{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.3298833037812874, "eval_steps": 500, "global_step": 40000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 9.999999832181266e-06, "loss": 99.1391, "step": 10 }, { "epoch": 0.0, "learning_rate": 9.999999328725077e-06, "loss": 64.5758, "step": 20 }, { "epoch": 0.0, "learning_rate": 9.999998489631465e-06, "loss": 37.0066, "step": 30 }, { "epoch": 0.0, "learning_rate": 9.999997314900488e-06, "loss": 27.2098, "step": 40 }, { "epoch": 0.0, "learning_rate": 9.999995804532221e-06, "loss": 18.9039, "step": 50 }, { "epoch": 0.0, "learning_rate": 9.999993958526771e-06, "loss": 16.8481, "step": 60 }, { "epoch": 0.0, "learning_rate": 9.999991776884258e-06, "loss": 15.1406, "step": 70 }, { "epoch": 0.0, "learning_rate": 9.99998925960483e-06, "loss": 13.9292, "step": 80 }, { "epoch": 0.0, "learning_rate": 9.999986406688656e-06, "loss": 13.6947, "step": 90 }, { "epoch": 0.0, "learning_rate": 9.999983218135925e-06, "loss": 12.7382, "step": 100 }, { "epoch": 0.0, "learning_rate": 9.999979693946855e-06, "loss": 13.3305, "step": 110 }, { "epoch": 0.0, "learning_rate": 9.99997583412168e-06, "loss": 12.3668, "step": 120 }, { "epoch": 0.0, "learning_rate": 9.99997163866066e-06, "loss": 12.27, "step": 130 }, { "epoch": 0.0, "learning_rate": 9.999967107564078e-06, "loss": 11.8517, "step": 140 }, { "epoch": 0.0, "learning_rate": 9.999962240832235e-06, "loss": 11.7888, "step": 150 }, { "epoch": 0.0, "learning_rate": 9.999957038465458e-06, "loss": 12.0075, "step": 160 }, { "epoch": 0.0, "learning_rate": 9.9999515004641e-06, "loss": 11.1305, "step": 170 }, { "epoch": 0.0, "learning_rate": 9.999945626828529e-06, "loss": 11.4643, "step": 180 }, { "epoch": 0.0, "learning_rate": 9.99993941755914e-06, "loss": 10.9472, "step": 190 }, { "epoch": 0.0, "learning_rate": 9.999932872656351e-06, "loss": 10.9608, "step": 200 }, { "epoch": 0.0, "learning_rate": 9.9999259921206e-06, "loss": 10.4347, "step": 210 }, { "epoch": 0.0, "learning_rate": 9.999918775952351e-06, "loss": 11.6568, "step": 220 }, { "epoch": 0.0, "learning_rate": 9.999911224152086e-06, "loss": 9.9892, "step": 230 }, { "epoch": 0.0, "learning_rate": 9.999903336720313e-06, "loss": 10.4669, "step": 240 }, { "epoch": 0.0, "learning_rate": 9.999895113657563e-06, "loss": 9.9074, "step": 250 }, { "epoch": 0.0, "learning_rate": 9.999886554964385e-06, "loss": 10.7965, "step": 260 }, { "epoch": 0.0, "learning_rate": 9.999877660641354e-06, "loss": 9.5332, "step": 270 }, { "epoch": 0.0, "learning_rate": 9.99986843068907e-06, "loss": 9.408, "step": 280 }, { "epoch": 0.0, "learning_rate": 9.99985886510815e-06, "loss": 9.7117, "step": 290 }, { "epoch": 0.0, "learning_rate": 9.999848963899235e-06, "loss": 9.9966, "step": 300 }, { "epoch": 0.0, "learning_rate": 9.999838727062993e-06, "loss": 10.2939, "step": 310 }, { "epoch": 0.0, "learning_rate": 9.999828154600108e-06, "loss": 10.1468, "step": 320 }, { "epoch": 0.0, "learning_rate": 9.999817246511292e-06, "loss": 9.7001, "step": 330 }, { "epoch": 0.0, "learning_rate": 9.999806002797277e-06, "loss": 9.5166, "step": 340 }, { "epoch": 0.0, "learning_rate": 9.999794423458817e-06, "loss": 9.8452, "step": 350 }, { "epoch": 0.0, "learning_rate": 9.999782508496688e-06, "loss": 9.6854, "step": 360 }, { "epoch": 0.0, "learning_rate": 9.999770257911692e-06, "loss": 9.0236, "step": 370 }, { "epoch": 0.0, "learning_rate": 9.99975767170465e-06, "loss": 9.0524, "step": 380 }, { "epoch": 0.0, "learning_rate": 9.99974474987641e-06, "loss": 9.4815, "step": 390 }, { "epoch": 0.0, "learning_rate": 9.999731492427835e-06, "loss": 8.9528, "step": 400 }, { "epoch": 0.0, "learning_rate": 9.999717899359817e-06, "loss": 8.805, "step": 410 }, { "epoch": 0.0, "learning_rate": 9.99970397067327e-06, "loss": 9.2981, "step": 420 }, { "epoch": 0.0, "learning_rate": 9.999689706369125e-06, "loss": 9.5023, "step": 430 }, { "epoch": 0.0, "learning_rate": 9.999675106448342e-06, "loss": 9.6937, "step": 440 }, { "epoch": 0.0, "learning_rate": 9.999660170911902e-06, "loss": 9.4832, "step": 450 }, { "epoch": 0.0, "learning_rate": 9.999644899760806e-06, "loss": 9.9977, "step": 460 }, { "epoch": 0.0, "learning_rate": 9.999629292996078e-06, "loss": 9.3421, "step": 470 }, { "epoch": 0.0, "learning_rate": 9.99961335061877e-06, "loss": 8.8915, "step": 480 }, { "epoch": 0.0, "learning_rate": 9.999597072629948e-06, "loss": 9.4162, "step": 490 }, { "epoch": 0.0, "learning_rate": 9.999580459030707e-06, "loss": 8.2067, "step": 500 }, { "epoch": 0.0, "learning_rate": 9.99956350982216e-06, "loss": 8.9235, "step": 510 }, { "epoch": 0.0, "learning_rate": 9.999546225005448e-06, "loss": 8.9816, "step": 520 }, { "epoch": 0.0, "learning_rate": 9.999528604581728e-06, "loss": 9.1177, "step": 530 }, { "epoch": 0.0, "learning_rate": 9.999510648552184e-06, "loss": 9.1034, "step": 540 }, { "epoch": 0.0, "learning_rate": 9.999492356918021e-06, "loss": 8.4282, "step": 550 }, { "epoch": 0.0, "learning_rate": 9.999473729680469e-06, "loss": 9.1486, "step": 560 }, { "epoch": 0.0, "learning_rate": 9.999454766840778e-06, "loss": 9.6939, "step": 570 }, { "epoch": 0.0, "learning_rate": 9.999435468400216e-06, "loss": 8.8328, "step": 580 }, { "epoch": 0.0, "learning_rate": 9.999415834360085e-06, "loss": 9.2537, "step": 590 }, { "epoch": 0.0, "learning_rate": 9.9993958647217e-06, "loss": 9.7711, "step": 600 }, { "epoch": 0.01, "learning_rate": 9.999375559486401e-06, "loss": 9.0733, "step": 610 }, { "epoch": 0.01, "learning_rate": 9.99935491865555e-06, "loss": 9.0194, "step": 620 }, { "epoch": 0.01, "learning_rate": 9.999333942230538e-06, "loss": 8.4513, "step": 630 }, { "epoch": 0.01, "learning_rate": 9.999312630212767e-06, "loss": 8.7859, "step": 640 }, { "epoch": 0.01, "learning_rate": 9.999290982603671e-06, "loss": 9.0826, "step": 650 }, { "epoch": 0.01, "learning_rate": 9.999268999404701e-06, "loss": 8.3879, "step": 660 }, { "epoch": 0.01, "learning_rate": 9.999246680617334e-06, "loss": 9.3509, "step": 670 }, { "epoch": 0.01, "learning_rate": 9.99922402624307e-06, "loss": 8.8033, "step": 680 }, { "epoch": 0.01, "learning_rate": 9.999201036283426e-06, "loss": 9.5108, "step": 690 }, { "epoch": 0.01, "learning_rate": 9.999177710739947e-06, "loss": 9.5319, "step": 700 }, { "epoch": 0.01, "learning_rate": 9.9991540496142e-06, "loss": 8.475, "step": 710 }, { "epoch": 0.01, "learning_rate": 9.999130052907771e-06, "loss": 8.3887, "step": 720 }, { "epoch": 0.01, "learning_rate": 9.999105720622273e-06, "loss": 8.8797, "step": 730 }, { "epoch": 0.01, "learning_rate": 9.999081052759338e-06, "loss": 9.4537, "step": 740 }, { "epoch": 0.01, "learning_rate": 9.999056049320621e-06, "loss": 8.9044, "step": 750 }, { "epoch": 0.01, "learning_rate": 9.999030710307803e-06, "loss": 8.3333, "step": 760 }, { "epoch": 0.01, "learning_rate": 9.999005035722584e-06, "loss": 8.7374, "step": 770 }, { "epoch": 0.01, "learning_rate": 9.998979025566686e-06, "loss": 7.9971, "step": 780 }, { "epoch": 0.01, "learning_rate": 9.998952679841856e-06, "loss": 8.7104, "step": 790 }, { "epoch": 0.01, "learning_rate": 9.998925998549864e-06, "loss": 10.1392, "step": 800 }, { "epoch": 0.01, "learning_rate": 9.998898981692498e-06, "loss": 9.0902, "step": 810 }, { "epoch": 0.01, "learning_rate": 9.998871629271574e-06, "loss": 8.5757, "step": 820 }, { "epoch": 0.01, "learning_rate": 9.998843941288928e-06, "loss": 8.9774, "step": 830 }, { "epoch": 0.01, "learning_rate": 9.998815917746416e-06, "loss": 8.5707, "step": 840 }, { "epoch": 0.01, "learning_rate": 9.998787558645921e-06, "loss": 9.297, "step": 850 }, { "epoch": 0.01, "learning_rate": 9.998758863989349e-06, "loss": 8.8131, "step": 860 }, { "epoch": 0.01, "learning_rate": 9.998729833778624e-06, "loss": 9.1968, "step": 870 }, { "epoch": 0.01, "learning_rate": 9.998700468015693e-06, "loss": 8.5377, "step": 880 }, { "epoch": 0.01, "learning_rate": 9.998670766702529e-06, "loss": 8.0083, "step": 890 }, { "epoch": 0.01, "learning_rate": 9.998640729841127e-06, "loss": 8.9833, "step": 900 }, { "epoch": 0.01, "learning_rate": 9.998610357433501e-06, "loss": 8.6782, "step": 910 }, { "epoch": 0.01, "learning_rate": 9.998579649481691e-06, "loss": 8.0428, "step": 920 }, { "epoch": 0.01, "learning_rate": 9.998548605987759e-06, "loss": 8.4037, "step": 930 }, { "epoch": 0.01, "learning_rate": 9.998517226953787e-06, "loss": 8.4013, "step": 940 }, { "epoch": 0.01, "learning_rate": 9.998485512381883e-06, "loss": 8.7569, "step": 950 }, { "epoch": 0.01, "learning_rate": 9.998453462274174e-06, "loss": 8.4717, "step": 960 }, { "epoch": 0.01, "learning_rate": 9.998421076632815e-06, "loss": 7.9827, "step": 970 }, { "epoch": 0.01, "learning_rate": 9.998388355459976e-06, "loss": 8.2216, "step": 980 }, { "epoch": 0.01, "learning_rate": 9.998355298757857e-06, "loss": 8.8212, "step": 990 }, { "epoch": 0.01, "learning_rate": 9.998321906528675e-06, "loss": 8.8272, "step": 1000 }, { "epoch": 0.01, "learning_rate": 9.99828817877467e-06, "loss": 8.8629, "step": 1010 }, { "epoch": 0.01, "learning_rate": 9.998254115498109e-06, "loss": 8.21, "step": 1020 }, { "epoch": 0.01, "learning_rate": 9.998219716701279e-06, "loss": 9.1343, "step": 1030 }, { "epoch": 0.01, "learning_rate": 9.998184982386484e-06, "loss": 8.2472, "step": 1040 }, { "epoch": 0.01, "learning_rate": 9.998149912556063e-06, "loss": 8.3801, "step": 1050 }, { "epoch": 0.01, "learning_rate": 9.998114507212365e-06, "loss": 8.3729, "step": 1060 }, { "epoch": 0.01, "learning_rate": 9.998078766357767e-06, "loss": 8.4559, "step": 1070 }, { "epoch": 0.01, "learning_rate": 9.99804268999467e-06, "loss": 8.2093, "step": 1080 }, { "epoch": 0.01, "learning_rate": 9.998006278125494e-06, "loss": 8.4312, "step": 1090 }, { "epoch": 0.01, "learning_rate": 9.997969530752685e-06, "loss": 8.6102, "step": 1100 }, { "epoch": 0.01, "learning_rate": 9.997932447878709e-06, "loss": 8.35, "step": 1110 }, { "epoch": 0.01, "learning_rate": 9.997895029506055e-06, "loss": 8.6922, "step": 1120 }, { "epoch": 0.01, "learning_rate": 9.997857275637235e-06, "loss": 8.7375, "step": 1130 }, { "epoch": 0.01, "learning_rate": 9.997819186274784e-06, "loss": 8.5644, "step": 1140 }, { "epoch": 0.01, "learning_rate": 9.997780761421257e-06, "loss": 8.0859, "step": 1150 }, { "epoch": 0.01, "learning_rate": 9.997742001079236e-06, "loss": 9.0077, "step": 1160 }, { "epoch": 0.01, "learning_rate": 9.99770290525132e-06, "loss": 8.995, "step": 1170 }, { "epoch": 0.01, "learning_rate": 9.997663473940136e-06, "loss": 7.9538, "step": 1180 }, { "epoch": 0.01, "learning_rate": 9.997623707148328e-06, "loss": 8.578, "step": 1190 }, { "epoch": 0.01, "learning_rate": 9.99758360487857e-06, "loss": 8.6668, "step": 1200 }, { "epoch": 0.01, "learning_rate": 9.997543167133547e-06, "loss": 8.1429, "step": 1210 }, { "epoch": 0.01, "learning_rate": 9.997502393915982e-06, "loss": 7.8296, "step": 1220 }, { "epoch": 0.01, "learning_rate": 9.997461285228605e-06, "loss": 7.7744, "step": 1230 }, { "epoch": 0.01, "learning_rate": 9.997419841074178e-06, "loss": 7.7537, "step": 1240 }, { "epoch": 0.01, "learning_rate": 9.997378061455484e-06, "loss": 8.2452, "step": 1250 }, { "epoch": 0.01, "learning_rate": 9.997335946375328e-06, "loss": 8.531, "step": 1260 }, { "epoch": 0.01, "learning_rate": 9.997293495836535e-06, "loss": 8.122, "step": 1270 }, { "epoch": 0.01, "learning_rate": 9.997250709841955e-06, "loss": 8.2261, "step": 1280 }, { "epoch": 0.01, "learning_rate": 9.997207588394461e-06, "loss": 7.9647, "step": 1290 }, { "epoch": 0.01, "learning_rate": 9.997164131496947e-06, "loss": 9.0417, "step": 1300 }, { "epoch": 0.01, "learning_rate": 9.99712033915233e-06, "loss": 8.851, "step": 1310 }, { "epoch": 0.01, "learning_rate": 9.99707621136355e-06, "loss": 8.0891, "step": 1320 }, { "epoch": 0.01, "learning_rate": 9.99703174813357e-06, "loss": 8.522, "step": 1330 }, { "epoch": 0.01, "learning_rate": 9.996986949465374e-06, "loss": 7.406, "step": 1340 }, { "epoch": 0.01, "learning_rate": 9.996941815361968e-06, "loss": 8.5289, "step": 1350 }, { "epoch": 0.01, "learning_rate": 9.996896345826383e-06, "loss": 8.2852, "step": 1360 }, { "epoch": 0.01, "learning_rate": 9.996850540861673e-06, "loss": 8.1178, "step": 1370 }, { "epoch": 0.01, "learning_rate": 9.99680440047091e-06, "loss": 8.3668, "step": 1380 }, { "epoch": 0.01, "learning_rate": 9.996757924657193e-06, "loss": 8.0, "step": 1390 }, { "epoch": 0.01, "learning_rate": 9.996711113423638e-06, "loss": 7.2943, "step": 1400 }, { "epoch": 0.01, "learning_rate": 9.996663966773393e-06, "loss": 7.8454, "step": 1410 }, { "epoch": 0.01, "learning_rate": 9.996616484709617e-06, "loss": 7.9649, "step": 1420 }, { "epoch": 0.01, "learning_rate": 9.996568667235504e-06, "loss": 8.9598, "step": 1430 }, { "epoch": 0.01, "learning_rate": 9.99652051435426e-06, "loss": 8.8281, "step": 1440 }, { "epoch": 0.01, "learning_rate": 9.996472026069118e-06, "loss": 7.5037, "step": 1450 }, { "epoch": 0.01, "learning_rate": 9.996423202383331e-06, "loss": 8.9229, "step": 1460 }, { "epoch": 0.01, "learning_rate": 9.996374043300179e-06, "loss": 8.3112, "step": 1470 }, { "epoch": 0.01, "learning_rate": 9.99632454882296e-06, "loss": 7.7392, "step": 1480 }, { "epoch": 0.01, "learning_rate": 9.996274718954998e-06, "loss": 8.6432, "step": 1490 }, { "epoch": 0.01, "learning_rate": 9.996224553699639e-06, "loss": 7.101, "step": 1500 }, { "epoch": 0.01, "learning_rate": 9.996174053060246e-06, "loss": 8.1319, "step": 1510 }, { "epoch": 0.01, "learning_rate": 9.996123217040213e-06, "loss": 8.7934, "step": 1520 }, { "epoch": 0.01, "learning_rate": 9.996072045642952e-06, "loss": 7.7222, "step": 1530 }, { "epoch": 0.01, "learning_rate": 9.996020538871897e-06, "loss": 7.8115, "step": 1540 }, { "epoch": 0.01, "learning_rate": 9.995968696730506e-06, "loss": 8.2136, "step": 1550 }, { "epoch": 0.01, "learning_rate": 9.995916519222258e-06, "loss": 8.6931, "step": 1560 }, { "epoch": 0.01, "learning_rate": 9.995864006350657e-06, "loss": 8.0203, "step": 1570 }, { "epoch": 0.01, "learning_rate": 9.995811158119228e-06, "loss": 8.4059, "step": 1580 }, { "epoch": 0.01, "learning_rate": 9.995757974531519e-06, "loss": 8.0458, "step": 1590 }, { "epoch": 0.01, "learning_rate": 9.995704455591096e-06, "loss": 8.2034, "step": 1600 }, { "epoch": 0.01, "learning_rate": 9.995650601301557e-06, "loss": 7.441, "step": 1610 }, { "epoch": 0.01, "learning_rate": 9.995596411666515e-06, "loss": 8.6135, "step": 1620 }, { "epoch": 0.01, "learning_rate": 9.995541886689605e-06, "loss": 7.6191, "step": 1630 }, { "epoch": 0.01, "learning_rate": 9.995487026374492e-06, "loss": 8.4708, "step": 1640 }, { "epoch": 0.01, "learning_rate": 9.995431830724858e-06, "loss": 7.152, "step": 1650 }, { "epoch": 0.01, "learning_rate": 9.995376299744404e-06, "loss": 8.6644, "step": 1660 }, { "epoch": 0.01, "learning_rate": 9.99532043343686e-06, "loss": 7.7558, "step": 1670 }, { "epoch": 0.01, "learning_rate": 9.995264231805976e-06, "loss": 7.9283, "step": 1680 }, { "epoch": 0.01, "learning_rate": 9.995207694855524e-06, "loss": 8.7824, "step": 1690 }, { "epoch": 0.01, "learning_rate": 9.995150822589301e-06, "loss": 7.4533, "step": 1700 }, { "epoch": 0.01, "learning_rate": 9.995093615011124e-06, "loss": 8.6312, "step": 1710 }, { "epoch": 0.01, "learning_rate": 9.995036072124834e-06, "loss": 7.781, "step": 1720 }, { "epoch": 0.01, "learning_rate": 9.99497819393429e-06, "loss": 7.9393, "step": 1730 }, { "epoch": 0.01, "learning_rate": 9.994919980443383e-06, "loss": 8.2448, "step": 1740 }, { "epoch": 0.01, "learning_rate": 9.994861431656016e-06, "loss": 7.3865, "step": 1750 }, { "epoch": 0.01, "learning_rate": 9.994802547576121e-06, "loss": 9.019, "step": 1760 }, { "epoch": 0.01, "learning_rate": 9.99474332820765e-06, "loss": 8.5024, "step": 1770 }, { "epoch": 0.01, "learning_rate": 9.994683773554578e-06, "loss": 7.8006, "step": 1780 }, { "epoch": 0.01, "learning_rate": 9.994623883620907e-06, "loss": 8.1917, "step": 1790 }, { "epoch": 0.01, "learning_rate": 9.99456365841065e-06, "loss": 7.8374, "step": 1800 }, { "epoch": 0.01, "learning_rate": 9.994503097927856e-06, "loss": 7.6676, "step": 1810 }, { "epoch": 0.02, "learning_rate": 9.994442202176587e-06, "loss": 8.0222, "step": 1820 }, { "epoch": 0.02, "learning_rate": 9.99438097116093e-06, "loss": 7.5274, "step": 1830 }, { "epoch": 0.02, "learning_rate": 9.994319404885e-06, "loss": 8.1651, "step": 1840 }, { "epoch": 0.02, "learning_rate": 9.994257503352925e-06, "loss": 7.8746, "step": 1850 }, { "epoch": 0.02, "learning_rate": 9.994195266568863e-06, "loss": 7.404, "step": 1860 }, { "epoch": 0.02, "learning_rate": 9.99413269453699e-06, "loss": 7.7201, "step": 1870 }, { "epoch": 0.02, "learning_rate": 9.994069787261507e-06, "loss": 7.8939, "step": 1880 }, { "epoch": 0.02, "learning_rate": 9.994006544746637e-06, "loss": 7.6505, "step": 1890 }, { "epoch": 0.02, "learning_rate": 9.993942966996625e-06, "loss": 8.2433, "step": 1900 }, { "epoch": 0.02, "learning_rate": 9.99387905401574e-06, "loss": 8.4485, "step": 1910 }, { "epoch": 0.02, "learning_rate": 9.993814805808271e-06, "loss": 7.9823, "step": 1920 }, { "epoch": 0.02, "learning_rate": 9.993750222378531e-06, "loss": 8.0406, "step": 1930 }, { "epoch": 0.02, "learning_rate": 9.993685303730858e-06, "loss": 7.6911, "step": 1940 }, { "epoch": 0.02, "learning_rate": 9.993620049869604e-06, "loss": 7.1978, "step": 1950 }, { "epoch": 0.02, "learning_rate": 9.993554460799153e-06, "loss": 8.0603, "step": 1960 }, { "epoch": 0.02, "learning_rate": 9.993488536523909e-06, "loss": 8.2429, "step": 1970 }, { "epoch": 0.02, "learning_rate": 9.993422277048295e-06, "loss": 8.5261, "step": 1980 }, { "epoch": 0.02, "learning_rate": 9.99335568237676e-06, "loss": 7.1479, "step": 1990 }, { "epoch": 0.02, "learning_rate": 9.993288752513772e-06, "loss": 7.6676, "step": 2000 }, { "epoch": 0.02, "learning_rate": 9.99322148746383e-06, "loss": 7.3667, "step": 2010 }, { "epoch": 0.02, "learning_rate": 9.993153887231441e-06, "loss": 7.9021, "step": 2020 }, { "epoch": 0.02, "learning_rate": 9.993085951821148e-06, "loss": 8.3443, "step": 2030 }, { "epoch": 0.02, "learning_rate": 9.993017681237512e-06, "loss": 6.9546, "step": 2040 }, { "epoch": 0.02, "learning_rate": 9.992949075485113e-06, "loss": 6.6433, "step": 2050 }, { "epoch": 0.02, "learning_rate": 9.992880134568559e-06, "loss": 7.4893, "step": 2060 }, { "epoch": 0.02, "learning_rate": 9.992810858492474e-06, "loss": 7.2082, "step": 2070 }, { "epoch": 0.02, "learning_rate": 9.992741247261513e-06, "loss": 7.7031, "step": 2080 }, { "epoch": 0.02, "learning_rate": 9.992671300880344e-06, "loss": 8.1719, "step": 2090 }, { "epoch": 0.02, "learning_rate": 9.992601019353665e-06, "loss": 7.5335, "step": 2100 }, { "epoch": 0.02, "learning_rate": 9.992530402686195e-06, "loss": 7.5347, "step": 2110 }, { "epoch": 0.02, "learning_rate": 9.992459450882674e-06, "loss": 6.8621, "step": 2120 }, { "epoch": 0.02, "learning_rate": 9.992388163947862e-06, "loss": 8.3943, "step": 2130 }, { "epoch": 0.02, "learning_rate": 9.992316541886546e-06, "loss": 8.5613, "step": 2140 }, { "epoch": 0.02, "learning_rate": 9.992244584703534e-06, "loss": 8.1814, "step": 2150 }, { "epoch": 0.02, "learning_rate": 9.992172292403657e-06, "loss": 8.6, "step": 2160 }, { "epoch": 0.02, "learning_rate": 9.992099664991766e-06, "loss": 8.1691, "step": 2170 }, { "epoch": 0.02, "learning_rate": 9.992026702472739e-06, "loss": 7.3712, "step": 2180 }, { "epoch": 0.02, "learning_rate": 9.99195340485147e-06, "loss": 7.6489, "step": 2190 }, { "epoch": 0.02, "learning_rate": 9.991879772132882e-06, "loss": 8.0067, "step": 2200 }, { "epoch": 0.02, "learning_rate": 9.99180580432192e-06, "loss": 8.2308, "step": 2210 }, { "epoch": 0.02, "learning_rate": 9.991731501423543e-06, "loss": 7.7733, "step": 2220 }, { "epoch": 0.02, "learning_rate": 9.991656863442743e-06, "loss": 7.6135, "step": 2230 }, { "epoch": 0.02, "learning_rate": 9.99158189038453e-06, "loss": 7.343, "step": 2240 }, { "epoch": 0.02, "learning_rate": 9.991506582253937e-06, "loss": 7.4861, "step": 2250 }, { "epoch": 0.02, "learning_rate": 9.991430939056015e-06, "loss": 7.2194, "step": 2260 }, { "epoch": 0.02, "learning_rate": 9.991354960795848e-06, "loss": 7.5862, "step": 2270 }, { "epoch": 0.02, "learning_rate": 9.991278647478533e-06, "loss": 7.2152, "step": 2280 }, { "epoch": 0.02, "learning_rate": 9.991201999109193e-06, "loss": 6.9332, "step": 2290 }, { "epoch": 0.02, "learning_rate": 9.991125015692973e-06, "loss": 7.2903, "step": 2300 }, { "epoch": 0.02, "learning_rate": 9.991047697235042e-06, "loss": 7.9582, "step": 2310 }, { "epoch": 0.02, "learning_rate": 9.99097004374059e-06, "loss": 8.7212, "step": 2320 }, { "epoch": 0.02, "learning_rate": 9.990892055214826e-06, "loss": 7.8376, "step": 2330 }, { "epoch": 0.02, "learning_rate": 9.990813731662989e-06, "loss": 7.1927, "step": 2340 }, { "epoch": 0.02, "learning_rate": 9.990735073090339e-06, "loss": 7.8084, "step": 2350 }, { "epoch": 0.02, "learning_rate": 9.990656079502149e-06, "loss": 7.7239, "step": 2360 }, { "epoch": 0.02, "learning_rate": 9.990576750903727e-06, "loss": 8.0591, "step": 2370 }, { "epoch": 0.02, "learning_rate": 9.990497087300396e-06, "loss": 8.4114, "step": 2380 }, { "epoch": 0.02, "learning_rate": 9.990417088697506e-06, "loss": 8.2667, "step": 2390 }, { "epoch": 0.02, "learning_rate": 9.990336755100425e-06, "loss": 7.3931, "step": 2400 }, { "epoch": 0.02, "learning_rate": 9.990256086514545e-06, "loss": 7.4332, "step": 2410 }, { "epoch": 0.02, "learning_rate": 9.990175082945284e-06, "loss": 7.7686, "step": 2420 }, { "epoch": 0.02, "learning_rate": 9.990093744398075e-06, "loss": 7.549, "step": 2430 }, { "epoch": 0.02, "learning_rate": 9.990012070878383e-06, "loss": 7.8469, "step": 2440 }, { "epoch": 0.02, "learning_rate": 9.989930062391686e-06, "loss": 7.8259, "step": 2450 }, { "epoch": 0.02, "learning_rate": 9.989847718943492e-06, "loss": 8.2772, "step": 2460 }, { "epoch": 0.02, "learning_rate": 9.98976504053933e-06, "loss": 7.5501, "step": 2470 }, { "epoch": 0.02, "learning_rate": 9.989682027184746e-06, "loss": 8.1939, "step": 2480 }, { "epoch": 0.02, "learning_rate": 9.989598678885314e-06, "loss": 8.2664, "step": 2490 }, { "epoch": 0.02, "learning_rate": 9.98951499564663e-06, "loss": 7.9766, "step": 2500 }, { "epoch": 0.02, "learning_rate": 9.98943097747431e-06, "loss": 7.4688, "step": 2510 }, { "epoch": 0.02, "learning_rate": 9.989346624373996e-06, "loss": 7.1791, "step": 2520 }, { "epoch": 0.02, "learning_rate": 9.989261936351348e-06, "loss": 7.9266, "step": 2530 }, { "epoch": 0.02, "learning_rate": 9.989176913412051e-06, "loss": 8.2659, "step": 2540 }, { "epoch": 0.02, "learning_rate": 9.989091555561815e-06, "loss": 7.4337, "step": 2550 }, { "epoch": 0.02, "learning_rate": 9.989005862806367e-06, "loss": 7.9989, "step": 2560 }, { "epoch": 0.02, "learning_rate": 9.98891983515146e-06, "loss": 7.8748, "step": 2570 }, { "epoch": 0.02, "learning_rate": 9.988833472602872e-06, "loss": 7.4328, "step": 2580 }, { "epoch": 0.02, "learning_rate": 9.988746775166395e-06, "loss": 7.9191, "step": 2590 }, { "epoch": 0.02, "learning_rate": 9.988659742847853e-06, "loss": 7.4705, "step": 2600 }, { "epoch": 0.02, "learning_rate": 9.988572375653084e-06, "loss": 7.2209, "step": 2610 }, { "epoch": 0.02, "learning_rate": 9.988484673587958e-06, "loss": 7.8061, "step": 2620 }, { "epoch": 0.02, "learning_rate": 9.988396636658359e-06, "loss": 7.8447, "step": 2630 }, { "epoch": 0.02, "learning_rate": 9.988308264870197e-06, "loss": 7.1011, "step": 2640 }, { "epoch": 0.02, "learning_rate": 9.988219558229404e-06, "loss": 7.6189, "step": 2650 }, { "epoch": 0.02, "learning_rate": 9.988130516741937e-06, "loss": 7.5161, "step": 2660 }, { "epoch": 0.02, "learning_rate": 9.988041140413768e-06, "loss": 8.1498, "step": 2670 }, { "epoch": 0.02, "learning_rate": 9.987951429250903e-06, "loss": 7.756, "step": 2680 }, { "epoch": 0.02, "learning_rate": 9.987861383259359e-06, "loss": 7.3466, "step": 2690 }, { "epoch": 0.02, "learning_rate": 9.987771002445183e-06, "loss": 7.614, "step": 2700 }, { "epoch": 0.02, "learning_rate": 9.987680286814442e-06, "loss": 8.0367, "step": 2710 }, { "epoch": 0.02, "learning_rate": 9.987589236373227e-06, "loss": 7.7962, "step": 2720 }, { "epoch": 0.02, "learning_rate": 9.987497851127646e-06, "loss": 7.3361, "step": 2730 }, { "epoch": 0.02, "learning_rate": 9.987406131083834e-06, "loss": 7.0697, "step": 2740 }, { "epoch": 0.02, "learning_rate": 9.987314076247951e-06, "loss": 7.7235, "step": 2750 }, { "epoch": 0.02, "learning_rate": 9.987221686626177e-06, "loss": 6.5847, "step": 2760 }, { "epoch": 0.02, "learning_rate": 9.987128962224708e-06, "loss": 7.5315, "step": 2770 }, { "epoch": 0.02, "learning_rate": 9.987035903049776e-06, "loss": 8.1905, "step": 2780 }, { "epoch": 0.02, "learning_rate": 9.98694250910762e-06, "loss": 7.3108, "step": 2790 }, { "epoch": 0.02, "learning_rate": 9.986848780404516e-06, "loss": 7.9112, "step": 2800 }, { "epoch": 0.02, "learning_rate": 9.986754716946752e-06, "loss": 7.4052, "step": 2810 }, { "epoch": 0.02, "learning_rate": 9.986660318740644e-06, "loss": 7.5958, "step": 2820 }, { "epoch": 0.02, "learning_rate": 9.986565585792527e-06, "loss": 7.4838, "step": 2830 }, { "epoch": 0.02, "learning_rate": 9.98647051810876e-06, "loss": 7.5265, "step": 2840 }, { "epoch": 0.02, "learning_rate": 9.986375115695729e-06, "loss": 7.2335, "step": 2850 }, { "epoch": 0.02, "learning_rate": 9.986279378559833e-06, "loss": 7.5092, "step": 2860 }, { "epoch": 0.02, "learning_rate": 9.986183306707499e-06, "loss": 7.2857, "step": 2870 }, { "epoch": 0.02, "learning_rate": 9.98608690014518e-06, "loss": 7.9391, "step": 2880 }, { "epoch": 0.02, "learning_rate": 9.985990158879345e-06, "loss": 7.9113, "step": 2890 }, { "epoch": 0.02, "learning_rate": 9.985893082916488e-06, "loss": 7.8935, "step": 2900 }, { "epoch": 0.02, "learning_rate": 9.985795672263126e-06, "loss": 7.9239, "step": 2910 }, { "epoch": 0.02, "learning_rate": 9.985697926925796e-06, "loss": 7.1782, "step": 2920 }, { "epoch": 0.02, "learning_rate": 9.985599846911061e-06, "loss": 6.9839, "step": 2930 }, { "epoch": 0.02, "learning_rate": 9.985501432225506e-06, "loss": 7.729, "step": 2940 }, { "epoch": 0.02, "learning_rate": 9.985402682875736e-06, "loss": 7.0516, "step": 2950 }, { "epoch": 0.02, "learning_rate": 9.98530359886838e-06, "loss": 7.0881, "step": 2960 }, { "epoch": 0.02, "learning_rate": 9.985204180210088e-06, "loss": 7.3982, "step": 2970 }, { "epoch": 0.02, "learning_rate": 9.985104426907536e-06, "loss": 6.7128, "step": 2980 }, { "epoch": 0.02, "learning_rate": 9.98500433896742e-06, "loss": 7.3192, "step": 2990 }, { "epoch": 0.02, "learning_rate": 9.984903916396456e-06, "loss": 7.3362, "step": 3000 }, { "epoch": 0.02, "learning_rate": 9.984803159201385e-06, "loss": 7.1151, "step": 3010 }, { "epoch": 0.02, "learning_rate": 9.984702067388976e-06, "loss": 7.4494, "step": 3020 }, { "epoch": 0.02, "learning_rate": 9.98460064096601e-06, "loss": 7.5787, "step": 3030 }, { "epoch": 0.03, "learning_rate": 9.984498879939297e-06, "loss": 7.257, "step": 3040 }, { "epoch": 0.03, "learning_rate": 9.984396784315668e-06, "loss": 7.6891, "step": 3050 }, { "epoch": 0.03, "learning_rate": 9.984294354101978e-06, "loss": 7.9083, "step": 3060 }, { "epoch": 0.03, "learning_rate": 9.9841915893051e-06, "loss": 7.8694, "step": 3070 }, { "epoch": 0.03, "learning_rate": 9.984088489931933e-06, "loss": 6.7238, "step": 3080 }, { "epoch": 0.03, "learning_rate": 9.9839850559894e-06, "loss": 7.0027, "step": 3090 }, { "epoch": 0.03, "learning_rate": 9.98388128748444e-06, "loss": 8.1018, "step": 3100 }, { "epoch": 0.03, "learning_rate": 9.983777184424024e-06, "loss": 8.0538, "step": 3110 }, { "epoch": 0.03, "learning_rate": 9.983672746815137e-06, "loss": 7.7142, "step": 3120 }, { "epoch": 0.03, "learning_rate": 9.983567974664788e-06, "loss": 8.1987, "step": 3130 }, { "epoch": 0.03, "learning_rate": 9.983462867980014e-06, "loss": 7.5198, "step": 3140 }, { "epoch": 0.03, "learning_rate": 9.98335742676787e-06, "loss": 7.3459, "step": 3150 }, { "epoch": 0.03, "learning_rate": 9.983251651035431e-06, "loss": 7.052, "step": 3160 }, { "epoch": 0.03, "learning_rate": 9.9831455407898e-06, "loss": 7.6776, "step": 3170 }, { "epoch": 0.03, "learning_rate": 9.983039096038099e-06, "loss": 7.3323, "step": 3180 }, { "epoch": 0.03, "learning_rate": 9.982932316787474e-06, "loss": 7.835, "step": 3190 }, { "epoch": 0.03, "learning_rate": 9.982825203045092e-06, "loss": 7.9045, "step": 3200 }, { "epoch": 0.03, "learning_rate": 9.982717754818142e-06, "loss": 7.2979, "step": 3210 }, { "epoch": 0.03, "learning_rate": 9.982609972113841e-06, "loss": 7.061, "step": 3220 }, { "epoch": 0.03, "learning_rate": 9.982501854939421e-06, "loss": 6.9841, "step": 3230 }, { "epoch": 0.03, "learning_rate": 9.982393403302141e-06, "loss": 7.4502, "step": 3240 }, { "epoch": 0.03, "learning_rate": 9.982284617209279e-06, "loss": 7.5982, "step": 3250 }, { "epoch": 0.03, "learning_rate": 9.982175496668139e-06, "loss": 6.6129, "step": 3260 }, { "epoch": 0.03, "learning_rate": 9.982066041686046e-06, "loss": 7.8783, "step": 3270 }, { "epoch": 0.03, "learning_rate": 9.981956252270349e-06, "loss": 7.5035, "step": 3280 }, { "epoch": 0.03, "learning_rate": 9.981846128428414e-06, "loss": 8.4631, "step": 3290 }, { "epoch": 0.03, "learning_rate": 9.981735670167637e-06, "loss": 7.4768, "step": 3300 }, { "epoch": 0.03, "learning_rate": 9.98162487749543e-06, "loss": 7.1797, "step": 3310 }, { "epoch": 0.03, "learning_rate": 9.981513750419232e-06, "loss": 7.9828, "step": 3320 }, { "epoch": 0.03, "learning_rate": 9.9814022889465e-06, "loss": 6.9073, "step": 3330 }, { "epoch": 0.03, "learning_rate": 9.98129049308472e-06, "loss": 7.3246, "step": 3340 }, { "epoch": 0.03, "learning_rate": 9.981178362841398e-06, "loss": 6.9554, "step": 3350 }, { "epoch": 0.03, "learning_rate": 9.981065898224054e-06, "loss": 6.8997, "step": 3360 }, { "epoch": 0.03, "learning_rate": 9.980953099240243e-06, "loss": 7.5347, "step": 3370 }, { "epoch": 0.03, "learning_rate": 9.980839965897536e-06, "loss": 8.1853, "step": 3380 }, { "epoch": 0.03, "learning_rate": 9.980726498203524e-06, "loss": 7.435, "step": 3390 }, { "epoch": 0.03, "learning_rate": 9.980612696165829e-06, "loss": 7.9127, "step": 3400 }, { "epoch": 0.03, "learning_rate": 9.980498559792086e-06, "loss": 7.305, "step": 3410 }, { "epoch": 0.03, "learning_rate": 9.980384089089961e-06, "loss": 7.2907, "step": 3420 }, { "epoch": 0.03, "learning_rate": 9.980269284067134e-06, "loss": 7.8238, "step": 3430 }, { "epoch": 0.03, "learning_rate": 9.980154144731312e-06, "loss": 7.4772, "step": 3440 }, { "epoch": 0.03, "learning_rate": 9.980038671090226e-06, "loss": 7.7437, "step": 3450 }, { "epoch": 0.03, "learning_rate": 9.979922863151628e-06, "loss": 7.3516, "step": 3460 }, { "epoch": 0.03, "learning_rate": 9.97980672092329e-06, "loss": 7.355, "step": 3470 }, { "epoch": 0.03, "learning_rate": 9.979690244413007e-06, "loss": 7.1702, "step": 3480 }, { "epoch": 0.03, "learning_rate": 9.9795734336286e-06, "loss": 7.7958, "step": 3490 }, { "epoch": 0.03, "learning_rate": 9.97945628857791e-06, "loss": 7.2236, "step": 3500 }, { "epoch": 0.03, "learning_rate": 9.979338809268802e-06, "loss": 7.2039, "step": 3510 }, { "epoch": 0.03, "learning_rate": 9.979220995709158e-06, "loss": 7.9259, "step": 3520 }, { "epoch": 0.03, "learning_rate": 9.979102847906892e-06, "loss": 7.8398, "step": 3530 }, { "epoch": 0.03, "learning_rate": 9.97898436586993e-06, "loss": 7.4478, "step": 3540 }, { "epoch": 0.03, "learning_rate": 9.978865549606228e-06, "loss": 7.4014, "step": 3550 }, { "epoch": 0.03, "learning_rate": 9.978746399123762e-06, "loss": 7.722, "step": 3560 }, { "epoch": 0.03, "learning_rate": 9.978626914430528e-06, "loss": 8.1063, "step": 3570 }, { "epoch": 0.03, "learning_rate": 9.97850709553455e-06, "loss": 7.0693, "step": 3580 }, { "epoch": 0.03, "learning_rate": 9.978386942443868e-06, "loss": 7.501, "step": 3590 }, { "epoch": 0.03, "learning_rate": 9.97826645516655e-06, "loss": 6.7009, "step": 3600 }, { "epoch": 0.03, "learning_rate": 9.978145633710683e-06, "loss": 7.2519, "step": 3610 }, { "epoch": 0.03, "learning_rate": 9.978024478084376e-06, "loss": 7.1002, "step": 3620 }, { "epoch": 0.03, "learning_rate": 9.977902988295764e-06, "loss": 6.5272, "step": 3630 }, { "epoch": 0.03, "learning_rate": 9.977781164353002e-06, "loss": 7.5908, "step": 3640 }, { "epoch": 0.03, "learning_rate": 9.977659006264268e-06, "loss": 7.2593, "step": 3650 }, { "epoch": 0.03, "learning_rate": 9.97753651403776e-06, "loss": 7.7882, "step": 3660 }, { "epoch": 0.03, "learning_rate": 9.977413687681704e-06, "loss": 7.0083, "step": 3670 }, { "epoch": 0.03, "learning_rate": 9.977290527204343e-06, "loss": 6.5609, "step": 3680 }, { "epoch": 0.03, "learning_rate": 9.977167032613943e-06, "loss": 8.277, "step": 3690 }, { "epoch": 0.03, "learning_rate": 9.977043203918797e-06, "loss": 7.4642, "step": 3700 }, { "epoch": 0.03, "learning_rate": 9.976919041127215e-06, "loss": 7.202, "step": 3710 }, { "epoch": 0.03, "learning_rate": 9.976794544247531e-06, "loss": 7.5123, "step": 3720 }, { "epoch": 0.03, "learning_rate": 9.976669713288106e-06, "loss": 7.0769, "step": 3730 }, { "epoch": 0.03, "learning_rate": 9.976544548257317e-06, "loss": 7.6905, "step": 3740 }, { "epoch": 0.03, "learning_rate": 9.976419049163566e-06, "loss": 7.6509, "step": 3750 }, { "epoch": 0.03, "learning_rate": 9.976293216015278e-06, "loss": 7.4508, "step": 3760 }, { "epoch": 0.03, "learning_rate": 9.976167048820898e-06, "loss": 6.994, "step": 3770 }, { "epoch": 0.03, "learning_rate": 9.9760405475889e-06, "loss": 7.6235, "step": 3780 }, { "epoch": 0.03, "learning_rate": 9.975913712327769e-06, "loss": 6.5067, "step": 3790 }, { "epoch": 0.03, "learning_rate": 9.975786543046025e-06, "loss": 7.7937, "step": 3800 }, { "epoch": 0.03, "learning_rate": 9.975659039752201e-06, "loss": 7.0586, "step": 3810 }, { "epoch": 0.03, "learning_rate": 9.975531202454858e-06, "loss": 7.0134, "step": 3820 }, { "epoch": 0.03, "learning_rate": 9.975403031162577e-06, "loss": 6.9457, "step": 3830 }, { "epoch": 0.03, "learning_rate": 9.975274525883961e-06, "loss": 6.9866, "step": 3840 }, { "epoch": 0.03, "learning_rate": 9.975145686627637e-06, "loss": 8.0182, "step": 3850 }, { "epoch": 0.03, "learning_rate": 9.975016513402252e-06, "loss": 7.0873, "step": 3860 }, { "epoch": 0.03, "learning_rate": 9.974887006216481e-06, "loss": 7.3393, "step": 3870 }, { "epoch": 0.03, "learning_rate": 9.974757165079014e-06, "loss": 6.9221, "step": 3880 }, { "epoch": 0.03, "learning_rate": 9.974626989998565e-06, "loss": 7.605, "step": 3890 }, { "epoch": 0.03, "learning_rate": 9.97449648098388e-06, "loss": 7.0438, "step": 3900 }, { "epoch": 0.03, "learning_rate": 9.974365638043712e-06, "loss": 7.3446, "step": 3910 }, { "epoch": 0.03, "learning_rate": 9.974234461186848e-06, "loss": 7.395, "step": 3920 }, { "epoch": 0.03, "learning_rate": 9.97410295042209e-06, "loss": 7.8512, "step": 3930 }, { "epoch": 0.03, "learning_rate": 9.973971105758273e-06, "loss": 7.1022, "step": 3940 }, { "epoch": 0.03, "learning_rate": 9.97383892720424e-06, "loss": 7.6017, "step": 3950 }, { "epoch": 0.03, "learning_rate": 9.973706414768868e-06, "loss": 7.57, "step": 3960 }, { "epoch": 0.03, "learning_rate": 9.97357356846105e-06, "loss": 8.2869, "step": 3970 }, { "epoch": 0.03, "learning_rate": 9.973440388289706e-06, "loss": 7.5194, "step": 3980 }, { "epoch": 0.03, "learning_rate": 9.973306874263775e-06, "loss": 7.111, "step": 3990 }, { "epoch": 0.03, "learning_rate": 9.973173026392218e-06, "loss": 7.7459, "step": 4000 }, { "epoch": 0.03, "learning_rate": 9.973038844684021e-06, "loss": 7.4221, "step": 4010 }, { "epoch": 0.03, "learning_rate": 9.972904329148192e-06, "loss": 7.0193, "step": 4020 }, { "epoch": 0.03, "learning_rate": 9.972769479793762e-06, "loss": 7.7305, "step": 4030 }, { "epoch": 0.03, "learning_rate": 9.972634296629779e-06, "loss": 7.3118, "step": 4040 }, { "epoch": 0.03, "learning_rate": 9.972498779665321e-06, "loss": 7.0704, "step": 4050 }, { "epoch": 0.03, "learning_rate": 9.972362928909482e-06, "loss": 8.1672, "step": 4060 }, { "epoch": 0.03, "learning_rate": 9.972226744371382e-06, "loss": 7.3044, "step": 4070 }, { "epoch": 0.03, "learning_rate": 9.972090226060168e-06, "loss": 7.4858, "step": 4080 }, { "epoch": 0.03, "learning_rate": 9.971953373984996e-06, "loss": 7.5737, "step": 4090 }, { "epoch": 0.03, "learning_rate": 9.971816188155056e-06, "loss": 7.1303, "step": 4100 }, { "epoch": 0.03, "learning_rate": 9.97167866857956e-06, "loss": 7.7467, "step": 4110 }, { "epoch": 0.03, "learning_rate": 9.971540815267735e-06, "loss": 7.6635, "step": 4120 }, { "epoch": 0.03, "learning_rate": 9.971402628228838e-06, "loss": 7.8511, "step": 4130 }, { "epoch": 0.03, "learning_rate": 9.97126410747214e-06, "loss": 7.2933, "step": 4140 }, { "epoch": 0.03, "learning_rate": 9.971125253006946e-06, "loss": 7.5836, "step": 4150 }, { "epoch": 0.03, "learning_rate": 9.970986064842571e-06, "loss": 7.2088, "step": 4160 }, { "epoch": 0.03, "learning_rate": 9.970846542988365e-06, "loss": 8.0864, "step": 4170 }, { "epoch": 0.03, "learning_rate": 9.970706687453689e-06, "loss": 6.8503, "step": 4180 }, { "epoch": 0.03, "learning_rate": 9.970566498247932e-06, "loss": 7.3713, "step": 4190 }, { "epoch": 0.03, "learning_rate": 9.970425975380504e-06, "loss": 8.0962, "step": 4200 }, { "epoch": 0.03, "learning_rate": 9.97028511886084e-06, "loss": 7.3701, "step": 4210 }, { "epoch": 0.03, "learning_rate": 9.970143928698393e-06, "loss": 7.1983, "step": 4220 }, { "epoch": 0.03, "learning_rate": 9.970002404902642e-06, "loss": 7.2478, "step": 4230 }, { "epoch": 0.03, "learning_rate": 9.969860547483088e-06, "loss": 8.19, "step": 4240 }, { "epoch": 0.04, "learning_rate": 9.969718356449252e-06, "loss": 7.7373, "step": 4250 }, { "epoch": 0.04, "learning_rate": 9.969575831810678e-06, "loss": 7.5701, "step": 4260 }, { "epoch": 0.04, "learning_rate": 9.969432973576936e-06, "loss": 7.1168, "step": 4270 }, { "epoch": 0.04, "learning_rate": 9.969289781757615e-06, "loss": 7.2015, "step": 4280 }, { "epoch": 0.04, "learning_rate": 9.969146256362326e-06, "loss": 7.2754, "step": 4290 }, { "epoch": 0.04, "learning_rate": 9.969002397400704e-06, "loss": 7.6981, "step": 4300 }, { "epoch": 0.04, "learning_rate": 9.968858204882407e-06, "loss": 7.4311, "step": 4310 }, { "epoch": 0.04, "learning_rate": 9.968713678817112e-06, "loss": 7.4371, "step": 4320 }, { "epoch": 0.04, "learning_rate": 9.968568819214522e-06, "loss": 7.4401, "step": 4330 }, { "epoch": 0.04, "learning_rate": 9.968423626084363e-06, "loss": 7.3939, "step": 4340 }, { "epoch": 0.04, "learning_rate": 9.968278099436376e-06, "loss": 7.2609, "step": 4350 }, { "epoch": 0.04, "learning_rate": 9.968132239280336e-06, "loss": 7.5615, "step": 4360 }, { "epoch": 0.04, "learning_rate": 9.96798604562603e-06, "loss": 7.8372, "step": 4370 }, { "epoch": 0.04, "learning_rate": 9.967839518483275e-06, "loss": 7.7795, "step": 4380 }, { "epoch": 0.04, "learning_rate": 9.967692657861902e-06, "loss": 7.4633, "step": 4390 }, { "epoch": 0.04, "learning_rate": 9.967545463771774e-06, "loss": 7.1363, "step": 4400 }, { "epoch": 0.04, "learning_rate": 9.967397936222773e-06, "loss": 6.8709, "step": 4410 }, { "epoch": 0.04, "learning_rate": 9.967250075224796e-06, "loss": 6.6448, "step": 4420 }, { "epoch": 0.04, "learning_rate": 9.967101880787774e-06, "loss": 7.0895, "step": 4430 }, { "epoch": 0.04, "learning_rate": 9.96695335292165e-06, "loss": 7.7764, "step": 4440 }, { "epoch": 0.04, "learning_rate": 9.966804491636402e-06, "loss": 6.9495, "step": 4450 }, { "epoch": 0.04, "learning_rate": 9.966655296942014e-06, "loss": 6.9843, "step": 4460 }, { "epoch": 0.04, "learning_rate": 9.966505768848507e-06, "loss": 7.3771, "step": 4470 }, { "epoch": 0.04, "learning_rate": 9.966355907365915e-06, "loss": 7.1972, "step": 4480 }, { "epoch": 0.04, "learning_rate": 9.9662057125043e-06, "loss": 7.045, "step": 4490 }, { "epoch": 0.04, "learning_rate": 9.966055184273742e-06, "loss": 7.1288, "step": 4500 }, { "epoch": 0.04, "learning_rate": 9.965904322684348e-06, "loss": 7.3145, "step": 4510 }, { "epoch": 0.04, "learning_rate": 9.965753127746245e-06, "loss": 7.5724, "step": 4520 }, { "epoch": 0.04, "learning_rate": 9.96560159946958e-06, "loss": 8.0071, "step": 4530 }, { "epoch": 0.04, "learning_rate": 9.965449737864526e-06, "loss": 6.7947, "step": 4540 }, { "epoch": 0.04, "learning_rate": 9.965297542941278e-06, "loss": 6.3782, "step": 4550 }, { "epoch": 0.04, "learning_rate": 9.965145014710052e-06, "loss": 7.0395, "step": 4560 }, { "epoch": 0.04, "learning_rate": 9.964992153181084e-06, "loss": 6.8789, "step": 4570 }, { "epoch": 0.04, "learning_rate": 9.96483895836464e-06, "loss": 7.0947, "step": 4580 }, { "epoch": 0.04, "learning_rate": 9.964685430271e-06, "loss": 7.1742, "step": 4590 }, { "epoch": 0.04, "learning_rate": 9.964531568910471e-06, "loss": 7.4616, "step": 4600 }, { "epoch": 0.04, "learning_rate": 9.964377374293383e-06, "loss": 7.3892, "step": 4610 }, { "epoch": 0.04, "learning_rate": 9.964222846430082e-06, "loss": 7.0683, "step": 4620 }, { "epoch": 0.04, "learning_rate": 9.964067985330947e-06, "loss": 6.51, "step": 4630 }, { "epoch": 0.04, "learning_rate": 9.963912791006372e-06, "loss": 7.1146, "step": 4640 }, { "epoch": 0.04, "learning_rate": 9.963757263466771e-06, "loss": 7.3508, "step": 4650 }, { "epoch": 0.04, "learning_rate": 9.963601402722587e-06, "loss": 6.8537, "step": 4660 }, { "epoch": 0.04, "learning_rate": 9.963445208784282e-06, "loss": 7.1176, "step": 4670 }, { "epoch": 0.04, "learning_rate": 9.963288681662342e-06, "loss": 7.3436, "step": 4680 }, { "epoch": 0.04, "learning_rate": 9.963131821367274e-06, "loss": 6.7248, "step": 4690 }, { "epoch": 0.04, "learning_rate": 9.962974627909605e-06, "loss": 7.0997, "step": 4700 }, { "epoch": 0.04, "learning_rate": 9.962817101299891e-06, "loss": 7.2486, "step": 4710 }, { "epoch": 0.04, "learning_rate": 9.962659241548702e-06, "loss": 7.3139, "step": 4720 }, { "epoch": 0.04, "learning_rate": 9.96250104866664e-06, "loss": 7.4697, "step": 4730 }, { "epoch": 0.04, "learning_rate": 9.962342522664319e-06, "loss": 7.1225, "step": 4740 }, { "epoch": 0.04, "learning_rate": 9.962183663552383e-06, "loss": 7.0565, "step": 4750 }, { "epoch": 0.04, "learning_rate": 9.962024471341497e-06, "loss": 7.2297, "step": 4760 }, { "epoch": 0.04, "learning_rate": 9.961864946042344e-06, "loss": 6.6926, "step": 4770 }, { "epoch": 0.04, "learning_rate": 9.961705087665635e-06, "loss": 7.1201, "step": 4780 }, { "epoch": 0.04, "learning_rate": 9.9615448962221e-06, "loss": 7.7415, "step": 4790 }, { "epoch": 0.04, "learning_rate": 9.961384371722493e-06, "loss": 7.0717, "step": 4800 }, { "epoch": 0.04, "learning_rate": 9.961223514177588e-06, "loss": 6.9712, "step": 4810 }, { "epoch": 0.04, "learning_rate": 9.961062323598185e-06, "loss": 8.0096, "step": 4820 }, { "epoch": 0.04, "learning_rate": 9.960900799995102e-06, "loss": 6.7267, "step": 4830 }, { "epoch": 0.04, "learning_rate": 9.960738943379182e-06, "loss": 7.5265, "step": 4840 }, { "epoch": 0.04, "learning_rate": 9.960576753761293e-06, "loss": 7.2657, "step": 4850 }, { "epoch": 0.04, "learning_rate": 9.96041423115232e-06, "loss": 7.5973, "step": 4860 }, { "epoch": 0.04, "learning_rate": 9.960251375563173e-06, "loss": 7.5398, "step": 4870 }, { "epoch": 0.04, "learning_rate": 9.960088187004783e-06, "loss": 6.5938, "step": 4880 }, { "epoch": 0.04, "learning_rate": 9.959924665488106e-06, "loss": 7.2152, "step": 4890 }, { "epoch": 0.04, "learning_rate": 9.95976081102412e-06, "loss": 7.0206, "step": 4900 }, { "epoch": 0.04, "learning_rate": 9.95959662362382e-06, "loss": 7.2249, "step": 4910 }, { "epoch": 0.04, "learning_rate": 9.959432103298232e-06, "loss": 7.3032, "step": 4920 }, { "epoch": 0.04, "learning_rate": 9.959267250058398e-06, "loss": 7.4783, "step": 4930 }, { "epoch": 0.04, "learning_rate": 9.959102063915383e-06, "loss": 7.5384, "step": 4940 }, { "epoch": 0.04, "learning_rate": 9.958936544880275e-06, "loss": 7.7788, "step": 4950 }, { "epoch": 0.04, "learning_rate": 9.958770692964188e-06, "loss": 7.0515, "step": 4960 }, { "epoch": 0.04, "learning_rate": 9.958604508178254e-06, "loss": 6.9126, "step": 4970 }, { "epoch": 0.04, "learning_rate": 9.958437990533628e-06, "loss": 6.9702, "step": 4980 }, { "epoch": 0.04, "learning_rate": 9.958271140041488e-06, "loss": 6.8909, "step": 4990 }, { "epoch": 0.04, "learning_rate": 9.958103956713035e-06, "loss": 7.3128, "step": 5000 }, { "epoch": 0.04, "learning_rate": 9.95793644055949e-06, "loss": 7.2242, "step": 5010 }, { "epoch": 0.04, "learning_rate": 9.9577685915921e-06, "loss": 6.6392, "step": 5020 }, { "epoch": 0.04, "learning_rate": 9.95760040982213e-06, "loss": 7.4619, "step": 5030 }, { "epoch": 0.04, "learning_rate": 9.957431895260873e-06, "loss": 7.2893, "step": 5040 }, { "epoch": 0.04, "learning_rate": 