{ "best_metric": null, "best_model_checkpoint": null, "epoch": 29.99870090503616, "eval_steps": 500, "global_step": 346380, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 1.4999999999999998e-06, "loss": 18.8583, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.9999999999999997e-06, "loss": 18.1984, "step": 20 }, { "epoch": 0.0, "learning_rate": 4.499999999999999e-06, "loss": 16.678, "step": 30 }, { "epoch": 0.0, "learning_rate": 5.999999999999999e-06, "loss": 14.9006, "step": 40 }, { "epoch": 0.0, "learning_rate": 7.499999999999999e-06, "loss": 12.7181, "step": 50 }, { "epoch": 0.01, "learning_rate": 8.999999999999999e-06, "loss": 10.6191, "step": 60 }, { "epoch": 0.01, "learning_rate": 1.05e-05, "loss": 10.0972, "step": 70 }, { "epoch": 0.01, "learning_rate": 1.1999999999999999e-05, "loss": 10.1492, "step": 80 }, { "epoch": 0.01, "learning_rate": 1.3499999999999998e-05, "loss": 10.0602, "step": 90 }, { "epoch": 0.01, "learning_rate": 1.4999999999999999e-05, "loss": 9.9927, "step": 100 }, { "epoch": 0.01, "learning_rate": 1.6499999999999998e-05, "loss": 9.8969, "step": 110 }, { "epoch": 0.01, "learning_rate": 1.7999999999999997e-05, "loss": 9.836, "step": 120 }, { "epoch": 0.01, "learning_rate": 1.95e-05, "loss": 9.8094, "step": 130 }, { "epoch": 0.01, "learning_rate": 2.1e-05, "loss": 9.6952, "step": 140 }, { "epoch": 0.01, "learning_rate": 2.2499999999999998e-05, "loss": 9.6581, "step": 150 }, { "epoch": 0.01, "learning_rate": 2.3999999999999997e-05, "loss": 9.4633, "step": 160 }, { "epoch": 0.01, "learning_rate": 2.55e-05, "loss": 9.3699, "step": 170 }, { "epoch": 0.02, "learning_rate": 2.6999999999999996e-05, "loss": 9.0112, "step": 180 }, { "epoch": 0.02, "learning_rate": 2.8499999999999998e-05, "loss": 8.8424, "step": 190 }, { "epoch": 0.02, "learning_rate": 2.9999999999999997e-05, "loss": 8.6897, "step": 200 }, { "epoch": 0.02, "learning_rate": 3.149999999999999e-05, "loss": 8.2824, "step": 210 }, { "epoch": 0.02, "learning_rate": 3.2999999999999996e-05, "loss": 8.0319, "step": 220 }, { "epoch": 0.02, "learning_rate": 3.45e-05, "loss": 7.5857, "step": 230 }, { "epoch": 0.02, "learning_rate": 3.5999999999999994e-05, "loss": 7.2037, "step": 240 }, { "epoch": 0.02, "learning_rate": 3.75e-05, "loss": 6.9506, "step": 250 }, { "epoch": 0.02, "learning_rate": 3.9e-05, "loss": 6.86, "step": 260 }, { "epoch": 0.02, "learning_rate": 4.05e-05, "loss": 6.7323, "step": 270 }, { "epoch": 0.02, "learning_rate": 4.2e-05, "loss": 6.5812, "step": 280 }, { "epoch": 0.03, "learning_rate": 4.3499999999999993e-05, "loss": 6.4457, "step": 290 }, { "epoch": 0.03, "learning_rate": 4.4999999999999996e-05, "loss": 6.404, "step": 300 }, { "epoch": 0.03, "learning_rate": 4.65e-05, "loss": 6.3427, "step": 310 }, { "epoch": 0.03, "learning_rate": 4.7999999999999994e-05, "loss": 6.2768, "step": 320 }, { "epoch": 0.03, "learning_rate": 4.95e-05, "loss": 6.3021, "step": 330 }, { "epoch": 0.03, "learning_rate": 5.1e-05, "loss": 6.1851, "step": 340 }, { "epoch": 0.03, "learning_rate": 5.2499999999999995e-05, "loss": 6.2159, "step": 350 }, { "epoch": 0.03, "learning_rate": 5.399999999999999e-05, "loss": 6.2137, "step": 360 }, { "epoch": 0.03, "learning_rate": 5.5499999999999994e-05, "loss": 6.0963, "step": 370 }, { "epoch": 0.03, "learning_rate": 5.6999999999999996e-05, "loss": 6.1693, "step": 380 }, { "epoch": 0.03, "learning_rate": 5.85e-05, "loss": 6.1917, "step": 390 }, { "epoch": 0.03, "learning_rate": 5.9999999999999995e-05, "loss": 6.158, "step": 400 }, { "epoch": 0.04, "learning_rate": 6.149999999999999e-05, "loss": 6.1419, "step": 410 }, { "epoch": 0.04, "learning_rate": 6.299999999999999e-05, "loss": 6.1728, "step": 420 }, { "epoch": 0.04, "learning_rate": 6.45e-05, "loss": 6.102, "step": 430 }, { "epoch": 0.04, "learning_rate": 6.599999999999999e-05, "loss": 6.0982, "step": 440 }, { "epoch": 0.04, "learning_rate": 6.75e-05, "loss": 6.0543, "step": 450 }, { "epoch": 0.04, "learning_rate": 6.9e-05, "loss": 6.1574, "step": 460 }, { "epoch": 0.04, "learning_rate": 7.049999999999999e-05, "loss": 6.1791, "step": 470 }, { "epoch": 0.04, "learning_rate": 7.199999999999999e-05, "loss": 6.0454, "step": 480 }, { "epoch": 0.04, "learning_rate": 7.35e-05, "loss": 6.0594, "step": 490 }, { "epoch": 0.04, "learning_rate": 7.5e-05, "loss": 5.9974, "step": 500 }, { "epoch": 0.04, "learning_rate": 7.499783161790216e-05, "loss": 6.1239, "step": 510 }, { "epoch": 0.05, "learning_rate": 7.499566323580433e-05, "loss": 6.0247, "step": 520 }, { "epoch": 0.05, "learning_rate": 7.499349485370648e-05, "loss": 6.0647, "step": 530 }, { "epoch": 0.05, "learning_rate": 7.499132647160865e-05, "loss": 6.0303, "step": 540 }, { "epoch": 0.05, "learning_rate": 7.49891580895108e-05, "loss": 5.9168, "step": 550 }, { "epoch": 0.05, "learning_rate": 7.498698970741296e-05, "loss": 6.0698, "step": 560 }, { "epoch": 0.05, "learning_rate": 7.498482132531513e-05, "loss": 6.0006, "step": 570 }, { "epoch": 0.05, "learning_rate": 7.49826529432173e-05, "loss": 5.9183, "step": 580 }, { "epoch": 0.05, "learning_rate": 7.498048456111946e-05, "loss": 5.8812, "step": 590 }, { "epoch": 0.05, "learning_rate": 7.497831617902162e-05, "loss": 5.9283, "step": 600 }, { "epoch": 0.05, "learning_rate": 7.497614779692378e-05, "loss": 5.857, "step": 610 }, { "epoch": 0.05, "learning_rate": 7.497397941482594e-05, "loss": 5.9092, "step": 620 }, { "epoch": 0.05, "learning_rate": 7.49718110327281e-05, "loss": 5.8752, "step": 630 }, { "epoch": 0.06, "learning_rate": 7.496964265063027e-05, "loss": 5.9765, "step": 640 }, { "epoch": 0.06, "learning_rate": 7.496747426853244e-05, "loss": 5.879, "step": 650 }, { "epoch": 0.06, "learning_rate": 7.49653058864346e-05, "loss": 5.7714, "step": 660 }, { "epoch": 0.06, "learning_rate": 7.496313750433675e-05, "loss": 5.7552, "step": 670 }, { "epoch": 0.06, "learning_rate": 7.496096912223892e-05, "loss": 5.9567, "step": 680 }, { "epoch": 0.06, "learning_rate": 7.495880074014109e-05, "loss": 5.6824, "step": 690 }, { "epoch": 0.06, "learning_rate": 7.495663235804324e-05, "loss": 5.8346, "step": 700 }, { "epoch": 0.06, "learning_rate": 7.49544639759454e-05, "loss": 5.8017, "step": 710 }, { "epoch": 0.06, "learning_rate": 7.495229559384757e-05, "loss": 5.7523, "step": 720 }, { "epoch": 0.06, "learning_rate": 7.495012721174974e-05, "loss": 5.7637, "step": 730 }, { "epoch": 0.06, "learning_rate": 7.494795882965189e-05, "loss": 5.7492, "step": 740 }, { "epoch": 0.06, "learning_rate": 7.494579044755406e-05, "loss": 5.756, "step": 750 }, { "epoch": 0.07, "learning_rate": 7.494362206545622e-05, "loss": 5.8157, "step": 760 }, { "epoch": 0.07, "learning_rate": 7.494145368335838e-05, "loss": 5.739, "step": 770 }, { "epoch": 0.07, "learning_rate": 7.493928530126054e-05, "loss": 5.8046, "step": 780 }, { "epoch": 0.07, "learning_rate": 7.493711691916271e-05, "loss": 5.8475, "step": 790 }, { "epoch": 0.07, "learning_rate": 7.493494853706488e-05, "loss": 5.6798, "step": 800 }, { "epoch": 0.07, "learning_rate": 7.493278015496703e-05, "loss": 5.7446, "step": 810 }, { "epoch": 0.07, "learning_rate": 7.49306117728692e-05, "loss": 5.6122, "step": 820 }, { "epoch": 0.07, "learning_rate": 7.492844339077136e-05, "loss": 5.772, "step": 830 }, { "epoch": 0.07, "learning_rate": 7.492627500867353e-05, "loss": 5.7689, "step": 840 }, { "epoch": 0.07, "learning_rate": 7.492410662657568e-05, "loss": 5.7019, "step": 850 }, { "epoch": 0.07, "learning_rate": 7.492193824447785e-05, "loss": 5.5278, "step": 860 }, { "epoch": 0.08, "learning_rate": 7.491976986238001e-05, "loss": 5.7049, "step": 870 }, { "epoch": 0.08, "learning_rate": 7.491760148028217e-05, "loss": 5.6011, "step": 880 }, { "epoch": 0.08, "learning_rate": 7.491543309818433e-05, "loss": 5.6569, "step": 890 }, { "epoch": 0.08, "learning_rate": 7.49132647160865e-05, "loss": 5.6797, "step": 900 }, { "epoch": 0.08, "learning_rate": 7.491109633398867e-05, "loss": 5.6013, "step": 910 }, { "epoch": 0.08, "learning_rate": 7.490892795189083e-05, "loss": 5.485, "step": 920 }, { "epoch": 0.08, "learning_rate": 7.490675956979299e-05, "loss": 5.4273, "step": 930 }, { "epoch": 0.08, "learning_rate": 7.490459118769515e-05, "loss": 5.4601, "step": 940 }, { "epoch": 0.08, "learning_rate": 7.49024228055973e-05, "loss": 5.5372, "step": 950 }, { "epoch": 0.08, "learning_rate": 7.490025442349947e-05, "loss": 5.5838, "step": 960 }, { "epoch": 0.08, "learning_rate": 7.489808604140164e-05, "loss": 5.4351, "step": 970 }, { "epoch": 0.08, "learning_rate": 7.48959176593038e-05, "loss": 5.5333, "step": 980 }, { "epoch": 0.09, "learning_rate": 7.489374927720597e-05, "loss": 5.3899, "step": 990 }, { "epoch": 0.09, "learning_rate": 7.489158089510812e-05, "loss": 5.3022, "step": 1000 }, { "epoch": 0.09, "learning_rate": 7.488941251301028e-05, "loss": 5.3262, "step": 1010 }, { "epoch": 0.09, "learning_rate": 7.488724413091244e-05, "loss": 5.4944, "step": 1020 }, { "epoch": 0.09, "learning_rate": 7.488507574881461e-05, "loss": 5.3458, "step": 1030 }, { "epoch": 0.09, "learning_rate": 7.488290736671678e-05, "loss": 5.3084, "step": 1040 }, { "epoch": 0.09, "learning_rate": 7.488073898461894e-05, "loss": 5.5479, "step": 1050 }, { "epoch": 0.09, "learning_rate": 7.48785706025211e-05, "loss": 5.2453, "step": 1060 }, { "epoch": 0.09, "learning_rate": 7.487640222042326e-05, "loss": 5.0319, "step": 1070 }, { "epoch": 0.09, "learning_rate": 7.487423383832541e-05, "loss": 5.2655, "step": 1080 }, { "epoch": 0.09, "learning_rate": 7.487206545622758e-05, "loss": 5.2834, "step": 1090 }, { "epoch": 0.1, "learning_rate": 7.486989707412975e-05, "loss": 5.4504, "step": 1100 }, { "epoch": 0.1, "learning_rate": 7.486772869203191e-05, "loss": 5.2936, "step": 1110 }, { "epoch": 0.1, "learning_rate": 7.486556030993408e-05, "loss": 5.1136, "step": 1120 }, { "epoch": 0.1, "learning_rate": 7.486339192783623e-05, "loss": 5.2435, "step": 1130 }, { "epoch": 0.1, "learning_rate": 7.48612235457384e-05, "loss": 5.1899, "step": 1140 }, { "epoch": 0.1, "learning_rate": 7.485905516364056e-05, "loss": 5.4103, "step": 1150 }, { "epoch": 0.1, "learning_rate": 7.485688678154272e-05, "loss": 5.1588, "step": 1160 }, { "epoch": 0.1, "learning_rate": 7.485471839944488e-05, "loss": 5.1584, "step": 1170 }, { "epoch": 0.1, "learning_rate": 7.485255001734705e-05, "loss": 4.9173, "step": 1180 }, { "epoch": 0.1, "learning_rate": 7.485038163524922e-05, "loss": 5.195, "step": 1190 }, { "epoch": 0.1, "learning_rate": 7.484821325315137e-05, "loss": 5.2081, "step": 1200 }, { "epoch": 0.1, "learning_rate": 7.484604487105354e-05, "loss": 5.1258, "step": 1210 }, { "epoch": 0.11, "learning_rate": 7.48438764889557e-05, "loss": 4.8151, "step": 1220 }, { "epoch": 0.11, "learning_rate": 7.484170810685787e-05, "loss": 5.0038, "step": 1230 }, { "epoch": 0.11, "learning_rate": 7.483953972476002e-05, "loss": 4.9418, "step": 1240 }, { "epoch": 0.11, "learning_rate": 7.483737134266219e-05, "loss": 4.9258, "step": 1250 }, { "epoch": 0.11, "learning_rate": 7.483520296056435e-05, "loss": 4.9788, "step": 1260 }, { "epoch": 0.11, "learning_rate": 7.483303457846651e-05, "loss": 5.1367, "step": 1270 }, { "epoch": 0.11, "learning_rate": 7.483086619636867e-05, "loss": 5.1771, "step": 1280 }, { "epoch": 0.11, "learning_rate": 7.482869781427084e-05, "loss": 4.9901, "step": 1290 }, { "epoch": 0.11, "learning_rate": 7.4826529432173e-05, "loss": 4.9196, "step": 1300 }, { "epoch": 0.11, "learning_rate": 7.482436105007516e-05, "loss": 4.98, "step": 1310 }, { "epoch": 0.11, "learning_rate": 7.482219266797733e-05, "loss": 4.9524, "step": 1320 }, { "epoch": 0.12, "learning_rate": 7.482002428587949e-05, "loss": 4.9602, "step": 1330 }, { "epoch": 0.12, "learning_rate": 7.481785590378165e-05, "loss": 4.9392, "step": 1340 }, { "epoch": 0.12, "learning_rate": 7.481568752168381e-05, "loss": 4.8363, "step": 1350 }, { "epoch": 0.12, "learning_rate": 7.481351913958598e-05, "loss": 4.8841, "step": 1360 }, { "epoch": 0.12, "learning_rate": 7.481135075748814e-05, "loss": 5.1635, "step": 1370 }, { "epoch": 0.12, "learning_rate": 7.480918237539031e-05, "loss": 4.886, "step": 1380 }, { "epoch": 0.12, "learning_rate": 7.480701399329246e-05, "loss": 4.69, "step": 1390 }, { "epoch": 0.12, "learning_rate": 7.480484561119463e-05, "loss": 4.9518, "step": 1400 }, { "epoch": 0.12, "learning_rate": 7.480267722909678e-05, "loss": 4.8937, "step": 1410 }, { "epoch": 0.12, "learning_rate": 7.480050884699895e-05, "loss": 4.6523, "step": 1420 }, { "epoch": 0.12, "learning_rate": 7.479834046490112e-05, "loss": 4.7557, "step": 1430 }, { "epoch": 0.12, "learning_rate": 7.479617208280328e-05, "loss": 4.8456, "step": 1440 }, { "epoch": 0.13, "learning_rate": 7.479400370070545e-05, "loss": 4.609, "step": 1450 }, { "epoch": 0.13, "learning_rate": 7.47918353186076e-05, "loss": 4.6028, "step": 1460 }, { "epoch": 0.13, "learning_rate": 7.478966693650977e-05, "loss": 4.6292, "step": 1470 }, { "epoch": 0.13, "learning_rate": 7.478749855441192e-05, "loss": 4.8444, "step": 1480 }, { "epoch": 0.13, "learning_rate": 7.478533017231409e-05, "loss": 4.6261, "step": 1490 }, { "epoch": 0.13, "learning_rate": 7.478316179021625e-05, "loss": 4.7111, "step": 1500 }, { "epoch": 0.13, "learning_rate": 7.478099340811842e-05, "loss": 4.7857, "step": 1510 }, { "epoch": 0.13, "learning_rate": 7.477882502602059e-05, "loss": 4.5204, "step": 1520 }, { "epoch": 0.13, "learning_rate": 7.477665664392274e-05, "loss": 4.627, "step": 1530 }, { "epoch": 0.13, "learning_rate": 7.47744882618249e-05, "loss": 4.5711, "step": 1540 }, { "epoch": 0.13, "learning_rate": 7.477231987972706e-05, "loss": 4.6654, "step": 1550 }, { "epoch": 0.14, "learning_rate": 7.477015149762922e-05, "loss": 4.6924, "step": 1560 }, { "epoch": 0.14, "learning_rate": 7.476798311553139e-05, "loss": 4.3533, "step": 1570 }, { "epoch": 0.14, "learning_rate": 7.476581473343356e-05, "loss": 4.3503, "step": 1580 }, { "epoch": 0.14, "learning_rate": 7.476364635133572e-05, "loss": 4.4145, "step": 1590 }, { "epoch": 0.14, "learning_rate": 7.476147796923788e-05, "loss": 4.4017, "step": 1600 }, { "epoch": 0.14, "learning_rate": 7.475930958714004e-05, "loss": 4.5338, "step": 1610 }, { "epoch": 0.14, "learning_rate": 7.47571412050422e-05, "loss": 4.4684, "step": 1620 }, { "epoch": 0.14, "learning_rate": 7.475497282294436e-05, "loss": 4.5775, "step": 1630 }, { "epoch": 0.14, "learning_rate": 7.475280444084653e-05, "loss": 4.3703, "step": 1640 }, { "epoch": 0.14, "learning_rate": 7.47506360587487e-05, "loss": 4.4034, "step": 1650 }, { "epoch": 0.14, "learning_rate": 7.474846767665086e-05, "loss": 4.4653, "step": 1660 }, { "epoch": 0.14, "learning_rate": 7.474629929455301e-05, "loss": 4.2406, "step": 1670 }, { "epoch": 0.15, "learning_rate": 7.474413091245518e-05, "loss": 4.2935, "step": 1680 }, { "epoch": 0.15, "learning_rate": 7.474196253035735e-05, "loss": 4.5922, "step": 1690 }, { "epoch": 0.15, "learning_rate": 7.47397941482595e-05, "loss": 4.2746, "step": 1700 }, { "epoch": 0.15, "learning_rate": 7.473762576616167e-05, "loss": 4.2702, "step": 1710 }, { "epoch": 0.15, "learning_rate": 7.473545738406383e-05, "loss": 4.2506, "step": 1720 }, { "epoch": 0.15, "learning_rate": 7.4733289001966e-05, "loss": 4.1589, "step": 1730 }, { "epoch": 0.15, "learning_rate": 7.473112061986815e-05, "loss": 4.3407, "step": 1740 }, { "epoch": 0.15, "learning_rate": 7.472895223777032e-05, "loss": 4.3728, "step": 1750 }, { "epoch": 0.15, "learning_rate": 7.472678385567248e-05, "loss": 4.4, "step": 1760 }, { "epoch": 0.15, "learning_rate": 7.472461547357465e-05, "loss": 4.1539, "step": 1770 }, { "epoch": 0.15, "learning_rate": 7.47224470914768e-05, "loss": 4.6651, "step": 1780 }, { "epoch": 0.16, "learning_rate": 7.472027870937897e-05, "loss": 4.351, "step": 1790 }, { "epoch": 0.16, "learning_rate": 7.471811032728114e-05, "loss": 4.4239, "step": 1800 }, { "epoch": 0.16, "learning_rate": 7.471594194518329e-05, "loss": 4.1925, "step": 1810 }, { "epoch": 0.16, "learning_rate": 7.471377356308546e-05, "loss": 4.1195, "step": 1820 }, { "epoch": 0.16, "learning_rate": 