      "step": 13137500
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 3.099229315255651e-05,
+      "loss": 2.0485,
+      "step": 13138000
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 3.099156950490923e-05,
+      "loss": 2.0627,
+      "step": 13138500
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 3.099084585726195e-05,
+      "loss": 2.0255,
+      "step": 13139000
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 3.0990122209614674e-05,
+      "loss": 2.0424,
+      "step": 13139500
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 3.09893985619674e-05,
+      "loss": 2.039,
+      "step": 13140000
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 3.098867491432012e-05,
+      "loss": 2.0338,
+      "step": 13140500
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 3.098795126667284e-05,
+      "loss": 2.051,
+      "step": 13141000
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 3.098722761902556e-05,
+      "loss": 2.0356,
+      "step": 13141500
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 3.098650397137829e-05,
+      "loss": 2.0686,
+      "step": 13142000
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 3.0985780323731015e-05,
+      "loss": 2.0545,
+      "step": 13142500
+    },
+    {
+      "epoch": 38.04,
+      "learning_rate": 3.098505667608374e-05,
+      "loss": 2.0543,
+      "step": 13143000
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 3.098433302843646e-05,
+      "loss": 2.0651,
+      "step": 13143500
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 3.098360938078918e-05,
+      "loss": 2.0674,
+      "step": 13144000
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 3.0982885733141904e-05,
+      "loss": 2.0535,
+      "step": 13144500
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 3.098216208549463e-05,
+      "loss": 2.0751,
+      "step": 13145000
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 3.098143988514265e-05,
+      "loss": 2.0583,
+      "step": 13145500
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 3.098071623749537e-05,
+      "loss": 2.0376,
+      "step": 13146000
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 3.097999258984809e-05,
+      "loss": 2.0326,
+      "step": 13146500
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 3.0979268942200815e-05,
+      "loss": 2.0381,
+      "step": 13147000
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 3.097854529455354e-05,
+      "loss": 2.0569,
+      "step": 13147500
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 3.0977821646906266e-05,
+      "loss": 2.0758,
+      "step": 13148000
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 3.097709799925899e-05,
+      "loss": 2.0258,
+      "step": 13148500
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 3.097637435161171e-05,
+      "loss": 2.076,
+      "step": 13149000
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 3.097565070396443e-05,
+      "loss": 2.0588,
+      "step": 13149500
+    },
+    {
+      "epoch": 38.06,
+      "learning_rate": 3.0974927056317155e-05,
+      "loss": 2.0564,
+      "step": 13150000
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 3.097420485596517e-05,
+      "loss": 2.0389,
+      "step": 13150500
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 3.097348265561319e-05,
+      "loss": 2.0427,
+      "step": 13151000
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 3.0972759007965915e-05,
+      "loss": 2.0358,
+      "step": 13151500
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 3.097203536031864e-05,
+      "loss": 2.0283,
+      "step": 13152000
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 3.097131315996666e-05,
+      "loss": 2.0518,
+      "step": 13152500
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 3.097058951231938e-05,
+      "loss": 2.0432,
+      "step": 13153000
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 3.0969865864672104e-05,
+      "loss": 2.0553,
+      "step": 13153500
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 3.0969142217024826e-05,
+      "loss": 2.0477,
+      "step": 13154000
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 3.096842001667284e-05,
+      "loss": 2.0603,
+      "step": 13154500
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 3.0967696369025564e-05,
+      "loss": 2.0692,
+      "step": 13155000
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 3.096697272137829e-05,
+      "loss": 2.0532,
+      "step": 13155500
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 3.0966249073731015e-05,
+      "loss": 2.0524,
+      "step": 13156000
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 3.096552542608374e-05,
+      "loss": 2.0634,
+      "step": 13156500
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 3.096480177843646e-05,
+      "loss": 2.0622,
+      "step": 13157000
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 3.0964079578084475e-05,
+      "loss": 2.0458,
+      "step": 13157500
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 3.09633559304372e-05,
+      "loss": 2.0406,
+      "step": 13158000
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 3.096263228278992e-05,
+      "loss": 2.0522,
+      "step": 13158500
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 3.096190863514264e-05,
+      "loss": 2.0351,
+      "step": 13159000
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 3.0961184987495364e-05,
+      "loss": 2.0601,
+      "step": 13159500
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 3.096046133984809e-05,
+      "loss": 2.0637,
+      "step": 13160000
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 3.0959737692200816e-05,
+      "loss": 2.0692,
+      "step": 13160500
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 3.0959014044553545e-05,
+      "loss": 2.039,
+      "step": 13161000
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 3.095829039690627e-05,
+      "loss": 2.041,
+      "step": 13161500
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 3.095756674925899e-05,
+      "loss": 2.0246,
+      "step": 13162000
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 3.095684310161171e-05,
+      "loss": 2.0709,
+      "step": 13162500
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 3.095612090125973e-05,
+      "loss": 2.0565,
+      "step": 13163000
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 3.095539725361245e-05,
+      "loss": 2.0677,
+      "step": 13163500
+    },
+    {
+      "epoch": 38.1,
+      "learning_rate": 3.095467360596517e-05,
+      "loss": 2.0613,
+      "step": 13164000
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 3.0953949958317894e-05,
+      "loss": 2.0654,
+      "step": 13164500
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 3.0953226310670616e-05,
+      "loss": 2.0533,
+      "step": 13165000
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 3.0952502663023345e-05,
+      "loss": 2.0604,
+      "step": 13165500
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 3.095177901537607e-05,
+      "loss": 2.0331,
+      "step": 13166000
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 3.095105536772879e-05,
+      "loss": 2.0613,
+      "step": 13166500
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 3.095033172008152e-05,
+      "loss": 2.0579,
+      "step": 13167000
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 3.0949609519729534e-05,
+      "loss": 2.0455,
+      "step": 13167500
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 3.0948885872082256e-05,
+      "loss": 2.0396,
+      "step": 13168000
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 3.094816222443498e-05,
+      "loss": 2.071,
+      "step": 13168500
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 3.09474385767877e-05,
+      "loss": 2.0295,
+      "step": 13169000
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 3.094671492914042e-05,
+      "loss": 2.0484,
+      "step": 13169500
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 3.0945991281493145e-05,
+      "loss": 2.0289,
+      "step": 13170000
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 3.094526763384587e-05,
+      "loss": 2.0656,
+      "step": 13170500
+    },
+    {
+      "epoch": 38.12,
+      "learning_rate": 3.0944543986198596e-05,
+      "loss": 2.0672,
+      "step": 13171000
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 3.094382178584661e-05,
+      "loss": 2.04,
+      "step": 13171500
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 3.0943098138199334e-05,
+      "loss": 2.0497,
+      "step": 13172000
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 3.0942374490552056e-05,
+      "loss": 2.0488,
+      "step": 13172500
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 3.094165084290478e-05,
+      "loss": 2.0531,
+      "step": 13173000
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 3.0940928642552794e-05,
+      "loss": 2.0341,
+      "step": 13173500
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 3.0940206442200816e-05,
+      "loss": 2.0768,
+      "step": 13174000
+    },
+    {
+      "epoch": 38.13,
+      "learning_rate": 3.0939482794553545e-05,
+      "loss": 2.0569,
+      "step": 13174500
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 3.093875914690627e-05,
+      "loss": 2.0517,
+      "step": 13175000
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 3.093803549925899e-05,
+      "loss": 2.0544,
+      "step": 13175500
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 3.093731185161171e-05,
+      "loss": 2.0465,
+      "step": 13176000
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 3.0936588203964434e-05,
+      "loss": 2.0513,
+      "step": 13176500
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 3.0935864556317157e-05,
+      "loss": 2.0594,
+      "step": 13177000
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 3.093514090866988e-05,
+      "loss": 2.0627,
+      "step": 13177500
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 3.09344172610226e-05,
+      "loss": 2.0459,
+      "step": 13178000
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 3.093369361337532e-05,
+      "loss": 2.0642,
+      "step": 13178500
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 3.0932969965728045e-05,
+      "loss": 2.05,
+      "step": 13179000
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 3.093224631808077e-05,
+      "loss": 2.0168,
+      "step": 13179500
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 3.09315226704335e-05,
+      "loss": 2.0754,
+      "step": 13180000
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 3.093079902278622e-05,
+      "loss": 2.0535,
+      "step": 13180500
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 3.0930076822434235e-05,
+      "loss": 2.0654,
+      "step": 13181000
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 3.092935317478696e-05,
+      "loss": 2.0595,
+      "step": 13181500
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 3.0928629527139686e-05,
+      "loss": 2.0687,
+      "step": 13182000
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 3.092790587949241e-05,
+      "loss": 2.0496,
+      "step": 13182500
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 3.0927183679140424e-05,
+      "loss": 2.0673,
+      "step": 13183000
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 3.0926461478788446e-05,
+      "loss": 2.036,
+      "step": 13183500
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 3.092573783114117e-05,
+      "loss": 2.0326,
+      "step": 13184000
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 3.092501418349389e-05,
+      "loss": 2.0425,
+      "step": 13184500
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 3.092429053584661e-05,
+      "loss": 2.0667,
+      "step": 13185000
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 3.0923566888199335e-05,
+      "loss": 2.0627,
+      "step": 13185500
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 3.092284324055206e-05,
+      "loss": 2.0512,
+      "step": 13186000
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 3.092211959290478e-05,
+      "loss": 2.0288,
+      "step": 13186500
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 3.09213959452575e-05,
+      "loss": 2.0557,
+      "step": 13187000
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 3.0920672297610224e-05,
+      "loss": 2.0422,
+      "step": 13187500
+    },
+    {
+      "epoch": 38.17,
+      "learning_rate": 3.0919948649962946e-05,
+      "loss": 2.0551,
+      "step": 13188000
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 3.091922500231567e-05,
+      "loss": 2.0567,
+      "step": 13188500
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 3.091850280196369e-05,
+      "loss": 2.0354,
+      "step": 13189000
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 3.091777915431642e-05,
+      "loss": 2.0652,
+      "step": 13189500
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 3.0917056953964435e-05,
+      "loss": 2.0318,
+      "step": 13190000
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 3.091633330631716e-05,
+      "loss": 2.0602,
+      "step": 13190500
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 3.091560965866988e-05,
+      "loss": 2.063,
+      "step": 13191000
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 3.09148860110226e-05,
+      "loss": 2.0532,
+      "step": 13191500
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 3.0914162363375324e-05,
+      "loss": 2.0853,
+      "step": 13192000
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 3.0913440163023346e-05,
+      "loss": 2.0645,
+      "step": 13192500
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 3.091271796267136e-05,
+      "loss": 2.0321,
+      "step": 13193000
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 3.0911994315024084e-05,
+      "loss": 2.0609,
+      "step": 13193500
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 3.0911270667376806e-05,
+      "loss": 2.0393,
+      "step": 13194000
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 3.091054701972953e-05,
+      "loss": 2.0503,
+      "step": 13194500
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 3.090982337208225e-05,
+      "loss": 2.0394,
+      "step": 13195000
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 3.090909972443497e-05,
+      "loss": 2.0653,
+      "step": 13195500
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 3.0908377524082995e-05,
+      "loss": 2.0732,
+      "step": 13196000
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 3.090765387643572e-05,
+      "loss": 2.0613,
+      "step": 13196500
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 3.090693022878845e-05,
+      "loss": 2.0329,
+      "step": 13197000
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 3.090620802843646e-05,
+      "loss": 2.0427,
+      "step": 13197500
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 3.0905484380789184e-05,
+      "loss": 2.0611,
+      "step": 13198000
+    },
+    {
+      "epoch": 38.2,
+      "learning_rate": 3.090476073314191e-05,
+      "loss": 2.064,
+      "step": 13198500
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 3.090403708549463e-05,
+      "loss": 2.032,
+      "step": 13199000
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 3.090331343784735e-05,
+      "loss": 2.027,
+      "step": 13199500
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 3.0902589790200073e-05,
+      "loss": 2.0515,
+      "step": 13200000
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 3.0901866142552796e-05,
+      "loss": 2.0472,
+      "step": 13200500
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 3.0901142494905525e-05,
+      "loss": 2.0527,
+      "step": 13201000
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 3.090041884725825e-05,
+      "loss": 2.0758,
+      "step": 13201500
+    },
+    {
+      "epoch": 38.21,
+      "learning_rate": 3.0899698094201556e-05,
+      "loss": 2.0322,
+      "step": 13202000
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 3.089897444655428e-05,
+      "loss": 2.085,
+      "step": 13202500
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 3.0898250798907e-05,
+      "loss": 2.0562,
+      "step": 13203000
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 3.089752715125972e-05,
+      "loss": 2.0361,
+      "step": 13203500
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 3.089680350361245e-05,
+      "loss": 2.0541,
+      "step": 13204000
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 3.0896079855965174e-05,
+      "loss": 2.0667,
+      "step": 13204500
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 3.08953562083179e-05,
+      "loss": 2.0458,
+      "step": 13205000
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 3.0894632560670625e-05,
+      "loss": 2.0516,
+      "step": 13205500
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 3.089390891302335e-05,
+      "loss": 2.0569,
+      "step": 13206000
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 3.089318526537607e-05,
+      "loss": 2.0643,
+      "step": 13206500
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 3.089246161772879e-05,
+      "loss": 2.0314,
+      "step": 13207000
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 3.0891737970081514e-05,
+      "loss": 2.0496,
+      "step": 13207500
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 3.0891014322434236e-05,
+      "loss": 2.0572,
+      "step": 13208000
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 3.089029067478696e-05,
+      "loss": 2.0461,
+      "step": 13208500
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 3.088956702713968e-05,
+      "loss": 2.0688,
+      "step": 13209000
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 3.08888433794924e-05,
+      "loss": 2.0458,
+      "step": 13209500
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 3.0888119731845125e-05,
+      "loss": 2.0616,
+      "step": 13210000
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 3.088739897878844e-05,
+      "loss": 2.0691,
+      "step": 13210500
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 3.088667533114116e-05,
+      "loss": 2.0724,
+      "step": 13211000
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 3.0885951683493885e-05,
+      "loss": 2.0821,
+      "step": 13211500
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 3.08852294831419e-05,
+      "loss": 2.0733,
+      "step": 13212000
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 3.088450583549463e-05,
+      "loss": 2.0852,
+      "step": 13212500
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 3.088378218784735e-05,
+      "loss": 2.0623,
+      "step": 13213000
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 3.0883059987495374e-05,
+      "loss": 2.062,
+      "step": 13213500
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 3.0882336339848097e-05,
+      "loss": 2.0368,
+      "step": 13214000
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 3.088161269220082e-05,
+      "loss": 2.0602,
+      "step": 13214500
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 3.088088904455354e-05,
+      "loss": 2.0582,
+      "step": 13215000
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 3.088016539690626e-05,
+      "loss": 2.0544,
+      "step": 13215500
+    },
+    {
+      "epoch": 38.25,
+      "learning_rate": 3.0879441749258985e-05,
+      "loss": 2.0493,
+      "step": 13216000
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 3.087871810161171e-05,
+      "loss": 2.0494,
+      "step": 13216500
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 3.087799445396443e-05,
+      "loss": 2.0367,
+      "step": 13217000
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 3.087727080631715e-05,
+      "loss": 2.0437,
+      "step": 13217500
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 3.0876547158669874e-05,
+      "loss": 2.0385,
+      "step": 13218000
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 3.08758249583179e-05,
+      "loss": 2.0361,
+      "step": 13218500
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 3.087510131067062e-05,
+      "loss": 2.0354,
+      "step": 13219000
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 3.087437766302335e-05,
+      "loss": 2.0591,
+      "step": 13219500
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 3.0873655462671364e-05,
+      "loss": 2.0355,
+      "step": 13220000
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 3.0872931815024086e-05,
+      "loss": 2.0462,
+      "step": 13220500
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 3.087220816737681e-05,
+      "loss": 2.0587,
+      "step": 13221000
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 3.087148451972953e-05,
+      "loss": 2.0351,
+      "step": 13221500
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 3.087076087208225e-05,
+      "loss": 2.0438,
+      "step": 13222000
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 3.0870037224434975e-05,
+      "loss": 2.0852,
+      "step": 13222500
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 3.0869313576787704e-05,
+      "loss": 2.064,
+      "step": 13223000