9.957263047919636e-06, "loss": 6.8977, "step": 5050 }, { "epoch": 0.04, "learning_rate": 9.957093867809757e-06, "loss": 7.059, "step": 5060 }, { "epoch": 0.04, "learning_rate": 9.956924354942592e-06, "loss": 7.1171, "step": 5070 }, { "epoch": 0.04, "learning_rate": 9.95675450932952e-06, "loss": 6.8879, "step": 5080 }, { "epoch": 0.04, "learning_rate": 9.956584330981943e-06, "loss": 6.6391, "step": 5090 }, { "epoch": 0.04, "learning_rate": 9.956413819911282e-06, "loss": 7.3774, "step": 5100 }, { "epoch": 0.04, "learning_rate": 9.956242976128984e-06, "loss": 6.8176, "step": 5110 }, { "epoch": 0.04, "learning_rate": 9.95607179964652e-06, "loss": 7.4752, "step": 5120 }, { "epoch": 0.04, "learning_rate": 9.955900290475376e-06, "loss": 7.0772, "step": 5130 }, { "epoch": 0.04, "learning_rate": 9.955728448627069e-06, "loss": 7.3474, "step": 5140 }, { "epoch": 0.04, "learning_rate": 9.955556274113134e-06, "loss": 6.864, "step": 5150 }, { "epoch": 0.04, "learning_rate": 9.955383766945126e-06, "loss": 6.7726, "step": 5160 }, { "epoch": 0.04, "learning_rate": 9.955210927134627e-06, "loss": 7.2079, "step": 5170 }, { "epoch": 0.04, "learning_rate": 9.95503775469324e-06, "loss": 7.3612, "step": 5180 }, { "epoch": 0.04, "learning_rate": 9.954864249632587e-06, "loss": 7.0475, "step": 5190 }, { "epoch": 0.04, "learning_rate": 9.954690411964315e-06, "loss": 7.3228, "step": 5200 }, { "epoch": 0.04, "learning_rate": 9.954516241700097e-06, "loss": 7.3021, "step": 5210 }, { "epoch": 0.04, "learning_rate": 9.954341738851623e-06, "loss": 7.3518, "step": 5220 }, { "epoch": 0.04, "learning_rate": 9.954166903430604e-06, "loss": 7.0864, "step": 5230 }, { "epoch": 0.04, "learning_rate": 9.953991735448781e-06, "loss": 6.8687, "step": 5240 }, { "epoch": 0.04, "learning_rate": 9.953816234917906e-06, "loss": 6.8228, "step": 5250 }, { "epoch": 0.04, "learning_rate": 9.953640401849767e-06, "loss": 8.3956, "step": 5260 }, { "epoch": 0.04, "learning_rate": 9.953464236256165e-06, "loss": 7.2605, "step": 5270 }, { "epoch": 0.04, "learning_rate": 9.953287738148923e-06, "loss": 6.5068, "step": 5280 }, { "epoch": 0.04, "learning_rate": 9.95311090753989e-06, "loss": 7.1747, "step": 5290 }, { "epoch": 0.04, "learning_rate": 9.95293374444094e-06, "loss": 6.3381, "step": 5300 }, { "epoch": 0.04, "learning_rate": 9.952756248863959e-06, "loss": 6.7142, "step": 5310 }, { "epoch": 0.04, "learning_rate": 9.952578420820868e-06, "loss": 6.5795, "step": 5320 }, { "epoch": 0.04, "learning_rate": 9.9524002603236e-06, "loss": 5.7024, "step": 5330 }, { "epoch": 0.04, "learning_rate": 9.952221767384115e-06, "loss": 8.0204, "step": 5340 }, { "epoch": 0.04, "learning_rate": 9.952042942014398e-06, "loss": 7.3602, "step": 5350 }, { "epoch": 0.04, "learning_rate": 9.951863784226448e-06, "loss": 6.4349, "step": 5360 }, { "epoch": 0.04, "learning_rate": 9.951684294032295e-06, "loss": 7.2605, "step": 5370 }, { "epoch": 0.04, "learning_rate": 9.951504471443987e-06, "loss": 7.1229, "step": 5380 }, { "epoch": 0.04, "learning_rate": 9.951324316473596e-06, "loss": 7.4539, "step": 5390 }, { "epoch": 0.04, "learning_rate": 9.951143829133213e-06, "loss": 7.6375, "step": 5400 }, { "epoch": 0.04, "learning_rate": 9.950963009434955e-06, "loss": 7.2036, "step": 5410 }, { "epoch": 0.04, "learning_rate": 9.95078185739096e-06, "loss": 6.8476, "step": 5420 }, { "epoch": 0.04, "learning_rate": 9.950600373013389e-06, "loss": 7.4792, "step": 5430 }, { "epoch": 0.04, "learning_rate": 9.950418556314422e-06, "loss": 7.3285, "step": 5440 }, { "epoch": 0.04, "learning_rate": 9.950236407306266e-06, "loss": 7.8843, "step": 5450 }, { "epoch": 0.05, "learning_rate": 9.95005392600115e-06, "loss": 7.1577, "step": 5460 }, { "epoch": 0.05, "learning_rate": 9.94987111241132e-06, "loss": 7.2949, "step": 5470 }, { "epoch": 0.05, "learning_rate": 9.949687966549049e-06, "loss": 7.2587, "step": 5480 }, { "epoch": 0.05, "learning_rate": 9.94950448842663e-06, "loss": 6.6391, "step": 5490 }, { "epoch": 0.05, "learning_rate": 9.949320678056382e-06, "loss": 6.4829, "step": 5500 }, { "epoch": 0.05, "learning_rate": 9.949136535450644e-06, "loss": 7.3321, "step": 5510 }, { "epoch": 0.05, "learning_rate": 9.948952060621773e-06, "loss": 7.5048, "step": 5520 }, { "epoch": 0.05, "learning_rate": 9.948767253582157e-06, "loss": 7.024, "step": 5530 }, { "epoch": 0.05, "learning_rate": 9.948582114344199e-06, "loss": 6.7223, "step": 5540 }, { "epoch": 0.05, "learning_rate": 9.948396642920327e-06, "loss": 7.3857, "step": 5550 }, { "epoch": 0.05, "learning_rate": 9.94821083932299e-06, "loss": 7.1798, "step": 5560 }, { "epoch": 0.05, "learning_rate": 9.948024703564664e-06, "loss": 6.8272, "step": 5570 }, { "epoch": 0.05, "learning_rate": 9.947838235657841e-06, "loss": 7.4581, "step": 5580 }, { "epoch": 0.05, "learning_rate": 9.94765143561504e-06, "loss": 7.0801, "step": 5590 }, { "epoch": 0.05, "learning_rate": 9.9474643034488e-06, "loss": 6.8333, "step": 5600 }, { "epoch": 0.05, "learning_rate": 9.94727683917168e-06, "loss": 7.0868, "step": 5610 }, { "epoch": 0.05, "learning_rate": 9.947089042796268e-06, "loss": 6.4297, "step": 5620 }, { "epoch": 0.05, "learning_rate": 9.94690091433517e-06, "loss": 7.5378, "step": 5630 }, { "epoch": 0.05, "learning_rate": 9.94671245380101e-06, "loss": 6.7419, "step": 5640 }, { "epoch": 0.05, "learning_rate": 9.946523661206443e-06, "loss": 6.2301, "step": 5650 }, { "epoch": 0.05, "learning_rate": 9.946334536564142e-06, "loss": 6.7965, "step": 5660 }, { "epoch": 0.05, "learning_rate": 9.9461450798868e-06, "loss": 6.6812, "step": 5670 }, { "epoch": 0.05, "learning_rate": 9.94595529118714e-06, "loss": 7.5431, "step": 5680 }, { "epoch": 0.05, "learning_rate": 9.945765170477895e-06, "loss": 6.795, "step": 5690 }, { "epoch": 0.05, "learning_rate": 9.94557471777183e-06, "loss": 6.9063, "step": 5700 }, { "epoch": 0.05, "learning_rate": 9.945383933081733e-06, "loss": 7.093, "step": 5710 }, { "epoch": 0.05, "learning_rate": 9.945192816420408e-06, "loss": 6.8102, "step": 5720 }, { "epoch": 0.05, "learning_rate": 9.945001367800684e-06, "loss": 6.3634, "step": 5730 }, { "epoch": 0.05, "learning_rate": 9.944809587235413e-06, "loss": 7.4217, "step": 5740 }, { "epoch": 0.05, "learning_rate": 9.944617474737469e-06, "loss": 7.7156, "step": 5750 }, { "epoch": 0.05, "learning_rate": 9.944425030319747e-06, "loss": 7.1197, "step": 5760 }, { "epoch": 0.05, "learning_rate": 9.944232253995167e-06, "loss": 6.6418, "step": 5770 }, { "epoch": 0.05, "learning_rate": 9.944039145776669e-06, "loss": 7.0049, "step": 5780 }, { "epoch": 0.05, "learning_rate": 9.943845705677214e-06, "loss": 7.0902, "step": 5790 }, { "epoch": 0.05, "learning_rate": 9.943651933709788e-06, "loss": 7.4111, "step": 5800 }, { "epoch": 0.05, "learning_rate": 9.943457829887403e-06, "loss": 6.5789, "step": 5810 }, { "epoch": 0.05, "learning_rate": 9.94326339422308e-06, "loss": 7.3982, "step": 5820 }, { "epoch": 0.05, "learning_rate": 9.943068626729876e-06, "loss": 6.3867, "step": 5830 }, { "epoch": 0.05, "learning_rate": 9.942873527420869e-06, "loss": 6.4358, "step": 5840 }, { "epoch": 0.05, "learning_rate": 9.942678096309149e-06, "loss": 6.1449, "step": 5850 }, { "epoch": 0.05, "learning_rate": 9.942482333407837e-06, "loss": 6.5979, "step": 5860 }, { "epoch": 0.05, "learning_rate": 9.942286238730076e-06, "loss": 6.1476, "step": 5870 }, { "epoch": 0.05, "learning_rate": 9.942089812289027e-06, "loss": 7.0635, "step": 5880 }, { "epoch": 0.05, "learning_rate": 9.941893054097875e-06, "loss": 7.2342, "step": 5890 }, { "epoch": 0.05, "learning_rate": 9.941695964169831e-06, "loss": 7.231, "step": 5900 }, { "epoch": 0.05, "learning_rate": 9.941498542518122e-06, "loss": 7.4901, "step": 5910 }, { "epoch": 0.05, "learning_rate": 9.941300789156003e-06, "loss": 7.1475, "step": 5920 }, { "epoch": 0.05, "learning_rate": 9.941102704096748e-06, "loss": 6.6626, "step": 5930 }, { "epoch": 0.05, "learning_rate": 9.940904287353651e-06, "loss": 6.6162, "step": 5940 }, { "epoch": 0.05, "learning_rate": 9.940705538940037e-06, "loss": 7.692, "step": 5950 }, { "epoch": 0.05, "learning_rate": 9.94050645886924e-06, "loss": 7.0178, "step": 5960 }, { "epoch": 0.05, "learning_rate": 9.940307047154632e-06, "loss": 6.5582, "step": 5970 }, { "epoch": 0.05, "learning_rate": 9.940107303809592e-06, "loss": 6.8362, "step": 5980 }, { "epoch": 0.05, "learning_rate": 9.939907228847533e-06, "loss": 7.2002, "step": 5990 }, { "epoch": 0.05, "learning_rate": 9.939706822281883e-06, "loss": 7.4561, "step": 6000 }, { "epoch": 0.05, "learning_rate": 9.939506084126096e-06, "loss": 7.3685, "step": 6010 }, { "epoch": 0.05, "learning_rate": 9.939305014393646e-06, "loss": 6.7709, "step": 6020 }, { "epoch": 0.05, "learning_rate": 9.939103613098032e-06, "loss": 6.8898, "step": 6030 }, { "epoch": 0.05, "learning_rate": 9.938901880252771e-06, "loss": 6.7448, "step": 6040 }, { "epoch": 0.05, "learning_rate": 9.938699815871407e-06, "loss": 8.3228, "step": 6050 }, { "epoch": 0.05, "learning_rate": 9.938497419967504e-06, "loss": 6.849, "step": 6060 }, { "epoch": 0.05, "learning_rate": 9.938294692554646e-06, "loss": 6.6746, "step": 6070 }, { "epoch": 0.05, "learning_rate": 9.938091633646443e-06, "loss": 6.9401, "step": 6080 }, { "epoch": 0.05, "learning_rate": 9.937888243256526e-06, "loss": 6.6807, "step": 6090 }, { "epoch": 0.05, "learning_rate": 9.93768452139855e-06, "loss": 6.9072, "step": 6100 }, { "epoch": 0.05, "learning_rate": 9.937480468086188e-06, "loss": 7.4432, "step": 6110 }, { "epoch": 0.05, "learning_rate": 9.937276083333138e-06, "loss": 7.0088, "step": 6120 }, { "epoch": 0.05, "learning_rate": 9.93707136715312e-06, "loss": 7.2001, "step": 6130 }, { "epoch": 0.05, "learning_rate": 9.936866319559874e-06, "loss": 7.2015, "step": 6140 }, { "epoch": 0.05, "learning_rate": 9.936660940567168e-06, "loss": 6.9338, "step": 6150 }, { "epoch": 0.05, "learning_rate": 9.936455230188788e-06, "loss": 6.9562, "step": 6160 }, { "epoch": 0.05, "learning_rate": 9.936249188438541e-06, "loss": 7.3782, "step": 6170 }, { "epoch": 0.05, "learning_rate": 9.936042815330258e-06, "loss": 7.0981, "step": 6180 }, { "epoch": 0.05, "learning_rate": 9.935836110877794e-06, "loss": 7.0212, "step": 6190 }, { "epoch": 0.05, "learning_rate": 9.935629075095023e-06, "loss": 6.839, "step": 6200 }, { "epoch": 0.05, "learning_rate": 9.935421707995844e-06, "loss": 7.3795, "step": 6210 }, { "epoch": 0.05, "learning_rate": 9.935214009594175e-06, "loss": 6.9564, "step": 6220 }, { "epoch": 0.05, "learning_rate": 9.935005979903962e-06, "loss": 6.857, "step": 6230 }, { "epoch": 0.05, "learning_rate": 9.934797618939165e-06, "loss": 7.0895, "step": 6240 }, { "epoch": 0.05, "learning_rate": 9.934588926713776e-06, "loss": 7.1592, "step": 6250 }, { "epoch": 0.05, "learning_rate": 9.934379903241798e-06, "loss": 7.1225, "step": 6260 }, { "epoch": 0.05, "learning_rate": 9.934170548537268e-06, "loss": 6.4613, "step": 6270 }, { "epoch": 0.05, "learning_rate": 9.933960862614234e-06, "loss": 7.2915, "step": 6280 }, { "epoch": 0.05, "learning_rate": 9.933750845486776e-06, "loss": 6.2184, "step": 6290 }, { "epoch": 0.05, "learning_rate": 9.93354049716899e-06, "loss": 7.7901, "step": 6300 }, { "epoch": 0.05, "learning_rate": 9.933329817674997e-06, "loss": 7.5837, "step": 6310 }, { "epoch": 0.05, "learning_rate": 9.933118807018937e-06, "loss": 7.0915, "step": 6320 }, { "epoch": 0.05, "learning_rate": 9.932907465214977e-06, "loss": 7.126, "step": 6330 }, { "epoch": 0.05, "learning_rate": 9.932695792277306e-06, "loss": 6.7402, "step": 6340 }, { "epoch": 0.05, "learning_rate": 9.932483788220126e-06, "loss": 7.5031, "step": 6350 }, { "epoch": 0.05, "learning_rate": 9.932271453057676e-06, "loss": 6.8347, "step": 6360 }, { "epoch": 0.05, "learning_rate": 9.932058786804205e-06, "loss": 6.829, "step": 6370 }, { "epoch": 0.05, "learning_rate": 9.93184578947399e-06, "loss": 6.6684, "step": 6380 }, { "epoch": 0.05, "learning_rate": 9.931632461081329e-06, "loss": 7.013, "step": 6390 }, { "epoch": 0.05, "learning_rate": 9.931418801640542e-06, "loss": 7.1999, "step": 6400 }, { "epoch": 0.05, "learning_rate": 9.931204811165972e-06, "loss": 6.3116, "step": 6410 }, { "epoch": 0.05, "learning_rate": 9.930990489671983e-06, "loss": 6.9823, "step": 6420 }, { "epoch": 0.05, "learning_rate": 9.930775837172962e-06, "loss": 6.8996, "step": 6430 }, { "epoch": 0.05, "learning_rate": 9.930560853683318e-06, "loss": 6.8404, "step": 6440 }, { "epoch": 0.05, "learning_rate": 9.930345539217483e-06, "loss": 7.0114, "step": 6450 }, { "epoch": 0.05, "learning_rate": 9.93012989378991e-06, "loss": 6.4016, "step": 6460 }, { "epoch": 0.05, "learning_rate": 9.929913917415075e-06, "loss": 7.2004, "step": 6470 }, { "epoch": 0.05, "learning_rate": 9.929697610107474e-06, "loss": 6.8805, "step": 6480 }, { "epoch": 0.05, "learning_rate": 9.92948097188163e-06, "loss": 6.6169, "step": 6490 }, { "epoch": 0.05, "learning_rate": 9.929264002752084e-06, "loss": 6.9283, "step": 6500 }, { "epoch": 0.05, "learning_rate": 9.9290467027334e-06, "loss": 6.726, "step": 6510 }, { "epoch": 0.05, "learning_rate": 9.928829071840168e-06, "loss": 6.2685, "step": 6520 }, { "epoch": 0.05, "learning_rate": 9.928611110086992e-06, "loss": 6.735, "step": 6530 }, { "epoch": 0.05, "learning_rate": 9.928392817488509e-06, "loss": 7.1224, "step": 6540 }, { "epoch": 0.05, "learning_rate": 9.928174194059366e-06, "loss": 6.4317, "step": 6550 }, { "epoch": 0.05, "learning_rate": 9.927955239814244e-06, "loss": 7.687, "step": 6560 }, { "epoch": 0.05, "learning_rate": 9.927735954767836e-06, "loss": 6.5593, "step": 6570 }, { "epoch": 0.05, "learning_rate": 9.927516338934869e-06, "loss": 7.063, "step": 6580 }, { "epoch": 0.05, "learning_rate": 9.927296392330077e-06, "loss": 7.3553, "step": 6590 }, { "epoch": 0.05, "learning_rate": 9.92707611496823e-06, "loss": 6.9652, "step": 6600 }, { "epoch": 0.05, "learning_rate": 9.926855506864114e-06, "loss": 7.1946, "step": 6610 }, { "epoch": 0.05, "learning_rate": 9.926634568032538e-06, "loss": 6.8457, "step": 6620 }, { "epoch": 0.05, "learning_rate": 9.926413298488332e-06, "loss": 7.6429, "step": 6630 }, { "epoch": 0.05, "learning_rate": 9.926191698246347e-06, "loss": 7.239, "step": 6640 }, { "epoch": 0.05, "learning_rate": 9.925969767321461e-06, "loss": 6.8714, "step": 6650 }, { "epoch": 0.05, "learning_rate": 9.925747505728574e-06, "loss": 7.4418, "step": 6660 }, { "epoch": 0.06, "learning_rate": 9.925524913482603e-06, "loss": 6.5029, "step": 6670 }, { "epoch": 0.06, "learning_rate": 9.92530199059849e-06, "loss": 7.2141, "step": 6680 }, { "epoch": 0.06, "learning_rate": 9.925078737091199e-06, "loss": 6.1117, "step": 6690 }, { "epoch": 0.06, "learning_rate": 9.924855152975718e-06, "loss": 6.1646, "step": 6700 }, { "epoch": 0.06, "learning_rate": 9.924631238267056e-06, "loss": 6.5004, "step": 6710 }, { "epoch": 0.06, "learning_rate": 9.92440699298024e-06, "loss": 7.5802, "step": 6720 }, { "epoch": 0.06, "learning_rate": 9.924182417130326e-06, "loss": 6.781, "step": 6730 }, { "epoch": 0.06, "learning_rate": 9.923957510732391e-06, "loss": 7.0276, "step": 6740 }, { "epoch": 0.06, "learning_rate": 9.92373227380153e-06, "loss": 6.4772, "step": 6750 }, { "epoch": 0.06, "learning_rate": 9.92350670635286e-06, "loss": 6.464, "step": 6760 }, { "epoch": 0.06, "learning_rate": 9.923280808401529e-06, "loss": 7.0374, "step": 6770 }, { "epoch": 0.06, "learning_rate": 9.923054579962696e-06, "loss": 7.4047, "step": 6780 }, { "epoch": 0.06, "learning_rate": 9.92282802105155e-06, "loss": 7.2648, "step": 6790 }, { "epoch": 0.06, "learning_rate": 9.922601131683298e-06, "loss": 6.5999, "step": 6800 }, { "epoch": 0.06, "learning_rate": 9.922373911873168e-06, "loss": 7.0598, "step": 6810 }, { "epoch": 0.06, "learning_rate": 9.922146361636418e-06, "loss": 7.005, "step": 6820 }, { "epoch": 0.06, "learning_rate": 9.92191848098832e-06, "loss": 7.0605, "step": 6830 }, { "epoch": 0.06, "learning_rate": 9.92169026994417e-06, "loss": 7.4502, "step": 6840 }, { "epoch": 0.06, "learning_rate": 9.92146172851929e-06, "loss": 6.8048, "step": 6850 }, { "epoch": 0.06, "learning_rate": 9.92123285672902e-06, "loss": 6.6115, "step": 6860 }, { "epoch": 0.06, "learning_rate": 9.921003654588722e-06, "loss": 6.5753, "step": 6870 }, { "epoch": 0.06, "learning_rate": 9.920774122113784e-06, "loss": 6.9603, "step": 6880 }, { "epoch": 0.06, "learning_rate": 9.920544259319612e-06, "loss": 6.6107, "step": 6890 }, { "epoch": 0.06, "learning_rate": 9.92031406622164e-06, "loss": 7.2407, "step": 6900 }, { "epoch": 0.06, "learning_rate": 9.920083542835317e-06, "loss": 6.8449, "step": 6910 }, { "epoch": 0.06, "learning_rate": 9.919852689176118e-06, "loss": 6.4711, "step": 6920 }, { "epoch": 0.06, "learning_rate": 9.919621505259539e-06, "loss": 7.0961, "step": 6930 }, { "epoch": 0.06, "learning_rate": 9.919389991101099e-06, "loss": 6.3896, "step": 6940 }, { "epoch": 0.06, "learning_rate": 9.919158146716339e-06, "loss": 7.4077, "step": 6950 }, { "epoch": 0.06, "learning_rate": 9.918925972120822e-06, "loss": 6.0565, "step": 6960 }, { "epoch": 0.06, "learning_rate": 9.918693467330137e-06, "loss": 7.4547, "step": 6970 }, { "epoch": 0.06, "learning_rate": 9.918460632359888e-06, "loss": 6.8291, "step": 6980 }, { "epoch": 0.06, "learning_rate": 9.918227467225703e-06, "loss": 6.4313, "step": 6990 }, { "epoch": 0.06, "learning_rate": 9.917993971943236e-06, "loss": 7.1031, "step": 7000 }, { "epoch": 0.06, "learning_rate": 9.91776014652816e-06, "loss": 6.9737, "step": 7010 }, { "epoch": 0.06, "learning_rate": 9.917525990996176e-06, "loss": 7.188, "step": 7020 }, { "epoch": 0.06, "learning_rate": 9.917291505362993e-06, "loss": 6.741, "step": 7030 }, { "epoch": 0.06, "learning_rate": 9.91705668964436e-06, "loss": 6.2666, "step": 7040 }, { "epoch": 0.06, "learning_rate": 9.916821543856034e-06, "loss": 6.7887, "step": 7050 }, { "epoch": 0.06, "learning_rate": 9.916586068013802e-06, "loss": 7.1627, "step": 7060 }, { "epoch": 0.06, "learning_rate": 9.916350262133474e-06, "loss": 6.517, "step": 7070 }, { "epoch": 0.06, "learning_rate": 9.916114126230872e-06, "loss": 6.0501, "step": 7080 }, { "epoch": 0.06, "learning_rate": 9.915877660321854e-06, "loss": 6.9389, "step": 7090 }, { "epoch": 0.06, "learning_rate": 9.915640864422288e-06, "loss": 6.7565, "step": 7100 }, { "epoch": 0.06, "learning_rate": 9.915403738548072e-06, "loss": 6.6617, "step": 7110 }, { "epoch": 0.06, "learning_rate": 9.915166282715125e-06, "loss": 6.6968, "step": 7120 }, { "epoch": 0.06, "learning_rate": 9.914928496939384e-06, "loss": 6.7324, "step": 7130 }, { "epoch": 0.06, "learning_rate": 9.914690381236815e-06, "loss": 7.5605, "step": 7140 }, { "epoch": 0.06, "learning_rate": 9.914451935623397e-06, "loss": 6.5119, "step": 7150 }, { "epoch": 0.06, "learning_rate": 9.914213160115139e-06, "loss": 6.6359, "step": 7160 }, { "epoch": 0.06, "learning_rate": 9.91397405472807e-06, "loss": 6.7115, "step": 7170 }, { "epoch": 0.06, "learning_rate": 9.91373461947824e-06, "loss": 6.5606, "step": 7180 }, { "epoch": 0.06, "learning_rate": 9.91349485438172e-06, "loss": 6.9101, "step": 7190 }, { "epoch": 0.06, "learning_rate": 9.913254759454607e-06, "loss": 7.012, "step": 7200 }, { "epoch": 0.06, "learning_rate": 9.913014334713017e-06, "loss": 6.4435, "step": 7210 }, { "epoch": 0.06, "learning_rate": 9.91277358017309e-06, "loss": 6.02, "step": 7220 }, { "epoch": 0.06, "learning_rate": 9.912532495850986e-06, "loss": 6.9012, "step": 7230 }, { "epoch": 0.06, "learning_rate": 9.912291081762888e-06, "loss": 6.1889, "step": 7240 }, { "epoch": 0.06, "learning_rate": 9.912049337925004e-06, "loss": 6.3024, "step": 7250 }, { "epoch": 0.06, "learning_rate": 9.91180726435356e-06, "loss": 6.9245, "step": 7260 }, { "epoch": 0.06, "learning_rate": 9.911564861064804e-06, "loss": 6.1587, "step": 7270 }, { "epoch": 0.06, "learning_rate": 9.911322128075012e-06, "loss": 6.5849, "step": 7280 }, { "epoch": 0.06, "learning_rate": 9.911079065400476e-06, "loss": 6.7292, "step": 7290 }, { "epoch": 0.06, "learning_rate": 9.91083567305751e-06, "loss": 7.2251, "step": 7300 }, { "epoch": 0.06, "learning_rate": 9.910591951062456e-06, "loss": 6.2586, "step": 7310 }, { "epoch": 0.06, "learning_rate": 9.910347899431672e-06, "loss": 6.5229, "step": 7320 }, { "epoch": 0.06, "learning_rate": 9.910103518181542e-06, "loss": 6.905, "step": 7330 }, { "epoch": 0.06, "learning_rate": 9.90985880732847e-06, "loss": 6.8754, "step": 7340 }, { "epoch": 0.06, "learning_rate": 9.909613766888882e-06, "loss": 7.4228, "step": 7350 }, { "epoch": 0.06, "learning_rate": 9.909368396879228e-06, "loss": 6.9978, "step": 7360 }, { "epoch": 0.06, "learning_rate": 9.90912269731598e-06, "loss": 6.5993, "step": 7370 }, { "epoch": 0.06, "learning_rate": 9.90887666821563e-06, "loss": 6.5102, "step": 7380 }, { "epoch": 0.06, "learning_rate": 9.908630309594693e-06, "loss": 7.0278, "step": 7390 }, { "epoch": 0.06, "learning_rate": 9.908383621469708e-06, "loss": 6.8091, "step": 7400 }, { "epoch": 0.06, "learning_rate": 9.908136603857233e-06, "loss": 7.5312, "step": 7410 }, { "epoch": 0.06, "learning_rate": 9.907889256773849e-06, "loss": 6.605, "step": 7420 }, { "epoch": 0.06, "learning_rate": 9.907641580236163e-06, "loss": 7.4207, "step": 7430 }, { "epoch": 0.06, "learning_rate": 9.907393574260798e-06, "loss": 7.2701, "step": 7440 }, { "epoch": 0.06, "learning_rate": 9.907145238864403e-06, "loss": 6.9415, "step": 7450 }, { "epoch": 0.06, "learning_rate": 9.906896574063645e-06, "loss": 7.0164, "step": 7460 }, { "epoch": 0.06, "learning_rate": 9.906647579875221e-06, "loss": 7.2242, "step": 7470 }, { "epoch": 0.06, "learning_rate": 9.906398256315846e-06, "loss": 6.0981, "step": 7480 }, { "epoch": 0.06, "learning_rate": 9.906148603402251e-06, "loss": 6.5004, "step": 7490 }, { "epoch": 0.06, "learning_rate": 9.905898621151199e-06, "loss": 6.7152, "step": 7500 }, { "epoch": 0.06, "learning_rate": 9.90564830957947e-06, "loss": 7.1982, "step": 7510 }, { "epoch": 0.06, "learning_rate": 9.905397668703864e-06, "loss": 6.4529, "step": 7520 }, { "epoch": 0.06, "learning_rate": 9.905146698541211e-06, "loss": 6.2713, "step": 7530 }, { "epoch": 0.06, "learning_rate": 9.904895399108352e-06, "loss": 6.5195, "step": 7540 }, { "epoch": 0.06, "learning_rate": 9.90464377042216e-06, "loss": 7.1455, "step": 7550 }, { "epoch": 0.06, "learning_rate": 9.904391812499526e-06, "loss": 6.6528, "step": 7560 }, { "epoch": 0.06, "learning_rate": 9.904139525357363e-06, "loss": 6.9228, "step": 7570 }, { "epoch": 0.06, "learning_rate": 9.903886909012605e-06, "loss": 6.8214, "step": 7580 }, { "epoch": 0.06, "learning_rate": 9.903633963482212e-06, "loss": 7.3434, "step": 7590 }, { "epoch": 0.06, "learning_rate": 9.90338068878316e-06, "loss": 7.1225, "step": 7600 }, { "epoch": 0.06, "learning_rate": 9.903127084932453e-06, "loss": 6.8145, "step": 7610 }, { "epoch": 0.06, "learning_rate": 9.902873151947115e-06, "loss": 6.3824, "step": 7620 }, { "epoch": 0.06, "learning_rate": 9.902618889844193e-06, "loss": 7.2431, "step": 7630 }, { "epoch": 0.06, "learning_rate": 9.902364298640753e-06, "loss": 6.8984, "step": 7640 }, { "epoch": 0.06, "learning_rate": 9.902109378353884e-06, "loss": 7.2011, "step": 7650 }, { "epoch": 0.06, "learning_rate": 9.9018541290007e-06, "loss": 6.8327, "step": 7660 }, { "epoch": 0.06, "learning_rate": 9.901598550598334e-06, "loss": 6.8536, "step": 7670 }, { "epoch": 0.06, "learning_rate": 9.901342643163944e-06, "loss": 6.6431, "step": 7680 }, { "epoch": 0.06, "learning_rate": 9.90108640671471e-06, "loss": 7.2962, "step": 7690 }, { "epoch": 0.06, "learning_rate": 9.900829841267829e-06, "loss": 6.6273, "step": 7700 }, { "epoch": 0.06, "learning_rate": 9.900572946840523e-06, "loss": 7.0557, "step": 7710 }, { "epoch": 0.06, "learning_rate": 9.90031572345004e-06, "loss": 7.2508, "step": 7720 }, { "epoch": 0.06, "learning_rate": 9.900058171113644e-06, "loss": 7.3558, "step": 7730 }, { "epoch": 0.06, "learning_rate": 9.899800289848626e-06, "loss": 6.9367, "step": 7740 }, { "epoch": 0.06, "learning_rate": 9.899542079672297e-06, "loss": 7.1656, "step": 7750 }, { "epoch": 0.06, "learning_rate": 9.899283540601989e-06, "loss": 7.0296, "step": 7760 }, { "epoch": 0.06, "learning_rate": 9.899024672655056e-06, "loss": 6.6687, "step": 7770 }, { "epoch": 0.06, "learning_rate": 9.898765475848876e-06, "loss": 5.9327, "step": 7780 }, { "epoch": 0.06, "learning_rate": 9.89850595020085e-06, "loss": 6.8456, "step": 7790 }, { "epoch": 0.06, "learning_rate": 9.898246095728396e-06, "loss": 7.3348, "step": 7800 }, { "epoch": 0.06, "learning_rate": 9.89798591244896e-06, "loss": 6.2856, "step": 7810 }, { "epoch": 0.06, "learning_rate": 9.897725400380004e-06, "loss": 7.0336, "step": 7820 }, { "epoch": 0.06, "learning_rate": 9.897464559539023e-06, "loss": 7.1459, "step": 7830 }, { "epoch": 0.06, "learning_rate": 9.897203389943519e-06, "loss": 7.1902, "step": 7840 }, { "epoch": 0.06, "learning_rate": 9.896941891611026e-06, "loss": 6.657, "step": 7850 }, { "epoch": 0.06, "learning_rate": 9.8966800645591e-06, "loss": 7.0426, "step": 7860 }, { "epoch": 0.06, "learning_rate": 9.896417908805312e-06, "loss": 6.9507, "step": 7870 }, { "epoch": 0.06, "learning_rate": 9.896155424367264e-06, "loss": 6.7302, "step": 7880 }, { "epoch": 0.07, "learning_rate": 9.895892611262576e-06, "loss": 7.4404, "step": 7890 }, { "epoch": 0.07, "learning_rate": 9.895629469508887e-06, "loss": 6.5296, "step": 7900 }, { "epoch": 0.07, "learning_rate": 9.895365999123865e-06, "loss": 6.344, "step": 7910 }, { "epoch": 0.07, "learning_rate": 9.895102200125191e-06, "loss": 6.5769, "step": 7920 }, { "epoch": 0.07, "learning_rate": 9.894838072530579e-06, "loss": 7.3423, "step": 7930 }, { "epoch": 0.07, "learning_rate": 9.894573616357754e-06, "loss": 6.4263, "step": 7940 }, { "epoch": 0.07, "learning_rate": 9.89430883162447e-06, "loss": 6.6694, "step": 7950 }, { "epoch": 0.07, "learning_rate": 9.894043718348503e-06, "loss": 6.0457, "step": 7960 }, { "epoch": 0.07, "learning_rate": 9.893778276547648e-06, "loss": 7.3297, "step": 7970 }, { "epoch": 0.07, "learning_rate": 9.893512506239724e-06, "loss": 6.5543, "step": 7980 }, { "epoch": 0.07, "learning_rate": 9.893246407442571e-06, "loss": 6.2466, "step": 7990 }, { "epoch": 0.07, "learning_rate": 9.89297998017405e-06, "loss": 6.0038, "step": 8000 }, { "epoch": 0.07, "learning_rate": 9.89271322445205e-06, "loss": 6.838, "step": 8010 }, { "epoch": 0.07, "learning_rate": 9.892446140294475e-06, "loss": 6.8697, "step": 8020 }, { "epoch": 0.07, "learning_rate": 9.892178727719252e-06, "loss": 6.7308, "step": 8030 }, { "epoch": 0.07, "learning_rate": 9.891910986744334e-06, "loss": 6.6499, "step": 8040 }, { "epoch": 0.07, "learning_rate": 9.891642917387694e-06, "loss": 6.5607, "step": 8050 }, { "epoch": 0.07, "learning_rate": 9.891374519667324e-06, "loss": 6.9583, "step": 8060 }, { "epoch": 0.07, "learning_rate": 9.891105793601243e-06, "loss": 6.7207, "step": 8070 }, { "epoch": 0.07, "learning_rate": 9.890836739207492e-06, "loss": 6.8669, "step": 8080 }, { "epoch": 0.07, "learning_rate": 9.890567356504129e-06, "loss": 6.7124, "step": 8090 }, { "epoch": 0.07, "learning_rate": 9.890297645509234e-06, "loss": 6.7478, "step": 8100 }, { "epoch": 0.07, "learning_rate": 9.890027606240919e-06, "loss": 6.3942, "step": 8110 }, { "epoch": 0.07, "learning_rate": 9.889757238717307e-06, "loss": 7.5618, "step": 8120 }, { "epoch": 0.07, "learning_rate": 9.88948654295655e-06, "loss": 6.6713, "step": 8130 }, { "epoch": 0.07, "learning_rate": 9.889215518976815e-06, "loss": 6.3347, "step": 8140 }, { "epoch": 0.07, "learning_rate": 9.888944166796296e-06, "loss": 7.1938, "step": 8150 }, { "epoch": 0.07, "learning_rate": 9.888672486433211e-06, "loss": 6.2991, "step": 8160 }, { "epoch": 0.07, "learning_rate": 9.888400477905796e-06, "loss": 7.0304, "step": 8170 }, { "epoch": 0.07, "learning_rate": 9.88812814123231e-06, "loss": 6.6373, "step": 8180 }, { "epoch": 0.07, "learning_rate": 9.887855476431033e-06, "loss": 7.6639, "step": 8190 }, { "epoch": 0.07, "learning_rate": 9.887582483520268e-06, "loss": 6.9472, "step": 8200 }, { "epoch": 0.07, "learning_rate": 9.887309162518344e-06, "loss": 6.8375, "step": 8210 }, { "epoch": 0.07, "learning_rate": 9.887035513443606e-06, "loss": 6.9862, "step": 8220 }, { "epoch": 0.07, "learning_rate": 9.886761536314421e-06, "loss": 6.5557, "step": 8230 }, { "epoch": 0.07, "learning_rate": 9.886487231149186e-06, "loss": 6.4412, "step": 8240 }, { "epoch": 0.07, "learning_rate": 9.886212597966309e-06, "loss": 6.8265, "step": 8250 }, { "epoch": 0.07, "learning_rate": 9.885937636784229e-06, "loss": 6.9954, "step": 8260 }, { "epoch": 0.07, "learning_rate": 9.885662347621402e-06, "loss": 6.6607, "step": 8270 }, { "epoch": 0.07, "learning_rate": 9.885386730496306e-06, "loss": 7.024, "step": 8280 }, { "epoch": 0.07, "learning_rate": 9.885110785427447e-06, "loss": 6.5453, "step": 8290 }, { "epoch": 0.07, "learning_rate": 9.884834512433343e-06, "loss": 7.1943, "step": 8300 }, { "epoch": 0.07, "learning_rate": 9.884557911532544e-06, "loss": 7.405, "step": 8310 }, { "epoch": 0.07, "learning_rate": 9.884280982743616e-06, "loss": 6.9033, "step": 8320 }, { "epoch": 0.07, "learning_rate": 9.884003726085145e-06, "loss": 7.0905, "step": 8330 }, { "epoch": 0.07, "learning_rate": 9.88372614157575e-06, "loss": 7.0258, "step": 8340 }, { "epoch": 0.07, "learning_rate": 9.883448229234057e-06, "loss": 6.8879, "step": 8350 }, { "epoch": 0.07, "learning_rate": 9.883169989078725e-06, "loss": 6.455, "step": 8360 }, { "epoch": 0.07, "learning_rate": 9.882891421128434e-06, "loss": 7.5013, "step": 8370 }, { "epoch": 0.07, "learning_rate": 9.882612525401878e-06, "loss": 7.0143, "step": 8380 }, { "epoch": 0.07, "learning_rate": 9.882333301917783e-06, "loss": 6.6277, "step": 8390 }, { "epoch": 0.07, "learning_rate": 9.88205375069489e-06, "loss": 6.636, "step": 8400 }, { "epoch": 0.07, "learning_rate": 9.881773871751968e-06, "loss": 6.727, "step": 8410 }, { "epoch": 0.07, "learning_rate": 9.8814936651078e-06, "loss": 6.4453, "step": 8420 }, { "epoch": 0.07, "learning_rate": 9.8812131307812e-06, "loss": 6.2422, "step": 8430 }, { "epoch": 0.07, "learning_rate": 9.880932268790994e-06, "loss": 6.783, "step": 8440 }, { "epoch": 0.07, "learning_rate": 9.880651079156043e-06, "loss": 7.1129, "step": 8450 }, { "epoch": 0.07, "learning_rate": 9.880369561895216e-06, "loss": 6.5638, "step": 8460 }, { "epoch": 0.07, "learning_rate": 9.880087717027413e-06, "loss": 6.7373, "step": 8470 }, { "epoch": 0.07, "learning_rate": 9.879805544571555e-06, "loss": 5.6904, "step": 8480 }, { "epoch": 0.07, "learning_rate": 9.879523044546581e-06, "loss": 6.7884, "step": 8490 }, { "epoch": 0.07, "learning_rate": 9.879240216971457e-06, "loss": 6.9103, "step": 8500 }, { "epoch": 0.07, "learning_rate": 9.878957061865166e-06, "loss": 6.6156, "step": 8510 }, { "epoch": 0.07, "learning_rate": 9.878673579246717e-06, "loss": 6.7371, "step": 8520 }, { "epoch": 0.07, "learning_rate": 9.878389769135139e-06, "loss": 6.6102, "step": 8530 }, { "epoch": 0.07, "learning_rate": 9.878105631549483e-06, "loss": 6.7654, "step": 8540 }, { "epoch": 0.07, "learning_rate": 9.877821166508825e-06, "loss": 6.6519, "step": 8550 }, { "epoch": 0.07, "learning_rate": 9.877536374032257e-06, "loss": 6.8022, "step": 8560 }, { "epoch": 0.07, "learning_rate": 9.877251254138897e-06, "loss": 6.3306, "step": 8570 }, { "epoch": 0.07, "learning_rate": 9.876965806847886e-06, "loss": 6.1556, "step": 8580 }, { "epoch": 0.07, "learning_rate": 9.876680032178385e-06, "loss": 7.6609, "step": 8590 }, { "epoch": 0.07, "learning_rate": 9.876393930149576e-06, "loss": 6.618, "step": 8600 }, { "epoch": 0.07, "learning_rate": 9.876107500780667e-06, "loss": 6.4699, "step": 8610 }, { "epoch": 0.07, "learning_rate": 9.875820744090883e-06, "loss": 7.1732, "step": 8620 }, { "epoch": 0.07, "learning_rate": 9.875533660099475e-06, "loss": 6.4687, "step": 8630 }, { "epoch": 0.07, "learning_rate": 9.875246248825709e-06, "loss": 6.7218, "step": 8640 }, { "epoch": 0.07, "learning_rate": 9.874958510288886e-06, "loss": 6.5402, "step": 8650 }, { "epoch": 0.07, "learning_rate": 9.874670444508317e-06, "loss": 6.1306, "step": 8660 }, { "epoch": 0.07, "learning_rate": 9.874382051503338e-06, "loss": 7.2664, "step": 8670 }, { "epoch": 0.07, "learning_rate": 9.87409333129331e-06, "loss": 7.0388, "step": 8680 }, { "epoch": 0.07, "learning_rate": 9.873804283897615e-06, "loss": 6.8184, "step": 8690 }, { "epoch": 0.07, "learning_rate": 9.873514909335652e-06, "loss": 7.0678, "step": 8700 }, { "epoch": 0.07, "learning_rate": 9.873225207626851e-06, "loss": 6.4324, "step": 8710 }, { "epoch": 0.07, "learning_rate": 9.872935178790656e-06, "loss": 6.9009, "step": 8720 }, { "epoch": 0.07, "learning_rate": 9.872644822846536e-06, "loss": 7.0063, "step": 8730 }, { "epoch": 0.07, "learning_rate": 9.872354139813984e-06, "loss": 6.3265, "step": 8740 }, { "epoch": 0.07, "learning_rate": 9.87206312971251e-06, "loss": 6.9925, "step": 8750 }, { "epoch": 0.07, "learning_rate": 9.871771792561651e-06, "loss": 5.9564, "step": 8760 }, { "epoch": 0.07, "learning_rate": 9.871480128380961e-06, "loss": 6.9855, "step": 8770 }, { "epoch": 0.07, "learning_rate": 9.871188137190022e-06, "loss": 7.2041, "step": 8780 }, { "epoch": 0.07, "learning_rate": 9.870895819008432e-06, "loss": 6.7055, "step": 8790 }, { "epoch": 0.07, "learning_rate": 9.870603173855818e-06, "loss": 6.6705, "step": 8800 }, { "epoch": 0.07, "learning_rate": 9.870310201751817e-06, "loss": 6.2887, "step": 8810 }, { "epoch": 0.07, "learning_rate": 9.870016902716101e-06, "loss": 6.8016, "step": 8820 }, { "epoch": 0.07, "learning_rate": 9.869723276768357e-06, "loss": 7.2325, "step": 8830 }, { "epoch": 0.07, "learning_rate": 9.869429323928296e-06, "loss": 6.8871, "step": 8840 }, { "epoch": 0.07, "learning_rate": 9.86913504421565e-06, "loss": 5.8455, "step": 8850 }, { "epoch": 0.07, "learning_rate": 9.868840437650173e-06, "loss": 6.9491, "step": 8860 }, { "epoch": 0.07, "learning_rate": 9.868545504251641e-06, "loss": 6.4126, "step": 8870 }, { "epoch": 0.07, "learning_rate": 9.868250244039853e-06, "loss": 6.7683, "step": 8880 }, { "epoch": 0.07, "learning_rate": 9.867954657034628e-06, "loss": 6.3585, "step": 8890 }, { "epoch": 0.07, "learning_rate": 9.867658743255808e-06, "loss": 7.5981, "step": 8900 }, { "epoch": 0.07, "learning_rate": 9.867362502723259e-06, "loss": 6.1316, "step": 8910 }, { "epoch": 0.07, "learning_rate": 9.867065935456863e-06, "loss": 7.42, "step": 8920 }, { "epoch": 0.07, "learning_rate": 9.866769041476532e-06, "loss": 6.5072, "step": 8930 }, { "epoch": 0.07, "learning_rate": 9.866471820802194e-06, "loss": 6.8787, "step": 8940 }, { "epoch": 0.07, "learning_rate": 9.8661742734538e-06, "loss": 6.5659, "step": 8950 }, { "epoch": 0.07, "learning_rate": 9.865876399451323e-06, "loss": 6.7948, "step": 8960 }, { "epoch": 0.07, "learning_rate": 9.865578198814763e-06, "loss": 6.5581, "step": 8970 }, { "epoch": 0.07, "learning_rate": 9.865279671564132e-06, "loss": 6.6691, "step": 8980 }, { "epoch": 0.07, "learning_rate": 9.864980817719471e-06, "loss": 7.2888, "step": 8990 }, { "epoch": 0.07, "learning_rate": 9.864681637300842e-06, "loss": 6.069, "step": 9000 }, { "epoch": 0.07, "learning_rate": 9.864382130328329e-06, "loss": 7.2605, "step": 9010 }, { "epoch": 0.07, "learning_rate": 9.864082296822036e-06, "loss": 6.6685, "step": 9020 }, { "epoch": 0.07, "learning_rate": 9.863782136802092e-06, "loss": 7.0808, "step": 9030 }, { "epoch": 0.07, "learning_rate": 9.863481650288642e-06, "loss": 7.049, "step": 9040 }, { "epoch": 0.07, "learning_rate": 9.86318083730186e-06, "loss": 7.617, "step": 9050 }, { "epoch": 0.07, "learning_rate": 9.862879697861939e-06, "loss": 6.5187, "step": 9060 }, { "epoch": 0.07, "learning_rate": 9.862578231989091e-06, "loss": 6.3905, "step": 9070 }, { "epoch": 0.07, "learning_rate": 9.862276439703556e-06, "loss": 7.2467, "step": 9080 }, { "epoch": 0.07, "learning_rate": 9.861974321025589e-06, "loss": 7.1708, "step": 9090 }, { "epoch": 0.08, "learning_rate": 9.861671875975473e-06, "loss": 6.7129, "step": 9100 }, { "epoch": 0.08, "learning_rate": 9.86136910457351e-06, "loss": 6.936, "step": 9110 }, { "epoch": 0.08, "learning_rate": 9.861066006840025e-06, "loss": 7.1603, "step": 9120 }, { "epoch": 0.08, "learning_rate": 9.860762582795364e-06, "loss": 6.1042, "step": 9130 }, { "epoch": 0.08, "learning_rate": 9.860458832459894e-06, "loss": 6.6888, "step": 9140 }, { "epoch": 0.08, "learning_rate": 9.860154755854004e-06, "loss": 6.7373, "step": 9150 }, { "epoch": 0.08, "learning_rate": 9.859850352998108e-06, "loss": 7.0042, "step": 9160 }, { "epoch": 0.08, "learning_rate": 9.85954562391264e-06, "loss": 6.9877, "step": 9170 }, { "epoch": 0.08, "learning_rate": 9.859240568618054e-06, "loss": 6.4049, "step": 9180 }, { "epoch": 0.08, "learning_rate": 9.858935187134829e-06, "loss": 6.8204, "step": 9190 }, { "epoch": 0.08, "learning_rate": 9.858629479483464e-06, "loss": 7.4292, "step": 9200 }, { "epoch": 0.08, "learning_rate": 9.85832344568448e-06, "loss": 6.6522, "step": 9210 }, { "epoch": 0.08, "learning_rate": 9.85801708575842e-06, "loss": 6.3582, "step": 9220 }, { "epoch": 0.08, "learning_rate": 9.85771039972585e-06, "loss": 6.0671, "step": 9230 }, { "epoch": 0.08, "learning_rate": 9.857403387607357e-06, "loss": 6.891, "step": 9240 }, { "epoch": 0.08, "learning_rate": 9.85709604942355e-06, "loss": 6.8922, "step": 9250 }, { "epoch": 0.08, "learning_rate": 9.856788385195059e-06, "loss": 6.1667, "step": 9260 }, { "epoch": 0.08, "learning_rate": 9.856480394942539e-06, "loss": 6.8034, "step": 9270 }, { "epoch": 0.08, "learning_rate": 9.856172078686661e-06, "loss": 6.8173, "step": 9280 }, { "epoch": 0.08, "learning_rate": 9.855863436448126e-06, "loss": 6.2372, "step": 9290 }, { "epoch": 0.08, "learning_rate": 9.855554468247647e-06, "loss": 6.6697, "step": 9300 }, { "epoch": 0.08, "learning_rate": 9.855245174105967e-06, "loss": 6.7832, "step": 9310 }, { "epoch": 0.08, "learning_rate": 9.85493555404385e-06, "loss": 6.8855, "step": 9320 }, { "epoch": 0.08, "learning_rate": 9.854625608082079e-06, "loss": 6.6976, "step": 9330 }, { "epoch": 0.08, "learning_rate": 9.854315336241457e-06, "loss": 7.4545, "step": 9340 }, { "epoch": 0.08, "learning_rate": 9.854004738542814e-06, "loss": 7.1847, "step": 9350 }, { "epoch": 0.08, "learning_rate": 9.853693815007002e-06, "loss": 6.8029, "step": 9360 }, { "epoch": 0.08, "learning_rate": 9.853382565654887e-06, "loss": 6.3733, "step": 9370 }, { "epoch": 0.08, "learning_rate": 9.853070990507367e-06, "loss": 6.761, "step": 9380 }, { "epoch": 0.08, "learning_rate": 9.852759089585357e-06, "loss": 6.5777, "step": 9390 }, { "epoch": 0.08, "learning_rate": 9.85244686290979e-06, "loss": 6.6708, "step": 9400 }, { "epoch": 0.08, "learning_rate": 9.852134310501631e-06, "loss": 6.493, "step": 9410 }, { "epoch": 0.08, "learning_rate": 9.851821432381856e-06, "loss": 6.4516, "step": 9420 }, { "epoch": 0.08, "learning_rate": 9.851508228571469e-06, "loss": 6.2217, "step": 9430 }, { "epoch": 0.08, "learning_rate": 9.851194699091496e-06, "loss": 6.3339, "step": 9440 }, { "epoch": 0.08, "learning_rate": 9.850880843962983e-06, "loss": 6.4633, "step": 9450 }, { "epoch": 0.08, "learning_rate": 9.850566663206995e-06, "loss": 6.6379, "step": 9460 }, { "epoch": 0.08, "learning_rate": 9.850252156844628e-06, "loss": 7.0508, "step": 9470 }, { "epoch": 0.08, "learning_rate": 9.84993732489699e-06, "loss": 6.6755, "step": 9480 }, { "epoch": 0.08, "learning_rate": 9.849622167385215e-06, "loss": 6.44, "step": 9490 }, { "epoch": 0.08, "learning_rate": 9.84930668433046e-06, "loss": 6.2783, "step": 9500 }, { "epoch": 0.08, "learning_rate": 9.848990875753903e-06, "loss": 6.6477, "step": 9510 }, { "epoch": 0.08, "learning_rate": 9.848674741676741e-06, "loss": 6.6813, "step": 9520 }, { "epoch": 0.08, "learning_rate": 9.848358282120199e-06, "loss": 6.404, "step": 9530 }, { "epoch": 0.08, "learning_rate": 9.848041497105515e-06, "loss": 6.769, "step": 9540 }, { "epoch": 0.08, "learning_rate": 9.84772438665396e-06, "loss": 6.1997, "step": 9550 }, { "epoch": 0.08, "learning_rate": 9.847406950786817e-06, "loss": 5.6585, "step": 9560 }, { "epoch": 0.08, "learning_rate": 9.847089189525397e-06, "loss": 7.2975, "step": 9570 }, { "epoch": 0.08, "learning_rate": 9.846771102891026e-06, "loss": 6.1315, "step": 9580 }, { "epoch": 0.08, "learning_rate": 9.846452690905061e-06, "loss": 6.1511, "step": 9590 }, { "epoch": 0.08, "learning_rate": 9.846133953588875e-06, "loss": 6.808, "step": 9600 }, { "epoch": 0.08, "learning_rate": 9.845814890963864e-06, "loss": 6.4127, "step": 9610 }, { "epoch": 0.08, "learning_rate": 9.845495503051443e-06, "loss": 6.9207, "step": 9620 }, { "epoch": 0.08, "learning_rate": 9.845175789873057e-06, "loss": 6.3364, "step": 9630 }, { "epoch": 0.08, "learning_rate": 9.844855751450163e-06, "loss": 6.7208, "step": 9640 }, { "epoch": 0.08, "learning_rate": 9.844535387804248e-06, "loss": 6.7256, "step": 9650 }, { "epoch": 0.08, "learning_rate": 9.844214698956814e-06, "loss": 6.8162, "step": 9660 }, { "epoch": 0.08, "learning_rate": 9.84389368492939e-06, "loss": 7.0284, "step": 9670 }, { "epoch": 0.08, "learning_rate": 9.843572345743521e-06, "loss": 6.5259, "step": 9680 }, { "epoch": 0.08, "learning_rate": 9.843250681420785e-06, "loss": 7.3396, "step": 9690 }, { "epoch": 0.08, "learning_rate": 9.842928691982769e-06, "loss": 6.6692, "step": 9700 }, { "epoch": 0.08, "learning_rate": 9.84260637745109e-06, "loss": 6.4917, "step": 9710 }, { "epoch": 0.08, "learning_rate": 9.84228373784738e-06, "loss": 6.4017, "step": 9720 }, { "epoch": 0.08, "learning_rate": 9.841960773193303e-06, "loss": 6.7186, "step": 9730 }, { "epoch": 0.08, "learning_rate": 9.841637483510535e-06, "loss": 6.9087, "step": 9740 }, { "epoch": 0.08, "learning_rate": 9.841313868820778e-06, "loss": 6.8888, "step": 9750 }, { "epoch": 0.08, "learning_rate": 9.840989929145757e-06, "loss": 7.2902, "step": 9760 }, { "epoch": 0.08, "learning_rate": 9.840665664507215e-06, "loss": 6.1428, "step": 9770 }, { "epoch": 0.08, "learning_rate": 9.84034107492692e-06, "loss": 6.1216, "step": 9780 }, { "epoch": 0.08, "learning_rate": 9.84001616042666e-06, "loss": 6.8331, "step": 9790 }, { "epoch": 0.08, "learning_rate": 9.839690921028249e-06, "loss": 7.6563, "step": 9800 }, { "epoch": 0.08, "learning_rate": 9.839365356753517e-06, "loss": 6.8403, "step": 9810 }, { "epoch": 0.08, "learning_rate": 9.839039467624317e-06, "loss": 6.9892, "step": 9820 }, { "epoch": 0.08, "learning_rate": 9.838713253662529e-06, "loss": 6.2581, "step": 9830 }, { "epoch": 0.08, "learning_rate": 9.838386714890047e-06, "loss": 6.9717, "step": 9840 }, { "epoch": 0.08, "learning_rate": 9.838059851328791e-06, "loss": 6.889, "step": 9850 }, { "epoch": 0.08, "learning_rate": 9.837732663000708e-06, "loss": 6.5774, "step": 9860 }, { "epoch": 0.08, "learning_rate": 9.837405149927753e-06, "loss": 7.3636, "step": 9870 }, { "epoch": 0.08, "learning_rate": 9.837077312131917e-06, "loss": 6.6574, "step": 9880 }, { "epoch": 0.08, "learning_rate": 9.836749149635207e-06, "loss": 6.1238, "step": 9890 }, { "epoch": 0.08, "learning_rate": 9.836420662459648e-06, "loss": 6.3936, "step": 9900 }, { "epoch": 0.08, "learning_rate": 9.836091850627292e-06, "loss": 6.8964, "step": 9910 }, { "epoch": 0.08, "learning_rate": 9.835762714160213e-06, "loss": 7.453, "step": 9920 }, { "epoch": 0.08, "learning_rate": 9.835433253080505e-06, "loss": 6.1682, "step": 9930 }, { "epoch": 0.08, "learning_rate": 9.83510346741028e-06, "loss": 6.3708, "step": 9940 }, { "epoch": 0.08, "learning_rate": 9.83477335717168e-06, "loss": 6.2002, "step": 9950 }, { "epoch": 0.08, "learning_rate": 9.834442922386864e-06, "loss": 6.4856, "step": 9960 }, { "epoch": 0.08, "learning_rate": 9.83411216307801e-06, "loss": 6.9631, "step": 9970 }, { "epoch": 0.08, "learning_rate": 9.833781079267326e-06, "loss": 7.2966, "step": 9980 }, { "epoch": 0.08, "learning_rate": 9.833449670977033e-06, "loss": 6.7957, "step": 9990 }, { "epoch": 0.08, "learning_rate": 9.833117938229377e-06, "loss": 7.3811, "step": 10000 }, { "epoch": 0.08, "learning_rate": 9.832785881046632e-06, "loss": 6.2964, "step": 10010 }, { "epoch": 0.08, "learning_rate": 9.832453499451081e-06, "loss": 6.5619, "step": 10020 }, { "epoch": 0.08, "learning_rate": 9.83212079346504e-06, "loss": 6.014, "step": 10030 }, { "epoch": 0.08, "learning_rate": 9.831787763110843e-06, "loss": 6.5502, "step": 10040 }, { "epoch": 0.08, "learning_rate": 9.831454408410844e-06, "loss": 6.9824, "step": 10050 }, { "epoch": 0.08, "learning_rate": 9.83112072938742e-06, "loss": 6.299, "step": 10060 }, { "epoch": 0.08, "learning_rate": 9.83078672606297e-06, "loss": 6.5957, "step": 10070 }, { "epoch": 0.08, "learning_rate": 9.830452398459917e-06, "loss": 6.6812, "step": 10080 }, { "epoch": 0.08, "learning_rate": 