7.471160518098762e-05, "loss": 4.2928, "step": 1830 }, { "epoch": 0.16, "learning_rate": 7.470943679888979e-05, "loss": 4.3644, "step": 1840 }, { "epoch": 0.16, "learning_rate": 7.470726841679194e-05, "loss": 4.1805, "step": 1850 }, { "epoch": 0.16, "learning_rate": 7.470510003469411e-05, "loss": 4.1852, "step": 1860 }, { "epoch": 0.16, "learning_rate": 7.470293165259627e-05, "loss": 4.3363, "step": 1870 }, { "epoch": 0.16, "learning_rate": 7.470076327049843e-05, "loss": 4.2637, "step": 1880 }, { "epoch": 0.16, "learning_rate": 7.46985948884006e-05, "loss": 4.1922, "step": 1890 }, { "epoch": 0.16, "learning_rate": 7.469642650630276e-05, "loss": 4.2863, "step": 1900 }, { "epoch": 0.17, "learning_rate": 7.469425812420493e-05, "loss": 4.1096, "step": 1910 }, { "epoch": 0.17, "learning_rate": 7.469208974210709e-05, "loss": 4.1397, "step": 1920 }, { "epoch": 0.17, "learning_rate": 7.468992136000925e-05, "loss": 3.9411, "step": 1930 }, { "epoch": 0.17, "learning_rate": 7.468775297791141e-05, "loss": 4.0327, "step": 1940 }, { "epoch": 0.17, "learning_rate": 7.468558459581356e-05, "loss": 4.2083, "step": 1950 }, { "epoch": 0.17, "learning_rate": 7.468341621371573e-05, "loss": 4.153, "step": 1960 }, { "epoch": 0.17, "learning_rate": 7.46812478316179e-05, "loss": 4.1819, "step": 1970 }, { "epoch": 0.17, "learning_rate": 7.467907944952006e-05, "loss": 4.1151, "step": 1980 }, { "epoch": 0.17, "learning_rate": 7.467691106742223e-05, "loss": 4.301, "step": 1990 }, { "epoch": 0.17, "learning_rate": 7.467474268532438e-05, "loss": 4.3091, "step": 2000 }, { "epoch": 0.17, "learning_rate": 7.467257430322655e-05, "loss": 4.1224, "step": 2010 }, { "epoch": 0.17, "learning_rate": 7.46704059211287e-05, "loss": 4.1797, "step": 2020 }, { "epoch": 0.18, "learning_rate": 7.466823753903087e-05, "loss": 4.2084, "step": 2030 }, { "epoch": 0.18, "learning_rate": 7.466606915693303e-05, "loss": 3.9976, "step": 2040 }, { "epoch": 0.18, "learning_rate": 7.46639007748352e-05, "loss": 4.0933, "step": 2050 }, { "epoch": 0.18, "learning_rate": 7.466173239273737e-05, "loss": 3.9961, "step": 2060 }, { "epoch": 0.18, "learning_rate": 7.465956401063952e-05, "loss": 3.8498, "step": 2070 }, { "epoch": 0.18, "learning_rate": 7.465739562854167e-05, "loss": 3.9387, "step": 2080 }, { "epoch": 0.18, "learning_rate": 7.465522724644384e-05, "loss": 3.9425, "step": 2090 }, { "epoch": 0.18, "learning_rate": 7.4653058864346e-05, "loss": 4.1571, "step": 2100 }, { "epoch": 0.18, "learning_rate": 7.465089048224817e-05, "loss": 3.9512, "step": 2110 }, { "epoch": 0.18, "learning_rate": 7.464872210015034e-05, "loss": 3.9681, "step": 2120 }, { "epoch": 0.18, "learning_rate": 7.464655371805249e-05, "loss": 3.8389, "step": 2130 }, { "epoch": 0.19, "learning_rate": 7.464438533595466e-05, "loss": 4.039, "step": 2140 }, { "epoch": 0.19, "learning_rate": 7.464221695385682e-05, "loss": 3.7907, "step": 2150 }, { "epoch": 0.19, "learning_rate": 7.464004857175898e-05, "loss": 3.9809, "step": 2160 }, { "epoch": 0.19, "learning_rate": 7.463788018966114e-05, "loss": 3.7039, "step": 2170 }, { "epoch": 0.19, "learning_rate": 7.463571180756331e-05, "loss": 3.7516, "step": 2180 }, { "epoch": 0.19, "learning_rate": 7.463354342546548e-05, "loss": 3.9935, "step": 2190 }, { "epoch": 0.19, "learning_rate": 7.463137504336763e-05, "loss": 3.9089, "step": 2200 }, { "epoch": 0.19, "learning_rate": 7.46292066612698e-05, "loss": 3.7331, "step": 2210 }, { "epoch": 0.19, "learning_rate": 7.462703827917196e-05, "loss": 3.9582, "step": 2220 }, { "epoch": 0.19, "learning_rate": 7.462486989707413e-05, "loss": 3.7926, "step": 2230 }, { "epoch": 0.19, "learning_rate": 7.462270151497628e-05, "loss": 3.8401, "step": 2240 }, { "epoch": 0.19, "learning_rate": 7.462053313287845e-05, "loss": 4.1245, "step": 2250 }, { "epoch": 0.2, "learning_rate": 7.461836475078061e-05, "loss": 3.8321, "step": 2260 }, { "epoch": 0.2, "learning_rate": 7.461619636868277e-05, "loss": 3.8349, "step": 2270 }, { "epoch": 0.2, "learning_rate": 7.461402798658493e-05, "loss": 3.7833, "step": 2280 }, { "epoch": 0.2, "learning_rate": 7.46118596044871e-05, "loss": 4.0642, "step": 2290 }, { "epoch": 0.2, "learning_rate": 7.460969122238927e-05, "loss": 4.0113, "step": 2300 }, { "epoch": 0.2, "learning_rate": 7.460752284029142e-05, "loss": 3.8905, "step": 2310 }, { "epoch": 0.2, "learning_rate": 7.460535445819359e-05, "loss": 3.7561, "step": 2320 }, { "epoch": 0.2, "learning_rate": 7.460318607609575e-05, "loss": 3.994, "step": 2330 }, { "epoch": 0.2, "learning_rate": 7.46010176939979e-05, "loss": 3.9728, "step": 2340 }, { "epoch": 0.2, "learning_rate": 7.459884931190007e-05, "loss": 3.5661, "step": 2350 }, { "epoch": 0.2, "learning_rate": 7.459668092980224e-05, "loss": 3.8083, "step": 2360 }, { "epoch": 0.21, "learning_rate": 7.45945125477044e-05, "loss": 3.7505, "step": 2370 }, { "epoch": 0.21, "learning_rate": 7.459234416560657e-05, "loss": 3.7702, "step": 2380 }, { "epoch": 0.21, "learning_rate": 7.459017578350872e-05, "loss": 3.6379, "step": 2390 }, { "epoch": 0.21, "learning_rate": 7.458800740141089e-05, "loss": 3.6239, "step": 2400 }, { "epoch": 0.21, "learning_rate": 7.458583901931304e-05, "loss": 3.6525, "step": 2410 }, { "epoch": 0.21, "learning_rate": 7.458367063721521e-05, "loss": 3.5079, "step": 2420 }, { "epoch": 0.21, "learning_rate": 7.458150225511738e-05, "loss": 3.9237, "step": 2430 }, { "epoch": 0.21, "learning_rate": 7.457933387301954e-05, "loss": 3.7548, "step": 2440 }, { "epoch": 0.21, "learning_rate": 7.457716549092171e-05, "loss": 3.3384, "step": 2450 }, { "epoch": 0.21, "learning_rate": 7.457499710882386e-05, "loss": 3.5203, "step": 2460 }, { "epoch": 0.21, "learning_rate": 7.457282872672603e-05, "loss": 3.6539, "step": 2470 }, { "epoch": 0.21, "learning_rate": 7.457066034462818e-05, "loss": 3.7093, "step": 2480 }, { "epoch": 0.22, "learning_rate": 7.456849196253035e-05, "loss": 3.5853, "step": 2490 }, { "epoch": 0.22, "learning_rate": 7.456632358043251e-05, "loss": 3.6316, "step": 2500 }, { "epoch": 0.22, "learning_rate": 7.456415519833468e-05, "loss": 3.6594, "step": 2510 }, { "epoch": 0.22, "learning_rate": 7.456198681623685e-05, "loss": 3.7069, "step": 2520 }, { "epoch": 0.22, "learning_rate": 7.4559818434139e-05, "loss": 3.7548, "step": 2530 }, { "epoch": 0.22, "learning_rate": 7.455765005204116e-05, "loss": 3.6331, "step": 2540 }, { "epoch": 0.22, "learning_rate": 7.455548166994332e-05, "loss": 3.6446, "step": 2550 }, { "epoch": 0.22, "learning_rate": 7.455331328784548e-05, "loss": 3.5858, "step": 2560 }, { "epoch": 0.22, "learning_rate": 7.455114490574765e-05, "loss": 3.8041, "step": 2570 }, { "epoch": 0.22, "learning_rate": 7.454897652364982e-05, "loss": 3.5973, "step": 2580 }, { "epoch": 0.22, "learning_rate": 7.454680814155198e-05, "loss": 3.5421, "step": 2590 }, { "epoch": 0.23, "learning_rate": 7.454463975945414e-05, "loss": 3.4146, "step": 2600 }, { "epoch": 0.23, "learning_rate": 7.45424713773563e-05, "loss": 3.4743, "step": 2610 }, { "epoch": 0.23, "learning_rate": 7.454030299525846e-05, "loss": 3.4382, "step": 2620 }, { "epoch": 0.23, "learning_rate": 7.453813461316062e-05, "loss": 3.5705, "step": 2630 }, { "epoch": 0.23, "learning_rate": 7.453596623106279e-05, "loss": 3.3237, "step": 2640 }, { "epoch": 0.23, "learning_rate": 7.453379784896495e-05, "loss": 3.7434, "step": 2650 }, { "epoch": 0.23, "learning_rate": 7.453162946686712e-05, "loss": 3.3854, "step": 2660 }, { "epoch": 0.23, "learning_rate": 7.452946108476927e-05, "loss": 3.4712, "step": 2670 }, { "epoch": 0.23, "learning_rate": 7.452729270267144e-05, "loss": 3.5223, "step": 2680 }, { "epoch": 0.23, "learning_rate": 7.45251243205736e-05, "loss": 3.5466, "step": 2690 }, { "epoch": 0.23, "learning_rate": 7.452295593847576e-05, "loss": 3.5547, "step": 2700 }, { "epoch": 0.23, "learning_rate": 7.452078755637793e-05, "loss": 3.6161, "step": 2710 }, { "epoch": 0.24, "learning_rate": 7.451861917428009e-05, "loss": 3.6327, "step": 2720 }, { "epoch": 0.24, "learning_rate": 7.451645079218226e-05, "loss": 3.2957, "step": 2730 }, { "epoch": 0.24, "learning_rate": 7.451428241008441e-05, "loss": 3.6022, "step": 2740 }, { "epoch": 0.24, "learning_rate": 7.451211402798658e-05, "loss": 3.6322, "step": 2750 }, { "epoch": 0.24, "learning_rate": 7.450994564588874e-05, "loss": 3.344, "step": 2760 }, { "epoch": 0.24, "learning_rate": 7.450777726379091e-05, "loss": 3.7069, "step": 2770 }, { "epoch": 0.24, "learning_rate": 7.450560888169306e-05, "loss": 3.5059, "step": 2780 }, { "epoch": 0.24, "learning_rate": 7.450344049959523e-05, "loss": 3.4515, "step": 2790 }, { "epoch": 0.24, "learning_rate": 7.45012721174974e-05, "loss": 3.5887, "step": 2800 }, { "epoch": 0.24, "learning_rate": 7.449910373539955e-05, "loss": 3.7024, "step": 2810 }, { "epoch": 0.24, "learning_rate": 7.449693535330172e-05, "loss": 3.3055, "step": 2820 }, { "epoch": 0.25, "learning_rate": 7.449476697120388e-05, "loss": 3.2593, "step": 2830 }, { "epoch": 0.25, "learning_rate": 7.449259858910605e-05, "loss": 3.399, "step": 2840 }, { "epoch": 0.25, "learning_rate": 7.44904302070082e-05, "loss": 3.4278, "step": 2850 }, { "epoch": 0.25, "learning_rate": 7.448826182491037e-05, "loss": 3.403, "step": 2860 }, { "epoch": 0.25, "learning_rate": 7.448609344281253e-05, "loss": 3.6006, "step": 2870 }, { "epoch": 0.25, "learning_rate": 7.448392506071469e-05, "loss": 3.5777, "step": 2880 }, { "epoch": 0.25, "learning_rate": 7.448175667861685e-05, "loss": 3.5498, "step": 2890 }, { "epoch": 0.25, "learning_rate": 7.447958829651902e-05, "loss": 3.2079, "step": 2900 }, { "epoch": 0.25, "learning_rate": 7.447741991442119e-05, "loss": 3.3489, "step": 2910 }, { "epoch": 0.25, "learning_rate": 7.447525153232335e-05, "loss": 3.5088, "step": 2920 }, { "epoch": 0.25, "learning_rate": 7.44730831502255e-05, "loss": 3.5346, "step": 2930 }, { "epoch": 0.25, "learning_rate": 7.447091476812767e-05, "loss": 3.4652, "step": 2940 }, { "epoch": 0.26, "learning_rate": 7.446874638602982e-05, "loss": 3.363, "step": 2950 }, { "epoch": 0.26, "learning_rate": 7.446657800393199e-05, "loss": 3.542, "step": 2960 }, { "epoch": 0.26, "learning_rate": 7.446440962183416e-05, "loss": 3.3968, "step": 2970 }, { "epoch": 0.26, "learning_rate": 7.446224123973632e-05, "loss": 3.555, "step": 2980 }, { "epoch": 0.26, "learning_rate": 7.446007285763849e-05, "loss": 3.4794, "step": 2990 }, { "epoch": 0.26, "learning_rate": 7.445790447554064e-05, "loss": 3.7936, "step": 3000 }, { "epoch": 0.26, "learning_rate": 7.445573609344281e-05, "loss": 3.325, "step": 3010 }, { "epoch": 0.26, "learning_rate": 7.445356771134496e-05, "loss": 3.3486, "step": 3020 }, { "epoch": 0.26, "learning_rate": 7.445139932924713e-05, "loss": 3.4112, "step": 3030 }, { "epoch": 0.26, "learning_rate": 7.44492309471493e-05, "loss": 3.2209, "step": 3040 }, { "epoch": 0.26, "learning_rate": 7.444706256505146e-05, "loss": 3.4518, "step": 3050 }, { "epoch": 0.27, "learning_rate": 7.444489418295363e-05, "loss": 3.0233, "step": 3060 }, { "epoch": 0.27, "learning_rate": 7.444272580085578e-05, "loss": 3.56, "step": 3070 }, { "epoch": 0.27, "learning_rate": 7.444055741875795e-05, "loss": 3.3346, "step": 3080 }, { "epoch": 0.27, "learning_rate": 7.44383890366601e-05, "loss": 3.1564, "step": 3090 }, { "epoch": 0.27, "learning_rate": 7.443622065456227e-05, "loss": 3.3377, "step": 3100 }, { "epoch": 0.27, "learning_rate": 7.443405227246443e-05, "loss": 3.3986, "step": 3110 }, { "epoch": 0.27, "learning_rate": 7.44318838903666e-05, "loss": 3.6383, "step": 3120 }, { "epoch": 0.27, "learning_rate": 7.442971550826877e-05, "loss": 3.5299, "step": 3130 }, { "epoch": 0.27, "learning_rate": 7.442754712617092e-05, "loss": 3.3486, "step": 3140 }, { "epoch": 0.27, "learning_rate": 7.442537874407308e-05, "loss": 3.3642, "step": 3150 }, { "epoch": 0.27, "learning_rate": 7.442321036197524e-05, "loss": 3.2285, "step": 3160 }, { "epoch": 0.27, "learning_rate": 7.44210419798774e-05, "loss": 3.4373, "step": 3170 }, { "epoch": 0.28, "learning_rate": 7.441887359777957e-05, "loss": 3.3355, "step": 3180 }, { "epoch": 0.28, "learning_rate": 7.441670521568174e-05, "loss": 3.3528, "step": 3190 }, { "epoch": 0.28, "learning_rate": 7.441453683358389e-05, "loss": 3.243, "step": 3200 }, { "epoch": 0.28, "learning_rate": 7.441236845148606e-05, "loss": 3.1817, "step": 3210 }, { "epoch": 0.28, "learning_rate": 7.441020006938822e-05, "loss": 3.1351, "step": 3220 }, { "epoch": 0.28, "learning_rate": 7.440803168729039e-05, "loss": 3.5019, "step": 3230 }, { "epoch": 0.28, "learning_rate": 7.440586330519254e-05, "loss": 3.1699, "step": 3240 }, { "epoch": 0.28, "learning_rate": 7.440369492309471e-05, "loss": 3.1525, "step": 3250 }, { "epoch": 0.28, "learning_rate": 7.440152654099687e-05, "loss": 3.2619, "step": 3260 }, { "epoch": 0.28, "learning_rate": 7.439935815889903e-05, "loss": 3.6294, "step": 3270 }, { "epoch": 0.28, "learning_rate": 7.439718977680119e-05, "loss": 3.2736, "step": 3280 }, { "epoch": 0.28, "learning_rate": 7.439502139470336e-05, "loss": 3.2921, "step": 3290 }, { "epoch": 0.29, "learning_rate": 7.439285301260553e-05, "loss": 2.9834, "step": 3300 }, { "epoch": 0.29, "learning_rate": 7.439068463050769e-05, "loss": 3.3976, "step": 3310 }, { "epoch": 0.29, "learning_rate": 7.438851624840985e-05, "loss": 3.3947, "step": 3320 }, { "epoch": 0.29, "learning_rate": 7.438634786631201e-05, "loss": 3.3761, "step": 3330 }, { "epoch": 0.29, "learning_rate": 7.438417948421416e-05, "loss": 3.2781, "step": 3340 }, { "epoch": 0.29, "learning_rate": 7.438201110211633e-05, "loss": 3.2493, "step": 3350 }, { "epoch": 0.29, "learning_rate": 7.43798427200185e-05, "loss": 3.2109, "step": 3360 }, { "epoch": 0.29, "learning_rate": 7.437767433792066e-05, "loss": 3.349, "step": 3370 }, { "epoch": 0.29, "learning_rate": 7.437550595582283e-05, "loss": 3.3225, "step": 3380 }, { "epoch": 0.29, "learning_rate": 7.437333757372498e-05, "loss": 3.2703, "step": 3390 }, { "epoch": 0.29, "learning_rate": 7.437116919162715e-05, "loss": 3.3758, "step": 3400 }, { "epoch": 0.3, "learning_rate": 7.43690008095293e-05, "loss": 3.2754, "step": 3410 }, { "epoch": 0.3, "learning_rate": 7.436683242743147e-05, "loss": 3.2345, "step": 3420 }, { "epoch": 0.3, "learning_rate": 7.436466404533363e-05, "loss": 3.1658, "step": 3430 }, { "epoch": 0.3, "learning_rate": 7.43624956632358e-05, "loss": 3.3213, "step": 3440 }, { "epoch": 0.3, "learning_rate": 7.436032728113797e-05, "loss": 3.3476, "step": 3450 }, { "epoch": 0.3, "learning_rate": 7.435815889904012e-05, "loss": 3.0984, "step": 3460 }, { "epoch": 0.3, "learning_rate": 7.435599051694229e-05, "loss": 3.1481, "step": 3470 }, { "epoch": 0.3, "learning_rate": 7.435382213484444e-05, "loss": 2.9932, "step": 3480 }, { "epoch": 0.3, "learning_rate": 7.43516537527466e-05, "loss": 3.1976, "step": 3490 }, { "epoch": 0.3, "learning_rate": 7.434948537064877e-05, "loss": 3.3327, "step": 3500 }, { "epoch": 0.3, "learning_rate": 7.434731698855094e-05, "loss": 3.2585, "step": 3510 }, { "epoch": 0.3, "learning_rate": 7.43451486064531e-05, "loss": 3.1494, "step": 3520 }, { "epoch": 0.31, "learning_rate": 7.434298022435526e-05, "loss": 3.051, "step": 3530 }, { "epoch": 0.31, "learning_rate": 7.434081184225742e-05, "loss": 3.2441, "step": 3540 }, { "epoch": 0.31, "learning_rate": 7.433864346015958e-05, "loss": 3.089, "step": 3550 }, { "epoch": 0.31, "learning_rate": 7.433647507806174e-05, "loss": 3.0438, "step": 3560 }, { "epoch": 0.31, "learning_rate": 7.433430669596391e-05, "loss": 3.1088, "step": 3570 }, { "epoch": 0.31, "learning_rate": 7.433213831386608e-05, "loss": 3.1263, "step": 3580 }, { "epoch": 0.31, "learning_rate": 7.432996993176824e-05, "loss": 3.1882, "step": 3590 }, { "epoch": 0.31, "learning_rate": 7.43278015496704e-05, "loss": 3.2919, "step": 3600 }, { "epoch": 0.31, "learning_rate": 7.432563316757256e-05, "loss": 3.2641, "step": 3610 }, { "epoch": 0.31, "learning_rate": 7.432346478547472e-05, "loss": 3.0463, "step": 3620 }, { "epoch": 0.31, "learning_rate": 7.432129640337688e-05, "loss": 