diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,8951 @@
+{
+  "best_metric": 0.35228549223252686,
+  "best_model_checkpoint": "esm2_t12_35M_lora_binding_sites_2023-09-24_04-19-04/checkpoint-148440",
+  "epoch": 2.0,
+  "eval_steps": 500,
+  "global_step": 296880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00037015670298819826,
+      "loss": 0.5221,
+      "step": 200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0003701563931502581,
+      "loss": 0.3727,
+      "step": 400
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00037015587606324446,
+      "loss": 0.3152,
+      "step": 600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00037015515172773665,
+      "loss": 0.2911,
+      "step": 800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003701542201445457,
+      "loss": 0.2675,
+      "step": 1000
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003701530813147147,
+      "loss": 0.2512,
+      "step": 1200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00037015173523951894,
+      "loss": 0.2466,
+      "step": 1400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00037015018192046586,
+      "loss": 0.2271,
+      "step": 1600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003701484213592947,
+      "loss": 0.2265,
+      "step": 1800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000370146453557977,
+      "loss": 0.2299,
+      "step": 2000
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0003701442899094131,
+      "loss": 0.22,
+      "step": 2200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003701419086708161,
+      "loss": 0.2157,
+      "step": 2400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00037013932019936536,
+      "loss": 0.215,
+      "step": 2600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003701365244979595,
+      "loss": 0.2036,
+      "step": 2800
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00037013352156972917,
+      "loss": 0.1992,
+      "step": 3000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00037013031141803696,
+      "loss": 0.2077,
+      "step": 3200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00037012689404647756,
+      "loss": 0.1882,
+      "step": 3400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0003701232694588778,
+      "loss": 0.1935,
+      "step": 3600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00037011943765929636,
+      "loss": 0.1966,
+      "step": 3800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003701154193624823,
+      "loss": 0.1852,
+      "step": 4000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003701111741880463,
+      "loss": 0.1871,
+      "step": 4200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00037010672181517294,
+      "loss": 0.1653,
+      "step": 4400
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003701020622488479,
+      "loss": 0.1865,
+      "step": 4600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00037009719549428895,
+      "loss": 0.1746,
+      "step": 4800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0003700921215569458,
+      "loss": 0.1787,
+      "step": 5000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00037008684044250026,
+      "loss": 0.1873,
+      "step": 5200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00037008135215686603,
+      "loss": 0.1742,
+      "step": 5400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003700756567061889,
+      "loss": 0.1777,
+      "step": 5600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003700697540968466,
+      "loss": 0.1792,
+      "step": 5800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00037006364433544883,
+      "loss": 0.1794,
+      "step": 6000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003700573274288372,
+      "loss": 0.1708,
+      "step": 6200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00037005080338408546,
+      "loss": 0.161,
+      "step": 6400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0003700440722084991,
+      "loss": 0.1661,
+      "step": 6600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00037003720431783943,
+      "loss": 0.1833,
+      "step": 6800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003700300609745444,
+      "loss": 0.1802,
+      "step": 7000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00037002271052364195,
+      "loss": 0.1557,
+      "step": 7200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003700151529733631,
+      "loss": 0.1763,
+      "step": 7400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003700073883321708,
+      "loss": 0.1661,
+      "step": 7600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0003699994166087597,
+      "loss": 0.152,
+      "step": 7800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00036999123781205663,
+      "loss": 0.1616,
+      "step": 8000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00036998285195121995,
+      "loss": 0.1552,
+      "step": 8200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00036997425903564034,
+      "loss": 0.1595,
+      "step": 8400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00036996545907493994,
+      "loss": 0.1567,
+      "step": 8600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0003699564520789729,
+      "loss": 0.1572,
+      "step": 8800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00036994723805782516,
+      "loss": 0.1628,
+      "step": 9000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00036993781702181466,
+      "loss": 0.1562,
+      "step": 9200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00036992818898149083,
+      "loss": 0.1583,
+      "step": 9400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00036991835394763526,
+      "loss": 0.1514,
+      "step": 9600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000369908311931261,
+      "loss": 0.1651,
+      "step": 9800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003698980629436131,
+      "loss": 0.1488,
+      "step": 10000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003698876069961683,
+      "loss": 0.1471,
+      "step": 10200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003698769441006351,
+      "loss": 0.1516,
+      "step": 10400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003698660742689538,
+      "loss": 0.1411,
+      "step": 10600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0003698549975132963,
+      "loss": 0.1573,
+      "step": 10800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00036984377077907405,
+      "loss": 0.1472,
+      "step": 11000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003698322812473697,
+      "loss": 0.1445,
+      "step": 11200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003698205848295303,
+      "loss": 0.1544,
+      "step": 11400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003698086815386536,
+      "loss": 0.1545,
+      "step": 11600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00036979657138806857,
+      "loss": 0.1405,
+      "step": 11800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00036978425439133624,
+      "loss": 0.1558,
+      "step": 12000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00036977173056224906,
+      "loss": 0.154,
+      "step": 12200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003697589999148311,
+      "loss": 0.1572,
+      "step": 12400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0003697460624633381,
+      "loss": 0.1462,
+      "step": 12600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003697329182222573,
+      "loss": 0.1413,
+      "step": 12800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003697195672063075,
+      "loss": 0.1513,
+      "step": 13000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00036970600943043915,
+      "loss": 0.1496,
+      "step": 13200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00036969231424668924,
+      "loss": 0.1537,
+      "step": 13400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003696783440303689,
+      "loss": 0.147,
+      "step": 13600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003696641671002913,
+      "loss": 0.153,
+      "step": 13800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0003696497834723318,
+      "loss": 0.1424,
+      "step": 14000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003696351931625971,
+      "loss": 0.1413,
+      "step": 14200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00036962039618742523,
+      "loss": 0.1382,
+      "step": 14400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00036960539256338583,
+      "loss": 0.1432,
+      "step": 14600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003695901823072799,
+      "loss": 0.1377,
+      "step": 14800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003695747654361396,
+      "loss": 0.1535,
+      "step": 15000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003695591419672288,
+      "loss": 0.1338,
+      "step": 15200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0003695433915821276,
+      "loss": 0.1478,
+      "step": 15400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003695273560031605,
+      "loss": 0.1427,
+      "step": 15600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00036951111387951155,
+      "loss": 0.149,
+      "step": 15800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00036949466522936866,
+      "loss": 0.1435,
+      "step": 16000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003694780100711509,
+      "loss": 0.1281,
+      "step": 16200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00036946123324535805,
+      "loss": 0.156,
+      "step": 16400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003694441661594781,
+      "loss": 0.1463,
+      "step": 16600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0003694268926220718,
+      "loss": 0.1403,
+      "step": 16800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00036940941265248186,
+      "loss": 0.1346,
+      "step": 17000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003693917262702823,
+      "loss": 0.1271,
+      "step": 17200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003693738334952782,
+      "loss": 0.1397,
+      "step": 17400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003693557343475058,
+      "loss": 0.1281,
+      "step": 17600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003693374288472324,
+      "loss": 0.1448,
+      "step": 17800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00036931891701495626,
+      "loss": 0.1354,
+      "step": 18000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003693001988714069,
+      "loss": 0.1408,
+      "step": 18200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0003692812744375448,
+      "loss": 0.131,
+      "step": 18400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003692621437345614,
+      "loss": 0.1347,
+      "step": 18600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00036924280678387906,
+      "loss": 0.1622,
+      "step": 18800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00036922326360715115,
+      "loss": 0.1395,
+      "step": 19000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003692035142262621,
+      "loss": 0.1348,
+      "step": 19200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00036918355866332706,
+      "loss": 0.1491,
+      "step": 19400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003691633969406922,
+      "loss": 0.1351,
+      "step": 19600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00036914302908093444,
+      "loss": 0.1387,
+      "step": 19800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0003691224551068615,
+      "loss": 0.1466,
+      "step": 20000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003691016750415121,
+      "loss": 0.1238,
+      "step": 20200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00036908068890815544,
+      "loss": 0.1465,
+      "step": 20400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00036905949673029186,
+      "loss": 0.1405,
+      "step": 20600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003690382060350824,
+      "loss": 0.1367,
+      "step": 20800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003690166028695521,
+      "loss": 0.138,
+      "step": 21000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00036899479373127795,
+      "loss": 0.1339,
+      "step": 21200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0003689727786446815,
+      "loss": 0.1398,
+      "step": 21400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003689505576344153,
+      "loss": 0.1278,
+      "step": 21600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003689281307253622,
+      "loss": 0.1472,
+      "step": 21800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00036890549794263563,
+      "loss": 0.1371,
+      "step": 22000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00036888265931157973,
+      "loss": 0.1401,
+      "step": 22200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000368859614857769,
+      "loss": 0.1349,
+      "step": 22400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003688363646070085,
+      "loss": 0.1389,
+      "step": 22600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0003688130263772538,
+      "loss": 0.1367,
+      "step": 22800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000368789365639588,
+      "loss": 0.138,
+      "step": 23000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000368765499183637,
+      "loss": 0.1284,
+      "step": 23200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00036874142703612636,
+      "loss": 0.1224,
+      "step": 23400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00036871714922401174,
+      "loss": 0.1329,
+      "step": 23600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00036869266577447944,
+      "loss": 0.1408,
+      "step": 23800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003686679767149457,
+      "loss": 0.1425,
+      "step": 24000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003686430820730573,
+      "loss": 0.1288,
+      "step": 24200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0003686179818766911,
+      "loss": 0.1299,
+      "step": 24400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00036859267615395395,
+      "loss": 0.1377,
+      "step": 24600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00036856729300041623,
+      "loss": 0.1319,
+      "step": 24800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00036854157733745503,
+      "loss": 0.1284,
+      "step": 25000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00036851565623368017,
+      "loss": 0.1223,
+      "step": 25200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003684895297181181,
+      "loss": 0.146,
+      "step": 25400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0003684631978200251,
+      "loss": 0.1486,
+      "step": 25600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00036843666056888733,
+      "loss": 0.1227,
+      "step": 25800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003684099179944211,
+      "loss": 0.1254,
+      "step": 26000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003683829701265725,
+      "loss": 0.1289,
+      "step": 26200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003683558169955177,
+      "loss": 0.129,
+      "step": 26400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00036832845863166245,
+      "loss": 0.1296,
+      "step": 26600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003683008950656426,
+      "loss": 0.1338,
+      "step": 26800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003682731263283235,
+      "loss": 0.1383,
+      "step": 27000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003682452928304226,
+      "loss": 0.1334,
+      "step": 27200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0003682171148694867,
+      "loss": 0.1387,
+      "step": 27400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003681887318310681,
+      "loss": 0.1242,
+      "step": 27600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036816014374694997,
+      "loss": 0.1278,
+      "step": 27800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000368131350649145,
+      "loss": 0.133,
+      "step": 28000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003681023525698955,
+      "loss": 0.1261,
+      "step": 28200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003680731495416734,
+      "loss": 0.1358,
+      "step": 28400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00036804374159717996,
+      "loss": 0.1362,
+      "step": 28600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0003680141287693461,
+      "loss": 0.1291,
+      "step": 28800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036798431109133197,
+      "loss": 0.1252,
+      "step": 29000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003679542885965272,
+      "loss": 0.1286,
+      "step": 29200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003679240613185508,
+      "loss": 0.1281,
+      "step": 29400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003678937819606453,
+      "loss": 0.1355,
+      "step": 29600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00036786314624159095,
+      "loss": 0.1357,
+      "step": 29800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003678323058414256,
+      "loss": 0.1293,
+      "step": 30000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003678012607946841,
+      "loss": 0.132,
+      "step": 30200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0003677700111361304,
+      "loss": 0.1237,
+      "step": 30400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003677385569007577,
+      "loss": 0.1306,
+      "step": 30600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003677068981237883,
+      "loss": 0.1231,
+      "step": 30800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036767519466573924,
+      "loss": 0.1405,
+      "step": 31000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003676431279344224,
+      "loss": 0.127,
+      "step": 31200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00036761085676836956,
+      "loss": 0.1336,
+      "step": 31400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003675783812037177,
+      "loss": 0.1346,
+      "step": 31600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0003675457012768328,
+      "loss": 0.117,
+      "step": 31800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00036751281702430955,
+      "loss": 0.1277,
+      "step": 32000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003674797284829716,
+      "loss": 0.1181,
+      "step": 32200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003674464356898712,
+      "loss": 0.124,
+      "step": 32400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003674129386822893,
+      "loss": 0.137,
+      "step": 32600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003673792374977358,
+      "loss": 0.1154,
+      "step": 32800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000367345332173949,
+      "loss": 0.1548,
+      "step": 33000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0003673112227488958,
+      "loss": 0.1304,
+      "step": 33200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003672769092607718,
+      "loss": 0.113,
+      "step": 33400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003672423917480009,
+      "loss": 0.1383,
+      "step": 33600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003672076702492356,
+      "loss": 0.1286,
+      "step": 33800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003671727448033568,
+      "loss": 0.1233,
+      "step": 34000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003671376154494739,
+      "loss": 0.1311,
+      "step": 34200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00036710228222692435,
+      "loss": 0.1342,
+      "step": 34400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003670669233674915,
+      "loss": 0.1317,
+      "step": 34600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0003670311835453819,
+      "loss": 0.1343,
+      "step": 34800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00036699523997378743,
+      "loss": 0.1177,
+      "step": 35000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00036695909269295746,
+      "loss": 0.1269,
+      "step": 35200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003669227417433694,
+      "loss": 0.1374,
+      "step": 35400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003668861871657289,
+      "loss": 0.1216,
+      "step": 35600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00036684942900096955,
+      "loss": 0.1234,
+      "step": 35800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00036681246729025277,
+      "loss": 0.1312,
+      "step": 36000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0003667753020749681,
+      "loss": 0.1219,
+      "step": 36200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003667379333967329,
+      "loss": 0.1231,
+      "step": 36400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00036670036129739236,
+      "loss": 0.1236,
+      "step": 36600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003666625858190194,
+      "loss": 0.1299,
+      "step": 36800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00036662460700391487,
+      "loss": 0.1284,
+      "step": 37000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003665864248946071,
+      "loss": 0.1276,
+      "step": 37200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00036654803953385206,
+      "loss": 0.1227,
+      "step": 37400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00036650945096463363,
