{ "best_metric": null, "best_model_checkpoint": null, "epoch": 4.9999899454035415, "global_step": 124321, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 1e-05, "loss": 10.7043, "step": 10 }, { "epoch": 0.0, "learning_rate": 2e-05, "loss": 10.6777, "step": 20 }, { "epoch": 0.0, "learning_rate": 3e-05, "loss": 10.6776, "step": 30 }, { "epoch": 0.0, "learning_rate": 4e-05, "loss": 10.6287, "step": 40 }, { "epoch": 0.0, "learning_rate": 5e-05, "loss": 10.581, "step": 50 }, { "epoch": 0.0, "learning_rate": 6e-05, "loss": 10.4726, "step": 60 }, { "epoch": 0.0, "learning_rate": 7.000000000000001e-05, "loss": 10.1574, "step": 70 }, { "epoch": 0.0, "learning_rate": 8e-05, "loss": 9.362, "step": 80 }, { "epoch": 0.0, "learning_rate": 8.999999999999999e-05, "loss": 8.4335, "step": 90 }, { "epoch": 0.0, "learning_rate": 0.0001, "loss": 7.9476, "step": 100 }, { "epoch": 0.0, "learning_rate": 0.00011, "loss": 7.3033, "step": 110 }, { "epoch": 0.0, "learning_rate": 0.00012, "loss": 6.9745, "step": 120 }, { "epoch": 0.01, "learning_rate": 0.00013000000000000002, "loss": 7.2853, "step": 130 }, { "epoch": 0.01, "learning_rate": 0.00014000000000000001, "loss": 7.025, "step": 140 }, { "epoch": 0.01, "learning_rate": 0.00015, "loss": 6.8469, "step": 150 }, { "epoch": 0.01, "learning_rate": 0.00016, "loss": 7.204, "step": 160 }, { "epoch": 0.01, "learning_rate": 0.00017, "loss": 6.9907, "step": 170 }, { "epoch": 0.01, "learning_rate": 0.00017999999999999998, "loss": 6.9468, "step": 180 }, { "epoch": 0.01, "learning_rate": 0.00019, "loss": 6.8738, "step": 190 }, { "epoch": 0.01, "learning_rate": 0.0002, "loss": 6.9808, "step": 200 }, { "epoch": 0.01, "learning_rate": 0.00021, "loss": 6.6993, "step": 210 }, { "epoch": 0.01, "learning_rate": 0.00022, "loss": 6.8427, "step": 220 }, { "epoch": 0.01, "learning_rate": 0.00023, "loss": 6.617, "step": 230 }, { "epoch": 0.01, "learning_rate": 0.00024, "loss": 7.0377, "step": 240 }, { "epoch": 0.01, "learning_rate": 0.00025, "loss": 6.8056, "step": 250 }, { "epoch": 0.01, "learning_rate": 0.00026000000000000003, "loss": 6.8114, "step": 260 }, { "epoch": 0.01, "learning_rate": 0.00027, "loss": 6.4411, "step": 270 }, { "epoch": 0.01, "learning_rate": 0.00028000000000000003, "loss": 6.5567, "step": 280 }, { "epoch": 0.01, "learning_rate": 0.00029, "loss": 6.7945, "step": 290 }, { "epoch": 0.01, "learning_rate": 0.0003, "loss": 6.439, "step": 300 }, { "epoch": 0.01, "learning_rate": 0.00031, "loss": 6.2834, "step": 310 }, { "epoch": 0.01, "learning_rate": 0.00032, "loss": 6.6685, "step": 320 }, { "epoch": 0.01, "learning_rate": 0.00033, "loss": 6.4431, "step": 330 }, { "epoch": 0.01, "learning_rate": 0.00034, "loss": 6.5047, "step": 340 }, { "epoch": 0.01, "learning_rate": 0.00035, "loss": 6.4539, "step": 350 }, { "epoch": 0.01, "learning_rate": 0.00035999999999999997, "loss": 6.5991, "step": 360 }, { "epoch": 0.01, "learning_rate": 0.00037, "loss": 6.1802, "step": 370 }, { "epoch": 0.02, "learning_rate": 0.00038, "loss": 6.4964, "step": 380 }, { "epoch": 0.02, "learning_rate": 0.00039000000000000005, "loss": 6.3887, "step": 390 }, { "epoch": 0.02, "learning_rate": 0.0004, "loss": 6.2293, "step": 400 }, { "epoch": 0.02, "learning_rate": 0.00041, "loss": 6.1274, "step": 410 }, { "epoch": 0.02, "learning_rate": 0.00042, "loss": 6.2553, "step": 420 }, { "epoch": 0.02, "learning_rate": 0.00043, "loss": 6.4652, "step": 430 }, { "epoch": 0.02, "learning_rate": 0.00044, "loss": 6.2114, "step": 440 }, { "epoch": 0.02, "learning_rate": 0.00045000000000000004, "loss": 5.9314, "step": 450 }, { "epoch": 0.02, "learning_rate": 0.00046, "loss": 6.1565, "step": 460 }, { "epoch": 0.02, "learning_rate": 0.00047, "loss": 6.0596, "step": 470 }, { "epoch": 0.02, "learning_rate": 0.00048, "loss": 6.0298, "step": 480 }, { "epoch": 0.02, "learning_rate": 0.00049, "loss": 5.9433, "step": 490 }, { "epoch": 0.02, "learning_rate": 0.0005, "loss": 5.9851, "step": 500 }, { "epoch": 0.02, "learning_rate": 0.0004999899351825757, "loss": 5.8544, "step": 510 }, { "epoch": 0.02, "learning_rate": 0.0004999798703651516, "loss": 5.8239, "step": 520 }, { "epoch": 0.02, "learning_rate": 0.0004999698055477273, "loss": 5.9386, "step": 530 }, { "epoch": 0.02, "learning_rate": 0.0004999597407303032, "loss": 5.8861, "step": 540 }, { "epoch": 0.02, "learning_rate": 0.0004999496759128789, "loss": 5.9269, "step": 550 }, { "epoch": 0.02, "learning_rate": 0.0004999396110954548, "loss": 5.7289, "step": 560 }, { "epoch": 0.02, "learning_rate": 0.0004999295462780305, "loss": 5.6925, "step": 570 }, { "epoch": 0.02, "learning_rate": 0.0004999194814606063, "loss": 5.8428, "step": 580 }, { "epoch": 0.02, "learning_rate": 0.0004999094166431821, "loss": 5.8405, "step": 590 }, { "epoch": 0.02, "learning_rate": 0.0004998993518257579, "loss": 5.5196, "step": 600 }, { "epoch": 0.02, "learning_rate": 0.0004998892870083337, "loss": 5.4851, "step": 610 }, { "epoch": 0.02, "learning_rate": 0.0004998792221909095, "loss": 5.6924, "step": 620 }, { "epoch": 0.03, "learning_rate": 0.0004998691573734852, "loss": 5.5769, "step": 630 }, { "epoch": 0.03, "learning_rate": 0.0004998590925560611, "loss": 5.6209, "step": 640 }, { "epoch": 0.03, "learning_rate": 0.0004998490277386368, "loss": 5.344, "step": 650 }, { "epoch": 0.03, "learning_rate": 0.0004998389629212127, "loss": 5.6723, "step": 660 }, { "epoch": 0.03, "learning_rate": 0.0004998288981037884, "loss": 5.5622, "step": 670 }, { "epoch": 0.03, "learning_rate": 0.0004998188332863641, "loss": 5.3601, "step": 680 }, { "epoch": 0.03, "learning_rate": 0.00049980876846894, "loss": 5.4225, "step": 690 }, { "epoch": 0.03, "learning_rate": 0.0004997987036515157, "loss": 5.5094, "step": 700 }, { "epoch": 0.03, "learning_rate": 0.0004997886388340916, "loss": 5.5905, "step": 710 }, { "epoch": 0.03, "learning_rate": 0.0004997785740166674, "loss": 5.4246, "step": 720 }, { "epoch": 0.03, "learning_rate": 0.0004997685091992431, "loss": 5.6965, "step": 730 }, { "epoch": 0.03, "learning_rate": 0.000499758444381819, "loss": 5.5467, "step": 740 }, { "epoch": 0.03, "learning_rate": 0.0004997483795643947, "loss": 5.494, "step": 750 }, { "epoch": 0.03, "learning_rate": 0.0004997383147469705, "loss": 5.2289, "step": 760 }, { "epoch": 0.03, "learning_rate": 0.0004997282499295463, "loss": 5.3773, "step": 770 }, { "epoch": 0.03, "learning_rate": 0.000499718185112122, "loss": 5.2676, "step": 780 }, { "epoch": 0.03, "learning_rate": 0.0004997081202946979, "loss": 5.2778, "step": 790 }, { "epoch": 0.03, "learning_rate": 0.0004996980554772736, "loss": 5.2564, "step": 800 }, { "epoch": 0.03, "learning_rate": 0.0004996879906598495, "loss": 5.2077, "step": 810 }, { "epoch": 0.03, "learning_rate": 0.0004996779258424252, "loss": 4.9609, "step": 820 }, { "epoch": 0.03, "learning_rate": 0.000499667861025001, "loss": 5.2737, "step": 830 }, { "epoch": 0.03, "learning_rate": 0.0004996577962075768, "loss": 5.0288, "step": 840 }, { "epoch": 0.03, "learning_rate": 0.0004996477313901525, "loss": 5.2016, "step": 850 }, { "epoch": 0.03, "learning_rate": 0.0004996376665727284, "loss": 5.1146, "step": 860 }, { "epoch": 0.03, "learning_rate": 0.0004996276017553042, "loss": 4.82, "step": 870 }, { "epoch": 0.04, "learning_rate": 0.00049961753693788, "loss": 5.0896, "step": 880 }, { "epoch": 0.04, "learning_rate": 0.0004996074721204558, "loss": 5.162, "step": 890 }, { "epoch": 0.04, "learning_rate": 0.0004995974073030315, "loss": 5.0342, "step": 900 }, { "epoch": 0.04, "learning_rate": 0.0004995873424856074, "loss": 4.7726, "step": 910 }, { "epoch": 0.04, "learning_rate": 0.0004995772776681831, "loss": 5.0794, "step": 920 }, { "epoch": 0.04, "learning_rate": 0.0004995672128507589, "loss": 4.9815, "step": 930 }, { "epoch": 0.04, "learning_rate": 0.0004995571480333347, "loss": 4.8256, "step": 940 }, { "epoch": 0.04, "learning_rate": 0.0004995470832159104, "loss": 5.1052, "step": 950 }, { "epoch": 0.04, "learning_rate": 0.0004995370183984863, "loss": 5.1086, "step": 960 }, { "epoch": 0.04, "learning_rate": 0.000499526953581062, "loss": 4.9837, "step": 970 }, { "epoch": 0.04, "learning_rate": 0.0004995168887636379, "loss": 4.7818, "step": 980 }, { "epoch": 0.04, "learning_rate": 0.0004995068239462136, "loss": 4.9077, "step": 990 }, { "epoch": 0.04, "learning_rate": 0.0004994967591287894, "loss": 4.6932, "step": 1000 }, { "epoch": 0.04, "learning_rate": 0.0004994866943113652, "loss": 5.0554, "step": 1010 }, { "epoch": 0.04, "learning_rate": 0.000499476629493941, "loss": 4.7956, "step": 1020 }, { "epoch": 0.04, "learning_rate": 0.0004994665646765168, "loss": 4.8894, "step": 1030 }, { "epoch": 0.04, "learning_rate": 0.0004994564998590926, "loss": 4.9363, "step": 1040 }, { "epoch": 0.04, "learning_rate": 0.0004994464350416683, "loss": 4.9249, "step": 1050 }, { "epoch": 0.04, "learning_rate": 0.0004994363702242442, "loss": 4.9123, "step": 1060 }, { "epoch": 0.04, "learning_rate": 0.0004994263054068199, "loss": 4.7132, "step": 1070 }, { "epoch": 0.04, "learning_rate": 0.0004994162405893958, "loss": 4.7995, "step": 1080 }, { "epoch": 0.04, "learning_rate": 0.0004994061757719715, "loss": 4.6651, "step": 1090 }, { "epoch": 0.04, "learning_rate": 0.0004993961109545472, "loss": 4.8011, "step": 1100 }, { "epoch": 0.04, "learning_rate": 0.0004993860461371231, "loss": 4.7827, "step": 1110 }, { "epoch": 0.05, "learning_rate": 0.0004993759813196988, "loss": 5.062, "step": 1120 }, { "epoch": 0.05, "learning_rate": 0.0004993659165022747, "loss": 4.5373, "step": 1130 }, { "epoch": 0.05, "learning_rate": 0.0004993558516848504, "loss": 4.4756, "step": 1140 }, { "epoch": 0.05, "learning_rate": 0.0004993457868674262, "loss": 4.7175, "step": 1150 }, { "epoch": 0.05, "learning_rate": 0.0004993357220500021, "loss": 4.9643, "step": 1160 }, { "epoch": 0.05, "learning_rate": 0.0004993256572325778, "loss": 5.218, "step": 1170 }, { "epoch": 0.05, "learning_rate": 0.0004993155924151536, "loss": 5.7388, "step": 1180 }, { "epoch": 0.05, "learning_rate": 0.0004993055275977294, "loss": 4.8866, "step": 1190 }, { "epoch": 0.05, "learning_rate": 0.0004992954627803051, "loss": 4.7099, "step": 1200 }, { "epoch": 0.05, "learning_rate": 0.000499285397962881, "loss": 4.5195, "step": 1210 }, { "epoch": 0.05, "learning_rate": 0.0004992753331454567, "loss": 4.6756, "step": 1220 }, { "epoch": 0.05, "learning_rate": 0.0004992652683280326, "loss": 4.6491, "step": 1230 }, { "epoch": 0.05, "learning_rate": 0.0004992552035106083, "loss": 4.8041, "step": 1240 }, { "epoch": 0.05, "learning_rate": 0.0004992451386931842, "loss": 4.5687, "step": 1250 }, { "epoch": 0.05, "learning_rate": 0.0004992350738757599, "loss": 4.4752, "step": 1260 }, { "epoch": 0.05, "learning_rate": 0.0004992250090583356, "loss": 4.5014, "step": 1270 }, { "epoch": 0.05, "learning_rate": 0.0004992149442409115, "loss": 4.3925, "step": 1280 }, { "epoch": 0.05, "learning_rate": 0.0004992048794234872, "loss": 4.4677, "step": 1290 }, { "epoch": 0.05, "learning_rate": 0.0004991948146060631, "loss": 4.5665, "step": 1300 }, { "epoch": 0.05, "learning_rate": 0.0004991847497886389, "loss": 4.5154, "step": 1310 }, { "epoch": 0.05, "learning_rate": 0.0004991746849712146, "loss": 4.7437, "step": 1320 }, { "epoch": 0.05, "learning_rate": 0.0004991646201537905, "loss": 4.5744, "step": 1330 }, { "epoch": 0.05, "learning_rate": 0.0004991545553363662, "loss": 4.3066, "step": 1340 }, { "epoch": 0.05, "learning_rate": 0.000499144490518942, "loss": 4.4891, "step": 1350 }, { "epoch": 0.05, "learning_rate": 0.0004991344257015178, "loss": 4.5493, "step": 1360 }, { "epoch": 0.06, "learning_rate": 0.0004991243608840935, "loss": 4.5137, "step": 1370 }, { "epoch": 0.06, "learning_rate": 0.0004991142960666694, "loss": 4.7195, "step": 1380 }, { "epoch": 0.06, "learning_rate": 0.0004991042312492451, "loss": 4.493, "step": 1390 }, { "epoch": 0.06, "learning_rate": 0.000499094166431821, "loss": 4.2165, "step": 1400 }, { "epoch": 0.06, "learning_rate": 0.0004990841016143967, "loss": 4.408, "step": 1410 }, { "epoch": 0.06, "learning_rate": 0.0004990740367969725, "loss": 4.3322, "step": 1420 }, { "epoch": 0.06, "learning_rate": 0.0004990639719795483, "loss": 4.5391, "step": 1430 }, { "epoch": 0.06, "learning_rate": 0.000499053907162124, "loss": 4.3628, "step": 1440 }, { "epoch": 0.06, "learning_rate": 0.0004990438423446999, "loss": 4.0576, "step": 1450 }, { "epoch": 0.06, "learning_rate": 0.0004990337775272757, "loss": 4.4032, "step": 1460 }, { "epoch": 0.06, "learning_rate": 0.0004990237127098514, "loss": 4.3908, "step": 1470 }, { "epoch": 0.06, "learning_rate": 0.0004990136478924273, "loss": 4.2756, "step": 1480 }, { "epoch": 0.06, "learning_rate": 0.000499003583075003, "loss": 4.1813, "step": 1490 }, { "epoch": 0.06, "learning_rate": 0.0004989935182575789, "loss": 4.2954, "step": 1500 }, { "epoch": 0.06, "learning_rate": 0.0004989834534401546, "loss": 4.0622, "step": 1510 }, { "epoch": 0.06, "learning_rate": 0.0004989733886227303, "loss": 4.113, "step": 1520 }, { "epoch": 0.06, "learning_rate": 0.0004989633238053062, "loss": 4.1197, "step": 1530 }, { "epoch": 0.06, "learning_rate": 0.0004989532589878819, "loss": 4.3373, "step": 1540 }, { "epoch": 0.06, "learning_rate": 0.0004989431941704578, "loss": 4.1861, "step": 1550 }, { "epoch": 0.06, "learning_rate": 0.0004989331293530335, "loss": 4.3061, "step": 1560 }, { "epoch": 0.06, "learning_rate": 0.0004989230645356094, "loss": 3.9928, "step": 1570 }, { "epoch": 0.06, "learning_rate": 0.0004989129997181851, "loss": 4.2371, "step": 1580 }, { "epoch": 0.06, "learning_rate": 0.0004989029349007609, "loss": 4.2351, "step": 1590 }, { "epoch": 0.06, "learning_rate": 0.0004988928700833367, "loss": 4.343, "step": 1600 }, { "epoch": 0.06, "learning_rate": 0.0004988828052659125, "loss": 4.0789, "step": 1610 }, { "epoch": 0.07, "learning_rate": 0.0004988727404484883, "loss": 4.2875, "step": 1620 }, { "epoch": 0.07, "learning_rate": 0.0004988626756310641, "loss": 3.9671, "step": 1630 }, { "epoch": 0.07, "learning_rate": 0.0004988526108136398, "loss": 4.2236, "step": 1640 }, { "epoch": 0.07, "learning_rate": 0.0004988425459962157, "loss": 3.9325, "step": 1650 }, { "epoch": 0.07, "learning_rate": 0.0004988324811787914, "loss": 4.1272, "step": 1660 }, { "epoch": 0.07, "learning_rate": 0.0004988224163613673, "loss": 4.2489, "step": 1670 }, { "epoch": 0.07, "learning_rate": 0.000498812351543943, "loss": 4.1515, "step": 1680 }, { "epoch": 0.07, "learning_rate": 0.0004988022867265187, "loss": 4.6586, "step": 1690 }, { "epoch": 0.07, "learning_rate": 0.0004987922219090946, "loss": 4.8059, "step": 1700 }, { "epoch": 0.07, "learning_rate": 0.0004987821570916703, "loss": 4.6982, "step": 1710 }, { "epoch": 0.07, "learning_rate": 0.0004987720922742462, "loss": 4.679, "step": 1720 }, { "epoch": 0.07, "learning_rate": 0.0004987620274568219, "loss": 4.9727, "step": 1730 }, { "epoch": 0.07, "learning_rate": 0.0004987519626393977, "loss": 5.0227, "step": 1740 }, { "epoch": 0.07, "learning_rate": 0.0004987418978219736, "loss": 5.4149, "step": 1750 }, { "epoch": 0.07, "learning_rate": 0.0004987318330045493, "loss": 5.1723, "step": 1760 }, { "epoch": 0.07, "learning_rate": 0.0004987217681871251, "loss": 5.2009, "step": 1770 }, { "epoch": 0.07, "learning_rate": 0.0004987117033697009, "loss": 5.0127, "step": 1780 }, { "epoch": 0.07, "learning_rate": 0.0004987016385522766, "loss": 4.5407, "step": 1790 }, { "epoch": 0.07, "learning_rate": 0.0004986915737348525, "loss": 4.54, "step": 1800 }, { "epoch": 0.07, "learning_rate": 0.0004986815089174282, "loss": 4.5878, "step": 1810 }, { "epoch": 0.07, "learning_rate": 0.0004986714441000041, "loss": 4.6118, "step": 1820 }, { "epoch": 0.07, "learning_rate": 0.0004986613792825798, "loss": 4.2691, "step": 1830 }, { "epoch": 0.07, "learning_rate": 0.0004986513144651557, "loss": 4.2872, "step": 1840 }, { "epoch": 0.07, "learning_rate": 0.0004986412496477314, "loss": 4.1103, "step": 1850 }, { "epoch": 0.07, "learning_rate": 0.0004986311848303071, "loss": 4.0249, "step": 1860 }, { "epoch": 0.08, "learning_rate": 0.000498621120012883, "loss": 4.5064, "step": 1870 }, { "epoch": 0.08, "learning_rate": 0.0004986110551954587, "loss": 4.3161, "step": 1880 }, { "epoch": 0.08, "learning_rate": 0.0004986009903780345, "loss": 4.308, "step": 1890 }, { "epoch": 0.08, "learning_rate": 0.0004985909255606104, "loss": 4.1388, "step": 1900 }, { "epoch": 0.08, "learning_rate": 0.0004985808607431861, "loss": 4.5835, "step": 1910 }, { "epoch": 0.08, "learning_rate": 0.000498570795925762, "loss": 4.0345, "step": 1920 }, { "epoch": 0.08, "learning_rate": 0.0004985607311083377, "loss": 4.1658, "step": 1930 }, { "epoch": 0.08, "learning_rate": 0.0004985506662909134, "loss": 4.1487, "step": 1940 }, { "epoch": 0.08, "learning_rate": 0.0004985406014734893, "loss": 4.3708, "step": 1950 }, { "epoch": 0.08, "learning_rate": 0.000498530536656065, "loss": 4.0366, "step": 1960 }, { "epoch": 0.08, "learning_rate": 0.0004985204718386409, "loss": 3.9569, "step": 1970 }, { "epoch": 0.08, "learning_rate": 0.0004985104070212166, "loss": 3.9947, "step": 1980 }, { "epoch": 0.08, "learning_rate": 0.0004985003422037925, "loss": 4.2372, "step": 1990 }, { "epoch": 0.08, "learning_rate": 0.0004984902773863682, "loss": 4.0002, "step": 2000 }, { "epoch": 0.08, "learning_rate": 0.000498480212568944, "loss": 3.9595, "step": 2010 }, { "epoch": 0.08, "learning_rate": 0.0004984701477515198, "loss": 4.0646, "step": 2020 }, { "epoch": 0.08, "learning_rate": 0.0004984600829340955, "loss": 4.2384, "step": 2030 }, { "epoch": 0.08, "learning_rate": 0.0004984500181166714, "loss": 3.8212, "step": 2040 }, { "epoch": 0.08, "learning_rate": 0.0004984399532992472, "loss": 3.6501, "step": 2050 }, { "epoch": 0.08, "learning_rate": 0.0004984298884818229, "loss": 3.9362, "step": 2060 }, { "epoch": 0.08, "learning_rate": 0.0004984198236643988, "loss": 4.0042, "step": 2070 }, { "epoch": 0.08, "learning_rate": 0.0004984097588469745, "loss": 3.8303, "step": 2080 }, { "epoch": 0.08, "learning_rate": 0.0004983996940295504, "loss": 3.8395, "step": 2090 }, { "epoch": 0.08, "learning_rate": 0.0004983896292121261, "loss": 3.7393, "step": 2100 }, { "epoch": 0.08, "learning_rate": 0.0004983795643947018, "loss": 3.8953, "step": 2110 }, { "epoch": 0.09, "learning_rate": 0.0004983694995772777, "loss": 3.6642, "step": 2120 }, { "epoch": 0.09, "learning_rate": 0.0004983594347598534, "loss": 3.8766, "step": 2130 }, { "epoch": 0.09, "learning_rate": 0.0004983493699424293, "loss": 3.6295, "step": 2140 }, { "epoch": 0.09, "learning_rate": 0.000498339305125005, "loss": 3.8133, "step": 2150 }, { "epoch": 0.09, "learning_rate": 0.0004983292403075808, "loss": 4.1285, "step": 2160 }, { "epoch": 0.09, "learning_rate": 0.0004983191754901566, "loss": 4.0588, "step": 2170 }, { "epoch": 0.09, "learning_rate": 0.0004983091106727324, "loss": 3.7453, "step": 2180 }, { "epoch": 0.09, "learning_rate": 0.0004982990458553083, "loss": 3.8906, "step": 2190 }, { "epoch": 0.09, "learning_rate": 0.000498288981037884, "loss": 4.0859, "step": 2200 }, { "epoch": 0.09, "learning_rate": 0.0004982789162204597, "loss": 3.9757, "step": 2210 }, { "epoch": 0.09, "learning_rate": 0.0004982688514030356, "loss": 4.1072, "step": 2220 }, { "epoch": 0.09, "learning_rate": 0.0004982587865856113, "loss": 3.7888, "step": 2230 }, { "epoch": 0.09, "learning_rate": 0.0004982487217681872, "loss": 3.8168, "step": 2240 }, { "epoch": 0.09, "learning_rate": 0.0004982386569507629, "loss": 3.6895, "step": 2250 }, { "epoch": 0.09, "learning_rate": 0.0004982285921333388, "loss": 3.9353, "step": 2260 }, { "epoch": 0.09, "learning_rate": 0.0004982185273159145, "loss": 3.9252, "step": 2270 }, { "epoch": 0.09, "learning_rate": 0.0004982084624984902, "loss": 3.8687, "step": 2280 }, { "epoch": 0.09, "learning_rate": 0.0004981983976810661, "loss": 3.7369, "step": 2290 }, { "epoch": 0.09, "learning_rate": 0.0004981883328636418, "loss": 3.8433, "step": 2300 }, { "epoch": 0.09, "learning_rate": 0.0004981782680462177, "loss": 3.8987, "step": 2310 }, { "epoch": 0.09, "learning_rate": 0.0004981682032287934, "loss": 3.6546, "step": 2320 }, { "epoch": 0.09, "learning_rate": 0.0004981581384113692, "loss": 4.0541, "step": 2330 }, { "epoch": 0.09, "learning_rate": 0.0004981480735939451, "loss": 3.5304, "step": 2340 }, { "epoch": 0.09, "learning_rate": 0.0004981380087765208, "loss": 3.5371, "step": 2350 }, { "epoch": 0.09, "learning_rate": 0.0004981279439590966, "loss": 3.9227, "step": 2360 }, { "epoch": 0.1, "learning_rate": 0.0004981178791416724, "loss": 3.7466, "step": 2370 }, { "epoch": 0.1, "learning_rate": 0.0004981078143242481, "loss": 3.9714, "step": 2380 }, { "epoch": 0.1, "learning_rate": 0.000498097749506824, "loss": 3.8155, "step": 2390 }, { "epoch": 0.1, "learning_rate": 0.0004980876846893997, "loss": 3.7942, "step": 2400 }, { "epoch": 0.1, "learning_rate": 0.0004980776198719756, "loss": 3.5684, "step": 2410 }, { "epoch": 0.1, "learning_rate": 0.0004980675550545513, "loss": 3.8792, "step": 2420 }, { "epoch": 0.1, "learning_rate": 0.0004980574902371271, "loss": 3.899, "step": 2430 }, { "epoch": 0.1, "learning_rate": 0.0004980474254197029, "loss": 3.9976, "step": 2440 }, { "epoch": 0.1, "learning_rate": 0.0004980373606022786, "loss": 3.855, "step": 2450 }, { "epoch": 0.1, "learning_rate": 0.0004980272957848545, "loss": 3.4133, "step": 2460 }, { "epoch": 0.1, "learning_rate": 0.0004980172309674302, "loss": 3.7825, "step": 2470 }, { "epoch": 0.1, "learning_rate": 0.000498007166150006, "loss": 3.6793, "step": 2480 }, { "epoch": 0.1, "learning_rate": 0.0004979971013325819, "loss": 3.8989, "step": 2490 }, { "epoch": 0.1, "learning_rate": 0.0004979870365151576, "loss": 3.9458, "step": 2500 }, { "epoch": 0.1, "learning_rate": 0.0004979769716977335, "loss": 4.0239, "step": 2510 }, { "epoch": 0.1, "learning_rate": 0.0004979669068803092, "loss": 3.7777, "step": 2520 }, { "epoch": 0.1, "learning_rate": 0.0004979568420628849, "loss": 4.1361, "step": 2530 }, { "epoch": 0.1, "learning_rate": 0.0004979467772454608, "loss": 4.6284, "step": 2540 }, { "epoch": 0.1, "learning_rate": 0.0004979367124280365, "loss": 4.1372, "step": 2550 }, { "epoch": 0.1, "learning_rate": 0.0004979266476106124, "loss": 4.5068, "step": 2560 }, { "epoch": 0.1, "learning_rate": 0.0004979165827931881, "loss": 4.0258, "step": 2570 }, { "epoch": 0.1, "learning_rate": 0.000497906517975764, "loss": 4.2993, "step": 2580 }, { "epoch": 0.1, "learning_rate": 0.0004978964531583397, "loss": 4.0526, "step": 2590 }, { "epoch": 0.1, "learning_rate": 0.0004978863883409155, "loss": 4.2795, "step": 2600 }, { "epoch": 0.1, "learning_rate": 0.0004978763235234913, "loss": 4.1905, "step": 2610 }, { "epoch": 0.11, "learning_rate": 0.0004978662587060671, "loss": 3.887, "step": 2620 }, { "epoch": 0.11, "learning_rate": 0.0004978561938886429, "loss": 4.1671, "step": 2630 }, { "epoch": 0.11, "learning_rate": 0.0004978461290712187, "loss": 3.8931, "step": 2640 }, { "epoch": 0.11, "learning_rate": 0.0004978360642537944, "loss": 4.3564, "step": 2650 }, { "epoch": 0.11, "learning_rate": 0.0004978259994363703, "loss": 4.1785, "step": 2660 }, { "epoch": 0.11, "learning_rate": 0.000497815934618946, "loss": 4.224, "step": 2670 }, { "epoch": 0.11, "learning_rate": 0.0004978058698015219, "loss": 3.7816, "step": 2680 }, { "epoch": 0.11, "learning_rate": 0.0004977958049840976, "loss": 3.817, "step": 2690 }, { "epoch": 0.11, "learning_rate": 0.0004977857401666733, "loss": 3.7505, "step": 2700 }, { "epoch": 0.11, "learning_rate": 0.0004977756753492492, "loss": 3.6608, "step": 2710 }, { "epoch": 0.11, "learning_rate": 0.0004977656105318249, "loss": 3.7537, "step": 2720 }, { "epoch": 0.11, "learning_rate": 0.0004977555457144008, "loss": 3.9693, "step": 2730 }, { "epoch": 0.11, "learning_rate": 0.0004977454808969765, "loss": 3.4175, "step": 2740 }, { "epoch": 0.11, "learning_rate": 0.0004977354160795523, "loss": 3.6201, "step": 2750 }, { "epoch": 0.11, "learning_rate": 0.0004977253512621281, "loss": 3.4799, "step": 2760 }, { "epoch": 0.11, "learning_rate": 0.0004977152864447039, "loss": 3.5987, "step": 2770 }, { "epoch": 0.11, "learning_rate": 0.0004977052216272798, "loss": 3.5622, "step": 2780 }, { "epoch": 0.11, "learning_rate": 0.0004976951568098555, "loss": 3.3876, "step": 2790 }, { "epoch": 0.11, "learning_rate": 0.0004976850919924312, "loss": 3.5733, "step": 2800 }, { "epoch": 0.11, "learning_rate": 0.0004976750271750071, "loss": 3.5008, "step": 2810 }, { "epoch": 0.11, "learning_rate": 0.0004976649623575828, "loss": 3.2841, "step": 2820 }, { "epoch": 0.11, "learning_rate": 0.0004976548975401587, "loss": 3.4872, "step": 2830 }, { "epoch": 0.11, "learning_rate": 0.0004976448327227344, "loss": 3.6131, "step": 2840 }, { "epoch": 0.11, "learning_rate": 0.0004976347679053102, "loss": 3.4651, "step": 2850 }, { "epoch": 0.12, "learning_rate": 0.000497624703087886, "loss": 3.3652, "step": 2860 }, { "epoch": 0.12, "learning_rate": 0.0004976146382704617, "loss": 3.4134, "step": 2870 }, { "epoch": 0.12, "learning_rate": 0.0004976045734530376, "loss": 3.6579, "step": 2880 }, { "epoch": 0.12, "learning_rate": 0.0004975945086356133, "loss": 3.5421, "step": 2890 }, { "epoch": 0.12, "learning_rate": 0.0004975844438181891, "loss": 3.5389, "step": 2900 }, { "epoch": 0.12, "learning_rate": 0.0004975743790007649, "loss": 3.3793, "step": 2910 }, { "epoch": 0.12, "learning_rate": 0.0004975643141833407, "loss": 3.4391, "step": 2920 }, { "epoch": 0.12, "learning_rate": 0.0004975542493659166, "loss": 3.5395, "step": 2930 }, { "epoch": 0.12, "learning_rate": 0.0004975441845484923, "loss": 3.6553, "step": 2940 }, { "epoch": 0.12, "learning_rate": 0.0004975341197310682, "loss": 3.1535, "step": 2950 }, { "epoch": 0.12, "learning_rate": 0.0004975240549136439, "loss": 3.5183, "step": 2960 }, { "epoch": 0.12, "learning_rate": 0.0004975139900962196, "loss": 3.3148, "step": 2970 }, { "epoch": 0.12, "learning_rate": 0.0004975039252787955, "loss": 3.3514, "step": 2980 }, { "epoch": 0.12, "learning_rate": 0.0004974938604613712, "loss": 3.4141, "step": 2990 }, { "epoch": 0.12, "learning_rate": 0.0004974837956439471, "loss": 3.6611, "step": 3000 }, { "epoch": 0.12, "learning_rate": 0.0004974737308265228, "loss": 3.5529, "step": 3010 }, { "epoch": 0.12, "learning_rate": 0.0004974636660090986, "loss": 3.7058, "step": 3020 }, { "epoch": 0.12, "learning_rate": 0.0004974536011916744, "loss": 3.4836, "step": 3030 }, { "epoch": 0.12, "learning_rate": 0.0004974435363742501, "loss": 3.6475, "step": 3040 }, { "epoch": 0.12, "learning_rate": 0.000497433471556826, "loss": 3.6912, "step": 3050 }, { "epoch": 0.12, "learning_rate": 0.0004974234067394018, "loss": 3.6976, "step": 3060 }, { "epoch": 0.12, "learning_rate": 0.0004974133419219775, "loss": 3.4766, "step": 3070 }, { "epoch": 0.12, "learning_rate": 0.0004974032771045534, "loss": 3.3256, "step": 3080 }, { "epoch": 0.12, "learning_rate": 0.0004973932122871291, "loss": 3.3051, "step": 3090 }, { "epoch": 0.12, "learning_rate": 0.000497383147469705, "loss": 3.2918, "step": 3100 }, { "epoch": 0.13, "learning_rate": 0.0004973730826522807, "loss": 3.3709, "step": 3110 }, { "epoch": 0.13, "learning_rate": 0.0004973630178348564, "loss": 3.4649, "step": 3120 }, { "epoch": 0.13, "learning_rate": 0.0004973529530174323, "loss": 3.3695, "step": 3130 }, { "epoch": 0.13, "learning_rate": 0.000497342888200008, "loss": 3.1171, "step": 3140 }, { "epoch": 0.13, "learning_rate": 0.0004973328233825839, "loss": 3.1544, "step": 3150 }, { "epoch": 0.13, "learning_rate": 0.0004973227585651596, "loss": 3.5125, "step": 3160 }, { "epoch": 0.13, "learning_rate": 0.0004973126937477354, "loss": 3.3241, "step": 3170 }, { "epoch": 0.13, "learning_rate": 0.0004973026289303112, "loss": 3.2988, "step": 3180 }, { "epoch": 0.13, "learning_rate": 0.000497292564112887, "loss": 3.2849, "step": 3190 }, { "epoch": 0.13, "learning_rate": 0.0004972824992954628, "loss": 3.3931, "step": 3200 }, { "epoch": 0.13, "learning_rate": 0.0004972724344780386, "loss": 3.2431, "step": 3210 }, { "epoch": 0.13, "learning_rate": 0.0004972623696606143, "loss": 3.1895, "step": 3220 }, { "epoch": 0.13, "learning_rate": 0.0004972523048431902, "loss": 3.5138, "step": 3230 }, { "epoch": 0.13, "learning_rate": 0.0004972422400257659, "loss": 3.235, "step": 3240 }, { "epoch": 0.13, "learning_rate": 0.0004972321752083418, "loss": 3.3493, "step": 3250 }, { "epoch": 0.13, "learning_rate": 0.0004972221103909175, "loss": 3.0921, "step": 3260 }, { "epoch": 0.13, "learning_rate": 0.0004972120455734934, "loss": 3.362, "step": 3270 }, { "epoch": 0.13, "learning_rate": 0.0004972019807560691, "loss": 3.3204, "step": 3280 }, { "epoch": 0.13, "learning_rate": 0.0004971919159386448, "loss": 3.4566, "step": 3290 }, { "epoch": 0.13, "learning_rate": 0.0004971818511212207, "loss": 3.4707, "step": 3300 }, { "epoch": 0.13, "learning_rate": 0.0004971717863037964, "loss": 3.2825, "step": 3310 }, { "epoch": 0.13, "learning_rate": 0.0004971617214863723, "loss": 3.625, "step": 3320 }, { "epoch": 0.13, "learning_rate": 0.000497151656668948, "loss": 3.1852, "step": 3330 }, { "epoch": 0.13, "learning_rate": 0.0004971415918515238, "loss": 3.4824, "step": 3340 }, { "epoch": 0.13, "learning_rate": 0.0004971315270340996, "loss": 3.1839, "step": 3350 }, { "epoch": 0.14, "learning_rate": 0.0004971214622166754, "loss": 3.205, "step": 3360 }, { "epoch": 0.14, "learning_rate": 0.0004971113973992513, "loss": 3.3706, "step": 3370 }, { "epoch": 0.14, "learning_rate": 0.000497101332581827, "loss": 3.2446, "step": 3380 }, { "epoch": 0.14, "learning_rate": 0.0004970912677644027, "loss": 3.428, "step": 3390 }, { "epoch": 0.14, "learning_rate": 0.0004970812029469786, "loss": 3.3282, "step": 3400 }, { "epoch": 0.14, "learning_rate": 0.0004970711381295543, "loss": 3.4579, "step": 3410 }, { "epoch": 0.14, "learning_rate": 0.0004970610733121302, "loss": 3.3511, "step": 3420 }, { "epoch": 0.14, "learning_rate": 0.0004970510084947059, "loss": 3.5192, "step": 3430 }, { "epoch": 0.14, "learning_rate": 0.0004970409436772817, "loss": 3.3623, "step": 3440 }, { "epoch": 0.14, "learning_rate": 0.0004970308788598575, "loss": 3.3895, "step": 3450 }, { "epoch": 0.14, "learning_rate": 0.0004970208140424332, "loss": 3.2469, "step": 3460 }, { "epoch": 0.14, "learning_rate": 0.0004970107492250091, "loss": 3.1397, "step": 3470 }, { "epoch": 0.14, "learning_rate": 0.0004970006844075848, "loss": 3.2481, "step": 3480 }, { "epoch": 0.14, "learning_rate": 0.0004969906195901606, "loss": 3.1872, "step": 3490 }, { "epoch": 0.14, "learning_rate": 0.0004969805547727365, "loss": 2.9192, "step": 3500 }, { "epoch": 0.14, "learning_rate": 0.0004969704899553122, "loss": 3.2793, "step": 3510 }, { "epoch": 0.14, "learning_rate": 0.0004969604251378881, "loss": 3.4971, "step": 3520 }, { "epoch": 0.14, "learning_rate": 0.0004969503603204638, "loss": 3.1852, "step": 3530 }, { "epoch": 0.14, "learning_rate": 0.0004969402955030396, "loss": 2.9807, "step": 3540 }, { "epoch": 0.14, "learning_rate": 0.0004969302306856154, "loss": 3.4011, "step": 3550 }, { "epoch": 0.14, "learning_rate": 0.0004969201658681911, "loss": 3.2266, "step": 3560 }, { "epoch": 0.14, "learning_rate": 0.000496910101050767, "loss": 2.8794, "step": 3570 }, { "epoch": 0.14, "learning_rate": 0.0004969000362333427, "loss": 3.3286, "step": 3580 }, { "epoch": 0.14, "learning_rate": 0.0004968899714159185, "loss": 3.3261, "step": 3590 }, { "epoch": 0.14, "learning_rate": 0.0004968799065984943, "loss": 3.139, "step": 3600 }, { "epoch": 0.15, "learning_rate": 0.0004968698417810701, "loss": 3.2977, "step": 3610 }, { "epoch": 0.15, "learning_rate": 0.0004968597769636459, "loss": 3.2808, "step": 3620 }, { "epoch": 0.15, "learning_rate": 0.0004968497121462216, "loss": 3.2529, "step": 3630 }, { "epoch": 0.15, "learning_rate": 0.0004968396473287974, "loss": 3.2002, "step": 3640 }, { "epoch": 0.15, "learning_rate": 0.0004968295825113733, "loss": 3.0244, "step": 3650 }, { "epoch": 0.15, "learning_rate": 0.000496819517693949, "loss": 3.1616, "step": 3660 }, { "epoch": 0.15, "learning_rate": 0.0004968094528765249, "loss": 3.4065, "step": 3670 }, { "epoch": 0.15, "learning_rate": 0.0004967993880591006, "loss": 3.2781, "step": 3680 }, { "epoch": 0.15, "learning_rate": 0.0004967893232416765, "loss": 3.3344, "step": 3690 }, { "epoch": 0.15, "learning_rate": 0.0004967792584242522, "loss": 4.0664, "step": 3700 }, { "epoch": 0.15, "learning_rate": 0.000496769193606828, "loss": 3.6834, "step": 3710 }, { "epoch": 0.15, "learning_rate": 0.0004967591287894038, "loss": 3.6978, "step": 3720 }, { "epoch": 0.15, "learning_rate": 0.0004967490639719795, "loss": 3.6311, "step": 3730 }, { "epoch": 0.15, "learning_rate": 0.0004967389991545554, "loss": 3.0278, "step": 3740 }, { "epoch": 0.15, "learning_rate": 0.0004967289343371311, "loss": 3.1963, "step": 3750 }, { "epoch": 0.15, "learning_rate": 0.0004967188695197069, "loss": 3.1915, "step": 3760 }, { "epoch": 0.15, "learning_rate": 0.0004967088047022827, "loss": 3.1057, "step": 3770 }, { "epoch": 0.15, "learning_rate": 0.0004966987398848585, "loss": 3.0061, "step": 3780 }, { "epoch": 0.15, "learning_rate": 0.0004966886750674344, "loss": 3.2814, "step": 3790 }, { "epoch": 0.15, "learning_rate": 0.0004966786102500101, "loss": 3.3398, "step": 3800 }, { "epoch": 0.15, "learning_rate": 0.0004966685454325858, "loss": 3.1964, "step": 3810 }, { "epoch": 0.15, "learning_rate": 0.0004966584806151617, "loss": 3.1746, "step": 3820 }, { "epoch": 0.15, "learning_rate": 0.0004966484157977374, "loss": 3.3023, "step": 3830 }, { "epoch": 0.15, "learning_rate": 0.0004966383509803133, "loss": 3.3625, "step": 3840 }, { "epoch": 0.15, "learning_rate": 0.000496628286162889, "loss": 2.9949, "step": 3850 }, { "epoch": 0.16, "learning_rate": 0.0004966182213454648, "loss": 3.5711, "step": 3860 }, { "epoch": 0.16, "learning_rate": 0.0004966081565280406, "loss": 3.2415, "step": 3870 }, { "epoch": 0.16, "learning_rate": 0.0004965980917106163, "loss": 3.1466, "step": 3880 }, { "epoch": 0.16, "learning_rate": 0.0004965880268931922, "loss": 3.1714, "step": 3890 }, { "epoch": 0.16, "learning_rate": 0.0004965779620757679, "loss": 3.1827, "step": 3900 }, { "epoch": 0.16, "learning_rate": 0.0004965678972583437, "loss": 2.8089, "step": 3910 }, { "epoch": 0.16, "learning_rate": 0.0004965578324409195, "loss": 2.978, "step": 3920 }, { "epoch": 0.16, "learning_rate": 0.0004965477676234953, "loss": 3.2167, "step": 3930 }, { "epoch": 0.16, "learning_rate": 0.0004965377028060712, "loss": 3.0611, "step": 3940 }, { "epoch": 0.16, "learning_rate": 0.0004965276379886469, "loss": 3.0129, "step": 3950 }, { "epoch": 0.16, "learning_rate": 0.0004965175731712228, "loss": 3.3542, "step": 3960 }, { "epoch": 0.16, "learning_rate": 0.0004965075083537985, "loss": 3.4148, "step": 3970 }, { "epoch": 0.16, "learning_rate": 0.0004964974435363742, "loss": 3.5862, "step": 3980 }, { "epoch": 0.16, "learning_rate": 0.0004964873787189501, "loss": 3.1365, "step": 3990 }, { "epoch": 0.16, "learning_rate": 0.0004964773139015258, "loss": 3.2045, "step": 4000 }, { "epoch": 0.16, "learning_rate": 0.0004964672490841017, "loss": 2.6069, "step": 4010 }, { "epoch": 0.16, "learning_rate": 0.0004964571842666774, "loss": 3.1084, "step": 4020 }, { "epoch": 0.16, "learning_rate": 0.0004964471194492532, "loss": 3.3933, "step": 4030 }, { "epoch": 0.16, "learning_rate": 0.000496437054631829, "loss": 3.4157, "step": 4040 }, { "epoch": 0.16, "learning_rate": 0.0004964269898144047, "loss": 3.0656, "step": 4050 }, { "epoch": 0.16, "learning_rate": 0.0004964169249969806, "loss": 3.2758, "step": 4060 }, { "epoch": 0.16, "learning_rate": 0.0004964068601795563, "loss": 2.9621, "step": 4070 }, { "epoch": 0.16, "learning_rate": 0.0004963967953621321, "loss": 2.926, "step": 4080 }, { "epoch": 0.16, "learning_rate": 0.000496386730544708, "loss": 3.1047, "step": 4090 }, { "epoch": 0.16, "learning_rate": 0.0004963766657272837, "loss": 2.9643, "step": 4100 }, { "epoch": 0.17, "learning_rate": 0.0004963666009098596, "loss": 3.1476, "step": 4110 }, { "epoch": 0.17, "learning_rate": 0.0004963565360924353, "loss": 2.8106, "step": 4120 }, { "epoch": 0.17, "learning_rate": 0.0004963464712750111, "loss": 2.6961, "step": 4130 }, { "epoch": 0.17, "learning_rate": 0.0004963364064575869, "loss": 2.8098, "step": 4140 }, { "epoch": 0.17, "learning_rate": 0.0004963263416401626, "loss": 2.8861, "step": 4150 }, { "epoch": 0.17, "learning_rate": 0.0004963162768227385, "loss": 3.0071, "step": 4160 }, { "epoch": 0.17, "learning_rate": 0.0004963062120053142, "loss": 3.2189, "step": 4170 }, { "epoch": 0.17, "learning_rate": 0.00049629614718789, "loss": 2.6315, "step": 4180 }, { "epoch": 0.17, "learning_rate": 0.0004962860823704658, "loss": 3.1194, "step": 4190 }, { "epoch": 0.17, "learning_rate": 0.0004962760175530416, "loss": 2.8986, "step": 4200 }, { "epoch": 0.17, "learning_rate": 0.0004962659527356174, "loss": 2.7889, "step": 4210 }, { "epoch": 0.17, "learning_rate": 0.0004962558879181931, "loss": 3.4869, "step": 4220 }, { "epoch": 0.17, "learning_rate": 0.0004962458231007689, "loss": 2.8804, "step": 4230 }, { "epoch": 0.17, "learning_rate": 0.0004962357582833448, "loss": 3.0753, "step": 4240 }, { "epoch": 0.17, "learning_rate": 0.0004962256934659205, "loss": 3.1096, "step": 4250 }, { "epoch": 0.17, "learning_rate": 0.0004962156286484964, "loss": 2.9593, "step": 4260 }, { "epoch": 0.17, "learning_rate": 0.0004962055638310721, "loss": 2.8227, "step": 4270 }, { "epoch": 0.17, "learning_rate": 0.000496195499013648, "loss": 3.1297, "step": 4280 }, { "epoch": 0.17, "learning_rate": 0.0004961854341962237, "loss": 2.8645, "step": 4290 }, { "epoch": 0.17, "learning_rate": 0.0004961753693787995, "loss": 2.9407, "step": 4300 }, { "epoch": 0.17, "learning_rate": 0.0004961653045613753, "loss": 2.8808, "step": 4310 }, { "epoch": 0.17, "learning_rate": 0.000496155239743951, "loss": 3.0871, "step": 4320 }, { "epoch": 0.17, "learning_rate": 0.0004961451749265269, "loss": 2.9011, "step": 4330 }, { "epoch": 0.17, "learning_rate": 0.0004961351101091026, "loss": 2.9874, "step": 4340 }, { "epoch": 0.17, "learning_rate": 0.0004961250452916784, "loss": 3.0849, "step": 4350 }, { "epoch": 0.18, "learning_rate": 0.0004961149804742542, "loss": 2.8904, "step": 4360 }, { "epoch": 0.18, "learning_rate": 0.00049610491565683, "loss": 2.9241, "step": 4370 }, { "epoch": 0.18, "learning_rate": 0.0004960948508394059, "loss": 3.1828, "step": 4380 }, { "epoch": 0.18, "learning_rate": 0.0004960847860219816, "loss": 2.9005, "step": 4390 }, { "epoch": 0.18, "learning_rate": 0.0004960747212045573, "loss": 2.9534, "step": 4400 }, { "epoch": 0.18, "learning_rate": 0.0004960646563871332, "loss": 3.047, "step": 4410 }, { "epoch": 0.18, "learning_rate": 0.0004960545915697089, "loss": 2.9061, "step": 4420 }, { "epoch": 0.18, "learning_rate": 0.0004960445267522848, "loss": 3.2624, "step": 4430 }, { "epoch": 0.18, "learning_rate": 0.0004960344619348605, "loss": 2.9217, "step": 4440 }, { "epoch": 0.18, "learning_rate": 0.0004960243971174363, "loss": 3.1217, "step": 4450 }, { "epoch": 0.18, "learning_rate": 0.0004960143323000121, "loss": 2.8333, "step": 4460 }, { "epoch": 0.18, "learning_rate": 0.0004960042674825879, "loss": 3.2877, "step": 4470 }, { "epoch": 0.18, "learning_rate": 0.0004959942026651637, "loss": 2.8856, "step": 4480 }, { "epoch": 0.18, "learning_rate": 0.0004959841378477394, "loss": 3.1463, "step": 4490 }, { "epoch": 0.18, "learning_rate": 0.0004959740730303152, "loss": 3.0235, "step": 4500 }, { "epoch": 0.18, "learning_rate": 0.000495964008212891, "loss": 3.1396, "step": 4510 }, { "epoch": 0.18, "learning_rate": 0.0004959539433954668, "loss": 2.8052, "step": 4520 }, { "epoch": 0.18, "learning_rate": 0.0004959438785780427, "loss": 3.1863, "step": 4530 }, { "epoch": 0.18, "learning_rate": 0.0004959338137606184, "loss": 2.768, "step": 4540 }, { "epoch": 0.18, "learning_rate": 0.0004959237489431942, "loss": 3.0024, "step": 4550 }, { "epoch": 0.18, "learning_rate": 0.00049591368412577, "loss": 2.9106, "step": 4560 }, { "epoch": 0.18, "learning_rate": 0.0004959036193083457, "loss": 2.956, "step": 4570 }, { "epoch": 0.18, "learning_rate": 0.0004958935544909216, "loss": 3.1474, "step": 4580 }, { "epoch": 0.18, "learning_rate": 0.0004958834896734973, "loss": 2.7926, "step": 4590 }, { "epoch": 0.19, "learning_rate": 0.0004958734248560731, "loss": 2.6942, "step": 4600 }, { "epoch": 0.19, "learning_rate": 0.0004958633600386489, "loss": 2.8558, "step": 4610 }, { "epoch": 0.19, "learning_rate": 0.0004958532952212247, "loss": 2.8449, "step": 4620 }, { "epoch": 0.19, "learning_rate": 0.0004958432304038005, "loss": 2.7286, "step": 4630 }, { "epoch": 0.19, "learning_rate": 0.0004958331655863762, "loss": 2.6975, "step": 4640 }, { "epoch": 0.19, "learning_rate": 0.000495823100768952, "loss": 2.9662, "step": 4650 }, { "epoch": 0.19, "learning_rate": 0.0004958130359515278, "loss": 3.0595, "step": 4660 }, { "epoch": 0.19, "learning_rate": 0.0004958029711341036, "loss": 2.7096, "step": 4670 }, { "epoch": 0.19, "learning_rate": 0.0004957929063166795, "loss": 2.9132, "step": 4680 }, { "epoch": 0.19, "learning_rate": 0.0004957828414992552, "loss": 2.8241, "step": 4690 }, { "epoch": 0.19, "learning_rate": 0.0004957727766818311, "loss": 3.0004, "step": 4700 }, { "epoch": 0.19, "learning_rate": 0.0004957627118644068, "loss": 2.9372, "step": 4710 }, { "epoch": 0.19, "learning_rate": 0.0004957526470469826, "loss": 2.8242, "step": 4720 }, { "epoch": 0.19, "learning_rate": 0.0004957425822295584, "loss": 2.7821, "step": 4730 }, { "epoch": 0.19, "learning_rate": 0.0004957325174121341, "loss": 3.0357, "step": 4740 }, { "epoch": 0.19, "learning_rate": 0.00049572245259471, "loss": 2.6544, "step": 4750 }, { "epoch": 0.19, "learning_rate": 0.0004957123877772857, "loss": 2.7951, "step": 4760 }, { "epoch": 0.19, "learning_rate": 0.0004957023229598615, "loss": 3.1265, "step": 4770 }, { "epoch": 0.19, "learning_rate": 0.0004956922581424373, "loss": 2.8536, "step": 4780 }, { "epoch": 0.19, "learning_rate": 0.0004956821933250131, "loss": 2.5293, "step": 4790 }, { "epoch": 0.19, "learning_rate": 0.0004956721285075889, "loss": 3.1884, "step": 4800 }, { "epoch": 0.19, "learning_rate": 0.0004956620636901646, "loss": 2.6987, "step": 4810 }, { "epoch": 0.19, "learning_rate": 0.0004956519988727404, "loss": 2.8872, "step": 4820 }, { "epoch": 0.19, "learning_rate": 0.0004956419340553163, "loss": 2.9701, "step": 4830 }, { "epoch": 0.19, "learning_rate": 0.000495631869237892, "loss": 2.9046, "step": 4840 }, { "epoch": 0.2, "learning_rate": 0.0004956218044204679, "loss": 2.7736, "step": 4850 }, { "epoch": 0.2, "learning_rate": 0.0004956117396030436, "loss": 2.7111, "step": 4860 }, { "epoch": 0.2, "learning_rate": 0.0004956016747856194, "loss": 2.9011, "step": 4870 }, { "epoch": 0.2, "learning_rate": 0.0004955916099681952, "loss": 2.9475, "step": 4880 }, { "epoch": 0.2, "learning_rate": 0.000495581545150771, "loss": 3.0267, "step": 4890 }, { "epoch": 0.2, "learning_rate": 0.0004955714803333468, "loss": 2.9197, "step": 4900 }, { "epoch": 0.2, "learning_rate": 0.0004955614155159225, "loss": 2.3555, "step": 4910 }, { "epoch": 0.2, "learning_rate": 0.0004955513506984983, "loss": 2.976, "step": 4920 }, { "epoch": 0.2, "learning_rate": 0.0004955412858810741, "loss": 3.1111, "step": 4930 }, { "epoch": 0.2, "learning_rate": 0.0004955312210636499, "loss": 3.0197, "step": 4940 }, { "epoch": 0.2, "learning_rate": 0.0004955211562462257, "loss": 2.7527, "step": 4950 }, { "epoch": 0.2, "learning_rate": 0.0004955110914288015, "loss": 2.9049, "step": 4960 }, { "epoch": 0.2, "learning_rate": 0.0004955010266113774, "loss": 2.7376, "step": 4970 }, { "epoch": 0.2, "learning_rate": 0.0004954909617939531, "loss": 2.9593, "step": 4980 }, { "epoch": 0.2, "learning_rate": 0.0004954808969765288, "loss": 2.5492, "step": 4990 }, { "epoch": 0.2, "learning_rate": 0.0004954708321591047, "loss": 2.8643, "step": 5000 }, { "epoch": 0.2, "learning_rate": 0.0004954607673416804, "loss": 2.7726, "step": 5010 }, { "epoch": 0.2, "learning_rate": 0.0004954507025242563, "loss": 2.7303, "step": 5020 }, { "epoch": 0.2, "learning_rate": 0.000495440637706832, "loss": 2.8117, "step": 5030 }, { "epoch": 0.2, "learning_rate": 0.0004954305728894078, "loss": 2.6733, "step": 5040 }, { "epoch": 0.2, "learning_rate": 0.0004954205080719836, "loss": 2.5375, "step": 5050 }, { "epoch": 0.2, "learning_rate": 0.0004954104432545594, "loss": 2.7863, "step": 5060 }, { "epoch": 0.2, "learning_rate": 0.0004954003784371352, "loss": 2.7684, "step": 5070 }, { "epoch": 0.2, "learning_rate": 0.0004953903136197109, "loss": 2.8229, "step": 5080 }, { "epoch": 0.2, "learning_rate": 0.0004953802488022867, "loss": 2.5866, "step": 5090 }, { "epoch": 0.21, "learning_rate": 0.0004953701839848625, "loss": 2.7255, "step": 5100 }, { "epoch": 0.21, "learning_rate": 0.0004953601191674383, "loss": 2.7529, "step": 5110 }, { "epoch": 0.21, "learning_rate": 0.0004953500543500142, "loss": 2.8561, "step": 5120 }, { "epoch": 0.21, "learning_rate": 0.0004953399895325899, "loss": 3.0261, "step": 5130 }, { "epoch": 0.21, "learning_rate": 0.0004953299247151657, "loss": 2.9733, "step": 5140 }, { "epoch": 0.21, "learning_rate": 0.0004953198598977415, "loss": 2.5709, "step": 5150 }, { "epoch": 0.21, "learning_rate": 0.0004953097950803172, "loss": 2.7104, "step": 5160 }, { "epoch": 0.21, "learning_rate": 0.0004952997302628931, "loss": 2.7946, "step": 5170 }, { "epoch": 0.21, "learning_rate": 0.0004952896654454688, "loss": 2.539, "step": 5180 }, { "epoch": 0.21, "learning_rate": 0.0004952796006280446, "loss": 2.9437, "step": 5190 }, { "epoch": 0.21, "learning_rate": 0.0004952695358106204, "loss": 2.8815, "step": 5200 }, { "epoch": 0.21, "learning_rate": 0.0004952594709931962, "loss": 2.5917, "step": 5210 }, { "epoch": 0.21, "learning_rate": 0.000495249406175772, "loss": 2.5003, "step": 5220 }, { "epoch": 0.21, "learning_rate": 0.0004952393413583478, "loss": 2.5421, "step": 5230 }, { "epoch": 0.21, "learning_rate": 0.0004952292765409235, "loss": 3.0234, "step": 5240 }, { "epoch": 0.21, "learning_rate": 0.0004952192117234993, "loss": 2.5699, "step": 5250 }, { "epoch": 0.21, "learning_rate": 0.0004952091469060751, "loss": 2.583, "step": 5260 }, { "epoch": 0.21, "learning_rate": 0.000495199082088651, "loss": 2.4337, "step": 5270 }, { "epoch": 0.21, "learning_rate": 0.0004951890172712267, "loss": 2.6464, "step": 5280 }, { "epoch": 0.21, "learning_rate": 0.0004951789524538025, "loss": 2.6126, "step": 5290 }, { "epoch": 0.21, "learning_rate": 0.0004951688876363783, "loss": 2.5432, "step": 5300 }, { "epoch": 0.21, "learning_rate": 0.0004951588228189541, "loss": 2.7288, "step": 5310 }, { "epoch": 0.21, "learning_rate": 0.0004951487580015299, "loss": 2.73, "step": 5320 }, { "epoch": 0.21, "learning_rate": 0.0004951386931841056, "loss": 2.5025, "step": 5330 }, { "epoch": 0.21, "learning_rate": 0.0004951286283666814, "loss": 2.4474, "step": 5340 }, { "epoch": 0.22, "learning_rate": 0.0004951185635492572, "loss": 2.6355, "step": 5350 }, { "epoch": 0.22, "learning_rate": 0.000495108498731833, "loss": 2.7917, "step": 5360 }, { "epoch": 0.22, "learning_rate": 0.0004950984339144088, "loss": 2.3573, "step": 5370 }, { "epoch": 0.22, "learning_rate": 0.0004950883690969846, "loss": 2.6335, "step": 5380 }, { "epoch": 0.22, "learning_rate": 0.0004950783042795603, "loss": 2.6518, "step": 5390 }, { "epoch": 0.22, "learning_rate": 0.0004950682394621362, "loss": 2.7296, "step": 5400 }, { "epoch": 0.22, "learning_rate": 0.0004950581746447119, "loss": 2.5178, "step": 5410 }, { "epoch": 0.22, "learning_rate": 0.0004950481098272878, "loss": 2.8853, "step": 5420 }, { "epoch": 0.22, "learning_rate": 0.0004950380450098635, "loss": 2.6441, "step": 5430 }, { "epoch": 0.22, "learning_rate": 0.0004950279801924394, "loss": 3.0134, "step": 5440 }, { "epoch": 0.22, "learning_rate": 0.0004950179153750151, "loss": 2.6097, "step": 5450 }, { "epoch": 0.22, "learning_rate": 0.0004950078505575909, "loss": 2.5236, "step": 5460 }, { "epoch": 0.22, "learning_rate": 0.0004949977857401667, "loss": 2.4144, "step": 5470 }, { "epoch": 0.22, "learning_rate": 0.0004949877209227425, "loss": 2.7146, "step": 5480 }, { "epoch": 0.22, "learning_rate": 0.0004949776561053183, "loss": 2.5774, "step": 5490 }, { "epoch": 0.22, "learning_rate": 0.000494967591287894, "loss": 2.5963, "step": 5500 }, { "epoch": 0.22, "learning_rate": 0.0004949575264704698, "loss": 2.8607, "step": 5510 }, { "epoch": 0.22, "learning_rate": 0.0004949474616530456, "loss": 2.9474, "step": 5520 }, { "epoch": 0.22, "learning_rate": 0.0004949373968356214, "loss": 2.5269, "step": 5530 }, { "epoch": 0.22, "learning_rate": 0.0004949273320181972, "loss": 2.5614, "step": 5540 }, { "epoch": 0.22, "learning_rate": 0.000494917267200773, "loss": 2.982, "step": 5550 }, { "epoch": 0.22, "learning_rate": 0.0004949072023833488, "loss": 2.6737, "step": 5560 }, { "epoch": 0.22, "learning_rate": 0.0004948971375659246, "loss": 2.775, "step": 5570 }, { "epoch": 0.22, "learning_rate": 0.0004948870727485003, "loss": 2.6179, "step": 5580 }, { "epoch": 0.22, "learning_rate": 0.0004948770079310762, "loss": 2.7976, "step": 5590 }, { "epoch": 0.23, "learning_rate": 0.0004948669431136519, "loss": 2.6672, "step": 5600 }, { "epoch": 0.23, "learning_rate": 0.0004948568782962277, "loss": 2.8466, "step": 5610 }, { "epoch": 0.23, "learning_rate": 0.0004948468134788035, "loss": 2.7651, "step": 5620 }, { "epoch": 0.23, "learning_rate": 0.0004948367486613793, "loss": 2.7245, "step": 5630 }, { "epoch": 0.23, "learning_rate": 0.0004948266838439551, "loss": 2.7182, "step": 5640 }, { "epoch": 0.23, "learning_rate": 0.0004948166190265309, "loss": 2.6112, "step": 5650 }, { "epoch": 0.23, "learning_rate": 0.0004948065542091066, "loss": 2.9692, "step": 5660 }, { "epoch": 0.23, "learning_rate": 0.0004947964893916824, "loss": 2.6909, "step": 5670 }, { "epoch": 0.23, "learning_rate": 0.0004947864245742582, "loss": 2.7655, "step": 5680 }, { "epoch": 0.23, "learning_rate": 0.000494776359756834, "loss": 2.4407, "step": 5690 }, { "epoch": 0.23, "learning_rate": 0.0004947662949394098, "loss": 2.4005, "step": 5700 }, { "epoch": 0.23, "learning_rate": 0.0004947562301219857, "loss": 2.9187, "step": 5710 }, { "epoch": 0.23, "learning_rate": 0.0004947461653045614, "loss": 2.6622, "step": 5720 }, { "epoch": 0.23, "learning_rate": 0.0004947361004871372, "loss": 2.8152, "step": 5730 }, { "epoch": 0.23, "learning_rate": 0.000494726035669713, "loss": 2.5448, "step": 5740 }, { "epoch": 0.23, "learning_rate": 0.0004947159708522887, "loss": 2.6227, "step": 5750 }, { "epoch": 0.23, "learning_rate": 0.0004947059060348646, "loss": 2.4048, "step": 5760 }, { "epoch": 0.23, "learning_rate": 0.0004946958412174403, "loss": 2.6114, "step": 5770 }, { "epoch": 0.23, "learning_rate": 0.0004946857764000161, "loss": 2.7011, "step": 5780 }, { "epoch": 0.23, "learning_rate": 0.0004946757115825919, "loss": 2.7589, "step": 5790 }, { "epoch": 0.23, "learning_rate": 0.0004946656467651677, "loss": 2.8323, "step": 5800 }, { "epoch": 0.23, "learning_rate": 0.0004946555819477435, "loss": 2.4696, "step": 5810 }, { "epoch": 0.23, "learning_rate": 0.0004946455171303193, "loss": 2.4456, "step": 5820 }, { "epoch": 0.23, "learning_rate": 0.000494635452312895, "loss": 2.5056, "step": 5830 }, { "epoch": 0.23, "learning_rate": 0.0004946253874954709, "loss": 2.823, "step": 5840 }, { "epoch": 0.24, "learning_rate": 0.0004946153226780466, "loss": 2.9519, "step": 5850 }, { "epoch": 0.24, "learning_rate": 0.0004946052578606225, "loss": 2.7386, "step": 5860 }, { "epoch": 0.24, "learning_rate": 0.0004945951930431982, "loss": 2.843, "step": 5870 }, { "epoch": 0.24, "learning_rate": 0.000494585128225774, "loss": 2.7688, "step": 5880 }, { "epoch": 0.24, "learning_rate": 0.0004945750634083498, "loss": 2.9375, "step": 5890 }, { "epoch": 0.24, "learning_rate": 0.0004945649985909256, "loss": 2.7765, "step": 5900 }, { "epoch": 0.24, "learning_rate": 0.0004945549337735014, "loss": 2.7435, "step": 5910 }, { "epoch": 0.24, "learning_rate": 0.0004945448689560771, "loss": 2.8087, "step": 5920 }, { "epoch": 0.24, "learning_rate": 0.0004945348041386529, "loss": 2.915, "step": 5930 }, { "epoch": 0.24, "learning_rate": 0.0004945247393212287, "loss": 2.6131, "step": 5940 }, { "epoch": 0.24, "learning_rate": 0.0004945146745038045, "loss": 2.8031, "step": 5950 }, { "epoch": 0.24, "learning_rate": 0.0004945046096863803, "loss": 2.6935, "step": 5960 }, { "epoch": 0.24, "learning_rate": 0.0004944945448689561, "loss": 2.9287, "step": 5970 }, { "epoch": 0.24, "learning_rate": 0.0004944844800515318, "loss": 2.953, "step": 5980 }, { "epoch": 0.24, "learning_rate": 0.0004944744152341077, "loss": 2.5577, "step": 5990 }, { "epoch": 0.24, "learning_rate": 0.0004944643504166834, "loss": 2.6617, "step": 6000 }, { "epoch": 0.24, "learning_rate": 0.0004944542855992593, "loss": 2.7828, "step": 6010 }, { "epoch": 0.24, "learning_rate": 0.000494444220781835, "loss": 2.6188, "step": 6020 }, { "epoch": 0.24, "learning_rate": 0.0004944341559644108, "loss": 2.5171, "step": 6030 }, { "epoch": 0.24, "learning_rate": 0.0004944240911469866, "loss": 2.5934, "step": 6040 }, { "epoch": 0.24, "learning_rate": 0.0004944140263295624, "loss": 2.6713, "step": 6050 }, { "epoch": 0.24, "learning_rate": 0.0004944039615121382, "loss": 2.5589, "step": 6060 }, { "epoch": 0.24, "learning_rate": 0.000494393896694714, "loss": 2.5145, "step": 6070 }, { "epoch": 0.24, "learning_rate": 0.0004943838318772897, "loss": 2.5469, "step": 6080 }, { "epoch": 0.24, "learning_rate": 0.0004943737670598655, "loss": 2.5896, "step": 6090 }, { "epoch": 0.25, "learning_rate": 0.0004943637022424413, "loss": 2.6307, "step": 6100 }, { "epoch": 0.25, "learning_rate": 0.0004943536374250171, "loss": 2.7047, "step": 6110 }, { "epoch": 0.25, "learning_rate": 0.0004943435726075929, "loss": 2.4709, "step": 6120 }, { "epoch": 0.25, "learning_rate": 0.0004943335077901688, "loss": 2.558, "step": 6130 }, { "epoch": 0.25, "learning_rate": 0.0004943234429727445, "loss": 2.7934, "step": 6140 }, { "epoch": 0.25, "learning_rate": 0.0004943133781553203, "loss": 2.5775, "step": 6150 }, { "epoch": 0.25, "learning_rate": 0.0004943033133378961, "loss": 2.4954, "step": 6160 }, { "epoch": 0.25, "learning_rate": 0.0004942932485204718, "loss": 2.4937, "step": 6170 }, { "epoch": 0.25, "learning_rate": 0.0004942831837030477, "loss": 2.4364, "step": 6180 }, { "epoch": 0.25, "learning_rate": 0.0004942731188856234, "loss": 2.4542, "step": 6190 }, { "epoch": 0.25, "learning_rate": 0.0004942630540681992, "loss": 2.8013, "step": 6200 }, { "epoch": 0.25, "learning_rate": 0.000494252989250775, "loss": 2.6823, "step": 6210 }, { "epoch": 0.25, "learning_rate": 0.0004942429244333508, "loss": 2.6875, "step": 6220 }, { "epoch": 0.25, "learning_rate": 0.0004942328596159266, "loss": 2.8049, "step": 6230 }, { "epoch": 0.25, "learning_rate": 0.0004942227947985024, "loss": 2.5667, "step": 6240 }, { "epoch": 0.25, "learning_rate": 0.0004942127299810781, "loss": 2.2779, "step": 6250 }, { "epoch": 0.25, "learning_rate": 0.0004942026651636539, "loss": 2.7376, "step": 6260 }, { "epoch": 0.25, "learning_rate": 0.0004941926003462297, "loss": 2.5986, "step": 6270 }, { "epoch": 0.25, "learning_rate": 0.0004941825355288056, "loss": 2.7602, "step": 6280 }, { "epoch": 0.25, "learning_rate": 0.0004941724707113813, "loss": 2.5512, "step": 6290 }, { "epoch": 0.25, "learning_rate": 0.0004941624058939571, "loss": 2.6205, "step": 6300 }, { "epoch": 0.25, "learning_rate": 0.0004941523410765329, "loss": 2.4691, "step": 6310 }, { "epoch": 0.25, "learning_rate": 0.0004941422762591087, "loss": 2.3391, "step": 6320 }, { "epoch": 0.25, "learning_rate": 0.0004941322114416845, "loss": 2.4048, "step": 6330 }, { "epoch": 0.25, "learning_rate": 0.0004941221466242602, "loss": 2.5862, "step": 6340 }, { "epoch": 0.26, "learning_rate": 0.000494112081806836, "loss": 2.5078, "step": 6350 }, { "epoch": 0.26, "learning_rate": 0.0004941020169894118, "loss": 2.5806, "step": 6360 }, { "epoch": 0.26, "learning_rate": 0.0004940919521719876, "loss": 2.7431, "step": 6370 }, { "epoch": 0.26, "learning_rate": 0.0004940818873545634, "loss": 2.4609, "step": 6380 }, { "epoch": 0.26, "learning_rate": 0.0004940718225371392, "loss": 2.6241, "step": 6390 }, { "epoch": 0.26, "learning_rate": 0.000494061757719715, "loss": 2.6445, "step": 6400 }, { "epoch": 0.26, "learning_rate": 0.0004940516929022908, "loss": 2.9056, "step": 6410 }, { "epoch": 0.26, "learning_rate": 0.0004940416280848665, "loss": 2.7153, "step": 6420 }, { "epoch": 0.26, "learning_rate": 0.0004940315632674424, "loss": 2.6533, "step": 6430 }, { "epoch": 0.26, "learning_rate": 0.0004940214984500181, "loss": 2.7628, "step": 6440 }, { "epoch": 0.26, "learning_rate": 0.000494011433632594, "loss": 2.5621, "step": 6450 }, { "epoch": 0.26, "learning_rate": 0.0004940013688151697, "loss": 2.5259, "step": 6460 }, { "epoch": 0.26, "learning_rate": 0.0004939913039977455, "loss": 2.6288, "step": 6470 }, { "epoch": 0.26, "learning_rate": 0.0004939812391803213, "loss": 2.7433, "step": 6480 }, { "epoch": 0.26, "learning_rate": 0.0004939711743628971, "loss": 2.6627, "step": 6490 }, { "epoch": 0.26, "learning_rate": 0.0004939611095454729, "loss": 2.5909, "step": 6500 }, { "epoch": 0.26, "learning_rate": 0.0004939510447280486, "loss": 2.6028, "step": 6510 }, { "epoch": 0.26, "learning_rate": 0.0004939409799106244, "loss": 2.3558, "step": 6520 }, { "epoch": 0.26, "learning_rate": 0.0004939309150932002, "loss": 2.4978, "step": 6530 }, { "epoch": 0.26, "learning_rate": 0.000493920850275776, "loss": 2.8378, "step": 6540 }, { "epoch": 0.26, "learning_rate": 0.0004939107854583518, "loss": 2.7598, "step": 6550 }, { "epoch": 0.26, "learning_rate": 0.0004939007206409276, "loss": 2.701, "step": 6560 }, { "epoch": 0.26, "learning_rate": 0.0004938906558235034, "loss": 2.7213, "step": 6570 }, { "epoch": 0.26, "learning_rate": 0.0004938805910060792, "loss": 2.6264, "step": 6580 }, { "epoch": 0.27, "learning_rate": 0.0004938705261886549, "loss": 2.3988, "step": 6590 }, { "epoch": 0.27, "learning_rate": 0.0004938604613712308, "loss": 2.3888, "step": 6600 }, { "epoch": 0.27, "learning_rate": 0.0004938503965538065, "loss": 2.6221, "step": 6610 }, { "epoch": 0.27, "learning_rate": 0.0004938403317363823, "loss": 2.7944, "step": 6620 }, { "epoch": 0.27, "learning_rate": 0.0004938302669189581, "loss": 2.6792, "step": 6630 }, { "epoch": 0.27, "learning_rate": 0.0004938202021015339, "loss": 2.3221, "step": 6640 }, { "epoch": 0.27, "learning_rate": 0.0004938101372841097, "loss": 2.9207, "step": 6650 }, { "epoch": 0.27, "learning_rate": 0.0004938000724666855, "loss": 2.4289, "step": 6660 }, { "epoch": 0.27, "learning_rate": 0.0004937900076492612, "loss": 2.7766, "step": 6670 }, { "epoch": 0.27, "learning_rate": 0.000493779942831837, "loss": 2.4586, "step": 6680 }, { "epoch": 0.27, "learning_rate": 0.0004937698780144128, "loss": 2.492, "step": 6690 }, { "epoch": 0.27, "learning_rate": 0.0004937598131969886, "loss": 2.3589, "step": 6700 }, { "epoch": 0.27, "learning_rate": 0.0004937497483795644, "loss": 2.5541, "step": 6710 }, { "epoch": 0.27, "learning_rate": 0.0004937396835621403, "loss": 2.5072, "step": 6720 }, { "epoch": 0.27, "learning_rate": 0.000493729618744716, "loss": 2.5264, "step": 6730 }, { "epoch": 0.27, "learning_rate": 0.0004937195539272918, "loss": 2.6815, "step": 6740 }, { "epoch": 0.27, "learning_rate": 0.0004937094891098676, "loss": 2.4314, "step": 6750 }, { "epoch": 0.27, "learning_rate": 0.0004936994242924433, "loss": 2.9705, "step": 6760 }, { "epoch": 0.27, "learning_rate": 0.0004936893594750192, "loss": 2.6916, "step": 6770 }, { "epoch": 0.27, "learning_rate": 0.0004936792946575949, "loss": 2.3482, "step": 6780 }, { "epoch": 0.27, "learning_rate": 0.0004936692298401707, "loss": 2.5663, "step": 6790 }, { "epoch": 0.27, "learning_rate": 0.0004936591650227465, "loss": 2.516, "step": 6800 }, { "epoch": 0.27, "learning_rate": 0.0004936491002053223, "loss": 2.6602, "step": 6810 }, { "epoch": 0.27, "learning_rate": 0.000493639035387898, "loss": 2.2737, "step": 6820 }, { "epoch": 0.27, "learning_rate": 0.0004936289705704739, "loss": 2.7228, "step": 6830 }, { "epoch": 0.28, "learning_rate": 0.0004936189057530496, "loss": 2.443, "step": 6840 }, { "epoch": 0.28, "learning_rate": 0.0004936088409356254, "loss": 2.5734, "step": 6850 }, { "epoch": 0.28, "learning_rate": 0.0004935987761182012, "loss": 2.3272, "step": 6860 }, { "epoch": 0.28, "learning_rate": 0.0004935887113007771, "loss": 2.4291, "step": 6870 }, { "epoch": 0.28, "learning_rate": 0.0004935786464833528, "loss": 2.2391, "step": 6880 }, { "epoch": 0.28, "learning_rate": 0.0004935685816659286, "loss": 2.4931, "step": 6890 }, { "epoch": 0.28, "learning_rate": 0.0004935585168485044, "loss": 2.5023, "step": 6900 }, { "epoch": 0.28, "learning_rate": 0.0004935484520310802, "loss": 2.3698, "step": 6910 }, { "epoch": 0.28, "learning_rate": 0.000493538387213656, "loss": 2.6116, "step": 6920 }, { "epoch": 0.28, "learning_rate": 0.0004935283223962317, "loss": 2.4749, "step": 6930 }, { "epoch": 0.28, "learning_rate": 0.0004935182575788075, "loss": 2.516, "step": 6940 }, { "epoch": 0.28, "learning_rate": 0.0004935081927613833, "loss": 2.2767, "step": 6950 }, { "epoch": 0.28, "learning_rate": 0.0004934981279439591, "loss": 2.4349, "step": 6960 }, { "epoch": 0.28, "learning_rate": 0.0004934880631265349, "loss": 2.4443, "step": 6970 }, { "epoch": 0.28, "learning_rate": 0.0004934779983091107, "loss": 2.5793, "step": 6980 }, { "epoch": 0.28, "learning_rate": 0.0004934679334916864, "loss": 2.442, "step": 6990 }, { "epoch": 0.28, "learning_rate": 0.0004934578686742623, "loss": 2.6277, "step": 7000 }, { "epoch": 0.28, "learning_rate": 0.000493447803856838, "loss": 2.6244, "step": 7010 }, { "epoch": 0.28, "learning_rate": 0.0004934377390394139, "loss": 2.2145, "step": 7020 }, { "epoch": 0.28, "learning_rate": 0.0004934276742219896, "loss": 3.0195, "step": 7030 }, { "epoch": 0.28, "learning_rate": 0.0004934176094045654, "loss": 2.4859, "step": 7040 }, { "epoch": 0.28, "learning_rate": 0.0004934075445871412, "loss": 2.2858, "step": 7050 }, { "epoch": 0.28, "learning_rate": 0.000493397479769717, "loss": 2.47, "step": 7060 }, { "epoch": 0.28, "learning_rate": 0.0004933874149522928, "loss": 2.5075, "step": 7070 }, { "epoch": 0.28, "learning_rate": 0.0004933773501348686, "loss": 2.6703, "step": 7080 }, { "epoch": 0.29, "learning_rate": 0.0004933672853174443, "loss": 2.4712, "step": 7090 }, { "epoch": 0.29, "learning_rate": 0.0004933572205000201, "loss": 2.2656, "step": 7100 }, { "epoch": 0.29, "learning_rate": 0.0004933471556825959, "loss": 2.4361, "step": 7110 }, { "epoch": 0.29, "learning_rate": 0.0004933370908651717, "loss": 2.5203, "step": 7120 }, { "epoch": 0.29, "learning_rate": 0.0004933270260477475, "loss": 2.1865, "step": 7130 }, { "epoch": 0.29, "learning_rate": 0.0004933169612303232, "loss": 2.5076, "step": 7140 }, { "epoch": 0.29, "learning_rate": 0.0004933068964128991, "loss": 2.3325, "step": 7150 }, { "epoch": 0.29, "learning_rate": 0.0004932968315954749, "loss": 2.3499, "step": 7160 }, { "epoch": 0.29, "learning_rate": 0.0004932867667780507, "loss": 2.5744, "step": 7170 }, { "epoch": 0.29, "learning_rate": 0.0004932767019606264, "loss": 2.5358, "step": 7180 }, { "epoch": 0.29, "learning_rate": 0.0004932666371432023, "loss": 2.5551, "step": 7190 }, { "epoch": 0.29, "learning_rate": 0.000493256572325778, "loss": 1.9699, "step": 7200 }, { "epoch": 0.29, "learning_rate": 0.0004932465075083538, "loss": 2.585, "step": 7210 }, { "epoch": 0.29, "learning_rate": 0.0004932364426909296, "loss": 2.5313, "step": 7220 }, { "epoch": 0.29, "learning_rate": 0.0004932263778735054, "loss": 2.4071, "step": 7230 }, { "epoch": 0.29, "learning_rate": 0.0004932163130560812, "loss": 2.4967, "step": 7240 }, { "epoch": 0.29, "learning_rate": 0.000493206248238657, "loss": 2.324, "step": 7250 }, { "epoch": 0.29, "learning_rate": 0.0004931961834212327, "loss": 2.6827, "step": 7260 }, { "epoch": 0.29, "learning_rate": 0.0004931861186038085, "loss": 2.501, "step": 7270 }, { "epoch": 0.29, "learning_rate": 0.0004931760537863843, "loss": 2.3014, "step": 7280 }, { "epoch": 0.29, "learning_rate": 0.0004931659889689601, "loss": 2.4266, "step": 7290 }, { "epoch": 0.29, "learning_rate": 0.0004931559241515359, "loss": 2.6119, "step": 7300 }, { "epoch": 0.29, "learning_rate": 0.0004931458593341117, "loss": 2.3546, "step": 7310 }, { "epoch": 0.29, "learning_rate": 0.0004931357945166875, "loss": 2.4759, "step": 7320 }, { "epoch": 0.29, "learning_rate": 0.0004931257296992633, "loss": 2.5739, "step": 7330 }, { "epoch": 0.3, "learning_rate": 0.0004931156648818391, "loss": 2.7196, "step": 7340 }, { "epoch": 0.3, "learning_rate": 0.0004931056000644148, "loss": 2.5023, "step": 7350 }, { "epoch": 0.3, "learning_rate": 0.0004930955352469906, "loss": 2.3362, "step": 7360 }, { "epoch": 0.3, "learning_rate": 0.0004930854704295664, "loss": 2.2402, "step": 7370 }, { "epoch": 0.3, "learning_rate": 0.0004930754056121422, "loss": 2.3844, "step": 7380 }, { "epoch": 0.3, "learning_rate": 0.000493065340794718, "loss": 2.4189, "step": 7390 }, { "epoch": 0.3, "learning_rate": 0.0004930552759772938, "loss": 2.5554, "step": 7400 }, { "epoch": 0.3, "learning_rate": 0.0004930452111598695, "loss": 2.3731, "step": 7410 }, { "epoch": 0.3, "learning_rate": 0.0004930351463424454, "loss": 2.4788, "step": 7420 }, { "epoch": 0.3, "learning_rate": 0.0004930250815250211, "loss": 2.3784, "step": 7430 }, { "epoch": 0.3, "learning_rate": 0.0004930150167075969, "loss": 2.3681, "step": 7440 }, { "epoch": 0.3, "learning_rate": 0.0004930049518901727, "loss": 2.3962, "step": 7450 }, { "epoch": 0.3, "learning_rate": 0.0004929948870727486, "loss": 2.4445, "step": 7460 }, { "epoch": 0.3, "learning_rate": 0.0004929848222553243, "loss": 2.5744, "step": 7470 }, { "epoch": 0.3, "learning_rate": 0.0004929747574379001, "loss": 2.5413, "step": 7480 }, { "epoch": 0.3, "learning_rate": 0.0004929646926204759, "loss": 2.3031, "step": 7490 }, { "epoch": 0.3, "learning_rate": 0.0004929546278030517, "loss": 2.5762, "step": 7500 }, { "epoch": 0.3, "learning_rate": 0.0004929445629856275, "loss": 2.471, "step": 7510 }, { "epoch": 0.3, "learning_rate": 0.0004929344981682032, "loss": 2.387, "step": 7520 }, { "epoch": 0.3, "learning_rate": 0.000492924433350779, "loss": 2.3443, "step": 7530 }, { "epoch": 0.3, "learning_rate": 0.0004929143685333548, "loss": 2.4017, "step": 7540 }, { "epoch": 0.3, "learning_rate": 0.0004929043037159306, "loss": 2.541, "step": 7550 }, { "epoch": 0.3, "learning_rate": 0.0004928942388985064, "loss": 2.6132, "step": 7560 }, { "epoch": 0.3, "learning_rate": 0.0004928841740810822, "loss": 2.3172, "step": 7570 }, { "epoch": 0.3, "learning_rate": 0.0004928741092636579, "loss": 2.55, "step": 7580 }, { "epoch": 0.31, "learning_rate": 0.0004928640444462338, "loss": 2.5584, "step": 7590 }, { "epoch": 0.31, "learning_rate": 0.0004928539796288095, "loss": 2.3453, "step": 7600 }, { "epoch": 0.31, "learning_rate": 0.0004928439148113854, "loss": 2.3702, "step": 7610 }, { "epoch": 0.31, "learning_rate": 0.0004928338499939611, "loss": 2.1864, "step": 7620 }, { "epoch": 0.31, "learning_rate": 0.0004928237851765369, "loss": 2.4291, "step": 7630 }, { "epoch": 0.31, "learning_rate": 0.0004928137203591127, "loss": 2.515, "step": 7640 }, { "epoch": 0.31, "learning_rate": 0.0004928036555416885, "loss": 2.102, "step": 7650 }, { "epoch": 0.31, "learning_rate": 0.0004927935907242643, "loss": 2.5789, "step": 7660 }, { "epoch": 0.31, "learning_rate": 0.0004927835259068401, "loss": 2.4497, "step": 7670 }, { "epoch": 0.31, "learning_rate": 0.0004927734610894158, "loss": 2.6077, "step": 7680 }, { "epoch": 0.31, "learning_rate": 0.0004927633962719916, "loss": 2.3852, "step": 7690 }, { "epoch": 0.31, "learning_rate": 0.0004927533314545674, "loss": 2.425, "step": 7700 }, { "epoch": 0.31, "learning_rate": 0.0004927432666371432, "loss": 2.3766, "step": 7710 }, { "epoch": 0.31, "learning_rate": 0.000492733201819719, "loss": 2.3403, "step": 7720 }, { "epoch": 0.31, "learning_rate": 0.0004927231370022947, "loss": 2.3715, "step": 7730 }, { "epoch": 0.31, "learning_rate": 0.0004927130721848706, "loss": 2.4242, "step": 7740 }, { "epoch": 0.31, "learning_rate": 0.0004927030073674464, "loss": 2.2349, "step": 7750 }, { "epoch": 0.31, "learning_rate": 0.0004926929425500222, "loss": 2.4143, "step": 7760 }, { "epoch": 0.31, "learning_rate": 0.0004926828777325979, "loss": 2.5214, "step": 7770 }, { "epoch": 0.31, "learning_rate": 0.0004926728129151737, "loss": 2.3148, "step": 7780 }, { "epoch": 0.31, "learning_rate": 0.0004926627480977495, "loss": 2.369, "step": 7790 }, { "epoch": 0.31, "learning_rate": 0.0004926526832803253, "loss": 2.1197, "step": 7800 }, { "epoch": 0.31, "learning_rate": 0.0004926426184629011, "loss": 2.173, "step": 7810 }, { "epoch": 0.31, "learning_rate": 0.0004926325536454769, "loss": 2.3753, "step": 7820 }, { "epoch": 0.31, "learning_rate": 0.0004926224888280526, "loss": 2.2626, "step": 7830 }, { "epoch": 0.32, "learning_rate": 0.0004926124240106285, "loss": 2.1603, "step": 7840 }, { "epoch": 0.32, "learning_rate": 0.0004926023591932042, "loss": 2.5039, "step": 7850 }, { "epoch": 0.32, "learning_rate": 0.00049259229437578, "loss": 2.233, "step": 7860 }, { "epoch": 0.32, "learning_rate": 0.0004925822295583558, "loss": 2.2497, "step": 7870 }, { "epoch": 0.32, "learning_rate": 0.0004925721647409315, "loss": 2.7173, "step": 7880 }, { "epoch": 0.32, "learning_rate": 0.0004925620999235074, "loss": 2.4346, "step": 7890 }, { "epoch": 0.32, "learning_rate": 0.0004925520351060832, "loss": 2.2807, "step": 7900 }, { "epoch": 0.32, "learning_rate": 0.000492541970288659, "loss": 2.1372, "step": 7910 }, { "epoch": 0.32, "learning_rate": 0.0004925319054712348, "loss": 2.5013, "step": 7920 }, { "epoch": 0.32, "learning_rate": 0.0004925218406538106, "loss": 2.1812, "step": 7930 }, { "epoch": 0.32, "learning_rate": 0.0004925117758363863, "loss": 2.2554, "step": 7940 }, { "epoch": 0.32, "learning_rate": 0.0004925017110189621, "loss": 2.346, "step": 7950 }, { "epoch": 0.32, "learning_rate": 0.0004924916462015379, "loss": 2.245, "step": 7960 }, { "epoch": 0.32, "learning_rate": 0.0004924815813841137, "loss": 2.3885, "step": 7970 }, { "epoch": 0.32, "learning_rate": 0.0004924715165666895, "loss": 2.5119, "step": 7980 }, { "epoch": 0.32, "learning_rate": 0.0004924614517492653, "loss": 2.2856, "step": 7990 }, { "epoch": 0.32, "learning_rate": 0.000492451386931841, "loss": 2.3597, "step": 8000 }, { "epoch": 0.32, "learning_rate": 0.0004924413221144169, "loss": 2.0986, "step": 8010 }, { "epoch": 0.32, "learning_rate": 0.0004924312572969926, "loss": 2.2249, "step": 8020 }, { "epoch": 0.32, "learning_rate": 0.0004924211924795685, "loss": 2.3988, "step": 8030 }, { "epoch": 0.32, "learning_rate": 0.0004924111276621442, "loss": 2.1121, "step": 8040 }, { "epoch": 0.32, "learning_rate": 0.00049240106284472, "loss": 2.2644, "step": 8050 }, { "epoch": 0.32, "learning_rate": 0.0004923909980272958, "loss": 2.2618, "step": 8060 }, { "epoch": 0.32, "learning_rate": 0.0004923809332098716, "loss": 2.5005, "step": 8070 }, { "epoch": 0.32, "learning_rate": 0.0004923708683924474, "loss": 2.1199, "step": 8080 }, { "epoch": 0.33, "learning_rate": 0.0004923608035750232, "loss": 2.4968, "step": 8090 }, { "epoch": 0.33, "learning_rate": 0.000492350738757599, "loss": 2.2024, "step": 8100 }, { "epoch": 0.33, "learning_rate": 0.0004923406739401747, "loss": 2.4467, "step": 8110 }, { "epoch": 0.33, "learning_rate": 0.0004923306091227505, "loss": 2.2767, "step": 8120 }, { "epoch": 0.33, "learning_rate": 0.0004923205443053263, "loss": 2.2488, "step": 8130 }, { "epoch": 0.33, "learning_rate": 0.0004923104794879021, "loss": 2.2506, "step": 8140 }, { "epoch": 0.33, "learning_rate": 0.0004923004146704778, "loss": 2.3258, "step": 8150 }, { "epoch": 0.33, "learning_rate": 0.0004922903498530537, "loss": 2.6362, "step": 8160 }, { "epoch": 0.33, "learning_rate": 0.0004922802850356294, "loss": 2.5746, "step": 8170 }, { "epoch": 0.33, "learning_rate": 0.0004922702202182053, "loss": 2.543, "step": 8180 }, { "epoch": 0.33, "learning_rate": 0.000492260155400781, "loss": 2.3461, "step": 8190 }, { "epoch": 0.33, "learning_rate": 0.0004922500905833569, "loss": 2.2811, "step": 8200 }, { "epoch": 0.33, "learning_rate": 0.0004922400257659326, "loss": 2.542, "step": 8210 }, { "epoch": 0.33, "learning_rate": 0.0004922299609485084, "loss": 2.2247, "step": 8220 }, { "epoch": 0.33, "learning_rate": 0.0004922198961310842, "loss": 2.4181, "step": 8230 }, { "epoch": 0.33, "learning_rate": 0.00049220983131366, "loss": 2.1891, "step": 8240 }, { "epoch": 0.33, "learning_rate": 0.0004921997664962358, "loss": 2.2038, "step": 8250 }, { "epoch": 0.33, "learning_rate": 0.0004921897016788116, "loss": 2.4079, "step": 8260 }, { "epoch": 0.33, "learning_rate": 0.0004921796368613873, "loss": 2.3315, "step": 8270 }, { "epoch": 0.33, "learning_rate": 0.0004921695720439631, "loss": 2.2933, "step": 8280 }, { "epoch": 0.33, "learning_rate": 0.0004921595072265389, "loss": 2.4307, "step": 8290 }, { "epoch": 0.33, "learning_rate": 0.0004921494424091147, "loss": 2.5057, "step": 8300 }, { "epoch": 0.33, "learning_rate": 0.0004921393775916905, "loss": 2.1871, "step": 8310 }, { "epoch": 0.33, "learning_rate": 0.0004921293127742662, "loss": 2.404, "step": 8320 }, { "epoch": 0.34, "learning_rate": 0.0004921192479568421, "loss": 2.2285, "step": 8330 }, { "epoch": 0.34, "learning_rate": 0.0004921091831394179, "loss": 2.3621, "step": 8340 }, { "epoch": 0.34, "learning_rate": 0.0004920991183219937, "loss": 2.0646, "step": 8350 }, { "epoch": 0.34, "learning_rate": 0.0004920890535045694, "loss": 2.2767, "step": 8360 }, { "epoch": 0.34, "learning_rate": 0.0004920789886871452, "loss": 2.5238, "step": 8370 }, { "epoch": 0.34, "learning_rate": 0.000492068923869721, "loss": 2.5591, "step": 8380 }, { "epoch": 0.34, "learning_rate": 0.0004920588590522968, "loss": 2.3224, "step": 8390 }, { "epoch": 0.34, "learning_rate": 0.0004920487942348726, "loss": 2.4829, "step": 8400 }, { "epoch": 0.34, "learning_rate": 0.0004920387294174484, "loss": 2.3757, "step": 8410 }, { "epoch": 0.34, "learning_rate": 0.0004920286646000241, "loss": 2.3735, "step": 8420 }, { "epoch": 0.34, "learning_rate": 0.0004920185997826, "loss": 2.1321, "step": 8430 }, { "epoch": 0.34, "learning_rate": 0.0004920085349651757, "loss": 2.4258, "step": 8440 }, { "epoch": 0.34, "learning_rate": 0.0004919984701477515, "loss": 2.374, "step": 8450 }, { "epoch": 0.34, "learning_rate": 0.0004919884053303273, "loss": 2.2347, "step": 8460 }, { "epoch": 0.34, "learning_rate": 0.0004919783405129032, "loss": 2.3047, "step": 8470 }, { "epoch": 0.34, "learning_rate": 0.0004919682756954789, "loss": 2.2334, "step": 8480 }, { "epoch": 0.34, "learning_rate": 0.0004919582108780547, "loss": 2.4211, "step": 8490 }, { "epoch": 0.34, "learning_rate": 0.0004919481460606305, "loss": 2.6326, "step": 8500 }, { "epoch": 0.34, "learning_rate": 0.0004919380812432063, "loss": 2.5208, "step": 8510 }, { "epoch": 0.34, "learning_rate": 0.000491928016425782, "loss": 2.5628, "step": 8520 }, { "epoch": 0.34, "learning_rate": 0.0004919179516083578, "loss": 2.4157, "step": 8530 }, { "epoch": 0.34, "learning_rate": 0.0004919078867909336, "loss": 2.3302, "step": 8540 }, { "epoch": 0.34, "learning_rate": 0.0004918978219735094, "loss": 2.2917, "step": 8550 }, { "epoch": 0.34, "learning_rate": 0.0004918877571560852, "loss": 2.2953, "step": 8560 }, { "epoch": 0.34, "learning_rate": 0.000491877692338661, "loss": 2.4308, "step": 8570 }, { "epoch": 0.35, "learning_rate": 0.0004918676275212368, "loss": 2.1004, "step": 8580 }, { "epoch": 0.35, "learning_rate": 0.0004918575627038125, "loss": 2.3588, "step": 8590 }, { "epoch": 0.35, "learning_rate": 0.0004918474978863884, "loss": 2.6143, "step": 8600 }, { "epoch": 0.35, "learning_rate": 0.0004918374330689641, "loss": 2.4716, "step": 8610 }, { "epoch": 0.35, "learning_rate": 0.00049182736825154, "loss": 2.3831, "step": 8620 }, { "epoch": 0.35, "learning_rate": 0.0004918173034341157, "loss": 2.1031, "step": 8630 }, { "epoch": 0.35, "learning_rate": 0.0004918072386166915, "loss": 2.4872, "step": 8640 }, { "epoch": 0.35, "learning_rate": 0.0004917971737992673, "loss": 2.394, "step": 8650 }, { "epoch": 0.35, "learning_rate": 0.0004917871089818431, "loss": 2.4447, "step": 8660 }, { "epoch": 0.35, "learning_rate": 0.0004917770441644189, "loss": 2.4611, "step": 8670 }, { "epoch": 0.35, "learning_rate": 0.0004917669793469947, "loss": 2.2007, "step": 8680 }, { "epoch": 0.35, "learning_rate": 0.0004917569145295704, "loss": 2.5684, "step": 8690 }, { "epoch": 0.35, "learning_rate": 0.0004917468497121462, "loss": 1.9168, "step": 8700 }, { "epoch": 0.35, "learning_rate": 0.000491736784894722, "loss": 2.3424, "step": 8710 }, { "epoch": 0.35, "learning_rate": 0.0004917267200772978, "loss": 2.3632, "step": 8720 }, { "epoch": 0.35, "learning_rate": 0.0004917166552598736, "loss": 2.3249, "step": 8730 }, { "epoch": 0.35, "learning_rate": 0.0004917065904424493, "loss": 2.2782, "step": 8740 }, { "epoch": 0.35, "learning_rate": 0.0004916965256250252, "loss": 2.2461, "step": 8750 }, { "epoch": 0.35, "learning_rate": 0.0004916864608076009, "loss": 2.2951, "step": 8760 }, { "epoch": 0.35, "learning_rate": 0.0004916763959901768, "loss": 2.279, "step": 8770 }, { "epoch": 0.35, "learning_rate": 0.0004916663311727526, "loss": 2.5309, "step": 8780 }, { "epoch": 0.35, "learning_rate": 0.0004916562663553283, "loss": 2.1996, "step": 8790 }, { "epoch": 0.35, "learning_rate": 0.0004916462015379041, "loss": 2.2941, "step": 8800 }, { "epoch": 0.35, "learning_rate": 0.0004916361367204799, "loss": 2.055, "step": 8810 }, { "epoch": 0.35, "learning_rate": 0.0004916260719030557, "loss": 2.404, "step": 8820 }, { "epoch": 0.36, "learning_rate": 0.0004916160070856315, "loss": 2.217, "step": 8830 }, { "epoch": 0.36, "learning_rate": 0.0004916059422682072, "loss": 2.1873, "step": 8840 }, { "epoch": 0.36, "learning_rate": 0.0004915958774507831, "loss": 2.3694, "step": 8850 }, { "epoch": 0.36, "learning_rate": 0.0004915858126333588, "loss": 2.2926, "step": 8860 }, { "epoch": 0.36, "learning_rate": 0.0004915757478159346, "loss": 2.3942, "step": 8870 }, { "epoch": 0.36, "learning_rate": 0.0004915656829985104, "loss": 2.2458, "step": 8880 }, { "epoch": 0.36, "learning_rate": 0.0004915556181810861, "loss": 2.2844, "step": 8890 }, { "epoch": 0.36, "learning_rate": 0.000491545553363662, "loss": 2.2027, "step": 8900 }, { "epoch": 0.36, "learning_rate": 0.0004915354885462378, "loss": 2.4155, "step": 8910 }, { "epoch": 0.36, "learning_rate": 0.0004915254237288136, "loss": 2.279, "step": 8920 }, { "epoch": 0.36, "learning_rate": 0.0004915153589113894, "loss": 2.5484, "step": 8930 }, { "epoch": 0.36, "learning_rate": 0.0004915052940939652, "loss": 2.2881, "step": 8940 }, { "epoch": 0.36, "learning_rate": 0.0004914952292765409, "loss": 2.1909, "step": 8950 }, { "epoch": 0.36, "learning_rate": 0.0004914851644591167, "loss": 2.2659, "step": 8960 }, { "epoch": 0.36, "learning_rate": 0.0004914750996416925, "loss": 2.1849, "step": 8970 }, { "epoch": 0.36, "learning_rate": 0.0004914650348242683, "loss": 2.3004, "step": 8980 }, { "epoch": 0.36, "learning_rate": 0.0004914549700068441, "loss": 2.3298, "step": 8990 }, { "epoch": 0.36, "learning_rate": 0.0004914449051894199, "loss": 2.2865, "step": 9000 }, { "epoch": 0.36, "learning_rate": 0.0004914348403719956, "loss": 2.1817, "step": 9010 }, { "epoch": 0.36, "learning_rate": 0.0004914247755545715, "loss": 1.982, "step": 9020 }, { "epoch": 0.36, "learning_rate": 0.0004914147107371472, "loss": 2.5559, "step": 9030 }, { "epoch": 0.36, "learning_rate": 0.000491404645919723, "loss": 2.2968, "step": 9040 }, { "epoch": 0.36, "learning_rate": 0.0004913945811022988, "loss": 2.4041, "step": 9050 }, { "epoch": 0.36, "learning_rate": 0.0004913845162848746, "loss": 2.3396, "step": 9060 }, { "epoch": 0.36, "learning_rate": 0.0004913744514674504, "loss": 2.3593, "step": 9070 }, { "epoch": 0.37, "learning_rate": 0.0004913643866500262, "loss": 2.3778, "step": 9080 }, { "epoch": 0.37, "learning_rate": 0.000491354321832602, "loss": 2.1736, "step": 9090 }, { "epoch": 0.37, "learning_rate": 0.0004913442570151778, "loss": 2.1308, "step": 9100 }, { "epoch": 0.37, "learning_rate": 0.0004913341921977535, "loss": 2.1716, "step": 9110 }, { "epoch": 0.37, "learning_rate": 0.0004913241273803293, "loss": 2.2161, "step": 9120 }, { "epoch": 0.37, "learning_rate": 0.0004913140625629051, "loss": 2.3516, "step": 9130 }, { "epoch": 0.37, "learning_rate": 0.0004913039977454809, "loss": 1.9502, "step": 9140 }, { "epoch": 0.37, "learning_rate": 0.0004912939329280567, "loss": 2.2894, "step": 9150 }, { "epoch": 0.37, "learning_rate": 0.0004912838681106324, "loss": 2.208, "step": 9160 }, { "epoch": 0.37, "learning_rate": 0.0004912738032932083, "loss": 2.3593, "step": 9170 }, { "epoch": 0.37, "learning_rate": 0.000491263738475784, "loss": 2.1145, "step": 9180 }, { "epoch": 0.37, "learning_rate": 0.0004912536736583599, "loss": 2.1143, "step": 9190 }, { "epoch": 0.37, "learning_rate": 0.0004912436088409357, "loss": 2.1958, "step": 9200 }, { "epoch": 0.37, "learning_rate": 0.0004912335440235115, "loss": 2.4044, "step": 9210 }, { "epoch": 0.37, "learning_rate": 0.0004912234792060872, "loss": 2.2699, "step": 9220 }, { "epoch": 0.37, "learning_rate": 0.000491213414388663, "loss": 2.3393, "step": 9230 }, { "epoch": 0.37, "learning_rate": 0.0004912033495712388, "loss": 2.1848, "step": 9240 }, { "epoch": 0.37, "learning_rate": 0.0004911932847538146, "loss": 2.0695, "step": 9250 }, { "epoch": 0.37, "learning_rate": 0.0004911832199363904, "loss": 2.1712, "step": 9260 }, { "epoch": 0.37, "learning_rate": 0.0004911731551189662, "loss": 2.4471, "step": 9270 }, { "epoch": 0.37, "learning_rate": 0.0004911630903015419, "loss": 2.194, "step": 9280 }, { "epoch": 0.37, "learning_rate": 0.0004911530254841177, "loss": 2.3332, "step": 9290 }, { "epoch": 0.37, "learning_rate": 0.0004911429606666935, "loss": 1.9933, "step": 9300 }, { "epoch": 0.37, "learning_rate": 0.0004911328958492693, "loss": 2.213, "step": 9310 }, { "epoch": 0.37, "learning_rate": 0.0004911228310318451, "loss": 2.313, "step": 9320 }, { "epoch": 0.38, "learning_rate": 0.0004911127662144208, "loss": 2.3731, "step": 9330 }, { "epoch": 0.38, "learning_rate": 0.0004911027013969967, "loss": 2.1257, "step": 9340 }, { "epoch": 0.38, "learning_rate": 0.0004910926365795725, "loss": 2.1567, "step": 9350 }, { "epoch": 0.38, "learning_rate": 0.0004910825717621483, "loss": 2.5776, "step": 9360 }, { "epoch": 0.38, "learning_rate": 0.0004910725069447241, "loss": 2.2349, "step": 9370 }, { "epoch": 0.38, "learning_rate": 0.0004910624421272998, "loss": 2.2219, "step": 9380 }, { "epoch": 0.38, "learning_rate": 0.0004910523773098756, "loss": 2.2955, "step": 9390 }, { "epoch": 0.38, "learning_rate": 0.0004910423124924514, "loss": 1.8715, "step": 9400 }, { "epoch": 0.38, "learning_rate": 0.0004910322476750272, "loss": 2.2134, "step": 9410 }, { "epoch": 0.38, "learning_rate": 0.000491022182857603, "loss": 2.3646, "step": 9420 }, { "epoch": 0.38, "learning_rate": 0.0004910121180401787, "loss": 2.2502, "step": 9430 }, { "epoch": 0.38, "learning_rate": 0.0004910020532227546, "loss": 2.0158, "step": 9440 }, { "epoch": 0.38, "learning_rate": 0.0004909919884053303, "loss": 2.5502, "step": 9450 }, { "epoch": 0.38, "learning_rate": 0.0004909819235879061, "loss": 2.2385, "step": 9460 }, { "epoch": 0.38, "learning_rate": 0.0004909718587704819, "loss": 2.33, "step": 9470 }, { "epoch": 0.38, "learning_rate": 0.0004909617939530576, "loss": 1.7928, "step": 9480 }, { "epoch": 0.38, "learning_rate": 0.0004909517291356335, "loss": 2.3433, "step": 9490 }, { "epoch": 0.38, "learning_rate": 0.0004909416643182093, "loss": 2.22, "step": 9500 }, { "epoch": 0.38, "learning_rate": 0.0004909315995007851, "loss": 2.4014, "step": 9510 }, { "epoch": 0.38, "learning_rate": 0.0004909215346833609, "loss": 2.221, "step": 9520 }, { "epoch": 0.38, "learning_rate": 0.0004909114698659366, "loss": 2.179, "step": 9530 }, { "epoch": 0.38, "learning_rate": 0.0004909014050485125, "loss": 2.0783, "step": 9540 }, { "epoch": 0.38, "learning_rate": 0.0004908913402310882, "loss": 2.0789, "step": 9550 }, { "epoch": 0.38, "learning_rate": 0.000490881275413664, "loss": 2.2039, "step": 9560 }, { "epoch": 0.38, "learning_rate": 0.0004908712105962398, "loss": 2.6313, "step": 9570 }, { "epoch": 0.39, "learning_rate": 0.0004908611457788155, "loss": 2.1243, "step": 9580 }, { "epoch": 0.39, "learning_rate": 0.0004908510809613914, "loss": 2.208, "step": 9590 }, { "epoch": 0.39, "learning_rate": 0.0004908410161439671, "loss": 1.9917, "step": 9600 }, { "epoch": 0.39, "learning_rate": 0.000490830951326543, "loss": 2.2798, "step": 9610 }, { "epoch": 0.39, "learning_rate": 0.0004908208865091187, "loss": 2.263, "step": 9620 }, { "epoch": 0.39, "learning_rate": 0.0004908108216916944, "loss": 2.1527, "step": 9630 }, { "epoch": 0.39, "learning_rate": 0.0004908007568742703, "loss": 2.262, "step": 9640 }, { "epoch": 0.39, "learning_rate": 0.0004907906920568461, "loss": 2.1024, "step": 9650 }, { "epoch": 0.39, "learning_rate": 0.0004907806272394219, "loss": 2.0496, "step": 9660 }, { "epoch": 0.39, "learning_rate": 0.0004907705624219977, "loss": 2.0043, "step": 9670 }, { "epoch": 0.39, "learning_rate": 0.0004907604976045735, "loss": 2.0987, "step": 9680 }, { "epoch": 0.39, "learning_rate": 0.0004907504327871493, "loss": 2.4551, "step": 9690 }, { "epoch": 0.39, "learning_rate": 0.000490740367969725, "loss": 2.3206, "step": 9700 }, { "epoch": 0.39, "learning_rate": 0.0004907303031523008, "loss": 2.1437, "step": 9710 }, { "epoch": 0.39, "learning_rate": 0.0004907202383348766, "loss": 1.7433, "step": 9720 }, { "epoch": 0.39, "learning_rate": 0.0004907101735174524, "loss": 1.8741, "step": 9730 }, { "epoch": 0.39, "learning_rate": 0.0004907001087000282, "loss": 2.3249, "step": 9740 }, { "epoch": 0.39, "learning_rate": 0.0004906900438826039, "loss": 2.2164, "step": 9750 }, { "epoch": 0.39, "learning_rate": 0.0004906799790651798, "loss": 2.3751, "step": 9760 }, { "epoch": 0.39, "learning_rate": 0.0004906699142477555, "loss": 2.0879, "step": 9770 }, { "epoch": 0.39, "learning_rate": 0.0004906598494303314, "loss": 1.9773, "step": 9780 }, { "epoch": 0.39, "learning_rate": 0.0004906497846129072, "loss": 2.3339, "step": 9790 }, { "epoch": 0.39, "learning_rate": 0.0004906397197954829, "loss": 2.4582, "step": 9800 }, { "epoch": 0.39, "learning_rate": 0.0004906296549780587, "loss": 2.2991, "step": 9810 }, { "epoch": 0.39, "learning_rate": 0.0004906195901606345, "loss": 2.321, "step": 9820 }, { "epoch": 0.4, "learning_rate": 0.0004906095253432103, "loss": 2.1508, "step": 9830 }, { "epoch": 0.4, "learning_rate": 0.0004905994605257861, "loss": 2.067, "step": 9840 }, { "epoch": 0.4, "learning_rate": 0.0004905893957083618, "loss": 2.1424, "step": 9850 }, { "epoch": 0.4, "learning_rate": 0.0004905793308909377, "loss": 2.1746, "step": 9860 }, { "epoch": 0.4, "learning_rate": 0.0004905692660735134, "loss": 2.3252, "step": 9870 }, { "epoch": 0.4, "learning_rate": 0.0004905592012560892, "loss": 2.0605, "step": 9880 }, { "epoch": 0.4, "learning_rate": 0.000490549136438665, "loss": 2.0683, "step": 9890 }, { "epoch": 0.4, "learning_rate": 0.0004905390716212407, "loss": 2.0821, "step": 9900 }, { "epoch": 0.4, "learning_rate": 0.0004905290068038166, "loss": 2.3987, "step": 9910 }, { "epoch": 0.4, "learning_rate": 0.0004905189419863923, "loss": 2.3039, "step": 9920 }, { "epoch": 0.4, "learning_rate": 0.0004905088771689682, "loss": 2.3865, "step": 9930 }, { "epoch": 0.4, "learning_rate": 0.000490498812351544, "loss": 2.2176, "step": 9940 }, { "epoch": 0.4, "learning_rate": 0.0004904887475341198, "loss": 2.2591, "step": 9950 }, { "epoch": 0.4, "learning_rate": 0.0004904786827166956, "loss": 2.307, "step": 9960 }, { "epoch": 0.4, "learning_rate": 0.0004904686178992713, "loss": 2.0173, "step": 9970 }, { "epoch": 0.4, "learning_rate": 0.0004904585530818471, "loss": 2.3171, "step": 9980 }, { "epoch": 0.4, "learning_rate": 0.0004904484882644229, "loss": 1.8602, "step": 9990 }, { "epoch": 0.4, "learning_rate": 0.0004904384234469987, "loss": 2.4278, "step": 10000 }, { "epoch": 0.4, "learning_rate": 0.0004904283586295745, "loss": 2.0543, "step": 10010 }, { "epoch": 0.4, "learning_rate": 0.0004904182938121502, "loss": 1.9762, "step": 10020 }, { "epoch": 0.4, "learning_rate": 0.0004904082289947261, "loss": 2.1327, "step": 10030 }, { "epoch": 0.4, "learning_rate": 0.0004903981641773018, "loss": 2.2153, "step": 10040 }, { "epoch": 0.4, "learning_rate": 0.0004903880993598776, "loss": 2.242, "step": 10050 }, { "epoch": 0.4, "learning_rate": 0.0004903780345424534, "loss": 2.0695, "step": 10060 }, { "epoch": 0.4, "learning_rate": 0.0004903679697250291, "loss": 2.0612, "step": 10070 }, { "epoch": 0.41, "learning_rate": 0.000490357904907605, "loss": 1.9479, "step": 10080 }, { "epoch": 0.41, "learning_rate": 0.0004903478400901808, "loss": 1.9894, "step": 10090 }, { "epoch": 0.41, "learning_rate": 0.0004903377752727566, "loss": 2.155, "step": 10100 }, { "epoch": 0.41, "learning_rate": 0.0004903277104553324, "loss": 2.1849, "step": 10110 }, { "epoch": 0.41, "learning_rate": 0.0004903176456379081, "loss": 2.1042, "step": 10120 }, { "epoch": 0.41, "learning_rate": 0.000490307580820484, "loss": 2.0932, "step": 10130 }, { "epoch": 0.41, "learning_rate": 0.0004902975160030597, "loss": 2.4404, "step": 10140 }, { "epoch": 0.41, "learning_rate": 0.0004902874511856355, "loss": 2.307, "step": 10150 }, { "epoch": 0.41, "learning_rate": 0.0004902773863682113, "loss": 2.4304, "step": 10160 }, { "epoch": 0.41, "learning_rate": 0.000490267321550787, "loss": 2.1231, "step": 10170 }, { "epoch": 0.41, "learning_rate": 0.0004902572567333629, "loss": 2.1831, "step": 10180 }, { "epoch": 0.41, "learning_rate": 0.0004902471919159386, "loss": 2.2814, "step": 10190 }, { "epoch": 0.41, "learning_rate": 0.0004902371270985145, "loss": 2.325, "step": 10200 }, { "epoch": 0.41, "learning_rate": 0.0004902270622810902, "loss": 2.1718, "step": 10210 }, { "epoch": 0.41, "learning_rate": 0.0004902169974636659, "loss": 1.9397, "step": 10220 }, { "epoch": 0.41, "learning_rate": 0.0004902069326462418, "loss": 2.3329, "step": 10230 }, { "epoch": 0.41, "learning_rate": 0.0004901968678288176, "loss": 2.1522, "step": 10240 }, { "epoch": 0.41, "learning_rate": 0.0004901868030113934, "loss": 2.2915, "step": 10250 }, { "epoch": 0.41, "learning_rate": 0.0004901767381939692, "loss": 2.2034, "step": 10260 }, { "epoch": 0.41, "learning_rate": 0.000490166673376545, "loss": 1.9563, "step": 10270 }, { "epoch": 0.41, "learning_rate": 0.0004901566085591208, "loss": 2.42, "step": 10280 }, { "epoch": 0.41, "learning_rate": 0.0004901465437416965, "loss": 2.1224, "step": 10290 }, { "epoch": 0.41, "learning_rate": 0.0004901364789242724, "loss": 2.2209, "step": 10300 }, { "epoch": 0.41, "learning_rate": 0.0004901264141068481, "loss": 1.8007, "step": 10310 }, { "epoch": 0.42, "learning_rate": 0.0004901163492894238, "loss": 1.9038, "step": 10320 }, { "epoch": 0.42, "learning_rate": 0.0004901062844719997, "loss": 2.2045, "step": 10330 }, { "epoch": 0.42, "learning_rate": 0.0004900962196545754, "loss": 2.5673, "step": 10340 }, { "epoch": 0.42, "learning_rate": 0.0004900861548371513, "loss": 2.5877, "step": 10350 }, { "epoch": 0.42, "learning_rate": 0.000490076090019727, "loss": 1.876, "step": 10360 }, { "epoch": 0.42, "learning_rate": 0.0004900660252023029, "loss": 2.1265, "step": 10370 }, { "epoch": 0.42, "learning_rate": 0.0004900559603848787, "loss": 2.0862, "step": 10380 }, { "epoch": 0.42, "learning_rate": 0.0004900458955674544, "loss": 2.1646, "step": 10390 }, { "epoch": 0.42, "learning_rate": 0.0004900358307500302, "loss": 2.225, "step": 10400 }, { "epoch": 0.42, "learning_rate": 0.000490025765932606, "loss": 1.9061, "step": 10410 }, { "epoch": 0.42, "learning_rate": 0.0004900157011151818, "loss": 2.0915, "step": 10420 }, { "epoch": 0.42, "learning_rate": 0.0004900056362977576, "loss": 2.0474, "step": 10430 }, { "epoch": 0.42, "learning_rate": 0.0004899955714803333, "loss": 2.0684, "step": 10440 }, { "epoch": 0.42, "learning_rate": 0.0004899855066629092, "loss": 2.3363, "step": 10450 }, { "epoch": 0.42, "learning_rate": 0.0004899754418454849, "loss": 2.1289, "step": 10460 }, { "epoch": 0.42, "learning_rate": 0.0004899653770280607, "loss": 2.2649, "step": 10470 }, { "epoch": 0.42, "learning_rate": 0.0004899553122106365, "loss": 2.1245, "step": 10480 }, { "epoch": 0.42, "learning_rate": 0.0004899452473932122, "loss": 2.2529, "step": 10490 }, { "epoch": 0.42, "learning_rate": 0.0004899351825757881, "loss": 2.0672, "step": 10500 }, { "epoch": 0.42, "learning_rate": 0.0004899251177583638, "loss": 2.3478, "step": 10510 }, { "epoch": 0.42, "learning_rate": 0.0004899150529409397, "loss": 2.2066, "step": 10520 }, { "epoch": 0.42, "learning_rate": 0.0004899049881235155, "loss": 2.0606, "step": 10530 }, { "epoch": 0.42, "learning_rate": 0.0004898949233060912, "loss": 2.1634, "step": 10540 }, { "epoch": 0.42, "learning_rate": 0.0004898848584886671, "loss": 2.113, "step": 10550 }, { "epoch": 0.42, "learning_rate": 0.0004898747936712428, "loss": 2.0042, "step": 10560 }, { "epoch": 0.43, "learning_rate": 0.0004898647288538186, "loss": 2.1195, "step": 10570 }, { "epoch": 0.43, "learning_rate": 0.0004898546640363944, "loss": 2.0584, "step": 10580 }, { "epoch": 0.43, "learning_rate": 0.0004898445992189701, "loss": 2.0794, "step": 10590 }, { "epoch": 0.43, "learning_rate": 0.000489834534401546, "loss": 2.0563, "step": 10600 }, { "epoch": 0.43, "learning_rate": 0.0004898244695841217, "loss": 2.1912, "step": 10610 }, { "epoch": 0.43, "learning_rate": 0.0004898144047666976, "loss": 1.9231, "step": 10620 }, { "epoch": 0.43, "learning_rate": 0.0004898043399492733, "loss": 1.95, "step": 10630 }, { "epoch": 0.43, "learning_rate": 0.000489794275131849, "loss": 2.2288, "step": 10640 }, { "epoch": 0.43, "learning_rate": 0.0004897842103144249, "loss": 2.0275, "step": 10650 }, { "epoch": 0.43, "learning_rate": 0.0004897741454970006, "loss": 2.0479, "step": 10660 }, { "epoch": 0.43, "learning_rate": 0.0004897640806795765, "loss": 2.0195, "step": 10670 }, { "epoch": 0.43, "learning_rate": 0.0004897540158621523, "loss": 2.3588, "step": 10680 }, { "epoch": 0.43, "learning_rate": 0.000489743951044728, "loss": 1.9931, "step": 10690 }, { "epoch": 0.43, "learning_rate": 0.0004897338862273039, "loss": 2.3159, "step": 10700 }, { "epoch": 0.43, "learning_rate": 0.0004897238214098796, "loss": 2.0517, "step": 10710 }, { "epoch": 0.43, "learning_rate": 0.0004897137565924555, "loss": 2.2703, "step": 10720 }, { "epoch": 0.43, "learning_rate": 0.0004897036917750312, "loss": 2.0634, "step": 10730 }, { "epoch": 0.43, "learning_rate": 0.000489693626957607, "loss": 2.11, "step": 10740 }, { "epoch": 0.43, "learning_rate": 0.0004896835621401828, "loss": 2.0561, "step": 10750 }, { "epoch": 0.43, "learning_rate": 0.0004896734973227585, "loss": 1.8435, "step": 10760 }, { "epoch": 0.43, "learning_rate": 0.0004896634325053344, "loss": 2.1758, "step": 10770 }, { "epoch": 0.43, "learning_rate": 0.0004896533676879101, "loss": 2.0039, "step": 10780 }, { "epoch": 0.43, "learning_rate": 0.000489643302870486, "loss": 2.2628, "step": 10790 }, { "epoch": 0.43, "learning_rate": 0.0004896332380530617, "loss": 2.2963, "step": 10800 }, { "epoch": 0.43, "learning_rate": 0.0004896231732356375, "loss": 2.0557, "step": 10810 }, { "epoch": 0.44, "learning_rate": 0.0004896131084182133, "loss": 2.1834, "step": 10820 }, { "epoch": 0.44, "learning_rate": 0.0004896030436007891, "loss": 2.1172, "step": 10830 }, { "epoch": 0.44, "learning_rate": 0.0004895929787833649, "loss": 2.204, "step": 10840 }, { "epoch": 0.44, "learning_rate": 0.0004895829139659407, "loss": 2.3621, "step": 10850 }, { "epoch": 0.44, "learning_rate": 0.0004895728491485164, "loss": 1.9346, "step": 10860 }, { "epoch": 0.44, "learning_rate": 0.0004895627843310923, "loss": 2.0509, "step": 10870 }, { "epoch": 0.44, "learning_rate": 0.000489552719513668, "loss": 2.0259, "step": 10880 }, { "epoch": 0.44, "learning_rate": 0.0004895426546962439, "loss": 1.9321, "step": 10890 }, { "epoch": 0.44, "learning_rate": 0.0004895325898788196, "loss": 2.1098, "step": 10900 }, { "epoch": 0.44, "learning_rate": 0.0004895225250613953, "loss": 2.0922, "step": 10910 }, { "epoch": 0.44, "learning_rate": 0.0004895124602439712, "loss": 2.0729, "step": 10920 }, { "epoch": 0.44, "learning_rate": 0.0004895023954265469, "loss": 2.1344, "step": 10930 }, { "epoch": 0.44, "learning_rate": 0.0004894923306091228, "loss": 1.8649, "step": 10940 }, { "epoch": 0.44, "learning_rate": 0.0004894822657916985, "loss": 1.9395, "step": 10950 }, { "epoch": 0.44, "learning_rate": 0.0004894722009742744, "loss": 2.0547, "step": 10960 }, { "epoch": 0.44, "learning_rate": 0.0004894621361568502, "loss": 2.1718, "step": 10970 }, { "epoch": 0.44, "learning_rate": 0.0004894520713394259, "loss": 2.0574, "step": 10980 }, { "epoch": 0.44, "learning_rate": 0.0004894420065220017, "loss": 2.4383, "step": 10990 }, { "epoch": 0.44, "learning_rate": 0.0004894319417045775, "loss": 2.1409, "step": 11000 }, { "epoch": 0.44, "learning_rate": 0.0004894218768871533, "loss": 2.0225, "step": 11010 }, { "epoch": 0.44, "learning_rate": 0.0004894118120697291, "loss": 2.1119, "step": 11020 }, { "epoch": 0.44, "learning_rate": 0.0004894017472523048, "loss": 2.0663, "step": 11030 }, { "epoch": 0.44, "learning_rate": 0.0004893916824348807, "loss": 2.0266, "step": 11040 }, { "epoch": 0.44, "learning_rate": 0.0004893816176174564, "loss": 2.0785, "step": 11050 }, { "epoch": 0.44, "learning_rate": 0.0004893715528000323, "loss": 1.7514, "step": 11060 }, { "epoch": 0.45, "learning_rate": 0.000489361487982608, "loss": 1.9264, "step": 11070 }, { "epoch": 0.45, "learning_rate": 0.0004893514231651837, "loss": 2.3436, "step": 11080 }, { "epoch": 0.45, "learning_rate": 0.0004893413583477596, "loss": 2.2181, "step": 11090 }, { "epoch": 0.45, "learning_rate": 0.0004893312935303354, "loss": 1.9963, "step": 11100 }, { "epoch": 0.45, "learning_rate": 0.0004893212287129112, "loss": 1.7513, "step": 11110 }, { "epoch": 0.45, "learning_rate": 0.000489311163895487, "loss": 2.1345, "step": 11120 }, { "epoch": 0.45, "learning_rate": 0.0004893010990780627, "loss": 2.0405, "step": 11130 }, { "epoch": 0.45, "learning_rate": 0.0004892910342606386, "loss": 2.1917, "step": 11140 }, { "epoch": 0.45, "learning_rate": 0.0004892809694432143, "loss": 2.2256, "step": 11150 }, { "epoch": 0.45, "learning_rate": 0.0004892709046257901, "loss": 1.8225, "step": 11160 }, { "epoch": 0.45, "learning_rate": 0.0004892608398083659, "loss": 2.1369, "step": 11170 }, { "epoch": 0.45, "learning_rate": 0.0004892507749909416, "loss": 1.9055, "step": 11180 }, { "epoch": 0.45, "learning_rate": 0.0004892407101735175, "loss": 2.3791, "step": 11190 }, { "epoch": 0.45, "learning_rate": 0.0004892306453560932, "loss": 2.131, "step": 11200 }, { "epoch": 0.45, "learning_rate": 0.0004892205805386691, "loss": 1.8755, "step": 11210 }, { "epoch": 0.45, "learning_rate": 0.0004892105157212448, "loss": 2.0911, "step": 11220 }, { "epoch": 0.45, "learning_rate": 0.0004892004509038205, "loss": 2.0773, "step": 11230 }, { "epoch": 0.45, "learning_rate": 0.0004891903860863964, "loss": 2.0573, "step": 11240 }, { "epoch": 0.45, "learning_rate": 0.0004891803212689722, "loss": 2.0697, "step": 11250 }, { "epoch": 0.45, "learning_rate": 0.000489170256451548, "loss": 2.2512, "step": 11260 }, { "epoch": 0.45, "learning_rate": 0.0004891601916341238, "loss": 1.7506, "step": 11270 }, { "epoch": 0.45, "learning_rate": 0.0004891501268166995, "loss": 2.3606, "step": 11280 }, { "epoch": 0.45, "learning_rate": 0.0004891400619992754, "loss": 2.1051, "step": 11290 }, { "epoch": 0.45, "learning_rate": 0.0004891299971818511, "loss": 1.887, "step": 11300 }, { "epoch": 0.45, "learning_rate": 0.000489119932364427, "loss": 2.269, "step": 11310 }, { "epoch": 0.46, "learning_rate": 0.0004891098675470027, "loss": 1.9119, "step": 11320 }, { "epoch": 0.46, "learning_rate": 0.0004890998027295784, "loss": 2.1062, "step": 11330 }, { "epoch": 0.46, "learning_rate": 0.0004890897379121543, "loss": 2.2169, "step": 11340 }, { "epoch": 0.46, "learning_rate": 0.00048907967309473, "loss": 2.2375, "step": 11350 }, { "epoch": 0.46, "learning_rate": 0.0004890696082773059, "loss": 2.1326, "step": 11360 }, { "epoch": 0.46, "learning_rate": 0.0004890595434598816, "loss": 2.0213, "step": 11370 }, { "epoch": 0.46, "learning_rate": 0.0004890494786424575, "loss": 1.9366, "step": 11380 }, { "epoch": 0.46, "learning_rate": 0.0004890394138250332, "loss": 2.0064, "step": 11390 }, { "epoch": 0.46, "learning_rate": 0.000489029349007609, "loss": 1.95, "step": 11400 }, { "epoch": 0.46, "learning_rate": 0.0004890192841901848, "loss": 1.9042, "step": 11410 }, { "epoch": 0.46, "learning_rate": 0.0004890092193727606, "loss": 2.0705, "step": 11420 }, { "epoch": 0.46, "learning_rate": 0.0004889991545553364, "loss": 2.0905, "step": 11430 }, { "epoch": 0.46, "learning_rate": 0.0004889890897379122, "loss": 2.1766, "step": 11440 }, { "epoch": 0.46, "learning_rate": 0.0004889790249204879, "loss": 1.9563, "step": 11450 }, { "epoch": 0.46, "learning_rate": 0.0004889689601030638, "loss": 1.9749, "step": 11460 }, { "epoch": 0.46, "learning_rate": 0.0004889588952856395, "loss": 1.7153, "step": 11470 }, { "epoch": 0.46, "learning_rate": 0.0004889488304682154, "loss": 2.0418, "step": 11480 }, { "epoch": 0.46, "learning_rate": 0.0004889387656507911, "loss": 1.8087, "step": 11490 }, { "epoch": 0.46, "learning_rate": 0.0004889287008333668, "loss": 2.1519, "step": 11500 }, { "epoch": 0.46, "learning_rate": 0.0004889186360159427, "loss": 2.1254, "step": 11510 }, { "epoch": 0.46, "learning_rate": 0.0004889085711985184, "loss": 2.0297, "step": 11520 }, { "epoch": 0.46, "learning_rate": 0.0004888985063810943, "loss": 2.1274, "step": 11530 }, { "epoch": 0.46, "learning_rate": 0.0004888884415636701, "loss": 2.1058, "step": 11540 }, { "epoch": 0.46, "learning_rate": 0.0004888783767462458, "loss": 2.1044, "step": 11550 }, { "epoch": 0.46, "learning_rate": 0.0004888683119288217, "loss": 2.1373, "step": 11560 }, { "epoch": 0.47, "learning_rate": 0.0004888582471113974, "loss": 2.0307, "step": 11570 }, { "epoch": 0.47, "learning_rate": 0.0004888481822939732, "loss": 1.9797, "step": 11580 }, { "epoch": 0.47, "learning_rate": 0.000488838117476549, "loss": 1.9104, "step": 11590 }, { "epoch": 0.47, "learning_rate": 0.0004888280526591247, "loss": 1.9053, "step": 11600 }, { "epoch": 0.47, "learning_rate": 0.0004888179878417006, "loss": 2.001, "step": 11610 }, { "epoch": 0.47, "learning_rate": 0.0004888079230242763, "loss": 2.186, "step": 11620 }, { "epoch": 0.47, "learning_rate": 0.0004887978582068522, "loss": 2.0639, "step": 11630 }, { "epoch": 0.47, "learning_rate": 0.0004887877933894279, "loss": 1.7103, "step": 11640 }, { "epoch": 0.47, "learning_rate": 0.0004887777285720038, "loss": 2.1196, "step": 11650 }, { "epoch": 0.47, "learning_rate": 0.0004887676637545795, "loss": 2.2138, "step": 11660 }, { "epoch": 0.47, "learning_rate": 0.0004887575989371552, "loss": 2.083, "step": 11670 }, { "epoch": 0.47, "learning_rate": 0.0004887475341197311, "loss": 1.9826, "step": 11680 }, { "epoch": 0.47, "learning_rate": 0.0004887374693023069, "loss": 1.8794, "step": 11690 }, { "epoch": 0.47, "learning_rate": 0.0004887274044848827, "loss": 1.7787, "step": 11700 }, { "epoch": 0.47, "learning_rate": 0.0004887173396674585, "loss": 1.942, "step": 11710 }, { "epoch": 0.47, "learning_rate": 0.0004887072748500342, "loss": 1.9009, "step": 11720 }, { "epoch": 0.47, "learning_rate": 0.0004886972100326101, "loss": 2.2062, "step": 11730 }, { "epoch": 0.47, "learning_rate": 0.0004886871452151858, "loss": 2.1199, "step": 11740 }, { "epoch": 0.47, "learning_rate": 0.0004886770803977616, "loss": 1.8937, "step": 11750 }, { "epoch": 0.47, "learning_rate": 0.0004886670155803374, "loss": 2.0491, "step": 11760 }, { "epoch": 0.47, "learning_rate": 0.0004886569507629131, "loss": 2.1474, "step": 11770 }, { "epoch": 0.47, "learning_rate": 0.000488646885945489, "loss": 2.0, "step": 11780 }, { "epoch": 0.47, "learning_rate": 0.0004886368211280647, "loss": 1.9578, "step": 11790 }, { "epoch": 0.47, "learning_rate": 0.0004886267563106406, "loss": 2.1987, "step": 11800 }, { "epoch": 0.47, "learning_rate": 0.0004886166914932163, "loss": 1.8193, "step": 11810 }, { "epoch": 0.48, "learning_rate": 0.0004886066266757921, "loss": 2.1743, "step": 11820 }, { "epoch": 0.48, "learning_rate": 0.0004885965618583679, "loss": 1.9253, "step": 11830 }, { "epoch": 0.48, "learning_rate": 0.0004885864970409437, "loss": 1.738, "step": 11840 }, { "epoch": 0.48, "learning_rate": 0.0004885764322235195, "loss": 2.0852, "step": 11850 }, { "epoch": 0.48, "learning_rate": 0.0004885663674060953, "loss": 2.1407, "step": 11860 }, { "epoch": 0.48, "learning_rate": 0.000488556302588671, "loss": 1.9191, "step": 11870 }, { "epoch": 0.48, "learning_rate": 0.0004885462377712469, "loss": 2.1676, "step": 11880 }, { "epoch": 0.48, "learning_rate": 0.0004885361729538226, "loss": 1.9691, "step": 11890 }, { "epoch": 0.48, "learning_rate": 0.0004885261081363985, "loss": 1.9414, "step": 11900 }, { "epoch": 0.48, "learning_rate": 0.0004885160433189742, "loss": 2.1411, "step": 11910 }, { "epoch": 0.48, "learning_rate": 0.0004885059785015499, "loss": 2.1547, "step": 11920 }, { "epoch": 0.48, "learning_rate": 0.0004884959136841258, "loss": 2.2308, "step": 11930 }, { "epoch": 0.48, "learning_rate": 0.0004884858488667015, "loss": 1.9581, "step": 11940 }, { "epoch": 0.48, "learning_rate": 0.0004884757840492774, "loss": 1.6692, "step": 11950 }, { "epoch": 0.48, "learning_rate": 0.0004884657192318531, "loss": 1.9452, "step": 11960 }, { "epoch": 0.48, "learning_rate": 0.000488455654414429, "loss": 2.0408, "step": 11970 }, { "epoch": 0.48, "learning_rate": 0.0004884455895970048, "loss": 2.0736, "step": 11980 }, { "epoch": 0.48, "learning_rate": 0.0004884355247795805, "loss": 1.8271, "step": 11990 }, { "epoch": 0.48, "learning_rate": 0.0004884254599621563, "loss": 2.0949, "step": 12000 }, { "epoch": 0.48, "learning_rate": 0.0004884153951447321, "loss": 1.877, "step": 12010 }, { "epoch": 0.48, "learning_rate": 0.0004884053303273078, "loss": 1.9128, "step": 12020 }, { "epoch": 0.48, "learning_rate": 0.0004883952655098837, "loss": 2.1044, "step": 12030 }, { "epoch": 0.48, "learning_rate": 0.0004883852006924594, "loss": 2.3039, "step": 12040 }, { "epoch": 0.48, "learning_rate": 0.0004883751358750353, "loss": 2.1712, "step": 12050 }, { "epoch": 0.49, "learning_rate": 0.000488365071057611, "loss": 1.8963, "step": 12060 }, { "epoch": 0.49, "learning_rate": 0.0004883550062401869, "loss": 1.9782, "step": 12070 }, { "epoch": 0.49, "learning_rate": 0.0004883449414227626, "loss": 1.7759, "step": 12080 }, { "epoch": 0.49, "learning_rate": 0.0004883348766053383, "loss": 1.6043, "step": 12090 }, { "epoch": 0.49, "learning_rate": 0.0004883248117879142, "loss": 2.1684, "step": 12100 }, { "epoch": 0.49, "learning_rate": 0.0004883147469704899, "loss": 2.0127, "step": 12110 }, { "epoch": 0.49, "learning_rate": 0.0004883046821530658, "loss": 1.939, "step": 12120 }, { "epoch": 0.49, "learning_rate": 0.0004882946173356416, "loss": 1.9024, "step": 12130 }, { "epoch": 0.49, "learning_rate": 0.0004882845525182173, "loss": 2.0238, "step": 12140 }, { "epoch": 0.49, "learning_rate": 0.00048827448770079313, "loss": 2.0207, "step": 12150 }, { "epoch": 0.49, "learning_rate": 0.0004882644228833689, "loss": 1.883, "step": 12160 }, { "epoch": 0.49, "learning_rate": 0.0004882543580659447, "loss": 2.1111, "step": 12170 }, { "epoch": 0.49, "learning_rate": 0.0004882442932485205, "loss": 2.1242, "step": 12180 }, { "epoch": 0.49, "learning_rate": 0.0004882342284310963, "loss": 2.0652, "step": 12190 }, { "epoch": 0.49, "learning_rate": 0.0004882241636136721, "loss": 1.9518, "step": 12200 }, { "epoch": 0.49, "learning_rate": 0.0004882140987962478, "loss": 1.9636, "step": 12210 }, { "epoch": 0.49, "learning_rate": 0.0004882040339788236, "loss": 2.0799, "step": 12220 }, { "epoch": 0.49, "learning_rate": 0.0004881939691613994, "loss": 2.027, "step": 12230 }, { "epoch": 0.49, "learning_rate": 0.0004881839043439752, "loss": 1.7992, "step": 12240 }, { "epoch": 0.49, "learning_rate": 0.000488173839526551, "loss": 1.9444, "step": 12250 }, { "epoch": 0.49, "learning_rate": 0.0004881637747091268, "loss": 2.0981, "step": 12260 }, { "epoch": 0.49, "learning_rate": 0.0004881537098917026, "loss": 2.036, "step": 12270 }, { "epoch": 0.49, "learning_rate": 0.0004881436450742784, "loss": 2.104, "step": 12280 }, { "epoch": 0.49, "learning_rate": 0.00048813358025685415, "loss": 1.8222, "step": 12290 }, { "epoch": 0.49, "learning_rate": 0.00048812351543942994, "loss": 1.7819, "step": 12300 }, { "epoch": 0.5, "learning_rate": 0.00048811345062200573, "loss": 1.9096, "step": 12310 }, { "epoch": 0.5, "learning_rate": 0.0004881033858045815, "loss": 1.9119, "step": 12320 }, { "epoch": 0.5, "learning_rate": 0.0004880933209871573, "loss": 1.846, "step": 12330 }, { "epoch": 0.5, "learning_rate": 0.0004880832561697331, "loss": 2.0277, "step": 12340 }, { "epoch": 0.5, "learning_rate": 0.0004880731913523089, "loss": 1.7877, "step": 12350 }, { "epoch": 0.5, "learning_rate": 0.0004880631265348847, "loss": 1.9702, "step": 12360 }, { "epoch": 0.5, "learning_rate": 0.0004880530617174604, "loss": 2.1775, "step": 12370 }, { "epoch": 0.5, "learning_rate": 0.0004880429969000362, "loss": 1.9069, "step": 12380 }, { "epoch": 0.5, "learning_rate": 0.000488032932082612, "loss": 1.7466, "step": 12390 }, { "epoch": 0.5, "learning_rate": 0.0004880228672651878, "loss": 1.8462, "step": 12400 }, { "epoch": 0.5, "learning_rate": 0.0004880128024477636, "loss": 2.0301, "step": 12410 }, { "epoch": 0.5, "learning_rate": 0.0004880027376303394, "loss": 1.9287, "step": 12420 }, { "epoch": 0.5, "learning_rate": 0.0004879926728129152, "loss": 2.0466, "step": 12430 }, { "epoch": 0.5, "learning_rate": 0.000487982607995491, "loss": 1.9292, "step": 12440 }, { "epoch": 0.5, "learning_rate": 0.0004879725431780668, "loss": 2.0283, "step": 12450 }, { "epoch": 0.5, "learning_rate": 0.00048796247836064253, "loss": 2.1234, "step": 12460 }, { "epoch": 0.5, "learning_rate": 0.0004879524135432183, "loss": 1.836, "step": 12470 }, { "epoch": 0.5, "learning_rate": 0.0004879423487257941, "loss": 2.2312, "step": 12480 }, { "epoch": 0.5, "learning_rate": 0.0004879322839083699, "loss": 2.0556, "step": 12490 }, { "epoch": 0.5, "learning_rate": 0.0004879222190909457, "loss": 2.1697, "step": 12500 }, { "epoch": 0.5, "learning_rate": 0.0004879121542735215, "loss": 1.7774, "step": 12510 }, { "epoch": 0.5, "learning_rate": 0.0004879020894560973, "loss": 2.0372, "step": 12520 }, { "epoch": 0.5, "learning_rate": 0.00048789202463867307, "loss": 2.0841, "step": 12530 }, { "epoch": 0.5, "learning_rate": 0.0004878819598212488, "loss": 1.9536, "step": 12540 }, { "epoch": 0.5, "learning_rate": 0.0004878718950038246, "loss": 1.8446, "step": 12550 }, { "epoch": 0.51, "learning_rate": 0.0004878618301864004, "loss": 1.9007, "step": 12560 }, { "epoch": 0.51, "learning_rate": 0.00048785176536897623, "loss": 2.3276, "step": 12570 }, { "epoch": 0.51, "learning_rate": 0.000487841700551552, "loss": 1.887, "step": 12580 }, { "epoch": 0.51, "learning_rate": 0.0004878316357341278, "loss": 1.9815, "step": 12590 }, { "epoch": 0.51, "learning_rate": 0.0004878215709167036, "loss": 2.0301, "step": 12600 }, { "epoch": 0.51, "learning_rate": 0.0004878115060992794, "loss": 2.1099, "step": 12610 }, { "epoch": 0.51, "learning_rate": 0.0004878014412818552, "loss": 1.8849, "step": 12620 }, { "epoch": 0.51, "learning_rate": 0.0004877913764644309, "loss": 2.0442, "step": 12630 }, { "epoch": 0.51, "learning_rate": 0.0004877813116470067, "loss": 1.8596, "step": 12640 }, { "epoch": 0.51, "learning_rate": 0.0004877712468295825, "loss": 2.0962, "step": 12650 }, { "epoch": 0.51, "learning_rate": 0.0004877611820121583, "loss": 1.8844, "step": 12660 }, { "epoch": 0.51, "learning_rate": 0.0004877511171947341, "loss": 2.0696, "step": 12670 }, { "epoch": 0.51, "learning_rate": 0.0004877410523773099, "loss": 2.0096, "step": 12680 }, { "epoch": 0.51, "learning_rate": 0.00048773098755988567, "loss": 1.8393, "step": 12690 }, { "epoch": 0.51, "learning_rate": 0.00048772092274246146, "loss": 2.1667, "step": 12700 }, { "epoch": 0.51, "learning_rate": 0.00048771085792503725, "loss": 1.9371, "step": 12710 }, { "epoch": 0.51, "learning_rate": 0.00048770079310761304, "loss": 2.0923, "step": 12720 }, { "epoch": 0.51, "learning_rate": 0.00048769072829018883, "loss": 2.1232, "step": 12730 }, { "epoch": 0.51, "learning_rate": 0.0004876806634727646, "loss": 1.9858, "step": 12740 }, { "epoch": 0.51, "learning_rate": 0.0004876705986553404, "loss": 1.973, "step": 12750 }, { "epoch": 0.51, "learning_rate": 0.0004876605338379162, "loss": 2.1107, "step": 12760 }, { "epoch": 0.51, "learning_rate": 0.000487650469020492, "loss": 1.8528, "step": 12770 }, { "epoch": 0.51, "learning_rate": 0.0004876404042030678, "loss": 2.1044, "step": 12780 }, { "epoch": 0.51, "learning_rate": 0.0004876303393856436, "loss": 2.0002, "step": 12790 }, { "epoch": 0.51, "learning_rate": 0.0004876202745682193, "loss": 1.7462, "step": 12800 }, { "epoch": 0.52, "learning_rate": 0.0004876102097507951, "loss": 2.0793, "step": 12810 }, { "epoch": 0.52, "learning_rate": 0.0004876001449333709, "loss": 2.3611, "step": 12820 }, { "epoch": 0.52, "learning_rate": 0.0004875900801159467, "loss": 1.6893, "step": 12830 }, { "epoch": 0.52, "learning_rate": 0.0004875800152985225, "loss": 2.1756, "step": 12840 }, { "epoch": 0.52, "learning_rate": 0.00048756995048109827, "loss": 1.962, "step": 12850 }, { "epoch": 0.52, "learning_rate": 0.0004875598856636741, "loss": 1.9084, "step": 12860 }, { "epoch": 0.52, "learning_rate": 0.0004875498208462499, "loss": 1.9002, "step": 12870 }, { "epoch": 0.52, "learning_rate": 0.00048753975602882564, "loss": 1.6363, "step": 12880 }, { "epoch": 0.52, "learning_rate": 0.00048752969121140143, "loss": 1.9845, "step": 12890 }, { "epoch": 0.52, "learning_rate": 0.0004875196263939772, "loss": 2.2484, "step": 12900 }, { "epoch": 0.52, "learning_rate": 0.000487509561576553, "loss": 1.9526, "step": 12910 }, { "epoch": 0.52, "learning_rate": 0.0004874994967591288, "loss": 2.0265, "step": 12920 }, { "epoch": 0.52, "learning_rate": 0.0004874894319417046, "loss": 1.9307, "step": 12930 }, { "epoch": 0.52, "learning_rate": 0.0004874793671242804, "loss": 1.9587, "step": 12940 }, { "epoch": 0.52, "learning_rate": 0.0004874693023068562, "loss": 1.7881, "step": 12950 }, { "epoch": 0.52, "learning_rate": 0.00048745923748943197, "loss": 1.8663, "step": 12960 }, { "epoch": 0.52, "learning_rate": 0.0004874491726720077, "loss": 2.3479, "step": 12970 }, { "epoch": 0.52, "learning_rate": 0.0004874391078545835, "loss": 1.7585, "step": 12980 }, { "epoch": 0.52, "learning_rate": 0.0004874290430371593, "loss": 1.9414, "step": 12990 }, { "epoch": 0.52, "learning_rate": 0.00048741897821973513, "loss": 2.0035, "step": 13000 }, { "epoch": 0.52, "learning_rate": 0.0004874089134023109, "loss": 2.0453, "step": 13010 }, { "epoch": 0.52, "learning_rate": 0.0004873988485848867, "loss": 1.8765, "step": 13020 }, { "epoch": 0.52, "learning_rate": 0.0004873887837674625, "loss": 1.8118, "step": 13030 }, { "epoch": 0.52, "learning_rate": 0.0004873787189500383, "loss": 2.0195, "step": 13040 }, { "epoch": 0.52, "learning_rate": 0.00048736865413261403, "loss": 1.9551, "step": 13050 }, { "epoch": 0.53, "learning_rate": 0.0004873585893151898, "loss": 1.9029, "step": 13060 }, { "epoch": 0.53, "learning_rate": 0.0004873485244977656, "loss": 1.8362, "step": 13070 }, { "epoch": 0.53, "learning_rate": 0.0004873384596803414, "loss": 1.9738, "step": 13080 }, { "epoch": 0.53, "learning_rate": 0.0004873283948629172, "loss": 1.9715, "step": 13090 }, { "epoch": 0.53, "learning_rate": 0.000487318330045493, "loss": 1.8444, "step": 13100 }, { "epoch": 0.53, "learning_rate": 0.0004873082652280688, "loss": 2.0444, "step": 13110 }, { "epoch": 0.53, "learning_rate": 0.00048729820041064456, "loss": 1.9516, "step": 13120 }, { "epoch": 0.53, "learning_rate": 0.0004872881355932203, "loss": 1.9446, "step": 13130 }, { "epoch": 0.53, "learning_rate": 0.0004872780707757961, "loss": 1.9636, "step": 13140 }, { "epoch": 0.53, "learning_rate": 0.00048726800595837194, "loss": 2.1325, "step": 13150 }, { "epoch": 0.53, "learning_rate": 0.0004872579411409477, "loss": 2.1091, "step": 13160 }, { "epoch": 0.53, "learning_rate": 0.0004872478763235235, "loss": 1.8848, "step": 13170 }, { "epoch": 0.53, "learning_rate": 0.0004872378115060993, "loss": 1.8311, "step": 13180 }, { "epoch": 0.53, "learning_rate": 0.0004872277466886751, "loss": 2.3148, "step": 13190 }, { "epoch": 0.53, "learning_rate": 0.0004872176818712509, "loss": 1.7549, "step": 13200 }, { "epoch": 0.53, "learning_rate": 0.0004872076170538267, "loss": 1.9264, "step": 13210 }, { "epoch": 0.53, "learning_rate": 0.0004871975522364024, "loss": 1.9194, "step": 13220 }, { "epoch": 0.53, "learning_rate": 0.0004871874874189782, "loss": 2.155, "step": 13230 }, { "epoch": 0.53, "learning_rate": 0.000487177422601554, "loss": 2.1746, "step": 13240 }, { "epoch": 0.53, "learning_rate": 0.0004871673577841298, "loss": 2.1459, "step": 13250 }, { "epoch": 0.53, "learning_rate": 0.0004871572929667056, "loss": 2.0727, "step": 13260 }, { "epoch": 0.53, "learning_rate": 0.00048714722814928137, "loss": 1.9896, "step": 13270 }, { "epoch": 0.53, "learning_rate": 0.00048713716333185716, "loss": 1.7041, "step": 13280 }, { "epoch": 0.53, "learning_rate": 0.00048712709851443295, "loss": 1.8043, "step": 13290 }, { "epoch": 0.53, "learning_rate": 0.0004871170336970088, "loss": 1.7209, "step": 13300 }, { "epoch": 0.54, "learning_rate": 0.00048710696887958453, "loss": 1.8524, "step": 13310 }, { "epoch": 0.54, "learning_rate": 0.0004870969040621603, "loss": 1.7134, "step": 13320 }, { "epoch": 0.54, "learning_rate": 0.0004870868392447361, "loss": 1.7852, "step": 13330 }, { "epoch": 0.54, "learning_rate": 0.0004870767744273119, "loss": 2.0386, "step": 13340 }, { "epoch": 0.54, "learning_rate": 0.0004870667096098877, "loss": 2.1857, "step": 13350 }, { "epoch": 0.54, "learning_rate": 0.0004870566447924635, "loss": 1.7999, "step": 13360 }, { "epoch": 0.54, "learning_rate": 0.0004870465799750393, "loss": 2.1127, "step": 13370 }, { "epoch": 0.54, "learning_rate": 0.00048703651515761507, "loss": 1.6574, "step": 13380 }, { "epoch": 0.54, "learning_rate": 0.0004870264503401908, "loss": 2.0575, "step": 13390 }, { "epoch": 0.54, "learning_rate": 0.0004870163855227666, "loss": 2.0399, "step": 13400 }, { "epoch": 0.54, "learning_rate": 0.0004870063207053424, "loss": 1.9221, "step": 13410 }, { "epoch": 0.54, "learning_rate": 0.0004869962558879182, "loss": 1.9665, "step": 13420 }, { "epoch": 0.54, "learning_rate": 0.00048698619107049397, "loss": 1.8696, "step": 13430 }, { "epoch": 0.54, "learning_rate": 0.0004869761262530698, "loss": 1.6954, "step": 13440 }, { "epoch": 0.54, "learning_rate": 0.0004869660614356456, "loss": 1.9037, "step": 13450 }, { "epoch": 0.54, "learning_rate": 0.0004869559966182214, "loss": 2.3042, "step": 13460 }, { "epoch": 0.54, "learning_rate": 0.00048694593180079713, "loss": 2.1022, "step": 13470 }, { "epoch": 0.54, "learning_rate": 0.0004869358669833729, "loss": 1.7964, "step": 13480 }, { "epoch": 0.54, "learning_rate": 0.0004869258021659487, "loss": 1.8282, "step": 13490 }, { "epoch": 0.54, "learning_rate": 0.0004869157373485245, "loss": 2.0984, "step": 13500 }, { "epoch": 0.54, "learning_rate": 0.0004869056725311003, "loss": 2.0527, "step": 13510 }, { "epoch": 0.54, "learning_rate": 0.0004868956077136761, "loss": 2.0976, "step": 13520 }, { "epoch": 0.54, "learning_rate": 0.0004868855428962519, "loss": 1.7107, "step": 13530 }, { "epoch": 0.54, "learning_rate": 0.00048687547807882767, "loss": 2.1218, "step": 13540 }, { "epoch": 0.54, "learning_rate": 0.00048686541326140346, "loss": 1.9202, "step": 13550 }, { "epoch": 0.55, "learning_rate": 0.0004868553484439792, "loss": 1.8839, "step": 13560 }, { "epoch": 0.55, "learning_rate": 0.000486845283626555, "loss": 2.081, "step": 13570 }, { "epoch": 0.55, "learning_rate": 0.0004868352188091308, "loss": 2.0436, "step": 13580 }, { "epoch": 0.55, "learning_rate": 0.0004868251539917066, "loss": 1.7938, "step": 13590 }, { "epoch": 0.55, "learning_rate": 0.0004868150891742824, "loss": 1.5098, "step": 13600 }, { "epoch": 0.55, "learning_rate": 0.0004868050243568582, "loss": 1.7629, "step": 13610 }, { "epoch": 0.55, "learning_rate": 0.000486794959539434, "loss": 1.8019, "step": 13620 }, { "epoch": 0.55, "learning_rate": 0.0004867848947220098, "loss": 2.2842, "step": 13630 }, { "epoch": 0.55, "learning_rate": 0.0004867748299045855, "loss": 1.9939, "step": 13640 }, { "epoch": 0.55, "learning_rate": 0.0004867647650871613, "loss": 1.8444, "step": 13650 }, { "epoch": 0.55, "learning_rate": 0.0004867547002697371, "loss": 1.8553, "step": 13660 }, { "epoch": 0.55, "learning_rate": 0.0004867446354523129, "loss": 1.7718, "step": 13670 }, { "epoch": 0.55, "learning_rate": 0.0004867345706348887, "loss": 2.0135, "step": 13680 }, { "epoch": 0.55, "learning_rate": 0.0004867245058174645, "loss": 2.004, "step": 13690 }, { "epoch": 0.55, "learning_rate": 0.00048671444100004027, "loss": 1.921, "step": 13700 }, { "epoch": 0.55, "learning_rate": 0.00048670437618261606, "loss": 1.9268, "step": 13710 }, { "epoch": 0.55, "learning_rate": 0.00048669431136519185, "loss": 2.0479, "step": 13720 }, { "epoch": 0.55, "learning_rate": 0.0004866842465477676, "loss": 1.7869, "step": 13730 }, { "epoch": 0.55, "learning_rate": 0.00048667418173034343, "loss": 1.9199, "step": 13740 }, { "epoch": 0.55, "learning_rate": 0.0004866641169129192, "loss": 1.9595, "step": 13750 }, { "epoch": 0.55, "learning_rate": 0.000486654052095495, "loss": 1.7283, "step": 13760 }, { "epoch": 0.55, "learning_rate": 0.0004866439872780708, "loss": 1.8304, "step": 13770 }, { "epoch": 0.55, "learning_rate": 0.0004866339224606466, "loss": 1.9594, "step": 13780 }, { "epoch": 0.55, "learning_rate": 0.0004866238576432224, "loss": 1.6093, "step": 13790 }, { "epoch": 0.56, "learning_rate": 0.0004866137928257982, "loss": 1.7316, "step": 13800 }, { "epoch": 0.56, "learning_rate": 0.0004866037280083739, "loss": 1.9997, "step": 13810 }, { "epoch": 0.56, "learning_rate": 0.0004865936631909497, "loss": 1.9447, "step": 13820 }, { "epoch": 0.56, "learning_rate": 0.0004865835983735255, "loss": 1.8499, "step": 13830 }, { "epoch": 0.56, "learning_rate": 0.0004865735335561013, "loss": 1.8471, "step": 13840 }, { "epoch": 0.56, "learning_rate": 0.0004865634687386771, "loss": 2.0152, "step": 13850 }, { "epoch": 0.56, "learning_rate": 0.00048655340392125286, "loss": 1.8033, "step": 13860 }, { "epoch": 0.56, "learning_rate": 0.00048654333910382866, "loss": 2.1614, "step": 13870 }, { "epoch": 0.56, "learning_rate": 0.0004865332742864045, "loss": 2.076, "step": 13880 }, { "epoch": 0.56, "learning_rate": 0.0004865232094689803, "loss": 1.7559, "step": 13890 }, { "epoch": 0.56, "learning_rate": 0.00048651314465155603, "loss": 1.936, "step": 13900 }, { "epoch": 0.56, "learning_rate": 0.0004865030798341318, "loss": 1.9432, "step": 13910 }, { "epoch": 0.56, "learning_rate": 0.0004864930150167076, "loss": 2.0539, "step": 13920 }, { "epoch": 0.56, "learning_rate": 0.0004864829501992834, "loss": 1.9337, "step": 13930 }, { "epoch": 0.56, "learning_rate": 0.0004864728853818592, "loss": 1.9267, "step": 13940 }, { "epoch": 0.56, "learning_rate": 0.000486462820564435, "loss": 1.6931, "step": 13950 }, { "epoch": 0.56, "learning_rate": 0.00048645275574701077, "loss": 2.1381, "step": 13960 }, { "epoch": 0.56, "learning_rate": 0.00048644269092958656, "loss": 1.782, "step": 13970 }, { "epoch": 0.56, "learning_rate": 0.0004864326261121623, "loss": 1.991, "step": 13980 }, { "epoch": 0.56, "learning_rate": 0.0004864225612947381, "loss": 1.9735, "step": 13990 }, { "epoch": 0.56, "learning_rate": 0.0004864124964773139, "loss": 1.7841, "step": 14000 }, { "epoch": 0.56, "learning_rate": 0.00048640243165988967, "loss": 2.1236, "step": 14010 }, { "epoch": 0.56, "learning_rate": 0.00048639236684246546, "loss": 1.8322, "step": 14020 }, { "epoch": 0.56, "learning_rate": 0.0004863823020250413, "loss": 1.9079, "step": 14030 }, { "epoch": 0.56, "learning_rate": 0.0004863722372076171, "loss": 1.6909, "step": 14040 }, { "epoch": 0.57, "learning_rate": 0.0004863621723901929, "loss": 1.9651, "step": 14050 }, { "epoch": 0.57, "learning_rate": 0.0004863521075727687, "loss": 1.7435, "step": 14060 }, { "epoch": 0.57, "learning_rate": 0.0004863420427553444, "loss": 1.9103, "step": 14070 }, { "epoch": 0.57, "learning_rate": 0.0004863319779379202, "loss": 1.7262, "step": 14080 }, { "epoch": 0.57, "learning_rate": 0.000486321913120496, "loss": 1.925, "step": 14090 }, { "epoch": 0.57, "learning_rate": 0.0004863118483030718, "loss": 1.8458, "step": 14100 }, { "epoch": 0.57, "learning_rate": 0.0004863017834856476, "loss": 1.8989, "step": 14110 }, { "epoch": 0.57, "learning_rate": 0.00048629171866822337, "loss": 2.0843, "step": 14120 }, { "epoch": 0.57, "learning_rate": 0.00048628165385079916, "loss": 2.0066, "step": 14130 }, { "epoch": 0.57, "learning_rate": 0.00048627158903337495, "loss": 1.8879, "step": 14140 }, { "epoch": 0.57, "learning_rate": 0.0004862615242159507, "loss": 1.6671, "step": 14150 }, { "epoch": 0.57, "learning_rate": 0.0004862514593985265, "loss": 1.8595, "step": 14160 }, { "epoch": 0.57, "learning_rate": 0.0004862413945811023, "loss": 2.0377, "step": 14170 }, { "epoch": 0.57, "learning_rate": 0.0004862313297636781, "loss": 2.0894, "step": 14180 }, { "epoch": 0.57, "learning_rate": 0.0004862212649462539, "loss": 1.8646, "step": 14190 }, { "epoch": 0.57, "learning_rate": 0.0004862112001288297, "loss": 1.7608, "step": 14200 }, { "epoch": 0.57, "learning_rate": 0.0004862011353114055, "loss": 1.9393, "step": 14210 }, { "epoch": 0.57, "learning_rate": 0.0004861910704939813, "loss": 1.9398, "step": 14220 }, { "epoch": 0.57, "learning_rate": 0.000486181005676557, "loss": 2.0566, "step": 14230 }, { "epoch": 0.57, "learning_rate": 0.0004861709408591328, "loss": 2.004, "step": 14240 }, { "epoch": 0.57, "learning_rate": 0.0004861608760417086, "loss": 2.0372, "step": 14250 }, { "epoch": 0.57, "learning_rate": 0.0004861508112242844, "loss": 1.9339, "step": 14260 }, { "epoch": 0.57, "learning_rate": 0.0004861407464068602, "loss": 1.8763, "step": 14270 }, { "epoch": 0.57, "learning_rate": 0.00048613068158943597, "loss": 1.6165, "step": 14280 }, { "epoch": 0.57, "learning_rate": 0.00048612061677201176, "loss": 1.9583, "step": 14290 }, { "epoch": 0.58, "learning_rate": 0.00048611055195458755, "loss": 1.7174, "step": 14300 }, { "epoch": 0.58, "learning_rate": 0.00048610048713716334, "loss": 1.7457, "step": 14310 }, { "epoch": 0.58, "learning_rate": 0.00048609042231973913, "loss": 1.9323, "step": 14320 }, { "epoch": 0.58, "learning_rate": 0.0004860803575023149, "loss": 2.203, "step": 14330 }, { "epoch": 0.58, "learning_rate": 0.0004860702926848907, "loss": 2.1473, "step": 14340 }, { "epoch": 0.58, "learning_rate": 0.0004860602278674665, "loss": 1.7308, "step": 14350 }, { "epoch": 0.58, "learning_rate": 0.0004860501630500423, "loss": 1.9809, "step": 14360 }, { "epoch": 0.58, "learning_rate": 0.0004860400982326181, "loss": 1.8911, "step": 14370 }, { "epoch": 0.58, "learning_rate": 0.0004860300334151939, "loss": 1.97, "step": 14380 }, { "epoch": 0.58, "learning_rate": 0.00048601996859776967, "loss": 2.1417, "step": 14390 }, { "epoch": 0.58, "learning_rate": 0.0004860099037803454, "loss": 1.8857, "step": 14400 }, { "epoch": 0.58, "learning_rate": 0.0004859998389629212, "loss": 1.9284, "step": 14410 }, { "epoch": 0.58, "learning_rate": 0.000485989774145497, "loss": 1.9178, "step": 14420 }, { "epoch": 0.58, "learning_rate": 0.0004859797093280728, "loss": 1.6986, "step": 14430 }, { "epoch": 0.58, "learning_rate": 0.00048596964451064857, "loss": 2.2542, "step": 14440 }, { "epoch": 0.58, "learning_rate": 0.00048595957969322436, "loss": 1.9097, "step": 14450 }, { "epoch": 0.58, "learning_rate": 0.00048594951487580015, "loss": 1.8503, "step": 14460 }, { "epoch": 0.58, "learning_rate": 0.000485939450058376, "loss": 1.8366, "step": 14470 }, { "epoch": 0.58, "learning_rate": 0.0004859293852409518, "loss": 1.7788, "step": 14480 }, { "epoch": 0.58, "learning_rate": 0.0004859193204235275, "loss": 1.9314, "step": 14490 }, { "epoch": 0.58, "learning_rate": 0.0004859092556061033, "loss": 2.0521, "step": 14500 }, { "epoch": 0.58, "learning_rate": 0.0004858991907886791, "loss": 2.0993, "step": 14510 }, { "epoch": 0.58, "learning_rate": 0.0004858891259712549, "loss": 1.849, "step": 14520 }, { "epoch": 0.58, "learning_rate": 0.0004858790611538307, "loss": 1.9687, "step": 14530 }, { "epoch": 0.58, "learning_rate": 0.0004858689963364065, "loss": 1.8287, "step": 14540 }, { "epoch": 0.59, "learning_rate": 0.00048585893151898227, "loss": 1.9574, "step": 14550 }, { "epoch": 0.59, "learning_rate": 0.00048584886670155806, "loss": 1.8921, "step": 14560 }, { "epoch": 0.59, "learning_rate": 0.0004858388018841338, "loss": 2.012, "step": 14570 }, { "epoch": 0.59, "learning_rate": 0.0004858287370667096, "loss": 1.9787, "step": 14580 }, { "epoch": 0.59, "learning_rate": 0.0004858186722492854, "loss": 2.071, "step": 14590 }, { "epoch": 0.59, "learning_rate": 0.00048580860743186117, "loss": 1.6576, "step": 14600 }, { "epoch": 0.59, "learning_rate": 0.000485798542614437, "loss": 2.1682, "step": 14610 }, { "epoch": 0.59, "learning_rate": 0.0004857884777970128, "loss": 1.9008, "step": 14620 }, { "epoch": 0.59, "learning_rate": 0.0004857784129795886, "loss": 1.9339, "step": 14630 }, { "epoch": 0.59, "learning_rate": 0.0004857683481621644, "loss": 1.7259, "step": 14640 }, { "epoch": 0.59, "learning_rate": 0.0004857582833447402, "loss": 2.0303, "step": 14650 }, { "epoch": 0.59, "learning_rate": 0.0004857482185273159, "loss": 2.2405, "step": 14660 }, { "epoch": 0.59, "learning_rate": 0.0004857381537098917, "loss": 1.9241, "step": 14670 }, { "epoch": 0.59, "learning_rate": 0.0004857280888924675, "loss": 1.7413, "step": 14680 }, { "epoch": 0.59, "learning_rate": 0.0004857180240750433, "loss": 2.3401, "step": 14690 }, { "epoch": 0.59, "learning_rate": 0.0004857079592576191, "loss": 1.6948, "step": 14700 }, { "epoch": 0.59, "learning_rate": 0.00048569789444019486, "loss": 1.7984, "step": 14710 }, { "epoch": 0.59, "learning_rate": 0.00048568782962277066, "loss": 1.8975, "step": 14720 }, { "epoch": 0.59, "learning_rate": 0.00048567776480534645, "loss": 1.8085, "step": 14730 }, { "epoch": 0.59, "learning_rate": 0.0004856676999879222, "loss": 1.7341, "step": 14740 }, { "epoch": 0.59, "learning_rate": 0.000485657635170498, "loss": 1.868, "step": 14750 }, { "epoch": 0.59, "learning_rate": 0.0004856475703530738, "loss": 2.2121, "step": 14760 }, { "epoch": 0.59, "learning_rate": 0.0004856375055356496, "loss": 2.1691, "step": 14770 }, { "epoch": 0.59, "learning_rate": 0.0004856274407182254, "loss": 1.8309, "step": 14780 }, { "epoch": 0.59, "learning_rate": 0.0004856173759008012, "loss": 2.0141, "step": 14790 }, { "epoch": 0.6, "learning_rate": 0.000485607311083377, "loss": 1.9064, "step": 14800 }, { "epoch": 0.6, "learning_rate": 0.00048559724626595277, "loss": 1.8915, "step": 14810 }, { "epoch": 0.6, "learning_rate": 0.00048558718144852856, "loss": 1.9869, "step": 14820 }, { "epoch": 0.6, "learning_rate": 0.0004855771166311043, "loss": 2.0594, "step": 14830 }, { "epoch": 0.6, "learning_rate": 0.0004855670518136801, "loss": 1.9373, "step": 14840 }, { "epoch": 0.6, "learning_rate": 0.0004855569869962559, "loss": 1.8697, "step": 14850 }, { "epoch": 0.6, "learning_rate": 0.00048554692217883167, "loss": 1.6404, "step": 14860 }, { "epoch": 0.6, "learning_rate": 0.00048553685736140746, "loss": 1.7586, "step": 14870 }, { "epoch": 0.6, "learning_rate": 0.00048552679254398325, "loss": 1.7396, "step": 14880 }, { "epoch": 0.6, "learning_rate": 0.00048551672772655904, "loss": 1.961, "step": 14890 }, { "epoch": 0.6, "learning_rate": 0.00048550666290913484, "loss": 1.8829, "step": 14900 }, { "epoch": 0.6, "learning_rate": 0.0004854965980917106, "loss": 1.8752, "step": 14910 }, { "epoch": 0.6, "learning_rate": 0.0004854865332742864, "loss": 1.9727, "step": 14920 }, { "epoch": 0.6, "learning_rate": 0.0004854764684568622, "loss": 1.7971, "step": 14930 }, { "epoch": 0.6, "learning_rate": 0.000485466403639438, "loss": 1.8099, "step": 14940 }, { "epoch": 0.6, "learning_rate": 0.0004854563388220138, "loss": 1.9864, "step": 14950 }, { "epoch": 0.6, "learning_rate": 0.0004854462740045896, "loss": 2.3245, "step": 14960 }, { "epoch": 0.6, "learning_rate": 0.00048543620918716537, "loss": 1.9412, "step": 14970 }, { "epoch": 0.6, "learning_rate": 0.00048542614436974116, "loss": 1.7299, "step": 14980 }, { "epoch": 0.6, "learning_rate": 0.0004854160795523169, "loss": 1.9903, "step": 14990 }, { "epoch": 0.6, "learning_rate": 0.0004854060147348927, "loss": 1.6841, "step": 15000 }, { "epoch": 0.6, "learning_rate": 0.0004853959499174685, "loss": 1.7963, "step": 15010 }, { "epoch": 0.6, "learning_rate": 0.00048538588510004427, "loss": 2.0326, "step": 15020 }, { "epoch": 0.6, "learning_rate": 0.00048537582028262006, "loss": 1.6005, "step": 15030 }, { "epoch": 0.6, "learning_rate": 0.00048536575546519585, "loss": 1.972, "step": 15040 }, { "epoch": 0.61, "learning_rate": 0.0004853556906477717, "loss": 1.9913, "step": 15050 }, { "epoch": 0.61, "learning_rate": 0.0004853456258303475, "loss": 2.1929, "step": 15060 }, { "epoch": 0.61, "learning_rate": 0.0004853355610129233, "loss": 1.9452, "step": 15070 }, { "epoch": 0.61, "learning_rate": 0.000485325496195499, "loss": 1.8828, "step": 15080 }, { "epoch": 0.61, "learning_rate": 0.0004853154313780748, "loss": 1.8793, "step": 15090 }, { "epoch": 0.61, "learning_rate": 0.0004853053665606506, "loss": 1.7939, "step": 15100 }, { "epoch": 0.61, "learning_rate": 0.0004852953017432264, "loss": 2.0722, "step": 15110 }, { "epoch": 0.61, "learning_rate": 0.0004852852369258022, "loss": 1.9821, "step": 15120 }, { "epoch": 0.61, "learning_rate": 0.00048527517210837797, "loss": 1.7426, "step": 15130 }, { "epoch": 0.61, "learning_rate": 0.00048526510729095376, "loss": 1.7645, "step": 15140 }, { "epoch": 0.61, "learning_rate": 0.00048525504247352955, "loss": 1.998, "step": 15150 }, { "epoch": 0.61, "learning_rate": 0.0004852449776561053, "loss": 1.9249, "step": 15160 }, { "epoch": 0.61, "learning_rate": 0.0004852349128386811, "loss": 1.7817, "step": 15170 }, { "epoch": 0.61, "learning_rate": 0.00048522484802125687, "loss": 1.9913, "step": 15180 }, { "epoch": 0.61, "learning_rate": 0.00048521478320383266, "loss": 1.9193, "step": 15190 }, { "epoch": 0.61, "learning_rate": 0.0004852047183864085, "loss": 1.9439, "step": 15200 }, { "epoch": 0.61, "learning_rate": 0.0004851946535689843, "loss": 1.7809, "step": 15210 }, { "epoch": 0.61, "learning_rate": 0.0004851845887515601, "loss": 1.8761, "step": 15220 }, { "epoch": 0.61, "learning_rate": 0.0004851745239341359, "loss": 2.0928, "step": 15230 }, { "epoch": 0.61, "learning_rate": 0.00048516445911671167, "loss": 1.8177, "step": 15240 }, { "epoch": 0.61, "learning_rate": 0.0004851543942992874, "loss": 1.6594, "step": 15250 }, { "epoch": 0.61, "learning_rate": 0.0004851443294818632, "loss": 1.6137, "step": 15260 }, { "epoch": 0.61, "learning_rate": 0.000485134264664439, "loss": 1.6941, "step": 15270 }, { "epoch": 0.61, "learning_rate": 0.0004851241998470148, "loss": 1.9267, "step": 15280 }, { "epoch": 0.61, "learning_rate": 0.00048511413502959057, "loss": 1.921, "step": 15290 }, { "epoch": 0.62, "learning_rate": 0.00048510407021216636, "loss": 1.7261, "step": 15300 }, { "epoch": 0.62, "learning_rate": 0.00048509400539474215, "loss": 1.725, "step": 15310 }, { "epoch": 0.62, "learning_rate": 0.00048508394057731794, "loss": 1.7928, "step": 15320 }, { "epoch": 0.62, "learning_rate": 0.0004850738757598937, "loss": 2.0527, "step": 15330 }, { "epoch": 0.62, "learning_rate": 0.0004850638109424695, "loss": 1.7998, "step": 15340 }, { "epoch": 0.62, "learning_rate": 0.0004850537461250453, "loss": 2.0464, "step": 15350 }, { "epoch": 0.62, "learning_rate": 0.0004850436813076211, "loss": 1.7816, "step": 15360 }, { "epoch": 0.62, "learning_rate": 0.0004850336164901969, "loss": 1.9433, "step": 15370 }, { "epoch": 0.62, "learning_rate": 0.0004850235516727727, "loss": 1.9593, "step": 15380 }, { "epoch": 0.62, "learning_rate": 0.0004850134868553485, "loss": 2.0583, "step": 15390 }, { "epoch": 0.62, "learning_rate": 0.00048500342203792427, "loss": 1.9331, "step": 15400 }, { "epoch": 0.62, "learning_rate": 0.00048499335722050006, "loss": 1.6606, "step": 15410 }, { "epoch": 0.62, "learning_rate": 0.0004849832924030758, "loss": 2.0746, "step": 15420 }, { "epoch": 0.62, "learning_rate": 0.0004849732275856516, "loss": 1.9877, "step": 15430 }, { "epoch": 0.62, "learning_rate": 0.0004849631627682274, "loss": 1.8733, "step": 15440 }, { "epoch": 0.62, "learning_rate": 0.00048495309795080317, "loss": 1.6895, "step": 15450 }, { "epoch": 0.62, "learning_rate": 0.00048494303313337896, "loss": 1.6948, "step": 15460 }, { "epoch": 0.62, "learning_rate": 0.00048493296831595475, "loss": 2.0404, "step": 15470 }, { "epoch": 0.62, "learning_rate": 0.00048492290349853054, "loss": 1.9845, "step": 15480 }, { "epoch": 0.62, "learning_rate": 0.0004849128386811064, "loss": 1.884, "step": 15490 }, { "epoch": 0.62, "learning_rate": 0.0004849027738636821, "loss": 2.1338, "step": 15500 }, { "epoch": 0.62, "learning_rate": 0.0004848927090462579, "loss": 1.7893, "step": 15510 }, { "epoch": 0.62, "learning_rate": 0.0004848826442288337, "loss": 1.89, "step": 15520 }, { "epoch": 0.62, "learning_rate": 0.0004848725794114095, "loss": 1.8016, "step": 15530 }, { "epoch": 0.62, "learning_rate": 0.0004848625145939853, "loss": 2.0533, "step": 15540 }, { "epoch": 0.63, "learning_rate": 0.0004848524497765611, "loss": 1.9745, "step": 15550 }, { "epoch": 0.63, "learning_rate": 0.00048484238495913686, "loss": 1.9057, "step": 15560 }, { "epoch": 0.63, "learning_rate": 0.00048483232014171266, "loss": 1.8018, "step": 15570 }, { "epoch": 0.63, "learning_rate": 0.00048482225532428845, "loss": 1.8207, "step": 15580 }, { "epoch": 0.63, "learning_rate": 0.0004848121905068642, "loss": 1.8095, "step": 15590 }, { "epoch": 0.63, "learning_rate": 0.00048480212568944, "loss": 1.9882, "step": 15600 }, { "epoch": 0.63, "learning_rate": 0.00048479206087201576, "loss": 1.7908, "step": 15610 }, { "epoch": 0.63, "learning_rate": 0.00048478199605459155, "loss": 2.1624, "step": 15620 }, { "epoch": 0.63, "learning_rate": 0.00048477193123716735, "loss": 2.2082, "step": 15630 }, { "epoch": 0.63, "learning_rate": 0.0004847618664197432, "loss": 1.8421, "step": 15640 }, { "epoch": 0.63, "learning_rate": 0.000484751801602319, "loss": 1.8518, "step": 15650 }, { "epoch": 0.63, "learning_rate": 0.00048474173678489477, "loss": 2.0076, "step": 15660 }, { "epoch": 0.63, "learning_rate": 0.0004847316719674705, "loss": 1.6358, "step": 15670 }, { "epoch": 0.63, "learning_rate": 0.0004847216071500463, "loss": 2.206, "step": 15680 }, { "epoch": 0.63, "learning_rate": 0.0004847115423326221, "loss": 1.9489, "step": 15690 }, { "epoch": 0.63, "learning_rate": 0.0004847014775151979, "loss": 2.0472, "step": 15700 }, { "epoch": 0.63, "learning_rate": 0.00048469141269777367, "loss": 1.8162, "step": 15710 }, { "epoch": 0.63, "learning_rate": 0.00048468134788034946, "loss": 1.7066, "step": 15720 }, { "epoch": 0.63, "learning_rate": 0.00048467128306292525, "loss": 1.8839, "step": 15730 }, { "epoch": 0.63, "learning_rate": 0.00048466121824550104, "loss": 1.7294, "step": 15740 }, { "epoch": 0.63, "learning_rate": 0.0004846511534280768, "loss": 1.6528, "step": 15750 }, { "epoch": 0.63, "learning_rate": 0.00048464108861065257, "loss": 1.8269, "step": 15760 }, { "epoch": 0.63, "learning_rate": 0.00048463102379322836, "loss": 2.1569, "step": 15770 }, { "epoch": 0.63, "learning_rate": 0.0004846209589758042, "loss": 2.1395, "step": 15780 }, { "epoch": 0.64, "learning_rate": 0.00048461089415838, "loss": 1.8658, "step": 15790 }, { "epoch": 0.64, "learning_rate": 0.0004846008293409558, "loss": 2.0625, "step": 15800 }, { "epoch": 0.64, "learning_rate": 0.0004845907645235316, "loss": 1.7178, "step": 15810 }, { "epoch": 0.64, "learning_rate": 0.00048458069970610737, "loss": 2.1079, "step": 15820 }, { "epoch": 0.64, "learning_rate": 0.00048457063488868316, "loss": 