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 3.0868589929140426e-05,
+      "loss": 2.0496,
+      "step": 13223500
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 3.086786628149315e-05,
+      "loss": 2.063,
+      "step": 13224000
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 3.086714263384587e-05,
+      "loss": 2.0544,
+      "step": 13224500
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 3.0866420433493886e-05,
+      "loss": 2.0686,
+      "step": 13225000
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 3.086569678584661e-05,
+      "loss": 2.0512,
+      "step": 13225500
+    },
+    {
+      "epoch": 38.28,
+      "learning_rate": 3.086497313819933e-05,
+      "loss": 2.0555,
+      "step": 13226000
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 3.086424949055205e-05,
+      "loss": 2.0608,
+      "step": 13226500
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 3.086352584290478e-05,
+      "loss": 2.0683,
+      "step": 13227000
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 3.0862803642552804e-05,
+      "loss": 2.0672,
+      "step": 13227500
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 3.0862079994905526e-05,
+      "loss": 2.0568,
+      "step": 13228000
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 3.086135634725825e-05,
+      "loss": 2.0552,
+      "step": 13228500
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 3.086063269961097e-05,
+      "loss": 2.0666,
+      "step": 13229000
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 3.085990905196369e-05,
+      "loss": 2.0701,
+      "step": 13229500
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 3.0859185404316415e-05,
+      "loss": 2.0415,
+      "step": 13230000
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 3.085846175666914e-05,
+      "loss": 2.0781,
+      "step": 13230500
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 3.085773810902186e-05,
+      "loss": 2.0327,
+      "step": 13231000
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 3.085701446137458e-05,
+      "loss": 2.0478,
+      "step": 13231500
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 3.0856292261022604e-05,
+      "loss": 2.0515,
+      "step": 13232000
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 3.0855568613375326e-05,
+      "loss": 2.0453,
+      "step": 13232500
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 3.085484496572805e-05,
+      "loss": 2.0562,
+      "step": 13233000
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 3.085412131808077e-05,
+      "loss": 2.0693,
+      "step": 13233500
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 3.085339767043349e-05,
+      "loss": 2.05,
+      "step": 13234000
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 3.0852675470081515e-05,
+      "loss": 2.0636,
+      "step": 13234500
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 3.085195182243424e-05,
+      "loss": 2.0556,
+      "step": 13235000
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 3.085122817478696e-05,
+      "loss": 2.0486,
+      "step": 13235500
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 3.085050597443498e-05,
+      "loss": 2.0879,
+      "step": 13236000
+    },
+    {
+      "epoch": 38.31,
+      "learning_rate": 3.0849782326787704e-05,
+      "loss": 2.0543,
+      "step": 13236500
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 3.084905867914043e-05,
+      "loss": 2.0437,
+      "step": 13237000
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 3.084833503149315e-05,
+      "loss": 2.0697,
+      "step": 13237500
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 3.084761138384587e-05,
+      "loss": 2.0279,
+      "step": 13238000
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 3.0846887736198593e-05,
+      "loss": 2.0518,
+      "step": 13238500
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 3.0846164088551316e-05,
+      "loss": 2.0708,
+      "step": 13239000
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 3.084544044090404e-05,
+      "loss": 2.0497,
+      "step": 13239500
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 3.0844718240552053e-05,
+      "loss": 2.0397,
+      "step": 13240000
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 3.084399459290478e-05,
+      "loss": 2.0493,
+      "step": 13240500
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 3.0843270945257505e-05,
+      "loss": 2.0672,
+      "step": 13241000
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 3.084254729761023e-05,
+      "loss": 2.0457,
+      "step": 13241500
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 3.0841823649962956e-05,
+      "loss": 2.0482,
+      "step": 13242000
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 3.084110000231568e-05,
+      "loss": 2.0619,
+      "step": 13242500
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 3.08403763546684e-05,
+      "loss": 2.0684,
+      "step": 13243000
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 3.083965270702112e-05,
+      "loss": 2.0272,
+      "step": 13243500
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 3.0838929059373845e-05,
+      "loss": 2.0715,
+      "step": 13244000
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 3.083820541172657e-05,
+      "loss": 2.0691,
+      "step": 13244500
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 3.083748176407929e-05,
+      "loss": 2.0417,
+      "step": 13245000
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 3.0836759563727305e-05,
+      "loss": 2.0654,
+      "step": 13245500
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 3.083603591608003e-05,
+      "loss": 2.0686,
+      "step": 13246000
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 3.083531371572805e-05,
+      "loss": 2.0601,
+      "step": 13246500
+    },
+    {
+      "epoch": 38.34,
+      "learning_rate": 3.083459006808077e-05,
+      "loss": 2.0734,
+      "step": 13247000
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 3.0833866420433494e-05,
+      "loss": 2.0459,
+      "step": 13247500
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 3.0833142772786216e-05,
+      "loss": 2.0708,
+      "step": 13248000
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 3.083242057243423e-05,
+      "loss": 2.0681,
+      "step": 13248500
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 3.0831696924786954e-05,
+      "loss": 2.0499,
+      "step": 13249000
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 3.083097327713968e-05,
+      "loss": 2.035,
+      "step": 13249500
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 3.0830249629492405e-05,
+      "loss": 2.0272,
+      "step": 13250000
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 3.0829525981845134e-05,
+      "loss": 2.0333,
+      "step": 13250500
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 3.0828802334197856e-05,
+      "loss": 2.0619,
+      "step": 13251000
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 3.082807868655058e-05,
+      "loss": 2.0427,
+      "step": 13251500
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 3.0827356486198594e-05,
+      "loss": 2.0532,
+      "step": 13252000
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 3.0826632838551316e-05,
+      "loss": 2.0505,
+      "step": 13252500
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 3.082590919090404e-05,
+      "loss": 2.0603,
+      "step": 13253000
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 3.0825188437847354e-05,
+      "loss": 2.0683,
+      "step": 13253500
+    },
+    {
+      "epoch": 38.36,
+      "learning_rate": 3.0824464790200077e-05,
+      "loss": 2.0352,
+      "step": 13254000
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 3.08237411425528e-05,
+      "loss": 2.049,
+      "step": 13254500
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 3.082301749490552e-05,
+      "loss": 2.052,
+      "step": 13255000
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 3.082229384725824e-05,
+      "loss": 2.0657,
+      "step": 13255500
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 3.0821570199610966e-05,
+      "loss": 2.0643,
+      "step": 13256000
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 3.082084655196369e-05,
+      "loss": 2.0702,
+      "step": 13256500
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 3.082012290431642e-05,
+      "loss": 2.0555,
+      "step": 13257000
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 3.081939925666914e-05,
+      "loss": 2.0724,
+      "step": 13257500
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 3.081867560902186e-05,
+      "loss": 2.0567,
+      "step": 13258000
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 3.0817951961374583e-05,
+      "loss": 2.0504,
+      "step": 13258500
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 3.0817228313727306e-05,
+      "loss": 2.0539,
+      "step": 13259000
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 3.0816504666080035e-05,
+      "loss": 2.0519,
+      "step": 13259500
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 3.081578101843276e-05,
+      "loss": 2.0304,
+      "step": 13260000
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 3.081505737078548e-05,
+      "loss": 2.0753,
+      "step": 13260500
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 3.08143337231382e-05,
+      "loss": 2.0772,
+      "step": 13261000
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 3.0813610075490924e-05,
+      "loss": 2.0915,
+      "step": 13261500
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 3.0812886427843646e-05,
+      "loss": 2.0526,
+      "step": 13262000
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 3.081216278019637e-05,
+      "loss": 2.0552,
+      "step": 13262500
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 3.081143913254909e-05,
+      "loss": 2.0381,
+      "step": 13263000
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 3.081071548490181e-05,
+      "loss": 2.0515,
+      "step": 13263500
+    },
+    {
+      "epoch": 38.39,
+      "learning_rate": 3.080999183725454e-05,
+      "loss": 2.0301,
+      "step": 13264000
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 3.080926963690256e-05,
+      "loss": 2.0425,
+      "step": 13264500
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 3.0808545989255286e-05,
+      "loss": 2.0313,
+      "step": 13265000
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 3.080782234160801e-05,
+      "loss": 2.0483,
+      "step": 13265500
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 3.080709869396073e-05,
+      "loss": 2.0788,
+      "step": 13266000
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 3.080637504631345e-05,
+      "loss": 2.0464,
+      "step": 13266500
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 3.0805651398666175e-05,
+      "loss": 2.0619,
+      "step": 13267000
+    },
+    {
+      "epoch": 38.4,
+      "learning_rate": 3.08049277510189e-05,
+      "loss": 2.0564,
+      "step": 13267500
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 3.080420410337162e-05,
+      "loss": 2.0287,
+      "step": 13268000
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 3.080348045572434e-05,
+      "loss": 2.0649,
+      "step": 13268500
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 3.080275970266766e-05,
+      "loss": 2.0546,
+      "step": 13269000
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 3.080203605502038e-05,
+      "loss": 2.0475,
+      "step": 13269500
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 3.08013124073731e-05,
+      "loss": 2.0733,
+      "step": 13270000
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 3.0800588759725824e-05,
+      "loss": 2.0732,
+      "step": 13270500
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 3.0799865112078546e-05,
+      "loss": 2.0585,
+      "step": 13271000
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 3.079914435902186e-05,
+      "loss": 2.0621,
+      "step": 13271500
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 3.0798420711374584e-05,
+      "loss": 2.0497,
+      "step": 13272000
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 3.079769706372731e-05,
+      "loss": 2.0607,
+      "step": 13272500
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 3.0796973416080035e-05,
+      "loss": 2.0391,
+      "step": 13273000
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 3.079624976843276e-05,
+      "loss": 2.08,
+      "step": 13273500
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 3.079552612078548e-05,
+      "loss": 2.0456,
+      "step": 13274000
+    },
+    {
+      "epoch": 38.42,
+      "learning_rate": 3.07948024731382e-05,
+      "loss": 2.0453,
+      "step": 13274500
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 3.0794078825490924e-05,
+      "loss": 2.0495,
+      "step": 13275000
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 3.079335662513894e-05,
+      "loss": 2.0519,
+      "step": 13275500
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 3.079263297749166e-05,
+      "loss": 2.0603,
+      "step": 13276000
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 3.0791910777139685e-05,
+      "loss": 2.0596,
+      "step": 13276500
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 3.079118712949241e-05,
+      "loss": 2.0757,
+      "step": 13277000
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 3.079046348184513e-05,
+      "loss": 2.0742,
+      "step": 13277500
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 3.078973983419785e-05,
+      "loss": 2.0715,
+      "step": 13278000
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 3.0789016186550573e-05,
+      "loss": 2.0789,
+      "step": 13278500
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 3.07882925389033e-05,
+      "loss": 2.0439,
+      "step": 13279000
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 3.0787568891256025e-05,
+      "loss": 2.0533,
+      "step": 13279500
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 3.078684524360875e-05,
+      "loss": 2.0731,
+      "step": 13280000
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 3.078612159596147e-05,
+      "loss": 2.0751,
+      "step": 13280500
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 3.078539794831419e-05,
+      "loss": 2.0428,
+      "step": 13281000
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 3.0784674300666914e-05,
+      "loss": 2.0797,
+      "step": 13281500
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 3.0783950653019636e-05,
+      "loss": 2.08,
+      "step": 13282000
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 3.078322845266766e-05,
+      "loss": 2.0331,
+      "step": 13282500
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 3.078250480502038e-05,
+      "loss": 2.0424,
+      "step": 13283000
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 3.07817811573731e-05,
+      "loss": 2.0764,
+      "step": 13283500
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 3.078105895702112e-05,
+      "loss": 2.064,
+      "step": 13284000
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 3.078033530937384e-05,
+      "loss": 2.069,
+      "step": 13284500
+    },
+    {
+      "epoch": 38.45,
+      "learning_rate": 3.077961166172656e-05,
+      "loss": 2.0596,
+      "step": 13285000
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 3.0778888014079285e-05,
+      "loss": 2.0456,
+      "step": 13285500
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 3.0778164366432014e-05,
+      "loss": 2.0444,
+      "step": 13286000
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 3.0777440718784736e-05,
+      "loss": 2.0559,
+      "step": 13286500
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 3.0776717071137465e-05,
+      "loss": 2.0447,
+      "step": 13287000
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 3.077599342349019e-05,
+      "loss": 2.0494,
+      "step": 13287500
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 3.0775272670433496e-05,
+      "loss": 2.0521,
+      "step": 13288000
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 3.077454902278622e-05,
+      "loss": 2.0433,
+      "step": 13288500
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 3.077382537513894e-05,
+      "loss": 2.0397,
+      "step": 13289000
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 3.077310172749166e-05,
+      "loss": 2.0406,
+      "step": 13289500
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 3.0772378079844385e-05,
+      "loss": 2.0451,
+      "step": 13290000
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 3.077165587949241e-05,
+      "loss": 2.0591,
+      "step": 13290500
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 3.077093223184513e-05,
+      "loss": 2.049,
+      "step": 13291000
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 3.077020858419785e-05,
+      "loss": 2.0776,
+      "step": 13291500
+    },
+    {
+      "epoch": 38.47,
+      "learning_rate": 3.0769484936550574e-05,
+      "loss": 2.0652,
+      "step": 13292000
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 3.0768761288903297e-05,
+      "loss": 2.0747,
+      "step": 13292500
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 3.076803764125602e-05,
+      "loss": 2.056,
+      "step": 13293000
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 3.076731399360874e-05,
+      "loss": 2.0539,
+      "step": 13293500
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 3.076659034596147e-05,
+      "loss": 2.05,
+      "step": 13294000
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 3.0765868145609486e-05,
+      "loss": 2.0765,
+      "step": 13294500
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 3.0765144497962215e-05,
+      "loss": 2.0531,
+      "step": 13295000
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 3.076442085031494e-05,
+      "loss": 2.0293,
+      "step": 13295500
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 3.076369720266766e-05,
+      "loss": 2.0571,
+      "step": 13296000
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 3.076297355502038e-05,
+      "loss": 2.0532,
+      "step": 13296500
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 3.0762249907373103e-05,
+      "loss": 2.0564,
+      "step": 13297000
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 3.076152770702112e-05,
+      "loss": 2.0338,
+      "step": 13297500
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 3.076080405937384e-05,
+      "loss": 2.0317,
+      "step": 13298000
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 3.0760080411726564e-05,
+      "loss": 2.052,
+      "step": 13298500
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 3.0759356764079286e-05,
+      "loss": 2.0607,
+      "step": 13299000
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 3.0758633116432015e-05,
+      "loss": 2.0648,
+      "step": 13299500
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 3.075790946878474e-05,
+      "loss": 2.0747,
+      "step": 13300000
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 3.075718726843275e-05,
+      "loss": 2.0699,
+      "step": 13300500
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 3.0756463620785475e-05,
+      "loss": 2.0846,
+      "step": 13301000
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 3.0755739973138204e-05,
+      "loss": 2.059,
+      "step": 13301500
+    },
+    {
+      "epoch": 38.5,
+      "learning_rate": 3.0755016325490926e-05,
+      "loss": 2.09,
+      "step": 13302000
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 3.075429267784365e-05,
+      "loss": 2.0614,
+      "step": 13302500
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 3.075356903019637e-05,
+      "loss": 2.0398,
+      "step": 13303000
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 3.075284538254909e-05,
+      "loss": 2.0585,
+      "step": 13303500
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 3.0752121734901815e-05,
+      "loss": 2.0643,
+      "step": 13304000
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 3.075139808725454e-05,
+      "loss": 2.0609,
+      "step": 13304500
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 3.0750674439607266e-05,
+      "loss": 2.0429,
+      "step": 13305000
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 3.074995079195999e-05,
+      "loss": 2.0607,
+      "step": 13305500
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 3.0749228591608004e-05,
+      "loss": 2.0784,
+      "step": 13306000
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 3.0748504943960726e-05,
+      "loss": 2.0561,
+      "step": 13306500
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 3.074778129631345e-05,
+      "loss": 2.0799,
+      "step": 13307000
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 3.074705764866617e-05,
+      "loss": 2.0671,
+      "step": 13307500
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 3.074633400101889e-05,
+      "loss": 2.06,
+      "step": 13308000
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 3.074561035337162e-05,
+      "loss": 2.0535,
+      "step": 13308500
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 3.074488815301964e-05,