9.830117746600702e-06, "loss": 6.671, "step": 10090 }, { "epoch": 0.08, "learning_rate": 9.829782770507789e-06, "loss": 6.2986, "step": 10100 }, { "epoch": 0.08, "learning_rate": 9.829447470203664e-06, "loss": 6.0837, "step": 10110 }, { "epoch": 0.08, "learning_rate": 9.829111845710838e-06, "loss": 6.6279, "step": 10120 }, { "epoch": 0.08, "learning_rate": 9.828775897051835e-06, "loss": 6.4508, "step": 10130 }, { "epoch": 0.08, "learning_rate": 9.82843962424921e-06, "loss": 6.971, "step": 10140 }, { "epoch": 0.08, "learning_rate": 9.828103027325537e-06, "loss": 6.2785, "step": 10150 }, { "epoch": 0.08, "learning_rate": 9.827766106303409e-06, "loss": 7.0219, "step": 10160 }, { "epoch": 0.08, "learning_rate": 9.827428861205443e-06, "loss": 6.1527, "step": 10170 }, { "epoch": 0.08, "learning_rate": 9.827091292054278e-06, "loss": 6.5985, "step": 10180 }, { "epoch": 0.08, "learning_rate": 9.826753398872574e-06, "loss": 6.6864, "step": 10190 }, { "epoch": 0.08, "learning_rate": 9.826415181683011e-06, "loss": 7.2026, "step": 10200 }, { "epoch": 0.08, "learning_rate": 9.826076640508297e-06, "loss": 5.7704, "step": 10210 }, { "epoch": 0.08, "learning_rate": 9.825737775371154e-06, "loss": 7.3879, "step": 10220 }, { "epoch": 0.08, "learning_rate": 9.825398586294328e-06, "loss": 7.121, "step": 10230 }, { "epoch": 0.08, "learning_rate": 9.825059073300591e-06, "loss": 6.6974, "step": 10240 }, { "epoch": 0.08, "learning_rate": 9.824719236412733e-06, "loss": 6.1271, "step": 10250 }, { "epoch": 0.08, "learning_rate": 9.824379075653565e-06, "loss": 6.6709, "step": 10260 }, { "epoch": 0.08, "learning_rate": 9.824038591045923e-06, "loss": 5.9571, "step": 10270 }, { "epoch": 0.08, "learning_rate": 9.823697782612662e-06, "loss": 6.5483, "step": 10280 }, { "epoch": 0.08, "learning_rate": 9.823356650376659e-06, "loss": 6.4958, "step": 10290 }, { "epoch": 0.08, "learning_rate": 9.823015194360814e-06, "loss": 6.6114, "step": 10300 }, { "epoch": 0.09, "learning_rate": 9.822673414588048e-06, "loss": 6.7402, "step": 10310 }, { "epoch": 0.09, "learning_rate": 9.822331311081304e-06, "loss": 6.3849, "step": 10320 }, { "epoch": 0.09, "learning_rate": 9.821988883863546e-06, "loss": 6.3938, "step": 10330 }, { "epoch": 0.09, "learning_rate": 9.821646132957762e-06, "loss": 5.9312, "step": 10340 }, { "epoch": 0.09, "learning_rate": 9.821303058386957e-06, "loss": 6.7169, "step": 10350 }, { "epoch": 0.09, "learning_rate": 9.820959660174163e-06, "loss": 6.7544, "step": 10360 }, { "epoch": 0.09, "learning_rate": 9.820615938342431e-06, "loss": 6.0127, "step": 10370 }, { "epoch": 0.09, "learning_rate": 9.820271892914834e-06, "loss": 6.7232, "step": 10380 }, { "epoch": 0.09, "learning_rate": 9.819927523914468e-06, "loss": 7.188, "step": 10390 }, { "epoch": 0.09, "learning_rate": 9.819582831364448e-06, "loss": 6.2976, "step": 10400 }, { "epoch": 0.09, "learning_rate": 9.819237815287911e-06, "loss": 6.9584, "step": 10410 }, { "epoch": 0.09, "learning_rate": 9.818892475708021e-06, "loss": 6.5673, "step": 10420 }, { "epoch": 0.09, "learning_rate": 9.818546812647958e-06, "loss": 6.6771, "step": 10430 }, { "epoch": 0.09, "learning_rate": 9.818200826130924e-06, "loss": 6.5268, "step": 10440 }, { "epoch": 0.09, "learning_rate": 9.817854516180146e-06, "loss": 6.9058, "step": 10450 }, { "epoch": 0.09, "learning_rate": 9.817507882818869e-06, "loss": 6.7307, "step": 10460 }, { "epoch": 0.09, "learning_rate": 9.817160926070364e-06, "loss": 6.2422, "step": 10470 }, { "epoch": 0.09, "learning_rate": 9.816813645957921e-06, "loss": 6.3289, "step": 10480 }, { "epoch": 0.09, "learning_rate": 9.816466042504852e-06, "loss": 6.936, "step": 10490 }, { "epoch": 0.09, "learning_rate": 9.816118115734489e-06, "loss": 6.733, "step": 10500 }, { "epoch": 0.09, "learning_rate": 9.815769865670188e-06, "loss": 6.6059, "step": 10510 }, { "epoch": 0.09, "learning_rate": 9.815421292335327e-06, "loss": 6.9804, "step": 10520 }, { "epoch": 0.09, "learning_rate": 9.815072395753307e-06, "loss": 6.2933, "step": 10530 }, { "epoch": 0.09, "learning_rate": 9.814723175947545e-06, "loss": 7.0108, "step": 10540 }, { "epoch": 0.09, "learning_rate": 9.814373632941484e-06, "loss": 6.8929, "step": 10550 }, { "epoch": 0.09, "learning_rate": 9.814023766758587e-06, "loss": 6.7401, "step": 10560 }, { "epoch": 0.09, "learning_rate": 9.813673577422342e-06, "loss": 7.4121, "step": 10570 }, { "epoch": 0.09, "learning_rate": 9.813323064956256e-06, "loss": 7.0111, "step": 10580 }, { "epoch": 0.09, "learning_rate": 9.812972229383857e-06, "loss": 6.3543, "step": 10590 }, { "epoch": 0.09, "learning_rate": 9.812621070728695e-06, "loss": 6.0304, "step": 10600 }, { "epoch": 0.09, "learning_rate": 9.812269589014344e-06, "loss": 6.6703, "step": 10610 }, { "epoch": 0.09, "learning_rate": 9.811917784264397e-06, "loss": 7.2593, "step": 10620 }, { "epoch": 0.09, "learning_rate": 9.811565656502473e-06, "loss": 6.3205, "step": 10630 }, { "epoch": 0.09, "learning_rate": 9.811213205752203e-06, "loss": 6.218, "step": 10640 }, { "epoch": 0.09, "learning_rate": 9.810860432037254e-06, "loss": 7.1351, "step": 10650 }, { "epoch": 0.09, "learning_rate": 9.8105073353813e-06, "loss": 6.512, "step": 10660 }, { "epoch": 0.09, "learning_rate": 9.810153915808048e-06, "loss": 6.9207, "step": 10670 }, { "epoch": 0.09, "learning_rate": 9.809800173341217e-06, "loss": 7.086, "step": 10680 }, { "epoch": 0.09, "learning_rate": 9.809446108004558e-06, "loss": 7.2192, "step": 10690 }, { "epoch": 0.09, "learning_rate": 9.809091719821839e-06, "loss": 6.2585, "step": 10700 }, { "epoch": 0.09, "learning_rate": 9.808737008816845e-06, "loss": 6.6726, "step": 10710 }, { "epoch": 0.09, "learning_rate": 9.808381975013389e-06, "loss": 6.6107, "step": 10720 }, { "epoch": 0.09, "learning_rate": 9.808026618435302e-06, "loss": 6.9034, "step": 10730 }, { "epoch": 0.09, "learning_rate": 9.807670939106442e-06, "loss": 6.1505, "step": 10740 }, { "epoch": 0.09, "learning_rate": 9.807314937050681e-06, "loss": 6.4607, "step": 10750 }, { "epoch": 0.09, "learning_rate": 9.80695861229192e-06, "loss": 6.3515, "step": 10760 }, { "epoch": 0.09, "learning_rate": 9.806601964854073e-06, "loss": 7.0099, "step": 10770 }, { "epoch": 0.09, "learning_rate": 9.806244994761085e-06, "loss": 6.2417, "step": 10780 }, { "epoch": 0.09, "learning_rate": 9.805887702036919e-06, "loss": 6.4342, "step": 10790 }, { "epoch": 0.09, "learning_rate": 9.805530086705555e-06, "loss": 6.8398, "step": 10800 }, { "epoch": 0.09, "learning_rate": 9.805172148791002e-06, "loss": 6.7818, "step": 10810 }, { "epoch": 0.09, "learning_rate": 9.804813888317288e-06, "loss": 6.612, "step": 10820 }, { "epoch": 0.09, "learning_rate": 9.804455305308461e-06, "loss": 7.0109, "step": 10830 }, { "epoch": 0.09, "learning_rate": 9.804096399788591e-06, "loss": 6.9158, "step": 10840 }, { "epoch": 0.09, "learning_rate": 9.803737171781773e-06, "loss": 6.4467, "step": 10850 }, { "epoch": 0.09, "learning_rate": 9.803377621312118e-06, "loss": 7.0119, "step": 10860 }, { "epoch": 0.09, "learning_rate": 9.803017748403762e-06, "loss": 6.3766, "step": 10870 }, { "epoch": 0.09, "learning_rate": 9.802657553080865e-06, "loss": 6.8032, "step": 10880 }, { "epoch": 0.09, "learning_rate": 9.802297035367604e-06, "loss": 6.2644, "step": 10890 }, { "epoch": 0.09, "learning_rate": 9.801936195288181e-06, "loss": 6.123, "step": 10900 }, { "epoch": 0.09, "learning_rate": 9.801575032866817e-06, "loss": 6.7611, "step": 10910 }, { "epoch": 0.09, "learning_rate": 9.801213548127755e-06, "loss": 6.6199, "step": 10920 }, { "epoch": 0.09, "learning_rate": 9.800851741095264e-06, "loss": 7.1014, "step": 10930 }, { "epoch": 0.09, "learning_rate": 9.800489611793626e-06, "loss": 7.1531, "step": 10940 }, { "epoch": 0.09, "learning_rate": 9.800127160247155e-06, "loss": 6.5111, "step": 10950 }, { "epoch": 0.09, "learning_rate": 9.799764386480179e-06, "loss": 6.3612, "step": 10960 }, { "epoch": 0.09, "learning_rate": 9.799401290517052e-06, "loss": 6.5275, "step": 10970 }, { "epoch": 0.09, "learning_rate": 9.799037872382145e-06, "loss": 6.7814, "step": 10980 }, { "epoch": 0.09, "learning_rate": 9.798674132099852e-06, "loss": 6.6009, "step": 10990 }, { "epoch": 0.09, "learning_rate": 9.798310069694596e-06, "loss": 6.9123, "step": 11000 }, { "epoch": 0.09, "learning_rate": 9.79794568519081e-06, "loss": 6.685, "step": 11010 }, { "epoch": 0.09, "learning_rate": 9.797580978612957e-06, "loss": 6.5985, "step": 11020 }, { "epoch": 0.09, "learning_rate": 9.797215949985519e-06, "loss": 7.1132, "step": 11030 }, { "epoch": 0.09, "learning_rate": 9.796850599332998e-06, "loss": 6.007, "step": 11040 }, { "epoch": 0.09, "learning_rate": 9.79648492667992e-06, "loss": 5.892, "step": 11050 }, { "epoch": 0.09, "learning_rate": 9.79611893205083e-06, "loss": 5.6392, "step": 11060 }, { "epoch": 0.09, "learning_rate": 9.795752615470298e-06, "loss": 6.861, "step": 11070 }, { "epoch": 0.09, "learning_rate": 9.795385976962914e-06, "loss": 6.59, "step": 11080 }, { "epoch": 0.09, "learning_rate": 9.79501901655329e-06, "loss": 6.8319, "step": 11090 }, { "epoch": 0.09, "learning_rate": 9.794651734266057e-06, "loss": 6.0592, "step": 11100 }, { "epoch": 0.09, "learning_rate": 9.794284130125874e-06, "loss": 6.6191, "step": 11110 }, { "epoch": 0.09, "learning_rate": 9.79391620415741e-06, "loss": 6.9582, "step": 11120 }, { "epoch": 0.09, "learning_rate": 9.793547956385369e-06, "loss": 6.0166, "step": 11130 }, { "epoch": 0.09, "learning_rate": 9.793179386834469e-06, "loss": 6.5062, "step": 11140 }, { "epoch": 0.09, "learning_rate": 9.79281049552945e-06, "loss": 6.3142, "step": 11150 }, { "epoch": 0.09, "learning_rate": 9.792441282495078e-06, "loss": 6.0395, "step": 11160 }, { "epoch": 0.09, "learning_rate": 9.792071747756134e-06, "loss": 6.2711, "step": 11170 }, { "epoch": 0.09, "learning_rate": 9.791701891337423e-06, "loss": 6.3159, "step": 11180 }, { "epoch": 0.09, "learning_rate": 9.791331713263777e-06, "loss": 6.9982, "step": 11190 }, { "epoch": 0.09, "learning_rate": 9.790961213560042e-06, "loss": 6.2258, "step": 11200 }, { "epoch": 0.09, "learning_rate": 9.79059039225109e-06, "loss": 6.539, "step": 11210 }, { "epoch": 0.09, "learning_rate": 9.790219249361812e-06, "loss": 6.667, "step": 11220 }, { "epoch": 0.09, "learning_rate": 9.789847784917122e-06, "loss": 6.7366, "step": 11230 }, { "epoch": 0.09, "learning_rate": 9.789475998941957e-06, "loss": 6.5822, "step": 11240 }, { "epoch": 0.09, "learning_rate": 9.789103891461273e-06, "loss": 6.4815, "step": 11250 }, { "epoch": 0.09, "learning_rate": 9.788731462500047e-06, "loss": 6.4355, "step": 11260 }, { "epoch": 0.09, "learning_rate": 9.788358712083283e-06, "loss": 5.2298, "step": 11270 }, { "epoch": 0.09, "learning_rate": 9.787985640236e-06, "loss": 6.218, "step": 11280 }, { "epoch": 0.09, "learning_rate": 9.787612246983244e-06, "loss": 6.0976, "step": 11290 }, { "epoch": 0.09, "learning_rate": 9.787238532350076e-06, "loss": 6.6952, "step": 11300 }, { "epoch": 0.09, "learning_rate": 9.786864496361586e-06, "loss": 6.1875, "step": 11310 }, { "epoch": 0.09, "learning_rate": 9.786490139042879e-06, "loss": 6.4138, "step": 11320 }, { "epoch": 0.09, "learning_rate": 9.786115460419087e-06, "loss": 6.3444, "step": 11330 }, { "epoch": 0.09, "learning_rate": 9.785740460515362e-06, "loss": 6.4799, "step": 11340 }, { "epoch": 0.09, "learning_rate": 9.785365139356873e-06, "loss": 6.437, "step": 11350 }, { "epoch": 0.09, "learning_rate": 9.78498949696882e-06, "loss": 6.538, "step": 11360 }, { "epoch": 0.09, "learning_rate": 9.784613533376413e-06, "loss": 6.2996, "step": 11370 }, { "epoch": 0.09, "learning_rate": 9.784237248604893e-06, "loss": 6.8085, "step": 11380 }, { "epoch": 0.09, "learning_rate": 9.783860642679518e-06, "loss": 5.9983, "step": 11390 }, { "epoch": 0.09, "learning_rate": 9.783483715625571e-06, "loss": 6.5107, "step": 11400 }, { "epoch": 0.09, "learning_rate": 9.78310646746835e-06, "loss": 6.3371, "step": 11410 }, { "epoch": 0.09, "learning_rate": 9.782728898233182e-06, "loss": 6.5141, "step": 11420 }, { "epoch": 0.09, "learning_rate": 9.78235100794541e-06, "loss": 6.8506, "step": 11430 }, { "epoch": 0.09, "learning_rate": 9.781972796630403e-06, "loss": 6.6375, "step": 11440 }, { "epoch": 0.09, "learning_rate": 9.781594264313548e-06, "loss": 6.0445, "step": 11450 }, { "epoch": 0.09, "learning_rate": 9.781215411020256e-06, "loss": 6.9692, "step": 11460 }, { "epoch": 0.09, "learning_rate": 9.780836236775954e-06, "loss": 6.5275, "step": 11470 }, { "epoch": 0.09, "learning_rate": 9.780456741606102e-06, "loss": 6.6202, "step": 11480 }, { "epoch": 0.09, "learning_rate": 9.78007692553617e-06, "loss": 6.0488, "step": 11490 }, { "epoch": 0.09, "learning_rate": 9.779696788591658e-06, "loss": 6.4865, "step": 11500 }, { "epoch": 0.09, "learning_rate": 9.779316330798077e-06, "loss": 6.9584, "step": 11510 }, { "epoch": 0.1, "learning_rate": 9.778935552180973e-06, "loss": 6.6342, "step": 11520 }, { "epoch": 0.1, "learning_rate": 9.778554452765903e-06, "loss": 6.596, "step": 11530 }, { "epoch": 0.1, "learning_rate": 9.77817303257845e-06, "loss": 6.5538, "step": 11540 }, { "epoch": 0.1, "learning_rate": 9.777791291644218e-06, "loss": 6.7598, "step": 11550 }, { "epoch": 0.1, "learning_rate": 9.777409229988832e-06, "loss": 6.0378, "step": 11560 }, { "epoch": 0.1, "learning_rate": 9.77702684763794e-06, "loss": 6.0812, "step": 11570 }, { "epoch": 0.1, "learning_rate": 9.776644144617209e-06, "loss": 6.983, "step": 11580 }, { "epoch": 0.1, "learning_rate": 9.776261120952329e-06, "loss": 6.7173, "step": 11590 }, { "epoch": 0.1, "learning_rate": 9.775877776669012e-06, "loss": 6.0504, "step": 11600 }, { "epoch": 0.1, "learning_rate": 9.77549411179299e-06, "loss": 5.989, "step": 11610 }, { "epoch": 0.1, "learning_rate": 9.775110126350019e-06, "loss": 6.2456, "step": 11620 }, { "epoch": 0.1, "learning_rate": 9.774725820365873e-06, "loss": 6.1234, "step": 11630 }, { "epoch": 0.1, "learning_rate": 9.774341193866352e-06, "loss": 6.4135, "step": 11640 }, { "epoch": 0.1, "learning_rate": 9.773956246877273e-06, "loss": 6.5527, "step": 11650 }, { "epoch": 0.1, "learning_rate": 9.773570979424478e-06, "loss": 6.4617, "step": 11660 }, { "epoch": 0.1, "learning_rate": 9.773185391533826e-06, "loss": 6.379, "step": 11670 }, { "epoch": 0.1, "learning_rate": 9.772799483231205e-06, "loss": 6.1349, "step": 11680 }, { "epoch": 0.1, "learning_rate": 9.772413254542518e-06, "loss": 7.1162, "step": 11690 }, { "epoch": 0.1, "learning_rate": 9.77202670549369e-06, "loss": 7.3327, "step": 11700 }, { "epoch": 0.1, "learning_rate": 9.77163983611067e-06, "loss": 5.9598, "step": 11710 }, { "epoch": 0.1, "learning_rate": 9.77125264641943e-06, "loss": 6.6929, "step": 11720 }, { "epoch": 0.1, "learning_rate": 9.770865136445958e-06, "loss": 6.6547, "step": 11730 }, { "epoch": 0.1, "learning_rate": 9.770477306216269e-06, "loss": 6.5115, "step": 11740 }, { "epoch": 0.1, "learning_rate": 9.770089155756394e-06, "loss": 6.5791, "step": 11750 }, { "epoch": 0.1, "learning_rate": 9.769700685092391e-06, "loss": 6.6984, "step": 11760 }, { "epoch": 0.1, "learning_rate": 9.769311894250335e-06, "loss": 7.0699, "step": 11770 }, { "epoch": 0.1, "learning_rate": 9.768922783256329e-06, "loss": 6.2291, "step": 11780 }, { "epoch": 0.1, "learning_rate": 9.768533352136487e-06, "loss": 6.4108, "step": 11790 }, { "epoch": 0.1, "learning_rate": 9.768143600916955e-06, "loss": 6.6121, "step": 11800 }, { "epoch": 0.1, "learning_rate": 9.767753529623894e-06, "loss": 6.8472, "step": 11810 }, { "epoch": 0.1, "learning_rate": 9.76736313828349e-06, "loss": 6.0079, "step": 11820 }, { "epoch": 0.1, "learning_rate": 9.766972426921946e-06, "loss": 5.9357, "step": 11830 }, { "epoch": 0.1, "learning_rate": 9.766581395565493e-06, "loss": 6.594, "step": 11840 }, { "epoch": 0.1, "learning_rate": 9.766190044240377e-06, "loss": 7.1796, "step": 11850 }, { "epoch": 0.1, "learning_rate": 9.765798372972873e-06, "loss": 6.899, "step": 11860 }, { "epoch": 0.1, "learning_rate": 9.765406381789267e-06, "loss": 6.4994, "step": 11870 }, { "epoch": 0.1, "learning_rate": 9.765014070715878e-06, "loss": 6.4831, "step": 11880 }, { "epoch": 0.1, "learning_rate": 9.764621439779034e-06, "loss": 6.2491, "step": 11890 }, { "epoch": 0.1, "learning_rate": 9.764228489005099e-06, "loss": 6.6146, "step": 11900 }, { "epoch": 0.1, "learning_rate": 9.763835218420446e-06, "loss": 6.1712, "step": 11910 }, { "epoch": 0.1, "learning_rate": 9.763441628051474e-06, "loss": 6.0751, "step": 11920 }, { "epoch": 0.1, "learning_rate": 9.763047717924606e-06, "loss": 6.1697, "step": 11930 }, { "epoch": 0.1, "learning_rate": 9.762653488066285e-06, "loss": 6.8719, "step": 11940 }, { "epoch": 0.1, "learning_rate": 9.762258938502971e-06, "loss": 6.5651, "step": 11950 }, { "epoch": 0.1, "learning_rate": 9.761864069261151e-06, "loss": 6.3883, "step": 11960 }, { "epoch": 0.1, "learning_rate": 9.761468880367333e-06, "loss": 6.3915, "step": 11970 }, { "epoch": 0.1, "learning_rate": 9.761073371848045e-06, "loss": 6.657, "step": 11980 }, { "epoch": 0.1, "learning_rate": 9.760677543729832e-06, "loss": 6.9187, "step": 11990 }, { "epoch": 0.1, "learning_rate": 9.76028139603927e-06, "loss": 6.2563, "step": 12000 }, { "epoch": 0.1, "learning_rate": 9.759884928802949e-06, "loss": 6.8625, "step": 12010 }, { "epoch": 0.1, "learning_rate": 9.759488142047485e-06, "loss": 6.7109, "step": 12020 }, { "epoch": 0.1, "learning_rate": 9.75909103579951e-06, "loss": 6.4676, "step": 12030 }, { "epoch": 0.1, "learning_rate": 9.758693610085685e-06, "loss": 5.6545, "step": 12040 }, { "epoch": 0.1, "learning_rate": 9.758295864932684e-06, "loss": 6.3573, "step": 12050 }, { "epoch": 0.1, "learning_rate": 9.757897800367209e-06, "loss": 7.137, "step": 12060 }, { "epoch": 0.1, "learning_rate": 9.75749941641598e-06, "loss": 7.007, "step": 12070 }, { "epoch": 0.1, "learning_rate": 9.757100713105739e-06, "loss": 6.3254, "step": 12080 }, { "epoch": 0.1, "learning_rate": 9.756701690463253e-06, "loss": 6.5765, "step": 12090 }, { "epoch": 0.1, "learning_rate": 9.756302348515303e-06, "loss": 7.0515, "step": 12100 }, { "epoch": 0.1, "learning_rate": 9.7559026872887e-06, "loss": 6.3884, "step": 12110 }, { "epoch": 0.1, "learning_rate": 9.755502706810269e-06, "loss": 6.6603, "step": 12120 }, { "epoch": 0.1, "learning_rate": 9.755102407106862e-06, "loss": 6.5244, "step": 12130 }, { "epoch": 0.1, "learning_rate": 9.75470178820535e-06, "loss": 6.6166, "step": 12140 }, { "epoch": 0.1, "learning_rate": 9.754300850132623e-06, "loss": 6.4439, "step": 12150 }, { "epoch": 0.1, "learning_rate": 9.753899592915596e-06, "loss": 6.5215, "step": 12160 }, { "epoch": 0.1, "learning_rate": 9.753498016581207e-06, "loss": 6.4873, "step": 12170 }, { "epoch": 0.1, "learning_rate": 9.753096121156412e-06, "loss": 6.0224, "step": 12180 }, { "epoch": 0.1, "learning_rate": 9.752693906668185e-06, "loss": 6.863, "step": 12190 }, { "epoch": 0.1, "learning_rate": 9.75229137314353e-06, "loss": 6.61, "step": 12200 }, { "epoch": 0.1, "learning_rate": 9.751888520609468e-06, "loss": 6.4127, "step": 12210 }, { "epoch": 0.1, "learning_rate": 9.751485349093039e-06, "loss": 6.9714, "step": 12220 }, { "epoch": 0.1, "learning_rate": 9.751081858621309e-06, "loss": 6.4993, "step": 12230 }, { "epoch": 0.1, "learning_rate": 9.750678049221362e-06, "loss": 6.5054, "step": 12240 }, { "epoch": 0.1, "learning_rate": 9.750273920920305e-06, "loss": 6.3386, "step": 12250 }, { "epoch": 0.1, "learning_rate": 9.749869473745266e-06, "loss": 6.2669, "step": 12260 }, { "epoch": 0.1, "learning_rate": 9.749464707723395e-06, "loss": 6.2725, "step": 12270 }, { "epoch": 0.1, "learning_rate": 9.749059622881865e-06, "loss": 6.4354, "step": 12280 }, { "epoch": 0.1, "learning_rate": 9.748654219247864e-06, "loss": 6.5044, "step": 12290 }, { "epoch": 0.1, "learning_rate": 9.748248496848609e-06, "loss": 6.7863, "step": 12300 }, { "epoch": 0.1, "learning_rate": 9.747842455711334e-06, "loss": 6.5698, "step": 12310 }, { "epoch": 0.1, "learning_rate": 9.747436095863294e-06, "loss": 6.8278, "step": 12320 }, { "epoch": 0.1, "learning_rate": 9.74702941733177e-06, "loss": 6.6152, "step": 12330 }, { "epoch": 0.1, "learning_rate": 9.746622420144059e-06, "loss": 6.7152, "step": 12340 }, { "epoch": 0.1, "learning_rate": 9.746215104327482e-06, "loss": 6.4546, "step": 12350 }, { "epoch": 0.1, "learning_rate": 9.745807469909383e-06, "loss": 6.3419, "step": 12360 }, { "epoch": 0.1, "learning_rate": 9.745399516917122e-06, "loss": 6.4755, "step": 12370 }, { "epoch": 0.1, "learning_rate": 9.744991245378089e-06, "loss": 6.3812, "step": 12380 }, { "epoch": 0.1, "learning_rate": 9.744582655319683e-06, "loss": 6.8784, "step": 12390 }, { "epoch": 0.1, "learning_rate": 9.744173746769339e-06, "loss": 6.8027, "step": 12400 }, { "epoch": 0.1, "learning_rate": 9.743764519754502e-06, "loss": 5.7975, "step": 12410 }, { "epoch": 0.1, "learning_rate": 9.743354974302642e-06, "loss": 6.2552, "step": 12420 }, { "epoch": 0.1, "learning_rate": 9.742945110441252e-06, "loss": 6.8516, "step": 12430 }, { "epoch": 0.1, "learning_rate": 9.742534928197847e-06, "loss": 7.171, "step": 12440 }, { "epoch": 0.1, "learning_rate": 9.742124427599959e-06, "loss": 5.9673, "step": 12450 }, { "epoch": 0.1, "learning_rate": 9.741713608675143e-06, "loss": 6.1619, "step": 12460 }, { "epoch": 0.1, "learning_rate": 9.741302471450978e-06, "loss": 6.3516, "step": 12470 }, { "epoch": 0.1, "learning_rate": 9.740891015955062e-06, "loss": 5.9412, "step": 12480 }, { "epoch": 0.1, "learning_rate": 9.740479242215018e-06, "loss": 6.1842, "step": 12490 }, { "epoch": 0.1, "learning_rate": 9.740067150258481e-06, "loss": 6.3028, "step": 12500 }, { "epoch": 0.1, "learning_rate": 9.739654740113119e-06, "loss": 7.2655, "step": 12510 }, { "epoch": 0.1, "learning_rate": 9.739242011806614e-06, "loss": 6.6183, "step": 12520 }, { "epoch": 0.1, "learning_rate": 9.738828965366673e-06, "loss": 6.54, "step": 12530 }, { "epoch": 0.1, "learning_rate": 9.738415600821021e-06, "loss": 6.3919, "step": 12540 }, { "epoch": 0.1, "learning_rate": 9.738001918197405e-06, "loss": 6.5399, "step": 12550 }, { "epoch": 0.1, "learning_rate": 9.737587917523597e-06, "loss": 6.1216, "step": 12560 }, { "epoch": 0.1, "learning_rate": 9.737173598827388e-06, "loss": 6.3144, "step": 12570 }, { "epoch": 0.1, "learning_rate": 9.736758962136588e-06, "loss": 6.4465, "step": 12580 }, { "epoch": 0.1, "learning_rate": 9.736344007479032e-06, "loss": 5.8549, "step": 12590 }, { "epoch": 0.1, "learning_rate": 9.735928734882574e-06, "loss": 6.1208, "step": 12600 }, { "epoch": 0.1, "learning_rate": 9.735513144375091e-06, "loss": 6.1644, "step": 12610 }, { "epoch": 0.1, "learning_rate": 9.735097235984481e-06, "loss": 6.7593, "step": 12620 }, { "epoch": 0.1, "learning_rate": 9.734681009738663e-06, "loss": 6.5941, "step": 12630 }, { "epoch": 0.1, "learning_rate": 9.734264465665574e-06, "loss": 6.2649, "step": 12640 }, { "epoch": 0.1, "learning_rate": 9.73384760379318e-06, "loss": 6.2271, "step": 12650 }, { "epoch": 0.1, "learning_rate": 9.73343042414946e-06, "loss": 6.4917, "step": 12660 }, { "epoch": 0.1, "learning_rate": 9.733012926762421e-06, "loss": 7.057, "step": 12670 }, { "epoch": 0.1, "learning_rate": 9.732595111660088e-06, "loss": 5.82, "step": 12680 }, { "epoch": 0.1, "learning_rate": 9.732176978870508e-06, "loss": 6.8837, "step": 12690 }, { "epoch": 0.1, "learning_rate": 9.731758528421747e-06, "loss": 6.3355, "step": 12700 }, { "epoch": 0.1, "learning_rate": 9.731339760341898e-06, "loss": 6.0429, "step": 12710 }, { "epoch": 0.1, "learning_rate": 9.73092067465907e-06, "loss": 6.4915, "step": 12720 }, { "epoch": 0.1, "learning_rate": 9.730501271401393e-06, "loss": 6.2806, "step": 12730 }, { "epoch": 0.11, "learning_rate": 9.730081550597025e-06, "loss": 5.7617, "step": 12740 }, { "epoch": 0.11, "learning_rate": 9.729661512274138e-06, "loss": 7.194, "step": 12750 }, { "epoch": 0.11, "learning_rate": 9.72924115646093e-06, "loss": 7.1925, "step": 12760 }, { "epoch": 0.11, "learning_rate": 9.728820483185615e-06, "loss": 6.0618, "step": 12770 }, { "epoch": 0.11, "learning_rate": 9.728399492476433e-06, "loss": 6.2375, "step": 12780 }, { "epoch": 0.11, "learning_rate": 9.727978184361648e-06, "loss": 6.7445, "step": 12790 }, { "epoch": 0.11, "learning_rate": 9.727556558869536e-06, "loss": 6.3334, "step": 12800 }, { "epoch": 0.11, "learning_rate": 9.727134616028403e-06, "loss": 6.2404, "step": 12810 }, { "epoch": 0.11, "learning_rate": 9.72671235586657e-06, "loss": 6.7364, "step": 12820 }, { "epoch": 0.11, "learning_rate": 9.726289778412386e-06, "loss": 6.2448, "step": 12830 }, { "epoch": 0.11, "learning_rate": 9.725866883694215e-06, "loss": 6.4532, "step": 12840 }, { "epoch": 0.11, "learning_rate": 9.725443671740447e-06, "loss": 6.3544, "step": 12850 }, { "epoch": 0.11, "learning_rate": 9.72502014257949e-06, "loss": 5.8891, "step": 12860 }, { "epoch": 0.11, "learning_rate": 9.724596296239772e-06, "loss": 6.2817, "step": 12870 }, { "epoch": 0.11, "learning_rate": 9.724172132749747e-06, "loss": 6.3557, "step": 12880 }, { "epoch": 0.11, "learning_rate": 9.72374765213789e-06, "loss": 6.1793, "step": 12890 }, { "epoch": 0.11, "learning_rate": 9.723322854432692e-06, "loss": 6.8174, "step": 12900 }, { "epoch": 0.11, "learning_rate": 9.72289773966267e-06, "loss": 6.148, "step": 12910 }, { "epoch": 0.11, "learning_rate": 9.722472307856361e-06, "loss": 6.2875, "step": 12920 }, { "epoch": 0.11, "learning_rate": 9.722046559042321e-06, "loss": 5.9226, "step": 12930 }, { "epoch": 0.11, "learning_rate": 9.721620493249135e-06, "loss": 6.3422, "step": 12940 }, { "epoch": 0.11, "learning_rate": 9.721194110505399e-06, "loss": 6.3776, "step": 12950 }, { "epoch": 0.11, "learning_rate": 9.720767410839736e-06, "loss": 6.5978, "step": 12960 }, { "epoch": 0.11, "learning_rate": 9.720340394280788e-06, "loss": 6.1446, "step": 12970 }, { "epoch": 0.11, "learning_rate": 9.719913060857222e-06, "loss": 6.1129, "step": 12980 }, { "epoch": 0.11, "learning_rate": 9.719485410597722e-06, "loss": 6.6818, "step": 12990 }, { "epoch": 0.11, "learning_rate": 9.719057443530997e-06, "loss": 6.8776, "step": 13000 }, { "epoch": 0.11, "learning_rate": 9.718629159685775e-06, "loss": 6.4471, "step": 13010 }, { "epoch": 0.11, "learning_rate": 9.718200559090803e-06, "loss": 6.4443, "step": 13020 }, { "epoch": 0.11, "learning_rate": 9.717771641774854e-06, "loss": 5.7274, "step": 13030 }, { "epoch": 0.11, "learning_rate": 9.71734240776672e-06, "loss": 6.1611, "step": 13040 }, { "epoch": 0.11, "learning_rate": 9.716912857095215e-06, "loss": 6.4658, "step": 13050 }, { "epoch": 0.11, "learning_rate": 9.716482989789171e-06, "loss": 6.3404, "step": 13060 }, { "epoch": 0.11, "learning_rate": 9.71605280587745e-06, "loss": 6.5029, "step": 13070 }, { "epoch": 0.11, "learning_rate": 9.71562230538892e-06, "loss": 6.1935, "step": 13080 }, { "epoch": 0.11, "learning_rate": 9.715191488352489e-06, "loss": 6.5686, "step": 13090 }, { "epoch": 0.11, "learning_rate": 9.71476035479707e-06, "loss": 7.2021, "step": 13100 }, { "epoch": 0.11, "learning_rate": 9.714328904751605e-06, "loss": 6.5691, "step": 13110 }, { "epoch": 0.11, "learning_rate": 9.71389713824506e-06, "loss": 6.259, "step": 13120 }, { "epoch": 0.11, "learning_rate": 9.713465055306412e-06, "loss": 6.8634, "step": 13130 }, { "epoch": 0.11, "learning_rate": 9.713032655964672e-06, "loss": 6.3201, "step": 13140 }, { "epoch": 0.11, "learning_rate": 9.712599940248862e-06, "loss": 6.5028, "step": 13150 }, { "epoch": 0.11, "learning_rate": 9.71216690818803e-06, "loss": 6.8221, "step": 13160 }, { "epoch": 0.11, "learning_rate": 9.711733559811245e-06, "loss": 6.4904, "step": 13170 }, { "epoch": 0.11, "learning_rate": 9.711299895147597e-06, "loss": 6.2836, "step": 13180 }, { "epoch": 0.11, "learning_rate": 9.710865914226195e-06, "loss": 6.7688, "step": 13190 }, { "epoch": 0.11, "learning_rate": 9.710431617076173e-06, "loss": 6.0001, "step": 13200 }, { "epoch": 0.11, "learning_rate": 9.709997003726683e-06, "loss": 6.3308, "step": 13210 }, { "epoch": 0.11, "learning_rate": 9.7095620742069e-06, "loss": 5.9234, "step": 13220 }, { "epoch": 0.11, "learning_rate": 9.709126828546019e-06, "loss": 6.6713, "step": 13230 }, { "epoch": 0.11, "learning_rate": 9.708691266773257e-06, "loss": 6.5581, "step": 13240 }, { "epoch": 0.11, "learning_rate": 9.708255388917853e-06, "loss": 7.0288, "step": 13250 }, { "epoch": 0.11, "learning_rate": 9.707819195009067e-06, "loss": 6.3329, "step": 13260 }, { "epoch": 0.11, "learning_rate": 9.707382685076179e-06, "loss": 5.978, "step": 13270 }, { "epoch": 0.11, "learning_rate": 9.706945859148488e-06, "loss": 6.6733, "step": 13280 }, { "epoch": 0.11, "learning_rate": 9.706508717255322e-06, "loss": 5.6011, "step": 13290 }, { "epoch": 0.11, "learning_rate": 9.706071259426021e-06, "loss": 6.5873, "step": 13300 }, { "epoch": 0.11, "learning_rate": 9.705633485689953e-06, "loss": 6.5061, "step": 13310 }, { "epoch": 0.11, "learning_rate": 9.705195396076505e-06, "loss": 6.5683, "step": 13320 }, { "epoch": 0.11, "learning_rate": 9.704756990615083e-06, "loss": 6.3001, "step": 13330 }, { "epoch": 0.11, "learning_rate": 9.704318269335117e-06, "loss": 5.639, "step": 13340 }, { "epoch": 0.11, "learning_rate": 9.703879232266058e-06, "loss": 7.0834, "step": 13350 }, { "epoch": 0.11, "learning_rate": 9.703439879437374e-06, "loss": 6.716, "step": 13360 }, { "epoch": 0.11, "learning_rate": 9.703000210878561e-06, "loss": 6.7711, "step": 13370 }, { "epoch": 0.11, "learning_rate": 9.702560226619133e-06, "loss": 6.2725, "step": 13380 }, { "epoch": 0.11, "learning_rate": 9.702119926688625e-06, "loss": 6.0542, "step": 13390 }, { "epoch": 0.11, "learning_rate": 9.70167931111659e-06, "loss": 6.751, "step": 13400 }, { "epoch": 0.11, "learning_rate": 9.701238379932609e-06, "loss": 6.6445, "step": 13410 }, { "epoch": 0.11, "learning_rate": 9.700797133166277e-06, "loss": 6.6536, "step": 13420 }, { "epoch": 0.11, "learning_rate": 9.70035557084722e-06, "loss": 6.4889, "step": 13430 }, { "epoch": 0.11, "learning_rate": 9.699913693005072e-06, "loss": 5.9565, "step": 13440 }, { "epoch": 0.11, "learning_rate": 9.6994714996695e-06, "loss": 6.0434, "step": 13450 }, { "epoch": 0.11, "learning_rate": 9.699028990870183e-06, "loss": 6.7661, "step": 13460 }, { "epoch": 0.11, "learning_rate": 9.698586166636831e-06, "loss": 7.1257, "step": 13470 }, { "epoch": 0.11, "learning_rate": 9.698143026999166e-06, "loss": 6.8586, "step": 13480 }, { "epoch": 0.11, "learning_rate": 9.697699571986935e-06, "loss": 6.1258, "step": 13490 }, { "epoch": 0.11, "learning_rate": 9.697255801629907e-06, "loss": 6.2181, "step": 13500 }, { "epoch": 0.11, "learning_rate": 9.69681171595787e-06, "loss": 7.1012, "step": 13510 }, { "epoch": 0.11, "learning_rate": 9.696367315000637e-06, "loss": 6.6316, "step": 13520 }, { "epoch": 0.11, "learning_rate": 9.695922598788036e-06, "loss": 6.0656, "step": 13530 }, { "epoch": 0.11, "learning_rate": 9.695477567349923e-06, "loss": 5.721, "step": 13540 }, { "epoch": 0.11, "learning_rate": 9.695032220716168e-06, "loss": 6.5476, "step": 13550 }, { "epoch": 0.11, "learning_rate": 9.69458655891667e-06, "loss": 5.6092, "step": 13560 }, { "epoch": 0.11, "learning_rate": 9.694140581981343e-06, "loss": 6.5496, "step": 13570 }, { "epoch": 0.11, "learning_rate": 9.693694289940124e-06, "loss": 6.0118, "step": 13580 }, { "epoch": 0.11, "learning_rate": 9.693247682822972e-06, "loss": 6.3133, "step": 13590 }, { "epoch": 0.11, "learning_rate": 9.692800760659869e-06, "loss": 6.2632, "step": 13600 }, { "epoch": 0.11, "learning_rate": 9.69235352348081e-06, "loss": 6.6471, "step": 13610 }, { "epoch": 0.11, "learning_rate": 9.691905971315823e-06, "loss": 5.6706, "step": 13620 }, { "epoch": 0.11, "learning_rate": 9.691458104194948e-06, "loss": 6.8701, "step": 13630 }, { "epoch": 0.11, "learning_rate": 9.69100992214825e-06, "loss": 7.0005, "step": 13640 }, { "epoch": 0.11, "learning_rate": 9.690561425205813e-06, "loss": 5.9729, "step": 13650 }, { "epoch": 0.11, "learning_rate": 9.690112613397745e-06, "loss": 6.3108, "step": 13660 }, { "epoch": 0.11, "learning_rate": 9.689663486754173e-06, "loss": 6.7369, "step": 13670 }, { "epoch": 0.11, "learning_rate": 9.689214045305248e-06, "loss": 7.0514, "step": 13680 }, { "epoch": 0.11, "learning_rate": 9.688764289081135e-06, "loss": 6.3187, "step": 13690 }, { "epoch": 0.11, "learning_rate": 9.68831421811203e-06, "loss": 6.512, "step": 13700 }, { "epoch": 0.11, "learning_rate": 9.68786383242814e-06, "loss": 6.7201, "step": 13710 }, { "epoch": 0.11, "learning_rate": 9.687413132059704e-06, "loss": 5.9603, "step": 13720 }, { "epoch": 0.11, "learning_rate": 9.686962117036972e-06, "loss": 6.3921, "step": 13730 }, { "epoch": 0.11, "learning_rate": 9.686510787390222e-06, "loss": 6.342, "step": 13740 }, { "epoch": 0.11, "learning_rate": 9.686059143149749e-06, "loss": 5.9, "step": 13750 }, { "epoch": 0.11, "learning_rate": 9.68560718434587e-06, "loss": 6.4174, "step": 13760 }, { "epoch": 0.11, "learning_rate": 9.685154911008927e-06, "loss": 6.3148, "step": 13770 }, { "epoch": 0.11, "learning_rate": 9.68470232316928e-06, "loss": 6.0562, "step": 13780 }, { "epoch": 0.11, "learning_rate": 9.684249420857305e-06, "loss": 6.8038, "step": 13790 }, { "epoch": 0.11, "learning_rate": 9.683796204103411e-06, "loss": 5.8693, "step": 13800 }, { "epoch": 0.11, "learning_rate": 9.683342672938014e-06, "loss": 6.5584, "step": 13810 }, { "epoch": 0.11, "learning_rate": 9.682888827391565e-06, "loss": 6.1516, "step": 13820 }, { "epoch": 0.11, "learning_rate": 9.682434667494526e-06, "loss": 6.4072, "step": 13830 }, { "epoch": 0.11, "learning_rate": 9.681980193277386e-06, "loss": 6.4953, "step": 13840 }, { "epoch": 0.11, "learning_rate": 9.681525404770647e-06, "loss": 6.2193, "step": 13850 }, { "epoch": 0.11, "learning_rate": 9.681070302004845e-06, "loss": 5.9307, "step": 13860 }, { "epoch": 0.11, "learning_rate": 9.680614885010527e-06, "loss": 7.1018, "step": 13870 }, { "epoch": 0.11, "learning_rate": 9.680159153818263e-06, "loss": 6.2863, "step": 13880 }, { "epoch": 0.11, "learning_rate": 9.679703108458645e-06, "loss": 6.3605, "step": 13890 }, { "epoch": 0.11, "learning_rate": 9.679246748962288e-06, "loss": 6.6927, "step": 13900 }, { "epoch": 0.11, "learning_rate": 9.678790075359826e-06, "loss": 6.2655, "step": 13910 }, { "epoch": 0.11, "learning_rate": 9.678333087681914e-06, "loss": 6.1696, "step": 13920 }, { "epoch": 0.11, "learning_rate": 9.677875785959228e-06, "loss": 5.6327, "step": 13930 }, { "epoch": 0.11, "learning_rate": 9.677418170222464e-06, "loss": 6.5959, "step": 13940 }, { "epoch": 0.12, "learning_rate": 9.676960240502344e-06, "loss": 6.6046, "step": 13950 }, { "epoch": 0.12, "learning_rate": 9.676501996829605e-06, "loss": 6.5749, "step": 13960 }, { "epoch": 0.12, "learning_rate": 9.67604343923501e-06, "loss": 6.4392, "step": 13970 }, { "epoch": 0.12, "learning_rate": 9.675584567749338e-06, "loss": 6.7073, "step": 13980 }, { "epoch": 0.12, "learning_rate": 9.675125382403393e-06, "loss": 6.5281, "step": 13990 }, { "epoch": 0.12, "learning_rate": 9.674665883228e-06, "loss": 6.4552, "step": 14000 }, { "epoch": 0.12, "learning_rate": 9.674206070254005e-06, "loss": 6.4212, "step": 14010 }, { "epoch": 0.12, "learning_rate": 9.673745943512272e-06, "loss": 6.6803, "step": 14020 }, { "epoch": 0.12, "learning_rate": 9.673285503033687e-06, "loss": 6.5761, "step": 14030 }, { "epoch": 0.12, "learning_rate": 9.672824748849161e-06, "loss": 6.0542, "step": 14040 }, { "epoch": 0.12, "learning_rate": 9.672363680989622e-06, "loss": 6.8083, "step": 14050 }, { "epoch": 0.12, "learning_rate": 9.67190229948602e-06, "loss": 6.2529, "step": 14060 }, { "epoch": 0.12, "learning_rate": 9.671440604369329e-06, "loss": 6.4605, "step": 14070 }, { "epoch": 0.12, "learning_rate": 9.670978595670537e-06, "loss": 6.3891, "step": 14080 }, { "epoch": 0.12, "learning_rate": 9.67051627342066e-06, "loss": 6.0127, "step": 14090 }, { "epoch": 0.12, "learning_rate": 9.670053637650733e-06, "loss": 6.0455, "step": 14100 }, { "epoch": 0.12, "learning_rate": 9.66959068839181e-06, "loss": 5.9731, "step": 14110 }, { "epoch": 0.12, "learning_rate": 9.66912742567497e-06, "loss": 6.4377, "step": 14120 }, { "epoch": 0.12, "learning_rate": 9.668663849531308e-06, "loss": 6.2265, "step": 14130 }, { "epoch": 0.12, "learning_rate": 9.668199959991944e-06, "loss": 5.7971, "step": 14140 }, { "epoch": 0.12, "learning_rate": 9.66773575708802e-06, "loss": 6.4006, "step": 14150 }, { "epoch": 0.12, "learning_rate": 9.66727124085069e-06, "loss": 6.3652, "step": 14160 }, { "epoch": 0.12, "learning_rate": 9.666806411311144e-06, "loss": 6.0073, "step": 14170 }, { "epoch": 0.12, "learning_rate": 9.666341268500579e-06, "loss": 7.1072, "step": 14180 }, { "epoch": 0.12, "learning_rate": 9.665875812450222e-06, "loss": 6.5689, "step": 14190 }, { "epoch": 0.12, "learning_rate": 9.665410043191316e-06, "loss": 6.7703, "step": 14200 }, { "epoch": 0.12, "learning_rate": 9.664943960755129e-06, "loss": 6.0196, "step": 14210 }, { "epoch": 0.12, "learning_rate": 9.664477565172945e-06, "loss": 6.1114, "step": 14220 }, { "epoch": 0.12, "learning_rate": 9.664010856476075e-06, "loss": 6.8358, "step": 14230 }, { "epoch": 0.12, "learning_rate": 9.663543834695845e-06, "loss": 5.707, "step": 14240 }, { "epoch": 0.12, "learning_rate": 9.663076499863609e-06, "loss": 6.171, "step": 14250 }, { "epoch": 0.12, "learning_rate": 9.662608852010734e-06, "loss": 6.7076, "step": 14260 }, { "epoch": 0.12, "learning_rate": 9.662140891168615e-06, "loss": 6.7144, "step": 14270 }, { "epoch": 0.12, "learning_rate": 9.661672617368662e-06, "loss": 6.1291, "step": 14280 }, { "epoch": 0.12, "learning_rate": 9.661204030642312e-06, "loss": 6.2339, "step": 14290 }, { "epoch": 0.12, "learning_rate": 9.66073513102102e-06, "loss": 6.4012, "step": 14300 }, { "epoch": 0.12, "learning_rate": 9.660265918536257e-06, "loss": 6.6452, "step": 14310 }, { "epoch": 0.12, "learning_rate": 9.659796393219527e-06, "loss": 6.3273, "step": 14320 }, { "epoch": 0.12, "learning_rate": 9.659326555102346e-06, "loss": 6.5717, "step": 14330 }, { "epoch": 0.12, "learning_rate": 9.65885640421625e-06, "loss": 5.7329, "step": 14340 }, { "epoch": 0.12, "learning_rate": 9.658385940592801e-06, "loss": 6.4545, "step": 14350 }, { "epoch": 0.12, "learning_rate": 9.65791516426358e-06, "loss": 6.1697, "step": 14360 }, { "epoch": 0.12, "learning_rate": 9.657444075260192e-06, "loss": 6.1927, "step": 14370 }, { "epoch": 0.12, "learning_rate": 9.656972673614255e-06, "loss": 6.7866, "step": 14380 }, { "epoch": 0.12, "learning_rate": 9.656500959357415e-06, "loss": 5.9683, "step": 14390 }, { "epoch": 0.12, "learning_rate": 9.656028932521337e-06, "loss": 5.9157, "step": 14400 }, { "epoch": 0.12, "learning_rate": 9.655556593137709e-06, "loss": 5.8327, "step": 14410 }, { "epoch": 0.12, "learning_rate": 9.655083941238234e-06, "loss": 6.2445, "step": 14420 }, { "epoch": 0.12, "learning_rate": 9.654610976854644e-06, "loss": 6.0892, "step": 14430 }, { "epoch": 0.12, "learning_rate": 9.654137700018684e-06, "loss": 5.7366, "step": 14440 }, { "epoch": 0.12, "learning_rate": 9.653664110762128e-06, "loss": 6.7209, "step": 14450 }, { "epoch": 0.12, "learning_rate": 9.653190209116763e-06, "loss": 6.3733, "step": 14460 }, { "epoch": 0.12, "learning_rate": 9.652715995114404e-06, "loss": 6.3636, "step": 14470 }, { "epoch": 0.12, "learning_rate": 9.652241468786882e-06, "loss": 6.4046, "step": 14480 }, { "epoch": 0.12, "learning_rate": 9.651766630166051e-06, "loss": 6.5162, "step": 14490 }, { "epoch": 0.12, "learning_rate": 9.651291479283783e-06, "loss": 5.9903, "step": 14500 }, { "epoch": 0.12, "learning_rate": 9.650816016171981e-06, "loss": 6.3501, "step": 14510 }, { "epoch": 0.12, "learning_rate": 9.650340240862556e-06, "loss": 6.3986, "step": 14520 }, { "epoch": 0.12, "learning_rate": 9.649864153387445e-06, "loss": 7.6799, "step": 14530 }, { "epoch": 0.12, "learning_rate": 9.649387753778612e-06, "loss": 6.0048, "step": 14540 }, { "epoch": 0.12, "learning_rate": 9.64891104206803e-06, "loss": 6.1062, "step": 14550 }, { "epoch": 0.12, "learning_rate": 9.648434018287703e-06, "loss": 6.4333, "step": 14560 }, { "epoch": 0.12, "learning_rate": 9.647956682469651e-06, "loss": 6.4099, "step": 14570 }, { "epoch": 0.12, "learning_rate": 9.647479034645917e-06, "loss": 6.6059, "step": 14580 }, { "epoch": 0.12, "learning_rate": 9.647001074848567e-06, "loss": 6.6573, "step": 14590 }, { "epoch": 0.12, "learning_rate": 9.64652280310968e-06, "loss": 6.3868, "step": 14600 }, { "epoch": 0.12, "learning_rate": 9.646044219461365e-06, "loss": 6.7693, "step": 14610 }, { "epoch": 0.12, "learning_rate": 9.645565323935745e-06, "loss": 5.9339, "step": 14620 }, { "epoch": 0.12, "learning_rate": 9.645086116564973e-06, "loss": 5.9203, "step": 14630 }, { "epoch": 0.12, "learning_rate": 9.64460659738121e-06, "loss": 5.4229, "step": 14640 }, { "epoch": 0.12, "learning_rate": 9.644126766416648e-06, "loss": 6.171, "step": 14650 }, { "epoch": 0.12, "learning_rate": 9.643646623703498e-06, "loss": 6.5461, "step": 14660 }, { "epoch": 0.12, "learning_rate": 9.64316616927399e-06, "loss": 6.1718, "step": 14670 }, { "epoch": 0.12, "learning_rate": 9.642685403160375e-06, "loss": 5.6497, "step": 14680 }, { "epoch": 0.12, "learning_rate": 9.642204325394923e-06, "loss": 6.3084, "step": 14690 }, { "epoch": 0.12, "learning_rate": 9.641722936009935e-06, "loss": 6.3063, "step": 14700 }, { "epoch": 0.12, "learning_rate": 9.641241235037717e-06, "loss": 6.313, "step": 14710 }, { "epoch": 0.12, "learning_rate": 9.640759222510609e-06, "loss": 6.971, "step": 14720 }, { "epoch": 0.12, "learning_rate": 9.640276898460967e-06, "loss": 6.0585, "step": 14730 }, { "epoch": 0.12, "learning_rate": 9.639794262921169e-06, "loss": 6.4615, "step": 14740 }, { "epoch": 0.12, "learning_rate": 9.639311315923612e-06, "loss": 6.1044, "step": 14750 }, { "epoch": 0.12, "learning_rate": 9.638828057500713e-06, "loss": 5.8194, "step": 14760 }, { "epoch": 0.12, "learning_rate": 9.638344487684914e-06, "loss": 6.7079, "step": 14770 }, { "epoch": 0.12, "learning_rate": 9.637860606508677e-06, "loss": 6.5039, "step": 14780 }, { "epoch": 0.12, "learning_rate": 9.637376414004483e-06, "loss": 6.6242, "step": 14790 }, { "epoch": 0.12, "learning_rate": 9.63689191020483e-06, "loss": 5.813, "step": 14800 }, { "epoch": 0.12, "learning_rate": 9.63640709514225e-06, "loss": 5.5681, "step": 14810 }, { "epoch": 0.12, "learning_rate": 9.635921968849281e-06, "loss": 6.3972, "step": 14820 }, { "epoch": 0.12, "learning_rate": 9.63543653135849e-06, "loss": 6.4077, "step": 14830 }, { "epoch": 0.12, "learning_rate": 9.634950782702464e-06, "loss": 6.8424, "step": 14840 }, { "epoch": 0.12, "learning_rate": 9.63446472291381e-06, "loss": 6.0362, "step": 14850 }, { "epoch": 0.12, "learning_rate": 9.633978352025156e-06, "loss": 5.9675, "step": 14860 }, { "epoch": 0.12, "learning_rate": 9.633491670069149e-06, "loss": 6.4634, "step": 14870 }, { "epoch": 0.12, "learning_rate": 9.63300467707846e-06, "loss": 5.9647, "step": 14880 }, { "epoch": 0.12, "learning_rate": 9.632517373085781e-06, "loss": 6.1054, "step": 14890 }, { "epoch": 0.12, "learning_rate": 9.632029758123822e-06, "loss": 6.5513, "step": 14900 }, { "epoch": 0.12, "learning_rate": 9.631541832225315e-06, "loss": 6.1608, "step": 14910 }, { "epoch": 0.12, "learning_rate": 9.631053595423015e-06, "loss": 6.3216, "step": 14920 }, { "epoch": 0.12, "learning_rate": 9.630565047749693e-06, "loss": 6.5167, "step": 14930 }, { "epoch": 0.12, "learning_rate": 9.630076189238147e-06, "loss": 6.0896, "step": 14940 }, { "epoch": 0.12, "learning_rate": 9.629587019921191e-06, "loss": 6.6915, "step": 14950 }, { "epoch": 0.12, "learning_rate": 9.629097539831664e-06, "loss": 6.044, "step": 14960 }, { "epoch": 0.12, "learning_rate": 9.628607749002421e-06, "loss": 6.5043, "step": 14970 }, { "epoch": 0.12, "learning_rate": 9.628117647466344e-06, "loss": 6.5433, "step": 14980 }, { "epoch": 0.12, "learning_rate": 9.627627235256326e-06, "loss": 6.6105, "step": 14990 }, { "epoch": 0.12, "learning_rate": 9.627136512405293e-06, "loss": 6.9223, "step": 15000 }, { "epoch": 0.12, "learning_rate": 9.626645478946184e-06, "loss": 6.5159, "step": 15010 }, { "epoch": 0.12, "learning_rate": 9.62615413491196e-06, "loss": 6.0351, "step": 15020 }, { "epoch": 0.12, "learning_rate": 9.625662480335606e-06, "loss": 6.7188, "step": 15030 }, { "epoch": 0.12, "learning_rate": 9.625170515250123e-06, "loss": 6.2485, "step": 15040 }, { "epoch": 0.12, "learning_rate": 9.624678239688536e-06, "loss": 6.2657, "step": 15050 }, { "epoch": 0.12, "learning_rate": 9.624185653683891e-06, "loss": 6.043, "step": 15060 }, { "epoch": 0.12, "learning_rate": 9.623692757269255e-06, "loss": 6.4555, "step": 15070 }, { "epoch": 0.12, "learning_rate": 9.62319955047771e-06, "loss": 5.9708, "step": 15080 }, { "epoch": 