3.1743, "step": 3630 }, { "epoch": 0.32, "learning_rate": 7.431912802127905e-05, "loss": 3.0994, "step": 3640 }, { "epoch": 0.32, "learning_rate": 7.431695963918121e-05, "loss": 3.1548, "step": 3650 }, { "epoch": 0.32, "learning_rate": 7.431479125708338e-05, "loss": 3.1535, "step": 3660 }, { "epoch": 0.32, "learning_rate": 7.431262287498553e-05, "loss": 3.251, "step": 3670 }, { "epoch": 0.32, "learning_rate": 7.43104544928877e-05, "loss": 3.2449, "step": 3680 }, { "epoch": 0.32, "learning_rate": 7.430828611078987e-05, "loss": 2.9014, "step": 3690 }, { "epoch": 0.32, "learning_rate": 7.430611772869202e-05, "loss": 3.2802, "step": 3700 }, { "epoch": 0.32, "learning_rate": 7.430394934659419e-05, "loss": 3.0916, "step": 3710 }, { "epoch": 0.32, "learning_rate": 7.430178096449635e-05, "loss": 3.3312, "step": 3720 }, { "epoch": 0.32, "learning_rate": 7.429961258239852e-05, "loss": 3.117, "step": 3730 }, { "epoch": 0.32, "learning_rate": 7.429744420030067e-05, "loss": 2.9901, "step": 3740 }, { "epoch": 0.32, "learning_rate": 7.429527581820284e-05, "loss": 3.0755, "step": 3750 }, { "epoch": 0.33, "learning_rate": 7.4293107436105e-05, "loss": 3.1105, "step": 3760 }, { "epoch": 0.33, "learning_rate": 7.429093905400717e-05, "loss": 2.9298, "step": 3770 }, { "epoch": 0.33, "learning_rate": 7.428877067190932e-05, "loss": 2.9732, "step": 3780 }, { "epoch": 0.33, "learning_rate": 7.428660228981149e-05, "loss": 3.2697, "step": 3790 }, { "epoch": 0.33, "learning_rate": 7.428443390771366e-05, "loss": 2.9716, "step": 3800 }, { "epoch": 0.33, "learning_rate": 7.428226552561581e-05, "loss": 3.0698, "step": 3810 }, { "epoch": 0.33, "learning_rate": 7.428009714351798e-05, "loss": 3.1655, "step": 3820 }, { "epoch": 0.33, "learning_rate": 7.427792876142014e-05, "loss": 2.9083, "step": 3830 }, { "epoch": 0.33, "learning_rate": 7.427576037932231e-05, "loss": 3.1177, "step": 3840 }, { "epoch": 0.33, "learning_rate": 7.427359199722446e-05, "loss": 3.2873, "step": 3850 }, { "epoch": 0.33, "learning_rate": 7.427142361512663e-05, "loss": 3.0279, "step": 3860 }, { "epoch": 0.34, "learning_rate": 7.42692552330288e-05, "loss": 3.1992, "step": 3870 }, { "epoch": 0.34, "learning_rate": 7.426708685093095e-05, "loss": 2.9868, "step": 3880 }, { "epoch": 0.34, "learning_rate": 7.426491846883311e-05, "loss": 3.0982, "step": 3890 }, { "epoch": 0.34, "learning_rate": 7.426275008673528e-05, "loss": 3.0561, "step": 3900 }, { "epoch": 0.34, "learning_rate": 7.426058170463745e-05, "loss": 2.9747, "step": 3910 }, { "epoch": 0.34, "learning_rate": 7.425841332253961e-05, "loss": 3.1103, "step": 3920 }, { "epoch": 0.34, "learning_rate": 7.425624494044176e-05, "loss": 3.2518, "step": 3930 }, { "epoch": 0.34, "learning_rate": 7.425407655834393e-05, "loss": 3.203, "step": 3940 }, { "epoch": 0.34, "learning_rate": 7.425190817624608e-05, "loss": 3.0603, "step": 3950 }, { "epoch": 0.34, "learning_rate": 7.424973979414825e-05, "loss": 3.0884, "step": 3960 }, { "epoch": 0.34, "learning_rate": 7.424757141205042e-05, "loss": 3.2292, "step": 3970 }, { "epoch": 0.34, "learning_rate": 7.424540302995258e-05, "loss": 3.2534, "step": 3980 }, { "epoch": 0.35, "learning_rate": 7.424323464785475e-05, "loss": 3.2484, "step": 3990 }, { "epoch": 0.35, "learning_rate": 7.42410662657569e-05, "loss": 3.2668, "step": 4000 }, { "epoch": 0.35, "learning_rate": 7.423889788365907e-05, "loss": 2.9588, "step": 4010 }, { "epoch": 0.35, "learning_rate": 7.423672950156122e-05, "loss": 2.965, "step": 4020 }, { "epoch": 0.35, "learning_rate": 7.423456111946339e-05, "loss": 2.8498, "step": 4030 }, { "epoch": 0.35, "learning_rate": 7.423239273736555e-05, "loss": 3.3157, "step": 4040 }, { "epoch": 0.35, "learning_rate": 7.423022435526772e-05, "loss": 2.9723, "step": 4050 }, { "epoch": 0.35, "learning_rate": 7.422805597316989e-05, "loss": 2.9596, "step": 4060 }, { "epoch": 0.35, "learning_rate": 7.422588759107204e-05, "loss": 3.0471, "step": 4070 }, { "epoch": 0.35, "learning_rate": 7.42237192089742e-05, "loss": 3.1012, "step": 4080 }, { "epoch": 0.35, "learning_rate": 7.422155082687636e-05, "loss": 3.0595, "step": 4090 }, { "epoch": 0.36, "learning_rate": 7.421938244477853e-05, "loss": 3.2645, "step": 4100 }, { "epoch": 0.36, "learning_rate": 7.421721406268069e-05, "loss": 2.9406, "step": 4110 }, { "epoch": 0.36, "learning_rate": 7.421504568058286e-05, "loss": 3.3189, "step": 4120 }, { "epoch": 0.36, "learning_rate": 7.421287729848502e-05, "loss": 3.0824, "step": 4130 }, { "epoch": 0.36, "learning_rate": 7.421070891638718e-05, "loss": 3.0649, "step": 4140 }, { "epoch": 0.36, "learning_rate": 7.420854053428934e-05, "loss": 3.2602, "step": 4150 }, { "epoch": 0.36, "learning_rate": 7.42063721521915e-05, "loss": 3.0296, "step": 4160 }, { "epoch": 0.36, "learning_rate": 7.420420377009366e-05, "loss": 2.9469, "step": 4170 }, { "epoch": 0.36, "learning_rate": 7.420203538799583e-05, "loss": 3.05, "step": 4180 }, { "epoch": 0.36, "learning_rate": 7.4199867005898e-05, "loss": 3.0725, "step": 4190 }, { "epoch": 0.36, "learning_rate": 7.419769862380016e-05, "loss": 2.9409, "step": 4200 }, { "epoch": 0.36, "learning_rate": 7.419553024170232e-05, "loss": 2.9826, "step": 4210 }, { "epoch": 0.37, "learning_rate": 7.419336185960448e-05, "loss": 2.9678, "step": 4220 }, { "epoch": 0.37, "learning_rate": 7.419119347750665e-05, "loss": 2.8676, "step": 4230 }, { "epoch": 0.37, "learning_rate": 7.41890250954088e-05, "loss": 2.8583, "step": 4240 }, { "epoch": 0.37, "learning_rate": 7.418685671331097e-05, "loss": 3.0534, "step": 4250 }, { "epoch": 0.37, "learning_rate": 7.418468833121313e-05, "loss": 3.0366, "step": 4260 }, { "epoch": 0.37, "learning_rate": 7.418251994911529e-05, "loss": 3.0822, "step": 4270 }, { "epoch": 0.37, "learning_rate": 7.418035156701745e-05, "loss": 3.3481, "step": 4280 }, { "epoch": 0.37, "learning_rate": 7.417818318491962e-05, "loss": 2.9979, "step": 4290 }, { "epoch": 0.37, "learning_rate": 7.417601480282179e-05, "loss": 3.1324, "step": 4300 }, { "epoch": 0.37, "learning_rate": 7.417384642072395e-05, "loss": 2.7938, "step": 4310 }, { "epoch": 0.37, "learning_rate": 7.41716780386261e-05, "loss": 3.0856, "step": 4320 }, { "epoch": 0.38, "learning_rate": 7.416950965652827e-05, "loss": 2.8916, "step": 4330 }, { "epoch": 0.38, "learning_rate": 7.416734127443042e-05, "loss": 2.9335, "step": 4340 }, { "epoch": 0.38, "learning_rate": 7.416517289233259e-05, "loss": 2.7747, "step": 4350 }, { "epoch": 0.38, "learning_rate": 7.416300451023476e-05, "loss": 3.2759, "step": 4360 }, { "epoch": 0.38, "learning_rate": 7.416083612813692e-05, "loss": 2.8455, "step": 4370 }, { "epoch": 0.38, "learning_rate": 7.415866774603909e-05, "loss": 3.0305, "step": 4380 }, { "epoch": 0.38, "learning_rate": 7.415649936394124e-05, "loss": 2.9684, "step": 4390 }, { "epoch": 0.38, "learning_rate": 7.415433098184341e-05, "loss": 3.1002, "step": 4400 }, { "epoch": 0.38, "learning_rate": 7.415216259974556e-05, "loss": 2.8322, "step": 4410 }, { "epoch": 0.38, "learning_rate": 7.414999421764773e-05, "loss": 2.9863, "step": 4420 }, { "epoch": 0.38, "learning_rate": 7.41478258355499e-05, "loss": 2.8815, "step": 4430 }, { "epoch": 0.38, "learning_rate": 7.414565745345206e-05, "loss": 2.9254, "step": 4440 }, { "epoch": 0.39, "learning_rate": 7.414348907135423e-05, "loss": 2.8573, "step": 4450 }, { "epoch": 0.39, "learning_rate": 7.414132068925638e-05, "loss": 3.0256, "step": 4460 }, { "epoch": 0.39, "learning_rate": 7.413915230715855e-05, "loss": 3.1364, "step": 4470 }, { "epoch": 0.39, "learning_rate": 7.41369839250607e-05, "loss": 3.224, "step": 4480 }, { "epoch": 0.39, "learning_rate": 7.413481554296287e-05, "loss": 3.1553, "step": 4490 }, { "epoch": 0.39, "learning_rate": 7.413264716086503e-05, "loss": 3.104, "step": 4500 }, { "epoch": 0.39, "learning_rate": 7.41304787787672e-05, "loss": 2.783, "step": 4510 }, { "epoch": 0.39, "learning_rate": 7.412831039666936e-05, "loss": 2.939, "step": 4520 }, { "epoch": 0.39, "learning_rate": 7.412614201457152e-05, "loss": 2.7808, "step": 4530 }, { "epoch": 0.39, "learning_rate": 7.412397363247368e-05, "loss": 3.0611, "step": 4540 }, { "epoch": 0.39, "learning_rate": 7.412180525037584e-05, "loss": 2.8822, "step": 4550 }, { "epoch": 0.39, "learning_rate": 7.4119636868278e-05, "loss": 3.2499, "step": 4560 }, { "epoch": 0.4, "learning_rate": 7.411746848618017e-05, "loss": 2.8341, "step": 4570 }, { "epoch": 0.4, "learning_rate": 7.411530010408234e-05, "loss": 3.0496, "step": 4580 }, { "epoch": 0.4, "learning_rate": 7.41131317219845e-05, "loss": 2.8537, "step": 4590 }, { "epoch": 0.4, "learning_rate": 7.411096333988666e-05, "loss": 2.8864, "step": 4600 }, { "epoch": 0.4, "learning_rate": 7.410879495778882e-05, "loss": 2.9079, "step": 4610 }, { "epoch": 0.4, "learning_rate": 7.410662657569097e-05, "loss": 3.091, "step": 4620 }, { "epoch": 0.4, "learning_rate": 7.410445819359314e-05, "loss": 2.9631, "step": 4630 }, { "epoch": 0.4, "learning_rate": 7.410228981149531e-05, "loss": 3.1094, "step": 4640 }, { "epoch": 0.4, "learning_rate": 7.410012142939747e-05, "loss": 2.8326, "step": 4650 }, { "epoch": 0.4, "learning_rate": 7.409795304729964e-05, "loss": 3.1073, "step": 4660 }, { "epoch": 0.4, "learning_rate": 7.409578466520179e-05, "loss": 2.8182, "step": 4670 }, { "epoch": 0.41, "learning_rate": 7.409361628310396e-05, "loss": 3.0326, "step": 4680 }, { "epoch": 0.41, "learning_rate": 7.409144790100613e-05, "loss": 3.0936, "step": 4690 }, { "epoch": 0.41, "learning_rate": 7.408927951890828e-05, "loss": 2.8344, "step": 4700 }, { "epoch": 0.41, "learning_rate": 7.408711113681045e-05, "loss": 3.1854, "step": 4710 }, { "epoch": 0.41, "learning_rate": 7.408494275471261e-05, "loss": 2.9086, "step": 4720 }, { "epoch": 0.41, "learning_rate": 7.408277437261478e-05, "loss": 2.9008, "step": 4730 }, { "epoch": 0.41, "learning_rate": 7.408060599051693e-05, "loss": 3.03, "step": 4740 }, { "epoch": 0.41, "learning_rate": 7.40784376084191e-05, "loss": 2.7896, "step": 4750 }, { "epoch": 0.41, "learning_rate": 7.407626922632126e-05, "loss": 3.2811, "step": 4760 }, { "epoch": 0.41, "learning_rate": 7.407410084422343e-05, "loss": 2.9749, "step": 4770 }, { "epoch": 0.41, "learning_rate": 7.407193246212558e-05, "loss": 3.037, "step": 4780 }, { "epoch": 0.41, "learning_rate": 7.406976408002775e-05, "loss": 2.7812, "step": 4790 }, { "epoch": 0.42, "learning_rate": 7.406759569792992e-05, "loss": 2.6845, "step": 4800 }, { "epoch": 0.42, "learning_rate": 7.406542731583207e-05, "loss": 3.0626, "step": 4810 }, { "epoch": 0.42, "learning_rate": 7.406325893373423e-05, "loss": 3.165, "step": 4820 }, { "epoch": 0.42, "learning_rate": 7.40610905516364e-05, "loss": 3.0081, "step": 4830 }, { "epoch": 0.42, "learning_rate": 7.405892216953857e-05, "loss": 2.7491, "step": 4840 }, { "epoch": 0.42, "learning_rate": 7.405675378744073e-05, "loss": 2.8929, "step": 4850 }, { "epoch": 0.42, "learning_rate": 7.405458540534289e-05, "loss": 3.0552, "step": 4860 }, { "epoch": 0.42, "learning_rate": 7.405241702324505e-05, "loss": 2.8105, "step": 4870 }, { "epoch": 0.42, "learning_rate": 7.40502486411472e-05, "loss": 2.9323, "step": 4880 }, { "epoch": 0.42, "learning_rate": 7.404808025904937e-05, "loss": 2.8548, "step": 4890 }, { "epoch": 0.42, "learning_rate": 7.404591187695154e-05, "loss": 2.678, "step": 4900 }, { "epoch": 0.43, "learning_rate": 7.40437434948537e-05, "loss": 3.014, "step": 4910 }, { "epoch": 0.43, "learning_rate": 7.404157511275587e-05, "loss": 2.7644, "step": 4920 }, { "epoch": 0.43, "learning_rate": 7.403940673065802e-05, "loss": 2.9546, "step": 4930 }, { "epoch": 0.43, "learning_rate": 7.403723834856019e-05, "loss": 2.9442, "step": 4940 }, { "epoch": 0.43, "learning_rate": 7.403506996646234e-05, "loss": 3.0603, "step": 4950 }, { "epoch": 0.43, "learning_rate": 7.403290158436451e-05, "loss": 2.8858, "step": 4960 }, { "epoch": 0.43, "learning_rate": 7.403073320226668e-05, "loss": 2.825, "step": 4970 }, { "epoch": 0.43, "learning_rate": 7.402856482016884e-05, "loss": 2.6631, "step": 4980 }, { "epoch": 0.43, "learning_rate": 7.402639643807101e-05, "loss": 2.912, "step": 4990 }, { "epoch": 0.43, "learning_rate": 7.402422805597316e-05, "loss": 3.0257, "step": 5000 }, { "epoch": 0.43, "learning_rate": 7.402205967387533e-05, "loss": 3.0001, "step": 5010 }, { "epoch": 0.43, "learning_rate": 7.401989129177748e-05, "loss": 2.7976, "step": 5020 }, { "epoch": 0.44, "learning_rate": 7.401772290967965e-05, "loss": 2.7679, "step": 5030 }, { "epoch": 0.44, "learning_rate": 7.401555452758181e-05, "loss": 3.0049, "step": 5040 }, { "epoch": 0.44, "learning_rate": 7.401338614548398e-05, "loss": 2.8595, "step": 5050 }, { "epoch": 0.44, "learning_rate": 7.401121776338615e-05, "loss": 2.8325, "step": 5060 }, { "epoch": 0.44, "learning_rate": 7.40090493812883e-05, "loss": 2.9403, "step": 5070 }, { "epoch": 0.44, "learning_rate": 7.400688099919047e-05, "loss": 2.8569, "step": 5080 }, { "epoch": 0.44, "learning_rate": 7.400471261709262e-05, "loss": 2.8426, "step": 5090 }, { "epoch": 0.44, "learning_rate": 7.400254423499479e-05, "loss": 2.8097, "step": 5100 }, { "epoch": 0.44, "learning_rate": 7.400037585289695e-05, "loss": 2.836, "step": 5110 }, { "epoch": 0.44, "learning_rate": 7.399820747079912e-05, "loss": 3.0055, "step": 5120 }, { "epoch": 0.44, "learning_rate": 7.399603908870128e-05, "loss": 2.9383, "step": 5130 }, { "epoch": 0.45, "learning_rate": 7.399387070660344e-05, "loss": 2.5624, "step": 5140 }, { "epoch": 0.45, "learning_rate": 7.39917023245056e-05, "loss": 2.8693, "step": 5150 }, { "epoch": 0.45, "learning_rate": 7.398953394240776e-05, "loss": 2.5125, "step": 5160 }, { "epoch": 0.45, "learning_rate": 7.398736556030992e-05, "loss": 3.0624, "step": 5170 }, { "epoch": 0.45, "learning_rate": 7.398519717821209e-05, "loss": 2.9647, "step": 5180 }, { "epoch": 0.45, "learning_rate": 7.398302879611426e-05, "loss": 2.6308, "step": 5190 }, { "epoch": 0.45, "learning_rate": 7.398086041401642e-05, "loss": 2.8746, "step": 5200 }, { "epoch": 0.45, "learning_rate": 7.397869203191857e-05, "loss": 2.8454, "step": 5210 }, { "epoch": 0.45, "learning_rate": 7.397652364982074e-05, "loss": 2.6212, "step": 5220 }, { "epoch": 0.45, "learning_rate": 7.397435526772291e-05, "loss": 2.91, "step": 5230 }, { "epoch": 0.45, "learning_rate": 7.397218688562506e-05, "loss": 3.1641, "step": 5240 }, { "epoch": 0.45, "learning_rate": 7.397001850352723e-05, "loss": 2.8644, "step": 5250 }, { "epoch": 0.46, "learning_rate": 7.39678501214294e-05, "loss": 2.7666, "step": 5260 }, { "epoch": 0.46, "learning_rate": 7.396568173933156e-05, "loss": 2.5286, "step": 5270 }, { "epoch": 0.46, "learning_rate": 7.396351335723371e-05, "loss": 2.9986, "step": 5280 }, { "epoch": 0.46, "learning_rate": 7.396134497513588e-05, "loss": 2.8511, "step": 5290 }, { "epoch": 0.46, "learning_rate": 7.395917659303805e-05, "loss": 2.6174, "step": 5300 }, { "epoch": 0.46, "learning_rate": 7.395700821094021e-05, "loss": 2.9677, "step": 5310 }, { "epoch": 0.46, "learning_rate": 7.395483982884236e-05, "loss": 2.6912, "step": 5320 }, { "epoch": 0.46, "learning_rate": 7.395267144674453e-05, "loss": 2.9514, "step": 5330 }, { "epoch": 0.46, "learning_rate": 7.395050306464668e-05, "loss": 3.1262, "step": 5340 }, { "epoch": 0.46, "learning_rate": 7.394833468254885e-05, "loss": 2.7491, "step": 5350 }, { "epoch": 0.46, "learning_rate": 7.394616630045102e-05, "loss": 2.8441, "step": 5360 }, { "epoch": 0.47, "learning_rate": 7.394399791835318e-05, "loss": 2.6192, "step": 5370 }, { "epoch": 0.47, "learning_rate": 7.394182953625535e-05, "loss": 3.0125, "step": 5380 }, { "epoch": 0.47, "learning_rate": 7.39396611541575e-05, "loss": 2.8341, "step": 5390 }, { "epoch": 0.47, "learning_rate": 7.393749277205967e-05, "loss": 2.7614, "step": 5400 }, { "epoch": 0.47, "learning_rate": 7.393532438996182e-05, "loss": 2.8003, "step": 5410 }, { "epoch": 0.47, "learning_rate": 7.393315600786399e-05, "loss": 2.572, "step": 5420 }, { "epoch": 0.47, "learning_rate": 7.393098762576615e-05, "loss": 2.8475, "step": 5430 }, { "epoch": 0.47, "learning_rate": 7.392881924366832e-05, "loss": 2.7744, "step": 5440 }, { "epoch": 0.47, "learning_rate": 7.392665086157049e-05, "loss": 2.9314, "step": 5450 }, { "epoch": 0.47, "learning_rate": 7.392448247947264e-05, "loss": 2.6637, "step": 5460 }, { "epoch": 0.47, "learning_rate": 7.39223140973748e-05, "loss": 2.8796, "step": 5470 }, { "epoch": 0.47, "learning_rate": 7.392014571527696e-05, "loss": 2.7235, "step": 5480 }, { "epoch": 0.48, "learning_rate": 7.391797733317913e-05, "loss": 2.7658, "step": 5490 }, { "epoch": 0.48, "learning_rate": 7.391580895108129e-05, "loss": 2.6113, "step": 5500 }, { "epoch": 0.48, "learning_rate": 7.391364056898346e-05, "loss": 2.8148, "step": 5510 }, { "epoch": 0.48, "learning_rate": 7.391147218688562e-05, "loss": 2.8047, "step": 5520 }, { "epoch": 0.48, "learning_rate": 7.390930380478778e-05, "loss": 2.7904, "step": 5530 }, { "epoch": 0.48, "learning_rate": 7.390713542268994e-05, "loss": 2.8902, "step": 5540 }, { "epoch": 0.48, "learning_rate": 7.39049670405921e-05, "loss": 2.9984, "step": 5550 }, { "epoch": 0.48, "learning_rate": 7.390279865849426e-05, "loss": 2.5971, "step": 5560 }, { "epoch": 0.48, "learning_rate": 7.390063027639643e-05, "loss": 2.7006, "step": 5570 }, { "epoch": 0.48, "learning_rate": 7.38984618942986e-05, "loss": 2.831, "step": 5580 }, { "epoch": 0.48, "learning_rate": 7.389629351220076e-05, "loss": 2.934, "step": 5590 }, { "epoch": 0.48, "learning_rate": 7.389412513010292e-05, "loss": 3.0801, "step": 5600 }, { "epoch": 0.49, "learning_rate": 7.389195674800508e-05, "loss": 2.8833, "step": 5610 }, { "epoch": 0.49, "learning_rate": 7.388978836590725e-05, "loss": 2.8294, "step": 5620 }, { "epoch": 0.49, "learning_rate": 7.38876199838094e-05, "loss": 2.799, "step": 5630 }, { "epoch": 0.49, "learning_rate": 7.388545160171157e-05, "loss": 2.6687, "step": 5640 }, { "epoch": 0.49, "learning_rate": 7.388328321961373e-05, "loss": 2.9306, "step": 5650 }, { "epoch": 0.49, "learning_rate": 7.38811148375159e-05, "loss": 2.8017, "step": 5660 }, { "epoch": 0.49, "learning_rate": 7.387894645541805e-05, "loss": 2.8187, "step": 5670 }, { "epoch": 0.49, "learning_rate": 7.387677807332022e-05, "loss": 2.7951, "step": 5680 }, { "epoch": 0.49, "learning_rate": 7.387460969122239e-05, "loss": 2.9067, "step": 5690 }, { "epoch": 0.49, "learning_rate": 7.387244130912454e-05, "loss": 2.6148, "step": 5700 }, { "epoch": 0.49, "learning_rate": 7.38702729270267e-05, "loss": 2.7119, "step": 5710 }, { "epoch": 0.5, "learning_rate": 7.386810454492887e-05, "loss": 2.759, "step": 5720 }, { "epoch": 0.5, "learning_rate": 7.386593616283104e-05, "loss": 2.8031, "step": 5730 }, { "epoch": 0.5, "learning_rate": 7.386376778073319e-05, "loss": 2.8266, "step": 5740 }, { "epoch": 0.5, "learning_rate": 7.386159939863536e-05, "loss": 2.7777, "step": 5750 }, { "epoch": 0.5, "learning_rate": 7.385943101653752e-05, "loss": 2.6982, "step": 5760 }, { "epoch": 0.5, "learning_rate": 7.385726263443969e-05, "loss": 2.7839, "step": 5770 }, { "epoch": 0.5, "learning_rate": 7.385509425234184e-05, "loss": 2.7013, "step": 5780 }, { "epoch": 0.5, "learning_rate": 7.385292587024401e-05, "loss": 2.739, "step": 5790 }, { "epoch": 0.5, "learning_rate": 7.385075748814618e-05, "loss": 2.7656, "step": 5800 }, { "epoch": 0.5, "learning_rate": 7.384858910604833e-05, "loss": 2.6513, "step": 5810 }, { "epoch": 0.5, "learning_rate": 7.38464207239505e-05, "loss": 2.563, "step": 5820 }, { "epoch": 0.5, "learning_rate": 7.384425234185266e-05, "loss": 2.8056, "step": 5830 }, { "epoch": 0.51, "learning_rate": 7.384208395975483e-05, "loss": 2.9981, "step": 5840 }, { "epoch": 0.51, "learning_rate": 7.3839915577657e-05, "loss": 2.9362, "step": 5850 }, { "epoch": 0.51, "learning_rate": 7.383774719555915e-05, "loss": 2.7083, "step": 5860 }, { "epoch": 0.51, "learning_rate": 7.383557881346131e-05, "loss": 2.9388, "step": 5870 }, { "epoch": 0.51, "learning_rate": 7.383341043136347e-05, "loss": 2.5626, "step": 5880 }, { "epoch": 0.51, "learning_rate": 7.383124204926563e-05, "loss": 2.8984, "step": 5890 }, { "epoch": 0.51, "learning_rate": 7.38290736671678e-05, "loss": 2.5998, "step": 5900 }, { "epoch": 0.51, "learning_rate": 7.382690528506996e-05, "loss": 2.44, "step": 5910 }, { "epoch": 0.51, "learning_rate": 7.382473690297213e-05, "loss": 2.7593, "step": 5920 }, { "epoch": 0.51, "learning_rate": 7.382256852087428e-05, "loss": 2.782, "step": 5930 }, { "epoch": 0.51, "learning_rate": 7.382040013877645e-05, "loss": 2.7126, "step": 5940 }, { "epoch": 0.52, "learning_rate": 7.38182317566786e-05, "loss": 2.6602, "step": 5950 }, { "epoch": 0.52, "learning_rate": 7.381606337458077e-05, "loss": 2.7454, "step": 5960 }, { "epoch": 0.52, "learning_rate": 7.381389499248294e-05, "loss": 2.6183, "step": 5970 }, { "epoch": 0.52, "learning_rate": 7.38117266103851e-05, "loss": 2.8427, "step": 5980 }, { "epoch": 0.52, "learning_rate": 7.380955822828727e-05, "loss": 2.9194, "step": 5990 }, { "epoch": 0.52, "learning_rate": 7.380738984618942e-05, "loss": 2.7572, "step": 6000 }, { "epoch": 0.52, "learning_rate": 7.380522146409159e-05, "loss": 2.6735, "step": 6010 }, { "epoch": 0.52, "learning_rate": 7.380305308199374e-05, "loss": 2.7409, "step": 6020 }, { "epoch": 0.52, "learning_rate": 7.380088469989591e-05, "loss": 2.5057, "step": 6030 }, { "epoch": 0.52, "learning_rate": 7.379871631779807e-05, "loss": 2.7142, "step": 6040 }, { "epoch": 0.52, "learning_rate": 7.379654793570024e-05, "loss": 2.876, "step": 6050 }, { "epoch": 0.52, "learning_rate": 7.37943795536024e-05, "loss": 2.7192, "step": 6060 }, { "epoch": 0.53, "learning_rate": 7.379221117150456e-05, "loss": 2.7241, "step": 6070 }, { "epoch": 0.53, "learning_rate": 7.379004278940673e-05, "loss": 2.8252, "step": 6080 }, { "epoch": 0.53, "learning_rate": 7.378787440730888e-05, "loss": 2.8167, "step": 6090 }, { "epoch": 0.53, "learning_rate": 7.378570602521105e-05, "loss": 2.749, "step": 6100 }, { "epoch": 0.53, "learning_rate": 7.378353764311321e-05, "loss": 2.687, "step": 6110 }, { "epoch": 0.53, "learning_rate": 7.378136926101538e-05, "loss": 2.6954, "step": 6120 }, { "epoch": 0.53, "learning_rate": 7.377920087891754e-05, "loss": 2.7197, "step": 6130 }, { "epoch": 0.53, "learning_rate": 7.37770324968197e-05, "loss": 2.6594, "step": 6140 }, { "epoch": 0.53, "learning_rate": 7.377486411472186e-05, "loss": 2.7222, "step": 6150 }, { "epoch": 0.53, "learning_rate": 7.377269573262402e-05, "loss": 2.4306, "step": 6160 }, { "epoch": 0.53, "learning_rate": 7.377052735052618e-05, "loss": 2.4076, "step": 6170 }, { "epoch": 0.54, "learning_rate": 7.376835896842835e-05, "loss": 2.7324, "step": 6180 }, { "epoch": 0.54, "learning_rate": 7.376619058633052e-05, "loss": 2.7341, "step": 6190 }, { "epoch": 0.54, "learning_rate": 7.376402220423268e-05, "loss": 3.1398, "step": 6200 }, { "epoch": 0.54, "learning_rate": 7.376185382213483e-05, "loss": 2.6369, "step": 6210 }, { "epoch": 0.54, "learning_rate": 7.3759685440037e-05, "loss": 2.7612, "step": 6220 }, { "epoch": 0.54, "learning_rate": 7.375751705793917e-05, "loss": 2.7983, "step": 6230 }, { "epoch": 0.54, "learning_rate": 7.375534867584132e-05, "loss": 2.744, "step": 6240 }, { "epoch": 0.54, "learning_rate": 7.375318029374349e-05, "loss": 2.6777, "step": 6250 }, { "epoch": 0.54, "learning_rate": 7.375101191164565e-05, "loss": 2.6174, "step": 6260 }, { "epoch": 0.54, "learning_rate": 7.374884352954782e-05, "loss": 2.7295, "step": 6270 }, { "epoch": 0.54, "learning_rate": 7.374667514744997e-05, "loss": 2.7771, "step": 6280 }, { "epoch": 0.54, "learning_rate": 7.374450676535214e-05, "loss": 2.6298, "step": 6290 }, { "epoch": 0.55, "learning_rate": 7.37423383832543e-05, "loss": 2.7491, "step": 6300 }, { "epoch": 0.55, "learning_rate": 7.374017000115647e-05, "loss": 2.6193, "step": 6310 }, { "epoch": 0.55, "learning_rate": 7.373800161905862e-05, "loss": 2.7825, "step": 6320 }, { "epoch": 0.55, "learning_rate": 7.373583323696079e-05, "loss": 2.7135, "step": 6330 }, { "epoch": 0.55, "learning_rate": 7.373366485486296e-05, "loss": 2.6877, "step": 6340 }, { "epoch": 0.55, "learning_rate": 7.373149647276511e-05, "loss": 2.9027, "step": 6350 }, { "epoch": 0.55, "learning_rate": 7.372932809066728e-05, "loss": 2.6651, "step": 6360 }, { "epoch": 0.55, "learning_rate": 7.372715970856944e-05, "loss": 2.7372, "step": 6370 }, { "epoch": 0.55, "learning_rate": 7.372499132647161e-05, "loss": 2.7515, "step": 6380 }, { "epoch": 0.55, "learning_rate": 7.372282294437378e-05, "loss": 2.7448, "step": 6390 }, { "epoch": 0.55, "learning_rate": 7.372065456227593e-05, "loss": 2.7507, "step": 6400 }, { "epoch": 0.56, "learning_rate": 7.371848618017808e-05, "loss": 2.5188, "step": 6410 }, { "epoch": 0.56, "learning_rate": 7.371631779808025e-05, "loss": 2.9014, "step": 6420 }, { "epoch": 0.56, "learning_rate": 7.371414941598241e-05, "loss": 2.7717, "step": 6430 }, { "epoch": 0.56, "learning_rate": 7.371198103388458e-05, "loss": 2.7297, "step": 6440 }, { "epoch": 0.56, "learning_rate": 7.370981265178675e-05, "loss": 2.828, "step": 6450 }, { "epoch": 0.56, "learning_rate": 7.37076442696889e-05, "loss": 2.8723, "step": 6460 }, { "epoch": 0.56, "learning_rate": 7.370547588759107e-05, "loss": 2.669, "step": 6470 }, { "epoch": 0.56, "learning_rate": 7.370330750549322e-05, "loss": 2.6188, "step": 6480 }, { "epoch": 0.56, "learning_rate": 7.370113912339539e-05, "loss": 2.5543, "step": 6490 }, { "epoch": 0.56, "learning_rate": 7.369897074129755e-05, "loss": 2.7378, "step": 6500 }, { "epoch": 0.56, "learning_rate": 7.369680235919972e-05, "loss": 2.3497, "step": 6510 }, { "epoch": 0.56, "learning_rate": 7.369463397710188e-05, "loss": 2.7275, "step": 6520 }, { "epoch": 0.57, "learning_rate": 7.369246559500404e-05, "loss": 2.6205, "step": 6530 }, { "epoch": 0.57, "learning_rate": 7.36902972129062e-05, "loss": 2.7103, "step": 6540 }, { "epoch": 0.57, "learning_rate": 7.368812883080836e-05, "loss": 2.6769, "step": 6550 }, { "epoch": 0.57, "learning_rate": 7.368596044871052e-05, "loss": 2.6082, "step": 6560 }, { "epoch": 0.57, "learning_rate": 7.368379206661269e-05, "loss": 2.936, "step": 6570 }, { "epoch": 0.57, "learning_rate": 7.368162368451486e-05, "loss": 2.7365, "step": 6580 }, { "epoch": 0.57, "learning_rate": 7.367945530241702e-05, "loss": 2.7131, "step": 6590 }, { "epoch": 0.57, "learning_rate": 7.367728692031917e-05, "loss": 2.7131, "step": 6600 }, { "epoch": 0.57, "learning_rate": 7.367511853822134e-05, "loss": 2.5484, "step": 6610 }, { "epoch": 0.57, "learning_rate": 7.367295015612351e-05, "loss": 2.8306, "step": 6620 }, { "epoch": 0.57, "learning_rate": 7.367078177402566e-05, "loss": 2.6053, "step": 6630 }, { "epoch": 0.58, "learning_rate": 7.366861339192783e-05, "loss": 2.5171, "step": 6640 }, { "epoch": 0.58, "learning_rate": 7.366644500983e-05, "loss": 2.7263, "step": 6650 }, { "epoch": 0.58, "learning_rate": 7.366427662773216e-05, "loss": 2.6017, "step": 6660 }, { "epoch": 0.58, "learning_rate": 7.366210824563431e-05, "loss": 2.8052, "step": 6670 }, { "epoch": 0.58, "learning_rate": 7.365993986353648e-05, "loss": 2.5411, "step": 6680 }, { "epoch": 0.58, "learning_rate": 7.365777148143865e-05, "loss": 2.735, "step": 6690 }, { "epoch": 0.58, "learning_rate": 7.36556030993408e-05, "loss": 2.3427, "step": 6700 }, { "epoch": 0.58, "learning_rate": 7.365343471724296e-05, "loss": 2.7947, "step": 6710 }, { "epoch": 0.58, "learning_rate": 7.365126633514513e-05, "loss": 2.7307, "step": 6720 }, { "epoch": 0.58, "learning_rate": 7.36490979530473e-05, "loss": 2.7084, "step": 6730 }, { "epoch": 0.58, "learning_rate": 7.364692957094945e-05, "loss": 2.6025, "step": 6740 }, { "epoch": 0.58, "learning_rate": 7.364476118885162e-05, "loss": 2.5918, "step": 6750 }, { "epoch": 0.59, "learning_rate": 7.364259280675378e-05, "loss": 2.8496, "step": 6760 }, { "epoch": 0.59, "learning_rate": 7.364042442465595e-05, "loss": 2.6947, "step": 6770 }, { "epoch": 0.59, "learning_rate": 7.36382560425581e-05, "loss": 2.7205, "step": 6780 }, { "epoch": 0.59, "learning_rate": 7.363608766046027e-05, "loss": 2.8079, "step": 6790 }, { "epoch": 0.59, "learning_rate": 7.363391927836243e-05, "loss": 2.5102, "step": 6800 }, { "epoch": 0.59, "learning_rate": 7.363175089626459e-05, "loss": 2.712, "step": 6810 }, { "epoch": 0.59, "learning_rate": 7.362958251416675e-05, "loss": 2.8943, "step": 6820 }, { "epoch": 0.59, "learning_rate": 7.362741413206892e-05, "loss": 2.4493, "step": 6830 }, { "epoch": 0.59, "learning_rate": 7.362524574997109e-05, "loss": 2.7023, "step": 6840 }, { "epoch": 0.59, "learning_rate": 7.362307736787325e-05, "loss": 2.5408, "step": 6850 }, { "epoch": 0.59, "learning_rate": 7.36209089857754e-05, "loss": 2.6312, "step": 6860 }, { "epoch": 0.59, "learning_rate": 7.361874060367757e-05, "loss": 2.728, "step": 6870 }, { "epoch": 0.6, "learning_rate": 7.361657222157973e-05, "loss": 2.6076, "step": 6880 }, { "epoch": 0.6, "learning_rate": 7.361440383948189e-05, "loss": 2.8936, "step": 6890 }, { "epoch": 0.6, "learning_rate": 7.361223545738406e-05, "loss": 2.7143, "step": 6900 }, { "epoch": 0.6, "learning_rate": 7.361006707528622e-05, "loss": 2.5887, "step": 6910 }, { "epoch": 0.6, "learning_rate": 7.360789869318839e-05, "loss": 2.5424, "step": 6920 }, { "epoch": 0.6, "learning_rate": 7.360573031109054e-05, "loss": 2.5649, "step": 6930 }, { "epoch": 0.6, "learning_rate": 7.360356192899271e-05, "loss": 2.9085, "step": 6940 }, { "epoch": 0.6, "learning_rate": 7.360139354689486e-05, "loss": 2.7691, "step": 6950 }, { "epoch": 0.6, "learning_rate": 7.359922516479703e-05, "loss": 2.646, "step": 6960 }, { "epoch": 0.6, "learning_rate": 7.35970567826992e-05, "loss": 2.6739, "step": 6970 }, { "epoch": 0.6, "learning_rate": 7.359488840060136e-05, "loss": 2.3309, "step": 6980 }, { "epoch": 0.61, "learning_rate": 7.359272001850353e-05, "loss": 2.4501, "step": 6990 }, { "epoch": 0.61, "learning_rate": 7.359055163640568e-05, "loss": 2.8763, "step": 7000 }, { "epoch": 0.61, "learning_rate": 7.358838325430785e-05, "loss": 2.4711, "step": 7010 }, { "epoch": 0.61, "learning_rate": 7.358621487221e-05, "loss": 2.4631, "step": 7020 }, { "epoch": 0.61, "learning_rate": 7.358404649011217e-05, "loss": 2.5148, "step": 7030 }, { "epoch": 0.61, "learning_rate": 7.358187810801433e-05, "loss": 2.8142, "step": 7040 }, { "epoch": 0.61, "learning_rate": 7.35797097259165e-05, "loss": 2.5286, "step": 7050 }, { "epoch": 0.61, "learning_rate": 7.357754134381867e-05, "loss": 2.8427, "step": 7060 }, { "epoch": 0.61, "learning_rate": 7.357537296172082e-05, "loss": 2.8113, "step": 7070 }, { "epoch": 0.61, "learning_rate": 7.357320457962299e-05, "loss": 2.5782, "step": 7080 }, { "epoch": 0.61, "learning_rate": 7.357103619752514e-05, "loss": 2.5264, "step": 7090 }, { "epoch": 0.61, "learning_rate": 7.35688678154273e-05, "loss": 2.7426, "step": 7100 }, { "epoch": 0.62, "learning_rate": 7.356669943332947e-05, "loss": 2.5477, "step": 7110 }, { "epoch": 0.62, "learning_rate": 7.356453105123164e-05, "loss": 2.7004, "step": 7120 }, { "epoch": 0.62, "learning_rate": 7.35623626691338e-05, "loss": 2.6102, "step": 7130 }, { "epoch": 0.62, "learning_rate": 7.356019428703596e-05, "loss": 2.5253, "step": 7140 }, { "epoch": 0.62, "learning_rate": 7.355802590493812e-05, "loss": 2.4509, "step": 7150 }, { "epoch": 0.62, "learning_rate": 7.355585752284029e-05, "loss": 2.745, "step": 7160 }, { "epoch": 0.62, "learning_rate": 7.355368914074244e-05, "loss": 2.5795, "step": 7170 }, { "epoch": 0.62, "learning_rate": 7.355152075864461e-05, "loss": 2.713, "step": 7180 }, { "epoch": 0.62, "learning_rate": 7.354935237654678e-05, "loss": 2.6368, "step": 7190 }, { "epoch": 0.62, "learning_rate": 7.354718399444894e-05, "loss": 2.5619, "step": 7200 }, { "epoch": 0.62, "learning_rate": 7.35450156123511e-05, "loss": 2.5486, "step": 7210 }, { "epoch": 0.63, "learning_rate": 7.354284723025326e-05, "loss": 2.4031, "step": 7220 }, { "epoch": 0.63, "learning_rate": 7.354067884815543e-05, "loss": 2.5905, "step": 7230 }, { "epoch": 0.63, "learning_rate": 