+      "loss": 2.0417,
+      "step": 13309000
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 3.0744164505372367e-05,
+      "loss": 2.0841,
+      "step": 13309500
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 3.074344085772509e-05,
+      "loss": 2.0737,
+      "step": 13310000
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 3.074271721007781e-05,
+      "loss": 2.0857,
+      "step": 13310500
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 3.074199356243053e-05,
+      "loss": 2.0534,
+      "step": 13311000
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 3.074127136207855e-05,
+      "loss": 2.0517,
+      "step": 13311500
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 3.074054771443127e-05,
+      "loss": 2.0632,
+      "step": 13312000
+    },
+    {
+      "epoch": 38.53,
+      "learning_rate": 3.073982406678399e-05,
+      "loss": 2.0564,
+      "step": 13312500
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 3.0739100419136715e-05,
+      "loss": 2.0527,
+      "step": 13313000
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 3.073837677148944e-05,
+      "loss": 2.0491,
+      "step": 13313500
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 3.073765312384217e-05,
+      "loss": 2.0512,
+      "step": 13314000
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 3.073692947619489e-05,
+      "loss": 2.044,
+      "step": 13314500
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 3.0736207275842904e-05,
+      "loss": 2.0395,
+      "step": 13315000
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 3.073548507549092e-05,
+      "loss": 2.0561,
+      "step": 13315500
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 3.073476287513894e-05,
+      "loss": 2.0429,
+      "step": 13316000
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 3.0734039227491665e-05,
+      "loss": 2.0684,
+      "step": 13316500
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 3.0733315579844394e-05,
+      "loss": 2.0406,
+      "step": 13317000
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 3.0732591932197116e-05,
+      "loss": 2.058,
+      "step": 13317500
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 3.073186828454984e-05,
+      "loss": 2.056,
+      "step": 13318000
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 3.073114463690256e-05,
+      "loss": 2.059,
+      "step": 13318500
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 3.0730422436550576e-05,
+      "loss": 2.0307,
+      "step": 13319000
+    },
+    {
+      "epoch": 38.55,
+      "learning_rate": 3.07296987889033e-05,
+      "loss": 2.0742,
+      "step": 13319500
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 3.072897514125602e-05,
+      "loss": 2.0551,
+      "step": 13320000
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 3.072825149360874e-05,
+      "loss": 2.0496,
+      "step": 13320500
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 3.0727527845961465e-05,
+      "loss": 2.0521,
+      "step": 13321000
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 3.0726804198314194e-05,
+      "loss": 2.0481,
+      "step": 13321500
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 3.0726080550666916e-05,
+      "loss": 2.0861,
+      "step": 13322000
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 3.072535835031493e-05,
+      "loss": 2.0653,
+      "step": 13322500
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 3.0724634702667654e-05,
+      "loss": 2.0816,
+      "step": 13323000
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 3.0723911055020376e-05,
+      "loss": 2.0799,
+      "step": 13323500
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 3.0723187407373105e-05,
+      "loss": 2.0555,
+      "step": 13324000
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 3.072246520702112e-05,
+      "loss": 2.0456,
+      "step": 13324500
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 3.072174155937384e-05,
+      "loss": 2.052,
+      "step": 13325000
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 3.0721017911726565e-05,
+      "loss": 2.0454,
+      "step": 13325500
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 3.0720294264079294e-05,
+      "loss": 2.068,
+      "step": 13326000
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 3.0719570616432016e-05,
+      "loss": 2.0844,
+      "step": 13326500
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 3.071884696878474e-05,
+      "loss": 2.075,
+      "step": 13327000
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 3.071812332113746e-05,
+      "loss": 2.0452,
+      "step": 13327500
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 3.071739967349018e-05,
+      "loss": 2.042,
+      "step": 13328000
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 3.0716676025842905e-05,
+      "loss": 2.0775,
+      "step": 13328500
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 3.071595382549092e-05,
+      "loss": 2.0476,
+      "step": 13329000
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 3.071523017784364e-05,
+      "loss": 2.0604,
+      "step": 13329500
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 3.0714506530196365e-05,
+      "loss": 2.0683,
+      "step": 13330000
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 3.0713782882549094e-05,
+      "loss": 2.0844,
+      "step": 13330500
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 3.0713059234901817e-05,
+      "loss": 2.0661,
+      "step": 13331000
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 3.0712335587254546e-05,
+      "loss": 2.0787,
+      "step": 13331500
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 3.071161338690256e-05,
+      "loss": 2.0716,
+      "step": 13332000
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 3.071089118655058e-05,
+      "loss": 2.0437,
+      "step": 13332500
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 3.07101675389033e-05,
+      "loss": 2.0636,
+      "step": 13333000
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 3.070944389125602e-05,
+      "loss": 2.0619,
+      "step": 13333500
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 3.070872024360874e-05,
+      "loss": 2.0577,
+      "step": 13334000
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 3.070799659596147e-05,
+      "loss": 2.0594,
+      "step": 13334500
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 3.0707272948314195e-05,
+      "loss": 2.0463,
+      "step": 13335000
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 3.070654930066692e-05,
+      "loss": 2.0843,
+      "step": 13335500
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 3.070582565301964e-05,
+      "loss": 2.0713,
+      "step": 13336000
+    },
+    {
+      "epoch": 38.6,
+      "learning_rate": 3.070510200537236e-05,
+      "loss": 2.0505,
+      "step": 13336500
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 3.0704378357725084e-05,
+      "loss": 2.0519,
+      "step": 13337000
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 3.0703654710077806e-05,
+      "loss": 2.0589,
+      "step": 13337500
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 3.070293106243053e-05,
+      "loss": 2.0445,
+      "step": 13338000
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 3.070220741478326e-05,
+      "loss": 2.0442,
+      "step": 13338500
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 3.070148376713598e-05,
+      "loss": 2.0602,
+      "step": 13339000
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 3.07007601194887e-05,
+      "loss": 2.068,
+      "step": 13339500
+    },
+    {
+      "epoch": 38.61,
+      "learning_rate": 3.0700036471841424e-05,
+      "loss": 2.0673,
+      "step": 13340000
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 3.0699314271489446e-05,
+      "loss": 2.0824,
+      "step": 13340500
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 3.069859062384217e-05,
+      "loss": 2.0991,
+      "step": 13341000
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 3.069786697619489e-05,
+      "loss": 2.0638,
+      "step": 13341500
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 3.069714332854761e-05,
+      "loss": 2.0481,
+      "step": 13342000
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 3.0696419680900335e-05,
+      "loss": 2.0697,
+      "step": 13342500
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 3.069569603325306e-05,
+      "loss": 2.0409,
+      "step": 13343000
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 3.069497238560578e-05,
+      "loss": 2.0568,
+      "step": 13343500
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 3.06942487379585e-05,
+      "loss": 2.0711,
+      "step": 13344000
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 3.0693525090311224e-05,
+      "loss": 2.0633,
+      "step": 13344500
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 3.0692801442663946e-05,
+      "loss": 2.0586,
+      "step": 13345000
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 3.069207779501667e-05,
+      "loss": 2.0658,
+      "step": 13345500
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 3.06913555946647e-05,
+      "loss": 2.0636,
+      "step": 13346000
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 3.069063194701742e-05,
+      "loss": 2.0782,
+      "step": 13346500
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 3.068990829937014e-05,
+      "loss": 2.0658,
+      "step": 13347000
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 3.0689184651722864e-05,
+      "loss": 2.0522,
+      "step": 13347500
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 3.0688461004075586e-05,
+      "loss": 2.0667,
+      "step": 13348000
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 3.068773735642831e-05,
+      "loss": 2.0709,
+      "step": 13348500
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 3.068701370878103e-05,
+      "loss": 2.0724,
+      "step": 13349000
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 3.068629006113375e-05,
+      "loss": 2.0489,
+      "step": 13349500
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 3.068556786078177e-05,
+      "loss": 2.0258,
+      "step": 13350000
+    },
+    {
+      "epoch": 38.64,
+      "learning_rate": 3.068484566042979e-05,
+      "loss": 2.072,
+      "step": 13350500
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 3.068412201278251e-05,
+      "loss": 2.0442,
+      "step": 13351000
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 3.0683398365135235e-05,
+      "loss": 2.08,
+      "step": 13351500
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 3.068267471748796e-05,
+      "loss": 2.0274,
+      "step": 13352000
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 3.068195106984068e-05,
+      "loss": 2.0581,
+      "step": 13352500
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 3.06812274221934e-05,
+      "loss": 2.0515,
+      "step": 13353000
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 3.0680505221841425e-05,
+      "loss": 2.0643,
+      "step": 13353500
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 3.067978157419415e-05,
+      "loss": 2.0556,
+      "step": 13354000
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 3.0679057926546876e-05,
+      "loss": 2.0379,
+      "step": 13354500
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 3.06783342788996e-05,
+      "loss": 2.0551,
+      "step": 13355000
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 3.067761063125232e-05,
+      "loss": 2.0355,
+      "step": 13355500
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 3.067688698360504e-05,
+      "loss": 2.0577,
+      "step": 13356000
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 3.0676163335957765e-05,
+      "loss": 2.0598,
+      "step": 13356500
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 3.067543968831049e-05,
+      "loss": 2.0515,
+      "step": 13357000
+    },
+    {
+      "epoch": 38.66,
+      "learning_rate": 3.067471604066321e-05,
+      "loss": 2.0491,
+      "step": 13357500
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 3.0673993840311225e-05,
+      "loss": 2.0634,
+      "step": 13358000
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 3.067327163995925e-05,
+      "loss": 2.0648,
+      "step": 13358500
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 3.067254799231197e-05,
+      "loss": 2.0815,
+      "step": 13359000
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 3.067182434466469e-05,
+      "loss": 2.065,
+      "step": 13359500
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 3.0671100697017414e-05,
+      "loss": 2.0336,
+      "step": 13360000
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 3.067037849666543e-05,
+      "loss": 2.0702,
+      "step": 13360500
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 3.066965484901816e-05,
+      "loss": 2.0538,
+      "step": 13361000
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 3.066893120137088e-05,
+      "loss": 2.0539,
+      "step": 13361500
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 3.06682075537236e-05,
+      "loss": 2.0563,
+      "step": 13362000
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 3.0667483906076325e-05,
+      "loss": 2.0446,
+      "step": 13362500
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 3.066676025842905e-05,
+      "loss": 2.0498,
+      "step": 13363000
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 3.0666036610781776e-05,
+      "loss": 2.0612,
+      "step": 13363500
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 3.06653129631345e-05,
+      "loss": 2.0389,
+      "step": 13364000
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 3.066458931548722e-05,
+      "loss": 2.0357,
+      "step": 13364500
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 3.066386566783994e-05,
+      "loss": 2.0613,
+      "step": 13365000
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 3.066314491478325e-05,
+      "loss": 2.0441,
+      "step": 13365500
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 3.0662421267135974e-05,
+      "loss": 2.073,
+      "step": 13366000
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 3.0661697619488696e-05,
+      "loss": 2.052,
+      "step": 13366500
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 3.0660973971841425e-05,
+      "loss": 2.0683,
+      "step": 13367000
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 3.066025032419415e-05,
+      "loss": 2.0435,
+      "step": 13367500
+    },
+    {
+      "epoch": 38.69,
+      "learning_rate": 3.065952667654687e-05,
+      "loss": 2.0527,
+      "step": 13368000
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 3.065880447619489e-05,
+      "loss": 2.0305,
+      "step": 13368500
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 3.0658080828547614e-05,
+      "loss": 2.0688,
+      "step": 13369000
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 3.0657357180900337e-05,
+      "loss": 2.0767,
+      "step": 13369500
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 3.065663498054835e-05,
+      "loss": 2.0627,
+      "step": 13370000
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 3.0655911332901074e-05,
+      "loss": 2.0416,
+      "step": 13370500
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 3.06551876852538e-05,
+      "loss": 2.0779,
+      "step": 13371000
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 3.0654464037606526e-05,
+      "loss": 2.0519,
+      "step": 13371500
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 3.065374038995925e-05,
+      "loss": 2.0567,
+      "step": 13372000
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 3.065301674231197e-05,
+      "loss": 2.066,
+      "step": 13372500
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 3.065229309466469e-05,
+      "loss": 2.0439,
+      "step": 13373000
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 3.0651569447017415e-05,
+      "loss": 2.0605,
+      "step": 13373500
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 3.065084579937014e-05,
+      "loss": 2.047,
+      "step": 13374000
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 3.065012215172286e-05,
+      "loss": 2.0521,
+      "step": 13374500
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 3.064939850407558e-05,
+      "loss": 2.0632,
+      "step": 13375000
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 3.0648676303723604e-05,
+      "loss": 2.0725,
+      "step": 13375500
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 3.0647952656076326e-05,
+      "loss": 2.0763,
+      "step": 13376000
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 3.064722900842905e-05,
+      "loss": 2.0506,
+      "step": 13376500
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 3.064650536078178e-05,
+      "loss": 2.0585,
+      "step": 13377000
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 3.06457817131345e-05,
+      "loss": 2.032,
+      "step": 13377500
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 3.064505806548722e-05,
+      "loss": 2.0603,
+      "step": 13378000
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 3.0644334417839944e-05,
+      "loss": 2.0777,
+      "step": 13378500
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 3.0643610770192666e-05,
+      "loss": 2.0752,
+      "step": 13379000
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 3.064288712254539e-05,
+      "loss": 2.0585,
+      "step": 13379500
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 3.0642164922193404e-05,
+      "loss": 2.0523,
+      "step": 13380000
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 3.0641441274546126e-05,
+      "loss": 2.0607,
+      "step": 13380500
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 3.064071762689885e-05,
+      "loss": 2.0686,
+      "step": 13381000
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 3.063999397925158e-05,
+      "loss": 2.0713,
+      "step": 13381500
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 3.06392703316043e-05,
+      "loss": 2.0735,
+      "step": 13382000
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 3.063854668395702e-05,
+      "loss": 2.066,
+      "step": 13382500
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 3.063782303630975e-05,
+      "loss": 2.0782,
+      "step": 13383000
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 3.0637100835957766e-05,
+      "loss": 2.0647,
+      "step": 13383500
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 3.063637718831049e-05,
+      "loss": 2.0673,
+      "step": 13384000
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 3.063565354066321e-05,
+      "loss": 2.0723,
+      "step": 13384500
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 3.063492989301593e-05,
+      "loss": 2.0579,
+      "step": 13385000
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 3.0634206245368655e-05,
+      "loss": 2.0686,
+      "step": 13385500
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 3.063348259772138e-05,
+      "loss": 2.0435,
+      "step": 13386000
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 3.06327589500741e-05,
+      "loss": 2.0718,
+      "step": 13386500
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 3.063203530242683e-05,
+      "loss": 2.0637,
+      "step": 13387000
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 3.0631313102074844e-05,
+      "loss": 2.0764,
+      "step": 13387500
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 3.063059234901815e-05,
+      "loss": 2.0644,
+      "step": 13388000
+    },
+    {
+      "epoch": 38.75,
+      "learning_rate": 3.0629868701370875e-05,
+      "loss": 2.0564,
+      "step": 13388500
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 3.06291465010189e-05,
+      "loss": 2.0602,
+      "step": 13389000
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 3.062842285337162e-05,
+      "loss": 2.0628,
+      "step": 13389500
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 3.062769920572434e-05,
+      "loss": 2.0628,
+      "step": 13390000
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 3.0626975558077064e-05,
+      "loss": 2.0665,
+      "step": 13390500
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 3.0626251910429793e-05,
+      "loss": 2.067,
+      "step": 13391000
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 3.0625528262782516e-05,
+      "loss": 2.0661,
+      "step": 13391500
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 3.062480461513524e-05,
+      "loss": 2.0535,
+      "step": 13392000
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 3.062408096748796e-05,
+      "loss": 2.0737,
+      "step": 13392500
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 3.062335731984068e-05,
+      "loss": 2.1037,
+      "step": 13393000
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 3.0622633672193405e-05,
+      "loss": 2.0591,
+      "step": 13393500
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 3.062191002454613e-05,
+      "loss": 2.0623,
+      "step": 13394000
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 3.0621186376898856e-05,
+      "loss": 2.0556,
+      "step": 13394500