+      "loss": 0.1275,
+      "step": 37600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0003664706592301629,
+      "loss": 0.1338,
+      "step": 37800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00036643166437387857,
+      "loss": 0.1248,
+      "step": 38000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00036639246643944696,
+      "loss": 0.1234,
+      "step": 38200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00036635306547076144,
+      "loss": 0.1274,
+      "step": 38400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003663134615119431,
+      "loss": 0.1211,
+      "step": 38600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00036627365460734014,
+      "loss": 0.1162,
+      "step": 38800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0003662336448015281,
+      "loss": 0.1093,
+      "step": 39000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00036619363370715154,
+      "loss": 0.1281,
+      "step": 39200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000366153219247501,
+      "loss": 0.1143,
+      "step": 39400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036611260202150397,
+      "loss": 0.1301,
+      "step": 39600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003660717820746435,
+      "loss": 0.1312,
+      "step": 39800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0003660307594526293,
+      "loss": 0.1249,
+      "step": 40000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036598953420139834,
+      "loss": 0.13,
+      "step": 40200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036594810636711434,
+      "loss": 0.1215,
+      "step": 40400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000365906475996168,
+      "loss": 0.1347,
+      "step": 40600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00036586464313517655,
+      "loss": 0.1219,
+      "step": 40800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036582260783098424,
+      "loss": 0.1228,
+      "step": 41000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000365780370130662,
+      "loss": 0.1263,
+      "step": 41200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003657379300815072,
+      "loss": 0.1278,
+      "step": 41400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036569528773104393,
+      "loss": 0.1191,
+      "step": 41600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003656526578530694,
+      "loss": 0.1304,
+      "step": 41800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00036560961205437574,
+      "loss": 0.1327,
+      "step": 42000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0003655663640980633,
+      "loss": 0.1281,
+      "step": 42200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003655229140325608,
+      "loss": 0.1266,
+      "step": 42400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003654792619065234,
+      "loss": 0.1142,
+      "step": 42600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003654354077688324,
+      "loss": 0.1087,
+      "step": 42800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003653913516685954,
+      "loss": 0.1197,
+      "step": 43000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036534709365514605,
+      "loss": 0.1249,
+      "step": 43200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0003653026337780443,
+      "loss": 0.1217,
+      "step": 43400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00036525797208707584,
+      "loss": 0.1167,
+      "step": 43600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0003652133334513314,
+      "loss": 0.1265,
+      "step": 43800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00036516826929133405,
+      "loss": 0.1287,
+      "step": 44000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00036512300346793045,
+      "loss": 0.1254,
+      "step": 44200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00036507753603180906,
+      "loss": 0.1234,
+      "step": 44400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000365031867033884,
+      "loss": 0.1256,
+      "step": 44600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00036498599652529513,
+      "loss": 0.1295,
+      "step": 44800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00036493992455740784,
+      "loss": 0.129,
+      "step": 45000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00036489365118181326,
+      "loss": 0.1252,
+      "step": 45200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000364847176450328,
+      "loss": 0.1276,
+      "step": 45400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00036480050041499414,
+      "loss": 0.1102,
+      "step": 45600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003647536231280792,
+      "loss": 0.125,
+      "step": 45800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00036470678053490165,
+      "loss": 0.1421,
+      "step": 46000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00036465950190812875,
+      "loss": 0.1281,
+      "step": 46200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000364612022187664,
+      "loss": 0.1147,
+      "step": 46400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0003645643414266747,
+      "loss": 0.1263,
+      "step": 46600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003645164596785535,
+      "loss": 0.1255,
+      "step": 46800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00036446837699691823,
+      "loss": 0.1227,
+      "step": 47000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00036442009343561136,
+      "loss": 0.1171,
+      "step": 47200
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00036437160904870047,
+      "loss": 0.1179,
+      "step": 47400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00036432292389047827,
+      "loss": 0.1258,
+      "step": 47600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003642740380154618,
+      "loss": 0.1297,
+      "step": 47800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00036422495147839327,
+      "loss": 0.1272,
+      "step": 48000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0003641756643342394,
+      "loss": 0.1152,
+      "step": 48200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003641264245754532,
+      "loss": 0.1162,
+      "step": 48400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00036407673738527173,
+      "loss": 0.1195,
+      "step": 48600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00036402684975397404,
+      "loss": 0.1252,
+      "step": 48800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003639767617374241,
+      "loss": 0.1175,
+      "step": 49000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00036392647339170993,
+      "loss": 0.1249,
+      "step": 49200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003638759847731442,
+      "loss": 0.1199,
+      "step": 49400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0003638252959382637,
+      "loss": 0.122,
+      "step": 49600
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00036377440694382947,
+      "loss": 0.1109,
+      "step": 49800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003637233178468266,
+      "loss": 0.1171,
+      "step": 50000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00036367202870446437,
+      "loss": 0.1213,
+      "step": 50200
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003636207975172021,
+      "loss": 0.1156,
+      "step": 50400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003635691094561523,
+      "loss": 0.1257,
+      "step": 50600
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00036351722152242464,
+      "loss": 0.1179,
+      "step": 50800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003634651337741228,
+      "loss": 0.1287,
+      "step": 51000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0003634128462695743,
+      "loss": 0.1182,
+      "step": 51200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003633606219999926,
+      "loss": 0.1371,
+      "step": 51400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00036330793615687614,
+      "loss": 0.1055,
+      "step": 51600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00036325505073354187,
+      "loss": 0.1216,
+      "step": 51800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00036320196578921055,
+      "loss": 0.1207,
+      "step": 52000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0003631486813833263,
+      "loss": 0.131,
+      "step": 52200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00036309519757555674,
+      "loss": 0.12,
+      "step": 52400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00036304151442579253,
+      "loss": 0.1199,
+      "step": 52600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0003629876319941478,
+      "loss": 0.116,
+      "step": 52800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00036293355034095967,
+      "loss": 0.1134,
+      "step": 53000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00036287926952678845,
+      "loss": 0.1133,
+      "step": 53200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00036282478961241737,
+      "loss": 0.1254,
+      "step": 53400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00036277011065885274,
+      "loss": 0.1207,
+      "step": 53600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00036271523272732363,
+      "loss": 0.1222,
+      "step": 53800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00036266015587928203,
+      "loss": 0.1161,
+      "step": 54000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003626048801764025,
+      "loss": 0.1216,
+      "step": 54200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00036254940568058254,
+      "loss": 0.1253,
+      "step": 54400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000362493732453942,
+      "loss": 0.1154,
+      "step": 54600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00036243786055882343,
+      "loss": 0.1163,
+      "step": 54800
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00036238179005779176,
+      "loss": 0.1154,
+      "step": 55000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00036232552101363446,
+      "loss": 0.1115,
+      "step": 55200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0003622693363205972,
+      "loss": 0.1203,
+      "step": 55400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00036221267137136654,
+      "loss": 0.1241,
+      "step": 55600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00036215580806838815,
+      "loss": 0.1113,
+      "step": 55800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003620987464753372,
+      "loss": 0.1275,
+      "step": 56000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003620417734481885,
+      "loss": 0.1238,
+      "step": 56200
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0003619843164575562,
+      "loss": 0.1158,
+      "step": 56400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00036192666136888646,
+      "loss": 0.113,
+      "step": 56600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000361868808246741,
+      "loss": 0.114,
+      "step": 56800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00036181075715590336,
+      "loss": 0.1304,
+      "step": 57000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00036175250816137877,
+      "loss": 0.1105,
+      "step": 57200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003616940613283941,
+      "loss": 0.1187,
+      "step": 57400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00036163541672239765,
+      "loss": 0.1166,
+      "step": 57600
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00036157657440905933,
+      "loss": 0.1209,
+      "step": 57800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00036151753445427023,
+      "loss": 0.1363,
+      "step": 58000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00036145829692414304,
+      "loss": 0.1096,
+      "step": 58200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0003613988618850114,
+      "loss": 0.1188,
+      "step": 58400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00036133922940343035,
+      "loss": 0.1083,
+      "step": 58600
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0003612793995461759,
+      "loss": 0.1197,
+      "step": 58800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00036121937238024515,
+      "loss": 0.1279,
+      "step": 59000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00036115914797285614,
+      "loss": 0.1189,
+      "step": 59200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00036109872639144766,
+      "loss": 0.1356,
+      "step": 59400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00036103810770367947,
+      "loss": 0.1222,
+      "step": 59600
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00036097729197743193,
+      "loss": 0.1201,
+      "step": 59800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000360916279280806,
+      "loss": 0.1156,
+      "step": 60000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003608550696821234,
+      "loss": 0.1168,
+      "step": 60200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00036079366324992627,
+      "loss": 0.1174,
+      "step": 60400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00036073236855829993,
+      "loss": 0.1202,
+      "step": 60600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00036067056964888823,
+      "loss": 0.1073,
+      "step": 60800
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00036060857411256375,
+      "loss": 0.1288,
+      "step": 61000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003605463820187487,
+      "loss": 0.1174,
+      "step": 61200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003604839934370854,
+      "loss": 0.1136,
+      "step": 61400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0003604214084374361,
+      "loss": 0.112,
+      "step": 61600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00036035894148492016,
+      "loss": 0.1091,
+      "step": 61800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003602959648409783,
+      "loss": 0.1329,
+      "step": 62000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00036023279198960366,
+      "loss": 0.1262,
+      "step": 62200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003601694230015367,
+      "loss": 0.1231,
+      "step": 62400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00036010585794773764,
+      "loss": 0.1107,
+      "step": 62600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003600420968993861,
+      "loss": 0.1299,
+      "step": 62800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0003599781399278814,
+      "loss": 0.1161,
+      "step": 63000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00035991398710484205,
+      "loss": 0.115,
+      "step": 63200
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00035984963850210587,
+      "loss": 0.1223,
+      "step": 63400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003597850941917301,
+      "loss": 0.1154,
+      "step": 63600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00035972035424599094,
+      "loss": 0.1167,
+      "step": 63800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003596554187373838,
+      "loss": 0.1129,
+      "step": 64000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000359590287738623,
+      "loss": 0.1255,
+      "step": 64200
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0003595249613226417,
+      "loss": 0.1212,
+      "step": 64400
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003594594395625922,
+      "loss": 0.1202,
+      "step": 64600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003593937225318451,
+      "loss": 0.1145,
+      "step": 64800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00035932781030399,
+      "loss": 0.1076,
+      "step": 65000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003592617029528349,
+      "loss": 0.1167,
+      "step": 65200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00035919540055240643,
+      "loss": 0.1182,
+      "step": 65400
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00035912890317694947,
+      "loss": 0.1249,
+      "step": 65600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00035906221090092747,
+      "loss": 0.1267,
+      "step": 65800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0003589953237990218,
+      "loss": 0.1111,
+      "step": 66000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003589282419461323,
+      "loss": 0.1155,
+      "step": 66200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003588613022841523,
+      "loss": 0.1156,
+      "step": 66400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003587938321276814,
+      "loss": 0.1381,
+      "step": 66600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003587261674458557,
+      "loss": 0.1141,
+      "step": 66800
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00035865830831444557,
+      "loss": 0.1039,
+      "step": 67000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003585902548094393,
+      "loss": 0.1329,
+      "step": 67200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0003585220070070426,
+      "loss": 0.1255,
+      "step": 67400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00035845356498367904,
+      "loss": 0.12,
+      "step": 67600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00035838492881598946,
+      "loss": 0.1142,
+      "step": 67800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00035831609858083215,
+      "loss": 0.1199,
+      "step": 68000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003582470743552828,
+      "loss": 0.1206,
+      "step": 68200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00035817820278955816,
+      "loss": 0.1298,
+      "step": 68400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0003581087917843051,
+      "loss": 0.1344,
+      "step": 68600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00035803918702080075,
+      "loss": 0.127,
+      "step": 68800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00035796938857698805,
+      "loss": 0.1171,
+      "step": 69000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00035789939653102667,
+      "loss": 0.1142,
+      "step": 69200
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003578292109612934,
+      "loss": 0.119,
+      "step": 69400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003577588319463815,
+      "loss": 0.1232,
+      "step": 69600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003576882595651009,
+      "loss": 0.1206,
+      "step": 69800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003576174938964781,
+      "loss": 0.1093,
+      "step": 70000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00035754653501975596,
+      "loss": 0.1352,
+      "step": 70200
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0003574753830143938,
+      "loss": 0.1207,
+      "step": 70400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00035740403796006713,
+      "loss": 0.1077,
+      "step": 70600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00035733249993666767,
+      "loss": 0.1159,
+      "step": 70800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00035726112815854356,
+      "loss": 0.1196,
+      "step": 71000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003571892054013809,
+      "loss": 0.1162,
+      "step": 71200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0003571170899157135,
+      "loss": 0.1439,
+      "step": 71400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00035704478178229574,
+      "loss": 0.1113,
+      "step": 71600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00035697228108209786,
+      "loss": 0.1195,
+      "step": 71800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00035689958789630546,
+      "loss": 0.1121,
+      "step": 72000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003568267023063201,
+      "loss": 0.1024,
+      "step": 72200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003567536243937585,
+      "loss": 0.1103,
+      "step": 72400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00035668035424045273,
+      "loss": 0.1306,
+      "step": 72600
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003566068919284503,
+      "loss": 0.1145,
+      "step": 72800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003565332375400137,
+      "loss": 0.1157,
+      "step": 73000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0003564593911576208,
+      "loss": 0.119,
+      "step": 73200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00035638535286396404,
+      "loss": 0.1117,
+      "step": 73400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003563114943695969,
+      "loss": 0.117,
+      "step": 73600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00035623707346086925,
+      "loss": 0.108,
+      "step": 73800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00035616246088982775,
+      "loss": 0.106,
+      "step": 74000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003560876567400229,
+      "loss": 0.1316,
+      "step": 74200
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003560126610952201,
+      "loss": 0.1121,
+      "step": 74400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0003559374740393988,
+      "loss": 0.124,
+      "step": 74600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000355862095656753,
+      "loss": 0.1163,
+      "step": 74800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003557865260316909,
+      "loss": 0.1191,
+      "step": 75000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003557107652488348,
+      "loss": 0.1189,
