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": 20.0,
-  "global_step": 6909440,
+  "epoch": 30.0,
+  "global_step": 10364160,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -83134,11 +83134,41581 @@
       "eval_samples_per_second": 1321.142,
       "eval_steps_per_second": 55.048,
       "step": 6909440
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 4.000354732076695e-05,
+      "loss": 2.1754,
+      "step": 6909500
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 4.0002823673119674e-05,
+      "loss": 2.1769,
+      "step": 6910000
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 4.00021000254724e-05,
+      "loss": 2.175,
+      "step": 6910500
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 4.0001376377825125e-05,
+      "loss": 2.1955,
+      "step": 6911000
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 4.000065273017785e-05,
+      "loss": 2.1907,
+      "step": 6911500
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 3.999992908253057e-05,
+      "loss": 2.2082,
+      "step": 6912000
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 3.999920543488329e-05,
+      "loss": 2.1514,
+      "step": 6912500
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 3.9998481787236014e-05,
+      "loss": 2.2034,
+      "step": 6913000
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 3.999775958688403e-05,
+      "loss": 2.1719,
+      "step": 6913500
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 3.999703593923675e-05,
+      "loss": 2.1982,
+      "step": 6914000
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 3.9996312291589474e-05,
+      "loss": 2.1762,
+      "step": 6914500
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 3.99955886439422e-05,
+      "loss": 2.1888,
+      "step": 6915000
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 3.9994866443590225e-05,
+      "loss": 2.1714,
+      "step": 6915500
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 3.9994145690533534e-05,
+      "loss": 2.1863,
+      "step": 6916000
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 3.9993422042886256e-05,
+      "loss": 2.1737,
+      "step": 6916500
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 3.999269839523898e-05,
+      "loss": 2.1762,
+      "step": 6917000
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 3.99919747475917e-05,
+      "loss": 2.1594,
+      "step": 6917500
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 3.999125109994443e-05,
+      "loss": 2.1686,
+      "step": 6918000
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 3.999052745229715e-05,
+      "loss": 2.1785,
+      "step": 6918500
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 3.9989803804649874e-05,
+      "loss": 2.195,
+      "step": 6919000
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 3.99890801570026e-05,
+      "loss": 2.1416,
+      "step": 6919500
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 3.998835795665061e-05,
+      "loss": 2.1669,
+      "step": 6920000
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 3.9987634309003334e-05,
+      "loss": 2.202,
+      "step": 6920500
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 3.998691066135606e-05,
+      "loss": 2.199,
+      "step": 6921000
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 3.998618701370878e-05,
+      "loss": 2.159,
+      "step": 6921500
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 3.99854633660615e-05,
+      "loss": 2.1523,
+      "step": 6922000
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 3.9984741165709523e-05,
+      "loss": 2.1608,
+      "step": 6922500
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 3.9984017518062246e-05,
+      "loss": 2.1697,
+      "step": 6923000
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 3.9983293870414975e-05,
+      "loss": 2.2001,
+      "step": 6923500
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 3.99825702227677e-05,
+      "loss": 2.2104,
+      "step": 6924000
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 3.998184802241571e-05,
+      "loss": 2.1931,
+      "step": 6924500
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 3.9981124374768435e-05,
+      "loss": 2.191,
+      "step": 6925000
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 3.998040072712116e-05,
+      "loss": 2.2201,
+      "step": 6925500
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 3.997967707947388e-05,
+      "loss": 2.2081,
+      "step": 6926000
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 3.99789534318266e-05,
+      "loss": 2.2025,
+      "step": 6926500
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 3.997822978417933e-05,
+      "loss": 2.1786,
+      "step": 6927000
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 3.997750613653205e-05,
+      "loss": 2.1673,
+      "step": 6927500
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 3.9976782488884775e-05,
+      "loss": 2.1964,
+      "step": 6928000
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 3.997606028853279e-05,
+      "loss": 2.1654,
+      "step": 6928500
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 3.997533664088551e-05,
+      "loss": 2.1721,
+      "step": 6929000
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 3.9974612993238235e-05,
+      "loss": 2.1773,
+      "step": 6929500
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 3.997388934559096e-05,
+      "loss": 2.1785,
+      "step": 6930000
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 3.9973165697943686e-05,
+      "loss": 2.1914,
+      "step": 6930500
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 3.997244205029641e-05,
+      "loss": 2.1868,
+      "step": 6931000
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 3.997171840264913e-05,
+      "loss": 2.1811,
+      "step": 6931500
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 3.997099620229715e-05,
+      "loss": 2.1854,
+      "step": 6932000
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 3.9970272554649875e-05,
+      "loss": 2.1944,
+      "step": 6932500
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 3.99695489070026e-05,
+      "loss": 2.1912,
+      "step": 6933000
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 3.996882525935532e-05,
+      "loss": 2.1516,
+      "step": 6933500
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 3.996810161170804e-05,
+      "loss": 2.2039,
+      "step": 6934000
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 3.9967377964060764e-05,
+      "loss": 2.1754,
+      "step": 6934500
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 3.9966654316413486e-05,
+      "loss": 2.182,
+      "step": 6935000
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 3.996593066876621e-05,
+      "loss": 2.1789,
+      "step": 6935500
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 3.996520702111893e-05,
+      "loss": 2.1669,
+      "step": 6936000
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 3.996448482076695e-05,
+      "loss": 2.1872,
+      "step": 6936500
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 3.9963761173119675e-05,
+      "loss": 2.1969,
+      "step": 6937000
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 3.99630375254724e-05,
+      "loss": 2.1933,
+      "step": 6937500
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 3.996231532512042e-05,
+      "loss": 2.1991,
+      "step": 6938000
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 3.996159167747314e-05,
+      "loss": 2.1975,
+      "step": 6938500
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 3.9960868029825864e-05,
+      "loss": 2.1958,
+      "step": 6939000
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 3.996014438217859e-05,
+      "loss": 2.1826,
+      "step": 6939500
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 3.995942073453131e-05,
+      "loss": 2.1876,
+      "step": 6940000
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 3.995869708688403e-05,
+      "loss": 2.197,
+      "step": 6940500
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 3.995797343923675e-05,
+      "loss": 2.173,
+      "step": 6941000
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 3.995724979158948e-05,
+      "loss": 2.1933,
+      "step": 6941500
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 3.9956526143942205e-05,
+      "loss": 2.1779,
+      "step": 6942000
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 3.995580249629493e-05,
+      "loss": 2.1874,
+      "step": 6942500
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 3.995507884864765e-05,
+      "loss": 2.1934,
+      "step": 6943000
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 3.995435520100037e-05,
+      "loss": 2.2024,
+      "step": 6943500
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 3.995363300064839e-05,
+      "loss": 2.1691,
+      "step": 6944000
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 3.995290935300111e-05,
+      "loss": 2.2027,
+      "step": 6944500
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 3.995218570535383e-05,
+      "loss": 2.1865,
+      "step": 6945000
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 3.995146205770656e-05,
+      "loss": 2.1712,
+      "step": 6945500
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 3.995073841005928e-05,
+      "loss": 2.1801,
+      "step": 6946000
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 3.9950016209707305e-05,
+      "loss": 2.1931,
+      "step": 6946500
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 3.994929256206003e-05,
+      "loss": 2.1908,
+      "step": 6947000
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 3.994856891441275e-05,
+      "loss": 2.2075,
+      "step": 6947500
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 3.994784526676547e-05,
+      "loss": 2.1888,
+      "step": 6948000
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 3.994712306641349e-05,
+      "loss": 2.185,
+      "step": 6948500
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 3.994640086606151e-05,
+      "loss": 2.1887,
+      "step": 6949000
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 3.994567721841423e-05,
+      "loss": 2.162,
+      "step": 6949500
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 3.9944953570766954e-05,
+      "loss": 2.1662,
+      "step": 6950000
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 3.994423281771026e-05,
+      "loss": 2.1872,
+      "step": 6950500
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 3.9943509170062985e-05,
+      "loss": 2.1945,
+      "step": 6951000
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 3.994278552241571e-05,
+      "loss": 2.1905,
+      "step": 6951500
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 3.994206187476843e-05,
+      "loss": 2.1907,
+      "step": 6952000
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 3.994133822712116e-05,
+      "loss": 2.1891,
+      "step": 6952500
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 3.994061457947388e-05,
+      "loss": 2.165,
+      "step": 6953000
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 3.99398923791219e-05,
+      "loss": 2.1748,
+      "step": 6953500
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 3.993917017876992e-05,
+      "loss": 2.1902,
+      "step": 6954000
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 3.993844653112264e-05,
+      "loss": 2.1953,
+      "step": 6954500
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 3.993772288347536e-05,
+      "loss": 2.1898,
+      "step": 6955000
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 3.9936999235828085e-05,
+      "loss": 2.1984,
+      "step": 6955500
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 3.993627558818081e-05,
+      "loss": 2.1697,
+      "step": 6956000
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 3.9935551940533537e-05,
+      "loss": 2.1823,
+      "step": 6956500
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 3.993482829288626e-05,
+      "loss": 2.207,
+      "step": 6957000
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 3.993410464523898e-05,
+      "loss": 2.1829,
+      "step": 6957500
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 3.99333809975917e-05,
+      "loss": 2.1809,
+      "step": 6958000
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 3.993265879723972e-05,
+      "loss": 2.213,
+      "step": 6958500
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 3.993193514959244e-05,
+      "loss": 2.1971,
+      "step": 6959000
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 3.993121150194516e-05,
+      "loss": 2.1764,
+      "step": 6959500
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 3.9930487854297886e-05,
+      "loss": 2.1732,
+      "step": 6960000
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 3.992976420665061e-05,
+      "loss": 2.2331,
+      "step": 6960500
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 3.992904200629864e-05,
+      "loss": 2.1961,
+      "step": 6961000
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 3.992831835865136e-05,
+      "loss": 2.1823,
+      "step": 6961500
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 3.992759471100408e-05,
+      "loss": 2.1727,
+      "step": 6962000
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 3.9926871063356804e-05,
+      "loss": 2.1899,
+      "step": 6962500
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 3.9926147415709526e-05,
+      "loss": 2.1896,
+      "step": 6963000
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 3.992542376806225e-05,
+      "loss": 2.1848,
+      "step": 6963500
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 3.992470012041497e-05,
+      "loss": 2.1608,
+      "step": 6964000
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 3.992397647276769e-05,
+      "loss": 2.1977,
+      "step": 6964500
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 3.9923252825120415e-05,
+      "loss": 2.2078,
+      "step": 6965000
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 3.992252917747314e-05,
+      "loss": 2.1759,
+      "step": 6965500
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 3.992180552982586e-05,
+      "loss": 2.1903,
+      "step": 6966000
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 3.992108188217858e-05,
+      "loss": 2.1703,
+      "step": 6966500
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 3.992035823453131e-05,
+      "loss": 2.1843,
+      "step": 6967000
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 3.9919636034179326e-05,
+      "loss": 2.1936,
+      "step": 6967500
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 3.9918912386532055e-05,
+      "loss": 2.185,
+      "step": 6968000
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 3.991818873888478e-05,
+      "loss": 2.1762,
+      "step": 6968500
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 3.99174650912375e-05,
+      "loss": 2.1699,
+      "step": 6969000
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 3.991674144359022e-05,
+      "loss": 2.1831,
+      "step": 6969500
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 3.9916017795942944e-05,
+      "loss": 2.1411,
+      "step": 6970000
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 3.9915294148295666e-05,
+      "loss": 2.1941,
+      "step": 6970500
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 3.991457050064839e-05,
+      "loss": 2.1852,
+      "step": 6971000
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 3.991384830029641e-05,
+      "loss": 2.1898,
+      "step": 6971500
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 3.991312465264913e-05,
+      "loss": 2.1809,
+      "step": 6972000
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 3.9912401005001855e-05,
+      "loss": 2.1772,
+      "step": 6972500
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 3.991167880464987e-05,
+      "loss": 2.179,
+      "step": 6973000
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 3.9910956604297886e-05,
+      "loss": 2.1684,
+      "step": 6973500
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 3.991023295665061e-05,
+      "loss": 2.2027,
+      "step": 6974000
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 3.990950930900334e-05,
+      "loss": 2.1633,
+      "step": 6974500
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 3.990878566135606e-05,
+      "loss": 2.1725,
+      "step": 6975000
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 3.9908063461004075e-05,
+      "loss": 2.1788,
+      "step": 6975500
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 3.9907339813356804e-05,
+      "loss": 2.1985,
+      "step": 6976000
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 3.990661616570953e-05,
+      "loss": 2.1897,
+      "step": 6976500
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 3.990589251806225e-05,
+      "loss": 2.1774,
+      "step": 6977000
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 3.990516887041497e-05,
+      "loss": 2.166,
+      "step": 6977500
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 3.990444522276769e-05,
+      "loss": 2.2045,
+      "step": 6978000
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 3.9903721575120416e-05,
+      "loss": 2.202,
+      "step": 6978500
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 3.990299937476844e-05,
+      "loss": 2.1984,
+      "step": 6979000
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 3.990227572712116e-05,
+      "loss": 2.191,
+      "step": 6979500
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 3.990155207947388e-05,
+      "loss": 2.1859,
+      "step": 6980000
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 3.9900828431826605e-05,
+      "loss": 2.1798,
+      "step": 6980500
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 3.990010478417933e-05,
+      "loss": 2.1933,
+      "step": 6981000
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 3.989938113653205e-05,
+      "loss": 2.1866,
+      "step": 6981500
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 3.989865748888477e-05,
+      "loss": 2.1655,
+      "step": 6982000
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 3.9897933841237494e-05,
+      "loss": 2.1898,
+      "step": 6982500
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 3.989721164088551e-05,
+      "loss": 2.1785,
+      "step": 6983000
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 3.989648799323824e-05,
+      "loss": 2.1879,
+      "step": 6983500
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 3.989576434559096e-05,
+      "loss": 2.1776,
+      "step": 6984000
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 3.989504069794369e-05,
+      "loss": 2.1721,
+      "step": 6984500
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 3.989431705029641e-05,
+      "loss": 2.1749,
+      "step": 6985000
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 3.9893593402649134e-05,
+      "loss": 2.1844,
+      "step": 6985500
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 3.9892869755001856e-05,
+      "loss": 2.1808,
+      "step": 6986000
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 3.989214610735458e-05,
+      "loss": 2.1985,
+      "step": 6986500
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 3.9891423907002594e-05,
+      "loss": 2.1776,
+      "step": 6987000
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 3.9890700259355316e-05,
+      "loss": 2.2041,
+      "step": 6987500
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 3.988997661170804e-05,
+      "loss": 2.1811,
+      "step": 6988000
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 3.988925296406076e-05,
+      "loss": 2.1988,
+      "step": 6988500
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 3.988852931641349e-05,
+      "loss": 2.218,
+      "step": 6989000
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 3.9887807116061505e-05,
+      "loss": 2.1924,
+      "step": 6989500
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 3.988708346841423e-05,
+      "loss": 2.1868,
+      "step": 6990000
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 3.9886359820766956e-05,
+      "loss": 2.1678,
+      "step": 6990500
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 3.988563617311968e-05,
+      "loss": 2.1913,
+      "step": 6991000
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 3.98849125254724e-05,
+      "loss": 2.2153,
+      "step": 6991500
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 3.988418887782512e-05,
+      "loss": 2.1932,
+      "step": 6992000
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 3.988346667747314e-05,
+      "loss": 2.1721,
+      "step": 6992500
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 3.988274302982586e-05,
+      "loss": 2.1697,
+      "step": 6993000
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 3.988201938217859e-05,
+      "loss": 2.1609,
+      "step": 6993500
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 3.988129573453131e-05,
+      "loss": 2.1694,
+      "step": 6994000
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 3.9880572086884034e-05,
+      "loss": 2.1845,
+      "step": 6994500
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 3.9879848439236757e-05,
+      "loss": 2.1812,
+      "step": 6995000
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 3.987912479158948e-05,
+      "loss": 2.1769,
+      "step": 6995500
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 3.98784011439422e-05,
+      "loss": 2.207,
+      "step": 6996000
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 3.987767749629492e-05,
+      "loss": 2.1728,
+      "step": 6996500
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 3.9876953848647645e-05,
+      "loss": 2.2241,
+      "step": 6997000
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 3.987623164829566e-05,
+      "loss": 2.209,
+      "step": 6997500
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 3.987550944794369e-05,
+      "loss": 2.185,
+      "step": 6998000
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 3.9874787247591706e-05,
+      "loss": 2.175,
+      "step": 6998500
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 3.987406359994443e-05,
+      "loss": 2.195,
+      "step": 6999000
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 3.987333995229715e-05,
+      "loss": 2.1832,
+      "step": 6999500
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 3.987261630464987e-05,
+      "loss": 2.1897,
+      "step": 7000000
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 3.9871892657002595e-05,
+      "loss": 2.1876,
+      "step": 7000500
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 3.987117045665062e-05,
+      "loss": 2.202,
+      "step": 7001000
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 3.987044680900334e-05,
+      "loss": 2.1758,
+      "step": 7001500
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 3.9869724608651355e-05,
+      "loss": 2.1852,
+      "step": 7002000
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 3.986900096100408e-05,
+      "loss": 2.1671,
+      "step": 7002500
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 3.98682773133568e-05,
+      "loss": 2.1718,
+      "step": 7003000
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 3.986755366570952e-05,
+      "loss": 2.1998,
+      "step": 7003500
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 3.9866830018062244e-05,
+      "loss": 2.1956,
+      "step": 7004000
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 3.9866106370414966e-05,
+      "loss": 2.2042,
+      "step": 7004500
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 3.986538272276769e-05,
+      "loss": 2.1867,
+      "step": 7005000
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 3.986465907512042e-05,
+      "loss": 2.1871,
+      "step": 7005500
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 3.986393542747314e-05,
+      "loss": 2.1908,
+      "step": 7006000
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 3.986321177982587e-05,
+      "loss": 2.1811,
+      "step": 7006500
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 3.986248813217859e-05,
+      "loss": 2.1708,
+      "step": 7007000
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 3.9861765931826606e-05,
+      "loss": 2.1658,
+      "step": 7007500
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 3.986104228417933e-05,
+      "loss": 2.2123,
+      "step": 7008000
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 3.986031863653205e-05,
+      "loss": 2.1934,
+      "step": 7008500
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 3.985959498888477e-05,
+      "loss": 2.1881,
+      "step": 7009000
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 3.9858871341237495e-05,
+      "loss": 2.1759,
+      "step": 7009500
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 3.985814769359022e-05,
+      "loss": 2.1932,
+      "step": 7010000
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 3.985742404594294e-05,
+      "loss": 2.1948,
+      "step": 7010500
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 3.985670184559096e-05,
+      "loss": 2.1971,
+      "step": 7011000
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 3.9855978197943684e-05,
+      "loss": 2.2012,
+      "step": 7011500
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 3.98552559975917e-05,
+      "loss": 2.1917,
+      "step": 7012000
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 3.985453234994442e-05,
+      "loss": 2.1855,
+      "step": 7012500
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 3.9853808702297144e-05,
+      "loss": 2.1852,
+      "step": 7013000
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 3.985308505464987e-05,
+      "loss": 2.2068,
+      "step": 7013500
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 3.9852361407002595e-05,
+      "loss": 2.2028,
+      "step": 7014000
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 3.985163775935532e-05,
+      "loss": 2.1731,
+      "step": 7014500
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 3.985091411170804e-05,
+      "loss": 2.18,
+      "step": 7015000
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 3.985019046406077e-05,
+      "loss": 2.1947,
+      "step": 7015500
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 3.9849468263708784e-05,
+      "loss": 2.2185,
+      "step": 7016000
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 3.984874461606151e-05,
+      "loss": 2.1597,
+      "step": 7016500
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 3.984802096841423e-05,
+      "loss": 2.1912,
+      "step": 7017000
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 3.984729732076695e-05,
+      "loss": 2.1821,
+      "step": 7017500
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 3.984657367311967e-05,
+      "loss": 2.1817,
+      "step": 7018000
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 3.9845850025472396e-05,
+      "loss": 2.1752,
+      "step": 7018500
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 3.984512782512042e-05,
+      "loss": 2.2003,
+      "step": 7019000
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 3.984440417747314e-05,
+      "loss": 2.1833,
+      "step": 7019500
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 3.984368052982586e-05,
+      "loss": 2.1761,
+      "step": 7020000
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 3.984295688217859e-05,
+      "loss": 2.1692,
+      "step": 7020500
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 3.9842233234531314e-05,
+      "loss": 2.1949,
+      "step": 7021000
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 3.9841509586884036e-05,
+      "loss": 2.1685,
+      "step": 7021500
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 3.984078593923676e-05,
+      "loss": 2.1944,
+      "step": 7022000
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 3.984006229158948e-05,
+      "loss": 2.1733,
+      "step": 7022500
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 3.98393386439422e-05,
+      "loss": 2.1836,
+      "step": 7023000
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 3.983861644359022e-05,
+      "loss": 2.1937,
+      "step": 7023500
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 3.983789279594295e-05,
+      "loss": 2.2019,
+      "step": 7024000
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 3.983716914829567e-05,
+      "loss": 2.185,
+      "step": 7024500
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 3.983644550064839e-05,
+      "loss": 2.2017,
+      "step": 7025000
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 3.9835721853001114e-05,
+      "loss": 2.1772,
+      "step": 7025500
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 3.9834998205353836e-05,
+      "loss": 2.1999,
+      "step": 7026000
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 3.983427600500185e-05,
+      "loss": 2.1927,
+      "step": 7026500
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 3.9833552357354574e-05,
+      "loss": 2.1623,
+      "step": 7027000
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 3.9832830157002596e-05,
+      "loss": 2.1911,
+      "step": 7027500
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 3.983210650935532e-05,
+      "loss": 2.1765,
+      "step": 7028000
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 3.983138286170805e-05,
+      "loss": 2.1804,
+      "step": 7028500
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 3.983065921406077e-05,
+      "loss": 2.1886,
+      "step": 7029000
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 3.9829937013708785e-05,
+      "loss": 2.1703,
+      "step": 7029500
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 3.98292148133568e-05,
+      "loss": 2.1797,
+      "step": 7030000
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 3.982849116570952e-05,
+      "loss": 2.2094,
+      "step": 7030500
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 3.9827767518062245e-05,
+      "loss": 2.191,
+      "step": 7031000
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 3.982704387041497e-05,
+      "loss": 2.1692,
+      "step": 7031500
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 3.9826320222767697e-05,
+      "loss": 2.2012,
+      "step": 7032000
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 3.982559657512042e-05,
+      "loss": 2.1667,
+      "step": 7032500
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 3.982487292747314e-05,
+      "loss": 2.1905,
+      "step": 7033000
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 3.982414927982586e-05,
+      "loss": 2.1896,
+      "step": 7033500
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 3.9823425632178585e-05,
+      "loss": 2.1945,
+      "step": 7034000
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 3.982270198453131e-05,
+      "loss": 2.1896,
+      "step": 7034500
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 3.982197833688403e-05,
+      "loss": 2.2126,
+      "step": 7035000
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 3.982125468923676e-05,
+      "loss": 2.2072,
+      "step": 7035500
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 3.982053104158948e-05,
+      "loss": 2.1824,
+      "step": 7036000
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 3.98198073939422e-05,
+      "loss": 2.1857,
+      "step": 7036500
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 3.981908519359022e-05,
+      "loss": 2.1857,
+      "step": 7037000
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 3.981836154594295e-05,
+      "loss": 2.1818,
+      "step": 7037500
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 3.981763789829567e-05,
+      "loss": 2.174,
+      "step": 7038000
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 3.981691425064839e-05,
+      "loss": 2.154,
+      "step": 7038500
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 3.9816190603001115e-05,
+      "loss": 2.1931,
+      "step": 7039000
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 3.981546695535384e-05,
+      "loss": 2.1898,
+      "step": 7039500
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 3.981474330770656e-05,
+      "loss": 2.1961,
+      "step": 7040000
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 3.981401966005928e-05,
+      "loss": 2.1729,
+      "step": 7040500
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 3.9813296012412004e-05,
+      "loss": 2.1951,
+      "step": 7041000
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 3.9812572364764726e-05,
+      "loss": 2.1889,
+      "step": 7041500
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 3.981184871711745e-05,
+      "loss": 2.1871,
+      "step": 7042000
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 3.981112651676547e-05,
+      "loss": 2.2107,
+      "step": 7042500
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 3.98104028691182e-05,
+      "loss": 2.1899,
+      "step": 7043000
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 3.9809680668766215e-05,
+      "loss": 2.1718,
+      "step": 7043500
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 3.980895702111894e-05,
+      "loss": 2.1903,
+      "step": 7044000
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 3.980823337347166e-05,
+      "loss": 2.1603,
+      "step": 7044500
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 3.980750972582438e-05,
+      "loss": 2.1761,
+      "step": 7045000
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 3.9806786078177104e-05,
+      "loss": 2.1583,
+      "step": 7045500
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 3.9806062430529826e-05,
+      "loss": 2.1942,
+      "step": 7046000
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 3.980534023017785e-05,
+      "loss": 2.1957,
+      "step": 7046500
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 3.980461658253057e-05,
+      "loss": 2.194,
+      "step": 7047000
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 3.980389293488329e-05,
+      "loss": 2.1757,
+      "step": 7047500
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 3.9803169287236015e-05,
+      "loss": 2.1898,
+      "step": 7048000
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 3.980244563958874e-05,
+      "loss": 2.2045,
+      "step": 7048500
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 3.980172343923675e-05,
+      "loss": 2.1692,
+      "step": 7049000
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 3.9800999791589475e-05,
+      "loss": 2.1962,
+      "step": 7049500
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 3.98002761439422e-05,
+      "loss": 2.1654,
+      "step": 7050000
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 3.9799552496294926e-05,
+      "loss": 2.1826,
+      "step": 7050500
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 3.979882884864765e-05,
+      "loss": 2.189,
+      "step": 7051000
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 3.979810520100037e-05,
+      "loss": 2.203,
+      "step": 7051500
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 3.97973815533531e-05,
+      "loss": 2.1781,
+      "step": 7052000
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 3.979665790570582e-05,
+      "loss": 2.1805,
+      "step": 7052500
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 3.9795934258058544e-05,
+      "loss": 2.1746,
+      "step": 7053000
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 3.979521205770656e-05,
+      "loss": 2.1918,
+      "step": 7053500
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 3.979448841005928e-05,
+      "loss": 2.2041,
+      "step": 7054000
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 3.9793764762412004e-05,
+      "loss": 2.2015,
+      "step": 7054500
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 3.9793041114764727e-05,
+      "loss": 2.1931,
+      "step": 7055000
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 3.979231746711745e-05,
+      "loss": 2.1616,
+      "step": 7055500
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 3.979159381947017e-05,
+      "loss": 2.1781,
+      "step": 7056000
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 3.9790871619118193e-05,
+      "loss": 2.2021,
+      "step": 7056500
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 3.9790147971470916e-05,
+      "loss": 2.2019,
+      "step": 7057000
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 3.9789424323823645e-05,
+      "loss": 2.1837,
+      "step": 7057500
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 3.978870067617637e-05,
+      "loss": 2.1752,
+      "step": 7058000
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 3.978797702852909e-05,
+      "loss": 2.1906,
+      "step": 7058500
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 3.9787254828177105e-05,
+      "loss": 2.2004,
+      "step": 7059000
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 3.978653118052983e-05,
+      "loss": 2.1731,
+      "step": 7059500
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 3.978580753288255e-05,
+      "loss": 2.1845,
+      "step": 7060000
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 3.978508388523527e-05,
+      "loss": 2.1708,
+      "step": 7060500
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 3.9784361684883294e-05,
+      "loss": 2.2081,
+      "step": 7061000
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 3.978363948453131e-05,
+      "loss": 2.1721,
+      "step": 7061500
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 3.9782917284179325e-05,
+      "loss": 2.2073,
+      "step": 7062000
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 3.978219363653205e-05,
+      "loss": 2.1906,
+      "step": 7062500
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 3.9781469988884776e-05,
+      "loss": 2.1897,
+      "step": 7063000
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 3.97807463412375e-05,
+      "loss": 2.203,
+      "step": 7063500
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 3.978002269359022e-05,
+      "loss": 2.1499,
+      "step": 7064000
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 3.977929904594294e-05,
+      "loss": 2.2014,
+      "step": 7064500
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 3.9778575398295665e-05,
+      "loss": 2.1797,
+      "step": 7065000
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 3.9777851750648394e-05,
+      "loss": 2.1962,
+      "step": 7065500
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 3.9777128103001116e-05,
+      "loss": 2.1929,
+      "step": 7066000
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 3.977640590264913e-05,
+      "loss": 2.1796,
+      "step": 7066500
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 3.9775682255001854e-05,
+      "loss": 2.1689,
+      "step": 7067000
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 3.9774958607354576e-05,
+      "loss": 2.2087,
+      "step": 7067500
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 3.97742349597073e-05,
+      "loss": 2.2148,
+      "step": 7068000
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 3.977351131206003e-05,
+      "loss": 2.1938,
+      "step": 7068500
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 3.977278766441275e-05,
+      "loss": 2.2049,
+      "step": 7069000
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 3.977206401676547e-05,
+      "loss": 2.192,
+      "step": 7069500
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 3.977134181641349e-05,
+      "loss": 2.1895,
+      "step": 7070000
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 3.977061816876621e-05,
+      "loss": 2.1864,
+      "step": 7070500
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 3.976989452111893e-05,
+      "loss": 2.1879,
+      "step": 7071000
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 3.9769170873471654e-05,
+      "loss": 2.2107,
+      "step": 7071500
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 3.9768448673119677e-05,
+      "loss": 2.2035,
+      "step": 7072000
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 3.97677250254724e-05,
+      "loss": 2.1805,
+      "step": 7072500
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 3.976700137782513e-05,
+      "loss": 2.1792,
+      "step": 7073000
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 3.976627773017785e-05,
+      "loss": 2.1931,
+      "step": 7073500
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 3.9765555529825866e-05,
+      "loss": 2.2143,
+      "step": 7074000
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 3.976483188217859e-05,
+      "loss": 2.195,
+      "step": 7074500
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 3.976410823453131e-05,
+      "loss": 2.187,
+      "step": 7075000
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 3.976338458688403e-05,
+      "loss": 2.2038,
+      "step": 7075500
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 3.9762660939236755e-05,
+      "loss": 2.1757,
+      "step": 7076000
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 3.976193729158948e-05,
+      "loss": 2.2237,
+      "step": 7076500
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 3.97612136439422e-05,
+      "loss": 2.1807,
+      "step": 7077000
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 3.976048999629493e-05,
+      "loss": 2.1902,
+      "step": 7077500
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 3.975976634864765e-05,
+      "loss": 2.2015,
+      "step": 7078000
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 3.975904270100037e-05,
+      "loss": 2.1906,
+      "step": 7078500
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 3.975832050064839e-05,
+      "loss": 2.2108,
+      "step": 7079000
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 3.975759685300111e-05,
+      "loss": 2.2003,
+      "step": 7079500
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 3.975687320535383e-05,
+      "loss": 2.2145,
+      "step": 7080000
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 3.975614955770656e-05,
+      "loss": 2.2078,
+      "step": 7080500
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 3.9755425910059284e-05,
+      "loss": 2.2111,
+      "step": 7081000
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 3.9754703709707306e-05,
+      "loss": 2.1868,
+      "step": 7081500
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 3.975398006206003e-05,
+      "loss": 2.1746,
+      "step": 7082000
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 3.975325641441275e-05,
+      "loss": 2.1731,
+      "step": 7082500
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 3.975253276676547e-05,
+      "loss": 2.1799,
+      "step": 7083000
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 3.975181056641349e-05,
+      "loss": 2.1802,
+      "step": 7083500
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 3.975108691876621e-05,
+      "loss": 2.192,
+      "step": 7084000
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 3.975036327111893e-05,
+      "loss": 2.2028,
+      "step": 7084500
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 3.9749639623471655e-05,
+      "loss": 2.1753,
+      "step": 7085000
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 3.974891597582438e-05,
+      "loss": 2.1644,
+      "step": 7085500
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 3.9748192328177106e-05,
+      "loss": 2.1746,
+      "step": 7086000
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 3.974746868052983e-05,
+      "loss": 2.1841,
+      "step": 7086500
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 3.974674792747314e-05,
+      "loss": 2.2011,
+      "step": 7087000
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 3.974602427982586e-05,
+      "loss": 2.1821,
+      "step": 7087500
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 3.974530063217859e-05,
+      "loss": 2.1987,
+      "step": 7088000
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 3.974457698453131e-05,
+      "loss": 2.1808,
+      "step": 7088500
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 3.974385333688403e-05,
+      "loss": 2.2231,
+      "step": 7089000
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 3.9743129689236755e-05,
+      "loss": 2.2048,
+      "step": 7089500
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 3.974240604158948e-05,
+      "loss": 2.1878,
+      "step": 7090000
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 3.9741682393942207e-05,
+      "loss": 2.1903,
+      "step": 7090500
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 3.974096019359022e-05,
+      "loss": 2.2219,
+      "step": 7091000
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 3.9740236545942944e-05,
+      "loss": 2.1865,
+      "step": 7091500
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 3.9739512898295667e-05,
+      "loss": 2.1993,
+      "step": 7092000
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 3.973879069794368e-05,
+      "loss": 2.1897,
+      "step": 7092500
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 3.9738067050296404e-05,
+      "loss": 2.1883,
+      "step": 7093000
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 3.9737343402649127e-05,
+      "loss": 2.1794,
+      "step": 7093500
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 3.9736619755001856e-05,
+      "loss": 2.1779,
+      "step": 7094000
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 3.973589610735458e-05,
+      "loss": 2.1929,
+      "step": 7094500
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 3.97351724597073e-05,
+      "loss": 2.1846,
+      "step": 7095000
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 3.973444881206003e-05,
+      "loss": 2.1924,
+      "step": 7095500
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 3.9733726611708045e-05,
+      "loss": 2.1688,
+      "step": 7096000
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 3.973300296406077e-05,
+      "loss": 2.2002,
+      "step": 7096500
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 3.973227931641349e-05,
+      "loss": 2.1993,
+      "step": 7097000
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 3.973155566876621e-05,
+      "loss": 2.1994,
+      "step": 7097500
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 3.9730832021118934e-05,
+      "loss": 2.1679,
+      "step": 7098000
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 3.9730108373471656e-05,
+      "loss": 2.1774,
+      "step": 7098500
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 3.972938472582438e-05,
+      "loss": 2.1987,
+      "step": 7099000
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 3.972866107817711e-05,
+      "loss": 2.1917,
+      "step": 7099500
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 3.972793743052983e-05,
+      "loss": 2.2247,
+      "step": 7100000
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 3.972721378288255e-05,
+      "loss": 2.1916,
+      "step": 7100500
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 3.9726490135235274e-05,
+      "loss": 2.2065,
+      "step": 7101000
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 3.9725766487587996e-05,
+      "loss": 2.1739,
+      "step": 7101500
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 3.972504283994072e-05,
+      "loss": 2.1685,
+      "step": 7102000
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 3.972431919229345e-05,
+      "loss": 2.2143,
+      "step": 7102500
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 3.972359699194146e-05,
+      "loss": 2.1757,
+      "step": 7103000
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 3.9722873344294185e-05,
+      "loss": 2.2041,
+      "step": 7103500
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 3.972214969664691e-05,
+      "loss": 2.1868,
+      "step": 7104000
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 3.972142604899963e-05,
+      "loss": 2.1786,
+      "step": 7104500
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 3.972070240135236e-05,
+      "loss": 2.1694,
+      "step": 7105000
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 3.971997875370508e-05,
+      "loss": 2.1881,
+      "step": 7105500
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 3.97192551060578e-05,
+      "loss": 2.1967,
+      "step": 7106000
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 3.971853290570582e-05,
+      "loss": 2.1878,
+      "step": 7106500
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 3.9717810705353834e-05,
+      "loss": 2.1783,
+      "step": 7107000
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 3.9717087057706556e-05,
+      "loss": 2.1804,
+      "step": 7107500
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 3.971636341005928e-05,
+      "loss": 2.1845,
+      "step": 7108000
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 3.971563976241201e-05,
+      "loss": 2.179,
+      "step": 7108500
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 3.971491611476473e-05,
+      "loss": 2.178,
+      "step": 7109000
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 3.971419246711745e-05,
+      "loss": 2.1881,
+      "step": 7109500
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 3.971346881947018e-05,
+      "loss": 2.1806,
+      "step": 7110000
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 3.97127451718229e-05,
+      "loss": 2.1826,
+      "step": 7110500
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 3.9712021524175626e-05,
+      "loss": 2.1917,
+      "step": 7111000
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 3.971129787652835e-05,
+      "loss": 2.1799,
+      "step": 7111500
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 3.971057567617636e-05,
+      "loss": 2.2012,
+      "step": 7112000
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 3.9709853475824386e-05,
+      "loss": 2.1971,
+      "step": 7112500
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 3.970912982817711e-05,
+      "loss": 2.1988,
+      "step": 7113000
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 3.970840618052983e-05,
+      "loss": 2.1677,
+      "step": 7113500
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 3.970768253288255e-05,
+      "loss": 2.2067,
+      "step": 7114000
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 3.9706958885235275e-05,
+      "loss": 2.1956,
+      "step": 7114500
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 3.9706235237588e-05,
+      "loss": 2.1921,
+      "step": 7115000
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 3.970551158994072e-05,
+      "loss": 2.1682,
+      "step": 7115500
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 3.9704789389588735e-05,
+      "loss": 2.1757,
+      "step": 7116000
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 3.970406574194146e-05,
+      "loss": 2.2008,
+      "step": 7116500
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 3.9703342094294186e-05,
+      "loss": 2.2106,
+      "step": 7117000
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 3.970261844664691e-05,
+      "loss": 2.1789,
+      "step": 7117500
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 3.970189479899964e-05,
+      "loss": 2.2042,
+      "step": 7118000
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 3.970117115135236e-05,
+      "loss": 2.1878,
+      "step": 7118500
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 3.970044750370508e-05,
+      "loss": 2.1809,
+      "step": 7119000
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 3.9699723856057804e-05,
+      "loss": 2.1625,
+      "step": 7119500
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 3.969900165570582e-05,
+      "loss": 2.1908,
+      "step": 7120000
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 3.969827800805854e-05,
+      "loss": 2.209,
+      "step": 7120500
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 3.9697554360411264e-05,
+      "loss": 2.192,
+      "step": 7121000
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 3.9696830712763986e-05,
+      "loss": 2.187,
+      "step": 7121500
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 3.969610706511671e-05,
+      "loss": 2.2004,
+      "step": 7122000
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 3.969538486476473e-05,
+      "loss": 2.1871,
+      "step": 7122500
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 3.9694662664412746e-05,
+      "loss": 2.1944,
+      "step": 7123000
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 3.969393901676547e-05,
+      "loss": 2.1808,
+      "step": 7123500
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 3.9693216816413484e-05,
+      "loss": 2.207,
+      "step": 7124000
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 3.9692493168766206e-05,
+      "loss": 2.1933,
+      "step": 7124500
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 3.9691769521118935e-05,
+      "loss": 2.2003,
+      "step": 7125000
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 3.969104587347166e-05,
+      "loss": 2.1714,
+      "step": 7125500
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 3.9690322225824386e-05,
+      "loss": 2.1927,
+      "step": 7126000
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 3.968959857817711e-05,
+      "loss": 2.1798,
+      "step": 7126500
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 3.968887493052983e-05,
+      "loss": 2.1889,
+      "step": 7127000
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 3.968815128288255e-05,
+      "loss": 2.1951,
+      "step": 7127500
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 3.9687427635235275e-05,
+      "loss": 2.1823,
+      "step": 7128000
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 3.9686703987588e-05,
+      "loss": 2.1805,
+      "step": 7128500
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 3.968598033994072e-05,
+      "loss": 2.2102,
+      "step": 7129000
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 3.968525669229344e-05,
+      "loss": 2.1742,
+      "step": 7129500
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 3.9684533044646164e-05,
+      "loss": 2.1756,
+      "step": 7130000
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 3.9683809396998887e-05,
+      "loss": 2.1813,
+      "step": 7130500
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 3.968308574935161e-05,
+      "loss": 2.1899,
+      "step": 7131000
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 3.968236210170434e-05,
+      "loss": 2.1827,
+      "step": 7131500
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 3.968163845405706e-05,
+      "loss": 2.1851,
+      "step": 7132000
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 3.968091480640979e-05,
+      "loss": 2.1713,
+      "step": 7132500
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 3.9680192606057805e-05,
+      "loss": 2.2002,
+      "step": 7133000
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 3.967947040570582e-05,
+      "loss": 2.1963,
+      "step": 7133500
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 3.967874675805854e-05,
+      "loss": 2.1959,
+      "step": 7134000
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 3.9678023110411265e-05,
+      "loss": 2.1962,
+      "step": 7134500
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 3.967729946276399e-05,
+      "loss": 2.1916,
+      "step": 7135000
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 3.967657581511671e-05,
+      "loss": 2.1807,
+      "step": 7135500
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 3.967585216746944e-05,
+      "loss": 2.2127,
+      "step": 7136000
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 3.967512851982216e-05,
+      "loss": 2.1899,
+      "step": 7136500
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 3.967440487217488e-05,
+      "loss": 2.1724,
+      "step": 7137000
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 3.9673681224527605e-05,
+      "loss": 2.2133,
+      "step": 7137500
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 3.967295757688033e-05,
+      "loss": 2.1579,
+      "step": 7138000
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 3.967223392923305e-05,
+      "loss": 2.1966,
+      "step": 7138500
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 3.967151028158577e-05,
+      "loss": 2.1854,
+      "step": 7139000
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 3.96707866339385e-05,
+      "loss": 2.2129,
+      "step": 7139500
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 3.9670064433586516e-05,
+      "loss": 2.1787,
+      "step": 7140000
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 3.966934078593924e-05,
+      "loss": 2.1755,
+      "step": 7140500
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 3.966861713829196e-05,
+      "loss": 2.1842,
+      "step": 7141000
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 3.966789493793998e-05,
+      "loss": 2.1688,
+      "step": 7141500
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 3.9667171290292705e-05,
+      "loss": 2.1846,
+      "step": 7142000
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 3.966644764264543e-05,
+      "loss": 2.1878,
+      "step": 7142500
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 3.966572399499815e-05,
+      "loss": 2.2005,
+      "step": 7143000
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 3.966500034735087e-05,
+      "loss": 2.1735,
+      "step": 7143500
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 3.9664276699703594e-05,
+      "loss": 2.1811,
+      "step": 7144000
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 3.9663553052056316e-05,
+      "loss": 2.1703,
+      "step": 7144500
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 3.966282940440904e-05,
+      "loss": 2.2097,
+      "step": 7145000
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 3.966210575676176e-05,
+      "loss": 2.1886,
+      "step": 7145500
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 3.966138210911449e-05,
+      "loss": 2.1999,
+      "step": 7146000
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 3.9660659908762505e-05,
+      "loss": 2.1974,
+      "step": 7146500
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 3.9659936261115234e-05,
+      "loss": 2.1876,
+      "step": 7147000
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 3.965921406076325e-05,
+      "loss": 2.1884,
+      "step": 7147500
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 3.965849041311597e-05,
+      "loss": 2.1904,
+      "step": 7148000
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 3.965776821276399e-05,
+      "loss": 2.1854,
+      "step": 7148500
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 3.965704456511672e-05,
+      "loss": 2.189,
+      "step": 7149000
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 3.965632091746944e-05,
+      "loss": 2.1912,
+      "step": 7149500
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 3.965559726982216e-05,
+      "loss": 2.1867,
+      "step": 7150000
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 3.965487362217488e-05,
+      "loss": 2.1754,
+      "step": 7150500
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 3.9654149974527606e-05,
+      "loss": 2.1954,
+      "step": 7151000
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 3.965342632688033e-05,
+      "loss": 2.1747,
+      "step": 7151500
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 3.965270267923305e-05,
+      "loss": 2.1966,
+      "step": 7152000
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 3.965197903158577e-05,
+      "loss": 2.2262,
+      "step": 7152500
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 3.9651255383938494e-05,
+      "loss": 2.1811,
+      "step": 7153000
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 3.965053173629122e-05,
+      "loss": 2.2038,
+      "step": 7153500
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 3.964980808864394e-05,
+      "loss": 2.1834,
+      "step": 7154000
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 3.964908444099667e-05,
+      "loss": 2.1938,
+      "step": 7154500
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 3.964836079334939e-05,
+      "loss": 2.1805,
+      "step": 7155000
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 3.964763714570211e-05,
+      "loss": 2.1713,
+      "step": 7155500
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 3.9646914945350135e-05,
+      "loss": 2.2008,
+      "step": 7156000
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 3.964619274499815e-05,
+      "loss": 2.2036,
+      "step": 7156500
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 3.964546909735087e-05,
+      "loss": 2.2091,
+      "step": 7157000
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 3.964474689699889e-05,
+      "loss": 2.2039,
+      "step": 7157500
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 3.964402324935162e-05,
+      "loss": 2.2009,
+      "step": 7158000
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 3.964329960170434e-05,
+      "loss": 2.1991,
+      "step": 7158500
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 3.964257595405706e-05,
+      "loss": 2.1926,
+      "step": 7159000
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 3.9641852306409784e-05,
+      "loss": 2.1953,
+      "step": 7159500
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 3.9641128658762506e-05,
+      "loss": 2.1746,
+      "step": 7160000
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 3.964040501111523e-05,
+      "loss": 2.1994,
+      "step": 7160500
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 3.963968136346795e-05,
+      "loss": 2.183,
+      "step": 7161000
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 3.9638959163115966e-05,
+      "loss": 2.1718,
+      "step": 7161500
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 3.963823551546869e-05,
+      "loss": 2.2027,
+      "step": 7162000
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 3.963751186782142e-05,
+      "loss": 2.2022,
+      "step": 7162500
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 3.963678822017414e-05,
+      "loss": 2.1776,
+      "step": 7163000
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 3.963606457252687e-05,
+      "loss": 2.1903,
+      "step": 7163500
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 3.9635342372174884e-05,
+      "loss": 2.193,
+      "step": 7164000
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 3.9634618724527606e-05,
+      "loss": 2.2214,
+      "step": 7164500
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 3.963389507688033e-05,
+      "loss": 2.1937,
+      "step": 7165000
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 3.963317142923305e-05,
+      "loss": 2.1901,
+      "step": 7165500
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 3.963244778158577e-05,
+      "loss": 2.1934,
+      "step": 7166000
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 3.963172558123379e-05,
+      "loss": 2.1893,
+      "step": 7166500
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 3.963100193358652e-05,
+      "loss": 2.1762,
+      "step": 7167000
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 3.963027828593924e-05,
+      "loss": 2.1888,
+      "step": 7167500
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 3.962955463829196e-05,
+      "loss": 2.2067,
+      "step": 7168000
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 3.9628830990644684e-05,
+      "loss": 2.199,
+      "step": 7168500
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 3.96281087902927e-05,
+      "loss": 2.2017,
+      "step": 7169000
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 3.962738514264542e-05,
+      "loss": 2.1802,
+      "step": 7169500
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 3.962666149499815e-05,
+      "loss": 2.2185,
+      "step": 7170000
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 3.962593784735087e-05,
+      "loss": 2.185,
+      "step": 7170500
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 3.9625214199703596e-05,
+      "loss": 2.1653,
+      "step": 7171000
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 3.962449055205632e-05,
+      "loss": 2.2023,
+      "step": 7171500
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 3.962376690440904e-05,
+      "loss": 2.1924,
+      "step": 7172000
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 3.962304325676177e-05,
+      "loss": 2.2058,
+      "step": 7172500
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 3.962231960911449e-05,
+      "loss": 2.1963,
+      "step": 7173000
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 3.9621595961467214e-05,
+      "loss": 2.1966,
+      "step": 7173500
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 3.9620872313819936e-05,
+      "loss": 2.2057,
+      "step": 7174000
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 3.962014866617266e-05,
+      "loss": 2.1787,
+      "step": 7174500
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 3.9619426465820674e-05,
+      "loss": 2.198,
+      "step": 7175000
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 3.961870426546869e-05,
+      "loss": 2.1983,
+      "step": 7175500
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 3.961798206511671e-05,
+      "loss": 2.1938,
+      "step": 7176000
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 3.9617258417469434e-05,
+      "loss": 2.2027,
+      "step": 7176500
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 3.9616534769822156e-05,
+      "loss": 2.1954,
+      "step": 7177000
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 3.9615811122174885e-05,
+      "loss": 2.2239,
+      "step": 7177500
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 3.961508747452761e-05,
+      "loss": 2.1961,
+      "step": 7178000
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 3.961436382688033e-05,
+      "loss": 2.1883,
+      "step": 7178500
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 3.9613641626528345e-05,
+      "loss": 2.1904,
+      "step": 7179000
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 3.961291797888107e-05,
+      "loss": 2.2041,
+      "step": 7179500
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 3.9612194331233796e-05,
+      "loss": 2.1883,
+      "step": 7180000
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 3.961147068358652e-05,
+      "loss": 2.1977,
+      "step": 7180500
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 3.961074703593924e-05,
+      "loss": 2.194,
+      "step": 7181000
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 3.961002338829196e-05,
+      "loss": 2.1984,
+      "step": 7181500
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 3.9609299740644685e-05,
+      "loss": 2.1997,
+      "step": 7182000
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 3.96085775402927e-05,
+      "loss": 2.207,
+      "step": 7182500
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 3.960785389264542e-05,
+      "loss": 2.1786,
+      "step": 7183000
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 3.9607130244998145e-05,
+      "loss": 2.2072,
+      "step": 7183500
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 3.960640659735087e-05,
+      "loss": 2.1652,
+      "step": 7184000
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 3.960568439699889e-05,
+      "loss": 2.1916,
+      "step": 7184500
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 3.960496074935162e-05,
+      "loss": 2.1847,
+      "step": 7185000
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 3.960423710170434e-05,
+      "loss": 2.2219,
+      "step": 7185500
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 3.960351345405706e-05,
+      "loss": 2.1979,
+      "step": 7186000
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 3.9602789806409785e-05,
+      "loss": 2.1909,
+      "step": 7186500
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 3.960206615876251e-05,
+      "loss": 2.1851,
+      "step": 7187000
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 3.960134251111523e-05,
+      "loss": 2.2095,
+      "step": 7187500
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 3.960061886346795e-05,
+      "loss": 2.1729,
+      "step": 7188000
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 3.959989811041127e-05,
+      "loss": 2.2107,
+      "step": 7188500
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 3.959917446276399e-05,
+      "loss": 2.1775,
+      "step": 7189000
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 3.959845081511671e-05,
+      "loss": 2.1794,
+      "step": 7189500
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 3.9597727167469434e-05,
+      "loss": 2.1903,
+      "step": 7190000
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 3.959700351982216e-05,
+      "loss": 2.1969,
+      "step": 7190500
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 3.959627987217488e-05,
+      "loss": 2.2004,
+      "step": 7191000
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 3.95955562245276e-05,
+      "loss": 2.1858,
+      "step": 7191500
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 3.9594832576880323e-05,
+      "loss": 2.1923,
+      "step": 7192000
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 3.959410892923305e-05,
+      "loss": 2.1947,
+      "step": 7192500
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 3.9593385281585775e-05,
+      "loss": 2.1651,
+      "step": 7193000
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 3.95926616339385e-05,
+      "loss": 2.1864,
+      "step": 7193500
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 3.959193943358652e-05,
+      "loss": 2.1864,
+      "step": 7194000
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 3.959121578593924e-05,
+      "loss": 2.1666,
+      "step": 7194500
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 3.9590492138291964e-05,
+      "loss": 2.1699,
+      "step": 7195000
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 3.9589768490644686e-05,
+      "loss": 2.1882,
+      "step": 7195500
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 3.958904484299741e-05,
+      "loss": 2.1871,
+      "step": 7196000
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 3.9588322642645424e-05,
+      "loss": 2.1892,
+      "step": 7196500
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 3.9587598994998146e-05,
+      "loss": 2.2012,
+      "step": 7197000
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 3.958687534735087e-05,
+      "loss": 2.1832,
+      "step": 7197500
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 3.9586154594294184e-05,
+      "loss": 2.1717,
+      "step": 7198000
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 3.9585430946646906e-05,
+      "loss": 2.1965,
+      "step": 7198500
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 3.958470729899963e-05,
+      "loss": 2.2071,
+      "step": 7199000
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 3.958398365135235e-05,
+      "loss": 2.1903,
+      "step": 7199500
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 3.958326000370508e-05,
+      "loss": 2.1991,
+      "step": 7200000
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 3.95825363560578e-05,
+      "loss": 2.1824,
+      "step": 7200500
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 3.9581812708410524e-05,
+      "loss": 2.2161,
+      "step": 7201000
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 3.9581089060763246e-05,
+      "loss": 2.1802,
+      "step": 7201500
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 3.958036541311597e-05,
+      "loss": 2.2092,
+      "step": 7202000
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 3.95796417654687e-05,
+      "loss": 2.2037,
+      "step": 7202500
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 3.957891956511671e-05,
+      "loss": 2.2011,
+      "step": 7203000
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 3.9578195917469435e-05,
+      "loss": 2.2149,
+      "step": 7203500
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 3.957747371711745e-05,
+      "loss": 2.2025,
+      "step": 7204000
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 3.957675006947017e-05,
+      "loss": 2.1809,
+      "step": 7204500
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 3.9576027869118195e-05,
+      "loss": 2.1722,
+      "step": 7205000
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 3.957530422147092e-05,
+      "loss": 2.1973,
+      "step": 7205500
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 3.957458057382364e-05,
+      "loss": 2.1982,
+      "step": 7206000
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 3.957385692617636e-05,
+      "loss": 2.2047,
+      "step": 7206500
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 3.9573133278529084e-05,
+      "loss": 2.1824,
+      "step": 7207000
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 3.957240963088181e-05,
+      "loss": 2.1721,
+      "step": 7207500
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 3.9571685983234536e-05,
+      "loss": 2.1989,
+      "step": 7208000
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 3.957096233558726e-05,
+      "loss": 2.1662,
+      "step": 7208500
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 3.957023868793998e-05,
+      "loss": 2.1649,
+      "step": 7209000
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 3.95695150402927e-05,
+      "loss": 2.1563,
+      "step": 7209500
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 3.9568791392645424e-05,
+      "loss": 2.1625,
+      "step": 7210000
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 3.956806774499815e-05,
+      "loss": 2.1936,
+      "step": 7210500
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 3.956734554464617e-05,
+      "loss": 2.1889,
+      "step": 7211000
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 3.956662189699889e-05,
+      "loss": 2.2026,
+      "step": 7211500
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 3.9565898249351614e-05,
+      "loss": 2.1658,
+      "step": 7212000
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 3.9565174601704336e-05,
+      "loss": 2.181,
+      "step": 7212500
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 3.956445095405706e-05,
+      "loss": 2.2005,
+      "step": 7213000
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 3.956372730640978e-05,
+      "loss": 2.1768,
+      "step": 7213500
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 3.95630036587625e-05,
+      "loss": 2.1973,
+      "step": 7214000
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 3.9562281458410525e-05,
+      "loss": 2.1814,
+      "step": 7214500
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 3.956155781076325e-05,
+      "loss": 2.1893,
+      "step": 7215000
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 3.9560834163115976e-05,
+      "loss": 2.2131,
+      "step": 7215500
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 3.95601105154687e-05,
+      "loss": 2.2068,
+      "step": 7216000
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 3.955938686782142e-05,
+      "loss": 2.1922,
+      "step": 7216500
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 3.955866322017414e-05,
+      "loss": 2.1701,
+      "step": 7217000
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 3.9557939572526865e-05,
+      "loss": 2.1862,
+      "step": 7217500
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 3.955721592487959e-05,
+      "loss": 2.187,
+      "step": 7218000
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 3.955649227723231e-05,
+      "loss": 2.1803,
+      "step": 7218500
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 3.9555770076880325e-05,
+      "loss": 2.1939,
+      "step": 7219000
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 3.955504642923305e-05,
+      "loss": 2.1774,
+      "step": 7219500
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 3.955432422888107e-05,
+      "loss": 2.1759,
+      "step": 7220000
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 3.955360058123379e-05,
+      "loss": 2.2085,
+      "step": 7220500
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 3.9552876933586514e-05,
+      "loss": 2.1986,
+      "step": 7221000
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 3.9552153285939236e-05,
+      "loss": 2.2012,
+      "step": 7221500
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 3.955142963829196e-05,
+      "loss": 2.1802,
+      "step": 7222000
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 3.955070599064469e-05,
+      "loss": 2.2077,
+      "step": 7222500
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 3.95499837902927e-05,
+      "loss": 2.1825,
+      "step": 7223000
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 3.9549260142645425e-05,
+      "loss": 2.1874,
+      "step": 7223500
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 3.954853649499815e-05,
+      "loss": 2.1706,
+      "step": 7224000
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 3.9547812847350877e-05,
+      "loss": 2.1958,
+      "step": 7224500
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 3.95470891997036e-05,
+      "loss": 2.1795,
+      "step": 7225000
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 3.954636555205632e-05,
+      "loss": 2.1917,
+      "step": 7225500
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 3.954564190440904e-05,
+      "loss": 2.1903,
+      "step": 7226000
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 3.9544918256761765e-05,
+      "loss": 2.2002,
+      "step": 7226500
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 3.954419460911449e-05,
+      "loss": 2.1824,
+      "step": 7227000
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 3.954347096146721e-05,
+      "loss": 2.1861,
+      "step": 7227500
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 3.954274731381993e-05,
+      "loss": 2.1993,
+      "step": 7228000
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 3.9542023666172654e-05,
+      "loss": 2.2133,
+      "step": 7228500
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 3.954130001852538e-05,
+      "loss": 2.1809,
+      "step": 7229000
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 3.9540576370878106e-05,
+      "loss": 2.1779,
+      "step": 7229500
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 3.953985417052613e-05,
+      "loss": 2.1869,
+      "step": 7230000
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 3.953913052287885e-05,
+      "loss": 2.1685,
+      "step": 7230500
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 3.9538408322526866e-05,
+      "loss": 2.1714,
+      "step": 7231000
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 3.953768612217488e-05,
+      "loss": 2.1683,
+      "step": 7231500
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 3.9536962474527604e-05,
+      "loss": 2.1793,
+      "step": 7232000
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 3.9536238826880326e-05,
+      "loss": 2.1777,
+      "step": 7232500
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 3.953551517923305e-05,
+      "loss": 2.1841,
+      "step": 7233000
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 3.953479153158578e-05,
+      "loss": 2.2017,
+      "step": 7233500
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 3.953406933123379e-05,
+      "loss": 2.1756,
+      "step": 7234000
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 3.9533345683586515e-05,
+      "loss": 2.2047,
+      "step": 7234500
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 3.953262203593924e-05,
+      "loss": 2.1831,
+      "step": 7235000
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 3.953189838829196e-05,
+      "loss": 2.1749,
+      "step": 7235500
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 3.953117474064468e-05,
+      "loss": 2.194,
+      "step": 7236000
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 3.9530451092997404e-05,
+      "loss": 2.2104,
+      "step": 7236500
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 3.952972744535013e-05,
+      "loss": 2.2044,
+      "step": 7237000
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 3.9529003797702855e-05,
+      "loss": 2.1808,
+      "step": 7237500
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 3.952828159735088e-05,
+      "loss": 2.2064,
+      "step": 7238000
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 3.952755939699889e-05,
+      "loss": 2.1714,
+      "step": 7238500
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 3.9526835749351615e-05,
+      "loss": 2.1923,
+      "step": 7239000
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 3.952611210170434e-05,
+      "loss": 2.1771,
+      "step": 7239500
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 3.952538845405706e-05,
+      "loss": 2.191,
+      "step": 7240000
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 3.952466480640978e-05,
+      "loss": 2.1641,
+      "step": 7240500
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 3.9523941158762504e-05,
+      "loss": 2.1744,
+      "step": 7241000
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 3.9523217511115226e-05,
+      "loss": 2.2073,
+      "step": 7241500
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 3.9522493863467955e-05,
+      "loss": 2.1801,
+      "step": 7242000
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 3.952177021582068e-05,
+      "loss": 2.1941,
+      "step": 7242500
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 3.95210465681734e-05,
+      "loss": 2.186,
+      "step": 7243000
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 3.952032292052612e-05,
+      "loss": 2.178,
+      "step": 7243500
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 3.951960072017414e-05,
+      "loss": 2.1826,
+      "step": 7244000
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 3.9518877072526867e-05,
+      "loss": 2.1881,
+      "step": 7244500
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 3.951815342487959e-05,
+      "loss": 2.1954,
+      "step": 7245000
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 3.951742977723231e-05,
+      "loss": 2.1761,
+      "step": 7245500
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 3.951670612958503e-05,
+      "loss": 2.1957,
+      "step": 7246000
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 3.9515982481937755e-05,
+      "loss": 2.2072,
+      "step": 7246500
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 3.951526028158578e-05,
+      "loss": 2.1965,
+      "step": 7247000
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 3.95145366339385e-05,
+      "loss": 2.1911,
+      "step": 7247500
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 3.951381298629122e-05,
+      "loss": 2.2166,
+      "step": 7248000
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 3.9513089338643945e-05,
+      "loss": 2.2259,
+      "step": 7248500
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 3.951236569099667e-05,
+      "loss": 2.1975,
+      "step": 7249000
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 3.951164204334939e-05,
+      "loss": 2.188,
+      "step": 7249500
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 3.951091839570211e-05,
+      "loss": 2.2196,
+      "step": 7250000
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 3.9510194748054833e-05,
+      "loss": 2.2185,
+      "step": 7250500
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 3.950947399499815e-05,
+      "loss": 2.1839,
+      "step": 7251000
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 3.950875034735087e-05,
+      "loss": 2.168,
+      "step": 7251500
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 3.9508026699703594e-05,
+      "loss": 2.1618,
+      "step": 7252000
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 3.950730305205632e-05,
+      "loss": 2.1876,
+      "step": 7252500
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 3.950658085170434e-05,
+      "loss": 2.2013,
+      "step": 7253000
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 3.950585720405706e-05,
+      "loss": 2.2074,
+      "step": 7253500
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 3.950513355640978e-05,
+      "loss": 2.1825,
+      "step": 7254000
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 3.9504409908762505e-05,
+      "loss": 2.1969,
+      "step": 7254500
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.6605211557582317,
+      "eval_accuracy_mlm": 0.6243918605421886,
+      "eval_accuracy_nsp": 0.8541291425481111,
+      "eval_loss": 2.224362373352051,
+      "eval_runtime": 331.3457,
+      "eval_samples_per_second": 1317.011,
+      "eval_steps_per_second": 54.876,
+      "step": 7254912
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 3.950368626111523e-05,
+      "loss": 2.1811,
+      "step": 7255000
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 3.9502962613467956e-05,
+      "loss": 2.1592,
+      "step": 7255500
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 3.950223896582068e-05,
+      "loss": 2.1583,
+      "step": 7256000
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 3.95015153181734e-05,
+      "loss": 2.1849,
+      "step": 7256500
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 3.950079167052612e-05,
+      "loss": 2.1725,
+      "step": 7257000
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 3.950006947017414e-05,
+      "loss": 2.1677,
+      "step": 7257500
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 3.9499347269822154e-05,
+      "loss": 2.167,
+      "step": 7258000
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 3.949862362217488e-05,
+      "loss": 2.1521,
+      "step": 7258500
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 3.9497899974527605e-05,
+      "loss": 2.1557,
+      "step": 7259000
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 3.949717632688033e-05,
+      "loss": 2.1681,
+      "step": 7259500
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 3.9496452679233056e-05,
+      "loss": 2.1724,
+      "step": 7260000
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 3.949572903158578e-05,
+      "loss": 2.1526,
+      "step": 7260500
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 3.94950053839385e-05,
+      "loss": 2.1771,
+      "step": 7261000
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 3.949428173629122e-05,
+      "loss": 2.1626,
+      "step": 7261500
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 3.9493558088643945e-05,
+      "loss": 2.1695,
+      "step": 7262000
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 3.949283444099667e-05,
+      "loss": 2.1735,
+      "step": 7262500
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 3.949211079334939e-05,
+      "loss": 2.1681,
+      "step": 7263000
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 3.949138714570211e-05,
+      "loss": 2.17,
+      "step": 7263500
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 3.949066494535013e-05,
+      "loss": 2.1688,
+      "step": 7264000
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 3.9489941297702857e-05,
+      "loss": 2.1476,
+      "step": 7264500
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 3.948921765005558e-05,
+      "loss": 2.1606,
+      "step": 7265000
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 3.94884940024083e-05,
+      "loss": 2.1699,
+      "step": 7265500
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 3.948777180205632e-05,
+      "loss": 2.1625,
+      "step": 7266000
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 3.948704815440904e-05,
+      "loss": 2.1701,
+      "step": 7266500
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 3.9486325954057054e-05,
+      "loss": 2.155,
+      "step": 7267000
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 3.9485603753705084e-05,
+      "loss": 2.1811,
+      "step": 7267500
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 3.9484880106057806e-05,
+      "loss": 2.1702,
+      "step": 7268000
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 3.948415790570582e-05,
+      "loss": 2.1634,
+      "step": 7268500
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 3.9483434258058544e-05,
+      "loss": 2.1607,
+      "step": 7269000
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 3.9482710610411266e-05,
+      "loss": 2.164,
+      "step": 7269500
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 3.948198696276399e-05,
+      "loss": 2.1654,
+      "step": 7270000
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 3.948126331511671e-05,
+      "loss": 2.1881,
+      "step": 7270500
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 3.948053966746943e-05,
+      "loss": 2.1687,
+      "step": 7271000
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 3.9479816019822155e-05,
+      "loss": 2.1866,
+      "step": 7271500
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 3.9479092372174884e-05,
+      "loss": 2.1686,
+      "step": 7272000
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 3.9478368724527606e-05,
+      "loss": 2.1756,
+      "step": 7272500
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 3.947764507688033e-05,
+      "loss": 2.1684,
+      "step": 7273000
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 3.9476922876528344e-05,
+      "loss": 2.155,
+      "step": 7273500
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 3.9476199228881066e-05,
+      "loss": 2.1698,
+      "step": 7274000
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 3.947547558123379e-05,
+      "loss": 2.1788,
+      "step": 7274500
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 3.947475193358652e-05,
+      "loss": 2.1493,
+      "step": 7275000
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 3.947402828593924e-05,
+      "loss": 2.1658,
+      "step": 7275500
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 3.947330463829196e-05,
+      "loss": 2.1779,
+      "step": 7276000
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 3.9472580990644684e-05,
+      "loss": 2.1535,
+      "step": 7276500
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 3.9471857342997406e-05,
+      "loss": 2.1848,
+      "step": 7277000
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 3.9471133695350135e-05,
+      "loss": 2.1536,
+      "step": 7277500
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 3.947041149499815e-05,
+      "loss": 2.164,
+      "step": 7278000
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 3.946968784735087e-05,
+      "loss": 2.1742,
+      "step": 7278500
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 3.9468964199703595e-05,
+      "loss": 2.1703,
+      "step": 7279000
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 3.946824199935161e-05,
+      "loss": 2.1969,
+      "step": 7279500
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 3.946751835170433e-05,
+      "loss": 2.1698,
+      "step": 7280000
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 3.9466794704057055e-05,
+      "loss": 2.1673,
+      "step": 7280500
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 3.9466071056409784e-05,
+      "loss": 2.1703,
+      "step": 7281000
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 3.9465347408762506e-05,
+      "loss": 2.1922,
+      "step": 7281500
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 3.946462520841052e-05,
+      "loss": 2.1833,
+      "step": 7282000
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 3.946390156076325e-05,
+      "loss": 2.1768,
+      "step": 7282500
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 3.9463179360411267e-05,
+      "loss": 2.1654,
+      "step": 7283000
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 3.946245571276399e-05,
+      "loss": 2.1596,
+      "step": 7283500
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 3.946173206511671e-05,
+      "loss": 2.1583,
+      "step": 7284000
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 3.946100841746943e-05,
+      "loss": 2.1923,
+      "step": 7284500
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 3.946028476982216e-05,
+      "loss": 2.191,
+      "step": 7285000
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 3.9459561122174885e-05,
+      "loss": 2.1561,
+      "step": 7285500
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 3.94588389218229e-05,
+      "loss": 2.193,
+      "step": 7286000
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 3.945811527417562e-05,
+      "loss": 2.1985,
+      "step": 7286500
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 3.9457391626528345e-05,
+      "loss": 2.1653,
+      "step": 7287000
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 3.945666797888107e-05,
+      "loss": 2.162,
+      "step": 7287500
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 3.945594433123379e-05,
+      "loss": 2.1915,
+      "step": 7288000
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 3.945522068358651e-05,
+      "loss": 2.182,
+      "step": 7288500
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 3.9454497035939233e-05,
+      "loss": 2.1659,
+      "step": 7289000
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 3.945377338829196e-05,
+      "loss": 2.1941,
+      "step": 7289500
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 3.9453049740644685e-05,
+      "loss": 2.1575,
+      "step": 7290000
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 3.9452326092997414e-05,
+      "loss": 2.1747,
+      "step": 7290500
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 3.9451602445350136e-05,
+      "loss": 2.1626,
+      "step": 7291000
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 3.945087879770286e-05,
+      "loss": 2.1849,
+      "step": 7291500
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 3.945015515005558e-05,
+      "loss": 2.1592,
+      "step": 7292000
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 3.9449432949703596e-05,
+      "loss": 2.1604,
+      "step": 7292500
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 3.944870930205632e-05,
+      "loss": 2.1631,
+      "step": 7293000
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 3.944798565440904e-05,
+      "loss": 2.1726,
+      "step": 7293500
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 3.944726200676176e-05,
+      "loss": 2.1853,
+      "step": 7294000
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 3.9446539806409785e-05,
+      "loss": 2.1589,
+      "step": 7294500
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 3.944581615876251e-05,
+      "loss": 2.1762,
+      "step": 7295000
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 3.944509251111523e-05,
+      "loss": 2.1781,
+      "step": 7295500
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 3.944436886346795e-05,
+      "loss": 2.1752,
+      "step": 7296000
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 3.9443645215820674e-05,
+      "loss": 2.183,
+      "step": 7296500
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 3.94429215681734e-05,
+      "loss": 2.1667,
+      "step": 7297000
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 3.9442197920526125e-05,
+      "loss": 2.1706,
+      "step": 7297500
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 3.944147572017414e-05,
+      "loss": 2.1725,
+      "step": 7298000
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 3.944075207252686e-05,
+      "loss": 2.1753,
+      "step": 7298500
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 3.9440028424879585e-05,
+      "loss": 2.1702,
+      "step": 7299000
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 3.9439304777232314e-05,
+      "loss": 2.1848,
+      "step": 7299500
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 3.943858257688033e-05,
+      "loss": 2.1862,
+      "step": 7300000
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 3.943785892923305e-05,
+      "loss": 2.19,
+      "step": 7300500
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 3.943713672888107e-05,
+      "loss": 2.1662,
+      "step": 7301000
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 3.943641308123379e-05,
+      "loss": 2.1805,
+      "step": 7301500
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 3.943568943358651e-05,
+      "loss": 2.1605,
+      "step": 7302000
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 3.9434965785939234e-05,
+      "loss": 2.1541,
+      "step": 7302500
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 3.943424213829196e-05,
+      "loss": 2.1838,
+      "step": 7303000
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 3.9433518490644686e-05,
+      "loss": 2.1809,
+      "step": 7303500
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 3.943279484299741e-05,
+      "loss": 2.2009,
+      "step": 7304000
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 3.943207119535014e-05,
+      "loss": 2.1774,
+      "step": 7304500
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 3.943134754770286e-05,
+      "loss": 2.1718,
+      "step": 7305000
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 3.943062390005558e-05,
+      "loss": 2.1606,
+      "step": 7305500
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 3.9429900252408303e-05,
+      "loss": 2.1617,
+      "step": 7306000
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 3.9429176604761026e-05,
+      "loss": 2.15,
+      "step": 7306500
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 3.942845295711375e-05,
+      "loss": 2.1653,
+      "step": 7307000
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 3.942772930946647e-05,
+      "loss": 2.1616,
+      "step": 7307500
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 3.942700566181919e-05,
+      "loss": 2.1741,
+      "step": 7308000
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 3.9426283461467215e-05,
+      "loss": 2.1527,
+      "step": 7308500
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 3.942555981381994e-05,
+      "loss": 2.1781,
+      "step": 7309000
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 3.942483616617266e-05,
+      "loss": 2.1946,
+      "step": 7309500
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 3.942411251852538e-05,
+      "loss": 2.1558,
+      "step": 7310000
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 3.9423388870878104e-05,
+      "loss": 2.1607,
+      "step": 7310500
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 3.9422665223230826e-05,
+      "loss": 2.1964,
+      "step": 7311000
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 3.942194157558355e-05,
+      "loss": 2.1828,
+      "step": 7311500
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 3.942121792793628e-05,
+      "loss": 2.1622,
+      "step": 7312000
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 3.9420494280289e-05,
+      "loss": 2.1892,
+      "step": 7312500
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 3.9419772079937015e-05,
+      "loss": 2.1681,
+      "step": 7313000
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 3.941904843228974e-05,
+      "loss": 2.1733,
+      "step": 7313500
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 3.9418324784642466e-05,
+      "loss": 2.1789,
+      "step": 7314000
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 3.941760113699519e-05,
+      "loss": 2.1653,
+      "step": 7314500
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 3.9416878936643204e-05,
+      "loss": 2.1745,
+      "step": 7315000
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 3.9416155288995926e-05,
+      "loss": 2.1559,
+      "step": 7315500
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 3.941543164134865e-05,
+      "loss": 2.1852,
+      "step": 7316000
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 3.9414709440996664e-05,
+      "loss": 2.1647,
+      "step": 7316500
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 3.9413985793349386e-05,
+      "loss": 2.1863,
+      "step": 7317000
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 3.941326359299741e-05,
+      "loss": 2.1804,
+      "step": 7317500
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 3.941253994535013e-05,
+      "loss": 2.1875,
+      "step": 7318000
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 3.941181629770285e-05,
+      "loss": 2.1557,
+      "step": 7318500
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 3.9411092650055575e-05,
+      "loss": 2.1733,
+      "step": 7319000
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 3.9410369002408304e-05,
+      "loss": 2.1816,
+      "step": 7319500
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 3.9409645354761026e-05,
+      "loss": 2.1725,
+      "step": 7320000
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 3.940892170711375e-05,
+      "loss": 2.137,
+      "step": 7320500
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 3.940819805946647e-05,
+      "loss": 2.1726,
+      "step": 7321000
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 3.940747441181919e-05,
+      "loss": 2.1622,
+      "step": 7321500
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 3.9406750764171915e-05,
+      "loss": 2.1717,
+      "step": 7322000
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 3.940602711652464e-05,
+      "loss": 2.1578,
+      "step": 7322500
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 3.940530346887737e-05,
+      "loss": 2.1802,
+      "step": 7323000
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 3.940457982123009e-05,
+      "loss": 2.1847,
+      "step": 7323500
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 3.940385617358281e-05,
+      "loss": 2.1939,
+      "step": 7324000
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 3.940313252593553e-05,
+      "loss": 2.1643,
+      "step": 7324500
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 3.9402408878288256e-05,
+      "loss": 2.1956,
+      "step": 7325000
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 3.9401688125231564e-05,
+      "loss": 2.1687,
+      "step": 7325500
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 3.9400964477584293e-05,
+      "loss": 2.1542,
+      "step": 7326000
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 3.9400240829937016e-05,
+      "loss": 2.1785,
+      "step": 7326500
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 3.939951718228974e-05,
+      "loss": 2.1795,
+      "step": 7327000
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 3.939879353464247e-05,
+      "loss": 2.1856,
+      "step": 7327500
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 3.939806988699519e-05,
+      "loss": 2.1624,
+      "step": 7328000
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 3.939734623934791e-05,
+      "loss": 2.1646,
+      "step": 7328500
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 3.9396622591700634e-05,
+      "loss": 2.1746,
+      "step": 7329000
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 3.9395898944053356e-05,
+      "loss": 2.1867,
+      "step": 7329500
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 3.939517529640608e-05,
+      "loss": 2.1711,
+      "step": 7330000
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 3.93944516487588e-05,
+      "loss": 2.1838,
+      "step": 7330500
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 3.939372800111152e-05,
+      "loss": 2.169,
+      "step": 7331000
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 3.9393004353464245e-05,
+      "loss": 2.1612,
+      "step": 7331500
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 3.939228360040756e-05,
+      "loss": 2.1424,
+      "step": 7332000
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 3.939155995276028e-05,
+      "loss": 2.1659,
+      "step": 7332500
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 3.9390836305113005e-05,
+      "loss": 2.1778,
+      "step": 7333000
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 3.939011410476102e-05,
+      "loss": 2.1708,
+      "step": 7333500
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 3.938939045711374e-05,
+      "loss": 2.1804,
+      "step": 7334000
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 3.938866680946647e-05,
+      "loss": 2.2183,
+      "step": 7334500
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 3.9387943161819194e-05,
+      "loss": 2.1728,
+      "step": 7335000
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 3.9387219514171916e-05,
+      "loss": 2.1556,
+      "step": 7335500
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 3.9386495866524645e-05,
+      "loss": 2.1651,
+      "step": 7336000
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 3.938577221887737e-05,
+      "loss": 2.1995,
+      "step": 7336500
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 3.938505001852538e-05,
+      "loss": 2.1532,
+      "step": 7337000
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 3.9384326370878105e-05,
+      "loss": 2.177,
+      "step": 7337500
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 3.938360272323083e-05,
+      "loss": 2.1766,
+      "step": 7338000
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 3.938287907558355e-05,
+      "loss": 2.1945,
+      "step": 7338500
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 3.938215542793627e-05,
+      "loss": 2.1851,
+      "step": 7339000
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 3.9381431780288994e-05,
+      "loss": 2.177,
+      "step": 7339500
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 3.9380709579937017e-05,
+      "loss": 2.1821,