1.9669, "step": 15830 }, { "epoch": 0.64, "learning_rate": 0.0004845605700712589, "loss": 1.8993, "step": 15840 }, { "epoch": 0.64, "learning_rate": 0.0004845505052538347, "loss": 1.8126, "step": 15850 }, { "epoch": 0.64, "learning_rate": 0.0004845404404364105, "loss": 1.7919, "step": 15860 }, { "epoch": 0.64, "learning_rate": 0.00048453037561898627, "loss": 1.845, "step": 15870 }, { "epoch": 0.64, "learning_rate": 0.00048452031080156206, "loss": 1.7272, "step": 15880 }, { "epoch": 0.64, "learning_rate": 0.00048451024598413785, "loss": 1.702, "step": 15890 }, { "epoch": 0.64, "learning_rate": 0.00048450018116671364, "loss": 1.815, "step": 15900 }, { "epoch": 0.64, "learning_rate": 0.00048449011634928943, "loss": 1.9861, "step": 15910 }, { "epoch": 0.64, "learning_rate": 0.00048448005153186517, "loss": 2.1954, "step": 15920 }, { "epoch": 0.64, "learning_rate": 0.000484469986714441, "loss": 1.7178, "step": 15930 }, { "epoch": 0.64, "learning_rate": 0.0004844599218970168, "loss": 1.6031, "step": 15940 }, { "epoch": 0.64, "learning_rate": 0.0004844498570795926, "loss": 1.862, "step": 15950 }, { "epoch": 0.64, "learning_rate": 0.0004844397922621684, "loss": 1.881, "step": 15960 }, { "epoch": 0.64, "learning_rate": 0.0004844297274447442, "loss": 1.7968, "step": 15970 }, { "epoch": 0.64, "learning_rate": 0.00048441966262731997, "loss": 1.8059, "step": 15980 }, { "epoch": 0.64, "learning_rate": 0.00048440959780989576, "loss": 1.6031, "step": 15990 }, { "epoch": 0.64, "learning_rate": 0.00048439953299247155, "loss": 1.9433, "step": 16000 }, { "epoch": 0.64, "learning_rate": 0.0004843894681750473, "loss": 1.8747, "step": 16010 }, { "epoch": 0.64, "learning_rate": 0.0004843794033576231, "loss": 1.7626, "step": 16020 }, { "epoch": 0.64, "learning_rate": 0.00048436933854019887, "loss": 1.8141, "step": 16030 }, { "epoch": 0.65, "learning_rate": 0.00048435927372277466, "loss": 1.7076, "step": 16040 }, { "epoch": 0.65, "learning_rate": 0.00048434920890535045, "loss": 1.6972, "step": 16050 }, { "epoch": 0.65, "learning_rate": 0.00048433914408792624, "loss": 1.829, "step": 16060 }, { "epoch": 0.65, "learning_rate": 0.00048432907927050203, "loss": 2.0835, "step": 16070 }, { "epoch": 0.65, "learning_rate": 0.0004843190144530779, "loss": 1.9304, "step": 16080 }, { "epoch": 0.65, "learning_rate": 0.0004843089496356536, "loss": 1.7761, "step": 16090 }, { "epoch": 0.65, "learning_rate": 0.0004842988848182294, "loss": 1.9326, "step": 16100 }, { "epoch": 0.65, "learning_rate": 0.0004842888200008052, "loss": 1.9927, "step": 16110 }, { "epoch": 0.65, "learning_rate": 0.000484278755183381, "loss": 1.7849, "step": 16120 }, { "epoch": 0.65, "learning_rate": 0.0004842686903659568, "loss": 1.9347, "step": 16130 }, { "epoch": 0.65, "learning_rate": 0.00048425862554853257, "loss": 1.6361, "step": 16140 }, { "epoch": 0.65, "learning_rate": 0.00048424856073110836, "loss": 1.7399, "step": 16150 }, { "epoch": 0.65, "learning_rate": 0.00048423849591368415, "loss": 1.7962, "step": 16160 }, { "epoch": 0.65, "learning_rate": 0.00048422843109625994, "loss": 1.7624, "step": 16170 }, { "epoch": 0.65, "learning_rate": 0.0004842183662788357, "loss": 1.8159, "step": 16180 }, { "epoch": 0.65, "learning_rate": 0.00048420830146141147, "loss": 1.9411, "step": 16190 }, { "epoch": 0.65, "learning_rate": 0.00048419823664398726, "loss": 1.7124, "step": 16200 }, { "epoch": 0.65, "learning_rate": 0.00048418817182656305, "loss": 2.0175, "step": 16210 }, { "epoch": 0.65, "learning_rate": 0.0004841781070091389, "loss": 1.788, "step": 16220 }, { "epoch": 0.65, "learning_rate": 0.0004841680421917147, "loss": 2.0307, "step": 16230 }, { "epoch": 0.65, "learning_rate": 0.0004841579773742905, "loss": 1.9877, "step": 16240 }, { "epoch": 0.65, "learning_rate": 0.00048414791255686627, "loss": 1.5304, "step": 16250 }, { "epoch": 0.65, "learning_rate": 0.000484137847739442, "loss": 1.7035, "step": 16260 }, { "epoch": 0.65, "learning_rate": 0.0004841277829220178, "loss": 1.8168, "step": 16270 }, { "epoch": 0.65, "learning_rate": 0.0004841177181045936, "loss": 1.9671, "step": 16280 }, { "epoch": 0.66, "learning_rate": 0.0004841076532871694, "loss": 1.7568, "step": 16290 }, { "epoch": 0.66, "learning_rate": 0.00048409758846974517, "loss": 1.8284, "step": 16300 }, { "epoch": 0.66, "learning_rate": 0.00048408752365232096, "loss": 1.8298, "step": 16310 }, { "epoch": 0.66, "learning_rate": 0.00048407745883489675, "loss": 2.1575, "step": 16320 }, { "epoch": 0.66, "learning_rate": 0.00048406739401747254, "loss": 1.9678, "step": 16330 }, { "epoch": 0.66, "learning_rate": 0.00048405732920004833, "loss": 1.8881, "step": 16340 }, { "epoch": 0.66, "learning_rate": 0.00048404726438262407, "loss": 1.9412, "step": 16350 }, { "epoch": 0.66, "learning_rate": 0.00048403719956519986, "loss": 1.7638, "step": 16360 }, { "epoch": 0.66, "learning_rate": 0.0004840271347477757, "loss": 1.8463, "step": 16370 }, { "epoch": 0.66, "learning_rate": 0.0004840170699303515, "loss": 2.049, "step": 16380 }, { "epoch": 0.66, "learning_rate": 0.0004840070051129273, "loss": 1.6454, "step": 16390 }, { "epoch": 0.66, "learning_rate": 0.0004839969402955031, "loss": 1.7696, "step": 16400 }, { "epoch": 0.66, "learning_rate": 0.00048398687547807886, "loss": 1.8236, "step": 16410 }, { "epoch": 0.66, "learning_rate": 0.00048397681066065466, "loss": 1.8404, "step": 16420 }, { "epoch": 0.66, "learning_rate": 0.0004839667458432304, "loss": 2.0463, "step": 16430 }, { "epoch": 0.66, "learning_rate": 0.0004839566810258062, "loss": 1.8416, "step": 16440 }, { "epoch": 0.66, "learning_rate": 0.000483946616208382, "loss": 1.6671, "step": 16450 }, { "epoch": 0.66, "learning_rate": 0.00048393655139095776, "loss": 1.9683, "step": 16460 }, { "epoch": 0.66, "learning_rate": 0.00048392648657353355, "loss": 1.8757, "step": 16470 }, { "epoch": 0.66, "learning_rate": 0.00048391642175610935, "loss": 2.0688, "step": 16480 }, { "epoch": 0.66, "learning_rate": 0.00048390635693868514, "loss": 1.7276, "step": 16490 }, { "epoch": 0.66, "learning_rate": 0.0004838962921212609, "loss": 1.8936, "step": 16500 }, { "epoch": 0.66, "learning_rate": 0.00048388622730383677, "loss": 1.8468, "step": 16510 }, { "epoch": 0.66, "learning_rate": 0.0004838761624864125, "loss": 1.7973, "step": 16520 }, { "epoch": 0.66, "learning_rate": 0.0004838660976689883, "loss": 1.8092, "step": 16530 }, { "epoch": 0.67, "learning_rate": 0.0004838560328515641, "loss": 1.9355, "step": 16540 }, { "epoch": 0.67, "learning_rate": 0.0004838459680341399, "loss": 1.8285, "step": 16550 }, { "epoch": 0.67, "learning_rate": 0.00048383590321671567, "loss": 1.8845, "step": 16560 }, { "epoch": 0.67, "learning_rate": 0.00048382583839929146, "loss": 1.827, "step": 16570 }, { "epoch": 0.67, "learning_rate": 0.00048381577358186725, "loss": 1.5896, "step": 16580 }, { "epoch": 0.67, "learning_rate": 0.00048380570876444304, "loss": 1.9087, "step": 16590 }, { "epoch": 0.67, "learning_rate": 0.0004837956439470188, "loss": 1.7902, "step": 16600 }, { "epoch": 0.67, "learning_rate": 0.00048378557912959457, "loss": 2.1414, "step": 16610 }, { "epoch": 0.67, "learning_rate": 0.00048377551431217036, "loss": 1.8227, "step": 16620 }, { "epoch": 0.67, "learning_rate": 0.00048376544949474615, "loss": 1.497, "step": 16630 }, { "epoch": 0.67, "learning_rate": 0.00048375538467732194, "loss": 1.8182, "step": 16640 }, { "epoch": 0.67, "learning_rate": 0.00048374531985989773, "loss": 2.1482, "step": 16650 }, { "epoch": 0.67, "learning_rate": 0.0004837352550424736, "loss": 1.7278, "step": 16660 }, { "epoch": 0.67, "learning_rate": 0.00048372519022504937, "loss": 1.728, "step": 16670 }, { "epoch": 0.67, "learning_rate": 0.0004837151254076251, "loss": 1.9379, "step": 16680 }, { "epoch": 0.67, "learning_rate": 0.0004837050605902009, "loss": 1.9461, "step": 16690 }, { "epoch": 0.67, "learning_rate": 0.0004836949957727767, "loss": 1.7552, "step": 16700 }, { "epoch": 0.67, "learning_rate": 0.0004836849309553525, "loss": 1.6722, "step": 16710 }, { "epoch": 0.67, "learning_rate": 0.00048367486613792827, "loss": 1.7166, "step": 16720 }, { "epoch": 0.67, "learning_rate": 0.00048366480132050406, "loss": 1.7394, "step": 16730 }, { "epoch": 0.67, "learning_rate": 0.00048365473650307985, "loss": 1.4972, "step": 16740 }, { "epoch": 0.67, "learning_rate": 0.00048364467168565564, "loss": 1.9085, "step": 16750 }, { "epoch": 0.67, "learning_rate": 0.00048363460686823143, "loss": 1.8095, "step": 16760 }, { "epoch": 0.67, "learning_rate": 0.00048362454205080717, "loss": 1.8597, "step": 16770 }, { "epoch": 0.67, "learning_rate": 0.00048361447723338296, "loss": 1.9253, "step": 16780 }, { "epoch": 0.68, "learning_rate": 0.00048360441241595875, "loss": 1.7286, "step": 16790 }, { "epoch": 0.68, "learning_rate": 0.00048359434759853454, "loss": 1.8431, "step": 16800 }, { "epoch": 0.68, "learning_rate": 0.0004835842827811104, "loss": 2.0535, "step": 16810 }, { "epoch": 0.68, "learning_rate": 0.0004835742179636862, "loss": 1.6999, "step": 16820 }, { "epoch": 0.68, "learning_rate": 0.00048356415314626197, "loss": 1.7267, "step": 16830 }, { "epoch": 0.68, "learning_rate": 0.00048355408832883776, "loss": 1.8767, "step": 16840 }, { "epoch": 0.68, "learning_rate": 0.0004835440235114135, "loss": 1.7969, "step": 16850 }, { "epoch": 0.68, "learning_rate": 0.0004835339586939893, "loss": 2.2174, "step": 16860 }, { "epoch": 0.68, "learning_rate": 0.0004835238938765651, "loss": 1.9706, "step": 16870 }, { "epoch": 0.68, "learning_rate": 0.00048351382905914087, "loss": 1.8788, "step": 16880 }, { "epoch": 0.68, "learning_rate": 0.00048350376424171666, "loss": 1.9011, "step": 16890 }, { "epoch": 0.68, "learning_rate": 0.00048349369942429245, "loss": 1.9367, "step": 16900 }, { "epoch": 0.68, "learning_rate": 0.00048348363460686824, "loss": 2.0555, "step": 16910 }, { "epoch": 0.68, "learning_rate": 0.00048347356978944403, "loss": 1.8989, "step": 16920 }, { "epoch": 0.68, "learning_rate": 0.0004834635049720198, "loss": 1.9596, "step": 16930 }, { "epoch": 0.68, "learning_rate": 0.00048345344015459556, "loss": 1.7319, "step": 16940 }, { "epoch": 0.68, "learning_rate": 0.0004834433753371714, "loss": 1.8301, "step": 16950 }, { "epoch": 0.68, "learning_rate": 0.0004834333105197472, "loss": 1.7964, "step": 16960 }, { "epoch": 0.68, "learning_rate": 0.000483423245702323, "loss": 1.7045, "step": 16970 }, { "epoch": 0.68, "learning_rate": 0.0004834131808848988, "loss": 1.6711, "step": 16980 }, { "epoch": 0.68, "learning_rate": 0.00048340311606747457, "loss": 1.4703, "step": 16990 }, { "epoch": 0.68, "learning_rate": 0.00048339305125005036, "loss": 1.6727, "step": 17000 }, { "epoch": 0.68, "learning_rate": 0.00048338298643262615, "loss": 1.9951, "step": 17010 }, { "epoch": 0.68, "learning_rate": 0.0004833729216152019, "loss": 2.0152, "step": 17020 }, { "epoch": 0.68, "learning_rate": 0.0004833628567977777, "loss": 2.078, "step": 17030 }, { "epoch": 0.69, "learning_rate": 0.00048335279198035347, "loss": 1.7349, "step": 17040 }, { "epoch": 0.69, "learning_rate": 0.00048334272716292926, "loss": 1.6933, "step": 17050 }, { "epoch": 0.69, "learning_rate": 0.00048333266234550505, "loss": 1.6734, "step": 17060 }, { "epoch": 0.69, "learning_rate": 0.00048332259752808084, "loss": 2.1157, "step": 17070 }, { "epoch": 0.69, "learning_rate": 0.00048331253271065663, "loss": 1.7727, "step": 17080 }, { "epoch": 0.69, "learning_rate": 0.0004833024678932324, "loss": 1.6493, "step": 17090 }, { "epoch": 0.69, "learning_rate": 0.00048329240307580827, "loss": 1.8108, "step": 17100 }, { "epoch": 0.69, "learning_rate": 0.000483282338258384, "loss": 1.6493, "step": 17110 }, { "epoch": 0.69, "learning_rate": 0.0004832722734409598, "loss": 2.0447, "step": 17120 }, { "epoch": 0.69, "learning_rate": 0.0004832622086235356, "loss": 1.657, "step": 17130 }, { "epoch": 0.69, "learning_rate": 0.0004832521438061114, "loss": 1.93, "step": 17140 }, { "epoch": 0.69, "learning_rate": 0.00048324207898868717, "loss": 1.7016, "step": 17150 }, { "epoch": 0.69, "learning_rate": 0.00048323201417126296, "loss": 1.9902, "step": 17160 }, { "epoch": 0.69, "learning_rate": 0.00048322194935383875, "loss": 2.1395, "step": 17170 }, { "epoch": 0.69, "learning_rate": 0.00048321188453641454, "loss": 1.6877, "step": 17180 }, { "epoch": 0.69, "learning_rate": 0.0004832018197189903, "loss": 1.7607, "step": 17190 }, { "epoch": 0.69, "learning_rate": 0.00048319175490156607, "loss": 1.6473, "step": 17200 }, { "epoch": 0.69, "learning_rate": 0.00048318169008414186, "loss": 1.8976, "step": 17210 }, { "epoch": 0.69, "learning_rate": 0.00048317162526671765, "loss": 1.7666, "step": 17220 }, { "epoch": 0.69, "learning_rate": 0.00048316156044929344, "loss": 1.7896, "step": 17230 }, { "epoch": 0.69, "learning_rate": 0.0004831514956318693, "loss": 1.9153, "step": 17240 }, { "epoch": 0.69, "learning_rate": 0.0004831414308144451, "loss": 1.8009, "step": 17250 }, { "epoch": 0.69, "learning_rate": 0.00048313136599702086, "loss": 1.8651, "step": 17260 }, { "epoch": 0.69, "learning_rate": 0.00048312130117959665, "loss": 2.0885, "step": 17270 }, { "epoch": 0.69, "learning_rate": 0.0004831112363621724, "loss": 1.7851, "step": 17280 }, { "epoch": 0.7, "learning_rate": 0.0004831011715447482, "loss": 1.9004, "step": 17290 }, { "epoch": 0.7, "learning_rate": 0.000483091106727324, "loss": 1.8078, "step": 17300 }, { "epoch": 0.7, "learning_rate": 0.00048308104190989976, "loss": 1.8924, "step": 17310 }, { "epoch": 0.7, "learning_rate": 0.00048307097709247555, "loss": 1.893, "step": 17320 }, { "epoch": 0.7, "learning_rate": 0.00048306091227505135, "loss": 1.7764, "step": 17330 }, { "epoch": 0.7, "learning_rate": 0.00048305084745762714, "loss": 1.683, "step": 17340 }, { "epoch": 0.7, "learning_rate": 0.0004830407826402029, "loss": 1.8697, "step": 17350 }, { "epoch": 0.7, "learning_rate": 0.00048303071782277866, "loss": 1.5998, "step": 17360 }, { "epoch": 0.7, "learning_rate": 0.00048302065300535445, "loss": 1.765, "step": 17370 }, { "epoch": 0.7, "learning_rate": 0.00048301058818793024, "loss": 2.092, "step": 17380 }, { "epoch": 0.7, "learning_rate": 0.0004830005233705061, "loss": 1.6577, "step": 17390 }, { "epoch": 0.7, "learning_rate": 0.0004829904585530819, "loss": 1.8183, "step": 17400 }, { "epoch": 0.7, "learning_rate": 0.00048298039373565767, "loss": 1.941, "step": 17410 }, { "epoch": 0.7, "learning_rate": 0.00048297032891823346, "loss": 1.6042, "step": 17420 }, { "epoch": 0.7, "learning_rate": 0.00048296026410080925, "loss": 1.8024, "step": 17430 }, { "epoch": 0.7, "learning_rate": 0.000482950199283385, "loss": 1.7865, "step": 17440 }, { "epoch": 0.7, "learning_rate": 0.0004829401344659608, "loss": 1.7009, "step": 17450 }, { "epoch": 0.7, "learning_rate": 0.00048293006964853657, "loss": 1.8255, "step": 17460 }, { "epoch": 0.7, "learning_rate": 0.00048292000483111236, "loss": 2.076, "step": 17470 }, { "epoch": 0.7, "learning_rate": 0.00048290994001368815, "loss": 1.8095, "step": 17480 }, { "epoch": 0.7, "learning_rate": 0.00048289987519626394, "loss": 1.6137, "step": 17490 }, { "epoch": 0.7, "learning_rate": 0.00048288981037883973, "loss": 1.5967, "step": 17500 }, { "epoch": 0.7, "learning_rate": 0.0004828797455614155, "loss": 1.7821, "step": 17510 }, { "epoch": 0.7, "learning_rate": 0.0004828696807439913, "loss": 2.0203, "step": 17520 }, { "epoch": 0.71, "learning_rate": 0.00048285961592656705, "loss": 2.004, "step": 17530 }, { "epoch": 0.71, "learning_rate": 0.0004828495511091429, "loss": 1.6868, "step": 17540 }, { "epoch": 0.71, "learning_rate": 0.0004828394862917187, "loss": 2.0279, "step": 17550 }, { "epoch": 0.71, "learning_rate": 0.0004828294214742945, "loss": 1.787, "step": 17560 }, { "epoch": 0.71, "learning_rate": 0.00048281935665687027, "loss": 1.6828, "step": 17570 }, { "epoch": 0.71, "learning_rate": 0.00048280929183944606, "loss": 1.7493, "step": 17580 }, { "epoch": 0.71, "learning_rate": 0.00048279922702202185, "loss": 1.9144, "step": 17590 }, { "epoch": 0.71, "learning_rate": 0.00048278916220459764, "loss": 1.8619, "step": 17600 }, { "epoch": 0.71, "learning_rate": 0.0004827790973871734, "loss": 1.6323, "step": 17610 }, { "epoch": 0.71, "learning_rate": 0.00048276903256974917, "loss": 1.6624, "step": 17620 }, { "epoch": 0.71, "learning_rate": 0.00048275896775232496, "loss": 1.9818, "step": 17630 }, { "epoch": 0.71, "learning_rate": 0.00048274890293490075, "loss": 1.5518, "step": 17640 }, { "epoch": 0.71, "learning_rate": 0.00048273883811747654, "loss": 2.1152, "step": 17650 }, { "epoch": 0.71, "learning_rate": 0.00048272877330005233, "loss": 1.7639, "step": 17660 }, { "epoch": 0.71, "learning_rate": 0.0004827187084826281, "loss": 2.1161, "step": 17670 }, { "epoch": 0.71, "learning_rate": 0.00048270864366520397, "loss": 1.9258, "step": 17680 }, { "epoch": 0.71, "learning_rate": 0.00048269857884777976, "loss": 1.584, "step": 17690 }, { "epoch": 0.71, "learning_rate": 0.0004826885140303555, "loss": 1.7846, "step": 17700 }, { "epoch": 0.71, "learning_rate": 0.0004826784492129313, "loss": 1.9284, "step": 17710 }, { "epoch": 0.71, "learning_rate": 0.0004826683843955071, "loss": 1.7664, "step": 17720 }, { "epoch": 0.71, "learning_rate": 0.00048265831957808287, "loss": 1.6784, "step": 17730 }, { "epoch": 0.71, "learning_rate": 0.00048264825476065866, "loss": 2.0149, "step": 17740 }, { "epoch": 0.71, "learning_rate": 0.00048263818994323445, "loss": 1.8208, "step": 17750 }, { "epoch": 0.71, "learning_rate": 0.00048262812512581024, "loss": 1.7533, "step": 17760 }, { "epoch": 0.71, "learning_rate": 0.00048261806030838603, "loss": 1.7818, "step": 17770 }, { "epoch": 0.72, "learning_rate": 0.00048260799549096177, "loss": 2.1214, "step": 17780 }, { "epoch": 0.72, "learning_rate": 0.00048259793067353756, "loss": 1.712, "step": 17790 }, { "epoch": 0.72, "learning_rate": 0.00048258786585611335, "loss": 1.6832, "step": 17800 }, { "epoch": 0.72, "learning_rate": 0.00048257780103868914, "loss": 1.8102, "step": 17810 }, { "epoch": 0.72, "learning_rate": 0.00048256773622126493, "loss": 1.9097, "step": 17820 }, { "epoch": 0.72, "learning_rate": 0.0004825576714038408, "loss": 2.0503, "step": 17830 }, { "epoch": 0.72, "learning_rate": 0.00048254760658641657, "loss": 2.0554, "step": 17840 }, { "epoch": 0.72, "learning_rate": 0.00048253754176899236, "loss": 1.7626, "step": 17850 }, { "epoch": 0.72, "learning_rate": 0.00048252747695156815, "loss": 1.7512, "step": 17860 }, { "epoch": 0.72, "learning_rate": 0.0004825174121341439, "loss": 1.8043, "step": 17870 }, { "epoch": 0.72, "learning_rate": 0.0004825073473167197, "loss": 2.2834, "step": 17880 }, { "epoch": 0.72, "learning_rate": 0.00048249728249929547, "loss": 1.7889, "step": 17890 }, { "epoch": 0.72, "learning_rate": 0.00048248721768187126, "loss": 1.7395, "step": 17900 }, { "epoch": 0.72, "learning_rate": 0.00048247715286444705, "loss": 1.9782, "step": 17910 }, { "epoch": 0.72, "learning_rate": 0.00048246708804702284, "loss": 1.9057, "step": 17920 }, { "epoch": 0.72, "learning_rate": 0.00048245702322959863, "loss": 1.6195, "step": 17930 }, { "epoch": 0.72, "learning_rate": 0.0004824469584121744, "loss": 1.898, "step": 17940 }, { "epoch": 0.72, "learning_rate": 0.00048243689359475016, "loss": 1.9314, "step": 17950 }, { "epoch": 0.72, "learning_rate": 0.00048242682877732595, "loss": 2.0167, "step": 17960 }, { "epoch": 0.72, "learning_rate": 0.00048241676395990174, "loss": 1.9475, "step": 17970 }, { "epoch": 0.72, "learning_rate": 0.0004824066991424776, "loss": 1.7567, "step": 17980 }, { "epoch": 0.72, "learning_rate": 0.0004823966343250534, "loss": 1.8644, "step": 17990 }, { "epoch": 0.72, "learning_rate": 0.00048238656950762917, "loss": 1.9192, "step": 18000 }, { "epoch": 0.72, "learning_rate": 0.00048237650469020496, "loss": 1.8763, "step": 18010 }, { "epoch": 0.72, "learning_rate": 0.00048236643987278075, "loss": 1.8109, "step": 18020 }, { "epoch": 0.73, "learning_rate": 0.00048235637505535654, "loss": 1.7719, "step": 18030 }, { "epoch": 0.73, "learning_rate": 0.0004823463102379323, "loss": 2.0447, "step": 18040 }, { "epoch": 0.73, "learning_rate": 0.00048233624542050806, "loss": 1.8905, "step": 18050 }, { "epoch": 0.73, "learning_rate": 0.00048232618060308386, "loss": 1.5347, "step": 18060 }, { "epoch": 0.73, "learning_rate": 0.00048231611578565965, "loss": 1.5872, "step": 18070 }, { "epoch": 0.73, "learning_rate": 0.00048230605096823544, "loss": 2.0056, "step": 18080 }, { "epoch": 0.73, "learning_rate": 0.00048229598615081123, "loss": 1.6712, "step": 18090 }, { "epoch": 0.73, "learning_rate": 0.000482285921333387, "loss": 1.6708, "step": 18100 }, { "epoch": 0.73, "learning_rate": 0.0004822758565159628, "loss": 1.529, "step": 18110 }, { "epoch": 0.73, "learning_rate": 0.0004822657916985386, "loss": 1.7579, "step": 18120 }, { "epoch": 0.73, "learning_rate": 0.0004822557268811144, "loss": 2.1259, "step": 18130 }, { "epoch": 0.73, "learning_rate": 0.0004822456620636902, "loss": 1.5353, "step": 18140 }, { "epoch": 0.73, "learning_rate": 0.00048223559724626597, "loss": 1.7709, "step": 18150 }, { "epoch": 0.73, "learning_rate": 0.00048222553242884176, "loss": 1.6349, "step": 18160 }, { "epoch": 0.73, "learning_rate": 0.00048221546761141755, "loss": 1.8151, "step": 18170 }, { "epoch": 0.73, "learning_rate": 0.00048220540279399335, "loss": 1.6714, "step": 18180 }, { "epoch": 0.73, "learning_rate": 0.00048219533797656914, "loss": 1.7126, "step": 18190 }, { "epoch": 0.73, "learning_rate": 0.00048218527315914487, "loss": 1.6558, "step": 18200 }, { "epoch": 0.73, "learning_rate": 0.00048217520834172066, "loss": 1.8959, "step": 18210 }, { "epoch": 0.73, "learning_rate": 0.00048216514352429645, "loss": 1.627, "step": 18220 }, { "epoch": 0.73, "learning_rate": 0.00048215507870687224, "loss": 1.6755, "step": 18230 }, { "epoch": 0.73, "learning_rate": 0.00048214501388944804, "loss": 2.0706, "step": 18240 }, { "epoch": 0.73, "learning_rate": 0.0004821349490720238, "loss": 1.7946, "step": 18250 }, { "epoch": 0.73, "learning_rate": 0.0004821248842545996, "loss": 1.6522, "step": 18260 }, { "epoch": 0.73, "learning_rate": 0.00048211481943717546, "loss": 1.8199, "step": 18270 }, { "epoch": 0.74, "learning_rate": 0.00048210475461975125, "loss": 1.6883, "step": 18280 }, { "epoch": 0.74, "learning_rate": 0.000482094689802327, "loss": 1.7381, "step": 18290 }, { "epoch": 0.74, "learning_rate": 0.0004820846249849028, "loss": 1.8715, "step": 18300 }, { "epoch": 0.74, "learning_rate": 0.00048207456016747857, "loss": 1.7519, "step": 18310 }, { "epoch": 0.74, "learning_rate": 0.00048206449535005436, "loss": 1.8746, "step": 18320 }, { "epoch": 0.74, "learning_rate": 0.00048205443053263015, "loss": 2.0285, "step": 18330 }, { "epoch": 0.74, "learning_rate": 0.00048204436571520594, "loss": 1.8198, "step": 18340 }, { "epoch": 0.74, "learning_rate": 0.00048203430089778173, "loss": 1.6866, "step": 18350 }, { "epoch": 0.74, "learning_rate": 0.0004820242360803575, "loss": 1.758, "step": 18360 }, { "epoch": 0.74, "learning_rate": 0.00048201417126293326, "loss": 1.5404, "step": 18370 }, { "epoch": 0.74, "learning_rate": 0.00048200410644550905, "loss": 1.8676, "step": 18380 }, { "epoch": 0.74, "learning_rate": 0.00048199404162808484, "loss": 1.8984, "step": 18390 }, { "epoch": 0.74, "learning_rate": 0.00048198397681066063, "loss": 1.7859, "step": 18400 }, { "epoch": 0.74, "learning_rate": 0.0004819739119932365, "loss": 1.7791, "step": 18410 }, { "epoch": 0.74, "learning_rate": 0.00048196384717581227, "loss": 1.8137, "step": 18420 }, { "epoch": 0.74, "learning_rate": 0.00048195378235838806, "loss": 1.6941, "step": 18430 }, { "epoch": 0.74, "learning_rate": 0.00048194371754096385, "loss": 1.8341, "step": 18440 }, { "epoch": 0.74, "learning_rate": 0.00048193365272353964, "loss": 1.7917, "step": 18450 }, { "epoch": 0.74, "learning_rate": 0.0004819235879061154, "loss": 1.7591, "step": 18460 }, { "epoch": 0.74, "learning_rate": 0.00048191352308869117, "loss": 1.9212, "step": 18470 }, { "epoch": 0.74, "learning_rate": 0.00048190345827126696, "loss": 1.7169, "step": 18480 }, { "epoch": 0.74, "learning_rate": 0.00048189339345384275, "loss": 1.9072, "step": 18490 }, { "epoch": 0.74, "learning_rate": 0.00048188332863641854, "loss": 1.8044, "step": 18500 }, { "epoch": 0.74, "learning_rate": 0.00048187326381899433, "loss": 2.1617, "step": 18510 }, { "epoch": 0.74, "learning_rate": 0.0004818631990015701, "loss": 1.9245, "step": 18520 }, { "epoch": 0.75, "learning_rate": 0.0004818531341841459, "loss": 1.5919, "step": 18530 }, { "epoch": 0.75, "learning_rate": 0.00048184306936672165, "loss": 1.8121, "step": 18540 }, { "epoch": 0.75, "learning_rate": 0.00048183300454929744, "loss": 1.7711, "step": 18550 }, { "epoch": 0.75, "learning_rate": 0.0004818229397318733, "loss": 1.7926, "step": 18560 }, { "epoch": 0.75, "learning_rate": 0.0004818128749144491, "loss": 1.6298, "step": 18570 }, { "epoch": 0.75, "learning_rate": 0.00048180281009702487, "loss": 2.0697, "step": 18580 }, { "epoch": 0.75, "learning_rate": 0.00048179274527960066, "loss": 1.9688, "step": 18590 }, { "epoch": 0.75, "learning_rate": 0.00048178268046217645, "loss": 1.9623, "step": 18600 }, { "epoch": 0.75, "learning_rate": 0.00048177261564475224, "loss": 1.8524, "step": 18610 }, { "epoch": 0.75, "learning_rate": 0.00048176255082732803, "loss": 1.8327, "step": 18620 }, { "epoch": 0.75, "learning_rate": 0.00048175248600990377, "loss": 1.681, "step": 18630 }, { "epoch": 0.75, "learning_rate": 0.00048174242119247956, "loss": 1.4845, "step": 18640 }, { "epoch": 0.75, "learning_rate": 0.00048173235637505535, "loss": 1.793, "step": 18650 }, { "epoch": 0.75, "learning_rate": 0.00048172229155763114, "loss": 1.877, "step": 18660 }, { "epoch": 0.75, "learning_rate": 0.00048171222674020693, "loss": 1.8931, "step": 18670 }, { "epoch": 0.75, "learning_rate": 0.0004817021619227827, "loss": 1.5923, "step": 18680 }, { "epoch": 0.75, "learning_rate": 0.0004816920971053585, "loss": 1.7834, "step": 18690 }, { "epoch": 0.75, "learning_rate": 0.0004816820322879343, "loss": 1.8277, "step": 18700 }, { "epoch": 0.75, "learning_rate": 0.0004816719674705101, "loss": 1.7228, "step": 18710 }, { "epoch": 0.75, "learning_rate": 0.0004816619026530859, "loss": 1.6846, "step": 18720 }, { "epoch": 0.75, "learning_rate": 0.0004816518378356617, "loss": 1.6484, "step": 18730 }, { "epoch": 0.75, "learning_rate": 0.00048164177301823747, "loss": 1.8958, "step": 18740 }, { "epoch": 0.75, "learning_rate": 0.00048163170820081326, "loss": 1.9437, "step": 18750 }, { "epoch": 0.75, "learning_rate": 0.00048162164338338905, "loss": 1.7417, "step": 18760 }, { "epoch": 0.75, "learning_rate": 0.00048161157856596484, "loss": 1.6294, "step": 18770 }, { "epoch": 0.76, "learning_rate": 0.00048160151374854063, "loss": 1.7759, "step": 18780 }, { "epoch": 0.76, "learning_rate": 0.0004815914489311164, "loss": 1.5215, "step": 18790 }, { "epoch": 0.76, "learning_rate": 0.00048158138411369216, "loss": 1.6817, "step": 18800 }, { "epoch": 0.76, "learning_rate": 0.00048157131929626795, "loss": 2.0583, "step": 18810 }, { "epoch": 0.76, "learning_rate": 0.00048156125447884374, "loss": 1.5656, "step": 18820 }, { "epoch": 0.76, "learning_rate": 0.00048155118966141953, "loss": 1.6977, "step": 18830 }, { "epoch": 0.76, "learning_rate": 0.0004815411248439953, "loss": 1.7604, "step": 18840 }, { "epoch": 0.76, "learning_rate": 0.00048153106002657116, "loss": 1.7303, "step": 18850 }, { "epoch": 0.76, "learning_rate": 0.00048152099520914696, "loss": 1.5572, "step": 18860 }, { "epoch": 0.76, "learning_rate": 0.00048151093039172275, "loss": 1.768, "step": 18870 }, { "epoch": 0.76, "learning_rate": 0.0004815008655742985, "loss": 1.6795, "step": 18880 }, { "epoch": 0.76, "learning_rate": 0.0004814908007568743, "loss": 1.701, "step": 18890 }, { "epoch": 0.76, "learning_rate": 0.00048148073593945006, "loss": 1.8594, "step": 18900 }, { "epoch": 0.76, "learning_rate": 0.00048147067112202586, "loss": 1.7513, "step": 18910 }, { "epoch": 0.76, "learning_rate": 0.00048146060630460165, "loss": 1.7535, "step": 18920 }, { "epoch": 0.76, "learning_rate": 0.00048145054148717744, "loss": 1.8751, "step": 18930 }, { "epoch": 0.76, "learning_rate": 0.00048144047666975323, "loss": 1.7944, "step": 18940 }, { "epoch": 0.76, "learning_rate": 0.000481430411852329, "loss": 2.2894, "step": 18950 }, { "epoch": 0.76, "learning_rate": 0.00048142034703490476, "loss": 1.6832, "step": 18960 }, { "epoch": 0.76, "learning_rate": 0.00048141028221748055, "loss": 2.0795, "step": 18970 }, { "epoch": 0.76, "learning_rate": 0.00048140021740005634, "loss": 1.717, "step": 18980 }, { "epoch": 0.76, "learning_rate": 0.00048139015258263213, "loss": 1.7547, "step": 18990 }, { "epoch": 0.76, "learning_rate": 0.00048138008776520797, "loss": 1.7335, "step": 19000 }, { "epoch": 0.76, "learning_rate": 0.00048137002294778376, "loss": 1.4654, "step": 19010 }, { "epoch": 0.76, "learning_rate": 0.00048135995813035955, "loss": 1.6673, "step": 19020 }, { "epoch": 0.77, "learning_rate": 0.00048134989331293534, "loss": 1.7162, "step": 19030 }, { "epoch": 0.77, "learning_rate": 0.00048133982849551114, "loss": 1.8885, "step": 19040 }, { "epoch": 0.77, "learning_rate": 0.00048132976367808687, "loss": 1.712, "step": 19050 }, { "epoch": 0.77, "learning_rate": 0.00048131969886066266, "loss": 1.7632, "step": 19060 }, { "epoch": 0.77, "learning_rate": 0.00048130963404323845, "loss": 1.7822, "step": 19070 }, { "epoch": 0.77, "learning_rate": 0.00048129956922581424, "loss": 1.6597, "step": 19080 }, { "epoch": 0.77, "learning_rate": 0.00048128950440839004, "loss": 1.7172, "step": 19090 }, { "epoch": 0.77, "learning_rate": 0.0004812794395909658, "loss": 1.727, "step": 19100 }, { "epoch": 0.77, "learning_rate": 0.0004812693747735416, "loss": 1.9093, "step": 19110 }, { "epoch": 0.77, "learning_rate": 0.0004812593099561174, "loss": 1.4905, "step": 19120 }, { "epoch": 0.77, "learning_rate": 0.00048124924513869314, "loss": 1.5172, "step": 19130 }, { "epoch": 0.77, "learning_rate": 0.00048123918032126893, "loss": 1.7992, "step": 19140 }, { "epoch": 0.77, "learning_rate": 0.0004812291155038448, "loss": 1.8188, "step": 19150 }, { "epoch": 0.77, "learning_rate": 0.00048121905068642057, "loss": 1.7363, "step": 19160 }, { "epoch": 0.77, "learning_rate": 0.00048120898586899636, "loss": 1.8568, "step": 19170 }, { "epoch": 0.77, "learning_rate": 0.00048119892105157215, "loss": 1.6988, "step": 19180 }, { "epoch": 0.77, "learning_rate": 0.00048118885623414794, "loss": 1.65, "step": 19190 }, { "epoch": 0.77, "learning_rate": 0.00048117879141672373, "loss": 1.4405, "step": 19200 }, { "epoch": 0.77, "learning_rate": 0.0004811687265992995, "loss": 2.0498, "step": 19210 }, { "epoch": 0.77, "learning_rate": 0.00048115866178187526, "loss": 1.895, "step": 19220 }, { "epoch": 0.77, "learning_rate": 0.00048114859696445105, "loss": 2.0904, "step": 19230 }, { "epoch": 0.77, "learning_rate": 0.00048113853214702684, "loss": 1.6916, "step": 19240 }, { "epoch": 0.77, "learning_rate": 0.00048112846732960263, "loss": 1.7728, "step": 19250 }, { "epoch": 0.77, "learning_rate": 0.0004811184025121784, "loss": 1.7652, "step": 19260 }, { "epoch": 0.78, "learning_rate": 0.0004811083376947542, "loss": 1.6244, "step": 19270 }, { "epoch": 0.78, "learning_rate": 0.00048109827287733, "loss": 1.8306, "step": 19280 }, { "epoch": 0.78, "learning_rate": 0.00048108820805990585, "loss": 1.6771, "step": 19290 }, { "epoch": 0.78, "learning_rate": 0.0004810781432424816, "loss": 1.9315, "step": 19300 }, { "epoch": 0.78, "learning_rate": 0.0004810680784250574, "loss": 1.8113, "step": 19310 }, { "epoch": 0.78, "learning_rate": 0.00048105801360763317, "loss": 1.9442, "step": 19320 }, { "epoch": 0.78, "learning_rate": 0.00048104794879020896, "loss": 1.6195, "step": 19330 }, { "epoch": 0.78, "learning_rate": 0.00048103788397278475, "loss": 1.4383, "step": 19340 }, { "epoch": 0.78, "learning_rate": 0.00048102781915536054, "loss": 1.9225, "step": 19350 }, { "epoch": 0.78, "learning_rate": 0.00048101775433793633, "loss": 1.837, "step": 19360 }, { "epoch": 0.78, "learning_rate": 0.0004810076895205121, "loss": 2.1779, "step": 19370 }, { "epoch": 0.78, "learning_rate": 0.0004809976247030879, "loss": 1.7187, "step": 19380 }, { "epoch": 0.78, "learning_rate": 0.00048098755988566365, "loss": 1.6668, "step": 19390 }, { "epoch": 0.78, "learning_rate": 0.00048097749506823944, "loss": 1.7257, "step": 19400 }, { "epoch": 0.78, "learning_rate": 0.00048096743025081523, "loss": 1.9988, "step": 19410 }, { "epoch": 0.78, "learning_rate": 0.000480957365433391, "loss": 1.7581, "step": 19420 }, { "epoch": 0.78, "learning_rate": 0.0004809473006159668, "loss": 2.1415, "step": 19430 }, { "epoch": 0.78, "learning_rate": 0.00048093723579854266, "loss": 1.5141, "step": 19440 }, { "epoch": 0.78, "learning_rate": 0.00048092717098111845, "loss": 1.9565, "step": 19450 }, { "epoch": 0.78, "learning_rate": 0.00048091710616369424, "loss": 1.8374, "step": 19460 }, { "epoch": 0.78, "learning_rate": 0.00048090704134627, "loss": 1.6256, "step": 19470 }, { "epoch": 0.78, "learning_rate": 0.00048089697652884577, "loss": 1.7761, "step": 19480 }, { "epoch": 0.78, "learning_rate": 0.00048088691171142156, "loss": 1.6999, "step": 19490 }, { "epoch": 0.78, "learning_rate": 0.00048087684689399735, "loss": 1.5439, "step": 19500 }, { "epoch": 0.78, "learning_rate": 0.00048086678207657314, "loss": 2.0201, "step": 19510 }, { "epoch": 0.79, "learning_rate": 0.00048085671725914893, "loss": 1.7964, "step": 19520 }, { "epoch": 0.79, "learning_rate": 0.0004808466524417247, "loss": 1.7661, "step": 19530 }, { "epoch": 0.79, "learning_rate": 0.0004808365876243005, "loss": 1.872, "step": 19540 }, { "epoch": 0.79, "learning_rate": 0.0004808265228068763, "loss": 1.7224, "step": 19550 }, { "epoch": 0.79, "learning_rate": 0.00048081645798945204, "loss": 1.743, "step": 19560 }, { "epoch": 0.79, "learning_rate": 0.00048080639317202783, "loss": 1.5759, "step": 19570 }, { "epoch": 0.79, "learning_rate": 0.0004807963283546037, "loss": 1.7173, "step": 19580 }, { "epoch": 0.79, "learning_rate": 0.00048078626353717947, "loss": 1.8918, "step": 19590 }, { "epoch": 0.79, "learning_rate": 0.00048077619871975526, "loss": 1.6998, "step": 19600 }, { "epoch": 0.79, "learning_rate": 0.00048076613390233105, "loss": 1.625, "step": 19610 }, { "epoch": 0.79, "learning_rate": 0.00048075606908490684, "loss": 1.7129, "step": 19620 }, { "epoch": 0.79, "learning_rate": 0.00048074600426748263, "loss": 1.6339, "step": 19630 }, { "epoch": 0.79, "learning_rate": 0.00048073593945005837, "loss": 1.9245, "step": 19640 }, { "epoch": 0.79, "learning_rate": 0.00048072587463263416, "loss": 1.6181, "step": 19650 }, { "epoch": 0.79, "learning_rate": 0.00048071580981520995, "loss": 1.7219, "step": 19660 }, { "epoch": 0.79, "learning_rate": 0.00048070574499778574, "loss": 1.6806, "step": 19670 }, { "epoch": 0.79, "learning_rate": 0.00048069568018036153, "loss": 2.0566, "step": 19680 }, { "epoch": 0.79, "learning_rate": 0.0004806856153629373, "loss": 1.7745, "step": 19690 }, { "epoch": 0.79, "learning_rate": 0.0004806755505455131, "loss": 1.805, "step": 19700 }, { "epoch": 0.79, "learning_rate": 0.0004806654857280889, "loss": 1.6893, "step": 19710 }, { "epoch": 0.79, "learning_rate": 0.00048065542091066464, "loss": 1.5767, "step": 19720 }, { "epoch": 0.79, "learning_rate": 0.0004806453560932405, "loss": 1.6838, "step": 19730 }, { "epoch": 0.79, "learning_rate": 0.0004806352912758163, "loss": 2.2081, "step": 19740 }, { "epoch": 0.79, "learning_rate": 0.00048062522645839206, "loss": 1.7813, "step": 19750 }, { "epoch": 0.79, "learning_rate": 0.00048061516164096786, "loss": 1.9077, "step": 19760 }, { "epoch": 0.8, "learning_rate": 0.00048060509682354365, "loss": 1.82, "step": 19770 }, { "epoch": 0.8, "learning_rate": 0.00048059503200611944, "loss": 1.9815, "step": 19780 }, { "epoch": 0.8, "learning_rate": 0.00048058496718869523, "loss": 1.942, "step": 19790 }, { "epoch": 0.8, "learning_rate": 0.000480574902371271, "loss": 1.8082, "step": 19800 }, { "epoch": 0.8, "learning_rate": 0.00048056483755384675, "loss": 1.6412, "step": 19810 }, { "epoch": 0.8, "learning_rate": 0.00048055477273642255, "loss": 1.8106, "step": 19820 }, { "epoch": 0.8, "learning_rate": 0.00048054470791899834, "loss": 1.7661, "step": 19830 }, { "epoch": 0.8, "learning_rate": 0.00048053464310157413, "loss": 1.9013, "step": 19840 }, { "epoch": 0.8, "learning_rate": 0.0004805245782841499, "loss": 1.6859, "step": 19850 }, { "epoch": 0.8, "learning_rate": 0.0004805145134667257, "loss": 1.5068, "step": 19860 }, { "epoch": 0.8, "learning_rate": 0.0004805044486493015, "loss": 1.8843, "step": 19870 }, { "epoch": 0.8, "learning_rate": 0.00048049438383187734, "loss": 2.0435, "step": 19880 }, { "epoch": 0.8, "learning_rate": 0.00048048431901445314, "loss": 1.833, "step": 19890 }, { "epoch": 0.8, "learning_rate": 0.00048047425419702887, "loss": 1.8238, "step": 19900 }, { "epoch": 0.8, "learning_rate": 0.00048046418937960466, "loss": 1.9807, "step": 19910 }, { "epoch": 0.8, "learning_rate": 0.00048045412456218045, "loss": 1.5295, "step": 19920 }, { "epoch": 0.8, "learning_rate": 0.00048044405974475624, "loss": 1.8328, "step": 19930 }, { "epoch": 0.8, "learning_rate": 0.00048043399492733204, "loss": 1.611, "step": 19940 }, { "epoch": 0.8, "learning_rate": 0.0004804239301099078, "loss": 1.7014, "step": 19950 }, { "epoch": 0.8, "learning_rate": 0.0004804138652924836, "loss": 1.8727, "step": 19960 }, { "epoch": 0.8, "learning_rate": 0.0004804038004750594, "loss": 1.8132, "step": 19970 }, { "epoch": 0.8, "learning_rate": 0.00048039373565763514, "loss": 1.6293, "step": 19980 }, { "epoch": 0.8, "learning_rate": 0.00048038367084021093, "loss": 1.5967, "step": 19990 }, { "epoch": 0.8, "learning_rate": 0.0004803736060227867, "loss": 1.6824, "step": 20000 }, { "epoch": 0.8, "learning_rate": 0.0004803635412053625, "loss": 1.7679, "step": 20010 }, { "epoch": 0.81, "learning_rate": 0.00048035347638793836, "loss": 1.7272, "step": 20020 }, { "epoch": 0.81, "learning_rate": 0.00048034341157051415, "loss": 1.587, "step": 20030 }, { "epoch": 0.81, "learning_rate": 0.00048033334675308994, "loss": 1.7784, "step": 20040 }, { "epoch": 0.81, "learning_rate": 0.00048032328193566573, "loss": 1.7871, "step": 20050 }, { "epoch": 0.81, "learning_rate": 0.00048031321711824147, "loss": 1.879, "step": 20060 }, { "epoch": 0.81, "learning_rate": 0.00048030315230081726, "loss": 1.6503, "step": 20070 }, { "epoch": 0.81, "learning_rate": 0.00048029308748339305, "loss": 1.6911, "step": 20080 }, { "epoch": 0.81, "learning_rate": 0.00048028302266596884, "loss": 1.6983, "step": 20090 }, { "epoch": 0.81, "learning_rate": 0.00048027295784854463, "loss": 1.6236, "step": 20100 }, { "epoch": 0.81, "learning_rate": 0.0004802628930311204, "loss": 1.5174, "step": 20110 }, { "epoch": 0.81, "learning_rate": 0.0004802528282136962, "loss": 1.7465, "step": 20120 }, { "epoch": 0.81, "learning_rate": 0.000480242763396272, "loss": 1.7551, "step": 20130 }, { "epoch": 0.81, "learning_rate": 0.0004802326985788478, "loss": 1.8189, "step": 20140 }, { "epoch": 0.81, "learning_rate": 0.00048022263376142353, "loss": 1.7066, "step": 20150 }, { "epoch": 0.81, "learning_rate": 0.0004802125689439993, "loss": 1.8718, "step": 20160 }, { "epoch": 0.81, "learning_rate": 0.00048020250412657517, "loss": 1.6306, "step": 20170 }, { "epoch": 0.81, "learning_rate": 0.00048019243930915096, "loss": 1.8379, "step": 20180 }, { "epoch": 0.81, "learning_rate": 0.00048018237449172675, "loss": 1.722, "step": 20190 }, { "epoch": 0.81, "learning_rate": 0.00048017230967430254, "loss": 1.8455, "step": 20200 }, { "epoch": 0.81, "learning_rate": 0.00048016224485687833, "loss": 1.8338, "step": 20210 }, { "epoch": 0.81, "learning_rate": 0.0004801521800394541, "loss": 1.7534, "step": 20220 }, { "epoch": 0.81, "learning_rate": 0.00048014211522202986, "loss": 1.7872, "step": 20230 }, { "epoch": 0.81, "learning_rate": 0.00048013205040460565, "loss": 1.3111, "step": 20240 }, { "epoch": 0.81, "learning_rate": 0.00048012198558718144, "loss": 1.6073, "step": 20250 }, { "epoch": 0.81, "learning_rate": 0.00048011192076975723, "loss": 1.6809, "step": 20260 }, { "epoch": 0.82, "learning_rate": 0.000480101855952333, "loss": 1.9215, "step": 20270 }, { "epoch": 0.82, "learning_rate": 0.0004800917911349088, "loss": 1.844, "step": 20280 }, { "epoch": 0.82, "learning_rate": 0.0004800817263174846, "loss": 1.8182, "step": 20290 }, { "epoch": 0.82, "learning_rate": 0.0004800716615000604, "loss": 2.0789, "step": 20300 }, { "epoch": 0.82, "learning_rate": 0.00048006159668263624, "loss": 1.903, "step": 20310 }, { "epoch": 0.82, "learning_rate": 0.000480051531865212, "loss": 1.787, "step": 20320 }, { "epoch": 0.82, "learning_rate": 0.00048004146704778777, "loss": 2.1562, "step": 20330 }, { "epoch": 0.82, "learning_rate": 0.00048003140223036356, "loss": 1.6782, "step": 20340 }, { "epoch": 0.82, "learning_rate": 0.00048002133741293935, "loss": 1.7682, "step": 20350 }, { "epoch": 0.82, "learning_rate": 0.00048001127259551514, "loss": 1.6954, "step": 20360 }, { "epoch": 0.82, "learning_rate": 0.00048000120777809093, "loss": 1.7341, "step": 20370 }, { "epoch": 0.82, "learning_rate": 0.0004799911429606667, "loss": 1.6433, "step": 20380 }, { "epoch": 0.82, "learning_rate": 0.0004799810781432425, "loss": 1.6669, "step": 20390 }, { "epoch": 0.82, "learning_rate": 0.00047997101332581825, "loss": 2.1767, "step": 20400 }, { "epoch": 0.82, "learning_rate": 0.00047996094850839404, "loss": 1.711, "step": 20410 }, { "epoch": 0.82, "learning_rate": 0.00047995088369096983, "loss": 1.8623, "step": 20420 }, { "epoch": 0.82, "learning_rate": 0.0004799408188735456, "loss": 1.7723, "step": 20430 }, { "epoch": 0.82, "learning_rate": 0.0004799307540561214, "loss": 1.9092, "step": 20440 }, { "epoch": 0.82, "learning_rate": 0.0004799206892386972, "loss": 1.5786, "step": 20450 }, { "epoch": 0.82, "learning_rate": 0.00047991062442127305, "loss": 2.0927, "step": 20460 }, { "epoch": 0.82, "learning_rate": 0.00047990055960384884, "loss": 1.9337, "step": 20470 }, { "epoch": 0.82, "learning_rate": 0.00047989049478642463, "loss": 1.8774, "step": 20480 }, { "epoch": 0.82, "learning_rate": 0.00047988042996900037, "loss": 1.7828, "step": 20490 }, { "epoch": 0.82, "learning_rate": 0.00047987036515157616, "loss": 2.1553, "step": 20500 }, { "epoch": 0.82, "learning_rate": 0.00047986030033415195, "loss": 1.8453, "step": 20510 }, { "epoch": 0.83, "learning_rate": 0.00047985023551672774, "loss": 1.731, "step": 20520 }, { "epoch": 0.83, "learning_rate": 0.00047984017069930353, "loss": 1.6836, "step": 20530 }, { "epoch": 0.83, "learning_rate": 0.0004798301058818793, "loss": 1.9146, "step": 20540 }, { "epoch": 0.83, "learning_rate": 0.0004798200410644551, "loss": 1.8436, "step": 20550 }, { "epoch": 0.83, "learning_rate": 0.0004798099762470309, "loss": 1.4822, "step": 20560 }, { "epoch": 0.83, "learning_rate": 0.00047979991142960664, "loss": 1.8033, "step": 20570 }, { "epoch": 0.83, "learning_rate": 0.00047978984661218243, "loss": 1.9664, "step": 20580 }, { "epoch": 0.83, "learning_rate": 0.0004797797817947582, "loss": 1.7018, "step": 20590 }, { "epoch": 0.83, "learning_rate": 0.000479769716977334, "loss": 1.7057, "step": 20600 }, { "epoch": 0.83, "learning_rate": 0.00047975965215990985, "loss": 1.6388, "step": 20610 }, { "epoch": 0.83, "learning_rate": 0.00047974958734248565, "loss": 1.4324, "step": 20620 }, { "epoch": 0.83, "learning_rate": 0.00047973952252506144, "loss": 1.9808, "step": 20630 }, { "epoch": 0.83, "learning_rate": 0.00047972945770763723, "loss": 1.6138, "step": 20640 }, { "epoch": 0.83, "learning_rate": 0.000479719392890213, "loss": 1.5845, "step": 20650 }, { "epoch": 0.83, "learning_rate": 0.00047970932807278875, "loss": 1.8947, "step": 20660 }, { "epoch": 0.83, "learning_rate": 0.00047969926325536455, "loss": 1.6063, "step": 20670 }, { "epoch": 0.83, "learning_rate": 0.00047968919843794034, "loss": 1.5234, "step": 20680 }, { "epoch": 0.83, "learning_rate": 0.0004796791336205161, "loss": 2.1175, "step": 20690 }, { "epoch": 0.83, "learning_rate": 0.0004796690688030919, "loss": 1.6978, "step": 20700 }, { "epoch": 0.83, "learning_rate": 0.0004796590039856677, "loss": 1.3993, "step": 20710 }, { "epoch": 0.83, "learning_rate": 0.0004796489391682435, "loss": 1.8978, "step": 20720 }, { "epoch": 0.83, "learning_rate": 0.0004796388743508193, "loss": 1.7337, "step": 20730 }, { "epoch": 0.83, "learning_rate": 0.000479628809533395, "loss": 1.8939, "step": 20740 }, { "epoch": 0.83, "learning_rate": 0.00047961874471597087, "loss": 1.7016, "step": 20750 }, { "epoch": 0.83, "learning_rate": 0.00047960867989854666, "loss": 1.8759, "step": 20760 }, { "epoch": 0.84, "learning_rate": 0.00047959861508112245, "loss": 1.6484, "step": 20770 }, { "epoch": 0.84, "learning_rate": 0.00047958855026369824, "loss": 1.7847, "step": 20780 }, { "epoch": 0.84, "learning_rate": 0.00047957848544627403, "loss": 1.8138, "step": 20790 }, { "epoch": 0.84, "learning_rate": 0.0004795684206288498, "loss": 1.8714, "step": 20800 }, { "epoch": 0.84, "learning_rate": 0.0004795583558114256, "loss": 1.7602, "step": 20810 }, { "epoch": 0.84, "learning_rate": 0.00047954829099400135, "loss": 2.0186, "step": 20820 }, { "epoch": 0.84, "learning_rate": 0.00047953822617657714, "loss": 1.6146, "step": 20830 }, { "epoch": 0.84, "learning_rate": 0.00047952816135915293, "loss": 1.7971, "step": 20840 }, { "epoch": 0.84, "learning_rate": 0.0004795180965417287, "loss": 1.6289, "step": 20850 }, { "epoch": 0.84, "learning_rate": 0.0004795080317243045, "loss": 1.792, "step": 20860 }, { "epoch": 0.84, "learning_rate": 0.0004794979669068803, "loss": 1.9309, "step": 20870 }, { "epoch": 0.84, "learning_rate": 0.0004794879020894561, "loss": 1.9408, "step": 20880 }, { "epoch": 0.84, "learning_rate": 0.0004794778372720319, "loss": 1.6818, "step": 20890 }, { "epoch": 0.84, "learning_rate": 0.00047946777245460773, "loss": 1.685, "step": 20900 }, { "epoch": 0.84, "learning_rate": 0.00047945770763718347, "loss": 1.8424, "step": 20910 }, { "epoch": 0.84, "learning_rate": 0.00047944764281975926, "loss": 1.9869, "step": 20920 }, { "epoch": 0.84, "learning_rate": 0.00047943757800233505, "loss": 1.6208, "step": 20930 }, { "epoch": 0.84, "learning_rate": 0.00047942751318491084, "loss": 1.5154, "step": 20940 }, { "epoch": 0.84, "learning_rate": 0.00047941744836748663, "loss": 1.6919, "step": 20950 }, { "epoch": 0.84, "learning_rate": 0.0004794073835500624, "loss": 1.7821, "step": 20960 }, { "epoch": 0.84, "learning_rate": 0.0004793973187326382, "loss": 1.8127, "step": 20970 }, { "epoch": 0.84, "learning_rate": 0.000479387253915214, "loss": 1.6685, "step": 20980 }, { "epoch": 0.84, "learning_rate": 0.00047937718909778974, "loss": 2.1184, "step": 20990 }, { "epoch": 0.84, "learning_rate": 0.00047936712428036553, "loss": 2.0116, "step": 21000 }, { "epoch": 0.84, "learning_rate": 0.0004793570594629413, "loss": 1.6976, "step": 21010 }, { "epoch": 0.85, "learning_rate": 0.0004793469946455171, "loss": 1.8305, "step": 21020 }, { "epoch": 0.85, "learning_rate": 0.0004793369298280929, "loss": 1.9742, "step": 21030 }, { "epoch": 0.85, "learning_rate": 0.0004793268650106687, "loss": 1.5566, "step": 21040 }, { "epoch": 0.85, "learning_rate": 0.00047931680019324454, "loss": 1.8222, "step": 21050 }, { "epoch": 0.85, "learning_rate": 0.00047930673537582033, "loss": 1.6456, "step": 21060 }, { "epoch": 0.85, "learning_rate": 0.0004792966705583961, "loss": 1.4809, "step": 21070 }, { "epoch": 0.85, "learning_rate": 0.00047928660574097186, "loss": 1.4938, "step": 21080 }, { "epoch": 0.85, "learning_rate": 0.00047927654092354765, "loss": 1.632, "step": 21090 }, { "epoch": 0.85, "learning_rate": 0.00047926647610612344, "loss": 1.7399, "step": 21100 }, { "epoch": 0.85, "learning_rate": 0.00047925641128869923, "loss": 1.7733, "step": 21110 }, { "epoch": 0.85, "learning_rate": 0.000479246346471275, "loss": 1.7004, "step": 21120 }, { "epoch": 0.85, "learning_rate": 0.0004792362816538508, "loss": 1.9659, "step": 21130 }, { "epoch": 0.85, "learning_rate": 0.0004792262168364266, "loss": 1.7212, "step": 21140 }, { "epoch": 0.85, "learning_rate": 0.0004792161520190024, "loss": 1.7535, "step": 21150 }, { "epoch": 0.85, "learning_rate": 0.00047920608720157813, "loss": 1.2716, "step": 21160 }, { "epoch": 0.85, "learning_rate": 0.0004791960223841539, "loss": 1.5689, "step": 21170 }, { "epoch": 0.85, "learning_rate": 0.0004791859575667297, "loss": 1.69, "step": 21180 }, { "epoch": 0.85, "learning_rate": 0.00047917589274930556, "loss": 1.4733, "step": 21190 }, { "epoch": 0.85, "learning_rate": 0.00047916582793188135, "loss": 1.6065, "step": 21200 }, { "epoch": 0.85, "learning_rate": 0.00047915576311445714, "loss": 1.9175, "step": 21210 }, { "epoch": 0.85, "learning_rate": 0.00047914569829703293, "loss": 1.881, "step": 21220 }, { "epoch": 0.85, "learning_rate": 0.0004791356334796087, "loss": 1.8635, "step": 21230 }, { "epoch": 0.85, "learning_rate": 0.0004791255686621845, "loss": 1.8855, "step": 21240 }, { "epoch": 0.85, "learning_rate": 0.00047911550384476025, "loss": 1.9013, "step": 21250 }, { "epoch": 0.86, "learning_rate": 0.00047910543902733604, "loss": 1.658, "step": 21260 }, { "epoch": 0.86, "learning_rate": 0.00047909537420991183, "loss": 1.5384, "step": 21270 }, { "epoch": 0.86, "learning_rate": 0.0004790853093924876, "loss": 1.8411, "step": 21280 }, { "epoch": 0.86, "learning_rate": 0.0004790752445750634, "loss": 1.6321, "step": 21290 }, { "epoch": 0.86, "learning_rate": 0.0004790651797576392, "loss": 1.7887, "step": 21300 }, { "epoch": 0.86, "learning_rate": 0.000479055114940215, "loss": 1.9156, "step": 21310 }, { "epoch": 0.86, "learning_rate": 0.0004790450501227908, "loss": 1.521, "step": 21320 }, { "epoch": 0.86, "learning_rate": 0.0004790349853053665, "loss": 1.7718, "step": 21330 }, { "epoch": 0.86, "learning_rate": 0.00047902492048794237, "loss": 1.6953, "step": 21340 }, { "epoch": 0.86, "learning_rate": 0.00047901485567051816, "loss": 1.6565, "step": 21350 }, { "epoch": 0.86, "learning_rate": 0.00047900479085309395, "loss": 1.8862, "step": 21360 }, { "epoch": 0.86, "learning_rate": 0.00047899472603566974, "loss": 1.8039, "step": 21370 }, { "epoch": 0.86, "learning_rate": 0.00047898466121824553, "loss": 1.8222, "step": 21380 }, { "epoch": 0.86, "learning_rate": 0.0004789745964008213, "loss": 1.7612, "step": 21390 }, { "epoch": 0.86, "learning_rate": 0.0004789645315833971, "loss": 1.7311, "step": 21400 }, { "epoch": 0.86, "learning_rate": 0.0004789544667659729, "loss": 1.6631, "step": 21410 }, { "epoch": 0.86, "learning_rate": 0.00047894440194854864, "loss": 1.6175, "step": 21420 }, { "epoch": 0.86, "learning_rate": 0.00047893433713112443, "loss": 1.5755, "step": 21430 }, { "epoch": 0.86, "learning_rate": 0.0004789242723137002, "loss": 1.6503, "step": 21440 }, { "epoch": 0.86, "learning_rate": 0.000478914207496276, "loss": 1.8421, "step": 21450 }, { "epoch": 0.86, "learning_rate": 0.0004789041426788518, "loss": 1.6434, "step": 21460 }, { "epoch": 0.86, "learning_rate": 0.0004788940778614276, "loss": 1.976, "step": 21470 }, { "epoch": 0.86, "learning_rate": 0.00047888401304400344, "loss": 1.8875, "step": 21480 }, { "epoch": 0.86, "learning_rate": 0.0004788739482265792, "loss": 1.9005, "step": 21490 }, { "epoch": 0.86, "learning_rate": 0.00047886388340915496, "loss": 1.7676, "step": 21500 }, { "epoch": 0.87, "learning_rate": 0.00047885381859173075, "loss": 2.0477, "step": 21510 }, { "epoch": 0.87, "learning_rate": 0.00047884375377430655, "loss": 1.6056, "step": 21520 }, { "epoch": 0.87, "learning_rate": 0.00047883368895688234, "loss": 1.882, "step": 21530 }, { "epoch": 0.87, "learning_rate": 0.0004788236241394581, "loss": 1.6605, "step": 21540 }, { "epoch": 0.87, "learning_rate": 0.0004788135593220339, "loss": 1.7753, "step": 21550 }, { "epoch": 0.87, "learning_rate": 0.0004788034945046097, "loss": 1.7728, "step": 21560 }, { "epoch": 0.87, "learning_rate": 0.0004787934296871855, "loss": 1.8129, "step": 21570 }, { "epoch": 0.87, "learning_rate": 0.00047878336486976124, "loss": 1.919, "step": 21580 }, { "epoch": 0.87, "learning_rate": 0.000478773300052337, "loss": 1.627, "step": 21590 }, { "epoch": 0.87, "learning_rate": 0.0004787632352349128, "loss": 1.9084, "step": 21600 }, { "epoch": 0.87, "learning_rate": 0.0004787531704174886, "loss": 1.6461, "step": 21610 }, { "epoch": 0.87, "learning_rate": 0.0004787431056000644, "loss": 1.5312, "step": 21620 }, { "epoch": 0.87, "learning_rate": 0.00047873304078264024, "loss": 1.5644, "step": 21630 }, { "epoch": 0.87, "learning_rate": 0.00047872297596521603, "loss": 1.5431, "step": 21640 }, { "epoch": 0.87, "learning_rate": 0.0004787129111477918, "loss": 1.637, "step": 21650 }, { "epoch": 0.87, "learning_rate": 0.0004787028463303676, "loss": 1.7385, "step": 21660 }, { "epoch": 0.87, "learning_rate": 0.00047869278151294335, "loss": 1.6636, "step": 21670 }, { "epoch": 0.87, "learning_rate": 0.00047868271669551914, "loss": 1.9381, "step": 21680 }, { "epoch": 0.87, "learning_rate": 0.00047867265187809493, "loss": 1.7134, "step": 21690 }, { "epoch": 0.87, "learning_rate": 0.0004786625870606707, "loss": 1.7439, "step": 21700 }, { "epoch": 0.87, "learning_rate": 0.0004786525222432465, "loss": 1.9095, "step": 21710 }, { "epoch": 0.87, "learning_rate": 0.0004786424574258223, "loss": 1.7254, "step": 21720 }, { "epoch": 0.87, "learning_rate": 0.0004786323926083981, "loss": 1.5285, "step": 21730 }, { "epoch": 0.87, "learning_rate": 0.0004786223277909739, "loss": 1.7002, "step": 21740 }, { "epoch": 0.87, "learning_rate": 0.0004786122629735496, "loss": 1.74, "step": 21750 }, { "epoch": 0.88, "learning_rate": 0.0004786021981561254, "loss": 1.5203, "step": 21760 }, { "epoch": 0.88, "learning_rate": 0.0004785921333387012, "loss": 1.7974, "step": 21770 }, { "epoch": 0.88, "learning_rate": 0.00047858206852127705, "loss": 1.575, "step": 21780 }, { "epoch": 0.88, "learning_rate": 0.00047857200370385284, "loss": 1.9031, "step": 21790 }, { "epoch": 0.88, "learning_rate": 0.00047856193888642863, "loss": 1.6598, "step": 21800 }, { "epoch": 0.88, "learning_rate": 0.0004785518740690044, "loss": 2.0106, "step": 21810 }, { "epoch": 0.88, "learning_rate": 0.0004785418092515802, "loss": 2.0707, "step": 21820 }, { "epoch": 0.88, "learning_rate": 0.000478531744434156, "loss": 1.7599, "step": 21830 }, { "epoch": 0.88, "learning_rate": 0.00047852167961673174, "loss": 1.6055, "step": 21840 }, { "epoch": 0.88, "learning_rate": 0.00047851161479930753, "loss": 1.8424, "step": 21850 }, { "epoch": 0.88, "learning_rate": 0.0004785015499818833, "loss": 1.8192, "step": 21860 }, { "epoch": 0.88, "learning_rate": 0.0004784914851644591, "loss": 1.899, "step": 21870 }, { "epoch": 0.88, "learning_rate": 0.0004784814203470349, "loss": 1.5418, "step": 21880 }, { "epoch": 0.88, "learning_rate": 0.0004784713555296107, "loss": 1.4351, "step": 21890 }, { "epoch": 0.88, "learning_rate": 0.0004784612907121865, "loss": 1.8633, "step": 21900 }, { "epoch": 0.88, "learning_rate": 0.0004784512258947623, "loss": 1.7044, "step": 21910 }, { "epoch": 0.88, "learning_rate": 0.00047844116107733807, "loss": 1.5058, "step": 21920 }, { "epoch": 0.88, "learning_rate": 0.00047843109625991386, "loss": 1.6865, "step": 21930 }, { "epoch": 0.88, "learning_rate": 0.00047842103144248965, "loss": 1.6514, "step": 21940 }, { "epoch": 0.88, "learning_rate": 0.00047841096662506544, "loss": 1.7775, "step": 21950 }, { "epoch": 0.88, "learning_rate": 0.00047840090180764123, "loss": 1.9235, "step": 21960 }, { "epoch": 0.88, "learning_rate": 0.000478390836990217, "loss": 1.8824, "step": 21970 }, { "epoch": 0.88, "learning_rate": 0.0004783807721727928, "loss": 1.6864, "step": 21980 }, { "epoch": 0.88, "learning_rate": 0.0004783707073553686, "loss": 1.6942, "step": 21990 }, { "epoch": 0.88, "learning_rate": 0.0004783606425379444, "loss": 1.6904, "step": 22000 }, { "epoch": 0.89, "learning_rate": 0.00047835057772052013, "loss": 1.6608, "step": 22010 }, { "epoch": 0.89, "learning_rate": 0.0004783405129030959, "loss": 1.6601, "step": 22020 }, { "epoch": 0.89, "learning_rate": 0.0004783304480856717, "loss": 1.9251, "step": 22030 }, { "epoch": 0.89, "learning_rate": 0.0004783203832682475, "loss": 1.7733, "step": 22040 }, { "epoch": 0.89, "learning_rate": 0.0004783103184508233, "loss": 1.7641, "step": 22050 }, { "epoch": 0.89, "learning_rate": 0.0004783002536333991, "loss": 1.9394, "step": 22060 }, { "epoch": 0.89, "learning_rate": 0.00047829018881597493, "loss": 1.6357, "step": 22070 }, { "epoch": 0.89, "learning_rate": 0.0004782801239985507, "loss": 1.6878, "step": 22080 }, { "epoch": 0.89, "learning_rate": 0.00047827005918112646, "loss": 1.8697, "step": 22090 }, { "epoch": 0.89, "learning_rate": 0.00047825999436370225, "loss": 1.5597, "step": 22100 }, { "epoch": 0.89, "learning_rate": 0.00047824992954627804, "loss": 1.7534, "step": 22110 }, { "epoch": 0.89, "learning_rate": 0.00047823986472885383, "loss": 1.6329, "step": 22120 }, { "epoch": 0.89, "learning_rate": 0.0004782297999114296, "loss": 1.6507, "step": 22130 }, { "epoch": 0.89, "learning_rate": 0.0004782197350940054, "loss": 1.8439, "step": 22140 }, { "epoch": 0.89, "learning_rate": 0.0004782096702765812, "loss": 1.5683, "step": 22150 }, { "epoch": 0.89, "learning_rate": 0.000478199605459157, "loss": 1.8665, "step": 22160 }, { "epoch": 0.89, "learning_rate": 0.0004781895406417328, "loss": 1.8846, "step": 22170 }, { "epoch": 0.89, "learning_rate": 0.0004781794758243085, "loss": 1.8566, "step": 22180 }, { "epoch": 0.89, "learning_rate": 0.0004781694110068843, "loss": 2.0938, "step": 22190 }, { "epoch": 0.89, "learning_rate": 0.0004781593461894601, "loss": 1.823, "step": 22200 }, { "epoch": 0.89, "learning_rate": 0.0004781492813720359, "loss": 1.8707, "step": 22210 }, { "epoch": 0.89, "learning_rate": 0.00047813921655461174, "loss": 1.7154, "step": 22220 }, { "epoch": 0.89, "learning_rate": 0.00047812915173718753, "loss": 1.6801, "step": 22230 }, { "epoch": 0.89, "learning_rate": 0.0004781190869197633, "loss": 1.573, "step": 22240 }, { "epoch": 0.89, "learning_rate": 0.0004781090221023391, "loss": 1.9459, "step": 22250 }, { "epoch": 0.9, "learning_rate": 0.00047809895728491485, "loss": 1.5922, "step": 22260 }, { "epoch": 0.9, "learning_rate": 0.00047808889246749064, "loss": 1.7235, "step": 22270 }, { "epoch": 0.9, "learning_rate": 0.00047807882765006643, "loss": 1.7203, "step": 22280 }, { "epoch": 0.9, "learning_rate": 0.0004780687628326422, "loss": 1.7952, "step": 22290 }, { "epoch": 0.9, "learning_rate": 0.000478058698015218, "loss": 1.7374, "step": 22300 }, { "epoch": 0.9, "learning_rate": 0.0004780486331977938, "loss": 1.6529, "step": 22310 }, { "epoch": 0.9, "learning_rate": 0.0004780385683803696, "loss": 1.8565, "step": 22320 }, { "epoch": 0.9, "learning_rate": 0.0004780285035629454, "loss": 1.7476, "step": 22330 }, { "epoch": 0.9, "learning_rate": 0.0004780184387455211, "loss": 1.7401, "step": 22340 }, { "epoch": 0.9, "learning_rate": 0.0004780083739280969, "loss": 1.6309, "step": 22350 }, { "epoch": 0.9, "learning_rate": 0.00047799830911067275, "loss": 1.6223, "step": 22360 }, { "epoch": 0.9, "learning_rate": 0.00047798824429324854, "loss": 1.6872, "step": 22370 }, { "epoch": 0.9, "learning_rate": 0.00047797817947582434, "loss": 1.623, "step": 22380 }, { "epoch": 0.9, "learning_rate": 0.0004779681146584001, "loss": 1.7352, "step": 22390 }, { "epoch": 0.9, "learning_rate": 0.0004779580498409759, "loss": 1.534, "step": 22400 }, { "epoch": 0.9, "learning_rate": 0.0004779479850235517, "loss": 1.9479, "step": 22410 }, { "epoch": 0.9, "learning_rate": 0.0004779379202061275, "loss": 1.8718, "step": 22420 }, { "epoch": 0.9, "learning_rate": 0.00047792785538870324, "loss": 1.6931, "step": 22430 }, { "epoch": 0.9, "learning_rate": 0.000477917790571279, "loss": 1.5843, "step": 22440 }, { "epoch": 0.9, "learning_rate": 0.0004779077257538548, "loss": 1.6997, "step": 22450 }, { "epoch": 0.9, "learning_rate": 0.0004778976609364306, "loss": 1.7302, "step": 22460 }, { "epoch": 0.9, "learning_rate": 0.0004778875961190064, "loss": 1.8374, "step": 22470 }, { "epoch": 0.9, "learning_rate": 0.0004778775313015822, "loss": 1.5689, "step": 22480 }, { "epoch": 0.9, "learning_rate": 0.000477867466484158, "loss": 1.6944, "step": 22490 }, { "epoch": 0.9, "learning_rate": 0.00047785740166673377, "loss": 1.5113, "step": 22500 }, { "epoch": 0.91, "learning_rate": 0.00047784733684930956, "loss": 1.7286, "step": 22510 }, { "epoch": 0.91, "learning_rate": 0.00047783727203188535, "loss": 1.6209, "step": 22520 }, { "epoch": 0.91, "learning_rate": 0.00047782720721446114, "loss": 1.7001, "step": 22530 }, { "epoch": 0.91, "learning_rate": 0.00047781714239703693, "loss": 1.6824, "step": 22540 }, { "epoch": 0.91, "learning_rate": 0.0004778070775796127, "loss": 1.6417, "step": 22550 }, { "epoch": 0.91, "learning_rate": 0.0004777970127621885, "loss": 1.6594, "step": 22560 }, { "epoch": 0.91, "learning_rate": 0.0004777869479447643, "loss": 1.2907, "step": 22570 }, { "epoch": 0.91, "learning_rate": 0.0004777768831273401, "loss": 1.7024, "step": 22580 }, { "epoch": 0.91, "learning_rate": 0.0004777668183099159, "loss": 1.7542, "step": 22590 }, { "epoch": 0.91, "learning_rate": 0.0004777567534924916, "loss": 1.9451, "step": 22600 }, { "epoch": 0.91, "learning_rate": 0.0004777466886750674, "loss": 1.8588, "step": 22610 }, { "epoch": 0.91, "learning_rate": 0.0004777366238576432, "loss": 1.7312, "step": 22620 }, { "epoch": 0.91, "learning_rate": 0.000477726559040219, "loss": 1.7507, "step": 22630 }, { "epoch": 0.91, "learning_rate": 0.0004777164942227948, "loss": 1.4893, "step": 22640 }, { "epoch": 0.91, "learning_rate": 0.00047770642940537063, "loss": 1.5942, "step": 22650 }, { "epoch": 0.91, "learning_rate": 0.0004776963645879464, "loss": 1.4801, "step": 22660 }, { "epoch": 0.91, "learning_rate": 0.0004776862997705222, "loss": 1.5994, "step": 22670 }, { "epoch": 0.91, "learning_rate": 0.00047767623495309795, "loss": 1.7879, "step": 22680 }, { "epoch": 0.91, "learning_rate": 0.00047766617013567374, "loss": 1.4551, "step": 22690 }, { "epoch": 0.91, "learning_rate": 0.00047765610531824953, "loss": 1.7744, "step": 22700 }, { "epoch": 0.91, "learning_rate": 0.0004776460405008253, "loss": 1.7043, "step": 22710 }, { "epoch": 0.91, "learning_rate": 0.0004776359756834011, "loss": 1.5656, "step": 22720 }, { "epoch": 0.91, "learning_rate": 0.0004776259108659769, "loss": 1.7037, "step": 22730 }, { "epoch": 0.91, "learning_rate": 0.0004776158460485527, "loss": 1.7154, "step": 22740 }, { "epoch": 0.91, "learning_rate": 0.0004776057812311285, "loss": 1.5815, "step": 22750 }, { "epoch": 0.92, "learning_rate": 0.0004775957164137043, "loss": 1.9043, "step": 22760 }, { "epoch": 0.92, "learning_rate": 0.00047758565159628, "loss": 1.7611, "step": 22770 }, { "epoch": 0.92, "learning_rate": 0.0004775755867788558, "loss": 1.5935, "step": 22780 }, { "epoch": 0.92, "learning_rate": 0.0004775655219614316, "loss": 1.7751, "step": 22790 }, { "epoch": 0.92, "learning_rate": 0.00047755545714400744, "loss": 1.67, "step": 22800 }, { "epoch": 0.92, "learning_rate": 0.00047754539232658323, "loss": 1.6061, "step": 22810 }, { "epoch": 0.92, "learning_rate": 0.000477535327509159, "loss": 1.8593, "step": 22820 }, { "epoch": 0.92, "learning_rate": 0.0004775252626917348, "loss": 1.741, "step": 22830 }, { "epoch": 0.92, "learning_rate": 0.0004775151978743106, "loss": 1.8082, "step": 22840 }, { "epoch": 0.92, "learning_rate": 0.00047750513305688634, "loss": 2.0207, "step": 22850 }, { "epoch": 0.92, "learning_rate": 0.00047749506823946213, "loss": 1.557, "step": 22860 }, { "epoch": 0.92, "learning_rate": 0.0004774850034220379, "loss": 1.7656, "step": 22870 }, { "epoch": 0.92, "learning_rate": 0.0004774749386046137, "loss": 1.2291, "step": 22880 }, { "epoch": 0.92, "learning_rate": 0.0004774648737871895, "loss": 1.8986, "step": 22890 }, { "epoch": 0.92, "learning_rate": 0.0004774548089697653, "loss": 1.9997, "step": 22900 }, { "epoch": 0.92, "learning_rate": 0.0004774447441523411, "loss": 1.6599, "step": 22910 }, { "epoch": 0.92, "learning_rate": 0.0004774346793349169, "loss": 1.6168, "step": 22920 }, { "epoch": 0.92, "learning_rate": 0.00047742461451749267, "loss": 1.6789, "step": 22930 }, { "epoch": 0.92, "learning_rate": 0.0004774145497000684, "loss": 1.6699, "step": 22940 }, { "epoch": 0.92, "learning_rate": 0.00047740448488264425, "loss": 1.6745, "step": 22950 }, { "epoch": 0.92, "learning_rate": 0.00047739442006522004, "loss": 1.4786, "step": 22960 }, { "epoch": 0.92, "learning_rate": 0.00047738435524779583, "loss": 1.6596, "step": 22970 }, { "epoch": 0.92, "learning_rate": 0.0004773742904303716, "loss": 1.7191, "step": 22980 }, { "epoch": 0.92, "learning_rate": 0.0004773642256129474, "loss": 1.6848, "step": 22990 }, { "epoch": 0.93, "learning_rate": 0.0004773541607955232, "loss": 1.6693, "step": 23000 }, { "epoch": 0.93, "learning_rate": 0.000477344095978099, "loss": 1.6479, "step": 23010 }, { "epoch": 0.93, "learning_rate": 0.00047733403116067473, "loss": 1.5552, "step": 23020 }, { "epoch": 0.93, "learning_rate": 0.0004773239663432505, "loss": 1.5248, "step": 23030 }, { "epoch": 0.93, "learning_rate": 0.0004773139015258263, "loss": 1.8279, "step": 23040 }, { "epoch": 0.93, "learning_rate": 0.0004773038367084021, "loss": 1.7001, "step": 23050 }, { "epoch": 0.93, "learning_rate": 0.0004772937718909779, "loss": 1.4535, "step": 23060 }, { "epoch": 0.93, "learning_rate": 0.0004772837070735537, "loss": 1.8536, "step": 23070 }, { "epoch": 0.93, "learning_rate": 0.0004772736422561295, "loss": 1.605, "step": 23080 }, { "epoch": 0.93, "learning_rate": 0.0004772635774387053, "loss": 1.6134, "step": 23090 }, { "epoch": 0.93, "learning_rate": 0.0004772535126212811, "loss": 1.7528, "step": 23100 }, { "epoch": 0.93, "learning_rate": 0.00047724344780385685, "loss": 1.7289, "step": 23110 }, { "epoch": 0.93, "learning_rate": 0.00047723338298643264, "loss": 1.663, "step": 23120 }, { "epoch": 0.93, "learning_rate": 0.00047722331816900843, "loss": 1.8607, "step": 23130 }, { "epoch": 0.93, "learning_rate": 0.0004772132533515842, "loss": 1.6549, "step": 23140 }, { "epoch": 0.93, "learning_rate": 0.00047720318853416, "loss": 1.7039, "step": 23150 }, { "epoch": 0.93, "learning_rate": 0.0004771931237167358, "loss": 1.9024, "step": 23160 }, { "epoch": 0.93, "learning_rate": 0.0004771830588993116, "loss": 1.8529, "step": 23170 }, { "epoch": 0.93, "learning_rate": 0.0004771729940818874, "loss": 1.6393, "step": 23180 }, { "epoch": 0.93, "learning_rate": 0.0004771629292644631, "loss": 1.7634, "step": 23190 }, { "epoch": 0.93, "learning_rate": 0.0004771528644470389, "loss": 1.5781, "step": 23200 }, { "epoch": 0.93, "learning_rate": 0.0004771427996296147, "loss": 1.9613, "step": 23210 }, { "epoch": 0.93, "learning_rate": 0.0004771327348121905, "loss": 1.6326, "step": 23220 }, { "epoch": 0.93, "learning_rate": 0.0004771226699947663, "loss": 1.4988, "step": 23230 }, { "epoch": 0.93, "learning_rate": 0.0004771126051773421, "loss": 2.2214, "step": 23240 }, { "epoch": 0.94, "learning_rate": 0.0004771025403599179, "loss": 1.6668, "step": 23250 }, { "epoch": 0.94, "learning_rate": 0.0004770924755424937, "loss": 1.6557, "step": 23260 }, { "epoch": 0.94, "learning_rate": 0.00047708241072506944, "loss": 1.653, "step": 23270 }, { "epoch": 0.94, "learning_rate": 0.00047707234590764524, "loss": 1.9763, "step": 23280 }, { "epoch": 0.94, "learning_rate": 0.000477062281090221, "loss": 1.683, "step": 23290 }, { "epoch": 0.94, "learning_rate": 0.0004770522162727968, "loss": 1.7593, "step": 23300 }, { "epoch": 0.94, "learning_rate": 0.0004770421514553726, "loss": 1.7505, "step": 23310 }, { "epoch": 0.94, "learning_rate": 0.0004770320866379484, "loss": 1.8922, "step": 23320 }, { "epoch": 0.94, "learning_rate": 0.0004770220218205242, "loss": 2.1026, "step": 23330 }, { "epoch": 0.94, "learning_rate": 0.0004770119570031, "loss": 1.8504, "step": 23340 }, { "epoch": 0.94, "learning_rate": 0.00047700189218567577, "loss": 1.5745, "step": 23350 }, { "epoch": 0.94, "learning_rate": 0.0004769918273682515, "loss": 1.9392, "step": 23360 }, { "epoch": 0.94, "learning_rate": 0.0004769817625508273, "loss": 1.7622, "step": 23370 }, { "epoch": 0.94, "learning_rate": 0.00047697169773340314, "loss": 1.6497, "step": 23380 }, { "epoch": 0.94, "learning_rate": 0.00047696163291597893, "loss": 1.4901, "step": 23390 }, { "epoch": 0.94, "learning_rate": 0.0004769515680985547, "loss": 1.9966, "step": 23400 }, { "epoch": 0.94, "learning_rate": 0.0004769415032811305, "loss": 1.5271, "step": 23410 }, { "epoch": 0.94, "learning_rate": 0.0004769314384637063, "loss": 1.6609, "step": 23420 }, { "epoch": 0.94, "learning_rate": 0.0004769213736462821, "loss": 1.7354, "step": 23430 }, { "epoch": 0.94, "learning_rate": 0.00047691130882885783, "loss": 1.5543, "step": 23440 }, { "epoch": 0.94, "learning_rate": 0.0004769012440114336, "loss": 1.971, "step": 23450 }, { "epoch": 0.94, "learning_rate": 0.0004768911791940094, "loss": 1.5681, "step": 23460 }, { "epoch": 0.94, "learning_rate": 0.0004768811143765852, "loss": 2.0374, "step": 23470 }, { "epoch": 0.94, "learning_rate": 0.000476871049559161, "loss": 1.686, "step": 23480 }, { "epoch": 0.94, "learning_rate": 0.0004768609847417368, "loss": 1.5195, "step": 23490 }, { "epoch": 0.95, "learning_rate": 0.0004768509199243126, "loss": 1.5206, "step": 23500 }, { "epoch": 0.95, "learning_rate": 0.00047684085510688837, "loss": 1.6958, "step": 23510 }, { "epoch": 0.95, "learning_rate": 0.00047683079028946416, "loss": 1.7635, "step": 23520 }, { "epoch": 0.95, "learning_rate": 0.00047682072547203995, "loss": 1.8178, "step": 23530 }, { "epoch": 0.95, "learning_rate": 0.00047681066065461574, "loss": 1.6473, "step": 23540 }, { "epoch": 0.95, "learning_rate": 0.00047680059583719153, "loss": 1.8284, "step": 23550 }, { "epoch": 0.95, "learning_rate": 0.0004767905310197673, "loss": 1.6852, "step": 23560 }, { "epoch": 0.95, "learning_rate": 0.0004767804662023431, "loss": 1.6959, "step": 23570 }, { "epoch": 0.95, "learning_rate": 0.0004767704013849189, "loss": 1.5884, "step": 23580 }, { "epoch": 0.95, "learning_rate": 0.0004767603365674947, "loss": 2.0183, "step": 23590 }, { "epoch": 0.95, "learning_rate": 0.0004767502717500705, "loss": 1.5797, "step": 23600 }, { "epoch": 0.95, "learning_rate": 0.0004767402069326462, "loss": 1.6873, "step": 23610 }, { "epoch": 0.95, "learning_rate": 0.000476730142115222, "loss": 1.7041, "step": 23620 }, { "epoch": 0.95, "learning_rate": 0.0004767200772977978, "loss": 1.6943, "step": 23630 }, { "epoch": 0.95, "learning_rate": 0.0004767100124803736, "loss": 1.5537, "step": 23640 }, { "epoch": 0.95, "learning_rate": 0.0004766999476629494, "loss": 2.1316, "step": 23650 }, { "epoch": 0.95, "learning_rate": 0.0004766898828455252, "loss": 1.5615, "step": 23660 }, { "epoch": 0.95, "learning_rate": 0.00047667981802810097, "loss": 1.9485, "step": 23670 }, { "epoch": 0.95, "learning_rate": 0.0004766697532106768, "loss": 1.5253, "step": 23680 }, { "epoch": 0.95, "learning_rate": 0.0004766596883932526, "loss": 1.598, "step": 23690 }, { "epoch": 0.95, "learning_rate": 0.00047664962357582834, "loss": 1.5006, "step": 23700 }, { "epoch": 0.95, "learning_rate": 0.00047663955875840413, "loss": 1.6845, "step": 23710 }, { "epoch": 0.95, "learning_rate": 0.0004766294939409799, "loss": 1.6009, "step": 23720 }, { "epoch": 0.95, "learning_rate": 0.0004766194291235557, "loss": 1.7124, "step": 23730 }, { "epoch": 0.95, "learning_rate": 0.0004766093643061315, "loss": 1.7869, "step": 23740 }, { "epoch": 0.96, "learning_rate": 0.0004765992994887073, "loss": 1.7395, "step": 23750 }, { "epoch": 0.96, "learning_rate": 0.0004765892346712831, "loss": 1.7974, "step": 23760 }, { "epoch": 0.96, "learning_rate": 0.0004765791698538589, "loss": 1.5015, "step": 23770 }, { "epoch": 0.96, "learning_rate": 0.0004765691050364346, "loss": 1.5939, "step": 23780 }, { "epoch": 0.96, "learning_rate": 0.0004765590402190104, "loss": 1.6755, "step": 23790 }, { "epoch": 0.96, "learning_rate": 0.0004765489754015862, "loss": 1.4104, "step": 23800 }, { "epoch": 0.96, "learning_rate": 0.000476538910584162, "loss": 1.6016, "step": 23810 }, { "epoch": 0.96, "learning_rate": 0.00047652884576673783, "loss": 1.5687, "step": 23820 }, { "epoch": 0.96, "learning_rate": 0.0004765187809493136, "loss": 1.739, "step": 23830 }, { "epoch": 0.96, "learning_rate": 0.0004765087161318894, "loss": 1.6579, "step": 23840 }, { "epoch": 0.96, "learning_rate": 0.0004764986513144652, "loss": 1.6476, "step": 23850 }, { "epoch": 0.96, "learning_rate": 0.000476488586497041, "loss": 1.5353, "step": 23860 }, { "epoch": 0.96, "learning_rate": 0.00047647852167961673, "loss": 1.687, "step": 23870 }, { "epoch": 0.96, "learning_rate": 0.0004764684568621925, "loss": 1.9383, "step": 23880 }, { "epoch": 0.96, "learning_rate": 0.0004764583920447683, "loss": 1.6774, "step": 23890 }, { "epoch": 0.96, "learning_rate": 0.0004764483272273441, "loss": 1.5171, "step": 23900 }, { "epoch": 0.96, "learning_rate": 0.0004764382624099199, "loss": 1.539, "step": 23910 }, { "epoch": 0.96, "learning_rate": 0.0004764281975924957, "loss": 1.4872, "step": 23920 }, { "epoch": 0.96, "learning_rate": 0.0004764181327750715, "loss": 1.7563, "step": 23930 }, { "epoch": 0.96, "learning_rate": 0.00047640806795764726, "loss": 1.8445, "step": 23940 }, { "epoch": 0.96, "learning_rate": 0.000476398003140223, "loss": 1.7864, "step": 23950 }, { "epoch": 0.96, "learning_rate": 0.0004763879383227988, "loss": 1.7819, "step": 23960 }, { "epoch": 0.96, "learning_rate": 0.00047637787350537464, "loss": 1.5611, "step": 23970 }, { "epoch": 0.96, "learning_rate": 0.00047636780868795043, "loss": 1.8754, "step": 23980 }, { "epoch": 0.96, "learning_rate": 0.0004763577438705262, "loss": 1.9849, "step": 23990 }, { "epoch": 0.97, "learning_rate": 0.000476347679053102, "loss": 1.8905, "step": 24000 }, { "epoch": 0.97, "learning_rate": 0.0004763376142356778, "loss": 1.6368, "step": 24010 }, { "epoch": 0.97, "learning_rate": 0.0004763275494182536, "loss": 1.7349, "step": 24020 }, { "epoch": 0.97, "learning_rate": 0.0004763174846008293, "loss": 1.807, "step": 24030 }, { "epoch": 0.97, "learning_rate": 0.0004763074197834051, "loss": 1.7254, "step": 24040 }, { "epoch": 0.97, "learning_rate": 0.0004762973549659809, "loss": 1.9493, "step": 24050 }, { "epoch": 0.97, "learning_rate": 0.0004762872901485567, "loss": 1.5725, "step": 24060 }, { "epoch": 0.97, "learning_rate": 0.0004762772253311325, "loss": 1.6519, "step": 24070 }, { "epoch": 0.97, "learning_rate": 0.0004762671605137083, "loss": 1.6335, "step": 24080 }, { "epoch": 0.97, "learning_rate": 0.00047625709569628407, "loss": 1.6181, "step": 24090 }, { "epoch": 0.97, "learning_rate": 0.00047624703087885986, "loss": 1.9877, "step": 24100 }, { "epoch": 0.97, "learning_rate": 0.00047623696606143565, "loss": 1.5691, "step": 24110 }, { "epoch": 0.97, "learning_rate": 0.00047622690124401144, "loss": 1.6642, "step": 24120 }, { "epoch": 0.97, "learning_rate": 0.00047621683642658723, "loss": 1.7522, "step": 24130 }, { "epoch": 0.97, "learning_rate": 0.000476206771609163, "loss": 1.8215, "step": 24140 }, { "epoch": 0.97, "learning_rate": 0.0004761967067917388, "loss": 1.6878, "step": 24150 }, { "epoch": 0.97, "learning_rate": 0.0004761866419743146, "loss": 1.7221, "step": 24160 }, { "epoch": 0.97, "learning_rate": 0.0004761765771568904, "loss": 1.675, "step": 24170 }, { "epoch": 0.97, "learning_rate": 0.0004761665123394662, "loss": 1.7387, "step": 24180 }, { "epoch": 0.97, "learning_rate": 0.000476156447522042, "loss": 1.3828, "step": 24190 }, { "epoch": 0.97, "learning_rate": 0.0004761463827046177, "loss": 1.8446, "step": 24200 }, { "epoch": 0.97, "learning_rate": 0.0004761363178871935, "loss": 1.6308, "step": 24210 }, { "epoch": 0.97, "learning_rate": 0.0004761262530697693, "loss": 1.6777, "step": 24220 }, { "epoch": 0.97, "learning_rate": 0.0004761161882523451, "loss": 1.762, "step": 24230 }, { "epoch": 0.97, "learning_rate": 0.0004761061234349209, "loss": 1.6339, "step": 24240 }, { "epoch": 0.98, "learning_rate": 0.00047609605861749667, "loss": 1.7809, "step": 24250 }, { "epoch": 0.98, "learning_rate": 0.0004760859938000725, "loss": 1.7576, "step": 24260 }, { "epoch": 0.98, "learning_rate": 0.0004760759289826483, "loss": 1.7916, "step": 24270 }, { "epoch": 0.98, "learning_rate": 0.0004760658641652241, "loss": 1.568, "step": 24280 }, { "epoch": 0.98, "learning_rate": 0.00047605579934779983, "loss": 1.2234, "step": 24290 }, { "epoch": 0.98, "learning_rate": 0.0004760457345303756, "loss": 1.3311, "step": 24300 }, { "epoch": 0.98, "learning_rate": 0.0004760356697129514, "loss": 1.6431, "step": 24310 }, { "epoch": 0.98, "learning_rate": 0.0004760256048955272, "loss": 1.4113, "step": 24320 }, { "epoch": 0.98, "learning_rate": 0.000476015540078103, "loss": 1.5412, "step": 24330 }, { "epoch": 0.98, "learning_rate": 0.0004760054752606788, "loss": 1.6648, "step": 24340 }, { "epoch": 0.98, "learning_rate": 0.0004759954104432546, "loss": 1.5412, "step": 24350 }, { "epoch": 0.98, "learning_rate": 0.00047598534562583037, "loss": 1.6115, "step": 24360 }, { "epoch": 0.98, "learning_rate": 0.0004759752808084061, "loss": 1.7025, "step": 24370 }, { "epoch": 0.98, "learning_rate": 0.0004759652159909819, "loss": 1.5887, "step": 24380 }, { "epoch": 0.98, "learning_rate": 0.0004759551511735577, "loss": 1.7808, "step": 24390 }, { "epoch": 0.98, "learning_rate": 0.0004759450863561335, "loss": 1.9056, "step": 24400 }, { "epoch": 0.98, "learning_rate": 0.0004759350215387093, "loss": 1.6435, "step": 24410 }, { "epoch": 0.98, "learning_rate": 0.0004759249567212851, "loss": 1.7277, "step": 24420 }, { "epoch": 0.98, "learning_rate": 0.0004759148919038609, "loss": 1.9069, "step": 24430 }, { "epoch": 0.98, "learning_rate": 0.0004759048270864367, "loss": 1.6058, "step": 24440 }, { "epoch": 0.98, "learning_rate": 0.0004758947622690125, "loss": 1.7079, "step": 24450 }, { "epoch": 0.98, "learning_rate": 0.0004758846974515882, "loss": 1.7479, "step": 24460 }, { "epoch": 0.98, "learning_rate": 0.000475874632634164, "loss": 1.6151, "step": 24470 }, { "epoch": 0.98, "learning_rate": 0.0004758645678167398, "loss": 1.9701, "step": 24480 }, { "epoch": 0.98, "learning_rate": 0.0004758545029993156, "loss": 1.6308, "step": 24490 }, { "epoch": 0.99, "learning_rate": 0.0004758444381818914, "loss": 1.4165, "step": 24500 }, { "epoch": 0.99, "learning_rate": 0.0004758343733644672, "loss": 1.6522, "step": 24510 }, { "epoch": 0.99, "learning_rate": 0.00047582430854704297, "loss": 1.7222, "step": 24520 }, { "epoch": 0.99, "learning_rate": 0.00047581424372961876, "loss": 1.4974, "step": 24530 }, { "epoch": 0.99, "learning_rate": 0.0004758041789121945, "loss": 1.5964, "step": 24540 }, { "epoch": 0.99, "learning_rate": 0.00047579411409477034, "loss": 1.7655, "step": 24550 }, { "epoch": 0.99, "learning_rate": 0.00047578404927734613, "loss": 1.7705, "step": 24560 }, { "epoch": 0.99, "learning_rate": 0.0004757739844599219, "loss": 1.6745, "step": 24570 }, { "epoch": 0.99, "learning_rate": 0.0004757639196424977, "loss": 1.652, "step": 24580 }, { "epoch": 0.99, "learning_rate": 0.0004757538548250735, "loss": 1.9004, "step": 24590 }, { "epoch": 0.99, "learning_rate": 0.0004757437900076493, "loss": 1.6223, "step": 24600 }, { "epoch": 0.99, "learning_rate": 0.0004757337251902251, "loss": 1.8116, "step": 24610 }, { "epoch": 0.99, "learning_rate": 0.0004757236603728009, "loss": 1.7116, "step": 24620 }, { "epoch": 0.99, "learning_rate": 0.0004757135955553766, "loss": 1.809, "step": 24630 }, { "epoch": 0.99, "learning_rate": 0.0004757035307379524, "loss": 1.7221, "step": 24640 }, { "epoch": 0.99, "learning_rate": 0.0004756934659205282, "loss": 1.436, "step": 24650 }, { "epoch": 0.99, "learning_rate": 0.000475683401103104, "loss": 1.542, "step": 24660 }, { "epoch": 0.99, "learning_rate": 0.0004756733362856798, "loss": 1.6264, "step": 24670 }, { "epoch": 0.99, "learning_rate": 0.00047566327146825557, "loss": 1.818, "step": 24680 }, { "epoch": 0.99, "learning_rate": 0.00047565320665083136, "loss": 1.6045, "step": 24690 }, { "epoch": 0.99, "learning_rate": 0.0004756431418334072, "loss": 1.5889, "step": 24700 }, { "epoch": 0.99, "learning_rate": 0.00047563307701598294, "loss": 1.6305, "step": 24710 }, { "epoch": 0.99, "learning_rate": 0.00047562301219855873, "loss": 1.6165, "step": 24720 }, { "epoch": 0.99, "learning_rate": 0.0004756129473811345, "loss": 1.5506, "step": 24730 }, { "epoch": 1.0, "learning_rate": 0.0004756028825637103, "loss": 1.4672, "step": 24740 }, { "epoch": 1.0, "learning_rate": 0.0004755928177462861, "loss": 1.7052, "step": 24750 }, { "epoch": 1.0, "learning_rate": 0.0004755827529288619, "loss": 1.7579, "step": 24760 }, { "epoch": 1.0, "learning_rate": 0.0004755726881114377, "loss": 1.6645, "step": 24770 }, { "epoch": 1.0, "learning_rate": 0.0004755626232940135, "loss": 1.5457, "step": 24780 }, { "epoch": 1.0, "learning_rate": 0.0004755525584765892, "loss": 1.5194, "step": 24790 }, { "epoch": 1.0, "learning_rate": 0.000475542493659165, "loss": 1.614, "step": 24800 }, { "epoch": 1.0, "learning_rate": 0.0004755324288417408, "loss": 1.7339, "step": 24810 }, { "epoch": 1.0, "learning_rate": 0.0004755223640243166, "loss": 1.4166, "step": 24820 }, { "epoch": 1.0, "learning_rate": 0.00047551229920689237, "loss": 1.5849, "step": 24830 }, { "epoch": 1.0, "learning_rate": 0.00047550223438946816, "loss": 1.5521, "step": 24840 }, { "epoch": 1.0, "learning_rate": 0.000475492169572044, "loss": 1.3629, "step": 24850 }, { "epoch": 1.0, "learning_rate": 0.0004754821047546198, "loss": 1.742, "step": 24860 }, { "epoch": 1.0, "eval_Bleu_1": 0.012076940508404375, "eval_Bleu_2": 1.602388757671511e-11, "eval_Bleu_3": 2.0142764772967107e-14, "eval_Bleu_4": 7.848614688061354e-16, "eval_ROUGE_L": 0.02656501355770769, "eval_cer": 0.7333478381265082, "eval_em": 0.25107204116638077, "eval_f1": 0.35158651774170213, "eval_loss": 1.953871250152588, "eval_runtime": 2101.7999, "eval_samples_per_second": 13.314, "eval_steps_per_second": 4.438, "eval_wer": 0.8833244053385609, "step": 24864 }, { "epoch": 1.0, "learning_rate": 0.0004754720399371956, "loss": 1.422, "step": 24870 }, { "epoch": 1.0, "learning_rate": 0.0004754619751197713, "loss": 1.519, "step": 24880 }, { "epoch": 1.0, "learning_rate": 0.0004754519103023471, "loss": 1.4818, "step": 24890 }, { "epoch": 1.0, "learning_rate": 0.0004754418454849229, "loss": 1.3813, "step": 24900 }, { "epoch": 1.0, "learning_rate": 0.0004754317806674987, "loss": 1.4157, "step": 24910 }, { "epoch": 1.0, "learning_rate": 0.0004754217158500745, "loss": 1.4756, "step": 24920 }, { "epoch": 1.0, "learning_rate": 0.0004754116510326503, "loss": 1.3879, "step": 24930 }, { "epoch": 1.0, "learning_rate": 0.00047540158621522607, "loss": 1.3514, "step": 24940 }, { "epoch": 1.0, "learning_rate": 0.00047539152139780186, "loss": 1.4703, "step": 24950 }, { "epoch": 1.0, "learning_rate": 0.0004753814565803776, "loss": 1.3097, "step": 24960 }, { "epoch": 1.0, "learning_rate": 0.0004753713917629534, "loss": 1.4286, "step": 24970 }, { "epoch": 1.0, "learning_rate": 0.0004753613269455292, "loss": 1.4178, "step": 24980 }, { "epoch": 1.01, "learning_rate": 0.000475351262128105, "loss": 1.4288, "step": 24990 }, { "epoch": 1.01, "learning_rate": 0.0004753411973106808, "loss": 1.6343, "step": 25000 }, { "epoch": 1.01, "learning_rate": 0.0004753311324932566, "loss": 1.3633, "step": 25010 }, { "epoch": 1.01, "learning_rate": 0.0004753210676758324, "loss": 1.4687, "step": 25020 }, { "epoch": 1.01, "learning_rate": 0.0004753110028584082, "loss": 1.587, "step": 25030 }, { "epoch": 1.01, "learning_rate": 0.000475300938040984, "loss": 1.5639, "step": 25040 }, { "epoch": 1.01, "learning_rate": 0.0004752908732235597, "loss": 1.6916, "step": 25050 }, { "epoch": 1.01, "learning_rate": 0.0004752808084061355, "loss": 1.5941, "step": 25060 }, { "epoch": 1.01, "learning_rate": 0.0004752707435887113, "loss": 1.2074, "step": 25070 }, { "epoch": 1.01, "learning_rate": 0.0004752606787712871, "loss": 1.6055, "step": 25080 }, { "epoch": 1.01, "learning_rate": 0.0004752506139538629, "loss": 1.2088, "step": 25090 }, { "epoch": 1.01, "learning_rate": 0.00047524054913643867, "loss": 1.4725, "step": 25100 }, { "epoch": 1.01, "learning_rate": 0.00047523048431901446, "loss": 1.2743, "step": 25110 }, { "epoch": 1.01, "learning_rate": 0.00047522041950159025, "loss": 1.504, "step": 25120 }, { "epoch": 1.01, "learning_rate": 0.000475210354684166, "loss": 1.3928, "step": 25130 }, { "epoch": 1.01, "learning_rate": 0.00047520028986674183, "loss": 1.3143, "step": 25140 }, { "epoch": 1.01, "learning_rate": 0.0004751902250493176, "loss": 1.3822, "step": 25150 }, { "epoch": 1.01, "learning_rate": 0.0004751801602318934, "loss": 1.2492, "step": 25160 }, { "epoch": 1.01, "learning_rate": 0.0004751700954144692, "loss": 1.537, "step": 25170 }, { "epoch": 1.01, "learning_rate": 0.000475160030597045, "loss": 1.6099, "step": 25180 }, { "epoch": 1.01, "learning_rate": 0.0004751499657796208, "loss": 1.4253, "step": 25190 }, { "epoch": 1.01, "learning_rate": 0.0004751399009621966, "loss": 1.4419, "step": 25200 }, { "epoch": 1.01, "learning_rate": 0.00047512983614477237, "loss": 1.3805, "step": 25210 }, { "epoch": 1.01, "learning_rate": 0.0004751197713273481, "loss": 1.4732, "step": 25220 }, { "epoch": 1.01, "learning_rate": 0.0004751097065099239, "loss": 1.3147, "step": 25230 }, { "epoch": 1.02, "learning_rate": 0.0004750996416924997, "loss": 1.3007, "step": 25240 }, { "epoch": 1.02, "learning_rate": 0.0004750895768750755, "loss": 1.6616, "step": 25250 }, { "epoch": 1.02, "learning_rate": 0.00047507951205765127, "loss": 1.3181, "step": 25260 }, { "epoch": 1.02, "learning_rate": 0.00047506944724022706, "loss": 1.2949, "step": 25270 }, { "epoch": 1.02, "learning_rate": 0.00047505938242280285, "loss": 1.4854, "step": 25280 }, { "epoch": 1.02, "learning_rate": 0.0004750493176053787, "loss": 1.4572, "step": 25290 }, { "epoch": 1.02, "learning_rate": 0.00047503925278795443, "loss": 1.2928, "step": 25300 }, { "epoch": 1.02, "learning_rate": 0.0004750291879705302, "loss": 1.6013, "step": 25310 }, { "epoch": 1.02, "learning_rate": 0.000475019123153106, "loss": 1.3621, "step": 25320 }, { "epoch": 1.02, "learning_rate": 0.0004750090583356818, "loss": 1.3978, "step": 25330 }, { "epoch": 1.02, "learning_rate": 0.0004749989935182576, "loss": 1.53, "step": 25340 }, { "epoch": 1.02, "learning_rate": 0.0004749889287008334, "loss": 1.2601, "step": 25350 }, { "epoch": 1.02, "learning_rate": 0.0004749788638834092, "loss": 1.5585, "step": 25360 }, { "epoch": 1.02, "learning_rate": 0.00047496879906598497, "loss": 1.6874, "step": 25370 }, { "epoch": 1.02, "learning_rate": 0.00047495873424856076, "loss": 1.4817, "step": 25380 }, { "epoch": 1.02, "learning_rate": 0.0004749486694311365, "loss": 1.2825, "step": 25390 }, { "epoch": 1.02, "learning_rate": 0.0004749386046137123, "loss": 1.3287, "step": 25400 }, { "epoch": 1.02, "learning_rate": 0.0004749285397962881, "loss": 1.4568, "step": 25410 }, { "epoch": 1.02, "learning_rate": 0.00047491847497886387, "loss": 1.5011, "step": 25420 }, { "epoch": 1.02, "learning_rate": 0.0004749084101614397, "loss": 1.669, "step": 25430 }, { "epoch": 1.02, "learning_rate": 0.0004748983453440155, "loss": 1.4099, "step": 25440 }, { "epoch": 1.02, "learning_rate": 0.0004748882805265913, "loss": 1.0933, "step": 25450 }, { "epoch": 1.02, "learning_rate": 0.0004748782157091671, "loss": 1.4318, "step": 25460 }, { "epoch": 1.02, "learning_rate": 0.0004748681508917428, "loss": 1.823, "step": 25470 }, { "epoch": 1.02, "learning_rate": 0.0004748580860743186, "loss": 1.5735, "step": 25480 }, { "epoch": 1.03, "learning_rate": 0.0004748480212568944, "loss": 1.4384, "step": 25490 }, { "epoch": 1.03, "learning_rate": 0.0004748379564394702, "loss": 1.4862, "step": 25500 }, { "epoch": 1.03, "learning_rate": 0.000474827891622046, "loss": 1.3125, "step": 25510 }, { "epoch": 1.03, "learning_rate": 0.0004748178268046218, "loss": 1.6341, "step": 25520 }, { "epoch": 1.03, "learning_rate": 0.00047480776198719757, "loss": 1.5662, "step": 25530 }, { "epoch": 1.03, "learning_rate": 0.00047479769716977336, "loss": 1.3421, "step": 25540 }, { "epoch": 1.03, "learning_rate": 0.0004747876323523491, "loss": 1.3611, "step": 25550 }, { "epoch": 1.03, "learning_rate": 0.0004747775675349249, "loss": 1.4828, "step": 25560 }, { "epoch": 1.03, "learning_rate": 0.0004747675027175007, "loss": 1.4826, "step": 25570 }, { "epoch": 1.03, "learning_rate": 0.0004747574379000765, "loss": 1.6922, "step": 25580 }, { "epoch": 1.03, "learning_rate": 0.0004747473730826523, "loss": 1.7231, "step": 25590 }, { "epoch": 1.03, "learning_rate": 0.0004747373082652281, "loss": 1.6205, "step": 25600 }, { "epoch": 1.03, "learning_rate": 0.0004747272434478039, "loss": 1.2912, "step": 25610 }, { "epoch": 1.03, "learning_rate": 0.0004747171786303797, "loss": 1.2118, "step": 25620 }, { "epoch": 1.03, "learning_rate": 0.0004747071138129555, "loss": 1.6091, "step": 25630 }, { "epoch": 1.03, "learning_rate": 0.0004746970489955312, "loss": 1.4228, "step": 25640 }, { "epoch": 1.03, "learning_rate": 0.000474686984178107, "loss": 1.5534, "step": 25650 }, { "epoch": 1.03, "learning_rate": 0.0004746769193606828, "loss": 1.3556, "step": 25660 }, { "epoch": 1.03, "learning_rate": 0.0004746668545432586, "loss": 1.7386, "step": 25670 }, { "epoch": 1.03, "learning_rate": 0.00047465678972583437, "loss": 1.303, "step": 25680 }, { "epoch": 1.03, "learning_rate": 0.00047464672490841016, "loss": 1.5665, "step": 25690 }, { "epoch": 1.03, "learning_rate": 0.00047463666009098595, "loss": 1.3185, "step": 25700 }, { "epoch": 1.03, "learning_rate": 0.00047462659527356174, "loss": 1.3916, "step": 25710 }, { "epoch": 1.03, "learning_rate": 0.0004746165304561376, "loss": 1.4055, "step": 25720 }, { "epoch": 1.03, "learning_rate": 0.0004746064656387133, "loss": 1.3743, "step": 25730 }, { "epoch": 1.04, "learning_rate": 0.0004745964008212891, "loss": 1.4578, "step": 25740 }, { "epoch": 1.04, "learning_rate": 0.0004745863360038649, "loss": 1.6626, "step": 25750 }, { "epoch": 1.04, "learning_rate": 0.0004745762711864407, "loss": 1.6407, "step": 25760 }, { "epoch": 1.04, "learning_rate": 0.0004745662063690165, "loss": 1.5687, "step": 25770 }, { "epoch": 1.04, "learning_rate": 0.0004745561415515923, "loss": 1.3836, "step": 25780 }, { "epoch": 1.04, "learning_rate": 0.00047454607673416807, "loss": 1.4669, "step": 25790 }, { "epoch": 1.04, "learning_rate": 0.00047453601191674386, "loss": 1.3104, "step": 25800 }, { "epoch": 1.04, "learning_rate": 0.0004745259470993196, "loss": 1.4286, "step": 25810 }, { "epoch": 1.04, "learning_rate": 0.0004745158822818954, "loss": 1.5331, "step": 25820 }, { "epoch": 1.04, "learning_rate": 0.0004745058174644712, "loss": 1.6243, "step": 25830 }, { "epoch": 1.04, "learning_rate": 0.00047449575264704697, "loss": 1.5257, "step": 25840 }, { "epoch": 1.04, "learning_rate": 0.00047448568782962276, "loss": 1.3282, "step": 25850 }, { "epoch": 1.04, "learning_rate": 0.00047447562301219855, "loss": 1.3839, "step": 25860 }, { "epoch": 1.04, "learning_rate": 0.0004744655581947744, "loss": 1.5093, "step": 25870 }, { "epoch": 1.04, "learning_rate": 0.0004744554933773502, "loss": 1.5517, "step": 25880 }, { "epoch": 1.04, "learning_rate": 0.0004744454285599259, "loss": 1.4118, "step": 25890 }, { "epoch": 1.04, "learning_rate": 0.0004744353637425017, "loss": 1.4488, "step": 25900 }, { "epoch": 1.04, "learning_rate": 0.0004744252989250775, "loss": 1.5484, "step": 25910 }, { "epoch": 1.04, "learning_rate": 0.0004744152341076533, "loss": 1.8612, "step": 25920 }, { "epoch": 1.04, "learning_rate": 0.0004744051692902291, "loss": 1.3881, "step": 25930 }, { "epoch": 1.04, "learning_rate": 0.0004743951044728049, "loss": 1.3759, "step": 25940 }, { "epoch": 1.04, "learning_rate": 0.00047438503965538067, "loss": 1.5876, "step": 25950 }, { "epoch": 1.04, "learning_rate": 0.00047437497483795646, "loss": 1.6126, "step": 25960 }, { "epoch": 1.04, "learning_rate": 0.00047436491002053225, "loss": 1.1723, "step": 25970 }, { "epoch": 1.04, "learning_rate": 0.000474354845203108, "loss": 1.27, "step": 25980 }, { "epoch": 1.05, "learning_rate": 0.0004743447803856838, "loss": 1.6077, "step": 25990 }, { "epoch": 1.05, "learning_rate": 0.00047433471556825957, "loss": 1.5963, "step": 26000 }, { "epoch": 1.05, "learning_rate": 0.00047432465075083536, "loss": 1.4107, "step": 26010 }, { "epoch": 1.05, "learning_rate": 0.0004743145859334112, "loss": 1.5967, "step": 26020 }, { "epoch": 1.05, "learning_rate": 0.000474304521115987, "loss": 1.3958, "step": 26030 }, { "epoch": 1.05, "learning_rate": 0.0004742944562985628, "loss": 1.2988, "step": 26040 }, { "epoch": 1.05, "learning_rate": 0.0004742843914811386, "loss": 1.3662, "step": 26050 }, { "epoch": 1.05, "learning_rate": 0.0004742743266637143, "loss": 1.4304, "step": 26060 }, { "epoch": 1.05, "learning_rate": 0.0004742642618462901, "loss": 1.7465, "step": 26070 }, { "epoch": 1.05, "learning_rate": 0.0004742541970288659, "loss": 1.3954, "step": 26080 }, { "epoch": 1.05, "learning_rate": 0.0004742441322114417, "loss": 1.4877, "step": 26090 }, { "epoch": 1.05, "learning_rate": 0.0004742340673940175, "loss": 1.5327, "step": 26100 }, { "epoch": 1.05, "learning_rate": 0.00047422400257659327, "loss": 1.5709, "step": 26110 }, { "epoch": 1.05, "learning_rate": 0.00047421393775916906, "loss": 1.3153, "step": 26120 }, { "epoch": 1.05, "learning_rate": 0.00047420387294174485, "loss": 1.7447, "step": 26130 }, { "epoch": 1.05, "learning_rate": 0.00047419380812432064, "loss": 1.3909, "step": 26140 }, { "epoch": 1.05, "learning_rate": 0.0004741837433068964, "loss": 1.5552, "step": 26150 }, { "epoch": 1.05, "learning_rate": 0.0004741736784894722, "loss": 1.2446, "step": 26160 }, { "epoch": 1.05, "learning_rate": 0.000474163613672048, "loss": 1.354, "step": 26170 }, { "epoch": 1.05, "learning_rate": 0.0004741535488546238, "loss": 1.537, "step": 26180 }, { "epoch": 1.05, "learning_rate": 0.0004741434840371996, "loss": 1.2775, "step": 26190 }, { "epoch": 1.05, "learning_rate": 0.0004741334192197754, "loss": 1.431, "step": 26200 }, { "epoch": 1.05, "learning_rate": 0.0004741233544023512, "loss": 1.5627, "step": 26210 }, { "epoch": 1.05, "learning_rate": 0.00047411328958492697, "loss": 1.3363, "step": 26220 }, { "epoch": 1.05, "learning_rate": 0.0004741032247675027, "loss": 1.3288, "step": 26230 }, { "epoch": 1.06, "learning_rate": 0.0004740931599500785, "loss": 1.4369, "step": 26240 }, { "epoch": 1.06, "learning_rate": 0.0004740830951326543, "loss": 1.604, "step": 26250 }, { "epoch": 1.06, "learning_rate": 0.0004740730303152301, "loss": 1.5884, "step": 26260 }, { "epoch": 1.06, "learning_rate": 0.00047406296549780587, "loss": 1.8012, "step": 26270 }, { "epoch": 1.06, "learning_rate": 0.00047405290068038166, "loss": 1.4752, "step": 26280 }, { "epoch": 1.06, "learning_rate": 0.00047404283586295745, "loss": 1.3582, "step": 26290 }, { "epoch": 1.06, "learning_rate": 0.00047403277104553324, "loss": 1.605, "step": 26300 }, { "epoch": 1.06, "learning_rate": 0.0004740227062281091, "loss": 1.5177, "step": 26310 }, { "epoch": 1.06, "learning_rate": 0.0004740126414106848, "loss": 1.3385, "step": 26320 }, { "epoch": 1.06, "learning_rate": 0.0004740025765932606, "loss": 1.5794, "step": 26330 }, { "epoch": 1.06, "learning_rate": 0.0004739925117758364, "loss": 1.5487, "step": 26340 }, { "epoch": 1.06, "learning_rate": 0.0004739824469584122, "loss": 1.3364, "step": 26350 }, { "epoch": 1.06, "learning_rate": 0.000473972382140988, "loss": 1.5962, "step": 26360 }, { "epoch": 1.06, "learning_rate": 0.0004739623173235638, "loss": 1.5542, "step": 26370 }, { "epoch": 1.06, "learning_rate": 0.00047395225250613956, "loss": 1.539, "step": 26380 }, { "epoch": 1.06, "learning_rate": 0.00047394218768871536, "loss": 1.2921, "step": 26390 }, { "epoch": 1.06, "learning_rate": 0.0004739321228712911, "loss": 1.4366, "step": 26400 }, { "epoch": 1.06, "learning_rate": 0.0004739220580538669, "loss": 1.4299, "step": 26410 }, { "epoch": 1.06, "learning_rate": 0.0004739119932364427, "loss": 1.6881, "step": 26420 }, { "epoch": 1.06, "learning_rate": 0.00047390192841901846, "loss": 1.5784, "step": 26430 }, { "epoch": 1.06, "learning_rate": 0.00047389186360159426, "loss": 1.6895, "step": 26440 }, { "epoch": 1.06, "learning_rate": 0.00047388179878417005, "loss": 1.0998, "step": 26450 }, { "epoch": 1.06, "learning_rate": 0.0004738717339667459, "loss": 1.7443, "step": 26460 }, { "epoch": 1.06, "learning_rate": 0.0004738616691493217, "loss": 1.363, "step": 26470 }, { "epoch": 1.06, "learning_rate": 0.00047385160433189747, "loss": 1.8515, "step": 26480 }, { "epoch": 1.07, "learning_rate": 0.0004738415395144732, "loss": 1.4368, "step": 26490 }, { "epoch": 1.07, "learning_rate": 0.000473831474697049, "loss": 1.5629, "step": 26500 }, { "epoch": 1.07, "learning_rate": 0.0004738214098796248, "loss": 1.4903, "step": 26510 }, { "epoch": 1.07, "learning_rate": 0.0004738113450622006, "loss": 1.5778, "step": 26520 }, { "epoch": 1.07, "learning_rate": 0.00047380128024477637, "loss": 1.4567, "step": 26530 }, { "epoch": 1.07, "learning_rate": 0.00047379121542735216, "loss": 1.5258, "step": 26540 }, { "epoch": 1.07, "learning_rate": 0.00047378115060992795, "loss": 1.3212, "step": 26550 }, { "epoch": 1.07, "learning_rate": 0.00047377108579250374, "loss": 1.5626, "step": 26560 }, { "epoch": 1.07, "learning_rate": 0.0004737610209750795, "loss": 1.399, "step": 26570 }, { "epoch": 1.07, "learning_rate": 0.00047375095615765527, "loss": 1.427, "step": 26580 }, { "epoch": 1.07, "learning_rate": 0.00047374089134023106, "loss": 1.686, "step": 26590 }, { "epoch": 1.07, "learning_rate": 0.0004737308265228069, "loss": 1.1385, "step": 26600 }, { "epoch": 1.07, "learning_rate": 0.0004737207617053827, "loss": 1.414, "step": 26610 }, { "epoch": 1.07, "learning_rate": 0.0004737106968879585, "loss": 1.6941, "step": 26620 }, { "epoch": 1.07, "learning_rate": 0.0004737006320705343, "loss": 1.7361, "step": 26630 }, { "epoch": 1.07, "learning_rate": 0.00047369056725311007, "loss": 1.546, "step": 26640 }, { "epoch": 1.07, "learning_rate": 0.0004736805024356858, "loss": 1.4707, "step": 26650 }, { "epoch": 1.07, "learning_rate": 0.0004736704376182616, "loss": 1.4678, "step": 26660 }, { "epoch": 1.07, "learning_rate": 0.0004736603728008374, "loss": 1.546, "step": 26670 }, { "epoch": 1.07, "learning_rate": 0.0004736503079834132, "loss": 1.7655, "step": 26680 }, { "epoch": 1.07, "learning_rate": 0.00047364024316598897, "loss": 1.316, "step": 26690 }, { "epoch": 1.07, "learning_rate": 0.00047363017834856476, "loss": 1.4104, "step": 26700 }, { "epoch": 1.07, "learning_rate": 0.00047362011353114055, "loss": 1.4336, "step": 26710 }, { "epoch": 1.07, "learning_rate": 0.00047361004871371634, "loss": 1.5164, "step": 26720 }, { "epoch": 1.08, "learning_rate": 0.00047359998389629213, "loss": 1.5085, "step": 26730 }, { "epoch": 1.08, "learning_rate": 0.00047358991907886787, "loss": 1.4908, "step": 26740 }, { "epoch": 1.08, "learning_rate": 0.0004735798542614437, "loss": 1.4711, "step": 26750 }, { "epoch": 1.08, "learning_rate": 0.0004735697894440195, "loss": 1.4664, "step": 26760 }, { "epoch": 1.08, "learning_rate": 0.0004735597246265953, "loss": 1.5793, "step": 26770 }, { "epoch": 1.08, "learning_rate": 0.0004735496598091711, "loss": 1.514, "step": 26780 }, { "epoch": 1.08, "learning_rate": 0.0004735395949917469, "loss": 1.4015, "step": 26790 }, { "epoch": 1.08, "learning_rate": 0.00047352953017432267, "loss": 1.6744, "step": 26800 }, { "epoch": 1.08, "learning_rate": 0.00047351946535689846, "loss": 1.4214, "step": 26810 }, { "epoch": 1.08, "learning_rate": 0.0004735094005394742, "loss": 1.5143, "step": 26820 }, { "epoch": 1.08, "learning_rate": 0.00047349933572205, "loss": 1.3093, "step": 26830 }, { "epoch": 1.08, "learning_rate": 0.0004734892709046258, "loss": 1.535, "step": 26840 }, { "epoch": 1.08, "learning_rate": 0.00047347920608720157, "loss": 1.5338, "step": 26850 }, { "epoch": 1.08, "learning_rate": 0.00047346914126977736, "loss": 1.6742, "step": 26860 }, { "epoch": 1.08, "learning_rate": 0.00047345907645235315, "loss": 1.4246, "step": 26870 }, { "epoch": 1.08, "learning_rate": 0.00047344901163492894, "loss": 1.5743, "step": 26880 }, { "epoch": 1.08, "learning_rate": 0.0004734389468175048, "loss": 1.4983, "step": 26890 }, { "epoch": 1.08, "learning_rate": 0.0004734288820000806, "loss": 1.4015, "step": 26900 }, { "epoch": 1.08, "learning_rate": 0.0004734188171826563, "loss": 1.4393, "step": 26910 }, { "epoch": 1.08, "learning_rate": 0.0004734087523652321, "loss": 1.5586, "step": 26920 }, { "epoch": 1.08, "learning_rate": 0.0004733986875478079, "loss": 1.4197, "step": 26930 }, { "epoch": 1.08, "learning_rate": 0.0004733886227303837, "loss": 1.5601, "step": 26940 }, { "epoch": 1.08, "learning_rate": 0.0004733785579129595, "loss": 1.3993, "step": 26950 }, { "epoch": 1.08, "learning_rate": 0.00047336849309553527, "loss": 1.3364, "step": 26960 }, { "epoch": 1.08, "learning_rate": 0.00047335842827811106, "loss": 1.4233, "step": 26970 }, { "epoch": 1.09, "learning_rate": 0.00047334836346068685, "loss": 1.7687, "step": 26980 }, { "epoch": 1.09, "learning_rate": 0.0004733382986432626, "loss": 1.4851, "step": 26990 }, { "epoch": 1.09, "learning_rate": 0.0004733282338258384, "loss": 1.5848, "step": 27000 }, { "epoch": 1.09, "learning_rate": 0.00047331816900841417, "loss": 1.4114, "step": 27010 }, { "epoch": 1.09, "learning_rate": 0.00047330810419098996, "loss": 1.5024, "step": 27020 }, { "epoch": 1.09, "learning_rate": 0.00047329803937356575, "loss": 1.5207, "step": 27030 }, { "epoch": 1.09, "learning_rate": 0.0004732879745561416, "loss": 1.4234, "step": 27040 }, { "epoch": 1.09, "learning_rate": 0.0004732779097387174, "loss": 1.5018, "step": 27050 }, { "epoch": 1.09, "learning_rate": 0.0004732678449212932, "loss": 1.2615, "step": 27060 }, { "epoch": 1.09, "learning_rate": 0.00047325778010386897, "loss": 1.3951, "step": 27070 }, { "epoch": 1.09, "learning_rate": 0.0004732477152864447, "loss": 1.4408, "step": 27080 }, { "epoch": 1.09, "learning_rate": 0.0004732376504690205, "loss": 1.3961, "step": 27090 }, { "epoch": 1.09, "learning_rate": 0.0004732275856515963, "loss": 1.4578, "step": 27100 }, { "epoch": 1.09, "learning_rate": 0.0004732175208341721, "loss": 1.2485, "step": 27110 }, { "epoch": 1.09, "learning_rate": 0.00047320745601674787, "loss": 1.6309, "step": 27120 }, { "epoch": 1.09, "learning_rate": 0.00047319739119932366, "loss": 1.4588, "step": 27130 }, { "epoch": 1.09, "learning_rate": 0.00047318732638189945, "loss": 1.2251, "step": 27140 }, { "epoch": 1.09, "learning_rate": 0.00047317726156447524, "loss": 1.5545, "step": 27150 }, { "epoch": 1.09, "learning_rate": 0.000473167196747051, "loss": 1.4056, "step": 27160 }, { "epoch": 1.09, "learning_rate": 0.00047315713192962677, "loss": 1.6127, "step": 27170 }, { "epoch": 1.09, "learning_rate": 0.00047314706711220256, "loss": 1.4823, "step": 27180 }, { "epoch": 1.09, "learning_rate": 0.0004731370022947784, "loss": 1.4123, "step": 27190 }, { "epoch": 1.09, "learning_rate": 0.0004731269374773542, "loss": 1.657, "step": 27200 }, { "epoch": 1.09, "learning_rate": 0.00047311687265993, "loss": 1.3849, "step": 27210 }, { "epoch": 1.09, "learning_rate": 0.0004731068078425058, "loss": 1.2584, "step": 27220 }, { "epoch": 1.1, "learning_rate": 0.00047309674302508156, "loss": 1.5592, "step": 27230 }, { "epoch": 1.1, "learning_rate": 0.00047308667820765736, "loss": 1.6348, "step": 27240 }, { "epoch": 1.1, "learning_rate": 0.0004730766133902331, "loss": 1.3654, "step": 27250 }, { "epoch": 1.1, "learning_rate": 0.0004730665485728089, "loss": 1.6996, "step": 27260 }, { "epoch": 1.1, "learning_rate": 0.0004730564837553847, "loss": 1.3472, "step": 27270 }, { "epoch": 1.1, "learning_rate": 0.00047304641893796046, "loss": 1.6805, "step": 27280 }, { "epoch": 1.1, "learning_rate": 0.00047303635412053626, "loss": 1.7085, "step": 27290 }, { "epoch": 1.1, "learning_rate": 0.00047302628930311205, "loss": 1.4632, "step": 27300 }, { "epoch": 1.1, "learning_rate": 0.00047301622448568784, "loss": 1.6248, "step": 27310 }, { "epoch": 1.1, "learning_rate": 0.00047300615966826363, "loss": 1.3791, "step": 27320 }, { "epoch": 1.1, "learning_rate": 0.0004729960948508394, "loss": 1.7396, "step": 27330 }, { "epoch": 1.1, "learning_rate": 0.0004729860300334152, "loss": 1.5432, "step": 27340 }, { "epoch": 1.1, "learning_rate": 0.000472975965215991, "loss": 1.5339, "step": 27350 }, { "epoch": 1.1, "learning_rate": 0.0004729659003985668, "loss": 1.6097, "step": 27360 }, { "epoch": 1.1, "learning_rate": 0.0004729558355811426, "loss": 1.7725, "step": 27370 }, { "epoch": 1.1, "learning_rate": 0.00047294577076371837, "loss": 1.6271, "step": 27380 }, { "epoch": 1.1, "learning_rate": 0.00047293570594629416, "loss": 1.5172, "step": 27390 }, { "epoch": 1.1, "learning_rate": 0.00047292564112886995, "loss": 1.4826, "step": 27400 }, { "epoch": 1.1, "learning_rate": 0.0004729155763114457, "loss": 1.5931, "step": 27410 }, { "epoch": 1.1, "learning_rate": 0.0004729055114940215, "loss": 1.5443, "step": 27420 }, { "epoch": 1.1, "learning_rate": 0.00047289544667659727, "loss": 1.4895, "step": 27430 }, { "epoch": 1.1, "learning_rate": 0.00047288538185917306, "loss": 1.4184, "step": 27440 }, { "epoch": 1.1, "learning_rate": 0.00047287531704174885, "loss": 1.3197, "step": 27450 }, { "epoch": 1.1, "learning_rate": 0.00047286525222432464, "loss": 1.3181, "step": 27460 }, { "epoch": 1.1, "learning_rate": 0.00047285518740690043, "loss": 1.4281, "step": 27470 }, { "epoch": 1.11, "learning_rate": 0.0004728451225894763, "loss": 1.3757, "step": 27480 }, { "epoch": 1.11, "learning_rate": 0.00047283505777205207, "loss": 1.4848, "step": 27490 }, { "epoch": 1.11, "learning_rate": 0.0004728249929546278, "loss": 1.566, "step": 27500 }, { "epoch": 1.11, "learning_rate": 0.0004728149281372036, "loss": 1.6154, "step": 27510 }, { "epoch": 1.11, "learning_rate": 0.0004728048633197794, "loss": 1.6386, "step": 27520 }, { "epoch": 1.11, "learning_rate": 0.0004727947985023552, "loss": 1.6289, "step": 27530 }, { "epoch": 1.11, "learning_rate": 0.00047278473368493097, "loss": 1.3257, "step": 27540 }, { "epoch": 1.11, "learning_rate": 0.00047277466886750676, "loss": 1.6459, "step": 27550 }, { "epoch": 1.11, "learning_rate": 0.00047276460405008255, "loss": 1.5711, "step": 27560 }, { "epoch": 1.11, "learning_rate": 0.00047275453923265834, "loss": 1.4661, "step": 27570 }, { "epoch": 1.11, "learning_rate": 0.0004727444744152341, "loss": 1.4065, "step": 27580 }, { "epoch": 1.11, "learning_rate": 0.00047273440959780987, "loss": 1.4408, "step": 27590 }, { "epoch": 1.11, "learning_rate": 0.00047272434478038566, "loss": 1.8493, "step": 27600 }, { "epoch": 1.11, "learning_rate": 0.00047271427996296145, "loss": 1.3239, "step": 27610 }, { "epoch": 1.11, "learning_rate": 0.0004727042151455373, "loss": 1.905, "step": 27620 }, { "epoch": 1.11, "learning_rate": 0.0004726941503281131, "loss": 1.3425, "step": 27630 }, { "epoch": 1.11, "learning_rate": 0.0004726840855106889, "loss": 1.5857, "step": 27640 }, { "epoch": 1.11, "learning_rate": 0.00047267402069326467, "loss": 1.3558, "step": 27650 }, { "epoch": 1.11, "learning_rate": 0.00047266395587584046, "loss": 1.404, "step": 27660 }, { "epoch": 1.11, "learning_rate": 0.0004726538910584162, "loss": 1.3906, "step": 27670 }, { "epoch": 1.11, "learning_rate": 0.000472643826240992, "loss": 1.3443, "step": 27680 }, { "epoch": 1.11, "learning_rate": 0.0004726337614235678, "loss": 1.5504, "step": 27690 }, { "epoch": 1.11, "learning_rate": 0.00047262369660614357, "loss": 1.4417, "step": 27700 }, { "epoch": 1.11, "learning_rate": 0.00047261363178871936, "loss": 1.3425, "step": 27710 }, { "epoch": 1.11, "learning_rate": 0.00047260356697129515, "loss": 1.3782, "step": 27720 }, { "epoch": 1.12, "learning_rate": 0.00047259350215387094, "loss": 1.6421, "step": 27730 }, { "epoch": 1.12, "learning_rate": 0.00047258343733644673, "loss": 1.4135, "step": 27740 }, { "epoch": 1.12, "learning_rate": 0.00047257337251902247, "loss": 1.5289, "step": 27750 }, { "epoch": 1.12, "learning_rate": 0.00047256330770159826, "loss": 1.48, "step": 27760 }, { "epoch": 1.12, "learning_rate": 0.0004725532428841741, "loss": 1.4218, "step": 27770 }, { "epoch": 1.12, "learning_rate": 0.0004725431780667499, "loss": 1.5959, "step": 27780 }, { "epoch": 1.12, "learning_rate": 0.0004725331132493257, "loss": 1.5554, "step": 27790 }, { "epoch": 1.12, "learning_rate": 0.0004725230484319015, "loss": 1.7281, "step": 27800 }, { "epoch": 1.12, "learning_rate": 0.00047251298361447727, "loss": 1.623, "step": 27810 }, { "epoch": 1.12, "learning_rate": 0.00047250291879705306, "loss": 1.4874, "step": 27820 }, { "epoch": 1.12, "learning_rate": 0.00047249285397962885, "loss": 1.4861, "step": 27830 }, { "epoch": 1.12, "learning_rate": 0.0004724827891622046, "loss": 1.6925, "step": 27840 }, { "epoch": 1.12, "learning_rate": 0.0004724727243447804, "loss": 1.4776, "step": 27850 }, { "epoch": 1.12, "learning_rate": 0.00047246265952735617, "loss": 1.2687, "step": 27860 }, { "epoch": 1.12, "learning_rate": 0.00047245259470993196, "loss": 1.7279, "step": 27870 }, { "epoch": 1.12, "learning_rate": 0.00047244252989250775, "loss": 1.2106, "step": 27880 }, { "epoch": 1.12, "learning_rate": 0.00047243246507508354, "loss": 1.4596, "step": 27890 }, { "epoch": 1.12, "learning_rate": 0.00047242240025765933, "loss": 1.4657, "step": 27900 }, { "epoch": 1.12, "learning_rate": 0.0004724123354402351, "loss": 1.3794, "step": 27910 }, { "epoch": 1.12, "learning_rate": 0.0004724022706228109, "loss": 1.3493, "step": 27920 }, { "epoch": 1.12, "learning_rate": 0.0004723922058053867, "loss": 1.4827, "step": 27930 }, { "epoch": 1.12, "learning_rate": 0.0004723821409879625, "loss": 1.4581, "step": 27940 }, { "epoch": 1.12, "learning_rate": 0.0004723720761705383, "loss": 1.4436, "step": 27950 }, { "epoch": 1.12, "learning_rate": 0.0004723620113531141, "loss": 1.4575, "step": 27960 }, { "epoch": 1.12, "learning_rate": 0.00047235194653568987, "loss": 1.4509, "step": 27970 }, { "epoch": 1.13, "learning_rate": 0.00047234188171826566, "loss": 1.6756, "step": 27980 }, { "epoch": 1.13, "learning_rate": 0.00047233181690084145, "loss": 1.5472, "step": 27990 }, { "epoch": 1.13, "learning_rate": 0.00047232175208341724, "loss": 1.6094, "step": 28000 }, { "epoch": 1.13, "learning_rate": 0.000472311687265993, "loss": 1.4588, "step": 28010 }, { "epoch": 1.13, "learning_rate": 0.00047230162244856877, "loss": 1.5717, "step": 28020 }, { "epoch": 1.13, "learning_rate": 0.00047229155763114456, "loss": 1.61, "step": 28030 }, { "epoch": 1.13, "learning_rate": 0.00047228149281372035, "loss": 1.9231, "step": 28040 }, { "epoch": 1.13, "learning_rate": 0.00047227142799629614, "loss": 1.2877, "step": 28050 }, { "epoch": 1.13, "learning_rate": 0.000472261363178872, "loss": 1.6165, "step": 28060 }, { "epoch": 1.13, "learning_rate": 0.0004722512983614478, "loss": 1.2683, "step": 28070 }, { "epoch": 1.13, "learning_rate": 0.00047224123354402356, "loss": 1.4839, "step": 28080 }, { "epoch": 1.13, "learning_rate": 0.0004722311687265993, "loss": 1.5313, "step": 28090 }, { "epoch": 1.13, "learning_rate": 0.0004722211039091751, "loss": 1.4766, "step": 28100 }, { "epoch": 1.13, "learning_rate": 0.0004722110390917509, "loss": 1.2508, "step": 28110 }, { "epoch": 1.13, "learning_rate": 0.0004722009742743267, "loss": 1.3317, "step": 28120 }, { "epoch": 1.13, "learning_rate": 0.00047219090945690246, "loss": 1.5589, "step": 28130 }, { "epoch": 1.13, "learning_rate": 0.00047218084463947825, "loss": 1.4653, "step": 28140 }, { "epoch": 1.13, "learning_rate": 0.00047217077982205405, "loss": 1.5783, "step": 28150 }, { "epoch": 1.13, "learning_rate": 0.00047216071500462984, "loss": 1.4684, "step": 28160 }, { "epoch": 1.13, "learning_rate": 0.0004721506501872056, "loss": 1.724, "step": 28170 }, { "epoch": 1.13, "learning_rate": 0.00047214058536978136, "loss": 1.3797, "step": 28180 }, { "epoch": 1.13, "learning_rate": 0.00047213052055235715, "loss": 1.3664, "step": 28190 }, { "epoch": 1.13, "learning_rate": 0.00047212045573493295, "loss": 1.4458, "step": 28200 }, { "epoch": 1.13, "learning_rate": 0.0004721103909175088, "loss": 1.4674, "step": 28210 }, { "epoch": 1.13, "learning_rate": 0.0004721003261000846, "loss": 1.1655, "step": 28220 }, { "epoch": 1.14, "learning_rate": 0.00047209026128266037, "loss": 1.2912, "step": 28230 }, { "epoch": 1.14, "learning_rate": 0.00047208019646523616, "loss": 1.6434, "step": 28240 }, { "epoch": 1.14, "learning_rate": 0.00047207013164781195, "loss": 1.2809, "step": 28250 }, { "epoch": 1.14, "learning_rate": 0.0004720600668303877, "loss": 1.2924, "step": 28260 }, { "epoch": 1.14, "learning_rate": 0.0004720500020129635, "loss": 1.6326, "step": 28270 }, { "epoch": 1.14, "learning_rate": 0.00047203993719553927, "loss": 1.2144, "step": 28280 }, { "epoch": 1.14, "learning_rate": 0.00047202987237811506, "loss": 1.587, "step": 28290 }, { "epoch": 1.14, "learning_rate": 0.00047201980756069085, "loss": 1.6197, "step": 28300 }, { "epoch": 1.14, "learning_rate": 0.00047200974274326664, "loss": 1.5011, "step": 28310 }, { "epoch": 1.14, "learning_rate": 0.00047199967792584243, "loss": 1.3576, "step": 28320 }, { "epoch": 1.14, "learning_rate": 0.0004719896131084182, "loss": 1.4118, "step": 28330 }, { "epoch": 1.14, "learning_rate": 0.00047197954829099396, "loss": 1.2279, "step": 28340 }, { "epoch": 1.14, "learning_rate": 0.00047196948347356975, "loss": 1.3879, "step": 28350 }, { "epoch": 1.14, "learning_rate": 0.0004719594186561456, "loss": 1.5567, "step": 28360 }, { "epoch": 1.14, "learning_rate": 0.0004719493538387214, "loss": 1.5062, "step": 28370 }, { "epoch": 1.14, "learning_rate": 0.0004719392890212972, "loss": 1.7716, "step": 28380 }, { "epoch": 1.14, "learning_rate": 0.00047192922420387297, "loss": 1.5156, "step": 28390 }, { "epoch": 1.14, "learning_rate": 0.00047191915938644876, "loss": 1.4215, "step": 28400 }, { "epoch": 1.14, "learning_rate": 0.00047190909456902455, "loss": 1.5533, "step": 28410 }, { "epoch": 1.14, "learning_rate": 0.00047189902975160034, "loss": 1.4573, "step": 28420 }, { "epoch": 1.14, "learning_rate": 0.0004718889649341761, "loss": 1.2336, "step": 28430 }, { "epoch": 1.14, "learning_rate": 0.00047187890011675187, "loss": 1.4063, "step": 28440 }, { "epoch": 1.14, "learning_rate": 0.00047186883529932766, "loss": 1.6445, "step": 28450 }, { "epoch": 1.14, "learning_rate": 0.00047185877048190345, "loss": 1.6403, "step": 28460 }, { "epoch": 1.15, "learning_rate": 0.00047184870566447924, "loss": 1.6439, "step": 28470 }, { "epoch": 1.15, "learning_rate": 0.00047183864084705503, "loss": 1.5115, "step": 28480 }, { "epoch": 1.15, "learning_rate": 0.0004718285760296308, "loss": 1.5325, "step": 28490 }, { "epoch": 1.15, "learning_rate": 0.00047181851121220667, "loss": 1.5516, "step": 28500 }, { "epoch": 1.15, "learning_rate": 0.0004718084463947824, "loss": 1.5568, "step": 28510 }, { "epoch": 1.15, "learning_rate": 0.0004717983815773582, "loss": 1.4976, "step": 28520 }, { "epoch": 1.15, "learning_rate": 0.000471788316759934, "loss": 1.2376, "step": 28530 }, { "epoch": 1.15, "learning_rate": 0.0004717782519425098, "loss": 1.3787, "step": 28540 }, { "epoch": 1.15, "learning_rate": 0.00047176818712508557, "loss": 1.3884, "step": 28550 }, { "epoch": 1.15, "learning_rate": 0.00047175812230766136, "loss": 1.2688, "step": 28560 }, { "epoch": 1.15, "learning_rate": 0.00047174805749023715, "loss": 1.2779, "step": 28570 }, { "epoch": 1.15, "learning_rate": 0.00047173799267281294, "loss": 1.431, "step": 28580 }, { "epoch": 1.15, "learning_rate": 0.00047172792785538873, "loss": 1.4224, "step": 28590 }, { "epoch": 1.15, "learning_rate": 0.00047171786303796447, "loss": 1.3907, "step": 28600 }, { "epoch": 1.15, "learning_rate": 0.00047170779822054026, "loss": 1.3553, "step": 28610 }, { "epoch": 1.15, "learning_rate": 0.00047169773340311605, "loss": 1.4272, "step": 28620 }, { "epoch": 1.15, "learning_rate": 0.00047168766858569184, "loss": 1.3113, "step": 28630 }, { "epoch": 1.15, "learning_rate": 0.00047167760376826763, "loss": 1.4849, "step": 28640 }, { "epoch": 1.15, "learning_rate": 0.0004716675389508435, "loss": 1.2562, "step": 28650 }, { "epoch": 1.15, "learning_rate": 0.00047165747413341927, "loss": 1.19, "step": 28660 }, { "epoch": 1.15, "learning_rate": 0.00047164740931599506, "loss": 1.473, "step": 28670 }, { "epoch": 1.15, "learning_rate": 0.0004716373444985708, "loss": 1.8834, "step": 28680 }, { "epoch": 1.15, "learning_rate": 0.0004716272796811466, "loss": 1.4728, "step": 28690 }, { "epoch": 1.15, "learning_rate": 0.0004716172148637224, "loss": 1.4539, "step": 28700 }, { "epoch": 1.15, "learning_rate": 0.00047160715004629817, "loss": 1.4364, "step": 28710 }, { "epoch": 1.16, "learning_rate": 0.00047159708522887396, "loss": 1.4014, "step": 28720 }, { "epoch": 1.16, "learning_rate": 0.00047158702041144975, "loss": 1.3096, "step": 28730 }, { "epoch": 1.16, "learning_rate": 0.00047157695559402554, "loss": 1.2194, "step": 28740 }, { "epoch": 1.16, "learning_rate": 0.00047156689077660133, "loss": 1.2987, "step": 28750 }, { "epoch": 1.16, "learning_rate": 0.0004715568259591771, "loss": 1.6778, "step": 28760 }, { "epoch": 1.16, "learning_rate": 0.00047154676114175286, "loss": 1.4213, "step": 28770 }, { "epoch": 1.16, "learning_rate": 0.00047153669632432865, "loss": 1.642, "step": 28780 }, { "epoch": 1.16, "learning_rate": 0.0004715266315069045, "loss": 1.5881, "step": 28790 }, { "epoch": 1.16, "learning_rate": 0.0004715165666894803, "loss": 1.6068, "step": 28800 }, { "epoch": 1.16, "learning_rate": 0.0004715065018720561, "loss": 1.7535, "step": 28810 }, { "epoch": 1.16, "learning_rate": 0.00047149643705463187, "loss": 1.4442, "step": 28820 }, { "epoch": 1.16, "learning_rate": 0.00047148637223720766, "loss": 1.3696, "step": 28830 }, { "epoch": 1.16, "learning_rate": 0.00047147630741978345, "loss": 1.5332, "step": 28840 }, { "epoch": 1.16, "learning_rate": 0.0004714662426023592, "loss": 1.5843, "step": 28850 }, { "epoch": 1.16, "learning_rate": 0.000471456177784935, "loss": 1.3897, "step": 28860 }, { "epoch": 1.16, "learning_rate": 0.00047144611296751077, "loss": 1.4139, "step": 28870 }, { "epoch": 1.16, "learning_rate": 0.00047143604815008656, "loss": 1.3989, "step": 28880 }, { "epoch": 1.16, "learning_rate": 0.00047142598333266235, "loss": 1.514, "step": 28890 }, { "epoch": 1.16, "learning_rate": 0.00047141591851523814, "loss": 1.5986, "step": 28900 }, { "epoch": 1.16, "learning_rate": 0.00047140585369781393, "loss": 1.6292, "step": 28910 }, { "epoch": 1.16, "learning_rate": 0.0004713957888803897, "loss": 1.5445, "step": 28920 }, { "epoch": 1.16, "learning_rate": 0.00047138572406296546, "loss": 1.3814, "step": 28930 }, { "epoch": 1.16, "learning_rate": 0.0004713756592455413, "loss": 1.498, "step": 28940 }, { "epoch": 1.16, "learning_rate": 0.0004713655944281171, "loss": 1.5817, "step": 28950 }, { "epoch": 1.16, "learning_rate": 0.0004713555296106929, "loss": 1.7437, "step": 28960 }, { "epoch": 1.17, "learning_rate": 0.0004713454647932687, "loss": 1.7958, "step": 28970 }, { "epoch": 1.17, "learning_rate": 0.00047133539997584446, "loss": 1.4468, "step": 28980 }, { "epoch": 1.17, "learning_rate": 0.00047132533515842025, "loss": 1.5619, "step": 28990 }, { "epoch": 1.17, "learning_rate": 0.00047131527034099605, "loss": 1.3904, "step": 29000 }, { "epoch": 1.17, "learning_rate": 0.00047130520552357184, "loss": 1.582, "step": 29010 }, { "epoch": 1.17, "learning_rate": 0.00047129514070614757, "loss": 1.5358, "step": 29020 }, { "epoch": 1.17, "learning_rate": 0.00047128507588872336, "loss": 1.5259, "step": 29030 }, { "epoch": 1.17, "learning_rate": 0.00047127501107129915, "loss": 1.5518, "step": 29040 }, { "epoch": 1.17, "learning_rate": 0.00047126494625387495, "loss": 1.4949, "step": 29050 }, { "epoch": 1.17, "learning_rate": 0.00047125488143645074, "loss": 1.6099, "step": 29060 }, { "epoch": 1.17, "learning_rate": 0.0004712448166190265, "loss": 1.5319, "step": 29070 }, { "epoch": 1.17, "learning_rate": 0.0004712347518016023, "loss": 1.4444, "step": 29080 }, { "epoch": 1.17, "learning_rate": 0.00047122468698417816, "loss": 1.4577, "step": 29090 }, { "epoch": 1.17, "learning_rate": 0.0004712146221667539, "loss": 1.5392, "step": 29100 }, { "epoch": 1.17, "learning_rate": 0.0004712045573493297, "loss": 1.3797, "step": 29110 }, { "epoch": 1.17, "learning_rate": 0.0004711944925319055, "loss": 1.421, "step": 29120 }, { "epoch": 1.17, "learning_rate": 0.00047118442771448127, "loss": 1.9299, "step": 29130 }, { "epoch": 1.17, "learning_rate": 0.00047117436289705706, "loss": 1.4935, "step": 29140 }, { "epoch": 1.17, "learning_rate": 0.00047116429807963285, "loss": 1.7057, "step": 29150 }, { "epoch": 1.17, "learning_rate": 0.00047115423326220864, "loss": 1.6403, "step": 29160 }, { "epoch": 1.17, "learning_rate": 0.00047114416844478443, "loss": 1.5588, "step": 29170 }, { "epoch": 1.17, "learning_rate": 0.0004711341036273602, "loss": 1.4988, "step": 29180 }, { "epoch": 1.17, "learning_rate": 0.00047112403880993596, "loss": 1.3708, "step": 29190 }, { "epoch": 1.17, "learning_rate": 0.00047111397399251175, "loss": 1.3476, "step": 29200 }, { "epoch": 1.17, "learning_rate": 0.00047110390917508754, "loss": 1.3332, "step": 29210 }, { "epoch": 1.18, "learning_rate": 0.00047109384435766333, "loss": 1.5945, "step": 29220 }, { "epoch": 1.18, "learning_rate": 0.0004710837795402392, "loss": 1.6119, "step": 29230 }, { "epoch": 1.18, "learning_rate": 0.00047107371472281497, "loss": 1.473, "step": 29240 }, { "epoch": 1.18, "learning_rate": 0.00047106364990539076, "loss": 1.338, "step": 29250 }, { "epoch": 1.18, "learning_rate": 0.00047105358508796655, "loss": 1.2697, "step": 29260 }, { "epoch": 1.18, "learning_rate": 0.0004710435202705423, "loss": 1.6157, "step": 29270 }, { "epoch": 1.18, "learning_rate": 0.0004710334554531181, "loss": 1.3811, "step": 29280 }, { "epoch": 1.18, "learning_rate": 0.00047102339063569387, "loss": 1.3819, "step": 29290 }, { "epoch": 1.18, "learning_rate": 0.00047101332581826966, "loss": 1.482, "step": 29300 }, { "epoch": 1.18, "learning_rate": 0.00047100326100084545, "loss": 1.5472, "step": 29310 }, { "epoch": 1.18, "learning_rate": 0.00047099319618342124, "loss": 1.5266, "step": 29320 }, { "epoch": 1.18, "learning_rate": 0.00047098313136599703, "loss": 1.5484, "step": 29330 }, { "epoch": 1.18, "learning_rate": 0.0004709730665485728, "loss": 1.6003, "step": 29340 }, { "epoch": 1.18, "learning_rate": 0.0004709630017311486, "loss": 1.5013, "step": 29350 }, { "epoch": 1.18, "learning_rate": 0.00047095293691372435, "loss": 1.6307, "step": 29360 }, { "epoch": 1.18, "learning_rate": 0.00047094287209630014, "loss": 1.3132, "step": 29370 }, { "epoch": 1.18, "learning_rate": 0.000470932807278876, "loss": 1.3115, "step": 29380 }, { "epoch": 1.18, "learning_rate": 0.0004709227424614518, "loss": 1.3951, "step": 29390 }, { "epoch": 1.18, "learning_rate": 0.00047091267764402757, "loss": 1.1739, "step": 29400 }, { "epoch": 1.18, "learning_rate": 0.00047090261282660336, "loss": 1.5394, "step": 29410 }, { "epoch": 1.18, "learning_rate": 0.00047089254800917915, "loss": 1.4232, "step": 29420 }, { "epoch": 1.18, "learning_rate": 0.00047088248319175494, "loss": 1.6469, "step": 29430 }, { "epoch": 1.18, "learning_rate": 0.0004708724183743307, "loss": 1.64, "step": 29440 }, { "epoch": 1.18, "learning_rate": 0.00047086235355690647, "loss": 1.5247, "step": 29450 }, { "epoch": 1.18, "learning_rate": 0.00047085228873948226, "loss": 1.4262, "step": 29460 }, { "epoch": 1.19, "learning_rate": 0.00047084222392205805, "loss": 1.3894, "step": 29470 }, { "epoch": 1.19, "learning_rate": 0.00047083215910463384, "loss": 1.331, "step": 29480 }, { "epoch": 1.19, "learning_rate": 0.00047082209428720963, "loss": 1.5819, "step": 29490 }, { "epoch": 1.19, "learning_rate": 0.0004708120294697854, "loss": 1.2435, "step": 29500 }, { "epoch": 1.19, "learning_rate": 0.0004708019646523612, "loss": 1.4457, "step": 29510 }, { "epoch": 1.19, "learning_rate": 0.000470791899834937, "loss": 1.507, "step": 29520 }, { "epoch": 1.19, "learning_rate": 0.0004707818350175128, "loss": 1.6136, "step": 29530 }, { "epoch": 1.19, "learning_rate": 0.0004707717702000886, "loss": 1.3666, "step": 29540 }, { "epoch": 1.19, "learning_rate": 0.0004707617053826644, "loss": 1.5042, "step": 29550 }, { "epoch": 1.19, "learning_rate": 0.00047075164056524017, "loss": 1.5812, "step": 29560 }, { "epoch": 1.19, "learning_rate": 0.00047074157574781596, "loss": 1.4797, "step": 29570 }, { "epoch": 1.19, "learning_rate": 0.00047073151093039175, "loss": 1.3498, "step": 29580 }, { "epoch": 1.19, "learning_rate": 0.00047072144611296754, "loss": 1.485, "step": 29590 }, { "epoch": 1.19, "learning_rate": 0.00047071138129554333, "loss": 1.6109, "step": 29600 }, { "epoch": 1.19, "learning_rate": 0.00047070131647811907, "loss": 1.3294, "step": 29610 }, { "epoch": 1.19, "learning_rate": 0.00047069125166069486, "loss": 1.4031, "step": 29620 }, { "epoch": 1.19, "learning_rate": 0.00047068118684327065, "loss": 1.5697, "step": 29630 }, { "epoch": 1.19, "learning_rate": 0.00047067112202584644, "loss": 1.4667, "step": 29640 }, { "epoch": 1.19, "learning_rate": 0.00047066105720842223, "loss": 1.4573, "step": 29650 }, { "epoch": 1.19, "learning_rate": 0.000470650992390998, "loss": 1.2555, "step": 29660 }, { "epoch": 1.19, "learning_rate": 0.00047064092757357387, "loss": 1.5052, "step": 29670 }, { "epoch": 1.19, "learning_rate": 0.00047063086275614966, "loss": 1.3375, "step": 29680 }, { "epoch": 1.19, "learning_rate": 0.00047062079793872545, "loss": 1.6971, "step": 29690 }, { "epoch": 1.19, "learning_rate": 0.0004706107331213012, "loss": 1.1402, "step": 29700 }, { "epoch": 1.19, "learning_rate": 0.000470600668303877, "loss": 1.5115, "step": 29710 }, { "epoch": 1.2, "learning_rate": 0.00047059060348645276, "loss": 1.3668, "step": 29720 }, { "epoch": 1.2, "learning_rate": 0.00047058053866902856, "loss": 1.4549, "step": 29730 }, { "epoch": 1.2, "learning_rate": 0.00047057047385160435, "loss": 1.6632, "step": 29740 }, { "epoch": 1.2, "learning_rate": 0.00047056040903418014, "loss": 1.6999, "step": 29750 }, { "epoch": 1.2, "learning_rate": 0.00047055034421675593, "loss": 1.7318, "step": 29760 }, { "epoch": 1.2, "learning_rate": 0.0004705402793993317, "loss": 1.5774, "step": 29770 }, { "epoch": 1.2, "learning_rate": 0.00047053021458190746, "loss": 1.3345, "step": 29780 }, { "epoch": 1.2, "learning_rate": 0.00047052014976448325, "loss": 1.6078, "step": 29790 }, { "epoch": 1.2, "learning_rate": 0.00047051008494705904, "loss": 1.4336, "step": 29800 }, { "epoch": 1.2, "learning_rate": 0.00047050002012963483, "loss": 1.6146, "step": 29810 }, { "epoch": 1.2, "learning_rate": 0.00047048995531221067, "loss": 1.8434, "step": 29820 }, { "epoch": 1.2, "learning_rate": 0.00047047989049478646, "loss": 1.7142, "step": 29830 }, { "epoch": 1.2, "learning_rate": 0.00047046982567736225, "loss": 1.3157, "step": 29840 }, { "epoch": 1.2, "learning_rate": 0.00047045976085993805, "loss": 1.4793, "step": 29850 }, { "epoch": 1.2, "learning_rate": 0.0004704496960425138, "loss": 1.3909, "step": 29860 }, { "epoch": 1.2, "learning_rate": 0.00047043963122508957, "loss": 1.5977, "step": 29870 }, { "epoch": 1.2, "learning_rate": 0.00047042956640766536, "loss": 1.3981, "step": 29880 }, { "epoch": 1.2, "learning_rate": 0.00047041950159024115, "loss": 1.3828, "step": 29890 }, { "epoch": 1.2, "learning_rate": 0.00047040943677281694, "loss": 1.5481, "step": 29900 }, { "epoch": 1.2, "learning_rate": 0.00047039937195539274, "loss": 1.7224, "step": 29910 }, { "epoch": 1.2, "learning_rate": 0.0004703893071379685, "loss": 1.4761, "step": 29920 }, { "epoch": 1.2, "learning_rate": 0.0004703792423205443, "loss": 1.5224, "step": 29930 }, { "epoch": 1.2, "learning_rate": 0.0004703691775031201, "loss": 1.294, "step": 29940 }, { "epoch": 1.2, "learning_rate": 0.00047035911268569584, "loss": 1.4172, "step": 29950 }, { "epoch": 1.2, "learning_rate": 0.0004703490478682717, "loss": 1.5934, "step": 29960 }, { "epoch": 1.21, "learning_rate": 0.0004703389830508475, "loss": 1.5394, "step": 29970 }, { "epoch": 1.21, "learning_rate": 0.00047032891823342327, "loss": 1.5359, "step": 29980 }, { "epoch": 1.21, "learning_rate": 0.00047031885341599906, "loss": 1.57, "step": 29990 }, { "epoch": 1.21, "learning_rate": 0.00047030878859857485, "loss": 1.3643, "step": 30000 }, { "epoch": 1.21, "learning_rate": 0.00047029872378115064, "loss": 1.5797, "step": 30010 }, { "epoch": 1.21, "learning_rate": 0.00047028865896372643, "loss": 1.6861, "step": 30020 }, { "epoch": 1.21, "learning_rate": 0.00047027859414630217, "loss": 1.4, "step": 30030 }, { "epoch": 1.21, "learning_rate": 0.00047026852932887796, "loss": 1.3787, "step": 30040 }, { "epoch": 1.21, "learning_rate": 0.00047025846451145375, "loss": 1.8403, "step": 30050 }, { "epoch": 1.21, "learning_rate": 0.00047024839969402954, "loss": 1.5878, "step": 30060 }, { "epoch": 1.21, "learning_rate": 0.00047023833487660533, "loss": 1.383, "step": 30070 }, { "epoch": 1.21, "learning_rate": 0.0004702282700591811, "loss": 1.4468, "step": 30080 }, { "epoch": 1.21, "learning_rate": 0.0004702182052417569, "loss": 1.5482, "step": 30090 }, { "epoch": 1.21, "learning_rate": 0.0004702081404243327, "loss": 1.3763, "step": 30100 }, { "epoch": 1.21, "learning_rate": 0.00047019807560690855, "loss": 1.5266, "step": 30110 }, { "epoch": 1.21, "learning_rate": 0.0004701880107894843, "loss": 1.4679, "step": 30120 }, { "epoch": 1.21, "learning_rate": 0.0004701779459720601, "loss": 1.2531, "step": 30130 }, { "epoch": 1.21, "learning_rate": 0.00047016788115463587, "loss": 1.4402, "step": 30140 }, { "epoch": 1.21, "learning_rate": 0.00047015781633721166, "loss": 1.6563, "step": 30150 }, { "epoch": 1.21, "learning_rate": 0.00047014775151978745, "loss": 1.4519, "step": 30160 }, { "epoch": 1.21, "learning_rate": 0.00047013768670236324, "loss": 1.5339, "step": 30170 }, { "epoch": 1.21, "learning_rate": 0.00047012762188493903, "loss": 1.411, "step": 30180 }, { "epoch": 1.21, "learning_rate": 0.0004701175570675148, "loss": 1.7301, "step": 30190 }, { "epoch": 1.21, "learning_rate": 0.00047010749225009056, "loss": 1.5449, "step": 30200 }, { "epoch": 1.21, "learning_rate": 0.00047009742743266635, "loss": 1.4901, "step": 30210 }, { "epoch": 1.22, "learning_rate": 0.00047008736261524214, "loss": 1.414, "step": 30220 }, { "epoch": 1.22, "learning_rate": 0.00047007729779781793, "loss": 1.6475, "step": 30230 }, { "epoch": 1.22, "learning_rate": 0.0004700672329803937, "loss": 1.3205, "step": 30240 }, { "epoch": 1.22, "learning_rate": 0.0004700571681629695, "loss": 1.6907, "step": 30250 }, { "epoch": 1.22, "learning_rate": 0.00047004710334554536, "loss": 1.3467, "step": 30260 }, { "epoch": 1.22, "learning_rate": 0.00047003703852812115, "loss": 1.6047, "step": 30270 }, { "epoch": 1.22, "learning_rate": 0.00047002697371069694, "loss": 1.3012, "step": 30280 }, { "epoch": 1.22, "learning_rate": 0.0004700169088932727, "loss": 1.6536, "step": 30290 }, { "epoch": 1.22, "learning_rate": 0.00047000684407584847, "loss": 1.5518, "step": 30300 }, { "epoch": 1.22, "learning_rate": 0.00046999677925842426, "loss": 1.3834, "step": 30310 }, { "epoch": 1.22, "learning_rate": 0.00046998671444100005, "loss": 1.4063, "step": 30320 }, { "epoch": 1.22, "learning_rate": 0.00046997664962357584, "loss": 1.3477, "step": 30330 }, { "epoch": 1.22, "learning_rate": 0.00046996658480615163, "loss": 1.6191, "step": 30340 }, { "epoch": 1.22, "learning_rate": 0.0004699565199887274, "loss": 1.5494, "step": 30350 }, { "epoch": 1.22, "learning_rate": 0.0004699464551713032, "loss": 1.459, "step": 30360 }, { "epoch": 1.22, "learning_rate": 0.00046993639035387895, "loss": 1.4312, "step": 30370 }, { "epoch": 1.22, "learning_rate": 0.00046992632553645474, "loss": 1.5875, "step": 30380 }, { "epoch": 1.22, "learning_rate": 0.00046991626071903053, "loss": 1.3777, "step": 30390 }, { "epoch": 1.22, "learning_rate": 0.0004699061959016064, "loss": 1.3254, "step": 30400 }, { "epoch": 1.22, "learning_rate": 0.00046989613108418217, "loss": 1.4332, "step": 30410 }, { "epoch": 1.22, "learning_rate": 0.00046988606626675796, "loss": 1.2834, "step": 30420 }, { "epoch": 1.22, "learning_rate": 0.00046987600144933375, "loss": 1.5196, "step": 30430 }, { "epoch": 1.22, "learning_rate": 0.00046986593663190954, "loss": 1.4114, "step": 30440 }, { "epoch": 1.22, "learning_rate": 0.00046985587181448533, "loss": 1.541, "step": 30450 }, { "epoch": 1.23, "learning_rate": 0.00046984580699706107, "loss": 1.5207, "step": 30460 }, { "epoch": 1.23, "learning_rate": 0.00046983574217963686, "loss": 1.5318, "step": 30470 }, { "epoch": 1.23, "learning_rate": 0.00046982567736221265, "loss": 1.5108, "step": 30480 }, { "epoch": 1.23, "learning_rate": 0.00046981561254478844, "loss": 1.3666, "step": 30490 }, { "epoch": 1.23, "learning_rate": 0.00046980554772736423, "loss": 1.4047, "step": 30500 }, { "epoch": 1.23, "learning_rate": 0.00046979548290994, "loss": 1.5576, "step": 30510 }, { "epoch": 1.23, "learning_rate": 0.0004697854180925158, "loss": 1.529, "step": 30520 }, { "epoch": 1.23, "learning_rate": 0.0004697753532750916, "loss": 1.6402, "step": 30530 }, { "epoch": 1.23, "learning_rate": 0.00046976528845766734, "loss": 1.4153, "step": 30540 }, { "epoch": 1.23, "learning_rate": 0.0004697552236402432, "loss": 1.4309, "step": 30550 }, { "epoch": 1.23, "learning_rate": 0.000469745158822819, "loss": 1.4751, "step": 30560 }, { "epoch": 1.23, "learning_rate": 0.00046973509400539476, "loss": 1.7574, "step": 30570 }, { "epoch": 1.23, "learning_rate": 0.00046972502918797056, "loss": 1.5249, "step": 30580 }, { "epoch": 1.23, "learning_rate": 0.00046971496437054635, "loss": 1.235, "step": 30590 }, { "epoch": 1.23, "learning_rate": 0.00046970489955312214, "loss": 1.5093, "step": 30600 }, { "epoch": 1.23, "learning_rate": 0.00046969483473569793, "loss": 1.6574, "step": 30610 }, { "epoch": 1.23, "learning_rate": 0.00046968476991827366, "loss": 1.5013, "step": 30620 }, { "epoch": 1.23, "learning_rate": 0.00046967470510084946, "loss": 1.5479, "step": 30630 }, { "epoch": 1.23, "learning_rate": 0.00046966464028342525, "loss": 1.5716, "step": 30640 }, { "epoch": 1.23, "learning_rate": 0.00046965457546600104, "loss": 1.5506, "step": 30650 }, { "epoch": 1.23, "learning_rate": 0.00046964451064857683, "loss": 1.6437, "step": 30660 }, { "epoch": 1.23, "learning_rate": 0.0004696344458311526, "loss": 1.5367, "step": 30670 }, { "epoch": 1.23, "learning_rate": 0.0004696243810137284, "loss": 1.3649, "step": 30680 }, { "epoch": 1.23, "learning_rate": 0.00046961431619630425, "loss": 1.5863, "step": 30690 }, { "epoch": 1.23, "learning_rate": 0.00046960425137888004, "loss": 1.6068, "step": 30700 }, { "epoch": 1.24, "learning_rate": 0.0004695941865614558, "loss": 1.2975, "step": 30710 }, { "epoch": 1.24, "learning_rate": 0.00046958412174403157, "loss": 1.4047, "step": 30720 }, { "epoch": 1.24, "learning_rate": 0.00046957405692660736, "loss": 1.7152, "step": 30730 }, { "epoch": 1.24, "learning_rate": 0.00046956399210918315, "loss": 1.4142, "step": 30740 }, { "epoch": 1.24, "learning_rate": 0.00046955392729175894, "loss": 1.4311, "step": 30750 }, { "epoch": 1.24, "learning_rate": 0.00046954386247433474, "loss": 1.2632, "step": 30760 }, { "epoch": 1.24, "learning_rate": 0.0004695337976569105, "loss": 1.8612, "step": 30770 }, { "epoch": 1.24, "learning_rate": 0.0004695237328394863, "loss": 1.3971, "step": 30780 }, { "epoch": 1.24, "learning_rate": 0.00046951366802206205, "loss": 1.6284, "step": 30790 }, { "epoch": 1.24, "learning_rate": 0.00046950360320463784, "loss": 1.4627, "step": 30800 }, { "epoch": 1.24, "learning_rate": 0.00046949353838721364, "loss": 1.4895, "step": 30810 }, { "epoch": 1.24, "learning_rate": 0.0004694834735697894, "loss": 1.3914, "step": 30820 }, { "epoch": 1.24, "learning_rate": 0.0004694734087523652, "loss": 1.3402, "step": 30830 }, { "epoch": 1.24, "learning_rate": 0.00046946334393494106, "loss": 1.4549, "step": 30840 }, { "epoch": 1.24, "learning_rate": 0.00046945327911751685, "loss": 1.3991, "step": 30850 }, { "epoch": 1.24, "learning_rate": 0.00046944321430009264, "loss": 1.4695, "step": 30860 }, { "epoch": 1.24, "learning_rate": 0.00046943314948266843, "loss": 1.4447, "step": 30870 }, { "epoch": 1.24, "learning_rate": 0.00046942308466524417, "loss": 1.3593, "step": 30880 }, { "epoch": 1.24, "learning_rate": 0.00046941301984781996, "loss": 1.5175, "step": 30890 }, { "epoch": 1.24, "learning_rate": 0.00046940295503039575, "loss": 1.5117, "step": 30900 }, { "epoch": 1.24, "learning_rate": 0.00046939289021297154, "loss": 1.5652, "step": 30910 }, { "epoch": 1.24, "learning_rate": 0.00046938282539554733, "loss": 1.2087, "step": 30920 }, { "epoch": 1.24, "learning_rate": 0.0004693727605781231, "loss": 1.384, "step": 30930 }, { "epoch": 1.24, "learning_rate": 0.0004693626957606989, "loss": 1.5001, "step": 30940 }, { "epoch": 1.24, "learning_rate": 0.0004693526309432747, "loss": 1.5428, "step": 30950 }, { "epoch": 1.25, "learning_rate": 0.00046934256612585044, "loss": 1.3514, "step": 30960 }, { "epoch": 1.25, "learning_rate": 0.00046933250130842623, "loss": 1.6336, "step": 30970 }, { "epoch": 1.25, "learning_rate": 0.000469322436491002, "loss": 1.2357, "step": 30980 }, { "epoch": 1.25, "learning_rate": 0.00046931237167357787, "loss": 1.4134, "step": 30990 }, { "epoch": 1.25, "learning_rate": 0.00046930230685615366, "loss": 1.6222, "step": 31000 }, { "epoch": 1.25, "learning_rate": 0.00046929224203872945, "loss": 1.5473, "step": 31010 }, { "epoch": 1.25, "learning_rate": 0.00046928217722130524, "loss": 1.6036, "step": 31020 }, { "epoch": 1.25, "learning_rate": 0.00046927211240388103, "loss": 1.5475, "step": 31030 }, { "epoch": 1.25, "learning_rate": 0.0004692620475864568, "loss": 1.5004, "step": 31040 }, { "epoch": 1.25, "learning_rate": 0.00046925198276903256, "loss": 1.595, "step": 31050 }, { "epoch": 1.25, "learning_rate": 0.00046924191795160835, "loss": 1.6755, "step": 31060 }, { "epoch": 1.25, "learning_rate": 0.00046923185313418414, "loss": 1.4367, "step": 31070 }, { "epoch": 1.25, "learning_rate": 0.00046922178831675993, "loss": 1.5279, "step": 31080 }, { "epoch": 1.25, "learning_rate": 0.0004692117234993357, "loss": 1.4549, "step": 31090 }, { "epoch": 1.25, "learning_rate": 0.0004692016586819115, "loss": 1.0738, "step": 31100 }, { "epoch": 1.25, "learning_rate": 0.0004691915938644873, "loss": 1.164, "step": 31110 }, { "epoch": 1.25, "learning_rate": 0.0004691815290470631, "loss": 1.4934, "step": 31120 }, { "epoch": 1.25, "learning_rate": 0.0004691714642296389, "loss": 1.4828, "step": 31130 }, { "epoch": 1.25, "learning_rate": 0.0004691613994122147, "loss": 1.5291, "step": 31140 }, { "epoch": 1.25, "learning_rate": 0.00046915133459479047, "loss": 1.5071, "step": 31150 }, { "epoch": 1.25, "learning_rate": 0.00046914126977736626, "loss": 1.2898, "step": 31160 }, { "epoch": 1.25, "learning_rate": 0.00046913120495994205, "loss": 1.4837, "step": 31170 }, { "epoch": 1.25, "learning_rate": 0.00046912114014251784, "loss": 1.4189, "step": 31180 }, { "epoch": 1.25, "learning_rate": 0.00046911107532509363, "loss": 1.3991, "step": 31190 }, { "epoch": 1.25, "learning_rate": 0.0004691010105076694, "loss": 1.6071, "step": 31200 }, { "epoch": 1.26, "learning_rate": 0.0004690909456902452, "loss": 1.4312, "step": 31210 }, { "epoch": 1.26, "learning_rate": 0.00046908088087282095, "loss": 1.8714, "step": 31220 }, { "epoch": 1.26, "learning_rate": 0.00046907081605539674, "loss": 1.4755, "step": 31230 }, { "epoch": 1.26, "learning_rate": 0.00046906075123797253, "loss": 1.6335, "step": 31240 }, { "epoch": 1.26, "learning_rate": 0.0004690506864205483, "loss": 1.4482, "step": 31250 }, { "epoch": 1.26, "learning_rate": 0.0004690406216031241, "loss": 1.5182, "step": 31260 }, { "epoch": 1.26, "learning_rate": 0.0004690305567856999, "loss": 1.4951, "step": 31270 }, { "epoch": 1.26, "learning_rate": 0.00046902049196827575, "loss": 1.5238, "step": 31280 }, { "epoch": 1.26, "learning_rate": 0.00046901042715085154, "loss": 1.5245, "step": 31290 }, { "epoch": 1.26, "learning_rate": 0.0004690003623334273, "loss": 1.7071, "step": 31300 }, { "epoch": 1.26, "learning_rate": 0.00046899029751600307, "loss": 1.477, "step": 31310 }, { "epoch": 1.26, "learning_rate": 0.00046898023269857886, "loss": 1.4845, "step": 31320 }, { "epoch": 1.26, "learning_rate": 0.00046897016788115465, "loss": 1.663, "step": 31330 }, { "epoch": 1.26, "learning_rate": 0.00046896010306373044, "loss": 1.6962, "step": 31340 }, { "epoch": 