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 3.062046272925158e-05,
+      "loss": 2.0579,
+      "step": 13395000
+    },
+    {
+      "epoch": 38.77,
+      "learning_rate": 3.06197390816043e-05,
+      "loss": 2.0707,
+      "step": 13395500
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 3.061901543395702e-05,
+      "loss": 2.0692,
+      "step": 13396000
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 3.0618291786309745e-05,
+      "loss": 2.0577,
+      "step": 13396500
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 3.061756813866247e-05,
+      "loss": 2.0556,
+      "step": 13397000
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 3.061684449101519e-05,
+      "loss": 2.0594,
+      "step": 13397500
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 3.061612084336792e-05,
+      "loss": 2.0434,
+      "step": 13398000
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 3.061539719572064e-05,
+      "loss": 2.0713,
+      "step": 13398500
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 3.0614674995368656e-05,
+      "loss": 2.0428,
+      "step": 13399000
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 3.061395279501668e-05,
+      "loss": 2.0674,
+      "step": 13399500
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 3.06132291473694e-05,
+      "loss": 2.0491,
+      "step": 13400000
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 3.0612506947017416e-05,
+      "loss": 2.0754,
+      "step": 13400500
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 3.061178329937014e-05,
+      "loss": 2.046,
+      "step": 13401000
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 3.061105965172286e-05,
+      "loss": 2.0644,
+      "step": 13401500
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 3.061033600407558e-05,
+      "loss": 2.0762,
+      "step": 13402000
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 3.0609612356428305e-05,
+      "loss": 2.0738,
+      "step": 13402500
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 3.060888870878103e-05,
+      "loss": 2.0544,
+      "step": 13403000
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 3.0608165061133756e-05,
+      "loss": 2.0888,
+      "step": 13403500
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 3.060744141348648e-05,
+      "loss": 2.0842,
+      "step": 13404000
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 3.0606719213134494e-05,
+      "loss": 2.0739,
+      "step": 13404500
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 3.0605995565487216e-05,
+      "loss": 2.0733,
+      "step": 13405000
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 3.060527191783994e-05,
+      "loss": 2.0834,
+      "step": 13405500
+    },
+    {
+      "epoch": 38.8,
+      "learning_rate": 3.060454827019267e-05,
+      "loss": 2.0404,
+      "step": 13406000
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 3.060382462254539e-05,
+      "loss": 2.061,
+      "step": 13406500
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 3.060310097489811e-05,
+      "loss": 2.0695,
+      "step": 13407000
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 3.0602377327250834e-05,
+      "loss": 2.0769,
+      "step": 13407500
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 3.0601653679603557e-05,
+      "loss": 2.055,
+      "step": 13408000
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 3.060093003195628e-05,
+      "loss": 2.065,
+      "step": 13408500
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 3.060020638430901e-05,
+      "loss": 2.0527,
+      "step": 13409000
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 3.059948273666173e-05,
+      "loss": 2.0642,
+      "step": 13409500
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 3.0598760536309746e-05,
+      "loss": 2.0539,
+      "step": 13410000
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 3.059803688866247e-05,
+      "loss": 2.0639,
+      "step": 13410500
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 3.059731324101519e-05,
+      "loss": 2.0612,
+      "step": 13411000
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 3.0596591040663206e-05,
+      "loss": 2.0314,
+      "step": 13411500
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 3.059586739301593e-05,
+      "loss": 2.0619,
+      "step": 13412000
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 3.059514374536866e-05,
+      "loss": 2.0707,
+      "step": 13412500
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 3.059442009772138e-05,
+      "loss": 2.0615,
+      "step": 13413000
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 3.059369645007411e-05,
+      "loss": 2.0606,
+      "step": 13413500
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 3.059297280242683e-05,
+      "loss": 2.0612,
+      "step": 13414000
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 3.059224915477955e-05,
+      "loss": 2.0758,
+      "step": 13414500
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 3.0591525507132275e-05,
+      "loss": 2.0699,
+      "step": 13415000
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 3.059080330678029e-05,
+      "loss": 2.06,
+      "step": 13415500
+    },
+    {
+      "epoch": 38.83,
+      "learning_rate": 3.059007965913301e-05,
+      "loss": 2.0431,
+      "step": 13416000
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 3.0589357458781035e-05,
+      "loss": 2.0779,
+      "step": 13416500
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 3.058863381113376e-05,
+      "loss": 2.0723,
+      "step": 13417000
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 3.058791016348648e-05,
+      "loss": 2.0556,
+      "step": 13417500
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 3.05871865158392e-05,
+      "loss": 2.0642,
+      "step": 13418000
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 3.0586462868191924e-05,
+      "loss": 2.0555,
+      "step": 13418500
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 3.0585739220544646e-05,
+      "loss": 2.0786,
+      "step": 13419000
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 3.058501702019266e-05,
+      "loss": 2.0398,
+      "step": 13419500
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 3.0584293372545384e-05,
+      "loss": 2.0641,
+      "step": 13420000
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 3.058356972489811e-05,
+      "loss": 2.0487,
+      "step": 13420500
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 3.0582846077250835e-05,
+      "loss": 2.0748,
+      "step": 13421000
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 3.058212387689886e-05,
+      "loss": 2.0593,
+      "step": 13421500
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 3.058140022925158e-05,
+      "loss": 2.0644,
+      "step": 13422000
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 3.0580678028899595e-05,
+      "loss": 2.0489,
+      "step": 13422500
+    },
+    {
+      "epoch": 38.85,
+      "learning_rate": 3.057995438125232e-05,
+      "loss": 2.0609,
+      "step": 13423000
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 3.057923073360504e-05,
+      "loss": 2.0646,
+      "step": 13423500
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 3.0578508533253055e-05,
+      "loss": 2.0458,
+      "step": 13424000
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 3.0577784885605784e-05,
+      "loss": 2.0459,
+      "step": 13424500
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 3.0577061237958506e-05,
+      "loss": 2.0729,
+      "step": 13425000
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 3.057633759031123e-05,
+      "loss": 2.0748,
+      "step": 13425500
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 3.057561394266395e-05,
+      "loss": 2.0495,
+      "step": 13426000
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 3.057489029501667e-05,
+      "loss": 2.0574,
+      "step": 13426500
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 3.0574166647369395e-05,
+      "loss": 2.0593,
+      "step": 13427000
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 3.057344299972212e-05,
+      "loss": 2.0513,
+      "step": 13427500
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 3.057271935207485e-05,
+      "loss": 2.0602,
+      "step": 13428000
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 3.057199570442757e-05,
+      "loss": 2.1,
+      "step": 13428500
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 3.057127205678029e-05,
+      "loss": 2.0673,
+      "step": 13429000
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 3.057054985642831e-05,
+      "loss": 2.0414,
+      "step": 13429500
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 3.0569826208781036e-05,
+      "loss": 2.0775,
+      "step": 13430000
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 3.056910400842905e-05,
+      "loss": 2.0351,
+      "step": 13430500
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 3.0568380360781773e-05,
+      "loss": 2.0714,
+      "step": 13431000
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 3.0567656713134496e-05,
+      "loss": 2.0858,
+      "step": 13431500
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 3.056693306548722e-05,
+      "loss": 2.0657,
+      "step": 13432000
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 3.056620941783994e-05,
+      "loss": 2.0344,
+      "step": 13432500
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 3.056548577019266e-05,
+      "loss": 2.0396,
+      "step": 13433000
+    },
+    {
+      "epoch": 38.88,
+      "learning_rate": 3.0564762122545385e-05,
+      "loss": 2.0637,
+      "step": 13433500
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 3.056403992219341e-05,
+      "loss": 2.0691,
+      "step": 13434000
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 3.056331627454613e-05,
+      "loss": 2.0481,
+      "step": 13434500
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 3.056259262689885e-05,
+      "loss": 2.068,
+      "step": 13435000
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 3.056186897925158e-05,
+      "loss": 2.0777,
+      "step": 13435500
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 3.05611453316043e-05,
+      "loss": 2.0628,
+      "step": 13436000
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 3.0560421683957025e-05,
+      "loss": 2.0459,
+      "step": 13436500
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 3.055969803630975e-05,
+      "loss": 2.0672,
+      "step": 13437000
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 3.055897438866247e-05,
+      "loss": 2.0674,
+      "step": 13437500
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 3.055825074101519e-05,
+      "loss": 2.0755,
+      "step": 13438000
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 3.0557527093367914e-05,
+      "loss": 2.0703,
+      "step": 13438500
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 3.0556803445720636e-05,
+      "loss": 2.0767,
+      "step": 13439000
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 3.055607979807336e-05,
+      "loss": 2.0619,
+      "step": 13439500
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 3.055535615042609e-05,
+      "loss": 2.058,
+      "step": 13440000
+    },
+    {
+      "epoch": 38.9,
+      "learning_rate": 3.055463250277881e-05,
+      "loss": 2.0372,
+      "step": 13440500
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 3.055390885513153e-05,
+      "loss": 2.0554,
+      "step": 13441000
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 3.0553185207484254e-05,
+      "loss": 2.0502,
+      "step": 13441500
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 3.0552461559836976e-05,
+      "loss": 2.066,
+      "step": 13442000
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 3.055173935948499e-05,
+      "loss": 2.0376,
+      "step": 13442500
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 3.055101571183772e-05,
+      "loss": 2.0557,
+      "step": 13443000
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 3.055029206419044e-05,
+      "loss": 2.0859,
+      "step": 13443500
+    },
+    {
+      "epoch": 38.91,
+      "learning_rate": 3.0549568416543165e-05,
+      "loss": 2.0752,
+      "step": 13444000
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 3.054884476889589e-05,
+      "loss": 2.0856,
+      "step": 13444500
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 3.054812112124861e-05,
+      "loss": 2.0807,
+      "step": 13445000
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 3.054739892089663e-05,
+      "loss": 2.0709,
+      "step": 13445500
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 3.0546675273249354e-05,
+      "loss": 2.0738,
+      "step": 13446000
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 3.054595307289737e-05,
+      "loss": 2.0543,
+      "step": 13446500
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 3.054522942525009e-05,
+      "loss": 2.0774,
+      "step": 13447000
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 3.0544505777602814e-05,
+      "loss": 2.0466,
+      "step": 13447500
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 3.0543782129955537e-05,
+      "loss": 2.0429,
+      "step": 13448000
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 3.054305848230826e-05,
+      "loss": 2.0772,
+      "step": 13448500
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 3.054233483466099e-05,
+      "loss": 2.0737,
+      "step": 13449000
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 3.054161118701371e-05,
+      "loss": 2.0603,
+      "step": 13449500
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 3.054088753936644e-05,
+      "loss": 2.0852,
+      "step": 13450000
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 3.054016389171916e-05,
+      "loss": 2.0691,
+      "step": 13450500
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 3.0539440244071884e-05,
+      "loss": 2.0429,
+      "step": 13451000
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 3.05387180437199e-05,
+      "loss": 2.0535,
+      "step": 13451500
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 3.053799439607262e-05,
+      "loss": 2.0606,
+      "step": 13452000
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 3.0537270748425344e-05,
+      "loss": 2.0684,
+      "step": 13452500
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 3.0536547100778066e-05,
+      "loss": 2.0523,
+      "step": 13453000
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 3.053582490042609e-05,
+      "loss": 2.0418,
+      "step": 13453500
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 3.0535102700074104e-05,
+      "loss": 2.0794,
+      "step": 13454000
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 3.0534379052426826e-05,
+      "loss": 2.0526,
+      "step": 13454500
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 3.053365540477955e-05,
+      "loss": 2.0535,
+      "step": 13455000
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 3.053293175713227e-05,
+      "loss": 2.0471,
+      "step": 13455500
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 3.053220810948499e-05,
+      "loss": 2.0699,
+      "step": 13456000
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 3.0531484461837715e-05,
+      "loss": 2.0878,
+      "step": 13456500
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 3.053076081419044e-05,
+      "loss": 2.0693,
+      "step": 13457000
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 3.0530037166543166e-05,
+      "loss": 2.0726,
+      "step": 13457500
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 3.052931496619119e-05,
+      "loss": 2.056,
+      "step": 13458000
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 3.052859131854391e-05,
+      "loss": 2.0871,
+      "step": 13458500
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 3.0527869118191926e-05,
+      "loss": 2.0695,
+      "step": 13459000
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 3.052714547054465e-05,
+      "loss": 2.0558,
+      "step": 13459500
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 3.052642182289737e-05,
+      "loss": 2.0603,
+      "step": 13460000
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 3.052569817525009e-05,
+      "loss": 2.0449,
+      "step": 13460500
+    },
+    {
+      "epoch": 38.96,
+      "learning_rate": 3.0524974527602815e-05,
+      "loss": 2.0627,
+      "step": 13461000
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 3.052425087995554e-05,
+      "loss": 2.0594,
+      "step": 13461500
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 3.052352867960356e-05,
+      "loss": 2.0689,
+      "step": 13462000
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 3.0522806479251575e-05,
+      "loss": 2.083,
+      "step": 13462500
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 3.05220828316043e-05,
+      "loss": 2.0561,
+      "step": 13463000
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 3.052135918395702e-05,
+      "loss": 2.0645,
+      "step": 13463500
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 3.052063553630974e-05,
+      "loss": 2.0761,
+      "step": 13464000
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 3.0519911888662464e-05,
+      "loss": 2.0684,
+      "step": 13464500
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 3.0519188241015186e-05,
+      "loss": 2.0678,
+      "step": 13465000
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 3.0518464593367915e-05,
+      "loss": 2.0667,
+      "step": 13465500
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 3.051774094572064e-05,
+      "loss": 2.0659,
+      "step": 13466000
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 3.0517017298073363e-05,
+      "loss": 2.0566,
+      "step": 13466500
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 3.0516293650426086e-05,
+      "loss": 2.0433,
+      "step": 13467000
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 3.0515571450074104e-05,
+      "loss": 2.0649,
+      "step": 13467500
+    },
+    {
+      "epoch": 38.98,
+      "learning_rate": 3.0514847802426827e-05,
+      "loss": 2.0619,
+      "step": 13468000
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 3.051412415477955e-05,
+      "loss": 2.0805,
+      "step": 13468500
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 3.0513401954427568e-05,
+      "loss": 2.065,
+      "step": 13469000
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 3.051267830678029e-05,
+      "loss": 2.0444,
+      "step": 13469500
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 3.0511954659133012e-05,
+      "loss": 2.0311,
+      "step": 13470000
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 3.0511231011485735e-05,
+      "loss": 2.069,
+      "step": 13470500
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 3.051050736383846e-05,
+      "loss": 2.0379,
+      "step": 13471000
+    },
+    {
+      "epoch": 38.99,
+      "learning_rate": 3.0509785163486476e-05,
+      "loss": 2.0784,
+      "step": 13471500
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 3.0509061515839198e-05,
+      "loss": 2.074,
+      "step": 13472000
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 3.050833786819192e-05,
+      "loss": 2.0628,
+      "step": 13472500
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 3.050761566783994e-05,
+      "loss": 2.0457,
+      "step": 13473000
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy": 0.670671881118002,
+      "eval_accuracy_mlm": 0.6356850853404965,
+      "eval_accuracy_nsp": 0.8582585142511446,
+      "eval_loss": 2.165445327758789,
+      "eval_runtime": 331.7981,
+      "eval_samples_per_second": 1315.216,
+      "eval_steps_per_second": 54.801,
+      "step": 13473408
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 3.0506892020192668e-05,
+      "loss": 2.067,
+      "step": 13473500
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 3.050616837254539e-05,
+      "loss": 2.0359,
+      "step": 13474000
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 3.0505444724898113e-05,
+      "loss": 2.0454,
+      "step": 13474500
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 3.0504721077250835e-05,
+      "loss": 2.0357,
+      "step": 13475000
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 3.050399742960356e-05,
+      "loss": 2.0514,
+      "step": 13475500
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 3.0503273781956283e-05,
+      "loss": 2.0459,
+      "step": 13476000
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 3.0502550134309005e-05,
+      "loss": 2.0284,
+      "step": 13476500
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 3.0501826486661727e-05,
+      "loss": 2.0664,
+      "step": 13477000
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 3.050110283901445e-05,
+      "loss": 2.0234,
+      "step": 13477500
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 3.050037919136717e-05,
+      "loss": 2.0301,
+      "step": 13478000
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 3.0499655543719897e-05,
+      "loss": 2.0341,
+      "step": 13478500
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 3.049893189607262e-05,
+      "loss": 2.0509,