+      "step": 75200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00035563481339302114,
+      "loss": 0.1118,
+      "step": 75400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00035555867054930034,
+      "loss": 0.1242,
+      "step": 75600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0003554823368029366,
+      "loss": 0.1064,
+      "step": 75800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00035540581223940794,
+      "loss": 0.1078,
+      "step": 76000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000355329096944406,
+      "loss": 0.1145,
+      "step": 76200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00035525219100383614,
+      "loss": 0.122,
+      "step": 76400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00035517548046019095,
+      "loss": 0.1132,
+      "step": 76600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003550981944392052,
+      "loss": 0.1183,
+      "step": 76800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00035502071803121456,
+      "loss": 0.1086,
+      "step": 77000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003549430513229765,
+      "loss": 0.11,
+      "step": 77200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003548651944014617,
+      "loss": 0.1144,
+      "step": 77400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003547871473538539,
+      "loss": 0.1114,
+      "step": 77600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0003547089102675495,
+      "loss": 0.1131,
+      "step": 77800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000354630483230158,
+      "loss": 0.1178,
+      "step": 78000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003545518663295014,
+      "loss": 0.1184,
+      "step": 78200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003544730596536143,
+      "loss": 0.1131,
+      "step": 78400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000354394063290744,
+      "loss": 0.1103,
+      "step": 78600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00035431487732935006,
+      "loss": 0.1099,
+      "step": 78800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00035423589920671934,
+      "loss": 0.124,
+      "step": 79000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0003541563352613894,
+      "loss": 0.1216,
+      "step": 79200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00035407658198374195,
+      "loss": 0.1153,
+      "step": 79400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00035399663946308424,
+      "loss": 0.107,
+      "step": 79600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00035391650778893543,
+      "loss": 0.1105,
+      "step": 79800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00035383618705102626,
+      "loss": 0.1361,
+      "step": 80000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00035375567733929953,
+      "loss": 0.1221,
+      "step": 80200
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003536749787439093,
+      "loss": 0.1085,
+      "step": 80400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0003535940913552215,
+      "loss": 0.1115,
+      "step": 80600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00035351301526381307,
+      "loss": 0.1107,
+      "step": 80800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00035343175056047263,
+      "loss": 0.1211,
+      "step": 81000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003533502973361997,
+      "loss": 0.117,
+      "step": 81200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00035326865568220515,
+      "loss": 0.1154,
+      "step": 81400
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0003531868256899108,
+      "loss": 0.1079,
+      "step": 81600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00035310480745094907,
+      "loss": 0.1115,
+      "step": 81800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00035302260105716366,
+      "loss": 0.1247,
+      "step": 82000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00035294061904054505,
+      "loss": 0.119,
+      "step": 82200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00035285803755310787,
+      "loss": 0.1219,
+      "step": 82400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00035277526818717824,
+      "loss": 0.1113,
+      "step": 82600
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00035269231103544056,
+      "loss": 0.1112,
+      "step": 82800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00035260958238174535,
+      "loss": 0.1099,
+      "step": 83000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00035252625087505373,
+      "loss": 0.1519,
+      "step": 83200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00035244273186140206,
+      "loss": 0.1178,
+      "step": 83400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0003523590254343144,
+      "loss": 0.1095,
+      "step": 83600
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00035227513168752453,
+      "loss": 0.1161,
+      "step": 83800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003521910507149763,
+      "loss": 0.1091,
+      "step": 84000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00035210678261082275,
+      "loss": 0.1176,
+      "step": 84200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00035202232746942697,
+      "loss": 0.1197,
+      "step": 84400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003519376853853612,
+      "loss": 0.1163,
+      "step": 84600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003518528564534069,
+      "loss": 0.1152,
+      "step": 84800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003517678407685551,
+      "loss": 0.1086,
+      "step": 85000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0003516826384260057,
+      "loss": 0.1023,
+      "step": 85200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00035159724952116777,
+      "loss": 0.1111,
+      "step": 85400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003515116741496592,
+      "loss": 0.1195,
+      "step": 85600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003514259124073067,
+      "loss": 0.1031,
+      "step": 85800
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00035133996439014564,
+      "loss": 0.1232,
+      "step": 86000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00035125383019442015,
+      "loss": 0.1214,
+      "step": 86200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003511675099165826,
+      "loss": 0.1129,
+      "step": 86400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0003510810036532938,
+      "loss": 0.1312,
+      "step": 86600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00035099431150142305,
+      "loss": 0.1087,
+      "step": 86800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00035090743355804756,
+      "loss": 0.1117,
+      "step": 87000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003508203699204526,
+      "loss": 0.1073,
+      "step": 87200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003507331206861315,
+      "loss": 0.1101,
+      "step": 87400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000350646123587719,
+      "loss": 0.1253,
+      "step": 87600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0003505585043800183,
+      "loss": 0.1035,
+      "step": 87800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00035047069986882655,
+      "loss": 0.1128,
+      "step": 88000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00035038271015246694,
+      "loss": 0.1128,
+      "step": 88200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003502945353294695,
+      "loss": 0.1097,
+      "step": 88400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035020617549857193,
+      "loss": 0.1153,
+      "step": 88600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035011763075871896,
+      "loss": 0.1261,
+      "step": 88800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00035002890120906235,
+      "loss": 0.1111,
+      "step": 89000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00034993998694896083,
+      "loss": 0.121,
+      "step": 89200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00034985088807798,
+      "loss": 0.1327,
+      "step": 89400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0003497616046958922,
+      "loss": 0.1166,
+      "step": 89600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00034967213690267624,
+      "loss": 0.1173,
+      "step": 89800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003495824847985177,
+      "loss": 0.1138,
+      "step": 90000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003494926484838084,
+      "loss": 0.1134,
+      "step": 90200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034940262805914626,
+      "loss": 0.1032,
+      "step": 90400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003493124236253357,
+      "loss": 0.1209,
+      "step": 90600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003492224876824008,
+      "loss": 0.1307,
+      "step": 90800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0003491319164523126,
+      "loss": 0.1144,
+      "step": 91000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00034904116151621703,
+      "loss": 0.1174,
+      "step": 91200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00034895067812499057,
+      "loss": 0.1099,
+      "step": 91400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00034885955699922493,
+      "loss": 0.1104,
+      "step": 91600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00034876825247243814,
+      "loss": 0.1115,
+      "step": 91800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003486767646468723,
+      "loss": 0.1149,
+      "step": 92000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0003485850936249749,
+      "loss": 0.1115,
+      "step": 92200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00034849323950939855,
+      "loss": 0.124,
+      "step": 92400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.00034840120240300075,
+      "loss": 0.1309,
+      "step": 92600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003483089824088442,
+      "loss": 0.1079,
+      "step": 92800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000348216579630196,
+      "loss": 0.1093,
+      "step": 93000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00034812399417052833,
+      "loss": 0.1159,
+      "step": 93200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00034803122613351773,
+      "loss": 0.1172,
+      "step": 93400
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00034793827562304517,
+      "loss": 0.1127,
+      "step": 93600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.000347845142743196,
+      "loss": 0.1152,
+      "step": 93800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0003477518275982598,
+      "loss": 0.1123,
+      "step": 94000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.00034765833029273027,
+      "loss": 0.1321,
+      "step": 94200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00034756465093130503,
+      "loss": 0.1116,
+      "step": 94400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003474707896188856,
+      "loss": 0.1063,
+      "step": 94600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00034737721712853624,
+      "loss": 0.1339,
+      "step": 94800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003472829931380883,
+      "loss": 0.1165,
+      "step": 95000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003471885875120447,
+      "loss": 0.1187,
+      "step": 95200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003470940003561202,
+      "loss": 0.1201,
+      "step": 95400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0003469992317762327,
+      "loss": 0.1115,
+      "step": 95600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003469042818785036,
+      "loss": 0.1031,
+      "step": 95800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000346809150769257,
+      "loss": 0.1212,
+      "step": 96000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00034671383855502014,
+      "loss": 0.1148,
+      "step": 96200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.000346618345342523,
+      "loss": 0.1273,
+      "step": 96400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003465226712386982,
+      "loss": 0.1204,
+      "step": 96600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00034642681635068095,
+      "loss": 0.1101,
+      "step": 96800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0003463307807858089,
+      "loss": 0.1171,
+      "step": 97000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00034623456465162207,
+      "loss": 0.1197,
+      "step": 97200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00034613816805586253,
+      "loss": 0.1153,
+      "step": 97400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003460415911064746,
+      "loss": 0.1155,
+      "step": 97600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00034594483391160453,
+      "loss": 0.1134,
+      "step": 97800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003458478965796003,
+      "loss": 0.1058,
+      "step": 98000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00034575077921901166,
+      "loss": 0.1172,
+      "step": 98200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0003456534819385901,
+      "loss": 0.1182,
+      "step": 98400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.00034555600484728827,
+      "loss": 0.1244,
+      "step": 98600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003454583480542605,
+      "loss": 0.1117,
+      "step": 98800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.000345360511668862,
+      "loss": 0.1088,
+      "step": 99000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003452624958006496,
+      "loss": 0.1136,
+      "step": 99200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003451643005593805,
+      "loss": 0.1148,
+      "step": 99400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003450659260550133,
+      "loss": 0.105,
+      "step": 99600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.00034496786561145394,
+      "loss": 0.1195,
+      "step": 99800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0003448691338065063,
+      "loss": 0.1182,
+      "step": 100000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003447702230689861,
+      "loss": 0.1163,
+      "step": 100200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.000344671133509653,
+      "loss": 0.1245,
+      "step": 100400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003445718652394668,
+      "loss": 0.1046,
+      "step": 100600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003444724183695874,
+      "loss": 0.1182,
+      "step": 100800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003443727930113748,
+      "loss": 0.1207,
+      "step": 101000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00034427298927638874,
+      "loss": 0.1079,
+      "step": 101200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003441730072763888,
+      "loss": 0.104,
+      "step": 101400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0003440728471233341,
+      "loss": 0.1102,
+      "step": 101600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.000343973011063044,
+      "loss": 0.1252,
+      "step": 101800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003438724958299181,
+      "loss": 0.1109,
+      "step": 102000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00034377230668762864,
+      "loss": 0.1208,
+      "step": 102200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00034367143682240804,
+      "loss": 0.1033,
+      "step": 102400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003435703893657879,
+      "loss": 0.1112,
+      "step": 102600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0003434691644309205,
+      "loss": 0.1153,
+      "step": 102800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00034336776213115676,
+      "loss": 0.1053,
+      "step": 103000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003432661825800463,
+      "loss": 0.1188,
+      "step": 103200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00034316442589133725,
+      "loss": 0.1069,
+      "step": 103400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.000343062492178976,
+      "loss": 0.1195,
+      "step": 103600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00034296038155710726,
+      "loss": 0.1099,
+      "step": 103800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003428580941400738,
+      "loss": 0.1185,
+      "step": 104000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003427556300424164,
+      "loss": 0.1247,
+      "step": 104200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0003426529893788736,
+      "loss": 0.111,
+      "step": 104400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00034255017226438175,
+      "loss": 0.1043,
+      "step": 104600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00034244717881407483,
+      "loss": 0.1183,
+      "step": 104800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00034234452542979544,
+      "loss": 0.1165,
+      "step": 105000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00034224118053428676,
+      "loss": 0.1178,
+      "step": 105200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0003421376596489696,
+      "loss": 0.1283,
+      "step": 105400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00034203396288976624,
+      "loss": 0.1091,
+      "step": 105600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00034193009037279546,
+      "loss": 0.1101,
+      "step": 105800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.00034182604221437305,
+      "loss": 0.1147,
+      "step": 106000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003417218185310113,
+      "loss": 0.1052,
+      "step": 106200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00034161741943941935,
+      "loss": 0.1089,
+      "step": 106400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003415128450565025,
+      "loss": 0.1223,
+      "step": 106600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00034140809549936245,
+      "loss": 0.1067,
+      "step": 106800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00034130317088529706,
+      "loss": 0.1182,
+      "step": 107000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0003411980713318002,
+      "loss": 0.1026,
+      "step": 107200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00034109279695656154,
+      "loss": 0.114,
+      "step": 107400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00034098734787746675,
+      "loss": 0.1046,
+      "step": 107600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00034088172421259687,
+      "loss": 0.114,
+      "step": 107800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00034077592608022865,
+      "loss": 0.1079,
+      "step": 108000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00034066995359883405,
+      "loss": 0.1134,
+      "step": 108200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00034056380688708037,
+      "loss": 0.1188,
+      "step": 108400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00034045748606383004,
+      "loss": 0.1123,
+      "step": 108600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.00034035099124814027,
+      "loss": 0.109,
+      "step": 108800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0003402443225592633,
+      "loss": 0.1105,
+      "step": 109000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00034013748011664594,
+      "loss": 0.117,
+      "step": 109200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00034003099955202997,
+      "loss": 0.1196,
+      "step": 109400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003399238108283236,
+      "loss": 0.112,
+      "step": 109600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00033981644870978374,
+      "loss": 0.1267,
+      "step": 109800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.000339709451424421,
+      "loss": 0.1266,
+      "step": 110000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003396017437425494,
+      "loss": 0.1283,
+      "step": 110200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0003394938630264929,
+      "loss": 0.1104,
+      "step": 110400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003393858093970554,
+      "loss": 0.1183,
+      "step": 110600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003392775829752349,
+      "loss": 0.1186,
+      "step": 110800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00033916918388222233,
+      "loss": 0.1141,
+      "step": 111000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003390606122394025,
+      "loss": 0.114,
+      "step": 111200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.000338951868168353,
+      "loss": 0.1064,
+      "step": 111400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00033884295179084477,
+      "loss": 0.1095,
+      "step": 111600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00033873386322884173,
+      "loss": 0.1064,
+      "step": 111800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0003386246026045003,
+      "loss": 0.1207,
+      "step": 112000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00033851517004016994,
+      "loss": 0.1147,
+      "step": 112200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003384055656583925,
+      "loss": 0.1238,
+      "step": 112400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003382957895819021,
+      "loss": 0.1208,
+      "step": 112600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00033818584193362537,
+      "loss": 0.1013,
+      "step": 112800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00033807572283668094,