+      "step": 7340000
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 3.937998737958503e-05,
+      "loss": 2.1676,
+      "step": 7340500
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 3.9379263731937754e-05,
+      "loss": 2.1636,
+      "step": 7341000
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 3.9378540084290477e-05,
+      "loss": 2.1845,
+      "step": 7341500
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 3.9377816436643206e-05,
+      "loss": 2.1416,
+      "step": 7342000
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 3.937709278899593e-05,
+      "loss": 2.1879,
+      "step": 7342500
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 3.937636914134865e-05,
+      "loss": 2.1821,
+      "step": 7343000
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 3.937564549370137e-05,
+      "loss": 2.1761,
+      "step": 7343500
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 3.9374923293349395e-05,
+      "loss": 2.1915,
+      "step": 7344000
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 3.937419964570212e-05,
+      "loss": 2.1481,
+      "step": 7344500
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 3.937347599805484e-05,
+      "loss": 2.1783,
+      "step": 7345000
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 3.937275235040756e-05,
+      "loss": 2.1802,
+      "step": 7345500
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 3.9372028702760284e-05,
+      "loss": 2.1797,
+      "step": 7346000
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 3.9371305055113006e-05,
+      "loss": 2.1631,
+      "step": 7346500
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 3.937058140746573e-05,
+      "loss": 2.1934,
+      "step": 7347000
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 3.936985775981845e-05,
+      "loss": 2.1966,
+      "step": 7347500
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 3.9369135559466466e-05,
+      "loss": 2.1479,
+      "step": 7348000
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 3.9368411911819195e-05,
+      "loss": 2.1707,
+      "step": 7348500
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 3.936768826417192e-05,
+      "loss": 2.1543,
+      "step": 7349000
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 3.9366964616524646e-05,
+      "loss": 2.1768,
+      "step": 7349500
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 3.936624096887737e-05,
+      "loss": 2.2039,
+      "step": 7350000
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 3.936551732123009e-05,
+      "loss": 2.1764,
+      "step": 7350500
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 3.936479367358281e-05,
+      "loss": 2.2059,
+      "step": 7351000
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 3.936407147323083e-05,
+      "loss": 2.1905,
+      "step": 7351500
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 3.936334782558355e-05,
+      "loss": 2.2004,
+      "step": 7352000
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 3.936262417793627e-05,
+      "loss": 2.19,
+      "step": 7352500
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 3.9361900530288995e-05,
+      "loss": 2.1772,
+      "step": 7353000
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 3.936117688264172e-05,
+      "loss": 2.1938,
+      "step": 7353500
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 3.9360453234994446e-05,
+      "loss": 2.1745,
+      "step": 7354000
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 3.935972958734717e-05,
+      "loss": 2.1877,
+      "step": 7354500
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 3.935900593969989e-05,
+      "loss": 2.1442,
+      "step": 7355000
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 3.9358283739347906e-05,
+      "loss": 2.1824,
+      "step": 7355500
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 3.935756009170063e-05,
+      "loss": 2.1821,
+      "step": 7356000
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 3.935683644405336e-05,
+      "loss": 2.1921,
+      "step": 7356500
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 3.935611279640608e-05,
+      "loss": 2.1854,
+      "step": 7357000
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 3.9355390596054095e-05,
+      "loss": 2.1697,
+      "step": 7357500
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 3.9354666948406824e-05,
+      "loss": 2.1799,
+      "step": 7358000
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 3.9353943300759547e-05,
+      "loss": 2.1927,
+      "step": 7358500
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 3.935321965311227e-05,
+      "loss": 2.1768,
+      "step": 7359000
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 3.935249890005558e-05,
+      "loss": 2.1827,
+      "step": 7359500
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 3.935177669970359e-05,
+      "loss": 2.1646,
+      "step": 7360000
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 3.935105305205632e-05,
+      "loss": 2.1761,
+      "step": 7360500
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 3.9350329404409044e-05,
+      "loss": 2.1737,
+      "step": 7361000
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 3.934960575676177e-05,
+      "loss": 2.1669,
+      "step": 7361500
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 3.934888210911449e-05,
+      "loss": 2.1957,
+      "step": 7362000
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 3.9348159908762504e-05,
+      "loss": 2.2002,
+      "step": 7362500
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 3.934743626111523e-05,
+      "loss": 2.2162,
+      "step": 7363000
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 3.934671261346795e-05,
+      "loss": 2.2243,
+      "step": 7363500
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 3.934598896582067e-05,
+      "loss": 2.2011,
+      "step": 7364000
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 3.93452653181734e-05,
+      "loss": 2.1808,
+      "step": 7364500
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 3.934454167052612e-05,
+      "loss": 2.1931,
+      "step": 7365000
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 3.9343818022878845e-05,
+      "loss": 2.2043,
+      "step": 7365500
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 3.9343094375231574e-05,
+      "loss": 2.1645,
+      "step": 7366000
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 3.934237217487959e-05,
+      "loss": 2.1788,
+      "step": 7366500
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 3.934164852723231e-05,
+      "loss": 2.1872,
+      "step": 7367000
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 3.9340924879585034e-05,
+      "loss": 2.1893,
+      "step": 7367500
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 3.9340201231937756e-05,
+      "loss": 2.154,
+      "step": 7368000
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 3.933947903158577e-05,
+      "loss": 2.1713,
+      "step": 7368500
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 3.9338755383938494e-05,
+      "loss": 2.19,
+      "step": 7369000
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 3.933803173629122e-05,
+      "loss": 2.1574,
+      "step": 7369500
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 3.933730953593924e-05,
+      "loss": 2.167,
+      "step": 7370000
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 3.933658588829196e-05,
+      "loss": 2.189,
+      "step": 7370500
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 3.933586224064468e-05,
+      "loss": 2.1435,
+      "step": 7371000
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 3.9335138592997405e-05,
+      "loss": 2.21,
+      "step": 7371500
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 3.9334414945350134e-05,
+      "loss": 2.1827,
+      "step": 7372000
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 3.9333691297702856e-05,
+      "loss": 2.1837,
+      "step": 7372500
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 3.933296765005558e-05,
+      "loss": 2.1621,
+      "step": 7373000
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 3.93322440024083e-05,
+      "loss": 2.1642,
+      "step": 7373500
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 3.933152035476102e-05,
+      "loss": 2.1759,
+      "step": 7374000
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 3.9330796707113745e-05,
+      "loss": 2.1893,
+      "step": 7374500
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 3.9330073059466474e-05,
+      "loss": 2.1808,
+      "step": 7375000
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 3.9329349411819196e-05,
+      "loss": 2.1829,
+      "step": 7375500
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 3.932862576417192e-05,
+      "loss": 2.1685,
+      "step": 7376000
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 3.9327903563819934e-05,
+      "loss": 2.1903,
+      "step": 7376500
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 3.9327179916172656e-05,
+      "loss": 2.1887,
+      "step": 7377000
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 3.932645626852538e-05,
+      "loss": 2.1607,
+      "step": 7377500
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 3.93257326208781e-05,
+      "loss": 2.179,
+      "step": 7378000
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 3.932501042052612e-05,
+      "loss": 2.1683,
+      "step": 7378500
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 3.9324286772878845e-05,
+      "loss": 2.1657,
+      "step": 7379000
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 3.9323563125231574e-05,
+      "loss": 2.1847,
+      "step": 7379500
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 3.93228394775843e-05,
+      "loss": 2.1889,
+      "step": 7380000
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 3.932211582993702e-05,
+      "loss": 2.1722,
+      "step": 7380500
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 3.932139218228974e-05,
+      "loss": 2.164,
+      "step": 7381000
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 3.932066853464246e-05,
+      "loss": 2.1855,
+      "step": 7381500
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 3.9319944886995186e-05,
+      "loss": 2.1724,
+      "step": 7382000
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 3.931922123934791e-05,
+      "loss": 2.1618,
+      "step": 7382500
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 3.931849759170063e-05,
+      "loss": 2.1968,
+      "step": 7383000
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 3.931777539134865e-05,
+      "loss": 2.1908,
+      "step": 7383500
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 3.931705463829196e-05,
+      "loss": 2.1698,
+      "step": 7384000
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 3.9316330990644684e-05,
+      "loss": 2.1771,
+      "step": 7384500
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 3.9315607342997406e-05,
+      "loss": 2.1714,
+      "step": 7385000
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 3.931488369535013e-05,
+      "loss": 2.1799,
+      "step": 7385500
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 3.931416004770285e-05,
+      "loss": 2.166,
+      "step": 7386000
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 3.931343640005557e-05,
+      "loss": 2.1757,
+      "step": 7386500
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 3.93127127524083e-05,
+      "loss": 2.1987,
+      "step": 7387000
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 3.9311989104761024e-05,
+      "loss": 2.1875,
+      "step": 7387500
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 3.931126545711375e-05,
+      "loss": 2.1956,
+      "step": 7388000
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 3.9310541809466475e-05,
+      "loss": 2.1745,
+      "step": 7388500
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 3.93098181618192e-05,
+      "loss": 2.1945,
+      "step": 7389000
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 3.930909451417192e-05,
+      "loss": 2.1807,
+      "step": 7389500
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 3.930837086652464e-05,
+      "loss": 2.1674,
+      "step": 7390000
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 3.930764866617266e-05,
+      "loss": 2.1812,
+      "step": 7390500
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 3.930692501852538e-05,
+      "loss": 2.1649,
+      "step": 7391000
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 3.93062028181734e-05,
+      "loss": 2.1796,
+      "step": 7391500
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 3.9305479170526124e-05,
+      "loss": 2.1525,
+      "step": 7392000
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 3.9304755522878846e-05,
+      "loss": 2.1747,
+      "step": 7392500
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 3.930403187523157e-05,
+      "loss": 2.1768,
+      "step": 7393000
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 3.930330822758429e-05,
+      "loss": 2.1695,
+      "step": 7393500
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 3.930258457993701e-05,
+      "loss": 2.1586,
+      "step": 7394000
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 3.930186093228974e-05,
+      "loss": 2.1701,
+      "step": 7394500
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 3.9301137284642464e-05,
+      "loss": 2.1704,
+      "step": 7395000
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 3.9300413636995186e-05,
+      "loss": 2.1689,
+      "step": 7395500
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 3.929968998934791e-05,
+      "loss": 2.1827,
+      "step": 7396000
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 3.929896634170063e-05,
+      "loss": 2.1671,
+      "step": 7396500
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 3.929824269405335e-05,
+      "loss": 2.1846,
+      "step": 7397000
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 3.9297519046406075e-05,
+      "loss": 2.183,
+      "step": 7397500
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 3.9296795398758804e-05,
+      "loss": 2.1863,
+      "step": 7398000
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 3.9296071751111527e-05,
+      "loss": 2.1783,
+      "step": 7398500
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 3.929534810346425e-05,
+      "loss": 2.181,
+      "step": 7399000
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 3.929462445581697e-05,
+      "loss": 2.1829,
+      "step": 7399500
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 3.929390080816969e-05,
+      "loss": 2.1929,
+      "step": 7400000
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 3.9293177160522415e-05,
+      "loss": 2.1762,
+      "step": 7400500
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 3.929245351287514e-05,
+      "loss": 2.1899,
+      "step": 7401000
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 3.929173131252316e-05,
+      "loss": 2.1811,
+      "step": 7401500
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 3.929100766487588e-05,
+      "loss": 2.1795,
+      "step": 7402000
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 3.9290285464523905e-05,
+      "loss": 2.1811,
+      "step": 7402500
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 3.928956181687663e-05,
+      "loss": 2.1707,
+      "step": 7403000
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 3.928883816922935e-05,
+      "loss": 2.2064,
+      "step": 7403500
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 3.928811452158207e-05,
+      "loss": 2.1536,
+      "step": 7404000
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 3.9287390873934794e-05,
+      "loss": 2.2146,
+      "step": 7404500
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 3.92866701208781e-05,
+      "loss": 2.1962,
+      "step": 7405000
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 3.9285946473230825e-05,
+      "loss": 2.1564,
+      "step": 7405500
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 3.9285222825583554e-05,
+      "loss": 2.1938,
+      "step": 7406000
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 3.9284499177936276e-05,
+      "loss": 2.1974,
+      "step": 7406500
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 3.9283775530289e-05,
+      "loss": 2.1517,
+      "step": 7407000
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 3.928305188264172e-05,
+      "loss": 2.2067,
+      "step": 7407500
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 3.928232823499444e-05,
+      "loss": 2.1722,
+      "step": 7408000
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 3.9281604587347165e-05,
+      "loss": 2.1874,
+      "step": 7408500
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 3.9280880939699894e-05,
+      "loss": 2.1932,
+      "step": 7409000
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 3.9280157292052616e-05,
+      "loss": 2.1972,
+      "step": 7409500
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 3.927943364440534e-05,
+      "loss": 2.172,
+      "step": 7410000
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 3.9278711444053354e-05,
+      "loss": 2.1714,
+      "step": 7410500
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 3.9277987796406076e-05,
+      "loss": 2.1979,
+      "step": 7411000
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 3.9277264148758805e-05,
+      "loss": 2.2076,
+      "step": 7411500
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 3.927654050111153e-05,
+      "loss": 2.1689,
+      "step": 7412000
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 3.927581685346425e-05,
+      "loss": 2.1666,
+      "step": 7412500
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 3.9275094653112265e-05,
+      "loss": 2.1898,
+      "step": 7413000
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 3.927437100546499e-05,
+      "loss": 2.1852,
+      "step": 7413500
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 3.927364735781771e-05,
+      "loss": 2.1855,
+      "step": 7414000
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 3.927292371017043e-05,
+      "loss": 2.171,
+      "step": 7414500
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 3.9272200062523154e-05,
+      "loss": 2.1626,
+      "step": 7415000
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 3.9271476414875876e-05,
+      "loss": 2.1961,
+      "step": 7415500
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 3.92707542145239e-05,
+      "loss": 2.1506,
+      "step": 7416000
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 3.927003056687663e-05,
+      "loss": 2.2037,
+      "step": 7416500
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 3.926930691922935e-05,
+      "loss": 2.1585,
+      "step": 7417000
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 3.9268584718877365e-05,
+      "loss": 2.1756,
+      "step": 7417500
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 3.926786107123009e-05,
+      "loss": 2.1634,
+      "step": 7418000
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 3.926713742358281e-05,
+      "loss": 2.1433,
+      "step": 7418500
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 3.926641377593553e-05,
+      "loss": 2.1699,
+      "step": 7419000
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 3.9265691575583554e-05,
+      "loss": 2.168,
+      "step": 7419500
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 3.926496792793628e-05,
+      "loss": 2.2164,
+      "step": 7420000
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 3.9264244280289e-05,
+      "loss": 2.1976,
+      "step": 7420500
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 3.926352063264172e-05,
+      "loss": 2.1949,
+      "step": 7421000
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 3.926279843228974e-05,
+      "loss": 2.1881,
+      "step": 7421500
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 3.926207478464246e-05,
+      "loss": 2.183,
+      "step": 7422000
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 3.926135113699518e-05,
+      "loss": 2.21,
+      "step": 7422500
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 3.9260627489347903e-05,
+      "loss": 2.1824,
+      "step": 7423000
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 3.925990384170063e-05,
+      "loss": 2.1955,
+      "step": 7423500
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 3.9259180194053355e-05,
+      "loss": 2.1647,
+      "step": 7424000
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 3.9258456546406084e-05,
+      "loss": 2.173,
+      "step": 7424500
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 3.92577343460541e-05,
+      "loss": 2.197,
+      "step": 7425000
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 3.9257012145702115e-05,
+      "loss": 2.1935,
+      "step": 7425500
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 3.925628849805484e-05,
+      "loss": 2.1915,
+      "step": 7426000
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 3.925556485040756e-05,
+      "loss": 2.1668,
+      "step": 7426500
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 3.925484120276028e-05,
+      "loss": 2.1476,
+      "step": 7427000
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 3.9254117555113004e-05,
+      "loss": 2.2069,
+      "step": 7427500
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 3.925339390746573e-05,
+      "loss": 2.2022,
+      "step": 7428000
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 3.9252670259818455e-05,
+      "loss": 2.1494,
+      "step": 7428500
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 3.925194661217118e-05,
+      "loss": 2.1633,
+      "step": 7429000
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 3.92512229645239e-05,
+      "loss": 2.1751,
+      "step": 7429500
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 3.925049931687662e-05,
+      "loss": 2.1869,
+      "step": 7430000
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 3.9249775669229344e-05,
+      "loss": 2.1915,
+      "step": 7430500
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 3.9249052021582066e-05,
+      "loss": 2.1894,
+      "step": 7431000
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 3.9248328373934795e-05,
+      "loss": 2.1581,
+      "step": 7431500
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 3.924760472628752e-05,
+      "loss": 2.1725,
+      "step": 7432000
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 3.924688107864024e-05,
+      "loss": 2.2109,
+      "step": 7432500
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 3.924615743099296e-05,
+      "loss": 2.2039,
+      "step": 7433000
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 3.9245433783345684e-05,
+      "loss": 2.1859,
+      "step": 7433500
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 3.9244710135698406e-05,
+      "loss": 2.1771,
+      "step": 7434000
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 3.9243986488051135e-05,
+      "loss": 2.2055,
+      "step": 7434500
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 3.924326284040386e-05,
+      "loss": 2.1934,
+      "step": 7435000
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 3.924253919275658e-05,
+      "loss": 2.1834,
+      "step": 7435500
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 3.92418155451093e-05,
+      "loss": 2.1704,
+      "step": 7436000
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 3.924109334475732e-05,
+      "loss": 2.1682,
+      "step": 7436500
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 3.924036969711004e-05,
+      "loss": 2.1772,
+      "step": 7437000
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 3.923964604946276e-05,
+      "loss": 2.1624,
+      "step": 7437500
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 3.9238922401815484e-05,
+      "loss": 2.1864,
+      "step": 7438000
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 3.923820020146351e-05,
+      "loss": 2.179,
+      "step": 7438500
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 3.9237476553816236e-05,
+      "loss": 2.1987,
+      "step": 7439000
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 3.923675290616896e-05,
+      "loss": 2.2051,
+      "step": 7439500
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 3.923602925852168e-05,
+      "loss": 2.1821,
+      "step": 7440000
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 3.92353056108744e-05,
+      "loss": 2.1852,
+      "step": 7440500
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 3.9234581963227125e-05,
+      "loss": 2.1968,
+      "step": 7441000
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 3.923385831557985e-05,
+      "loss": 2.1977,
+      "step": 7441500
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 3.923313466793257e-05,
+      "loss": 2.2089,
+      "step": 7442000
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 3.923241102028529e-05,
+      "loss": 2.1745,
+      "step": 7442500
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 3.923168881993331e-05,
+      "loss": 2.178,
+      "step": 7443000
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 3.9230965172286036e-05,
+      "loss": 2.1747,
+      "step": 7443500
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 3.923024152463876e-05,
+      "loss": 2.1972,
+      "step": 7444000
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 3.922951787699148e-05,
+      "loss": 2.1925,
+      "step": 7444500
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 3.9228795676639496e-05,
+      "loss": 2.1751,
+      "step": 7445000
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 3.922807202899222e-05,
+      "loss": 2.1494,
+      "step": 7445500
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 3.922734838134495e-05,
+      "loss": 2.1624,
+      "step": 7446000
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 3.922662762828826e-05,
+      "loss": 2.163,
+      "step": 7446500
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 3.9225903980640985e-05,
+      "loss": 2.179,
+      "step": 7447000
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 3.9225181780289e-05,
+      "loss": 2.2142,
+      "step": 7447500
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 3.922445813264172e-05,
+      "loss": 2.1795,
+      "step": 7448000
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 3.9223734484994445e-05,
+      "loss": 2.1742,
+      "step": 7448500
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 3.922301083734717e-05,
+      "loss": 2.1664,
+      "step": 7449000
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 3.922228718969989e-05,
+      "loss": 2.1848,
+      "step": 7449500
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 3.922156354205261e-05,
+      "loss": 2.1794,
+      "step": 7450000
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 3.9220841341700634e-05,
+      "loss": 2.199,
+      "step": 7450500
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 3.9220117694053356e-05,
+      "loss": 2.1618,
+      "step": 7451000
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 3.921939404640608e-05,
+      "loss": 2.1731,
+      "step": 7451500
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 3.92186703987588e-05,
+      "loss": 2.1871,
+      "step": 7452000
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 3.921794675111152e-05,
+      "loss": 2.1632,
+      "step": 7452500
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 3.9217223103464245e-05,
+      "loss": 2.1524,
+      "step": 7453000
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 3.921649945581697e-05,
+      "loss": 2.1799,
+      "step": 7453500
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 3.9215775808169696e-05,
+      "loss": 2.2059,
+      "step": 7454000
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 3.921505216052242e-05,
+      "loss": 2.178,
+      "step": 7454500
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 3.921432851287514e-05,
+      "loss": 2.1574,
+      "step": 7455000
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 3.921360486522786e-05,
+      "loss": 2.1919,
+      "step": 7455500
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 3.9212881217580585e-05,
+      "loss": 2.1674,
+      "step": 7456000
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 3.9212157569933314e-05,
+      "loss": 2.188,
+      "step": 7456500
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 3.921143392228604e-05,
+      "loss": 2.191,
+      "step": 7457000
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 3.921071027463876e-05,
+      "loss": 2.1843,
+      "step": 7457500
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 3.920998662699148e-05,
+      "loss": 2.1864,
+      "step": 7458000
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 3.92092629793442e-05,
+      "loss": 2.2013,
+      "step": 7458500
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 3.920854222628751e-05,
+      "loss": 2.2088,
+      "step": 7459000
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 3.9207820025935535e-05,
+      "loss": 2.182,
+      "step": 7459500
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 3.920709637828826e-05,
+      "loss": 2.1812,
+      "step": 7460000
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 3.920637273064098e-05,
+      "loss": 2.2006,
+      "step": 7460500
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 3.92056490829937e-05,
+      "loss": 2.1763,
+      "step": 7461000
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 3.920492543534643e-05,
+      "loss": 2.1878,
+      "step": 7461500
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 3.9204203234994446e-05,
+      "loss": 2.1814,
+      "step": 7462000
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 3.920347958734717e-05,
+      "loss": 2.1871,
+      "step": 7462500
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 3.920275593969989e-05,
+      "loss": 2.1993,
+      "step": 7463000
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 3.920203229205261e-05,
+      "loss": 2.2042,
+      "step": 7463500
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 3.9201308644405335e-05,
+      "loss": 2.1742,
+      "step": 7464000
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 3.9200584996758064e-05,
+      "loss": 2.1776,
+      "step": 7464500
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 3.9199861349110786e-05,
+      "loss": 2.1785,
+      "step": 7465000
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 3.919913770146351e-05,
+      "loss": 2.1707,
+      "step": 7465500
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 3.919841405381623e-05,
+      "loss": 2.1471,
+      "step": 7466000
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 3.919769040616895e-05,
+      "loss": 2.1765,
+      "step": 7466500
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 3.9196966758521675e-05,
+      "loss": 2.1801,
+      "step": 7467000
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 3.91962431108744e-05,
+      "loss": 2.1702,
+      "step": 7467500
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 3.919551946322712e-05,
+      "loss": 2.1616,
+      "step": 7468000
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 3.919479581557985e-05,
+      "loss": 2.1438,
+      "step": 7468500
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 3.9194073615227864e-05,
+      "loss": 2.179,
+      "step": 7469000
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 3.9193349967580586e-05,
+      "loss": 2.171,
+      "step": 7469500
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 3.9192626319933315e-05,
+      "loss": 2.1588,
+      "step": 7470000
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 3.919190411958133e-05,
+      "loss": 2.1717,
+      "step": 7470500
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 3.919118047193405e-05,
+      "loss": 2.1756,
+      "step": 7471000
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 3.9190456824286775e-05,
+      "loss": 2.1669,
+      "step": 7471500
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 3.918973462393479e-05,
+      "loss": 2.156,
+      "step": 7472000
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 3.918901097628751e-05,
+      "loss": 2.1952,
+      "step": 7472500
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 3.9188288775935535e-05,
+      "loss": 2.1767,
+      "step": 7473000
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 3.918756512828826e-05,
+      "loss": 2.1598,
+      "step": 7473500
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 3.918684148064098e-05,
+      "loss": 2.1992,
+      "step": 7474000
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 3.91861178329937e-05,
+      "loss": 2.1774,
+      "step": 7474500
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 3.9185394185346424e-05,
+      "loss": 2.164,
+      "step": 7475000
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 3.9184670537699146e-05,
+      "loss": 2.1937,
+      "step": 7475500
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 3.918394689005187e-05,
+      "loss": 2.172,
+      "step": 7476000
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 3.91832232424046e-05,
+      "loss": 2.1713,
+      "step": 7476500
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 3.918249959475732e-05,
+      "loss": 2.1821,
+      "step": 7477000
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 3.918177594711004e-05,
+      "loss": 2.1878,
+      "step": 7477500
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 3.9181052299462764e-05,
+      "loss": 2.1677,
+      "step": 7478000
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 3.918032865181549e-05,
+      "loss": 2.1788,
+      "step": 7478500
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 3.9179605004168216e-05,
+      "loss": 2.173,
+      "step": 7479000
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 3.917888135652094e-05,
+      "loss": 2.158,
+      "step": 7479500
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 3.917815770887366e-05,
+      "loss": 2.1874,
+      "step": 7480000
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 3.917743406122638e-05,
+      "loss": 2.162,
+      "step": 7480500
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 3.9176710413579105e-05,
+      "loss": 2.1589,
+      "step": 7481000
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 3.917598821322712e-05,
+      "loss": 2.1896,
+      "step": 7481500
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 3.917526456557984e-05,
+      "loss": 2.1629,
+      "step": 7482000
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 3.9174540917932565e-05,
+      "loss": 2.1803,
+      "step": 7482500
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 3.917381727028529e-05,
+      "loss": 2.1863,
+      "step": 7483000
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 3.9173095069933316e-05,
+      "loss": 2.1846,
+      "step": 7483500
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 3.917237142228604e-05,
+      "loss": 2.1606,
+      "step": 7484000
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 3.917164777463876e-05,
+      "loss": 2.2064,
+      "step": 7484500
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 3.917092412699148e-05,
+      "loss": 2.1954,
+      "step": 7485000
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 3.9170200479344205e-05,
+      "loss": 2.2058,
+      "step": 7485500
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 3.916947683169693e-05,
+      "loss": 2.1758,
+      "step": 7486000
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 3.916875318404965e-05,
+      "loss": 2.1934,
+      "step": 7486500
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 3.916802953640237e-05,
+      "loss": 2.1885,
+      "step": 7487000
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 3.9167307336050394e-05,
+      "loss": 2.1658,
+      "step": 7487500
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 3.916658513569841e-05,
+      "loss": 2.1712,
+      "step": 7488000
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 3.916586148805113e-05,
+      "loss": 2.1918,
+      "step": 7488500
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 3.9165137840403854e-05,
+      "loss": 2.1972,
+      "step": 7489000
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 3.916441564005187e-05,
+      "loss": 2.1542,
+      "step": 7489500
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 3.916369199240459e-05,
+      "loss": 2.1748,
+      "step": 7490000
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 3.9162969792052614e-05,
+      "loss": 2.1849,
+      "step": 7490500
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 3.9162246144405336e-05,
+      "loss": 2.1759,
+      "step": 7491000
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 3.9161522496758065e-05,
+      "loss": 2.2153,
+      "step": 7491500
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 3.916079884911079e-05,
+      "loss": 2.1853,
+      "step": 7492000
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 3.916007520146351e-05,
+      "loss": 2.1965,
+      "step": 7492500
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 3.915935155381623e-05,
+      "loss": 2.1753,
+      "step": 7493000
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 3.9158627906168954e-05,
+      "loss": 2.1816,
+      "step": 7493500
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 3.9157904258521677e-05,
+      "loss": 2.1816,
+      "step": 7494000
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 3.91571806108744e-05,
+      "loss": 2.2008,
+      "step": 7494500
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 3.915645696322712e-05,
+      "loss": 2.1804,
+      "step": 7495000
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 3.915573331557984e-05,
+      "loss": 2.1724,
+      "step": 7495500
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 3.9155011115227866e-05,
+      "loss": 2.1961,
+      "step": 7496000
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 3.915428746758059e-05,
+      "loss": 2.1982,
+      "step": 7496500
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 3.915356381993331e-05,
+      "loss": 2.1869,
+      "step": 7497000
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 3.915284017228603e-05,
+      "loss": 2.1745,
+      "step": 7497500
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 3.9152116524638754e-05,
+      "loss": 2.1755,
+      "step": 7498000
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 3.9151392876991483e-05,
+      "loss": 2.1728,
+      "step": 7498500
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 3.9150669229344206e-05,
+      "loss": 2.1969,
+      "step": 7499000
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 3.914994558169693e-05,
+      "loss": 2.1751,
+      "step": 7499500
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 3.914922193404965e-05,
+      "loss": 2.1822,
+      "step": 7500000
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 3.914849828640237e-05,
+      "loss": 2.1737,
+      "step": 7500500
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 3.9147774638755095e-05,
+      "loss": 2.1687,
+      "step": 7501000
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 3.914705099110782e-05,
+      "loss": 2.1637,
+      "step": 7501500
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 3.9146327343460546e-05,
+      "loss": 2.1725,
+      "step": 7502000
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 3.914560369581327e-05,
+      "loss": 2.1874,
+      "step": 7502500
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 3.914488004816599e-05,
+      "loss": 2.1632,
+      "step": 7503000
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 3.9144157847814006e-05,
+      "loss": 2.1687,
+      "step": 7503500
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 3.914343564746202e-05,
+      "loss": 2.1903,
+      "step": 7504000
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 3.9142711999814744e-05,
+      "loss": 2.1855,
+      "step": 7504500
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 3.9141989799462766e-05,
+      "loss": 2.1708,
+      "step": 7505000
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 3.914126615181549e-05,
+      "loss": 2.1862,
+      "step": 7505500
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 3.914054250416822e-05,
+      "loss": 2.1881,
+      "step": 7506000
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 3.913981885652094e-05,
+      "loss": 2.1584,
+      "step": 7506500
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 3.913909520887366e-05,
+      "loss": 2.186,
+      "step": 7507000
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 3.9138371561226384e-05,
+      "loss": 2.1854,
+      "step": 7507500
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 3.9137647913579106e-05,
+      "loss": 2.2044,
+      "step": 7508000
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 3.913692426593183e-05,
+      "loss": 2.1793,
+      "step": 7508500
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 3.913620061828455e-05,
+      "loss": 2.1838,
+      "step": 7509000
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 3.913547697063727e-05,
+      "loss": 2.1703,
+      "step": 7509500
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 3.9134753322989995e-05,
+      "loss": 2.1703,
+      "step": 7510000
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 3.913402967534272e-05,
+      "loss": 2.189,
+      "step": 7510500
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 3.9133306027695446e-05,
+      "loss": 2.1984,
+      "step": 7511000
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 3.913258238004817e-05,
+      "loss": 2.1858,
+      "step": 7511500
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 3.9131860179696184e-05,
+      "loss": 2.1745,
+      "step": 7512000
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 3.9131136532048906e-05,
+      "loss": 2.1422,
+      "step": 7512500
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 3.913041288440163e-05,
+      "loss": 2.1975,
+      "step": 7513000
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 3.912968923675436e-05,
+      "loss": 2.1808,
+      "step": 7513500
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 3.912896703640237e-05,
+      "loss": 2.1855,
+      "step": 7514000
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 3.9128243388755095e-05,
+      "loss": 2.18,
+      "step": 7514500
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 3.912751974110782e-05,
+      "loss": 2.1831,
+      "step": 7515000
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 3.912679609346055e-05,
+      "loss": 2.198,
+      "step": 7515500
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 3.912607244581327e-05,
+      "loss": 2.1494,
+      "step": 7516000
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 3.9125350245461284e-05,
+      "loss": 2.1882,
+      "step": 7516500
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 3.912462659781401e-05,
+      "loss": 2.1999,
+      "step": 7517000
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 3.912390439746202e-05,
+      "loss": 2.1839,
+      "step": 7517500
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 3.9123180749814744e-05,
+      "loss": 2.178,
+      "step": 7518000
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 3.9122457102167474e-05,
+      "loss": 2.1761,
+      "step": 7518500
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 3.9121733454520196e-05,
+      "loss": 2.1891,
+      "step": 7519000
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 3.912100980687292e-05,
+      "loss": 2.213,
+      "step": 7519500
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 3.912028615922564e-05,
+      "loss": 2.1675,
+      "step": 7520000
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 3.911956251157836e-05,
+      "loss": 2.1744,
+      "step": 7520500
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 3.9118840311226385e-05,
+      "loss": 2.1674,
+      "step": 7521000
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 3.911811666357911e-05,
+      "loss": 2.1704,
+      "step": 7521500
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 3.911739301593183e-05,
+      "loss": 2.1946,
+      "step": 7522000
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 3.911666936828455e-05,
+      "loss": 2.194,
+      "step": 7522500
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 3.9115945720637274e-05,
+      "loss": 2.1863,
+      "step": 7523000
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 3.9115222072989996e-05,
+      "loss": 2.1545,
+      "step": 7523500
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 3.9114498425342725e-05,
+      "loss": 2.1872,
+      "step": 7524000
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 3.911377477769545e-05,
+      "loss": 2.1743,
+      "step": 7524500
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 3.9113054024638756e-05,
+      "loss": 2.1915,
+      "step": 7525000
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 3.911233037699148e-05,
+      "loss": 2.207,
+      "step": 7525500
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 3.91116067293442e-05,
+      "loss": 2.1635,
+      "step": 7526000
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 3.911088308169692e-05,
+      "loss": 2.1887,
+      "step": 7526500
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 3.9110160881344945e-05,
+      "loss": 2.1717,
+      "step": 7527000
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 3.910943723369767e-05,
+      "loss": 2.1795,
+      "step": 7527500
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 3.910871358605039e-05,
+      "loss": 2.1882,
+      "step": 7528000
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 3.910798993840312e-05,
+      "loss": 2.1758,
+      "step": 7528500
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 3.910726629075584e-05,
+      "loss": 2.1844,
+      "step": 7529000
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 3.910654264310856e-05,
+      "loss": 2.1769,
+      "step": 7529500
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 3.9105818995461285e-05,
+      "loss": 2.1757,
+      "step": 7530000
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 3.910509534781401e-05,
+      "loss": 2.166,
+      "step": 7530500
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 3.910437170016673e-05,
+      "loss": 2.1631,
+      "step": 7531000
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 3.910364805251945e-05,
+      "loss": 2.1754,
+      "step": 7531500
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 3.9102924404872174e-05,
+      "loss": 2.1829,
+      "step": 7532000
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 3.9102200757224896e-05,
+      "loss": 2.1779,
+      "step": 7532500
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 3.9101477109577625e-05,
+      "loss": 2.2018,
+      "step": 7533000
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 3.910075346193035e-05,
+      "loss": 2.1823,
+      "step": 7533500
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 3.910002981428307e-05,
+      "loss": 2.1655,
+      "step": 7534000
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 3.909930616663579e-05,
+      "loss": 2.1836,
+      "step": 7534500
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 3.9098582518988514e-05,
+      "loss": 2.1751,
+      "step": 7535000
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 3.9097858871341243e-05,
+      "loss": 2.1929,
+      "step": 7535500
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 3.909713667098926e-05,
+      "loss": 2.182,
+      "step": 7536000
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 3.909641302334198e-05,
+      "loss": 2.1838,
+      "step": 7536500
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 3.909569082299e-05,
+      "loss": 2.1738,
+      "step": 7537000
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 3.9094967175342726e-05,
+      "loss": 2.1727,
+      "step": 7537500
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 3.909424352769545e-05,
+      "loss": 2.1715,
+      "step": 7538000
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 3.909351988004817e-05,
+      "loss": 2.168,
+      "step": 7538500
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 3.909279623240089e-05,
+      "loss": 2.1789,
+      "step": 7539000
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 3.9092072584753615e-05,
+      "loss": 2.1871,
+      "step": 7539500
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 3.909134893710634e-05,
+      "loss": 2.1779,
+      "step": 7540000
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 3.909062528945906e-05,
+      "loss": 2.1981,
+      "step": 7540500
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 3.908990164181178e-05,
+      "loss": 2.1884,
+      "step": 7541000
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 3.90891794414598e-05,
+      "loss": 2.1672,
+      "step": 7541500
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 3.9088455793812526e-05,
+      "loss": 2.1553,
+      "step": 7542000
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 3.908773214616525e-05,
+      "loss": 2.1748,
+      "step": 7542500
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 3.908700849851798e-05,
+      "loss": 2.1748,
+      "step": 7543000
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 3.90862848508707e-05,
+      "loss": 2.1851,
+      "step": 7543500
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 3.908556120322342e-05,
+      "loss": 2.2001,
+      "step": 7544000
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 3.908483900287144e-05,
+      "loss": 2.1648,
+      "step": 7544500
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 3.908411535522416e-05,
+      "loss": 2.1833,
+      "step": 7545000
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 3.908339170757688e-05,
+      "loss": 2.1634,
+      "step": 7545500
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 3.9082668059929604e-05,
+      "loss": 2.2034,
+      "step": 7546000
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 3.9081944412282326e-05,
+      "loss": 2.1534,
+      "step": 7546500
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 3.908122076463505e-05,
+      "loss": 2.2166,
+      "step": 7547000
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 3.908049711698778e-05,
+      "loss": 2.1725,
+      "step": 7547500
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 3.907977491663579e-05,
+      "loss": 2.1905,
+      "step": 7548000
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 3.9079051268988515e-05,
+      "loss": 2.2052,
+      "step": 7548500
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 3.907832762134124e-05,
+      "loss": 2.2163,
+      "step": 7549000
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 3.907760397369396e-05,
+      "loss": 2.1789,
+      "step": 7549500
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 3.9076881773341975e-05,
+      "loss": 2.1577,
+      "step": 7550000
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 3.907615957299e-05,
+      "loss": 2.196,
+      "step": 7550500
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 3.9075435925342727e-05,
+      "loss": 2.173,
+      "step": 7551000
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 3.907471227769545e-05,
+      "loss": 2.193,
+      "step": 7551500
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 3.9073990077343464e-05,
+      "loss": 2.1713,
+      "step": 7552000
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 3.9073266429696187e-05,
+      "loss": 2.1857,
+      "step": 7552500
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 3.907254278204891e-05,
+      "loss": 2.1811,
+      "step": 7553000
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 3.907181913440163e-05,
+      "loss": 2.1852,
+      "step": 7553500
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 3.907109548675435e-05,
+      "loss": 2.1536,
+      "step": 7554000
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 3.9070371839107075e-05,
+      "loss": 2.1752,
+      "step": 7554500
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 3.9069648191459805e-05,
+      "loss": 2.1734,
+      "step": 7555000
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 3.906892454381253e-05,
+      "loss": 2.2085,
+      "step": 7555500
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 3.906820234346054e-05,
+      "loss": 2.2047,
+      "step": 7556000
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 3.9067478695813265e-05,
+      "loss": 2.1934,
+      "step": 7556500
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 3.906675504816599e-05,
+      "loss": 2.2059,
+      "step": 7557000
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 3.9066032847814e-05,
+      "loss": 2.1902,
+      "step": 7557500
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 3.9065309200166725e-05,
+      "loss": 2.1782,
+      "step": 7558000
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 3.9064585552519454e-05,
+      "loss": 2.1751,
+      "step": 7558500
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 3.9063861904872176e-05,
+      "loss": 2.1642,
+      "step": 7559000
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 3.9063138257224905e-05,
+      "loss": 2.1571,
+      "step": 7559500
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 3.906241460957763e-05,
+      "loss": 2.1717,
+      "step": 7560000
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 3.906169096193035e-05,
+      "loss": 2.1746,
+      "step": 7560500
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 3.906096731428307e-05,
+      "loss": 2.1788,
+      "step": 7561000
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 3.9060243666635794e-05,
+      "loss": 2.1843,
+      "step": 7561500
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 3.9059520018988516e-05,
+      "loss": 2.1483,
+      "step": 7562000
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 3.905879637134124e-05,
+      "loss": 2.1761,
+      "step": 7562500
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 3.905807272369396e-05,
+      "loss": 2.1484,
+      "step": 7563000
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 3.905734907604668e-05,
+      "loss": 2.1707,
+      "step": 7563500
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 3.9056626875694705e-05,
+      "loss": 2.1508,
+      "step": 7564000
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 3.905590322804743e-05,
+      "loss": 2.2057,
+      "step": 7564500
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 3.905517958040015e-05,
+      "loss": 2.1926,
+      "step": 7565000
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 3.905445593275288e-05,
+      "loss": 2.1812,
+      "step": 7565500
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 3.90537322851056e-05,
+      "loss": 2.2041,
+      "step": 7566000
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 3.9053010084753616e-05,
+      "loss": 2.1778,
+      "step": 7566500
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 3.905228643710634e-05,
+      "loss": 2.1626,
+      "step": 7567000
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 3.905156278945906e-05,
+      "loss": 2.1675,
+      "step": 7567500
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 3.905083914181178e-05,
+      "loss": 2.1802,
+      "step": 7568000
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 3.9050115494164505e-05,
+      "loss": 2.1899,
+      "step": 7568500
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 3.904939184651723e-05,
+      "loss": 2.1737,
+      "step": 7569000
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 3.9048668198869956e-05,
+      "loss": 2.1874,
+      "step": 7569500
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 3.904794455122268e-05,
+      "loss": 2.1981,
+      "step": 7570000
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 3.90472209035754e-05,
+      "loss": 2.1752,
+      "step": 7570500
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 3.9046498703223416e-05,
+      "loss": 2.2074,
+      "step": 7571000
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 3.904577650287143e-05,
+      "loss": 2.1879,
+      "step": 7571500
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 3.9045052855224154e-05,
+      "loss": 2.18,
+      "step": 7572000
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 3.9044329207576876e-05,
+      "loss": 2.1933,
+      "step": 7572500
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 3.9043605559929606e-05,
+      "loss": 2.1751,
+      "step": 7573000
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 3.904288191228233e-05,
+      "loss": 2.1637,
+      "step": 7573500
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 3.904215826463506e-05,
+      "loss": 2.1561,
+      "step": 7574000
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 3.904143461698778e-05,
+      "loss": 2.1432,
+      "step": 7574500
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 3.90407109693405e-05,
+      "loss": 2.1915,
+      "step": 7575000
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 3.9039987321693223e-05,
+      "loss": 2.2253,
+      "step": 7575500
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 3.903926512134124e-05,
+      "loss": 2.1731,
+      "step": 7576000
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 3.903854147369396e-05,
+      "loss": 2.1778,
+      "step": 7576500
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 3.9037817826046683e-05,
+      "loss": 2.17,
+      "step": 7577000
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 3.9037094178399406e-05,
+      "loss": 2.2105,
+      "step": 7577500
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 3.903637053075213e-05,
+      "loss": 2.1666,
+      "step": 7578000
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 3.903564688310486e-05,
+      "loss": 2.1828,
+      "step": 7578500
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 3.903492323545758e-05,
+      "loss": 2.192,
+      "step": 7579000
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 3.90341995878103e-05,
+      "loss": 2.1726,
+      "step": 7579500
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 3.903347594016303e-05,
+      "loss": 2.1655,
+      "step": 7580000
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 3.9032753739811046e-05,
+      "loss": 2.1777,
+      "step": 7580500
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 3.903203009216377e-05,
+      "loss": 2.1563,
+      "step": 7581000
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 3.903130644451649e-05,
+      "loss": 2.1907,
+      "step": 7581500
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 3.9030584244164506e-05,
+      "loss": 2.1966,
+      "step": 7582000
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 3.902986059651723e-05,
+      "loss": 2.1691,
+      "step": 7582500
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 3.902913694886996e-05,
+      "loss": 2.1687,
+      "step": 7583000
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 3.902841330122268e-05,
+      "loss": 2.1792,
+      "step": 7583500
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 3.9027691100870695e-05,
+      "loss": 2.1698,
+      "step": 7584000
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 3.902696745322342e-05,
+      "loss": 2.1715,
+      "step": 7584500
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 3.902624380557614e-05,
+      "loss": 2.1834,
+      "step": 7585000
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 3.902552015792886e-05,
+      "loss": 2.1836,
+      "step": 7585500
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 3.9024797957576884e-05,
+      "loss": 2.1816,
+      "step": 7586000
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 3.9024074309929606e-05,
+      "loss": 2.1679,
+      "step": 7586500
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 3.902335210957762e-05,
+      "loss": 2.1657,
+      "step": 7587000
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 3.9022628461930344e-05,
+      "loss": 2.1966,
+      "step": 7587500
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 3.902190481428307e-05,
+      "loss": 2.1623,
+      "step": 7588000
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 3.9021181166635795e-05,
+      "loss": 2.1577,
+      "step": 7588500
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 3.902045896628381e-05,
+      "loss": 2.2023,
+      "step": 7589000
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 3.901973531863653e-05,
+      "loss": 2.1655,
+      "step": 7589500
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 3.9019011670989255e-05,
+      "loss": 2.1756,
+      "step": 7590000
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 3.901828947063728e-05,
+      "loss": 2.1583,
+      "step": 7590500
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 3.901756582299e-05,
+      "loss": 2.1931,
+      "step": 7591000
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 3.901684217534272e-05,
+      "loss": 2.1851,
+      "step": 7591500
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 3.9016118527695444e-05,
+      "loss": 2.1946,
+      "step": 7592000
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 3.901539488004817e-05,
+      "loss": 2.1678,
+      "step": 7592500
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 3.901467123240089e-05,
+      "loss": 2.1697,
+      "step": 7593000
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 3.901394758475361e-05,
+      "loss": 2.1563,
+      "step": 7593500
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 3.9013225384401633e-05,
+      "loss": 2.1698,
+      "step": 7594000
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 3.9012501736754356e-05,
+      "loss": 2.1727,
+      "step": 7594500
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 3.901177808910708e-05,
+      "loss": 2.1725,
+      "step": 7595000
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 3.901105444145981e-05,
+      "loss": 2.171,
+      "step": 7595500
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 3.901033079381253e-05,
+      "loss": 2.1822,
+      "step": 7596000
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 3.900960714616525e-05,
+      "loss": 2.1791,
+      "step": 7596500
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 3.9008883498517974e-05,
+      "loss": 2.1904,
+      "step": 7597000
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 3.9008159850870696e-05,
+      "loss": 2.1842,
+      "step": 7597500
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 3.900743620322342e-05,
+      "loss": 2.1409,
+      "step": 7598000
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 3.9006714002871434e-05,
+      "loss": 2.1635,
+      "step": 7598500
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 3.9005990355224156e-05,
+      "loss": 2.1908,
+      "step": 7599000
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 3.9005266707576885e-05,
+      "loss": 2.179,
+      "step": 7599500
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 3.900454305992961e-05,
+      "loss": 2.1825,
+      "step": 7600000
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.6618738376038161,
+      "eval_accuracy_mlm": 0.6256257501839781,
+      "eval_accuracy_nsp": 0.8561273734721095,
+      "eval_loss": 2.2176835536956787,
+      "eval_runtime": 331.6809,
+      "eval_samples_per_second": 1315.68,
+      "eval_steps_per_second": 54.821,
+      "step": 7600384
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 3.900381941228233e-05,
+      "loss": 2.1843,
+      "step": 7600500
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 3.900309576463505e-05,
+      "loss": 2.1525,
+      "step": 7601000
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 3.900237356428307e-05,
+      "loss": 2.1665,
+      "step": 7601500
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 3.900164991663579e-05,
+      "loss": 2.1588,
+      "step": 7602000
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 3.900092626898851e-05,
+      "loss": 2.1481,
+      "step": 7602500
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 3.900020262134124e-05,
+      "loss": 2.1454,
+      "step": 7603000
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 3.899947897369396e-05,
+      "loss": 2.1571,
+      "step": 7603500
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 3.8998755326046685e-05,
+      "loss": 2.1507,
+      "step": 7604000
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 3.899803312569471e-05,
+      "loss": 2.1619,
+      "step": 7604500
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 3.899730947804743e-05,
+      "loss": 2.1628,
+      "step": 7605000
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 3.899658583040015e-05,
+      "loss": 2.1625,
+      "step": 7605500
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 3.8995862182752874e-05,
+      "loss": 2.1462,
+      "step": 7606000
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 3.899513998240089e-05,
+      "loss": 2.1624,
+      "step": 7606500
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 3.899441778204891e-05,
+      "loss": 2.1682,
+      "step": 7607000
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 3.8993694134401634e-05,
+      "loss": 2.1578,
+      "step": 7607500
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 3.8992970486754356e-05,
+      "loss": 2.1436,
+      "step": 7608000
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 3.899224683910708e-05,
+      "loss": 2.1512,
+      "step": 7608500
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 3.89915231914598e-05,
+      "loss": 2.1432,
+      "step": 7609000
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 3.899079954381252e-05,
+      "loss": 2.1692,
+      "step": 7609500
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 3.8990075896165245e-05,
+      "loss": 2.1634,
+      "step": 7610000
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 3.8989352248517974e-05,
+      "loss": 2.1887,
+      "step": 7610500
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 3.898863004816599e-05,
+      "loss": 2.1697,
+      "step": 7611000
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 3.898790640051871e-05,
+      "loss": 2.1452,
+      "step": 7611500
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 3.8987182752871434e-05,
+      "loss": 2.1169,
+      "step": 7612000
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 3.8986459105224163e-05,
+      "loss": 2.1549,
+      "step": 7612500
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 3.898573690487218e-05,
+      "loss": 2.1482,
+      "step": 7613000
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 3.89850132572249e-05,
+      "loss": 2.1528,
+      "step": 7613500
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 3.8984289609577623e-05,
+      "loss": 2.1622,
+      "step": 7614000
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 3.8983565961930346e-05,
+      "loss": 2.1683,
+      "step": 7614500
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 3.898284231428307e-05,
+      "loss": 2.1819,
+      "step": 7615000
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 3.898211866663579e-05,
+      "loss": 2.1598,
+      "step": 7615500
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 3.898139501898851e-05,
+      "loss": 2.1426,
+      "step": 7616000
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 3.8980671371341235e-05,
+      "loss": 2.1521,
+      "step": 7616500
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 3.8979947723693964e-05,
+      "loss": 2.1691,
+      "step": 7617000
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 3.897922552334198e-05,
+      "loss": 2.1551,
+      "step": 7617500
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 3.897850187569471e-05,
+      "loss": 2.156,
+      "step": 7618000
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 3.8977779675342724e-05,
+      "loss": 2.1732,
+      "step": 7618500
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 3.8977056027695446e-05,
+      "loss": 2.1632,
+      "step": 7619000
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 3.897633238004817e-05,
+      "loss": 2.1695,
+      "step": 7619500
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 3.897560873240089e-05,
+      "loss": 2.1729,
+      "step": 7620000
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 3.8974887979344206e-05,
+      "loss": 2.1893,
+      "step": 7620500
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 3.897416433169693e-05,
+      "loss": 2.1657,
+      "step": 7621000
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 3.897344068404965e-05,
+      "loss": 2.1659,
+      "step": 7621500
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 3.897271703640237e-05,
+      "loss": 2.1542,
+      "step": 7622000
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 3.8971993388755095e-05,
+      "loss": 2.1733,
+      "step": 7622500
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 3.897127118840311e-05,
+      "loss": 2.1531,
+      "step": 7623000
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 3.897054754075584e-05,
+      "loss": 2.155,
+      "step": 7623500
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 3.896982389310856e-05,
+      "loss": 2.1917,
+      "step": 7624000
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 3.8969100245461284e-05,
+      "loss": 2.1622,
+      "step": 7624500
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 3.8968376597814006e-05,
+      "loss": 2.1771,
+      "step": 7625000
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 3.896765295016673e-05,
+      "loss": 2.1242,
+      "step": 7625500
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 3.896692930251946e-05,
+      "loss": 2.1531,
+      "step": 7626000
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 3.896620565487218e-05,
+      "loss": 2.1461,
+      "step": 7626500
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 3.89654820072249e-05,
+      "loss": 2.1461,
+      "step": 7627000
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 3.896475980687292e-05,
+      "loss": 2.1567,
+      "step": 7627500
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 3.896403615922564e-05,
+      "loss": 2.157,
+      "step": 7628000
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 3.896331395887366e-05,
+      "loss": 2.1644,
+      "step": 7628500
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 3.8962590311226384e-05,
+      "loss": 2.1523,
+      "step": 7629000
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 3.89618681108744e-05,
+      "loss": 2.1733,
+      "step": 7629500
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 3.896114446322712e-05,
+      "loss": 2.1623,
+      "step": 7630000
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 3.8960420815579844e-05,
+      "loss": 2.164,
+      "step": 7630500
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 3.895969716793257e-05,
+      "loss": 2.1664,
+      "step": 7631000
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 3.895897352028529e-05,
+      "loss": 2.1459,
+      "step": 7631500
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 3.895824987263801e-05,
+      "loss": 2.157,
+      "step": 7632000
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 3.895752622499074e-05,
+      "loss": 2.1557,
+      "step": 7632500
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 3.895680257734346e-05,
+      "loss": 2.1615,
+      "step": 7633000
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 3.895607892969619e-05,
+      "loss": 2.1445,
+      "step": 7633500
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 3.8955355282048914e-05,
+      "loss": 2.1793,
+      "step": 7634000
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 3.8954631634401636e-05,
+      "loss": 2.1503,
+      "step": 7634500
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 3.895390798675436e-05,
+      "loss": 2.1457,
+      "step": 7635000
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 3.895318433910708e-05,
+      "loss": 2.1628,
+      "step": 7635500
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 3.8952462138755096e-05,
+      "loss": 2.1451,
+      "step": 7636000
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 3.895173849110782e-05,
+      "loss": 2.1431,
+      "step": 7636500
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 3.895101484346054e-05,
+      "loss": 2.1634,
+      "step": 7637000
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 3.895029119581326e-05,
+      "loss": 2.1614,
+      "step": 7637500
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 3.894956754816599e-05,
+      "loss": 2.163,
+      "step": 7638000
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 3.89488467951093e-05,
+      "loss": 2.1657,
+      "step": 7638500
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 3.894812314746202e-05,
+      "loss": 2.1856,
+      "step": 7639000
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 3.8947399499814745e-05,
+      "loss": 2.1445,
+      "step": 7639500
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 3.894667585216747e-05,
+      "loss": 2.192,
+      "step": 7640000
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 3.894595220452019e-05,
+      "loss": 2.1555,
+      "step": 7640500
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 3.894522855687292e-05,
+      "loss": 2.1486,
+      "step": 7641000
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 3.894450490922564e-05,
+      "loss": 2.1745,
+      "step": 7641500
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 3.894378126157836e-05,
+      "loss": 2.1469,
+      "step": 7642000
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 3.8943059061226385e-05,
+      "loss": 2.1672,
+      "step": 7642500
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 3.89423368608744e-05,
+      "loss": 2.1663,
+      "step": 7643000
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 3.894161321322712e-05,
+      "loss": 2.1638,
+      "step": 7643500
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 3.8940889565579845e-05,
+      "loss": 2.1588,
+      "step": 7644000
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 3.894016591793257e-05,
+      "loss": 2.163,
+      "step": 7644500
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 3.893944227028529e-05,
+      "loss": 2.1767,
+      "step": 7645000
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 3.893871862263801e-05,
+      "loss": 2.171,
+      "step": 7645500
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 3.893799497499074e-05,
+      "loss": 2.1679,
+      "step": 7646000
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 3.893727132734346e-05,
+      "loss": 2.1847,
+      "step": 7646500
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 3.8936547679696185e-05,
+      "loss": 2.1575,
+      "step": 7647000
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 3.893582403204891e-05,
+      "loss": 2.1725,
+      "step": 7647500
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 3.893510038440163e-05,
+      "loss": 2.1721,
+      "step": 7648000
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 3.893437673675436e-05,
+      "loss": 2.1598,
+      "step": 7648500
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 3.893365308910708e-05,
+      "loss": 2.1628,
+      "step": 7649000
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 3.89329308887551e-05,
+      "loss": 2.1704,
+      "step": 7649500
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 3.893220724110782e-05,
+      "loss": 2.1719,
+      "step": 7650000
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 3.893148359346054e-05,
+      "loss": 2.1402,
+      "step": 7650500
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 3.893075994581326e-05,
+      "loss": 2.1626,
+      "step": 7651000
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 3.893003629816599e-05,
+      "loss": 2.1506,
+      "step": 7651500
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 3.8929312650518715e-05,
+      "loss": 2.1639,
+      "step": 7652000
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 3.892858900287144e-05,
+      "loss": 2.1733,
+      "step": 7652500
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 3.892786535522416e-05,
+      "loss": 2.1566,
+      "step": 7653000
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 3.8927143154872175e-05,
+      "loss": 2.1792,
+      "step": 7653500
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 3.89264195072249e-05,
+      "loss": 2.1474,
+      "step": 7654000
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 3.892569585957762e-05,
+      "loss": 2.1741,
+      "step": 7654500
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 3.892497221193034e-05,
+      "loss": 2.1564,
+      "step": 7655000
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 3.892424856428307e-05,
+      "loss": 2.1549,
+      "step": 7655500
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 3.8923527811226386e-05,
+      "loss": 2.1689,
+      "step": 7656000
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 3.892280416357911e-05,
+      "loss": 2.1525,
+      "step": 7656500
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 3.892208051593183e-05,
+      "loss": 2.1423,
+      "step": 7657000
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 3.892135686828455e-05,
+      "loss": 2.1789,
+      "step": 7657500
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 3.8920633220637275e-05,
+      "loss": 2.1807,
+      "step": 7658000
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 3.891990957299e-05,
+      "loss": 2.1792,
+      "step": 7658500
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 3.891918592534272e-05,
+      "loss": 2.1756,
+      "step": 7659000
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 3.891846227769544e-05,
+      "loss": 2.1662,
+      "step": 7659500
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 3.891773863004817e-05,
+      "loss": 2.1724,
+      "step": 7660000
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 3.891701498240089e-05,
+      "loss": 2.147,
+      "step": 7660500
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 3.8916291334753615e-05,
+      "loss": 2.1614,
+      "step": 7661000
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 3.891556913440163e-05,
+      "loss": 2.1831,
+      "step": 7661500
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 3.891484548675435e-05,
+      "loss": 2.1466,
+      "step": 7662000
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 3.8914121839107075e-05,
+      "loss": 2.1531,
+      "step": 7662500
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 3.8913398191459804e-05,
+      "loss": 2.1472,
+      "step": 7663000
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 3.8912674543812526e-05,
+      "loss": 2.1663,
+      "step": 7663500
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 3.891195089616525e-05,
+      "loss": 2.1634,
+      "step": 7664000
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 3.891122869581327e-05,
+      "loss": 2.1659,
+      "step": 7664500
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 3.891050504816599e-05,
+      "loss": 2.164,
+      "step": 7665000
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 3.8909781400518715e-05,
+      "loss": 2.1755,
+      "step": 7665500
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 3.890905775287144e-05,
+      "loss": 2.1684,
+      "step": 7666000
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 3.890833555251945e-05,
+      "loss": 2.1736,
+      "step": 7666500
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 3.8907611904872175e-05,
+      "loss": 2.1987,
+      "step": 7667000
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 3.89068882572249e-05,
+      "loss": 2.1778,
+      "step": 7667500
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 3.890616460957762e-05,
+      "loss": 2.1562,
+      "step": 7668000
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 3.890544096193034e-05,
+      "loss": 2.1766,
+      "step": 7668500
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 3.890471731428307e-05,
+      "loss": 2.1803,
+      "step": 7669000
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 3.890399366663579e-05,
+      "loss": 2.1537,
+      "step": 7669500
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 3.8903270018988516e-05,
+      "loss": 2.1523,
+      "step": 7670000
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 3.8902546371341245e-05,
+      "loss": 2.1493,
+      "step": 7670500
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 3.8901825618284553e-05,
+      "loss": 2.1617,
+      "step": 7671000
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 3.8901101970637276e-05,
+      "loss": 2.1887,
+      "step": 7671500
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 3.890037832299e-05,
+      "loss": 2.1946,
+      "step": 7672000
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 3.889965467534272e-05,
+      "loss": 2.1536,
+      "step": 7672500
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 3.889893102769544e-05,
+      "loss": 2.1721,
+      "step": 7673000
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 3.889820738004817e-05,
+      "loss": 2.1846,
+      "step": 7673500
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 3.889748517969619e-05,
+      "loss": 2.1547,
+      "step": 7674000
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 3.889676153204891e-05,
+      "loss": 2.1656,
+      "step": 7674500
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 3.889603788440163e-05,
+      "loss": 2.1508,
+      "step": 7675000
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 3.889531568404965e-05,
+      "loss": 2.182,
+      "step": 7675500
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 3.889459203640237e-05,
+      "loss": 2.1695,
+      "step": 7676000
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 3.889386838875509e-05,
+      "loss": 2.1854,
+      "step": 7676500
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 3.889314474110782e-05,
+      "loss": 2.1706,
+      "step": 7677000
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 3.889242109346054e-05,
+      "loss": 2.1534,
+      "step": 7677500
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 3.889169744581327e-05,
+      "loss": 2.175,
+      "step": 7678000
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 3.8890973798165994e-05,
+      "loss": 2.1651,
+      "step": 7678500
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 3.8890250150518716e-05,
+      "loss": 2.1777,
+      "step": 7679000
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 3.888952650287144e-05,
+      "loss": 2.1544,
+      "step": 7679500
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 3.888880285522416e-05,
+      "loss": 2.1813,
+      "step": 7680000
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 3.888807920757688e-05,
+      "loss": 2.1685,
+      "step": 7680500
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 3.8887355559929605e-05,
+      "loss": 2.1689,
+      "step": 7681000
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 3.888663335957762e-05,
+      "loss": 2.1667,
+      "step": 7681500
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 3.888591115922564e-05,
+      "loss": 2.157,
+      "step": 7682000
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 3.8885187511578365e-05,
+      "loss": 2.1786,
+      "step": 7682500
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 3.888446386393109e-05,
+      "loss": 2.1723,
+      "step": 7683000
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 3.888374021628381e-05,
+      "loss": 2.1667,
+      "step": 7683500
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 3.888301656863653e-05,
+      "loss": 2.1395,
+      "step": 7684000
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 3.8882292920989254e-05,
+      "loss": 2.1671,
+      "step": 7684500
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 3.8881569273341976e-05,
+      "loss": 2.1683,
+      "step": 7685000
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 3.8880845625694705e-05,
+      "loss": 2.1823,
+      "step": 7685500
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 3.888012197804743e-05,
+      "loss": 2.179,
+      "step": 7686000
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 3.887939833040015e-05,
+      "loss": 2.1606,
+      "step": 7686500
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 3.887867613004817e-05,
+      "loss": 2.1783,
+      "step": 7687000
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 3.887795392969619e-05,
+      "loss": 2.1652,
+      "step": 7687500
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 3.887723028204891e-05,
+      "loss": 2.1588,
+      "step": 7688000
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 3.887650663440163e-05,
+      "loss": 2.1628,
+      "step": 7688500
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 3.8875782986754354e-05,
+      "loss": 2.171,
+      "step": 7689000
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 3.887505933910708e-05,
+      "loss": 2.168,
+      "step": 7689500
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 3.88743356914598e-05,
+      "loss": 2.1617,
+      "step": 7690000
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 3.887361204381252e-05,
+      "loss": 2.1307,
+      "step": 7690500
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 3.887288839616525e-05,
+      "loss": 2.1626,
+      "step": 7691000
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 3.887216474851797e-05,
+      "loss": 2.149,
+      "step": 7691500
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 3.887144254816599e-05,
+      "loss": 2.1757,
+      "step": 7692000
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 3.887071890051871e-05,
+      "loss": 2.1794,
+      "step": 7692500
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 3.886999525287144e-05,
+      "loss": 2.1467,
+      "step": 7693000
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 3.886927160522416e-05,
+      "loss": 2.1599,
+      "step": 7693500
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 3.8868547957576884e-05,
+      "loss": 2.157,
+      "step": 7694000
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 3.88678257572249e-05,
+      "loss": 2.1619,
+      "step": 7694500
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 3.886710210957762e-05,
+      "loss": 2.1628,
+      "step": 7695000
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 3.886637846193035e-05,
+      "loss": 2.1698,
+      "step": 7695500
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 3.886565481428307e-05,
+      "loss": 2.161,
+      "step": 7696000
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 3.8864931166635795e-05,
+      "loss": 2.1825,
+      "step": 7696500
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 3.886420896628381e-05,
+      "loss": 2.1718,
+      "step": 7697000
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 3.886348531863653e-05,
+      "loss": 2.1692,
+      "step": 7697500
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 3.8862761670989255e-05,
+      "loss": 2.148,
+      "step": 7698000
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 3.886203802334198e-05,
+      "loss": 2.1812,
+      "step": 7698500
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 3.886131582299e-05,
+      "loss": 2.1719,
+      "step": 7699000
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 3.886059217534272e-05,
+      "loss": 2.1468,
+      "step": 7699500
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 3.885986997499074e-05,
+      "loss": 2.1779,
+      "step": 7700000
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 3.885914632734346e-05,
+      "loss": 2.1632,
+      "step": 7700500
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 3.885842267969619e-05,
+      "loss": 2.1364,
+      "step": 7701000
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 3.885769903204891e-05,
+      "loss": 2.1426,
+      "step": 7701500
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 3.885697538440163e-05,
+      "loss": 2.1761,
+      "step": 7702000
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 3.8856251736754355e-05,
+      "loss": 2.1409,
+      "step": 7702500
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 3.885552808910708e-05,
+      "loss": 2.1685,
+      "step": 7703000
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 3.88548044414598e-05,
+      "loss": 2.1555,
+      "step": 7703500
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 3.885408079381252e-05,
+      "loss": 2.1595,
+      "step": 7704000
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 3.885335714616525e-05,
+      "loss": 2.1716,
+      "step": 7704500
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 3.8852634945813267e-05,
+      "loss": 2.1818,
+      "step": 7705000
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 3.885191129816599e-05,
+      "loss": 2.1685,
+      "step": 7705500
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 3.885118765051871e-05,
+      "loss": 2.1947,
+      "step": 7706000
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 3.885046400287143e-05,
+      "loss": 2.1836,
+      "step": 7706500
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 3.8849740355224155e-05,
+      "loss": 2.1672,
+      "step": 7707000
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 3.884901670757688e-05,
+      "loss": 2.1584,
+      "step": 7707500
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 3.884829305992961e-05,
+      "loss": 2.1617,
+      "step": 7708000
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 3.884756941228233e-05,
+      "loss": 2.1605,
+      "step": 7708500
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 3.884684576463505e-05,
+      "loss": 2.1945,
+      "step": 7709000
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 3.8846123564283073e-05,
+      "loss": 2.1833,
+      "step": 7709500
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 3.8845399916635796e-05,
+      "loss": 2.1624,
+      "step": 7710000
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 3.884467626898852e-05,
+      "loss": 2.1593,
+      "step": 7710500
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 3.884395262134124e-05,
+      "loss": 2.1691,
+      "step": 7711000
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 3.884322897369396e-05,
+      "loss": 2.1955,
+      "step": 7711500
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 3.8842505326046685e-05,
+      "loss": 2.1726,
+      "step": 7712000
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 3.884178167839941e-05,
+      "loss": 2.1731,
+      "step": 7712500
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 3.884105803075213e-05,
+      "loss": 2.1569,
+      "step": 7713000
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 3.884033583040015e-05,
+      "loss": 2.1857,
+      "step": 7713500
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 3.8839612182752874e-05,
+      "loss": 2.144,
+      "step": 7714000
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 3.883888998240089e-05,
+      "loss": 2.1481,
+      "step": 7714500
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 3.883816633475361e-05,
+      "loss": 2.1781,
+      "step": 7715000
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 3.883744268710634e-05,
+      "loss": 2.1697,
+      "step": 7715500
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 3.8836720486754356e-05,
+      "loss": 2.1551,
+      "step": 7716000
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 3.883599828640238e-05,
+      "loss": 2.1846,
+      "step": 7716500
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 3.88352746387551e-05,
+      "loss": 2.1514,
+      "step": 7717000
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 3.883455099110782e-05,
+      "loss": 2.1658,
+      "step": 7717500
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 3.8833827343460545e-05,
+      "loss": 2.1655,
+      "step": 7718000
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 3.883310369581327e-05,
+      "loss": 2.1625,
+      "step": 7718500
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 3.883238004816599e-05,
+      "loss": 2.174,
+      "step": 7719000
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 3.883165640051871e-05,
+      "loss": 2.1712,
+      "step": 7719500
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 3.883093420016673e-05,
+      "loss": 2.1678,
+      "step": 7720000
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 3.883021055251945e-05,
+      "loss": 2.1753,
+      "step": 7720500
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 3.882948690487218e-05,
+      "loss": 2.18,
+      "step": 7721000
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 3.88287632572249e-05,
+      "loss": 2.1806,
+      "step": 7721500
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 3.882803960957762e-05,
+      "loss": 2.1847,
+      "step": 7722000
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 3.8827315961930345e-05,
+      "loss": 2.1838,
+      "step": 7722500
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 3.882659376157836e-05,
+      "loss": 2.1834,
+      "step": 7723000
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 3.882587156122638e-05,
+      "loss": 2.1826,
+      "step": 7723500
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 3.8825147913579105e-05,
+      "loss": 2.1659,
+      "step": 7724000
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 3.882442426593183e-05,
+      "loss": 2.1518,
+      "step": 7724500
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 3.882370061828455e-05,
+      "loss": 2.1595,
+      "step": 7725000
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 3.882297697063728e-05,
+      "loss": 2.1806,
+      "step": 7725500
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 3.882225332299e-05,
+      "loss": 2.19,
+      "step": 7726000
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 3.882152967534272e-05,
+      "loss": 2.159,
+      "step": 7726500
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 3.8820806027695446e-05,
+      "loss": 2.153,
+      "step": 7727000
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 3.882008238004817e-05,
+      "loss": 2.1653,
+      "step": 7727500
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 3.881936017969618e-05,
+      "loss": 2.1531,
+      "step": 7728000
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 3.8818636532048906e-05,
+      "loss": 2.1792,
+      "step": 7728500
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 3.881791288440163e-05,
+      "loss": 2.1612,
+      "step": 7729000
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 3.881718923675435e-05,
+      "loss": 2.1576,
+      "step": 7729500
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 3.881646558910708e-05,
+      "loss": 2.1608,
+      "step": 7730000
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 3.88157419414598e-05,
+      "loss": 2.165,
+      "step": 7730500
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 3.881501829381253e-05,
+      "loss": 2.1592,
+      "step": 7731000
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 3.8814296093460546e-05,
+      "loss": 2.1864,
+      "step": 7731500
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 3.881357244581327e-05,
+      "loss": 2.1649,
+      "step": 7732000
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 3.881284879816599e-05,
+      "loss": 2.1827,
+      "step": 7732500
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 3.881212515051871e-05,
+      "loss": 2.1716,
+      "step": 7733000
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 3.8811401502871435e-05,
+      "loss": 2.195,
+      "step": 7733500
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 3.881067930251945e-05,
+      "loss": 2.1707,
+      "step": 7734000
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 3.880995565487218e-05,
+      "loss": 2.1745,
+      "step": 7734500
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 3.88092320072249e-05,
+      "loss": 2.1814,
+      "step": 7735000
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 3.8808508359577624e-05,
+      "loss": 2.1654,
+      "step": 7735500
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 3.8807784711930346e-05,
+      "loss": 2.1683,
+      "step": 7736000
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 3.880706106428307e-05,
+      "loss": 2.1475,
+      "step": 7736500
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 3.8806338863931084e-05,
+      "loss": 2.1745,
+      "step": 7737000
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 3.8805615216283806e-05,
+      "loss": 2.1819,
+      "step": 7737500
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 3.8804891568636535e-05,
+      "loss": 2.1592,
+      "step": 7738000
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 3.880416792098926e-05,
+      "loss": 2.1607,
+      "step": 7738500
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 3.880344427334198e-05,
+      "loss": 2.1712,
+      "step": 7739000
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 3.88027206256947e-05,
+      "loss": 2.1652,
+      "step": 7739500
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 3.880199697804743e-05,
+      "loss": 2.1734,
+      "step": 7740000
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 3.880127333040015e-05,
+      "loss": 2.1768,
+      "step": 7740500
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 3.8800549682752875e-05,
+      "loss": 2.1662,
+      "step": 7741000
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 3.87998260351056e-05,
+      "loss": 2.1706,
+      "step": 7741500
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 3.879910238745832e-05,
+      "loss": 2.1749,
+      "step": 7742000
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 3.879837873981104e-05,
+      "loss": 2.1996,
+      "step": 7742500
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 3.8797655092163764e-05,
+      "loss": 2.1773,
+      "step": 7743000
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 3.8796931444516486e-05,
+      "loss": 2.1481,
+      "step": 7743500
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 3.879620779686921e-05,
+      "loss": 2.1503,
+      "step": 7744000
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 3.879548414922193e-05,
+      "loss": 2.1451,
+      "step": 7744500
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 3.879476050157466e-05,
+      "loss": 2.1784,
+      "step": 7745000
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 3.879403830122268e-05,
+      "loss": 2.1722,
+      "step": 7745500
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 3.87933161008707e-05,
+      "loss": 2.158,
+      "step": 7746000
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 3.879259245322342e-05,
+      "loss": 2.1522,
+      "step": 7746500
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 3.879186880557614e-05,
+      "loss": 2.166,
+      "step": 7747000
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 3.879114660522416e-05,
+      "loss": 2.1847,
+      "step": 7747500
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 3.879042295757688e-05,
+      "loss": 2.1777,
+      "step": 7748000
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 3.878969930992961e-05,
+      "loss": 2.18,
+      "step": 7748500
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 3.878897566228233e-05,
+      "loss": 2.1532,
+      "step": 7749000
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 3.878825346193035e-05,
+      "loss": 2.1702,
+      "step": 7749500
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 3.878752981428307e-05,
+      "loss": 2.1611,
+      "step": 7750000
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 3.878680616663579e-05,
+      "loss": 2.1769,
+      "step": 7750500
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 3.8786082518988514e-05,
+      "loss": 2.157,
+      "step": 7751000
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 3.878536031863653e-05,
+      "loss": 2.1711,
+      "step": 7751500
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 3.878463667098926e-05,
+      "loss": 2.1618,
+      "step": 7752000
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 3.878391302334198e-05,
+      "loss": 2.1628,
+      "step": 7752500
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 3.878318937569471e-05,
+      "loss": 2.1816,
+      "step": 7753000
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 3.878246572804743e-05,
+      "loss": 2.1909,
+      "step": 7753500
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 3.8781742080400154e-05,
+      "loss": 2.1519,
+      "step": 7754000
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 3.8781018432752876e-05,
+      "loss": 2.1684,
+      "step": 7754500
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 3.87802947851056e-05,
+      "loss": 2.1574,
+      "step": 7755000
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 3.877957113745832e-05,
+      "loss": 2.1792,
+      "step": 7755500
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 3.8778848937106336e-05,
+      "loss": 2.1617,
+      "step": 7756000
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 3.877812673675436e-05,
+      "loss": 2.1458,
+      "step": 7756500
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 3.877740308910708e-05,
+      "loss": 2.1746,
+      "step": 7757000
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 3.87766794414598e-05,
+      "loss": 2.1571,
+      "step": 7757500
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 3.8775955793812525e-05,
+      "loss": 2.1687,
+      "step": 7758000
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 3.877523214616525e-05,
+      "loss": 2.1448,
+      "step": 7758500
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 3.877450849851797e-05,
+      "loss": 2.1768,
+      "step": 7759000
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 3.877378485087069e-05,
+      "loss": 2.18,
+      "step": 7759500
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 3.877306265051871e-05,
+      "loss": 2.1709,
+      "step": 7760000
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 3.8772339002871436e-05,
+      "loss": 2.1592,
+      "step": 7760500
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 3.877161535522416e-05,
+      "loss": 2.1725,
+      "step": 7761000
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 3.877089170757688e-05,
+      "loss": 2.1799,
+      "step": 7761500
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 3.877016805992961e-05,
+      "loss": 2.167,
+      "step": 7762000
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 3.8769445859577625e-05,
+      "loss": 2.168,
+      "step": 7762500
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 3.876872221193035e-05,
+      "loss": 2.1792,
+      "step": 7763000
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 3.876799856428307e-05,
+      "loss": 2.1752,
+      "step": 7763500
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 3.876727491663579e-05,
+      "loss": 2.1552,
+      "step": 7764000
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 3.8766551268988514e-05,
+      "loss": 2.1467,
+      "step": 7764500
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 3.876582906863653e-05,
+      "loss": 2.1835,
+      "step": 7765000
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 3.876510542098926e-05,
+      "loss": 2.1845,
+      "step": 7765500
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 3.876438177334198e-05,
+      "loss": 2.1669,
+      "step": 7766000
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 3.8763658125694703e-05,
+      "loss": 2.1936,
+      "step": 7766500
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 3.876293592534272e-05,
+      "loss": 2.1729,
+      "step": 7767000
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 3.876221227769544e-05,
+      "loss": 2.1918,
+      "step": 7767500
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 3.876148863004817e-05,
+      "loss": 2.1522,
+      "step": 7768000
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 3.876076498240089e-05,
+      "loss": 2.175,
+      "step": 7768500
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 3.8760041334753615e-05,
+      "loss": 2.1644,
+      "step": 7769000
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 3.875931913440164e-05,
+      "loss": 2.1704,
+      "step": 7769500
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 3.875859548675436e-05,
+      "loss": 2.1685,
+      "step": 7770000
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 3.875787183910708e-05,
+      "loss": 2.1617,
+      "step": 7770500
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 3.8757148191459804e-05,
+      "loss": 2.1854,
+      "step": 7771000
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 3.8756424543812526e-05,
+      "loss": 2.1846,
+      "step": 7771500
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 3.875570089616525e-05,
+      "loss": 2.1777,
+      "step": 7772000
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 3.875497724851797e-05,