+      "step": 13479000
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 3.0498208248425342e-05,
+      "loss": 2.0225,
+      "step": 13479500
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 3.049748604807336e-05,
+      "loss": 2.0345,
+      "step": 13480000
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 3.0496762400426086e-05,
+      "loss": 2.0305,
+      "step": 13480500
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 3.0496038752778812e-05,
+      "loss": 2.0365,
+      "step": 13481000
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 3.0495315105131534e-05,
+      "loss": 2.049,
+      "step": 13481500
+    },
+    {
+      "epoch": 39.02,
+      "learning_rate": 3.0494591457484256e-05,
+      "loss": 2.0357,
+      "step": 13482000
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 3.049386780983698e-05,
+      "loss": 2.0503,
+      "step": 13482500
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 3.04931441621897e-05,
+      "loss": 2.0263,
+      "step": 13483000
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 3.0492420514542423e-05,
+      "loss": 2.0413,
+      "step": 13483500
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 3.0491698314190442e-05,
+      "loss": 2.0307,
+      "step": 13484000
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 3.049097611383846e-05,
+      "loss": 2.0425,
+      "step": 13484500
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 3.0490252466191183e-05,
+      "loss": 2.0416,
+      "step": 13485000
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 3.0489528818543905e-05,
+      "loss": 2.0431,
+      "step": 13485500
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 3.0488805170896628e-05,
+      "loss": 2.0321,
+      "step": 13486000
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 3.048808152324935e-05,
+      "loss": 2.0438,
+      "step": 13486500
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 3.0487357875602076e-05,
+      "loss": 2.0413,
+      "step": 13487000
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 3.048663567525009e-05,
+      "loss": 2.0389,
+      "step": 13487500
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 3.048591202760282e-05,
+      "loss": 2.0395,
+      "step": 13488000
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 3.0485188379955542e-05,
+      "loss": 2.0539,
+      "step": 13488500
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 3.0484464732308265e-05,
+      "loss": 2.0372,
+      "step": 13489000
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 3.0483741084660987e-05,
+      "loss": 2.0397,
+      "step": 13489500
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 3.0483017437013712e-05,
+      "loss": 2.0421,
+      "step": 13490000
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 3.0482293789366435e-05,
+      "loss": 2.0502,
+      "step": 13490500
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 3.0481570141719157e-05,
+      "loss": 2.0402,
+      "step": 13491000
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 3.0480847941367176e-05,
+      "loss": 2.0459,
+      "step": 13491500
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 3.0480124293719898e-05,
+      "loss": 2.0406,
+      "step": 13492000
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 3.0479402093367914e-05,
+      "loss": 2.0685,
+      "step": 13492500
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 3.047867844572064e-05,
+      "loss": 2.037,
+      "step": 13493000
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 3.047795479807336e-05,
+      "loss": 2.0194,
+      "step": 13493500
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 3.0477231150426084e-05,
+      "loss": 2.0433,
+      "step": 13494000
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 3.0476507502778806e-05,
+      "loss": 2.0437,
+      "step": 13494500
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 3.0475785302426825e-05,
+      "loss": 2.0862,
+      "step": 13495000
+    },
+    {
+      "epoch": 39.06,
+      "learning_rate": 3.047506165477955e-05,
+      "loss": 2.0431,
+      "step": 13495500
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 3.047433945442757e-05,
+      "loss": 2.0625,
+      "step": 13496000
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 3.0473615806780292e-05,
+      "loss": 2.0451,
+      "step": 13496500
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 3.0472892159133014e-05,
+      "loss": 2.0657,
+      "step": 13497000
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 3.047216851148574e-05,
+      "loss": 2.062,
+      "step": 13497500
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 3.0471444863838462e-05,
+      "loss": 2.0422,
+      "step": 13498000
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 3.0470721216191184e-05,
+      "loss": 2.038,
+      "step": 13498500
+    },
+    {
+      "epoch": 39.07,
+      "learning_rate": 3.0469997568543906e-05,
+      "loss": 2.0319,
+      "step": 13499000
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 3.0469275368191925e-05,
+      "loss": 2.0661,
+      "step": 13499500
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 3.0468551720544647e-05,
+      "loss": 2.0323,
+      "step": 13500000
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 3.046782807289737e-05,
+      "loss": 2.0137,
+      "step": 13500500
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 3.0467104425250092e-05,
+      "loss": 2.0332,
+      "step": 13501000
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 3.0466380777602814e-05,
+      "loss": 2.0544,
+      "step": 13501500
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 3.046565712995554e-05,
+      "loss": 2.0399,
+      "step": 13502000
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 3.0464933482308262e-05,
+      "loss": 2.029,
+      "step": 13502500
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 3.046420983466099e-05,
+      "loss": 2.0543,
+      "step": 13503000
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 3.0463486187013713e-05,
+      "loss": 2.0439,
+      "step": 13503500
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 3.0462762539366435e-05,
+      "loss": 2.0558,
+      "step": 13504000
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 3.0462038891719158e-05,
+      "loss": 2.0375,
+      "step": 13504500
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 3.0461316691367177e-05,
+      "loss": 2.0737,
+      "step": 13505000
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 3.04605930437199e-05,
+      "loss": 2.0258,
+      "step": 13505500
+    },
+    {
+      "epoch": 39.09,
+      "learning_rate": 3.045986939607262e-05,
+      "loss": 2.0499,
+      "step": 13506000
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 3.0459145748425343e-05,
+      "loss": 2.0614,
+      "step": 13506500
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 3.0458422100778066e-05,
+      "loss": 2.0475,
+      "step": 13507000
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 3.045769845313079e-05,
+      "loss": 2.0333,
+      "step": 13507500
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 3.0456976252778807e-05,
+      "loss": 2.0399,
+      "step": 13508000
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 3.0456254052426826e-05,
+      "loss": 2.0643,
+      "step": 13508500
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 3.0455530404779548e-05,
+      "loss": 2.0295,
+      "step": 13509000
+    },
+    {
+      "epoch": 39.1,
+      "learning_rate": 3.045480675713227e-05,
+      "loss": 2.0685,
+      "step": 13509500
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 3.0454083109484992e-05,
+      "loss": 2.0386,
+      "step": 13510000
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 3.0453360909133015e-05,
+      "loss": 2.0454,
+      "step": 13510500
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 3.045263726148574e-05,
+      "loss": 2.0556,
+      "step": 13511000
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 3.0451913613838463e-05,
+      "loss": 2.0348,
+      "step": 13511500
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 3.0451189966191185e-05,
+      "loss": 2.0393,
+      "step": 13512000
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 3.0450466318543907e-05,
+      "loss": 2.0387,
+      "step": 13512500
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 3.044974267089663e-05,
+      "loss": 2.06,
+      "step": 13513000
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 3.0449019023249355e-05,
+      "loss": 2.0253,
+      "step": 13513500
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 3.0448295375602077e-05,
+      "loss": 2.0646,
+      "step": 13514000
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 3.04475717279548e-05,
+      "loss": 2.0578,
+      "step": 13514500
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 3.044684808030752e-05,
+      "loss": 2.031,
+      "step": 13515000
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 3.0446124432660244e-05,
+      "loss": 2.0369,
+      "step": 13515500
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 3.0445400785012966e-05,
+      "loss": 2.0339,
+      "step": 13516000
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 3.0444677137365692e-05,
+      "loss": 2.0486,
+      "step": 13516500
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 3.0443954937013707e-05,
+      "loss": 2.0413,
+      "step": 13517000
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 3.0443232736661726e-05,
+      "loss": 2.0709,
+      "step": 13517500
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 3.044251053630975e-05,
+      "loss": 2.0424,
+      "step": 13518000
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 3.044178688866247e-05,
+      "loss": 2.0397,
+      "step": 13518500
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 3.0441063241015193e-05,
+      "loss": 2.0279,
+      "step": 13519000
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 3.044033959336792e-05,
+      "loss": 2.0272,
+      "step": 13519500
+    },
+    {
+      "epoch": 39.13,
+      "learning_rate": 3.043961594572064e-05,
+      "loss": 2.0383,
+      "step": 13520000
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 3.0438892298073363e-05,
+      "loss": 2.0444,
+      "step": 13520500
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 3.0438168650426085e-05,
+      "loss": 2.0415,
+      "step": 13521000
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 3.0437445002778808e-05,
+      "loss": 2.0605,
+      "step": 13521500
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 3.043672135513153e-05,
+      "loss": 2.0253,
+      "step": 13522000
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 3.0435997707484255e-05,
+      "loss": 2.0229,
+      "step": 13522500
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 3.0435274059836978e-05,
+      "loss": 2.0311,
+      "step": 13523000
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 3.04345504121897e-05,
+      "loss": 2.0445,
+      "step": 13523500
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 3.0433826764542422e-05,
+      "loss": 2.0631,
+      "step": 13524000
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 3.0433103116895144e-05,
+      "loss": 2.036,
+      "step": 13524500
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 3.0432379469247873e-05,
+      "loss": 2.0316,
+      "step": 13525000
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 3.0431655821600596e-05,
+      "loss": 2.0671,
+      "step": 13525500
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 3.0430933621248615e-05,
+      "loss": 2.0393,
+      "step": 13526000
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 3.0430209973601337e-05,
+      "loss": 2.0367,
+      "step": 13526500
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 3.042948922054465e-05,
+      "loss": 2.0505,
+      "step": 13527000
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 3.042876557289737e-05,
+      "loss": 2.0367,
+      "step": 13527500
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 3.042804337254539e-05,
+      "loss": 2.0679,
+      "step": 13528000
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 3.0427319724898112e-05,
+      "loss": 2.0318,
+      "step": 13528500
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 3.0426596077250835e-05,
+      "loss": 2.0507,
+      "step": 13529000
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 3.0425872429603557e-05,
+      "loss": 2.0602,
+      "step": 13529500
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 3.042514878195628e-05,
+      "loss": 2.0265,
+      "step": 13530000
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 3.0424426581604298e-05,
+      "loss": 2.0723,
+      "step": 13530500
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 3.042370293395702e-05,
+      "loss": 2.0363,
+      "step": 13531000
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 3.0422979286309743e-05,
+      "loss": 2.0536,
+      "step": 13531500
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 3.0422255638662468e-05,
+      "loss": 2.0248,
+      "step": 13532000
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 3.042153199101519e-05,
+      "loss": 2.0305,
+      "step": 13532500
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 3.042080834336792e-05,
+      "loss": 2.0487,
+      "step": 13533000
+    },
+    {
+      "epoch": 39.17,
+      "learning_rate": 3.042008469572064e-05,
+      "loss": 2.0231,
+      "step": 13533500
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 3.0419361048073364e-05,
+      "loss": 2.0644,
+      "step": 13534000
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 3.0418637400426086e-05,
+      "loss": 2.0479,
+      "step": 13534500
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 3.041791375277881e-05,
+      "loss": 2.0669,
+      "step": 13535000
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 3.041719010513153e-05,
+      "loss": 2.0484,
+      "step": 13535500
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 3.041646790477955e-05,
+      "loss": 2.0485,
+      "step": 13536000
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 3.0415744257132272e-05,
+      "loss": 2.0443,
+      "step": 13536500
+    },
+    {
+      "epoch": 39.18,
+      "learning_rate": 3.0415020609484994e-05,
+      "loss": 2.0371,
+      "step": 13537000
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 3.041429696183772e-05,
+      "loss": 2.0536,
+      "step": 13537500
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 3.0413573314190442e-05,
+      "loss": 2.0338,
+      "step": 13538000
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 3.0412849666543164e-05,
+      "loss": 2.045,
+      "step": 13538500
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 3.0412127466191183e-05,
+      "loss": 2.0274,
+      "step": 13539000
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 3.0411403818543905e-05,
+      "loss": 2.0597,
+      "step": 13539500
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 3.0410680170896627e-05,
+      "loss": 2.0434,
+      "step": 13540000
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 3.0409956523249356e-05,
+      "loss": 2.0411,
+      "step": 13540500
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 3.040923287560208e-05,
+      "loss": 2.0563,
+      "step": 13541000
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 3.04085092279548e-05,
+      "loss": 2.0552,
+      "step": 13541500
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 3.0407785580307523e-05,
+      "loss": 2.0558,
+      "step": 13542000
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 3.0407061932660245e-05,
+      "loss": 2.0372,
+      "step": 13542500
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 3.040633828501297e-05,
+      "loss": 2.0438,
+      "step": 13543000
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 3.0405614637365693e-05,
+      "loss": 2.0439,
+      "step": 13543500
+    },
+    {
+      "epoch": 39.2,
+      "learning_rate": 3.0404890989718416e-05,
+      "loss": 2.0479,
+      "step": 13544000
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 3.0404168789366434e-05,
+      "loss": 2.0549,
+      "step": 13544500
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 3.040344658901445e-05,
+      "loss": 2.0674,
+      "step": 13545000
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 3.040272438866247e-05,
+      "loss": 2.0384,
+      "step": 13545500
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 3.040200074101519e-05,
+      "loss": 2.0668,
+      "step": 13546000
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 3.0401278540663207e-05,
+      "loss": 2.0592,
+      "step": 13546500
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 3.0400554893015932e-05,
+      "loss": 2.0595,
+      "step": 13547000
+    },
+    {
+      "epoch": 39.21,
+      "learning_rate": 3.0399831245368655e-05,
+      "loss": 2.0357,
+      "step": 13547500
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 3.0399107597721384e-05,
+      "loss": 2.0719,
+      "step": 13548000
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 3.0398383950074106e-05,
+      "loss": 2.0378,
+      "step": 13548500
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 3.0397660302426828e-05,
+      "loss": 2.0603,
+      "step": 13549000
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 3.039693665477955e-05,
+      "loss": 2.0602,
+      "step": 13549500
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 3.0396213007132273e-05,
+      "loss": 2.0576,
+      "step": 13550000
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 3.0395489359484998e-05,
+      "loss": 2.0485,
+      "step": 13550500
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 3.039476571183772e-05,
+      "loss": 2.052,
+      "step": 13551000
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 3.0394042064190443e-05,
+      "loss": 2.0601,
+      "step": 13551500
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 3.0393319863838458e-05,
+      "loss": 2.0523,
+      "step": 13552000
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 3.0392596216191184e-05,
+      "loss": 2.0463,
+      "step": 13552500
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 3.0391872568543906e-05,
+      "loss": 2.0385,
+      "step": 13553000
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 3.0391148920896628e-05,
+      "loss": 2.0348,
+      "step": 13553500
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 3.039042527324935e-05,
+      "loss": 2.0455,
+      "step": 13554000
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 3.0389701625602073e-05,
+      "loss": 2.0388,
+      "step": 13554500
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 3.0388977977954802e-05,
+      "loss": 2.0338,
+      "step": 13555000
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 3.0388254330307524e-05,
+      "loss": 2.0377,
+      "step": 13555500
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 3.038753068266025e-05,
+      "loss": 2.0357,
+      "step": 13556000
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 3.0386807035012972e-05,
+      "loss": 2.0475,
+      "step": 13556500
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 3.0386083387365694e-05,
+      "loss": 2.0504,
+      "step": 13557000
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 3.0385359739718416e-05,
+      "loss": 2.0359,
+      "step": 13557500
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 3.038463609207114e-05,
+      "loss": 2.055,
+      "step": 13558000
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 3.0383913891719157e-05,
+      "loss": 2.0547,
+      "step": 13558500
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 3.038319024407188e-05,
+      "loss": 2.045,
+      "step": 13559000
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 3.03824680437199e-05,
+      "loss": 2.0613,
+      "step": 13559500
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 3.038174439607262e-05,
+      "loss": 2.0347,
+      "step": 13560000
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 3.0381020748425343e-05,
+      "loss": 2.0447,
+      "step": 13560500
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 3.0380297100778065e-05,
+      "loss": 2.0493,
+      "step": 13561000
+    },
+    {
+      "epoch": 39.25,
+      "learning_rate": 3.0379573453130788e-05,
+      "loss": 2.0416,
+      "step": 13561500
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 3.037884980548351e-05,
+      "loss": 2.0547,
+      "step": 13562000
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 3.037812615783624e-05,
+      "loss": 2.0542,
+      "step": 13562500
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 3.0377403957484258e-05,
+      "loss": 2.0555,
+      "step": 13563000
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 3.037668030983698e-05,
+      "loss": 2.0608,
+      "step": 13563500
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 3.0375956662189702e-05,
+      "loss": 2.0355,
+      "step": 13564000
+    },
+    {
+      "epoch": 39.26,
+      "learning_rate": 3.0375233014542424e-05,