0.12, "learning_rate": 9.622706033342371e-06, "loss": 6.4678, "step": 15090 }, { "epoch": 0.12, "learning_rate": 9.622212205896363e-06, "loss": 6.6896, "step": 15100 }, { "epoch": 0.12, "learning_rate": 9.621718068172832e-06, "loss": 6.7853, "step": 15110 }, { "epoch": 0.12, "learning_rate": 9.621223620204954e-06, "loss": 6.7508, "step": 15120 }, { "epoch": 0.12, "learning_rate": 9.620728862025916e-06, "loss": 6.3508, "step": 15130 }, { "epoch": 0.12, "learning_rate": 9.620233793668933e-06, "loss": 6.9548, "step": 15140 }, { "epoch": 0.12, "learning_rate": 9.619738415167237e-06, "loss": 5.9917, "step": 15150 }, { "epoch": 0.13, "learning_rate": 9.619242726554078e-06, "loss": 6.2051, "step": 15160 }, { "epoch": 0.13, "learning_rate": 9.618746727862736e-06, "loss": 6.0252, "step": 15170 }, { "epoch": 0.13, "learning_rate": 9.6182504191265e-06, "loss": 6.4987, "step": 15180 }, { "epoch": 0.13, "learning_rate": 9.61775380037869e-06, "loss": 6.8654, "step": 15190 }, { "epoch": 0.13, "learning_rate": 9.617256871652642e-06, "loss": 6.3167, "step": 15200 }, { "epoch": 0.13, "learning_rate": 9.616759632981716e-06, "loss": 6.1057, "step": 15210 }, { "epoch": 0.13, "learning_rate": 9.616262084399285e-06, "loss": 6.5653, "step": 15220 }, { "epoch": 0.13, "learning_rate": 9.615764225938751e-06, "loss": 6.7167, "step": 15230 }, { "epoch": 0.13, "learning_rate": 9.615266057633536e-06, "loss": 6.0372, "step": 15240 }, { "epoch": 0.13, "learning_rate": 9.614767579517077e-06, "loss": 5.5193, "step": 15250 }, { "epoch": 0.13, "learning_rate": 9.614268791622838e-06, "loss": 5.848, "step": 15260 }, { "epoch": 0.13, "learning_rate": 9.6137696939843e-06, "loss": 6.1961, "step": 15270 }, { "epoch": 0.13, "learning_rate": 9.613270286634968e-06, "loss": 6.1664, "step": 15280 }, { "epoch": 0.13, "learning_rate": 9.612770569608364e-06, "loss": 6.3774, "step": 15290 }, { "epoch": 0.13, "learning_rate": 9.612270542938035e-06, "loss": 5.4234, "step": 15300 }, { "epoch": 0.13, "learning_rate": 9.611770206657546e-06, "loss": 6.1481, "step": 15310 }, { "epoch": 0.13, "learning_rate": 9.61126956080048e-06, "loss": 5.8617, "step": 15320 }, { "epoch": 0.13, "learning_rate": 9.610768605400448e-06, "loss": 6.3251, "step": 15330 }, { "epoch": 0.13, "learning_rate": 9.610267340491078e-06, "loss": 6.3627, "step": 15340 }, { "epoch": 0.13, "learning_rate": 9.609765766106013e-06, "loss": 5.8928, "step": 15350 }, { "epoch": 0.13, "learning_rate": 9.60926388227893e-06, "loss": 5.7295, "step": 15360 }, { "epoch": 0.13, "learning_rate": 9.608761689043515e-06, "loss": 6.7369, "step": 15370 }, { "epoch": 0.13, "learning_rate": 9.60825918643348e-06, "loss": 6.1276, "step": 15380 }, { "epoch": 0.13, "learning_rate": 9.607756374482556e-06, "loss": 6.3328, "step": 15390 }, { "epoch": 0.13, "learning_rate": 9.607253253224496e-06, "loss": 6.7382, "step": 15400 }, { "epoch": 0.13, "learning_rate": 9.606749822693074e-06, "loss": 7.0521, "step": 15410 }, { "epoch": 0.13, "learning_rate": 9.606246082922081e-06, "loss": 6.0396, "step": 15420 }, { "epoch": 0.13, "learning_rate": 9.605742033945338e-06, "loss": 6.8157, "step": 15430 }, { "epoch": 0.13, "learning_rate": 9.605237675796676e-06, "loss": 6.4278, "step": 15440 }, { "epoch": 0.13, "learning_rate": 9.604733008509947e-06, "loss": 6.372, "step": 15450 }, { "epoch": 0.13, "learning_rate": 9.604228032119039e-06, "loss": 6.0642, "step": 15460 }, { "epoch": 0.13, "learning_rate": 9.60372274665784e-06, "loss": 6.3368, "step": 15470 }, { "epoch": 0.13, "learning_rate": 9.603217152160273e-06, "loss": 6.1459, "step": 15480 }, { "epoch": 0.13, "learning_rate": 9.60271124866028e-06, "loss": 5.5309, "step": 15490 }, { "epoch": 0.13, "learning_rate": 9.602205036191813e-06, "loss": 6.359, "step": 15500 }, { "epoch": 0.13, "learning_rate": 9.601698514788859e-06, "loss": 6.0406, "step": 15510 }, { "epoch": 0.13, "learning_rate": 9.60119168448542e-06, "loss": 6.3776, "step": 15520 }, { "epoch": 0.13, "learning_rate": 9.600684545315514e-06, "loss": 6.6905, "step": 15530 }, { "epoch": 0.13, "learning_rate": 9.600177097313188e-06, "loss": 5.9575, "step": 15540 }, { "epoch": 0.13, "learning_rate": 9.599669340512502e-06, "loss": 6.256, "step": 15550 }, { "epoch": 0.13, "learning_rate": 9.599161274947543e-06, "loss": 6.6221, "step": 15560 }, { "epoch": 0.13, "learning_rate": 9.598652900652417e-06, "loss": 6.9258, "step": 15570 }, { "epoch": 0.13, "learning_rate": 9.598144217661244e-06, "loss": 6.7363, "step": 15580 }, { "epoch": 0.13, "learning_rate": 9.597635226008179e-06, "loss": 6.3662, "step": 15590 }, { "epoch": 0.13, "learning_rate": 9.597125925727385e-06, "loss": 5.8699, "step": 15600 }, { "epoch": 0.13, "learning_rate": 9.59661631685305e-06, "loss": 6.4427, "step": 15610 }, { "epoch": 0.13, "learning_rate": 9.596106399419383e-06, "loss": 6.1958, "step": 15620 }, { "epoch": 0.13, "learning_rate": 9.595596173460614e-06, "loss": 6.0528, "step": 15630 }, { "epoch": 0.13, "learning_rate": 9.595085639010993e-06, "loss": 5.7121, "step": 15640 }, { "epoch": 0.13, "learning_rate": 9.59457479610479e-06, "loss": 6.3628, "step": 15650 }, { "epoch": 0.13, "learning_rate": 9.594063644776296e-06, "loss": 6.1215, "step": 15660 }, { "epoch": 0.13, "learning_rate": 9.593552185059827e-06, "loss": 6.07, "step": 15670 }, { "epoch": 0.13, "learning_rate": 9.593040416989713e-06, "loss": 6.3629, "step": 15680 }, { "epoch": 0.13, "learning_rate": 9.592528340600308e-06, "loss": 5.8222, "step": 15690 }, { "epoch": 0.13, "learning_rate": 9.592015955925987e-06, "loss": 6.9316, "step": 15700 }, { "epoch": 0.13, "learning_rate": 9.591503263001144e-06, "loss": 6.9101, "step": 15710 }, { "epoch": 0.13, "learning_rate": 9.590990261860198e-06, "loss": 7.121, "step": 15720 }, { "epoch": 0.13, "learning_rate": 9.590476952537582e-06, "loss": 6.8031, "step": 15730 }, { "epoch": 0.13, "learning_rate": 9.589963335067753e-06, "loss": 6.2942, "step": 15740 }, { "epoch": 0.13, "learning_rate": 9.589449409485192e-06, "loss": 6.7447, "step": 15750 }, { "epoch": 0.13, "learning_rate": 9.588935175824393e-06, "loss": 6.201, "step": 15760 }, { "epoch": 0.13, "learning_rate": 9.588420634119882e-06, "loss": 6.5905, "step": 15770 }, { "epoch": 0.13, "learning_rate": 9.587905784406193e-06, "loss": 6.7995, "step": 15780 }, { "epoch": 0.13, "learning_rate": 9.587390626717887e-06, "loss": 6.3148, "step": 15790 }, { "epoch": 0.13, "learning_rate": 9.586875161089549e-06, "loss": 5.9388, "step": 15800 }, { "epoch": 0.13, "learning_rate": 9.586359387555776e-06, "loss": 5.4104, "step": 15810 }, { "epoch": 0.13, "learning_rate": 9.585843306151195e-06, "loss": 6.0367, "step": 15820 }, { "epoch": 0.13, "learning_rate": 9.585326916910447e-06, "loss": 6.3543, "step": 15830 }, { "epoch": 0.13, "learning_rate": 9.584810219868195e-06, "loss": 6.9452, "step": 15840 }, { "epoch": 0.13, "learning_rate": 9.584293215059126e-06, "loss": 6.0734, "step": 15850 }, { "epoch": 0.13, "learning_rate": 9.583775902517945e-06, "loss": 6.5714, "step": 15860 }, { "epoch": 0.13, "learning_rate": 9.583258282279375e-06, "loss": 6.1431, "step": 15870 }, { "epoch": 0.13, "learning_rate": 9.582740354378165e-06, "loss": 6.5943, "step": 15880 }, { "epoch": 0.13, "learning_rate": 9.582222118849083e-06, "loss": 6.4599, "step": 15890 }, { "epoch": 0.13, "learning_rate": 9.581703575726914e-06, "loss": 6.2359, "step": 15900 }, { "epoch": 0.13, "learning_rate": 9.581184725046468e-06, "loss": 5.9677, "step": 15910 }, { "epoch": 0.13, "learning_rate": 9.580665566842577e-06, "loss": 5.6332, "step": 15920 }, { "epoch": 0.13, "learning_rate": 9.580146101150085e-06, "loss": 6.7686, "step": 15930 }, { "epoch": 0.13, "learning_rate": 9.579626328003868e-06, "loss": 6.1504, "step": 15940 }, { "epoch": 0.13, "learning_rate": 9.579106247438813e-06, "loss": 5.3625, "step": 15950 }, { "epoch": 0.13, "learning_rate": 9.578585859489833e-06, "loss": 5.8942, "step": 15960 }, { "epoch": 0.13, "learning_rate": 9.578065164191862e-06, "loss": 6.3635, "step": 15970 }, { "epoch": 0.13, "learning_rate": 9.57754416157985e-06, "loss": 6.3212, "step": 15980 }, { "epoch": 0.13, "learning_rate": 9.577022851688772e-06, "loss": 6.4019, "step": 15990 }, { "epoch": 0.13, "learning_rate": 9.576501234553623e-06, "loss": 6.6171, "step": 16000 }, { "epoch": 0.13, "learning_rate": 9.575979310209417e-06, "loss": 6.5219, "step": 16010 }, { "epoch": 0.13, "learning_rate": 9.575457078691191e-06, "loss": 6.2247, "step": 16020 }, { "epoch": 0.13, "learning_rate": 9.574934540034e-06, "loss": 5.9623, "step": 16030 }, { "epoch": 0.13, "learning_rate": 9.574411694272918e-06, "loss": 7.0051, "step": 16040 }, { "epoch": 0.13, "learning_rate": 9.573888541443047e-06, "loss": 6.3386, "step": 16050 }, { "epoch": 0.13, "learning_rate": 9.573365081579501e-06, "loss": 6.568, "step": 16060 }, { "epoch": 0.13, "learning_rate": 9.572841314717422e-06, "loss": 6.1139, "step": 16070 }, { "epoch": 0.13, "learning_rate": 9.572317240891967e-06, "loss": 6.5346, "step": 16080 }, { "epoch": 0.13, "learning_rate": 9.571792860138316e-06, "loss": 5.9908, "step": 16090 }, { "epoch": 0.13, "learning_rate": 9.57126817249167e-06, "loss": 6.8558, "step": 16100 }, { "epoch": 0.13, "learning_rate": 9.570743177987249e-06, "loss": 6.2324, "step": 16110 }, { "epoch": 0.13, "learning_rate": 9.570217876660295e-06, "loss": 6.4444, "step": 16120 }, { "epoch": 0.13, "learning_rate": 9.56969226854607e-06, "loss": 6.2591, "step": 16130 }, { "epoch": 0.13, "learning_rate": 9.56916635367986e-06, "loss": 6.223, "step": 16140 }, { "epoch": 0.13, "learning_rate": 9.568640132096963e-06, "loss": 6.6914, "step": 16150 }, { "epoch": 0.13, "learning_rate": 9.568113603832705e-06, "loss": 6.0392, "step": 16160 }, { "epoch": 0.13, "learning_rate": 9.56758676892243e-06, "loss": 6.2301, "step": 16170 }, { "epoch": 0.13, "learning_rate": 9.567059627401507e-06, "loss": 6.7116, "step": 16180 }, { "epoch": 0.13, "learning_rate": 9.566532179305316e-06, "loss": 5.7683, "step": 16190 }, { "epoch": 0.13, "learning_rate": 9.566004424669266e-06, "loss": 6.5423, "step": 16200 }, { "epoch": 0.13, "learning_rate": 9.565476363528785e-06, "loss": 6.7029, "step": 16210 }, { "epoch": 0.13, "learning_rate": 9.564947995919317e-06, "loss": 6.2306, "step": 16220 }, { "epoch": 0.13, "learning_rate": 9.564419321876335e-06, "loss": 6.0164, "step": 16230 }, { "epoch": 0.13, "learning_rate": 9.563890341435325e-06, "loss": 6.523, "step": 16240 }, { "epoch": 0.13, "learning_rate": 9.563361054631794e-06, "loss": 6.4399, "step": 16250 }, { "epoch": 0.13, "learning_rate": 9.562831461501274e-06, "loss": 6.4875, "step": 16260 }, { "epoch": 0.13, "learning_rate": 9.562301562079315e-06, "loss": 6.3356, "step": 16270 }, { "epoch": 0.13, "learning_rate": 9.561771356401487e-06, "loss": 6.4586, "step": 16280 }, { "epoch": 0.13, "learning_rate": 9.561240844503384e-06, "loss": 6.0144, "step": 16290 }, { "epoch": 0.13, "learning_rate": 9.560710026420615e-06, "loss": 6.1158, "step": 16300 }, { "epoch": 0.13, "learning_rate": 9.560178902188812e-06, "loss": 6.1322, "step": 16310 }, { "epoch": 0.13, "learning_rate": 9.559647471843631e-06, "loss": 5.5444, "step": 16320 }, { "epoch": 0.13, "learning_rate": 9.559115735420743e-06, "loss": 6.6331, "step": 16330 }, { "epoch": 0.13, "learning_rate": 9.558583692955844e-06, "loss": 6.0943, "step": 16340 }, { "epoch": 0.13, "learning_rate": 9.558051344484647e-06, "loss": 6.9481, "step": 16350 }, { "epoch": 0.13, "learning_rate": 9.557518690042889e-06, "loss": 6.3376, "step": 16360 }, { "epoch": 0.14, "learning_rate": 9.556985729666325e-06, "loss": 5.8536, "step": 16370 }, { "epoch": 0.14, "learning_rate": 9.55645246339073e-06, "loss": 6.1013, "step": 16380 }, { "epoch": 0.14, "learning_rate": 9.555918891251903e-06, "loss": 6.4571, "step": 16390 }, { "epoch": 0.14, "learning_rate": 9.55538501328566e-06, "loss": 6.4749, "step": 16400 }, { "epoch": 0.14, "learning_rate": 9.55485082952784e-06, "loss": 6.5222, "step": 16410 }, { "epoch": 0.14, "learning_rate": 9.5543163400143e-06, "loss": 6.7155, "step": 16420 }, { "epoch": 0.14, "learning_rate": 9.55378154478092e-06, "loss": 6.0568, "step": 16430 }, { "epoch": 0.14, "learning_rate": 9.553246443863599e-06, "loss": 6.2224, "step": 16440 }, { "epoch": 0.14, "learning_rate": 9.552711037298257e-06, "loss": 6.7842, "step": 16450 }, { "epoch": 0.14, "learning_rate": 9.552175325120833e-06, "loss": 6.4365, "step": 16460 }, { "epoch": 0.14, "learning_rate": 9.551639307367291e-06, "loss": 6.1685, "step": 16470 }, { "epoch": 0.14, "learning_rate": 9.551102984073611e-06, "loss": 6.506, "step": 16480 }, { "epoch": 0.14, "learning_rate": 9.550566355275796e-06, "loss": 6.2681, "step": 16490 }, { "epoch": 0.14, "learning_rate": 9.550029421009865e-06, "loss": 6.1159, "step": 16500 }, { "epoch": 0.14, "learning_rate": 9.549492181311865e-06, "loss": 5.7845, "step": 16510 }, { "epoch": 0.14, "learning_rate": 9.54895463621786e-06, "loss": 5.8015, "step": 16520 }, { "epoch": 0.14, "learning_rate": 9.54841678576393e-06, "loss": 6.4017, "step": 16530 }, { "epoch": 0.14, "learning_rate": 9.547878629986184e-06, "loss": 6.525, "step": 16540 }, { "epoch": 0.14, "learning_rate": 9.547340168920745e-06, "loss": 6.475, "step": 16550 }, { "epoch": 0.14, "learning_rate": 9.546801402603756e-06, "loss": 6.3476, "step": 16560 }, { "epoch": 0.14, "learning_rate": 9.546262331071387e-06, "loss": 6.0086, "step": 16570 }, { "epoch": 0.14, "learning_rate": 9.545722954359822e-06, "loss": 6.4521, "step": 16580 }, { "epoch": 0.14, "learning_rate": 9.545183272505271e-06, "loss": 6.1249, "step": 16590 }, { "epoch": 0.14, "learning_rate": 9.544643285543957e-06, "loss": 6.8428, "step": 16600 }, { "epoch": 0.14, "learning_rate": 9.544102993512133e-06, "loss": 5.8829, "step": 16610 }, { "epoch": 0.14, "learning_rate": 9.543562396446064e-06, "loss": 6.2599, "step": 16620 }, { "epoch": 0.14, "learning_rate": 9.54302149438204e-06, "loss": 7.013, "step": 16630 }, { "epoch": 0.14, "learning_rate": 9.542480287356369e-06, "loss": 6.2635, "step": 16640 }, { "epoch": 0.14, "learning_rate": 9.541938775405383e-06, "loss": 6.2717, "step": 16650 }, { "epoch": 0.14, "learning_rate": 9.541396958565432e-06, "loss": 6.9572, "step": 16660 }, { "epoch": 0.14, "learning_rate": 9.540854836872887e-06, "loss": 6.4323, "step": 16670 }, { "epoch": 0.14, "learning_rate": 9.540312410364136e-06, "loss": 6.507, "step": 16680 }, { "epoch": 0.14, "learning_rate": 9.539769679075594e-06, "loss": 5.786, "step": 16690 }, { "epoch": 0.14, "learning_rate": 9.539226643043694e-06, "loss": 6.5614, "step": 16700 }, { "epoch": 0.14, "learning_rate": 9.538683302304887e-06, "loss": 6.7405, "step": 16710 }, { "epoch": 0.14, "learning_rate": 9.538139656895646e-06, "loss": 6.4437, "step": 16720 }, { "epoch": 0.14, "learning_rate": 9.537595706852464e-06, "loss": 5.5011, "step": 16730 }, { "epoch": 0.14, "learning_rate": 9.537051452211856e-06, "loss": 6.0814, "step": 16740 }, { "epoch": 0.14, "learning_rate": 9.536506893010355e-06, "loss": 6.4854, "step": 16750 }, { "epoch": 0.14, "learning_rate": 9.53596202928452e-06, "loss": 6.1098, "step": 16760 }, { "epoch": 0.14, "learning_rate": 9.535416861070922e-06, "loss": 6.5878, "step": 16770 }, { "epoch": 0.14, "learning_rate": 9.534871388406156e-06, "loss": 6.1166, "step": 16780 }, { "epoch": 0.14, "learning_rate": 9.534325611326844e-06, "loss": 6.4738, "step": 16790 }, { "epoch": 0.14, "learning_rate": 9.533779529869618e-06, "loss": 6.1007, "step": 16800 }, { "epoch": 0.14, "learning_rate": 9.533233144071137e-06, "loss": 6.2552, "step": 16810 }, { "epoch": 0.14, "learning_rate": 9.532686453968077e-06, "loss": 5.8754, "step": 16820 }, { "epoch": 0.14, "learning_rate": 9.532139459597134e-06, "loss": 6.4393, "step": 16830 }, { "epoch": 0.14, "learning_rate": 9.531592160995032e-06, "loss": 6.2587, "step": 16840 }, { "epoch": 0.14, "learning_rate": 9.531044558198506e-06, "loss": 6.5004, "step": 16850 }, { "epoch": 0.14, "learning_rate": 9.530496651244316e-06, "loss": 6.2693, "step": 16860 }, { "epoch": 0.14, "learning_rate": 9.529948440169241e-06, "loss": 6.754, "step": 16870 }, { "epoch": 0.14, "learning_rate": 9.529399925010083e-06, "loss": 6.1816, "step": 16880 }, { "epoch": 0.14, "learning_rate": 9.52885110580366e-06, "loss": 5.2962, "step": 16890 }, { "epoch": 0.14, "learning_rate": 9.528301982586814e-06, "loss": 6.8618, "step": 16900 }, { "epoch": 0.14, "learning_rate": 9.527752555396406e-06, "loss": 6.8784, "step": 16910 }, { "epoch": 0.14, "learning_rate": 9.52720282426932e-06, "loss": 5.8303, "step": 16920 }, { "epoch": 0.14, "learning_rate": 9.526652789242454e-06, "loss": 6.0075, "step": 16930 }, { "epoch": 0.14, "learning_rate": 9.526102450352733e-06, "loss": 6.7433, "step": 16940 }, { "epoch": 0.14, "learning_rate": 9.525551807637098e-06, "loss": 5.9563, "step": 16950 }, { "epoch": 0.14, "learning_rate": 9.525000861132515e-06, "loss": 6.2847, "step": 16960 }, { "epoch": 0.14, "learning_rate": 9.524449610875966e-06, "loss": 6.0202, "step": 16970 }, { "epoch": 0.14, "learning_rate": 9.523898056904454e-06, "loss": 6.323, "step": 16980 }, { "epoch": 0.14, "learning_rate": 9.523346199255006e-06, "loss": 5.7249, "step": 16990 }, { "epoch": 0.14, "learning_rate": 9.522794037964664e-06, "loss": 6.1751, "step": 17000 }, { "epoch": 0.14, "learning_rate": 9.522241573070497e-06, "loss": 6.3995, "step": 17010 }, { "epoch": 0.14, "learning_rate": 9.521688804609588e-06, "loss": 7.0487, "step": 17020 }, { "epoch": 0.14, "learning_rate": 9.521135732619042e-06, "loss": 6.1471, "step": 17030 }, { "epoch": 0.14, "learning_rate": 9.520582357135986e-06, "loss": 5.9161, "step": 17040 }, { "epoch": 0.14, "learning_rate": 9.52002867819757e-06, "loss": 6.1465, "step": 17050 }, { "epoch": 0.14, "learning_rate": 9.519474695840956e-06, "loss": 6.0492, "step": 17060 }, { "epoch": 0.14, "learning_rate": 9.518920410103334e-06, "loss": 5.7289, "step": 17070 }, { "epoch": 0.14, "learning_rate": 9.518365821021913e-06, "loss": 6.5029, "step": 17080 }, { "epoch": 0.14, "learning_rate": 9.517810928633919e-06, "loss": 6.0918, "step": 17090 }, { "epoch": 0.14, "learning_rate": 9.517255732976601e-06, "loss": 6.1753, "step": 17100 }, { "epoch": 0.14, "learning_rate": 9.516700234087228e-06, "loss": 6.0443, "step": 17110 }, { "epoch": 0.14, "learning_rate": 9.516144432003092e-06, "loss": 6.7854, "step": 17120 }, { "epoch": 0.14, "learning_rate": 9.515588326761497e-06, "loss": 6.2938, "step": 17130 }, { "epoch": 0.14, "learning_rate": 9.515031918399777e-06, "loss": 6.6337, "step": 17140 }, { "epoch": 0.14, "learning_rate": 9.514475206955282e-06, "loss": 6.2669, "step": 17150 }, { "epoch": 0.14, "learning_rate": 9.51391819246538e-06, "loss": 5.785, "step": 17160 }, { "epoch": 0.14, "learning_rate": 9.513360874967467e-06, "loss": 6.2549, "step": 17170 }, { "epoch": 0.14, "learning_rate": 9.51280325449895e-06, "loss": 5.9376, "step": 17180 }, { "epoch": 0.14, "learning_rate": 9.512245331097263e-06, "loss": 5.7299, "step": 17190 }, { "epoch": 0.14, "learning_rate": 9.511687104799855e-06, "loss": 6.4711, "step": 17200 }, { "epoch": 0.14, "learning_rate": 9.511128575644202e-06, "loss": 5.617, "step": 17210 }, { "epoch": 0.14, "learning_rate": 9.510569743667794e-06, "loss": 6.2935, "step": 17220 }, { "epoch": 0.14, "learning_rate": 9.510010608908145e-06, "loss": 5.6228, "step": 17230 }, { "epoch": 0.14, "learning_rate": 9.509451171402789e-06, "loss": 6.0667, "step": 17240 }, { "epoch": 0.14, "learning_rate": 9.508891431189278e-06, "loss": 6.7139, "step": 17250 }, { "epoch": 0.14, "learning_rate": 9.508331388305185e-06, "loss": 6.5745, "step": 17260 }, { "epoch": 0.14, "learning_rate": 9.50777104278811e-06, "loss": 6.7852, "step": 17270 }, { "epoch": 0.14, "learning_rate": 9.50721039467566e-06, "loss": 5.7563, "step": 17280 }, { "epoch": 0.14, "learning_rate": 9.506649444005476e-06, "loss": 5.9556, "step": 17290 }, { "epoch": 0.14, "learning_rate": 9.50608819081521e-06, "loss": 5.9633, "step": 17300 }, { "epoch": 0.14, "learning_rate": 9.50552663514254e-06, "loss": 5.921, "step": 17310 }, { "epoch": 0.14, "learning_rate": 9.504964777025159e-06, "loss": 5.6698, "step": 17320 }, { "epoch": 0.14, "learning_rate": 9.504402616500785e-06, "loss": 6.021, "step": 17330 }, { "epoch": 0.14, "learning_rate": 9.503840153607153e-06, "loss": 6.0231, "step": 17340 }, { "epoch": 0.14, "learning_rate": 9.50327738838202e-06, "loss": 6.1535, "step": 17350 }, { "epoch": 0.14, "learning_rate": 9.502714320863165e-06, "loss": 5.8285, "step": 17360 }, { "epoch": 0.14, "learning_rate": 9.502150951088382e-06, "loss": 6.3964, "step": 17370 }, { "epoch": 0.14, "learning_rate": 9.501587279095491e-06, "loss": 6.0944, "step": 17380 }, { "epoch": 0.14, "learning_rate": 9.501023304922329e-06, "loss": 6.7453, "step": 17390 }, { "epoch": 0.14, "learning_rate": 9.500459028606755e-06, "loss": 6.3534, "step": 17400 }, { "epoch": 0.14, "learning_rate": 9.499894450186648e-06, "loss": 6.7022, "step": 17410 }, { "epoch": 0.14, "learning_rate": 9.499329569699905e-06, "loss": 5.5776, "step": 17420 }, { "epoch": 0.14, "learning_rate": 9.498764387184445e-06, "loss": 6.0179, "step": 17430 }, { "epoch": 0.14, "learning_rate": 9.498198902678207e-06, "loss": 5.7474, "step": 17440 }, { "epoch": 0.14, "learning_rate": 9.497633116219153e-06, "loss": 5.9471, "step": 17450 }, { "epoch": 0.14, "learning_rate": 9.497067027845262e-06, "loss": 5.7323, "step": 17460 }, { "epoch": 0.14, "learning_rate": 9.496500637594532e-06, "loss": 5.1723, "step": 17470 }, { "epoch": 0.14, "learning_rate": 9.495933945504985e-06, "loss": 6.6303, "step": 17480 }, { "epoch": 0.14, "learning_rate": 9.495366951614661e-06, "loss": 6.0336, "step": 17490 }, { "epoch": 0.14, "learning_rate": 9.494799655961622e-06, "loss": 5.9829, "step": 17500 }, { "epoch": 0.14, "learning_rate": 9.494232058583947e-06, "loss": 5.815, "step": 17510 }, { "epoch": 0.14, "learning_rate": 9.493664159519741e-06, "loss": 6.1694, "step": 17520 }, { "epoch": 0.14, "learning_rate": 9.493095958807122e-06, "loss": 6.3654, "step": 17530 }, { "epoch": 0.14, "learning_rate": 9.492527456484236e-06, "loss": 6.5538, "step": 17540 }, { "epoch": 0.14, "learning_rate": 9.49195865258924e-06, "loss": 6.3851, "step": 17550 }, { "epoch": 0.14, "learning_rate": 9.491389547160318e-06, "loss": 6.4404, "step": 17560 }, { "epoch": 0.14, "learning_rate": 9.490820140235675e-06, "loss": 5.6737, "step": 17570 }, { "epoch": 0.14, "learning_rate": 9.490250431853532e-06, "loss": 6.1935, "step": 17580 }, { "epoch": 0.15, "learning_rate": 9.489680422052132e-06, "loss": 6.1981, "step": 17590 }, { "epoch": 0.15, "learning_rate": 9.48911011086974e-06, "loss": 5.7988, "step": 17600 }, { "epoch": 0.15, "learning_rate": 9.488539498344635e-06, "loss": 6.3997, "step": 17610 }, { "epoch": 0.15, "learning_rate": 9.487968584515128e-06, "loss": 6.5971, "step": 17620 }, { "epoch": 0.15, "learning_rate": 9.487397369419537e-06, "loss": 6.6141, "step": 17630 }, { "epoch": 0.15, "learning_rate": 9.48682585309621e-06, "loss": 6.0743, "step": 17640 }, { "epoch": 0.15, "learning_rate": 9.486254035583509e-06, "loss": 6.6594, "step": 17650 }, { "epoch": 0.15, "learning_rate": 9.485681916919817e-06, "loss": 6.1286, "step": 17660 }, { "epoch": 0.15, "learning_rate": 9.485109497143545e-06, "loss": 6.0008, "step": 17670 }, { "epoch": 0.15, "learning_rate": 9.484536776293113e-06, "loss": 6.4315, "step": 17680 }, { "epoch": 0.15, "learning_rate": 9.483963754406969e-06, "loss": 6.0936, "step": 17690 }, { "epoch": 0.15, "learning_rate": 9.483390431523575e-06, "loss": 6.3142, "step": 17700 }, { "epoch": 0.15, "learning_rate": 9.482816807681422e-06, "loss": 6.3266, "step": 17710 }, { "epoch": 0.15, "learning_rate": 9.482242882919011e-06, "loss": 6.0344, "step": 17720 }, { "epoch": 0.15, "learning_rate": 9.48166865727487e-06, "loss": 6.0118, "step": 17730 }, { "epoch": 0.15, "learning_rate": 9.481094130787545e-06, "loss": 5.7229, "step": 17740 }, { "epoch": 0.15, "learning_rate": 9.480519303495605e-06, "loss": 6.5476, "step": 17750 }, { "epoch": 0.15, "learning_rate": 9.479944175437633e-06, "loss": 6.2084, "step": 17760 }, { "epoch": 0.15, "learning_rate": 9.479368746652238e-06, "loss": 6.5008, "step": 17770 }, { "epoch": 0.15, "learning_rate": 9.478793017178047e-06, "loss": 6.8064, "step": 17780 }, { "epoch": 0.15, "learning_rate": 9.478216987053706e-06, "loss": 5.9039, "step": 17790 }, { "epoch": 0.15, "learning_rate": 9.477640656317885e-06, "loss": 6.7213, "step": 17800 }, { "epoch": 0.15, "learning_rate": 9.477064025009269e-06, "loss": 6.0476, "step": 17810 }, { "epoch": 0.15, "learning_rate": 9.476487093166566e-06, "loss": 6.3539, "step": 17820 }, { "epoch": 0.15, "learning_rate": 9.475909860828507e-06, "loss": 6.5475, "step": 17830 }, { "epoch": 0.15, "learning_rate": 9.475332328033837e-06, "loss": 5.806, "step": 17840 }, { "epoch": 0.15, "learning_rate": 9.474754494821325e-06, "loss": 7.1027, "step": 17850 }, { "epoch": 0.15, "learning_rate": 9.474176361229761e-06, "loss": 5.9355, "step": 17860 }, { "epoch": 0.15, "learning_rate": 9.47359792729795e-06, "loss": 6.4536, "step": 17870 }, { "epoch": 0.15, "learning_rate": 9.473019193064725e-06, "loss": 6.3954, "step": 17880 }, { "epoch": 0.15, "learning_rate": 9.472440158568934e-06, "loss": 6.5164, "step": 17890 }, { "epoch": 0.15, "learning_rate": 9.471860823849444e-06, "loss": 6.2613, "step": 17900 }, { "epoch": 0.15, "learning_rate": 9.471281188945148e-06, "loss": 6.0053, "step": 17910 }, { "epoch": 0.15, "learning_rate": 9.470701253894951e-06, "loss": 7.2422, "step": 17920 }, { "epoch": 0.15, "learning_rate": 9.470121018737785e-06, "loss": 5.821, "step": 17930 }, { "epoch": 0.15, "learning_rate": 9.4695404835126e-06, "loss": 6.1803, "step": 17940 }, { "epoch": 0.15, "learning_rate": 9.468959648258365e-06, "loss": 6.503, "step": 17950 }, { "epoch": 0.15, "learning_rate": 9.468378513014069e-06, "loss": 6.1656, "step": 17960 }, { "epoch": 0.15, "learning_rate": 9.467797077818726e-06, "loss": 6.116, "step": 17970 }, { "epoch": 0.15, "learning_rate": 9.467215342711363e-06, "loss": 6.0137, "step": 17980 }, { "epoch": 0.15, "learning_rate": 9.466633307731031e-06, "loss": 6.1521, "step": 17990 }, { "epoch": 0.15, "learning_rate": 9.4660509729168e-06, "loss": 6.2656, "step": 18000 }, { "epoch": 0.15, "learning_rate": 9.465468338307763e-06, "loss": 5.5998, "step": 18010 }, { "epoch": 0.15, "learning_rate": 9.464885403943028e-06, "loss": 6.1451, "step": 18020 }, { "epoch": 0.15, "learning_rate": 9.464302169861728e-06, "loss": 6.4675, "step": 18030 }, { "epoch": 0.15, "learning_rate": 9.463718636103012e-06, "loss": 7.1617, "step": 18040 }, { "epoch": 0.15, "learning_rate": 9.463134802706054e-06, "loss": 5.904, "step": 18050 }, { "epoch": 0.15, "learning_rate": 9.462550669710042e-06, "loss": 6.6504, "step": 18060 }, { "epoch": 0.15, "learning_rate": 9.46196623715419e-06, "loss": 6.2341, "step": 18070 }, { "epoch": 0.15, "learning_rate": 9.461381505077729e-06, "loss": 5.8894, "step": 18080 }, { "epoch": 0.15, "learning_rate": 9.460796473519908e-06, "loss": 6.1011, "step": 18090 }, { "epoch": 0.15, "learning_rate": 9.460211142520003e-06, "loss": 5.8591, "step": 18100 }, { "epoch": 0.15, "learning_rate": 9.459625512117302e-06, "loss": 6.411, "step": 18110 }, { "epoch": 0.15, "learning_rate": 9.459039582351122e-06, "loss": 6.206, "step": 18120 }, { "epoch": 0.15, "learning_rate": 9.458453353260789e-06, "loss": 6.2839, "step": 18130 }, { "epoch": 0.15, "learning_rate": 9.457866824885656e-06, "loss": 6.2451, "step": 18140 }, { "epoch": 0.15, "learning_rate": 9.457279997265099e-06, "loss": 6.6604, "step": 18150 }, { "epoch": 0.15, "learning_rate": 9.456692870438507e-06, "loss": 6.4123, "step": 18160 }, { "epoch": 0.15, "learning_rate": 9.456105444445294e-06, "loss": 6.6185, "step": 18170 }, { "epoch": 0.15, "learning_rate": 9.455517719324891e-06, "loss": 6.6717, "step": 18180 }, { "epoch": 0.15, "learning_rate": 9.454929695116752e-06, "loss": 5.8354, "step": 18190 }, { "epoch": 0.15, "learning_rate": 9.45434137186035e-06, "loss": 6.5452, "step": 18200 }, { "epoch": 0.15, "learning_rate": 9.453752749595176e-06, "loss": 6.3914, "step": 18210 }, { "epoch": 0.15, "learning_rate": 9.453163828360743e-06, "loss": 6.4117, "step": 18220 }, { "epoch": 0.15, "learning_rate": 9.452574608196584e-06, "loss": 6.1831, "step": 18230 }, { "epoch": 0.15, "learning_rate": 9.451985089142251e-06, "loss": 6.2835, "step": 18240 }, { "epoch": 0.15, "learning_rate": 9.451395271237321e-06, "loss": 5.7988, "step": 18250 }, { "epoch": 0.15, "learning_rate": 9.450805154521381e-06, "loss": 6.4303, "step": 18260 }, { "epoch": 0.15, "learning_rate": 9.450214739034049e-06, "loss": 6.4673, "step": 18270 }, { "epoch": 0.15, "learning_rate": 9.449624024814954e-06, "loss": 6.8202, "step": 18280 }, { "epoch": 0.15, "learning_rate": 9.449033011903753e-06, "loss": 6.099, "step": 18290 }, { "epoch": 0.15, "learning_rate": 9.448441700340117e-06, "loss": 5.9492, "step": 18300 }, { "epoch": 0.15, "learning_rate": 9.44785009016374e-06, "loss": 6.1029, "step": 18310 }, { "epoch": 0.15, "learning_rate": 9.447258181414335e-06, "loss": 5.7402, "step": 18320 }, { "epoch": 0.15, "learning_rate": 9.446665974131635e-06, "loss": 5.7932, "step": 18330 }, { "epoch": 0.15, "learning_rate": 9.446073468355395e-06, "loss": 5.5669, "step": 18340 }, { "epoch": 0.15, "learning_rate": 9.445480664125386e-06, "loss": 6.5584, "step": 18350 }, { "epoch": 0.15, "learning_rate": 9.444887561481403e-06, "loss": 6.6723, "step": 18360 }, { "epoch": 0.15, "learning_rate": 9.444294160463259e-06, "loss": 5.8518, "step": 18370 }, { "epoch": 0.15, "learning_rate": 9.443700461110788e-06, "loss": 6.2776, "step": 18380 }, { "epoch": 0.15, "learning_rate": 9.443106463463844e-06, "loss": 6.474, "step": 18390 }, { "epoch": 0.15, "learning_rate": 9.442512167562298e-06, "loss": 6.2478, "step": 18400 }, { "epoch": 0.15, "learning_rate": 9.441917573446047e-06, "loss": 6.2301, "step": 18410 }, { "epoch": 0.15, "learning_rate": 9.441322681155003e-06, "loss": 5.8905, "step": 18420 }, { "epoch": 0.15, "learning_rate": 9.4407274907291e-06, "loss": 6.6157, "step": 18430 }, { "epoch": 0.15, "learning_rate": 9.440132002208292e-06, "loss": 7.1674, "step": 18440 }, { "epoch": 0.15, "learning_rate": 9.43953621563255e-06, "loss": 6.0376, "step": 18450 }, { "epoch": 0.15, "learning_rate": 9.43894013104187e-06, "loss": 5.9779, "step": 18460 }, { "epoch": 0.15, "learning_rate": 9.438343748476266e-06, "loss": 6.0982, "step": 18470 }, { "epoch": 0.15, "learning_rate": 9.437747067975771e-06, "loss": 6.1039, "step": 18480 }, { "epoch": 0.15, "learning_rate": 9.437150089580438e-06, "loss": 6.2033, "step": 18490 }, { "epoch": 0.15, "learning_rate": 9.436552813330342e-06, "loss": 5.6813, "step": 18500 }, { "epoch": 0.15, "learning_rate": 9.435955239265575e-06, "loss": 6.1824, "step": 18510 }, { "epoch": 0.15, "learning_rate": 9.435357367426252e-06, "loss": 6.3195, "step": 18520 }, { "epoch": 0.15, "learning_rate": 9.434759197852506e-06, "loss": 6.3702, "step": 18530 }, { "epoch": 0.15, "learning_rate": 9.434160730584492e-06, "loss": 6.1147, "step": 18540 }, { "epoch": 0.15, "learning_rate": 9.43356196566238e-06, "loss": 6.2187, "step": 18550 }, { "epoch": 0.15, "learning_rate": 9.43296290312637e-06, "loss": 5.7554, "step": 18560 }, { "epoch": 0.15, "learning_rate": 9.43236354301667e-06, "loss": 6.2765, "step": 18570 }, { "epoch": 0.15, "learning_rate": 9.431763885373513e-06, "loss": 5.9244, "step": 18580 }, { "epoch": 0.15, "learning_rate": 9.431163930237157e-06, "loss": 6.4509, "step": 18590 }, { "epoch": 0.15, "learning_rate": 9.430563677647872e-06, "loss": 6.329, "step": 18600 }, { "epoch": 0.15, "learning_rate": 9.429963127645954e-06, "loss": 6.0746, "step": 18610 }, { "epoch": 0.15, "learning_rate": 9.429362280271714e-06, "loss": 6.5715, "step": 18620 }, { "epoch": 0.15, "learning_rate": 9.428761135565486e-06, "loss": 6.492, "step": 18630 }, { "epoch": 0.15, "learning_rate": 9.428159693567625e-06, "loss": 6.1052, "step": 18640 }, { "epoch": 0.15, "learning_rate": 9.427557954318503e-06, "loss": 5.9523, "step": 18650 }, { "epoch": 0.15, "learning_rate": 9.426955917858512e-06, "loss": 6.0625, "step": 18660 }, { "epoch": 0.15, "learning_rate": 9.426353584228068e-06, "loss": 6.4292, "step": 18670 }, { "epoch": 0.15, "learning_rate": 9.425750953467601e-06, "loss": 5.3806, "step": 18680 }, { "epoch": 0.15, "learning_rate": 9.425148025617568e-06, "loss": 6.2264, "step": 18690 }, { "epoch": 0.15, "learning_rate": 9.424544800718437e-06, "loss": 6.2262, "step": 18700 }, { "epoch": 0.15, "learning_rate": 9.423941278810705e-06, "loss": 5.996, "step": 18710 }, { "epoch": 0.15, "learning_rate": 9.423337459934885e-06, "loss": 5.8733, "step": 18720 }, { "epoch": 0.15, "learning_rate": 9.422733344131506e-06, "loss": 5.1343, "step": 18730 }, { "epoch": 0.15, "learning_rate": 9.422128931441126e-06, "loss": 6.3548, "step": 18740 }, { "epoch": 0.15, "learning_rate": 9.421524221904313e-06, "loss": 6.0361, "step": 18750 }, { "epoch": 0.15, "learning_rate": 9.42091921556166e-06, "loss": 6.1955, "step": 18760 }, { "epoch": 0.15, "learning_rate": 9.420313912453785e-06, "loss": 6.9867, "step": 18770 }, { "epoch": 0.15, "learning_rate": 9.419708312621317e-06, "loss": 6.3771, "step": 18780 }, { "epoch": 0.15, "learning_rate": 9.419102416104905e-06, "loss": 6.1763, "step": 18790 }, { "epoch": 0.16, "learning_rate": 9.418496222945227e-06, "loss": 6.0603, "step": 18800 }, { "epoch": 0.16, "learning_rate": 9.41788973318297e-06, "loss": 5.7174, "step": 18810 }, { "epoch": 0.16, "learning_rate": 9.41728294685885e-06, "loss": 5.9477, "step": 18820 }, { "epoch": 0.16, "learning_rate": 9.416675864013599e-06, "loss": 6.1167, "step": 18830 }, { "epoch": 0.16, "learning_rate": 9.416068484687965e-06, "loss": 6.049, "step": 18840 }, { "epoch": 0.16, "learning_rate": 9.415460808922725e-06, "loss": 6.156, "step": 18850 }, { "epoch": 0.16, "learning_rate": 9.414852836758667e-06, "loss": 6.8452, "step": 18860 }, { "epoch": 0.16, "learning_rate": 9.414244568236604e-06, "loss": 5.8758, "step": 18870 }, { "epoch": 0.16, "learning_rate": 9.413636003397369e-06, "loss": 6.5873, "step": 18880 }, { "epoch": 0.16, "learning_rate": 9.41302714228181e-06, "loss": 5.8926, "step": 18890 }, { "epoch": 0.16, "learning_rate": 9.412417984930802e-06, "loss": 5.5854, "step": 18900 }, { "epoch": 0.16, "learning_rate": 9.411808531385234e-06, "loss": 6.4783, "step": 18910 }, { "epoch": 0.16, "learning_rate": 9.411198781686016e-06, "loss": 6.899, "step": 18920 }, { "epoch": 0.16, "learning_rate": 9.410588735874083e-06, "loss": 5.6761, "step": 18930 }, { "epoch": 0.16, "learning_rate": 9.409978393990382e-06, "loss": 5.9307, "step": 18940 }, { "epoch": 0.16, "learning_rate": 9.409367756075884e-06, "loss": 5.779, "step": 18950 }, { "epoch": 0.16, "learning_rate": 9.408756822171583e-06, "loss": 6.2827, "step": 18960 }, { "epoch": 0.16, "learning_rate": 9.408145592318487e-06, "loss": 6.1148, "step": 18970 }, { "epoch": 0.16, "learning_rate": 9.407534066557625e-06, "loss": 6.026, "step": 18980 }, { "epoch": 0.16, "learning_rate": 9.406922244930051e-06, "loss": 6.9763, "step": 18990 }, { "epoch": 0.16, "learning_rate": 9.40631012747683e-06, "loss": 6.6704, "step": 19000 }, { "epoch": 0.16, "learning_rate": 9.405697714239056e-06, "loss": 7.141, "step": 19010 }, { "epoch": 0.16, "learning_rate": 9.405085005257836e-06, "loss": 6.2847, "step": 19020 }, { "epoch": 0.16, "learning_rate": 9.404472000574303e-06, "loss": 6.0164, "step": 19030 }, { "epoch": 0.16, "learning_rate": 9.403858700229605e-06, "loss": 6.0271, "step": 19040 }, { "epoch": 0.16, "learning_rate": 9.403245104264909e-06, "loss": 5.713, "step": 19050 }, { "epoch": 0.16, "learning_rate": 9.402631212721407e-06, "loss": 5.8123, "step": 19060 }, { "epoch": 0.16, "learning_rate": 9.402017025640306e-06, "loss": 5.9796, "step": 19070 }, { "epoch": 0.16, "learning_rate": 9.401402543062838e-06, "loss": 6.0602, "step": 19080 }, { "epoch": 0.16, "learning_rate": 9.400787765030247e-06, "loss": 6.2995, "step": 19090 }, { "epoch": 0.16, "learning_rate": 9.400172691583806e-06, "loss": 5.6984, "step": 19100 }, { "epoch": 0.16, "learning_rate": 9.3995573227648e-06, "loss": 6.1589, "step": 19110 }, { "epoch": 0.16, "learning_rate": 9.398941658614539e-06, "loss": 6.3778, "step": 19120 }, { "epoch": 0.16, "learning_rate": 9.39832569917435e-06, "loss": 5.9774, "step": 19130 }, { "epoch": 0.16, "learning_rate": 9.397709444485582e-06, "loss": 6.2054, "step": 19140 }, { "epoch": 0.16, "learning_rate": 9.397092894589602e-06, "loss": 6.4322, "step": 19150 }, { "epoch": 0.16, "learning_rate": 9.396476049527797e-06, "loss": 6.3988, "step": 19160 }, { "epoch": 0.16, "learning_rate": 9.395858909341575e-06, "loss": 5.8821, "step": 19170 }, { "epoch": 0.16, "learning_rate": 9.395241474072362e-06, "loss": 6.2851, "step": 19180 }, { "epoch": 0.16, "learning_rate": 9.394623743761606e-06, "loss": 6.0723, "step": 19190 }, { "epoch": 0.16, "learning_rate": 9.394005718450773e-06, "loss": 6.6229, "step": 19200 }, { "epoch": 0.16, "learning_rate": 9.39338739818135e-06, "loss": 5.8828, "step": 19210 }, { "epoch": 0.16, "learning_rate": 9.392768782994842e-06, "loss": 5.7854, "step": 19220 }, { "epoch": 0.16, "learning_rate": 9.392149872932779e-06, "loss": 6.1576, "step": 19230 }, { "epoch": 0.16, "learning_rate": 9.391530668036703e-06, "loss": 6.2918, "step": 19240 }, { "epoch": 0.16, "learning_rate": 9.39091116834818e-06, "loss": 6.5254, "step": 19250 }, { "epoch": 0.16, "learning_rate": 9.390291373908796e-06, "loss": 6.2291, "step": 19260 }, { "epoch": 0.16, "learning_rate": 9.389671284760158e-06, "loss": 6.4293, "step": 19270 }, { "epoch": 0.16, "learning_rate": 9.389050900943888e-06, "loss": 6.5261, "step": 19280 }, { "epoch": 0.16, "learning_rate": 9.388430222501632e-06, "loss": 6.0966, "step": 19290 }, { "epoch": 0.16, "learning_rate": 9.387809249475055e-06, "loss": 5.8488, "step": 19300 }, { "epoch": 0.16, "learning_rate": 9.387187981905842e-06, "loss": 5.8413, "step": 19310 }, { "epoch": 0.16, "learning_rate": 9.386566419835697e-06, "loss": 6.625, "step": 19320 }, { "epoch": 0.16, "learning_rate": 9.385944563306342e-06, "loss": 6.4747, "step": 19330 }, { "epoch": 0.16, "learning_rate": 9.385322412359522e-06, "loss": 6.0054, "step": 19340 }, { "epoch": 0.16, "learning_rate": 9.384699967037e-06, "loss": 5.8288, "step": 19350 }, { "epoch": 0.16, "learning_rate": 9.384077227380561e-06, "loss": 5.8726, "step": 19360 }, { "epoch": 0.16, "learning_rate": 9.383454193432005e-06, "loss": 5.8586, "step": 19370 }, { "epoch": 0.16, "learning_rate": 9.382830865233157e-06, "loss": 6.2528, "step": 19380 }, { "epoch": 0.16, "learning_rate": 9.382207242825857e-06, "loss": 6.8322, "step": 19390 }, { "epoch": 0.16, "learning_rate": 9.381583326251972e-06, "loss": 6.3257, "step": 19400 }, { "epoch": 0.16, "learning_rate": 9.380959115553378e-06, "loss": 6.7861, "step": 19410 }, { "epoch": 0.16, "learning_rate": 9.380334610771982e-06, "loss": 6.6071, "step": 19420 }, { "epoch": 0.16, "learning_rate": 9.379709811949702e-06, "loss": 6.4599, "step": 19430 }, { "epoch": 0.16, "learning_rate": 9.37908471912848e-06, "loss": 5.8492, "step": 19440 }, { "epoch": 0.16, "learning_rate": 9.378459332350276e-06, "loss": 6.042, "step": 19450 }, { "epoch": 0.16, "learning_rate": 9.377833651657074e-06, "loss": 6.0787, "step": 19460 }, { "epoch": 0.16, "learning_rate": 9.377207677090872e-06, "loss": 6.7646, "step": 19470 }, { "epoch": 0.16, "learning_rate": 9.376581408693687e-06, "loss": 6.8895, "step": 19480 }, { "epoch": 0.16, "learning_rate": 9.375954846507564e-06, "loss": 5.888, "step": 19490 }, { "epoch": 0.16, "learning_rate": 9.375327990574563e-06, "loss": 6.9242, "step": 19500 }, { "epoch": 0.16, "learning_rate": 9.374700840936756e-06, "loss": 6.2941, "step": 19510 }, { "epoch": 0.16, "learning_rate": 9.374073397636253e-06, "loss": 5.834, "step": 19520 }, { "epoch": 0.16, "learning_rate": 9.373445660715161e-06, "loss": 5.7813, "step": 19530 }, { "epoch": 0.16, "learning_rate": 9.372817630215626e-06, "loss": 6.2084, "step": 19540 }, { "epoch": 0.16, "learning_rate": 9.372189306179805e-06, "loss": 6.6227, "step": 19550 }, { "epoch": 0.16, "learning_rate": 9.371560688649872e-06, "loss": 5.8606, "step": 19560 }, { "epoch": 0.16, "learning_rate": 9.37093177766803e-06, "loss": 5.7067, "step": 19570 }, { "epoch": 0.16, "learning_rate": 9.370302573276493e-06, "loss": 6.0368, "step": 19580 }, { "epoch": 0.16, "learning_rate": 9.369673075517497e-06, "loss": 6.0146, "step": 19590 }, { "epoch": 0.16, "learning_rate": 9.369043284433303e-06, "loss": 5.6878, "step": 19600 }, { "epoch": 0.16, "learning_rate": 9.368413200066182e-06, "loss": 6.1409, "step": 19610 }, { "epoch": 0.16, "learning_rate": 9.367782822458434e-06, "loss": 6.4919, "step": 19620 }, { "epoch": 0.16, "learning_rate": 9.367152151652372e-06, "loss": 6.0632, "step": 19630 }, { "epoch": 0.16, "learning_rate": 9.366521187690333e-06, "loss": 5.9995, "step": 19640 }, { "epoch": 0.16, "learning_rate": 9.365889930614672e-06, "loss": 6.2916, "step": 19650 }, { "epoch": 0.16, "learning_rate": 9.365258380467762e-06, "loss": 6.0754, "step": 19660 }, { "epoch": 0.16, "learning_rate": 9.364626537291999e-06, "loss": 6.0058, "step": 19670 }, { "epoch": 0.16, "learning_rate": 9.363994401129797e-06, "loss": 6.3099, "step": 19680 }, { "epoch": 0.16, "learning_rate": 9.363361972023588e-06, "loss": 6.0217, "step": 19690 }, { "epoch": 0.16, "learning_rate": 9.362729250015827e-06, "loss": 5.9986, "step": 19700 }, { "epoch": 0.16, "learning_rate": 9.362096235148988e-06, "loss": 5.8779, "step": 19710 }, { "epoch": 0.16, "learning_rate": 9.36146292746556e-06, "loss": 6.2527, "step": 19720 }, { "epoch": 0.16, "learning_rate": 9.360829327008062e-06, "loss": 5.9587, "step": 19730 }, { "epoch": 0.16, "learning_rate": 9.360195433819019e-06, "loss": 5.6058, "step": 19740 }, { "epoch": 0.16, "learning_rate": 9.359561247940985e-06, "loss": 6.7259, "step": 19750 }, { "epoch": 0.16, "learning_rate": 9.358926769416532e-06, "loss": 5.7634, "step": 19760 }, { "epoch": 0.16, "learning_rate": 9.358291998288251e-06, "loss": 5.5594, "step": 19770 }, { "epoch": 0.16, "learning_rate": 9.357656934598754e-06, "loss": 5.9904, "step": 19780 }, { "epoch": 0.16, "learning_rate": 9.357021578390667e-06, "loss": 6.1408, "step": 19790 }, { "epoch": 0.16, "learning_rate": 9.356385929706644e-06, "loss": 6.2926, "step": 19800 }, { "epoch": 0.16, "learning_rate": 9.355749988589353e-06, "loss": 6.4899, "step": 19810 }, { "epoch": 0.16, "learning_rate": 9.355113755081482e-06, "loss": 5.9069, "step": 19820 }, { "epoch": 0.16, "learning_rate": 9.35447722922574e-06, "loss": 5.8312, "step": 19830 }, { "epoch": 0.16, "learning_rate": 9.353840411064858e-06, "loss": 6.0257, "step": 19840 }, { "epoch": 0.16, "learning_rate": 9.353203300641582e-06, "loss": 6.2781, "step": 19850 }, { "epoch": 0.16, "learning_rate": 9.35256589799868e-06, "loss": 5.9349, "step": 19860 }, { "epoch": 0.16, "learning_rate": 9.351928203178938e-06, "loss": 6.5633, "step": 19870 }, { "epoch": 0.16, "learning_rate": 9.351290216225166e-06, "loss": 6.5059, "step": 19880 }, { "epoch": 0.16, "learning_rate": 9.350651937180186e-06, "loss": 6.2228, "step": 19890 }, { "epoch": 0.16, "learning_rate": 9.350013366086848e-06, "loss": 6.4018, "step": 19900 }, { "epoch": 0.16, "learning_rate": 9.349374502988015e-06, "loss": 6.0398, "step": 19910 }, { "epoch": 0.16, "learning_rate": 9.348735347926574e-06, "loss": 6.8657, "step": 19920 }, { "epoch": 0.16, "learning_rate": 9.348095900945428e-06, "loss": 6.0936, "step": 19930 }, { "epoch": 0.16, "learning_rate": 9.347456162087504e-06, "loss": 5.9887, "step": 19940 }, { "epoch": 0.16, "learning_rate": 9.346816131395746e-06, "loss": 5.8403, "step": 19950 }, { "epoch": 0.16, "learning_rate": 9.346175808913114e-06, "loss": 5.8753, "step": 19960 }, { "epoch": 0.16, "learning_rate": 9.345535194682596e-06, "loss": 6.2367, "step": 19970 }, { "epoch": 0.16, "learning_rate": 9.344894288747192e-06, "loss": 6.1435, "step": 19980 }, { "epoch": 0.16, "learning_rate": 9.344253091149925e-06, "loss": 6.0118, "step": 19990 }, { "epoch": 0.16, "learning_rate": 9.343611601933835e-06, "loss": 6.2273, "step": 20000 }, { "epoch": 0.17, "learning_rate": 9.342969821141988e-06, "loss": 6.3326, "step": 20010 }, { "epoch": 0.17, "learning_rate": 9.342327748817462e-06, "loss": 5.7464, "step": 20020 }, { "epoch": 0.17, "learning_rate": 9.341685385003359e-06, "loss": 6.7596, "step": 20030 }, { "epoch": 0.17, "learning_rate": 9.341042729742797e-06, "loss": 6.5729, "step": 20040 }, { "epoch": 0.17, "learning_rate": 9.340399783078917e-06, "loss": 5.5704, "step": 20050 }, { "epoch": 0.17, "learning_rate": 9.339756545054881e-06, "loss": 6.3316, "step": 20060 }, { "epoch": 0.17, "learning_rate": 