+      "loss": 2.1667,
+      "step": 7772500
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 3.8754256495461286e-05,
+      "loss": 2.1533,
+      "step": 7773000
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 3.875353284781401e-05,
+      "loss": 2.1757,
+      "step": 7773500
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 3.875280920016673e-05,
+      "loss": 2.1791,
+      "step": 7774000
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 3.875208555251945e-05,
+      "loss": 2.1627,
+      "step": 7774500
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 3.8751361904872175e-05,
+      "loss": 2.1564,
+      "step": 7775000
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 3.8750638257224904e-05,
+      "loss": 2.2073,
+      "step": 7775500
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 3.8749914609577626e-05,
+      "loss": 2.1602,
+      "step": 7776000
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 3.874919240922564e-05,
+      "loss": 2.2062,
+      "step": 7776500
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 3.8748468761578364e-05,
+      "loss": 2.1559,
+      "step": 7777000
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 3.8747745113931086e-05,
+      "loss": 2.162,
+      "step": 7777500
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 3.874702146628381e-05,
+      "loss": 2.1691,
+      "step": 7778000
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 3.874629781863654e-05,
+      "loss": 2.1844,
+      "step": 7778500
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 3.874557417098926e-05,
+      "loss": 2.1498,
+      "step": 7779000
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 3.874485052334198e-05,
+      "loss": 2.1582,
+      "step": 7779500
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 3.8744126875694704e-05,
+      "loss": 2.188,
+      "step": 7780000
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 3.8743403228047426e-05,
+      "loss": 2.176,
+      "step": 7780500
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 3.874267958040015e-05,
+      "loss": 2.1601,
+      "step": 7781000
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 3.874195593275287e-05,
+      "loss": 2.1624,
+      "step": 7781500
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 3.874123228510559e-05,
+      "loss": 2.1615,
+      "step": 7782000
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 3.8740508637458315e-05,
+      "loss": 2.1485,
+      "step": 7782500
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 3.873978643710634e-05,
+      "loss": 2.1623,
+      "step": 7783000
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 3.873906278945906e-05,
+      "loss": 2.1976,
+      "step": 7783500
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 3.873834058910708e-05,
+      "loss": 2.1814,
+      "step": 7784000
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 3.8737616941459804e-05,
+      "loss": 2.1771,
+      "step": 7784500
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 3.873689329381253e-05,
+      "loss": 2.1463,
+      "step": 7785000
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 3.873616964616525e-05,
+      "loss": 2.1693,
+      "step": 7785500
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 3.873544599851797e-05,
+      "loss": 2.189,
+      "step": 7786000
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 3.873472379816599e-05,
+      "loss": 2.1608,
+      "step": 7786500
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 3.873400015051871e-05,
+      "loss": 2.167,
+      "step": 7787000
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 3.873327650287144e-05,
+      "loss": 2.1885,
+      "step": 7787500
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 3.873255285522416e-05,
+      "loss": 2.1841,
+      "step": 7788000
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 3.873182920757688e-05,
+      "loss": 2.1374,
+      "step": 7788500
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 3.8731105559929605e-05,
+      "loss": 2.1823,
+      "step": 7789000
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 3.873038191228233e-05,
+      "loss": 2.186,
+      "step": 7789500
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 3.872965826463505e-05,
+      "loss": 2.1766,
+      "step": 7790000
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 3.872893461698778e-05,
+      "loss": 2.1684,
+      "step": 7790500
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 3.8728212416635794e-05,
+      "loss": 2.1664,
+      "step": 7791000
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 3.8727488768988516e-05,
+      "loss": 2.1762,
+      "step": 7791500
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 3.872676512134124e-05,
+      "loss": 2.1775,
+      "step": 7792000
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 3.872604292098926e-05,
+      "loss": 2.1771,
+      "step": 7792500
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 3.872531927334198e-05,
+      "loss": 2.1838,
+      "step": 7793000
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 3.8724595625694705e-05,
+      "loss": 2.1716,
+      "step": 7793500
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 3.872387342534272e-05,
+      "loss": 2.1574,
+      "step": 7794000
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 3.872314977769544e-05,
+      "loss": 2.1901,
+      "step": 7794500
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 3.8722426130048165e-05,
+      "loss": 2.1842,
+      "step": 7795000
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 3.872170248240089e-05,
+      "loss": 2.1366,
+      "step": 7795500
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 3.8720978834753616e-05,
+      "loss": 2.1581,
+      "step": 7796000
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 3.872025663440163e-05,
+      "loss": 2.1783,
+      "step": 7796500
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 3.8719532986754354e-05,
+      "loss": 2.1952,
+      "step": 7797000
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 3.871881078640237e-05,
+      "loss": 2.1839,
+      "step": 7797500
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 3.871808713875509e-05,
+      "loss": 2.1815,
+      "step": 7798000
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 3.871736349110782e-05,
+      "loss": 2.1613,
+      "step": 7798500
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 3.871663984346054e-05,
+      "loss": 2.1996,
+      "step": 7799000
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 3.8715916195813265e-05,
+      "loss": 2.1669,
+      "step": 7799500
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 3.871519254816599e-05,
+      "loss": 2.2006,
+      "step": 7800000
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 3.8714468900518717e-05,
+      "loss": 2.152,
+      "step": 7800500
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 3.871374525287144e-05,
+      "loss": 2.1787,
+      "step": 7801000
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 3.871302160522416e-05,
+      "loss": 2.1958,
+      "step": 7801500
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 3.871229795757688e-05,
+      "loss": 2.1455,
+      "step": 7802000
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 3.8711574309929605e-05,
+      "loss": 2.1723,
+      "step": 7802500
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 3.871085066228233e-05,
+      "loss": 2.1608,
+      "step": 7803000
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 3.871012701463505e-05,
+      "loss": 2.1673,
+      "step": 7803500
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 3.8709404814283066e-05,
+      "loss": 2.141,
+      "step": 7804000
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 3.870868116663579e-05,
+      "loss": 2.1554,
+      "step": 7804500
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 3.870795751898852e-05,
+      "loss": 2.1863,
+      "step": 7805000
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 3.870723387134124e-05,
+      "loss": 2.1736,
+      "step": 7805500
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 3.870651022369397e-05,
+      "loss": 2.1445,
+      "step": 7806000
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 3.8705788023341984e-05,
+      "loss": 2.1581,
+      "step": 7806500
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 3.8705064375694706e-05,
+      "loss": 2.1994,
+      "step": 7807000
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 3.870434072804743e-05,
+      "loss": 2.1474,
+      "step": 7807500
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 3.870361708040015e-05,
+      "loss": 2.1766,
+      "step": 7808000
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 3.870289343275287e-05,
+      "loss": 2.1509,
+      "step": 7808500
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 3.8702169785105595e-05,
+      "loss": 2.1494,
+      "step": 7809000
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 3.870144613745832e-05,
+      "loss": 2.181,
+      "step": 7809500
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 3.870072248981104e-05,
+      "loss": 2.1766,
+      "step": 7810000
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 3.869999884216377e-05,
+      "loss": 2.1602,
+      "step": 7810500
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 3.869927519451649e-05,
+      "loss": 2.1538,
+      "step": 7811000
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 3.869855154686921e-05,
+      "loss": 2.1746,
+      "step": 7811500
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 3.8697833688403115e-05,
+      "loss": 2.1944,
+      "step": 7812000
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 3.869711004075584e-05,
+      "loss": 2.1946,
+      "step": 7812500
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 3.869638639310856e-05,
+      "loss": 2.1922,
+      "step": 7813000
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 3.869566274546129e-05,
+      "loss": 2.1595,
+      "step": 7813500
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 3.869493909781401e-05,
+      "loss": 2.1679,
+      "step": 7814000
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 3.8694216897462026e-05,
+      "loss": 2.201,
+      "step": 7814500
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 3.869349324981475e-05,
+      "loss": 2.1789,
+      "step": 7815000
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 3.869276960216747e-05,
+      "loss": 2.1825,
+      "step": 7815500
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 3.869204595452019e-05,
+      "loss": 2.17,
+      "step": 7816000
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 3.8691322306872915e-05,
+      "loss": 2.1745,
+      "step": 7816500
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 3.8690598659225644e-05,
+      "loss": 2.1779,
+      "step": 7817000
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 3.8689875011578366e-05,
+      "loss": 2.1683,
+      "step": 7817500
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 3.868915136393109e-05,
+      "loss": 2.1643,
+      "step": 7818000
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 3.868842771628381e-05,
+      "loss": 2.1589,
+      "step": 7818500
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 3.868770406863653e-05,
+      "loss": 2.1922,
+      "step": 7819000
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 3.8686980420989255e-05,
+      "loss": 2.1725,
+      "step": 7819500
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 3.868625677334198e-05,
+      "loss": 2.1735,
+      "step": 7820000
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 3.8685533125694707e-05,
+      "loss": 2.1706,
+      "step": 7820500
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 3.868480947804743e-05,
+      "loss": 2.1985,
+      "step": 7821000
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 3.8684087277695444e-05,
+      "loss": 2.1585,
+      "step": 7821500
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 3.8683363630048167e-05,
+      "loss": 2.1734,
+      "step": 7822000
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 3.868264142969619e-05,
+      "loss": 2.179,
+      "step": 7822500
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 3.868191778204891e-05,
+      "loss": 2.184,
+      "step": 7823000
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 3.8681194134401633e-05,
+      "loss": 2.1694,
+      "step": 7823500
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 3.8680470486754356e-05,
+      "loss": 2.1732,
+      "step": 7824000
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 3.867974683910708e-05,
+      "loss": 2.1599,
+      "step": 7824500
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 3.86790231914598e-05,
+      "loss": 2.1651,
+      "step": 7825000
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 3.867829954381252e-05,
+      "loss": 2.1613,
+      "step": 7825500
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 3.867757879075584e-05,
+      "loss": 2.1681,
+      "step": 7826000
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 3.867685514310856e-05,
+      "loss": 2.1918,
+      "step": 7826500
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 3.867613149546128e-05,
+      "loss": 2.1695,
+      "step": 7827000
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 3.8675407847814005e-05,
+      "loss": 2.1918,
+      "step": 7827500
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 3.867468420016673e-05,
+      "loss": 2.1921,
+      "step": 7828000
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 3.8673960552519456e-05,
+      "loss": 2.18,
+      "step": 7828500
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 3.867323690487218e-05,
+      "loss": 2.1813,
+      "step": 7829000
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 3.8672514704520194e-05,
+      "loss": 2.1916,
+      "step": 7829500
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 3.8671791056872916e-05,
+      "loss": 2.1613,
+      "step": 7830000
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 3.8671067409225645e-05,
+      "loss": 2.1668,
+      "step": 7830500
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 3.867034376157837e-05,
+      "loss": 2.185,
+      "step": 7831000
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 3.866962011393109e-05,
+      "loss": 2.1922,
+      "step": 7831500
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 3.866889646628381e-05,
+      "loss": 2.1818,
+      "step": 7832000
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 3.8668172818636534e-05,
+      "loss": 2.1637,
+      "step": 7832500
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 3.8667449170989256e-05,
+      "loss": 2.1558,
+      "step": 7833000
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 3.866672552334198e-05,
+      "loss": 2.1697,
+      "step": 7833500
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 3.86660018756947e-05,
+      "loss": 2.155,
+      "step": 7834000
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 3.866527822804742e-05,
+      "loss": 2.2049,
+      "step": 7834500
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 3.8664554580400145e-05,
+      "loss": 2.1896,
+      "step": 7835000
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 3.8663830932752874e-05,
+      "loss": 2.1517,
+      "step": 7835500
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 3.8663107285105596e-05,
+      "loss": 2.1678,
+      "step": 7836000
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 3.866238508475362e-05,
+      "loss": 2.1557,
+      "step": 7836500
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 3.8661662884401634e-05,
+      "loss": 2.1618,
+      "step": 7837000
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 3.8660939236754356e-05,
+      "loss": 2.1806,
+      "step": 7837500
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 3.866021558910708e-05,
+      "loss": 2.1697,
+      "step": 7838000
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 3.86594919414598e-05,
+      "loss": 2.1788,
+      "step": 7838500
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 3.865876829381252e-05,
+      "loss": 2.1632,
+      "step": 7839000
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 3.8658044646165245e-05,
+      "loss": 2.1906,
+      "step": 7839500
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 3.865732099851797e-05,
+      "loss": 2.1871,
+      "step": 7840000
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 3.8656597350870697e-05,
+      "loss": 2.1827,
+      "step": 7840500
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 3.865587370322342e-05,
+      "loss": 2.1924,
+      "step": 7841000
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 3.865515005557614e-05,
+      "loss": 2.1582,
+      "step": 7841500
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 3.865442640792886e-05,
+      "loss": 2.1586,
+      "step": 7842000
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 3.865370420757688e-05,
+      "loss": 2.1745,
+      "step": 7842500
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 3.865298055992961e-05,
+      "loss": 2.1928,
+      "step": 7843000
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 3.865225691228233e-05,
+      "loss": 2.1668,
+      "step": 7843500
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 3.865153326463505e-05,
+      "loss": 2.1708,
+      "step": 7844000
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 3.8650809616987775e-05,
+      "loss": 2.155,
+      "step": 7844500
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 3.86500874166358e-05,
+      "loss": 2.1662,
+      "step": 7845000
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 3.864936376898852e-05,
+      "loss": 2.1554,
+      "step": 7845500
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 3.864864012134124e-05,
+      "loss": 2.1818,
+      "step": 7846000
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 3.8647916473693964e-05,
+      "loss": 2.1443,
+      "step": 7846500
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 3.8647192826046686e-05,
+      "loss": 2.1567,
+      "step": 7847000
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 3.864646917839941e-05,
+      "loss": 2.1653,
+      "step": 7847500
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 3.8645746978047424e-05,
+      "loss": 2.1748,
+      "step": 7848000
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 3.8645023330400146e-05,
+      "loss": 2.1683,
+      "step": 7848500
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 3.864429968275287e-05,
+      "loss": 2.1805,
+      "step": 7849000
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 3.86435760351056e-05,
+      "loss": 2.1885,
+      "step": 7849500
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 3.864285238745832e-05,
+      "loss": 2.1443,
+      "step": 7850000
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 3.864212873981105e-05,
+      "loss": 2.171,
+      "step": 7850500
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 3.864140509216377e-05,
+      "loss": 2.1731,
+      "step": 7851000
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 3.864068144451649e-05,
+      "loss": 2.1554,
+      "step": 7851500
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 3.8639957796869215e-05,
+      "loss": 2.16,
+      "step": 7852000
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 3.863923559651723e-05,
+      "loss": 2.1844,
+      "step": 7852500
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 3.863851194886995e-05,
+      "loss": 2.1833,
+      "step": 7853000
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 3.8637788301222675e-05,
+      "loss": 2.1734,
+      "step": 7853500
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 3.86370646535754e-05,
+      "loss": 2.1769,
+      "step": 7854000
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 3.863634390051871e-05,
+      "loss": 2.1706,
+      "step": 7854500
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 3.8635620252871435e-05,
+      "loss": 2.1889,
+      "step": 7855000
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 3.863489660522416e-05,
+      "loss": 2.1743,
+      "step": 7855500
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 3.863417295757688e-05,
+      "loss": 2.1509,
+      "step": 7856000
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 3.86334493099296e-05,
+      "loss": 2.1788,
+      "step": 7856500
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 3.8632725662282324e-05,
+      "loss": 2.1895,
+      "step": 7857000
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 3.8632002014635046e-05,
+      "loss": 2.1767,
+      "step": 7857500
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 3.8631278366987775e-05,
+      "loss": 2.1783,
+      "step": 7858000
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 3.86305561666358e-05,
+      "loss": 2.167,
+      "step": 7858500
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 3.862983251898852e-05,
+      "loss": 2.1495,
+      "step": 7859000
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 3.862910887134124e-05,
+      "loss": 2.1717,
+      "step": 7859500
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 3.862838667098926e-05,
+      "loss": 2.177,
+      "step": 7860000
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 3.862766302334198e-05,
+      "loss": 2.1671,
+      "step": 7860500
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 3.86269393756947e-05,
+      "loss": 2.1805,
+      "step": 7861000
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 3.8626215728047424e-05,
+      "loss": 2.171,
+      "step": 7861500
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 3.862549208040015e-05,
+      "loss": 2.1875,
+      "step": 7862000
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 3.862476988004817e-05,
+      "loss": 2.1924,
+      "step": 7862500
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 3.862404623240089e-05,
+      "loss": 2.1711,
+      "step": 7863000
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 3.8623322584753613e-05,
+      "loss": 2.172,
+      "step": 7863500
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 3.8622598937106336e-05,
+      "loss": 2.1415,
+      "step": 7864000
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 3.862187528945906e-05,
+      "loss": 2.1526,
+      "step": 7864500
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 3.862115164181178e-05,
+      "loss": 2.1722,
+      "step": 7865000
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 3.862042799416451e-05,
+      "loss": 2.174,
+      "step": 7865500
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 3.861970434651723e-05,
+      "loss": 2.1651,
+      "step": 7866000
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 3.8618980698869954e-05,
+      "loss": 2.1486,
+      "step": 7866500
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 3.8618257051222676e-05,
+      "loss": 2.1296,
+      "step": 7867000
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 3.86175334035754e-05,
+      "loss": 2.1898,
+      "step": 7867500
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 3.861680975592813e-05,
+      "loss": 2.1796,
+      "step": 7868000
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 3.861608755557614e-05,
+      "loss": 2.1765,
+      "step": 7868500
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 3.8615363907928865e-05,
+      "loss": 2.1769,
+      "step": 7869000
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 3.861464026028159e-05,
+      "loss": 2.1778,
+      "step": 7869500
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 3.861391661263431e-05,
+      "loss": 2.1837,
+      "step": 7870000
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 3.861319296498703e-05,
+      "loss": 2.1686,
+      "step": 7870500
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 3.8612469317339754e-05,
+      "loss": 2.1737,
+      "step": 7871000
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 3.8611745669692476e-05,
+      "loss": 2.1786,
+      "step": 7871500
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 3.86110220220452e-05,
+      "loss": 2.1693,
+      "step": 7872000
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 3.861029982169322e-05,
+      "loss": 2.1827,
+      "step": 7872500
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 3.860957617404595e-05,
+      "loss": 2.165,
+      "step": 7873000
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 3.860885252639867e-05,
+      "loss": 2.1688,
+      "step": 7873500
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 3.860813032604669e-05,
+      "loss": 2.1755,
+      "step": 7874000
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 3.860740667839941e-05,
+      "loss": 2.1641,
+      "step": 7874500
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 3.860668303075213e-05,
+      "loss": 2.1942,
+      "step": 7875000
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 3.8605959383104854e-05,
+      "loss": 2.1957,
+      "step": 7875500
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 3.8605235735457576e-05,
+      "loss": 2.1693,
+      "step": 7876000
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 3.86045135351056e-05,
+      "loss": 2.1746,
+      "step": 7876500
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 3.8603791334753614e-05,
+      "loss": 2.147,
+      "step": 7877000
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 3.8603067687106336e-05,
+      "loss": 2.1774,
+      "step": 7877500
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 3.860234403945906e-05,
+      "loss": 2.175,
+      "step": 7878000
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 3.860162039181178e-05,
+      "loss": 2.1627,
+      "step": 7878500
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 3.86008967441645e-05,
+      "loss": 2.1579,
+      "step": 7879000
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 3.8600174543812526e-05,
+      "loss": 2.1922,
+      "step": 7879500
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 3.859945089616525e-05,
+      "loss": 2.1699,
+      "step": 7880000
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 3.859872724851798e-05,
+      "loss": 2.1505,
+      "step": 7880500
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 3.85980036008707e-05,
+      "loss": 2.1709,
+      "step": 7881000
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 3.859727995322342e-05,
+      "loss": 2.1527,
+      "step": 7881500
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 3.8596556305576143e-05,
+      "loss": 2.1695,
+      "step": 7882000
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 3.8595832657928866e-05,
+      "loss": 2.1839,
+      "step": 7882500
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 3.859510901028159e-05,
+      "loss": 2.1994,
+      "step": 7883000
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 3.859438536263431e-05,
+      "loss": 2.1735,
+      "step": 7883500
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 3.8593664609577626e-05,
+      "loss": 2.1946,
+      "step": 7884000
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 3.859294096193035e-05,
+      "loss": 2.1678,
+      "step": 7884500
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 3.859221731428307e-05,
+      "loss": 2.1744,
+      "step": 7885000
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 3.8591495113931086e-05,
+      "loss": 2.1577,
+      "step": 7885500
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 3.859077146628381e-05,
+      "loss": 2.1584,
+      "step": 7886000
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 3.859004781863653e-05,
+      "loss": 2.1772,
+      "step": 7886500
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 3.858932417098925e-05,
+      "loss": 2.1732,
+      "step": 7887000
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 3.8588600523341975e-05,
+      "loss": 2.1525,
+      "step": 7887500
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 3.8587876875694704e-05,
+      "loss": 2.1585,
+      "step": 7888000
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 3.8587153228047426e-05,
+      "loss": 2.1552,
+      "step": 7888500
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 3.8586429580400155e-05,
+      "loss": 2.1765,
+      "step": 7889000
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 3.858570593275288e-05,
+      "loss": 2.1544,
+      "step": 7889500
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 3.85849822851056e-05,
+      "loss": 2.1748,
+      "step": 7890000
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 3.858425863745832e-05,
+      "loss": 2.1783,
+      "step": 7890500
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 3.8583534989811044e-05,
+      "loss": 2.1664,
+      "step": 7891000
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 3.858281278945906e-05,
+      "loss": 2.1422,
+      "step": 7891500
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 3.858208914181178e-05,
+      "loss": 2.1638,
+      "step": 7892000
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 3.8581366941459804e-05,
+      "loss": 2.1768,
+      "step": 7892500
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 3.8580643293812526e-05,
+      "loss": 2.1856,
+      "step": 7893000
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 3.857991964616525e-05,
+      "loss": 2.1817,
+      "step": 7893500
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 3.857919599851797e-05,
+      "loss": 2.188,
+      "step": 7894000
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 3.857847235087069e-05,
+      "loss": 2.1709,
+      "step": 7894500
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 3.8577748703223415e-05,
+      "loss": 2.1758,
+      "step": 7895000
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 3.8577025055576144e-05,
+      "loss": 2.1729,
+      "step": 7895500
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 3.8576301407928866e-05,
+      "loss": 2.1412,
+      "step": 7896000
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 3.857557776028159e-05,
+      "loss": 2.1444,
+      "step": 7896500
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 3.857485411263431e-05,
+      "loss": 2.1683,
+      "step": 7897000
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 3.8574131912282327e-05,
+      "loss": 2.1503,
+      "step": 7897500
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 3.8573408264635056e-05,
+      "loss": 2.1566,
+      "step": 7898000
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 3.857268461698778e-05,
+      "loss": 2.2027,
+      "step": 7898500
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 3.85719609693405e-05,
+      "loss": 2.1658,
+      "step": 7899000
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 3.8571238768988516e-05,
+      "loss": 2.1976,
+      "step": 7899500
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 3.857051512134124e-05,
+      "loss": 2.1542,
+      "step": 7900000
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 3.856979147369396e-05,
+      "loss": 2.1515,
+      "step": 7900500
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 3.856906782604668e-05,
+      "loss": 2.1669,
+      "step": 7901000
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 3.8568345625694705e-05,
+      "loss": 2.196,
+      "step": 7901500
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 3.856762197804743e-05,
+      "loss": 2.1748,
+      "step": 7902000
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 3.856689833040015e-05,
+      "loss": 2.1874,
+      "step": 7902500
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 3.856617613004817e-05,
+      "loss": 2.1558,
+      "step": 7903000
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 3.8565452482400894e-05,
+      "loss": 2.166,
+      "step": 7903500
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 3.8564728834753616e-05,
+      "loss": 2.1717,
+      "step": 7904000
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 3.856400518710634e-05,
+      "loss": 2.1853,
+      "step": 7904500
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 3.856328153945906e-05,
+      "loss": 2.1431,
+      "step": 7905000
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 3.856255789181178e-05,
+      "loss": 2.1779,
+      "step": 7905500
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 3.8561834244164505e-05,
+      "loss": 2.1609,
+      "step": 7906000
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 3.856111059651723e-05,
+      "loss": 2.1711,
+      "step": 7906500
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 3.856038839616525e-05,
+      "loss": 2.1644,
+      "step": 7907000
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 3.855966474851797e-05,
+      "loss": 2.1475,
+      "step": 7907500
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 3.855894254816599e-05,
+      "loss": 2.188,
+      "step": 7908000
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 3.855821890051871e-05,
+      "loss": 2.1722,
+      "step": 7908500
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 3.855749525287143e-05,
+      "loss": 2.1819,
+      "step": 7909000
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 3.8556771605224154e-05,
+      "loss": 2.1751,
+      "step": 7909500
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 3.855604795757688e-05,
+      "loss": 2.1745,
+      "step": 7910000
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 3.8555325757224905e-05,
+      "loss": 2.1765,
+      "step": 7910500
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 3.855460210957763e-05,
+      "loss": 2.173,
+      "step": 7911000
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 3.855387846193035e-05,
+      "loss": 2.1691,
+      "step": 7911500
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 3.855315481428307e-05,
+      "loss": 2.1403,
+      "step": 7912000
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 3.8552431166635794e-05,
+      "loss": 2.1962,
+      "step": 7912500
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 3.8551707518988516e-05,
+      "loss": 2.1799,
+      "step": 7913000
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 3.855098387134124e-05,
+      "loss": 2.1699,
+      "step": 7913500
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 3.855026022369396e-05,
+      "loss": 2.1623,
+      "step": 7914000
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 3.854953657604668e-05,
+      "loss": 2.1735,
+      "step": 7914500
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 3.854881582299e-05,
+      "loss": 2.1808,
+      "step": 7915000
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 3.854809217534272e-05,
+      "loss": 2.1754,
+      "step": 7915500
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 3.854736852769544e-05,
+      "loss": 2.1588,
+      "step": 7916000
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 3.854664632734346e-05,
+      "loss": 2.1595,
+      "step": 7916500
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 3.854592267969618e-05,
+      "loss": 2.156,
+      "step": 7917000
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 3.85451990320489e-05,
+      "loss": 2.1631,
+      "step": 7917500
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 3.854447538440163e-05,
+      "loss": 2.1753,
+      "step": 7918000
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 3.8543751736754354e-05,
+      "loss": 2.2077,
+      "step": 7918500
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 3.8543028089107083e-05,
+      "loss": 2.1626,
+      "step": 7919000
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 3.8542304441459806e-05,
+      "loss": 2.1937,
+      "step": 7919500
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 3.854158224110782e-05,
+      "loss": 2.159,
+      "step": 7920000
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 3.8540858593460543e-05,
+      "loss": 2.1952,
+      "step": 7920500
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 3.8540134945813266e-05,
+      "loss": 2.1522,
+      "step": 7921000
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 3.853941129816599e-05,
+      "loss": 2.1482,
+      "step": 7921500
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 3.853868765051871e-05,
+      "loss": 2.2005,
+      "step": 7922000
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 3.853796400287143e-05,
+      "loss": 2.1829,
+      "step": 7922500
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 3.8537240355224155e-05,
+      "loss": 2.1448,
+      "step": 7923000
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 3.8536516707576884e-05,
+      "loss": 2.176,
+      "step": 7923500
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 3.8535793059929606e-05,
+      "loss": 2.1648,
+      "step": 7924000
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 3.853506941228233e-05,
+      "loss": 2.1883,
+      "step": 7924500
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 3.853434576463505e-05,
+      "loss": 2.1918,
+      "step": 7925000
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 3.853362211698778e-05,
+      "loss": 2.1574,
+      "step": 7925500
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 3.85328984693405e-05,
+      "loss": 2.1816,
+      "step": 7926000
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 3.8532174821693224e-05,
+      "loss": 2.1578,
+      "step": 7926500
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 3.8531451174045946e-05,
+      "loss": 2.1797,
+      "step": 7927000
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 3.853072752639867e-05,
+      "loss": 2.1987,
+      "step": 7927500
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 3.853000387875139e-05,
+      "loss": 2.1994,
+      "step": 7928000
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 3.8529281678399406e-05,
+      "loss": 2.1749,
+      "step": 7928500
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 3.852855947804743e-05,
+      "loss": 2.1965,
+      "step": 7929000
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 3.852783583040015e-05,
+      "loss": 2.1691,
+      "step": 7929500
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 3.852711218275287e-05,
+      "loss": 2.1685,
+      "step": 7930000
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 3.8526388535105595e-05,
+      "loss": 2.1688,
+      "step": 7930500
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 3.852566488745832e-05,
+      "loss": 2.176,
+      "step": 7931000
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 3.852494123981104e-05,
+      "loss": 2.1398,
+      "step": 7931500
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 3.852421759216376e-05,
+      "loss": 2.1552,
+      "step": 7932000
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 3.8523495391811784e-05,
+      "loss": 2.1962,
+      "step": 7932500
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 3.8522771744164506e-05,
+      "loss": 2.1589,
+      "step": 7933000
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 3.8522048096517235e-05,
+      "loss": 2.128,
+      "step": 7933500
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 3.852132444886996e-05,
+      "loss": 2.1852,
+      "step": 7934000
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 3.852060224851797e-05,
+      "loss": 2.178,
+      "step": 7934500
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 3.8519878600870695e-05,
+      "loss": 2.178,
+      "step": 7935000
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 3.851915495322342e-05,
+      "loss": 2.1823,
+      "step": 7935500
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 3.851843275287143e-05,
+      "loss": 2.1736,
+      "step": 7936000
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 3.851770910522416e-05,
+      "loss": 2.163,
+      "step": 7936500
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 3.8516985457576884e-05,
+      "loss": 2.1661,
+      "step": 7937000
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 3.85162632572249e-05,
+      "loss": 2.1854,
+      "step": 7937500
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 3.851553960957762e-05,
+      "loss": 2.167,
+      "step": 7938000
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 3.8514815961930344e-05,
+      "loss": 2.156,
+      "step": 7938500
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 3.851409231428307e-05,
+      "loss": 2.203,
+      "step": 7939000
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 3.851336866663579e-05,
+      "loss": 2.1684,
+      "step": 7939500
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 3.851264501898851e-05,
+      "loss": 2.1749,
+      "step": 7940000
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 3.851192137134124e-05,
+      "loss": 2.1663,
+      "step": 7940500
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 3.851119772369396e-05,
+      "loss": 2.1829,
+      "step": 7941000
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 3.8510474076046685e-05,
+      "loss": 2.1789,
+      "step": 7941500
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 3.8509750428399414e-05,
+      "loss": 2.1861,
+      "step": 7942000
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 3.8509026780752136e-05,
+      "loss": 2.1714,
+      "step": 7942500
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 3.850830313310486e-05,
+      "loss": 2.1863,
+      "step": 7943000
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 3.850757948545758e-05,
+      "loss": 2.1632,
+      "step": 7943500
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 3.85068558378103e-05,
+      "loss": 2.1801,
+      "step": 7944000
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 3.8506132190163025e-05,
+      "loss": 2.2011,
+      "step": 7944500
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 3.850540854251575e-05,
+      "loss": 2.1393,
+      "step": 7945000
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 3.850468634216376e-05,
+      "loss": 2.1625,
+      "step": 7945500
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.6629433126284148,
+      "eval_accuracy_mlm": 0.6268137506413546,
+      "eval_accuracy_nsp": 0.8565788086693891,
+      "eval_loss": 2.2072818279266357,
+      "eval_runtime": 331.6784,
+      "eval_samples_per_second": 1315.69,
+      "eval_steps_per_second": 54.821,
+      "step": 7945856
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 3.8503962694516485e-05,
+      "loss": 2.1442,
+      "step": 7946000
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 3.850324049416451e-05,
+      "loss": 2.1303,
+      "step": 7946500
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 3.850251684651723e-05,
+      "loss": 2.1414,
+      "step": 7947000
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 3.8501794646165245e-05,
+      "loss": 2.1315,
+      "step": 7947500
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 3.8501070998517974e-05,
+      "loss": 2.127,
+      "step": 7948000
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 3.8500347350870696e-05,
+      "loss": 2.1598,
+      "step": 7948500
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 3.849962370322342e-05,
+      "loss": 2.1488,
+      "step": 7949000
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 3.849890005557614e-05,
+      "loss": 2.1516,
+      "step": 7949500
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 3.849817640792886e-05,
+      "loss": 2.1613,
+      "step": 7950000
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 3.8497454207576885e-05,
+      "loss": 2.1363,
+      "step": 7950500
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 3.849673055992961e-05,
+      "loss": 2.1572,
+      "step": 7951000
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 3.849600691228233e-05,
+      "loss": 2.181,
+      "step": 7951500
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 3.849528326463505e-05,
+      "loss": 2.1382,
+      "step": 7952000
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 3.8494559616987774e-05,
+      "loss": 2.1549,
+      "step": 7952500
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 3.8493835969340496e-05,
+      "loss": 2.1579,
+      "step": 7953000
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 3.849311232169322e-05,
+      "loss": 2.1404,
+      "step": 7953500
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 3.849238867404594e-05,
+      "loss": 2.1772,
+      "step": 7954000
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 3.849166647369396e-05,
+      "loss": 2.1351,
+      "step": 7954500
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 3.8490942826046685e-05,
+      "loss": 2.1548,
+      "step": 7955000
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 3.849022062569471e-05,
+      "loss": 2.1422,
+      "step": 7955500
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 3.848949697804743e-05,
+      "loss": 2.1416,
+      "step": 7956000
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 3.848877333040015e-05,
+      "loss": 2.1436,
+      "step": 7956500
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 3.8488049682752874e-05,
+      "loss": 2.1563,
+      "step": 7957000
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 3.84873260351056e-05,
+      "loss": 2.1466,
+      "step": 7957500
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 3.848660238745832e-05,
+      "loss": 2.168,
+      "step": 7958000
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 3.848587873981104e-05,
+      "loss": 2.1611,
+      "step": 7958500
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 3.848515509216376e-05,
+      "loss": 2.1472,
+      "step": 7959000
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 3.8484431444516486e-05,
+      "loss": 2.1322,
+      "step": 7959500
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 3.848370924416451e-05,
+      "loss": 2.154,
+      "step": 7960000
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 3.848298559651723e-05,
+      "loss": 2.1513,
+      "step": 7960500
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 3.8482263396165246e-05,
+      "loss": 2.1473,
+      "step": 7961000
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 3.848153974851797e-05,
+      "loss": 2.1557,
+      "step": 7961500
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 3.848081610087069e-05,
+      "loss": 2.1321,
+      "step": 7962000
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 3.848009245322341e-05,
+      "loss": 2.1495,
+      "step": 7962500
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 3.847936880557614e-05,
+      "loss": 2.1675,
+      "step": 7963000
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 3.8478645157928864e-05,
+      "loss": 2.1607,
+      "step": 7963500
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 3.847792151028159e-05,
+      "loss": 2.1607,
+      "step": 7964000
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 3.8477197862634315e-05,
+      "loss": 2.1329,
+      "step": 7964500
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 3.847647421498704e-05,
+      "loss": 2.1613,
+      "step": 7965000
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 3.847575056733976e-05,
+      "loss": 2.1416,
+      "step": 7965500
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 3.847502691969248e-05,
+      "loss": 2.1727,
+      "step": 7966000
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 3.8474303272045204e-05,
+      "loss": 2.1365,
+      "step": 7966500
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 3.847358107169322e-05,
+      "loss": 2.1673,
+      "step": 7967000
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 3.847285742404594e-05,
+      "loss": 2.1507,
+      "step": 7967500
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 3.8472135223693964e-05,
+      "loss": 2.1445,
+      "step": 7968000
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 3.8471411576046686e-05,
+      "loss": 2.1591,
+      "step": 7968500
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 3.847068792839941e-05,
+      "loss": 2.1565,
+      "step": 7969000
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 3.846996428075213e-05,
+      "loss": 2.1592,
+      "step": 7969500
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 3.8469242080400146e-05,
+      "loss": 2.1612,
+      "step": 7970000
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 3.846851988004817e-05,
+      "loss": 2.1491,
+      "step": 7970500
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 3.846779623240089e-05,
+      "loss": 2.1527,
+      "step": 7971000
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 3.846707258475361e-05,
+      "loss": 2.1514,
+      "step": 7971500
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 3.846634893710634e-05,
+      "loss": 2.1395,
+      "step": 7972000
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 3.8465625289459064e-05,
+      "loss": 2.166,
+      "step": 7972500
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 3.8464901641811787e-05,
+      "loss": 2.1407,
+      "step": 7973000
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 3.846417799416451e-05,
+      "loss": 2.153,
+      "step": 7973500
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 3.8463455793812524e-05,
+      "loss": 2.1546,
+      "step": 7974000
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 3.8462732146165247e-05,
+      "loss": 2.1605,
+      "step": 7974500
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 3.846200849851797e-05,
+      "loss": 2.1606,
+      "step": 7975000
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 3.846128485087069e-05,
+      "loss": 2.1628,
+      "step": 7975500
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 3.846056120322341e-05,
+      "loss": 2.1589,
+      "step": 7976000
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 3.845983755557614e-05,
+      "loss": 2.1561,
+      "step": 7976500
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 3.8459113907928864e-05,
+      "loss": 2.1478,
+      "step": 7977000
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 3.845839026028159e-05,
+      "loss": 2.1129,
+      "step": 7977500
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 3.8457666612634316e-05,
+      "loss": 2.1418,
+      "step": 7978000
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 3.845694296498704e-05,
+      "loss": 2.1404,
+      "step": 7978500
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 3.845621931733976e-05,
+      "loss": 2.1566,
+      "step": 7979000
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 3.845549566969248e-05,
+      "loss": 2.164,
+      "step": 7979500
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 3.8454772022045205e-05,
+      "loss": 2.1501,
+      "step": 7980000
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 3.845404982169322e-05,
+      "loss": 2.1566,
+      "step": 7980500
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 3.845332617404594e-05,
+      "loss": 2.1451,
+      "step": 7981000
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 3.8452602526398665e-05,
+      "loss": 2.1506,
+      "step": 7981500
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 3.8451878878751394e-05,
+      "loss": 2.1511,
+      "step": 7982000
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 3.8451155231104116e-05,
+      "loss": 2.1718,
+      "step": 7982500
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 3.845043158345684e-05,
+      "loss": 2.1625,
+      "step": 7983000
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 3.844970793580956e-05,
+      "loss": 2.1542,
+      "step": 7983500
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 3.8448985735457576e-05,
+      "loss": 2.1534,
+      "step": 7984000
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 3.84482620878103e-05,
+      "loss": 2.1445,
+      "step": 7984500
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 3.844753844016303e-05,
+      "loss": 2.1608,
+      "step": 7985000
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 3.844681623981104e-05,
+      "loss": 2.1556,
+      "step": 7985500
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 3.8446094039459065e-05,
+      "loss": 2.1433,
+      "step": 7986000
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 3.844537039181179e-05,
+      "loss": 2.1676,
+      "step": 7986500
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 3.844464674416451e-05,
+      "loss": 2.1478,
+      "step": 7987000
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 3.844392309651723e-05,
+      "loss": 2.1672,
+      "step": 7987500
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 3.8443199448869954e-05,
+      "loss": 2.1471,
+      "step": 7988000
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 3.8442475801222676e-05,
+      "loss": 2.1514,
+      "step": 7988500
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 3.84417521535754e-05,
+      "loss": 2.1411,
+      "step": 7989000
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 3.844102995322342e-05,
+      "loss": 2.1434,
+      "step": 7989500
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 3.844030630557614e-05,
+      "loss": 2.1495,
+      "step": 7990000
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 3.8439582657928865e-05,
+      "loss": 2.1476,
+      "step": 7990500
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 3.843885901028159e-05,
+      "loss": 2.1709,
+      "step": 7991000
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 3.84381368099296e-05,
+      "loss": 2.1532,
+      "step": 7991500
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 3.8437413162282325e-05,
+      "loss": 2.1552,
+      "step": 7992000
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 3.843668951463505e-05,
+      "loss": 2.1778,
+      "step": 7992500
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 3.8435965866987777e-05,
+      "loss": 2.1377,
+      "step": 7993000
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 3.84352422193405e-05,
+      "loss": 2.1502,
+      "step": 7993500
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 3.843451857169322e-05,
+      "loss": 2.1729,
+      "step": 7994000
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 3.843379637134124e-05,
+      "loss": 2.1515,
+      "step": 7994500
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 3.843307417098926e-05,
+      "loss": 2.1511,
+      "step": 7995000
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 3.843235052334198e-05,
+      "loss": 2.1653,
+      "step": 7995500
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 3.84316268756947e-05,
+      "loss": 2.1508,
+      "step": 7996000
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 3.8430903228047426e-05,
+      "loss": 2.1203,
+      "step": 7996500
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 3.843017958040015e-05,
+      "loss": 2.155,
+      "step": 7997000
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 3.842945593275287e-05,
+      "loss": 2.1523,
+      "step": 7997500
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 3.842873228510559e-05,
+      "loss": 2.1477,
+      "step": 7998000
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 3.842800863745832e-05,
+      "loss": 2.1658,
+      "step": 7998500
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 3.8427284989811044e-05,
+      "loss": 2.1433,
+      "step": 7999000
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 3.8426561342163766e-05,
+      "loss": 2.1297,
+      "step": 7999500
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 3.8425837694516495e-05,
+      "loss": 2.1559,
+      "step": 8000000
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 3.842511404686922e-05,
+      "loss": 2.1479,
+      "step": 8000500
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 3.842439039922194e-05,
+      "loss": 2.1705,
+      "step": 8001000
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 3.842366675157466e-05,
+      "loss": 2.163,
+      "step": 8001500
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 3.8422943103927384e-05,
+      "loss": 2.1501,
+      "step": 8002000
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 3.8422219456280106e-05,
+      "loss": 2.1433,
+      "step": 8002500
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 3.842149580863283e-05,
+      "loss": 2.1643,
+      "step": 8003000
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 3.842077216098555e-05,
+      "loss": 2.1548,
+      "step": 8003500
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 3.842004851333827e-05,
+      "loss": 2.1574,
+      "step": 8004000
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 3.8419324865690995e-05,
+      "loss": 2.1732,
+      "step": 8004500
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 3.841860121804372e-05,
+      "loss": 2.159,
+      "step": 8005000
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 3.841787901769174e-05,
+      "loss": 2.1464,
+      "step": 8005500
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 3.841715537004446e-05,
+      "loss": 2.1482,
+      "step": 8006000
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 3.8416431722397184e-05,
+      "loss": 2.1453,
+      "step": 8006500
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 3.8415708074749906e-05,
+      "loss": 2.1489,
+      "step": 8007000
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 3.8414984427102635e-05,
+      "loss": 2.1436,
+      "step": 8007500
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 3.841426222675065e-05,
+      "loss": 2.1502,
+      "step": 8008000
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 3.841353857910337e-05,
+      "loss": 2.1435,
+      "step": 8008500
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 3.8412814931456095e-05,
+      "loss": 2.1633,
+      "step": 8009000
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 3.8412091283808824e-05,
+      "loss": 2.1467,
+      "step": 8009500
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 3.8411367636161546e-05,
+      "loss": 2.1665,
+      "step": 8010000
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 3.841064543580956e-05,
+      "loss": 2.1544,
+      "step": 8010500
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 3.840992323545758e-05,
+      "loss": 2.1397,
+      "step": 8011000
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 3.84091995878103e-05,
+      "loss": 2.1601,
+      "step": 8011500
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 3.840847594016302e-05,
+      "loss": 2.1606,
+      "step": 8012000
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 3.8407752292515744e-05,
+      "loss": 2.188,
+      "step": 8012500
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 3.840702864486847e-05,
+      "loss": 2.1765,
+      "step": 8013000
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 3.8406304997221195e-05,
+      "loss": 2.1394,
+      "step": 8013500
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 3.8405584244164504e-05,
+      "loss": 2.1532,
+      "step": 8014000
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 3.8404860596517227e-05,
+      "loss": 2.1509,
+      "step": 8014500
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 3.840413694886995e-05,
+      "loss": 2.1659,
+      "step": 8015000
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 3.840341330122268e-05,
+      "loss": 2.1431,
+      "step": 8015500
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 3.84026896535754e-05,
+      "loss": 2.1528,
+      "step": 8016000
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 3.840196600592812e-05,
+      "loss": 2.1248,
+      "step": 8016500
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 3.8401242358280845e-05,
+      "loss": 2.1558,
+      "step": 8017000
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 3.8400518710633574e-05,
+      "loss": 2.1714,
+      "step": 8017500
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 3.839979651028159e-05,
+      "loss": 2.1743,
+      "step": 8018000
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 3.83990757572249e-05,
+      "loss": 2.1425,
+      "step": 8018500
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 3.839835210957762e-05,
+      "loss": 2.1629,
+      "step": 8019000
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 3.839762846193035e-05,
+      "loss": 2.1936,
+      "step": 8019500
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 3.8396906261578365e-05,
+      "loss": 2.1583,
+      "step": 8020000
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 3.839618261393109e-05,
+      "loss": 2.1329,
+      "step": 8020500
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 3.839545896628381e-05,
+      "loss": 2.1771,
+      "step": 8021000
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 3.839473531863653e-05,
+      "loss": 2.133,
+      "step": 8021500
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 3.8394011670989254e-05,
+      "loss": 2.1578,
+      "step": 8022000
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 3.8393288023341976e-05,
+      "loss": 2.177,
+      "step": 8022500
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 3.8392564375694705e-05,
+      "loss": 2.1474,
+      "step": 8023000
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 3.839184072804743e-05,
+      "loss": 2.1756,
+      "step": 8023500
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 3.839111708040015e-05,
+      "loss": 2.1576,
+      "step": 8024000
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 3.839039488004817e-05,
+      "loss": 2.1588,
+      "step": 8024500
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 3.8389671232400894e-05,
+      "loss": 2.1762,
+      "step": 8025000
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 3.8388947584753616e-05,
+      "loss": 2.1525,
+      "step": 8025500
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 3.838822538440163e-05,
+      "loss": 2.1526,
+      "step": 8026000
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 3.838750318404965e-05,
+      "loss": 2.1667,
+      "step": 8026500
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 3.8386779536402376e-05,
+      "loss": 2.1483,
+      "step": 8027000
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 3.83860558887551e-05,
+      "loss": 2.1566,
+      "step": 8027500
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 3.838533224110782e-05,
+      "loss": 2.1473,
+      "step": 8028000
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 3.838460859346054e-05,
+      "loss": 2.1534,
+      "step": 8028500
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 3.8383884945813265e-05,
+      "loss": 2.1436,
+      "step": 8029000
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 3.838316129816599e-05,
+      "loss": 2.1343,
+      "step": 8029500
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 3.838243765051871e-05,
+      "loss": 2.1352,
+      "step": 8030000
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 3.838171400287144e-05,
+      "loss": 2.1471,
+      "step": 8030500
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 3.838099035522416e-05,
+      "loss": 2.1695,
+      "step": 8031000
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 3.838026670757688e-05,
+      "loss": 2.1491,
+      "step": 8031500
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 3.8379543059929605e-05,
+      "loss": 2.1922,
+      "step": 8032000
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 3.837881941228233e-05,
+      "loss": 2.1257,
+      "step": 8032500
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 3.837809576463505e-05,
+      "loss": 2.1526,
+      "step": 8033000
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 3.837737211698777e-05,
+      "loss": 2.1614,
+      "step": 8033500
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 3.83766484693405e-05,
+      "loss": 2.1471,
+      "step": 8034000
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 3.8375924821693223e-05,
+      "loss": 2.1583,
+      "step": 8034500
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 3.8375201174045946e-05,
+      "loss": 2.1541,
+      "step": 8035000
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 3.837447752639867e-05,
+      "loss": 2.1617,
+      "step": 8035500
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 3.837375387875139e-05,
+      "loss": 2.1525,
+      "step": 8036000
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 3.837303023110411e-05,
+      "loss": 2.1786,
+      "step": 8036500
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 3.8372306583456835e-05,
+      "loss": 2.1409,
+      "step": 8037000
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 3.837158438310486e-05,
+      "loss": 2.1678,
+      "step": 8037500
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 3.837086073545758e-05,
+      "loss": 2.175,
+      "step": 8038000
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 3.83701370878103e-05,
+      "loss": 2.1549,
+      "step": 8038500
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 3.8369414887458324e-05,
+      "loss": 2.1561,
+      "step": 8039000
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 3.8368691239811046e-05,
+      "loss": 2.1541,
+      "step": 8039500
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 3.836796759216377e-05,
+      "loss": 2.1452,
+      "step": 8040000
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 3.836724394451649e-05,
+      "loss": 2.1547,
+      "step": 8040500
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 3.836652029686921e-05,
+      "loss": 2.159,
+      "step": 8041000
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 3.8365796649221935e-05,
+      "loss": 2.162,
+      "step": 8041500
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 3.836507300157466e-05,
+      "loss": 2.1683,
+      "step": 8042000
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 3.836434935392738e-05,
+      "loss": 2.1655,
+      "step": 8042500
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 3.83636257062801e-05,
+      "loss": 2.1738,
+      "step": 8043000
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 3.8362903505928124e-05,
+      "loss": 2.1574,
+      "step": 8043500
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 3.836218130557614e-05,
+      "loss": 2.1842,
+      "step": 8044000
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 3.836145765792886e-05,
+      "loss": 2.156,
+      "step": 8044500
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 3.8360734010281584e-05,
+      "loss": 2.1711,
+      "step": 8045000
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 3.836001036263431e-05,
+      "loss": 2.1611,
+      "step": 8045500
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 3.8359286714987035e-05,
+      "loss": 2.1576,
+      "step": 8046000
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 3.835856306733976e-05,
+      "loss": 2.1464,
+      "step": 8046500
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 3.835783941969248e-05,
+      "loss": 2.1397,
+      "step": 8047000
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 3.83571172193405e-05,
+      "loss": 2.1666,
+      "step": 8047500
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 3.8356393571693224e-05,
+      "loss": 2.1435,
+      "step": 8048000
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 3.8355669924045946e-05,
+      "loss": 2.1609,
+      "step": 8048500
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 3.835494627639867e-05,
+      "loss": 2.1652,
+      "step": 8049000
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 3.835422262875139e-05,
+      "loss": 2.1409,
+      "step": 8049500
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 3.835349898110411e-05,
+      "loss": 2.1627,
+      "step": 8050000
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 3.8352775333456835e-05,
+      "loss": 2.1641,
+      "step": 8050500
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 3.835205168580956e-05,
+      "loss": 2.1255,
+      "step": 8051000
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 3.835132948545758e-05,
+      "loss": 2.1319,
+      "step": 8051500
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 3.83506058378103e-05,
+      "loss": 2.1567,
+      "step": 8052000
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 3.834988219016303e-05,
+      "loss": 2.1554,
+      "step": 8052500
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 3.8349158542515753e-05,
+      "loss": 2.16,
+      "step": 8053000
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 3.834843634216377e-05,
+      "loss": 2.1413,
+      "step": 8053500
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 3.834771269451649e-05,
+      "loss": 2.1649,
+      "step": 8054000
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 3.8346989046869213e-05,
+      "loss": 2.1811,
+      "step": 8054500
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 3.8346265399221936e-05,
+      "loss": 2.1672,
+      "step": 8055000
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 3.834554175157466e-05,
+      "loss": 2.1631,
+      "step": 8055500
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 3.834481810392738e-05,
+      "loss": 2.152,
+      "step": 8056000
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 3.83440959035754e-05,
+      "loss": 2.1567,
+      "step": 8056500
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 3.8343372255928125e-05,
+      "loss": 2.1679,
+      "step": 8057000
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 3.834264860828085e-05,
+      "loss": 2.193,
+      "step": 8057500
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 3.834192496063357e-05,
+      "loss": 2.1605,
+      "step": 8058000
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 3.834120131298629e-05,
+      "loss": 2.1514,
+      "step": 8058500
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 3.83404805599296e-05,
+      "loss": 2.1341,
+      "step": 8059000
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 3.833975691228233e-05,
+      "loss": 2.1524,
+      "step": 8059500
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 3.833903326463505e-05,
+      "loss": 2.1455,
+      "step": 8060000
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 3.833830961698778e-05,
+      "loss": 2.1356,
+      "step": 8060500
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 3.83375859693405e-05,
+      "loss": 2.142,
+      "step": 8061000
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 3.8336862321693225e-05,
+      "loss": 2.1608,
+      "step": 8061500
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 3.833613867404595e-05,
+      "loss": 2.1584,
+      "step": 8062000
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 3.833541502639867e-05,
+      "loss": 2.1618,
+      "step": 8062500
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 3.833469137875139e-05,
+      "loss": 2.152,
+      "step": 8063000
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 3.8333967731104114e-05,
+      "loss": 2.1822,
+      "step": 8063500
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 3.833324553075213e-05,
+      "loss": 2.1364,
+      "step": 8064000
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 3.833252188310485e-05,
+      "loss": 2.1742,
+      "step": 8064500
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 3.833179823545758e-05,
+      "loss": 2.1271,
+      "step": 8065000
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 3.8331076035105596e-05,
+      "loss": 2.1706,
+      "step": 8065500
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 3.833035238745832e-05,
+      "loss": 2.1534,
+      "step": 8066000
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 3.832962873981104e-05,
+      "loss": 2.188,
+      "step": 8066500
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 3.832890509216376e-05,
+      "loss": 2.143,
+      "step": 8067000
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 3.832818144451649e-05,
+      "loss": 2.1687,
+      "step": 8067500
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 3.8327457796869214e-05,
+      "loss": 2.1594,
+      "step": 8068000
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 3.8326734149221936e-05,
+      "loss": 2.1665,
+      "step": 8068500
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 3.832601050157466e-05,
+      "loss": 2.1698,
+      "step": 8069000
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 3.832528685392738e-05,
+      "loss": 2.146,
+      "step": 8069500
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 3.83245646535754e-05,
+      "loss": 2.1528,
+      "step": 8070000
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 3.8323841005928125e-05,
+      "loss": 2.1544,
+      "step": 8070500
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 3.832311735828085e-05,
+      "loss": 2.1589,
+      "step": 8071000
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 3.832239515792886e-05,
+      "loss": 2.1878,
+      "step": 8071500
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 3.8321671510281586e-05,
+      "loss": 2.1357,
+      "step": 8072000
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 3.832094786263431e-05,
+      "loss": 2.1545,
+      "step": 8072500
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 3.832022421498703e-05,
+      "loss": 2.156,
+      "step": 8073000
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 3.831950056733975e-05,
+      "loss": 2.1636,
+      "step": 8073500
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 3.831877691969248e-05,
+      "loss": 2.1689,
+      "step": 8074000
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 3.831805616663579e-05,
+      "loss": 2.1667,
+      "step": 8074500
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 3.831733251898851e-05,
+      "loss": 2.1807,
+      "step": 8075000
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 3.831660887134124e-05,
+      "loss": 2.1491,
+      "step": 8075500
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 3.8315885223693964e-05,
+      "loss": 2.1855,
+      "step": 8076000
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 3.8315161576046686e-05,
+      "loss": 2.1419,
+      "step": 8076500
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 3.831443792839941e-05,
+      "loss": 2.1618,
+      "step": 8077000
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 3.831371428075213e-05,
+      "loss": 2.1637,
+      "step": 8077500
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 3.831299063310486e-05,
+      "loss": 2.1749,
+      "step": 8078000
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 3.831226698545758e-05,
+      "loss": 2.1335,
+      "step": 8078500
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 3.8311543337810304e-05,
+      "loss": 2.1451,
+      "step": 8079000
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 3.8310819690163026e-05,
+      "loss": 2.1549,
+      "step": 8079500
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 3.831009604251575e-05,
+      "loss": 2.131,
+      "step": 8080000
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 3.830937239486847e-05,
+      "loss": 2.1704,
+      "step": 8080500
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 3.830864874722119e-05,
+      "loss": 2.1779,
+      "step": 8081000
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 3.8307925099573915e-05,
+      "loss": 2.1493,
+      "step": 8081500
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 3.830720145192664e-05,
+      "loss": 2.1648,
+      "step": 8082000
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 3.8306477804279366e-05,
+      "loss": 2.1504,
+      "step": 8082500
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 3.830575415663209e-05,
+      "loss": 2.1686,
+      "step": 8083000
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 3.830503050898481e-05,
+      "loss": 2.1385,
+      "step": 8083500
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 3.830430686133753e-05,
+      "loss": 2.1396,
+      "step": 8084000
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 3.830358610828085e-05,
+      "loss": 2.1813,
+      "step": 8084500
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 3.830286246063357e-05,
+      "loss": 2.1622,
+      "step": 8085000
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 3.830213881298629e-05,
+      "loss": 2.1518,
+      "step": 8085500
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 3.8301415165339015e-05,
+      "loss": 2.164,
+      "step": 8086000
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 3.830069151769174e-05,
+      "loss": 2.1583,
+      "step": 8086500
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 3.829996787004446e-05,
+      "loss": 2.1553,
+      "step": 8087000
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 3.829924422239718e-05,
+      "loss": 2.1529,
+      "step": 8087500
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 3.829852057474991e-05,
+      "loss": 2.1819,
+      "step": 8088000
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 3.829779692710263e-05,
+      "loss": 2.1666,
+      "step": 8088500
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 3.8297073279455355e-05,
+      "loss": 2.1452,
+      "step": 8089000
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 3.8296349631808084e-05,
+      "loss": 2.1529,
+      "step": 8089500
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 3.829562598416081e-05,
+      "loss": 2.178,
+      "step": 8090000
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 3.829490233651353e-05,
+      "loss": 2.1613,
+      "step": 8090500
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 3.8294180136161544e-05,
+      "loss": 2.1622,
+      "step": 8091000
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 3.829345648851427e-05,
+      "loss": 2.1375,
+      "step": 8091500
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 3.829273284086699e-05,
+      "loss": 2.1878,
+      "step": 8092000
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 3.829201064051501e-05,
+      "loss": 2.1726,
+      "step": 8092500
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 3.8291286992867733e-05,
+      "loss": 2.1887,
+      "step": 8093000
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 3.829056479251575e-05,
+      "loss": 2.1584,
+      "step": 8093500
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 3.828984114486847e-05,
+      "loss": 2.1612,
+      "step": 8094000
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 3.8289117497221193e-05,
+      "loss": 2.156,
+      "step": 8094500
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 3.8288393849573916e-05,
+      "loss": 2.1661,
+      "step": 8095000
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 3.828767020192664e-05,
+      "loss": 2.1471,
+      "step": 8095500
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 3.828694655427936e-05,
+      "loss": 2.1523,
+      "step": 8096000
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 3.828622290663208e-05,
+      "loss": 2.1495,
+      "step": 8096500
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 3.8285500706280105e-05,
+      "loss": 2.1502,
+      "step": 8097000
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 3.8284777058632834e-05,
+      "loss": 2.1731,
+      "step": 8097500
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 3.8284053410985556e-05,
+      "loss": 2.1594,
+      "step": 8098000
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 3.828332976333828e-05,
+      "loss": 2.1675,
+      "step": 8098500
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 3.8282606115691e-05,
+      "loss": 2.1627,
+      "step": 8099000
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 3.828188246804372e-05,
+      "loss": 2.1564,
+      "step": 8099500
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 3.8281158820396445e-05,
+      "loss": 2.1466,
+      "step": 8100000
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 3.828043806733976e-05,
+      "loss": 2.1667,
+      "step": 8100500
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 3.827971441969248e-05,
+      "loss": 2.1631,
+      "step": 8101000
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 3.8278990772045205e-05,
+      "loss": 2.1652,
+      "step": 8101500
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 3.827826712439793e-05,
+      "loss": 2.1661,
+      "step": 8102000
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 3.827754347675065e-05,
+      "loss": 2.1744,
+      "step": 8102500
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 3.827681982910337e-05,
+      "loss": 2.1489,
+      "step": 8103000
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 3.8276096181456094e-05,
+      "loss": 2.1685,
+      "step": 8103500
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 3.8275372533808816e-05,
+      "loss": 2.166,
+      "step": 8104000
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 3.827464888616154e-05,
+      "loss": 2.1722,
+      "step": 8104500
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 3.827392523851427e-05,
+      "loss": 2.1875,
+      "step": 8105000
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 3.827320303816228e-05,
+      "loss": 2.1437,
+      "step": 8105500
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 3.827247939051501e-05,
+      "loss": 2.1674,
+      "step": 8106000
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 3.8271755742867734e-05,
+      "loss": 2.1638,
+      "step": 8106500
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 3.8271032095220456e-05,
+      "loss": 2.1623,
+      "step": 8107000
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 3.827030844757318e-05,
+      "loss": 2.1803,
+      "step": 8107500
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 3.82695847999259e-05,
+      "loss": 2.1605,
+      "step": 8108000
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 3.8268862599573917e-05,
+      "loss": 2.1829,
+      "step": 8108500
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 3.826813895192664e-05,
+      "loss": 2.1731,
+      "step": 8109000
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 3.826741530427936e-05,
+      "loss": 2.1725,
+      "step": 8109500
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 3.826669165663208e-05,
+      "loss": 2.1808,
+      "step": 8110000
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 3.826596800898481e-05,
+      "loss": 2.1528,
+      "step": 8110500
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 3.8265244361337534e-05,
+      "loss": 2.1356,
+      "step": 8111000
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 3.826452071369026e-05,
+      "loss": 2.1422,
+      "step": 8111500
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 3.8263797066042986e-05,
+      "loss": 2.1799,
+      "step": 8112000
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 3.8263074865691e-05,
+      "loss": 2.1856,
+      "step": 8112500
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 3.8262351218043723e-05,
+      "loss": 2.1643,
+      "step": 8113000
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 3.8261627570396446e-05,
+      "loss": 2.1486,
+      "step": 8113500
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 3.826090392274917e-05,
+      "loss": 2.1454,
+      "step": 8114000
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 3.826018027510189e-05,
+      "loss": 2.1712,
+      "step": 8114500
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 3.825945662745461e-05,
+      "loss": 2.1691,
+      "step": 8115000
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 3.8258732979807335e-05,
+      "loss": 2.1571,
+      "step": 8115500
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 3.825801077945536e-05,
+      "loss": 2.1648,
+      "step": 8116000
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 3.825728713180808e-05,
+      "loss": 2.1826,
+      "step": 8116500
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 3.82565634841608e-05,
+      "loss": 2.1806,
+      "step": 8117000
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 3.8255839836513524e-05,
+      "loss": 2.1382,
+      "step": 8117500
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 3.8255116188866246e-05,
+      "loss": 2.1702,
+      "step": 8118000
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 3.825439254121897e-05,
+      "loss": 2.1434,
+      "step": 8118500
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 3.825367034086699e-05,
+      "loss": 2.1758,
+      "step": 8119000
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 3.825294669321971e-05,
+      "loss": 2.1716,
+      "step": 8119500
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 3.825222304557244e-05,
+      "loss": 2.1926,
+      "step": 8120000
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 3.8251499397925164e-05,
+      "loss": 2.1539,
+      "step": 8120500
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 3.8250775750277886e-05,
+      "loss": 2.1306,
+      "step": 8121000
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 3.825005210263061e-05,
+      "loss": 2.1608,
+      "step": 8121500
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 3.8249329902278624e-05,
+      "loss": 2.1581,
+      "step": 8122000
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 3.824860770192664e-05,
+      "loss": 2.1733,
+      "step": 8122500
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 3.824788405427936e-05,
+      "loss": 2.1464,
+      "step": 8123000
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 3.824716040663209e-05,
+      "loss": 2.169,
+      "step": 8123500
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 3.824643675898481e-05,
+      "loss": 2.1347,
+      "step": 8124000
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 3.8245713111337535e-05,
+      "loss": 2.1534,
+      "step": 8124500
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 3.824499091098555e-05,
+      "loss": 2.1684,
+      "step": 8125000
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 3.824426726333827e-05,
+      "loss": 2.1778,
+      "step": 8125500
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 3.8243543615690995e-05,
+      "loss": 2.1803,
+      "step": 8126000
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 3.824281996804372e-05,
+      "loss": 2.1998,
+      "step": 8126500
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 3.824209632039644e-05,
+      "loss": 2.1521,
+      "step": 8127000
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 3.824137267274917e-05,
+      "loss": 2.1643,
+      "step": 8127500
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 3.824064902510189e-05,
+      "loss": 2.1551,
+      "step": 8128000
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 3.823992537745461e-05,
+      "loss": 2.1777,
+      "step": 8128500
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 3.8239203177102636e-05,
+      "loss": 2.1739,
+      "step": 8129000
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 3.823847952945536e-05,
+      "loss": 2.1587,
+      "step": 8129500
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 3.823775732910337e-05,
+      "loss": 2.1541,
+      "step": 8130000
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 3.8237033681456096e-05,
+      "loss": 2.1632,
+      "step": 8130500
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 3.823631003380882e-05,
+      "loss": 2.1414,
+      "step": 8131000
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 3.823558638616154e-05,
+      "loss": 2.1897,
+      "step": 8131500
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 3.823486273851426e-05,
+      "loss": 2.15,
+      "step": 8132000
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 3.823413909086699e-05,
+      "loss": 2.1688,
+      "step": 8132500
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 3.8233415443219714e-05,
+      "loss": 2.1456,
+      "step": 8133000
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 3.8232691795572436e-05,
+      "loss": 2.1744,
+      "step": 8133500
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 3.823196814792516e-05,
+      "loss": 2.1384,
+      "step": 8134000
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 3.823124450027789e-05,
+      "loss": 2.1718,
+      "step": 8134500
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 3.823052085263061e-05,
+      "loss": 2.1671,
+      "step": 8135000
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 3.822979720498333e-05,
+      "loss": 2.1511,
+      "step": 8135500
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 3.8229073557336054e-05,
+      "loss": 2.1734,
+      "step": 8136000
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 3.822835135698407e-05,
+      "loss": 2.1548,
+      "step": 8136500
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 3.822762770933679e-05,
+      "loss": 2.1718,
+      "step": 8137000
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 3.8226904061689514e-05,
+      "loss": 2.1741,
+      "step": 8137500
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 3.822618041404224e-05,
+      "loss": 2.1555,
+      "step": 8138000
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 3.8225456766394965e-05,
+      "loss": 2.1442,
+      "step": 8138500
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 3.822473311874769e-05,
+      "loss": 2.1682,
+      "step": 8139000
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 3.822400947110041e-05,
+      "loss": 2.1603,
+      "step": 8139500
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 3.8223287270748425e-05,
+      "loss": 2.146,
+      "step": 8140000
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 3.822256362310115e-05,
+      "loss": 2.1843,
+      "step": 8140500
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 3.822183997545387e-05,
+      "loss": 2.1576,
+      "step": 8141000
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 3.822111632780659e-05,
+      "loss": 2.1579,
+      "step": 8141500
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 3.8220394127454614e-05,
+      "loss": 2.1499,
+      "step": 8142000
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 3.821967047980734e-05,
+      "loss": 2.1718,
+      "step": 8142500
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 3.8218946832160065e-05,
+      "loss": 2.1596,
+      "step": 8143000
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 3.821822318451279e-05,
+      "loss": 2.1691,
+      "step": 8143500
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 3.821749953686551e-05,
+      "loss": 2.1697,
+      "step": 8144000
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 3.821677588921823e-05,
+      "loss": 2.1672,
+      "step": 8144500
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 3.821605368886625e-05,
+      "loss": 2.1731,
+      "step": 8145000
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 3.821533004121897e-05,
+      "loss": 2.167,
+      "step": 8145500
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 3.821460639357169e-05,
+      "loss": 2.1822,
+      "step": 8146000
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 3.8213882745924414e-05,
+      "loss": 2.1531,
+      "step": 8146500
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 3.821315909827714e-05,
+      "loss": 2.1819,
+      "step": 8147000
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 3.8212435450629865e-05,
+      "loss": 2.1773,
+      "step": 8147500
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 3.821171325027788e-05,
+      "loss": 2.1347,
+      "step": 8148000
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 3.82109896026306e-05,
+      "loss": 2.1964,
+      "step": 8148500
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 3.8210265954983325e-05,
+      "loss": 2.1456,
+      "step": 8149000
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 3.8209542307336054e-05,
+      "loss": 2.166,
+      "step": 8149500
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 3.820881865968878e-05,
+      "loss": 2.1605,
+      "step": 8150000
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 3.820809645933679e-05,
+      "loss": 2.1586,
+      "step": 8150500
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 3.8207374258984815e-05,
+      "loss": 2.1788,
+      "step": 8151000
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 3.820665061133754e-05,
+      "loss": 2.1458,
+      "step": 8151500
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 3.820592696369026e-05,
+      "loss": 2.1648,
+      "step": 8152000
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 3.820520331604298e-05,
+      "loss": 2.1648,
+      "step": 8152500
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 3.8204479668395704e-05,
+      "loss": 2.181,
+      "step": 8153000
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 3.8203756020748426e-05,
+      "loss": 2.1581,
+      "step": 8153500
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 3.820303237310115e-05,
+      "loss": 2.1623,
+      "step": 8154000
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 3.820230872545387e-05,
+      "loss": 2.1895,
+      "step": 8154500
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 3.820158507780659e-05,
+      "loss": 2.1604,
+      "step": 8155000
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 3.820086143015932e-05,
+      "loss": 2.1569,
+      "step": 8155500
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 3.8200137782512044e-05,
+      "loss": 2.1513,
+      "step": 8156000
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 3.819941558216006e-05,
+      "loss": 2.1729,
+      "step": 8156500
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 3.819869193451279e-05,
+      "loss": 2.1657,
+      "step": 8157000
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 3.819796828686551e-05,
+      "loss": 2.1643,
+      "step": 8157500
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 3.819724463921823e-05,
+      "loss": 2.1793,
+      "step": 8158000
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 3.8196520991570955e-05,
+      "loss": 2.1531,
+      "step": 8158500
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 3.819579879121897e-05,
+      "loss": 2.1506,
+      "step": 8159000
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 3.819507514357169e-05,
+      "loss": 2.1548,
+      "step": 8159500
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 3.819435149592442e-05,
+      "loss": 2.1676,
+      "step": 8160000
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 3.8193627848277144e-05,
+      "loss": 2.1442,
+      "step": 8160500
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 3.8192904200629866e-05,
+      "loss": 2.1657,
+      "step": 8161000
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 3.819218055298259e-05,
+      "loss": 2.1817,
+      "step": 8161500
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 3.819145690533531e-05,
+      "loss": 2.1683,
+      "step": 8162000
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 3.819073325768803e-05,
+      "loss": 2.1698,
+      "step": 8162500
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 3.8190009610040755e-05,
+      "loss": 2.1682,
+      "step": 8163000
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 3.818928596239348e-05,
+      "loss": 2.1602,
+      "step": 8163500
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 3.818856376204149e-05,
+      "loss": 2.1651,
+      "step": 8164000
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 3.818784011439422e-05,
+      "loss": 2.1676,
+      "step": 8164500
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 3.8187116466746944e-05,
+      "loss": 2.1715,
+      "step": 8165000
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 3.818639281909967e-05,
+      "loss": 2.1686,
+      "step": 8165500
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 3.8185669171452395e-05,
+      "loss": 2.1636,
+      "step": 8166000
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 3.818494552380512e-05,
+      "loss": 2.1742,
+      "step": 8166500
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 3.818422332345313e-05,
+      "loss": 2.1632,
+      "step": 8167000
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 3.8183499675805855e-05,
+      "loss": 2.1405,
+      "step": 8167500
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 3.818277602815858e-05,
+      "loss": 2.1556,
+      "step": 8168000
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 3.81820523805113e-05,
+      "loss": 2.1711,
+      "step": 8168500
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 3.818133018015932e-05,
+      "loss": 2.1692,
+      "step": 8169000
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 3.8180606532512045e-05,
+      "loss": 2.1519,
+      "step": 8169500
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 3.817988288486477e-05,
+      "loss": 2.1584,
+      "step": 8170000
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 3.817915923721749e-05,
+      "loss": 2.1754,
+      "step": 8170500
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 3.817843558957021e-05,
+      "loss": 2.1513,
+      "step": 8171000
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 3.817771194192294e-05,
+      "loss": 2.1734,
+      "step": 8171500
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 3.8176989741570956e-05,
+      "loss": 2.1654,
+      "step": 8172000
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 3.817626609392368e-05,
+      "loss": 2.1565,
+      "step": 8172500
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 3.81755424462764e-05,
+      "loss": 2.1423,
+      "step": 8173000
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 3.817481879862912e-05,
+      "loss": 2.1542,
+      "step": 8173500
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 3.8174096598277145e-05,
+      "loss": 2.1716,
+      "step": 8174000
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 3.817337295062987e-05,
+      "loss": 2.1389,
+      "step": 8174500
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 3.817265075027788e-05,
+      "loss": 2.1575,
+      "step": 8175000
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 3.8171927102630605e-05,
+      "loss": 2.1508,
+      "step": 8175500
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 3.817120345498333e-05,
+      "loss": 2.176,
+      "step": 8176000
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 3.817047980733605e-05,
+      "loss": 2.1682,
+      "step": 8176500
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 3.816975615968877e-05,
+      "loss": 2.1711,
+      "step": 8177000
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 3.8169033959336794e-05,
+      "loss": 2.1749,
+      "step": 8177500
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 3.816831175898481e-05,
+      "loss": 2.1356,
+      "step": 8178000
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 3.816758811133753e-05,
+      "loss": 2.1597,
+      "step": 8178500
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 3.8166864463690254e-05,
+      "loss": 2.1596,
+      "step": 8179000
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 3.816614081604298e-05,
+      "loss": 2.1462,
+      "step": 8179500
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 3.8165417168395705e-05,
+      "loss": 2.183,
+      "step": 8180000
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 3.816469352074843e-05,
+      "loss": 2.1572,
+      "step": 8180500
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 3.816396987310115e-05,
+      "loss": 2.1823,
+      "step": 8181000
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 3.816324622545387e-05,
+      "loss": 2.165,
+      "step": 8181500
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 3.81625225778066e-05,
+      "loss": 2.1467,
+      "step": 8182000
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 3.816179893015932e-05,
+      "loss": 2.1667,
+      "step": 8182500
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 3.8161075282512045e-05,
+      "loss": 2.1691,
+      "step": 8183000
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 3.816035163486477e-05,
+      "loss": 2.1658,
+      "step": 8183500
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 3.815962798721749e-05,
+      "loss": 2.1463,
+      "step": 8184000
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 3.815890433957021e-05,
+      "loss": 2.1441,
+      "step": 8184500
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 3.8158180691922934e-05,
+      "loss": 2.1675,
+      "step": 8185000
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 3.8157457044275656e-05,
+      "loss": 2.1554,
+      "step": 8185500
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 3.815673484392367e-05,
+      "loss": 2.1299,
+      "step": 8186000
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 3.81560111962764e-05,
+      "loss": 2.1299,
+      "step": 8186500
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 3.815528754862912e-05,
+      "loss": 2.1451,
+      "step": 8187000
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 3.815456390098185e-05,
+      "loss": 2.1558,
+      "step": 8187500
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 3.8153840253334575e-05,
+      "loss": 2.1581,
+      "step": 8188000
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 3.81531166056873e-05,
+      "loss": 2.1481,
+      "step": 8188500
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 3.815239440533531e-05,
+      "loss": 2.1685,
+      "step": 8189000
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 3.8151670757688035e-05,
+      "loss": 2.165,
+      "step": 8189500
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 3.815094855733605e-05,
+      "loss": 2.1563,
+      "step": 8190000
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 3.815022490968877e-05,
+      "loss": 2.1851,
+      "step": 8190500
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 3.81495012620415e-05,
+      "loss": 2.1733,
+      "step": 8191000
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 3.8148777614394224e-05,
+      "loss": 2.153,
+      "step": 8191500
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 3.8148053966746946e-05,
+      "loss": 2.1626,
+      "step": 8192000
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 3.814733176639496e-05,
+      "loss": 2.1456,
+      "step": 8192500
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 3.814660956604298e-05,
+      "loss": 2.1472,
+      "step": 8193000
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 3.81458859183957e-05,
+      "loss": 2.1673,
+      "step": 8193500
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 3.814516371804372e-05,
+      "loss": 2.1704,
+      "step": 8194000
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 3.8144440070396444e-05,
+      "loss": 2.1626,
+      "step": 8194500
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 3.8143717870044466e-05,
+      "loss": 2.1555,
+      "step": 8195000
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 3.814299422239719e-05,
+      "loss": 2.1598,
+      "step": 8195500
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 3.814227057474991e-05,
+      "loss": 2.1894,
+      "step": 8196000
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 3.814154692710263e-05,
+      "loss": 2.1539,
+      "step": 8196500
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 3.8140823279455355e-05,
+      "loss": 2.1522,
+      "step": 8197000
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 3.814009963180808e-05,
+      "loss": 2.1872,
+      "step": 8197500
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 3.81393759841608e-05,
+      "loss": 2.1452,
+      "step": 8198000
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 3.813865233651353e-05,
+      "loss": 2.1418,
+      "step": 8198500
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 3.813792868886625e-05,
+      "loss": 2.1533,
+      "step": 8199000
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 3.813720504121897e-05,
+      "loss": 2.1705,
+      "step": 8199500
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 3.8136481393571695e-05,
+      "loss": 2.1527,
+      "step": 8200000
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 3.813575774592442e-05,
+      "loss": 2.1689,
+      "step": 8200500
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 3.813503409827714e-05,
+      "loss": 2.1685,
+      "step": 8201000
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 3.813431045062986e-05,
+      "loss": 2.1583,
+      "step": 8201500
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 3.813358680298259e-05,
+      "loss": 2.1638,
+      "step": 8202000
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 3.813286315533531e-05,
+      "loss": 2.1887,
+      "step": 8202500
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 3.8132139507688035e-05,
+      "loss": 2.1681,
+      "step": 8203000
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 3.813141586004076e-05,
+      "loss": 2.1706,
+      "step": 8203500
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 3.813069221239348e-05,
+      "loss": 2.1541,
+      "step": 8204000
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 3.81299700120415e-05,
+      "loss": 2.1435,
+      "step": 8204500