+      "loss": 2.0527,
+      "step": 13564500
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 3.037450936689515e-05,
+      "loss": 2.0336,
+      "step": 13565000
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 3.0373785719247872e-05,
+      "loss": 2.0376,
+      "step": 13565500
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 3.0373062071600595e-05,
+      "loss": 2.0461,
+      "step": 13566000
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 3.0372338423953317e-05,
+      "loss": 2.0568,
+      "step": 13566500
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 3.037161477630604e-05,
+      "loss": 2.0285,
+      "step": 13567000
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 3.037089112865876e-05,
+      "loss": 2.0646,
+      "step": 13567500
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 3.0370167481011487e-05,
+      "loss": 2.035,
+      "step": 13568000
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 3.036944383336421e-05,
+      "loss": 2.0536,
+      "step": 13568500
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 3.0368721633012225e-05,
+      "loss": 2.0171,
+      "step": 13569000
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 3.036799798536495e-05,
+      "loss": 2.0602,
+      "step": 13569500
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 3.0367274337717676e-05,
+      "loss": 2.0534,
+      "step": 13570000
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 3.03665506900704e-05,
+      "loss": 2.0448,
+      "step": 13570500
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 3.0365827042423124e-05,
+      "loss": 2.0476,
+      "step": 13571000
+    },
+    {
+      "epoch": 39.28,
+      "learning_rate": 3.0365103394775846e-05,
+      "loss": 2.0513,
+      "step": 13571500
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 3.036438119442386e-05,
+      "loss": 2.0524,
+      "step": 13572000
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 3.0363657546776587e-05,
+      "loss": 2.0585,
+      "step": 13572500
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 3.036293389912931e-05,
+      "loss": 2.0389,
+      "step": 13573000
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 3.036221025148203e-05,
+      "loss": 2.0696,
+      "step": 13573500
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 3.0361486603834754e-05,
+      "loss": 2.0436,
+      "step": 13574000
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 3.0360765850778066e-05,
+      "loss": 2.0545,
+      "step": 13574500
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 3.036004220313079e-05,
+      "loss": 2.0642,
+      "step": 13575000
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 3.0359318555483514e-05,
+      "loss": 2.0497,
+      "step": 13575500
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 3.0358594907836236e-05,
+      "loss": 2.0778,
+      "step": 13576000
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 3.035787126018896e-05,
+      "loss": 2.0471,
+      "step": 13576500
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 3.035714761254168e-05,
+      "loss": 2.0699,
+      "step": 13577000
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 3.035642396489441e-05,
+      "loss": 2.048,
+      "step": 13577500
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 3.0355701764542425e-05,
+      "loss": 2.0522,
+      "step": 13578000
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 3.035497811689515e-05,
+      "loss": 2.0436,
+      "step": 13578500
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 3.0354254469247873e-05,
+      "loss": 2.0499,
+      "step": 13579000
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 3.0353530821600595e-05,
+      "loss": 2.0429,
+      "step": 13579500
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 3.0352807173953318e-05,
+      "loss": 2.0255,
+      "step": 13580000
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 3.035208352630604e-05,
+      "loss": 2.049,
+      "step": 13580500
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 3.035136132595406e-05,
+      "loss": 2.0651,
+      "step": 13581000
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 3.035063767830678e-05,
+      "loss": 2.0712,
+      "step": 13581500
+    },
+    {
+      "epoch": 39.31,
+      "learning_rate": 3.0349914030659503e-05,
+      "loss": 2.0539,
+      "step": 13582000
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 3.0349190383012225e-05,
+      "loss": 2.0669,
+      "step": 13582500
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 3.034846673536495e-05,
+      "loss": 2.0493,
+      "step": 13583000
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 3.0347744535012967e-05,
+      "loss": 2.0554,
+      "step": 13583500
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 3.034702088736569e-05,
+      "loss": 2.0468,
+      "step": 13584000
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 3.0346297239718415e-05,
+      "loss": 2.0488,
+      "step": 13584500
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 3.034557359207114e-05,
+      "loss": 2.0582,
+      "step": 13585000
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 3.034485139171916e-05,
+      "loss": 2.0764,
+      "step": 13585500
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 3.034412774407188e-05,
+      "loss": 2.0536,
+      "step": 13586000
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 3.0343404096424604e-05,
+      "loss": 2.0244,
+      "step": 13586500
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 3.034268044877733e-05,
+      "loss": 2.0689,
+      "step": 13587000
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 3.034195680113005e-05,
+      "loss": 2.0549,
+      "step": 13587500
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 3.0341233153482774e-05,
+      "loss": 2.0379,
+      "step": 13588000
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 3.0340509505835496e-05,
+      "loss": 2.0471,
+      "step": 13588500
+    },
+    {
+      "epoch": 39.33,
+      "learning_rate": 3.0339787305483515e-05,
+      "loss": 2.0456,
+      "step": 13589000
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 3.0339063657836237e-05,
+      "loss": 2.0388,
+      "step": 13589500
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 3.033834001018896e-05,
+      "loss": 2.0565,
+      "step": 13590000
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 3.033761636254168e-05,
+      "loss": 2.0467,
+      "step": 13590500
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 3.0336892714894404e-05,
+      "loss": 2.0282,
+      "step": 13591000
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 3.033616906724713e-05,
+      "loss": 2.0444,
+      "step": 13591500
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 3.033544541959985e-05,
+      "loss": 2.0598,
+      "step": 13592000
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 3.0334721771952577e-05,
+      "loss": 2.0655,
+      "step": 13592500
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 3.0333998124305303e-05,
+      "loss": 2.0385,
+      "step": 13593000
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 3.033327592395332e-05,
+      "loss": 2.0475,
+      "step": 13593500
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 3.033255227630604e-05,
+      "loss": 2.071,
+      "step": 13594000
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 3.033183007595406e-05,
+      "loss": 2.0423,
+      "step": 13594500
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 3.0331106428306782e-05,
+      "loss": 2.0534,
+      "step": 13595000
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 3.0330382780659504e-05,
+      "loss": 2.0352,
+      "step": 13595500
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 3.032965913301223e-05,
+      "loss": 2.0534,
+      "step": 13596000
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 3.0328935485364952e-05,
+      "loss": 2.053,
+      "step": 13596500
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 3.0328211837717674e-05,
+      "loss": 2.0586,
+      "step": 13597000
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 3.0327488190070396e-05,
+      "loss": 2.0464,
+      "step": 13597500
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 3.032676454242312e-05,
+      "loss": 2.0724,
+      "step": 13598000
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 3.0326042342071138e-05,
+      "loss": 2.0686,
+      "step": 13598500
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 3.032531869442386e-05,
+      "loss": 2.0552,
+      "step": 13599000
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 3.0324595046776582e-05,
+      "loss": 2.0301,
+      "step": 13599500
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 3.032387139912931e-05,
+      "loss": 2.042,
+      "step": 13600000
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 3.0323147751482033e-05,
+      "loss": 2.0598,
+      "step": 13600500
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 3.0322425551130052e-05,
+      "loss": 2.033,
+      "step": 13601000
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 3.0321703350778068e-05,
+      "loss": 2.0295,
+      "step": 13601500
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 3.0320979703130793e-05,
+      "loss": 2.0363,
+      "step": 13602000
+    },
+    {
+      "epoch": 39.37,
+      "learning_rate": 3.0320256055483516e-05,
+      "loss": 2.0476,
+      "step": 13602500
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 3.0319532407836238e-05,
+      "loss": 2.0525,
+      "step": 13603000
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 3.031880876018896e-05,
+      "loss": 2.0729,
+      "step": 13603500
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 3.0318085112541682e-05,
+      "loss": 2.0565,
+      "step": 13604000
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 3.0317361464894405e-05,
+      "loss": 2.0689,
+      "step": 13604500
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 3.031663781724713e-05,
+      "loss": 2.0834,
+      "step": 13605000
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 3.0315914169599852e-05,
+      "loss": 2.0377,
+      "step": 13605500
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 3.0315191969247868e-05,
+      "loss": 2.0466,
+      "step": 13606000
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 3.0314468321600594e-05,
+      "loss": 2.0518,
+      "step": 13606500
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 3.0313744673953316e-05,
+      "loss": 2.0336,
+      "step": 13607000
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 3.0313021026306045e-05,
+      "loss": 2.0384,
+      "step": 13607500
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 3.031229882595406e-05,
+      "loss": 2.0586,
+      "step": 13608000
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 3.0311575178306783e-05,
+      "loss": 2.066,
+      "step": 13608500
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 3.0310851530659505e-05,
+      "loss": 2.0432,
+      "step": 13609000
+    },
+    {
+      "epoch": 39.39,
+      "learning_rate": 3.031012788301223e-05,
+      "loss": 2.0442,
+      "step": 13609500
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 3.0309404235364953e-05,
+      "loss": 2.0527,
+      "step": 13610000
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 3.0308682035012968e-05,
+      "loss": 2.0526,
+      "step": 13610500
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 3.0307958387365694e-05,
+      "loss": 2.0414,
+      "step": 13611000
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 3.0307234739718416e-05,
+      "loss": 2.0252,
+      "step": 13611500
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 3.030651109207114e-05,
+      "loss": 2.0355,
+      "step": 13612000
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 3.030578744442386e-05,
+      "loss": 2.0518,
+      "step": 13612500
+    },
+    {
+      "epoch": 39.4,
+      "learning_rate": 3.0305063796776583e-05,
+      "loss": 2.0556,
+      "step": 13613000
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 3.0304341596424602e-05,
+      "loss": 2.0421,
+      "step": 13613500
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 3.0303617948777324e-05,
+      "loss": 2.0582,
+      "step": 13614000
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 3.0302894301130046e-05,
+      "loss": 2.0427,
+      "step": 13614500
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 3.0302170653482775e-05,
+      "loss": 2.0524,
+      "step": 13615000
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 3.0301447005835497e-05,
+      "loss": 2.0416,
+      "step": 13615500
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 3.0300724805483516e-05,
+      "loss": 2.0515,
+      "step": 13616000
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 3.030000115783624e-05,
+      "loss": 2.0326,
+      "step": 13616500
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 3.029927751018896e-05,
+      "loss": 2.0558,
+      "step": 13617000
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 3.0298553862541683e-05,
+      "loss": 2.0582,
+      "step": 13617500
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 3.0297831662189702e-05,
+      "loss": 2.0616,
+      "step": 13618000
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 3.0297108014542424e-05,
+      "loss": 2.0451,
+      "step": 13618500
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 3.0296384366895147e-05,
+      "loss": 2.0612,
+      "step": 13619000
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 3.029566071924787e-05,
+      "loss": 2.0445,
+      "step": 13619500
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 3.0294937071600594e-05,
+      "loss": 2.0662,
+      "step": 13620000
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 3.0294213423953317e-05,
+      "loss": 2.0599,
+      "step": 13620500
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 3.029348977630604e-05,
+      "loss": 2.0414,
+      "step": 13621000
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 3.029276612865876e-05,
+      "loss": 2.0573,
+      "step": 13621500
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 3.0292042481011483e-05,
+      "loss": 2.0239,
+      "step": 13622000
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 3.0291318833364212e-05,
+      "loss": 2.043,
+      "step": 13622500
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 3.0290598080307525e-05,
+      "loss": 2.0685,
+      "step": 13623000
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 3.0289874432660247e-05,
+      "loss": 2.0441,
+      "step": 13623500
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 3.0289150785012972e-05,
+      "loss": 2.0487,
+      "step": 13624000
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 3.0288427137365695e-05,
+      "loss": 2.0342,
+      "step": 13624500
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 3.028770493701371e-05,
+      "loss": 2.0286,
+      "step": 13625000
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 3.0286981289366432e-05,
+      "loss": 2.0082,
+      "step": 13625500
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 3.0286257641719158e-05,
+      "loss": 2.022,
+      "step": 13626000
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 3.028553399407188e-05,
+      "loss": 2.0388,
+      "step": 13626500
+    },
+    {
+      "epoch": 39.44,
+      "learning_rate": 3.0284810346424603e-05,
+      "loss": 2.0221,
+      "step": 13627000
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 3.0284086698777325e-05,
+      "loss": 2.0518,
+      "step": 13627500
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 3.0283363051130047e-05,
+      "loss": 2.0633,
+      "step": 13628000
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 3.0282639403482773e-05,
+      "loss": 2.029,
+      "step": 13628500
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 3.0281917203130788e-05,
+      "loss": 2.0316,
+      "step": 13629000
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 3.028119355548351e-05,
+      "loss": 2.0457,
+      "step": 13629500
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 3.028046990783624e-05,
+      "loss": 2.0704,
+      "step": 13630000
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 3.027974626018896e-05,
+      "loss": 2.0682,
+      "step": 13630500
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 3.0279022612541684e-05,
+      "loss": 2.0457,
+      "step": 13631000
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 3.0278301859484996e-05,
+      "loss": 2.0317,
+      "step": 13631500
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 3.0277578211837722e-05,
+      "loss": 2.0449,
+      "step": 13632000
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 3.0276854564190444e-05,
+      "loss": 2.0408,
+      "step": 13632500
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 3.0276130916543166e-05,
+      "loss": 2.0607,
+      "step": 13633000
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 3.0275408716191185e-05,
+      "loss": 2.0793,
+      "step": 13633500
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 3.0274685068543907e-05,
+      "loss": 2.0515,
+      "step": 13634000
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 3.027396142089663e-05,
+      "loss": 2.0483,
+      "step": 13634500
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 3.0273237773249352e-05,
+      "loss": 2.0518,
+      "step": 13635000
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 3.0272514125602074e-05,
+      "loss": 2.0527,
+      "step": 13635500
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 3.0271790477954796e-05,
+      "loss": 2.0736,
+      "step": 13636000
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 3.0271066830307522e-05,
+      "loss": 2.0561,
+      "step": 13636500
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 3.0270343182660244e-05,
+      "loss": 2.0623,
+      "step": 13637000
+    },
+    {
+      "epoch": 39.47,
+      "learning_rate": 3.026962098230826e-05,
+      "loss": 2.0547,
+      "step": 13637500
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 3.026889733466099e-05,
+      "loss": 2.0223,
+      "step": 13638000
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 3.026817368701371e-05,
+      "loss": 2.0437,
+      "step": 13638500
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 3.0267450039366437e-05,
+      "loss": 2.0585,
+      "step": 13639000
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 3.026672639171916e-05,
+      "loss": 2.0743,
+      "step": 13639500
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 3.026600274407188e-05,
+      "loss": 2.0545,
+      "step": 13640000
+    },
+    {
+      "epoch": 39.48,
+      "learning_rate": 3.0265280543719897e-05,
+      "loss": 2.0243,
+      "step": 13640500
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 3.0264556896072622e-05,
+      "loss": 2.0414,
+      "step": 13641000
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 3.0263833248425345e-05,
+      "loss": 2.0474,
+      "step": 13641500
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 3.0263109600778067e-05,
+      "loss": 2.0397,
+      "step": 13642000
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 3.026238595313079e-05,
+      "loss": 2.045,
+      "step": 13642500
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 3.026166230548351e-05,
+      "loss": 2.0526,
+      "step": 13643000
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 3.0260938657836237e-05,
+      "loss": 2.0616,
+      "step": 13643500
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 3.026021501018896e-05,
+      "loss": 2.0591,
+      "step": 13644000
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 3.025949136254168e-05,
+      "loss": 2.0612,
+      "step": 13644500
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 3.025876771489441e-05,
+      "loss": 2.0501,
+      "step": 13645000
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 3.0258044067247133e-05,
+      "loss": 2.0457,
+      "step": 13645500
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 3.0257321866895148e-05,
+      "loss": 2.0516,
+      "step": 13646000
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 3.0256598219247874e-05,
+      "loss": 2.0463,
+      "step": 13646500
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 3.0255874571600596e-05,
+      "loss": 2.0763,
+      "step": 13647000
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 3.0255150923953318e-05,
+      "loss": 2.0634,
+      "step": 13647500
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 3.025442727630604e-05,
+      "loss": 2.0368,
+      "step": 13648000
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 3.0253703628658763e-05,
+      "loss": 2.0761,
+      "step": 13648500
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 3.0252979981011488e-05,
+      "loss": 2.0394,
+      "step": 13649000
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 3.025225633336421e-05,
+      "loss": 2.0557,
+      "step": 13649500
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 3.0251534133012226e-05,