+      "loss": 0.1144,
+      "step": 113000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0003379654324143793,
+      "loss": 0.1062,
+      "step": 113200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00033785497079022296,
+      "loss": 0.1134,
+      "step": 113400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00033774433808790615,
+      "loss": 0.1087,
+      "step": 113600
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003376335344313145,
+      "loss": 0.1135,
+      "step": 113800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003375225599445254,
+      "loss": 0.1042,
+      "step": 114000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003374114147518071,
+      "loss": 0.1081,
+      "step": 114200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003373006559806056,
+      "loss": 0.1206,
+      "step": 114400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.00033718917060157275,
+      "loss": 0.1112,
+      "step": 114600
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003370775148899379,
+      "loss": 0.1075,
+      "step": 114800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0003369656889707324,
+      "loss": 0.1091,
+      "step": 115000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003368536929691781,
+      "loss": 0.1104,
+      "step": 115200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003367415270106874,
+      "loss": 0.1055,
+      "step": 115400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00033662919122086295,
+      "loss": 0.11,
+      "step": 115600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00033651668572549757,
+      "loss": 0.1178,
+      "step": 115800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003364040106505742,
+      "loss": 0.1063,
+      "step": 116000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00033629116612226564,
+      "loss": 0.1232,
+      "step": 116200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0003361781522669344,
+      "loss": 0.1143,
+      "step": 116400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003360655355470882,
+      "loss": 0.1233,
+      "step": 116600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003359521842626105,
+      "loss": 0.1191,
+      "step": 116800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00033583866403069976,
+      "loss": 0.1114,
+      "step": 117000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003357249749784751,
+      "loss": 0.1075,
+      "step": 117200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003356111172332447,
+      "loss": 0.1084,
+      "step": 117400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003354970909225057,
+      "loss": 0.1037,
+      "step": 117600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0003353828961739441,
+      "loss": 0.1187,
+      "step": 117800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00033526853311543423,
+      "loss": 0.1078,
+      "step": 118000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00033515400187503924,
+      "loss": 0.1083,
+      "step": 118200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00033503930258101024,
+      "loss": 0.1163,
+      "step": 118400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003349244353617869,
+      "loss": 0.101,
+      "step": 118600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00033480997593825657,
+      "loss": 0.1162,
+      "step": 118800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00033469477409273286,
+      "loss": 0.1088,
+      "step": 119000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00033457940470781543,
+      "loss": 0.1168,
+      "step": 119200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0003344638679126941,
+      "loss": 0.1154,
+      "step": 119400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003343481638367461,
+      "loss": 0.1118,
+      "step": 119600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003342322926095362,
+      "loss": 0.1446,
+      "step": 119800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003341162543608161,
+      "loss": 0.1079,
+      "step": 120000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003340000492205248,
+      "loss": 0.1049,
+      "step": 120200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0003338836773187878,
+      "loss": 0.1123,
+      "step": 120400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00033376713878591777,
+      "loss": 0.1151,
+      "step": 120600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.00033365043375241366,
+      "loss": 0.0998,
+      "step": 120800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.000333533562348961,
+      "loss": 0.1189,
+      "step": 121000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003334165247064315,
+      "loss": 0.118,
+      "step": 121200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003332993209558833,
+      "loss": 0.1189,
+      "step": 121400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003331819512285602,
+      "loss": 0.113,
+      "step": 121600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003330650037460781,
+      "loss": 0.1162,
+      "step": 121800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00033294730328792147,
+      "loss": 0.1105,
+      "step": 122000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0003328294372471769,
+      "loss": 0.1041,
+      "step": 122200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00033271140575583016,
+      "loss": 0.1064,
+      "step": 122400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.000332593208946052,
+      "loss": 0.1035,
+      "step": 122600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003324748469501984,
+      "loss": 0.1061,
+      "step": 122800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00033235631990081034,
+      "loss": 0.1199,
+      "step": 123000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00033223762793061363,
+      "loss": 0.1074,
+      "step": 123200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00033211877117251865,
+      "loss": 0.1195,
+      "step": 123400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00033199974975962047,
+      "loss": 0.1132,
+      "step": 123600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0003318811601638971,
+      "loss": 0.1206,
+      "step": 123800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003317618106630228,
+      "loss": 0.1195,
+      "step": 124000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003316428948842102,
+      "loss": 0.1057,
+      "step": 124200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.000331523217827277,
+      "loss": 0.1067,
+      "step": 124400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00033140337678243725,
+      "loss": 0.1185,
+      "step": 124600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0003312833718838879,
+      "loss": 0.1089,
+      "step": 124800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00033116320326600986,
+      "loss": 0.104,
+      "step": 125000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.000331042871063367,
+      "loss": 0.1254,
+      "step": 125200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.00033092237541070647,
+      "loss": 0.1085,
+      "step": 125400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003308017164429586,
+      "loss": 0.1132,
+      "step": 125600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00033068089429523636,
+      "loss": 0.0999,
+      "step": 125800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00033055990910283567,
+      "loss": 0.106,
+      "step": 126000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00033043876100123486,
+      "loss": 0.1066,
+      "step": 126200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0003303174501260947,
+      "loss": 0.1109,
+      "step": 126400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00033019597661325845,
+      "loss": 0.1097,
+      "step": 126600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00033007434059875116,
+      "loss": 0.1142,
+      "step": 126800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00032995254221878013,
+      "loss": 0.1156,
+      "step": 127000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00032983058160973426,
+      "loss": 0.1143,
+      "step": 127200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00032970845890818425,
+      "loss": 0.1193,
+      "step": 127400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00032958617425088226,
+      "loss": 0.1052,
+      "step": 127600
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00032946372777476184,
+      "loss": 0.1078,
+      "step": 127800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003293411196169377,
+      "loss": 0.1017,
+      "step": 128000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00032921834991470565,
+      "loss": 0.1029,
+      "step": 128200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0003290960338623602,
+      "loss": 0.0962,
+      "step": 128400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00032897294228992666,
+      "loss": 0.1078,
+      "step": 128600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00032884968958536763,
+      "loss": 0.1129,
+      "step": 128800
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00032872627588670074,
+      "loss": 0.1107,
+      "step": 129000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00032860270133212367,
+      "loss": 0.1232,
+      "step": 129200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0003284789660600145,
+      "loss": 0.1185,
+      "step": 129400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.00032835507020893096,
+      "loss": 0.111,
+      "step": 129600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.000328231013917611,
+      "loss": 0.1082,
+      "step": 129800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00032810741880645453,
+      "loss": 0.1128,
+      "step": 130000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.000327983042852058,
+      "loss": 0.1224,
+      "step": 130200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0003278585068740186,
+      "loss": 0.1127,
+      "step": 130400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00032773381101179096,
+      "loss": 0.1122,
+      "step": 130600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00032760895540500854,
+      "loss": 0.103,
+      "step": 130800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.000327483940193484,
+      "loss": 0.1129,
+      "step": 131000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00032735876551720825,
+      "loss": 0.1118,
+      "step": 131200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00032723343151635126,
+      "loss": 0.1141,
+      "step": 131400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00032710793833126114,
+      "loss": 0.1043,
+      "step": 131600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003269822861024642,
+      "loss": 0.1223,
+      "step": 131800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003268564749706651,
+      "loss": 0.1129,
+      "step": 132000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0003267305050767462,
+      "loss": 0.1159,
+      "step": 132200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00032660437656176775,
+      "loss": 0.1026,
+      "step": 132400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00032647808956696767,
+      "loss": 0.1107,
+      "step": 132600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00032635164423376107,
+      "loss": 0.1001,
+      "step": 132800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003262250407037408,
+      "loss": 0.1046,
+      "step": 133000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003260982791186765,
+      "loss": 0.1198,
+      "step": 133200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.000325971359620515,
+      "loss": 0.1114,
+      "step": 133400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003258442823513799,
+      "loss": 0.1076,
+      "step": 133600
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00032571704745357134,
+      "loss": 0.1073,
+      "step": 133800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00032558965506956623,
+      "loss": 0.1153,
+      "step": 134000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0003254621053420176,
+      "loss": 0.111,
+      "step": 134200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00032533503733919646,
+      "loss": 0.1122,
+      "step": 134400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003252071741381574,
+      "loss": 0.1042,
+      "step": 134600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003250791540218744,
+      "loss": 0.1103,
+      "step": 134800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003249509771337035,
+      "loss": 0.11,
+      "step": 135000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003248226436171763,
+      "loss": 0.1291,
+      "step": 135200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003246941536159999,
+      "loss": 0.1095,
+      "step": 135400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003245655072740564,
+      "loss": 0.1084,
+      "step": 135600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0003244367047354031,
+      "loss": 0.0981,
+      "step": 135800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00032430839132516974,
+      "loss": 0.1288,
+      "step": 136000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00032417927760514917,
+      "loss": 0.1036,
+      "step": 136200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003240500081209162,
+      "loss": 0.1092,
+      "step": 136400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00032392058301722597,
+      "loss": 0.1103,
+      "step": 136600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.00032379100243900774,
+      "loss": 0.1238,
+      "step": 136800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.000323661266531365,
+      "loss": 0.1125,
+      "step": 137000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0003235313754395751,
+      "loss": 0.1181,
+      "step": 137200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003234013293090893,
+      "loss": 0.1107,
+      "step": 137400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00032327112828553224,
+      "loss": 0.1189,
+      "step": 137600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00032314077251470224,
+      "loss": 0.1177,
+      "step": 137800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00032301026214257077,
+      "loss": 0.1147,
+      "step": 138000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00032287959731528245,
+      "loss": 0.1098,
+      "step": 138200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00032274877817915483,
+      "loss": 0.1114,
+      "step": 138400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0003226178048806784,
+      "loss": 0.1136,
+      "step": 138600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.000322486677566516,
+      "loss": 0.1111,
+      "step": 138800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003223553963835033,
+      "loss": 0.1106,
+      "step": 139000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003222239614786479,
+      "loss": 0.1227,
+      "step": 139200
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003220923729991298,
+      "loss": 0.1115,
+      "step": 139400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00032196063109230086,
+      "loss": 0.119,
+      "step": 139600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003218287359056848,
+      "loss": 0.111,
+      "step": 139800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00032169668758697683,
+      "loss": 0.1146,
+      "step": 140000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0003215651476708621,
+      "loss": 0.0971,
+      "step": 140200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003214327942955547,
+      "loss": 0.1157,
+      "step": 140400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003213002882315282,
+      "loss": 0.1192,
+      "step": 140600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00032116762962716206,
+      "loss": 0.1189,
+      "step": 140800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003210348186310064,
+      "loss": 0.113,
+      "step": 141000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003209018553917821,
+      "loss": 0.1198,
+      "step": 141200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.00032076874005838044,
+      "loss": 0.1082,
+      "step": 141400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0003206354727798631,
+      "loss": 0.1188,
+      "step": 141600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00032050205370546186,
+      "loss": 0.112,
+      "step": 141800
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003203684829845785,
+      "loss": 0.1139,
+      "step": 142000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003202347607667846,
+      "loss": 0.1087,
+      "step": 142200
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00032010222668614745,
+      "loss": 0.1351,
+      "step": 142400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00031996820343515567,
+      "loss": 0.1126,
+      "step": 142600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00031983402913548365,
+      "loss": 0.1261,
+      "step": 142800
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003196997039373788,
+      "loss": 0.1123,
+      "step": 143000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0003195652279912577,
+      "loss": 0.1252,
+      "step": 143200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00031943060144770555,
+      "loss": 0.1031,
+      "step": 143400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00031929582445747623,
+      "loss": 0.1078,
+      "step": 143600
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003191608971714921,
+      "loss": 0.1157,
+      "step": 143800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00031902581974084374,
+      "loss": 0.107,
+      "step": 144000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00031889059231679,
+      "loss": 0.1033,
+      "step": 144200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00031875521505075774,
+      "loss": 0.1127,
+      "step": 144400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0003186196880943413,
+      "loss": 0.119,
+      "step": 144600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00031848401159930296,
+      "loss": 0.1187,
+      "step": 144800
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00031834818571757227,
+      "loss": 0.1131,
+      "step": 145000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003182122106012462,
+      "loss": 0.1141,
+      "step": 145200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00031807608640258873,
+      "loss": 0.1158,
+      "step": 145400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00031793981327403085,
+      "loss": 0.112,
+      "step": 145600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003178033913681702,
+      "loss": 0.1089,
+      "step": 145800
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003176668208377711,
+      "loss": 0.1117,
+      "step": 146000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0003175301018357644,
+      "loss": 0.1141,
+      "step": 146200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.000317393234515247,
+      "loss": 0.1145,
+      "step": 146400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003172562190294821,
+      "loss": 0.1055,
+      "step": 146600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003171197417173119,
+      "loss": 0.1074,
+      "step": 146800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003169824311004134,
+      "loss": 0.1166,
+      "step": 147000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003168449727782822,
+      "loss": 0.1075,
+      "step": 147200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.00031670736690484315,
+      "loss": 0.0993,
+      "step": 147400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0003165696136341865,
+      "loss": 0.1093,
+      "step": 147600
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003164317131205673,
+      "loss": 0.109,
+      "step": 147800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00031629366551840567,
+      "loss": 0.1129,
+      "step": 148000
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00031615547098228636,
+      "loss": 0.1125,
+      "step": 148200
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003160171296669587,
+      "loss": 0.1057,
+      "step": 148400
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.9420512498988455,
+      "eval_auc": 0.8522129916967822,
+      "eval_f1": 0.35228549223252686,
+      "eval_loss": 0.4365999102592468,
+      "eval_mcc": 0.39725936154292724,
+      "eval_precision": 0.2294216317399737,
+      "eval_recall": 0.7584797572261781,
+      "eval_runtime": 9338.8173,
+      "eval_samples_per_second": 23.624,
+      "eval_steps_per_second": 3.937,
+      "step": 148440
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003158786417273362,
+      "loss": 0.1028,
+      "step": 148600
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003157407008546266,
+      "loss": 0.096,
+      "step": 148800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0003156019208629957,
+      "loss": 0.1202,
+      "step": 149000
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0003154629947120181,
+      "loss": 0.1025,
+      "step": 149200
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0003153239225572624,
+      "loss": 0.1131,
+      "step": 149400
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0003151847045544605,