9.339113015713866e-06, "loss": 5.4413, "step": 20070 }, { "epoch": 0.17, "learning_rate": 9.338469195099068e-06, "loss": 6.6633, "step": 20080 }, { "epoch": 0.17, "learning_rate": 9.337825083253709e-06, "loss": 6.4725, "step": 20090 }, { "epoch": 0.17, "learning_rate": 9.337180680221024e-06, "loss": 6.2611, "step": 20100 }, { "epoch": 0.17, "learning_rate": 9.336535986044273e-06, "loss": 6.6629, "step": 20110 }, { "epoch": 0.17, "learning_rate": 9.33589100076673e-06, "loss": 5.8969, "step": 20120 }, { "epoch": 0.17, "learning_rate": 9.33524572443169e-06, "loss": 6.6024, "step": 20130 }, { "epoch": 0.17, "learning_rate": 9.334600157082474e-06, "loss": 6.6339, "step": 20140 }, { "epoch": 0.17, "learning_rate": 9.333954298762413e-06, "loss": 5.9669, "step": 20150 }, { "epoch": 0.17, "learning_rate": 9.333308149514862e-06, "loss": 6.1973, "step": 20160 }, { "epoch": 0.17, "learning_rate": 9.332661709383198e-06, "loss": 6.14, "step": 20170 }, { "epoch": 0.17, "learning_rate": 9.332014978410813e-06, "loss": 7.2527, "step": 20180 }, { "epoch": 0.17, "learning_rate": 9.331367956641121e-06, "loss": 6.192, "step": 20190 }, { "epoch": 0.17, "learning_rate": 9.330720644117555e-06, "loss": 6.4064, "step": 20200 }, { "epoch": 0.17, "learning_rate": 9.330073040883567e-06, "loss": 5.7625, "step": 20210 }, { "epoch": 0.17, "learning_rate": 9.329425146982626e-06, "loss": 5.8944, "step": 20220 }, { "epoch": 0.17, "learning_rate": 9.328776962458231e-06, "loss": 6.6667, "step": 20230 }, { "epoch": 0.17, "learning_rate": 9.328128487353887e-06, "loss": 6.3105, "step": 20240 }, { "epoch": 0.17, "learning_rate": 9.327479721713126e-06, "loss": 6.0903, "step": 20250 }, { "epoch": 0.17, "learning_rate": 9.3268306655795e-06, "loss": 6.0788, "step": 20260 }, { "epoch": 0.17, "learning_rate": 9.326181318996574e-06, "loss": 5.7799, "step": 20270 }, { "epoch": 0.17, "learning_rate": 9.325531682007943e-06, "loss": 6.6599, "step": 20280 }, { "epoch": 0.17, "learning_rate": 9.324881754657209e-06, "loss": 6.0525, "step": 20290 }, { "epoch": 0.17, "learning_rate": 9.324231536988005e-06, "loss": 6.3228, "step": 20300 }, { "epoch": 0.17, "learning_rate": 9.323581029043977e-06, "loss": 6.9823, "step": 20310 }, { "epoch": 0.17, "learning_rate": 9.32293023086879e-06, "loss": 6.4386, "step": 20320 }, { "epoch": 0.17, "learning_rate": 9.322279142506134e-06, "loss": 5.7906, "step": 20330 }, { "epoch": 0.17, "learning_rate": 9.321627763999711e-06, "loss": 5.7021, "step": 20340 }, { "epoch": 0.17, "learning_rate": 9.32097609539325e-06, "loss": 5.5198, "step": 20350 }, { "epoch": 0.17, "learning_rate": 9.320324136730495e-06, "loss": 5.7538, "step": 20360 }, { "epoch": 0.17, "learning_rate": 9.319671888055207e-06, "loss": 5.9602, "step": 20370 }, { "epoch": 0.17, "learning_rate": 9.319019349411174e-06, "loss": 5.6362, "step": 20380 }, { "epoch": 0.17, "learning_rate": 9.3183665208422e-06, "loss": 5.7609, "step": 20390 }, { "epoch": 0.17, "learning_rate": 9.317713402392102e-06, "loss": 6.89, "step": 20400 }, { "epoch": 0.17, "learning_rate": 9.317059994104727e-06, "loss": 6.2784, "step": 20410 }, { "epoch": 0.17, "learning_rate": 9.316406296023935e-06, "loss": 5.8542, "step": 20420 }, { "epoch": 0.17, "learning_rate": 9.315752308193609e-06, "loss": 5.9013, "step": 20430 }, { "epoch": 0.17, "learning_rate": 9.315098030657647e-06, "loss": 5.9656, "step": 20440 }, { "epoch": 0.17, "learning_rate": 9.31444346345997e-06, "loss": 6.422, "step": 20450 }, { "epoch": 0.17, "learning_rate": 9.313788606644518e-06, "loss": 5.9131, "step": 20460 }, { "epoch": 0.17, "learning_rate": 9.31313346025525e-06, "loss": 6.0115, "step": 20470 }, { "epoch": 0.17, "learning_rate": 9.312478024336142e-06, "loss": 6.381, "step": 20480 }, { "epoch": 0.17, "learning_rate": 9.311822298931195e-06, "loss": 6.0607, "step": 20490 }, { "epoch": 0.17, "learning_rate": 9.311166284084424e-06, "loss": 6.273, "step": 20500 }, { "epoch": 0.17, "learning_rate": 9.310509979839868e-06, "loss": 6.3661, "step": 20510 }, { "epoch": 0.17, "learning_rate": 9.309853386241578e-06, "loss": 5.7992, "step": 20520 }, { "epoch": 0.17, "learning_rate": 9.309196503333636e-06, "loss": 5.4779, "step": 20530 }, { "epoch": 0.17, "learning_rate": 9.308539331160134e-06, "loss": 6.3165, "step": 20540 }, { "epoch": 0.17, "learning_rate": 9.307881869765185e-06, "loss": 5.4488, "step": 20550 }, { "epoch": 0.17, "learning_rate": 9.307224119192923e-06, "loss": 6.6858, "step": 20560 }, { "epoch": 0.17, "learning_rate": 9.306566079487503e-06, "loss": 6.2886, "step": 20570 }, { "epoch": 0.17, "learning_rate": 9.305907750693098e-06, "loss": 5.8909, "step": 20580 }, { "epoch": 0.17, "learning_rate": 9.305249132853897e-06, "loss": 6.7175, "step": 20590 }, { "epoch": 0.17, "learning_rate": 9.304590226014111e-06, "loss": 5.6567, "step": 20600 }, { "epoch": 0.17, "learning_rate": 9.303931030217976e-06, "loss": 6.6367, "step": 20610 }, { "epoch": 0.17, "learning_rate": 9.303271545509738e-06, "loss": 6.2455, "step": 20620 }, { "epoch": 0.17, "learning_rate": 9.302611771933668e-06, "loss": 6.433, "step": 20630 }, { "epoch": 0.17, "learning_rate": 9.301951709534051e-06, "loss": 5.7255, "step": 20640 }, { "epoch": 0.17, "learning_rate": 9.301291358355202e-06, "loss": 5.8661, "step": 20650 }, { "epoch": 0.17, "learning_rate": 9.300630718441444e-06, "loss": 7.1444, "step": 20660 }, { "epoch": 0.17, "learning_rate": 9.299969789837126e-06, "loss": 6.8187, "step": 20670 }, { "epoch": 0.17, "learning_rate": 9.299308572586615e-06, "loss": 6.2042, "step": 20680 }, { "epoch": 0.17, "learning_rate": 9.298647066734295e-06, "loss": 7.0006, "step": 20690 }, { "epoch": 0.17, "learning_rate": 9.297985272324571e-06, "loss": 6.076, "step": 20700 }, { "epoch": 0.17, "learning_rate": 9.297323189401869e-06, "loss": 5.9383, "step": 20710 }, { "epoch": 0.17, "learning_rate": 9.296660818010633e-06, "loss": 5.8848, "step": 20720 }, { "epoch": 0.17, "learning_rate": 9.295998158195327e-06, "loss": 5.9503, "step": 20730 }, { "epoch": 0.17, "learning_rate": 9.295335210000431e-06, "loss": 5.9935, "step": 20740 }, { "epoch": 0.17, "learning_rate": 9.29467197347045e-06, "loss": 6.166, "step": 20750 }, { "epoch": 0.17, "learning_rate": 9.294008448649903e-06, "loss": 6.3066, "step": 20760 }, { "epoch": 0.17, "learning_rate": 9.293344635583332e-06, "loss": 6.1206, "step": 20770 }, { "epoch": 0.17, "learning_rate": 9.292680534315299e-06, "loss": 5.8039, "step": 20780 }, { "epoch": 0.17, "learning_rate": 9.292016144890378e-06, "loss": 5.9026, "step": 20790 }, { "epoch": 0.17, "learning_rate": 9.291351467353173e-06, "loss": 5.6863, "step": 20800 }, { "epoch": 0.17, "learning_rate": 9.2906865017483e-06, "loss": 6.3012, "step": 20810 }, { "epoch": 0.17, "learning_rate": 9.290021248120397e-06, "loss": 6.1853, "step": 20820 }, { "epoch": 0.17, "learning_rate": 9.289355706514123e-06, "loss": 6.1139, "step": 20830 }, { "epoch": 0.17, "learning_rate": 9.288689876974149e-06, "loss": 6.003, "step": 20840 }, { "epoch": 0.17, "learning_rate": 9.288023759545175e-06, "loss": 6.3195, "step": 20850 }, { "epoch": 0.17, "learning_rate": 9.287357354271912e-06, "loss": 5.5651, "step": 20860 }, { "epoch": 0.17, "learning_rate": 9.286690661199099e-06, "loss": 5.9906, "step": 20870 }, { "epoch": 0.17, "learning_rate": 9.286023680371483e-06, "loss": 6.253, "step": 20880 }, { "epoch": 0.17, "learning_rate": 9.285356411833842e-06, "loss": 5.6769, "step": 20890 }, { "epoch": 0.17, "learning_rate": 9.284688855630969e-06, "loss": 5.5239, "step": 20900 }, { "epoch": 0.17, "learning_rate": 9.284021011807668e-06, "loss": 5.777, "step": 20910 }, { "epoch": 0.17, "learning_rate": 9.283352880408776e-06, "loss": 6.1874, "step": 20920 }, { "epoch": 0.17, "learning_rate": 9.282684461479143e-06, "loss": 6.1376, "step": 20930 }, { "epoch": 0.17, "learning_rate": 9.282015755063636e-06, "loss": 6.2061, "step": 20940 }, { "epoch": 0.17, "learning_rate": 9.281346761207145e-06, "loss": 6.6216, "step": 20950 }, { "epoch": 0.17, "learning_rate": 9.280677479954577e-06, "loss": 6.0007, "step": 20960 }, { "epoch": 0.17, "learning_rate": 9.280007911350858e-06, "loss": 5.7717, "step": 20970 }, { "epoch": 0.17, "learning_rate": 9.279338055440936e-06, "loss": 5.9187, "step": 20980 }, { "epoch": 0.17, "learning_rate": 9.27866791226978e-06, "loss": 6.3223, "step": 20990 }, { "epoch": 0.17, "learning_rate": 9.277997481882368e-06, "loss": 5.9818, "step": 21000 }, { "epoch": 0.17, "learning_rate": 9.277326764323708e-06, "loss": 5.6991, "step": 21010 }, { "epoch": 0.17, "learning_rate": 9.276655759638824e-06, "loss": 6.5049, "step": 21020 }, { "epoch": 0.17, "learning_rate": 9.27598446787276e-06, "loss": 6.1748, "step": 21030 }, { "epoch": 0.17, "learning_rate": 9.275312889070576e-06, "loss": 6.3604, "step": 21040 }, { "epoch": 0.17, "learning_rate": 9.274641023277353e-06, "loss": 5.1777, "step": 21050 }, { "epoch": 0.17, "learning_rate": 9.273968870538193e-06, "loss": 6.2468, "step": 21060 }, { "epoch": 0.17, "learning_rate": 9.273296430898216e-06, "loss": 5.6939, "step": 21070 }, { "epoch": 0.17, "learning_rate": 9.27262370440256e-06, "loss": 6.0974, "step": 21080 }, { "epoch": 0.17, "learning_rate": 9.271950691096382e-06, "loss": 6.6625, "step": 21090 }, { "epoch": 0.17, "learning_rate": 9.271277391024867e-06, "loss": 5.852, "step": 21100 }, { "epoch": 0.17, "learning_rate": 9.270603804233203e-06, "loss": 6.6029, "step": 21110 }, { "epoch": 0.17, "learning_rate": 9.26992993076661e-06, "loss": 6.2744, "step": 21120 }, { "epoch": 0.17, "learning_rate": 9.269255770670325e-06, "loss": 6.4953, "step": 21130 }, { "epoch": 0.17, "learning_rate": 9.2685813239896e-06, "loss": 5.9079, "step": 21140 }, { "epoch": 0.17, "learning_rate": 9.26790659076971e-06, "loss": 6.4613, "step": 21150 }, { "epoch": 0.17, "learning_rate": 9.26723157105595e-06, "loss": 6.0113, "step": 21160 }, { "epoch": 0.17, "learning_rate": 9.266556264893628e-06, "loss": 5.9158, "step": 21170 }, { "epoch": 0.17, "learning_rate": 9.265880672328081e-06, "loss": 5.9776, "step": 21180 }, { "epoch": 0.17, "learning_rate": 9.265204793404654e-06, "loss": 6.1819, "step": 21190 }, { "epoch": 0.17, "learning_rate": 9.264528628168721e-06, "loss": 5.9035, "step": 21200 }, { "epoch": 0.17, "learning_rate": 9.263852176665672e-06, "loss": 6.385, "step": 21210 }, { "epoch": 0.18, "learning_rate": 9.263175438940913e-06, "loss": 6.231, "step": 21220 }, { "epoch": 0.18, "learning_rate": 9.26249841503987e-06, "loss": 6.2243, "step": 21230 }, { "epoch": 0.18, "learning_rate": 9.261821105007995e-06, "loss": 5.9389, "step": 21240 }, { "epoch": 0.18, "learning_rate": 9.261143508890752e-06, "loss": 6.0635, "step": 21250 }, { "epoch": 0.18, "learning_rate": 9.260465626733625e-06, "loss": 6.0586, "step": 21260 }, { "epoch": 0.18, "learning_rate": 9.259787458582118e-06, "loss": 6.0682, "step": 21270 }, { "epoch": 0.18, "learning_rate": 9.259109004481759e-06, "loss": 5.5996, "step": 21280 }, { "epoch": 0.18, "learning_rate": 9.258430264478087e-06, "loss": 6.2454, "step": 21290 }, { "epoch": 0.18, "learning_rate": 9.257751238616665e-06, "loss": 5.9725, "step": 21300 }, { "epoch": 0.18, "learning_rate": 9.257071926943073e-06, "loss": 6.6304, "step": 21310 }, { "epoch": 0.18, "learning_rate": 9.256392329502916e-06, "loss": 6.2687, "step": 21320 }, { "epoch": 0.18, "learning_rate": 9.255712446341809e-06, "loss": 6.3354, "step": 21330 }, { "epoch": 0.18, "learning_rate": 9.255032277505391e-06, "loss": 5.8786, "step": 21340 }, { "epoch": 0.18, "learning_rate": 9.254351823039324e-06, "loss": 6.365, "step": 21350 }, { "epoch": 0.18, "learning_rate": 9.253671082989281e-06, "loss": 5.8193, "step": 21360 }, { "epoch": 0.18, "learning_rate": 9.252990057400962e-06, "loss": 5.8417, "step": 21370 }, { "epoch": 0.18, "learning_rate": 9.25230874632008e-06, "loss": 5.4916, "step": 21380 }, { "epoch": 0.18, "learning_rate": 9.25162714979237e-06, "loss": 6.1608, "step": 21390 }, { "epoch": 0.18, "learning_rate": 9.250945267863586e-06, "loss": 5.8644, "step": 21400 }, { "epoch": 0.18, "learning_rate": 9.250263100579502e-06, "loss": 6.3254, "step": 21410 }, { "epoch": 0.18, "learning_rate": 9.24958064798591e-06, "loss": 5.9571, "step": 21420 }, { "epoch": 0.18, "learning_rate": 9.248897910128618e-06, "loss": 6.2476, "step": 21430 }, { "epoch": 0.18, "learning_rate": 9.248214887053462e-06, "loss": 6.1361, "step": 21440 }, { "epoch": 0.18, "learning_rate": 9.247531578806287e-06, "loss": 6.3226, "step": 21450 }, { "epoch": 0.18, "learning_rate": 9.246847985432963e-06, "loss": 5.8751, "step": 21460 }, { "epoch": 0.18, "learning_rate": 9.24616410697938e-06, "loss": 5.9978, "step": 21470 }, { "epoch": 0.18, "learning_rate": 9.245479943491444e-06, "loss": 6.3691, "step": 21480 }, { "epoch": 0.18, "learning_rate": 9.244795495015082e-06, "loss": 6.2703, "step": 21490 }, { "epoch": 0.18, "learning_rate": 9.244110761596236e-06, "loss": 6.0261, "step": 21500 }, { "epoch": 0.18, "learning_rate": 9.243425743280875e-06, "loss": 5.8357, "step": 21510 }, { "epoch": 0.18, "learning_rate": 9.242740440114977e-06, "loss": 6.271, "step": 21520 }, { "epoch": 0.18, "learning_rate": 9.242054852144549e-06, "loss": 5.632, "step": 21530 }, { "epoch": 0.18, "learning_rate": 9.24136897941561e-06, "loss": 6.742, "step": 21540 }, { "epoch": 0.18, "learning_rate": 9.240682821974205e-06, "loss": 5.8143, "step": 21550 }, { "epoch": 0.18, "learning_rate": 9.239996379866391e-06, "loss": 6.8346, "step": 21560 }, { "epoch": 0.18, "learning_rate": 9.239309653138247e-06, "loss": 6.4629, "step": 21570 }, { "epoch": 0.18, "learning_rate": 9.238622641835872e-06, "loss": 6.2634, "step": 21580 }, { "epoch": 0.18, "learning_rate": 9.237935346005384e-06, "loss": 6.3319, "step": 21590 }, { "epoch": 0.18, "learning_rate": 9.237247765692918e-06, "loss": 6.369, "step": 21600 }, { "epoch": 0.18, "learning_rate": 9.23655990094463e-06, "loss": 6.5128, "step": 21610 }, { "epoch": 0.18, "learning_rate": 9.235871751806696e-06, "loss": 6.5394, "step": 21620 }, { "epoch": 0.18, "learning_rate": 9.235183318325306e-06, "loss": 6.0872, "step": 21630 }, { "epoch": 0.18, "learning_rate": 9.234494600546678e-06, "loss": 5.9056, "step": 21640 }, { "epoch": 0.18, "learning_rate": 9.23380559851704e-06, "loss": 6.5651, "step": 21650 }, { "epoch": 0.18, "learning_rate": 9.233116312282647e-06, "loss": 6.3686, "step": 21660 }, { "epoch": 0.18, "learning_rate": 9.232426741889763e-06, "loss": 6.0019, "step": 21670 }, { "epoch": 0.18, "learning_rate": 9.23173688738468e-06, "loss": 6.7747, "step": 21680 }, { "epoch": 0.18, "learning_rate": 9.23104674881371e-06, "loss": 6.2979, "step": 21690 }, { "epoch": 0.18, "learning_rate": 9.230356326223176e-06, "loss": 6.3564, "step": 21700 }, { "epoch": 0.18, "learning_rate": 9.229665619659424e-06, "loss": 5.8545, "step": 21710 }, { "epoch": 0.18, "learning_rate": 9.22897462916882e-06, "loss": 5.6506, "step": 21720 }, { "epoch": 0.18, "learning_rate": 9.228283354797752e-06, "loss": 6.3974, "step": 21730 }, { "epoch": 0.18, "learning_rate": 9.227591796592618e-06, "loss": 6.1559, "step": 21740 }, { "epoch": 0.18, "learning_rate": 9.226899954599847e-06, "loss": 6.4802, "step": 21750 }, { "epoch": 0.18, "learning_rate": 9.226207828865873e-06, "loss": 5.9435, "step": 21760 }, { "epoch": 0.18, "learning_rate": 9.225515419437162e-06, "loss": 6.0788, "step": 21770 }, { "epoch": 0.18, "learning_rate": 9.224822726360192e-06, "loss": 6.3299, "step": 21780 }, { "epoch": 0.18, "learning_rate": 9.224129749681463e-06, "loss": 6.737, "step": 21790 }, { "epoch": 0.18, "learning_rate": 9.223436489447491e-06, "loss": 5.8221, "step": 21800 }, { "epoch": 0.18, "learning_rate": 9.222742945704812e-06, "loss": 6.553, "step": 21810 }, { "epoch": 0.18, "learning_rate": 9.222049118499985e-06, "loss": 6.3604, "step": 21820 }, { "epoch": 0.18, "learning_rate": 9.221355007879585e-06, "loss": 6.3899, "step": 21830 }, { "epoch": 0.18, "learning_rate": 9.220660613890202e-06, "loss": 6.9371, "step": 21840 }, { "epoch": 0.18, "learning_rate": 9.219965936578452e-06, "loss": 6.3136, "step": 21850 }, { "epoch": 0.18, "learning_rate": 9.219270975990965e-06, "loss": 6.0438, "step": 21860 }, { "epoch": 0.18, "learning_rate": 9.218575732174395e-06, "loss": 6.3048, "step": 21870 }, { "epoch": 0.18, "learning_rate": 9.217880205175409e-06, "loss": 6.2559, "step": 21880 }, { "epoch": 0.18, "learning_rate": 9.217184395040696e-06, "loss": 6.2365, "step": 21890 }, { "epoch": 0.18, "learning_rate": 9.216488301816966e-06, "loss": 6.0017, "step": 21900 }, { "epoch": 0.18, "learning_rate": 9.215791925550946e-06, "loss": 6.7826, "step": 21910 }, { "epoch": 0.18, "learning_rate": 9.21509526628938e-06, "loss": 5.8095, "step": 21920 }, { "epoch": 0.18, "learning_rate": 9.214398324079033e-06, "loss": 6.5045, "step": 21930 }, { "epoch": 0.18, "learning_rate": 9.213701098966692e-06, "loss": 5.4787, "step": 21940 }, { "epoch": 0.18, "learning_rate": 9.213003590999156e-06, "loss": 5.9376, "step": 21950 }, { "epoch": 0.18, "learning_rate": 9.21230580022325e-06, "loss": 5.9164, "step": 21960 }, { "epoch": 0.18, "learning_rate": 9.211607726685816e-06, "loss": 6.8194, "step": 21970 }, { "epoch": 0.18, "learning_rate": 9.210909370433708e-06, "loss": 5.9571, "step": 21980 }, { "epoch": 0.18, "learning_rate": 9.21021073151381e-06, "loss": 6.3713, "step": 21990 }, { "epoch": 0.18, "learning_rate": 9.20951180997302e-06, "loss": 5.7391, "step": 22000 }, { "epoch": 0.18, "learning_rate": 9.208812605858252e-06, "loss": 5.8281, "step": 22010 }, { "epoch": 0.18, "learning_rate": 9.208113119216445e-06, "loss": 5.7797, "step": 22020 }, { "epoch": 0.18, "learning_rate": 9.20741335009455e-06, "loss": 6.2241, "step": 22030 }, { "epoch": 0.18, "learning_rate": 9.206713298539542e-06, "loss": 5.8059, "step": 22040 }, { "epoch": 0.18, "learning_rate": 9.206012964598416e-06, "loss": 5.9748, "step": 22050 }, { "epoch": 0.18, "learning_rate": 9.205312348318182e-06, "loss": 5.9092, "step": 22060 }, { "epoch": 0.18, "learning_rate": 9.204611449745872e-06, "loss": 5.8941, "step": 22070 }, { "epoch": 0.18, "learning_rate": 9.203910268928533e-06, "loss": 6.355, "step": 22080 }, { "epoch": 0.18, "learning_rate": 9.203208805913235e-06, "loss": 5.9448, "step": 22090 }, { "epoch": 0.18, "learning_rate": 9.202507060747066e-06, "loss": 6.7445, "step": 22100 }, { "epoch": 0.18, "learning_rate": 9.20180503347713e-06, "loss": 6.1377, "step": 22110 }, { "epoch": 0.18, "learning_rate": 9.201102724150555e-06, "loss": 6.0992, "step": 22120 }, { "epoch": 0.18, "learning_rate": 9.200400132814484e-06, "loss": 6.4729, "step": 22130 }, { "epoch": 0.18, "learning_rate": 9.199697259516082e-06, "loss": 5.9001, "step": 22140 }, { "epoch": 0.18, "learning_rate": 9.198994104302527e-06, "loss": 6.2624, "step": 22150 }, { "epoch": 0.18, "learning_rate": 9.198290667221026e-06, "loss": 6.3072, "step": 22160 }, { "epoch": 0.18, "learning_rate": 9.197586948318792e-06, "loss": 6.3211, "step": 22170 }, { "epoch": 0.18, "learning_rate": 9.19688294764307e-06, "loss": 6.3028, "step": 22180 }, { "epoch": 0.18, "learning_rate": 9.196178665241115e-06, "loss": 6.4846, "step": 22190 }, { "epoch": 0.18, "learning_rate": 9.195474101160204e-06, "loss": 6.0823, "step": 22200 }, { "epoch": 0.18, "learning_rate": 9.194769255447631e-06, "loss": 5.6557, "step": 22210 }, { "epoch": 0.18, "learning_rate": 9.194064128150714e-06, "loss": 5.813, "step": 22220 }, { "epoch": 0.18, "learning_rate": 9.193358719316783e-06, "loss": 6.3142, "step": 22230 }, { "epoch": 0.18, "learning_rate": 9.192653028993193e-06, "loss": 6.5885, "step": 22240 }, { "epoch": 0.18, "learning_rate": 9.191947057227316e-06, "loss": 5.4047, "step": 22250 }, { "epoch": 0.18, "learning_rate": 9.191240804066537e-06, "loss": 6.4628, "step": 22260 }, { "epoch": 0.18, "learning_rate": 9.190534269558268e-06, "loss": 6.1739, "step": 22270 }, { "epoch": 0.18, "learning_rate": 9.18982745374994e-06, "loss": 6.5567, "step": 22280 }, { "epoch": 0.18, "learning_rate": 9.189120356688996e-06, "loss": 5.893, "step": 22290 }, { "epoch": 0.18, "learning_rate": 9.188412978422902e-06, "loss": 5.9359, "step": 22300 }, { "epoch": 0.18, "learning_rate": 9.187705318999142e-06, "loss": 5.9187, "step": 22310 }, { "epoch": 0.18, "learning_rate": 9.186997378465223e-06, "loss": 6.1872, "step": 22320 }, { "epoch": 0.18, "learning_rate": 9.186289156868664e-06, "loss": 5.7119, "step": 22330 }, { "epoch": 0.18, "learning_rate": 9.185580654257006e-06, "loss": 6.455, "step": 22340 }, { "epoch": 0.18, "learning_rate": 9.18487187067781e-06, "loss": 5.7933, "step": 22350 }, { "epoch": 0.18, "learning_rate": 9.184162806178656e-06, "loss": 6.5529, "step": 22360 }, { "epoch": 0.18, "learning_rate": 9.183453460807139e-06, "loss": 6.4143, "step": 22370 }, { "epoch": 0.18, "learning_rate": 9.182743834610879e-06, "loss": 5.5529, "step": 22380 }, { "epoch": 0.18, "learning_rate": 9.182033927637509e-06, "loss": 6.4674, "step": 22390 }, { "epoch": 0.18, "learning_rate": 9.181323739934682e-06, "loss": 5.7729, "step": 22400 }, { "epoch": 0.18, "learning_rate": 9.180613271550074e-06, "loss": 5.9974, "step": 22410 }, { "epoch": 0.18, "learning_rate": 9.179902522531376e-06, "loss": 5.5125, "step": 22420 }, { "epoch": 0.18, "learning_rate": 9.179191492926298e-06, "loss": 6.0524, "step": 22430 }, { "epoch": 0.19, "learning_rate": 9.178480182782572e-06, "loss": 5.5325, "step": 22440 }, { "epoch": 0.19, "learning_rate": 9.177768592147944e-06, "loss": 6.1146, "step": 22450 }, { "epoch": 0.19, "learning_rate": 9.177056721070182e-06, "loss": 6.5086, "step": 22460 }, { "epoch": 0.19, "learning_rate": 9.176344569597071e-06, "loss": 5.9288, "step": 22470 }, { "epoch": 0.19, "learning_rate": 9.175632137776419e-06, "loss": 6.2931, "step": 22480 }, { "epoch": 0.19, "learning_rate": 9.174919425656048e-06, "loss": 6.7465, "step": 22490 }, { "epoch": 0.19, "learning_rate": 9.1742064332838e-06, "loss": 5.7781, "step": 22500 }, { "epoch": 0.19, "learning_rate": 9.173493160707536e-06, "loss": 6.314, "step": 22510 }, { "epoch": 0.19, "learning_rate": 9.172779607975137e-06, "loss": 5.7898, "step": 22520 }, { "epoch": 0.19, "learning_rate": 9.172065775134503e-06, "loss": 5.9337, "step": 22530 }, { "epoch": 0.19, "learning_rate": 9.17135166223355e-06, "loss": 6.0079, "step": 22540 }, { "epoch": 0.19, "learning_rate": 9.170637269320217e-06, "loss": 6.1393, "step": 22550 }, { "epoch": 0.19, "learning_rate": 9.169922596442455e-06, "loss": 5.6974, "step": 22560 }, { "epoch": 0.19, "learning_rate": 9.169207643648244e-06, "loss": 6.1292, "step": 22570 }, { "epoch": 0.19, "learning_rate": 9.168492410985575e-06, "loss": 5.1907, "step": 22580 }, { "epoch": 0.19, "learning_rate": 9.167776898502457e-06, "loss": 6.1254, "step": 22590 }, { "epoch": 0.19, "learning_rate": 9.167061106246921e-06, "loss": 6.0553, "step": 22600 }, { "epoch": 0.19, "learning_rate": 9.16634503426702e-06, "loss": 7.0814, "step": 22610 }, { "epoch": 0.19, "learning_rate": 9.165628682610819e-06, "loss": 6.0114, "step": 22620 }, { "epoch": 0.19, "learning_rate": 9.164912051326404e-06, "loss": 5.9431, "step": 22630 }, { "epoch": 0.19, "learning_rate": 9.164195140461885e-06, "loss": 6.2301, "step": 22640 }, { "epoch": 0.19, "learning_rate": 9.163477950065384e-06, "loss": 6.5089, "step": 22650 }, { "epoch": 0.19, "learning_rate": 9.162760480185042e-06, "loss": 6.4454, "step": 22660 }, { "epoch": 0.19, "learning_rate": 9.162042730869026e-06, "loss": 5.7281, "step": 22670 }, { "epoch": 0.19, "learning_rate": 9.161324702165511e-06, "loss": 5.7764, "step": 22680 }, { "epoch": 0.19, "learning_rate": 9.1606063941227e-06, "loss": 6.5572, "step": 22690 }, { "epoch": 0.19, "learning_rate": 9.159887806788812e-06, "loss": 6.4843, "step": 22700 }, { "epoch": 0.19, "learning_rate": 9.15916894021208e-06, "loss": 5.611, "step": 22710 }, { "epoch": 0.19, "learning_rate": 9.158449794440764e-06, "loss": 6.7448, "step": 22720 }, { "epoch": 0.19, "learning_rate": 9.157730369523135e-06, "loss": 6.1341, "step": 22730 }, { "epoch": 0.19, "learning_rate": 9.15701066550749e-06, "loss": 5.9208, "step": 22740 }, { "epoch": 0.19, "learning_rate": 9.156290682442137e-06, "loss": 6.4079, "step": 22750 }, { "epoch": 0.19, "learning_rate": 9.155570420375407e-06, "loss": 6.0154, "step": 22760 }, { "epoch": 0.19, "learning_rate": 9.154849879355652e-06, "loss": 6.2184, "step": 22770 }, { "epoch": 0.19, "learning_rate": 9.154129059431239e-06, "loss": 6.0682, "step": 22780 }, { "epoch": 0.19, "learning_rate": 9.153407960650556e-06, "loss": 6.1247, "step": 22790 }, { "epoch": 0.19, "learning_rate": 9.152686583062005e-06, "loss": 5.8268, "step": 22800 }, { "epoch": 0.19, "learning_rate": 9.151964926714014e-06, "loss": 6.142, "step": 22810 }, { "epoch": 0.19, "learning_rate": 9.151242991655022e-06, "loss": 5.9229, "step": 22820 }, { "epoch": 0.19, "learning_rate": 9.150520777933497e-06, "loss": 5.2103, "step": 22830 }, { "epoch": 0.19, "learning_rate": 9.149798285597913e-06, "loss": 6.0981, "step": 22840 }, { "epoch": 0.19, "learning_rate": 9.149075514696772e-06, "loss": 5.8915, "step": 22850 }, { "epoch": 0.19, "learning_rate": 9.148352465278592e-06, "loss": 5.9787, "step": 22860 }, { "epoch": 0.19, "learning_rate": 9.147629137391906e-06, "loss": 6.4316, "step": 22870 }, { "epoch": 0.19, "learning_rate": 9.146905531085276e-06, "loss": 6.7204, "step": 22880 }, { "epoch": 0.19, "learning_rate": 9.14618164640727e-06, "loss": 6.4394, "step": 22890 }, { "epoch": 0.19, "learning_rate": 9.145457483406483e-06, "loss": 6.04, "step": 22900 }, { "epoch": 0.19, "learning_rate": 9.144733042131527e-06, "loss": 6.3108, "step": 22910 }, { "epoch": 0.19, "learning_rate": 9.14400832263103e-06, "loss": 6.4633, "step": 22920 }, { "epoch": 0.19, "learning_rate": 9.14328332495364e-06, "loss": 5.7987, "step": 22930 }, { "epoch": 0.19, "learning_rate": 9.142558049148028e-06, "loss": 5.9734, "step": 22940 }, { "epoch": 0.19, "learning_rate": 9.141832495262877e-06, "loss": 5.7832, "step": 22950 }, { "epoch": 0.19, "learning_rate": 9.141106663346891e-06, "loss": 5.9838, "step": 22960 }, { "epoch": 0.19, "learning_rate": 9.140380553448795e-06, "loss": 5.9285, "step": 22970 }, { "epoch": 0.19, "learning_rate": 9.13965416561733e-06, "loss": 6.0314, "step": 22980 }, { "epoch": 0.19, "learning_rate": 9.138927499901258e-06, "loss": 5.9712, "step": 22990 }, { "epoch": 0.19, "learning_rate": 9.138200556349358e-06, "loss": 6.2164, "step": 23000 }, { "epoch": 0.19, "learning_rate": 9.137473335010426e-06, "loss": 5.9542, "step": 23010 }, { "epoch": 0.19, "learning_rate": 9.136745835933278e-06, "loss": 6.5379, "step": 23020 }, { "epoch": 0.19, "learning_rate": 9.136018059166752e-06, "loss": 6.3683, "step": 23030 }, { "epoch": 0.19, "learning_rate": 9.135290004759701e-06, "loss": 6.527, "step": 23040 }, { "epoch": 0.19, "learning_rate": 9.134561672760998e-06, "loss": 6.1044, "step": 23050 }, { "epoch": 0.19, "learning_rate": 9.133833063219533e-06, "loss": 6.4031, "step": 23060 }, { "epoch": 0.19, "learning_rate": 9.133104176184215e-06, "loss": 6.0543, "step": 23070 }, { "epoch": 0.19, "learning_rate": 9.132375011703973e-06, "loss": 6.3553, "step": 23080 }, { "epoch": 0.19, "learning_rate": 9.131645569827757e-06, "loss": 5.6761, "step": 23090 }, { "epoch": 0.19, "learning_rate": 9.130915850604526e-06, "loss": 6.291, "step": 23100 }, { "epoch": 0.19, "learning_rate": 9.130185854083272e-06, "loss": 5.6516, "step": 23110 }, { "epoch": 0.19, "learning_rate": 9.129455580312992e-06, "loss": 5.8514, "step": 23120 }, { "epoch": 0.19, "learning_rate": 9.128725029342712e-06, "loss": 7.1484, "step": 23130 }, { "epoch": 0.19, "learning_rate": 9.127994201221465e-06, "loss": 5.7537, "step": 23140 }, { "epoch": 0.19, "learning_rate": 9.127263095998318e-06, "loss": 6.7515, "step": 23150 }, { "epoch": 0.19, "learning_rate": 9.126531713722341e-06, "loss": 5.4681, "step": 23160 }, { "epoch": 0.19, "learning_rate": 9.125800054442638e-06, "loss": 6.3764, "step": 23170 }, { "epoch": 0.19, "learning_rate": 9.125068118208316e-06, "loss": 5.6501, "step": 23180 }, { "epoch": 0.19, "learning_rate": 9.12433590506851e-06, "loss": 6.1375, "step": 23190 }, { "epoch": 0.19, "learning_rate": 9.123603415072373e-06, "loss": 6.1448, "step": 23200 }, { "epoch": 0.19, "learning_rate": 9.122870648269075e-06, "loss": 6.1104, "step": 23210 }, { "epoch": 0.19, "learning_rate": 9.122137604707805e-06, "loss": 5.6837, "step": 23220 }, { "epoch": 0.19, "learning_rate": 9.12140428443777e-06, "loss": 6.04, "step": 23230 }, { "epoch": 0.19, "learning_rate": 9.120670687508195e-06, "loss": 6.1244, "step": 23240 }, { "epoch": 0.19, "learning_rate": 9.119936813968324e-06, "loss": 5.7747, "step": 23250 }, { "epoch": 0.19, "learning_rate": 9.119202663867423e-06, "loss": 5.773, "step": 23260 }, { "epoch": 0.19, "learning_rate": 9.118468237254772e-06, "loss": 6.2728, "step": 23270 }, { "epoch": 0.19, "learning_rate": 9.11773353417967e-06, "loss": 5.3907, "step": 23280 }, { "epoch": 0.19, "learning_rate": 9.11699855469144e-06, "loss": 6.6242, "step": 23290 }, { "epoch": 0.19, "learning_rate": 9.116263298839414e-06, "loss": 6.0644, "step": 23300 }, { "epoch": 0.19, "learning_rate": 9.115527766672951e-06, "loss": 6.4422, "step": 23310 }, { "epoch": 0.19, "learning_rate": 9.114791958241424e-06, "loss": 6.1078, "step": 23320 }, { "epoch": 0.19, "learning_rate": 9.114055873594227e-06, "loss": 6.6922, "step": 23330 }, { "epoch": 0.19, "learning_rate": 9.113319512780771e-06, "loss": 5.7583, "step": 23340 }, { "epoch": 0.19, "learning_rate": 9.112582875850488e-06, "loss": 6.454, "step": 23350 }, { "epoch": 0.19, "learning_rate": 9.111845962852823e-06, "loss": 6.6392, "step": 23360 }, { "epoch": 0.19, "learning_rate": 9.111108773837246e-06, "loss": 6.1218, "step": 23370 }, { "epoch": 0.19, "learning_rate": 9.110371308853241e-06, "loss": 6.069, "step": 23380 }, { "epoch": 0.19, "learning_rate": 9.109633567950313e-06, "loss": 6.3569, "step": 23390 }, { "epoch": 0.19, "learning_rate": 9.108895551177983e-06, "loss": 5.8274, "step": 23400 }, { "epoch": 0.19, "learning_rate": 9.108157258585795e-06, "loss": 5.5101, "step": 23410 }, { "epoch": 0.19, "learning_rate": 9.107418690223308e-06, "loss": 5.5356, "step": 23420 }, { "epoch": 0.19, "learning_rate": 9.106679846140098e-06, "loss": 5.3822, "step": 23430 }, { "epoch": 0.19, "learning_rate": 9.105940726385765e-06, "loss": 6.1402, "step": 23440 }, { "epoch": 0.19, "learning_rate": 9.105201331009922e-06, "loss": 5.6062, "step": 23450 }, { "epoch": 0.19, "learning_rate": 9.104461660062202e-06, "loss": 5.2639, "step": 23460 }, { "epoch": 0.19, "learning_rate": 9.103721713592262e-06, "loss": 6.224, "step": 23470 }, { "epoch": 0.19, "learning_rate": 9.102981491649766e-06, "loss": 6.0308, "step": 23480 }, { "epoch": 0.19, "learning_rate": 9.102240994284411e-06, "loss": 5.8429, "step": 23490 }, { "epoch": 0.19, "learning_rate": 9.101500221545897e-06, "loss": 6.6583, "step": 23500 }, { "epoch": 0.19, "learning_rate": 9.100759173483952e-06, "loss": 5.6521, "step": 23510 }, { "epoch": 0.19, "learning_rate": 9.100017850148327e-06, "loss": 6.8877, "step": 23520 }, { "epoch": 0.19, "learning_rate": 9.099276251588777e-06, "loss": 6.2859, "step": 23530 }, { "epoch": 0.19, "learning_rate": 9.098534377855088e-06, "loss": 5.9479, "step": 23540 }, { "epoch": 0.19, "learning_rate": 9.097792228997059e-06, "loss": 5.424, "step": 23550 }, { "epoch": 0.19, "learning_rate": 9.097049805064511e-06, "loss": 6.5659, "step": 23560 }, { "epoch": 0.19, "learning_rate": 9.096307106107277e-06, "loss": 5.7207, "step": 23570 }, { "epoch": 0.19, "learning_rate": 9.095564132175214e-06, "loss": 5.9942, "step": 23580 }, { "epoch": 0.19, "learning_rate": 9.094820883318198e-06, "loss": 5.9544, "step": 23590 }, { "epoch": 0.19, "learning_rate": 9.09407735958612e-06, "loss": 6.0569, "step": 23600 }, { "epoch": 0.19, "learning_rate": 9.09333356102889e-06, "loss": 6.3161, "step": 23610 }, { "epoch": 0.19, "learning_rate": 9.092589487696438e-06, "loss": 6.3186, "step": 23620 }, { "epoch": 0.19, "learning_rate": 9.09184513963871e-06, "loss": 6.021, "step": 23630 }, { "epoch": 0.19, "learning_rate": 9.091100516905678e-06, "loss": 5.5556, "step": 23640 }, { "epoch": 0.2, "learning_rate": 9.09035561954732e-06, "loss": 5.6534, "step": 23650 }, { "epoch": 0.2, "learning_rate": 9.089610447613642e-06, "loss": 5.9104, "step": 23660 }, { "epoch": 0.2, "learning_rate": 9.088865001154666e-06, "loss": 6.4387, "step": 23670 }, { "epoch": 0.2, "learning_rate": 9.088119280220431e-06, "loss": 6.1835, "step": 23680 }, { "epoch": 0.2, "learning_rate": 9.087373284860997e-06, "loss": 6.5686, "step": 23690 }, { "epoch": 0.2, "learning_rate": 9.086627015126437e-06, "loss": 5.6617, "step": 23700 }, { "epoch": 0.2, "learning_rate": 9.085880471066851e-06, "loss": 6.1813, "step": 23710 }, { "epoch": 0.2, "learning_rate": 9.08513365273235e-06, "loss": 5.6463, "step": 23720 }, { "epoch": 0.2, "learning_rate": 9.084386560173063e-06, "loss": 5.1213, "step": 23730 }, { "epoch": 0.2, "learning_rate": 9.08363919343915e-06, "loss": 5.437, "step": 23740 }, { "epoch": 0.2, "learning_rate": 9.082891552580769e-06, "loss": 6.4527, "step": 23750 }, { "epoch": 0.2, "learning_rate": 9.082143637648114e-06, "loss": 5.8226, "step": 23760 }, { "epoch": 0.2, "learning_rate": 9.081395448691388e-06, "loss": 5.8692, "step": 23770 }, { "epoch": 0.2, "learning_rate": 9.080646985760815e-06, "loss": 5.6792, "step": 23780 }, { "epoch": 0.2, "learning_rate": 9.079898248906637e-06, "loss": 6.0535, "step": 23790 }, { "epoch": 0.2, "learning_rate": 9.079149238179119e-06, "loss": 6.0857, "step": 23800 }, { "epoch": 0.2, "learning_rate": 9.078399953628534e-06, "loss": 6.5403, "step": 23810 }, { "epoch": 0.2, "learning_rate": 9.077650395305184e-06, "loss": 5.8833, "step": 23820 }, { "epoch": 0.2, "learning_rate": 9.076900563259384e-06, "loss": 6.1417, "step": 23830 }, { "epoch": 0.2, "learning_rate": 9.076150457541468e-06, "loss": 6.2624, "step": 23840 }, { "epoch": 0.2, "learning_rate": 9.07540007820179e-06, "loss": 5.9619, "step": 23850 }, { "epoch": 0.2, "learning_rate": 9.074649425290718e-06, "loss": 6.9333, "step": 23860 }, { "epoch": 0.2, "learning_rate": 9.073898498858644e-06, "loss": 5.8128, "step": 23870 }, { "epoch": 0.2, "learning_rate": 9.073147298955974e-06, "loss": 6.3184, "step": 23880 }, { "epoch": 0.2, "learning_rate": 9.072395825633137e-06, "loss": 5.6081, "step": 23890 }, { "epoch": 0.2, "learning_rate": 9.071644078940574e-06, "loss": 6.7297, "step": 23900 }, { "epoch": 0.2, "learning_rate": 9.070892058928751e-06, "loss": 6.1404, "step": 23910 }, { "epoch": 0.2, "learning_rate": 9.070139765648148e-06, "loss": 6.026, "step": 23920 }, { "epoch": 0.2, "learning_rate": 9.069387199149265e-06, "loss": 6.0899, "step": 23930 }, { "epoch": 0.2, "learning_rate": 9.068634359482617e-06, "loss": 6.3358, "step": 23940 }, { "epoch": 0.2, "learning_rate": 9.067881246698744e-06, "loss": 5.5737, "step": 23950 }, { "epoch": 0.2, "learning_rate": 9.067127860848199e-06, "loss": 6.1716, "step": 23960 }, { "epoch": 0.2, "learning_rate": 9.066374201981555e-06, "loss": 6.148, "step": 23970 }, { "epoch": 0.2, "learning_rate": 9.065620270149401e-06, "loss": 5.4797, "step": 23980 }, { "epoch": 0.2, "learning_rate": 9.064866065402354e-06, "loss": 5.4559, "step": 23990 }, { "epoch": 0.2, "learning_rate": 9.064111587791033e-06, "loss": 6.1122, "step": 24000 }, { "epoch": 0.2, "learning_rate": 9.063356837366089e-06, "loss": 6.3451, "step": 24010 }, { "epoch": 0.2, "learning_rate": 9.062601814178184e-06, "loss": 5.7569, "step": 24020 }, { "epoch": 0.2, "learning_rate": 9.061846518278003e-06, "loss": 6.5554, "step": 24030 }, { "epoch": 0.2, "learning_rate": 9.061090949716247e-06, "loss": 6.2992, "step": 24040 }, { "epoch": 0.2, "learning_rate": 9.060335108543636e-06, "loss": 6.2578, "step": 24050 }, { "epoch": 0.2, "learning_rate": 9.059578994810905e-06, "loss": 5.9388, "step": 24060 }, { "epoch": 0.2, "learning_rate": 9.05882260856881e-06, "loss": 5.7203, "step": 24070 }, { "epoch": 0.2, "learning_rate": 9.058065949868129e-06, "loss": 5.9073, "step": 24080 }, { "epoch": 0.2, "learning_rate": 9.05730901875965e-06, "loss": 5.8509, "step": 24090 }, { "epoch": 0.2, "learning_rate": 9.05655181529419e-06, "loss": 5.5812, "step": 24100 }, { "epoch": 0.2, "learning_rate": 9.055794339522572e-06, "loss": 6.441, "step": 24110 }, { "epoch": 0.2, "learning_rate": 9.055036591495646e-06, "loss": 5.6772, "step": 24120 }, { "epoch": 0.2, "learning_rate": 9.054278571264279e-06, "loss": 5.4802, "step": 24130 }, { "epoch": 0.2, "learning_rate": 9.053520278879352e-06, "loss": 5.714, "step": 24140 }, { "epoch": 0.2, "learning_rate": 9.05276171439177e-06, "loss": 6.6651, "step": 24150 }, { "epoch": 0.2, "learning_rate": 9.052002877852452e-06, "loss": 5.7039, "step": 24160 }, { "epoch": 0.2, "learning_rate": 9.051243769312336e-06, "loss": 5.9676, "step": 24170 }, { "epoch": 0.2, "learning_rate": 9.050484388822382e-06, "loss": 6.285, "step": 24180 }, { "epoch": 0.2, "learning_rate": 9.049724736433565e-06, "loss": 5.9451, "step": 24190 }, { "epoch": 0.2, "learning_rate": 9.048964812196875e-06, "loss": 5.8472, "step": 24200 }, { "epoch": 0.2, "learning_rate": 9.048204616163325e-06, "loss": 6.1299, "step": 24210 }, { "epoch": 0.2, "learning_rate": 9.047444148383948e-06, "loss": 6.0987, "step": 24220 }, { "epoch": 0.2, "learning_rate": 9.046683408909788e-06, "loss": 6.3969, "step": 24230 }, { "epoch": 0.2, "learning_rate": 9.045922397791918e-06, "loss": 6.6177, "step": 24240 }, { "epoch": 0.2, "learning_rate": 9.045161115081416e-06, "loss": 6.1003, "step": 24250 }, { "epoch": 0.2, "learning_rate": 9.044399560829386e-06, "loss": 6.2482, "step": 24260 }, { "epoch": 0.2, "learning_rate": 9.043637735086954e-06, "loss": 5.8629, "step": 24270 }, { "epoch": 0.2, "learning_rate": 9.042875637905255e-06, "loss": 5.9511, "step": 24280 }, { "epoch": 0.2, "learning_rate": 9.042113269335449e-06, "loss": 5.8747, "step": 24290 }, { "epoch": 0.2, "learning_rate": 9.041350629428709e-06, "loss": 6.3251, "step": 24300 }, { "epoch": 0.2, "learning_rate": 9.040587718236232e-06, "loss": 5.9127, "step": 24310 }, { "epoch": 0.2, "learning_rate": 9.039824535809229e-06, "loss": 6.232, "step": 24320 }, { "epoch": 0.2, "learning_rate": 9.039061082198931e-06, "loss": 6.3542, "step": 24330 }, { "epoch": 0.2, "learning_rate": 9.038297357456587e-06, "loss": 6.1406, "step": 24340 }, { "epoch": 0.2, "learning_rate": 9.037533361633463e-06, "loss": 5.7295, "step": 24350 }, { "epoch": 0.2, "learning_rate": 9.036769094780844e-06, "loss": 5.8455, "step": 24360 }, { "epoch": 0.2, "learning_rate": 9.036004556950035e-06, "loss": 6.3165, "step": 24370 }, { "epoch": 0.2, "learning_rate": 9.035239748192355e-06, "loss": 6.2861, "step": 24380 }, { "epoch": 0.2, "learning_rate": 9.034474668559148e-06, "loss": 6.2367, "step": 24390 }, { "epoch": 0.2, "learning_rate": 9.033709318101767e-06, "loss": 5.8365, "step": 24400 }, { "epoch": 0.2, "learning_rate": 9.03294369687159e-06, "loss": 5.2213, "step": 24410 }, { "epoch": 0.2, "learning_rate": 9.03217780492001e-06, "loss": 6.4012, "step": 24420 }, { "epoch": 0.2, "learning_rate": 9.031411642298443e-06, "loss": 6.1671, "step": 24430 }, { "epoch": 0.2, "learning_rate": 9.030645209058317e-06, "loss": 5.8819, "step": 24440 }, { "epoch": 0.2, "learning_rate": 9.02987850525108e-06, "loss": 6.7307, "step": 24450 }, { "epoch": 0.2, "learning_rate": 9.029111530928201e-06, "loss": 6.1529, "step": 24460 }, { "epoch": 0.2, "learning_rate": 9.028344286141163e-06, "loss": 5.8512, "step": 24470 }, { "epoch": 0.2, "learning_rate": 9.027576770941471e-06, "loss": 5.6531, "step": 24480 }, { "epoch": 0.2, "learning_rate": 9.026808985380644e-06, "loss": 5.403, "step": 24490 }, { "epoch": 0.2, "learning_rate": 9.026040929510225e-06, "loss": 6.2772, "step": 24500 }, { "epoch": 0.2, "learning_rate": 9.025272603381768e-06, "loss": 5.5274, "step": 24510 }, { "epoch": 0.2, "learning_rate": 9.024504007046852e-06, "loss": 6.0714, "step": 24520 }, { "epoch": 0.2, "learning_rate": 9.02373514055707e-06, "loss": 5.7681, "step": 24530 }, { "epoch": 0.2, "learning_rate": 9.022966003964033e-06, "loss": 5.9957, "step": 24540 }, { "epoch": 0.2, "learning_rate": 9.02219659731937e-06, "loss": 5.9327, "step": 24550 }, { "epoch": 0.2, "learning_rate": 9.021426920674735e-06, "loss": 6.0693, "step": 24560 }, { "epoch": 0.2, "learning_rate": 9.020656974081787e-06, "loss": 6.405, "step": 24570 }, { "epoch": 0.2, "learning_rate": 9.019886757592216e-06, "loss": 5.9296, "step": 24580 }, { "epoch": 0.2, "learning_rate": 9.019116271257722e-06, "loss": 5.5674, "step": 24590 }, { "epoch": 0.2, "learning_rate": 9.018345515130027e-06, "loss": 6.0513, "step": 24600 }, { "epoch": 0.2, "learning_rate": 9.01757448926087e-06, "loss": 6.29, "step": 24610 }, { "epoch": 0.2, "learning_rate": 9.016803193702008e-06, "loss": 5.7475, "step": 24620 }, { "epoch": 0.2, "learning_rate": 9.016031628505214e-06, "loss": 5.8751, "step": 24630 }, { "epoch": 0.2, "learning_rate": 9.015259793722284e-06, "loss": 5.5854, "step": 24640 }, { "epoch": 0.2, "learning_rate": 9.01448768940503e-06, "loss": 6.0836, "step": 24650 }, { "epoch": 0.2, "learning_rate": 9.013715315605278e-06, "loss": 6.1369, "step": 24660 }, { "epoch": 0.2, "learning_rate": 9.012942672374878e-06, "loss": 6.3742, "step": 24670 }, { "epoch": 0.2, "learning_rate": 9.012169759765694e-06, "loss": 6.1905, "step": 24680 }, { "epoch": 0.2, "learning_rate": 9.011396577829612e-06, "loss": 5.856, "step": 24690 }, { "epoch": 0.2, "learning_rate": 9.010623126618531e-06, "loss": 6.0686, "step": 24700 }, { "epoch": 0.2, "learning_rate": 9.009849406184374e-06, "loss": 6.0561, "step": 24710 }, { "epoch": 0.2, "learning_rate": 9.009075416579075e-06, "loss": 7.2052, "step": 24720 }, { "epoch": 0.2, "learning_rate": 9.008301157854594e-06, "loss": 6.2691, "step": 24730 }, { "epoch": 0.2, "learning_rate": 9.007526630062902e-06, "loss": 5.9812, "step": 24740 }, { "epoch": 0.2, "learning_rate": 9.006751833255993e-06, "loss": 6.2109, "step": 24750 }, { "epoch": 0.2, "learning_rate": 9.005976767485877e-06, "loss": 6.2255, "step": 24760 }, { "epoch": 0.2, "learning_rate": 9.00520143280458e-06, "loss": 5.6291, "step": 24770 }, { "epoch": 0.2, "learning_rate": 9.00442582926415e-06, "loss": 5.5652, "step": 24780 }, { "epoch": 0.2, "learning_rate": 9.003649956916653e-06, "loss": 6.0265, "step": 24790 }, { "epoch": 0.2, "learning_rate": 9.002873815814168e-06, "loss": 6.171, "step": 24800 }, { "epoch": 0.2, "learning_rate": 9.002097406008799e-06, "loss": 6.1083, "step": 24810 }, { "epoch": 0.2, "learning_rate": 9.001320727552661e-06, "loss": 5.8505, "step": 24820 }, { "epoch": 0.2, "learning_rate": 9.000543780497892e-06, "loss": 6.1932, "step": 24830 }, { "epoch": 0.2, "learning_rate": 8.999766564896645e-06, "loss": 6.6487, "step": 24840 }, { "epoch": 0.2, "learning_rate": 8.998989080801097e-06, "loss": 5.8818, "step": 24850 }, { "epoch": 0.21, "learning_rate": 8.998211328263434e-06, "loss": 6.3259, "step": 24860 }, { "epoch": 0.21, "learning_rate": 8.997433307335867e-06, "loss": 6.3726, "step": 24870 }, { "epoch": 0.21, "learning_rate": 8.996655018070621e-06, "loss": 6.0502, "step": 24880 }, { "epoch": 0.21, "learning_rate": 8.99587646051994e-06, "loss": 5.7175, "step": 24890 }, { "epoch": 0.21, "learning_rate": 8.99509763473609e-06, "loss": 6.0258, "step": 24900 }, { "epoch": 0.21, "learning_rate": 8.99431854077135e-06, "loss": 5.6599, "step": 24910 }, { "epoch": 0.21, "learning_rate": 8.993539178678017e-06, "loss": 6.6631, "step": 24920 }, { "epoch": 0.21, "learning_rate": 8.992759548508409e-06, "loss": 6.0284, "step": 24930 }, { "epoch": 0.21, "learning_rate": 8.99197965031486e-06, "loss": 5.8616, "step": 24940 }, { "epoch": 0.21, "learning_rate": 8.991199484149725e-06, "loss": 6.1619, "step": 24950 }, { "epoch": 0.21, "learning_rate": 8.990419050065371e-06, "loss": 5.8619, "step": 24960 }, { "epoch": 0.21, "learning_rate": 8.989638348114187e-06, "loss": 5.8396, "step": 24970 }, { "epoch": 0.21, "learning_rate": 8.988857378348583e-06, "loss": 6.3226, "step": 24980 }, { "epoch": 0.21, "learning_rate": 8.98807614082098e-06, "loss": 5.5756, "step": 24990 }, { "epoch": 0.21, "learning_rate": 8.987294635583823e-06, "loss": 6.9664, "step": 25000 }, { "epoch": 0.21, "learning_rate": 8.986512862689571e-06, "loss": 5.8978, "step": 25010 }, { "epoch": 0.21, "learning_rate": 8.985730822190702e-06, "loss": 5.8688, "step": 25020 }, { "epoch": 0.21, "learning_rate": 8.984948514139714e-06, "loss": 6.4379, "step": 25030 }, { "epoch": 0.21, "learning_rate": 8.98416593858912e-06, "loss": 5.9329, "step": 25040 }, { "epoch": 0.21, "learning_rate": 8.983383095591454e-06, "loss": 6.4858, "step": 25050 }, { "epoch": 0.21, "learning_rate": 8.982599985199266e-06, "loss": 6.2924, "step": 25060 }, { "epoch": 0.21, "learning_rate": 8.981816607465121e-06, "loss": 6.0344, "step": 25070 }, { "epoch": 0.21, "learning_rate": 8.98103296244161e-06, "loss": 5.3008, "step": 25080 }, { "epoch": 0.21, "learning_rate": 8.980249050181332e-06, "loss": 5.9928, "step": 25090 }, { "epoch": 0.21, "learning_rate": 8.979464870736914e-06, "loss": 5.8741, "step": 25100 }, { "epoch": 0.21, "learning_rate": 8.978680424160992e-06, "loss": 6.0305, "step": 25110 }, { "epoch": 0.21, "learning_rate": 8.977895710506227e-06, "loss": 5.7128, "step": 25120 }, { "epoch": 0.21, "learning_rate": 8.977110729825295e-06, "loss": 5.8334, "step": 25130 }, { "epoch": 0.21, "learning_rate": 8.976325482170887e-06, "loss": 6.2167, "step": 25140 }, { "epoch": 0.21, "learning_rate": 8.975539967595716e-06, "loss": 6.0498, "step": 25150 }, { "epoch": 0.21, "learning_rate": 8.974754186152512e-06, "loss": 5.8735, "step": 25160 }, { "epoch": 0.21, "learning_rate": 8.973968137894023e-06, "loss": 6.4913, "step": 25170 }, { "epoch": 0.21, "learning_rate": 8.973181822873012e-06, "loss": 5.7263, "step": 25180 }, { "epoch": 0.21, "learning_rate": 8.972395241142266e-06, "loss": 6.0301, "step": 25190 }, { "epoch": 0.21, "learning_rate": 8.971608392754582e-06, "loss": 6.5942, "step": 25200 }, { "epoch": 0.21, "learning_rate": 8.970821277762784e-06, "loss": 5.936, "step": 25210 }, { "epoch": 0.21, "learning_rate": 8.970033896219706e-06, "loss": 6.3393, "step": 25220 }, { "epoch": 0.21, "learning_rate": 8.969246248178204e-06, "loss": 6.7427, "step": 25230 }, { "epoch": 0.21, "learning_rate": 8.968458333691148e-06, "loss": 5.9128, "step": 25240 }, { "epoch": 0.21, "learning_rate": 8.967670152811434e-06, "loss": 5.6228, "step": 25250 }, { "epoch": 0.21, "learning_rate": 8.966881705591966e-06, "loss": 6.1541, "step": 25260 }, { "epoch": 0.21, "learning_rate": 8.966092992085671e-06, "loss": 6.2378, "step": 25270 }, { "epoch": 0.21, "learning_rate": 8.965304012345496e-06, "loss": 6.2559, "step": 25280 }, { "epoch": 0.21, "learning_rate": 8.964514766424402e-06, "loss": 6.0823, "step": 25290 }, { "epoch": 0.21, "learning_rate": 8.963725254375368e-06, "loss": 6.0831, "step": 25300 }, { "epoch": 0.21, "learning_rate": 8.962935476251396e-06, "loss": 6.2425, "step": 25310 }, { "epoch": 0.21, "learning_rate": 8.962145432105494e-06, "loss": 6.3642, "step": 25320 }, { "epoch": 0.21, "learning_rate": 8.961355121990702e-06, "loss": 6.3109, "step": 25330 }, { "epoch": 0.21, "learning_rate": 8.96056454596007e-06, "loss": 6.4324, "step": 25340 }, { "epoch": 0.21, "learning_rate": 8.95977370406667e-06, "loss": 5.758, "step": 25350 }, { "epoch": 0.21, "learning_rate": 8.958982596363583e-06, "loss": 6.2059, "step": 25360 }, { "epoch": 0.21, "learning_rate": 8.958191222903919e-06, "loss": 6.6498, "step": 25370 }, { "epoch": 0.21, "learning_rate": 8.9573995837408e-06, "loss": 5.589, "step": 25380 }, { "epoch": 0.21, "learning_rate": 8.956607678927366e-06, "loss": 5.8853, "step": 25390 }, { "epoch": 0.21, "learning_rate": 8.955815508516775e-06, "loss": 6.069, "step": 25400 }, { "epoch": 0.21, "learning_rate": 8.955023072562203e-06, "loss": 6.0769, "step": 25410 }, { "epoch": 0.21, "learning_rate": 8.954230371116847e-06, "loss": 5.413, "step": 25420 }, { "epoch": 0.21, "learning_rate": 8.953437404233919e-06, "loss": 6.5655, "step": 25430 }, { "epoch": 0.21, "learning_rate": 8.952644171966645e-06, "loss": 6.5267, "step": 25440 }, { "epoch": 0.21, "learning_rate": 8.951850674368276e-06, "loss": 6.3105, "step": 25450 }, { "epoch": 0.21, "learning_rate": 8.951056911492078e-06, "loss": 5.805, "step": 25460 }, { "epoch": 0.21, "learning_rate": 8.950262883391332e-06, "loss": 5.014, "step": 25470 }, { "epoch": 0.21, "learning_rate": 8.949468590119339e-06, "loss": 5.5789, "step": 25480 }, { "epoch": 0.21, "learning_rate": 8.94867403172942e-06, "loss": 5.8552, "step": 25490 }, { "epoch": 0.21, "learning_rate": 8.94787920827491e-06, "loss": 5.5878, "step": 25500 }, { "epoch": 0.21, "learning_rate": 8.947084119809165e-06, "loss": 6.486, "step": 25510 }, { "epoch": 0.21, "learning_rate": 8.946288766385556e-06, "loss": 5.7768, "step": 25520 }, { "epoch": 0.21, "learning_rate": 8.945493148057477e-06, "loss": 6.1337, "step": 25530 }, { "epoch": 0.21, "learning_rate": 8.944697264878329e-06, "loss": 5.4814, "step": 25540 }, { "epoch": 0.21, "learning_rate": 8.943901116901538e-06, "loss": 6.1183, "step": 25550 }, { "epoch": 0.21, "learning_rate": 8.943104704180554e-06, "loss": 5.6401, "step": 25560 }, { "epoch": 0.21, "learning_rate": 8.942308026768836e-06, "loss": 5.8772, "step": 25570 }, { "epoch": 0.21, "learning_rate": 8.94151108471986e-06, "loss": 6.2572, "step": 25580 }, { "epoch": 0.21, "learning_rate": 8.940713878087124e-06, "loss": 6.0479, "step": 25590 }, { "epoch": 0.21, "learning_rate": 8.939916406924144e-06, "loss": 6.0678, "step": 25600 }, { "epoch": 0.21, "learning_rate": 8.93911867128445e-06, "loss": 7.0598, "step": 25610 }, { "epoch": 0.21, "learning_rate": 8.938320671221595e-06, "loss": 6.387, "step": 25620 }, { "epoch": 0.21, "learning_rate": 8.937522406789144e-06, "loss": 6.5202, "step": 25630 }, { "epoch": 0.21, "learning_rate": 8.93672387804068e-06, "loss": 5.6231, "step": 25640 }, { "epoch": 0.21, "learning_rate": 8.935925085029813e-06, "loss": 5.9028, "step": 25650 }, { "epoch": 0.21, "learning_rate": 8.93512602781016e-06, "loss": 6.1721, "step": 25660 }, { "epoch": 0.21, "learning_rate": 8.93432670643536e-06, "loss": 5.9942, "step": 25670 }, { "epoch": 0.21, "learning_rate": 8.93352712095907e-06, "loss": 5.982, "step": 25680 }, { "epoch": 0.21, "learning_rate": 8.932727271434965e-06, "loss": 6.4169, "step": 25690 }, { "epoch": 0.21, "learning_rate": 8.931927157916734e-06, "loss": 5.9326, "step": 25700 }, { "epoch": 0.21, "learning_rate": 8.931126780458088e-06, "loss": 6.2034, "step": 25710 }, { "epoch": 0.21, "learning_rate": 8.930326139112755e-06, "loss": 6.2826, "step": 25720 }, { "epoch": 0.21, "learning_rate": 8.929525233934479e-06, "loss": 6.4906, "step": 25730 }, { "epoch": 0.21, "learning_rate": 8.928724064977026e-06, "loss": 6.1382, "step": 25740 }, { "epoch": 0.21, "learning_rate": 8.927922632294172e-06, "loss": 5.9045, "step": 25750 }, { "epoch": 0.21, "learning_rate": 8.927120935939719e-06, "loss": 6.0596, "step": 25760 }, { "epoch": 0.21, "learning_rate": 8.926318975967478e-06, "loss": 6.458, "step": 25770 }, { "epoch": 0.21, "learning_rate": 8.925516752431286e-06, "loss": 6.0813, "step": 25780 }, { "epoch": 0.21, "learning_rate": 8.924714265384997e-06, "loss": 5.7709, "step": 25790 }, { "epoch": 0.21, "learning_rate": 8.923911514882475e-06, "loss": 6.0713, "step": 25800 }, { "epoch": 0.21, "learning_rate": 8.923108500977607e-06, "loss": 6.1969, "step": 25810 }, { "epoch": 0.21, "learning_rate": 8.922305223724301e-06, "loss": 6.6268, "step": 25820 }, { "epoch": 0.21, "learning_rate": 8.921501683176474e-06, "loss": 5.8642, "step": 25830 }, { "epoch": 0.21, "learning_rate": 8.920697879388072e-06, "loss": 6.333, "step": 25840 }, { "epoch": 0.21, "learning_rate": 8.919893812413045e-06, "loss": 6.3258, "step": 25850 }, { "epoch": 0.21, "learning_rate": 8.919089482305374e-06, "loss": 5.4248, "step": 25860 }, { "epoch": 0.21, "learning_rate": 8.918284889119048e-06, "loss": 5.6758, "step": 25870 }, { "epoch": 0.21, "learning_rate": 8.91748003290808e-06, "loss": 6.7266, "step": 25880 }, { "epoch": 0.21, "learning_rate": 8.916674913726495e-06, "loss": 5.7047, "step": 25890 }, { "epoch": 0.21, "learning_rate": 8.915869531628341e-06, "loss": 6.3579, "step": 25900 }, { "epoch": 0.21, "learning_rate": 8.915063886667681e-06, "loss": 5.6604, "step": 25910 }, { "epoch": 0.21, "learning_rate": 8.914257978898595e-06, "loss": 6.4062, "step": 25920 }, { "epoch": 0.21, "learning_rate": 8.913451808375181e-06, "loss": 5.8744, "step": 25930 }, { "epoch": 0.21, "learning_rate": 8.912645375151558e-06, "loss": 5.3737, "step": 25940 }, { "epoch": 0.21, "learning_rate": 8.911838679281859e-06, "loss": 6.1902, "step": 25950 }, { "epoch": 0.21, "learning_rate": 8.911031720820232e-06, "loss": 6.2459, "step": 25960 }, { "epoch": 0.21, "learning_rate": 8.91022449982085e-06, "loss": 5.0189, "step": 25970 }, { "epoch": 0.21, "learning_rate": 8.909417016337898e-06, "loss": 6.0813, "step": 25980 }, { "epoch": 0.21, "learning_rate": 8.908609270425581e-06, "loss": 6.2218, "step": 25990 }, { "epoch": 0.21, "learning_rate": 8.90780126213812e-06, "loss": 6.2107, "step": 26000 }, { "epoch": 0.21, "learning_rate": 8.906992991529757e-06, "loss": 6.4679, "step": 26010 }, { "epoch": 0.21, "learning_rate": 8.906184458654747e-06, "loss": 6.0975, "step": 26020 }, { "epoch": 0.21, "learning_rate": 8.905375663567364e-06, "loss": 6.2614, "step": 26030 }, { "epoch": 0.21, "learning_rate": 8.904566606321903e-06, "loss": 6.4026, "step": 26040 }, { "epoch": 0.21, "learning_rate": 8.90375728697267e-06, "loss": 5.9203, "step": 26050 }, { "epoch": 0.21, "learning_rate": 8.902947705573997e-06, "loss": 5.6168, "step": 26060 }, { "epoch": 0.22, "learning_rate": 8.902137862180228e-06, "loss": 5.6505, "step": 26070 }, { "epoch": 0.22, "learning_rate": 8.901327756845725e-06, "loss": 5.6371, "step": 26080 }, { "epoch": 0.22, "learning_rate": 8.900517389624868e-06, "loss": 5.8461, "step": 26090 }, { "epoch": 0.22, "learning_rate": 8.899706760572056e-06, "loss": 5.7265, "step": 26100 }, { "epoch": 0.22, "learning_rate": 8.8988958697417e-06, "loss": 5.8154, "step": 26110 }, { "epoch": 0.22, "learning_rate": 8.898084717188243e-06, "loss": 5.3438, "step": 26120 }, { "epoch": 0.22, "learning_rate": 8.897273302966124e-06, "loss": 6.1839, "step": 26130 }, { "epoch": 0.22, "learning_rate": 8.896461627129821e-06, "loss": 6.4748, "step": 26140 }, { "epoch": 0.22, "learning_rate": 8.895649689733814e-06, "loss": 6.5143, "step": 26150 }, { "epoch": 0.22, "learning_rate": 8.894837490832609e-06, "loss": 5.4248, "step": 26160 }, { "epoch": 0.22, "learning_rate": 8.894025030480724e-06, "loss": 5.8218, "step": 26170 }, { "epoch": 0.22, "learning_rate": 8.893212308732699e-06, "loss": 6.536, "step": 26180 }, { "epoch": 0.22, "learning_rate": 8.892399325643091e-06, "loss": 6.4618, "step": 26190 }, { "epoch": 0.22, "learning_rate": 8.891586081266472e-06, "loss": 5.955, "step": 26200 }, { "epoch": 0.22, "learning_rate": 8.890772575657435e-06, "loss": 5.643, "step": 26210 }, { "epoch": 0.22, "learning_rate": 8.889958808870587e-06, "loss": 5.9269, "step": 26220 }, { "epoch": 0.22, "learning_rate": 8.889144780960552e-06, "loss": 5.7336, "step": 26230 }, { "epoch": 0.22, "learning_rate": 8.88833049198198e-06, "loss": 5.5807, "step": 26240 }, { "epoch": 0.22, "learning_rate": 8.887515941989527e-06, "loss": 6.3115, "step": 26250 }, { "epoch": 0.22, "learning_rate": 8.88670113103787e-06, "loss": 5.3909, "step": 26260 }, { "epoch": 0.22, "learning_rate": 8.885886059181712e-06, "loss": 6.6804, "step": 26270 }, { "epoch": 0.22, "learning_rate": 8.885070726475761e-06, "loss": 6.0175, "step": 26280 }, { "epoch": 0.22, "learning_rate": 8.884255132974751e-06, "loss": 5.8419, "step": 26290 }, { "epoch": 0.22, "learning_rate": 8.883439278733429e-06, "loss": 5.7217, "step": 26300 }, { "epoch": 0.22, "learning_rate": 8.882623163806563e-06, "loss": 6.199, "step": 26310 }, { "epoch": 0.22, "learning_rate": 8.881806788248935e-06, "loss": 6.2925, "step": 26320 }, { "epoch": 0.22, "learning_rate": 8.880990152115348e-06, "loss": 6.095, "step": 26330 }, { "epoch": 0.22, "learning_rate": 8.880173255460618e-06, "loss": 6.0086, "step": 26340 }, { "epoch": 0.22, "learning_rate": 8.879356098339585e-06, "loss": 5.9708, "step": 26350 }, { "epoch": 0.22, "learning_rate": 8.8785386808071e-06, "loss": 6.2779, "step": 26360 }, { "epoch": 0.22, "learning_rate": 8.877721002918036e-06, "loss": 5.9538, "step": 26370 }, { "epoch": 0.22, "learning_rate": 8.876903064727281e-06, "loss": 6.4848, "step": 26380 }, { "epoch": 0.22, "learning_rate": 8.87608486628974e-06, "loss": 5.9976, "step": 26390 }, { "epoch": 0.22, "learning_rate": 8.875266407660337e-06, "loss": 5.7614, "step": 26400 }, { "epoch": 0.22, "learning_rate": 8.874447688894013e-06, "loss": 5.5174, "step": 26410 }, { "epoch": 0.22, "learning_rate": 8.873628710045728e-06, "loss": 5.9811, "step": 26420 }, { "epoch": 0.22, "learning_rate": 8.872809471170458e-06, "loss": 6.5441, "step": 26430 }, { "epoch": 0.22, "learning_rate": 8.871989972323193e-06, "loss": 5.5751, "step": 26440 }, { "epoch": 0.22, "learning_rate": 8.871170213558948e-06, "loss": 5.897, "step": 26450 }, { "epoch": 0.22, "learning_rate": 8.87035019493275e-06, "loss": 5.1083, "step": 26460 }, { "epoch": 0.22, "learning_rate": 8.869529916499644e-06, "loss": 6.3198, "step": 26470 }, { "epoch": 0.22, "learning_rate": 8.868709378314692e-06, "loss": 6.7616, "step": 26480 }, { "epoch": 0.22, "learning_rate": 8.86788858043298e-06, "loss": 5.5861, "step": 26490 }, { "epoch": 0.22, "learning_rate": 8.8670675229096e-06, "loss": 5.7065, "step": 26500 }, { "epoch": 0.22, "learning_rate": 8.86624620579967e-06, "loss": 5.7659, "step": 26510 }, { "epoch": 0.22, "learning_rate": 8.865424629158325e-06, "loss": 6.1715, "step": 26520 }, { "epoch": 0.22, "learning_rate": 8.864602793040713e-06, "loss": 5.7696, "step": 26530 }, { "epoch": 0.22, "learning_rate": 8.863780697502001e-06, "loss": 5.731, "step": 26540 }, { "epoch": 0.22, "learning_rate": 8.862958342597376e-06, "loss": 6.0413, "step": 26550 }, { "epoch": 0.22, "learning_rate": 8.86213572838204e-06, "loss": 5.9124, "step": 26560 }, { "epoch": 0.22, "learning_rate": 8.861312854911213e-06, "loss": 5.7193, "step": 26570 }, { "epoch": 0.22, "learning_rate": 8.860489722240133e-06, "loss": 6.1644, "step": 26580 }, { "epoch": 0.22, "learning_rate": 8.859666330424054e-06, "loss": 6.0668, "step": 26590 }, { "epoch": 0.22, "learning_rate": 8.858842679518249e-06, "loss": 5.8596, "step": 26600 }, { "epoch": 0.22, "learning_rate": 8.858018769578005e-06, "loss": 6.0984, "step": 26610 }, { "epoch": 0.22, "learning_rate": 8.857194600658635e-06, "loss": 6.3325, "step": 26620 }, { "epoch": 0.22, "learning_rate": 8.856370172815457e-06, "loss": 5.8296, "step": 26630 }, { "epoch": 0.22, "learning_rate": 8.855545486103816e-06, "loss": 5.8917, "step": 26640 }, { "epoch": 0.22, "learning_rate": 8.854720540579071e-06, "loss": 6.2683, "step": 26650 }, { "epoch": 0.22, "learning_rate": 8.853895336296596e-06, "loss": 5.6851, "step": 26660 }, { "epoch": 0.22, "learning_rate": 8.853069873311789e-06, "loss": 6.474, "step": 26670 }, { "epoch": 0.22, "learning_rate": 8.852244151680058e-06, "loss": 5.71, "step": 26680 }, { "epoch": 0.22, "learning_rate": 8.851418171456833e-06, "loss": 6.1659, "step": 26690 }, { "epoch": 0.22, "learning_rate": 8.850591932697562e-06, "loss": 5.2618, "step": 26700 }, { "epoch": 0.22, "learning_rate": 8.849765435457704e-06, "loss": 5.8806, "step": 26710 }, { "epoch": 0.22, "learning_rate": 8.848938679792743e-06, "loss": 5.8451, "step": 26720 }, { "epoch": 0.22, "learning_rate": 8.848111665758175e-06, "loss": 6.1379, "step": 26730 }, { "epoch": 0.22, "learning_rate": 8.847284393409516e-06, "loss": 6.1768, "step": 26740 }, { "epoch": 0.22, "learning_rate": 8.846456862802299e-06, "loss": 6.0801, "step": 26750 }, { "epoch": 0.22, "learning_rate": 8.845629073992075e-06, "loss": 5.8457, "step": 26760 }, { "epoch": 0.22, "learning_rate": 8.84480102703441e-06, "loss": 6.7029, "step": 26770 }, { "epoch": 0.22, "learning_rate": 8.84397272198489e-06, "loss": 6.4729, "step": 26780 }, { "epoch": 0.22, "learning_rate": 8.843144158899114e-06, "loss": 6.0583, "step": 26790 }, { "epoch": 0.22, "learning_rate": 8.842315337832705e-06, "loss": 5.149, "step": 26800 }, { "epoch": 0.22, "learning_rate": 8.841486258841298e-06, "loss": 5.5228, "step": 26810 }, { "epoch": 0.22, "learning_rate": 8.840656921980546e-06, "loss": 5.7403, "step": 26820 }, { "epoch": 0.22, "learning_rate": 8.839827327306123e-06, "loss": 5.7389, "step": 26830 }, { "epoch": 0.22, "learning_rate": 8.838997474873716e-06, "loss": 5.6576, "step": 26840 }, { "epoch": 0.22, "learning_rate": 8.838167364739032e-06, "loss": 6.5826, "step": 26850 }, { "epoch": 0.22, "learning_rate": 8.837336996957792e-06, "loss": 6.1797, "step": 26860 }, { "epoch": 0.22, "learning_rate": 8.836506371585738e-06, "loss": 6.1976, "step": 26870 }, { "epoch": 0.22, "learning_rate": 8.835675488678628e-06, "loss": 6.655, "step": 26880 }, { "epoch": 0.22, "learning_rate": 8.834844348292236e-06, "loss": 5.6178, "step": 26890 }, { "epoch": 0.22, "learning_rate": 8.834012950482355e-06, "loss": 5.862, "step": 26900 }, { "epoch": 0.22, "learning_rate": 8.833181295304795e-06, "loss": 5.5962, "step": 26910 }, { "epoch": 0.22, "learning_rate": 8.832349382815381e-06, "loss": 5.4294, "step": 26920 }, { "epoch": 0.22, "learning_rate": 8.83151721306996e-06, "loss": 5.6207, "step": 26930 }, { "epoch": 0.22, "learning_rate": 8.830684786124393e-06, "loss": 5.7894, "step": 26940 }, { "epoch": 0.22, "learning_rate": 8.829852102034558e-06, "loss": 5.6179, "step": 26950 }, { "epoch": 0.22, "learning_rate": 8.829019160856351e-06, "loss": 6.3895, "step": 26960 }, { "epoch": 0.22, "learning_rate": 8.828185962645684e-06, "loss": 5.6703, "step": 26970 }, { "epoch": 0.22, "learning_rate": 8.82735250745849e-06, "loss": 5.5463, "step": 26980 }, { "epoch": 0.22, "learning_rate": 8.826518795350715e-06, "loss": 5.827, "step": 26990 }, { "epoch": 0.22, "learning_rate": 8.825684826378325e-06, "loss": 6.1257, "step": 27000 }, { "epoch": 0.22, "learning_rate": 8.824850600597301e-06, "loss": 5.4886, "step": 27010 }, { "epoch": 0.22, "learning_rate": 8.824016118063643e-06, "loss": 6.0499, "step": 27020 }, { "epoch": 0.22, "learning_rate": 8.82318137883337e-06, "loss": 5.8143, "step": 27030 }, { "epoch": 0.22, "learning_rate": 8.822346382962512e-06, "loss": 5.9813, "step": 27040 }, { "epoch": 0.22, "learning_rate": 8.821511130507122e-06, "loss": 5.78, "step": 27050 }, { "epoch": 0.22, "learning_rate": 8.82067562152327e-06, "loss": 6.3056, "step": 27060 }, { "epoch": 0.22, "learning_rate": 8.819839856067038e-06, "loss": 6.4306, "step": 27070 }, { "epoch": 0.22, "learning_rate": 8.819003834194533e-06, "loss": 5.9354, "step": 27080 }, { "epoch": 0.22, "learning_rate": 8.818167555961872e-06, "loss": 6.2549, "step": 27090 }, { "epoch": 0.22, "learning_rate": 8.817331021425194e-06, "loss": 6.5248, "step": 27100 }, { "epoch": 0.22, "learning_rate": 8.816494230640652e-06, "loss": 6.1148, "step": 27110 }, { "epoch": 0.22, "learning_rate": 8.815657183664419e-06, "loss": 5.3404, "step": 27120 }, { "epoch": 0.22, "learning_rate": 8.814819880552683e-06, "loss": 5.7878, "step": 27130 }, { "epoch": 0.22, "learning_rate": 8.81398232136165e-06, "loss": 5.5469, "step": 27140 }, { "epoch": 0.22, "learning_rate": 8.813144506147544e-06, "loss": 7.1228, "step": 27150 }, { "epoch": 0.22, "learning_rate": 8.812306434966606e-06, "loss": 5.8994, "step": 27160 }, { "epoch": 0.22, "learning_rate": 8.811468107875092e-06, "loss": 5.9253, "step": 27170 }, { "epoch": 0.22, "learning_rate": 8.810629524929278e-06, "loss": 6.2172, "step": 27180 }, { "epoch": 0.22, "learning_rate": 8.809790686185454e-06, "loss": 5.4202, "step": 27190 }, { "epoch": 0.22, "learning_rate": 8.80895159169993e-06, "loss": 5.8664, "step": 27200 }, { "epoch": 0.22, "learning_rate": 8.808112241529035e-06, "loss": 6.4053, "step": 27210 }, { "epoch": 0.22, "learning_rate": 8.80727263572911e-06, "loss": 6.4077, "step": 27220 }, { "epoch": 0.22, "learning_rate": 8.806432774356514e-06, "loss": 6.1464, "step": 27230 }, { "epoch": 0.22, "learning_rate": 8.805592657467628e-06, "loss": 6.4341, "step": 27240 }, { "epoch": 0.22, "learning_rate": 8.804752285118847e-06, "loss": 5.6697, "step": 27250 }, { "epoch": 0.22, "learning_rate": 8.803911657366579e-06, "loss": 5.7591, "step": 27260 }, { "epoch": 0.22, "learning_rate": 8.803070774267259e-06, "loss": 5.7971, "step": 27270 }, { "epoch": 0.22, "learning_rate": 8.802229635877328e-06, "loss": 6.2265, "step": 27280 }, { "epoch": 0.23, "learning_rate": 8.801388242253253e-06, "loss": 5.5332, "step": 27290 }, { "epoch": 0.23, "learning_rate": 8.800546593451514e-06, "loss": 5.7818, "step": 27300 }, { "epoch": 0.23, "learning_rate": 8.799704689528608e-06, "loss": 5.8859, "step": 27310 }, { "epoch": 0.23, "learning_rate": 8.79886253054105e-06, "loss": 6.0762, "step": 27320 }, { "epoch": 0.23, "learning_rate": 8.798020116545371e-06, "loss": 6.3453, "step": 27330 }, { "epoch": 0.23, "learning_rate": 8.797177447598125e-06, "loss": 5.6239, "step": 27340 }, { "epoch": 0.23, "learning_rate": 8.796334523755871e-06, "loss": 6.5767, "step": 27350 }, { "epoch": 0.23, "learning_rate": 8.795491345075198e-06, "loss": 5.8273, "step": 27360 }, { "epoch": 0.23, "learning_rate": 8.794647911612703e-06, "loss": 6.1261, "step": 27370 }, { "epoch": 0.23, "learning_rate": 8.793804223425006e-06, "loss": 6.9533, "step": 27380 }, { "epoch": 0.23, "learning_rate": 8.792960280568741e-06, "loss": 6.0218, "step": 27390 }, { "epoch": 0.23, "learning_rate": 8.79211608310056e-06, "loss": 6.2324, "step": 27400 }, { "epoch": 0.23, "learning_rate": 8.791271631077131e-06, "loss": 5.6643, "step": 27410 }, { "epoch": 0.23, "learning_rate": 8.79042692455514e-06, "loss": 5.2897, "step": 27420 }, { "epoch": 0.23, "learning_rate": 8.789581963591289e-06, "loss": 5.9617, "step": 27430 }, { "epoch": 0.23, "learning_rate": 8.788736748242302e-06, "loss": 5.9643, "step": 27440 }, { "epoch": 0.23, "learning_rate": 8.787891278564911e-06, "loss": 6.1385, "step": 27450 }, { "epoch": 0.23, "learning_rate": 8.787045554615875e-06, "loss": 6.1483, "step": 27460 }, { "epoch": 0.23, "learning_rate": 8.786199576451963e-06, "loss": 5.8961, "step": 27470 }, { "epoch": 0.23, "learning_rate": 8.785353344129963e-06, "loss": 6.0032, "step": 27480 }, { "epoch": 0.23, "learning_rate": 8.78450685770668e-06, "loss": 6.1888, "step": 27490 }, { "epoch": 0.23, "learning_rate": 8.783660117238939e-06, "loss": 6.0801, "step": 27500 }, { "epoch": 0.23, "learning_rate": 8.782813122783579e-06, "loss": 5.9844, "step": 27510 }, { "epoch": 0.23, "learning_rate": 8.781965874397455e-06, "loss": 5.7378, "step": 27520 }, { "epoch": 0.23, "learning_rate": 8.781118372137441e-06, "loss": 6.1635, "step": 27530 }, { "epoch": 0.23, "learning_rate": 8.780270616060428e-06, "loss": 6.3016, "step": 27540 }, { "epoch": 0.23, "learning_rate": 8.779422606223325e-06, "loss": 5.4822, "step": 27550 }, { "epoch": 0.23, "learning_rate": 8.778574342683055e-06, "loss": 6.2223, "step": 27560 }, { "epoch": 0.23, "learning_rate": 8.77772582549656e-06, "loss": 5.563, "step": 27570 }, { "epoch": 0.23, "learning_rate": 8.7768770547208e-06, "loss": 6.079, "step": 27580 }, { "epoch": 0.23, "learning_rate": 8.776028030412749e-06, "loss": 6.0672, "step": 27590 }, { "epoch": 0.23, "learning_rate": 8.775178752629403e-06, "loss": 6.2394, "step": 27600 }, { "epoch": 0.23, "learning_rate": 8.774329221427768e-06, "loss": 5.1184, "step": 27610 }, { "epoch": 0.23, "learning_rate": 8.773479436864873e-06, "loss": 5.8571, "step": 27620 }, { "epoch": 0.23, "learning_rate": 8.772629398997762e-06, "loss": 5.717, "step": 27630 }, { "epoch": 0.23, "learning_rate": 8.771779107883495e-06, "loss": 5.8789, "step": 27640 }, { "epoch": 0.23, "learning_rate": 8.770928563579153e-06, "loss": 6.1137, "step": 27650 }, { "epoch": 0.23, "learning_rate": 8.770077766141827e-06, "loss": 5.7531, "step": 27660 }, { "epoch": 0.23, "learning_rate": 8.76922671562863e-06, "loss": 6.0683, "step": 27670 }, { "epoch": 0.23, "learning_rate": 8.76837541209669e-06, "loss": 6.3452, "step": 27680 }, { "epoch": 0.23, "learning_rate": 8.767523855603155e-06, "loss": 5.8728, "step": 27690 }, { "epoch": 0.23, "learning_rate": 8.766672046205189e-06, "loss": 6.2414, "step": 27700 }, { "epoch": 0.23, "learning_rate": 8.765819983959967e-06, "loss": 6.1733, "step": 27710 }, { "epoch": 0.23, "learning_rate": 8.764967668924691e-06, "loss": 6.2257, "step": 27720 }, { "epoch": 0.23, "learning_rate": 8.764115101156572e-06, "loss": 6.0032, "step": 27730 }, { "epoch": 0.23, "learning_rate": 8.763262280712838e-06, "loss": 6.5907, "step": 27740 }, { "epoch": 0.23, "learning_rate": 8.762409207650744e-06, "loss": 6.2074, "step": 27750 }, { "epoch": 0.23, "learning_rate": 8.761555882027549e-06, "loss": 5.6272, "step": 27760 }, { "epoch": 0.23, "learning_rate": 8.760702303900535e-06, "loss": 5.2447, "step": 27770 }, { "epoch": 0.23, "learning_rate": 8.759848473327002e-06, "loss": 5.9997, "step": 27780 }, { "epoch": 0.23, "learning_rate": 8.758994390364265e-06, "loss": 5.6028, "step": 27790 }, { "epoch": 0.23, "learning_rate": 8.758140055069657e-06, "loss": 6.1832, "step": 27800 }, { "epoch": 0.23, "learning_rate": 8.757285467500527e-06, "loss": 5.863, "step": 27810 }, { "epoch": 0.23, "learning_rate": 8.75643062771424e-06, "loss": 6.5489, "step": 27820 }, { "epoch": 0.23, "learning_rate": 8.755575535768183e-06, "loss": 6.1964, "step": 27830 }, { "epoch": 0.23, "learning_rate": 8.754720191719752e-06, "loss": 5.6463, "step": 27840 }, { "epoch": 0.23, "learning_rate": 8.753864595626366e-06, "loss": 6.2522, "step": 27850 }, { "epoch": 0.23, "learning_rate": 8.753008747545459e-06, "loss": 5.8049, "step": 27860 }, { "epoch": 0.23, "learning_rate": 8.752152647534483e-06, "loss": 5.5881, "step": 27870 }, { "epoch": 0.23, "learning_rate": 8.751296295650904e-06, "loss": 5.4482, "step": 27880 }, { "epoch": 0.23, "learning_rate": 8.750439691952206e-06, "loss": 6.177, "step": 27890 }, { "epoch": 0.23, "learning_rate": 8.749582836495894e-06, "loss": 5.2685, "step": 27900 }, { "epoch": 0.23, "learning_rate": 8.748725729339483e-06, "loss": 5.668, "step": 27910 }, { "epoch": 0.23, "learning_rate": 8.747868370540511e-06, "loss": 6.2312, "step": 27920 }, { "epoch": 0.23, "learning_rate": 8.747010760156529e-06, "loss": 6.408, "step": 27930 }, { "epoch": 0.23, "learning_rate": 8.746152898245106e-06, "loss": 5.433, "step": 27940 }, { "epoch": 0.23, "learning_rate": 8.745294784863828e-06, "loss": 6.0229, "step": 27950 }, { "epoch": 0.23, "learning_rate": 8.744436420070302e-06, "loss": 5.8175, "step": 27960 }, { "epoch": 0.23, "learning_rate": 8.743577803922142e-06, "loss": 5.9663, "step": 27970 }, { "epoch": 0.23, "learning_rate": 8.742718936476988e-06, "loss": 6.1724, "step": 27980 }, { "epoch": 0.23, "learning_rate": 8.741859817792494e-06, "loss": 6.516, "step": 27990 }, { "epoch": 0.23, "learning_rate": 8.741000447926328e-06, "loss": 6.2506, "step": 28000 }, { "epoch": 0.23, "learning_rate": 8.740140826936182e-06, "loss": 6.4293, "step": 28010 }, { "epoch": 0.23, "learning_rate": 8.739280954879752e-06, "loss": 6.0745, "step": 28020 }, { "epoch": 0.23, "learning_rate": 8.738420831814766e-06, "loss": 5.595, "step": 28030 }, { "epoch": 0.23, "learning_rate": 8.737560457798961e-06, "loss": 6.5117, "step": 28040 }, { "epoch": 0.23, "learning_rate": 8.73669983289009e-06, "loss": 5.7699, "step": 28050 }, { "epoch": 0.23, "learning_rate": 8.735838957145927e-06, "loss": 5.9057, "step": 28060 }, { "epoch": 0.23, "learning_rate": 8.734977830624258e-06, "loss": 6.3869, "step": 28070 }, { "epoch": 0.23, "learning_rate": 8.734116453382888e-06, "loss": 5.8622, "step": 28080 }, { "epoch": 0.23, "learning_rate": 8.73325482547964e-06, "loss": 5.5707, "step": 28090 }, { "epoch": 0.23, "learning_rate": 8.732392946972355e-06, "loss": 6.2165, "step": 28100 }, { "epoch": 0.23, "learning_rate": 8.731530817918885e-06, "loss": 5.6226, "step": 28110 }, { "epoch": 0.23, "learning_rate": 8.730668438377106e-06, "loss": 5.3118, "step": 28120 }, { "epoch": 0.23, "learning_rate": 8.729805808404903e-06, "loss": 6.5013, "step": 28130 }, { "epoch": 0.23, "learning_rate": 8.728942928060187e-06, "loss": 5.4136, "step": 28140 }, { "epoch": 0.23, "learning_rate": 8.728079797400878e-06, "loss": 6.0333, "step": 28150 }, { "epoch": 0.23, "learning_rate": 8.727216416484918e-06, "loss": 5.7529, "step": 28160 }, { "epoch": 0.23, "learning_rate": 8.726352785370262e-06, "loss": 5.8201, "step": 28170 }, { "epoch": 0.23, "learning_rate": 8.725488904114882e-06, "loss": 5.9181, "step": 28180 }, { "epoch": 0.23, "learning_rate": 8.72462477277677e-06, "loss": 5.1402, "step": 28190 }, { "epoch": 0.23, "learning_rate": 8.723760391413935e-06, "loss": 6.3761, "step": 28200 }, { "epoch": 0.23, "learning_rate": 8.722895760084397e-06, "loss": 5.8307, "step": 28210 }, { "epoch": 0.23, "learning_rate": 8.7220308788462e-06, "loss": 5.3366, "step": 28220 }, { "epoch": 0.23, "learning_rate": 8.721165747757399e-06, "loss": 5.6899, "step": 28230 }, { "epoch": 0.23, "learning_rate": 8.720300366876067e-06, "loss": 5.9321, "step": 28240 }, { "epoch": 0.23, "learning_rate": 8.719434736260297e-06, "loss": 5.7363, "step": 28250 }, { "epoch": 0.23, "learning_rate": 8.718568855968195e-06, "loss": 6.4123, "step": 28260 }, { "epoch": 0.23, "learning_rate": 8.717702726057889e-06, "loss": 5.7218, "step": 28270 }, { "epoch": 0.23, "learning_rate": 8.716836346587516e-06, "loss": 5.7848, "step": 28280 }, { "epoch": 0.23, "learning_rate": 8.715969717615235e-06, "loss": 5.6052, "step": 28290 }, { "epoch": 0.23, "learning_rate": 8.715102839199222e-06, "loss": 5.5504, "step": 28300 }, { "epoch": 0.23, "learning_rate": 8.714235711397666e-06, "loss": 5.7886, "step": 28310 }, { "epoch": 0.23, "learning_rate": 8.713368334268777e-06, "loss": 5.8696, "step": 28320 }, { "epoch": 0.23, "learning_rate": 8.71250070787078e-06, "loss": 5.9511, "step": 28330 }, { "epoch": 0.23, "learning_rate": 8.711632832261916e-06, "loss": 6.0051, "step": 28340 }, { "epoch": 0.23, "learning_rate": 8.710764707500444e-06, "loss": 5.7274, "step": 28350 }, { "epoch": 0.23, "learning_rate": 8.709896333644637e-06, "loss": 6.0546, "step": 28360 }, { "epoch": 0.23, "learning_rate": 8.709027710752788e-06, "loss": 5.9046, "step": 28370 }, { "epoch": 0.23, "learning_rate": 8.708158838883205e-06, "loss": 5.9917, "step": 28380 }, { "epoch": 0.23, "learning_rate": 8.707289718094214e-06, "loss": 6.0681, "step": 28390 }, { "epoch": 0.23, "learning_rate": 8.706420348444159e-06, "loss": 6.5182, "step": 28400 }, { "epoch": 0.23, "learning_rate": 8.705550729991392e-06, "loss": 6.4888, "step": 28410 }, { "epoch": 0.23, "learning_rate": 8.704680862794296e-06, "loss": 5.8839, "step": 28420 }, { "epoch": 0.23, "learning_rate": 8.70381074691126e-06, "loss": 5.8576, "step": 28430 }, { "epoch": 0.23, "learning_rate": 8.702940382400688e-06, "loss": 6.153, "step": 28440 }, { "epoch": 0.23, "learning_rate": 8.702069769321013e-06, "loss": 6.1636, "step": 28450 }, { "epoch": 0.23, "learning_rate": 8.701198907730673e-06, "loss": 6.232, "step": 28460 }, { "epoch": 0.23, "learning_rate": 8.700327797688129e-06, "loss": 6.0982, "step": 28470 }, { "epoch": 0.23, "learning_rate": 8.699456439251852e-06, "loss": 5.6638, "step": 28480 }, { "epoch": 0.23, "learning_rate": 8.69858483248034e-06, "loss": 5.6299, "step": 28490 }, { "epoch": 0.24, "learning_rate": 8.697712977432095e-06, "loss": 6.195, "step": 28500 }, { "epoch": 0.24, "learning_rate": 8.69684087416565e-06, "loss": 5.8176, "step": 28510 }, { "epoch": 0.24, "learning_rate": 8.695968522739541e-06, "loss": 6.7242, "step": 28520 }, { "epoch": 0.24, "learning_rate": 8.69509592321233e-06, "loss": 6.3246, "step": 28530 }, { "epoch": 0.24, "learning_rate": 8.694223075642592e-06, "loss": 6.4703, "step": 28540 }, { "epoch": 0.24, "learning_rate": 8.693349980088917e-06, "loss": 6.0678, "step": 28550 }, { "epoch": 0.24, "learning_rate": 8.692476636609917e-06, "loss": 5.7372, "step": 28560 }, { "epoch": 0.24, "learning_rate": 8.691603045264214e-06, "loss": 6.0466, "step": 28570 }, { "epoch": 0.24, "learning_rate": 8.690729206110454e-06, "loss": 5.5836, "step": 28580 }, { "epoch": 0.24, "learning_rate": 8.689855119207292e-06, "loss": 5.8633, "step": 28590 }, { "epoch": 0.24, "learning_rate": 8.688980784613403e-06, "loss": 6.0917, "step": 28600 }, { "epoch": 0.24, "learning_rate": 8.688106202387483e-06, "loss": 5.7191, "step": 28610 }, { "epoch": 0.24, "learning_rate": 8.687231372588237e-06, "loss": 6.0251, "step": 28620 }, { "epoch": 0.24, "learning_rate": 8.68635629527439e-06, "loss": 5.9386, "step": 28630 }, { "epoch": 0.24, "learning_rate": 8.685480970504686e-06, "loss": 6.0744, "step": 28640 }, { "epoch": 0.24, "learning_rate": 8.684605398337882e-06, "loss": 5.8715, "step": 28650 }, { "epoch": 0.24, "learning_rate": 8.683729578832754e-06, "loss": 5.6806, "step": 28660 }, { "epoch": 0.24, "learning_rate": 8.68285351204809e-06, "loss": 5.7914, "step": 28670 }, { "epoch": 0.24, "learning_rate": 8.681977198042703e-06, "loss": 5.7826, "step": 28680 }, { "epoch": 0.24, "learning_rate": 8.681100636875415e-06, "loss": 5.6151, "step": 28690 }, { "epoch": 0.24, "learning_rate": 8.68022382860507e-06, "loss": 5.7995, "step": 28700 }, { "epoch": 0.24, "learning_rate": 8.679346773290522e-06, "loss": 5.7083, "step": 28710 }, { "epoch": 0.24, "learning_rate": 8.678469470990646e-06, "loss": 6.2337, "step": 28720 }, { "epoch": 0.24, "learning_rate": 8.677591921764338e-06, "loss": 5.7314, "step": 28730 }, { "epoch": 0.24, "learning_rate": 8.6767141256705e-06, "loss": 5.774, "step": 28740 }, { "epoch": 0.24, "learning_rate": 8.67583608276806e-06, "loss": 5.5971, "step": 28750 }, { "epoch": 0.24, "learning_rate": 8.674957793115958e-06, "loss": 5.6923, "step": 28760 }, { "epoch": 0.24, "learning_rate": 8.674079256773147e-06, "loss": 5.6916, "step": 28770 }, { "epoch": 0.24, "learning_rate": 8.673200473798608e-06, "loss": 5.6633, "step": 28780 }, { "epoch": 0.24, "learning_rate": 8.672321444251326e-06, "loss": 5.7366, "step": 28790 }, { "epoch": 0.24, "learning_rate": 8.671442168190312e-06, "loss": 5.3066, "step": 28800 }, { "epoch": 0.24, "learning_rate": 8.670562645674587e-06, "loss": 6.5399, "step": 28810 }, { "epoch": 0.24, "learning_rate": 8.66968287676319e-06, "loss": 5.8797, "step": 28820 }, { "epoch": 0.24, "learning_rate": 8.668802861515182e-06, "loss": 5.8404, "step": 28830 }, { "epoch": 0.24, "learning_rate": 8.667922599989634e-06, "loss": 5.8472, "step": 28840 }, { "epoch": 0.24, "learning_rate": 8.667042092245635e-06, "loss": 5.6344, "step": 28850 }, { "epoch": 0.24, "learning_rate": 8.666161338342293e-06, "loss": 5.807, "step": 28860 }, { "epoch": 0.24, "learning_rate": 8.665280338338729e-06, "loss": 6.1478, "step": 28870 }, { "epoch": 0.24, "learning_rate": 8.664399092294082e-06, "loss": 5.6142, "step": 28880 }, { "epoch": 0.24, "learning_rate": 8.663517600267509e-06, "loss": 5.3683, "step": 28890 }, { "epoch": 0.24, "learning_rate": 8.662635862318183e-06, "loss": 5.9933, "step": 28900 }, { "epoch": 0.24, "learning_rate": 8.661753878505292e-06, "loss": 6.1135, "step": 28910 }, { "epoch": 0.24, "learning_rate": 8.66087164888804e-06, "loss": 5.4925, "step": 28920 }, { "epoch": 0.24, "learning_rate": 8.659989173525651e-06, "loss": 5.6088, "step": 28930 }, { "epoch": 0.24, "learning_rate": 8.659106452477363e-06, "loss": 5.3841, "step": 28940 }, { "epoch": 0.24, "learning_rate": 8.65822348580243e-06, "loss": 5.6668, "step": 28950 }, { "epoch": 0.24, "learning_rate": 8.657340273560126e-06, "loss": 5.7294, "step": 28960 }, { "epoch": 0.24, "learning_rate": 8.656456815809734e-06, "loss": 5.8854, "step": 28970 }, { "epoch": 0.24, "learning_rate": 8.655573112610562e-06, "loss": 6.2777, "step": 28980 }, { "epoch": 0.24, "learning_rate": 8.65468916402193e-06, "loss": 6.0326, "step": 28990 }, { "epoch": 0.24, "learning_rate": 8.653804970103175e-06, "loss": 5.949, "step": 29000 }, { "epoch": 0.24, "learning_rate": 8.652920530913649e-06, "loss": 5.8391, "step": 29010 }, { "epoch": 0.24, "learning_rate": 8.652035846512725e-06, "loss": 5.5997, "step": 29020 }, { "epoch": 0.24, "learning_rate": 8.651150916959789e-06, "loss": 6.269, "step": 29030 }, { "epoch": 0.24, "learning_rate": 8.650265742314244e-06, "loss": 6.3121, "step": 29040 }, { "epoch": 0.24, "learning_rate": 8.649380322635508e-06, "loss": 5.5911, "step": 29050 }, { "epoch": 0.24, "learning_rate": 8.648494657983019e-06, "loss": 5.6887, "step": 29060 }, { "epoch": 0.24, "learning_rate": 8.647608748416227e-06, "loss": 5.7869, "step": 29070 }, { "epoch": 0.24, "learning_rate": 8.646722593994605e-06, "loss": 5.7182, "step": 29080 }, { "epoch": 0.24, "learning_rate": 8.645836194777635e-06, "loss": 6.5608, "step": 29090 }, { "epoch": 0.24, "learning_rate": 8.644949550824819e-06, "loss": 5.901, "step": 29100 }, { "epoch": 0.24, "learning_rate": 8.644062662195676e-06, "loss": 6.1163, "step": 29110 }, { "epoch": 0.24, "learning_rate": 8.643175528949741e-06, "loss": 6.1036, "step": 29120 }, { "epoch": 0.24, "learning_rate": 8.642288151146562e-06, "loss": 6.4316, "step": 29130 }, { "epoch": 0.24, "learning_rate": 8.641400528845712e-06, "loss": 6.3261, "step": 29140 }, { "epoch": 0.24, "learning_rate": 8.64051266210677e-06, "loss": 5.9155, "step": 29150 }, { "epoch": 0.24, "learning_rate": 8.63962455098934e-06, "loss": 5.6055, "step": 29160 }, { "epoch": 0.24, "learning_rate": 8.638736195553034e-06, "loss": 5.7039, "step": 29170 }, { "epoch": 0.24, "learning_rate": 8.637847595857491e-06, "loss": 5.7344, "step": 29180 }, { "epoch": 0.24, "learning_rate": 8.636958751962354e-06, "loss": 5.65, "step": 29190 }, { "epoch": 0.24, "learning_rate": 8.636069663927293e-06, "loss": 6.324, "step": 29200 }, { "epoch": 0.24, "learning_rate": 8.635180331811991e-06, "loss": 5.6489, "step": 29210 }, { "epoch": 0.24, "learning_rate": 8.634290755676143e-06, "loss": 5.2774, "step": 29220 }, { "epoch": 0.24, "learning_rate": 8.633400935579469e-06, "loss": 6.2213, "step": 29230 }, { "epoch": 0.24, "learning_rate": 8.632510871581695e-06, "loss": 5.8852, "step": 29240 }, { "epoch": 0.24, "learning_rate": 8.63162056374257e-06, "loss": 5.4934, "step": 29250 }, { "epoch": 0.24, "learning_rate": 8.63073001212186e-06, "loss": 5.8361, "step": 29260 }, { "epoch": 0.24, "learning_rate": 8.629839216779346e-06, "loss": 6.1251, "step": 29270 }, { "epoch": 0.24, "learning_rate": 8.628948177774824e-06, "loss": 5.0745, "step": 29280 }, { "epoch": 0.24, "learning_rate": 8.628056895168106e-06, "loss": 6.4017, "step": 29290 }, { "epoch": 0.24, "learning_rate": 8.627165369019021e-06, "loss": 5.5288, "step": 29300 }, { "epoch": 0.24, "learning_rate": 8.626273599387419e-06, "loss": 5.6593, "step": 29310 }, { "epoch": 0.24, "learning_rate": 8.625381586333157e-06, "loss": 5.3479, "step": 29320 }, { "epoch": 0.24, "learning_rate": 8.62448932991612e-06, "loss": 5.7672, "step": 29330 }, { "epoch": 0.24, "learning_rate": 8.623596830196194e-06, "loss": 5.7578, "step": 29340 }, { "epoch": 0.24, "learning_rate": 8.6227040872333e-06, "loss": 6.1668, "step": 29350 }, { "epoch": 0.24, "learning_rate": 8.621811101087358e-06, "loss": 5.9201, "step": 29360 }, { "epoch": 0.24, "learning_rate": 8.620917871818316e-06, "loss": 6.8095, "step": 29370 }, { "epoch": 0.24, "learning_rate": 8.620024399486133e-06, "loss": 6.1668, "step": 29380 }, { "epoch": 0.24, "learning_rate": 8.619130684150787e-06, "loss": 6.154, "step": 29390 }, { "epoch": 0.24, "learning_rate": 8.618236725872269e-06, "loss": 6.2774, "step": 29400 }, { "epoch": 0.24, "learning_rate": 8.617342524710588e-06, "loss": 5.9622, "step": 29410 }, { "epoch": 0.24, "learning_rate": 8.61644808072577e-06, "loss": 5.7593, "step": 29420 }, { "epoch": 0.24, "learning_rate": 8.615553393977857e-06, "loss": 5.7751, "step": 29430 }, { "epoch": 0.24, "learning_rate": 8.614658464526908e-06, "loss": 5.9578, "step": 29440 }, { "epoch": 0.24, "learning_rate": 8.613763292432998e-06, "loss": 5.8718, "step": 29450 }, { "epoch": 0.24, "learning_rate": 8.612867877756213e-06, "loss": 5.7517, "step": 29460 }, { "epoch": 0.24, "learning_rate": 8.611972220556666e-06, "loss": 5.9671, "step": 29470 }, { "epoch": 0.24, "learning_rate": 8.611076320894475e-06, "loss": 5.8168, "step": 29480 }, { "epoch": 0.24, "learning_rate": 8.610180178829783e-06, "loss": 6.2134, "step": 29490 }, { "epoch": 0.24, "learning_rate": 8.609283794422744e-06, "loss": 5.6385, "step": 29500 }, { "epoch": 0.24, "learning_rate": 8.608387167733532e-06, "loss": 5.8309, "step": 29510 }, { "epoch": 0.24, "learning_rate": 8.607490298822332e-06, "loss": 5.5461, "step": 29520 }, { "epoch": 0.24, "learning_rate": 8.606593187749353e-06, "loss": 6.1364, "step": 29530 }, { "epoch": 0.24, "learning_rate": 8.605695834574811e-06, "loss": 6.1011, "step": 29540 }, { "epoch": 0.24, "learning_rate": 8.604798239358947e-06, "loss": 5.9105, "step": 29550 }, { "epoch": 0.24, "learning_rate": 8.603900402162011e-06, "loss": 5.9562, "step": 29560 }, { "epoch": 0.24, "learning_rate": 8.603002323044276e-06, "loss": 5.6909, "step": 29570 }, { "epoch": 0.24, "learning_rate": 8.602104002066025e-06, "loss": 6.1028, "step": 29580 }, { "epoch": 0.24, "learning_rate": 8.60120543928756e-06, "loss": 6.0892, "step": 29590 }, { "epoch": 0.24, "learning_rate": 8.600306634769202e-06, "loss": 5.8671, "step": 29600 }, { "epoch": 0.24, "learning_rate": 8.599407588571282e-06, "loss": 6.5069, "step": 29610 }, { "epoch": 0.24, "learning_rate": 8.598508300754157e-06, "loss": 5.4891, "step": 29620 }, { "epoch": 0.24, "learning_rate": 8.597608771378186e-06, "loss": 6.0135, "step": 29630 }, { "epoch": 0.24, "learning_rate": 8.596709000503754e-06, "loss": 5.1789, "step": 29640 }, { "epoch": 0.24, "learning_rate": 8.595808988191266e-06, "loss": 5.8708, "step": 29650 }, { "epoch": 0.24, "learning_rate": 8.59490873450113e-06, "loss": 5.6298, "step": 29660 }, { "epoch": 0.24, "learning_rate": 8.594008239493785e-06, "loss": 6.0596, "step": 29670 }, { "epoch": 0.24, "learning_rate": 8.593107503229672e-06, "loss": 6.339, "step": 29680 }, { "epoch": 0.24, "learning_rate": 8.592206525769261e-06, "loss": 5.4859, "step": 29690 }, { "epoch": 0.24, "learning_rate": 8.59130530717303e-06, "loss": 5.8563, "step": 29700 }, { "epoch": 0.25, "learning_rate": 8.590403847501473e-06, "loss": 6.1651, "step": 29710 }, { "epoch": 0.25, "learning_rate": 8.589502146815108e-06, "loss": 5.3467, "step": 29720 }, { "epoch": 0.25, "learning_rate": 8.588600205174457e-06, "loss": 5.956, "step": 29730 }, { "epoch": 0.25, "learning_rate": 8.587698022640071e-06, "loss": 5.854, "step": 29740 }, { "epoch": 0.25, "learning_rate": 8.586795599272512e-06, "loss": 5.784, "step": 29750 }, { "epoch": 0.25, "learning_rate": 8.585892935132353e-06, "loss": 5.4276, "step": 29760 }, { "epoch": 0.25, "learning_rate": 8.584990030280189e-06, "loss": 5.0438, "step": 29770 }, { "epoch": 0.25, "learning_rate": 8.58408688477663e-06, "loss": 6.6993, "step": 29780 }, { "epoch": 0.25, "learning_rate": 8.583183498682304e-06, "loss": 5.6712, "step": 29790 }, { "epoch": 0.25, "learning_rate": 8.582279872057848e-06, "loss": 6.2185, "step": 29800 }, { "epoch": 0.25, "learning_rate": 8.581376004963925e-06, "loss": 5.8574, "step": 29810 }, { "epoch": 0.25, "learning_rate": 8.580471897461207e-06, "loss": 6.4299, "step": 29820 }, { "epoch": 0.25, "learning_rate": 8.579567549610385e-06, "loss": 5.7545, "step": 29830 }, { "epoch": 0.25, "learning_rate": 8.578662961472166e-06, "loss": 6.078, "step": 29840 }, { "epoch": 0.25, "learning_rate": 8.577758133107273e-06, "loss": 5.9932, "step": 29850 }, { "epoch": 0.25, "learning_rate": 8.576853064576444e-06, "loss": 5.6888, "step": 29860 }, { "epoch": 0.25, "learning_rate": 8.575947755940433e-06, "loss": 5.9076, "step": 29870 }, { "epoch": 0.25, "learning_rate": 8.575042207260014e-06, "loss": 6.2831, "step": 29880 }, { "epoch": 0.25, "learning_rate": 8.574136418595971e-06, "loss": 6.1363, "step": 29890 }, { "epoch": 0.25, "learning_rate": 8.57323039000911e-06, "loss": 6.505, "step": 29900 }, { "epoch": 0.25, "learning_rate": 8.57232412156025e-06, "loss": 5.1913, "step": 29910 }, { "epoch": 0.25, "learning_rate": 8.571417613310225e-06, "loss": 5.9217, "step": 29920 }, { "epoch": 0.25, "learning_rate": 8.570510865319888e-06, "loss": 6.0391, "step": 29930 }, { "epoch": 0.25, "learning_rate": 8.569603877650103e-06, "loss": 6.1672, "step": 29940 }, { "epoch": 0.25, "learning_rate": 8.56869665036176e-06, "loss": 6.1328, "step": 29950 }, { "epoch": 0.25, "learning_rate": 8.567789183515756e-06, "loss": 5.8125, "step": 29960 }, { "epoch": 0.25, "learning_rate": 8.566881477173006e-06, "loss": 6.4841, "step": 29970 }, { "epoch": 0.25, "learning_rate": 8.565973531394444e-06, "loss": 5.9904, "step": 29980 }, { "epoch": 0.25, "learning_rate": 8.565065346241015e-06, "loss": 6.5378, "step": 29990 }, { "epoch": 0.25, "learning_rate": 8.564156921773688e-06, "loss": 6.0011, "step": 30000 }, { "epoch": 0.25, "learning_rate": 8.563248258053438e-06, "loss": 5.9951, "step": 30010 }, { "epoch": 0.25, "learning_rate": 8.562339355141265e-06, "loss": 5.8575, "step": 30020 }, { "epoch": 0.25, "learning_rate": 8.561430213098178e-06, "loss": 5.6084, "step": 30030 }, { "epoch": 0.25, "learning_rate": 8.56052083198521e-06, "loss": 6.8777, "step": 30040 }, { "epoch": 0.25, "learning_rate": 