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 3.8129246364394224e-05,
+      "loss": 2.17,
+      "step": 8205000
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 3.8128522716746947e-05,
+      "loss": 2.1855,
+      "step": 8205500
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 3.812779906909967e-05,
+      "loss": 2.1642,
+      "step": 8206000
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 3.812707542145239e-05,
+      "loss": 2.1594,
+      "step": 8206500
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 3.812635177380511e-05,
+      "loss": 2.1621,
+      "step": 8207000
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 3.8125628126157836e-05,
+      "loss": 2.1387,
+      "step": 8207500
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 3.812490447851056e-05,
+      "loss": 2.1518,
+      "step": 8208000
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 3.812418083086328e-05,
+      "loss": 2.1803,
+      "step": 8208500
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 3.812345718321601e-05,
+      "loss": 2.1733,
+      "step": 8209000
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 3.8122734982864025e-05,
+      "loss": 2.1558,
+      "step": 8209500
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 3.812201278251205e-05,
+      "loss": 2.1262,
+      "step": 8210000
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 3.812128913486477e-05,
+      "loss": 2.1506,
+      "step": 8210500
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 3.812056548721749e-05,
+      "loss": 2.1664,
+      "step": 8211000
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 3.8119841839570214e-05,
+      "loss": 2.1344,
+      "step": 8211500
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 3.8119118191922936e-05,
+      "loss": 2.1596,
+      "step": 8212000
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 3.811839454427566e-05,
+      "loss": 2.1568,
+      "step": 8212500
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 3.811767089662838e-05,
+      "loss": 2.1704,
+      "step": 8213000
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 3.8116950143571696e-05,
+      "loss": 2.1406,
+      "step": 8213500
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 3.811622649592442e-05,
+      "loss": 2.1838,
+      "step": 8214000
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 3.811550284827714e-05,
+      "loss": 2.1785,
+      "step": 8214500
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 3.811477920062986e-05,
+      "loss": 2.1736,
+      "step": 8215000
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 3.8114055552982585e-05,
+      "loss": 2.1585,
+      "step": 8215500
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 3.811333190533531e-05,
+      "loss": 2.1485,
+      "step": 8216000
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 3.811260825768803e-05,
+      "loss": 2.1771,
+      "step": 8216500
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 3.811188461004076e-05,
+      "loss": 2.1565,
+      "step": 8217000
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 3.811116096239348e-05,
+      "loss": 2.1305,
+      "step": 8217500
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 3.81104373147462e-05,
+      "loss": 2.1484,
+      "step": 8218000
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 3.810971366709893e-05,
+      "loss": 2.1589,
+      "step": 8218500
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 3.8108990019451654e-05,
+      "loss": 2.1727,
+      "step": 8219000
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 3.810826926639496e-05,
+      "loss": 2.1719,
+      "step": 8219500
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 3.8107545618747685e-05,
+      "loss": 2.1549,
+      "step": 8220000
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 3.810682197110041e-05,
+      "loss": 2.1434,
+      "step": 8220500
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 3.810609832345313e-05,
+      "loss": 2.1574,
+      "step": 8221000
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 3.810537467580585e-05,
+      "loss": 2.1528,
+      "step": 8221500
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 3.810465102815858e-05,
+      "loss": 2.163,
+      "step": 8222000
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 3.81039273805113e-05,
+      "loss": 2.1647,
+      "step": 8222500
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 3.8103203732864025e-05,
+      "loss": 2.1577,
+      "step": 8223000
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 3.810248008521675e-05,
+      "loss": 2.1901,
+      "step": 8223500
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 3.8101756437569477e-05,
+      "loss": 2.1336,
+      "step": 8224000
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 3.81010327899222e-05,
+      "loss": 2.1838,
+      "step": 8224500
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 3.810030914227492e-05,
+      "loss": 2.1539,
+      "step": 8225000
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 3.809958694192294e-05,
+      "loss": 2.1486,
+      "step": 8225500
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 3.809886329427566e-05,
+      "loss": 2.1652,
+      "step": 8226000
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 3.809813964662838e-05,
+      "loss": 2.1783,
+      "step": 8226500
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 3.80974159989811e-05,
+      "loss": 2.1726,
+      "step": 8227000
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 3.809669235133383e-05,
+      "loss": 2.1881,
+      "step": 8227500
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 3.8095968703686555e-05,
+      "loss": 2.1718,
+      "step": 8228000
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 3.809524505603928e-05,
+      "loss": 2.1634,
+      "step": 8228500
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 3.8094521408392e-05,
+      "loss": 2.1803,
+      "step": 8229000
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 3.809379776074472e-05,
+      "loss": 2.1676,
+      "step": 8229500
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 3.8093074113097444e-05,
+      "loss": 2.1476,
+      "step": 8230000
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 3.8092350465450166e-05,
+      "loss": 2.1511,
+      "step": 8230500
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 3.809162826509818e-05,
+      "loss": 2.152,
+      "step": 8231000
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 3.809090461745091e-05,
+      "loss": 2.151,
+      "step": 8231500
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 3.809018096980363e-05,
+      "loss": 2.1798,
+      "step": 8232000
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 3.8089457322156355e-05,
+      "loss": 2.154,
+      "step": 8232500
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 3.8088733674509084e-05,
+      "loss": 2.176,
+      "step": 8233000
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 3.80880114741571e-05,
+      "loss": 2.1372,
+      "step": 8233500
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 3.808728782650982e-05,
+      "loss": 2.1523,
+      "step": 8234000
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 3.8086564178862544e-05,
+      "loss": 2.1458,
+      "step": 8234500
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 3.8085840531215266e-05,
+      "loss": 2.1608,
+      "step": 8235000
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 3.808511688356799e-05,
+      "loss": 2.1817,
+      "step": 8235500
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 3.808439323592071e-05,
+      "loss": 2.1701,
+      "step": 8236000
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 3.808367103556873e-05,
+      "loss": 2.1682,
+      "step": 8236500
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 3.808294883521675e-05,
+      "loss": 2.1494,
+      "step": 8237000
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 3.8082226634864764e-05,
+      "loss": 2.1437,
+      "step": 8237500
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 3.8081502987217486e-05,
+      "loss": 2.1672,
+      "step": 8238000
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 3.808077933957021e-05,
+      "loss": 2.1494,
+      "step": 8238500
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 3.808005569192294e-05,
+      "loss": 2.1635,
+      "step": 8239000
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 3.807933349157096e-05,
+      "loss": 2.1551,
+      "step": 8239500
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 3.807860984392368e-05,
+      "loss": 2.1856,
+      "step": 8240000
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 3.8077886196276404e-05,
+      "loss": 2.1547,
+      "step": 8240500
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 3.8077162548629126e-05,
+      "loss": 2.1465,
+      "step": 8241000
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 3.807643890098185e-05,
+      "loss": 2.1627,
+      "step": 8241500
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 3.807571525333457e-05,
+      "loss": 2.1772,
+      "step": 8242000
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 3.807499160568729e-05,
+      "loss": 2.1372,
+      "step": 8242500
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 3.8074267958040015e-05,
+      "loss": 2.1626,
+      "step": 8243000
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 3.807354431039274e-05,
+      "loss": 2.1571,
+      "step": 8243500
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 3.807282066274546e-05,
+      "loss": 2.1608,
+      "step": 8244000
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 3.807209701509818e-05,
+      "loss": 2.1878,
+      "step": 8244500
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 3.8071373367450904e-05,
+      "loss": 2.1578,
+      "step": 8245000
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 3.807065116709893e-05,
+      "loss": 2.1823,
+      "step": 8245500
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 3.806992896674694e-05,
+      "loss": 2.1539,
+      "step": 8246000
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 3.806920531909967e-05,
+      "loss": 2.181,
+      "step": 8246500
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 3.8068481671452393e-05,
+      "loss": 2.1546,
+      "step": 8247000
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 3.806775947110041e-05,
+      "loss": 2.1418,
+      "step": 8247500
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 3.806703582345313e-05,
+      "loss": 2.1687,
+      "step": 8248000
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 3.806631217580586e-05,
+      "loss": 2.1731,
+      "step": 8248500
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 3.806558852815858e-05,
+      "loss": 2.18,
+      "step": 8249000
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 3.8064864880511305e-05,
+      "loss": 2.1846,
+      "step": 8249500
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 3.806414123286403e-05,
+      "loss": 2.1556,
+      "step": 8250000
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 3.806341758521675e-05,
+      "loss": 2.1636,
+      "step": 8250500
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 3.806269393756947e-05,
+      "loss": 2.1484,
+      "step": 8251000
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 3.8061970289922194e-05,
+      "loss": 2.141,
+      "step": 8251500
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 3.8061246642274916e-05,
+      "loss": 2.1538,
+      "step": 8252000
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 3.806052299462764e-05,
+      "loss": 2.1404,
+      "step": 8252500
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 3.805980079427566e-05,
+      "loss": 2.1597,
+      "step": 8253000
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 3.805907714662838e-05,
+      "loss": 2.1495,
+      "step": 8253500
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 3.805835349898111e-05,
+      "loss": 2.1631,
+      "step": 8254000
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 3.8057629851333834e-05,
+      "loss": 2.161,
+      "step": 8254500
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 3.8056906203686556e-05,
+      "loss": 2.1557,
+      "step": 8255000
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 3.805618255603928e-05,
+      "loss": 2.1608,
+      "step": 8255500
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 3.8055458908392e-05,
+      "loss": 2.1414,
+      "step": 8256000
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 3.805473526074472e-05,
+      "loss": 2.156,
+      "step": 8256500
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 3.8054011613097445e-05,
+      "loss": 2.1595,
+      "step": 8257000
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 3.805328796545017e-05,
+      "loss": 2.1569,
+      "step": 8257500
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 3.805256431780289e-05,
+      "loss": 2.1546,
+      "step": 8258000
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 3.805184067015561e-05,
+      "loss": 2.1463,
+      "step": 8258500
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 3.8051118469803634e-05,
+      "loss": 2.1783,
+      "step": 8259000
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 3.8050394822156356e-05,
+      "loss": 2.1647,
+      "step": 8259500
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 3.804967262180437e-05,
+      "loss": 2.1527,
+      "step": 8260000
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 3.804895042145239e-05,
+      "loss": 2.1926,
+      "step": 8260500
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 3.804822677380511e-05,
+      "loss": 2.1554,
+      "step": 8261000
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 3.804750312615784e-05,
+      "loss": 2.1503,
+      "step": 8261500
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 3.804677947851056e-05,
+      "loss": 2.1311,
+      "step": 8262000
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 3.804605583086328e-05,
+      "loss": 2.1653,
+      "step": 8262500
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 3.804533218321601e-05,
+      "loss": 2.1615,
+      "step": 8263000
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 3.8044608535568734e-05,
+      "loss": 2.1605,
+      "step": 8263500
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 3.804388488792146e-05,
+      "loss": 2.1875,
+      "step": 8264000
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 3.804316124027418e-05,
+      "loss": 2.1689,
+      "step": 8264500
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 3.80424375926269e-05,
+      "loss": 2.1572,
+      "step": 8265000
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 3.804171539227492e-05,
+      "loss": 2.1518,
+      "step": 8265500
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 3.804099174462764e-05,
+      "loss": 2.1638,
+      "step": 8266000
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 3.804026809698036e-05,
+      "loss": 2.1563,
+      "step": 8266500
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 3.8039545896628383e-05,
+      "loss": 2.191,
+      "step": 8267000
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 3.8038822248981106e-05,
+      "loss": 2.15,
+      "step": 8267500
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 3.803809860133383e-05,
+      "loss": 2.1866,
+      "step": 8268000
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 3.803737495368655e-05,
+      "loss": 2.164,
+      "step": 8268500
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 3.803665130603928e-05,
+      "loss": 2.1533,
+      "step": 8269000
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 3.8035927658392e-05,
+      "loss": 2.1681,
+      "step": 8269500
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 3.8035204010744724e-05,
+      "loss": 2.1631,
+      "step": 8270000
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 3.803448181039274e-05,
+      "loss": 2.1604,
+      "step": 8270500
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 3.803375816274546e-05,
+      "loss": 2.172,
+      "step": 8271000
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 3.8033034515098184e-05,
+      "loss": 2.1431,
+      "step": 8271500
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 3.803231086745091e-05,
+      "loss": 2.1359,
+      "step": 8272000
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 3.803158866709893e-05,
+      "loss": 2.1798,
+      "step": 8272500
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 3.803086501945165e-05,
+      "loss": 2.1435,
+      "step": 8273000
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 3.803014137180437e-05,
+      "loss": 2.1544,
+      "step": 8273500
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 3.8029417724157095e-05,
+      "loss": 2.1764,
+      "step": 8274000
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 3.802869407650982e-05,
+      "loss": 2.1387,
+      "step": 8274500
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 3.802797042886254e-05,
+      "loss": 2.1784,
+      "step": 8275000
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 3.802724678121526e-05,
+      "loss": 2.1562,
+      "step": 8275500
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 3.8026523133567984e-05,
+      "loss": 2.1731,
+      "step": 8276000
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 3.802579948592071e-05,
+      "loss": 2.1758,
+      "step": 8276500
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 3.8025077285568735e-05,
+      "loss": 2.1705,
+      "step": 8277000
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 3.802435363792146e-05,
+      "loss": 2.1576,
+      "step": 8277500
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 3.802362999027418e-05,
+      "loss": 2.1841,
+      "step": 8278000
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 3.80229063426269e-05,
+      "loss": 2.1704,
+      "step": 8278500
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 3.802218414227492e-05,
+      "loss": 2.1693,
+      "step": 8279000
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 3.802146049462764e-05,
+      "loss": 2.1698,
+      "step": 8279500
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 3.802073684698036e-05,
+      "loss": 2.1376,
+      "step": 8280000
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 3.802001319933309e-05,
+      "loss": 2.1496,
+      "step": 8280500
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 3.801928955168581e-05,
+      "loss": 2.1751,
+      "step": 8281000
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 3.801856735133383e-05,
+      "loss": 2.1772,
+      "step": 8281500
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 3.801784370368655e-05,
+      "loss": 2.1202,
+      "step": 8282000
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 3.801712005603927e-05,
+      "loss": 2.1796,
+      "step": 8282500
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 3.8016396408391995e-05,
+      "loss": 2.1609,
+      "step": 8283000
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 3.801567276074472e-05,
+      "loss": 2.1416,
+      "step": 8283500
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 3.801495056039274e-05,
+      "loss": 2.1584,
+      "step": 8284000
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 3.801422836004076e-05,
+      "loss": 2.1732,
+      "step": 8284500
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 3.8013504712393485e-05,
+      "loss": 2.1624,
+      "step": 8285000
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 3.801278106474621e-05,
+      "loss": 2.1438,
+      "step": 8285500
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 3.801205741709893e-05,
+      "loss": 2.1604,
+      "step": 8286000
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 3.801133376945165e-05,
+      "loss": 2.1592,
+      "step": 8286500
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 3.8010610121804374e-05,
+      "loss": 2.1515,
+      "step": 8287000
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 3.8009886474157096e-05,
+      "loss": 2.1628,
+      "step": 8287500
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 3.800916282650982e-05,
+      "loss": 2.1588,
+      "step": 8288000
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 3.800843917886254e-05,
+      "loss": 2.1775,
+      "step": 8288500
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 3.800771553121526e-05,
+      "loss": 2.1663,
+      "step": 8289000
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 3.800699188356799e-05,
+      "loss": 2.1489,
+      "step": 8289500
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 3.8006268235920714e-05,
+      "loss": 2.1821,
+      "step": 8290000
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 3.8005544588273436e-05,
+      "loss": 2.15,
+      "step": 8290500
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 3.8004820940626165e-05,
+      "loss": 2.1397,
+      "step": 8291000
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.6629079772621248,
+      "eval_accuracy_mlm": 0.6270114574392218,
+      "eval_accuracy_nsp": 0.8555338622228944,
+      "eval_loss": 2.2102925777435303,
+      "eval_runtime": 331.9145,
+      "eval_samples_per_second": 1314.754,
+      "eval_steps_per_second": 54.782,
+      "step": 8291328
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 3.800409874027418e-05,
+      "loss": 2.1363,
+      "step": 8291500
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 3.80033750926269e-05,
+      "loss": 2.1397,
+      "step": 8292000
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 3.8002651444979625e-05,
+      "loss": 2.1261,
+      "step": 8292500
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 3.800192779733235e-05,
+      "loss": 2.1153,
+      "step": 8293000
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 3.800120559698036e-05,
+      "loss": 2.1465,
+      "step": 8293500
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 3.800048194933309e-05,
+      "loss": 2.1581,
+      "step": 8294000
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 3.7999758301685814e-05,
+      "loss": 2.1546,
+      "step": 8294500
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 3.7999034654038536e-05,
+      "loss": 2.1238,
+      "step": 8295000
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 3.799831245368655e-05,
+      "loss": 2.1491,
+      "step": 8295500
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 3.7997588806039274e-05,
+      "loss": 2.1534,
+      "step": 8296000
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 3.7996865158391996e-05,
+      "loss": 2.1569,
+      "step": 8296500
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 3.799614151074472e-05,
+      "loss": 2.162,
+      "step": 8297000
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 3.799541786309744e-05,
+      "loss": 2.1316,
+      "step": 8297500
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 3.799469566274546e-05,
+      "loss": 2.1299,
+      "step": 8298000
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 3.7993972015098185e-05,
+      "loss": 2.1527,
+      "step": 8298500
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 3.799324981474621e-05,
+      "loss": 2.1406,
+      "step": 8299000
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 3.799252616709893e-05,
+      "loss": 2.1478,
+      "step": 8299500
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 3.799180251945165e-05,
+      "loss": 2.1649,
+      "step": 8300000
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 3.7991078871804374e-05,
+      "loss": 2.127,
+      "step": 8300500
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 3.7990355224157097e-05,
+      "loss": 2.1331,
+      "step": 8301000
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 3.798963157650982e-05,
+      "loss": 2.1607,
+      "step": 8301500
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 3.798890792886254e-05,
+      "loss": 2.1315,
+      "step": 8302000
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 3.798818428121526e-05,
+      "loss": 2.1581,
+      "step": 8302500
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 3.798746063356799e-05,
+      "loss": 2.1367,
+      "step": 8303000
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 3.798673843321601e-05,
+      "loss": 2.1365,
+      "step": 8303500
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 3.798601478556873e-05,
+      "loss": 2.1485,
+      "step": 8304000
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 3.798529113792145e-05,
+      "loss": 2.1408,
+      "step": 8304500
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 3.7984567490274175e-05,
+      "loss": 2.1714,
+      "step": 8305000
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 3.79838438426269e-05,
+      "loss": 2.1472,
+      "step": 8305500
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 3.798312019497962e-05,
+      "loss": 2.1618,
+      "step": 8306000
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 3.798239654733235e-05,
+      "loss": 2.1577,
+      "step": 8306500
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 3.798167289968507e-05,
+      "loss": 2.1283,
+      "step": 8307000
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 3.798094925203779e-05,
+      "loss": 2.1504,
+      "step": 8307500
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 3.798022849898111e-05,
+      "loss": 2.1256,
+      "step": 8308000
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 3.7979506298629124e-05,
+      "loss": 2.1087,
+      "step": 8308500
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 3.7978782650981846e-05,
+      "loss": 2.1414,
+      "step": 8309000
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 3.797805900333457e-05,
+      "loss": 2.1258,
+      "step": 8309500
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 3.797733535568729e-05,
+      "loss": 2.1683,
+      "step": 8310000
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 3.797661170804002e-05,
+      "loss": 2.1434,
+      "step": 8310500
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 3.797588806039274e-05,
+      "loss": 2.1418,
+      "step": 8311000
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 3.7975164412745464e-05,
+      "loss": 2.1152,
+      "step": 8311500
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 3.7974440765098186e-05,
+      "loss": 2.1548,
+      "step": 8312000
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 3.79737185647462e-05,
+      "loss": 2.1485,
+      "step": 8312500
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 3.7972994917098924e-05,
+      "loss": 2.141,
+      "step": 8313000
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 3.7972271269451646e-05,
+      "loss": 2.1313,
+      "step": 8313500
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 3.7971547621804375e-05,
+      "loss": 2.1218,
+      "step": 8314000
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 3.797082542145239e-05,
+      "loss": 2.156,
+      "step": 8314500
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 3.797010177380512e-05,
+      "loss": 2.1518,
+      "step": 8315000
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 3.796937812615784e-05,
+      "loss": 2.1754,
+      "step": 8315500
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 3.7968654478510564e-05,
+      "loss": 2.162,
+      "step": 8316000
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 3.7967930830863286e-05,
+      "loss": 2.1538,
+      "step": 8316500
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 3.796720718321601e-05,
+      "loss": 2.133,
+      "step": 8317000
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 3.7966484982864024e-05,
+      "loss": 2.1546,
+      "step": 8317500
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 3.7965761335216746e-05,
+      "loss": 2.1505,
+      "step": 8318000
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 3.796503768756947e-05,
+      "loss": 2.1261,
+      "step": 8318500
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 3.796431403992219e-05,
+      "loss": 2.1514,
+      "step": 8319000
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 3.796359039227492e-05,
+      "loss": 2.1407,
+      "step": 8319500
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 3.7962868191922935e-05,
+      "loss": 2.1468,
+      "step": 8320000
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 3.796214454427566e-05,
+      "loss": 2.1241,
+      "step": 8320500
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 3.796142089662838e-05,
+      "loss": 2.1218,
+      "step": 8321000
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 3.796069724898111e-05,
+      "loss": 2.1301,
+      "step": 8321500
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 3.795997360133383e-05,
+      "loss": 2.1362,
+      "step": 8322000
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 3.795925140098185e-05,
+      "loss": 2.161,
+      "step": 8322500
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 3.795852775333457e-05,
+      "loss": 2.154,
+      "step": 8323000
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 3.795780555298259e-05,
+      "loss": 2.1502,
+      "step": 8323500
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 3.7957081905335313e-05,
+      "loss": 2.1301,
+      "step": 8324000
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 3.7956358257688036e-05,
+      "loss": 2.1366,
+      "step": 8324500
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 3.795563461004076e-05,
+      "loss": 2.1506,
+      "step": 8325000
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 3.795491096239348e-05,
+      "loss": 2.1507,
+      "step": 8325500
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 3.79541873147462e-05,
+      "loss": 2.1491,
+      "step": 8326000
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 3.7953463667098925e-05,
+      "loss": 2.1503,
+      "step": 8326500
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 3.795274001945165e-05,
+      "loss": 2.1587,
+      "step": 8327000
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 3.795201637180437e-05,
+      "loss": 2.1578,
+      "step": 8327500
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 3.795129417145239e-05,
+      "loss": 2.1488,
+      "step": 8328000
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 3.7950570523805114e-05,
+      "loss": 2.1581,
+      "step": 8328500
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 3.794984687615784e-05,
+      "loss": 2.1502,
+      "step": 8329000
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 3.7949123228510565e-05,
+      "loss": 2.1713,
+      "step": 8329500
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 3.794840102815858e-05,
+      "loss": 2.1542,
+      "step": 8330000
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 3.79476773805113e-05,
+      "loss": 2.138,
+      "step": 8330500
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 3.7946953732864025e-05,
+      "loss": 2.1514,
+      "step": 8331000
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 3.794623008521675e-05,
+      "loss": 2.1401,
+      "step": 8331500
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 3.794550643756947e-05,
+      "loss": 2.1389,
+      "step": 8332000
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 3.79447827899222e-05,
+      "loss": 2.1453,
+      "step": 8332500
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 3.794405914227492e-05,
+      "loss": 2.1255,
+      "step": 8333000
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 3.794333549462764e-05,
+      "loss": 2.1395,
+      "step": 8333500
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 3.7942611846980365e-05,
+      "loss": 2.1536,
+      "step": 8334000
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 3.794188819933309e-05,
+      "loss": 2.1509,
+      "step": 8334500
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 3.7941167446276396e-05,
+      "loss": 2.145,
+      "step": 8335000
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 3.794044379862912e-05,
+      "loss": 2.1287,
+      "step": 8335500
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 3.793972159827714e-05,
+      "loss": 2.1377,
+      "step": 8336000
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 3.793899795062986e-05,
+      "loss": 2.1452,
+      "step": 8336500
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 3.793827430298259e-05,
+      "loss": 2.1314,
+      "step": 8337000
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 3.7937550655335314e-05,
+      "loss": 2.1252,
+      "step": 8337500
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 3.7936827007688037e-05,
+      "loss": 2.18,
+      "step": 8338000
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 3.793610336004076e-05,
+      "loss": 2.145,
+      "step": 8338500
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 3.793537971239348e-05,
+      "loss": 2.1529,
+      "step": 8339000
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 3.79346560647462e-05,
+      "loss": 2.1396,
+      "step": 8339500
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 3.7933932417098925e-05,
+      "loss": 2.1374,
+      "step": 8340000
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 3.793320876945165e-05,
+      "loss": 2.1499,
+      "step": 8340500
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 3.793248656909967e-05,
+      "loss": 2.1372,
+      "step": 8341000
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 3.793176292145239e-05,
+      "loss": 2.1557,
+      "step": 8341500
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 3.7931039273805114e-05,
+      "loss": 2.1511,
+      "step": 8342000
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 3.793031562615784e-05,
+      "loss": 2.1651,
+      "step": 8342500
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 3.792959342580585e-05,
+      "loss": 2.1475,
+      "step": 8343000
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 3.7928869778158575e-05,
+      "loss": 2.1299,
+      "step": 8343500
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 3.7928146130511304e-05,
+      "loss": 2.1563,
+      "step": 8344000
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 3.7927422482864026e-05,
+      "loss": 2.1483,
+      "step": 8344500
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 3.792669883521675e-05,
+      "loss": 2.154,
+      "step": 8345000
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 3.792597518756947e-05,
+      "loss": 2.1513,
+      "step": 8345500
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 3.79252515399222e-05,
+      "loss": 2.131,
+      "step": 8346000
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 3.792452789227492e-05,
+      "loss": 2.136,
+      "step": 8346500
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 3.792380569192294e-05,
+      "loss": 2.1469,
+      "step": 8347000
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 3.792308204427566e-05,
+      "loss": 2.1459,
+      "step": 8347500
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 3.792235839662838e-05,
+      "loss": 2.1374,
+      "step": 8348000
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 3.7921634748981104e-05,
+      "loss": 2.1338,
+      "step": 8348500
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 3.7920911101333826e-05,
+      "loss": 2.1399,
+      "step": 8349000
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 3.792018745368655e-05,
+      "loss": 2.1486,
+      "step": 8349500
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 3.791946380603927e-05,
+      "loss": 2.1324,
+      "step": 8350000
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 3.7918740158392e-05,
+      "loss": 2.1359,
+      "step": 8350500
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 3.791801651074472e-05,
+      "loss": 2.1622,
+      "step": 8351000
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 3.7917294310392744e-05,
+      "loss": 2.1389,
+      "step": 8351500
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 3.7916570662745466e-05,
+      "loss": 2.147,
+      "step": 8352000
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 3.791584701509819e-05,
+      "loss": 2.1553,
+      "step": 8352500
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 3.791512336745091e-05,
+      "loss": 2.1577,
+      "step": 8353000
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 3.791439971980363e-05,
+      "loss": 2.151,
+      "step": 8353500
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 3.7913676072156355e-05,
+      "loss": 2.1455,
+      "step": 8354000
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 3.791295242450908e-05,
+      "loss": 2.1295,
+      "step": 8354500
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 3.79122287768618e-05,
+      "loss": 2.1686,
+      "step": 8355000
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 3.791150512921452e-05,
+      "loss": 2.1363,
+      "step": 8355500
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 3.791078148156725e-05,
+      "loss": 2.1717,
+      "step": 8356000
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 3.791005783391997e-05,
+      "loss": 2.1328,
+      "step": 8356500
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 3.790933708086328e-05,
+      "loss": 2.1363,
+      "step": 8357000
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 3.7908613433216004e-05,
+      "loss": 2.1395,
+      "step": 8357500
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 3.7907889785568726e-05,
+      "loss": 2.1482,
+      "step": 8358000
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 3.790716613792145e-05,
+      "loss": 2.1789,
+      "step": 8358500
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 3.790644249027418e-05,
+      "loss": 2.1602,
+      "step": 8359000
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 3.79057188426269e-05,
+      "loss": 2.1632,
+      "step": 8359500
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 3.790499519497963e-05,
+      "loss": 2.1464,
+      "step": 8360000
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 3.790427154733235e-05,
+      "loss": 2.144,
+      "step": 8360500
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 3.7903547899685073e-05,
+      "loss": 2.1492,
+      "step": 8361000
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 3.790282569933309e-05,
+      "loss": 2.1499,
+      "step": 8361500
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 3.790210205168581e-05,
+      "loss": 2.152,
+      "step": 8362000
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 3.7901378404038533e-05,
+      "loss": 2.1433,
+      "step": 8362500
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 3.7900654756391256e-05,
+      "loss": 2.1359,
+      "step": 8363000
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 3.789993110874398e-05,
+      "loss": 2.1379,
+      "step": 8363500
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 3.7899208908392e-05,
+      "loss": 2.1399,
+      "step": 8364000
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 3.789848526074472e-05,
+      "loss": 2.1406,
+      "step": 8364500
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 3.7897761613097445e-05,
+      "loss": 2.1426,
+      "step": 8365000
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 3.7897040860040754e-05,
+      "loss": 2.125,
+      "step": 8365500
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 3.7896317212393476e-05,
+      "loss": 2.1415,
+      "step": 8366000
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 3.7895593564746205e-05,
+      "loss": 2.143,
+      "step": 8366500
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 3.789486991709893e-05,
+      "loss": 2.1408,
+      "step": 8367000
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 3.789414626945165e-05,
+      "loss": 2.1394,
+      "step": 8367500
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 3.789342262180438e-05,
+      "loss": 2.1476,
+      "step": 8368000
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 3.78926989741571e-05,
+      "loss": 2.1441,
+      "step": 8368500
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 3.789197532650982e-05,
+      "loss": 2.153,
+      "step": 8369000
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 3.7891251678862545e-05,
+      "loss": 2.1368,
+      "step": 8369500
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 3.789052803121527e-05,
+      "loss": 2.1512,
+      "step": 8370000
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 3.788980438356799e-05,
+      "loss": 2.1435,
+      "step": 8370500
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 3.788908073592071e-05,
+      "loss": 2.1428,
+      "step": 8371000
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 3.7888357088273434e-05,
+      "loss": 2.1456,
+      "step": 8371500
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 3.7887633440626156e-05,
+      "loss": 2.1354,
+      "step": 8372000
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 3.788690979297888e-05,
+      "loss": 2.1351,
+      "step": 8372500
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 3.78861861453316e-05,
+      "loss": 2.1369,
+      "step": 8373000
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 3.788546249768433e-05,
+      "loss": 2.1479,
+      "step": 8373500
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 3.788474029733235e-05,
+      "loss": 2.1492,
+      "step": 8374000
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 3.7884016649685074e-05,
+      "loss": 2.1351,
+      "step": 8374500
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 3.7883293002037796e-05,
+      "loss": 2.1281,
+      "step": 8375000
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 3.788256935439052e-05,
+      "loss": 2.1383,
+      "step": 8375500
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 3.788184570674324e-05,
+      "loss": 2.1312,
+      "step": 8376000
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 3.7881123506391256e-05,
+      "loss": 2.1453,
+      "step": 8376500
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 3.788039985874398e-05,
+      "loss": 2.1364,
+      "step": 8377000
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 3.78796762110967e-05,
+      "loss": 2.1765,
+      "step": 8377500
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 3.787895256344943e-05,
+      "loss": 2.1426,
+      "step": 8378000
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 3.7878230363097445e-05,
+      "loss": 2.1264,
+      "step": 8378500
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 3.787750816274546e-05,
+      "loss": 2.1494,
+      "step": 8379000
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 3.7876785962393477e-05,
+      "loss": 2.145,
+      "step": 8379500
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 3.7876062314746206e-05,
+      "loss": 2.13,
+      "step": 8380000
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 3.787533866709893e-05,
+      "loss": 2.1509,
+      "step": 8380500
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 3.787461501945165e-05,
+      "loss": 2.1585,
+      "step": 8381000
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 3.787389137180438e-05,
+      "loss": 2.146,
+      "step": 8381500
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 3.78731677241571e-05,
+      "loss": 2.1405,
+      "step": 8382000
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 3.787244552380512e-05,
+      "loss": 2.1364,
+      "step": 8382500
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 3.787172187615784e-05,
+      "loss": 2.1505,
+      "step": 8383000
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 3.787099822851056e-05,
+      "loss": 2.1586,
+      "step": 8383500
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 3.7870274580863284e-05,
+      "loss": 2.1387,
+      "step": 8384000
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 3.7869550933216006e-05,
+      "loss": 2.1475,
+      "step": 8384500
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 3.786882728556873e-05,
+      "loss": 2.1466,
+      "step": 8385000
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 3.786810363792146e-05,
+      "loss": 2.1278,
+      "step": 8385500
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 3.786737999027418e-05,
+      "loss": 2.1429,
+      "step": 8386000
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 3.78666563426269e-05,
+      "loss": 2.1506,
+      "step": 8386500
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 3.7865932694979624e-05,
+      "loss": 2.1569,
+      "step": 8387000
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 3.7865209047332346e-05,
+      "loss": 2.1404,
+      "step": 8387500
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 3.786448539968507e-05,
+      "loss": 2.1653,
+      "step": 8388000
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 3.786376464662838e-05,
+      "loss": 2.1593,
+      "step": 8388500
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 3.7863040998981106e-05,
+      "loss": 2.1451,
+      "step": 8389000
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 3.786231735133383e-05,
+      "loss": 2.1534,
+      "step": 8389500
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 3.786159370368656e-05,
+      "loss": 2.1465,
+      "step": 8390000
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 3.786087005603928e-05,
+      "loss": 2.1331,
+      "step": 8390500
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 3.7860146408392e-05,
+      "loss": 2.1543,
+      "step": 8391000
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 3.7859422760744724e-05,
+      "loss": 2.1694,
+      "step": 8391500
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 3.7858699113097446e-05,
+      "loss": 2.1353,
+      "step": 8392000
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 3.785797546545017e-05,
+      "loss": 2.1492,
+      "step": 8392500
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 3.785725181780289e-05,
+      "loss": 2.1494,
+      "step": 8393000
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 3.785652817015561e-05,
+      "loss": 2.1425,
+      "step": 8393500
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 3.7855804522508335e-05,
+      "loss": 2.1467,
+      "step": 8394000
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 3.785508232215636e-05,
+      "loss": 2.1427,
+      "step": 8394500
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 3.785435867450908e-05,
+      "loss": 2.1347,
+      "step": 8395000
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 3.78536350268618e-05,
+      "loss": 2.1628,
+      "step": 8395500
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 3.785291282650982e-05,
+      "loss": 2.1373,
+      "step": 8396000
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 3.7852189178862547e-05,
+      "loss": 2.1618,
+      "step": 8396500
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 3.785146697851056e-05,
+      "loss": 2.1584,
+      "step": 8397000
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 3.7850743330863284e-05,
+      "loss": 2.1398,
+      "step": 8397500
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 3.7850019683216007e-05,
+      "loss": 2.1548,
+      "step": 8398000
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 3.784929603556873e-05,
+      "loss": 2.1873,
+      "step": 8398500
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 3.784857238792146e-05,
+      "loss": 2.141,
+      "step": 8399000
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 3.784784874027418e-05,
+      "loss": 2.1438,
+      "step": 8399500
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 3.78471250926269e-05,
+      "loss": 2.156,
+      "step": 8400000
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 3.7846401444979625e-05,
+      "loss": 2.1435,
+      "step": 8400500
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 3.784567779733235e-05,
+      "loss": 2.1652,
+      "step": 8401000
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 3.784495414968507e-05,
+      "loss": 2.1485,
+      "step": 8401500
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 3.7844231949333085e-05,
+      "loss": 2.1684,
+      "step": 8402000
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 3.784350830168581e-05,
+      "loss": 2.1509,
+      "step": 8402500
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 3.784278465403853e-05,
+      "loss": 2.141,
+      "step": 8403000
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 3.784206100639126e-05,
+      "loss": 2.1532,
+      "step": 8403500
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 3.784133735874398e-05,
+      "loss": 2.1508,
+      "step": 8404000
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 3.784061371109671e-05,
+      "loss": 2.1503,
+      "step": 8404500
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 3.783989006344943e-05,
+      "loss": 2.1654,
+      "step": 8405000
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 3.7839166415802154e-05,
+      "loss": 2.1598,
+      "step": 8405500
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 3.7838442768154876e-05,
+      "loss": 2.1418,
+      "step": 8406000
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 3.78377191205076e-05,
+      "loss": 2.1587,
+      "step": 8406500
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 3.783699547286032e-05,
+      "loss": 2.1497,
+      "step": 8407000
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 3.7836273272508336e-05,
+      "loss": 2.1368,
+      "step": 8407500
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 3.783554962486106e-05,
+      "loss": 2.1602,
+      "step": 8408000
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 3.783482742450908e-05,
+      "loss": 2.1556,
+      "step": 8408500
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 3.78341037768618e-05,
+      "loss": 2.1496,
+      "step": 8409000
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 3.783338157650982e-05,
+      "loss": 2.1522,
+      "step": 8409500
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 3.783265792886254e-05,
+      "loss": 2.1293,
+      "step": 8410000
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 3.783193428121526e-05,
+      "loss": 2.1526,
+      "step": 8410500
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 3.7831210633567985e-05,
+      "loss": 2.1673,
+      "step": 8411000
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 3.7830486985920714e-05,
+      "loss": 2.1437,
+      "step": 8411500
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 3.7829763338273436e-05,
+      "loss": 2.15,
+      "step": 8412000
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 3.782903969062616e-05,
+      "loss": 2.1523,
+      "step": 8412500
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 3.782831749027418e-05,
+      "loss": 2.1173,
+      "step": 8413000
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 3.78275938426269e-05,
+      "loss": 2.1669,
+      "step": 8413500
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 3.782687164227492e-05,
+      "loss": 2.1469,
+      "step": 8414000
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 3.782614799462764e-05,
+      "loss": 2.1393,
+      "step": 8414500
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 3.782542434698036e-05,
+      "loss": 2.1285,
+      "step": 8415000
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 3.7824700699333085e-05,
+      "loss": 2.1708,
+      "step": 8415500
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 3.782397705168581e-05,
+      "loss": 2.1573,
+      "step": 8416000
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 3.782325485133383e-05,
+      "loss": 2.1815,
+      "step": 8416500
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 3.782253120368655e-05,
+      "loss": 2.1518,
+      "step": 8417000
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 3.7821807556039274e-05,
+      "loss": 2.1402,
+      "step": 8417500
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 3.7821083908392e-05,
+      "loss": 2.1504,
+      "step": 8418000
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 3.782036026074472e-05,
+      "loss": 2.1609,
+      "step": 8418500
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 3.781963661309745e-05,
+      "loss": 2.1662,
+      "step": 8419000
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 3.781891296545017e-05,
+      "loss": 2.1435,
+      "step": 8419500
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 3.781818931780289e-05,
+      "loss": 2.1426,
+      "step": 8420000
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 3.7817465670155615e-05,
+      "loss": 2.1394,
+      "step": 8420500
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 3.781674202250834e-05,
+      "loss": 2.1541,
+      "step": 8421000
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 3.781601837486106e-05,
+      "loss": 2.1589,
+      "step": 8421500
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 3.781529472721379e-05,
+      "loss": 2.1702,
+      "step": 8422000
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 3.781457107956651e-05,
+      "loss": 2.1515,
+      "step": 8422500
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 3.781384743191923e-05,
+      "loss": 2.1466,
+      "step": 8423000
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 3.7813123784271955e-05,
+      "loss": 2.1413,
+      "step": 8423500
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 3.781240013662468e-05,
+      "loss": 2.1335,
+      "step": 8424000
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 3.78116764889774e-05,
+      "loss": 2.1279,
+      "step": 8424500
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 3.781095284133012e-05,
+      "loss": 2.1441,
+      "step": 8425000
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 3.7810229193682844e-05,
+      "loss": 2.1484,
+      "step": 8425500
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 3.7809506993330866e-05,
+      "loss": 2.152,
+      "step": 8426000
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 3.780878334568359e-05,
+      "loss": 2.1665,
+      "step": 8426500
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 3.780805969803631e-05,
+      "loss": 2.1696,
+      "step": 8427000
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 3.780733749768433e-05,
+      "loss": 2.1409,
+      "step": 8427500
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 3.780661529733235e-05,
+      "loss": 2.1578,
+      "step": 8428000
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 3.780589164968507e-05,
+      "loss": 2.163,
+      "step": 8428500
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 3.780516800203779e-05,
+      "loss": 2.1503,
+      "step": 8429000
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 3.7804444354390515e-05,
+      "loss": 2.1647,
+      "step": 8429500
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 3.780372070674324e-05,
+      "loss": 2.1648,
+      "step": 8430000
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 3.780299705909596e-05,
+      "loss": 2.1327,
+      "step": 8430500
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 3.780227341144869e-05,
+      "loss": 2.1431,
+      "step": 8431000
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 3.780154976380141e-05,
+      "loss": 2.1654,
+      "step": 8431500
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 3.780082611615413e-05,
+      "loss": 2.153,
+      "step": 8432000
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 3.7800102468506855e-05,
+      "loss": 2.1527,
+      "step": 8432500
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 3.779937882085958e-05,
+      "loss": 2.147,
+      "step": 8433000
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 3.77986566205076e-05,
+      "loss": 2.1484,
+      "step": 8433500
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 3.779793297286032e-05,
+      "loss": 2.1595,
+      "step": 8434000
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 3.7797209325213044e-05,
+      "loss": 2.1551,
+      "step": 8434500
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 3.7796485677565767e-05,
+      "loss": 2.1628,
+      "step": 8435000
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 3.779576202991849e-05,
+      "loss": 2.1486,
+      "step": 8435500
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 3.779503838227121e-05,
+      "loss": 2.1471,
+      "step": 8436000
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 3.779431473462394e-05,
+      "loss": 2.1442,
+      "step": 8436500
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 3.779359108697666e-05,
+      "loss": 2.1657,
+      "step": 8437000
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 3.7792867439329384e-05,
+      "loss": 2.1852,
+      "step": 8437500
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 3.779214379168211e-05,
+      "loss": 2.1666,
+      "step": 8438000
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 3.779142014403483e-05,
+      "loss": 2.1416,
+      "step": 8438500
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 3.779069649638755e-05,
+      "loss": 2.157,
+      "step": 8439000
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 3.7789972848740273e-05,
+      "loss": 2.1641,
+      "step": 8439500
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 3.778925209568359e-05,
+      "loss": 2.1303,
+      "step": 8440000
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 3.778852844803631e-05,
+      "loss": 2.1406,
+      "step": 8440500
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 3.778780480038904e-05,
+      "loss": 2.1593,
+      "step": 8441000
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 3.778708115274176e-05,
+      "loss": 2.1471,
+      "step": 8441500
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 3.7786357505094485e-05,
+      "loss": 2.1387,
+      "step": 8442000
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 3.778563385744721e-05,
+      "loss": 2.1386,
+      "step": 8442500
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 3.778491165709522e-05,
+      "loss": 2.148,
+      "step": 8443000
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 3.7784188009447945e-05,
+      "loss": 2.1408,
+      "step": 8443500
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 3.778346436180067e-05,
+      "loss": 2.1449,
+      "step": 8444000
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 3.778274071415339e-05,
+      "loss": 2.1464,
+      "step": 8444500
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 3.778201851380141e-05,
+      "loss": 2.1662,
+      "step": 8445000
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 3.7781294866154134e-05,
+      "loss": 2.1528,
+      "step": 8445500
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 3.7780571218506856e-05,
+      "loss": 2.1194,
+      "step": 8446000
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 3.777984901815487e-05,
+      "loss": 2.1515,
+      "step": 8446500
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 3.7779125370507594e-05,
+      "loss": 2.1565,
+      "step": 8447000
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 3.7778401722860316e-05,
+      "loss": 2.1509,
+      "step": 8447500
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 3.777767807521304e-05,
+      "loss": 2.157,
+      "step": 8448000
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 3.777695442756577e-05,
+      "loss": 2.1513,
+      "step": 8448500
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 3.777623077991849e-05,
+      "loss": 2.1479,
+      "step": 8449000
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 3.777550713227121e-05,
+      "loss": 2.159,
+      "step": 8449500
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 3.777478348462394e-05,
+      "loss": 2.1536,
+      "step": 8450000
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 3.7774061284271956e-05,
+      "loss": 2.1424,
+      "step": 8450500
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 3.777333763662468e-05,
+      "loss": 2.1505,
+      "step": 8451000
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 3.77726139889774e-05,
+      "loss": 2.1396,
+      "step": 8451500
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 3.7771891788625416e-05,
+      "loss": 2.1605,
+      "step": 8452000
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 3.777116814097814e-05,
+      "loss": 2.1354,
+      "step": 8452500
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 3.777044449333087e-05,
+      "loss": 2.1648,
+      "step": 8453000
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 3.776972084568359e-05,
+      "loss": 2.1515,
+      "step": 8453500
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 3.776899719803631e-05,
+      "loss": 2.1708,
+      "step": 8454000
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 3.7768273550389034e-05,
+      "loss": 2.1387,
+      "step": 8454500
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 3.7767549902741757e-05,
+      "loss": 2.1645,
+      "step": 8455000
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 3.7766826255094486e-05,
+      "loss": 2.1645,
+      "step": 8455500
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 3.776610260744721e-05,
+      "loss": 2.1446,
+      "step": 8456000
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 3.776537895979993e-05,
+      "loss": 2.1566,
+      "step": 8456500
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 3.776465531215265e-05,
+      "loss": 2.1508,
+      "step": 8457000
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 3.7763931664505374e-05,
+      "loss": 2.1704,
+      "step": 8457500
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 3.77632080168581e-05,
+      "loss": 2.1594,
+      "step": 8458000
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 3.776248436921082e-05,
+      "loss": 2.1443,
+      "step": 8458500
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 3.776176072156354e-05,
+      "loss": 2.1675,
+      "step": 8459000
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 3.7761037073916263e-05,
+      "loss": 2.159,
+      "step": 8459500
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 3.776031342626899e-05,
+      "loss": 2.1595,
+      "step": 8460000
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 3.7759589778621715e-05,
+      "loss": 2.1358,
+      "step": 8460500
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 3.775886613097444e-05,
+      "loss": 2.1686,
+      "step": 8461000
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 3.775814248332716e-05,
+      "loss": 2.1126,
+      "step": 8461500
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 3.775741883567988e-05,
+      "loss": 2.126,
+      "step": 8462000
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 3.775669808262319e-05,
+      "loss": 2.1548,
+      "step": 8462500
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 3.775597443497592e-05,
+      "loss": 2.1439,
+      "step": 8463000
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 3.775525078732864e-05,
+      "loss": 2.1254,
+      "step": 8463500
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 3.775452713968137e-05,
+      "loss": 2.1541,
+      "step": 8464000
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 3.775380349203409e-05,
+      "loss": 2.1311,
+      "step": 8464500
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 3.7753079844386815e-05,
+      "loss": 2.1514,
+      "step": 8465000
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 3.775235764403483e-05,
+      "loss": 2.1432,
+      "step": 8465500
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 3.775163399638755e-05,
+      "loss": 2.1862,
+      "step": 8466000
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 3.7750910348740275e-05,
+      "loss": 2.1487,
+      "step": 8466500
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 3.7750186701093e-05,
+      "loss": 2.1608,
+      "step": 8467000
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 3.774946305344572e-05,
+      "loss": 2.1695,
+      "step": 8467500
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 3.774874085309374e-05,
+      "loss": 2.168,
+      "step": 8468000
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 3.7748017205446464e-05,
+      "loss": 2.1523,
+      "step": 8468500
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 3.774729500509448e-05,
+      "loss": 2.14,
+      "step": 8469000
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 3.77465713574472e-05,
+      "loss": 2.1614,
+      "step": 8469500
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 3.7745847709799924e-05,
+      "loss": 2.1773,
+      "step": 8470000
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 3.774512406215265e-05,
+      "loss": 2.165,
+      "step": 8470500
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 3.774440186180067e-05,
+      "loss": 2.1444,
+      "step": 8471000
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 3.774367821415339e-05,
+      "loss": 2.1514,
+      "step": 8471500
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 3.774295456650612e-05,
+      "loss": 2.1656,
+      "step": 8472000
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 3.774223091885884e-05,
+      "loss": 2.1391,
+      "step": 8472500
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 3.774150871850686e-05,
+      "loss": 2.1551,
+      "step": 8473000
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 3.774078507085958e-05,
+      "loss": 2.1434,
+      "step": 8473500
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 3.77400614232123e-05,
+      "loss": 2.1786,
+      "step": 8474000
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 3.7739337775565024e-05,
+      "loss": 2.1434,
+      "step": 8474500
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 3.7738614127917747e-05,
+      "loss": 2.1644,
+      "step": 8475000
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 3.773789048027047e-05,
+      "loss": 2.1711,
+      "step": 8475500
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 3.773716683262319e-05,
+      "loss": 2.1603,
+      "step": 8476000
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 3.773644463227121e-05,
+      "loss": 2.1337,
+      "step": 8476500
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 3.7735720984623936e-05,
+      "loss": 2.1638,
+      "step": 8477000
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 3.773499733697666e-05,
+      "loss": 2.1511,
+      "step": 8477500
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 3.773427368932939e-05,
+      "loss": 2.1592,
+      "step": 8478000
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 3.773355004168211e-05,
+      "loss": 2.156,
+      "step": 8478500
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 3.773282639403483e-05,
+      "loss": 2.1665,
+      "step": 8479000
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 3.7732102746387554e-05,
+      "loss": 2.1623,
+      "step": 8479500
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 3.773138054603557e-05,
+      "loss": 2.1508,
+      "step": 8480000
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 3.773065689838829e-05,
+      "loss": 2.1446,
+      "step": 8480500
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 3.772993325074102e-05,
+      "loss": 2.1516,
+      "step": 8481000
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 3.772920960309374e-05,
+      "loss": 2.174,
+      "step": 8481500
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 3.772848740274176e-05,
+      "loss": 2.1404,
+      "step": 8482000
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 3.772776375509448e-05,
+      "loss": 2.1727,
+      "step": 8482500
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 3.77270401074472e-05,
+      "loss": 2.1517,
+      "step": 8483000
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 3.7726316459799925e-05,
+      "loss": 2.1576,
+      "step": 8483500
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 3.772559281215265e-05,
+      "loss": 2.1743,
+      "step": 8484000
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 3.772486916450537e-05,
+      "loss": 2.1554,
+      "step": 8484500
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 3.772414696415339e-05,
+      "loss": 2.1434,
+      "step": 8485000
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 3.772342331650612e-05,
+      "loss": 2.1564,
+      "step": 8485500
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 3.772269966885884e-05,
+      "loss": 2.1354,
+      "step": 8486000
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 3.772197746850686e-05,
+      "loss": 2.1493,
+      "step": 8486500
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 3.772125382085958e-05,
+      "loss": 2.1307,
+      "step": 8487000
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 3.77205301732123e-05,
+      "loss": 2.1443,
+      "step": 8487500
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 3.7719806525565025e-05,
+      "loss": 2.142,
+      "step": 8488000
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 3.771908287791775e-05,
+      "loss": 2.1597,
+      "step": 8488500
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 3.771835923027047e-05,
+      "loss": 2.1447,
+      "step": 8489000
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 3.77176355826232e-05,
+      "loss": 2.1432,
+      "step": 8489500
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 3.771691193497592e-05,
+      "loss": 2.1555,
+      "step": 8490000
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 3.771618828732864e-05,
+      "loss": 2.1501,
+      "step": 8490500
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 3.7715464639681365e-05,
+      "loss": 2.1475,
+      "step": 8491000
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 3.771474099203409e-05,
+      "loss": 2.1381,
+      "step": 8491500
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 3.771401734438681e-05,
+      "loss": 2.1692,
+      "step": 8492000
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 3.771329369673953e-05,
+      "loss": 2.1397,
+      "step": 8492500
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 3.7712571496387554e-05,
+      "loss": 2.1638,
+      "step": 8493000
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 3.7711847848740277e-05,
+      "loss": 2.1483,
+      "step": 8493500
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 3.7711124201093e-05,
+      "loss": 2.1694,
+      "step": 8494000
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 3.771040055344572e-05,
+      "loss": 2.1654,
+      "step": 8494500
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 3.770967835309374e-05,
+      "loss": 2.151,
+      "step": 8495000
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 3.7708954705446466e-05,
+      "loss": 2.1478,
+      "step": 8495500
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 3.770823105779919e-05,
+      "loss": 2.1552,
+      "step": 8496000
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 3.770750741015191e-05,
+      "loss": 2.1453,
+      "step": 8496500
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 3.770678376250463e-05,
+      "loss": 2.1508,
+      "step": 8497000
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 3.770606156215265e-05,
+      "loss": 2.1658,
+      "step": 8497500
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 3.770533791450537e-05,
+      "loss": 2.1358,
+      "step": 8498000
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 3.770461571415339e-05,
+      "loss": 2.1402,
+      "step": 8498500
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 3.770389351380141e-05,
+      "loss": 2.1474,
+      "step": 8499000
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 3.770316986615413e-05,
+      "loss": 2.1489,
+      "step": 8499500
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 3.770244621850685e-05,
+      "loss": 2.1612,
+      "step": 8500000
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 3.7701722570859575e-05,
+      "loss": 2.1618,
+      "step": 8500500
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 3.7700998923212304e-05,
+      "loss": 2.1287,
+      "step": 8501000
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 3.7700275275565026e-05,
+      "loss": 2.1393,
+      "step": 8501500
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 3.769955162791775e-05,
+      "loss": 2.1764,
+      "step": 8502000
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 3.769882798027047e-05,
+      "loss": 2.1439,
+      "step": 8502500
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 3.76981043326232e-05,
+      "loss": 2.1417,
+      "step": 8503000
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 3.7697382132271215e-05,
+      "loss": 2.1384,
+      "step": 8503500
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 3.769665848462394e-05,
+      "loss": 2.1684,
+      "step": 8504000
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 3.769593483697666e-05,
+      "loss": 2.1463,
+      "step": 8504500
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 3.769521118932938e-05,
+      "loss": 2.152,
+      "step": 8505000
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 3.76944889889774e-05,
+      "loss": 2.1421,
+      "step": 8505500
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 3.769376534133012e-05,
+      "loss": 2.153,
+      "step": 8506000
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 3.769304169368285e-05,
+      "loss": 2.168,
+      "step": 8506500
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 3.769231804603557e-05,
+      "loss": 2.1493,
+      "step": 8507000
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 3.769159439838829e-05,
+      "loss": 2.1357,
+      "step": 8507500
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 3.769087075074102e-05,
+      "loss": 2.1444,
+      "step": 8508000
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 3.7690147103093744e-05,
+      "loss": 2.1287,
+      "step": 8508500
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 3.7689423455446466e-05,
+      "loss": 2.1488,
+      "step": 8509000
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 3.768869980779919e-05,
+      "loss": 2.1404,
+      "step": 8509500
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 3.768797616015191e-05,
+      "loss": 2.146,
+      "step": 8510000
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 3.768725251250463e-05,
+      "loss": 2.147,
+      "step": 8510500
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 3.768653031215265e-05,
+      "loss": 2.1665,
+      "step": 8511000
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 3.768580666450537e-05,
+      "loss": 2.1381,
+      "step": 8511500
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 3.76850830168581e-05,
+      "loss": 2.1553,
+      "step": 8512000
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 3.768435936921082e-05,
+      "loss": 2.1499,
+      "step": 8512500
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 3.7683635721563544e-05,
+      "loss": 2.1427,
+      "step": 8513000
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 3.7682912073916267e-05,
+      "loss": 2.1481,
+      "step": 8513500
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 3.768218842626899e-05,
+      "loss": 2.1519,
+      "step": 8514000
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 3.768146477862171e-05,
+      "loss": 2.1449,
+      "step": 8514500
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 3.768074113097443e-05,
+      "loss": 2.1476,
+      "step": 8515000
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 3.7680018930622456e-05,
+      "loss": 2.1553,
+      "step": 8515500
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 3.767929528297518e-05,
+      "loss": 2.1688,
+      "step": 8516000
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 3.76785716353279e-05,
+      "loss": 2.1618,
+      "step": 8516500
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 3.767784798768062e-05,
+      "loss": 2.1477,
+      "step": 8517000
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 3.767712434003335e-05,
+      "loss": 2.1426,
+      "step": 8517500
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 3.767640213968137e-05,
+      "loss": 2.1434,
+      "step": 8518000
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 3.767567849203409e-05,
+      "loss": 2.1435,
+      "step": 8518500
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 3.767495484438681e-05,
+      "loss": 2.1395,
+      "step": 8519000
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 3.7674231196739534e-05,
+      "loss": 2.1539,
+      "step": 8519500
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 3.7673507549092256e-05,
+      "loss": 2.1318,
+      "step": 8520000
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 3.767278534874028e-05,
+      "loss": 2.1463,
+      "step": 8520500
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 3.7672061701093e-05,
+      "loss": 2.1514,
+      "step": 8521000
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 3.767133805344572e-05,
+      "loss": 2.135,
+      "step": 8521500
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 3.7670614405798445e-05,
+      "loss": 2.1493,
+      "step": 8522000
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 3.766989075815117e-05,
+      "loss": 2.1363,
+      "step": 8522500
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 3.766916855779919e-05,
+      "loss": 2.1642,
+      "step": 8523000
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 3.766844491015191e-05,
+      "loss": 2.1185,
+      "step": 8523500
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 3.7667721262504634e-05,
+      "loss": 2.1613,
+      "step": 8524000
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 3.7666997614857356e-05,
+      "loss": 2.1347,
+      "step": 8524500
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 3.766627541450538e-05,
+      "loss": 2.1434,
+      "step": 8525000
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 3.76655517668581e-05,
+      "loss": 2.1502,
+      "step": 8525500
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 3.766482811921082e-05,
+      "loss": 2.1379,
+      "step": 8526000
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 3.7664104471563545e-05,
+      "loss": 2.1669,
+      "step": 8526500
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 3.766338082391627e-05,
+      "loss": 2.1514,
+      "step": 8527000
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 3.766265862356428e-05,
+      "loss": 2.1491,
+      "step": 8527500
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 3.76619364232123e-05,
+      "loss": 2.1686,
+      "step": 8528000
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 3.766121277556503e-05,
+      "loss": 2.1302,
+      "step": 8528500
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 3.766048912791775e-05,
+      "loss": 2.1454,
+      "step": 8529000
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 3.765976548027047e-05,
+      "loss": 2.1506,
+      "step": 8529500
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 3.7659041832623194e-05,
+      "loss": 2.1558,
+      "step": 8530000
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 3.765831818497592e-05,
+      "loss": 2.1551,
+      "step": 8530500
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 3.7657594537328645e-05,
+      "loss": 2.1655,
+      "step": 8531000
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 3.765687088968137e-05,
+      "loss": 2.137,
+      "step": 8531500
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 3.765614724203409e-05,
+      "loss": 2.1513,
+      "step": 8532000
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 3.7655425041682105e-05,
+      "loss": 2.1449,
+      "step": 8532500
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 3.765470139403483e-05,
+      "loss": 2.1463,
+      "step": 8533000
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 3.765398064097814e-05,
+      "loss": 2.1611,
+      "step": 8533500
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 3.7653256993330866e-05,
+      "loss": 2.1431,
+      "step": 8534000
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 3.765253334568359e-05,
+      "loss": 2.1596,
+      "step": 8534500
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 3.765180969803631e-05,
+      "loss": 2.1665,
+      "step": 8535000
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 3.765108605038903e-05,
+      "loss": 2.1813,
+      "step": 8535500
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 3.7650362402741755e-05,
+      "loss": 2.1393,
+      "step": 8536000
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 3.764963875509448e-05,
+      "loss": 2.1481,
+      "step": 8536500
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 3.764891800203779e-05,
+      "loss": 2.1378,
+      "step": 8537000
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 3.7648194354390515e-05,
+      "loss": 2.1382,
+      "step": 8537500
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 3.764747070674324e-05,
+      "loss": 2.1477,
+      "step": 8538000
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 3.7646747059095966e-05,
+      "loss": 2.1378,
+      "step": 8538500
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 3.764602341144869e-05,
+      "loss": 2.1487,
+      "step": 8539000
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 3.764529976380141e-05,
+      "loss": 2.1462,
+      "step": 8539500
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 3.764457611615413e-05,
+      "loss": 2.1527,
+      "step": 8540000
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 3.7643852468506855e-05,
+      "loss": 2.1642,
+      "step": 8540500
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 3.764312882085958e-05,
+      "loss": 2.1475,
+      "step": 8541000
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 3.7642405173212306e-05,
+      "loss": 2.1504,
+      "step": 8541500
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 3.764168152556503e-05,
+      "loss": 2.1524,
+      "step": 8542000
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 3.764095787791775e-05,
+      "loss": 2.1634,
+      "step": 8542500
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 3.764023423027047e-05,
+      "loss": 2.1607,
+      "step": 8543000
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 3.7639510582623195e-05,
+      "loss": 2.1791,
+      "step": 8543500
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 3.763878693497592e-05,
+      "loss": 2.1314,
+      "step": 8544000
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 3.763806328732864e-05,
+      "loss": 2.1775,
+      "step": 8544500
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 3.763733963968136e-05,
+      "loss": 2.1709,
+      "step": 8545000
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 3.763661599203409e-05,
+      "loss": 2.1546,
+      "step": 8545500
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 3.7635893791682106e-05,
+      "loss": 2.1463,
+      "step": 8546000
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 3.763517014403483e-05,
+      "loss": 2.1397,
+      "step": 8546500
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 3.763444649638756e-05,
+      "loss": 2.1586,
+      "step": 8547000
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 3.763372284874028e-05,
+      "loss": 2.1403,
+      "step": 8547500
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 3.7632999201093e-05,
+      "loss": 2.1604,
+      "step": 8548000
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 3.7632275553445724e-05,
+      "loss": 2.1639,
+      "step": 8548500
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 3.7631551905798446e-05,
+      "loss": 2.1615,
+      "step": 8549000
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 3.763082970544646e-05,
+      "loss": 2.1237,
+      "step": 8549500
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 3.7630106057799184e-05,
+      "loss": 2.1689,
+      "step": 8550000
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 3.7629382410151906e-05,
+      "loss": 2.1586,
+      "step": 8550500
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 3.762866020979993e-05,
+      "loss": 2.1449,
+      "step": 8551000
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 3.7627938009447944e-05,
+      "loss": 2.1342,
+      "step": 8551500
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 3.7627214361800667e-05,
+      "loss": 2.1597,
+      "step": 8552000
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 3.762649071415339e-05,
+      "loss": 2.1683,
+      "step": 8552500
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 3.762576706650612e-05,
+      "loss": 2.1685,
+      "step": 8553000
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 3.762504341885884e-05,
+      "loss": 2.1642,
+      "step": 8553500
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 3.762431977121156e-05,
+      "loss": 2.1676,
+      "step": 8554000
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 3.7623596123564285e-05,
+      "loss": 2.1489,
+      "step": 8554500
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 3.762287247591701e-05,
+      "loss": 2.167,
+      "step": 8555000
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 3.762214882826973e-05,
+      "loss": 2.1369,
+      "step": 8555500
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 3.762142518062246e-05,
+      "loss": 2.133,
+      "step": 8556000
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 3.7620702980270474e-05,
+      "loss": 2.1603,
+      "step": 8556500
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 3.7619979332623196e-05,
+      "loss": 2.1641,
+      "step": 8557000
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 3.761925568497592e-05,
+      "loss": 2.1751,
+      "step": 8557500
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 3.761853203732864e-05,
+      "loss": 2.1658,
+      "step": 8558000
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 3.7617809836976656e-05,
+      "loss": 2.1656,
+      "step": 8558500
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 3.761708618932938e-05,
+      "loss": 2.1233,
+      "step": 8559000
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 3.761636254168211e-05,
+      "loss": 2.1824,
+      "step": 8559500
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 3.761563889403483e-05,
+      "loss": 2.1388,
+      "step": 8560000
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 3.761491524638756e-05,
+      "loss": 2.1483,
+      "step": 8560500
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 3.761419159874028e-05,
+      "loss": 2.1481,
+      "step": 8561000
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 3.7613467951093e-05,
+      "loss": 2.1261,
+      "step": 8561500
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 3.7612744303445725e-05,
+      "loss": 2.162,
+      "step": 8562000
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 3.761202065579845e-05,
+      "loss": 2.1458,
+      "step": 8562500
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 3.761129700815117e-05,
+      "loss": 2.1722,
+      "step": 8563000
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 3.761057336050389e-05,
+      "loss": 2.1785,
+      "step": 8563500
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 3.7609849712856614e-05,
+      "loss": 2.144,
+      "step": 8564000
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 3.7609126065209336e-05,
+      "loss": 2.143,
+      "step": 8564500
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 3.760840386485736e-05,
+      "loss": 2.1617,
+      "step": 8565000
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 3.760768021721008e-05,
+      "loss": 2.1707,
+      "step": 8565500
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 3.7606958016858096e-05,
+      "loss": 2.1646,
+      "step": 8566000
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 3.760623436921082e-05,
+      "loss": 2.1984,
+      "step": 8566500
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 3.760551072156354e-05,
+      "loss": 2.1484,
+      "step": 8567000
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 3.760478707391626e-05,
+      "loss": 2.1529,
+      "step": 8567500
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 3.760406342626899e-05,
+      "loss": 2.1374,
+      "step": 8568000
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 3.7603339778621714e-05,
+      "loss": 2.1576,
+      "step": 8568500
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 3.7602616130974436e-05,
+      "loss": 2.1527,
+      "step": 8569000
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 3.760189248332716e-05,
+      "loss": 2.1506,
+      "step": 8569500
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 3.760116883567988e-05,
+      "loss": 2.1429,
+      "step": 8570000
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 3.760044518803261e-05,
+      "loss": 2.1583,
+      "step": 8570500
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 3.7599722987680626e-05,
+      "loss": 2.1784,
+      "step": 8571000
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 3.759899934003335e-05,
+      "loss": 2.1536,
+      "step": 8571500
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 3.759827569238607e-05,
+      "loss": 2.1325,
+      "step": 8572000
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 3.759755204473879e-05,
+      "loss": 2.164,
+      "step": 8572500
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 3.7596828397091514e-05,
+      "loss": 2.1706,
+      "step": 8573000
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 3.759610474944424e-05,
+      "loss": 2.1727,
+      "step": 8573500
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 3.759538110179696e-05,
+      "loss": 2.1567,
+      "step": 8574000
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 3.759465745414968e-05,
+      "loss": 2.1439,
+      "step": 8574500
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 3.759393525379771e-05,
+      "loss": 2.1419,
+      "step": 8575000
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 3.759321160615043e-05,
+      "loss": 2.1888,
+      "step": 8575500
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 3.759248940579845e-05,
+      "loss": 2.1594,
+      "step": 8576000
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 3.7591767205446464e-05,
+      "loss": 2.1384,
+      "step": 8576500
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 3.7591043557799186e-05,
+      "loss": 2.1556,
+      "step": 8577000
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 3.759031991015191e-05,
+      "loss": 2.1577,
+      "step": 8577500
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 3.758959626250464e-05,
+      "loss": 2.1828,
+      "step": 8578000
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 3.758887406215265e-05,
+      "loss": 2.1814,
+      "step": 8578500
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 3.7588150414505375e-05,
+      "loss": 2.1376,
+      "step": 8579000
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 3.75874267668581e-05,
+      "loss": 2.1499,
+      "step": 8579500
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 3.758670311921082e-05,
+      "loss": 2.1818,
+      "step": 8580000
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 3.758597947156354e-05,
+      "loss": 2.1828,
+      "step": 8580500
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 3.7585255823916264e-05,
+      "loss": 2.1494,
+      "step": 8581000
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 3.7584532176268986e-05,
+      "loss": 2.165,
+      "step": 8581500
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 3.758380852862171e-05,
+      "loss": 2.1683,
+      "step": 8582000
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 3.758308488097444e-05,
+      "loss": 2.1592,
+      "step": 8582500
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 3.758236123332716e-05,
+      "loss": 2.1569,
+      "step": 8583000
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 3.758163903297518e-05,
+      "loss": 2.1516,
+      "step": 8583500
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 3.7580915385327904e-05,
+      "loss": 2.1631,
+      "step": 8584000
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 3.7580191737680626e-05,
+      "loss": 2.1678,
+      "step": 8584500
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 3.757946809003335e-05,
+      "loss": 2.1427,
+      "step": 8585000
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 3.757874444238607e-05,
+      "loss": 2.1494,
+      "step": 8585500
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 3.757802079473879e-05,
+      "loss": 2.1409,
+      "step": 8586000
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 3.7577297147091515e-05,
+      "loss": 2.1849,
+      "step": 8586500
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 3.757657494673954e-05,
+      "loss": 2.1366,
+      "step": 8587000
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 3.757585274638755e-05,
+      "loss": 2.1416,
+      "step": 8587500
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 3.757513054603557e-05,
+      "loss": 2.1363,
+      "step": 8588000
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 3.757440689838829e-05,
+      "loss": 2.143,
+      "step": 8588500
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 3.757368325074101e-05,
+      "loss": 2.1744,
+      "step": 8589000
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 3.7572959603093735e-05,
+      "loss": 2.1523,
+      "step": 8589500
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 3.757223595544646e-05,
+      "loss": 2.1579,
+      "step": 8590000
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 3.757151230779919e-05,
+      "loss": 2.1664,
+      "step": 8590500
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 3.757078866015191e-05,
+      "loss": 2.1494,
+      "step": 8591000
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 3.757006501250464e-05,
+      "loss": 2.1543,
+      "step": 8591500
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 3.756934136485736e-05,
+      "loss": 2.1459,
+      "step": 8592000
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 3.756861771721008e-05,
+      "loss": 2.1324,
+      "step": 8592500
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 3.7567894069562805e-05,
+      "loss": 2.1464,
+      "step": 8593000
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 3.756717042191553e-05,
+      "loss": 2.1465,
+      "step": 8593500
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 3.756644677426825e-05,
+      "loss": 2.1682,
+      "step": 8594000
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 3.756572312662097e-05,
+      "loss": 2.156,
+      "step": 8594500
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 3.7564999478973694e-05,
+      "loss": 2.1662,
+      "step": 8595000
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 3.7564275831326416e-05,
+      "loss": 2.1568,
+      "step": 8595500
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 3.756355363097444e-05,
+      "loss": 2.1578,
+      "step": 8596000
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 3.7562831430622454e-05,
+      "loss": 2.1709,
+      "step": 8596500
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 3.7562107782975176e-05,
+      "loss": 2.1561,
+      "step": 8597000
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 3.75613841353279e-05,
+      "loss": 2.1699,
+      "step": 8597500
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 3.756066048768063e-05,
+      "loss": 2.1833,
+      "step": 8598000
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 3.755993684003335e-05,
+      "loss": 2.1327,
+      "step": 8598500
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 3.755921319238607e-05,
+      "loss": 2.1584,
+      "step": 8599000
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 3.7558489544738794e-05,
+      "loss": 2.1395,
+      "step": 8599500
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 3.7557767344386816e-05,
+      "loss": 2.1771,
+      "step": 8600000
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 3.755704369673954e-05,
+      "loss": 2.1429,
+      "step": 8600500
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 3.7556321496387554e-05,
+      "loss": 2.1833,
+      "step": 8601000
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 3.755559929603557e-05,
+      "loss": 2.18,
+      "step": 8601500
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 3.755487564838829e-05,
+      "loss": 2.1448,
+      "step": 8602000
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 3.7554152000741014e-05,
+      "loss": 2.1571,
+      "step": 8602500
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 3.7553428353093736e-05,
+      "loss": 2.1701,
+      "step": 8603000
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 3.7552704705446465e-05,
+      "loss": 2.17,
+      "step": 8603500
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 3.755198250509448e-05,
+      "loss": 2.1698,
+      "step": 8604000
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 3.75512588574472e-05,
+      "loss": 2.1715,
+      "step": 8604500
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 3.7550535209799925e-05,
+      "loss": 2.1627,
+      "step": 8605000
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 3.7549811562152654e-05,
+      "loss": 2.1445,
+      "step": 8605500
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 3.7549087914505376e-05,
+      "loss": 2.1485,
+      "step": 8606000
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 3.75483642668581e-05,
+      "loss": 2.1535,
+      "step": 8606500
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 3.754764061921082e-05,
+      "loss": 2.165,
+      "step": 8607000
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 3.754691697156354e-05,
+      "loss": 2.1476,
+      "step": 8607500
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 3.7546193323916265e-05,
+      "loss": 2.159,
+      "step": 8608000
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 3.754546967626899e-05,
+      "loss": 2.1687,
+      "step": 8608500
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 3.754474602862172e-05,
+      "loss": 2.1579,
+      "step": 8609000
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 3.754402238097444e-05,
+      "loss": 2.1337,
+      "step": 8609500
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 3.754329873332716e-05,
+      "loss": 2.1646,
+      "step": 8610000
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 3.754257508567988e-05,
+      "loss": 2.175,
+      "step": 8610500
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 3.7541851438032606e-05,
+      "loss": 2.1481,
+      "step": 8611000
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 3.754112779038533e-05,
+      "loss": 2.1402,
+      "step": 8611500
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 3.754040414273805e-05,
+      "loss": 2.1396,
+      "step": 8612000
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 3.7539683389681366e-05,
+      "loss": 2.1379,
+      "step": 8612500
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 3.753895974203409e-05,
+      "loss": 2.127,
+      "step": 8613000
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 3.753823609438682e-05,
+      "loss": 2.1571,
+      "step": 8613500
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 3.753751244673954e-05,
+      "loss": 2.16,
+      "step": 8614000
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 3.753678879909226e-05,
+      "loss": 2.1639,
+      "step": 8614500
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 3.7536065151444984e-05,
+      "loss": 2.1724,
+      "step": 8615000
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 3.7535342951093e-05,
+      "loss": 2.1632,
+      "step": 8615500
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 3.753461930344572e-05,
+      "loss": 2.1744,
+      "step": 8616000
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 3.753389710309374e-05,
+      "loss": 2.1434,
+      "step": 8616500
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 3.7533173455446466e-05,
+      "loss": 2.156,
+      "step": 8617000
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 3.753244980779919e-05,
+      "loss": 2.1484,
+      "step": 8617500
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 3.753172616015191e-05,
+      "loss": 2.1475,
+      "step": 8618000
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 3.753100251250463e-05,
+      "loss": 2.1588,
+      "step": 8618500
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 3.7530278864857355e-05,
+      "loss": 2.1377,
+      "step": 8619000
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 3.752955521721008e-05,
+      "loss": 2.1535,
+      "step": 8619500
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 3.75288315695628e-05,
+      "loss": 2.1499,
+      "step": 8620000
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 3.752810792191553e-05,
+      "loss": 2.1498,
+      "step": 8620500
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 3.752738427426825e-05,
+      "loss": 2.1387,
+      "step": 8621000
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 3.752666062662097e-05,
+      "loss": 2.1541,
+      "step": 8621500
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 3.7525936978973695e-05,
+      "loss": 2.1232,
+      "step": 8622000
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 3.752521477862172e-05,
+      "loss": 2.1425,
+      "step": 8622500
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 3.752449113097444e-05,
+      "loss": 2.1524,
+      "step": 8623000
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 3.7523768930622455e-05,
+      "loss": 2.1705,
+      "step": 8623500
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 3.752304528297518e-05,
+      "loss": 2.1598,
+      "step": 8624000
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 3.75223216353279e-05,
+      "loss": 2.1421,
+      "step": 8624500
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 3.752159798768062e-05,
+      "loss": 2.1577,
+      "step": 8625000
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 3.7520874340033344e-05,
+      "loss": 2.1655,
+      "step": 8625500
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 3.7520150692386066e-05,
+      "loss": 2.1316,
+      "step": 8626000
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 3.751942704473879e-05,
+      "loss": 2.1615,
+      "step": 8626500
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 3.751870339709152e-05,
+      "loss": 2.1835,
+      "step": 8627000
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 3.751798119673953e-05,
+      "loss": 2.1476,
+      "step": 8627500
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 3.751725754909226e-05,
+      "loss": 2.1477,
+      "step": 8628000
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 3.751653534874028e-05,
+      "loss": 2.1672,
+      "step": 8628500
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 3.7515811701093e-05,
+      "loss": 2.1265,
+      "step": 8629000
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 3.751508805344572e-05,
+      "loss": 2.1589,
+      "step": 8629500
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 3.7514364405798444e-05,
+      "loss": 2.153,
+      "step": 8630000
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 3.751364075815117e-05,
+      "loss": 2.1568,
+      "step": 8630500
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 3.751291855779919e-05,
+      "loss": 2.1553,
+      "step": 8631000
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 3.751219491015191e-05,
+      "loss": 2.1537,
+      "step": 8631500
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 3.7511471262504633e-05,
+      "loss": 2.1553,
+      "step": 8632000
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 3.7510747614857356e-05,
+      "loss": 2.1513,
+      "step": 8632500
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 3.751002396721008e-05,
+      "loss": 2.1253,
+      "step": 8633000
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 3.75093003195628e-05,
+      "loss": 2.1523,
+      "step": 8633500
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 3.7508578119210816e-05,
+      "loss": 2.1388,
+      "step": 8634000
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 3.7507854471563545e-05,
+      "loss": 2.1643,
+      "step": 8634500
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 3.750713082391627e-05,
+      "loss": 2.1456,
+      "step": 8635000
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 3.7506407176268996e-05,
+      "loss": 2.1356,
+      "step": 8635500
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 3.750568352862172e-05,
+      "loss": 2.151,
+      "step": 8636000
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 3.750495988097444e-05,
+      "loss": 2.1853,
+      "step": 8636500
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.6643911972760643,
+      "eval_accuracy_mlm": 0.6288164705108058,