+      "loss": 0.1144,
+      "step": 149600
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0003150453408595079,
+      "loss": 0.1147,
+      "step": 149800
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.000314905831628463,
+      "loss": 0.1024,
+      "step": 150000
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.00031476617701754745,
+      "loss": 0.1044,
+      "step": 150200
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0003146263771831454,
+      "loss": 0.097,
+      "step": 150400
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0003144864322818037,
+      "loss": 0.1143,
+      "step": 150600
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0003143463424702319,
+      "loss": 0.1225,
+      "step": 150800
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0003142061079053015,
+      "loss": 0.1045,
+      "step": 151000
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00031406572874404616,
+      "loss": 0.1082,
+      "step": 151200
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0003139252051436616,
+      "loss": 0.1185,
+      "step": 151400
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00031378453726150514,
+      "loss": 0.1067,
+      "step": 151600
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00031364372525509573,
+      "loss": 0.1033,
+      "step": 151800
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00031350347441983444,
+      "loss": 0.11,
+      "step": 152000
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0003133623753567721,
+      "loss": 0.0992,
+      "step": 152200
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0003132211326421909,
+      "loss": 0.1083,
+      "step": 152400
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0003130797464342533,
+      "loss": 0.1068,
+      "step": 152600
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00031293821689128273,
+      "loss": 0.1233,
+      "step": 152800
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.000312796544171763,
+      "loss": 0.109,
+      "step": 153000
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0003126547284343383,
+      "loss": 0.1109,
+      "step": 153200
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0003125127698378129,
+      "loss": 0.107,
+      "step": 153400
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.000312370668541151,
+      "loss": 0.1118,
+      "step": 153600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0003122284247034767,
+      "loss": 0.1025,
+      "step": 153800
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00031208603848407383,
+      "loss": 0.1085,
+      "step": 154000
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00031194422303810766,
+      "loss": 0.1027,
+      "step": 154200
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.000311801553243652,
+      "loss": 0.1262,
+      "step": 154400
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0003116587415454755,
+      "loss": 0.1146,
+      "step": 154600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.00031151578810349756,
+      "loss": 0.1027,
+      "step": 154800
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0003113726930777965,
+      "loss": 0.1051,
+      "step": 155000
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0003112294566286092,
+      "loss": 0.1045,
+      "step": 155200
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0003110860789163309,
+      "loss": 0.1087,
+      "step": 155400
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0003109425601015149,
+      "loss": 0.1286,
+      "step": 155600
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00031079890034487265,
+      "loss": 0.112,
+      "step": 155800
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0003106558191598876,
+      "loss": 0.1005,
+      "step": 156000
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0003105118787050569,
+      "loss": 0.1044,
+      "step": 156200
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00031036779779067405,
+      "loss": 0.1101,
+      "step": 156400
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00031022357657807976,
+      "loss": 0.1105,
+      "step": 156600
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00031007921522877214,
+      "loss": 0.1006,
+      "step": 156800
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00030993471390440585,
+      "loss": 0.1214,
+      "step": 157000
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0003097900727667927,
+      "loss": 0.1065,
+      "step": 157200
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00030964529197790067,
+      "loss": 0.112,
+      "step": 157400
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00030950037169985437,
+      "loss": 0.1212,
+      "step": 157600
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00030935531209493457,
+      "loss": 0.1087,
+      "step": 157800
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00030921011332557806,
+      "loss": 0.1061,
+      "step": 158000
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0003090647755543773,
+      "loss": 0.109,
+      "step": 158200
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00030891929894408065,
+      "loss": 0.1149,
+      "step": 158400
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00030877368365759167,
+      "loss": 0.1195,
+      "step": 158600
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00030862792985796966,
+      "loss": 0.117,
+      "step": 158800
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0003084820377084285,
+      "loss": 0.1091,
+      "step": 159000
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00030833600737233726,
+      "loss": 0.108,
+      "step": 159200
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0003081898390132197,
+      "loss": 0.1114,
+      "step": 159400
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00030804353279475426,
+      "loss": 0.1096,
+      "step": 159600
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0003078970888807736,
+      "loss": 0.1054,
+      "step": 159800
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0003077505074352645,
+      "loss": 0.1122,
+      "step": 160000
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.000307603788622368,
+      "loss": 0.1065,
+      "step": 160200
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00030745693260637885,
+      "loss": 0.1131,
+      "step": 160400
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0003073106748576299,
+      "loss": 0.1154,
+      "step": 160600
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.00030716354561291433,
+      "loss": 0.1215,
+      "step": 160800
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0003070162796580872,
+      "loss": 0.0988,
+      "step": 161000
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0003068688771580559,
+      "loss": 0.113,
+      "step": 161200
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00030672133827788073,
+      "loss": 0.121,
+      "step": 161400
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00030657366318277474,
+      "loss": 0.0969,
+      "step": 161600
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00030642659143197647,
+      "loss": 0.1109,
+      "step": 161800
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0003062786450822661,
+      "loss": 0.1125,
+      "step": 162000
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0003061305630133498,
+      "loss": 0.1032,
+      "step": 162200
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.00030598234539104877,
+      "loss": 0.1153,
+      "step": 162400
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00030583399238133616,
+      "loss": 0.1126,
+      "step": 162600
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00030568550415033653,
+      "loss": 0.1225,
+      "step": 162800
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00030553688086432617,
+      "loss": 0.1078,
+      "step": 163000
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00030538812268973226,
+      "loss": 0.1095,
+      "step": 163200
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00030523922979313326,
+      "loss": 0.1162,
+      "step": 163400
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00030509020234125854,
+      "loss": 0.1042,
+      "step": 163600
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0003049410405009878,
+      "loss": 0.116,
+      "step": 163800
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0003047917444393516,
+      "loss": 0.1032,
+      "step": 164000
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0003046423143235306,
+      "loss": 0.1167,
+      "step": 164200
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00030449275032085564,
+      "loss": 0.112,
+      "step": 164400
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00030434305259880747,
+      "loss": 0.1068,
+      "step": 164600
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00030419322132501666,
+      "loss": 0.1163,
+      "step": 164800
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00030404325666726306,
+      "loss": 0.1113,
+      "step": 165000
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00030389315879347626,
+      "loss": 0.1142,
+      "step": 165200
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0003037429278717348,
+      "loss": 0.1199,
+      "step": 165400
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0003035925640702662,
+      "loss": 0.1094,
+      "step": 165600
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00030344206755744674,
+      "loss": 0.1238,
+      "step": 165800
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0003032921919765011,
+      "loss": 0.1096,
+      "step": 166000
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0003031414312081544,
+      "loss": 0.1103,
+      "step": 166200
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00030299053823363247,
+      "loss": 0.1067,
+      "step": 166400
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00030283951322190416,
+      "loss": 0.1044,
+      "step": 166600
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00030268835634208633,
+      "loss": 0.1238,
+      "step": 166800
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00030253706776344337,
+      "loss": 0.1184,
+      "step": 167000
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0003023864050828265,
+      "loss": 0.1143,
+      "step": 167200
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0003022348542712936,
+      "loss": 0.1058,
+      "step": 167400
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0003020831722687641,
+      "loss": 0.099,
+      "step": 167600
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00030193135924509056,
+      "loss": 0.1164,
+      "step": 167800
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.000301779415370272,
+      "loss": 0.1132,
+      "step": 168000
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0003016273408144543,
+      "loss": 0.1027,
+      "step": 168200
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.00030147513574792944,
+      "loss": 0.122,
+      "step": 168400
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0003013228003411356,
+      "loss": 0.1127,
+      "step": 168600
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.000301170334764657,
+      "loss": 0.112,
+      "step": 168800
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0003010177391892235,
+      "loss": 0.1098,
+      "step": 169000
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00030086501378571057,
+      "loss": 0.112,
+      "step": 169200
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0003007121587251391,
+      "loss": 0.1037,
+      "step": 169400
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0003005591741786752,
+      "loss": 0.1179,
+      "step": 169600
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0003004060603176298,
+      "loss": 0.1041,
+      "step": 169800
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0003002528173134589,
+      "loss": 0.1142,
+      "step": 170000
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0003000994453377629,
+      "loss": 0.1118,
+      "step": 170200
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00029994594456228665,
+      "loss": 0.109,
+      "step": 170400
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00029979231515891937,
+      "loss": 0.1139,
+      "step": 170600
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002996385572996941,
+      "loss": 0.1119,
+      "step": 170800
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.000299484671156788,
+      "loss": 0.1058,
+      "step": 171000
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002993314272921847,
+      "loss": 0.1129,
+      "step": 171200
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00029917728573828744,
+      "loss": 0.1008,
+      "step": 171400
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002990237880813783,
+      "loss": 0.1019,
+      "step": 171600
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00029886939180346847,
+      "loss": 0.111,
+      "step": 171800
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000298714868103184,
+      "loss": 0.1101,
+      "step": 172000
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00029856021715355947,
+      "loss": 0.1147,
+      "step": 172200
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00029840543912777203,
+      "loss": 0.1142,
+      "step": 172400
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000298250534199141,
+      "loss": 0.1117,
+      "step": 172600
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002980955025411278,
+      "loss": 0.1095,
+      "step": 172800
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00029794034432733597,
+      "loss": 0.1034,
+      "step": 173000
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00029778505973151055,
+      "loss": 0.1097,
+      "step": 173200
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00029762964892753816,
+      "loss": 0.108,
+      "step": 173400
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002974741120894467,
+      "loss": 0.1098,
+      "step": 173600
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.000297319228017684,
+      "loss": 0.1179,
+      "step": 173800
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00029716344026199713,
+      "loss": 0.1175,
+      "step": 174000
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00029700752699424865,
+      "loss": 0.1041,
+      "step": 174200
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00029685148838902903,
+      "loss": 0.1099,
+      "step": 174400
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00029669532462106926,
+      "loss": 0.1064,
+      "step": 174600
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00029653903586524053,
+      "loss": 0.1046,
+      "step": 174800
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00029638262229655384,
+      "loss": 0.109,
+      "step": 175000
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002962260840901601,
+      "loss": 0.1278,
+      "step": 175200
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002960694214213497,
+      "loss": 0.1119,
+      "step": 175400
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002959126344655524,
+      "loss": 0.108,
+      "step": 175600
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002957557233983373,
+      "loss": 0.101,
+      "step": 175800
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002955986883954122,
+      "loss": 0.1151,
+      "step": 176000
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00029544152963262386,
+      "loss": 0.1182,
+      "step": 176200
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002952842472859576,
+      "loss": 0.0995,
+      "step": 176400
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002951268415315372,
+      "loss": 0.1064,
+      "step": 176600
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002949693125456244,
+      "loss": 0.1202,
+      "step": 176800
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002948116605046191,
+      "loss": 0.1103,
+      "step": 177000
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002946538855850592,
+      "loss": 0.1041,
+      "step": 177200
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029449598796361975,
+      "loss": 0.1136,
+      "step": 177400
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002943379678171135,
+      "loss": 0.1018,
+      "step": 177600
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029418061633901157,
+      "loss": 0.1199,
+      "step": 177800
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029402235228377286,
+      "loss": 0.1152,
+      "step": 178000
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002938639662338413,
+      "loss": 0.121,
+      "step": 178200
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00029370545836657664,
+      "loss": 0.1021,
+      "step": 178400
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002935468288594748,
+      "loss": 0.1046,
+      "step": 178600
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002933880778901681,
+      "loss": 0.1073,
+      "step": 178800
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002932292056364248,
+      "loss": 0.1162,
+      "step": 179000
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002930702122761489,
+      "loss": 0.1048,
+      "step": 179200
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002929110979873801,
+      "loss": 0.107,
+      "step": 179400
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00029275186294829356,
+      "loss": 0.1152,
+      "step": 179600
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002925925073371995,
+      "loss": 0.1175,
+      "step": 179800
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002924338290117502,
+      "loss": 0.1251,
+      "step": 180000
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002922742333927425,
+      "loss": 0.109,
+      "step": 180200
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00029211451773657356,
+      "loss": 0.1099,
+      "step": 180400
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002919546822220921,
+      "loss": 0.124,
+      "step": 180600
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00029179472702828063,
+      "loss": 0.1024,
+      "step": 180800
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.000291634652334256,
+      "loss": 0.1051,
+      "step": 181000
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00029147445831926867,
+      "loss": 0.1062,
+      "step": 181200
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002913141451627029,
+      "loss": 0.1134,
+      "step": 181400
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002911537130440763,
+      "loss": 0.1158,
+      "step": 181600
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002909931621430395,
+      "loss": 0.1146,
+      "step": 181800
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002908324926393765,
+      "loss": 0.1031,
+      "step": 182000
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002906717047130038,
+      "loss": 0.113,
+      "step": 182200
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.00029051079854397065,
+      "loss": 0.1105,
+      "step": 182400
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002903497743124587,
+      "loss": 0.1074,
+      "step": 182600
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002901886321987817,
+      "loss": 0.1044,
+      "step": 182800
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002900273723833857,
+      "loss": 0.104,
+      "step": 183000
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002898659950468481,
+      "loss": 0.1385,
+      "step": 183200
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00028970450036987824,
+      "loss": 0.1143,
+      "step": 183400
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002895428885333168,
+      "loss": 0.1072,
+      "step": 183600
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00028938196865289624,
+      "loss": 0.1125,
+      "step": 183800
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.000289220123623735,
+      "loss": 0.1072,
+      "step": 184000
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00028905816197738384,
+      "loss": 0.1123,
+      "step": 184200
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00028889608389520645,
+      "loss": 0.1132,
+      "step": 184400
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002887338895586967,
+      "loss": 0.1261,
+      "step": 184600
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002885715791494787,
+      "loss": 0.1122,
+      "step": 184800
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00028840915284930656,
+      "loss": 0.1035,
+      "step": 185000
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00028824661084006416,
+      "loss": 0.1153,
+      "step": 185200
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002880839533037649,
+      "loss": 0.1025,
+      "step": 185400
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00028792118042255164,
+      "loss": 0.1182,
+      "step": 185600
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002877582923786962,
+      "loss": 0.1138,
+      "step": 185800
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002875952893545997,
+      "loss": 0.0991,
+      "step": 186000
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00028743217153279167,
+      "loss": 0.1069,
+      "step": 186200
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00028726893909593044,
+      "loss": 0.1092,
+      "step": 186400
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002871055922268025,
+      "loss": 0.1178,
+      "step": 186600
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002869421311083226,