1.26, "learning_rate": 0.00046895003824630623, "loss": 1.2405, "step": 31350 }, { "epoch": 1.26, "learning_rate": 0.000468939973428882, "loss": 1.4614, "step": 31360 }, { "epoch": 1.26, "learning_rate": 0.0004689299086114578, "loss": 1.4529, "step": 31370 }, { "epoch": 1.26, "learning_rate": 0.00046891984379403355, "loss": 1.7292, "step": 31380 }, { "epoch": 1.26, "learning_rate": 0.00046890977897660934, "loss": 1.3661, "step": 31390 }, { "epoch": 1.26, "learning_rate": 0.00046889971415918513, "loss": 1.4247, "step": 31400 }, { "epoch": 1.26, "learning_rate": 0.0004688896493417609, "loss": 1.49, "step": 31410 }, { "epoch": 1.26, "learning_rate": 0.0004688795845243367, "loss": 1.6583, "step": 31420 }, { "epoch": 1.26, "learning_rate": 0.00046886951970691256, "loss": 1.6202, "step": 31430 }, { "epoch": 1.26, "learning_rate": 0.00046885945488948835, "loss": 1.5987, "step": 31440 }, { "epoch": 1.26, "learning_rate": 0.00046884939007206414, "loss": 1.6899, "step": 31450 }, { "epoch": 1.27, "learning_rate": 0.00046883932525463993, "loss": 1.3255, "step": 31460 }, { "epoch": 1.27, "learning_rate": 0.00046882926043721566, "loss": 1.3282, "step": 31470 }, { "epoch": 1.27, "learning_rate": 0.00046881919561979145, "loss": 1.527, "step": 31480 }, { "epoch": 1.27, "learning_rate": 0.00046880913080236725, "loss": 1.2398, "step": 31490 }, { "epoch": 1.27, "learning_rate": 0.00046879906598494304, "loss": 1.4951, "step": 31500 }, { "epoch": 1.27, "learning_rate": 0.00046878900116751883, "loss": 1.4987, "step": 31510 }, { "epoch": 1.27, "learning_rate": 0.0004687789363500946, "loss": 1.6443, "step": 31520 }, { "epoch": 1.27, "learning_rate": 0.0004687688715326704, "loss": 1.5343, "step": 31530 }, { "epoch": 1.27, "learning_rate": 0.0004687588067152462, "loss": 1.4593, "step": 31540 }, { "epoch": 1.27, "learning_rate": 0.00046874874189782194, "loss": 1.2343, "step": 31550 }, { "epoch": 1.27, "learning_rate": 0.0004687386770803977, "loss": 1.3284, "step": 31560 }, { "epoch": 1.27, "learning_rate": 0.00046872861226297357, "loss": 1.2696, "step": 31570 }, { "epoch": 1.27, "learning_rate": 0.00046871854744554936, "loss": 1.5588, "step": 31580 }, { "epoch": 1.27, "learning_rate": 0.00046870848262812515, "loss": 1.6664, "step": 31590 }, { "epoch": 1.27, "learning_rate": 0.00046869841781070094, "loss": 1.7526, "step": 31600 }, { "epoch": 1.27, "learning_rate": 0.00046868835299327674, "loss": 1.5445, "step": 31610 }, { "epoch": 1.27, "learning_rate": 0.0004686782881758525, "loss": 1.6242, "step": 31620 }, { "epoch": 1.27, "learning_rate": 0.0004686682233584283, "loss": 1.5181, "step": 31630 }, { "epoch": 1.27, "learning_rate": 0.00046865815854100405, "loss": 1.5471, "step": 31640 }, { "epoch": 1.27, "learning_rate": 0.00046864809372357984, "loss": 1.2552, "step": 31650 }, { "epoch": 1.27, "learning_rate": 0.00046863802890615563, "loss": 1.5951, "step": 31660 }, { "epoch": 1.27, "learning_rate": 0.0004686279640887314, "loss": 1.4635, "step": 31670 }, { "epoch": 1.27, "learning_rate": 0.0004686178992713072, "loss": 1.4796, "step": 31680 }, { "epoch": 1.27, "learning_rate": 0.000468607834453883, "loss": 1.561, "step": 31690 }, { "epoch": 1.27, "learning_rate": 0.0004685977696364588, "loss": 1.3958, "step": 31700 }, { "epoch": 1.28, "learning_rate": 0.0004685877048190346, "loss": 1.2455, "step": 31710 }, { "epoch": 1.28, "learning_rate": 0.0004685776400016104, "loss": 1.4618, "step": 31720 }, { "epoch": 1.28, "learning_rate": 0.00046856757518418617, "loss": 1.1827, "step": 31730 }, { "epoch": 1.28, "learning_rate": 0.00046855751036676196, "loss": 1.5519, "step": 31740 }, { "epoch": 1.28, "learning_rate": 0.00046854744554933775, "loss": 1.4478, "step": 31750 }, { "epoch": 1.28, "learning_rate": 0.00046853738073191354, "loss": 1.3835, "step": 31760 }, { "epoch": 1.28, "learning_rate": 0.00046852731591448933, "loss": 1.7259, "step": 31770 }, { "epoch": 1.28, "learning_rate": 0.0004685172510970651, "loss": 1.3811, "step": 31780 }, { "epoch": 1.28, "learning_rate": 0.0004685071862796409, "loss": 1.7901, "step": 31790 }, { "epoch": 1.28, "learning_rate": 0.0004684971214622167, "loss": 1.4364, "step": 31800 }, { "epoch": 1.28, "learning_rate": 0.00046848705664479244, "loss": 1.4233, "step": 31810 }, { "epoch": 1.28, "learning_rate": 0.00046847699182736823, "loss": 1.3783, "step": 31820 }, { "epoch": 1.28, "learning_rate": 0.000468466927009944, "loss": 1.3649, "step": 31830 }, { "epoch": 1.28, "learning_rate": 0.0004684568621925198, "loss": 2.0107, "step": 31840 }, { "epoch": 1.28, "learning_rate": 0.0004684467973750956, "loss": 1.4146, "step": 31850 }, { "epoch": 1.28, "learning_rate": 0.00046843673255767145, "loss": 1.4473, "step": 31860 }, { "epoch": 1.28, "learning_rate": 0.00046842666774024724, "loss": 1.4559, "step": 31870 }, { "epoch": 1.28, "learning_rate": 0.00046841660292282303, "loss": 1.4761, "step": 31880 }, { "epoch": 1.28, "learning_rate": 0.00046840653810539877, "loss": 1.4301, "step": 31890 }, { "epoch": 1.28, "learning_rate": 0.00046839647328797456, "loss": 1.4492, "step": 31900 }, { "epoch": 1.28, "learning_rate": 0.00046838640847055035, "loss": 1.9098, "step": 31910 }, { "epoch": 1.28, "learning_rate": 0.00046837634365312614, "loss": 1.5622, "step": 31920 }, { "epoch": 1.28, "learning_rate": 0.00046836627883570193, "loss": 1.5812, "step": 31930 }, { "epoch": 1.28, "learning_rate": 0.0004683562140182777, "loss": 1.5894, "step": 31940 }, { "epoch": 1.28, "learning_rate": 0.0004683461492008535, "loss": 1.7342, "step": 31950 }, { "epoch": 1.29, "learning_rate": 0.0004683360843834293, "loss": 1.562, "step": 31960 }, { "epoch": 1.29, "learning_rate": 0.0004683260195660051, "loss": 1.3973, "step": 31970 }, { "epoch": 1.29, "learning_rate": 0.00046831595474858083, "loss": 1.7244, "step": 31980 }, { "epoch": 1.29, "learning_rate": 0.0004683058899311566, "loss": 1.49, "step": 31990 }, { "epoch": 1.29, "learning_rate": 0.0004682958251137324, "loss": 1.504, "step": 32000 }, { "epoch": 1.29, "learning_rate": 0.00046828576029630826, "loss": 1.5439, "step": 32010 }, { "epoch": 1.29, "learning_rate": 0.00046827569547888405, "loss": 1.5638, "step": 32020 }, { "epoch": 1.29, "learning_rate": 0.00046826563066145984, "loss": 1.317, "step": 32030 }, { "epoch": 1.29, "learning_rate": 0.00046825556584403563, "loss": 1.485, "step": 32040 }, { "epoch": 1.29, "learning_rate": 0.0004682455010266114, "loss": 1.4856, "step": 32050 }, { "epoch": 1.29, "learning_rate": 0.00046823543620918716, "loss": 1.4138, "step": 32060 }, { "epoch": 1.29, "learning_rate": 0.00046822537139176295, "loss": 1.7877, "step": 32070 }, { "epoch": 1.29, "learning_rate": 0.00046821530657433874, "loss": 1.2049, "step": 32080 }, { "epoch": 1.29, "learning_rate": 0.00046820524175691453, "loss": 1.4026, "step": 32090 }, { "epoch": 1.29, "learning_rate": 0.0004681951769394903, "loss": 1.3838, "step": 32100 }, { "epoch": 1.29, "learning_rate": 0.0004681851121220661, "loss": 1.3881, "step": 32110 }, { "epoch": 1.29, "learning_rate": 0.0004681750473046419, "loss": 1.575, "step": 32120 }, { "epoch": 1.29, "learning_rate": 0.0004681649824872177, "loss": 1.4554, "step": 32130 }, { "epoch": 1.29, "learning_rate": 0.00046815491766979343, "loss": 1.4757, "step": 32140 }, { "epoch": 1.29, "learning_rate": 0.0004681448528523692, "loss": 1.2818, "step": 32150 }, { "epoch": 1.29, "learning_rate": 0.00046813478803494507, "loss": 1.661, "step": 32160 }, { "epoch": 1.29, "learning_rate": 0.00046812472321752086, "loss": 1.4488, "step": 32170 }, { "epoch": 1.29, "learning_rate": 0.00046811465840009665, "loss": 1.2464, "step": 32180 }, { "epoch": 1.29, "learning_rate": 0.00046810459358267244, "loss": 1.2996, "step": 32190 }, { "epoch": 1.3, "learning_rate": 0.00046809452876524823, "loss": 1.4172, "step": 32200 }, { "epoch": 1.3, "learning_rate": 0.000468084463947824, "loss": 1.3983, "step": 32210 }, { "epoch": 1.3, "learning_rate": 0.0004680743991303998, "loss": 1.2381, "step": 32220 }, { "epoch": 1.3, "learning_rate": 0.00046806433431297555, "loss": 1.6427, "step": 32230 }, { "epoch": 1.3, "learning_rate": 0.00046805426949555134, "loss": 1.5171, "step": 32240 }, { "epoch": 1.3, "learning_rate": 0.00046804420467812713, "loss": 1.6218, "step": 32250 }, { "epoch": 1.3, "learning_rate": 0.0004680341398607029, "loss": 1.5099, "step": 32260 }, { "epoch": 1.3, "learning_rate": 0.0004680240750432787, "loss": 1.4637, "step": 32270 }, { "epoch": 1.3, "learning_rate": 0.0004680140102258545, "loss": 1.3752, "step": 32280 }, { "epoch": 1.3, "learning_rate": 0.0004680039454084303, "loss": 1.5616, "step": 32290 }, { "epoch": 1.3, "learning_rate": 0.00046799388059100614, "loss": 1.4687, "step": 32300 }, { "epoch": 1.3, "learning_rate": 0.0004679838157735819, "loss": 1.4375, "step": 32310 }, { "epoch": 1.3, "learning_rate": 0.00046797375095615766, "loss": 1.6019, "step": 32320 }, { "epoch": 1.3, "learning_rate": 0.00046796368613873345, "loss": 1.7638, "step": 32330 }, { "epoch": 1.3, "learning_rate": 0.00046795362132130925, "loss": 1.7985, "step": 32340 }, { "epoch": 1.3, "learning_rate": 0.00046794355650388504, "loss": 1.4728, "step": 32350 }, { "epoch": 1.3, "learning_rate": 0.0004679334916864608, "loss": 1.6285, "step": 32360 }, { "epoch": 1.3, "learning_rate": 0.0004679234268690366, "loss": 1.4542, "step": 32370 }, { "epoch": 1.3, "learning_rate": 0.0004679133620516124, "loss": 1.3173, "step": 32380 }, { "epoch": 1.3, "learning_rate": 0.0004679032972341882, "loss": 1.6973, "step": 32390 }, { "epoch": 1.3, "learning_rate": 0.00046789323241676394, "loss": 1.3706, "step": 32400 }, { "epoch": 1.3, "learning_rate": 0.0004678831675993397, "loss": 1.4884, "step": 32410 }, { "epoch": 1.3, "learning_rate": 0.0004678731027819155, "loss": 1.5854, "step": 32420 }, { "epoch": 1.3, "learning_rate": 0.0004678630379644913, "loss": 1.7609, "step": 32430 }, { "epoch": 1.3, "learning_rate": 0.0004678529731470671, "loss": 1.5095, "step": 32440 }, { "epoch": 1.31, "learning_rate": 0.00046784290832964294, "loss": 1.1473, "step": 32450 }, { "epoch": 1.31, "learning_rate": 0.00046783284351221873, "loss": 1.5666, "step": 32460 }, { "epoch": 1.31, "learning_rate": 0.0004678227786947945, "loss": 1.4913, "step": 32470 }, { "epoch": 1.31, "learning_rate": 0.00046781271387737026, "loss": 1.5256, "step": 32480 }, { "epoch": 1.31, "learning_rate": 0.00046780264905994605, "loss": 1.6519, "step": 32490 }, { "epoch": 1.31, "learning_rate": 0.00046779258424252184, "loss": 1.6656, "step": 32500 }, { "epoch": 1.31, "learning_rate": 0.00046778251942509763, "loss": 1.5864, "step": 32510 }, { "epoch": 1.31, "learning_rate": 0.0004677724546076734, "loss": 1.3819, "step": 32520 }, { "epoch": 1.31, "learning_rate": 0.0004677623897902492, "loss": 1.4123, "step": 32530 }, { "epoch": 1.31, "learning_rate": 0.000467752324972825, "loss": 1.4645, "step": 32540 }, { "epoch": 1.31, "learning_rate": 0.0004677422601554008, "loss": 1.435, "step": 32550 }, { "epoch": 1.31, "learning_rate": 0.0004677321953379766, "loss": 1.3692, "step": 32560 }, { "epoch": 1.31, "learning_rate": 0.0004677221305205523, "loss": 1.5429, "step": 32570 }, { "epoch": 1.31, "learning_rate": 0.0004677120657031281, "loss": 1.5531, "step": 32580 }, { "epoch": 1.31, "learning_rate": 0.0004677020008857039, "loss": 1.375, "step": 32590 }, { "epoch": 1.31, "learning_rate": 0.00046769193606827975, "loss": 1.453, "step": 32600 }, { "epoch": 1.31, "learning_rate": 0.00046768187125085554, "loss": 1.5293, "step": 32610 }, { "epoch": 1.31, "learning_rate": 0.00046767180643343133, "loss": 1.4054, "step": 32620 }, { "epoch": 1.31, "learning_rate": 0.0004676617416160071, "loss": 1.3343, "step": 32630 }, { "epoch": 1.31, "learning_rate": 0.0004676516767985829, "loss": 1.5477, "step": 32640 }, { "epoch": 1.31, "learning_rate": 0.00046764161198115865, "loss": 1.6562, "step": 32650 }, { "epoch": 1.31, "learning_rate": 0.00046763154716373444, "loss": 1.684, "step": 32660 }, { "epoch": 1.31, "learning_rate": 0.00046762148234631023, "loss": 1.5111, "step": 32670 }, { "epoch": 1.31, "learning_rate": 0.000467611417528886, "loss": 1.4364, "step": 32680 }, { "epoch": 1.31, "learning_rate": 0.0004676013527114618, "loss": 1.464, "step": 32690 }, { "epoch": 1.32, "learning_rate": 0.0004675912878940376, "loss": 1.3998, "step": 32700 }, { "epoch": 1.32, "learning_rate": 0.0004675812230766134, "loss": 1.3817, "step": 32710 }, { "epoch": 1.32, "learning_rate": 0.0004675711582591892, "loss": 1.455, "step": 32720 }, { "epoch": 1.32, "learning_rate": 0.000467561093441765, "loss": 1.3109, "step": 32730 }, { "epoch": 1.32, "learning_rate": 0.00046755102862434077, "loss": 1.3453, "step": 32740 }, { "epoch": 1.32, "learning_rate": 0.00046754096380691656, "loss": 1.0217, "step": 32750 }, { "epoch": 1.32, "learning_rate": 0.00046753089898949235, "loss": 1.6901, "step": 32760 }, { "epoch": 1.32, "learning_rate": 0.00046752083417206814, "loss": 1.3344, "step": 32770 }, { "epoch": 1.32, "learning_rate": 0.00046751076935464393, "loss": 1.2605, "step": 32780 }, { "epoch": 1.32, "learning_rate": 0.0004675007045372197, "loss": 1.325, "step": 32790 }, { "epoch": 1.32, "learning_rate": 0.0004674906397197955, "loss": 1.516, "step": 32800 }, { "epoch": 1.32, "learning_rate": 0.0004674805749023713, "loss": 1.4372, "step": 32810 }, { "epoch": 1.32, "learning_rate": 0.00046747051008494704, "loss": 1.6028, "step": 32820 }, { "epoch": 1.32, "learning_rate": 0.00046746044526752283, "loss": 1.3416, "step": 32830 }, { "epoch": 1.32, "learning_rate": 0.0004674503804500986, "loss": 1.4119, "step": 32840 }, { "epoch": 1.32, "learning_rate": 0.0004674403156326744, "loss": 1.6345, "step": 32850 }, { "epoch": 1.32, "learning_rate": 0.0004674302508152502, "loss": 1.4501, "step": 32860 }, { "epoch": 1.32, "learning_rate": 0.000467420185997826, "loss": 1.4208, "step": 32870 }, { "epoch": 1.32, "learning_rate": 0.0004674101211804018, "loss": 1.5226, "step": 32880 }, { "epoch": 1.32, "learning_rate": 0.00046740005636297763, "loss": 1.5063, "step": 32890 }, { "epoch": 1.32, "learning_rate": 0.0004673899915455534, "loss": 1.5267, "step": 32900 }, { "epoch": 1.32, "learning_rate": 0.00046737992672812916, "loss": 1.5212, "step": 32910 }, { "epoch": 1.32, "learning_rate": 0.00046736986191070495, "loss": 1.7657, "step": 32920 }, { "epoch": 1.32, "learning_rate": 0.00046735979709328074, "loss": 1.3322, "step": 32930 }, { "epoch": 1.32, "learning_rate": 0.00046734973227585653, "loss": 1.4432, "step": 32940 }, { "epoch": 1.33, "learning_rate": 0.0004673396674584323, "loss": 1.5508, "step": 32950 }, { "epoch": 1.33, "learning_rate": 0.0004673296026410081, "loss": 1.4144, "step": 32960 }, { "epoch": 1.33, "learning_rate": 0.0004673195378235839, "loss": 1.5107, "step": 32970 }, { "epoch": 1.33, "learning_rate": 0.0004673094730061597, "loss": 1.4879, "step": 32980 }, { "epoch": 1.33, "learning_rate": 0.00046729940818873543, "loss": 1.4285, "step": 32990 }, { "epoch": 1.33, "learning_rate": 0.0004672893433713112, "loss": 1.2835, "step": 33000 }, { "epoch": 1.33, "learning_rate": 0.000467279278553887, "loss": 1.3883, "step": 33010 }, { "epoch": 1.33, "learning_rate": 0.0004672692137364628, "loss": 1.515, "step": 33020 }, { "epoch": 1.33, "learning_rate": 0.00046725914891903865, "loss": 1.3835, "step": 33030 }, { "epoch": 1.33, "learning_rate": 0.00046724908410161444, "loss": 1.6371, "step": 33040 }, { "epoch": 1.33, "learning_rate": 0.00046723901928419023, "loss": 1.3405, "step": 33050 }, { "epoch": 1.33, "learning_rate": 0.000467228954466766, "loss": 1.3808, "step": 33060 }, { "epoch": 1.33, "learning_rate": 0.00046721888964934176, "loss": 1.9553, "step": 33070 }, { "epoch": 1.33, "learning_rate": 0.00046720882483191755, "loss": 1.3048, "step": 33080 }, { "epoch": 1.33, "learning_rate": 0.00046719876001449334, "loss": 1.6423, "step": 33090 }, { "epoch": 1.33, "learning_rate": 0.00046718869519706913, "loss": 1.1335, "step": 33100 }, { "epoch": 1.33, "learning_rate": 0.0004671786303796449, "loss": 1.4725, "step": 33110 }, { "epoch": 1.33, "learning_rate": 0.0004671685655622207, "loss": 1.534, "step": 33120 }, { "epoch": 1.33, "learning_rate": 0.0004671585007447965, "loss": 1.2776, "step": 33130 }, { "epoch": 1.33, "learning_rate": 0.0004671484359273723, "loss": 1.4098, "step": 33140 }, { "epoch": 1.33, "learning_rate": 0.0004671383711099481, "loss": 1.4084, "step": 33150 }, { "epoch": 1.33, "learning_rate": 0.0004671283062925238, "loss": 1.387, "step": 33160 }, { "epoch": 1.33, "learning_rate": 0.0004671182414750996, "loss": 1.4721, "step": 33170 }, { "epoch": 1.33, "learning_rate": 0.00046710817665767545, "loss": 1.4597, "step": 33180 }, { "epoch": 1.33, "learning_rate": 0.00046709811184025125, "loss": 1.2449, "step": 33190 }, { "epoch": 1.34, "learning_rate": 0.00046708804702282704, "loss": 1.6395, "step": 33200 }, { "epoch": 1.34, "learning_rate": 0.0004670779822054028, "loss": 1.3386, "step": 33210 }, { "epoch": 1.34, "learning_rate": 0.0004670679173879786, "loss": 1.1396, "step": 33220 }, { "epoch": 1.34, "learning_rate": 0.0004670578525705544, "loss": 1.6167, "step": 33230 }, { "epoch": 1.34, "learning_rate": 0.00046704778775313014, "loss": 1.3124, "step": 33240 }, { "epoch": 1.34, "learning_rate": 0.00046703772293570594, "loss": 1.4136, "step": 33250 }, { "epoch": 1.34, "learning_rate": 0.0004670276581182817, "loss": 1.3085, "step": 33260 }, { "epoch": 1.34, "learning_rate": 0.0004670175933008575, "loss": 1.3435, "step": 33270 }, { "epoch": 1.34, "learning_rate": 0.0004670075284834333, "loss": 1.6435, "step": 33280 }, { "epoch": 1.34, "learning_rate": 0.0004669974636660091, "loss": 1.161, "step": 33290 }, { "epoch": 1.34, "learning_rate": 0.0004669873988485849, "loss": 1.6687, "step": 33300 }, { "epoch": 1.34, "learning_rate": 0.0004669773340311607, "loss": 1.6194, "step": 33310 }, { "epoch": 1.34, "learning_rate": 0.00046696726921373647, "loss": 1.2503, "step": 33320 }, { "epoch": 1.34, "learning_rate": 0.00046695720439631226, "loss": 1.5898, "step": 33330 }, { "epoch": 1.34, "learning_rate": 0.00046694713957888805, "loss": 1.5588, "step": 33340 }, { "epoch": 1.34, "learning_rate": 0.00046693707476146384, "loss": 1.5056, "step": 33350 }, { "epoch": 1.34, "learning_rate": 0.00046692700994403963, "loss": 1.3873, "step": 33360 }, { "epoch": 1.34, "learning_rate": 0.0004669169451266154, "loss": 1.1706, "step": 33370 }, { "epoch": 1.34, "learning_rate": 0.0004669068803091912, "loss": 1.5666, "step": 33380 }, { "epoch": 1.34, "learning_rate": 0.000466896815491767, "loss": 1.5391, "step": 33390 }, { "epoch": 1.34, "learning_rate": 0.0004668867506743428, "loss": 1.5989, "step": 33400 }, { "epoch": 1.34, "learning_rate": 0.00046687668585691853, "loss": 1.4326, "step": 33410 }, { "epoch": 1.34, "learning_rate": 0.0004668666210394943, "loss": 1.3117, "step": 33420 }, { "epoch": 1.34, "learning_rate": 0.0004668565562220701, "loss": 1.659, "step": 33430 }, { "epoch": 1.34, "learning_rate": 0.0004668464914046459, "loss": 1.4273, "step": 33440 }, { "epoch": 1.35, "learning_rate": 0.0004668364265872217, "loss": 1.7859, "step": 33450 }, { "epoch": 1.35, "learning_rate": 0.0004668263617697975, "loss": 1.3847, "step": 33460 }, { "epoch": 1.35, "learning_rate": 0.00046681629695237333, "loss": 1.2412, "step": 33470 }, { "epoch": 1.35, "learning_rate": 0.0004668062321349491, "loss": 1.558, "step": 33480 }, { "epoch": 1.35, "learning_rate": 0.0004667961673175249, "loss": 1.3632, "step": 33490 }, { "epoch": 1.35, "learning_rate": 0.00046678610250010065, "loss": 1.4701, "step": 33500 }, { "epoch": 1.35, "learning_rate": 0.00046677603768267644, "loss": 1.4697, "step": 33510 }, { "epoch": 1.35, "learning_rate": 0.00046676597286525223, "loss": 1.5543, "step": 33520 }, { "epoch": 1.35, "learning_rate": 0.000466755908047828, "loss": 1.3538, "step": 33530 }, { "epoch": 1.35, "learning_rate": 0.0004667458432304038, "loss": 1.5487, "step": 33540 }, { "epoch": 1.35, "learning_rate": 0.0004667357784129796, "loss": 1.422, "step": 33550 }, { "epoch": 1.35, "learning_rate": 0.0004667257135955554, "loss": 1.5703, "step": 33560 }, { "epoch": 1.35, "learning_rate": 0.0004667156487781312, "loss": 1.6325, "step": 33570 }, { "epoch": 1.35, "learning_rate": 0.0004667055839607069, "loss": 1.7053, "step": 33580 }, { "epoch": 1.35, "learning_rate": 0.0004666955191432827, "loss": 1.4952, "step": 33590 }, { "epoch": 1.35, "learning_rate": 0.0004666854543258585, "loss": 1.5614, "step": 33600 }, { "epoch": 1.35, "learning_rate": 0.0004666753895084343, "loss": 1.2268, "step": 33610 }, { "epoch": 1.35, "learning_rate": 0.00046666532469101014, "loss": 1.4735, "step": 33620 }, { "epoch": 1.35, "learning_rate": 0.00046665525987358593, "loss": 1.7238, "step": 33630 }, { "epoch": 1.35, "learning_rate": 0.0004666451950561617, "loss": 1.471, "step": 33640 }, { "epoch": 1.35, "learning_rate": 0.0004666351302387375, "loss": 1.7486, "step": 33650 }, { "epoch": 1.35, "learning_rate": 0.0004666250654213133, "loss": 1.3785, "step": 33660 }, { "epoch": 1.35, "learning_rate": 0.00046661500060388904, "loss": 1.087, "step": 33670 }, { "epoch": 1.35, "learning_rate": 0.00046660493578646483, "loss": 1.3574, "step": 33680 }, { "epoch": 1.35, "learning_rate": 0.0004665948709690406, "loss": 1.4848, "step": 33690 }, { "epoch": 1.36, "learning_rate": 0.0004665848061516164, "loss": 1.5418, "step": 33700 }, { "epoch": 1.36, "learning_rate": 0.0004665747413341922, "loss": 1.4022, "step": 33710 }, { "epoch": 1.36, "learning_rate": 0.000466564676516768, "loss": 1.3759, "step": 33720 }, { "epoch": 1.36, "learning_rate": 0.0004665546116993438, "loss": 1.5805, "step": 33730 }, { "epoch": 1.36, "learning_rate": 0.0004665445468819196, "loss": 1.3856, "step": 33740 }, { "epoch": 1.36, "learning_rate": 0.0004665344820644953, "loss": 1.6122, "step": 33750 }, { "epoch": 1.36, "learning_rate": 0.00046652441724707116, "loss": 1.6912, "step": 33760 }, { "epoch": 1.36, "learning_rate": 0.00046651435242964695, "loss": 1.4421, "step": 33770 }, { "epoch": 1.36, "learning_rate": 0.00046650428761222274, "loss": 1.4644, "step": 33780 }, { "epoch": 1.36, "learning_rate": 0.00046649422279479853, "loss": 1.3355, "step": 33790 }, { "epoch": 1.36, "learning_rate": 0.0004664841579773743, "loss": 1.5833, "step": 33800 }, { "epoch": 1.36, "learning_rate": 0.0004664740931599501, "loss": 1.3316, "step": 33810 }, { "epoch": 1.36, "learning_rate": 0.0004664640283425259, "loss": 1.391, "step": 33820 }, { "epoch": 1.36, "learning_rate": 0.00046645396352510164, "loss": 1.4264, "step": 33830 }, { "epoch": 1.36, "learning_rate": 0.00046644389870767743, "loss": 1.3209, "step": 33840 }, { "epoch": 1.36, "learning_rate": 0.0004664338338902532, "loss": 1.3251, "step": 33850 }, { "epoch": 1.36, "learning_rate": 0.000466423769072829, "loss": 1.6028, "step": 33860 }, { "epoch": 1.36, "learning_rate": 0.0004664137042554048, "loss": 1.5238, "step": 33870 }, { "epoch": 1.36, "learning_rate": 0.0004664036394379806, "loss": 1.4265, "step": 33880 }, { "epoch": 1.36, "learning_rate": 0.0004663935746205564, "loss": 1.4611, "step": 33890 }, { "epoch": 1.36, "learning_rate": 0.0004663835098031322, "loss": 1.7129, "step": 33900 }, { "epoch": 1.36, "learning_rate": 0.000466373444985708, "loss": 1.3338, "step": 33910 }, { "epoch": 1.36, "learning_rate": 0.00046636338016828376, "loss": 1.6003, "step": 33920 }, { "epoch": 1.36, "learning_rate": 0.00046635331535085955, "loss": 1.4766, "step": 33930 }, { "epoch": 1.37, "learning_rate": 0.00046634325053343534, "loss": 1.5665, "step": 33940 }, { "epoch": 1.37, "learning_rate": 0.00046633318571601113, "loss": 1.4521, "step": 33950 }, { "epoch": 1.37, "learning_rate": 0.0004663231208985869, "loss": 1.5044, "step": 33960 }, { "epoch": 1.37, "learning_rate": 0.0004663130560811627, "loss": 1.4044, "step": 33970 }, { "epoch": 1.37, "learning_rate": 0.0004663029912637385, "loss": 1.5686, "step": 33980 }, { "epoch": 1.37, "learning_rate": 0.0004662929264463143, "loss": 1.4006, "step": 33990 }, { "epoch": 1.37, "learning_rate": 0.00046628286162889003, "loss": 1.5731, "step": 34000 }, { "epoch": 1.37, "learning_rate": 0.0004662727968114658, "loss": 1.6657, "step": 34010 }, { "epoch": 1.37, "learning_rate": 0.0004662627319940416, "loss": 1.4704, "step": 34020 }, { "epoch": 1.37, "learning_rate": 0.0004662526671766174, "loss": 1.5126, "step": 34030 }, { "epoch": 1.37, "learning_rate": 0.0004662426023591932, "loss": 1.2998, "step": 34040 }, { "epoch": 1.37, "learning_rate": 0.000466232537541769, "loss": 1.2524, "step": 34050 }, { "epoch": 1.37, "learning_rate": 0.0004662224727243448, "loss": 1.3434, "step": 34060 }, { "epoch": 1.37, "learning_rate": 0.0004662124079069206, "loss": 1.2801, "step": 34070 }, { "epoch": 1.37, "learning_rate": 0.0004662023430894964, "loss": 1.516, "step": 34080 }, { "epoch": 1.37, "learning_rate": 0.00046619227827207214, "loss": 1.5924, "step": 34090 }, { "epoch": 1.37, "learning_rate": 0.00046618221345464794, "loss": 1.2831, "step": 34100 }, { "epoch": 1.37, "learning_rate": 0.0004661721486372237, "loss": 1.352, "step": 34110 }, { "epoch": 1.37, "learning_rate": 0.0004661620838197995, "loss": 1.5248, "step": 34120 }, { "epoch": 1.37, "learning_rate": 0.0004661520190023753, "loss": 1.5671, "step": 34130 }, { "epoch": 1.37, "learning_rate": 0.0004661419541849511, "loss": 1.7444, "step": 34140 }, { "epoch": 1.37, "learning_rate": 0.0004661318893675269, "loss": 1.7313, "step": 34150 }, { "epoch": 1.37, "learning_rate": 0.0004661218245501027, "loss": 1.5775, "step": 34160 }, { "epoch": 1.37, "learning_rate": 0.0004661117597326784, "loss": 1.6445, "step": 34170 }, { "epoch": 1.37, "learning_rate": 0.0004661016949152542, "loss": 1.5267, "step": 34180 }, { "epoch": 1.38, "learning_rate": 0.00046609163009783, "loss": 1.6285, "step": 34190 }, { "epoch": 1.38, "learning_rate": 0.00046608156528040584, "loss": 1.5999, "step": 34200 }, { "epoch": 1.38, "learning_rate": 0.00046607150046298163, "loss": 1.2489, "step": 34210 }, { "epoch": 1.38, "learning_rate": 0.0004660614356455574, "loss": 1.4281, "step": 34220 }, { "epoch": 1.38, "learning_rate": 0.0004660513708281332, "loss": 1.7239, "step": 34230 }, { "epoch": 1.38, "learning_rate": 0.000466041306010709, "loss": 1.7922, "step": 34240 }, { "epoch": 1.38, "learning_rate": 0.0004660312411932848, "loss": 1.3626, "step": 34250 }, { "epoch": 1.38, "learning_rate": 0.00046602117637586053, "loss": 1.2382, "step": 34260 }, { "epoch": 1.38, "learning_rate": 0.0004660111115584363, "loss": 1.3773, "step": 34270 }, { "epoch": 1.38, "learning_rate": 0.0004660010467410121, "loss": 1.6336, "step": 34280 }, { "epoch": 1.38, "learning_rate": 0.0004659909819235879, "loss": 1.6313, "step": 34290 }, { "epoch": 1.38, "learning_rate": 0.0004659809171061637, "loss": 1.5732, "step": 34300 }, { "epoch": 1.38, "learning_rate": 0.0004659708522887395, "loss": 1.4226, "step": 34310 }, { "epoch": 1.38, "learning_rate": 0.0004659607874713153, "loss": 1.5035, "step": 34320 }, { "epoch": 1.38, "learning_rate": 0.00046595072265389107, "loss": 1.8273, "step": 34330 }, { "epoch": 1.38, "learning_rate": 0.0004659406578364668, "loss": 1.4332, "step": 34340 }, { "epoch": 1.38, "learning_rate": 0.00046593059301904265, "loss": 1.6402, "step": 34350 }, { "epoch": 1.38, "learning_rate": 0.00046592052820161844, "loss": 1.5412, "step": 34360 }, { "epoch": 1.38, "learning_rate": 0.00046591046338419423, "loss": 1.5304, "step": 34370 }, { "epoch": 1.38, "learning_rate": 0.00046590039856677, "loss": 1.3894, "step": 34380 }, { "epoch": 1.38, "learning_rate": 0.0004658903337493458, "loss": 1.5483, "step": 34390 }, { "epoch": 1.38, "learning_rate": 0.0004658802689319216, "loss": 1.4649, "step": 34400 }, { "epoch": 1.38, "learning_rate": 0.0004658702041144974, "loss": 1.6837, "step": 34410 }, { "epoch": 1.38, "learning_rate": 0.0004658601392970732, "loss": 1.5019, "step": 34420 }, { "epoch": 1.38, "learning_rate": 0.0004658500744796489, "loss": 1.3949, "step": 34430 }, { "epoch": 1.39, "learning_rate": 0.0004658400096622247, "loss": 1.4127, "step": 34440 }, { "epoch": 1.39, "learning_rate": 0.0004658299448448005, "loss": 1.5204, "step": 34450 }, { "epoch": 1.39, "learning_rate": 0.0004658198800273763, "loss": 1.2738, "step": 34460 }, { "epoch": 1.39, "learning_rate": 0.0004658098152099521, "loss": 1.657, "step": 34470 }, { "epoch": 1.39, "learning_rate": 0.0004657997503925279, "loss": 1.6716, "step": 34480 }, { "epoch": 1.39, "learning_rate": 0.00046578968557510367, "loss": 1.5894, "step": 34490 }, { "epoch": 1.39, "learning_rate": 0.0004657796207576795, "loss": 1.4113, "step": 34500 }, { "epoch": 1.39, "learning_rate": 0.00046576955594025525, "loss": 1.3749, "step": 34510 }, { "epoch": 1.39, "learning_rate": 0.00046575949112283104, "loss": 1.5666, "step": 34520 }, { "epoch": 1.39, "learning_rate": 0.00046574942630540683, "loss": 1.4802, "step": 34530 }, { "epoch": 1.39, "learning_rate": 0.0004657393614879826, "loss": 1.7317, "step": 34540 }, { "epoch": 1.39, "learning_rate": 0.0004657292966705584, "loss": 1.5558, "step": 34550 }, { "epoch": 1.39, "learning_rate": 0.0004657192318531342, "loss": 1.3453, "step": 34560 }, { "epoch": 1.39, "learning_rate": 0.00046570916703571, "loss": 1.6852, "step": 34570 }, { "epoch": 1.39, "learning_rate": 0.0004656991022182858, "loss": 1.4636, "step": 34580 }, { "epoch": 1.39, "learning_rate": 0.0004656890374008615, "loss": 1.2332, "step": 34590 }, { "epoch": 1.39, "learning_rate": 0.0004656789725834373, "loss": 1.7058, "step": 34600 }, { "epoch": 1.39, "learning_rate": 0.0004656689077660131, "loss": 1.5405, "step": 34610 }, { "epoch": 1.39, "learning_rate": 0.0004656588429485889, "loss": 1.3329, "step": 34620 }, { "epoch": 1.39, "learning_rate": 0.0004656487781311647, "loss": 1.5842, "step": 34630 }, { "epoch": 1.39, "learning_rate": 0.00046563871331374053, "loss": 1.3755, "step": 34640 }, { "epoch": 1.39, "learning_rate": 0.0004656286484963163, "loss": 1.6003, "step": 34650 }, { "epoch": 1.39, "learning_rate": 0.0004656185836788921, "loss": 1.4051, "step": 34660 }, { "epoch": 1.39, "learning_rate": 0.0004656085188614679, "loss": 1.4908, "step": 34670 }, { "epoch": 1.39, "learning_rate": 0.00046559845404404364, "loss": 1.6503, "step": 34680 }, { "epoch": 1.4, "learning_rate": 0.00046558838922661943, "loss": 1.7375, "step": 34690 }, { "epoch": 1.4, "learning_rate": 0.0004655783244091952, "loss": 1.3201, "step": 34700 }, { "epoch": 1.4, "learning_rate": 0.000465568259591771, "loss": 1.3831, "step": 34710 }, { "epoch": 1.4, "learning_rate": 0.0004655581947743468, "loss": 1.708, "step": 34720 }, { "epoch": 1.4, "learning_rate": 0.0004655481299569226, "loss": 1.3548, "step": 34730 }, { "epoch": 1.4, "learning_rate": 0.0004655380651394984, "loss": 1.4736, "step": 34740 }, { "epoch": 1.4, "learning_rate": 0.0004655280003220742, "loss": 1.7211, "step": 34750 }, { "epoch": 1.4, "learning_rate": 0.0004655179355046499, "loss": 1.3724, "step": 34760 }, { "epoch": 1.4, "learning_rate": 0.0004655078706872257, "loss": 1.4633, "step": 34770 }, { "epoch": 1.4, "learning_rate": 0.0004654978058698015, "loss": 1.3845, "step": 34780 }, { "epoch": 1.4, "learning_rate": 0.00046548774105237734, "loss": 1.4298, "step": 34790 }, { "epoch": 1.4, "learning_rate": 0.00046547767623495313, "loss": 1.3247, "step": 34800 }, { "epoch": 1.4, "learning_rate": 0.0004654676114175289, "loss": 1.6524, "step": 34810 }, { "epoch": 1.4, "learning_rate": 0.0004654575466001047, "loss": 1.1181, "step": 34820 }, { "epoch": 1.4, "learning_rate": 0.0004654474817826805, "loss": 1.7522, "step": 34830 }, { "epoch": 1.4, "learning_rate": 0.0004654374169652563, "loss": 1.3551, "step": 34840 }, { "epoch": 1.4, "learning_rate": 0.00046542735214783203, "loss": 1.6473, "step": 34850 }, { "epoch": 1.4, "learning_rate": 0.0004654172873304078, "loss": 1.4606, "step": 34860 }, { "epoch": 1.4, "learning_rate": 0.0004654072225129836, "loss": 1.8224, "step": 34870 }, { "epoch": 1.4, "learning_rate": 0.0004653971576955594, "loss": 1.4589, "step": 34880 }, { "epoch": 1.4, "learning_rate": 0.0004653870928781352, "loss": 1.5953, "step": 34890 }, { "epoch": 1.4, "learning_rate": 0.000465377028060711, "loss": 1.3382, "step": 34900 }, { "epoch": 1.4, "learning_rate": 0.00046536696324328677, "loss": 1.61, "step": 34910 }, { "epoch": 1.4, "learning_rate": 0.00046535689842586256, "loss": 1.7125, "step": 34920 }, { "epoch": 1.4, "learning_rate": 0.00046534683360843835, "loss": 1.3851, "step": 34930 }, { "epoch": 1.41, "learning_rate": 0.00046533676879101414, "loss": 1.4695, "step": 34940 }, { "epoch": 1.41, "learning_rate": 0.00046532670397358994, "loss": 1.2941, "step": 34950 }, { "epoch": 1.41, "learning_rate": 0.0004653166391561657, "loss": 1.5191, "step": 34960 }, { "epoch": 1.41, "learning_rate": 0.0004653065743387415, "loss": 1.4383, "step": 34970 }, { "epoch": 1.41, "learning_rate": 0.0004652965095213173, "loss": 1.4916, "step": 34980 }, { "epoch": 1.41, "learning_rate": 0.0004652864447038931, "loss": 1.6508, "step": 34990 }, { "epoch": 1.41, "learning_rate": 0.0004652763798864689, "loss": 1.3317, "step": 35000 }, { "epoch": 1.41, "learning_rate": 0.0004652663150690447, "loss": 1.478, "step": 35010 }, { "epoch": 1.41, "learning_rate": 0.0004652562502516204, "loss": 1.5816, "step": 35020 }, { "epoch": 1.41, "learning_rate": 0.0004652461854341962, "loss": 1.4855, "step": 35030 }, { "epoch": 1.41, "learning_rate": 0.000465236120616772, "loss": 1.3781, "step": 35040 }, { "epoch": 1.41, "learning_rate": 0.0004652260557993478, "loss": 1.2513, "step": 35050 }, { "epoch": 1.41, "learning_rate": 0.0004652159909819236, "loss": 1.411, "step": 35060 }, { "epoch": 1.41, "learning_rate": 0.00046520592616449937, "loss": 1.6063, "step": 35070 }, { "epoch": 1.41, "learning_rate": 0.0004651958613470752, "loss": 1.4723, "step": 35080 }, { "epoch": 1.41, "learning_rate": 0.000465185796529651, "loss": 1.6229, "step": 35090 }, { "epoch": 1.41, "learning_rate": 0.00046517573171222674, "loss": 1.3946, "step": 35100 }, { "epoch": 1.41, "learning_rate": 0.00046516566689480253, "loss": 1.1963, "step": 35110 }, { "epoch": 1.41, "learning_rate": 0.0004651556020773783, "loss": 1.529, "step": 35120 }, { "epoch": 1.41, "learning_rate": 0.0004651455372599541, "loss": 1.6123, "step": 35130 }, { "epoch": 1.41, "learning_rate": 0.0004651354724425299, "loss": 1.3844, "step": 35140 }, { "epoch": 1.41, "learning_rate": 0.0004651254076251057, "loss": 1.5796, "step": 35150 }, { "epoch": 1.41, "learning_rate": 0.0004651153428076815, "loss": 1.5443, "step": 35160 }, { "epoch": 1.41, "learning_rate": 0.0004651052779902573, "loss": 1.6023, "step": 35170 }, { "epoch": 1.41, "learning_rate": 0.00046509521317283307, "loss": 1.3033, "step": 35180 }, { "epoch": 1.42, "learning_rate": 0.0004650851483554088, "loss": 1.7883, "step": 35190 }, { "epoch": 1.42, "learning_rate": 0.0004650750835379846, "loss": 1.6801, "step": 35200 }, { "epoch": 1.42, "learning_rate": 0.0004650650187205604, "loss": 1.0667, "step": 35210 }, { "epoch": 1.42, "learning_rate": 0.0004650549539031362, "loss": 1.3868, "step": 35220 }, { "epoch": 1.42, "learning_rate": 0.000465044889085712, "loss": 1.376, "step": 35230 }, { "epoch": 1.42, "learning_rate": 0.0004650348242682878, "loss": 1.3685, "step": 35240 }, { "epoch": 1.42, "learning_rate": 0.0004650247594508636, "loss": 2.0172, "step": 35250 }, { "epoch": 1.42, "learning_rate": 0.0004650146946334394, "loss": 1.5438, "step": 35260 }, { "epoch": 1.42, "learning_rate": 0.00046500462981601513, "loss": 1.5906, "step": 35270 }, { "epoch": 1.42, "learning_rate": 0.0004649945649985909, "loss": 1.4462, "step": 35280 }, { "epoch": 1.42, "learning_rate": 0.0004649845001811667, "loss": 1.6944, "step": 35290 }, { "epoch": 1.42, "learning_rate": 0.0004649744353637425, "loss": 1.2147, "step": 35300 }, { "epoch": 1.42, "learning_rate": 0.0004649643705463183, "loss": 1.314, "step": 35310 }, { "epoch": 1.42, "learning_rate": 0.0004649543057288941, "loss": 1.6662, "step": 35320 }, { "epoch": 1.42, "learning_rate": 0.0004649442409114699, "loss": 1.486, "step": 35330 }, { "epoch": 1.42, "learning_rate": 0.00046493417609404567, "loss": 1.6636, "step": 35340 }, { "epoch": 1.42, "learning_rate": 0.0004649241112766214, "loss": 1.4056, "step": 35350 }, { "epoch": 1.42, "learning_rate": 0.0004649140464591972, "loss": 1.6116, "step": 35360 }, { "epoch": 1.42, "learning_rate": 0.00046490398164177304, "loss": 1.5842, "step": 35370 }, { "epoch": 1.42, "learning_rate": 0.00046489391682434883, "loss": 1.42, "step": 35380 }, { "epoch": 1.42, "learning_rate": 0.0004648838520069246, "loss": 1.4587, "step": 35390 }, { "epoch": 1.42, "learning_rate": 0.0004648737871895004, "loss": 1.4955, "step": 35400 }, { "epoch": 1.42, "learning_rate": 0.0004648637223720762, "loss": 1.3622, "step": 35410 }, { "epoch": 1.42, "learning_rate": 0.000464853657554652, "loss": 1.4842, "step": 35420 }, { "epoch": 1.42, "learning_rate": 0.0004648435927372278, "loss": 1.5476, "step": 35430 }, { "epoch": 1.43, "learning_rate": 0.0004648335279198035, "loss": 1.2514, "step": 35440 }, { "epoch": 1.43, "learning_rate": 0.0004648234631023793, "loss": 1.4776, "step": 35450 }, { "epoch": 1.43, "learning_rate": 0.0004648133982849551, "loss": 1.4991, "step": 35460 }, { "epoch": 1.43, "learning_rate": 0.0004648033334675309, "loss": 1.5792, "step": 35470 }, { "epoch": 1.43, "learning_rate": 0.0004647932686501067, "loss": 1.7914, "step": 35480 }, { "epoch": 1.43, "learning_rate": 0.0004647832038326825, "loss": 1.5965, "step": 35490 }, { "epoch": 1.43, "learning_rate": 0.00046477313901525827, "loss": 1.1593, "step": 35500 }, { "epoch": 1.43, "learning_rate": 0.00046476307419783406, "loss": 1.5145, "step": 35510 }, { "epoch": 1.43, "learning_rate": 0.0004647530093804099, "loss": 1.4931, "step": 35520 }, { "epoch": 1.43, "learning_rate": 0.00046474294456298564, "loss": 1.3488, "step": 35530 }, { "epoch": 1.43, "learning_rate": 0.00046473287974556143, "loss": 1.3295, "step": 35540 }, { "epoch": 1.43, "learning_rate": 0.0004647228149281372, "loss": 1.7341, "step": 35550 }, { "epoch": 1.43, "learning_rate": 0.000464712750110713, "loss": 1.4646, "step": 35560 }, { "epoch": 1.43, "learning_rate": 0.0004647026852932888, "loss": 1.4374, "step": 35570 }, { "epoch": 1.43, "learning_rate": 0.0004646926204758646, "loss": 1.6407, "step": 35580 }, { "epoch": 1.43, "learning_rate": 0.0004646825556584404, "loss": 1.6111, "step": 35590 }, { "epoch": 1.43, "learning_rate": 0.0004646724908410162, "loss": 1.3556, "step": 35600 }, { "epoch": 1.43, "learning_rate": 0.0004646624260235919, "loss": 1.5718, "step": 35610 }, { "epoch": 1.43, "learning_rate": 0.0004646523612061677, "loss": 1.4146, "step": 35620 }, { "epoch": 1.43, "learning_rate": 0.0004646422963887435, "loss": 1.3445, "step": 35630 }, { "epoch": 1.43, "learning_rate": 0.0004646322315713193, "loss": 1.5992, "step": 35640 }, { "epoch": 1.43, "learning_rate": 0.0004646221667538951, "loss": 1.1719, "step": 35650 }, { "epoch": 1.43, "learning_rate": 0.00046461210193647086, "loss": 1.4951, "step": 35660 }, { "epoch": 1.43, "learning_rate": 0.0004646020371190467, "loss": 1.6268, "step": 35670 }, { "epoch": 1.43, "learning_rate": 0.0004645919723016225, "loss": 1.6092, "step": 35680 }, { "epoch": 1.44, "learning_rate": 0.00046458190748419824, "loss": 1.5187, "step": 35690 }, { "epoch": 1.44, "learning_rate": 0.000464571842666774, "loss": 1.5749, "step": 35700 }, { "epoch": 1.44, "learning_rate": 0.0004645617778493498, "loss": 1.5125, "step": 35710 }, { "epoch": 1.44, "learning_rate": 0.0004645517130319256, "loss": 1.6716, "step": 35720 }, { "epoch": 1.44, "learning_rate": 0.0004645416482145014, "loss": 1.5108, "step": 35730 }, { "epoch": 1.44, "learning_rate": 0.0004645315833970772, "loss": 1.5988, "step": 35740 }, { "epoch": 1.44, "learning_rate": 0.000464521518579653, "loss": 1.5421, "step": 35750 }, { "epoch": 1.44, "learning_rate": 0.00046451145376222877, "loss": 1.6296, "step": 35760 }, { "epoch": 1.44, "learning_rate": 0.00046450138894480456, "loss": 1.6397, "step": 35770 }, { "epoch": 1.44, "learning_rate": 0.0004644913241273803, "loss": 1.3047, "step": 35780 }, { "epoch": 1.44, "learning_rate": 0.0004644812593099561, "loss": 1.4022, "step": 35790 }, { "epoch": 1.44, "learning_rate": 0.0004644711944925319, "loss": 1.6371, "step": 35800 }, { "epoch": 1.44, "learning_rate": 0.0004644611296751077, "loss": 1.5644, "step": 35810 }, { "epoch": 1.44, "learning_rate": 0.0004644510648576835, "loss": 1.6331, "step": 35820 }, { "epoch": 1.44, "learning_rate": 0.0004644410000402593, "loss": 1.4136, "step": 35830 }, { "epoch": 1.44, "learning_rate": 0.0004644309352228351, "loss": 1.5711, "step": 35840 }, { "epoch": 1.44, "learning_rate": 0.0004644208704054109, "loss": 1.2954, "step": 35850 }, { "epoch": 1.44, "learning_rate": 0.0004644108055879866, "loss": 1.4602, "step": 35860 }, { "epoch": 1.44, "learning_rate": 0.0004644007407705624, "loss": 1.4733, "step": 35870 }, { "epoch": 1.44, "learning_rate": 0.0004643906759531382, "loss": 1.3806, "step": 35880 }, { "epoch": 1.44, "learning_rate": 0.000464380611135714, "loss": 1.5993, "step": 35890 }, { "epoch": 1.44, "learning_rate": 0.0004643705463182898, "loss": 1.5249, "step": 35900 }, { "epoch": 1.44, "learning_rate": 0.0004643604815008656, "loss": 1.4029, "step": 35910 }, { "epoch": 1.44, "learning_rate": 0.00046435041668344137, "loss": 1.2719, "step": 35920 }, { "epoch": 1.45, "learning_rate": 0.00046434035186601716, "loss": 1.5475, "step": 35930 }, { "epoch": 1.45, "learning_rate": 0.00046433028704859295, "loss": 1.6132, "step": 35940 }, { "epoch": 1.45, "learning_rate": 0.0004643202222311687, "loss": 1.497, "step": 35950 }, { "epoch": 1.45, "learning_rate": 0.00046431015741374453, "loss": 1.5359, "step": 35960 }, { "epoch": 1.45, "learning_rate": 0.0004643000925963203, "loss": 1.3835, "step": 35970 }, { "epoch": 1.45, "learning_rate": 0.0004642900277788961, "loss": 1.5268, "step": 35980 }, { "epoch": 1.45, "learning_rate": 0.0004642799629614719, "loss": 1.4829, "step": 35990 }, { "epoch": 1.45, "learning_rate": 0.0004642698981440477, "loss": 1.6537, "step": 36000 }, { "epoch": 1.45, "learning_rate": 0.0004642598333266235, "loss": 1.458, "step": 36010 }, { "epoch": 1.45, "learning_rate": 0.0004642497685091993, "loss": 1.6264, "step": 36020 }, { "epoch": 1.45, "learning_rate": 0.000464239703691775, "loss": 1.5177, "step": 36030 }, { "epoch": 1.45, "learning_rate": 0.0004642296388743508, "loss": 1.586, "step": 36040 }, { "epoch": 1.45, "learning_rate": 0.0004642195740569266, "loss": 1.494, "step": 36050 }, { "epoch": 1.45, "learning_rate": 0.0004642095092395024, "loss": 1.4115, "step": 36060 }, { "epoch": 1.45, "learning_rate": 0.0004641994444220782, "loss": 1.4646, "step": 36070 }, { "epoch": 1.45, "learning_rate": 0.00046418937960465397, "loss": 1.6155, "step": 36080 }, { "epoch": 1.45, "learning_rate": 0.00046417931478722976, "loss": 1.4057, "step": 36090 }, { "epoch": 1.45, "learning_rate": 0.0004641692499698056, "loss": 1.549, "step": 36100 }, { "epoch": 1.45, "learning_rate": 0.0004641591851523814, "loss": 1.3887, "step": 36110 }, { "epoch": 1.45, "learning_rate": 0.00046414912033495713, "loss": 1.4975, "step": 36120 }, { "epoch": 1.45, "learning_rate": 0.0004641390555175329, "loss": 1.6548, "step": 36130 }, { "epoch": 1.45, "learning_rate": 0.0004641289907001087, "loss": 1.5373, "step": 36140 }, { "epoch": 1.45, "learning_rate": 0.0004641189258826845, "loss": 1.2928, "step": 36150 }, { "epoch": 1.45, "learning_rate": 0.0004641088610652603, "loss": 1.4875, "step": 36160 }, { "epoch": 1.45, "learning_rate": 0.0004640987962478361, "loss": 1.5171, "step": 36170 }, { "epoch": 1.46, "learning_rate": 0.0004640887314304119, "loss": 1.511, "step": 36180 }, { "epoch": 1.46, "learning_rate": 0.00046407866661298767, "loss": 1.419, "step": 36190 }, { "epoch": 1.46, "learning_rate": 0.0004640686017955634, "loss": 1.1607, "step": 36200 }, { "epoch": 1.46, "learning_rate": 0.0004640585369781392, "loss": 1.4263, "step": 36210 }, { "epoch": 1.46, "learning_rate": 0.000464048472160715, "loss": 1.6549, "step": 36220 }, { "epoch": 1.46, "learning_rate": 0.0004640384073432908, "loss": 1.2982, "step": 36230 }, { "epoch": 1.46, "learning_rate": 0.00046402834252586657, "loss": 1.5673, "step": 36240 }, { "epoch": 1.46, "learning_rate": 0.0004640182777084424, "loss": 1.4528, "step": 36250 }, { "epoch": 1.46, "learning_rate": 0.0004640082128910182, "loss": 1.5323, "step": 36260 }, { "epoch": 1.46, "learning_rate": 0.000463998148073594, "loss": 1.5546, "step": 36270 }, { "epoch": 1.46, "learning_rate": 0.0004639880832561698, "loss": 1.4337, "step": 36280 }, { "epoch": 1.46, "learning_rate": 0.0004639780184387455, "loss": 1.805, "step": 36290 }, { "epoch": 1.46, "learning_rate": 0.0004639679536213213, "loss": 1.3342, "step": 36300 }, { "epoch": 1.46, "learning_rate": 0.0004639578888038971, "loss": 1.3957, "step": 36310 }, { "epoch": 1.46, "learning_rate": 0.0004639478239864729, "loss": 1.1685, "step": 36320 }, { "epoch": 1.46, "learning_rate": 0.0004639377591690487, "loss": 1.5755, "step": 36330 }, { "epoch": 1.46, "learning_rate": 0.0004639276943516245, "loss": 1.4991, "step": 36340 }, { "epoch": 1.46, "learning_rate": 0.00046391762953420027, "loss": 1.6451, "step": 36350 }, { "epoch": 1.46, "learning_rate": 0.00046390756471677606, "loss": 1.6525, "step": 36360 }, { "epoch": 1.46, "learning_rate": 0.0004638974998993518, "loss": 1.292, "step": 36370 }, { "epoch": 1.46, "learning_rate": 0.0004638874350819276, "loss": 1.3728, "step": 36380 }, { "epoch": 1.46, "learning_rate": 0.0004638773702645034, "loss": 1.3769, "step": 36390 }, { "epoch": 1.46, "learning_rate": 0.0004638673054470792, "loss": 1.5499, "step": 36400 }, { "epoch": 1.46, "learning_rate": 0.000463857240629655, "loss": 1.5654, "step": 36410 }, { "epoch": 1.46, "learning_rate": 0.0004638471758122308, "loss": 1.688, "step": 36420 }, { "epoch": 1.47, "learning_rate": 0.0004638371109948066, "loss": 1.2719, "step": 36430 }, { "epoch": 1.47, "learning_rate": 0.0004638270461773824, "loss": 1.3357, "step": 36440 }, { "epoch": 1.47, "learning_rate": 0.0004638169813599581, "loss": 1.583, "step": 36450 }, { "epoch": 1.47, "learning_rate": 0.0004638069165425339, "loss": 1.7364, "step": 36460 }, { "epoch": 1.47, "learning_rate": 0.0004637968517251097, "loss": 1.3349, "step": 36470 }, { "epoch": 1.47, "learning_rate": 0.0004637867869076855, "loss": 1.2724, "step": 36480 }, { "epoch": 1.47, "learning_rate": 0.0004637767220902613, "loss": 1.3271, "step": 36490 }, { "epoch": 1.47, "learning_rate": 0.0004637666572728371, "loss": 1.3687, "step": 36500 }, { "epoch": 1.47, "learning_rate": 0.00046375659245541286, "loss": 1.4205, "step": 36510 }, { "epoch": 1.47, "learning_rate": 0.00046374652763798865, "loss": 1.5239, "step": 36520 }, { "epoch": 1.47, "learning_rate": 0.00046373646282056445, "loss": 1.4521, "step": 36530 }, { "epoch": 1.47, "learning_rate": 0.00046372639800314024, "loss": 1.5783, "step": 36540 }, { "epoch": 1.47, "learning_rate": 0.000463716333185716, "loss": 1.725, "step": 36550 }, { "epoch": 1.47, "learning_rate": 0.0004637062683682918, "loss": 1.333, "step": 36560 }, { "epoch": 1.47, "learning_rate": 0.0004636962035508676, "loss": 1.1385, "step": 36570 }, { "epoch": 1.47, "learning_rate": 0.0004636861387334434, "loss": 1.6828, "step": 36580 }, { "epoch": 1.47, "learning_rate": 0.0004636760739160192, "loss": 1.4501, "step": 36590 }, { "epoch": 1.47, "learning_rate": 0.000463666009098595, "loss": 1.494, "step": 36600 }, { "epoch": 1.47, "learning_rate": 0.00046365594428117077, "loss": 1.4151, "step": 36610 }, { "epoch": 1.47, "learning_rate": 0.0004636458794637465, "loss": 1.2878, "step": 36620 }, { "epoch": 1.47, "learning_rate": 0.0004636358146463223, "loss": 1.2669, "step": 36630 }, { "epoch": 1.47, "learning_rate": 0.0004636257498288981, "loss": 1.2984, "step": 36640 }, { "epoch": 1.47, "learning_rate": 0.0004636156850114739, "loss": 1.1532, "step": 36650 }, { "epoch": 1.47, "learning_rate": 0.00046360562019404967, "loss": 1.4026, "step": 36660 }, { "epoch": 1.47, "learning_rate": 0.00046359555537662546, "loss": 1.4297, "step": 36670 }, { "epoch": 1.48, "learning_rate": 0.00046358549055920125, "loss": 1.3253, "step": 36680 }, { "epoch": 1.48, "learning_rate": 0.0004635754257417771, "loss": 1.7859, "step": 36690 }, { "epoch": 1.48, "learning_rate": 0.0004635653609243529, "loss": 1.4161, "step": 36700 }, { "epoch": 1.48, "learning_rate": 0.0004635552961069286, "loss": 1.4715, "step": 36710 }, { "epoch": 1.48, "learning_rate": 0.0004635452312895044, "loss": 1.3222, "step": 36720 }, { "epoch": 1.48, "learning_rate": 0.0004635351664720802, "loss": 1.5298, "step": 36730 }, { "epoch": 1.48, "learning_rate": 0.000463525101654656, "loss": 1.5334, "step": 36740 }, { "epoch": 1.48, "learning_rate": 0.0004635150368372318, "loss": 1.432, "step": 36750 }, { "epoch": 1.48, "learning_rate": 0.0004635049720198076, "loss": 1.5576, "step": 36760 }, { "epoch": 1.48, "learning_rate": 0.00046349490720238337, "loss": 1.5743, "step": 36770 }, { "epoch": 1.48, "learning_rate": 0.00046348484238495916, "loss": 1.5208, "step": 36780 }, { "epoch": 1.48, "learning_rate": 0.0004634747775675349, "loss": 1.5377, "step": 36790 }, { "epoch": 1.48, "learning_rate": 0.0004634647127501107, "loss": 1.2863, "step": 36800 }, { "epoch": 1.48, "learning_rate": 0.0004634546479326865, "loss": 1.5333, "step": 36810 }, { "epoch": 1.48, "learning_rate": 0.00046344458311526227, "loss": 1.3121, "step": 36820 }, { "epoch": 1.48, "learning_rate": 0.00046343451829783806, "loss": 1.584, "step": 36830 }, { "epoch": 1.48, "learning_rate": 0.0004634244534804139, "loss": 1.5616, "step": 36840 }, { "epoch": 1.48, "learning_rate": 0.0004634143886629897, "loss": 1.2332, "step": 36850 }, { "epoch": 1.48, "learning_rate": 0.0004634043238455655, "loss": 1.444, "step": 36860 }, { "epoch": 1.48, "learning_rate": 0.0004633942590281413, "loss": 1.7226, "step": 36870 }, { "epoch": 1.48, "learning_rate": 0.000463384194210717, "loss": 1.4966, "step": 36880 }, { "epoch": 1.48, "learning_rate": 0.0004633741293932928, "loss": 1.6144, "step": 36890 }, { "epoch": 1.48, "learning_rate": 0.0004633640645758686, "loss": 1.7415, "step": 36900 }, { "epoch": 1.48, "learning_rate": 0.0004633539997584444, "loss": 1.3623, "step": 36910 }, { "epoch": 1.48, "learning_rate": 0.0004633439349410202, "loss": 1.4576, "step": 36920 }, { "epoch": 1.49, "learning_rate": 0.00046333387012359597, "loss": 1.6627, "step": 36930 }, { "epoch": 1.49, "learning_rate": 0.00046332380530617176, "loss": 1.3977, "step": 36940 }, { "epoch": 1.49, "learning_rate": 0.00046331374048874755, "loss": 1.2279, "step": 36950 }, { "epoch": 1.49, "learning_rate": 0.0004633036756713233, "loss": 1.3941, "step": 36960 }, { "epoch": 1.49, "learning_rate": 0.0004632936108538991, "loss": 1.5486, "step": 36970 }, { "epoch": 1.49, "learning_rate": 0.0004632835460364749, "loss": 1.4175, "step": 36980 }, { "epoch": 1.49, "learning_rate": 0.0004632734812190507, "loss": 1.3364, "step": 36990 }, { "epoch": 1.49, "learning_rate": 0.0004632634164016265, "loss": 1.6674, "step": 37000 }, { "epoch": 1.49, "learning_rate": 0.0004632533515842023, "loss": 1.3945, "step": 37010 }, { "epoch": 1.49, "learning_rate": 0.0004632432867667781, "loss": 1.6013, "step": 37020 }, { "epoch": 1.49, "learning_rate": 0.0004632332219493539, "loss": 1.2579, "step": 37030 }, { "epoch": 1.49, "learning_rate": 0.00046322315713192967, "loss": 1.5093, "step": 37040 }, { "epoch": 1.49, "learning_rate": 0.0004632130923145054, "loss": 1.5919, "step": 37050 }, { "epoch": 1.49, "learning_rate": 0.0004632030274970812, "loss": 1.5198, "step": 37060 }, { "epoch": 1.49, "learning_rate": 0.000463192962679657, "loss": 1.5025, "step": 37070 }, { "epoch": 1.49, "learning_rate": 0.0004631828978622328, "loss": 1.4254, "step": 37080 }, { "epoch": 1.49, "learning_rate": 0.00046317283304480857, "loss": 1.6202, "step": 37090 }, { "epoch": 1.49, "learning_rate": 0.00046316276822738436, "loss": 1.5864, "step": 37100 }, { "epoch": 1.49, "learning_rate": 0.00046315270340996015, "loss": 1.9564, "step": 37110 }, { "epoch": 1.49, "learning_rate": 0.00046314263859253594, "loss": 1.6114, "step": 37120 }, { "epoch": 1.49, "learning_rate": 0.00046313257377511173, "loss": 1.4373, "step": 37130 }, { "epoch": 1.49, "learning_rate": 0.0004631225089576875, "loss": 1.4727, "step": 37140 }, { "epoch": 1.49, "learning_rate": 0.0004631124441402633, "loss": 1.3804, "step": 37150 }, { "epoch": 1.49, "learning_rate": 0.0004631023793228391, "loss": 1.5692, "step": 37160 }, { "epoch": 1.49, "learning_rate": 0.0004630923145054149, "loss": 1.1982, "step": 37170 }, { "epoch": 1.5, "learning_rate": 0.0004630822496879907, "loss": 1.6851, "step": 37180 }, { "epoch": 1.5, "learning_rate": 0.0004630721848705665, "loss": 1.5161, "step": 37190 }, { "epoch": 1.5, "learning_rate": 0.00046306212005314227, "loss": 1.3202, "step": 37200 }, { "epoch": 1.5, "learning_rate": 0.000463052055235718, "loss": 1.4883, "step": 37210 }, { "epoch": 1.5, "learning_rate": 0.0004630419904182938, "loss": 1.1532, "step": 37220 }, { "epoch": 1.5, "learning_rate": 0.0004630319256008696, "loss": 1.4891, "step": 37230 }, { "epoch": 1.5, "learning_rate": 0.0004630218607834454, "loss": 1.4221, "step": 37240 }, { "epoch": 1.5, "learning_rate": 0.00046301179596602116, "loss": 1.2797, "step": 37250 }, { "epoch": 1.5, "learning_rate": 0.00046300173114859696, "loss": 1.6062, "step": 37260 }, { "epoch": 1.5, "learning_rate": 0.0004629916663311728, "loss": 1.6662, "step": 37270 }, { "epoch": 1.5, "learning_rate": 0.0004629816015137486, "loss": 1.6949, "step": 37280 }, { "epoch": 1.5, "learning_rate": 0.0004629715366963244, "loss": 1.6344, "step": 37290 }, { "epoch": 1.5, "learning_rate": 0.0004629614718789001, "loss": 1.3135, "step": 37300 }, { "epoch": 1.5, "learning_rate": 0.0004629514070614759, "loss": 1.3549, "step": 37310 }, { "epoch": 1.5, "learning_rate": 0.0004629413422440517, "loss": 1.5029, "step": 37320 }, { "epoch": 1.5, "learning_rate": 0.0004629312774266275, "loss": 1.4946, "step": 37330 }, { "epoch": 1.5, "learning_rate": 0.0004629212126092033, "loss": 1.3679, "step": 37340 }, { "epoch": 1.5, "learning_rate": 0.00046291114779177907, "loss": 1.4709, "step": 37350 }, { "epoch": 1.5, "learning_rate": 0.00046290108297435486, "loss": 1.5761, "step": 37360 }, { "epoch": 1.5, "learning_rate": 0.00046289101815693065, "loss": 1.4473, "step": 37370 }, { "epoch": 1.5, "learning_rate": 0.0004628809533395064, "loss": 1.5474, "step": 37380 }, { "epoch": 1.5, "learning_rate": 0.0004628708885220822, "loss": 1.4475, "step": 37390 }, { "epoch": 1.5, "learning_rate": 0.00046286082370465797, "loss": 1.4418, "step": 37400 }, { "epoch": 1.5, "learning_rate": 0.00046285075888723376, "loss": 1.2351, "step": 37410 }, { "epoch": 1.5, "learning_rate": 0.0004628406940698096, "loss": 1.1869, "step": 37420 }, { "epoch": 1.51, "learning_rate": 0.0004628306292523854, "loss": 1.4745, "step": 37430 }, { "epoch": 1.51, "learning_rate": 0.0004628205644349612, "loss": 1.512, "step": 37440 }, { "epoch": 1.51, "learning_rate": 0.000462810499617537, "loss": 1.6528, "step": 37450 }, { "epoch": 1.51, "learning_rate": 0.00046280043480011277, "loss": 1.4721, "step": 37460 }, { "epoch": 1.51, "learning_rate": 0.0004627903699826885, "loss": 1.3659, "step": 37470 }, { "epoch": 1.51, "learning_rate": 0.0004627803051652643, "loss": 1.4284, "step": 37480 }, { "epoch": 1.51, "learning_rate": 0.0004627702403478401, "loss": 1.3238, "step": 37490 }, { "epoch": 1.51, "learning_rate": 0.0004627601755304159, "loss": 1.5606, "step": 37500 }, { "epoch": 1.51, "learning_rate": 0.00046275011071299167, "loss": 1.3613, "step": 37510 }, { "epoch": 1.51, "learning_rate": 0.00046274004589556746, "loss": 1.5554, "step": 37520 }, { "epoch": 1.51, "learning_rate": 0.00046272998107814325, "loss": 1.4505, "step": 37530 }, { "epoch": 1.51, "learning_rate": 0.00046271991626071904, "loss": 1.4192, "step": 37540 }, { "epoch": 1.51, "learning_rate": 0.0004627098514432948, "loss": 1.6332, "step": 37550 }, { "epoch": 1.51, "learning_rate": 0.00046269978662587057, "loss": 1.449, "step": 37560 }, { "epoch": 1.51, "learning_rate": 0.0004626897218084464, "loss": 1.6057, "step": 37570 }, { "epoch": 1.51, "learning_rate": 0.0004626796569910222, "loss": 1.4979, "step": 37580 }, { "epoch": 1.51, "learning_rate": 0.000462669592173598, "loss": 1.579, "step": 37590 }, { "epoch": 1.51, "learning_rate": 0.0004626595273561738, "loss": 1.4465, "step": 37600 }, { "epoch": 1.51, "learning_rate": 0.0004626494625387496, "loss": 1.5504, "step": 37610 }, { "epoch": 1.51, "learning_rate": 0.00046263939772132537, "loss": 1.2437, "step": 37620 }, { "epoch": 1.51, "learning_rate": 0.00046262933290390116, "loss": 1.8515, "step": 37630 }, { "epoch": 1.51, "learning_rate": 0.0004626192680864769, "loss": 1.58, "step": 37640 }, { "epoch": 1.51, "learning_rate": 0.0004626092032690527, "loss": 1.5586, "step": 37650 }, { "epoch": 1.51, "learning_rate": 0.0004625991384516285, "loss": 1.5193, "step": 37660 }, { "epoch": 1.52, "learning_rate": 0.00046258907363420427, "loss": 1.5839, "step": 37670 }, { "epoch": 1.52, "learning_rate": 0.00046257900881678006, "loss": 1.3562, "step": 37680 }, { "epoch": 1.52, "learning_rate": 0.00046256894399935585, "loss": 1.6313, "step": 37690 }, { "epoch": 1.52, "learning_rate": 0.00046255887918193164, "loss": 1.7592, "step": 37700 }, { "epoch": 1.52, "learning_rate": 0.0004625488143645075, "loss": 1.3552, "step": 37710 }, { "epoch": 1.52, "learning_rate": 0.0004625387495470832, "loss": 1.4484, "step": 37720 }, { "epoch": 1.52, "learning_rate": 0.000462528684729659, "loss": 1.3235, "step": 37730 }, { "epoch": 1.52, "learning_rate": 0.0004625186199122348, "loss": 1.1155, "step": 37740 }, { "epoch": 1.52, "learning_rate": 0.0004625085550948106, "loss": 1.1976, "step": 37750 }, { "epoch": 1.52, "learning_rate": 0.0004624984902773864, "loss": 1.4623, "step": 37760 }, { "epoch": 1.52, "learning_rate": 0.0004624884254599622, "loss": 1.2633, "step": 37770 }, { "epoch": 1.52, "learning_rate": 0.00046247836064253797, "loss": 1.5735, "step": 37780 }, { "epoch": 1.52, "learning_rate": 0.00046246829582511376, "loss": 1.4274, "step": 37790 }, { "epoch": 1.52, "learning_rate": 0.00046245823100768955, "loss": 1.3433, "step": 37800 }, { "epoch": 1.52, "learning_rate": 0.0004624481661902653, "loss": 1.5163, "step": 37810 }, { "epoch": 1.52, "learning_rate": 0.0004624381013728411, "loss": 1.6308, "step": 37820 }, { "epoch": 1.52, "learning_rate": 0.00046242803655541687, "loss": 1.797, "step": 37830 }, { "epoch": 1.52, "learning_rate": 0.00046241797173799266, "loss": 1.4832, "step": 37840 }, { "epoch": 1.52, "learning_rate": 0.00046240790692056845, "loss": 1.2002, "step": 37850 }, { "epoch": 1.52, "learning_rate": 0.0004623978421031443, "loss": 1.2894, "step": 37860 }, { "epoch": 1.52, "learning_rate": 0.0004623877772857201, "loss": 1.6296, "step": 37870 }, { "epoch": 1.52, "learning_rate": 0.0004623777124682959, "loss": 1.4834, "step": 37880 }, { "epoch": 1.52, "learning_rate": 0.0004623676476508716, "loss": 1.4324, "step": 37890 }, { "epoch": 1.52, "learning_rate": 0.0004623575828334474, "loss": 1.3876, "step": 37900 }, { "epoch": 1.52, "learning_rate": 0.0004623475180160232, "loss": 1.4145, "step": 37910 }, { "epoch": 1.53, "learning_rate": 0.000462337453198599, "loss": 1.4469, "step": 37920 }, { "epoch": 1.53, "learning_rate": 0.0004623273883811748, "loss": 1.6423, "step": 37930 }, { "epoch": 1.53, "learning_rate": 0.00046231732356375057, "loss": 1.487, "step": 37940 }, { "epoch": 1.53, "learning_rate": 0.00046230725874632636, "loss": 1.4224, "step": 37950 }, { "epoch": 1.53, "learning_rate": 0.00046229719392890215, "loss": 1.5824, "step": 37960 }, { "epoch": 1.53, "learning_rate": 0.0004622871291114779, "loss": 1.4537, "step": 37970 }, { "epoch": 1.53, "learning_rate": 0.0004622770642940537, "loss": 1.6034, "step": 37980 }, { "epoch": 1.53, "learning_rate": 0.00046226699947662947, "loss": 1.444, "step": 37990 }, { "epoch": 1.53, "learning_rate": 0.0004622569346592053, "loss": 1.3555, "step": 38000 }, { "epoch": 1.53, "learning_rate": 0.0004622468698417811, "loss": 1.6905, "step": 38010 }, { "epoch": 1.53, "learning_rate": 0.0004622368050243569, "loss": 1.4988, "step": 38020 }, { "epoch": 1.53, "learning_rate": 0.0004622267402069327, "loss": 1.5772, "step": 38030 }, { "epoch": 1.53, "learning_rate": 0.0004622166753895085, "loss": 1.4108, "step": 38040 }, { "epoch": 1.53, "learning_rate": 0.00046220661057208426, "loss": 1.4209, "step": 38050 }, { "epoch": 1.53, "learning_rate": 0.00046219654575466, "loss": 1.4007, "step": 38060 }, { "epoch": 1.53, "learning_rate": 0.0004621864809372358, "loss": 1.577, "step": 38070 }, { "epoch": 1.53, "learning_rate": 0.0004621764161198116, "loss": 1.4924, "step": 38080 }, { "epoch": 1.53, "learning_rate": 0.0004621663513023874, "loss": 1.4418, "step": 38090 }, { "epoch": 1.53, "learning_rate": 0.00046215628648496316, "loss": 1.4412, "step": 38100 }, { "epoch": 1.53, "learning_rate": 0.00046214622166753896, "loss": 1.436, "step": 38110 }, { "epoch": 1.53, "learning_rate": 0.00046213615685011475, "loss": 1.3472, "step": 38120 }, { "epoch": 1.53, "learning_rate": 0.00046212609203269054, "loss": 1.3906, "step": 38130 }, { "epoch": 1.53, "learning_rate": 0.0004621160272152663, "loss": 1.2427, "step": 38140 }, { "epoch": 1.53, "learning_rate": 0.0004621059623978421, "loss": 1.6007, "step": 38150 }, { "epoch": 1.53, "learning_rate": 0.0004620958975804179, "loss": 1.392, "step": 38160 }, { "epoch": 1.54, "learning_rate": 0.0004620858327629937, "loss": 1.354, "step": 38170 }, { "epoch": 1.54, "learning_rate": 0.0004620757679455695, "loss": 1.4901, "step": 38180 }, { "epoch": 1.54, "learning_rate": 0.0004620657031281453, "loss": 1.2481, "step": 38190 }, { "epoch": 1.54, "learning_rate": 0.00046205563831072107, "loss": 1.3789, "step": 38200 }, { "epoch": 1.54, "learning_rate": 0.00046204557349329686, "loss": 1.6226, "step": 38210 }, { "epoch": 1.54, "learning_rate": 0.00046203550867587265, "loss": 1.532, "step": 38220 }, { "epoch": 1.54, "learning_rate": 0.0004620254438584484, "loss": 1.432, "step": 38230 }, { "epoch": 1.54, "learning_rate": 0.0004620153790410242, "loss": 1.561, "step": 38240 }, { "epoch": 1.54, "learning_rate": 0.00046200531422359997, "loss": 1.7064, "step": 38250 }, { "epoch": 1.54, "learning_rate": 0.00046199524940617576, "loss": 1.2849, "step": 38260 }, { "epoch": 1.54, "learning_rate": 0.00046198518458875155, "loss": 1.3694, "step": 38270 }, { "epoch": 1.54, "learning_rate": 0.00046197511977132734, "loss": 1.4259, "step": 38280 }, { "epoch": 1.54, "learning_rate": 0.00046196505495390314, "loss": 1.4617, "step": 38290 }, { "epoch": 1.54, "learning_rate": 0.000461954990136479, "loss": 1.3952, "step": 38300 }, { "epoch": 1.54, "learning_rate": 0.0004619449253190547, "loss": 1.5083, "step": 38310 }, { "epoch": 1.54, "learning_rate": 0.0004619348605016305, "loss": 1.2579, "step": 38320 }, { "epoch": 1.54, "learning_rate": 0.0004619247956842063, "loss": 1.3434, "step": 38330 }, { "epoch": 1.54, "learning_rate": 0.0004619147308667821, "loss": 1.4106, "step": 38340 }, { "epoch": 1.54, "learning_rate": 0.0004619046660493579, "loss": 1.3961, "step": 38350 }, { "epoch": 1.54, "learning_rate": 0.00046189460123193367, "loss": 1.4316, "step": 38360 }, { "epoch": 1.54, "learning_rate": 0.00046188453641450946, "loss": 1.3013, "step": 38370 }, { "epoch": 1.54, "learning_rate": 0.00046187447159708525, "loss": 1.4389, "step": 38380 }, { "epoch": 1.54, "learning_rate": 0.00046186440677966104, "loss": 1.3298, "step": 38390 }, { "epoch": 1.54, "learning_rate": 0.0004618543419622368, "loss": 1.5387, "step": 38400 }, { "epoch": 1.54, "learning_rate": 0.00046184427714481257, "loss": 1.5188, "step": 38410 }, { "epoch": 1.55, "learning_rate": 0.00046183421232738836, "loss": 1.5521, "step": 38420 }, { "epoch": 1.55, "learning_rate": 0.00046182414750996415, "loss": 1.4444, "step": 38430 }, { "epoch": 1.55, "learning_rate": 0.00046181408269254, "loss": 1.5849, "step": 38440 }, { "epoch": 1.55, "learning_rate": 0.0004618040178751158, "loss": 1.5979, "step": 38450 }, { "epoch": 1.55, "learning_rate": 0.0004617939530576916, "loss": 1.5963, "step": 38460 }, { "epoch": 1.55, "learning_rate": 0.00046178388824026737, "loss": 1.679, "step": 38470 }, { "epoch": 1.55, "learning_rate": 0.0004617738234228431, "loss": 1.5465, "step": 38480 }, { "epoch": 1.55, "learning_rate": 0.0004617637586054189, "loss": 1.3949, "step": 38490 }, { "epoch": 1.55, "learning_rate": 0.0004617536937879947, "loss": 1.5211, "step": 38500 }, { "epoch": 1.55, "learning_rate": 0.0004617436289705705, "loss": 1.633, "step": 38510 }, { "epoch": 1.55, "learning_rate": 0.00046173356415314627, "loss": 1.5478, "step": 38520 }, { "epoch": 1.55, "learning_rate": 0.00046172349933572206, "loss": 1.4783, "step": 38530 }, { "epoch": 1.55, "learning_rate": 0.00046171343451829785, "loss": 1.2804, "step": 38540 }, { "epoch": 1.55, "learning_rate": 0.00046170336970087364, "loss": 1.6425, "step": 38550 }, { "epoch": 1.55, "learning_rate": 0.00046169330488344943, "loss": 1.5715, "step": 38560 }, { "epoch": 1.55, "learning_rate": 0.00046168324006602517, "loss": 1.6009, "step": 38570 }, { "epoch": 1.55, "learning_rate": 0.00046167317524860096, "loss": 1.5723, "step": 38580 }, { "epoch": 1.55, "learning_rate": 0.0004616631104311768, "loss": 1.5008, "step": 38590 }, { "epoch": 1.55, "learning_rate": 0.0004616530456137526, "loss": 1.6249, "step": 38600 }, { "epoch": 1.55, "learning_rate": 0.0004616429807963284, "loss": 1.2835, "step": 38610 }, { "epoch": 1.55, "learning_rate": 0.0004616329159789042, "loss": 1.5547, "step": 38620 }, { "epoch": 1.55, "learning_rate": 0.00046162285116147997, "loss": 1.5491, "step": 38630 }, { "epoch": 1.55, "learning_rate": 0.00046161278634405576, "loss": 1.811, "step": 38640 }, { "epoch": 1.55, "learning_rate": 0.0004616027215266315, "loss": 1.3042, "step": 38650 }, { "epoch": 1.55, "learning_rate": 0.0004615926567092073, "loss": 1.5596, "step": 38660 }, { "epoch": 1.56, "learning_rate": 0.0004615825918917831, "loss": 1.5662, "step": 38670 }, { "epoch": 1.56, "learning_rate": 0.00046157252707435887, "loss": 1.3496, "step": 38680 }, { "epoch": 1.56, "learning_rate": 0.00046156246225693466, "loss": 1.3861, "step": 38690 }, { "epoch": 1.56, "learning_rate": 0.00046155239743951045, "loss": 1.4788, "step": 38700 }, { "epoch": 1.56, "learning_rate": 0.00046154233262208624, "loss": 1.585, "step": 38710 }, { "epoch": 1.56, "learning_rate": 0.00046153226780466203, "loss": 1.4971, "step": 38720 }, { "epoch": 1.56, "learning_rate": 0.00046152220298723777, "loss": 1.4125, "step": 38730 }, { "epoch": 1.56, "learning_rate": 0.0004615121381698136, "loss": 1.526, "step": 38740 }, { "epoch": 1.56, "learning_rate": 0.0004615020733523894, "loss": 1.3611, "step": 38750 }, { "epoch": 1.56, "learning_rate": 0.0004614920085349652, "loss": 1.3087, "step": 38760 }, { "epoch": 1.56, "learning_rate": 0.000461481943717541, "loss": 1.4793, "step": 38770 }, { "epoch": 1.56, "learning_rate": 0.0004614718789001168, "loss": 1.5039, "step": 38780 }, { "epoch": 1.56, "learning_rate": 0.00046146181408269257, "loss": 1.4416, "step": 38790 }, { "epoch": 1.56, "learning_rate": 0.00046145174926526836, "loss": 1.3911, "step": 38800 }, { "epoch": 1.56, "learning_rate": 0.00046144168444784415, "loss": 1.2, "step": 38810 }, { "epoch": 1.56, "learning_rate": 0.0004614316196304199, "loss": 1.5063, "step": 38820 }, { "epoch": 1.56, "learning_rate": 0.0004614215548129957, "loss": 1.5174, "step": 38830 }, { "epoch": 1.56, "learning_rate": 0.00046141148999557147, "loss": 1.4166, "step": 38840 }, { "epoch": 1.56, "learning_rate": 0.00046140142517814726, "loss": 1.4177, "step": 38850 }, { "epoch": 1.56, "learning_rate": 0.00046139136036072305, "loss": 1.5289, "step": 38860 }, { "epoch": 1.56, "learning_rate": 0.00046138129554329884, "loss": 1.7258, "step": 38870 }, { "epoch": 1.56, "learning_rate": 0.0004613712307258747, "loss": 1.4397, "step": 38880 }, { "epoch": 1.56, "learning_rate": 0.0004613611659084505, "loss": 1.5537, "step": 38890 }, { "epoch": 1.56, "learning_rate": 0.0004613511010910262, "loss": 1.5626, "step": 38900 }, { "epoch": 1.56, "learning_rate": 0.000461341036273602, "loss": 1.4307, "step": 38910 }, { "epoch": 1.57, "learning_rate": 0.0004613309714561778, "loss": 1.5132, "step": 38920 }, { "epoch": 1.57, "learning_rate": 0.0004613209066387536, "loss": 1.4203, "step": 38930 }, { "epoch": 1.57, "learning_rate": 0.0004613108418213294, "loss": 1.3646, "step": 38940 }, { "epoch": 1.57, "learning_rate": 0.00046130077700390516, "loss": 1.4172, "step": 38950 }, { "epoch": 1.57, "learning_rate": 0.00046129071218648096, "loss": 1.5854, "step": 38960 }, { "epoch": 1.57, "learning_rate": 0.00046128064736905675, "loss": 1.423, "step": 38970 }, { "epoch": 1.57, "learning_rate": 0.00046127058255163254, "loss": 1.4941, "step": 38980 }, { "epoch": 1.57, "learning_rate": 0.0004612605177342083, "loss": 1.403, "step": 38990 }, { "epoch": 1.57, "learning_rate": 0.00046125045291678406, "loss": 1.4995, "step": 39000 }, { "epoch": 1.57, "learning_rate": 0.00046124038809935985, "loss": 1.4625, "step": 39010 }, { "epoch": 1.57, "learning_rate": 0.00046123032328193565, "loss": 1.3906, "step": 39020 }, { "epoch": 1.57, "learning_rate": 0.0004612202584645115, "loss": 1.5874, "step": 39030 }, { "epoch": 1.57, "learning_rate": 0.0004612101936470873, "loss": 1.3289, "step": 39040 }, { "epoch": 1.57, "learning_rate": 0.00046120012882966307, "loss": 1.4967, "step": 39050 }, { "epoch": 1.57, "learning_rate": 0.00046119006401223886, "loss": 1.5952, "step": 39060 }, { "epoch": 1.57, "learning_rate": 0.0004611799991948146, "loss": 1.5318, "step": 39070 }, { "epoch": 1.57, "learning_rate": 0.0004611699343773904, "loss": 1.4395, "step": 39080 }, { "epoch": 1.57, "learning_rate": 0.0004611598695599662, "loss": 1.3812, "step": 39090 }, { "epoch": 1.57, "learning_rate": 0.00046114980474254197, "loss": 1.37, "step": 39100 }, { "epoch": 1.57, "learning_rate": 0.00046113973992511776, "loss": 1.4961, "step": 39110 }, { "epoch": 1.57, "learning_rate": 0.00046112967510769355, "loss": 1.6999, "step": 39120 }, { "epoch": 1.57, "learning_rate": 0.00046111961029026934, "loss": 1.4611, "step": 39130 }, { "epoch": 1.57, "learning_rate": 0.00046110954547284514, "loss": 1.2725, "step": 39140 }, { "epoch": 1.57, "learning_rate": 0.0004610994806554209, "loss": 1.6006, "step": 39150 }, { "epoch": 1.57, "learning_rate": 0.00046108941583799666, "loss": 1.7031, "step": 39160 }, { "epoch": 1.58, "learning_rate": 0.0004610793510205725, "loss": 1.324, "step": 39170 }, { "epoch": 1.58, "learning_rate": 0.0004610692862031483, "loss": 1.6586, "step": 39180 }, { "epoch": 1.58, "learning_rate": 0.0004610592213857241, "loss": 1.4541, "step": 39190 }, { "epoch": 1.58, "learning_rate": 0.0004610491565682999, "loss": 1.2846, "step": 39200 }, { "epoch": 1.58, "learning_rate": 0.00046103909175087567, "loss": 1.4965, "step": 39210 }, { "epoch": 1.58, "learning_rate": 0.00046102902693345146, "loss": 1.39, "step": 39220 }, { "epoch": 1.58, "learning_rate": 0.00046101896211602725, "loss": 1.3694, "step": 39230 }, { "epoch": 1.58, "learning_rate": 0.000461008897298603, "loss": 1.7636, "step": 39240 }, { "epoch": 1.58, "learning_rate": 0.0004609988324811788, "loss": 1.426, "step": 39250 }, { "epoch": 1.58, "learning_rate": 0.00046098876766375457, "loss": 1.3953, "step": 39260 }, { "epoch": 1.58, "learning_rate": 0.00046097870284633036, "loss": 1.3295, "step": 39270 }, { "epoch": 1.58, "learning_rate": 0.00046096863802890615, "loss": 1.289, "step": 39280 }, { "epoch": 1.58, "learning_rate": 0.00046095857321148194, "loss": 1.5146, "step": 39290 }, { "epoch": 1.58, "learning_rate": 0.00046094850839405773, "loss": 1.4584, "step": 39300 }, { "epoch": 1.58, "learning_rate": 0.0004609384435766335, "loss": 1.2572, "step": 39310 }, { "epoch": 1.58, "learning_rate": 0.00046092837875920937, "loss": 1.2595, "step": 39320 }, { "epoch": 1.58, "learning_rate": 0.0004609183139417851, "loss": 1.7078, "step": 39330 }, { "epoch": 1.58, "learning_rate": 0.0004609082491243609, "loss": 1.4986, "step": 39340 }, { "epoch": 1.58, "learning_rate": 0.0004608981843069367, "loss": 1.4098, "step": 39350 }, { "epoch": 1.58, "learning_rate": 0.0004608881194895125, "loss": 1.458, "step": 39360 }, { "epoch": 1.58, "learning_rate": 0.00046087805467208827, "loss": 1.4227, "step": 39370 }, { "epoch": 1.58, "learning_rate": 0.00046086798985466406, "loss": 1.4918, "step": 39380 }, { "epoch": 1.58, "learning_rate": 0.00046085792503723985, "loss": 1.554, "step": 39390 }, { "epoch": 1.58, "learning_rate": 0.00046084786021981564, "loss": 1.1701, "step": 39400 }, { "epoch": 1.59, "learning_rate": 0.0004608377954023914, "loss": 1.4631, "step": 39410 }, { "epoch": 1.59, "learning_rate": 0.00046082773058496717, "loss": 1.6675, "step": 39420 }, { "epoch": 1.59, "learning_rate": 0.00046081766576754296, "loss": 1.7616, "step": 39430 }, { "epoch": 1.59, "learning_rate": 0.00046080760095011875, "loss": 1.5018, "step": 39440 }, { "epoch": 1.59, "learning_rate": 0.00046079753613269454, "loss": 1.5506, "step": 39450 }, { "epoch": 1.59, "learning_rate": 0.00046078747131527033, "loss": 1.6332, "step": 39460 }, { "epoch": 1.59, "learning_rate": 0.0004607774064978462, "loss": 1.6529, "step": 39470 }, { "epoch": 1.59, "learning_rate": 0.00046076734168042197, "loss": 1.2659, "step": 39480 }, { "epoch": 1.59, "learning_rate": 0.00046075727686299776, "loss": 1.6944, "step": 39490 }, { "epoch": 1.59, "learning_rate": 0.0004607472120455735, "loss": 1.5443, "step": 39500 }, { "epoch": 1.59, "learning_rate": 0.0004607371472281493, "loss": 1.6269, "step": 39510 }, { "epoch": 1.59, "learning_rate": 0.0004607270824107251, "loss": 1.3403, "step": 39520 }, { "epoch": 1.59, "learning_rate": 0.00046071701759330087, "loss": 1.3808, "step": 39530 }, { "epoch": 1.59, "learning_rate": 0.00046070695277587666, "loss": 1.5072, "step": 39540 }, { "epoch": 1.59, "learning_rate": 0.00046069688795845245, "loss": 1.7299, "step": 39550 }, { "epoch": 1.59, "learning_rate": 0.00046068682314102824, "loss": 1.6339, "step": 39560 }, { "epoch": 1.59, "learning_rate": 0.00046067675832360403, "loss": 1.4146, "step": 39570 }, { "epoch": 1.59, "learning_rate": 0.00046066669350617977, "loss": 1.5462, "step": 39580 }, { "epoch": 1.59, "learning_rate": 0.00046065662868875556, "loss": 1.2715, "step": 39590 }, { "epoch": 1.59, "learning_rate": 0.00046064656387133135, "loss": 1.496, "step": 39600 }, { "epoch": 1.59, "learning_rate": 0.0004606364990539072, "loss": 1.6645, "step": 39610 }, { "epoch": 1.59, "learning_rate": 0.000460626434236483, "loss": 1.1418, "step": 39620 }, { "epoch": 1.59, "learning_rate": 0.0004606163694190588, "loss": 1.4844, "step": 39630 }, { "epoch": 1.59, "learning_rate": 0.00046060630460163457, "loss": 1.4169, "step": 39640 }, { "epoch": 1.59, "learning_rate": 0.00046059623978421036, "loss": 1.3424, "step": 39650 }, { "epoch": 1.6, "learning_rate": 0.0004605861749667861, "loss": 1.4005, "step": 39660 }, { "epoch": 1.6, "learning_rate": 0.0004605761101493619, "loss": 1.4422, "step": 39670 }, { "epoch": 1.6, "learning_rate": 0.0004605660453319377, "loss": 1.481, "step": 39680 }, { "epoch": 1.6, "learning_rate": 0.00046055598051451347, "loss": 1.5575, "step": 39690 }, { "epoch": 1.6, "learning_rate": 0.00046054591569708926, "loss": 1.2836, "step": 39700 }, { "epoch": 1.6, "learning_rate": 0.00046053585087966505, "loss": 1.6127, "step": 39710 }, { "epoch": 1.6, "learning_rate": 0.00046052578606224084, "loss": 1.1913, "step": 39720 }, { "epoch": 1.6, "learning_rate": 0.00046051572124481663, "loss": 1.2969, "step": 39730 }, { "epoch": 1.6, "learning_rate": 0.0004605056564273924, "loss": 1.1796, "step": 39740 }, { "epoch": 1.6, "learning_rate": 0.00046049559160996816, "loss": 1.7808, "step": 39750 }, { "epoch": 1.6, "learning_rate": 0.000460485526792544, "loss": 1.5169, "step": 39760 }, { "epoch": 1.6, "learning_rate": 0.0004604754619751198, "loss": 1.3202, "step": 39770 }, { "epoch": 1.6, "learning_rate": 0.0004604653971576956, "loss": 1.5114, "step": 39780 }, { "epoch": 1.6, "learning_rate": 0.0004604553323402714, "loss": 1.5249, "step": 39790 }, { "epoch": 1.6, "learning_rate": 0.00046044526752284716, "loss": 1.7391, "step": 39800 }, { "epoch": 1.6, "learning_rate": 0.00046043520270542295, "loss": 1.2518, "step": 39810 }, { "epoch": 1.6, "learning_rate": 0.00046042513788799875, "loss": 1.3982, "step": 39820 }, { "epoch": 1.6, "learning_rate": 0.0004604150730705745, "loss": 1.4308, "step": 39830 }, { "epoch": 1.6, "learning_rate": 0.0004604050082531503, "loss": 1.3947, "step": 39840 }, { "epoch": 1.6, "learning_rate": 0.00046039494343572606, "loss": 1.4263, "step": 39850 }, { "epoch": 1.6, "learning_rate": 0.00046038487861830185, "loss": 1.4216, "step": 39860 }, { "epoch": 1.6, "learning_rate": 0.00046037481380087765, "loss": 1.5126, "step": 39870 }, { "epoch": 1.6, "learning_rate": 0.00046036474898345344, "loss": 1.1955, "step": 39880 }, { "epoch": 1.6, "learning_rate": 0.0004603546841660292, "loss": 1.3981, "step": 39890 }, { "epoch": 1.6, "learning_rate": 0.000460344619348605, "loss": 1.4281, "step": 39900 }, { "epoch": 1.61, "learning_rate": 0.00046033455453118086, "loss": 1.5438, "step": 39910 }, { "epoch": 1.61, "learning_rate": 0.0004603244897137566, "loss": 1.4784, "step": 39920 }, { "epoch": 1.61, "learning_rate": 0.0004603144248963324, "loss": 1.2957, "step": 39930 }, { "epoch": 1.61, "learning_rate": 0.0004603043600789082, "loss": 1.1858, "step": 39940 }, { "epoch": 1.61, "learning_rate": 0.00046029429526148397, "loss": 1.4853, "step": 39950 }, { "epoch": 1.61, "learning_rate": 0.00046028423044405976, "loss": 1.6715, "step": 39960 }, { "epoch": 1.61, "learning_rate": 0.00046027416562663555, "loss": 1.4694, "step": 39970 }, { "epoch": 1.61, "learning_rate": 0.00046026410080921134, "loss": 1.4714, "step": 39980 }, { "epoch": 1.61, "learning_rate": 0.00046025403599178713, "loss": 1.6751, "step": 39990 }, { "epoch": 1.61, "learning_rate": 0.00046024397117436287, "loss": 1.5144, "step": 40000 }, { "epoch": 1.61, "learning_rate": 0.00046023390635693866, "loss": 1.6444, "step": 40010 }, { "epoch": 1.61, "learning_rate": 0.00046022384153951445, "loss": 1.6046, "step": 40020 }, { "epoch": 1.61, "learning_rate": 0.00046021377672209024, "loss": 1.4346, "step": 40030 }, { "epoch": 1.61, "learning_rate": 0.00046020371190466603, "loss": 1.6259, "step": 40040 }, { "epoch": 1.61, "learning_rate": 0.0004601936470872419, "loss": 1.6169, "step": 40050 }, { "epoch": 1.61, "learning_rate": 0.00046018358226981767, "loss": 1.3557, "step": 40060 }, { "epoch": 1.61, "learning_rate": 0.00046017351745239346, "loss": 1.2489, "step": 40070 }, { "epoch": 1.61, "learning_rate": 0.00046016345263496925, "loss": 1.2926, "step": 40080 }, { "epoch": 1.61, "learning_rate": 0.000460153387817545, "loss": 1.4623, "step": 40090 }, { "epoch": 1.61, "learning_rate": 0.0004601433230001208, "loss": 1.1965, "step": 40100 }, { "epoch": 1.61, "learning_rate": 0.00046013325818269657, "loss": 1.5744, "step": 40110 }, { "epoch": 1.61, "learning_rate": 0.00046012319336527236, "loss": 1.5729, "step": 40120 }, { "epoch": 1.61, "learning_rate": 0.00046011312854784815, "loss": 1.587, "step": 40130 }, { "epoch": 1.61, "learning_rate": 0.00046010306373042394, "loss": 1.4962, "step": 40140 }, { "epoch": 1.61, "learning_rate": 0.00046009299891299973, "loss": 1.5996, "step": 40150 }, { "epoch": 1.62, "learning_rate": 0.0004600829340955755, "loss": 1.4841, "step": 40160 }, { "epoch": 1.62, "learning_rate": 0.00046007286927815126, "loss": 1.4569, "step": 40170 }, { "epoch": 1.62, "learning_rate": 0.00046006280446072705, "loss": 1.4963, "step": 40180 }, { "epoch": 1.62, "learning_rate": 0.00046005273964330284, "loss": 1.5542, "step": 40190 }, { "epoch": 1.62, "learning_rate": 0.0004600426748258787, "loss": 1.5038, "step": 40200 }, { "epoch": 1.62, "learning_rate": 0.0004600326100084545, "loss": 1.7104, "step": 40210 }, { "epoch": 1.62, "learning_rate": 0.00046002254519103027, "loss": 1.3376, "step": 40220 }, { "epoch": 1.62, "learning_rate": 0.00046001248037360606, "loss": 1.4676, "step": 40230 }, { "epoch": 1.62, "learning_rate": 0.00046000241555618185, "loss": 1.4998, "step": 40240 }, { "epoch": 1.62, "learning_rate": 0.00045999235073875764, "loss": 1.7289, "step": 40250 }, { "epoch": 1.62, "learning_rate": 0.0004599822859213334, "loss": 1.3458, "step": 40260 }, { "epoch": 1.62, "learning_rate": 0.00045997222110390917, "loss": 1.404, "step": 40270 }, { "epoch": 1.62, "learning_rate": 0.00045996215628648496, "loss": 1.5989, "step": 40280 }, { "epoch": 1.62, "learning_rate": 0.00045995209146906075, "loss": 1.6319, "step": 40290 }, { "epoch": 1.62, "learning_rate": 0.00045994202665163654, "loss": 1.6473, "step": 40300 }, { "epoch": 1.62, "learning_rate": 0.00045993196183421233, "loss": 1.3603, "step": 40310 }, { "epoch": 1.62, "learning_rate": 0.0004599218970167881, "loss": 1.4758, "step": 40320 }, { "epoch": 1.62, "learning_rate": 0.0004599118321993639, "loss": 1.5747, "step": 40330 }, { "epoch": 1.62, "learning_rate": 0.0004599017673819397, "loss": 1.5106, "step": 40340 }, { "epoch": 1.62, "learning_rate": 0.0004598917025645155, "loss": 1.6277, "step": 40350 }, { "epoch": 1.62, "learning_rate": 0.0004598816377470913, "loss": 1.4202, "step": 40360 }, { "epoch": 1.62, "learning_rate": 0.0004598715729296671, "loss": 1.3326, "step": 40370 }, { "epoch": 1.62, "learning_rate": 0.00045986150811224287, "loss": 1.2436, "step": 40380 }, { "epoch": 1.62, "learning_rate": 0.00045985144329481866, "loss": 1.4719, "step": 40390 }, { "epoch": 1.62, "learning_rate": 0.00045984137847739445, "loss": 1.6727, "step": 40400 }, { "epoch": 1.63, "learning_rate": 0.00045983131365997024, "loss": 1.3532, "step": 40410 }, { "epoch": 1.63, "learning_rate": 0.000459821248842546, "loss": 1.4004, "step": 40420 }, { "epoch": 1.63, "learning_rate": 0.00045981118402512177, "loss": 1.3322, "step": 40430 }, { "epoch": 1.63, "learning_rate": 0.00045980111920769756, "loss": 1.642, "step": 40440 }, { "epoch": 1.63, "learning_rate": 0.00045979105439027335, "loss": 1.4442, "step": 40450 }, { "epoch": 1.63, "learning_rate": 0.00045978098957284914, "loss": 1.6052, "step": 40460 }, { "epoch": 1.63, "learning_rate": 0.00045977092475542493, "loss": 1.6557, "step": 40470 }, { "epoch": 1.63, "learning_rate": 0.0004597608599380007, "loss": 1.5262, "step": 40480 }, { "epoch": 1.63, "learning_rate": 0.00045975079512057657, "loss": 1.2683, "step": 40490 }, { "epoch": 1.63, "learning_rate": 0.00045974073030315236, "loss": 1.1603, "step": 40500 }, { "epoch": 1.63, "learning_rate": 0.0004597306654857281, "loss": 1.5193, "step": 40510 }, { "epoch": 1.63, "learning_rate": 0.0004597206006683039, "loss": 1.5421, "step": 40520 }, { "epoch": 1.63, "learning_rate": 0.0004597105358508797, "loss": 1.5818, "step": 40530 }, { "epoch": 1.63, "learning_rate": 0.00045970047103345547, "loss": 1.5843, "step": 40540 }, { "epoch": 1.63, "learning_rate": 0.00045969040621603126, "loss": 1.5908, "step": 40550 }, { "epoch": 1.63, "learning_rate": 0.00045968034139860705, "loss": 1.5739, "step": 40560 }, { "epoch": 1.63, "learning_rate": 0.00045967027658118284, "loss": 1.5113, "step": 40570 }, { "epoch": 1.63, "learning_rate": 0.00045966021176375863, "loss": 1.4769, "step": 40580 }, { "epoch": 1.63, "learning_rate": 0.00045965014694633436, "loss": 1.547, "step": 40590 }, { "epoch": 1.63, "learning_rate": 0.00045964008212891016, "loss": 1.5172, "step": 40600 }, { "epoch": 1.63, "learning_rate": 0.00045963001731148595, "loss": 1.4768, "step": 40610 }, { "epoch": 1.63, "learning_rate": 0.00045961995249406174, "loss": 1.3153, "step": 40620 }, { "epoch": 1.63, "learning_rate": 0.00045960988767663753, "loss": 1.2945, "step": 40630 }, { "epoch": 1.63, "learning_rate": 0.0004595998228592134, "loss": 1.3946, "step": 40640 }, { "epoch": 1.63, "learning_rate": 0.00045958975804178916, "loss": 1.684, "step": 40650 }, { "epoch": 1.64, "learning_rate": 0.00045957969322436495, "loss": 1.6089, "step": 40660 }, { "epoch": 1.64, "learning_rate": 0.00045956962840694075, "loss": 1.3624, "step": 40670 }, { "epoch": 1.64, "learning_rate": 0.0004595595635895165, "loss": 1.3695, "step": 40680 }, { "epoch": 1.64, "learning_rate": 0.00045954949877209227, "loss": 1.5175, "step": 40690 }, { "epoch": 1.64, "learning_rate": 0.00045953943395466806, "loss": 1.6141, "step": 40700 }, { "epoch": 1.64, "learning_rate": 0.00045952936913724385, "loss": 1.2591, "step": 40710 }, { "epoch": 1.64, "learning_rate": 0.00045951930431981965, "loss": 1.4664, "step": 40720 }, { "epoch": 1.64, "learning_rate": 0.00045950923950239544, "loss": 1.4664, "step": 40730 }, { "epoch": 1.64, "learning_rate": 0.0004594991746849712, "loss": 1.5204, "step": 40740 }, { "epoch": 1.64, "learning_rate": 0.000459489109867547, "loss": 1.4817, "step": 40750 }, { "epoch": 1.64, "learning_rate": 0.00045947904505012275, "loss": 1.4977, "step": 40760 }, { "epoch": 1.64, "learning_rate": 0.00045946898023269854, "loss": 1.2201, "step": 40770 }, { "epoch": 1.64, "learning_rate": 0.0004594589154152744, "loss": 1.2481, "step": 40780 }, { "epoch": 1.64, "learning_rate": 0.0004594488505978502, "loss": 1.3494, "step": 40790 }, { "epoch": 1.64, "learning_rate": 0.00045943878578042597, "loss": 1.6208, "step": 40800 }, { "epoch": 1.64, "learning_rate": 0.00045942872096300176, "loss": 1.5573, "step": 40810 }, { "epoch": 1.64, "learning_rate": 0.00045941865614557755, "loss": 1.41, "step": 40820 }, { "epoch": 1.64, "learning_rate": 0.00045940859132815334, "loss": 1.6072, "step": 40830 }, { "epoch": 1.64, "learning_rate": 0.00045939852651072913, "loss": 1.3921, "step": 40840 }, { "epoch": 1.64, "learning_rate": 0.00045938846169330487, "loss": 1.3344, "step": 40850 }, { "epoch": 1.64, "learning_rate": 0.00045937839687588066, "loss": 1.5781, "step": 40860 }, { "epoch": 1.64, "learning_rate": 0.00045936833205845645, "loss": 1.1728, "step": 40870 }, { "epoch": 1.64, "learning_rate": 0.00045935826724103224, "loss": 1.4861, "step": 40880 }, { "epoch": 1.64, "learning_rate": 0.00045934820242360803, "loss": 1.3896, "step": 40890 }, { "epoch": 1.64, "learning_rate": 0.0004593381376061838, "loss": 1.3422, "step": 40900 }, { "epoch": 1.65, "learning_rate": 0.0004593280727887596, "loss": 1.2358, "step": 40910 }, { "epoch": 1.65, "learning_rate": 0.0004593180079713354, "loss": 1.5695, "step": 40920 }, { "epoch": 1.65, "learning_rate": 0.0004593079431539112, "loss": 1.2918, "step": 40930 }, { "epoch": 1.65, "learning_rate": 0.000459297878336487, "loss": 1.2096, "step": 40940 }, { "epoch": 1.65, "learning_rate": 0.0004592878135190628, "loss": 1.419, "step": 40950 }, { "epoch": 1.65, "learning_rate": 0.00045927774870163857, "loss": 1.3147, "step": 40960 }, { "epoch": 1.65, "learning_rate": 0.00045926768388421436, "loss": 1.3128, "step": 40970 }, { "epoch": 1.65, "learning_rate": 0.00045925761906679015, "loss": 1.3932, "step": 40980 }, { "epoch": 1.65, "learning_rate": 0.00045924755424936594, "loss": 1.6871, "step": 40990 }, { "epoch": 1.65, "learning_rate": 0.00045923748943194173, "loss": 1.523, "step": 41000 }, { "epoch": 1.65, "learning_rate": 0.0004592274246145175, "loss": 1.5555, "step": 41010 }, { "epoch": 1.65, "learning_rate": 0.00045921735979709326, "loss": 1.4156, "step": 41020 }, { "epoch": 1.65, "learning_rate": 0.00045920729497966905, "loss": 1.2901, "step": 41030 }, { "epoch": 1.65, "learning_rate": 0.00045919723016224484, "loss": 1.4381, "step": 41040 }, { "epoch": 1.65, "learning_rate": 0.00045918716534482063, "loss": 1.4286, "step": 41050 }, { "epoch": 1.65, "learning_rate": 0.0004591771005273964, "loss": 1.194, "step": 41060 }, { "epoch": 1.65, "learning_rate": 0.00045916703570997227, "loss": 1.7646, "step": 41070 }, { "epoch": 1.65, "learning_rate": 0.00045915697089254806, "loss": 1.3692, "step": 41080 }, { "epoch": 1.65, "learning_rate": 0.00045914690607512385, "loss": 1.4973, "step": 41090 }, { "epoch": 1.65, "learning_rate": 0.0004591368412576996, "loss": 1.4712, "step": 41100 }, { "epoch": 1.65, "learning_rate": 0.0004591267764402754, "loss": 1.5482, "step": 41110 }, { "epoch": 1.65, "learning_rate": 0.00045911671162285117, "loss": 1.5263, "step": 41120 }, { "epoch": 1.65, "learning_rate": 0.00045910664680542696, "loss": 1.3912, "step": 41130 }, { "epoch": 1.65, "learning_rate": 0.00045909658198800275, "loss": 1.6768, "step": 41140 }, { "epoch": 1.65, "learning_rate": 0.00045908651717057854, "loss": 1.4003, "step": 41150 }, { "epoch": 1.66, "learning_rate": 0.00045907645235315433, "loss": 1.3555, "step": 41160 }, { "epoch": 1.66, "learning_rate": 0.0004590663875357301, "loss": 1.1579, "step": 41170 }, { "epoch": 1.66, "learning_rate": 0.00045905632271830586, "loss": 1.4157, "step": 41180 }, { "epoch": 1.66, "learning_rate": 0.00045904625790088165, "loss": 1.577, "step": 41190 }, { "epoch": 1.66, "learning_rate": 0.00045903619308345744, "loss": 1.3207, "step": 41200 }, { "epoch": 1.66, "learning_rate": 0.00045902612826603323, "loss": 1.5637, "step": 41210 }, { "epoch": 1.66, "learning_rate": 0.0004590160634486091, "loss": 1.144, "step": 41220 }, { "epoch": 1.66, "learning_rate": 0.00045900599863118487, "loss": 1.4702, "step": 41230 }, { "epoch": 1.66, "learning_rate": 0.00045899593381376066, "loss": 1.6863, "step": 41240 }, { "epoch": 1.66, "learning_rate": 0.00045898586899633645, "loss": 1.5707, "step": 41250 }, { "epoch": 1.66, "learning_rate": 0.00045897580417891224, "loss": 1.4974, "step": 41260 }, { "epoch": 1.66, "learning_rate": 0.000458965739361488, "loss": 1.591, "step": 41270 }, { "epoch": 1.66, "learning_rate": 0.00045895567454406377, "loss": 1.659, "step": 41280 }, { "epoch": 1.66, "learning_rate": 0.00045894560972663956, "loss": 1.493, "step": 41290 }, { "epoch": 1.66, "learning_rate": 0.00045893554490921535, "loss": 1.3641, "step": 41300 }, { "epoch": 1.66, "learning_rate": 0.00045892548009179114, "loss": 1.6029, "step": 41310 }, { "epoch": 1.66, "learning_rate": 0.00045891541527436693, "loss": 1.5027, "step": 41320 }, { "epoch": 1.66, "learning_rate": 0.0004589053504569427, "loss": 1.3901, "step": 41330 }, { "epoch": 1.66, "learning_rate": 0.0004588952856395185, "loss": 1.6504, "step": 41340 }, { "epoch": 1.66, "learning_rate": 0.00045888522082209425, "loss": 1.2593, "step": 41350 }, { "epoch": 1.66, "learning_rate": 0.00045887515600467004, "loss": 1.3614, "step": 41360 }, { "epoch": 1.66, "learning_rate": 0.0004588650911872459, "loss": 1.3223, "step": 41370 }, { "epoch": 1.66, "learning_rate": 0.0004588550263698217, "loss": 1.5313, "step": 41380 }, { "epoch": 1.66, "learning_rate": 0.00045884496155239746, "loss": 1.4883, "step": 41390 }, { "epoch": 1.67, "learning_rate": 0.00045883489673497326, "loss": 1.3346, "step": 41400 }, { "epoch": 1.67, "learning_rate": 0.00045882483191754905, "loss": 1.4145, "step": 41410 }, { "epoch": 1.67, "learning_rate": 0.00045881476710012484, "loss": 1.4058, "step": 41420 }, { "epoch": 1.67, "learning_rate": 0.00045880470228270063, "loss": 1.4022, "step": 41430 }, { "epoch": 1.67, "learning_rate": 0.00045879463746527636, "loss": 1.206, "step": 41440 }, { "epoch": 1.67, "learning_rate": 0.00045878457264785216, "loss": 1.3046, "step": 41450 }, { "epoch": 1.67, "learning_rate": 0.00045877450783042795, "loss": 1.2218, "step": 41460 }, { "epoch": 1.67, "learning_rate": 0.00045876444301300374, "loss": 1.5, "step": 41470 }, { "epoch": 1.67, "learning_rate": 0.00045875437819557953, "loss": 1.638, "step": 41480 }, { "epoch": 1.67, "learning_rate": 0.0004587443133781553, "loss": 1.4743, "step": 41490 }, { "epoch": 1.67, "learning_rate": 0.0004587342485607311, "loss": 1.3447, "step": 41500 }, { "epoch": 1.67, "learning_rate": 0.00045872418374330695, "loss": 1.3473, "step": 41510 }, { "epoch": 1.67, "learning_rate": 0.0004587141189258827, "loss": 1.4942, "step": 41520 }, { "epoch": 1.67, "learning_rate": 0.0004587040541084585, "loss": 1.37, "step": 41530 }, { "epoch": 1.67, "learning_rate": 0.00045869398929103427, "loss": 1.4073, "step": 41540 }, { "epoch": 1.67, "learning_rate": 0.00045868392447361006, "loss": 1.5633, "step": 41550 }, { "epoch": 1.67, "learning_rate": 0.00045867385965618585, "loss": 1.505, "step": 41560 }, { "epoch": 1.67, "learning_rate": 0.00045866379483876164, "loss": 1.5637, "step": 41570 }, { "epoch": 1.67, "learning_rate": 0.00045865373002133744, "loss": 1.4828, "step": 41580 }, { "epoch": 1.67, "learning_rate": 0.0004586436652039132, "loss": 1.6504, "step": 41590 }, { "epoch": 1.67, "learning_rate": 0.000458633600386489, "loss": 1.252, "step": 41600 }, { "epoch": 1.67, "learning_rate": 0.00045862353556906475, "loss": 1.3278, "step": 41610 }, { "epoch": 1.67, "learning_rate": 0.00045861347075164054, "loss": 1.351, "step": 41620 }, { "epoch": 1.67, "learning_rate": 0.00045860340593421634, "loss": 1.5329, "step": 41630 }, { "epoch": 1.67, "learning_rate": 0.0004585933411167921, "loss": 1.3842, "step": 41640 }, { "epoch": 1.68, "learning_rate": 0.0004585832762993679, "loss": 1.5488, "step": 41650 }, { "epoch": 1.68, "learning_rate": 0.00045857321148194376, "loss": 1.7002, "step": 41660 }, { "epoch": 1.68, "learning_rate": 0.00045856314666451955, "loss": 1.4547, "step": 41670 }, { "epoch": 1.68, "learning_rate": 0.00045855308184709534, "loss": 1.2813, "step": 41680 }, { "epoch": 1.68, "learning_rate": 0.0004585430170296711, "loss": 1.1628, "step": 41690 }, { "epoch": 1.68, "learning_rate": 0.00045853295221224687, "loss": 1.3888, "step": 41700 }, { "epoch": 1.68, "learning_rate": 0.00045852288739482266, "loss": 1.6734, "step": 41710 }, { "epoch": 1.68, "learning_rate": 0.00045851282257739845, "loss": 1.3153, "step": 41720 }, { "epoch": 1.68, "learning_rate": 0.00045850275775997424, "loss": 1.4497, "step": 41730 }, { "epoch": 1.68, "learning_rate": 0.00045849269294255003, "loss": 1.674, "step": 41740 }, { "epoch": 1.68, "learning_rate": 0.0004584826281251258, "loss": 1.2876, "step": 41750 }, { "epoch": 1.68, "learning_rate": 0.0004584725633077016, "loss": 1.3444, "step": 41760 }, { "epoch": 1.68, "learning_rate": 0.0004584624984902774, "loss": 1.7097, "step": 41770 }, { "epoch": 1.68, "learning_rate": 0.00045845243367285314, "loss": 1.412, "step": 41780 }, { "epoch": 1.68, "learning_rate": 0.00045844236885542893, "loss": 1.084, "step": 41790 }, { "epoch": 1.68, "learning_rate": 0.0004584323040380047, "loss": 1.3943, "step": 41800 }, { "epoch": 1.68, "learning_rate": 0.00045842223922058057, "loss": 1.5517, "step": 41810 }, { "epoch": 1.68, "learning_rate": 0.00045841217440315636, "loss": 1.6665, "step": 41820 }, { "epoch": 1.68, "learning_rate": 0.00045840210958573215, "loss": 1.2057, "step": 41830 }, { "epoch": 1.68, "learning_rate": 0.00045839204476830794, "loss": 1.3782, "step": 41840 }, { "epoch": 1.68, "learning_rate": 0.00045838197995088373, "loss": 1.5744, "step": 41850 }, { "epoch": 1.68, "learning_rate": 0.00045837191513345947, "loss": 1.3696, "step": 41860 }, { "epoch": 1.68, "learning_rate": 0.00045836185031603526, "loss": 1.5293, "step": 41870 }, { "epoch": 1.68, "learning_rate": 0.00045835178549861105, "loss": 1.4898, "step": 41880 }, { "epoch": 1.68, "learning_rate": 0.00045834172068118684, "loss": 1.6937, "step": 41890 }, { "epoch": 1.69, "learning_rate": 0.00045833165586376263, "loss": 1.4181, "step": 41900 }, { "epoch": 1.69, "learning_rate": 0.0004583215910463384, "loss": 1.3647, "step": 41910 }, { "epoch": 1.69, "learning_rate": 0.0004583115262289142, "loss": 1.3799, "step": 41920 }, { "epoch": 1.69, "learning_rate": 0.00045830146141149, "loss": 1.5918, "step": 41930 }, { "epoch": 1.69, "learning_rate": 0.00045829139659406574, "loss": 1.3965, "step": 41940 }, { "epoch": 1.69, "learning_rate": 0.0004582813317766416, "loss": 1.3343, "step": 41950 }, { "epoch": 1.69, "learning_rate": 0.0004582712669592174, "loss": 1.6772, "step": 41960 }, { "epoch": 1.69, "learning_rate": 0.00045826120214179317, "loss": 1.2516, "step": 41970 }, { "epoch": 1.69, "learning_rate": 0.00045825113732436896, "loss": 1.3189, "step": 41980 }, { "epoch": 1.69, "learning_rate": 0.00045824107250694475, "loss": 1.3387, "step": 41990 }, { "epoch": 1.69, "learning_rate": 0.00045823100768952054, "loss": 1.6179, "step": 42000 }, { "epoch": 1.69, "learning_rate": 0.00045822094287209633, "loss": 1.1511, "step": 42010 }, { "epoch": 1.69, "learning_rate": 0.0004582108780546721, "loss": 1.3812, "step": 42020 }, { "epoch": 1.69, "learning_rate": 0.00045820081323724786, "loss": 1.5789, "step": 42030 }, { "epoch": 1.69, "learning_rate": 0.00045819074841982365, "loss": 1.4455, "step": 42040 }, { "epoch": 1.69, "learning_rate": 0.00045818068360239944, "loss": 1.4097, "step": 42050 }, { "epoch": 1.69, "learning_rate": 0.00045817061878497523, "loss": 1.2144, "step": 42060 }, { "epoch": 1.69, "learning_rate": 0.000458160553967551, "loss": 1.4488, "step": 42070 }, { "epoch": 1.69, "learning_rate": 0.0004581504891501268, "loss": 1.5339, "step": 42080 }, { "epoch": 1.69, "learning_rate": 0.0004581404243327026, "loss": 1.6984, "step": 42090 }, { "epoch": 1.69, "learning_rate": 0.00045813035951527845, "loss": 1.4471, "step": 42100 }, { "epoch": 1.69, "learning_rate": 0.00045812029469785424, "loss": 1.6165, "step": 42110 }, { "epoch": 1.69, "learning_rate": 0.00045811022988043, "loss": 1.2739, "step": 42120 }, { "epoch": 1.69, "learning_rate": 0.00045810016506300577, "loss": 1.3737, "step": 42130 }, { "epoch": 1.69, "learning_rate": 0.00045809010024558156, "loss": 1.5886, "step": 42140 }, { "epoch": 1.7, "learning_rate": 0.00045808003542815735, "loss": 1.5781, "step": 42150 }, { "epoch": 1.7, "learning_rate": 0.00045806997061073314, "loss": 1.5691, "step": 42160 }, { "epoch": 1.7, "learning_rate": 0.00045805990579330893, "loss": 1.429, "step": 42170 }, { "epoch": 1.7, "learning_rate": 0.0004580498409758847, "loss": 1.3379, "step": 42180 }, { "epoch": 1.7, "learning_rate": 0.0004580397761584605, "loss": 1.1566, "step": 42190 }, { "epoch": 1.7, "learning_rate": 0.00045802971134103625, "loss": 1.5287, "step": 42200 }, { "epoch": 1.7, "learning_rate": 0.00045801964652361204, "loss": 1.4307, "step": 42210 }, { "epoch": 1.7, "learning_rate": 0.00045800958170618783, "loss": 1.4152, "step": 42220 }, { "epoch": 1.7, "learning_rate": 0.0004579995168887636, "loss": 1.3688, "step": 42230 }, { "epoch": 1.7, "learning_rate": 0.00045798945207133946, "loss": 1.4128, "step": 42240 }, { "epoch": 1.7, "learning_rate": 0.00045797938725391526, "loss": 1.4766, "step": 42250 }, { "epoch": 1.7, "learning_rate": 0.00045796932243649105, "loss": 1.5441, "step": 42260 }, { "epoch": 1.7, "learning_rate": 0.00045795925761906684, "loss": 1.4269, "step": 42270 }, { "epoch": 1.7, "learning_rate": 0.0004579491928016426, "loss": 1.5015, "step": 42280 }, { "epoch": 1.7, "learning_rate": 0.00045793912798421836, "loss": 1.1886, "step": 42290 }, { "epoch": 1.7, "learning_rate": 0.00045792906316679416, "loss": 1.3366, "step": 42300 }, { "epoch": 1.7, "learning_rate": 0.00045791899834936995, "loss": 1.4426, "step": 42310 }, { "epoch": 1.7, "learning_rate": 0.00045790893353194574, "loss": 1.4043, "step": 42320 }, { "epoch": 1.7, "learning_rate": 0.00045789886871452153, "loss": 1.4718, "step": 42330 }, { "epoch": 1.7, "learning_rate": 0.0004578888038970973, "loss": 1.5251, "step": 42340 }, { "epoch": 1.7, "learning_rate": 0.0004578787390796731, "loss": 1.3548, "step": 42350 }, { "epoch": 1.7, "learning_rate": 0.0004578686742622489, "loss": 1.3633, "step": 42360 }, { "epoch": 1.7, "learning_rate": 0.00045785860944482464, "loss": 1.3656, "step": 42370 }, { "epoch": 1.7, "learning_rate": 0.00045784854462740043, "loss": 1.3448, "step": 42380 }, { "epoch": 1.7, "learning_rate": 0.00045783847980997627, "loss": 1.355, "step": 42390 }, { "epoch": 1.71, "learning_rate": 0.00045782841499255206, "loss": 1.4619, "step": 42400 }, { "epoch": 1.71, "learning_rate": 0.00045781835017512785, "loss": 1.6015, "step": 42410 }, { "epoch": 1.71, "learning_rate": 0.00045780828535770364, "loss": 1.6506, "step": 42420 }, { "epoch": 1.71, "learning_rate": 0.00045779822054027944, "loss": 1.3765, "step": 42430 }, { "epoch": 1.71, "learning_rate": 0.0004577881557228552, "loss": 1.4412, "step": 42440 }, { "epoch": 1.71, "learning_rate": 0.00045777809090543096, "loss": 1.4222, "step": 42450 }, { "epoch": 1.71, "learning_rate": 0.00045776802608800675, "loss": 1.3726, "step": 42460 }, { "epoch": 1.71, "learning_rate": 0.00045775796127058254, "loss": 1.3529, "step": 42470 }, { "epoch": 1.71, "learning_rate": 0.00045774789645315834, "loss": 1.2873, "step": 42480 }, { "epoch": 1.71, "learning_rate": 0.0004577378316357341, "loss": 1.6219, "step": 42490 }, { "epoch": 1.71, "learning_rate": 0.0004577277668183099, "loss": 1.373, "step": 42500 }, { "epoch": 1.71, "learning_rate": 0.0004577177020008857, "loss": 1.3833, "step": 42510 }, { "epoch": 1.71, "learning_rate": 0.0004577076371834615, "loss": 1.3519, "step": 42520 }, { "epoch": 1.71, "learning_rate": 0.0004576975723660373, "loss": 1.3118, "step": 42530 }, { "epoch": 1.71, "learning_rate": 0.0004576875075486131, "loss": 1.4676, "step": 42540 }, { "epoch": 1.71, "learning_rate": 0.00045767744273118887, "loss": 1.2803, "step": 42550 }, { "epoch": 1.71, "learning_rate": 0.00045766737791376466, "loss": 1.6362, "step": 42560 }, { "epoch": 1.71, "learning_rate": 0.00045765731309634045, "loss": 1.3425, "step": 42570 }, { "epoch": 1.71, "learning_rate": 0.00045764724827891624, "loss": 1.6348, "step": 42580 }, { "epoch": 1.71, "learning_rate": 0.00045763718346149203, "loss": 1.6002, "step": 42590 }, { "epoch": 1.71, "learning_rate": 0.0004576271186440678, "loss": 1.3626, "step": 42600 }, { "epoch": 1.71, "learning_rate": 0.0004576170538266436, "loss": 1.4266, "step": 42610 }, { "epoch": 1.71, "learning_rate": 0.00045760698900921935, "loss": 1.2339, "step": 42620 }, { "epoch": 1.71, "learning_rate": 0.00045759692419179514, "loss": 1.7489, "step": 42630 }, { "epoch": 1.71, "learning_rate": 0.00045758685937437093, "loss": 1.2268, "step": 42640 }, { "epoch": 1.72, "learning_rate": 0.0004575767945569467, "loss": 1.4911, "step": 42650 }, { "epoch": 1.72, "learning_rate": 0.0004575667297395225, "loss": 1.3515, "step": 42660 }, { "epoch": 1.72, "learning_rate": 0.0004575566649220983, "loss": 1.6867, "step": 42670 }, { "epoch": 1.72, "learning_rate": 0.00045754660010467415, "loss": 1.1972, "step": 42680 }, { "epoch": 1.72, "learning_rate": 0.00045753653528724994, "loss": 1.5373, "step": 42690 }, { "epoch": 1.72, "learning_rate": 0.00045752647046982573, "loss": 1.6574, "step": 42700 }, { "epoch": 1.72, "learning_rate": 0.00045751640565240147, "loss": 1.3013, "step": 42710 }, { "epoch": 1.72, "learning_rate": 0.00045750634083497726, "loss": 1.3771, "step": 42720 }, { "epoch": 1.72, "learning_rate": 0.00045749627601755305, "loss": 1.2386, "step": 42730 }, { "epoch": 1.72, "learning_rate": 0.00045748621120012884, "loss": 1.4041, "step": 42740 }, { "epoch": 1.72, "learning_rate": 0.00045747614638270463, "loss": 1.3653, "step": 42750 }, { "epoch": 1.72, "learning_rate": 0.0004574660815652804, "loss": 1.3129, "step": 42760 }, { "epoch": 1.72, "learning_rate": 0.0004574560167478562, "loss": 1.3982, "step": 42770 }, { "epoch": 1.72, "learning_rate": 0.000457445951930432, "loss": 1.5066, "step": 42780 }, { "epoch": 1.72, "learning_rate": 0.00045743588711300774, "loss": 1.3732, "step": 42790 }, { "epoch": 1.72, "learning_rate": 0.00045742582229558353, "loss": 1.4496, "step": 42800 }, { "epoch": 1.72, "learning_rate": 0.0004574157574781593, "loss": 1.4651, "step": 42810 }, { "epoch": 1.72, "learning_rate": 0.0004574056926607351, "loss": 1.6273, "step": 42820 }, { "epoch": 1.72, "learning_rate": 0.00045739562784331096, "loss": 1.762, "step": 42830 }, { "epoch": 1.72, "learning_rate": 0.00045738556302588675, "loss": 1.2764, "step": 42840 }, { "epoch": 1.72, "learning_rate": 0.00045737549820846254, "loss": 1.5536, "step": 42850 }, { "epoch": 1.72, "learning_rate": 0.00045736543339103833, "loss": 1.609, "step": 42860 }, { "epoch": 1.72, "learning_rate": 0.0004573553685736141, "loss": 1.5619, "step": 42870 }, { "epoch": 1.72, "learning_rate": 0.00045734530375618986, "loss": 1.4133, "step": 42880 }, { "epoch": 1.72, "learning_rate": 0.00045733523893876565, "loss": 1.6628, "step": 42890 }, { "epoch": 1.73, "learning_rate": 0.00045732517412134144, "loss": 1.9489, "step": 42900 }, { "epoch": 1.73, "learning_rate": 0.00045731510930391723, "loss": 1.4125, "step": 42910 }, { "epoch": 1.73, "learning_rate": 0.000457305044486493, "loss": 1.4565, "step": 42920 }, { "epoch": 1.73, "learning_rate": 0.0004572949796690688, "loss": 1.5001, "step": 42930 }, { "epoch": 1.73, "learning_rate": 0.0004572849148516446, "loss": 1.6602, "step": 42940 }, { "epoch": 1.73, "learning_rate": 0.0004572748500342204, "loss": 1.341, "step": 42950 }, { "epoch": 1.73, "learning_rate": 0.00045726478521679613, "loss": 1.1952, "step": 42960 }, { "epoch": 1.73, "learning_rate": 0.0004572547203993719, "loss": 1.3794, "step": 42970 }, { "epoch": 1.73, "learning_rate": 0.00045724465558194777, "loss": 1.5226, "step": 42980 }, { "epoch": 1.73, "learning_rate": 0.00045723459076452356, "loss": 1.4178, "step": 42990 }, { "epoch": 1.73, "learning_rate": 0.00045722452594709935, "loss": 1.3508, "step": 43000 }, { "epoch": 1.73, "learning_rate": 0.00045721446112967514, "loss": 1.5112, "step": 43010 }, { "epoch": 1.73, "learning_rate": 0.00045720439631225093, "loss": 1.4738, "step": 43020 }, { "epoch": 1.73, "learning_rate": 0.0004571943314948267, "loss": 1.4678, "step": 43030 }, { "epoch": 1.73, "learning_rate": 0.00045718426667740246, "loss": 1.2443, "step": 43040 }, { "epoch": 1.73, "learning_rate": 0.00045717420185997825, "loss": 1.4928, "step": 43050 }, { "epoch": 1.73, "learning_rate": 0.00045716413704255404, "loss": 1.2254, "step": 43060 }, { "epoch": 1.73, "learning_rate": 0.00045715407222512983, "loss": 1.5129, "step": 43070 }, { "epoch": 1.73, "learning_rate": 0.0004571440074077056, "loss": 1.447, "step": 43080 }, { "epoch": 1.73, "learning_rate": 0.0004571339425902814, "loss": 1.6737, "step": 43090 }, { "epoch": 1.73, "learning_rate": 0.0004571238777728572, "loss": 1.4606, "step": 43100 }, { "epoch": 1.73, "learning_rate": 0.000457113812955433, "loss": 1.5057, "step": 43110 }, { "epoch": 1.73, "learning_rate": 0.00045710374813800884, "loss": 1.5616, "step": 43120 }, { "epoch": 1.73, "learning_rate": 0.0004570936833205846, "loss": 1.6395, "step": 43130 }, { "epoch": 1.74, "learning_rate": 0.00045708361850316036, "loss": 1.387, "step": 43140 }, { "epoch": 1.74, "learning_rate": 0.00045707355368573615, "loss": 1.599, "step": 43150 }, { "epoch": 1.74, "learning_rate": 0.00045706348886831195, "loss": 1.5052, "step": 43160 }, { "epoch": 1.74, "learning_rate": 0.00045705342405088774, "loss": 1.6717, "step": 43170 }, { "epoch": 1.74, "learning_rate": 0.00045704335923346353, "loss": 1.3978, "step": 43180 }, { "epoch": 1.74, "learning_rate": 0.0004570332944160393, "loss": 1.4504, "step": 43190 }, { "epoch": 1.74, "learning_rate": 0.0004570232295986151, "loss": 1.4329, "step": 43200 }, { "epoch": 1.74, "learning_rate": 0.00045701316478119085, "loss": 1.4295, "step": 43210 }, { "epoch": 1.74, "learning_rate": 0.00045700309996376664, "loss": 1.4104, "step": 43220 }, { "epoch": 1.74, "learning_rate": 0.0004569930351463424, "loss": 1.4473, "step": 43230 }, { "epoch": 1.74, "learning_rate": 0.0004569829703289182, "loss": 1.5601, "step": 43240 }, { "epoch": 1.74, "learning_rate": 0.000456972905511494, "loss": 1.3688, "step": 43250 }, { "epoch": 1.74, "learning_rate": 0.0004569628406940698, "loss": 1.2956, "step": 43260 }, { "epoch": 1.74, "learning_rate": 0.00045695277587664564, "loss": 1.3465, "step": 43270 }, { "epoch": 1.74, "learning_rate": 0.00045694271105922144, "loss": 1.3675, "step": 43280 }, { "epoch": 1.74, "learning_rate": 0.0004569326462417972, "loss": 1.5327, "step": 43290 }, { "epoch": 1.74, "learning_rate": 0.00045692258142437296, "loss": 1.4564, "step": 43300 }, { "epoch": 1.74, "learning_rate": 0.00045691251660694875, "loss": 1.7983, "step": 43310 }, { "epoch": 1.74, "learning_rate": 0.00045690245178952454, "loss": 1.4906, "step": 43320 }, { "epoch": 1.74, "learning_rate": 0.00045689238697210033, "loss": 1.3205, "step": 43330 }, { "epoch": 1.74, "learning_rate": 0.0004568823221546761, "loss": 1.5586, "step": 43340 }, { "epoch": 1.74, "learning_rate": 0.0004568722573372519, "loss": 1.6249, "step": 43350 }, { "epoch": 1.74, "learning_rate": 0.0004568621925198277, "loss": 1.7255, "step": 43360 }, { "epoch": 1.74, "learning_rate": 0.0004568521277024035, "loss": 1.1687, "step": 43370 }, { "epoch": 1.74, "learning_rate": 0.00045684206288497923, "loss": 1.14, "step": 43380 }, { "epoch": 1.75, "learning_rate": 0.000456831998067555, "loss": 1.2484, "step": 43390 }, { "epoch": 1.75, "learning_rate": 0.0004568219332501308, "loss": 1.5762, "step": 43400 }, { "epoch": 1.75, "learning_rate": 0.00045681186843270666, "loss": 1.3362, "step": 43410 }, { "epoch": 1.75, "learning_rate": 0.00045680180361528245, "loss": 1.612, "step": 43420 }, { "epoch": 1.75, "learning_rate": 0.00045679173879785824, "loss": 1.3388, "step": 43430 }, { "epoch": 1.75, "learning_rate": 0.00045678167398043403, "loss": 1.5833, "step": 43440 }, { "epoch": 1.75, "learning_rate": 0.0004567716091630098, "loss": 1.157, "step": 43450 }, { "epoch": 1.75, "learning_rate": 0.0004567615443455856, "loss": 1.5578, "step": 43460 }, { "epoch": 1.75, "learning_rate": 0.00045675147952816135, "loss": 1.2453, "step": 43470 }, { "epoch": 1.75, "learning_rate": 0.00045674141471073714, "loss": 1.4565, "step": 43480 }, { "epoch": 1.75, "learning_rate": 0.00045673134989331293, "loss": 1.6303, "step": 43490 }, { "epoch": 1.75, "learning_rate": 0.0004567212850758887, "loss": 1.5227, "step": 43500 }, { "epoch": 1.75, "learning_rate": 0.0004567112202584645, "loss": 1.3921, "step": 43510 }, { "epoch": 1.75, "learning_rate": 0.0004567011554410403, "loss": 1.4515, "step": 43520 }, { "epoch": 1.75, "learning_rate": 0.0004566910906236161, "loss": 1.3405, "step": 43530 }, { "epoch": 1.75, "learning_rate": 0.0004566810258061919, "loss": 1.2626, "step": 43540 }, { "epoch": 1.75, "learning_rate": 0.0004566709609887676, "loss": 1.1044, "step": 43550 }, { "epoch": 1.75, "learning_rate": 0.00045666089617134347, "loss": 1.411, "step": 43560 }, { "epoch": 1.75, "learning_rate": 0.00045665083135391926, "loss": 1.2823, "step": 43570 }, { "epoch": 1.75, "learning_rate": 0.00045664076653649505, "loss": 1.684, "step": 43580 }, { "epoch": 1.75, "learning_rate": 0.00045663070171907084, "loss": 1.3891, "step": 43590 }, { "epoch": 1.75, "learning_rate": 0.00045662063690164663, "loss": 1.582, "step": 43600 }, { "epoch": 1.75, "learning_rate": 0.0004566105720842224, "loss": 1.3528, "step": 43610 }, { "epoch": 1.75, "learning_rate": 0.0004566005072667982, "loss": 1.5075, "step": 43620 }, { "epoch": 1.75, "learning_rate": 0.000456590442449374, "loss": 1.2499, "step": 43630 }, { "epoch": 1.76, "learning_rate": 0.00045658037763194974, "loss": 1.2979, "step": 43640 }, { "epoch": 1.76, "learning_rate": 0.00045657031281452553, "loss": 1.4075, "step": 43650 }, { "epoch": 1.76, "learning_rate": 0.0004565602479971013, "loss": 1.7069, "step": 43660 }, { "epoch": 1.76, "learning_rate": 0.0004565501831796771, "loss": 1.3855, "step": 43670 }, { "epoch": 1.76, "learning_rate": 0.0004565401183622529, "loss": 1.3325, "step": 43680 }, { "epoch": 1.76, "learning_rate": 0.0004565300535448287, "loss": 1.3259, "step": 43690 }, { "epoch": 1.76, "learning_rate": 0.0004565199887274045, "loss": 1.5108, "step": 43700 }, { "epoch": 1.76, "learning_rate": 0.00045650992390998033, "loss": 1.514, "step": 43710 }, { "epoch": 1.76, "learning_rate": 0.00045649985909255607, "loss": 1.4303, "step": 43720 }, { "epoch": 1.76, "learning_rate": 0.00045648979427513186, "loss": 1.7544, "step": 43730 }, { "epoch": 1.76, "learning_rate": 0.00045647972945770765, "loss": 1.3252, "step": 43740 }, { "epoch": 1.76, "learning_rate": 0.00045646966464028344, "loss": 1.6833, "step": 43750 }, { "epoch": 1.76, "learning_rate": 0.00045645959982285923, "loss": 1.491, "step": 43760 }, { "epoch": 1.76, "learning_rate": 0.000456449535005435, "loss": 1.3664, "step": 43770 }, { "epoch": 1.76, "learning_rate": 0.0004564394701880108, "loss": 1.5269, "step": 43780 }, { "epoch": 1.76, "learning_rate": 0.0004564294053705866, "loss": 1.4491, "step": 43790 }, { "epoch": 1.76, "learning_rate": 0.00045641934055316234, "loss": 1.2749, "step": 43800 }, { "epoch": 1.76, "learning_rate": 0.00045640927573573813, "loss": 1.6278, "step": 43810 }, { "epoch": 1.76, "learning_rate": 0.0004563992109183139, "loss": 1.5781, "step": 43820 }, { "epoch": 1.76, "learning_rate": 0.0004563891461008897, "loss": 1.5394, "step": 43830 }, { "epoch": 1.76, "learning_rate": 0.0004563790812834655, "loss": 1.6145, "step": 43840 }, { "epoch": 1.76, "learning_rate": 0.00045636901646604135, "loss": 1.2602, "step": 43850 }, { "epoch": 1.76, "learning_rate": 0.00045635895164861714, "loss": 1.5619, "step": 43860 }, { "epoch": 1.76, "learning_rate": 0.00045634888683119293, "loss": 1.5061, "step": 43870 }, { "epoch": 1.76, "learning_rate": 0.0004563388220137687, "loss": 1.4982, "step": 43880 }, { "epoch": 1.77, "learning_rate": 0.00045632875719634446, "loss": 1.2743, "step": 43890 }, { "epoch": 1.77, "learning_rate": 0.00045631869237892025, "loss": 1.5102, "step": 43900 }, { "epoch": 1.77, "learning_rate": 0.00045630862756149604, "loss": 1.5816, "step": 43910 }, { "epoch": 1.77, "learning_rate": 0.00045629856274407183, "loss": 1.2869, "step": 43920 }, { "epoch": 1.77, "learning_rate": 0.0004562884979266476, "loss": 1.5321, "step": 43930 }, { "epoch": 1.77, "learning_rate": 0.0004562784331092234, "loss": 1.4156, "step": 43940 }, { "epoch": 1.77, "learning_rate": 0.0004562683682917992, "loss": 1.4088, "step": 43950 }, { "epoch": 1.77, "learning_rate": 0.000456258303474375, "loss": 1.3652, "step": 43960 }, { "epoch": 1.77, "learning_rate": 0.00045624823865695073, "loss": 1.5593, "step": 43970 }, { "epoch": 1.77, "learning_rate": 0.0004562381738395265, "loss": 1.3126, "step": 43980 }, { "epoch": 1.77, "learning_rate": 0.0004562281090221023, "loss": 1.4309, "step": 43990 }, { "epoch": 1.77, "learning_rate": 0.00045621804420467815, "loss": 1.1814, "step": 44000 }, { "epoch": 1.77, "learning_rate": 0.00045620797938725395, "loss": 1.4535, "step": 44010 }, { "epoch": 1.77, "learning_rate": 0.00045619791456982974, "loss": 1.4917, "step": 44020 }, { "epoch": 1.77, "learning_rate": 0.0004561878497524055, "loss": 1.3221, "step": 44030 }, { "epoch": 1.77, "learning_rate": 0.0004561777849349813, "loss": 1.378, "step": 44040 }, { "epoch": 1.77, "learning_rate": 0.0004561677201175571, "loss": 1.2789, "step": 44050 }, { "epoch": 1.77, "learning_rate": 0.00045615765530013285, "loss": 1.2836, "step": 44060 }, { "epoch": 1.77, "learning_rate": 0.00045614759048270864, "loss": 1.4756, "step": 44070 }, { "epoch": 1.77, "learning_rate": 0.0004561375256652844, "loss": 1.4047, "step": 44080 }, { "epoch": 1.77, "learning_rate": 0.0004561274608478602, "loss": 1.3967, "step": 44090 }, { "epoch": 1.77, "learning_rate": 0.000456117396030436, "loss": 1.5464, "step": 44100 }, { "epoch": 1.77, "learning_rate": 0.0004561073312130118, "loss": 1.2976, "step": 44110 }, { "epoch": 1.77, "learning_rate": 0.0004560972663955876, "loss": 1.4996, "step": 44120 }, { "epoch": 1.77, "learning_rate": 0.0004560872015781634, "loss": 1.3799, "step": 44130 }, { "epoch": 1.78, "learning_rate": 0.0004560771367607391, "loss": 1.5455, "step": 44140 }, { "epoch": 1.78, "learning_rate": 0.00045606707194331496, "loss": 1.6261, "step": 44150 }, { "epoch": 1.78, "learning_rate": 0.00045605700712589075, "loss": 1.5539, "step": 44160 }, { "epoch": 1.78, "learning_rate": 0.00045604694230846654, "loss": 1.5994, "step": 44170 }, { "epoch": 1.78, "learning_rate": 0.00045603687749104233, "loss": 1.3882, "step": 44180 }, { "epoch": 1.78, "learning_rate": 0.0004560268126736181, "loss": 1.235, "step": 44190 }, { "epoch": 1.78, "learning_rate": 0.0004560167478561939, "loss": 1.3527, "step": 44200 }, { "epoch": 1.78, "learning_rate": 0.0004560066830387697, "loss": 1.3451, "step": 44210 }, { "epoch": 1.78, "learning_rate": 0.0004559966182213455, "loss": 1.2001, "step": 44220 }, { "epoch": 1.78, "learning_rate": 0.00045598655340392123, "loss": 1.5629, "step": 44230 }, { "epoch": 1.78, "learning_rate": 0.000455976488586497, "loss": 1.2858, "step": 44240 }, { "epoch": 1.78, "learning_rate": 0.0004559664237690728, "loss": 1.5969, "step": 44250 }, { "epoch": 1.78, "learning_rate": 0.0004559563589516486, "loss": 1.3156, "step": 44260 }, { "epoch": 1.78, "learning_rate": 0.0004559462941342244, "loss": 1.4309, "step": 44270 }, { "epoch": 1.78, "learning_rate": 0.0004559362293168002, "loss": 1.3529, "step": 44280 }, { "epoch": 1.78, "learning_rate": 0.00045592616449937603, "loss": 1.3734, "step": 44290 }, { "epoch": 1.78, "learning_rate": 0.0004559160996819518, "loss": 1.3153, "step": 44300 }, { "epoch": 1.78, "learning_rate": 0.00045590603486452756, "loss": 1.3072, "step": 44310 }, { "epoch": 1.78, "learning_rate": 0.00045589597004710335, "loss": 1.4496, "step": 44320 }, { "epoch": 1.78, "learning_rate": 0.00045588590522967914, "loss": 1.416, "step": 44330 }, { "epoch": 1.78, "learning_rate": 0.00045587584041225493, "loss": 1.5824, "step": 44340 }, { "epoch": 1.78, "learning_rate": 0.0004558657755948307, "loss": 1.6299, "step": 44350 }, { "epoch": 1.78, "learning_rate": 0.0004558557107774065, "loss": 1.6159, "step": 44360 }, { "epoch": 1.78, "learning_rate": 0.0004558456459599823, "loss": 1.258, "step": 44370 }, { "epoch": 1.78, "learning_rate": 0.0004558355811425581, "loss": 1.4139, "step": 44380 }, { "epoch": 1.79, "learning_rate": 0.0004558255163251339, "loss": 1.4363, "step": 44390 }, { "epoch": 1.79, "learning_rate": 0.0004558154515077096, "loss": 1.155, "step": 44400 }, { "epoch": 1.79, "learning_rate": 0.0004558053866902854, "loss": 1.4271, "step": 44410 }, { "epoch": 1.79, "learning_rate": 0.0004557953218728612, "loss": 1.495, "step": 44420 }, { "epoch": 1.79, "learning_rate": 0.000455785257055437, "loss": 1.4429, "step": 44430 }, { "epoch": 1.79, "learning_rate": 0.00045577519223801284, "loss": 1.5063, "step": 44440 }, { "epoch": 1.79, "learning_rate": 0.00045576512742058863, "loss": 1.3864, "step": 44450 }, { "epoch": 1.79, "learning_rate": 0.0004557550626031644, "loss": 1.4752, "step": 44460 }, { "epoch": 1.79, "learning_rate": 0.0004557449977857402, "loss": 1.655, "step": 44470 }, { "epoch": 1.79, "learning_rate": 0.00045573493296831595, "loss": 1.4233, "step": 44480 }, { "epoch": 1.79, "learning_rate": 0.00045572486815089174, "loss": 1.438, "step": 44490 }, { "epoch": 1.79, "learning_rate": 0.00045571480333346753, "loss": 1.7341, "step": 44500 }, { "epoch": 1.79, "learning_rate": 0.0004557047385160433, "loss": 1.2828, "step": 44510 }, { "epoch": 1.79, "learning_rate": 0.0004556946736986191, "loss": 1.5035, "step": 44520 }, { "epoch": 1.79, "learning_rate": 0.0004556846088811949, "loss": 1.4824, "step": 44530 }, { "epoch": 1.79, "learning_rate": 0.0004556745440637707, "loss": 1.4001, "step": 44540 }, { "epoch": 1.79, "learning_rate": 0.0004556644792463465, "loss": 1.5237, "step": 44550 }, { "epoch": 1.79, "learning_rate": 0.0004556544144289222, "loss": 1.2141, "step": 44560 }, { "epoch": 1.79, "learning_rate": 0.000455644349611498, "loss": 1.4563, "step": 44570 }, { "epoch": 1.79, "learning_rate": 0.00045563428479407386, "loss": 1.3396, "step": 44580 }, { "epoch": 1.79, "learning_rate": 0.00045562421997664965, "loss": 1.3854, "step": 44590 }, { "epoch": 1.79, "learning_rate": 0.00045561415515922544, "loss": 1.3254, "step": 44600 }, { "epoch": 1.79, "learning_rate": 0.00045560409034180123, "loss": 1.3818, "step": 44610 }, { "epoch": 1.79, "learning_rate": 0.000455594025524377, "loss": 1.3772, "step": 44620 }, { "epoch": 1.79, "learning_rate": 0.0004555839607069528, "loss": 1.2477, "step": 44630 }, { "epoch": 1.8, "learning_rate": 0.0004555738958895286, "loss": 1.2783, "step": 44640 }, { "epoch": 1.8, "learning_rate": 0.00045556383107210434, "loss": 1.341, "step": 44650 }, { "epoch": 1.8, "learning_rate": 0.00045555376625468013, "loss": 1.4415, "step": 44660 }, { "epoch": 1.8, "learning_rate": 0.0004555437014372559, "loss": 1.4811, "step": 44670 }, { "epoch": 1.8, "learning_rate": 0.0004555336366198317, "loss": 1.5386, "step": 44680 }, { "epoch": 1.8, "learning_rate": 0.0004555235718024075, "loss": 1.3676, "step": 44690 }, { "epoch": 1.8, "learning_rate": 0.0004555135069849833, "loss": 1.3131, "step": 44700 }, { "epoch": 1.8, "learning_rate": 0.0004555034421675591, "loss": 1.4877, "step": 44710 }, { "epoch": 1.8, "learning_rate": 0.0004554933773501349, "loss": 1.444, "step": 44720 }, { "epoch": 1.8, "learning_rate": 0.00045548331253271067, "loss": 1.286, "step": 44730 }, { "epoch": 1.8, "learning_rate": 0.00045547324771528646, "loss": 1.2577, "step": 44740 }, { "epoch": 1.8, "learning_rate": 0.00045546318289786225, "loss": 1.5244, "step": 44750 }, { "epoch": 1.8, "learning_rate": 0.00045545311808043804, "loss": 1.5115, "step": 44760 }, { "epoch": 1.8, "learning_rate": 0.00045544305326301383, "loss": 1.4914, "step": 44770 }, { "epoch": 1.8, "learning_rate": 0.0004554329884455896, "loss": 1.6182, "step": 44780 }, { "epoch": 1.8, "learning_rate": 0.0004554229236281654, "loss": 1.5349, "step": 44790 }, { "epoch": 1.8, "learning_rate": 0.0004554128588107412, "loss": 1.2765, "step": 44800 }, { "epoch": 1.8, "learning_rate": 0.000455402793993317, "loss": 1.5186, "step": 44810 }, { "epoch": 1.8, "learning_rate": 0.00045539272917589273, "loss": 1.5094, "step": 44820 }, { "epoch": 1.8, "learning_rate": 0.0004553826643584685, "loss": 1.3151, "step": 44830 }, { "epoch": 1.8, "learning_rate": 0.0004553725995410443, "loss": 1.4954, "step": 44840 }, { "epoch": 1.8, "learning_rate": 0.0004553625347236201, "loss": 1.4388, "step": 44850 }, { "epoch": 1.8, "learning_rate": 0.0004553524699061959, "loss": 1.5249, "step": 44860 }, { "epoch": 1.8, "learning_rate": 0.0004553424050887717, "loss": 1.5125, "step": 44870 }, { "epoch": 1.81, "learning_rate": 0.0004553323402713475, "loss": 1.4182, "step": 44880 }, { "epoch": 1.81, "learning_rate": 0.0004553222754539233, "loss": 1.3622, "step": 44890 }, { "epoch": 1.81, "learning_rate": 0.00045531221063649905, "loss": 1.2538, "step": 44900 }, { "epoch": 1.81, "learning_rate": 0.00045530214581907484, "loss": 1.3372, "step": 44910 }, { "epoch": 1.81, "learning_rate": 0.00045529208100165064, "loss": 1.5536, "step": 44920 }, { "epoch": 1.81, "learning_rate": 0.0004552820161842264, "loss": 1.4602, "step": 44930 }, { "epoch": 1.81, "learning_rate": 0.0004552719513668022, "loss": 1.2397, "step": 44940 }, { "epoch": 1.81, "learning_rate": 0.000455261886549378, "loss": 1.4527, "step": 44950 }, { "epoch": 1.81, "learning_rate": 0.0004552518217319538, "loss": 1.5982, "step": 44960 }, { "epoch": 1.81, "learning_rate": 0.0004552417569145296, "loss": 1.9173, "step": 44970 }, { "epoch": 1.81, "learning_rate": 0.0004552316920971054, "loss": 1.2298, "step": 44980 }, { "epoch": 1.81, "learning_rate": 0.0004552216272796811, "loss": 1.4485, "step": 44990 }, { "epoch": 1.81, "learning_rate": 0.0004552115624622569, "loss": 1.3171, "step": 45000 }, { "epoch": 1.81, "learning_rate": 0.0004552014976448327, "loss": 1.628, "step": 45010 }, { "epoch": 1.81, "learning_rate": 0.00045519143282740854, "loss": 1.629, "step": 45020 }, { "epoch": 1.81, "learning_rate": 0.00045518136800998433, "loss": 1.6892, "step": 45030 }, { "epoch": 1.81, "learning_rate": 0.0004551713031925601, "loss": 1.5108, "step": 45040 }, { "epoch": 1.81, "learning_rate": 0.0004551612383751359, "loss": 1.5331, "step": 45050 }, { "epoch": 1.81, "learning_rate": 0.0004551511735577117, "loss": 1.316, "step": 45060 }, { "epoch": 1.81, "learning_rate": 0.00045514110874028744, "loss": 1.3776, "step": 45070 }, { "epoch": 1.81, "learning_rate": 0.00045513104392286323, "loss": 1.4636, "step": 45080 }, { "epoch": 1.81, "learning_rate": 0.000455120979105439, "loss": 1.4923, "step": 45090 }, { "epoch": 1.81, "learning_rate": 0.0004551109142880148, "loss": 1.4988, "step": 45100 }, { "epoch": 1.81, "learning_rate": 0.0004551008494705906, "loss": 1.3097, "step": 45110 }, { "epoch": 1.81, "learning_rate": 0.0004550907846531664, "loss": 1.5288, "step": 45120 }, { "epoch": 1.82, "learning_rate": 0.0004550807198357422, "loss": 1.4763, "step": 45130 }, { "epoch": 1.82, "learning_rate": 0.000455070655018318, "loss": 1.4789, "step": 45140 }, { "epoch": 1.82, "learning_rate": 0.00045506059020089377, "loss": 1.3256, "step": 45150 }, { "epoch": 1.82, "learning_rate": 0.0004550505253834695, "loss": 1.418, "step": 45160 }, { "epoch": 1.82, "learning_rate": 0.00045504046056604535, "loss": 1.3796, "step": 45170 }, { "epoch": 1.82, "learning_rate": 0.00045503039574862114, "loss": 1.4958, "step": 45180 }, { "epoch": 1.82, "learning_rate": 0.00045502033093119693, "loss": 1.4192, "step": 45190 }, { "epoch": 1.82, "learning_rate": 0.0004550102661137727, "loss": 1.4143, "step": 45200 }, { "epoch": 1.82, "learning_rate": 0.0004550002012963485, "loss": 1.3938, "step": 45210 }, { "epoch": 1.82, "learning_rate": 0.0004549901364789243, "loss": 1.393, "step": 45220 }, { "epoch": 1.82, "learning_rate": 0.0004549800716615001, "loss": 1.3421, "step": 45230 }, { "epoch": 1.82, "learning_rate": 0.00045497000684407583, "loss": 1.232, "step": 45240 }, { "epoch": 1.82, "learning_rate": 0.0004549599420266516, "loss": 1.5965, "step": 45250 }, { "epoch": 1.82, "learning_rate": 0.0004549498772092274, "loss": 1.4582, "step": 45260 }, { "epoch": 1.82, "learning_rate": 0.0004549398123918032, "loss": 1.7602, "step": 45270 }, { "epoch": 1.82, "learning_rate": 0.000454929747574379, "loss": 1.3928, "step": 45280 }, { "epoch": 1.82, "learning_rate": 0.0004549196827569548, "loss": 1.5673, "step": 45290 }, { "epoch": 1.82, "learning_rate": 0.0004549096179395306, "loss": 1.3891, "step": 45300 }, { "epoch": 1.82, "learning_rate": 0.0004548995531221064, "loss": 1.6443, "step": 45310 }, { "epoch": 1.82, "learning_rate": 0.0004548894883046822, "loss": 1.5159, "step": 45320 }, { "epoch": 1.82, "learning_rate": 0.00045487942348725795, "loss": 1.4166, "step": 45330 }, { "epoch": 1.82, "learning_rate": 0.00045486935866983374, "loss": 1.2809, "step": 45340 }, { "epoch": 1.82, "learning_rate": 0.00045485929385240953, "loss": 1.5216, "step": 45350 }, { "epoch": 1.82, "learning_rate": 0.0004548492290349853, "loss": 1.3842, "step": 45360 }, { "epoch": 1.82, "learning_rate": 0.0004548391642175611, "loss": 1.7422, "step": 45370 }, { "epoch": 1.83, "learning_rate": 0.0004548290994001369, "loss": 1.4539, "step": 45380 }, { "epoch": 1.83, "learning_rate": 0.0004548190345827127, "loss": 1.5146, "step": 45390 }, { "epoch": 1.83, "learning_rate": 0.0004548089697652885, "loss": 1.3149, "step": 45400 }, { "epoch": 1.83, "learning_rate": 0.0004547989049478642, "loss": 1.5384, "step": 45410 }, { "epoch": 1.83, "learning_rate": 0.00045478884013044, "loss": 1.5568, "step": 45420 }, { "epoch": 1.83, "learning_rate": 0.0004547787753130158, "loss": 1.2306, "step": 45430 }, { "epoch": 1.83, "learning_rate": 0.0004547687104955916, "loss": 1.7789, "step": 45440 }, { "epoch": 1.83, "learning_rate": 0.0004547586456781674, "loss": 1.3488, "step": 45450 }, { "epoch": 1.83, "learning_rate": 0.00045474858086074323, "loss": 1.5614, "step": 45460 }, { "epoch": 1.83, "learning_rate": 0.000454738516043319, "loss": 1.399, "step": 45470 }, { "epoch": 1.83, "learning_rate": 0.0004547284512258948, "loss": 1.3248, "step": 45480 }, { "epoch": 1.83, "learning_rate": 0.00045471838640847055, "loss": 1.3346, "step": 45490 }, { "epoch": 1.83, "learning_rate": 0.00045470832159104634, "loss": 1.6294, "step": 45500 }, { "epoch": 1.83, "learning_rate": 0.00045469825677362213, "loss": 1.385, "step": 45510 }, { "epoch": 1.83, "learning_rate": 0.0004546881919561979, "loss": 1.2578, "step": 45520 }, { "epoch": 1.83, "learning_rate": 0.0004546781271387737, "loss": 1.4841, "step": 45530 }, { "epoch": 1.83, "learning_rate": 0.0004546680623213495, "loss": 1.1994, "step": 45540 }, { "epoch": 1.83, "learning_rate": 0.0004546579975039253, "loss": 1.3705, "step": 45550 }, { "epoch": 1.83, "learning_rate": 0.0004546479326865011, "loss": 1.7655, "step": 45560 }, { "epoch": 1.83, "learning_rate": 0.0004546378678690769, "loss": 1.3156, "step": 45570 }, { "epoch": 1.83, "learning_rate": 0.0004546278030516526, "loss": 1.5036, "step": 45580 }, { "epoch": 1.83, "learning_rate": 0.0004546177382342284, "loss": 1.3819, "step": 45590 }, { "epoch": 1.83, "learning_rate": 0.0004546076734168042, "loss": 1.5358, "step": 45600 }, { "epoch": 1.83, "learning_rate": 0.00045459760859938004, "loss": 1.5014, "step": 45610 }, { "epoch": 1.83, "learning_rate": 0.00045458754378195583, "loss": 1.2051, "step": 45620 }, { "epoch": 1.84, "learning_rate": 0.0004545774789645316, "loss": 1.3865, "step": 45630 }, { "epoch": 1.84, "learning_rate": 0.0004545674141471074, "loss": 1.6086, "step": 45640 }, { "epoch": 1.84, "learning_rate": 0.0004545573493296832, "loss": 1.6975, "step": 45650 }, { "epoch": 1.84, "learning_rate": 0.00045454728451225894, "loss": 1.5659, "step": 45660 }, { "epoch": 1.84, "learning_rate": 0.00045453721969483473, "loss": 1.2637, "step": 45670 }, { "epoch": 1.84, "learning_rate": 0.0004545271548774105, "loss": 1.6152, "step": 45680 }, { "epoch": 1.84, "learning_rate": 0.0004545170900599863, "loss": 1.424, "step": 45690 }, { "epoch": 1.84, "learning_rate": 0.0004545070252425621, "loss": 1.3324, "step": 45700 }, { "epoch": 1.84, "learning_rate": 0.0004544969604251379, "loss": 1.1791, "step": 45710 }, { "epoch": 1.84, "learning_rate": 0.0004544868956077137, "loss": 1.4745, "step": 45720 }, { "epoch": 1.84, "learning_rate": 0.00045447683079028947, "loss": 1.6022, "step": 45730 }, { "epoch": 1.84, "learning_rate": 0.00045446676597286526, "loss": 1.4834, "step": 45740 }, { "epoch": 1.84, "learning_rate": 0.00045445670115544105, "loss": 1.3155, "step": 45750 }, { "epoch": 1.84, "learning_rate": 0.00045444663633801684, "loss": 1.3427, "step": 45760 }, { "epoch": 1.84, "learning_rate": 0.00045443657152059264, "loss": 1.3566, "step": 45770 }, { "epoch": 1.84, "learning_rate": 0.0004544265067031684, "loss": 1.4851, "step": 45780 }, { "epoch": 1.84, "learning_rate": 0.0004544164418857442, "loss": 1.6705, "step": 45790 }, { "epoch": 1.84, "learning_rate": 0.00045440637706832, "loss": 1.4424, "step": 45800 }, { "epoch": 1.84, "learning_rate": 0.0004543963122508958, "loss": 1.6364, "step": 45810 }, { "epoch": 1.84, "learning_rate": 0.0004543862474334716, "loss": 1.4224, "step": 45820 }, { "epoch": 1.84, "learning_rate": 0.0004543761826160473, "loss": 1.3757, "step": 45830 }, { "epoch": 1.84, "learning_rate": 0.0004543661177986231, "loss": 1.6069, "step": 45840 }, { "epoch": 1.84, "learning_rate": 0.0004543560529811989, "loss": 1.5949, "step": 45850 }, { "epoch": 1.84, "learning_rate": 0.0004543459881637747, "loss": 1.2374, "step": 45860 }, { "epoch": 1.84, "learning_rate": 0.0004543359233463505, "loss": 1.4748, "step": 45870 }, { "epoch": 1.85, "learning_rate": 0.0004543258585289263, "loss": 1.7006, "step": 45880 }, { "epoch": 1.85, "learning_rate": 0.00045431579371150207, "loss": 1.4064, "step": 45890 }, { "epoch": 1.85, "learning_rate": 0.0004543057288940779, "loss": 1.392, "step": 45900 }, { "epoch": 1.85, "learning_rate": 0.0004542956640766537, "loss": 1.4843, "step": 45910 }, { "epoch": 1.85, "learning_rate": 0.00045428559925922944, "loss": 1.8168, "step": 45920 }, { "epoch": 1.85, "learning_rate": 0.00045427553444180523, "loss": 1.2047, "step": 45930 }, { "epoch": 1.85, "learning_rate": 0.000454265469624381, "loss": 1.4158, "step": 45940 }, { "epoch": 1.85, "learning_rate": 0.0004542554048069568, "loss": 1.5045, "step": 45950 }, { "epoch": 1.85, "learning_rate": 0.0004542453399895326, "loss": 1.3994, "step": 45960 }, { "epoch": 1.85, "learning_rate": 0.0004542352751721084, "loss": 1.2381, "step": 45970 }, { "epoch": 1.85, "learning_rate": 0.0004542252103546842, "loss": 1.1077, "step": 45980 }, { "epoch": 1.85, "learning_rate": 0.00045421514553726, "loss": 1.2202, "step": 45990 }, { "epoch": 1.85, "learning_rate": 0.0004542050807198357, "loss": 1.5082, "step": 46000 }, { "epoch": 1.85, "learning_rate": 0.0004541950159024115, "loss": 1.3573, "step": 46010 }, { "epoch": 1.85, "learning_rate": 0.0004541849510849873, "loss": 1.4747, "step": 46020 }, { "epoch": 1.85, "learning_rate": 0.0004541748862675631, "loss": 1.3901, "step": 46030 }, { "epoch": 1.85, "learning_rate": 0.0004541648214501389, "loss": 1.5028, "step": 46040 }, { "epoch": 1.85, "learning_rate": 0.0004541547566327147, "loss": 1.5162, "step": 46050 }, { "epoch": 1.85, "learning_rate": 0.0004541446918152905, "loss": 1.3294, "step": 46060 }, { "epoch": 1.85, "learning_rate": 0.0004541346269978663, "loss": 1.6254, "step": 46070 }, { "epoch": 1.85, "learning_rate": 0.0004541245621804421, "loss": 1.5051, "step": 46080 }, { "epoch": 1.85, "learning_rate": 0.00045411449736301783, "loss": 1.2313, "step": 46090 }, { "epoch": 1.85, "learning_rate": 0.0004541044325455936, "loss": 1.3551, "step": 46100 }, { "epoch": 1.85, "learning_rate": 0.0004540943677281694, "loss": 1.3241, "step": 46110 }, { "epoch": 1.85, "learning_rate": 0.0004540843029107452, "loss": 1.4083, "step": 46120 }, { "epoch": 1.86, "learning_rate": 0.000454074238093321, "loss": 1.6705, "step": 46130 }, { "epoch": 1.86, "learning_rate": 0.0004540641732758968, "loss": 1.3139, "step": 46140 }, { "epoch": 1.86, "learning_rate": 0.0004540541084584726, "loss": 1.4204, "step": 46150 }, { "epoch": 1.86, "learning_rate": 0.00045404404364104837, "loss": 1.2105, "step": 46160 }, { "epoch": 1.86, "learning_rate": 0.0004540339788236241, "loss": 1.6478, "step": 46170 }, { "epoch": 1.86, "learning_rate": 0.0004540239140061999, "loss": 1.4761, "step": 46180 }, { "epoch": 1.86, "learning_rate": 0.00045401384918877574, "loss": 1.3999, "step": 46190 }, { "epoch": 1.86, "learning_rate": 0.00045400378437135153, "loss": 1.2904, "step": 46200 }, { "epoch": 1.86, "learning_rate": 0.0004539937195539273, "loss": 1.4496, "step": 46210 }, { "epoch": 1.86, "learning_rate": 0.0004539836547365031, "loss": 1.5157, "step": 46220 }, { "epoch": 1.86, "learning_rate": 0.0004539735899190789, "loss": 1.3637, "step": 46230 }, { "epoch": 1.86, "learning_rate": 0.0004539635251016547, "loss": 1.3612, "step": 46240 }, { "epoch": 1.86, "learning_rate": 0.00045395346028423043, "loss": 1.475, "step": 46250 }, { "epoch": 1.86, "learning_rate": 0.0004539433954668062, "loss": 1.6795, "step": 46260 }, { "epoch": 1.86, "learning_rate": 0.000453933330649382, "loss": 1.1383, "step": 46270 }, { "epoch": 1.86, "learning_rate": 0.0004539232658319578, "loss": 1.2316, "step": 46280 }, { "epoch": 1.86, "learning_rate": 0.0004539132010145336, "loss": 1.5378, "step": 46290 }, { "epoch": 1.86, "learning_rate": 0.0004539031361971094, "loss": 1.3437, "step": 46300 }, { "epoch": 1.86, "learning_rate": 0.0004538930713796852, "loss": 1.8107, "step": 46310 }, { "epoch": 1.86, "learning_rate": 0.00045388300656226097, "loss": 1.4942, "step": 46320 }, { "epoch": 1.86, "learning_rate": 0.00045387294174483676, "loss": 1.3404, "step": 46330 }, { "epoch": 1.86, "learning_rate": 0.00045386287692741255, "loss": 1.2261, "step": 46340 }, { "epoch": 1.86, "learning_rate": 0.00045385281210998834, "loss": 1.3491, "step": 46350 }, { "epoch": 1.86, "learning_rate": 0.00045384274729256413, "loss": 1.4145, "step": 46360 }, { "epoch": 1.86, "learning_rate": 0.0004538326824751399, "loss": 1.5861, "step": 46370 }, { "epoch": 1.87, "learning_rate": 0.0004538226176577157, "loss": 1.5182, "step": 46380 }, { "epoch": 1.87, "learning_rate": 0.0004538125528402915, "loss": 1.5277, "step": 46390 }, { "epoch": 1.87, "learning_rate": 0.0004538024880228673, "loss": 1.6941, "step": 46400 }, { "epoch": 1.87, "learning_rate": 0.0004537924232054431, "loss": 1.4839, "step": 46410 }, { "epoch": 1.87, "learning_rate": 0.0004537823583880188, "loss": 1.4351, "step": 46420 }, { "epoch": 1.87, "learning_rate": 0.0004537722935705946, "loss": 1.4675, "step": 46430 }, { "epoch": 1.87, "learning_rate": 0.0004537622287531704, "loss": 1.3893, "step": 46440 }, { "epoch": 1.87, "learning_rate": 0.0004537521639357462, "loss": 1.3476, "step": 46450 }, { "epoch": 1.87, "learning_rate": 0.000453742099118322, "loss": 1.5172, "step": 46460 }, { "epoch": 1.87, "learning_rate": 0.0004537320343008978, "loss": 1.5038, "step": 46470 }, { "epoch": 1.87, "learning_rate": 0.0004537219694834736, "loss": 1.4545, "step": 46480 }, { "epoch": 1.87, "learning_rate": 0.0004537119046660494, "loss": 1.3035, "step": 46490 }, { "epoch": 1.87, "learning_rate": 0.0004537018398486252, "loss": 1.2655, "step": 46500 }, { "epoch": 1.87, "learning_rate": 0.00045369177503120094, "loss": 1.5154, "step": 46510 }, { "epoch": 1.87, "learning_rate": 0.00045368171021377673, "loss": 1.5288, "step": 46520 }, { "epoch": 1.87, "learning_rate": 0.0004536716453963525, "loss": 1.276, "step": 46530 }, { "epoch": 1.87, "learning_rate": 0.0004536615805789283, "loss": 1.3421, "step": 46540 }, { "epoch": 1.87, "learning_rate": 0.0004536515157615041, "loss": 1.5032, "step": 46550 }, { "epoch": 1.87, "learning_rate": 0.0004536414509440799, "loss": 1.255, "step": 46560 }, { "epoch": 1.87, "learning_rate": 0.0004536313861266557, "loss": 1.3183, "step": 46570 }, { "epoch": 1.87, "learning_rate": 0.00045362132130923147, "loss": 1.5507, "step": 46580 }, { "epoch": 1.87, "learning_rate": 0.0004536112564918072, "loss": 1.4041, "step": 46590 }, { "epoch": 1.87, "learning_rate": 0.000453601191674383, "loss": 1.5486, "step": 46600 }, { "epoch": 1.87, "learning_rate": 0.0004535911268569588, "loss": 1.6507, "step": 46610 }, { "epoch": 1.87, "learning_rate": 0.0004535810620395346, "loss": 1.3245, "step": 46620 }, { "epoch": 1.88, "learning_rate": 0.0004535709972221104, "loss": 1.2715, "step": 46630 }, { "epoch": 1.88, "learning_rate": 0.0004535609324046862, "loss": 1.2891, "step": 46640 }, { "epoch": 1.88, "learning_rate": 0.000453550867587262, "loss": 1.4328, "step": 46650 }, { "epoch": 1.88, "learning_rate": 0.0004535408027698378, "loss": 1.4675, "step": 46660 }, { "epoch": 1.88, "learning_rate": 0.0004535307379524136, "loss": 1.3848, "step": 46670 }, { "epoch": 1.88, "learning_rate": 0.0004535206731349893, "loss": 1.7282, "step": 46680 }, { "epoch": 1.88, "learning_rate": 0.0004535106083175651, "loss": 1.7564, "step": 46690 }, { "epoch": 1.88, "learning_rate": 0.0004535005435001409, "loss": 1.5339, "step": 46700 }, { "epoch": 1.88, "learning_rate": 0.0004534904786827167, "loss": 1.5032, "step": 46710 }, { "epoch": 1.88, "learning_rate": 0.0004534804138652925, "loss": 1.5058, "step": 46720 }, { "epoch": 1.88, "learning_rate": 0.0004534703490478683, "loss": 1.5093, "step": 46730 }, { "epoch": 1.88, "learning_rate": 0.00045346028423044407, "loss": 1.5479, "step": 46740 }, { "epoch": 1.88, "learning_rate": 0.00045345021941301986, "loss": 1.6685, "step": 46750 }, { "epoch": 1.88, "learning_rate": 0.0004534401545955956, "loss": 1.2076, "step": 46760 }, { "epoch": 1.88, "learning_rate": 0.0004534300897781714, "loss": 1.4517, "step": 46770 }, { "epoch": 1.88, "learning_rate": 0.00045342002496074723, "loss": 1.2174, "step": 46780 }, { "epoch": 1.88, "learning_rate": 0.000453409960143323, "loss": 1.4921, "step": 46790 }, { "epoch": 1.88, "learning_rate": 0.0004533998953258988, "loss": 1.3234, "step": 46800 }, { "epoch": 1.88, "learning_rate": 0.0004533898305084746, "loss": 1.3201, "step": 46810 }, { "epoch": 1.88, "learning_rate": 0.0004533797656910504, "loss": 1.4028, "step": 46820 }, { "epoch": 1.88, "learning_rate": 0.0004533697008736262, "loss": 1.2764, "step": 46830 }, { "epoch": 1.88, "learning_rate": 0.000453359636056202, "loss": 1.6598, "step": 46840 }, { "epoch": 1.88, "learning_rate": 0.0004533495712387777, "loss": 1.3471, "step": 46850 }, { "epoch": 1.88, "learning_rate": 0.0004533395064213535, "loss": 1.4519, "step": 46860 }, { "epoch": 1.89, "learning_rate": 0.0004533294416039293, "loss": 1.5156, "step": 46870 }, { "epoch": 1.89, "learning_rate": 0.0004533193767865051, "loss": 1.2422, "step": 46880 }, { "epoch": 1.89, "learning_rate": 0.0004533093119690809, "loss": 1.4925, "step": 46890 }, { "epoch": 1.89, "learning_rate": 0.00045329924715165667, "loss": 1.4046, "step": 46900 }, { "epoch": 1.89, "learning_rate": 0.00045328918233423246, "loss": 1.3567, "step": 46910 }, { "epoch": 1.89, "learning_rate": 0.0004532791175168083, "loss": 1.3415, "step": 46920 }, { "epoch": 1.89, "learning_rate": 