7.353851046605758e-05, "loss": 2.6423, "step": 7240 }, { "epoch": 0.63, "learning_rate": 7.353634208395975e-05, "loss": 2.5614, "step": 7250 }, { "epoch": 0.63, "learning_rate": 7.353417370186191e-05, "loss": 2.5984, "step": 7260 }, { "epoch": 0.63, "learning_rate": 7.353200531976408e-05, "loss": 2.6916, "step": 7270 }, { "epoch": 0.63, "learning_rate": 7.352983693766623e-05, "loss": 2.7093, "step": 7280 }, { "epoch": 0.63, "learning_rate": 7.35276685555684e-05, "loss": 2.59, "step": 7290 }, { "epoch": 0.63, "learning_rate": 7.352550017347056e-05, "loss": 2.5747, "step": 7300 }, { "epoch": 0.63, "learning_rate": 7.352333179137273e-05, "loss": 2.524, "step": 7310 }, { "epoch": 0.63, "learning_rate": 7.352116340927488e-05, "loss": 2.3918, "step": 7320 }, { "epoch": 0.63, "learning_rate": 7.351899502717705e-05, "loss": 2.5548, "step": 7330 }, { "epoch": 0.64, "learning_rate": 7.351682664507922e-05, "loss": 2.5692, "step": 7340 }, { "epoch": 0.64, "learning_rate": 7.351465826298137e-05, "loss": 2.5674, "step": 7350 }, { "epoch": 0.64, "learning_rate": 7.351248988088354e-05, "loss": 2.5107, "step": 7360 }, { "epoch": 0.64, "learning_rate": 7.35103214987857e-05, "loss": 2.5247, "step": 7370 }, { "epoch": 0.64, "learning_rate": 7.350815311668787e-05, "loss": 2.6253, "step": 7380 }, { "epoch": 0.64, "learning_rate": 7.350598473459004e-05, "loss": 2.6666, "step": 7390 }, { "epoch": 0.64, "learning_rate": 7.350381635249219e-05, "loss": 2.6775, "step": 7400 }, { "epoch": 0.64, "learning_rate": 7.350164797039435e-05, "loss": 2.6195, "step": 7410 }, { "epoch": 0.64, "learning_rate": 7.349947958829651e-05, "loss": 2.495, "step": 7420 }, { "epoch": 0.64, "learning_rate": 7.349731120619867e-05, "loss": 2.7041, "step": 7430 }, { "epoch": 0.64, "learning_rate": 7.349514282410084e-05, "loss": 2.8191, "step": 7440 }, { "epoch": 0.65, "learning_rate": 7.3492974442003e-05, "loss": 2.8075, "step": 7450 }, { "epoch": 0.65, "learning_rate": 7.349080605990517e-05, "loss": 2.7427, "step": 7460 }, { "epoch": 0.65, "learning_rate": 7.348863767780733e-05, "loss": 2.5691, "step": 7470 }, { "epoch": 0.65, "learning_rate": 7.348646929570948e-05, "loss": 2.6108, "step": 7480 }, { "epoch": 0.65, "learning_rate": 7.348430091361164e-05, "loss": 2.5559, "step": 7490 }, { "epoch": 0.65, "learning_rate": 7.348213253151381e-05, "loss": 2.7071, "step": 7500 }, { "epoch": 0.65, "learning_rate": 7.347996414941598e-05, "loss": 2.5024, "step": 7510 }, { "epoch": 0.65, "learning_rate": 7.347779576731814e-05, "loss": 2.7041, "step": 7520 }, { "epoch": 0.65, "learning_rate": 7.34756273852203e-05, "loss": 2.553, "step": 7530 }, { "epoch": 0.65, "learning_rate": 7.347345900312246e-05, "loss": 2.5618, "step": 7540 }, { "epoch": 0.65, "learning_rate": 7.347129062102462e-05, "loss": 2.6617, "step": 7550 }, { "epoch": 0.65, "learning_rate": 7.346912223892678e-05, "loss": 2.7241, "step": 7560 }, { "epoch": 0.66, "learning_rate": 7.346695385682895e-05, "loss": 2.4777, "step": 7570 }, { "epoch": 0.66, "learning_rate": 7.346478547473112e-05, "loss": 2.6245, "step": 7580 }, { "epoch": 0.66, "learning_rate": 7.346261709263328e-05, "loss": 2.5701, "step": 7590 }, { "epoch": 0.66, "learning_rate": 7.346044871053543e-05, "loss": 2.6401, "step": 7600 }, { "epoch": 0.66, "learning_rate": 7.34582803284376e-05, "loss": 2.3486, "step": 7610 }, { "epoch": 0.66, "learning_rate": 7.345611194633977e-05, "loss": 2.4001, "step": 7620 }, { "epoch": 0.66, "learning_rate": 7.345394356424192e-05, "loss": 2.5041, "step": 7630 }, { "epoch": 0.66, "learning_rate": 7.345177518214409e-05, "loss": 2.5868, "step": 7640 }, { "epoch": 0.66, "learning_rate": 7.344960680004625e-05, "loss": 2.5642, "step": 7650 }, { "epoch": 0.66, "learning_rate": 7.344743841794842e-05, "loss": 2.6328, "step": 7660 }, { "epoch": 0.66, "learning_rate": 7.344527003585057e-05, "loss": 2.5773, "step": 7670 }, { "epoch": 0.67, "learning_rate": 7.344310165375274e-05, "loss": 2.4989, "step": 7680 }, { "epoch": 0.67, "learning_rate": 7.34409332716549e-05, "loss": 2.5158, "step": 7690 }, { "epoch": 0.67, "learning_rate": 7.343876488955706e-05, "loss": 2.4615, "step": 7700 }, { "epoch": 0.67, "learning_rate": 7.343659650745922e-05, "loss": 2.6328, "step": 7710 }, { "epoch": 0.67, "learning_rate": 7.343442812536139e-05, "loss": 2.5019, "step": 7720 }, { "epoch": 0.67, "learning_rate": 7.343225974326356e-05, "loss": 2.6813, "step": 7730 }, { "epoch": 0.67, "learning_rate": 7.343009136116571e-05, "loss": 2.4017, "step": 7740 }, { "epoch": 0.67, "learning_rate": 7.342792297906788e-05, "loss": 2.7138, "step": 7750 }, { "epoch": 0.67, "learning_rate": 7.342575459697004e-05, "loss": 2.7688, "step": 7760 }, { "epoch": 0.67, "learning_rate": 7.342358621487221e-05, "loss": 2.5782, "step": 7770 }, { "epoch": 0.67, "learning_rate": 7.342141783277436e-05, "loss": 2.5307, "step": 7780 }, { "epoch": 0.67, "learning_rate": 7.341924945067653e-05, "loss": 2.6103, "step": 7790 }, { "epoch": 0.68, "learning_rate": 7.34170810685787e-05, "loss": 2.6142, "step": 7800 }, { "epoch": 0.68, "learning_rate": 7.341491268648085e-05, "loss": 2.4974, "step": 7810 }, { "epoch": 0.68, "learning_rate": 7.341274430438301e-05, "loss": 2.7161, "step": 7820 }, { "epoch": 0.68, "learning_rate": 7.341057592228518e-05, "loss": 2.6442, "step": 7830 }, { "epoch": 0.68, "learning_rate": 7.340840754018735e-05, "loss": 2.4472, "step": 7840 }, { "epoch": 0.68, "learning_rate": 7.340623915808951e-05, "loss": 2.6063, "step": 7850 }, { "epoch": 0.68, "learning_rate": 7.340407077599167e-05, "loss": 2.7867, "step": 7860 }, { "epoch": 0.68, "learning_rate": 7.340190239389383e-05, "loss": 2.4145, "step": 7870 }, { "epoch": 0.68, "learning_rate": 7.339973401179599e-05, "loss": 2.5193, "step": 7880 }, { "epoch": 0.68, "learning_rate": 7.339756562969815e-05, "loss": 2.8357, "step": 7890 }, { "epoch": 0.68, "learning_rate": 7.339539724760032e-05, "loss": 2.8346, "step": 7900 }, { "epoch": 0.69, "learning_rate": 7.339322886550248e-05, "loss": 2.4574, "step": 7910 }, { "epoch": 0.69, "learning_rate": 7.339106048340465e-05, "loss": 2.5875, "step": 7920 }, { "epoch": 0.69, "learning_rate": 7.33888921013068e-05, "loss": 2.6348, "step": 7930 }, { "epoch": 0.69, "learning_rate": 7.338672371920897e-05, "loss": 2.5875, "step": 7940 }, { "epoch": 0.69, "learning_rate": 7.338455533711112e-05, "loss": 2.7627, "step": 7950 }, { "epoch": 0.69, "learning_rate": 7.338238695501329e-05, "loss": 2.3992, "step": 7960 }, { "epoch": 0.69, "learning_rate": 7.338021857291546e-05, "loss": 2.5626, "step": 7970 }, { "epoch": 0.69, "learning_rate": 7.337805019081762e-05, "loss": 2.5854, "step": 7980 }, { "epoch": 0.69, "learning_rate": 7.337588180871979e-05, "loss": 2.3504, "step": 7990 }, { "epoch": 0.69, "learning_rate": 7.337371342662194e-05, "loss": 2.5677, "step": 8000 }, { "epoch": 0.69, "learning_rate": 7.337154504452411e-05, "loss": 2.5634, "step": 8010 }, { "epoch": 0.69, "learning_rate": 7.336937666242626e-05, "loss": 2.526, "step": 8020 }, { "epoch": 0.7, "learning_rate": 7.336720828032843e-05, "loss": 2.6569, "step": 8030 }, { "epoch": 0.7, "learning_rate": 7.336503989823059e-05, "loss": 2.4486, "step": 8040 }, { "epoch": 0.7, "learning_rate": 7.336287151613276e-05, "loss": 2.4336, "step": 8050 }, { "epoch": 0.7, "learning_rate": 7.336070313403493e-05, "loss": 2.5843, "step": 8060 }, { "epoch": 0.7, "learning_rate": 7.335853475193708e-05, "loss": 2.5285, "step": 8070 }, { "epoch": 0.7, "learning_rate": 7.335636636983925e-05, "loss": 2.6196, "step": 8080 }, { "epoch": 0.7, "learning_rate": 7.33541979877414e-05, "loss": 2.351, "step": 8090 }, { "epoch": 0.7, "learning_rate": 7.335202960564356e-05, "loss": 2.4833, "step": 8100 }, { "epoch": 0.7, "learning_rate": 7.334986122354573e-05, "loss": 2.6733, "step": 8110 }, { "epoch": 0.7, "learning_rate": 7.33476928414479e-05, "loss": 2.5761, "step": 8120 }, { "epoch": 0.7, "learning_rate": 7.334552445935006e-05, "loss": 2.7597, "step": 8130 }, { "epoch": 0.7, "learning_rate": 7.334335607725222e-05, "loss": 2.5386, "step": 8140 }, { "epoch": 0.71, "learning_rate": 7.334118769515438e-05, "loss": 2.6154, "step": 8150 }, { "epoch": 0.71, "learning_rate": 7.333901931305655e-05, "loss": 2.2955, "step": 8160 }, { "epoch": 0.71, "learning_rate": 7.33368509309587e-05, "loss": 2.37, "step": 8170 }, { "epoch": 0.71, "learning_rate": 7.333468254886087e-05, "loss": 2.4759, "step": 8180 }, { "epoch": 0.71, "learning_rate": 7.333251416676303e-05, "loss": 2.2924, "step": 8190 }, { "epoch": 0.71, "learning_rate": 7.33303457846652e-05, "loss": 2.4127, "step": 8200 }, { "epoch": 0.71, "learning_rate": 7.332817740256735e-05, "loss": 2.4053, "step": 8210 }, { "epoch": 0.71, "learning_rate": 7.332600902046952e-05, "loss": 2.5437, "step": 8220 }, { "epoch": 0.71, "learning_rate": 7.332384063837169e-05, "loss": 2.6504, "step": 8230 }, { "epoch": 0.71, "learning_rate": 7.332167225627384e-05, "loss": 2.7329, "step": 8240 }, { "epoch": 0.71, "learning_rate": 7.3319503874176e-05, "loss": 2.4659, "step": 8250 }, { "epoch": 0.72, "learning_rate": 7.331733549207817e-05, "loss": 2.7041, "step": 8260 }, { "epoch": 0.72, "learning_rate": 7.331516710998034e-05, "loss": 2.4477, "step": 8270 }, { "epoch": 0.72, "learning_rate": 7.331299872788249e-05, "loss": 2.3842, "step": 8280 }, { "epoch": 0.72, "learning_rate": 7.331083034578466e-05, "loss": 2.6698, "step": 8290 }, { "epoch": 0.72, "learning_rate": 7.330866196368682e-05, "loss": 2.3953, "step": 8300 }, { "epoch": 0.72, "learning_rate": 7.330649358158899e-05, "loss": 2.415, "step": 8310 }, { "epoch": 0.72, "learning_rate": 7.330432519949114e-05, "loss": 2.4234, "step": 8320 }, { "epoch": 0.72, "learning_rate": 7.330215681739331e-05, "loss": 2.5922, "step": 8330 }, { "epoch": 0.72, "learning_rate": 7.329998843529548e-05, "loss": 2.4434, "step": 8340 }, { "epoch": 0.72, "learning_rate": 7.329782005319763e-05, "loss": 2.4847, "step": 8350 }, { "epoch": 0.72, "learning_rate": 7.32956516710998e-05, "loss": 2.3967, "step": 8360 }, { "epoch": 0.72, "learning_rate": 7.329348328900196e-05, "loss": 2.4669, "step": 8370 }, { "epoch": 0.73, "learning_rate": 7.329131490690413e-05, "loss": 2.3713, "step": 8380 }, { "epoch": 0.73, "learning_rate": 7.32891465248063e-05, "loss": 2.5006, "step": 8390 }, { "epoch": 0.73, "learning_rate": 7.328697814270845e-05, "loss": 2.4112, "step": 8400 }, { "epoch": 0.73, "learning_rate": 7.328480976061061e-05, "loss": 2.4651, "step": 8410 }, { "epoch": 0.73, "learning_rate": 7.328264137851277e-05, "loss": 2.7108, "step": 8420 }, { "epoch": 0.73, "learning_rate": 7.328047299641493e-05, "loss": 2.407, "step": 8430 }, { "epoch": 0.73, "learning_rate": 7.32783046143171e-05, "loss": 2.5256, "step": 8440 }, { "epoch": 0.73, "learning_rate": 7.327613623221927e-05, "loss": 2.4493, "step": 8450 }, { "epoch": 0.73, "learning_rate": 7.327396785012143e-05, "loss": 2.6493, "step": 8460 }, { "epoch": 0.73, "learning_rate": 7.327179946802359e-05, "loss": 2.4456, "step": 8470 }, { "epoch": 0.73, "learning_rate": 7.326963108592575e-05, "loss": 2.4589, "step": 8480 }, { "epoch": 0.74, "learning_rate": 7.32674627038279e-05, "loss": 2.4465, "step": 8490 }, { "epoch": 0.74, "learning_rate": 7.326529432173007e-05, "loss": 2.5213, "step": 8500 }, { "epoch": 0.74, "learning_rate": 7.326312593963224e-05, "loss": 2.3042, "step": 8510 }, { "epoch": 0.74, "learning_rate": 7.32609575575344e-05, "loss": 2.5674, "step": 8520 }, { "epoch": 0.74, "learning_rate": 7.325878917543657e-05, "loss": 2.682, "step": 8530 }, { "epoch": 0.74, "learning_rate": 7.325662079333872e-05, "loss": 2.5601, "step": 8540 }, { "epoch": 0.74, "learning_rate": 7.325445241124088e-05, "loss": 2.7501, "step": 8550 }, { "epoch": 0.74, "learning_rate": 7.325228402914304e-05, "loss": 2.3392, "step": 8560 }, { "epoch": 0.74, "learning_rate": 7.325011564704521e-05, "loss": 2.4445, "step": 8570 }, { "epoch": 0.74, "learning_rate": 7.324794726494738e-05, "loss": 2.5767, "step": 8580 }, { "epoch": 0.74, "learning_rate": 7.324577888284954e-05, "loss": 2.5104, "step": 8590 }, { "epoch": 0.74, "learning_rate": 7.32436105007517e-05, "loss": 2.5725, "step": 8600 }, { "epoch": 0.75, "learning_rate": 7.324144211865386e-05, "loss": 2.4908, "step": 8610 }, { "epoch": 0.75, "learning_rate": 7.323927373655603e-05, "loss": 2.4767, "step": 8620 }, { "epoch": 0.75, "learning_rate": 7.323710535445818e-05, "loss": 2.4381, "step": 8630 }, { "epoch": 0.75, "learning_rate": 7.323493697236035e-05, "loss": 2.5251, "step": 8640 }, { "epoch": 0.75, "learning_rate": 7.323276859026251e-05, "loss": 2.3338, "step": 8650 }, { "epoch": 0.75, "learning_rate": 7.323060020816468e-05, "loss": 2.5321, "step": 8660 }, { "epoch": 0.75, "learning_rate": 7.322843182606683e-05, "loss": 2.6396, "step": 8670 }, { "epoch": 0.75, "learning_rate": 7.3226263443969e-05, "loss": 2.4621, "step": 8680 }, { "epoch": 0.75, "learning_rate": 7.322409506187116e-05, "loss": 2.4663, "step": 8690 }, { "epoch": 0.75, "learning_rate": 7.322192667977333e-05, "loss": 2.3852, "step": 8700 }, { "epoch": 0.75, "learning_rate": 7.321975829767548e-05, "loss": 2.5342, "step": 8710 }, { "epoch": 0.76, "learning_rate": 7.321758991557765e-05, "loss": 2.4584, "step": 8720 }, { "epoch": 0.76, "learning_rate": 7.321542153347982e-05, "loss": 2.822, "step": 8730 }, { "epoch": 0.76, "learning_rate": 7.321325315138197e-05, "loss": 2.6256, "step": 8740 }, { "epoch": 0.76, "learning_rate": 7.321108476928414e-05, "loss": 2.4641, "step": 8750 }, { "epoch": 0.76, "learning_rate": 7.32089163871863e-05, "loss": 2.539, "step": 8760 }, { "epoch": 0.76, "learning_rate": 7.320674800508847e-05, "loss": 2.5972, "step": 8770 }, { "epoch": 0.76, "learning_rate": 7.320457962299062e-05, "loss": 2.4998, "step": 8780 }, { "epoch": 0.76, "learning_rate": 7.320241124089279e-05, "loss": 2.5182, "step": 8790 }, { "epoch": 0.76, "learning_rate": 7.320024285879495e-05, "loss": 2.5621, "step": 8800 }, { "epoch": 0.76, "learning_rate": 7.319807447669711e-05, "loss": 2.417, "step": 8810 }, { "epoch": 0.76, "learning_rate": 7.319590609459927e-05, "loss": 2.553, "step": 8820 }, { "epoch": 0.76, "learning_rate": 7.319373771250144e-05, "loss": 2.6057, "step": 8830 }, { "epoch": 0.77, "learning_rate": 7.31915693304036e-05, "loss": 2.3281, "step": 8840 }, { "epoch": 0.77, "learning_rate": 7.318940094830577e-05, "loss": 2.1437, "step": 8850 }, { "epoch": 0.77, "learning_rate": 7.318723256620793e-05, "loss": 2.3204, "step": 8860 }, { "epoch": 0.77, "learning_rate": 7.318506418411009e-05, "loss": 2.3575, "step": 8870 }, { "epoch": 0.77, "learning_rate": 7.318289580201224e-05, "loss": 2.4457, "step": 8880 }, { "epoch": 0.77, "learning_rate": 7.318072741991441e-05, "loss": 2.3519, "step": 8890 }, { "epoch": 0.77, "learning_rate": 7.317855903781658e-05, "loss": 2.5907, "step": 8900 }, { "epoch": 0.77, "learning_rate": 7.317639065571874e-05, "loss": 2.5826, "step": 8910 }, { "epoch": 0.77, "learning_rate": 7.317422227362091e-05, "loss": 2.4032, "step": 8920 }, { "epoch": 0.77, "learning_rate": 7.317205389152306e-05, "loss": 2.5346, "step": 8930 }, { "epoch": 0.77, "learning_rate": 7.316988550942523e-05, "loss": 2.4471, "step": 8940 }, { "epoch": 0.78, "learning_rate": 7.316771712732738e-05, "loss": 2.5844, "step": 8950 }, { "epoch": 0.78, "learning_rate": 7.316554874522955e-05, "loss": 2.5136, "step": 8960 }, { "epoch": 0.78, "learning_rate": 7.316338036313172e-05, "loss": 2.5352, "step": 8970 }, { "epoch": 0.78, "learning_rate": 7.316121198103388e-05, "loss": 2.5078, "step": 8980 }, { "epoch": 0.78, "learning_rate": 7.315904359893605e-05, "loss": 2.5292, "step": 8990 }, { "epoch": 0.78, "learning_rate": 7.31568752168382e-05, "loss": 2.6326, "step": 9000 }, { "epoch": 0.78, "learning_rate": 7.315470683474037e-05, "loss": 2.3852, "step": 9010 }, { "epoch": 0.78, "learning_rate": 7.315253845264252e-05, "loss": 2.3974, "step": 9020 }, { "epoch": 0.78, "learning_rate": 7.315037007054469e-05, "loss": 2.5917, "step": 9030 }, { "epoch": 0.78, "learning_rate": 7.314820168844685e-05, "loss": 