8.559611211863403e-06, "loss": 5.877, "step": 30050 }, { "epoch": 0.25, "learning_rate": 8.558701352793818e-06, "loss": 5.5299, "step": 30060 }, { "epoch": 0.25, "learning_rate": 8.55779125483753e-06, "loss": 5.5858, "step": 30070 }, { "epoch": 0.25, "learning_rate": 8.556880918055633e-06, "loss": 5.1916, "step": 30080 }, { "epoch": 0.25, "learning_rate": 8.555970342509235e-06, "loss": 5.9413, "step": 30090 }, { "epoch": 0.25, "learning_rate": 8.555059528259462e-06, "loss": 6.3026, "step": 30100 }, { "epoch": 0.25, "learning_rate": 8.554148475367453e-06, "loss": 6.0026, "step": 30110 }, { "epoch": 0.25, "learning_rate": 8.553237183894366e-06, "loss": 5.9117, "step": 30120 }, { "epoch": 0.25, "learning_rate": 8.552325653901373e-06, "loss": 6.0762, "step": 30130 }, { "epoch": 0.25, "learning_rate": 8.55141388544966e-06, "loss": 5.3896, "step": 30140 }, { "epoch": 0.25, "learning_rate": 8.550501878600439e-06, "loss": 5.4281, "step": 30150 }, { "epoch": 0.25, "learning_rate": 8.549589633414923e-06, "loss": 5.7646, "step": 30160 }, { "epoch": 0.25, "learning_rate": 8.548677149954355e-06, "loss": 6.0557, "step": 30170 }, { "epoch": 0.25, "learning_rate": 8.547764428279982e-06, "loss": 5.9068, "step": 30180 }, { "epoch": 0.25, "learning_rate": 8.546851468453076e-06, "loss": 6.5226, "step": 30190 }, { "epoch": 0.25, "learning_rate": 8.545938270534922e-06, "loss": 6.0716, "step": 30200 }, { "epoch": 0.25, "learning_rate": 8.54502483458682e-06, "loss": 5.3644, "step": 30210 }, { "epoch": 0.25, "learning_rate": 8.544111160670085e-06, "loss": 6.2694, "step": 30220 }, { "epoch": 0.25, "learning_rate": 8.543197248846054e-06, "loss": 5.8983, "step": 30230 }, { "epoch": 0.25, "learning_rate": 8.54228309917607e-06, "loss": 5.9776, "step": 30240 }, { "epoch": 0.25, "learning_rate": 8.541368711721502e-06, "loss": 5.9249, "step": 30250 }, { "epoch": 0.25, "learning_rate": 8.540454086543728e-06, "loss": 5.6661, "step": 30260 }, { "epoch": 0.25, "learning_rate": 8.539539223704146e-06, "loss": 6.2417, "step": 30270 }, { "epoch": 0.25, "learning_rate": 8.538624123264169e-06, "loss": 5.832, "step": 30280 }, { "epoch": 0.25, "learning_rate": 8.537708785285221e-06, "loss": 5.6686, "step": 30290 }, { "epoch": 0.25, "learning_rate": 8.536793209828754e-06, "loss": 6.0467, "step": 30300 }, { "epoch": 0.25, "learning_rate": 8.53587739695622e-06, "loss": 5.8167, "step": 30310 }, { "epoch": 0.25, "learning_rate": 8.5349613467291e-06, "loss": 5.8877, "step": 30320 }, { "epoch": 0.25, "learning_rate": 8.534045059208887e-06, "loss": 5.8232, "step": 30330 }, { "epoch": 0.25, "learning_rate": 8.533128534457086e-06, "loss": 5.3114, "step": 30340 }, { "epoch": 0.25, "learning_rate": 8.532211772535222e-06, "loss": 6.2921, "step": 30350 }, { "epoch": 0.25, "learning_rate": 8.531294773504835e-06, "loss": 6.1697, "step": 30360 }, { "epoch": 0.25, "learning_rate": 8.530377537427483e-06, "loss": 6.4208, "step": 30370 }, { "epoch": 0.25, "learning_rate": 8.529460064364733e-06, "loss": 5.6349, "step": 30380 }, { "epoch": 0.25, "learning_rate": 8.528542354378178e-06, "loss": 5.315, "step": 30390 }, { "epoch": 0.25, "learning_rate": 8.52762440752942e-06, "loss": 6.3596, "step": 30400 }, { "epoch": 0.25, "learning_rate": 8.526706223880075e-06, "loss": 5.5402, "step": 30410 }, { "epoch": 0.25, "learning_rate": 8.525787803491783e-06, "loss": 6.0998, "step": 30420 }, { "epoch": 0.25, "learning_rate": 8.524869146426191e-06, "loss": 6.1774, "step": 30430 }, { "epoch": 0.25, "learning_rate": 8.52395025274497e-06, "loss": 6.9572, "step": 30440 }, { "epoch": 0.25, "learning_rate": 8.5230311225098e-06, "loss": 5.9305, "step": 30450 }, { "epoch": 0.25, "learning_rate": 8.522111755782384e-06, "loss": 5.4999, "step": 30460 }, { "epoch": 0.25, "learning_rate": 8.521192152624431e-06, "loss": 6.1508, "step": 30470 }, { "epoch": 0.25, "learning_rate": 8.520272313097678e-06, "loss": 5.8146, "step": 30480 }, { "epoch": 0.25, "learning_rate": 8.519352237263865e-06, "loss": 5.258, "step": 30490 }, { "epoch": 0.25, "learning_rate": 8.518431925184758e-06, "loss": 6.1282, "step": 30500 }, { "epoch": 0.25, "learning_rate": 8.517511376922136e-06, "loss": 5.5414, "step": 30510 }, { "epoch": 0.25, "learning_rate": 8.516590592537792e-06, "loss": 5.7398, "step": 30520 }, { "epoch": 0.25, "learning_rate": 8.515669572093536e-06, "loss": 5.7043, "step": 30530 }, { "epoch": 0.25, "learning_rate": 8.514748315651193e-06, "loss": 6.0916, "step": 30540 }, { "epoch": 0.25, "learning_rate": 8.513826823272607e-06, "loss": 5.5419, "step": 30550 }, { "epoch": 0.25, "learning_rate": 8.512905095019632e-06, "loss": 5.8623, "step": 30560 }, { "epoch": 0.25, "learning_rate": 8.511983130954144e-06, "loss": 5.4868, "step": 30570 }, { "epoch": 0.25, "learning_rate": 8.511060931138031e-06, "loss": 6.3204, "step": 30580 }, { "epoch": 0.25, "learning_rate": 8.510138495633198e-06, "loss": 5.8859, "step": 30590 }, { "epoch": 0.25, "learning_rate": 8.509215824501567e-06, "loss": 5.8705, "step": 30600 }, { "epoch": 0.25, "learning_rate": 8.508292917805073e-06, "loss": 6.3778, "step": 30610 }, { "epoch": 0.25, "learning_rate": 8.50736977560567e-06, "loss": 6.0986, "step": 30620 }, { "epoch": 0.25, "learning_rate": 8.506446397965325e-06, "loss": 6.1583, "step": 30630 }, { "epoch": 0.25, "learning_rate": 8.50552278494602e-06, "loss": 5.9598, "step": 30640 }, { "epoch": 0.25, "learning_rate": 8.504598936609759e-06, "loss": 5.6415, "step": 30650 }, { "epoch": 0.25, "learning_rate": 8.503674853018555e-06, "loss": 6.0323, "step": 30660 }, { "epoch": 0.25, "learning_rate": 8.502750534234441e-06, "loss": 5.4258, "step": 30670 }, { "epoch": 0.25, "learning_rate": 8.501825980319466e-06, "loss": 6.5644, "step": 30680 }, { "epoch": 0.25, "learning_rate": 8.500901191335686e-06, "loss": 6.1465, "step": 30690 }, { "epoch": 0.25, "learning_rate": 8.499976167345185e-06, "loss": 5.9718, "step": 30700 }, { "epoch": 0.25, "learning_rate": 8.499050908410057e-06, "loss": 5.9446, "step": 30710 }, { "epoch": 0.25, "learning_rate": 8.498125414592412e-06, "loss": 5.6615, "step": 30720 }, { "epoch": 0.25, "learning_rate": 8.497199685954377e-06, "loss": 5.7589, "step": 30730 }, { "epoch": 0.25, "learning_rate": 8.496273722558091e-06, "loss": 6.4595, "step": 30740 }, { "epoch": 0.25, "learning_rate": 8.495347524465716e-06, "loss": 5.3682, "step": 30750 }, { "epoch": 0.25, "learning_rate": 8.494421091739422e-06, "loss": 6.3853, "step": 30760 }, { "epoch": 0.25, "learning_rate": 8.493494424441396e-06, "loss": 5.2975, "step": 30770 }, { "epoch": 0.25, "learning_rate": 8.492567522633849e-06, "loss": 5.6537, "step": 30780 }, { "epoch": 0.25, "learning_rate": 8.491640386378998e-06, "loss": 6.1518, "step": 30790 }, { "epoch": 0.25, "learning_rate": 8.490713015739079e-06, "loss": 6.2306, "step": 30800 }, { "epoch": 0.25, "learning_rate": 8.489785410776346e-06, "loss": 5.5383, "step": 30810 }, { "epoch": 0.25, "learning_rate": 8.488857571553064e-06, "loss": 6.2515, "step": 30820 }, { "epoch": 0.25, "learning_rate": 8.48792949813152e-06, "loss": 6.3491, "step": 30830 }, { "epoch": 0.25, "learning_rate": 8.487001190574008e-06, "loss": 6.2411, "step": 30840 }, { "epoch": 0.25, "learning_rate": 8.486072648942849e-06, "loss": 5.6804, "step": 30850 }, { "epoch": 0.25, "learning_rate": 8.485143873300371e-06, "loss": 5.6834, "step": 30860 }, { "epoch": 0.25, "learning_rate": 8.48421486370892e-06, "loss": 5.4902, "step": 30870 }, { "epoch": 0.25, "learning_rate": 8.48328562023086e-06, "loss": 5.7008, "step": 30880 }, { "epoch": 0.25, "learning_rate": 8.482356142928565e-06, "loss": 5.5454, "step": 30890 }, { "epoch": 0.25, "learning_rate": 8.48142643186443e-06, "loss": 6.2075, "step": 30900 }, { "epoch": 0.25, "learning_rate": 8.480496487100867e-06, "loss": 5.7698, "step": 30910 }, { "epoch": 0.25, "learning_rate": 8.479566308700298e-06, "loss": 5.8641, "step": 30920 }, { "epoch": 0.26, "learning_rate": 8.478635896725165e-06, "loss": 5.5428, "step": 30930 }, { "epoch": 0.26, "learning_rate": 8.477705251237923e-06, "loss": 6.4876, "step": 30940 }, { "epoch": 0.26, "learning_rate": 8.476774372301044e-06, "loss": 5.4925, "step": 30950 }, { "epoch": 0.26, "learning_rate": 8.475843259977016e-06, "loss": 5.5815, "step": 30960 }, { "epoch": 0.26, "learning_rate": 8.474911914328344e-06, "loss": 7.0016, "step": 30970 }, { "epoch": 0.26, "learning_rate": 8.473980335417542e-06, "loss": 6.4409, "step": 30980 }, { "epoch": 0.26, "learning_rate": 8.473048523307151e-06, "loss": 5.7836, "step": 30990 }, { "epoch": 0.26, "learning_rate": 8.472116478059715e-06, "loss": 6.2062, "step": 31000 }, { "epoch": 0.26, "learning_rate": 8.471184199737805e-06, "loss": 5.8315, "step": 31010 }, { "epoch": 0.26, "learning_rate": 8.470251688403998e-06, "loss": 5.9109, "step": 31020 }, { "epoch": 0.26, "learning_rate": 8.469318944120895e-06, "loss": 6.0234, "step": 31030 }, { "epoch": 0.26, "learning_rate": 8.468385966951107e-06, "loss": 6.3747, "step": 31040 }, { "epoch": 0.26, "learning_rate": 8.467452756957263e-06, "loss": 5.4756, "step": 31050 }, { "epoch": 0.26, "learning_rate": 8.466519314202006e-06, "loss": 6.1248, "step": 31060 }, { "epoch": 0.26, "learning_rate": 8.465585638747998e-06, "loss": 5.8459, "step": 31070 }, { "epoch": 0.26, "learning_rate": 8.46465173065791e-06, "loss": 5.706, "step": 31080 }, { "epoch": 0.26, "learning_rate": 8.463717589994436e-06, "loss": 5.8343, "step": 31090 }, { "epoch": 0.26, "learning_rate": 8.462783216820283e-06, "loss": 5.0276, "step": 31100 }, { "epoch": 0.26, "learning_rate": 8.461848611198174e-06, "loss": 6.0328, "step": 31110 }, { "epoch": 0.26, "learning_rate": 8.460913773190842e-06, "loss": 5.6228, "step": 31120 }, { "epoch": 0.26, "learning_rate": 8.459978702861045e-06, "loss": 6.6067, "step": 31130 }, { "epoch": 0.26, "learning_rate": 8.459043400271549e-06, "loss": 5.6933, "step": 31140 }, { "epoch": 0.26, "learning_rate": 8.45810786548514e-06, "loss": 5.7203, "step": 31150 }, { "epoch": 0.26, "learning_rate": 8.457172098564617e-06, "loss": 5.067, "step": 31160 }, { "epoch": 0.26, "learning_rate": 8.4562360995728e-06, "loss": 5.4249, "step": 31170 }, { "epoch": 0.26, "learning_rate": 8.455299868572513e-06, "loss": 5.9235, "step": 31180 }, { "epoch": 0.26, "learning_rate": 8.454363405626608e-06, "loss": 6.2476, "step": 31190 }, { "epoch": 0.26, "learning_rate": 8.453426710797946e-06, "loss": 5.4638, "step": 31200 }, { "epoch": 0.26, "learning_rate": 8.452489784149405e-06, "loss": 5.6858, "step": 31210 }, { "epoch": 0.26, "learning_rate": 8.451552625743878e-06, "loss": 5.8913, "step": 31220 }, { "epoch": 0.26, "learning_rate": 8.450615235644276e-06, "loss": 5.6661, "step": 31230 }, { "epoch": 0.26, "learning_rate": 8.44967761391352e-06, "loss": 5.0631, "step": 31240 }, { "epoch": 0.26, "learning_rate": 8.448739760614554e-06, "loss": 5.8298, "step": 31250 }, { "epoch": 0.26, "learning_rate": 8.447801675810333e-06, "loss": 6.0022, "step": 31260 }, { "epoch": 0.26, "learning_rate": 8.446863359563827e-06, "loss": 6.6641, "step": 31270 }, { "epoch": 0.26, "learning_rate": 8.445924811938021e-06, "loss": 6.1821, "step": 31280 }, { "epoch": 0.26, "learning_rate": 8.44498603299592e-06, "loss": 5.4998, "step": 31290 }, { "epoch": 0.26, "learning_rate": 8.444047022800546e-06, "loss": 5.5909, "step": 31300 }, { "epoch": 0.26, "learning_rate": 8.443107781414922e-06, "loss": 5.9145, "step": 31310 }, { "epoch": 0.26, "learning_rate": 8.442168308902107e-06, "loss": 6.4649, "step": 31320 }, { "epoch": 0.26, "learning_rate": 8.441228605325162e-06, "loss": 5.8732, "step": 31330 }, { "epoch": 0.26, "learning_rate": 8.440288670747163e-06, "loss": 6.5371, "step": 31340 }, { "epoch": 0.26, "learning_rate": 8.439348505231211e-06, "loss": 6.0609, "step": 31350 }, { "epoch": 0.26, "learning_rate": 8.438408108840412e-06, "loss": 5.6166, "step": 31360 }, { "epoch": 0.26, "learning_rate": 8.437467481637897e-06, "loss": 5.3521, "step": 31370 }, { "epoch": 0.26, "learning_rate": 8.436526623686807e-06, "loss": 7.0637, "step": 31380 }, { "epoch": 0.26, "learning_rate": 8.435585535050298e-06, "loss": 5.8795, "step": 31390 }, { "epoch": 0.26, "learning_rate": 8.434644215791543e-06, "loss": 5.5464, "step": 31400 }, { "epoch": 0.26, "learning_rate": 8.43370266597373e-06, "loss": 5.1991, "step": 31410 }, { "epoch": 0.26, "learning_rate": 8.432760885660064e-06, "loss": 6.0314, "step": 31420 }, { "epoch": 0.26, "learning_rate": 8.431818874913764e-06, "loss": 6.1051, "step": 31430 }, { "epoch": 0.26, "learning_rate": 8.430876633798065e-06, "loss": 5.8034, "step": 31440 }, { "epoch": 0.26, "learning_rate": 8.429934162376218e-06, "loss": 6.4696, "step": 31450 }, { "epoch": 0.26, "learning_rate": 8.428991460711487e-06, "loss": 5.7371, "step": 31460 }, { "epoch": 0.26, "learning_rate": 8.428048528867155e-06, "loss": 5.779, "step": 31470 }, { "epoch": 0.26, "learning_rate": 8.427105366906515e-06, "loss": 5.5996, "step": 31480 }, { "epoch": 0.26, "learning_rate": 8.426161974892886e-06, "loss": 5.5447, "step": 31490 }, { "epoch": 0.26, "learning_rate": 8.425218352889589e-06, "loss": 5.8051, "step": 31500 }, { "epoch": 0.26, "learning_rate": 8.42427450095997e-06, "loss": 6.0581, "step": 31510 }, { "epoch": 0.26, "learning_rate": 8.423330419167386e-06, "loss": 6.1937, "step": 31520 }, { "epoch": 0.26, "learning_rate": 8.422386107575211e-06, "loss": 5.7348, "step": 31530 }, { "epoch": 0.26, "learning_rate": 8.421441566246835e-06, "loss": 6.3624, "step": 31540 }, { "epoch": 0.26, "learning_rate": 8.420496795245665e-06, "loss": 5.5, "step": 31550 }, { "epoch": 0.26, "learning_rate": 8.419551794635117e-06, "loss": 6.2422, "step": 31560 }, { "epoch": 0.26, "learning_rate": 8.418606564478629e-06, "loss": 5.2824, "step": 31570 }, { "epoch": 0.26, "learning_rate": 8.417661104839649e-06, "loss": 5.298, "step": 31580 }, { "epoch": 0.26, "learning_rate": 8.416715415781648e-06, "loss": 5.3707, "step": 31590 }, { "epoch": 0.26, "learning_rate": 8.415769497368104e-06, "loss": 5.7332, "step": 31600 }, { "epoch": 0.26, "learning_rate": 8.414823349662515e-06, "loss": 5.7179, "step": 31610 }, { "epoch": 0.26, "learning_rate": 8.413876972728395e-06, "loss": 5.7249, "step": 31620 }, { "epoch": 0.26, "learning_rate": 8.41293036662927e-06, "loss": 6.0485, "step": 31630 }, { "epoch": 0.26, "learning_rate": 8.411983531428685e-06, "loss": 6.3642, "step": 31640 }, { "epoch": 0.26, "learning_rate": 8.411036467190197e-06, "loss": 5.7678, "step": 31650 }, { "epoch": 0.26, "learning_rate": 8.41008917397738e-06, "loss": 5.5008, "step": 31660 }, { "epoch": 0.26, "learning_rate": 8.409141651853825e-06, "loss": 5.8948, "step": 31670 }, { "epoch": 0.26, "learning_rate": 8.408193900883136e-06, "loss": 5.9333, "step": 31680 }, { "epoch": 0.26, "learning_rate": 8.407245921128933e-06, "loss": 5.8694, "step": 31690 }, { "epoch": 0.26, "learning_rate": 8.406297712654851e-06, "loss": 5.5681, "step": 31700 }, { "epoch": 0.26, "learning_rate": 8.405349275524544e-06, "loss": 6.2204, "step": 31710 }, { "epoch": 0.26, "learning_rate": 8.404400609801674e-06, "loss": 5.8438, "step": 31720 }, { "epoch": 0.26, "learning_rate": 8.403451715549925e-06, "loss": 5.9946, "step": 31730 }, { "epoch": 0.26, "learning_rate": 8.402502592832991e-06, "loss": 5.7841, "step": 31740 }, { "epoch": 0.26, "learning_rate": 8.401553241714588e-06, "loss": 6.1948, "step": 31750 }, { "epoch": 0.26, "learning_rate": 8.400603662258441e-06, "loss": 5.8614, "step": 31760 }, { "epoch": 0.26, "learning_rate": 8.399653854528296e-06, "loss": 5.8414, "step": 31770 }, { "epoch": 0.26, "learning_rate": 8.398703818587908e-06, "loss": 5.8706, "step": 31780 }, { "epoch": 0.26, "learning_rate": 8.397753554501051e-06, "loss": 6.1304, "step": 31790 }, { "epoch": 0.26, "learning_rate": 8.396803062331515e-06, "loss": 5.4132, "step": 31800 }, { "epoch": 0.26, "learning_rate": 8.395852342143103e-06, "loss": 6.3981, "step": 31810 }, { "epoch": 0.26, "learning_rate": 8.394901393999635e-06, "loss": 6.3001, "step": 31820 }, { "epoch": 0.26, "learning_rate": 8.393950217964946e-06, "loss": 6.0959, "step": 31830 }, { "epoch": 0.26, "learning_rate": 8.392998814102887e-06, "loss": 5.8578, "step": 31840 }, { "epoch": 0.26, "learning_rate": 8.392047182477321e-06, "loss": 5.6317, "step": 31850 }, { "epoch": 0.26, "learning_rate": 8.39109532315213e-06, "loss": 6.039, "step": 31860 }, { "epoch": 0.26, "learning_rate": 8.390143236191211e-06, "loss": 5.9227, "step": 31870 }, { "epoch": 0.26, "learning_rate": 8.389190921658473e-06, "loss": 5.84, "step": 31880 }, { "epoch": 0.26, "learning_rate": 8.388238379617844e-06, "loss": 6.6195, "step": 31890 }, { "epoch": 0.26, "learning_rate": 8.387285610133264e-06, "loss": 5.1062, "step": 31900 }, { "epoch": 0.26, "learning_rate": 8.386332613268693e-06, "loss": 5.6228, "step": 31910 }, { "epoch": 0.26, "learning_rate": 8.3853793890881e-06, "loss": 5.944, "step": 31920 }, { "epoch": 0.26, "learning_rate": 8.384425937655473e-06, "loss": 6.3399, "step": 31930 }, { "epoch": 0.26, "learning_rate": 8.383472259034819e-06, "loss": 6.3078, "step": 31940 }, { "epoch": 0.26, "learning_rate": 8.382518353290152e-06, "loss": 6.0821, "step": 31950 }, { "epoch": 0.26, "learning_rate": 8.381564220485508e-06, "loss": 5.9086, "step": 31960 }, { "epoch": 0.26, "learning_rate": 8.38060986068493e-06, "loss": 5.6265, "step": 31970 }, { "epoch": 0.26, "learning_rate": 8.379655273952489e-06, "loss": 5.5176, "step": 31980 }, { "epoch": 0.26, "learning_rate": 8.378700460352258e-06, "loss": 6.0117, "step": 31990 }, { "epoch": 0.26, "learning_rate": 8.377745419948335e-06, "loss": 5.649, "step": 32000 }, { "epoch": 0.26, "learning_rate": 8.376790152804829e-06, "loss": 5.5999, "step": 32010 }, { "epoch": 0.26, "learning_rate": 8.375834658985865e-06, "loss": 6.0963, "step": 32020 }, { "epoch": 0.26, "learning_rate": 8.374878938555579e-06, "loss": 6.5294, "step": 32030 }, { "epoch": 0.26, "learning_rate": 8.373922991578132e-06, "loss": 6.1584, "step": 32040 }, { "epoch": 0.26, "learning_rate": 8.37296681811769e-06, "loss": 6.5293, "step": 32050 }, { "epoch": 0.26, "learning_rate": 8.372010418238438e-06, "loss": 6.5969, "step": 32060 }, { "epoch": 0.26, "learning_rate": 8.37105379200458e-06, "loss": 6.0068, "step": 32070 }, { "epoch": 0.26, "learning_rate": 8.370096939480333e-06, "loss": 5.9076, "step": 32080 }, { "epoch": 0.26, "learning_rate": 8.369139860729922e-06, "loss": 6.2082, "step": 32090 }, { "epoch": 0.26, "learning_rate": 8.3681825558176e-06, "loss": 5.9559, "step": 32100 }, { "epoch": 0.26, "learning_rate": 8.367225024807623e-06, "loss": 6.2679, "step": 32110 }, { "epoch": 0.26, "learning_rate": 8.366267267764272e-06, "loss": 5.9392, "step": 32120 }, { "epoch": 0.26, "learning_rate": 8.365309284751837e-06, "loss": 5.9426, "step": 32130 }, { "epoch": 0.27, "learning_rate": 8.364351075834623e-06, "loss": 5.623, "step": 32140 }, { "epoch": 0.27, "learning_rate": 8.363392641076956e-06, "loss": 5.9165, "step": 32150 }, { "epoch": 0.27, "learning_rate": 8.36243398054317e-06, "loss": 6.0029, "step": 32160 }, { "epoch": 0.27, "learning_rate": 8.36147509429762e-06, "loss": 6.5581, "step": 32170 }, { "epoch": 0.27, "learning_rate": 8.360515982404674e-06, "loss": 5.8866, "step": 32180 }, { "epoch": 0.27, "learning_rate": 8.35955664492871e-06, "loss": 5.499, "step": 32190 }, { "epoch": 0.27, "learning_rate": 8.35859708193413e-06, "loss": 5.9022, "step": 32200 }, { "epoch": 0.27, "learning_rate": 8.357637293485346e-06, "loss": 5.671, "step": 32210 }, { "epoch": 0.27, "learning_rate": 8.356677279646789e-06, "loss": 5.989, "step": 32220 }, { "epoch": 0.27, "learning_rate": 8.355717040482898e-06, "loss": 6.1522, "step": 32230 }, { "epoch": 0.27, "learning_rate": 8.354756576058134e-06, "loss": 5.9516, "step": 32240 }, { "epoch": 0.27, "learning_rate": 8.35379588643697e-06, "loss": 5.6577, "step": 32250 }, { "epoch": 0.27, "learning_rate": 8.352834971683892e-06, "loss": 5.7335, "step": 32260 }, { "epoch": 0.27, "learning_rate": 8.351873831863409e-06, "loss": 6.1091, "step": 32270 }, { "epoch": 0.27, "learning_rate": 8.350912467040036e-06, "loss": 6.606, "step": 32280 }, { "epoch": 0.27, "learning_rate": 8.34995087727831e-06, "loss": 6.0491, "step": 32290 }, { "epoch": 0.27, "learning_rate": 8.348989062642776e-06, "loss": 5.926, "step": 32300 }, { "epoch": 0.27, "learning_rate": 8.348027023198002e-06, "loss": 5.274, "step": 32310 }, { "epoch": 0.27, "learning_rate": 8.347064759008563e-06, "loss": 6.5254, "step": 32320 }, { "epoch": 0.27, "learning_rate": 8.346102270139061e-06, "loss": 6.0481, "step": 32330 }, { "epoch": 0.27, "learning_rate": 8.345139556654097e-06, "loss": 5.6538, "step": 32340 }, { "epoch": 0.27, "learning_rate": 8.3441766186183e-06, "loss": 5.5194, "step": 32350 }, { "epoch": 0.27, "learning_rate": 8.34321345609631e-06, "loss": 6.0451, "step": 32360 }, { "epoch": 0.27, "learning_rate": 8.342250069152779e-06, "loss": 5.8863, "step": 32370 }, { "epoch": 0.27, "learning_rate": 8.34128645785238e-06, "loss": 5.6656, "step": 32380 }, { "epoch": 0.27, "learning_rate": 8.340322622259796e-06, "loss": 5.587, "step": 32390 }, { "epoch": 0.27, "learning_rate": 8.339358562439725e-06, "loss": 5.262, "step": 32400 }, { "epoch": 0.27, "learning_rate": 8.338394278456886e-06, "loss": 6.1648, "step": 32410 }, { "epoch": 0.27, "learning_rate": 8.337429770376005e-06, "loss": 6.1703, "step": 32420 }, { "epoch": 0.27, "learning_rate": 8.33646503826183e-06, "loss": 5.693, "step": 32430 }, { "epoch": 0.27, "learning_rate": 8.33550008217912e-06, "loss": 5.6437, "step": 32440 }, { "epoch": 0.27, "learning_rate": 8.334534902192649e-06, "loss": 6.1414, "step": 32450 }, { "epoch": 0.27, "learning_rate": 8.33356949836721e-06, "loss": 6.0545, "step": 32460 }, { "epoch": 0.27, "learning_rate": 8.332603870767604e-06, "loss": 5.5843, "step": 32470 }, { "epoch": 0.27, "learning_rate": 8.331638019458655e-06, "loss": 5.9915, "step": 32480 }, { "epoch": 0.27, "learning_rate": 8.330671944505195e-06, "loss": 5.7065, "step": 32490 }, { "epoch": 0.27, "learning_rate": 8.329705645972077e-06, "loss": 5.1181, "step": 32500 }, { "epoch": 0.27, "learning_rate": 8.328739123924165e-06, "loss": 5.3908, "step": 32510 }, { "epoch": 0.27, "learning_rate": 8.327772378426338e-06, "loss": 5.6708, "step": 32520 }, { "epoch": 0.27, "learning_rate": 8.326805409543495e-06, "loss": 6.1676, "step": 32530 }, { "epoch": 0.27, "learning_rate": 8.325838217340541e-06, "loss": 5.3527, "step": 32540 }, { "epoch": 0.27, "learning_rate": 8.324870801882405e-06, "loss": 5.8659, "step": 32550 }, { "epoch": 0.27, "learning_rate": 8.323903163234025e-06, "loss": 5.714, "step": 32560 }, { "epoch": 0.27, "learning_rate": 8.322935301460358e-06, "loss": 5.439, "step": 32570 }, { "epoch": 0.27, "learning_rate": 8.321967216626371e-06, "loss": 5.4281, "step": 32580 }, { "epoch": 0.27, "learning_rate": 8.320998908797055e-06, "loss": 5.6728, "step": 32590 }, { "epoch": 0.27, "learning_rate": 8.320030378037405e-06, "loss": 6.2254, "step": 32600 }, { "epoch": 0.27, "learning_rate": 8.319061624412438e-06, "loss": 5.5543, "step": 32610 }, { "epoch": 0.27, "learning_rate": 8.31809264798718e-06, "loss": 6.021, "step": 32620 }, { "epoch": 0.27, "learning_rate": 8.317123448826682e-06, "loss": 5.6381, "step": 32630 }, { "epoch": 0.27, "learning_rate": 8.316154026996001e-06, "loss": 6.1491, "step": 32640 }, { "epoch": 0.27, "learning_rate": 8.315184382560212e-06, "loss": 5.7344, "step": 32650 }, { "epoch": 0.27, "learning_rate": 8.314214515584407e-06, "loss": 5.7213, "step": 32660 }, { "epoch": 0.27, "learning_rate": 8.313244426133686e-06, "loss": 5.8855, "step": 32670 }, { "epoch": 0.27, "learning_rate": 8.312274114273173e-06, "loss": 5.6671, "step": 32680 }, { "epoch": 0.27, "learning_rate": 8.311303580068001e-06, "loss": 5.5793, "step": 32690 }, { "epoch": 0.27, "learning_rate": 8.31033282358332e-06, "loss": 5.6982, "step": 32700 }, { "epoch": 0.27, "learning_rate": 8.309361844884291e-06, "loss": 5.8228, "step": 32710 }, { "epoch": 0.27, "learning_rate": 8.3083906440361e-06, "loss": 6.4461, "step": 32720 }, { "epoch": 0.27, "learning_rate": 8.307419221103936e-06, "loss": 6.4559, "step": 32730 }, { "epoch": 0.27, "learning_rate": 8.30644757615301e-06, "loss": 5.613, "step": 32740 }, { "epoch": 0.27, "learning_rate": 8.305475709248544e-06, "loss": 5.6215, "step": 32750 }, { "epoch": 0.27, "learning_rate": 8.304503620455782e-06, "loss": 5.6684, "step": 32760 }, { "epoch": 0.27, "learning_rate": 8.303531309839974e-06, "loss": 5.4831, "step": 32770 }, { "epoch": 0.27, "learning_rate": 8.302558777466388e-06, "loss": 5.5208, "step": 32780 }, { "epoch": 0.27, "learning_rate": 8.30158602340031e-06, "loss": 5.6357, "step": 32790 }, { "epoch": 0.27, "learning_rate": 8.300613047707037e-06, "loss": 6.4841, "step": 32800 }, { "epoch": 0.27, "learning_rate": 8.299639850451885e-06, "loss": 6.08, "step": 32810 }, { "epoch": 0.27, "learning_rate": 8.29866643170018e-06, "loss": 5.3746, "step": 32820 }, { "epoch": 0.27, "learning_rate": 8.297692791517264e-06, "loss": 5.6708, "step": 32830 }, { "epoch": 0.27, "learning_rate": 8.296718929968498e-06, "loss": 6.1331, "step": 32840 }, { "epoch": 0.27, "learning_rate": 8.295744847119254e-06, "loss": 6.1045, "step": 32850 }, { "epoch": 0.27, "learning_rate": 8.294770543034918e-06, "loss": 5.7395, "step": 32860 }, { "epoch": 0.27, "learning_rate": 8.293796017780894e-06, "loss": 6.3299, "step": 32870 }, { "epoch": 0.27, "learning_rate": 8.292821271422601e-06, "loss": 5.8379, "step": 32880 }, { "epoch": 0.27, "learning_rate": 8.291846304025468e-06, "loss": 6.0219, "step": 32890 }, { "epoch": 0.27, "learning_rate": 8.290871115654944e-06, "loss": 6.0739, "step": 32900 }, { "epoch": 0.27, "learning_rate": 8.28989570637649e-06, "loss": 6.3651, "step": 32910 }, { "epoch": 0.27, "learning_rate": 8.288920076255584e-06, "loss": 5.8453, "step": 32920 }, { "epoch": 0.27, "learning_rate": 8.287944225357718e-06, "loss": 5.5812, "step": 32930 }, { "epoch": 0.27, "learning_rate": 8.286968153748396e-06, "loss": 5.4006, "step": 32940 }, { "epoch": 0.27, "learning_rate": 8.285991861493142e-06, "loss": 6.187, "step": 32950 }, { "epoch": 0.27, "learning_rate": 8.28501534865749e-06, "loss": 6.0937, "step": 32960 }, { "epoch": 0.27, "learning_rate": 8.28403861530699e-06, "loss": 6.0511, "step": 32970 }, { "epoch": 0.27, "learning_rate": 8.283061661507213e-06, "loss": 5.6073, "step": 32980 }, { "epoch": 0.27, "learning_rate": 8.282084487323733e-06, "loss": 5.8765, "step": 32990 }, { "epoch": 0.27, "learning_rate": 8.281107092822148e-06, "loss": 6.0144, "step": 33000 }, { "epoch": 0.27, "learning_rate": 8.280129478068069e-06, "loss": 6.0085, "step": 33010 }, { "epoch": 0.27, "learning_rate": 8.27915164312712e-06, "loss": 5.6433, "step": 33020 }, { "epoch": 0.27, "learning_rate": 8.278173588064939e-06, "loss": 5.6168, "step": 33030 }, { "epoch": 0.27, "learning_rate": 8.277195312947184e-06, "loss": 6.3771, "step": 33040 }, { "epoch": 0.27, "learning_rate": 8.276216817839519e-06, "loss": 6.5387, "step": 33050 }, { "epoch": 0.27, "learning_rate": 8.275238102807635e-06, "loss": 6.0451, "step": 33060 }, { "epoch": 0.27, "learning_rate": 8.274259167917224e-06, "loss": 5.5461, "step": 33070 }, { "epoch": 0.27, "learning_rate": 8.273280013234002e-06, "loss": 6.1437, "step": 33080 }, { "epoch": 0.27, "learning_rate": 8.272300638823698e-06, "loss": 5.8074, "step": 33090 }, { "epoch": 0.27, "learning_rate": 8.271321044752055e-06, "loss": 5.7906, "step": 33100 }, { "epoch": 0.27, "learning_rate": 8.270341231084828e-06, "loss": 5.9387, "step": 33110 }, { "epoch": 0.27, "learning_rate": 8.26936119788779e-06, "loss": 5.8424, "step": 33120 }, { "epoch": 0.27, "learning_rate": 8.268380945226733e-06, "loss": 5.4619, "step": 33130 }, { "epoch": 0.27, "learning_rate": 8.267400473167454e-06, "loss": 5.8643, "step": 33140 }, { "epoch": 0.27, "learning_rate": 8.266419781775772e-06, "loss": 6.4061, "step": 33150 }, { "epoch": 0.27, "learning_rate": 8.265438871117517e-06, "loss": 6.4202, "step": 33160 }, { "epoch": 0.27, "learning_rate": 8.264457741258533e-06, "loss": 6.5892, "step": 33170 }, { "epoch": 0.27, "learning_rate": 8.263476392264687e-06, "loss": 5.6135, "step": 33180 }, { "epoch": 0.27, "learning_rate": 8.262494824201848e-06, "loss": 5.4523, "step": 33190 }, { "epoch": 0.27, "learning_rate": 8.261513037135911e-06, "loss": 5.9368, "step": 33200 }, { "epoch": 0.27, "learning_rate": 8.260531031132779e-06, "loss": 5.4305, "step": 33210 }, { "epoch": 0.27, "learning_rate": 8.25954880625837e-06, "loss": 5.7179, "step": 33220 }, { "epoch": 0.27, "learning_rate": 8.258566362578619e-06, "loss": 6.4802, "step": 33230 }, { "epoch": 0.27, "learning_rate": 8.257583700159478e-06, "loss": 5.9189, "step": 33240 }, { "epoch": 0.27, "learning_rate": 8.256600819066908e-06, "loss": 6.1941, "step": 33250 }, { "epoch": 0.27, "learning_rate": 8.255617719366887e-06, "loss": 5.569, "step": 33260 }, { "epoch": 0.27, "learning_rate": 8.254634401125407e-06, "loss": 6.5721, "step": 33270 }, { "epoch": 0.27, "learning_rate": 8.25365086440848e-06, "loss": 5.7543, "step": 33280 }, { "epoch": 0.27, "learning_rate": 8.252667109282125e-06, "loss": 5.35, "step": 33290 }, { "epoch": 0.27, "learning_rate": 8.251683135812381e-06, "loss": 6.1579, "step": 33300 }, { "epoch": 0.27, "learning_rate": 8.250698944065296e-06, "loss": 5.875, "step": 33310 }, { "epoch": 0.27, "learning_rate": 8.24971453410694e-06, "loss": 6.4973, "step": 33320 }, { "epoch": 0.27, "learning_rate": 8.248729906003395e-06, "loss": 5.2769, "step": 33330 }, { "epoch": 0.27, "learning_rate": 8.24774505982075e-06, "loss": 5.3068, "step": 33340 }, { "epoch": 0.28, "learning_rate": 8.246759995625125e-06, "loss": 5.7596, "step": 33350 }, { "epoch": 0.28, "learning_rate": 8.245774713482635e-06, "loss": 5.4413, "step": 33360 }, { "epoch": 0.28, "learning_rate": 8.244789213459427e-06, "loss": 5.9666, "step": 33370 }, { "epoch": 0.28, "learning_rate": 8.243803495621651e-06, "loss": 6.0162, "step": 33380 }, { "epoch": 0.28, "learning_rate": 8.242817560035476e-06, "loss": 5.8845, "step": 33390 }, { "epoch": 0.28, "learning_rate": 8.24183140676709e-06, "loss": 5.0888, "step": 33400 }, { "epoch": 0.28, "learning_rate": 8.240845035882683e-06, "loss": 5.6598, "step": 33410 }, { "epoch": 0.28, "learning_rate": 8.239858447448474e-06, "loss": 5.637, "step": 33420 }, { "epoch": 0.28, "learning_rate": 8.238871641530688e-06, "loss": 5.8429, "step": 33430 }, { "epoch": 0.28, "learning_rate": 8.237884618195569e-06, "loss": 5.9923, "step": 33440 }, { "epoch": 0.28, "learning_rate": 8.236897377509368e-06, "loss": 5.7248, "step": 33450 }, { "epoch": 0.28, "learning_rate": 8.235909919538363e-06, "loss": 4.979, "step": 33460 }, { "epoch": 0.28, "learning_rate": 8.234922244348832e-06, "loss": 5.8234, "step": 33470 }, { "epoch": 0.28, "learning_rate": 8.233934352007081e-06, "loss": 6.2781, "step": 33480 }, { "epoch": 0.28, "learning_rate": 8.232946242579422e-06, "loss": 5.8263, "step": 33490 }, { "epoch": 0.28, "learning_rate": 8.231957916132187e-06, "loss": 5.6346, "step": 33500 }, { "epoch": 0.28, "learning_rate": 8.230969372731717e-06, "loss": 5.9279, "step": 33510 }, { "epoch": 0.28, "learning_rate": 8.229980612444373e-06, "loss": 5.5812, "step": 33520 }, { "epoch": 0.28, "learning_rate": 8.228991635336526e-06, "loss": 5.612, "step": 33530 }, { "epoch": 0.28, "learning_rate": 8.228002441474563e-06, "loss": 5.5706, "step": 33540 }, { "epoch": 0.28, "learning_rate": 8.227013030924887e-06, "loss": 5.6936, "step": 33550 }, { "epoch": 0.28, "learning_rate": 8.226023403753914e-06, "loss": 5.8701, "step": 33560 }, { "epoch": 0.28, "learning_rate": 8.225033560028078e-06, "loss": 5.7105, "step": 33570 }, { "epoch": 0.28, "learning_rate": 8.224043499813822e-06, "loss": 6.0913, "step": 33580 }, { "epoch": 0.28, "learning_rate": 8.223053223177607e-06, "loss": 5.6728, "step": 33590 }, { "epoch": 0.28, "learning_rate": 8.222062730185909e-06, "loss": 5.6061, "step": 33600 }, { "epoch": 0.28, "learning_rate": 8.221072020905214e-06, "loss": 6.4894, "step": 33610 }, { "epoch": 0.28, "learning_rate": 8.220081095402028e-06, "loss": 5.593, "step": 33620 }, { "epoch": 0.28, "learning_rate": 8.21908995374287e-06, "loss": 6.3143, "step": 33630 }, { "epoch": 0.28, "learning_rate": 8.218098595994272e-06, "loss": 5.4931, "step": 33640 }, { "epoch": 0.28, "learning_rate": 8.217107022222784e-06, "loss": 5.3829, "step": 33650 }, { "epoch": 0.28, "learning_rate": 8.216115232494962e-06, "loss": 5.8669, "step": 33660 }, { "epoch": 0.28, "learning_rate": 8.215123226877387e-06, "loss": 5.571, "step": 33670 }, { "epoch": 0.28, "learning_rate": 8.214131005436648e-06, "loss": 5.6468, "step": 33680 }, { "epoch": 0.28, "learning_rate": 8.213138568239352e-06, "loss": 5.9499, "step": 33690 }, { "epoch": 0.28, "learning_rate": 8.212145915352116e-06, "loss": 5.7901, "step": 33700 }, { "epoch": 0.28, "learning_rate": 8.211153046841578e-06, "loss": 5.3083, "step": 33710 }, { "epoch": 0.28, "learning_rate": 8.210159962774383e-06, "loss": 6.0829, "step": 33720 }, { "epoch": 0.28, "learning_rate": 8.209166663217197e-06, "loss": 5.6246, "step": 33730 }, { "epoch": 0.28, "learning_rate": 8.208173148236697e-06, "loss": 6.1778, "step": 33740 }, { "epoch": 0.28, "learning_rate": 8.207179417899574e-06, "loss": 5.2848, "step": 33750 }, { "epoch": 0.28, "learning_rate": 8.206185472272537e-06, "loss": 6.6079, "step": 33760 }, { "epoch": 0.28, "learning_rate": 8.205191311422303e-06, "loss": 5.672, "step": 33770 }, { "epoch": 0.28, "learning_rate": 8.204196935415613e-06, "loss": 5.3163, "step": 33780 }, { "epoch": 0.28, "learning_rate": 8.203202344319212e-06, "loss": 6.0562, "step": 33790 }, { "epoch": 0.28, "learning_rate": 8.20220753819987e-06, "loss": 5.636, "step": 33800 }, { "epoch": 0.28, "learning_rate": 8.201212517124358e-06, "loss": 6.0649, "step": 33810 }, { "epoch": 0.28, "learning_rate": 8.200217281159474e-06, "loss": 5.7598, "step": 33820 }, { "epoch": 0.28, "learning_rate": 8.199221830372028e-06, "loss": 6.031, "step": 33830 }, { "epoch": 0.28, "learning_rate": 8.198226164828837e-06, "loss": 5.509, "step": 33840 }, { "epoch": 0.28, "learning_rate": 8.197230284596742e-06, "loss": 5.9983, "step": 33850 }, { "epoch": 0.28, "learning_rate": 8.196234189742591e-06, "loss": 5.6731, "step": 33860 }, { "epoch": 0.28, "learning_rate": 8.195237880333252e-06, "loss": 6.0209, "step": 33870 }, { "epoch": 0.28, "learning_rate": 8.1942413564356e-06, "loss": 5.7663, "step": 33880 }, { "epoch": 0.28, "learning_rate": 8.193244618116534e-06, "loss": 6.0372, "step": 33890 }, { "epoch": 0.28, "learning_rate": 8.19224766544296e-06, "loss": 6.3111, "step": 33900 }, { "epoch": 0.28, "learning_rate": 8.191250498481802e-06, "loss": 6.0827, "step": 33910 }, { "epoch": 0.28, "learning_rate": 8.190253117299997e-06, "loss": 5.5465, "step": 33920 }, { "epoch": 0.28, "learning_rate": 8.189255521964496e-06, "loss": 5.9218, "step": 33930 }, { "epoch": 0.28, "learning_rate": 8.188257712542267e-06, "loss": 5.3253, "step": 33940 }, { "epoch": 0.28, "learning_rate": 8.187259689100288e-06, "loss": 5.996, "step": 33950 }, { "epoch": 0.28, "learning_rate": 8.186261451705557e-06, "loss": 5.7066, "step": 33960 }, { "epoch": 0.28, "learning_rate": 8.185263000425078e-06, "loss": 5.316, "step": 33970 }, { "epoch": 0.28, "learning_rate": 8.18426433532588e-06, "loss": 5.5278, "step": 33980 }, { "epoch": 0.28, "learning_rate": 8.183265456475e-06, "loss": 6.1389, "step": 33990 }, { "epoch": 0.28, "learning_rate": 8.182266363939486e-06, "loss": 5.79, "step": 34000 }, { "epoch": 0.28, "learning_rate": 8.181267057786408e-06, "loss": 5.8157, "step": 34010 }, { "epoch": 0.28, "learning_rate": 8.180267538082847e-06, "loss": 5.7646, "step": 34020 }, { "epoch": 0.28, "learning_rate": 8.179267804895897e-06, "loss": 5.3971, "step": 34030 }, { "epoch": 0.28, "learning_rate": 8.178267858292667e-06, "loss": 6.2966, "step": 34040 }, { "epoch": 0.28, "learning_rate": 8.177267698340286e-06, "loss": 5.8857, "step": 34050 }, { "epoch": 0.28, "learning_rate": 8.176267325105885e-06, "loss": 6.3153, "step": 34060 }, { "epoch": 0.28, "learning_rate": 8.175266738656621e-06, "loss": 5.982, "step": 34070 }, { "epoch": 0.28, "learning_rate": 8.174265939059658e-06, "loss": 5.9845, "step": 34080 }, { "epoch": 0.28, "learning_rate": 8.17326492638218e-06, "loss": 5.9848, "step": 34090 }, { "epoch": 0.28, "learning_rate": 8.172263700691384e-06, "loss": 5.5254, "step": 34100 }, { "epoch": 0.28, "learning_rate": 8.171262262054474e-06, "loss": 6.2372, "step": 34110 }, { "epoch": 0.28, "learning_rate": 8.170260610538678e-06, "loss": 5.9085, "step": 34120 }, { "epoch": 0.28, "learning_rate": 8.169258746211232e-06, "loss": 5.6583, "step": 34130 }, { "epoch": 0.28, "learning_rate": 8.168256669139392e-06, "loss": 6.249, "step": 34140 }, { "epoch": 0.28, "learning_rate": 8.167254379390424e-06, "loss": 6.1212, "step": 34150 }, { "epoch": 0.28, "learning_rate": 8.166251877031606e-06, "loss": 6.2213, "step": 34160 }, { "epoch": 0.28, "learning_rate": 8.165249162130238e-06, "loss": 5.8347, "step": 34170 }, { "epoch": 0.28, "learning_rate": 8.164246234753626e-06, "loss": 6.5223, "step": 34180 }, { "epoch": 0.28, "learning_rate": 8.163243094969096e-06, "loss": 6.2966, "step": 34190 }, { "epoch": 0.28, "learning_rate": 8.162239742843986e-06, "loss": 5.4536, "step": 34200 }, { "epoch": 0.28, "learning_rate": 8.161236178445647e-06, "loss": 5.8063, "step": 34210 }, { "epoch": 0.28, "learning_rate": 8.160232401841449e-06, "loss": 5.8875, "step": 34220 }, { "epoch": 0.28, "learning_rate": 8.15922841309877e-06, "loss": 6.1635, "step": 34230 }, { "epoch": 0.28, "learning_rate": 8.158224212285007e-06, "loss": 5.6576, "step": 34240 }, { "epoch": 0.28, "learning_rate": 8.157219799467566e-06, "loss": 6.1326, "step": 34250 }, { "epoch": 0.28, "learning_rate": 8.156215174713877e-06, "loss": 5.7258, "step": 34260 }, { "epoch": 0.28, "learning_rate": 8.155210338091373e-06, "loss": 5.9618, "step": 34270 }, { "epoch": 0.28, "learning_rate": 8.154205289667507e-06, "loss": 6.0896, "step": 34280 }, { "epoch": 0.28, "learning_rate": 8.153200029509747e-06, "loss": 6.1723, "step": 34290 }, { "epoch": 0.28, "learning_rate": 8.152194557685572e-06, "loss": 5.9427, "step": 34300 }, { "epoch": 0.28, "learning_rate": 8.15118887426248e-06, "loss": 5.4572, "step": 34310 }, { "epoch": 0.28, "learning_rate": 8.150182979307973e-06, "loss": 6.9564, "step": 34320 }, { "epoch": 0.28, "learning_rate": 8.149176872889581e-06, "loss": 5.5051, "step": 34330 }, { "epoch": 0.28, "learning_rate": 8.148170555074838e-06, "loss": 5.7802, "step": 34340 }, { "epoch": 0.28, "learning_rate": 8.147164025931297e-06, "loss": 5.9905, "step": 34350 }, { "epoch": 0.28, "learning_rate": 8.146157285526525e-06, "loss": 6.1021, "step": 34360 }, { "epoch": 0.28, "learning_rate": 8.145150333928099e-06, "loss": 5.8015, "step": 34370 }, { "epoch": 0.28, "learning_rate": 8.144143171203614e-06, "loss": 5.8082, "step": 34380 }, { "epoch": 0.28, "learning_rate": 8.14313579742068e-06, "loss": 6.1807, "step": 34390 }, { "epoch": 0.28, "learning_rate": 8.142128212646918e-06, "loss": 5.8979, "step": 34400 }, { "epoch": 0.28, "learning_rate": 8.141120416949963e-06, "loss": 5.5645, "step": 34410 }, { "epoch": 0.28, "learning_rate": 8.14011241039747e-06, "loss": 5.2419, "step": 34420 }, { "epoch": 0.28, "learning_rate": 8.1391041930571e-06, "loss": 6.2137, "step": 34430 }, { "epoch": 0.28, "learning_rate": 8.138095764996535e-06, "loss": 5.644, "step": 34440 }, { "epoch": 0.28, "learning_rate": 8.137087126283466e-06, "loss": 5.9477, "step": 34450 }, { "epoch": 0.28, "learning_rate": 8.136078276985601e-06, "loss": 6.0415, "step": 34460 }, { "epoch": 0.28, "learning_rate": 8.135069217170663e-06, "loss": 5.5472, "step": 34470 }, { "epoch": 0.28, "learning_rate": 8.134059946906387e-06, "loss": 5.8183, "step": 34480 }, { "epoch": 0.28, "learning_rate": 8.133050466260521e-06, "loss": 5.6647, "step": 34490 }, { "epoch": 0.28, "learning_rate": 8.132040775300833e-06, "loss": 5.0672, "step": 34500 }, { "epoch": 0.28, "learning_rate": 8.131030874095095e-06, "loss": 5.8081, "step": 34510 }, { "epoch": 0.28, "learning_rate": 8.130020762711103e-06, "loss": 6.304, "step": 34520 }, { "epoch": 0.28, "learning_rate": 8.129010441216666e-06, "loss": 5.4081, "step": 34530 }, { "epoch": 0.28, "learning_rate": 8.127999909679597e-06, "loss": 6.2428, "step": 34540 }, { "epoch": 0.28, "learning_rate": 8.126989168167736e-06, "loss": 5.5753, "step": 34550 }, { "epoch": 0.29, "learning_rate": 8.125978216748929e-06, "loss": 5.215, "step": 34560 }, { "epoch": 0.29, "learning_rate": 8.124967055491042e-06, "loss": 5.8906, "step": 34570 }, { "epoch": 0.29, "learning_rate": 8.123955684461948e-06, "loss": 6.0419, "step": 34580 }, { "epoch": 0.29, "learning_rate": 8.122944103729539e-06, "loss": 5.2361, "step": 34590 }, { "epoch": 0.29, "learning_rate": 8.12193231336172e-06, "loss": 5.9008, "step": 34600 }, { "epoch": 0.29, "learning_rate": 8.12092031342641e-06, "loss": 6.1338, "step": 34610 }, { "epoch": 0.29, "learning_rate": 8.119908103991541e-06, "loss": 6.4739, "step": 34620 }, { "epoch": 0.29, "learning_rate": 8.118895685125063e-06, "loss": 6.1209, "step": 34630 }, { "epoch": 0.29, "learning_rate": 8.117883056894935e-06, "loss": 5.6686, "step": 34640 }, { "epoch": 0.29, "learning_rate": 8.116870219369133e-06, "loss": 6.0164, "step": 34650 }, { "epoch": 0.29, "learning_rate": 8.115857172615644e-06, "loss": 5.8502, "step": 34660 }, { "epoch": 0.29, "learning_rate": 8.114843916702473e-06, "loss": 5.9603, "step": 34670 }, { "epoch": 0.29, "learning_rate": 8.113830451697638e-06, "loss": 5.6434, "step": 34680 }, { "epoch": 0.29, "learning_rate": 8.11281677766917e-06, "loss": 6.1244, "step": 34690 }, { "epoch": 0.29, "learning_rate": 8.111802894685113e-06, "loss": 5.1726, "step": 34700 }, { "epoch": 0.29, "learning_rate": 8.110788802813529e-06, "loss": 5.8417, "step": 34710 }, { "epoch": 0.29, "learning_rate": 8.10977450212249e-06, "loss": 5.19, "step": 34720 }, { "epoch": 0.29, "learning_rate": 8.10875999268008e-06, "loss": 5.4444, "step": 34730 }, { "epoch": 0.29, "learning_rate": 8.107745274554407e-06, "loss": 6.3707, "step": 34740 }, { "epoch": 0.29, "learning_rate": 8.106730347813584e-06, "loss": 5.6426, "step": 34750 }, { "epoch": 0.29, "learning_rate": 8.105715212525737e-06, "loss": 6.2456, "step": 34760 }, { "epoch": 0.29, "learning_rate": 8.104699868759013e-06, "loss": 5.9309, "step": 34770 }, { "epoch": 0.29, "learning_rate": 8.10368431658157e-06, "loss": 5.1171, "step": 34780 }, { "epoch": 0.29, "learning_rate": 8.102668556061579e-06, "loss": 6.2731, "step": 34790 }, { "epoch": 0.29, "learning_rate": 8.101652587267224e-06, "loss": 6.0819, "step": 34800 }, { "epoch": 0.29, "learning_rate": 8.100636410266705e-06, "loss": 5.7309, "step": 34810 }, { "epoch": 0.29, "learning_rate": 8.099620025128234e-06, "loss": 6.4177, "step": 34820 }, { "epoch": 0.29, "learning_rate": 8.098603431920042e-06, "loss": 5.7523, "step": 34830 }, { "epoch": 0.29, "learning_rate": 8.097586630710369e-06, "loss": 5.7629, "step": 34840 }, { "epoch": 0.29, "learning_rate": 8.096569621567468e-06, "loss": 6.2384, "step": 34850 }, { "epoch": 0.29, "learning_rate": 8.095552404559608e-06, "loss": 5.351, "step": 34860 }, { "epoch": 0.29, "learning_rate": 8.094534979755078e-06, "loss": 5.3183, "step": 34870 }, { "epoch": 0.29, "learning_rate": 8.093517347222169e-06, "loss": 5.6695, "step": 34880 }, { "epoch": 0.29, "learning_rate": 8.092499507029196e-06, "loss": 5.3809, "step": 34890 }, { "epoch": 0.29, "learning_rate": 8.091481459244481e-06, "loss": 5.4392, "step": 34900 }, { "epoch": 0.29, "learning_rate": 8.090463203936366e-06, "loss": 5.5665, "step": 34910 }, { "epoch": 0.29, "learning_rate": 8.0894447411732e-06, "loss": 5.8573, "step": 34920 }, { "epoch": 0.29, "learning_rate": 8.088426071023355e-06, "loss": 6.1008, "step": 34930 }, { "epoch": 0.29, "learning_rate": 8.087407193555208e-06, "loss": 5.653, "step": 34940 }, { "epoch": 0.29, "learning_rate": 8.086388108837156e-06, "loss": 5.8266, "step": 34950 }, { "epoch": 0.29, "learning_rate": 8.085368816937604e-06, "loss": 5.813, "step": 34960 }, { "epoch": 0.29, "learning_rate": 8.08434931792498e-06, "loss": 6.025, "step": 34970 }, { "epoch": 0.29, "learning_rate": 8.083329611867714e-06, "loss": 5.6053, "step": 34980 }, { "epoch": 0.29, "learning_rate": 8.08230969883426e-06, "loss": 5.8554, "step": 34990 }, { "epoch": 0.29, "learning_rate": 8.081289578893084e-06, "loss": 5.9654, "step": 35000 }, { "epoch": 0.29, "learning_rate": 8.08026925211266e-06, "loss": 5.4976, "step": 35010 }, { "epoch": 0.29, "learning_rate": 8.079248718561483e-06, "loss": 6.4285, "step": 35020 }, { "epoch": 0.29, "learning_rate": 8.078227978308057e-06, "loss": 5.3715, "step": 