+      "eval_accuracy_nsp": 0.855142007305459,
+      "eval_loss": 2.1985487937927246,
+      "eval_runtime": 331.5793,
+      "eval_samples_per_second": 1316.084,
+      "eval_steps_per_second": 54.838,
+      "step": 8636800
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 3.750423623332716e-05,
+      "loss": 2.1611,
+      "step": 8637000
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 3.7503512585679885e-05,
+      "loss": 2.1248,
+      "step": 8637500
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 3.750278893803261e-05,
+      "loss": 2.116,
+      "step": 8638000
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 3.750206529038533e-05,
+      "loss": 2.1289,
+      "step": 8638500
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 3.7501343090033345e-05,
+      "loss": 2.1352,
+      "step": 8639000
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 3.750061944238607e-05,
+      "loss": 2.1219,
+      "step": 8639500
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 3.7499895794738796e-05,
+      "loss": 2.1537,
+      "step": 8640000
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 3.749917214709152e-05,
+      "loss": 2.1422,
+      "step": 8640500
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 3.7498449946739534e-05,
+      "loss": 2.1252,
+      "step": 8641000
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 3.7497726299092256e-05,
+      "loss": 2.1215,
+      "step": 8641500
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 3.749700265144498e-05,
+      "loss": 2.109,
+      "step": 8642000
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 3.7496280451092994e-05,
+      "loss": 2.1269,
+      "step": 8642500
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 3.749555680344572e-05,
+      "loss": 2.113,
+      "step": 8643000
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 3.7494833155798445e-05,
+      "loss": 2.1451,
+      "step": 8643500
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 3.749410950815117e-05,
+      "loss": 2.1177,
+      "step": 8644000
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 3.749338730779919e-05,
+      "loss": 2.1566,
+      "step": 8644500
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 3.749266366015191e-05,
+      "loss": 2.1421,
+      "step": 8645000
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 3.7491940012504634e-05,
+      "loss": 2.1486,
+      "step": 8645500
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 3.7491216364857357e-05,
+      "loss": 2.1403,
+      "step": 8646000
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 3.749049271721008e-05,
+      "loss": 2.1392,
+      "step": 8646500
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 3.74897690695628e-05,
+      "loss": 2.1235,
+      "step": 8647000
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 3.748904542191552e-05,
+      "loss": 2.1229,
+      "step": 8647500
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 3.7488321774268245e-05,
+      "loss": 2.1268,
+      "step": 8648000
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 3.748759812662097e-05,
+      "loss": 2.1283,
+      "step": 8648500
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 3.748687592626899e-05,
+      "loss": 2.1199,
+      "step": 8649000
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 3.748615227862171e-05,
+      "loss": 2.1399,
+      "step": 8649500
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 3.748543007826973e-05,
+      "loss": 2.1184,
+      "step": 8650000
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 3.748470643062246e-05,
+      "loss": 2.1636,
+      "step": 8650500
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 3.748398278297518e-05,
+      "loss": 2.1227,
+      "step": 8651000
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 3.74832591353279e-05,
+      "loss": 2.1397,
+      "step": 8651500
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 3.7482535487680624e-05,
+      "loss": 2.1339,
+      "step": 8652000
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 3.7481811840033346e-05,
+      "loss": 2.1513,
+      "step": 8652500
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 3.748108819238607e-05,
+      "loss": 2.1268,
+      "step": 8653000
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 3.748036599203409e-05,
+      "loss": 2.1305,
+      "step": 8653500
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 3.747964234438681e-05,
+      "loss": 2.1514,
+      "step": 8654000
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 3.7478918696739535e-05,
+      "loss": 2.1455,
+      "step": 8654500
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 3.747819504909226e-05,
+      "loss": 2.1103,
+      "step": 8655000
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 3.747747284874027e-05,
+      "loss": 2.1387,
+      "step": 8655500
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 3.7476749201092995e-05,
+      "loss": 2.1374,
+      "step": 8656000
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 3.7476025553445724e-05,
+      "loss": 2.1478,
+      "step": 8656500
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 3.7475301905798446e-05,
+      "loss": 2.1449,
+      "step": 8657000
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 3.747457825815117e-05,
+      "loss": 2.165,
+      "step": 8657500
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 3.74738546105039e-05,
+      "loss": 2.1407,
+      "step": 8658000
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 3.747313096285662e-05,
+      "loss": 2.1436,
+      "step": 8658500
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 3.747240731520934e-05,
+      "loss": 2.1387,
+      "step": 8659000
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 3.7471683667562064e-05,
+      "loss": 2.1306,
+      "step": 8659500
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 3.7470960019914786e-05,
+      "loss": 2.133,
+      "step": 8660000
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 3.747023637226751e-05,
+      "loss": 2.146,
+      "step": 8660500
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 3.746951272462023e-05,
+      "loss": 2.132,
+      "step": 8661000
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 3.746878907697295e-05,
+      "loss": 2.1471,
+      "step": 8661500
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 3.7468065429325675e-05,
+      "loss": 2.1235,
+      "step": 8662000
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 3.74673432289737e-05,
+      "loss": 2.1359,
+      "step": 8662500
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 3.746661958132642e-05,
+      "loss": 2.1529,
+      "step": 8663000
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 3.7465897380974435e-05,
+      "loss": 2.1217,
+      "step": 8663500
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 3.746517518062245e-05,
+      "loss": 2.1262,
+      "step": 8664000
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 3.746445153297517e-05,
+      "loss": 2.1333,
+      "step": 8664500
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 3.7463727885327895e-05,
+      "loss": 2.1118,
+      "step": 8665000
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 3.7463004237680624e-05,
+      "loss": 2.1291,
+      "step": 8665500
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 3.7462280590033347e-05,
+      "loss": 2.1271,
+      "step": 8666000
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 3.7461556942386076e-05,
+      "loss": 2.1451,
+      "step": 8666500
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 3.74608332947388e-05,
+      "loss": 2.1524,
+      "step": 8667000
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 3.746010964709152e-05,
+      "loss": 2.1012,
+      "step": 8667500
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 3.745938599944424e-05,
+      "loss": 2.1293,
+      "step": 8668000
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 3.7458662351796964e-05,
+      "loss": 2.1278,
+      "step": 8668500
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 3.745793870414969e-05,
+      "loss": 2.1313,
+      "step": 8669000
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 3.74572165037977e-05,
+      "loss": 2.1278,
+      "step": 8669500
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 3.7456492856150425e-05,
+      "loss": 2.1348,
+      "step": 8670000
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 3.745576920850315e-05,
+      "loss": 2.1306,
+      "step": 8670500
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 3.7455045560855876e-05,
+      "loss": 2.1573,
+      "step": 8671000
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 3.74543219132086e-05,
+      "loss": 2.1213,
+      "step": 8671500
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 3.745359826556132e-05,
+      "loss": 2.1283,
+      "step": 8672000
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 3.745287461791405e-05,
+      "loss": 2.1506,
+      "step": 8672500
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 3.745215097026677e-05,
+      "loss": 2.155,
+      "step": 8673000
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 3.7451427322619494e-05,
+      "loss": 2.1521,
+      "step": 8673500
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 3.745070512226751e-05,
+      "loss": 2.1436,
+      "step": 8674000
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 3.744998147462023e-05,
+      "loss": 2.1416,
+      "step": 8674500
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 3.7449257826972954e-05,
+      "loss": 2.1242,
+      "step": 8675000
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 3.7448534179325676e-05,
+      "loss": 2.1427,
+      "step": 8675500
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 3.74478105316784e-05,
+      "loss": 2.164,
+      "step": 8676000
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 3.744708833132642e-05,
+      "loss": 2.1568,
+      "step": 8676500
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 3.744636468367914e-05,
+      "loss": 2.1389,
+      "step": 8677000
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 3.744564248332716e-05,
+      "loss": 2.1179,
+      "step": 8677500
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 3.744491883567988e-05,
+      "loss": 2.1256,
+      "step": 8678000
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 3.74441951880326e-05,
+      "loss": 2.1189,
+      "step": 8678500
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 3.7443471540385325e-05,
+      "loss": 2.1361,
+      "step": 8679000
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 3.744274789273805e-05,
+      "loss": 2.1518,
+      "step": 8679500
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 3.7442024245090776e-05,
+      "loss": 2.157,
+      "step": 8680000
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 3.74413005974435e-05,
+      "loss": 2.1375,
+      "step": 8680500
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 3.7440579844386814e-05,
+      "loss": 2.1262,
+      "step": 8681000
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 3.7439856196739536e-05,
+      "loss": 2.1367,
+      "step": 8681500
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 3.743913254909226e-05,
+      "loss": 2.1424,
+      "step": 8682000
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 3.743840890144498e-05,
+      "loss": 2.1308,
+      "step": 8682500
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 3.74376852537977e-05,
+      "loss": 2.1413,
+      "step": 8683000
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 3.7436961606150425e-05,
+      "loss": 2.1377,
+      "step": 8683500
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 3.743623795850315e-05,
+      "loss": 2.1568,
+      "step": 8684000
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 3.7435514310855877e-05,
+      "loss": 2.1175,
+      "step": 8684500
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 3.74347906632086e-05,
+      "loss": 2.1596,
+      "step": 8685000
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 3.743406701556132e-05,
+      "loss": 2.1602,
+      "step": 8685500
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 3.743334336791404e-05,
+      "loss": 2.1425,
+      "step": 8686000
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 3.7432619720266765e-05,
+      "loss": 2.1432,
+      "step": 8686500
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 3.743189607261949e-05,
+      "loss": 2.1264,
+      "step": 8687000
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 3.743117242497222e-05,
+      "loss": 2.1438,
+      "step": 8687500
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 3.743044877732494e-05,
+      "loss": 2.1249,
+      "step": 8688000
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 3.742972512967766e-05,
+      "loss": 2.1357,
+      "step": 8688500
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 3.7429001482030383e-05,
+      "loss": 2.1598,
+      "step": 8689000
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 3.7428277834383106e-05,
+      "loss": 2.1356,
+      "step": 8689500
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 3.742755418673583e-05,
+      "loss": 2.1433,
+      "step": 8690000
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 3.742683053908855e-05,
+      "loss": 2.1297,
+      "step": 8690500
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 3.742610833873657e-05,
+      "loss": 2.1297,
+      "step": 8691000
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 3.742538613838459e-05,
+      "loss": 2.1294,
+      "step": 8691500
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 3.742466249073731e-05,
+      "loss": 2.1544,
+      "step": 8692000
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 3.742393884309003e-05,
+      "loss": 2.1588,
+      "step": 8692500
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 3.7423215195442755e-05,
+      "loss": 2.1517,
+      "step": 8693000
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 3.742249299509078e-05,
+      "loss": 2.1401,
+      "step": 8693500
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 3.742177079473879e-05,
+      "loss": 2.1348,
+      "step": 8694000
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 3.7421047147091515e-05,
+      "loss": 2.119,
+      "step": 8694500
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 3.7420323499444244e-05,
+      "loss": 2.1308,
+      "step": 8695000
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 3.7419599851796966e-05,
+      "loss": 2.1231,
+      "step": 8695500
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 3.741887620414969e-05,
+      "loss": 2.1295,
+      "step": 8696000
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 3.741815255650241e-05,
+      "loss": 2.1423,
+      "step": 8696500
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 3.7417430356150426e-05,
+      "loss": 2.1238,
+      "step": 8697000
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 3.7416706708503155e-05,
+      "loss": 2.136,
+      "step": 8697500
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 3.741598306085588e-05,
+      "loss": 2.1424,
+      "step": 8698000
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 3.74152594132086e-05,
+      "loss": 2.1348,
+      "step": 8698500
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 3.741453576556132e-05,
+      "loss": 2.1387,
+      "step": 8699000
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 3.741381356520934e-05,
+      "loss": 2.1476,
+      "step": 8699500
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 3.741308991756206e-05,
+      "loss": 2.1415,
+      "step": 8700000
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 3.741236626991478e-05,
+      "loss": 2.1634,
+      "step": 8700500
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 3.7411642622267504e-05,
+      "loss": 2.1488,
+      "step": 8701000
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 3.7410918974620226e-05,
+      "loss": 2.1551,
+      "step": 8701500
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 3.7410195326972955e-05,
+      "loss": 2.1446,
+      "step": 8702000
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 3.740947167932568e-05,
+      "loss": 2.144,
+      "step": 8702500
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 3.7408748031678407e-05,
+      "loss": 2.1337,
+      "step": 8703000
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 3.740802583132642e-05,
+      "loss": 2.1356,
+      "step": 8703500
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 3.740730363097444e-05,
+      "loss": 2.1515,
+      "step": 8704000
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 3.740657998332716e-05,
+      "loss": 2.1283,
+      "step": 8704500
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 3.740585633567988e-05,
+      "loss": 2.1364,
+      "step": 8705000
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 3.7405132688032604e-05,
+      "loss": 2.1344,
+      "step": 8705500
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 3.740441048768063e-05,
+      "loss": 2.1311,
+      "step": 8706000
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 3.740368684003335e-05,
+      "loss": 2.1441,
+      "step": 8706500
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 3.7402964639681364e-05,
+      "loss": 2.1606,
+      "step": 8707000
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 3.740224099203409e-05,
+      "loss": 2.1437,
+      "step": 8707500
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 3.740151734438681e-05,
+      "loss": 2.1245,
+      "step": 8708000
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 3.740079369673953e-05,
+      "loss": 2.1367,
+      "step": 8708500
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 3.7400070049092253e-05,
+      "loss": 2.1456,
+      "step": 8709000
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 3.7399346401444976e-05,
+      "loss": 2.1287,
+      "step": 8709500
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 3.7398622753797705e-05,
+      "loss": 2.1102,
+      "step": 8710000
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 3.739789910615043e-05,
+      "loss": 2.1369,
+      "step": 8710500
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 3.7397175458503156e-05,
+      "loss": 2.1073,
+      "step": 8711000
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 3.739645181085588e-05,
+      "loss": 2.1512,
+      "step": 8711500
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 3.7395729610503894e-05,
+      "loss": 2.1504,
+      "step": 8712000
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 3.7395005962856616e-05,
+      "loss": 2.1513,
+      "step": 8712500
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 3.739428231520934e-05,
+      "loss": 2.1373,
+      "step": 8713000
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 3.739355866756206e-05,
+      "loss": 2.1274,
+      "step": 8713500
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 3.739283501991478e-05,
+      "loss": 2.1575,
+      "step": 8714000
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 3.7392112819562805e-05,
+      "loss": 2.1315,
+      "step": 8714500
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 3.739138917191553e-05,
+      "loss": 2.1499,
+      "step": 8715000
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 3.739066552426825e-05,
+      "loss": 2.1287,
+      "step": 8715500
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 3.738994187662097e-05,
+      "loss": 2.1483,
+      "step": 8716000
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 3.7389218228973694e-05,
+      "loss": 2.1349,
+      "step": 8716500
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 3.7388494581326416e-05,
+      "loss": 2.1438,
+      "step": 8717000
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 3.7387770933679145e-05,
+      "loss": 2.1411,
+      "step": 8717500
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 3.738704728603187e-05,
+      "loss": 2.1614,
+      "step": 8718000
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 3.738632363838459e-05,
+      "loss": 2.135,
+      "step": 8718500
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 3.738559999073731e-05,
+      "loss": 2.1449,
+      "step": 8719000
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 3.7384876343090034e-05,
+      "loss": 2.1572,
+      "step": 8719500
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 3.7384152695442756e-05,
+      "loss": 2.1376,
+      "step": 8720000
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 3.738343049509078e-05,
+      "loss": 2.1297,
+      "step": 8720500
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 3.73827068474435e-05,
+      "loss": 2.1333,
+      "step": 8721000
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 3.738198319979622e-05,
+      "loss": 2.1251,
+      "step": 8721500
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 3.738126099944424e-05,
+      "loss": 2.1384,
+      "step": 8722000
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 3.738053735179696e-05,
+      "loss": 2.13,
+      "step": 8722500
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 3.737981370414968e-05,
+      "loss": 2.157,
+      "step": 8723000
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 3.7379090056502405e-05,
+      "loss": 2.1382,
+      "step": 8723500
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 3.737836785615043e-05,
+      "loss": 2.1392,
+      "step": 8724000
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 3.737764420850315e-05,
+      "loss": 2.12,
+      "step": 8724500
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 3.737692056085588e-05,
+      "loss": 2.1294,
+      "step": 8725000
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 3.73761969132086e-05,
+      "loss": 2.1359,
+      "step": 8725500
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 3.7375473265561323e-05,
+      "loss": 2.1526,
+      "step": 8726000
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 3.7374749617914046e-05,
+      "loss": 2.1395,
+      "step": 8726500
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 3.737402597026677e-05,
+      "loss": 2.1501,
+      "step": 8727000
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 3.7373303769914783e-05,
+      "loss": 2.1576,
+      "step": 8727500
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 3.7372580122267506e-05,
+      "loss": 2.1446,
+      "step": 8728000
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 3.7371856474620235e-05,
+      "loss": 2.1384,
+      "step": 8728500
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 3.737113282697296e-05,
+      "loss": 2.1047,
+      "step": 8729000
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 3.737041062662097e-05,
+      "loss": 2.1382,
+      "step": 8729500
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 3.7369686978973695e-05,
+      "loss": 2.1428,
+      "step": 8730000
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 3.736896333132642e-05,
+      "loss": 2.1387,
+      "step": 8730500
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 3.736823968367914e-05,
+      "loss": 2.1367,
+      "step": 8731000
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 3.736751603603186e-05,
+      "loss": 2.122,
+      "step": 8731500
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 3.7366792388384584e-05,
+      "loss": 2.1418,
+      "step": 8732000
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 3.736606874073731e-05,
+      "loss": 2.117,
+      "step": 8732500
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 3.7365346540385335e-05,
+      "loss": 2.1505,
+      "step": 8733000
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 3.736462289273806e-05,
+      "loss": 2.1168,
+      "step": 8733500
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 3.736389924509078e-05,
+      "loss": 2.1482,
+      "step": 8734000
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 3.73631755974435e-05,
+      "loss": 2.1437,
+      "step": 8734500
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 3.7362451949796224e-05,
+      "loss": 2.1474,
+      "step": 8735000
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 3.7361728302148946e-05,
+      "loss": 2.1335,
+      "step": 8735500
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 3.736100465450167e-05,
+      "loss": 2.1469,
+      "step": 8736000
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 3.736028100685439e-05,
+      "loss": 2.1337,
+      "step": 8736500
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 3.735955735920711e-05,
+      "loss": 2.142,
+      "step": 8737000
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 3.7358833711559835e-05,
+      "loss": 2.1372,
+      "step": 8737500
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 3.735811006391256e-05,
+      "loss": 2.1352,
+      "step": 8738000
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 3.7357386416265286e-05,
+      "loss": 2.1173,
+      "step": 8738500
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 3.7356665663208595e-05,
+      "loss": 2.1318,
+      "step": 8739000
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 3.735594201556132e-05,
+      "loss": 2.1281,
+      "step": 8739500
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 3.7355218367914046e-05,
+      "loss": 2.1388,
+      "step": 8740000
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 3.735449472026677e-05,
+      "loss": 2.158,
+      "step": 8740500
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 3.7353772519914784e-05,
+      "loss": 2.1679,
+      "step": 8741000
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 3.7353048872267506e-05,
+      "loss": 2.1372,
+      "step": 8741500
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 3.7352325224620235e-05,
+      "loss": 2.1399,
+      "step": 8742000
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 3.735160157697296e-05,
+      "loss": 2.1442,
+      "step": 8742500
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 3.735087792932568e-05,
+      "loss": 2.143,
+      "step": 8743000
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 3.7350155728973695e-05,
+      "loss": 2.1479,
+      "step": 8743500
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 3.734943208132642e-05,
+      "loss": 2.1425,
+      "step": 8744000
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 3.734870843367914e-05,
+      "loss": 2.1431,
+      "step": 8744500
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 3.734798478603186e-05,
+      "loss": 2.1581,
+      "step": 8745000
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 3.7347261138384584e-05,
+      "loss": 2.1435,
+      "step": 8745500
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 3.734653749073731e-05,
+      "loss": 2.1275,
+      "step": 8746000
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 3.7345813843090036e-05,
+      "loss": 2.1488,
+      "step": 8746500
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 3.734509019544276e-05,
+      "loss": 2.1501,
+      "step": 8747000
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 3.734436654779549e-05,
+      "loss": 2.1434,
+      "step": 8747500
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 3.734364290014821e-05,
+      "loss": 2.1363,
+      "step": 8748000
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 3.7342920699796225e-05,
+      "loss": 2.1485,
+      "step": 8748500
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 3.734219849944424e-05,
+      "loss": 2.1238,
+      "step": 8749000
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 3.734147629909226e-05,
+      "loss": 2.1669,
+      "step": 8749500
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 3.7340752651444985e-05,
+      "loss": 2.1388,
+      "step": 8750000
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 3.734002900379771e-05,
+      "loss": 2.1322,
+      "step": 8750500
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 3.733930535615043e-05,
+      "loss": 2.1292,
+      "step": 8751000
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 3.733858170850315e-05,
+      "loss": 2.133,
+      "step": 8751500
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 3.733785950815117e-05,
+      "loss": 2.1313,
+      "step": 8752000
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 3.733713586050389e-05,
+      "loss": 2.1622,
+      "step": 8752500
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 3.733641221285661e-05,
+      "loss": 2.1408,
+      "step": 8753000
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 3.7335688565209334e-05,
+      "loss": 2.1379,
+      "step": 8753500
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 3.733496491756206e-05,
+      "loss": 2.1283,
+      "step": 8754000
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 3.733424271721008e-05,
+      "loss": 2.1493,
+      "step": 8754500
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 3.73335190695628e-05,
+      "loss": 2.1383,
+      "step": 8755000
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 3.733279542191553e-05,
+      "loss": 2.131,
+      "step": 8755500
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 3.733207177426825e-05,
+      "loss": 2.1597,
+      "step": 8756000
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 3.7331348126620974e-05,
+      "loss": 2.1396,
+      "step": 8756500
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 3.7330624478973696e-05,
+      "loss": 2.1267,
+      "step": 8757000
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 3.732990227862171e-05,
+      "loss": 2.151,
+      "step": 8757500
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 3.7329178630974434e-05,
+      "loss": 2.143,
+      "step": 8758000
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 3.732845498332716e-05,
+      "loss": 2.1423,
+      "step": 8758500
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 3.732773278297518e-05,
+      "loss": 2.1291,
+      "step": 8759000
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 3.73270091353279e-05,
+      "loss": 2.1456,
+      "step": 8759500
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 3.732628548768062e-05,
+      "loss": 2.1502,
+      "step": 8760000
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 3.7325561840033345e-05,
+      "loss": 2.1525,
+      "step": 8760500
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 3.732483819238607e-05,
+      "loss": 2.1447,
+      "step": 8761000
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 3.732411454473879e-05,
+      "loss": 2.1554,
+      "step": 8761500
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 3.732339089709151e-05,
+      "loss": 2.1539,
+      "step": 8762000
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 3.732266724944424e-05,
+      "loss": 2.1576,
+      "step": 8762500
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 3.732194360179696e-05,
+      "loss": 2.1255,
+      "step": 8763000
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 3.7321219954149686e-05,
+      "loss": 2.1441,
+      "step": 8763500
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 3.7320496306502415e-05,
+      "loss": 2.1539,
+      "step": 8764000
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 3.731977410615043e-05,
+      "loss": 2.1425,
+      "step": 8764500
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 3.731905045850315e-05,
+      "loss": 2.1528,
+      "step": 8765000
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 3.7318326810855875e-05,
+      "loss": 2.1503,
+      "step": 8765500
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 3.73176031632086e-05,
+      "loss": 2.1622,
+      "step": 8766000
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 3.731687951556132e-05,
+      "loss": 2.145,
+      "step": 8766500
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 3.731615586791404e-05,
+      "loss": 2.1288,
+      "step": 8767000
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 3.7315432220266763e-05,
+      "loss": 2.1321,
+      "step": 8767500
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 3.7314708572619486e-05,
+      "loss": 2.1285,
+      "step": 8768000
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 3.7313984924972215e-05,
+      "loss": 2.1213,
+      "step": 8768500
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 3.731326127732494e-05,
+      "loss": 2.1506,
+      "step": 8769000
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 3.731253907697295e-05,
+      "loss": 2.1513,
+      "step": 8769500
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 3.731181542932568e-05,
+      "loss": 2.1257,
+      "step": 8770000
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 3.7311091781678404e-05,
+      "loss": 2.1353,
+      "step": 8770500
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 3.7310368134031126e-05,
+      "loss": 2.1617,
+      "step": 8771000
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 3.730964593367914e-05,
+      "loss": 2.1375,
+      "step": 8771500
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 3.7308922286031864e-05,
+      "loss": 2.1609,
+      "step": 8772000
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 3.7308198638384586e-05,
+      "loss": 2.1546,
+      "step": 8772500
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 3.730747643803261e-05,
+      "loss": 2.1633,
+      "step": 8773000
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 3.730675279038533e-05,
+      "loss": 2.1565,
+      "step": 8773500
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 3.730602914273805e-05,
+      "loss": 2.1306,
+      "step": 8774000
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 3.7305305495090775e-05,
+      "loss": 2.1513,
+      "step": 8774500
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 3.73045818474435e-05,
+      "loss": 2.1453,
+      "step": 8775000
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 3.730385819979622e-05,
+      "loss": 2.1375,
+      "step": 8775500
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 3.730313455214894e-05,
+      "loss": 2.1445,
+      "step": 8776000
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 3.7302410904501664e-05,
+      "loss": 2.1442,
+      "step": 8776500
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 3.7301688704149686e-05,
+      "loss": 2.1414,
+      "step": 8777000
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 3.7300965056502415e-05,
+      "loss": 2.1414,
+      "step": 8777500
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 3.730024140885514e-05,
+      "loss": 2.1443,
+      "step": 8778000
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 3.729951776120786e-05,
+      "loss": 2.1391,
+      "step": 8778500
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 3.729879411356058e-05,
+      "loss": 2.1352,
+      "step": 8779000
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 3.7298070465913304e-05,
+      "loss": 2.1166,
+      "step": 8779500
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 3.7297346818266026e-05,
+      "loss": 2.146,
+      "step": 8780000
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 3.729662317061875e-05,
+      "loss": 2.1499,
+      "step": 8780500
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 3.729589952297147e-05,
+      "loss": 2.1289,
+      "step": 8781000
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 3.729517732261949e-05,
+      "loss": 2.1581,
+      "step": 8781500
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 3.7294453674972216e-05,
+      "loss": 2.1548,
+      "step": 8782000
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 3.729373002732494e-05,
+      "loss": 2.152,
+      "step": 8782500
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 3.729300637967766e-05,
+      "loss": 2.1315,
+      "step": 8783000
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 3.7292284179325676e-05,
+      "loss": 2.1271,
+      "step": 8783500
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 3.72915605316784e-05,
+      "loss": 2.1267,
+      "step": 8784000
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 3.729083688403112e-05,
+      "loss": 2.1561,
+      "step": 8784500
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 3.729011323638385e-05,
+      "loss": 2.1385,
+      "step": 8785000
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 3.7289391036031865e-05,
+      "loss": 2.1532,
+      "step": 8785500
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 3.7288667388384594e-05,
+      "loss": 2.1355,
+      "step": 8786000
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 3.728794518803261e-05,
+      "loss": 2.1468,
+      "step": 8786500
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 3.728722154038533e-05,
+      "loss": 2.1671,
+      "step": 8787000
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 3.7286497892738054e-05,
+      "loss": 2.1503,
+      "step": 8787500
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 3.728577569238607e-05,
+      "loss": 2.1667,
+      "step": 8788000
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 3.728505204473879e-05,
+      "loss": 2.1319,
+      "step": 8788500
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 3.7284328397091514e-05,
+      "loss": 2.1254,
+      "step": 8789000
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 3.728360474944424e-05,
+      "loss": 2.1396,
+      "step": 8789500
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 3.7282881101796965e-05,
+      "loss": 2.1369,
+      "step": 8790000
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 3.728215745414969e-05,
+      "loss": 2.1298,
+      "step": 8790500
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 3.728143380650241e-05,
+      "loss": 2.1316,
+      "step": 8791000
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 3.728071015885513e-05,
+      "loss": 2.1597,
+      "step": 8791500
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 3.7279986511207854e-05,
+      "loss": 2.1239,
+      "step": 8792000
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 3.727926286356058e-05,
+      "loss": 2.1307,
+      "step": 8792500
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 3.7278539215913305e-05,
+      "loss": 2.1287,
+      "step": 8793000
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 3.727781556826603e-05,
+      "loss": 2.1383,
+      "step": 8793500
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 3.727709192061875e-05,
+      "loss": 2.1386,
+      "step": 8794000
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 3.727636827297147e-05,
+      "loss": 2.1388,
+      "step": 8794500
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 3.7275644625324194e-05,
+      "loss": 2.1361,
+      "step": 8795000
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 3.7274920977676916e-05,
+      "loss": 2.1362,
+      "step": 8795500
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 3.7274197330029645e-05,
+      "loss": 2.1527,
+      "step": 8796000
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 3.727347368238237e-05,
+      "loss": 2.1298,
+      "step": 8796500
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 3.727275003473509e-05,
+      "loss": 2.1528,
+      "step": 8797000
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 3.727202638708781e-05,
+      "loss": 2.1531,
+      "step": 8797500
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 3.727130418673583e-05,
+      "loss": 2.1447,
+      "step": 8798000
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 3.727058343367914e-05,
+      "loss": 2.1628,
+      "step": 8798500
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 3.7269859786031865e-05,
+      "loss": 2.1406,
+      "step": 8799000
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 3.726913613838459e-05,
+      "loss": 2.1397,
+      "step": 8799500
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 3.726841249073732e-05,
+      "loss": 2.1713,
+      "step": 8800000
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 3.726769029038533e-05,
+      "loss": 2.126,
+      "step": 8800500
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 3.7266966642738054e-05,
+      "loss": 2.133,
+      "step": 8801000
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 3.726624299509078e-05,
+      "loss": 2.1465,
+      "step": 8801500
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 3.72655193474435e-05,
+      "loss": 2.1265,
+      "step": 8802000
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 3.726479714709152e-05,
+      "loss": 2.1425,
+      "step": 8802500
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 3.7264073499444243e-05,
+      "loss": 2.1311,
+      "step": 8803000
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 3.7263349851796966e-05,
+      "loss": 2.1497,
+      "step": 8803500
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 3.726262620414969e-05,
+      "loss": 2.1495,
+      "step": 8804000
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 3.726190255650241e-05,
+      "loss": 2.149,
+      "step": 8804500
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 3.726117890885513e-05,
+      "loss": 2.1636,
+      "step": 8805000
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 3.7260455261207855e-05,
+      "loss": 2.1421,
+      "step": 8805500
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 3.725973161356058e-05,
+      "loss": 2.1386,
+      "step": 8806000
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 3.72590079659133e-05,
+      "loss": 2.1569,
+      "step": 8806500
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 3.725828431826602e-05,
+      "loss": 2.1697,
+      "step": 8807000
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 3.725756067061875e-05,
+      "loss": 2.1843,
+      "step": 8807500
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 3.725683702297147e-05,
+      "loss": 2.1585,
+      "step": 8808000
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 3.7256113375324195e-05,
+      "loss": 2.1302,
+      "step": 8808500
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 3.725538972767692e-05,
+      "loss": 2.122,
+      "step": 8809000
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 3.7254666080029646e-05,
+      "loss": 2.1358,
+      "step": 8809500
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 3.725394243238237e-05,
+      "loss": 2.1437,
+      "step": 8810000
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 3.725321878473509e-05,
+      "loss": 2.138,
+      "step": 8810500
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 3.7252496584383106e-05,
+      "loss": 2.152,
+      "step": 8811000
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 3.725177438403112e-05,
+      "loss": 2.1411,
+      "step": 8811500
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 3.7251050736383844e-05,
+      "loss": 2.1417,
+      "step": 8812000
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 3.725032708873657e-05,
+      "loss": 2.1396,
+      "step": 8812500
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 3.7249603441089295e-05,
+      "loss": 2.127,
+      "step": 8813000
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 3.724887979344202e-05,
+      "loss": 2.1191,
+      "step": 8813500
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 3.724815759309003e-05,
+      "loss": 2.1389,
+      "step": 8814000
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 3.7247433945442755e-05,
+      "loss": 2.1538,
+      "step": 8814500
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 3.7246710297795484e-05,
+      "loss": 2.1251,
+      "step": 8815000
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 3.7245986650148206e-05,
+      "loss": 2.1593,
+      "step": 8815500
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 3.724526589709152e-05,
+      "loss": 2.1379,
+      "step": 8816000
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 3.724454369673954e-05,
+      "loss": 2.1711,
+      "step": 8816500
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 3.724382004909226e-05,
+      "loss": 2.1667,
+      "step": 8817000
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 3.724309640144498e-05,
+      "loss": 2.111,
+      "step": 8817500
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 3.7242372753797704e-05,
+      "loss": 2.1671,
+      "step": 8818000
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 3.7241649106150426e-05,
+      "loss": 2.136,
+      "step": 8818500
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 3.724092545850315e-05,
+      "loss": 2.1636,
+      "step": 8819000
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 3.724020181085587e-05,
+      "loss": 2.1358,
+      "step": 8819500
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 3.723947816320859e-05,
+      "loss": 2.1514,
+      "step": 8820000
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 3.723875451556132e-05,
+      "loss": 2.1552,
+      "step": 8820500
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 3.7238030867914044e-05,
+      "loss": 2.1541,
+      "step": 8821000
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 3.723730866756206e-05,
+      "loss": 2.1513,
+      "step": 8821500
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 3.723658501991478e-05,
+      "loss": 2.1404,
+      "step": 8822000
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 3.723586137226751e-05,
+      "loss": 2.1584,
+      "step": 8822500
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 3.723513917191553e-05,
+      "loss": 2.1296,
+      "step": 8823000
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 3.723441552426825e-05,
+      "loss": 2.1703,
+      "step": 8823500
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 3.723369187662097e-05,
+      "loss": 2.1471,
+      "step": 8824000
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 3.72329682289737e-05,
+      "loss": 2.1183,
+      "step": 8824500
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 3.7232246028621716e-05,
+      "loss": 2.1427,
+      "step": 8825000
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 3.723152238097444e-05,
+      "loss": 2.1526,
+      "step": 8825500
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 3.723079873332716e-05,
+      "loss": 2.1572,
+      "step": 8826000
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 3.723007508567988e-05,
+      "loss": 2.1235,
+      "step": 8826500
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 3.7229351438032605e-05,
+      "loss": 2.1413,
+      "step": 8827000
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 3.722862779038533e-05,
+      "loss": 2.1589,
+      "step": 8827500
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 3.722790414273805e-05,
+      "loss": 2.1673,
+      "step": 8828000
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 3.722718049509077e-05,
+      "loss": 2.1736,
+      "step": 8828500
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 3.72264568474435e-05,
+      "loss": 2.1289,
+      "step": 8829000
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 3.7225734647091516e-05,
+      "loss": 2.1274,
+      "step": 8829500
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 3.7225010999444245e-05,
+      "loss": 2.1674,
+      "step": 8830000
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 3.722428735179697e-05,
+      "loss": 2.1582,
+      "step": 8830500
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 3.722356370414969e-05,
+      "loss": 2.1191,
+      "step": 8831000
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 3.722284005650241e-05,
+      "loss": 2.1365,
+      "step": 8831500
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 3.7222116408855134e-05,
+      "loss": 2.1457,
+      "step": 8832000
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 3.7221392761207856e-05,
+      "loss": 2.1431,
+      "step": 8832500
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 3.722066911356058e-05,
+      "loss": 2.134,
+      "step": 8833000
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 3.72199469132086e-05,
+      "loss": 2.1304,
+      "step": 8833500
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 3.721922326556132e-05,
+      "loss": 2.1316,
+      "step": 8834000
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 3.7218499617914045e-05,
+      "loss": 2.1543,
+      "step": 8834500
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 3.721777597026677e-05,
+      "loss": 2.1429,
+      "step": 8835000
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 3.721705376991478e-05,
+      "loss": 2.1336,
+      "step": 8835500
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 3.7216330122267505e-05,
+      "loss": 2.1074,
+      "step": 8836000
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 3.721560647462023e-05,
+      "loss": 2.1454,
+      "step": 8836500
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 3.721488427426825e-05,
+      "loss": 2.1506,
+      "step": 8837000
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 3.721416062662097e-05,
+      "loss": 2.1578,
+      "step": 8837500
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 3.72134369789737e-05,
+      "loss": 2.1417,
+      "step": 8838000
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 3.721271333132642e-05,
+      "loss": 2.163,
+      "step": 8838500
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 3.7211989683679146e-05,
+      "loss": 2.1459,
+      "step": 8839000
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 3.721126603603187e-05,
+      "loss": 2.1443,
+      "step": 8839500
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 3.721054238838459e-05,
+      "loss": 2.1338,
+      "step": 8840000
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 3.720981874073731e-05,
+      "loss": 2.1389,
+      "step": 8840500
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 3.7209095093090034e-05,
+      "loss": 2.158,
+      "step": 8841000
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 3.720837144544276e-05,
+      "loss": 2.1521,
+      "step": 8841500
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 3.720764779779548e-05,
+      "loss": 2.133,
+      "step": 8842000
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 3.72069241501482e-05,
+      "loss": 2.1157,
+      "step": 8842500
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 3.7206201949796223e-05,
+      "loss": 2.1605,
+      "step": 8843000
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 3.7205478302148946e-05,
+      "loss": 2.1536,
+      "step": 8843500
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 3.720475465450167e-05,
+      "loss": 2.1313,
+      "step": 8844000
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 3.720403100685439e-05,
+      "loss": 2.1387,
+      "step": 8844500
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 3.720330735920712e-05,
+      "loss": 2.1446,
+      "step": 8845000
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 3.720258371155984e-05,
+      "loss": 2.134,
+      "step": 8845500
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 3.7201860063912564e-05,
+      "loss": 2.1326,
+      "step": 8846000
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 3.7201136416265286e-05,
+      "loss": 2.1532,
+      "step": 8846500
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 3.72004142159133e-05,
+      "loss": 2.1624,
+      "step": 8847000
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 3.7199692015561324e-05,
+      "loss": 2.136,
+      "step": 8847500
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 3.7198968367914046e-05,
+      "loss": 2.1259,
+      "step": 8848000
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 3.719824472026677e-05,
+      "loss": 2.1583,
+      "step": 8848500
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 3.719752107261949e-05,
+      "loss": 2.1116,
+      "step": 8849000
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 3.719679742497221e-05,
+      "loss": 2.1055,
+      "step": 8849500
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 3.7196073777324935e-05,
+      "loss": 2.1308,
+      "step": 8850000
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 3.719535012967766e-05,
+      "loss": 2.1281,
+      "step": 8850500
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 3.719462648203038e-05,
+      "loss": 2.1534,
+      "step": 8851000
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 3.71939042816784e-05,
+      "loss": 2.151,
+      "step": 8851500
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 3.719318208132642e-05,
+      "loss": 2.1499,
+      "step": 8852000
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 3.7192458433679146e-05,
+      "loss": 2.1329,
+      "step": 8852500
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 3.719173623332716e-05,
+      "loss": 2.1712,
+      "step": 8853000
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 3.7191012585679884e-05,
+      "loss": 2.1569,
+      "step": 8853500
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 3.7190288938032606e-05,
+      "loss": 2.1365,
+      "step": 8854000
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 3.718956673768063e-05,
+      "loss": 2.154,
+      "step": 8854500
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 3.718884309003335e-05,
+      "loss": 2.1352,
+      "step": 8855000
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 3.718811944238607e-05,
+      "loss": 2.1265,
+      "step": 8855500
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 3.7187395794738795e-05,
+      "loss": 2.1489,
+      "step": 8856000
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 3.718667214709152e-05,
+      "loss": 2.1435,
+      "step": 8856500
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 3.718594849944424e-05,
+      "loss": 2.157,
+      "step": 8857000
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 3.718522485179696e-05,
+      "loss": 2.1586,
+      "step": 8857500
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 3.7184501204149684e-05,
+      "loss": 2.1255,
+      "step": 8858000
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 3.7183777556502407e-05,
+      "loss": 2.1367,
+      "step": 8858500
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 3.718305390885513e-05,
+      "loss": 2.1572,
+      "step": 8859000
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 3.718233026120785e-05,
+      "loss": 2.1621,
+      "step": 8859500
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 3.718160806085588e-05,
+      "loss": 2.1704,
+      "step": 8860000
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 3.71808844132086e-05,
+      "loss": 2.1306,
+      "step": 8860500
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 3.7180160765561325e-05,
+      "loss": 2.1422,
+      "step": 8861000
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 3.717943856520934e-05,
+      "loss": 2.1629,
+      "step": 8861500
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 3.717871491756206e-05,
+      "loss": 2.1433,
+      "step": 8862000
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 3.7177991269914785e-05,
+      "loss": 2.1427,
+      "step": 8862500
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 3.71772690695628e-05,
+      "loss": 2.1365,
+      "step": 8863000
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 3.717654542191553e-05,
+      "loss": 2.1534,
+      "step": 8863500
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 3.717582177426825e-05,
+      "loss": 2.113,
+      "step": 8864000
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 3.7175098126620974e-05,
+      "loss": 2.1388,
+      "step": 8864500
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 3.7174374478973696e-05,
+      "loss": 2.162,
+      "step": 8865000
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 3.717365083132642e-05,
+      "loss": 2.134,
+      "step": 8865500
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 3.717292718367914e-05,
+      "loss": 2.1473,
+      "step": 8866000
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 3.717220353603186e-05,
+      "loss": 2.1371,
+      "step": 8866500
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 3.7171479888384585e-05,
+      "loss": 2.1553,
+      "step": 8867000
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 3.7170756240737314e-05,
+      "loss": 2.1349,
+      "step": 8867500
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 3.7170032593090036e-05,
+      "loss": 2.1482,
+      "step": 8868000
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 3.716931184003335e-05,
+      "loss": 2.1318,
+      "step": 8868500
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 3.7168588192386074e-05,
+      "loss": 2.1187,
+      "step": 8869000
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 3.7167864544738796e-05,
+      "loss": 2.133,
+      "step": 8869500
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 3.716714089709152e-05,
+      "loss": 2.152,
+      "step": 8870000
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 3.716641724944424e-05,
+      "loss": 2.1549,
+      "step": 8870500
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 3.716569360179696e-05,
+      "loss": 2.152,
+      "step": 8871000
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 3.7164969954149685e-05,
+      "loss": 2.1777,
+      "step": 8871500
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 3.716424630650241e-05,
+      "loss": 2.1192,
+      "step": 8872000
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 3.716352265885513e-05,
+      "loss": 2.1401,
+      "step": 8872500
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 3.716279901120785e-05,
+      "loss": 2.1459,
+      "step": 8873000
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 3.716207536356058e-05,
+      "loss": 2.1714,
+      "step": 8873500
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 3.71613517159133e-05,
+      "loss": 2.1425,
+      "step": 8874000
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 3.7160628068266025e-05,
+      "loss": 2.1509,
+      "step": 8874500
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 3.715990586791405e-05,
+      "loss": 2.1224,
+      "step": 8875000
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 3.715918222026677e-05,
+      "loss": 2.1457,
+      "step": 8875500
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 3.715845857261949e-05,
+      "loss": 2.172,
+      "step": 8876000
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 3.7157734924972214e-05,
+      "loss": 2.1284,
+      "step": 8876500
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 3.7157011277324937e-05,
+      "loss": 2.1368,
+      "step": 8877000
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 3.715628762967766e-05,
+      "loss": 2.1268,
+      "step": 8877500
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 3.715556398203038e-05,
+      "loss": 2.1363,
+      "step": 8878000
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 3.71548403343831e-05,
+      "loss": 2.1381,
+      "step": 8878500
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 3.715411668673583e-05,
+      "loss": 2.1409,
+      "step": 8879000
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 3.7153393039088554e-05,
+      "loss": 2.1436,
+      "step": 8879500
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 3.715267083873657e-05,
+      "loss": 2.1651,
+      "step": 8880000
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 3.715194719108929e-05,
+      "loss": 2.165,
+      "step": 8880500
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 3.715122499073731e-05,
+      "loss": 2.1547,
+      "step": 8881000
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 3.715050134309003e-05,
+      "loss": 2.1316,
+      "step": 8881500
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 3.714977769544275e-05,
+      "loss": 2.1693,
+      "step": 8882000
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 3.714905404779548e-05,
+      "loss": 2.1481,
+      "step": 8882500
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 3.7148330400148204e-05,
+      "loss": 2.1481,
+      "step": 8883000
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 3.7147608199796226e-05,
+      "loss": 2.1496,
+      "step": 8883500
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 3.714688599944424e-05,
+      "loss": 2.1588,
+      "step": 8884000
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 3.7146162351796964e-05,
+      "loss": 2.1363,
+      "step": 8884500
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 3.7145438704149686e-05,
+      "loss": 2.1599,
+      "step": 8885000
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 3.714471505650241e-05,
+      "loss": 2.1307,
+      "step": 8885500
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 3.714399140885513e-05,
+      "loss": 2.1512,
+      "step": 8886000
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 3.714326776120786e-05,
+      "loss": 2.1565,
+      "step": 8886500
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 3.7142545560855875e-05,
+      "loss": 2.1398,
+      "step": 8887000
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 3.71418219132086e-05,
+      "loss": 2.1457,
+      "step": 8887500
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 3.714109826556132e-05,
+      "loss": 2.127,
+      "step": 8888000
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 3.714037461791404e-05,
+      "loss": 2.1493,
+      "step": 8888500
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 3.7139650970266764e-05,
+      "loss": 2.1459,
+      "step": 8889000
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 3.7138927322619486e-05,
+      "loss": 2.1224,
+      "step": 8889500
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 3.7138203674972215e-05,
+      "loss": 2.1247,
+      "step": 8890000
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 3.713748292191553e-05,
+      "loss": 2.1521,
+      "step": 8890500
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 3.7136760721563546e-05,
+      "loss": 2.1532,
+      "step": 8891000
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 3.713603707391627e-05,
+      "loss": 2.159,
+      "step": 8891500
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 3.713531342626899e-05,
+      "loss": 2.1714,
+      "step": 8892000
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 3.713458977862171e-05,
+      "loss": 2.1552,
+      "step": 8892500
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 3.7133866130974435e-05,
+      "loss": 2.1712,
+      "step": 8893000
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 3.713314248332716e-05,
+      "loss": 2.133,
+      "step": 8893500
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 3.713241883567988e-05,
+      "loss": 2.1477,
+      "step": 8894000
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 3.713169518803261e-05,
+      "loss": 2.1126,
+      "step": 8894500
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 3.713097154038533e-05,
+      "loss": 2.158,
+      "step": 8895000
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 3.713024789273805e-05,
+      "loss": 2.1387,
+      "step": 8895500
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 3.7129524245090775e-05,
+      "loss": 2.1262,
+      "step": 8896000
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 3.71288005974435e-05,
+      "loss": 2.1486,
+      "step": 8896500
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 3.712807694979622e-05,
+      "loss": 2.127,
+      "step": 8897000
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 3.712735330214895e-05,
+      "loss": 2.1321,
+      "step": 8897500
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 3.712662965450167e-05,
+      "loss": 2.1612,
+      "step": 8898000
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 3.712590745414969e-05,
+      "loss": 2.1608,
+      "step": 8898500
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 3.712518380650241e-05,
+      "loss": 2.1371,
+      "step": 8899000
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 3.712446015885513e-05,
+      "loss": 2.1416,
+      "step": 8899500
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 3.712373651120786e-05,
+      "loss": 2.1373,
+      "step": 8900000
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 3.712301286356058e-05,
+      "loss": 2.1311,
+      "step": 8900500
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 3.7122289215913305e-05,
+      "loss": 2.1498,
+      "step": 8901000
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 3.712156556826603e-05,
+      "loss": 2.1789,
+      "step": 8901500
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 3.712084192061875e-05,
+      "loss": 2.1502,
+      "step": 8902000
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 3.712011827297147e-05,
+      "loss": 2.1536,
+      "step": 8902500
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 3.7119394625324194e-05,
+      "loss": 2.1377,
+      "step": 8903000
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 3.7118670977676916e-05,
+      "loss": 2.1456,
+      "step": 8903500
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 3.711794733002964e-05,
+      "loss": 2.1476,
+      "step": 8904000
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 3.711722368238237e-05,
+      "loss": 2.1421,
+      "step": 8904500
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 3.711650003473509e-05,
+      "loss": 2.1247,
+      "step": 8905000
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 3.711577638708781e-05,
+      "loss": 2.1566,
+      "step": 8905500
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 3.7115052739440534e-05,
+      "loss": 2.1444,
+      "step": 8906000
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 3.711432909179326e-05,
+      "loss": 2.1558,
+      "step": 8906500
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 3.711360689144128e-05,
+      "loss": 2.1424,
+      "step": 8907000
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 3.7112883243794e-05,
+      "loss": 2.1381,
+      "step": 8907500
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 3.711215959614672e-05,
+      "loss": 2.1233,
+      "step": 8908000
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 3.711143739579474e-05,
+      "loss": 2.136,
+      "step": 8908500
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 3.711071374814746e-05,
+      "loss": 2.137,
+      "step": 8909000
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 3.710999010050018e-05,
+      "loss": 2.1273,
+      "step": 8909500
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 3.710926645285291e-05,
+      "loss": 2.1615,
+      "step": 8910000
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 3.7108542805205634e-05,
+      "loss": 2.1267,
+      "step": 8910500
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 3.7107819157558356e-05,
+      "loss": 2.1354,
+      "step": 8911000
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 3.710709550991108e-05,
+      "loss": 2.1571,
+      "step": 8911500
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 3.710637186226381e-05,
+      "loss": 2.1581,
+      "step": 8912000
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 3.710564821461653e-05,
+      "loss": 2.1469,
+      "step": 8912500
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 3.710492456696925e-05,
+      "loss": 2.1563,
+      "step": 8913000
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 3.7104200919321974e-05,
+      "loss": 2.1527,
+      "step": 8913500
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 3.710347871896999e-05,
+      "loss": 2.1488,
+      "step": 8914000
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 3.710275651861801e-05,
+      "loss": 2.1524,
+      "step": 8914500
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 3.710203431826603e-05,
+      "loss": 2.1635,
+      "step": 8915000
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 3.710131067061875e-05,
+      "loss": 2.1471,
+      "step": 8915500
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 3.710058702297147e-05,
+      "loss": 2.1429,
+      "step": 8916000
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 3.709986482261949e-05,
+      "loss": 2.1334,
+      "step": 8916500
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 3.709914117497221e-05,
+      "loss": 2.1579,
+      "step": 8917000
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 3.709841752732494e-05,
+      "loss": 2.1411,
+      "step": 8917500
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 3.7097695326972954e-05,
+      "loss": 2.1445,
+      "step": 8918000
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 3.709697167932568e-05,
+      "loss": 2.1207,
+      "step": 8918500
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 3.70962480316784e-05,
+      "loss": 2.1472,
+      "step": 8919000
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 3.709552438403112e-05,
+      "loss": 2.1468,
+      "step": 8919500
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 3.709480073638385e-05,
+      "loss": 2.1532,
+      "step": 8920000
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 3.709407708873657e-05,
+      "loss": 2.1575,
+      "step": 8920500
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 3.7093353441089295e-05,
+      "loss": 2.1434,
+      "step": 8921000
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 3.709262979344202e-05,
+      "loss": 2.1485,
+      "step": 8921500
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 3.709190614579474e-05,
+      "loss": 2.167,
+      "step": 8922000
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 3.709118394544276e-05,
+      "loss": 2.1712,
+      "step": 8922500
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 3.7090460297795484e-05,
+      "loss": 2.1275,
+      "step": 8923000
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 3.7089736650148206e-05,
+      "loss": 2.1571,
+      "step": 8923500
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 3.708901300250093e-05,
+      "loss": 2.1436,
+      "step": 8924000
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 3.708828935485365e-05,
+      "loss": 2.157,
+      "step": 8924500
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 3.708756570720637e-05,
+      "loss": 2.141,
+      "step": 8925000
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 3.7086842059559095e-05,
+      "loss": 2.1461,
+      "step": 8925500
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 3.708611841191182e-05,
+      "loss": 2.1437,
+      "step": 8926000
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 3.708539476426454e-05,
+      "loss": 2.1508,
+      "step": 8926500
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 3.708467111661727e-05,
+      "loss": 2.1196,
+      "step": 8927000
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 3.708394891626529e-05,
+      "loss": 2.1668,
+      "step": 8927500
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 3.708322526861801e-05,
+      "loss": 2.1362,
+      "step": 8928000
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 3.7082501620970735e-05,
+      "loss": 2.133,
+      "step": 8928500
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 3.708177797332346e-05,
+      "loss": 2.1493,
+      "step": 8929000
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 3.708105432567618e-05,
+      "loss": 2.1433,
+      "step": 8929500
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 3.70803306780289e-05,
+      "loss": 2.1386,
+      "step": 8930000
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 3.7079607030381624e-05,
+      "loss": 2.128,
+      "step": 8930500
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 3.7078883382734346e-05,
+      "loss": 2.1333,
+      "step": 8931000
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 3.707815973508707e-05,
+      "loss": 2.1296,
+      "step": 8931500
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 3.707743608743979e-05,
+      "loss": 2.1245,
+      "step": 8932000
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 3.707671243979251e-05,
+      "loss": 2.1415,
+      "step": 8932500
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 3.7075990239440535e-05,
+      "loss": 2.1393,
+      "step": 8933000
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 3.707526803908855e-05,
+      "loss": 2.1125,
+      "step": 8933500
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 3.7074545838736566e-05,
+      "loss": 2.1401,
+      "step": 8934000
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 3.707382363838459e-05,
+      "loss": 2.1541,
+      "step": 8934500
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 3.707309999073732e-05,
+      "loss": 2.1542,
+      "step": 8935000
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 3.707237634309004e-05,
+      "loss": 2.1553,
+      "step": 8935500
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 3.707165269544276e-05,
+      "loss": 2.1308,
+      "step": 8936000
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 3.7070929047795484e-05,
+      "loss": 2.1555,
+      "step": 8936500
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 3.707020540014821e-05,
+      "loss": 2.1467,
+      "step": 8937000
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 3.706948175250093e-05,
+      "loss": 2.122,
+      "step": 8937500
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 3.706875810485365e-05,
+      "loss": 2.145,
+      "step": 8938000
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 3.7068034457206373e-05,
+      "loss": 2.1369,
+      "step": 8938500
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 3.7067310809559096e-05,
+      "loss": 2.158,
+      "step": 8939000
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 3.706658716191182e-05,
+      "loss": 2.157,
+      "step": 8939500
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 3.706586351426454e-05,
+      "loss": 2.1417,
+      "step": 8940000
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 3.706513986661726e-05,
+      "loss": 2.1469,
+      "step": 8940500
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 3.7064417666265285e-05,
+      "loss": 2.1528,
+      "step": 8941000
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 3.706369401861801e-05,
+      "loss": 2.1294,
+      "step": 8941500
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 3.7062970370970736e-05,
+      "loss": 2.1534,
+      "step": 8942000
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 3.706224672332346e-05,
+      "loss": 2.1097,
+      "step": 8942500
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 3.706152307567618e-05,
+      "loss": 2.1597,
+      "step": 8943000
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 3.70607994280289e-05,
+      "loss": 2.1183,
+      "step": 8943500
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 3.7060075780381625e-05,
+      "loss": 2.1356,
+      "step": 8944000
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 3.705935213273435e-05,
+      "loss": 2.1415,
+      "step": 8944500
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 3.705862848508707e-05,
+      "loss": 2.1458,
+      "step": 8945000
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 3.705790483743979e-05,
+      "loss": 2.1247,
+      "step": 8945500
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 3.7057182637087814e-05,
+      "loss": 2.1485,
+      "step": 8946000
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 3.7056458989440536e-05,
+      "loss": 2.1638,
+      "step": 8946500
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 3.705573534179326e-05,
+      "loss": 2.133,
+      "step": 8947000
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 3.705501169414598e-05,
+      "loss": 2.1538,
+      "step": 8947500
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 3.70542880464987e-05,
+      "loss": 2.1604,
+      "step": 8948000
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 3.7053564398851425e-05,
+      "loss": 2.1223,
+      "step": 8948500
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 3.7052840751204154e-05,
+      "loss": 2.1594,
+      "step": 8949000
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 3.7052117103556876e-05,
+      "loss": 2.1586,
+      "step": 8949500
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 3.70513934559096e-05,
+      "loss": 2.1465,
+      "step": 8950000
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 3.7050671255557614e-05,
+      "loss": 2.1271,
+      "step": 8950500
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 3.704994760791034e-05,
+      "loss": 2.1594,
+      "step": 8951000
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 3.7049223960263065e-05,
+      "loss": 2.1451,
+      "step": 8951500
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 3.704850031261579e-05,
+      "loss": 2.1443,
+      "step": 8952000
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 3.704777666496851e-05,
+      "loss": 2.1507,
+      "step": 8952500
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 3.704705301732123e-05,
+      "loss": 2.156,
+      "step": 8953000
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 3.704633226426454e-05,
+      "loss": 2.1362,
+      "step": 8953500
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 3.704560861661727e-05,
+      "loss": 2.1435,
+      "step": 8954000
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 3.704488496896999e-05,
+      "loss": 2.142,
+      "step": 8954500
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 3.7044161321322714e-05,
+      "loss": 2.1775,
+      "step": 8955000
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 3.704343767367544e-05,
+      "loss": 2.1431,
+      "step": 8955500
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 3.704271402602816e-05,
+      "loss": 2.125,
+      "step": 8956000
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 3.704199037838088e-05,
+      "loss": 2.1473,
+      "step": 8956500
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 3.7041268178028903e-05,
+      "loss": 2.1452,
+      "step": 8957000
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 3.7040544530381626e-05,
+      "loss": 2.1516,
+      "step": 8957500
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 3.703982088273435e-05,
+      "loss": 2.1366,
+      "step": 8958000
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 3.703909868238237e-05,
+      "loss": 2.1659,
+      "step": 8958500
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 3.703837503473509e-05,
+      "loss": 2.1461,
+      "step": 8959000
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 3.7037651387087815e-05,
+      "loss": 2.1582,
+      "step": 8959500
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 3.703692773944054e-05,
+      "loss": 2.1329,
+      "step": 8960000
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 3.703620409179326e-05,
+      "loss": 2.1535,
+      "step": 8960500
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 3.703548044414598e-05,
+      "loss": 2.1443,
+      "step": 8961000
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 3.7034756796498704e-05,
+      "loss": 2.1299,
+      "step": 8961500
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 3.7034033148851426e-05,
+      "loss": 2.1598,
+      "step": 8962000
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 3.703330950120415e-05,
+      "loss": 2.1196,
+      "step": 8962500
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 3.703258585355687e-05,
+      "loss": 2.1332,
+      "step": 8963000
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 3.703186220590959e-05,
+      "loss": 2.1577,
+      "step": 8963500
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 3.703113855826232e-05,
+      "loss": 2.1334,
+      "step": 8964000
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 3.7030416357910344e-05,
+      "loss": 2.1257,
+      "step": 8964500
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 3.7029692710263066e-05,
+      "loss": 2.143,
+      "step": 8965000
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 3.702897050991108e-05,
+      "loss": 2.152,
+      "step": 8965500
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 3.7028246862263804e-05,
+      "loss": 2.1524,
+      "step": 8966000
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 3.7027523214616526e-05,
+      "loss": 2.1357,
+      "step": 8966500
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 3.702679956696925e-05,
+      "loss": 2.1274,
+      "step": 8967000
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 3.702607591932197e-05,
+      "loss": 2.1353,
+      "step": 8967500
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 3.702535371896999e-05,
+      "loss": 2.1502,
+      "step": 8968000
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 3.7024630071322715e-05,
+      "loss": 2.1368,
+      "step": 8968500
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 3.702390642367544e-05,
+      "loss": 2.1393,
+      "step": 8969000
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 3.702318277602816e-05,
+      "loss": 2.1497,
+      "step": 8969500
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 3.702245912838088e-05,
+      "loss": 2.1291,
+      "step": 8970000
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 3.7021735480733604e-05,
+      "loss": 2.163,
+      "step": 8970500
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 3.702101328038162e-05,
+      "loss": 2.1246,
+      "step": 8971000
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 3.702029108002964e-05,
+      "loss": 2.1488,
+      "step": 8971500
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 3.701956743238237e-05,
+      "loss": 2.1501,
+      "step": 8972000
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 3.701884378473509e-05,
+      "loss": 2.1442,
+      "step": 8972500
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 3.7018120137087815e-05,
+      "loss": 2.137,
+      "step": 8973000
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 3.701739648944054e-05,
+      "loss": 2.1569,
+      "step": 8973500
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 3.701667284179326e-05,
+      "loss": 2.1253,
+      "step": 8974000
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 3.701594919414598e-05,
+      "loss": 2.131,
+      "step": 8974500
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 3.7015225546498704e-05,
+      "loss": 2.1437,
+      "step": 8975000
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 3.701450189885143e-05,
+      "loss": 2.1385,
+      "step": 8975500
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 3.701377969849944e-05,
+      "loss": 2.1423,
+      "step": 8976000
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 3.7013057498147465e-05,
+      "loss": 2.1507,
+      "step": 8976500
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 3.701233385050019e-05,
+      "loss": 2.1345,
+      "step": 8977000
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 3.701161020285291e-05,
+      "loss": 2.1249,
+      "step": 8977500
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 3.701088655520563e-05,
+      "loss": 2.1573,
+      "step": 8978000
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 3.701016435485365e-05,
+      "loss": 2.1351,
+      "step": 8978500
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 3.700944070720637e-05,
+      "loss": 2.1653,
+      "step": 8979000
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 3.70087170595591e-05,
+      "loss": 2.1664,
+      "step": 8979500
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 3.700799341191182e-05,
+      "loss": 2.1596,
+      "step": 8980000
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 3.700726976426455e-05,
+      "loss": 2.1665,
+      "step": 8980500
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 3.700654611661727e-05,
+      "loss": 2.1455,
+      "step": 8981000
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 3.7005822468969994e-05,
+      "loss": 2.1675,
+      "step": 8981500
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 3.7005098821322716e-05,
+      "loss": 2.1458,
+      "step": 8982000
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.6646410012811101,
+      "eval_accuracy_mlm": 0.6288016755078222,
+      "eval_accuracy_nsp": 0.8570210776697694,
+      "eval_loss": 2.196295738220215,
+      "eval_runtime": 331.5653,
+      "eval_samples_per_second": 1316.139,
+      "eval_steps_per_second": 54.84,
+      "step": 8982272
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 3.700437517367544e-05,
+      "loss": 2.115,
+      "step": 8982500
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 3.700365152602816e-05,
+      "loss": 2.1196,
+      "step": 8983000
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 3.700292787838088e-05,
+      "loss": 2.1178,
+      "step": 8983500
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 3.70022056780289e-05,
+      "loss": 2.1168,
+      "step": 8984000
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 3.700148203038162e-05,
+      "loss": 2.1063,
+      "step": 8984500
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 3.700075838273435e-05,
+      "loss": 2.1416,
+      "step": 8985000
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 3.700003473508707e-05,
+      "loss": 2.1263,
+      "step": 8985500
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 3.6999311087439794e-05,
+      "loss": 2.1203,
+      "step": 8986000
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 3.699858888708781e-05,
+      "loss": 2.1251,
+      "step": 8986500
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 3.699786523944054e-05,
+      "loss": 2.1247,
+      "step": 8987000
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 3.699714159179326e-05,
+      "loss": 2.1123,
+      "step": 8987500
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 3.699641794414598e-05,
+      "loss": 2.1132,
+      "step": 8988000
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 3.6995694296498705e-05,
+      "loss": 2.1001,
+      "step": 8988500
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 3.699497064885143e-05,
+      "loss": 2.1507,
+      "step": 8989000
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 3.699424844849945e-05,
+      "loss": 2.1417,
+      "step": 8989500
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 3.6993526248147465e-05,
+      "loss": 2.1089,
+      "step": 8990000
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 3.699280260050019e-05,
+      "loss": 2.117,
+      "step": 8990500
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 3.699207895285291e-05,
+      "loss": 2.1257,
+      "step": 8991000
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 3.699135530520563e-05,
+      "loss": 2.1017,
+      "step": 8991500
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 3.6990631657558354e-05,
+      "loss": 2.1304,
+      "step": 8992000
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 3.6989908009911077e-05,
+      "loss": 2.1561,
+      "step": 8992500
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 3.69891858095591e-05,
+      "loss": 2.1537,
+      "step": 8993000
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 3.698846216191182e-05,
+      "loss": 2.1263,
+      "step": 8993500
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 3.698774140885513e-05,
+      "loss": 2.1262,
+      "step": 8994000
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 3.698701776120785e-05,
+      "loss": 2.1102,
+      "step": 8994500
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 3.698629411356058e-05,
+      "loss": 2.1301,
+      "step": 8995000
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 3.6985570465913303e-05,
+      "loss": 2.1059,
+      "step": 8995500
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 3.6984846818266026e-05,
+      "loss": 2.1171,
+      "step": 8996000
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 3.698412317061875e-05,
+      "loss": 2.1184,
+      "step": 8996500
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 3.698339952297148e-05,
+      "loss": 2.1239,
+      "step": 8997000
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 3.69826758753242e-05,
+      "loss": 2.1337,
+      "step": 8997500
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 3.698195222767692e-05,
+      "loss": 2.1337,
+      "step": 8998000
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 3.6981228580029644e-05,
+      "loss": 2.1263,
+      "step": 8998500
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 3.6980504932382366e-05,
+      "loss": 2.1372,
+      "step": 8999000
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 3.697978128473509e-05,
+      "loss": 2.1075,
+      "step": 8999500
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 3.697905763708781e-05,
+      "loss": 2.1373,
+      "step": 9000000
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 3.697833398944053e-05,
+      "loss": 2.1132,
+      "step": 9000500
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 3.6977610341793255e-05,
+      "loss": 2.1306,
+      "step": 9001000
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 3.697688669414598e-05,
+      "loss": 2.1407,
+      "step": 9001500
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 3.6976164493794e-05,
+      "loss": 2.162,
+      "step": 9002000
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 3.697544229344202e-05,
+      "loss": 2.1394,
+      "step": 9002500
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 3.6974718645794744e-05,
+      "loss": 2.1344,
+      "step": 9003000
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 3.6973994998147466e-05,
+      "loss": 2.1266,
+      "step": 9003500
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 3.697327135050019e-05,
+      "loss": 2.1385,
+      "step": 9004000
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 3.697254770285291e-05,
+      "loss": 2.152,
+      "step": 9004500
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 3.697182405520563e-05,
+      "loss": 2.137,
+      "step": 9005000
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 3.6971100407558355e-05,
+      "loss": 2.1074,
+      "step": 9005500
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 3.697037675991108e-05,
+      "loss": 2.1188,
+      "step": 9006000
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 3.69696531122638e-05,
+      "loss": 2.1222,
+      "step": 9006500
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 3.696893091191182e-05,
+      "loss": 2.1264,
+      "step": 9007000
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 3.696820871155984e-05,
+      "loss": 2.1113,
+      "step": 9007500
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 3.696748506391256e-05,
+      "loss": 2.1249,
+      "step": 9008000
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 3.696676141626528e-05,
+      "loss": 2.1115,
+      "step": 9008500
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 3.6966037768618004e-05,
+      "loss": 2.1572,
+      "step": 9009000
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 3.696531412097073e-05,
+      "loss": 2.121,
+      "step": 9009500
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 3.696459192061875e-05,
+      "loss": 2.1272,
+      "step": 9010000
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 3.696386827297148e-05,
+      "loss": 2.1275,
+      "step": 9010500
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 3.69631446253242e-05,
+      "loss": 2.1338,
+      "step": 9011000
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 3.696242097767692e-05,
+      "loss": 2.1312,
+      "step": 9011500
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 3.6961697330029644e-05,
+      "loss": 2.1388,
+      "step": 9012000
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 3.696097368238237e-05,
+      "loss": 2.1167,
+      "step": 9012500
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 3.696025003473509e-05,
+      "loss": 2.1326,
+      "step": 9013000
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 3.695952638708781e-05,
+      "loss": 2.1263,
+      "step": 9013500
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 3.695880418673583e-05,
+      "loss": 2.1322,
+      "step": 9014000
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 3.695808053908855e-05,
+      "loss": 2.1408,
+      "step": 9014500
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 3.695735689144128e-05,
+      "loss": 2.1419,
+      "step": 9015000
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 3.6956634691089293e-05,
+      "loss": 2.1355,
+      "step": 9015500
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 3.6955911043442016e-05,
+      "loss": 2.1144,
+      "step": 9016000
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 3.695518739579474e-05,
+      "loss": 2.1168,
+      "step": 9016500
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 3.6954465195442753e-05,
+      "loss": 2.1385,
+      "step": 9017000
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 3.695374154779548e-05,
+      "loss": 2.1045,
+      "step": 9017500
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 3.6953017900148205e-05,
+      "loss": 2.1059,
+      "step": 9018000
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 3.695229425250093e-05,
+      "loss": 2.1003,
+      "step": 9018500
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 3.695157060485365e-05,
+      "loss": 2.1116,
+      "step": 9019000
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 3.695084695720638e-05,
+      "loss": 2.1263,
+      "step": 9019500
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 3.69501233095591e-05,
+      "loss": 2.1269,
+      "step": 9020000
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 3.694939966191182e-05,
+      "loss": 2.1252,
+      "step": 9020500
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 3.6948676014264545e-05,
+      "loss": 2.132,
+      "step": 9021000
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 3.694795236661727e-05,
+      "loss": 2.1331,
+      "step": 9021500
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 3.694723016626528e-05,
+      "loss": 2.1381,
+      "step": 9022000
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 3.6946506518618005e-05,
+      "loss": 2.118,
+      "step": 9022500
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 3.694578287097073e-05,
+      "loss": 2.1239,
+      "step": 9023000
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 3.694505922332345e-05,
+      "loss": 2.1418,
+      "step": 9023500
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 3.694433557567618e-05,
+      "loss": 2.1505,
+      "step": 9024000
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 3.69436119280289e-05,
+      "loss": 2.1188,
+      "step": 9024500
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 3.694288972767692e-05,
+      "loss": 2.1366,
+      "step": 9025000
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 3.6942166080029645e-05,
+      "loss": 2.1231,
+      "step": 9025500
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 3.694144243238237e-05,
+      "loss": 2.1257,
+      "step": 9026000
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 3.694071878473509e-05,
+      "loss": 2.1105,
+      "step": 9026500
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 3.693999513708781e-05,
+      "loss": 2.1502,
+      "step": 9027000
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 3.6939271489440534e-05,
+      "loss": 2.1258,
+      "step": 9027500
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 3.6938549289088556e-05,
+      "loss": 2.134,
+      "step": 9028000
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 3.693782564144128e-05,
+      "loss": 2.1355,
+      "step": 9028500
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 3.6937101993794e-05,
+      "loss": 2.1275,
+      "step": 9029000
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 3.693637834614672e-05,
+      "loss": 2.1374,
+      "step": 9029500
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 3.6935654698499445e-05,
+      "loss": 2.1248,
+      "step": 9030000
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 3.693493105085217e-05,
+      "loss": 2.1246,
+      "step": 9030500
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 3.693420740320489e-05,
+      "loss": 2.1231,
+      "step": 9031000
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 3.693348375555761e-05,
+      "loss": 2.1273,
+      "step": 9031500
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 3.6932761555205634e-05,
+      "loss": 2.1419,
+      "step": 9032000
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 3.693203790755836e-05,
+      "loss": 2.1537,
+      "step": 9032500
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 3.693131425991108e-05,
+      "loss": 2.1194,
+      "step": 9033000
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 3.693059061226381e-05,
+      "loss": 2.1415,
+      "step": 9033500
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 3.692986696461653e-05,
+      "loss": 2.119,
+      "step": 9034000
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 3.692914331696925e-05,
+      "loss": 2.1363,
+      "step": 9034500
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 3.6928419669321975e-05,
+      "loss": 2.1507,
+      "step": 9035000
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 3.69276960216747e-05,
+      "loss": 2.1318,
+      "step": 9035500
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 3.692697237402742e-05,
+      "loss": 2.1336,
+      "step": 9036000
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 3.6926250173675435e-05,
+      "loss": 2.1171,
+      "step": 9036500
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 3.692552652602816e-05,
+      "loss": 2.1344,
+      "step": 9037000
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 3.692480432567618e-05,
+      "loss": 2.1271,
+      "step": 9037500
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 3.69240806780289e-05,
+      "loss": 2.1277,
+      "step": 9038000
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 3.6923357030381624e-05,
+      "loss": 2.1138,
+      "step": 9038500
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 3.6922633382734346e-05,
+      "loss": 2.1341,
+      "step": 9039000
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 3.6921909735087075e-05,
+      "loss": 2.1369,
+      "step": 9039500
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 3.69211860874398e-05,
+      "loss": 2.1049,
+      "step": 9040000
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 3.692046243979252e-05,
+      "loss": 2.133,
+      "step": 9040500
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 3.6919740239440535e-05,
+      "loss": 2.1326,
+      "step": 9041000
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 3.691901659179326e-05,
+      "loss": 2.1246,
+      "step": 9041500
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 3.691829294414598e-05,
+      "loss": 2.1609,
+      "step": 9042000
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 3.691756929649871e-05,
+      "loss": 2.135,
+      "step": 9042500
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 3.691684564885143e-05,
+      "loss": 2.1292,
+      "step": 9043000
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 3.6916123448499446e-05,
+      "loss": 2.1372,
+      "step": 9043500
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 3.691539980085217e-05,
+      "loss": 2.1319,
+      "step": 9044000
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 3.691467615320489e-05,
+      "loss": 2.1338,
+      "step": 9044500
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 3.691395250555761e-05,
+      "loss": 2.1113,
+      "step": 9045000
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 3.6913228857910335e-05,
+      "loss": 2.1211,
+      "step": 9045500
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 3.691250665755836e-05,
+      "loss": 2.1096,
+      "step": 9046000
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 3.691178300991108e-05,
+      "loss": 2.1257,
+      "step": 9046500
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 3.691105936226381e-05,
+      "loss": 2.1303,
+      "step": 9047000
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 3.691033571461653e-05,
+      "loss": 2.1123,
+      "step": 9047500
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 3.6909613514264546e-05,
+      "loss": 2.1369,
+      "step": 9048000
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 3.690888986661727e-05,
+      "loss": 2.1272,
+      "step": 9048500
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 3.690816621896999e-05,
+      "loss": 2.1053,
+      "step": 9049000
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 3.690744257132271e-05,
+      "loss": 2.1362,
+      "step": 9049500
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 3.6906718923675435e-05,
+      "loss": 2.1486,
+      "step": 9050000
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 3.690599527602816e-05,
+      "loss": 2.1572,
+      "step": 9050500
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 3.690527162838088e-05,
+      "loss": 2.1528,
+      "step": 9051000
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 3.690454798073361e-05,
+      "loss": 2.1333,
+      "step": 9051500
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 3.690382433308633e-05,
+      "loss": 2.1269,
+      "step": 9052000
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 3.690310068543905e-05,
+      "loss": 2.1424,
+      "step": 9052500
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 3.6902377037791776e-05,
+      "loss": 2.1244,
+      "step": 9053000
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 3.69016533901445e-05,
+      "loss": 2.1345,
+      "step": 9053500
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 3.690092974249723e-05,
+      "loss": 2.1417,
+      "step": 9054000
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 3.690020609484995e-05,
+      "loss": 2.1199,
+      "step": 9054500
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 3.689948244720267e-05,
+      "loss": 2.1327,
+      "step": 9055000
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 3.6898758799555394e-05,
+      "loss": 2.1441,
+      "step": 9055500
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 3.689803804649871e-05,
+      "loss": 2.1296,
+      "step": 9056000
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 3.689731439885143e-05,
+      "loss": 2.1436,
+      "step": 9056500
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 3.689659219849945e-05,
+      "loss": 2.1082,
+      "step": 9057000
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 3.689586855085217e-05,
+      "loss": 2.1487,
+      "step": 9057500
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 3.689514490320489e-05,
+      "loss": 2.1502,
+      "step": 9058000
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 3.6894421255557614e-05,
+      "loss": 2.1373,
+      "step": 9058500
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 3.6893697607910336e-05,
+      "loss": 2.1291,
+      "step": 9059000
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 3.689297396026306e-05,
+      "loss": 2.1238,
+      "step": 9059500
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 3.689225031261578e-05,
+      "loss": 2.1491,
+      "step": 9060000
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 3.689152666496851e-05,
+      "loss": 2.1268,
+      "step": 9060500
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 3.689080301732123e-05,
+      "loss": 2.1219,
+      "step": 9061000
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 3.689007936967396e-05,
+      "loss": 2.134,
+      "step": 9061500
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 3.6889357169321976e-05,
+      "loss": 2.1225,
+      "step": 9062000
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 3.68886335216747e-05,
+      "loss": 2.1288,
+      "step": 9062500
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 3.688790987402742e-05,
+      "loss": 2.1168,
+      "step": 9063000
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 3.688718622638014e-05,
+      "loss": 2.1528,
+      "step": 9063500
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 3.6886462578732865e-05,
+      "loss": 2.1139,
+      "step": 9064000
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 3.688574037838089e-05,
+      "loss": 2.1238,
+      "step": 9064500
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 3.688501673073361e-05,
+      "loss": 2.1232,
+      "step": 9065000
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 3.6884294530381625e-05,
+      "loss": 2.1337,
+      "step": 9065500
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 3.688357233002964e-05,
+      "loss": 2.1341,
+      "step": 9066000
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 3.688284868238236e-05,
+      "loss": 2.1463,
+      "step": 9066500
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 3.6882125034735085e-05,
+      "loss": 2.1204,
+      "step": 9067000
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 3.688140138708781e-05,
+      "loss": 2.1451,
+      "step": 9067500
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 3.6880677739440537e-05,
+      "loss": 2.1229,