0.00045326905269938404, "loss": 1.3465, "step": 46930 }, { "epoch": 1.89, "learning_rate": 0.00045325898788195983, "loss": 1.4263, "step": 46940 }, { "epoch": 1.89, "learning_rate": 0.0004532489230645356, "loss": 1.4305, "step": 46950 }, { "epoch": 1.89, "learning_rate": 0.0004532388582471114, "loss": 1.3785, "step": 46960 }, { "epoch": 1.89, "learning_rate": 0.0004532287934296872, "loss": 1.3336, "step": 46970 }, { "epoch": 1.89, "learning_rate": 0.000453218728612263, "loss": 1.264, "step": 46980 }, { "epoch": 1.89, "learning_rate": 0.0004532086637948388, "loss": 1.5786, "step": 46990 }, { "epoch": 1.89, "learning_rate": 0.0004531985989774146, "loss": 1.4973, "step": 47000 }, { "epoch": 1.89, "learning_rate": 0.0004531885341599903, "loss": 1.1936, "step": 47010 }, { "epoch": 1.89, "learning_rate": 0.0004531784693425661, "loss": 1.556, "step": 47020 }, { "epoch": 1.89, "learning_rate": 0.0004531684045251419, "loss": 1.6538, "step": 47030 }, { "epoch": 1.89, "learning_rate": 0.0004531583397077177, "loss": 1.6264, "step": 47040 }, { "epoch": 1.89, "learning_rate": 0.0004531482748902935, "loss": 1.3754, "step": 47050 }, { "epoch": 1.89, "learning_rate": 0.00045313821007286927, "loss": 1.3228, "step": 47060 }, { "epoch": 1.89, "learning_rate": 0.0004531281452554451, "loss": 1.4158, "step": 47070 }, { "epoch": 1.89, "learning_rate": 0.0004531180804380209, "loss": 1.1811, "step": 47080 }, { "epoch": 1.89, "learning_rate": 0.0004531080156205967, "loss": 1.3082, "step": 47090 }, { "epoch": 1.89, "learning_rate": 0.00045309795080317243, "loss": 1.4724, "step": 47100 }, { "epoch": 1.89, "learning_rate": 0.0004530878859857482, "loss": 1.5491, "step": 47110 }, { "epoch": 1.9, "learning_rate": 0.000453077821168324, "loss": 1.6265, "step": 47120 }, { "epoch": 1.9, "learning_rate": 0.0004530677563508998, "loss": 1.2581, "step": 47130 }, { "epoch": 1.9, "learning_rate": 0.0004530576915334756, "loss": 1.3383, "step": 47140 }, { "epoch": 1.9, "learning_rate": 0.0004530476267160514, "loss": 1.5589, "step": 47150 }, { "epoch": 1.9, "learning_rate": 0.0004530375618986272, "loss": 1.3282, "step": 47160 }, { "epoch": 1.9, "learning_rate": 0.00045302749708120297, "loss": 1.4603, "step": 47170 }, { "epoch": 1.9, "learning_rate": 0.0004530174322637787, "loss": 1.2665, "step": 47180 }, { "epoch": 1.9, "learning_rate": 0.0004530073674463545, "loss": 1.4805, "step": 47190 }, { "epoch": 1.9, "learning_rate": 0.0004529973026289303, "loss": 1.4044, "step": 47200 }, { "epoch": 1.9, "learning_rate": 0.0004529872378115061, "loss": 1.3345, "step": 47210 }, { "epoch": 1.9, "learning_rate": 0.0004529771729940819, "loss": 1.5778, "step": 47220 }, { "epoch": 1.9, "learning_rate": 0.0004529671081766577, "loss": 1.3495, "step": 47230 }, { "epoch": 1.9, "learning_rate": 0.0004529570433592335, "loss": 1.2776, "step": 47240 }, { "epoch": 1.9, "learning_rate": 0.0004529469785418093, "loss": 1.4375, "step": 47250 }, { "epoch": 1.9, "learning_rate": 0.0004529369137243851, "loss": 1.3192, "step": 47260 }, { "epoch": 1.9, "learning_rate": 0.0004529268489069608, "loss": 1.4471, "step": 47270 }, { "epoch": 1.9, "learning_rate": 0.0004529167840895366, "loss": 1.1506, "step": 47280 }, { "epoch": 1.9, "learning_rate": 0.0004529067192721124, "loss": 1.2383, "step": 47290 }, { "epoch": 1.9, "learning_rate": 0.0004528966544546882, "loss": 1.4682, "step": 47300 }, { "epoch": 1.9, "learning_rate": 0.000452886589637264, "loss": 1.4128, "step": 47310 }, { "epoch": 1.9, "learning_rate": 0.0004528765248198398, "loss": 1.3896, "step": 47320 }, { "epoch": 1.9, "learning_rate": 0.00045286646000241556, "loss": 1.2425, "step": 47330 }, { "epoch": 1.9, "learning_rate": 0.00045285639518499135, "loss": 1.1173, "step": 47340 }, { "epoch": 1.9, "learning_rate": 0.0004528463303675671, "loss": 1.2135, "step": 47350 }, { "epoch": 1.9, "learning_rate": 0.00045283626555014294, "loss": 1.5316, "step": 47360 }, { "epoch": 1.91, "learning_rate": 0.00045282620073271873, "loss": 1.4923, "step": 47370 }, { "epoch": 1.91, "learning_rate": 0.0004528161359152945, "loss": 1.4021, "step": 47380 }, { "epoch": 1.91, "learning_rate": 0.0004528060710978703, "loss": 1.659, "step": 47390 }, { "epoch": 1.91, "learning_rate": 0.0004527960062804461, "loss": 1.5955, "step": 47400 }, { "epoch": 1.91, "learning_rate": 0.0004527859414630219, "loss": 1.2877, "step": 47410 }, { "epoch": 1.91, "learning_rate": 0.0004527758766455977, "loss": 1.4547, "step": 47420 }, { "epoch": 1.91, "learning_rate": 0.00045276581182817347, "loss": 1.5228, "step": 47430 }, { "epoch": 1.91, "learning_rate": 0.0004527557470107492, "loss": 1.5129, "step": 47440 }, { "epoch": 1.91, "learning_rate": 0.000452745682193325, "loss": 1.3642, "step": 47450 }, { "epoch": 1.91, "learning_rate": 0.0004527356173759008, "loss": 1.6076, "step": 47460 }, { "epoch": 1.91, "learning_rate": 0.0004527255525584766, "loss": 1.5509, "step": 47470 }, { "epoch": 1.91, "learning_rate": 0.00045271548774105237, "loss": 1.4671, "step": 47480 }, { "epoch": 1.91, "learning_rate": 0.00045270542292362816, "loss": 1.5487, "step": 47490 }, { "epoch": 1.91, "learning_rate": 0.00045269535810620395, "loss": 1.6844, "step": 47500 }, { "epoch": 1.91, "learning_rate": 0.0004526852932887798, "loss": 1.3961, "step": 47510 }, { "epoch": 1.91, "learning_rate": 0.00045267522847135553, "loss": 1.5304, "step": 47520 }, { "epoch": 1.91, "learning_rate": 0.0004526651636539313, "loss": 1.4573, "step": 47530 }, { "epoch": 1.91, "learning_rate": 0.0004526550988365071, "loss": 1.5523, "step": 47540 }, { "epoch": 1.91, "learning_rate": 0.0004526450340190829, "loss": 1.5149, "step": 47550 }, { "epoch": 1.91, "learning_rate": 0.0004526349692016587, "loss": 1.2447, "step": 47560 }, { "epoch": 1.91, "learning_rate": 0.0004526249043842345, "loss": 1.423, "step": 47570 }, { "epoch": 1.91, "learning_rate": 0.0004526148395668103, "loss": 1.1847, "step": 47580 }, { "epoch": 1.91, "learning_rate": 0.00045260477474938607, "loss": 1.4938, "step": 47590 }, { "epoch": 1.91, "learning_rate": 0.00045259470993196186, "loss": 1.4789, "step": 47600 }, { "epoch": 1.91, "learning_rate": 0.0004525846451145376, "loss": 1.476, "step": 47610 }, { "epoch": 1.92, "learning_rate": 0.0004525745802971134, "loss": 1.4321, "step": 47620 }, { "epoch": 1.92, "learning_rate": 0.0004525645154796892, "loss": 1.4104, "step": 47630 }, { "epoch": 1.92, "learning_rate": 0.00045255445066226497, "loss": 1.4929, "step": 47640 }, { "epoch": 1.92, "learning_rate": 0.0004525443858448408, "loss": 1.4688, "step": 47650 }, { "epoch": 1.92, "learning_rate": 0.0004525343210274166, "loss": 1.533, "step": 47660 }, { "epoch": 1.92, "learning_rate": 0.0004525242562099924, "loss": 1.5093, "step": 47670 }, { "epoch": 1.92, "learning_rate": 0.0004525141913925682, "loss": 1.4146, "step": 47680 }, { "epoch": 1.92, "learning_rate": 0.0004525041265751439, "loss": 1.4592, "step": 47690 }, { "epoch": 1.92, "learning_rate": 0.0004524940617577197, "loss": 1.586, "step": 47700 }, { "epoch": 1.92, "learning_rate": 0.0004524839969402955, "loss": 1.3736, "step": 47710 }, { "epoch": 1.92, "learning_rate": 0.0004524739321228713, "loss": 1.6687, "step": 47720 }, { "epoch": 1.92, "learning_rate": 0.0004524638673054471, "loss": 1.2367, "step": 47730 }, { "epoch": 1.92, "learning_rate": 0.0004524538024880229, "loss": 1.3978, "step": 47740 }, { "epoch": 1.92, "learning_rate": 0.00045244373767059867, "loss": 1.3335, "step": 47750 }, { "epoch": 1.92, "learning_rate": 0.00045243367285317446, "loss": 1.5093, "step": 47760 }, { "epoch": 1.92, "learning_rate": 0.0004524236080357502, "loss": 1.4875, "step": 47770 }, { "epoch": 1.92, "learning_rate": 0.000452413543218326, "loss": 1.4539, "step": 47780 }, { "epoch": 1.92, "learning_rate": 0.0004524034784009018, "loss": 1.4698, "step": 47790 }, { "epoch": 1.92, "learning_rate": 0.0004523934135834776, "loss": 1.4401, "step": 47800 }, { "epoch": 1.92, "learning_rate": 0.0004523833487660534, "loss": 1.6377, "step": 47810 }, { "epoch": 1.92, "learning_rate": 0.0004523732839486292, "loss": 1.4231, "step": 47820 }, { "epoch": 1.92, "learning_rate": 0.000452363219131205, "loss": 1.4548, "step": 47830 }, { "epoch": 1.92, "learning_rate": 0.0004523531543137808, "loss": 1.3288, "step": 47840 }, { "epoch": 1.92, "learning_rate": 0.0004523430894963566, "loss": 1.4661, "step": 47850 }, { "epoch": 1.92, "learning_rate": 0.0004523330246789323, "loss": 1.5043, "step": 47860 }, { "epoch": 1.93, "learning_rate": 0.0004523229598615081, "loss": 1.4234, "step": 47870 }, { "epoch": 1.93, "learning_rate": 0.0004523128950440839, "loss": 1.6619, "step": 47880 }, { "epoch": 1.93, "learning_rate": 0.0004523028302266597, "loss": 1.3334, "step": 47890 }, { "epoch": 1.93, "learning_rate": 0.0004522927654092355, "loss": 1.3357, "step": 47900 }, { "epoch": 1.93, "learning_rate": 0.00045228270059181127, "loss": 1.3601, "step": 47910 }, { "epoch": 1.93, "learning_rate": 0.00045227263577438706, "loss": 1.4999, "step": 47920 }, { "epoch": 1.93, "learning_rate": 0.00045226257095696285, "loss": 1.4803, "step": 47930 }, { "epoch": 1.93, "learning_rate": 0.0004522525061395386, "loss": 1.4617, "step": 47940 }, { "epoch": 1.93, "learning_rate": 0.00045224244132211443, "loss": 1.4535, "step": 47950 }, { "epoch": 1.93, "learning_rate": 0.0004522323765046902, "loss": 1.3526, "step": 47960 }, { "epoch": 1.93, "learning_rate": 0.000452222311687266, "loss": 1.235, "step": 47970 }, { "epoch": 1.93, "learning_rate": 0.0004522122468698418, "loss": 1.6589, "step": 47980 }, { "epoch": 1.93, "learning_rate": 0.0004522021820524176, "loss": 1.3584, "step": 47990 }, { "epoch": 1.93, "learning_rate": 0.0004521921172349934, "loss": 1.4518, "step": 48000 }, { "epoch": 1.93, "learning_rate": 0.0004521820524175692, "loss": 1.6237, "step": 48010 }, { "epoch": 1.93, "learning_rate": 0.00045217198760014497, "loss": 1.4016, "step": 48020 }, { "epoch": 1.93, "learning_rate": 0.0004521619227827207, "loss": 1.3696, "step": 48030 }, { "epoch": 1.93, "learning_rate": 0.0004521518579652965, "loss": 1.5607, "step": 48040 }, { "epoch": 1.93, "learning_rate": 0.0004521417931478723, "loss": 1.3193, "step": 48050 }, { "epoch": 1.93, "learning_rate": 0.0004521317283304481, "loss": 1.466, "step": 48060 }, { "epoch": 1.93, "learning_rate": 0.00045212166351302387, "loss": 1.5673, "step": 48070 }, { "epoch": 1.93, "learning_rate": 0.00045211159869559966, "loss": 1.3209, "step": 48080 }, { "epoch": 1.93, "learning_rate": 0.0004521015338781755, "loss": 1.7304, "step": 48090 }, { "epoch": 1.93, "learning_rate": 0.0004520914690607513, "loss": 1.3803, "step": 48100 }, { "epoch": 1.93, "learning_rate": 0.00045208140424332703, "loss": 1.7141, "step": 48110 }, { "epoch": 1.94, "learning_rate": 0.0004520713394259028, "loss": 1.145, "step": 48120 }, { "epoch": 1.94, "learning_rate": 0.0004520612746084786, "loss": 1.4325, "step": 48130 }, { "epoch": 1.94, "learning_rate": 0.0004520512097910544, "loss": 1.6921, "step": 48140 }, { "epoch": 1.94, "learning_rate": 0.0004520411449736302, "loss": 1.121, "step": 48150 }, { "epoch": 1.94, "learning_rate": 0.000452031080156206, "loss": 1.4955, "step": 48160 }, { "epoch": 1.94, "learning_rate": 0.0004520210153387818, "loss": 1.3311, "step": 48170 }, { "epoch": 1.94, "learning_rate": 0.00045201095052135756, "loss": 1.4325, "step": 48180 }, { "epoch": 1.94, "learning_rate": 0.00045200088570393335, "loss": 1.1332, "step": 48190 }, { "epoch": 1.94, "learning_rate": 0.0004519908208865091, "loss": 1.4945, "step": 48200 }, { "epoch": 1.94, "learning_rate": 0.0004519807560690849, "loss": 1.2792, "step": 48210 }, { "epoch": 1.94, "learning_rate": 0.00045197069125166067, "loss": 1.5696, "step": 48220 }, { "epoch": 1.94, "learning_rate": 0.00045196062643423646, "loss": 1.1216, "step": 48230 }, { "epoch": 1.94, "learning_rate": 0.0004519505616168123, "loss": 1.5381, "step": 48240 }, { "epoch": 1.94, "learning_rate": 0.0004519404967993881, "loss": 1.2068, "step": 48250 }, { "epoch": 1.94, "learning_rate": 0.0004519304319819639, "loss": 1.4692, "step": 48260 }, { "epoch": 1.94, "learning_rate": 0.0004519203671645397, "loss": 1.1952, "step": 48270 }, { "epoch": 1.94, "learning_rate": 0.0004519103023471154, "loss": 1.2414, "step": 48280 }, { "epoch": 1.94, "learning_rate": 0.0004519002375296912, "loss": 1.4459, "step": 48290 }, { "epoch": 1.94, "learning_rate": 0.000451890172712267, "loss": 1.5523, "step": 48300 }, { "epoch": 1.94, "learning_rate": 0.0004518801078948428, "loss": 1.0498, "step": 48310 }, { "epoch": 1.94, "learning_rate": 0.0004518700430774186, "loss": 1.3676, "step": 48320 }, { "epoch": 1.94, "learning_rate": 0.00045185997825999437, "loss": 1.48, "step": 48330 }, { "epoch": 1.94, "learning_rate": 0.00045184991344257016, "loss": 1.2817, "step": 48340 }, { "epoch": 1.94, "learning_rate": 0.00045183984862514595, "loss": 1.1568, "step": 48350 }, { "epoch": 1.94, "learning_rate": 0.00045182978380772174, "loss": 1.1178, "step": 48360 }, { "epoch": 1.95, "learning_rate": 0.0004518197189902975, "loss": 1.6523, "step": 48370 }, { "epoch": 1.95, "learning_rate": 0.0004518096541728733, "loss": 1.474, "step": 48380 }, { "epoch": 1.95, "learning_rate": 0.0004517995893554491, "loss": 1.5394, "step": 48390 }, { "epoch": 1.95, "learning_rate": 0.0004517895245380249, "loss": 1.4034, "step": 48400 }, { "epoch": 1.95, "learning_rate": 0.0004517794597206007, "loss": 1.2838, "step": 48410 }, { "epoch": 1.95, "learning_rate": 0.0004517693949031765, "loss": 1.3045, "step": 48420 }, { "epoch": 1.95, "learning_rate": 0.0004517593300857523, "loss": 1.4632, "step": 48430 }, { "epoch": 1.95, "learning_rate": 0.00045174926526832807, "loss": 1.4144, "step": 48440 }, { "epoch": 1.95, "learning_rate": 0.0004517392004509038, "loss": 1.1804, "step": 48450 }, { "epoch": 1.95, "learning_rate": 0.0004517291356334796, "loss": 1.3554, "step": 48460 }, { "epoch": 1.95, "learning_rate": 0.0004517190708160554, "loss": 1.2108, "step": 48470 }, { "epoch": 1.95, "learning_rate": 0.0004517090059986312, "loss": 1.0626, "step": 48480 }, { "epoch": 1.95, "learning_rate": 0.00045169894118120697, "loss": 1.4567, "step": 48490 }, { "epoch": 1.95, "learning_rate": 0.00045168887636378276, "loss": 1.2076, "step": 48500 }, { "epoch": 1.95, "learning_rate": 0.00045167881154635855, "loss": 1.395, "step": 48510 }, { "epoch": 1.95, "learning_rate": 0.00045166874672893434, "loss": 1.3869, "step": 48520 }, { "epoch": 1.95, "learning_rate": 0.0004516586819115102, "loss": 1.3178, "step": 48530 }, { "epoch": 1.95, "learning_rate": 0.0004516486170940859, "loss": 1.2324, "step": 48540 }, { "epoch": 1.95, "learning_rate": 0.0004516385522766617, "loss": 1.4271, "step": 48550 }, { "epoch": 1.95, "learning_rate": 0.0004516284874592375, "loss": 1.222, "step": 48560 }, { "epoch": 1.95, "learning_rate": 0.0004516184226418133, "loss": 1.2505, "step": 48570 }, { "epoch": 1.95, "learning_rate": 0.0004516083578243891, "loss": 1.3396, "step": 48580 }, { "epoch": 1.95, "learning_rate": 0.0004515982930069649, "loss": 1.3079, "step": 48590 }, { "epoch": 1.95, "learning_rate": 0.00045158822818954067, "loss": 1.353, "step": 48600 }, { "epoch": 1.96, "learning_rate": 0.00045157816337211646, "loss": 1.2913, "step": 48610 }, { "epoch": 1.96, "learning_rate": 0.0004515680985546922, "loss": 1.532, "step": 48620 }, { "epoch": 1.96, "learning_rate": 0.000451558033737268, "loss": 1.0972, "step": 48630 }, { "epoch": 1.96, "learning_rate": 0.0004515479689198438, "loss": 1.7614, "step": 48640 }, { "epoch": 1.96, "learning_rate": 0.00045153790410241957, "loss": 1.5906, "step": 48650 }, { "epoch": 1.96, "learning_rate": 0.00045152783928499536, "loss": 1.3568, "step": 48660 }, { "epoch": 1.96, "learning_rate": 0.00045151777446757115, "loss": 1.3295, "step": 48670 }, { "epoch": 1.96, "learning_rate": 0.000451507709650147, "loss": 1.3232, "step": 48680 }, { "epoch": 1.96, "learning_rate": 0.0004514976448327228, "loss": 1.5658, "step": 48690 }, { "epoch": 1.96, "learning_rate": 0.0004514875800152986, "loss": 1.718, "step": 48700 }, { "epoch": 1.96, "learning_rate": 0.0004514775151978743, "loss": 1.5614, "step": 48710 }, { "epoch": 1.96, "learning_rate": 0.0004514674503804501, "loss": 1.5651, "step": 48720 }, { "epoch": 1.96, "learning_rate": 0.0004514573855630259, "loss": 1.19, "step": 48730 }, { "epoch": 1.96, "learning_rate": 0.0004514473207456017, "loss": 1.3017, "step": 48740 }, { "epoch": 1.96, "learning_rate": 0.0004514372559281775, "loss": 1.6073, "step": 48750 }, { "epoch": 1.96, "learning_rate": 0.00045142719111075327, "loss": 1.372, "step": 48760 }, { "epoch": 1.96, "learning_rate": 0.00045141712629332906, "loss": 1.4823, "step": 48770 }, { "epoch": 1.96, "learning_rate": 0.00045140706147590485, "loss": 1.4223, "step": 48780 }, { "epoch": 1.96, "learning_rate": 0.0004513969966584806, "loss": 1.4399, "step": 48790 }, { "epoch": 1.96, "learning_rate": 0.0004513869318410564, "loss": 1.2727, "step": 48800 }, { "epoch": 1.96, "learning_rate": 0.00045137686702363217, "loss": 1.2672, "step": 48810 }, { "epoch": 1.96, "learning_rate": 0.000451366802206208, "loss": 1.4007, "step": 48820 }, { "epoch": 1.96, "learning_rate": 0.0004513567373887838, "loss": 1.3956, "step": 48830 }, { "epoch": 1.96, "learning_rate": 0.0004513466725713596, "loss": 1.4197, "step": 48840 }, { "epoch": 1.96, "learning_rate": 0.0004513366077539354, "loss": 1.4345, "step": 48850 }, { "epoch": 1.97, "learning_rate": 0.0004513265429365112, "loss": 1.3655, "step": 48860 }, { "epoch": 1.97, "learning_rate": 0.0004513164781190869, "loss": 1.1769, "step": 48870 }, { "epoch": 1.97, "learning_rate": 0.0004513064133016627, "loss": 1.2453, "step": 48880 }, { "epoch": 1.97, "learning_rate": 0.0004512963484842385, "loss": 1.5212, "step": 48890 }, { "epoch": 1.97, "learning_rate": 0.0004512862836668143, "loss": 1.1069, "step": 48900 }, { "epoch": 1.97, "learning_rate": 0.0004512762188493901, "loss": 1.297, "step": 48910 }, { "epoch": 1.97, "learning_rate": 0.00045126615403196586, "loss": 1.4267, "step": 48920 }, { "epoch": 1.97, "learning_rate": 0.00045125608921454166, "loss": 1.4856, "step": 48930 }, { "epoch": 1.97, "learning_rate": 0.00045124602439711745, "loss": 1.4194, "step": 48940 }, { "epoch": 1.97, "learning_rate": 0.00045123595957969324, "loss": 1.3462, "step": 48950 }, { "epoch": 1.97, "learning_rate": 0.000451225894762269, "loss": 1.0358, "step": 48960 }, { "epoch": 1.97, "learning_rate": 0.0004512158299448448, "loss": 1.2761, "step": 48970 }, { "epoch": 1.97, "learning_rate": 0.0004512057651274206, "loss": 1.2424, "step": 48980 }, { "epoch": 1.97, "learning_rate": 0.0004511957003099964, "loss": 1.2066, "step": 48990 }, { "epoch": 1.97, "learning_rate": 0.0004511856354925722, "loss": 1.3662, "step": 49000 }, { "epoch": 1.97, "learning_rate": 0.000451175570675148, "loss": 1.5081, "step": 49010 }, { "epoch": 1.97, "learning_rate": 0.00045116550585772377, "loss": 1.3012, "step": 49020 }, { "epoch": 1.97, "learning_rate": 0.00045115544104029956, "loss": 1.4885, "step": 49030 }, { "epoch": 1.97, "learning_rate": 0.0004511453762228753, "loss": 1.4723, "step": 49040 }, { "epoch": 1.97, "learning_rate": 0.0004511353114054511, "loss": 1.3875, "step": 49050 }, { "epoch": 1.97, "learning_rate": 0.0004511252465880269, "loss": 1.3469, "step": 49060 }, { "epoch": 1.97, "learning_rate": 0.00045111518177060267, "loss": 1.4569, "step": 49070 }, { "epoch": 1.97, "learning_rate": 0.00045110511695317846, "loss": 1.3205, "step": 49080 }, { "epoch": 1.97, "learning_rate": 0.00045109505213575425, "loss": 1.5516, "step": 49090 }, { "epoch": 1.97, "learning_rate": 0.00045108498731833004, "loss": 1.2548, "step": 49100 }, { "epoch": 1.98, "learning_rate": 0.00045107492250090584, "loss": 1.3317, "step": 49110 }, { "epoch": 1.98, "learning_rate": 0.0004510648576834817, "loss": 1.3727, "step": 49120 }, { "epoch": 1.98, "learning_rate": 0.0004510547928660574, "loss": 1.291, "step": 49130 }, { "epoch": 1.98, "learning_rate": 0.0004510447280486332, "loss": 1.386, "step": 49140 }, { "epoch": 1.98, "learning_rate": 0.000451034663231209, "loss": 1.4195, "step": 49150 }, { "epoch": 1.98, "learning_rate": 0.0004510245984137848, "loss": 1.5384, "step": 49160 }, { "epoch": 1.98, "learning_rate": 0.0004510145335963606, "loss": 1.3672, "step": 49170 }, { "epoch": 1.98, "learning_rate": 0.00045100446877893637, "loss": 1.4414, "step": 49180 }, { "epoch": 1.98, "learning_rate": 0.00045099440396151216, "loss": 1.6677, "step": 49190 }, { "epoch": 1.98, "learning_rate": 0.00045098433914408795, "loss": 1.3208, "step": 49200 }, { "epoch": 1.98, "learning_rate": 0.0004509742743266637, "loss": 1.3783, "step": 49210 }, { "epoch": 1.98, "learning_rate": 0.0004509642095092395, "loss": 1.5649, "step": 49220 }, { "epoch": 1.98, "learning_rate": 0.00045095414469181527, "loss": 1.587, "step": 49230 }, { "epoch": 1.98, "learning_rate": 0.00045094407987439106, "loss": 1.5869, "step": 49240 }, { "epoch": 1.98, "learning_rate": 0.00045093401505696685, "loss": 1.2418, "step": 49250 }, { "epoch": 1.98, "learning_rate": 0.0004509239502395427, "loss": 1.3612, "step": 49260 }, { "epoch": 1.98, "learning_rate": 0.0004509138854221185, "loss": 1.2955, "step": 49270 }, { "epoch": 1.98, "learning_rate": 0.0004509038206046943, "loss": 1.1217, "step": 49280 }, { "epoch": 1.98, "learning_rate": 0.00045089375578727007, "loss": 1.6974, "step": 49290 }, { "epoch": 1.98, "learning_rate": 0.0004508836909698458, "loss": 1.2139, "step": 49300 }, { "epoch": 1.98, "learning_rate": 0.0004508736261524216, "loss": 1.4889, "step": 49310 }, { "epoch": 1.98, "learning_rate": 0.0004508635613349974, "loss": 1.1958, "step": 49320 }, { "epoch": 1.98, "learning_rate": 0.0004508534965175732, "loss": 1.3976, "step": 49330 }, { "epoch": 1.98, "learning_rate": 0.00045084343170014897, "loss": 1.4294, "step": 49340 }, { "epoch": 1.98, "learning_rate": 0.00045083336688272476, "loss": 1.2763, "step": 49350 }, { "epoch": 1.99, "learning_rate": 0.00045082330206530055, "loss": 1.3162, "step": 49360 }, { "epoch": 1.99, "learning_rate": 0.00045081323724787634, "loss": 1.6841, "step": 49370 }, { "epoch": 1.99, "learning_rate": 0.0004508031724304521, "loss": 1.3497, "step": 49380 }, { "epoch": 1.99, "learning_rate": 0.00045079310761302787, "loss": 1.4399, "step": 49390 }, { "epoch": 1.99, "learning_rate": 0.00045078304279560366, "loss": 1.1128, "step": 49400 }, { "epoch": 1.99, "learning_rate": 0.0004507729779781795, "loss": 1.2639, "step": 49410 }, { "epoch": 1.99, "learning_rate": 0.0004507629131607553, "loss": 1.6651, "step": 49420 }, { "epoch": 1.99, "learning_rate": 0.0004507528483433311, "loss": 1.499, "step": 49430 }, { "epoch": 1.99, "learning_rate": 0.0004507427835259069, "loss": 1.3118, "step": 49440 }, { "epoch": 1.99, "learning_rate": 0.00045073271870848267, "loss": 1.3016, "step": 49450 }, { "epoch": 1.99, "learning_rate": 0.00045072265389105846, "loss": 1.3666, "step": 49460 }, { "epoch": 1.99, "learning_rate": 0.0004507125890736342, "loss": 1.3827, "step": 49470 }, { "epoch": 1.99, "learning_rate": 0.00045070252425621, "loss": 1.5074, "step": 49480 }, { "epoch": 1.99, "learning_rate": 0.0004506924594387858, "loss": 1.3027, "step": 49490 }, { "epoch": 1.99, "learning_rate": 0.00045068239462136157, "loss": 1.3606, "step": 49500 }, { "epoch": 1.99, "learning_rate": 0.00045067232980393736, "loss": 1.4945, "step": 49510 }, { "epoch": 1.99, "learning_rate": 0.00045066226498651315, "loss": 1.3869, "step": 49520 }, { "epoch": 1.99, "learning_rate": 0.00045065220016908894, "loss": 1.3251, "step": 49530 }, { "epoch": 1.99, "learning_rate": 0.00045064213535166473, "loss": 1.3124, "step": 49540 }, { "epoch": 1.99, "learning_rate": 0.0004506320705342405, "loss": 1.2682, "step": 49550 }, { "epoch": 1.99, "learning_rate": 0.0004506220057168163, "loss": 1.3506, "step": 49560 }, { "epoch": 1.99, "learning_rate": 0.0004506119408993921, "loss": 1.5518, "step": 49570 }, { "epoch": 1.99, "learning_rate": 0.0004506018760819679, "loss": 1.3974, "step": 49580 }, { "epoch": 1.99, "learning_rate": 0.0004505918112645437, "loss": 1.8569, "step": 49590 }, { "epoch": 1.99, "learning_rate": 0.0004505817464471195, "loss": 1.3021, "step": 49600 }, { "epoch": 2.0, "learning_rate": 0.00045057168162969527, "loss": 1.3834, "step": 49610 }, { "epoch": 2.0, "learning_rate": 0.00045056161681227106, "loss": 1.5737, "step": 49620 }, { "epoch": 2.0, "learning_rate": 0.0004505515519948468, "loss": 1.2319, "step": 49630 }, { "epoch": 2.0, "learning_rate": 0.0004505414871774226, "loss": 1.288, "step": 49640 }, { "epoch": 2.0, "learning_rate": 0.0004505314223599984, "loss": 1.6769, "step": 49650 }, { "epoch": 2.0, "learning_rate": 0.00045052135754257417, "loss": 1.512, "step": 49660 }, { "epoch": 2.0, "learning_rate": 0.00045051129272514996, "loss": 1.6431, "step": 49670 }, { "epoch": 2.0, "learning_rate": 0.00045050122790772575, "loss": 1.7044, "step": 49680 }, { "epoch": 2.0, "learning_rate": 0.00045049116309030154, "loss": 1.3799, "step": 49690 }, { "epoch": 2.0, "learning_rate": 0.0004504810982728774, "loss": 1.2493, "step": 49700 }, { "epoch": 2.0, "learning_rate": 0.0004504710334554532, "loss": 1.1867, "step": 49710 }, { "epoch": 2.0, "learning_rate": 0.0004504609686380289, "loss": 1.3953, "step": 49720 }, { "epoch": 2.0, "eval_Bleu_1": 0.013846765773380228, "eval_Bleu_2": 1.6245114600178832e-11, "eval_Bleu_3": 1.921868811980634e-14, "eval_Bleu_4": 7.137164172650854e-16, "eval_ROUGE_L": 0.03422804346656579, "eval_cer": 0.7392262352150006, "eval_em": 0.278552029731275, "eval_f1": 0.38559647991883905, "eval_loss": 1.7965617179870605, "eval_runtime": 2165.2556, "eval_samples_per_second": 12.924, "eval_steps_per_second": 4.308, "eval_wer": 0.8919522675207149, "step": 49728 }, { "epoch": 2.0, "learning_rate": 0.0004504509038206047, "loss": 1.6205, "step": 49730 }, { "epoch": 2.0, "learning_rate": 0.0004504408390031805, "loss": 1.1266, "step": 49740 }, { "epoch": 2.0, "learning_rate": 0.0004504307741857563, "loss": 1.4678, "step": 49750 }, { "epoch": 2.0, "learning_rate": 0.0004504207093683321, "loss": 1.2468, "step": 49760 }, { "epoch": 2.0, "learning_rate": 0.00045041064455090786, "loss": 1.1428, "step": 49770 }, { "epoch": 2.0, "learning_rate": 0.00045040057973348366, "loss": 1.0923, "step": 49780 }, { "epoch": 2.0, "learning_rate": 0.00045039051491605945, "loss": 1.2329, "step": 49790 }, { "epoch": 2.0, "learning_rate": 0.0004503804500986352, "loss": 1.1864, "step": 49800 }, { "epoch": 2.0, "learning_rate": 0.000450370385281211, "loss": 1.0864, "step": 49810 }, { "epoch": 2.0, "learning_rate": 0.00045036032046378676, "loss": 1.1989, "step": 49820 }, { "epoch": 2.0, "learning_rate": 0.00045035025564636256, "loss": 0.8668, "step": 49830 }, { "epoch": 2.0, "learning_rate": 0.00045034019082893835, "loss": 1.2617, "step": 49840 }, { "epoch": 2.0, "learning_rate": 0.0004503301260115142, "loss": 1.0186, "step": 49850 }, { "epoch": 2.01, "learning_rate": 0.00045032006119409, "loss": 1.3171, "step": 49860 }, { "epoch": 2.01, "learning_rate": 0.00045030999637666577, "loss": 1.1091, "step": 49870 }, { "epoch": 2.01, "learning_rate": 0.00045029993155924156, "loss": 1.2498, "step": 49880 }, { "epoch": 2.01, "learning_rate": 0.0004502898667418173, "loss": 1.2317, "step": 49890 }, { "epoch": 2.01, "learning_rate": 0.0004502798019243931, "loss": 1.2259, "step": 49900 }, { "epoch": 2.01, "learning_rate": 0.0004502697371069689, "loss": 1.1933, "step": 49910 }, { "epoch": 2.01, "learning_rate": 0.00045025967228954467, "loss": 1.3167, "step": 49920 }, { "epoch": 2.01, "learning_rate": 0.00045024960747212046, "loss": 1.29, "step": 49930 }, { "epoch": 2.01, "learning_rate": 0.00045023954265469625, "loss": 1.2762, "step": 49940 }, { "epoch": 2.01, "learning_rate": 0.00045022947783727204, "loss": 1.1693, "step": 49950 }, { "epoch": 2.01, "learning_rate": 0.00045021941301984784, "loss": 1.1846, "step": 49960 }, { "epoch": 2.01, "learning_rate": 0.00045020934820242357, "loss": 1.167, "step": 49970 }, { "epoch": 2.01, "learning_rate": 0.00045019928338499936, "loss": 1.0637, "step": 49980 }, { "epoch": 2.01, "learning_rate": 0.0004501892185675752, "loss": 1.387, "step": 49990 }, { "epoch": 2.01, "learning_rate": 0.000450179153750151, "loss": 1.1966, "step": 50000 }, { "epoch": 2.01, "learning_rate": 0.0004501690889327268, "loss": 1.0535, "step": 50010 }, { "epoch": 2.01, "learning_rate": 0.0004501590241153026, "loss": 1.1615, "step": 50020 }, { "epoch": 2.01, "learning_rate": 0.00045014895929787837, "loss": 1.2297, "step": 50030 }, { "epoch": 2.01, "learning_rate": 0.00045013889448045416, "loss": 1.2597, "step": 50040 }, { "epoch": 2.01, "learning_rate": 0.00045012882966302995, "loss": 1.2284, "step": 50050 }, { "epoch": 2.01, "learning_rate": 0.0004501187648456057, "loss": 0.8308, "step": 50060 }, { "epoch": 2.01, "learning_rate": 0.0004501087000281815, "loss": 1.3199, "step": 50070 }, { "epoch": 2.01, "learning_rate": 0.00045009863521075727, "loss": 1.178, "step": 50080 }, { "epoch": 2.01, "learning_rate": 0.00045008857039333306, "loss": 1.0795, "step": 50090 }, { "epoch": 2.01, "learning_rate": 0.00045007850557590885, "loss": 1.295, "step": 50100 }, { "epoch": 2.02, "learning_rate": 0.00045006844075848464, "loss": 1.2713, "step": 50110 }, { "epoch": 2.02, "learning_rate": 0.00045005837594106043, "loss": 1.0505, "step": 50120 }, { "epoch": 2.02, "learning_rate": 0.0004500483111236362, "loss": 1.1193, "step": 50130 }, { "epoch": 2.02, "learning_rate": 0.000450038246306212, "loss": 0.9743, "step": 50140 }, { "epoch": 2.02, "learning_rate": 0.0004500281814887878, "loss": 1.2922, "step": 50150 }, { "epoch": 2.02, "learning_rate": 0.0004500181166713636, "loss": 1.1802, "step": 50160 }, { "epoch": 2.02, "learning_rate": 0.0004500080518539394, "loss": 1.2226, "step": 50170 }, { "epoch": 2.02, "learning_rate": 0.0004499979870365152, "loss": 1.2664, "step": 50180 }, { "epoch": 2.02, "learning_rate": 0.00044998792221909097, "loss": 1.3732, "step": 50190 }, { "epoch": 2.02, "learning_rate": 0.00044997785740166676, "loss": 1.4478, "step": 50200 }, { "epoch": 2.02, "learning_rate": 0.00044996779258424255, "loss": 1.2585, "step": 50210 }, { "epoch": 2.02, "learning_rate": 0.00044995772776681834, "loss": 1.1667, "step": 50220 }, { "epoch": 2.02, "learning_rate": 0.0004499476629493941, "loss": 1.2892, "step": 50230 }, { "epoch": 2.02, "learning_rate": 0.00044993759813196987, "loss": 1.1581, "step": 50240 }, { "epoch": 2.02, "learning_rate": 0.00044992753331454566, "loss": 1.2344, "step": 50250 }, { "epoch": 2.02, "learning_rate": 0.00044991746849712145, "loss": 1.1809, "step": 50260 }, { "epoch": 2.02, "learning_rate": 0.00044990740367969724, "loss": 1.2808, "step": 50270 }, { "epoch": 2.02, "learning_rate": 0.00044989733886227303, "loss": 1.0728, "step": 50280 }, { "epoch": 2.02, "learning_rate": 0.0004498872740448489, "loss": 0.9773, "step": 50290 }, { "epoch": 2.02, "learning_rate": 0.00044987720922742467, "loss": 1.1822, "step": 50300 }, { "epoch": 2.02, "learning_rate": 0.0004498671444100004, "loss": 1.1082, "step": 50310 }, { "epoch": 2.02, "learning_rate": 0.0004498570795925762, "loss": 1.3174, "step": 50320 }, { "epoch": 2.02, "learning_rate": 0.000449847014775152, "loss": 0.9729, "step": 50330 }, { "epoch": 2.02, "learning_rate": 0.0004498369499577278, "loss": 1.3286, "step": 50340 }, { "epoch": 2.02, "learning_rate": 0.00044982688514030357, "loss": 1.27, "step": 50350 }, { "epoch": 2.03, "learning_rate": 0.00044981682032287936, "loss": 1.2896, "step": 50360 }, { "epoch": 2.03, "learning_rate": 0.00044980675550545515, "loss": 1.2467, "step": 50370 }, { "epoch": 2.03, "learning_rate": 0.00044979669068803094, "loss": 1.3331, "step": 50380 }, { "epoch": 2.03, "learning_rate": 0.0004497866258706067, "loss": 1.048, "step": 50390 }, { "epoch": 2.03, "learning_rate": 0.00044977656105318247, "loss": 1.1818, "step": 50400 }, { "epoch": 2.03, "learning_rate": 0.00044976649623575826, "loss": 1.1982, "step": 50410 }, { "epoch": 2.03, "learning_rate": 0.00044975643141833405, "loss": 1.1695, "step": 50420 }, { "epoch": 2.03, "learning_rate": 0.0004497463666009099, "loss": 1.2976, "step": 50430 }, { "epoch": 2.03, "learning_rate": 0.0004497363017834857, "loss": 1.1217, "step": 50440 }, { "epoch": 2.03, "learning_rate": 0.0004497262369660615, "loss": 1.0488, "step": 50450 }, { "epoch": 2.03, "learning_rate": 0.00044971617214863727, "loss": 1.0623, "step": 50460 }, { "epoch": 2.03, "learning_rate": 0.00044970610733121306, "loss": 1.1909, "step": 50470 }, { "epoch": 2.03, "learning_rate": 0.0004496960425137888, "loss": 0.9597, "step": 50480 }, { "epoch": 2.03, "learning_rate": 0.0004496859776963646, "loss": 1.0205, "step": 50490 }, { "epoch": 2.03, "learning_rate": 0.0004496759128789404, "loss": 1.3001, "step": 50500 }, { "epoch": 2.03, "learning_rate": 0.00044966584806151617, "loss": 1.1604, "step": 50510 }, { "epoch": 2.03, "learning_rate": 0.00044965578324409196, "loss": 1.119, "step": 50520 }, { "epoch": 2.03, "learning_rate": 0.00044964571842666775, "loss": 1.2212, "step": 50530 }, { "epoch": 2.03, "learning_rate": 0.00044963565360924354, "loss": 1.0153, "step": 50540 }, { "epoch": 2.03, "learning_rate": 0.00044962558879181933, "loss": 1.114, "step": 50550 }, { "epoch": 2.03, "learning_rate": 0.00044961552397439507, "loss": 1.0624, "step": 50560 }, { "epoch": 2.03, "learning_rate": 0.00044960545915697086, "loss": 0.9448, "step": 50570 }, { "epoch": 2.03, "learning_rate": 0.0004495953943395467, "loss": 1.1315, "step": 50580 }, { "epoch": 2.03, "learning_rate": 0.0004495853295221225, "loss": 1.1908, "step": 50590 }, { "epoch": 2.04, "learning_rate": 0.0004495752647046983, "loss": 1.1549, "step": 50600 }, { "epoch": 2.04, "learning_rate": 0.0004495651998872741, "loss": 1.1178, "step": 50610 }, { "epoch": 2.04, "learning_rate": 0.00044955513506984986, "loss": 1.208, "step": 50620 }, { "epoch": 2.04, "learning_rate": 0.00044954507025242566, "loss": 1.1551, "step": 50630 }, { "epoch": 2.04, "learning_rate": 0.00044953500543500145, "loss": 1.272, "step": 50640 }, { "epoch": 2.04, "learning_rate": 0.0004495249406175772, "loss": 0.8433, "step": 50650 }, { "epoch": 2.04, "learning_rate": 0.000449514875800153, "loss": 1.2803, "step": 50660 }, { "epoch": 2.04, "learning_rate": 0.00044950481098272876, "loss": 1.0602, "step": 50670 }, { "epoch": 2.04, "learning_rate": 0.00044949474616530455, "loss": 1.136, "step": 50680 }, { "epoch": 2.04, "learning_rate": 0.00044948468134788035, "loss": 1.0518, "step": 50690 }, { "epoch": 2.04, "learning_rate": 0.00044947461653045614, "loss": 1.2251, "step": 50700 }, { "epoch": 2.04, "learning_rate": 0.00044946455171303193, "loss": 1.4781, "step": 50710 }, { "epoch": 2.04, "learning_rate": 0.00044945448689560777, "loss": 1.1917, "step": 50720 }, { "epoch": 2.04, "learning_rate": 0.0004494444220781835, "loss": 1.1033, "step": 50730 }, { "epoch": 2.04, "learning_rate": 0.0004494343572607593, "loss": 1.2484, "step": 50740 }, { "epoch": 2.04, "learning_rate": 0.0004494242924433351, "loss": 1.1654, "step": 50750 }, { "epoch": 2.04, "learning_rate": 0.0004494142276259109, "loss": 1.2169, "step": 50760 }, { "epoch": 2.04, "learning_rate": 0.00044940416280848667, "loss": 1.253, "step": 50770 }, { "epoch": 2.04, "learning_rate": 0.00044939409799106246, "loss": 1.1377, "step": 50780 }, { "epoch": 2.04, "learning_rate": 0.00044938403317363825, "loss": 1.4181, "step": 50790 }, { "epoch": 2.04, "learning_rate": 0.00044937396835621404, "loss": 0.9349, "step": 50800 }, { "epoch": 2.04, "learning_rate": 0.00044936390353878984, "loss": 1.1418, "step": 50810 }, { "epoch": 2.04, "learning_rate": 0.00044935383872136557, "loss": 1.1994, "step": 50820 }, { "epoch": 2.04, "learning_rate": 0.00044934377390394136, "loss": 1.1757, "step": 50830 }, { "epoch": 2.04, "learning_rate": 0.00044933370908651715, "loss": 1.3737, "step": 50840 }, { "epoch": 2.05, "learning_rate": 0.00044932364426909294, "loss": 1.1106, "step": 50850 }, { "epoch": 2.05, "learning_rate": 0.00044931357945166873, "loss": 1.3317, "step": 50860 }, { "epoch": 2.05, "learning_rate": 0.0004493035146342446, "loss": 1.0548, "step": 50870 }, { "epoch": 2.05, "learning_rate": 0.00044929344981682037, "loss": 1.1853, "step": 50880 }, { "epoch": 2.05, "learning_rate": 0.00044928338499939616, "loss": 1.0989, "step": 50890 }, { "epoch": 2.05, "learning_rate": 0.0004492733201819719, "loss": 0.9968, "step": 50900 }, { "epoch": 2.05, "learning_rate": 0.0004492632553645477, "loss": 1.0869, "step": 50910 }, { "epoch": 2.05, "learning_rate": 0.0004492531905471235, "loss": 1.07, "step": 50920 }, { "epoch": 2.05, "learning_rate": 0.00044924312572969927, "loss": 1.3113, "step": 50930 }, { "epoch": 2.05, "learning_rate": 0.00044923306091227506, "loss": 1.3032, "step": 50940 }, { "epoch": 2.05, "learning_rate": 0.00044922299609485085, "loss": 0.9993, "step": 50950 }, { "epoch": 2.05, "learning_rate": 0.00044921293127742664, "loss": 1.0708, "step": 50960 }, { "epoch": 2.05, "learning_rate": 0.00044920286646000243, "loss": 1.3955, "step": 50970 }, { "epoch": 2.05, "learning_rate": 0.0004491928016425782, "loss": 1.1759, "step": 50980 }, { "epoch": 2.05, "learning_rate": 0.00044918273682515396, "loss": 1.096, "step": 50990 }, { "epoch": 2.05, "learning_rate": 0.00044917267200772975, "loss": 1.2455, "step": 51000 }, { "epoch": 2.05, "learning_rate": 0.00044916260719030554, "loss": 1.2511, "step": 51010 }, { "epoch": 2.05, "learning_rate": 0.0004491525423728814, "loss": 1.073, "step": 51020 }, { "epoch": 2.05, "learning_rate": 0.0004491424775554572, "loss": 1.1505, "step": 51030 }, { "epoch": 2.05, "learning_rate": 0.00044913241273803297, "loss": 1.154, "step": 51040 }, { "epoch": 2.05, "learning_rate": 0.00044912234792060876, "loss": 1.2953, "step": 51050 }, { "epoch": 2.05, "learning_rate": 0.00044911228310318455, "loss": 1.145, "step": 51060 }, { "epoch": 2.05, "learning_rate": 0.0004491022182857603, "loss": 1.2478, "step": 51070 }, { "epoch": 2.05, "learning_rate": 0.0004490921534683361, "loss": 1.3662, "step": 51080 }, { "epoch": 2.05, "learning_rate": 0.00044908208865091187, "loss": 1.1016, "step": 51090 }, { "epoch": 2.06, "learning_rate": 0.00044907202383348766, "loss": 1.111, "step": 51100 }, { "epoch": 2.06, "learning_rate": 0.00044906195901606345, "loss": 0.9415, "step": 51110 }, { "epoch": 2.06, "learning_rate": 0.00044905189419863924, "loss": 1.1282, "step": 51120 }, { "epoch": 2.06, "learning_rate": 0.00044904182938121503, "loss": 1.177, "step": 51130 }, { "epoch": 2.06, "learning_rate": 0.0004490317645637908, "loss": 0.9936, "step": 51140 }, { "epoch": 2.06, "learning_rate": 0.00044902169974636656, "loss": 1.2128, "step": 51150 }, { "epoch": 2.06, "learning_rate": 0.0004490116349289424, "loss": 1.3362, "step": 51160 }, { "epoch": 2.06, "learning_rate": 0.0004490015701115182, "loss": 1.2366, "step": 51170 }, { "epoch": 2.06, "learning_rate": 0.000448991505294094, "loss": 1.2503, "step": 51180 }, { "epoch": 2.06, "learning_rate": 0.0004489814404766698, "loss": 1.3114, "step": 51190 }, { "epoch": 2.06, "learning_rate": 0.00044897137565924557, "loss": 1.2567, "step": 51200 }, { "epoch": 2.06, "learning_rate": 0.00044896131084182136, "loss": 1.2209, "step": 51210 }, { "epoch": 2.06, "learning_rate": 0.00044895124602439715, "loss": 1.0145, "step": 51220 }, { "epoch": 2.06, "learning_rate": 0.00044894118120697294, "loss": 0.9764, "step": 51230 }, { "epoch": 2.06, "learning_rate": 0.0004489311163895487, "loss": 1.0538, "step": 51240 }, { "epoch": 2.06, "learning_rate": 0.00044892105157212447, "loss": 1.2075, "step": 51250 }, { "epoch": 2.06, "learning_rate": 0.00044891098675470026, "loss": 1.203, "step": 51260 }, { "epoch": 2.06, "learning_rate": 0.00044890092193727605, "loss": 1.1901, "step": 51270 }, { "epoch": 2.06, "learning_rate": 0.00044889085711985184, "loss": 1.2884, "step": 51280 }, { "epoch": 2.06, "learning_rate": 0.00044888079230242763, "loss": 1.0544, "step": 51290 }, { "epoch": 2.06, "learning_rate": 0.0004488707274850034, "loss": 1.0681, "step": 51300 }, { "epoch": 2.06, "learning_rate": 0.00044886066266757927, "loss": 1.3334, "step": 51310 }, { "epoch": 2.06, "learning_rate": 0.000448850597850155, "loss": 1.4082, "step": 51320 }, { "epoch": 2.06, "learning_rate": 0.0004488405330327308, "loss": 1.1598, "step": 51330 }, { "epoch": 2.06, "learning_rate": 0.0004488304682153066, "loss": 1.3022, "step": 51340 }, { "epoch": 2.07, "learning_rate": 0.0004488204033978824, "loss": 1.3521, "step": 51350 }, { "epoch": 2.07, "learning_rate": 0.00044881033858045817, "loss": 1.2016, "step": 51360 }, { "epoch": 2.07, "learning_rate": 0.00044880027376303396, "loss": 1.4464, "step": 51370 }, { "epoch": 2.07, "learning_rate": 0.00044879020894560975, "loss": 1.2255, "step": 51380 }, { "epoch": 2.07, "learning_rate": 0.00044878014412818554, "loss": 1.1473, "step": 51390 }, { "epoch": 2.07, "learning_rate": 0.00044877007931076133, "loss": 1.4618, "step": 51400 }, { "epoch": 2.07, "learning_rate": 0.00044876001449333707, "loss": 1.145, "step": 51410 }, { "epoch": 2.07, "learning_rate": 0.00044874994967591286, "loss": 1.2294, "step": 51420 }, { "epoch": 2.07, "learning_rate": 0.00044873988485848865, "loss": 1.1821, "step": 51430 }, { "epoch": 2.07, "learning_rate": 0.00044872982004106444, "loss": 1.0731, "step": 51440 }, { "epoch": 2.07, "learning_rate": 0.0004487197552236403, "loss": 1.3178, "step": 51450 }, { "epoch": 2.07, "learning_rate": 0.0004487096904062161, "loss": 1.2789, "step": 51460 }, { "epoch": 2.07, "learning_rate": 0.00044869962558879186, "loss": 1.1218, "step": 51470 }, { "epoch": 2.07, "learning_rate": 0.00044868956077136765, "loss": 1.4436, "step": 51480 }, { "epoch": 2.07, "learning_rate": 0.0004486794959539434, "loss": 1.0019, "step": 51490 }, { "epoch": 2.07, "learning_rate": 0.0004486694311365192, "loss": 1.1003, "step": 51500 }, { "epoch": 2.07, "learning_rate": 0.000448659366319095, "loss": 1.1142, "step": 51510 }, { "epoch": 2.07, "learning_rate": 0.00044864930150167076, "loss": 1.0045, "step": 51520 }, { "epoch": 2.07, "learning_rate": 0.00044863923668424655, "loss": 0.9833, "step": 51530 }, { "epoch": 2.07, "learning_rate": 0.00044862917186682235, "loss": 1.2384, "step": 51540 }, { "epoch": 2.07, "learning_rate": 0.00044861910704939814, "loss": 1.1758, "step": 51550 }, { "epoch": 2.07, "learning_rate": 0.0004486090422319739, "loss": 1.1495, "step": 51560 }, { "epoch": 2.07, "learning_rate": 0.0004485989774145497, "loss": 1.2142, "step": 51570 }, { "epoch": 2.07, "learning_rate": 0.00044858891259712545, "loss": 1.4206, "step": 51580 }, { "epoch": 2.07, "learning_rate": 0.00044857884777970125, "loss": 1.1072, "step": 51590 }, { "epoch": 2.08, "learning_rate": 0.0004485687829622771, "loss": 1.1774, "step": 51600 }, { "epoch": 2.08, "learning_rate": 0.0004485587181448529, "loss": 0.9278, "step": 51610 }, { "epoch": 2.08, "learning_rate": 0.00044854865332742867, "loss": 1.1686, "step": 51620 }, { "epoch": 2.08, "learning_rate": 0.00044853858851000446, "loss": 1.2468, "step": 51630 }, { "epoch": 2.08, "learning_rate": 0.00044852852369258025, "loss": 1.0165, "step": 51640 }, { "epoch": 2.08, "learning_rate": 0.00044851845887515604, "loss": 1.2273, "step": 51650 }, { "epoch": 2.08, "learning_rate": 0.0004485083940577318, "loss": 1.1271, "step": 51660 }, { "epoch": 2.08, "learning_rate": 0.00044849832924030757, "loss": 1.2254, "step": 51670 }, { "epoch": 2.08, "learning_rate": 0.00044848826442288336, "loss": 1.2904, "step": 51680 }, { "epoch": 2.08, "learning_rate": 0.00044847819960545915, "loss": 1.1302, "step": 51690 }, { "epoch": 2.08, "learning_rate": 0.00044846813478803494, "loss": 1.22, "step": 51700 }, { "epoch": 2.08, "learning_rate": 0.00044845806997061073, "loss": 1.2618, "step": 51710 }, { "epoch": 2.08, "learning_rate": 0.0004484480051531865, "loss": 1.1551, "step": 51720 }, { "epoch": 2.08, "learning_rate": 0.0004484379403357623, "loss": 1.3175, "step": 51730 }, { "epoch": 2.08, "learning_rate": 0.0004484278755183381, "loss": 1.228, "step": 51740 }, { "epoch": 2.08, "learning_rate": 0.0004484178107009139, "loss": 1.2115, "step": 51750 }, { "epoch": 2.08, "learning_rate": 0.0004484077458834897, "loss": 1.2541, "step": 51760 }, { "epoch": 2.08, "learning_rate": 0.0004483976810660655, "loss": 1.18, "step": 51770 }, { "epoch": 2.08, "learning_rate": 0.00044838761624864127, "loss": 1.2686, "step": 51780 }, { "epoch": 2.08, "learning_rate": 0.00044837755143121706, "loss": 1.251, "step": 51790 }, { "epoch": 2.08, "learning_rate": 0.00044836748661379285, "loss": 1.2917, "step": 51800 }, { "epoch": 2.08, "learning_rate": 0.00044835742179636864, "loss": 1.1793, "step": 51810 }, { "epoch": 2.08, "learning_rate": 0.00044834735697894443, "loss": 1.1705, "step": 51820 }, { "epoch": 2.08, "learning_rate": 0.00044833729216152017, "loss": 1.4586, "step": 51830 }, { "epoch": 2.08, "learning_rate": 0.00044832722734409596, "loss": 1.1776, "step": 51840 }, { "epoch": 2.09, "learning_rate": 0.00044831716252667175, "loss": 1.3747, "step": 51850 }, { "epoch": 2.09, "learning_rate": 0.00044830709770924754, "loss": 1.1583, "step": 51860 }, { "epoch": 2.09, "learning_rate": 0.00044829703289182333, "loss": 1.0821, "step": 51870 }, { "epoch": 2.09, "learning_rate": 0.0004482869680743991, "loss": 1.0099, "step": 51880 }, { "epoch": 2.09, "learning_rate": 0.00044827690325697497, "loss": 1.2788, "step": 51890 }, { "epoch": 2.09, "learning_rate": 0.00044826683843955076, "loss": 1.148, "step": 51900 }, { "epoch": 2.09, "learning_rate": 0.00044825677362212655, "loss": 1.269, "step": 51910 }, { "epoch": 2.09, "learning_rate": 0.0004482467088047023, "loss": 1.1437, "step": 51920 }, { "epoch": 2.09, "learning_rate": 0.0004482366439872781, "loss": 1.0251, "step": 51930 }, { "epoch": 2.09, "learning_rate": 0.00044822657916985387, "loss": 1.0343, "step": 51940 }, { "epoch": 2.09, "learning_rate": 0.00044821651435242966, "loss": 1.2487, "step": 51950 }, { "epoch": 2.09, "learning_rate": 0.00044820644953500545, "loss": 0.9714, "step": 51960 }, { "epoch": 2.09, "learning_rate": 0.00044819638471758124, "loss": 1.1463, "step": 51970 }, { "epoch": 2.09, "learning_rate": 0.00044818631990015703, "loss": 1.3226, "step": 51980 }, { "epoch": 2.09, "learning_rate": 0.0004481762550827328, "loss": 1.1998, "step": 51990 }, { "epoch": 2.09, "learning_rate": 0.00044816619026530856, "loss": 1.2461, "step": 52000 }, { "epoch": 2.09, "learning_rate": 0.00044815612544788435, "loss": 1.1825, "step": 52010 }, { "epoch": 2.09, "learning_rate": 0.00044814606063046014, "loss": 1.371, "step": 52020 }, { "epoch": 2.09, "learning_rate": 0.00044813599581303593, "loss": 1.1397, "step": 52030 }, { "epoch": 2.09, "learning_rate": 0.0004481259309956118, "loss": 1.3722, "step": 52040 }, { "epoch": 2.09, "learning_rate": 0.00044811586617818757, "loss": 0.9709, "step": 52050 }, { "epoch": 2.09, "learning_rate": 0.00044810580136076336, "loss": 1.3152, "step": 52060 }, { "epoch": 2.09, "learning_rate": 0.00044809573654333915, "loss": 1.292, "step": 52070 }, { "epoch": 2.09, "learning_rate": 0.0004480856717259149, "loss": 1.267, "step": 52080 }, { "epoch": 2.09, "learning_rate": 0.0004480756069084907, "loss": 1.0776, "step": 52090 }, { "epoch": 2.1, "learning_rate": 0.00044806554209106647, "loss": 1.3998, "step": 52100 }, { "epoch": 2.1, "learning_rate": 0.00044805547727364226, "loss": 1.1939, "step": 52110 }, { "epoch": 2.1, "learning_rate": 0.00044804541245621805, "loss": 1.1229, "step": 52120 }, { "epoch": 2.1, "learning_rate": 0.00044803534763879384, "loss": 1.0738, "step": 52130 }, { "epoch": 2.1, "learning_rate": 0.00044802528282136963, "loss": 1.2511, "step": 52140 }, { "epoch": 2.1, "learning_rate": 0.0004480152180039454, "loss": 1.3636, "step": 52150 }, { "epoch": 2.1, "learning_rate": 0.0004480051531865212, "loss": 1.2592, "step": 52160 }, { "epoch": 2.1, "learning_rate": 0.00044799508836909695, "loss": 1.0642, "step": 52170 }, { "epoch": 2.1, "learning_rate": 0.00044798502355167274, "loss": 1.1884, "step": 52180 }, { "epoch": 2.1, "learning_rate": 0.0004479749587342486, "loss": 1.2202, "step": 52190 }, { "epoch": 2.1, "learning_rate": 0.0004479648939168244, "loss": 1.3638, "step": 52200 }, { "epoch": 2.1, "learning_rate": 0.00044795482909940017, "loss": 1.2183, "step": 52210 }, { "epoch": 2.1, "learning_rate": 0.00044794476428197596, "loss": 1.1075, "step": 52220 }, { "epoch": 2.1, "learning_rate": 0.00044793469946455175, "loss": 1.0668, "step": 52230 }, { "epoch": 2.1, "learning_rate": 0.00044792463464712754, "loss": 1.1307, "step": 52240 }, { "epoch": 2.1, "learning_rate": 0.0004479145698297033, "loss": 1.1264, "step": 52250 }, { "epoch": 2.1, "learning_rate": 0.00044790450501227907, "loss": 1.3124, "step": 52260 }, { "epoch": 2.1, "learning_rate": 0.00044789444019485486, "loss": 1.1448, "step": 52270 }, { "epoch": 2.1, "learning_rate": 0.00044788437537743065, "loss": 1.2274, "step": 52280 }, { "epoch": 2.1, "learning_rate": 0.00044787431056000644, "loss": 1.3263, "step": 52290 }, { "epoch": 2.1, "learning_rate": 0.00044786424574258223, "loss": 1.3173, "step": 52300 }, { "epoch": 2.1, "learning_rate": 0.000447854180925158, "loss": 1.458, "step": 52310 }, { "epoch": 2.1, "learning_rate": 0.0004478441161077338, "loss": 1.3342, "step": 52320 }, { "epoch": 2.1, "learning_rate": 0.00044783405129030965, "loss": 1.3609, "step": 52330 }, { "epoch": 2.11, "learning_rate": 0.0004478239864728854, "loss": 1.2132, "step": 52340 }, { "epoch": 2.11, "learning_rate": 0.0004478139216554612, "loss": 1.3034, "step": 52350 }, { "epoch": 2.11, "learning_rate": 0.00044780385683803697, "loss": 1.1915, "step": 52360 }, { "epoch": 2.11, "learning_rate": 0.00044779379202061276, "loss": 0.9709, "step": 52370 }, { "epoch": 2.11, "learning_rate": 0.00044778372720318855, "loss": 1.1271, "step": 52380 }, { "epoch": 2.11, "learning_rate": 0.00044777366238576435, "loss": 1.3064, "step": 52390 }, { "epoch": 2.11, "learning_rate": 0.00044776359756834014, "loss": 1.2366, "step": 52400 }, { "epoch": 2.11, "learning_rate": 0.0004477535327509159, "loss": 1.2302, "step": 52410 }, { "epoch": 2.11, "learning_rate": 0.00044774346793349166, "loss": 1.4909, "step": 52420 }, { "epoch": 2.11, "learning_rate": 0.00044773340311606745, "loss": 1.1728, "step": 52430 }, { "epoch": 2.11, "learning_rate": 0.00044772333829864324, "loss": 1.3352, "step": 52440 }, { "epoch": 2.11, "learning_rate": 0.00044771327348121904, "loss": 1.0241, "step": 52450 }, { "epoch": 2.11, "learning_rate": 0.0004477032086637948, "loss": 1.0289, "step": 52460 }, { "epoch": 2.11, "learning_rate": 0.0004476931438463706, "loss": 1.4416, "step": 52470 }, { "epoch": 2.11, "learning_rate": 0.00044768307902894646, "loss": 1.1627, "step": 52480 }, { "epoch": 2.11, "learning_rate": 0.00044767301421152225, "loss": 1.0509, "step": 52490 }, { "epoch": 2.11, "learning_rate": 0.00044766294939409804, "loss": 1.0189, "step": 52500 }, { "epoch": 2.11, "learning_rate": 0.0004476528845766738, "loss": 1.2384, "step": 52510 }, { "epoch": 2.11, "learning_rate": 0.00044764281975924957, "loss": 1.3945, "step": 52520 }, { "epoch": 2.11, "learning_rate": 0.00044763275494182536, "loss": 1.2282, "step": 52530 }, { "epoch": 2.11, "learning_rate": 0.00044762269012440115, "loss": 1.0943, "step": 52540 }, { "epoch": 2.11, "learning_rate": 0.00044761262530697694, "loss": 1.1646, "step": 52550 }, { "epoch": 2.11, "learning_rate": 0.00044760256048955273, "loss": 1.1433, "step": 52560 }, { "epoch": 2.11, "learning_rate": 0.0004475924956721285, "loss": 1.2235, "step": 52570 }, { "epoch": 2.11, "learning_rate": 0.0004475824308547043, "loss": 1.3669, "step": 52580 }, { "epoch": 2.12, "learning_rate": 0.00044757236603728005, "loss": 1.2687, "step": 52590 }, { "epoch": 2.12, "learning_rate": 0.00044756230121985584, "loss": 1.2606, "step": 52600 }, { "epoch": 2.12, "learning_rate": 0.00044755223640243163, "loss": 1.6072, "step": 52610 }, { "epoch": 2.12, "learning_rate": 0.0004475421715850075, "loss": 1.0735, "step": 52620 }, { "epoch": 2.12, "learning_rate": 0.00044753210676758327, "loss": 1.2345, "step": 52630 }, { "epoch": 2.12, "learning_rate": 0.00044752204195015906, "loss": 1.0919, "step": 52640 }, { "epoch": 2.12, "learning_rate": 0.00044751197713273485, "loss": 1.3699, "step": 52650 }, { "epoch": 2.12, "learning_rate": 0.00044750191231531064, "loss": 1.2555, "step": 52660 }, { "epoch": 2.12, "learning_rate": 0.00044749184749788643, "loss": 1.0787, "step": 52670 }, { "epoch": 2.12, "learning_rate": 0.00044748178268046217, "loss": 1.3718, "step": 52680 }, { "epoch": 2.12, "learning_rate": 0.00044747171786303796, "loss": 1.2493, "step": 52690 }, { "epoch": 2.12, "learning_rate": 0.00044746165304561375, "loss": 1.2606, "step": 52700 }, { "epoch": 2.12, "learning_rate": 0.00044745158822818954, "loss": 1.278, "step": 52710 }, { "epoch": 2.12, "learning_rate": 0.00044744152341076533, "loss": 1.0956, "step": 52720 }, { "epoch": 2.12, "learning_rate": 0.0004474314585933411, "loss": 1.0961, "step": 52730 }, { "epoch": 2.12, "learning_rate": 0.0004474213937759169, "loss": 1.2913, "step": 52740 }, { "epoch": 2.12, "learning_rate": 0.0004474113289584927, "loss": 1.1834, "step": 52750 }, { "epoch": 2.12, "learning_rate": 0.00044740126414106844, "loss": 1.2844, "step": 52760 }, { "epoch": 2.12, "learning_rate": 0.0004473911993236443, "loss": 1.2843, "step": 52770 }, { "epoch": 2.12, "learning_rate": 0.0004473811345062201, "loss": 1.0257, "step": 52780 }, { "epoch": 2.12, "learning_rate": 0.00044737106968879587, "loss": 1.6532, "step": 52790 }, { "epoch": 2.12, "learning_rate": 0.00044736100487137166, "loss": 1.1227, "step": 52800 }, { "epoch": 2.12, "learning_rate": 0.00044735094005394745, "loss": 1.1332, "step": 52810 }, { "epoch": 2.12, "learning_rate": 0.00044734087523652324, "loss": 1.0617, "step": 52820 }, { "epoch": 2.12, "learning_rate": 0.00044733081041909903, "loss": 1.1921, "step": 52830 }, { "epoch": 2.13, "learning_rate": 0.00044732074560167477, "loss": 1.265, "step": 52840 }, { "epoch": 2.13, "learning_rate": 0.00044731068078425056, "loss": 1.1171, "step": 52850 }, { "epoch": 2.13, "learning_rate": 0.00044730061596682635, "loss": 1.2406, "step": 52860 }, { "epoch": 2.13, "learning_rate": 0.00044729055114940214, "loss": 1.267, "step": 52870 }, { "epoch": 2.13, "learning_rate": 0.00044728048633197793, "loss": 1.0524, "step": 52880 }, { "epoch": 2.13, "learning_rate": 0.0004472704215145537, "loss": 1.2899, "step": 52890 }, { "epoch": 2.13, "learning_rate": 0.0004472603566971295, "loss": 1.1865, "step": 52900 }, { "epoch": 2.13, "learning_rate": 0.0004472502918797053, "loss": 1.273, "step": 52910 }, { "epoch": 2.13, "learning_rate": 0.00044724022706228115, "loss": 1.256, "step": 52920 }, { "epoch": 2.13, "learning_rate": 0.0004472301622448569, "loss": 1.2706, "step": 52930 }, { "epoch": 2.13, "learning_rate": 0.0004472200974274327, "loss": 1.3009, "step": 52940 }, { "epoch": 2.13, "learning_rate": 0.00044721003261000847, "loss": 1.3894, "step": 52950 }, { "epoch": 2.13, "learning_rate": 0.00044719996779258426, "loss": 0.9921, "step": 52960 }, { "epoch": 2.13, "learning_rate": 0.00044718990297516005, "loss": 1.1056, "step": 52970 }, { "epoch": 2.13, "learning_rate": 0.00044717983815773584, "loss": 1.0192, "step": 52980 }, { "epoch": 2.13, "learning_rate": 0.00044716977334031163, "loss": 1.2835, "step": 52990 }, { "epoch": 2.13, "learning_rate": 0.0004471597085228874, "loss": 1.0351, "step": 53000 }, { "epoch": 2.13, "learning_rate": 0.00044714964370546316, "loss": 0.9965, "step": 53010 }, { "epoch": 2.13, "learning_rate": 0.00044713957888803895, "loss": 1.3192, "step": 53020 }, { "epoch": 2.13, "learning_rate": 0.00044712951407061474, "loss": 0.9403, "step": 53030 }, { "epoch": 2.13, "learning_rate": 0.00044711944925319053, "loss": 1.0406, "step": 53040 }, { "epoch": 2.13, "learning_rate": 0.0004471093844357663, "loss": 1.2251, "step": 53050 }, { "epoch": 2.13, "learning_rate": 0.00044709931961834217, "loss": 1.0067, "step": 53060 }, { "epoch": 2.13, "learning_rate": 0.00044708925480091796, "loss": 0.9438, "step": 53070 }, { "epoch": 2.13, "learning_rate": 0.00044707918998349375, "loss": 1.2472, "step": 53080 }, { "epoch": 2.14, "learning_rate": 0.00044706912516606954, "loss": 1.2003, "step": 53090 }, { "epoch": 2.14, "learning_rate": 0.0004470590603486453, "loss": 1.109, "step": 53100 }, { "epoch": 2.14, "learning_rate": 0.00044704899553122106, "loss": 1.1368, "step": 53110 }, { "epoch": 2.14, "learning_rate": 0.00044703893071379686, "loss": 1.1753, "step": 53120 }, { "epoch": 2.14, "learning_rate": 0.00044702886589637265, "loss": 1.2108, "step": 53130 }, { "epoch": 2.14, "learning_rate": 0.00044701880107894844, "loss": 1.2537, "step": 53140 }, { "epoch": 2.14, "learning_rate": 0.00044700873626152423, "loss": 1.1239, "step": 53150 }, { "epoch": 2.14, "learning_rate": 0.0004469986714441, "loss": 1.1356, "step": 53160 }, { "epoch": 2.14, "learning_rate": 0.0004469886066266758, "loss": 1.5232, "step": 53170 }, { "epoch": 2.14, "learning_rate": 0.00044697854180925155, "loss": 1.1717, "step": 53180 }, { "epoch": 2.14, "learning_rate": 0.00044696847699182734, "loss": 1.1356, "step": 53190 }, { "epoch": 2.14, "learning_rate": 0.00044695841217440313, "loss": 1.1322, "step": 53200 }, { "epoch": 2.14, "learning_rate": 0.00044694834735697897, "loss": 1.1619, "step": 53210 }, { "epoch": 2.14, "learning_rate": 0.00044693828253955476, "loss": 1.261, "step": 53220 }, { "epoch": 2.14, "learning_rate": 0.00044692821772213055, "loss": 1.2757, "step": 53230 }, { "epoch": 2.14, "learning_rate": 0.00044691815290470634, "loss": 1.3041, "step": 53240 }, { "epoch": 2.14, "learning_rate": 0.00044690808808728214, "loss": 1.3934, "step": 53250 }, { "epoch": 2.14, "learning_rate": 0.0004468980232698579, "loss": 1.2056, "step": 53260 }, { "epoch": 2.14, "learning_rate": 0.00044688795845243366, "loss": 1.2554, "step": 53270 }, { "epoch": 2.14, "learning_rate": 0.00044687789363500945, "loss": 1.0204, "step": 53280 }, { "epoch": 2.14, "learning_rate": 0.00044686782881758524, "loss": 1.3034, "step": 53290 }, { "epoch": 2.14, "learning_rate": 0.00044685776400016104, "loss": 1.1105, "step": 53300 }, { "epoch": 2.14, "learning_rate": 0.0004468476991827368, "loss": 1.4237, "step": 53310 }, { "epoch": 2.14, "learning_rate": 0.0004468376343653126, "loss": 1.1365, "step": 53320 }, { "epoch": 2.14, "learning_rate": 0.0004468275695478884, "loss": 1.4022, "step": 53330 }, { "epoch": 2.15, "learning_rate": 0.0004468175047304642, "loss": 1.1695, "step": 53340 }, { "epoch": 2.15, "learning_rate": 0.00044680743991303994, "loss": 1.1153, "step": 53350 }, { "epoch": 2.15, "learning_rate": 0.0004467973750956158, "loss": 1.1638, "step": 53360 }, { "epoch": 2.15, "learning_rate": 0.00044678731027819157, "loss": 0.9517, "step": 53370 }, { "epoch": 2.15, "learning_rate": 0.00044677724546076736, "loss": 1.2021, "step": 53380 }, { "epoch": 2.15, "learning_rate": 0.00044676718064334315, "loss": 1.109, "step": 53390 }, { "epoch": 2.15, "learning_rate": 0.00044675711582591894, "loss": 1.2508, "step": 53400 }, { "epoch": 2.15, "learning_rate": 0.00044674705100849473, "loss": 1.0082, "step": 53410 }, { "epoch": 2.15, "learning_rate": 0.0004467369861910705, "loss": 1.0341, "step": 53420 }, { "epoch": 2.15, "learning_rate": 0.0004467269213736463, "loss": 1.3095, "step": 53430 }, { "epoch": 2.15, "learning_rate": 0.00044671685655622205, "loss": 1.1441, "step": 53440 }, { "epoch": 2.15, "learning_rate": 0.00044670679173879784, "loss": 1.3398, "step": 53450 }, { "epoch": 2.15, "learning_rate": 0.00044669672692137363, "loss": 1.2456, "step": 53460 }, { "epoch": 2.15, "learning_rate": 0.0004466866621039494, "loss": 1.0498, "step": 53470 }, { "epoch": 2.15, "learning_rate": 0.0004466765972865252, "loss": 1.2448, "step": 53480 }, { "epoch": 2.15, "learning_rate": 0.000446666532469101, "loss": 1.1939, "step": 53490 }, { "epoch": 2.15, "learning_rate": 0.00044665646765167685, "loss": 1.2199, "step": 53500 }, { "epoch": 2.15, "learning_rate": 0.00044664640283425264, "loss": 1.0659, "step": 53510 }, { "epoch": 2.15, "learning_rate": 0.0004466363380168284, "loss": 1.1424, "step": 53520 }, { "epoch": 2.15, "learning_rate": 0.00044662627319940417, "loss": 0.9782, "step": 53530 }, { "epoch": 2.15, "learning_rate": 0.00044661620838197996, "loss": 1.2637, "step": 53540 }, { "epoch": 2.15, "learning_rate": 0.00044660614356455575, "loss": 1.1248, "step": 53550 }, { "epoch": 2.15, "learning_rate": 0.00044659607874713154, "loss": 1.3327, "step": 53560 }, { "epoch": 2.15, "learning_rate": 0.00044658601392970733, "loss": 1.157, "step": 53570 }, { "epoch": 2.15, "learning_rate": 0.0004465759491122831, "loss": 1.3933, "step": 53580 }, { "epoch": 2.16, "learning_rate": 0.0004465658842948589, "loss": 1.4448, "step": 53590 }, { "epoch": 2.16, "learning_rate": 0.00044655581947743465, "loss": 1.0891, "step": 53600 }, { "epoch": 2.16, "learning_rate": 0.00044654575466001044, "loss": 1.1877, "step": 53610 }, { "epoch": 2.16, "learning_rate": 0.00044653568984258623, "loss": 1.5196, "step": 53620 }, { "epoch": 2.16, "learning_rate": 0.000446525625025162, "loss": 1.2579, "step": 53630 }, { "epoch": 2.16, "learning_rate": 0.0004465155602077378, "loss": 1.426, "step": 53640 }, { "epoch": 2.16, "learning_rate": 0.00044650549539031366, "loss": 0.9541, "step": 53650 }, { "epoch": 2.16, "learning_rate": 0.00044649543057288945, "loss": 1.1549, "step": 53660 }, { "epoch": 2.16, "learning_rate": 0.00044648536575546524, "loss": 1.1597, "step": 53670 }, { "epoch": 2.16, "learning_rate": 0.00044647530093804103, "loss": 1.2276, "step": 53680 }, { "epoch": 2.16, "learning_rate": 0.00044646523612061677, "loss": 1.1797, "step": 53690 }, { "epoch": 2.16, "learning_rate": 0.00044645517130319256, "loss": 1.1148, "step": 53700 }, { "epoch": 2.16, "learning_rate": 0.00044644510648576835, "loss": 1.2412, "step": 53710 }, { "epoch": 2.16, "learning_rate": 0.00044643504166834414, "loss": 1.2852, "step": 53720 }, { "epoch": 2.16, "learning_rate": 0.00044642497685091993, "loss": 1.2113, "step": 53730 }, { "epoch": 2.16, "learning_rate": 0.0004464149120334957, "loss": 1.2588, "step": 53740 }, { "epoch": 2.16, "learning_rate": 0.0004464048472160715, "loss": 1.2423, "step": 53750 }, { "epoch": 2.16, "learning_rate": 0.0004463947823986473, "loss": 1.3991, "step": 53760 }, { "epoch": 2.16, "learning_rate": 0.00044638471758122304, "loss": 1.241, "step": 53770 }, { "epoch": 2.16, "learning_rate": 0.00044637465276379883, "loss": 1.284, "step": 53780 }, { "epoch": 2.16, "learning_rate": 0.0004463645879463747, "loss": 1.3864, "step": 53790 }, { "epoch": 2.16, "learning_rate": 0.00044635452312895047, "loss": 1.3536, "step": 53800 }, { "epoch": 2.16, "learning_rate": 0.00044634445831152626, "loss": 1.2651, "step": 53810 }, { "epoch": 2.16, "learning_rate": 0.00044633439349410205, "loss": 1.4844, "step": 53820 }, { "epoch": 2.16, "learning_rate": 0.00044632432867667784, "loss": 1.0497, "step": 53830 }, { "epoch": 2.17, "learning_rate": 0.00044631426385925363, "loss": 1.1331, "step": 53840 }, { "epoch": 2.17, "learning_rate": 0.0004463041990418294, "loss": 1.2628, "step": 53850 }, { "epoch": 2.17, "learning_rate": 0.00044629413422440516, "loss": 1.4144, "step": 53860 }, { "epoch": 2.17, "learning_rate": 0.00044628406940698095, "loss": 1.0768, "step": 53870 }, { "epoch": 2.17, "learning_rate": 0.00044627400458955674, "loss": 1.2915, "step": 53880 }, { "epoch": 2.17, "learning_rate": 0.00044626393977213253, "loss": 1.2972, "step": 53890 }, { "epoch": 2.17, "learning_rate": 0.0004462538749547083, "loss": 1.4624, "step": 53900 }, { "epoch": 2.17, "learning_rate": 0.0004462438101372841, "loss": 1.0928, "step": 53910 }, { "epoch": 2.17, "learning_rate": 0.0004462337453198599, "loss": 1.1139, "step": 53920 }, { "epoch": 2.17, "learning_rate": 0.0004462236805024357, "loss": 1.3086, "step": 53930 }, { "epoch": 2.17, "learning_rate": 0.0004462136156850115, "loss": 1.1203, "step": 53940 }, { "epoch": 2.17, "learning_rate": 0.0004462035508675873, "loss": 1.1395, "step": 53950 }, { "epoch": 2.17, "learning_rate": 0.00044619348605016306, "loss": 1.4049, "step": 53960 }, { "epoch": 2.17, "learning_rate": 0.00044618342123273886, "loss": 1.3097, "step": 53970 }, { "epoch": 2.17, "learning_rate": 0.00044617335641531465, "loss": 1.1087, "step": 53980 }, { "epoch": 2.17, "learning_rate": 0.00044616329159789044, "loss": 1.0867, "step": 53990 }, { "epoch": 2.17, "learning_rate": 0.00044615322678046623, "loss": 1.3317, "step": 54000 }, { "epoch": 2.17, "learning_rate": 0.000446143161963042, "loss": 1.0168, "step": 54010 }, { "epoch": 2.17, "learning_rate": 0.0004461330971456178, "loss": 1.2424, "step": 54020 }, { "epoch": 2.17, "learning_rate": 0.00044612303232819355, "loss": 1.0953, "step": 54030 }, { "epoch": 2.17, "learning_rate": 0.00044611296751076934, "loss": 1.0363, "step": 54040 }, { "epoch": 2.17, "learning_rate": 0.00044610290269334513, "loss": 1.013, "step": 54050 }, { "epoch": 2.17, "learning_rate": 0.0004460928378759209, "loss": 1.2045, "step": 54060 }, { "epoch": 2.17, "learning_rate": 0.0004460827730584967, "loss": 1.6176, "step": 54070 }, { "epoch": 2.18, "learning_rate": 0.0004460727082410725, "loss": 1.1945, "step": 54080 }, { "epoch": 2.18, "learning_rate": 0.00044606264342364834, "loss": 1.3262, "step": 54090 }, { "epoch": 2.18, "learning_rate": 0.00044605257860622414, "loss": 1.2586, "step": 54100 }, { "epoch": 2.18, "learning_rate": 0.00044604251378879987, "loss": 1.1956, "step": 54110 }, { "epoch": 2.18, "learning_rate": 0.00044603244897137566, "loss": 1.2982, "step": 54120 }, { "epoch": 2.18, "learning_rate": 0.00044602238415395145, "loss": 1.3473, "step": 54130 }, { "epoch": 2.18, "learning_rate": 0.00044601231933652724, "loss": 1.3402, "step": 54140 }, { "epoch": 2.18, "learning_rate": 0.00044600225451910304, "loss": 1.1366, "step": 54150 }, { "epoch": 2.18, "learning_rate": 0.0004459921897016788, "loss": 1.1992, "step": 54160 }, { "epoch": 2.18, "learning_rate": 0.0004459821248842546, "loss": 0.9987, "step": 54170 }, { "epoch": 2.18, "learning_rate": 0.0004459720600668304, "loss": 1.2758, "step": 54180 }, { "epoch": 2.18, "learning_rate": 0.0004459619952494062, "loss": 1.3639, "step": 54190 }, { "epoch": 2.18, "learning_rate": 0.00044595193043198193, "loss": 1.2675, "step": 54200 }, { "epoch": 2.18, "learning_rate": 0.0004459418656145577, "loss": 1.0625, "step": 54210 }, { "epoch": 2.18, "learning_rate": 0.0004459318007971335, "loss": 1.2187, "step": 54220 }, { "epoch": 2.18, "learning_rate": 0.00044592173597970936, "loss": 1.2842, "step": 54230 }, { "epoch": 2.18, "learning_rate": 0.00044591167116228515, "loss": 1.1204, "step": 54240 }, { "epoch": 2.18, "learning_rate": 0.00044590160634486094, "loss": 1.2295, "step": 54250 }, { "epoch": 2.18, "learning_rate": 0.00044589154152743673, "loss": 1.3341, "step": 54260 }, { "epoch": 2.18, "learning_rate": 0.0004458814767100125, "loss": 0.9573, "step": 54270 }, { "epoch": 2.18, "learning_rate": 0.00044587141189258826, "loss": 1.5228, "step": 54280 }, { "epoch": 2.18, "learning_rate": 0.00044586134707516405, "loss": 1.1209, "step": 54290 }, { "epoch": 2.18, "learning_rate": 0.00044585128225773984, "loss": 1.2621, "step": 54300 }, { "epoch": 2.18, "learning_rate": 0.00044584121744031563, "loss": 1.5072, "step": 54310 }, { "epoch": 2.18, "learning_rate": 0.0004458311526228914, "loss": 1.0475, "step": 54320 }, { "epoch": 2.19, "learning_rate": 0.0004458210878054672, "loss": 1.0641, "step": 54330 }, { "epoch": 2.19, "learning_rate": 0.000445811022988043, "loss": 1.3832, "step": 54340 }, { "epoch": 2.19, "learning_rate": 0.0004458009581706188, "loss": 1.2984, "step": 54350 }, { "epoch": 2.19, "learning_rate": 0.00044579089335319453, "loss": 1.4062, "step": 54360 }, { "epoch": 2.19, "learning_rate": 0.0004457808285357703, "loss": 1.2121, "step": 54370 }, { "epoch": 2.19, "learning_rate": 0.00044577076371834617, "loss": 1.338, "step": 54380 }, { "epoch": 2.19, "learning_rate": 0.00044576069890092196, "loss": 1.0896, "step": 54390 }, { "epoch": 2.19, "learning_rate": 0.00044575063408349775, "loss": 1.4131, "step": 54400 }, { "epoch": 2.19, "learning_rate": 0.00044574056926607354, "loss": 1.1885, "step": 54410 }, { "epoch": 2.19, "learning_rate": 0.00044573050444864933, "loss": 1.2767, "step": 54420 }, { "epoch": 2.19, "learning_rate": 0.0004457204396312251, "loss": 0.9818, "step": 54430 }, { "epoch": 2.19, "learning_rate": 0.0004457103748138009, "loss": 1.1702, "step": 54440 }, { "epoch": 2.19, "learning_rate": 0.00044570030999637665, "loss": 1.332, "step": 54450 }, { "epoch": 2.19, "learning_rate": 0.00044569024517895244, "loss": 1.2449, "step": 54460 }, { "epoch": 2.19, "learning_rate": 0.00044568018036152823, "loss": 1.1096, "step": 54470 }, { "epoch": 2.19, "learning_rate": 0.000445670115544104, "loss": 1.0629, "step": 54480 }, { "epoch": 2.19, "learning_rate": 0.0004456600507266798, "loss": 1.3275, "step": 54490 }, { "epoch": 2.19, "learning_rate": 0.0004456499859092556, "loss": 1.2089, "step": 54500 }, { "epoch": 2.19, "learning_rate": 0.0004456399210918314, "loss": 1.2279, "step": 54510 }, { "epoch": 2.19, "learning_rate": 0.0004456298562744072, "loss": 1.1322, "step": 54520 }, { "epoch": 2.19, "learning_rate": 0.000445619791456983, "loss": 1.2785, "step": 54530 }, { "epoch": 2.19, "learning_rate": 0.00044560972663955877, "loss": 1.1706, "step": 54540 }, { "epoch": 2.19, "learning_rate": 0.00044559966182213456, "loss": 1.2794, "step": 54550 }, { "epoch": 2.19, "learning_rate": 0.00044558959700471035, "loss": 1.1946, "step": 54560 }, { "epoch": 2.19, "learning_rate": 0.00044557953218728614, "loss": 1.1437, "step": 54570 }, { "epoch": 2.2, "learning_rate": 0.00044556946736986193, "loss": 1.258, "step": 54580 }, { "epoch": 2.2, "learning_rate": 0.0004455594025524377, "loss": 1.3146, "step": 54590 }, { "epoch": 2.2, "learning_rate": 0.0004455493377350135, "loss": 1.1528, "step": 54600 }, { "epoch": 2.2, "learning_rate": 0.0004455392729175893, "loss": 1.1198, "step": 54610 }, { "epoch": 2.2, "learning_rate": 0.00044552920810016504, "loss": 1.209, "step": 54620 }, { "epoch": 2.2, "learning_rate": 0.00044551914328274083, "loss": 1.408, "step": 54630 }, { "epoch": 2.2, "learning_rate": 0.0004455090784653166, "loss": 0.9091, "step": 54640 }, { "epoch": 2.2, "learning_rate": 0.0004454990136478924, "loss": 1.2353, "step": 54650 }, { "epoch": 2.2, "learning_rate": 0.0004454889488304682, "loss": 1.3471, "step": 54660 }, { "epoch": 2.2, "learning_rate": 0.00044547888401304405, "loss": 1.4187, "step": 54670 }, { "epoch": 2.2, "learning_rate": 0.00044546881919561984, "loss": 1.2161, "step": 54680 }, { "epoch": 2.2, "learning_rate": 0.00044545875437819563, "loss": 1.036, "step": 54690 }, { "epoch": 2.2, "learning_rate": 0.00044544868956077137, "loss": 1.4212, "step": 54700 }, { "epoch": 2.2, "learning_rate": 0.00044543862474334716, "loss": 1.1816, "step": 54710 }, { "epoch": 2.2, "learning_rate": 0.00044542855992592295, "loss": 1.1577, "step": 54720 }, { "epoch": 2.2, "learning_rate": 0.00044541849510849874, "loss": 1.3248, "step": 54730 }, { "epoch": 2.2, "learning_rate": 0.00044540843029107453, "loss": 1.0874, "step": 54740 }, { "epoch": 2.2, "learning_rate": 0.0004453983654736503, "loss": 1.4524, "step": 54750 }, { "epoch": 2.2, "learning_rate": 0.0004453883006562261, "loss": 1.1489, "step": 54760 }, { "epoch": 2.2, "learning_rate": 0.0004453782358388019, "loss": 1.3165, "step": 54770 }, { "epoch": 2.2, "learning_rate": 0.0004453681710213777, "loss": 1.1327, "step": 54780 }, { "epoch": 2.2, "learning_rate": 0.00044535810620395343, "loss": 1.1231, "step": 54790 }, { "epoch": 2.2, "learning_rate": 0.0004453480413865292, "loss": 1.0247, "step": 54800 }, { "epoch": 2.2, "learning_rate": 0.000445337976569105, "loss": 1.2493, "step": 54810 }, { "epoch": 2.2, "learning_rate": 0.00044532791175168086, "loss": 1.195, "step": 54820 }, { "epoch": 2.21, "learning_rate": 0.00044531784693425665, "loss": 1.151, "step": 54830 }, { "epoch": 2.21, "learning_rate": 0.00044530778211683244, "loss": 1.1072, "step": 54840 }, { "epoch": 2.21, "learning_rate": 0.00044529771729940823, "loss": 1.2331, "step": 54850 }, { "epoch": 2.21, "learning_rate": 0.000445287652481984, "loss": 1.2923, "step": 54860 }, { "epoch": 2.21, "learning_rate": 0.00044527758766455975, "loss": 1.3905, "step": 54870 }, { "epoch": 2.21, "learning_rate": 0.00044526752284713555, "loss": 1.1764, "step": 54880 }, { "epoch": 2.21, "learning_rate": 0.00044525745802971134, "loss": 1.1577, "step": 54890 }, { "epoch": 2.21, "learning_rate": 0.0004452473932122871, "loss": 1.3533, "step": 54900 }, { "epoch": 2.21, "learning_rate": 0.0004452373283948629, "loss": 1.0977, "step": 54910 }, { "epoch": 2.21, "learning_rate": 0.0004452272635774387, "loss": 1.441, "step": 54920 }, { "epoch": 2.21, "learning_rate": 0.0004452171987600145, "loss": 1.2625, "step": 54930 }, { "epoch": 2.21, "learning_rate": 0.0004452071339425903, "loss": 1.1383, "step": 54940 }, { "epoch": 2.21, "learning_rate": 0.0004451970691251661, "loss": 1.0203, "step": 54950 }, { "epoch": 2.21, "learning_rate": 0.00044518700430774187, "loss": 1.5457, "step": 54960 }, { "epoch": 2.21, "learning_rate": 0.00044517693949031766, "loss": 1.1905, "step": 54970 }, { "epoch": 2.21, "learning_rate": 0.00044516687467289345, "loss": 1.1101, "step": 54980 }, { "epoch": 2.21, "learning_rate": 0.00044515680985546924, "loss": 1.1863, "step": 54990 }, { "epoch": 2.21, "learning_rate": 0.00044514674503804503, "loss": 1.2521, "step": 55000 }, { "epoch": 2.21, "learning_rate": 0.0004451366802206208, "loss": 1.1549, "step": 55010 }, { "epoch": 2.21, "learning_rate": 0.0004451266154031966, "loss": 0.9971, "step": 55020 }, { "epoch": 2.21, "learning_rate": 0.0004451165505857724, "loss": 0.9349, "step": 55030 }, { "epoch": 2.21, "learning_rate": 0.00044510648576834814, "loss": 1.1409, "step": 55040 }, { "epoch": 2.21, "learning_rate": 0.00044509642095092393, "loss": 1.2704, "step": 55050 }, { "epoch": 2.21, "learning_rate": 0.0004450863561334997, "loss": 1.1611, "step": 55060 }, { "epoch": 2.21, "learning_rate": 0.0004450762913160755, "loss": 1.3019, "step": 55070 }, { "epoch": 2.22, "learning_rate": 0.0004450662264986513, "loss": 1.1225, "step": 55080 }, { "epoch": 2.22, "learning_rate": 0.0004450561616812271, "loss": 1.4679, "step": 55090 }, { "epoch": 2.22, "learning_rate": 0.0004450460968638029, "loss": 1.1694, "step": 55100 }, { "epoch": 2.22, "learning_rate": 0.00044503603204637873, "loss": 1.2158, "step": 55110 }, { "epoch": 2.22, "learning_rate": 0.0004450259672289545, "loss": 1.1522, "step": 55120 }, { "epoch": 2.22, "learning_rate": 0.00044501590241153026, "loss": 1.4366, "step": 55130 }, { "epoch": 2.22, "learning_rate": 0.00044500583759410605, "loss": 1.256, "step": 55140 }, { "epoch": 2.22, "learning_rate": 0.00044499577277668184, "loss": 1.5357, "step": 55150 }, { "epoch": 2.22, "learning_rate": 0.00044498570795925763, "loss": 1.0908, "step": 55160 }, { "epoch": 2.22, "learning_rate": 0.0004449756431418334, "loss": 1.305, "step": 55170 }, { "epoch": 2.22, "learning_rate": 0.0004449655783244092, "loss": 1.4287, "step": 55180 }, { "epoch": 2.22, "learning_rate": 0.000444955513506985, "loss": 1.2394, "step": 55190 }, { "epoch": 2.22, "learning_rate": 0.0004449454486895608, "loss": 1.441, "step": 55200 }, { "epoch": 2.22, "learning_rate": 0.00044493538387213653, "loss": 1.1587, "step": 55210 }, { "epoch": 2.22, "learning_rate": 0.0004449253190547123, "loss": 1.3986, "step": 55220 }, { "epoch": 2.22, "learning_rate": 0.0004449152542372881, "loss": 1.2471, "step": 55230 }, { "epoch": 2.22, "learning_rate": 0.0004449051894198639, "loss": 1.1619, "step": 55240 }, { "epoch": 2.22, "learning_rate": 0.0004448951246024397, "loss": 0.9993, "step": 55250 }, { "epoch": 2.22, "learning_rate": 0.00044488505978501554, "loss": 1.2574, "step": 55260 }, { "epoch": 2.22, "learning_rate": 0.00044487499496759133, "loss": 1.2947, "step": 55270 }, { "epoch": 2.22, "learning_rate": 0.0004448649301501671, "loss": 1.2547, "step": 55280 }, { "epoch": 2.22, "learning_rate": 0.00044485486533274286, "loss": 1.2242, "step": 55290 }, { "epoch": 2.22, "learning_rate": 0.00044484480051531865, "loss": 1.1863, "step": 55300 }, { "epoch": 2.22, "learning_rate": 0.00044483473569789444, "loss": 1.0317, "step": 55310 }, { "epoch": 2.22, "learning_rate": 0.00044482467088047023, "loss": 1.2926, "step": 55320 }, { "epoch": 2.23, "learning_rate": 0.000444814606063046, "loss": 1.286, "step": 55330 }, { "epoch": 2.23, "learning_rate": 0.0004448045412456218, "loss": 1.2351, "step": 55340 }, { "epoch": 2.23, "learning_rate": 0.0004447944764281976, "loss": 1.3114, "step": 55350 }, { "epoch": 2.23, "learning_rate": 0.0004447844116107734, "loss": 1.0828, "step": 55360 }, { "epoch": 2.23, "learning_rate": 0.0004447743467933492, "loss": 1.2953, "step": 55370 }, { "epoch": 2.23, "learning_rate": 0.0004447642819759249, "loss": 1.0531, "step": 55380 }, { "epoch": 2.23, "learning_rate": 0.0004447542171585007, "loss": 1.1461, "step": 55390 }, { "epoch": 2.23, "learning_rate": 0.00044474415234107656, "loss": 1.3014, "step": 55400 }, { "epoch": 2.23, "learning_rate": 0.00044473408752365235, "loss": 1.0327, "step": 55410 }, { "epoch": 2.23, "learning_rate": 0.00044472402270622814, "loss": 1.2079, "step": 55420 }, { "epoch": 2.23, "learning_rate": 0.00044471395788880393, "loss": 1.1386, "step": 55430 }, { "epoch": 2.23, "learning_rate": 0.0004447038930713797, "loss": 1.3542, "step": 55440 }, { "epoch": 2.23, "learning_rate": 0.0004446938282539555, "loss": 1.4391, "step": 55450 }, { "epoch": 2.23, "learning_rate": 0.00044468376343653125, "loss": 1.4656, "step": 55460 }, { "epoch": 2.23, "learning_rate": 0.00044467369861910704, "loss": 1.1196, "step": 55470 }, { "epoch": 2.23, "learning_rate": 0.00044466363380168283, "loss": 1.697, "step": 55480 }, { "epoch": 2.23, "learning_rate": 0.0004446535689842586, "loss": 1.0899, "step": 55490 }, { "epoch": 2.23, "learning_rate": 0.0004446435041668344, "loss": 1.2475, "step": 55500 }, { "epoch": 2.23, "learning_rate": 0.0004446334393494102, "loss": 1.1877, "step": 55510 }, { "epoch": 2.23, "learning_rate": 0.000444623374531986, "loss": 1.1415, "step": 55520 }, { "epoch": 2.23, "learning_rate": 0.0004446133097145618, "loss": 1.2522, "step": 55530 }, { "epoch": 2.23, "learning_rate": 0.0004446032448971376, "loss": 1.1531, "step": 55540 }, { "epoch": 2.23, "learning_rate": 0.00044459318007971337, "loss": 1.3869, "step": 55550 }, { "epoch": 2.23, "learning_rate": 0.00044458311526228916, "loss": 1.2067, "step": 55560 }, { "epoch": 2.23, "learning_rate": 0.00044457305044486495, "loss": 1.4163, "step": 55570 }, { "epoch": 2.24, "learning_rate": 0.00044456298562744074, "loss": 1.3871, "step": 55580 }, { "epoch": 2.24, "learning_rate": 0.00044455292081001653, "loss": 1.0558, "step": 55590 }, { "epoch": 2.24, "learning_rate": 0.0004445428559925923, "loss": 1.0669, "step": 55600 }, { "epoch": 2.24, "learning_rate": 0.0004445327911751681, "loss": 1.2751, "step": 55610 }, { "epoch": 2.24, "learning_rate": 0.0004445227263577439, "loss": 1.2678, "step": 55620 }, { "epoch": 2.24, "learning_rate": 0.00044451266154031964, "loss": 1.3135, "step": 55630 }, { "epoch": 2.24, "learning_rate": 0.00044450259672289543, "loss": 1.3047, "step": 55640 }, { "epoch": 2.24, "learning_rate": 0.0004444925319054712, "loss": 1.3571, "step": 55650 }, { "epoch": 2.24, "learning_rate": 0.000444482467088047, "loss": 1.3256, "step": 55660 }, { "epoch": 2.24, "learning_rate": 0.0004444724022706228, "loss": 1.2262, "step": 55670 }, { "epoch": 2.24, "learning_rate": 0.0004444623374531986, "loss": 1.4909, "step": 55680 }, { "epoch": 2.24, "learning_rate": 0.00044445227263577444, "loss": 1.5843, "step": 55690 }, { "epoch": 2.24, "learning_rate": 0.00044444220781835023, "loss": 1.3265, "step": 55700 }, { "epoch": 2.24, "learning_rate": 0.000444432143000926, "loss": 1.1686, "step": 55710 }, { "epoch": 2.24, "learning_rate": 0.00044442207818350175, "loss": 1.2614, "step": 55720 }, { "epoch": 2.24, "learning_rate": 0.00044441201336607755, "loss": 1.0662, "step": 55730 }, { "epoch": 2.24, "learning_rate": 0.00044440194854865334, "loss": 1.1983, "step": 55740 }, { "epoch": 2.24, "learning_rate": 0.0004443918837312291, "loss": 1.1833, "step": 55750 }, { "epoch": 2.24, "learning_rate": 0.0004443818189138049, "loss": 1.3675, "step": 55760 }, { "epoch": 2.24, "learning_rate": 0.0004443717540963807, "loss": 1.3453, "step": 55770 }, { "epoch": 2.24, "learning_rate": 0.0004443616892789565, "loss": 1.3571, "step": 55780 }, { "epoch": 2.24, "learning_rate": 0.0004443516244615323, "loss": 1.3968, "step": 55790 }, { "epoch": 2.24, "learning_rate": 0.000444341559644108, "loss": 1.2185, "step": 55800 }, { "epoch": 2.24, "learning_rate": 0.0004443314948266838, "loss": 1.0985, "step": 55810 }, { "epoch": 2.24, "learning_rate": 0.0004443214300092596, "loss": 1.2898, "step": 55820 }, { "epoch": 2.25, "learning_rate": 0.0004443113651918354, "loss": 1.3182, "step": 55830 }, { "epoch": 2.25, "learning_rate": 0.00044430130037441124, "loss": 1.2339, "step": 55840 }, { "epoch": 2.25, "learning_rate": 0.00044429123555698703, "loss": 1.1189, "step": 55850 }, { "epoch": 2.25, "learning_rate": 0.0004442811707395628, "loss": 1.0394, "step": 55860 }, { "epoch": 2.25, "learning_rate": 0.0004442711059221386, "loss": 1.1897, "step": 55870 }, { "epoch": 2.25, "learning_rate": 0.0004442610411047144, "loss": 1.4646, "step": 55880 }, { "epoch": 2.25, "learning_rate": 0.00044425097628729014, "loss": 1.3026, "step": 55890 }, { "epoch": 2.25, "learning_rate": 0.00044424091146986593, "loss": 1.2863, "step": 55900 }, { "epoch": 2.25, "learning_rate": 0.0004442308466524417, "loss": 1.2976, "step": 55910 }, { "epoch": 2.25, "learning_rate": 0.0004442207818350175, "loss": 1.0639, "step": 55920 }, { "epoch": 2.25, "learning_rate": 0.0004442107170175933, "loss": 1.2494, "step": 55930 }, { "epoch": 2.25, "learning_rate": 0.0004442006522001691, "loss": 1.3074, "step": 55940 }, { "epoch": 2.25, "learning_rate": 0.0004441905873827449, "loss": 1.2158, "step": 55950 }, { "epoch": 2.25, "learning_rate": 0.0004441805225653207, "loss": 1.4376, "step": 55960 }, { "epoch": 2.25, "learning_rate": 0.0004441704577478964, "loss": 1.2578, "step": 55970 }, { "epoch": 2.25, "learning_rate": 0.0004441603929304722, "loss": 1.1307, "step": 55980 }, { "epoch": 2.25, "learning_rate": 0.00044415032811304805, "loss": 1.2417, "step": 55990 }, { "epoch": 2.25, "learning_rate": 0.00044414026329562384, "loss": 1.2824, "step": 56000 }, { "epoch": 2.25, "learning_rate": 0.00044413019847819963, "loss": 1.1052, "step": 56010 }, { "epoch": 2.25, "learning_rate": 0.0004441201336607754, "loss": 1.2349, "step": 56020 }, { "epoch": 2.25, "learning_rate": 0.0004441100688433512, "loss": 1.0281, "step": 56030 }, { "epoch": 2.25, "learning_rate": 0.000444100004025927, "loss": 1.3616, "step": 56040 }, { "epoch": 2.25, "learning_rate": 0.00044408993920850274, "loss": 1.1912, "step": 56050 }, { "epoch": 2.25, "learning_rate": 0.00044407987439107853, "loss": 1.2266, "step": 56060 }, { "epoch": 2.26, "learning_rate": 0.0004440698095736543, "loss": 1.1873, "step": 56070 }, { "epoch": 2.26, "learning_rate": 0.0004440597447562301, "loss": 1.2263, "step": 56080 }, { "epoch": 2.26, "learning_rate": 0.0004440496799388059, "loss": 1.374, "step": 56090 }, { "epoch": 2.26, "learning_rate": 0.0004440396151213817, "loss": 1.1156, "step": 56100 }, { "epoch": 2.26, "learning_rate": 0.0004440295503039575, "loss": 1.3303, "step": 56110 }, { "epoch": 2.26, "learning_rate": 0.0004440194854865333, "loss": 1.2066, "step": 56120 }, { "epoch": 2.26, "learning_rate": 0.0004440094206691091, "loss": 1.3561, "step": 56130 }, { "epoch": 2.26, "learning_rate": 0.00044399935585168486, "loss": 1.414, "step": 56140 }, { "epoch": 2.26, "learning_rate": 0.00044398929103426065, "loss": 1.2228, "step": 56150 }, { "epoch": 2.26, "learning_rate": 0.00044397922621683644, "loss": 1.4442, "step": 56160 }, { "epoch": 2.26, "learning_rate": 0.00044396916139941223, "loss": 1.1246, "step": 56170 }, { "epoch": 2.26, "learning_rate": 0.000443959096581988, "loss": 1.3065, "step": 56180 }, { "epoch": 2.26, "learning_rate": 0.0004439490317645638, "loss": 1.2554, "step": 56190 }, { "epoch": 2.26, "learning_rate": 0.0004439389669471396, "loss": 1.1507, "step": 56200 }, { "epoch": 2.26, "learning_rate": 0.0004439289021297154, "loss": 1.2891, "step": 56210 }, { "epoch": 2.26, "learning_rate": 0.00044391883731229113, "loss": 1.2766, "step": 56220 }, { "epoch": 2.26, "learning_rate": 0.0004439087724948669, "loss": 1.1861, "step": 56230 }, { "epoch": 2.26, "learning_rate": 0.0004438987076774427, "loss": 1.1234, "step": 56240 }, { "epoch": 2.26, "learning_rate": 0.0004438886428600185, "loss": 1.2623, "step": 56250 }, { "epoch": 2.26, "learning_rate": 0.0004438785780425943, "loss": 1.4062, "step": 56260 }, { "epoch": 2.26, "learning_rate": 0.0004438685132251701, "loss": 1.3155, "step": 56270 }, { "epoch": 2.26, "learning_rate": 0.00044385844840774593, "loss": 1.0175, "step": 56280 }, { "epoch": 2.26, "learning_rate": 0.0004438483835903217, "loss": 1.0744, "step": 56290 }, { "epoch": 2.26, "learning_rate": 0.0004438383187728975, "loss": 1.3363, "step": 56300 }, { "epoch": 2.26, "learning_rate": 0.00044382825395547325, "loss": 1.1777, "step": 56310 }, { "epoch": 2.27, "learning_rate": 0.00044381818913804904, "loss": 1.1729, "step": 56320 }, { "epoch": 2.27, "learning_rate": 0.00044380812432062483, "loss": 1.3109, "step": 56330 }, { "epoch": 2.27, "learning_rate": 0.0004437980595032006, "loss": 1.2676, "step": 56340 }, { "epoch": 2.27, "learning_rate": 0.0004437879946857764, "loss": 1.1705, "step": 56350 }, { "epoch": 2.27, "learning_rate": 0.0004437779298683522, "loss": 1.1199, "step": 56360 }, { "epoch": 2.27, "learning_rate": 0.000443767865050928, "loss": 1.3477, "step": 56370 }, { "epoch": 2.27, "learning_rate": 0.0004437578002335038, "loss": 1.2526, "step": 56380 }, { "epoch": 2.27, "learning_rate": 0.0004437477354160795, "loss": 1.1777, "step": 56390 }, { "epoch": 2.27, "learning_rate": 0.0004437376705986553, "loss": 1.2179, "step": 56400 }, { "epoch": 2.27, "learning_rate": 0.0004437276057812311, "loss": 1.3489, "step": 56410 }, { "epoch": 2.27, "learning_rate": 0.0004437175409638069, "loss": 0.935, "step": 56420 }, { "epoch": 2.27, "learning_rate": 0.00044370747614638274, "loss": 1.3074, "step": 56430 }, { "epoch": 2.27, "learning_rate": 0.00044369741132895853, "loss": 1.3729, "step": 56440 }, { "epoch": 2.27, "learning_rate": 0.0004436873465115343, "loss": 1.2872, "step": 56450 }, { "epoch": 2.27, "learning_rate": 0.0004436772816941101, "loss": 1.3049, "step": 56460 }, { "epoch": 2.27, "learning_rate": 0.0004436672168766859, "loss": 1.2235, "step": 56470 }, { "epoch": 2.27, "learning_rate": 0.00044365715205926164, "loss": 1.2607, "step": 56480 }, { "epoch": 2.27, "learning_rate": 0.00044364708724183743, "loss": 1.2788, "step": 56490 }, { "epoch": 2.27, "learning_rate": 0.0004436370224244132, "loss": 1.1307, "step": 56500 }, { "epoch": 2.27, "learning_rate": 0.000443626957606989, "loss": 1.0763, "step": 56510 }, { "epoch": 2.27, "learning_rate": 0.0004436168927895648, "loss": 1.2511, "step": 56520 }, { "epoch": 2.27, "learning_rate": 0.0004436068279721406, "loss": 0.988, "step": 56530 }, { "epoch": 2.27, "learning_rate": 0.0004435967631547164, "loss": 1.088, "step": 56540 }, { "epoch": 2.27, "learning_rate": 0.00044358669833729217, "loss": 1.3172, "step": 56550 }, { "epoch": 2.27, "learning_rate": 0.0004435766335198679, "loss": 1.4772, "step": 56560 }, { "epoch": 2.28, "learning_rate": 0.00044356656870244375, "loss": 1.2874, "step": 56570 }, { "epoch": 2.28, "learning_rate": 0.00044355650388501955, "loss": 1.419, "step": 56580 }, { "epoch": 2.28, "learning_rate": 0.00044354643906759534, "loss": 1.1794, "step": 56590 }, { "epoch": 2.28, "learning_rate": 0.0004435363742501711, "loss": 1.3201, "step": 56600 }, { "epoch": 2.28, "learning_rate": 0.0004435263094327469, "loss": 1.2096, "step": 56610 }, { "epoch": 2.28, "learning_rate": 0.0004435162446153227, "loss": 1.2618, "step": 56620 }, { "epoch": 2.28, "learning_rate": 0.0004435061797978985, "loss": 1.3563, "step": 56630 }, { "epoch": 2.28, "learning_rate": 0.0004434961149804743, "loss": 1.37, "step": 56640 }, { "epoch": 2.28, "learning_rate": 0.00044348605016305, "loss": 1.0789, "step": 56650 }, { "epoch": 2.28, "learning_rate": 0.0004434759853456258, "loss": 1.2442, "step": 56660 }, { "epoch": 2.28, "learning_rate": 0.0004434659205282016, "loss": 1.0648, "step": 56670 }, { "epoch": 2.28, "learning_rate": 0.0004434558557107774, "loss": 1.6675, "step": 56680 }, { "epoch": 2.28, "learning_rate": 0.0004434457908933532, "loss": 1.3442, "step": 56690 }, { "epoch": 2.28, "learning_rate": 0.000443435726075929, "loss": 1.2865, "step": 56700 }, { "epoch": 2.28, "learning_rate": 0.00044342566125850477, "loss": 1.1053, "step": 56710 }, { "epoch": 2.28, "learning_rate": 0.0004434155964410806, "loss": 1.0864, "step": 56720 }, { "epoch": 2.28, "learning_rate": 0.00044340553162365635, "loss": 1.5099, "step": 56730 }, { "epoch": 2.28, "learning_rate": 0.00044339546680623214, "loss": 1.0385, "step": 56740 }, { "epoch": 2.28, "learning_rate": 0.00044338540198880793, "loss": 1.2026, "step": 56750 }, { "epoch": 2.28, "learning_rate": 0.0004433753371713837, "loss": 1.2656, "step": 56760 }, { "epoch": 2.28, "learning_rate": 0.0004433652723539595, "loss": 1.2557, "step": 56770 }, { "epoch": 2.28, "learning_rate": 0.0004433552075365353, "loss": 1.3943, "step": 56780 }, { "epoch": 2.28, "learning_rate": 0.0004433451427191111, "loss": 1.4465, "step": 56790 }, { "epoch": 2.28, "learning_rate": 0.0004433350779016869, "loss": 1.2598, "step": 56800 }, { "epoch": 2.28, "learning_rate": 0.0004433250130842626, "loss": 1.4175, "step": 56810 }, { "epoch": 2.29, "learning_rate": 0.0004433149482668384, "loss": 1.1625, "step": 56820 }, { "epoch": 2.29, "learning_rate": 0.0004433048834494142, "loss": 1.3471, "step": 56830 }, { "epoch": 2.29, "learning_rate": 0.00044329481863199, "loss": 1.2523, "step": 56840 }, { "epoch": 2.29, "learning_rate": 0.0004432847538145658, "loss": 1.185, "step": 56850 }, { "epoch": 2.29, "learning_rate": 0.00044327468899714163, "loss": 1.2072, "step": 56860 }, { "epoch": 2.29, "learning_rate": 0.0004432646241797174, "loss": 1.2534, "step": 56870 }, { "epoch": 2.29, "learning_rate": 0.0004432545593622932, "loss": 1.0533, "step": 56880 }, { "epoch": 2.29, "learning_rate": 0.000443244494544869, "loss": 1.1795, "step": 56890 }, { "epoch": 2.29, "learning_rate": 0.00044323442972744474, "loss": 1.1418, "step": 56900 }, { "epoch": 2.29, "learning_rate": 0.00044322436491002053, "loss": 1.0135, "step": 56910 }, { "epoch": 2.29, "learning_rate": 0.0004432143000925963, "loss": 1.4296, "step": 56920 }, { "epoch": 2.29, "learning_rate": 0.0004432042352751721, "loss": 1.278, "step": 56930 }, { "epoch": 2.29, "learning_rate": 0.0004431941704577479, "loss": 1.0997, "step": 56940 }, { "epoch": 2.29, "learning_rate": 0.0004431841056403237, "loss": 0.8755, "step": 56950 }, { "epoch": 2.29, "learning_rate": 0.0004431740408228995, "loss": 1.0717, "step": 56960 }, { "epoch": 2.29, "learning_rate": 0.0004431639760054753, "loss": 1.3214, "step": 56970 }, { "epoch": 2.29, "learning_rate": 0.000443153911188051, "loss": 1.1435, "step": 56980 }, { "epoch": 2.29, "learning_rate": 0.0004431438463706268, "loss": 1.4718, "step": 56990 }, { "epoch": 2.29, "learning_rate": 0.0004431337815532026, "loss": 1.2817, "step": 57000 }, { "epoch": 2.29, "learning_rate": 0.00044312371673577844, "loss": 1.1105, "step": 57010 }, { "epoch": 2.29, "learning_rate": 0.00044311365191835423, "loss": 1.1617, "step": 57020 }, { "epoch": 2.29, "learning_rate": 0.00044310358710093, "loss": 1.214, "step": 57030 }, { "epoch": 2.29, "learning_rate": 0.0004430935222835058, "loss": 1.0791, "step": 57040 }, { "epoch": 2.29, "learning_rate": 0.0004430834574660816, "loss": 1.0633, "step": 57050 }, { "epoch": 2.29, "learning_rate": 0.0004430733926486574, "loss": 1.3539, "step": 57060 }, { "epoch": 2.3, "learning_rate": 0.00044306332783123313, "loss": 1.3125, "step": 57070 }, { "epoch": 2.3, "learning_rate": 0.0004430532630138089, "loss": 1.3757, "step": 57080 }, { "epoch": 2.3, "learning_rate": 0.0004430431981963847, "loss": 1.3271, "step": 57090 }, { "epoch": 2.3, "learning_rate": 0.0004430331333789605, "loss": 1.2937, "step": 57100 }, { "epoch": 2.3, "learning_rate": 0.0004430230685615363, "loss": 1.2235, "step": 57110 }, { "epoch": 2.3, "learning_rate": 0.0004430130037441121, "loss": 1.1071, "step": 57120 }, { "epoch": 2.3, "learning_rate": 0.0004430029389266879, "loss": 1.3488, "step": 57130 }, { "epoch": 2.3, "learning_rate": 0.00044299287410926367, "loss": 1.1206, "step": 57140 }, { "epoch": 2.3, "learning_rate": 0.0004429828092918394, "loss": 1.4074, "step": 57150 }, { "epoch": 2.3, "learning_rate": 0.00044297274447441525, "loss": 1.0574, "step": 57160 }, { "epoch": 2.3, "learning_rate": 0.00044296267965699104, "loss": 1.2839, "step": 57170 }, { "epoch": 2.3, "learning_rate": 0.00044295261483956683, "loss": 1.1546, "step": 57180 }, { "epoch": 2.3, "learning_rate": 0.0004429425500221426, "loss": 1.3502, "step": 57190 }, { "epoch": 2.3, "learning_rate": 0.0004429324852047184, "loss": 1.4711, "step": 57200 }, { "epoch": 2.3, "learning_rate": 0.0004429224203872942, "loss": 1.3063, "step": 57210 }, { "epoch": 2.3, "learning_rate": 0.00044291235556987, "loss": 1.3304, "step": 57220 }, { "epoch": 2.3, "learning_rate": 0.0004429022907524458, "loss": 1.2887, "step": 57230 }, { "epoch": 2.3, "learning_rate": 0.0004428922259350215, "loss": 1.2327, "step": 57240 }, { "epoch": 2.3, "learning_rate": 0.0004428821611175973, "loss": 1.0584, "step": 57250 }, { "epoch": 2.3, "learning_rate": 0.0004428720963001731, "loss": 1.506, "step": 57260 }, { "epoch": 2.3, "learning_rate": 0.0004428620314827489, "loss": 1.1889, "step": 57270 }, { "epoch": 2.3, "learning_rate": 0.0004428519666653247, "loss": 1.2357, "step": 57280 }, { "epoch": 2.3, "learning_rate": 0.0004428419018479005, "loss": 1.0468, "step": 57290 }, { "epoch": 2.3, "learning_rate": 0.0004428318370304763, "loss": 1.3463, "step": 57300 }, { "epoch": 2.3, "learning_rate": 0.0004428217722130521, "loss": 1.1746, "step": 57310 }, { "epoch": 2.31, "learning_rate": 0.00044281170739562785, "loss": 1.0829, "step": 57320 }, { "epoch": 2.31, "learning_rate": 0.00044280164257820364, "loss": 1.1785, "step": 57330 }, { "epoch": 2.31, "learning_rate": 0.00044279157776077943, "loss": 1.0976, "step": 57340 }, { "epoch": 2.31, "learning_rate": 0.0004427815129433552, "loss": 1.1463, "step": 57350 }, { "epoch": 2.31, "learning_rate": 0.000442771448125931, "loss": 0.8799, "step": 57360 }, { "epoch": 2.31, "learning_rate": 0.0004427613833085068, "loss": 1.3317, "step": 57370 }, { "epoch": 2.31, "learning_rate": 0.0004427513184910826, "loss": 1.2916, "step": 57380 }, { "epoch": 2.31, "learning_rate": 0.0004427412536736584, "loss": 1.0417, "step": 57390 }, { "epoch": 2.31, "learning_rate": 0.00044273118885623417, "loss": 1.0686, "step": 57400 }, { "epoch": 2.31, "learning_rate": 0.0004427211240388099, "loss": 1.2602, "step": 57410 }, { "epoch": 2.31, "learning_rate": 0.0004427110592213857, "loss": 1.3349, "step": 57420 }, { "epoch": 2.31, "learning_rate": 0.0004427009944039615, "loss": 1.4638, "step": 57430 }, { "epoch": 2.31, "learning_rate": 0.0004426909295865373, "loss": 1.3837, "step": 57440 }, { "epoch": 2.31, "learning_rate": 0.0004426808647691131, "loss": 1.3165, "step": 57450 }, { "epoch": 2.31, "learning_rate": 0.0004426707999516889, "loss": 1.192, "step": 57460 }, { "epoch": 2.31, "learning_rate": 0.0004426607351342647, "loss": 1.4596, "step": 57470 }, { "epoch": 2.31, "learning_rate": 0.0004426506703168405, "loss": 1.0406, "step": 57480 }, { "epoch": 2.31, "learning_rate": 0.00044264060549941624, "loss": 1.0634, "step": 57490 }, { "epoch": 2.31, "learning_rate": 0.000442630540681992, "loss": 1.2911, "step": 57500 }, { "epoch": 2.31, "learning_rate": 0.0004426204758645678, "loss": 1.1894, "step": 57510 }, { "epoch": 2.31, "learning_rate": 0.0004426104110471436, "loss": 1.2376, "step": 57520 }, { "epoch": 2.31, "learning_rate": 0.0004426003462297194, "loss": 1.2377, "step": 57530 }, { "epoch": 2.31, "learning_rate": 0.0004425902814122952, "loss": 1.1502, "step": 57540 }, { "epoch": 2.31, "learning_rate": 0.000442580216594871, "loss": 1.4042, "step": 57550 }, { "epoch": 2.31, "learning_rate": 0.00044257015177744677, "loss": 1.2414, "step": 57560 }, { "epoch": 2.32, "learning_rate": 0.0004425600869600225, "loss": 1.3037, "step": 57570 }, { "epoch": 2.32, "learning_rate": 0.0004425500221425983, "loss": 1.2532, "step": 57580 }, { "epoch": 2.32, "learning_rate": 0.0004425399573251741, "loss": 1.1841, "step": 57590 }, { "epoch": 2.32, "learning_rate": 0.00044252989250774993, "loss": 1.2406, "step": 57600 }, { "epoch": 2.32, "learning_rate": 0.0004425198276903257, "loss": 1.5369, "step": 57610 }, { "epoch": 2.32, "learning_rate": 0.0004425097628729015, "loss": 1.1164, "step": 57620 }, { "epoch": 2.32, "learning_rate": 0.0004424996980554773, "loss": 1.2785, "step": 57630 }, { "epoch": 2.32, "learning_rate": 0.0004424896332380531, "loss": 1.415, "step": 57640 }, { "epoch": 2.32, "learning_rate": 0.0004424795684206289, "loss": 1.4812, "step": 57650 }, { "epoch": 2.32, "learning_rate": 0.0004424695036032046, "loss": 1.0633, "step": 57660 }, { "epoch": 2.32, "learning_rate": 0.0004424594387857804, "loss": 1.4534, "step": 57670 }, { "epoch": 2.32, "learning_rate": 0.0004424493739683562, "loss": 1.4042, "step": 57680 }, { "epoch": 2.32, "learning_rate": 0.000442439309150932, "loss": 1.1298, "step": 57690 }, { "epoch": 2.32, "learning_rate": 0.0004424292443335078, "loss": 1.2511, "step": 57700 }, { "epoch": 2.32, "learning_rate": 0.0004424191795160836, "loss": 1.2327, "step": 57710 }, { "epoch": 2.32, "learning_rate": 0.00044240911469865937, "loss": 1.1224, "step": 57720 }, { "epoch": 2.32, "learning_rate": 0.00044239904988123516, "loss": 1.2298, "step": 57730 }, { "epoch": 2.32, "learning_rate": 0.000442388985063811, "loss": 1.2478, "step": 57740 }, { "epoch": 2.32, "learning_rate": 0.00044237892024638674, "loss": 1.1196, "step": 57750 }, { "epoch": 2.32, "learning_rate": 0.00044236885542896253, "loss": 1.1281, "step": 57760 }, { "epoch": 2.32, "learning_rate": 0.0004423587906115383, "loss": 1.324, "step": 57770 }, { "epoch": 2.32, "learning_rate": 0.0004423487257941141, "loss": 1.1004, "step": 57780 }, { "epoch": 2.32, "learning_rate": 0.0004423386609766899, "loss": 1.2393, "step": 57790 }, { "epoch": 2.32, "learning_rate": 0.0004423285961592657, "loss": 1.1836, "step": 57800 }, { "epoch": 2.33, "learning_rate": 0.0004423185313418415, "loss": 1.4177, "step": 57810 }, { "epoch": 2.33, "learning_rate": 0.0004423084665244173, "loss": 1.3082, "step": 57820 }, { "epoch": 2.33, "learning_rate": 0.000442298401706993, "loss": 1.3738, "step": 57830 }, { "epoch": 2.33, "learning_rate": 0.0004422883368895688, "loss": 1.2551, "step": 57840 }, { "epoch": 2.33, "learning_rate": 0.0004422782720721446, "loss": 1.2361, "step": 57850 }, { "epoch": 2.33, "learning_rate": 0.0004422682072547204, "loss": 1.0992, "step": 57860 }, { "epoch": 2.33, "learning_rate": 0.0004422581424372962, "loss": 1.1226, "step": 57870 }, { "epoch": 2.33, "learning_rate": 0.00044224807761987197, "loss": 1.398, "step": 57880 }, { "epoch": 2.33, "learning_rate": 0.0004422380128024478, "loss": 1.2561, "step": 57890 }, { "epoch": 2.33, "learning_rate": 0.0004422279479850236, "loss": 1.2116, "step": 57900 }, { "epoch": 2.33, "learning_rate": 0.00044221788316759934, "loss": 1.1682, "step": 57910 }, { "epoch": 2.33, "learning_rate": 0.00044220781835017513, "loss": 1.2066, "step": 57920 }, { "epoch": 2.33, "learning_rate": 0.0004421977535327509, "loss": 1.0033, "step": 57930 }, { "epoch": 2.33, "learning_rate": 0.0004421876887153267, "loss": 1.1418, "step": 57940 }, { "epoch": 2.33, "learning_rate": 0.0004421776238979025, "loss": 1.3328, "step": 57950 }, { "epoch": 2.33, "learning_rate": 0.0004421675590804783, "loss": 1.0375, "step": 57960 }, { "epoch": 2.33, "learning_rate": 0.0004421574942630541, "loss": 1.1263, "step": 57970 }, { "epoch": 2.33, "learning_rate": 0.0004421474294456299, "loss": 1.4369, "step": 57980 }, { "epoch": 2.33, "learning_rate": 0.00044213736462820567, "loss": 1.2418, "step": 57990 }, { "epoch": 2.33, "learning_rate": 0.0004421272998107814, "loss": 1.1611, "step": 58000 }, { "epoch": 2.33, "learning_rate": 0.0004421172349933572, "loss": 1.3947, "step": 58010 }, { "epoch": 2.33, "learning_rate": 0.000442107170175933, "loss": 1.2887, "step": 58020 }, { "epoch": 2.33, "learning_rate": 0.00044209710535850883, "loss": 1.2447, "step": 58030 }, { "epoch": 2.33, "learning_rate": 0.0004420870405410846, "loss": 1.2178, "step": 58040 }, { "epoch": 2.33, "learning_rate": 0.0004420769757236604, "loss": 1.305, "step": 58050 }, { "epoch": 2.34, "learning_rate": 0.0004420669109062362, "loss": 1.283, "step": 58060 }, { "epoch": 2.34, "learning_rate": 0.000442056846088812, "loss": 1.1031, "step": 58070 }, { "epoch": 2.34, "learning_rate": 0.00044204678127138773, "loss": 1.3194, "step": 58080 }, { "epoch": 2.34, "learning_rate": 0.0004420367164539635, "loss": 1.2257, "step": 58090 }, { "epoch": 2.34, "learning_rate": 0.0004420266516365393, "loss": 1.0931, "step": 58100 }, { "epoch": 2.34, "learning_rate": 0.0004420165868191151, "loss": 1.1884, "step": 58110 }, { "epoch": 2.34, "learning_rate": 0.0004420065220016909, "loss": 1.2478, "step": 58120 }, { "epoch": 2.34, "learning_rate": 0.0004419964571842667, "loss": 1.2981, "step": 58130 }, { "epoch": 2.34, "learning_rate": 0.0004419863923668425, "loss": 0.9906, "step": 58140 }, { "epoch": 2.34, "learning_rate": 0.00044197632754941826, "loss": 1.3316, "step": 58150 }, { "epoch": 2.34, "learning_rate": 0.00044196626273199406, "loss": 1.3377, "step": 58160 }, { "epoch": 2.34, "learning_rate": 0.0004419561979145698, "loss": 1.2292, "step": 58170 }, { "epoch": 2.34, "learning_rate": 0.00044194613309714564, "loss": 1.3138, "step": 58180 }, { "epoch": 2.34, "learning_rate": 0.00044193606827972143, "loss": 1.0544, "step": 58190 }, { "epoch": 2.34, "learning_rate": 0.0004419260034622972, "loss": 1.2318, "step": 58200 }, { "epoch": 2.34, "learning_rate": 0.000441915938644873, "loss": 1.5302, "step": 58210 }, { "epoch": 2.34, "learning_rate": 0.0004419058738274488, "loss": 1.2462, "step": 58220 }, { "epoch": 2.34, "learning_rate": 0.0004418958090100246, "loss": 1.1701, "step": 58230 }, { "epoch": 2.34, "learning_rate": 0.0004418857441926004, "loss": 1.2585, "step": 58240 }, { "epoch": 2.34, "learning_rate": 0.0004418756793751761, "loss": 1.1565, "step": 58250 }, { "epoch": 2.34, "learning_rate": 0.0004418656145577519, "loss": 1.2022, "step": 58260 }, { "epoch": 2.34, "learning_rate": 0.0004418555497403277, "loss": 1.4217, "step": 58270 }, { "epoch": 2.34, "learning_rate": 0.0004418454849229035, "loss": 1.2073, "step": 58280 }, { "epoch": 2.34, "learning_rate": 0.0004418354201054793, "loss": 1.051, "step": 58290 }, { "epoch": 2.34, "learning_rate": 0.00044182535528805507, "loss": 1.0484, "step": 58300 }, { "epoch": 2.35, "learning_rate": 0.00044181529047063086, "loss": 1.1753, "step": 58310 }, { "epoch": 2.35, "learning_rate": 0.00044180522565320665, "loss": 1.0448, "step": 58320 }, { "epoch": 2.35, "learning_rate": 0.0004417951608357825, "loss": 1.3684, "step": 58330 }, { "epoch": 2.35, "learning_rate": 0.00044178509601835824, "loss": 1.248, "step": 58340 }, { "epoch": 2.35, "learning_rate": 0.000441775031200934, "loss": 1.1948, "step": 58350 }, { "epoch": 2.35, "learning_rate": 0.0004417649663835098, "loss": 1.2831, "step": 58360 }, { "epoch": 2.35, "learning_rate": 0.0004417549015660856, "loss": 1.0301, "step": 58370 }, { "epoch": 2.35, "learning_rate": 0.0004417448367486614, "loss": 1.0318, "step": 58380 }, { "epoch": 2.35, "learning_rate": 0.0004417347719312372, "loss": 1.2036, "step": 58390 }, { "epoch": 2.35, "learning_rate": 0.000441724707113813, "loss": 1.1209, "step": 58400 }, { "epoch": 2.35, "learning_rate": 0.00044171464229638877, "loss": 1.0976, "step": 58410 }, { "epoch": 2.35, "learning_rate": 0.0004417045774789645, "loss": 1.4823, "step": 58420 }, { "epoch": 2.35, "learning_rate": 0.0004416945126615403, "loss": 1.1516, "step": 58430 }, { "epoch": 2.35, "learning_rate": 0.0004416844478441161, "loss": 1.3852, "step": 58440 }, { "epoch": 2.35, "learning_rate": 0.0004416743830266919, "loss": 1.1937, "step": 58450 }, { "epoch": 2.35, "learning_rate": 0.00044166431820926767, "loss": 1.2374, "step": 58460 }, { "epoch": 2.35, "learning_rate": 0.0004416542533918435, "loss": 1.276, "step": 58470 }, { "epoch": 2.35, "learning_rate": 0.0004416441885744193, "loss": 1.0273, "step": 58480 }, { "epoch": 2.35, "learning_rate": 0.0004416341237569951, "loss": 1.3296, "step": 58490 }, { "epoch": 2.35, "learning_rate": 0.0004416240589395709, "loss": 1.3465, "step": 58500 }, { "epoch": 2.35, "learning_rate": 0.0004416139941221466, "loss": 1.2757, "step": 58510 }, { "epoch": 2.35, "learning_rate": 0.0004416039293047224, "loss": 1.1756, "step": 58520 }, { "epoch": 2.35, "learning_rate": 0.0004415938644872982, "loss": 1.3214, "step": 58530 }, { "epoch": 2.35, "learning_rate": 0.000441583799669874, "loss": 1.4014, "step": 58540 }, { "epoch": 2.35, "learning_rate": 0.0004415737348524498, "loss": 1.3011, "step": 58550 }, { "epoch": 2.36, "learning_rate": 0.0004415636700350256, "loss": 1.0041, "step": 58560 }, { "epoch": 2.36, "learning_rate": 0.00044155360521760137, "loss": 1.0178, "step": 58570 }, { "epoch": 2.36, "learning_rate": 0.00044154354040017716, "loss": 1.5187, "step": 58580 }, { "epoch": 2.36, "learning_rate": 0.0004415334755827529, "loss": 1.1733, "step": 58590 }, { "epoch": 2.36, "learning_rate": 0.0004415234107653287, "loss": 1.0152, "step": 58600 }, { "epoch": 2.36, "learning_rate": 0.0004415133459479045, "loss": 1.211, "step": 58610 }, { "epoch": 2.36, "learning_rate": 0.0004415032811304803, "loss": 1.0709, "step": 58620 }, { "epoch": 2.36, "learning_rate": 0.0004414932163130561, "loss": 1.1786, "step": 58630 }, { "epoch": 2.36, "learning_rate": 0.0004414831514956319, "loss": 1.2722, "step": 58640 }, { "epoch": 2.36, "learning_rate": 0.0004414730866782077, "loss": 1.1759, "step": 58650 }, { "epoch": 2.36, "learning_rate": 0.0004414630218607835, "loss": 1.4746, "step": 58660 }, { "epoch": 2.36, "learning_rate": 0.0004414529570433592, "loss": 1.2825, "step": 58670 }, { "epoch": 2.36, "learning_rate": 0.000441442892225935, "loss": 1.0993, "step": 58680 }, { "epoch": 2.36, "learning_rate": 0.0004414328274085108, "loss": 1.1243, "step": 58690 }, { "epoch": 2.36, "learning_rate": 0.0004414227625910866, "loss": 1.2841, "step": 58700 }, { "epoch": 2.36, "learning_rate": 0.0004414126977736624, "loss": 1.5036, "step": 58710 }, { "epoch": 2.36, "learning_rate": 0.0004414026329562382, "loss": 1.1605, "step": 58720 }, { "epoch": 2.36, "learning_rate": 0.00044139256813881397, "loss": 1.2009, "step": 58730 }, { "epoch": 2.36, "learning_rate": 0.00044138250332138976, "loss": 0.9913, "step": 58740 }, { "epoch": 2.36, "learning_rate": 0.00044137243850396555, "loss": 1.1085, "step": 58750 }, { "epoch": 2.36, "learning_rate": 0.00044136237368654134, "loss": 0.9746, "step": 58760 }, { "epoch": 2.36, "learning_rate": 0.00044135230886911713, "loss": 1.1834, "step": 58770 }, { "epoch": 2.36, "learning_rate": 0.0004413422440516929, "loss": 1.2652, "step": 58780 }, { "epoch": 2.36, "learning_rate": 0.0004413321792342687, "loss": 1.1931, "step": 58790 }, { "epoch": 2.36, "learning_rate": 0.0004413221144168445, "loss": 1.3345, "step": 58800 }, { "epoch": 2.37, "learning_rate": 0.0004413120495994203, "loss": 1.1755, "step": 58810 }, { "epoch": 2.37, "learning_rate": 0.0004413019847819961, "loss": 1.2342, "step": 58820 }, { "epoch": 2.37, "learning_rate": 0.0004412919199645719, "loss": 1.1399, "step": 58830 }, { "epoch": 2.37, "learning_rate": 0.0004412818551471476, "loss": 1.1792, "step": 58840 }, { "epoch": 2.37, "learning_rate": 0.0004412717903297234, "loss": 1.3317, "step": 58850 }, { "epoch": 2.37, "learning_rate": 0.0004412617255122992, "loss": 1.0568, "step": 58860 }, { "epoch": 2.37, "learning_rate": 0.000441251660694875, "loss": 1.2729, "step": 58870 }, { "epoch": 2.37, "learning_rate": 0.0004412415958774508, "loss": 1.2774, "step": 58880 }, { "epoch": 2.37, "learning_rate": 0.00044123153106002657, "loss": 1.2338, "step": 58890 }, { "epoch": 2.37, "learning_rate": 0.00044122146624260236, "loss": 1.2276, "step": 58900 }, { "epoch": 2.37, "learning_rate": 0.0004412114014251782, "loss": 1.2665, "step": 58910 }, { "epoch": 2.37, "learning_rate": 0.000441201336607754, "loss": 1.4846, "step": 58920 }, { "epoch": 2.37, "learning_rate": 0.00044119127179032973, "loss": 1.3029, "step": 58930 }, { "epoch": 2.37, "learning_rate": 0.0004411812069729055, "loss": 1.3043, "step": 58940 }, { "epoch": 2.37, "learning_rate": 0.0004411711421554813, "loss": 1.2136, "step": 58950 }, { "epoch": 2.37, "learning_rate": 0.0004411610773380571, "loss": 1.24, "step": 58960 }, { "epoch": 2.37, "learning_rate": 0.0004411510125206329, "loss": 1.2181, "step": 58970 }, { "epoch": 2.37, "learning_rate": 0.0004411409477032087, "loss": 1.2123, "step": 58980 }, { "epoch": 2.37, "learning_rate": 0.0004411308828857845, "loss": 1.4009, "step": 58990 }, { "epoch": 2.37, "learning_rate": 0.00044112081806836026, "loss": 1.3092, "step": 59000 }, { "epoch": 2.37, "learning_rate": 0.000441110753250936, "loss": 1.3188, "step": 59010 }, { "epoch": 2.37, "learning_rate": 0.0004411006884335118, "loss": 1.1311, "step": 59020 }, { "epoch": 2.37, "learning_rate": 0.0004410906236160876, "loss": 1.3104, "step": 59030 }, { "epoch": 2.37, "learning_rate": 0.0004410805587986634, "loss": 1.5547, "step": 59040 }, { "epoch": 2.37, "learning_rate": 0.00044107049398123916, "loss": 1.3721, "step": 59050 }, { "epoch": 2.38, "learning_rate": 0.000441060429163815, "loss": 1.2766, "step": 59060 }, { "epoch": 2.38, "learning_rate": 0.0004410503643463908, "loss": 1.2299, "step": 59070 }, { "epoch": 2.38, "learning_rate": 0.0004410402995289666, "loss": 1.1693, "step": 59080 }, { "epoch": 2.38, "learning_rate": 0.0004410302347115424, "loss": 1.285, "step": 59090 }, { "epoch": 2.38, "learning_rate": 0.0004410201698941181, "loss": 1.2731, "step": 59100 }, { "epoch": 2.38, "learning_rate": 0.0004410101050766939, "loss": 1.3855, "step": 59110 }, { "epoch": 2.38, "learning_rate": 0.0004410000402592697, "loss": 1.2135, "step": 59120 }, { "epoch": 2.38, "learning_rate": 0.0004409899754418455, "loss": 1.2375, "step": 59130 }, { "epoch": 2.38, "learning_rate": 0.0004409799106244213, "loss": 1.1272, "step": 59140 }, { "epoch": 2.38, "learning_rate": 0.00044096984580699707, "loss": 1.4232, "step": 59150 }, { "epoch": 2.38, "learning_rate": 0.00044095978098957286, "loss": 1.4774, "step": 59160 }, { "epoch": 2.38, "learning_rate": 0.00044094971617214865, "loss": 1.1037, "step": 59170 }, { "epoch": 2.38, "learning_rate": 0.0004409396513547244, "loss": 1.3158, "step": 59180 }, { "epoch": 2.38, "learning_rate": 0.0004409295865373002, "loss": 1.2704, "step": 59190 }, { "epoch": 2.38, "learning_rate": 0.000440919521719876, "loss": 1.1377, "step": 59200 }, { "epoch": 2.38, "learning_rate": 0.0004409094569024518, "loss": 1.0417, "step": 59210 }, { "epoch": 2.38, "learning_rate": 0.0004408993920850276, "loss": 1.4154, "step": 59220 }, { "epoch": 2.38, "learning_rate": 0.0004408893272676034, "loss": 1.1948, "step": 59230 }, { "epoch": 2.38, "learning_rate": 0.0004408792624501792, "loss": 1.193, "step": 59240 }, { "epoch": 2.38, "learning_rate": 0.000440869197632755, "loss": 1.1678, "step": 59250 }, { "epoch": 2.38, "learning_rate": 0.00044085913281533077, "loss": 1.4904, "step": 59260 }, { "epoch": 2.38, "learning_rate": 0.0004408490679979065, "loss": 1.326, "step": 59270 }, { "epoch": 2.38, "learning_rate": 0.0004408390031804823, "loss": 1.1099, "step": 59280 }, { "epoch": 2.38, "learning_rate": 0.0004408289383630581, "loss": 1.4393, "step": 59290 }, { "epoch": 2.38, "learning_rate": 0.0004408188735456339, "loss": 1.3095, "step": 59300 }, { "epoch": 2.39, "learning_rate": 0.00044080880872820967, "loss": 1.233, "step": 59310 }, { "epoch": 2.39, "learning_rate": 0.00044079874391078546, "loss": 1.1594, "step": 59320 }, { "epoch": 2.39, "learning_rate": 0.00044078867909336125, "loss": 1.3855, "step": 59330 }, { "epoch": 2.39, "learning_rate": 0.00044077861427593704, "loss": 1.557, "step": 59340 }, { "epoch": 2.39, "learning_rate": 0.00044076854945851283, "loss": 1.2278, "step": 59350 }, { "epoch": 2.39, "learning_rate": 0.0004407584846410886, "loss": 1.0838, "step": 59360 }, { "epoch": 2.39, "learning_rate": 0.0004407484198236644, "loss": 1.3229, "step": 59370 }, { "epoch": 2.39, "learning_rate": 0.0004407383550062402, "loss": 1.2541, "step": 59380 }, { "epoch": 2.39, "learning_rate": 0.000440728290188816, "loss": 1.0369, "step": 59390 }, { "epoch": 2.39, "learning_rate": 0.0004407182253713918, "loss": 1.208, "step": 59400 }, { "epoch": 2.39, "learning_rate": 0.0004407081605539676, "loss": 1.3545, "step": 59410 }, { "epoch": 2.39, "learning_rate": 0.00044069809573654337, "loss": 1.1822, "step": 59420 }, { "epoch": 2.39, "learning_rate": 0.0004406880309191191, "loss": 1.2839, "step": 59430 }, { "epoch": 2.39, "learning_rate": 0.0004406779661016949, "loss": 1.1689, "step": 59440 }, { "epoch": 2.39, "learning_rate": 0.0004406679012842707, "loss": 1.1787, "step": 59450 }, { "epoch": 2.39, "learning_rate": 0.0004406578364668465, "loss": 1.3104, "step": 59460 }, { "epoch": 2.39, "learning_rate": 0.00044064777164942227, "loss": 0.9487, "step": 59470 }, { "epoch": 2.39, "learning_rate": 0.00044063770683199806, "loss": 1.0954, "step": 59480 }, { "epoch": 2.39, "learning_rate": 0.00044062764201457385, "loss": 1.0815, "step": 59490 }, { "epoch": 2.39, "learning_rate": 0.0004406175771971497, "loss": 0.9928, "step": 59500 }, { "epoch": 2.39, "learning_rate": 0.0004406075123797255, "loss": 1.1267, "step": 59510 }, { "epoch": 2.39, "learning_rate": 0.0004405974475623012, "loss": 1.3569, "step": 59520 }, { "epoch": 2.39, "learning_rate": 0.000440587382744877, "loss": 0.9894, "step": 59530 }, { "epoch": 2.39, "learning_rate": 0.0004405773179274528, "loss": 1.1247, "step": 59540 }, { "epoch": 2.4, "learning_rate": 0.0004405672531100286, "loss": 1.1693, "step": 59550 }, { "epoch": 2.4, "learning_rate": 0.0004405571882926044, "loss": 1.448, "step": 59560 }, { "epoch": 2.4, "learning_rate": 0.0004405471234751802, "loss": 1.2521, "step": 59570 }, { "epoch": 2.4, "learning_rate": 0.00044053705865775597, "loss": 1.2838, "step": 59580 }, { "epoch": 2.4, "learning_rate": 0.00044052699384033176, "loss": 1.2549, "step": 59590 }, { "epoch": 2.4, "learning_rate": 0.0004405169290229075, "loss": 1.2582, "step": 59600 }, { "epoch": 2.4, "learning_rate": 0.0004405068642054833, "loss": 1.5033, "step": 59610 }, { "epoch": 2.4, "learning_rate": 0.0004404967993880591, "loss": 1.1894, "step": 59620 }, { "epoch": 2.4, "learning_rate": 0.00044048673457063487, "loss": 1.2321, "step": 59630 }, { "epoch": 2.4, "learning_rate": 0.0004404766697532107, "loss": 1.356, "step": 59640 }, { "epoch": 2.4, "learning_rate": 0.0004404666049357865, "loss": 1.2193, "step": 59650 }, { "epoch": 2.4, "learning_rate": 0.0004404565401183623, "loss": 1.3394, "step": 59660 }, { "epoch": 2.4, "learning_rate": 0.0004404464753009381, "loss": 1.209, "step": 59670 }, { "epoch": 2.4, "learning_rate": 0.0004404364104835139, "loss": 1.4024, "step": 59680 }, { "epoch": 2.4, "learning_rate": 0.0004404263456660896, "loss": 1.0967, "step": 59690 }, { "epoch": 2.4, "learning_rate": 0.0004404162808486654, "loss": 1.5408, "step": 59700 }, { "epoch": 2.4, "learning_rate": 0.0004404062160312412, "loss": 1.1897, "step": 59710 }, { "epoch": 2.4, "learning_rate": 0.000440396151213817, "loss": 1.3536, "step": 59720 }, { "epoch": 2.4, "learning_rate": 0.0004403860863963928, "loss": 1.4164, "step": 59730 }, { "epoch": 2.4, "learning_rate": 0.00044037602157896857, "loss": 1.2076, "step": 59740 }, { "epoch": 2.4, "learning_rate": 0.00044036595676154436, "loss": 1.4823, "step": 59750 }, { "epoch": 2.4, "learning_rate": 0.00044035589194412015, "loss": 1.3513, "step": 59760 }, { "epoch": 2.4, "learning_rate": 0.0004403458271266959, "loss": 1.1903, "step": 59770 }, { "epoch": 2.4, "learning_rate": 0.0004403357623092717, "loss": 1.4995, "step": 59780 }, { "epoch": 2.4, "learning_rate": 0.0004403256974918475, "loss": 1.5242, "step": 59790 }, { "epoch": 2.41, "learning_rate": 0.0004403156326744233, "loss": 1.4112, "step": 59800 }, { "epoch": 2.41, "learning_rate": 0.0004403055678569991, "loss": 1.1942, "step": 59810 }, { "epoch": 2.41, "learning_rate": 0.0004402955030395749, "loss": 1.272, "step": 59820 }, { "epoch": 2.41, "learning_rate": 0.0004402854382221507, "loss": 1.1561, "step": 59830 }, { "epoch": 2.41, "learning_rate": 0.0004402753734047265, "loss": 1.32, "step": 59840 }, { "epoch": 2.41, "learning_rate": 0.00044026530858730226, "loss": 1.0388, "step": 59850 }, { "epoch": 2.41, "learning_rate": 0.000440255243769878, "loss": 1.4657, "step": 59860 }, { "epoch": 2.41, "learning_rate": 0.0004402451789524538, "loss": 1.2051, "step": 59870 }, { "epoch": 2.41, "learning_rate": 0.0004402351141350296, "loss": 1.2031, "step": 59880 }, { "epoch": 2.41, "learning_rate": 0.00044022504931760537, "loss": 1.1371, "step": 59890 }, { "epoch": 2.41, "learning_rate": 0.00044021498450018116, "loss": 1.2557, "step": 59900 }, { "epoch": 2.41, "learning_rate": 0.00044020491968275695, "loss": 0.9678, "step": 59910 }, { "epoch": 2.41, "learning_rate": 0.00044019485486533275, "loss": 1.2493, "step": 59920 }, { "epoch": 2.41, "learning_rate": 0.0004401847900479086, "loss": 1.4902, "step": 59930 }, { "epoch": 2.41, "learning_rate": 0.0004401747252304843, "loss": 1.2069, "step": 59940 }, { "epoch": 2.41, "learning_rate": 0.0004401646604130601, "loss": 1.1113, "step": 59950 }, { "epoch": 2.41, "learning_rate": 0.0004401545955956359, "loss": 0.9869, "step": 59960 }, { "epoch": 2.41, "learning_rate": 0.0004401445307782117, "loss": 1.22, "step": 59970 }, { "epoch": 2.41, "learning_rate": 0.0004401344659607875, "loss": 1.2346, "step": 59980 }, { "epoch": 2.41, "learning_rate": 0.0004401244011433633, "loss": 1.3032, "step": 59990 }, { "epoch": 2.41, "learning_rate": 0.00044011433632593907, "loss": 1.1995, "step": 60000 }, { "epoch": 2.41, "learning_rate": 0.00044010427150851486, "loss": 1.2692, "step": 60010 }, { "epoch": 2.41, "learning_rate": 0.00044009420669109065, "loss": 1.3485, "step": 60020 }, { "epoch": 2.41, "learning_rate": 0.0004400841418736664, "loss": 1.3553, "step": 60030 }, { "epoch": 2.41, "learning_rate": 0.0004400740770562422, "loss": 1.2767, "step": 60040 }, { "epoch": 2.42, "learning_rate": 0.00044006401223881797, "loss": 1.1337, "step": 60050 }, { "epoch": 2.42, "learning_rate": 0.00044005394742139376, "loss": 1.3481, "step": 60060 }, { "epoch": 2.42, "learning_rate": 0.00044004388260396955, "loss": 1.3509, "step": 60070 }, { "epoch": 2.42, "learning_rate": 0.0004400338177865454, "loss": 1.1452, "step": 60080 }, { "epoch": 2.42, "learning_rate": 0.0004400237529691212, "loss": 1.2126, "step": 60090 }, { "epoch": 2.42, "learning_rate": 0.000440013688151697, "loss": 1.0956, "step": 60100 }, { "epoch": 2.42, "learning_rate": 0.0004400036233342727, "loss": 1.2987, "step": 60110 }, { "epoch": 2.42, "learning_rate": 0.0004399935585168485, "loss": 1.3191, "step": 60120 }, { "epoch": 2.42, "learning_rate": 0.0004399834936994243, "loss": 1.1357, "step": 60130 }, { "epoch": 2.42, "learning_rate": 0.0004399734288820001, "loss": 1.2805, "step": 60140 }, { "epoch": 2.42, "learning_rate": 0.0004399633640645759, "loss": 1.3111, "step": 60150 }, { "epoch": 2.42, "learning_rate": 0.00043995329924715167, "loss": 1.3945, "step": 60160 }, { "epoch": 2.42, "learning_rate": 0.00043994323442972746, "loss": 1.2485, "step": 60170 }, { "epoch": 2.42, "learning_rate": 0.00043993316961230325, "loss": 1.2946, "step": 60180 }, { "epoch": 2.42, "learning_rate": 0.000439923104794879, "loss": 1.2069, "step": 60190 }, { "epoch": 2.42, "learning_rate": 0.0004399130399774548, "loss": 1.5154, "step": 60200 }, { "epoch": 2.42, "learning_rate": 0.00043990297516003057, "loss": 1.1395, "step": 60210 }, { "epoch": 2.42, "learning_rate": 0.00043989291034260636, "loss": 1.295, "step": 60220 }, { "epoch": 2.42, "learning_rate": 0.0004398828455251822, "loss": 1.3688, "step": 60230 }, { "epoch": 2.42, "learning_rate": 0.000439872780707758, "loss": 1.1137, "step": 60240 }, { "epoch": 2.42, "learning_rate": 0.0004398627158903338, "loss": 1.2243, "step": 60250 }, { "epoch": 2.42, "learning_rate": 0.0004398526510729096, "loss": 1.5027, "step": 60260 }, { "epoch": 2.42, "learning_rate": 0.00043984258625548537, "loss": 1.0989, "step": 60270 }, { "epoch": 2.42, "learning_rate": 0.0004398325214380611, "loss": 1.3387, "step": 60280 }, { "epoch": 2.42, "learning_rate": 0.0004398224566206369, "loss": 1.0587, "step": 60290 }, { "epoch": 2.43, "learning_rate": 0.0004398123918032127, "loss": 1.2662, "step": 60300 }, { "epoch": 2.43, "learning_rate": 0.0004398023269857885, "loss": 1.1468, "step": 60310 }, { "epoch": 2.43, "learning_rate": 0.00043979226216836427, "loss": 1.4241, "step": 60320 }, { "epoch": 2.43, "learning_rate": 0.00043978219735094006, "loss": 1.1927, "step": 60330 }, { "epoch": 2.43, "learning_rate": 0.00043977213253351585, "loss": 1.1792, "step": 60340 }, { "epoch": 2.43, "learning_rate": 0.00043976206771609164, "loss": 1.3899, "step": 60350 }, { "epoch": 2.43, "learning_rate": 0.0004397520028986674, "loss": 1.4196, "step": 60360 }, { "epoch": 2.43, "learning_rate": 0.0004397419380812432, "loss": 1.3249, "step": 60370 }, { "epoch": 2.43, "learning_rate": 0.000439731873263819, "loss": 1.4011, "step": 60380 }, { "epoch": 2.43, "learning_rate": 0.0004397218084463948, "loss": 1.0524, "step": 60390 }, { "epoch": 2.43, "learning_rate": 0.0004397117436289706, "loss": 1.1636, "step": 60400 }, { "epoch": 2.43, "learning_rate": 0.0004397016788115464, "loss": 1.2065, "step": 60410 }, { "epoch": 2.43, "learning_rate": 0.0004396916139941222, "loss": 0.9836, "step": 60420 }, { "epoch": 2.43, "learning_rate": 0.00043968154917669797, "loss": 0.9997, "step": 60430 }, { "epoch": 2.43, "learning_rate": 0.00043967148435927376, "loss": 1.1283, "step": 60440 }, { "epoch": 2.43, "learning_rate": 0.0004396614195418495, "loss": 1.1428, "step": 60450 }, { "epoch": 2.43, "learning_rate": 0.0004396513547244253, "loss": 1.2645, "step": 60460 }, { "epoch": 2.43, "learning_rate": 0.0004396412899070011, "loss": 1.2792, "step": 60470 }, { "epoch": 2.43, "learning_rate": 0.00043963122508957687, "loss": 1.5219, "step": 60480 }, { "epoch": 2.43, "learning_rate": 0.00043962116027215266, "loss": 1.5573, "step": 60490 }, { "epoch": 2.43, "learning_rate": 0.00043961109545472845, "loss": 1.3048, "step": 60500 }, { "epoch": 2.43, "learning_rate": 0.00043960103063730424, "loss": 1.1349, "step": 60510 }, { "epoch": 2.43, "learning_rate": 0.0004395909658198801, "loss": 1.321, "step": 60520 }, { "epoch": 2.43, "learning_rate": 0.0004395809010024558, "loss": 1.4979, "step": 60530 }, { "epoch": 2.43, "learning_rate": 0.0004395708361850316, "loss": 1.2431, "step": 60540 }, { "epoch": 2.44, "learning_rate": 0.0004395607713676074, "loss": 1.3367, "step": 60550 }, { "epoch": 2.44, "learning_rate": 0.0004395507065501832, "loss": 1.4171, "step": 60560 }, { "epoch": 2.44, "learning_rate": 0.000439540641732759, "loss": 0.9373, "step": 60570 }, { "epoch": 2.44, "learning_rate": 0.0004395305769153348, "loss": 1.5037, "step": 60580 }, { "epoch": 2.44, "learning_rate": 0.00043952051209791056, "loss": 1.2965, "step": 60590 }, { "epoch": 2.44, "learning_rate": 0.00043951044728048636, "loss": 1.2457, "step": 60600 }, { "epoch": 2.44, "learning_rate": 0.00043950038246306215, "loss": 1.3163, "step": 60610 }, { "epoch": 2.44, "learning_rate": 0.0004394903176456379, "loss": 1.1803, "step": 60620 }, { "epoch": 2.44, "learning_rate": 0.0004394802528282137, "loss": 1.0643, "step": 60630 }, { "epoch": 2.44, "learning_rate": 0.00043947018801078946, "loss": 1.4152, "step": 60640 }, { "epoch": 2.44, "learning_rate": 0.00043946012319336526, "loss": 1.1294, "step": 60650 }, { "epoch": 2.44, "learning_rate": 0.00043945005837594105, "loss": 1.3084, "step": 60660 }, { "epoch": 2.44, "learning_rate": 0.0004394399935585169, "loss": 1.3881, "step": 60670 }, { "epoch": 2.44, "learning_rate": 0.0004394299287410927, "loss": 1.2033, "step": 60680 }, { "epoch": 2.44, "learning_rate": 0.00043941986392366847, "loss": 1.2139, "step": 60690 }, { "epoch": 2.44, "learning_rate": 0.0004394097991062442, "loss": 1.2632, "step": 60700 }, { "epoch": 2.44, "learning_rate": 0.00043939973428882, "loss": 1.2179, "step": 60710 }, { "epoch": 2.44, "learning_rate": 0.0004393896694713958, "loss": 1.2673, "step": 60720 }, { "epoch": 2.44, "learning_rate": 0.0004393796046539716, "loss": 1.2908, "step": 60730 }, { "epoch": 2.44, "learning_rate": 0.00043936953983654737, "loss": 1.1242, "step": 60740 }, { "epoch": 2.44, "learning_rate": 0.00043935947501912316, "loss": 1.1664, "step": 60750 }, { "epoch": 2.44, "learning_rate": 0.00043934941020169895, "loss": 1.3111, "step": 60760 }, { "epoch": 2.44, "learning_rate": 0.00043933934538427474, "loss": 1.3272, "step": 60770 }, { "epoch": 2.44, "learning_rate": 0.00043932928056685054, "loss": 1.2263, "step": 60780 }, { "epoch": 2.44, "learning_rate": 0.00043931921574942627, "loss": 1.2755, "step": 60790 }, { "epoch": 2.45, "learning_rate": 0.00043930915093200206, "loss": 1.3232, "step": 60800 }, { "epoch": 2.45, "learning_rate": 0.0004392990861145779, "loss": 1.2081, "step": 60810 }, { "epoch": 2.45, "learning_rate": 0.0004392890212971537, "loss": 1.2002, "step": 60820 }, { "epoch": 2.45, "learning_rate": 0.0004392789564797295, "loss": 1.3522, "step": 60830 }, { "epoch": 2.45, "learning_rate": 0.0004392688916623053, "loss": 1.127, "step": 60840 }, { "epoch": 2.45, "learning_rate": 0.00043925882684488107, "loss": 1.1645, "step": 60850 }, { "epoch": 2.45, "learning_rate": 0.00043924876202745686, "loss": 1.3043, "step": 60860 }, { "epoch": 2.45, "learning_rate": 0.0004392386972100326, "loss": 1.0987, "step": 60870 }, { "epoch": 2.45, "learning_rate": 0.0004392286323926084, "loss": 1.1994, "step": 60880 }, { "epoch": 2.45, "learning_rate": 0.0004392185675751842, "loss": 1.1828, "step": 60890 }, { "epoch": 2.45, "learning_rate": 0.00043920850275775997, "loss": 1.6592, "step": 60900 }, { "epoch": 2.45, "learning_rate": 0.00043919843794033576, "loss": 1.127, "step": 60910 }, { "epoch": 2.45, "learning_rate": 0.00043918837312291155, "loss": 1.2814, "step": 60920 }, { "epoch": 2.45, "learning_rate": 0.00043917830830548734, "loss": 1.3225, "step": 60930 }, { "epoch": 2.45, "learning_rate": 0.00043916824348806313, "loss": 1.4191, "step": 60940 }, { "epoch": 2.45, "learning_rate": 0.00043915817867063887, "loss": 1.4252, "step": 60950 }, { "epoch": 2.45, "learning_rate": 0.0004391481138532147, "loss": 1.3596, "step": 60960 }, { "epoch": 2.45, "learning_rate": 0.0004391380490357905, "loss": 1.1987, "step": 60970 }, { "epoch": 2.45, "learning_rate": 0.0004391279842183663, "loss": 1.1953, "step": 60980 }, { "epoch": 2.45, "learning_rate": 0.0004391179194009421, "loss": 1.4514, "step": 60990 }, { "epoch": 2.45, "learning_rate": 0.0004391078545835179, "loss": 1.204, "step": 61000 }, { "epoch": 2.45, "learning_rate": 0.00043909778976609367, "loss": 1.2978, "step": 61010 }, { "epoch": 2.45, "learning_rate": 0.00043908772494866946, "loss": 1.2856, "step": 61020 }, { "epoch": 2.45, "learning_rate": 0.00043907766013124525, "loss": 1.389, "step": 61030 }, { "epoch": 2.45, "learning_rate": 0.000439067595313821, "loss": 1.2084, "step": 61040 }, { "epoch": 2.46, "learning_rate": 0.0004390575304963968, "loss": 1.1635, "step": 61050 }, { "epoch": 2.46, "learning_rate": 0.00043904746567897257, "loss": 1.1263, "step": 61060 }, { "epoch": 2.46, "learning_rate": 0.00043903740086154836, "loss": 1.123, "step": 61070 }, { "epoch": 2.46, "learning_rate": 0.00043902733604412415, "loss": 1.2359, "step": 61080 }, { "epoch": 2.46, "learning_rate": 0.00043901727122669994, "loss": 1.0697, "step": 61090 }, { "epoch": 2.46, "learning_rate": 0.0004390072064092758, "loss": 1.1593, "step": 61100 }, { "epoch": 2.46, "learning_rate": 0.0004389971415918516, "loss": 1.2464, "step": 61110 }, { "epoch": 2.46, "learning_rate": 0.0004389870767744273, "loss": 1.3274, "step": 61120 }, { "epoch": 2.46, "learning_rate": 0.0004389770119570031, "loss": 1.12, "step": 61130 }, { "epoch": 2.46, "learning_rate": 0.0004389669471395789, "loss": 1.347, "step": 61140 }, { "epoch": 2.46, "learning_rate": 0.0004389568823221547, "loss": 1.1447, "step": 61150 }, { "epoch": 2.46, "learning_rate": 0.0004389468175047305, "loss": 1.2616, "step": 61160 }, { "epoch": 2.46, "learning_rate": 0.00043893675268730627, "loss": 1.2636, "step": 61170 }, { "epoch": 2.46, "learning_rate": 0.00043892668786988206, "loss": 1.5153, "step": 61180 }, { "epoch": 2.46, "learning_rate": 0.00043891662305245785, "loss": 1.2812, "step": 61190 }, { "epoch": 2.46, "learning_rate": 0.00043890655823503364, "loss": 1.2158, "step": 61200 }, { "epoch": 2.46, "learning_rate": 0.0004388964934176094, "loss": 1.3819, "step": 61210 }, { "epoch": 2.46, "learning_rate": 0.00043888642860018517, "loss": 1.4439, "step": 61220 }, { "epoch": 2.46, "learning_rate": 0.00043887636378276096, "loss": 1.2396, "step": 61230 }, { "epoch": 2.46, "learning_rate": 0.00043886629896533675, "loss": 1.4142, "step": 61240 }, { "epoch": 2.46, "learning_rate": 0.0004388562341479126, "loss": 1.4801, "step": 61250 }, { "epoch": 2.46, "learning_rate": 0.0004388461693304884, "loss": 0.9866, "step": 61260 }, { "epoch": 2.46, "learning_rate": 0.0004388361045130642, "loss": 1.2216, "step": 61270 }, { "epoch": 2.46, "learning_rate": 0.00043882603969563997, "loss": 1.1227, "step": 61280 }, { "epoch": 2.46, "learning_rate": 0.0004388159748782157, "loss": 1.3091, "step": 61290 }, { "epoch": 2.47, "learning_rate": 0.0004388059100607915, "loss": 1.3244, "step": 61300 }, { "epoch": 2.47, "learning_rate": 0.0004387958452433673, "loss": 1.0864, "step": 61310 }, { "epoch": 2.47, "learning_rate": 0.0004387857804259431, "loss": 1.0873, "step": 61320 }, { "epoch": 2.47, "learning_rate": 0.00043877571560851887, "loss": 1.1588, "step": 61330 }, { "epoch": 2.47, "learning_rate": 0.00043876565079109466, "loss": 1.2624, "step": 61340 }, { "epoch": 2.47, "learning_rate": 0.00043875558597367045, "loss": 1.2635, "step": 61350 }, { "epoch": 2.47, "learning_rate": 0.00043874552115624624, "loss": 1.319, "step": 61360 }, { "epoch": 2.47, "learning_rate": 0.00043873545633882203, "loss": 1.4491, "step": 61370 }, { "epoch": 2.47, "learning_rate": 0.00043872539152139777, "loss": 1.309, "step": 61380 }, { "epoch": 2.47, "learning_rate": 0.00043871532670397356, "loss": 1.0953, "step": 61390 }, { "epoch": 2.47, "learning_rate": 0.0004387052618865494, "loss": 1.2059, "step": 61400 }, { "epoch": 2.47, "learning_rate": 0.0004386951970691252, "loss": 1.2504, "step": 61410 }, { "epoch": 2.47, "learning_rate": 0.000438685132251701, "loss": 1.0374, "step": 61420 }, { "epoch": 2.47, "learning_rate": 0.0004386750674342768, "loss": 1.1976, "step": 61430 }, { "epoch": 2.47, "learning_rate": 0.00043866500261685256, "loss": 1.285, "step": 61440 }, { "epoch": 2.47, "learning_rate": 0.00043865493779942836, "loss": 1.2528, "step": 61450 }, { "epoch": 2.47, "learning_rate": 0.0004386448729820041, "loss": 1.1337, "step": 61460 }, { "epoch": 2.47, "learning_rate": 0.0004386348081645799, "loss": 1.124, "step": 61470 }, { "epoch": 2.47, "learning_rate": 0.0004386247433471557, "loss": 1.3355, "step": 61480 }, { "epoch": 2.47, "learning_rate": 0.00043861467852973146, "loss": 1.2197, "step": 61490 }, { "epoch": 2.47, "learning_rate": 0.00043860461371230726, "loss": 1.3379, "step": 61500 }, { "epoch": 2.47, "learning_rate": 0.00043859454889488305, "loss": 1.2812, "step": 61510 }, { "epoch": 2.47, "learning_rate": 0.00043858448407745884, "loss": 1.4418, "step": 61520 }, { "epoch": 2.47, "learning_rate": 0.00043857441926003463, "loss": 1.2015, "step": 61530 }, { "epoch": 2.48, "learning_rate": 0.00043856435444261047, "loss": 1.0161, "step": 61540 }, { "epoch": 2.48, "learning_rate": 0.0004385542896251862, "loss": 1.292, "step": 61550 }, { "epoch": 2.48, "learning_rate": 0.000438544224807762, "loss": 1.3147, "step": 61560 }, { "epoch": 2.48, "learning_rate": 0.0004385341599903378, "loss": 1.1204, "step": 61570 }, { "epoch": 2.48, "learning_rate": 0.0004385240951729136, "loss": 1.3378, "step": 61580 }, { "epoch": 2.48, "learning_rate": 0.00043851403035548937, "loss": 1.1952, "step": 61590 }, { "epoch": 2.48, "learning_rate": 0.00043850396553806516, "loss": 1.4639, "step": 61600 }, { "epoch": 2.48, "learning_rate": 0.00043849390072064095, "loss": 1.175, "step": 61610 }, { "epoch": 2.48, "learning_rate": 0.00043848383590321674, "loss": 1.4802, "step": 61620 }, { "epoch": 2.48, "learning_rate": 0.0004384737710857925, "loss": 1.3019, "step": 61630 }, { "epoch": 2.48, "learning_rate": 0.00043846370626836827, "loss": 1.4279, "step": 61640 }, { "epoch": 2.48, "learning_rate": 0.00043845364145094406, "loss": 1.2238, "step": 61650 }, { "epoch": 2.48, "learning_rate": 0.00043844357663351985, "loss": 1.3737, "step": 61660 }, { "epoch": 2.48, "learning_rate": 0.00043843351181609564, "loss": 1.2633, "step": 61670 }, { "epoch": 2.48, "learning_rate": 0.00043842344699867144, "loss": 1.1597, "step": 61680 }, { "epoch": 2.48, "learning_rate": 0.0004384133821812473, "loss": 0.8779, "step": 61690 }, { "epoch": 2.48, "learning_rate": 0.00043840331736382307, "loss": 1.0793, "step": 61700 }, { "epoch": 2.48, "learning_rate": 0.00043839325254639886, "loss": 1.2819, "step": 61710 }, { "epoch": 2.48, "learning_rate": 0.0004383831877289746, "loss": 1.2926, "step": 61720 }, { "epoch": 2.48, "learning_rate": 0.0004383731229115504, "loss": 1.2498, "step": 61730 }, { "epoch": 2.48, "learning_rate": 0.0004383630580941262, "loss": 1.474, "step": 61740 }, { "epoch": 2.48, "learning_rate": 0.00043835299327670197, "loss": 1.372, "step": 61750 }, { "epoch": 2.48, "learning_rate": 0.00043834292845927776, "loss": 1.2173, "step": 61760 }, { "epoch": 2.48, "learning_rate": 0.00043833286364185355, "loss": 1.1696, "step": 61770 }, { "epoch": 2.48, "learning_rate": 0.00043832279882442934, "loss": 1.1438, "step": 61780 }, { "epoch": 2.49, "learning_rate": 0.00043831273400700513, "loss": 1.4701, "step": 61790 }, { "epoch": 2.49, "learning_rate": 0.00043830266918958087, "loss": 1.0699, "step": 61800 }, { "epoch": 2.49, "learning_rate": 0.00043829260437215666, "loss": 1.2563, "step": 61810 }, { "epoch": 2.49, "learning_rate": 0.00043828253955473245, "loss": 1.2043, "step": 61820 }, { "epoch": 2.49, "learning_rate": 0.0004382724747373083, "loss": 1.3034, "step": 61830 }, { "epoch": 2.49, "learning_rate": 0.0004382624099198841, "loss": 1.1753, "step": 61840 }, { "epoch": 2.49, "learning_rate": 0.0004382523451024599, "loss": 1.348, "step": 61850 }, { "epoch": 2.49, "learning_rate": 0.00043824228028503567, "loss": 1.0893, "step": 61860 }, { "epoch": 2.49, "learning_rate": 0.00043823221546761146, "loss": 1.2364, "step": 61870 }, { "epoch": 2.49, "learning_rate": 0.0004382221506501872, "loss": 1.2126, "step": 61880 }, { "epoch": 2.49, "learning_rate": 0.000438212085832763, "loss": 1.112, "step": 61890 }, { "epoch": 2.49, "learning_rate": 0.0004382020210153388, "loss": 1.2519, "step": 61900 }, { "epoch": 2.49, "learning_rate": 0.00043819195619791457, "loss": 1.5005, "step": 61910 }, { "epoch": 2.49, "learning_rate": 0.00043818189138049036, "loss": 1.1635, "step": 61920 }, { "epoch": 2.49, "learning_rate": 0.00043817182656306615, "loss": 1.396, "step": 61930 }, { "epoch": 2.49, "learning_rate": 0.00043816176174564194, "loss": 1.1748, "step": 61940 }, { "epoch": 2.49, "learning_rate": 0.00043815169692821773, "loss": 1.3944, "step": 61950 }, { "epoch": 2.49, "learning_rate": 0.0004381416321107935, "loss": 1.5547, "step": 61960 }, { "epoch": 2.49, "learning_rate": 0.00043813156729336926, "loss": 1.1937, "step": 61970 }, { "epoch": 2.49, "learning_rate": 0.0004381215024759451, "loss": 1.3292, "step": 61980 }, { "epoch": 2.49, "learning_rate": 0.0004381114376585209, "loss": 1.2905, "step": 61990 }, { "epoch": 2.49, "learning_rate": 0.0004381013728410967, "loss": 1.195, "step": 62000 }, { "epoch": 2.49, "learning_rate": 0.0004380913080236725, "loss": 1.1708, "step": 62010 }, { "epoch": 2.49, "learning_rate": 0.00043808124320624827, "loss": 1.5789, "step": 62020 }, { "epoch": 2.49, "learning_rate": 0.00043807117838882406, "loss": 1.1597, "step": 62030 }, { "epoch": 2.5, "learning_rate": 0.00043806111357139985, "loss": 1.3157, "step": 62040 }, { "epoch": 2.5, "learning_rate": 0.0004380510487539756, "loss": 1.5075, "step": 62050 }, { "epoch": 2.5, "learning_rate": 0.0004380409839365514, "loss": 1.3708, "step": 62060 }, { "epoch": 2.5, "learning_rate": 0.00043803091911912717, "loss": 1.3076, "step": 62070 }, { "epoch": 2.5, "learning_rate": 0.00043802085430170296, "loss": 1.1407, "step": 62080 }, { "epoch": 2.5, "learning_rate": 0.00043801078948427875, "loss": 1.1321, "step": 62090 }, { "epoch": 2.5, "learning_rate": 0.00043800072466685454, "loss": 1.1594, "step": 62100 }, { "epoch": 2.5, "learning_rate": 0.00043799065984943033, "loss": 1.1331, "step": 62110 }, { "epoch": 2.5, "learning_rate": 0.0004379805950320061, "loss": 1.4953, "step": 62120 }, { "epoch": 2.5, "learning_rate": 0.00043797053021458197, "loss": 1.1717, "step": 62130 }, { "epoch": 2.5, "learning_rate": 0.0004379604653971577, "loss": 1.3456, "step": 62140 }, { "epoch": 2.5, "learning_rate": 0.0004379504005797335, "loss": 1.1641, "step": 62150 }, { "epoch": 2.5, "learning_rate": 0.0004379403357623093, "loss": 1.2878, "step": 62160 }, { "epoch": 2.5, "learning_rate": 0.0004379302709448851, "loss": 1.3216, "step": 62170 }, { "epoch": 2.5, "learning_rate": 0.00043792020612746087, "loss": 1.1628, "step": 62180 }, { "epoch": 2.5, "learning_rate": 0.00043791014131003666, "loss": 1.3035, "step": 62190 }, { "epoch": 2.5, "learning_rate": 0.00043790007649261245, "loss": 1.1348, "step": 62200 }, { "epoch": 2.5, "learning_rate": 0.00043789001167518824, "loss": 1.0726, "step": 62210 }, { "epoch": 2.5, "learning_rate": 0.000437879946857764, "loss": 1.2304, "step": 62220 }, { "epoch": 2.5, "learning_rate": 0.00043786988204033977, "loss": 1.2324, "step": 62230 }, { "epoch": 2.5, "learning_rate": 0.00043785981722291556, "loss": 1.1968, "step": 62240 }, { "epoch": 2.5, "learning_rate": 0.00043784975240549135, "loss": 1.4652, "step": 62250 }, { "epoch": 2.5, "learning_rate": 0.00043783968758806714, "loss": 1.2268, "step": 62260 }, { "epoch": 2.5, "learning_rate": 0.000437829622770643, "loss": 1.212, "step": 62270 }, { "epoch": 2.5, "learning_rate": 0.0004378195579532188, "loss": 1.1272, "step": 62280 }, { "epoch": 2.51, "learning_rate": 0.00043780949313579456, "loss": 1.1962, "step": 62290 }, { "epoch": 2.51, "learning_rate": 0.00043779942831837036, "loss": 0.9908, "step": 62300 }, { "epoch": 2.51, "learning_rate": 0.0004377893635009461, "loss": 1.2072, "step": 62310 }, { "epoch": 2.51, "learning_rate": 0.0004377792986835219, "loss": 1.3935, "step": 62320 }, { "epoch": 2.51, "learning_rate": 0.0004377692338660977, "loss": 1.5153, "step": 62330 }, { "epoch": 2.51, "learning_rate": 0.00043775916904867346, "loss": 1.287, "step": 62340 }, { "epoch": 2.51, "learning_rate": 0.00043774910423124925, "loss": 1.2685, "step": 62350 }, { "epoch": 2.51, "learning_rate": 0.00043773903941382505, "loss": 1.2018, "step": 62360 }, { "epoch": 2.51, "learning_rate": 0.00043772897459640084, "loss": 1.0966, "step": 62370 }, { "epoch": 2.51, "learning_rate": 0.00043771890977897663, "loss": 1.5829, "step": 62380 }, { "epoch": 2.51, "learning_rate": 0.00043770884496155236, "loss": 1.2767, "step": 62390 }, { "epoch": 2.51, "learning_rate": 0.00043769878014412815, "loss": 1.3486, "step": 62400 }, { "epoch": 2.51, "learning_rate": 0.00043768871532670395, "loss": 1.3155, "step": 62410 }, { "epoch": 2.51, "learning_rate": 0.0004376786505092798, "loss": 1.2753, "step": 62420 }, { "epoch": 2.51, "learning_rate": 0.0004376685856918556, "loss": 1.2436, "step": 62430 }, { "epoch": 2.51, "learning_rate": 0.00043765852087443137, "loss": 1.4815, "step": 62440 }, { "epoch": 2.51, "learning_rate": 0.00043764845605700716, "loss": 1.0211, "step": 62450 }, { "epoch": 2.51, "learning_rate": 