+      "loss": 0.1065,
+      "step": 186800
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00028677855592353335,
+      "loss": 0.0977,
+      "step": 187000
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002866148668556052,
+      "loss": 0.1074,
+      "step": 187200
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00028645106408783596,
+      "loss": 0.1101,
+      "step": 187400
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002862871478036509,
+      "loss": 0.0997,
+      "step": 187600
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002861231181866024,
+      "loss": 0.1043,
+      "step": 187800
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00028595897542036964,
+      "loss": 0.1088,
+      "step": 188000
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00028579554124811386,
+      "loss": 0.109,
+      "step": 188200
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00028563117329850646,
+      "loss": 0.1188,
+      "step": 188400
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00028546669275059144,
+      "loss": 0.1289,
+      "step": 188600
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00028530209978855283,
+      "loss": 0.1117,
+      "step": 188800
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002851373945967009,
+      "loss": 0.1109,
+      "step": 189000
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002849725773594712,
+      "loss": 0.1077,
+      "step": 189200
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00028480764826142516,
+      "loss": 0.1115,
+      "step": 189400
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002846426074872491,
+      "loss": 0.1147,
+      "step": 189600
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00028447745522175455,
+      "loss": 0.1163,
+      "step": 189800
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.000284312191649878,
+      "loss": 0.1165,
+      "step": 190000
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00028414681695668046,
+      "loss": 0.1177,
+      "step": 190200
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00028398133132734733,
+      "loss": 0.0974,
+      "step": 190400
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002838165632042751,
+      "loss": 0.1099,
+      "step": 190600
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00028365085681108935,
+      "loss": 0.1195,
+      "step": 190800
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00028348504003714085,
+      "loss": 0.1002,
+      "step": 191000
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002833191130681101,
+      "loss": 0.1065,
+      "step": 191200
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002831530760898009,
+      "loss": 0.1089,
+      "step": 191400
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002829869292881404,
+      "loss": 0.1148,
+      "step": 191600
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002828215044037883,
+      "loss": 0.1049,
+      "step": 191800
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00028265513906049043,
+      "loss": 0.1026,
+      "step": 192000
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002824886644514278,
+      "loss": 0.1162,
+      "step": 192200
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002823220807630174,
+      "loss": 0.1066,
+      "step": 192400
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002821553881817987,
+      "loss": 0.1169,
+      "step": 192600
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00028198858689443274,
+      "loss": 0.1052,
+      "step": 192800
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00028182167708770255,
+      "loss": 0.1017,
+      "step": 193000
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00028165465894851256,
+      "loss": 0.1143,
+      "step": 193200
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002814875326638886,
+      "loss": 0.1133,
+      "step": 193400
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002813202984209775,
+      "loss": 0.1013,
+      "step": 193600
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00028115295640704705,
+      "loss": 0.1057,
+      "step": 193800
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00028098550680948564,
+      "loss": 0.1137,
+      "step": 194000
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002808179498158023,
+      "loss": 0.1064,
+      "step": 194200
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00028065112420100753,
+      "loss": 0.1063,
+      "step": 194400
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002804833535127243,
+      "loss": 0.1058,
+      "step": 194600
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002803154759906268,
+      "loss": 0.1047,
+      "step": 194800
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00028014749182270334,
+      "loss": 0.1154,
+      "step": 195000
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00027997940119706133,
+      "loss": 0.1168,
+      "step": 195200
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00027981120430192757,
+      "loss": 0.1151,
+      "step": 195400
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002796429013256477,
+      "loss": 0.1142,
+      "step": 195600
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002794744924566864,
+      "loss": 0.1051,
+      "step": 195800
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00027930597788362665,
+      "loss": 0.1081,
+      "step": 196000
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00027913735779516995,
+      "loss": 0.1047,
+      "step": 196200
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002789686323801359,
+      "loss": 0.1242,
+      "step": 196400
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002787998018274622,
+      "loss": 0.1033,
+      "step": 196600
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.000278630866326204,
+      "loss": 0.1124,
+      "step": 196800
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002784618260655342,
+      "loss": 0.1067,
+      "step": 197000
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002782926812347429,
+      "loss": 0.1173,
+      "step": 197200
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00027812343202323735,
+      "loss": 0.1127,
+      "step": 197400
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002779540786205417,
+      "loss": 0.1032,
+      "step": 197600
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002777846212162966,
+      "loss": 0.1048,
+      "step": 197800
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.00027761506000025947,
+      "loss": 0.1208,
+      "step": 198000
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00027744539516230373,
+      "loss": 0.1019,
+      "step": 198200
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002772756268924189,
+      "loss": 0.1051,
+      "step": 198400
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00027710660499476826,
+      "loss": 0.0973,
+      "step": 198600
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00027693663094624163,
+      "loss": 0.1141,
+      "step": 198800
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002767665540354966,
+      "loss": 0.1098,
+      "step": 199000
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00027659637445298413,
+      "loss": 0.1083,
+      "step": 199200
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002764269440541954,
+      "loss": 0.1212,
+      "step": 199400
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002762565602109388,
+      "loss": 0.1119,
+      "step": 199600
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002760860742670024,
+      "loss": 0.101,
+      "step": 199800
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002759154864132951,
+      "loss": 0.1161,
+      "step": 200000
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002757447968408403,
+      "loss": 0.1007,
+      "step": 200200
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002755740057407747,
+      "loss": 0.1083,
+      "step": 200400
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00027540311330434927,
+      "loss": 0.1048,
+      "step": 200600
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.000275232119722928,
+      "loss": 0.1031,
+      "step": 200800
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00027506102518798847,
+      "loss": 0.1104,
+      "step": 201000
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00027488982989112105,
+      "loss": 0.1104,
+      "step": 201200
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00027471853402402923,
+      "loss": 0.1052,
+      "step": 201400
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00027454713777852875,
+      "loss": 0.1087,
+      "step": 201600
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00027437649907760433,
+      "loss": 0.115,
+      "step": 201800
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002742049031506784,
+      "loss": 0.1065,
+      "step": 202000
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002740332074205042,
+      "loss": 0.1195,
+      "step": 202200
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00027386141207934555,
+      "loss": 0.1128,
+      "step": 202400
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002736895173195777,
+      "loss": 0.1072,
+      "step": 202600
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002735175233336872,
+      "loss": 0.1028,
+      "step": 202800
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002733454303142717,
+      "loss": 0.1085,
+      "step": 203000
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00027317323845403986,
+      "loss": 0.1047,
+      "step": 203200
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.000273000947945811,
+      "loss": 0.1193,
+      "step": 203400
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002728285589825147,
+      "loss": 0.1041,
+      "step": 203600
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002726560717571911,
+      "loss": 0.1032,
+      "step": 203800
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00027248348646299013,
+      "loss": 0.0963,
+      "step": 204000
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00027231080329317167,
+      "loss": 0.107,
+      "step": 204200
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00027213802244110513,
+      "loss": 0.1048,
+      "step": 204400
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002719651441002693,
+      "loss": 0.1097,
+      "step": 204600
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00027179216846425223,
+      "loss": 0.1019,
+      "step": 204800
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00027161909572675087,
+      "loss": 0.1078,
+      "step": 205000
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00027144592608157085,
+      "loss": 0.1092,
+      "step": 205200
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002712726597226264,
+      "loss": 0.1169,
+      "step": 205400
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00027110016389810513,
+      "loss": 0.1222,
+      "step": 205600
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00027092670517495234,
+      "loss": 0.1128,
+      "step": 205800
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00027075315031945495,
+      "loss": 0.1243,
+      "step": 206000
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00027057949952595886,
+      "loss": 0.108,
+      "step": 206200
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002704057529889168,
+      "loss": 0.1073,
+      "step": 206400
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00027023191090288904,
+      "loss": 0.1142,
+      "step": 206600
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002700579734625427,
+      "loss": 0.1108,
+      "step": 206800
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002698839408626519,
+      "loss": 0.1077,
+      "step": 207000
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.000269709813298097,
+      "loss": 0.1101,
+      "step": 207200
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002695355909638649,
+      "loss": 0.1073,
+      "step": 207400
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002693612740550486,
+      "loss": 0.106,
+      "step": 207600
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00026918686276684704,
+      "loss": 0.1054,
+      "step": 207800
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00026901235729456465,
+      "loss": 0.1073,
+      "step": 208000
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00026883775783361163,
+      "loss": 0.1018,
+      "step": 208200
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002686630645795032,
+      "loss": 0.1091,
+      "step": 208400
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00026848827772785964,
+      "loss": 0.1111,
+      "step": 208600
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002683133974744061,
+      "loss": 0.1087,
+      "step": 208800
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00026813842401497234,
+      "loss": 0.1138,
+      "step": 209000
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00026796335754549237,
+      "loss": 0.1044,
+      "step": 209200
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00026778819826200456,
+      "loss": 0.1088,
+      "step": 209400
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00026761294636065095,
+      "loss": 0.1114,
+      "step": 209600
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00026743760203767745,
+      "loss": 0.1001,
+      "step": 209800
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.00026726216548943345,
+      "loss": 0.1039,
+      "step": 210000
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002670866369123717,
+      "loss": 0.1047,
+      "step": 210200
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00026691101650304773,
+      "loss": 0.1054,
+      "step": 210400
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002667353044581202,
+      "loss": 0.1058,
+      "step": 210600
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00026655950097435025,
+      "loss": 0.1053,
+      "step": 210800
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002663844859488688,
+      "loss": 0.1081,
+      "step": 211000
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002662085006328417,
+      "loss": 0.1077,
+      "step": 211200
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002660324244678835,
+      "loss": 0.1008,
+      "step": 211400
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00026585625765116307,
+      "loss": 0.108,
+      "step": 211600
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00026568088189098604,
+      "loss": 0.1066,
+      "step": 211800
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00026550453481344815,
+      "loss": 0.1103,
+      "step": 212000
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002653280976752753,
+      "loss": 0.1218,
+      "step": 212200
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00026515157067404056,
+      "loss": 0.0986,
+      "step": 212400
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.00026497495400741773,
+      "loss": 0.1041,
+      "step": 212600
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002647982478731809,
+      "loss": 0.1071,
+      "step": 212800
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002646214524692044,
+      "loss": 0.1149,
+      "step": 213000
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00026444456799346256,
+      "loss": 0.1116,
+      "step": 213200
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002642675946440293,
+      "loss": 0.1057,
+      "step": 213400
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002640905326190783,
+      "loss": 0.1118,
+      "step": 213600
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00026391338211688236,
+      "loss": 0.1243,
+      "step": 213800
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00026373614333581336,
+      "loss": 0.102,
+      "step": 214000
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00026355881647434213,
+      "loss": 0.1077,
+      "step": 214200
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002633814017310381,
+      "loss": 0.1235,
+      "step": 214400
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.000263203899304569,
+      "loss": 0.0982,
+      "step": 214600
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.000263026309393701,
+      "loss": 0.1053,
+      "step": 214800
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00026284863219729797,
+      "loss": 0.1,
+      "step": 215000
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002626708679143215,
+      "loss": 0.109,
+      "step": 215200
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00026249301674383113,
+      "loss": 0.1043,
+      "step": 215400
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002623150788849831,
+      "loss": 0.1044,
+      "step": 215600
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002621379448735828,
+      "loss": 0.1312,
+      "step": 215800
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00026195983466683015,
+      "loss": 0.1166,
+      "step": 216000
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00026178252956408584,
+      "loss": 0.1259,
+      "step": 216200
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002616042478032307,
+      "loss": 0.1008,
+      "step": 216400
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002614258803492553,
+      "loss": 0.1057,
+      "step": 216600
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002612474274018945,
+      "loss": 0.1101,
+      "step": 216800
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002610688891609786,
+      "loss": 0.1088,
+      "step": 217000
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00026089026582643357,
+      "loss": 0.1161,
+      "step": 217200
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002607115575982805,
+      "loss": 0.107,
+      "step": 217400
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00026053276467663576,
+      "loss": 0.1081,
+      "step": 217600
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002603538872617103,
+      "loss": 0.1039,
+      "step": 217800
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00026017492555381,
+      "loss": 0.1083,
+      "step": 218000
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002599958797533348,
+      "loss": 0.1198,
+      "step": 218200
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002598167500607792,
+      "loss": 0.1207,
+      "step": 218400
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002596375366767311,
+      "loss": 0.1063,
+      "step": 218600
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002594591364935983,
+      "loss": 0.1176,
+      "step": 218800
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00025927975674465545,
+      "loss": 0.1,
+      "step": 219000
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00025910029390554173,
+      "loss": 0.1113,
+      "step": 219200
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002589207481772184,
+      "loss": 0.1117,
+      "step": 219400
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002587411197607395,
+      "loss": 0.1175,
+      "step": 219600
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002585614088572518,
+      "loss": 0.1,
+      "step": 219800
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00025838161566799425,
+      "loss": 0.1118,
+      "step": 220000
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002582017403942981,
+      "loss": 0.1157,
+      "step": 220200
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002580217832375865,
+      "loss": 0.1029,
+      "step": 220400
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002578417443993741,
+      "loss": 0.1049,
+      "step": 220600
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002576616240812673,
+      "loss": 0.0959,
+      "step": 220800
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002574814224849635,
+      "loss": 0.1038,
+      "step": 221000
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002573011398122514,
+      "loss": 0.1141,
+      "step": 221200
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00025712077626501007,
+      "loss": 0.1005,
+      "step": 221400
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002569403320452095,
+      "loss": 0.1049,
+      "step": 221600
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.00025675980735491,
+      "loss": 0.1105,
+      "step": 221800
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00025657920239626175,
+      "loss": 0.1046,
+      "step": 222000
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002563985173715049,
+      "loss": 0.1049,
+      "step": 222200
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00025621775248296953,
+      "loss": 0.1063,
+      "step": 222400
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002560369079330747,
+      "loss": 0.1137,
+      "step": 222600
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00025585598392432904,
+      "loss": 0.1066,
+      "step": 222800
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002556758858724697,
+      "loss": 0.1117,
+      "step": 223000
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002554957095545989,
+      "loss": 0.1273,
+      "step": 223200
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00025531454917274415,
+      "loss": 0.1036,
+      "step": 223400
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.000255133310140931,
+      "loss": 0.116,
+      "step": 223600
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00025495199266210964,
+      "loss": 0.0995,
+      "step": 223800
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002547705969393181,