+      "step": 9068000
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 3.687995409179326e-05,
+      "loss": 2.1478,
+      "step": 9068500
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 3.687923044414598e-05,
+      "loss": 2.131,
+      "step": 9069000
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 3.687850679649871e-05,
+      "loss": 2.1204,
+      "step": 9069500
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 3.6877784596146726e-05,
+      "loss": 2.1131,
+      "step": 9070000
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 3.687706094849945e-05,
+      "loss": 2.1423,
+      "step": 9070500
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 3.687633874814746e-05,
+      "loss": 2.1051,
+      "step": 9071000
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 3.6875616547795486e-05,
+      "loss": 2.133,
+      "step": 9071500
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 3.687489290014821e-05,
+      "loss": 2.1328,
+      "step": 9072000
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 3.687416925250093e-05,
+      "loss": 2.144,
+      "step": 9072500
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 3.687344560485365e-05,
+      "loss": 2.1069,
+      "step": 9073000
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 3.6872721957206375e-05,
+      "loss": 2.1227,
+      "step": 9073500
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 3.68719983095591e-05,
+      "loss": 2.1207,
+      "step": 9074000
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 3.687127466191182e-05,
+      "loss": 2.1541,
+      "step": 9074500
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 3.687055101426454e-05,
+      "loss": 2.1103,
+      "step": 9075000
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 3.6869827366617264e-05,
+      "loss": 2.1335,
+      "step": 9075500
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 3.6869103718969986e-05,
+      "loss": 2.1551,
+      "step": 9076000
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 3.686838007132271e-05,
+      "loss": 2.1337,
+      "step": 9076500
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 3.686765642367544e-05,
+      "loss": 2.1507,
+      "step": 9077000
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 3.686693277602816e-05,
+      "loss": 2.1285,
+      "step": 9077500
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 3.686620912838089e-05,
+      "loss": 2.1198,
+      "step": 9078000
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 3.686548548073361e-05,
+      "loss": 2.1481,
+      "step": 9078500
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 3.6864763280381626e-05,
+      "loss": 2.1511,
+      "step": 9079000
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 3.686403963273435e-05,
+      "loss": 2.1319,
+      "step": 9079500
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 3.686331598508707e-05,
+      "loss": 2.1197,
+      "step": 9080000
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 3.686259233743979e-05,
+      "loss": 2.1341,
+      "step": 9080500
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 3.6861868689792515e-05,
+      "loss": 2.1118,
+      "step": 9081000
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 3.686114504214524e-05,
+      "loss": 2.0959,
+      "step": 9081500
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 3.686042139449796e-05,
+      "loss": 2.1312,
+      "step": 9082000
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 3.685969774685069e-05,
+      "loss": 2.1285,
+      "step": 9082500
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 3.685897409920341e-05,
+      "loss": 2.1444,
+      "step": 9083000
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 3.685825045155613e-05,
+      "loss": 2.1235,
+      "step": 9083500
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 3.685752680390886e-05,
+      "loss": 2.1134,
+      "step": 9084000
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 3.6856803156261584e-05,
+      "loss": 2.1482,
+      "step": 9084500
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 3.68560809559096e-05,
+      "loss": 2.1447,
+      "step": 9085000
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 3.685535730826232e-05,
+      "loss": 2.1316,
+      "step": 9085500
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 3.6854633660615044e-05,
+      "loss": 2.1416,
+      "step": 9086000
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 3.6853910012967766e-05,
+      "loss": 2.123,
+      "step": 9086500
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 3.685318636532049e-05,
+      "loss": 2.1418,
+      "step": 9087000
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 3.685246271767321e-05,
+      "loss": 2.1396,
+      "step": 9087500
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 3.685173907002594e-05,
+      "loss": 2.1622,
+      "step": 9088000
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 3.6851016869673955e-05,
+      "loss": 2.1469,
+      "step": 9088500
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 3.685029322202668e-05,
+      "loss": 2.1323,
+      "step": 9089000
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 3.68495695743794e-05,
+      "loss": 2.1256,
+      "step": 9089500
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 3.684884592673212e-05,
+      "loss": 2.1258,
+      "step": 9090000
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 3.6848122279084844e-05,
+      "loss": 2.1376,
+      "step": 9090500
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 3.684739863143757e-05,
+      "loss": 2.1762,
+      "step": 9091000
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 3.684667643108559e-05,
+      "loss": 2.1337,
+      "step": 9091500
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 3.684595278343831e-05,
+      "loss": 2.1196,
+      "step": 9092000
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 3.6845230583086334e-05,
+      "loss": 2.1242,
+      "step": 9092500
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 3.684450838273435e-05,
+      "loss": 2.1347,
+      "step": 9093000
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 3.6843786182382365e-05,
+      "loss": 2.1126,
+      "step": 9093500
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 3.684306253473509e-05,
+      "loss": 2.153,
+      "step": 9094000
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 3.6842338887087816e-05,
+      "loss": 2.1585,
+      "step": 9094500
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 3.684161523944054e-05,
+      "loss": 2.1367,
+      "step": 9095000
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 3.684089159179326e-05,
+      "loss": 2.1191,
+      "step": 9095500
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 3.684016794414598e-05,
+      "loss": 2.1624,
+      "step": 9096000
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 3.6839445743794e-05,
+      "loss": 2.1598,
+      "step": 9096500
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 3.683872209614672e-05,
+      "loss": 2.1327,
+      "step": 9097000
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 3.683799844849944e-05,
+      "loss": 2.1274,
+      "step": 9097500
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 3.6837274800852165e-05,
+      "loss": 2.1354,
+      "step": 9098000
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 3.683655115320489e-05,
+      "loss": 2.1448,
+      "step": 9098500
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 3.6835827505557616e-05,
+      "loss": 2.1467,
+      "step": 9099000
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 3.683510385791034e-05,
+      "loss": 2.1147,
+      "step": 9099500
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 3.683438021026307e-05,
+      "loss": 2.152,
+      "step": 9100000
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 3.683365656261579e-05,
+      "loss": 2.137,
+      "step": 9100500
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 3.683293291496851e-05,
+      "loss": 2.1129,
+      "step": 9101000
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 3.6832209267321234e-05,
+      "loss": 2.1538,
+      "step": 9101500
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 3.6831485619673956e-05,
+      "loss": 2.1512,
+      "step": 9102000
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 3.683076197202668e-05,
+      "loss": 2.1229,
+      "step": 9102500
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 3.68300383243794e-05,
+      "loss": 2.1289,
+      "step": 9103000
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 3.682931467673212e-05,
+      "loss": 2.1404,
+      "step": 9103500
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 3.682859247638014e-05,
+      "loss": 2.1519,
+      "step": 9104000
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 3.682787027602816e-05,
+      "loss": 2.125,
+      "step": 9104500
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 3.682714662838088e-05,
+      "loss": 2.1236,
+      "step": 9105000
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 3.6826422980733605e-05,
+      "loss": 2.1367,
+      "step": 9105500
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 3.682569933308633e-05,
+      "loss": 2.1129,
+      "step": 9106000
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 3.6824975685439057e-05,
+      "loss": 2.136,
+      "step": 9106500
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 3.682425203779178e-05,
+      "loss": 2.1213,
+      "step": 9107000
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 3.68235283901445e-05,
+      "loss": 2.1441,
+      "step": 9107500
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 3.682280474249722e-05,
+      "loss": 2.1304,
+      "step": 9108000
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 3.6822081094849945e-05,
+      "loss": 2.1307,
+      "step": 9108500
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 3.682135889449797e-05,
+      "loss": 2.1587,
+      "step": 9109000
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 3.682063669414598e-05,
+      "loss": 2.1129,
+      "step": 9109500
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 3.6819913046498706e-05,
+      "loss": 2.1541,
+      "step": 9110000
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 3.681918939885143e-05,
+      "loss": 2.1313,
+      "step": 9110500
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 3.681846575120415e-05,
+      "loss": 2.1533,
+      "step": 9111000
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 3.681774210355687e-05,
+      "loss": 2.1137,
+      "step": 9111500
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 3.6817018455909595e-05,
+      "loss": 2.1638,
+      "step": 9112000
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 3.681629480826232e-05,
+      "loss": 2.1494,
+      "step": 9112500
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 3.681557116061504e-05,
+      "loss": 2.1222,
+      "step": 9113000
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 3.681484751296777e-05,
+      "loss": 2.1166,
+      "step": 9113500
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 3.681412386532049e-05,
+      "loss": 2.1556,
+      "step": 9114000
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 3.681340021767322e-05,
+      "loss": 2.1594,
+      "step": 9114500
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 3.681267657002594e-05,
+      "loss": 2.1309,
+      "step": 9115000
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 3.6811952922378664e-05,
+      "loss": 2.1399,
+      "step": 9115500
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 3.681123072202668e-05,
+      "loss": 2.1268,
+      "step": 9116000
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 3.68105070743794e-05,
+      "loss": 2.1392,
+      "step": 9116500
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 3.6809783426732124e-05,
+      "loss": 2.1116,
+      "step": 9117000
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 3.6809059779084846e-05,
+      "loss": 2.1291,
+      "step": 9117500
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 3.680833613143757e-05,
+      "loss": 2.1357,
+      "step": 9118000
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 3.680761248379029e-05,
+      "loss": 2.1135,
+      "step": 9118500
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 3.680689028343831e-05,
+      "loss": 2.1408,
+      "step": 9119000
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 3.680616808308633e-05,
+      "loss": 2.1234,
+      "step": 9119500
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 3.680544443543905e-05,
+      "loss": 2.114,
+      "step": 9120000
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 3.680472078779177e-05,
+      "loss": 2.1281,
+      "step": 9120500
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 3.6803997140144495e-05,
+      "loss": 2.128,
+      "step": 9121000
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 3.6803273492497224e-05,
+      "loss": 2.1517,
+      "step": 9121500
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 3.6802549844849946e-05,
+      "loss": 2.1459,
+      "step": 9122000
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 3.680182619720267e-05,
+      "loss": 2.1253,
+      "step": 9122500
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 3.680110254955539e-05,
+      "loss": 2.1332,
+      "step": 9123000
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 3.680037890190812e-05,
+      "loss": 2.1396,
+      "step": 9123500
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 3.6799656701556135e-05,
+      "loss": 2.1593,
+      "step": 9124000
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 3.679893450120415e-05,
+      "loss": 2.1489,
+      "step": 9124500
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 3.679821085355687e-05,
+      "loss": 2.1372,
+      "step": 9125000
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 3.6797487205909595e-05,
+      "loss": 2.1567,
+      "step": 9125500
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 3.679676355826232e-05,
+      "loss": 2.1535,
+      "step": 9126000
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 3.679604135791034e-05,
+      "loss": 2.1742,
+      "step": 9126500
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 3.6795319157558355e-05,
+      "loss": 2.1456,
+      "step": 9127000
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 3.679459550991108e-05,
+      "loss": 2.1333,
+      "step": 9127500
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 3.67938718622638e-05,
+      "loss": 2.151,
+      "step": 9128000
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 3.679314821461652e-05,
+      "loss": 2.1298,
+      "step": 9128500
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 3.6792424566969244e-05,
+      "loss": 2.1186,
+      "step": 9129000
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 3.6791700919321973e-05,
+      "loss": 2.1293,
+      "step": 9129500
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 3.6790977271674696e-05,
+      "loss": 2.1047,
+      "step": 9130000
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 3.679025362402742e-05,
+      "loss": 2.1281,
+      "step": 9130500
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 3.678953142367544e-05,
+      "loss": 2.1327,
+      "step": 9131000
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 3.678880777602816e-05,
+      "loss": 2.1541,
+      "step": 9131500
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 3.678808557567618e-05,
+      "loss": 2.1098,
+      "step": 9132000
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 3.67873619280289e-05,
+      "loss": 2.1419,
+      "step": 9132500
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 3.678663828038162e-05,
+      "loss": 2.1375,
+      "step": 9133000
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 3.6785914632734345e-05,
+      "loss": 2.1292,
+      "step": 9133500
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 3.678519098508707e-05,
+      "loss": 2.128,
+      "step": 9134000
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 3.6784467337439796e-05,
+      "loss": 2.1289,
+      "step": 9134500
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 3.678374368979252e-05,
+      "loss": 2.1182,
+      "step": 9135000
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 3.678302004214524e-05,
+      "loss": 2.1254,
+      "step": 9135500
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 3.678229639449796e-05,
+      "loss": 2.1327,
+      "step": 9136000
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 3.678157564144127e-05,
+      "loss": 2.1531,
+      "step": 9136500
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 3.6780851993794e-05,
+      "loss": 2.1163,
+      "step": 9137000
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 3.678012834614672e-05,
+      "loss": 2.1233,
+      "step": 9137500
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 3.6779404698499445e-05,
+      "loss": 2.1228,
+      "step": 9138000
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 3.6778681050852174e-05,
+      "loss": 2.133,
+      "step": 9138500
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 3.6777957403204896e-05,
+      "loss": 2.1509,
+      "step": 9139000
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 3.677723520285291e-05,
+      "loss": 2.1352,
+      "step": 9139500
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 3.6776511555205634e-05,
+      "loss": 2.1335,
+      "step": 9140000
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 3.6775787907558356e-05,
+      "loss": 2.1556,
+      "step": 9140500
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 3.677506570720637e-05,
+      "loss": 2.167,
+      "step": 9141000
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 3.6774343506854394e-05,
+      "loss": 2.1393,
+      "step": 9141500
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 3.6773619859207116e-05,
+      "loss": 2.1128,
+      "step": 9142000
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 3.677289621155984e-05,
+      "loss": 2.1315,
+      "step": 9142500
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 3.677217256391256e-05,
+      "loss": 2.1425,
+      "step": 9143000
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 3.677144891626528e-05,
+      "loss": 2.1512,
+      "step": 9143500
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 3.6770725268618005e-05,
+      "loss": 2.125,
+      "step": 9144000
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 3.6770001620970734e-05,
+      "loss": 2.1214,
+      "step": 9144500
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 3.6769277973323457e-05,
+      "loss": 2.1471,
+      "step": 9145000
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 3.676855432567618e-05,
+      "loss": 2.1413,
+      "step": 9145500
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 3.67678306780289e-05,
+      "loss": 2.1458,
+      "step": 9146000
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 3.676710703038162e-05,
+      "loss": 2.1184,
+      "step": 9146500
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 3.6766383382734345e-05,
+      "loss": 2.1495,
+      "step": 9147000
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 3.6765659735087075e-05,
+      "loss": 2.141,
+      "step": 9147500
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 3.67649360874398e-05,
+      "loss": 2.1262,
+      "step": 9148000
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 3.676421243979252e-05,
+      "loss": 2.1389,
+      "step": 9148500
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 3.676348879214524e-05,
+      "loss": 2.1343,
+      "step": 9149000
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 3.6762765144497963e-05,
+      "loss": 2.1418,
+      "step": 9149500
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 3.6762041496850686e-05,
+      "loss": 2.1259,
+      "step": 9150000
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 3.676131784920341e-05,
+      "loss": 2.1004,
+      "step": 9150500
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 3.6760595648851423e-05,
+      "loss": 2.1433,
+      "step": 9151000
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 3.675987200120415e-05,
+      "loss": 2.1264,
+      "step": 9151500
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 3.6759148353556875e-05,
+      "loss": 2.1295,
+      "step": 9152000
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 3.67584261532049e-05,
+      "loss": 2.1401,
+      "step": 9152500
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 3.675770250555762e-05,
+      "loss": 2.1281,
+      "step": 9153000
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 3.675697885791034e-05,
+      "loss": 2.1347,
+      "step": 9153500
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 3.6756255210263064e-05,
+      "loss": 2.1333,
+      "step": 9154000
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 3.6755531562615786e-05,
+      "loss": 2.1514,
+      "step": 9154500
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 3.675480791496851e-05,
+      "loss": 2.1318,
+      "step": 9155000
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 3.675408426732123e-05,
+      "loss": 2.1404,
+      "step": 9155500
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 3.675336061967395e-05,
+      "loss": 2.1406,
+      "step": 9156000
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 3.6752636972026675e-05,
+      "loss": 2.1468,
+      "step": 9156500
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 3.67519133243794e-05,
+      "loss": 2.1264,
+      "step": 9157000
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 3.6751189676732126e-05,
+      "loss": 2.1622,
+      "step": 9157500
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 3.675046602908485e-05,
+      "loss": 2.1551,
+      "step": 9158000
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 3.6749743828732864e-05,
+      "loss": 2.1614,
+      "step": 9158500
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 3.674902018108559e-05,
+      "loss": 2.1525,
+      "step": 9159000
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 3.6748296533438315e-05,
+      "loss": 2.1326,
+      "step": 9159500
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 3.674757288579104e-05,
+      "loss": 2.1278,
+      "step": 9160000
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 3.674684923814376e-05,
+      "loss": 2.1561,
+      "step": 9160500
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 3.6746127037791775e-05,
+      "loss": 2.1438,
+      "step": 9161000
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 3.67454033901445e-05,
+      "loss": 2.1258,
+      "step": 9161500
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 3.6744679742497226e-05,
+      "loss": 2.1378,
+      "step": 9162000
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 3.674395609484995e-05,
+      "loss": 2.147,
+      "step": 9162500
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 3.674323244720267e-05,
+      "loss": 2.131,
+      "step": 9163000
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 3.674250879955539e-05,
+      "loss": 2.1427,
+      "step": 9163500
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 3.674178659920341e-05,
+      "loss": 2.1377,
+      "step": 9164000
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 3.674106295155613e-05,
+      "loss": 2.1482,
+      "step": 9164500
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 3.674033930390885e-05,
+      "loss": 2.139,
+      "step": 9165000
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 3.6739615656261575e-05,
+      "loss": 2.1478,
+      "step": 9165500
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 3.67388920086143e-05,
+      "loss": 2.1425,
+      "step": 9166000
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 3.673816836096703e-05,
+      "loss": 2.1377,
+      "step": 9166500
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 3.673744471331975e-05,
+      "loss": 2.1237,
+      "step": 9167000
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 3.673672106567248e-05,
+      "loss": 2.1322,
+      "step": 9167500
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 3.67359974180252e-05,
+      "loss": 2.1451,
+      "step": 9168000
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 3.673527377037792e-05,
+      "loss": 2.1698,
+      "step": 9168500
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 3.6734550122730645e-05,
+      "loss": 2.1015,
+      "step": 9169000
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 3.673382647508337e-05,
+      "loss": 2.1358,
+      "step": 9169500
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 3.673310282743609e-05,
+      "loss": 2.1305,
+      "step": 9170000
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 3.6732380627084105e-05,
+      "loss": 2.1278,
+      "step": 9170500
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 3.673165697943683e-05,
+      "loss": 2.1284,
+      "step": 9171000
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 3.673093333178955e-05,
+      "loss": 2.1518,
+      "step": 9171500
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 3.673021113143757e-05,
+      "loss": 2.1475,
+      "step": 9172000
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 3.6729487483790294e-05,
+      "loss": 2.1455,
+      "step": 9172500
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 3.6728763836143016e-05,
+      "loss": 2.1337,
+      "step": 9173000
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 3.6728040188495745e-05,
+      "loss": 2.1241,
+      "step": 9173500
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 3.672731654084847e-05,
+      "loss": 2.1295,
+      "step": 9174000
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 3.672659289320119e-05,
+      "loss": 2.1186,
+      "step": 9174500
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 3.672586924555391e-05,
+      "loss": 2.1272,
+      "step": 9175000
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 3.672514704520193e-05,
+      "loss": 2.1527,
+      "step": 9175500
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 3.672442339755465e-05,
+      "loss": 2.1433,
+      "step": 9176000
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 3.672369974990738e-05,
+      "loss": 2.126,
+      "step": 9176500
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 3.67229761022601e-05,
+      "loss": 2.1539,
+      "step": 9177000
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 3.672225245461282e-05,
+      "loss": 2.1526,
+      "step": 9177500
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 3.6721528806965545e-05,
+      "loss": 2.1414,
+      "step": 9178000
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 3.672080660661356e-05,
+      "loss": 2.1362,
+      "step": 9178500
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 3.6720084406261576e-05,
+      "loss": 2.1084,
+      "step": 9179000
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 3.67193607586143e-05,
+      "loss": 2.1167,
+      "step": 9179500
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 3.671863711096703e-05,
+      "loss": 2.1127,
+      "step": 9180000
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 3.671791346331975e-05,
+      "loss": 2.1119,
+      "step": 9180500
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 3.671718981567247e-05,
+      "loss": 2.1605,
+      "step": 9181000
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 3.67164661680252e-05,
+      "loss": 2.1275,
+      "step": 9181500
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 3.6715743967673216e-05,
+      "loss": 2.1273,
+      "step": 9182000
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 3.671502032002594e-05,
+      "loss": 2.1093,
+      "step": 9182500
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 3.671429667237866e-05,
+      "loss": 2.1511,
+      "step": 9183000
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 3.671357302473138e-05,
+      "loss": 2.1362,
+      "step": 9183500
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 3.6712849377084105e-05,
+      "loss": 2.1469,
+      "step": 9184000
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 3.671212572943683e-05,
+      "loss": 2.1289,
+      "step": 9184500
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 3.671140208178955e-05,
+      "loss": 2.1461,
+      "step": 9185000
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 3.671067843414228e-05,
+      "loss": 2.1371,
+      "step": 9185500
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 3.6709956233790294e-05,
+      "loss": 2.128,
+      "step": 9186000
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 3.670923258614302e-05,
+      "loss": 2.154,
+      "step": 9186500
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 3.670850893849574e-05,
+      "loss": 2.1334,
+      "step": 9187000
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 3.670778529084846e-05,
+      "loss": 2.1331,
+      "step": 9187500
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 3.670706309049648e-05,
+      "loss": 2.1125,
+      "step": 9188000
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 3.670634233743979e-05,
+      "loss": 2.1372,
+      "step": 9188500
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 3.670561868979252e-05,
+      "loss": 2.1458,
+      "step": 9189000
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 3.6704895042145244e-05,
+      "loss": 2.141,
+      "step": 9189500
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 3.6704171394497966e-05,
+      "loss": 2.1386,
+      "step": 9190000
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 3.670344774685069e-05,
+      "loss": 2.1422,
+      "step": 9190500
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 3.670272409920341e-05,
+      "loss": 2.1495,
+      "step": 9191000
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 3.670200045155613e-05,
+      "loss": 2.1428,
+      "step": 9191500
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 3.6701276803908855e-05,
+      "loss": 2.1358,
+      "step": 9192000
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 3.670055315626158e-05,
+      "loss": 2.1554,
+      "step": 9192500
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 3.6699829508614306e-05,
+      "loss": 2.1434,
+      "step": 9193000
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 3.669910586096703e-05,
+      "loss": 2.1222,
+      "step": 9193500
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 3.669838221331975e-05,
+      "loss": 2.1239,
+      "step": 9194000
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 3.669765856567247e-05,
+      "loss": 2.115,
+      "step": 9194500
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 3.669693636532049e-05,
+      "loss": 2.1451,
+      "step": 9195000
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 3.669621271767321e-05,
+      "loss": 2.125,
+      "step": 9195500
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 3.6695491964616526e-05,
+      "loss": 2.1508,
+      "step": 9196000
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 3.669476831696925e-05,
+      "loss": 2.1543,
+      "step": 9196500
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 3.669404466932198e-05,
+      "loss": 2.1166,
+      "step": 9197000
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 3.66933210216747e-05,
+      "loss": 2.1467,
+      "step": 9197500
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 3.669259737402742e-05,
+      "loss": 2.1466,
+      "step": 9198000
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 3.6691873726380144e-05,
+      "loss": 2.1286,
+      "step": 9198500
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 3.6691150078732866e-05,
+      "loss": 2.1323,
+      "step": 9199000
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 3.669042643108559e-05,
+      "loss": 2.1461,
+      "step": 9199500
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 3.668970278343831e-05,
+      "loss": 2.1402,
+      "step": 9200000
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 3.668897913579103e-05,
+      "loss": 2.1385,
+      "step": 9200500
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 3.6688255488143755e-05,
+      "loss": 2.1449,
+      "step": 9201000
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 3.668753184049648e-05,
+      "loss": 2.1374,
+      "step": 9201500
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 3.6686808192849206e-05,
+      "loss": 2.1375,
+      "step": 9202000
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 3.668608454520193e-05,
+      "loss": 2.1437,
+      "step": 9202500
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 3.6685362344849944e-05,
+      "loss": 2.1153,
+      "step": 9203000
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 3.6684638697202667e-05,
+      "loss": 2.1451,
+      "step": 9203500
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 3.6683915049555396e-05,
+      "loss": 2.1506,
+      "step": 9204000
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 3.668319140190812e-05,
+      "loss": 2.1622,
+      "step": 9204500
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 3.668246775426084e-05,
+      "loss": 2.1542,
+      "step": 9205000
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 3.668174410661356e-05,
+      "loss": 2.1223,
+      "step": 9205500
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 3.6681020458966284e-05,
+      "loss": 2.1261,
+      "step": 9206000
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 3.668029681131901e-05,
+      "loss": 2.1192,
+      "step": 9206500
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 3.667957316367173e-05,
+      "loss": 2.1366,
+      "step": 9207000
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 3.667885096331975e-05,
+      "loss": 2.1092,
+      "step": 9207500
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 3.6678127315672473e-05,
+      "loss": 2.1258,
+      "step": 9208000
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 3.667740511532049e-05,
+      "loss": 2.1426,
+      "step": 9208500
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 3.667668146767321e-05,
+      "loss": 2.1382,
+      "step": 9209000
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 3.6675957820025934e-05,
+      "loss": 2.1398,
+      "step": 9209500
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 3.6675234172378656e-05,
+      "loss": 2.137,
+      "step": 9210000
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 3.667451197202668e-05,
+      "loss": 2.1328,
+      "step": 9210500
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 3.6673789771674694e-05,
+      "loss": 2.1133,
+      "step": 9211000
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 3.667306612402742e-05,
+      "loss": 2.1502,
+      "step": 9211500
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 3.6672342476380145e-05,
+      "loss": 2.1196,
+      "step": 9212000
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 3.667161882873287e-05,
+      "loss": 2.1227,
+      "step": 9212500
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 3.667089518108559e-05,
+      "loss": 2.1302,
+      "step": 9213000
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 3.667017153343831e-05,
+      "loss": 2.1368,
+      "step": 9213500
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 3.6669447885791034e-05,
+      "loss": 2.1432,
+      "step": 9214000
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 3.6668724238143756e-05,
+      "loss": 2.1367,
+      "step": 9214500
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 3.6668000590496485e-05,
+      "loss": 2.1125,
+      "step": 9215000
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 3.666727694284921e-05,
+      "loss": 2.1507,
+      "step": 9215500
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 3.666655474249722e-05,
+      "loss": 2.1429,
+      "step": 9216000
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 3.6665831094849945e-05,
+      "loss": 2.1362,
+      "step": 9216500
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 3.666510744720267e-05,
+      "loss": 2.1175,
+      "step": 9217000
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 3.666438379955539e-05,
+      "loss": 2.1568,
+      "step": 9217500
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 3.666366015190811e-05,
+      "loss": 2.1417,
+      "step": 9218000
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 3.6662936504260834e-05,
+      "loss": 2.1523,
+      "step": 9218500
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 3.666221285661356e-05,
+      "loss": 2.1162,
+      "step": 9219000
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 3.6661489208966285e-05,
+      "loss": 2.1588,
+      "step": 9219500
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 3.666076556131901e-05,
+      "loss": 2.1314,
+      "step": 9220000
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 3.6660041913671737e-05,
+      "loss": 2.1179,
+      "step": 9220500
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 3.665931826602446e-05,
+      "loss": 2.1386,
+      "step": 9221000
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 3.665859461837718e-05,
+      "loss": 2.1212,
+      "step": 9221500
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 3.66578709707299e-05,
+      "loss": 2.1295,
+      "step": 9222000
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 3.665714877037792e-05,
+      "loss": 2.1387,
+      "step": 9222500
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 3.665642512273064e-05,
+      "loss": 2.1558,
+      "step": 9223000
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 3.665570147508336e-05,
+      "loss": 2.1516,
+      "step": 9223500
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 3.6654977827436085e-05,
+      "loss": 2.1343,
+      "step": 9224000
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 3.66542570743794e-05,
+      "loss": 2.1452,
+      "step": 9224500
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 3.665353342673212e-05,
+      "loss": 2.1307,
+      "step": 9225000
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 3.6652809779084846e-05,
+      "loss": 2.1329,
+      "step": 9225500
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 3.665208613143757e-05,
+      "loss": 2.1144,
+      "step": 9226000
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 3.6651365378380883e-05,
+      "loss": 2.1593,
+      "step": 9226500
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 3.665064173073361e-05,
+      "loss": 2.1296,
+      "step": 9227000
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 3.6649918083086335e-05,
+      "loss": 2.153,
+      "step": 9227500
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 3.664919443543906e-05,
+      "loss": 2.145,
+      "step": 9228000
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 3.664847078779178e-05,
+      "loss": 2.1672,
+      "step": 9228500
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 3.66477471401445e-05,
+      "loss": 2.1151,
+      "step": 9229000
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 3.6647023492497224e-05,
+      "loss": 2.1376,
+      "step": 9229500
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 3.6646299844849946e-05,
+      "loss": 2.1237,
+      "step": 9230000
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 3.664557619720267e-05,
+      "loss": 2.1373,
+      "step": 9230500
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 3.664485254955539e-05,
+      "loss": 2.1437,
+      "step": 9231000
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 3.664412890190811e-05,
+      "loss": 2.1523,
+      "step": 9231500
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 3.6643405254260835e-05,
+      "loss": 2.149,
+      "step": 9232000
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 3.664268160661356e-05,
+      "loss": 2.1558,
+      "step": 9232500
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 3.6641957958966286e-05,
+      "loss": 2.1288,
+      "step": 9233000
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 3.664123431131901e-05,
+      "loss": 2.165,
+      "step": 9233500
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 3.664051066367174e-05,
+      "loss": 2.1393,
+      "step": 9234000
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 3.663978701602446e-05,
+      "loss": 2.1356,
+      "step": 9234500
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 3.6639064815672475e-05,
+      "loss": 2.1422,
+      "step": 9235000
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 3.66383411680252e-05,
+      "loss": 2.1455,
+      "step": 9235500
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 3.663761896767321e-05,
+      "loss": 2.1487,
+      "step": 9236000
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 3.6636895320025935e-05,
+      "loss": 2.1423,
+      "step": 9236500
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 3.6636171672378664e-05,
+      "loss": 2.122,
+      "step": 9237000
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 3.6635448024731386e-05,
+      "loss": 2.156,
+      "step": 9237500
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 3.663472437708411e-05,
+      "loss": 2.1476,
+      "step": 9238000
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 3.663400072943683e-05,
+      "loss": 2.1299,
+      "step": 9238500
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 3.6633278529084846e-05,
+      "loss": 2.1625,
+      "step": 9239000
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 3.663255488143757e-05,
+      "loss": 2.1428,
+      "step": 9239500
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 3.663183123379029e-05,
+      "loss": 2.1355,
+      "step": 9240000
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 3.663110758614301e-05,
+      "loss": 2.1319,
+      "step": 9240500
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 3.6630385385791035e-05,
+      "loss": 2.1423,
+      "step": 9241000
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 3.6629661738143764e-05,
+      "loss": 2.1486,
+      "step": 9241500
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 3.662893809049649e-05,
+      "loss": 2.1383,
+      "step": 9242000
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 3.662821444284921e-05,
+      "loss": 2.1648,
+      "step": 9242500
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 3.662749079520193e-05,
+      "loss": 2.136,
+      "step": 9243000
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 3.662676714755465e-05,
+      "loss": 2.1374,
+      "step": 9243500
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 3.6626043499907376e-05,
+      "loss": 2.1315,
+      "step": 9244000
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 3.66253198522601e-05,
+      "loss": 2.1656,
+      "step": 9244500
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 3.662459620461282e-05,
+      "loss": 2.1479,
+      "step": 9245000
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 3.6623874004260836e-05,
+      "loss": 2.1446,
+      "step": 9245500
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 3.6623150356613565e-05,
+      "loss": 2.135,
+      "step": 9246000
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 3.662242815626158e-05,
+      "loss": 2.1551,
+      "step": 9246500
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 3.66217045086143e-05,
+      "loss": 2.1248,
+      "step": 9247000
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 3.6620980860967025e-05,
+      "loss": 2.1632,
+      "step": 9247500
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 3.662025866061504e-05,
+      "loss": 2.1292,
+      "step": 9248000
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 3.661953501296776e-05,
+      "loss": 2.1409,
+      "step": 9248500
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 3.661881136532049e-05,
+      "loss": 2.1437,
+      "step": 9249000
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 3.6618087717673214e-05,
+      "loss": 2.1611,
+      "step": 9249500
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 3.6617364070025936e-05,
+      "loss": 2.1592,
+      "step": 9250000
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 3.6616640422378665e-05,
+      "loss": 2.1548,
+      "step": 9250500
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 3.661591677473139e-05,
+      "loss": 2.1295,
+      "step": 9251000
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 3.661519312708411e-05,
+      "loss": 2.1375,
+      "step": 9251500
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 3.661446947943683e-05,
+      "loss": 2.1253,
+      "step": 9252000
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 3.661374727908485e-05,
+      "loss": 2.133,
+      "step": 9252500
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 3.661302363143757e-05,
+      "loss": 2.1567,
+      "step": 9253000
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 3.661229998379029e-05,
+      "loss": 2.1385,
+      "step": 9253500
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 3.6611576336143014e-05,
+      "loss": 2.1184,
+      "step": 9254000
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 3.6610852688495736e-05,
+      "loss": 2.1171,
+      "step": 9254500
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 3.661013048814376e-05,
+      "loss": 2.1176,
+      "step": 9255000
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 3.660940684049648e-05,
+      "loss": 2.1424,
+      "step": 9255500
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 3.66086831928492e-05,
+      "loss": 2.1311,
+      "step": 9256000
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 3.660795954520193e-05,
+      "loss": 2.1362,
+      "step": 9256500
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 3.6607235897554654e-05,
+      "loss": 2.1427,
+      "step": 9257000
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 3.6606512249907376e-05,
+      "loss": 2.1485,
+      "step": 9257500
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 3.66057886022601e-05,
+      "loss": 2.1493,
+      "step": 9258000
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 3.660506495461282e-05,
+      "loss": 2.1304,
+      "step": 9258500
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 3.6604342754260836e-05,
+      "loss": 2.11,
+      "step": 9259000
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 3.6603619106613565e-05,
+      "loss": 2.1388,
+      "step": 9259500
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 3.660289545896629e-05,
+      "loss": 2.1235,
+      "step": 9260000
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 3.66021732586143e-05,
+      "loss": 2.1608,
+      "step": 9260500
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 3.6601449610967025e-05,
+      "loss": 2.1398,
+      "step": 9261000
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 3.660072741061504e-05,
+      "loss": 2.144,
+      "step": 9261500
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 3.660000376296776e-05,
+      "loss": 2.1471,
+      "step": 9262000
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 3.659928011532049e-05,
+      "loss": 2.1408,
+      "step": 9262500
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 3.6598556467673214e-05,
+      "loss": 2.1159,
+      "step": 9263000
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 3.659783282002594e-05,
+      "loss": 2.1462,
+      "step": 9263500
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 3.6597109172378666e-05,
+      "loss": 2.1333,
+      "step": 9264000
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 3.659638552473139e-05,
+      "loss": 2.1169,
+      "step": 9264500
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 3.659566187708411e-05,
+      "loss": 2.1264,
+      "step": 9265000
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 3.659493822943683e-05,
+      "loss": 2.1359,
+      "step": 9265500
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 3.6594214581789555e-05,
+      "loss": 2.1231,
+      "step": 9266000
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 3.659349093414228e-05,
+      "loss": 2.1545,
+      "step": 9266500
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 3.6592767286495e-05,
+      "loss": 2.1399,
+      "step": 9267000
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 3.6592045086143015e-05,
+      "loss": 2.1413,
+      "step": 9267500
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 3.6591321438495744e-05,
+      "loss": 2.133,
+      "step": 9268000
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 3.659059923814376e-05,
+      "loss": 2.1533,
+      "step": 9268500
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 3.6589877037791775e-05,
+      "loss": 2.1318,
+      "step": 9269000
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 3.65891533901445e-05,
+      "loss": 2.1441,
+      "step": 9269500
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 3.658842974249722e-05,
+      "loss": 2.1386,
+      "step": 9270000
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 3.658770609484994e-05,
+      "loss": 2.1183,
+      "step": 9270500
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 3.6586982447202664e-05,
+      "loss": 2.1395,
+      "step": 9271000
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 3.658625879955539e-05,
+      "loss": 2.1544,
+      "step": 9271500
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 3.6585535151908115e-05,
+      "loss": 2.1648,
+      "step": 9272000
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 3.6584811504260844e-05,
+      "loss": 2.1506,
+      "step": 9272500
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 3.6584087856613566e-05,
+      "loss": 2.1505,
+      "step": 9273000
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 3.658336420896629e-05,
+      "loss": 2.1548,
+      "step": 9273500
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 3.658264056131901e-05,
+      "loss": 2.1496,
+      "step": 9274000
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 3.658191691367173e-05,
+      "loss": 2.1281,
+      "step": 9274500
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 3.6581193266024455e-05,
+      "loss": 2.1403,
+      "step": 9275000
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 3.658046961837718e-05,
+      "loss": 2.1404,
+      "step": 9275500
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 3.65797459707299e-05,
+      "loss": 2.1401,
+      "step": 9276000
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 3.657902232308262e-05,
+      "loss": 2.1477,
+      "step": 9276500
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 3.6578298675435344e-05,
+      "loss": 2.1195,
+      "step": 9277000
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 3.6577575027788066e-05,
+      "loss": 2.1554,
+      "step": 9277500
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 3.6576851380140795e-05,
+      "loss": 2.1367,
+      "step": 9278000
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 3.6576130627084104e-05,
+      "loss": 2.1406,
+      "step": 9278500
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 3.657540697943683e-05,
+      "loss": 2.1552,
+      "step": 9279000
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 3.6574683331789555e-05,
+      "loss": 2.1598,
+      "step": 9279500
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 3.657395968414228e-05,
+      "loss": 2.1591,
+      "step": 9280000
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 3.6573236036495e-05,
+      "loss": 2.1188,
+      "step": 9280500
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 3.657251238884772e-05,
+      "loss": 2.1295,
+      "step": 9281000
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 3.6571788741200444e-05,
+      "loss": 2.1022,
+      "step": 9281500
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 3.6571065093553167e-05,
+      "loss": 2.1486,
+      "step": 9282000
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 3.6570341445905896e-05,
+      "loss": 2.1545,
+      "step": 9282500
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 3.656961779825862e-05,
+      "loss": 2.1382,
+      "step": 9283000
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 3.656889415061134e-05,
+      "loss": 2.1414,
+      "step": 9283500
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 3.6568171950259356e-05,
+      "loss": 2.1492,
+      "step": 9284000
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 3.656744974990737e-05,
+      "loss": 2.1552,
+      "step": 9284500
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 3.6566726102260093e-05,
+      "loss": 2.1419,
+      "step": 9285000
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 3.6566002454612816e-05,
+      "loss": 2.1349,
+      "step": 9285500
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 3.656528025426084e-05,
+      "loss": 2.1363,
+      "step": 9286000
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 3.656455660661357e-05,
+      "loss": 2.1527,
+      "step": 9286500
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 3.656383295896629e-05,
+      "loss": 2.1319,
+      "step": 9287000
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 3.6563110758614305e-05,
+      "loss": 2.1648,
+      "step": 9287500
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 3.656238711096703e-05,
+      "loss": 2.1269,
+      "step": 9288000
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 3.656166346331975e-05,
+      "loss": 2.1514,
+      "step": 9288500
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 3.656093981567247e-05,
+      "loss": 2.1763,
+      "step": 9289000
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 3.6560216168025194e-05,
+      "loss": 2.1329,
+      "step": 9289500
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 3.6559492520377916e-05,
+      "loss": 2.1321,
+      "step": 9290000
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 3.6558768872730645e-05,
+      "loss": 2.1622,
+      "step": 9290500
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 3.655804522508337e-05,
+      "loss": 2.1351,
+      "step": 9291000
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 3.655732157743609e-05,
+      "loss": 2.1534,
+      "step": 9291500
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 3.655659792978881e-05,
+      "loss": 2.1489,
+      "step": 9292000
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 3.6555874282141534e-05,
+      "loss": 2.1286,
+      "step": 9292500
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 3.6555150634494256e-05,
+      "loss": 2.1301,
+      "step": 9293000
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 3.6554426986846985e-05,
+      "loss": 2.1108,
+      "step": 9293500
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 3.655370333919971e-05,
+      "loss": 2.1345,
+      "step": 9294000
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 3.655298113884772e-05,
+      "loss": 2.1089,
+      "step": 9294500
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 3.6552257491200445e-05,
+      "loss": 2.157,
+      "step": 9295000
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 3.655153384355317e-05,
+      "loss": 2.1542,
+      "step": 9295500
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 3.655081164320119e-05,
+      "loss": 2.1379,
+      "step": 9296000
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 3.655008799555391e-05,
+      "loss": 2.153,
+      "step": 9296500
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 3.6549364347906634e-05,
+      "loss": 2.1381,
+      "step": 9297000
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 3.6548640700259356e-05,
+      "loss": 2.1364,
+      "step": 9297500
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 3.654791705261208e-05,
+      "loss": 2.1477,
+      "step": 9298000
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 3.65471934049648e-05,
+      "loss": 2.1424,
+      "step": 9298500
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 3.654646975731752e-05,
+      "loss": 2.1482,
+      "step": 9299000
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 3.6545746109670245e-05,
+      "loss": 2.1598,
+      "step": 9299500
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 3.654502246202297e-05,
+      "loss": 2.1278,
+      "step": 9300000
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 3.654430026167099e-05,
+      "loss": 2.1282,
+      "step": 9300500
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 3.654357661402372e-05,
+      "loss": 2.1273,
+      "step": 9301000
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 3.6542854413671735e-05,
+      "loss": 2.1472,
+      "step": 9301500
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 3.654213076602446e-05,
+      "loss": 2.157,
+      "step": 9302000
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 3.654140856567247e-05,
+      "loss": 2.1444,
+      "step": 9302500
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 3.6540684918025195e-05,
+      "loss": 2.1562,
+      "step": 9303000
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 3.6539961270377924e-05,
+      "loss": 2.1473,
+      "step": 9303500
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 3.6539237622730646e-05,
+      "loss": 2.1421,
+      "step": 9304000
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 3.6538516869673955e-05,
+      "loss": 2.16,
+      "step": 9304500
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 3.653779322202668e-05,
+      "loss": 2.1664,
+      "step": 9305000
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 3.65370695743794e-05,
+      "loss": 2.1723,
+      "step": 9305500
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 3.653634592673212e-05,
+      "loss": 2.1505,
+      "step": 9306000
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 3.6535622279084844e-05,
+      "loss": 2.124,
+      "step": 9306500
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 3.653489863143757e-05,
+      "loss": 2.1374,
+      "step": 9307000
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 3.6534174983790295e-05,
+      "loss": 2.1466,
+      "step": 9307500
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 3.653345133614302e-05,
+      "loss": 2.1059,
+      "step": 9308000
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 3.6532727688495746e-05,
+      "loss": 2.1284,
+      "step": 9308500
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 3.653200404084847e-05,
+      "loss": 2.1206,
+      "step": 9309000
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 3.653128039320119e-05,
+      "loss": 2.1431,
+      "step": 9309500
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 3.653055674555391e-05,
+      "loss": 2.1342,
+      "step": 9310000
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 3.6529833097906635e-05,
+      "loss": 2.1258,
+      "step": 9310500
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 3.652910945025936e-05,
+      "loss": 2.1348,
+      "step": 9311000
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 3.652838580261208e-05,
+      "loss": 2.1432,
+      "step": 9311500
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 3.65276621549648e-05,
+      "loss": 2.1449,
+      "step": 9312000
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 3.6526939954612824e-05,
+      "loss": 2.1341,
+      "step": 9312500
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 3.6526216306965546e-05,
+      "loss": 2.1419,
+      "step": 9313000
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 3.652549265931827e-05,
+      "loss": 2.1376,
+      "step": 9313500
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 3.6524770458966284e-05,
+      "loss": 2.1499,
+      "step": 9314000
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 3.6524046811319006e-05,
+      "loss": 2.1365,
+      "step": 9314500
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 3.652332316367173e-05,
+      "loss": 2.1465,
+      "step": 9315000
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 3.6522600963319744e-05,
+      "loss": 2.1642,
+      "step": 9315500
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 3.652187731567247e-05,
+      "loss": 2.1571,
+      "step": 9316000
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 3.6521153668025195e-05,
+      "loss": 2.1414,
+      "step": 9316500
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 3.6520430020377924e-05,
+      "loss": 2.1382,
+      "step": 9317000
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 3.6519706372730647e-05,
+      "loss": 2.1122,
+      "step": 9317500
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 3.651898272508337e-05,
+      "loss": 2.1298,
+      "step": 9318000
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 3.651825907743609e-05,
+      "loss": 2.152,
+      "step": 9318500
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 3.651753542978881e-05,
+      "loss": 2.1294,
+      "step": 9319000
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 3.6516811782141535e-05,
+      "loss": 2.1448,
+      "step": 9319500
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 3.651608813449426e-05,
+      "loss": 2.1297,
+      "step": 9320000
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 3.651536448684698e-05,
+      "loss": 2.1291,
+      "step": 9320500
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 3.65146408391997e-05,
+      "loss": 2.1351,
+      "step": 9321000
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 3.6513917191552424e-05,
+      "loss": 2.1276,
+      "step": 9321500
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 3.651319354390515e-05,
+      "loss": 2.1426,
+      "step": 9322000
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 3.651247134355317e-05,
+      "loss": 2.1344,
+      "step": 9322500
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 3.6511749143201185e-05,
+      "loss": 2.1516,
+      "step": 9323000
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 3.6511025495553914e-05,
+      "loss": 2.1658,
+      "step": 9323500
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 3.6510301847906636e-05,
+      "loss": 2.1379,
+      "step": 9324000
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 3.650957820025936e-05,
+      "loss": 2.1527,
+      "step": 9324500
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 3.650885455261208e-05,
+      "loss": 2.11,
+      "step": 9325000
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 3.65081309049648e-05,
+      "loss": 2.1295,
+      "step": 9325500
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 3.6507407257317525e-05,
+      "loss": 2.1344,
+      "step": 9326000
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 3.650668360967025e-05,
+      "loss": 2.1454,
+      "step": 9326500
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 3.6505959962022976e-05,
+      "loss": 2.1264,
+      "step": 9327000
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 3.650523776167099e-05,
+      "loss": 2.1357,
+      "step": 9327500
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy": 0.6654050335667157,
+      "eval_accuracy_mlm": 0.6298091509434769,
+      "eval_accuracy_nsp": 0.8562923650162929,
+      "eval_loss": 2.195469617843628,
+      "eval_runtime": 332.0987,
+      "eval_samples_per_second": 1314.025,
+      "eval_steps_per_second": 54.752,
+      "step": 9327744
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 3.6504514114023714e-05,
+      "loss": 2.1241,
+      "step": 9328000
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 3.6503790466376436e-05,
+      "loss": 2.1168,
+      "step": 9328500
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 3.650306681872916e-05,
+      "loss": 2.1252,
+      "step": 9329000
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 3.6502344618377174e-05,
+      "loss": 2.1047,
+      "step": 9329500
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 3.6501622418025196e-05,
+      "loss": 2.1268,
+      "step": 9330000
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 3.650089877037792e-05,
+      "loss": 2.1355,
+      "step": 9330500
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 3.650017512273065e-05,
+      "loss": 2.112,
+      "step": 9331000
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 3.649945147508337e-05,
+      "loss": 2.1301,
+      "step": 9331500
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 3.649872782743609e-05,
+      "loss": 2.1215,
+      "step": 9332000
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 3.6498004179788814e-05,
+      "loss": 2.1114,
+      "step": 9332500
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.6497280532141536e-05,
+      "loss": 2.127,
+      "step": 9333000
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.649655688449426e-05,
+      "loss": 2.107,
+      "step": 9333500
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.649583323684698e-05,
+      "loss": 2.1007,
+      "step": 9334000
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.64951095891997e-05,
+      "loss": 2.1159,
+      "step": 9334500
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.6494385941552425e-05,
+      "loss": 2.1118,
+      "step": 9335000
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.6493662293905154e-05,
+      "loss": 2.1256,
+      "step": 9335500
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.6492938646257876e-05,
+      "loss": 2.1073,
+      "step": 9336000
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.649221644590589e-05,
+      "loss": 2.0944,
+      "step": 9336500
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.6491492798258614e-05,
+      "loss": 2.0876,
+      "step": 9337000
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.6490769150611336e-05,
+      "loss": 2.1152,
+      "step": 9337500
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.649004695025935e-05,
+      "loss": 2.1064,
+      "step": 9338000
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.648932330261208e-05,
+      "loss": 2.1165,
+      "step": 9338500
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.64885996549648e-05,
+      "loss": 2.1109,
+      "step": 9339000
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.6487876007317526e-05,
+      "loss": 2.1144,
+      "step": 9339500
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.6487152359670255e-05,
+      "loss": 2.1415,
+      "step": 9340000
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.648642871202298e-05,
+      "loss": 2.1132,
+      "step": 9340500
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.64857050643757e-05,
+      "loss": 2.1361,
+      "step": 9341000
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.648498141672842e-05,
+      "loss": 2.1143,
+      "step": 9341500
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.648425921637644e-05,
+      "loss": 2.1237,
+      "step": 9342000
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.648353556872916e-05,
+      "loss": 2.1159,
+      "step": 9342500
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.648281192108188e-05,
+      "loss": 2.1212,
+      "step": 9343000
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.6482089720729904e-05,
+      "loss": 2.1147,
+      "step": 9343500
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.6481366073082626e-05,
+      "loss": 2.0983,
+      "step": 9344000
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.648064242543535e-05,
+      "loss": 2.1101,
+      "step": 9344500
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.647991877778807e-05,
+      "loss": 2.1243,
+      "step": 9345000
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.647919513014079e-05,
+      "loss": 2.12,
+      "step": 9345500
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.64784758243794e-05,
+      "loss": 2.1144,
+      "step": 9346000
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.647775217673213e-05,
+      "loss": 2.1069,
+      "step": 9346500
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.647702852908485e-05,
+      "loss": 2.1268,
+      "step": 9347000
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.6476304881437575e-05,
+      "loss": 2.141,
+      "step": 9347500
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.64755812337903e-05,
+      "loss": 2.1165,
+      "step": 9348000
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.647485758614302e-05,
+      "loss": 2.1145,
+      "step": 9348500
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.647413393849574e-05,
+      "loss": 2.1054,
+      "step": 9349000
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.6473410290848464e-05,
+      "loss": 2.1317,
+      "step": 9349500
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.6472686643201186e-05,
+      "loss": 2.1535,
+      "step": 9350000
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.647196299555391e-05,
+      "loss": 2.1028,
+      "step": 9350500
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.647123934790663e-05,
+      "loss": 2.1305,
+      "step": 9351000
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.647051570025935e-05,
+      "loss": 2.1478,
+      "step": 9351500
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.6469792052612075e-05,
+      "loss": 2.1249,
+      "step": 9352000
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.6469068404964804e-05,
+      "loss": 2.1451,
+      "step": 9352500
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.6468344757317526e-05,
+      "loss": 2.1325,
+      "step": 9353000
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.6467621109670255e-05,
+      "loss": 2.1186,
+      "step": 9353500
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.646689890931827e-05,
+      "loss": 2.1638,
+      "step": 9354000
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.646617526167099e-05,
+      "loss": 2.1217,
+      "step": 9354500
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.6465451614023715e-05,
+      "loss": 2.1206,
+      "step": 9355000
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.646472796637644e-05,
+      "loss": 2.1012,
+      "step": 9355500
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.646400576602445e-05,
+      "loss": 2.1042,
+      "step": 9356000
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.6463283565672475e-05,
+      "loss": 2.1169,
+      "step": 9356500
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.64625599180252e-05,
+      "loss": 2.1303,
+      "step": 9357000
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.646183627037792e-05,
+      "loss": 2.0985,
+      "step": 9357500
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.646111262273064e-05,
+      "loss": 2.1195,
+      "step": 9358000
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.6460388975083364e-05,
+      "loss": 2.1092,
+      "step": 9358500
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.645966532743609e-05,
+      "loss": 2.1248,
+      "step": 9359000
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.645894167978881e-05,
+      "loss": 2.1193,
+      "step": 9359500
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.645821803214153e-05,
+      "loss": 2.1186,
+      "step": 9360000
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.645749438449425e-05,
+      "loss": 2.1365,
+      "step": 9360500
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.645677073684698e-05,
+      "loss": 2.1207,
+      "step": 9361000
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.6456048536495005e-05,
+      "loss": 2.1427,
+      "step": 9361500
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.645532488884773e-05,
+      "loss": 2.1196,
+      "step": 9362000
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.645460124120045e-05,
+      "loss": 2.1274,
+      "step": 9362500
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.645387759355317e-05,
+      "loss": 2.0993,
+      "step": 9363000
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.6453153945905894e-05,
+      "loss": 2.1174,
+      "step": 9363500
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.6452430298258616e-05,
+      "loss": 2.1102,
+      "step": 9364000
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.645170665061134e-05,
+      "loss": 2.1362,
+      "step": 9364500
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.6450984450259354e-05,
+      "loss": 2.1262,
+      "step": 9365000
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.6450262249907376e-05,
+      "loss": 2.1112,
+      "step": 9365500
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.64495386022601e-05,
+      "loss": 2.1292,
+      "step": 9366000
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.644881495461282e-05,
+      "loss": 2.0959,
+      "step": 9366500
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.644809130696554e-05,
+      "loss": 2.1159,
+      "step": 9367000
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.6447367659318265e-05,
+      "loss": 2.126,
+      "step": 9367500
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.644664401167099e-05,
+      "loss": 2.1283,
+      "step": 9368000
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.6445920364023716e-05,
+      "loss": 2.1089,
+      "step": 9368500
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.644519671637644e-05,
+      "loss": 2.1168,
+      "step": 9369000
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.644447306872916e-05,
+      "loss": 2.1102,
+      "step": 9369500
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.644375086837718e-05,
+      "loss": 2.1243,
+      "step": 9370000
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.6443027220729905e-05,
+      "loss": 2.1364,
+      "step": 9370500
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.644230502037792e-05,
+      "loss": 2.1184,
+      "step": 9371000
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.644158137273064e-05,
+      "loss": 2.133,
+      "step": 9371500
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.6440857725083365e-05,
+      "loss": 2.1167,
+      "step": 9372000
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.644013407743609e-05,
+      "loss": 2.1125,
+      "step": 9372500
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.643941042978881e-05,
+      "loss": 2.1,
+      "step": 9373000
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.643868678214153e-05,
+      "loss": 2.1185,
+      "step": 9373500
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.6437963134494254e-05,
+      "loss": 2.1121,
+      "step": 9374000
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.643723948684698e-05,
+      "loss": 2.112,
+      "step": 9374500
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.6436515839199705e-05,
+      "loss": 2.1383,
+      "step": 9375000
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.643579219155243e-05,
+      "loss": 2.1072,
+      "step": 9375500
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.643506854390516e-05,
+      "loss": 2.1162,
+      "step": 9376000
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.643434489625788e-05,
+      "loss": 2.1288,
+      "step": 9376500
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.64336212486106e-05,
+      "loss": 2.1308,
+      "step": 9377000
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.643289760096332e-05,
+      "loss": 2.1348,
+      "step": 9377500
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.643217540061134e-05,
+      "loss": 2.107,
+      "step": 9378000
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.643145175296406e-05,
+      "loss": 2.1144,
+      "step": 9378500
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.6430729552612083e-05,
+      "loss": 2.1105,
+      "step": 9379000
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.6430005904964806e-05,
+      "loss": 2.1332,
+      "step": 9379500
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.642928225731753e-05,
+      "loss": 2.111,
+      "step": 9380000
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.642855860967025e-05,
+      "loss": 2.1396,
+      "step": 9380500
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.642783496202297e-05,
+      "loss": 2.1539,
+      "step": 9381000
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.6427111314375695e-05,
+      "loss": 2.1049,
+      "step": 9381500
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.642638766672842e-05,
+      "loss": 2.1525,
+      "step": 9382000
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.642566401908114e-05,
+      "loss": 2.116,
+      "step": 9382500
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.6424941818729155e-05,
+      "loss": 2.1464,
+      "step": 9383000
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.6424218171081884e-05,
+      "loss": 2.1183,
+      "step": 9383500
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.6423495970729906e-05,
+      "loss": 2.1141,
+      "step": 9384000
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.642277232308263e-05,
+      "loss": 2.1176,
+      "step": 9384500
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.642204867543535e-05,
+      "loss": 2.1113,
+      "step": 9385000
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.642132502778807e-05,
+      "loss": 2.1593,
+      "step": 9385500
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.6420601380140795e-05,
+      "loss": 2.134,
+      "step": 9386000
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.641987773249352e-05,
+      "loss": 2.0956,
+      "step": 9386500
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.641915553214153e-05,
+      "loss": 2.1253,
+      "step": 9387000
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.641843188449426e-05,
+      "loss": 2.1252,
+      "step": 9387500
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.6417708236846984e-05,
+      "loss": 2.1253,
+      "step": 9388000
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.6416986036495e-05,
+      "loss": 2.1338,
+      "step": 9388500
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.6416263836143015e-05,
+      "loss": 2.1018,
+      "step": 9389000
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.641554018849574e-05,
+      "loss": 2.1206,
+      "step": 9389500
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.641481654084846e-05,
+      "loss": 2.1274,
+      "step": 9390000
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.641409289320118e-05,
+      "loss": 2.1407,
+      "step": 9390500
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.641336924555391e-05,
+      "loss": 2.1033,
+      "step": 9391000
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.641264559790663e-05,
+      "loss": 2.1043,
+      "step": 9391500
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.641192195025936e-05,
+      "loss": 2.1227,
+      "step": 9392000
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.6411198302612084e-05,
+      "loss": 2.1133,
+      "step": 9392500
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.6410474654964806e-05,
+      "loss": 2.1042,
+      "step": 9393000
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.640975100731753e-05,
+      "loss": 2.1152,
+      "step": 9393500
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.640902735967025e-05,
+      "loss": 2.1534,
+      "step": 9394000
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.640830371202297e-05,
+      "loss": 2.1233,
+      "step": 9394500
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.640758151167099e-05,
+      "loss": 2.1142,
+      "step": 9395000
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.640685786402371e-05,
+      "loss": 2.137,
+      "step": 9395500
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.640613421637643e-05,
+      "loss": 2.1133,
+      "step": 9396000
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.640541056872916e-05,
+      "loss": 2.1185,
+      "step": 9396500
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.6404686921081884e-05,
+      "loss": 2.1367,
+      "step": 9397000
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.640396327343461e-05,
+      "loss": 2.1468,
+      "step": 9397500
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.6403239625787336e-05,
+      "loss": 2.1351,
+      "step": 9398000
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.640251597814006e-05,
+      "loss": 2.1358,
+      "step": 9398500
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.640179233049278e-05,
+      "loss": 2.1508,
+      "step": 9399000
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.64010686828455e-05,
+      "loss": 2.112,
+      "step": 9399500
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.6400345035198225e-05,
+      "loss": 2.1166,
+      "step": 9400000
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.639962138755095e-05,
+      "loss": 2.1465,
+      "step": 9400500
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.639889773990367e-05,
+      "loss": 2.1014,
+      "step": 9401000
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.639817409225639e-05,
+      "loss": 2.1333,
+      "step": 9401500
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.6397450444609114e-05,
+      "loss": 2.1416,
+      "step": 9402000
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.6396728244257136e-05,
+      "loss": 2.1028,
+      "step": 9402500
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.639600459660986e-05,
+      "loss": 2.136,
+      "step": 9403000
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.639528094896258e-05,
+      "loss": 2.1353,
+      "step": 9403500
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.63945573013153e-05,
+      "loss": 2.1107,
+      "step": 9404000
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.6393833653668025e-05,
+      "loss": 2.0979,
+      "step": 9404500
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.639311145331604e-05,
+      "loss": 2.1314,
+      "step": 9405000
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.639238780566877e-05,
+      "loss": 2.1138,
+      "step": 9405500
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.6391665605316785e-05,
+      "loss": 2.1193,
+      "step": 9406000
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.6390941957669514e-05,
+      "loss": 2.1203,
+      "step": 9406500
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.6390218310022236e-05,
+      "loss": 2.1295,
+      "step": 9407000
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.638949466237496e-05,
+      "loss": 2.1479,
+      "step": 9407500
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.638877101472768e-05,
+      "loss": 2.1429,
+      "step": 9408000
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.6388048814375696e-05,
+      "loss": 2.1206,
+      "step": 9408500
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.638732516672842e-05,
+      "loss": 2.1061,
+      "step": 9409000
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.638660151908114e-05,
+      "loss": 2.1364,
+      "step": 9409500
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.638587787143386e-05,
+      "loss": 2.1165,
+      "step": 9410000
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.6385154223786585e-05,
+      "loss": 2.1244,
+      "step": 9410500
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.638443202343461e-05,
+      "loss": 2.1306,
+      "step": 9411000
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.638370982308262e-05,
+      "loss": 2.1197,
+      "step": 9411500
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.6382986175435345e-05,
+      "loss": 2.1139,
+      "step": 9412000
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.638226252778807e-05,
+      "loss": 2.1179,
+      "step": 9412500
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.638153888014079e-05,
+      "loss": 2.1289,
+      "step": 9413000
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.638081523249352e-05,
+      "loss": 2.1565,
+      "step": 9413500
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.638009303214154e-05,
+      "loss": 2.117,
+      "step": 9414000
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.637936938449426e-05,
+      "loss": 2.1167,
+      "step": 9414500
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.6378645736846986e-05,
+      "loss": 2.1219,
+      "step": 9415000
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.637792208919971e-05,
+      "loss": 2.1162,
+      "step": 9415500
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.637719844155243e-05,
+      "loss": 2.1259,
+      "step": 9416000
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.637647479390515e-05,
+      "loss": 2.1318,
+      "step": 9416500
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.6375751146257874e-05,
+      "loss": 2.1223,
+      "step": 9417000
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.63750274986106e-05,
+      "loss": 2.1176,
+      "step": 9417500
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.637430385096332e-05,
+      "loss": 2.1136,
+      "step": 9418000
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.637358020331604e-05,
+      "loss": 2.1365,
+      "step": 9418500
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.6372856555668763e-05,
+      "loss": 2.1179,
+      "step": 9419000
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.6372132908021486e-05,
+      "loss": 2.1056,
+      "step": 9419500
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.6371409260374215e-05,
+      "loss": 2.1146,
+      "step": 9420000
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.637068561272694e-05,
+      "loss": 2.1341,
+      "step": 9420500
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.636996341237496e-05,
+      "loss": 2.1421,
+      "step": 9421000
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.636923976472768e-05,
+      "loss": 2.1355,
+      "step": 9421500
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.6368516117080404e-05,
+      "loss": 2.1241,
+      "step": 9422000
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.6367792469433126e-05,
+      "loss": 2.1196,
+      "step": 9422500
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.636706882178585e-05,
+      "loss": 2.1396,
+      "step": 9423000
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.6366346621433864e-05,
+      "loss": 2.1212,
+      "step": 9423500
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.636562297378659e-05,
+      "loss": 2.1376,
+      "step": 9424000
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.6364899326139315e-05,
+      "loss": 2.124,
+      "step": 9424500
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.636417567849204e-05,
+      "loss": 2.1347,
+      "step": 9425000
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.636345203084476e-05,
+      "loss": 2.1274,
+      "step": 9425500
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.636272838319748e-05,
+      "loss": 2.1162,
+      "step": 9426000
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.6362004735550204e-05,
+      "loss": 2.1105,
+      "step": 9426500
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.636128253519822e-05,
+      "loss": 2.1147,
+      "step": 9427000
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.636056033484624e-05,
+      "loss": 2.1417,
+      "step": 9427500
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.6359836687198964e-05,
+      "loss": 2.1242,
+      "step": 9428000
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.635911303955169e-05,
+      "loss": 2.1308,
+      "step": 9428500
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.6358389391904415e-05,
+      "loss": 2.1121,
+      "step": 9429000
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.635766574425714e-05,
+      "loss": 2.141,
+      "step": 9429500
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.635694209660986e-05,
+      "loss": 2.1219,
+      "step": 9430000
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.635621844896258e-05,
+      "loss": 2.1323,
+      "step": 9430500
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.6355494801315304e-05,
+      "loss": 2.1395,
+      "step": 9431000
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.6354771153668026e-05,
+      "loss": 2.1123,
+      "step": 9431500
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.635404750602075e-05,
+      "loss": 2.1056,
+      "step": 9432000
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.635332385837347e-05,
+      "loss": 2.1416,
+      "step": 9432500
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.635260165802149e-05,
+      "loss": 2.1166,
+      "step": 9433000
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.635187945766951e-05,
+      "loss": 2.1069,
+      "step": 9433500
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.635115581002223e-05,
+      "loss": 2.1161,
+      "step": 9434000
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.635043216237495e-05,
+      "loss": 2.1283,
+      "step": 9434500
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.6349708514727675e-05,
+      "loss": 2.1177,
+      "step": 9435000
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.6348984867080404e-05,
+      "loss": 2.1167,
+      "step": 9435500
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.634826121943313e-05,
+      "loss": 2.1189,
+      "step": 9436000
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.634753757178585e-05,
+      "loss": 2.135,
+      "step": 9436500
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.634681392413857e-05,
+      "loss": 2.1005,
+      "step": 9437000
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.6346090276491293e-05,
+      "loss": 2.144,
+      "step": 9437500
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.6345366628844016e-05,
+      "loss": 2.1205,
+      "step": 9438000
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.634464587578733e-05,
+      "loss": 2.1247,
+      "step": 9438500
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.6343922228140054e-05,
+      "loss": 2.1106,
+      "step": 9439000
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.6343198580492776e-05,
+      "loss": 2.1286,
+      "step": 9439500
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.63424749328455e-05,
+      "loss": 2.131,
+      "step": 9440000
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 3.634175128519822e-05,
+      "loss": 2.1254,
+      "step": 9440500
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 3.634102763755094e-05,
+      "loss": 2.1117,
+      "step": 9441000
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 3.6340303989903665e-05,
+      "loss": 2.1368,
+      "step": 9441500
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 3.6339580342256394e-05,
+      "loss": 2.1311,
+      "step": 9442000
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 3.633885814190441e-05,
+      "loss": 2.1257,
+      "step": 9442500
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 3.633813449425714e-05,
+      "loss": 2.1567,
+      "step": 9443000
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.633741084660986e-05,
+      "loss": 2.1161,
+      "step": 9443500
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.633668719896258e-05,
+      "loss": 2.1364,
+      "step": 9444000
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.6335963551315305e-05,
+      "loss": 2.1235,
+      "step": 9444500
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.633523990366803e-05,
+      "loss": 2.1249,
+      "step": 9445000
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.633451625602075e-05,
+      "loss": 2.1353,
+      "step": 9445500
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.6333794055668765e-05,
+      "loss": 2.1088,
+      "step": 9446000
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.6333070408021494e-05,
+      "loss": 2.1205,
+      "step": 9446500
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.6332346760374216e-05,
+      "loss": 2.1349,
+      "step": 9447000
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.633162311272694e-05,
+      "loss": 2.1224,
+      "step": 9447500
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.633089946507966e-05,
+      "loss": 2.1554,
+      "step": 9448000
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.633017581743238e-05,
+      "loss": 2.1129,
+      "step": 9448500
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.6329452169785105e-05,
+      "loss": 2.1289,
+      "step": 9449000
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.632872852213783e-05,
+      "loss": 2.1087,
+      "step": 9449500
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.6328004874490556e-05,
+      "loss": 2.1153,
+      "step": 9450000
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.632728122684328e-05,
+      "loss": 2.1417,
+      "step": 9450500
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.6326557579196e-05,
+      "loss": 2.1387,
+      "step": 9451000
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.632583393154872e-05,
+      "loss": 2.1071,
+      "step": 9451500
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.632511317849204e-05,
+      "loss": 2.1391,
+      "step": 9452000
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.632438953084476e-05,
+      "loss": 2.1154,
+      "step": 9452500
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.632366588319748e-05,
+      "loss": 2.0988,
+      "step": 9453000
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.6322942235550205e-05,
+      "loss": 2.1395,
+      "step": 9453500
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.632222003519822e-05,
+      "loss": 2.1016,
+      "step": 9454000
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.632149638755094e-05,
+      "loss": 2.1325,
+      "step": 9454500
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.632077273990367e-05,
+      "loss": 2.134,
+      "step": 9455000
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.632005053955169e-05,
+      "loss": 2.1332,
+      "step": 9455500
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.631932689190441e-05,
+      "loss": 2.1176,
+      "step": 9456000
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.631860324425713e-05,
+      "loss": 2.1223,
+      "step": 9456500
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.6317879596609855e-05,
+      "loss": 2.144,
+      "step": 9457000
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.631715594896258e-05,
+      "loss": 2.1235,
+      "step": 9457500
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.6316432301315306e-05,
+      "loss": 2.1245,
+      "step": 9458000
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.631570865366803e-05,
+      "loss": 2.104,
+      "step": 9458500
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.631498500602075e-05,
+      "loss": 2.118,
+      "step": 9459000
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.631426135837347e-05,
+      "loss": 2.149,
+      "step": 9459500
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.6313537710726195e-05,
+      "loss": 2.1454,
+      "step": 9460000
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.6312814063078924e-05,
+      "loss": 2.1441,
+      "step": 9460500
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.6312090415431646e-05,
+      "loss": 2.1428,
+      "step": 9461000
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.631136676778437e-05,
+      "loss": 2.1154,
+      "step": 9461500
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.631064312013709e-05,
+      "loss": 2.1374,
+      "step": 9462000
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.6309920919785106e-05,
+      "loss": 2.1253,
+      "step": 9462500
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.630919727213783e-05,
+      "loss": 2.1299,
+      "step": 9463000
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.630847362449055e-05,
+      "loss": 2.1074,
+      "step": 9463500
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.630774997684327e-05,
+      "loss": 2.1154,
+      "step": 9464000
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.6307026329195995e-05,
+      "loss": 2.1189,
+      "step": 9464500
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.6306302681548724e-05,
+      "loss": 2.1306,
+      "step": 9465000
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.6305580481196746e-05,
+      "loss": 2.1135,
+      "step": 9465500
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.630485828084476e-05,
+      "loss": 2.1419,
+      "step": 9466000
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.6304134633197484e-05,
+      "loss": 2.1205,
+      "step": 9466500
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.6303410985550206e-05,
+      "loss": 2.123,
+      "step": 9467000
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.630268733790293e-05,
+      "loss": 2.1353,
+      "step": 9467500
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.630196369025565e-05,
+      "loss": 2.1059,
+      "step": 9468000
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.630124004260837e-05,
+      "loss": 2.1037,
+      "step": 9468500
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.6300516394961095e-05,
+      "loss": 2.1309,
+      "step": 9469000
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.629979564190441e-05,
+      "loss": 2.1089,
+      "step": 9469500
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.629907199425713e-05,
+      "loss": 2.1391,
+      "step": 9470000
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.6298348346609855e-05,
+      "loss": 2.1275,
+      "step": 9470500
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.629762469896258e-05,
+      "loss": 2.1259,
+      "step": 9471000
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.62969010513153e-05,
+      "loss": 2.1009,
+      "step": 9471500
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.629617885096332e-05,
+      "loss": 2.1243,
+      "step": 9472000
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.6295455203316044e-05,
+      "loss": 2.1183,
+      "step": 9472500
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.629473155566877e-05,
+      "loss": 2.1456,
+      "step": 9473000
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.6294007908021496e-05,
+      "loss": 2.1257,
+      "step": 9473500
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.629328426037422e-05,
+      "loss": 2.1111,
+      "step": 9474000
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.629256061272694e-05,
+      "loss": 2.1249,
+      "step": 9474500
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.629183696507966e-05,
+      "loss": 2.1227,
+      "step": 9475000
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.629111476472768e-05,
+      "loss": 2.1397,
+      "step": 9475500
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.62903911170804e-05,
+      "loss": 2.1201,
+      "step": 9476000
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.628966746943312e-05,
+      "loss": 2.1195,
+      "step": 9476500
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.628894382178585e-05,
+      "loss": 2.1175,
+      "step": 9477000
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.6288220174138574e-05,
+      "loss": 2.1034,
+      "step": 9477500
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.6287496526491296e-05,
+      "loss": 2.1252,
+      "step": 9478000
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.628677287884402e-05,
+      "loss": 2.1167,
+      "step": 9478500
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.628604923119674e-05,
+      "loss": 2.1445,
+      "step": 9479000
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.628532558354946e-05,
+      "loss": 2.1118,
+      "step": 9479500
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.628460193590219e-05,
+      "loss": 2.1121,
+      "step": 9480000
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.6283878288254914e-05,
+      "loss": 2.1105,
+      "step": 9480500
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.6283154640607636e-05,
+      "loss": 2.114,
+      "step": 9481000
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.628243099296036e-05,
+      "loss": 2.1317,
+      "step": 9481500
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.628170734531308e-05,
+      "loss": 2.1198,
+      "step": 9482000
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.6280985144961096e-05,
+      "loss": 2.1091,
+      "step": 9482500
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.6280261497313825e-05,
+      "loss": 2.1437,
+      "step": 9483000
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.627953929696184e-05,
+      "loss": 2.1233,
+      "step": 9483500
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.627881564931456e-05,
+      "loss": 2.1414,
+      "step": 9484000
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.6278092001667285e-05,
+      "loss": 2.1299,
+      "step": 9484500
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.62773698013153e-05,
+      "loss": 2.1404,
+      "step": 9485000
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.627664615366802e-05,
+      "loss": 2.1494,
+      "step": 9485500
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.627592250602075e-05,
+      "loss": 2.1407,
+      "step": 9486000
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.6275198858373474e-05,
+      "loss": 2.1245,
+      "step": 9486500
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.6274475210726196e-05,
+      "loss": 2.1194,
+      "step": 9487000
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.627375156307892e-05,
+      "loss": 2.1281,
+      "step": 9487500
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.627302791543165e-05,
+      "loss": 2.132,
+      "step": 9488000
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.627230426778437e-05,
+      "loss": 2.1419,
+      "step": 9488500
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.627158062013709e-05,
+      "loss": 2.1498,
+      "step": 9489000
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.6270856972489814e-05,
+      "loss": 2.1208,
+      "step": 9489500
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.627013477213783e-05,
+      "loss": 2.1102,
+      "step": 9490000
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.626941257178585e-05,
+      "loss": 2.14,
+      "step": 9490500
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.6268688924138574e-05,
+      "loss": 2.1083,
+      "step": 9491000
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.6267965276491297e-05,
+      "loss": 2.1337,
+      "step": 9491500
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.626724162884402e-05,
+      "loss": 2.1272,
+      "step": 9492000
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.626651798119674e-05,
+      "loss": 2.097,
+      "step": 9492500
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.626579433354946e-05,
+      "loss": 2.128,
+      "step": 9493000
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.6265070685902186e-05,
+      "loss": 2.1427,
+      "step": 9493500
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.626434703825491e-05,
+      "loss": 2.1121,
+      "step": 9494000
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.626362339060763e-05,
+      "loss": 2.1649,
+      "step": 9494500
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.626290119025565e-05,
+      "loss": 2.082,
+      "step": 9495000
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.6262177542608375e-05,
+      "loss": 2.1075,
+      "step": 9495500
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.6261453894961104e-05,
+      "loss": 2.1376,
+      "step": 9496000
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.6260730247313826e-05,
+      "loss": 2.1288,
+      "step": 9496500
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.626000659966655e-05,
+      "loss": 2.156,
+      "step": 9497000
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.625928295201927e-05,
+      "loss": 2.1263,
+      "step": 9497500
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.625855930437199e-05,
+      "loss": 2.1282,
+      "step": 9498000
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.625783710402001e-05,
+      "loss": 2.1351,
+      "step": 9498500
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.625711345637273e-05,
+      "loss": 2.1434,
+      "step": 9499000
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.625638980872545e-05,
+      "loss": 2.1384,
+      "step": 9499500
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.6255666161078175e-05,
+      "loss": 2.1443,
+      "step": 9500000
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.6254942513430904e-05,
+      "loss": 2.1507,
+      "step": 9500500
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.6254218865783626e-05,
+      "loss": 2.1403,
+      "step": 9501000
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.625349666543164e-05,
+      "loss": 2.1203,
+      "step": 9501500
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.625277446507966e-05,
+      "loss": 2.1091,
+      "step": 9502000
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.625205081743238e-05,
+      "loss": 2.1346,
+      "step": 9502500
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.625132716978511e-05,