2.5714, "step": 9040 }, { "epoch": 0.78, "learning_rate": 7.314603330634902e-05, "loss": 2.3732, "step": 9050 }, { "epoch": 0.78, "learning_rate": 7.314386492425119e-05, "loss": 2.4496, "step": 9060 }, { "epoch": 0.79, "learning_rate": 7.314169654215334e-05, "loss": 2.4925, "step": 9070 }, { "epoch": 0.79, "learning_rate": 7.31395281600555e-05, "loss": 2.5439, "step": 9080 }, { "epoch": 0.79, "learning_rate": 7.313735977795766e-05, "loss": 2.4479, "step": 9090 }, { "epoch": 0.79, "learning_rate": 7.313519139585982e-05, "loss": 2.7665, "step": 9100 }, { "epoch": 0.79, "learning_rate": 7.313302301376199e-05, "loss": 2.409, "step": 9110 }, { "epoch": 0.79, "learning_rate": 7.313085463166416e-05, "loss": 2.4556, "step": 9120 }, { "epoch": 0.79, "learning_rate": 7.312868624956632e-05, "loss": 2.5622, "step": 9130 }, { "epoch": 0.79, "learning_rate": 7.312651786746848e-05, "loss": 2.4869, "step": 9140 }, { "epoch": 0.79, "learning_rate": 7.312434948537064e-05, "loss": 2.5899, "step": 9150 }, { "epoch": 0.79, "learning_rate": 7.312218110327281e-05, "loss": 2.5157, "step": 9160 }, { "epoch": 0.79, "learning_rate": 7.312001272117496e-05, "loss": 2.5473, "step": 9170 }, { "epoch": 0.8, "learning_rate": 7.311784433907713e-05, "loss": 2.4717, "step": 9180 }, { "epoch": 0.8, "learning_rate": 7.31156759569793e-05, "loss": 2.508, "step": 9190 }, { "epoch": 0.8, "learning_rate": 7.311350757488146e-05, "loss": 2.4095, "step": 9200 }, { "epoch": 0.8, "learning_rate": 7.311133919278361e-05, "loss": 2.2594, "step": 9210 }, { "epoch": 0.8, "learning_rate": 7.310917081068578e-05, "loss": 2.5682, "step": 9220 }, { "epoch": 0.8, "learning_rate": 7.310700242858795e-05, "loss": 2.3355, "step": 9230 }, { "epoch": 0.8, "learning_rate": 7.31048340464901e-05, "loss": 2.5447, "step": 9240 }, { "epoch": 0.8, "learning_rate": 7.310266566439227e-05, "loss": 2.411, "step": 9250 }, { "epoch": 0.8, "learning_rate": 7.310049728229443e-05, "loss": 2.6761, "step": 9260 }, { "epoch": 0.8, "learning_rate": 7.30983289001966e-05, "loss": 2.463, "step": 9270 }, { "epoch": 0.8, "learning_rate": 7.309616051809875e-05, "loss": 2.4703, "step": 9280 }, { "epoch": 0.8, "learning_rate": 7.309399213600092e-05, "loss": 2.4207, "step": 9290 }, { "epoch": 0.81, "learning_rate": 7.309182375390308e-05, "loss": 2.5605, "step": 9300 }, { "epoch": 0.81, "learning_rate": 7.308965537180525e-05, "loss": 2.6386, "step": 9310 }, { "epoch": 0.81, "learning_rate": 7.30874869897074e-05, "loss": 2.4549, "step": 9320 }, { "epoch": 0.81, "learning_rate": 7.308531860760957e-05, "loss": 2.5876, "step": 9330 }, { "epoch": 0.81, "learning_rate": 7.308315022551174e-05, "loss": 2.5122, "step": 9340 }, { "epoch": 0.81, "learning_rate": 7.308098184341389e-05, "loss": 2.4106, "step": 9350 }, { "epoch": 0.81, "learning_rate": 7.307881346131606e-05, "loss": 2.4654, "step": 9360 }, { "epoch": 0.81, "learning_rate": 7.307664507921822e-05, "loss": 2.5514, "step": 9370 }, { "epoch": 0.81, "learning_rate": 7.307447669712039e-05, "loss": 2.5201, "step": 9380 }, { "epoch": 0.81, "learning_rate": 7.307230831502255e-05, "loss": 2.2305, "step": 9390 }, { "epoch": 0.81, "learning_rate": 7.307013993292471e-05, "loss": 2.54, "step": 9400 }, { "epoch": 0.81, "learning_rate": 7.306797155082687e-05, "loss": 2.2885, "step": 9410 }, { "epoch": 0.82, "learning_rate": 7.306580316872903e-05, "loss": 2.3299, "step": 9420 }, { "epoch": 0.82, "learning_rate": 7.306363478663119e-05, "loss": 2.5243, "step": 9430 }, { "epoch": 0.82, "learning_rate": 7.306146640453336e-05, "loss": 2.2799, "step": 9440 }, { "epoch": 0.82, "learning_rate": 7.305929802243553e-05, "loss": 2.284, "step": 9450 }, { "epoch": 0.82, "learning_rate": 7.305712964033769e-05, "loss": 2.4211, "step": 9460 }, { "epoch": 0.82, "learning_rate": 7.305496125823985e-05, "loss": 2.4308, "step": 9470 }, { "epoch": 0.82, "learning_rate": 7.305279287614201e-05, "loss": 2.4251, "step": 9480 }, { "epoch": 0.82, "learning_rate": 7.305062449404416e-05, "loss": 2.434, "step": 9490 }, { "epoch": 0.82, "learning_rate": 7.304845611194633e-05, "loss": 2.5911, "step": 9500 }, { "epoch": 0.82, "learning_rate": 7.30462877298485e-05, "loss": 2.3312, "step": 9510 }, { "epoch": 0.82, "learning_rate": 7.304411934775066e-05, "loss": 2.563, "step": 9520 }, { "epoch": 0.83, "learning_rate": 7.304195096565283e-05, "loss": 2.3293, "step": 9530 }, { "epoch": 0.83, "learning_rate": 7.303978258355498e-05, "loss": 2.4045, "step": 9540 }, { "epoch": 0.83, "learning_rate": 7.303761420145715e-05, "loss": 2.4978, "step": 9550 }, { "epoch": 0.83, "learning_rate": 7.30354458193593e-05, "loss": 2.5774, "step": 9560 }, { "epoch": 0.83, "learning_rate": 7.303327743726147e-05, "loss": 2.3215, "step": 9570 }, { "epoch": 0.83, "learning_rate": 7.303110905516363e-05, "loss": 2.4437, "step": 9580 }, { "epoch": 0.83, "learning_rate": 7.30289406730658e-05, "loss": 2.4149, "step": 9590 }, { "epoch": 0.83, "learning_rate": 7.302677229096797e-05, "loss": 2.4484, "step": 9600 }, { "epoch": 0.83, "learning_rate": 7.302460390887012e-05, "loss": 2.6747, "step": 9610 }, { "epoch": 0.83, "learning_rate": 7.302243552677229e-05, "loss": 2.5537, "step": 9620 }, { "epoch": 0.83, "learning_rate": 7.302026714467444e-05, "loss": 2.4839, "step": 9630 }, { "epoch": 0.83, "learning_rate": 7.30180987625766e-05, "loss": 2.4304, "step": 9640 }, { "epoch": 0.84, "learning_rate": 7.301593038047877e-05, "loss": 2.4255, "step": 9650 }, { "epoch": 0.84, "learning_rate": 7.301376199838094e-05, "loss": 2.707, "step": 9660 }, { "epoch": 0.84, "learning_rate": 7.301159361628309e-05, "loss": 2.4639, "step": 9670 }, { "epoch": 0.84, "learning_rate": 7.300942523418526e-05, "loss": 2.549, "step": 9680 }, { "epoch": 0.84, "learning_rate": 7.300725685208742e-05, "loss": 2.5638, "step": 9690 }, { "epoch": 0.84, "learning_rate": 7.300508846998959e-05, "loss": 2.4262, "step": 9700 }, { "epoch": 0.84, "learning_rate": 7.300292008789174e-05, "loss": 2.4608, "step": 9710 }, { "epoch": 0.84, "learning_rate": 7.300075170579391e-05, "loss": 2.2461, "step": 9720 }, { "epoch": 0.84, "learning_rate": 7.299858332369608e-05, "loss": 2.6344, "step": 9730 }, { "epoch": 0.84, "learning_rate": 7.299641494159823e-05, "loss": 2.5389, "step": 9740 }, { "epoch": 0.84, "learning_rate": 7.29942465595004e-05, "loss": 2.5669, "step": 9750 }, { "epoch": 0.85, "learning_rate": 7.299207817740256e-05, "loss": 2.4162, "step": 9760 }, { "epoch": 0.85, "learning_rate": 7.298990979530473e-05, "loss": 2.3988, "step": 9770 }, { "epoch": 0.85, "learning_rate": 7.298774141320688e-05, "loss": 2.5187, "step": 9780 }, { "epoch": 0.85, "learning_rate": 7.298557303110905e-05, "loss": 2.4163, "step": 9790 }, { "epoch": 0.85, "learning_rate": 7.298340464901121e-05, "loss": 2.6093, "step": 9800 }, { "epoch": 0.85, "learning_rate": 7.298123626691337e-05, "loss": 2.6156, "step": 9810 }, { "epoch": 0.85, "learning_rate": 7.297906788481553e-05, "loss": 2.4556, "step": 9820 }, { "epoch": 0.85, "learning_rate": 7.29768995027177e-05, "loss": 2.67, "step": 9830 }, { "epoch": 0.85, "learning_rate": 7.297473112061987e-05, "loss": 2.4632, "step": 9840 }, { "epoch": 0.85, "learning_rate": 7.297256273852203e-05, "loss": 2.5759, "step": 9850 }, { "epoch": 0.85, "learning_rate": 7.297039435642419e-05, "loss": 2.2062, "step": 9860 }, { "epoch": 0.85, "learning_rate": 7.296822597432635e-05, "loss": 2.6903, "step": 9870 }, { "epoch": 0.86, "learning_rate": 7.29660575922285e-05, "loss": 2.4141, "step": 9880 }, { "epoch": 0.86, "learning_rate": 7.296388921013067e-05, "loss": 2.4677, "step": 9890 }, { "epoch": 0.86, "learning_rate": 7.296172082803284e-05, "loss": 2.3188, "step": 9900 }, { "epoch": 0.86, "learning_rate": 7.2959552445935e-05, "loss": 2.4032, "step": 9910 }, { "epoch": 0.86, "learning_rate": 7.295738406383717e-05, "loss": 2.4201, "step": 9920 }, { "epoch": 0.86, "learning_rate": 7.295521568173932e-05, "loss": 2.3471, "step": 9930 }, { "epoch": 0.86, "learning_rate": 7.295304729964149e-05, "loss": 2.6448, "step": 9940 }, { "epoch": 0.86, "learning_rate": 7.295087891754364e-05, "loss": 2.3977, "step": 9950 }, { "epoch": 0.86, "learning_rate": 7.294871053544581e-05, "loss": 2.5776, "step": 9960 }, { "epoch": 0.86, "learning_rate": 7.294654215334797e-05, "loss": 2.6108, "step": 9970 }, { "epoch": 0.86, "learning_rate": 7.294437377125014e-05, "loss": 2.2477, "step": 9980 }, { "epoch": 0.87, "learning_rate": 7.294220538915231e-05, "loss": 2.327, "step": 9990 }, { "epoch": 0.87, "learning_rate": 7.294003700705446e-05, "loss": 2.2713, "step": 10000 }, { "epoch": 0.87, "learning_rate": 7.293786862495663e-05, "loss": 2.3322, "step": 10010 }, { "epoch": 0.87, "learning_rate": 7.293570024285878e-05, "loss": 2.3208, "step": 10020 }, { "epoch": 0.87, "learning_rate": 7.293353186076095e-05, "loss": 2.3658, "step": 10030 }, { "epoch": 0.87, "learning_rate": 7.293136347866311e-05, "loss": 2.6031, "step": 10040 }, { "epoch": 0.87, "learning_rate": 7.292919509656528e-05, "loss": 2.328, "step": 10050 }, { "epoch": 0.87, "learning_rate": 7.292702671446745e-05, "loss": 2.3588, "step": 10060 }, { "epoch": 0.87, "learning_rate": 7.29248583323696e-05, "loss": 2.5174, "step": 10070 }, { "epoch": 0.87, "learning_rate": 7.292268995027176e-05, "loss": 2.384, "step": 10080 }, { "epoch": 0.87, "learning_rate": 7.292052156817392e-05, "loss": 2.4347, "step": 10090 }, { "epoch": 0.87, "learning_rate": 7.291835318607608e-05, "loss": 2.3358, "step": 10100 }, { "epoch": 0.88, "learning_rate": 7.291618480397825e-05, "loss": 2.6614, "step": 10110 }, { "epoch": 0.88, "learning_rate": 7.291401642188042e-05, "loss": 2.7445, "step": 10120 }, { "epoch": 0.88, "learning_rate": 7.291184803978258e-05, "loss": 2.5789, "step": 10130 }, { "epoch": 0.88, "learning_rate": 7.290967965768474e-05, "loss": 2.3187, "step": 10140 }, { "epoch": 0.88, "learning_rate": 7.29075112755869e-05, "loss": 2.7078, "step": 10150 }, { "epoch": 0.88, "learning_rate": 7.290534289348907e-05, "loss": 2.4168, "step": 10160 }, { "epoch": 0.88, "learning_rate": 7.290317451139122e-05, "loss": 2.5226, "step": 10170 }, { "epoch": 0.88, "learning_rate": 7.290100612929339e-05, "loss": 2.5073, "step": 10180 }, { "epoch": 0.88, "learning_rate": 7.289883774719555e-05, "loss": 2.195, "step": 10190 }, { "epoch": 0.88, "learning_rate": 7.289666936509772e-05, "loss": 2.3014, "step": 10200 }, { "epoch": 0.88, "learning_rate": 7.289450098299987e-05, "loss": 2.4434, "step": 10210 }, { "epoch": 0.89, "learning_rate": 7.289233260090204e-05, "loss": 2.2289, "step": 10220 }, { "epoch": 0.89, "learning_rate": 7.28901642188042e-05, "loss": 2.3865, "step": 10230 }, { "epoch": 0.89, "learning_rate": 7.288799583670637e-05, "loss": 2.4298, "step": 10240 }, { "epoch": 0.89, "learning_rate": 7.288582745460853e-05, "loss": 2.3905, "step": 10250 }, { "epoch": 0.89, "learning_rate": 7.288365907251069e-05, "loss": 2.4478, "step": 10260 }, { "epoch": 0.89, "learning_rate": 7.288149069041286e-05, "loss": 2.1611, "step": 10270 }, { "epoch": 0.89, "learning_rate": 7.287932230831501e-05, "loss": 2.3296, "step": 10280 }, { "epoch": 0.89, "learning_rate": 7.287715392621718e-05, "loss": 2.2856, "step": 10290 }, { "epoch": 0.89, "learning_rate": 7.287498554411934e-05, "loss": 2.5335, "step": 10300 }, { "epoch": 0.89, "learning_rate": 7.287281716202151e-05, "loss": 2.4135, "step": 10310 }, { "epoch": 0.89, "learning_rate": 7.287064877992366e-05, "loss": 2.196, "step": 10320 }, { "epoch": 0.89, "learning_rate": 7.286848039782583e-05, "loss": 2.2275, "step": 10330 }, { "epoch": 0.9, "learning_rate": 7.2866312015728e-05, "loss": 2.3034, "step": 10340 }, { "epoch": 0.9, "learning_rate": 7.286414363363015e-05, "loss": 2.6955, "step": 10350 }, { "epoch": 0.9, "learning_rate": 7.286197525153232e-05, "loss": 2.4161, "step": 10360 }, { "epoch": 0.9, "learning_rate": 7.285980686943448e-05, "loss": 2.3176, "step": 10370 }, { "epoch": 0.9, "learning_rate": 7.285763848733665e-05, "loss": 2.2054, "step": 10380 }, { "epoch": 0.9, "learning_rate": 7.285547010523881e-05, "loss": 2.426, "step": 10390 }, { "epoch": 0.9, "learning_rate": 7.285330172314097e-05, "loss": 2.2805, "step": 10400 }, { "epoch": 0.9, "learning_rate": 7.285113334104313e-05, "loss": 2.2381, "step": 10410 }, { "epoch": 0.9, "learning_rate": 7.284896495894529e-05, "loss": 2.4698, "step": 10420 }, { "epoch": 0.9, "learning_rate": 7.284679657684745e-05, "loss": 2.2853, "step": 10430 }, { "epoch": 0.9, "learning_rate": 7.284462819474962e-05, "loss": 2.2152, "step": 10440 }, { "epoch": 0.91, "learning_rate": 7.284245981265179e-05, "loss": 2.3083, "step": 10450 }, { "epoch": 0.91, "learning_rate": 7.284029143055395e-05, "loss": 2.4998, "step": 10460 }, { "epoch": 0.91, "learning_rate": 7.28381230484561e-05, "loss": 2.3028, "step": 10470 }, { "epoch": 0.91, "learning_rate": 7.283595466635827e-05, "loss": 2.4158, "step": 10480 }, { "epoch": 0.91, "learning_rate": 7.283378628426042e-05, "loss": 2.2088, "step": 10490 }, { "epoch": 0.91, "learning_rate": 7.283161790216259e-05, "loss": 2.3, "step": 10500 }, { "epoch": 0.91, "learning_rate": 7.282944952006476e-05, "loss": 2.1619, "step": 10510 }, { "epoch": 0.91, "learning_rate": 7.282728113796692e-05, "loss": 2.3271, "step": 10520 }, { "epoch": 0.91, "learning_rate": 7.282511275586909e-05, "loss": 2.4837, "step": 10530 }, { "epoch": 0.91, "learning_rate": 7.282294437377124e-05, "loss": 2.3912, "step": 10540 }, { "epoch": 0.91, "learning_rate": 7.282077599167341e-05, "loss": 2.6127, "step": 10550 }, { "epoch": 0.91, "learning_rate": 7.281860760957556e-05, "loss": 2.3377, "step": 10560 }, { "epoch": 0.92, "learning_rate": 7.281643922747773e-05, "loss": 2.4499, "step": 10570 }, { "epoch": 0.92, "learning_rate": 7.28142708453799e-05, "loss": 2.2509, "step": 10580 }, { "epoch": 0.92, "learning_rate": 7.281210246328206e-05, "loss": 2.2825, "step": 10590 }, { "epoch": 0.92, "learning_rate": 7.280993408118423e-05, "loss": 2.3526, "step": 10600 }, { "epoch": 0.92, "learning_rate": 7.280776569908638e-05, "loss": 2.5947, "step": 10610 }, { "epoch": 0.92, "learning_rate": 7.280559731698855e-05, "loss": 2.3514, "step": 10620 }, { "epoch": 0.92, "learning_rate": 7.28034289348907e-05, "loss": 2.2662, "step": 10630 }, { "epoch": 0.92, "learning_rate": 7.280126055279287e-05, "loss": 2.3863, "step": 10640 }, { "epoch": 0.92, "learning_rate": 7.279909217069503e-05, "loss": 2.2697, "step": 10650 }, { "epoch": 0.92, "learning_rate": 7.27969237885972e-05, "loss": 2.365, "step": 10660 }, { "epoch": 0.92, "learning_rate": 7.279475540649936e-05, "loss": 2.2899, "step": 10670 }, { "epoch": 0.92, "learning_rate": 7.279258702440152e-05, "loss": 2.4646, "step": 10680 }, { "epoch": 0.93, "learning_rate": 7.279041864230368e-05, "loss": 2.447, "step": 10690 }, { "epoch": 0.93, "learning_rate": 7.278825026020585e-05, "loss": 2.5025, "step": 10700 }, { "epoch": 0.93, "learning_rate": 7.2786081878108e-05, "loss": 2.4206, "step": 10710 }, { "epoch": 0.93, "learning_rate": 7.278391349601017e-05, "loss": 2.331, "step": 10720 }, { "epoch": 0.93, "learning_rate": 7.278174511391234e-05, "loss": 2.5421, "step": 10730 }, { "epoch": 0.93, "learning_rate": 7.277957673181449e-05, "loss": 2.293, "step": 10740 }, { "epoch": 0.93, "learning_rate": 7.277740834971666e-05, "loss": 2.7538, "step": 10750 }, { "epoch": 0.93, "learning_rate": 7.277523996761882e-05, "loss": 2.3863, "step": 10760 }, { "epoch": 0.93, "learning_rate": 7.277307158552099e-05, "loss": 2.3471, "step": 10770 }, { "epoch": 0.93, "learning_rate": 7.277090320342314e-05, "loss": 2.5381, "step": 10780 }, { "epoch": 0.93, "learning_rate": 7.276873482132531e-05, "loss": 2.4188, "step": 10790 }, { "epoch": 0.94, "learning_rate": 7.276656643922747e-05, "loss": 2.4586, "step": 10800 }, { "epoch": 0.94, "learning_rate": 7.276439805712963e-05, "loss": 2.3782, "step": 10810 }, { "epoch": 0.94, "learning_rate": 7.276222967503179e-05, "loss": 2.3006, "step": 10820 }, { "epoch": 0.94, "learning_rate": 7.276006129293396e-05, "loss": 2.3748, "step": 10830 }, { "epoch": 0.94, "learning_rate": 