+      "loss": 2.0434,
+      "step": 13650000
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 3.025081337995554e-05,
+      "loss": 2.0481,
+      "step": 13650500
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 3.025008973230826e-05,
+      "loss": 2.0571,
+      "step": 13651000
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 3.0249366084660986e-05,
+      "loss": 2.0574,
+      "step": 13651500
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 3.024864243701371e-05,
+      "loss": 2.0454,
+      "step": 13652000
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 3.0247920236661724e-05,
+      "loss": 2.0788,
+      "step": 13652500
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 3.0247196589014453e-05,
+      "loss": 2.0495,
+      "step": 13653000
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 3.0246472941367175e-05,
+      "loss": 2.0395,
+      "step": 13653500
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 3.02457492937199e-05,
+      "loss": 2.0243,
+      "step": 13654000
+    },
+    {
+      "epoch": 39.52,
+      "learning_rate": 3.0245025646072623e-05,
+      "loss": 2.0571,
+      "step": 13654500
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 3.0244301998425345e-05,
+      "loss": 2.0565,
+      "step": 13655000
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 3.0243578350778068e-05,
+      "loss": 2.0479,
+      "step": 13655500
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 3.024285470313079e-05,
+      "loss": 2.0359,
+      "step": 13656000
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 3.0242131055483512e-05,
+      "loss": 2.0444,
+      "step": 13656500
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 3.0241407407836238e-05,
+      "loss": 2.0549,
+      "step": 13657000
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 3.024068376018896e-05,
+      "loss": 2.0387,
+      "step": 13657500
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 3.0239960112541682e-05,
+      "loss": 2.0752,
+      "step": 13658000
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 3.02392379121897e-05,
+      "loss": 2.0787,
+      "step": 13658500
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 3.0238514264542423e-05,
+      "loss": 2.0489,
+      "step": 13659000
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 3.0237790616895146e-05,
+      "loss": 2.0379,
+      "step": 13659500
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 3.0237066969247875e-05,
+      "loss": 2.0568,
+      "step": 13660000
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 3.0236343321600597e-05,
+      "loss": 2.0679,
+      "step": 13660500
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 3.023561967395332e-05,
+      "loss": 2.0428,
+      "step": 13661000
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 3.023489602630604e-05,
+      "loss": 2.0529,
+      "step": 13661500
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 3.0234172378658763e-05,
+      "loss": 2.0355,
+      "step": 13662000
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 3.0233450178306782e-05,
+      "loss": 2.0726,
+      "step": 13662500
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 3.0232726530659505e-05,
+      "loss": 2.0441,
+      "step": 13663000
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 3.0232002883012227e-05,
+      "loss": 2.0284,
+      "step": 13663500
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 3.0231279235364952e-05,
+      "loss": 2.0535,
+      "step": 13664000
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 3.0230557035012968e-05,
+      "loss": 2.048,
+      "step": 13664500
+    },
+    {
+      "epoch": 39.55,
+      "learning_rate": 3.022983338736569e-05,
+      "loss": 2.0665,
+      "step": 13665000
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 3.0229109739718413e-05,
+      "loss": 2.0609,
+      "step": 13665500
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 3.022838753936643e-05,
+      "loss": 2.0495,
+      "step": 13666000
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 3.0227663891719154e-05,
+      "loss": 2.0404,
+      "step": 13666500
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 3.0226940244071876e-05,
+      "loss": 2.0411,
+      "step": 13667000
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 3.0226216596424605e-05,
+      "loss": 2.0258,
+      "step": 13667500
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 3.0225492948777327e-05,
+      "loss": 2.0681,
+      "step": 13668000
+    },
+    {
+      "epoch": 39.56,
+      "learning_rate": 3.0224769301130053e-05,
+      "loss": 2.0536,
+      "step": 13668500
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 3.0224045653482775e-05,
+      "loss": 2.0712,
+      "step": 13669000
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 3.0223322005835497e-05,
+      "loss": 2.0581,
+      "step": 13669500
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 3.022259835818822e-05,
+      "loss": 2.0685,
+      "step": 13670000
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 3.022187615783624e-05,
+      "loss": 2.0571,
+      "step": 13670500
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 3.022115251018896e-05,
+      "loss": 2.0593,
+      "step": 13671000
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 3.0220428862541683e-05,
+      "loss": 2.0456,
+      "step": 13671500
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 3.0219705214894405e-05,
+      "loss": 2.0521,
+      "step": 13672000
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 3.0218981567247127e-05,
+      "loss": 2.0733,
+      "step": 13672500
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 3.0218257919599853e-05,
+      "loss": 2.0551,
+      "step": 13673000
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 3.0217534271952575e-05,
+      "loss": 2.0261,
+      "step": 13673500
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 3.0216810624305297e-05,
+      "loss": 2.0587,
+      "step": 13674000
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 3.0216086976658026e-05,
+      "loss": 2.0436,
+      "step": 13674500
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 3.021536332901075e-05,
+      "loss": 2.0973,
+      "step": 13675000
+    },
+    {
+      "epoch": 39.58,
+      "learning_rate": 3.0214641128658768e-05,
+      "loss": 2.0221,
+      "step": 13675500
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 3.021391748101149e-05,
+      "loss": 2.07,
+      "step": 13676000
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 3.0213193833364212e-05,
+      "loss": 2.0494,
+      "step": 13676500
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 3.0212470185716934e-05,
+      "loss": 2.0657,
+      "step": 13677000
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 3.0211746538069657e-05,
+      "loss": 2.0572,
+      "step": 13677500
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 3.021102289042238e-05,
+      "loss": 2.0581,
+      "step": 13678000
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 3.0210299242775104e-05,
+      "loss": 2.0645,
+      "step": 13678500
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 3.0209575595127827e-05,
+      "loss": 2.0328,
+      "step": 13679000
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 3.020885194748055e-05,
+      "loss": 2.0698,
+      "step": 13679500
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 3.0208129747128568e-05,
+      "loss": 2.0627,
+      "step": 13680000
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 3.0207407546776583e-05,
+      "loss": 2.055,
+      "step": 13680500
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 3.0206683899129306e-05,
+      "loss": 2.0599,
+      "step": 13681000
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 3.0205960251482028e-05,
+      "loss": 2.0349,
+      "step": 13681500
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 3.0205238051130047e-05,
+      "loss": 2.0645,
+      "step": 13682000
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 3.0204514403482776e-05,
+      "loss": 2.0492,
+      "step": 13682500
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 3.020379220313079e-05,
+      "loss": 2.0427,
+      "step": 13683000
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 3.0203068555483517e-05,
+      "loss": 2.0413,
+      "step": 13683500
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 3.020234490783624e-05,
+      "loss": 2.0575,
+      "step": 13684000
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 3.020162126018896e-05,
+      "loss": 2.0501,
+      "step": 13684500
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 3.020089905983698e-05,
+      "loss": 2.0483,
+      "step": 13685000
+    },
+    {
+      "epoch": 39.61,
+      "learning_rate": 3.0200175412189703e-05,
+      "loss": 2.0336,
+      "step": 13685500
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 3.0199451764542425e-05,
+      "loss": 2.0262,
+      "step": 13686000
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 3.0198728116895147e-05,
+      "loss": 2.0406,
+      "step": 13686500
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 3.019800446924787e-05,
+      "loss": 2.0149,
+      "step": 13687000
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 3.019728082160059e-05,
+      "loss": 2.0811,
+      "step": 13687500
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 3.0196557173953317e-05,
+      "loss": 2.0493,
+      "step": 13688000
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 3.019583352630604e-05,
+      "loss": 2.0499,
+      "step": 13688500
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 3.019510987865876e-05,
+      "loss": 2.0718,
+      "step": 13689000
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 3.0194386231011484e-05,
+      "loss": 2.0569,
+      "step": 13689500
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 3.0193662583364213e-05,
+      "loss": 2.0568,
+      "step": 13690000
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 3.0192938935716935e-05,
+      "loss": 2.0452,
+      "step": 13690500
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 3.0192216735364954e-05,
+      "loss": 2.0453,
+      "step": 13691000
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 3.019149453501297e-05,
+      "loss": 2.0511,
+      "step": 13691500
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 3.0190770887365695e-05,
+      "loss": 2.0402,
+      "step": 13692000
+    },
+    {
+      "epoch": 39.63,
+      "learning_rate": 3.019004868701371e-05,
+      "loss": 2.0575,
+      "step": 13692500
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 3.0189325039366433e-05,
+      "loss": 2.0482,
+      "step": 13693000
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 3.0188601391719155e-05,
+      "loss": 2.0625,
+      "step": 13693500
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 3.018787774407188e-05,
+      "loss": 2.0696,
+      "step": 13694000
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 3.0187154096424603e-05,
+      "loss": 2.0476,
+      "step": 13694500
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 3.0186430448777325e-05,
+      "loss": 2.07,
+      "step": 13695000
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 3.0185706801130048e-05,
+      "loss": 2.0452,
+      "step": 13695500
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 3.018498315348277e-05,
+      "loss": 2.0586,
+      "step": 13696000
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 3.0184259505835495e-05,
+      "loss": 2.0492,
+      "step": 13696500
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 3.0183535858188218e-05,
+      "loss": 2.0514,
+      "step": 13697000
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 3.0182812210540943e-05,
+      "loss": 2.0562,
+      "step": 13697500
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 3.018208856289367e-05,
+      "loss": 2.0712,
+      "step": 13698000
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 3.0181366362541684e-05,
+      "loss": 2.0622,
+      "step": 13698500
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 3.0180644162189703e-05,
+      "loss": 2.0478,
+      "step": 13699000
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 3.0179920514542426e-05,
+      "loss": 2.0524,
+      "step": 13699500
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 3.0179196866895148e-05,
+      "loss": 2.0614,
+      "step": 13700000
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 3.0178474666543167e-05,
+      "loss": 2.0506,
+      "step": 13700500
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 3.017775101889589e-05,
+      "loss": 2.0362,
+      "step": 13701000
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 3.017702737124861e-05,
+      "loss": 2.0628,
+      "step": 13701500
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 3.017630517089663e-05,
+      "loss": 2.0461,
+      "step": 13702000
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 3.0175581523249352e-05,
+      "loss": 2.0537,
+      "step": 13702500
+    },
+    {
+      "epoch": 39.66,
+      "learning_rate": 3.0174857875602075e-05,
+      "loss": 2.0636,
+      "step": 13703000
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 3.0174134227954797e-05,
+      "loss": 2.063,
+      "step": 13703500
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 3.017341058030752e-05,
+      "loss": 2.0412,
+      "step": 13704000
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 3.0172686932660245e-05,
+      "loss": 2.0472,
+      "step": 13704500
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 3.017196328501297e-05,
+      "loss": 2.0674,
+      "step": 13705000
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 3.0171239637365696e-05,
+      "loss": 2.0362,
+      "step": 13705500
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 3.0170515989718418e-05,
+      "loss": 2.0616,
+      "step": 13706000
+    },
+    {
+      "epoch": 39.67,
+      "learning_rate": 3.016979234207114e-05,
+      "loss": 2.0448,
+      "step": 13706500
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 3.0169068694423863e-05,
+      "loss": 2.0565,
+      "step": 13707000
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 3.0168345046776585e-05,
+      "loss": 2.0572,
+      "step": 13707500
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 3.0167621399129307e-05,
+      "loss": 2.0461,
+      "step": 13708000
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 3.0166897751482033e-05,
+      "loss": 2.0424,
+      "step": 13708500
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 3.0166174103834755e-05,
+      "loss": 2.0469,
+      "step": 13709000
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 3.0165450456187477e-05,
+      "loss": 2.0424,
+      "step": 13709500
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 3.01647268085402e-05,
+      "loss": 2.0461,
+      "step": 13710000
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 3.016400460818822e-05,
+      "loss": 2.0473,
+      "step": 13710500
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 3.0163282407836234e-05,
+      "loss": 2.0533,
+      "step": 13711000
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 3.016255876018896e-05,
+      "loss": 2.0687,
+      "step": 13711500
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 3.0161835112541682e-05,
+      "loss": 2.0482,
+      "step": 13712000
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 3.016111146489441e-05,
+      "loss": 2.0454,
+      "step": 13712500
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 3.0160389264542426e-05,
+      "loss": 2.0473,
+      "step": 13713000
+    },
+    {
+      "epoch": 39.69,
+      "learning_rate": 3.015966561689515e-05,
+      "loss": 2.0696,
+      "step": 13713500
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 3.0158943416543168e-05,
+      "loss": 2.0453,
+      "step": 13714000
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 3.015821976889589e-05,
+      "loss": 2.0533,
+      "step": 13714500
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 3.0157496121248612e-05,
+      "loss": 2.0645,
+      "step": 13715000
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 3.0156772473601334e-05,
+      "loss": 2.0718,
+      "step": 13715500
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 3.015604882595406e-05,
+      "loss": 2.0466,
+      "step": 13716000
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 3.0155325178306782e-05,
+      "loss": 2.0337,
+      "step": 13716500
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 3.0154601530659504e-05,
+      "loss": 2.0442,
+      "step": 13717000
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 3.0153877883012227e-05,
+      "loss": 2.0722,
+      "step": 13717500
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 3.015315423536495e-05,
+      "loss": 2.0482,
+      "step": 13718000
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 3.0152432035012968e-05,
+      "loss": 2.0397,
+      "step": 13718500
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 3.015170838736569e-05,
+      "loss": 2.049,
+      "step": 13719000
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 3.0150984739718412e-05,
+      "loss": 2.0454,
+      "step": 13719500
+    },
+    {
+      "epoch": 39.71,
+      "learning_rate": 3.015026109207114e-05,
+      "loss": 2.0836,
+      "step": 13720000
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 3.014953889171916e-05,
+      "loss": 2.057,
+      "step": 13720500
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 3.0148815244071883e-05,
+      "loss": 2.0813,
+      "step": 13721000
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 3.0148091596424605e-05,
+      "loss": 2.0463,
+      "step": 13721500
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 3.0147367948777327e-05,
+      "loss": 2.0317,
+      "step": 13722000
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 3.014664430113005e-05,
+      "loss": 2.0703,
+      "step": 13722500
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 3.0145920653482775e-05,
+      "loss": 2.0424,
+      "step": 13723000
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 3.0145197005835497e-05,
+      "loss": 2.0494,
+      "step": 13723500
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 3.014447335818822e-05,
+      "loss": 2.0561,
+      "step": 13724000
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 3.014374971054094e-05,
+      "loss": 2.0504,
+      "step": 13724500
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 3.0143026062893664e-05,
+      "loss": 2.0603,
+      "step": 13725000
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 3.0142302415246386e-05,
+      "loss": 2.0517,
+      "step": 13725500
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 3.014157876759911e-05,
+      "loss": 2.066,
+      "step": 13726000
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 3.0140855119951834e-05,
+      "loss": 2.0545,
+      "step": 13726500
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 3.0140131472304563e-05,
+      "loss": 2.0828,
+      "step": 13727000
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 3.0139407824657285e-05,
+      "loss": 2.0596,
+      "step": 13727500
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 3.01386856243053e-05,
+      "loss": 2.0505,
+      "step": 13728000
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 3.0137961976658023e-05,
+      "loss": 2.045,
+      "step": 13728500
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 3.013723832901075e-05,
+      "loss": 2.0467,
+      "step": 13729000
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 3.013651468136347e-05,
+      "loss": 2.0589,
+      "step": 13729500
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 3.0135791033716193e-05,
+      "loss": 2.0485,
+      "step": 13730000
+    },
+    {
+      "epoch": 39.74,
+      "learning_rate": 3.0135068833364212e-05,
+      "loss": 2.0648,
+      "step": 13730500
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 3.0134345185716934e-05,
+      "loss": 2.0425,
+      "step": 13731000
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 3.0133621538069656e-05,
+      "loss": 2.0805,
+      "step": 13731500
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 3.013289789042238e-05,
+      "loss": 2.0398,
+      "step": 13732000
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 3.01321742427751e-05,
+      "loss": 2.0479,
+      "step": 13732500
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 3.0131450595127823e-05,
+      "loss": 2.0431,
+      "step": 13733000
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 3.0130729842071135e-05,
+      "loss": 2.0371,
+      "step": 13733500
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 3.013000619442386e-05,
+      "loss": 2.0449,
+      "step": 13734000
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 3.0129282546776583e-05,
+      "loss": 2.0439,
+      "step": 13734500
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 3.0128558899129312e-05,
+      "loss": 2.0607,
+      "step": 13735000