+      "loss": 2.135,
+      "step": 9503000
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.625060352213783e-05,
+      "loss": 2.108,
+      "step": 9503500
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.624987987449055e-05,
+      "loss": 2.1312,
+      "step": 9504000
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.6249156226843275e-05,
+      "loss": 2.1144,
+      "step": 9504500
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.6248432579196004e-05,
+      "loss": 2.1127,
+      "step": 9505000
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.6247708931548726e-05,
+      "loss": 2.1425,
+      "step": 9505500
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.624698673119674e-05,
+      "loss": 2.1135,
+      "step": 9506000
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.6246263083549464e-05,
+      "loss": 2.1232,
+      "step": 9506500
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.6245539435902186e-05,
+      "loss": 2.112,
+      "step": 9507000
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.624481578825491e-05,
+      "loss": 2.1397,
+      "step": 9507500
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.624409214060763e-05,
+      "loss": 2.1157,
+      "step": 9508000
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.624336849296035e-05,
+      "loss": 2.133,
+      "step": 9508500
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.6242644845313075e-05,
+      "loss": 2.1287,
+      "step": 9509000
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.6241921197665804e-05,
+      "loss": 2.1079,
+      "step": 9509500
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.6241197550018526e-05,
+      "loss": 2.1054,
+      "step": 9510000
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.6240473902371256e-05,
+      "loss": 2.1095,
+      "step": 9510500
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.623975025472398e-05,
+      "loss": 2.1208,
+      "step": 9511000
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.62390266070767e-05,
+      "loss": 2.1257,
+      "step": 9511500
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.623830295942942e-05,
+      "loss": 2.1262,
+      "step": 9512000
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.6237579311782144e-05,
+      "loss": 2.1256,
+      "step": 9512500
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.623685711143016e-05,
+      "loss": 2.122,
+      "step": 9513000
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.623613346378288e-05,
+      "loss": 2.124,
+      "step": 9513500
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.6235409816135604e-05,
+      "loss": 2.1347,
+      "step": 9514000
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.623468761578363e-05,
+      "loss": 2.1202,
+      "step": 9514500
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.623396396813635e-05,
+      "loss": 2.1505,
+      "step": 9515000
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.623324032048907e-05,
+      "loss": 2.1059,
+      "step": 9515500
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.6232516672841793e-05,
+      "loss": 2.1142,
+      "step": 9516000
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.6231793025194516e-05,
+      "loss": 2.1231,
+      "step": 9516500
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.623106937754724e-05,
+      "loss": 2.1407,
+      "step": 9517000
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.6230348624490554e-05,
+      "loss": 2.1399,
+      "step": 9517500
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.622962497684328e-05,
+      "loss": 2.1157,
+      "step": 9518000
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.6228901329196005e-05,
+      "loss": 2.1261,
+      "step": 9518500
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.622817768154873e-05,
+      "loss": 2.1342,
+      "step": 9519000
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.622745403390145e-05,
+      "loss": 2.1429,
+      "step": 9519500
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.622673038625417e-05,
+      "loss": 2.1262,
+      "step": 9520000
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.6226006738606894e-05,
+      "loss": 2.108,
+      "step": 9520500
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.6225283090959616e-05,
+      "loss": 2.1221,
+      "step": 9521000
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.622455944331234e-05,
+      "loss": 2.1313,
+      "step": 9521500
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.622383579566506e-05,
+      "loss": 2.1431,
+      "step": 9522000
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.622311214801778e-05,
+      "loss": 2.146,
+      "step": 9522500
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.6222389947665805e-05,
+      "loss": 2.1275,
+      "step": 9523000
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.622166630001853e-05,
+      "loss": 2.1547,
+      "step": 9523500
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.622094265237125e-05,
+      "loss": 2.1156,
+      "step": 9524000
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.6220220452019265e-05,
+      "loss": 2.1408,
+      "step": 9524500
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.6219496804371994e-05,
+      "loss": 2.1132,
+      "step": 9525000
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.6218773156724716e-05,
+      "loss": 2.1366,
+      "step": 9525500
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.621804950907744e-05,
+      "loss": 2.1272,
+      "step": 9526000
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.621732586143016e-05,
+      "loss": 2.1233,
+      "step": 9526500
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.621660221378288e-05,
+      "loss": 2.134,
+      "step": 9527000
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.6215878566135605e-05,
+      "loss": 2.1163,
+      "step": 9527500
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.6215154918488334e-05,
+      "loss": 2.1026,
+      "step": 9528000
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.6214431270841057e-05,
+      "loss": 2.1383,
+      "step": 9528500
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.621370907048907e-05,
+      "loss": 2.1563,
+      "step": 9529000
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.621298687013709e-05,
+      "loss": 2.118,
+      "step": 9529500
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.621226322248981e-05,
+      "loss": 2.1579,
+      "step": 9530000
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.621153957484253e-05,
+      "loss": 2.1466,
+      "step": 9530500
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.6210815927195254e-05,
+      "loss": 2.1263,
+      "step": 9531000
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.621009227954798e-05,
+      "loss": 2.1511,
+      "step": 9531500
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.6209368631900706e-05,
+      "loss": 2.1201,
+      "step": 9532000
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.6208644984253435e-05,
+      "loss": 2.0989,
+      "step": 9532500
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.620792133660616e-05,
+      "loss": 2.1107,
+      "step": 9533000
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.620719768895888e-05,
+      "loss": 2.1143,
+      "step": 9533500
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.6206475488606895e-05,
+      "loss": 2.1139,
+      "step": 9534000
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.620575328825491e-05,
+      "loss": 2.1432,
+      "step": 9534500
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.620502964060763e-05,
+      "loss": 2.1453,
+      "step": 9535000
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.6204307440255655e-05,
+      "loss": 2.148,
+      "step": 9535500
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.620358379260838e-05,
+      "loss": 2.1096,
+      "step": 9536000
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.62028601449611e-05,
+      "loss": 2.1463,
+      "step": 9536500
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.620213649731382e-05,
+      "loss": 2.1185,
+      "step": 9537000
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.6201412849666544e-05,
+      "loss": 2.1344,
+      "step": 9537500
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.6200689202019266e-05,
+      "loss": 2.1196,
+      "step": 9538000
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.619996555437199e-05,
+      "loss": 2.1241,
+      "step": 9538500
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.619924190672471e-05,
+      "loss": 2.1032,
+      "step": 9539000
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.619851970637273e-05,
+      "loss": 2.1291,
+      "step": 9539500
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.619779605872546e-05,
+      "loss": 2.1353,
+      "step": 9540000
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.619707385837348e-05,
+      "loss": 2.1274,
+      "step": 9540500
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.61963502107262e-05,
+      "loss": 2.1299,
+      "step": 9541000
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.619562656307892e-05,
+      "loss": 2.1234,
+      "step": 9541500
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.6194902915431644e-05,
+      "loss": 2.1596,
+      "step": 9542000
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.6194179267784366e-05,
+      "loss": 2.1227,
+      "step": 9542500
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.619345706743238e-05,
+      "loss": 2.1253,
+      "step": 9543000
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.619273341978511e-05,
+      "loss": 2.1426,
+      "step": 9543500
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.619200977213783e-05,
+      "loss": 2.1342,
+      "step": 9544000
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.6191286124490555e-05,
+      "loss": 2.1213,
+      "step": 9544500
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.619056392413857e-05,
+      "loss": 2.1249,
+      "step": 9545000
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.618984027649129e-05,
+      "loss": 2.1537,
+      "step": 9545500
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.6189116628844015e-05,
+      "loss": 2.1371,
+      "step": 9546000
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.618839442849203e-05,
+      "loss": 2.1244,
+      "step": 9546500
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.618767078084476e-05,
+      "loss": 2.1411,
+      "step": 9547000
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.618694713319748e-05,
+      "loss": 2.1214,
+      "step": 9547500
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.6186224932845504e-05,
+      "loss": 2.1341,
+      "step": 9548000
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.6185501285198227e-05,
+      "loss": 2.137,
+      "step": 9548500
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.618477763755095e-05,
+      "loss": 2.157,
+      "step": 9549000
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.618405398990367e-05,
+      "loss": 2.1509,
+      "step": 9549500
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.618333034225639e-05,
+      "loss": 2.1422,
+      "step": 9550000
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.6182606694609116e-05,
+      "loss": 2.1156,
+      "step": 9550500
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.618188304696184e-05,
+      "loss": 2.1268,
+      "step": 9551000
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.618116084660986e-05,
+      "loss": 2.1137,
+      "step": 9551500
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.618043719896258e-05,
+      "loss": 2.1199,
+      "step": 9552000
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.6179713551315305e-05,
+      "loss": 2.1365,
+      "step": 9552500
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.617898990366803e-05,
+      "loss": 2.1416,
+      "step": 9553000
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.617826625602075e-05,
+      "loss": 2.136,
+      "step": 9553500
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.6177544055668765e-05,
+      "loss": 2.1532,
+      "step": 9554000
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.617682040802149e-05,
+      "loss": 2.129,
+      "step": 9554500
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.617609676037421e-05,
+      "loss": 2.1294,
+      "step": 9555000
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.617537311272694e-05,
+      "loss": 2.1228,
+      "step": 9555500
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.617464946507966e-05,
+      "loss": 2.1441,
+      "step": 9556000
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.617392581743238e-05,
+      "loss": 2.1113,
+      "step": 9556500
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.617320216978511e-05,
+      "loss": 2.1351,
+      "step": 9557000
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.6172478522137834e-05,
+      "loss": 2.1173,
+      "step": 9557500
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.6171754874490556e-05,
+      "loss": 2.1271,
+      "step": 9558000
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.617103122684328e-05,
+      "loss": 2.1384,
+      "step": 9558500
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.6170307579196e-05,
+      "loss": 2.1536,
+      "step": 9559000
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.616958393154872e-05,
+      "loss": 2.1392,
+      "step": 9559500
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.6168860283901445e-05,
+      "loss": 2.1494,
+      "step": 9560000
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.616813663625417e-05,
+      "loss": 2.1201,
+      "step": 9560500
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.616741443590218e-05,
+      "loss": 2.1564,
+      "step": 9561000
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.616669078825491e-05,
+      "loss": 2.1187,
+      "step": 9561500
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.6165967140607634e-05,
+      "loss": 2.1649,
+      "step": 9562000
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.616524349296036e-05,
+      "loss": 2.1708,
+      "step": 9562500
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.6164519845313085e-05,
+      "loss": 2.1444,
+      "step": 9563000
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.616379619766581e-05,
+      "loss": 2.1387,
+      "step": 9563500
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.616307255001853e-05,
+      "loss": 2.1207,
+      "step": 9564000
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.616234890237125e-05,
+      "loss": 2.1487,
+      "step": 9564500
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.6161625254723974e-05,
+      "loss": 2.1186,
+      "step": 9565000
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.6160901607076696e-05,
+      "loss": 2.1254,
+      "step": 9565500
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.616017940672471e-05,
+      "loss": 2.1133,
+      "step": 9566000
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.6159455759077434e-05,
+      "loss": 2.1251,
+      "step": 9566500
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.615873211143016e-05,
+      "loss": 2.1488,
+      "step": 9567000
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.6158008463782885e-05,
+      "loss": 2.1059,
+      "step": 9567500
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.615728481613561e-05,
+      "loss": 2.1446,
+      "step": 9568000
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.615656116848833e-05,
+      "loss": 2.1164,
+      "step": 9568500
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.615584041543164e-05,
+      "loss": 2.1339,
+      "step": 9569000
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.615511676778436e-05,
+      "loss": 2.1331,
+      "step": 9569500
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.615439312013709e-05,
+      "loss": 2.1381,
+      "step": 9570000
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.615366947248981e-05,
+      "loss": 2.1504,
+      "step": 9570500
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.615294582484254e-05,
+      "loss": 2.1334,
+      "step": 9571000
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.6152222177195263e-05,
+      "loss": 2.1107,
+      "step": 9571500
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.6151498529547986e-05,
+      "loss": 2.1297,
+      "step": 9572000
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.6150776329196e-05,
+      "loss": 2.1263,
+      "step": 9572500
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.6150052681548724e-05,
+      "loss": 2.1453,
+      "step": 9573000
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.6149329033901446e-05,
+      "loss": 2.1341,
+      "step": 9573500
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.614860538625417e-05,
+      "loss": 2.1402,
+      "step": 9574000
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.614788318590219e-05,
+      "loss": 2.1449,
+      "step": 9574500
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.614715953825491e-05,
+      "loss": 2.1377,
+      "step": 9575000
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.614643733790293e-05,
+      "loss": 2.113,
+      "step": 9575500
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.614571369025565e-05,
+      "loss": 2.1473,
+      "step": 9576000
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.614499004260837e-05,
+      "loss": 2.1225,
+      "step": 9576500
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.6144266394961095e-05,
+      "loss": 2.1249,
+      "step": 9577000
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.6143542747313824e-05,
+      "loss": 2.1205,
+      "step": 9577500
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.6142819099666546e-05,
+      "loss": 2.1276,
+      "step": 9578000
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.614209545201927e-05,
+      "loss": 2.1068,
+      "step": 9578500
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.614137180437199e-05,
+      "loss": 2.1187,
+      "step": 9579000
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.614064815672471e-05,
+      "loss": 2.1292,
+      "step": 9579500
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.613992450907744e-05,
+      "loss": 2.1332,
+      "step": 9580000
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.6139200861430164e-05,
+      "loss": 2.1407,
+      "step": 9580500
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.6138477213782886e-05,
+      "loss": 2.1521,
+      "step": 9581000
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.61377550134309e-05,
+      "loss": 2.1114,
+      "step": 9581500
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.6137031365783624e-05,
+      "loss": 2.1432,
+      "step": 9582000
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.6136307718136346e-05,
+      "loss": 2.1442,
+      "step": 9582500
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.613558407048907e-05,
+      "loss": 2.1274,
+      "step": 9583000
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.613486187013709e-05,
+      "loss": 2.1041,
+      "step": 9583500
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.613413822248981e-05,
+      "loss": 2.1089,
+      "step": 9584000
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.6133414574842535e-05,
+      "loss": 2.1186,
+      "step": 9584500
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.6132690927195264e-05,
+      "loss": 2.1394,
+      "step": 9585000
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.6131967279547987e-05,
+      "loss": 2.1353,
+      "step": 9585500
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.613124363190071e-05,
+      "loss": 2.1187,
+      "step": 9586000
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.613051998425343e-05,
+      "loss": 2.1251,
+      "step": 9586500
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.612979633660615e-05,
+      "loss": 2.1643,
+      "step": 9587000
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.6129072688958875e-05,
+      "loss": 2.1161,
+      "step": 9587500
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.61283490413116e-05,
+      "loss": 2.1215,
+      "step": 9588000
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.612762539366432e-05,
+      "loss": 2.1219,
+      "step": 9588500
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 3.612690174601704e-05,
+      "loss": 2.1098,
+      "step": 9589000
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 3.6126178098369764e-05,
+      "loss": 2.1078,
+      "step": 9589500
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 3.612545445072249e-05,
+      "loss": 2.1452,
+      "step": 9590000
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 3.612473225037051e-05,
+      "loss": 2.1479,
+      "step": 9590500
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 3.612400860272323e-05,
+      "loss": 2.1195,
+      "step": 9591000
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 3.612328640237125e-05,
+      "loss": 2.1112,
+      "step": 9591500
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 3.612256275472397e-05,
+      "loss": 2.1198,
+      "step": 9592000
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 3.61218391070767e-05,
+      "loss": 2.1215,
+      "step": 9592500
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 3.612111545942942e-05,
+      "loss": 2.1268,
+      "step": 9593000
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 3.612039181178214e-05,
+      "loss": 2.1409,
+      "step": 9593500
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 3.6119669611430165e-05,
+      "loss": 2.1392,
+      "step": 9594000
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 3.611894596378289e-05,
+      "loss": 2.1109,
+      "step": 9594500
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 3.611822231613561e-05,
+      "loss": 2.1669,
+      "step": 9595000
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 3.611749866848833e-05,
+      "loss": 2.1258,
+      "step": 9595500
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 3.6116775020841054e-05,
+      "loss": 2.15,
+      "step": 9596000
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 3.6116051373193776e-05,
+      "loss": 2.1218,
+      "step": 9596500
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 3.61153277255465e-05,
+      "loss": 2.1138,
+      "step": 9597000
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 3.611460407789922e-05,
+      "loss": 2.1274,
+      "step": 9597500
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 3.611388043025194e-05,
+      "loss": 2.0926,
+      "step": 9598000
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 3.6113156782604665e-05,
+      "loss": 2.1139,
+      "step": 9598500
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 3.611243458225269e-05,
+      "loss": 2.1334,
+      "step": 9599000
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 3.6111710934605416e-05,
+      "loss": 2.1543,
+      "step": 9599500
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 3.6110990181548725e-05,
+      "loss": 2.107,
+      "step": 9600000
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 3.611026653390145e-05,
+      "loss": 2.1197,
+      "step": 9600500
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 3.610954288625417e-05,
+      "loss": 2.123,
+      "step": 9601000
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 3.610881923860689e-05,
+      "loss": 2.1228,
+      "step": 9601500
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 3.610809559095962e-05,
+      "loss": 2.1338,
+      "step": 9602000
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 3.610737194331234e-05,
+      "loss": 2.1162,
+      "step": 9602500
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 3.6106648295665065e-05,
+      "loss": 2.1359,
+      "step": 9603000
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 3.610592464801779e-05,
+      "loss": 2.1264,
+      "step": 9603500
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 3.610520100037051e-05,
+      "loss": 2.1134,
+      "step": 9604000
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 3.610447735272323e-05,
+      "loss": 2.1314,
+      "step": 9604500
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 3.610375515237125e-05,
+      "loss": 2.1377,
+      "step": 9605000
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 3.610303150472397e-05,
+      "loss": 2.1348,
+      "step": 9605500
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 3.610230785707669e-05,
+      "loss": 2.1335,
+      "step": 9606000
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 3.610158420942942e-05,
+      "loss": 2.1015,
+      "step": 9606500
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 3.610086056178214e-05,
+      "loss": 2.1356,
+      "step": 9607000
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 3.610013691413487e-05,
+      "loss": 2.1391,
+      "step": 9607500
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 3.6099413266487594e-05,
+      "loss": 2.1297,
+      "step": 9608000
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 3.609868961884032e-05,
+      "loss": 2.1328,
+      "step": 9608500
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 3.609796597119304e-05,
+      "loss": 2.1067,
+      "step": 9609000
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 3.6097243770841055e-05,
+      "loss": 2.15,
+      "step": 9609500
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 3.609652157048907e-05,
+      "loss": 2.1284,
+      "step": 9610000
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 3.609579792284179e-05,
+      "loss": 2.1375,
+      "step": 9610500
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 3.609507427519452e-05,
+      "loss": 2.1493,
+      "step": 9611000
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 3.6094350627547244e-05,
+      "loss": 2.1137,
+      "step": 9611500
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 3.609362842719526e-05,
+      "loss": 2.1391,
+      "step": 9612000
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 3.609290477954798e-05,
+      "loss": 2.1063,
+      "step": 9612500
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 3.6092181131900704e-05,
+      "loss": 2.1398,
+      "step": 9613000
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 3.6091457484253426e-05,
+      "loss": 2.1307,
+      "step": 9613500
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 3.609073383660615e-05,
+      "loss": 2.1197,
+      "step": 9614000
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 3.609001018895887e-05,
+      "loss": 2.1372,
+      "step": 9614500
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 3.60892865413116e-05,
+      "loss": 2.1388,
+      "step": 9615000
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 3.608856289366432e-05,
+      "loss": 2.1187,
+      "step": 9615500
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 3.6087839246017044e-05,
+      "loss": 2.1265,
+      "step": 9616000
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 3.608711559836977e-05,
+      "loss": 2.1271,
+      "step": 9616500
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 3.6086391950722495e-05,
+      "loss": 2.1419,
+      "step": 9617000
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 3.608566830307522e-05,
+      "loss": 2.1247,
+      "step": 9617500
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 3.608494610272323e-05,
+      "loss": 2.1075,
+      "step": 9618000
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 3.6084222455075955e-05,
+      "loss": 2.1124,
+      "step": 9618500
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 3.608349880742868e-05,
+      "loss": 2.1225,
+      "step": 9619000
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 3.60827751597814e-05,
+      "loss": 2.1242,
+      "step": 9619500
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 3.608205151213412e-05,
+      "loss": 2.1294,
+      "step": 9620000
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 3.6081327864486844e-05,
+      "loss": 2.13,
+      "step": 9620500
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 3.608060421683957e-05,
+      "loss": 2.1577,
+      "step": 9621000
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 3.6079880569192295e-05,
+      "loss": 2.1386,
+      "step": 9621500
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 3.6079156921545024e-05,
+      "loss": 2.1183,
+      "step": 9622000
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 3.6078433273897746e-05,
+      "loss": 2.1544,
+      "step": 9622500
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 3.607770962625047e-05,
+      "loss": 2.1098,
+      "step": 9623000
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 3.607698597860319e-05,
+      "loss": 2.1281,
+      "step": 9623500
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 3.6076263778251206e-05,
+      "loss": 2.1417,
+      "step": 9624000
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 3.607554013060393e-05,
+      "loss": 2.1485,
+      "step": 9624500
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 3.607481648295665e-05,
+      "loss": 2.1342,
+      "step": 9625000
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 3.607409283530937e-05,
+      "loss": 2.1379,
+      "step": 9625500
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 3.6073369187662095e-05,
+      "loss": 2.1298,
+      "step": 9626000
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 3.607264698731012e-05,
+      "loss": 2.1281,
+      "step": 9626500
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 3.607192478695813e-05,
+      "loss": 2.1611,
+      "step": 9627000
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 3.6071201139310855e-05,
+      "loss": 2.1368,
+      "step": 9627500
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 3.607047749166358e-05,
+      "loss": 2.1243,
+      "step": 9628000
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 3.60697538440163e-05,
+      "loss": 2.1295,
+      "step": 9628500
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 3.606903019636902e-05,
+      "loss": 2.1619,
+      "step": 9629000
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 3.606830654872175e-05,
+      "loss": 2.1307,
+      "step": 9629500
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 3.6067582901074473e-05,
+      "loss": 2.1454,
+      "step": 9630000
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 3.6066860700722496e-05,
+      "loss": 2.1367,
+      "step": 9630500
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 3.606613705307522e-05,
+      "loss": 2.1338,
+      "step": 9631000
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 3.606541340542794e-05,
+      "loss": 2.1179,
+      "step": 9631500
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 3.606468975778066e-05,
+      "loss": 2.1327,
+      "step": 9632000
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 3.6063966110133385e-05,
+      "loss": 2.1236,
+      "step": 9632500
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 3.606324246248611e-05,
+      "loss": 2.122,
+      "step": 9633000
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 3.606251881483883e-05,
+      "loss": 2.1371,
+      "step": 9633500
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 3.606179516719155e-05,
+      "loss": 2.1099,
+      "step": 9634000
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 3.6061071519544274e-05,
+      "loss": 2.1343,
+      "step": 9634500
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 3.6060347871896996e-05,
+      "loss": 2.134,
+      "step": 9635000
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 3.6059624224249725e-05,
+      "loss": 2.1288,
+      "step": 9635500
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 3.605890057660245e-05,
+      "loss": 2.1352,
+      "step": 9636000
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 3.6058176928955176e-05,
+      "loss": 2.107,
+      "step": 9636500
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 3.605745472860319e-05,
+      "loss": 2.1166,
+      "step": 9637000
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 3.605673252825121e-05,
+      "loss": 2.1099,
+      "step": 9637500
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 3.605600888060393e-05,
+      "loss": 2.1298,
+      "step": 9638000
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 3.605528523295665e-05,
+      "loss": 2.1625,
+      "step": 9638500
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 3.6054561585309374e-05,
+      "loss": 2.1411,
+      "step": 9639000
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 3.6053837937662096e-05,
+      "loss": 2.1446,
+      "step": 9639500
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 3.6053114290014825e-05,
+      "loss": 2.1408,
+      "step": 9640000
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 3.605239208966284e-05,
+      "loss": 2.0942,
+      "step": 9640500
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 3.605166844201556e-05,
+      "loss": 2.1201,
+      "step": 9641000
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 3.6050944794368285e-05,
+      "loss": 2.1447,
+      "step": 9641500
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 3.605022114672101e-05,
+      "loss": 2.107,
+      "step": 9642000
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 3.604949894636902e-05,
+      "loss": 2.1312,
+      "step": 9642500
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 3.604877529872175e-05,
+      "loss": 2.1211,
+      "step": 9643000
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 3.6048051651074474e-05,
+      "loss": 2.1299,
+      "step": 9643500
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 3.604732945072249e-05,
+      "loss": 2.1183,
+      "step": 9644000
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 3.604660580307522e-05,
+      "loss": 2.1179,
+      "step": 9644500
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 3.604588215542794e-05,
+      "loss": 2.1398,
+      "step": 9645000
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 3.604515850778066e-05,
+      "loss": 2.1322,
+      "step": 9645500
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 3.6044434860133386e-05,
+      "loss": 2.1187,
+      "step": 9646000
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 3.60437126597814e-05,
+      "loss": 2.1373,
+      "step": 9646500
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 3.604298901213412e-05,
+      "loss": 2.1187,
+      "step": 9647000
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 3.604226536448685e-05,
+      "loss": 2.1388,
+      "step": 9647500
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 3.6041541716839575e-05,
+      "loss": 2.1416,
+      "step": 9648000
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 3.604081951648759e-05,
+      "loss": 2.1082,
+      "step": 9648500
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 3.6040097316135606e-05,
+      "loss": 2.1092,
+      "step": 9649000
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 3.603937366848833e-05,
+      "loss": 2.1464,
+      "step": 9649500
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 3.603865002084105e-05,
+      "loss": 2.1291,
+      "step": 9650000
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 3.603792637319377e-05,
+      "loss": 2.1387,
+      "step": 9650500
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 3.60372027255465e-05,
+      "loss": 2.1787,
+      "step": 9651000
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 3.6036479077899224e-05,
+      "loss": 2.1305,
+      "step": 9651500
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 3.603575543025195e-05,
+      "loss": 2.1446,
+      "step": 9652000
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 3.603503322989997e-05,
+      "loss": 2.1378,
+      "step": 9652500
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 3.603430958225269e-05,
+      "loss": 2.1284,
+      "step": 9653000
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 3.603358593460541e-05,
+      "loss": 2.123,
+      "step": 9653500
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 3.6032862286958135e-05,
+      "loss": 2.1125,
+      "step": 9654000
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 3.603213863931086e-05,
+      "loss": 2.1343,
+      "step": 9654500
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 3.603141499166358e-05,
+      "loss": 2.1392,
+      "step": 9655000
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 3.60306913440163e-05,
+      "loss": 2.1062,
+      "step": 9655500
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 3.6029967696369024e-05,
+      "loss": 2.1388,
+      "step": 9656000
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 3.602924404872175e-05,
+      "loss": 2.1242,
+      "step": 9656500
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 3.6028520401074475e-05,
+      "loss": 2.1361,
+      "step": 9657000
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 3.60277967534272e-05,
+      "loss": 2.1455,
+      "step": 9657500
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 3.602707455307521e-05,
+      "loss": 2.1104,
+      "step": 9658000
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 3.602635235272323e-05,
+      "loss": 2.1188,
+      "step": 9658500
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 3.602562870507595e-05,
+      "loss": 2.1517,
+      "step": 9659000
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 3.602490505742868e-05,
+      "loss": 2.1288,
+      "step": 9659500
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 3.60241814097814e-05,
+      "loss": 2.1118,
+      "step": 9660000
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 3.6023457762134124e-05,
+      "loss": 2.1258,
+      "step": 9660500
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 3.602273411448685e-05,
+      "loss": 2.1268,
+      "step": 9661000
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 3.6022010466839575e-05,
+      "loss": 2.1035,
+      "step": 9661500
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 3.60212868191923e-05,
+      "loss": 2.1396,
+      "step": 9662000
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 3.602056461884031e-05,
+      "loss": 2.1087,
+      "step": 9662500
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 3.6019840971193035e-05,
+      "loss": 2.1088,
+      "step": 9663000
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 3.601911732354576e-05,
+      "loss": 2.1218,
+      "step": 9663500
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 3.601839367589848e-05,
+      "loss": 2.1318,
+      "step": 9664000
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 3.60176700282512e-05,
+      "loss": 2.1261,
+      "step": 9664500
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 3.6016946380603924e-05,
+      "loss": 2.1323,
+      "step": 9665000
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 3.601622273295665e-05,
+      "loss": 2.1172,
+      "step": 9665500
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 3.6015499085309376e-05,
+      "loss": 2.1254,
+      "step": 9666000
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 3.60147754376621e-05,
+      "loss": 2.1237,
+      "step": 9666500
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 3.601405179001483e-05,
+      "loss": 2.1486,
+      "step": 9667000
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 3.601332814236755e-05,
+      "loss": 2.1296,
+      "step": 9667500
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 3.601260449472027e-05,
+      "loss": 2.1071,
+      "step": 9668000
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 3.601188229436829e-05,
+      "loss": 2.1483,
+      "step": 9668500
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 3.60111600940163e-05,
+      "loss": 2.1151,
+      "step": 9669000
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 3.601043644636903e-05,
+      "loss": 2.1529,
+      "step": 9669500
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 3.6009712798721754e-05,
+      "loss": 2.1361,
+      "step": 9670000
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 3.6008989151074476e-05,
+      "loss": 2.1446,
+      "step": 9670500
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 3.60082655034272e-05,
+      "loss": 2.1146,
+      "step": 9671000
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 3.600754185577992e-05,
+      "loss": 2.1497,
+      "step": 9671500
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 3.600681820813264e-05,
+      "loss": 2.1167,
+      "step": 9672000
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 3.600609600778066e-05,
+      "loss": 2.1237,
+      "step": 9672500
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 3.600537236013338e-05,
+      "loss": 2.1563,
+      "step": 9673000
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.6667694438971397,
+      "eval_accuracy_mlm": 0.6312277095277679,
+      "eval_accuracy_nsp": 0.8573396030120123,
+      "eval_loss": 2.182663917541504,
+      "eval_runtime": 331.438,
+      "eval_samples_per_second": 1316.645,
+      "eval_steps_per_second": 54.861,
+      "step": 9673216
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 3.60046487124861e-05,
+      "loss": 2.1061,
+      "step": 9673500
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 3.600392506483883e-05,
+      "loss": 2.1082,
+      "step": 9674000
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 3.6003201417191554e-05,
+      "loss": 2.1198,
+      "step": 9674500
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 3.600247776954428e-05,
+      "loss": 2.1286,
+      "step": 9675000
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 3.6001754121897005e-05,
+      "loss": 2.1053,
+      "step": 9675500
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 3.600103047424973e-05,
+      "loss": 2.1165,
+      "step": 9676000
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 3.600030827389774e-05,
+      "loss": 2.0749,
+      "step": 9676500
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 3.5999584626250465e-05,
+      "loss": 2.107,
+      "step": 9677000
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 3.599886097860319e-05,
+      "loss": 2.1012,
+      "step": 9677500
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 3.599813733095591e-05,
+      "loss": 2.074,
+      "step": 9678000
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 3.599741513060393e-05,
+      "loss": 2.0736,
+      "step": 9678500
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 3.5996691482956654e-05,
+      "loss": 2.1095,
+      "step": 9679000
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 3.5995967835309376e-05,
+      "loss": 2.1076,
+      "step": 9679500
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 3.59952441876621e-05,
+      "loss": 2.0927,
+      "step": 9680000
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 3.599452054001482e-05,
+      "loss": 2.0745,
+      "step": 9680500
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 3.599379689236754e-05,
+      "loss": 2.1202,
+      "step": 9681000
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 3.599307324472027e-05,
+      "loss": 2.1223,
+      "step": 9681500
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 3.599235104436829e-05,
+      "loss": 2.0964,
+      "step": 9682000
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 3.599162739672101e-05,
+      "loss": 2.099,
+      "step": 9682500
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 3.599090374907373e-05,
+      "loss": 2.1141,
+      "step": 9683000
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 3.5990180101426454e-05,
+      "loss": 2.1131,
+      "step": 9683500
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 3.598945645377918e-05,
+      "loss": 2.1317,
+      "step": 9684000
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 3.59887342534272e-05,
+      "loss": 2.0862,
+      "step": 9684500
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 3.598801060577992e-05,
+      "loss": 2.1147,
+      "step": 9685000
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 3.598728695813264e-05,
+      "loss": 2.0986,
+      "step": 9685500
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 3.5986563310485366e-05,
+      "loss": 2.1117,
+      "step": 9686000
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 3.598583966283809e-05,
+      "loss": 2.1287,
+      "step": 9686500
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 3.59851174624861e-05,
+      "loss": 2.1126,
+      "step": 9687000
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 3.598439381483883e-05,
+      "loss": 2.097,
+      "step": 9687500
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 3.598367161448685e-05,
+      "loss": 2.1014,
+      "step": 9688000
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 3.598294796683957e-05,
+      "loss": 2.1132,
+      "step": 9688500
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 3.598222431919229e-05,
+      "loss": 2.0837,
+      "step": 9689000
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 3.598150067154502e-05,
+      "loss": 2.1264,
+      "step": 9689500
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 3.5980777023897744e-05,
+      "loss": 2.1075,
+      "step": 9690000
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 3.5980053376250466e-05,
+      "loss": 2.1201,
+      "step": 9690500
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 3.597932972860319e-05,
+      "loss": 2.1236,
+      "step": 9691000
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 3.597860608095591e-05,
+      "loss": 2.1086,
+      "step": 9691500
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 3.597788243330863e-05,
+      "loss": 2.1238,
+      "step": 9692000
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 3.597716168025195e-05,
+      "loss": 2.1132,
+      "step": 9692500
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 3.597643803260467e-05,
+      "loss": 2.1133,
+      "step": 9693000
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 3.597571438495739e-05,
+      "loss": 2.1301,
+      "step": 9693500
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 3.597499218460541e-05,
+      "loss": 2.1253,
+      "step": 9694000
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 3.597426853695813e-05,
+      "loss": 2.1069,
+      "step": 9694500
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 3.597354488931086e-05,
+      "loss": 2.1051,
+      "step": 9695000
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 3.597282124166358e-05,
+      "loss": 2.1015,
+      "step": 9695500
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 3.5972097594016304e-05,
+      "loss": 2.1002,
+      "step": 9696000
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 3.5971373946369026e-05,
+      "loss": 2.1073,
+      "step": 9696500
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 3.5970650298721755e-05,
+      "loss": 2.113,
+      "step": 9697000
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 3.596992665107448e-05,
+      "loss": 2.1286,
+      "step": 9697500
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 3.59692030034272e-05,
+      "loss": 2.0978,
+      "step": 9698000
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 3.596847935577992e-05,
+      "loss": 2.1021,
+      "step": 9698500
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 3.5967755708132644e-05,
+      "loss": 2.1171,
+      "step": 9699000
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 3.5967032060485366e-05,
+      "loss": 2.1204,
+      "step": 9699500
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 3.596630841283809e-05,
+      "loss": 2.1066,
+      "step": 9700000
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 3.596558476519081e-05,
+      "loss": 2.1341,
+      "step": 9700500
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 3.596486111754353e-05,
+      "loss": 2.0861,
+      "step": 9701000
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 3.5964137469896255e-05,
+      "loss": 2.1171,
+      "step": 9701500
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 3.5963413822248984e-05,
+      "loss": 2.1173,
+      "step": 9702000
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 3.5962690174601707e-05,
+      "loss": 2.1187,
+      "step": 9702500
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 3.596196652695443e-05,
+      "loss": 2.1059,
+      "step": 9703000
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 3.596124287930715e-05,
+      "loss": 2.1111,
+      "step": 9703500
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 3.596051923165988e-05,
+      "loss": 2.0957,
+      "step": 9704000
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 3.59597955840126e-05,
+      "loss": 2.0935,
+      "step": 9704500
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 3.595907338366062e-05,
+      "loss": 2.1162,
+      "step": 9705000
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 3.595834973601334e-05,
+      "loss": 2.0782,
+      "step": 9705500
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 3.595762608836606e-05,
+      "loss": 2.1191,
+      "step": 9706000
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 3.5956902440718784e-05,
+      "loss": 2.111,
+      "step": 9706500
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 3.595617879307151e-05,
+      "loss": 2.1414,
+      "step": 9707000
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 3.5955455145424236e-05,
+      "loss": 2.1114,
+      "step": 9707500
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 3.595473294507225e-05,
+      "loss": 2.1124,
+      "step": 9708000
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 3.5954009297424974e-05,
+      "loss": 2.122,
+      "step": 9708500
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 3.5953285649777696e-05,
+      "loss": 2.1112,
+      "step": 9709000
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 3.595256200213042e-05,
+      "loss": 2.1096,
+      "step": 9709500
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 3.5951839801778434e-05,
+      "loss": 2.0982,
+      "step": 9710000
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 3.595111615413116e-05,
+      "loss": 2.1295,
+      "step": 9710500
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 3.5950392506483885e-05,
+      "loss": 2.1374,
+      "step": 9711000
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 3.594967030613191e-05,
+      "loss": 2.1336,
+      "step": 9711500
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 3.594894665848463e-05,
+      "loss": 2.0976,
+      "step": 9712000
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 3.594822301083735e-05,
+      "loss": 2.1118,
+      "step": 9712500
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 3.5947499363190074e-05,
+      "loss": 2.1132,
+      "step": 9713000
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 3.5946775715542796e-05,
+      "loss": 2.1001,
+      "step": 9713500
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 3.594605351519081e-05,
+      "loss": 2.1274,
+      "step": 9714000
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 3.5945329867543534e-05,
+      "loss": 2.1312,
+      "step": 9714500
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 3.594460621989626e-05,
+      "loss": 2.0991,
+      "step": 9715000
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 3.5943882572248985e-05,
+      "loss": 2.1227,
+      "step": 9715500
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 3.594315892460171e-05,
+      "loss": 2.1156,
+      "step": 9716000
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 3.594243527695443e-05,
+      "loss": 2.1033,
+      "step": 9716500
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 3.594171162930715e-05,
+      "loss": 2.1374,
+      "step": 9717000
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 3.5940987981659874e-05,
+      "loss": 2.1077,
+      "step": 9717500
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 3.594026578130789e-05,
+      "loss": 2.1252,
+      "step": 9718000
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 3.593954213366061e-05,
+      "loss": 2.116,
+      "step": 9718500
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 3.593881848601334e-05,
+      "loss": 2.1305,
+      "step": 9719000
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 3.593809483836606e-05,
+      "loss": 2.1143,
+      "step": 9719500
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 3.5937371190718785e-05,
+      "loss": 2.1179,
+      "step": 9720000
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 3.593664899036681e-05,
+      "loss": 2.1189,
+      "step": 9720500
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 3.593592534271953e-05,
+      "loss": 2.1159,
+      "step": 9721000
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 3.593520169507225e-05,
+      "loss": 2.106,
+      "step": 9721500
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 3.5934478047424974e-05,
+      "loss": 2.1107,
+      "step": 9722000
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 3.5933754399777697e-05,
+      "loss": 2.1076,
+      "step": 9722500
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 3.593303219942571e-05,
+      "loss": 2.0999,
+      "step": 9723000
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 3.5932308551778434e-05,
+      "loss": 2.1012,
+      "step": 9723500
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 3.593158490413116e-05,
+      "loss": 2.1255,
+      "step": 9724000
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 3.5930861256483886e-05,
+      "loss": 2.1432,
+      "step": 9724500
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 3.593013760883661e-05,
+      "loss": 2.1175,
+      "step": 9725000
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 3.592941396118933e-05,
+      "loss": 2.0825,
+      "step": 9725500
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 3.592869031354205e-05,
+      "loss": 2.1055,
+      "step": 9726000
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 3.592796666589478e-05,
+      "loss": 2.0972,
+      "step": 9726500
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 3.59272444655428e-05,
+      "loss": 2.1253,
+      "step": 9727000
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 3.592652226519081e-05,
+      "loss": 2.1051,
+      "step": 9727500
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 3.5925798617543535e-05,
+      "loss": 2.113,
+      "step": 9728000
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 3.5925074969896264e-05,
+      "loss": 2.1197,
+      "step": 9728500
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 3.5924351322248986e-05,
+      "loss": 2.1027,
+      "step": 9729000
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 3.592362767460171e-05,
+      "loss": 2.1086,
+      "step": 9729500
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 3.592290402695443e-05,
+      "loss": 2.1291,
+      "step": 9730000
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 3.592218037930715e-05,
+      "loss": 2.1228,
+      "step": 9730500
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 3.592145817895517e-05,
+      "loss": 2.0911,
+      "step": 9731000
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 3.592073453130789e-05,
+      "loss": 2.0891,
+      "step": 9731500
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 3.592001088366061e-05,
+      "loss": 2.1078,
+      "step": 9732000
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 3.5919287236013335e-05,
+      "loss": 2.1269,
+      "step": 9732500
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 3.5918563588366064e-05,
+      "loss": 2.113,
+      "step": 9733000
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 3.5917839940718786e-05,
+      "loss": 2.1318,
+      "step": 9733500
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 3.5917116293071515e-05,
+      "loss": 2.1302,
+      "step": 9734000
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 3.591639264542424e-05,
+      "loss": 2.1092,
+      "step": 9734500
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 3.591566899777696e-05,
+      "loss": 2.1128,
+      "step": 9735000
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 3.591494535012968e-05,
+      "loss": 2.1336,
+      "step": 9735500
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 3.5914221702482404e-05,
+      "loss": 2.129,
+      "step": 9736000
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 3.591349950213042e-05,
+      "loss": 2.1164,
+      "step": 9736500
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 3.591277585448314e-05,
+      "loss": 2.1208,
+      "step": 9737000
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 3.5912052206835864e-05,
+      "loss": 2.1181,
+      "step": 9737500
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 3.5911330006483886e-05,
+      "loss": 2.1108,
+      "step": 9738000
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 3.591060635883661e-05,
+      "loss": 2.1379,
+      "step": 9738500
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 3.5909884158484624e-05,
+      "loss": 2.1188,
+      "step": 9739000
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 3.5909160510837346e-05,
+      "loss": 2.1071,
+      "step": 9739500
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 3.590843686319007e-05,
+      "loss": 2.1368,
+      "step": 9740000
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 3.590771321554279e-05,
+      "loss": 2.1228,
+      "step": 9740500
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 3.590698956789551e-05,
+      "loss": 2.0931,
+      "step": 9741000
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 3.590626736754354e-05,
+      "loss": 2.1131,
+      "step": 9741500
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 3.5905543719896264e-05,
+      "loss": 2.1081,
+      "step": 9742000
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 3.590482007224899e-05,
+      "loss": 2.1264,
+      "step": 9742500
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 3.590409642460171e-05,
+      "loss": 2.1113,
+      "step": 9743000
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 3.590337277695443e-05,
+      "loss": 2.1298,
+      "step": 9743500
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 3.590264912930715e-05,
+      "loss": 2.0918,
+      "step": 9744000
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 3.5901925481659876e-05,
+      "loss": 2.1193,
+      "step": 9744500
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 3.590120328130789e-05,
+      "loss": 2.1092,
+      "step": 9745000
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 3.5900481080955914e-05,
+      "loss": 2.1235,
+      "step": 9745500
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 3.5899757433308636e-05,
+      "loss": 2.133,
+      "step": 9746000
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 3.589903378566136e-05,
+      "loss": 2.1233,
+      "step": 9746500
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 3.589831013801408e-05,
+      "loss": 2.1275,
+      "step": 9747000
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 3.58975864903668e-05,
+      "loss": 2.1009,
+      "step": 9747500
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 3.5896862842719525e-05,
+      "loss": 2.0962,
+      "step": 9748000
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 3.589613919507225e-05,
+      "loss": 2.1337,
+      "step": 9748500
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 3.5895415547424976e-05,
+      "loss": 2.1135,
+      "step": 9749000
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 3.58946918997777e-05,
+      "loss": 2.1099,
+      "step": 9749500
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 3.589396825213042e-05,
+      "loss": 2.1018,
+      "step": 9750000
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 3.589324460448314e-05,
+      "loss": 2.1461,
+      "step": 9750500
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 3.5892520956835865e-05,
+      "loss": 2.123,
+      "step": 9751000
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 3.5891797309188594e-05,
+      "loss": 2.1269,
+      "step": 9751500
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 3.5891073661541316e-05,
+      "loss": 2.1197,
+      "step": 9752000
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 3.589035001389404e-05,
+      "loss": 2.1178,
+      "step": 9752500
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 3.588962636624676e-05,
+      "loss": 2.0853,
+      "step": 9753000
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 3.588890271859948e-05,
+      "loss": 2.0897,
+      "step": 9753500
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 3.5888179070952205e-05,
+      "loss": 2.1092,
+      "step": 9754000
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 3.588745542330493e-05,
+      "loss": 2.1079,
+      "step": 9754500
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 3.588673177565765e-05,
+      "loss": 2.12,
+      "step": 9755000
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 3.5886009575305665e-05,
+      "loss": 2.1188,
+      "step": 9755500
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 3.5885285927658394e-05,
+      "loss": 2.1172,
+      "step": 9756000
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 3.5884562280011116e-05,
+      "loss": 2.0877,
+      "step": 9756500
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 3.5883838632363845e-05,
+      "loss": 2.1093,
+      "step": 9757000
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 3.588311498471657e-05,
+      "loss": 2.0726,
+      "step": 9757500
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 3.588239278436458e-05,
+      "loss": 2.1199,
+      "step": 9758000
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 3.58816705840126e-05,
+      "loss": 2.1303,
+      "step": 9758500
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 3.588094693636532e-05,
+      "loss": 2.1251,
+      "step": 9759000
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 3.588022328871804e-05,
+      "loss": 2.1037,
+      "step": 9759500
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 3.5879501088366065e-05,
+      "loss": 2.1147,
+      "step": 9760000
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 3.587877744071879e-05,
+      "loss": 2.1048,
+      "step": 9760500
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 3.587805379307151e-05,
+      "loss": 2.1051,
+      "step": 9761000
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 3.587733014542423e-05,
+      "loss": 2.1207,
+      "step": 9761500
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 3.5876606497776954e-05,
+      "loss": 2.108,
+      "step": 9762000
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 3.587588285012968e-05,
+      "loss": 2.1296,
+      "step": 9762500
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 3.58751592024824e-05,
+      "loss": 2.1368,
+      "step": 9763000
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 3.587443555483513e-05,
+      "loss": 2.1373,
+      "step": 9763500
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 3.587371190718785e-05,
+      "loss": 2.1152,
+      "step": 9764000
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 3.5872991154131166e-05,
+      "loss": 2.0997,
+      "step": 9764500
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 3.587226750648389e-05,
+      "loss": 2.1188,
+      "step": 9765000
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 3.587154385883661e-05,
+      "loss": 2.121,
+      "step": 9765500
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 3.587082021118933e-05,
+      "loss": 2.1049,
+      "step": 9766000
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 3.587009801083735e-05,
+      "loss": 2.1265,
+      "step": 9766500
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 3.586937436319007e-05,
+      "loss": 2.1336,
+      "step": 9767000
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 3.586865071554279e-05,
+      "loss": 2.118,
+      "step": 9767500
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 3.586792706789552e-05,
+      "loss": 2.0857,
+      "step": 9768000
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 3.5867203420248244e-05,
+      "loss": 2.1091,
+      "step": 9768500
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 3.5866479772600966e-05,
+      "loss": 2.1117,
+      "step": 9769000
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 3.586575612495369e-05,
+      "loss": 2.1078,
+      "step": 9769500
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 3.586503247730641e-05,
+      "loss": 2.1325,
+      "step": 9770000
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 3.5864310276954426e-05,
+      "loss": 2.0993,
+      "step": 9770500
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 3.586358662930715e-05,
+      "loss": 2.1342,
+      "step": 9771000
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 3.586286298165988e-05,
+      "loss": 2.1069,
+      "step": 9771500
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 3.58621393340126e-05,
+      "loss": 2.1118,
+      "step": 9772000
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 3.586141568636532e-05,
+      "loss": 2.1175,
+      "step": 9772500
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 3.5860692038718044e-05,
+      "loss": 2.098,
+      "step": 9773000
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 3.585996839107077e-05,
+      "loss": 2.1196,
+      "step": 9773500
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 3.5859244743423495e-05,
+      "loss": 2.1388,
+      "step": 9774000
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 3.585852109577622e-05,
+      "loss": 2.1055,
+      "step": 9774500
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 3.585779889542423e-05,
+      "loss": 2.1197,
+      "step": 9775000
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 3.585707669507225e-05,
+      "loss": 2.1116,
+      "step": 9775500
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 3.585635304742497e-05,
+      "loss": 2.1101,
+      "step": 9776000
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 3.585562939977769e-05,
+      "loss": 2.1035,
+      "step": 9776500
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 3.585490575213042e-05,
+      "loss": 2.1427,
+      "step": 9777000
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 3.585418355177844e-05,
+      "loss": 2.1214,
+      "step": 9777500
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 3.585345990413116e-05,
+      "loss": 2.1162,
+      "step": 9778000
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 3.585273625648388e-05,
+      "loss": 2.114,
+      "step": 9778500
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 3.5852014056131904e-05,
+      "loss": 2.0958,
+      "step": 9779000
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 3.5851290408484627e-05,
+      "loss": 2.1082,
+      "step": 9779500
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 3.585056676083735e-05,
+      "loss": 2.1168,
+      "step": 9780000
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 3.584984456048537e-05,
+      "loss": 2.1255,
+      "step": 9780500
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 3.584912091283809e-05,
+      "loss": 2.1245,
+      "step": 9781000
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 3.5848397265190816e-05,
+      "loss": 2.1295,
+      "step": 9781500
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 3.584767361754354e-05,
+      "loss": 2.1004,
+      "step": 9782000
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 3.584694996989626e-05,
+      "loss": 2.1135,
+      "step": 9782500
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 3.584622632224898e-05,
+      "loss": 2.1314,
+      "step": 9783000
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 3.5845502674601705e-05,
+      "loss": 2.1205,
+      "step": 9783500
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 3.584477902695443e-05,
+      "loss": 2.1112,
+      "step": 9784000
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 3.584405682660244e-05,
+      "loss": 2.1346,
+      "step": 9784500
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 3.584333317895517e-05,
+      "loss": 2.1227,
+      "step": 9785000
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 3.5842609531307894e-05,
+      "loss": 2.1237,
+      "step": 9785500
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 3.5841885883660616e-05,
+      "loss": 2.1023,
+      "step": 9786000
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 3.5841162236013345e-05,
+      "loss": 2.1134,
+      "step": 9786500
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 3.584043858836607e-05,
+      "loss": 2.1003,
+      "step": 9787000
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 3.583971494071879e-05,
+      "loss": 2.0957,
+      "step": 9787500
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 3.583899129307151e-05,
+      "loss": 2.1128,
+      "step": 9788000
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 3.583826909271953e-05,
+      "loss": 2.1329,
+      "step": 9788500
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 3.583754544507225e-05,
+      "loss": 2.1094,
+      "step": 9789000
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 3.583682179742497e-05,
+      "loss": 2.1169,
+      "step": 9789500
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 3.5836098149777694e-05,
+      "loss": 2.135,
+      "step": 9790000
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 3.583537450213042e-05,
+      "loss": 2.1215,
+      "step": 9790500
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 3.5834650854483145e-05,
+      "loss": 2.0952,
+      "step": 9791000
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 3.583392720683587e-05,
+      "loss": 2.1154,
+      "step": 9791500
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 3.583320355918859e-05,
+      "loss": 2.1077,
+      "step": 9792000
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 3.583247991154131e-05,
+      "loss": 2.0838,
+      "step": 9792500
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 3.5831756263894034e-05,
+      "loss": 2.1259,
+      "step": 9793000
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 3.583103261624676e-05,
+      "loss": 2.1493,
+      "step": 9793500
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 3.5830308968599485e-05,
+      "loss": 2.1401,
+      "step": 9794000
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 3.582958532095221e-05,
+      "loss": 2.0956,
+      "step": 9794500
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 3.582886167330493e-05,
+      "loss": 2.1065,
+      "step": 9795000
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 3.582813802565765e-05,
+      "loss": 2.1162,
+      "step": 9795500
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 3.5827414378010374e-05,
+      "loss": 2.1199,
+      "step": 9796000
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 3.5826690730363096e-05,
+      "loss": 2.1204,
+      "step": 9796500
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 3.5825967082715825e-05,
+      "loss": 2.1336,
+      "step": 9797000
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 3.582524343506855e-05,
+      "loss": 2.0934,
+      "step": 9797500
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 3.582452123471656e-05,
+      "loss": 2.1321,
+      "step": 9798000
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 3.5823797587069285e-05,
+      "loss": 2.1188,
+      "step": 9798500
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 3.582307393942201e-05,
+      "loss": 2.1139,
+      "step": 9799000
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 3.582235029177473e-05,
+      "loss": 2.1445,
+      "step": 9799500
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 3.5821628091422745e-05,
+      "loss": 2.1099,
+      "step": 9800000
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 3.5820904443775474e-05,
+      "loss": 2.1213,
+      "step": 9800500
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 3.58201807961282e-05,
+      "loss": 2.1135,
+      "step": 9801000
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 3.581945859577622e-05,
+      "loss": 2.1255,
+      "step": 9801500
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 3.581873494812894e-05,
+      "loss": 2.1185,
+      "step": 9802000
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 3.5818011300481663e-05,
+      "loss": 2.1169,
+      "step": 9802500
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 3.5817287652834386e-05,
+      "loss": 2.1245,
+      "step": 9803000
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 3.581656400518711e-05,
+      "loss": 2.113,
+      "step": 9803500
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 3.581584035753983e-05,
+      "loss": 2.1206,
+      "step": 9804000
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 3.581511670989255e-05,
+      "loss": 2.1102,
+      "step": 9804500
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 3.5814393062245275e-05,
+      "loss": 2.1223,
+      "step": 9805000
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 3.5813669414598e-05,
+      "loss": 2.1247,
+      "step": 9805500
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 3.5812945766950726e-05,
+      "loss": 2.1378,
+      "step": 9806000
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 3.581222211930345e-05,
+      "loss": 2.1345,
+      "step": 9806500
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 3.581149847165617e-05,
+      "loss": 2.1241,
+      "step": 9807000
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 3.581077482400889e-05,
+      "loss": 2.1054,
+      "step": 9807500
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 3.581005262365691e-05,
+      "loss": 2.1356,
+      "step": 9808000
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 3.580932897600964e-05,
+      "loss": 2.1182,
+      "step": 9808500
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 3.580860677565765e-05,
+      "loss": 2.1131,
+      "step": 9809000
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 3.5807883128010375e-05,
+      "loss": 2.1035,
+      "step": 9809500
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 3.58071594803631e-05,
+      "loss": 2.1201,
+      "step": 9810000
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 3.5806435832715826e-05,
+      "loss": 2.1187,
+      "step": 9810500
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 3.580571218506855e-05,
+      "loss": 2.131,
+      "step": 9811000
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 3.580498853742127e-05,
+      "loss": 2.126,
+      "step": 9811500
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 3.580426488977399e-05,
+      "loss": 2.1069,
+      "step": 9812000
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 3.5803541242126715e-05,
+      "loss": 2.1087,
+      "step": 9812500
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 3.580281904177473e-05,
+      "loss": 2.1309,
+      "step": 9813000
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 3.580209539412745e-05,
+      "loss": 2.1127,
+      "step": 9813500
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 3.5801371746480175e-05,
+      "loss": 2.1318,
+      "step": 9814000
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 3.58006480988329e-05,
+      "loss": 2.1047,
+      "step": 9814500
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 3.579992734577621e-05,
+      "loss": 2.1024,
+      "step": 9815000
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 3.5799203698128935e-05,
+      "loss": 2.1306,
+      "step": 9815500
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 3.5798480050481664e-05,
+      "loss": 2.1236,
+      "step": 9816000
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 3.5797756402834386e-05,
+      "loss": 2.1389,
+      "step": 9816500
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 3.579703275518711e-05,
+      "loss": 2.1108,
+      "step": 9817000
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 3.5796310554835124e-05,
+      "loss": 2.1106,
+      "step": 9817500
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 3.579558690718785e-05,
+      "loss": 2.1048,
+      "step": 9818000
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 3.5794863259540576e-05,
+      "loss": 2.1417,
+      "step": 9818500
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 3.57941396118933e-05,
+      "loss": 2.1071,
+      "step": 9819000
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 3.579341741154131e-05,
+      "loss": 2.1094,
+      "step": 9819500
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 3.5792693763894036e-05,
+      "loss": 2.1025,
+      "step": 9820000
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 3.579197011624676e-05,
+      "loss": 2.118,
+      "step": 9820500
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 3.579124791589477e-05,
+      "loss": 2.1428,
+      "step": 9821000
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 3.57905242682475e-05,
+      "loss": 2.1052,
+      "step": 9821500
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 3.5789800620600225e-05,
+      "loss": 2.1111,
+      "step": 9822000
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 3.578907697295295e-05,
+      "loss": 2.1208,
+      "step": 9822500
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 3.578835332530567e-05,
+      "loss": 2.1222,
+      "step": 9823000
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 3.57876296776584e-05,
+      "loss": 2.1196,
+      "step": 9823500
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 3.578690603001112e-05,
+      "loss": 2.1002,
+      "step": 9824000
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 3.578618238236384e-05,
+      "loss": 2.1224,
+      "step": 9824500
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 3.5785458734716565e-05,
+      "loss": 2.1259,
+      "step": 9825000
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 3.578473508706929e-05,
+      "loss": 2.1108,
+      "step": 9825500
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 3.578401143942201e-05,
+      "loss": 2.1137,
+      "step": 9826000
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 3.578328779177473e-05,
+      "loss": 2.1241,
+      "step": 9826500
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 3.5782564144127454e-05,
+      "loss": 2.0793,
+      "step": 9827000
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 3.5781840496480176e-05,
+      "loss": 2.1182,
+      "step": 9827500
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 3.5781116848832905e-05,
+      "loss": 2.0881,
+      "step": 9828000
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 3.5780396095776214e-05,
+      "loss": 2.1403,
+      "step": 9828500
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 3.5779672448128936e-05,
+      "loss": 2.0986,
+      "step": 9829000
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 3.577894880048166e-05,
+      "loss": 2.1291,
+      "step": 9829500
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 3.577822515283438e-05,
+      "loss": 2.0975,
+      "step": 9830000
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 3.57775015051871e-05,
+      "loss": 2.1168,
+      "step": 9830500
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 3.577677785753983e-05,
+      "loss": 2.1261,
+      "step": 9831000
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 3.5776054209892554e-05,
+      "loss": 2.1234,
+      "step": 9831500
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 3.5775330562245276e-05,
+      "loss": 2.1235,
+      "step": 9832000
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 3.5774606914598005e-05,
+      "loss": 2.1073,
+      "step": 9832500
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 3.577388326695073e-05,
+      "loss": 2.1105,
+      "step": 9833000
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 3.577315961930345e-05,
+      "loss": 2.1124,
+      "step": 9833500
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 3.577243597165617e-05,
+      "loss": 2.1292,
+      "step": 9834000
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 3.5771712324008894e-05,
+      "loss": 2.1294,
+      "step": 9834500
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 3.5770988676361616e-05,
+      "loss": 2.1268,
+      "step": 9835000
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 3.577026502871434e-05,
+      "loss": 2.1317,
+      "step": 9835500
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 3.576954138106706e-05,
+      "loss": 2.105,
+      "step": 9836000
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 3.576881773341978e-05,
+      "loss": 2.1292,
+      "step": 9836500
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 3.5768095533067805e-05,
+      "loss": 2.1137,
+      "step": 9837000
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 3.576737333271582e-05,
+      "loss": 2.1355,
+      "step": 9837500
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 3.576664968506854e-05,
+      "loss": 2.1267,
+      "step": 9838000
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 3.576592603742127e-05,
+      "loss": 2.1257,
+      "step": 9838500
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 3.5765202389773994e-05,
+      "loss": 2.1247,
+      "step": 9839000
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 3.576448018942201e-05,
+      "loss": 2.1351,
+      "step": 9839500
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 3.576375798907003e-05,
+      "loss": 2.1,
+      "step": 9840000
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 3.5763034341422755e-05,
+      "loss": 2.093,
+      "step": 9840500
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 3.576231069377548e-05,
+      "loss": 2.1247,
+      "step": 9841000
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 3.57615870461282e-05,
+      "loss": 2.112,
+      "step": 9841500
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 3.576086339848092e-05,
+      "loss": 2.1111,
+      "step": 9842000
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 3.5760139750833644e-05,
+      "loss": 2.1213,
+      "step": 9842500
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 3.575941755048166e-05,
+      "loss": 2.1113,
+      "step": 9843000
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 3.575869535012968e-05,
+      "loss": 2.1223,
+      "step": 9843500
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 3.5757971702482404e-05,
+      "loss": 2.1162,
+      "step": 9844000
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 3.5757248054835126e-05,
+      "loss": 2.1232,
+      "step": 9844500
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 3.575652440718785e-05,
+      "loss": 2.1208,
+      "step": 9845000
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 3.5755802206835864e-05,
+      "loss": 2.0977,
+      "step": 9845500
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 3.575507855918859e-05,
+      "loss": 2.1307,
+      "step": 9846000
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 3.5754354911541315e-05,
+      "loss": 2.1225,
+      "step": 9846500
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 3.575363126389404e-05,
+      "loss": 2.1274,
+      "step": 9847000
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 3.575290906354205e-05,
+      "loss": 2.1438,
+      "step": 9847500
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 3.575218541589478e-05,
+      "loss": 2.1223,
+      "step": 9848000
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 3.5751461768247504e-05,
+      "loss": 2.1202,
+      "step": 9848500
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 3.5750738120600226e-05,
+      "loss": 2.1022,
+      "step": 9849000
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 3.575001447295295e-05,
+      "loss": 2.104,
+      "step": 9849500
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 3.574929082530567e-05,
+      "loss": 2.1203,
+      "step": 9850000
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 3.574856717765839e-05,
+      "loss": 2.1342,
+      "step": 9850500
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 3.5747843530011115e-05,
+      "loss": 2.1152,
+      "step": 9851000
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 3.574711988236384e-05,
+      "loss": 2.12,
+      "step": 9851500
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 3.574639623471656e-05,
+      "loss": 2.1258,
+      "step": 9852000
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 3.574567258706928e-05,
+      "loss": 2.1044,
+      "step": 9852500
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 3.5744948939422004e-05,
+      "loss": 2.137,
+      "step": 9853000
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 3.574422529177473e-05,
+      "loss": 2.1273,
+      "step": 9853500
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 3.5743501644127455e-05,
+      "loss": 2.1254,
+      "step": 9854000
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 3.574277944377548e-05,
+      "loss": 2.1192,
+      "step": 9854500
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 3.57420557961282e-05,
+      "loss": 2.1,
+      "step": 9855000
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 3.574133214848092e-05,
+      "loss": 2.1237,
+      "step": 9855500
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 3.5740608500833644e-05,
+      "loss": 2.1088,
+      "step": 9856000
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 3.573988919507225e-05,
+      "loss": 2.1301,
+      "step": 9856500
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 3.5739165547424976e-05,
+      "loss": 2.1057,
+      "step": 9857000
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 3.57384418997777e-05,
+      "loss": 2.1023,
+      "step": 9857500
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 3.573771825213042e-05,
+      "loss": 2.1206,
+      "step": 9858000
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 3.5736996051778436e-05,
+      "loss": 2.1206,
+      "step": 9858500
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 3.573627385142646e-05,
+      "loss": 2.1222,
+      "step": 9859000
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 3.573555020377918e-05,
+      "loss": 2.1742,
+      "step": 9859500
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 3.57348265561319e-05,
+      "loss": 2.127,
+      "step": 9860000
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 3.5734102908484625e-05,
+      "loss": 2.1315,
+      "step": 9860500
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 3.573337926083735e-05,
+      "loss": 2.1113,
+      "step": 9861000
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 3.5732655613190076e-05,
+      "loss": 2.1281,
+      "step": 9861500
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 3.57319319655428e-05,
+      "loss": 2.1248,
+      "step": 9862000
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 3.573120831789552e-05,
+      "loss": 2.1267,
+      "step": 9862500
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 3.573048467024824e-05,
+      "loss": 2.1038,
+      "step": 9863000
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 3.572976246989626e-05,
+      "loss": 2.1268,
+      "step": 9863500
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 3.572903882224898e-05,
+      "loss": 2.1396,
+      "step": 9864000
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 3.572831517460171e-05,
+      "loss": 2.112,
+      "step": 9864500
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 3.572759152695443e-05,
+      "loss": 2.1244,
+      "step": 9865000
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 3.5726867879307154e-05,
+      "loss": 2.1293,
+      "step": 9865500
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 3.572614567895517e-05,
+      "loss": 2.1077,
+      "step": 9866000
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 3.572542203130789e-05,
+      "loss": 2.099,
+      "step": 9866500
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 3.572469983095591e-05,
+      "loss": 2.1153,
+      "step": 9867000
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 3.5723976183308636e-05,
+      "loss": 2.1222,
+      "step": 9867500
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 3.572325253566136e-05,
+      "loss": 2.1436,
+      "step": 9868000
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 3.572252888801408e-05,
+      "loss": 2.118,
+      "step": 9868500
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 3.572180524036681e-05,
+      "loss": 2.1236,
+      "step": 9869000
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 3.5721083040014825e-05,
+      "loss": 2.1346,
+      "step": 9869500
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 3.572035939236755e-05,
+      "loss": 2.1186,
+      "step": 9870000
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 3.571963574472027e-05,
+      "loss": 2.1223,
+      "step": 9870500
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 3.571891209707299e-05,
+      "loss": 2.109,
+      "step": 9871000
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 3.5718188449425714e-05,
+      "loss": 2.133,
+      "step": 9871500
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 3.5717464801778436e-05,
+      "loss": 2.1274,
+      "step": 9872000
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 3.571674115413116e-05,
+      "loss": 2.1256,
+      "step": 9872500
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 3.571601750648389e-05,
+      "loss": 2.1406,
+      "step": 9873000
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 3.571529385883661e-05,
+      "loss": 2.1038,
+      "step": 9873500
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 3.5714571658484625e-05,
+      "loss": 2.1238,
+      "step": 9874000
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 3.571384801083735e-05,
+      "loss": 2.111,
+      "step": 9874500
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 3.571312436319007e-05,
+      "loss": 2.1085,
+      "step": 9875000
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 3.5712402162838085e-05,
+      "loss": 2.1222,
+      "step": 9875500
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 3.571167996248611e-05,
+      "loss": 2.1288,
+      "step": 9876000
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 3.571095631483883e-05,
+      "loss": 2.1315,
+      "step": 9876500
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 3.571023266719156e-05,
+      "loss": 2.1071,
+      "step": 9877000
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 3.570950901954428e-05,
+      "loss": 2.1149,
+      "step": 9877500
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 3.5708785371897003e-05,
+      "loss": 2.1025,
+      "step": 9878000
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 3.5708061724249726e-05,
+      "loss": 2.1029,
+      "step": 9878500
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 3.570733807660245e-05,
+      "loss": 2.1142,
+      "step": 9879000
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 3.570661442895517e-05,
+      "loss": 2.1267,
+      "step": 9879500
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 3.570589078130789e-05,
+      "loss": 2.1202,
+      "step": 9880000
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 3.5705167133660615e-05,
+      "loss": 2.1123,
+      "step": 9880500
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 3.570444348601334e-05,
+      "loss": 2.1053,
+      "step": 9881000
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 3.570371983836606e-05,
+      "loss": 2.1314,
+      "step": 9881500
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 3.570299619071879e-05,
+      "loss": 2.1404,
+      "step": 9882000
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 3.570227254307151e-05,
+      "loss": 2.1321,
+      "step": 9882500
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 3.570154889542423e-05,
+      "loss": 2.1232,
+      "step": 9883000
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 3.5700825247776955e-05,
+      "loss": 2.1228,
+      "step": 9883500
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 3.5700101600129684e-05,
+      "loss": 2.1087,
+      "step": 9884000
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 3.5699377952482406e-05,
+      "loss": 2.0966,
+      "step": 9884500
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 3.569865430483513e-05,
+      "loss": 2.1453,
+      "step": 9885000
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 3.569793065718785e-05,
+      "loss": 2.1107,
+      "step": 9885500
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 3.5697208456835866e-05,
+      "loss": 2.1266,
+      "step": 9886000
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 3.569648480918859e-05,
+      "loss": 2.1393,
+      "step": 9886500
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 3.569576116154131e-05,
+      "loss": 2.1222,
+      "step": 9887000
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 3.569503896118933e-05,
+      "loss": 2.1214,
+      "step": 9887500
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 3.5694315313542055e-05,
+      "loss": 2.116,
+      "step": 9888000
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 3.569359166589478e-05,
+      "loss": 2.1235,
+      "step": 9888500
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 3.56928680182475e-05,
+      "loss": 2.1251,
+      "step": 9889000
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 3.569214437060022e-05,
+      "loss": 2.1259,
+      "step": 9889500
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 3.5691420722952944e-05,
+      "loss": 2.1277,
+      "step": 9890000
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 3.5690697075305666e-05,
+      "loss": 2.1083,
+      "step": 9890500
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 3.568997487495369e-05,
+      "loss": 2.089,
+      "step": 9891000
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 3.568925122730641e-05,
+      "loss": 2.12,
+      "step": 9891500
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 3.568852757965914e-05,
+      "loss": 2.1131,
+      "step": 9892000
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 3.568780393201186e-05,
+      "loss": 2.1152,
+      "step": 9892500
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 3.5687080284364584e-05,
+      "loss": 2.096,
+      "step": 9893000
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 3.5686356636717307e-05,
+      "loss": 2.1328,
+      "step": 9893500
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 3.568563298907003e-05,
+      "loss": 2.0819,
+      "step": 9894000
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 3.568490934142275e-05,
+      "loss": 2.119,
+      "step": 9894500
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 3.5684187141070767e-05,
+      "loss": 2.1089,
+      "step": 9895000
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 3.568346349342349e-05,
+      "loss": 2.1064,
+      "step": 9895500
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 3.568273984577621e-05,
+      "loss": 2.1283,
+      "step": 9896000
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 3.568201619812894e-05,
+      "loss": 2.1276,
+      "step": 9896500
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 3.568129255048166e-05,
+      "loss": 2.1333,
+      "step": 9897000
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 3.5680568902834384e-05,
+      "loss": 2.1138,
+      "step": 9897500
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 3.567984525518711e-05,
+      "loss": 2.1138,
+      "step": 9898000
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 3.5679121607539836e-05,
+      "loss": 2.1133,
+      "step": 9898500
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 3.567839795989256e-05,
+      "loss": 2.1412,
+      "step": 9899000
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 3.5677675759540574e-05,
+      "loss": 2.1368,
+      "step": 9899500
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 3.567695355918859e-05,
+      "loss": 2.1011,
+      "step": 9900000
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 3.567622991154131e-05,
+      "loss": 2.1404,
+      "step": 9900500
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 3.567550626389404e-05,
+      "loss": 2.1189,
+      "step": 9901000
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 3.567478261624676e-05,
+      "loss": 2.1181,
+      "step": 9901500
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 3.5674058968599485e-05,
+      "loss": 2.1057,
+      "step": 9902000
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 3.567333532095221e-05,
+      "loss": 2.1118,
+      "step": 9902500
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 3.567261167330493e-05,
+      "loss": 2.1242,
+      "step": 9903000
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 3.567188802565765e-05,
+      "loss": 2.0965,
+      "step": 9903500
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 3.5671164378010374e-05,
+      "loss": 2.116,
+      "step": 9904000
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 3.567044217765839e-05,
+      "loss": 2.1028,
+      "step": 9904500
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 3.566971997730641e-05,
+      "loss": 2.1047,
+      "step": 9905000
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 3.5668996329659134e-05,
+      "loss": 2.1127,
+      "step": 9905500
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 3.566827412930715e-05,
+      "loss": 2.1076,
+      "step": 9906000
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 3.566755048165988e-05,
+      "loss": 2.1361,
+      "step": 9906500
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 3.56668268340126e-05,
+      "loss": 2.1333,
+      "step": 9907000
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 3.5666104633660616e-05,
+      "loss": 2.1383,
+      "step": 9907500
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 3.566538098601334e-05,
+      "loss": 2.1064,
+      "step": 9908000
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 3.566465733836607e-05,
+      "loss": 2.1328,
+      "step": 9908500
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 3.566393369071879e-05,
+      "loss": 2.0844,
+      "step": 9909000
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 3.566321004307151e-05,
+      "loss": 2.1031,
+      "step": 9909500
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 3.566248784271953e-05,
+      "loss": 2.1222,
+      "step": 9910000
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 3.566176419507225e-05,
+      "loss": 2.1328,
+      "step": 9910500
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 3.5661041994720265e-05,
+      "loss": 2.1261,
+      "step": 9911000
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 3.566031834707299e-05,
+      "loss": 2.1293,
+      "step": 9911500
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 3.5659594699425716e-05,
+      "loss": 2.1336,
+      "step": 9912000
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 3.565887105177844e-05,
+      "loss": 2.1166,
+      "step": 9912500
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 3.565814740413116e-05,
+      "loss": 2.1113,
+      "step": 9913000
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 3.565742375648388e-05,
+      "loss": 2.106,
+      "step": 9913500
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 3.565670010883661e-05,
+      "loss": 2.1276,
+      "step": 9914000
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 3.5655976461189334e-05,
+      "loss": 2.1252,
+      "step": 9914500
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 3.565525281354206e-05,
+      "loss": 2.1281,
+      "step": 9915000
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 3.565452916589478e-05,
+      "loss": 2.1347,
+      "step": 9915500
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 3.56538055182475e-05,
+      "loss": 2.1117,
+      "step": 9916000
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 3.565308187060022e-05,
+      "loss": 2.153,
+      "step": 9916500
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 3.5652358222952946e-05,
+      "loss": 2.1481,
+      "step": 9917000
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 3.565163457530567e-05,
+      "loss": 2.1251,
+      "step": 9917500
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 3.565091237495369e-05,
+      "loss": 2.1152,
+      "step": 9918000
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 3.565018872730641e-05,
+      "loss": 2.1146,
+      "step": 9918500
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 3.5649465079659135e-05,
+      "loss": 2.1237,
+      "step": 9919000
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 3.564874143201186e-05,
+      "loss": 2.1236,
+      "step": 9919500
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 3.564801778436458e-05,
+      "loss": 2.1139,
+      "step": 9920000
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 3.5647295584012595e-05,
+      "loss": 2.123,
+      "step": 9920500
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 3.564657338366062e-05,
+      "loss": 2.1333,
+      "step": 9921000
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 3.564584973601334e-05,
+      "loss": 2.1068,
+      "step": 9921500
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 3.564512608836607e-05,
+      "loss": 2.1423,
+      "step": 9922000
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 3.564440244071879e-05,
+      "loss": 2.1118,
+      "step": 9922500
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 3.564367879307151e-05,
+      "loss": 2.1491,
+      "step": 9923000
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 3.5642955145424235e-05,
+      "loss": 2.1252,
+      "step": 9923500
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 3.564223149777696e-05,
+      "loss": 2.1035,
+      "step": 9924000
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 3.564150785012968e-05,
+      "loss": 2.1048,
+      "step": 9924500
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 3.56407842024824e-05,
+      "loss": 2.1005,
+      "step": 9925000
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 3.5640060554835124e-05,
+      "loss": 2.1277,
+      "step": 9925500
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 3.5639336907187846e-05,
+      "loss": 2.1255,
+      "step": 9926000
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 3.563861325954057e-05,
+      "loss": 2.125,
+      "step": 9926500
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 3.563788961189329e-05,
+      "loss": 2.1176,
+      "step": 9927000
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 3.563716596424602e-05,
+      "loss": 2.1107,
+      "step": 9927500
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 3.563644231659874e-05,
+      "loss": 2.1106,
+      "step": 9928000
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 3.5635720116246764e-05,
+      "loss": 2.1156,
+      "step": 9928500
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 3.5634996468599486e-05,
+      "loss": 2.1397,
+      "step": 9929000
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 3.563427282095221e-05,
+      "loss": 2.1293,
+      "step": 9929500
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 3.563354917330493e-05,
+      "loss": 2.1244,
+      "step": 9930000
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 3.5632826972952946e-05,
+      "loss": 2.1279,
+      "step": 9930500
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 3.563210477260097e-05,
+      "loss": 2.109,
+      "step": 9931000
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 3.563138112495369e-05,
+      "loss": 2.1182,
+      "step": 9931500
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 3.563065747730641e-05,
+      "loss": 2.1414,
+      "step": 9932000
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 3.5629933829659135e-05,
+      "loss": 2.1029,
+      "step": 9932500
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 3.562921018201186e-05,
+      "loss": 2.1087,
+      "step": 9933000
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 3.562848653436458e-05,
+      "loss": 2.1121,
+      "step": 9933500
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 3.56277628867173e-05,
+      "loss": 2.1282,
+      "step": 9934000
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 3.5627039239070024e-05,
+      "loss": 2.1274,
+      "step": 9934500
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 3.5626315591422747e-05,
+      "loss": 2.1175,
+      "step": 9935000
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 3.562559194377547e-05,
+      "loss": 2.1191,
+      "step": 9935500
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 3.56248682961282e-05,
+      "loss": 2.0979,
+      "step": 9936000
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 3.562414609577622e-05,
+      "loss": 2.1195,