7.275789291083613e-05, "loss": 2.4911, "step": 10840 }, { "epoch": 0.94, "learning_rate": 7.275572452873829e-05, "loss": 2.6194, "step": 10850 }, { "epoch": 0.94, "learning_rate": 7.275355614664044e-05, "loss": 2.0594, "step": 10860 }, { "epoch": 0.94, "learning_rate": 7.275138776454261e-05, "loss": 2.4816, "step": 10870 }, { "epoch": 0.94, "learning_rate": 7.274921938244476e-05, "loss": 2.2823, "step": 10880 }, { "epoch": 0.94, "learning_rate": 7.274705100034693e-05, "loss": 2.5641, "step": 10890 }, { "epoch": 0.94, "learning_rate": 7.27448826182491e-05, "loss": 2.3601, "step": 10900 }, { "epoch": 0.94, "learning_rate": 7.274271423615126e-05, "loss": 2.2152, "step": 10910 }, { "epoch": 0.95, "learning_rate": 7.274054585405343e-05, "loss": 2.4627, "step": 10920 }, { "epoch": 0.95, "learning_rate": 7.273837747195558e-05, "loss": 2.2316, "step": 10930 }, { "epoch": 0.95, "learning_rate": 7.273620908985775e-05, "loss": 2.2364, "step": 10940 }, { "epoch": 0.95, "learning_rate": 7.27340407077599e-05, "loss": 2.6584, "step": 10950 }, { "epoch": 0.95, "learning_rate": 7.273187232566207e-05, "loss": 2.4595, "step": 10960 }, { "epoch": 0.95, "learning_rate": 7.272970394356423e-05, "loss": 2.3589, "step": 10970 }, { "epoch": 0.95, "learning_rate": 7.27275355614664e-05, "loss": 2.2422, "step": 10980 }, { "epoch": 0.95, "learning_rate": 7.272536717936857e-05, "loss": 2.4273, "step": 10990 }, { "epoch": 0.95, "learning_rate": 7.272319879727072e-05, "loss": 2.3288, "step": 11000 }, { "epoch": 0.95, "learning_rate": 7.272103041517289e-05, "loss": 2.3508, "step": 11010 }, { "epoch": 0.95, "learning_rate": 7.271886203307504e-05, "loss": 2.3355, "step": 11020 }, { "epoch": 0.96, "learning_rate": 7.27166936509772e-05, "loss": 2.7278, "step": 11030 }, { "epoch": 0.96, "learning_rate": 7.271452526887937e-05, "loss": 2.2532, "step": 11040 }, { "epoch": 0.96, "learning_rate": 7.271235688678154e-05, "loss": 2.3818, "step": 11050 }, { "epoch": 0.96, "learning_rate": 7.27101885046837e-05, "loss": 2.6214, "step": 11060 }, { "epoch": 0.96, "learning_rate": 7.270802012258586e-05, "loss": 2.2967, "step": 11070 }, { "epoch": 0.96, "learning_rate": 7.270585174048802e-05, "loss": 2.32, "step": 11080 }, { "epoch": 0.96, "learning_rate": 7.270368335839018e-05, "loss": 2.3077, "step": 11090 }, { "epoch": 0.96, "learning_rate": 7.270151497629234e-05, "loss": 2.3925, "step": 11100 }, { "epoch": 0.96, "learning_rate": 7.269934659419451e-05, "loss": 2.0784, "step": 11110 }, { "epoch": 0.96, "learning_rate": 7.269717821209668e-05, "loss": 2.234, "step": 11120 }, { "epoch": 0.96, "learning_rate": 7.269500982999884e-05, "loss": 2.6075, "step": 11130 }, { "epoch": 0.96, "learning_rate": 7.2692841447901e-05, "loss": 2.5371, "step": 11140 }, { "epoch": 0.97, "learning_rate": 7.269067306580316e-05, "loss": 2.4469, "step": 11150 }, { "epoch": 0.97, "learning_rate": 7.268850468370533e-05, "loss": 2.4983, "step": 11160 }, { "epoch": 0.97, "learning_rate": 7.268633630160748e-05, "loss": 2.4924, "step": 11170 }, { "epoch": 0.97, "learning_rate": 7.268416791950965e-05, "loss": 2.3893, "step": 11180 }, { "epoch": 0.97, "learning_rate": 7.268199953741181e-05, "loss": 2.1413, "step": 11190 }, { "epoch": 0.97, "learning_rate": 7.267983115531398e-05, "loss": 2.5606, "step": 11200 }, { "epoch": 0.97, "learning_rate": 7.267766277321613e-05, "loss": 2.3859, "step": 11210 }, { "epoch": 0.97, "learning_rate": 7.26754943911183e-05, "loss": 2.3262, "step": 11220 }, { "epoch": 0.97, "learning_rate": 7.267332600902047e-05, "loss": 2.5273, "step": 11230 }, { "epoch": 0.97, "learning_rate": 7.267115762692263e-05, "loss": 2.3642, "step": 11240 }, { "epoch": 0.97, "learning_rate": 7.266898924482479e-05, "loss": 2.1965, "step": 11250 }, { "epoch": 0.98, "learning_rate": 7.266682086272695e-05, "loss": 2.4616, "step": 11260 }, { "epoch": 0.98, "learning_rate": 7.266465248062912e-05, "loss": 2.4421, "step": 11270 }, { "epoch": 0.98, "learning_rate": 7.266248409853127e-05, "loss": 2.3068, "step": 11280 }, { "epoch": 0.98, "learning_rate": 7.266031571643344e-05, "loss": 2.351, "step": 11290 }, { "epoch": 0.98, "learning_rate": 7.26581473343356e-05, "loss": 2.2534, "step": 11300 }, { "epoch": 0.98, "learning_rate": 7.265597895223777e-05, "loss": 2.5675, "step": 11310 }, { "epoch": 0.98, "learning_rate": 7.265381057013992e-05, "loss": 2.2698, "step": 11320 }, { "epoch": 0.98, "learning_rate": 7.265164218804209e-05, "loss": 2.383, "step": 11330 }, { "epoch": 0.98, "learning_rate": 7.264947380594426e-05, "loss": 2.1802, "step": 11340 }, { "epoch": 0.98, "learning_rate": 7.264730542384641e-05, "loss": 2.1892, "step": 11350 }, { "epoch": 0.98, "learning_rate": 7.264513704174857e-05, "loss": 2.2678, "step": 11360 }, { "epoch": 0.98, "learning_rate": 7.264296865965074e-05, "loss": 2.1562, "step": 11370 }, { "epoch": 0.99, "learning_rate": 7.264080027755291e-05, "loss": 1.9664, "step": 11380 }, { "epoch": 0.99, "learning_rate": 7.263863189545507e-05, "loss": 2.3075, "step": 11390 }, { "epoch": 0.99, "learning_rate": 7.263646351335723e-05, "loss": 2.3432, "step": 11400 }, { "epoch": 0.99, "learning_rate": 7.263429513125939e-05, "loss": 2.2382, "step": 11410 }, { "epoch": 0.99, "learning_rate": 7.263212674916155e-05, "loss": 2.3751, "step": 11420 }, { "epoch": 0.99, "learning_rate": 7.262995836706371e-05, "loss": 2.4887, "step": 11430 }, { "epoch": 0.99, "learning_rate": 7.262778998496588e-05, "loss": 2.6273, "step": 11440 }, { "epoch": 0.99, "learning_rate": 7.262562160286805e-05, "loss": 2.1826, "step": 11450 }, { "epoch": 0.99, "learning_rate": 7.262345322077021e-05, "loss": 2.3884, "step": 11460 }, { "epoch": 0.99, "learning_rate": 7.262128483867236e-05, "loss": 2.3594, "step": 11470 }, { "epoch": 0.99, "learning_rate": 7.261911645657453e-05, "loss": 2.3217, "step": 11480 }, { "epoch": 1.0, "learning_rate": 7.261694807447668e-05, "loss": 2.2663, "step": 11490 }, { "epoch": 1.0, "learning_rate": 7.261477969237885e-05, "loss": 2.468, "step": 11500 }, { "epoch": 1.0, "learning_rate": 7.261261131028102e-05, "loss": 2.2943, "step": 11510 }, { "epoch": 1.0, "learning_rate": 7.261044292818318e-05, "loss": 2.5087, "step": 11520 }, { "epoch": 1.0, "learning_rate": 7.260827454608535e-05, "loss": 2.5396, "step": 11530 }, { "epoch": 1.0, "learning_rate": 7.26061061639875e-05, "loss": 2.3572, "step": 11540 }, { "epoch": 1.0, "eval_cer": 0.5563959113468189, "eval_em": 0, "eval_f1": 0, "eval_loss": 2.063237190246582, "eval_runtime": 285.0649, "eval_samples_per_second": 6.802, "eval_steps_per_second": 0.852, "eval_wer": 1.0, "step": 11546 }, { "epoch": 1.0, "learning_rate": 7.260393778188967e-05, "loss": 2.3242, "step": 11550 }, { "epoch": 1.0, "learning_rate": 7.260176939979182e-05, "loss": 2.372, "step": 11560 }, { "epoch": 1.0, "learning_rate": 7.259960101769399e-05, "loss": 2.2653, "step": 11570 }, { "epoch": 1.0, "learning_rate": 7.259743263559615e-05, "loss": 2.3281, "step": 11580 }, { "epoch": 1.0, "learning_rate": 7.259526425349832e-05, "loss": 2.3027, "step": 11590 }, { "epoch": 1.0, "learning_rate": 7.259309587140049e-05, "loss": 2.3061, "step": 11600 }, { "epoch": 1.01, "learning_rate": 7.259092748930264e-05, "loss": 2.3538, "step": 11610 }, { "epoch": 1.01, "learning_rate": 7.25887591072048e-05, "loss": 2.3014, "step": 11620 }, { "epoch": 1.01, "learning_rate": 7.258659072510696e-05, "loss": 2.3987, "step": 11630 }, { "epoch": 1.01, "learning_rate": 7.258442234300913e-05, "loss": 2.2108, "step": 11640 }, { "epoch": 1.01, "learning_rate": 7.258225396091129e-05, "loss": 2.3365, "step": 11650 }, { "epoch": 1.01, "learning_rate": 7.258008557881346e-05, "loss": 2.3254, "step": 11660 }, { "epoch": 1.01, "learning_rate": 7.257791719671562e-05, "loss": 2.3328, "step": 11670 }, { "epoch": 1.01, "learning_rate": 7.257574881461778e-05, "loss": 2.5662, "step": 11680 }, { "epoch": 1.01, "learning_rate": 7.257358043251994e-05, "loss": 2.2034, "step": 11690 }, { "epoch": 1.01, "learning_rate": 7.257141205042211e-05, "loss": 2.234, "step": 11700 }, { "epoch": 1.01, "learning_rate": 7.256924366832426e-05, "loss": 2.1269, "step": 11710 }, { "epoch": 1.02, "learning_rate": 7.256707528622643e-05, "loss": 2.5595, "step": 11720 }, { "epoch": 1.02, "learning_rate": 7.25649069041286e-05, "loss": 2.3009, "step": 11730 }, { "epoch": 1.02, "learning_rate": 7.256273852203076e-05, "loss": 2.4485, "step": 11740 }, { "epoch": 1.02, "learning_rate": 7.256057013993291e-05, "loss": 2.3818, "step": 11750 }, { "epoch": 1.02, "learning_rate": 7.255840175783508e-05, "loss": 2.4514, "step": 11760 }, { "epoch": 1.02, "learning_rate": 7.255623337573725e-05, "loss": 2.3976, "step": 11770 }, { "epoch": 1.02, "learning_rate": 7.255406499363941e-05, "loss": 2.3529, "step": 11780 }, { "epoch": 1.02, "learning_rate": 7.255189661154157e-05, "loss": 2.2323, "step": 11790 }, { "epoch": 1.02, "learning_rate": 7.254972822944373e-05, "loss": 2.3197, "step": 11800 }, { "epoch": 1.02, "learning_rate": 7.254755984734589e-05, "loss": 2.2392, "step": 11810 }, { "epoch": 1.02, "learning_rate": 7.254539146524805e-05, "loss": 2.2482, "step": 11820 }, { "epoch": 1.02, "learning_rate": 7.254322308315022e-05, "loss": 2.5098, "step": 11830 }, { "epoch": 1.03, "learning_rate": 7.254105470105239e-05, "loss": 2.414, "step": 11840 }, { "epoch": 1.03, "learning_rate": 7.253888631895455e-05, "loss": 2.2325, "step": 11850 }, { "epoch": 1.03, "learning_rate": 7.25367179368567e-05, "loss": 2.3714, "step": 11860 }, { "epoch": 1.03, "learning_rate": 7.253454955475887e-05, "loss": 2.1954, "step": 11870 }, { "epoch": 1.03, "learning_rate": 7.253238117266102e-05, "loss": 2.1671, "step": 11880 }, { "epoch": 1.03, "learning_rate": 7.253021279056319e-05, "loss": 2.3915, "step": 11890 }, { "epoch": 1.03, "learning_rate": 7.252804440846536e-05, "loss": 2.5285, "step": 11900 }, { "epoch": 1.03, "learning_rate": 7.252587602636752e-05, "loss": 2.2772, "step": 11910 }, { "epoch": 1.03, "learning_rate": 7.252370764426969e-05, "loss": 2.2163, "step": 11920 }, { "epoch": 1.03, "learning_rate": 7.252153926217184e-05, "loss": 2.1866, "step": 11930 }, { "epoch": 1.03, "learning_rate": 7.251937088007401e-05, "loss": 2.3302, "step": 11940 }, { "epoch": 1.03, "learning_rate": 7.251720249797616e-05, "loss": 2.0387, "step": 11950 }, { "epoch": 1.04, "learning_rate": 7.251503411587833e-05, "loss": 2.4906, "step": 11960 }, { "epoch": 1.04, "learning_rate": 7.25128657337805e-05, "loss": 2.2662, "step": 11970 }, { "epoch": 1.04, "learning_rate": 7.251069735168266e-05, "loss": 2.1155, "step": 11980 }, { "epoch": 1.04, "learning_rate": 7.250852896958483e-05, "loss": 2.3574, "step": 11990 }, { "epoch": 1.04, "learning_rate": 7.250636058748698e-05, "loss": 2.004, "step": 12000 }, { "epoch": 1.04, "learning_rate": 7.250419220538915e-05, "loss": 1.9951, "step": 12010 }, { "epoch": 1.04, "learning_rate": 7.25020238232913e-05, "loss": 2.3132, "step": 12020 }, { "epoch": 1.04, "learning_rate": 7.249985544119347e-05, "loss": 2.2264, "step": 12030 }, { "epoch": 1.04, "learning_rate": 7.249768705909563e-05, "loss": 2.4038, "step": 12040 }, { "epoch": 1.04, "learning_rate": 7.24955186769978e-05, "loss": 2.156, "step": 12050 }, { "epoch": 1.04, "learning_rate": 7.249335029489996e-05, "loss": 2.0514, "step": 12060 }, { "epoch": 1.05, "learning_rate": 7.249118191280212e-05, "loss": 2.1029, "step": 12070 }, { "epoch": 1.05, "learning_rate": 7.248901353070428e-05, "loss": 2.3221, "step": 12080 }, { "epoch": 1.05, "learning_rate": 7.248684514860644e-05, "loss": 2.3531, "step": 12090 }, { "epoch": 1.05, "learning_rate": 7.24846767665086e-05, "loss": 2.2629, "step": 12100 }, { "epoch": 1.05, "learning_rate": 7.248250838441077e-05, "loss": 2.2139, "step": 12110 }, { "epoch": 1.05, "learning_rate": 7.248034000231294e-05, "loss": 2.2371, "step": 12120 }, { "epoch": 1.05, "learning_rate": 7.24781716202151e-05, "loss": 2.2491, "step": 12130 }, { "epoch": 1.05, "learning_rate": 7.247600323811726e-05, "loss": 2.2006, "step": 12140 }, { "epoch": 1.05, "learning_rate": 7.247383485601942e-05, "loss": 2.4795, "step": 12150 }, { "epoch": 1.05, "learning_rate": 7.247166647392159e-05, "loss": 2.1825, "step": 12160 }, { "epoch": 1.05, "learning_rate": 7.246949809182374e-05, "loss": 2.3763, "step": 12170 }, { "epoch": 1.05, "learning_rate": 7.246732970972591e-05, "loss": 2.0824, "step": 12180 }, { "epoch": 1.06, "learning_rate": 7.246516132762807e-05, "loss": 2.154, "step": 12190 }, { "epoch": 1.06, "learning_rate": 7.246299294553024e-05, "loss": 2.2186, "step": 12200 }, { "epoch": 1.06, "learning_rate": 7.246082456343239e-05, "loss": 2.2157, "step": 12210 }, { "epoch": 1.06, "learning_rate": 7.245865618133456e-05, "loss": 1.9759, "step": 12220 }, { "epoch": 1.06, "learning_rate": 7.245648779923673e-05, "loss": 2.2342, "step": 12230 }, { "epoch": 1.06, "learning_rate": 7.245431941713889e-05, "loss": 2.3008, "step": 12240 }, { "epoch": 1.06, "learning_rate": 7.245215103504104e-05, "loss": 2.3261, "step": 12250 }, { "epoch": 1.06, "learning_rate": 7.244998265294321e-05, "loss": 2.4928, "step": 12260 }, { "epoch": 1.06, "learning_rate": 7.244781427084538e-05, "loss": 2.2851, "step": 12270 }, { "epoch": 1.06, "learning_rate": 7.244564588874753e-05, "loss": 2.3432, "step": 12280 }, { "epoch": 1.06, "learning_rate": 7.24434775066497e-05, "loss": 2.2331, "step": 12290 }, { "epoch": 1.07, "learning_rate": 7.244130912455186e-05, "loss": 2.1339, "step": 12300 }, { "epoch": 1.07, "learning_rate": 7.243914074245403e-05, "loss": 2.3545, "step": 12310 }, { "epoch": 1.07, "learning_rate": 7.243697236035618e-05, "loss": 2.3472, "step": 12320 }, { "epoch": 1.07, "learning_rate": 7.243480397825835e-05, "loss": 2.1639, "step": 12330 }, { "epoch": 1.07, "learning_rate": 7.243263559616052e-05, "loss": 2.5696, "step": 12340 }, { "epoch": 1.07, "learning_rate": 7.243046721406267e-05, "loss": 2.406, "step": 12350 }, { "epoch": 1.07, "learning_rate": 7.242829883196483e-05, "loss": 2.3381, "step": 12360 }, { "epoch": 1.07, "learning_rate": 7.2426130449867e-05, "loss": 2.2713, "step": 12370 }, { "epoch": 1.07, "learning_rate": 7.242396206776917e-05, "loss": 2.1591, "step": 12380 }, { "epoch": 1.07, "learning_rate": 7.242179368567133e-05, "loss": 2.2541, "step": 12390 }, { "epoch": 1.07, "learning_rate": 7.241962530357349e-05, "loss": 2.4114, "step": 12400 }, { "epoch": 1.07, "learning_rate": 7.241745692147565e-05, "loss": 2.3038, "step": 12410 }, { "epoch": 1.08, "learning_rate": 7.24152885393778e-05, "loss": 2.2576, "step": 12420 }, { "epoch": 1.08, "learning_rate": 7.241312015727997e-05, "loss": 2.3649, "step": 12430 }, { "epoch": 1.08, "learning_rate": 7.241095177518214e-05, "loss": 2.162, "step": 12440 }, { "epoch": 1.08, "learning_rate": 7.24087833930843e-05, "loss": 2.4632, "step": 12450 }, { "epoch": 1.08, "learning_rate": 7.240661501098647e-05, "loss": 2.3664, "step": 12460 }, { "epoch": 1.08, "learning_rate": 7.240444662888862e-05, "loss": 2.1529, "step": 12470 }, { "epoch": 1.08, "learning_rate": 7.240227824679079e-05, "loss": 2.4455, "step": 12480 }, { "epoch": 1.08, "learning_rate": 7.240010986469294e-05, "loss": 2.3604, "step": 12490 }, { "epoch": 1.08, "learning_rate": 7.239794148259511e-05, "loss": 2.3725, "step": 12500 }, { "epoch": 1.08, "learning_rate": 7.239577310049728e-05, "loss": 2.3069, "step": 12510 }, { "epoch": 1.08, "learning_rate": 7.239360471839944e-05, "loss": 2.1104, "step": 12520 }, { "epoch": 1.09, "learning_rate": 7.239143633630161e-05, "loss": 2.2989, "step": 12530 }, { "epoch": 1.09, "learning_rate": 7.238926795420376e-05, "loss": 2.182, "step": 12540 }, { "epoch": 1.09, "learning_rate": 7.238709957210593e-05, "loss": 2.2397, "step": 12550 }, { "epoch": 1.09, "learning_rate": 7.238493119000808e-05, "loss": 2.1972, "step": 12560 }, { "epoch": 1.09, "learning_rate": 7.238276280791025e-05, "loss": 2.2661, "step": 12570 }, { "epoch": 1.09, "learning_rate": 7.238059442581241e-05, "loss": 2.1699, "step": 12580 }, { "epoch": 1.09, "learning_rate": 7.237842604371458e-05, "loss": 2.2986, "step": 