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 3.0127835251482034e-05,
+      "loss": 2.0745,
+      "step": 13735500
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 3.0127111603834757e-05,
+      "loss": 2.0807,
+      "step": 13736000
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 3.012638795618748e-05,
+      "loss": 2.049,
+      "step": 13736500
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 3.0125665755835498e-05,
+      "loss": 2.0525,
+      "step": 13737000
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 3.012494210818822e-05,
+      "loss": 2.0582,
+      "step": 13737500
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 3.0124218460540942e-05,
+      "loss": 2.0532,
+      "step": 13738000
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 3.0123494812893665e-05,
+      "loss": 2.0567,
+      "step": 13738500
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 3.0122771165246387e-05,
+      "loss": 2.0426,
+      "step": 13739000
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 3.0122047517599112e-05,
+      "loss": 2.0665,
+      "step": 13739500
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 3.0121325317247128e-05,
+      "loss": 2.0633,
+      "step": 13740000
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 3.012060166959985e-05,
+      "loss": 2.0289,
+      "step": 13740500
+    },
+    {
+      "epoch": 39.77,
+      "learning_rate": 3.0119878021952576e-05,
+      "loss": 2.0387,
+      "step": 13741000
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 3.011915582160059e-05,
+      "loss": 2.0418,
+      "step": 13741500
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 3.0118432173953314e-05,
+      "loss": 2.041,
+      "step": 13742000
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 3.0117708526306043e-05,
+      "loss": 2.0174,
+      "step": 13742500
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 3.0116984878658765e-05,
+      "loss": 2.0437,
+      "step": 13743000
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 3.011626123101149e-05,
+      "loss": 2.0643,
+      "step": 13743500
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 3.0115537583364213e-05,
+      "loss": 2.0686,
+      "step": 13744000
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 3.0114813935716935e-05,
+      "loss": 2.0469,
+      "step": 13744500
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 3.011409173536495e-05,
+      "loss": 2.0761,
+      "step": 13745000
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 3.0113368087717676e-05,
+      "loss": 2.0775,
+      "step": 13745500
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 3.01126444400704e-05,
+      "loss": 2.0508,
+      "step": 13746000
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 3.011192079242312e-05,
+      "loss": 2.0474,
+      "step": 13746500
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 3.0111197144775843e-05,
+      "loss": 2.0491,
+      "step": 13747000
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 3.0110473497128565e-05,
+      "loss": 2.0442,
+      "step": 13747500
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 3.010974984948129e-05,
+      "loss": 2.0625,
+      "step": 13748000
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 3.0109026201834013e-05,
+      "loss": 2.0459,
+      "step": 13748500
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 3.010830400148203e-05,
+      "loss": 2.0337,
+      "step": 13749000
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 3.010758035383475e-05,
+      "loss": 2.0451,
+      "step": 13749500
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 3.010685670618748e-05,
+      "loss": 2.0486,
+      "step": 13750000
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 3.0106133058540202e-05,
+      "loss": 2.0596,
+      "step": 13750500
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 3.0105409410892928e-05,
+      "loss": 2.0416,
+      "step": 13751000
+    },
+    {
+      "epoch": 39.8,
+      "learning_rate": 3.010468576324565e-05,
+      "loss": 2.0561,
+      "step": 13751500
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 3.0103962115598372e-05,
+      "loss": 2.0574,
+      "step": 13752000
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 3.0103238467951094e-05,
+      "loss": 2.0529,
+      "step": 13752500
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 3.0102514820303816e-05,
+      "loss": 2.0485,
+      "step": 13753000
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 3.0101792619951835e-05,
+      "loss": 2.0502,
+      "step": 13753500
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 3.0101068972304558e-05,
+      "loss": 2.0579,
+      "step": 13754000
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 3.010034532465728e-05,
+      "loss": 2.0877,
+      "step": 13754500
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 3.0099621677010002e-05,
+      "loss": 2.0587,
+      "step": 13755000
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 3.009889947665802e-05,
+      "loss": 2.075,
+      "step": 13755500
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 3.0098175829010743e-05,
+      "loss": 2.0603,
+      "step": 13756000
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 3.0097452181363466e-05,
+      "loss": 2.0598,
+      "step": 13756500
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 3.0096728533716195e-05,
+      "loss": 2.0419,
+      "step": 13757000
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 3.0096004886068917e-05,
+      "loss": 2.0328,
+      "step": 13757500
+    },
+    {
+      "epoch": 39.82,
+      "learning_rate": 3.0095281238421642e-05,
+      "loss": 2.0643,
+      "step": 13758000
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 3.0094557590774365e-05,
+      "loss": 2.0506,
+      "step": 13758500
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 3.0093833943127087e-05,
+      "loss": 2.0588,
+      "step": 13759000
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 3.009311029547981e-05,
+      "loss": 2.037,
+      "step": 13759500
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 3.009238664783253e-05,
+      "loss": 2.0431,
+      "step": 13760000
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 3.0091663000185254e-05,
+      "loss": 2.0282,
+      "step": 13760500
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 3.009093935253798e-05,
+      "loss": 2.052,
+      "step": 13761000
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 3.0090217152185995e-05,
+      "loss": 2.0635,
+      "step": 13761500
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 3.0089494951834014e-05,
+      "loss": 2.078,
+      "step": 13762000
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 3.0088771304186736e-05,
+      "loss": 2.0675,
+      "step": 13762500
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 3.0088047656539458e-05,
+      "loss": 2.0581,
+      "step": 13763000
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 3.008732400889218e-05,
+      "loss": 2.0869,
+      "step": 13763500
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 3.0086600361244906e-05,
+      "loss": 2.0296,
+      "step": 13764000
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 3.008587671359763e-05,
+      "loss": 2.0463,
+      "step": 13764500
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 3.0085153065950354e-05,
+      "loss": 2.0711,
+      "step": 13765000
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 3.008442941830308e-05,
+      "loss": 2.0339,
+      "step": 13765500
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 3.0083707217951095e-05,
+      "loss": 2.0474,
+      "step": 13766000
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 3.0082983570303817e-05,
+      "loss": 2.0473,
+      "step": 13766500
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 3.0082259922656543e-05,
+      "loss": 2.0689,
+      "step": 13767000
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 3.0081536275009265e-05,
+      "loss": 2.0479,
+      "step": 13767500
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 3.0080812627361987e-05,
+      "loss": 2.0298,
+      "step": 13768000
+    },
+    {
+      "epoch": 39.85,
+      "learning_rate": 3.008008897971471e-05,
+      "loss": 2.0623,
+      "step": 13768500
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 3.0079365332067432e-05,
+      "loss": 2.0611,
+      "step": 13769000
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 3.0078641684420154e-05,
+      "loss": 2.0763,
+      "step": 13769500
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 3.0077919484068173e-05,
+      "loss": 2.0557,
+      "step": 13770000
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 3.0077197283716192e-05,
+      "loss": 2.0324,
+      "step": 13770500
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 3.0076473636068914e-05,
+      "loss": 2.0462,
+      "step": 13771000
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 3.0075749988421636e-05,
+      "loss": 2.0435,
+      "step": 13771500
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 3.0075026340774365e-05,
+      "loss": 2.0594,
+      "step": 13772000
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 3.0074302693127088e-05,
+      "loss": 2.0472,
+      "step": 13772500
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 3.007357904547981e-05,
+      "loss": 2.0491,
+      "step": 13773000
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 3.007285684512783e-05,
+      "loss": 2.06,
+      "step": 13773500
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 3.007213319748055e-05,
+      "loss": 2.0596,
+      "step": 13774000
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 3.0071409549833273e-05,
+      "loss": 2.0684,
+      "step": 13774500
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 3.0070685902185996e-05,
+      "loss": 2.0829,
+      "step": 13775000
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 3.0069962254538718e-05,
+      "loss": 2.0599,
+      "step": 13775500
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 3.0069238606891443e-05,
+      "loss": 2.0512,
+      "step": 13776000
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 3.006851640653946e-05,
+      "loss": 2.0426,
+      "step": 13776500
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 3.006779275889218e-05,
+      "loss": 2.034,
+      "step": 13777000
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 3.0067069111244907e-05,
+      "loss": 2.0528,
+      "step": 13777500
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 3.006634546359763e-05,
+      "loss": 2.0671,
+      "step": 13778000
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 3.0065623263245645e-05,
+      "loss": 2.0662,
+      "step": 13778500
+    },
+    {
+      "epoch": 39.88,
+      "learning_rate": 3.006489961559837e-05,
+      "loss": 2.0341,
+      "step": 13779000
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 3.0064175967951096e-05,
+      "loss": 2.0566,
+      "step": 13779500
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 3.006345232030382e-05,
+      "loss": 2.0755,
+      "step": 13780000
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 3.0062730119951837e-05,
+      "loss": 2.0419,
+      "step": 13780500
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 3.006200647230456e-05,
+      "loss": 2.0668,
+      "step": 13781000
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 3.006128282465728e-05,
+      "loss": 2.0723,
+      "step": 13781500
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 3.0060559177010007e-05,
+      "loss": 2.0432,
+      "step": 13782000
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 3.005983552936273e-05,
+      "loss": 2.0508,
+      "step": 13782500
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 3.005911188171545e-05,
+      "loss": 2.0564,
+      "step": 13783000
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 3.005838968136347e-05,
+      "loss": 2.0695,
+      "step": 13783500
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 3.0057666033716193e-05,
+      "loss": 2.0339,
+      "step": 13784000
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 3.0056942386068915e-05,
+      "loss": 2.0665,
+      "step": 13784500
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 3.0056218738421637e-05,
+      "loss": 2.0369,
+      "step": 13785000
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 3.005549509077436e-05,
+      "loss": 2.081,
+      "step": 13785500
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 3.0054771443127082e-05,
+      "loss": 2.0483,
+      "step": 13786000
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 3.00540492427751e-05,
+      "loss": 2.0534,
+      "step": 13786500
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 3.005332559512783e-05,
+      "loss": 2.0654,
+      "step": 13787000
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 3.0052601947480552e-05,
+      "loss": 2.056,
+      "step": 13787500
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 3.0051878299833274e-05,
+      "loss": 2.0607,
+      "step": 13788000
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 3.0051154652185996e-05,
+      "loss": 2.0395,
+      "step": 13788500
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 3.0050431004538722e-05,
+      "loss": 2.057,
+      "step": 13789000
+    },
+    {
+      "epoch": 39.91,
+      "learning_rate": 3.0049707356891444e-05,
+      "loss": 2.056,
+      "step": 13789500
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 3.0048983709244166e-05,
+      "loss": 2.0287,
+      "step": 13790000
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 3.0048261508892185e-05,
+      "loss": 2.0396,
+      "step": 13790500
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 3.0047537861244908e-05,
+      "loss": 2.0423,
+      "step": 13791000
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 3.004681421359763e-05,
+      "loss": 2.0727,
+      "step": 13791500
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 3.0046090565950352e-05,
+      "loss": 2.031,
+      "step": 13792000
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 3.0045366918303074e-05,
+      "loss": 2.0786,
+      "step": 13792500
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 3.0044643270655797e-05,
+      "loss": 2.0731,
+      "step": 13793000
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 3.0043921070303815e-05,
+      "loss": 2.0495,
+      "step": 13793500
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 3.0043197422656538e-05,
+      "loss": 2.0576,
+      "step": 13794000
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 3.004247522230456e-05,
+      "loss": 2.0412,
+      "step": 13794500
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 3.0041751574657286e-05,
+      "loss": 2.0529,
+      "step": 13795000
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 3.0041027927010008e-05,
+      "loss": 2.0499,
+      "step": 13795500
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 3.004030427936273e-05,
+      "loss": 2.0686,
+      "step": 13796000
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 3.0039580631715452e-05,
+      "loss": 2.0635,
+      "step": 13796500
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 3.0038856984068175e-05,
+      "loss": 2.0713,
+      "step": 13797000
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 3.0038133336420897e-05,
+      "loss": 2.048,
+      "step": 13797500
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 3.0037409688773622e-05,
+      "loss": 2.0541,
+      "step": 13798000
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 3.0036687488421638e-05,
+      "loss": 2.0624,
+      "step": 13798500
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 3.003596384077436e-05,
+      "loss": 2.0601,
+      "step": 13799000
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 3.0035240193127086e-05,
+      "loss": 2.0536,
+      "step": 13799500
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 3.00345179927751e-05,
+      "loss": 2.0662,
+      "step": 13800000
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 3.0033794345127824e-05,
+      "loss": 2.0678,
+      "step": 13800500
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 3.0033070697480546e-05,
+      "loss": 2.0624,
+      "step": 13801000
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 3.003234704983327e-05,
+      "loss": 2.0618,
+      "step": 13801500
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 3.0031623402185997e-05,
+      "loss": 2.0309,
+      "step": 13802000
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 3.0030899754538723e-05,
+      "loss": 2.0441,
+      "step": 13802500
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 3.0030176106891445e-05,
+      "loss": 2.0305,
+      "step": 13803000
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 3.0029452459244167e-05,
+      "loss": 2.0778,
+      "step": 13803500
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 3.002872881159689e-05,
+      "loss": 2.0494,
+      "step": 13804000
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 3.0028005163949612e-05,
+      "loss": 2.025,
+      "step": 13804500
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 3.002728296359763e-05,
+      "loss": 2.0497,
+      "step": 13805000
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 3.0026559315950353e-05,
+      "loss": 2.0632,
+      "step": 13805500
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 3.0025837115598372e-05,
+      "loss": 2.0593,
+      "step": 13806000
+    },
+    {
+      "epoch": 39.96,
+      "learning_rate": 3.0025113467951094e-05,
+      "loss": 2.0818,
+      "step": 13806500
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 3.0024389820303816e-05,
+      "loss": 2.0342,
+      "step": 13807000
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 3.0023667619951835e-05,
+      "loss": 2.0495,
+      "step": 13807500
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 3.0022943972304557e-05,
+      "loss": 2.0523,
+      "step": 13808000
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 3.002222032465728e-05,
+      "loss": 2.0411,
+      "step": 13808500
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 3.0021496677010002e-05,
+      "loss": 2.0529,
+      "step": 13809000
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 3.002077302936273e-05,
+      "loss": 2.047,
+      "step": 13809500
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 3.0020049381715453e-05,
+      "loss": 2.036,
+      "step": 13810000
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 3.0019325734068175e-05,
+      "loss": 2.0671,
+      "step": 13810500
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 3.00186020864209e-05,
+      "loss": 2.0604,
+      "step": 13811000
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 3.0017878438773623e-05,
+      "loss": 2.0671,
+      "step": 13811500
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 3.0017154791126345e-05,
+      "loss": 2.0569,
+      "step": 13812000
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 3.001643259077436e-05,
+      "loss": 2.0469,
+      "step": 13812500
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 3.001571039042238e-05,
+      "loss": 2.0436,
+      "step": 13813000
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 3.0014986742775102e-05,
+      "loss": 2.0437,
+      "step": 13813500
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 3.0014263095127824e-05,
+      "loss": 2.0407,
+      "step": 13814000
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 3.001353944748055e-05,
+      "loss": 2.0631,
+      "step": 13814500
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 3.0012815799833272e-05,
+      "loss": 2.0607,
+      "step": 13815000
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 3.0012092152185995e-05,
+      "loss": 2.0498,
+      "step": 13815500
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 3.0011368504538717e-05,
+      "loss": 2.062,
+      "step": 13816000
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 3.001064485689144e-05,
+      "loss": 2.0513,
+      "step": 13816500
+    },
+    {
+      "epoch": 39.99,
+      "learning_rate": 3.0009922656539465e-05,
+      "loss": 2.0367,
+      "step": 13817000
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 3.0009199008892187e-05,
+      "loss": 2.0887,
+      "step": 13817500
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 3.000847536124491e-05,
+      "loss": 2.0401,
+      "step": 13818000
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 3.000775171359763e-05,
+      "loss": 2.0678,
+      "step": 13818500
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.6710043969127615,
+      "eval_accuracy_mlm": 0.6363494969737239,
+      "eval_accuracy_nsp": 0.8568583776748109,
+      "eval_loss": 2.165112018585205,
+      "eval_runtime": 331.1724,
+      "eval_samples_per_second": 1317.7,
+      "eval_steps_per_second": 54.905,
+      "step": 13818880
     }
   ],
   "max_steps": 34547200,
   "num_train_epochs": 100,
-  "total_flos": 1.4199760659998734e+19,
+  "total_flos": 1.893300081164393e+19,
   "trial_name": null,
   "trial_params": null
 }