0.00043763839123958295, "loss": 1.235, "step": 62460 }, { "epoch": 2.51, "learning_rate": 0.00043762832642215874, "loss": 1.4147, "step": 62470 }, { "epoch": 2.51, "learning_rate": 0.0004376182616047345, "loss": 1.4006, "step": 62480 }, { "epoch": 2.51, "learning_rate": 0.00043760819678731027, "loss": 1.0904, "step": 62490 }, { "epoch": 2.51, "learning_rate": 0.00043759813196988606, "loss": 1.3038, "step": 62500 }, { "epoch": 2.51, "learning_rate": 0.00043758806715246185, "loss": 1.1303, "step": 62510 }, { "epoch": 2.51, "learning_rate": 0.00043757800233503764, "loss": 1.2989, "step": 62520 }, { "epoch": 2.51, "learning_rate": 0.00043756793751761343, "loss": 1.2867, "step": 62530 }, { "epoch": 2.52, "learning_rate": 0.0004375578727001892, "loss": 1.4966, "step": 62540 }, { "epoch": 2.52, "learning_rate": 0.000437547807882765, "loss": 1.0531, "step": 62550 }, { "epoch": 2.52, "learning_rate": 0.00043753774306534075, "loss": 1.3877, "step": 62560 }, { "epoch": 2.52, "learning_rate": 0.0004375276782479166, "loss": 1.1807, "step": 62570 }, { "epoch": 2.52, "learning_rate": 0.0004375176134304924, "loss": 1.1668, "step": 62580 }, { "epoch": 2.52, "learning_rate": 0.0004375075486130682, "loss": 1.1332, "step": 62590 }, { "epoch": 2.52, "learning_rate": 0.00043749748379564397, "loss": 1.4412, "step": 62600 }, { "epoch": 2.52, "learning_rate": 0.00043748741897821976, "loss": 1.3233, "step": 62610 }, { "epoch": 2.52, "learning_rate": 0.00043747735416079555, "loss": 1.3185, "step": 62620 }, { "epoch": 2.52, "learning_rate": 0.00043746728934337134, "loss": 1.3936, "step": 62630 }, { "epoch": 2.52, "learning_rate": 0.0004374572245259471, "loss": 1.4013, "step": 62640 }, { "epoch": 2.52, "learning_rate": 0.00043744715970852287, "loss": 1.0018, "step": 62650 }, { "epoch": 2.52, "learning_rate": 0.00043743709489109866, "loss": 1.3264, "step": 62660 }, { "epoch": 2.52, "learning_rate": 0.00043742703007367445, "loss": 1.421, "step": 62670 }, { "epoch": 2.52, "learning_rate": 0.00043741696525625024, "loss": 1.0697, "step": 62680 }, { "epoch": 2.52, "learning_rate": 0.00043740690043882603, "loss": 1.2869, "step": 62690 }, { "epoch": 2.52, "learning_rate": 0.0004373968356214018, "loss": 1.6505, "step": 62700 }, { "epoch": 2.52, "learning_rate": 0.00043738677080397767, "loss": 1.462, "step": 62710 }, { "epoch": 2.52, "learning_rate": 0.00043737670598655346, "loss": 1.4143, "step": 62720 }, { "epoch": 2.52, "learning_rate": 0.0004373666411691292, "loss": 1.3037, "step": 62730 }, { "epoch": 2.52, "learning_rate": 0.000437356576351705, "loss": 1.0952, "step": 62740 }, { "epoch": 2.52, "learning_rate": 0.0004373465115342808, "loss": 1.1762, "step": 62750 }, { "epoch": 2.52, "learning_rate": 0.00043733644671685657, "loss": 1.1607, "step": 62760 }, { "epoch": 2.52, "learning_rate": 0.00043732638189943236, "loss": 1.3574, "step": 62770 }, { "epoch": 2.52, "learning_rate": 0.00043731631708200815, "loss": 1.5411, "step": 62780 }, { "epoch": 2.53, "learning_rate": 0.00043730625226458394, "loss": 1.0902, "step": 62790 }, { "epoch": 2.53, "learning_rate": 0.00043729618744715973, "loss": 1.2093, "step": 62800 }, { "epoch": 2.53, "learning_rate": 0.00043728612262973547, "loss": 1.1965, "step": 62810 }, { "epoch": 2.53, "learning_rate": 0.00043727605781231126, "loss": 1.4396, "step": 62820 }, { "epoch": 2.53, "learning_rate": 0.00043726599299488705, "loss": 1.1661, "step": 62830 }, { "epoch": 2.53, "learning_rate": 0.00043725592817746284, "loss": 1.5922, "step": 62840 }, { "epoch": 2.53, "learning_rate": 0.00043724586336003863, "loss": 1.2146, "step": 62850 }, { "epoch": 2.53, "learning_rate": 0.0004372357985426145, "loss": 1.5899, "step": 62860 }, { "epoch": 2.53, "learning_rate": 0.00043722573372519027, "loss": 1.3424, "step": 62870 }, { "epoch": 2.53, "learning_rate": 0.00043721566890776606, "loss": 1.2884, "step": 62880 }, { "epoch": 2.53, "learning_rate": 0.00043720560409034185, "loss": 1.4319, "step": 62890 }, { "epoch": 2.53, "learning_rate": 0.0004371955392729176, "loss": 1.1947, "step": 62900 }, { "epoch": 2.53, "learning_rate": 0.0004371854744554934, "loss": 1.2851, "step": 62910 }, { "epoch": 2.53, "learning_rate": 0.00043717540963806917, "loss": 1.1917, "step": 62920 }, { "epoch": 2.53, "learning_rate": 0.00043716534482064496, "loss": 1.1235, "step": 62930 }, { "epoch": 2.53, "learning_rate": 0.00043715528000322075, "loss": 0.9981, "step": 62940 }, { "epoch": 2.53, "learning_rate": 0.00043714521518579654, "loss": 1.0027, "step": 62950 }, { "epoch": 2.53, "learning_rate": 0.00043713515036837233, "loss": 1.0449, "step": 62960 }, { "epoch": 2.53, "learning_rate": 0.0004371250855509481, "loss": 1.3833, "step": 62970 }, { "epoch": 2.53, "learning_rate": 0.00043711502073352386, "loss": 1.1795, "step": 62980 }, { "epoch": 2.53, "learning_rate": 0.00043710495591609965, "loss": 1.4395, "step": 62990 }, { "epoch": 2.53, "learning_rate": 0.0004370948910986755, "loss": 1.4532, "step": 63000 }, { "epoch": 2.53, "learning_rate": 0.0004370848262812513, "loss": 1.1533, "step": 63010 }, { "epoch": 2.53, "learning_rate": 0.0004370747614638271, "loss": 1.228, "step": 63020 }, { "epoch": 2.53, "learning_rate": 0.00043706469664640287, "loss": 1.2235, "step": 63030 }, { "epoch": 2.54, "learning_rate": 0.00043705463182897866, "loss": 1.1291, "step": 63040 }, { "epoch": 2.54, "learning_rate": 0.00043704456701155445, "loss": 1.367, "step": 63050 }, { "epoch": 2.54, "learning_rate": 0.00043703450219413024, "loss": 1.323, "step": 63060 }, { "epoch": 2.54, "learning_rate": 0.000437024437376706, "loss": 1.0086, "step": 63070 }, { "epoch": 2.54, "learning_rate": 0.00043701437255928177, "loss": 1.1619, "step": 63080 }, { "epoch": 2.54, "learning_rate": 0.00043700430774185756, "loss": 1.3134, "step": 63090 }, { "epoch": 2.54, "learning_rate": 0.00043699424292443335, "loss": 1.4567, "step": 63100 }, { "epoch": 2.54, "learning_rate": 0.00043698417810700914, "loss": 1.1916, "step": 63110 }, { "epoch": 2.54, "learning_rate": 0.00043697411328958493, "loss": 1.1126, "step": 63120 }, { "epoch": 2.54, "learning_rate": 0.0004369640484721607, "loss": 1.3781, "step": 63130 }, { "epoch": 2.54, "learning_rate": 0.0004369539836547365, "loss": 1.3215, "step": 63140 }, { "epoch": 2.54, "learning_rate": 0.0004369439188373123, "loss": 1.2446, "step": 63150 }, { "epoch": 2.54, "learning_rate": 0.0004369338540198881, "loss": 1.2651, "step": 63160 }, { "epoch": 2.54, "learning_rate": 0.0004369237892024639, "loss": 1.5314, "step": 63170 }, { "epoch": 2.54, "learning_rate": 0.0004369137243850397, "loss": 1.2188, "step": 63180 }, { "epoch": 2.54, "learning_rate": 0.00043690365956761546, "loss": 1.484, "step": 63190 }, { "epoch": 2.54, "learning_rate": 0.00043689359475019125, "loss": 1.2234, "step": 63200 }, { "epoch": 2.54, "learning_rate": 0.00043688352993276705, "loss": 1.097, "step": 63210 }, { "epoch": 2.54, "learning_rate": 0.00043687346511534284, "loss": 1.175, "step": 63220 }, { "epoch": 2.54, "learning_rate": 0.0004368634002979186, "loss": 1.0474, "step": 63230 }, { "epoch": 2.54, "learning_rate": 0.00043685333548049436, "loss": 1.2779, "step": 63240 }, { "epoch": 2.54, "learning_rate": 0.00043684327066307015, "loss": 1.1173, "step": 63250 }, { "epoch": 2.54, "learning_rate": 0.00043683320584564595, "loss": 1.299, "step": 63260 }, { "epoch": 2.54, "learning_rate": 0.00043682314102822174, "loss": 1.2431, "step": 63270 }, { "epoch": 2.55, "learning_rate": 0.0004368130762107975, "loss": 1.3099, "step": 63280 }, { "epoch": 2.55, "learning_rate": 0.0004368030113933733, "loss": 1.1917, "step": 63290 }, { "epoch": 2.55, "learning_rate": 0.00043679294657594916, "loss": 1.2434, "step": 63300 }, { "epoch": 2.55, "learning_rate": 0.00043678288175852495, "loss": 1.0858, "step": 63310 }, { "epoch": 2.55, "learning_rate": 0.0004367728169411007, "loss": 1.6648, "step": 63320 }, { "epoch": 2.55, "learning_rate": 0.0004367627521236765, "loss": 1.3465, "step": 63330 }, { "epoch": 2.55, "learning_rate": 0.00043675268730625227, "loss": 1.4164, "step": 63340 }, { "epoch": 2.55, "learning_rate": 0.00043674262248882806, "loss": 1.3178, "step": 63350 }, { "epoch": 2.55, "learning_rate": 0.00043673255767140385, "loss": 1.3335, "step": 63360 }, { "epoch": 2.55, "learning_rate": 0.00043672249285397964, "loss": 1.3557, "step": 63370 }, { "epoch": 2.55, "learning_rate": 0.00043671242803655543, "loss": 1.1394, "step": 63380 }, { "epoch": 2.55, "learning_rate": 0.0004367023632191312, "loss": 1.6099, "step": 63390 }, { "epoch": 2.55, "learning_rate": 0.00043669229840170696, "loss": 1.4254, "step": 63400 }, { "epoch": 2.55, "learning_rate": 0.00043668223358428275, "loss": 1.1455, "step": 63410 }, { "epoch": 2.55, "learning_rate": 0.00043667216876685854, "loss": 1.3452, "step": 63420 }, { "epoch": 2.55, "learning_rate": 0.00043666210394943433, "loss": 1.382, "step": 63430 }, { "epoch": 2.55, "learning_rate": 0.0004366520391320102, "loss": 1.3096, "step": 63440 }, { "epoch": 2.55, "learning_rate": 0.00043664197431458597, "loss": 1.1205, "step": 63450 }, { "epoch": 2.55, "learning_rate": 0.00043663190949716176, "loss": 1.2541, "step": 63460 }, { "epoch": 2.55, "learning_rate": 0.00043662184467973755, "loss": 1.2476, "step": 63470 }, { "epoch": 2.55, "learning_rate": 0.00043661177986231334, "loss": 1.1908, "step": 63480 }, { "epoch": 2.55, "learning_rate": 0.0004366017150448891, "loss": 1.3013, "step": 63490 }, { "epoch": 2.55, "learning_rate": 0.00043659165022746487, "loss": 1.3811, "step": 63500 }, { "epoch": 2.55, "learning_rate": 0.00043658158541004066, "loss": 1.3764, "step": 63510 }, { "epoch": 2.55, "learning_rate": 0.00043657152059261645, "loss": 1.287, "step": 63520 }, { "epoch": 2.56, "learning_rate": 0.00043656145577519224, "loss": 1.2975, "step": 63530 }, { "epoch": 2.56, "learning_rate": 0.00043655139095776803, "loss": 1.0316, "step": 63540 }, { "epoch": 2.56, "learning_rate": 0.0004365413261403438, "loss": 1.2044, "step": 63550 }, { "epoch": 2.56, "learning_rate": 0.0004365312613229196, "loss": 1.181, "step": 63560 }, { "epoch": 2.56, "learning_rate": 0.00043652119650549535, "loss": 1.1915, "step": 63570 }, { "epoch": 2.56, "learning_rate": 0.00043651113168807114, "loss": 1.2676, "step": 63580 }, { "epoch": 2.56, "learning_rate": 0.000436501066870647, "loss": 1.3871, "step": 63590 }, { "epoch": 2.56, "learning_rate": 0.0004364910020532228, "loss": 1.1867, "step": 63600 }, { "epoch": 2.56, "learning_rate": 0.00043648093723579857, "loss": 1.1185, "step": 63610 }, { "epoch": 2.56, "learning_rate": 0.00043647087241837436, "loss": 1.2004, "step": 63620 }, { "epoch": 2.56, "learning_rate": 0.00043646080760095015, "loss": 1.2187, "step": 63630 }, { "epoch": 2.56, "learning_rate": 0.00043645074278352594, "loss": 1.5803, "step": 63640 }, { "epoch": 2.56, "learning_rate": 0.00043644067796610173, "loss": 1.2624, "step": 63650 }, { "epoch": 2.56, "learning_rate": 0.00043643061314867747, "loss": 1.2976, "step": 63660 }, { "epoch": 2.56, "learning_rate": 0.00043642054833125326, "loss": 1.2838, "step": 63670 }, { "epoch": 2.56, "learning_rate": 0.00043641048351382905, "loss": 1.2459, "step": 63680 }, { "epoch": 2.56, "learning_rate": 0.00043640041869640484, "loss": 1.1253, "step": 63690 }, { "epoch": 2.56, "learning_rate": 0.00043639035387898063, "loss": 1.2939, "step": 63700 }, { "epoch": 2.56, "learning_rate": 0.0004363802890615564, "loss": 1.2279, "step": 63710 }, { "epoch": 2.56, "learning_rate": 0.0004363702242441322, "loss": 0.9794, "step": 63720 }, { "epoch": 2.56, "learning_rate": 0.000436360159426708, "loss": 1.1066, "step": 63730 }, { "epoch": 2.56, "learning_rate": 0.0004363500946092838, "loss": 1.1735, "step": 63740 }, { "epoch": 2.56, "learning_rate": 0.0004363400297918596, "loss": 1.2832, "step": 63750 }, { "epoch": 2.56, "learning_rate": 0.0004363299649744354, "loss": 1.237, "step": 63760 }, { "epoch": 2.56, "learning_rate": 0.00043631990015701117, "loss": 1.0705, "step": 63770 }, { "epoch": 2.57, "learning_rate": 0.00043630983533958696, "loss": 1.4068, "step": 63780 }, { "epoch": 2.57, "learning_rate": 0.00043629977052216275, "loss": 1.3432, "step": 63790 }, { "epoch": 2.57, "learning_rate": 0.00043628970570473854, "loss": 0.987, "step": 63800 }, { "epoch": 2.57, "learning_rate": 0.00043627964088731433, "loss": 1.3617, "step": 63810 }, { "epoch": 2.57, "learning_rate": 0.0004362695760698901, "loss": 1.2721, "step": 63820 }, { "epoch": 2.57, "learning_rate": 0.00043625951125246586, "loss": 1.4504, "step": 63830 }, { "epoch": 2.57, "learning_rate": 0.00043624944643504165, "loss": 1.147, "step": 63840 }, { "epoch": 2.57, "learning_rate": 0.00043623938161761744, "loss": 1.0707, "step": 63850 }, { "epoch": 2.57, "learning_rate": 0.00043622931680019323, "loss": 1.4692, "step": 63860 }, { "epoch": 2.57, "learning_rate": 0.000436219251982769, "loss": 1.2326, "step": 63870 }, { "epoch": 2.57, "learning_rate": 0.00043620918716534487, "loss": 1.0793, "step": 63880 }, { "epoch": 2.57, "learning_rate": 0.00043619912234792066, "loss": 1.2319, "step": 63890 }, { "epoch": 2.57, "learning_rate": 0.00043618905753049645, "loss": 1.5179, "step": 63900 }, { "epoch": 2.57, "learning_rate": 0.0004361789927130722, "loss": 1.148, "step": 63910 }, { "epoch": 2.57, "learning_rate": 0.000436168927895648, "loss": 1.4453, "step": 63920 }, { "epoch": 2.57, "learning_rate": 0.00043615886307822377, "loss": 1.4398, "step": 63930 }, { "epoch": 2.57, "learning_rate": 0.00043614879826079956, "loss": 1.366, "step": 63940 }, { "epoch": 2.57, "learning_rate": 0.00043613873344337535, "loss": 1.1631, "step": 63950 }, { "epoch": 2.57, "learning_rate": 0.00043612866862595114, "loss": 1.3371, "step": 63960 }, { "epoch": 2.57, "learning_rate": 0.00043611860380852693, "loss": 1.1058, "step": 63970 }, { "epoch": 2.57, "learning_rate": 0.0004361085389911027, "loss": 1.2238, "step": 63980 }, { "epoch": 2.57, "learning_rate": 0.0004360984741736785, "loss": 1.4046, "step": 63990 }, { "epoch": 2.57, "learning_rate": 0.00043608840935625425, "loss": 1.4406, "step": 64000 }, { "epoch": 2.57, "learning_rate": 0.00043607834453883004, "loss": 1.3037, "step": 64010 }, { "epoch": 2.57, "learning_rate": 0.00043606827972140583, "loss": 1.2134, "step": 64020 }, { "epoch": 2.58, "learning_rate": 0.0004360582149039817, "loss": 1.0652, "step": 64030 }, { "epoch": 2.58, "learning_rate": 0.00043604815008655746, "loss": 1.0747, "step": 64040 }, { "epoch": 2.58, "learning_rate": 0.00043603808526913325, "loss": 1.0782, "step": 64050 }, { "epoch": 2.58, "learning_rate": 0.00043602802045170905, "loss": 1.2928, "step": 64060 }, { "epoch": 2.58, "learning_rate": 0.00043601795563428484, "loss": 1.5477, "step": 64070 }, { "epoch": 2.58, "learning_rate": 0.00043600789081686057, "loss": 1.1749, "step": 64080 }, { "epoch": 2.58, "learning_rate": 0.00043599782599943636, "loss": 1.5074, "step": 64090 }, { "epoch": 2.58, "learning_rate": 0.00043598776118201215, "loss": 1.0316, "step": 64100 }, { "epoch": 2.58, "learning_rate": 0.00043597769636458794, "loss": 1.2556, "step": 64110 }, { "epoch": 2.58, "learning_rate": 0.00043596763154716374, "loss": 1.3151, "step": 64120 }, { "epoch": 2.58, "learning_rate": 0.0004359575667297395, "loss": 1.1934, "step": 64130 }, { "epoch": 2.58, "learning_rate": 0.0004359475019123153, "loss": 1.2792, "step": 64140 }, { "epoch": 2.58, "learning_rate": 0.0004359374370948911, "loss": 1.2004, "step": 64150 }, { "epoch": 2.58, "learning_rate": 0.00043592737227746684, "loss": 1.2908, "step": 64160 }, { "epoch": 2.58, "learning_rate": 0.0004359173074600427, "loss": 0.9603, "step": 64170 }, { "epoch": 2.58, "learning_rate": 0.0004359072426426185, "loss": 1.3278, "step": 64180 }, { "epoch": 2.58, "learning_rate": 0.00043589717782519427, "loss": 0.9919, "step": 64190 }, { "epoch": 2.58, "learning_rate": 0.00043588711300777006, "loss": 1.1326, "step": 64200 }, { "epoch": 2.58, "learning_rate": 0.00043587704819034585, "loss": 1.2794, "step": 64210 }, { "epoch": 2.58, "learning_rate": 0.00043586698337292164, "loss": 1.1647, "step": 64220 }, { "epoch": 2.58, "learning_rate": 0.00043585691855549743, "loss": 1.1622, "step": 64230 }, { "epoch": 2.58, "learning_rate": 0.0004358468537380732, "loss": 1.1824, "step": 64240 }, { "epoch": 2.58, "learning_rate": 0.00043583678892064896, "loss": 1.4586, "step": 64250 }, { "epoch": 2.58, "learning_rate": 0.00043582672410322475, "loss": 1.3918, "step": 64260 }, { "epoch": 2.58, "learning_rate": 0.00043581665928580054, "loss": 1.121, "step": 64270 }, { "epoch": 2.59, "learning_rate": 0.00043580659446837633, "loss": 1.4518, "step": 64280 }, { "epoch": 2.59, "learning_rate": 0.0004357965296509521, "loss": 1.2143, "step": 64290 }, { "epoch": 2.59, "learning_rate": 0.0004357864648335279, "loss": 1.1521, "step": 64300 }, { "epoch": 2.59, "learning_rate": 0.0004357764000161037, "loss": 1.2625, "step": 64310 }, { "epoch": 2.59, "learning_rate": 0.00043576633519867955, "loss": 1.5325, "step": 64320 }, { "epoch": 2.59, "learning_rate": 0.00043575627038125534, "loss": 1.2299, "step": 64330 }, { "epoch": 2.59, "learning_rate": 0.0004357462055638311, "loss": 1.3118, "step": 64340 }, { "epoch": 2.59, "learning_rate": 0.00043573614074640687, "loss": 1.2734, "step": 64350 }, { "epoch": 2.59, "learning_rate": 0.00043572607592898266, "loss": 0.8586, "step": 64360 }, { "epoch": 2.59, "learning_rate": 0.00043571601111155845, "loss": 1.32, "step": 64370 }, { "epoch": 2.59, "learning_rate": 0.00043570594629413424, "loss": 1.1305, "step": 64380 }, { "epoch": 2.59, "learning_rate": 0.00043569588147671003, "loss": 1.5073, "step": 64390 }, { "epoch": 2.59, "learning_rate": 0.0004356858166592858, "loss": 1.3153, "step": 64400 }, { "epoch": 2.59, "learning_rate": 0.0004356757518418616, "loss": 1.3254, "step": 64410 }, { "epoch": 2.59, "learning_rate": 0.00043566568702443735, "loss": 1.4305, "step": 64420 }, { "epoch": 2.59, "learning_rate": 0.00043565562220701314, "loss": 1.2682, "step": 64430 }, { "epoch": 2.59, "learning_rate": 0.00043564555738958893, "loss": 1.2098, "step": 64440 }, { "epoch": 2.59, "learning_rate": 0.0004356354925721647, "loss": 1.2225, "step": 64450 }, { "epoch": 2.59, "learning_rate": 0.0004356254277547405, "loss": 1.3459, "step": 64460 }, { "epoch": 2.59, "learning_rate": 0.00043561536293731636, "loss": 1.2261, "step": 64470 }, { "epoch": 2.59, "learning_rate": 0.00043560529811989215, "loss": 1.5363, "step": 64480 }, { "epoch": 2.59, "learning_rate": 0.00043559523330246794, "loss": 1.3251, "step": 64490 }, { "epoch": 2.59, "learning_rate": 0.0004355851684850437, "loss": 1.4389, "step": 64500 }, { "epoch": 2.59, "learning_rate": 0.00043557510366761947, "loss": 1.3105, "step": 64510 }, { "epoch": 2.59, "learning_rate": 0.00043556503885019526, "loss": 1.4637, "step": 64520 }, { "epoch": 2.6, "learning_rate": 0.00043555497403277105, "loss": 1.5429, "step": 64530 }, { "epoch": 2.6, "learning_rate": 0.00043554490921534684, "loss": 1.4549, "step": 64540 }, { "epoch": 2.6, "learning_rate": 0.00043553484439792263, "loss": 1.5332, "step": 64550 }, { "epoch": 2.6, "learning_rate": 0.0004355247795804984, "loss": 1.1254, "step": 64560 }, { "epoch": 2.6, "learning_rate": 0.0004355147147630742, "loss": 1.1014, "step": 64570 }, { "epoch": 2.6, "learning_rate": 0.00043550464994565, "loss": 1.2793, "step": 64580 }, { "epoch": 2.6, "learning_rate": 0.00043549458512822574, "loss": 1.2456, "step": 64590 }, { "epoch": 2.6, "learning_rate": 0.00043548452031080153, "loss": 1.2284, "step": 64600 }, { "epoch": 2.6, "learning_rate": 0.0004354744554933774, "loss": 1.2726, "step": 64610 }, { "epoch": 2.6, "learning_rate": 0.00043546439067595317, "loss": 1.2522, "step": 64620 }, { "epoch": 2.6, "learning_rate": 0.00043545432585852896, "loss": 1.1593, "step": 64630 }, { "epoch": 2.6, "learning_rate": 0.00043544426104110475, "loss": 1.1992, "step": 64640 }, { "epoch": 2.6, "learning_rate": 0.00043543419622368054, "loss": 1.2496, "step": 64650 }, { "epoch": 2.6, "learning_rate": 0.00043542413140625633, "loss": 1.4045, "step": 64660 }, { "epoch": 2.6, "learning_rate": 0.00043541406658883207, "loss": 1.0217, "step": 64670 }, { "epoch": 2.6, "learning_rate": 0.00043540400177140786, "loss": 1.2751, "step": 64680 }, { "epoch": 2.6, "learning_rate": 0.00043539393695398365, "loss": 1.2523, "step": 64690 }, { "epoch": 2.6, "learning_rate": 0.00043538387213655944, "loss": 1.1279, "step": 64700 }, { "epoch": 2.6, "learning_rate": 0.00043537380731913523, "loss": 1.2128, "step": 64710 }, { "epoch": 2.6, "learning_rate": 0.000435363742501711, "loss": 1.2003, "step": 64720 }, { "epoch": 2.6, "learning_rate": 0.0004353536776842868, "loss": 1.2083, "step": 64730 }, { "epoch": 2.6, "learning_rate": 0.0004353436128668626, "loss": 1.2945, "step": 64740 }, { "epoch": 2.6, "learning_rate": 0.0004353335480494384, "loss": 1.2461, "step": 64750 }, { "epoch": 2.6, "learning_rate": 0.0004353234832320142, "loss": 1.2802, "step": 64760 }, { "epoch": 2.6, "learning_rate": 0.00043531341841459, "loss": 1.3644, "step": 64770 }, { "epoch": 2.61, "learning_rate": 0.00043530335359716576, "loss": 1.2391, "step": 64780 }, { "epoch": 2.61, "learning_rate": 0.00043529328877974156, "loss": 1.3204, "step": 64790 }, { "epoch": 2.61, "learning_rate": 0.00043528322396231735, "loss": 1.2551, "step": 64800 }, { "epoch": 2.61, "learning_rate": 0.00043527315914489314, "loss": 1.506, "step": 64810 }, { "epoch": 2.61, "learning_rate": 0.00043526309432746893, "loss": 1.3625, "step": 64820 }, { "epoch": 2.61, "learning_rate": 0.0004352530295100447, "loss": 1.0902, "step": 64830 }, { "epoch": 2.61, "learning_rate": 0.00043524296469262046, "loss": 1.2109, "step": 64840 }, { "epoch": 2.61, "learning_rate": 0.00043523289987519625, "loss": 1.3423, "step": 64850 }, { "epoch": 2.61, "learning_rate": 0.00043522283505777204, "loss": 1.2231, "step": 64860 }, { "epoch": 2.61, "learning_rate": 0.00043521277024034783, "loss": 1.0502, "step": 64870 }, { "epoch": 2.61, "learning_rate": 0.0004352027054229236, "loss": 1.4417, "step": 64880 }, { "epoch": 2.61, "learning_rate": 0.0004351926406054994, "loss": 1.1811, "step": 64890 }, { "epoch": 2.61, "learning_rate": 0.0004351825757880752, "loss": 1.3319, "step": 64900 }, { "epoch": 2.61, "learning_rate": 0.00043517251097065105, "loss": 1.0745, "step": 64910 }, { "epoch": 2.61, "learning_rate": 0.00043516244615322684, "loss": 1.4172, "step": 64920 }, { "epoch": 2.61, "learning_rate": 0.00043515238133580257, "loss": 1.3387, "step": 64930 }, { "epoch": 2.61, "learning_rate": 0.00043514231651837836, "loss": 1.2048, "step": 64940 }, { "epoch": 2.61, "learning_rate": 0.00043513225170095415, "loss": 1.4653, "step": 64950 }, { "epoch": 2.61, "learning_rate": 0.00043512218688352994, "loss": 1.3682, "step": 64960 }, { "epoch": 2.61, "learning_rate": 0.00043511212206610574, "loss": 1.1889, "step": 64970 }, { "epoch": 2.61, "learning_rate": 0.0004351020572486815, "loss": 1.1103, "step": 64980 }, { "epoch": 2.61, "learning_rate": 0.0004350919924312573, "loss": 1.0514, "step": 64990 }, { "epoch": 2.61, "learning_rate": 0.0004350819276138331, "loss": 1.1933, "step": 65000 }, { "epoch": 2.61, "learning_rate": 0.00043507186279640884, "loss": 1.3032, "step": 65010 }, { "epoch": 2.61, "learning_rate": 0.00043506179797898464, "loss": 1.1221, "step": 65020 }, { "epoch": 2.62, "learning_rate": 0.0004350517331615604, "loss": 1.2248, "step": 65030 }, { "epoch": 2.62, "learning_rate": 0.0004350416683441362, "loss": 1.3811, "step": 65040 }, { "epoch": 2.62, "learning_rate": 0.00043503160352671206, "loss": 1.5026, "step": 65050 }, { "epoch": 2.62, "learning_rate": 0.00043502153870928785, "loss": 1.2814, "step": 65060 }, { "epoch": 2.62, "learning_rate": 0.00043501147389186364, "loss": 1.0509, "step": 65070 }, { "epoch": 2.62, "learning_rate": 0.00043500140907443943, "loss": 1.3133, "step": 65080 }, { "epoch": 2.62, "learning_rate": 0.0004349913442570152, "loss": 1.196, "step": 65090 }, { "epoch": 2.62, "learning_rate": 0.00043498127943959096, "loss": 1.1793, "step": 65100 }, { "epoch": 2.62, "learning_rate": 0.00043497121462216675, "loss": 1.1134, "step": 65110 }, { "epoch": 2.62, "learning_rate": 0.00043496114980474254, "loss": 1.0284, "step": 65120 }, { "epoch": 2.62, "learning_rate": 0.00043495108498731833, "loss": 1.3646, "step": 65130 }, { "epoch": 2.62, "learning_rate": 0.0004349410201698941, "loss": 1.5575, "step": 65140 }, { "epoch": 2.62, "learning_rate": 0.0004349309553524699, "loss": 1.1785, "step": 65150 }, { "epoch": 2.62, "learning_rate": 0.0004349208905350457, "loss": 1.201, "step": 65160 }, { "epoch": 2.62, "learning_rate": 0.0004349108257176215, "loss": 1.1564, "step": 65170 }, { "epoch": 2.62, "learning_rate": 0.00043490076090019723, "loss": 1.4111, "step": 65180 }, { "epoch": 2.62, "learning_rate": 0.000434890696082773, "loss": 1.382, "step": 65190 }, { "epoch": 2.62, "learning_rate": 0.00043488063126534887, "loss": 1.4133, "step": 65200 }, { "epoch": 2.62, "learning_rate": 0.00043487056644792466, "loss": 1.1231, "step": 65210 }, { "epoch": 2.62, "learning_rate": 0.00043486050163050045, "loss": 1.179, "step": 65220 }, { "epoch": 2.62, "learning_rate": 0.00043485043681307624, "loss": 1.3395, "step": 65230 }, { "epoch": 2.62, "learning_rate": 0.00043484037199565203, "loss": 1.2524, "step": 65240 }, { "epoch": 2.62, "learning_rate": 0.0004348303071782278, "loss": 1.1838, "step": 65250 }, { "epoch": 2.62, "learning_rate": 0.00043482024236080356, "loss": 1.1457, "step": 65260 }, { "epoch": 2.63, "learning_rate": 0.00043481017754337935, "loss": 1.2911, "step": 65270 }, { "epoch": 2.63, "learning_rate": 0.00043480011272595514, "loss": 1.2742, "step": 65280 }, { "epoch": 2.63, "learning_rate": 0.00043479004790853093, "loss": 1.2417, "step": 65290 }, { "epoch": 2.63, "learning_rate": 0.0004347799830911067, "loss": 1.1347, "step": 65300 }, { "epoch": 2.63, "learning_rate": 0.0004347699182736825, "loss": 1.3896, "step": 65310 }, { "epoch": 2.63, "learning_rate": 0.0004347598534562583, "loss": 1.2202, "step": 65320 }, { "epoch": 2.63, "learning_rate": 0.0004347497886388341, "loss": 1.2333, "step": 65330 }, { "epoch": 2.63, "learning_rate": 0.00043473972382140994, "loss": 1.3963, "step": 65340 }, { "epoch": 2.63, "learning_rate": 0.0004347296590039857, "loss": 1.3848, "step": 65350 }, { "epoch": 2.63, "learning_rate": 0.00043471959418656147, "loss": 1.2384, "step": 65360 }, { "epoch": 2.63, "learning_rate": 0.00043470952936913726, "loss": 1.3029, "step": 65370 }, { "epoch": 2.63, "learning_rate": 0.00043469946455171305, "loss": 0.9843, "step": 65380 }, { "epoch": 2.63, "learning_rate": 0.00043468939973428884, "loss": 1.3688, "step": 65390 }, { "epoch": 2.63, "learning_rate": 0.00043467933491686463, "loss": 1.2389, "step": 65400 }, { "epoch": 2.63, "learning_rate": 0.0004346692700994404, "loss": 1.4614, "step": 65410 }, { "epoch": 2.63, "learning_rate": 0.0004346592052820162, "loss": 1.5113, "step": 65420 }, { "epoch": 2.63, "learning_rate": 0.00043464914046459195, "loss": 1.132, "step": 65430 }, { "epoch": 2.63, "learning_rate": 0.00043463907564716774, "loss": 1.3244, "step": 65440 }, { "epoch": 2.63, "learning_rate": 0.00043462901082974353, "loss": 1.3841, "step": 65450 }, { "epoch": 2.63, "learning_rate": 0.0004346189460123193, "loss": 1.221, "step": 65460 }, { "epoch": 2.63, "learning_rate": 0.0004346088811948951, "loss": 1.0361, "step": 65470 }, { "epoch": 2.63, "learning_rate": 0.0004345988163774709, "loss": 1.2979, "step": 65480 }, { "epoch": 2.63, "learning_rate": 0.00043458875156004675, "loss": 1.5237, "step": 65490 }, { "epoch": 2.63, "learning_rate": 0.00043457868674262254, "loss": 1.3126, "step": 65500 }, { "epoch": 2.63, "learning_rate": 0.00043456862192519833, "loss": 1.1844, "step": 65510 }, { "epoch": 2.64, "learning_rate": 0.00043455855710777407, "loss": 0.9853, "step": 65520 }, { "epoch": 2.64, "learning_rate": 0.00043454849229034986, "loss": 1.2591, "step": 65530 }, { "epoch": 2.64, "learning_rate": 0.00043453842747292565, "loss": 1.1044, "step": 65540 }, { "epoch": 2.64, "learning_rate": 0.00043452836265550144, "loss": 1.2489, "step": 65550 }, { "epoch": 2.64, "learning_rate": 0.00043451829783807723, "loss": 1.3032, "step": 65560 }, { "epoch": 2.64, "learning_rate": 0.000434508233020653, "loss": 1.3684, "step": 65570 }, { "epoch": 2.64, "learning_rate": 0.0004344981682032288, "loss": 1.3699, "step": 65580 }, { "epoch": 2.64, "learning_rate": 0.0004344881033858046, "loss": 1.2761, "step": 65590 }, { "epoch": 2.64, "learning_rate": 0.00043447803856838034, "loss": 1.238, "step": 65600 }, { "epoch": 2.64, "learning_rate": 0.00043446797375095613, "loss": 1.0145, "step": 65610 }, { "epoch": 2.64, "learning_rate": 0.0004344579089335319, "loss": 1.2243, "step": 65620 }, { "epoch": 2.64, "learning_rate": 0.0004344478441161077, "loss": 1.2619, "step": 65630 }, { "epoch": 2.64, "learning_rate": 0.00043443777929868356, "loss": 1.2062, "step": 65640 }, { "epoch": 2.64, "learning_rate": 0.00043442771448125935, "loss": 1.418, "step": 65650 }, { "epoch": 2.64, "learning_rate": 0.00043441764966383514, "loss": 1.1005, "step": 65660 }, { "epoch": 2.64, "learning_rate": 0.00043440758484641093, "loss": 1.4333, "step": 65670 }, { "epoch": 2.64, "learning_rate": 0.0004343975200289867, "loss": 1.0191, "step": 65680 }, { "epoch": 2.64, "learning_rate": 0.00043438745521156246, "loss": 1.2645, "step": 65690 }, { "epoch": 2.64, "learning_rate": 0.00043437739039413825, "loss": 1.1416, "step": 65700 }, { "epoch": 2.64, "learning_rate": 0.00043436732557671404, "loss": 1.177, "step": 65710 }, { "epoch": 2.64, "learning_rate": 0.00043435726075928983, "loss": 1.1562, "step": 65720 }, { "epoch": 2.64, "learning_rate": 0.0004343471959418656, "loss": 1.0733, "step": 65730 }, { "epoch": 2.64, "learning_rate": 0.0004343371311244414, "loss": 1.0324, "step": 65740 }, { "epoch": 2.64, "learning_rate": 0.0004343270663070172, "loss": 1.3071, "step": 65750 }, { "epoch": 2.64, "learning_rate": 0.000434317001489593, "loss": 1.363, "step": 65760 }, { "epoch": 2.65, "learning_rate": 0.00043430693667216873, "loss": 1.3555, "step": 65770 }, { "epoch": 2.65, "learning_rate": 0.00043429687185474457, "loss": 1.3939, "step": 65780 }, { "epoch": 2.65, "learning_rate": 0.00043428680703732036, "loss": 1.2117, "step": 65790 }, { "epoch": 2.65, "learning_rate": 0.00043427674221989615, "loss": 1.0276, "step": 65800 }, { "epoch": 2.65, "learning_rate": 0.00043426667740247194, "loss": 1.2874, "step": 65810 }, { "epoch": 2.65, "learning_rate": 0.00043425661258504774, "loss": 1.2434, "step": 65820 }, { "epoch": 2.65, "learning_rate": 0.0004342465477676235, "loss": 1.2652, "step": 65830 }, { "epoch": 2.65, "learning_rate": 0.0004342364829501993, "loss": 1.4596, "step": 65840 }, { "epoch": 2.65, "learning_rate": 0.0004342264181327751, "loss": 1.0548, "step": 65850 }, { "epoch": 2.65, "learning_rate": 0.00043421635331535084, "loss": 1.4163, "step": 65860 }, { "epoch": 2.65, "learning_rate": 0.00043420628849792663, "loss": 1.4767, "step": 65870 }, { "epoch": 2.65, "learning_rate": 0.0004341962236805024, "loss": 1.2584, "step": 65880 }, { "epoch": 2.65, "learning_rate": 0.0004341861588630782, "loss": 1.2202, "step": 65890 }, { "epoch": 2.65, "learning_rate": 0.000434176094045654, "loss": 1.3085, "step": 65900 }, { "epoch": 2.65, "learning_rate": 0.0004341660292282298, "loss": 1.1357, "step": 65910 }, { "epoch": 2.65, "learning_rate": 0.0004341559644108056, "loss": 1.0886, "step": 65920 }, { "epoch": 2.65, "learning_rate": 0.00043414589959338143, "loss": 1.2277, "step": 65930 }, { "epoch": 2.65, "learning_rate": 0.00043413583477595717, "loss": 1.4443, "step": 65940 }, { "epoch": 2.65, "learning_rate": 0.00043412576995853296, "loss": 1.1509, "step": 65950 }, { "epoch": 2.65, "learning_rate": 0.00043411570514110875, "loss": 1.1873, "step": 65960 }, { "epoch": 2.65, "learning_rate": 0.00043410564032368454, "loss": 1.2175, "step": 65970 }, { "epoch": 2.65, "learning_rate": 0.00043409557550626033, "loss": 1.4914, "step": 65980 }, { "epoch": 2.65, "learning_rate": 0.0004340855106888361, "loss": 1.2391, "step": 65990 }, { "epoch": 2.65, "learning_rate": 0.0004340754458714119, "loss": 1.2994, "step": 66000 }, { "epoch": 2.65, "learning_rate": 0.0004340653810539877, "loss": 1.2994, "step": 66010 }, { "epoch": 2.66, "learning_rate": 0.00043405531623656344, "loss": 1.0442, "step": 66020 }, { "epoch": 2.66, "learning_rate": 0.00043404525141913923, "loss": 1.5283, "step": 66030 }, { "epoch": 2.66, "learning_rate": 0.000434035186601715, "loss": 1.3323, "step": 66040 }, { "epoch": 2.66, "learning_rate": 0.0004340251217842908, "loss": 1.3206, "step": 66050 }, { "epoch": 2.66, "learning_rate": 0.0004340150569668666, "loss": 1.2418, "step": 66060 }, { "epoch": 2.66, "learning_rate": 0.00043400499214944245, "loss": 1.2757, "step": 66070 }, { "epoch": 2.66, "learning_rate": 0.00043399492733201824, "loss": 1.1634, "step": 66080 }, { "epoch": 2.66, "learning_rate": 0.00043398486251459403, "loss": 1.335, "step": 66090 }, { "epoch": 2.66, "learning_rate": 0.0004339747976971698, "loss": 1.3881, "step": 66100 }, { "epoch": 2.66, "learning_rate": 0.00043396473287974556, "loss": 1.1831, "step": 66110 }, { "epoch": 2.66, "learning_rate": 0.00043395466806232135, "loss": 1.3638, "step": 66120 }, { "epoch": 2.66, "learning_rate": 0.00043394460324489714, "loss": 1.3295, "step": 66130 }, { "epoch": 2.66, "learning_rate": 0.00043393453842747293, "loss": 1.126, "step": 66140 }, { "epoch": 2.66, "learning_rate": 0.0004339244736100487, "loss": 1.1811, "step": 66150 }, { "epoch": 2.66, "learning_rate": 0.0004339144087926245, "loss": 1.3659, "step": 66160 }, { "epoch": 2.66, "learning_rate": 0.0004339043439752003, "loss": 1.1669, "step": 66170 }, { "epoch": 2.66, "learning_rate": 0.0004338942791577761, "loss": 1.2575, "step": 66180 }, { "epoch": 2.66, "learning_rate": 0.00043388421434035183, "loss": 0.8893, "step": 66190 }, { "epoch": 2.66, "learning_rate": 0.0004338741495229276, "loss": 1.2741, "step": 66200 }, { "epoch": 2.66, "learning_rate": 0.0004338640847055034, "loss": 1.3231, "step": 66210 }, { "epoch": 2.66, "learning_rate": 0.00043385401988807926, "loss": 1.3009, "step": 66220 }, { "epoch": 2.66, "learning_rate": 0.00043384395507065505, "loss": 1.1971, "step": 66230 }, { "epoch": 2.66, "learning_rate": 0.00043383389025323084, "loss": 1.366, "step": 66240 }, { "epoch": 2.66, "learning_rate": 0.00043382382543580663, "loss": 1.5437, "step": 66250 }, { "epoch": 2.66, "learning_rate": 0.0004338137606183824, "loss": 1.4154, "step": 66260 }, { "epoch": 2.67, "learning_rate": 0.0004338036958009582, "loss": 1.2985, "step": 66270 }, { "epoch": 2.67, "learning_rate": 0.00043379363098353395, "loss": 1.127, "step": 66280 }, { "epoch": 2.67, "learning_rate": 0.00043378356616610974, "loss": 1.3801, "step": 66290 }, { "epoch": 2.67, "learning_rate": 0.00043377350134868553, "loss": 1.1533, "step": 66300 }, { "epoch": 2.67, "learning_rate": 0.0004337634365312613, "loss": 1.3776, "step": 66310 }, { "epoch": 2.67, "learning_rate": 0.0004337533717138371, "loss": 1.2563, "step": 66320 }, { "epoch": 2.67, "learning_rate": 0.0004337433068964129, "loss": 1.1293, "step": 66330 }, { "epoch": 2.67, "learning_rate": 0.0004337332420789887, "loss": 1.4084, "step": 66340 }, { "epoch": 2.67, "learning_rate": 0.0004337231772615645, "loss": 1.2064, "step": 66350 }, { "epoch": 2.67, "learning_rate": 0.0004337131124441402, "loss": 1.1217, "step": 66360 }, { "epoch": 2.67, "learning_rate": 0.00043370304762671607, "loss": 1.3037, "step": 66370 }, { "epoch": 2.67, "learning_rate": 0.00043369298280929186, "loss": 0.996, "step": 66380 }, { "epoch": 2.67, "learning_rate": 0.00043368291799186765, "loss": 1.0075, "step": 66390 }, { "epoch": 2.67, "learning_rate": 0.00043367285317444344, "loss": 1.5379, "step": 66400 }, { "epoch": 2.67, "learning_rate": 0.00043366278835701923, "loss": 1.5023, "step": 66410 }, { "epoch": 2.67, "learning_rate": 0.000433652723539595, "loss": 1.1812, "step": 66420 }, { "epoch": 2.67, "learning_rate": 0.0004336426587221708, "loss": 1.2094, "step": 66430 }, { "epoch": 2.67, "learning_rate": 0.0004336325939047466, "loss": 1.2851, "step": 66440 }, { "epoch": 2.67, "learning_rate": 0.00043362252908732234, "loss": 1.4245, "step": 66450 }, { "epoch": 2.67, "learning_rate": 0.00043361246426989813, "loss": 1.2602, "step": 66460 }, { "epoch": 2.67, "learning_rate": 0.0004336023994524739, "loss": 1.5121, "step": 66470 }, { "epoch": 2.67, "learning_rate": 0.0004335923346350497, "loss": 1.1117, "step": 66480 }, { "epoch": 2.67, "learning_rate": 0.0004335822698176255, "loss": 1.3782, "step": 66490 }, { "epoch": 2.67, "learning_rate": 0.0004335722050002013, "loss": 1.1146, "step": 66500 }, { "epoch": 2.67, "learning_rate": 0.00043356214018277714, "loss": 1.2842, "step": 66510 }, { "epoch": 2.68, "learning_rate": 0.00043355207536535293, "loss": 1.2108, "step": 66520 }, { "epoch": 2.68, "learning_rate": 0.00043354201054792866, "loss": 1.2838, "step": 66530 }, { "epoch": 2.68, "learning_rate": 0.00043353194573050445, "loss": 1.1989, "step": 66540 }, { "epoch": 2.68, "learning_rate": 0.00043352188091308025, "loss": 1.1808, "step": 66550 }, { "epoch": 2.68, "learning_rate": 0.00043351181609565604, "loss": 1.1828, "step": 66560 }, { "epoch": 2.68, "learning_rate": 0.00043350175127823183, "loss": 1.3334, "step": 66570 }, { "epoch": 2.68, "learning_rate": 0.0004334916864608076, "loss": 1.0904, "step": 66580 }, { "epoch": 2.68, "learning_rate": 0.0004334816216433834, "loss": 1.1503, "step": 66590 }, { "epoch": 2.68, "learning_rate": 0.0004334715568259592, "loss": 1.1718, "step": 66600 }, { "epoch": 2.68, "learning_rate": 0.000433461492008535, "loss": 1.082, "step": 66610 }, { "epoch": 2.68, "learning_rate": 0.0004334514271911107, "loss": 1.4057, "step": 66620 }, { "epoch": 2.68, "learning_rate": 0.0004334413623736865, "loss": 1.3751, "step": 66630 }, { "epoch": 2.68, "learning_rate": 0.0004334312975562623, "loss": 1.2903, "step": 66640 }, { "epoch": 2.68, "learning_rate": 0.0004334212327388381, "loss": 1.3934, "step": 66650 }, { "epoch": 2.68, "learning_rate": 0.00043341116792141394, "loss": 1.2374, "step": 66660 }, { "epoch": 2.68, "learning_rate": 0.00043340110310398974, "loss": 1.2292, "step": 66670 }, { "epoch": 2.68, "learning_rate": 0.0004333910382865655, "loss": 1.2748, "step": 66680 }, { "epoch": 2.68, "learning_rate": 0.0004333809734691413, "loss": 1.1589, "step": 66690 }, { "epoch": 2.68, "learning_rate": 0.00043337090865171705, "loss": 1.1782, "step": 66700 }, { "epoch": 2.68, "learning_rate": 0.00043336084383429284, "loss": 1.1879, "step": 66710 }, { "epoch": 2.68, "learning_rate": 0.00043335077901686863, "loss": 1.4267, "step": 66720 }, { "epoch": 2.68, "learning_rate": 0.0004333407141994444, "loss": 1.2381, "step": 66730 }, { "epoch": 2.68, "learning_rate": 0.0004333306493820202, "loss": 1.2671, "step": 66740 }, { "epoch": 2.68, "learning_rate": 0.000433320584564596, "loss": 1.2097, "step": 66750 }, { "epoch": 2.68, "learning_rate": 0.0004333105197471718, "loss": 1.3291, "step": 66760 }, { "epoch": 2.69, "learning_rate": 0.0004333004549297476, "loss": 1.378, "step": 66770 }, { "epoch": 2.69, "learning_rate": 0.0004332903901123233, "loss": 1.2065, "step": 66780 }, { "epoch": 2.69, "learning_rate": 0.0004332803252948991, "loss": 1.2838, "step": 66790 }, { "epoch": 2.69, "learning_rate": 0.0004332702604774749, "loss": 1.3223, "step": 66800 }, { "epoch": 2.69, "learning_rate": 0.00043326019566005075, "loss": 1.5067, "step": 66810 }, { "epoch": 2.69, "learning_rate": 0.00043325013084262654, "loss": 1.1996, "step": 66820 }, { "epoch": 2.69, "learning_rate": 0.00043324006602520233, "loss": 1.1655, "step": 66830 }, { "epoch": 2.69, "learning_rate": 0.0004332300012077781, "loss": 1.5289, "step": 66840 }, { "epoch": 2.69, "learning_rate": 0.0004332199363903539, "loss": 1.4466, "step": 66850 }, { "epoch": 2.69, "learning_rate": 0.0004332098715729297, "loss": 1.4151, "step": 66860 }, { "epoch": 2.69, "learning_rate": 0.00043319980675550544, "loss": 1.2738, "step": 66870 }, { "epoch": 2.69, "learning_rate": 0.00043318974193808123, "loss": 1.3459, "step": 66880 }, { "epoch": 2.69, "learning_rate": 0.000433179677120657, "loss": 1.4736, "step": 66890 }, { "epoch": 2.69, "learning_rate": 0.0004331696123032328, "loss": 1.3205, "step": 66900 }, { "epoch": 2.69, "learning_rate": 0.0004331595474858086, "loss": 1.1276, "step": 66910 }, { "epoch": 2.69, "learning_rate": 0.0004331494826683844, "loss": 1.4145, "step": 66920 }, { "epoch": 2.69, "learning_rate": 0.0004331394178509602, "loss": 1.3747, "step": 66930 }, { "epoch": 2.69, "learning_rate": 0.000433129353033536, "loss": 1.2789, "step": 66940 }, { "epoch": 2.69, "learning_rate": 0.00043311928821611177, "loss": 1.2615, "step": 66950 }, { "epoch": 2.69, "learning_rate": 0.00043310922339868756, "loss": 1.2397, "step": 66960 }, { "epoch": 2.69, "learning_rate": 0.00043309915858126335, "loss": 1.0915, "step": 66970 }, { "epoch": 2.69, "learning_rate": 0.00043308909376383914, "loss": 1.3705, "step": 66980 }, { "epoch": 2.69, "learning_rate": 0.00043307902894641493, "loss": 1.4753, "step": 66990 }, { "epoch": 2.69, "learning_rate": 0.0004330689641289907, "loss": 1.1972, "step": 67000 }, { "epoch": 2.7, "learning_rate": 0.0004330588993115665, "loss": 1.3735, "step": 67010 }, { "epoch": 2.7, "learning_rate": 0.0004330488344941423, "loss": 1.0804, "step": 67020 }, { "epoch": 2.7, "learning_rate": 0.0004330387696767181, "loss": 1.3732, "step": 67030 }, { "epoch": 2.7, "learning_rate": 0.00043302870485929383, "loss": 1.2396, "step": 67040 }, { "epoch": 2.7, "learning_rate": 0.0004330186400418696, "loss": 1.2394, "step": 67050 }, { "epoch": 2.7, "learning_rate": 0.0004330085752244454, "loss": 1.466, "step": 67060 }, { "epoch": 2.7, "learning_rate": 0.0004329985104070212, "loss": 1.3333, "step": 67070 }, { "epoch": 2.7, "learning_rate": 0.000432988445589597, "loss": 1.3558, "step": 67080 }, { "epoch": 2.7, "learning_rate": 0.0004329783807721728, "loss": 1.5086, "step": 67090 }, { "epoch": 2.7, "learning_rate": 0.00043296831595474863, "loss": 1.4331, "step": 67100 }, { "epoch": 2.7, "learning_rate": 0.0004329582511373244, "loss": 1.2875, "step": 67110 }, { "epoch": 2.7, "learning_rate": 0.00043294818631990016, "loss": 1.3019, "step": 67120 }, { "epoch": 2.7, "learning_rate": 0.00043293812150247595, "loss": 1.4925, "step": 67130 }, { "epoch": 2.7, "learning_rate": 0.00043292805668505174, "loss": 1.3639, "step": 67140 }, { "epoch": 2.7, "learning_rate": 0.00043291799186762753, "loss": 1.1087, "step": 67150 }, { "epoch": 2.7, "learning_rate": 0.0004329079270502033, "loss": 1.2408, "step": 67160 }, { "epoch": 2.7, "learning_rate": 0.0004328978622327791, "loss": 1.0491, "step": 67170 }, { "epoch": 2.7, "learning_rate": 0.0004328877974153549, "loss": 1.3041, "step": 67180 }, { "epoch": 2.7, "learning_rate": 0.0004328777325979307, "loss": 1.4477, "step": 67190 }, { "epoch": 2.7, "learning_rate": 0.0004328676677805065, "loss": 1.1962, "step": 67200 }, { "epoch": 2.7, "learning_rate": 0.0004328576029630822, "loss": 1.4226, "step": 67210 }, { "epoch": 2.7, "learning_rate": 0.000432847538145658, "loss": 1.5101, "step": 67220 }, { "epoch": 2.7, "learning_rate": 0.0004328374733282338, "loss": 1.4197, "step": 67230 }, { "epoch": 2.7, "learning_rate": 0.00043282740851080965, "loss": 1.36, "step": 67240 }, { "epoch": 2.7, "learning_rate": 0.00043281734369338544, "loss": 1.2673, "step": 67250 }, { "epoch": 2.71, "learning_rate": 0.00043280727887596123, "loss": 1.0057, "step": 67260 }, { "epoch": 2.71, "learning_rate": 0.000432797214058537, "loss": 1.2765, "step": 67270 }, { "epoch": 2.71, "learning_rate": 0.0004327871492411128, "loss": 1.3288, "step": 67280 }, { "epoch": 2.71, "learning_rate": 0.00043277708442368855, "loss": 1.2873, "step": 67290 }, { "epoch": 2.71, "learning_rate": 0.00043276701960626434, "loss": 1.2178, "step": 67300 }, { "epoch": 2.71, "learning_rate": 0.00043275695478884013, "loss": 1.2639, "step": 67310 }, { "epoch": 2.71, "learning_rate": 0.0004327468899714159, "loss": 1.377, "step": 67320 }, { "epoch": 2.71, "learning_rate": 0.0004327368251539917, "loss": 1.2356, "step": 67330 }, { "epoch": 2.71, "learning_rate": 0.0004327267603365675, "loss": 1.1942, "step": 67340 }, { "epoch": 2.71, "learning_rate": 0.0004327166955191433, "loss": 1.2119, "step": 67350 }, { "epoch": 2.71, "learning_rate": 0.0004327066307017191, "loss": 1.1867, "step": 67360 }, { "epoch": 2.71, "learning_rate": 0.0004326965658842949, "loss": 1.0928, "step": 67370 }, { "epoch": 2.71, "learning_rate": 0.0004326865010668706, "loss": 1.124, "step": 67380 }, { "epoch": 2.71, "learning_rate": 0.00043267643624944645, "loss": 1.1706, "step": 67390 }, { "epoch": 2.71, "learning_rate": 0.00043266637143202225, "loss": 1.4759, "step": 67400 }, { "epoch": 2.71, "learning_rate": 0.00043265630661459804, "loss": 1.41, "step": 67410 }, { "epoch": 2.71, "learning_rate": 0.0004326462417971738, "loss": 1.3115, "step": 67420 }, { "epoch": 2.71, "learning_rate": 0.0004326361769797496, "loss": 1.1039, "step": 67430 }, { "epoch": 2.71, "learning_rate": 0.0004326261121623254, "loss": 1.4445, "step": 67440 }, { "epoch": 2.71, "learning_rate": 0.0004326160473449012, "loss": 1.1529, "step": 67450 }, { "epoch": 2.71, "learning_rate": 0.00043260598252747694, "loss": 1.1902, "step": 67460 }, { "epoch": 2.71, "learning_rate": 0.0004325959177100527, "loss": 1.1748, "step": 67470 }, { "epoch": 2.71, "learning_rate": 0.0004325858528926285, "loss": 1.5833, "step": 67480 }, { "epoch": 2.71, "learning_rate": 0.0004325757880752043, "loss": 1.354, "step": 67490 }, { "epoch": 2.71, "learning_rate": 0.0004325657232577801, "loss": 1.3162, "step": 67500 }, { "epoch": 2.72, "learning_rate": 0.0004325556584403559, "loss": 1.2246, "step": 67510 }, { "epoch": 2.72, "learning_rate": 0.0004325455936229317, "loss": 1.3514, "step": 67520 }, { "epoch": 2.72, "learning_rate": 0.00043253552880550747, "loss": 1.1176, "step": 67530 }, { "epoch": 2.72, "learning_rate": 0.0004325254639880833, "loss": 1.3464, "step": 67540 }, { "epoch": 2.72, "learning_rate": 0.00043251539917065905, "loss": 1.1375, "step": 67550 }, { "epoch": 2.72, "learning_rate": 0.00043250533435323484, "loss": 1.3126, "step": 67560 }, { "epoch": 2.72, "learning_rate": 0.00043249526953581063, "loss": 1.5288, "step": 67570 }, { "epoch": 2.72, "learning_rate": 0.0004324852047183864, "loss": 1.2613, "step": 67580 }, { "epoch": 2.72, "learning_rate": 0.0004324751399009622, "loss": 1.2326, "step": 67590 }, { "epoch": 2.72, "learning_rate": 0.000432465075083538, "loss": 1.0201, "step": 67600 }, { "epoch": 2.72, "learning_rate": 0.0004324550102661138, "loss": 1.3019, "step": 67610 }, { "epoch": 2.72, "learning_rate": 0.0004324449454486896, "loss": 1.3768, "step": 67620 }, { "epoch": 2.72, "learning_rate": 0.0004324348806312653, "loss": 1.4769, "step": 67630 }, { "epoch": 2.72, "learning_rate": 0.0004324248158138411, "loss": 1.3061, "step": 67640 }, { "epoch": 2.72, "learning_rate": 0.0004324147509964169, "loss": 1.2053, "step": 67650 }, { "epoch": 2.72, "learning_rate": 0.0004324046861789927, "loss": 1.0372, "step": 67660 }, { "epoch": 2.72, "learning_rate": 0.0004323946213615685, "loss": 1.3108, "step": 67670 }, { "epoch": 2.72, "learning_rate": 0.00043238455654414433, "loss": 1.5334, "step": 67680 }, { "epoch": 2.72, "learning_rate": 0.0004323744917267201, "loss": 1.3934, "step": 67690 }, { "epoch": 2.72, "learning_rate": 0.0004323644269092959, "loss": 1.3724, "step": 67700 }, { "epoch": 2.72, "learning_rate": 0.00043235436209187165, "loss": 1.7711, "step": 67710 }, { "epoch": 2.72, "learning_rate": 0.00043234429727444744, "loss": 1.3205, "step": 67720 }, { "epoch": 2.72, "learning_rate": 0.00043233423245702323, "loss": 1.2934, "step": 67730 }, { "epoch": 2.72, "learning_rate": 0.000432324167639599, "loss": 1.7171, "step": 67740 }, { "epoch": 2.72, "learning_rate": 0.0004323141028221748, "loss": 1.3508, "step": 67750 }, { "epoch": 2.73, "learning_rate": 0.0004323040380047506, "loss": 1.1566, "step": 67760 }, { "epoch": 2.73, "learning_rate": 0.0004322939731873264, "loss": 1.4329, "step": 67770 }, { "epoch": 2.73, "learning_rate": 0.0004322839083699022, "loss": 1.2565, "step": 67780 }, { "epoch": 2.73, "learning_rate": 0.000432273843552478, "loss": 1.3758, "step": 67790 }, { "epoch": 2.73, "learning_rate": 0.0004322637787350537, "loss": 1.2907, "step": 67800 }, { "epoch": 2.73, "learning_rate": 0.0004322537139176295, "loss": 1.1111, "step": 67810 }, { "epoch": 2.73, "learning_rate": 0.0004322436491002053, "loss": 1.231, "step": 67820 }, { "epoch": 2.73, "learning_rate": 0.00043223358428278114, "loss": 1.1405, "step": 67830 }, { "epoch": 2.73, "learning_rate": 0.00043222351946535693, "loss": 1.0685, "step": 67840 }, { "epoch": 2.73, "learning_rate": 0.0004322134546479327, "loss": 1.1864, "step": 67850 }, { "epoch": 2.73, "learning_rate": 0.0004322033898305085, "loss": 1.1157, "step": 67860 }, { "epoch": 2.73, "learning_rate": 0.0004321933250130843, "loss": 1.3863, "step": 67870 }, { "epoch": 2.73, "learning_rate": 0.00043218326019566004, "loss": 1.5183, "step": 67880 }, { "epoch": 2.73, "learning_rate": 0.00043217319537823583, "loss": 1.4119, "step": 67890 }, { "epoch": 2.73, "learning_rate": 0.0004321631305608116, "loss": 1.2187, "step": 67900 }, { "epoch": 2.73, "learning_rate": 0.0004321530657433874, "loss": 1.4004, "step": 67910 }, { "epoch": 2.73, "learning_rate": 0.0004321430009259632, "loss": 1.2849, "step": 67920 }, { "epoch": 2.73, "learning_rate": 0.000432132936108539, "loss": 1.1736, "step": 67930 }, { "epoch": 2.73, "learning_rate": 0.0004321228712911148, "loss": 1.378, "step": 67940 }, { "epoch": 2.73, "learning_rate": 0.0004321128064736906, "loss": 1.165, "step": 67950 }, { "epoch": 2.73, "learning_rate": 0.00043210274165626637, "loss": 1.4266, "step": 67960 }, { "epoch": 2.73, "learning_rate": 0.0004320926768388421, "loss": 1.1617, "step": 67970 }, { "epoch": 2.73, "learning_rate": 0.00043208261202141795, "loss": 1.3177, "step": 67980 }, { "epoch": 2.73, "learning_rate": 0.00043207254720399374, "loss": 1.2354, "step": 67990 }, { "epoch": 2.73, "learning_rate": 0.00043206248238656953, "loss": 1.2462, "step": 68000 }, { "epoch": 2.74, "learning_rate": 0.0004320524175691453, "loss": 1.1758, "step": 68010 }, { "epoch": 2.74, "learning_rate": 0.0004320423527517211, "loss": 1.2201, "step": 68020 }, { "epoch": 2.74, "learning_rate": 0.0004320322879342969, "loss": 1.2973, "step": 68030 }, { "epoch": 2.74, "learning_rate": 0.0004320222231168727, "loss": 1.3582, "step": 68040 }, { "epoch": 2.74, "learning_rate": 0.00043201215829944843, "loss": 1.1208, "step": 68050 }, { "epoch": 2.74, "learning_rate": 0.0004320020934820242, "loss": 1.1441, "step": 68060 }, { "epoch": 2.74, "learning_rate": 0.0004319920286646, "loss": 1.2442, "step": 68070 }, { "epoch": 2.74, "learning_rate": 0.0004319819638471758, "loss": 1.2315, "step": 68080 }, { "epoch": 2.74, "learning_rate": 0.0004319718990297516, "loss": 1.195, "step": 68090 }, { "epoch": 2.74, "learning_rate": 0.0004319618342123274, "loss": 1.1279, "step": 68100 }, { "epoch": 2.74, "learning_rate": 0.0004319517693949032, "loss": 1.4213, "step": 68110 }, { "epoch": 2.74, "learning_rate": 0.000431941704577479, "loss": 1.281, "step": 68120 }, { "epoch": 2.74, "learning_rate": 0.0004319316397600548, "loss": 1.3941, "step": 68130 }, { "epoch": 2.74, "learning_rate": 0.00043192157494263055, "loss": 1.2501, "step": 68140 }, { "epoch": 2.74, "learning_rate": 0.00043191151012520634, "loss": 1.1994, "step": 68150 }, { "epoch": 2.74, "learning_rate": 0.00043190144530778213, "loss": 1.3204, "step": 68160 }, { "epoch": 2.74, "learning_rate": 0.0004318913804903579, "loss": 1.5526, "step": 68170 }, { "epoch": 2.74, "learning_rate": 0.0004318813156729337, "loss": 1.1829, "step": 68180 }, { "epoch": 2.74, "learning_rate": 0.0004318712508555095, "loss": 1.1717, "step": 68190 }, { "epoch": 2.74, "learning_rate": 0.0004318611860380853, "loss": 1.1673, "step": 68200 }, { "epoch": 2.74, "learning_rate": 0.0004318511212206611, "loss": 1.4756, "step": 68210 }, { "epoch": 2.74, "learning_rate": 0.0004318410564032368, "loss": 1.4095, "step": 68220 }, { "epoch": 2.74, "learning_rate": 0.0004318309915858126, "loss": 1.5967, "step": 68230 }, { "epoch": 2.74, "learning_rate": 0.0004318209267683884, "loss": 1.0555, "step": 68240 }, { "epoch": 2.74, "learning_rate": 0.0004318108619509642, "loss": 1.1971, "step": 68250 }, { "epoch": 2.75, "learning_rate": 0.00043180079713354, "loss": 1.2895, "step": 68260 }, { "epoch": 2.75, "learning_rate": 0.0004317907323161158, "loss": 1.2215, "step": 68270 }, { "epoch": 2.75, "learning_rate": 0.0004317806674986916, "loss": 1.3957, "step": 68280 }, { "epoch": 2.75, "learning_rate": 0.0004317706026812674, "loss": 1.3499, "step": 68290 }, { "epoch": 2.75, "learning_rate": 0.0004317605378638432, "loss": 1.4077, "step": 68300 }, { "epoch": 2.75, "learning_rate": 0.00043175047304641894, "loss": 1.3336, "step": 68310 }, { "epoch": 2.75, "learning_rate": 0.0004317404082289947, "loss": 0.945, "step": 68320 }, { "epoch": 2.75, "learning_rate": 0.0004317303434115705, "loss": 1.5455, "step": 68330 }, { "epoch": 2.75, "learning_rate": 0.0004317202785941463, "loss": 1.4107, "step": 68340 }, { "epoch": 2.75, "learning_rate": 0.0004317102137767221, "loss": 1.406, "step": 68350 }, { "epoch": 2.75, "learning_rate": 0.0004317001489592979, "loss": 1.2568, "step": 68360 }, { "epoch": 2.75, "learning_rate": 0.0004316900841418737, "loss": 1.2601, "step": 68370 }, { "epoch": 2.75, "learning_rate": 0.00043168001932444947, "loss": 1.0503, "step": 68380 }, { "epoch": 2.75, "learning_rate": 0.0004316699545070252, "loss": 1.4072, "step": 68390 }, { "epoch": 2.75, "learning_rate": 0.000431659889689601, "loss": 1.2663, "step": 68400 }, { "epoch": 2.75, "learning_rate": 0.00043164982487217684, "loss": 1.4013, "step": 68410 }, { "epoch": 2.75, "learning_rate": 0.00043163976005475263, "loss": 1.3663, "step": 68420 }, { "epoch": 2.75, "learning_rate": 0.0004316296952373284, "loss": 1.337, "step": 68430 }, { "epoch": 2.75, "learning_rate": 0.0004316196304199042, "loss": 1.3358, "step": 68440 }, { "epoch": 2.75, "learning_rate": 0.00043160956560248, "loss": 1.4251, "step": 68450 }, { "epoch": 2.75, "learning_rate": 0.0004315995007850558, "loss": 1.2944, "step": 68460 }, { "epoch": 2.75, "learning_rate": 0.00043158943596763153, "loss": 1.4145, "step": 68470 }, { "epoch": 2.75, "learning_rate": 0.0004315793711502073, "loss": 1.1659, "step": 68480 }, { "epoch": 2.75, "learning_rate": 0.0004315693063327831, "loss": 1.4042, "step": 68490 }, { "epoch": 2.75, "learning_rate": 0.0004315592415153589, "loss": 1.1661, "step": 68500 }, { "epoch": 2.76, "learning_rate": 0.0004315491766979347, "loss": 1.3599, "step": 68510 }, { "epoch": 2.76, "learning_rate": 0.0004315391118805105, "loss": 1.3128, "step": 68520 }, { "epoch": 2.76, "learning_rate": 0.0004315290470630863, "loss": 1.2475, "step": 68530 }, { "epoch": 2.76, "learning_rate": 0.00043151898224566207, "loss": 1.3329, "step": 68540 }, { "epoch": 2.76, "learning_rate": 0.00043150891742823786, "loss": 1.5017, "step": 68550 }, { "epoch": 2.76, "learning_rate": 0.00043149885261081365, "loss": 0.975, "step": 68560 }, { "epoch": 2.76, "learning_rate": 0.00043148878779338944, "loss": 1.1292, "step": 68570 }, { "epoch": 2.76, "learning_rate": 0.00043147872297596523, "loss": 1.2657, "step": 68580 }, { "epoch": 2.76, "learning_rate": 0.000431468658158541, "loss": 1.0409, "step": 68590 }, { "epoch": 2.76, "learning_rate": 0.0004314585933411168, "loss": 1.3453, "step": 68600 }, { "epoch": 2.76, "learning_rate": 0.0004314485285236926, "loss": 0.9857, "step": 68610 }, { "epoch": 2.76, "learning_rate": 0.0004314384637062684, "loss": 1.1782, "step": 68620 }, { "epoch": 2.76, "learning_rate": 0.0004314283988888442, "loss": 1.2594, "step": 68630 }, { "epoch": 2.76, "learning_rate": 0.0004314183340714199, "loss": 1.233, "step": 68640 }, { "epoch": 2.76, "learning_rate": 0.0004314082692539957, "loss": 1.2784, "step": 68650 }, { "epoch": 2.76, "learning_rate": 0.0004313982044365715, "loss": 1.2774, "step": 68660 }, { "epoch": 2.76, "learning_rate": 0.0004313881396191473, "loss": 1.2586, "step": 68670 }, { "epoch": 2.76, "learning_rate": 0.0004313780748017231, "loss": 1.0176, "step": 68680 }, { "epoch": 2.76, "learning_rate": 0.0004313680099842989, "loss": 1.3011, "step": 68690 }, { "epoch": 2.76, "learning_rate": 0.00043135794516687467, "loss": 1.3266, "step": 68700 }, { "epoch": 2.76, "learning_rate": 0.0004313478803494505, "loss": 1.3707, "step": 68710 }, { "epoch": 2.76, "learning_rate": 0.0004313378155320263, "loss": 1.3232, "step": 68720 }, { "epoch": 2.76, "learning_rate": 0.00043132775071460204, "loss": 1.1921, "step": 68730 }, { "epoch": 2.76, "learning_rate": 0.00043131768589717783, "loss": 1.3801, "step": 68740 }, { "epoch": 2.77, "learning_rate": 0.0004313076210797536, "loss": 1.2144, "step": 68750 }, { "epoch": 2.77, "learning_rate": 0.0004312975562623294, "loss": 1.1933, "step": 68760 }, { "epoch": 2.77, "learning_rate": 0.0004312874914449052, "loss": 1.0497, "step": 68770 }, { "epoch": 2.77, "learning_rate": 0.000431277426627481, "loss": 1.1924, "step": 68780 }, { "epoch": 2.77, "learning_rate": 0.0004312673618100568, "loss": 0.9755, "step": 68790 }, { "epoch": 2.77, "learning_rate": 0.0004312572969926326, "loss": 1.3477, "step": 68800 }, { "epoch": 2.77, "learning_rate": 0.0004312472321752083, "loss": 1.1901, "step": 68810 }, { "epoch": 2.77, "learning_rate": 0.0004312371673577841, "loss": 1.2321, "step": 68820 }, { "epoch": 2.77, "learning_rate": 0.0004312271025403599, "loss": 1.3417, "step": 68830 }, { "epoch": 2.77, "learning_rate": 0.0004312170377229357, "loss": 1.0787, "step": 68840 }, { "epoch": 2.77, "learning_rate": 0.00043120697290551153, "loss": 1.311, "step": 68850 }, { "epoch": 2.77, "learning_rate": 0.0004311969080880873, "loss": 1.1579, "step": 68860 }, { "epoch": 2.77, "learning_rate": 0.0004311868432706631, "loss": 1.3089, "step": 68870 }, { "epoch": 2.77, "learning_rate": 0.0004311767784532389, "loss": 1.3645, "step": 68880 }, { "epoch": 2.77, "learning_rate": 0.0004311667136358147, "loss": 1.3564, "step": 68890 }, { "epoch": 2.77, "learning_rate": 0.00043115664881839043, "loss": 1.3567, "step": 68900 }, { "epoch": 2.77, "learning_rate": 0.0004311465840009662, "loss": 1.4149, "step": 68910 }, { "epoch": 2.77, "learning_rate": 0.000431136519183542, "loss": 1.2338, "step": 68920 }, { "epoch": 2.77, "learning_rate": 0.0004311264543661178, "loss": 1.3189, "step": 68930 }, { "epoch": 2.77, "learning_rate": 0.0004311163895486936, "loss": 1.1499, "step": 68940 }, { "epoch": 2.77, "learning_rate": 0.0004311063247312694, "loss": 1.1233, "step": 68950 }, { "epoch": 2.77, "learning_rate": 0.0004310962599138452, "loss": 1.1054, "step": 68960 }, { "epoch": 2.77, "learning_rate": 0.00043108619509642096, "loss": 1.2086, "step": 68970 }, { "epoch": 2.77, "learning_rate": 0.0004310761302789967, "loss": 1.3375, "step": 68980 }, { "epoch": 2.77, "learning_rate": 0.0004310660654615725, "loss": 1.2514, "step": 68990 }, { "epoch": 2.78, "learning_rate": 0.00043105600064414834, "loss": 1.2563, "step": 69000 }, { "epoch": 2.78, "learning_rate": 0.00043104593582672413, "loss": 1.6569, "step": 69010 }, { "epoch": 2.78, "learning_rate": 0.0004310358710092999, "loss": 1.4875, "step": 69020 }, { "epoch": 2.78, "learning_rate": 0.0004310258061918757, "loss": 1.4713, "step": 69030 }, { "epoch": 2.78, "learning_rate": 0.0004310157413744515, "loss": 1.3225, "step": 69040 }, { "epoch": 2.78, "learning_rate": 0.0004310056765570273, "loss": 1.2754, "step": 69050 }, { "epoch": 2.78, "learning_rate": 0.0004309956117396031, "loss": 1.4675, "step": 69060 }, { "epoch": 2.78, "learning_rate": 0.0004309855469221788, "loss": 1.476, "step": 69070 }, { "epoch": 2.78, "learning_rate": 0.0004309754821047546, "loss": 1.369, "step": 69080 }, { "epoch": 2.78, "learning_rate": 0.0004309654172873304, "loss": 1.2931, "step": 69090 }, { "epoch": 2.78, "learning_rate": 0.0004309553524699062, "loss": 1.3745, "step": 69100 }, { "epoch": 2.78, "learning_rate": 0.000430945287652482, "loss": 1.2895, "step": 69110 }, { "epoch": 2.78, "learning_rate": 0.00043093522283505777, "loss": 1.3139, "step": 69120 }, { "epoch": 2.78, "learning_rate": 0.00043092515801763356, "loss": 1.4233, "step": 69130 }, { "epoch": 2.78, "learning_rate": 0.0004309150932002094, "loss": 1.1598, "step": 69140 }, { "epoch": 2.78, "learning_rate": 0.00043090502838278514, "loss": 1.3358, "step": 69150 }, { "epoch": 2.78, "learning_rate": 0.00043089496356536094, "loss": 1.3776, "step": 69160 }, { "epoch": 2.78, "learning_rate": 0.0004308848987479367, "loss": 1.3382, "step": 69170 }, { "epoch": 2.78, "learning_rate": 0.0004308748339305125, "loss": 1.2438, "step": 69180 }, { "epoch": 2.78, "learning_rate": 0.0004308647691130883, "loss": 1.0602, "step": 69190 }, { "epoch": 2.78, "learning_rate": 0.0004308547042956641, "loss": 1.2672, "step": 69200 }, { "epoch": 2.78, "learning_rate": 0.0004308446394782399, "loss": 1.2788, "step": 69210 }, { "epoch": 2.78, "learning_rate": 0.0004308345746608157, "loss": 1.116, "step": 69220 }, { "epoch": 2.78, "learning_rate": 0.0004308245098433914, "loss": 1.3361, "step": 69230 }, { "epoch": 2.78, "learning_rate": 0.0004308144450259672, "loss": 1.3148, "step": 69240 }, { "epoch": 2.79, "learning_rate": 0.000430804380208543, "loss": 1.0328, "step": 69250 }, { "epoch": 2.79, "learning_rate": 0.0004307943153911188, "loss": 1.2208, "step": 69260 }, { "epoch": 2.79, "learning_rate": 0.0004307842505736946, "loss": 1.1403, "step": 69270 }, { "epoch": 2.79, "learning_rate": 0.00043077418575627037, "loss": 1.0747, "step": 69280 }, { "epoch": 2.79, "learning_rate": 0.0004307641209388462, "loss": 1.3613, "step": 69290 }, { "epoch": 2.79, "learning_rate": 0.000430754056121422, "loss": 1.4512, "step": 69300 }, { "epoch": 2.79, "learning_rate": 0.0004307439913039978, "loss": 1.1504, "step": 69310 }, { "epoch": 2.79, "learning_rate": 0.00043073392648657353, "loss": 1.2312, "step": 69320 }, { "epoch": 2.79, "learning_rate": 0.0004307238616691493, "loss": 1.3248, "step": 69330 }, { "epoch": 2.79, "learning_rate": 0.0004307137968517251, "loss": 0.9842, "step": 69340 }, { "epoch": 2.79, "learning_rate": 0.0004307037320343009, "loss": 1.1026, "step": 69350 }, { "epoch": 2.79, "learning_rate": 0.0004306936672168767, "loss": 1.1865, "step": 69360 }, { "epoch": 2.79, "learning_rate": 0.0004306836023994525, "loss": 1.1101, "step": 69370 }, { "epoch": 2.79, "learning_rate": 0.0004306735375820283, "loss": 1.1136, "step": 69380 }, { "epoch": 2.79, "learning_rate": 0.00043066347276460407, "loss": 1.3905, "step": 69390 }, { "epoch": 2.79, "learning_rate": 0.0004306534079471798, "loss": 1.2983, "step": 69400 }, { "epoch": 2.79, "learning_rate": 0.0004306433431297556, "loss": 1.4034, "step": 69410 }, { "epoch": 2.79, "learning_rate": 0.0004306332783123314, "loss": 1.2571, "step": 69420 }, { "epoch": 2.79, "learning_rate": 0.0004306232134949072, "loss": 1.1085, "step": 69430 }, { "epoch": 2.79, "learning_rate": 0.000430613148677483, "loss": 1.1209, "step": 69440 }, { "epoch": 2.79, "learning_rate": 0.0004306030838600588, "loss": 1.172, "step": 69450 }, { "epoch": 2.79, "learning_rate": 0.0004305930190426346, "loss": 1.2441, "step": 69460 }, { "epoch": 2.79, "learning_rate": 0.0004305829542252104, "loss": 1.1158, "step": 69470 }, { "epoch": 2.79, "learning_rate": 0.0004305728894077862, "loss": 1.0549, "step": 69480 }, { "epoch": 2.79, "learning_rate": 0.0004305628245903619, "loss": 1.3791, "step": 69490 }, { "epoch": 2.8, "learning_rate": 0.0004305527597729377, "loss": 1.165, "step": 69500 }, { "epoch": 2.8, "learning_rate": 0.0004305426949555135, "loss": 1.2248, "step": 69510 }, { "epoch": 2.8, "learning_rate": 0.0004305326301380893, "loss": 1.2821, "step": 69520 }, { "epoch": 2.8, "learning_rate": 0.0004305225653206651, "loss": 1.3451, "step": 69530 }, { "epoch": 2.8, "learning_rate": 0.0004305125005032409, "loss": 1.275, "step": 69540 }, { "epoch": 2.8, "learning_rate": 0.00043050243568581667, "loss": 1.3207, "step": 69550 }, { "epoch": 2.8, "learning_rate": 0.00043049237086839246, "loss": 1.3792, "step": 69560 }, { "epoch": 2.8, "learning_rate": 0.0004304823060509682, "loss": 1.2155, "step": 69570 }, { "epoch": 2.8, "learning_rate": 0.00043047224123354404, "loss": 1.3621, "step": 69580 }, { "epoch": 2.8, "learning_rate": 0.00043046217641611983, "loss": 1.0454, "step": 69590 }, { "epoch": 2.8, "learning_rate": 0.0004304521115986956, "loss": 1.3884, "step": 69600 }, { "epoch": 2.8, "learning_rate": 0.0004304420467812714, "loss": 1.1682, "step": 69610 }, { "epoch": 2.8, "learning_rate": 0.0004304319819638472, "loss": 1.1893, "step": 69620 }, { "epoch": 2.8, "learning_rate": 0.000430421917146423, "loss": 1.2542, "step": 69630 }, { "epoch": 2.8, "learning_rate": 0.0004304118523289988, "loss": 1.0776, "step": 69640 }, { "epoch": 2.8, "learning_rate": 0.0004304017875115746, "loss": 1.2368, "step": 69650 }, { "epoch": 2.8, "learning_rate": 0.0004303917226941503, "loss": 1.4187, "step": 69660 }, { "epoch": 2.8, "learning_rate": 0.0004303816578767261, "loss": 1.1115, "step": 69670 }, { "epoch": 2.8, "learning_rate": 0.0004303715930593019, "loss": 1.1506, "step": 69680 }, { "epoch": 2.8, "learning_rate": 0.0004303615282418777, "loss": 1.2337, "step": 69690 }, { "epoch": 2.8, "learning_rate": 0.0004303514634244535, "loss": 1.6217, "step": 69700 }, { "epoch": 2.8, "learning_rate": 0.00043034139860702927, "loss": 1.0111, "step": 69710 }, { "epoch": 2.8, "learning_rate": 0.00043033133378960506, "loss": 1.1038, "step": 69720 }, { "epoch": 2.8, "learning_rate": 0.0004303212689721809, "loss": 1.2029, "step": 69730 }, { "epoch": 2.8, "learning_rate": 0.00043031120415475664, "loss": 1.1801, "step": 69740 }, { "epoch": 2.81, "learning_rate": 0.00043030113933733243, "loss": 1.4683, "step": 69750 }, { "epoch": 2.81, "learning_rate": 0.0004302910745199082, "loss": 1.0622, "step": 69760 }, { "epoch": 2.81, "learning_rate": 0.000430281009702484, "loss": 1.5418, "step": 69770 }, { "epoch": 2.81, "learning_rate": 0.0004302709448850598, "loss": 1.3988, "step": 69780 }, { "epoch": 2.81, "learning_rate": 0.0004302608800676356, "loss": 1.3075, "step": 69790 }, { "epoch": 2.81, "learning_rate": 0.0004302508152502114, "loss": 1.0924, "step": 69800 }, { "epoch": 2.81, "learning_rate": 0.0004302407504327872, "loss": 1.2231, "step": 69810 }, { "epoch": 2.81, "learning_rate": 0.00043023068561536296, "loss": 1.2945, "step": 69820 }, { "epoch": 2.81, "learning_rate": 0.0004302206207979387, "loss": 1.4141, "step": 69830 }, { "epoch": 2.81, "learning_rate": 0.0004302105559805145, "loss": 1.254, "step": 69840 }, { "epoch": 2.81, "learning_rate": 0.0004302004911630903, "loss": 1.3022, "step": 69850 }, { "epoch": 2.81, "learning_rate": 0.0004301904263456661, "loss": 1.4275, "step": 69860 }, { "epoch": 2.81, "learning_rate": 0.00043018036152824186, "loss": 1.425, "step": 69870 }, { "epoch": 2.81, "learning_rate": 0.0004301702967108177, "loss": 1.1339, "step": 69880 }, { "epoch": 2.81, "learning_rate": 0.0004301602318933935, "loss": 1.7016, "step": 69890 }, { "epoch": 2.81, "learning_rate": 0.0004301501670759693, "loss": 1.2656, "step": 69900 }, { "epoch": 2.81, "learning_rate": 0.00043014010225854503, "loss": 1.153, "step": 69910 }, { "epoch": 2.81, "learning_rate": 0.0004301300374411208, "loss": 1.6219, "step": 69920 }, { "epoch": 2.81, "learning_rate": 0.0004301199726236966, "loss": 1.4261, "step": 69930 }, { "epoch": 2.81, "learning_rate": 0.0004301099078062724, "loss": 1.3121, "step": 69940 }, { "epoch": 2.81, "learning_rate": 0.0004300998429888482, "loss": 1.1964, "step": 69950 }, { "epoch": 2.81, "learning_rate": 0.000430089778171424, "loss": 1.0639, "step": 69960 }, { "epoch": 2.81, "learning_rate": 0.00043007971335399977, "loss": 0.8652, "step": 69970 }, { "epoch": 2.81, "learning_rate": 0.00043006964853657556, "loss": 1.2627, "step": 69980 }, { "epoch": 2.81, "learning_rate": 0.0004300595837191513, "loss": 0.7851, "step": 69990 }, { "epoch": 2.82, "learning_rate": 0.0004300495189017271, "loss": 1.1526, "step": 70000 }, { "epoch": 2.82, "learning_rate": 0.0004300394540843029, "loss": 1.4554, "step": 70010 }, { "epoch": 2.82, "learning_rate": 0.0004300293892668787, "loss": 1.5625, "step": 70020 }, { "epoch": 2.82, "learning_rate": 0.0004300193244494545, "loss": 1.4202, "step": 70030 }, { "epoch": 2.82, "learning_rate": 0.0004300092596320303, "loss": 1.1881, "step": 70040 }, { "epoch": 2.82, "learning_rate": 0.0004299991948146061, "loss": 1.7378, "step": 70050 }, { "epoch": 2.82, "learning_rate": 0.0004299891299971819, "loss": 1.3002, "step": 70060 }, { "epoch": 2.82, "learning_rate": 0.0004299790651797577, "loss": 1.4013, "step": 70070 }, { "epoch": 2.82, "learning_rate": 0.0004299690003623334, "loss": 1.2471, "step": 70080 }, { "epoch": 2.82, "learning_rate": 0.0004299589355449092, "loss": 1.4001, "step": 70090 }, { "epoch": 2.82, "learning_rate": 0.000429948870727485, "loss": 1.0188, "step": 70100 }, { "epoch": 2.82, "learning_rate": 0.0004299388059100608, "loss": 1.4508, "step": 70110 }, { "epoch": 2.82, "learning_rate": 0.0004299287410926366, "loss": 1.3054, "step": 70120 }, { "epoch": 2.82, "learning_rate": 0.00042991867627521237, "loss": 1.2075, "step": 70130 }, { "epoch": 2.82, "learning_rate": 0.00042990861145778816, "loss": 1.0421, "step": 70140 }, { "epoch": 2.82, "learning_rate": 0.00042989854664036395, "loss": 1.3503, "step": 70150 }, { "epoch": 2.82, "learning_rate": 0.0004298884818229397, "loss": 1.2912, "step": 70160 }, { "epoch": 2.82, "learning_rate": 0.00042987841700551553, "loss": 1.3312, "step": 70170 }, { "epoch": 2.82, "learning_rate": 0.0004298683521880913, "loss": 1.1436, "step": 70180 }, { "epoch": 2.82, "learning_rate": 0.0004298582873706671, "loss": 1.1246, "step": 70190 }, { "epoch": 2.82, "learning_rate": 0.0004298482225532429, "loss": 1.1986, "step": 70200 }, { "epoch": 2.82, "learning_rate": 0.0004298381577358187, "loss": 1.1821, "step": 70210 }, { "epoch": 2.82, "learning_rate": 0.0004298280929183945, "loss": 1.343, "step": 70220 }, { "epoch": 2.82, "learning_rate": 0.0004298180281009703, "loss": 1.4309, "step": 70230 }, { "epoch": 2.82, "learning_rate": 0.00042980796328354607, "loss": 1.1961, "step": 70240 }, { "epoch": 2.83, "learning_rate": 0.0004297978984661218, "loss": 1.2351, "step": 70250 }, { "epoch": 2.83, "learning_rate": 0.0004297878336486976, "loss": 1.5832, "step": 70260 }, { "epoch": 2.83, "learning_rate": 0.0004297777688312734, "loss": 1.2302, "step": 70270 }, { "epoch": 2.83, "learning_rate": 0.0004297677040138492, "loss": 1.3962, "step": 70280 }, { "epoch": 2.83, "learning_rate": 0.00042975763919642497, "loss": 1.3475, "step": 70290 }, { "epoch": 2.83, "learning_rate": 0.00042974757437900076, "loss": 1.2024, "step": 70300 }, { "epoch": 2.83, "learning_rate": 0.0004297375095615766, "loss": 1.1973, "step": 70310 }, { "epoch": 2.83, "learning_rate": 0.0004297274447441524, "loss": 1.2808, "step": 70320 }, { "epoch": 2.83, "learning_rate": 0.00042971737992672813, "loss": 1.4403, "step": 70330 }, { "epoch": 2.83, "learning_rate": 0.0004297073151093039, "loss": 1.6514, "step": 70340 }, { "epoch": 2.83, "learning_rate": 0.0004296972502918797, "loss": 1.1314, "step": 70350 }, { "epoch": 2.83, "learning_rate": 0.0004296871854744555, "loss": 1.2477, "step": 70360 }, { "epoch": 2.83, "learning_rate": 0.0004296771206570313, "loss": 1.2911, "step": 70370 }, { "epoch": 2.83, "learning_rate": 0.0004296670558396071, "loss": 1.4418, "step": 70380 }, { "epoch": 2.83, "learning_rate": 0.0004296569910221829, "loss": 1.4997, "step": 70390 }, { "epoch": 2.83, "learning_rate": 0.00042964692620475867, "loss": 1.1468, "step": 70400 }, { "epoch": 2.83, "learning_rate": 0.00042963686138733446, "loss": 1.1711, "step": 70410 }, { "epoch": 2.83, "learning_rate": 0.0004296267965699102, "loss": 1.2599, "step": 70420 }, { "epoch": 2.83, "learning_rate": 0.000429616731752486, "loss": 1.1542, "step": 70430 }, { "epoch": 2.83, "learning_rate": 0.0004296066669350618, "loss": 1.1051, "step": 70440 }, { "epoch": 2.83, "learning_rate": 0.00042959660211763757, "loss": 1.3577, "step": 70450 }, { "epoch": 2.83, "learning_rate": 0.0004295865373002134, "loss": 1.493, "step": 70460 }, { "epoch": 2.83, "learning_rate": 0.0004295764724827892, "loss": 1.2094, "step": 70470 }, { "epoch": 2.83, "learning_rate": 0.000429566407665365, "loss": 1.2709, "step": 70480 }, { "epoch": 2.83, "learning_rate": 0.0004295563428479408, "loss": 1.4614, "step": 70490 }, { "epoch": 2.84, "learning_rate": 0.0004295462780305165, "loss": 1.4685, "step": 70500 }, { "epoch": 2.84, "learning_rate": 0.0004295362132130923, "loss": 1.3573, "step": 70510 }, { "epoch": 2.84, "learning_rate": 0.0004295261483956681, "loss": 1.4166, "step": 70520 }, { "epoch": 2.84, "learning_rate": 0.0004295160835782439, "loss": 1.1172, "step": 70530 }, { "epoch": 2.84, "learning_rate": 0.0004295060187608197, "loss": 1.1727, "step": 70540 }, { "epoch": 2.84, "learning_rate": 0.0004294959539433955, "loss": 1.3582, "step": 70550 }, { "epoch": 2.84, "learning_rate": 0.00042948588912597127, "loss": 1.3325, "step": 70560 }, { "epoch": 2.84, "learning_rate": 0.00042947582430854706, "loss": 1.108, "step": 70570 }, { "epoch": 2.84, "learning_rate": 0.00042946575949112285, "loss": 1.3193, "step": 70580 }, { "epoch": 2.84, "learning_rate": 0.0004294556946736986, "loss": 1.1668, "step": 70590 }, { "epoch": 2.84, "learning_rate": 0.0004294456298562744, "loss": 1.1122, "step": 70600 }, { "epoch": 2.84, "learning_rate": 0.0004294355650388502, "loss": 1.3208, "step": 70610 }, { "epoch": 2.84, "learning_rate": 0.000429425500221426, "loss": 1.3102, "step": 70620 }, { "epoch": 2.84, "learning_rate": 0.0004294154354040018, "loss": 1.5789, "step": 70630 }, { "epoch": 2.84, "learning_rate": 0.0004294053705865776, "loss": 1.1908, "step": 70640 }, { "epoch": 2.84, "learning_rate": 0.0004293953057691534, "loss": 1.3524, "step": 70650 }, { "epoch": 2.84, "learning_rate": 0.0004293852409517292, "loss": 1.2007, "step": 70660 }, { "epoch": 2.84, "learning_rate": 0.0004293751761343049, "loss": 1.2536, "step": 70670 }, { "epoch": 2.84, "learning_rate": 0.0004293651113168807, "loss": 1.3842, "step": 70680 }, { "epoch": 2.84, "learning_rate": 0.0004293550464994565, "loss": 1.3229, "step": 70690 }, { "epoch": 2.84, "learning_rate": 0.0004293449816820323, "loss": 1.1905, "step": 70700 }, { "epoch": 2.84, "learning_rate": 0.0004293349168646081, "loss": 1.4696, "step": 70710 }, { "epoch": 2.84, "learning_rate": 0.00042932485204718386, "loss": 1.5998, "step": 70720 }, { "epoch": 2.84, "learning_rate": 0.00042931478722975965, "loss": 1.3962, "step": 70730 }, { "epoch": 2.85, "learning_rate": 0.00042930472241233545, "loss": 1.2057, "step": 70740 }, { "epoch": 2.85, "learning_rate": 0.0004292946575949113, "loss": 1.3551, "step": 70750 }, { "epoch": 2.85, "learning_rate": 0.000429284592777487, "loss": 1.3308, "step": 70760 }, { "epoch": 2.85, "learning_rate": 0.0004292745279600628, "loss": 1.2883, "step": 70770 }, { "epoch": 2.85, "learning_rate": 0.0004292644631426386, "loss": 1.3447, "step": 70780 }, { "epoch": 2.85, "learning_rate": 0.0004292543983252144, "loss": 1.4499, "step": 70790 }, { "epoch": 2.85, "learning_rate": 0.0004292443335077902, "loss": 1.1409, "step": 70800 }, { "epoch": 2.85, "learning_rate": 0.000429234268690366, "loss": 1.4528, "step": 70810 }, { "epoch": 2.85, "learning_rate": 0.00042922420387294177, "loss": 1.2007, "step": 70820 }, { "epoch": 2.85, "learning_rate": 0.00042921413905551756, "loss": 1.3581, "step": 70830 }, { "epoch": 2.85, "learning_rate": 0.0004292040742380933, "loss": 1.4313, "step": 70840 }, { "epoch": 2.85, "learning_rate": 0.0004291940094206691, "loss": 1.3704, "step": 70850 }, { "epoch": 2.85, "learning_rate": 0.0004291839446032449, "loss": 1.1098, "step": 70860 }, { "epoch": 2.85, "learning_rate": 0.00042917387978582067, "loss": 1.4069, "step": 70870 }, { "epoch": 2.85, "learning_rate": 0.00042916381496839646, "loss": 1.4089, "step": 70880 }, { "epoch": 2.85, "learning_rate": 0.00042915375015097225, "loss": 1.3782, "step": 70890 }, { "epoch": 2.85, "learning_rate": 0.0004291436853335481, "loss": 1.205, "step": 70900 }, { "epoch": 2.85, "learning_rate": 0.0004291336205161239, "loss": 1.3827, "step": 70910 }, { "epoch": 2.85, "learning_rate": 0.0004291235556986997, "loss": 1.1169, "step": 70920 }, { "epoch": 2.85, "learning_rate": 0.0004291134908812754, "loss": 1.338, "step": 70930 }, { "epoch": 2.85, "learning_rate": 0.0004291034260638512, "loss": 1.2404, "step": 70940 }, { "epoch": 2.85, "learning_rate": 0.000429093361246427, "loss": 1.2775, "step": 70950 }, { "epoch": 2.85, "learning_rate": 0.0004290832964290028, "loss": 1.3594, "step": 70960 }, { "epoch": 2.85, "learning_rate": 0.0004290732316115786, "loss": 1.402, "step": 70970 }, { "epoch": 2.85, "learning_rate": 0.00042906316679415437, "loss": 1.2919, "step": 70980 }, { "epoch": 2.86, "learning_rate": 0.00042905310197673016, "loss": 1.1792, "step": 70990 }, { "epoch": 2.86, "learning_rate": 0.00042904303715930595, "loss": 1.2441, "step": 71000 }, { "epoch": 2.86, "learning_rate": 0.0004290329723418817, "loss": 1.2791, "step": 71010 }, { "epoch": 2.86, "learning_rate": 0.0004290229075244575, "loss": 1.1476, "step": 71020 }, { "epoch": 2.86, "learning_rate": 0.00042901284270703327, "loss": 1.1697, "step": 71030 }, { "epoch": 2.86, "learning_rate": 0.00042900277788960906, "loss": 1.2953, "step": 71040 }, { "epoch": 2.86, "learning_rate": 0.0004289927130721849, "loss": 1.6653, "step": 71050 }, { "epoch": 2.86, "learning_rate": 0.0004289826482547607, "loss": 1.2398, "step": 71060 }, { "epoch": 2.86, "learning_rate": 0.0004289725834373365, "loss": 1.3195, "step": 71070 }, { "epoch": 2.86, "learning_rate": 0.0004289625186199123, "loss": 1.1769, "step": 71080 }, { "epoch": 2.86, "learning_rate": 0.000428952453802488, "loss": 1.0791, "step": 71090 }, { "epoch": 2.86, "learning_rate": 0.0004289423889850638, "loss": 1.2937, "step": 71100 }, { "epoch": 2.86, "learning_rate": 0.0004289323241676396, "loss": 0.9859, "step": 71110 }, { "epoch": 2.86, "learning_rate": 0.0004289222593502154, "loss": 1.2058, "step": 71120 }, { "epoch": 2.86, "learning_rate": 0.0004289121945327912, "loss": 1.2978, "step": 71130 }, { "epoch": 2.86, "learning_rate": 0.00042890212971536697, "loss": 1.3921, "step": 71140 }, { "epoch": 2.86, "learning_rate": 0.00042889206489794276, "loss": 1.2656, "step": 71150 }, { "epoch": 2.86, "learning_rate": 0.00042888200008051855, "loss": 1.3415, "step": 71160 }, { "epoch": 2.86, "learning_rate": 0.00042887193526309434, "loss": 1.0859, "step": 71170 }, { "epoch": 2.86, "learning_rate": 0.0004288618704456701, "loss": 1.3279, "step": 71180 }, { "epoch": 2.86, "learning_rate": 0.0004288518056282459, "loss": 1.2593, "step": 71190 }, { "epoch": 2.86, "learning_rate": 0.0004288417408108217, "loss": 1.5373, "step": 71200 }, { "epoch": 2.86, "learning_rate": 0.0004288316759933975, "loss": 1.247, "step": 71210 }, { "epoch": 2.86, "learning_rate": 0.0004288216111759733, "loss": 1.0885, "step": 71220 }, { "epoch": 2.86, "learning_rate": 0.0004288115463585491, "loss": 1.6123, "step": 71230 }, { "epoch": 2.87, "learning_rate": 0.0004288014815411249, "loss": 1.1132, "step": 71240 }, { "epoch": 2.87, "learning_rate": 0.00042879141672370067, "loss": 1.2538, "step": 71250 }, { "epoch": 2.87, "learning_rate": 0.0004287813519062764, "loss": 1.4884, "step": 71260 }, { "epoch": 2.87, "learning_rate": 0.0004287712870888522, "loss": 1.4278, "step": 71270 }, { "epoch": 2.87, "learning_rate": 0.000428761222271428, "loss": 1.1713, "step": 71280 }, { "epoch": 2.87, "learning_rate": 0.0004287511574540038, "loss": 1.1201, "step": 71290 }, { "epoch": 2.87, "learning_rate": 0.00042874109263657957, "loss": 1.0205, "step": 71300 }, { "epoch": 2.87, "learning_rate": 0.00042873102781915536, "loss": 1.4857, "step": 71310 }, { "epoch": 2.87, "learning_rate": 0.00042872096300173115, "loss": 1.302, "step": 71320 }, { "epoch": 2.87, "learning_rate": 0.00042871089818430694, "loss": 1.4195, "step": 71330 }, { "epoch": 2.87, "learning_rate": 0.0004287008333668828, "loss": 1.3905, "step": 71340 }, { "epoch": 2.87, "learning_rate": 0.0004286907685494585, "loss": 1.3765, "step": 71350 }, { "epoch": 2.87, "learning_rate": 0.0004286807037320343, "loss": 1.2693, "step": 71360 }, { "epoch": 2.87, "learning_rate": 0.0004286706389146101, "loss": 1.4027, "step": 71370 }, { "epoch": 2.87, "learning_rate": 0.0004286605740971859, "loss": 1.2113, "step": 71380 }, { "epoch": 2.87, "learning_rate": 0.0004286505092797617, "loss": 1.2917, "step": 71390 }, { "epoch": 2.87, "learning_rate": 0.0004286404444623375, "loss": 1.3475, "step": 71400 }, { "epoch": 2.87, "learning_rate": 0.00042863037964491327, "loss": 1.0927, "step": 71410 }, { "epoch": 2.87, "learning_rate": 0.00042862031482748906, "loss": 1.5709, "step": 71420 }, { "epoch": 2.87, "learning_rate": 0.0004286102500100648, "loss": 1.1282, "step": 71430 }, { "epoch": 2.87, "learning_rate": 0.0004286001851926406, "loss": 1.492, "step": 71440 }, { "epoch": 2.87, "learning_rate": 0.0004285901203752164, "loss": 1.2934, "step": 71450 }, { "epoch": 2.87, "learning_rate": 0.00042858005555779217, "loss": 1.207, "step": 71460 }, { "epoch": 2.87, "learning_rate": 0.00042856999074036796, "loss": 1.4485, "step": 71470 }, { "epoch": 2.87, "learning_rate": 0.0004285599259229438, "loss": 1.1605, "step": 71480 }, { "epoch": 2.88, "learning_rate": 0.0004285498611055196, "loss": 1.525, "step": 71490 }, { "epoch": 2.88, "learning_rate": 0.0004285397962880954, "loss": 1.2744, "step": 71500 }, { "epoch": 2.88, "learning_rate": 0.0004285297314706712, "loss": 1.7365, "step": 71510 }, { "epoch": 2.88, "learning_rate": 0.0004285196666532469, "loss": 1.2667, "step": 71520 }, { "epoch": 2.88, "learning_rate": 0.0004285096018358227, "loss": 1.1877, "step": 71530 }, { "epoch": 2.88, "learning_rate": 0.0004284995370183985, "loss": 1.4046, "step": 71540 }, { "epoch": 2.88, "learning_rate": 0.0004284894722009743, "loss": 1.4038, "step": 71550 }, { "epoch": 2.88, "learning_rate": 0.00042847940738355007, "loss": 1.1545, "step": 71560 }, { "epoch": 2.88, "learning_rate": 0.00042846934256612586, "loss": 1.3958, "step": 71570 }, { "epoch": 2.88, "learning_rate": 0.00042845927774870165, "loss": 1.2341, "step": 71580 }, { "epoch": 2.88, "learning_rate": 0.00042844921293127745, "loss": 1.1234, "step": 71590 }, { "epoch": 2.88, "learning_rate": 0.0004284391481138532, "loss": 1.2765, "step": 71600 }, { "epoch": 2.88, "learning_rate": 0.00042842908329642897, "loss": 1.2435, "step": 71610 }, { "epoch": 2.88, "learning_rate": 0.00042841901847900476, "loss": 1.5595, "step": 71620 }, { "epoch": 2.88, "learning_rate": 0.0004284089536615806, "loss": 1.0983, "step": 71630 }, { "epoch": 2.88, "learning_rate": 0.0004283988888441564, "loss": 1.1567, "step": 71640 }, { "epoch": 2.88, "learning_rate": 0.0004283888240267322, "loss": 1.2936, "step": 71650 }, { "epoch": 2.88, "learning_rate": 0.000428378759209308, "loss": 1.1465, "step": 71660 }, { "epoch": 2.88, "learning_rate": 0.00042836869439188377, "loss": 1.2249, "step": 71670 }, { "epoch": 2.88, "learning_rate": 0.00042835862957445956, "loss": 1.3765, "step": 71680 }, { "epoch": 2.88, "learning_rate": 0.0004283485647570353, "loss": 1.0865, "step": 71690 }, { "epoch": 2.88, "learning_rate": 0.0004283384999396111, "loss": 1.5668, "step": 71700 }, { "epoch": 2.88, "learning_rate": 0.0004283284351221869, "loss": 1.2825, "step": 71710 }, { "epoch": 2.88, "learning_rate": 0.00042831837030476267, "loss": 1.1193, "step": 71720 }, { "epoch": 2.88, "learning_rate": 0.00042830830548733846, "loss": 1.3873, "step": 71730 }, { "epoch": 2.89, "learning_rate": 0.00042829824066991425, "loss": 1.2841, "step": 71740 }, { "epoch": 2.89, "learning_rate": 0.00042828817585249004, "loss": 1.1442, "step": 71750 }, { "epoch": 2.89, "learning_rate": 0.00042827811103506583, "loss": 1.2223, "step": 71760 }, { "epoch": 2.89, "learning_rate": 0.00042826804621764157, "loss": 1.1678, "step": 71770 }, { "epoch": 2.89, "learning_rate": 0.0004282579814002174, "loss": 1.1671, "step": 71780 }, { "epoch": 2.89, "learning_rate": 0.0004282479165827932, "loss": 1.3327, "step": 71790 }, { "epoch": 2.89, "learning_rate": 0.000428237851765369, "loss": 1.2325, "step": 71800 }, { "epoch": 2.89, "learning_rate": 0.0004282277869479448, "loss": 1.3303, "step": 71810 }, { "epoch": 2.89, "learning_rate": 0.0004282177221305206, "loss": 1.4034, "step": 71820 }, { "epoch": 2.89, "learning_rate": 0.00042820765731309637, "loss": 1.1596, "step": 71830 }, { "epoch": 2.89, "learning_rate": 0.00042819759249567216, "loss": 1.5593, "step": 71840 }, { "epoch": 2.89, "learning_rate": 0.0004281875276782479, "loss": 1.1357, "step": 71850 }, { "epoch": 2.89, "learning_rate": 0.0004281774628608237, "loss": 1.1322, "step": 71860 }, { "epoch": 2.89, "learning_rate": 0.0004281673980433995, "loss": 1.2778, "step": 71870 }, { "epoch": 2.89, "learning_rate": 0.00042815733322597527, "loss": 1.2809, "step": 71880 }, { "epoch": 2.89, "learning_rate": 0.00042814726840855106, "loss": 1.0889, "step": 71890 }, { "epoch": 2.89, "learning_rate": 0.00042813720359112685, "loss": 1.2258, "step": 71900 }, { "epoch": 2.89, "learning_rate": 0.00042812713877370264, "loss": 1.1052, "step": 71910 }, { "epoch": 2.89, "learning_rate": 0.0004281170739562785, "loss": 0.9645, "step": 71920 }, { "epoch": 2.89, "learning_rate": 0.0004281070091388543, "loss": 1.4287, "step": 71930 }, { "epoch": 2.89, "learning_rate": 0.00042809694432143, "loss": 1.3466, "step": 71940 }, { "epoch": 2.89, "learning_rate": 0.0004280868795040058, "loss": 1.3498, "step": 71950 }, { "epoch": 2.89, "learning_rate": 0.0004280768146865816, "loss": 1.2839, "step": 71960 }, { "epoch": 2.89, "learning_rate": 0.0004280667498691574, "loss": 1.234, "step": 71970 }, { "epoch": 2.89, "learning_rate": 0.0004280566850517332, "loss": 1.3474, "step": 71980 }, { "epoch": 2.9, "learning_rate": 0.00042804662023430897, "loss": 1.2703, "step": 71990 }, { "epoch": 2.9, "learning_rate": 0.00042803655541688476, "loss": 1.3241, "step": 72000 }, { "epoch": 2.9, "learning_rate": 0.00042802649059946055, "loss": 1.2768, "step": 72010 }, { "epoch": 2.9, "learning_rate": 0.0004280164257820363, "loss": 1.1326, "step": 72020 }, { "epoch": 2.9, "learning_rate": 0.0004280063609646121, "loss": 1.4211, "step": 72030 }, { "epoch": 2.9, "learning_rate": 0.00042799629614718787, "loss": 1.2877, "step": 72040 }, { "epoch": 2.9, "learning_rate": 0.00042798623132976366, "loss": 1.0935, "step": 72050 }, { "epoch": 2.9, "learning_rate": 0.00042797616651233945, "loss": 1.3197, "step": 72060 }, { "epoch": 2.9, "learning_rate": 0.0004279661016949153, "loss": 1.3654, "step": 72070 }, { "epoch": 2.9, "learning_rate": 0.0004279560368774911, "loss": 1.1486, "step": 72080 }, { "epoch": 2.9, "learning_rate": 0.0004279459720600669, "loss": 1.2661, "step": 72090 }, { "epoch": 2.9, "learning_rate": 0.00042793590724264267, "loss": 1.1711, "step": 72100 }, { "epoch": 2.9, "learning_rate": 0.0004279258424252184, "loss": 1.5735, "step": 72110 }, { "epoch": 2.9, "learning_rate": 0.0004279157776077942, "loss": 1.3696, "step": 72120 }, { "epoch": 2.9, "learning_rate": 0.00042790571279037, "loss": 1.3401, "step": 72130 }, { "epoch": 2.9, "learning_rate": 0.0004278956479729458, "loss": 1.4852, "step": 72140 }, { "epoch": 2.9, "learning_rate": 0.00042788558315552157, "loss": 1.2988, "step": 72150 }, { "epoch": 2.9, "learning_rate": 0.00042787551833809736, "loss": 1.2829, "step": 72160 }, { "epoch": 2.9, "learning_rate": 0.00042786545352067315, "loss": 1.2591, "step": 72170 }, { "epoch": 2.9, "learning_rate": 0.00042785538870324894, "loss": 1.2602, "step": 72180 }, { "epoch": 2.9, "learning_rate": 0.0004278453238858247, "loss": 1.4631, "step": 72190 }, { "epoch": 2.9, "learning_rate": 0.00042783525906840047, "loss": 1.2812, "step": 72200 }, { "epoch": 2.9, "learning_rate": 0.00042782519425097626, "loss": 1.1261, "step": 72210 }, { "epoch": 2.9, "learning_rate": 0.0004278151294335521, "loss": 1.2904, "step": 72220 }, { "epoch": 2.9, "learning_rate": 0.0004278050646161279, "loss": 1.2673, "step": 72230 }, { "epoch": 2.91, "learning_rate": 0.0004277949997987037, "loss": 1.3615, "step": 72240 }, { "epoch": 2.91, "learning_rate": 0.0004277849349812795, "loss": 1.085, "step": 72250 }, { "epoch": 2.91, "learning_rate": 0.00042777487016385527, "loss": 1.1811, "step": 72260 }, { "epoch": 2.91, "learning_rate": 0.00042776480534643106, "loss": 1.4898, "step": 72270 }, { "epoch": 2.91, "learning_rate": 0.0004277547405290068, "loss": 1.2177, "step": 72280 }, { "epoch": 2.91, "learning_rate": 0.0004277446757115826, "loss": 1.0859, "step": 72290 }, { "epoch": 2.91, "learning_rate": 0.0004277346108941584, "loss": 1.2317, "step": 72300 }, { "epoch": 2.91, "learning_rate": 0.00042772454607673416, "loss": 1.5696, "step": 72310 }, { "epoch": 2.91, "learning_rate": 0.00042771448125930996, "loss": 1.2015, "step": 72320 }, { "epoch": 2.91, "learning_rate": 0.00042770441644188575, "loss": 1.293, "step": 72330 }, { "epoch": 2.91, "learning_rate": 0.00042769435162446154, "loss": 1.2944, "step": 72340 }, { "epoch": 2.91, "learning_rate": 0.00042768428680703733, "loss": 1.2615, "step": 72350 }, { "epoch": 2.91, "learning_rate": 0.0004276742219896131, "loss": 1.0394, "step": 72360 }, { "epoch": 2.91, "learning_rate": 0.0004276641571721889, "loss": 1.4327, "step": 72370 }, { "epoch": 2.91, "learning_rate": 0.0004276540923547647, "loss": 1.201, "step": 72380 }, { "epoch": 2.91, "learning_rate": 0.0004276440275373405, "loss": 1.4023, "step": 72390 }, { "epoch": 2.91, "learning_rate": 0.0004276339627199163, "loss": 1.2894, "step": 72400 }, { "epoch": 2.91, "learning_rate": 0.00042762389790249207, "loss": 1.2829, "step": 72410 }, { "epoch": 2.91, "learning_rate": 0.00042761383308506786, "loss": 1.2455, "step": 72420 }, { "epoch": 2.91, "learning_rate": 0.00042760376826764365, "loss": 1.2755, "step": 72430 }, { "epoch": 2.91, "learning_rate": 0.00042759370345021944, "loss": 1.3826, "step": 72440 }, { "epoch": 2.91, "learning_rate": 0.0004275836386327952, "loss": 1.5725, "step": 72450 }, { "epoch": 2.91, "learning_rate": 0.00042757357381537097, "loss": 1.159, "step": 72460 }, { "epoch": 2.91, "learning_rate": 0.00042756350899794676, "loss": 1.182, "step": 72470 }, { "epoch": 2.92, "learning_rate": 0.00042755344418052255, "loss": 1.3773, "step": 72480 }, { "epoch": 2.92, "learning_rate": 0.00042754337936309834, "loss": 1.3474, "step": 72490 }, { "epoch": 2.92, "learning_rate": 0.00042753331454567414, "loss": 1.2851, "step": 72500 }, { "epoch": 2.92, "learning_rate": 0.00042752324972825, "loss": 1.1245, "step": 72510 }, { "epoch": 2.92, "learning_rate": 0.00042751318491082577, "loss": 1.1737, "step": 72520 }, { "epoch": 2.92, "learning_rate": 0.0004275031200934015, "loss": 1.1641, "step": 72530 }, { "epoch": 2.92, "learning_rate": 0.0004274930552759773, "loss": 1.5165, "step": 72540 }, { "epoch": 2.92, "learning_rate": 0.0004274829904585531, "loss": 1.4, "step": 72550 }, { "epoch": 2.92, "learning_rate": 0.0004274729256411289, "loss": 1.215, "step": 72560 }, { "epoch": 2.92, "learning_rate": 0.00042746286082370467, "loss": 1.1018, "step": 72570 }, { "epoch": 2.92, "learning_rate": 0.00042745279600628046, "loss": 1.4739, "step": 72580 }, { "epoch": 2.92, "learning_rate": 0.00042744273118885625, "loss": 1.2126, "step": 72590 }, { "epoch": 2.92, "learning_rate": 0.00042743266637143204, "loss": 1.2175, "step": 72600 }, { "epoch": 2.92, "learning_rate": 0.0004274226015540078, "loss": 1.367, "step": 72610 }, { "epoch": 2.92, "learning_rate": 0.00042741253673658357, "loss": 1.0425, "step": 72620 }, { "epoch": 2.92, "learning_rate": 0.00042740247191915936, "loss": 1.2655, "step": 72630 }, { "epoch": 2.92, "learning_rate": 0.00042739240710173515, "loss": 1.3622, "step": 72640 }, { "epoch": 2.92, "learning_rate": 0.000427382342284311, "loss": 1.218, "step": 72650 }, { "epoch": 2.92, "learning_rate": 0.0004273722774668868, "loss": 1.3427, "step": 72660 }, { "epoch": 2.92, "learning_rate": 0.0004273622126494626, "loss": 1.2982, "step": 72670 }, { "epoch": 2.92, "learning_rate": 0.00042735214783203837, "loss": 1.1059, "step": 72680 }, { "epoch": 2.92, "learning_rate": 0.00042734208301461416, "loss": 1.2891, "step": 72690 }, { "epoch": 2.92, "learning_rate": 0.0004273320181971899, "loss": 1.3255, "step": 72700 }, { "epoch": 2.92, "learning_rate": 0.0004273219533797657, "loss": 1.304, "step": 72710 }, { "epoch": 2.92, "learning_rate": 0.0004273118885623415, "loss": 1.1481, "step": 72720 }, { "epoch": 2.93, "learning_rate": 0.00042730182374491727, "loss": 1.2613, "step": 72730 }, { "epoch": 2.93, "learning_rate": 0.00042729175892749306, "loss": 1.2722, "step": 72740 }, { "epoch": 2.93, "learning_rate": 0.00042728169411006885, "loss": 1.096, "step": 72750 }, { "epoch": 2.93, "learning_rate": 0.00042727162929264464, "loss": 1.1251, "step": 72760 }, { "epoch": 2.93, "learning_rate": 0.00042726156447522043, "loss": 1.2565, "step": 72770 }, { "epoch": 2.93, "learning_rate": 0.00042725149965779617, "loss": 1.2147, "step": 72780 }, { "epoch": 2.93, "learning_rate": 0.00042724143484037196, "loss": 1.3669, "step": 72790 }, { "epoch": 2.93, "learning_rate": 0.0004272313700229478, "loss": 1.376, "step": 72800 }, { "epoch": 2.93, "learning_rate": 0.0004272213052055236, "loss": 1.274, "step": 72810 }, { "epoch": 2.93, "learning_rate": 0.0004272112403880994, "loss": 1.2577, "step": 72820 }, { "epoch": 2.93, "learning_rate": 0.0004272011755706752, "loss": 1.4524, "step": 72830 }, { "epoch": 2.93, "learning_rate": 0.00042719111075325097, "loss": 1.3178, "step": 72840 }, { "epoch": 2.93, "learning_rate": 0.00042718104593582676, "loss": 1.2706, "step": 72850 }, { "epoch": 2.93, "learning_rate": 0.00042717098111840255, "loss": 1.1603, "step": 72860 }, { "epoch": 2.93, "learning_rate": 0.0004271609163009783, "loss": 1.5552, "step": 72870 }, { "epoch": 2.93, "learning_rate": 0.0004271508514835541, "loss": 1.3589, "step": 72880 }, { "epoch": 2.93, "learning_rate": 0.00042714078666612987, "loss": 1.4928, "step": 72890 }, { "epoch": 2.93, "learning_rate": 0.00042713072184870566, "loss": 1.2095, "step": 72900 }, { "epoch": 2.93, "learning_rate": 0.00042712065703128145, "loss": 1.606, "step": 72910 }, { "epoch": 2.93, "learning_rate": 0.00042711059221385724, "loss": 1.1787, "step": 72920 }, { "epoch": 2.93, "learning_rate": 0.00042710052739643303, "loss": 1.2098, "step": 72930 }, { "epoch": 2.93, "learning_rate": 0.0004270904625790088, "loss": 1.1842, "step": 72940 }, { "epoch": 2.93, "learning_rate": 0.0004270803977615846, "loss": 1.2236, "step": 72950 }, { "epoch": 2.93, "learning_rate": 0.0004270703329441604, "loss": 1.153, "step": 72960 }, { "epoch": 2.93, "learning_rate": 0.0004270602681267362, "loss": 1.0669, "step": 72970 }, { "epoch": 2.94, "learning_rate": 0.000427050203309312, "loss": 1.4646, "step": 72980 }, { "epoch": 2.94, "learning_rate": 0.0004270401384918878, "loss": 0.8925, "step": 72990 }, { "epoch": 2.94, "learning_rate": 0.00042703007367446357, "loss": 1.3977, "step": 73000 }, { "epoch": 2.94, "learning_rate": 0.00042702000885703936, "loss": 1.2048, "step": 73010 }, { "epoch": 2.94, "learning_rate": 0.00042700994403961515, "loss": 1.1965, "step": 73020 }, { "epoch": 2.94, "learning_rate": 0.00042699987922219094, "loss": 1.174, "step": 73030 }, { "epoch": 2.94, "learning_rate": 0.0004269898144047667, "loss": 1.5716, "step": 73040 }, { "epoch": 2.94, "learning_rate": 0.00042697974958734247, "loss": 1.1435, "step": 73050 }, { "epoch": 2.94, "learning_rate": 0.00042696968476991826, "loss": 1.4278, "step": 73060 }, { "epoch": 2.94, "learning_rate": 0.00042695961995249405, "loss": 1.5153, "step": 73070 }, { "epoch": 2.94, "learning_rate": 0.00042694955513506984, "loss": 1.4208, "step": 73080 }, { "epoch": 2.94, "learning_rate": 0.0004269394903176457, "loss": 1.6951, "step": 73090 }, { "epoch": 2.94, "learning_rate": 0.0004269294255002215, "loss": 1.4433, "step": 73100 }, { "epoch": 2.94, "learning_rate": 0.00042691936068279726, "loss": 1.2979, "step": 73110 }, { "epoch": 2.94, "learning_rate": 0.000426909295865373, "loss": 1.3292, "step": 73120 }, { "epoch": 2.94, "learning_rate": 0.0004268992310479488, "loss": 1.5296, "step": 73130 }, { "epoch": 2.94, "learning_rate": 0.0004268891662305246, "loss": 1.37, "step": 73140 }, { "epoch": 2.94, "learning_rate": 0.0004268791014131004, "loss": 1.3052, "step": 73150 }, { "epoch": 2.94, "learning_rate": 0.00042686903659567616, "loss": 1.3562, "step": 73160 }, { "epoch": 2.94, "learning_rate": 0.00042685897177825196, "loss": 1.218, "step": 73170 }, { "epoch": 2.94, "learning_rate": 0.00042684890696082775, "loss": 1.4464, "step": 73180 }, { "epoch": 2.94, "learning_rate": 0.00042683884214340354, "loss": 1.0912, "step": 73190 }, { "epoch": 2.94, "learning_rate": 0.0004268287773259793, "loss": 1.2703, "step": 73200 }, { "epoch": 2.94, "learning_rate": 0.00042681871250855506, "loss": 1.1426, "step": 73210 }, { "epoch": 2.94, "learning_rate": 0.00042680864769113086, "loss": 1.7352, "step": 73220 }, { "epoch": 2.95, "learning_rate": 0.00042679858287370665, "loss": 1.3925, "step": 73230 }, { "epoch": 2.95, "learning_rate": 0.0004267885180562825, "loss": 1.2677, "step": 73240 }, { "epoch": 2.95, "learning_rate": 0.0004267784532388583, "loss": 0.9102, "step": 73250 }, { "epoch": 2.95, "learning_rate": 0.00042676838842143407, "loss": 1.1873, "step": 73260 }, { "epoch": 2.95, "learning_rate": 0.00042675832360400986, "loss": 1.0656, "step": 73270 }, { "epoch": 2.95, "learning_rate": 0.00042674825878658565, "loss": 1.267, "step": 73280 }, { "epoch": 2.95, "learning_rate": 0.0004267381939691614, "loss": 1.0807, "step": 73290 }, { "epoch": 2.95, "learning_rate": 0.0004267281291517372, "loss": 1.469, "step": 73300 }, { "epoch": 2.95, "learning_rate": 0.00042671806433431297, "loss": 1.29, "step": 73310 }, { "epoch": 2.95, "learning_rate": 0.00042670799951688876, "loss": 1.0908, "step": 73320 }, { "epoch": 2.95, "learning_rate": 0.00042669793469946455, "loss": 1.094, "step": 73330 }, { "epoch": 2.95, "learning_rate": 0.00042668786988204034, "loss": 1.4303, "step": 73340 }, { "epoch": 2.95, "learning_rate": 0.00042667780506461614, "loss": 1.1165, "step": 73350 }, { "epoch": 2.95, "learning_rate": 0.0004266677402471919, "loss": 1.3084, "step": 73360 }, { "epoch": 2.95, "learning_rate": 0.00042665767542976766, "loss": 1.4193, "step": 73370 }, { "epoch": 2.95, "learning_rate": 0.0004266476106123435, "loss": 1.17, "step": 73380 }, { "epoch": 2.95, "learning_rate": 0.0004266375457949193, "loss": 1.2272, "step": 73390 }, { "epoch": 2.95, "learning_rate": 0.0004266274809774951, "loss": 1.1222, "step": 73400 }, { "epoch": 2.95, "learning_rate": 0.0004266174161600709, "loss": 1.3453, "step": 73410 }, { "epoch": 2.95, "learning_rate": 0.00042660735134264667, "loss": 1.2658, "step": 73420 }, { "epoch": 2.95, "learning_rate": 0.00042659728652522246, "loss": 1.375, "step": 73430 }, { "epoch": 2.95, "learning_rate": 0.00042658722170779825, "loss": 1.2885, "step": 73440 }, { "epoch": 2.95, "learning_rate": 0.00042657715689037404, "loss": 1.165, "step": 73450 }, { "epoch": 2.95, "learning_rate": 0.0004265670920729498, "loss": 1.3715, "step": 73460 }, { "epoch": 2.95, "learning_rate": 0.00042655702725552557, "loss": 1.272, "step": 73470 }, { "epoch": 2.96, "learning_rate": 0.00042654696243810136, "loss": 1.1831, "step": 73480 }, { "epoch": 2.96, "learning_rate": 0.00042653689762067715, "loss": 1.3601, "step": 73490 }, { "epoch": 2.96, "learning_rate": 0.00042652683280325294, "loss": 1.2727, "step": 73500 }, { "epoch": 2.96, "learning_rate": 0.00042651676798582873, "loss": 1.3293, "step": 73510 }, { "epoch": 2.96, "learning_rate": 0.0004265067031684045, "loss": 1.0653, "step": 73520 }, { "epoch": 2.96, "learning_rate": 0.00042649663835098037, "loss": 1.5086, "step": 73530 }, { "epoch": 2.96, "learning_rate": 0.0004264865735335561, "loss": 1.449, "step": 73540 }, { "epoch": 2.96, "learning_rate": 0.0004264765087161319, "loss": 1.1596, "step": 73550 }, { "epoch": 2.96, "learning_rate": 0.0004264664438987077, "loss": 1.3579, "step": 73560 }, { "epoch": 2.96, "learning_rate": 0.0004264563790812835, "loss": 1.089, "step": 73570 }, { "epoch": 2.96, "learning_rate": 0.00042644631426385927, "loss": 1.0959, "step": 73580 }, { "epoch": 2.96, "learning_rate": 0.00042643624944643506, "loss": 1.1534, "step": 73590 }, { "epoch": 2.96, "learning_rate": 0.00042642618462901085, "loss": 1.2774, "step": 73600 }, { "epoch": 2.96, "learning_rate": 0.00042641611981158664, "loss": 1.1123, "step": 73610 }, { "epoch": 2.96, "learning_rate": 0.00042640605499416243, "loss": 1.3238, "step": 73620 }, { "epoch": 2.96, "learning_rate": 0.00042639599017673817, "loss": 1.3251, "step": 73630 }, { "epoch": 2.96, "learning_rate": 0.00042638592535931396, "loss": 1.2336, "step": 73640 }, { "epoch": 2.96, "learning_rate": 0.00042637586054188975, "loss": 1.4405, "step": 73650 }, { "epoch": 2.96, "learning_rate": 0.00042636579572446554, "loss": 1.2985, "step": 73660 }, { "epoch": 2.96, "learning_rate": 0.00042635573090704133, "loss": 1.3871, "step": 73670 }, { "epoch": 2.96, "learning_rate": 0.0004263456660896172, "loss": 1.2334, "step": 73680 }, { "epoch": 2.96, "learning_rate": 0.00042633560127219297, "loss": 1.2415, "step": 73690 }, { "epoch": 2.96, "learning_rate": 0.00042632553645476876, "loss": 1.4566, "step": 73700 }, { "epoch": 2.96, "learning_rate": 0.0004263154716373445, "loss": 1.3197, "step": 73710 }, { "epoch": 2.96, "learning_rate": 0.0004263054068199203, "loss": 1.3387, "step": 73720 }, { "epoch": 2.97, "learning_rate": 0.0004262953420024961, "loss": 1.3416, "step": 73730 }, { "epoch": 2.97, "learning_rate": 0.00042628527718507187, "loss": 1.1111, "step": 73740 }, { "epoch": 2.97, "learning_rate": 0.00042627521236764766, "loss": 1.1484, "step": 73750 }, { "epoch": 2.97, "learning_rate": 0.00042626514755022345, "loss": 1.2999, "step": 73760 }, { "epoch": 2.97, "learning_rate": 0.00042625508273279924, "loss": 1.3361, "step": 73770 }, { "epoch": 2.97, "learning_rate": 0.00042624501791537503, "loss": 1.3267, "step": 73780 }, { "epoch": 2.97, "learning_rate": 0.0004262349530979508, "loss": 1.3185, "step": 73790 }, { "epoch": 2.97, "learning_rate": 0.00042622488828052656, "loss": 1.1243, "step": 73800 }, { "epoch": 2.97, "learning_rate": 0.00042621482346310235, "loss": 1.4371, "step": 73810 }, { "epoch": 2.97, "learning_rate": 0.0004262047586456782, "loss": 1.2917, "step": 73820 }, { "epoch": 2.97, "learning_rate": 0.000426194693828254, "loss": 1.4072, "step": 73830 }, { "epoch": 2.97, "learning_rate": 0.0004261846290108298, "loss": 1.2841, "step": 73840 }, { "epoch": 2.97, "learning_rate": 0.00042617456419340557, "loss": 1.099, "step": 73850 }, { "epoch": 2.97, "learning_rate": 0.00042616449937598136, "loss": 1.1165, "step": 73860 }, { "epoch": 2.97, "learning_rate": 0.00042615443455855715, "loss": 1.2739, "step": 73870 }, { "epoch": 2.97, "learning_rate": 0.0004261443697411329, "loss": 1.0629, "step": 73880 }, { "epoch": 2.97, "learning_rate": 0.0004261343049237087, "loss": 0.961, "step": 73890 }, { "epoch": 2.97, "learning_rate": 0.00042612424010628447, "loss": 1.574, "step": 73900 }, { "epoch": 2.97, "learning_rate": 0.00042611417528886026, "loss": 1.174, "step": 73910 }, { "epoch": 2.97, "learning_rate": 0.00042610411047143605, "loss": 1.3059, "step": 73920 }, { "epoch": 2.97, "learning_rate": 0.00042609404565401184, "loss": 1.3167, "step": 73930 }, { "epoch": 2.97, "learning_rate": 0.00042608398083658763, "loss": 1.2116, "step": 73940 }, { "epoch": 2.97, "learning_rate": 0.0004260739160191634, "loss": 1.4202, "step": 73950 }, { "epoch": 2.97, "learning_rate": 0.00042606385120173916, "loss": 1.263, "step": 73960 }, { "epoch": 2.97, "learning_rate": 0.000426053786384315, "loss": 1.3382, "step": 73970 }, { "epoch": 2.98, "learning_rate": 0.0004260437215668908, "loss": 1.1998, "step": 73980 }, { "epoch": 2.98, "learning_rate": 0.0004260336567494666, "loss": 1.0869, "step": 73990 }, { "epoch": 2.98, "learning_rate": 0.0004260235919320424, "loss": 1.2812, "step": 74000 }, { "epoch": 2.98, "learning_rate": 0.00042601352711461816, "loss": 1.1718, "step": 74010 }, { "epoch": 2.98, "learning_rate": 0.00042600346229719396, "loss": 1.0888, "step": 74020 }, { "epoch": 2.98, "learning_rate": 0.00042599339747976975, "loss": 1.0831, "step": 74030 }, { "epoch": 2.98, "learning_rate": 0.00042598333266234554, "loss": 1.2707, "step": 74040 }, { "epoch": 2.98, "learning_rate": 0.0004259732678449213, "loss": 1.2458, "step": 74050 }, { "epoch": 2.98, "learning_rate": 0.00042596320302749706, "loss": 1.2271, "step": 74060 }, { "epoch": 2.98, "learning_rate": 0.00042595313821007285, "loss": 1.3965, "step": 74070 }, { "epoch": 2.98, "learning_rate": 0.00042594307339264865, "loss": 1.399, "step": 74080 }, { "epoch": 2.98, "learning_rate": 0.00042593300857522444, "loss": 1.2487, "step": 74090 }, { "epoch": 2.98, "learning_rate": 0.00042592294375780023, "loss": 1.2081, "step": 74100 }, { "epoch": 2.98, "learning_rate": 0.000425912878940376, "loss": 1.0343, "step": 74110 }, { "epoch": 2.98, "learning_rate": 0.00042590281412295186, "loss": 1.1866, "step": 74120 }, { "epoch": 2.98, "learning_rate": 0.00042589274930552765, "loss": 1.1753, "step": 74130 }, { "epoch": 2.98, "learning_rate": 0.0004258826844881034, "loss": 1.1474, "step": 74140 }, { "epoch": 2.98, "learning_rate": 0.0004258726196706792, "loss": 1.374, "step": 74150 }, { "epoch": 2.98, "learning_rate": 0.00042586255485325497, "loss": 1.0593, "step": 74160 }, { "epoch": 2.98, "learning_rate": 0.00042585249003583076, "loss": 1.0763, "step": 74170 }, { "epoch": 2.98, "learning_rate": 0.00042584242521840655, "loss": 1.1913, "step": 74180 }, { "epoch": 2.98, "learning_rate": 0.00042583236040098234, "loss": 1.2792, "step": 74190 }, { "epoch": 2.98, "learning_rate": 0.00042582229558355813, "loss": 1.2136, "step": 74200 }, { "epoch": 2.98, "learning_rate": 0.0004258122307661339, "loss": 1.2351, "step": 74210 }, { "epoch": 2.99, "learning_rate": 0.00042580216594870966, "loss": 1.091, "step": 74220 }, { "epoch": 2.99, "learning_rate": 0.00042579210113128545, "loss": 1.4096, "step": 74230 }, { "epoch": 2.99, "learning_rate": 0.00042578203631386124, "loss": 1.5041, "step": 74240 }, { "epoch": 2.99, "learning_rate": 0.00042577197149643703, "loss": 1.2958, "step": 74250 }, { "epoch": 2.99, "learning_rate": 0.0004257619066790129, "loss": 1.2171, "step": 74260 }, { "epoch": 2.99, "learning_rate": 0.00042575184186158867, "loss": 1.3154, "step": 74270 }, { "epoch": 2.99, "learning_rate": 0.00042574177704416446, "loss": 1.23, "step": 74280 }, { "epoch": 2.99, "learning_rate": 0.00042573171222674025, "loss": 1.0696, "step": 74290 }, { "epoch": 2.99, "learning_rate": 0.000425721647409316, "loss": 1.2633, "step": 74300 }, { "epoch": 2.99, "learning_rate": 0.0004257115825918918, "loss": 1.3734, "step": 74310 }, { "epoch": 2.99, "learning_rate": 0.00042570151777446757, "loss": 1.2103, "step": 74320 }, { "epoch": 2.99, "learning_rate": 0.00042569145295704336, "loss": 1.1487, "step": 74330 }, { "epoch": 2.99, "learning_rate": 0.00042568138813961915, "loss": 1.2235, "step": 74340 }, { "epoch": 2.99, "learning_rate": 0.00042567132332219494, "loss": 1.4716, "step": 74350 }, { "epoch": 2.99, "learning_rate": 0.00042566125850477073, "loss": 1.23, "step": 74360 }, { "epoch": 2.99, "learning_rate": 0.0004256511936873465, "loss": 1.268, "step": 74370 }, { "epoch": 2.99, "learning_rate": 0.0004256411288699223, "loss": 1.5709, "step": 74380 }, { "epoch": 2.99, "learning_rate": 0.00042563106405249805, "loss": 1.2348, "step": 74390 }, { "epoch": 2.99, "learning_rate": 0.00042562099923507384, "loss": 1.4306, "step": 74400 }, { "epoch": 2.99, "learning_rate": 0.0004256109344176497, "loss": 1.0707, "step": 74410 }, { "epoch": 2.99, "learning_rate": 0.0004256008696002255, "loss": 1.2217, "step": 74420 }, { "epoch": 2.99, "learning_rate": 0.00042559080478280127, "loss": 1.2765, "step": 74430 }, { "epoch": 2.99, "learning_rate": 0.00042558073996537706, "loss": 1.0457, "step": 74440 }, { "epoch": 2.99, "learning_rate": 0.00042557067514795285, "loss": 1.4137, "step": 74450 }, { "epoch": 2.99, "learning_rate": 0.00042556061033052864, "loss": 1.1462, "step": 74460 }, { "epoch": 3.0, "learning_rate": 0.0004255505455131044, "loss": 1.4976, "step": 74470 }, { "epoch": 3.0, "learning_rate": 0.00042554048069568017, "loss": 1.2083, "step": 74480 }, { "epoch": 3.0, "learning_rate": 0.00042553041587825596, "loss": 1.1824, "step": 74490 }, { "epoch": 3.0, "learning_rate": 0.00042552035106083175, "loss": 1.2259, "step": 74500 }, { "epoch": 3.0, "learning_rate": 0.00042551028624340754, "loss": 1.4188, "step": 74510 }, { "epoch": 3.0, "learning_rate": 0.00042550022142598333, "loss": 1.2643, "step": 74520 }, { "epoch": 3.0, "learning_rate": 0.0004254901566085591, "loss": 1.1109, "step": 74530 }, { "epoch": 3.0, "learning_rate": 0.0004254800917911349, "loss": 1.0653, "step": 74540 }, { "epoch": 3.0, "learning_rate": 0.00042547002697371076, "loss": 1.1017, "step": 74550 }, { "epoch": 3.0, "learning_rate": 0.0004254599621562865, "loss": 1.2962, "step": 74560 }, { "epoch": 3.0, "learning_rate": 0.0004254498973388623, "loss": 1.1268, "step": 74570 }, { "epoch": 3.0, "learning_rate": 0.0004254398325214381, "loss": 1.2758, "step": 74580 }, { "epoch": 3.0, "learning_rate": 0.00042542976770401387, "loss": 1.1774, "step": 74590 }, { "epoch": 3.0, "eval_Bleu_1": 0.014737476420037534, "eval_Bleu_2": 1.7458215473869327e-11, "eval_Bleu_3": 2.0907460590113877e-14, "eval_Bleu_4": 7.871360392208877e-16, "eval_ROUGE_L": 0.034671021617610334, "eval_cer": 0.6949424423434472, "eval_em": 0.30467409948542024, "eval_f1": 0.4115594211393248, "eval_loss": 1.7662725448608398, "eval_runtime": 2104.9733, "eval_samples_per_second": 13.294, "eval_steps_per_second": 4.431, "eval_wer": 0.834417989719509, "step": 74592 }, { "epoch": 3.0, "learning_rate": 0.00042541970288658966, "loss": 1.108, "step": 74600 }, { "epoch": 3.0, "learning_rate": 0.00042540963806916545, "loss": 0.9498, "step": 74610 }, { "epoch": 3.0, "learning_rate": 0.00042539957325174124, "loss": 1.1082, "step": 74620 }, { "epoch": 3.0, "learning_rate": 0.00042538950843431703, "loss": 1.0058, "step": 74630 }, { "epoch": 3.0, "learning_rate": 0.00042537944361689277, "loss": 0.9914, "step": 74640 }, { "epoch": 3.0, "learning_rate": 0.00042536937879946856, "loss": 0.9857, "step": 74650 }, { "epoch": 3.0, "learning_rate": 0.00042535931398204435, "loss": 0.8691, "step": 74660 }, { "epoch": 3.0, "learning_rate": 0.00042534924916462014, "loss": 1.0425, "step": 74670 }, { "epoch": 3.0, "learning_rate": 0.00042533918434719593, "loss": 0.9532, "step": 74680 }, { "epoch": 3.0, "learning_rate": 0.0004253291195297717, "loss": 0.8294, "step": 74690 }, { "epoch": 3.0, "learning_rate": 0.00042531905471234757, "loss": 0.8567, "step": 74700 }, { "epoch": 3.0, "learning_rate": 0.00042530898989492336, "loss": 1.1286, "step": 74710 }, { "epoch": 3.01, "learning_rate": 0.00042529892507749915, "loss": 1.1202, "step": 74720 }, { "epoch": 3.01, "learning_rate": 0.0004252888602600749, "loss": 1.1575, "step": 74730 }, { "epoch": 3.01, "learning_rate": 0.0004252787954426507, "loss": 1.1685, "step": 74740 }, { "epoch": 3.01, "learning_rate": 0.00042526873062522647, "loss": 0.8844, "step": 74750 }, { "epoch": 3.01, "learning_rate": 0.00042525866580780226, "loss": 1.0582, "step": 74760 }, { "epoch": 3.01, "learning_rate": 0.00042524860099037805, "loss": 1.0865, "step": 74770 }, { "epoch": 3.01, "learning_rate": 0.00042523853617295384, "loss": 1.1037, "step": 74780 }, { "epoch": 3.01, "learning_rate": 0.00042522847135552963, "loss": 1.0217, "step": 74790 }, { "epoch": 3.01, "learning_rate": 0.0004252184065381054, "loss": 0.9827, "step": 74800 }, { "epoch": 3.01, "learning_rate": 0.00042520834172068116, "loss": 0.9933, "step": 74810 }, { "epoch": 3.01, "learning_rate": 0.00042519827690325695, "loss": 1.1105, "step": 74820 }, { "epoch": 3.01, "learning_rate": 0.00042518821208583274, "loss": 1.2178, "step": 74830 }, { "epoch": 3.01, "learning_rate": 0.00042517814726840853, "loss": 1.3624, "step": 74840 }, { "epoch": 3.01, "learning_rate": 0.0004251680824509844, "loss": 1.0613, "step": 74850 }, { "epoch": 3.01, "learning_rate": 0.00042515801763356016, "loss": 0.9937, "step": 74860 }, { "epoch": 3.01, "learning_rate": 0.00042514795281613595, "loss": 1.1014, "step": 74870 }, { "epoch": 3.01, "learning_rate": 0.00042513788799871175, "loss": 0.9762, "step": 74880 }, { "epoch": 3.01, "learning_rate": 0.00042512782318128754, "loss": 1.0085, "step": 74890 }, { "epoch": 3.01, "learning_rate": 0.0004251177583638633, "loss": 1.0235, "step": 74900 }, { "epoch": 3.01, "learning_rate": 0.00042510769354643906, "loss": 1.0472, "step": 74910 }, { "epoch": 3.01, "learning_rate": 0.00042509762872901485, "loss": 0.931, "step": 74920 }, { "epoch": 3.01, "learning_rate": 0.00042508756391159065, "loss": 0.871, "step": 74930 }, { "epoch": 3.01, "learning_rate": 0.00042507749909416644, "loss": 0.9524, "step": 74940 }, { "epoch": 3.01, "learning_rate": 0.0004250674342767422, "loss": 1.0703, "step": 74950 }, { "epoch": 3.01, "learning_rate": 0.000425057369459318, "loss": 1.281, "step": 74960 }, { "epoch": 3.02, "learning_rate": 0.0004250473046418938, "loss": 1.1748, "step": 74970 }, { "epoch": 3.02, "learning_rate": 0.00042503723982446955, "loss": 0.9145, "step": 74980 }, { "epoch": 3.02, "learning_rate": 0.0004250271750070454, "loss": 1.0194, "step": 74990 }, { "epoch": 3.02, "learning_rate": 0.0004250171101896212, "loss": 0.925, "step": 75000 }, { "epoch": 3.02, "learning_rate": 0.00042500704537219697, "loss": 0.9794, "step": 75010 }, { "epoch": 3.02, "learning_rate": 0.00042499698055477276, "loss": 1.248, "step": 75020 }, { "epoch": 3.02, "learning_rate": 0.00042498691573734855, "loss": 1.0079, "step": 75030 }, { "epoch": 3.02, "learning_rate": 0.00042497685091992434, "loss": 0.985, "step": 75040 }, { "epoch": 3.02, "learning_rate": 0.00042496678610250013, "loss": 1.0175, "step": 75050 }, { "epoch": 3.02, "learning_rate": 0.00042495672128507587, "loss": 0.9578, "step": 75060 }, { "epoch": 3.02, "learning_rate": 0.00042494665646765166, "loss": 0.9449, "step": 75070 }, { "epoch": 3.02, "learning_rate": 0.00042493659165022745, "loss": 1.0958, "step": 75080 }, { "epoch": 3.02, "learning_rate": 0.00042492652683280324, "loss": 0.8689, "step": 75090 }, { "epoch": 3.02, "learning_rate": 0.00042491646201537903, "loss": 1.137, "step": 75100 }, { "epoch": 3.02, "learning_rate": 0.0004249063971979548, "loss": 1.1251, "step": 75110 }, { "epoch": 3.02, "learning_rate": 0.0004248963323805306, "loss": 1.1463, "step": 75120 }, { "epoch": 3.02, "learning_rate": 0.0004248862675631064, "loss": 0.8954, "step": 75130 }, { "epoch": 3.02, "learning_rate": 0.00042487620274568225, "loss": 1.1125, "step": 75140 }, { "epoch": 3.02, "learning_rate": 0.000424866137928258, "loss": 1.0494, "step": 75150 }, { "epoch": 3.02, "learning_rate": 0.0004248560731108338, "loss": 1.2059, "step": 75160 }, { "epoch": 3.02, "learning_rate": 0.00042484600829340957, "loss": 0.9759, "step": 75170 }, { "epoch": 3.02, "learning_rate": 0.00042483594347598536, "loss": 1.0408, "step": 75180 }, { "epoch": 3.02, "learning_rate": 0.00042482587865856115, "loss": 1.2989, "step": 75190 }, { "epoch": 3.02, "learning_rate": 0.00042481581384113694, "loss": 0.9967, "step": 75200 }, { "epoch": 3.02, "learning_rate": 0.00042480574902371273, "loss": 1.0987, "step": 75210 }, { "epoch": 3.03, "learning_rate": 0.0004247956842062885, "loss": 0.9818, "step": 75220 }, { "epoch": 3.03, "learning_rate": 0.00042478561938886426, "loss": 1.2382, "step": 75230 }, { "epoch": 3.03, "learning_rate": 0.00042477555457144005, "loss": 0.9304, "step": 75240 }, { "epoch": 3.03, "learning_rate": 0.00042476548975401584, "loss": 1.1784, "step": 75250 }, { "epoch": 3.03, "learning_rate": 0.00042475542493659163, "loss": 1.1696, "step": 75260 }, { "epoch": 3.03, "learning_rate": 0.0004247453601191674, "loss": 1.0574, "step": 75270 }, { "epoch": 3.03, "learning_rate": 0.0004247352953017432, "loss": 0.9373, "step": 75280 }, { "epoch": 3.03, "learning_rate": 0.00042472523048431906, "loss": 1.0768, "step": 75290 }, { "epoch": 3.03, "learning_rate": 0.00042471516566689485, "loss": 1.1141, "step": 75300 }, { "epoch": 3.03, "learning_rate": 0.00042470510084947064, "loss": 1.0392, "step": 75310 }, { "epoch": 3.03, "learning_rate": 0.0004246950360320464, "loss": 0.9259, "step": 75320 }, { "epoch": 3.03, "learning_rate": 0.00042468497121462217, "loss": 1.1036, "step": 75330 }, { "epoch": 3.03, "learning_rate": 0.00042467490639719796, "loss": 1.218, "step": 75340 }, { "epoch": 3.03, "learning_rate": 0.00042466484157977375, "loss": 0.8061, "step": 75350 }, { "epoch": 3.03, "learning_rate": 0.00042465477676234954, "loss": 0.8694, "step": 75360 }, { "epoch": 3.03, "learning_rate": 0.00042464471194492533, "loss": 1.1645, "step": 75370 }, { "epoch": 3.03, "learning_rate": 0.0004246346471275011, "loss": 0.9812, "step": 75380 }, { "epoch": 3.03, "learning_rate": 0.0004246245823100769, "loss": 0.9339, "step": 75390 }, { "epoch": 3.03, "learning_rate": 0.00042461451749265265, "loss": 1.0593, "step": 75400 }, { "epoch": 3.03, "learning_rate": 0.00042460445267522844, "loss": 1.0985, "step": 75410 }, { "epoch": 3.03, "learning_rate": 0.00042459438785780423, "loss": 1.1718, "step": 75420 }, { "epoch": 3.03, "learning_rate": 0.0004245843230403801, "loss": 0.8199, "step": 75430 }, { "epoch": 3.03, "learning_rate": 0.00042457425822295587, "loss": 0.9612, "step": 75440 }, { "epoch": 3.03, "learning_rate": 0.00042456419340553166, "loss": 1.1862, "step": 75450 }, { "epoch": 3.03, "learning_rate": 0.00042455412858810745, "loss": 1.0115, "step": 75460 }, { "epoch": 3.04, "learning_rate": 0.00042454406377068324, "loss": 1.1856, "step": 75470 }, { "epoch": 3.04, "learning_rate": 0.00042453399895325903, "loss": 1.1152, "step": 75480 }, { "epoch": 3.04, "learning_rate": 0.00042452393413583477, "loss": 1.0403, "step": 75490 }, { "epoch": 3.04, "learning_rate": 0.00042451386931841056, "loss": 1.0057, "step": 75500 }, { "epoch": 3.04, "learning_rate": 0.00042450380450098635, "loss": 1.0292, "step": 75510 }, { "epoch": 3.04, "learning_rate": 0.00042449373968356214, "loss": 0.9449, "step": 75520 }, { "epoch": 3.04, "learning_rate": 0.00042448367486613793, "loss": 1.2301, "step": 75530 }, { "epoch": 3.04, "learning_rate": 0.0004244736100487137, "loss": 0.8447, "step": 75540 }, { "epoch": 3.04, "learning_rate": 0.0004244635452312895, "loss": 1.0265, "step": 75550 }, { "epoch": 3.04, "learning_rate": 0.0004244534804138653, "loss": 1.2302, "step": 75560 }, { "epoch": 3.04, "learning_rate": 0.00042444341559644104, "loss": 1.0443, "step": 75570 }, { "epoch": 3.04, "learning_rate": 0.0004244333507790169, "loss": 1.0412, "step": 75580 }, { "epoch": 3.04, "learning_rate": 0.0004244232859615927, "loss": 0.8623, "step": 75590 }, { "epoch": 3.04, "learning_rate": 0.00042441322114416847, "loss": 0.9931, "step": 75600 }, { "epoch": 3.04, "learning_rate": 0.00042440315632674426, "loss": 1.136, "step": 75610 }, { "epoch": 3.04, "learning_rate": 0.00042439309150932005, "loss": 0.9906, "step": 75620 }, { "epoch": 3.04, "learning_rate": 0.00042438302669189584, "loss": 0.8863, "step": 75630 }, { "epoch": 3.04, "learning_rate": 0.00042437296187447163, "loss": 0.8397, "step": 75640 }, { "epoch": 3.04, "learning_rate": 0.0004243628970570474, "loss": 0.9796, "step": 75650 }, { "epoch": 3.04, "learning_rate": 0.00042435283223962316, "loss": 0.8947, "step": 75660 }, { "epoch": 3.04, "learning_rate": 0.00042434276742219895, "loss": 1.0625, "step": 75670 }, { "epoch": 3.04, "learning_rate": 0.00042433270260477474, "loss": 1.1921, "step": 75680 }, { "epoch": 3.04, "learning_rate": 0.00042432263778735053, "loss": 1.1228, "step": 75690 }, { "epoch": 3.04, "learning_rate": 0.0004243125729699263, "loss": 0.9213, "step": 75700 }, { "epoch": 3.04, "learning_rate": 0.0004243025081525021, "loss": 0.9995, "step": 75710 }, { "epoch": 3.05, "learning_rate": 0.00042429244333507795, "loss": 1.1093, "step": 75720 }, { "epoch": 3.05, "learning_rate": 0.00042428237851765375, "loss": 1.1437, "step": 75730 }, { "epoch": 3.05, "learning_rate": 0.0004242723137002295, "loss": 1.2063, "step": 75740 }, { "epoch": 3.05, "learning_rate": 0.00042426224888280527, "loss": 1.2033, "step": 75750 }, { "epoch": 3.05, "learning_rate": 0.00042425218406538106, "loss": 0.9458, "step": 75760 }, { "epoch": 3.05, "learning_rate": 0.00042424211924795685, "loss": 1.1335, "step": 75770 }, { "epoch": 3.05, "learning_rate": 0.00042423205443053265, "loss": 1.2646, "step": 75780 }, { "epoch": 3.05, "learning_rate": 0.00042422198961310844, "loss": 1.0805, "step": 75790 }, { "epoch": 3.05, "learning_rate": 0.0004242119247956842, "loss": 0.8772, "step": 75800 }, { "epoch": 3.05, "learning_rate": 0.00042420185997826, "loss": 1.2611, "step": 75810 }, { "epoch": 3.05, "learning_rate": 0.00042419179516083575, "loss": 0.8545, "step": 75820 }, { "epoch": 3.05, "learning_rate": 0.00042418173034341154, "loss": 0.9209, "step": 75830 }, { "epoch": 3.05, "learning_rate": 0.00042417166552598734, "loss": 1.0501, "step": 75840 }, { "epoch": 3.05, "learning_rate": 0.0004241616007085631, "loss": 1.169, "step": 75850 }, { "epoch": 3.05, "learning_rate": 0.0004241515358911389, "loss": 0.922, "step": 75860 }, { "epoch": 3.05, "learning_rate": 0.00042414147107371476, "loss": 0.9944, "step": 75870 }, { "epoch": 3.05, "learning_rate": 0.00042413140625629055, "loss": 1.2032, "step": 75880 }, { "epoch": 3.05, "learning_rate": 0.00042412134143886634, "loss": 1.0033, "step": 75890 }, { "epoch": 3.05, "learning_rate": 0.00042411127662144213, "loss": 0.9859, "step": 75900 }, { "epoch": 3.05, "learning_rate": 0.00042410121180401787, "loss": 1.0663, "step": 75910 }, { "epoch": 3.05, "learning_rate": 0.00042409114698659366, "loss": 1.0949, "step": 75920 }, { "epoch": 3.05, "learning_rate": 0.00042408108216916945, "loss": 0.9499, "step": 75930 }, { "epoch": 3.05, "learning_rate": 0.00042407101735174524, "loss": 1.0061, "step": 75940 }, { "epoch": 3.05, "learning_rate": 0.00042406095253432103, "loss": 1.0082, "step": 75950 }, { "epoch": 3.05, "learning_rate": 0.0004240508877168968, "loss": 0.9991, "step": 75960 }, { "epoch": 3.06, "learning_rate": 0.0004240408228994726, "loss": 0.8937, "step": 75970 }, { "epoch": 3.06, "learning_rate": 0.0004240307580820484, "loss": 0.988, "step": 75980 }, { "epoch": 3.06, "learning_rate": 0.00042402069326462414, "loss": 1.2596, "step": 75990 }, { "epoch": 3.06, "learning_rate": 0.00042401062844719993, "loss": 1.1387, "step": 76000 }, { "epoch": 3.06, "learning_rate": 0.0004240005636297757, "loss": 1.046, "step": 76010 }, { "epoch": 3.06, "learning_rate": 0.00042399049881235157, "loss": 1.1496, "step": 76020 }, { "epoch": 3.06, "learning_rate": 0.00042398043399492736, "loss": 1.0929, "step": 76030 }, { "epoch": 3.06, "learning_rate": 0.00042397036917750315, "loss": 1.0829, "step": 76040 }, { "epoch": 3.06, "learning_rate": 0.00042396030436007894, "loss": 0.9903, "step": 76050 }, { "epoch": 3.06, "learning_rate": 0.00042395023954265473, "loss": 1.0449, "step": 76060 }, { "epoch": 3.06, "learning_rate": 0.0004239401747252305, "loss": 1.0305, "step": 76070 }, { "epoch": 3.06, "learning_rate": 0.00042393010990780626, "loss": 0.9796, "step": 76080 }, { "epoch": 3.06, "learning_rate": 0.00042392004509038205, "loss": 1.2807, "step": 76090 }, { "epoch": 3.06, "learning_rate": 0.00042390998027295784, "loss": 1.0717, "step": 76100 }, { "epoch": 3.06, "learning_rate": 0.00042389991545553363, "loss": 0.827, "step": 76110 }, { "epoch": 3.06, "learning_rate": 0.0004238898506381094, "loss": 1.129, "step": 76120 }, { "epoch": 3.06, "learning_rate": 0.0004238797858206852, "loss": 1.0317, "step": 76130 }, { "epoch": 3.06, "learning_rate": 0.000423869721003261, "loss": 0.9708, "step": 76140 }, { "epoch": 3.06, "learning_rate": 0.0004238596561858368, "loss": 0.9432, "step": 76150 }, { "epoch": 3.06, "learning_rate": 0.0004238495913684126, "loss": 0.9824, "step": 76160 }, { "epoch": 3.06, "learning_rate": 0.0004238395265509884, "loss": 0.8689, "step": 76170 }, { "epoch": 3.06, "learning_rate": 0.00042382946173356417, "loss": 1.0192, "step": 76180 }, { "epoch": 3.06, "learning_rate": 0.00042381939691613996, "loss": 1.1727, "step": 76190 }, { "epoch": 3.06, "learning_rate": 0.00042380933209871575, "loss": 0.9051, "step": 76200 }, { "epoch": 3.07, "learning_rate": 0.00042379926728129154, "loss": 1.0097, "step": 76210 }, { "epoch": 3.07, "learning_rate": 0.00042378920246386733, "loss": 1.148, "step": 76220 }, { "epoch": 3.07, "learning_rate": 0.0004237791376464431, "loss": 1.2302, "step": 76230 }, { "epoch": 3.07, "learning_rate": 0.0004237690728290189, "loss": 0.9052, "step": 76240 }, { "epoch": 3.07, "learning_rate": 0.00042375900801159465, "loss": 1.1534, "step": 76250 }, { "epoch": 3.07, "learning_rate": 0.00042374894319417044, "loss": 1.1938, "step": 76260 }, { "epoch": 3.07, "learning_rate": 0.00042373887837674623, "loss": 1.0127, "step": 76270 }, { "epoch": 3.07, "learning_rate": 0.000423728813559322, "loss": 1.112, "step": 76280 }, { "epoch": 3.07, "learning_rate": 0.0004237187487418978, "loss": 1.1278, "step": 76290 }, { "epoch": 3.07, "learning_rate": 0.0004237086839244736, "loss": 1.206, "step": 76300 }, { "epoch": 3.07, "learning_rate": 0.00042369861910704945, "loss": 1.1598, "step": 76310 }, { "epoch": 3.07, "learning_rate": 0.00042368855428962524, "loss": 1.1838, "step": 76320 }, { "epoch": 3.07, "learning_rate": 0.000423678489472201, "loss": 1.0739, "step": 76330 }, { "epoch": 3.07, "learning_rate": 0.00042366842465477677, "loss": 0.9261, "step": 76340 }, { "epoch": 3.07, "learning_rate": 0.00042365835983735256, "loss": 0.8256, "step": 76350 }, { "epoch": 3.07, "learning_rate": 0.00042364829501992835, "loss": 1.1388, "step": 76360 }, { "epoch": 3.07, "learning_rate": 0.00042363823020250414, "loss": 1.2144, "step": 76370 }, { "epoch": 3.07, "learning_rate": 0.00042362816538507993, "loss": 0.8913, "step": 76380 }, { "epoch": 3.07, "learning_rate": 0.0004236181005676557, "loss": 0.9368, "step": 76390 }, { "epoch": 3.07, "learning_rate": 0.0004236080357502315, "loss": 0.9703, "step": 76400 }, { "epoch": 3.07, "learning_rate": 0.0004235979709328073, "loss": 1.1097, "step": 76410 }, { "epoch": 3.07, "learning_rate": 0.00042358790611538304, "loss": 0.9733, "step": 76420 }, { "epoch": 3.07, "learning_rate": 0.00042357784129795883, "loss": 1.0121, "step": 76430 }, { "epoch": 3.07, "learning_rate": 0.0004235677764805346, "loss": 0.9372, "step": 76440 }, { "epoch": 3.07, "learning_rate": 0.00042355771166311046, "loss": 0.9474, "step": 76450 }, { "epoch": 3.08, "learning_rate": 0.00042354764684568626, "loss": 0.875, "step": 76460 }, { "epoch": 3.08, "learning_rate": 0.00042353758202826205, "loss": 1.0381, "step": 76470 }, { "epoch": 3.08, "learning_rate": 0.00042352751721083784, "loss": 1.0181, "step": 76480 }, { "epoch": 3.08, "learning_rate": 0.00042351745239341363, "loss": 0.9247, "step": 76490 }, { "epoch": 3.08, "learning_rate": 0.00042350738757598936, "loss": 1.0527, "step": 76500 }, { "epoch": 3.08, "learning_rate": 0.00042349732275856516, "loss": 1.0714, "step": 76510 }, { "epoch": 3.08, "learning_rate": 0.00042348725794114095, "loss": 1.002, "step": 76520 }, { "epoch": 3.08, "learning_rate": 0.00042347719312371674, "loss": 1.0767, "step": 76530 }, { "epoch": 3.08, "learning_rate": 0.00042346712830629253, "loss": 1.0487, "step": 76540 }, { "epoch": 3.08, "learning_rate": 0.0004234570634888683, "loss": 1.0017, "step": 76550 }, { "epoch": 3.08, "learning_rate": 0.0004234469986714441, "loss": 1.2075, "step": 76560 }, { "epoch": 3.08, "learning_rate": 0.0004234369338540199, "loss": 1.0495, "step": 76570 }, { "epoch": 3.08, "learning_rate": 0.00042342686903659564, "loss": 0.9439, "step": 76580 }, { "epoch": 3.08, "learning_rate": 0.00042341680421917143, "loss": 1.0732, "step": 76590 }, { "epoch": 3.08, "learning_rate": 0.00042340673940174727, "loss": 1.022, "step": 76600 }, { "epoch": 3.08, "learning_rate": 0.00042339667458432306, "loss": 1.0874, "step": 76610 }, { "epoch": 3.08, "learning_rate": 0.00042338660976689885, "loss": 0.9517, "step": 76620 }, { "epoch": 3.08, "learning_rate": 0.00042337654494947464, "loss": 1.074, "step": 76630 }, { "epoch": 3.08, "learning_rate": 0.00042336648013205044, "loss": 1.0516, "step": 76640 }, { "epoch": 3.08, "learning_rate": 0.0004233564153146262, "loss": 1.0888, "step": 76650 }, { "epoch": 3.08, "learning_rate": 0.000423346350497202, "loss": 0.8707, "step": 76660 }, { "epoch": 3.08, "learning_rate": 0.00042333628567977775, "loss": 0.9296, "step": 76670 }, { "epoch": 3.08, "learning_rate": 0.00042332622086235354, "loss": 1.0296, "step": 76680 }, { "epoch": 3.08, "learning_rate": 0.00042331615604492934, "loss": 0.9478, "step": 76690 }, { "epoch": 3.08, "learning_rate": 0.0004233060912275051, "loss": 0.9954, "step": 76700 }, { "epoch": 3.09, "learning_rate": 0.0004232960264100809, "loss": 0.9967, "step": 76710 }, { "epoch": 3.09, "learning_rate": 0.0004232859615926567, "loss": 1.268, "step": 76720 }, { "epoch": 3.09, "learning_rate": 0.0004232758967752325, "loss": 0.9955, "step": 76730 }, { "epoch": 3.09, "learning_rate": 0.0004232658319578083, "loss": 0.8315, "step": 76740 }, { "epoch": 3.09, "learning_rate": 0.0004232557671403841, "loss": 1.1828, "step": 76750 }, { "epoch": 3.09, "learning_rate": 0.00042324570232295987, "loss": 1.2687, "step": 76760 }, { "epoch": 3.09, "learning_rate": 0.00042323563750553566, "loss": 1.1772, "step": 76770 }, { "epoch": 3.09, "learning_rate": 0.00042322557268811145, "loss": 1.3521, "step": 76780 }, { "epoch": 3.09, "learning_rate": 0.00042321550787068724, "loss": 1.2302, "step": 76790 }, { "epoch": 3.09, "learning_rate": 0.00042320544305326303, "loss": 0.8279, "step": 76800 }, { "epoch": 3.09, "learning_rate": 0.0004231953782358388, "loss": 1.153, "step": 76810 }, { "epoch": 3.09, "learning_rate": 0.0004231853134184146, "loss": 1.0315, "step": 76820 }, { "epoch": 3.09, "learning_rate": 0.0004231752486009904, "loss": 0.9533, "step": 76830 }, { "epoch": 3.09, "learning_rate": 0.00042316518378356614, "loss": 1.0894, "step": 76840 }, { "epoch": 3.09, "learning_rate": 0.00042315511896614193, "loss": 0.8687, "step": 76850 }, { "epoch": 3.09, "learning_rate": 0.0004231450541487177, "loss": 1.0194, "step": 76860 }, { "epoch": 3.09, "learning_rate": 0.0004231349893312935, "loss": 1.0118, "step": 76870 }, { "epoch": 3.09, "learning_rate": 0.0004231249245138693, "loss": 0.9772, "step": 76880 }, { "epoch": 3.09, "learning_rate": 0.00042311485969644515, "loss": 0.9943, "step": 76890 }, { "epoch": 3.09, "learning_rate": 0.00042310479487902094, "loss": 1.1078, "step": 76900 }, { "epoch": 3.09, "learning_rate": 0.00042309473006159673, "loss": 1.1345, "step": 76910 }, { "epoch": 3.09, "learning_rate": 0.00042308466524417247, "loss": 1.1404, "step": 76920 }, { "epoch": 3.09, "learning_rate": 0.00042307460042674826, "loss": 1.1108, "step": 76930 }, { "epoch": 3.09, "learning_rate": 0.00042306453560932405, "loss": 1.0306, "step": 76940 }, { "epoch": 3.09, "learning_rate": 0.00042305447079189984, "loss": 1.0644, "step": 76950 }, { "epoch": 3.1, "learning_rate": 0.00042304440597447563, "loss": 1.1457, "step": 76960 }, { "epoch": 3.1, "learning_rate": 0.0004230343411570514, "loss": 0.9326, "step": 76970 }, { "epoch": 3.1, "learning_rate": 0.0004230242763396272, "loss": 1.0211, "step": 76980 }, { "epoch": 3.1, "learning_rate": 0.000423014211522203, "loss": 0.9928, "step": 76990 }, { "epoch": 3.1, "learning_rate": 0.0004230041467047788, "loss": 1.0875, "step": 77000 }, { "epoch": 3.1, "learning_rate": 0.00042299408188735453, "loss": 1.21, "step": 77010 }, { "epoch": 3.1, "learning_rate": 0.0004229840170699303, "loss": 0.9334, "step": 77020 }, { "epoch": 3.1, "learning_rate": 0.0004229739522525061, "loss": 0.9998, "step": 77030 }, { "epoch": 3.1, "learning_rate": 0.00042296388743508196, "loss": 1.105, "step": 77040 }, { "epoch": 3.1, "learning_rate": 0.00042295382261765775, "loss": 0.971, "step": 77050 }, { "epoch": 3.1, "learning_rate": 0.00042294375780023354, "loss": 1.0882, "step": 77060 }, { "epoch": 3.1, "learning_rate": 0.00042293369298280933, "loss": 1.1047, "step": 77070 }, { "epoch": 3.1, "learning_rate": 0.0004229236281653851, "loss": 0.9491, "step": 77080 }, { "epoch": 3.1, "learning_rate": 0.00042291356334796086, "loss": 0.9775, "step": 77090 }, { "epoch": 3.1, "learning_rate": 0.00042290349853053665, "loss": 0.9806, "step": 77100 }, { "epoch": 3.1, "learning_rate": 0.00042289343371311244, "loss": 1.2431, "step": 77110 }, { "epoch": 3.1, "learning_rate": 0.00042288336889568823, "loss": 1.0012, "step": 77120 }, { "epoch": 3.1, "learning_rate": 0.000422873304078264, "loss": 0.9962, "step": 77130 }, { "epoch": 3.1, "learning_rate": 0.0004228632392608398, "loss": 1.0532, "step": 77140 }, { "epoch": 3.1, "learning_rate": 0.0004228531744434156, "loss": 0.8916, "step": 77150 }, { "epoch": 3.1, "learning_rate": 0.0004228431096259914, "loss": 1.0002, "step": 77160 }, { "epoch": 3.1, "learning_rate": 0.0004228330448085672, "loss": 1.0463, "step": 77170 }, { "epoch": 3.1, "learning_rate": 0.0004228229799911429, "loss": 1.2477, "step": 77180 }, { "epoch": 3.1, "learning_rate": 0.00042281291517371877, "loss": 1.0073, "step": 77190 }, { "epoch": 3.1, "learning_rate": 0.00042280285035629456, "loss": 1.0069, "step": 77200 }, { "epoch": 3.11, "learning_rate": 0.00042279278553887035, "loss": 1.0217, "step": 77210 }, { "epoch": 3.11, "learning_rate": 0.00042278272072144614, "loss": 1.0212, "step": 77220 }, { "epoch": 3.11, "learning_rate": 0.00042277265590402193, "loss": 1.0798, "step": 77230 }, { "epoch": 3.11, "learning_rate": 0.0004227625910865977, "loss": 1.1789, "step": 77240 }, { "epoch": 3.11, "learning_rate": 0.0004227525262691735, "loss": 1.2054, "step": 77250 }, { "epoch": 3.11, "learning_rate": 0.00042274246145174925, "loss": 0.9147, "step": 77260 }, { "epoch": 3.11, "learning_rate": 0.00042273239663432504, "loss": 1.1439, "step": 77270 }, { "epoch": 3.11, "learning_rate": 0.00042272233181690083, "loss": 1.1044, "step": 77280 }, { "epoch": 3.11, "learning_rate": 0.0004227122669994766, "loss": 0.8868, "step": 77290 }, { "epoch": 3.11, "learning_rate": 0.0004227022021820524, "loss": 1.1099, "step": 77300 }, { "epoch": 3.11, "learning_rate": 0.0004226921373646282, "loss": 0.981, "step": 77310 }, { "epoch": 3.11, "learning_rate": 0.000422682072547204, "loss": 1.0461, "step": 77320 }, { "epoch": 3.11, "learning_rate": 0.00042267200772977984, "loss": 1.1983, "step": 77330 }, { "epoch": 3.11, "learning_rate": 0.00042266194291235563, "loss": 0.9679, "step": 77340 }, { "epoch": 3.11, "learning_rate": 0.00042265187809493136, "loss": 1.0912, "step": 77350 }, { "epoch": 3.11, "learning_rate": 0.00042264181327750716, "loss": 1.2312, "step": 77360 }, { "epoch": 3.11, "learning_rate": 0.00042263174846008295, "loss": 1.2326, "step": 77370 }, { "epoch": 3.11, "learning_rate": 0.00042262168364265874, "loss": 1.165, "step": 77380 }, { "epoch": 3.11, "learning_rate": 0.00042261161882523453, "loss": 1.1591, "step": 77390 }, { "epoch": 3.11, "learning_rate": 0.0004226015540078103, "loss": 0.8989, "step": 77400 }, { "epoch": 3.11, "learning_rate": 0.0004225914891903861, "loss": 1.1306, "step": 77410 }, { "epoch": 3.11, "learning_rate": 0.0004225814243729619, "loss": 0.9196, "step": 77420 }, { "epoch": 3.11, "learning_rate": 0.00042257135955553764, "loss": 1.2368, "step": 77430 }, { "epoch": 3.11, "learning_rate": 0.00042256129473811343, "loss": 1.0357, "step": 77440 }, { "epoch": 3.11, "learning_rate": 0.0004225512299206892, "loss": 1.0748, "step": 77450 }, { "epoch": 3.12, "learning_rate": 0.000422541165103265, "loss": 1.0537, "step": 77460 }, { "epoch": 3.12, "learning_rate": 0.0004225311002858408, "loss": 0.8898, "step": 77470 }, { "epoch": 3.12, "learning_rate": 0.00042252103546841664, "loss": 1.0676, "step": 77480 }, { "epoch": 3.12, "learning_rate": 0.00042251097065099244, "loss": 0.9944, "step": 77490 }, { "epoch": 3.12, "learning_rate": 0.0004225009058335682, "loss": 1.261, "step": 77500 }, { "epoch": 3.12, "learning_rate": 0.00042249084101614396, "loss": 1.1605, "step": 77510 }, { "epoch": 3.12, "learning_rate": 0.00042248077619871975, "loss": 1.12, "step": 77520 }, { "epoch": 3.12, "learning_rate": 0.00042247071138129554, "loss": 1.201, "step": 77530 }, { "epoch": 3.12, "learning_rate": 0.00042246064656387134, "loss": 1.1427, "step": 77540 }, { "epoch": 3.12, "learning_rate": 0.0004224505817464471, "loss": 1.1063, "step": 77550 }, { "epoch": 3.12, "learning_rate": 0.0004224405169290229, "loss": 1.1226, "step": 77560 }, { "epoch": 3.12, "learning_rate": 0.0004224304521115987, "loss": 0.923, "step": 77570 }, { "epoch": 3.12, "learning_rate": 0.0004224203872941745, "loss": 0.9057, "step": 77580 }, { "epoch": 3.12, "learning_rate": 0.0004224103224767503, "loss": 1.0658, "step": 77590 }, { "epoch": 3.12, "learning_rate": 0.000422400257659326, "loss": 0.9658, "step": 77600 }, { "epoch": 3.12, "learning_rate": 0.0004223901928419018, "loss": 1.0112, "step": 77610 }, { "epoch": 3.12, "learning_rate": 0.00042238012802447766, "loss": 0.8441, "step": 77620 }, { "epoch": 3.12, "learning_rate": 0.00042237006320705345, "loss": 1.4143, "step": 77630 }, { "epoch": 3.12, "learning_rate": 0.00042235999838962924, "loss": 1.1667, "step": 77640 }, { "epoch": 3.12, "learning_rate": 0.00042234993357220503, "loss": 1.0569, "step": 77650 }, { "epoch": 3.12, "learning_rate": 0.0004223398687547808, "loss": 1.0101, "step": 77660 }, { "epoch": 3.12, "learning_rate": 0.0004223298039373566, "loss": 1.1267, "step": 77670 }, { "epoch": 3.12, "learning_rate": 0.00042231973911993235, "loss": 1.208, "step": 77680 }, { "epoch": 3.12, "learning_rate": 0.00042230967430250814, "loss": 1.0208, "step": 77690 }, { "epoch": 3.12, "learning_rate": 0.00042229960948508393, "loss": 1.2161, "step": 77700 }, { "epoch": 3.13, "learning_rate": 0.0004222895446676597, "loss": 1.2359, "step": 77710 }, { "epoch": 3.13, "learning_rate": 0.0004222794798502355, "loss": 1.112, "step": 77720 }, { "epoch": 3.13, "learning_rate": 0.0004222694150328113, "loss": 1.1359, "step": 77730 }, { "epoch": 3.13, "learning_rate": 0.0004222593502153871, "loss": 0.9422, "step": 77740 }, { "epoch": 3.13, "learning_rate": 0.0004222492853979629, "loss": 1.1665, "step": 77750 }, { "epoch": 3.13, "learning_rate": 0.0004222392205805387, "loss": 1.0182, "step": 77760 }, { "epoch": 3.13, "learning_rate": 0.00042222915576311447, "loss": 0.9789, "step": 77770 }, { "epoch": 3.13, "learning_rate": 0.00042221909094569026, "loss": 0.9626, "step": 77780 }, { "epoch": 3.13, "learning_rate": 0.00042220902612826605, "loss": 0.8967, "step": 77790 }, { "epoch": 3.13, "learning_rate": 0.00042219896131084184, "loss": 1.0016, "step": 77800 }, { "epoch": 3.13, "learning_rate": 0.00042218889649341763, "loss": 1.1171, "step": 77810 }, { "epoch": 3.13, "learning_rate": 0.0004221788316759934, "loss": 1.202, "step": 77820 }, { "epoch": 3.13, "learning_rate": 0.0004221687668585692, "loss": 0.9458, "step": 77830 }, { "epoch": 3.13, "learning_rate": 0.000422158702041145, "loss": 1.2488, "step": 77840 }, { "epoch": 3.13, "learning_rate": 0.00042214863722372074, "loss": 0.9813, "step": 77850 }, { "epoch": 3.13, "learning_rate": 0.00042213857240629653, "loss": 0.8966, "step": 77860 }, { "epoch": 3.13, "learning_rate": 0.0004221285075888723, "loss": 1.1441, "step": 77870 }, { "epoch": 3.13, "learning_rate": 0.0004221184427714481, "loss": 1.1961, "step": 77880 }, { "epoch": 3.13, "learning_rate": 0.0004221083779540239, "loss": 1.0642, "step": 77890 }, { "epoch": 3.13, "learning_rate": 0.0004220983131365997, "loss": 0.8916, "step": 77900 }, { "epoch": 3.13, "learning_rate": 0.0004220882483191755, "loss": 1.0044, "step": 77910 }, { "epoch": 3.13, "learning_rate": 0.00042207818350175133, "loss": 1.1788, "step": 77920 }, { "epoch": 3.13, "learning_rate": 0.0004220681186843271, "loss": 1.034, "step": 77930 }, { "epoch": 3.13, "learning_rate": 0.00042205805386690286, "loss": 1.1064, "step": 77940 }, { "epoch": 3.14, "learning_rate": 0.00042204798904947865, "loss": 0.9695, "step": 77950 }, { "epoch": 3.14, "learning_rate": 0.00042203792423205444, "loss": 1.1035, "step": 77960 }, { "epoch": 3.14, "learning_rate": 0.00042202785941463023, "loss": 1.1349, "step": 77970 }, { "epoch": 3.14, "learning_rate": 0.000422017794597206, "loss": 1.1421, "step": 77980 }, { "epoch": 3.14, "learning_rate": 0.0004220077297797818, "loss": 0.9401, "step": 77990 }, { "epoch": 3.14, "learning_rate": 0.0004219976649623576, "loss": 1.1919, "step": 78000 }, { "epoch": 3.14, "learning_rate": 0.0004219876001449334, "loss": 0.8553, "step": 78010 }, { "epoch": 3.14, "learning_rate": 0.00042197753532750913, "loss": 0.922, "step": 78020 }, { "epoch": 3.14, "learning_rate": 0.0004219674705100849, "loss": 1.0934, "step": 78030 }, { "epoch": 3.14, "learning_rate": 0.0004219574056926607, "loss": 1.1114, "step": 78040 }, { "epoch": 3.14, "learning_rate": 0.0004219473408752365, "loss": 0.9302, "step": 78050 }, { "epoch": 3.14, "learning_rate": 0.00042193727605781235, "loss": 1.0274, "step": 78060 }, { "epoch": 3.14, "learning_rate": 0.00042192721124038814, "loss": 0.978, "step": 78070 }, { "epoch": 3.14, "learning_rate": 0.00042191714642296393, "loss": 1.0801, "step": 78080 }, { "epoch": 3.14, "learning_rate": 0.0004219070816055397, "loss": 1.2871, "step": 78090 }, { "epoch": 3.14, "learning_rate": 0.0004218970167881155, "loss": 0.9124, "step": 78100 }, { "epoch": 3.14, "learning_rate": 0.00042188695197069125, "loss": 0.9709, "step": 78110 }, { "epoch": 3.14, "learning_rate": 0.00042187688715326704, "loss": 0.8863, "step": 78120 }, { "epoch": 3.14, "learning_rate": 0.00042186682233584283, "loss": 0.9925, "step": 78130 }, { "epoch": 3.14, "learning_rate": 0.0004218567575184186, "loss": 1.1884, "step": 78140 }, { "epoch": 3.14, "learning_rate": 0.0004218466927009944, "loss": 0.8926, "step": 78150 }, { "epoch": 3.14, "learning_rate": 0.0004218366278835702, "loss": 1.1023, "step": 78160 }, { "epoch": 3.14, "learning_rate": 0.000421826563066146, "loss": 1.1661, "step": 78170 }, { "epoch": 3.14, "learning_rate": 0.0004218164982487218, "loss": 0.9529, "step": 78180 }, { "epoch": 3.14, "learning_rate": 0.0004218064334312975, "loss": 1.194, "step": 78190 }, { "epoch": 3.15, "learning_rate": 0.0004217963686138733, "loss": 1.3797, "step": 78200 }, { "epoch": 3.15, "learning_rate": 0.00042178630379644915, "loss": 0.8952, "step": 78210 }, { "epoch": 3.15, "learning_rate": 0.00042177623897902495, "loss": 0.9295, "step": 78220 }, { "epoch": 3.15, "learning_rate": 0.00042176617416160074, "loss": 1.0041, "step": 78230 }, { "epoch": 3.15, "learning_rate": 0.00042175610934417653, "loss": 0.9105, "step": 78240 }, { "epoch": 3.15, "learning_rate": 0.0004217460445267523, "loss": 1.1591, "step": 78250 }, { "epoch": 3.15, "learning_rate": 0.0004217359797093281, "loss": 1.2337, "step": 78260 }, { "epoch": 3.15, "learning_rate": 0.00042172591489190385, "loss": 0.9653, "step": 78270 }, { "epoch": 3.15, "learning_rate": 0.00042171585007447964, "loss": 1.0123, "step": 78280 }, { "epoch": 3.15, "learning_rate": 0.0004217057852570554, "loss": 1.0209, "step": 78290 }, { "epoch": 3.15, "learning_rate": 0.0004216957204396312, "loss": 1.0381, "step": 78300 }, { "epoch": 3.15, "learning_rate": 0.000421685655622207, "loss": 1.046, "step": 78310 }, { "epoch": 3.15, "learning_rate": 0.0004216755908047828, "loss": 1.1271, "step": 78320 }, { "epoch": 3.15, "learning_rate": 0.0004216655259873586, "loss": 1.0674, "step": 78330 }, { "epoch": 3.15, "learning_rate": 0.0004216554611699344, "loss": 0.8919, "step": 78340 }, { "epoch": 3.15, "learning_rate": 0.00042164539635251017, "loss": 0.8352, "step": 78350 }, { "epoch": 3.15, "learning_rate": 0.00042163533153508596, "loss": 1.1655, "step": 78360 }, { "epoch": 3.15, "learning_rate": 0.00042162526671766175, "loss": 1.1134, "step": 78370 }, { "epoch": 3.15, "learning_rate": 0.00042161520190023754, "loss": 0.9367, "step": 78380 }, { "epoch": 3.15, "learning_rate": 0.00042160513708281333, "loss": 1.0855, "step": 78390 }, { "epoch": 3.15, "learning_rate": 0.0004215950722653891, "loss": 0.965, "step": 78400 }, { "epoch": 3.15, "learning_rate": 0.0004215850074479649, "loss": 0.9753, "step": 78410 }, { "epoch": 3.15, "learning_rate": 0.0004215749426305407, "loss": 1.0646, "step": 78420 }, { "epoch": 3.15, "learning_rate": 0.0004215648778131165, "loss": 0.9852, "step": 78430 }, { "epoch": 3.15, "learning_rate": 0.00042155481299569223, "loss": 1.139, "step": 78440 }, { "epoch": 3.16, "learning_rate": 0.000421544748178268, "loss": 1.1623, "step": 78450 }, { "epoch": 3.16, "learning_rate": 0.0004215346833608438, "loss": 1.2701, "step": 78460 }, { "epoch": 3.16, "learning_rate": 0.0004215246185434196, "loss": 1.1196, "step": 78470 }, { "epoch": 3.16, "learning_rate": 0.0004215145537259954, "loss": 1.0904, "step": 78480 }, { "epoch": 3.16, "learning_rate": 0.0004215044889085712, "loss": 1.2099, "step": 78490 }, { "epoch": 3.16, "learning_rate": 0.00042149442409114703, "loss": 1.1007, "step": 78500 }, { "epoch": 3.16, "learning_rate": 0.0004214843592737228, "loss": 1.1353, "step": 78510 }, { "epoch": 3.16, "learning_rate": 0.0004214742944562986, "loss": 1.2653, "step": 78520 }, { "epoch": 3.16, "learning_rate": 0.00042146422963887435, "loss": 0.8733, "step": 78530 }, { "epoch": 3.16, "learning_rate": 0.00042145416482145014, "loss": 1.0786, "step": 78540 }, { "epoch": 3.16, "learning_rate": 0.00042144410000402593, "loss": 1.1602, "step": 78550 }, { "epoch": 3.16, "learning_rate": 0.0004214340351866017, "loss": 0.9714, "step": 78560 }, { "epoch": 3.16, "learning_rate": 0.0004214239703691775, "loss": 1.0282, "step": 78570 }, { "epoch": 3.16, "learning_rate": 0.0004214139055517533, "loss": 1.0285, "step": 78580 }, { "epoch": 3.16, "learning_rate": 0.0004214038407343291, "loss": 0.828, "step": 78590 }, { "epoch": 3.16, "learning_rate": 0.0004213937759169049, "loss": 1.0838, "step": 78600 }, { "epoch": 3.16, "learning_rate": 0.0004213837110994806, "loss": 1.0498, "step": 78610 }, { "epoch": 3.16, "learning_rate": 0.0004213736462820564, "loss": 1.1663, "step": 78620 }, { "epoch": 3.16, "learning_rate": 0.0004213635814646322, "loss": 0.9955, "step": 78630 }, { "epoch": 3.16, "learning_rate": 0.000421353516647208, "loss": 0.9179, "step": 78640 }, { "epoch": 3.16, "learning_rate": 0.00042134345182978384, "loss": 1.126, "step": 78650 }, { "epoch": 3.16, "learning_rate": 0.00042133338701235963, "loss": 1.1552, "step": 78660 }, { "epoch": 3.16, "learning_rate": 0.0004213233221949354, "loss": 1.0604, "step": 78670 }, { "epoch": 3.16, "learning_rate": 0.0004213132573775112, "loss": 1.0943, "step": 78680 }, { "epoch": 3.16, "learning_rate": 0.000421303192560087, "loss": 1.219, "step": 78690 }, { "epoch": 3.17, "learning_rate": 0.00042129312774266274, "loss": 0.7862, "step": 78700 }, { "epoch": 3.17, "learning_rate": 0.00042128306292523853, "loss": 1.0274, "step": 78710 }, { "epoch": 3.17, "learning_rate": 0.0004212729981078143, "loss": 1.1597, "step": 78720 }, { "epoch": 3.17, "learning_rate": 0.0004212629332903901, "loss": 1.1559, "step": 78730 }, { "epoch": 3.17, "learning_rate": 0.0004212528684729659, "loss": 1.0477, "step": 78740 }, { "epoch": 3.17, "learning_rate": 0.0004212428036555417, "loss": 1.0052, "step": 78750 }, { "epoch": 3.17, "learning_rate": 0.0004212327388381175, "loss": 1.1609, "step": 78760 }, { "epoch": 3.17, "learning_rate": 0.0004212226740206933, "loss": 0.976, "step": 78770 }, { "epoch": 3.17, "learning_rate": 0.000421212609203269, "loss": 1.0426, "step": 78780 }, { "epoch": 3.17, "learning_rate": 0.00042120254438584486, "loss": 1.004, "step": 78790 }, { "epoch": 3.17, "learning_rate": 0.00042119247956842065, "loss": 1.0876, "step": 78800 }, { "epoch": 3.17, "learning_rate": 0.00042118241475099644, "loss": 1.1355, "step": 78810 }, { "epoch": 3.17, "learning_rate": 0.00042117234993357223, "loss": 1.0525, "step": 78820 }, { "epoch": 3.17, "learning_rate": 0.000421162285116148, "loss": 1.2283, "step": 78830 }, { "epoch": 3.17, "learning_rate": 0.0004211522202987238, "loss": 1.0934, "step": 78840 }, { "epoch": 3.17, "learning_rate": 0.0004211421554812996, "loss": 0.9805, "step": 78850 }, { "epoch": 3.17, "learning_rate": 0.0004211320906638754, "loss": 1.0576, "step": 78860 }, { "epoch": 3.17, "learning_rate": 0.00042112202584645113, "loss": 0.8999, "step": 78870 }, { "epoch": 3.17, "learning_rate": 0.0004211119610290269, "loss": 0.79, "step": 78880 }, { "epoch": 3.17, "learning_rate": 0.0004211018962116027, "loss": 1.0715, "step": 78890 }, { "epoch": 3.17, "learning_rate": 0.0004210918313941785, "loss": 1.0557, "step": 78900 }, { "epoch": 3.17, "learning_rate": 0.0004210817665767543, "loss": 1.0837, "step": 78910 }, { "epoch": 3.17, "learning_rate": 0.0004210717017593301, "loss": 1.0154, "step": 78920 }, { "epoch": 3.17, "learning_rate": 0.0004210616369419059, "loss": 1.176, "step": 78930 }, { "epoch": 3.17, "learning_rate": 0.0004210515721244817, "loss": 1.112, "step": 78940 }, { "epoch": 3.18, "learning_rate": 0.00042104150730705746, "loss": 0.9279, "step": 78950 }, { "epoch": 3.18, "learning_rate": 0.00042103144248963325, "loss": 1.2389, "step": 78960 }, { "epoch": 3.18, "learning_rate": 0.00042102137767220904, "loss": 1.0748, "step": 78970 }, { "epoch": 3.18, "learning_rate": 0.00042101131285478483, "loss": 1.1279, "step": 78980 }, { "epoch": 3.18, "learning_rate": 0.0004210012480373606, "loss": 0.8563, "step": 78990 }, { "epoch": 3.18, "learning_rate": 0.0004209911832199364, "loss": 0.9488, "step": 79000 }, { "epoch": 3.18, "learning_rate": 0.0004209811184025122, "loss": 1.0677, "step": 79010 }, { "epoch": 3.18, "learning_rate": 0.000420971053585088, "loss": 1.0811, "step": 79020 }, { "epoch": 3.18, "learning_rate": 0.00042096098876766373, "loss": 1.1045, "step": 79030 }, { "epoch": 3.18, "learning_rate": 0.0004209509239502395, "loss": 1.0478, "step": 79040 }, { "epoch": 3.18, "learning_rate": 0.0004209408591328153, "loss": 1.046, "step": 79050 }, { "epoch": 3.18, "learning_rate": 0.0004209307943153911, "loss": 1.0023, "step": 79060 }, { "epoch": 3.18, "learning_rate": 0.0004209207294979669, "loss": 0.8823, "step": 79070 }, { "epoch": 3.18, "learning_rate": 0.0004209106646805427, "loss": 1.2669, "step": 79080 }, { "epoch": 3.18, "learning_rate": 0.0004209005998631185, "loss": 1.0712, "step": 79090 }, { "epoch": 3.18, "learning_rate": 0.0004208905350456943, "loss": 0.9993, "step": 79100 }, { "epoch": 3.18, "learning_rate": 0.0004208804702282701, "loss": 1.1307, "step": 79110 }, { "epoch": 3.18, "learning_rate": 0.00042087040541084585, "loss": 1.0781, "step": 79120 }, { "epoch": 3.18, "learning_rate": 0.00042086034059342164, "loss": 0.9004, "step": 79130 }, { "epoch": 3.18, "learning_rate": 0.0004208502757759974, "loss": 1.0236, "step": 79140 }, { "epoch": 3.18, "learning_rate": 0.0004208402109585732, "loss": 1.0622, "step": 79150 }, { "epoch": 3.18, "learning_rate": 0.000420830146141149, "loss": 1.0716, "step": 79160 }, { "epoch": 3.18, "learning_rate": 0.0004208200813237248, "loss": 0.9439, "step": 79170 }, { "epoch": 3.18, "learning_rate": 0.0004208100165063006, "loss": 0.9491, "step": 79180 }, { "epoch": 3.18, "learning_rate": 0.0004207999516888764, "loss": 1.2303, "step": 79190 }, { "epoch": 3.19, "learning_rate": 0.0004207898868714521, "loss": 1.1249, "step": 79200 }, { "epoch": 3.19, "learning_rate": 0.0004207798220540279, "loss": 1.0217, "step": 79210 }, { "epoch": 3.19, "learning_rate": 0.0004207697572366037, "loss": 1.2518, "step": 79220 }, { "epoch": 3.19, "learning_rate": 0.00042075969241917954, "loss": 0.9836, "step": 79230 }, { "epoch": 3.19, "learning_rate": 0.00042074962760175533, "loss": 0.9095, "step": 79240 }, { "epoch": 3.19, "learning_rate": 0.0004207395627843311, "loss": 0.9559, "step": 79250 }, { "epoch": 3.19, "learning_rate": 0.0004207294979669069, "loss": 0.9436, "step": 79260 }, { "epoch": 3.19, "learning_rate": 0.0004207194331494827, "loss": 1.1047, "step": 79270 }, { "epoch": 3.19, "learning_rate": 0.0004207093683320585, "loss": 1.185, "step": 79280 }, { "epoch": 3.19, "learning_rate": 0.00042069930351463423, "loss": 1.0915, "step": 79290 }, { "epoch": 3.19, "learning_rate": 0.00042068923869721, "loss": 1.1621, "step": 79300 }, { "epoch": 3.19, "learning_rate": 0.0004206791738797858, "loss": 1.1442, "step": 79310 }, { "epoch": 3.19, "learning_rate": 0.0004206691090623616, "loss": 0.9757, "step": 79320 }, { "epoch": 3.19, "learning_rate": 0.0004206590442449374, "loss": 1.0952, "step": 79330 }, { "epoch": 3.19, "learning_rate": 0.0004206489794275132, "loss": 1.0336, "step": 79340 }, { "epoch": 3.19, "learning_rate": 0.000420638914610089, "loss": 1.1858, "step": 79350 }, { "epoch": 3.19, "learning_rate": 0.00042062884979266477, "loss": 0.8765, "step": 79360 }, { "epoch": 3.19, "learning_rate": 0.0004206187849752405, "loss": 1.0829, "step": 79370 }, { "epoch": 3.19, "learning_rate": 0.00042060872015781635, "loss": 1.1603, "step": 79380 }, { "epoch": 3.19, "learning_rate": 0.00042059865534039214, "loss": 1.11, "step": 79390 }, { "epoch": 3.19, "learning_rate": 0.00042058859052296793, "loss": 1.0916, "step": 79400 }, { "epoch": 3.19, "learning_rate": 0.0004205785257055437, "loss": 1.0506, "step": 79410 }, { "epoch": 3.19, "learning_rate": 0.0004205684608881195, "loss": 1.0167, "step": 79420 }, { "epoch": 3.19, "learning_rate": 0.0004205583960706953, "loss": 1.0678, "step": 79430 }, { "epoch": 3.19, "learning_rate": 0.0004205483312532711, "loss": 0.9805, "step": 79440 }, { "epoch": 3.2, "learning_rate": 0.0004205382664358469, "loss": 1.1686, "step": 79450 }, { "epoch": 3.2, "learning_rate": 0.0004205282016184226, "loss": 1.0096, "step": 79460 }, { "epoch": 3.2, "learning_rate": 0.0004205181368009984, "loss": 1.1097, "step": 79470 }, { "epoch": 3.2, "learning_rate": 0.0004205080719835742, "loss": 0.8692, "step": 79480 }, { "epoch": 3.2, "learning_rate": 0.00042049800716615, "loss": 0.9428, "step": 79490 }, { "epoch": 3.2, "learning_rate": 0.0004204879423487258, "loss": 1.0456, "step": 79500 }, { "epoch": 3.2, "learning_rate": 0.0004204778775313016, "loss": 1.1532, "step": 79510 }, { "epoch": 3.2, "learning_rate": 0.0004204678127138774, "loss": 0.923, "step": 79520 }, { "epoch": 3.2, "learning_rate": 0.0004204577478964532, "loss": 0.7911, "step": 79530 }, { "epoch": 3.2, "learning_rate": 0.00042044768307902895, "loss": 0.9979, "step": 79540 }, { "epoch": 3.2, "learning_rate": 0.00042043761826160474, "loss": 1.0097, "step": 79550 }, { "epoch": 3.2, "learning_rate": 0.00042042755344418053, "loss": 0.9715, "step": 79560 }, { "epoch": 3.2, "learning_rate": 0.0004204174886267563, "loss": 1.0585, "step": 79570 }, { "epoch": 3.2, "learning_rate": 0.0004204074238093321, "loss": 0.9226, "step": 79580 }, { "epoch": 3.2, "learning_rate": 0.0004203973589919079, "loss": 1.2905, "step": 79590 }, { "epoch": 3.2, "learning_rate": 0.0004203872941744837, "loss": 1.1381, "step": 79600 }, { "epoch": 3.2, "learning_rate": 0.0004203772293570595, "loss": 1.018, "step": 79610 }, { "epoch": 3.2, "learning_rate": 0.0004203671645396353, "loss": 1.0845, "step": 79620 }, { "epoch": 3.2, "learning_rate": 0.000420357099722211, "loss": 0.9861, "step": 79630 }, { "epoch": 3.2, "learning_rate": 0.0004203470349047868, "loss": 1.3228, "step": 79640 }, { "epoch": 3.2, "learning_rate": 0.0004203369700873626, "loss": 1.2366, "step": 79650 }, { "epoch": 3.2, "learning_rate": 0.0004203269052699384, "loss": 1.1381, "step": 79660 }, { "epoch": 3.2, "learning_rate": 0.00042031684045251423, "loss": 1.1311, "step": 79670 }, { "epoch": 3.2, "learning_rate": 0.00042030677563509, "loss": 1.2134, "step": 79680 }, { "epoch": 3.21, "learning_rate": 0.0004202967108176658, "loss": 1.3666, "step": 79690 }, { "epoch": 3.21, "learning_rate": 0.0004202866460002416, "loss": 0.9949, "step": 79700 }, { "epoch": 3.21, "learning_rate": 0.00042027658118281734, "loss": 1.0409, "step": 79710 }, { "epoch": 3.21, "learning_rate": 0.00042026651636539313, "loss": 1.076, "step": 79720 }, { "epoch": 3.21, "learning_rate": 0.0004202564515479689, "loss": 1.1224, "step": 79730 }, { "epoch": 3.21, "learning_rate": 0.0004202463867305447, "loss": 1.1339, "step": 79740 }, { "epoch": 3.21, "learning_rate": 0.0004202363219131205, "loss": 1.0421, "step": 79750 }, { "epoch": 3.21, "learning_rate": 0.0004202262570956963, "loss": 1.1177, "step": 79760 }, { "epoch": 3.21, "learning_rate": 0.0004202161922782721, "loss": 1.2026, "step": 79770 }, { "epoch": 3.21, "learning_rate": 0.0004202061274608479, "loss": 1.0825, "step": 79780 }, { "epoch": 3.21, "learning_rate": 0.0004201960626434236, "loss": 1.077, "step": 79790 }, { "epoch": 3.21, "learning_rate": 0.0004201859978259994, "loss": 1.2022, "step": 79800 }, { "epoch": 3.21, "learning_rate": 0.0004201759330085752, "loss": 1.4696, "step": 79810 }, { "epoch": 3.21, "learning_rate": 0.00042016586819115104, "loss": 1.0981, "step": 79820 }, { "epoch": 3.21, "learning_rate": 0.00042015580337372683, "loss": 1.3206, "step": 79830 }, { "epoch": 3.21, "learning_rate": 0.0004201457385563026, "loss": 1.1446, "step": 79840 }, { "epoch": 3.21, "learning_rate": 0.0004201356737388784, "loss": 0.9225, "step": 79850 }, { "epoch": 3.21, "learning_rate": 0.0004201256089214542, "loss": 1.0768, "step": 79860 }, { "epoch": 3.21, "learning_rate": 0.00042011554410403, "loss": 1.154, "step": 79870 }, { "epoch": 3.21, "learning_rate": 0.00042010547928660573, "loss": 1.01, "step": 79880 }, { "epoch": 3.21, "learning_rate": 0.0004200954144691815, "loss": 1.1422, "step": 79890 }, { "epoch": 3.21, "learning_rate": 0.0004200853496517573, "loss": 1.3073, "step": 79900 }, { "epoch": 3.21, "learning_rate": 0.0004200752848343331, "loss": 0.8028, "step": 79910 }, { "epoch": 3.21, "learning_rate": 0.0004200652200169089, "loss": 0.9587, "step": 79920 }, { "epoch": 3.21, "learning_rate": 0.0004200551551994847, "loss": 1.2632, "step": 79930 }, { "epoch": 3.22, "learning_rate": 0.00042004509038206047, "loss": 1.1001, "step": 79940 }, { "epoch": 3.22, "learning_rate": 0.00042003502556463626, "loss": 1.1175, "step": 79950 }, { "epoch": 3.22, "learning_rate": 0.0004200249607472121, "loss": 1.0298, "step": 79960 }, { "epoch": 3.22, "learning_rate": 0.00042001489592978784, "loss": 1.1977, "step": 79970 }, { "epoch": 3.22, "learning_rate": 0.00042000483111236364, "loss": 1.0849, "step": 79980 }, { "epoch": 3.22, "learning_rate": 0.0004199947662949394, "loss": 1.0122, "step": 79990 }, { "epoch": 3.22, "learning_rate": 0.0004199847014775152, "loss": 1.032, "step": 80000 }, { "epoch": 3.22, "learning_rate": 0.000419974636660091, "loss": 0.8908, "step": 80010 }, { "epoch": 3.22, "learning_rate": 0.0004199645718426668, "loss": 0.9661, "step": 80020 }, { "epoch": 3.22, "learning_rate": 0.0004199545070252426, "loss": 1.2543, "step": 80030 }, { "epoch": 3.22, "learning_rate": 0.0004199444422078184, "loss": 1.1301, "step": 80040 }, { "epoch": 3.22, "learning_rate": 0.0004199343773903941, "loss": 1.1078, "step": 80050 }, { "epoch": 3.22, "learning_rate": 0.0004199243125729699, "loss": 1.0406, "step": 80060 }, { "epoch": 3.22, "learning_rate": 0.0004199142477555457, "loss": 1.1265, "step": 80070 }, { "epoch": 3.22, "learning_rate": 0.0004199041829381215, "loss": 1.1891, "step": 80080 }, { "epoch": 3.22, "learning_rate": 0.0004198941181206973, "loss": 0.9702, "step": 80090 }, { "epoch": 3.22, "learning_rate": 0.00041988405330327307, "loss": 1.2809, "step": 80100 }, { "epoch": 3.22, "learning_rate": 0.0004198739884858489, "loss": 1.1125, "step": 80110 }, { "epoch": 3.22, "learning_rate": 0.0004198639236684247, "loss": 1.1744, "step": 80120 }, { "epoch": 3.22, "learning_rate": 0.00041985385885100044, "loss": 1.0771, "step": 80130 }, { "epoch": 3.22, "learning_rate": 0.00041984379403357623, "loss": 1.0518, "step": 80140 }, { "epoch": 3.22, "learning_rate": 0.000419833729216152, "loss": 1.2273, "step": 80150 }, { "epoch": 3.22, "learning_rate": 0.0004198236643987278, "loss": 1.2448, "step": 80160 }, { "epoch": 3.22, "learning_rate": 0.0004198135995813036, "loss": 1.254, "step": 80170 }, { "epoch": 3.22, "learning_rate": 0.0004198035347638794, "loss": 1.2779, "step": 80180 }, { "epoch": 3.23, "learning_rate": 0.0004197934699464552, "loss": 0.9612, "step": 80190 }, { "epoch": 3.23, "learning_rate": 0.000419783405129031, "loss": 0.9015, "step": 80200 }, { "epoch": 3.23, "learning_rate": 0.00041977334031160677, "loss": 0.8428, "step": 80210 }, { "epoch": 3.23, "learning_rate": 0.0004197632754941825, "loss": 1.0333, "step": 80220 }, { "epoch": 3.23, "learning_rate": 0.0004197532106767583, "loss": 1.0149, "step": 80230 }, { "epoch": 3.23, "learning_rate": 0.0004197431458593341, "loss": 1.3656, "step": 80240 }, { "epoch": 3.23, "learning_rate": 0.0004197330810419099, "loss": 1.1989, "step": 80250 }, { "epoch": 3.23, "learning_rate": 0.0004197230162244857, "loss": 1.1005, "step": 80260 }, { "epoch": 3.23, "learning_rate": 0.0004197129514070615, "loss": 0.8264, "step": 80270 }, { "epoch": 3.23, "learning_rate": 0.0004197028865896373, "loss": 1.0456, "step": 80280 }, { "epoch": 3.23, "learning_rate": 0.0004196928217722131, "loss": 1.23, "step": 80290 }, { "epoch": 3.23, "learning_rate": 0.00041968275695478883, "loss": 1.1468, "step": 80300 }, { "epoch": 3.23, "learning_rate": 0.0004196726921373646, "loss": 1.1925, "step": 80310 }, { "epoch": 3.23, "learning_rate": 0.0004196626273199404, "loss": 1.078, "step": 80320 }, { "epoch": 3.23, "learning_rate": 0.0004196525625025162, "loss": 0.9951, "step": 80330 }, { "epoch": 3.23, "learning_rate": 0.000419642497685092, "loss": 1.0706, "step": 80340 }, { "epoch": 3.23, "learning_rate": 0.0004196324328676678, "loss": 1.0143, "step": 80350 }, { "epoch": 3.23, "learning_rate": 0.0004196223680502436, "loss": 1.1559, "step": 80360 }, { "epoch": 3.23, "learning_rate": 0.00041961230323281937, "loss": 1.0305, "step": 80370 }, { "epoch": 3.23, "learning_rate": 0.00041960223841539516, "loss": 0.9983, "step": 80380 }, { "epoch": 3.23, "learning_rate": 0.0004195921735979709, "loss": 0.9779, "step": 80390 }, { "epoch": 3.23, "learning_rate": 0.00041958210878054674, "loss": 1.3674, "step": 80400 }, { "epoch": 3.23, "learning_rate": 0.00041957204396312253, "loss": 0.9684, "step": 80410 }, { "epoch": 3.23, "learning_rate": 0.0004195619791456983, "loss": 1.2394, "step": 80420 }, { "epoch": 3.23, "learning_rate": 0.0004195519143282741, "loss": 1.1156, "step": 80430 }, { "epoch": 3.24, "learning_rate": 0.0004195418495108499, "loss": 0.93, "step": 80440 }, { "epoch": 3.24, "learning_rate": 0.0004195317846934257, "loss": 1.1062, "step": 80450 }, { "epoch": 3.24, "learning_rate": 0.0004195217198760015, "loss": 1.0048, "step": 80460 }, { "epoch": 3.24, "learning_rate": 0.0004195116550585772, "loss": 1.0258, "step": 80470 }, { "epoch": 3.24, "learning_rate": 0.000419501590241153, "loss": 1.0324, "step": 80480 }, { "epoch": 3.24, "learning_rate": 0.0004194915254237288, "loss": 1.0859, "step": 80490 }, { "epoch": 3.24, "learning_rate": 0.0004194814606063046, "loss": 1.0424, "step": 80500 }, { "epoch": 3.24, "learning_rate": 0.0004194713957888804, "loss": 1.0439, "step": 80510 }, { "epoch": 3.24, "learning_rate": 0.0004194613309714562, "loss": 1.227, "step": 80520 }, { "epoch": 3.24, "learning_rate": 0.00041945126615403197, "loss": 1.0607, "step": 80530 }, { "epoch": 3.24, "learning_rate": 0.00041944120133660776, "loss": 0.7862, "step": 80540 }, { "epoch": 3.24, "learning_rate": 0.0004194311365191836, "loss": 1.1195, "step": 80550 }, { "epoch": 3.24, "learning_rate": 0.00041942107170175934, "loss": 1.1257, "step": 80560 }, { "epoch": 3.24, "learning_rate": 0.00041941100688433513, "loss": 1.043, "step": 80570 }, { "epoch": 3.24, "learning_rate": 0.0004194009420669109, "loss": 0.9891, "step": 80580 }, { "epoch": 3.24, "learning_rate": 0.0004193908772494867, "loss": 1.0519, "step": 80590 }, { "epoch": 3.24, "learning_rate": 0.0004193808124320625, "loss": 1.1817, "step": 80600 }, { "epoch": 3.24, "learning_rate": 0.0004193707476146383, "loss": 0.9721, "step": 80610 }, { "epoch": 3.24, "learning_rate": 0.0004193606827972141, "loss": 1.1494, "step": 80620 }, { "epoch": 3.24, "learning_rate": 0.0004193506179797899, "loss": 1.0363, "step": 80630 }, { "epoch": 3.24, "learning_rate": 0.0004193405531623656, "loss": 0.9817, "step": 80640 }, { "epoch": 3.24, "learning_rate": 0.0004193304883449414, "loss": 1.032, "step": 80650 }, { "epoch": 3.24, "learning_rate": 0.0004193204235275172, "loss": 1.2597, "step": 80660 }, { "epoch": 3.24, "learning_rate": 0.000419310358710093, "loss": 1.0554, "step": 80670 }, { "epoch": 3.24, "learning_rate": 0.0004193002938926688, "loss": 1.3587, "step": 80680 }, { "epoch": 3.25, "learning_rate": 0.0004192902290752446, "loss": 1.1971, "step": 80690 }, { "epoch": 3.25, "learning_rate": 0.0004192801642578204, "loss": 1.1032, "step": 80700 }, { "epoch": 3.25, "learning_rate": 0.0004192700994403962, "loss": 1.2457, "step": 80710 }, { "epoch": 3.25, "learning_rate": 0.000419260034622972, "loss": 1.1596, "step": 80720 }, { "epoch": 3.25, "learning_rate": 0.00041924996980554773, "loss": 0.8782, "step": 80730 }, { "epoch": 3.25, "learning_rate": 0.0004192399049881235, "loss": 1.0105, "step": 80740 }, { "epoch": 3.25, "learning_rate": 0.0004192298401706993, "loss": 0.9618, "step": 80750 }, { "epoch": 3.25, "learning_rate": 0.0004192197753532751, "loss": 0.9831, "step": 80760 }, { "epoch": 3.25, "learning_rate": 0.0004192097105358509, "loss": 1.0094, "step": 80770 }, { "epoch": 3.25, "learning_rate": 0.0004191996457184267, "loss": 1.0288, "step": 80780 }, { "epoch": 3.25, "learning_rate": 0.00041918958090100247, "loss": 1.1575, "step": 80790 }, { "epoch": 3.25, "learning_rate": 0.00041917951608357826, "loss": 1.126, "step": 80800 }, { "epoch": 3.25, "learning_rate": 0.000419169451266154, "loss": 1.0389, "step": 80810 }, { "epoch": 3.25, "learning_rate": 0.0004191593864487298, "loss": 1.0497, "step": 80820 }, { "epoch": 3.25, "learning_rate": 0.0004191493216313056, "loss": 1.1654, "step": 80830 }, { "epoch": 3.25, "learning_rate": 0.0004191392568138814, "loss": 1.3716, "step": 80840 }, { "epoch": 3.25, "learning_rate": 0.0004191291919964572, "loss": 1.1308, "step": 80850 }, { "epoch": 3.25, "learning_rate": 0.000419119127179033, "loss": 1.2996, "step": 80860 }, { "epoch": 3.25, "learning_rate": 0.0004191090623616088, "loss": 1.0008, "step": 80870 }, { "epoch": 3.25, "learning_rate": 0.0004190989975441846, "loss": 1.2154, "step": 80880 }, { "epoch": 3.25, "learning_rate": 0.0004190889327267603, "loss": 1.1596, "step": 80890 }, { "epoch": 3.25, "learning_rate": 0.0004190788679093361, "loss": 1.0623, "step": 80900 }, { "epoch": 3.25, "learning_rate": 0.0004190688030919119, "loss": 1.0529, "step": 80910 }, { "epoch": 3.25, "learning_rate": 0.0004190587382744877, "loss": 0.8232, "step": 80920 }, { "epoch": 3.25, "learning_rate": 0.0004190486734570635, "loss": 1.2394, "step": 80930 }, { "epoch": 3.26, "learning_rate": 0.0004190386086396393, "loss": 1.1873, "step": 80940 }, { "epoch": 3.26, "learning_rate": 0.00041902854382221507, "loss": 1.1381, "step": 80950 }, { "epoch": 3.26, "learning_rate": 0.00041901847900479086, "loss": 1.0776, "step": 80960 }, { "epoch": 3.26, "learning_rate": 0.00041900841418736665, "loss": 1.212, "step": 80970 }, { "epoch": 3.26, "learning_rate": 0.0004189983493699424, "loss": 1.0875, "step": 80980 }, { "epoch": 3.26, "learning_rate": 0.00041898828455251823, "loss": 0.9263, "step": 80990 }, { "epoch": 3.26, "learning_rate": 0.000418978219735094, "loss": 0.8537, "step": 81000 }, { "epoch": 3.26, "learning_rate": 0.0004189681549176698, "loss": 0.9957, "step": 81010 }, { "epoch": 3.26, "learning_rate": 0.0004189580901002456, "loss": 1.0732, "step": 81020 }, { "epoch": 3.26, "learning_rate": 0.0004189480252828214, "loss": 1.1176, "step": 81030 }, { "epoch": 3.26, "learning_rate": 0.0004189379604653972, "loss": 1.0477, "step": 81040 }, { "epoch": 3.26, "learning_rate": 0.000418927895647973, "loss": 1.0882, "step": 81050 }, { "epoch": 3.26, "learning_rate": 0.0004189178308305487, "loss": 1.1858, "step": 81060 }, { "epoch": 3.26, "learning_rate": 0.0004189077660131245, "loss": 0.8784, "step": 81070 }, { "epoch": 3.26, "learning_rate": 0.0004188977011957003, "loss": 1.2782, "step": 81080 }, { "epoch": 3.26, "learning_rate": 0.0004188876363782761, "loss": 1.0516, "step": 81090 }, { "epoch": 3.26, "learning_rate": 0.0004188775715608519, "loss": 1.0459, "step": 81100 }, { "epoch": 3.26, "learning_rate": 0.00041886750674342767, "loss": 1.3093, "step": 81110 }, { "epoch": 3.26, "learning_rate": 0.00041885744192600346, "loss": 1.0982, "step": 81120 }, { "epoch": 3.26, "learning_rate": 0.0004188473771085793, "loss": 1.1238, "step": 81130 }, { "epoch": 3.26, "learning_rate": 0.0004188373122911551, "loss": 1.1498, "step": 81140 }, { "epoch": 3.26, "learning_rate": 0.00041882724747373083, "loss": 1.322, "step": 81150 }, { "epoch": 3.26, "learning_rate": 0.0004188171826563066, "loss": 1.3344, "step": 81160 }, { "epoch": 3.26, "learning_rate": 0.0004188071178388824, "loss": 1.0309, "step": 81170 }, { "epoch": 3.26, "learning_rate": 0.0004187970530214582, "loss": 1.2405, "step": 81180 }, { "epoch": 3.27, "learning_rate": 0.000418786988204034, "loss": 1.1874, "step": 81190 }, { "epoch": 3.27, "learning_rate": 0.0004187769233866098, "loss": 1.0495, "step": 81200 }, { "epoch": 3.27, "learning_rate": 0.0004187668585691856, "loss": 1.1281, "step": 81210 }, { "epoch": 3.27, "learning_rate": 0.00041875679375176137, "loss": 0.9701, "step": 81220 }, { "epoch": 3.27, "learning_rate": 0.0004187467289343371, "loss": 0.9886, "step": 81230 }, { "epoch": 3.27, "learning_rate": 0.0004187366641169129, "loss": 1.0931, "step": 81240 }, { "epoch": 3.27, "learning_rate": 0.0004187265992994887, "loss": 1.1873, "step": 81250 }, { "epoch": 3.27, "learning_rate": 0.0004187165344820645, "loss": 1.0275, "step": 81260 }, { "epoch": 3.27, "learning_rate": 0.00041870646966464027, "loss": 1.2524, "step": 81270 }, { "epoch": 3.27, "learning_rate": 0.0004186964048472161, "loss": 1.1905, "step": 81280 }, { "epoch": 3.27, "learning_rate": 0.0004186863400297919, "loss": 1.3395, "step": 81290 }, { "epoch": 3.27, "learning_rate": 0.0004186762752123677, "loss": 0.9558, "step": 81300 }, { "epoch": 3.27, "learning_rate": 0.0004186662103949435, "loss": 1.2478, "step": 81310 }, { "epoch": 3.27, "learning_rate": 0.0004186561455775192, "loss": 1.2391, "step": 81320 }, { "epoch": 3.27, "learning_rate": 0.000418646080760095, "loss": 0.9441, "step": 81330 }, { "epoch": 3.27, "learning_rate": 0.0004186360159426708, "loss": 1.3604, "step": 81340 }, { "epoch": 3.27, "learning_rate": 0.0004186259511252466, "loss": 1.0805, "step": 81350 }, { "epoch": 3.27, "learning_rate": 0.0004186158863078224, "loss": 1.0262, "step": 81360 }, { "epoch": 3.27, "learning_rate": 0.0004186058214903982, "loss": 1.1491, "step": 81370 }, { "epoch": 3.27, "learning_rate": 0.00041859575667297397, "loss": 1.079, "step": 81380 }, { "epoch": 3.27, "learning_rate": 0.00041858569185554976, "loss": 1.0114, "step": 81390 }, { "epoch": 3.27, "learning_rate": 0.0004185756270381255, "loss": 1.1415, "step": 81400 }, { "epoch": 3.27, "learning_rate": 0.0004185655622207013, "loss": 0.964, "step": 81410 }, { "epoch": 3.27, "learning_rate": 0.0004185554974032771, "loss": 1.0978, "step": 81420 }, { "epoch": 3.27, "learning_rate": 0.0004185454325858529, "loss": 0.9259, "step": 81430 }, { "epoch": 3.28, "learning_rate": 0.0004185353677684287, "loss": 0.9443, "step": 81440 }, { "epoch": 3.28, "learning_rate": 0.0004185253029510045, "loss": 1.0097, "step": 81450 }, { "epoch": 3.28, "learning_rate": 0.0004185152381335803, "loss": 1.1459, "step": 81460 }, { "epoch": 3.28, "learning_rate": 0.0004185051733161561, "loss": 1.2025, "step": 81470 }, { "epoch": 3.28, "learning_rate": 0.0004184951084987319, "loss": 1.0265, "step": 81480 }, { "epoch": 3.28, "learning_rate": 0.0004184850436813076, "loss": 1.0181, "step": 81490 }, { "epoch": 3.28, "learning_rate": 0.0004184749788638834, "loss": 1.1383, "step": 81500 }, { "epoch": 3.28, "learning_rate": 0.0004184649140464592, "loss": 0.9259, "step": 81510 }, { "epoch": 3.28, "learning_rate": 0.000418454849229035, "loss": 1.3201, "step": 81520 }, { "epoch": 3.28, "learning_rate": 0.0004184447844116108, "loss": 1.0479, "step": 81530 }, { "epoch": 3.28, "learning_rate": 0.00041843471959418656, "loss": 1.0753, "step": 81540 }, { "epoch": 3.28, "learning_rate": 0.00041842465477676235, "loss": 1.2457, "step": 81550 }, { "epoch": 3.28, "learning_rate": 0.00041841458995933815, "loss": 1.0181, "step": 81560 }, { "epoch": 3.28, "learning_rate": 0.00041840452514191394, "loss": 1.3089, "step": 81570 }, { "epoch": 3.28, "learning_rate": 0.00041839446032448973, "loss": 1.3212, "step": 81580 }, { "epoch": 3.28, "learning_rate": 0.0004183843955070655, "loss": 1.0865, "step": 81590 }, { "epoch": 3.28, "learning_rate": 0.0004183743306896413, "loss": 1.2222, "step": 81600 }, { "epoch": 3.28, "learning_rate": 0.0004183642658722171, "loss": 1.1511, "step": 81610 }, { "epoch": 3.28, "learning_rate": 0.0004183542010547929, "loss": 1.2893, "step": 81620 }, { "epoch": 3.28, "learning_rate": 0.0004183441362373687, "loss": 1.0166, "step": 81630 }, { "epoch": 3.28, "learning_rate": 0.00041833407141994447, "loss": 1.123, "step": 81640 }, { "epoch": 3.28, "learning_rate": 0.0004183240066025202, "loss": 1.1731, "step": 81650 }, { "epoch": 3.28, "learning_rate": 0.000418313941785096, "loss": 1.0687, "step": 81660 }, { "epoch": 3.28, "learning_rate": 0.0004183038769676718, "loss": 1.1448, "step": 81670 }, { "epoch": 3.29, "learning_rate": 0.0004182938121502476, "loss": 1.0318, "step": 81680 }, { "epoch": 3.29, "learning_rate": 0.00041828374733282337, "loss": 1.0163, "step": 81690 }, { "epoch": 3.29, "learning_rate": 0.00041827368251539916, "loss": 0.9283, "step": 81700 }, { "epoch": 3.29, "learning_rate": 0.00041826361769797495, "loss": 1.1287, "step": 81710 }, { "epoch": 3.29, "learning_rate": 0.0004182535528805508, "loss": 0.9753, "step": 81720 }, { "epoch": 3.29, "learning_rate": 0.0004182434880631266, "loss": 1.1982, "step": 81730 }, { "epoch": 3.29, "learning_rate": 0.0004182334232457023, "loss": 1.1515, "step": 81740 }, { "epoch": 3.29, "learning_rate": 0.0004182233584282781, "loss": 0.9146, "step": 81750 }, { "epoch": 3.29, "learning_rate": 0.0004182132936108539, "loss": 1.0416, "step": 81760 }, { "epoch": 3.29, "learning_rate": 0.0004182032287934297, "loss": 1.0378, "step": 81770 }, { "epoch": 3.29, "learning_rate": 0.0004181931639760055, "loss": 1.1297, "step": 81780 }, { "epoch": 3.29, "learning_rate": 0.0004181830991585813, "loss": 1.1904, "step": 81790 }, { "epoch": 3.29, "learning_rate": 0.00041817303434115707, "loss": 0.9873, "step": 81800 }, { "epoch": 3.29, "learning_rate": 0.00041816296952373286, "loss": 1.1366, "step": 81810 }, { "epoch": 3.29, "learning_rate": 0.0004181529047063086, "loss": 1.1485, "step": 81820 }, { "epoch": 3.29, "learning_rate": 0.0004181428398888844, "loss": 1.1007, "step": 81830 }, { "epoch": 3.29, "learning_rate": 0.0004181327750714602, "loss": 1.1939, "step": 81840 }, { "epoch": 3.29, "learning_rate": 0.00041812271025403597, "loss": 1.1501, "step": 81850 }, { "epoch": 3.29, "learning_rate": 0.0004181126454366118, "loss": 1.145, "step": 81860 }, { "epoch": 3.29, "learning_rate": 0.0004181025806191876, "loss": 1.1725, "step": 81870 }, { "epoch": 3.29, "learning_rate": 0.0004180925158017634, "loss": 1.1702, "step": 81880 }, { "epoch": 3.29, "learning_rate": 0.0004180824509843392, "loss": 1.1224, "step": 81890 }, { "epoch": 3.29, "learning_rate": 0.000418072386166915, "loss": 1.0526, "step": 81900 }, { "epoch": 3.29, "learning_rate": 0.0004180623213494907, "loss": 1.2527, "step": 81910 }, { "epoch": 3.29, "learning_rate": 0.0004180522565320665, "loss": 1.0431, "step": 81920 }, { "epoch": 3.3, "learning_rate": 0.0004180421917146423, "loss": 1.0886, "step": 81930 }, { "epoch": 3.3, "learning_rate": 0.0004180321268972181, "loss": 1.1799, "step": 81940 }, { "epoch": 3.3, "learning_rate": 0.0004180220620797939, "loss": 0.9663, "step": 81950 }, { "epoch": 3.3, "learning_rate": 0.00041801199726236967, "loss": 0.7861, "step": 81960 }, { "epoch": 3.3, "learning_rate": 0.00041800193244494546, "loss": 0.9852, "step": 81970 }, { "epoch": 3.3, "learning_rate": 0.00041799186762752125, "loss": 1.0111, "step": 81980 }, { "epoch": 3.3, "learning_rate": 0.000417981802810097, "loss": 1.0001, "step": 81990 }, { "epoch": 3.3, "learning_rate": 0.0004179717379926728, "loss": 1.2744, "step": 82000 }, { "epoch": 3.3, "learning_rate": 0.0004179616731752486, "loss": 1.1582, "step": 82010 }, { "epoch": 3.3, "learning_rate": 0.0004179516083578244, "loss": 1.0851, "step": 82020 }, { "epoch": 3.3, "learning_rate": 0.0004179415435404002, "loss": 1.11, "step": 82030 }, { "epoch": 3.3, "learning_rate": 0.000417931478722976, "loss": 1.0514, "step": 82040 }, { "epoch": 3.3, "learning_rate": 0.0004179214139055518, "loss": 0.9404, "step": 82050 }, { "epoch": 3.3, "learning_rate": 0.0004179113490881276, "loss": 1.1419, "step": 82060 }, { "epoch": 3.3, "learning_rate": 0.00041790128427070337, "loss": 1.0939, "step": 82070 }, { "epoch": 3.3, "learning_rate": 0.0004178912194532791, "loss": 1.092, "step": 82080 }, { "epoch": 3.3, "learning_rate": 0.0004178811546358549, "loss": 1.2469, "step": 82090 }, { "epoch": 3.3, "learning_rate": 0.0004178710898184307, "loss": 1.0399, "step": 82100 }, { "epoch": 3.3, "learning_rate": 0.0004178610250010065, "loss": 1.1211, "step": 82110 }, { "epoch": 3.3, "learning_rate": 0.00041785096018358227, "loss": 0.992, "step": 82120 }, { "epoch": 3.3, "learning_rate": 0.00041784089536615806, "loss": 1.1558, "step": 82130 }, { "epoch": 3.3, "learning_rate": 0.00041783083054873385, "loss": 0.897, "step": 82140 }, { "epoch": 3.3, "learning_rate": 0.00041782076573130964, "loss": 1.206, "step": 82150 }, { "epoch": 3.3, "learning_rate": 0.00041781070091388543, "loss": 1.3478, "step": 82160 }, { "epoch": 3.3, "learning_rate": 0.0004178006360964612, "loss": 1.0445, "step": 82170 }, { "epoch": 3.31, "learning_rate": 0.000417790571279037, "loss": 0.8109, "step": 82180 }, { "epoch": 3.31, "learning_rate": 0.0004177805064616128, "loss": 1.1098, "step": 82190 }, { "epoch": 3.31, "learning_rate": 0.0004177704416441886, "loss": 1.0497, "step": 82200 }, { "epoch": 3.31, "learning_rate": 0.0004177603768267644, "loss": 0.9652, "step": 82210 }, { "epoch": 3.31, "learning_rate": 0.0004177503120093402, "loss": 0.9436, "step": 82220 }, { "epoch": 3.31, "learning_rate": 0.00041774024719191597, "loss": 0.8379, "step": 82230 }, { "epoch": 3.31, "learning_rate": 0.00041773018237449176, "loss": 1.2349, "step": 82240 }, { "epoch": 3.31, "learning_rate": 0.0004177201175570675, "loss": 1.1516, "step": 82250 }, { "epoch": 3.31, "learning_rate": 0.0004177100527396433, "loss": 1.0305, "step": 82260 }, { "epoch": 3.31, "learning_rate": 0.0004176999879222191, "loss": 1.0532, "step": 82270 }, { "epoch": 3.31, "learning_rate": 0.00041768992310479487, "loss": 0.8986, "step": 82280 }, { "epoch": 3.31, "learning_rate": 0.00041767985828737066, "loss": 0.9984, "step": 82290 }, { "epoch": 3.31, "learning_rate": 0.0004176697934699465, "loss": 1.1393, "step": 82300 }, { "epoch": 3.31, "learning_rate": 0.0004176597286525223, "loss": 0.9505, "step": 82310 }, { "epoch": 3.31, "learning_rate": 0.0004176496638350981, "loss": 1.1222, "step": 82320 }, { "epoch": 3.31, "learning_rate": 0.0004176395990176738, "loss": 1.3629, "step": 82330 }, { "epoch": 3.31, "learning_rate": 0.0004176295342002496, "loss": 1.1969, "step": 82340 }, { "epoch": 3.31, "learning_rate": 0.0004176194693828254, "loss": 0.9923, "step": 82350 }, { "epoch": 3.31, "learning_rate": 0.0004176094045654012, "loss": 1.0384, "step": 82360 }, { "epoch": 3.31, "learning_rate": 0.000417599339747977, "loss": 1.025, "step": 82370 }, { "epoch": 3.31, "learning_rate": 0.0004175892749305528, "loss": 0.9299, "step": 82380 }, { "epoch": 3.31, "learning_rate": 0.00041757921011312856, "loss": 0.8349, "step": 82390 }, { "epoch": 3.31, "learning_rate": 0.00041756914529570435, "loss": 1.0517, "step": 82400 }, { "epoch": 3.31, "learning_rate": 0.0004175590804782801, "loss": 1.1784, "step": 82410 }, { "epoch": 3.31, "learning_rate": 0.0004175490156608559, "loss": 1.0626, "step": 82420 }, { "epoch": 3.32, "learning_rate": 0.00041753895084343167, "loss": 1.0195, "step": 82430 }, { "epoch": 3.32, "learning_rate": 0.00041752888602600746, "loss": 1.2938, "step": 82440 }, { "epoch": 3.32, "learning_rate": 0.0004175188212085833, "loss": 1.2188, "step": 82450 }, { "epoch": 3.32, "learning_rate": 0.0004175087563911591, "loss": 1.1232, "step": 82460 }, { "epoch": 3.32, "learning_rate": 0.0004174986915737349, "loss": 1.15, "step": 82470 }, { "epoch": 3.32, "learning_rate": 0.0004174886267563107, "loss": 1.0945, "step": 82480 }, { "epoch": 3.32, "learning_rate": 0.00041747856193888647, "loss": 1.2239, "step": 82490 }, { "epoch": 3.32, "learning_rate": 0.0004174684971214622, "loss": 0.974, "step": 82500 }, { "epoch": 3.32, "learning_rate": 0.000417458432304038, "loss": 0.9669, "step": 82510 }, { "epoch": 3.32, "learning_rate": 0.0004174483674866138, "loss": 1.1237, "step": 82520 }, { "epoch": 3.32, "learning_rate": 0.0004174383026691896, "loss": 1.0665, "step": 82530 }, { "epoch": 3.32, "learning_rate": 0.00041742823785176537, "loss": 0.9608, "step": 82540 }, { "epoch": 3.32, "learning_rate": 0.00041741817303434116, "loss": 1.2497, "step": 82550 }, { "epoch": 3.32, "learning_rate": 0.00041740810821691695, "loss": 1.1475, "step": 82560 }, { "epoch": 3.32, "learning_rate": 0.00041739804339949274, "loss": 1.0367, "step": 82570 }, { "epoch": 3.32, "learning_rate": 0.0004173879785820685, "loss": 1.2019, "step": 82580 }, { "epoch": 3.32, "learning_rate": 0.00041737791376464427, "loss": 1.0261, "step": 82590 }, { "epoch": 3.32, "learning_rate": 0.0004173678489472201, "loss": 1.1538, "step": 82600 }, { "epoch": 3.32, "learning_rate": 0.0004173577841297959, "loss": 1.2691, "step": 82610 }, { "epoch": 3.32, "learning_rate": 0.0004173477193123717, "loss": 0.9846, "step": 82620 }, { "epoch": 3.32, "learning_rate": 0.0004173376544949475, "loss": 0.9114, "step": 82630 }, { "epoch": 3.32, "learning_rate": 0.0004173275896775233, "loss": 0.9888, "step": 82640 }, { "epoch": 3.32, "learning_rate": 0.00041731752486009907, "loss": 1.1712, "step": 82650 }, { "epoch": 3.32, "learning_rate": 0.00041730746004267486, "loss": 1.1725, "step": 82660 }, { "epoch": 3.32, "learning_rate": 0.0004172973952252506, "loss": 0.8587, "step": 82670 }, { "epoch": 3.33, "learning_rate": 0.0004172873304078264, "loss": 0.9447, "step": 82680 }, { "epoch": 3.33, "learning_rate": 0.0004172772655904022, "loss": 0.9717, "step": 82690 }, { "epoch": 3.33, "learning_rate": 0.00041726720077297797, "loss": 1.0061, "step": 82700 }, { "epoch": 3.33, "learning_rate": 0.00041725713595555376, "loss": 1.1816, "step": 82710 }, { "epoch": 3.33, "learning_rate": 0.00041724707113812955, "loss": 1.1221, "step": 82720 }, { "epoch": 3.33, "learning_rate": 0.00041723700632070534, "loss": 1.2696, "step": 82730 }, { "epoch": 3.33, "learning_rate": 0.0004172269415032812, "loss": 1.2063, "step": 82740 }, { "epoch": 3.33, "learning_rate": 0.0004172168766858569, "loss": 1.2103, "step": 82750 }, { "epoch": 3.33, "learning_rate": 0.0004172068118684327, "loss": 1.1054, "step": 82760 }, { "epoch": 3.33, "learning_rate": 0.0004171967470510085, "loss": 1.4132, "step": 82770 }, { "epoch": 3.33, "learning_rate": 0.0004171866822335843, "loss": 0.932, "step": 82780 }, { "epoch": 3.33, "learning_rate": 0.0004171766174161601, "loss": 1.1006, "step": 82790 }, { "epoch": 3.33, "learning_rate": 0.0004171665525987359, "loss": 1.0757, "step": 82800 }, { "epoch": 3.33, "learning_rate": 0.00041715648778131167, "loss": 1.0334, "step": 82810 }, { "epoch": 3.33, "learning_rate": 0.00041714642296388746, "loss": 1.0974, "step": 82820 }, { "epoch": 3.33, "learning_rate": 0.00041713635814646325, "loss": 1.1727, "step": 82830 }, { "epoch": 3.33, "learning_rate": 0.000417126293329039, "loss": 1.0282, "step": 82840 }, { "epoch": 3.33, "learning_rate": 0.0004171162285116148, "loss": 0.977, "step": 82850 }, { "epoch": 3.33, "learning_rate": 0.00041710616369419057, "loss": 1.2192, "step": 82860 }, { "epoch": 3.33, "learning_rate": 0.00041709609887676636, "loss": 1.2805, "step": 82870 }, { "epoch": 3.33, "learning_rate": 0.00041708603405934215, "loss": 1.1828, "step": 82880 }, { "epoch": 3.33, "learning_rate": 0.000417075969241918, "loss": 1.1298, "step": 82890 }, { "epoch": 3.33, "learning_rate": 0.0004170659044244938, "loss": 1.319, "step": 82900 }, { "epoch": 3.33, "learning_rate": 0.0004170558396070696, "loss": 1.0848, "step": 82910 }, { "epoch": 3.33, "learning_rate": 0.0004170457747896453, "loss": 1.1397, "step": 82920 }, { "epoch": 3.34, "learning_rate": 0.0004170357099722211, "loss": 1.0934, "step": 82930 }, { "epoch": 3.34, "learning_rate": 0.0004170256451547969, "loss": 1.0296, "step": 82940 }, { "epoch": 3.34, "learning_rate": 0.0004170155803373727, "loss": 1.2339, "step": 82950 }, { "epoch": 3.34, "learning_rate": 0.0004170055155199485, "loss": 1.0861, "step": 82960 }, { "epoch": 3.34, "learning_rate": 0.00041699545070252427, "loss": 1.2128, "step": 82970 }, { "epoch": 3.34, "learning_rate": 0.00041698538588510006, "loss": 1.1938, "step": 82980 }, { "epoch": 3.34, "learning_rate": 0.00041697532106767585, "loss": 0.964, "step": 82990 }, { "epoch": 3.34, "learning_rate": 0.00041696525625025164, "loss": 1.3136, "step": 83000 }, { "epoch": 3.34, "learning_rate": 0.0004169551914328274, "loss": 1.1139, "step": 83010 }, { "epoch": 3.34, "learning_rate": 0.00041694512661540317, "loss": 1.0416, "step": 83020 }, { "epoch": 3.34, "learning_rate": 0.000416935061797979, "loss": 1.1751, "step": 83030 }, { "epoch": 3.34, "learning_rate": 0.0004169249969805548, "loss": 1.0981, "step": 83040 }, { "epoch": 3.34, "learning_rate": 0.0004169149321631306, "loss": 1.0755, "step": 83050 }, { "epoch": 3.34, "learning_rate": 0.0004169048673457064, "loss": 1.3561, "step": 83060 }, { "epoch": 3.34, "learning_rate": 0.0004168948025282822, "loss": 1.1604, "step": 83070 }, { "epoch": 3.34, "learning_rate": 0.00041688473771085797, "loss": 1.1843, "step": 83080 }, { "epoch": 3.34, "learning_rate": 0.0004168746728934337, "loss": 1.1906, "step": 83090 }, { "epoch": 3.34, "learning_rate": 0.0004168646080760095, "loss": 1.1531, "step": 83100 }, { "epoch": 3.34, "learning_rate": 0.0004168545432585853, "loss": 0.9563, "step": 83110 }, { "epoch": 3.34, "learning_rate": 0.0004168444784411611, "loss": 1.1165, "step": 83120 }, { "epoch": 3.34, "learning_rate": 0.00041683441362373687, "loss": 1.09, "step": 83130 }, { "epoch": 3.34, "learning_rate": 0.00041682434880631266, "loss": 1.2307, "step": 83140 }, { "epoch": 3.34, "learning_rate": 0.00041681428398888845, "loss": 1.182, "step": 83150 }, { "epoch": 3.34, "learning_rate": 0.00041680421917146424, "loss": 1.0361, "step": 83160 }, { "epoch": 3.34, "learning_rate": 0.00041679415435404, "loss": 1.0528, "step": 83170 }, { "epoch": 3.35, "learning_rate": 0.0004167840895366158, "loss": 1.1462, "step": 83180 }, { "epoch": 3.35, "learning_rate": 0.0004167740247191916, "loss": 1.1959, "step": 83190 }, { "epoch": 3.35, "learning_rate": 0.0004167639599017674, "loss": 1.0374, "step": 83200 }, { "epoch": 3.35, "learning_rate": 0.0004167538950843432, "loss": 1.0128, "step": 83210 }, { "epoch": 3.35, "learning_rate": 0.000416743830266919, "loss": 1.1162, "step": 83220 }, { "epoch": 3.35, "learning_rate": 0.0004167337654494948, "loss": 1.2008, "step": 83230 }, { "epoch": 3.35, "learning_rate": 0.00041672370063207056, "loss": 1.0192, "step": 83240 }, { "epoch": 3.35, "learning_rate": 0.00041671363581464635, "loss": 1.3539, "step": 83250 }, { "epoch": 3.35, "learning_rate": 0.0004167035709972221, "loss": 0.9493, "step": 83260 }, { "epoch": 3.35, "learning_rate": 0.0004166935061797979, "loss": 1.0914, "step": 83270 }, { "epoch": 3.35, "learning_rate": 0.00041668344136237367, "loss": 0.982, "step": 83280 }, { "epoch": 3.35, "learning_rate": 0.00041667337654494946, "loss": 1.1651, "step": 83290 }, { "epoch": 3.35, "learning_rate": 0.00041666331172752525, "loss": 0.9883, "step": 83300 }, { "epoch": 3.35, "learning_rate": 0.00041665324691010104, "loss": 1.0995, "step": 83310 }, { "epoch": 3.35, "learning_rate": 0.00041664318209267684, "loss": 0.9901, "step": 83320 }, { "epoch": 3.35, "learning_rate": 0.0004166331172752527, "loss": 1.2016, "step": 83330 }, { "epoch": 3.35, "learning_rate": 0.0004166230524578284, "loss": 0.8873, "step": 83340 }, { "epoch": 3.35, "learning_rate": 0.0004166129876404042, "loss": 1.0894, "step": 83350 }, { "epoch": 3.35, "learning_rate": 0.00041660292282298, "loss": 1.1502, "step": 83360 }, { "epoch": 3.35, "learning_rate": 0.0004165928580055558, "loss": 1.0012, "step": 83370 }, { "epoch": 3.35, "learning_rate": 0.0004165827931881316, "loss": 0.9569, "step": 83380 }, { "epoch": 3.35, "learning_rate": 0.00041657272837070737, "loss": 1.0474, "step": 83390 }, { "epoch": 3.35, "learning_rate": 0.00041656266355328316, "loss": 1.126, "step": 83400 }, { "epoch": 3.35, "learning_rate": 0.00041655259873585895, "loss": 1.0009, "step": 83410 }, { "epoch": 3.36, "learning_rate": 0.00041654253391843474, "loss": 1.4356, "step": 83420 }, { "epoch": 3.36, "learning_rate": 0.0004165324691010105, "loss": 1.236, "step": 83430 }, { "epoch": 3.36, "learning_rate": 0.00041652240428358627, "loss": 1.2234, "step": 83440 }, { "epoch": 3.36, "learning_rate": 0.00041651233946616206, "loss": 1.1666, "step": 83450 }, { "epoch": 3.36, "learning_rate": 0.00041650227464873785, "loss": 1.1474, "step": 83460 }, { "epoch": 3.36, "learning_rate": 0.0004164922098313137, "loss": 1.02, "step": 83470 }, { "epoch": 3.36, "learning_rate": 0.0004164821450138895, "loss": 0.8583, "step": 83480 }, { "epoch": 3.36, "learning_rate": 0.0004164720801964653, "loss": 1.1324, "step": 83490 }, { "epoch": 3.36, "learning_rate": 0.00041646201537904107, "loss": 1.0252, "step": 83500 }, { "epoch": 3.36, "learning_rate": 0.0004164519505616168, "loss": 0.9435, "step": 83510 }, { "epoch": 3.36, "learning_rate": 0.0004164418857441926, "loss": 1.0094, "step": 83520 }, { "epoch": 3.36, "learning_rate": 0.0004164318209267684, "loss": 1.134, "step": 83530 }, { "epoch": 3.36, "learning_rate": 0.0004164217561093442, "loss": 1.1458, "step": 83540 }, { "epoch": 3.36, "learning_rate": 0.00041641169129191997, "loss": 1.1964, "step": 83550 }, { "epoch": 3.36, "learning_rate": 0.00041640162647449576, "loss": 1.4507, "step": 83560 }, { "epoch": 3.36, "learning_rate": 0.00041639156165707155, "loss": 1.2457, "step": 83570 }, { "epoch": 3.36, "learning_rate": 0.00041638149683964734, "loss": 1.266, "step": 83580 }, { "epoch": 3.36, "learning_rate": 0.00041637143202222313, "loss": 1.0238, "step": 83590 }, { "epoch": 3.36, "learning_rate": 0.00041636136720479887, "loss": 1.4196, "step": 83600 }, { "epoch": 3.36, "learning_rate": 0.00041635130238737466, "loss": 1.1884, "step": 83610 }, { "epoch": 3.36, "learning_rate": 0.0004163412375699505, "loss": 1.0176, "step": 83620 }, { "epoch": 3.36, "learning_rate": 0.0004163311727525263, "loss": 0.9898, "step": 83630 }, { "epoch": 3.36, "learning_rate": 0.0004163211079351021, "loss": 1.0795, "step": 83640 }, { "epoch": 3.36, "learning_rate": 0.0004163110431176779, "loss": 1.0444, "step": 83650 }, { "epoch": 3.36, "learning_rate": 0.00041630097830025367, "loss": 1.0859, "step": 83660 }, { "epoch": 3.37, "learning_rate": 0.00041629091348282946, "loss": 1.0458, "step": 83670 }, { "epoch": 3.37, "learning_rate": 0.0004162808486654052, "loss": 1.0305, "step": 83680 }, { "epoch": 3.37, "learning_rate": 0.000416270783847981, "loss": 0.9191, "step": 83690 }, { "epoch": 3.37, "learning_rate": 0.0004162607190305568, "loss": 0.9901, "step": 83700 }, { "epoch": 3.37, "learning_rate": 0.00041625065421313257, "loss": 1.153, "step": 83710 }, { "epoch": 3.37, "learning_rate": 0.00041624058939570836, "loss": 1.3294, "step": 83720 }, { "epoch": 3.37, "learning_rate": 0.00041623052457828415, "loss": 1.417, "step": 83730 }, { "epoch": 3.37, "learning_rate": 0.00041622045976085994, "loss": 1.0994, "step": 83740 }, { "epoch": 3.37, "learning_rate": 0.00041621039494343573, "loss": 1.3394, "step": 83750 }, { "epoch": 3.37, "learning_rate": 0.0004162003301260116, "loss": 1.0534, "step": 83760 }, { "epoch": 3.37, "learning_rate": 0.0004161902653085873, "loss": 1.0309, "step": 83770 }, { "epoch": 3.37, "learning_rate": 0.0004161802004911631, "loss": 1.2132, "step": 83780 }, { "epoch": 3.37, "learning_rate": 0.0004161701356737389, "loss": 1.0204, "step": 83790 }, { "epoch": 3.37, "learning_rate": 0.0004161600708563147, "loss": 1.2937, "step": 83800 }, { "epoch": 3.37, "learning_rate": 0.0004161500060388905, "loss": 1.0332, "step": 83810 }, { "epoch": 3.37, "learning_rate": 0.00041613994122146627, "loss": 1.0133, "step": 83820 }, { "epoch": 3.37, "learning_rate": 0.00041612987640404206, "loss": 0.8892, "step": 83830 }, { "epoch": 3.37, "learning_rate": 0.00041611981158661785, "loss": 1.2155, "step": 83840 }, { "epoch": 3.37, "learning_rate": 0.0004161097467691936, "loss": 1.1698, "step": 83850 }, { "epoch": 3.37, "learning_rate": 0.0004160996819517694, "loss": 1.151, "step": 83860 }, { "epoch": 3.37, "learning_rate": 0.00041608961713434517, "loss": 1.3241, "step": 83870 }, { "epoch": 3.37, "learning_rate": 0.00041607955231692096, "loss": 1.188, "step": 83880 }, { "epoch": 3.37, "learning_rate": 0.00041606948749949675, "loss": 1.0598, "step": 83890 }, { "epoch": 3.37, "learning_rate": 0.00041605942268207254, "loss": 1.0885, "step": 83900 }, { "epoch": 3.37, "learning_rate": 0.0004160493578646484, "loss": 1.01, "step": 83910 }, { "epoch": 3.38, "learning_rate": 0.0004160392930472242, "loss": 0.9904, "step": 83920 }, { "epoch": 3.38, "learning_rate": 0.00041602922822979997, "loss": 0.9804, "step": 83930 }, { "epoch": 3.38, "learning_rate": 0.0004160191634123757, "loss": 1.0683, "step": 83940 }, { "epoch": 3.38, "learning_rate": 0.0004160090985949515, "loss": 1.3353, "step": 83950 }, { "epoch": 3.38, "learning_rate": 0.0004159990337775273, "loss": 1.1289, "step": 83960 }, { "epoch": 3.38, "learning_rate": 0.0004159889689601031, "loss": 1.0086, "step": 83970 }, { "epoch": 3.38, "learning_rate": 0.00041597890414267886, "loss": 0.9014, "step": 83980 }, { "epoch": 3.38, "learning_rate": 0.00041596883932525466, "loss": 1.0494, "step": 83990 }, { "epoch": 3.38, "learning_rate": 0.00041595877450783045, "loss": 1.0003, "step": 84000 }, { "epoch": 3.38, "learning_rate": 0.00041594870969040624, "loss": 1.125, "step": 84010 }, { "epoch": 3.38, "learning_rate": 0.000415938644872982, "loss": 1.1028, "step": 84020 }, { "epoch": 3.38, "learning_rate": 0.00041592858005555776, "loss": 1.1808, "step": 84030 }, { "epoch": 3.38, "learning_rate": 0.00041591851523813356, "loss": 1.1027, "step": 84040 }, { "epoch": 3.38, "learning_rate": 0.00041590845042070935, "loss": 1.259, "step": 84050 }, { "epoch": 3.38, "learning_rate": 0.0004158983856032852, "loss": 1.0879, "step": 84060 }, { "epoch": 3.38, "learning_rate": 0.000415888320785861, "loss": 0.9143, "step": 84070 }, { "epoch": 3.38, "learning_rate": 0.00041587825596843677, "loss": 1.1474, "step": 84080 }, { "epoch": 3.38, "learning_rate": 0.00041586819115101256, "loss": 1.0867, "step": 84090 }, { "epoch": 3.38, "learning_rate": 0.0004158581263335883, "loss": 1.0287, "step": 84100 }, { "epoch": 3.38, "learning_rate": 0.0004158480615161641, "loss": 0.8985, "step": 84110 }, { "epoch": 3.38, "learning_rate": 0.0004158379966987399, "loss": 1.2902, "step": 84120 }, { "epoch": 3.38, "learning_rate": 0.00041582793188131567, "loss": 1.0164, "step": 84130 }, { "epoch": 3.38, "learning_rate": 0.00041581786706389146, "loss": 1.3146, "step": 84140 }, { "epoch": 3.38, "learning_rate": 0.00041580780224646725, "loss": 1.1531, "step": 84150 }, { "epoch": 3.38, "learning_rate": 0.00041579773742904304, "loss": 1.2413, "step": 84160 }, { "epoch": 3.39, "learning_rate": 0.00041578767261161884, "loss": 1.089, "step": 84170 }, { "epoch": 3.39, "learning_rate": 0.0004157776077941946, "loss": 1.3312, "step": 84180 }, { "epoch": 3.39, "learning_rate": 0.00041576754297677036, "loss": 1.1665, "step": 84190 }, { "epoch": 3.39, "learning_rate": 0.0004157574781593462, "loss": 1.0744, "step": 84200 }, { "epoch": 3.39, "learning_rate": 0.000415747413341922, "loss": 1.1191, "step": 84210 }, { "epoch": 3.39, "learning_rate": 0.0004157373485244978, "loss": 0.9254, "step": 84220 }, { "epoch": 3.39, "learning_rate": 0.0004157272837070736, "loss": 1.1351, "step": 84230 }, { "epoch": 3.39, "learning_rate": 0.00041571721888964937, "loss": 1.0098, "step": 84240 }, { "epoch": 3.39, "learning_rate": 0.00041570715407222516, "loss": 1.0529, "step": 84250 }, { "epoch": 3.39, "learning_rate": 0.00041569708925480095, "loss": 1.3193, "step": 84260 }, { "epoch": 3.39, "learning_rate": 0.0004156870244373767, "loss": 0.8638, "step": 84270 }, { "epoch": 3.39, "learning_rate": 0.0004156769596199525, "loss": 1.0814, "step": 84280 }, { "epoch": 3.39, "learning_rate": 0.00041566689480252827, "loss": 1.0403, "step": 84290 }, { "epoch": 3.39, "learning_rate": 0.00041565682998510406, "loss": 1.1545, "step": 84300 }, { "epoch": 3.39, "learning_rate": 0.00041564676516767985, "loss": 1.2017, "step": 84310 }, { "epoch": 3.39, "learning_rate": 0.00041563670035025564, "loss": 1.0215, "step": 84320 }, { "epoch": 3.39, "learning_rate": 0.00041562663553283143, "loss": 0.9138, "step": 84330 }, { "epoch": 3.39, "learning_rate": 0.0004156165707154072, "loss": 0.9617, "step": 84340 }, { "epoch": 3.39, "learning_rate": 0.00041560650589798307, "loss": 1.1183, "step": 84350 }, { "epoch": 3.39, "learning_rate": 0.0004155964410805588, "loss": 1.2023, "step": 84360 }, { "epoch": 3.39, "learning_rate": 0.0004155863762631346, "loss": 1.1246, "step": 84370 }, { "epoch": 3.39, "learning_rate": 0.0004155763114457104, "loss": 1.089, "step": 84380 }, { "epoch": 3.39, "learning_rate": 0.0004155662466282862, "loss": 1.0047, "step": 84390 }, { "epoch": 3.39, "learning_rate": 0.00041555618181086197, "loss": 1.1727, "step": 84400 }, { "epoch": 3.39, "learning_rate": 0.00041554611699343776, "loss": 1.2118, "step": 84410 }, { "epoch": 3.4, "learning_rate": 0.00041553605217601355, "loss": 1.2786, "step": 84420 }, { "epoch": 3.4, "learning_rate": 0.00041552598735858934, "loss": 1.0636, "step": 84430 }, { "epoch": 3.4, "learning_rate": 0.0004155159225411651, "loss": 1.217, "step": 84440 }, { "epoch": 3.4, "learning_rate": 0.00041550585772374087, "loss": 0.9104, "step": 84450 }, { "epoch": 3.4, "learning_rate": 0.00041549579290631666, "loss": 1.0766, "step": 84460 }, { "epoch": 3.4, "learning_rate": 0.00041548572808889245, "loss": 1.1931, "step": 84470 }, { "epoch": 3.4, "learning_rate": 0.00041547566327146824, "loss": 1.0167, "step": 84480 }, { "epoch": 3.4, "learning_rate": 0.00041546559845404403, "loss": 0.9179, "step": 84490 }, { "epoch": 3.4, "learning_rate": 0.0004154555336366199, "loss": 0.8488, "step": 84500 }, { "epoch": 3.4, "learning_rate": 0.00041544546881919567, "loss": 1.0601, "step": 84510 }, { "epoch": 3.4, "learning_rate": 0.00041543540400177146, "loss": 1.0493, "step": 84520 }, { "epoch": 3.4, "learning_rate": 0.0004154253391843472, "loss": 1.0749, "step": 84530 }, { "epoch": 3.4, "learning_rate": 0.000415415274366923, "loss": 1.0372, "step": 84540 }, { "epoch": 3.4, "learning_rate": 0.0004154052095494988, "loss": 1.0568, "step": 84550 }, { "epoch": 3.4, "learning_rate": 0.00041539514473207457, "loss": 1.2905, "step": 84560 }, { "epoch": 3.4, "learning_rate": 0.00041538507991465036, "loss": 1.2629, "step": 84570 }, { "epoch": 3.4, "learning_rate": 0.00041537501509722615, "loss": 0.9757, "step": 84580 }, { "epoch": 3.4, "learning_rate": 0.00041536495027980194, "loss": 1.0458, "step": 84590 }, { "epoch": 3.4, "learning_rate": 0.00041535488546237773, "loss": 1.2071, "step": 84600 }, { "epoch": 3.4, "learning_rate": 0.00041534482064495347, "loss": 1.1003, "step": 84610 }, { "epoch": 3.4, "learning_rate": 0.00041533475582752926, "loss": 0.8454, "step": 84620 }, { "epoch": 3.4, "learning_rate": 0.00041532469101010505, "loss": 1.2178, "step": 84630 }, { "epoch": 3.4, "learning_rate": 0.0004153146261926809, "loss": 1.1268, "step": 84640 }, { "epoch": 3.4, "learning_rate": 0.0004153045613752567, "loss": 1.0755, "step": 84650 }, { "epoch": 3.4, "learning_rate": 0.0004152944965578325, "loss": 1.0526, "step": 84660 }, { "epoch": 3.41, "learning_rate": 0.00041528443174040827, "loss": 1.0937, "step": 84670 }, { "epoch": 3.41, "learning_rate": 0.00041527436692298406, "loss": 1.1424, "step": 84680 }, { "epoch": 3.41, "learning_rate": 0.00041526430210555985, "loss": 1.0464, "step": 84690 }, { "epoch": 3.41, "learning_rate": 0.0004152542372881356, "loss": 1.0241, "step": 84700 }, { "epoch": 3.41, "learning_rate": 0.0004152441724707114, "loss": 1.0934, "step": 84710 }, { "epoch": 3.41, "learning_rate": 0.00041523410765328717, "loss": 1.0895, "step": 84720 }, { "epoch": 3.41, "learning_rate": 0.00041522404283586296, "loss": 1.025, "step": 84730 }, { "epoch": 3.41, "learning_rate": 0.00041521397801843875, "loss": 1.2265, "step": 84740 }, { "epoch": 3.41, "learning_rate": 0.00041520391320101454, "loss": 1.2322, "step": 84750 }, { "epoch": 3.41, "learning_rate": 0.00041519384838359033, "loss": 1.1131, "step": 84760 }, { "epoch": 3.41, "learning_rate": 0.0004151837835661661, "loss": 0.9506, "step": 84770 }, { "epoch": 3.41, "learning_rate": 0.00041517371874874186, "loss": 1.1562, "step": 84780 }, { "epoch": 3.41, "learning_rate": 0.0004151636539313177, "loss": 0.991, "step": 84790 }, { "epoch": 3.41, "learning_rate": 0.0004151535891138935, "loss": 1.0959, "step": 84800 }, { "epoch": 3.41, "learning_rate": 0.0004151435242964693, "loss": 1.1275, "step": 84810 }, { "epoch": 3.41, "learning_rate": 0.0004151334594790451, "loss": 1.2031, "step": 84820 }, { "epoch": 3.41, "learning_rate": 0.00041512339466162086, "loss": 0.9258, "step": 84830 }, { "epoch": 3.41, "learning_rate": 0.00041511332984419666, "loss": 1.2665, "step": 84840 }, { "epoch": 3.41, "learning_rate": 0.00041510326502677245, "loss": 1.4065, "step": 84850 }, { "epoch": 3.41, "learning_rate": 0.0004150932002093482, "loss": 1.1054, "step": 84860 }, { "epoch": 3.41, "learning_rate": 0.000415083135391924, "loss": 0.9202, "step": 84870 }, { "epoch": 3.41, "learning_rate": 0.00041507307057449976, "loss": 1.057, "step": 84880 }, { "epoch": 3.41, "learning_rate": 0.00041506300575707556, "loss": 1.1863, "step": 84890 }, { "epoch": 3.41, "learning_rate": 0.00041505294093965135, "loss": 1.2008, "step": 84900 }, { "epoch": 3.41, "learning_rate": 0.00041504287612222714, "loss": 1.449, "step": 84910 }, { "epoch": 3.42, "learning_rate": 0.00041503281130480293, "loss": 1.1904, "step": 84920 }, { "epoch": 3.42, "learning_rate": 0.00041502274648737877, "loss": 1.0577, "step": 84930 }, { "epoch": 3.42, "learning_rate": 0.00041501268166995456, "loss": 1.1975, "step": 84940 }, { "epoch": 3.42, "learning_rate": 0.0004150026168525303, "loss": 1.2054, "step": 84950 }, { "epoch": 3.42, "learning_rate": 0.0004149925520351061, "loss": 1.2113, "step": 84960 }, { "epoch": 3.42, "learning_rate": 0.0004149824872176819, "loss": 0.9421, "step": 84970 }, { "epoch": 3.42, "learning_rate": 0.00041497242240025767, "loss": 1.2339, "step": 84980 }, { "epoch": 3.42, "learning_rate": 0.00041496235758283346, "loss": 1.1521, "step": 84990 }, { "epoch": 3.42, "learning_rate": 0.00041495229276540925, "loss": 1.0737, "step": 85000 }, { "epoch": 3.42, "learning_rate": 0.00041494222794798504, "loss": 1.161, "step": 85010 }, { "epoch": 3.42, "learning_rate": 0.00041493216313056084, "loss": 1.197, "step": 85020 }, { "epoch": 3.42, "learning_rate": 0.00041492209831313657, "loss": 1.285, "step": 85030 }, { "epoch": 3.42, "learning_rate": 0.00041491203349571236, "loss": 1.0727, "step": 85040 }, { "epoch": 3.42, "learning_rate": 0.00041490196867828815, "loss": 1.214, "step": 85050 }, { "epoch": 3.42, "learning_rate": 0.00041489190386086394, "loss": 1.074, "step": 85060 }, { "epoch": 3.42, "learning_rate": 0.00041488183904343973, "loss": 1.1846, "step": 85070 }, { "epoch": 3.42, "learning_rate": 0.0004148717742260156, "loss": 0.9478, "step": 85080 }, { "epoch": 3.42, "learning_rate": 0.00041486170940859137, "loss": 0.9267, "step": 85090 }, { "epoch": 3.42, "learning_rate": 0.00041485164459116716, "loss": 1.2061, "step": 85100 }, { "epoch": 3.42, "learning_rate": 0.00041484157977374295, "loss": 1.2983, "step": 85110 }, { "epoch": 3.42, "learning_rate": 0.0004148315149563187, "loss": 1.1507, "step": 85120 }, { "epoch": 3.42, "learning_rate": 0.0004148214501388945, "loss": 1.1882, "step": 85130 }, { "epoch": 3.42, "learning_rate": 0.00041481138532147027, "loss": 1.1745, "step": 85140 }, { "epoch": 3.42, "learning_rate": 0.00041480132050404606, "loss": 1.1309, "step": 85150 }, { "epoch": 3.42, "learning_rate": 0.00041479125568662185, "loss": 1.0414, "step": 85160 }, { "epoch": 3.43, "learning_rate": 0.00041478119086919764, "loss": 1.0922, "step": 85170 }, { "epoch": 3.43, "learning_rate": 0.00041477112605177343, "loss": 1.1815, "step": 85180 }, { "epoch": 3.43, "learning_rate": 0.0004147610612343492, "loss": 1.1359, "step": 85190 }, { "epoch": 3.43, "learning_rate": 0.00041475099641692496, "loss": 0.9875, "step": 85200 }, { "epoch": 3.43, "learning_rate": 0.00041474093159950075, "loss": 1.0979, "step": 85210 }, { "epoch": 3.43, "learning_rate": 0.00041473086678207654, "loss": 1.1763, "step": 85220 }, { "epoch": 3.43, "learning_rate": 0.0004147208019646524, "loss": 1.1273, "step": 85230 }, { "epoch": 3.43, "learning_rate": 0.0004147107371472282, "loss": 1.0437, "step": 85240 }, { "epoch": 3.43, "learning_rate": 0.00041470067232980397, "loss": 1.071, "step": 85250 }, { "epoch": 3.43, "learning_rate": 0.00041469060751237976, "loss": 1.0472, "step": 85260 }, { "epoch": 3.43, "learning_rate": 0.00041468054269495555, "loss": 1.2304, "step": 85270 }, { "epoch": 3.43, "learning_rate": 0.00041467047787753134, "loss": 1.0379, "step": 85280 }, { "epoch": 3.43, "learning_rate": 0.0004146604130601071, "loss": 1.067, "step": 85290 }, { "epoch": 3.43, "learning_rate": 0.00041465034824268287, "loss": 0.9841, "step": 85300 }, { "epoch": 3.43, "learning_rate": 0.00041464028342525866, "loss": 1.2852, "step": 85310 }, { "epoch": 3.43, "learning_rate": 0.00041463021860783445, "loss": 0.9418, "step": 85320 }, { "epoch": 3.43, "learning_rate": 0.00041462015379041024, "loss": 1.007, "step": 85330 }, { "epoch": 3.43, "learning_rate": 0.00041461008897298603, "loss": 1.1071, "step": 85340 }, { "epoch": 3.43, "learning_rate": 0.0004146000241555618, "loss": 1.032, "step": 85350 }, { "epoch": 3.43, "learning_rate": 0.0004145899593381376, "loss": 0.9496, "step": 85360 }, { "epoch": 3.43, "learning_rate": 0.0004145798945207134, "loss": 1.2198, "step": 85370 }, { "epoch": 3.43, "learning_rate": 0.0004145698297032892, "loss": 1.0275, "step": 85380 }, { "epoch": 3.43, "learning_rate": 0.000414559764885865, "loss": 1.128, "step": 85390 }, { "epoch": 3.43, "learning_rate": 0.0004145497000684408, "loss": 1.1518, "step": 85400 }, { "epoch": 3.44, "learning_rate": 0.00041453963525101657, "loss": 0.9812, "step": 85410 }, { "epoch": 3.44, "learning_rate": 0.00041452957043359236, "loss": 1.1148, "step": 85420 }, { "epoch": 3.44, "learning_rate": 0.00041451950561616815, "loss": 1.159, "step": 85430 }, { "epoch": 3.44, "learning_rate": 0.00041450944079874394, "loss": 1.1771, "step": 85440 }, { "epoch": 3.44, "learning_rate": 0.00041449937598131973, "loss": 1.1305, "step": 85450 }, { "epoch": 3.44, "learning_rate": 0.00041448931116389547, "loss": 1.2301, "step": 85460 }, { "epoch": 3.44, "learning_rate": 0.00041447924634647126, "loss": 1.2791, "step": 85470 }, { "epoch": 3.44, "learning_rate": 0.00041446918152904705, "loss": 1.1565, "step": 85480 }, { "epoch": 3.44, "learning_rate": 0.00041445911671162284, "loss": 1.0009, "step": 85490 }, { "epoch": 3.44, "learning_rate": 0.00041444905189419863, "loss": 1.2935, "step": 85500 }, { "epoch": 3.44, "learning_rate": 0.0004144389870767744, "loss": 0.9774, "step": 85510 }, { "epoch": 3.44, "learning_rate": 0.00041442892225935027, "loss": 1.1755, "step": 85520 }, { "epoch": 3.44, "learning_rate": 0.00041441885744192606, "loss": 1.1878, "step": 85530 }, { "epoch": 3.44, "learning_rate": 0.0004144087926245018, "loss": 1.2342, "step": 85540 }, { "epoch": 3.44, "learning_rate": 0.0004143987278070776, "loss": 1.0158, "step": 85550 }, { "epoch": 3.44, "learning_rate": 0.0004143886629896534, "loss": 0.9633, "step": 85560 }, { "epoch": 3.44, "learning_rate": 0.00041437859817222917, "loss": 1.1693, "step": 85570 }, { "epoch": 3.44, "learning_rate": 0.00041436853335480496, "loss": 1.0919, "step": 85580 }, { "epoch": 3.44, "learning_rate": 0.00041435846853738075, "loss": 1.0894, "step": 85590 }, { "epoch": 3.44, "learning_rate": 0.00041434840371995654, "loss": 1.2067, "step": 85600 }, { "epoch": 3.44, "learning_rate": 0.00041433833890253233, "loss": 1.2006, "step": 85610 }, { "epoch": 3.44, "learning_rate": 0.00041432827408510807, "loss": 1.4941, "step": 85620 }, { "epoch": 3.44, "learning_rate": 0.00041431820926768386, "loss": 1.2714, "step": 85630 }, { "epoch": 3.44, "learning_rate": 0.00041430814445025965, "loss": 1.4357, "step": 85640 }, { "epoch": 3.44, "learning_rate": 0.00041429807963283544, "loss": 1.2303, "step": 85650 }, { "epoch": 3.45, "learning_rate": 0.00041428801481541123, "loss": 1.0659, "step": 85660 }, { "epoch": 3.45, "learning_rate": 0.0004142779499979871, "loss": 1.1045, "step": 85670 }, { "epoch": 3.45, "learning_rate": 0.00041426788518056286, "loss": 1.3111, "step": 85680 }, { "epoch": 3.45, "learning_rate": 0.00041425782036313866, "loss": 0.996, "step": 85690 }, { "epoch": 3.45, "learning_rate": 0.00041424775554571445, "loss": 1.0538, "step": 85700 }, { "epoch": 3.45, "learning_rate": 0.0004142376907282902, "loss": 1.0171, "step": 85710 }, { "epoch": 3.45, "learning_rate": 0.000414227625910866, "loss": 0.9297, "step": 85720 }, { "epoch": 3.45, "learning_rate": 0.00041421756109344176, "loss": 1.0591, "step": 85730 }, { "epoch": 3.45, "learning_rate": 0.00041420749627601755, "loss": 1.1204, "step": 85740 }, { "epoch": 3.45, "learning_rate": 0.00041419743145859335, "loss": 1.2092, "step": 85750 }, { "epoch": 3.45, "learning_rate": 0.00041418736664116914, "loss": 1.1569, "step": 85760 }, { "epoch": 3.45, "learning_rate": 0.00041417730182374493, "loss": 1.2601, "step": 85770 }, { "epoch": 3.45, "learning_rate": 0.0004141672370063207, "loss": 1.0448, "step": 85780 }, { "epoch": 3.45, "learning_rate": 0.00041415717218889645, "loss": 0.9072, "step": 85790 }, { "epoch": 3.45, "learning_rate": 0.00041414710737147225, "loss": 0.9552, "step": 85800 }, { "epoch": 3.45, "learning_rate": 0.0004141370425540481, "loss": 1.3518, "step": 85810 }, { "epoch": 3.45, "learning_rate": 0.0004141269777366239, "loss": 1.1044, "step": 85820 }, { "epoch": 3.45, "learning_rate": 0.00041411691291919967, "loss": 0.9978, "step": 85830 }, { "epoch": 3.45, "learning_rate": 0.00041410684810177546, "loss": 1.0562, "step": 85840 }, { "epoch": 3.45, "learning_rate": 0.00041409678328435125, "loss": 1.1876, "step": 85850 }, { "epoch": 3.45, "learning_rate": 0.00041408671846692704, "loss": 1.0939, "step": 85860 }, { "epoch": 3.45, "learning_rate": 0.00041407665364950284, "loss": 1.2207, "step": 85870 }, { "epoch": 3.45, "learning_rate": 0.00041406658883207857, "loss": 1.0119, "step": 85880 }, { "epoch": 3.45, "learning_rate": 0.00041405652401465436, "loss": 0.829, "step": 85890 }, { "epoch": 3.45, "learning_rate": 0.00041404645919723015, "loss": 1.119, "step": 85900 }, { "epoch": 3.46, "learning_rate": 0.00041403639437980594, "loss": 1.0741, "step": 85910 }, { "epoch": 3.46, "learning_rate": 0.00041402632956238173, "loss": 0.9863, "step": 85920 }, { "epoch": 3.46, "learning_rate": 0.0004140162647449575, "loss": 1.084, "step": 85930 }, { "epoch": 3.46, "learning_rate": 0.0004140061999275333, "loss": 1.0616, "step": 85940 }, { "epoch": 3.46, "learning_rate": 0.0004139961351101091, "loss": 1.1786, "step": 85950 }, { "epoch": 3.46, "learning_rate": 0.0004139860702926849, "loss": 1.0318, "step": 85960 }, { "epoch": 3.46, "learning_rate": 0.0004139760054752607, "loss": 1.1434, "step": 85970 }, { "epoch": 3.46, "learning_rate": 0.0004139659406578365, "loss": 1.0419, "step": 85980 }, { "epoch": 3.46, "learning_rate": 0.00041395587584041227, "loss": 1.3471, "step": 85990 }, { "epoch": 3.46, "learning_rate": 0.00041394581102298806, "loss": 0.9317, "step": 86000 }, { "epoch": 3.46, "learning_rate": 0.00041393574620556385, "loss": 1.0092, "step": 86010 }, { "epoch": 3.46, "learning_rate": 0.00041392568138813964, "loss": 1.0775, "step": 86020 }, { "epoch": 3.46, "learning_rate": 0.00041391561657071543, "loss": 1.0473, "step": 86030 }, { "epoch": 3.46, "learning_rate": 0.0004139055517532912, "loss": 0.9663, "step": 86040 }, { "epoch": 3.46, "learning_rate": 0.00041389548693586696, "loss": 1.3082, "step": 86050 }, { "epoch": 3.46, "learning_rate": 0.00041388542211844275, "loss": 1.1611, "step": 86060 }, { "epoch": 3.46, "learning_rate": 0.00041387535730101854, "loss": 1.1107, "step": 86070 }, { "epoch": 3.46, "learning_rate": 0.00041386529248359433, "loss": 1.099, "step": 86080 }, { "epoch": 3.46, "learning_rate": 0.0004138552276661701, "loss": 1.2737, "step": 86090 }, { "epoch": 3.46, "learning_rate": 0.00041384516284874597, "loss": 1.1044, "step": 86100 }, { "epoch": 3.46, "learning_rate": 0.00041383509803132176, "loss": 1.0941, "step": 86110 }, { "epoch": 3.46, "learning_rate": 0.00041382503321389755, "loss": 0.9275, "step": 86120 }, { "epoch": 3.46, "learning_rate": 0.0004138149683964733, "loss": 0.9989, "step": 86130 }, { "epoch": 3.46, "learning_rate": 0.0004138049035790491, "loss": 1.0391, "step": 86140 }, { "epoch": 3.46, "learning_rate": 0.00041379483876162487, "loss": 1.2661, "step": 86150 }, { "epoch": 3.47, "learning_rate": 0.00041378477394420066, "loss": 1.1069, "step": 86160 }, { "epoch": 3.47, "learning_rate": 0.00041377470912677645, "loss": 1.2134, "step": 86170 }, { "epoch": 3.47, "learning_rate": 0.00041376464430935224, "loss": 1.0825, "step": 86180 }, { "epoch": 3.47, "learning_rate": 0.00041375457949192803, "loss": 1.1256, "step": 86190 }, { "epoch": 3.47, "learning_rate": 0.0004137445146745038, "loss": 1.1349, "step": 86200 }, { "epoch": 3.47, "learning_rate": 0.0004137344498570796, "loss": 1.1163, "step": 86210 }, { "epoch": 3.47, "learning_rate": 0.00041372438503965535, "loss": 1.1927, "step": 86220 }, { "epoch": 3.47, "learning_rate": 0.00041371432022223114, "loss": 1.3521, "step": 86230 }, { "epoch": 3.47, "learning_rate": 0.00041370425540480693, "loss": 1.2978, "step": 86240 }, { "epoch": 3.47, "learning_rate": 0.0004136941905873828, "loss": 1.3289, "step": 86250 }, { "epoch": 3.47, "learning_rate": 0.00041368412576995857, "loss": 1.2153, "step": 86260 }, { "epoch": 3.47, "learning_rate": 0.00041367406095253436, "loss": 1.2769, "step": 86270 }, { "epoch": 3.47, "learning_rate": 0.00041366399613511015, "loss": 1.1275, "step": 86280 }, { "epoch": 3.47, "learning_rate": 0.00041365393131768594, "loss": 1.1301, "step": 86290 }, { "epoch": 3.47, "learning_rate": 0.0004136438665002617, "loss": 1.2108, "step": 86300 }, { "epoch": 3.47, "learning_rate": 0.00041363380168283747, "loss": 1.0722, "step": 86310 }, { "epoch": 3.47, "learning_rate": 0.00041362373686541326, "loss": 0.9911, "step": 86320 }, { "epoch": 3.47, "learning_rate": 0.00041361367204798905, "loss": 1.085, "step": 86330 }, { "epoch": 3.47, "learning_rate": 0.00041360360723056484, "loss": 1.0983, "step": 86340 }, { "epoch": 3.47, "learning_rate": 0.00041359354241314063, "loss": 1.1645, "step": 86350 }, { "epoch": 3.47, "learning_rate": 0.0004135834775957164, "loss": 1.0984, "step": 86360 }, { "epoch": 3.47, "learning_rate": 0.0004135734127782922, "loss": 1.1837, "step": 86370 }, { "epoch": 3.47, "learning_rate": 0.00041356334796086795, "loss": 1.12, "step": 86380 }, { "epoch": 3.47, "learning_rate": 0.00041355328314344374, "loss": 1.1938, "step": 86390 }, { "epoch": 3.47, "learning_rate": 0.0004135432183260196, "loss": 1.2088, "step": 86400 }, { "epoch": 3.48, "learning_rate": 0.0004135331535085954, "loss": 1.2065, "step": 86410 }, { "epoch": 3.48, "learning_rate": 0.00041352308869117117, "loss": 1.1952, "step": 86420 }, { "epoch": 3.48, "learning_rate": 0.00041351302387374696, "loss": 1.0632, "step": 86430 }, { "epoch": 3.48, "learning_rate": 0.00041350295905632275, "loss": 1.1891, "step": 86440 }, { "epoch": 3.48, "learning_rate": 0.00041349289423889854, "loss": 1.1666, "step": 86450 }, { "epoch": 3.48, "learning_rate": 0.00041348282942147433, "loss": 1.1859, "step": 86460 }, { "epoch": 3.48, "learning_rate": 0.00041347276460405007, "loss": 1.166, "step": 86470 }, { "epoch": 3.48, "learning_rate": 0.00041346269978662586, "loss": 1.0365, "step": 86480 }, { "epoch": 3.48, "learning_rate": 0.00041345263496920165, "loss": 0.9961, "step": 86490 }, { "epoch": 3.48, "learning_rate": 0.00041344257015177744, "loss": 1.0053, "step": 86500 }, { "epoch": 3.48, "learning_rate": 0.00041343250533435323, "loss": 1.0179, "step": 86510 }, { "epoch": 3.48, "learning_rate": 0.000413422440516929, "loss": 1.272, "step": 86520 }, { "epoch": 3.48, "learning_rate": 0.0004134123756995048, "loss": 1.3371, "step": 86530 }, { "epoch": 3.48, "learning_rate": 0.00041340231088208065, "loss": 1.0428, "step": 86540 }, { "epoch": 3.48, "learning_rate": 0.00041339224606465645, "loss": 0.9133, "step": 86550 }, { "epoch": 3.48, "learning_rate": 0.0004133821812472322, "loss": 1.2503, "step": 86560 }, { "epoch": 3.48, "learning_rate": 0.000413372116429808, "loss": 0.9726, "step": 86570 }, { "epoch": 3.48, "learning_rate": 0.00041336205161238376, "loss": 1.3065, "step": 86580 }, { "epoch": 3.48, "learning_rate": 0.00041335198679495955, "loss": 1.034, "step": 86590 }, { "epoch": 3.48, "learning_rate": 0.00041334192197753535, "loss": 1.0573, "step": 86600 }, { "epoch": 3.48, "learning_rate": 0.00041333185716011114, "loss": 1.2827, "step": 86610 }, { "epoch": 3.48, "learning_rate": 0.0004133217923426869, "loss": 0.9462, "step": 86620 }, { "epoch": 3.48, "learning_rate": 0.0004133117275252627, "loss": 1.2828, "step": 86630 }, { "epoch": 3.48, "learning_rate": 0.00041330166270783845, "loss": 1.1406, "step": 86640 }, { "epoch": 3.48, "learning_rate": 0.00041329159789041425, "loss": 1.1282, "step": 86650 }, { "epoch": 3.49, "learning_rate": 0.00041328153307299004, "loss": 1.2309, "step": 86660 }, { "epoch": 3.49, "learning_rate": 0.0004132714682555658, "loss": 1.1069, "step": 86670 }, { "epoch": 3.49, "learning_rate": 0.0004132614034381416, "loss": 1.0198, "step": 86680 }, { "epoch": 3.49, "learning_rate": 0.00041325133862071746, "loss": 1.1641, "step": 86690 }, { "epoch": 3.49, "learning_rate": 0.00041324127380329325, "loss": 1.1669, "step": 86700 }, { "epoch": 3.49, "learning_rate": 0.00041323120898586904, "loss": 1.2657, "step": 86710 }, { "epoch": 3.49, "learning_rate": 0.0004132211441684448, "loss": 1.2421, "step": 86720 }, { "epoch": 3.49, "learning_rate": 0.00041321107935102057, "loss": 1.0236, "step": 86730 }, { "epoch": 3.49, "learning_rate": 0.00041320101453359636, "loss": 1.2481, "step": 86740 }, { "epoch": 3.49, "learning_rate": 0.00041319094971617215, "loss": 1.2699, "step": 86750 }, { "epoch": 3.49, "learning_rate": 0.00041318088489874794, "loss": 0.9828, "step": 86760 }, { "epoch": 3.49, "learning_rate": 0.00041317082008132373, "loss": 1.0708, "step": 86770 }, { "epoch": 3.49, "learning_rate": 0.0004131607552638995, "loss": 1.3159, "step": 86780 }, { "epoch": 3.49, "learning_rate": 0.0004131506904464753, "loss": 1.1772, "step": 86790 }, { "epoch": 3.49, "learning_rate": 0.0004131406256290511, "loss": 0.9394, "step": 86800 }, { "epoch": 3.49, "learning_rate": 0.00041313056081162684, "loss": 1.1669, "step": 86810 }, { "epoch": 3.49, "learning_rate": 0.00041312049599420263, "loss": 1.0025, "step": 86820 }, { "epoch": 3.49, "learning_rate": 0.0004131104311767785, "loss": 0.9628, "step": 86830 }, { "epoch": 3.49, "learning_rate": 0.00041310036635935427, "loss": 0.9959, "step": 86840 }, { "epoch": 3.49, "learning_rate": 0.00041309030154193006, "loss": 0.998, "step": 86850 }, { "epoch": 3.49, "learning_rate": 0.00041308023672450585, "loss": 1.0557, "step": 86860 }, { "epoch": 3.49, "learning_rate": 0.00041307017190708164, "loss": 0.9866, "step": 86870 }, { "epoch": 3.49, "learning_rate": 0.00041306010708965743, "loss": 0.9897, "step": 86880 }, { "epoch": 3.49, "learning_rate": 0.00041305004227223317, "loss": 1.1898, "step": 86890 }, { "epoch": 3.49, "learning_rate": 0.00041303997745480896, "loss": 1.2319, "step": 86900 }, { "epoch": 3.5, "learning_rate": 0.00041302991263738475, "loss": 1.0914, "step": 86910 }, { "epoch": 3.5, "learning_rate": 0.00041301984781996054, "loss": 1.3158, "step": 86920 }, { "epoch": 3.5, "learning_rate": 0.00041300978300253633, "loss": 1.0619, "step": 86930 }, { "epoch": 3.5, "learning_rate": 0.0004129997181851121, "loss": 0.9767, "step": 86940 }, { "epoch": 3.5, "learning_rate": 0.0004129896533676879, "loss": 1.2695, "step": 86950 }, { "epoch": 3.5, "learning_rate": 0.0004129795885502637, "loss": 1.3431, "step": 86960 }, { "epoch": 3.5, "learning_rate": 0.0004129695237328395, "loss": 1.0842, "step": 86970 }, { "epoch": 3.5, "learning_rate": 0.0004129594589154153, "loss": 1.0671, "step": 86980 }, { "epoch": 3.5, "learning_rate": 0.0004129493940979911, "loss": 0.9847, "step": 86990 }, { "epoch": 3.5, "learning_rate": 0.00041293932928056687, "loss": 1.0229, "step": 87000 }, { "epoch": 3.5, "learning_rate": 0.00041292926446314266, "loss": 1.1706, "step": 87010 }, { "epoch": 3.5, "learning_rate": 0.00041291919964571845, "loss": 1.2821, "step": 87020 }, { "epoch": 3.5, "learning_rate": 0.00041290913482829424, "loss": 1.1943, "step": 87030 }, { "epoch": 3.5, "learning_rate": 0.00041289907001087003, "loss": 0.9773, "step": 87040 }, { "epoch": 3.5, "learning_rate": 0.0004128890051934458, "loss": 1.0721, "step": 87050 }, { "epoch": 3.5, "learning_rate": 0.00041287894037602156, "loss": 1.0053, "step": 87060 }, { "epoch": 3.5, "learning_rate": 0.00041286887555859735, "loss": 1.2669, "step": 87070 }, { "epoch": 3.5, "learning_rate": 0.00041285881074117314, "loss": 1.0044, "step": 87080 }, { "epoch": 3.5, "learning_rate": 0.00041284874592374893, "loss": 1.0781, "step": 87090 }, { "epoch": 3.5, "learning_rate": 0.0004128386811063247, "loss": 1.301, "step": 87100 }, { "epoch": 3.5, "learning_rate": 0.0004128286162889005, "loss": 1.1212, "step": 87110 }, { "epoch": 3.5, "learning_rate": 0.0004128185514714763, "loss": 1.212, "step": 87120 }, { "epoch": 3.5, "learning_rate": 0.00041280848665405215, "loss": 1.0372, "step": 87130 }, { "epoch": 3.5, "learning_rate": 0.00041279842183662794, "loss": 1.0914, "step": 87140 }, { "epoch": 3.51, "learning_rate": 0.0004127883570192037, "loss": 1.0368, "step": 87150 }, { "epoch": 3.51, "learning_rate": 0.00041277829220177947, "loss": 0.9901, "step": 87160 }, { "epoch": 3.51, "learning_rate": 0.00041276822738435526, "loss": 1.0449, "step": 87170 }, { "epoch": 3.51, "learning_rate": 0.00041275816256693105, "loss": 1.092, "step": 87180 }, { "epoch": 3.51, "learning_rate": 0.00041274809774950684, "loss": 1.0192, "step": 87190 }, { "epoch": 3.51, "learning_rate": 0.00041273803293208263, "loss": 1.0027, "step": 87200 }, { "epoch": 3.51, "learning_rate": 0.0004127279681146584, "loss": 1.1855, "step": 87210 }, { "epoch": 3.51, "learning_rate": 0.0004127179032972342, "loss": 1.4183, "step": 87220 }, { "epoch": 3.51, "learning_rate": 0.00041270783847980995, "loss": 1.3935, "step": 87230 }, { "epoch": 3.51, "learning_rate": 0.00041269777366238574, "loss": 1.2199, "step": 87240 }, { "epoch": 3.51, "learning_rate": 0.00041268770884496153, "loss": 1.3338, "step": 87250 }, { "epoch": 3.51, "learning_rate": 0.0004126776440275373, "loss": 1.0597, "step": 87260 }, { "epoch": 3.51, "learning_rate": 0.00041266757921011317, "loss": 1.3211, "step": 87270 }, { "epoch": 3.51, "learning_rate": 0.00041265751439268896, "loss": 0.9597, "step": 87280 }, { "epoch": 3.51, "learning_rate": 0.00041264744957526475, "loss": 1.0405, "step": 87290 }, { "epoch": 3.51, "learning_rate": 0.00041263738475784054, "loss": 1.0046, "step": 87300 }, { "epoch": 3.51, "learning_rate": 0.00041262731994041633, "loss": 1.132, "step": 87310 }, { "epoch": 3.51, "learning_rate": 0.00041261725512299206, "loss": 1.0486, "step": 87320 }, { "epoch": 3.51, "learning_rate": 0.00041260719030556786, "loss": 1.2908, "step": 87330 }, { "epoch": 3.51, "learning_rate": 0.00041259712548814365, "loss": 1.0979, "step": 87340 }, { "epoch": 3.51, "learning_rate": 0.00041258706067071944, "loss": 0.9397, "step": 87350 }, { "epoch": 3.51, "learning_rate": 0.00041257699585329523, "loss": 1.1962, "step": 87360 }, { "epoch": 3.51, "learning_rate": 0.000412566931035871, "loss": 1.0583, "step": 87370 }, { "epoch": 3.51, "learning_rate": 0.0004125568662184468, "loss": 0.9144, "step": 87380 }, { "epoch": 3.51, "learning_rate": 0.0004125468014010226, "loss": 1.1035, "step": 87390 }, { "epoch": 3.52, "learning_rate": 0.00041253673658359834, "loss": 1.1159, "step": 87400 }, { "epoch": 3.52, "learning_rate": 0.00041252667176617413, "loss": 1.1367, "step": 87410 }, { "epoch": 3.52, "learning_rate": 0.00041251660694874997, "loss": 1.0237, "step": 87420 }, { "epoch": 3.52, "learning_rate": 0.00041250654213132576, "loss": 1.3219, "step": 87430 }, { "epoch": 3.52, "learning_rate": 0.00041249647731390155, "loss": 1.01, "step": 87440 }, { "epoch": 3.52, "learning_rate": 0.00041248641249647735, "loss": 1.0704, "step": 87450 }, { "epoch": 3.52, "learning_rate": 0.00041247634767905314, "loss": 1.1408, "step": 87460 }, { "epoch": 3.52, "learning_rate": 0.0004124662828616289, "loss": 1.1085, "step": 87470 }, { "epoch": 3.52, "learning_rate": 0.00041245621804420466, "loss": 0.9397, "step": 87480 }, { "epoch": 3.52, "learning_rate": 0.00041244615322678045, "loss": 1.1712, "step": 87490 }, { "epoch": 3.52, "learning_rate": 0.00041243608840935624, "loss": 1.4227, "step": 87500 }, { "epoch": 3.52, "learning_rate": 0.00041242602359193204, "loss": 1.2987, "step": 87510 }, { "epoch": 3.52, "learning_rate": 0.0004124159587745078, "loss": 1.0807, "step": 87520 }, { "epoch": 3.52, "learning_rate": 0.0004124058939570836, "loss": 0.9421, "step": 87530 }, { "epoch": 3.52, "learning_rate": 0.0004123958291396594, "loss": 0.9888, "step": 87540 }, { "epoch": 3.52, "learning_rate": 0.0004123857643222352, "loss": 1.0967, "step": 87550 }, { "epoch": 3.52, "learning_rate": 0.000412375699504811, "loss": 1.0408, "step": 87560 }, { "epoch": 3.52, "learning_rate": 0.0004123656346873868, "loss": 1.2366, "step": 87570 }, { "epoch": 3.52, "learning_rate": 0.00041235556986996257, "loss": 1.2543, "step": 87580 }, { "epoch": 3.52, "learning_rate": 0.00041234550505253836, "loss": 1.0783, "step": 87590 }, { "epoch": 3.52, "learning_rate": 0.00041233544023511415, "loss": 1.0506, "step": 87600 }, { "epoch": 3.52, "learning_rate": 0.00041232537541768994, "loss": 1.2155, "step": 87610 }, { "epoch": 3.52, "learning_rate": 0.00041231531060026573, "loss": 1.3121, "step": 87620 }, { "epoch": 3.52, "learning_rate": 0.0004123052457828415, "loss": 1.1687, "step": 87630 }, { "epoch": 3.52, "learning_rate": 0.0004122951809654173, "loss": 1.2397, "step": 87640 }, { "epoch": 3.53, "learning_rate": 0.00041228511614799305, "loss": 1.2015, "step": 87650 }, { "epoch": 3.53, "learning_rate": 0.00041227505133056884, "loss": 1.0277, "step": 87660 }, { "epoch": 3.53, "learning_rate": 0.00041226498651314463, "loss": 1.321, "step": 87670 }, { "epoch": 3.53, "learning_rate": 0.0004122549216957204, "loss": 1.2026, "step": 87680 }, { "epoch": 3.53, "learning_rate": 0.0004122448568782962, "loss": 1.1959, "step": 87690 }, { "epoch": 3.53, "learning_rate": 0.000412234792060872, "loss": 1.3199, "step": 87700 }, { "epoch": 3.53, "learning_rate": 0.00041222472724344785, "loss": 1.0821, "step": 87710 }, { "epoch": 3.53, "learning_rate": 0.00041221466242602364, "loss": 0.9299, "step": 87720 }, { "epoch": 3.53, "learning_rate": 0.00041220459760859943, "loss": 1.2725, "step": 87730 }, { "epoch": 3.53, "learning_rate": 0.00041219453279117517, "loss": 1.0898, "step": 87740 }, { "epoch": 3.53, "learning_rate": 0.00041218446797375096, "loss": 1.1487, "step": 87750 }, { "epoch": 3.53, "learning_rate": 0.00041217440315632675, "loss": 1.0023, "step": 87760 }, { "epoch": 3.53, "learning_rate": 0.00041216433833890254, "loss": 1.2372, "step": 87770 }, { "epoch": 3.53, "learning_rate": 0.00041215427352147833, "loss": 0.922, "step": 87780 }, { "epoch": 3.53, "learning_rate": 0.0004121442087040541, "loss": 1.0401, "step": 87790 }, { "epoch": 3.53, "learning_rate": 0.0004121341438866299, "loss": 1.2543, "step": 87800 }, { "epoch": 3.53, "learning_rate": 0.0004121240790692057, "loss": 1.0017, "step": 87810 }, { "epoch": 3.53, "learning_rate": 0.00041211401425178144, "loss": 1.1221, "step": 87820 }, { "epoch": 3.53, "learning_rate": 0.00041210394943435723, "loss": 0.955, "step": 87830 }, { "epoch": 3.53, "learning_rate": 0.000412093884616933, "loss": 0.9549, "step": 87840 }, { "epoch": 3.53, "learning_rate": 0.0004120838197995088, "loss": 1.1781, "step": 87850 }, { "epoch": 3.53, "learning_rate": 0.00041207375498208466, "loss": 1.0776, "step": 87860 }, { "epoch": 3.53, "learning_rate": 0.00041206369016466045, "loss": 1.166, "step": 87870 }, { "epoch": 3.53, "learning_rate": 0.00041205362534723624, "loss": 0.9552, "step": 87880 }, { "epoch": 3.53, "learning_rate": 0.00041204356052981203, "loss": 1.138, "step": 87890 }, { "epoch": 3.54, "learning_rate": 0.0004120334957123878, "loss": 1.1085, "step": 87900 }, { "epoch": 3.54, "learning_rate": 0.00041202343089496356, "loss": 1.0506, "step": 87910 }, { "epoch": 3.54, "learning_rate": 0.00041201336607753935, "loss": 0.9072, "step": 87920 }, { "epoch": 3.54, "learning_rate": 0.00041200330126011514, "loss": 1.115, "step": 87930 }, { "epoch": 3.54, "learning_rate": 0.00041199323644269093, "loss": 1.1354, "step": 87940 }, { "epoch": 3.54, "learning_rate": 0.0004119831716252667, "loss": 1.1887, "step": 87950 }, { "epoch": 3.54, "learning_rate": 0.0004119731068078425, "loss": 1.2308, "step": 87960 }, { "epoch": 3.54, "learning_rate": 0.0004119630419904183, "loss": 1.2187, "step": 87970 }, { "epoch": 3.54, "learning_rate": 0.0004119529771729941, "loss": 1.4444, "step": 87980 }, { "epoch": 3.54, "learning_rate": 0.00041194291235556983, "loss": 1.1357, "step": 87990 }, { "epoch": 3.54, "learning_rate": 0.0004119328475381457, "loss": 1.2327, "step": 88000 }, { "epoch": 3.54, "learning_rate": 0.00041192278272072147, "loss": 1.5093, "step": 88010 }, { "epoch": 3.54, "learning_rate": 0.00041191271790329726, "loss": 1.1067, "step": 88020 }, { "epoch": 3.54, "learning_rate": 0.00041190265308587305, "loss": 1.0693, "step": 88030 }, { "epoch": 3.54, "learning_rate": 0.00041189258826844884, "loss": 1.0535, "step": 88040 }, { "epoch": 3.54, "learning_rate": 0.00041188252345102463, "loss": 1.2825, "step": 88050 }, { "epoch": 3.54, "learning_rate": 0.0004118724586336004, "loss": 1.1148, "step": 88060 }, { "epoch": 3.54, "learning_rate": 0.0004118623938161762, "loss": 1.3019, "step": 88070 }, { "epoch": 3.54, "learning_rate": 0.00041185232899875195, "loss": 0.9748, "step": 88080 }, { "epoch": 3.54, "learning_rate": 0.00041184226418132774, "loss": 1.0219, "step": 88090 }, { "epoch": 3.54, "learning_rate": 0.00041183219936390353, "loss": 1.0817, "step": 88100 }, { "epoch": 3.54, "learning_rate": 0.0004118221345464793, "loss": 1.0292, "step": 88110 }, { "epoch": 3.54, "learning_rate": 0.0004118120697290551, "loss": 1.0549, "step": 88120 }, { "epoch": 3.54, "learning_rate": 0.0004118020049116309, "loss": 1.2548, "step": 88130 }, { "epoch": 3.54, "learning_rate": 0.0004117919400942067, "loss": 1.3831, "step": 88140 }, { "epoch": 3.55, "learning_rate": 0.00041178187527678254, "loss": 1.1576, "step": 88150 }, { "epoch": 3.55, "learning_rate": 0.0004117718104593583, "loss": 1.1668, "step": 88160 }, { "epoch": 3.55, "learning_rate": 0.00041176174564193406, "loss": 1.0435, "step": 88170 }, { "epoch": 3.55, "learning_rate": 0.00041175168082450986, "loss": 1.0523, "step": 88180 }, { "epoch": 3.55, "learning_rate": 0.00041174161600708565, "loss": 1.1376, "step": 88190 }, { "epoch": 3.55, "learning_rate": 0.00041173155118966144, "loss": 1.1739, "step": 88200 }, { "epoch": 3.55, "learning_rate": 0.00041172148637223723, "loss": 1.0217, "step": 88210 }, { "epoch": 3.55, "learning_rate": 0.000411711421554813, "loss": 1.1606, "step": 88220 }, { "epoch": 3.55, "learning_rate": 0.0004117013567373888, "loss": 1.2156, "step": 88230 }, { "epoch": 3.55, "learning_rate": 0.00041169129191996455, "loss": 1.329, "step": 88240 }, { "epoch": 3.55, "learning_rate": 0.00041168122710254034, "loss": 1.0156, "step": 88250 }, { "epoch": 3.55, "learning_rate": 0.00041167116228511613, "loss": 1.3083, "step": 88260 }, { "epoch": 3.55, "learning_rate": 0.0004116610974676919, "loss": 1.2931, "step": 88270 }, { "epoch": 3.55, "learning_rate": 0.0004116510326502677, "loss": 1.0886, "step": 88280 }, { "epoch": 3.55, "learning_rate": 0.0004116409678328435, "loss": 1.4007, "step": 88290 }, { "epoch": 3.55, "learning_rate": 0.00041163090301541934, "loss": 1.0859, "step": 88300 }, { "epoch": 3.55, "learning_rate": 0.00041162083819799514, "loss": 1.0498, "step": 88310 }, { "epoch": 3.55, "learning_rate": 0.0004116107733805709, "loss": 1.1441, "step": 88320 }, { "epoch": 3.55, "learning_rate": 0.00041160070856314666, "loss": 1.0758, "step": 88330 }, { "epoch": 3.55, "learning_rate": 0.00041159064374572245, "loss": 1.016, "step": 88340 }, { "epoch": 3.55, "learning_rate": 0.00041158057892829824, "loss": 1.261, "step": 88350 }, { "epoch": 3.55, "learning_rate": 0.00041157051411087404, "loss": 1.2224, "step": 88360 }, { "epoch": 3.55, "learning_rate": 0.0004115604492934498, "loss": 1.0966, "step": 88370 }, { "epoch": 3.55, "learning_rate": 0.0004115503844760256, "loss": 1.2405, "step": 88380 }, { "epoch": 3.55, "learning_rate": 0.0004115403196586014, "loss": 1.135, "step": 88390 }, { "epoch": 3.56, "learning_rate": 0.0004115302548411772, "loss": 1.1959, "step": 88400 }, { "epoch": 3.56, "learning_rate": 0.00041152019002375294, "loss": 0.9924, "step": 88410 }, { "epoch": 3.56, "learning_rate": 0.0004115101252063287, "loss": 1.0111, "step": 88420 }, { "epoch": 3.56, "learning_rate": 0.0004115000603889045, "loss": 1.207, "step": 88430 }, { "epoch": 3.56, "learning_rate": 0.00041148999557148036, "loss": 1.0073, "step": 88440 }, { "epoch": 3.56, "learning_rate": 0.00041147993075405615, "loss": 1.2029, "step": 88450 }, { "epoch": 3.56, "learning_rate": 0.00041146986593663194, "loss": 1.0852, "step": 88460 }, { "epoch": 3.56, "learning_rate": 0.00041145980111920773, "loss": 1.1984, "step": 88470 }, { "epoch": 3.56, "learning_rate": 0.0004114497363017835, "loss": 0.8981, "step": 88480 }, { "epoch": 3.56, "learning_rate": 0.0004114396714843593, "loss": 1.1171, "step": 88490 }, { "epoch": 3.56, "learning_rate": 0.00041142960666693505, "loss": 1.2858, "step": 88500 }, { "epoch": 3.56, "learning_rate": 0.00041141954184951084, "loss": 1.0217, "step": 88510 }, { "epoch": 3.56, "learning_rate": 0.00041140947703208663, "loss": 1.16, "step": 88520 }, { "epoch": 3.56, "learning_rate": 0.0004113994122146624, "loss": 1.4383, "step": 88530 }, { "epoch": 3.56, "learning_rate": 0.0004113893473972382, "loss": 0.9437, "step": 88540 }, { "epoch": 3.56, "learning_rate": 0.000411379282579814, "loss": 1.1859, "step": 88550 }, { "epoch": 3.56, "learning_rate": 0.0004113692177623898, "loss": 1.1397, "step": 88560 }, { "epoch": 3.56, "learning_rate": 0.0004113591529449656, "loss": 1.1199, "step": 88570 }, { "epoch": 3.56, "learning_rate": 0.0004113490881275413, "loss": 1.1102, "step": 88580 }, { "epoch": 3.56, "learning_rate": 0.00041133902331011717, "loss": 1.2724, "step": 88590 }, { "epoch": 3.56, "learning_rate": 0.00041132895849269296, "loss": 1.2631, "step": 88600 }, { "epoch": 3.56, "learning_rate": 0.00041131889367526875, "loss": 1.1479, "step": 88610 }, { "epoch": 3.56, "learning_rate": 0.00041130882885784454, "loss": 1.1144, "step": 88620 }, { "epoch": 3.56, "learning_rate": 0.00041129876404042033, "loss": 0.9615, "step": 88630 }, { "epoch": 3.56, "learning_rate": 0.0004112886992229961, "loss": 1.264, "step": 88640 }, { "epoch": 3.57, "learning_rate": 0.0004112786344055719, "loss": 1.0121, "step": 88650 }, { "epoch": 3.57, "learning_rate": 0.0004112685695881477, "loss": 1.4637, "step": 88660 }, { "epoch": 3.57, "learning_rate": 0.00041125850477072344, "loss": 1.032, "step": 88670 }, { "epoch": 3.57, "learning_rate": 0.00041124843995329923, "loss": 1.2419, "step": 88680 }, { "epoch": 3.57, "learning_rate": 0.000411238375135875, "loss": 1.1068, "step": 88690 }, { "epoch": 3.57, "learning_rate": 0.0004112283103184508, "loss": 1.2024, "step": 88700 }, { "epoch": 3.57, "learning_rate": 0.0004112182455010266, "loss": 1.1842, "step": 88710 }, { "epoch": 3.57, "learning_rate": 0.0004112081806836024, "loss": 1.2911, "step": 88720 }, { "epoch": 3.57, "learning_rate": 0.0004111981158661782, "loss": 1.1272, "step": 88730 }, { "epoch": 3.57, "learning_rate": 0.00041118805104875403, "loss": 1.0912, "step": 88740 }, { "epoch": 3.57, "learning_rate": 0.00041117798623132977, "loss": 0.9929, "step": 88750 }, { "epoch": 3.57, "learning_rate": 0.00041116792141390556, "loss": 1.1384, "step": 88760 }, { "epoch": 3.57, "learning_rate": 0.00041115785659648135, "loss": 0.9775, "step": 88770 }, { "epoch": 3.57, "learning_rate": 0.00041114779177905714, "loss": 1.1497, "step": 88780 }, { "epoch": 3.57, "learning_rate": 0.00041113772696163293, "loss": 1.0645, "step": 88790 }, { "epoch": 3.57, "learning_rate": 0.0004111276621442087, "loss": 1.1884, "step": 88800 }, { "epoch": 3.57, "learning_rate": 0.0004111175973267845, "loss": 1.0589, "step": 88810 }, { "epoch": 3.57, "learning_rate": 0.0004111075325093603, "loss": 1.4983, "step": 88820 }, { "epoch": 3.57, "learning_rate": 0.0004110974676919361, "loss": 0.9884, "step": 88830 }, { "epoch": 3.57, "learning_rate": 0.00041108740287451183, "loss": 1.1226, "step": 88840 }, { "epoch": 3.57, "learning_rate": 0.0004110773380570876, "loss": 1.124, "step": 88850 }, { "epoch": 3.57, "learning_rate": 0.0004110672732396634, "loss": 1.3193, "step": 88860 }, { "epoch": 3.57, "learning_rate": 0.0004110572084222392, "loss": 1.3009, "step": 88870 }, { "epoch": 3.57, "learning_rate": 0.00041104714360481505, "loss": 1.1195, "step": 88880 }, { "epoch": 3.58, "learning_rate": 0.00041103707878739084, "loss": 1.1283, "step": 88890 }, { "epoch": 3.58, "learning_rate": 0.00041102701396996663, "loss": 1.142, "step": 88900 }, { "epoch": 3.58, "learning_rate": 0.0004110169491525424, "loss": 1.2239, "step": 88910 }, { "epoch": 3.58, "learning_rate": 0.00041100688433511816, "loss": 1.1547, "step": 88920 }, { "epoch": 3.58, "learning_rate": 0.00041099681951769395, "loss": 1.1282, "step": 88930 }, { "epoch": 3.58, "learning_rate": 0.00041098675470026974, "loss": 1.2648, "step": 88940 }, { "epoch": 3.58, "learning_rate": 0.00041097668988284553, "loss": 1.1468, "step": 88950 }, { "epoch": 3.58, "learning_rate": 0.0004109666250654213, "loss": 1.2002, "step": 88960 }, { "epoch": 3.58, "learning_rate": 0.0004109565602479971, "loss": 1.2178, "step": 88970 }, { "epoch": 3.58, "learning_rate": 0.0004109464954305729, "loss": 1.3501, "step": 88980 }, { "epoch": 3.58, "learning_rate": 0.0004109364306131487, "loss": 1.1165, "step": 88990 }, { "epoch": 3.58, "learning_rate": 0.00041092636579572443, "loss": 1.098, "step": 89000 }, { "epoch": 3.58, "learning_rate": 0.0004109163009783002, "loss": 0.9706, "step": 89010 }, { "epoch": 3.58, "learning_rate": 0.000410906236160876, "loss": 1.2416, "step": 89020 }, { "epoch": 3.58, "learning_rate": 0.00041089617134345186, "loss": 1.3945, "step": 89030 }, { "epoch": 3.58, "learning_rate": 0.00041088610652602765, "loss": 1.2517, "step": 89040 }, { "epoch": 3.58, "learning_rate": 0.00041087604170860344, "loss": 1.0214, "step": 89050 }, { "epoch": 3.58, "learning_rate": 0.00041086597689117923, "loss": 1.0628, "step": 89060 }, { "epoch": 3.58, "learning_rate": 0.000410855912073755, "loss": 1.2221, "step": 89070 }, { "epoch": 3.58, "learning_rate": 0.0004108458472563308, "loss": 0.9949, "step": 89080 }, { "epoch": 3.58, "learning_rate": 0.00041083578243890655, "loss": 1.0842, "step": 89090 }, { "epoch": 3.58, "learning_rate": 0.00041082571762148234, "loss": 1.3419, "step": 89100 }, { "epoch": 3.58, "learning_rate": 0.00041081565280405813, "loss": 1.1019, "step": 89110 }, { "epoch": 3.58, "learning_rate": 0.0004108055879866339, "loss": 1.0661, "step": 89120 }, { "epoch": 3.58, "learning_rate": 0.0004107955231692097, "loss": 0.9627, "step": 89130 }, { "epoch": 3.59, "learning_rate": 0.0004107854583517855, "loss": 1.2396, "step": 89140 }, { "epoch": 3.59, "learning_rate": 0.0004107753935343613, "loss": 0.9031, "step": 89150 }, { "epoch": 3.59, "learning_rate": 0.0004107653287169371, "loss": 1.0456, "step": 89160 }, { "epoch": 3.59, "learning_rate": 0.00041075526389951287, "loss": 1.342, "step": 89170 }, { "epoch": 3.59, "learning_rate": 0.00041074519908208866, "loss": 1.024, "step": 89180 }, { "epoch": 3.59, "learning_rate": 0.00041073513426466445, "loss": 1.086, "step": 89190 }, { "epoch": 3.59, "learning_rate": 0.00041072506944724024, "loss": 1.2748, "step": 89200 }, { "epoch": 3.59, "learning_rate": 0.00041071500462981604, "loss": 1.1669, "step": 89210 }, { "epoch": 3.59, "learning_rate": 0.0004107049398123918, "loss": 1.0446, "step": 89220 }, { "epoch": 3.59, "learning_rate": 0.0004106948749949676, "loss": 1.1701, "step": 89230 }, { "epoch": 3.59, "learning_rate": 0.0004106848101775434, "loss": 0.8499, "step": 89240 }, { "epoch": 3.59, "learning_rate": 0.0004106747453601192, "loss": 1.4432, "step": 89250 }, { "epoch": 3.59, "learning_rate": 0.00041066468054269493, "loss": 0.9929, "step": 89260 }, { "epoch": 3.59, "learning_rate": 0.0004106546157252707, "loss": 1.0682, "step": 89270 }, { "epoch": 3.59, "learning_rate": 0.0004106445509078465, "loss": 1.182, "step": 89280 }, { "epoch": 3.59, "learning_rate": 0.0004106344860904223, "loss": 1.1086, "step": 89290 }, { "epoch": 3.59, "learning_rate": 0.0004106244212729981, "loss": 1.0968, "step": 89300 }, { "epoch": 3.59, "learning_rate": 0.0004106143564555739, "loss": 1.0029, "step": 89310 }, { "epoch": 3.59, "learning_rate": 0.00041060429163814973, "loss": 1.1245, "step": 89320 }, { "epoch": 3.59, "learning_rate": 0.0004105942268207255, "loss": 1.1879, "step": 89330 }, { "epoch": 3.59, "learning_rate": 0.00041058416200330126, "loss": 1.0585, "step": 89340 }, { "epoch": 3.59, "learning_rate": 0.00041057409718587705, "loss": 1.1504, "step": 89350 }, { "epoch": 3.59, "learning_rate": 0.00041056403236845284, "loss": 1.3366, "step": 89360 }, { "epoch": 3.59, "learning_rate": 0.00041055396755102863, "loss": 1.2, "step": 89370 }, { "epoch": 3.59, "learning_rate": 0.0004105439027336044, "loss": 1.2466, "step": 89380 }, { "epoch": 3.6, "learning_rate": 0.0004105338379161802, "loss": 1.5074, "step": 89390 }, { "epoch": 3.6, "learning_rate": 0.000410523773098756, "loss": 1.0448, "step": 89400 }, { "epoch": 3.6, "learning_rate": 0.0004105137082813318, "loss": 1.3034, "step": 89410 }, { "epoch": 3.6, "learning_rate": 0.0004105036434639076, "loss": 1.3105, "step": 89420 }, { "epoch": 3.6, "learning_rate": 0.0004104935786464833, "loss": 1.0462, "step": 89430 }, { "epoch": 3.6, "learning_rate": 0.0004104835138290591, "loss": 1.1337, "step": 89440 }, { "epoch": 3.6, "learning_rate": 0.0004104734490116349, "loss": 0.9251, "step": 89450 }, { "epoch": 3.6, "learning_rate": 0.0004104633841942107, "loss": 1.3647, "step": 89460 }, { "epoch": 3.6, "learning_rate": 0.00041045331937678654, "loss": 1.0487, "step": 89470 }, { "epoch": 3.6, "learning_rate": 0.00041044325455936233, "loss": 1.3438, "step": 89480 }, { "epoch": 3.6, "learning_rate": 0.0004104331897419381, "loss": 1.2163, "step": 89490 }, { "epoch": 3.6, "learning_rate": 0.0004104231249245139, "loss": 1.3531, "step": 89500 }, { "epoch": 3.6, "learning_rate": 0.00041041306010708965, "loss": 1.1924, "step": 89510 }, { "epoch": 3.6, "learning_rate": 0.00041040299528966544, "loss": 1.1617, "step": 89520 }, { "epoch": 3.6, "learning_rate": 0.00041039293047224123, "loss": 1.1077, "step": 89530 }, { "epoch": 3.6, "learning_rate": 0.000410382865654817, "loss": 0.9675, "step": 89540 }, { "epoch": 3.6, "learning_rate": 0.0004103728008373928, "loss": 1.0139, "step": 89550 }, { "epoch": 3.6, "learning_rate": 0.0004103627360199686, "loss": 1.2722, "step": 89560 }, { "epoch": 3.6, "learning_rate": 0.0004103526712025444, "loss": 1.1185, "step": 89570 }, { "epoch": 3.6, "learning_rate": 0.0004103426063851202, "loss": 0.9342, "step": 89580 }, { "epoch": 3.6, "learning_rate": 0.000410332541567696, "loss": 1.006, "step": 89590 }, { "epoch": 3.6, "learning_rate": 0.0004103224767502717, "loss": 1.2555, "step": 89600 }, { "epoch": 3.6, "learning_rate": 0.00041031241193284756, "loss": 1.2787, "step": 89610 }, { "epoch": 3.6, "learning_rate": 0.00041030234711542335, "loss": 1.2721, "step": 89620 }, { "epoch": 3.6, "learning_rate": 0.00041029228229799914, "loss": 1.18, "step": 89630 }, { "epoch": 3.61, "learning_rate": 0.00041028221748057493, "loss": 1.3915, "step": 89640 }, { "epoch": 3.61, "learning_rate": 0.0004102721526631507, "loss": 1.2962, "step": 89650 }, { "epoch": 3.61, "learning_rate": 0.0004102620878457265, "loss": 1.209, "step": 89660 }, { "epoch": 3.61, "learning_rate": 0.0004102520230283023, "loss": 1.2756, "step": 89670 }, { "epoch": 3.61, "learning_rate": 0.00041024195821087804, "loss": 1.1052, "step": 89680 }, { "epoch": 3.61, "learning_rate": 0.00041023189339345383, "loss": 1.1737, "step": 89690 }, { "epoch": 3.61, "learning_rate": 0.0004102218285760296, "loss": 1.2368, "step": 89700 }, { "epoch": 3.61, "learning_rate": 0.0004102117637586054, "loss": 1.1199, "step": 89710 }, { "epoch": 3.61, "learning_rate": 0.0004102016989411812, "loss": 1.193, "step": 89720 }, { "epoch": 3.61, "learning_rate": 0.000410191634123757, "loss": 1.1631, "step": 89730 }, { "epoch": 3.61, "learning_rate": 0.0004101815693063328, "loss": 1.0266, "step": 89740 }, { "epoch": 3.61, "learning_rate": 0.0004101715044889086, "loss": 1.211, "step": 89750 }, { "epoch": 3.61, "learning_rate": 0.0004101614396714844, "loss": 0.9116, "step": 89760 }, { "epoch": 3.61, "learning_rate": 0.00041015137485406016, "loss": 1.0632, "step": 89770 }, { "epoch": 3.61, "learning_rate": 0.00041014131003663595, "loss": 1.0698, "step": 89780 }, { "epoch": 3.61, "learning_rate": 0.00041013124521921174, "loss": 0.9976, "step": 89790 }, { "epoch": 3.61, "learning_rate": 0.00041012118040178753, "loss": 1.0475, "step": 89800 }, { "epoch": 3.61, "learning_rate": 0.0004101111155843633, "loss": 1.1469, "step": 89810 }, { "epoch": 3.61, "learning_rate": 0.0004101010507669391, "loss": 1.2413, "step": 89820 }, { "epoch": 3.61, "learning_rate": 0.0004100909859495149, "loss": 1.1365, "step": 89830 }, { "epoch": 3.61, "learning_rate": 0.0004100809211320907, "loss": 1.3159, "step": 89840 }, { "epoch": 3.61, "learning_rate": 0.00041007085631466643, "loss": 1.306, "step": 89850 }, { "epoch": 3.61, "learning_rate": 0.0004100607914972422, "loss": 1.0668, "step": 89860 }, { "epoch": 3.61, "learning_rate": 0.000410050726679818, "loss": 1.2618, "step": 89870 }, { "epoch": 3.61, "learning_rate": 0.0004100406618623938, "loss": 1.3632, "step": 89880 }, { "epoch": 3.62, "learning_rate": 0.0004100305970449696, "loss": 1.1989, "step": 89890 }, { "epoch": 3.62, "learning_rate": 0.00041002053222754544, "loss": 1.1775, "step": 89900 }, { "epoch": 3.62, "learning_rate": 0.00041001046741012123, "loss": 1.1699, "step": 89910 }, { "epoch": 3.62, "learning_rate": 0.000410000402592697, "loss": 1.2936, "step": 89920 }, { "epoch": 3.62, "learning_rate": 0.00040999033777527275, "loss": 1.0279, "step": 89930 }, { "epoch": 3.62, "learning_rate": 0.00040998027295784855, "loss": 1.1977, "step": 89940 }, { "epoch": 3.62, "learning_rate": 0.00040997020814042434, "loss": 1.1563, "step": 89950 }, { "epoch": 3.62, "learning_rate": 0.0004099601433230001, "loss": 1.1229, "step": 89960 }, { "epoch": 3.62, "learning_rate": 0.0004099500785055759, "loss": 1.0213, "step": 89970 }, { "epoch": 3.62, "learning_rate": 0.0004099400136881517, "loss": 1.2609, "step": 89980 }, { "epoch": 3.62, "learning_rate": 0.0004099299488707275, "loss": 1.0893, "step": 89990 }, { "epoch": 3.62, "learning_rate": 0.0004099198840533033, "loss": 1.1817, "step": 90000 }, { "epoch": 3.62, "learning_rate": 0.0004099098192358791, "loss": 0.9354, "step": 90010 }, { "epoch": 3.62, "learning_rate": 0.0004098997544184548, "loss": 1.0863, "step": 90020 }, { "epoch": 3.62, "learning_rate": 0.0004098896896010306, "loss": 1.2509, "step": 90030 }, { "epoch": 3.62, "learning_rate": 0.0004098796247836064, "loss": 1.2671, "step": 90040 }, { "epoch": 3.62, "learning_rate": 0.00040986955996618224, "loss": 1.0611, "step": 90050 }, { "epoch": 3.62, "learning_rate": 0.00040985949514875803, "loss": 0.9924, "step": 90060 }, { "epoch": 3.62, "learning_rate": 0.0004098494303313338, "loss": 1.0662, "step": 90070 }, { "epoch": 3.62, "learning_rate": 0.0004098393655139096, "loss": 1.1398, "step": 90080 }, { "epoch": 3.62, "learning_rate": 0.0004098293006964854, "loss": 1.2402, "step": 90090 }, { "epoch": 3.62, "learning_rate": 0.00040981923587906114, "loss": 1.1513, "step": 90100 }, { "epoch": 3.62, "learning_rate": 0.00040980917106163693, "loss": 1.0782, "step": 90110 }, { "epoch": 3.62, "learning_rate": 0.0004097991062442127, "loss": 1.1611, "step": 90120 }, { "epoch": 3.62, "learning_rate": 0.0004097890414267885, "loss": 0.9797, "step": 90130 }, { "epoch": 3.63, "learning_rate": 0.0004097789766093643, "loss": 0.889, "step": 90140 }, { "epoch": 3.63, "learning_rate": 0.0004097689117919401, "loss": 1.0368, "step": 90150 }, { "epoch": 3.63, "learning_rate": 0.0004097588469745159, "loss": 1.1309, "step": 90160 }, { "epoch": 3.63, "learning_rate": 0.0004097487821570917, "loss": 1.2213, "step": 90170 }, { "epoch": 3.63, "learning_rate": 0.00040973871733966747, "loss": 0.9716, "step": 90180 }, { "epoch": 3.63, "learning_rate": 0.0004097286525222432, "loss": 1.2033, "step": 90190 }, { "epoch": 3.63, "learning_rate": 0.00040971858770481905, "loss": 1.1768, "step": 90200 }, { "epoch": 3.63, "learning_rate": 0.00040970852288739484, "loss": 1.1307, "step": 90210 }, { "epoch": 3.63, "learning_rate": 0.00040969845806997063, "loss": 1.1556, "step": 90220 }, { "epoch": 3.63, "learning_rate": 0.0004096883932525464, "loss": 1.0893, "step": 90230 }, { "epoch": 3.63, "learning_rate": 0.0004096783284351222, "loss": 1.1469, "step": 90240 }, { "epoch": 3.63, "learning_rate": 0.000409668263617698, "loss": 1.011, "step": 90250 }, { "epoch": 3.63, "learning_rate": 0.0004096581988002738, "loss": 1.1122, "step": 90260 }, { "epoch": 3.63, "learning_rate": 0.00040964813398284953, "loss": 1.3123, "step": 90270 }, { "epoch": 3.63, "learning_rate": 0.0004096380691654253, "loss": 1.1166, "step": 90280 }, { "epoch": 3.63, "learning_rate": 0.0004096280043480011, "loss": 1.1701, "step": 90290 }, { "epoch": 3.63, "learning_rate": 0.0004096179395305769, "loss": 1.0031, "step": 90300 }, { "epoch": 3.63, "learning_rate": 0.0004096078747131527, "loss": 1.0201, "step": 90310 }, { "epoch": 3.63, "learning_rate": 0.0004095978098957285, "loss": 1.0165, "step": 90320 }, { "epoch": 3.63, "learning_rate": 0.0004095877450783043, "loss": 1.2252, "step": 90330 }, { "epoch": 3.63, "learning_rate": 0.0004095776802608801, "loss": 1.1327, "step": 90340 }, { "epoch": 3.63, "learning_rate": 0.0004095676154434559, "loss": 1.0991, "step": 90350 }, { "epoch": 3.63, "learning_rate": 0.00040955755062603165, "loss": 1.0874, "step": 90360 }, { "epoch": 3.63, "learning_rate": 0.00040954748580860744, "loss": 1.028, "step": 90370 }, { "epoch": 3.63, "learning_rate": 0.00040953742099118323, "loss": 0.9627, "step": 90380 }, { "epoch": 3.64, "learning_rate": 0.000409527356173759, "loss": 1.1132, "step": 90390 }, { "epoch": 3.64, "learning_rate": 0.0004095172913563348, "loss": 1.2377, "step": 90400 }, { "epoch": 3.64, "learning_rate": 0.0004095072265389106, "loss": 1.1164, "step": 90410 }, { "epoch": 3.64, "learning_rate": 0.0004094971617214864, "loss": 1.3133, "step": 90420 }, { "epoch": 3.64, "learning_rate": 0.0004094870969040622, "loss": 1.145, "step": 90430 }, { "epoch": 3.64, "learning_rate": 0.0004094770320866379, "loss": 1.266, "step": 90440 }, { "epoch": 3.64, "learning_rate": 0.0004094669672692137, "loss": 1.1854, "step": 90450 }, { "epoch": 3.64, "learning_rate": 0.0004094569024517895, "loss": 1.1575, "step": 90460 }, { "epoch": 3.64, "learning_rate": 0.0004094468376343653, "loss": 1.2442, "step": 90470 }, { "epoch": 3.64, "learning_rate": 0.0004094367728169411, "loss": 1.0295, "step": 90480 }, { "epoch": 3.64, "learning_rate": 0.00040942670799951693, "loss": 1.2767, "step": 90490 }, { "epoch": 3.64, "learning_rate": 0.0004094166431820927, "loss": 1.1957, "step": 90500 }, { "epoch": 3.64, "learning_rate": 0.0004094065783646685, "loss": 0.9066, "step": 90510 }, { "epoch": 3.64, "learning_rate": 0.0004093965135472443, "loss": 1.2913, "step": 90520 }, { "epoch": 3.64, "learning_rate": 0.00040938644872982004, "loss": 1.1295, "step": 90530 }, { "epoch": 3.64, "learning_rate": 0.00040937638391239583, "loss": 1.2994, "step": 90540 }, { "epoch": 3.64, "learning_rate": 0.0004093663190949716, "loss": 1.2843, "step": 90550 }, { "epoch": 3.64, "learning_rate": 0.0004093562542775474, "loss": 1.2502, "step": 90560 }, { "epoch": 3.64, "learning_rate": 0.0004093461894601232, "loss": 1.173, "step": 90570 }, { "epoch": 3.64, "learning_rate": 0.000409336124642699, "loss": 1.1139, "step": 90580 }, { "epoch": 3.64, "learning_rate": 0.0004093260598252748, "loss": 1.3567, "step": 90590 }, { "epoch": 3.64, "learning_rate": 0.0004093159950078506, "loss": 1.2813, "step": 90600 }, { "epoch": 3.64, "learning_rate": 0.0004093059301904263, "loss": 1.1498, "step": 90610 }, { "epoch": 3.64, "learning_rate": 0.0004092958653730021, "loss": 1.2893, "step": 90620 }, { "epoch": 3.64, "learning_rate": 0.0004092858005555779, "loss": 1.0296, "step": 90630 }, { "epoch": 3.65, "learning_rate": 0.00040927573573815374, "loss": 1.2602, "step": 90640 }, { "epoch": 3.65, "learning_rate": 0.00040926567092072953, "loss": 1.1449, "step": 90650 }, { "epoch": 3.65, "learning_rate": 0.0004092556061033053, "loss": 1.0532, "step": 90660 }, { "epoch": 3.65, "learning_rate": 0.0004092455412858811, "loss": 1.1654, "step": 90670 }, { "epoch": 3.65, "learning_rate": 0.0004092354764684569, "loss": 0.9888, "step": 90680 }, { "epoch": 3.65, "learning_rate": 0.00040922541165103264, "loss": 1.0192, "step": 90690 }, { "epoch": 3.65, "learning_rate": 0.00040921534683360843, "loss": 1.0118, "step": 90700 }, { "epoch": 3.65, "learning_rate": 0.0004092052820161842, "loss": 1.1557, "step": 90710 }, { "epoch": 3.65, "learning_rate": 0.00040919521719876, "loss": 0.9978, "step": 90720 }, { "epoch": 3.65, "learning_rate": 0.0004091851523813358, "loss": 1.2164, "step": 90730 }, { "epoch": 3.65, "learning_rate": 0.0004091750875639116, "loss": 1.2279, "step": 90740 }, { "epoch": 3.65, "learning_rate": 0.0004091650227464874, "loss": 1.041, "step": 90750 }, { "epoch": 3.65, "learning_rate": 0.00040915495792906317, "loss": 1.2666, "step": 90760 }, { "epoch": 3.65, "learning_rate": 0.00040914489311163896, "loss": 1.1201, "step": 90770 }, { "epoch": 3.65, "learning_rate": 0.00040913482829421475, "loss": 1.1139, "step": 90780 }, { "epoch": 3.65, "learning_rate": 0.00040912476347679055, "loss": 1.022, "step": 90790 }, { "epoch": 3.65, "learning_rate": 0.00040911469865936634, "loss": 1.2554, "step": 90800 }, { "epoch": 3.65, "learning_rate": 0.0004091046338419421, "loss": 1.0916, "step": 90810 }, { "epoch": 3.65, "learning_rate": 0.0004090945690245179, "loss": 1.1726, "step": 90820 }, { "epoch": 3.65, "learning_rate": 0.0004090845042070937, "loss": 1.1825, "step": 90830 }, { "epoch": 3.65, "learning_rate": 0.0004090744393896695, "loss": 1.2018, "step": 90840 }, { "epoch": 3.65, "learning_rate": 0.0004090643745722453, "loss": 0.93, "step": 90850 }, { "epoch": 3.65, "learning_rate": 0.000409054309754821, "loss": 1.08, "step": 90860 }, { "epoch": 3.65, "learning_rate": 0.0004090442449373968, "loss": 1.3261, "step": 90870 }, { "epoch": 3.66, "learning_rate": 0.0004090341801199726, "loss": 1.1533, "step": 90880 }, { "epoch": 3.66, "learning_rate": 0.0004090241153025484, "loss": 1.1267, "step": 90890 }, { "epoch": 3.66, "learning_rate": 0.0004090140504851242, "loss": 1.275, "step": 90900 }, { "epoch": 3.66, "learning_rate": 0.0004090039856677, "loss": 1.008, "step": 90910 }, { "epoch": 3.66, "learning_rate": 0.00040899392085027577, "loss": 1.1254, "step": 90920 }, { "epoch": 3.66, "learning_rate": 0.0004089838560328516, "loss": 1.0888, "step": 90930 }, { "epoch": 3.66, "learning_rate": 0.0004089737912154274, "loss": 1.1153, "step": 90940 }, { "epoch": 3.66, "learning_rate": 0.00040896372639800314, "loss": 1.2158, "step": 90950 }, { "epoch": 3.66, "learning_rate": 0.00040895366158057893, "loss": 1.2258, "step": 90960 }, { "epoch": 3.66, "learning_rate": 0.0004089435967631547, "loss": 1.3164, "step": 90970 }, { "epoch": 3.66, "learning_rate": 0.0004089335319457305, "loss": 1.1512, "step": 90980 }, { "epoch": 3.66, "learning_rate": 0.0004089234671283063, "loss": 1.2228, "step": 90990 }, { "epoch": 3.66, "learning_rate": 0.0004089134023108821, "loss": 1.066, "step": 91000 }, { "epoch": 3.66, "learning_rate": 0.0004089033374934579, "loss": 0.9511, "step": 91010 }, { "epoch": 3.66, "learning_rate": 0.0004088932726760337, "loss": 1.2158, "step": 91020 }, { "epoch": 3.66, "learning_rate": 0.0004088832078586094, "loss": 1.1701, "step": 91030 }, { "epoch": 3.66, "learning_rate": 0.0004088731430411852, "loss": 1.0903, "step": 91040 }, { "epoch": 3.66, "learning_rate": 0.000408863078223761, "loss": 1.0114, "step": 91050 }, { "epoch": 3.66, "learning_rate": 0.0004088530134063368, "loss": 1.1253, "step": 91060 }, { "epoch": 3.66, "learning_rate": 0.00040884294858891263, "loss": 1.1307, "step": 91070 }, { "epoch": 3.66, "learning_rate": 0.0004088328837714884, "loss": 0.9755, "step": 91080 }, { "epoch": 3.66, "learning_rate": 0.0004088228189540642, "loss": 1.3121, "step": 91090 }, { "epoch": 3.66, "learning_rate": 0.00040881275413664, "loss": 1.1162, "step": 91100 }, { "epoch": 3.66, "learning_rate": 0.0004088026893192158, "loss": 1.078, "step": 91110 }, { "epoch": 3.66, "learning_rate": 0.00040879262450179153, "loss": 1.1008, "step": 91120 }, { "epoch": 3.67, "learning_rate": 0.0004087825596843673, "loss": 1.216, "step": 91130 }, { "epoch": 3.67, "learning_rate": 0.0004087724948669431, "loss": 1.0268, "step": 91140 }, { "epoch": 3.67, "learning_rate": 0.0004087624300495189, "loss": 1.2933, "step": 91150 }, { "epoch": 3.67, "learning_rate": 0.0004087523652320947, "loss": 0.9749, "step": 91160 }, { "epoch": 3.67, "learning_rate": 0.0004087423004146705, "loss": 1.0066, "step": 91170 }, { "epoch": 3.67, "learning_rate": 0.0004087322355972463, "loss": 1.4761, "step": 91180 }, { "epoch": 3.67, "learning_rate": 0.00040872217077982207, "loss": 1.1276, "step": 91190 }, { "epoch": 3.67, "learning_rate": 0.0004087121059623978, "loss": 1.1973, "step": 91200 }, { "epoch": 3.67, "learning_rate": 0.0004087020411449736, "loss": 0.9681, "step": 91210 }, { "epoch": 3.67, "learning_rate": 0.00040869197632754944, "loss": 0.9476, "step": 91220 }, { "epoch": 3.67, "learning_rate": 0.00040868191151012523, "loss": 1.0576, "step": 91230 }, { "epoch": 3.67, "learning_rate": 0.000408671846692701, "loss": 1.1048, "step": 91240 }, { "epoch": 3.67, "learning_rate": 0.0004086617818752768, "loss": 1.3088, "step": 91250 }, { "epoch": 3.67, "learning_rate": 0.0004086517170578526, "loss": 1.1901, "step": 91260 }, { "epoch": 3.67, "learning_rate": 0.0004086416522404284, "loss": 1.2138, "step": 91270 }, { "epoch": 3.67, "learning_rate": 0.0004086315874230042, "loss": 1.0948, "step": 91280 }, { "epoch": 3.67, "learning_rate": 0.0004086215226055799, "loss": 1.1091, "step": 91290 }, { "epoch": 3.67, "learning_rate": 0.0004086114577881557, "loss": 1.1857, "step": 91300 }, { "epoch": 3.67, "learning_rate": 0.0004086013929707315, "loss": 1.1743, "step": 91310 }, { "epoch": 3.67, "learning_rate": 0.0004085913281533073, "loss": 1.098, "step": 91320 }, { "epoch": 3.67, "learning_rate": 0.0004085812633358831, "loss": 1.2313, "step": 91330 }, { "epoch": 3.67, "learning_rate": 0.0004085711985184589, "loss": 1.1387, "step": 91340 }, { "epoch": 3.67, "learning_rate": 0.00040856113370103467, "loss": 1.2793, "step": 91350 }, { "epoch": 3.67, "learning_rate": 0.00040855106888361046, "loss": 1.0511, "step": 91360 }, { "epoch": 3.67, "learning_rate": 0.00040854100406618625, "loss": 1.2723, "step": 91370 }, { "epoch": 3.68, "learning_rate": 0.00040853093924876204, "loss": 0.9568, "step": 91380 }, { "epoch": 3.68, "learning_rate": 0.00040852087443133783, "loss": 1.4127, "step": 91390 }, { "epoch": 3.68, "learning_rate": 0.0004085108096139136, "loss": 1.0394, "step": 91400 }, { "epoch": 3.68, "learning_rate": 0.0004085007447964894, "loss": 1.2725, "step": 91410 }, { "epoch": 3.68, "learning_rate": 0.0004084906799790652, "loss": 1.3059, "step": 91420 }, { "epoch": 3.68, "learning_rate": 0.000408480615161641, "loss": 1.2746, "step": 91430 }, { "epoch": 3.68, "learning_rate": 0.0004084705503442168, "loss": 1.099, "step": 91440 }, { "epoch": 3.68, "learning_rate": 0.0004084604855267925, "loss": 1.101, "step": 91450 }, { "epoch": 3.68, "learning_rate": 0.0004084504207093683, "loss": 1.3332, "step": 91460 }, { "epoch": 3.68, "learning_rate": 0.0004084403558919441, "loss": 1.2174, "step": 91470 }, { "epoch": 3.68, "learning_rate": 0.0004084302910745199, "loss": 1.0429, "step": 91480 }, { "epoch": 3.68, "learning_rate": 0.0004084202262570957, "loss": 1.2987, "step": 91490 }, { "epoch": 3.68, "learning_rate": 0.0004084101614396715, "loss": 1.2194, "step": 91500 }, { "epoch": 3.68, "learning_rate": 0.0004084000966222473, "loss": 1.282, "step": 91510 }, { "epoch": 3.68, "learning_rate": 0.0004083900318048231, "loss": 1.0678, "step": 91520 }, { "epoch": 3.68, "learning_rate": 0.0004083799669873989, "loss": 1.225, "step": 91530 }, { "epoch": 3.68, "learning_rate": 0.00040836990216997464, "loss": 1.1648, "step": 91540 }, { "epoch": 3.68, "learning_rate": 0.00040835983735255043, "loss": 1.0113, "step": 91550 }, { "epoch": 3.68, "learning_rate": 0.0004083497725351262, "loss": 0.9631, "step": 91560 }, { "epoch": 3.68, "learning_rate": 0.000408339707717702, "loss": 1.3162, "step": 91570 }, { "epoch": 3.68, "learning_rate": 0.0004083296429002778, "loss": 1.1755, "step": 91580 }, { "epoch": 3.68, "learning_rate": 0.0004083195780828536, "loss": 1.2501, "step": 91590 }, { "epoch": 3.68, "learning_rate": 0.0004083095132654294, "loss": 1.0626, "step": 91600 }, { "epoch": 3.68, "learning_rate": 0.00040829944844800517, "loss": 1.1317, "step": 91610 }, { "epoch": 3.68, "learning_rate": 0.0004082893836305809, "loss": 1.0684, "step": 91620 }, { "epoch": 3.69, "learning_rate": 0.0004082793188131567, "loss": 1.0618, "step": 91630 }, { "epoch": 3.69, "learning_rate": 0.0004082692539957325, "loss": 1.0118, "step": 91640 }, { "epoch": 3.69, "learning_rate": 0.0004082591891783083, "loss": 1.0429, "step": 91650 }, { "epoch": 3.69, "learning_rate": 0.0004082491243608841, "loss": 1.0059, "step": 91660 }, { "epoch": 3.69, "learning_rate": 0.0004082390595434599, "loss": 0.9338, "step": 91670 }, { "epoch": 3.69, "learning_rate": 0.0004082289947260357, "loss": 1.025, "step": 91680 }, { "epoch": 3.69, "learning_rate": 0.0004082189299086115, "loss": 1.0692, "step": 91690 }, { "epoch": 3.69, "learning_rate": 0.0004082088650911873, "loss": 1.1053, "step": 91700 }, { "epoch": 3.69, "learning_rate": 0.000408198800273763, "loss": 1.2299, "step": 91710 }, { "epoch": 3.69, "learning_rate": 0.0004081887354563388, "loss": 1.3403, "step": 91720 }, { "epoch": 3.69, "learning_rate": 0.0004081786706389146, "loss": 1.1275, "step": 91730 }, { "epoch": 3.69, "learning_rate": 0.0004081686058214904, "loss": 1.2205, "step": 91740 }, { "epoch": 3.69, "learning_rate": 0.0004081585410040662, "loss": 1.1634, "step": 91750 }, { "epoch": 3.69, "learning_rate": 0.000408148476186642, "loss": 1.1148, "step": 91760 }, { "epoch": 3.69, "learning_rate": 0.00040813841136921777, "loss": 1.1122, "step": 91770 }, { "epoch": 3.69, "learning_rate": 0.00040812834655179356, "loss": 0.9264, "step": 91780 }, { "epoch": 3.69, "learning_rate": 0.0004081182817343693, "loss": 1.1149, "step": 91790 }, { "epoch": 3.69, "learning_rate": 0.0004081082169169451, "loss": 1.0536, "step": 91800 }, { "epoch": 3.69, "learning_rate": 0.00040809815209952093, "loss": 1.0587, "step": 91810 }, { "epoch": 3.69, "learning_rate": 0.0004080880872820967, "loss": 1.2573, "step": 91820 }, { "epoch": 3.69, "learning_rate": 0.0004080780224646725, "loss": 1.2829, "step": 91830 }, { "epoch": 3.69, "learning_rate": 0.0004080679576472483, "loss": 1.1244, "step": 91840 }, { "epoch": 3.69, "learning_rate": 0.0004080578928298241, "loss": 1.0612, "step": 91850 }, { "epoch": 3.69, "learning_rate": 0.0004080478280123999, "loss": 0.9719, "step": 91860 }, { "epoch": 3.69, "learning_rate": 0.0004080377631949757, "loss": 1.0611, "step": 91870 }, { "epoch": 3.7, "learning_rate": 0.0004080276983775514, "loss": 0.9606, "step": 91880 }, { "epoch": 3.7, "learning_rate": 0.0004080176335601272, "loss": 1.146, "step": 91890 }, { "epoch": 3.7, "learning_rate": 0.000408007568742703, "loss": 1.1621, "step": 91900 }, { "epoch": 3.7, "learning_rate": 0.0004079975039252788, "loss": 1.0954, "step": 91910 }, { "epoch": 3.7, "learning_rate": 0.0004079874391078546, "loss": 1.0378, "step": 91920 }, { "epoch": 3.7, "learning_rate": 0.00040797737429043037, "loss": 1.3703, "step": 91930 }, { "epoch": 3.7, "learning_rate": 0.00040796730947300616, "loss": 1.171, "step": 91940 }, { "epoch": 3.7, "learning_rate": 0.000407957244655582, "loss": 1.0538, "step": 91950 }, { "epoch": 3.7, "learning_rate": 0.00040794717983815774, "loss": 1.2309, "step": 91960 }, { "epoch": 3.7, "learning_rate": 0.00040793711502073353, "loss": 1.1608, "step": 91970 }, { "epoch": 3.7, "learning_rate": 0.0004079270502033093, "loss": 1.2649, "step": 91980 }, { "epoch": 3.7, "learning_rate": 0.0004079169853858851, "loss": 1.0594, "step": 91990 }, { "epoch": 3.7, "learning_rate": 0.0004079069205684609, "loss": 1.1013, "step": 92000 }, { "epoch": 3.7, "learning_rate": 0.0004078968557510367, "loss": 1.1924, "step": 92010 }, { "epoch": 3.7, "learning_rate": 0.0004078867909336125, "loss": 1.081, "step": 92020 }, { "epoch": 3.7, "learning_rate": 0.0004078767261161883, "loss": 1.195, "step": 92030 }, { "epoch": 3.7, "learning_rate": 0.00040786666129876407, "loss": 0.9843, "step": 92040 }, { "epoch": 3.7, "learning_rate": 0.0004078565964813398, "loss": 0.9857, "step": 92050 }, { "epoch": 3.7, "learning_rate": 0.0004078465316639156, "loss": 0.8265, "step": 92060 }, { "epoch": 3.7, "learning_rate": 0.0004078364668464914, "loss": 1.1037, "step": 92070 }, { "epoch": 3.7, "learning_rate": 0.0004078264020290672, "loss": 1.2235, "step": 92080 }, { "epoch": 3.7, "learning_rate": 0.00040781633721164297, "loss": 1.1368, "step": 92090 }, { "epoch": 3.7, "learning_rate": 0.0004078062723942188, "loss": 1.3155, "step": 92100 }, { "epoch": 3.7, "learning_rate": 0.0004077962075767946, "loss": 1.3004, "step": 92110 }, { "epoch": 3.7, "learning_rate": 0.0004077861427593704, "loss": 1.1008, "step": 92120 }, { "epoch": 3.71, "learning_rate": 0.00040777607794194613, "loss": 0.9741, "step": 92130 }, { "epoch": 3.71, "learning_rate": 0.0004077660131245219, "loss": 1.4594, "step": 92140 }, { "epoch": 3.71, "learning_rate": 0.0004077559483070977, "loss": 1.3692, "step": 92150 }, { "epoch": 3.71, "learning_rate": 0.0004077458834896735, "loss": 1.0583, "step": 92160 }, { "epoch": 3.71, "learning_rate": 0.0004077358186722493, "loss": 1.076, "step": 92170 }, { "epoch": 3.71, "learning_rate": 0.0004077257538548251, "loss": 1.1304, "step": 92180 }, { "epoch": 3.71, "learning_rate": 0.0004077156890374009, "loss": 1.1306, "step": 92190 }, { "epoch": 3.71, "learning_rate": 0.00040770562421997667, "loss": 1.3409, "step": 92200 }, { "epoch": 3.71, "learning_rate": 0.0004076955594025524, "loss": 1.1185, "step": 92210 }, { "epoch": 3.71, "learning_rate": 0.0004076854945851282, "loss": 0.9232, "step": 92220 }, { "epoch": 3.71, "learning_rate": 0.000407675429767704, "loss": 1.1598, "step": 92230 }, { "epoch": 3.71, "learning_rate": 0.00040766536495027983, "loss": 1.2646, "step": 92240 }, { "epoch": 3.71, "learning_rate": 0.0004076553001328556, "loss": 1.0825, "step": 92250 }, { "epoch": 3.71, "learning_rate": 0.0004076452353154314, "loss": 1.0161, "step": 92260 }, { "epoch": 3.71, "learning_rate": 0.0004076351704980072, "loss": 1.254, "step": 92270 }, { "epoch": 3.71, "learning_rate": 0.000407625105680583, "loss": 1.0513, "step": 92280 }, { "epoch": 3.71, "learning_rate": 0.0004076150408631588, "loss": 1.0098, "step": 92290 }, { "epoch": 3.71, "learning_rate": 0.0004076049760457345, "loss": 1.0744, "step": 92300 }, { "epoch": 3.71, "learning_rate": 0.0004075949112283103, "loss": 1.1387, "step": 92310 }, { "epoch": 3.71, "learning_rate": 0.0004075848464108861, "loss": 1.0394, "step": 92320 }, { "epoch": 3.71, "learning_rate": 0.0004075747815934619, "loss": 0.9142, "step": 92330 }, { "epoch": 3.71, "learning_rate": 0.0004075647167760377, "loss": 1.3356, "step": 92340 }, { "epoch": 3.71, "learning_rate": 0.0004075546519586135, "loss": 1.4589, "step": 92350 }, { "epoch": 3.71, "learning_rate": 0.00040754458714118926, "loss": 1.2553, "step": 92360 }, { "epoch": 3.71, "learning_rate": 0.00040753452232376506, "loss": 1.1669, "step": 92370 }, { "epoch": 3.72, "learning_rate": 0.0004075244575063408, "loss": 1.167, "step": 92380 }, { "epoch": 3.72, "learning_rate": 0.00040751439268891664, "loss": 1.4514, "step": 92390 }, { "epoch": 3.72, "learning_rate": 0.00040750432787149243, "loss": 1.2162, "step": 92400 }, { "epoch": 3.72, "learning_rate": 0.0004074942630540682, "loss": 1.2785, "step": 92410 }, { "epoch": 3.72, "learning_rate": 0.000407484198236644, "loss": 1.2436, "step": 92420 }, { "epoch": 3.72, "learning_rate": 0.0004074741334192198, "loss": 1.2453, "step": 92430 }, { "epoch": 3.72, "learning_rate": 0.0004074640686017956, "loss": 1.1963, "step": 92440 }, { "epoch": 3.72, "learning_rate": 0.0004074540037843714, "loss": 1.2112, "step": 92450 }, { "epoch": 3.72, "learning_rate": 0.00040744393896694717, "loss": 1.1955, "step": 92460 }, { "epoch": 3.72, "learning_rate": 0.0004074338741495229, "loss": 0.9716, "step": 92470 }, { "epoch": 3.72, "learning_rate": 0.0004074238093320987, "loss": 1.1583, "step": 92480 }, { "epoch": 3.72, "learning_rate": 0.0004074137445146745, "loss": 1.2301, "step": 92490 }, { "epoch": 3.72, "learning_rate": 0.0004074036796972503, "loss": 1.0642, "step": 92500 }, { "epoch": 3.72, "learning_rate": 0.00040739361487982607, "loss": 1.0096, "step": 92510 }, { "epoch": 3.72, "learning_rate": 0.00040738355006240186, "loss": 1.1964, "step": 92520 }, { "epoch": 3.72, "learning_rate": 0.00040737348524497765, "loss": 1.1852, "step": 92530 }, { "epoch": 3.72, "learning_rate": 0.0004073634204275535, "loss": 1.0492, "step": 92540 }, { "epoch": 3.72, "learning_rate": 0.00040735335561012924, "loss": 1.2657, "step": 92550 }, { "epoch": 3.72, "learning_rate": 0.000407343290792705, "loss": 1.2553, "step": 92560 }, { "epoch": 3.72, "learning_rate": 0.0004073332259752808, "loss": 1.0302, "step": 92570 }, { "epoch": 3.72, "learning_rate": 0.0004073231611578566, "loss": 1.2596, "step": 92580 }, { "epoch": 3.72, "learning_rate": 0.0004073130963404324, "loss": 1.0825, "step": 92590 }, { "epoch": 3.72, "learning_rate": 0.0004073030315230082, "loss": 0.9168, "step": 92600 }, { "epoch": 3.72, "learning_rate": 0.000407292966705584, "loss": 1.159, "step": 92610 }, { "epoch": 3.73, "learning_rate": 0.00040728290188815977, "loss": 1.2766, "step": 92620 }, { "epoch": 3.73, "learning_rate": 0.00040727283707073556, "loss": 1.1166, "step": 92630 }, { "epoch": 3.73, "learning_rate": 0.0004072627722533113, "loss": 1.2495, "step": 92640 }, { "epoch": 3.73, "learning_rate": 0.0004072527074358871, "loss": 1.323, "step": 92650 }, { "epoch": 3.73, "learning_rate": 0.0004072426426184629, "loss": 1.3316, "step": 92660 }, { "epoch": 3.73, "learning_rate": 0.00040723257780103867, "loss": 1.3049, "step": 92670 }, { "epoch": 3.73, "learning_rate": 0.0004072225129836145, "loss": 1.0778, "step": 92680 }, { "epoch": 3.73, "learning_rate": 0.0004072124481661903, "loss": 0.9264, "step": 92690 }, { "epoch": 3.73, "learning_rate": 0.0004072023833487661, "loss": 1.1359, "step": 92700 }, { "epoch": 3.73, "learning_rate": 0.0004071923185313419, "loss": 1.2631, "step": 92710 }, { "epoch": 3.73, "learning_rate": 0.0004071822537139176, "loss": 1.0434, "step": 92720 }, { "epoch": 3.73, "learning_rate": 0.0004071721888964934, "loss": 1.2191, "step": 92730 }, { "epoch": 3.73, "learning_rate": 0.0004071621240790692, "loss": 1.2795, "step": 92740 }, { "epoch": 3.73, "learning_rate": 0.000407152059261645, "loss": 1.2671, "step": 92750 }, { "epoch": 3.73, "learning_rate": 0.0004071419944442208, "loss": 0.8795, "step": 92760 }, { "epoch": 3.73, "learning_rate": 0.0004071319296267966, "loss": 1.1175, "step": 92770 }, { "epoch": 3.73, "learning_rate": 0.00040712186480937237, "loss": 1.109, "step": 92780 }, { "epoch": 3.73, "learning_rate": 0.00040711179999194816, "loss": 1.2816, "step": 92790 }, { "epoch": 3.73, "learning_rate": 0.00040710173517452395, "loss": 1.0963, "step": 92800 }, { "epoch": 3.73, "learning_rate": 0.0004070916703570997, "loss": 1.3775, "step": 92810 }, { "epoch": 3.73, "learning_rate": 0.0004070816055396755, "loss": 1.1275, "step": 92820 }, { "epoch": 3.73, "learning_rate": 0.0004070715407222513, "loss": 1.1293, "step": 92830 }, { "epoch": 3.73, "learning_rate": 0.0004070614759048271, "loss": 1.276, "step": 92840 }, { "epoch": 3.73, "learning_rate": 0.0004070514110874029, "loss": 1.0589, "step": 92850 }, { "epoch": 3.73, "learning_rate": 0.0004070413462699787, "loss": 1.2878, "step": 92860 }, { "epoch": 3.74, "learning_rate": 0.0004070312814525545, "loss": 1.3308, "step": 92870 }, { "epoch": 3.74, "learning_rate": 0.0004070212166351303, "loss": 1.2953, "step": 92880 }, { "epoch": 3.74, "learning_rate": 0.000407011151817706, "loss": 1.402, "step": 92890 }, { "epoch": 3.74, "learning_rate": 0.0004070010870002818, "loss": 0.9282, "step": 92900 }, { "epoch": 3.74, "learning_rate": 0.0004069910221828576, "loss": 1.1341, "step": 92910 }, { "epoch": 3.74, "learning_rate": 0.0004069809573654334, "loss": 1.0733, "step": 92920 }, { "epoch": 3.74, "learning_rate": 0.0004069708925480092, "loss": 0.9843, "step": 92930 }, { "epoch": 3.74, "learning_rate": 0.00040696082773058497, "loss": 1.2557, "step": 92940 }, { "epoch": 3.74, "learning_rate": 0.00040695076291316076, "loss": 0.9233, "step": 92950 }, { "epoch": 3.74, "learning_rate": 0.00040694069809573655, "loss": 1.1377, "step": 92960 }, { "epoch": 3.74, "learning_rate": 0.0004069306332783123, "loss": 1.2219, "step": 92970 }, { "epoch": 3.74, "learning_rate": 0.00040692056846088813, "loss": 1.1883, "step": 92980 }, { "epoch": 3.74, "learning_rate": 0.0004069105036434639, "loss": 1.108, "step": 92990 }, { "epoch": 3.74, "learning_rate": 0.0004069004388260397, "loss": 0.998, "step": 93000 }, { "epoch": 3.74, "learning_rate": 0.0004068903740086155, "loss": 0.9684, "step": 93010 }, { "epoch": 3.74, "learning_rate": 0.0004068803091911913, "loss": 1.1378, "step": 93020 }, { "epoch": 3.74, "learning_rate": 0.0004068702443737671, "loss": 1.0187, "step": 93030 }, { "epoch": 3.74, "learning_rate": 0.0004068601795563429, "loss": 1.1481, "step": 93040 }, { "epoch": 3.74, "learning_rate": 0.00040685011473891867, "loss": 1.1179, "step": 93050 }, { "epoch": 3.74, "learning_rate": 0.0004068400499214944, "loss": 1.1847, "step": 93060 }, { "epoch": 3.74, "learning_rate": 0.0004068299851040702, "loss": 0.9753, "step": 93070 }, { "epoch": 3.74, "learning_rate": 0.000406819920286646, "loss": 1.1152, "step": 93080 }, { "epoch": 3.74, "learning_rate": 0.0004068098554692218, "loss": 1.2831, "step": 93090 }, { "epoch": 3.74, "learning_rate": 0.00040679979065179757, "loss": 1.1914, "step": 93100 }, { "epoch": 3.74, "learning_rate": 0.00040678972583437336, "loss": 1.3103, "step": 93110 }, { "epoch": 3.75, "learning_rate": 0.0004067796610169492, "loss": 1.2144, "step": 93120 }, { "epoch": 3.75, "learning_rate": 0.000406769596199525, "loss": 0.9628, "step": 93130 }, { "epoch": 3.75, "learning_rate": 0.0004067595313821008, "loss": 1.1197, "step": 93140 }, { "epoch": 3.75, "learning_rate": 0.0004067494665646765, "loss": 1.1437, "step": 93150 }, { "epoch": 3.75, "learning_rate": 0.0004067394017472523, "loss": 1.4101, "step": 93160 }, { "epoch": 3.75, "learning_rate": 0.0004067293369298281, "loss": 1.1946, "step": 93170 }, { "epoch": 3.75, "learning_rate": 0.0004067192721124039, "loss": 1.2047, "step": 93180 }, { "epoch": 3.75, "learning_rate": 0.0004067092072949797, "loss": 1.2326, "step": 93190 }, { "epoch": 3.75, "learning_rate": 0.0004066991424775555, "loss": 1.101, "step": 93200 }, { "epoch": 3.75, "learning_rate": 0.00040668907766013126, "loss": 1.3178, "step": 93210 }, { "epoch": 3.75, "learning_rate": 0.00040667901284270706, "loss": 1.247, "step": 93220 }, { "epoch": 3.75, "learning_rate": 0.0004066689480252828, "loss": 1.1642, "step": 93230 }, { "epoch": 3.75, "learning_rate": 0.0004066588832078586, "loss": 1.2736, "step": 93240 }, { "epoch": 3.75, "learning_rate": 0.0004066488183904344, "loss": 0.83, "step": 93250 }, { "epoch": 3.75, "learning_rate": 0.00040663875357301016, "loss": 1.2687, "step": 93260 }, { "epoch": 3.75, "learning_rate": 0.000406628688755586, "loss": 1.1663, "step": 93270 }, { "epoch": 3.75, "learning_rate": 0.0004066186239381618, "loss": 0.9494, "step": 93280 }, { "epoch": 3.75, "learning_rate": 0.0004066085591207376, "loss": 1.0621, "step": 93290 }, { "epoch": 3.75, "learning_rate": 0.0004065984943033134, "loss": 0.9636, "step": 93300 }, { "epoch": 3.75, "learning_rate": 0.0004065884294858891, "loss": 1.2606, "step": 93310 }, { "epoch": 3.75, "learning_rate": 0.0004065783646684649, "loss": 1.159, "step": 93320 }, { "epoch": 3.75, "learning_rate": 0.0004065682998510407, "loss": 1.0584, "step": 93330 }, { "epoch": 3.75, "learning_rate": 0.0004065582350336165, "loss": 1.0793, "step": 93340 }, { "epoch": 3.75, "learning_rate": 0.0004065481702161923, "loss": 1.2161, "step": 93350 }, { "epoch": 3.75, "learning_rate": 0.00040653810539876807, "loss": 1.0052, "step": 93360 }, { "epoch": 3.76, "learning_rate": 0.00040652804058134386, "loss": 1.3023, "step": 93370 }, { "epoch": 3.76, "learning_rate": 0.00040651797576391965, "loss": 1.1, "step": 93380 }, { "epoch": 3.76, "learning_rate": 0.00040650791094649544, "loss": 1.2282, "step": 93390 }, { "epoch": 3.76, "learning_rate": 0.0004064978461290712, "loss": 1.1483, "step": 93400 }, { "epoch": 3.76, "learning_rate": 0.000406487781311647, "loss": 0.9391, "step": 93410 }, { "epoch": 3.76, "learning_rate": 0.0004064777164942228, "loss": 1.0753, "step": 93420 }, { "epoch": 3.76, "learning_rate": 0.0004064676516767986, "loss": 1.0876, "step": 93430 }, { "epoch": 3.76, "learning_rate": 0.0004064575868593744, "loss": 1.3337, "step": 93440 }, { "epoch": 3.76, "learning_rate": 0.0004064475220419502, "loss": 1.0357, "step": 93450 }, { "epoch": 3.76, "learning_rate": 0.000406437457224526, "loss": 1.2299, "step": 93460 }, { "epoch": 3.76, "learning_rate": 0.00040642739240710177, "loss": 1.3275, "step": 93470 }, { "epoch": 3.76, "learning_rate": 0.0004064173275896775, "loss": 1.3652, "step": 93480 }, { "epoch": 3.76, "learning_rate": 0.0004064072627722533, "loss": 1.3818, "step": 93490 }, { "epoch": 3.76, "learning_rate": 0.0004063971979548291, "loss": 1.0835, "step": 93500 }, { "epoch": 3.76, "learning_rate": 0.0004063871331374049, "loss": 1.0814, "step": 93510 }, { "epoch": 3.76, "learning_rate": 0.00040637706831998067, "loss": 0.9519, "step": 93520 }, { "epoch": 3.76, "learning_rate": 0.00040636700350255646, "loss": 1.3357, "step": 93530 }, { "epoch": 3.76, "learning_rate": 0.00040635693868513225, "loss": 1.1915, "step": 93540 }, { "epoch": 3.76, "learning_rate": 0.00040634687386770804, "loss": 1.0043, "step": 93550 }, { "epoch": 3.76, "learning_rate": 0.0004063368090502839, "loss": 1.1193, "step": 93560 }, { "epoch": 3.76, "learning_rate": 0.0004063267442328596, "loss": 1.1433, "step": 93570 }, { "epoch": 3.76, "learning_rate": 0.0004063166794154354, "loss": 1.3036, "step": 93580 }, { "epoch": 3.76, "learning_rate": 0.0004063066145980112, "loss": 1.1702, "step": 93590 }, { "epoch": 3.76, "learning_rate": 0.000406296549780587, "loss": 0.9001, "step": 93600 }, { "epoch": 3.76, "learning_rate": 0.0004062864849631628, "loss": 1.0738, "step": 93610 }, { "epoch": 3.77, "learning_rate": 0.0004062764201457386, "loss": 1.2492, "step": 93620 }, { "epoch": 3.77, "learning_rate": 0.00040626635532831437, "loss": 1.1785, "step": 93630 }, { "epoch": 3.77, "learning_rate": 0.00040625629051089016, "loss": 1.0435, "step": 93640 }, { "epoch": 3.77, "learning_rate": 0.0004062462256934659, "loss": 1.1003, "step": 93650 }, { "epoch": 3.77, "learning_rate": 0.0004062361608760417, "loss": 1.2118, "step": 93660 }, { "epoch": 3.77, "learning_rate": 0.0004062260960586175, "loss": 1.1474, "step": 93670 }, { "epoch": 3.77, "learning_rate": 0.00040621603124119327, "loss": 1.1673, "step": 93680 }, { "epoch": 3.77, "learning_rate": 0.00040620596642376906, "loss": 1.0948, "step": 93690 }, { "epoch": 3.77, "learning_rate": 0.00040619590160634485, "loss": 1.1673, "step": 93700 }, { "epoch": 3.77, "learning_rate": 0.0004061858367889207, "loss": 1.1443, "step": 93710 }, { "epoch": 3.77, "learning_rate": 0.0004061757719714965, "loss": 1.0274, "step": 93720 }, { "epoch": 3.77, "learning_rate": 0.0004061657071540723, "loss": 0.9821, "step": 93730 }, { "epoch": 3.77, "learning_rate": 0.000406155642336648, "loss": 1.0269, "step": 93740 }, { "epoch": 3.77, "learning_rate": 0.0004061455775192238, "loss": 0.9055, "step": 93750 }, { "epoch": 3.77, "learning_rate": 0.0004061355127017996, "loss": 1.2525, "step": 93760 }, { "epoch": 3.77, "learning_rate": 0.0004061254478843754, "loss": 1.1935, "step": 93770 }, { "epoch": 3.77, "learning_rate": 0.0004061153830669512, "loss": 0.988, "step": 93780 }, { "epoch": 3.77, "learning_rate": 0.00040610531824952697, "loss": 1.461, "step": 93790 }, { "epoch": 3.77, "learning_rate": 0.00040609525343210276, "loss": 1.1111, "step": 93800 }, { "epoch": 3.77, "learning_rate": 0.00040608518861467855, "loss": 1.4119, "step": 93810 }, { "epoch": 3.77, "learning_rate": 0.0004060751237972543, "loss": 1.1091, "step": 93820 }, { "epoch": 3.77, "learning_rate": 0.0004060650589798301, "loss": 1.0577, "step": 93830 }, { "epoch": 3.77, "learning_rate": 0.00040605499416240587, "loss": 1.1073, "step": 93840 }, { "epoch": 3.77, "learning_rate": 0.0004060449293449817, "loss": 1.2646, "step": 93850 }, { "epoch": 3.77, "learning_rate": 0.0004060348645275575, "loss": 1.1407, "step": 93860 }, { "epoch": 3.78, "learning_rate": 0.0004060247997101333, "loss": 1.2284, "step": 93870 }, { "epoch": 3.78, "learning_rate": 0.0004060147348927091, "loss": 1.169, "step": 93880 }, { "epoch": 3.78, "learning_rate": 0.0004060046700752849, "loss": 1.1505, "step": 93890 }, { "epoch": 3.78, "learning_rate": 0.0004059946052578606, "loss": 1.1883, "step": 93900 }, { "epoch": 3.78, "learning_rate": 0.0004059845404404364, "loss": 1.1346, "step": 93910 }, { "epoch": 3.78, "learning_rate": 0.0004059744756230122, "loss": 0.9784, "step": 93920 }, { "epoch": 3.78, "learning_rate": 0.000405964410805588, "loss": 1.0667, "step": 93930 }, { "epoch": 3.78, "learning_rate": 0.0004059543459881638, "loss": 0.974, "step": 93940 }, { "epoch": 3.78, "learning_rate": 0.00040594428117073957, "loss": 1.1175, "step": 93950 }, { "epoch": 3.78, "learning_rate": 0.00040593421635331536, "loss": 1.1271, "step": 93960 }, { "epoch": 3.78, "learning_rate": 0.00040592415153589115, "loss": 1.2466, "step": 93970 }, { "epoch": 3.78, "learning_rate": 0.00040591408671846694, "loss": 1.2862, "step": 93980 }, { "epoch": 3.78, "learning_rate": 0.0004059040219010427, "loss": 1.1574, "step": 93990 }, { "epoch": 3.78, "learning_rate": 0.0004058939570836185, "loss": 1.1612, "step": 94000 }, { "epoch": 3.78, "learning_rate": 0.0004058838922661943, "loss": 1.0497, "step": 94010 }, { "epoch": 3.78, "learning_rate": 0.0004058738274487701, "loss": 0.8634, "step": 94020 }, { "epoch": 3.78, "learning_rate": 0.0004058637626313459, "loss": 1.0779, "step": 94030 }, { "epoch": 3.78, "learning_rate": 0.0004058536978139217, "loss": 1.1709, "step": 94040 }, { "epoch": 3.78, "learning_rate": 0.0004058436329964975, "loss": 1.2021, "step": 94050 }, { "epoch": 3.78, "learning_rate": 0.00040583356817907326, "loss": 0.9397, "step": 94060 }, { "epoch": 3.78, "learning_rate": 0.000405823503361649, "loss": 1.0955, "step": 94070 }, { "epoch": 3.78, "learning_rate": 0.0004058134385442248, "loss": 1.2912, "step": 94080 }, { "epoch": 3.78, "learning_rate": 0.0004058033737268006, "loss": 1.3062, "step": 94090 }, { "epoch": 3.78, "learning_rate": 0.0004057933089093764, "loss": 1.2659, "step": 94100 }, { "epoch": 3.78, "learning_rate": 0.00040578324409195216, "loss": 1.2423, "step": 94110 }, { "epoch": 3.79, "learning_rate": 0.00040577317927452795, "loss": 1.1483, "step": 94120 }, { "epoch": 3.79, "learning_rate": 0.00040576311445710375, "loss": 1.1955, "step": 94130 }, { "epoch": 3.79, "learning_rate": 0.0004057530496396796, "loss": 0.9857, "step": 94140 }, { "epoch": 3.79, "learning_rate": 0.0004057429848222554, "loss": 1.1914, "step": 94150 }, { "epoch": 3.79, "learning_rate": 0.0004057329200048311, "loss": 0.923, "step": 94160 }, { "epoch": 3.79, "learning_rate": 0.0004057228551874069, "loss": 1.1, "step": 94170 }, { "epoch": 3.79, "learning_rate": 0.0004057127903699827, "loss": 1.3493, "step": 94180 }, { "epoch": 3.79, "learning_rate": 0.0004057027255525585, "loss": 1.2427, "step": 94190 }, { "epoch": 3.79, "learning_rate": 0.0004056926607351343, "loss": 1.2305, "step": 94200 }, { "epoch": 3.79, "learning_rate": 0.00040568259591771007, "loss": 1.1142, "step": 94210 }, { "epoch": 3.79, "learning_rate": 0.00040567253110028586, "loss": 1.0815, "step": 94220 }, { "epoch": 3.79, "learning_rate": 0.00040566246628286165, "loss": 1.0362, "step": 94230 }, { "epoch": 3.79, "learning_rate": 0.0004056524014654374, "loss": 1.1875, "step": 94240 }, { "epoch": 3.79, "learning_rate": 0.0004056423366480132, "loss": 1.4145, "step": 94250 }, { "epoch": 3.79, "learning_rate": 0.00040563227183058897, "loss": 1.2133, "step": 94260 }, { "epoch": 3.79, "learning_rate": 0.00040562220701316476, "loss": 1.4207, "step": 94270 }, { "epoch": 3.79, "learning_rate": 0.00040561214219574055, "loss": 1.1853, "step": 94280 }, { "epoch": 3.79, "learning_rate": 0.0004056020773783164, "loss": 1.0075, "step": 94290 }, { "epoch": 3.79, "learning_rate": 0.0004055920125608922, "loss": 1.3091, "step": 94300 }, { "epoch": 3.79, "learning_rate": 0.000405581947743468, "loss": 0.8871, "step": 94310 }, { "epoch": 3.79, "learning_rate": 0.00040557188292604377, "loss": 1.2199, "step": 94320 }, { "epoch": 3.79, "learning_rate": 0.0004055618181086195, "loss": 1.1168, "step": 94330 }, { "epoch": 3.79, "learning_rate": 0.0004055517532911953, "loss": 1.3357, "step": 94340 }, { "epoch": 3.79, "learning_rate": 0.0004055416884737711, "loss": 1.0252, "step": 94350 }, { "epoch": 3.8, "learning_rate": 0.0004055316236563469, "loss": 1.249, "step": 94360 }, { "epoch": 3.8, "learning_rate": 0.00040552155883892267, "loss": 1.0322, "step": 94370 }, { "epoch": 3.8, "learning_rate": 0.00040551149402149846, "loss": 1.1283, "step": 94380 }, { "epoch": 3.8, "learning_rate": 0.00040550142920407425, "loss": 1.2317, "step": 94390 }, { "epoch": 3.8, "learning_rate": 0.00040549136438665004, "loss": 1.2336, "step": 94400 }, { "epoch": 3.8, "learning_rate": 0.0004054812995692258, "loss": 1.338, "step": 94410 }, { "epoch": 3.8, "learning_rate": 0.00040547123475180157, "loss": 1.1699, "step": 94420 }, { "epoch": 3.8, "learning_rate": 0.00040546116993437736, "loss": 1.0197, "step": 94430 }, { "epoch": 3.8, "learning_rate": 0.0004054511051169532, "loss": 1.5022, "step": 94440 }, { "epoch": 3.8, "learning_rate": 0.000405441040299529, "loss": 0.9241, "step": 94450 }, { "epoch": 3.8, "learning_rate": 0.0004054309754821048, "loss": 1.1783, "step": 94460 }, { "epoch": 3.8, "learning_rate": 0.0004054209106646806, "loss": 1.5571, "step": 94470 }, { "epoch": 3.8, "learning_rate": 0.00040541084584725637, "loss": 1.2121, "step": 94480 }, { "epoch": 3.8, "learning_rate": 0.00040540078102983216, "loss": 1.1834, "step": 94490 }, { "epoch": 3.8, "learning_rate": 0.0004053907162124079, "loss": 1.2971, "step": 94500 }, { "epoch": 3.8, "learning_rate": 0.0004053806513949837, "loss": 1.2164, "step": 94510 }, { "epoch": 3.8, "learning_rate": 0.0004053705865775595, "loss": 1.0307, "step": 94520 }, { "epoch": 3.8, "learning_rate": 0.00040536052176013527, "loss": 1.2169, "step": 94530 }, { "epoch": 3.8, "learning_rate": 0.00040535045694271106, "loss": 1.006, "step": 94540 }, { "epoch": 3.8, "learning_rate": 0.00040534039212528685, "loss": 1.0656, "step": 94550 }, { "epoch": 3.8, "learning_rate": 0.00040533032730786264, "loss": 1.0319, "step": 94560 }, { "epoch": 3.8, "learning_rate": 0.00040532026249043843, "loss": 1.2206, "step": 94570 }, { "epoch": 3.8, "learning_rate": 0.0004053101976730142, "loss": 1.2802, "step": 94580 }, { "epoch": 3.8, "learning_rate": 0.00040530013285559, "loss": 1.1385, "step": 94590 }, { "epoch": 3.8, "learning_rate": 0.0004052900680381658, "loss": 1.5894, "step": 94600 }, { "epoch": 3.81, "learning_rate": 0.0004052800032207416, "loss": 1.2466, "step": 94610 }, { "epoch": 3.81, "learning_rate": 0.0004052699384033174, "loss": 1.2199, "step": 94620 }, { "epoch": 3.81, "learning_rate": 0.0004052598735858932, "loss": 1.0591, "step": 94630 }, { "epoch": 3.81, "learning_rate": 0.00040524980876846897, "loss": 1.2197, "step": 94640 }, { "epoch": 3.81, "learning_rate": 0.00040523974395104476, "loss": 1.1506, "step": 94650 }, { "epoch": 3.81, "learning_rate": 0.0004052296791336205, "loss": 1.0774, "step": 94660 }, { "epoch": 3.81, "learning_rate": 0.0004052196143161963, "loss": 1.3107, "step": 94670 }, { "epoch": 3.81, "learning_rate": 0.0004052095494987721, "loss": 1.0266, "step": 94680 }, { "epoch": 3.81, "learning_rate": 0.00040519948468134787, "loss": 1.1543, "step": 94690 }, { "epoch": 3.81, "learning_rate": 0.00040518941986392366, "loss": 0.9857, "step": 94700 }, { "epoch": 3.81, "learning_rate": 0.00040517935504649945, "loss": 1.1878, "step": 94710 }, { "epoch": 3.81, "learning_rate": 0.00040516929022907524, "loss": 1.1578, "step": 94720 }, { "epoch": 3.81, "learning_rate": 0.0004051592254116511, "loss": 1.2358, "step": 94730 }, { "epoch": 3.81, "learning_rate": 0.0004051491605942269, "loss": 1.1632, "step": 94740 }, { "epoch": 3.81, "learning_rate": 0.0004051390957768026, "loss": 1.0881, "step": 94750 }, { "epoch": 3.81, "learning_rate": 0.0004051290309593784, "loss": 1.0933, "step": 94760 }, { "epoch": 3.81, "learning_rate": 0.0004051189661419542, "loss": 1.0036, "step": 94770 }, { "epoch": 3.81, "learning_rate": 0.00040510890132453, "loss": 1.1844, "step": 94780 }, { "epoch": 3.81, "learning_rate": 0.0004050988365071058, "loss": 1.1709, "step": 94790 }, { "epoch": 3.81, "learning_rate": 0.00040508877168968157, "loss": 1.1003, "step": 94800 }, { "epoch": 3.81, "learning_rate": 0.00040507870687225736, "loss": 1.1332, "step": 94810 }, { "epoch": 3.81, "learning_rate": 0.00040506864205483315, "loss": 1.0599, "step": 94820 }, { "epoch": 3.81, "learning_rate": 0.0004050585772374089, "loss": 1.1324, "step": 94830 }, { "epoch": 3.81, "learning_rate": 0.0004050485124199847, "loss": 0.9411, "step": 94840 }, { "epoch": 3.81, "learning_rate": 0.00040503844760256046, "loss": 1.0374, "step": 94850 }, { "epoch": 3.82, "learning_rate": 0.00040502838278513626, "loss": 1.3244, "step": 94860 }, { "epoch": 3.82, "learning_rate": 0.00040501831796771205, "loss": 1.2977, "step": 94870 }, { "epoch": 3.82, "learning_rate": 0.0004050082531502879, "loss": 1.2636, "step": 94880 }, { "epoch": 3.82, "learning_rate": 0.0004049981883328637, "loss": 1.1608, "step": 94890 }, { "epoch": 3.82, "learning_rate": 0.0004049881235154395, "loss": 1.3302, "step": 94900 }, { "epoch": 3.82, "learning_rate": 0.00040497805869801526, "loss": 1.387, "step": 94910 }, { "epoch": 3.82, "learning_rate": 0.000404967993880591, "loss": 1.274, "step": 94920 }, { "epoch": 3.82, "learning_rate": 0.0004049579290631668, "loss": 1.1466, "step": 94930 }, { "epoch": 3.82, "learning_rate": 0.0004049478642457426, "loss": 1.1891, "step": 94940 }, { "epoch": 3.82, "learning_rate": 0.00040493779942831837, "loss": 1.1318, "step": 94950 }, { "epoch": 3.82, "learning_rate": 0.00040492773461089416, "loss": 1.2248, "step": 94960 }, { "epoch": 3.82, "learning_rate": 0.00040491766979346995, "loss": 1.4799, "step": 94970 }, { "epoch": 3.82, "learning_rate": 0.00040490760497604575, "loss": 0.9449, "step": 94980 }, { "epoch": 3.82, "learning_rate": 0.00040489754015862154, "loss": 1.2151, "step": 94990 }, { "epoch": 3.82, "learning_rate": 0.00040488747534119727, "loss": 0.9968, "step": 95000 }, { "epoch": 3.82, "learning_rate": 0.00040487741052377306, "loss": 1.1172, "step": 95010 }, { "epoch": 3.82, "learning_rate": 0.0004048673457063489, "loss": 1.0425, "step": 95020 }, { "epoch": 3.82, "learning_rate": 0.0004048572808889247, "loss": 1.2662, "step": 95030 }, { "epoch": 3.82, "learning_rate": 0.0004048472160715005, "loss": 1.1596, "step": 95040 }, { "epoch": 3.82, "learning_rate": 0.0004048371512540763, "loss": 0.9331, "step": 95050 }, { "epoch": 3.82, "learning_rate": 0.00040482708643665207, "loss": 1.3735, "step": 95060 }, { "epoch": 3.82, "learning_rate": 0.00040481702161922786, "loss": 1.2535, "step": 95070 }, { "epoch": 3.82, "learning_rate": 0.00040480695680180365, "loss": 1.0698, "step": 95080 }, { "epoch": 3.82, "learning_rate": 0.0004047968919843794, "loss": 1.2106, "step": 95090 }, { "epoch": 3.82, "learning_rate": 0.0004047868271669552, "loss": 1.1619, "step": 95100 }, { "epoch": 3.83, "learning_rate": 0.00040477676234953097, "loss": 1.2181, "step": 95110 }, { "epoch": 3.83, "learning_rate": 0.00040476669753210676, "loss": 1.2112, "step": 95120 }, { "epoch": 3.83, "learning_rate": 0.00040475663271468255, "loss": 1.1809, "step": 95130 }, { "epoch": 3.83, "learning_rate": 0.00040474656789725834, "loss": 1.2723, "step": 95140 }, { "epoch": 3.83, "learning_rate": 0.00040473650307983413, "loss": 1.3786, "step": 95150 }, { "epoch": 3.83, "learning_rate": 0.0004047264382624099, "loss": 0.9592, "step": 95160 }, { "epoch": 3.83, "learning_rate": 0.0004047163734449857, "loss": 1.1863, "step": 95170 }, { "epoch": 3.83, "learning_rate": 0.0004047063086275615, "loss": 1.0448, "step": 95180 }, { "epoch": 3.83, "learning_rate": 0.0004046962438101373, "loss": 1.0078, "step": 95190 }, { "epoch": 3.83, "learning_rate": 0.0004046861789927131, "loss": 1.1302, "step": 95200 }, { "epoch": 3.83, "learning_rate": 0.0004046761141752889, "loss": 0.9352, "step": 95210 }, { "epoch": 3.83, "learning_rate": 0.00040466604935786467, "loss": 1.0128, "step": 95220 }, { "epoch": 3.83, "learning_rate": 0.00040465598454044046, "loss": 1.0036, "step": 95230 }, { "epoch": 3.83, "learning_rate": 0.00040464591972301625, "loss": 1.048, "step": 95240 }, { "epoch": 3.83, "learning_rate": 0.00040463585490559204, "loss": 1.1624, "step": 95250 }, { "epoch": 3.83, "learning_rate": 0.0004046257900881678, "loss": 1.0478, "step": 95260 }, { "epoch": 3.83, "learning_rate": 0.00040461572527074357, "loss": 1.1413, "step": 95270 }, { "epoch": 3.83, "learning_rate": 0.00040460566045331936, "loss": 1.1364, "step": 95280 }, { "epoch": 3.83, "learning_rate": 0.00040459559563589515, "loss": 1.1755, "step": 95290 }, { "epoch": 3.83, "learning_rate": 0.00040458553081847094, "loss": 0.9963, "step": 95300 }, { "epoch": 3.83, "learning_rate": 0.0004045754660010468, "loss": 1.3838, "step": 95310 }, { "epoch": 3.83, "learning_rate": 0.0004045654011836226, "loss": 1.5393, "step": 95320 }, { "epoch": 3.83, "learning_rate": 0.00040455533636619837, "loss": 1.0635, "step": 95330 }, { "epoch": 3.83, "learning_rate": 0.0004045452715487741, "loss": 1.0878, "step": 95340 }, { "epoch": 3.83, "learning_rate": 0.0004045352067313499, "loss": 1.0243, "step": 95350 }, { "epoch": 3.84, "learning_rate": 0.0004045251419139257, "loss": 1.1274, "step": 95360 }, { "epoch": 3.84, "learning_rate": 0.0004045150770965015, "loss": 1.1714, "step": 95370 }, { "epoch": 3.84, "learning_rate": 0.00040450501227907727, "loss": 1.0926, "step": 95380 }, { "epoch": 3.84, "learning_rate": 0.00040449494746165306, "loss": 1.3279, "step": 95390 }, { "epoch": 3.84, "learning_rate": 0.00040448488264422885, "loss": 1.1584, "step": 95400 }, { "epoch": 3.84, "learning_rate": 0.00040447481782680464, "loss": 1.1368, "step": 95410 }, { "epoch": 3.84, "learning_rate": 0.0004044647530093804, "loss": 1.2142, "step": 95420 }, { "epoch": 3.84, "learning_rate": 0.00040445468819195617, "loss": 1.1526, "step": 95430 }, { "epoch": 3.84, "learning_rate": 0.00040444462337453196, "loss": 1.2608, "step": 95440 }, { "epoch": 3.84, "learning_rate": 0.00040443455855710775, "loss": 1.0605, "step": 95450 }, { "epoch": 3.84, "learning_rate": 0.0004044244937396836, "loss": 1.1476, "step": 95460 }, { "epoch": 3.84, "learning_rate": 0.0004044144289222594, "loss": 1.131, "step": 95470 }, { "epoch": 3.84, "learning_rate": 0.0004044043641048352, "loss": 1.1184, "step": 95480 }, { "epoch": 3.84, "learning_rate": 0.00040439429928741097, "loss": 0.9441, "step": 95490 }, { "epoch": 3.84, "learning_rate": 0.00040438423446998676, "loss": 1.2133, "step": 95500 }, { "epoch": 3.84, "learning_rate": 0.0004043741696525625, "loss": 1.0349, "step": 95510 }, { "epoch": 3.84, "learning_rate": 0.0004043641048351383, "loss": 1.1367, "step": 95520 }, { "epoch": 3.84, "learning_rate": 0.0004043540400177141, "loss": 1.1099, "step": 95530 }, { "epoch": 3.84, "learning_rate": 0.00040434397520028987, "loss": 1.2071, "step": 95540 }, { "epoch": 3.84, "learning_rate": 0.00040433391038286566, "loss": 1.2228, "step": 95550 }, { "epoch": 3.84, "learning_rate": 0.00040432384556544145, "loss": 1.3348, "step": 95560 }, { "epoch": 3.84, "learning_rate": 0.00040431378074801724, "loss": 1.1148, "step": 95570 }, { "epoch": 3.84, "learning_rate": 0.00040430371593059303, "loss": 1.2966, "step": 95580 }, { "epoch": 3.84, "learning_rate": 0.00040429365111316877, "loss": 1.2181, "step": 95590 }, { "epoch": 3.84, "learning_rate": 0.00040428358629574456, "loss": 1.0675, "step": 95600 }, { "epoch": 3.85, "learning_rate": 0.0004042735214783204, "loss": 1.0307, "step": 95610 }, { "epoch": 3.85, "learning_rate": 0.0004042634566608962, "loss": 1.1573, "step": 95620 }, { "epoch": 3.85, "learning_rate": 0.000404253391843472, "loss": 1.245, "step": 95630 }, { "epoch": 3.85, "learning_rate": 0.0004042433270260478, "loss": 1.0525, "step": 95640 }, { "epoch": 3.85, "learning_rate": 0.00040423326220862356, "loss": 1.173, "step": 95650 }, { "epoch": 3.85, "learning_rate": 0.00040422319739119936, "loss": 1.0302, "step": 95660 }, { "epoch": 3.85, "learning_rate": 0.00040421313257377515, "loss": 1.2017, "step": 95670 }, { "epoch": 3.85, "learning_rate": 0.0004042030677563509, "loss": 1.0351, "step": 95680 }, { "epoch": 3.85, "learning_rate": 0.0004041930029389267, "loss": 1.353, "step": 95690 }, { "epoch": 3.85, "learning_rate": 0.00040418293812150246, "loss": 1.1305, "step": 95700 }, { "epoch": 3.85, "learning_rate": 0.00040417287330407826, "loss": 1.0681, "step": 95710 }, { "epoch": 3.85, "learning_rate": 0.00040416280848665405, "loss": 1.0573, "step": 95720 }, { "epoch": 3.85, "learning_rate": 0.00040415274366922984, "loss": 0.9148, "step": 95730 }, { "epoch": 3.85, "learning_rate": 0.00040414267885180563, "loss": 1.2531, "step": 95740 }, { "epoch": 3.85, "learning_rate": 0.00040413261403438147, "loss": 1.0593, "step": 95750 }, { "epoch": 3.85, "learning_rate": 0.0004041225492169572, "loss": 1.3586, "step": 95760 }, { "epoch": 3.85, "learning_rate": 0.000404112484399533, "loss": 1.0663, "step": 95770 }, { "epoch": 3.85, "learning_rate": 0.0004041024195821088, "loss": 1.322, "step": 95780 }, { "epoch": 3.85, "learning_rate": 0.0004040923547646846, "loss": 1.1686, "step": 95790 }, { "epoch": 3.85, "learning_rate": 0.00040408228994726037, "loss": 1.1436, "step": 95800 }, { "epoch": 3.85, "learning_rate": 0.00040407222512983616, "loss": 0.9613, "step": 95810 }, { "epoch": 3.85, "learning_rate": 0.00040406216031241195, "loss": 1.5407, "step": 95820 }, { "epoch": 3.85, "learning_rate": 0.00040405209549498774, "loss": 1.33, "step": 95830 }, { "epoch": 3.85, "learning_rate": 0.00040404203067756354, "loss": 1.1579, "step": 95840 }, { "epoch": 3.85, "learning_rate": 0.00040403196586013927, "loss": 1.1215, "step": 95850 }, { "epoch": 3.86, "learning_rate": 0.00040402190104271506, "loss": 1.4942, "step": 95860 }, { "epoch": 3.86, "learning_rate": 0.00040401183622529085, "loss": 0.9471, "step": 95870 }, { "epoch": 3.86, "learning_rate": 0.00040400177140786664, "loss": 1.0611, "step": 95880 }, { "epoch": 3.86, "learning_rate": 0.00040399170659044244, "loss": 1.2063, "step": 95890 }, { "epoch": 3.86, "learning_rate": 0.0004039816417730183, "loss": 1.192, "step": 95900 }, { "epoch": 3.86, "learning_rate": 0.00040397157695559407, "loss": 1.3241, "step": 95910 }, { "epoch": 3.86, "learning_rate": 0.00040396151213816986, "loss": 1.0242, "step": 95920 }, { "epoch": 3.86, "learning_rate": 0.0004039514473207456, "loss": 1.0189, "step": 95930 }, { "epoch": 3.86, "learning_rate": 0.0004039413825033214, "loss": 1.2715, "step": 95940 }, { "epoch": 3.86, "learning_rate": 0.0004039313176858972, "loss": 0.9612, "step": 95950 }, { "epoch": 3.86, "learning_rate": 0.00040392125286847297, "loss": 1.1317, "step": 95960 }, { "epoch": 3.86, "learning_rate": 0.00040391118805104876, "loss": 1.0675, "step": 95970 }, { "epoch": 3.86, "learning_rate": 0.00040390112323362455, "loss": 1.3193, "step": 95980 }, { "epoch": 3.86, "learning_rate": 0.00040389105841620034, "loss": 1.0296, "step": 95990 }, { "epoch": 3.86, "learning_rate": 0.00040388099359877613, "loss": 1.0318, "step": 96000 }, { "epoch": 3.86, "learning_rate": 0.0004038709287813519, "loss": 1.1403, "step": 96010 }, { "epoch": 3.86, "learning_rate": 0.00040386086396392766, "loss": 1.1522, "step": 96020 }, { "epoch": 3.86, "learning_rate": 0.00040385079914650345, "loss": 1.1213, "step": 96030 }, { "epoch": 3.86, "learning_rate": 0.00040384073432907924, "loss": 1.1674, "step": 96040 }, { "epoch": 3.86, "learning_rate": 0.0004038306695116551, "loss": 1.1785, "step": 96050 }, { "epoch": 3.86, "learning_rate": 0.0004038206046942309, "loss": 1.1998, "step": 96060 }, { "epoch": 3.86, "learning_rate": 0.00040381053987680667, "loss": 1.0484, "step": 96070 }, { "epoch": 3.86, "learning_rate": 0.00040380047505938246, "loss": 0.9995, "step": 96080 }, { "epoch": 3.86, "learning_rate": 0.00040379041024195825, "loss": 1.201, "step": 96090 }, { "epoch": 3.86, "learning_rate": 0.000403780345424534, "loss": 1.1961, "step": 96100 }, { "epoch": 3.87, "learning_rate": 0.0004037702806071098, "loss": 1.0412, "step": 96110 }, { "epoch": 3.87, "learning_rate": 0.00040376021578968557, "loss": 1.4587, "step": 96120 }, { "epoch": 3.87, "learning_rate": 0.00040375015097226136, "loss": 1.1204, "step": 96130 }, { "epoch": 3.87, "learning_rate": 0.00040374008615483715, "loss": 1.1077, "step": 96140 }, { "epoch": 3.87, "learning_rate": 0.00040373002133741294, "loss": 1.1594, "step": 96150 }, { "epoch": 3.87, "learning_rate": 0.00040371995651998873, "loss": 1.3314, "step": 96160 }, { "epoch": 3.87, "learning_rate": 0.0004037098917025645, "loss": 1.2153, "step": 96170 }, { "epoch": 3.87, "learning_rate": 0.00040369982688514026, "loss": 1.1411, "step": 96180 }, { "epoch": 3.87, "learning_rate": 0.0004036897620677161, "loss": 1.052, "step": 96190 }, { "epoch": 3.87, "learning_rate": 0.0004036796972502919, "loss": 1.1531, "step": 96200 }, { "epoch": 3.87, "learning_rate": 0.0004036696324328677, "loss": 1.0277, "step": 96210 }, { "epoch": 3.87, "learning_rate": 0.0004036595676154435, "loss": 1.3133, "step": 96220 }, { "epoch": 3.87, "learning_rate": 0.00040364950279801927, "loss": 1.141, "step": 96230 }, { "epoch": 3.87, "learning_rate": 0.00040363943798059506, "loss": 1.1518, "step": 96240 }, { "epoch": 3.87, "learning_rate": 0.00040362937316317085, "loss": 1.2011, "step": 96250 }, { "epoch": 3.87, "learning_rate": 0.00040361930834574664, "loss": 1.1234, "step": 96260 }, { "epoch": 3.87, "learning_rate": 0.0004036092435283224, "loss": 1.0298, "step": 96270 }, { "epoch": 3.87, "learning_rate": 0.00040359917871089817, "loss": 1.1541, "step": 96280 }, { "epoch": 3.87, "learning_rate": 0.00040358911389347396, "loss": 1.2404, "step": 96290 }, { "epoch": 3.87, "learning_rate": 0.00040357904907604975, "loss": 1.0605, "step": 96300 }, { "epoch": 3.87, "learning_rate": 0.00040356898425862554, "loss": 0.938, "step": 96310 }, { "epoch": 3.87, "learning_rate": 0.00040355891944120133, "loss": 1.0427, "step": 96320 }, { "epoch": 3.87, "learning_rate": 0.0004035488546237771, "loss": 1.3369, "step": 96330 }, { "epoch": 3.87, "learning_rate": 0.00040353878980635297, "loss": 1.2498, "step": 96340 }, { "epoch": 3.88, "learning_rate": 0.00040352872498892876, "loss": 1.2607, "step": 96350 }, { "epoch": 3.88, "learning_rate": 0.0004035186601715045, "loss": 1.2502, "step": 96360 }, { "epoch": 3.88, "learning_rate": 0.0004035085953540803, "loss": 1.0665, "step": 96370 }, { "epoch": 3.88, "learning_rate": 0.0004034985305366561, "loss": 0.9611, "step": 96380 }, { "epoch": 3.88, "learning_rate": 0.00040348846571923187, "loss": 1.0634, "step": 96390 }, { "epoch": 3.88, "learning_rate": 0.00040347840090180766, "loss": 1.2266, "step": 96400 }, { "epoch": 3.88, "learning_rate": 0.00040346833608438345, "loss": 0.9848, "step": 96410 }, { "epoch": 3.88, "learning_rate": 0.00040345827126695924, "loss": 1.2587, "step": 96420 }, { "epoch": 3.88, "learning_rate": 0.00040344820644953503, "loss": 1.2459, "step": 96430 }, { "epoch": 3.88, "learning_rate": 0.00040343814163211077, "loss": 1.105, "step": 96440 }, { "epoch": 3.88, "learning_rate": 0.00040342807681468656, "loss": 1.1077, "step": 96450 }, { "epoch": 3.88, "learning_rate": 0.00040341801199726235, "loss": 1.0607, "step": 96460 }, { "epoch": 3.88, "learning_rate": 0.00040340794717983814, "loss": 1.1118, "step": 96470 }, { "epoch": 3.88, "learning_rate": 0.000403397882362414, "loss": 1.0264, "step": 96480 }, { "epoch": 3.88, "learning_rate": 0.0004033878175449898, "loss": 1.2724, "step": 96490 }, { "epoch": 3.88, "learning_rate": 0.00040337775272756556, "loss": 1.1556, "step": 96500 }, { "epoch": 3.88, "learning_rate": 0.00040336768791014136, "loss": 1.0679, "step": 96510 }, { "epoch": 3.88, "learning_rate": 0.0004033576230927171, "loss": 1.0281, "step": 96520 }, { "epoch": 3.88, "learning_rate": 0.0004033475582752929, "loss": 1.2757, "step": 96530 }, { "epoch": 3.88, "learning_rate": 0.0004033374934578687, "loss": 1.2701, "step": 96540 }, { "epoch": 3.88, "learning_rate": 0.00040332742864044446, "loss": 0.9734, "step": 96550 }, { "epoch": 3.88, "learning_rate": 0.00040331736382302026, "loss": 1.0175, "step": 96560 }, { "epoch": 3.88, "learning_rate": 0.00040330729900559605, "loss": 1.0967, "step": 96570 }, { "epoch": 3.88, "learning_rate": 0.00040329723418817184, "loss": 1.1026, "step": 96580 }, { "epoch": 3.88, "learning_rate": 0.00040328716937074763, "loss": 1.3711, "step": 96590 }, { "epoch": 3.89, "learning_rate": 0.0004032771045533234, "loss": 1.246, "step": 96600 }, { "epoch": 3.89, "learning_rate": 0.00040326703973589915, "loss": 1.2109, "step": 96610 }, { "epoch": 3.89, "learning_rate": 0.00040325697491847495, "loss": 1.1305, "step": 96620 }, { "epoch": 3.89, "learning_rate": 0.0004032469101010508, "loss": 1.1487, "step": 96630 }, { "epoch": 3.89, "learning_rate": 0.0004032368452836266, "loss": 1.0369, "step": 96640 }, { "epoch": 3.89, "learning_rate": 0.00040322678046620237, "loss": 0.96, "step": 96650 }, { "epoch": 3.89, "learning_rate": 0.00040321671564877816, "loss": 0.9846, "step": 96660 }, { "epoch": 3.89, "learning_rate": 0.00040320665083135395, "loss": 1.1207, "step": 96670 }, { "epoch": 3.89, "learning_rate": 0.00040319658601392974, "loss": 1.2486, "step": 96680 }, { "epoch": 3.89, "learning_rate": 0.0004031865211965055, "loss": 1.0816, "step": 96690 }, { "epoch": 3.89, "learning_rate": 0.00040317645637908127, "loss": 1.0269, "step": 96700 }, { "epoch": 3.89, "learning_rate": 0.00040316639156165706, "loss": 1.0841, "step": 96710 }, { "epoch": 3.89, "learning_rate": 0.00040315632674423285, "loss": 0.9721, "step": 96720 }, { "epoch": 3.89, "learning_rate": 0.00040314626192680864, "loss": 1.1826, "step": 96730 }, { "epoch": 3.89, "learning_rate": 0.00040313619710938444, "loss": 1.1808, "step": 96740 }, { "epoch": 3.89, "learning_rate": 0.0004031261322919602, "loss": 1.1142, "step": 96750 }, { "epoch": 3.89, "learning_rate": 0.000403116067474536, "loss": 1.1404, "step": 96760 }, { "epoch": 3.89, "learning_rate": 0.0004031060026571118, "loss": 1.3099, "step": 96770 }, { "epoch": 3.89, "learning_rate": 0.0004030959378396876, "loss": 0.9851, "step": 96780 }, { "epoch": 3.89, "learning_rate": 0.0004030858730222634, "loss": 1.133, "step": 96790 }, { "epoch": 3.89, "learning_rate": 0.0004030758082048392, "loss": 1.1358, "step": 96800 }, { "epoch": 3.89, "learning_rate": 0.00040306574338741497, "loss": 1.1231, "step": 96810 }, { "epoch": 3.89, "learning_rate": 0.00040305567856999076, "loss": 1.0981, "step": 96820 }, { "epoch": 3.89, "learning_rate": 0.00040304561375256655, "loss": 1.3192, "step": 96830 }, { "epoch": 3.89, "learning_rate": 0.00040303554893514234, "loss": 1.0051, "step": 96840 }, { "epoch": 3.9, "learning_rate": 0.00040302548411771813, "loss": 1.2558, "step": 96850 }, { "epoch": 3.9, "learning_rate": 0.00040301541930029387, "loss": 1.2474, "step": 96860 }, { "epoch": 3.9, "learning_rate": 0.00040300535448286966, "loss": 1.214, "step": 96870 }, { "epoch": 3.9, "learning_rate": 0.00040299528966544545, "loss": 1.3026, "step": 96880 }, { "epoch": 3.9, "learning_rate": 0.00040298522484802124, "loss": 1.1895, "step": 96890 }, { "epoch": 3.9, "learning_rate": 0.00040297516003059703, "loss": 1.1813, "step": 96900 }, { "epoch": 3.9, "learning_rate": 0.0004029650952131728, "loss": 1.2118, "step": 96910 }, { "epoch": 3.9, "learning_rate": 0.00040295503039574867, "loss": 1.3171, "step": 96920 }, { "epoch": 3.9, "learning_rate": 0.00040294496557832446, "loss": 1.3301, "step": 96930 }, { "epoch": 3.9, "learning_rate": 0.00040293490076090025, "loss": 1.1397, "step": 96940 }, { "epoch": 3.9, "learning_rate": 0.000402924835943476, "loss": 1.0908, "step": 96950 }, { "epoch": 3.9, "learning_rate": 0.0004029147711260518, "loss": 1.0543, "step": 96960 }, { "epoch": 3.9, "learning_rate": 0.00040290470630862757, "loss": 1.1876, "step": 96970 }, { "epoch": 3.9, "learning_rate": 0.00040289464149120336, "loss": 1.1084, "step": 96980 }, { "epoch": 3.9, "learning_rate": 0.00040288457667377915, "loss": 1.1848, "step": 96990 }, { "epoch": 3.9, "learning_rate": 0.00040287451185635494, "loss": 1.1216, "step": 97000 }, { "epoch": 3.9, "learning_rate": 0.00040286444703893073, "loss": 1.1239, "step": 97010 }, { "epoch": 3.9, "learning_rate": 0.0004028543822215065, "loss": 1.0874, "step": 97020 }, { "epoch": 3.9, "learning_rate": 0.00040284431740408226, "loss": 1.0981, "step": 97030 }, { "epoch": 3.9, "learning_rate": 0.00040283425258665805, "loss": 0.9799, "step": 97040 }, { "epoch": 3.9, "learning_rate": 0.00040282418776923384, "loss": 0.989, "step": 97050 }, { "epoch": 3.9, "learning_rate": 0.00040281412295180963, "loss": 1.0378, "step": 97060 }, { "epoch": 3.9, "learning_rate": 0.0004028040581343855, "loss": 1.0624, "step": 97070 }, { "epoch": 3.9, "learning_rate": 0.00040279399331696127, "loss": 1.0019, "step": 97080 }, { "epoch": 3.9, "learning_rate": 0.00040278392849953706, "loss": 1.161, "step": 97090 }, { "epoch": 3.91, "learning_rate": 0.00040277386368211285, "loss": 1.0674, "step": 97100 }, { "epoch": 3.91, "learning_rate": 0.00040276379886468864, "loss": 1.1156, "step": 97110 }, { "epoch": 3.91, "learning_rate": 0.0004027537340472644, "loss": 0.9089, "step": 97120 }, { "epoch": 3.91, "learning_rate": 0.00040274366922984017, "loss": 1.223, "step": 97130 }, { "epoch": 3.91, "learning_rate": 0.00040273360441241596, "loss": 1.1249, "step": 97140 }, { "epoch": 3.91, "learning_rate": 0.00040272353959499175, "loss": 1.3317, "step": 97150 }, { "epoch": 3.91, "learning_rate": 0.00040271347477756754, "loss": 1.1854, "step": 97160 }, { "epoch": 3.91, "learning_rate": 0.00040270340996014333, "loss": 1.3749, "step": 97170 }, { "epoch": 3.91, "learning_rate": 0.0004026933451427191, "loss": 0.7831, "step": 97180 }, { "epoch": 3.91, "learning_rate": 0.0004026832803252949, "loss": 1.3451, "step": 97190 }, { "epoch": 3.91, "learning_rate": 0.00040267321550787065, "loss": 1.1097, "step": 97200 }, { "epoch": 3.91, "learning_rate": 0.0004026631506904465, "loss": 1.1405, "step": 97210 }, { "epoch": 3.91, "learning_rate": 0.0004026530858730223, "loss": 0.8407, "step": 97220 }, { "epoch": 3.91, "learning_rate": 0.0004026430210555981, "loss": 1.0702, "step": 97230 }, { "epoch": 3.91, "learning_rate": 0.00040263295623817387, "loss": 0.9878, "step": 97240 }, { "epoch": 3.91, "learning_rate": 0.00040262289142074966, "loss": 1.1633, "step": 97250 }, { "epoch": 3.91, "learning_rate": 0.00040261282660332545, "loss": 1.2416, "step": 97260 }, { "epoch": 3.91, "learning_rate": 0.00040260276178590124, "loss": 0.9571, "step": 97270 }, { "epoch": 3.91, "learning_rate": 0.000402592696968477, "loss": 0.9799, "step": 97280 }, { "epoch": 3.91, "learning_rate": 0.00040258263215105277, "loss": 1.118, "step": 97290 }, { "epoch": 3.91, "learning_rate": 0.00040257256733362856, "loss": 1.0853, "step": 97300 }, { "epoch": 3.91, "learning_rate": 0.00040256250251620435, "loss": 1.372, "step": 97310 }, { "epoch": 3.91, "learning_rate": 0.00040255243769878014, "loss": 1.1423, "step": 97320 }, { "epoch": 3.91, "learning_rate": 0.00040254237288135593, "loss": 1.1956, "step": 97330 }, { "epoch": 3.91, "learning_rate": 0.0004025323080639317, "loss": 1.2536, "step": 97340 }, { "epoch": 3.92, "learning_rate": 0.0004025222432465075, "loss": 1.1518, "step": 97350 }, { "epoch": 3.92, "learning_rate": 0.00040251217842908336, "loss": 1.338, "step": 97360 }, { "epoch": 3.92, "learning_rate": 0.0004025021136116591, "loss": 1.293, "step": 97370 }, { "epoch": 3.92, "learning_rate": 0.0004024920487942349, "loss": 1.1473, "step": 97380 }, { "epoch": 3.92, "learning_rate": 0.0004024819839768107, "loss": 1.1982, "step": 97390 }, { "epoch": 3.92, "learning_rate": 0.00040247191915938646, "loss": 1.3391, "step": 97400 }, { "epoch": 3.92, "learning_rate": 0.00040246185434196225, "loss": 0.9392, "step": 97410 }, { "epoch": 3.92, "learning_rate": 0.00040245178952453805, "loss": 1.1466, "step": 97420 }, { "epoch": 3.92, "learning_rate": 0.00040244172470711384, "loss": 1.0626, "step": 97430 }, { "epoch": 3.92, "learning_rate": 0.00040243165988968963, "loss": 1.0696, "step": 97440 }, { "epoch": 3.92, "learning_rate": 0.00040242159507226536, "loss": 1.1544, "step": 97450 }, { "epoch": 3.92, "learning_rate": 0.00040241153025484115, "loss": 0.9007, "step": 97460 }, { "epoch": 3.92, "learning_rate": 0.00040240146543741695, "loss": 1.0281, "step": 97470 }, { "epoch": 3.92, "learning_rate": 0.00040239140061999274, "loss": 1.1969, "step": 97480 }, { "epoch": 3.92, "learning_rate": 0.0004023813358025685, "loss": 0.9938, "step": 97490 }, { "epoch": 3.92, "learning_rate": 0.0004023712709851443, "loss": 1.248, "step": 97500 }, { "epoch": 3.92, "learning_rate": 0.00040236120616772016, "loss": 1.1851, "step": 97510 }, { "epoch": 3.92, "learning_rate": 0.00040235114135029595, "loss": 1.1838, "step": 97520 }, { "epoch": 3.92, "learning_rate": 0.00040234107653287174, "loss": 1.0235, "step": 97530 }, { "epoch": 3.92, "learning_rate": 0.0004023310117154475, "loss": 1.1578, "step": 97540 }, { "epoch": 3.92, "learning_rate": 0.00040232094689802327, "loss": 1.1133, "step": 97550 }, { "epoch": 3.92, "learning_rate": 0.00040231088208059906, "loss": 1.2995, "step": 97560 }, { "epoch": 3.92, "learning_rate": 0.00040230081726317485, "loss": 1.231, "step": 97570 }, { "epoch": 3.92, "learning_rate": 0.00040229075244575064, "loss": 1.2747, "step": 97580 }, { "epoch": 3.92, "learning_rate": 0.00040228068762832643, "loss": 1.2948, "step": 97590 }, { "epoch": 3.93, "learning_rate": 0.0004022706228109022, "loss": 1.2106, "step": 97600 }, { "epoch": 3.93, "learning_rate": 0.000402260557993478, "loss": 1.3471, "step": 97610 }, { "epoch": 3.93, "learning_rate": 0.00040225049317605375, "loss": 1.2945, "step": 97620 }, { "epoch": 3.93, "learning_rate": 0.00040224042835862954, "loss": 1.0677, "step": 97630 }, { "epoch": 3.93, "learning_rate": 0.00040223036354120533, "loss": 1.0227, "step": 97640 }, { "epoch": 3.93, "learning_rate": 0.0004022202987237812, "loss": 1.2025, "step": 97650 }, { "epoch": 3.93, "learning_rate": 0.00040221023390635697, "loss": 1.2205, "step": 97660 }, { "epoch": 3.93, "learning_rate": 0.00040220016908893276, "loss": 1.1256, "step": 97670 }, { "epoch": 3.93, "learning_rate": 0.00040219010427150855, "loss": 1.0924, "step": 97680 }, { "epoch": 3.93, "learning_rate": 0.00040218003945408434, "loss": 1.2727, "step": 97690 }, { "epoch": 3.93, "learning_rate": 0.00040216997463666013, "loss": 1.4737, "step": 97700 }, { "epoch": 3.93, "learning_rate": 0.00040215990981923587, "loss": 1.2208, "step": 97710 }, { "epoch": 3.93, "learning_rate": 0.00040214984500181166, "loss": 1.3158, "step": 97720 }, { "epoch": 3.93, "learning_rate": 0.00040213978018438745, "loss": 1.2669, "step": 97730 }, { "epoch": 3.93, "learning_rate": 0.00040212971536696324, "loss": 1.0107, "step": 97740 }, { "epoch": 3.93, "learning_rate": 0.00040211965054953903, "loss": 1.1175, "step": 97750 }, { "epoch": 3.93, "learning_rate": 0.0004021095857321148, "loss": 1.092, "step": 97760 }, { "epoch": 3.93, "learning_rate": 0.0004020995209146906, "loss": 1.2122, "step": 97770 }, { "epoch": 3.93, "learning_rate": 0.0004020894560972664, "loss": 1.4323, "step": 97780 }, { "epoch": 3.93, "learning_rate": 0.00040207939127984214, "loss": 1.132, "step": 97790 }, { "epoch": 3.93, "learning_rate": 0.000402069326462418, "loss": 1.1812, "step": 97800 }, { "epoch": 3.93, "learning_rate": 0.0004020592616449938, "loss": 1.0435, "step": 97810 }, { "epoch": 3.93, "learning_rate": 0.00040204919682756957, "loss": 1.044, "step": 97820 }, { "epoch": 3.93, "learning_rate": 0.00040203913201014536, "loss": 1.2351, "step": 97830 }, { "epoch": 3.93, "learning_rate": 0.00040202906719272115, "loss": 1.2063, "step": 97840 }, { "epoch": 3.94, "learning_rate": 0.00040201900237529694, "loss": 1.1502, "step": 97850 }, { "epoch": 3.94, "learning_rate": 0.00040200893755787273, "loss": 1.014, "step": 97860 }, { "epoch": 3.94, "learning_rate": 0.0004019988727404485, "loss": 1.1814, "step": 97870 }, { "epoch": 3.94, "learning_rate": 0.00040198880792302426, "loss": 1.1639, "step": 97880 }, { "epoch": 3.94, "learning_rate": 0.00040197874310560005, "loss": 0.9632, "step": 97890 }, { "epoch": 3.94, "learning_rate": 0.00040196867828817584, "loss": 1.1963, "step": 97900 }, { "epoch": 3.94, "learning_rate": 0.00040195861347075163, "loss": 0.9874, "step": 97910 }, { "epoch": 3.94, "learning_rate": 0.0004019485486533274, "loss": 1.1587, "step": 97920 }, { "epoch": 3.94, "learning_rate": 0.0004019384838359032, "loss": 1.0538, "step": 97930 }, { "epoch": 3.94, "learning_rate": 0.000401928419018479, "loss": 1.4715, "step": 97940 }, { "epoch": 3.94, "learning_rate": 0.00040191835420105485, "loss": 1.3343, "step": 97950 }, { "epoch": 3.94, "learning_rate": 0.0004019082893836306, "loss": 0.9186, "step": 97960 }, { "epoch": 3.94, "learning_rate": 0.0004018982245662064, "loss": 1.3198, "step": 97970 }, { "epoch": 3.94, "learning_rate": 0.00040188815974878217, "loss": 1.1192, "step": 97980 }, { "epoch": 3.94, "learning_rate": 0.00040187809493135796, "loss": 1.0665, "step": 97990 }, { "epoch": 3.94, "learning_rate": 0.00040186803011393375, "loss": 1.1929, "step": 98000 }, { "epoch": 3.94, "learning_rate": 0.00040185796529650954, "loss": 1.1306, "step": 98010 }, { "epoch": 3.94, "learning_rate": 0.00040184790047908533, "loss": 0.9881, "step": 98020 }, { "epoch": 3.94, "learning_rate": 0.0004018378356616611, "loss": 1.0744, "step": 98030 }, { "epoch": 3.94, "learning_rate": 0.00040182777084423686, "loss": 1.2279, "step": 98040 }, { "epoch": 3.94, "learning_rate": 0.00040181770602681265, "loss": 1.2104, "step": 98050 }, { "epoch": 3.94, "learning_rate": 0.00040180764120938844, "loss": 1.1971, "step": 98060 }, { "epoch": 3.94, "learning_rate": 0.00040179757639196423, "loss": 0.9676, "step": 98070 }, { "epoch": 3.94, "learning_rate": 0.00040178751157454, "loss": 1.1452, "step": 98080 }, { "epoch": 3.95, "learning_rate": 0.00040177744675711587, "loss": 1.0648, "step": 98090 }, { "epoch": 3.95, "learning_rate": 0.00040176738193969166, "loss": 1.3313, "step": 98100 }, { "epoch": 3.95, "learning_rate": 0.00040175731712226745, "loss": 1.0862, "step": 98110 }, { "epoch": 3.95, "learning_rate": 0.00040174725230484324, "loss": 1.496, "step": 98120 }, { "epoch": 3.95, "learning_rate": 0.000401737187487419, "loss": 1.1645, "step": 98130 }, { "epoch": 3.95, "learning_rate": 0.00040172712266999477, "loss": 1.1382, "step": 98140 }, { "epoch": 3.95, "learning_rate": 0.00040171705785257056, "loss": 1.1424, "step": 98150 }, { "epoch": 3.95, "learning_rate": 0.00040170699303514635, "loss": 1.2275, "step": 98160 }, { "epoch": 3.95, "learning_rate": 0.00040169692821772214, "loss": 1.2313, "step": 98170 }, { "epoch": 3.95, "learning_rate": 0.00040168686340029793, "loss": 1.0829, "step": 98180 }, { "epoch": 3.95, "learning_rate": 0.0004016767985828737, "loss": 1.1836, "step": 98190 }, { "epoch": 3.95, "learning_rate": 0.0004016667337654495, "loss": 1.3405, "step": 98200 }, { "epoch": 3.95, "learning_rate": 0.00040165666894802525, "loss": 1.2805, "step": 98210 }, { "epoch": 3.95, "learning_rate": 0.00040164660413060104, "loss": 1.0345, "step": 98220 }, { "epoch": 3.95, "learning_rate": 0.00040163653931317683, "loss": 1.2001, "step": 98230 }, { "epoch": 3.95, "learning_rate": 0.0004016264744957527, "loss": 0.984, "step": 98240 }, { "epoch": 3.95, "learning_rate": 0.00040161640967832846, "loss": 1.1116, "step": 98250 }, { "epoch": 3.95, "learning_rate": 0.00040160634486090425, "loss": 1.2979, "step": 98260 }, { "epoch": 3.95, "learning_rate": 0.00040159628004348005, "loss": 1.3909, "step": 98270 }, { "epoch": 3.95, "learning_rate": 0.00040158621522605584, "loss": 1.2458, "step": 98280 }, { "epoch": 3.95, "learning_rate": 0.0004015761504086316, "loss": 1.1104, "step": 98290 }, { "epoch": 3.95, "learning_rate": 0.00040156608559120736, "loss": 1.2613, "step": 98300 }, { "epoch": 3.95, "learning_rate": 0.00040155602077378315, "loss": 1.1098, "step": 98310 }, { "epoch": 3.95, "learning_rate": 0.00040154595595635895, "loss": 1.175, "step": 98320 }, { "epoch": 3.95, "learning_rate": 0.00040153589113893474, "loss": 0.9547, "step": 98330 }, { "epoch": 3.96, "learning_rate": 0.0004015258263215105, "loss": 1.1822, "step": 98340 }, { "epoch": 3.96, "learning_rate": 0.0004015157615040863, "loss": 0.9516, "step": 98350 }, { "epoch": 3.96, "learning_rate": 0.0004015056966866621, "loss": 1.0911, "step": 98360 }, { "epoch": 3.96, "learning_rate": 0.0004014956318692379, "loss": 1.364, "step": 98370 }, { "epoch": 3.96, "learning_rate": 0.0004014855670518137, "loss": 1.0119, "step": 98380 }, { "epoch": 3.96, "learning_rate": 0.0004014755022343895, "loss": 1.1973, "step": 98390 }, { "epoch": 3.96, "learning_rate": 0.00040146543741696527, "loss": 1.1507, "step": 98400 }, { "epoch": 3.96, "learning_rate": 0.00040145537259954106, "loss": 0.9315, "step": 98410 }, { "epoch": 3.96, "learning_rate": 0.00040144530778211685, "loss": 1.2632, "step": 98420 }, { "epoch": 3.96, "learning_rate": 0.00040143524296469264, "loss": 1.1496, "step": 98430 }, { "epoch": 3.96, "learning_rate": 0.00040142517814726843, "loss": 0.9159, "step": 98440 }, { "epoch": 3.96, "learning_rate": 0.0004014151133298442, "loss": 1.1814, "step": 98450 }, { "epoch": 3.96, "learning_rate": 0.00040140504851242, "loss": 1.1747, "step": 98460 }, { "epoch": 3.96, "learning_rate": 0.00040139498369499575, "loss": 1.0406, "step": 98470 }, { "epoch": 3.96, "learning_rate": 0.00040138491887757154, "loss": 1.1826, "step": 98480 }, { "epoch": 3.96, "learning_rate": 0.00040137485406014733, "loss": 1.2047, "step": 98490 }, { "epoch": 3.96, "learning_rate": 0.0004013647892427231, "loss": 1.0255, "step": 98500 }, { "epoch": 3.96, "learning_rate": 0.0004013547244252989, "loss": 1.2487, "step": 98510 }, { "epoch": 3.96, "learning_rate": 0.0004013446596078747, "loss": 0.9437, "step": 98520 }, { "epoch": 3.96, "learning_rate": 0.00040133459479045055, "loss": 1.2141, "step": 98530 }, { "epoch": 3.96, "learning_rate": 0.00040132452997302634, "loss": 1.1807, "step": 98540 }, { "epoch": 3.96, "learning_rate": 0.0004013144651556021, "loss": 1.1509, "step": 98550 }, { "epoch": 3.96, "learning_rate": 0.00040130440033817787, "loss": 1.2014, "step": 98560 }, { "epoch": 3.96, "learning_rate": 0.00040129433552075366, "loss": 0.9102, "step": 98570 }, { "epoch": 3.96, "learning_rate": 0.00040128427070332945, "loss": 1.2663, "step": 98580 }, { "epoch": 3.97, "learning_rate": 0.00040127420588590524, "loss": 1.1539, "step": 98590 }, { "epoch": 3.97, "learning_rate": 0.00040126414106848103, "loss": 1.188, "step": 98600 }, { "epoch": 3.97, "learning_rate": 0.0004012540762510568, "loss": 1.3464, "step": 98610 }, { "epoch": 3.97, "learning_rate": 0.0004012440114336326, "loss": 1.2978, "step": 98620 }, { "epoch": 3.97, "learning_rate": 0.0004012339466162084, "loss": 1.1821, "step": 98630 }, { "epoch": 3.97, "learning_rate": 0.00040122388179878414, "loss": 1.1233, "step": 98640 }, { "epoch": 3.97, "learning_rate": 0.00040121381698135993, "loss": 1.1292, "step": 98650 }, { "epoch": 3.97, "learning_rate": 0.0004012037521639357, "loss": 1.0771, "step": 98660 }, { "epoch": 3.97, "learning_rate": 0.0004011936873465115, "loss": 1.2195, "step": 98670 }, { "epoch": 3.97, "learning_rate": 0.00040118362252908736, "loss": 1.0684, "step": 98680 }, { "epoch": 3.97, "learning_rate": 0.00040117355771166315, "loss": 1.1894, "step": 98690 }, { "epoch": 3.97, "learning_rate": 0.00040116349289423894, "loss": 1.0397, "step": 98700 }, { "epoch": 3.97, "learning_rate": 0.00040115342807681473, "loss": 1.2843, "step": 98710 }, { "epoch": 3.97, "learning_rate": 0.00040114336325939047, "loss": 1.0136, "step": 98720 }, { "epoch": 3.97, "learning_rate": 0.00040113329844196626, "loss": 1.1429, "step": 98730 }, { "epoch": 3.97, "learning_rate": 0.00040112323362454205, "loss": 1.1041, "step": 98740 }, { "epoch": 3.97, "learning_rate": 0.00040111316880711784, "loss": 1.0698, "step": 98750 }, { "epoch": 3.97, "learning_rate": 0.00040110310398969363, "loss": 1.3925, "step": 98760 }, { "epoch": 3.97, "learning_rate": 0.0004010930391722694, "loss": 1.4096, "step": 98770 }, { "epoch": 3.97, "learning_rate": 0.0004010829743548452, "loss": 1.1806, "step": 98780 }, { "epoch": 3.97, "learning_rate": 0.000401072909537421, "loss": 1.1701, "step": 98790 }, { "epoch": 3.97, "learning_rate": 0.00040106284471999674, "loss": 1.3398, "step": 98800 }, { "epoch": 3.97, "learning_rate": 0.00040105277990257253, "loss": 1.2258, "step": 98810 }, { "epoch": 3.97, "learning_rate": 0.0004010427150851484, "loss": 1.0766, "step": 98820 }, { "epoch": 3.97, "learning_rate": 0.00040103265026772417, "loss": 1.0587, "step": 98830 }, { "epoch": 3.98, "learning_rate": 0.00040102258545029996, "loss": 1.2707, "step": 98840 }, { "epoch": 3.98, "learning_rate": 0.00040101252063287575, "loss": 1.3885, "step": 98850 }, { "epoch": 3.98, "learning_rate": 0.00040100245581545154, "loss": 1.1187, "step": 98860 }, { "epoch": 3.98, "learning_rate": 0.00040099239099802733, "loss": 1.2176, "step": 98870 }, { "epoch": 3.98, "learning_rate": 0.0004009823261806031, "loss": 1.0885, "step": 98880 }, { "epoch": 3.98, "learning_rate": 0.00040097226136317886, "loss": 1.1058, "step": 98890 }, { "epoch": 3.98, "learning_rate": 0.00040096219654575465, "loss": 1.0451, "step": 98900 }, { "epoch": 3.98, "learning_rate": 0.00040095213172833044, "loss": 1.1293, "step": 98910 }, { "epoch": 3.98, "learning_rate": 0.00040094206691090623, "loss": 1.0655, "step": 98920 }, { "epoch": 3.98, "learning_rate": 0.000400932002093482, "loss": 1.181, "step": 98930 }, { "epoch": 3.98, "learning_rate": 0.0004009219372760578, "loss": 1.161, "step": 98940 }, { "epoch": 3.98, "learning_rate": 0.0004009118724586336, "loss": 1.3601, "step": 98950 }, { "epoch": 3.98, "learning_rate": 0.0004009018076412094, "loss": 1.1771, "step": 98960 }, { "epoch": 3.98, "learning_rate": 0.0004008917428237852, "loss": 1.3111, "step": 98970 }, { "epoch": 3.98, "learning_rate": 0.000400881678006361, "loss": 1.0306, "step": 98980 }, { "epoch": 3.98, "learning_rate": 0.00040087161318893677, "loss": 1.096, "step": 98990 }, { "epoch": 3.98, "learning_rate": 0.00040086154837151256, "loss": 1.2457, "step": 99000 }, { "epoch": 3.98, "learning_rate": 0.00040085148355408835, "loss": 1.3637, "step": 99010 }, { "epoch": 3.98, "learning_rate": 0.00040084141873666414, "loss": 1.022, "step": 99020 }, { "epoch": 3.98, "learning_rate": 0.00040083135391923993, "loss": 1.2285, "step": 99030 }, { "epoch": 3.98, "learning_rate": 0.0004008212891018157, "loss": 1.4043, "step": 99040 }, { "epoch": 3.98, "learning_rate": 0.0004008112242843915, "loss": 1.0117, "step": 99050 }, { "epoch": 3.98, "learning_rate": 0.00040080115946696725, "loss": 1.1754, "step": 99060 }, { "epoch": 3.98, "learning_rate": 0.00040079109464954304, "loss": 1.148, "step": 99070 }, { "epoch": 3.98, "learning_rate": 0.00040078102983211883, "loss": 1.2974, "step": 99080 }, { "epoch": 3.99, "learning_rate": 0.0004007709650146946, "loss": 1.2768, "step": 99090 }, { "epoch": 3.99, "learning_rate": 0.0004007609001972704, "loss": 1.2624, "step": 99100 }, { "epoch": 3.99, "learning_rate": 0.0004007508353798462, "loss": 1.0935, "step": 99110 }, { "epoch": 3.99, "learning_rate": 0.00040074077056242205, "loss": 1.1023, "step": 99120 }, { "epoch": 3.99, "learning_rate": 0.00040073070574499784, "loss": 1.0885, "step": 99130 }, { "epoch": 3.99, "learning_rate": 0.00040072064092757357, "loss": 1.1464, "step": 99140 }, { "epoch": 3.99, "learning_rate": 0.00040071057611014936, "loss": 0.9916, "step": 99150 }, { "epoch": 3.99, "learning_rate": 0.00040070051129272515, "loss": 1.2652, "step": 99160 }, { "epoch": 3.99, "learning_rate": 0.00040069044647530094, "loss": 1.0787, "step": 99170 }, { "epoch": 3.99, "learning_rate": 0.00040068038165787674, "loss": 1.1993, "step": 99180 }, { "epoch": 3.99, "learning_rate": 0.0004006703168404525, "loss": 1.3184, "step": 99190 }, { "epoch": 3.99, "learning_rate": 0.0004006602520230283, "loss": 1.1926, "step": 99200 }, { "epoch": 3.99, "learning_rate": 0.0004006501872056041, "loss": 0.9925, "step": 99210 }, { "epoch": 3.99, "learning_rate": 0.0004006401223881799, "loss": 1.1001, "step": 99220 }, { "epoch": 3.99, "learning_rate": 0.00040063005757075564, "loss": 1.2335, "step": 99230 }, { "epoch": 3.99, "learning_rate": 0.0004006199927533314, "loss": 1.1535, "step": 99240 }, { "epoch": 3.99, "learning_rate": 0.0004006099279359072, "loss": 1.2041, "step": 99250 }, { "epoch": 3.99, "learning_rate": 0.00040059986311848306, "loss": 1.2252, "step": 99260 }, { "epoch": 3.99, "learning_rate": 0.00040058979830105885, "loss": 1.0336, "step": 99270 }, { "epoch": 3.99, "learning_rate": 0.00040057973348363464, "loss": 1.0705, "step": 99280 }, { "epoch": 3.99, "learning_rate": 0.00040056966866621043, "loss": 0.9134, "step": 99290 }, { "epoch": 3.99, "learning_rate": 0.0004005596038487862, "loss": 1.1003, "step": 99300 }, { "epoch": 3.99, "learning_rate": 0.00040054953903136196, "loss": 1.1663, "step": 99310 }, { "epoch": 3.99, "learning_rate": 0.00040053947421393775, "loss": 1.0961, "step": 99320 }, { "epoch": 3.99, "learning_rate": 0.00040052940939651354, "loss": 1.1326, "step": 99330 }, { "epoch": 4.0, "learning_rate": 0.00040051934457908933, "loss": 1.2174, "step": 99340 }, { "epoch": 4.0, "learning_rate": 0.0004005092797616651, "loss": 1.1255, "step": 99350 }, { "epoch": 4.0, "learning_rate": 0.0004004992149442409, "loss": 1.2206, "step": 99360 }, { "epoch": 4.0, "learning_rate": 0.0004004891501268167, "loss": 1.2352, "step": 99370 }, { "epoch": 4.0, "learning_rate": 0.0004004790853093925, "loss": 1.1218, "step": 99380 }, { "epoch": 4.0, "learning_rate": 0.0004004690204919683, "loss": 1.2938, "step": 99390 }, { "epoch": 4.0, "learning_rate": 0.000400458955674544, "loss": 1.2264, "step": 99400 }, { "epoch": 4.0, "learning_rate": 0.00040044889085711987, "loss": 1.2493, "step": 99410 }, { "epoch": 4.0, "learning_rate": 0.00040043882603969566, "loss": 1.08, "step": 99420 }, { "epoch": 4.0, "learning_rate": 0.00040042876122227145, "loss": 1.0216, "step": 99430 }, { "epoch": 4.0, "learning_rate": 0.00040041869640484724, "loss": 1.3213, "step": 99440 }, { "epoch": 4.0, "learning_rate": 0.00040040863158742303, "loss": 1.1371, "step": 99450 }, { "epoch": 4.0, "eval_Bleu_1": 0.013593270461350957, "eval_Bleu_2": 1.6090820540115143e-11, "eval_Bleu_3": 1.9167699839965318e-14, "eval_Bleu_4": 7.16570669708969e-16, "eval_ROUGE_L": 0.03360663812308151, "eval_cer": 0.7014755330511492, "eval_em": 0.3156089193825043, "eval_f1": 0.4235784840838624, "eval_loss": 1.7718961238861084, "eval_runtime": 2155.778, "eval_samples_per_second": 12.981, "eval_steps_per_second": 4.327, "eval_wer": 0.8482886743637379, "step": 99457 }, { "epoch": 4.0, "learning_rate": 0.0004003985667699988, "loss": 0.9698, "step": 99460 }, { "epoch": 4.0, "learning_rate": 0.0004003885019525746, "loss": 0.9213, "step": 99470 }, { "epoch": 4.0, "learning_rate": 0.00040037843713515035, "loss": 1.0345, "step": 99480 }, { "epoch": 4.0, "learning_rate": 0.00040036837231772614, "loss": 0.8721, "step": 99490 }, { "epoch": 4.0, "learning_rate": 0.00040035830750030193, "loss": 0.8644, "step": 99500 }, { "epoch": 4.0, "learning_rate": 0.0004003482426828777, "loss": 0.8688, "step": 99510 }, { "epoch": 4.0, "learning_rate": 0.0004003381778654535, "loss": 0.8634, "step": 99520 }, { "epoch": 4.0, "learning_rate": 0.0004003281130480293, "loss": 0.8846, "step": 99530 }, { "epoch": 4.0, "learning_rate": 0.0004003180482306051, "loss": 0.8496, "step": 99540 }, { "epoch": 4.0, "learning_rate": 0.00040030798341318094, "loss": 0.6611, "step": 99550 }, { "epoch": 4.0, "learning_rate": 0.00040029791859575673, "loss": 0.9481, "step": 99560 }, { "epoch": 4.0, "learning_rate": 0.00040028785377833247, "loss": 0.9549, "step": 99570 }, { "epoch": 4.0, "learning_rate": 0.00040027778896090826, "loss": 0.7185, "step": 99580 }, { "epoch": 4.01, "learning_rate": 0.00040026772414348405, "loss": 1.0468, "step": 99590 }, { "epoch": 4.01, "learning_rate": 0.00040025765932605984, "loss": 1.0305, "step": 99600 }, { "epoch": 4.01, "learning_rate": 0.00040024759450863563, "loss": 0.7819, "step": 99610 }, { "epoch": 4.01, "learning_rate": 0.0004002375296912114, "loss": 0.7694, "step": 99620 }, { "epoch": 4.01, "learning_rate": 0.0004002274648737872, "loss": 0.9994, "step": 99630 }, { "epoch": 4.01, "learning_rate": 0.000400217400056363, "loss": 0.6983, "step": 99640 }, { "epoch": 4.01, "learning_rate": 0.00040020733523893874, "loss": 0.8098, "step": 99650 }, { "epoch": 4.01, "learning_rate": 0.00040019727042151453, "loss": 0.9344, "step": 99660 }, { "epoch": 4.01, "learning_rate": 0.0004001872056040903, "loss": 0.9656, "step": 99670 }, { "epoch": 4.01, "learning_rate": 0.0004001771407866661, "loss": 0.9972, "step": 99680 }, { "epoch": 4.01, "learning_rate": 0.0004001670759692419, "loss": 1.1113, "step": 99690 }, { "epoch": 4.01, "learning_rate": 0.00040015701115181775, "loss": 0.8704, "step": 99700 }, { "epoch": 4.01, "learning_rate": 0.00040014694633439354, "loss": 0.9147, "step": 99710 }, { "epoch": 4.01, "learning_rate": 0.00040013688151696933, "loss": 0.7803, "step": 99720 }, { "epoch": 4.01, "learning_rate": 0.00040012681669954507, "loss": 0.8446, "step": 99730 }, { "epoch": 4.01, "learning_rate": 0.00040011675188212086, "loss": 1.0096, "step": 99740 }, { "epoch": 4.01, "learning_rate": 0.00040010668706469665, "loss": 0.7601, "step": 99750 }, { "epoch": 4.01, "learning_rate": 0.00040009662224727244, "loss": 0.981, "step": 99760 }, { "epoch": 4.01, "learning_rate": 0.00040008655742984823, "loss": 0.9999, "step": 99770 }, { "epoch": 4.01, "learning_rate": 0.000400076492612424, "loss": 0.9989, "step": 99780 }, { "epoch": 4.01, "learning_rate": 0.0004000664277949998, "loss": 0.96, "step": 99790 }, { "epoch": 4.01, "learning_rate": 0.0004000563629775756, "loss": 0.8034, "step": 99800 }, { "epoch": 4.01, "learning_rate": 0.0004000462981601514, "loss": 0.6789, "step": 99810 }, { "epoch": 4.01, "learning_rate": 0.00040003623334272713, "loss": 0.8347, "step": 99820 }, { "epoch": 4.02, "learning_rate": 0.0004000261685253029, "loss": 1.0312, "step": 99830 }, { "epoch": 4.02, "learning_rate": 0.0004000161037078787, "loss": 0.8704, "step": 99840 }, { "epoch": 4.02, "learning_rate": 0.00040000603889045456, "loss": 0.987, "step": 99850 }, { "epoch": 4.02, "learning_rate": 0.00039999597407303035, "loss": 0.9653, "step": 99860 }, { "epoch": 4.02, "learning_rate": 0.00039998590925560614, "loss": 1.0014, "step": 99870 }, { "epoch": 4.02, "learning_rate": 0.00039997584443818193, "loss": 0.7411, "step": 99880 }, { "epoch": 4.02, "learning_rate": 0.0003999657796207577, "loss": 0.9416, "step": 99890 }, { "epoch": 4.02, "learning_rate": 0.00039995571480333346, "loss": 0.8838, "step": 99900 }, { "epoch": 4.02, "learning_rate": 0.00039994564998590925, "loss": 0.8067, "step": 99910 }, { "epoch": 4.02, "learning_rate": 0.00039993558516848504, "loss": 0.6025, "step": 99920 }, { "epoch": 4.02, "learning_rate": 0.00039992552035106083, "loss": 1.1184, "step": 99930 }, { "epoch": 4.02, "learning_rate": 0.0003999154555336366, "loss": 1.0763, "step": 99940 }, { "epoch": 4.02, "learning_rate": 0.0003999053907162124, "loss": 0.8304, "step": 99950 }, { "epoch": 4.02, "learning_rate": 0.0003998953258987882, "loss": 1.0206, "step": 99960 }, { "epoch": 4.02, "learning_rate": 0.000399885261081364, "loss": 0.7993, "step": 99970 }, { "epoch": 4.02, "learning_rate": 0.0003998751962639398, "loss": 0.8889, "step": 99980 }, { "epoch": 4.02, "learning_rate": 0.00039986513144651557, "loss": 0.9681, "step": 99990 }, { "epoch": 4.02, "learning_rate": 0.00039985506662909136, "loss": 0.7795, "step": 100000 }, { "epoch": 4.02, "learning_rate": 0.00039984500181166715, "loss": 0.8797, "step": 100010 }, { "epoch": 4.02, "learning_rate": 0.00039983493699424294, "loss": 1.154, "step": 100020 }, { "epoch": 4.02, "learning_rate": 0.00039982487217681874, "loss": 0.7898, "step": 100030 }, { "epoch": 4.02, "learning_rate": 0.0003998148073593945, "loss": 0.8858, "step": 100040 }, { "epoch": 4.02, "learning_rate": 0.0003998047425419703, "loss": 1.0475, "step": 100050 }, { "epoch": 4.02, "learning_rate": 0.0003997946777245461, "loss": 0.8041, "step": 100060 }, { "epoch": 4.02, "learning_rate": 0.00039978461290712184, "loss": 0.8299, "step": 100070 }, { "epoch": 4.03, "learning_rate": 0.00039977454808969764, "loss": 0.9655, "step": 100080 }, { "epoch": 4.03, "learning_rate": 0.0003997644832722734, "loss": 1.0665, "step": 100090 }, { "epoch": 4.03, "learning_rate": 0.0003997544184548492, "loss": 0.7826, "step": 100100 }, { "epoch": 4.03, "learning_rate": 0.000399744353637425, "loss": 0.8651, "step": 100110 }, { "epoch": 4.03, "learning_rate": 0.0003997342888200008, "loss": 0.7539, "step": 100120 }, { "epoch": 4.03, "learning_rate": 0.0003997242240025766, "loss": 0.7867, "step": 100130 }, { "epoch": 4.03, "learning_rate": 0.00039971415918515243, "loss": 0.7873, "step": 100140 }, { "epoch": 4.03, "learning_rate": 0.0003997040943677282, "loss": 0.8605, "step": 100150 }, { "epoch": 4.03, "learning_rate": 0.00039969402955030396, "loss": 0.9501, "step": 100160 }, { "epoch": 4.03, "learning_rate": 0.00039968396473287975, "loss": 0.9455, "step": 100170 }, { "epoch": 4.03, "learning_rate": 0.00039967389991545554, "loss": 0.6964, "step": 100180 }, { "epoch": 4.03, "learning_rate": 0.00039966383509803133, "loss": 0.9512, "step": 100190 }, { "epoch": 4.03, "learning_rate": 0.0003996537702806071, "loss": 0.6899, "step": 100200 }, { "epoch": 4.03, "learning_rate": 0.0003996437054631829, "loss": 0.9712, "step": 100210 }, { "epoch": 4.03, "learning_rate": 0.0003996336406457587, "loss": 0.7169, "step": 100220 }, { "epoch": 4.03, "learning_rate": 0.0003996235758283345, "loss": 1.033, "step": 100230 }, { "epoch": 4.03, "learning_rate": 0.00039961351101091023, "loss": 0.9207, "step": 100240 }, { "epoch": 4.03, "learning_rate": 0.000399603446193486, "loss": 0.8779, "step": 100250 }, { "epoch": 4.03, "learning_rate": 0.0003995933813760618, "loss": 1.0272, "step": 100260 }, { "epoch": 4.03, "learning_rate": 0.0003995833165586376, "loss": 0.9387, "step": 100270 }, { "epoch": 4.03, "learning_rate": 0.00039957325174121345, "loss": 1.0187, "step": 100280 }, { "epoch": 4.03, "learning_rate": 0.00039956318692378924, "loss": 0.8848, "step": 100290 }, { "epoch": 4.03, "learning_rate": 0.00039955312210636503, "loss": 0.9666, "step": 100300 }, { "epoch": 4.03, "learning_rate": 0.0003995430572889408, "loss": 0.9008, "step": 100310 }, { "epoch": 4.03, "learning_rate": 0.0003995329924715166, "loss": 0.8164, "step": 100320 }, { "epoch": 4.04, "learning_rate": 0.00039952292765409235, "loss": 0.8632, "step": 100330 }, { "epoch": 4.04, "learning_rate": 0.00039951286283666814, "loss": 0.8335, "step": 100340 }, { "epoch": 4.04, "learning_rate": 0.00039950279801924393, "loss": 0.9395, "step": 100350 }, { "epoch": 4.04, "learning_rate": 0.0003994927332018197, "loss": 0.9048, "step": 100360 }, { "epoch": 4.04, "learning_rate": 0.0003994826683843955, "loss": 0.8915, "step": 100370 }, { "epoch": 4.04, "learning_rate": 0.0003994726035669713, "loss": 1.0007, "step": 100380 }, { "epoch": 4.04, "learning_rate": 0.0003994625387495471, "loss": 0.9525, "step": 100390 }, { "epoch": 4.04, "learning_rate": 0.0003994524739321229, "loss": 0.8988, "step": 100400 }, { "epoch": 4.04, "learning_rate": 0.0003994424091146986, "loss": 1.0165, "step": 100410 }, { "epoch": 4.04, "learning_rate": 0.0003994323442972744, "loss": 0.9591, "step": 100420 }, { "epoch": 4.04, "learning_rate": 0.00039942227947985026, "loss": 0.801, "step": 100430 }, { "epoch": 4.04, "learning_rate": 0.00039941221466242605, "loss": 0.7991, "step": 100440 }, { "epoch": 4.04, "learning_rate": 0.00039940214984500184, "loss": 1.1575, "step": 100450 }, { "epoch": 4.04, "learning_rate": 0.00039939208502757763, "loss": 0.892, "step": 100460 }, { "epoch": 4.04, "learning_rate": 0.0003993820202101534, "loss": 0.9574, "step": 100470 }, { "epoch": 4.04, "learning_rate": 0.0003993719553927292, "loss": 0.8558, "step": 100480 }, { "epoch": 4.04, "learning_rate": 0.00039936189057530495, "loss": 0.8243, "step": 100490 }, { "epoch": 4.04, "learning_rate": 0.00039935182575788074, "loss": 0.7632, "step": 100500 }, { "epoch": 4.04, "learning_rate": 0.00039934176094045653, "loss": 1.0394, "step": 100510 }, { "epoch": 4.04, "learning_rate": 0.0003993316961230323, "loss": 0.9544, "step": 100520 }, { "epoch": 4.04, "learning_rate": 0.0003993216313056081, "loss": 0.9701, "step": 100530 }, { "epoch": 4.04, "learning_rate": 0.0003993115664881839, "loss": 0.7878, "step": 100540 }, { "epoch": 4.04, "learning_rate": 0.0003993015016707597, "loss": 0.8523, "step": 100550 }, { "epoch": 4.04, "learning_rate": 0.0003992914368533355, "loss": 0.964, "step": 100560 }, { "epoch": 4.04, "learning_rate": 0.0003992813720359113, "loss": 1.0023, "step": 100570 }, { "epoch": 4.05, "learning_rate": 0.00039927130721848707, "loss": 0.7259, "step": 100580 }, { "epoch": 4.05, "learning_rate": 0.00039926124240106286, "loss": 0.9446, "step": 100590 }, { "epoch": 4.05, "learning_rate": 0.00039925117758363865, "loss": 0.9114, "step": 100600 }, { "epoch": 4.05, "learning_rate": 0.00039924111276621444, "loss": 0.9531, "step": 100610 }, { "epoch": 4.05, "learning_rate": 0.00039923104794879023, "loss": 0.9248, "step": 100620 }, { "epoch": 4.05, "learning_rate": 0.000399220983131366, "loss": 0.9843, "step": 100630 }, { "epoch": 4.05, "learning_rate": 0.0003992109183139418, "loss": 0.9701, "step": 100640 }, { "epoch": 4.05, "learning_rate": 0.0003992008534965176, "loss": 0.983, "step": 100650 }, { "epoch": 4.05, "learning_rate": 0.00039919078867909334, "loss": 0.9534, "step": 100660 }, { "epoch": 4.05, "learning_rate": 0.00039918072386166913, "loss": 0.9859, "step": 100670 }, { "epoch": 4.05, "learning_rate": 0.0003991706590442449, "loss": 0.8557, "step": 100680 }, { "epoch": 4.05, "learning_rate": 0.0003991605942268207, "loss": 1.2523, "step": 100690 }, { "epoch": 4.05, "learning_rate": 0.0003991505294093965, "loss": 0.8414, "step": 100700 }, { "epoch": 4.05, "learning_rate": 0.0003991404645919723, "loss": 0.8426, "step": 100710 }, { "epoch": 4.05, "learning_rate": 0.00039913039977454814, "loss": 1.0002, "step": 100720 }, { "epoch": 4.05, "learning_rate": 0.00039912033495712393, "loss": 0.928, "step": 100730 }, { "epoch": 4.05, "learning_rate": 0.0003991102701396997, "loss": 1.019, "step": 100740 }, { "epoch": 4.05, "learning_rate": 0.00039910020532227545, "loss": 0.817, "step": 100750 }, { "epoch": 4.05, "learning_rate": 0.00039909014050485125, "loss": 0.8573, "step": 100760 }, { "epoch": 4.05, "learning_rate": 0.00039908007568742704, "loss": 1.0506, "step": 100770 }, { "epoch": 4.05, "learning_rate": 0.00039907001087000283, "loss": 1.0443, "step": 100780 }, { "epoch": 4.05, "learning_rate": 0.0003990599460525786, "loss": 0.7874, "step": 100790 }, { "epoch": 4.05, "learning_rate": 0.0003990498812351544, "loss": 0.772, "step": 100800 }, { "epoch": 4.05, "learning_rate": 0.0003990398164177302, "loss": 0.9543, "step": 100810 }, { "epoch": 4.05, "learning_rate": 0.000399029751600306, "loss": 0.7879, "step": 100820 }, { "epoch": 4.06, "learning_rate": 0.0003990196867828817, "loss": 0.9191, "step": 100830 }, { "epoch": 4.06, "learning_rate": 0.0003990096219654575, "loss": 0.8418, "step": 100840 }, { "epoch": 4.06, "learning_rate": 0.0003989995571480333, "loss": 1.0181, "step": 100850 }, { "epoch": 4.06, "learning_rate": 0.0003989894923306091, "loss": 0.7844, "step": 100860 }, { "epoch": 4.06, "learning_rate": 0.00039897942751318494, "loss": 0.8072, "step": 100870 }, { "epoch": 4.06, "learning_rate": 0.00039896936269576074, "loss": 0.7595, "step": 100880 }, { "epoch": 4.06, "learning_rate": 0.0003989592978783365, "loss": 0.7719, "step": 100890 }, { "epoch": 4.06, "learning_rate": 0.0003989492330609123, "loss": 0.8359, "step": 100900 }, { "epoch": 4.06, "learning_rate": 0.0003989391682434881, "loss": 1.036, "step": 100910 }, { "epoch": 4.06, "learning_rate": 0.00039892910342606384, "loss": 0.9655, "step": 100920 }, { "epoch": 4.06, "learning_rate": 0.00039891903860863963, "loss": 0.8938, "step": 100930 }, { "epoch": 4.06, "learning_rate": 0.0003989089737912154, "loss": 0.8734, "step": 100940 }, { "epoch": 4.06, "learning_rate": 0.0003988989089737912, "loss": 0.9412, "step": 100950 }, { "epoch": 4.06, "learning_rate": 0.000398888844156367, "loss": 0.9864, "step": 100960 }, { "epoch": 4.06, "learning_rate": 0.0003988787793389428, "loss": 0.946, "step": 100970 }, { "epoch": 4.06, "learning_rate": 0.0003988687145215186, "loss": 0.7976, "step": 100980 }, { "epoch": 4.06, "learning_rate": 0.0003988586497040944, "loss": 0.8617, "step": 100990 }, { "epoch": 4.06, "learning_rate": 0.0003988485848866701, "loss": 0.9622, "step": 101000 }, { "epoch": 4.06, "learning_rate": 0.0003988385200692459, "loss": 0.7736, "step": 101010 }, { "epoch": 4.06, "learning_rate": 0.00039882845525182175, "loss": 0.9116, "step": 101020 }, { "epoch": 4.06, "learning_rate": 0.00039881839043439754, "loss": 0.7928, "step": 101030 }, { "epoch": 4.06, "learning_rate": 0.00039880832561697333, "loss": 1.27, "step": 101040 }, { "epoch": 4.06, "learning_rate": 0.0003987982607995491, "loss": 1.2241, "step": 101050 }, { "epoch": 4.06, "learning_rate": 0.0003987881959821249, "loss": 0.6813, "step": 101060 }, { "epoch": 4.06, "learning_rate": 0.0003987781311647007, "loss": 0.9695, "step": 101070 }, { "epoch": 4.07, "learning_rate": 0.0003987680663472765, "loss": 1.0807, "step": 101080 }, { "epoch": 4.07, "learning_rate": 0.00039875800152985223, "loss": 0.8497, "step": 101090 }, { "epoch": 4.07, "learning_rate": 0.000398747936712428, "loss": 0.944, "step": 101100 }, { "epoch": 4.07, "learning_rate": 0.0003987378718950038, "loss": 0.9158, "step": 101110 }, { "epoch": 4.07, "learning_rate": 0.0003987278070775796, "loss": 0.9099, "step": 101120 }, { "epoch": 4.07, "learning_rate": 0.0003987177422601554, "loss": 0.9577, "step": 101130 }, { "epoch": 4.07, "learning_rate": 0.0003987076774427312, "loss": 0.8214, "step": 101140 }, { "epoch": 4.07, "learning_rate": 0.000398697612625307, "loss": 0.7699, "step": 101150 }, { "epoch": 4.07, "learning_rate": 0.0003986875478078828, "loss": 1.0826, "step": 101160 }, { "epoch": 4.07, "learning_rate": 0.00039867748299045856, "loss": 1.11, "step": 101170 }, { "epoch": 4.07, "learning_rate": 0.00039866741817303435, "loss": 1.0109, "step": 101180 }, { "epoch": 4.07, "learning_rate": 0.00039865735335561014, "loss": 0.8466, "step": 101190 }, { "epoch": 4.07, "learning_rate": 0.00039864728853818593, "loss": 0.9038, "step": 101200 }, { "epoch": 4.07, "learning_rate": 0.0003986372237207617, "loss": 0.749, "step": 101210 }, { "epoch": 4.07, "learning_rate": 0.0003986271589033375, "loss": 0.9364, "step": 101220 }, { "epoch": 4.07, "learning_rate": 0.0003986170940859133, "loss": 0.9518, "step": 101230 }, { "epoch": 4.07, "learning_rate": 0.0003986070292684891, "loss": 0.9417, "step": 101240 }, { "epoch": 4.07, "learning_rate": 0.00039859696445106483, "loss": 0.8427, "step": 101250 }, { "epoch": 4.07, "learning_rate": 0.0003985868996336406, "loss": 0.8157, "step": 101260 }, { "epoch": 4.07, "learning_rate": 0.0003985768348162164, "loss": 0.8414, "step": 101270 }, { "epoch": 4.07, "learning_rate": 0.0003985667699987922, "loss": 1.0728, "step": 101280 }, { "epoch": 4.07, "learning_rate": 0.000398556705181368, "loss": 0.7586, "step": 101290 }, { "epoch": 4.07, "learning_rate": 0.0003985466403639438, "loss": 0.8142, "step": 101300 }, { "epoch": 4.07, "learning_rate": 0.00039853657554651963, "loss": 1.0268, "step": 101310 }, { "epoch": 4.07, "learning_rate": 0.0003985265107290954, "loss": 0.8514, "step": 101320 }, { "epoch": 4.08, "learning_rate": 0.0003985164459116712, "loss": 0.8364, "step": 101330 }, { "epoch": 4.08, "learning_rate": 0.00039850638109424695, "loss": 0.9804, "step": 101340 }, { "epoch": 4.08, "learning_rate": 0.00039849631627682274, "loss": 0.8561, "step": 101350 }, { "epoch": 4.08, "learning_rate": 0.00039848625145939853, "loss": 0.8215, "step": 101360 }, { "epoch": 4.08, "learning_rate": 0.0003984761866419743, "loss": 0.9759, "step": 101370 }, { "epoch": 4.08, "learning_rate": 0.0003984661218245501, "loss": 0.8816, "step": 101380 }, { "epoch": 4.08, "learning_rate": 0.0003984560570071259, "loss": 0.9023, "step": 101390 }, { "epoch": 4.08, "learning_rate": 0.0003984459921897017, "loss": 0.7592, "step": 101400 }, { "epoch": 4.08, "learning_rate": 0.0003984359273722775, "loss": 0.8736, "step": 101410 }, { "epoch": 4.08, "learning_rate": 0.0003984258625548532, "loss": 1.2112, "step": 101420 }, { "epoch": 4.08, "learning_rate": 0.000398415797737429, "loss": 0.7781, "step": 101430 }, { "epoch": 4.08, "learning_rate": 0.0003984057329200048, "loss": 0.8768, "step": 101440 }, { "epoch": 4.08, "learning_rate": 0.00039839566810258065, "loss": 0.8568, "step": 101450 }, { "epoch": 4.08, "learning_rate": 0.00039838560328515644, "loss": 0.8341, "step": 101460 }, { "epoch": 4.08, "learning_rate": 0.00039837553846773223, "loss": 1.0201, "step": 101470 }, { "epoch": 4.08, "learning_rate": 0.000398365473650308, "loss": 0.8224, "step": 101480 }, { "epoch": 4.08, "learning_rate": 0.0003983554088328838, "loss": 0.986, "step": 101490 }, { "epoch": 4.08, "learning_rate": 0.0003983453440154596, "loss": 0.736, "step": 101500 }, { "epoch": 4.08, "learning_rate": 0.00039833527919803534, "loss": 0.9201, "step": 101510 }, { "epoch": 4.08, "learning_rate": 0.00039832521438061113, "loss": 1.0392, "step": 101520 }, { "epoch": 4.08, "learning_rate": 0.0003983151495631869, "loss": 1.023, "step": 101530 }, { "epoch": 4.08, "learning_rate": 0.0003983050847457627, "loss": 0.8821, "step": 101540 }, { "epoch": 4.08, "learning_rate": 0.0003982950199283385, "loss": 0.8794, "step": 101550 }, { "epoch": 4.08, "learning_rate": 0.0003982849551109143, "loss": 0.8687, "step": 101560 }, { "epoch": 4.08, "learning_rate": 0.0003982748902934901, "loss": 0.8745, "step": 101570 }, { "epoch": 4.09, "learning_rate": 0.0003982648254760659, "loss": 0.8599, "step": 101580 }, { "epoch": 4.09, "learning_rate": 0.0003982547606586416, "loss": 0.9415, "step": 101590 }, { "epoch": 4.09, "learning_rate": 0.00039824469584121745, "loss": 0.7521, "step": 101600 }, { "epoch": 4.09, "learning_rate": 0.00039823463102379325, "loss": 0.7936, "step": 101610 }, { "epoch": 4.09, "learning_rate": 0.00039822456620636904, "loss": 0.8973, "step": 101620 }, { "epoch": 4.09, "learning_rate": 0.0003982145013889448, "loss": 0.8934, "step": 101630 }, { "epoch": 4.09, "learning_rate": 0.0003982044365715206, "loss": 0.9484, "step": 101640 }, { "epoch": 4.09, "learning_rate": 0.0003981943717540964, "loss": 0.9996, "step": 101650 }, { "epoch": 4.09, "learning_rate": 0.0003981843069366722, "loss": 0.9042, "step": 101660 }, { "epoch": 4.09, "learning_rate": 0.000398174242119248, "loss": 1.08, "step": 101670 }, { "epoch": 4.09, "learning_rate": 0.0003981641773018237, "loss": 0.9342, "step": 101680 }, { "epoch": 4.09, "learning_rate": 0.0003981541124843995, "loss": 0.841, "step": 101690 }, { "epoch": 4.09, "learning_rate": 0.0003981440476669753, "loss": 0.9891, "step": 101700 }, { "epoch": 4.09, "learning_rate": 0.0003981339828495511, "loss": 0.9265, "step": 101710 }, { "epoch": 4.09, "learning_rate": 0.0003981239180321269, "loss": 0.8461, "step": 101720 }, { "epoch": 4.09, "learning_rate": 0.0003981138532147027, "loss": 1.1197, "step": 101730 }, { "epoch": 4.09, "learning_rate": 0.00039810378839727847, "loss": 0.8367, "step": 101740 }, { "epoch": 4.09, "learning_rate": 0.0003980937235798543, "loss": 0.7991, "step": 101750 }, { "epoch": 4.09, "learning_rate": 0.00039808365876243005, "loss": 0.8809, "step": 101760 }, { "epoch": 4.09, "learning_rate": 0.00039807359394500584, "loss": 0.8727, "step": 101770 }, { "epoch": 4.09, "learning_rate": 0.00039806352912758163, "loss": 1.1376, "step": 101780 }, { "epoch": 4.09, "learning_rate": 0.0003980534643101574, "loss": 0.7376, "step": 101790 }, { "epoch": 4.09, "learning_rate": 0.0003980433994927332, "loss": 0.9566, "step": 101800 }, { "epoch": 4.09, "learning_rate": 0.000398033334675309, "loss": 0.9586, "step": 101810 }, { "epoch": 4.1, "learning_rate": 0.0003980232698578848, "loss": 0.7985, "step": 101820 }, { "epoch": 4.1, "learning_rate": 0.0003980132050404606, "loss": 1.0996, "step": 101830 }, { "epoch": 4.1, "learning_rate": 0.0003980031402230364, "loss": 0.8601, "step": 101840 }, { "epoch": 4.1, "learning_rate": 0.0003979930754056121, "loss": 0.9301, "step": 101850 }, { "epoch": 4.1, "learning_rate": 0.0003979830105881879, "loss": 1.1577, "step": 101860 }, { "epoch": 4.1, "learning_rate": 0.0003979729457707637, "loss": 1.0023, "step": 101870 }, { "epoch": 4.1, "learning_rate": 0.0003979628809533395, "loss": 0.9298, "step": 101880 }, { "epoch": 4.1, "learning_rate": 0.00039795281613591533, "loss": 0.8275, "step": 101890 }, { "epoch": 4.1, "learning_rate": 0.0003979427513184911, "loss": 0.8376, "step": 101900 }, { "epoch": 4.1, "learning_rate": 0.0003979326865010669, "loss": 0.9605, "step": 101910 }, { "epoch": 4.1, "learning_rate": 0.0003979226216836427, "loss": 0.8523, "step": 101920 }, { "epoch": 4.1, "learning_rate": 0.00039791255686621844, "loss": 0.9784, "step": 101930 }, { "epoch": 4.1, "learning_rate": 0.00039790249204879423, "loss": 0.618, "step": 101940 }, { "epoch": 4.1, "learning_rate": 0.00039789242723137, "loss": 0.9479, "step": 101950 }, { "epoch": 4.1, "learning_rate": 0.0003978823624139458, "loss": 0.9593, "step": 101960 }, { "epoch": 4.1, "learning_rate": 0.0003978722975965216, "loss": 0.8363, "step": 101970 }, { "epoch": 4.1, "learning_rate": 0.0003978622327790974, "loss": 0.9838, "step": 101980 }, { "epoch": 4.1, "learning_rate": 0.0003978521679616732, "loss": 0.7879, "step": 101990 }, { "epoch": 4.1, "learning_rate": 0.000397842103144249, "loss": 1.102, "step": 102000 }, { "epoch": 4.1, "learning_rate": 0.0003978320383268247, "loss": 1.0787, "step": 102010 }, { "epoch": 4.1, "learning_rate": 0.0003978219735094005, "loss": 0.9202, "step": 102020 }, { "epoch": 4.1, "learning_rate": 0.0003978119086919763, "loss": 0.8578, "step": 102030 }, { "epoch": 4.1, "learning_rate": 0.00039780184387455214, "loss": 0.8817, "step": 102040 }, { "epoch": 4.1, "learning_rate": 0.00039779177905712793, "loss": 0.9429, "step": 102050 }, { "epoch": 4.1, "learning_rate": 0.0003977817142397037, "loss": 0.9837, "step": 102060 }, { "epoch": 4.11, "learning_rate": 0.0003977716494222795, "loss": 0.9929, "step": 102070 }, { "epoch": 4.11, "learning_rate": 0.0003977615846048553, "loss": 0.9054, "step": 102080 }, { "epoch": 4.11, "learning_rate": 0.0003977515197874311, "loss": 0.9796, "step": 102090 }, { "epoch": 4.11, "learning_rate": 0.00039774145497000683, "loss": 0.6752, "step": 102100 }, { "epoch": 4.11, "learning_rate": 0.0003977313901525826, "loss": 0.9606, "step": 102110 }, { "epoch": 4.11, "learning_rate": 0.0003977213253351584, "loss": 1.2447, "step": 102120 }, { "epoch": 4.11, "learning_rate": 0.0003977112605177342, "loss": 0.9955, "step": 102130 }, { "epoch": 4.11, "learning_rate": 0.00039770119570031, "loss": 0.8685, "step": 102140 }, { "epoch": 4.11, "learning_rate": 0.0003976911308828858, "loss": 0.9097, "step": 102150 }, { "epoch": 4.11, "learning_rate": 0.0003976810660654616, "loss": 1.0287, "step": 102160 }, { "epoch": 4.11, "learning_rate": 0.00039767100124803737, "loss": 0.7321, "step": 102170 }, { "epoch": 4.11, "learning_rate": 0.0003976609364306131, "loss": 0.8816, "step": 102180 }, { "epoch": 4.11, "learning_rate": 0.00039765087161318895, "loss": 0.7281, "step": 102190 }, { "epoch": 4.11, "learning_rate": 0.00039764080679576474, "loss": 1.1577, "step": 102200 }, { "epoch": 4.11, "learning_rate": 0.00039763074197834053, "loss": 0.9431, "step": 102210 }, { "epoch": 4.11, "learning_rate": 0.0003976206771609163, "loss": 0.9893, "step": 102220 }, { "epoch": 4.11, "learning_rate": 0.0003976106123434921, "loss": 0.8438, "step": 102230 }, { "epoch": 4.11, "learning_rate": 0.0003976005475260679, "loss": 0.9432, "step": 102240 }, { "epoch": 4.11, "learning_rate": 0.0003975904827086437, "loss": 0.6701, "step": 102250 }, { "epoch": 4.11, "learning_rate": 0.0003975804178912195, "loss": 1.0792, "step": 102260 }, { "epoch": 4.11, "learning_rate": 0.0003975703530737952, "loss": 0.8788, "step": 102270 }, { "epoch": 4.11, "learning_rate": 0.000397560288256371, "loss": 0.9849, "step": 102280 }, { "epoch": 4.11, "learning_rate": 0.0003975502234389468, "loss": 0.8347, "step": 102290 }, { "epoch": 4.11, "learning_rate": 0.0003975401586215226, "loss": 0.9831, "step": 102300 }, { "epoch": 4.11, "learning_rate": 0.0003975300938040984, "loss": 0.8756, "step": 102310 }, { "epoch": 4.12, "learning_rate": 0.0003975200289866742, "loss": 0.9968, "step": 102320 }, { "epoch": 4.12, "learning_rate": 0.00039750996416925, "loss": 1.0175, "step": 102330 }, { "epoch": 4.12, "learning_rate": 0.0003974998993518258, "loss": 1.0521, "step": 102340 }, { "epoch": 4.12, "learning_rate": 0.00039748983453440155, "loss": 0.8692, "step": 102350 }, { "epoch": 4.12, "learning_rate": 0.00039747976971697734, "loss": 0.9605, "step": 102360 }, { "epoch": 4.12, "learning_rate": 0.00039746970489955313, "loss": 0.9844, "step": 102370 }, { "epoch": 4.12, "learning_rate": 0.0003974596400821289, "loss": 0.8785, "step": 102380 }, { "epoch": 4.12, "learning_rate": 0.0003974495752647047, "loss": 1.0155, "step": 102390 }, { "epoch": 4.12, "learning_rate": 0.0003974395104472805, "loss": 1.0366, "step": 102400 }, { "epoch": 4.12, "learning_rate": 0.0003974294456298563, "loss": 1.0247, "step": 102410 }, { "epoch": 4.12, "learning_rate": 0.0003974193808124321, "loss": 1.1602, "step": 102420 }, { "epoch": 4.12, "learning_rate": 0.0003974093159950079, "loss": 0.8226, "step": 102430 }, { "epoch": 4.12, "learning_rate": 0.0003973992511775836, "loss": 0.9689, "step": 102440 }, { "epoch": 4.12, "learning_rate": 0.0003973891863601594, "loss": 1.0404, "step": 102450 }, { "epoch": 4.12, "learning_rate": 0.0003973791215427352, "loss": 1.2805, "step": 102460 }, { "epoch": 4.12, "learning_rate": 0.000397369056725311, "loss": 1.0492, "step": 102470 }, { "epoch": 4.12, "learning_rate": 0.0003973589919078868, "loss": 0.9595, "step": 102480 }, { "epoch": 4.12, "learning_rate": 0.0003973489270904626, "loss": 0.9989, "step": 102490 }, { "epoch": 4.12, "learning_rate": 0.0003973388622730384, "loss": 0.902, "step": 102500 }, { "epoch": 4.12, "learning_rate": 0.0003973287974556142, "loss": 0.902, "step": 102510 }, { "epoch": 4.12, "learning_rate": 0.00039731873263818994, "loss": 0.9215, "step": 102520 }, { "epoch": 4.12, "learning_rate": 0.0003973086678207657, "loss": 0.8334, "step": 102530 }, { "epoch": 4.12, "learning_rate": 0.0003972986030033415, "loss": 0.9205, "step": 102540 }, { "epoch": 4.12, "learning_rate": 0.0003972885381859173, "loss": 1.0599, "step": 102550 }, { "epoch": 4.12, "learning_rate": 0.0003972784733684931, "loss": 1.1129, "step": 102560 }, { "epoch": 4.13, "learning_rate": 0.0003972684085510689, "loss": 0.687, "step": 102570 }, { "epoch": 4.13, "learning_rate": 0.0003972583437336447, "loss": 0.8893, "step": 102580 }, { "epoch": 4.13, "learning_rate": 0.00039724827891622047, "loss": 0.8371, "step": 102590 }, { "epoch": 4.13, "learning_rate": 0.00039723821409879626, "loss": 0.8428, "step": 102600 }, { "epoch": 4.13, "learning_rate": 0.000397228149281372, "loss": 1.0418, "step": 102610 }, { "epoch": 4.13, "learning_rate": 0.00039721808446394784, "loss": 0.9359, "step": 102620 }, { "epoch": 4.13, "learning_rate": 0.00039720801964652363, "loss": 0.8584, "step": 102630 }, { "epoch": 4.13, "learning_rate": 0.0003971979548290994, "loss": 0.9642, "step": 102640 }, { "epoch": 4.13, "learning_rate": 0.0003971878900116752, "loss": 0.8528, "step": 102650 }, { "epoch": 4.13, "learning_rate": 0.000397177825194251, "loss": 0.8041, "step": 102660 }, { "epoch": 4.13, "learning_rate": 0.0003971677603768268, "loss": 0.9286, "step": 102670 }, { "epoch": 4.13, "learning_rate": 0.0003971576955594026, "loss": 1.0941, "step": 102680 }, { "epoch": 4.13, "learning_rate": 0.0003971476307419783, "loss": 1.0922, "step": 102690 }, { "epoch": 4.13, "learning_rate": 0.0003971375659245541, "loss": 1.1589, "step": 102700 }, { "epoch": 4.13, "learning_rate": 0.0003971275011071299, "loss": 0.8631, "step": 102710 }, { "epoch": 4.13, "learning_rate": 0.0003971174362897057, "loss": 1.0441, "step": 102720 }, { "epoch": 4.13, "learning_rate": 0.0003971073714722815, "loss": 1.1881, "step": 102730 }, { "epoch": 4.13, "learning_rate": 0.0003970973066548573, "loss": 0.988, "step": 102740 }, { "epoch": 4.13, "learning_rate": 0.00039708724183743307, "loss": 0.7384, "step": 102750 }, { "epoch": 4.13, "learning_rate": 0.00039707717702000886, "loss": 1.0504, "step": 102760 }, { "epoch": 4.13, "learning_rate": 0.0003970671122025847, "loss": 1.0458, "step": 102770 }, { "epoch": 4.13, "learning_rate": 0.00039705704738516044, "loss": 1.0962, "step": 102780 }, { "epoch": 4.13, "learning_rate": 0.00039704698256773623, "loss": 1.0201, "step": 102790 }, { "epoch": 4.13, "learning_rate": 0.000397036917750312, "loss": 0.9127, "step": 102800 }, { "epoch": 4.13, "learning_rate": 0.0003970268529328878, "loss": 1.0139, "step": 102810 }, { "epoch": 4.14, "learning_rate": 0.0003970167881154636, "loss": 0.9564, "step": 102820 }, { "epoch": 4.14, "learning_rate": 0.0003970067232980394, "loss": 0.9405, "step": 102830 }, { "epoch": 4.14, "learning_rate": 0.0003969966584806152, "loss": 0.7752, "step": 102840 }, { "epoch": 4.14, "learning_rate": 0.000396986593663191, "loss": 0.8959, "step": 102850 }, { "epoch": 4.14, "learning_rate": 0.0003969765288457667, "loss": 0.7493, "step": 102860 }, { "epoch": 4.14, "learning_rate": 0.0003969664640283425, "loss": 0.9118, "step": 102870 }, { "epoch": 4.14, "learning_rate": 0.0003969563992109183, "loss": 0.9527, "step": 102880 }, { "epoch": 4.14, "learning_rate": 0.0003969463343934941, "loss": 1.0358, "step": 102890 }, { "epoch": 4.14, "learning_rate": 0.0003969362695760699, "loss": 1.0064, "step": 102900 }, { "epoch": 4.14, "learning_rate": 0.00039692620475864567, "loss": 1.0132, "step": 102910 }, { "epoch": 4.14, "learning_rate": 0.0003969161399412215, "loss": 0.7862, "step": 102920 }, { "epoch": 4.14, "learning_rate": 0.0003969060751237973, "loss": 1.0164, "step": 102930 }, { "epoch": 4.14, "learning_rate": 0.0003968960103063731, "loss": 1.1659, "step": 102940 }, { "epoch": 4.14, "learning_rate": 0.00039688594548894883, "loss": 0.9738, "step": 102950 }, { "epoch": 4.14, "learning_rate": 0.0003968758806715246, "loss": 1.1983, "step": 102960 }, { "epoch": 4.14, "learning_rate": 0.0003968658158541004, "loss": 0.8191, "step": 102970 }, { "epoch": 4.14, "learning_rate": 0.0003968557510366762, "loss": 0.9933, "step": 102980 }, { "epoch": 4.14, "learning_rate": 0.000396845686219252, "loss": 0.9228, "step": 102990 }, { "epoch": 4.14, "learning_rate": 0.0003968356214018278, "loss": 1.04, "step": 103000 }, { "epoch": 4.14, "learning_rate": 0.0003968255565844036, "loss": 0.7358, "step": 103010 }, { "epoch": 4.14, "learning_rate": 0.00039681549176697937, "loss": 1.0652, "step": 103020 }, { "epoch": 4.14, "learning_rate": 0.0003968054269495551, "loss": 0.846, "step": 103030 }, { "epoch": 4.14, "learning_rate": 0.0003967953621321309, "loss": 0.9573, "step": 103040 }, { "epoch": 4.14, "learning_rate": 0.0003967852973147067, "loss": 0.9961, "step": 103050 }, { "epoch": 4.14, "learning_rate": 0.00039677523249728253, "loss": 0.9743, "step": 103060 }, { "epoch": 4.15, "learning_rate": 0.0003967651676798583, "loss": 0.9117, "step": 103070 }, { "epoch": 4.15, "learning_rate": 0.0003967551028624341, "loss": 0.8707, "step": 103080 }, { "epoch": 4.15, "learning_rate": 0.0003967450380450099, "loss": 0.9539, "step": 103090 }, { "epoch": 4.15, "learning_rate": 0.0003967349732275857, "loss": 0.9774, "step": 103100 }, { "epoch": 4.15, "learning_rate": 0.00039672490841016143, "loss": 0.9582, "step": 103110 }, { "epoch": 4.15, "learning_rate": 0.0003967148435927372, "loss": 1.053, "step": 103120 }, { "epoch": 4.15, "learning_rate": 0.000396704778775313, "loss": 0.8253, "step": 103130 }, { "epoch": 4.15, "learning_rate": 0.0003966947139578888, "loss": 0.8722, "step": 103140 }, { "epoch": 4.15, "learning_rate": 0.0003966846491404646, "loss": 1.0003, "step": 103150 }, { "epoch": 4.15, "learning_rate": 0.0003966745843230404, "loss": 1.0077, "step": 103160 }, { "epoch": 4.15, "learning_rate": 0.0003966645195056162, "loss": 1.1202, "step": 103170 }, { "epoch": 4.15, "learning_rate": 0.00039665445468819196, "loss": 0.7647, "step": 103180 }, { "epoch": 4.15, "learning_rate": 0.00039664438987076776, "loss": 0.8429, "step": 103190 }, { "epoch": 4.15, "learning_rate": 0.0003966343250533435, "loss": 1.1246, "step": 103200 }, { "epoch": 4.15, "learning_rate": 0.00039662426023591934, "loss": 1.0693, "step": 103210 }, { "epoch": 4.15, "learning_rate": 0.00039661419541849513, "loss": 0.8857, "step": 103220 }, { "epoch": 4.15, "learning_rate": 0.0003966041306010709, "loss": 1.0516, "step": 103230 }, { "epoch": 4.15, "learning_rate": 0.0003965940657836467, "loss": 0.8478, "step": 103240 }, { "epoch": 4.15, "learning_rate": 0.0003965840009662225, "loss": 1.0255, "step": 103250 }, { "epoch": 4.15, "learning_rate": 0.0003965739361487983, "loss": 0.9447, "step": 103260 }, { "epoch": 4.15, "learning_rate": 0.0003965638713313741, "loss": 0.86, "step": 103270 }, { "epoch": 4.15, "learning_rate": 0.0003965538065139498, "loss": 0.9277, "step": 103280 }, { "epoch": 4.15, "learning_rate": 0.0003965437416965256, "loss": 1.0413, "step": 103290 }, { "epoch": 4.15, "learning_rate": 0.0003965336768791014, "loss": 0.9433, "step": 103300 }, { "epoch": 4.15, "learning_rate": 0.0003965236120616772, "loss": 0.8484, "step": 103310 }, { "epoch": 4.16, "learning_rate": 0.000396513547244253, "loss": 1.0114, "step": 103320 }, { "epoch": 4.16, "learning_rate": 0.00039650348242682877, "loss": 0.9565, "step": 103330 }, { "epoch": 4.16, "learning_rate": 0.00039649341760940456, "loss": 1.0371, "step": 103340 }, { "epoch": 4.16, "learning_rate": 0.00039648335279198035, "loss": 1.0095, "step": 103350 }, { "epoch": 4.16, "learning_rate": 0.0003964732879745562, "loss": 1.0739, "step": 103360 }, { "epoch": 4.16, "learning_rate": 0.00039646322315713194, "loss": 1.1021, "step": 103370 }, { "epoch": 4.16, "learning_rate": 0.0003964531583397077, "loss": 1.0252, "step": 103380 }, { "epoch": 4.16, "learning_rate": 0.0003964430935222835, "loss": 0.8769, "step": 103390 }, { "epoch": 4.16, "learning_rate": 0.0003964330287048593, "loss": 0.8593, "step": 103400 }, { "epoch": 4.16, "learning_rate": 0.0003964229638874351, "loss": 1.0257, "step": 103410 }, { "epoch": 4.16, "learning_rate": 0.0003964128990700109, "loss": 0.825, "step": 103420 }, { "epoch": 4.16, "learning_rate": 0.0003964028342525867, "loss": 0.8134, "step": 103430 }, { "epoch": 4.16, "learning_rate": 0.00039639276943516247, "loss": 0.9475, "step": 103440 }, { "epoch": 4.16, "learning_rate": 0.0003963827046177382, "loss": 0.9466, "step": 103450 }, { "epoch": 4.16, "learning_rate": 0.000396372639800314, "loss": 1.1118, "step": 103460 }, { "epoch": 4.16, "learning_rate": 0.0003963625749828898, "loss": 1.0519, "step": 103470 }, { "epoch": 4.16, "learning_rate": 0.0003963525101654656, "loss": 0.8974, "step": 103480 }, { "epoch": 4.16, "learning_rate": 0.00039634244534804137, "loss": 0.8415, "step": 103490 }, { "epoch": 4.16, "learning_rate": 0.0003963323805306172, "loss": 0.9224, "step": 103500 }, { "epoch": 4.16, "learning_rate": 0.000396322315713193, "loss": 0.9767, "step": 103510 }, { "epoch": 4.16, "learning_rate": 0.0003963122508957688, "loss": 1.0714, "step": 103520 }, { "epoch": 4.16, "learning_rate": 0.0003963021860783446, "loss": 0.862, "step": 103530 }, { "epoch": 4.16, "learning_rate": 0.0003962921212609203, "loss": 1.0161, "step": 103540 }, { "epoch": 4.16, "learning_rate": 0.0003962820564434961, "loss": 1.1136, "step": 103550 }, { "epoch": 4.17, "learning_rate": 0.0003962719916260719, "loss": 0.8491, "step": 103560 }, { "epoch": 4.17, "learning_rate": 0.0003962619268086477, "loss": 1.1858, "step": 103570 }, { "epoch": 4.17, "learning_rate": 0.0003962518619912235, "loss": 1.0504, "step": 103580 }, { "epoch": 4.17, "learning_rate": 0.0003962417971737993, "loss": 1.2164, "step": 103590 }, { "epoch": 4.17, "learning_rate": 0.00039623173235637507, "loss": 0.9848, "step": 103600 }, { "epoch": 4.17, "learning_rate": 0.00039622166753895086, "loss": 0.9433, "step": 103610 }, { "epoch": 4.17, "learning_rate": 0.0003962116027215266, "loss": 0.9492, "step": 103620 }, { "epoch": 4.17, "learning_rate": 0.0003962015379041024, "loss": 1.2142, "step": 103630 }, { "epoch": 4.17, "learning_rate": 0.0003961914730866782, "loss": 0.9461, "step": 103640 }, { "epoch": 4.17, "learning_rate": 0.000396181408269254, "loss": 0.8612, "step": 103650 }, { "epoch": 4.17, "learning_rate": 0.0003961713434518298, "loss": 0.9324, "step": 103660 }, { "epoch": 4.17, "learning_rate": 0.0003961612786344056, "loss": 0.972, "step": 103670 }, { "epoch": 4.17, "learning_rate": 0.0003961512138169814, "loss": 0.9261, "step": 103680 }, { "epoch": 4.17, "learning_rate": 0.0003961411489995572, "loss": 0.9834, "step": 103690 }, { "epoch": 4.17, "learning_rate": 0.000396131084182133, "loss": 1.1492, "step": 103700 }, { "epoch": 4.17, "learning_rate": 0.0003961210193647087, "loss": 1.05, "step": 103710 }, { "epoch": 4.17, "learning_rate": 0.0003961109545472845, "loss": 1.0001, "step": 103720 }, { "epoch": 4.17, "learning_rate": 0.0003961008897298603, "loss": 0.9361, "step": 103730 }, { "epoch": 4.17, "learning_rate": 0.0003960908249124361, "loss": 0.7637, "step": 103740 }, { "epoch": 4.17, "learning_rate": 0.0003960807600950119, "loss": 0.7973, "step": 103750 }, { "epoch": 4.17, "learning_rate": 0.00039607069527758767, "loss": 1.003, "step": 103760 }, { "epoch": 4.17, "learning_rate": 0.00039606063046016346, "loss": 0.8681, "step": 103770 }, { "epoch": 4.17, "learning_rate": 0.00039605056564273925, "loss": 0.7759, "step": 103780 }, { "epoch": 4.17, "learning_rate": 0.00039604050082531504, "loss": 1.2307, "step": 103790 }, { "epoch": 4.17, "learning_rate": 0.00039603043600789083, "loss": 1.0799, "step": 103800 }, { "epoch": 4.18, "learning_rate": 0.0003960203711904666, "loss": 0.8079, "step": 103810 }, { "epoch": 4.18, "learning_rate": 0.0003960103063730424, "loss": 1.2079, "step": 103820 }, { "epoch": 4.18, "learning_rate": 0.0003960002415556182, "loss": 0.7509, "step": 103830 }, { "epoch": 4.18, "learning_rate": 0.000395990176738194, "loss": 0.7731, "step": 103840 }, { "epoch": 4.18, "learning_rate": 0.0003959801119207698, "loss": 0.9989, "step": 103850 }, { "epoch": 4.18, "learning_rate": 0.0003959700471033456, "loss": 1.0552, "step": 103860 }, { "epoch": 4.18, "learning_rate": 0.0003959599822859213, "loss": 1.0674, "step": 103870 }, { "epoch": 4.18, "learning_rate": 0.0003959499174684971, "loss": 1.0501, "step": 103880 }, { "epoch": 4.18, "learning_rate": 0.0003959398526510729, "loss": 1.0515, "step": 103890 }, { "epoch": 4.18, "learning_rate": 0.0003959297878336487, "loss": 0.8474, "step": 103900 }, { "epoch": 4.18, "learning_rate": 0.0003959197230162245, "loss": 1.0377, "step": 103910 }, { "epoch": 4.18, "learning_rate": 0.00039590965819880027, "loss": 1.0075, "step": 103920 }, { "epoch": 4.18, "learning_rate": 0.00039589959338137606, "loss": 0.9122, "step": 103930 }, { "epoch": 4.18, "learning_rate": 0.0003958895285639519, "loss": 1.1282, "step": 103940 }, { "epoch": 4.18, "learning_rate": 0.0003958794637465277, "loss": 1.0958, "step": 103950 }, { "epoch": 4.18, "learning_rate": 0.00039586939892910343, "loss": 0.9727, "step": 103960 }, { "epoch": 4.18, "learning_rate": 0.0003958593341116792, "loss": 0.975, "step": 103970 }, { "epoch": 4.18, "learning_rate": 0.000395849269294255, "loss": 1.0539, "step": 103980 }, { "epoch": 4.18, "learning_rate": 0.0003958392044768308, "loss": 0.9516, "step": 103990 }, { "epoch": 4.18, "learning_rate": 0.0003958291396594066, "loss": 1.0144, "step": 104000 }, { "epoch": 4.18, "learning_rate": 0.0003958190748419824, "loss": 1.2234, "step": 104010 }, { "epoch": 4.18, "learning_rate": 0.0003958090100245582, "loss": 0.9167, "step": 104020 }, { "epoch": 4.18, "learning_rate": 0.00039579894520713396, "loss": 0.8509, "step": 104030 }, { "epoch": 4.18, "learning_rate": 0.0003957888803897097, "loss": 1.1098, "step": 104040 }, { "epoch": 4.18, "learning_rate": 0.0003957788155722855, "loss": 0.8809, "step": 104050 }, { "epoch": 4.19, "learning_rate": 0.0003957687507548613, "loss": 1.0512, "step": 104060 }, { "epoch": 4.19, "learning_rate": 0.0003957586859374371, "loss": 1.1167, "step": 104070 }, { "epoch": 4.19, "learning_rate": 0.00039574862112001286, "loss": 1.0754, "step": 104080 }, { "epoch": 4.19, "learning_rate": 0.0003957385563025887, "loss": 0.8562, "step": 104090 }, { "epoch": 4.19, "learning_rate": 0.0003957284914851645, "loss": 0.8439, "step": 104100 }, { "epoch": 4.19, "learning_rate": 0.0003957184266677403, "loss": 1.0118, "step": 104110 }, { "epoch": 4.19, "learning_rate": 0.0003957083618503161, "loss": 1.1606, "step": 104120 }, { "epoch": 4.19, "learning_rate": 0.0003956982970328918, "loss": 0.9417, "step": 104130 }, { "epoch": 4.19, "learning_rate": 0.0003956882322154676, "loss": 0.9749, "step": 104140 }, { "epoch": 4.19, "learning_rate": 0.0003956781673980434, "loss": 0.9669, "step": 104150 }, { "epoch": 4.19, "learning_rate": 0.0003956681025806192, "loss": 1.0411, "step": 104160 }, { "epoch": 4.19, "learning_rate": 0.000395658037763195, "loss": 0.8656, "step": 104170 }, { "epoch": 4.19, "learning_rate": 0.00039564797294577077, "loss": 0.9037, "step": 104180 }, { "epoch": 4.19, "learning_rate": 0.00039563790812834656, "loss": 1.1469, "step": 104190 }, { "epoch": 4.19, "learning_rate": 0.00039562784331092235, "loss": 0.9919, "step": 104200 }, { "epoch": 4.19, "learning_rate": 0.0003956177784934981, "loss": 1.0136, "step": 104210 }, { "epoch": 4.19, "learning_rate": 0.0003956077136760739, "loss": 1.0195, "step": 104220 }, { "epoch": 4.19, "learning_rate": 0.0003955976488586497, "loss": 0.9758, "step": 104230 }, { "epoch": 4.19, "learning_rate": 0.0003955875840412255, "loss": 0.9311, "step": 104240 }, { "epoch": 4.19, "learning_rate": 0.0003955775192238013, "loss": 1.0084, "step": 104250 }, { "epoch": 4.19, "learning_rate": 0.0003955674544063771, "loss": 0.9411, "step": 104260 }, { "epoch": 4.19, "learning_rate": 0.0003955573895889529, "loss": 0.9145, "step": 104270 }, { "epoch": 4.19, "learning_rate": 0.0003955473247715287, "loss": 0.9028, "step": 104280 }, { "epoch": 4.19, "learning_rate": 0.00039553725995410447, "loss": 0.9524, "step": 104290 }, { "epoch": 4.19, "learning_rate": 0.0003955271951366802, "loss": 1.0573, "step": 104300 }, { "epoch": 4.2, "learning_rate": 0.000395517130319256, "loss": 0.9605, "step": 104310 }, { "epoch": 4.2, "learning_rate": 0.0003955070655018318, "loss": 0.9007, "step": 104320 }, { "epoch": 4.2, "learning_rate": 0.0003954970006844076, "loss": 1.2186, "step": 104330 }, { "epoch": 4.2, "learning_rate": 0.00039548693586698337, "loss": 0.9996, "step": 104340 }, { "epoch": 4.2, "learning_rate": 0.00039547687104955916, "loss": 1.0507, "step": 104350 }, { "epoch": 4.2, "learning_rate": 0.00039546680623213495, "loss": 0.9507, "step": 104360 }, { "epoch": 4.2, "learning_rate": 0.00039545674141471074, "loss": 1.1079, "step": 104370 }, { "epoch": 4.2, "learning_rate": 0.00039544667659728653, "loss": 0.9813, "step": 104380 }, { "epoch": 4.2, "learning_rate": 0.0003954366117798623, "loss": 1.0086, "step": 104390 }, { "epoch": 4.2, "learning_rate": 0.0003954265469624381, "loss": 0.9724, "step": 104400 }, { "epoch": 4.2, "learning_rate": 0.0003954164821450139, "loss": 1.1007, "step": 104410 }, { "epoch": 4.2, "learning_rate": 0.0003954064173275897, "loss": 0.9187, "step": 104420 }, { "epoch": 4.2, "learning_rate": 0.0003953963525101655, "loss": 0.9661, "step": 104430 }, { "epoch": 4.2, "learning_rate": 0.0003953862876927413, "loss": 1.0695, "step": 104440 }, { "epoch": 4.2, "learning_rate": 0.00039537622287531707, "loss": 0.9257, "step": 104450 }, { "epoch": 4.2, "learning_rate": 0.00039536615805789286, "loss": 0.858, "step": 104460 }, { "epoch": 4.2, "learning_rate": 0.0003953560932404686, "loss": 0.8623, "step": 104470 }, { "epoch": 4.2, "learning_rate": 0.0003953460284230444, "loss": 1.0406, "step": 104480 }, { "epoch": 4.2, "learning_rate": 0.0003953359636056202, "loss": 1.1293, "step": 104490 }, { "epoch": 4.2, "learning_rate": 0.00039532589878819597, "loss": 0.9764, "step": 104500 }, { "epoch": 4.2, "learning_rate": 0.00039531583397077176, "loss": 0.9589, "step": 104510 }, { "epoch": 4.2, "learning_rate": 0.0003953057691533476, "loss": 0.898, "step": 104520 }, { "epoch": 4.2, "learning_rate": 0.0003952957043359234, "loss": 0.8688, "step": 104530 }, { "epoch": 4.2, "learning_rate": 0.0003952856395184992, "loss": 0.8564, "step": 104540 }, { "epoch": 4.2, "learning_rate": 0.0003952755747010749, "loss": 0.8251, "step": 104550 }, { "epoch": 4.21, "learning_rate": 0.0003952655098836507, "loss": 0.8497, "step": 104560 }, { "epoch": 4.21, "learning_rate": 0.0003952554450662265, "loss": 0.9294, "step": 104570 }, { "epoch": 4.21, "learning_rate": 0.0003952453802488023, "loss": 0.9518, "step": 104580 }, { "epoch": 4.21, "learning_rate": 0.0003952353154313781, "loss": 0.8117, "step": 104590 }, { "epoch": 4.21, "learning_rate": 0.0003952252506139539, "loss": 1.1006, "step": 104600 }, { "epoch": 4.21, "learning_rate": 0.00039521518579652967, "loss": 1.2261, "step": 104610 }, { "epoch": 4.21, "learning_rate": 0.00039520512097910546, "loss": 1.051, "step": 104620 }, { "epoch": 4.21, "learning_rate": 0.0003951950561616812, "loss": 0.8751, "step": 104630 }, { "epoch": 4.21, "learning_rate": 0.000395184991344257, "loss": 1.0231, "step": 104640 }, { "epoch": 4.21, "learning_rate": 0.0003951749265268328, "loss": 1.2486, "step": 104650 }, { "epoch": 4.21, "learning_rate": 0.00039516486170940857, "loss": 0.9683, "step": 104660 }, { "epoch": 4.21, "learning_rate": 0.0003951547968919844, "loss": 0.7784, "step": 104670 }, { "epoch": 4.21, "learning_rate": 0.0003951447320745602, "loss": 0.7874, "step": 104680 }, { "epoch": 4.21, "learning_rate": 0.000395134667257136, "loss": 0.9537, "step": 104690 }, { "epoch": 4.21, "learning_rate": 0.0003951246024397118, "loss": 0.7642, "step": 104700 }, { "epoch": 4.21, "learning_rate": 0.0003951145376222876, "loss": 1.1267, "step": 104710 }, { "epoch": 4.21, "learning_rate": 0.0003951044728048633, "loss": 0.8796, "step": 104720 }, { "epoch": 4.21, "learning_rate": 0.0003950944079874391, "loss": 1.0462, "step": 104730 }, { "epoch": 4.21, "learning_rate": 0.0003950843431700149, "loss": 0.7449, "step": 104740 }, { "epoch": 4.21, "learning_rate": 0.0003950742783525907, "loss": 0.7496, "step": 104750 }, { "epoch": 4.21, "learning_rate": 0.0003950642135351665, "loss": 1.1334, "step": 104760 }, { "epoch": 4.21, "learning_rate": 0.00039505414871774227, "loss": 1.0173, "step": 104770 }, { "epoch": 4.21, "learning_rate": 0.00039504408390031806, "loss": 0.9723, "step": 104780 }, { "epoch": 4.21, "learning_rate": 0.00039503401908289385, "loss": 1.0467, "step": 104790 }, { "epoch": 4.21, "learning_rate": 0.0003950239542654696, "loss": 1.0291, "step": 104800 }, { "epoch": 4.22, "learning_rate": 0.0003950138894480454, "loss": 0.9058, "step": 104810 }, { "epoch": 4.22, "learning_rate": 0.0003950038246306212, "loss": 1.0555, "step": 104820 }, { "epoch": 4.22, "learning_rate": 0.000394993759813197, "loss": 1.1695, "step": 104830 }, { "epoch": 4.22, "learning_rate": 0.0003949836949957728, "loss": 1.0127, "step": 104840 }, { "epoch": 4.22, "learning_rate": 0.0003949736301783486, "loss": 1.1441, "step": 104850 }, { "epoch": 4.22, "learning_rate": 0.0003949635653609244, "loss": 0.7747, "step": 104860 }, { "epoch": 4.22, "learning_rate": 0.0003949535005435002, "loss": 0.8218, "step": 104870 }, { "epoch": 4.22, "learning_rate": 0.00039494343572607596, "loss": 0.8132, "step": 104880 }, { "epoch": 4.22, "learning_rate": 0.0003949333709086517, "loss": 1.0272, "step": 104890 }, { "epoch": 4.22, "learning_rate": 0.0003949233060912275, "loss": 0.7564, "step": 104900 }, { "epoch": 4.22, "learning_rate": 0.0003949132412738033, "loss": 0.9176, "step": 104910 }, { "epoch": 4.22, "learning_rate": 0.0003949031764563791, "loss": 0.8043, "step": 104920 }, { "epoch": 4.22, "learning_rate": 0.00039489311163895486, "loss": 0.8099, "step": 104930 }, { "epoch": 4.22, "learning_rate": 0.00039488304682153065, "loss": 0.9292, "step": 104940 }, { "epoch": 4.22, "learning_rate": 0.00039487298200410645, "loss": 1.1126, "step": 104950 }, { "epoch": 4.22, "learning_rate": 0.0003948629171866823, "loss": 1.0294, "step": 104960 }, { "epoch": 4.22, "learning_rate": 0.00039485285236925803, "loss": 0.755, "step": 104970 }, { "epoch": 4.22, "learning_rate": 0.0003948427875518338, "loss": 0.6977, "step": 104980 }, { "epoch": 4.22, "learning_rate": 0.0003948327227344096, "loss": 1.1078, "step": 104990 }, { "epoch": 4.22, "learning_rate": 0.0003948226579169854, "loss": 0.9335, "step": 105000 }, { "epoch": 4.22, "learning_rate": 0.0003948125930995612, "loss": 1.1926, "step": 105010 }, { "epoch": 4.22, "learning_rate": 0.000394802528282137, "loss": 0.9542, "step": 105020 }, { "epoch": 4.22, "learning_rate": 0.00039479246346471277, "loss": 0.9796, "step": 105030 }, { "epoch": 4.22, "learning_rate": 0.00039478239864728856, "loss": 0.8123, "step": 105040 }, { "epoch": 4.22, "learning_rate": 0.00039477233382986435, "loss": 1.012, "step": 105050 }, { "epoch": 4.23, "learning_rate": 0.0003947622690124401, "loss": 0.9873, "step": 105060 }, { "epoch": 4.23, "learning_rate": 0.0003947522041950159, "loss": 0.917, "step": 105070 }, { "epoch": 4.23, "learning_rate": 0.00039474213937759167, "loss": 0.8023, "step": 105080 }, { "epoch": 4.23, "learning_rate": 0.00039473207456016746, "loss": 1.0457, "step": 105090 }, { "epoch": 4.23, "learning_rate": 0.00039472200974274325, "loss": 0.9488, "step": 105100 }, { "epoch": 4.23, "learning_rate": 0.0003947119449253191, "loss": 1.1359, "step": 105110 }, { "epoch": 4.23, "learning_rate": 0.0003947018801078949, "loss": 1.0717, "step": 105120 }, { "epoch": 4.23, "learning_rate": 0.0003946918152904707, "loss": 1.0039, "step": 105130 }, { "epoch": 4.23, "learning_rate": 0.0003946817504730464, "loss": 1.2071, "step": 105140 }, { "epoch": 4.23, "learning_rate": 0.0003946716856556222, "loss": 1.1617, "step": 105150 }, { "epoch": 4.23, "learning_rate": 0.000394661620838198, "loss": 0.9041, "step": 105160 }, { "epoch": 4.23, "learning_rate": 0.0003946515560207738, "loss": 0.9538, "step": 105170 }, { "epoch": 4.23, "learning_rate": 0.0003946414912033496, "loss": 0.8966, "step": 105180 }, { "epoch": 4.23, "learning_rate": 0.00039463142638592537, "loss": 0.884, "step": 105190 }, { "epoch": 4.23, "learning_rate": 0.00039462136156850116, "loss": 0.9187, "step": 105200 }, { "epoch": 4.23, "learning_rate": 0.00039461129675107695, "loss": 0.8212, "step": 105210 }, { "epoch": 4.23, "learning_rate": 0.00039460123193365274, "loss": 0.9928, "step": 105220 }, { "epoch": 4.23, "learning_rate": 0.0003945911671162285, "loss": 1.1264, "step": 105230 }, { "epoch": 4.23, "learning_rate": 0.00039458110229880427, "loss": 1.0793, "step": 105240 }, { "epoch": 4.23, "learning_rate": 0.00039457103748138006, "loss": 0.9381, "step": 105250 }, { "epoch": 4.23, "learning_rate": 0.0003945609726639559, "loss": 1.1608, "step": 105260 }, { "epoch": 4.23, "learning_rate": 0.0003945509078465317, "loss": 1.1816, "step": 105270 }, { "epoch": 4.23, "learning_rate": 0.0003945408430291075, "loss": 0.9814, "step": 105280 }, { "epoch": 4.23, "learning_rate": 0.0003945307782116833, "loss": 1.0983, "step": 105290 }, { "epoch": 4.23, "learning_rate": 0.00039452071339425907, "loss": 0.769, "step": 105300 }, { "epoch": 4.24, "learning_rate": 0.0003945106485768348, "loss": 1.0075, "step": 105310 }, { "epoch": 4.24, "learning_rate": 0.0003945005837594106, "loss": 0.8871, "step": 105320 }, { "epoch": 4.24, "learning_rate": 0.0003944905189419864, "loss": 1.2145, "step": 105330 }, { "epoch": 4.24, "learning_rate": 0.0003944804541245622, "loss": 0.995, "step": 105340 }, { "epoch": 4.24, "learning_rate": 0.00039447038930713797, "loss": 0.8186, "step": 105350 }, { "epoch": 4.24, "learning_rate": 0.00039446032448971376, "loss": 1.0781, "step": 105360 }, { "epoch": 4.24, "learning_rate": 0.00039445025967228955, "loss": 0.856, "step": 105370 }, { "epoch": 4.24, "learning_rate": 0.00039444019485486534, "loss": 0.9726, "step": 105380 }, { "epoch": 4.24, "learning_rate": 0.0003944301300374411, "loss": 0.936, "step": 105390 }, { "epoch": 4.24, "learning_rate": 0.0003944200652200169, "loss": 0.8985, "step": 105400 }, { "epoch": 4.24, "learning_rate": 0.0003944100004025927, "loss": 0.9022, "step": 105410 }, { "epoch": 4.24, "learning_rate": 0.0003943999355851685, "loss": 0.9648, "step": 105420 }, { "epoch": 4.24, "learning_rate": 0.0003943898707677443, "loss": 1.0224, "step": 105430 }, { "epoch": 4.24, "learning_rate": 0.0003943798059503201, "loss": 1.0576, "step": 105440 }, { "epoch": 4.24, "learning_rate": 0.0003943697411328959, "loss": 1.0041, "step": 105450 }, { "epoch": 4.24, "learning_rate": 0.00039435967631547167, "loss": 0.9114, "step": 105460 }, { "epoch": 4.24, "learning_rate": 0.00039434961149804746, "loss": 0.9202, "step": 105470 }, { "epoch": 4.24, "learning_rate": 0.0003943395466806232, "loss": 0.8707, "step": 105480 }, { "epoch": 4.24, "learning_rate": 0.000394329481863199, "loss": 0.8268, "step": 105490 }, { "epoch": 4.24, "learning_rate": 0.0003943194170457748, "loss": 1.0722, "step": 105500 }, { "epoch": 4.24, "learning_rate": 0.00039430935222835057, "loss": 0.9996, "step": 105510 }, { "epoch": 4.24, "learning_rate": 0.00039429928741092636, "loss": 0.7361, "step": 105520 }, { "epoch": 4.24, "learning_rate": 0.00039428922259350215, "loss": 0.9849, "step": 105530 }, { "epoch": 4.24, "learning_rate": 0.00039427915777607794, "loss": 0.9866, "step": 105540 }, { "epoch": 4.25, "learning_rate": 0.0003942690929586538, "loss": 0.9912, "step": 105550 }, { "epoch": 4.25, "learning_rate": 0.0003942590281412295, "loss": 0.979, "step": 105560 }, { "epoch": 4.25, "learning_rate": 0.0003942489633238053, "loss": 1.0685, "step": 105570 }, { "epoch": 4.25, "learning_rate": 0.0003942388985063811, "loss": 0.8588, "step": 105580 }, { "epoch": 4.25, "learning_rate": 0.0003942288336889569, "loss": 1.2369, "step": 105590 }, { "epoch": 4.25, "learning_rate": 0.0003942187688715327, "loss": 0.8026, "step": 105600 }, { "epoch": 4.25, "learning_rate": 0.0003942087040541085, "loss": 1.2635, "step": 105610 }, { "epoch": 4.25, "learning_rate": 0.00039419863923668427, "loss": 1.3864, "step": 105620 }, { "epoch": 4.25, "learning_rate": 0.00039418857441926006, "loss": 1.057, "step": 105630 }, { "epoch": 4.25, "learning_rate": 0.00039417850960183585, "loss": 1.0669, "step": 105640 }, { "epoch": 4.25, "learning_rate": 0.0003941684447844116, "loss": 1.0186, "step": 105650 }, { "epoch": 4.25, "learning_rate": 0.0003941583799669874, "loss": 1.0948, "step": 105660 }, { "epoch": 4.25, "learning_rate": 0.00039414831514956317, "loss": 1.1443, "step": 105670 }, { "epoch": 4.25, "learning_rate": 0.00039413825033213896, "loss": 0.7714, "step": 105680 }, { "epoch": 4.25, "learning_rate": 0.0003941281855147148, "loss": 1.0032, "step": 105690 }, { "epoch": 4.25, "learning_rate": 0.0003941181206972906, "loss": 0.85, "step": 105700 }, { "epoch": 4.25, "learning_rate": 0.0003941080558798664, "loss": 1.0781, "step": 105710 }, { "epoch": 4.25, "learning_rate": 0.0003940979910624422, "loss": 0.9063, "step": 105720 }, { "epoch": 4.25, "learning_rate": 0.0003940879262450179, "loss": 1.1006, "step": 105730 }, { "epoch": 4.25, "learning_rate": 0.0003940778614275937, "loss": 1.0739, "step": 105740 }, { "epoch": 4.25, "learning_rate": 0.0003940677966101695, "loss": 0.9704, "step": 105750 }, { "epoch": 4.25, "learning_rate": 0.0003940577317927453, "loss": 0.8747, "step": 105760 }, { "epoch": 4.25, "learning_rate": 0.0003940476669753211, "loss": 1.0903, "step": 105770 }, { "epoch": 4.25, "learning_rate": 0.00039403760215789686, "loss": 0.9662, "step": 105780 }, { "epoch": 4.25, "learning_rate": 0.00039402753734047265, "loss": 0.9124, "step": 105790 }, { "epoch": 4.26, "learning_rate": 0.00039401747252304845, "loss": 0.9645, "step": 105800 }, { "epoch": 4.26, "learning_rate": 0.00039400740770562424, "loss": 0.9195, "step": 105810 }, { "epoch": 4.26, "learning_rate": 0.00039399734288819997, "loss": 0.887, "step": 105820 }, { "epoch": 4.26, "learning_rate": 0.00039398727807077576, "loss": 0.9065, "step": 105830 }, { "epoch": 4.26, "learning_rate": 0.0003939772132533516, "loss": 0.9572, "step": 105840 }, { "epoch": 4.26, "learning_rate": 0.0003939671484359274, "loss": 0.91, "step": 105850 }, { "epoch": 4.26, "learning_rate": 0.0003939570836185032, "loss": 1.159, "step": 105860 }, { "epoch": 4.26, "learning_rate": 0.000393947018801079, "loss": 0.9835, "step": 105870 }, { "epoch": 4.26, "learning_rate": 0.00039393695398365477, "loss": 0.9159, "step": 105880 }, { "epoch": 4.26, "learning_rate": 0.00039392688916623056, "loss": 1.1243, "step": 105890 }, { "epoch": 4.26, "learning_rate": 0.0003939168243488063, "loss": 1.032, "step": 105900 }, { "epoch": 4.26, "learning_rate": 0.0003939067595313821, "loss": 0.7323, "step": 105910 }, { "epoch": 4.26, "learning_rate": 0.0003938966947139579, "loss": 1.0299, "step": 105920 }, { "epoch": 4.26, "learning_rate": 0.00039388662989653367, "loss": 0.8907, "step": 105930 }, { "epoch": 4.26, "learning_rate": 0.00039387656507910946, "loss": 0.7748, "step": 105940 }, { "epoch": 4.26, "learning_rate": 0.00039386650026168525, "loss": 1.0943, "step": 105950 }, { "epoch": 4.26, "learning_rate": 0.00039385643544426104, "loss": 1.1798, "step": 105960 }, { "epoch": 4.26, "learning_rate": 0.00039384637062683683, "loss": 1.2033, "step": 105970 }, { "epoch": 4.26, "learning_rate": 0.0003938363058094126, "loss": 1.018, "step": 105980 }, { "epoch": 4.26, "learning_rate": 0.0003938262409919884, "loss": 1.0998, "step": 105990 }, { "epoch": 4.26, "learning_rate": 0.0003938161761745642, "loss": 1.0382, "step": 106000 }, { "epoch": 4.26, "learning_rate": 0.00039380611135714, "loss": 1.151, "step": 106010 }, { "epoch": 4.26, "learning_rate": 0.0003937960465397158, "loss": 0.9081, "step": 106020 }, { "epoch": 4.26, "learning_rate": 0.0003937859817222916, "loss": 0.9696, "step": 106030 }, { "epoch": 4.26, "learning_rate": 0.00039377591690486737, "loss": 0.9543, "step": 106040 }, { "epoch": 4.27, "learning_rate": 0.00039376585208744316, "loss": 1.0499, "step": 106050 }, { "epoch": 4.27, "learning_rate": 0.00039375578727001895, "loss": 0.8898, "step": 106060 }, { "epoch": 4.27, "learning_rate": 0.0003937457224525947, "loss": 0.8915, "step": 106070 }, { "epoch": 4.27, "learning_rate": 0.0003937356576351705, "loss": 0.9896, "step": 106080 }, { "epoch": 4.27, "learning_rate": 0.00039372559281774627, "loss": 1.0171, "step": 106090 }, { "epoch": 4.27, "learning_rate": 0.00039371552800032206, "loss": 1.1801, "step": 106100 }, { "epoch": 4.27, "learning_rate": 0.00039370546318289785, "loss": 0.9206, "step": 106110 }, { "epoch": 4.27, "learning_rate": 0.00039369539836547364, "loss": 0.9677, "step": 106120 }, { "epoch": 4.27, "learning_rate": 0.0003936853335480495, "loss": 0.7984, "step": 106130 }, { "epoch": 4.27, "learning_rate": 0.0003936752687306253, "loss": 0.9105, "step": 106140 }, { "epoch": 4.27, "learning_rate": 0.00039366520391320107, "loss": 1.2197, "step": 106150 }, { "epoch": 4.27, "learning_rate": 0.0003936551390957768, "loss": 1.0385, "step": 106160 }, { "epoch": 4.27, "learning_rate": 0.0003936450742783526, "loss": 0.9613, "step": 106170 }, { "epoch": 4.27, "learning_rate": 0.0003936350094609284, "loss": 0.9703, "step": 106180 }, { "epoch": 4.27, "learning_rate": 0.0003936249446435042, "loss": 0.8957, "step": 106190 }, { "epoch": 4.27, "learning_rate": 0.00039361487982607997, "loss": 1.1679, "step": 106200 }, { "epoch": 4.27, "learning_rate": 0.00039360481500865576, "loss": 1.0943, "step": 106210 }, { "epoch": 4.27, "learning_rate": 0.00039359475019123155, "loss": 1.0828, "step": 106220 }, { "epoch": 4.27, "learning_rate": 0.00039358468537380734, "loss": 1.1431, "step": 106230 }, { "epoch": 4.27, "learning_rate": 0.0003935746205563831, "loss": 0.9832, "step": 106240 }, { "epoch": 4.27, "learning_rate": 0.00039356455573895887, "loss": 0.9772, "step": 106250 }, { "epoch": 4.27, "learning_rate": 0.00039355449092153466, "loss": 0.928, "step": 106260 }, { "epoch": 4.27, "learning_rate": 0.00039354442610411045, "loss": 0.8984, "step": 106270 }, { "epoch": 4.27, "learning_rate": 0.0003935343612866863, "loss": 1.1714, "step": 106280 }, { "epoch": 4.27, "learning_rate": 0.0003935242964692621, "loss": 1.1446, "step": 106290 }, { "epoch": 4.28, "learning_rate": 0.0003935142316518379, "loss": 0.9108, "step": 106300 }, { "epoch": 4.28, "learning_rate": 0.00039350416683441367, "loss": 1.1543, "step": 106310 }, { "epoch": 4.28, "learning_rate": 0.0003934941020169894, "loss": 1.2129, "step": 106320 }, { "epoch": 4.28, "learning_rate": 0.0003934840371995652, "loss": 1.1208, "step": 106330 }, { "epoch": 4.28, "learning_rate": 0.000393473972382141, "loss": 1.093, "step": 106340 }, { "epoch": 4.28, "learning_rate": 0.0003934639075647168, "loss": 1.0172, "step": 106350 }, { "epoch": 4.28, "learning_rate": 0.00039345384274729257, "loss": 1.0255, "step": 106360 }, { "epoch": 4.28, "learning_rate": 0.00039344377792986836, "loss": 0.8687, "step": 106370 }, { "epoch": 4.28, "learning_rate": 0.00039343371311244415, "loss": 1.1583, "step": 106380 }, { "epoch": 4.28, "learning_rate": 0.00039342364829501994, "loss": 0.8192, "step": 106390 }, { "epoch": 4.28, "learning_rate": 0.00039341358347759573, "loss": 0.851, "step": 106400 }, { "epoch": 4.28, "learning_rate": 0.00039340351866017147, "loss": 1.1343, "step": 106410 }, { "epoch": 4.28, "learning_rate": 0.00039339345384274726, "loss": 0.8284, "step": 106420 }, { "epoch": 4.28, "learning_rate": 0.0003933833890253231, "loss": 0.9657, "step": 106430 }, { "epoch": 4.28, "learning_rate": 0.0003933733242078989, "loss": 1.0053, "step": 106440 }, { "epoch": 4.28, "learning_rate": 0.0003933632593904747, "loss": 0.9765, "step": 106450 }, { "epoch": 4.28, "learning_rate": 0.0003933531945730505, "loss": 0.8498, "step": 106460 }, { "epoch": 4.28, "learning_rate": 0.00039334312975562627, "loss": 1.1064, "step": 106470 }, { "epoch": 4.28, "learning_rate": 0.00039333306493820206, "loss": 0.9496, "step": 106480 }, { "epoch": 4.28, "learning_rate": 0.0003933230001207778, "loss": 0.9585, "step": 106490 }, { "epoch": 4.28, "learning_rate": 0.0003933129353033536, "loss": 1.1191, "step": 106500 }, { "epoch": 4.28, "learning_rate": 0.0003933028704859294, "loss": 0.9944, "step": 106510 }, { "epoch": 4.28, "learning_rate": 0.00039329280566850516, "loss": 0.9002, "step": 106520 }, { "epoch": 4.28, "learning_rate": 0.00039328274085108096, "loss": 1.0472, "step": 106530 }, { "epoch": 4.28, "learning_rate": 0.00039327267603365675, "loss": 1.0118, "step": 106540 }, { "epoch": 4.29, "learning_rate": 0.00039326261121623254, "loss": 0.9681, "step": 106550 }, { "epoch": 4.29, "learning_rate": 0.00039325254639880833, "loss": 1.1185, "step": 106560 }, { "epoch": 4.29, "learning_rate": 0.0003932424815813842, "loss": 1.0064, "step": 106570 }, { "epoch": 4.29, "learning_rate": 0.0003932324167639599, "loss": 1.268, "step": 106580 }, { "epoch": 4.29, "learning_rate": 0.0003932223519465357, "loss": 1.0146, "step": 106590 }, { "epoch": 4.29, "learning_rate": 0.0003932122871291115, "loss": 0.9412, "step": 106600 }, { "epoch": 4.29, "learning_rate": 0.0003932022223116873, "loss": 1.2643, "step": 106610 }, { "epoch": 4.29, "learning_rate": 0.00039319215749426307, "loss": 0.9479, "step": 106620 }, { "epoch": 4.29, "learning_rate": 0.00039318209267683886, "loss": 0.8172, "step": 106630 }, { "epoch": 4.29, "learning_rate": 0.00039317202785941465, "loss": 1.2756, "step": 106640 }, { "epoch": 4.29, "learning_rate": 0.00039316196304199045, "loss": 0.9753, "step": 106650 }, { "epoch": 4.29, "learning_rate": 0.0003931518982245662, "loss": 1.0494, "step": 106660 }, { "epoch": 4.29, "learning_rate": 0.00039314183340714197, "loss": 0.8815, "step": 106670 }, { "epoch": 4.29, "learning_rate": 0.00039313176858971776, "loss": 1.0346, "step": 106680 }, { "epoch": 4.29, "learning_rate": 0.00039312170377229355, "loss": 1.1707, "step": 106690 }, { "epoch": 4.29, "learning_rate": 0.00039311163895486934, "loss": 0.8788, "step": 106700 }, { "epoch": 4.29, "learning_rate": 0.00039310157413744514, "loss": 0.8168, "step": 106710 }, { "epoch": 4.29, "learning_rate": 0.000393091509320021, "loss": 0.8724, "step": 106720 }, { "epoch": 4.29, "learning_rate": 0.00039308144450259677, "loss": 1.1474, "step": 106730 }, { "epoch": 4.29, "learning_rate": 0.00039307137968517256, "loss": 1.0984, "step": 106740 }, { "epoch": 4.29, "learning_rate": 0.0003930613148677483, "loss": 1.0807, "step": 106750 }, { "epoch": 4.29, "learning_rate": 0.0003930512500503241, "loss": 1.0103, "step": 106760 }, { "epoch": 4.29, "learning_rate": 0.0003930411852328999, "loss": 0.7917, "step": 106770 }, { "epoch": 4.29, "learning_rate": 0.00039303112041547567, "loss": 1.1091, "step": 106780 }, { "epoch": 4.29, "learning_rate": 0.00039302105559805146, "loss": 0.8878, "step": 106790 }, { "epoch": 4.3, "learning_rate": 0.00039301099078062725, "loss": 0.6509, "step": 106800 }, { "epoch": 4.3, "learning_rate": 0.00039300092596320304, "loss": 0.8622, "step": 106810 }, { "epoch": 4.3, "learning_rate": 0.00039299086114577883, "loss": 1.1994, "step": 106820 }, { "epoch": 4.3, "learning_rate": 0.00039298079632835457, "loss": 0.8608, "step": 106830 }, { "epoch": 4.3, "learning_rate": 0.00039297073151093036, "loss": 1.0058, "step": 106840 }, { "epoch": 4.3, "learning_rate": 0.00039296066669350615, "loss": 1.114, "step": 106850 }, { "epoch": 4.3, "learning_rate": 0.000392950601876082, "loss": 1.0894, "step": 106860 }, { "epoch": 4.3, "learning_rate": 0.0003929405370586578, "loss": 1.0208, "step": 106870 }, { "epoch": 4.3, "learning_rate": 0.0003929304722412336, "loss": 0.9407, "step": 106880 }, { "epoch": 4.3, "learning_rate": 0.00039292040742380937, "loss": 0.874, "step": 106890 }, { "epoch": 4.3, "learning_rate": 0.00039291034260638516, "loss": 1.0623, "step": 106900 }, { "epoch": 4.3, "learning_rate": 0.00039290027778896095, "loss": 0.9603, "step": 106910 }, { "epoch": 4.3, "learning_rate": 0.0003928902129715367, "loss": 1.0529, "step": 106920 }, { "epoch": 4.3, "learning_rate": 0.0003928801481541125, "loss": 1.2013, "step": 106930 }, { "epoch": 4.3, "learning_rate": 0.00039287008333668827, "loss": 1.0716, "step": 106940 }, { "epoch": 4.3, "learning_rate": 0.00039286001851926406, "loss": 0.951, "step": 106950 }, { "epoch": 4.3, "learning_rate": 0.00039284995370183985, "loss": 0.9603, "step": 106960 }, { "epoch": 4.3, "learning_rate": 0.00039283988888441564, "loss": 1.1903, "step": 106970 }, { "epoch": 4.3, "learning_rate": 0.00039282982406699143, "loss": 1.0301, "step": 106980 }, { "epoch": 4.3, "learning_rate": 0.0003928197592495672, "loss": 1.0433, "step": 106990 }, { "epoch": 4.3, "learning_rate": 0.00039280969443214296, "loss": 0.7212, "step": 107000 }, { "epoch": 4.3, "learning_rate": 0.0003927996296147188, "loss": 1.0097, "step": 107010 }, { "epoch": 4.3, "learning_rate": 0.0003927895647972946, "loss": 0.778, "step": 107020 }, { "epoch": 4.3, "learning_rate": 0.0003927794999798704, "loss": 0.8281, "step": 107030 }, { "epoch": 4.3, "learning_rate": 0.0003927694351624462, "loss": 0.9058, "step": 107040 }, { "epoch": 4.31, "learning_rate": 0.00039275937034502197, "loss": 1.0035, "step": 107050 }, { "epoch": 4.31, "learning_rate": 0.00039274930552759776, "loss": 1.0262, "step": 107060 }, { "epoch": 4.31, "learning_rate": 0.00039273924071017355, "loss": 0.9788, "step": 107070 }, { "epoch": 4.31, "learning_rate": 0.0003927291758927493, "loss": 0.8426, "step": 107080 }, { "epoch": 4.31, "learning_rate": 0.0003927191110753251, "loss": 1.1035, "step": 107090 }, { "epoch": 4.31, "learning_rate": 0.00039270904625790087, "loss": 0.8486, "step": 107100 }, { "epoch": 4.31, "learning_rate": 0.00039269898144047666, "loss": 0.9214, "step": 107110 }, { "epoch": 4.31, "learning_rate": 0.00039268891662305245, "loss": 0.9741, "step": 107120 }, { "epoch": 4.31, "learning_rate": 0.00039267885180562824, "loss": 0.7671, "step": 107130 }, { "epoch": 4.31, "learning_rate": 0.00039266878698820403, "loss": 0.8269, "step": 107140 }, { "epoch": 4.31, "learning_rate": 0.0003926587221707798, "loss": 1.0316, "step": 107150 }, { "epoch": 4.31, "learning_rate": 0.00039264865735335567, "loss": 1.1259, "step": 107160 }, { "epoch": 4.31, "learning_rate": 0.0003926385925359314, "loss": 0.8753, "step": 107170 }, { "epoch": 4.31, "learning_rate": 0.0003926285277185072, "loss": 1.032, "step": 107180 }, { "epoch": 4.31, "learning_rate": 0.000392618462901083, "loss": 1.0665, "step": 107190 }, { "epoch": 4.31, "learning_rate": 0.0003926083980836588, "loss": 0.8707, "step": 107200 }, { "epoch": 4.31, "learning_rate": 0.00039259833326623457, "loss": 0.7662, "step": 107210 }, { "epoch": 4.31, "learning_rate": 0.00039258826844881036, "loss": 0.8952, "step": 107220 }, { "epoch": 4.31, "learning_rate": 0.00039257820363138615, "loss": 1.021, "step": 107230 }, { "epoch": 4.31, "learning_rate": 0.00039256813881396194, "loss": 0.868, "step": 107240 }, { "epoch": 4.31, "learning_rate": 0.0003925580739965377, "loss": 1.0655, "step": 107250 }, { "epoch": 4.31, "learning_rate": 0.00039254800917911347, "loss": 1.082, "step": 107260 }, { "epoch": 4.31, "learning_rate": 0.00039253794436168926, "loss": 1.0405, "step": 107270 }, { "epoch": 4.31, "learning_rate": 0.00039252787954426505, "loss": 1.0042, "step": 107280 }, { "epoch": 4.32, "learning_rate": 0.00039251781472684084, "loss": 1.0702, "step": 107290 }, { "epoch": 4.32, "learning_rate": 0.0003925077499094167, "loss": 0.7609, "step": 107300 }, { "epoch": 4.32, "learning_rate": 0.0003924976850919925, "loss": 1.103, "step": 107310 }, { "epoch": 4.32, "learning_rate": 0.00039248762027456826, "loss": 0.9951, "step": 107320 }, { "epoch": 4.32, "learning_rate": 0.00039247755545714406, "loss": 1.098, "step": 107330 }, { "epoch": 4.32, "learning_rate": 0.0003924674906397198, "loss": 1.0899, "step": 107340 }, { "epoch": 4.32, "learning_rate": 0.0003924574258222956, "loss": 0.9309, "step": 107350 }, { "epoch": 4.32, "learning_rate": 0.0003924473610048714, "loss": 0.8686, "step": 107360 }, { "epoch": 4.32, "learning_rate": 0.00039243729618744716, "loss": 1.0381, "step": 107370 }, { "epoch": 4.32, "learning_rate": 0.00039242723137002296, "loss": 1.2168, "step": 107380 }, { "epoch": 4.32, "learning_rate": 0.00039241716655259875, "loss": 1.0662, "step": 107390 }, { "epoch": 4.32, "learning_rate": 0.00039240710173517454, "loss": 1.0492, "step": 107400 }, { "epoch": 4.32, "learning_rate": 0.00039239703691775033, "loss": 1.1304, "step": 107410 }, { "epoch": 4.32, "learning_rate": 0.00039238697210032606, "loss": 0.8756, "step": 107420 }, { "epoch": 4.32, "learning_rate": 0.00039237690728290186, "loss": 0.7789, "step": 107430 }, { "epoch": 4.32, "learning_rate": 0.00039236684246547765, "loss": 1.1072, "step": 107440 }, { "epoch": 4.32, "learning_rate": 0.0003923567776480535, "loss": 0.8153, "step": 107450 }, { "epoch": 4.32, "learning_rate": 0.0003923467128306293, "loss": 1.0684, "step": 107460 }, { "epoch": 4.32, "learning_rate": 0.00039233664801320507, "loss": 1.0022, "step": 107470 }, { "epoch": 4.32, "learning_rate": 0.00039232658319578086, "loss": 0.8869, "step": 107480 }, { "epoch": 4.32, "learning_rate": 0.00039231651837835665, "loss": 0.9894, "step": 107490 }, { "epoch": 4.32, "learning_rate": 0.00039230645356093244, "loss": 1.0347, "step": 107500 }, { "epoch": 4.32, "learning_rate": 0.0003922963887435082, "loss": 1.0202, "step": 107510 }, { "epoch": 4.32, "learning_rate": 0.00039228632392608397, "loss": 0.8654, "step": 107520 }, { "epoch": 4.32, "learning_rate": 0.00039227625910865976, "loss": 0.9701, "step": 107530 }, { "epoch": 4.33, "learning_rate": 0.00039226619429123555, "loss": 1.0058, "step": 107540 }, { "epoch": 4.33, "learning_rate": 0.00039225612947381134, "loss": 0.9581, "step": 107550 }, { "epoch": 4.33, "learning_rate": 0.00039224606465638714, "loss": 1.0543, "step": 107560 }, { "epoch": 4.33, "learning_rate": 0.0003922359998389629, "loss": 1.1655, "step": 107570 }, { "epoch": 4.33, "learning_rate": 0.0003922259350215387, "loss": 0.9212, "step": 107580 }, { "epoch": 4.33, "learning_rate": 0.0003922158702041145, "loss": 1.2437, "step": 107590 }, { "epoch": 4.33, "learning_rate": 0.0003922058053866903, "loss": 0.7471, "step": 107600 }, { "epoch": 4.33, "learning_rate": 0.0003921957405692661, "loss": 1.031, "step": 107610 }, { "epoch": 4.33, "learning_rate": 0.0003921856757518419, "loss": 0.8466, "step": 107620 }, { "epoch": 4.33, "learning_rate": 0.00039217561093441767, "loss": 0.942, "step": 107630 }, { "epoch": 4.33, "learning_rate": 0.00039216554611699346, "loss": 0.7389, "step": 107640 }, { "epoch": 4.33, "learning_rate": 0.00039215548129956925, "loss": 0.8112, "step": 107650 }, { "epoch": 4.33, "learning_rate": 0.00039214541648214504, "loss": 1.0544, "step": 107660 }, { "epoch": 4.33, "learning_rate": 0.00039213535166472083, "loss": 0.989, "step": 107670 }, { "epoch": 4.33, "learning_rate": 0.00039212528684729657, "loss": 0.813, "step": 107680 }, { "epoch": 4.33, "learning_rate": 0.00039211522202987236, "loss": 1.0054, "step": 107690 }, { "epoch": 4.33, "learning_rate": 0.00039210515721244815, "loss": 0.9596, "step": 107700 }, { "epoch": 4.33, "learning_rate": 0.00039209509239502394, "loss": 1.0843, "step": 107710 }, { "epoch": 4.33, "learning_rate": 0.00039208502757759973, "loss": 1.0972, "step": 107720 }, { "epoch": 4.33, "learning_rate": 0.0003920749627601755, "loss": 0.7787, "step": 107730 }, { "epoch": 4.33, "learning_rate": 0.00039206489794275137, "loss": 0.7603, "step": 107740 }, { "epoch": 4.33, "learning_rate": 0.00039205483312532716, "loss": 0.8811, "step": 107750 }, { "epoch": 4.33, "learning_rate": 0.0003920447683079029, "loss": 1.0613, "step": 107760 }, { "epoch": 4.33, "learning_rate": 0.0003920347034904787, "loss": 0.9276, "step": 107770 }, { "epoch": 4.33, "learning_rate": 0.0003920246386730545, "loss": 1.0698, "step": 107780 }, { "epoch": 4.34, "learning_rate": 0.00039201457385563027, "loss": 0.862, "step": 107790 }, { "epoch": 4.34, "learning_rate": 0.00039200450903820606, "loss": 1.0882, "step": 107800 }, { "epoch": 4.34, "learning_rate": 0.00039199444422078185, "loss": 0.9615, "step": 107810 }, { "epoch": 4.34, "learning_rate": 0.00039198437940335764, "loss": 1.0136, "step": 107820 }, { "epoch": 4.34, "learning_rate": 0.00039197431458593343, "loss": 0.8982, "step": 107830 }, { "epoch": 4.34, "learning_rate": 0.00039196424976850917, "loss": 1.0163, "step": 107840 }, { "epoch": 4.34, "learning_rate": 0.00039195418495108496, "loss": 1.1894, "step": 107850 }, { "epoch": 4.34, "learning_rate": 0.00039194412013366075, "loss": 1.0184, "step": 107860 }, { "epoch": 4.34, "learning_rate": 0.00039193405531623654, "loss": 1.1228, "step": 107870 }, { "epoch": 4.34, "learning_rate": 0.00039192399049881233, "loss": 1.1907, "step": 107880 }, { "epoch": 4.34, "learning_rate": 0.0003919139256813882, "loss": 0.8912, "step": 107890 }, { "epoch": 4.34, "learning_rate": 0.00039190386086396397, "loss": 1.183, "step": 107900 }, { "epoch": 4.34, "learning_rate": 0.00039189379604653976, "loss": 1.1623, "step": 107910 }, { "epoch": 4.34, "learning_rate": 0.00039188373122911555, "loss": 1.0542, "step": 107920 }, { "epoch": 4.34, "learning_rate": 0.0003918736664116913, "loss": 0.9495, "step": 107930 }, { "epoch": 4.34, "learning_rate": 0.0003918636015942671, "loss": 0.9583, "step": 107940 }, { "epoch": 4.34, "learning_rate": 0.00039185353677684287, "loss": 0.9451, "step": 107950 }, { "epoch": 4.34, "learning_rate": 0.00039184347195941866, "loss": 1.0599, "step": 107960 }, { "epoch": 4.34, "learning_rate": 0.00039183340714199445, "loss": 1.1333, "step": 107970 }, { "epoch": 4.34, "learning_rate": 0.00039182334232457024, "loss": 1.1074, "step": 107980 }, { "epoch": 4.34, "learning_rate": 0.00039181327750714603, "loss": 1.1096, "step": 107990 }, { "epoch": 4.34, "learning_rate": 0.0003918032126897218, "loss": 1.0978, "step": 108000 }, { "epoch": 4.34, "learning_rate": 0.00039179314787229756, "loss": 1.0266, "step": 108010 }, { "epoch": 4.34, "learning_rate": 0.00039178308305487335, "loss": 0.9342, "step": 108020 }, { "epoch": 4.34, "learning_rate": 0.0003917730182374492, "loss": 1.0298, "step": 108030 }, { "epoch": 4.35, "learning_rate": 0.000391762953420025, "loss": 0.994, "step": 108040 }, { "epoch": 4.35, "learning_rate": 0.0003917528886026008, "loss": 0.8858, "step": 108050 }, { "epoch": 4.35, "learning_rate": 0.00039174282378517657, "loss": 0.7724, "step": 108060 }, { "epoch": 4.35, "learning_rate": 0.00039173275896775236, "loss": 0.931, "step": 108070 }, { "epoch": 4.35, "learning_rate": 0.00039172269415032815, "loss": 1.0114, "step": 108080 }, { "epoch": 4.35, "learning_rate": 0.00039171262933290394, "loss": 1.0799, "step": 108090 }, { "epoch": 4.35, "learning_rate": 0.0003917025645154797, "loss": 0.8498, "step": 108100 }, { "epoch": 4.35, "learning_rate": 0.00039169249969805547, "loss": 0.9609, "step": 108110 }, { "epoch": 4.35, "learning_rate": 0.00039168243488063126, "loss": 1.1577, "step": 108120 }, { "epoch": 4.35, "learning_rate": 0.00039167237006320705, "loss": 0.8609, "step": 108130 }, { "epoch": 4.35, "learning_rate": 0.00039166230524578284, "loss": 0.9868, "step": 108140 }, { "epoch": 4.35, "learning_rate": 0.00039165224042835863, "loss": 1.1775, "step": 108150 }, { "epoch": 4.35, "learning_rate": 0.0003916421756109344, "loss": 1.0618, "step": 108160 }, { "epoch": 4.35, "learning_rate": 0.0003916321107935102, "loss": 0.9171, "step": 108170 }, { "epoch": 4.35, "learning_rate": 0.000391622045976086, "loss": 0.902, "step": 108180 }, { "epoch": 4.35, "learning_rate": 0.0003916119811586618, "loss": 0.8763, "step": 108190 }, { "epoch": 4.35, "learning_rate": 0.0003916019163412376, "loss": 0.8499, "step": 108200 }, { "epoch": 4.35, "learning_rate": 0.0003915918515238134, "loss": 0.8967, "step": 108210 }, { "epoch": 4.35, "learning_rate": 0.00039158178670638916, "loss": 1.0708, "step": 108220 }, { "epoch": 4.35, "learning_rate": 0.00039157172188896496, "loss": 0.9732, "step": 108230 }, { "epoch": 4.35, "learning_rate": 0.00039156165707154075, "loss": 0.8303, "step": 108240 }, { "epoch": 4.35, "learning_rate": 0.00039155159225411654, "loss": 1.1143, "step": 108250 }, { "epoch": 4.35, "learning_rate": 0.00039154152743669233, "loss": 1.0261, "step": 108260 }, { "epoch": 4.35, "learning_rate": 0.00039153146261926806, "loss": 1.0046, "step": 108270 }, { "epoch": 4.35, "learning_rate": 0.00039152139780184385, "loss": 0.8585, "step": 108280 }, { "epoch": 4.36, "learning_rate": 0.00039151133298441965, "loss": 0.9732, "step": 108290 }, { "epoch": 4.36, "learning_rate": 0.00039150126816699544, "loss": 0.9556, "step": 108300 }, { "epoch": 4.36, "learning_rate": 0.00039149120334957123, "loss": 1.0081, "step": 108310 }, { "epoch": 4.36, "learning_rate": 0.000391481138532147, "loss": 0.9543, "step": 108320 }, { "epoch": 4.36, "learning_rate": 0.00039147107371472286, "loss": 0.9585, "step": 108330 }, { "epoch": 4.36, "learning_rate": 0.00039146100889729865, "loss": 1.058, "step": 108340 }, { "epoch": 4.36, "learning_rate": 0.0003914509440798744, "loss": 0.9769, "step": 108350 }, { "epoch": 4.36, "learning_rate": 0.0003914408792624502, "loss": 1.0117, "step": 108360 }, { "epoch": 4.36, "learning_rate": 0.00039143081444502597, "loss": 1.0111, "step": 108370 }, { "epoch": 4.36, "learning_rate": 0.00039142074962760176, "loss": 1.0966, "step": 108380 }, { "epoch": 4.36, "learning_rate": 0.00039141068481017755, "loss": 1.1978, "step": 108390 }, { "epoch": 4.36, "learning_rate": 0.00039140061999275334, "loss": 1.0851, "step": 108400 }, { "epoch": 4.36, "learning_rate": 0.00039139055517532914, "loss": 0.9689, "step": 108410 }, { "epoch": 4.36, "learning_rate": 0.0003913804903579049, "loss": 0.8171, "step": 108420 }, { "epoch": 4.36, "learning_rate": 0.0003913704255404807, "loss": 0.9864, "step": 108430 }, { "epoch": 4.36, "learning_rate": 0.00039136036072305645, "loss": 0.9316, "step": 108440 }, { "epoch": 4.36, "learning_rate": 0.00039135029590563224, "loss": 1.0138, "step": 108450 }, { "epoch": 4.36, "learning_rate": 0.00039134023108820803, "loss": 0.8723, "step": 108460 }, { "epoch": 4.36, "learning_rate": 0.0003913301662707839, "loss": 1.1049, "step": 108470 }, { "epoch": 4.36, "learning_rate": 0.00039132010145335967, "loss": 0.978, "step": 108480 }, { "epoch": 4.36, "learning_rate": 0.00039131003663593546, "loss": 0.9737, "step": 108490 }, { "epoch": 4.36, "learning_rate": 0.00039129997181851125, "loss": 1.1245, "step": 108500 }, { "epoch": 4.36, "learning_rate": 0.00039128990700108704, "loss": 0.98, "step": 108510 }, { "epoch": 4.36, "learning_rate": 0.0003912798421836628, "loss": 1.0261, "step": 108520 }, { "epoch": 4.36, "learning_rate": 0.00039126977736623857, "loss": 0.901, "step": 108530 }, { "epoch": 4.37, "learning_rate": 0.00039125971254881436, "loss": 1.0871, "step": 108540 }, { "epoch": 4.37, "learning_rate": 0.00039124964773139015, "loss": 0.997, "step": 108550 }, { "epoch": 4.37, "learning_rate": 0.00039123958291396594, "loss": 1.1101, "step": 108560 }, { "epoch": 4.37, "learning_rate": 0.00039122951809654173, "loss": 1.2204, "step": 108570 }, { "epoch": 4.37, "learning_rate": 0.0003912194532791175, "loss": 0.873, "step": 108580 }, { "epoch": 4.37, "learning_rate": 0.0003912093884616933, "loss": 1.1933, "step": 108590 }, { "epoch": 4.37, "learning_rate": 0.00039119932364426905, "loss": 0.7856, "step": 108600 }, { "epoch": 4.37, "learning_rate": 0.00039118925882684484, "loss": 0.8218, "step": 108610 }, { "epoch": 4.37, "learning_rate": 0.0003911791940094207, "loss": 0.9159, "step": 108620 }, { "epoch": 4.37, "learning_rate": 0.0003911691291919965, "loss": 1.0089, "step": 108630 }, { "epoch": 4.37, "learning_rate": 0.00039115906437457227, "loss": 0.8743, "step": 108640 }, { "epoch": 4.37, "learning_rate": 0.00039114899955714806, "loss": 0.9509, "step": 108650 }, { "epoch": 4.37, "learning_rate": 0.00039113893473972385, "loss": 1.0067, "step": 108660 }, { "epoch": 4.37, "learning_rate": 0.00039112886992229964, "loss": 0.8713, "step": 108670 }, { "epoch": 4.37, "learning_rate": 0.00039111880510487543, "loss": 0.9128, "step": 108680 }, { "epoch": 4.37, "learning_rate": 0.00039110874028745117, "loss": 0.9266, "step": 108690 }, { "epoch": 4.37, "learning_rate": 0.00039109867547002696, "loss": 0.9968, "step": 108700 }, { "epoch": 4.37, "learning_rate": 0.00039108861065260275, "loss": 0.849, "step": 108710 }, { "epoch": 4.37, "learning_rate": 0.00039107854583517854, "loss": 0.9238, "step": 108720 }, { "epoch": 4.37, "learning_rate": 0.00039106848101775433, "loss": 0.9375, "step": 108730 }, { "epoch": 4.37, "learning_rate": 0.0003910584162003301, "loss": 1.1573, "step": 108740 }, { "epoch": 4.37, "learning_rate": 0.0003910483513829059, "loss": 1.0672, "step": 108750 }, { "epoch": 4.37, "learning_rate": 0.00039103828656548176, "loss": 0.8422, "step": 108760 }, { "epoch": 4.37, "learning_rate": 0.00039102822174805755, "loss": 1.0403, "step": 108770 }, { "epoch": 4.37, "learning_rate": 0.0003910181569306333, "loss": 1.1067, "step": 108780 }, { "epoch": 4.38, "learning_rate": 0.0003910080921132091, "loss": 0.9894, "step": 108790 }, { "epoch": 4.38, "learning_rate": 0.00039099802729578487, "loss": 0.796, "step": 108800 }, { "epoch": 4.38, "learning_rate": 0.00039098796247836066, "loss": 0.8884, "step": 108810 }, { "epoch": 4.38, "learning_rate": 0.00039097789766093645, "loss": 1.1459, "step": 108820 }, { "epoch": 4.38, "learning_rate": 0.00039096783284351224, "loss": 1.2764, "step": 108830 }, { "epoch": 4.38, "learning_rate": 0.00039095776802608803, "loss": 1.1099, "step": 108840 }, { "epoch": 4.38, "learning_rate": 0.0003909477032086638, "loss": 1.1219, "step": 108850 }, { "epoch": 4.38, "learning_rate": 0.00039093763839123956, "loss": 0.9459, "step": 108860 }, { "epoch": 4.38, "learning_rate": 0.00039092757357381535, "loss": 0.9791, "step": 108870 }, { "epoch": 4.38, "learning_rate": 0.00039091750875639114, "loss": 0.9248, "step": 108880 }, { "epoch": 4.38, "learning_rate": 0.00039090744393896693, "loss": 1.127, "step": 108890 }, { "epoch": 4.38, "learning_rate": 0.0003908973791215427, "loss": 1.0829, "step": 108900 }, { "epoch": 4.38, "learning_rate": 0.00039088731430411857, "loss": 0.9429, "step": 108910 }, { "epoch": 4.38, "learning_rate": 0.00039087724948669436, "loss": 1.1842, "step": 108920 }, { "epoch": 4.38, "learning_rate": 0.00039086718466927015, "loss": 0.9054, "step": 108930 }, { "epoch": 4.38, "learning_rate": 0.0003908571198518459, "loss": 1.1903, "step": 108940 }, { "epoch": 4.38, "learning_rate": 0.0003908470550344217, "loss": 1.0011, "step": 108950 }, { "epoch": 4.38, "learning_rate": 0.00039083699021699747, "loss": 0.7337, "step": 108960 }, { "epoch": 4.38, "learning_rate": 0.00039082692539957326, "loss": 0.9992, "step": 108970 }, { "epoch": 4.38, "learning_rate": 0.00039081686058214905, "loss": 1.044, "step": 108980 }, { "epoch": 4.38, "learning_rate": 0.00039080679576472484, "loss": 1.1369, "step": 108990 }, { "epoch": 4.38, "learning_rate": 0.00039079673094730063, "loss": 1.0901, "step": 109000 }, { "epoch": 4.38, "learning_rate": 0.0003907866661298764, "loss": 0.8553, "step": 109010 }, { "epoch": 4.38, "learning_rate": 0.0003907766013124522, "loss": 0.8973, "step": 109020 }, { "epoch": 4.39, "learning_rate": 0.00039076653649502795, "loss": 0.9193, "step": 109030 }, { "epoch": 4.39, "learning_rate": 0.00039075647167760374, "loss": 1.0587, "step": 109040 }, { "epoch": 4.39, "learning_rate": 0.00039074640686017953, "loss": 0.9377, "step": 109050 }, { "epoch": 4.39, "learning_rate": 0.0003907363420427554, "loss": 0.9258, "step": 109060 }, { "epoch": 4.39, "learning_rate": 0.00039072627722533116, "loss": 0.9519, "step": 109070 }, { "epoch": 4.39, "learning_rate": 0.00039071621240790695, "loss": 1.0828, "step": 109080 }, { "epoch": 4.39, "learning_rate": 0.00039070614759048275, "loss": 0.7853, "step": 109090 }, { "epoch": 4.39, "learning_rate": 0.00039069608277305854, "loss": 0.8668, "step": 109100 }, { "epoch": 4.39, "learning_rate": 0.0003906860179556343, "loss": 1.097, "step": 109110 }, { "epoch": 4.39, "learning_rate": 0.00039067595313821006, "loss": 1.0055, "step": 109120 }, { "epoch": 4.39, "learning_rate": 0.00039066588832078585, "loss": 0.9469, "step": 109130 }, { "epoch": 4.39, "learning_rate": 0.00039065582350336165, "loss": 1.0608, "step": 109140 }, { "epoch": 4.39, "learning_rate": 0.00039064575868593744, "loss": 0.9165, "step": 109150 }, { "epoch": 4.39, "learning_rate": 0.0003906356938685132, "loss": 1.1183, "step": 109160 }, { "epoch": 4.39, "learning_rate": 0.000390625629051089, "loss": 0.9984, "step": 109170 }, { "epoch": 4.39, "learning_rate": 0.0003906155642336648, "loss": 0.9025, "step": 109180 }, { "epoch": 4.39, "learning_rate": 0.0003906054994162406, "loss": 0.8128, "step": 109190 }, { "epoch": 4.39, "learning_rate": 0.0003905954345988164, "loss": 1.0845, "step": 109200 }, { "epoch": 4.39, "learning_rate": 0.0003905853697813922, "loss": 0.958, "step": 109210 }, { "epoch": 4.39, "learning_rate": 0.00039057530496396797, "loss": 1.1404, "step": 109220 }, { "epoch": 4.39, "learning_rate": 0.00039056524014654376, "loss": 1.0271, "step": 109230 }, { "epoch": 4.39, "learning_rate": 0.00039055517532911955, "loss": 1.0005, "step": 109240 }, { "epoch": 4.39, "learning_rate": 0.00039054511051169534, "loss": 1.2314, "step": 109250 }, { "epoch": 4.39, "learning_rate": 0.00039053504569427113, "loss": 0.9889, "step": 109260 }, { "epoch": 4.39, "learning_rate": 0.0003905249808768469, "loss": 1.3569, "step": 109270 }, { "epoch": 4.4, "learning_rate": 0.00039051491605942266, "loss": 1.1446, "step": 109280 }, { "epoch": 4.4, "learning_rate": 0.00039050485124199845, "loss": 1.1133, "step": 109290 }, { "epoch": 4.4, "learning_rate": 0.00039049478642457424, "loss": 1.1208, "step": 109300 }, { "epoch": 4.4, "learning_rate": 0.00039048472160715003, "loss": 0.9601, "step": 109310 }, { "epoch": 4.4, "learning_rate": 0.0003904746567897258, "loss": 0.8692, "step": 109320 }, { "epoch": 4.4, "learning_rate": 0.0003904645919723016, "loss": 1.2071, "step": 109330 }, { "epoch": 4.4, "learning_rate": 0.0003904545271548774, "loss": 1.0688, "step": 109340 }, { "epoch": 4.4, "learning_rate": 0.00039044446233745325, "loss": 1.028, "step": 109350 }, { "epoch": 4.4, "learning_rate": 0.00039043439752002904, "loss": 0.9558, "step": 109360 }, { "epoch": 4.4, "learning_rate": 0.0003904243327026048, "loss": 0.9481, "step": 109370 }, { "epoch": 4.4, "learning_rate": 0.00039041426788518057, "loss": 1.2613, "step": 109380 }, { "epoch": 4.4, "learning_rate": 0.00039040420306775636, "loss": 0.8931, "step": 109390 }, { "epoch": 4.4, "learning_rate": 0.00039039413825033215, "loss": 1.1077, "step": 109400 }, { "epoch": 4.4, "learning_rate": 0.00039038407343290794, "loss": 0.9326, "step": 109410 }, { "epoch": 4.4, "learning_rate": 0.00039037400861548373, "loss": 1.1111, "step": 109420 }, { "epoch": 4.4, "learning_rate": 0.0003903639437980595, "loss": 0.8971, "step": 109430 }, { "epoch": 4.4, "learning_rate": 0.0003903538789806353, "loss": 1.0169, "step": 109440 }, { "epoch": 4.4, "learning_rate": 0.00039034381416321105, "loss": 0.8352, "step": 109450 }, { "epoch": 4.4, "learning_rate": 0.00039033374934578684, "loss": 1.0451, "step": 109460 }, { "epoch": 4.4, "learning_rate": 0.00039032368452836263, "loss": 0.8657, "step": 109470 }, { "epoch": 4.4, "learning_rate": 0.0003903136197109384, "loss": 0.9653, "step": 109480 }, { "epoch": 4.4, "learning_rate": 0.0003903035548935142, "loss": 0.9854, "step": 109490 }, { "epoch": 4.4, "learning_rate": 0.00039029349007609006, "loss": 0.9564, "step": 109500 }, { "epoch": 4.4, "learning_rate": 0.00039028342525866585, "loss": 0.9921, "step": 109510 }, { "epoch": 4.4, "learning_rate": 0.00039027336044124164, "loss": 0.9912, "step": 109520 }, { "epoch": 4.41, "learning_rate": 0.00039026329562381743, "loss": 0.9483, "step": 109530 }, { "epoch": 4.41, "learning_rate": 0.00039025323080639317, "loss": 1.0209, "step": 109540 }, { "epoch": 4.41, "learning_rate": 0.00039024316598896896, "loss": 1.1527, "step": 109550 }, { "epoch": 4.41, "learning_rate": 0.00039023310117154475, "loss": 1.0429, "step": 109560 }, { "epoch": 4.41, "learning_rate": 0.00039022303635412054, "loss": 0.9331, "step": 109570 }, { "epoch": 4.41, "learning_rate": 0.00039021297153669633, "loss": 1.0694, "step": 109580 }, { "epoch": 4.41, "learning_rate": 0.0003902029067192721, "loss": 1.0257, "step": 109590 }, { "epoch": 4.41, "learning_rate": 0.0003901928419018479, "loss": 1.0098, "step": 109600 }, { "epoch": 4.41, "learning_rate": 0.0003901827770844237, "loss": 1.1098, "step": 109610 }, { "epoch": 4.41, "learning_rate": 0.00039017271226699944, "loss": 0.9107, "step": 109620 }, { "epoch": 4.41, "learning_rate": 0.00039016264744957523, "loss": 1.1455, "step": 109630 }, { "epoch": 4.41, "learning_rate": 0.0003901525826321511, "loss": 0.9841, "step": 109640 }, { "epoch": 4.41, "learning_rate": 0.00039014251781472687, "loss": 0.909, "step": 109650 }, { "epoch": 4.41, "learning_rate": 0.00039013245299730266, "loss": 0.9522, "step": 109660 }, { "epoch": 4.41, "learning_rate": 0.00039012238817987845, "loss": 0.9016, "step": 109670 }, { "epoch": 4.41, "learning_rate": 0.00039011232336245424, "loss": 1.0022, "step": 109680 }, { "epoch": 4.41, "learning_rate": 0.00039010225854503003, "loss": 1.221, "step": 109690 }, { "epoch": 4.41, "learning_rate": 0.00039009219372760577, "loss": 1.0067, "step": 109700 }, { "epoch": 4.41, "learning_rate": 0.00039008212891018156, "loss": 0.9874, "step": 109710 }, { "epoch": 4.41, "learning_rate": 0.00039007206409275735, "loss": 0.8972, "step": 109720 }, { "epoch": 4.41, "learning_rate": 0.00039006199927533314, "loss": 0.9329, "step": 109730 }, { "epoch": 4.41, "learning_rate": 0.00039005193445790893, "loss": 0.9249, "step": 109740 }, { "epoch": 4.41, "learning_rate": 0.0003900418696404847, "loss": 0.9822, "step": 109750 }, { "epoch": 4.41, "learning_rate": 0.0003900318048230605, "loss": 1.1455, "step": 109760 }, { "epoch": 4.41, "learning_rate": 0.0003900217400056363, "loss": 1.1475, "step": 109770 }, { "epoch": 4.42, "learning_rate": 0.0003900116751882121, "loss": 1.2385, "step": 109780 }, { "epoch": 4.42, "learning_rate": 0.0003900016103707879, "loss": 0.7315, "step": 109790 }, { "epoch": 4.42, "learning_rate": 0.0003899915455533637, "loss": 1.0851, "step": 109800 }, { "epoch": 4.42, "learning_rate": 0.00038998148073593947, "loss": 1.1242, "step": 109810 }, { "epoch": 4.42, "learning_rate": 0.00038997141591851526, "loss": 0.8169, "step": 109820 }, { "epoch": 4.42, "learning_rate": 0.00038996135110109105, "loss": 1.0817, "step": 109830 }, { "epoch": 4.42, "learning_rate": 0.00038995128628366684, "loss": 1.0316, "step": 109840 }, { "epoch": 4.42, "learning_rate": 0.00038994122146624263, "loss": 0.983, "step": 109850 }, { "epoch": 4.42, "learning_rate": 0.0003899311566488184, "loss": 1.2439, "step": 109860 }, { "epoch": 4.42, "learning_rate": 0.00038992109183139416, "loss": 1.024, "step": 109870 }, { "epoch": 4.42, "learning_rate": 0.00038991102701396995, "loss": 0.8765, "step": 109880 }, { "epoch": 4.42, "learning_rate": 0.00038990096219654574, "loss": 0.8521, "step": 109890 }, { "epoch": 4.42, "learning_rate": 0.00038989089737912153, "loss": 1.0925, "step": 109900 }, { "epoch": 4.42, "learning_rate": 0.0003898808325616973, "loss": 1.0787, "step": 109910 }, { "epoch": 4.42, "learning_rate": 0.0003898707677442731, "loss": 1.1053, "step": 109920 }, { "epoch": 4.42, "learning_rate": 0.00038986070292684895, "loss": 1.0969, "step": 109930 }, { "epoch": 4.42, "learning_rate": 0.00038985063810942475, "loss": 0.8943, "step": 109940 }, { "epoch": 4.42, "learning_rate": 0.00038984057329200054, "loss": 0.9051, "step": 109950 }, { "epoch": 4.42, "learning_rate": 0.00038983050847457627, "loss": 0.8038, "step": 109960 }, { "epoch": 4.42, "learning_rate": 0.00038982044365715206, "loss": 1.1136, "step": 109970 }, { "epoch": 4.42, "learning_rate": 0.00038981037883972785, "loss": 0.9755, "step": 109980 }, { "epoch": 4.42, "learning_rate": 0.00038980031402230365, "loss": 0.9067, "step": 109990 }, { "epoch": 4.42, "learning_rate": 0.00038979024920487944, "loss": 1.1277, "step": 110000 }, { "epoch": 4.42, "learning_rate": 0.0003897801843874552, "loss": 0.8162, "step": 110010 }, { "epoch": 4.42, "learning_rate": 0.000389770119570031, "loss": 0.8019, "step": 110020 }, { "epoch": 4.43, "learning_rate": 0.0003897600547526068, "loss": 1.1543, "step": 110030 }, { "epoch": 4.43, "learning_rate": 0.00038974998993518254, "loss": 1.4445, "step": 110040 }, { "epoch": 4.43, "learning_rate": 0.00038973992511775834, "loss": 0.9677, "step": 110050 }, { "epoch": 4.43, "learning_rate": 0.0003897298603003341, "loss": 0.9753, "step": 110060 }, { "epoch": 4.43, "learning_rate": 0.0003897197954829099, "loss": 1.0393, "step": 110070 }, { "epoch": 4.43, "learning_rate": 0.00038970973066548576, "loss": 0.9799, "step": 110080 }, { "epoch": 4.43, "learning_rate": 0.00038969966584806155, "loss": 1.1731, "step": 110090 }, { "epoch": 4.43, "learning_rate": 0.00038968960103063734, "loss": 1.0405, "step": 110100 }, { "epoch": 4.43, "learning_rate": 0.00038967953621321313, "loss": 0.9974, "step": 110110 }, { "epoch": 4.43, "learning_rate": 0.0003896694713957889, "loss": 1.1468, "step": 110120 }, { "epoch": 4.43, "learning_rate": 0.00038965940657836466, "loss": 0.9889, "step": 110130 }, { "epoch": 4.43, "learning_rate": 0.00038964934176094045, "loss": 0.9882, "step": 110140 }, { "epoch": 4.43, "learning_rate": 0.00038963927694351624, "loss": 0.8345, "step": 110150 }, { "epoch": 4.43, "learning_rate": 0.00038962921212609203, "loss": 1.0183, "step": 110160 }, { "epoch": 4.43, "learning_rate": 0.0003896191473086678, "loss": 0.9322, "step": 110170 }, { "epoch": 4.43, "learning_rate": 0.0003896090824912436, "loss": 0.9227, "step": 110180 }, { "epoch": 4.43, "learning_rate": 0.0003895990176738194, "loss": 1.1428, "step": 110190 }, { "epoch": 4.43, "learning_rate": 0.0003895889528563952, "loss": 0.9292, "step": 110200 }, { "epoch": 4.43, "learning_rate": 0.00038957888803897093, "loss": 0.8193, "step": 110210 }, { "epoch": 4.43, "learning_rate": 0.0003895688232215467, "loss": 1.0233, "step": 110220 }, { "epoch": 4.43, "learning_rate": 0.00038955875840412257, "loss": 1.1475, "step": 110230 }, { "epoch": 4.43, "learning_rate": 0.00038954869358669836, "loss": 0.8881, "step": 110240 }, { "epoch": 4.43, "learning_rate": 0.00038953862876927415, "loss": 0.9497, "step": 110250 }, { "epoch": 4.43, "learning_rate": 0.00038952856395184994, "loss": 0.9023, "step": 110260 }, { "epoch": 4.43, "learning_rate": 0.00038951849913442573, "loss": 1.1463, "step": 110270 }, { "epoch": 4.44, "learning_rate": 0.0003895084343170015, "loss": 0.9366, "step": 110280 }, { "epoch": 4.44, "learning_rate": 0.0003894983694995773, "loss": 1.2943, "step": 110290 }, { "epoch": 4.44, "learning_rate": 0.00038948830468215305, "loss": 0.9572, "step": 110300 }, { "epoch": 4.44, "learning_rate": 0.00038947823986472884, "loss": 0.9562, "step": 110310 }, { "epoch": 4.44, "learning_rate": 0.00038946817504730463, "loss": 0.8035, "step": 110320 }, { "epoch": 4.44, "learning_rate": 0.0003894581102298804, "loss": 1.3213, "step": 110330 }, { "epoch": 4.44, "learning_rate": 0.0003894480454124562, "loss": 1.0057, "step": 110340 }, { "epoch": 4.44, "learning_rate": 0.000389437980595032, "loss": 1.2364, "step": 110350 }, { "epoch": 4.44, "learning_rate": 0.0003894279157776078, "loss": 1.1513, "step": 110360 }, { "epoch": 4.44, "learning_rate": 0.00038941785096018364, "loss": 0.9822, "step": 110370 }, { "epoch": 4.44, "learning_rate": 0.0003894077861427594, "loss": 1.0186, "step": 110380 }, { "epoch": 4.44, "learning_rate": 0.00038939772132533517, "loss": 0.9429, "step": 110390 }, { "epoch": 4.44, "learning_rate": 0.00038938765650791096, "loss": 0.8592, "step": 110400 }, { "epoch": 4.44, "learning_rate": 0.00038937759169048675, "loss": 1.0402, "step": 110410 }, { "epoch": 4.44, "learning_rate": 0.00038936752687306254, "loss": 0.9387, "step": 110420 }, { "epoch": 4.44, "learning_rate": 0.00038935746205563833, "loss": 1.0128, "step": 110430 }, { "epoch": 4.44, "learning_rate": 0.0003893473972382141, "loss": 1.013, "step": 110440 }, { "epoch": 4.44, "learning_rate": 0.0003893373324207899, "loss": 1.0117, "step": 110450 }, { "epoch": 4.44, "learning_rate": 0.00038932726760336565, "loss": 1.0292, "step": 110460 }, { "epoch": 4.44, "learning_rate": 0.00038931720278594144, "loss": 0.9073, "step": 110470 }, { "epoch": 4.44, "learning_rate": 0.00038930713796851723, "loss": 1.1947, "step": 110480 }, { "epoch": 4.44, "learning_rate": 0.000389297073151093, "loss": 1.0263, "step": 110490 }, { "epoch": 4.44, "learning_rate": 0.0003892870083336688, "loss": 1.201, "step": 110500 }, { "epoch": 4.44, "learning_rate": 0.0003892769435162446, "loss": 1.0809, "step": 110510 }, { "epoch": 4.44, "learning_rate": 0.00038926687869882045, "loss": 1.1238, "step": 110520 }, { "epoch": 4.45, "learning_rate": 0.00038925681388139624, "loss": 0.9936, "step": 110530 }, { "epoch": 4.45, "learning_rate": 0.00038924674906397203, "loss": 1.0064, "step": 110540 }, { "epoch": 4.45, "learning_rate": 0.00038923668424654777, "loss": 1.41, "step": 110550 }, { "epoch": 4.45, "learning_rate": 0.00038922661942912356, "loss": 0.9605, "step": 110560 }, { "epoch": 4.45, "learning_rate": 0.00038921655461169935, "loss": 1.0703, "step": 110570 }, { "epoch": 4.45, "learning_rate": 0.00038920648979427514, "loss": 1.2211, "step": 110580 }, { "epoch": 4.45, "learning_rate": 0.00038919642497685093, "loss": 0.9739, "step": 110590 }, { "epoch": 4.45, "learning_rate": 0.0003891863601594267, "loss": 0.8296, "step": 110600 }, { "epoch": 4.45, "learning_rate": 0.0003891762953420025, "loss": 1.0541, "step": 110610 }, { "epoch": 4.45, "learning_rate": 0.0003891662305245783, "loss": 1.0649, "step": 110620 }, { "epoch": 4.45, "learning_rate": 0.00038915616570715404, "loss": 1.1566, "step": 110630 }, { "epoch": 4.45, "learning_rate": 0.00038914610088972983, "loss": 0.8696, "step": 110640 }, { "epoch": 4.45, "learning_rate": 0.0003891360360723056, "loss": 0.9759, "step": 110650 }, { "epoch": 4.45, "learning_rate": 0.0003891259712548814, "loss": 1.0509, "step": 110660 }, { "epoch": 4.45, "learning_rate": 0.00038911590643745726, "loss": 1.1671, "step": 110670 }, { "epoch": 4.45, "learning_rate": 0.00038910584162003305, "loss": 0.9003, "step": 110680 }, { "epoch": 4.45, "learning_rate": 0.00038909577680260884, "loss": 1.2194, "step": 110690 }, { "epoch": 4.45, "learning_rate": 0.00038908571198518463, "loss": 0.9321, "step": 110700 }, { "epoch": 4.45, "learning_rate": 0.0003890756471677604, "loss": 1.0574, "step": 110710 }, { "epoch": 4.45, "learning_rate": 0.00038906558235033616, "loss": 0.8544, "step": 110720 }, { "epoch": 4.45, "learning_rate": 0.00038905551753291195, "loss": 1.2527, "step": 110730 }, { "epoch": 4.45, "learning_rate": 0.00038904545271548774, "loss": 1.035, "step": 110740 }, { "epoch": 4.45, "learning_rate": 0.00038903538789806353, "loss": 0.8273, "step": 110750 }, { "epoch": 4.45, "learning_rate": 0.0003890253230806393, "loss": 1.002, "step": 110760 }, { "epoch": 4.45, "learning_rate": 0.0003890152582632151, "loss": 1.1376, "step": 110770 }, { "epoch": 4.46, "learning_rate": 0.0003890051934457909, "loss": 0.9219, "step": 110780 }, { "epoch": 4.46, "learning_rate": 0.0003889951286283667, "loss": 1.0778, "step": 110790 }, { "epoch": 4.46, "learning_rate": 0.00038898506381094243, "loss": 1.0815, "step": 110800 }, { "epoch": 4.46, "learning_rate": 0.00038897499899351827, "loss": 1.0316, "step": 110810 }, { "epoch": 4.46, "learning_rate": 0.00038896493417609406, "loss": 0.877, "step": 110820 }, { "epoch": 4.46, "learning_rate": 0.00038895486935866985, "loss": 0.8679, "step": 110830 }, { "epoch": 4.46, "learning_rate": 0.00038894480454124564, "loss": 0.9075, "step": 110840 }, { "epoch": 4.46, "learning_rate": 0.00038893473972382144, "loss": 1.2192, "step": 110850 }, { "epoch": 4.46, "learning_rate": 0.0003889246749063972, "loss": 1.0482, "step": 110860 }, { "epoch": 4.46, "learning_rate": 0.000388914610088973, "loss": 1.3808, "step": 110870 }, { "epoch": 4.46, "learning_rate": 0.0003889045452715488, "loss": 0.859, "step": 110880 }, { "epoch": 4.46, "learning_rate": 0.00038889448045412454, "loss": 1.009, "step": 110890 }, { "epoch": 4.46, "learning_rate": 0.00038888441563670034, "loss": 0.9338, "step": 110900 }, { "epoch": 4.46, "learning_rate": 0.0003888743508192761, "loss": 1.1655, "step": 110910 }, { "epoch": 4.46, "learning_rate": 0.0003888642860018519, "loss": 0.8113, "step": 110920 }, { "epoch": 4.46, "learning_rate": 0.0003888542211844277, "loss": 1.1633, "step": 110930 }, { "epoch": 4.46, "learning_rate": 0.0003888441563670035, "loss": 1.196, "step": 110940 }, { "epoch": 4.46, "learning_rate": 0.0003888340915495793, "loss": 0.9461, "step": 110950 }, { "epoch": 4.46, "learning_rate": 0.00038882402673215513, "loss": 0.9034, "step": 110960 }, { "epoch": 4.46, "learning_rate": 0.00038881396191473087, "loss": 1.1916, "step": 110970 }, { "epoch": 4.46, "learning_rate": 0.00038880389709730666, "loss": 1.1274, "step": 110980 }, { "epoch": 4.46, "learning_rate": 0.00038879383227988245, "loss": 0.8569, "step": 110990 }, { "epoch": 4.46, "learning_rate": 0.00038878376746245824, "loss": 1.0563, "step": 111000 }, { "epoch": 4.46, "learning_rate": 0.00038877370264503403, "loss": 0.9204, "step": 111010 }, { "epoch": 4.47, "learning_rate": 0.0003887636378276098, "loss": 0.9199, "step": 111020 }, { "epoch": 4.47, "learning_rate": 0.0003887535730101856, "loss": 1.0206, "step": 111030 }, { "epoch": 4.47, "learning_rate": 0.0003887435081927614, "loss": 1.0682, "step": 111040 }, { "epoch": 4.47, "learning_rate": 0.0003887334433753372, "loss": 0.8119, "step": 111050 }, { "epoch": 4.47, "learning_rate": 0.00038872337855791293, "loss": 1.0204, "step": 111060 }, { "epoch": 4.47, "learning_rate": 0.0003887133137404887, "loss": 1.0314, "step": 111070 }, { "epoch": 4.47, "learning_rate": 0.0003887032489230645, "loss": 0.9319, "step": 111080 }, { "epoch": 4.47, "learning_rate": 0.0003886931841056403, "loss": 0.9066, "step": 111090 }, { "epoch": 4.47, "learning_rate": 0.00038868311928821615, "loss": 1.0633, "step": 111100 }, { "epoch": 4.47, "learning_rate": 0.00038867305447079194, "loss": 0.89, "step": 111110 }, { "epoch": 4.47, "learning_rate": 0.00038866298965336773, "loss": 1.0034, "step": 111120 }, { "epoch": 4.47, "learning_rate": 0.0003886529248359435, "loss": 1.1816, "step": 111130 }, { "epoch": 4.47, "learning_rate": 0.00038864286001851926, "loss": 1.0406, "step": 111140 }, { "epoch": 4.47, "learning_rate": 0.00038863279520109505, "loss": 1.0203, "step": 111150 }, { "epoch": 4.47, "learning_rate": 0.00038862273038367084, "loss": 0.9331, "step": 111160 }, { "epoch": 4.47, "learning_rate": 0.00038861266556624663, "loss": 1.0129, "step": 111170 }, { "epoch": 4.47, "learning_rate": 0.0003886026007488224, "loss": 1.0748, "step": 111180 }, { "epoch": 4.47, "learning_rate": 0.0003885925359313982, "loss": 0.8884, "step": 111190 }, { "epoch": 4.47, "learning_rate": 0.000388582471113974, "loss": 0.9649, "step": 111200 }, { "epoch": 4.47, "learning_rate": 0.0003885724062965498, "loss": 0.9931, "step": 111210 }, { "epoch": 4.47, "learning_rate": 0.00038856234147912553, "loss": 0.9562, "step": 111220 }, { "epoch": 4.47, "learning_rate": 0.0003885522766617013, "loss": 1.2719, "step": 111230 }, { "epoch": 4.47, "learning_rate": 0.0003885422118442771, "loss": 1.0807, "step": 111240 }, { "epoch": 4.47, "learning_rate": 0.00038853214702685296, "loss": 1.1583, "step": 111250 }, { "epoch": 4.47, "learning_rate": 0.00038852208220942875, "loss": 1.1086, "step": 111260 }, { "epoch": 4.48, "learning_rate": 0.00038851201739200454, "loss": 1.0244, "step": 111270 }, { "epoch": 4.48, "learning_rate": 0.00038850195257458033, "loss": 1.1375, "step": 111280 }, { "epoch": 4.48, "learning_rate": 0.0003884918877571561, "loss": 1.139, "step": 111290 }, { "epoch": 4.48, "learning_rate": 0.0003884818229397319, "loss": 0.8515, "step": 111300 }, { "epoch": 4.48, "learning_rate": 0.00038847175812230765, "loss": 1.248, "step": 111310 }, { "epoch": 4.48, "learning_rate": 0.00038846169330488344, "loss": 0.8778, "step": 111320 }, { "epoch": 4.48, "learning_rate": 0.00038845162848745923, "loss": 1.0795, "step": 111330 }, { "epoch": 4.48, "learning_rate": 0.000388441563670035, "loss": 0.9891, "step": 111340 }, { "epoch": 4.48, "learning_rate": 0.0003884314988526108, "loss": 1.1927, "step": 111350 }, { "epoch": 4.48, "learning_rate": 0.0003884214340351866, "loss": 1.1341, "step": 111360 }, { "epoch": 4.48, "learning_rate": 0.0003884113692177624, "loss": 0.8649, "step": 111370 }, { "epoch": 4.48, "learning_rate": 0.0003884013044003382, "loss": 0.9563, "step": 111380 }, { "epoch": 4.48, "learning_rate": 0.0003883912395829139, "loss": 0.8394, "step": 111390 }, { "epoch": 4.48, "learning_rate": 0.00038838117476548977, "loss": 0.8524, "step": 111400 }, { "epoch": 4.48, "learning_rate": 0.00038837110994806556, "loss": 1.0339, "step": 111410 }, { "epoch": 4.48, "learning_rate": 0.00038836104513064135, "loss": 1.1186, "step": 111420 }, { "epoch": 4.48, "learning_rate": 0.00038835098031321714, "loss": 1.0924, "step": 111430 }, { "epoch": 4.48, "learning_rate": 0.00038834091549579293, "loss": 1.0419, "step": 111440 }, { "epoch": 4.48, "learning_rate": 0.0003883308506783687, "loss": 0.8532, "step": 111450 }, { "epoch": 4.48, "learning_rate": 0.0003883207858609445, "loss": 1.0384, "step": 111460 }, { "epoch": 4.48, "learning_rate": 0.0003883107210435203, "loss": 0.8994, "step": 111470 }, { "epoch": 4.48, "learning_rate": 0.00038830065622609604, "loss": 1.1645, "step": 111480 }, { "epoch": 4.48, "learning_rate": 0.00038829059140867183, "loss": 1.0144, "step": 111490 }, { "epoch": 4.48, "learning_rate": 0.0003882805265912476, "loss": 1.0114, "step": 111500 }, { "epoch": 4.48, "learning_rate": 0.0003882704617738234, "loss": 0.8619, "step": 111510 }, { "epoch": 4.49, "learning_rate": 0.0003882603969563992, "loss": 0.8736, "step": 111520 }, { "epoch": 4.49, "learning_rate": 0.000388250332138975, "loss": 1.3077, "step": 111530 }, { "epoch": 4.49, "learning_rate": 0.00038824026732155084, "loss": 0.9139, "step": 111540 }, { "epoch": 4.49, "learning_rate": 0.00038823020250412663, "loss": 0.9027, "step": 111550 }, { "epoch": 4.49, "learning_rate": 0.00038822013768670236, "loss": 1.0723, "step": 111560 }, { "epoch": 4.49, "learning_rate": 0.00038821007286927816, "loss": 1.077, "step": 111570 }, { "epoch": 4.49, "learning_rate": 0.00038820000805185395, "loss": 1.0286, "step": 111580 }, { "epoch": 4.49, "learning_rate": 0.00038818994323442974, "loss": 1.1989, "step": 111590 }, { "epoch": 4.49, "learning_rate": 0.00038817987841700553, "loss": 1.0265, "step": 111600 }, { "epoch": 4.49, "learning_rate": 0.0003881698135995813, "loss": 1.0644, "step": 111610 }, { "epoch": 4.49, "learning_rate": 0.0003881597487821571, "loss": 1.2367, "step": 111620 }, { "epoch": 4.49, "learning_rate": 0.0003881496839647329, "loss": 1.0173, "step": 111630 }, { "epoch": 4.49, "learning_rate": 0.0003881396191473087, "loss": 1.3249, "step": 111640 }, { "epoch": 4.49, "learning_rate": 0.00038812955432988443, "loss": 1.0644, "step": 111650 }, { "epoch": 4.49, "learning_rate": 0.0003881194895124602, "loss": 1.0079, "step": 111660 }, { "epoch": 4.49, "learning_rate": 0.000388109424695036, "loss": 0.8407, "step": 111670 }, { "epoch": 4.49, "learning_rate": 0.0003880993598776118, "loss": 0.9599, "step": 111680 }, { "epoch": 4.49, "learning_rate": 0.00038808929506018764, "loss": 1.0354, "step": 111690 }, { "epoch": 4.49, "learning_rate": 0.00038807923024276344, "loss": 1.0121, "step": 111700 }, { "epoch": 4.49, "learning_rate": 0.0003880691654253392, "loss": 0.9805, "step": 111710 }, { "epoch": 4.49, "learning_rate": 0.000388059100607915, "loss": 0.982, "step": 111720 }, { "epoch": 4.49, "learning_rate": 0.00038804903579049075, "loss": 1.0537, "step": 111730 }, { "epoch": 4.49, "learning_rate": 0.00038803897097306654, "loss": 0.9152, "step": 111740 }, { "epoch": 4.49, "learning_rate": 0.00038802890615564234, "loss": 1.0761, "step": 111750 }, { "epoch": 4.49, "learning_rate": 0.0003880188413382181, "loss": 1.031, "step": 111760 }, { "epoch": 4.5, "learning_rate": 0.0003880087765207939, "loss": 1.2367, "step": 111770 }, { "epoch": 4.5, "learning_rate": 0.0003879987117033697, "loss": 1.0737, "step": 111780 }, { "epoch": 4.5, "learning_rate": 0.0003879886468859455, "loss": 1.1617, "step": 111790 }, { "epoch": 4.5, "learning_rate": 0.0003879785820685213, "loss": 1.2358, "step": 111800 }, { "epoch": 4.5, "learning_rate": 0.0003879685172510971, "loss": 0.9429, "step": 111810 }, { "epoch": 4.5, "learning_rate": 0.0003879584524336728, "loss": 0.9187, "step": 111820 }, { "epoch": 4.5, "learning_rate": 0.00038794838761624866, "loss": 1.0481, "step": 111830 }, { "epoch": 4.5, "learning_rate": 0.00038793832279882445, "loss": 0.7823, "step": 111840 }, { "epoch": 4.5, "learning_rate": 0.00038792825798140024, "loss": 0.9441, "step": 111850 }, { "epoch": 4.5, "learning_rate": 0.00038791819316397603, "loss": 0.9928, "step": 111860 }, { "epoch": 4.5, "learning_rate": 0.0003879081283465518, "loss": 0.8997, "step": 111870 }, { "epoch": 4.5, "learning_rate": 0.0003878980635291276, "loss": 0.9948, "step": 111880 }, { "epoch": 4.5, "learning_rate": 0.0003878879987117034, "loss": 0.993, "step": 111890 }, { "epoch": 4.5, "learning_rate": 0.00038787793389427914, "loss": 0.9489, "step": 111900 }, { "epoch": 4.5, "learning_rate": 0.00038786786907685493, "loss": 0.9793, "step": 111910 }, { "epoch": 4.5, "learning_rate": 0.0003878578042594307, "loss": 1.0275, "step": 111920 }, { "epoch": 4.5, "learning_rate": 0.0003878477394420065, "loss": 1.0463, "step": 111930 }, { "epoch": 4.5, "learning_rate": 0.0003878376746245823, "loss": 0.8092, "step": 111940 }, { "epoch": 4.5, "learning_rate": 0.0003878276098071581, "loss": 1.3361, "step": 111950 }, { "epoch": 4.5, "learning_rate": 0.0003878175449897339, "loss": 0.868, "step": 111960 }, { "epoch": 4.5, "learning_rate": 0.0003878074801723097, "loss": 1.0672, "step": 111970 }, { "epoch": 4.5, "learning_rate": 0.0003877974153548855, "loss": 0.9123, "step": 111980 }, { "epoch": 4.5, "learning_rate": 0.00038778735053746126, "loss": 0.8359, "step": 111990 }, { "epoch": 4.5, "learning_rate": 0.00038777728572003705, "loss": 1.0667, "step": 112000 }, { "epoch": 4.5, "learning_rate": 0.00038776722090261284, "loss": 0.9843, "step": 112010 }, { "epoch": 4.51, "learning_rate": 0.00038775715608518863, "loss": 1.0193, "step": 112020 }, { "epoch": 4.51, "learning_rate": 0.0003877470912677644, "loss": 1.0639, "step": 112030 }, { "epoch": 4.51, "learning_rate": 0.0003877370264503402, "loss": 0.8504, "step": 112040 }, { "epoch": 4.51, "learning_rate": 0.000387726961632916, "loss": 1.0248, "step": 112050 }, { "epoch": 4.51, "learning_rate": 0.0003877168968154918, "loss": 1.1327, "step": 112060 }, { "epoch": 4.51, "learning_rate": 0.00038770683199806753, "loss": 1.0025, "step": 112070 }, { "epoch": 4.51, "learning_rate": 0.0003876967671806433, "loss": 0.8657, "step": 112080 }, { "epoch": 4.51, "learning_rate": 0.0003876867023632191, "loss": 1.1522, "step": 112090 }, { "epoch": 4.51, "learning_rate": 0.0003876766375457949, "loss": 1.035, "step": 112100 }, { "epoch": 4.51, "learning_rate": 0.0003876665727283707, "loss": 0.9485, "step": 112110 }, { "epoch": 4.51, "learning_rate": 0.0003876565079109465, "loss": 1.006, "step": 112120 }, { "epoch": 4.51, "learning_rate": 0.00038764644309352233, "loss": 0.8882, "step": 112130 }, { "epoch": 4.51, "learning_rate": 0.0003876363782760981, "loss": 1.0854, "step": 112140 }, { "epoch": 4.51, "learning_rate": 0.00038762631345867386, "loss": 0.8829, "step": 112150 }, { "epoch": 4.51, "learning_rate": 0.00038761624864124965, "loss": 1.2249, "step": 112160 }, { "epoch": 4.51, "learning_rate": 0.00038760618382382544, "loss": 1.4083, "step": 112170 }, { "epoch": 4.51, "learning_rate": 0.00038759611900640123, "loss": 1.0606, "step": 112180 }, { "epoch": 4.51, "learning_rate": 0.000387586054188977, "loss": 1.1711, "step": 112190 }, { "epoch": 4.51, "learning_rate": 0.0003875759893715528, "loss": 1.0634, "step": 112200 }, { "epoch": 4.51, "learning_rate": 0.0003875659245541286, "loss": 1.1109, "step": 112210 }, { "epoch": 4.51, "learning_rate": 0.0003875558597367044, "loss": 1.0671, "step": 112220 }, { "epoch": 4.51, "learning_rate": 0.0003875457949192802, "loss": 1.0178, "step": 112230 }, { "epoch": 4.51, "learning_rate": 0.0003875357301018559, "loss": 1.0143, "step": 112240 }, { "epoch": 4.51, "learning_rate": 0.0003875256652844317, "loss": 0.7731, "step": 112250 }, { "epoch": 4.51, "learning_rate": 0.0003875156004670075, "loss": 1.1031, "step": 112260 }, { "epoch": 4.52, "learning_rate": 0.00038750553564958335, "loss": 0.8956, "step": 112270 }, { "epoch": 4.52, "learning_rate": 0.00038749547083215914, "loss": 1.1818, "step": 112280 }, { "epoch": 4.52, "learning_rate": 0.00038748540601473493, "loss": 0.9043, "step": 112290 }, { "epoch": 4.52, "learning_rate": 0.0003874753411973107, "loss": 1.0865, "step": 112300 }, { "epoch": 4.52, "learning_rate": 0.0003874652763798865, "loss": 0.9672, "step": 112310 }, { "epoch": 4.52, "learning_rate": 0.00038745521156246225, "loss": 1.0156, "step": 112320 }, { "epoch": 4.52, "learning_rate": 0.00038744514674503804, "loss": 0.8623, "step": 112330 }, { "epoch": 4.52, "learning_rate": 0.00038743508192761383, "loss": 1.1298, "step": 112340 }, { "epoch": 4.52, "learning_rate": 0.0003874250171101896, "loss": 1.2436, "step": 112350 }, { "epoch": 4.52, "learning_rate": 0.0003874149522927654, "loss": 1.0902, "step": 112360 }, { "epoch": 4.52, "learning_rate": 0.0003874048874753412, "loss": 0.9906, "step": 112370 }, { "epoch": 4.52, "learning_rate": 0.000387394822657917, "loss": 0.9371, "step": 112380 }, { "epoch": 4.52, "learning_rate": 0.0003873847578404928, "loss": 1.1111, "step": 112390 }, { "epoch": 4.52, "learning_rate": 0.0003873746930230686, "loss": 1.2168, "step": 112400 }, { "epoch": 4.52, "learning_rate": 0.0003873646282056443, "loss": 0.9668, "step": 112410 }, { "epoch": 4.52, "learning_rate": 0.00038735456338822016, "loss": 0.9764, "step": 112420 }, { "epoch": 4.52, "learning_rate": 0.00038734449857079595, "loss": 1.0575, "step": 112430 }, { "epoch": 4.52, "learning_rate": 0.00038733443375337174, "loss": 1.0253, "step": 112440 }, { "epoch": 4.52, "learning_rate": 0.00038732436893594753, "loss": 1.0563, "step": 112450 }, { "epoch": 4.52, "learning_rate": 0.0003873143041185233, "loss": 0.8655, "step": 112460 }, { "epoch": 4.52, "learning_rate": 0.0003873042393010991, "loss": 1.1646, "step": 112470 }, { "epoch": 4.52, "learning_rate": 0.0003872941744836749, "loss": 1.2232, "step": 112480 }, { "epoch": 4.52, "learning_rate": 0.00038728410966625064, "loss": 0.9676, "step": 112490 }, { "epoch": 4.52, "learning_rate": 0.00038727404484882643, "loss": 1.0112, "step": 112500 }, { "epoch": 4.52, "learning_rate": 0.0003872639800314022, "loss": 0.897, "step": 112510 }, { "epoch": 4.53, "learning_rate": 0.000387253915213978, "loss": 1.052, "step": 112520 }, { "epoch": 4.53, "learning_rate": 0.0003872438503965538, "loss": 1.0086, "step": 112530 }, { "epoch": 4.53, "learning_rate": 0.0003872337855791296, "loss": 1.0677, "step": 112540 }, { "epoch": 4.53, "learning_rate": 0.0003872237207617054, "loss": 1.0506, "step": 112550 }, { "epoch": 4.53, "learning_rate": 0.00038721365594428117, "loss": 0.9855, "step": 112560 }, { "epoch": 4.53, "learning_rate": 0.000387203591126857, "loss": 1.1678, "step": 112570 }, { "epoch": 4.53, "learning_rate": 0.00038719352630943275, "loss": 1.2369, "step": 112580 }, { "epoch": 4.53, "learning_rate": 0.00038718346149200854, "loss": 0.8712, "step": 112590 }, { "epoch": 4.53, "learning_rate": 0.00038717339667458433, "loss": 0.9973, "step": 112600 }, { "epoch": 4.53, "learning_rate": 0.0003871633318571601, "loss": 0.9709, "step": 112610 }, { "epoch": 4.53, "learning_rate": 0.0003871532670397359, "loss": 1.1492, "step": 112620 }, { "epoch": 4.53, "learning_rate": 0.0003871432022223117, "loss": 0.9278, "step": 112630 }, { "epoch": 4.53, "learning_rate": 0.0003871331374048875, "loss": 1.0874, "step": 112640 }, { "epoch": 4.53, "learning_rate": 0.0003871230725874633, "loss": 0.9872, "step": 112650 }, { "epoch": 4.53, "learning_rate": 0.000387113007770039, "loss": 0.9726, "step": 112660 }, { "epoch": 4.53, "learning_rate": 0.0003871029429526148, "loss": 1.075, "step": 112670 }, { "epoch": 4.53, "learning_rate": 0.0003870928781351906, "loss": 0.936, "step": 112680 }, { "epoch": 4.53, "learning_rate": 0.0003870828133177664, "loss": 1.1035, "step": 112690 }, { "epoch": 4.53, "learning_rate": 0.0003870727485003422, "loss": 0.9695, "step": 112700 }, { "epoch": 4.53, "learning_rate": 0.00038706268368291803, "loss": 0.9043, "step": 112710 }, { "epoch": 4.53, "learning_rate": 0.0003870526188654938, "loss": 1.1806, "step": 112720 }, { "epoch": 4.53, "learning_rate": 0.0003870425540480696, "loss": 0.8814, "step": 112730 }, { "epoch": 4.53, "learning_rate": 0.0003870324892306454, "loss": 0.9528, "step": 112740 }, { "epoch": 4.53, "learning_rate": 0.00038702242441322114, "loss": 0.9327, "step": 112750 }, { "epoch": 4.54, "learning_rate": 0.00038701235959579693, "loss": 1.2507, "step": 112760 }, { "epoch": 4.54, "learning_rate": 0.0003870022947783727, "loss": 0.9576, "step": 112770 }, { "epoch": 4.54, "learning_rate": 0.0003869922299609485, "loss": 0.9495, "step": 112780 }, { "epoch": 4.54, "learning_rate": 0.0003869821651435243, "loss": 0.9885, "step": 112790 }, { "epoch": 4.54, "learning_rate": 0.0003869721003261001, "loss": 0.9744, "step": 112800 }, { "epoch": 4.54, "learning_rate": 0.0003869620355086759, "loss": 1.2958, "step": 112810 }, { "epoch": 4.54, "learning_rate": 0.0003869519706912517, "loss": 1.221, "step": 112820 }, { "epoch": 4.54, "learning_rate": 0.0003869419058738274, "loss": 1.1351, "step": 112830 }, { "epoch": 4.54, "learning_rate": 0.0003869318410564032, "loss": 0.8126, "step": 112840 }, { "epoch": 4.54, "learning_rate": 0.000386921776238979, "loss": 1.117, "step": 112850 }, { "epoch": 4.54, "learning_rate": 0.00038691171142155484, "loss": 0.9813, "step": 112860 }, { "epoch": 4.54, "learning_rate": 0.00038690164660413063, "loss": 1.1665, "step": 112870 }, { "epoch": 4.54, "learning_rate": 0.0003868915817867064, "loss": 1.0358, "step": 112880 }, { "epoch": 4.54, "learning_rate": 0.0003868815169692822, "loss": 1.0244, "step": 112890 }, { "epoch": 4.54, "learning_rate": 0.000386871452151858, "loss": 1.0105, "step": 112900 }, { "epoch": 4.54, "learning_rate": 0.00038686138733443374, "loss": 0.9411, "step": 112910 }, { "epoch": 4.54, "learning_rate": 0.00038685132251700953, "loss": 0.9306, "step": 112920 }, { "epoch": 4.54, "learning_rate": 0.0003868412576995853, "loss": 1.0225, "step": 112930 }, { "epoch": 4.54, "learning_rate": 0.0003868311928821611, "loss": 1.1094, "step": 112940 }, { "epoch": 4.54, "learning_rate": 0.0003868211280647369, "loss": 1.0509, "step": 112950 }, { "epoch": 4.54, "learning_rate": 0.0003868110632473127, "loss": 1.1641, "step": 112960 }, { "epoch": 4.54, "learning_rate": 0.0003868009984298885, "loss": 1.072, "step": 112970 }, { "epoch": 4.54, "learning_rate": 0.0003867909336124643, "loss": 0.9799, "step": 112980 }, { "epoch": 4.54, "learning_rate": 0.00038678086879504007, "loss": 0.8818, "step": 112990 }, { "epoch": 4.54, "learning_rate": 0.00038677080397761586, "loss": 0.8796, "step": 113000 }, { "epoch": 4.55, "learning_rate": 0.00038676073916019165, "loss": 0.9572, "step": 113010 }, { "epoch": 4.55, "learning_rate": 0.00038675067434276744, "loss": 0.9962, "step": 113020 }, { "epoch": 4.55, "learning_rate": 0.00038674060952534323, "loss": 1.069, "step": 113030 }, { "epoch": 4.55, "learning_rate": 0.000386730544707919, "loss": 0.924, "step": 113040 }, { "epoch": 4.55, "learning_rate": 0.0003867204798904948, "loss": 1.1405, "step": 113050 }, { "epoch": 4.55, "learning_rate": 0.0003867104150730706, "loss": 0.9617, "step": 113060 }, { "epoch": 4.55, "learning_rate": 0.0003867003502556464, "loss": 1.0969, "step": 113070 }, { "epoch": 4.55, "learning_rate": 0.00038669028543822213, "loss": 1.0961, "step": 113080 }, { "epoch": 4.55, "learning_rate": 0.0003866802206207979, "loss": 0.9219, "step": 113090 }, { "epoch": 4.55, "learning_rate": 0.0003866701558033737, "loss": 1.2216, "step": 113100 }, { "epoch": 4.55, "learning_rate": 0.0003866600909859495, "loss": 1.0362, "step": 113110 }, { "epoch": 4.55, "learning_rate": 0.0003866500261685253, "loss": 0.9646, "step": 113120 }, { "epoch": 4.55, "learning_rate": 0.0003866399613511011, "loss": 0.9491, "step": 113130 }, { "epoch": 4.55, "learning_rate": 0.0003866298965336769, "loss": 1.0286, "step": 113140 }, { "epoch": 4.55, "learning_rate": 0.0003866198317162527, "loss": 1.0449, "step": 113150 }, { "epoch": 4.55, "learning_rate": 0.0003866097668988285, "loss": 0.8704, "step": 113160 }, { "epoch": 4.55, "learning_rate": 0.00038659970208140425, "loss": 0.953, "step": 113170 }, { "epoch": 4.55, "learning_rate": 0.00038658963726398004, "loss": 0.8686, "step": 113180 }, { "epoch": 4.55, "learning_rate": 0.00038657957244655583, "loss": 1.0215, "step": 113190 }, { "epoch": 4.55, "learning_rate": 0.0003865695076291316, "loss": 1.1363, "step": 113200 }, { "epoch": 4.55, "learning_rate": 0.0003865594428117074, "loss": 0.8427, "step": 113210 }, { "epoch": 4.55, "learning_rate": 0.0003865493779942832, "loss": 1.1044, "step": 113220 }, { "epoch": 4.55, "learning_rate": 0.000386539313176859, "loss": 0.9732, "step": 113230 }, { "epoch": 4.55, "learning_rate": 0.0003865292483594348, "loss": 0.9717, "step": 113240 }, { "epoch": 4.55, "learning_rate": 0.0003865191835420105, "loss": 1.1732, "step": 113250 }, { "epoch": 4.56, "learning_rate": 0.0003865091187245863, "loss": 1.2355, "step": 113260 }, { "epoch": 4.56, "learning_rate": 0.0003864990539071621, "loss": 0.964, "step": 113270 }, { "epoch": 4.56, "learning_rate": 0.0003864889890897379, "loss": 0.9312, "step": 113280 }, { "epoch": 4.56, "learning_rate": 0.0003864789242723137, "loss": 0.9817, "step": 113290 }, { "epoch": 4.56, "learning_rate": 0.00038646885945488953, "loss": 1.0367, "step": 113300 }, { "epoch": 4.56, "learning_rate": 0.0003864587946374653, "loss": 0.9787, "step": 113310 }, { "epoch": 4.56, "learning_rate": 0.0003864487298200411, "loss": 1.2317, "step": 113320 }, { "epoch": 4.56, "learning_rate": 0.0003864386650026169, "loss": 0.9147, "step": 113330 }, { "epoch": 4.56, "learning_rate": 0.00038642860018519264, "loss": 1.133, "step": 113340 }, { "epoch": 4.56, "learning_rate": 0.0003864185353677684, "loss": 0.8694, "step": 113350 }, { "epoch": 4.56, "learning_rate": 0.0003864084705503442, "loss": 1.3071, "step": 113360 }, { "epoch": 4.56, "learning_rate": 0.00038639840573292, "loss": 1.0349, "step": 113370 }, { "epoch": 4.56, "learning_rate": 0.0003863883409154958, "loss": 1.1235, "step": 113380 }, { "epoch": 4.56, "learning_rate": 0.0003863782760980716, "loss": 1.1116, "step": 113390 }, { "epoch": 4.56, "learning_rate": 0.0003863682112806474, "loss": 0.9121, "step": 113400 }, { "epoch": 4.56, "learning_rate": 0.00038635814646322317, "loss": 1.0394, "step": 113410 }, { "epoch": 4.56, "learning_rate": 0.0003863480816457989, "loss": 0.986, "step": 113420 }, { "epoch": 4.56, "learning_rate": 0.0003863380168283747, "loss": 1.0782, "step": 113430 }, { "epoch": 4.56, "learning_rate": 0.00038632795201095054, "loss": 0.8398, "step": 113440 }, { "epoch": 4.56, "learning_rate": 0.00038631788719352633, "loss": 1.1741, "step": 113450 }, { "epoch": 4.56, "learning_rate": 0.0003863078223761021, "loss": 1.2116, "step": 113460 }, { "epoch": 4.56, "learning_rate": 0.0003862977575586779, "loss": 0.9062, "step": 113470 }, { "epoch": 4.56, "learning_rate": 0.0003862876927412537, "loss": 0.9396, "step": 113480 }, { "epoch": 4.56, "learning_rate": 0.0003862776279238295, "loss": 1.0362, "step": 113490 }, { "epoch": 4.56, "learning_rate": 0.0003862675631064053, "loss": 1.1961, "step": 113500 }, { "epoch": 4.57, "learning_rate": 0.000386257498288981, "loss": 1.16, "step": 113510 }, { "epoch": 4.57, "learning_rate": 0.0003862474334715568, "loss": 1.1896, "step": 113520 }, { "epoch": 4.57, "learning_rate": 0.0003862373686541326, "loss": 1.0022, "step": 113530 }, { "epoch": 4.57, "learning_rate": 0.0003862273038367084, "loss": 1.0051, "step": 113540 }, { "epoch": 4.57, "learning_rate": 0.0003862172390192842, "loss": 1.1089, "step": 113550 }, { "epoch": 4.57, "learning_rate": 0.00038620717420186, "loss": 1.0314, "step": 113560 }, { "epoch": 4.57, "learning_rate": 0.00038619710938443577, "loss": 0.9109, "step": 113570 }, { "epoch": 4.57, "learning_rate": 0.00038618704456701156, "loss": 0.9268, "step": 113580 }, { "epoch": 4.57, "learning_rate": 0.00038617697974958735, "loss": 1.0573, "step": 113590 }, { "epoch": 4.57, "learning_rate": 0.00038616691493216314, "loss": 0.9901, "step": 113600 }, { "epoch": 4.57, "learning_rate": 0.00038615685011473893, "loss": 0.8606, "step": 113610 }, { "epoch": 4.57, "learning_rate": 0.0003861467852973147, "loss": 1.0137, "step": 113620 }, { "epoch": 4.57, "learning_rate": 0.0003861367204798905, "loss": 0.9343, "step": 113630 }, { "epoch": 4.57, "learning_rate": 0.0003861266556624663, "loss": 0.9547, "step": 113640 }, { "epoch": 4.57, "learning_rate": 0.0003861165908450421, "loss": 1.0957, "step": 113650 }, { "epoch": 4.57, "learning_rate": 0.0003861065260276179, "loss": 1.0357, "step": 113660 }, { "epoch": 4.57, "learning_rate": 0.0003860964612101936, "loss": 0.9989, "step": 113670 }, { "epoch": 4.57, "learning_rate": 0.0003860863963927694, "loss": 0.8935, "step": 113680 }, { "epoch": 4.57, "learning_rate": 0.0003860763315753452, "loss": 0.7334, "step": 113690 }, { "epoch": 4.57, "learning_rate": 0.000386066266757921, "loss": 1.2205, "step": 113700 }, { "epoch": 4.57, "learning_rate": 0.0003860562019404968, "loss": 1.1752, "step": 113710 }, { "epoch": 4.57, "learning_rate": 0.0003860461371230726, "loss": 0.9064, "step": 113720 }, { "epoch": 4.57, "learning_rate": 0.00038603607230564837, "loss": 0.9349, "step": 113730 }, { "epoch": 4.57, "learning_rate": 0.0003860260074882242, "loss": 1.0618, "step": 113740 }, { "epoch": 4.57, "learning_rate": 0.0003860159426708, "loss": 0.8858, "step": 113750 }, { "epoch": 4.58, "learning_rate": 0.00038600587785337574, "loss": 0.9082, "step": 113760 }, { "epoch": 4.58, "learning_rate": 0.00038599581303595153, "loss": 1.0781, "step": 113770 }, { "epoch": 4.58, "learning_rate": 0.0003859857482185273, "loss": 1.0307, "step": 113780 }, { "epoch": 4.58, "learning_rate": 0.0003859756834011031, "loss": 1.284, "step": 113790 }, { "epoch": 4.58, "learning_rate": 0.0003859656185836789, "loss": 1.0728, "step": 113800 }, { "epoch": 4.58, "learning_rate": 0.0003859555537662547, "loss": 1.0212, "step": 113810 }, { "epoch": 4.58, "learning_rate": 0.0003859454889488305, "loss": 0.929, "step": 113820 }, { "epoch": 4.58, "learning_rate": 0.0003859354241314063, "loss": 1.0112, "step": 113830 }, { "epoch": 4.58, "learning_rate": 0.000385925359313982, "loss": 1.169, "step": 113840 }, { "epoch": 4.58, "learning_rate": 0.0003859152944965578, "loss": 0.847, "step": 113850 }, { "epoch": 4.58, "learning_rate": 0.0003859052296791336, "loss": 1.1602, "step": 113860 }, { "epoch": 4.58, "learning_rate": 0.0003858951648617094, "loss": 1.0204, "step": 113870 }, { "epoch": 4.58, "learning_rate": 0.00038588510004428523, "loss": 0.8866, "step": 113880 }, { "epoch": 4.58, "learning_rate": 0.000385875035226861, "loss": 1.0065, "step": 113890 }, { "epoch": 4.58, "learning_rate": 0.0003858649704094368, "loss": 1.1159, "step": 113900 }, { "epoch": 4.58, "learning_rate": 0.0003858549055920126, "loss": 0.8511, "step": 113910 }, { "epoch": 4.58, "learning_rate": 0.0003858448407745884, "loss": 0.9748, "step": 113920 }, { "epoch": 4.58, "learning_rate": 0.00038583477595716413, "loss": 0.908, "step": 113930 }, { "epoch": 4.58, "learning_rate": 0.0003858247111397399, "loss": 1.0436, "step": 113940 }, { "epoch": 4.58, "learning_rate": 0.0003858146463223157, "loss": 1.0952, "step": 113950 }, { "epoch": 4.58, "learning_rate": 0.0003858045815048915, "loss": 1.1164, "step": 113960 }, { "epoch": 4.58, "learning_rate": 0.0003857945166874673, "loss": 0.9348, "step": 113970 }, { "epoch": 4.58, "learning_rate": 0.0003857844518700431, "loss": 0.8005, "step": 113980 }, { "epoch": 4.58, "learning_rate": 0.0003857743870526189, "loss": 1.0992, "step": 113990 }, { "epoch": 4.58, "learning_rate": 0.00038576432223519467, "loss": 0.9331, "step": 114000 }, { "epoch": 4.59, "learning_rate": 0.0003857542574177704, "loss": 1.3045, "step": 114010 }, { "epoch": 4.59, "learning_rate": 0.0003857441926003462, "loss": 1.0272, "step": 114020 }, { "epoch": 4.59, "learning_rate": 0.00038573412778292204, "loss": 1.2135, "step": 114030 }, { "epoch": 4.59, "learning_rate": 0.00038572406296549783, "loss": 1.31, "step": 114040 }, { "epoch": 4.59, "learning_rate": 0.0003857139981480736, "loss": 0.8769, "step": 114050 }, { "epoch": 4.59, "learning_rate": 0.0003857039333306494, "loss": 0.992, "step": 114060 }, { "epoch": 4.59, "learning_rate": 0.0003856938685132252, "loss": 1.106, "step": 114070 }, { "epoch": 4.59, "learning_rate": 0.000385683803695801, "loss": 1.0472, "step": 114080 }, { "epoch": 4.59, "learning_rate": 0.0003856737388783768, "loss": 1.1411, "step": 114090 }, { "epoch": 4.59, "learning_rate": 0.0003856636740609525, "loss": 1.1967, "step": 114100 }, { "epoch": 4.59, "learning_rate": 0.0003856536092435283, "loss": 1.3163, "step": 114110 }, { "epoch": 4.59, "learning_rate": 0.0003856435444261041, "loss": 1.064, "step": 114120 }, { "epoch": 4.59, "learning_rate": 0.0003856334796086799, "loss": 1.0836, "step": 114130 }, { "epoch": 4.59, "learning_rate": 0.0003856234147912557, "loss": 1.0388, "step": 114140 }, { "epoch": 4.59, "learning_rate": 0.00038561334997383147, "loss": 1.0445, "step": 114150 }, { "epoch": 4.59, "learning_rate": 0.00038560328515640726, "loss": 1.0685, "step": 114160 }, { "epoch": 4.59, "learning_rate": 0.0003855932203389831, "loss": 1.3191, "step": 114170 }, { "epoch": 4.59, "learning_rate": 0.00038558315552155885, "loss": 1.1799, "step": 114180 }, { "epoch": 4.59, "learning_rate": 0.00038557309070413464, "loss": 0.997, "step": 114190 }, { "epoch": 4.59, "learning_rate": 0.0003855630258867104, "loss": 0.8738, "step": 114200 }, { "epoch": 4.59, "learning_rate": 0.0003855529610692862, "loss": 0.9409, "step": 114210 }, { "epoch": 4.59, "learning_rate": 0.000385542896251862, "loss": 0.8349, "step": 114220 }, { "epoch": 4.59, "learning_rate": 0.0003855328314344378, "loss": 1.0051, "step": 114230 }, { "epoch": 4.59, "learning_rate": 0.0003855227666170136, "loss": 1.0006, "step": 114240 }, { "epoch": 4.59, "learning_rate": 0.0003855127017995894, "loss": 1.0584, "step": 114250 }, { "epoch": 4.6, "learning_rate": 0.00038550263698216517, "loss": 1.1488, "step": 114260 }, { "epoch": 4.6, "learning_rate": 0.0003854925721647409, "loss": 0.9865, "step": 114270 }, { "epoch": 4.6, "learning_rate": 0.0003854825073473167, "loss": 0.9824, "step": 114280 }, { "epoch": 4.6, "learning_rate": 0.0003854724425298925, "loss": 0.9233, "step": 114290 }, { "epoch": 4.6, "learning_rate": 0.0003854623777124683, "loss": 0.9001, "step": 114300 }, { "epoch": 4.6, "learning_rate": 0.00038545231289504407, "loss": 1.0865, "step": 114310 }, { "epoch": 4.6, "learning_rate": 0.0003854422480776199, "loss": 1.1065, "step": 114320 }, { "epoch": 4.6, "learning_rate": 0.0003854321832601957, "loss": 0.9415, "step": 114330 }, { "epoch": 4.6, "learning_rate": 0.0003854221184427715, "loss": 0.937, "step": 114340 }, { "epoch": 4.6, "learning_rate": 0.00038541205362534723, "loss": 1.013, "step": 114350 }, { "epoch": 4.6, "learning_rate": 0.000385401988807923, "loss": 0.9879, "step": 114360 }, { "epoch": 4.6, "learning_rate": 0.0003853919239904988, "loss": 1.048, "step": 114370 }, { "epoch": 4.6, "learning_rate": 0.0003853818591730746, "loss": 0.9387, "step": 114380 }, { "epoch": 4.6, "learning_rate": 0.0003853717943556504, "loss": 0.9281, "step": 114390 }, { "epoch": 4.6, "learning_rate": 0.0003853617295382262, "loss": 1.0856, "step": 114400 }, { "epoch": 4.6, "learning_rate": 0.000385351664720802, "loss": 0.8968, "step": 114410 }, { "epoch": 4.6, "learning_rate": 0.00038534159990337777, "loss": 0.9843, "step": 114420 }, { "epoch": 4.6, "learning_rate": 0.0003853315350859535, "loss": 0.9851, "step": 114430 }, { "epoch": 4.6, "learning_rate": 0.0003853214702685293, "loss": 1.0289, "step": 114440 }, { "epoch": 4.6, "learning_rate": 0.0003853114054511051, "loss": 1.0804, "step": 114450 }, { "epoch": 4.6, "learning_rate": 0.0003853013406336809, "loss": 0.9683, "step": 114460 }, { "epoch": 4.6, "learning_rate": 0.0003852912758162567, "loss": 1.025, "step": 114470 }, { "epoch": 4.6, "learning_rate": 0.0003852812109988325, "loss": 0.9565, "step": 114480 }, { "epoch": 4.6, "learning_rate": 0.0003852711461814083, "loss": 0.984, "step": 114490 }, { "epoch": 4.61, "learning_rate": 0.0003852610813639841, "loss": 0.9715, "step": 114500 }, { "epoch": 4.61, "learning_rate": 0.0003852510165465599, "loss": 1.1394, "step": 114510 }, { "epoch": 4.61, "learning_rate": 0.0003852409517291356, "loss": 1.026, "step": 114520 }, { "epoch": 4.61, "learning_rate": 0.0003852308869117114, "loss": 1.0402, "step": 114530 }, { "epoch": 4.61, "learning_rate": 0.0003852208220942872, "loss": 0.9016, "step": 114540 }, { "epoch": 4.61, "learning_rate": 0.000385210757276863, "loss": 1.0612, "step": 114550 }, { "epoch": 4.61, "learning_rate": 0.0003852006924594388, "loss": 0.8473, "step": 114560 }, { "epoch": 4.61, "learning_rate": 0.0003851906276420146, "loss": 0.9152, "step": 114570 }, { "epoch": 4.61, "learning_rate": 0.00038518056282459037, "loss": 1.1615, "step": 114580 }, { "epoch": 4.61, "learning_rate": 0.00038517049800716616, "loss": 1.2494, "step": 114590 }, { "epoch": 4.61, "learning_rate": 0.0003851604331897419, "loss": 1.0966, "step": 114600 }, { "epoch": 4.61, "learning_rate": 0.00038515036837231774, "loss": 1.1368, "step": 114610 }, { "epoch": 4.61, "learning_rate": 0.00038514030355489353, "loss": 1.0519, "step": 114620 }, { "epoch": 4.61, "learning_rate": 0.0003851302387374693, "loss": 1.0854, "step": 114630 }, { "epoch": 4.61, "learning_rate": 0.0003851201739200451, "loss": 1.0559, "step": 114640 }, { "epoch": 4.61, "learning_rate": 0.0003851101091026209, "loss": 1.0043, "step": 114650 }, { "epoch": 4.61, "learning_rate": 0.0003851000442851967, "loss": 1.0461, "step": 114660 }, { "epoch": 4.61, "learning_rate": 0.0003850899794677725, "loss": 0.9801, "step": 114670 }, { "epoch": 4.61, "learning_rate": 0.0003850799146503483, "loss": 0.9758, "step": 114680 }, { "epoch": 4.61, "learning_rate": 0.000385069849832924, "loss": 1.1335, "step": 114690 }, { "epoch": 4.61, "learning_rate": 0.0003850597850154998, "loss": 1.2347, "step": 114700 }, { "epoch": 4.61, "learning_rate": 0.0003850497201980756, "loss": 1.0729, "step": 114710 }, { "epoch": 4.61, "learning_rate": 0.0003850396553806514, "loss": 0.8453, "step": 114720 }, { "epoch": 4.61, "learning_rate": 0.0003850295905632272, "loss": 1.179, "step": 114730 }, { "epoch": 4.61, "learning_rate": 0.00038501952574580297, "loss": 0.8571, "step": 114740 }, { "epoch": 4.62, "learning_rate": 0.00038500946092837876, "loss": 0.9818, "step": 114750 }, { "epoch": 4.62, "learning_rate": 0.0003849993961109546, "loss": 1.1224, "step": 114760 }, { "epoch": 4.62, "learning_rate": 0.00038498933129353034, "loss": 1.0878, "step": 114770 }, { "epoch": 4.62, "learning_rate": 0.00038497926647610613, "loss": 0.8184, "step": 114780 }, { "epoch": 4.62, "learning_rate": 0.0003849692016586819, "loss": 1.1062, "step": 114790 }, { "epoch": 4.62, "learning_rate": 0.0003849591368412577, "loss": 1.0488, "step": 114800 }, { "epoch": 4.62, "learning_rate": 0.0003849490720238335, "loss": 1.0383, "step": 114810 }, { "epoch": 4.62, "learning_rate": 0.0003849390072064093, "loss": 1.2852, "step": 114820 }, { "epoch": 4.62, "learning_rate": 0.0003849289423889851, "loss": 1.029, "step": 114830 }, { "epoch": 4.62, "learning_rate": 0.0003849188775715609, "loss": 0.994, "step": 114840 }, { "epoch": 4.62, "learning_rate": 0.00038490881275413666, "loss": 1.0669, "step": 114850 }, { "epoch": 4.62, "learning_rate": 0.0003848987479367124, "loss": 1.0422, "step": 114860 }, { "epoch": 4.62, "learning_rate": 0.0003848886831192882, "loss": 1.1244, "step": 114870 }, { "epoch": 4.62, "learning_rate": 0.000384878618301864, "loss": 1.0654, "step": 114880 }, { "epoch": 4.62, "learning_rate": 0.0003848685534844398, "loss": 1.0984, "step": 114890 }, { "epoch": 4.62, "learning_rate": 0.0003848584886670156, "loss": 1.0819, "step": 114900 }, { "epoch": 4.62, "learning_rate": 0.0003848484238495914, "loss": 1.1352, "step": 114910 }, { "epoch": 4.62, "learning_rate": 0.0003848383590321672, "loss": 1.1005, "step": 114920 }, { "epoch": 4.62, "learning_rate": 0.000384828294214743, "loss": 0.9516, "step": 114930 }, { "epoch": 4.62, "learning_rate": 0.00038481822939731873, "loss": 1.2804, "step": 114940 }, { "epoch": 4.62, "learning_rate": 0.0003848081645798945, "loss": 1.0778, "step": 114950 }, { "epoch": 4.62, "learning_rate": 0.0003847980997624703, "loss": 1.1018, "step": 114960 }, { "epoch": 4.62, "learning_rate": 0.0003847880349450461, "loss": 0.9921, "step": 114970 }, { "epoch": 4.62, "learning_rate": 0.0003847779701276219, "loss": 1.0479, "step": 114980 }, { "epoch": 4.62, "learning_rate": 0.0003847679053101977, "loss": 1.212, "step": 114990 }, { "epoch": 4.63, "learning_rate": 0.00038475784049277347, "loss": 1.2207, "step": 115000 }, { "epoch": 4.63, "learning_rate": 0.00038474777567534926, "loss": 1.0192, "step": 115010 }, { "epoch": 4.63, "learning_rate": 0.00038473771085792505, "loss": 1.2994, "step": 115020 }, { "epoch": 4.63, "learning_rate": 0.0003847276460405008, "loss": 1.1129, "step": 115030 }, { "epoch": 4.63, "learning_rate": 0.0003847175812230766, "loss": 1.0476, "step": 115040 }, { "epoch": 4.63, "learning_rate": 0.0003847075164056524, "loss": 1.3267, "step": 115050 }, { "epoch": 4.63, "learning_rate": 0.0003846974515882282, "loss": 1.1828, "step": 115060 }, { "epoch": 4.63, "learning_rate": 0.000384687386770804, "loss": 1.0677, "step": 115070 }, { "epoch": 4.63, "learning_rate": 0.0003846773219533798, "loss": 1.2286, "step": 115080 }, { "epoch": 4.63, "learning_rate": 0.0003846672571359556, "loss": 0.9318, "step": 115090 }, { "epoch": 4.63, "learning_rate": 0.0003846571923185314, "loss": 1.1968, "step": 115100 }, { "epoch": 4.63, "learning_rate": 0.0003846471275011071, "loss": 1.0433, "step": 115110 }, { "epoch": 4.63, "learning_rate": 0.0003846370626836829, "loss": 1.1575, "step": 115120 }, { "epoch": 4.63, "learning_rate": 0.0003846269978662587, "loss": 1.2095, "step": 115130 }, { "epoch": 4.63, "learning_rate": 0.0003846169330488345, "loss": 0.9916, "step": 115140 }, { "epoch": 4.63, "learning_rate": 0.0003846068682314103, "loss": 1.1019, "step": 115150 }, { "epoch": 4.63, "learning_rate": 0.00038459680341398607, "loss": 1.2201, "step": 115160 }, { "epoch": 4.63, "learning_rate": 0.00038458673859656186, "loss": 1.1262, "step": 115170 }, { "epoch": 4.63, "learning_rate": 0.00038457667377913765, "loss": 1.2615, "step": 115180 }, { "epoch": 4.63, "learning_rate": 0.0003845666089617134, "loss": 0.9307, "step": 115190 }, { "epoch": 4.63, "learning_rate": 0.00038455654414428923, "loss": 0.9978, "step": 115200 }, { "epoch": 4.63, "learning_rate": 0.000384546479326865, "loss": 1.0079, "step": 115210 }, { "epoch": 4.63, "learning_rate": 0.0003845364145094408, "loss": 1.1618, "step": 115220 }, { "epoch": 4.63, "learning_rate": 0.0003845263496920166, "loss": 0.9443, "step": 115230 }, { "epoch": 4.63, "learning_rate": 0.0003845162848745924, "loss": 1.0435, "step": 115240 }, { "epoch": 4.64, "learning_rate": 0.0003845062200571682, "loss": 0.9697, "step": 115250 }, { "epoch": 4.64, "learning_rate": 0.000384496155239744, "loss": 1.0308, "step": 115260 }, { "epoch": 4.64, "learning_rate": 0.00038448609042231977, "loss": 0.8621, "step": 115270 }, { "epoch": 4.64, "learning_rate": 0.0003844760256048955, "loss": 1.0362, "step": 115280 }, { "epoch": 4.64, "learning_rate": 0.0003844659607874713, "loss": 1.0015, "step": 115290 }, { "epoch": 4.64, "learning_rate": 0.0003844558959700471, "loss": 1.1214, "step": 115300 }, { "epoch": 4.64, "learning_rate": 0.0003844458311526229, "loss": 0.9415, "step": 115310 }, { "epoch": 4.64, "learning_rate": 0.00038443576633519867, "loss": 1.1859, "step": 115320 }, { "epoch": 4.64, "learning_rate": 0.00038442570151777446, "loss": 0.873, "step": 115330 }, { "epoch": 4.64, "learning_rate": 0.0003844156367003503, "loss": 1.1335, "step": 115340 }, { "epoch": 4.64, "learning_rate": 0.0003844055718829261, "loss": 1.0446, "step": 115350 }, { "epoch": 4.64, "learning_rate": 0.00038439550706550183, "loss": 0.8611, "step": 115360 }, { "epoch": 4.64, "learning_rate": 0.0003843854422480776, "loss": 1.0607, "step": 115370 }, { "epoch": 4.64, "learning_rate": 0.0003843753774306534, "loss": 0.8472, "step": 115380 }, { "epoch": 4.64, "learning_rate": 0.0003843653126132292, "loss": 1.3686, "step": 115390 }, { "epoch": 4.64, "learning_rate": 0.000384355247795805, "loss": 1.2944, "step": 115400 }, { "epoch": 4.64, "learning_rate": 0.0003843451829783808, "loss": 0.9565, "step": 115410 }, { "epoch": 4.64, "learning_rate": 0.0003843351181609566, "loss": 1.0289, "step": 115420 }, { "epoch": 4.64, "learning_rate": 0.00038432505334353237, "loss": 1.1377, "step": 115430 }, { "epoch": 4.64, "learning_rate": 0.00038431498852610816, "loss": 0.8908, "step": 115440 }, { "epoch": 4.64, "learning_rate": 0.0003843049237086839, "loss": 1.2941, "step": 115450 }, { "epoch": 4.64, "learning_rate": 0.0003842948588912597, "loss": 1.043, "step": 115460 }, { "epoch": 4.64, "learning_rate": 0.0003842847940738355, "loss": 1.1737, "step": 115470 }, { "epoch": 4.64, "learning_rate": 0.00038427472925641127, "loss": 1.0329, "step": 115480 }, { "epoch": 4.64, "learning_rate": 0.0003842646644389871, "loss": 1.2884, "step": 115490 }, { "epoch": 4.65, "learning_rate": 0.0003842545996215629, "loss": 0.9868, "step": 115500 }, { "epoch": 4.65, "learning_rate": 0.0003842445348041387, "loss": 0.8142, "step": 115510 }, { "epoch": 4.65, "learning_rate": 0.0003842344699867145, "loss": 0.9978, "step": 115520 }, { "epoch": 4.65, "learning_rate": 0.0003842244051692902, "loss": 1.0745, "step": 115530 }, { "epoch": 4.65, "learning_rate": 0.000384214340351866, "loss": 1.1169, "step": 115540 }, { "epoch": 4.65, "learning_rate": 0.0003842042755344418, "loss": 1.1797, "step": 115550 }, { "epoch": 4.65, "learning_rate": 0.0003841942107170176, "loss": 1.0876, "step": 115560 }, { "epoch": 4.65, "learning_rate": 0.0003841841458995934, "loss": 1.2414, "step": 115570 }, { "epoch": 4.65, "learning_rate": 0.0003841740810821692, "loss": 1.0086, "step": 115580 }, { "epoch": 4.65, "learning_rate": 0.00038416401626474497, "loss": 0.9964, "step": 115590 }, { "epoch": 4.65, "learning_rate": 0.00038415395144732076, "loss": 0.8479, "step": 115600 }, { "epoch": 4.65, "learning_rate": 0.00038414388662989655, "loss": 1.1383, "step": 115610 }, { "epoch": 4.65, "learning_rate": 0.0003841338218124723, "loss": 0.894, "step": 115620 }, { "epoch": 4.65, "learning_rate": 0.0003841237569950481, "loss": 1.091, "step": 115630 }, { "epoch": 4.65, "learning_rate": 0.0003841136921776239, "loss": 1.1554, "step": 115640 }, { "epoch": 4.65, "learning_rate": 0.0003841036273601997, "loss": 1.1041, "step": 115650 }, { "epoch": 4.65, "learning_rate": 0.0003840935625427755, "loss": 1.1966, "step": 115660 }, { "epoch": 4.65, "learning_rate": 0.0003840834977253513, "loss": 0.9145, "step": 115670 }, { "epoch": 4.65, "learning_rate": 0.0003840734329079271, "loss": 1.0529, "step": 115680 }, { "epoch": 4.65, "learning_rate": 0.0003840633680905029, "loss": 1.0447, "step": 115690 }, { "epoch": 4.65, "learning_rate": 0.0003840533032730786, "loss": 1.1613, "step": 115700 }, { "epoch": 4.65, "learning_rate": 0.0003840432384556544, "loss": 0.914, "step": 115710 }, { "epoch": 4.65, "learning_rate": 0.0003840331736382302, "loss": 1.0033, "step": 115720 }, { "epoch": 4.65, "learning_rate": 0.000384023108820806, "loss": 0.9618, "step": 115730 }, { "epoch": 4.65, "learning_rate": 0.0003840130440033818, "loss": 1.1712, "step": 115740 }, { "epoch": 4.66, "learning_rate": 0.00038400297918595756, "loss": 1.0203, "step": 115750 }, { "epoch": 4.66, "learning_rate": 0.00038399291436853336, "loss": 1.2071, "step": 115760 }, { "epoch": 4.66, "learning_rate": 0.00038398284955110915, "loss": 1.145, "step": 115770 }, { "epoch": 4.66, "learning_rate": 0.000383972784733685, "loss": 1.1269, "step": 115780 }, { "epoch": 4.66, "learning_rate": 0.00038396271991626073, "loss": 1.2738, "step": 115790 }, { "epoch": 4.66, "learning_rate": 0.0003839526550988365, "loss": 1.0075, "step": 115800 }, { "epoch": 4.66, "learning_rate": 0.0003839425902814123, "loss": 1.134, "step": 115810 }, { "epoch": 4.66, "learning_rate": 0.0003839325254639881, "loss": 1.344, "step": 115820 }, { "epoch": 4.66, "learning_rate": 0.0003839224606465639, "loss": 1.011, "step": 115830 }, { "epoch": 4.66, "learning_rate": 0.0003839123958291397, "loss": 1.0798, "step": 115840 }, { "epoch": 4.66, "learning_rate": 0.00038390233101171547, "loss": 1.0251, "step": 115850 }, { "epoch": 4.66, "learning_rate": 0.00038389226619429126, "loss": 0.9843, "step": 115860 }, { "epoch": 4.66, "learning_rate": 0.000383882201376867, "loss": 0.9673, "step": 115870 }, { "epoch": 4.66, "learning_rate": 0.0003838721365594428, "loss": 1.069, "step": 115880 }, { "epoch": 4.66, "learning_rate": 0.0003838620717420186, "loss": 1.0296, "step": 115890 }, { "epoch": 4.66, "learning_rate": 0.00038385200692459437, "loss": 0.961, "step": 115900 }, { "epoch": 4.66, "learning_rate": 0.00038384194210717016, "loss": 1.0323, "step": 115910 }, { "epoch": 4.66, "learning_rate": 0.00038383187728974595, "loss": 1.0416, "step": 115920 }, { "epoch": 4.66, "learning_rate": 0.0003838218124723218, "loss": 1.0116, "step": 115930 }, { "epoch": 4.66, "learning_rate": 0.0003838117476548976, "loss": 1.0974, "step": 115940 }, { "epoch": 4.66, "learning_rate": 0.0003838016828374734, "loss": 1.2632, "step": 115950 }, { "epoch": 4.66, "learning_rate": 0.0003837916180200491, "loss": 1.04, "step": 115960 }, { "epoch": 4.66, "learning_rate": 0.0003837815532026249, "loss": 1.0314, "step": 115970 }, { "epoch": 4.66, "learning_rate": 0.0003837714883852007, "loss": 0.902, "step": 115980 }, { "epoch": 4.66, "learning_rate": 0.0003837614235677765, "loss": 1.2116, "step": 115990 }, { "epoch": 4.67, "learning_rate": 0.0003837513587503523, "loss": 1.0076, "step": 116000 }, { "epoch": 4.67, "learning_rate": 0.00038374129393292807, "loss": 1.1554, "step": 116010 }, { "epoch": 4.67, "learning_rate": 0.00038373122911550386, "loss": 0.9392, "step": 116020 }, { "epoch": 4.67, "learning_rate": 0.00038372116429807965, "loss": 1.0597, "step": 116030 }, { "epoch": 4.67, "learning_rate": 0.0003837110994806554, "loss": 1.069, "step": 116040 }, { "epoch": 4.67, "learning_rate": 0.0003837010346632312, "loss": 1.2509, "step": 116050 }, { "epoch": 4.67, "learning_rate": 0.00038369096984580697, "loss": 0.8719, "step": 116060 }, { "epoch": 4.67, "learning_rate": 0.0003836809050283828, "loss": 1.117, "step": 116070 }, { "epoch": 4.67, "learning_rate": 0.0003836708402109586, "loss": 0.9702, "step": 116080 }, { "epoch": 4.67, "learning_rate": 0.0003836607753935344, "loss": 1.1952, "step": 116090 }, { "epoch": 4.67, "learning_rate": 0.0003836507105761102, "loss": 0.9218, "step": 116100 }, { "epoch": 4.67, "learning_rate": 0.000383640645758686, "loss": 1.0983, "step": 116110 }, { "epoch": 4.67, "learning_rate": 0.0003836305809412617, "loss": 0.9185, "step": 116120 }, { "epoch": 4.67, "learning_rate": 0.0003836205161238375, "loss": 0.9902, "step": 116130 }, { "epoch": 4.67, "learning_rate": 0.0003836104513064133, "loss": 0.9499, "step": 116140 }, { "epoch": 4.67, "learning_rate": 0.0003836003864889891, "loss": 0.9454, "step": 116150 }, { "epoch": 4.67, "learning_rate": 0.0003835903216715649, "loss": 0.9443, "step": 116160 }, { "epoch": 4.67, "learning_rate": 0.00038358025685414067, "loss": 1.1047, "step": 116170 }, { "epoch": 4.67, "learning_rate": 0.00038357019203671646, "loss": 1.1583, "step": 116180 }, { "epoch": 4.67, "learning_rate": 0.00038356012721929225, "loss": 0.8981, "step": 116190 }, { "epoch": 4.67, "learning_rate": 0.00038355006240186804, "loss": 1.0001, "step": 116200 }, { "epoch": 4.67, "learning_rate": 0.0003835399975844438, "loss": 1.0516, "step": 116210 }, { "epoch": 4.67, "learning_rate": 0.0003835299327670196, "loss": 1.013, "step": 116220 }, { "epoch": 4.67, "learning_rate": 0.0003835198679495954, "loss": 1.0563, "step": 116230 }, { "epoch": 4.67, "learning_rate": 0.0003835098031321712, "loss": 0.9724, "step": 116240 }, { "epoch": 4.68, "learning_rate": 0.000383499738314747, "loss": 1.2196, "step": 116250 }, { "epoch": 4.68, "learning_rate": 0.0003834896734973228, "loss": 1.1748, "step": 116260 }, { "epoch": 4.68, "learning_rate": 0.0003834796086798986, "loss": 0.8003, "step": 116270 }, { "epoch": 4.68, "learning_rate": 0.00038346954386247437, "loss": 0.8818, "step": 116280 }, { "epoch": 4.68, "learning_rate": 0.0003834594790450501, "loss": 0.8901, "step": 116290 }, { "epoch": 4.68, "learning_rate": 0.0003834494142276259, "loss": 1.1518, "step": 116300 }, { "epoch": 4.68, "learning_rate": 0.0003834393494102017, "loss": 1.1005, "step": 116310 }, { "epoch": 4.68, "learning_rate": 0.0003834292845927775, "loss": 1.1644, "step": 116320 }, { "epoch": 4.68, "learning_rate": 0.00038341921977535327, "loss": 1.0701, "step": 116330 }, { "epoch": 4.68, "learning_rate": 0.00038340915495792906, "loss": 1.165, "step": 116340 }, { "epoch": 4.68, "learning_rate": 0.00038339909014050485, "loss": 1.1258, "step": 116350 }, { "epoch": 4.68, "learning_rate": 0.00038338902532308064, "loss": 0.9229, "step": 116360 }, { "epoch": 4.68, "learning_rate": 0.0003833789605056565, "loss": 0.914, "step": 116370 }, { "epoch": 4.68, "learning_rate": 0.0003833688956882322, "loss": 1.1037, "step": 116380 }, { "epoch": 4.68, "learning_rate": 0.000383358830870808, "loss": 0.9264, "step": 116390 }, { "epoch": 4.68, "learning_rate": 0.0003833487660533838, "loss": 0.9761, "step": 116400 }, { "epoch": 4.68, "learning_rate": 0.0003833387012359596, "loss": 0.945, "step": 116410 }, { "epoch": 4.68, "learning_rate": 0.0003833286364185354, "loss": 1.0464, "step": 116420 }, { "epoch": 4.68, "learning_rate": 0.0003833185716011112, "loss": 1.0939, "step": 116430 }, { "epoch": 4.68, "learning_rate": 0.00038330850678368697, "loss": 1.0283, "step": 116440 }, { "epoch": 4.68, "learning_rate": 0.00038329844196626276, "loss": 1.008, "step": 116450 }, { "epoch": 4.68, "learning_rate": 0.0003832883771488385, "loss": 0.9994, "step": 116460 }, { "epoch": 4.68, "learning_rate": 0.0003832783123314143, "loss": 0.9567, "step": 116470 }, { "epoch": 4.68, "learning_rate": 0.0003832682475139901, "loss": 0.8901, "step": 116480 }, { "epoch": 4.69, "learning_rate": 0.00038325818269656587, "loss": 1.1813, "step": 116490 }, { "epoch": 4.69, "learning_rate": 0.00038324811787914166, "loss": 0.9706, "step": 116500 }, { "epoch": 4.69, "learning_rate": 0.0003832380530617175, "loss": 0.9453, "step": 116510 }, { "epoch": 4.69, "learning_rate": 0.0003832279882442933, "loss": 0.9263, "step": 116520 }, { "epoch": 4.69, "learning_rate": 0.0003832179234268691, "loss": 1.0212, "step": 116530 }, { "epoch": 4.69, "learning_rate": 0.0003832078586094449, "loss": 0.8921, "step": 116540 }, { "epoch": 4.69, "learning_rate": 0.0003831977937920206, "loss": 0.9913, "step": 116550 }, { "epoch": 4.69, "learning_rate": 0.0003831877289745964, "loss": 1.0685, "step": 116560 }, { "epoch": 4.69, "learning_rate": 0.0003831776641571722, "loss": 0.9924, "step": 116570 }, { "epoch": 4.69, "learning_rate": 0.000383167599339748, "loss": 1.0323, "step": 116580 }, { "epoch": 4.69, "learning_rate": 0.0003831575345223238, "loss": 0.9715, "step": 116590 }, { "epoch": 4.69, "learning_rate": 0.00038314746970489956, "loss": 0.9819, "step": 116600 }, { "epoch": 4.69, "learning_rate": 0.00038313740488747535, "loss": 1.1062, "step": 116610 }, { "epoch": 4.69, "learning_rate": 0.00038312734007005115, "loss": 1.0248, "step": 116620 }, { "epoch": 4.69, "learning_rate": 0.0003831172752526269, "loss": 1.0675, "step": 116630 }, { "epoch": 4.69, "learning_rate": 0.0003831072104352027, "loss": 0.9554, "step": 116640 }, { "epoch": 4.69, "learning_rate": 0.00038309714561777846, "loss": 1.1836, "step": 116650 }, { "epoch": 4.69, "learning_rate": 0.0003830870808003543, "loss": 0.8859, "step": 116660 }, { "epoch": 4.69, "learning_rate": 0.0003830770159829301, "loss": 1.1626, "step": 116670 }, { "epoch": 4.69, "learning_rate": 0.0003830669511655059, "loss": 1.0346, "step": 116680 }, { "epoch": 4.69, "learning_rate": 0.0003830568863480817, "loss": 0.9867, "step": 116690 }, { "epoch": 4.69, "learning_rate": 0.00038304682153065747, "loss": 1.1515, "step": 116700 }, { "epoch": 4.69, "learning_rate": 0.00038303675671323326, "loss": 1.125, "step": 116710 }, { "epoch": 4.69, "learning_rate": 0.000383026691895809, "loss": 1.0402, "step": 116720 }, { "epoch": 4.69, "learning_rate": 0.0003830166270783848, "loss": 1.039, "step": 116730 }, { "epoch": 4.7, "learning_rate": 0.0003830065622609606, "loss": 1.1526, "step": 116740 }, { "epoch": 4.7, "learning_rate": 0.00038299649744353637, "loss": 0.9934, "step": 116750 }, { "epoch": 4.7, "learning_rate": 0.00038298643262611216, "loss": 1.1657, "step": 116760 }, { "epoch": 4.7, "learning_rate": 0.00038297636780868795, "loss": 1.0961, "step": 116770 }, { "epoch": 4.7, "learning_rate": 0.00038296630299126374, "loss": 1.1633, "step": 116780 }, { "epoch": 4.7, "learning_rate": 0.00038295623817383953, "loss": 1.112, "step": 116790 }, { "epoch": 4.7, "learning_rate": 0.00038294617335641527, "loss": 1.052, "step": 116800 }, { "epoch": 4.7, "learning_rate": 0.0003829361085389911, "loss": 1.0946, "step": 116810 }, { "epoch": 4.7, "learning_rate": 0.0003829260437215669, "loss": 0.8727, "step": 116820 }, { "epoch": 4.7, "learning_rate": 0.0003829159789041427, "loss": 1.0455, "step": 116830 }, { "epoch": 4.7, "learning_rate": 0.0003829059140867185, "loss": 1.0452, "step": 116840 }, { "epoch": 4.7, "learning_rate": 0.0003828958492692943, "loss": 0.9518, "step": 116850 }, { "epoch": 4.7, "learning_rate": 0.00038288578445187007, "loss": 1.0569, "step": 116860 }, { "epoch": 4.7, "learning_rate": 0.00038287571963444586, "loss": 1.1685, "step": 116870 }, { "epoch": 4.7, "learning_rate": 0.0003828656548170216, "loss": 1.007, "step": 116880 }, { "epoch": 4.7, "learning_rate": 0.0003828555899995974, "loss": 0.988, "step": 116890 }, { "epoch": 4.7, "learning_rate": 0.0003828455251821732, "loss": 0.9299, "step": 116900 }, { "epoch": 4.7, "learning_rate": 0.00038283546036474897, "loss": 0.9466, "step": 116910 }, { "epoch": 4.7, "learning_rate": 0.00038282539554732476, "loss": 1.1598, "step": 116920 }, { "epoch": 4.7, "learning_rate": 0.00038281533072990055, "loss": 1.2969, "step": 116930 }, { "epoch": 4.7, "learning_rate": 0.00038280526591247634, "loss": 1.0197, "step": 116940 }, { "epoch": 4.7, "learning_rate": 0.0003827952010950522, "loss": 1.0195, "step": 116950 }, { "epoch": 4.7, "learning_rate": 0.000382785136277628, "loss": 1.1993, "step": 116960 }, { "epoch": 4.7, "learning_rate": 0.0003827750714602037, "loss": 1.0579, "step": 116970 }, { "epoch": 4.7, "learning_rate": 0.0003827650066427795, "loss": 0.9798, "step": 116980 }, { "epoch": 4.71, "learning_rate": 0.0003827549418253553, "loss": 1.1509, "step": 116990 }, { "epoch": 4.71, "learning_rate": 0.0003827448770079311, "loss": 0.9927, "step": 117000 }, { "epoch": 4.71, "learning_rate": 0.0003827348121905069, "loss": 1.2082, "step": 117010 }, { "epoch": 4.71, "learning_rate": 0.00038272474737308267, "loss": 1.1382, "step": 117020 }, { "epoch": 4.71, "learning_rate": 0.00038271468255565846, "loss": 1.0169, "step": 117030 }, { "epoch": 4.71, "learning_rate": 0.00038270461773823425, "loss": 0.9637, "step": 117040 }, { "epoch": 4.71, "learning_rate": 0.00038269455292081, "loss": 0.9015, "step": 117050 }, { "epoch": 4.71, "learning_rate": 0.0003826844881033858, "loss": 1.2351, "step": 117060 }, { "epoch": 4.71, "learning_rate": 0.00038267442328596157, "loss": 1.0447, "step": 117070 }, { "epoch": 4.71, "learning_rate": 0.00038266435846853736, "loss": 1.0202, "step": 117080 }, { "epoch": 4.71, "learning_rate": 0.00038265429365111315, "loss": 1.0298, "step": 117090 }, { "epoch": 4.71, "learning_rate": 0.000382644228833689, "loss": 1.1539, "step": 117100 }, { "epoch": 4.71, "learning_rate": 0.0003826341640162648, "loss": 1.1284, "step": 117110 }, { "epoch": 4.71, "learning_rate": 0.0003826240991988406, "loss": 1.1233, "step": 117120 }, { "epoch": 4.71, "learning_rate": 0.00038261403438141637, "loss": 0.9331, "step": 117130 }, { "epoch": 4.71, "learning_rate": 0.0003826039695639921, "loss": 1.0712, "step": 117140 }, { "epoch": 4.71, "learning_rate": 0.0003825939047465679, "loss": 1.0528, "step": 117150 }, { "epoch": 4.71, "learning_rate": 0.0003825838399291437, "loss": 0.9917, "step": 117160 }, { "epoch": 4.71, "learning_rate": 0.0003825737751117195, "loss": 1.3144, "step": 117170 }, { "epoch": 4.71, "learning_rate": 0.00038256371029429527, "loss": 1.0638, "step": 117180 }, { "epoch": 4.71, "learning_rate": 0.00038255364547687106, "loss": 1.0936, "step": 117190 }, { "epoch": 4.71, "learning_rate": 0.00038254358065944685, "loss": 0.9543, "step": 117200 }, { "epoch": 4.71, "learning_rate": 0.00038253351584202264, "loss": 1.0667, "step": 117210 }, { "epoch": 4.71, "learning_rate": 0.0003825234510245984, "loss": 1.1291, "step": 117220 }, { "epoch": 4.71, "learning_rate": 0.00038251338620717417, "loss": 1.1634, "step": 117230 }, { "epoch": 4.72, "learning_rate": 0.00038250332138975, "loss": 1.0377, "step": 117240 }, { "epoch": 4.72, "learning_rate": 0.0003824932565723258, "loss": 0.9055, "step": 117250 }, { "epoch": 4.72, "learning_rate": 0.0003824831917549016, "loss": 1.0962, "step": 117260 }, { "epoch": 4.72, "learning_rate": 0.0003824731269374774, "loss": 1.0205, "step": 117270 }, { "epoch": 4.72, "learning_rate": 0.0003824630621200532, "loss": 1.0032, "step": 117280 }, { "epoch": 4.72, "learning_rate": 0.00038245299730262897, "loss": 1.1233, "step": 117290 }, { "epoch": 4.72, "learning_rate": 0.00038244293248520476, "loss": 1.3673, "step": 117300 }, { "epoch": 4.72, "learning_rate": 0.0003824328676677805, "loss": 1.3755, "step": 117310 }, { "epoch": 4.72, "learning_rate": 0.0003824228028503563, "loss": 0.87, "step": 117320 }, { "epoch": 4.72, "learning_rate": 0.0003824127380329321, "loss": 1.223, "step": 117330 }, { "epoch": 4.72, "learning_rate": 0.00038240267321550787, "loss": 1.1707, "step": 117340 }, { "epoch": 4.72, "learning_rate": 0.00038239260839808366, "loss": 0.9099, "step": 117350 }, { "epoch": 4.72, "learning_rate": 0.00038238254358065945, "loss": 0.961, "step": 117360 }, { "epoch": 4.72, "learning_rate": 0.00038237247876323524, "loss": 1.0702, "step": 117370 }, { "epoch": 4.72, "learning_rate": 0.00038236241394581103, "loss": 0.9515, "step": 117380 }, { "epoch": 4.72, "learning_rate": 0.0003823523491283868, "loss": 0.9828, "step": 117390 }, { "epoch": 4.72, "learning_rate": 0.0003823422843109626, "loss": 1.1247, "step": 117400 }, { "epoch": 4.72, "learning_rate": 0.0003823322194935384, "loss": 1.1672, "step": 117410 }, { "epoch": 4.72, "learning_rate": 0.0003823221546761142, "loss": 1.0082, "step": 117420 }, { "epoch": 4.72, "learning_rate": 0.00038231208985869, "loss": 1.124, "step": 117430 }, { "epoch": 4.72, "learning_rate": 0.0003823020250412658, "loss": 1.1217, "step": 117440 }, { "epoch": 4.72, "learning_rate": 0.00038229196022384156, "loss": 1.1481, "step": 117450 }, { "epoch": 4.72, "learning_rate": 0.00038228189540641735, "loss": 1.148, "step": 117460 }, { "epoch": 4.72, "learning_rate": 0.00038227183058899315, "loss": 1.3663, "step": 117470 }, { "epoch": 4.72, "learning_rate": 0.0003822617657715689, "loss": 1.0465, "step": 117480 }, { "epoch": 4.73, "learning_rate": 0.00038225170095414467, "loss": 0.999, "step": 117490 }, { "epoch": 4.73, "learning_rate": 0.00038224163613672046, "loss": 0.865, "step": 117500 }, { "epoch": 4.73, "learning_rate": 0.00038223157131929625, "loss": 0.9791, "step": 117510 }, { "epoch": 4.73, "learning_rate": 0.00038222150650187205, "loss": 0.9658, "step": 117520 }, { "epoch": 4.73, "learning_rate": 0.00038221144168444784, "loss": 1.0608, "step": 117530 }, { "epoch": 4.73, "learning_rate": 0.0003822013768670237, "loss": 1.0453, "step": 117540 }, { "epoch": 4.73, "learning_rate": 0.00038219131204959947, "loss": 1.0989, "step": 117550 }, { "epoch": 4.73, "learning_rate": 0.0003821812472321752, "loss": 1.0708, "step": 117560 }, { "epoch": 4.73, "learning_rate": 0.000382171182414751, "loss": 1.118, "step": 117570 }, { "epoch": 4.73, "learning_rate": 0.0003821611175973268, "loss": 0.9642, "step": 117580 }, { "epoch": 4.73, "learning_rate": 0.0003821510527799026, "loss": 1.1207, "step": 117590 }, { "epoch": 4.73, "learning_rate": 0.00038214098796247837, "loss": 0.9673, "step": 117600 }, { "epoch": 4.73, "learning_rate": 0.00038213092314505416, "loss": 1.1237, "step": 117610 }, { "epoch": 4.73, "learning_rate": 0.00038212085832762995, "loss": 1.3069, "step": 117620 }, { "epoch": 4.73, "learning_rate": 0.00038211079351020574, "loss": 0.911, "step": 117630 }, { "epoch": 4.73, "learning_rate": 0.0003821007286927815, "loss": 1.0553, "step": 117640 }, { "epoch": 4.73, "learning_rate": 0.00038209066387535727, "loss": 1.2772, "step": 117650 }, { "epoch": 4.73, "learning_rate": 0.00038208059905793306, "loss": 1.2015, "step": 117660 }, { "epoch": 4.73, "learning_rate": 0.00038207053424050885, "loss": 0.8271, "step": 117670 }, { "epoch": 4.73, "learning_rate": 0.0003820604694230847, "loss": 0.9876, "step": 117680 }, { "epoch": 4.73, "learning_rate": 0.0003820504046056605, "loss": 1.0175, "step": 117690 }, { "epoch": 4.73, "learning_rate": 0.0003820403397882363, "loss": 1.0972, "step": 117700 }, { "epoch": 4.73, "learning_rate": 0.00038203027497081207, "loss": 1.0032, "step": 117710 }, { "epoch": 4.73, "learning_rate": 0.00038202021015338786, "loss": 1.1024, "step": 117720 }, { "epoch": 4.73, "learning_rate": 0.0003820101453359636, "loss": 1.074, "step": 117730 }, { "epoch": 4.74, "learning_rate": 0.0003820000805185394, "loss": 1.2609, "step": 117740 }, { "epoch": 4.74, "learning_rate": 0.0003819900157011152, "loss": 1.1199, "step": 117750 }, { "epoch": 4.74, "learning_rate": 0.00038197995088369097, "loss": 0.9216, "step": 117760 }, { "epoch": 4.74, "learning_rate": 0.00038196988606626676, "loss": 1.1402, "step": 117770 }, { "epoch": 4.74, "learning_rate": 0.00038195982124884255, "loss": 1.085, "step": 117780 }, { "epoch": 4.74, "learning_rate": 0.00038194975643141834, "loss": 1.2203, "step": 117790 }, { "epoch": 4.74, "learning_rate": 0.00038193969161399413, "loss": 1.0199, "step": 117800 }, { "epoch": 4.74, "learning_rate": 0.00038192962679656987, "loss": 1.0199, "step": 117810 }, { "epoch": 4.74, "learning_rate": 0.00038191956197914566, "loss": 0.8708, "step": 117820 }, { "epoch": 4.74, "learning_rate": 0.0003819094971617215, "loss": 1.1552, "step": 117830 }, { "epoch": 4.74, "learning_rate": 0.0003818994323442973, "loss": 1.0798, "step": 117840 }, { "epoch": 4.74, "learning_rate": 0.0003818893675268731, "loss": 0.874, "step": 117850 }, { "epoch": 4.74, "learning_rate": 0.0003818793027094489, "loss": 1.0111, "step": 117860 }, { "epoch": 4.74, "learning_rate": 0.00038186923789202467, "loss": 1.0109, "step": 117870 }, { "epoch": 4.74, "learning_rate": 0.00038185917307460046, "loss": 1.069, "step": 117880 }, { "epoch": 4.74, "learning_rate": 0.00038184910825717625, "loss": 1.0224, "step": 117890 }, { "epoch": 4.74, "learning_rate": 0.000381839043439752, "loss": 1.0718, "step": 117900 }, { "epoch": 4.74, "learning_rate": 0.0003818289786223278, "loss": 1.2749, "step": 117910 }, { "epoch": 4.74, "learning_rate": 0.00038181891380490357, "loss": 1.1771, "step": 117920 }, { "epoch": 4.74, "learning_rate": 0.00038180884898747936, "loss": 0.9917, "step": 117930 }, { "epoch": 4.74, "learning_rate": 0.00038179878417005515, "loss": 1.0228, "step": 117940 }, { "epoch": 4.74, "learning_rate": 0.00038178871935263094, "loss": 1.2826, "step": 117950 }, { "epoch": 4.74, "learning_rate": 0.00038177865453520673, "loss": 1.2132, "step": 117960 }, { "epoch": 4.74, "learning_rate": 0.0003817685897177826, "loss": 1.02, "step": 117970 }, { "epoch": 4.74, "learning_rate": 0.0003817585249003583, "loss": 1.0899, "step": 117980 }, { "epoch": 4.75, "learning_rate": 0.0003817484600829341, "loss": 0.8075, "step": 117990 }, { "epoch": 4.75, "learning_rate": 0.0003817383952655099, "loss": 1.1402, "step": 118000 }, { "epoch": 4.75, "learning_rate": 0.0003817283304480857, "loss": 1.4627, "step": 118010 }, { "epoch": 4.75, "learning_rate": 0.0003817182656306615, "loss": 0.9646, "step": 118020 }, { "epoch": 4.75, "learning_rate": 0.00038170820081323727, "loss": 0.8831, "step": 118030 }, { "epoch": 4.75, "learning_rate": 0.00038169813599581306, "loss": 0.8106, "step": 118040 }, { "epoch": 4.75, "learning_rate": 0.00038168807117838885, "loss": 1.0003, "step": 118050 }, { "epoch": 4.75, "learning_rate": 0.00038167800636096464, "loss": 1.0837, "step": 118060 }, { "epoch": 4.75, "learning_rate": 0.0003816679415435404, "loss": 0.9104, "step": 118070 }, { "epoch": 4.75, "learning_rate": 0.00038165787672611617, "loss": 1.0813, "step": 118080 }, { "epoch": 4.75, "learning_rate": 0.00038164781190869196, "loss": 1.1651, "step": 118090 }, { "epoch": 4.75, "learning_rate": 0.00038163774709126775, "loss": 1.0193, "step": 118100 }, { "epoch": 4.75, "learning_rate": 0.00038162768227384354, "loss": 1.119, "step": 118110 }, { "epoch": 4.75, "learning_rate": 0.0003816176174564194, "loss": 1.1333, "step": 118120 }, { "epoch": 4.75, "learning_rate": 0.0003816075526389952, "loss": 1.1384, "step": 118130 }, { "epoch": 4.75, "learning_rate": 0.00038159748782157097, "loss": 1.2143, "step": 118140 }, { "epoch": 4.75, "learning_rate": 0.0003815874230041467, "loss": 0.9646, "step": 118150 }, { "epoch": 4.75, "learning_rate": 0.0003815773581867225, "loss": 0.9396, "step": 118160 }, { "epoch": 4.75, "learning_rate": 0.0003815672933692983, "loss": 1.1615, "step": 118170 }, { "epoch": 4.75, "learning_rate": 0.0003815572285518741, "loss": 1.1119, "step": 118180 }, { "epoch": 4.75, "learning_rate": 0.00038154716373444987, "loss": 1.1668, "step": 118190 }, { "epoch": 4.75, "learning_rate": 0.00038153709891702566, "loss": 0.9552, "step": 118200 }, { "epoch": 4.75, "learning_rate": 0.00038152703409960145, "loss": 1.1444, "step": 118210 }, { "epoch": 4.75, "learning_rate": 0.00038151696928217724, "loss": 1.0503, "step": 118220 }, { "epoch": 4.76, "learning_rate": 0.00038150690446475303, "loss": 1.053, "step": 118230 }, { "epoch": 4.76, "learning_rate": 0.00038149683964732876, "loss": 1.0376, "step": 118240 }, { "epoch": 4.76, "learning_rate": 0.00038148677482990456, "loss": 0.9409, "step": 118250 }, { "epoch": 4.76, "learning_rate": 0.00038147671001248035, "loss": 1.1633, "step": 118260 }, { "epoch": 4.76, "learning_rate": 0.0003814666451950562, "loss": 1.1541, "step": 118270 }, { "epoch": 4.76, "learning_rate": 0.000381456580377632, "loss": 1.0958, "step": 118280 }, { "epoch": 4.76, "learning_rate": 0.00038144651556020777, "loss": 1.108, "step": 118290 }, { "epoch": 4.76, "learning_rate": 0.00038143645074278356, "loss": 1.2352, "step": 118300 }, { "epoch": 4.76, "learning_rate": 0.00038142638592535935, "loss": 1.0474, "step": 118310 }, { "epoch": 4.76, "learning_rate": 0.0003814163211079351, "loss": 0.9973, "step": 118320 }, { "epoch": 4.76, "learning_rate": 0.0003814062562905109, "loss": 1.1601, "step": 118330 }, { "epoch": 4.76, "learning_rate": 0.00038139619147308667, "loss": 1.0228, "step": 118340 }, { "epoch": 4.76, "learning_rate": 0.00038138612665566246, "loss": 1.1003, "step": 118350 }, { "epoch": 4.76, "learning_rate": 0.00038137606183823825, "loss": 0.8538, "step": 118360 }, { "epoch": 4.76, "learning_rate": 0.00038136599702081404, "loss": 0.8584, "step": 118370 }, { "epoch": 4.76, "learning_rate": 0.00038135593220338984, "loss": 1.0059, "step": 118380 }, { "epoch": 4.76, "learning_rate": 0.0003813458673859656, "loss": 1.0775, "step": 118390 }, { "epoch": 4.76, "learning_rate": 0.00038133580256854136, "loss": 0.9877, "step": 118400 }, { "epoch": 4.76, "learning_rate": 0.0003813257377511172, "loss": 0.9866, "step": 118410 }, { "epoch": 4.76, "learning_rate": 0.000381315672933693, "loss": 1.1111, "step": 118420 }, { "epoch": 4.76, "learning_rate": 0.0003813056081162688, "loss": 1.1224, "step": 118430 }, { "epoch": 4.76, "learning_rate": 0.0003812955432988446, "loss": 0.9747, "step": 118440 }, { "epoch": 4.76, "learning_rate": 0.00038128547848142037, "loss": 1.0856, "step": 118450 }, { "epoch": 4.76, "learning_rate": 0.00038127541366399616, "loss": 1.0138, "step": 118460 }, { "epoch": 4.76, "learning_rate": 0.00038126534884657195, "loss": 0.8961, "step": 118470 }, { "epoch": 4.77, "learning_rate": 0.00038125528402914774, "loss": 1.1427, "step": 118480 }, { "epoch": 4.77, "learning_rate": 0.0003812452192117235, "loss": 0.8301, "step": 118490 }, { "epoch": 4.77, "learning_rate": 0.00038123515439429927, "loss": 1.0627, "step": 118500 }, { "epoch": 4.77, "learning_rate": 0.00038122508957687506, "loss": 1.0569, "step": 118510 }, { "epoch": 4.77, "learning_rate": 0.00038121502475945085, "loss": 0.8961, "step": 118520 }, { "epoch": 4.77, "learning_rate": 0.00038120495994202664, "loss": 1.0803, "step": 118530 }, { "epoch": 4.77, "learning_rate": 0.00038119489512460243, "loss": 1.0308, "step": 118540 }, { "epoch": 4.77, "learning_rate": 0.0003811848303071782, "loss": 0.95, "step": 118550 }, { "epoch": 4.77, "learning_rate": 0.00038117476548975407, "loss": 1.1399, "step": 118560 }, { "epoch": 4.77, "learning_rate": 0.00038116470067232986, "loss": 0.93, "step": 118570 }, { "epoch": 4.77, "learning_rate": 0.0003811546358549056, "loss": 1.0967, "step": 118580 }, { "epoch": 4.77, "learning_rate": 0.0003811445710374814, "loss": 1.0138, "step": 118590 }, { "epoch": 4.77, "learning_rate": 0.0003811345062200572, "loss": 1.1015, "step": 118600 }, { "epoch": 4.77, "learning_rate": 0.00038112444140263297, "loss": 1.1866, "step": 118610 }, { "epoch": 4.77, "learning_rate": 0.00038111437658520876, "loss": 1.0301, "step": 118620 }, { "epoch": 4.77, "learning_rate": 0.00038110431176778455, "loss": 1.1632, "step": 118630 }, { "epoch": 4.77, "learning_rate": 0.00038109424695036034, "loss": 0.9633, "step": 118640 }, { "epoch": 4.77, "learning_rate": 0.00038108418213293613, "loss": 0.8538, "step": 118650 }, { "epoch": 4.77, "learning_rate": 0.00038107411731551187, "loss": 0.995, "step": 118660 }, { "epoch": 4.77, "learning_rate": 0.00038106405249808766, "loss": 1.3446, "step": 118670 }, { "epoch": 4.77, "learning_rate": 0.00038105398768066345, "loss": 0.8962, "step": 118680 }, { "epoch": 4.77, "learning_rate": 0.00038104392286323924, "loss": 1.0565, "step": 118690 }, { "epoch": 4.77, "learning_rate": 0.00038103385804581503, "loss": 1.2828, "step": 118700 }, { "epoch": 4.77, "learning_rate": 0.0003810237932283909, "loss": 0.9782, "step": 118710 }, { "epoch": 4.77, "learning_rate": 0.00038101372841096667, "loss": 0.9063, "step": 118720 }, { "epoch": 4.78, "learning_rate": 0.00038100366359354246, "loss": 1.155, "step": 118730 }, { "epoch": 4.78, "learning_rate": 0.0003809935987761182, "loss": 1.0094, "step": 118740 }, { "epoch": 4.78, "learning_rate": 0.000380983533958694, "loss": 1.0164, "step": 118750 }, { "epoch": 4.78, "learning_rate": 0.0003809734691412698, "loss": 1.002, "step": 118760 }, { "epoch": 4.78, "learning_rate": 0.00038096340432384557, "loss": 0.8699, "step": 118770 }, { "epoch": 4.78, "learning_rate": 0.00038095333950642136, "loss": 1.0175, "step": 118780 }, { "epoch": 4.78, "learning_rate": 0.00038094327468899715, "loss": 0.7648, "step": 118790 }, { "epoch": 4.78, "learning_rate": 0.00038093320987157294, "loss": 1.0123, "step": 118800 }, { "epoch": 4.78, "learning_rate": 0.00038092314505414873, "loss": 1.3168, "step": 118810 }, { "epoch": 4.78, "learning_rate": 0.0003809130802367245, "loss": 1.0147, "step": 118820 }, { "epoch": 4.78, "learning_rate": 0.00038090301541930026, "loss": 1.0505, "step": 118830 }, { "epoch": 4.78, "learning_rate": 0.00038089295060187605, "loss": 1.1094, "step": 118840 }, { "epoch": 4.78, "learning_rate": 0.0003808828857844519, "loss": 1.0559, "step": 118850 }, { "epoch": 4.78, "learning_rate": 0.0003808728209670277, "loss": 1.1454, "step": 118860 }, { "epoch": 4.78, "learning_rate": 0.0003808627561496035, "loss": 1.0302, "step": 118870 }, { "epoch": 4.78, "learning_rate": 0.00038085269133217927, "loss": 1.0505, "step": 118880 }, { "epoch": 4.78, "learning_rate": 0.00038084262651475506, "loss": 1.0872, "step": 118890 }, { "epoch": 4.78, "learning_rate": 0.00038083256169733085, "loss": 1.318, "step": 118900 }, { "epoch": 4.78, "learning_rate": 0.0003808224968799066, "loss": 1.2812, "step": 118910 }, { "epoch": 4.78, "learning_rate": 0.0003808124320624824, "loss": 1.2927, "step": 118920 }, { "epoch": 4.78, "learning_rate": 0.00038080236724505817, "loss": 0.8869, "step": 118930 }, { "epoch": 4.78, "learning_rate": 0.00038079230242763396, "loss": 1.1222, "step": 118940 }, { "epoch": 4.78, "learning_rate": 0.00038078223761020975, "loss": 1.1922, "step": 118950 }, { "epoch": 4.78, "learning_rate": 0.00038077217279278554, "loss": 1.08, "step": 118960 }, { "epoch": 4.78, "learning_rate": 0.00038076210797536133, "loss": 1.2479, "step": 118970 }, { "epoch": 4.79, "learning_rate": 0.0003807520431579371, "loss": 1.3854, "step": 118980 }, { "epoch": 4.79, "learning_rate": 0.0003807419783405129, "loss": 0.9193, "step": 118990 }, { "epoch": 4.79, "learning_rate": 0.0003807319135230887, "loss": 1.1033, "step": 119000 }, { "epoch": 4.79, "learning_rate": 0.0003807218487056645, "loss": 1.1448, "step": 119010 }, { "epoch": 4.79, "learning_rate": 0.0003807117838882403, "loss": 0.9363, "step": 119020 }, { "epoch": 4.79, "learning_rate": 0.0003807017190708161, "loss": 1.1108, "step": 119030 }, { "epoch": 4.79, "learning_rate": 0.00038069165425339186, "loss": 0.862, "step": 119040 }, { "epoch": 4.79, "learning_rate": 0.00038068158943596766, "loss": 1.1089, "step": 119050 }, { "epoch": 4.79, "learning_rate": 0.00038067152461854345, "loss": 1.0607, "step": 119060 }, { "epoch": 4.79, "learning_rate": 0.00038066145980111924, "loss": 1.0395, "step": 119070 }, { "epoch": 4.79, "learning_rate": 0.000380651394983695, "loss": 1.0699, "step": 119080 }, { "epoch": 4.79, "learning_rate": 0.00038064133016627076, "loss": 0.8732, "step": 119090 }, { "epoch": 4.79, "learning_rate": 0.00038063126534884656, "loss": 1.1378, "step": 119100 }, { "epoch": 4.79, "learning_rate": 0.00038062120053142235, "loss": 1.2666, "step": 119110 }, { "epoch": 4.79, "learning_rate": 0.00038061113571399814, "loss": 0.8297, "step": 119120 }, { "epoch": 4.79, "learning_rate": 0.00038060107089657393, "loss": 1.144, "step": 119130 }, { "epoch": 4.79, "learning_rate": 0.00038059100607914977, "loss": 1.0095, "step": 119140 }, { "epoch": 4.79, "learning_rate": 0.00038058094126172556, "loss": 1.2531, "step": 119150 }, { "epoch": 4.79, "learning_rate": 0.00038057087644430135, "loss": 0.9876, "step": 119160 }, { "epoch": 4.79, "learning_rate": 0.0003805608116268771, "loss": 0.9858, "step": 119170 }, { "epoch": 4.79, "learning_rate": 0.0003805507468094529, "loss": 1.1428, "step": 119180 }, { "epoch": 4.79, "learning_rate": 0.00038054068199202867, "loss": 1.0787, "step": 119190 }, { "epoch": 4.79, "learning_rate": 0.00038053061717460446, "loss": 0.8476, "step": 119200 }, { "epoch": 4.79, "learning_rate": 0.00038052055235718025, "loss": 1.116, "step": 119210 }, { "epoch": 4.79, "learning_rate": 0.00038051048753975604, "loss": 1.2691, "step": 119220 }, { "epoch": 4.8, "learning_rate": 0.00038050042272233184, "loss": 1.1348, "step": 119230 }, { "epoch": 4.8, "learning_rate": 0.0003804903579049076, "loss": 1.0909, "step": 119240 }, { "epoch": 4.8, "learning_rate": 0.00038048029308748336, "loss": 1.1397, "step": 119250 }, { "epoch": 4.8, "learning_rate": 0.00038047022827005915, "loss": 0.9557, "step": 119260 }, { "epoch": 4.8, "learning_rate": 0.00038046016345263494, "loss": 1.1332, "step": 119270 }, { "epoch": 4.8, "learning_rate": 0.00038045009863521074, "loss": 0.9149, "step": 119280 }, { "epoch": 4.8, "learning_rate": 0.0003804400338177866, "loss": 0.8876, "step": 119290 }, { "epoch": 4.8, "learning_rate": 0.00038042996900036237, "loss": 1.2441, "step": 119300 }, { "epoch": 4.8, "learning_rate": 0.00038041990418293816, "loss": 1.096, "step": 119310 }, { "epoch": 4.8, "learning_rate": 0.00038040983936551395, "loss": 1.1819, "step": 119320 }, { "epoch": 4.8, "learning_rate": 0.00038039977454808974, "loss": 1.0836, "step": 119330 }, { "epoch": 4.8, "learning_rate": 0.0003803897097306655, "loss": 1.0857, "step": 119340 }, { "epoch": 4.8, "learning_rate": 0.00038037964491324127, "loss": 1.0735, "step": 119350 }, { "epoch": 4.8, "learning_rate": 0.00038036958009581706, "loss": 1.0632, "step": 119360 }, { "epoch": 4.8, "learning_rate": 0.00038035951527839285, "loss": 1.0294, "step": 119370 }, { "epoch": 4.8, "learning_rate": 0.00038034945046096864, "loss": 0.912, "step": 119380 }, { "epoch": 4.8, "learning_rate": 0.00038033938564354443, "loss": 0.8554, "step": 119390 }, { "epoch": 4.8, "learning_rate": 0.0003803293208261202, "loss": 0.8145, "step": 119400 }, { "epoch": 4.8, "learning_rate": 0.000380319256008696, "loss": 0.8505, "step": 119410 }, { "epoch": 4.8, "learning_rate": 0.00038030919119127175, "loss": 0.9612, "step": 119420 }, { "epoch": 4.8, "learning_rate": 0.00038029912637384754, "loss": 1.063, "step": 119430 }, { "epoch": 4.8, "learning_rate": 0.0003802890615564234, "loss": 1.0299, "step": 119440 }, { "epoch": 4.8, "learning_rate": 0.0003802789967389992, "loss": 0.9329, "step": 119450 }, { "epoch": 4.8, "learning_rate": 0.00038026893192157497, "loss": 1.0187, "step": 119460 }, { "epoch": 4.8, "learning_rate": 0.00038025886710415076, "loss": 0.9404, "step": 119470 }, { "epoch": 4.81, "learning_rate": 0.00038024880228672655, "loss": 0.9817, "step": 119480 }, { "epoch": 4.81, "learning_rate": 0.00038023873746930234, "loss": 1.0029, "step": 119490 }, { "epoch": 4.81, "learning_rate": 0.0003802286726518781, "loss": 1.1154, "step": 119500 }, { "epoch": 4.81, "learning_rate": 0.00038021860783445387, "loss": 0.9646, "step": 119510 }, { "epoch": 4.81, "learning_rate": 0.00038020854301702966, "loss": 1.0038, "step": 119520 }, { "epoch": 4.81, "learning_rate": 0.00038019847819960545, "loss": 1.084, "step": 119530 }, { "epoch": 4.81, "learning_rate": 0.00038018841338218124, "loss": 0.9632, "step": 119540 }, { "epoch": 4.81, "learning_rate": 0.00038017834856475703, "loss": 0.8424, "step": 119550 }, { "epoch": 4.81, "learning_rate": 0.0003801682837473328, "loss": 0.97, "step": 119560 }, { "epoch": 4.81, "learning_rate": 0.0003801582189299086, "loss": 1.0867, "step": 119570 }, { "epoch": 4.81, "learning_rate": 0.00038014815411248446, "loss": 1.0791, "step": 119580 }, { "epoch": 4.81, "learning_rate": 0.0003801380892950602, "loss": 1.2076, "step": 119590 }, { "epoch": 4.81, "learning_rate": 0.000380128024477636, "loss": 1.1368, "step": 119600 }, { "epoch": 4.81, "learning_rate": 0.0003801179596602118, "loss": 1.0404, "step": 119610 }, { "epoch": 4.81, "learning_rate": 0.00038010789484278757, "loss": 1.0655, "step": 119620 }, { "epoch": 4.81, "learning_rate": 0.00038009783002536336, "loss": 1.1875, "step": 119630 }, { "epoch": 4.81, "learning_rate": 0.00038008776520793915, "loss": 0.9959, "step": 119640 }, { "epoch": 4.81, "learning_rate": 0.00038007770039051494, "loss": 1.2693, "step": 119650 }, { "epoch": 4.81, "learning_rate": 0.00038006763557309073, "loss": 0.9234, "step": 119660 }, { "epoch": 4.81, "learning_rate": 0.00038005757075566647, "loss": 1.1705, "step": 119670 }, { "epoch": 4.81, "learning_rate": 0.00038004750593824226, "loss": 1.03, "step": 119680 }, { "epoch": 4.81, "learning_rate": 0.00038003744112081805, "loss": 1.0346, "step": 119690 }, { "epoch": 4.81, "learning_rate": 0.00038002737630339384, "loss": 0.9675, "step": 119700 }, { "epoch": 4.81, "learning_rate": 0.00038001731148596963, "loss": 1.0314, "step": 119710 }, { "epoch": 4.81, "learning_rate": 0.0003800072466685454, "loss": 1.1512, "step": 119720 }, { "epoch": 4.82, "learning_rate": 0.00037999718185112127, "loss": 1.0482, "step": 119730 }, { "epoch": 4.82, "learning_rate": 0.00037998711703369706, "loss": 1.2004, "step": 119740 }, { "epoch": 4.82, "learning_rate": 0.00037997705221627285, "loss": 1.1735, "step": 119750 }, { "epoch": 4.82, "learning_rate": 0.0003799669873988486, "loss": 1.2164, "step": 119760 }, { "epoch": 4.82, "learning_rate": 0.0003799569225814244, "loss": 0.9296, "step": 119770 }, { "epoch": 4.82, "learning_rate": 0.00037994685776400017, "loss": 1.0043, "step": 119780 }, { "epoch": 4.82, "learning_rate": 0.00037993679294657596, "loss": 1.0955, "step": 119790 }, { "epoch": 4.82, "learning_rate": 0.00037992672812915175, "loss": 1.0888, "step": 119800 }, { "epoch": 4.82, "learning_rate": 0.00037991666331172754, "loss": 1.1337, "step": 119810 }, { "epoch": 4.82, "learning_rate": 0.00037990659849430333, "loss": 1.0928, "step": 119820 }, { "epoch": 4.82, "learning_rate": 0.0003798965336768791, "loss": 1.0229, "step": 119830 }, { "epoch": 4.82, "learning_rate": 0.00037988646885945486, "loss": 0.9314, "step": 119840 }, { "epoch": 4.82, "learning_rate": 0.00037987640404203065, "loss": 1.2117, "step": 119850 }, { "epoch": 4.82, "learning_rate": 0.00037986633922460644, "loss": 1.1367, "step": 119860 }, { "epoch": 4.82, "learning_rate": 0.00037985627440718223, "loss": 0.9204, "step": 119870 }, { "epoch": 4.82, "learning_rate": 0.0003798462095897581, "loss": 1.1263, "step": 119880 }, { "epoch": 4.82, "learning_rate": 0.00037983614477233386, "loss": 1.0462, "step": 119890 }, { "epoch": 4.82, "learning_rate": 0.00037982607995490966, "loss": 1.0462, "step": 119900 }, { "epoch": 4.82, "learning_rate": 0.00037981601513748545, "loss": 0.9824, "step": 119910 }, { "epoch": 4.82, "learning_rate": 0.00037980595032006124, "loss": 1.1128, "step": 119920 }, { "epoch": 4.82, "learning_rate": 0.000379795885502637, "loss": 1.2492, "step": 119930 }, { "epoch": 4.82, "learning_rate": 0.00037978582068521276, "loss": 1.1384, "step": 119940 }, { "epoch": 4.82, "learning_rate": 0.00037977575586778856, "loss": 1.1858, "step": 119950 }, { "epoch": 4.82, "learning_rate": 0.00037976569105036435, "loss": 1.3317, "step": 119960 }, { "epoch": 4.82, "learning_rate": 0.00037975562623294014, "loss": 1.1044, "step": 119970 }, { "epoch": 4.83, "learning_rate": 0.00037974556141551593, "loss": 0.9446, "step": 119980 }, { "epoch": 4.83, "learning_rate": 0.0003797354965980917, "loss": 1.1088, "step": 119990 }, { "epoch": 4.83, "learning_rate": 0.0003797254317806675, "loss": 1.1363, "step": 120000 }, { "epoch": 4.83, "learning_rate": 0.00037971536696324325, "loss": 1.081, "step": 120010 }, { "epoch": 4.83, "learning_rate": 0.0003797053021458191, "loss": 1.2856, "step": 120020 }, { "epoch": 4.83, "learning_rate": 0.0003796952373283949, "loss": 1.17, "step": 120030 }, { "epoch": 4.83, "learning_rate": 0.00037968517251097067, "loss": 1.2061, "step": 120040 }, { "epoch": 4.83, "learning_rate": 0.00037967510769354646, "loss": 1.0784, "step": 120050 }, { "epoch": 4.83, "learning_rate": 0.00037966504287612225, "loss": 1.2963, "step": 120060 }, { "epoch": 4.83, "learning_rate": 0.00037965497805869804, "loss": 1.0906, "step": 120070 }, { "epoch": 4.83, "learning_rate": 0.00037964491324127384, "loss": 1.0645, "step": 120080 }, { "epoch": 4.83, "learning_rate": 0.0003796348484238496, "loss": 1.0222, "step": 120090 }, { "epoch": 4.83, "learning_rate": 0.00037962478360642536, "loss": 1.3294, "step": 120100 }, { "epoch": 4.83, "learning_rate": 0.00037961471878900115, "loss": 1.1977, "step": 120110 }, { "epoch": 4.83, "learning_rate": 0.00037960465397157694, "loss": 1.2115, "step": 120120 }, { "epoch": 4.83, "learning_rate": 0.00037959458915415273, "loss": 1.0165, "step": 120130 }, { "epoch": 4.83, "learning_rate": 0.0003795845243367285, "loss": 1.0253, "step": 120140 }, { "epoch": 4.83, "learning_rate": 0.0003795744595193043, "loss": 1.0053, "step": 120150 }, { "epoch": 4.83, "learning_rate": 0.0003795643947018801, "loss": 0.8989, "step": 120160 }, { "epoch": 4.83, "learning_rate": 0.00037955432988445595, "loss": 1.018, "step": 120170 }, { "epoch": 4.83, "learning_rate": 0.0003795442650670317, "loss": 0.8826, "step": 120180 }, { "epoch": 4.83, "learning_rate": 0.0003795342002496075, "loss": 1.0894, "step": 120190 }, { "epoch": 4.83, "learning_rate": 0.00037952413543218327, "loss": 1.0335, "step": 120200 }, { "epoch": 4.83, "learning_rate": 0.00037951407061475906, "loss": 1.0132, "step": 120210 }, { "epoch": 4.84, "learning_rate": 0.00037950400579733485, "loss": 1.1458, "step": 120220 }, { "epoch": 4.84, "learning_rate": 0.00037949394097991064, "loss": 0.9703, "step": 120230 }, { "epoch": 4.84, "learning_rate": 0.00037948387616248643, "loss": 1.0049, "step": 120240 }, { "epoch": 4.84, "learning_rate": 0.0003794738113450622, "loss": 1.1158, "step": 120250 }, { "epoch": 4.84, "learning_rate": 0.00037946374652763796, "loss": 1.0402, "step": 120260 }, { "epoch": 4.84, "learning_rate": 0.00037945368171021375, "loss": 1.2425, "step": 120270 }, { "epoch": 4.84, "learning_rate": 0.00037944361689278954, "loss": 1.0261, "step": 120280 }, { "epoch": 4.84, "learning_rate": 0.00037943355207536533, "loss": 0.9252, "step": 120290 }, { "epoch": 4.84, "learning_rate": 0.0003794234872579411, "loss": 1.1382, "step": 120300 }, { "epoch": 4.84, "learning_rate": 0.00037941342244051697, "loss": 1.0228, "step": 120310 }, { "epoch": 4.84, "learning_rate": 0.00037940335762309276, "loss": 1.2491, "step": 120320 }, { "epoch": 4.84, "learning_rate": 0.00037939329280566855, "loss": 1.0734, "step": 120330 }, { "epoch": 4.84, "learning_rate": 0.00037938322798824434, "loss": 1.1496, "step": 120340 }, { "epoch": 4.84, "learning_rate": 0.0003793731631708201, "loss": 1.0427, "step": 120350 }, { "epoch": 4.84, "learning_rate": 0.00037936309835339587, "loss": 1.1091, "step": 120360 }, { "epoch": 4.84, "learning_rate": 0.00037935303353597166, "loss": 1.1168, "step": 120370 }, { "epoch": 4.84, "learning_rate": 0.00037934296871854745, "loss": 1.2122, "step": 120380 }, { "epoch": 4.84, "learning_rate": 0.00037933290390112324, "loss": 1.1237, "step": 120390 }, { "epoch": 4.84, "learning_rate": 0.00037932283908369903, "loss": 1.1551, "step": 120400 }, { "epoch": 4.84, "learning_rate": 0.0003793127742662748, "loss": 1.0741, "step": 120410 }, { "epoch": 4.84, "learning_rate": 0.0003793027094488506, "loss": 1.0637, "step": 120420 }, { "epoch": 4.84, "learning_rate": 0.00037929264463142635, "loss": 1.3315, "step": 120430 }, { "epoch": 4.84, "learning_rate": 0.00037928257981400214, "loss": 1.2616, "step": 120440 }, { "epoch": 4.84, "learning_rate": 0.00037927251499657793, "loss": 0.9464, "step": 120450 }, { "epoch": 4.84, "learning_rate": 0.0003792624501791538, "loss": 1.0673, "step": 120460 }, { "epoch": 4.85, "learning_rate": 0.00037925238536172957, "loss": 1.2514, "step": 120470 }, { "epoch": 4.85, "learning_rate": 0.00037924232054430536, "loss": 1.1749, "step": 120480 }, { "epoch": 4.85, "learning_rate": 0.00037923225572688115, "loss": 1.0347, "step": 120490 }, { "epoch": 4.85, "learning_rate": 0.00037922219090945694, "loss": 1.0185, "step": 120500 }, { "epoch": 4.85, "learning_rate": 0.00037921212609203273, "loss": 1.1959, "step": 120510 }, { "epoch": 4.85, "learning_rate": 0.00037920206127460847, "loss": 0.9496, "step": 120520 }, { "epoch": 4.85, "learning_rate": 0.00037919199645718426, "loss": 1.0418, "step": 120530 }, { "epoch": 4.85, "learning_rate": 0.00037918193163976005, "loss": 0.9047, "step": 120540 }, { "epoch": 4.85, "learning_rate": 0.00037917186682233584, "loss": 0.8695, "step": 120550 }, { "epoch": 4.85, "learning_rate": 0.00037916180200491163, "loss": 1.1258, "step": 120560 }, { "epoch": 4.85, "learning_rate": 0.0003791517371874874, "loss": 0.8424, "step": 120570 }, { "epoch": 4.85, "learning_rate": 0.0003791416723700632, "loss": 1.1512, "step": 120580 }, { "epoch": 4.85, "learning_rate": 0.000379131607552639, "loss": 1.2445, "step": 120590 }, { "epoch": 4.85, "learning_rate": 0.00037912154273521474, "loss": 1.2898, "step": 120600 }, { "epoch": 4.85, "learning_rate": 0.0003791114779177906, "loss": 1.0639, "step": 120610 }, { "epoch": 4.85, "learning_rate": 0.0003791014131003664, "loss": 0.9587, "step": 120620 }, { "epoch": 4.85, "learning_rate": 0.00037909134828294217, "loss": 0.8425, "step": 120630 }, { "epoch": 4.85, "learning_rate": 0.00037908128346551796, "loss": 1.0334, "step": 120640 }, { "epoch": 4.85, "learning_rate": 0.00037907121864809375, "loss": 0.8251, "step": 120650 }, { "epoch": 4.85, "learning_rate": 0.00037906115383066954, "loss": 1.1348, "step": 120660 }, { "epoch": 4.85, "learning_rate": 0.00037905108901324533, "loss": 1.07, "step": 120670 }, { "epoch": 4.85, "learning_rate": 0.0003790410241958211, "loss": 0.8656, "step": 120680 }, { "epoch": 4.85, "learning_rate": 0.00037903095937839686, "loss": 0.9887, "step": 120690 }, { "epoch": 4.85, "learning_rate": 0.00037902089456097265, "loss": 1.5777, "step": 120700 }, { "epoch": 4.85, "learning_rate": 0.00037901082974354844, "loss": 0.8707, "step": 120710 }, { "epoch": 4.86, "learning_rate": 0.00037900076492612423, "loss": 1.1617, "step": 120720 }, { "epoch": 4.86, "learning_rate": 0.0003789907001087, "loss": 0.9526, "step": 120730 }, { "epoch": 4.86, "learning_rate": 0.0003789806352912758, "loss": 0.8868, "step": 120740 }, { "epoch": 4.86, "learning_rate": 0.00037897057047385166, "loss": 0.965, "step": 120750 }, { "epoch": 4.86, "learning_rate": 0.00037896050565642745, "loss": 1.1454, "step": 120760 }, { "epoch": 4.86, "learning_rate": 0.0003789504408390032, "loss": 1.3433, "step": 120770 }, { "epoch": 4.86, "learning_rate": 0.000378940376021579, "loss": 1.0276, "step": 120780 }, { "epoch": 4.86, "learning_rate": 0.00037893031120415476, "loss": 1.0459, "step": 120790 }, { "epoch": 4.86, "learning_rate": 0.00037892024638673055, "loss": 1.3359, "step": 120800 }, { "epoch": 4.86, "learning_rate": 0.00037891018156930635, "loss": 1.0158, "step": 120810 }, { "epoch": 4.86, "learning_rate": 0.00037890011675188214, "loss": 1.1009, "step": 120820 }, { "epoch": 4.86, "learning_rate": 0.00037889005193445793, "loss": 1.0751, "step": 120830 }, { "epoch": 4.86, "learning_rate": 0.0003788799871170337, "loss": 1.0006, "step": 120840 }, { "epoch": 4.86, "learning_rate": 0.0003788699222996095, "loss": 1.1261, "step": 120850 }, { "epoch": 4.86, "learning_rate": 0.00037885985748218525, "loss": 1.0372, "step": 120860 }, { "epoch": 4.86, "learning_rate": 0.00037884979266476104, "loss": 0.9595, "step": 120870 }, { "epoch": 4.86, "learning_rate": 0.0003788397278473368, "loss": 0.8933, "step": 120880 }, { "epoch": 4.86, "learning_rate": 0.0003788296630299126, "loss": 1.2022, "step": 120890 }, { "epoch": 4.86, "learning_rate": 0.00037881959821248846, "loss": 0.8756, "step": 120900 }, { "epoch": 4.86, "learning_rate": 0.00037880953339506425, "loss": 1.262, "step": 120910 }, { "epoch": 4.86, "learning_rate": 0.00037879946857764004, "loss": 0.9729, "step": 120920 }, { "epoch": 4.86, "learning_rate": 0.00037878940376021583, "loss": 1.0882, "step": 120930 }, { "epoch": 4.86, "learning_rate": 0.00037877933894279157, "loss": 1.2978, "step": 120940 }, { "epoch": 4.86, "learning_rate": 0.00037876927412536736, "loss": 1.1814, "step": 120950 }, { "epoch": 4.86, "learning_rate": 0.00037875920930794315, "loss": 0.9718, "step": 120960 }, { "epoch": 4.87, "learning_rate": 0.00037874914449051894, "loss": 0.8662, "step": 120970 }, { "epoch": 4.87, "learning_rate": 0.00037873907967309473, "loss": 1.0119, "step": 120980 }, { "epoch": 4.87, "learning_rate": 0.0003787290148556705, "loss": 1.0918, "step": 120990 }, { "epoch": 4.87, "learning_rate": 0.0003787189500382463, "loss": 0.9788, "step": 121000 }, { "epoch": 4.87, "learning_rate": 0.0003787088852208221, "loss": 1.1173, "step": 121010 }, { "epoch": 4.87, "learning_rate": 0.00037869882040339784, "loss": 1.1312, "step": 121020 }, { "epoch": 4.87, "learning_rate": 0.00037868875558597363, "loss": 1.1974, "step": 121030 }, { "epoch": 4.87, "learning_rate": 0.0003786786907685494, "loss": 0.9969, "step": 121040 }, { "epoch": 4.87, "learning_rate": 0.00037866862595112527, "loss": 1.0598, "step": 121050 }, { "epoch": 4.87, "learning_rate": 0.00037865856113370106, "loss": 1.0601, "step": 121060 }, { "epoch": 4.87, "learning_rate": 0.00037864849631627685, "loss": 1.1493, "step": 121070 }, { "epoch": 4.87, "learning_rate": 0.00037863843149885264, "loss": 1.3045, "step": 121080 }, { "epoch": 4.87, "learning_rate": 0.00037862836668142843, "loss": 0.9868, "step": 121090 }, { "epoch": 4.87, "learning_rate": 0.0003786183018640042, "loss": 1.0794, "step": 121100 }, { "epoch": 4.87, "learning_rate": 0.00037860823704657996, "loss": 1.0791, "step": 121110 }, { "epoch": 4.87, "learning_rate": 0.00037859817222915575, "loss": 1.1277, "step": 121120 }, { "epoch": 4.87, "learning_rate": 0.00037858810741173154, "loss": 0.862, "step": 121130 }, { "epoch": 4.87, "learning_rate": 0.00037857804259430733, "loss": 1.1986, "step": 121140 }, { "epoch": 4.87, "learning_rate": 0.0003785679777768831, "loss": 1.2529, "step": 121150 }, { "epoch": 4.87, "learning_rate": 0.0003785579129594589, "loss": 0.954, "step": 121160 }, { "epoch": 4.87, "learning_rate": 0.0003785478481420347, "loss": 0.9675, "step": 121170 }, { "epoch": 4.87, "learning_rate": 0.0003785377833246105, "loss": 0.8681, "step": 121180 }, { "epoch": 4.87, "learning_rate": 0.0003785277185071863, "loss": 1.0989, "step": 121190 }, { "epoch": 4.87, "learning_rate": 0.0003785176536897621, "loss": 1.1175, "step": 121200 }, { "epoch": 4.87, "learning_rate": 0.00037850758887233787, "loss": 1.1836, "step": 121210 }, { "epoch": 4.88, "learning_rate": 0.00037849752405491366, "loss": 1.2337, "step": 121220 }, { "epoch": 4.88, "learning_rate": 0.00037848745923748945, "loss": 1.1428, "step": 121230 }, { "epoch": 4.88, "learning_rate": 0.00037847739442006524, "loss": 0.946, "step": 121240 }, { "epoch": 4.88, "learning_rate": 0.00037846732960264103, "loss": 0.9104, "step": 121250 }, { "epoch": 4.88, "learning_rate": 0.0003784572647852168, "loss": 1.1711, "step": 121260 }, { "epoch": 4.88, "learning_rate": 0.0003784471999677926, "loss": 0.9466, "step": 121270 }, { "epoch": 4.88, "learning_rate": 0.00037843713515036835, "loss": 0.9644, "step": 121280 }, { "epoch": 4.88, "learning_rate": 0.00037842707033294414, "loss": 1.1892, "step": 121290 }, { "epoch": 4.88, "learning_rate": 0.00037841700551551993, "loss": 1.0584, "step": 121300 }, { "epoch": 4.88, "learning_rate": 0.0003784069406980957, "loss": 1.0112, "step": 121310 }, { "epoch": 4.88, "learning_rate": 0.0003783968758806715, "loss": 1.0539, "step": 121320 }, { "epoch": 4.88, "learning_rate": 0.0003783868110632473, "loss": 1.0131, "step": 121330 }, { "epoch": 4.88, "learning_rate": 0.00037837674624582315, "loss": 0.8144, "step": 121340 }, { "epoch": 4.88, "learning_rate": 0.00037836668142839894, "loss": 1.0343, "step": 121350 }, { "epoch": 4.88, "learning_rate": 0.0003783566166109747, "loss": 1.01, "step": 121360 }, { "epoch": 4.88, "learning_rate": 0.00037834655179355047, "loss": 0.9556, "step": 121370 }, { "epoch": 4.88, "learning_rate": 0.00037833648697612626, "loss": 1.1572, "step": 121380 }, { "epoch": 4.88, "learning_rate": 0.00037832642215870205, "loss": 1.2137, "step": 121390 }, { "epoch": 4.88, "learning_rate": 0.00037831635734127784, "loss": 1.0548, "step": 121400 }, { "epoch": 4.88, "learning_rate": 0.00037830629252385363, "loss": 1.1255, "step": 121410 }, { "epoch": 4.88, "learning_rate": 0.0003782962277064294, "loss": 0.8996, "step": 121420 }, { "epoch": 4.88, "learning_rate": 0.0003782861628890052, "loss": 1.0362, "step": 121430 }, { "epoch": 4.88, "learning_rate": 0.000378276098071581, "loss": 0.9895, "step": 121440 }, { "epoch": 4.88, "learning_rate": 0.00037826603325415674, "loss": 0.9846, "step": 121450 }, { "epoch": 4.88, "learning_rate": 0.00037825596843673253, "loss": 0.8645, "step": 121460 }, { "epoch": 4.89, "learning_rate": 0.0003782459036193083, "loss": 1.2478, "step": 121470 }, { "epoch": 4.89, "learning_rate": 0.00037823583880188417, "loss": 1.2907, "step": 121480 }, { "epoch": 4.89, "learning_rate": 0.00037822577398445996, "loss": 0.952, "step": 121490 }, { "epoch": 4.89, "learning_rate": 0.00037821570916703575, "loss": 1.2395, "step": 121500 }, { "epoch": 4.89, "learning_rate": 0.00037820564434961154, "loss": 1.0366, "step": 121510 }, { "epoch": 4.89, "learning_rate": 0.00037819557953218733, "loss": 0.9641, "step": 121520 }, { "epoch": 4.89, "learning_rate": 0.00037818551471476307, "loss": 1.1314, "step": 121530 }, { "epoch": 4.89, "learning_rate": 0.00037817544989733886, "loss": 1.1609, "step": 121540 }, { "epoch": 4.89, "learning_rate": 0.00037816538507991465, "loss": 0.9748, "step": 121550 }, { "epoch": 4.89, "learning_rate": 0.00037815532026249044, "loss": 1.0484, "step": 121560 }, { "epoch": 4.89, "learning_rate": 0.00037814525544506623, "loss": 0.9574, "step": 121570 }, { "epoch": 4.89, "learning_rate": 0.000378135190627642, "loss": 1.0063, "step": 121580 }, { "epoch": 4.89, "learning_rate": 0.0003781251258102178, "loss": 1.2571, "step": 121590 }, { "epoch": 4.89, "learning_rate": 0.0003781150609927936, "loss": 1.154, "step": 121600 }, { "epoch": 4.89, "learning_rate": 0.0003781049961753694, "loss": 1.2301, "step": 121610 }, { "epoch": 4.89, "learning_rate": 0.00037809493135794513, "loss": 1.1671, "step": 121620 }, { "epoch": 4.89, "learning_rate": 0.000378084866540521, "loss": 1.119, "step": 121630 }, { "epoch": 4.89, "learning_rate": 0.00037807480172309676, "loss": 1.1657, "step": 121640 }, { "epoch": 4.89, "learning_rate": 0.00037806473690567255, "loss": 0.9864, "step": 121650 }, { "epoch": 4.89, "learning_rate": 0.00037805467208824835, "loss": 1.1236, "step": 121660 }, { "epoch": 4.89, "learning_rate": 0.00037804460727082414, "loss": 1.1102, "step": 121670 }, { "epoch": 4.89, "learning_rate": 0.0003780345424533999, "loss": 1.0065, "step": 121680 }, { "epoch": 4.89, "learning_rate": 0.0003780244776359757, "loss": 0.9538, "step": 121690 }, { "epoch": 4.89, "learning_rate": 0.00037801441281855145, "loss": 1.1225, "step": 121700 }, { "epoch": 4.89, "learning_rate": 0.00037800434800112724, "loss": 1.0606, "step": 121710 }, { "epoch": 4.9, "learning_rate": 0.00037799428318370304, "loss": 0.9437, "step": 121720 }, { "epoch": 4.9, "learning_rate": 0.0003779842183662788, "loss": 1.1071, "step": 121730 }, { "epoch": 4.9, "learning_rate": 0.0003779741535488546, "loss": 1.0831, "step": 121740 }, { "epoch": 4.9, "learning_rate": 0.0003779640887314304, "loss": 1.1327, "step": 121750 }, { "epoch": 4.9, "learning_rate": 0.0003779540239140062, "loss": 1.0686, "step": 121760 }, { "epoch": 4.9, "learning_rate": 0.000377943959096582, "loss": 1.0513, "step": 121770 }, { "epoch": 4.9, "learning_rate": 0.00037793389427915783, "loss": 0.9463, "step": 121780 }, { "epoch": 4.9, "learning_rate": 0.00037792382946173357, "loss": 1.1277, "step": 121790 }, { "epoch": 4.9, "learning_rate": 0.00037791376464430936, "loss": 1.1274, "step": 121800 }, { "epoch": 4.9, "learning_rate": 0.00037790369982688515, "loss": 1.104, "step": 121810 }, { "epoch": 4.9, "learning_rate": 0.00037789363500946094, "loss": 1.0527, "step": 121820 }, { "epoch": 4.9, "learning_rate": 0.00037788357019203673, "loss": 1.1163, "step": 121830 }, { "epoch": 4.9, "learning_rate": 0.0003778735053746125, "loss": 1.2786, "step": 121840 }, { "epoch": 4.9, "learning_rate": 0.0003778634405571883, "loss": 1.1111, "step": 121850 }, { "epoch": 4.9, "learning_rate": 0.0003778533757397641, "loss": 0.9129, "step": 121860 }, { "epoch": 4.9, "learning_rate": 0.00037784331092233984, "loss": 1.0178, "step": 121870 }, { "epoch": 4.9, "learning_rate": 0.00037783324610491563, "loss": 1.0599, "step": 121880 }, { "epoch": 4.9, "learning_rate": 0.0003778231812874914, "loss": 1.0852, "step": 121890 }, { "epoch": 4.9, "learning_rate": 0.0003778131164700672, "loss": 1.1324, "step": 121900 }, { "epoch": 4.9, "learning_rate": 0.000377803051652643, "loss": 0.9437, "step": 121910 }, { "epoch": 4.9, "learning_rate": 0.00037779298683521885, "loss": 0.9885, "step": 121920 }, { "epoch": 4.9, "learning_rate": 0.00037778292201779464, "loss": 1.0821, "step": 121930 }, { "epoch": 4.9, "learning_rate": 0.00037777285720037043, "loss": 1.1249, "step": 121940 }, { "epoch": 4.9, "learning_rate": 0.00037776279238294617, "loss": 1.1325, "step": 121950 }, { "epoch": 4.91, "learning_rate": 0.00037775272756552196, "loss": 0.8388, "step": 121960 }, { "epoch": 4.91, "learning_rate": 0.00037774266274809775, "loss": 1.0684, "step": 121970 }, { "epoch": 4.91, "learning_rate": 0.00037773259793067354, "loss": 1.141, "step": 121980 }, { "epoch": 4.91, "learning_rate": 0.00037772253311324933, "loss": 1.1283, "step": 121990 }, { "epoch": 4.91, "learning_rate": 0.0003777124682958251, "loss": 0.9995, "step": 122000 }, { "epoch": 4.91, "learning_rate": 0.0003777024034784009, "loss": 1.0808, "step": 122010 }, { "epoch": 4.91, "learning_rate": 0.0003776923386609767, "loss": 0.8413, "step": 122020 }, { "epoch": 4.91, "learning_rate": 0.0003776822738435525, "loss": 0.974, "step": 122030 }, { "epoch": 4.91, "learning_rate": 0.00037767220902612823, "loss": 1.0365, "step": 122040 }, { "epoch": 4.91, "learning_rate": 0.000377662144208704, "loss": 1.0876, "step": 122050 }, { "epoch": 4.91, "learning_rate": 0.0003776520793912798, "loss": 0.9985, "step": 122060 }, { "epoch": 4.91, "learning_rate": 0.00037764201457385566, "loss": 0.9955, "step": 122070 }, { "epoch": 4.91, "learning_rate": 0.00037763194975643145, "loss": 1.2067, "step": 122080 }, { "epoch": 4.91, "learning_rate": 0.00037762188493900724, "loss": 1.1513, "step": 122090 }, { "epoch": 4.91, "learning_rate": 0.00037761182012158303, "loss": 1.081, "step": 122100 }, { "epoch": 4.91, "learning_rate": 0.0003776017553041588, "loss": 1.0773, "step": 122110 }, { "epoch": 4.91, "learning_rate": 0.00037759169048673456, "loss": 1.0406, "step": 122120 }, { "epoch": 4.91, "learning_rate": 0.00037758162566931035, "loss": 0.9833, "step": 122130 }, { "epoch": 4.91, "learning_rate": 0.00037757156085188614, "loss": 0.9855, "step": 122140 }, { "epoch": 4.91, "learning_rate": 0.00037756149603446193, "loss": 1.105, "step": 122150 }, { "epoch": 4.91, "learning_rate": 0.0003775514312170377, "loss": 1.0778, "step": 122160 }, { "epoch": 4.91, "learning_rate": 0.0003775413663996135, "loss": 0.8861, "step": 122170 }, { "epoch": 4.91, "learning_rate": 0.0003775313015821893, "loss": 0.9491, "step": 122180 }, { "epoch": 4.91, "learning_rate": 0.0003775212367647651, "loss": 0.9382, "step": 122190 }, { "epoch": 4.91, "learning_rate": 0.0003775111719473409, "loss": 0.8808, "step": 122200 }, { "epoch": 4.92, "learning_rate": 0.0003775011071299167, "loss": 1.0928, "step": 122210 }, { "epoch": 4.92, "learning_rate": 0.00037749104231249247, "loss": 1.1532, "step": 122220 }, { "epoch": 4.92, "learning_rate": 0.00037748097749506826, "loss": 1.0264, "step": 122230 }, { "epoch": 4.92, "learning_rate": 0.00037747091267764405, "loss": 0.8479, "step": 122240 }, { "epoch": 4.92, "learning_rate": 0.00037746084786021984, "loss": 1.1303, "step": 122250 }, { "epoch": 4.92, "learning_rate": 0.00037745078304279563, "loss": 1.1058, "step": 122260 }, { "epoch": 4.92, "learning_rate": 0.0003774407182253714, "loss": 1.2791, "step": 122270 }, { "epoch": 4.92, "learning_rate": 0.0003774306534079472, "loss": 0.8524, "step": 122280 }, { "epoch": 4.92, "learning_rate": 0.00037742058859052295, "loss": 1.2539, "step": 122290 }, { "epoch": 4.92, "learning_rate": 0.00037741052377309874, "loss": 1.1507, "step": 122300 }, { "epoch": 4.92, "learning_rate": 0.00037740045895567453, "loss": 1.2876, "step": 122310 }, { "epoch": 4.92, "learning_rate": 0.0003773903941382503, "loss": 1.0326, "step": 122320 }, { "epoch": 4.92, "learning_rate": 0.0003773803293208261, "loss": 0.9514, "step": 122330 }, { "epoch": 4.92, "learning_rate": 0.0003773702645034019, "loss": 1.2722, "step": 122340 }, { "epoch": 4.92, "learning_rate": 0.0003773601996859777, "loss": 0.9047, "step": 122350 }, { "epoch": 4.92, "learning_rate": 0.00037735013486855354, "loss": 1.0729, "step": 122360 }, { "epoch": 4.92, "learning_rate": 0.00037734007005112933, "loss": 1.1421, "step": 122370 }, { "epoch": 4.92, "learning_rate": 0.00037733000523370506, "loss": 0.9165, "step": 122380 }, { "epoch": 4.92, "learning_rate": 0.00037731994041628086, "loss": 1.3227, "step": 122390 }, { "epoch": 4.92, "learning_rate": 0.00037730987559885665, "loss": 1.1255, "step": 122400 }, { "epoch": 4.92, "learning_rate": 0.00037729981078143244, "loss": 1.0446, "step": 122410 }, { "epoch": 4.92, "learning_rate": 0.00037728974596400823, "loss": 0.9351, "step": 122420 }, { "epoch": 4.92, "learning_rate": 0.000377279681146584, "loss": 1.0167, "step": 122430 }, { "epoch": 4.92, "learning_rate": 0.0003772696163291598, "loss": 1.1543, "step": 122440 }, { "epoch": 4.92, "learning_rate": 0.0003772595515117356, "loss": 1.1495, "step": 122450 }, { "epoch": 4.93, "learning_rate": 0.00037724948669431134, "loss": 1.2155, "step": 122460 }, { "epoch": 4.93, "learning_rate": 0.00037723942187688713, "loss": 1.1401, "step": 122470 }, { "epoch": 4.93, "learning_rate": 0.0003772293570594629, "loss": 1.0823, "step": 122480 }, { "epoch": 4.93, "learning_rate": 0.0003772192922420387, "loss": 0.8932, "step": 122490 }, { "epoch": 4.93, "learning_rate": 0.0003772092274246145, "loss": 1.2029, "step": 122500 }, { "epoch": 4.93, "learning_rate": 0.00037719916260719035, "loss": 1.2236, "step": 122510 }, { "epoch": 4.93, "learning_rate": 0.00037718909778976614, "loss": 0.976, "step": 122520 }, { "epoch": 4.93, "learning_rate": 0.0003771790329723419, "loss": 1.115, "step": 122530 }, { "epoch": 4.93, "learning_rate": 0.0003771689681549177, "loss": 1.1404, "step": 122540 }, { "epoch": 4.93, "learning_rate": 0.00037715890333749345, "loss": 1.1018, "step": 122550 }, { "epoch": 4.93, "learning_rate": 0.00037714883852006924, "loss": 0.9544, "step": 122560 }, { "epoch": 4.93, "learning_rate": 0.00037713877370264504, "loss": 1.0354, "step": 122570 }, { "epoch": 4.93, "learning_rate": 0.0003771287088852208, "loss": 1.1319, "step": 122580 }, { "epoch": 4.93, "learning_rate": 0.0003771186440677966, "loss": 1.0493, "step": 122590 }, { "epoch": 4.93, "learning_rate": 0.0003771085792503724, "loss": 1.3096, "step": 122600 }, { "epoch": 4.93, "learning_rate": 0.0003770985144329482, "loss": 1.1229, "step": 122610 }, { "epoch": 4.93, "learning_rate": 0.000377088449615524, "loss": 0.9701, "step": 122620 }, { "epoch": 4.93, "learning_rate": 0.0003770783847980997, "loss": 1.0202, "step": 122630 }, { "epoch": 4.93, "learning_rate": 0.0003770683199806755, "loss": 1.1339, "step": 122640 }, { "epoch": 4.93, "learning_rate": 0.00037705825516325136, "loss": 1.2154, "step": 122650 }, { "epoch": 4.93, "learning_rate": 0.00037704819034582715, "loss": 0.906, "step": 122660 }, { "epoch": 4.93, "learning_rate": 0.00037703812552840294, "loss": 1.2134, "step": 122670 }, { "epoch": 4.93, "learning_rate": 0.00037702806071097873, "loss": 0.9145, "step": 122680 }, { "epoch": 4.93, "learning_rate": 0.0003770179958935545, "loss": 1.1279, "step": 122690 }, { "epoch": 4.93, "learning_rate": 0.0003770079310761303, "loss": 0.9347, "step": 122700 }, { "epoch": 4.94, "learning_rate": 0.00037699786625870605, "loss": 1.1607, "step": 122710 }, { "epoch": 4.94, "learning_rate": 0.00037698780144128184, "loss": 1.0496, "step": 122720 }, { "epoch": 4.94, "learning_rate": 0.00037697773662385763, "loss": 1.066, "step": 122730 }, { "epoch": 4.94, "learning_rate": 0.0003769676718064334, "loss": 0.9197, "step": 122740 }, { "epoch": 4.94, "learning_rate": 0.0003769576069890092, "loss": 1.019, "step": 122750 }, { "epoch": 4.94, "learning_rate": 0.000376947542171585, "loss": 1.2685, "step": 122760 }, { "epoch": 4.94, "learning_rate": 0.0003769374773541608, "loss": 1.2083, "step": 122770 }, { "epoch": 4.94, "learning_rate": 0.0003769274125367366, "loss": 0.854, "step": 122780 }, { "epoch": 4.94, "learning_rate": 0.0003769173477193124, "loss": 0.9193, "step": 122790 }, { "epoch": 4.94, "learning_rate": 0.00037690728290188817, "loss": 1.1785, "step": 122800 }, { "epoch": 4.94, "learning_rate": 0.00037689721808446396, "loss": 1.1178, "step": 122810 }, { "epoch": 4.94, "learning_rate": 0.00037688715326703975, "loss": 1.0791, "step": 122820 }, { "epoch": 4.94, "learning_rate": 0.00037687708844961554, "loss": 0.9993, "step": 122830 }, { "epoch": 4.94, "learning_rate": 0.00037686702363219133, "loss": 1.0914, "step": 122840 }, { "epoch": 4.94, "learning_rate": 0.0003768569588147671, "loss": 0.8396, "step": 122850 }, { "epoch": 4.94, "learning_rate": 0.0003768468939973429, "loss": 0.9805, "step": 122860 }, { "epoch": 4.94, "learning_rate": 0.0003768368291799187, "loss": 1.3251, "step": 122870 }, { "epoch": 4.94, "learning_rate": 0.00037682676436249444, "loss": 1.0956, "step": 122880 }, { "epoch": 4.94, "learning_rate": 0.00037681669954507023, "loss": 1.1945, "step": 122890 }, { "epoch": 4.94, "learning_rate": 0.000376806634727646, "loss": 0.8151, "step": 122900 }, { "epoch": 4.94, "learning_rate": 0.0003767965699102218, "loss": 1.0253, "step": 122910 }, { "epoch": 4.94, "learning_rate": 0.0003767865050927976, "loss": 0.9224, "step": 122920 }, { "epoch": 4.94, "learning_rate": 0.0003767764402753734, "loss": 1.1075, "step": 122930 }, { "epoch": 4.94, "learning_rate": 0.0003767663754579492, "loss": 1.0409, "step": 122940 }, { "epoch": 4.94, "learning_rate": 0.00037675631064052503, "loss": 1.1817, "step": 122950 }, { "epoch": 4.95, "learning_rate": 0.0003767462458231008, "loss": 1.1389, "step": 122960 }, { "epoch": 4.95, "learning_rate": 0.00037673618100567656, "loss": 1.2619, "step": 122970 }, { "epoch": 4.95, "learning_rate": 0.00037672611618825235, "loss": 1.1051, "step": 122980 }, { "epoch": 4.95, "learning_rate": 0.00037671605137082814, "loss": 1.1625, "step": 122990 }, { "epoch": 4.95, "learning_rate": 0.00037670598655340393, "loss": 0.9418, "step": 123000 }, { "epoch": 4.95, "learning_rate": 0.0003766959217359797, "loss": 1.013, "step": 123010 }, { "epoch": 4.95, "learning_rate": 0.0003766858569185555, "loss": 1.0562, "step": 123020 }, { "epoch": 4.95, "learning_rate": 0.0003766757921011313, "loss": 0.8341, "step": 123030 }, { "epoch": 4.95, "learning_rate": 0.0003766657272837071, "loss": 1.0195, "step": 123040 }, { "epoch": 4.95, "learning_rate": 0.00037665566246628283, "loss": 0.9463, "step": 123050 }, { "epoch": 4.95, "learning_rate": 0.0003766455976488586, "loss": 1.0703, "step": 123060 }, { "epoch": 4.95, "learning_rate": 0.0003766355328314344, "loss": 1.0045, "step": 123070 }, { "epoch": 4.95, "learning_rate": 0.0003766254680140102, "loss": 1.048, "step": 123080 }, { "epoch": 4.95, "learning_rate": 0.00037661540319658605, "loss": 1.3187, "step": 123090 }, { "epoch": 4.95, "learning_rate": 0.00037660533837916184, "loss": 1.1993, "step": 123100 }, { "epoch": 4.95, "learning_rate": 0.00037659527356173763, "loss": 1.1298, "step": 123110 }, { "epoch": 4.95, "learning_rate": 0.0003765852087443134, "loss": 0.9189, "step": 123120 }, { "epoch": 4.95, "learning_rate": 0.0003765751439268892, "loss": 1.0052, "step": 123130 }, { "epoch": 4.95, "learning_rate": 0.00037656507910946495, "loss": 1.2217, "step": 123140 }, { "epoch": 4.95, "learning_rate": 0.00037655501429204074, "loss": 1.0691, "step": 123150 }, { "epoch": 4.95, "learning_rate": 0.00037654494947461653, "loss": 1.0129, "step": 123160 }, { "epoch": 4.95, "learning_rate": 0.0003765348846571923, "loss": 1.0283, "step": 123170 }, { "epoch": 4.95, "learning_rate": 0.0003765248198397681, "loss": 1.1396, "step": 123180 }, { "epoch": 4.95, "learning_rate": 0.0003765147550223439, "loss": 1.0488, "step": 123190 }, { "epoch": 4.95, "learning_rate": 0.0003765046902049197, "loss": 0.9967, "step": 123200 }, { "epoch": 4.96, "learning_rate": 0.0003764946253874955, "loss": 1.1055, "step": 123210 }, { "epoch": 4.96, "learning_rate": 0.0003764845605700712, "loss": 0.9938, "step": 123220 }, { "epoch": 4.96, "learning_rate": 0.000376474495752647, "loss": 0.9526, "step": 123230 }, { "epoch": 4.96, "learning_rate": 0.00037646443093522286, "loss": 1.0732, "step": 123240 }, { "epoch": 4.96, "learning_rate": 0.00037645436611779865, "loss": 1.1183, "step": 123250 }, { "epoch": 4.96, "learning_rate": 0.00037644430130037444, "loss": 1.0031, "step": 123260 }, { "epoch": 4.96, "learning_rate": 0.00037643423648295023, "loss": 1.166, "step": 123270 }, { "epoch": 4.96, "learning_rate": 0.000376424171665526, "loss": 1.1058, "step": 123280 }, { "epoch": 4.96, "learning_rate": 0.0003764141068481018, "loss": 1.1686, "step": 123290 }, { "epoch": 4.96, "learning_rate": 0.0003764040420306776, "loss": 0.9913, "step": 123300 }, { "epoch": 4.96, "learning_rate": 0.00037639397721325334, "loss": 1.123, "step": 123310 }, { "epoch": 4.96, "learning_rate": 0.00037638391239582913, "loss": 0.9896, "step": 123320 }, { "epoch": 4.96, "learning_rate": 0.0003763738475784049, "loss": 1.171, "step": 123330 }, { "epoch": 4.96, "learning_rate": 0.0003763637827609807, "loss": 1.0285, "step": 123340 }, { "epoch": 4.96, "learning_rate": 0.0003763537179435565, "loss": 0.7869, "step": 123350 }, { "epoch": 4.96, "learning_rate": 0.0003763436531261323, "loss": 0.9722, "step": 123360 }, { "epoch": 4.96, "learning_rate": 0.0003763335883087081, "loss": 1.1176, "step": 123370 }, { "epoch": 4.96, "learning_rate": 0.0003763235234912839, "loss": 1.1002, "step": 123380 }, { "epoch": 4.96, "learning_rate": 0.00037631345867385966, "loss": 1.3149, "step": 123390 }, { "epoch": 4.96, "learning_rate": 0.00037630339385643545, "loss": 1.0323, "step": 123400 }, { "epoch": 4.96, "learning_rate": 0.00037629332903901124, "loss": 1.3547, "step": 123410 }, { "epoch": 4.96, "learning_rate": 0.00037628326422158704, "loss": 0.982, "step": 123420 }, { "epoch": 4.96, "learning_rate": 0.0003762731994041628, "loss": 1.2098, "step": 123430 }, { "epoch": 4.96, "learning_rate": 0.0003762631345867386, "loss": 1.1025, "step": 123440 }, { "epoch": 4.96, "learning_rate": 0.0003762530697693144, "loss": 1.3097, "step": 123450 }, { "epoch": 4.97, "learning_rate": 0.0003762430049518902, "loss": 1.0838, "step": 123460 }, { "epoch": 4.97, "learning_rate": 0.00037623294013446593, "loss": 0.9884, "step": 123470 }, { "epoch": 4.97, "learning_rate": 0.0003762228753170417, "loss": 1.1198, "step": 123480 }, { "epoch": 4.97, "learning_rate": 0.0003762128104996175, "loss": 0.8794, "step": 123490 }, { "epoch": 4.97, "learning_rate": 0.0003762027456821933, "loss": 1.2285, "step": 123500 }, { "epoch": 4.97, "learning_rate": 0.0003761926808647691, "loss": 1.0977, "step": 123510 }, { "epoch": 4.97, "learning_rate": 0.0003761826160473449, "loss": 1.2762, "step": 123520 }, { "epoch": 4.97, "learning_rate": 0.00037617255122992073, "loss": 1.1613, "step": 123530 }, { "epoch": 4.97, "learning_rate": 0.0003761624864124965, "loss": 1.0524, "step": 123540 }, { "epoch": 4.97, "learning_rate": 0.0003761524215950723, "loss": 0.9975, "step": 123550 }, { "epoch": 4.97, "learning_rate": 0.00037614235677764805, "loss": 1.1204, "step": 123560 }, { "epoch": 4.97, "learning_rate": 0.00037613229196022384, "loss": 1.0615, "step": 123570 }, { "epoch": 4.97, "learning_rate": 0.00037612222714279963, "loss": 1.2037, "step": 123580 }, { "epoch": 4.97, "learning_rate": 0.0003761121623253754, "loss": 1.103, "step": 123590 }, { "epoch": 4.97, "learning_rate": 0.0003761020975079512, "loss": 1.1683, "step": 123600 }, { "epoch": 4.97, "learning_rate": 0.000376092032690527, "loss": 0.9731, "step": 123610 }, { "epoch": 4.97, "learning_rate": 0.0003760819678731028, "loss": 1.004, "step": 123620 }, { "epoch": 4.97, "learning_rate": 0.0003760719030556786, "loss": 1.2201, "step": 123630 }, { "epoch": 4.97, "learning_rate": 0.0003760618382382543, "loss": 1.072, "step": 123640 }, { "epoch": 4.97, "learning_rate": 0.0003760517734208301, "loss": 1.0039, "step": 123650 }, { "epoch": 4.97, "learning_rate": 0.0003760417086034059, "loss": 1.2606, "step": 123660 }, { "epoch": 4.97, "learning_rate": 0.0003760316437859817, "loss": 1.2677, "step": 123670 }, { "epoch": 4.97, "learning_rate": 0.00037602157896855754, "loss": 1.0379, "step": 123680 }, { "epoch": 4.97, "learning_rate": 0.00037601151415113333, "loss": 0.9472, "step": 123690 }, { "epoch": 4.98, "learning_rate": 0.0003760014493337091, "loss": 1.0544, "step": 123700 }, { "epoch": 4.98, "learning_rate": 0.0003759913845162849, "loss": 1.0083, "step": 123710 }, { "epoch": 4.98, "learning_rate": 0.0003759813196988607, "loss": 1.1733, "step": 123720 }, { "epoch": 4.98, "learning_rate": 0.00037597125488143644, "loss": 1.2663, "step": 123730 }, { "epoch": 4.98, "learning_rate": 0.00037596119006401223, "loss": 0.9782, "step": 123740 }, { "epoch": 4.98, "learning_rate": 0.000375951125246588, "loss": 1.2253, "step": 123750 }, { "epoch": 4.98, "learning_rate": 0.0003759410604291638, "loss": 1.289, "step": 123760 }, { "epoch": 4.98, "learning_rate": 0.0003759309956117396, "loss": 1.2602, "step": 123770 }, { "epoch": 4.98, "learning_rate": 0.0003759209307943154, "loss": 1.1448, "step": 123780 }, { "epoch": 4.98, "learning_rate": 0.0003759108659768912, "loss": 1.0775, "step": 123790 }, { "epoch": 4.98, "learning_rate": 0.000375900801159467, "loss": 1.0795, "step": 123800 }, { "epoch": 4.98, "learning_rate": 0.0003758907363420427, "loss": 1.1146, "step": 123810 }, { "epoch": 4.98, "learning_rate": 0.00037588067152461856, "loss": 1.0675, "step": 123820 }, { "epoch": 4.98, "learning_rate": 0.00037587060670719435, "loss": 1.2276, "step": 123830 }, { "epoch": 4.98, "learning_rate": 0.00037586054188977014, "loss": 1.0623, "step": 123840 }, { "epoch": 4.98, "learning_rate": 0.00037585047707234593, "loss": 1.056, "step": 123850 }, { "epoch": 4.98, "learning_rate": 0.0003758404122549217, "loss": 1.2199, "step": 123860 }, { "epoch": 4.98, "learning_rate": 0.0003758303474374975, "loss": 1.0802, "step": 123870 }, { "epoch": 4.98, "learning_rate": 0.0003758202826200733, "loss": 0.8998, "step": 123880 }, { "epoch": 4.98, "learning_rate": 0.0003758102178026491, "loss": 1.046, "step": 123890 }, { "epoch": 4.98, "learning_rate": 0.00037580015298522483, "loss": 0.9523, "step": 123900 }, { "epoch": 4.98, "learning_rate": 0.0003757900881678006, "loss": 1.2579, "step": 123910 }, { "epoch": 4.98, "learning_rate": 0.0003757800233503764, "loss": 1.0459, "step": 123920 }, { "epoch": 4.98, "learning_rate": 0.0003757699585329522, "loss": 0.8647, "step": 123930 }, { "epoch": 4.98, "learning_rate": 0.000375759893715528, "loss": 1.0066, "step": 123940 }, { "epoch": 4.99, "learning_rate": 0.0003757498288981038, "loss": 1.1295, "step": 123950 }, { "epoch": 4.99, "learning_rate": 0.0003757397640806796, "loss": 1.1639, "step": 123960 }, { "epoch": 4.99, "learning_rate": 0.0003757296992632554, "loss": 1.0368, "step": 123970 }, { "epoch": 4.99, "learning_rate": 0.00037571963444583116, "loss": 1.2417, "step": 123980 }, { "epoch": 4.99, "learning_rate": 0.00037570956962840695, "loss": 1.2238, "step": 123990 }, { "epoch": 4.99, "learning_rate": 0.00037569950481098274, "loss": 0.974, "step": 124000 }, { "epoch": 4.99, "learning_rate": 0.00037568943999355853, "loss": 0.8771, "step": 124010 }, { "epoch": 4.99, "learning_rate": 0.0003756793751761343, "loss": 1.3052, "step": 124020 }, { "epoch": 4.99, "learning_rate": 0.0003756693103587101, "loss": 1.118, "step": 124030 }, { "epoch": 4.99, "learning_rate": 0.0003756592455412859, "loss": 0.9119, "step": 124040 }, { "epoch": 4.99, "learning_rate": 0.0003756491807238617, "loss": 1.1551, "step": 124050 }, { "epoch": 4.99, "learning_rate": 0.0003756391159064375, "loss": 0.9509, "step": 124060 }, { "epoch": 4.99, "learning_rate": 0.0003756290510890132, "loss": 1.0763, "step": 124070 }, { "epoch": 4.99, "learning_rate": 0.000375618986271589, "loss": 1.1499, "step": 124080 }, { "epoch": 4.99, "learning_rate": 0.0003756089214541648, "loss": 1.1431, "step": 124090 }, { "epoch": 4.99, "learning_rate": 0.0003755988566367406, "loss": 1.0372, "step": 124100 }, { "epoch": 4.99, "learning_rate": 0.0003755887918193164, "loss": 0.8679, "step": 124110 }, { "epoch": 4.99, "learning_rate": 0.00037557872700189223, "loss": 1.0431, "step": 124120 }, { "epoch": 4.99, "learning_rate": 0.000375568662184468, "loss": 0.9634, "step": 124130 }, { "epoch": 4.99, "learning_rate": 0.0003755585973670438, "loss": 1.3384, "step": 124140 }, { "epoch": 4.99, "learning_rate": 0.00037554853254961955, "loss": 1.234, "step": 124150 }, { "epoch": 4.99, "learning_rate": 0.00037553846773219534, "loss": 1.1099, "step": 124160 }, { "epoch": 4.99, "learning_rate": 0.00037552840291477113, "loss": 1.1615, "step": 124170 }, { "epoch": 4.99, "learning_rate": 0.0003755183380973469, "loss": 1.0221, "step": 124180 }, { "epoch": 4.99, "learning_rate": 0.0003755082732799227, "loss": 1.158, "step": 124190 }, { "epoch": 5.0, "learning_rate": 0.0003754982084624985, "loss": 0.8807, "step": 124200 }, { "epoch": 5.0, "learning_rate": 0.0003754881436450743, "loss": 1.0552, "step": 124210 }, { "epoch": 5.0, "learning_rate": 0.0003754780788276501, "loss": 1.1293, "step": 124220 }, { "epoch": 5.0, "learning_rate": 0.0003754680140102258, "loss": 1.2243, "step": 124230 }, { "epoch": 5.0, "learning_rate": 0.0003754579491928016, "loss": 1.1243, "step": 124240 }, { "epoch": 5.0, "learning_rate": 0.0003754478843753774, "loss": 0.9413, "step": 124250 }, { "epoch": 5.0, "learning_rate": 0.00037543781955795324, "loss": 1.0482, "step": 124260 }, { "epoch": 5.0, "learning_rate": 0.00037542775474052904, "loss": 1.1715, "step": 124270 }, { "epoch": 5.0, "learning_rate": 0.0003754176899231048, "loss": 1.1489, "step": 124280 }, { "epoch": 5.0, "learning_rate": 0.0003754076251056806, "loss": 1.2025, "step": 124290 }, { "epoch": 5.0, "learning_rate": 0.0003753975602882564, "loss": 1.1736, "step": 124300 }, { "epoch": 5.0, "learning_rate": 0.0003753874954708322, "loss": 1.0315, "step": 124310 }, { "epoch": 5.0, "learning_rate": 0.00037537743065340793, "loss": 0.7704, "step": 124320 }, { "epoch": 5.0, "eval_Bleu_1": 0.016265628838520108, "eval_Bleu_2": 1.8153528512486202e-11, "eval_Bleu_3": 2.1350049727728716e-14, "eval_Bleu_4": 7.958823038617678e-16, "eval_ROUGE_L": 0.03830770696546531, "eval_cer": 0.6820582301515091, "eval_em": 0.32332761578044594, "eval_f1": 0.43246031181423683, "eval_loss": 1.7568819522857666, "eval_runtime": 2103.439, "eval_samples_per_second": 13.304, "eval_steps_per_second": 4.435, "eval_wer": 0.8240235243164385, "step": 124321 } ], "max_steps": 497280, "num_train_epochs": 20, "total_flos": 1.4845743813589862e+18, "trial_name": null, "trial_params": null }