35030 }, { "epoch": 0.29, "learning_rate": 8.077207031420901e-06, "loss": 5.4702, "step": 35040 }, { "epoch": 0.29, "learning_rate": 8.076185877968552e-06, "loss": 5.2696, "step": 35050 }, { "epoch": 0.29, "learning_rate": 8.075164518019557e-06, "loss": 5.6557, "step": 35060 }, { "epoch": 0.29, "learning_rate": 8.074142951642474e-06, "loss": 5.8271, "step": 35070 }, { "epoch": 0.29, "learning_rate": 8.07312117890588e-06, "loss": 6.0217, "step": 35080 }, { "epoch": 0.29, "learning_rate": 8.072099199878363e-06, "loss": 5.6132, "step": 35090 }, { "epoch": 0.29, "learning_rate": 8.07107701462853e-06, "loss": 5.9201, "step": 35100 }, { "epoch": 0.29, "learning_rate": 8.070054623224993e-06, "loss": 6.0165, "step": 35110 }, { "epoch": 0.29, "learning_rate": 8.069032025736382e-06, "loss": 5.4749, "step": 35120 }, { "epoch": 0.29, "learning_rate": 8.068009222231348e-06, "loss": 5.3654, "step": 35130 }, { "epoch": 0.29, "learning_rate": 8.066986212778543e-06, "loss": 6.0734, "step": 35140 }, { "epoch": 0.29, "learning_rate": 8.065962997446642e-06, "loss": 5.1346, "step": 35150 }, { "epoch": 0.29, "learning_rate": 8.064939576304327e-06, "loss": 6.2202, "step": 35160 }, { "epoch": 0.29, "learning_rate": 8.063915949420303e-06, "loss": 5.7449, "step": 35170 }, { "epoch": 0.29, "learning_rate": 8.062892116863281e-06, "loss": 5.7808, "step": 35180 }, { "epoch": 0.29, "learning_rate": 8.061868078701987e-06, "loss": 6.3101, "step": 35190 }, { "epoch": 0.29, "learning_rate": 8.060843835005166e-06, "loss": 6.1385, "step": 35200 }, { "epoch": 0.29, "learning_rate": 8.059819385841569e-06, "loss": 6.0249, "step": 35210 }, { "epoch": 0.29, "learning_rate": 8.058794731279966e-06, "loss": 6.0532, "step": 35220 }, { "epoch": 0.29, "learning_rate": 8.057769871389139e-06, "loss": 5.4055, "step": 35230 }, { "epoch": 0.29, "learning_rate": 8.056744806237886e-06, "loss": 5.107, "step": 35240 }, { "epoch": 0.29, "learning_rate": 8.055719535895015e-06, "loss": 5.1059, "step": 35250 }, { "epoch": 0.29, "learning_rate": 8.054694060429351e-06, "loss": 5.053, "step": 35260 }, { "epoch": 0.29, "learning_rate": 8.053668379909731e-06, "loss": 6.2579, "step": 35270 }, { "epoch": 0.29, "learning_rate": 8.05264249440501e-06, "loss": 6.1839, "step": 35280 }, { "epoch": 0.29, "learning_rate": 8.051616403984046e-06, "loss": 5.8778, "step": 35290 }, { "epoch": 0.29, "learning_rate": 8.050590108715723e-06, "loss": 6.0657, "step": 35300 }, { "epoch": 0.29, "learning_rate": 8.049563608668933e-06, "loss": 5.6857, "step": 35310 }, { "epoch": 0.29, "learning_rate": 8.048536903912581e-06, "loss": 6.1076, "step": 35320 }, { "epoch": 0.29, "learning_rate": 8.047509994515588e-06, "loss": 5.6529, "step": 35330 }, { "epoch": 0.29, "learning_rate": 8.046482880546887e-06, "loss": 4.9624, "step": 35340 }, { "epoch": 0.29, "learning_rate": 8.045455562075429e-06, "loss": 5.8665, "step": 35350 }, { "epoch": 0.29, "learning_rate": 8.04442803917017e-06, "loss": 6.3879, "step": 35360 }, { "epoch": 0.29, "learning_rate": 8.043400311900088e-06, "loss": 5.8495, "step": 35370 }, { "epoch": 0.29, "learning_rate": 8.042372380334171e-06, "loss": 6.3694, "step": 35380 }, { "epoch": 0.29, "learning_rate": 8.041344244541423e-06, "loss": 5.6089, "step": 35390 }, { "epoch": 0.29, "learning_rate": 8.040315904590858e-06, "loss": 5.9765, "step": 35400 }, { "epoch": 0.29, "learning_rate": 8.03928736055151e-06, "loss": 5.711, "step": 35410 }, { "epoch": 0.29, "learning_rate": 8.038258612492417e-06, "loss": 6.0806, "step": 35420 }, { "epoch": 0.29, "learning_rate": 8.037229660482639e-06, "loss": 5.7157, "step": 35430 }, { "epoch": 0.29, "learning_rate": 8.036200504591245e-06, "loss": 5.6026, "step": 35440 }, { "epoch": 0.29, "learning_rate": 8.035171144887323e-06, "loss": 5.6896, "step": 35450 }, { "epoch": 0.29, "learning_rate": 8.034141581439972e-06, "loss": 5.9523, "step": 35460 }, { "epoch": 0.29, "learning_rate": 8.0331118143183e-06, "loss": 5.9868, "step": 35470 }, { "epoch": 0.29, "learning_rate": 8.032081843591432e-06, "loss": 6.2481, "step": 35480 }, { "epoch": 0.29, "learning_rate": 8.031051669328514e-06, "loss": 5.7992, "step": 35490 }, { "epoch": 0.29, "learning_rate": 8.030021291598694e-06, "loss": 5.5293, "step": 35500 }, { "epoch": 0.29, "learning_rate": 8.028990710471136e-06, "loss": 6.0092, "step": 35510 }, { "epoch": 0.29, "learning_rate": 8.027959926015027e-06, "loss": 6.5339, "step": 35520 }, { "epoch": 0.29, "learning_rate": 8.026928938299556e-06, "loss": 5.3089, "step": 35530 }, { "epoch": 0.29, "learning_rate": 8.025897747393935e-06, "loss": 5.6462, "step": 35540 }, { "epoch": 0.29, "learning_rate": 8.02486635336738e-06, "loss": 5.7826, "step": 35550 }, { "epoch": 0.29, "learning_rate": 8.023834756289131e-06, "loss": 5.4208, "step": 35560 }, { "epoch": 0.29, "learning_rate": 8.022802956228433e-06, "loss": 5.6226, "step": 35570 }, { "epoch": 0.29, "learning_rate": 8.021770953254548e-06, "loss": 5.5406, "step": 35580 }, { "epoch": 0.29, "learning_rate": 8.020738747436757e-06, "loss": 5.7667, "step": 35590 }, { "epoch": 0.29, "learning_rate": 8.019706338844341e-06, "loss": 6.2809, "step": 35600 }, { "epoch": 0.29, "learning_rate": 8.018673727546608e-06, "loss": 5.0722, "step": 35610 }, { "epoch": 0.29, "learning_rate": 8.017640913612876e-06, "loss": 6.0921, "step": 35620 }, { "epoch": 0.29, "learning_rate": 8.016607897112474e-06, "loss": 6.41, "step": 35630 }, { "epoch": 0.29, "learning_rate": 8.015574678114742e-06, "loss": 5.6887, "step": 35640 }, { "epoch": 0.29, "learning_rate": 8.014541256689042e-06, "loss": 6.3468, "step": 35650 }, { "epoch": 0.29, "learning_rate": 8.013507632904744e-06, "loss": 5.2442, "step": 35660 }, { "epoch": 0.29, "learning_rate": 8.012473806831232e-06, "loss": 5.0792, "step": 35670 }, { "epoch": 0.29, "learning_rate": 8.011439778537903e-06, "loss": 5.3073, "step": 35680 }, { "epoch": 0.29, "learning_rate": 8.01040554809417e-06, "loss": 6.1627, "step": 35690 }, { "epoch": 0.29, "learning_rate": 8.009371115569459e-06, "loss": 5.8109, "step": 35700 }, { "epoch": 0.29, "learning_rate": 8.008336481033207e-06, "loss": 5.9379, "step": 35710 }, { "epoch": 0.29, "learning_rate": 8.00730164455487e-06, "loss": 5.9386, "step": 35720 }, { "epoch": 0.29, "learning_rate": 8.006266606203908e-06, "loss": 5.777, "step": 35730 }, { "epoch": 0.29, "learning_rate": 8.005231366049805e-06, "loss": 5.6304, "step": 35740 }, { "epoch": 0.29, "learning_rate": 8.004195924162056e-06, "loss": 5.1602, "step": 35750 }, { "epoch": 0.29, "learning_rate": 8.003160280610163e-06, "loss": 6.376, "step": 35760 }, { "epoch": 0.29, "learning_rate": 8.002124435463646e-06, "loss": 6.1083, "step": 35770 }, { "epoch": 0.3, "learning_rate": 8.001088388792042e-06, "loss": 5.9307, "step": 35780 }, { "epoch": 0.3, "learning_rate": 8.000052140664898e-06, "loss": 5.6811, "step": 35790 }, { "epoch": 0.3, "learning_rate": 7.999015691151772e-06, "loss": 6.4003, "step": 35800 }, { "epoch": 0.3, "learning_rate": 7.997979040322239e-06, "loss": 6.0256, "step": 35810 }, { "epoch": 0.3, "learning_rate": 7.99694218824589e-06, "loss": 5.8122, "step": 35820 }, { "epoch": 0.3, "learning_rate": 7.995905134992321e-06, "loss": 5.5551, "step": 35830 }, { "epoch": 0.3, "learning_rate": 7.994867880631151e-06, "loss": 6.4172, "step": 35840 }, { "epoch": 0.3, "learning_rate": 7.993830425232008e-06, "loss": 5.6266, "step": 35850 }, { "epoch": 0.3, "learning_rate": 7.992792768864531e-06, "loss": 5.1055, "step": 35860 }, { "epoch": 0.3, "learning_rate": 7.991754911598377e-06, "loss": 5.4398, "step": 35870 }, { "epoch": 0.3, "learning_rate": 7.990716853503215e-06, "loss": 5.189, "step": 35880 }, { "epoch": 0.3, "learning_rate": 7.989678594648727e-06, "loss": 5.9297, "step": 35890 }, { "epoch": 0.3, "learning_rate": 7.988640135104609e-06, "loss": 5.6624, "step": 35900 }, { "epoch": 0.3, "learning_rate": 7.987601474940568e-06, "loss": 5.7825, "step": 35910 }, { "epoch": 0.3, "learning_rate": 7.98656261422633e-06, "loss": 5.2879, "step": 35920 }, { "epoch": 0.3, "learning_rate": 7.98552355303163e-06, "loss": 5.5986, "step": 35930 }, { "epoch": 0.3, "learning_rate": 7.984484291426216e-06, "loss": 6.2317, "step": 35940 }, { "epoch": 0.3, "learning_rate": 7.983444829479854e-06, "loss": 6.1611, "step": 35950 }, { "epoch": 0.3, "learning_rate": 7.982405167262317e-06, "loss": 5.9168, "step": 35960 }, { "epoch": 0.3, "learning_rate": 7.981365304843397e-06, "loss": 5.9327, "step": 35970 }, { "epoch": 0.3, "learning_rate": 7.980325242292898e-06, "loss": 5.4992, "step": 35980 }, { "epoch": 0.3, "learning_rate": 7.979284979680636e-06, "loss": 5.212, "step": 35990 }, { "epoch": 0.3, "learning_rate": 7.978244517076438e-06, "loss": 5.8304, "step": 36000 }, { "epoch": 0.3, "learning_rate": 7.977203854550153e-06, "loss": 6.0946, "step": 36010 }, { "epoch": 0.3, "learning_rate": 7.976162992171634e-06, "loss": 6.3479, "step": 36020 }, { "epoch": 0.3, "learning_rate": 7.975121930010755e-06, "loss": 6.117, "step": 36030 }, { "epoch": 0.3, "learning_rate": 7.974080668137396e-06, "loss": 5.4586, "step": 36040 }, { "epoch": 0.3, "learning_rate": 7.973039206621458e-06, "loss": 5.1038, "step": 36050 }, { "epoch": 0.3, "learning_rate": 7.97199754553285e-06, "loss": 5.9724, "step": 36060 }, { "epoch": 0.3, "learning_rate": 7.970955684941496e-06, "loss": 5.5447, "step": 36070 }, { "epoch": 0.3, "learning_rate": 7.969913624917334e-06, "loss": 5.6432, "step": 36080 }, { "epoch": 0.3, "learning_rate": 7.968871365530312e-06, "loss": 6.2977, "step": 36090 }, { "epoch": 0.3, "learning_rate": 7.967828906850399e-06, "loss": 6.0608, "step": 36100 }, { "epoch": 0.3, "learning_rate": 7.96678624894757e-06, "loss": 5.8804, "step": 36110 }, { "epoch": 0.3, "learning_rate": 7.965743391891817e-06, "loss": 5.9494, "step": 36120 }, { "epoch": 0.3, "learning_rate": 7.964700335753144e-06, "loss": 5.8791, "step": 36130 }, { "epoch": 0.3, "learning_rate": 7.963657080601569e-06, "loss": 5.6163, "step": 36140 }, { "epoch": 0.3, "learning_rate": 7.962613626507121e-06, "loss": 5.791, "step": 36150 }, { "epoch": 0.3, "learning_rate": 7.961569973539846e-06, "loss": 5.5594, "step": 36160 }, { "epoch": 0.3, "learning_rate": 7.960526121769803e-06, "loss": 6.0587, "step": 36170 }, { "epoch": 0.3, "learning_rate": 7.959482071267062e-06, "loss": 6.3187, "step": 36180 }, { "epoch": 0.3, "learning_rate": 7.958437822101706e-06, "loss": 5.82, "step": 36190 }, { "epoch": 0.3, "learning_rate": 7.957393374343836e-06, "loss": 6.0267, "step": 36200 }, { "epoch": 0.3, "learning_rate": 7.956348728063561e-06, "loss": 5.8792, "step": 36210 }, { "epoch": 0.3, "learning_rate": 7.955303883331007e-06, "loss": 5.8762, "step": 36220 }, { "epoch": 0.3, "learning_rate": 7.954258840216308e-06, "loss": 4.9007, "step": 36230 }, { "epoch": 0.3, "learning_rate": 7.953213598789623e-06, "loss": 5.7985, "step": 36240 }, { "epoch": 0.3, "learning_rate": 7.952168159121108e-06, "loss": 5.7839, "step": 36250 }, { "epoch": 0.3, "learning_rate": 7.951122521280945e-06, "loss": 6.5567, "step": 36260 }, { "epoch": 0.3, "learning_rate": 7.950076685339323e-06, "loss": 6.106, "step": 36270 }, { "epoch": 0.3, "learning_rate": 7.949030651366448e-06, "loss": 5.8883, "step": 36280 }, { "epoch": 0.3, "learning_rate": 7.947984419432537e-06, "loss": 5.9423, "step": 36290 }, { "epoch": 0.3, "learning_rate": 7.946937989607821e-06, "loss": 5.7427, "step": 36300 }, { "epoch": 0.3, "learning_rate": 7.945891361962545e-06, "loss": 5.7921, "step": 36310 }, { "epoch": 0.3, "learning_rate": 7.944844536566967e-06, "loss": 5.6562, "step": 36320 }, { "epoch": 0.3, "learning_rate": 7.943797513491356e-06, "loss": 5.1672, "step": 36330 }, { "epoch": 0.3, "learning_rate": 7.942750292805996e-06, "loss": 6.0735, "step": 36340 }, { "epoch": 0.3, "learning_rate": 7.941702874581184e-06, "loss": 6.0619, "step": 36350 }, { "epoch": 0.3, "learning_rate": 7.940655258887232e-06, "loss": 5.9892, "step": 36360 }, { "epoch": 0.3, "learning_rate": 7.939607445794464e-06, "loss": 5.7461, "step": 36370 }, { "epoch": 0.3, "learning_rate": 7.938559435373217e-06, "loss": 5.7397, "step": 36380 }, { "epoch": 0.3, "learning_rate": 7.937511227693839e-06, "loss": 5.9411, "step": 36390 }, { "epoch": 0.3, "learning_rate": 7.936462822826696e-06, "loss": 6.3923, "step": 36400 }, { "epoch": 0.3, "learning_rate": 7.935414220842164e-06, "loss": 5.7649, "step": 36410 }, { "epoch": 0.3, "learning_rate": 7.934365421810633e-06, "loss": 5.8543, "step": 36420 }, { "epoch": 0.3, "learning_rate": 7.933316425802507e-06, "loss": 5.9961, "step": 36430 }, { "epoch": 0.3, "learning_rate": 7.9322672328882e-06, "loss": 5.5432, "step": 36440 }, { "epoch": 0.3, "learning_rate": 7.931217843138143e-06, "loss": 5.5024, "step": 36450 }, { "epoch": 0.3, "learning_rate": 7.930168256622782e-06, "loss": 5.9793, "step": 36460 }, { "epoch": 0.3, "learning_rate": 7.92911847341257e-06, "loss": 5.8896, "step": 36470 }, { "epoch": 0.3, "learning_rate": 7.928068493577976e-06, "loss": 6.4144, "step": 36480 }, { "epoch": 0.3, "learning_rate": 7.927018317189483e-06, "loss": 5.7431, "step": 36490 }, { "epoch": 0.3, "learning_rate": 7.925967944317588e-06, "loss": 5.5088, "step": 36500 }, { "epoch": 0.3, "learning_rate": 7.924917375032796e-06, "loss": 6.1647, "step": 36510 }, { "epoch": 0.3, "learning_rate": 7.923866609405636e-06, "loss": 6.1283, "step": 36520 }, { "epoch": 0.3, "learning_rate": 7.922815647506638e-06, "loss": 5.7741, "step": 36530 }, { "epoch": 0.3, "learning_rate": 7.921764489406351e-06, "loss": 6.0695, "step": 36540 }, { "epoch": 0.3, "learning_rate": 7.920713135175338e-06, "loss": 5.378, "step": 36550 }, { "epoch": 0.3, "learning_rate": 7.919661584884171e-06, "loss": 6.0021, "step": 36560 }, { "epoch": 0.3, "learning_rate": 7.918609838603442e-06, "loss": 5.2851, "step": 36570 }, { "epoch": 0.3, "learning_rate": 7.91755789640375e-06, "loss": 5.6004, "step": 36580 }, { "epoch": 0.3, "learning_rate": 7.91650575835571e-06, "loss": 6.268, "step": 36590 }, { "epoch": 0.3, "learning_rate": 7.915453424529946e-06, "loss": 5.8782, "step": 36600 }, { "epoch": 0.3, "learning_rate": 7.914400894997103e-06, "loss": 5.3896, "step": 36610 }, { "epoch": 0.3, "learning_rate": 7.913348169827833e-06, "loss": 5.6353, "step": 36620 }, { "epoch": 0.3, "learning_rate": 7.912295249092803e-06, "loss": 5.919, "step": 36630 }, { "epoch": 0.3, "learning_rate": 7.911242132862693e-06, "loss": 5.9656, "step": 36640 }, { "epoch": 0.3, "learning_rate": 7.910188821208195e-06, "loss": 5.7557, "step": 36650 }, { "epoch": 0.3, "learning_rate": 7.909135314200015e-06, "loss": 5.9494, "step": 36660 }, { "epoch": 0.3, "learning_rate": 7.908081611908876e-06, "loss": 6.0964, "step": 36670 }, { "epoch": 0.3, "learning_rate": 7.907027714405504e-06, "loss": 6.0058, "step": 36680 }, { "epoch": 0.3, "learning_rate": 7.90597362176065e-06, "loss": 5.448, "step": 36690 }, { "epoch": 0.3, "learning_rate": 7.904919334045071e-06, "loss": 5.9212, "step": 36700 }, { "epoch": 0.3, "learning_rate": 7.903864851329538e-06, "loss": 5.5198, "step": 36710 }, { "epoch": 0.3, "learning_rate": 7.902810173684835e-06, "loss": 5.8831, "step": 36720 }, { "epoch": 0.3, "learning_rate": 7.901755301181761e-06, "loss": 5.8437, "step": 36730 }, { "epoch": 0.3, "learning_rate": 7.900700233891127e-06, "loss": 5.4683, "step": 36740 }, { "epoch": 0.3, "learning_rate": 7.89964497188376e-06, "loss": 5.9745, "step": 36750 }, { "epoch": 0.3, "learning_rate": 7.898589515230491e-06, "loss": 6.3095, "step": 36760 }, { "epoch": 0.3, "learning_rate": 7.897533864002175e-06, "loss": 6.0184, "step": 36770 }, { "epoch": 0.3, "learning_rate": 7.896478018269672e-06, "loss": 6.0476, "step": 36780 }, { "epoch": 0.3, "learning_rate": 7.89542197810386e-06, "loss": 5.7105, "step": 36790 }, { "epoch": 0.3, "learning_rate": 7.89436574357563e-06, "loss": 5.8915, "step": 36800 }, { "epoch": 0.3, "learning_rate": 7.893309314755879e-06, "loss": 5.597, "step": 36810 }, { "epoch": 0.3, "learning_rate": 7.892252691715527e-06, "loss": 6.7144, "step": 36820 }, { "epoch": 0.3, "learning_rate": 7.891195874525502e-06, "loss": 6.3941, "step": 36830 }, { "epoch": 0.3, "learning_rate": 7.890138863256745e-06, "loss": 5.1622, "step": 36840 }, { "epoch": 0.3, "learning_rate": 7.889081657980211e-06, "loss": 5.5468, "step": 36850 }, { "epoch": 0.3, "learning_rate": 7.888024258766865e-06, "loss": 5.2703, "step": 36860 }, { "epoch": 0.3, "learning_rate": 7.88696666568769e-06, "loss": 6.4302, "step": 36870 }, { "epoch": 0.3, "learning_rate": 7.88590887881368e-06, "loss": 5.8343, "step": 36880 }, { "epoch": 0.3, "learning_rate": 7.88485089821584e-06, "loss": 6.7095, "step": 36890 }, { "epoch": 0.3, "learning_rate": 7.88379272396519e-06, "loss": 5.6456, "step": 36900 }, { "epoch": 0.3, "learning_rate": 7.882734356132761e-06, "loss": 5.8303, "step": 36910 }, { "epoch": 0.3, "learning_rate": 7.881675794789603e-06, "loss": 5.7919, "step": 36920 }, { "epoch": 0.3, "learning_rate": 7.88061704000677e-06, "loss": 5.8358, "step": 36930 }, { "epoch": 0.3, "learning_rate": 7.879558091855338e-06, "loss": 5.3964, "step": 36940 }, { "epoch": 0.3, "learning_rate": 7.878498950406385e-06, "loss": 6.2264, "step": 36950 }, { "epoch": 0.3, "learning_rate": 7.877439615731014e-06, "loss": 5.7889, "step": 36960 }, { "epoch": 0.3, "learning_rate": 7.876380087900338e-06, "loss": 6.9241, "step": 36970 }, { "epoch": 0.3, "learning_rate": 7.875320366985473e-06, "loss": 5.4378, "step": 36980 }, { "epoch": 0.31, "learning_rate": 7.874260453057558e-06, "loss": 5.6956, "step": 36990 }, { "epoch": 0.31, "learning_rate": 7.873200346187744e-06, "loss": 5.9909, "step": 37000 }, { "epoch": 0.31, "learning_rate": 7.872140046447192e-06, "loss": 5.746, "step": 37010 }, { "epoch": 0.31, "learning_rate": 7.871079553907078e-06, "loss": 5.8479, "step": 37020 }, { "epoch": 0.31, "learning_rate": 7.870018868638591e-06, "loss": 5.7882, "step": 37030 }, { "epoch": 0.31, "learning_rate": 7.868957990712931e-06, "loss": 5.7529, "step": 37040 }, { "epoch": 0.31, "learning_rate": 7.867896920201312e-06, "loss": 5.2994, "step": 37050 }, { "epoch": 0.31, "learning_rate": 7.86683565717496e-06, "loss": 5.4185, "step": 37060 }, { "epoch": 0.31, "learning_rate": 7.865774201705117e-06, "loss": 5.9032, "step": 37070 }, { "epoch": 0.31, "learning_rate": 7.864712553863034e-06, "loss": 5.7519, "step": 37080 }, { "epoch": 0.31, "learning_rate": 7.863650713719979e-06, "loss": 5.4276, "step": 37090 }, { "epoch": 0.31, "learning_rate": 7.86258868134723e-06, "loss": 6.1194, "step": 37100 }, { "epoch": 0.31, "learning_rate": 7.861526456816078e-06, "loss": 6.0178, "step": 37110 }, { "epoch": 0.31, "learning_rate": 7.860464040197826e-06, "loss": 6.501, "step": 37120 }, { "epoch": 0.31, "learning_rate": 7.859401431563792e-06, "loss": 5.758, "step": 37130 }, { "epoch": 0.31, "learning_rate": 7.85833863098531e-06, "loss": 5.6339, "step": 37140 }, { "epoch": 0.31, "learning_rate": 7.857275638533717e-06, "loss": 6.1676, "step": 37150 }, { "epoch": 0.31, "learning_rate": 7.856212454280374e-06, "loss": 5.4288, "step": 37160 }, { "epoch": 0.31, "learning_rate": 7.855149078296649e-06, "loss": 6.3287, "step": 37170 }, { "epoch": 0.31, "learning_rate": 7.854085510653921e-06, "loss": 5.4784, "step": 37180 }, { "epoch": 0.31, "learning_rate": 7.853021751423587e-06, "loss": 5.6148, "step": 37190 }, { "epoch": 0.31, "learning_rate": 7.851957800677054e-06, "loss": 5.893, "step": 37200 }, { "epoch": 0.31, "learning_rate": 7.850893658485743e-06, "loss": 5.5093, "step": 37210 }, { "epoch": 0.31, "learning_rate": 7.849829324921086e-06, "loss": 5.6381, "step": 37220 }, { "epoch": 0.31, "learning_rate": 7.84876480005453e-06, "loss": 5.8014, "step": 37230 }, { "epoch": 0.31, "learning_rate": 7.847700083957532e-06, "loss": 5.4829, "step": 37240 }, { "epoch": 0.31, "learning_rate": 7.846635176701567e-06, "loss": 5.4089, "step": 37250 }, { "epoch": 0.31, "learning_rate": 7.845570078358116e-06, "loss": 5.7599, "step": 37260 }, { "epoch": 0.31, "learning_rate": 7.84450478899868e-06, "loss": 5.6362, "step": 37270 }, { "epoch": 0.31, "learning_rate": 7.843439308694766e-06, "loss": 5.6385, "step": 37280 }, { "epoch": 0.31, "learning_rate": 7.8423736375179e-06, "loss": 6.2706, "step": 37290 }, { "epoch": 0.31, "learning_rate": 7.841307775539616e-06, "loss": 5.5633, "step": 37300 }, { "epoch": 0.31, "learning_rate": 7.840241722831461e-06, "loss": 5.8003, "step": 37310 }, { "epoch": 0.31, "learning_rate": 7.839175479464998e-06, "loss": 5.5779, "step": 37320 }, { "epoch": 0.31, "learning_rate": 7.838109045511804e-06, "loss": 5.4906, "step": 37330 }, { "epoch": 0.31, "learning_rate": 7.837042421043462e-06, "loss": 6.2033, "step": 37340 }, { "epoch": 0.31, "learning_rate": 7.835975606131574e-06, "loss": 6.2062, "step": 37350 }, { "epoch": 0.31, "learning_rate": 7.834908600847751e-06, "loss": 5.7145, "step": 37360 }, { "epoch": 0.31, "learning_rate": 7.83384140526362e-06, "loss": 6.2293, "step": 37370 }, { "epoch": 0.31, "learning_rate": 7.832774019450819e-06, "loss": 5.7647, "step": 37380 }, { "epoch": 0.31, "learning_rate": 7.831706443480997e-06, "loss": 5.8289, "step": 37390 }, { "epoch": 0.31, "learning_rate": 7.830638677425821e-06, "loss": 5.873, "step": 37400 }, { "epoch": 0.31, "learning_rate": 7.829570721356965e-06, "loss": 6.0151, "step": 37410 }, { "epoch": 0.31, "learning_rate": 7.828502575346118e-06, "loss": 5.9804, "step": 37420 }, { "epoch": 0.31, "learning_rate": 7.827434239464983e-06, "loss": 5.6401, "step": 37430 }, { "epoch": 0.31, "learning_rate": 7.826365713785276e-06, "loss": 6.2588, "step": 37440 }, { "epoch": 0.31, "learning_rate": 7.825296998378723e-06, "loss": 6.0578, "step": 37450 }, { "epoch": 0.31, "learning_rate": 7.824228093317061e-06, "loss": 6.3436, "step": 37460 }, { "epoch": 0.31, "learning_rate": 7.823158998672048e-06, "loss": 5.3612, "step": 37470 }, { "epoch": 0.31, "learning_rate": 7.82208971451545e-06, "loss": 5.8186, "step": 37480 }, { "epoch": 0.31, "learning_rate": 7.821020240919042e-06, "loss": 5.7686, "step": 37490 }, { "epoch": 0.31, "learning_rate": 7.819950577954613e-06, "loss": 5.7917, "step": 37500 }, { "epoch": 0.31, "learning_rate": 7.818880725693974e-06, "loss": 5.7956, "step": 37510 }, { "epoch": 0.31, "learning_rate": 7.817810684208934e-06, "loss": 5.7755, "step": 37520 }, { "epoch": 0.31, "learning_rate": 7.816740453571327e-06, "loss": 6.5172, "step": 37530 }, { "epoch": 0.31, "learning_rate": 7.815670033852995e-06, "loss": 5.7149, "step": 37540 }, { "epoch": 0.31, "learning_rate": 7.81459942512579e-06, "loss": 5.5722, "step": 37550 }, { "epoch": 0.31, "learning_rate": 7.813528627461583e-06, "loss": 5.5986, "step": 37560 }, { "epoch": 0.31, "learning_rate": 7.812457640932248e-06, "loss": 5.604, "step": 37570 }, { "epoch": 0.31, "learning_rate": 7.811386465609682e-06, "loss": 5.952, "step": 37580 }, { "epoch": 0.31, "learning_rate": 7.81031510156579e-06, "loss": 6.1049, "step": 37590 }, { "epoch": 0.31, "learning_rate": 7.809243548872487e-06, "loss": 5.8343, "step": 37600 }, { "epoch": 0.31, "learning_rate": 7.808171807601708e-06, "loss": 5.5382, "step": 37610 }, { "epoch": 0.31, "learning_rate": 7.807099877825394e-06, "loss": 6.0233, "step": 37620 }, { "epoch": 0.31, "learning_rate": 7.806027759615501e-06, "loss": 5.8387, "step": 37630 }, { "epoch": 0.31, "learning_rate": 7.804955453043999e-06, "loss": 5.2911, "step": 37640 }, { "epoch": 0.31, "learning_rate": 7.803882958182868e-06, "loss": 5.5954, "step": 37650 }, { "epoch": 0.31, "learning_rate": 7.8028102751041e-06, "loss": 5.7665, "step": 37660 }, { "epoch": 0.31, "learning_rate": 7.801737403879705e-06, "loss": 5.8385, "step": 37670 }, { "epoch": 0.31, "learning_rate": 7.800664344581702e-06, "loss": 5.5355, "step": 37680 }, { "epoch": 0.31, "learning_rate": 7.79959109728212e-06, "loss": 5.3721, "step": 37690 }, { "epoch": 0.31, "learning_rate": 7.798517662053005e-06, "loss": 6.5357, "step": 37700 }, { "epoch": 0.31, "learning_rate": 7.797444038966414e-06, "loss": 5.4599, "step": 37710 }, { "epoch": 0.31, "learning_rate": 7.796370228094416e-06, "loss": 5.8686, "step": 37720 }, { "epoch": 0.31, "learning_rate": 7.795296229509093e-06, "loss": 5.8253, "step": 37730 }, { "epoch": 0.31, "learning_rate": 7.794222043282543e-06, "loss": 5.9906, "step": 37740 }, { "epoch": 0.31, "learning_rate": 7.79314766948687e-06, "loss": 5.9131, "step": 37750 }, { "epoch": 0.31, "learning_rate": 7.792073108194196e-06, "loss": 6.3, "step": 37760 }, { "epoch": 0.31, "learning_rate": 7.790998359476651e-06, "loss": 6.4161, "step": 37770 }, { "epoch": 0.31, "learning_rate": 7.78992342340638e-06, "loss": 5.2382, "step": 37780 }, { "epoch": 0.31, "learning_rate": 7.788848300055545e-06, "loss": 5.5474, "step": 37790 }, { "epoch": 0.31, "learning_rate": 7.787772989496315e-06, "loss": 5.7667, "step": 37800 }, { "epoch": 0.31, "learning_rate": 7.786697491800871e-06, "loss": 5.4252, "step": 37810 }, { "epoch": 0.31, "learning_rate": 7.78562180704141e-06, "loss": 5.6741, "step": 37820 }, { "epoch": 0.31, "learning_rate": 7.784545935290138e-06, "loss": 6.4106, "step": 37830 }, { "epoch": 0.31, "learning_rate": 7.783469876619276e-06, "loss": 6.5459, "step": 37840 }, { "epoch": 0.31, "learning_rate": 7.78239363110106e-06, "loss": 5.6096, "step": 37850 }, { "epoch": 0.31, "learning_rate": 7.781317198807733e-06, "loss": 5.2115, "step": 37860 }, { "epoch": 0.31, "learning_rate": 7.780240579811554e-06, "loss": 5.925, "step": 37870 }, { "epoch": 0.31, "learning_rate": 7.779163774184792e-06, "loss": 5.6878, "step": 37880 }, { "epoch": 0.31, "learning_rate": 7.778086781999734e-06, "loss": 5.7323, "step": 37890 }, { "epoch": 0.31, "learning_rate": 7.777009603328673e-06, "loss": 5.8455, "step": 37900 }, { "epoch": 0.31, "learning_rate": 7.775932238243917e-06, "loss": 5.4051, "step": 37910 }, { "epoch": 0.31, "learning_rate": 7.774854686817788e-06, "loss": 5.6818, "step": 37920 }, { "epoch": 0.31, "learning_rate": 7.773776949122621e-06, "loss": 5.8068, "step": 37930 }, { "epoch": 0.31, "learning_rate": 7.772699025230758e-06, "loss": 5.3061, "step": 37940 }, { "epoch": 0.31, "learning_rate": 7.77162091521456e-06, "loss": 5.6327, "step": 37950 }, { "epoch": 0.31, "learning_rate": 7.770542619146398e-06, "loss": 5.4707, "step": 37960 }, { "epoch": 0.31, "learning_rate": 7.769464137098653e-06, "loss": 5.3039, "step": 37970 }, { "epoch": 0.31, "learning_rate": 7.768385469143722e-06, "loss": 5.6927, "step": 37980 }, { "epoch": 0.31, "learning_rate": 7.767306615354014e-06, "loss": 5.6313, "step": 37990 }, { "epoch": 0.31, "learning_rate": 7.76622757580195e-06, "loss": 5.9683, "step": 38000 }, { "epoch": 0.31, "learning_rate": 7.765148350559961e-06, "loss": 6.0909, "step": 38010 }, { "epoch": 0.31, "learning_rate": 7.764068939700495e-06, "loss": 5.7255, "step": 38020 }, { "epoch": 0.31, "learning_rate": 7.762989343296011e-06, "loss": 6.0946, "step": 38030 }, { "epoch": 0.31, "learning_rate": 7.761909561418974e-06, "loss": 5.118, "step": 38040 }, { "epoch": 0.31, "learning_rate": 7.760829594141874e-06, "loss": 6.2554, "step": 38050 }, { "epoch": 0.31, "learning_rate": 7.759749441537202e-06, "loss": 5.3452, "step": 38060 }, { "epoch": 0.31, "learning_rate": 7.758669103677467e-06, "loss": 6.2148, "step": 38070 }, { "epoch": 0.31, "learning_rate": 7.75758858063519e-06, "loss": 5.8272, "step": 38080 }, { "epoch": 0.31, "learning_rate": 7.756507872482906e-06, "loss": 5.3508, "step": 38090 }, { "epoch": 0.31, "learning_rate": 7.755426979293154e-06, "loss": 5.4908, "step": 38100 }, { "epoch": 0.31, "learning_rate": 7.754345901138497e-06, "loss": 6.6445, "step": 38110 }, { "epoch": 0.31, "learning_rate": 7.753264638091504e-06, "loss": 5.7758, "step": 38120 }, { "epoch": 0.31, "learning_rate": 7.752183190224759e-06, "loss": 6.079, "step": 38130 }, { "epoch": 0.31, "learning_rate": 7.751101557610851e-06, "loss": 5.3295, "step": 38140 }, { "epoch": 0.31, "learning_rate": 7.750019740322394e-06, "loss": 6.0832, "step": 38150 }, { "epoch": 0.31, "learning_rate": 7.748937738432003e-06, "loss": 5.8697, "step": 38160 }, { "epoch": 0.31, "learning_rate": 7.747855552012311e-06, "loss": 5.8368, "step": 38170 }, { "epoch": 0.31, "learning_rate": 7.746773181135964e-06, "loss": 6.0264, "step": 38180 }, { "epoch": 0.31, "learning_rate": 7.74569062587562e-06, "loss": 5.6721, "step": 38190 }, { "epoch": 0.32, "learning_rate": 7.744607886303947e-06, "loss": 5.9808, "step": 38200 }, { "epoch": 0.32, "learning_rate": 7.743524962493623e-06, "loss": 5.7829, "step": 38210 }, { "epoch": 0.32, "learning_rate": 7.742441854517347e-06, "loss": 5.5758, "step": 38220 }, { "epoch": 0.32, "learning_rate": 7.741358562447822e-06, "loss": 5.3196, "step": 38230 }, { "epoch": 0.32, "learning_rate": 7.74027508635777e-06, "loss": 5.879, "step": 38240 }, { "epoch": 0.32, "learning_rate": 7.739191426319918e-06, "loss": 5.7829, "step": 38250 }, { "epoch": 0.32, "learning_rate": 7.738107582407012e-06, "loss": 6.1302, "step": 38260 }, { "epoch": 0.32, "learning_rate": 7.737023554691808e-06, "loss": 5.6076, "step": 38270 }, { "epoch": 0.32, "learning_rate": 7.735939343247073e-06, "loss": 5.8387, "step": 38280 }, { "epoch": 0.32, "learning_rate": 7.734854948145586e-06, "loss": 6.1557, "step": 38290 }, { "epoch": 0.32, "learning_rate": 7.73377036946014e-06, "loss": 5.496, "step": 38300 }, { "epoch": 0.32, "learning_rate": 7.732685607263545e-06, "loss": 5.782, "step": 38310 }, { "epoch": 0.32, "learning_rate": 7.731600661628613e-06, "loss": 5.8573, "step": 38320 }, { "epoch": 0.32, "learning_rate": 7.730515532628177e-06, "loss": 5.3449, "step": 38330 }, { "epoch": 0.32, "learning_rate": 7.729430220335075e-06, "loss": 5.7768, "step": 38340 }, { "epoch": 0.32, "learning_rate": 7.728344724822165e-06, "loss": 5.904, "step": 38350 }, { "epoch": 0.32, "learning_rate": 7.727259046162312e-06, "loss": 5.3389, "step": 38360 }, { "epoch": 0.32, "learning_rate": 7.726173184428395e-06, "loss": 6.0489, "step": 38370 }, { "epoch": 0.32, "learning_rate": 7.725087139693307e-06, "loss": 5.7889, "step": 38380 }, { "epoch": 0.32, "learning_rate": 7.724000912029947e-06, "loss": 5.757, "step": 38390 }, { "epoch": 0.32, "learning_rate": 7.722914501511235e-06, "loss": 5.9066, "step": 38400 }, { "epoch": 0.32, "learning_rate": 7.721827908210098e-06, "loss": 5.4203, "step": 38410 }, { "epoch": 0.32, "learning_rate": 7.720741132199476e-06, "loss": 5.473, "step": 38420 }, { "epoch": 0.32, "learning_rate": 7.71965417355232e-06, "loss": 5.9945, "step": 38430 }, { "epoch": 0.32, "learning_rate": 7.718567032341597e-06, "loss": 5.1309, "step": 38440 }, { "epoch": 0.32, "learning_rate": 7.717479708640282e-06, "loss": 5.7993, "step": 38450 }, { "epoch": 0.32, "learning_rate": 7.716392202521369e-06, "loss": 4.9337, "step": 38460 }, { "epoch": 0.32, "learning_rate": 7.715304514057853e-06, "loss": 6.1877, "step": 38470 }, { "epoch": 0.32, "learning_rate": 7.714216643322754e-06, "loss": 5.0919, "step": 38480 }, { "epoch": 0.32, "learning_rate": 7.713128590389092e-06, "loss": 5.8474, "step": 38490 }, { "epoch": 0.32, "learning_rate": 7.712040355329909e-06, "loss": 5.5857, "step": 38500 }, { "epoch": 0.32, "learning_rate": 7.710951938218254e-06, "loss": 5.44, "step": 38510 }, { "epoch": 0.32, "learning_rate": 7.709863339127192e-06, "loss": 5.7048, "step": 38520 }, { "epoch": 0.32, "learning_rate": 7.708774558129796e-06, "loss": 5.898, "step": 38530 }, { "epoch": 0.32, "learning_rate": 7.707685595299153e-06, "loss": 5.5598, "step": 38540 }, { "epoch": 0.32, "learning_rate": 7.706596450708363e-06, "loss": 5.5249, "step": 38550 }, { "epoch": 0.32, "learning_rate": 7.705507124430538e-06, "loss": 5.8846, "step": 38560 }, { "epoch": 0.32, "learning_rate": 7.7044176165388e-06, "loss": 6.1083, "step": 38570 }, { "epoch": 0.32, "learning_rate": 7.703327927106287e-06, "loss": 5.9987, "step": 38580 }, { "epoch": 0.32, "learning_rate": 7.702238056206146e-06, "loss": 5.6633, "step": 38590 }, { "epoch": 0.32, "learning_rate": 7.701148003911537e-06, "loss": 6.2238, "step": 38600 }, { "epoch": 0.32, "learning_rate": 7.700057770295633e-06, "loss": 5.2979, "step": 38610 }, { "epoch": 0.32, "learning_rate": 7.69896735543162e-06, "loss": 5.2048, "step": 38620 }, { "epoch": 0.32, "learning_rate": 7.697876759392692e-06, "loss": 5.6529, "step": 38630 }, { "epoch": 0.32, "learning_rate": 7.69678598225206e-06, "loss": 5.8151, "step": 38640 }, { "epoch": 0.32, "learning_rate": 7.695695024082945e-06, "loss": 5.3813, "step": 38650 }, { "epoch": 0.32, "learning_rate": 7.69460388495858e-06, "loss": 5.4066, "step": 38660 }, { "epoch": 0.32, "learning_rate": 7.693512564952209e-06, "loss": 5.9237, "step": 38670 }, { "epoch": 0.32, "learning_rate": 7.692421064137091e-06, "loss": 5.7404, "step": 38680 }, { "epoch": 0.32, "learning_rate": 7.691329382586496e-06, "loss": 5.9527, "step": 38690 }, { "epoch": 0.32, "learning_rate": 7.690237520373706e-06, "loss": 5.816, "step": 38700 }, { "epoch": 0.32, "learning_rate": 7.689145477572014e-06, "loss": 6.1062, "step": 38710 }, { "epoch": 0.32, "learning_rate": 7.688053254254727e-06, "loss": 5.4317, "step": 38720 }, { "epoch": 0.32, "learning_rate": 7.68696085049516e-06, "loss": 5.4732, "step": 38730 }, { "epoch": 0.32, "learning_rate": 7.68586826636665e-06, "loss": 6.1516, "step": 38740 }, { "epoch": 0.32, "learning_rate": 7.684775501942534e-06, "loss": 5.4061, "step": 38750 }, { "epoch": 0.32, "learning_rate": 7.683682557296168e-06, "loss": 5.3262, "step": 38760 }, { "epoch": 0.32, "learning_rate": 7.682589432500918e-06, "loss": 5.5015, "step": 38770 }, { "epoch": 0.32, "learning_rate": 7.681496127630164e-06, "loss": 6.0925, "step": 38780 }, { "epoch": 0.32, "learning_rate": 7.680402642757293e-06, "loss": 5.2794, "step": 38790 }, { "epoch": 0.32, "learning_rate": 7.679308977955714e-06, "loss": 5.9243, "step": 38800 }, { "epoch": 0.32, "learning_rate": 7.67821513329884e-06, "loss": 5.7886, "step": 38810 }, { "epoch": 0.32, "learning_rate": 7.677121108860096e-06, "loss": 5.2997, "step": 38820 }, { "epoch": 0.32, "learning_rate": 7.676026904712923e-06, "loss": 5.3312, "step": 38830 }, { "epoch": 0.32, "learning_rate": 7.67493252093077e-06, "loss": 5.8174, "step": 38840 }, { "epoch": 0.32, "learning_rate": 7.673837957587104e-06, "loss": 5.9304, "step": 38850 }, { "epoch": 0.32, "learning_rate": 7.672743214755396e-06, "loss": 6.012, "step": 38860 }, { "epoch": 0.32, "learning_rate": 7.671648292509134e-06, "loss": 6.0686, "step": 38870 }, { "epoch": 0.32, "learning_rate": 7.670553190921823e-06, "loss": 5.9592, "step": 38880 }, { "epoch": 0.32, "learning_rate": 7.669457910066965e-06, "loss": 5.3524, "step": 38890 }, { "epoch": 0.32, "learning_rate": 7.668362450018092e-06, "loss": 5.6202, "step": 38900 }, { "epoch": 0.32, "learning_rate": 7.667266810848735e-06, "loss": 5.6713, "step": 38910 }, { "epoch": 0.32, "learning_rate": 7.666170992632442e-06, "loss": 5.6515, "step": 38920 }, { "epoch": 0.32, "learning_rate": 7.665074995442773e-06, "loss": 5.8867, "step": 38930 }, { "epoch": 0.32, "learning_rate": 7.663978819353301e-06, "loss": 5.2014, "step": 38940 }, { "epoch": 0.32, "learning_rate": 7.662882464437606e-06, "loss": 5.8869, "step": 38950 }, { "epoch": 0.32, "learning_rate": 7.661785930769289e-06, "loss": 5.7309, "step": 38960 }, { "epoch": 0.32, "learning_rate": 7.660689218421953e-06, "loss": 5.2178, "step": 38970 }, { "epoch": 0.32, "learning_rate": 7.659592327469217e-06, "loss": 6.1007, "step": 38980 }, { "epoch": 0.32, "learning_rate": 7.658495257984717e-06, "loss": 5.9613, "step": 38990 }, { "epoch": 0.32, "learning_rate": 7.657398010042092e-06, "loss": 5.6755, "step": 39000 }, { "epoch": 0.32, "learning_rate": 7.656300583715e-06, "loss": 5.6795, "step": 39010 }, { "epoch": 0.32, "learning_rate": 7.655202979077108e-06, "loss": 5.3642, "step": 39020 }, { "epoch": 0.32, "learning_rate": 7.654105196202094e-06, "loss": 5.624, "step": 39030 }, { "epoch": 0.32, "learning_rate": 7.653007235163653e-06, "loss": 5.9689, "step": 39040 }, { "epoch": 0.32, "learning_rate": 7.651909096035484e-06, "loss": 6.1009, "step": 39050 }, { "epoch": 0.32, "learning_rate": 7.650810778891307e-06, "loss": 5.6609, "step": 39060 }, { "epoch": 0.32, "learning_rate": 7.649712283804845e-06, "loss": 6.07, "step": 39070 }, { "epoch": 0.32, "learning_rate": 7.64861361084984e-06, "loss": 6.0348, "step": 39080 }, { "epoch": 0.32, "learning_rate": 7.64751476010004e-06, "loss": 5.1357, "step": 39090 }, { "epoch": 0.32, "learning_rate": 7.64641573162921e-06, "loss": 5.8563, "step": 39100 }, { "epoch": 0.32, "learning_rate": 7.645316525511128e-06, "loss": 5.3338, "step": 39110 }, { "epoch": 0.32, "learning_rate": 7.644217141819577e-06, "loss": 5.4969, "step": 39120 }, { "epoch": 0.32, "learning_rate": 7.643117580628359e-06, "loss": 5.1954, "step": 39130 }, { "epoch": 0.32, "learning_rate": 7.64201784201128e-06, "loss": 5.9265, "step": 39140 }, { "epoch": 0.32, "learning_rate": 7.640917926042166e-06, "loss": 5.4614, "step": 39150 }, { "epoch": 0.32, "learning_rate": 7.639817832794851e-06, "loss": 6.3298, "step": 39160 }, { "epoch": 0.32, "learning_rate": 7.638717562343183e-06, "loss": 5.951, "step": 39170 }, { "epoch": 0.32, "learning_rate": 7.637617114761019e-06, "loss": 6.0223, "step": 39180 }, { "epoch": 0.32, "learning_rate": 7.636516490122228e-06, "loss": 5.7492, "step": 39190 }, { "epoch": 0.32, "learning_rate": 7.635415688500696e-06, "loss": 6.1994, "step": 39200 }, { "epoch": 0.32, "learning_rate": 7.634314709970312e-06, "loss": 5.5332, "step": 39210 }, { "epoch": 0.32, "learning_rate": 7.633213554604984e-06, "loss": 5.4003, "step": 39220 }, { "epoch": 0.32, "learning_rate": 7.632112222478632e-06, "loss": 5.5456, "step": 39230 }, { "epoch": 0.32, "learning_rate": 7.631010713665183e-06, "loss": 5.4559, "step": 39240 }, { "epoch": 0.32, "learning_rate": 7.62990902823858e-06, "loss": 5.7086, "step": 39250 }, { "epoch": 0.32, "learning_rate": 7.628807166272776e-06, "loss": 5.2671, "step": 39260 }, { "epoch": 0.32, "learning_rate": 7.627705127841736e-06, "loss": 6.2033, "step": 39270 }, { "epoch": 0.32, "learning_rate": 7.626602913019437e-06, "loss": 5.8687, "step": 39280 }, { "epoch": 0.32, "learning_rate": 7.625500521879868e-06, "loss": 5.7953, "step": 39290 }, { "epoch": 0.32, "learning_rate": 7.624397954497029e-06, "loss": 5.4402, "step": 39300 }, { "epoch": 0.32, "learning_rate": 7.623295210944934e-06, "loss": 6.6032, "step": 39310 }, { "epoch": 0.32, "learning_rate": 7.622192291297609e-06, "loss": 5.3752, "step": 39320 }, { "epoch": 0.32, "learning_rate": 7.621089195629086e-06, "loss": 5.9303, "step": 39330 }, { "epoch": 0.32, "learning_rate": 7.619985924013415e-06, "loss": 5.8054, "step": 39340 }, { "epoch": 0.32, "learning_rate": 7.618882476524656e-06, "loss": 5.7795, "step": 39350 }, { "epoch": 0.32, "learning_rate": 7.617778853236882e-06, "loss": 5.4032, "step": 39360 }, { "epoch": 0.32, "learning_rate": 7.616675054224174e-06, "loss": 5.682, "step": 39370 }, { "epoch": 0.32, "learning_rate": 7.6155710795606284e-06, "loss": 5.768, "step": 39380 }, { "epoch": 0.32, "learning_rate": 7.614466929320352e-06, "loss": 5.7598, "step": 39390 }, { "epoch": 0.32, "learning_rate": 7.6133626035774645e-06, "loss": 6.2022, "step": 39400 }, { "epoch": 0.33, "learning_rate": 7.612258102406096e-06, "loss": 5.0735, "step": 39410 }, { "epoch": 0.33, "learning_rate": 7.611153425880387e-06, "loss": 5.7405, "step": 39420 }, { "epoch": 0.33, "learning_rate": 7.610048574074496e-06, "loss": 5.6671, "step": 39430 }, { "epoch": 0.33, "learning_rate": 7.608943547062585e-06, "loss": 6.0352, "step": 39440 }, { "epoch": 0.33, "learning_rate": 7.607838344918832e-06, "loss": 6.2165, "step": 39450 }, { "epoch": 0.33, "learning_rate": 7.6067329677174295e-06, "loss": 6.2595, "step": 39460 }, { "epoch": 0.33, "learning_rate": 7.605627415532575e-06, "loss": 5.8632, "step": 39470 }, { "epoch": 0.33, "learning_rate": 7.604521688438484e-06, "loss": 5.9067, "step": 39480 }, { "epoch": 0.33, "learning_rate": 7.60341578650938e-06, "loss": 5.2864, "step": 39490 }, { "epoch": 0.33, "learning_rate": 7.602309709819502e-06, "loss": 5.7121, "step": 39500 }, { "epoch": 0.33, "learning_rate": 7.601203458443094e-06, "loss": 5.5495, "step": 39510 }, { "epoch": 0.33, "learning_rate": 7.600097032454418e-06, "loss": 5.697, "step": 39520 }, { "epoch": 0.33, "learning_rate": 7.598990431927746e-06, "loss": 5.5341, "step": 39530 }, { "epoch": 0.33, "learning_rate": 7.597883656937359e-06, "loss": 6.1074, "step": 39540 }, { "epoch": 0.33, "learning_rate": 7.5967767075575545e-06, "loss": 6.2524, "step": 39550 }, { "epoch": 0.33, "learning_rate": 7.595669583862639e-06, "loss": 5.976, "step": 39560 }, { "epoch": 0.33, "learning_rate": 7.594562285926932e-06, "loss": 6.6125, "step": 39570 }, { "epoch": 0.33, "learning_rate": 7.593454813824759e-06, "loss": 5.5605, "step": 39580 }, { "epoch": 0.33, "learning_rate": 7.592347167630466e-06, "loss": 6.3904, "step": 39590 }, { "epoch": 0.33, "learning_rate": 7.5912393474184054e-06, "loss": 5.5687, "step": 39600 }, { "epoch": 0.33, "learning_rate": 7.590131353262943e-06, "loss": 5.373, "step": 39610 }, { "epoch": 0.33, "learning_rate": 7.589023185238455e-06, "loss": 6.9414, "step": 39620 }, { "epoch": 0.33, "learning_rate": 7.587914843419329e-06, "loss": 5.5072, "step": 39630 }, { "epoch": 0.33, "learning_rate": 7.586806327879968e-06, "loss": 6.399, "step": 39640 }, { "epoch": 0.33, "learning_rate": 7.5856976386947815e-06, "loss": 5.7422, "step": 39650 }, { "epoch": 0.33, "learning_rate": 7.584588775938193e-06, "loss": 5.2604, "step": 39660 }, { "epoch": 0.33, "learning_rate": 7.583479739684638e-06, "loss": 5.6107, "step": 39670 }, { "epoch": 0.33, "learning_rate": 7.5823705300085636e-06, "loss": 6.0458, "step": 39680 }, { "epoch": 0.33, "learning_rate": 7.58126114698443e-06, "loss": 5.1917, "step": 39690 }, { "epoch": 0.33, "learning_rate": 7.580151590686704e-06, "loss": 5.8694, "step": 39700 }, { "epoch": 0.33, "learning_rate": 7.579041861189871e-06, "loss": 5.5321, "step": 39710 }, { "epoch": 0.33, "learning_rate": 7.57793195856842e-06, "loss": 6.1177, "step": 39720 }, { "epoch": 0.33, "learning_rate": 7.576821882896858e-06, "loss": 5.6748, "step": 39730 }, { "epoch": 0.33, "learning_rate": 7.5757116342497026e-06, "loss": 5.5125, "step": 39740 }, { "epoch": 0.33, "learning_rate": 7.574601212701481e-06, "loss": 5.3698, "step": 39750 }, { "epoch": 0.33, "learning_rate": 7.573490618326735e-06, "loss": 5.7881, "step": 39760 }, { "epoch": 0.33, "learning_rate": 7.572379851200012e-06, "loss": 5.7441, "step": 39770 }, { "epoch": 0.33, "learning_rate": 7.571268911395879e-06, "loss": 5.9751, "step": 39780 }, { "epoch": 0.33, "learning_rate": 7.570157798988909e-06, "loss": 5.6601, "step": 39790 }, { "epoch": 0.33, "learning_rate": 7.5690465140536865e-06, "loss": 5.6234, "step": 39800 }, { "epoch": 0.33, "learning_rate": 7.5679350566648126e-06, "loss": 5.4494, "step": 39810 }, { "epoch": 0.33, "learning_rate": 7.566823426896893e-06, "loss": 5.6283, "step": 39820 }, { "epoch": 0.33, "learning_rate": 7.565711624824552e-06, "loss": 6.3774, "step": 39830 }, { "epoch": 0.33, "learning_rate": 7.564599650522421e-06, "loss": 5.8258, "step": 39840 }, { "epoch": 0.33, "learning_rate": 7.563487504065144e-06, "loss": 5.789, "step": 39850 }, { "epoch": 0.33, "learning_rate": 7.562375185527373e-06, "loss": 5.7317, "step": 39860 }, { "epoch": 0.33, "learning_rate": 7.561262694983782e-06, "loss": 6.0824, "step": 39870 }, { "epoch": 0.33, "learning_rate": 7.560150032509044e-06, "loss": 5.7727, "step": 39880 }, { "epoch": 0.33, "learning_rate": 7.559037198177852e-06, "loss": 5.918, "step": 39890 }, { "epoch": 0.33, "learning_rate": 7.557924192064908e-06, "loss": 5.9357, "step": 39900 }, { "epoch": 0.33, "learning_rate": 7.556811014244924e-06, "loss": 5.6139, "step": 39910 }, { "epoch": 0.33, "learning_rate": 7.555697664792626e-06, "loss": 5.7107, "step": 39920 }, { "epoch": 0.33, "learning_rate": 7.5545841437827485e-06, "loss": 5.8633, "step": 39930 }, { "epoch": 0.33, "learning_rate": 7.553470451290041e-06, "loss": 6.0826, "step": 39940 }, { "epoch": 0.33, "learning_rate": 7.5523565873892625e-06, "loss": 5.6602, "step": 39950 }, { "epoch": 0.33, "learning_rate": 7.551242552155182e-06, "loss": 6.1686, "step": 39960 }, { "epoch": 0.33, "learning_rate": 7.550128345662587e-06, "loss": 5.8249, "step": 39970 }, { "epoch": 0.33, "learning_rate": 7.549013967986267e-06, "loss": 5.3876, "step": 39980 }, { "epoch": 0.33, "learning_rate": 7.547899419201028e-06, "loss": 5.5031, "step": 39990 }, { "epoch": 0.33, "learning_rate": 7.5467846993816866e-06, "loss": 5.6345, "step": 40000 } ], "logging_steps": 10, "max_steps": 121255, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 2000, "total_flos": 2.1368236854723452e+23, "train_batch_size": 6, "trial_name": null, "trial_params": null }