+      "loss": 0.1075,
+      "step": 224000
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00025458912317568216,
+      "loss": 0.108,
+      "step": 224200
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002544075715744149,
+      "loss": 0.1033,
+      "step": 224400
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002542259423388165,
+      "loss": 0.1068,
+      "step": 224600
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00025404423567227416,
+      "loss": 0.117,
+      "step": 224800
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002538624517782618,
+      "loss": 0.1027,
+      "step": 225000
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00025368059086033976,
+      "loss": 0.1048,
+      "step": 225200
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002534995630015991,
+      "loss": 0.1265,
+      "step": 225400
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00025331754902945935,
+      "loss": 0.1161,
+      "step": 225600
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.00025313545864358826,
+      "loss": 0.0981,
+      "step": 225800
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002529532920478896,
+      "loss": 0.11,
+      "step": 226000
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.000252771049446352,
+      "loss": 0.1136,
+      "step": 226200
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002525887310430497,
+      "loss": 0.1101,
+      "step": 226400
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00025240633704214153,
+      "loss": 0.1082,
+      "step": 226600
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00025222386764787107,
+      "loss": 0.1127,
+      "step": 226800
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00025204132306456625,
+      "loss": 0.1061,
+      "step": 227000
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002518587034966394,
+      "loss": 0.1125,
+      "step": 227200
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00025167600914858645,
+      "loss": 0.1111,
+      "step": 227400
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002514932402249875,
+      "loss": 0.1103,
+      "step": 227600
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00025131039693050575,
+      "loss": 0.1084,
+      "step": 227800
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002511274794698881,
+      "loss": 0.1109,
+      "step": 228000
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.000250944488047964,
+      "loss": 0.1005,
+      "step": 228200
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002507614228696462,
+      "loss": 0.105,
+      "step": 228400
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002505782841399298,
+      "loss": 0.1013,
+      "step": 228600
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00025039507206389233,
+      "loss": 0.1049,
+      "step": 228800
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002502117868466933,
+      "loss": 0.0967,
+      "step": 229000
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002500284286935744,
+      "loss": 0.0953,
+      "step": 229200
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002498449978098587,
+      "loss": 0.0965,
+      "step": 229400
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00024966149440095104,
+      "loss": 0.1028,
+      "step": 229600
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002494779186723371,
+      "loss": 0.1118,
+      "step": 229800
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002492951892478414,
+      "loss": 0.1011,
+      "step": 230000
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002491114698556271,
+      "loss": 0.1075,
+      "step": 230200
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002489276787596204,
+      "loss": 0.0981,
+      "step": 230400
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002487438161656293,
+      "loss": 0.108,
+      "step": 230600
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002485598822795418,
+      "loss": 0.1076,
+      "step": 230800
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0002483758773073258,
+      "loss": 0.1204,
+      "step": 231000
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0002481918014550289,
+      "loss": 0.1068,
+      "step": 231200
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00024800765492877794,
+      "loss": 0.1066,
+      "step": 231400
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.000247823437934779,
+      "loss": 0.1072,
+      "step": 231600
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0002476391506793169,
+      "loss": 0.1118,
+      "step": 231800
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.00024745479336875533,
+      "loss": 0.1003,
+      "step": 232000
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0002472703662095364,
+      "loss": 0.1214,
+      "step": 232200
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002470858694081803,
+      "loss": 0.1063,
+      "step": 232400
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00024690222617484883,
+      "loss": 0.1117,
+      "step": 232600
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002467185144021892,
+      "loss": 0.1019,
+      "step": 232800
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002465338106035194,
+      "loss": 0.1018,
+      "step": 233000
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002463490379875025,
+      "loss": 0.1261,
+      "step": 233200
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002461641967610455,
+      "loss": 0.103,
+      "step": 233400
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002459792871311324,
+      "loss": 0.1017,
+      "step": 233600
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00024579430930482364,
+      "loss": 0.0997,
+      "step": 233800
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00024560926348925633,
+      "loss": 0.1041,
+      "step": 234000
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0002454241498916433,
+      "loss": 0.1037,
+      "step": 234200
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0002452389687192737,
+      "loss": 0.1114,
+      "step": 234400
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00024505372017951196,
+      "loss": 0.1064,
+      "step": 234600
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00024486840447979826,
+      "loss": 0.1128,
+      "step": 234800
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00024468302182764775,
+      "loss": 0.107,
+      "step": 235000
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0002444975724306508,
+      "loss": 0.1024,
+      "step": 235200
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002443120564964722,
+      "loss": 0.1176,
+      "step": 235400
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00024412647423285164,
+      "loss": 0.0995,
+      "step": 235600
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00024394082584760268,
+      "loss": 0.1047,
+      "step": 235800
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00024375511154861332,
+      "loss": 0.1159,
+      "step": 236000
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00024356933154384496,
+      "loss": 0.1087,
+      "step": 236200
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00024338348604133295,
+      "loss": 0.1051,
+      "step": 236400
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002431975752491857,
+      "loss": 0.1116,
+      "step": 236600
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00024301252941649857,
+      "loss": 0.1048,
+      "step": 236800
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00024282648899354658,
+      "loss": 0.0991,
+      "step": 237000
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00024264038390468088,
+      "loss": 0.1106,
+      "step": 237200
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00024245421435830072,
+      "loss": 0.0987,
+      "step": 237400
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00024226798056287748,
+      "loss": 0.1062,
+      "step": 237600
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00024208168272695445,
+      "loss": 0.106,
+      "step": 237800
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00024189532105914677,
+      "loss": 0.1128,
+      "step": 238000
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00024170889576814088,
+      "loss": 0.1094,
+      "step": 238200
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00024152240706269457,
+      "loss": 0.121,
+      "step": 238400
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00024133585515163664,
+      "loss": 0.1045,
+      "step": 238600
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00024114924024386668,
+      "loss": 0.1022,
+      "step": 238800
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00024096349609267103,
+      "loss": 0.1023,
+      "step": 239000
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00024077675613083096,
+      "loss": 0.1093,
+      "step": 239200
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.000240589953798354,
+      "loss": 0.1054,
+      "step": 239400
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00024040308930442034,
+      "loss": 0.1011,
+      "step": 239600
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0002402161628582794,
+      "loss": 0.0995,
+      "step": 239800
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00024002917466925026,
+      "loss": 0.1062,
+      "step": 240000
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0002398421249467209,
+      "loss": 0.1071,
+      "step": 240200
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00023965594960757828,
+      "loss": 0.1081,
+      "step": 240400
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00023946877775153958,
+      "loss": 0.1126,
+      "step": 240600
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00023928154498952938,
+      "loss": 0.1077,
+      "step": 240800
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00023909425153120962,
+      "loss": 0.1136,
+      "step": 241000
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0002389068975863103,
+      "loss": 0.1101,
+      "step": 241200
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00023871948336462903,
+      "loss": 0.0935,
+      "step": 241400
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00023853200907603106,
+      "loss": 0.1068,
+      "step": 241600
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002383444749304487,
+      "loss": 0.1098,
+      "step": 241800
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002381568811378815,
+      "loss": 0.1166,
+      "step": 242000
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002379692279083957,
+      "loss": 0.1049,
+      "step": 242200
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00023778151545212405,
+      "loss": 0.1021,
+      "step": 242400
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.00023759374397926578,
+      "loss": 0.1157,
+      "step": 242600
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00023740591370008605,
+      "loss": 0.0974,
+      "step": 242800
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00023721802482491592,
+      "loss": 0.1002,
+      "step": 243000
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00023703007756415207,
+      "loss": 0.0989,
+      "step": 243200
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0002368420721282566,
+      "loss": 0.1032,
+      "step": 243400
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00023665400872775662,
+      "loss": 0.1085,
+      "step": 243600
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00023646588757324441,
+      "loss": 0.1113,
+      "step": 243800
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00023627770887537663,
+      "loss": 0.1086,
+      "step": 244000
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00023608947284487457,
+      "loss": 0.0993,
+      "step": 244200
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002359021213000272,
+      "loss": 0.1198,
+      "step": 244400
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00023571377152070707,
+      "loss": 0.1126,
+      "step": 244600
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00023552536504024596,
+      "loss": 0.1061,
+      "step": 244800
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00023533690206961993,
+      "loss": 0.1046,
+      "step": 245000
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002351483828198688,
+      "loss": 0.1033,
+      "step": 245200
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00023495980750209484,
+      "loss": 0.1094,
+      "step": 245400
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00023477117632746366,
+      "loss": 0.0984,
+      "step": 245600
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00023458248950720297,
+      "loss": 0.1101,
+      "step": 245800
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00023439374725260314,
+      "loss": 0.108,
+      "step": 246000
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00023420494977501635,
+      "loss": 0.1068,
+      "step": 246200
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00023401704168479406,
+      "loss": 0.103,
+      "step": 246400
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002338281346690118,
+      "loss": 0.1124,
+      "step": 246600
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.00023363917306361173,
+      "loss": 0.103,
+      "step": 246800
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002334501570801918,
+      "loss": 0.1052,
+      "step": 247000
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00023326108693041077,
+      "loss": 0.1122,
+      "step": 247200
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002330719628259882,
+      "loss": 0.1069,
+      "step": 247400
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00023288278497870388,
+      "loss": 0.106,
+      "step": 247600
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002326935536003979,
+      "loss": 0.106,
+      "step": 247800
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002325042689029703,
+      "loss": 0.1027,
+      "step": 248000
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00023231493109838075,
+      "loss": 0.1063,
+      "step": 248200
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00023212648748337304,
+      "loss": 0.1053,
+      "step": 248400
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00023193704436346414,
+      "loss": 0.1191,
+      "step": 248600
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00023174849637970008,
+      "loss": 0.1165,
+      "step": 248800
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00023155894878878252,
+      "loss": 0.0973,
+      "step": 249000
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00023136934914926618,
+      "loss": 0.1023,
+      "step": 249200
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00023117969767346322,
+      "loss": 0.1,
+      "step": 249400
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00023098999457374426,
+      "loss": 0.1107,
+      "step": 249600
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00023080024006253745,
+      "loss": 0.1013,
+      "step": 249800
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00023061043435232856,
+      "loss": 0.0996,
+      "step": 250000
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00023042057765566083,
+      "loss": 0.1205,
+      "step": 250200
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0002302306701851344,
+      "loss": 0.115,
+      "step": 250400
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0002300407121534064,
+      "loss": 0.0969,
+      "step": 250600
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0002298507037731906,
+      "loss": 0.1042,
+      "step": 250800
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.000229660645257257,
+      "loss": 0.112,
+      "step": 251000
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0002294705368184319,
+      "loss": 0.1101,
+      "step": 251200
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00022928037866959736,
+      "loss": 0.1112,
+      "step": 251400
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.00022909017102369127,
+      "loss": 0.0958,
+      "step": 251600
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0002288999140937068,
+      "loss": 0.115,
+      "step": 251800
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00022870960809269237,
+      "loss": 0.0982,
+      "step": 252000
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00022852020512922758,
+      "loss": 0.101,
+      "step": 252200
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00022832980186821158,
+      "loss": 0.0993,
+      "step": 252400
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0002281393501745734,
+      "loss": 0.1095,
+      "step": 252600
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00022794885026157955,
+      "loss": 0.102,
+      "step": 252800
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00022775830234255057,
+      "loss": 0.0989,
+      "step": 253000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00022756770663086073,
+      "loss": 0.104,
+      "step": 253200
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0002273770633399378,
+      "loss": 0.1069,
+      "step": 253400
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00022718637268326274,
+      "loss": 0.1034,
+      "step": 253600
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00022699563487436978,
+      "loss": 0.1017,
+      "step": 253800
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00022680485012684578,
+      "loss": 0.1064,
+      "step": 254000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0002266140186543302,
+      "loss": 0.1003,
+      "step": 254200
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00022642409517577717,
+      "loss": 0.1237,
+      "step": 254400
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00022623317112536182,
+      "loss": 0.1315,
+      "step": 254600
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00022604220099011713,
+      "loss": 0.1103,
+      "step": 254800
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00022585118498389016,
+      "loss": 0.1142,
+      "step": 255000
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00022566012332057937,
+      "loss": 0.1058,
+      "step": 255200
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00022546901621413423,
+      "loss": 0.098,
+      "step": 255400
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00022527786387855522,
+      "loss": 0.1109,
+      "step": 255600
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00022508666652789347,
+      "loss": 0.1164,
+      "step": 255800
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002248954243762504,
+      "loss": 0.1031,
+      "step": 256000
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0002247041376377777,
+      "loss": 0.123,
+      "step": 256200
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00022451280652667694,
+      "loss": 0.1113,
+      "step": 256400
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0002243214312571994,
+      "loss": 0.1076,
+      "step": 256600
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00022413001204364582,
+      "loss": 0.1129,
+      "step": 256800
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00022393854910036614,
+      "loss": 0.0994,
+      "step": 257000
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00022374800028194188,
+      "loss": 0.124,
+      "step": 257200
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00022355645073842626,
+      "loss": 0.0966,
+      "step": 257400
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00022336485810745454,
+      "loss": 0.1031,
+      "step": 257600
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00022317322260357087,
+      "loss": 0.1166,
+      "step": 257800
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00022298154444136742,
+      "loss": 0.1128,
+      "step": 258000
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00022278982383548408,
+      "loss": 0.1087,
+      "step": 258200
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0002225980610006082,
+      "loss": 0.1097,
+      "step": 258400
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00022240625615147454,
+      "loss": 0.104,
+      "step": 258600
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00022221440950286487,
+      "loss": 0.1104,
+      "step": 258800
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.00022202252126960782,
+      "loss": 0.1102,
+      "step": 259000
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00022183059166657837,
+      "loss": 0.1033,
+      "step": 259200
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0002216386209086982,
+      "loss": 0.1034,
+      "step": 259400
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0002214466092109346,
+      "loss": 0.1014,
+      "step": 259600
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00022125455678830122,
+      "loss": 0.1118,
+      "step": 259800
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0002210624638558569,
+      "loss": 0.1032,
+      "step": 260000
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00022087033062870605,
+      "loss": 0.0966,
+      "step": 260200
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00022067815732199807,
+      "loss": 0.107,
+      "step": 260400
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00022048690531558938,
+      "loss": 0.1213,
+      "step": 260600
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00022029465269310516,
+      "loss": 0.1062,
+      "step": 260800
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00022010236063570412,