+      "step": 9936500
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 3.562342244812894e-05,
+      "loss": 2.1039,
+      "step": 9937000
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 3.5622698800481665e-05,
+      "loss": 2.1042,
+      "step": 9937500
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 3.562197515283439e-05,
+      "loss": 2.1364,
+      "step": 9938000
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 3.562125150518711e-05,
+      "loss": 2.1089,
+      "step": 9938500
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 3.562052785753983e-05,
+      "loss": 2.1136,
+      "step": 9939000
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 3.5619804209892554e-05,
+      "loss": 2.1247,
+      "step": 9939500
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 3.5619080562245276e-05,
+      "loss": 2.1139,
+      "step": 9940000
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 3.5618356914598e-05,
+      "loss": 2.1322,
+      "step": 9940500
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 3.561763326695072e-05,
+      "loss": 2.116,
+      "step": 9941000
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 3.561690961930344e-05,
+      "loss": 2.1115,
+      "step": 9941500
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 3.561618597165617e-05,
+      "loss": 2.1234,
+      "step": 9942000
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 3.5615462324008894e-05,
+      "loss": 2.1311,
+      "step": 9942500
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 3.561474012365691e-05,
+      "loss": 2.1029,
+      "step": 9943000
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 3.561401792330493e-05,
+      "loss": 2.1162,
+      "step": 9943500
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 3.5613294275657654e-05,
+      "loss": 2.1395,
+      "step": 9944000
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 3.5612570628010376e-05,
+      "loss": 2.1428,
+      "step": 9944500
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 3.56118469803631e-05,
+      "loss": 2.1258,
+      "step": 9945000
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 3.561112333271582e-05,
+      "loss": 2.1098,
+      "step": 9945500
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 3.561039968506854e-05,
+      "loss": 2.1309,
+      "step": 9946000
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 3.560967603742127e-05,
+      "loss": 2.1278,
+      "step": 9946500
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 3.560895383706929e-05,
+      "loss": 2.116,
+      "step": 9947000
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 3.560823018942201e-05,
+      "loss": 2.1039,
+      "step": 9947500
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 3.5607507989070025e-05,
+      "loss": 2.1296,
+      "step": 9948000
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 3.560678434142275e-05,
+      "loss": 2.1257,
+      "step": 9948500
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 3.560606069377547e-05,
+      "loss": 2.1199,
+      "step": 9949000
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 3.56053370461282e-05,
+      "loss": 2.1199,
+      "step": 9949500
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 3.560461339848092e-05,
+      "loss": 2.1225,
+      "step": 9950000
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 3.560388975083364e-05,
+      "loss": 2.1296,
+      "step": 9950500
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 3.5603167550481665e-05,
+      "loss": 2.1256,
+      "step": 9951000
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 3.560244535012968e-05,
+      "loss": 2.1484,
+      "step": 9951500
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 3.56017217024824e-05,
+      "loss": 2.1169,
+      "step": 9952000
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 3.5600998054835125e-05,
+      "loss": 2.1363,
+      "step": 9952500
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 3.560027440718785e-05,
+      "loss": 2.1287,
+      "step": 9953000
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 3.559955075954057e-05,
+      "loss": 2.1115,
+      "step": 9953500
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 3.55988271118933e-05,
+      "loss": 2.1599,
+      "step": 9954000
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 3.559810346424602e-05,
+      "loss": 2.1142,
+      "step": 9954500
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 3.5597379816598743e-05,
+      "loss": 2.1238,
+      "step": 9955000
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 3.5596656168951466e-05,
+      "loss": 2.1118,
+      "step": 9955500
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 3.559593252130419e-05,
+      "loss": 2.1253,
+      "step": 9956000
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 3.5595210320952203e-05,
+      "loss": 2.1195,
+      "step": 9956500
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 3.5594486673304926e-05,
+      "loss": 2.1301,
+      "step": 9957000
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 3.559376302565765e-05,
+      "loss": 2.1329,
+      "step": 9957500
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 3.559303937801037e-05,
+      "loss": 2.1453,
+      "step": 9958000
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 3.55923157303631e-05,
+      "loss": 2.1479,
+      "step": 9958500
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 3.559159208271582e-05,
+      "loss": 2.1278,
+      "step": 9959000
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 3.559086843506855e-05,
+      "loss": 2.119,
+      "step": 9959500
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 3.559014478742127e-05,
+      "loss": 2.1224,
+      "step": 9960000
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 3.5589421139773995e-05,
+      "loss": 2.137,
+      "step": 9960500
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 3.558869893942201e-05,
+      "loss": 2.1381,
+      "step": 9961000
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 3.558797529177473e-05,
+      "loss": 2.1382,
+      "step": 9961500
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 3.5587251644127455e-05,
+      "loss": 2.1392,
+      "step": 9962000
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 3.558652799648018e-05,
+      "loss": 2.1117,
+      "step": 9962500
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 3.55858043488329e-05,
+      "loss": 2.1267,
+      "step": 9963000
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 3.558508070118562e-05,
+      "loss": 2.0989,
+      "step": 9963500
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 3.5584358500833644e-05,
+      "loss": 2.1455,
+      "step": 9964000
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 3.5583634853186366e-05,
+      "loss": 2.1252,
+      "step": 9964500
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 3.558291265283438e-05,
+      "loss": 2.1478,
+      "step": 9965000
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 3.5582189005187104e-05,
+      "loss": 2.1452,
+      "step": 9965500
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 3.558146535753983e-05,
+      "loss": 2.1156,
+      "step": 9966000
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 3.5580741709892555e-05,
+      "loss": 2.1415,
+      "step": 9966500
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 3.558001806224528e-05,
+      "loss": 2.1377,
+      "step": 9967000
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 3.5579294414598e-05,
+      "loss": 2.1316,
+      "step": 9967500
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 3.557857076695072e-05,
+      "loss": 2.1107,
+      "step": 9968000
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 3.557784711930345e-05,
+      "loss": 2.1122,
+      "step": 9968500
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 3.557712347165617e-05,
+      "loss": 2.1062,
+      "step": 9969000
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 3.5576399824008895e-05,
+      "loss": 2.1113,
+      "step": 9969500
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 3.557567762365691e-05,
+      "loss": 2.1405,
+      "step": 9970000
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 3.557495397600963e-05,
+      "loss": 2.1206,
+      "step": 9970500
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 3.557423177565765e-05,
+      "loss": 2.1295,
+      "step": 9971000
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 3.557350812801038e-05,
+      "loss": 2.1024,
+      "step": 9971500
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 3.55727844803631e-05,
+      "loss": 2.1213,
+      "step": 9972000
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 3.557206083271582e-05,
+      "loss": 2.1113,
+      "step": 9972500
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 3.5571337185068544e-05,
+      "loss": 2.1388,
+      "step": 9973000
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 3.5570613537421273e-05,
+      "loss": 2.1143,
+      "step": 9973500
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 3.5569889889773996e-05,
+      "loss": 2.1291,
+      "step": 9974000
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 3.556916624212672e-05,
+      "loss": 2.119,
+      "step": 9974500
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 3.556844259447944e-05,
+      "loss": 2.1275,
+      "step": 9975000
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 3.556771894683216e-05,
+      "loss": 2.1332,
+      "step": 9975500
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 3.5566995299184885e-05,
+      "loss": 2.1147,
+      "step": 9976000
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 3.556627165153761e-05,
+      "loss": 2.149,
+      "step": 9976500
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 3.556554945118563e-05,
+      "loss": 2.1211,
+      "step": 9977000
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 3.556482580353835e-05,
+      "loss": 2.1241,
+      "step": 9977500
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 3.5564102155891074e-05,
+      "loss": 2.1157,
+      "step": 9978000
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 3.5563378508243796e-05,
+      "loss": 2.1271,
+      "step": 9978500
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 3.556265630789181e-05,
+      "loss": 2.144,
+      "step": 9979000
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 3.556193555483513e-05,
+      "loss": 2.1351,
+      "step": 9979500
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 3.556121190718785e-05,
+      "loss": 2.1219,
+      "step": 9980000
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 3.5560489706835865e-05,
+      "loss": 2.1085,
+      "step": 9980500
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 3.5559766059188594e-05,
+      "loss": 2.1248,
+      "step": 9981000
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 3.5559042411541316e-05,
+      "loss": 2.1212,
+      "step": 9981500
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 3.555831876389404e-05,
+      "loss": 2.1092,
+      "step": 9982000
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 3.555759511624676e-05,
+      "loss": 2.1262,
+      "step": 9982500
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 3.555687146859948e-05,
+      "loss": 2.1341,
+      "step": 9983000
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 3.5556147820952205e-05,
+      "loss": 2.1081,
+      "step": 9983500
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 3.555542417330493e-05,
+      "loss": 2.1099,
+      "step": 9984000
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 3.555470052565765e-05,
+      "loss": 2.1126,
+      "step": 9984500
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 3.555397832530567e-05,
+      "loss": 2.1272,
+      "step": 9985000
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 3.5553254677658394e-05,
+      "loss": 2.1062,
+      "step": 9985500
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 3.5552531030011116e-05,
+      "loss": 2.118,
+      "step": 9986000
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 3.555180738236384e-05,
+      "loss": 2.1222,
+      "step": 9986500
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 3.555108373471656e-05,
+      "loss": 2.1528,
+      "step": 9987000
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 3.555036008706928e-05,
+      "loss": 2.1166,
+      "step": 9987500
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 3.5549636439422005e-05,
+      "loss": 2.0957,
+      "step": 9988000
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 3.5548912791774734e-05,
+      "loss": 2.1213,
+      "step": 9988500
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 3.554819059142275e-05,
+      "loss": 2.1021,
+      "step": 9989000
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 3.554746694377548e-05,
+      "loss": 2.1138,
+      "step": 9989500
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 3.55467432961282e-05,
+      "loss": 2.1155,
+      "step": 9990000
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 3.554601964848092e-05,
+      "loss": 2.1273,
+      "step": 9990500
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 3.5545296000833645e-05,
+      "loss": 2.1018,
+      "step": 9991000
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 3.554457235318637e-05,
+      "loss": 2.1229,
+      "step": 9991500
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 3.554384870553909e-05,
+      "loss": 2.1173,
+      "step": 9992000
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 3.554312505789181e-05,
+      "loss": 2.1118,
+      "step": 9992500
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 3.5542401410244534e-05,
+      "loss": 2.1104,
+      "step": 9993000
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 3.554167920989255e-05,
+      "loss": 2.1203,
+      "step": 9993500
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 3.554095700954057e-05,
+      "loss": 2.1198,
+      "step": 9994000
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 3.5540233361893295e-05,
+      "loss": 2.12,
+      "step": 9994500
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 3.5539512608836603e-05,
+      "loss": 2.1148,
+      "step": 9995000
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 3.5538788961189326e-05,
+      "loss": 2.1509,
+      "step": 9995500
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 3.5538065313542055e-05,
+      "loss": 2.1294,
+      "step": 9996000
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 3.553734166589478e-05,
+      "loss": 2.1334,
+      "step": 9996500
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 3.5536618018247506e-05,
+      "loss": 2.1393,
+      "step": 9997000
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 3.553589437060023e-05,
+      "loss": 2.1287,
+      "step": 9997500
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 3.553517072295295e-05,
+      "loss": 2.1171,
+      "step": 9998000
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 3.553444707530567e-05,
+      "loss": 2.1068,
+      "step": 9998500
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 3.5533723427658395e-05,
+      "loss": 2.1243,
+      "step": 9999000
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 3.553299978001112e-05,
+      "loss": 2.1243,
+      "step": 9999500
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 3.553227613236384e-05,
+      "loss": 2.1165,
+      "step": 10000000
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 3.553155248471656e-05,
+      "loss": 2.1262,
+      "step": 10000500
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 3.5530828837069284e-05,
+      "loss": 2.1372,
+      "step": 10001000
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 3.5530105189422006e-05,
+      "loss": 2.113,
+      "step": 10001500
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 3.552938154177473e-05,
+      "loss": 2.1161,
+      "step": 10002000
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 3.552865789412746e-05,
+      "loss": 2.117,
+      "step": 10002500
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 3.552793424648018e-05,
+      "loss": 2.1401,
+      "step": 10003000
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 3.55272120461282e-05,
+      "loss": 2.1165,
+      "step": 10003500
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 3.5526488398480924e-05,
+      "loss": 2.1269,
+      "step": 10004000
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 3.5525764750833646e-05,
+      "loss": 2.1217,
+      "step": 10004500
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 3.552504110318637e-05,
+      "loss": 2.1015,
+      "step": 10005000
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 3.5524318902834384e-05,
+      "loss": 2.1069,
+      "step": 10005500
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 3.5523595255187106e-05,
+      "loss": 2.1272,
+      "step": 10006000
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 3.552287160753983e-05,
+      "loss": 2.1312,
+      "step": 10006500
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 3.552214795989256e-05,
+      "loss": 2.123,
+      "step": 10007000
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 3.552142431224528e-05,
+      "loss": 2.1673,
+      "step": 10007500
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 3.5520702111893295e-05,
+      "loss": 2.1334,
+      "step": 10008000
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 3.551997846424602e-05,
+      "loss": 2.1039,
+      "step": 10008500
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 3.551925481659874e-05,
+      "loss": 2.134,
+      "step": 10009000
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 3.551853116895146e-05,
+      "loss": 2.1271,
+      "step": 10009500
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 3.5517807521304184e-05,
+      "loss": 2.1175,
+      "step": 10010000
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 3.5517085320952207e-05,
+      "loss": 2.123,
+      "step": 10010500
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 3.551636167330493e-05,
+      "loss": 2.1119,
+      "step": 10011000
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 3.551563802565766e-05,
+      "loss": 2.1212,
+      "step": 10011500
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 3.551491437801038e-05,
+      "loss": 2.1231,
+      "step": 10012000
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 3.55141907303631e-05,
+      "loss": 2.1192,
+      "step": 10012500
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 3.5513467082715825e-05,
+      "loss": 2.1198,
+      "step": 10013000
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 3.551274343506855e-05,
+      "loss": 2.104,
+      "step": 10013500
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 3.551201978742127e-05,
+      "loss": 2.1169,
+      "step": 10014000
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 3.551129613977399e-05,
+      "loss": 2.1155,
+      "step": 10014500
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 3.551057393942201e-05,
+      "loss": 2.1271,
+      "step": 10015000
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 3.550985173907003e-05,
+      "loss": 2.1298,
+      "step": 10015500
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 3.5509129538718045e-05,
+      "loss": 2.1075,
+      "step": 10016000
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 3.550840589107077e-05,
+      "loss": 2.1213,
+      "step": 10016500
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 3.550768369071878e-05,
+      "loss": 2.1133,
+      "step": 10017000
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 3.5506960043071505e-05,
+      "loss": 2.1327,
+      "step": 10017500
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 3.5506236395424234e-05,
+      "loss": 2.1346,
+      "step": 10018000
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 3.5505512747776956e-05,
+      "loss": 2.1271,
+      "step": 10018500
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy": 0.6662792394368734,
+      "eval_accuracy_mlm": 0.6307335576560936,
+      "eval_accuracy_nsp": 0.856817129788765,
+      "eval_loss": 2.1900088787078857,
+      "eval_runtime": 331.3704,
+      "eval_samples_per_second": 1316.913,
+      "eval_steps_per_second": 54.872,
+      "step": 10018688
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 3.5504789100129685e-05,
+      "loss": 2.0982,
+      "step": 10019000
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 3.550406545248241e-05,
+      "loss": 2.1009,
+      "step": 10019500
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 3.550334180483513e-05,
+      "loss": 2.1048,
+      "step": 10020000
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 3.550261815718785e-05,
+      "loss": 2.1176,
+      "step": 10020500
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 3.5501894509540574e-05,
+      "loss": 2.0853,
+      "step": 10021000
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 3.5501170861893296e-05,
+      "loss": 2.0882,
+      "step": 10021500
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 3.550044866154131e-05,
+      "loss": 2.0952,
+      "step": 10022000
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 3.5499726461189334e-05,
+      "loss": 2.1082,
+      "step": 10022500
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 3.5499002813542056e-05,
+      "loss": 2.0918,
+      "step": 10023000
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 3.549827916589478e-05,
+      "loss": 2.0879,
+      "step": 10023500
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 3.54975555182475e-05,
+      "loss": 2.1145,
+      "step": 10024000
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 3.5496833317895516e-05,
+      "loss": 2.1052,
+      "step": 10024500
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 3.549610967024824e-05,
+      "loss": 2.1257,
+      "step": 10025000
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 3.549538602260096e-05,
+      "loss": 2.092,
+      "step": 10025500
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 3.549466237495368e-05,
+      "loss": 2.1166,
+      "step": 10026000
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 3.549393872730641e-05,
+      "loss": 2.0745,
+      "step": 10026500
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 3.5493216526954434e-05,
+      "loss": 2.1014,
+      "step": 10027000
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 3.5492492879307157e-05,
+      "loss": 2.0886,
+      "step": 10027500
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 3.549176923165988e-05,
+      "loss": 2.0981,
+      "step": 10028000
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 3.54910455840126e-05,
+      "loss": 2.082,
+      "step": 10028500
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 3.549032193636532e-05,
+      "loss": 2.1012,
+      "step": 10029000
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 3.5489598288718045e-05,
+      "loss": 2.0996,
+      "step": 10029500
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 3.548887464107077e-05,
+      "loss": 2.0904,
+      "step": 10030000
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 3.548815099342349e-05,
+      "loss": 2.101,
+      "step": 10030500
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 3.5487428793071506e-05,
+      "loss": 2.1211,
+      "step": 10031000
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 3.5486705145424235e-05,
+      "loss": 2.1169,
+      "step": 10031500
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 3.548598439236754e-05,
+      "loss": 2.0889,
+      "step": 10032000
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 3.5485260744720266e-05,
+      "loss": 2.1029,
+      "step": 10032500
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 3.548453709707299e-05,
+      "loss": 2.098,
+      "step": 10033000
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 3.548381344942571e-05,
+      "loss": 2.0799,
+      "step": 10033500
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 3.548308980177844e-05,
+      "loss": 2.0904,
+      "step": 10034000
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 3.548236760142646e-05,
+      "loss": 2.0925,
+      "step": 10034500
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 3.5481643953779184e-05,
+      "loss": 2.1226,
+      "step": 10035000
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 3.5480920306131906e-05,
+      "loss": 2.1151,
+      "step": 10035500
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 3.548019665848463e-05,
+      "loss": 2.0998,
+      "step": 10036000
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 3.547947301083735e-05,
+      "loss": 2.12,
+      "step": 10036500
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 3.547874936319007e-05,
+      "loss": 2.1126,
+      "step": 10037000
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 3.5478025715542795e-05,
+      "loss": 2.084,
+      "step": 10037500
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 3.547730206789552e-05,
+      "loss": 2.1149,
+      "step": 10038000
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 3.547657842024824e-05,
+      "loss": 2.104,
+      "step": 10038500
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 3.547585477260096e-05,
+      "loss": 2.104,
+      "step": 10039000
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 3.5475131124953684e-05,
+      "loss": 2.0888,
+      "step": 10039500
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 3.547440747730641e-05,
+      "loss": 2.1084,
+      "step": 10040000
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 3.5473683829659135e-05,
+      "loss": 2.0827,
+      "step": 10040500
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 3.547296162930715e-05,
+      "loss": 2.1294,
+      "step": 10041000
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 3.547223798165988e-05,
+      "loss": 2.1124,
+      "step": 10041500
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 3.54715143340126e-05,
+      "loss": 2.0882,
+      "step": 10042000
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 3.5470790686365324e-05,
+      "loss": 2.0722,
+      "step": 10042500
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 3.547006848601334e-05,
+      "loss": 2.1021,
+      "step": 10043000
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 3.546934483836606e-05,
+      "loss": 2.0901,
+      "step": 10043500
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 3.5468621190718784e-05,
+      "loss": 2.1086,
+      "step": 10044000
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 3.546789754307151e-05,
+      "loss": 2.1175,
+      "step": 10044500
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 3.546717534271953e-05,
+      "loss": 2.108,
+      "step": 10045000
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 3.546645169507225e-05,
+      "loss": 2.0909,
+      "step": 10045500
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 3.5465729494720266e-05,
+      "loss": 2.0755,
+      "step": 10046000
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 3.546500584707299e-05,
+      "loss": 2.0929,
+      "step": 10046500
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 3.546428219942571e-05,
+      "loss": 2.1164,
+      "step": 10047000
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 3.546355855177843e-05,
+      "loss": 2.1038,
+      "step": 10047500
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 3.546283490413116e-05,
+      "loss": 2.0862,
+      "step": 10048000
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 3.5462111256483884e-05,
+      "loss": 2.1005,
+      "step": 10048500
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 3.546138760883661e-05,
+      "loss": 2.0959,
+      "step": 10049000
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 3.5460663961189336e-05,
+      "loss": 2.1033,
+      "step": 10049500
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 3.545994031354206e-05,
+      "loss": 2.1339,
+      "step": 10050000
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 3.545921666589478e-05,
+      "loss": 2.0974,
+      "step": 10050500
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 3.54584930182475e-05,
+      "loss": 2.1015,
+      "step": 10051000
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 3.5457769370600225e-05,
+      "loss": 2.1039,
+      "step": 10051500
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 3.545704572295295e-05,
+      "loss": 2.0994,
+      "step": 10052000
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 3.545632352260096e-05,
+      "loss": 2.0851,
+      "step": 10052500
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 3.545560276954428e-05,
+      "loss": 2.0928,
+      "step": 10053000
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 3.5454879121897e-05,
+      "loss": 2.094,
+      "step": 10053500
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 3.5454156921545016e-05,
+      "loss": 2.1112,
+      "step": 10054000
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 3.545343327389774e-05,
+      "loss": 2.0995,
+      "step": 10054500
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 3.545270962625046e-05,
+      "loss": 2.0966,
+      "step": 10055000
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 3.545198597860319e-05,
+      "loss": 2.0848,
+      "step": 10055500
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 3.545126233095591e-05,
+      "loss": 2.0994,
+      "step": 10056000
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 3.545053868330864e-05,
+      "loss": 2.1201,
+      "step": 10056500
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 3.544981503566136e-05,
+      "loss": 2.1009,
+      "step": 10057000
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 3.544909283530938e-05,
+      "loss": 2.0938,
+      "step": 10057500
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 3.54483691876621e-05,
+      "loss": 2.1009,
+      "step": 10058000
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 3.544764554001482e-05,
+      "loss": 2.1127,
+      "step": 10058500
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 3.5446921892367545e-05,
+      "loss": 2.1195,
+      "step": 10059000
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 3.544619824472027e-05,
+      "loss": 2.1151,
+      "step": 10059500
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 3.544547459707299e-05,
+      "loss": 2.1096,
+      "step": 10060000
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 3.544475094942571e-05,
+      "loss": 2.0873,
+      "step": 10060500
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 3.5444028749073734e-05,
+      "loss": 2.1079,
+      "step": 10061000
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 3.5443305101426456e-05,
+      "loss": 2.0924,
+      "step": 10061500
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 3.544258145377918e-05,
+      "loss": 2.1137,
+      "step": 10062000
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 3.54418578061319e-05,
+      "loss": 2.0996,
+      "step": 10062500
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 3.544113415848462e-05,
+      "loss": 2.1074,
+      "step": 10063000
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 3.5440410510837345e-05,
+      "loss": 2.0977,
+      "step": 10063500
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 3.5439686863190074e-05,
+      "loss": 2.097,
+      "step": 10064000
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 3.5438963215542796e-05,
+      "loss": 2.1042,
+      "step": 10064500
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 3.543823956789552e-05,
+      "loss": 2.0965,
+      "step": 10065000
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 3.543751592024824e-05,
+      "loss": 2.1078,
+      "step": 10065500
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 3.543679227260096e-05,
+      "loss": 2.1071,
+      "step": 10066000
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 3.543606862495369e-05,
+      "loss": 2.093,
+      "step": 10066500
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 3.5435344977306414e-05,
+      "loss": 2.1206,
+      "step": 10067000
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 3.5434621329659137e-05,
+      "loss": 2.129,
+      "step": 10067500
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 3.543389912930715e-05,
+      "loss": 2.1102,
+      "step": 10068000
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 3.5433175481659874e-05,
+      "loss": 2.0843,
+      "step": 10068500
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 3.54324518340126e-05,
+      "loss": 2.1039,
+      "step": 10069000
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 3.543172818636532e-05,
+      "loss": 2.1112,
+      "step": 10069500
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 3.543100453871804e-05,
+      "loss": 2.1044,
+      "step": 10070000
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 3.543028089107076e-05,
+      "loss": 2.0929,
+      "step": 10070500
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 3.5429558690718786e-05,
+      "loss": 2.1076,
+      "step": 10071000
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 3.5428835043071515e-05,
+      "loss": 2.1004,
+      "step": 10071500
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 3.542811139542424e-05,
+      "loss": 2.1186,
+      "step": 10072000
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 3.542738919507225e-05,
+      "loss": 2.1283,
+      "step": 10072500
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 3.5426665547424975e-05,
+      "loss": 2.087,
+      "step": 10073000
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 3.54259418997777e-05,
+      "loss": 2.0945,
+      "step": 10073500
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 3.542521825213042e-05,
+      "loss": 2.0937,
+      "step": 10074000
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 3.542449460448314e-05,
+      "loss": 2.1217,
+      "step": 10074500
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 3.5423770956835864e-05,
+      "loss": 2.1126,
+      "step": 10075000
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 3.542304730918859e-05,
+      "loss": 2.1206,
+      "step": 10075500
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 3.542232510883661e-05,
+      "loss": 2.1067,
+      "step": 10076000
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 3.542160146118933e-05,
+      "loss": 2.1164,
+      "step": 10076500
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 3.542087781354205e-05,
+      "loss": 2.0922,
+      "step": 10077000
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 3.5420154165894775e-05,
+      "loss": 2.0974,
+      "step": 10077500
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 3.54194305182475e-05,
+      "loss": 2.0901,
+      "step": 10078000
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 3.5418706870600226e-05,
+      "loss": 2.1171,
+      "step": 10078500
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 3.541798322295295e-05,
+      "loss": 2.1219,
+      "step": 10079000
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 3.541725957530567e-05,
+      "loss": 2.1036,
+      "step": 10079500
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 3.541653592765839e-05,
+      "loss": 2.0928,
+      "step": 10080000
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 3.5415812280011115e-05,
+      "loss": 2.0988,
+      "step": 10080500
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 3.5415088632363844e-05,
+      "loss": 2.108,
+      "step": 10081000
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 3.5414364984716566e-05,
+      "loss": 2.0931,
+      "step": 10081500
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 3.541364133706929e-05,
+      "loss": 2.0936,
+      "step": 10082000
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 3.5412919136717304e-05,
+      "loss": 2.1037,
+      "step": 10082500
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 3.5412195489070026e-05,
+      "loss": 2.1157,
+      "step": 10083000
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 3.541147184142275e-05,
+      "loss": 2.1112,
+      "step": 10083500
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 3.541074819377547e-05,
+      "loss": 2.1344,
+      "step": 10084000
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 3.541002599342349e-05,
+      "loss": 2.1111,
+      "step": 10084500
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 3.540930379307151e-05,
+      "loss": 2.0956,
+      "step": 10085000
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 3.540858014542423e-05,
+      "loss": 2.1105,
+      "step": 10085500
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 3.540785649777695e-05,
+      "loss": 2.0927,
+      "step": 10086000
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 3.540713285012968e-05,
+      "loss": 2.0944,
+      "step": 10086500
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 3.5406409202482404e-05,
+      "loss": 2.1158,
+      "step": 10087000
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 3.540568555483513e-05,
+      "loss": 2.1223,
+      "step": 10087500
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 3.540496190718785e-05,
+      "loss": 2.1121,
+      "step": 10088000
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 3.5404239706835864e-05,
+      "loss": 2.0998,
+      "step": 10088500
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 3.540351750648389e-05,
+      "loss": 2.1016,
+      "step": 10089000
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 3.540279385883661e-05,
+      "loss": 2.1194,
+      "step": 10089500
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 3.540207021118933e-05,
+      "loss": 2.1171,
+      "step": 10090000
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 3.5401346563542053e-05,
+      "loss": 2.1074,
+      "step": 10090500
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 3.5400622915894776e-05,
+      "loss": 2.1131,
+      "step": 10091000
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 3.53998992682475e-05,
+      "loss": 2.1268,
+      "step": 10091500
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 3.539917562060022e-05,
+      "loss": 2.098,
+      "step": 10092000
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 3.539845197295294e-05,
+      "loss": 2.1144,
+      "step": 10092500
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 3.5397729772600965e-05,
+      "loss": 2.1159,
+      "step": 10093000
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 3.539700757224898e-05,
+      "loss": 2.1003,
+      "step": 10093500
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 3.539628392460171e-05,
+      "loss": 2.1299,
+      "step": 10094000
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 3.539556027695443e-05,
+      "loss": 2.0894,
+      "step": 10094500
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 3.5394836629307154e-05,
+      "loss": 2.1106,
+      "step": 10095000
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 3.5394112981659876e-05,
+      "loss": 2.11,
+      "step": 10095500
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 3.53933893340126e-05,
+      "loss": 2.0997,
+      "step": 10096000
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 3.539266568636532e-05,
+      "loss": 2.1033,
+      "step": 10096500
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 3.539194203871804e-05,
+      "loss": 2.1105,
+      "step": 10097000
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 3.539121839107077e-05,
+      "loss": 2.1027,
+      "step": 10097500
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 3.5390494743423494e-05,
+      "loss": 2.1187,
+      "step": 10098000
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 3.5389771095776216e-05,
+      "loss": 2.1438,
+      "step": 10098500
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 3.538904744812894e-05,
+      "loss": 2.131,
+      "step": 10099000
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 3.538832380048166e-05,
+      "loss": 2.1234,
+      "step": 10099500
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 3.538760015283438e-05,
+      "loss": 2.1057,
+      "step": 10100000
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 3.53868779524824e-05,
+      "loss": 2.1114,
+      "step": 10100500
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 3.538615430483513e-05,
+      "loss": 2.1141,
+      "step": 10101000
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 3.538543065718785e-05,
+      "loss": 2.116,
+      "step": 10101500
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 3.538470700954057e-05,
+      "loss": 2.1367,
+      "step": 10102000
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 3.5383983361893294e-05,
+      "loss": 2.0767,
+      "step": 10102500
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 3.538325971424602e-05,
+      "loss": 2.1188,
+      "step": 10103000
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 3.538253751389404e-05,
+      "loss": 2.1192,
+      "step": 10103500
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 3.538181386624676e-05,
+      "loss": 2.1221,
+      "step": 10104000
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 3.538109021859948e-05,
+      "loss": 2.1275,
+      "step": 10104500
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 3.5380366570952205e-05,
+      "loss": 2.1072,
+      "step": 10105000
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 3.537964292330493e-05,
+      "loss": 2.1255,
+      "step": 10105500
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 3.537891927565765e-05,
+      "loss": 2.1018,
+      "step": 10106000
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 3.537819562801037e-05,
+      "loss": 2.1032,
+      "step": 10106500
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 3.5377471980363094e-05,
+      "loss": 2.1093,
+      "step": 10107000
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 3.537674833271582e-05,
+      "loss": 2.1085,
+      "step": 10107500
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 3.5376024685068546e-05,
+      "loss": 2.1094,
+      "step": 10108000
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 3.5375301037421275e-05,
+      "loss": 2.0979,
+      "step": 10108500
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 3.5374577389774e-05,
+      "loss": 2.0836,
+      "step": 10109000
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 3.537385374212672e-05,
+      "loss": 2.0949,
+      "step": 10109500
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 3.537313009447944e-05,
+      "loss": 2.1178,
+      "step": 10110000
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 3.5372406446832164e-05,
+      "loss": 2.0883,
+      "step": 10110500
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 3.5371682799184886e-05,
+      "loss": 2.1264,
+      "step": 10111000
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 3.53709605988329e-05,
+      "loss": 2.09,
+      "step": 10111500
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 3.5370236951185624e-05,
+      "loss": 2.0995,
+      "step": 10112000
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 3.5369513303538346e-05,
+      "loss": 2.1285,
+      "step": 10112500
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 3.536879255048166e-05,
+      "loss": 2.1151,
+      "step": 10113000
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 3.5368068902834384e-05,
+      "loss": 2.0835,
+      "step": 10113500
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 3.5367345255187106e-05,
+      "loss": 2.1137,
+      "step": 10114000
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 3.536662160753983e-05,
+      "loss": 2.1143,
+      "step": 10114500
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 3.536589795989255e-05,
+      "loss": 2.116,
+      "step": 10115000
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 3.536517575954057e-05,
+      "loss": 2.1189,
+      "step": 10115500
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 3.5364453559188595e-05,
+      "loss": 2.1276,
+      "step": 10116000
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 3.536372991154132e-05,
+      "loss": 2.0934,
+      "step": 10116500
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 3.536300626389404e-05,
+      "loss": 2.1313,
+      "step": 10117000
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 3.536228261624676e-05,
+      "loss": 2.0721,
+      "step": 10117500
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 3.5361558968599484e-05,
+      "loss": 2.0895,
+      "step": 10118000
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 3.5360835320952206e-05,
+      "loss": 2.1166,
+      "step": 10118500
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 3.536011167330493e-05,
+      "loss": 2.0963,
+      "step": 10119000
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 3.535938802565765e-05,
+      "loss": 2.122,
+      "step": 10119500
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 3.535866437801037e-05,
+      "loss": 2.1048,
+      "step": 10120000
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 3.5357940730363095e-05,
+      "loss": 2.1212,
+      "step": 10120500
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 3.5357217082715824e-05,
+      "loss": 2.0951,
+      "step": 10121000
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 3.5356493435068546e-05,
+      "loss": 2.0906,
+      "step": 10121500
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 3.535576978742127e-05,
+      "loss": 2.1043,
+      "step": 10122000
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 3.5355047587069284e-05,
+      "loss": 2.0728,
+      "step": 10122500
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 3.53543253867173e-05,
+      "loss": 2.0867,
+      "step": 10123000
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 3.535360173907003e-05,
+      "loss": 2.0928,
+      "step": 10123500
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 3.535287809142275e-05,
+      "loss": 2.1054,
+      "step": 10124000
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 3.535215444377547e-05,
+      "loss": 2.1018,
+      "step": 10124500
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 3.535143369071879e-05,
+      "loss": 2.1008,
+      "step": 10125000
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 3.535071004307151e-05,
+      "loss": 2.105,
+      "step": 10125500
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 3.534998639542423e-05,
+      "loss": 2.1305,
+      "step": 10126000
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 3.5349262747776956e-05,
+      "loss": 2.1059,
+      "step": 10126500
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 3.534853910012968e-05,
+      "loss": 2.1162,
+      "step": 10127000
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 3.53478154524824e-05,
+      "loss": 2.121,
+      "step": 10127500
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 3.534709180483512e-05,
+      "loss": 2.0983,
+      "step": 10128000
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 3.534636815718785e-05,
+      "loss": 2.1087,
+      "step": 10128500
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 3.534564595683587e-05,
+      "loss": 2.1009,
+      "step": 10129000
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 3.534492230918859e-05,
+      "loss": 2.1082,
+      "step": 10129500
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 3.534419866154131e-05,
+      "loss": 2.1222,
+      "step": 10130000
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 3.5343475013894034e-05,
+      "loss": 2.0979,
+      "step": 10130500
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 3.534275136624676e-05,
+      "loss": 2.093,
+      "step": 10131000
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 3.534202916589478e-05,
+      "loss": 2.1366,
+      "step": 10131500
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 3.53413055182475e-05,
+      "loss": 2.0985,
+      "step": 10132000
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 3.534058187060022e-05,
+      "loss": 2.1058,
+      "step": 10132500
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 3.533985822295295e-05,
+      "loss": 2.1194,
+      "step": 10133000
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 3.5339134575305674e-05,
+      "loss": 2.0976,
+      "step": 10133500
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 3.5338410927658396e-05,
+      "loss": 2.0871,
+      "step": 10134000
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 3.533768728001112e-05,
+      "loss": 2.1024,
+      "step": 10134500
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 3.533696363236384e-05,
+      "loss": 2.1072,
+      "step": 10135000
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 3.533623998471656e-05,
+      "loss": 2.1013,
+      "step": 10135500
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 3.533551778436458e-05,
+      "loss": 2.105,
+      "step": 10136000
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 3.53347941367173e-05,
+      "loss": 2.1342,
+      "step": 10136500
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 3.533407193636532e-05,
+      "loss": 2.1156,
+      "step": 10137000
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 3.5333348288718045e-05,
+      "loss": 2.103,
+      "step": 10137500
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 3.533262608836606e-05,
+      "loss": 2.1184,
+      "step": 10138000
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 3.533190244071878e-05,
+      "loss": 2.1141,
+      "step": 10138500
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 3.533117879307151e-05,
+      "loss": 2.1148,
+      "step": 10139000
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 3.5330455145424234e-05,
+      "loss": 2.1212,
+      "step": 10139500
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 3.5329731497776956e-05,
+      "loss": 2.1252,
+      "step": 10140000
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 3.532900785012968e-05,
+      "loss": 2.1205,
+      "step": 10140500
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 3.53282842024824e-05,
+      "loss": 2.1031,
+      "step": 10141000
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 3.532756055483512e-05,
+      "loss": 2.109,
+      "step": 10141500
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 3.532683690718785e-05,
+      "loss": 2.0919,
+      "step": 10142000
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 3.5326113259540574e-05,
+      "loss": 2.1027,
+      "step": 10142500
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 3.5325389611893297e-05,
+      "loss": 2.1076,
+      "step": 10143000
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 3.532466596424602e-05,
+      "loss": 2.1344,
+      "step": 10143500
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 3.5323943763894034e-05,
+      "loss": 2.1049,
+      "step": 10144000
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 3.5323220116246757e-05,
+      "loss": 2.0946,
+      "step": 10144500
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 3.532249646859948e-05,
+      "loss": 2.1283,
+      "step": 10145000
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 3.53217728209522e-05,
+      "loss": 2.1056,
+      "step": 10145500
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 3.532104917330493e-05,
+      "loss": 2.0956,
+      "step": 10146000
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 3.532032552565765e-05,
+      "loss": 2.1071,
+      "step": 10146500
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 3.5319601878010374e-05,
+      "loss": 2.1017,
+      "step": 10147000
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 3.5318878230363103e-05,
+      "loss": 2.1078,
+      "step": 10147500
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 3.5318154582715826e-05,
+      "loss": 2.0792,
+      "step": 10148000
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 3.531743093506855e-05,
+      "loss": 2.1029,
+      "step": 10148500
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 3.531670728742127e-05,
+      "loss": 2.1167,
+      "step": 10149000
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 3.5315985087069286e-05,
+      "loss": 2.1294,
+      "step": 10149500
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 3.531526143942201e-05,
+      "loss": 2.1129,
+      "step": 10150000
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 3.531453779177473e-05,
+      "loss": 2.1165,
+      "step": 10150500
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 3.531381414412745e-05,
+      "loss": 2.1068,
+      "step": 10151000
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 3.5313090496480175e-05,
+      "loss": 2.1112,
+      "step": 10151500
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 3.5312366848832904e-05,
+      "loss": 2.1092,
+      "step": 10152000
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 3.5311643201185626e-05,
+      "loss": 2.1378,
+      "step": 10152500
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 3.5310919553538355e-05,
+      "loss": 2.1203,
+      "step": 10153000
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 3.531019590589108e-05,
+      "loss": 2.0985,
+      "step": 10153500
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 3.530947370553909e-05,
+      "loss": 2.0924,
+      "step": 10154000
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 3.5308750057891815e-05,
+      "loss": 2.099,
+      "step": 10154500
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 3.530802641024454e-05,
+      "loss": 2.1033,
+      "step": 10155000
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 3.530730276259726e-05,
+      "loss": 2.1017,
+      "step": 10155500
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 3.5306580562245275e-05,
+      "loss": 2.0951,
+      "step": 10156000
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 3.5305856914598004e-05,
+      "loss": 2.1089,
+      "step": 10156500
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 3.5305133266950726e-05,
+      "loss": 2.1044,
+      "step": 10157000
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 3.530440961930345e-05,
+      "loss": 2.103,
+      "step": 10157500
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 3.530368597165617e-05,
+      "loss": 2.1001,
+      "step": 10158000
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 3.5302963771304186e-05,
+      "loss": 2.1024,
+      "step": 10158500
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 3.53022430182475e-05,
+      "loss": 2.1164,
+      "step": 10159000
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 3.5301519370600224e-05,
+      "loss": 2.1045,
+      "step": 10159500
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 3.5300795722952946e-05,
+      "loss": 2.1154,
+      "step": 10160000
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 3.530007207530567e-05,
+      "loss": 2.0998,
+      "step": 10160500
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 3.52993484276584e-05,
+      "loss": 2.1152,
+      "step": 10161000
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 3.529862478001112e-05,
+      "loss": 2.0962,
+      "step": 10161500
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 3.529790113236384e-05,
+      "loss": 2.1202,
+      "step": 10162000
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 3.5297177484716564e-05,
+      "loss": 2.1352,
+      "step": 10162500
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 3.5296453837069287e-05,
+      "loss": 2.1282,
+      "step": 10163000
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 3.529573018942201e-05,
+      "loss": 2.1083,
+      "step": 10163500
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 3.529500654177473e-05,
+      "loss": 2.1109,
+      "step": 10164000
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 3.529428289412745e-05,
+      "loss": 2.0879,
+      "step": 10164500
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 3.529355924648018e-05,
+      "loss": 2.0928,
+      "step": 10165000
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 3.52928370461282e-05,
+      "loss": 2.126,
+      "step": 10165500
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 3.529211484577621e-05,
+      "loss": 2.1132,
+      "step": 10166000
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 3.5291391198128936e-05,
+      "loss": 2.1046,
+      "step": 10166500
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 3.529066755048166e-05,
+      "loss": 2.1224,
+      "step": 10167000
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 3.528994390283438e-05,
+      "loss": 2.1248,
+      "step": 10167500
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 3.52892202551871e-05,
+      "loss": 2.0978,
+      "step": 10168000
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 3.528849805483513e-05,
+      "loss": 2.11,
+      "step": 10168500
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 3.528777585448315e-05,
+      "loss": 2.1371,
+      "step": 10169000
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 3.528705220683587e-05,
+      "loss": 2.1095,
+      "step": 10169500
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 3.528632855918859e-05,
+      "loss": 2.1371,
+      "step": 10170000
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 3.5285604911541314e-05,
+      "loss": 2.1178,
+      "step": 10170500
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 3.5284881263894036e-05,
+      "loss": 2.1129,
+      "step": 10171000
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 3.528415761624676e-05,
+      "loss": 2.1197,
+      "step": 10171500
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 3.528343396859948e-05,
+      "loss": 2.1004,
+      "step": 10172000
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 3.528271466283809e-05,
+      "loss": 2.106,
+      "step": 10172500
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 3.528199101519081e-05,
+      "loss": 2.1093,
+      "step": 10173000
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 3.5281267367543534e-05,
+      "loss": 2.1193,
+      "step": 10173500
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 3.5280543719896256e-05,
+      "loss": 2.0901,
+      "step": 10174000
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 3.527982007224898e-05,
+      "loss": 2.1182,
+      "step": 10174500
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 3.527909642460171e-05,
+      "loss": 2.1215,
+      "step": 10175000
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 3.527837277695443e-05,
+      "loss": 2.0913,
+      "step": 10175500
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 3.527764912930715e-05,
+      "loss": 2.1154,
+      "step": 10176000
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 3.527692548165988e-05,
+      "loss": 2.0852,
+      "step": 10176500
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 3.52762018340126e-05,
+      "loss": 2.0995,
+      "step": 10177000
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 3.5275478186365325e-05,
+      "loss": 2.1336,
+      "step": 10177500
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 3.527475453871805e-05,
+      "loss": 2.1288,
+      "step": 10178000
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 3.527403089107077e-05,
+      "loss": 2.1193,
+      "step": 10178500
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 3.527330724342349e-05,
+      "loss": 2.1213,
+      "step": 10179000
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 3.5272583595776214e-05,
+      "loss": 2.1178,
+      "step": 10179500
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 3.5271859948128936e-05,
+      "loss": 2.1031,
+      "step": 10180000
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 3.527113630048166e-05,
+      "loss": 2.1057,
+      "step": 10180500
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 3.527041265283438e-05,
+      "loss": 2.0922,
+      "step": 10181000
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 3.526968900518711e-05,
+      "loss": 2.0927,
+      "step": 10181500
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 3.5268966804835125e-05,
+      "loss": 2.1036,
+      "step": 10182000
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 3.526824315718785e-05,
+      "loss": 2.1107,
+      "step": 10182500
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 3.526751950954057e-05,
+      "loss": 2.1316,
+      "step": 10183000
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 3.52667958618933e-05,
+      "loss": 2.1473,
+      "step": 10183500
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 3.526607221424602e-05,
+      "loss": 2.1143,
+      "step": 10184000
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 3.526535001389404e-05,
+      "loss": 2.1082,
+      "step": 10184500
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 3.526462636624676e-05,
+      "loss": 2.1121,
+      "step": 10185000
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 3.526390271859948e-05,
+      "loss": 2.0948,
+      "step": 10185500
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 3.526317907095221e-05,
+      "loss": 2.1285,
+      "step": 10186000
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 3.526245542330493e-05,
+      "loss": 2.1175,
+      "step": 10186500
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 3.526173322295295e-05,
+      "loss": 2.1139,
+      "step": 10187000
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 3.526100957530567e-05,
+      "loss": 2.0972,
+      "step": 10187500
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 3.526028592765839e-05,
+      "loss": 2.1153,
+      "step": 10188000
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 3.5259562280011115e-05,
+      "loss": 2.0871,
+      "step": 10188500
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 3.525883863236384e-05,
+      "loss": 2.104,
+      "step": 10189000
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 3.525811498471656e-05,
+      "loss": 2.1058,
+      "step": 10189500
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 3.525739133706928e-05,
+      "loss": 2.1092,
+      "step": 10190000
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 3.5256669136717304e-05,
+      "loss": 2.1488,
+      "step": 10190500
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 3.525594693636532e-05,
+      "loss": 2.1031,
+      "step": 10191000
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 3.525522473601334e-05,
+      "loss": 2.1264,
+      "step": 10191500
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 3.5254501088366064e-05,
+      "loss": 2.1004,
+      "step": 10192000
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 3.5253777440718786e-05,
+      "loss": 2.1073,
+      "step": 10192500
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 3.525305379307151e-05,
+      "loss": 2.1287,
+      "step": 10193000
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 3.525233014542423e-05,
+      "loss": 2.1277,
+      "step": 10193500
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 3.525160649777696e-05,
+      "loss": 2.1333,
+      "step": 10194000
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 3.525088285012968e-05,
+      "loss": 2.0879,
+      "step": 10194500
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 3.52501606497777e-05,
+      "loss": 2.1002,
+      "step": 10195000
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 3.524943700213042e-05,
+      "loss": 2.1175,
+      "step": 10195500
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 3.524871335448314e-05,
+      "loss": 2.1135,
+      "step": 10196000
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 3.5247989706835864e-05,
+      "loss": 2.1047,
+      "step": 10196500
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 3.5247266059188586e-05,
+      "loss": 2.1317,
+      "step": 10197000
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 3.524654241154131e-05,
+      "loss": 2.1054,
+      "step": 10197500
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 3.524581876389403e-05,
+      "loss": 2.1159,
+      "step": 10198000
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 3.524509511624676e-05,
+      "loss": 2.107,
+      "step": 10198500
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 3.524437146859948e-05,
+      "loss": 2.1379,
+      "step": 10199000
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 3.524364782095221e-05,
+      "loss": 2.1186,
+      "step": 10199500
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 3.524292417330493e-05,
+      "loss": 2.1054,
+      "step": 10200000
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 3.5242200525657655e-05,
+      "loss": 2.1386,
+      "step": 10200500
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 3.524147687801038e-05,
+      "loss": 2.0921,
+      "step": 10201000
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 3.524075467765839e-05,
+      "loss": 2.1101,
+      "step": 10201500
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 3.5240031030011115e-05,
+      "loss": 2.1308,
+      "step": 10202000
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 3.523930882965914e-05,
+      "loss": 2.1321,
+      "step": 10202500
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 3.523858518201186e-05,
+      "loss": 2.102,
+      "step": 10203000
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 3.523786153436458e-05,
+      "loss": 2.1208,
+      "step": 10203500
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 3.5237137886717304e-05,
+      "loss": 2.1341,
+      "step": 10204000
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 3.523641423907003e-05,
+      "loss": 2.1154,
+      "step": 10204500
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 3.523569059142275e-05,
+      "loss": 2.1375,
+      "step": 10205000
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 3.5234968391070765e-05,
+      "loss": 2.1052,
+      "step": 10205500
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 3.5234244743423494e-05,
+      "loss": 2.1009,
+      "step": 10206000
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 3.5233521095776216e-05,
+      "loss": 2.0865,
+      "step": 10206500
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 3.523279744812894e-05,
+      "loss": 2.1158,
+      "step": 10207000
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 3.523207380048166e-05,
+      "loss": 2.1036,
+      "step": 10207500
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 3.523135015283439e-05,
+      "loss": 2.0814,
+      "step": 10208000
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 3.523062650518711e-05,
+      "loss": 2.1023,
+      "step": 10208500
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 3.5229902857539834e-05,
+      "loss": 2.1272,
+      "step": 10209000
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 3.5229179209892556e-05,
+      "loss": 2.1229,
+      "step": 10209500
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 3.522845556224528e-05,
+      "loss": 2.1319,
+      "step": 10210000
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 3.5227731914598e-05,
+      "loss": 2.0992,
+      "step": 10210500
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 3.5227009714246016e-05,
+      "loss": 2.1075,
+      "step": 10211000
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 3.522628606659874e-05,
+      "loss": 2.1068,
+      "step": 10211500
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 3.522556241895146e-05,
+      "loss": 2.092,
+      "step": 10212000
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 3.522483877130419e-05,
+      "loss": 2.1199,
+      "step": 10212500
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 3.522411512365691e-05,
+      "loss": 2.1159,
+      "step": 10213000
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 3.522339147600964e-05,
+      "loss": 2.1518,
+      "step": 10213500
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 3.5222669275657656e-05,
+      "loss": 2.1365,
+      "step": 10214000
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 3.522194562801038e-05,
+      "loss": 2.1085,
+      "step": 10214500
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 3.5221223427658394e-05,
+      "loss": 2.0976,
+      "step": 10215000
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 3.5220499780011116e-05,
+      "loss": 2.1006,
+      "step": 10215500
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 3.521977613236384e-05,
+      "loss": 2.1127,
+      "step": 10216000
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 3.521905248471656e-05,
+      "loss": 2.1273,
+      "step": 10216500
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 3.521833028436458e-05,
+      "loss": 2.1,
+      "step": 10217000
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 3.5217606636717305e-05,
+      "loss": 2.1045,
+      "step": 10217500
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 3.521688298907003e-05,
+      "loss": 2.1024,
+      "step": 10218000
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 3.521615934142275e-05,
+      "loss": 2.1136,
+      "step": 10218500
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 3.521543569377547e-05,
+      "loss": 2.0962,
+      "step": 10219000
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 3.5214712046128194e-05,
+      "loss": 2.1329,
+      "step": 10219500
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 3.5213988398480916e-05,
+      "loss": 2.1099,
+      "step": 10220000
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 3.521326475083364e-05,
+      "loss": 2.0751,
+      "step": 10220500
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 3.521254110318637e-05,
+      "loss": 2.1279,
+      "step": 10221000
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 3.521181890283439e-05,
+      "loss": 2.1184,
+      "step": 10221500
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 3.521109525518711e-05,
+      "loss": 2.108,
+      "step": 10222000
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 3.5210371607539834e-05,
+      "loss": 2.1202,
+      "step": 10222500
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 3.520964795989256e-05,
+      "loss": 2.1204,
+      "step": 10223000
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 3.520892431224528e-05,
+      "loss": 2.116,
+      "step": 10223500
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 3.5208200664598e-05,
+      "loss": 2.0814,
+      "step": 10224000
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 3.5207477016950723e-05,
+      "loss": 2.1033,
+      "step": 10224500
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 3.5206753369303446e-05,
+      "loss": 2.1197,
+      "step": 10225000
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 3.520602972165617e-05,
+      "loss": 2.1186,
+      "step": 10225500
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 3.520530607400889e-05,
+      "loss": 2.1058,
+      "step": 10226000
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 3.520458242636161e-05,
+      "loss": 2.1148,
+      "step": 10226500
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 3.5203860226009635e-05,
+      "loss": 2.1362,
+      "step": 10227000
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 3.520313802565765e-05,
+      "loss": 2.1019,
+      "step": 10227500
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 3.520241437801037e-05,
+      "loss": 2.1289,
+      "step": 10228000
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 3.52016907303631e-05,
+      "loss": 2.0993,
+      "step": 10228500
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 3.520096853001112e-05,
+      "loss": 2.1159,
+      "step": 10229000
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 3.520024488236384e-05,
+      "loss": 2.1293,
+      "step": 10229500
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 3.519952123471656e-05,
+      "loss": 2.1072,
+      "step": 10230000
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 3.519879758706929e-05,
+      "loss": 2.1008,
+      "step": 10230500
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 3.519807393942201e-05,
+      "loss": 2.1046,
+      "step": 10231000
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 3.5197350291774735e-05,
+      "loss": 2.0959,
+      "step": 10231500
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 3.519662664412746e-05,
+      "loss": 2.132,
+      "step": 10232000
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 3.519590299648018e-05,
+      "loss": 2.1051,
+      "step": 10232500
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 3.51951793488329e-05,
+      "loss": 2.1322,
+      "step": 10233000
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 3.5194455701185624e-05,
+      "loss": 2.0929,
+      "step": 10233500
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 3.5193732053538346e-05,
+      "loss": 2.1258,
+      "step": 10234000
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 3.519300840589107e-05,
+      "loss": 2.1223,
+      "step": 10234500
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 3.519228475824379e-05,
+      "loss": 2.1199,
+      "step": 10235000
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 3.519156111059652e-05,
+      "loss": 2.1067,
+      "step": 10235500
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 3.519083746294924e-05,
+      "loss": 2.108,
+      "step": 10236000
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 3.5190113815301964e-05,
+      "loss": 2.1081,
+      "step": 10236500
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 3.518939016765469e-05,
+      "loss": 2.1255,
+      "step": 10237000
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 3.5188666520007415e-05,
+      "loss": 2.1109,
+      "step": 10237500
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 3.518794287236014e-05,
+      "loss": 2.1141,
+      "step": 10238000
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 3.518721922471286e-05,
+      "loss": 2.105,
+      "step": 10238500
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 3.5186497024360875e-05,
+      "loss": 2.106,
+      "step": 10239000
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 3.51857733767136e-05,
+      "loss": 2.1004,
+      "step": 10239500
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 3.518505117636161e-05,
+      "loss": 2.098,
+      "step": 10240000
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 3.5184328976009635e-05,
+      "loss": 2.1135,
+      "step": 10240500
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 3.518360532836236e-05,
+      "loss": 2.1278,
+      "step": 10241000
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 3.518288312801037e-05,
+      "loss": 2.1082,
+      "step": 10241500
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 3.518216092765839e-05,
+      "loss": 2.1071,
+      "step": 10242000
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 3.518143728001112e-05,
+      "loss": 2.1356,
+      "step": 10242500
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 3.518071363236384e-05,
+      "loss": 2.1168,
+      "step": 10243000
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 3.517998998471657e-05,
+      "loss": 2.1038,
+      "step": 10243500
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 3.517926633706929e-05,
+      "loss": 2.1229,
+      "step": 10244000
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 3.5178542689422014e-05,
+      "loss": 2.1182,
+      "step": 10244500
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 3.5177819041774736e-05,
+      "loss": 2.1204,
+      "step": 10245000
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 3.517709539412746e-05,
+      "loss": 2.13,
+      "step": 10245500
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 3.517637174648018e-05,
+      "loss": 2.126,
+      "step": 10246000
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 3.51756480988329e-05,
+      "loss": 2.1271,
+      "step": 10246500
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 3.5174924451185625e-05,
+      "loss": 2.1293,
+      "step": 10247000
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 3.517420080353835e-05,
+      "loss": 2.1183,
+      "step": 10247500
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 3.517347715589107e-05,
+      "loss": 2.0908,
+      "step": 10248000
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 3.517275350824379e-05,
+      "loss": 2.1158,
+      "step": 10248500
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 3.5172031307891814e-05,
+      "loss": 2.123,
+      "step": 10249000
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 3.5171307660244536e-05,
+      "loss": 2.1139,
+      "step": 10249500
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 3.517058401259726e-05,
+      "loss": 2.1062,
+      "step": 10250000
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 3.516986036494999e-05,
+      "loss": 2.1003,
+      "step": 10250500
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 3.516913671730271e-05,
+      "loss": 2.1267,
+      "step": 10251000
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 3.5168414516950725e-05,
+      "loss": 2.1204,
+      "step": 10251500
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 3.516769086930345e-05,
+      "loss": 2.0959,
+      "step": 10252000
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 3.516696866895147e-05,
+      "loss": 2.096,
+      "step": 10252500
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 3.516624502130419e-05,
+      "loss": 2.1229,
+      "step": 10253000
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 3.5165521373656914e-05,
+      "loss": 2.1272,
+      "step": 10253500
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 3.5164797726009636e-05,
+      "loss": 2.0996,
+      "step": 10254000
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 3.516407407836236e-05,
+      "loss": 2.1232,
+      "step": 10254500
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 3.5163351878010374e-05,
+      "loss": 2.1184,
+      "step": 10255000
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 3.5162628230363096e-05,
+      "loss": 2.1159,
+      "step": 10255500
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 3.516190458271582e-05,
+      "loss": 2.1139,
+      "step": 10256000
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 3.516118093506854e-05,
+      "loss": 2.1479,
+      "step": 10256500
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 3.516045728742127e-05,
+      "loss": 2.0913,
+      "step": 10257000
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 3.515973363977399e-05,
+      "loss": 2.0999,
+      "step": 10257500
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 3.515900999212672e-05,
+      "loss": 2.108,
+      "step": 10258000
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 3.515828634447944e-05,
+      "loss": 2.1207,
+      "step": 10258500
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 3.5157562696832165e-05,
+      "loss": 2.1045,
+      "step": 10259000
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 3.515683904918489e-05,
+      "loss": 2.0791,
+      "step": 10259500
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 3.515611540153761e-05,
+      "loss": 2.0841,
+      "step": 10260000
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 3.515539175389033e-05,
+      "loss": 2.0996,
+      "step": 10260500
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 3.515466955353835e-05,
+      "loss": 2.1166,
+      "step": 10261000
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 3.515394590589107e-05,
+      "loss": 2.1033,
+      "step": 10261500
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 3.515322225824379e-05,
+      "loss": 2.1085,
+      "step": 10262000
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.515249861059652e-05,
+      "loss": 2.0997,
+      "step": 10262500
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.5151774962949243e-05,
+      "loss": 2.0946,
+      "step": 10263000
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.515105276259726e-05,
+      "loss": 2.1382,
+      "step": 10263500
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.515032911494998e-05,
+      "loss": 2.1323,
+      "step": 10264000
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.5149605467302703e-05,
+      "loss": 2.1098,
+      "step": 10264500
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.5148881819655426e-05,
+      "loss": 2.0981,
+      "step": 10265000
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.5148158172008155e-05,
+      "loss": 2.1097,
+      "step": 10265500
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.514743452436088e-05,
+      "loss": 2.1366,
+      "step": 10266000
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.51467108767136e-05,
+      "loss": 2.0885,
+      "step": 10266500
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.514598722906632e-05,
+      "loss": 2.1093,
+      "step": 10267000
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.5145265028714344e-05,
+      "loss": 2.1117,
+      "step": 10267500
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.514454282836236e-05,
+      "loss": 2.1112,
+      "step": 10268000
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.514381918071508e-05,
+      "loss": 2.1242,
+      "step": 10268500
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.5143095533067804e-05,
+      "loss": 2.1179,
+      "step": 10269000
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.5142371885420526e-05,
+      "loss": 2.0993,
+      "step": 10269500
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.514164968506855e-05,
+      "loss": 2.0976,
+      "step": 10270000
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.514092603742127e-05,
+      "loss": 2.1208,
+      "step": 10270500
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.514020238977399e-05,
+      "loss": 2.118,
+      "step": 10271000
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.5139478742126715e-05,
+      "loss": 2.116,
+      "step": 10271500
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.513875509447944e-05,
+      "loss": 2.1124,
+      "step": 10272000
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.513803144683216e-05,
+      "loss": 2.0891,
+      "step": 10272500
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.513730779918489e-05,
+      "loss": 2.1068,
+      "step": 10273000
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.513658415153761e-05,
+      "loss": 2.1024,
+      "step": 10273500
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.513586050389033e-05,
+      "loss": 2.1281,
+      "step": 10274000
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.513513830353835e-05,
+      "loss": 2.1109,
+      "step": 10274500
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.513441465589107e-05,
+      "loss": 2.0959,
+      "step": 10275000
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.51336910082438e-05,
+      "loss": 2.1353,
+      "step": 10275500
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.5132968807891815e-05,
+      "loss": 2.141,
+      "step": 10276000
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.513224516024454e-05,
+      "loss": 2.1089,
+      "step": 10276500
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.513152151259726e-05,
+      "loss": 2.0886,
+      "step": 10277000
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.513079786494998e-05,
+      "loss": 2.1127,
+      "step": 10277500
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.5130074217302704e-05,
+      "loss": 2.1377,
+      "step": 10278000
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.5129350569655427e-05,
+      "loss": 2.1022,
+      "step": 10278500
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.512862692200815e-05,
+      "loss": 2.1141,
+      "step": 10279000
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.512790327436087e-05,
+      "loss": 2.1152,
+      "step": 10279500
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.51271796267136e-05,
+      "loss": 2.1037,
+      "step": 10280000
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.512645742636162e-05,
+      "loss": 2.0929,
+      "step": 10280500
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.5125733778714345e-05,
+      "loss": 2.109,
+      "step": 10281000
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.512501013106707e-05,
+      "loss": 2.1228,
+      "step": 10281500
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.512428648341979e-05,
+      "loss": 2.1201,
+      "step": 10282000
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.512356283577251e-05,
+      "loss": 2.1225,
+      "step": 10282500
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.512284208271582e-05,
+      "loss": 2.1054,
+      "step": 10283000
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.512211843506855e-05,
+      "loss": 2.1151,
+      "step": 10283500
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.512139478742127e-05,
+      "loss": 2.1429,
+      "step": 10284000
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.512067258706929e-05,
+      "loss": 2.1193,
+      "step": 10284500
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.511994893942201e-05,
+      "loss": 2.0916,
+      "step": 10285000
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.511922529177473e-05,
+      "loss": 2.11,
+      "step": 10285500
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.5118501644127454e-05,
+      "loss": 2.1206,
+      "step": 10286000
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 3.5117777996480176e-05,
+      "loss": 2.1103,
+      "step": 10286500
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 3.51170543488329e-05,
+      "loss": 2.1012,
+      "step": 10287000
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 3.511633070118562e-05,
+      "loss": 2.0877,
+      "step": 10287500
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 3.511560705353835e-05,
+      "loss": 2.0958,
+      "step": 10288000
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 3.511488485318637e-05,
+      "loss": 2.0903,
+      "step": 10288500
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 3.5114161205539094e-05,
+      "loss": 2.1312,
+      "step": 10289000
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 3.5113437557891816e-05,
+      "loss": 2.1131,
+      "step": 10289500
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 3.511271391024454e-05,
+      "loss": 2.127,
+      "step": 10290000
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 3.511199026259726e-05,
+      "loss": 2.0806,
+      "step": 10290500
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 3.511126661494998e-05,
+      "loss": 2.1315,
+      "step": 10291000
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 3.5110542967302705e-05,
+      "loss": 2.1237,
+      "step": 10291500
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 3.510981931965543e-05,
+      "loss": 2.1159,
+      "step": 10292000
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 3.510909567200815e-05,
+      "loss": 2.0942,
+      "step": 10292500
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 3.510837202436087e-05,
+      "loss": 2.119,
+      "step": 10293000
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 3.51076483767136e-05,
+      "loss": 2.1045,
+      "step": 10293500
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 3.5106926176361616e-05,
+      "loss": 2.1205,
+      "step": 10294000
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 3.510620252871434e-05,
+      "loss": 2.0996,
+      "step": 10294500
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 3.510547888106706e-05,
+      "loss": 2.121,
+      "step": 10295000
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 3.510475523341979e-05,
+      "loss": 2.0856,
+      "step": 10295500
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 3.510403158577251e-05,
+      "loss": 2.1039,
+      "step": 10296000
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 3.510330938542053e-05,
+      "loss": 2.1253,
+      "step": 10296500
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 3.510258573777325e-05,
+      "loss": 2.1182,
+      "step": 10297000
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 3.510186209012597e-05,
+      "loss": 2.1173,
+      "step": 10297500
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 3.51011384424787e-05,
+      "loss": 2.1053,
+      "step": 10298000
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 3.510041479483142e-05,
+      "loss": 2.124,
+      "step": 10298500
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 3.5099691147184146e-05,
+      "loss": 2.0907,
+      "step": 10299000
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 3.509896749953687e-05,
+      "loss": 2.1057,
+      "step": 10299500
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 3.509824385188959e-05,
+      "loss": 2.1031,
+      "step": 10300000
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 3.509752020424231e-05,
+      "loss": 2.1287,
+      "step": 10300500
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 3.5096796556595034e-05,
+      "loss": 2.093,
+      "step": 10301000
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 3.509607435624305e-05,
+      "loss": 2.0915,
+      "step": 10301500
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 3.509535070859577e-05,
+      "loss": 2.1291,
+      "step": 10302000
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 3.50946270609485e-05,
+      "loss": 2.0945,
+      "step": 10302500
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 3.5093903413301224e-05,
+      "loss": 2.1464,
+      "step": 10303000
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 3.509317976565395e-05,
+      "loss": 2.0889,
+      "step": 10303500
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 3.5092456118006675e-05,
+      "loss": 2.1135,
+      "step": 10304000
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 3.509173391765469e-05,
+      "loss": 2.0937,
+      "step": 10304500
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 3.509101027000741e-05,
+      "loss": 2.0971,
+      "step": 10305000
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 3.5090286622360135e-05,
+      "loss": 2.0857,
+      "step": 10305500
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 3.508956297471286e-05,
+      "loss": 2.1109,
+      "step": 10306000
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 3.508883932706558e-05,
+      "loss": 2.0777,
+      "step": 10306500
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 3.50881156794183e-05,
+      "loss": 2.1212,
+      "step": 10307000
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 3.5087392031771024e-05,
+      "loss": 2.1044,
+      "step": 10307500
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 3.5086669831419046e-05,
+      "loss": 2.1279,
+      "step": 10308000
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 3.508594618377177e-05,
+      "loss": 2.114,
+      "step": 10308500
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 3.508522253612449e-05,
+      "loss": 2.1412,
+      "step": 10309000
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 3.508449888847721e-05,
+      "loss": 2.1124,
+      "step": 10309500
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 3.508377668812523e-05,
+      "loss": 2.1197,
+      "step": 10310000
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 3.508305304047796e-05,
+      "loss": 2.1028,
+      "step": 10310500
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 3.508233084012598e-05,
+      "loss": 2.1179,
+      "step": 10311000
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 3.50816071924787e-05,
+      "loss": 2.1142,
+      "step": 10311500
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 3.5080883544831424e-05,
+      "loss": 2.1107,
+      "step": 10312000
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 3.5080159897184146e-05,
+      "loss": 2.0965,
+      "step": 10312500
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 3.507943624953687e-05,
+      "loss": 2.1237,
+      "step": 10313000
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 3.507871260188959e-05,
+      "loss": 2.1081,
+      "step": 10313500
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 3.507798895424231e-05,
+      "loss": 2.0986,
+      "step": 10314000
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 3.5077265306595035e-05,
+      "loss": 2.1146,
+      "step": 10314500
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 3.507654310624305e-05,
+      "loss": 2.0997,
+      "step": 10315000
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 3.507582090589107e-05,
+      "loss": 2.0968,
+      "step": 10315500
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 3.5075097258243795e-05,
+      "loss": 2.1059,
+      "step": 10316000
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 3.507437361059652e-05,
+      "loss": 2.1141,
+      "step": 10316500
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 3.507364996294924e-05,
+      "loss": 2.13,
+      "step": 10317000
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 3.507292631530196e-05,
+      "loss": 2.126,
+      "step": 10317500
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 3.507220266765469e-05,
+      "loss": 2.1293,
+      "step": 10318000
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 3.507147902000741e-05,
+      "loss": 2.1024,
+      "step": 10318500
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 3.5070755372360136e-05,
+      "loss": 2.0927,
+      "step": 10319000
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 3.507003317200815e-05,
+      "loss": 2.1222,
+      "step": 10319500
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 3.506930952436088e-05,
+      "loss": 2.1126,
+      "step": 10320000
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 3.50685858767136e-05,
+      "loss": 2.124,
+      "step": 10320500
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 3.5067862229066325e-05,
+      "loss": 2.1235,
+      "step": 10321000
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 3.506713858141905e-05,
+      "loss": 2.0973,
+      "step": 10321500
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 3.506641493377177e-05,
+      "loss": 2.1164,
+      "step": 10322000
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 3.5065692733419785e-05,
+      "loss": 2.1189,
+      "step": 10322500
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 3.50649705330678e-05,
+      "loss": 2.095,
+      "step": 10323000
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 3.506424688542053e-05,
+      "loss": 2.1269,
+      "step": 10323500
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 3.506352323777325e-05,
+      "loss": 2.107,
+      "step": 10324000
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 3.5062799590125974e-05,
+      "loss": 2.0958,
+      "step": 10324500
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 3.5062075942478696e-05,
+      "loss": 2.1201,
+      "step": 10325000
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 3.5061352294831425e-05,
+      "loss": 2.1084,
+      "step": 10325500
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 3.506062864718415e-05,
+      "loss": 2.1035,
+      "step": 10326000
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 3.505990499953687e-05,
+      "loss": 2.1009,
+      "step": 10326500
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 3.505918135188959e-05,
+      "loss": 2.0851,
+      "step": 10327000
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 3.5058457704242314e-05,
+      "loss": 2.0958,
+      "step": 10327500
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 3.5057734056595036e-05,
+      "loss": 2.0774,
+      "step": 10328000
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 3.505701040894776e-05,
+      "loss": 2.1328,
+      "step": 10328500
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 3.505628820859578e-05,
+      "loss": 2.1031,
+      "step": 10329000
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 3.5055566008243796e-05,
+      "loss": 2.1249,
+      "step": 10329500
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 3.505484236059652e-05,
+      "loss": 2.0982,
+      "step": 10330000
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 3.505411871294924e-05,
+      "loss": 2.0988,
+      "step": 10330500
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 3.505339506530196e-05,
+      "loss": 2.1075,
+      "step": 10331000
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 3.5052671417654685e-05,
+      "loss": 2.1253,
+      "step": 10331500
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 3.505194777000741e-05,
+      "loss": 2.1128,
+      "step": 10332000
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 3.505122412236013e-05,
+      "loss": 2.1027,
+      "step": 10332500
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 3.505050047471286e-05,
+      "loss": 2.1191,
+      "step": 10333000
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 3.504977682706558e-05,
+      "loss": 2.0969,
+      "step": 10333500
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 3.50490531794183e-05,
+      "loss": 2.11,
+      "step": 10334000
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 3.504832953177103e-05,
+      "loss": 2.1086,
+      "step": 10334500
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 3.5047605884123754e-05,
+      "loss": 2.0911,
+      "step": 10335000
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 3.5046882236476477e-05,
+      "loss": 2.1363,
+      "step": 10335500
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 3.50461585888292e-05,
+      "loss": 2.1077,
+      "step": 10336000
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 3.504543783577251e-05,
+      "loss": 2.1242,
+      "step": 10336500
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 3.504471418812523e-05,
+      "loss": 2.1258,
+      "step": 10337000
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 3.504399054047796e-05,
+      "loss": 2.1157,
+      "step": 10337500
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 3.504326689283068e-05,
+      "loss": 2.1037,
+      "step": 10338000
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 3.50425432451834e-05,
+      "loss": 2.1157,
+      "step": 10338500
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 3.5041819597536126e-05,
+      "loss": 2.1323,
+      "step": 10339000
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 3.504109594988885e-05,
+      "loss": 2.1242,
+      "step": 10339500
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 3.504037230224158e-05,
+      "loss": 2.1248,
+      "step": 10340000
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 3.50396486545943e-05,
+      "loss": 2.1205,
+      "step": 10340500
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 3.503892500694702e-05,
+      "loss": 2.1261,
+      "step": 10341000
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 3.503820425389033e-05,
+      "loss": 2.1112,
+      "step": 10341500
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 3.503748060624306e-05,
+      "loss": 2.1174,
+      "step": 10342000
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 3.503675695859578e-05,
+      "loss": 2.1267,
+      "step": 10342500
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 3.5036033310948504e-05,
+      "loss": 2.1154,
+      "step": 10343000
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 3.5035309663301226e-05,
+      "loss": 2.1113,
+      "step": 10343500
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 3.503458601565395e-05,
+      "loss": 2.0955,
+      "step": 10344000
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 3.503386236800667e-05,
+      "loss": 2.1247,
+      "step": 10344500
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 3.5033140167654686e-05,
+      "loss": 2.0924,
+      "step": 10345000
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 3.503241796730271e-05,
+      "loss": 2.1218,
+      "step": 10345500
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 3.503169431965543e-05,
+      "loss": 2.1285,
+      "step": 10346000
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 3.503097067200815e-05,
+      "loss": 2.1402,
+      "step": 10346500
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 3.5030247024360875e-05,
+      "loss": 2.1073,
+      "step": 10347000
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 3.50295233767136e-05,
+      "loss": 2.1025,
+      "step": 10347500
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 3.5028799729066326e-05,
+      "loss": 2.0933,
+      "step": 10348000
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 3.502807752871434e-05,
+      "loss": 2.1174,
+      "step": 10348500
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 3.5027353881067064e-05,
+      "loss": 2.1067,
+      "step": 10349000
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 3.5026630233419786e-05,
+      "loss": 2.1269,
+      "step": 10349500
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 3.502590658577251e-05,
+      "loss": 2.1089,
+      "step": 10350000
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 3.502518293812523e-05,
+      "loss": 2.1373,
+      "step": 10350500
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 3.502445929047796e-05,
+      "loss": 2.1092,
+      "step": 10351000
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 3.502373564283068e-05,
+      "loss": 2.1041,
+      "step": 10351500
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 3.5023011995183404e-05,
+      "loss": 2.1282,
+      "step": 10352000
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 3.5022288347536126e-05,
+      "loss": 2.1508,
+      "step": 10352500
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 3.502156469988885e-05,
+      "loss": 2.1082,
+      "step": 10353000
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 3.502084105224157e-05,
+      "loss": 2.1203,
+      "step": 10353500
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 3.502011740459429e-05,
+      "loss": 2.125,
+      "step": 10354000
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 3.5019393756947015e-05,
+      "loss": 2.1222,
+      "step": 10354500
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 3.5018670109299744e-05,
+      "loss": 2.1277,
+      "step": 10355000
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 3.5017946461652467e-05,
+      "loss": 2.1167,
+      "step": 10355500
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 3.501722281400519e-05,
+      "loss": 2.1318,
+      "step": 10356000
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 3.50165035082438e-05,
+      "loss": 2.1085,
+      "step": 10356500
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 3.501577986059652e-05,
+      "loss": 2.1203,
+      "step": 10357000
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 3.501505621294924e-05,
+      "loss": 2.1128,
+      "step": 10357500
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 3.5014332565301964e-05,
+      "loss": 2.1036,
+      "step": 10358000
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 3.501360891765469e-05,
+      "loss": 2.1303,
+      "step": 10358500
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 3.501288527000741e-05,
+      "loss": 2.1141,
+      "step": 10359000
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 3.501216162236013e-05,
+      "loss": 2.1066,
+      "step": 10359500
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 3.501143797471286e-05,
+      "loss": 2.1166,
+      "step": 10360000
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 3.501071432706558e-05,
+      "loss": 2.0972,
+      "step": 10360500
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 3.50099921267136e-05,
+      "loss": 2.1028,
+      "step": 10361000
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 3.500926847906632e-05,
+      "loss": 2.1232,
+      "step": 10361500
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 3.500854483141904e-05,
+      "loss": 2.1076,
+      "step": 10362000
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 3.5007821183771765e-05,
+      "loss": 2.1259,
+      "step": 10362500
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 3.5007097536124494e-05,
+      "loss": 2.1214,
+      "step": 10363000
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 3.5006373888477216e-05,
+      "loss": 2.1293,
+      "step": 10363500
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 3.500565024082994e-05,
+      "loss": 2.0928,
+      "step": 10364000
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.6665779107038252,
+      "eval_accuracy_mlm": 0.6310435936946435,
+      "eval_accuracy_nsp": 0.8573464776596866,
+      "eval_loss": 2.1807310581207275,
+      "eval_runtime": 331.5713,
+      "eval_samples_per_second": 1316.115,
+      "eval_steps_per_second": 54.839,
+      "step": 10364160
     }
   ],
   "max_steps": 34547200,
   "num_train_epochs": 100,
-  "total_flos": 9.466493397162928e+18,
+  "total_flos": 1.4199760659998734e+19,
   "trial_name": null,
   "trial_params": null
 }