+      "loss": 0.0971,
+      "step": 261000
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00021991002935871357,
+      "loss": 0.1026,
+      "step": 261200
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0002197176590775048,
+      "loss": 0.1056,
+      "step": 261400
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00021952525000749266,
+      "loss": 0.1076,
+      "step": 261600
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.00021933280236413567,
+      "loss": 0.1034,
+      "step": 261800
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00021914031636293516,
+      "loss": 0.105,
+      "step": 262000
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00021894779221943593,
+      "loss": 0.1006,
+      "step": 262200
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00021875523014922496,
+      "loss": 0.0986,
+      "step": 262400
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0002185626303679321,
+      "loss": 0.1038,
+      "step": 262600
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00021837095637052567,
+      "loss": 0.1153,
+      "step": 262800
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00021817828199998843,
+      "loss": 0.1022,
+      "step": 263000
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00021798557056443188,
+      "loss": 0.1056,
+      "step": 263200
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00021779282227965291,
+      "loss": 0.1112,
+      "step": 263400
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00021760003736148975,
+      "loss": 0.1254,
+      "step": 263600
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0002174072160258217,
+      "loss": 0.0986,
+      "step": 263800
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00021721435848856872,
+      "loss": 0.1134,
+      "step": 264000
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00021702146496569135,
+      "loss": 0.1026,
+      "step": 264200
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00021682853567319053,
+      "loss": 0.1051,
+      "step": 264400
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00021663557082710706,
+      "loss": 0.1032,
+      "step": 264600
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.00021644257064352173,
+      "loss": 0.1088,
+      "step": 264800
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00021624953533855483,
+      "loss": 0.1064,
+      "step": 265000
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00021605646512836595,
+      "loss": 0.1037,
+      "step": 265200
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00021586336022915377,
+      "loss": 0.1051,
+      "step": 265400
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00021567022085715594,
+      "loss": 0.1153,
+      "step": 265600
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00021547704722864853,
+      "loss": 0.1095,
+      "step": 265800
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0002152838395599461,
+      "loss": 0.1001,
+      "step": 266000
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00021509059806740124,
+      "loss": 0.1076,
+      "step": 266200
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0002148973229674045,
+      "loss": 0.1002,
+      "step": 266400
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00021470401447638398,
+      "loss": 0.1111,
+      "step": 266600
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002145106728108052,
+      "loss": 0.1106,
+      "step": 266800
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00021431729818717092,
+      "loss": 0.1043,
+      "step": 267000
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002141248579399327,
+      "loss": 0.1007,
+      "step": 267200
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00021393141821192865,
+      "loss": 0.0938,
+      "step": 267400
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002137379461745144,
+      "loss": 0.112,
+      "step": 267600
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002135444420443386,
+      "loss": 0.097,
+      "step": 267800
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002133509060380859,
+      "loss": 0.0971,
+      "step": 268000
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00021315733837247666,
+      "loss": 0.1002,
+      "step": 268200
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00021296373926426643,
+      "loss": 0.1044,
+      "step": 268400
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002127701089302464,
+      "loss": 0.1069,
+      "step": 268600
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00021257644758724233,
+      "loss": 0.1053,
+      "step": 268800
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00021238275545211504,
+      "loss": 0.0994,
+      "step": 269000
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00021218903274175944,
+      "loss": 0.1059,
+      "step": 269200
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00021199527967310516,
+      "loss": 0.1064,
+      "step": 269400
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00021180246545378291,
+      "loss": 0.1124,
+      "step": 269600
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00021160865246853668,
+      "loss": 0.1018,
+      "step": 269800
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00021141480977489752,
+      "loss": 0.1083,
+      "step": 270000
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00021122093758992914,
+      "loss": 0.1062,
+      "step": 270200
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00021102703613072832,
+      "loss": 0.107,
+      "step": 270400
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00021083310561442455,
+      "loss": 0.1007,
+      "step": 270600
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00021063914625817995,
+      "loss": 0.1065,
+      "step": 270800
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00021044612828992364,
+      "loss": 0.1126,
+      "step": 271000
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00021025211204689966,
+      "loss": 0.1022,
+      "step": 271200
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00021005806761452758,
+      "loss": 0.0992,
+      "step": 271400
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00020986399521009683,
+      "loss": 0.0991,
+      "step": 271600
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00020966989505092856,
+      "loss": 0.1072,
+      "step": 271800
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00020947576735437463,
+      "loss": 0.0948,
+      "step": 272000
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.00020928161233781808,
+      "loss": 0.1053,
+      "step": 272200
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00020908743021867218,
+      "loss": 0.1007,
+      "step": 272400
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00020889322121438094,
+      "loss": 0.0949,
+      "step": 272600
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00020869898554241816,
+      "loss": 0.1083,
+      "step": 272800
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0002085047234202877,
+      "loss": 0.1076,
+      "step": 273000
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0002083114065721905,
+      "loss": 0.1129,
+      "step": 273200
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0002081170923318883,
+      "loss": 0.0989,
+      "step": 273400
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00020792275229301839,
+      "loss": 0.1126,
+      "step": 273600
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0002077283866732015,
+      "loss": 0.1027,
+      "step": 273800
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00020753399569008686,
+      "loss": 0.1086,
+      "step": 274000
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00020733957956135224,
+      "loss": 0.1178,
+      "step": 274200
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00020714513850470342,
+      "loss": 0.1053,
+      "step": 274400
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00020695067273787424,
+      "loss": 0.0955,
+      "step": 274600
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00020675618247862612,
+      "loss": 0.1083,
+      "step": 274800
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0002065616679447479,
+      "loss": 0.1019,
+      "step": 275000
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.00020636712935405562,
+      "loss": 0.1158,
+      "step": 275200
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00020617256692439238,
+      "loss": 0.1004,
+      "step": 275400
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002059789538622786,
+      "loss": 0.1307,
+      "step": 275600
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002057843445247827,
+      "loss": 0.1028,
+      "step": 275800
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00020558971200091428,
+      "loss": 0.1094,
+      "step": 276000
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002053950565086214,
+      "loss": 0.1136,
+      "step": 276200
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00020520037826587812,
+      "loss": 0.104,
+      "step": 276400
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002050056774906837,
+      "loss": 0.108,
+      "step": 276600
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.00020481095440106272,
+      "loss": 0.1083,
+      "step": 276800
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00020461620921506486,
+      "loss": 0.1,
+      "step": 277000
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002044214421507644,
+      "loss": 0.1045,
+      "step": 277200
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00020422665342626015,
+      "loss": 0.1069,
+      "step": 277400
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00020403184325967527,
+      "loss": 0.1211,
+      "step": 277600
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00020383701186915681,
+      "loss": 0.0911,
+      "step": 277800
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00020364215947287566,
+      "loss": 0.1028,
+      "step": 278000
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00020344728628902614,
+      "loss": 0.1103,
+      "step": 278200
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00020325239253582604,
+      "loss": 0.0972,
+      "step": 278400
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0002030574784315159,
+      "loss": 0.1101,
+      "step": 278600
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0002028625441943593,
+      "loss": 0.1062,
+      "step": 278800
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00020266759004264228,
+      "loss": 0.1075,
+      "step": 279000
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00020247261619467312,
+      "loss": 0.0943,
+      "step": 279200
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00020227762286878224,
+      "loss": 0.0966,
+      "step": 279400
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0002020835853937961,
+      "loss": 0.105,
+      "step": 279600
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0002018885538618027,
+      "loss": 0.0999,
+      "step": 279800
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00020169350350591656,
+      "loss": 0.1132,
+      "step": 280000
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00020149940993528039,
+      "loss": 0.105,
+      "step": 280200
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00020130432267826973,
+      "loss": 0.1023,
+      "step": 280400
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0002011092172515844,
+      "loss": 0.0991,
+      "step": 280600
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00020091409387370222,
+      "loss": 0.1192,
+      "step": 280800
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00020071895276312101,
+      "loss": 0.1115,
+      "step": 281000
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00020052379413835846,
+      "loss": 0.1081,
+      "step": 281200
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00020032861821795184,
+      "loss": 0.1058,
+      "step": 281400
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0002001334252204578,
+      "loss": 0.1074,
+      "step": 281600
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00019993821536445222,
+      "loss": 0.1086,
+      "step": 281800
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00019974298886852966,
+      "loss": 0.1091,
+      "step": 282000
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001995477459513035,
+      "loss": 0.1083,
+      "step": 282200
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0001993524868314054,
+      "loss": 0.1123,
+      "step": 282400
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00019915721172748514,
+      "loss": 0.0978,
+      "step": 282600
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0001989628973514115,
+      "loss": 0.1267,
+      "step": 282800
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00019876759101265725,
+      "loss": 0.0962,
+      "step": 283000
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00019857226934484322,
+      "loss": 0.1077,
+      "step": 283200
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00019837693256668934,
+      "loss": 0.1028,
+      "step": 283400
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00019818158089693235,
+      "loss": 0.1053,
+      "step": 283600
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00019798621455432573,
+      "loss": 0.104,
+      "step": 283800
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00019779083375763941,
+      "loss": 0.0913,
+      "step": 284000
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00019759543872565946,
+      "loss": 0.1018,
+      "step": 284200
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0001974000296771879,
+      "loss": 0.1131,
+      "step": 284400
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00019720460683104248,
+      "loss": 0.1039,
+      "step": 284600
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00019700917040605631,
+      "loss": 0.1166,
+      "step": 284800
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00019681469790287367,
+      "loss": 0.1165,
+      "step": 285000
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00019661923504192736,
+      "loss": 0.1031,
+      "step": 285200
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00019642375925763573,
+      "loss": 0.1168,
+      "step": 285400
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.00019622827076889117,
+      "loss": 0.1137,
+      "step": 285600
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00019603276979460036,
+      "loss": 0.102,
+      "step": 285800
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00019583725655368395,
+      "loss": 0.1076,
+      "step": 286000
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00019564173126507638,
+      "loss": 0.1038,
+      "step": 286200
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0001954461941477256,
+      "loss": 0.1135,
+      "step": 286400
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00019525064542059252,
+      "loss": 0.1059,
+      "step": 286600
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00019505508530265155,
+      "loss": 0.1097,
+      "step": 286800
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00019485951401288927,
+      "loss": 0.1102,
+      "step": 287000
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00019466393177030537,
+      "loss": 0.1019,
+      "step": 287200
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00019446833879391132,
+      "loss": 0.095,
+      "step": 287400
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00019427273530273097,
+      "loss": 0.1002,
+      "step": 287600
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00019407712151579958,
+      "loss": 0.1048,
+      "step": 287800
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0001938814976521643,
+      "loss": 0.0957,
+      "step": 288000
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0001936858639308833,
+      "loss": 0.1086,
+      "step": 288200
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00019349022057102595,
+      "loss": 0.1094,
+      "step": 288400
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00019329456779167232,
+      "loss": 0.1095,
+      "step": 288600
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00019309988414433548,
+      "loss": 0.1254,
+      "step": 288800
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00019290421322763307,
+      "loss": 0.1082,
+      "step": 289000
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00019270853354764138,
+      "loss": 0.1007,
+      "step": 289200
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0001925128453234813,
+      "loss": 0.109,
+      "step": 289400
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00019231714877428302,
+      "loss": 0.1117,
+      "step": 289600
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00019212144411918632,
+      "loss": 0.1033,
+      "step": 289800
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00019192573157733988,
+      "loss": 0.1046,
+      "step": 290000
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00019173001136790129,
+      "loss": 0.1005,
+      "step": 290200
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0001915342837100367,
+      "loss": 0.1151,
+      "step": 290400
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00019133854882292064,
+      "loss": 0.0974,
+      "step": 290600
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0001911428069257357,
+      "loss": 0.1024,
+      "step": 290800
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0001909470582376723,
+      "loss": 0.1075,
+      "step": 291000
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00019075130297792843,
+      "loss": 0.1117,
+      "step": 291200
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0001905555413657096,
+      "loss": 0.1029,
+      "step": 291400
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00019035977362022831,
+      "loss": 0.1056,
+      "step": 291600
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00019016399996070392,
+      "loss": 0.1045,
+      "step": 291800
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00018996822060636244,
+      "loss": 0.1058,
+      "step": 292000
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00018977243577643626,
+      "loss": 0.106,
+      "step": 292200
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00018957664569016394,
+      "loss": 0.098,
+      "step": 292400
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00018938085056678978,
+      "loss": 0.1008,
+      "step": 292600
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00018918602963689183,
+      "loss": 0.1052,
+      "step": 292800
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.00018899022511951726,
+      "loss": 0.1062,
+      "step": 293000
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00018879441622171074,
+      "loss": 0.1033,
+      "step": 293200
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0001885986031627376,
+      "loss": 0.0977,
+      "step": 293400
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00018840278616186814,
+      "loss": 0.0991,
+      "step": 293600
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00018820696543837674,
+      "loss": 0.1059,
+      "step": 293800
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00018801114121154224,
+      "loss": 0.1171,
+      "step": 294000
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00018781629284600801,
+      "loss": 0.1001,
+      "step": 294200
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00018762046228511725,
+      "loss": 0.1031,
+      "step": 294400
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00018742462887764543,
+      "loss": 0.1137,
+      "step": 294600
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00018722879284288572,
+      "loss": 0.1026,
+      "step": 294800
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0001870329544001338,
+      "loss": 0.0991,
+      "step": 295000
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00018683711376868837,
+      "loss": 0.1004,
+      "step": 295200
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0001866412711678503,
+      "loss": 0.101,
+      "step": 295400
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00018644542681692293,
+      "loss": 0.0959,
+      "step": 295600
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00018624958093521127,
+      "loss": 0.1299,
+      "step": 295800
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0001860537337420224,
+      "loss": 0.1124,
+      "step": 296000
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00018585788545666447,
+      "loss": 0.0973,
+      "step": 296200
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00018566203629844722,
+      "loss": 0.1098,
+      "step": 296400
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.00018546618648668112,
+      "loss": 0.0972,
+      "step": 296600
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0001852713154926249,
+      "loss": 0.108,
+      "step": 296800
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.9392292210730847,
+      "eval_auc": 0.8528852594008389,
+      "eval_f1": 0.34279097031908445,
+      "eval_loss": 0.45720356702804565,
+      "eval_mcc": 0.3902260284864879,
+      "eval_precision": 0.22106795124992942,
+      "eval_recall": 0.7627978181868509,
+      "eval_runtime": 9308.0675,
+      "eval_samples_per_second": 23.702,
+      "eval_steps_per_second": 3.95,
+      "step": 296880
+    }
+  ],
+  "logging_steps": 200,
+  "max_steps": 593760,
+  "num_train_epochs": 4,
+  "save_steps": 500,
+  "total_flos": 3.568751227922958e+17,
+  "trial_name": null,
+  "trial_params": null
+}