12590 }, { "epoch": 1.09, "learning_rate": 7.237625766161675e-05, "loss": 2.1467, "step": 12600 }, { "epoch": 1.09, "learning_rate": 7.23740892795189e-05, "loss": 2.0823, "step": 12610 }, { "epoch": 1.09, "learning_rate": 7.237192089742107e-05, "loss": 2.2643, "step": 12620 }, { "epoch": 1.09, "learning_rate": 7.236975251532322e-05, "loss": 2.3259, "step": 12630 }, { "epoch": 1.09, "learning_rate": 7.236758413322539e-05, "loss": 2.1863, "step": 12640 }, { "epoch": 1.1, "learning_rate": 7.236541575112755e-05, "loss": 2.2864, "step": 12650 }, { "epoch": 1.1, "learning_rate": 7.236324736902972e-05, "loss": 2.2679, "step": 12660 }, { "epoch": 1.1, "learning_rate": 7.236107898693188e-05, "loss": 2.3555, "step": 12670 }, { "epoch": 1.1, "learning_rate": 7.235891060483404e-05, "loss": 2.2431, "step": 12680 }, { "epoch": 1.1, "learning_rate": 7.23567422227362e-05, "loss": 2.3403, "step": 12690 }, { "epoch": 1.1, "learning_rate": 7.235457384063837e-05, "loss": 2.1672, "step": 12700 }, { "epoch": 1.1, "learning_rate": 7.235240545854052e-05, "loss": 2.1562, "step": 12710 }, { "epoch": 1.1, "learning_rate": 7.235023707644269e-05, "loss": 2.394, "step": 12720 }, { "epoch": 1.1, "learning_rate": 7.234806869434486e-05, "loss": 2.2972, "step": 12730 }, { "epoch": 1.1, "learning_rate": 7.234590031224702e-05, "loss": 2.2802, "step": 12740 }, { "epoch": 1.1, "learning_rate": 7.234373193014917e-05, "loss": 2.2916, "step": 12750 }, { "epoch": 1.11, "learning_rate": 7.234156354805134e-05, "loss": 2.482, "step": 12760 }, { "epoch": 1.11, "learning_rate": 7.233939516595351e-05, "loss": 2.282, "step": 12770 }, { "epoch": 1.11, "learning_rate": 7.233722678385567e-05, "loss": 2.1639, "step": 12780 }, { "epoch": 1.11, "learning_rate": 7.233505840175783e-05, "loss": 2.4264, "step": 12790 }, { "epoch": 1.11, "learning_rate": 7.233289001965999e-05, "loss": 2.3386, "step": 12800 }, { "epoch": 1.11, "learning_rate": 7.233072163756216e-05, "loss": 1.861, "step": 12810 }, { "epoch": 1.11, "learning_rate": 7.232855325546431e-05, "loss": 2.2051, "step": 12820 }, { "epoch": 1.11, "learning_rate": 7.232638487336648e-05, "loss": 2.1565, "step": 12830 }, { "epoch": 1.11, "learning_rate": 7.232421649126865e-05, "loss": 1.9876, "step": 12840 }, { "epoch": 1.11, "learning_rate": 7.232204810917081e-05, "loss": 2.3944, "step": 12850 }, { "epoch": 1.11, "learning_rate": 7.231987972707296e-05, "loss": 2.485, "step": 12860 }, { "epoch": 1.11, "learning_rate": 7.231771134497513e-05, "loss": 2.2623, "step": 12870 }, { "epoch": 1.12, "learning_rate": 7.231554296287728e-05, "loss": 2.339, "step": 12880 }, { "epoch": 1.12, "learning_rate": 7.231337458077945e-05, "loss": 2.4339, "step": 12890 }, { "epoch": 1.12, "learning_rate": 7.231120619868162e-05, "loss": 2.1471, "step": 12900 }, { "epoch": 1.12, "learning_rate": 7.230903781658378e-05, "loss": 2.2941, "step": 12910 }, { "epoch": 1.12, "learning_rate": 7.230686943448595e-05, "loss": 2.3087, "step": 12920 }, { "epoch": 1.12, "learning_rate": 7.23047010523881e-05, "loss": 2.0894, "step": 12930 }, { "epoch": 1.12, "learning_rate": 7.230253267029027e-05, "loss": 2.3048, "step": 12940 }, { "epoch": 1.12, "learning_rate": 7.230036428819242e-05, "loss": 2.4382, "step": 12950 }, { "epoch": 1.12, "learning_rate": 7.229819590609459e-05, "loss": 2.4554, "step": 12960 }, { "epoch": 1.12, "learning_rate": 7.229602752399675e-05, "loss": 2.3955, "step": 12970 }, { "epoch": 1.12, "learning_rate": 7.229385914189892e-05, "loss": 2.5451, "step": 12980 }, { "epoch": 1.13, "learning_rate": 7.229169075980109e-05, "loss": 2.1321, "step": 12990 }, { "epoch": 1.13, "learning_rate": 7.228952237770324e-05, "loss": 2.1825, "step": 13000 }, { "epoch": 1.13, "learning_rate": 7.22873539956054e-05, "loss": 2.105, "step": 13010 }, { "epoch": 1.13, "learning_rate": 7.228518561350756e-05, "loss": 2.5431, "step": 13020 }, { "epoch": 1.13, "learning_rate": 7.228301723140973e-05, "loss": 2.067, "step": 13030 }, { "epoch": 1.13, "learning_rate": 7.228084884931189e-05, "loss": 2.3499, "step": 13040 }, { "epoch": 1.13, "learning_rate": 7.227868046721406e-05, "loss": 2.1598, "step": 13050 }, { "epoch": 1.13, "learning_rate": 7.227651208511622e-05, "loss": 2.3941, "step": 13060 }, { "epoch": 1.13, "learning_rate": 7.227434370301838e-05, "loss": 2.5236, "step": 13070 }, { "epoch": 1.13, "learning_rate": 7.227217532092054e-05, "loss": 2.352, "step": 13080 }, { "epoch": 1.13, "learning_rate": 7.22700069388227e-05, "loss": 2.3686, "step": 13090 }, { "epoch": 1.13, "learning_rate": 7.226783855672486e-05, "loss": 2.2647, "step": 13100 }, { "epoch": 1.14, "learning_rate": 7.226567017462703e-05, "loss": 2.0961, "step": 13110 }, { "epoch": 1.14, "learning_rate": 7.22635017925292e-05, "loss": 2.3604, "step": 13120 }, { "epoch": 1.14, "learning_rate": 7.226133341043136e-05, "loss": 2.009, "step": 13130 }, { "epoch": 1.14, "learning_rate": 7.225916502833351e-05, "loss": 2.3779, "step": 13140 }, { "epoch": 1.14, "learning_rate": 7.225699664623568e-05, "loss": 2.0044, "step": 13150 }, { "epoch": 1.14, "learning_rate": 7.225482826413785e-05, "loss": 2.314, "step": 13160 }, { "epoch": 1.14, "learning_rate": 7.225265988204e-05, "loss": 2.2335, "step": 13170 }, { "epoch": 1.14, "learning_rate": 7.225049149994217e-05, "loss": 2.1952, "step": 13180 }, { "epoch": 1.14, "learning_rate": 7.224832311784433e-05, "loss": 2.3097, "step": 13190 }, { "epoch": 1.14, "learning_rate": 7.22461547357465e-05, "loss": 2.1328, "step": 13200 }, { "epoch": 1.14, "learning_rate": 7.224398635364865e-05, "loss": 2.1839, "step": 13210 }, { "epoch": 1.14, "learning_rate": 7.224181797155082e-05, "loss": 2.4987, "step": 13220 }, { "epoch": 1.15, "learning_rate": 7.223964958945299e-05, "loss": 2.2262, "step": 13230 }, { "epoch": 1.15, "learning_rate": 7.223748120735515e-05, "loss": 2.1396, "step": 13240 }, { "epoch": 1.15, "learning_rate": 7.22353128252573e-05, "loss": 2.1645, "step": 13250 }, { "epoch": 1.15, "learning_rate": 7.223314444315947e-05, "loss": 2.2137, "step": 13260 }, { "epoch": 1.15, "learning_rate": 7.223097606106164e-05, "loss": 1.9098, "step": 13270 }, { "epoch": 1.15, "learning_rate": 7.222880767896379e-05, "loss": 2.3809, "step": 13280 }, { "epoch": 1.15, "learning_rate": 7.222663929686596e-05, "loss": 2.1961, "step": 13290 }, { "epoch": 1.15, "learning_rate": 7.222447091476812e-05, "loss": 2.3911, "step": 13300 }, { "epoch": 1.15, "learning_rate": 7.222230253267029e-05, "loss": 2.2707, "step": 13310 }, { "epoch": 1.15, "learning_rate": 7.222013415057246e-05, "loss": 1.9873, "step": 13320 }, { "epoch": 1.15, "learning_rate": 7.221796576847461e-05, "loss": 2.1933, "step": 13330 }, { "epoch": 1.16, "learning_rate": 7.221579738637677e-05, "loss": 2.4865, "step": 13340 }, { "epoch": 1.16, "learning_rate": 7.221362900427893e-05, "loss": 2.4283, "step": 13350 }, { "epoch": 1.16, "learning_rate": 7.22114606221811e-05, "loss": 2.2129, "step": 13360 }, { "epoch": 1.16, "learning_rate": 7.220929224008326e-05, "loss": 2.2501, "step": 13370 }, { "epoch": 1.16, "learning_rate": 7.220712385798543e-05, "loss": 2.1765, "step": 13380 }, { "epoch": 1.16, "learning_rate": 7.22049554758876e-05, "loss": 2.4736, "step": 13390 }, { "epoch": 1.16, "learning_rate": 7.220278709378975e-05, "loss": 2.1525, "step": 13400 }, { "epoch": 1.16, "learning_rate": 7.220061871169191e-05, "loss": 2.3589, "step": 13410 }, { "epoch": 1.16, "learning_rate": 7.219845032959407e-05, "loss": 2.5337, "step": 13420 }, { "epoch": 1.16, "learning_rate": 7.219628194749623e-05, "loss": 2.49, "step": 13430 }, { "epoch": 1.16, "learning_rate": 7.21941135653984e-05, "loss": 2.2036, "step": 13440 }, { "epoch": 1.16, "learning_rate": 7.219194518330056e-05, "loss": 2.4184, "step": 13450 }, { "epoch": 1.17, "learning_rate": 7.218977680120273e-05, "loss": 2.0361, "step": 13460 }, { "epoch": 1.17, "learning_rate": 7.218760841910488e-05, "loss": 2.3077, "step": 13470 }, { "epoch": 1.17, "learning_rate": 7.218544003700705e-05, "loss": 1.9989, "step": 13480 }, { "epoch": 1.17, "learning_rate": 7.21832716549092e-05, "loss": 2.1538, "step": 13490 }, { "epoch": 1.17, "learning_rate": 7.218110327281137e-05, "loss": 2.087, "step": 13500 }, { "epoch": 1.17, "learning_rate": 7.217893489071354e-05, "loss": 2.2446, "step": 13510 }, { "epoch": 1.17, "learning_rate": 7.21767665086157e-05, "loss": 2.1652, "step": 13520 }, { "epoch": 1.17, "learning_rate": 7.217459812651787e-05, "loss": 2.1393, "step": 13530 }, { "epoch": 1.17, "learning_rate": 7.217242974442002e-05, "loss": 2.2459, "step": 13540 }, { "epoch": 1.17, "learning_rate": 7.217026136232219e-05, "loss": 2.307, "step": 13550 }, { "epoch": 1.17, "learning_rate": 7.216809298022434e-05, "loss": 2.1535, "step": 13560 }, { "epoch": 1.18, "learning_rate": 7.216592459812651e-05, "loss": 2.1709, "step": 13570 }, { "epoch": 1.18, "learning_rate": 7.216375621602867e-05, "loss": 2.2543, "step": 13580 }, { "epoch": 1.18, "learning_rate": 7.216158783393084e-05, "loss": 2.2291, "step": 13590 }, { "epoch": 1.18, "learning_rate": 7.2159419451833e-05, "loss": 2.3412, "step": 13600 }, { "epoch": 1.18, "learning_rate": 7.215725106973516e-05, "loss": 2.3428, "step": 13610 }, { "epoch": 1.18, "learning_rate": 7.215508268763733e-05, "loss": 2.3759, "step": 13620 }, { "epoch": 1.18, "learning_rate": 7.215291430553948e-05, "loss": 2.3345, "step": 13630 }, { "epoch": 1.18, "learning_rate": 7.215074592344164e-05, "loss": 2.3006, "step": 13640 }, { "epoch": 1.18, "learning_rate": 7.214857754134381e-05, "loss": 1.9563, "step": 13650 }, { "epoch": 1.18, "learning_rate": 7.214640915924598e-05, "loss": 2.1898, "step": 13660 }, { "epoch": 1.18, "learning_rate": 7.214424077714814e-05, "loss": 2.2925, "step": 13670 }, { "epoch": 1.18, "learning_rate": 7.21420723950503e-05, "loss": 2.0537, "step": 13680 }, { "epoch": 1.19, "learning_rate": 7.213990401295246e-05, "loss": 2.3641, "step": 13690 }, { "epoch": 1.19, "learning_rate": 7.213773563085463e-05, "loss": 2.1275, "step": 13700 }, { "epoch": 1.19, "learning_rate": 7.213556724875678e-05, "loss": 2.2752, "step": 13710 }, { "epoch": 1.19, "learning_rate": 7.213339886665895e-05, "loss": 2.2461, "step": 13720 }, { "epoch": 1.19, "learning_rate": 7.213123048456112e-05, "loss": 2.31, "step": 13730 }, { "epoch": 1.19, "learning_rate": 7.212906210246328e-05, "loss": 2.198, "step": 13740 }, { "epoch": 1.19, "learning_rate": 7.212689372036543e-05, "loss": 2.1453, "step": 13750 }, { "epoch": 1.19, "learning_rate": 7.21247253382676e-05, "loss": 2.2775, "step": 13760 }, { "epoch": 1.19, "learning_rate": 7.212255695616977e-05, "loss": 2.1323, "step": 13770 }, { "epoch": 1.19, "learning_rate": 7.212038857407193e-05, "loss": 2.301, "step": 13780 }, { "epoch": 1.19, "learning_rate": 7.211822019197409e-05, "loss": 2.246, "step": 13790 }, { "epoch": 1.2, "learning_rate": 7.211605180987625e-05, "loss": 2.1769, "step": 13800 }, { "epoch": 1.2, "learning_rate": 7.211388342777842e-05, "loss": 2.0959, "step": 13810 }, { "epoch": 1.2, "learning_rate": 7.211171504568057e-05, "loss": 2.1551, "step": 13820 }, { "epoch": 1.2, "learning_rate": 7.210954666358274e-05, "loss": 2.1787, "step": 13830 }, { "epoch": 1.2, "learning_rate": 7.21073782814849e-05, "loss": 2.2047, "step": 13840 }, { "epoch": 1.2, "learning_rate": 7.210520989938707e-05, "loss": 2.3396, "step": 13850 }, { "epoch": 1.2, "learning_rate": 7.210304151728922e-05, "loss": 2.4593, "step": 13860 }, { "epoch": 1.2, "learning_rate": 7.210087313519139e-05, "loss": 2.335, "step": 13870 }, { "epoch": 1.2, "learning_rate": 7.209870475309356e-05, "loss": 2.2371, "step": 13880 }, { "epoch": 1.2, "learning_rate": 7.209653637099571e-05, "loss": 2.3854, "step": 13890 }, { "epoch": 1.2, "learning_rate": 7.209436798889788e-05, "loss": 2.2064, "step": 13900 }, { "epoch": 1.2, "learning_rate": 7.209219960680004e-05, "loss": 2.0844, "step": 13910 }, { "epoch": 1.21, "learning_rate": 7.209003122470221e-05, "loss": 2.1732, "step": 13920 }, { "epoch": 1.21, "learning_rate": 7.208786284260438e-05, "loss": 2.1699, "step": 13930 }, { "epoch": 1.21, "learning_rate": 7.208569446050653e-05, "loss": 2.4865, "step": 13940 }, { "epoch": 1.21, "learning_rate": 7.208352607840868e-05, "loss": 2.2305, "step": 13950 }, { "epoch": 1.21, "learning_rate": 7.208135769631085e-05, "loss": 2.3679, "step": 13960 }, { "epoch": 1.21, "learning_rate": 7.207918931421301e-05, "loss": 2.2638, "step": 13970 }, { "epoch": 1.21, "learning_rate": 7.207702093211518e-05, "loss": 2.2441, "step": 13980 }, { "epoch": 1.21, "learning_rate": 7.207485255001735e-05, "loss": 2.1784, "step": 13990 }, { "epoch": 1.21, "learning_rate": 7.20726841679195e-05, "loss": 2.3226, "step": 14000 }, { "epoch": 1.21, "learning_rate": 7.207051578582167e-05, "loss": 2.0993, "step": 14010 }, { "epoch": 1.21, "learning_rate": 7.206834740372382e-05, "loss": 2.3361, "step": 14020 }, { "epoch": 1.22, "learning_rate": 7.206617902162598e-05, "loss": 2.2502, "step": 14030 }, { "epoch": 1.22, "learning_rate": 7.206401063952815e-05, "loss": 2.0878, "step": 14040 }, { "epoch": 1.22, "learning_rate": 7.206184225743032e-05, "loss": 2.2837, "step": 14050 }, { "epoch": 1.22, "learning_rate": 7.205967387533248e-05, "loss": 2.28, "step": 14060 }, { "epoch": 1.22, "learning_rate": 7.205750549323464e-05, "loss": 2.1929, "step": 14070 }, { "epoch": 1.22, "learning_rate": 7.20553371111368e-05, "loss": 2.3141, "step": 14080 }, { "epoch": 1.22, "learning_rate": 7.205316872903897e-05, "loss": 2.2442, "step": 14090 }, { "epoch": 1.22, "learning_rate": 7.205100034694112e-05, "loss": 2.1122, "step": 14100 }, { "epoch": 1.22, "learning_rate": 7.204883196484329e-05, "loss": 2.4443, "step": 14110 }, { "epoch": 1.22, "learning_rate": 7.204666358274546e-05, "loss": 2.0623, "step": 14120 }, { "epoch": 1.22, "learning_rate": 7.204449520064762e-05, "loss": 2.1528, "step": 14130 }, { "epoch": 1.22, "learning_rate": 7.204232681854977e-05, "loss": 2.2306, "step": 14140 }, { "epoch": 1.23, "learning_rate": 7.204015843645194e-05, "loss": 2.1971, "step": 14150 }, { "epoch": 1.23, "learning_rate": 7.203799005435411e-05, "loss": 1.9787, "step": 14160 }, { "epoch": 1.23, "learning_rate": 7.203582167225626e-05, "loss": 2.404, "step": 14170 }, { "epoch": 1.23, "learning_rate": 7.203365329015843e-05, "loss": 2.1898, "step": 14180 }, { "epoch": 1.23, "learning_rate": 7.203148490806059e-05, "loss": 2.1092, "step": 14190 }, { "epoch": 1.23, "learning_rate": 7.202931652596276e-05, "loss": 2.1696, "step": 14200 }, { "epoch": 1.23, "learning_rate": 7.202714814386491e-05, "loss": 2.3963, "step": 14210 }, { "epoch": 1.23, "learning_rate": 7.202497976176708e-05, "loss": 2.2315, "step": 14220 }, { "epoch": 1.23, "learning_rate": 7.202281137966924e-05, "loss": 2.0963, "step": 14230 }, { "epoch": 1.23, "learning_rate": 7.202064299757141e-05, "loss": 2.351, "step": 14240 }, { "epoch": 1.23, "learning_rate": 7.201847461547356e-05, "loss": 2.4555, "step": 14250 }, { "epoch": 1.24, "learning_rate": 7.201630623337573e-05, "loss": 2.1769, "step": 14260 }, { "epoch": 1.24, "learning_rate": 7.20141378512779e-05, "loss": 2.0947, "step": 14270 }, { "epoch": 1.24, "learning_rate": 7.201196946918005e-05, "loss": 2.3279, "step": 14280 }, { "epoch": 1.24, "learning_rate": 7.200980108708222e-05, "loss": 2.2115, "step": 14290 }, { "epoch": 1.24, "learning_rate": 7.200763270498438e-05, "loss": 2.3222, "step": 14300 }, { "epoch": 1.24, "learning_rate": 7.200546432288655e-05, "loss": 2.4364, "step": 14310 }, { "epoch": 1.24, "learning_rate": 7.200329594078872e-05, "loss": 2.1355, "step": 14320 }, { "epoch": 1.24, "learning_rate": 7.200112755869087e-05, "loss": 2.0993, "step": 14330 }, { "epoch": 1.24, "learning_rate": 7.199895917659303e-05, "loss": 2.1106, "step": 14340 }, { "epoch": 1.24, "learning_rate": 7.199679079449519e-05, "loss": 2.1805, "step": 14350 }, { "epoch": 1.24, "learning_rate": 7.199462241239735e-05, "loss": 2.488, "step": 14360 }, { "epoch": 1.24, "learning_rate": 7.199245403029952e-05, "loss": 2.2367, "step": 14370 }, { "epoch": 1.25, "learning_rate": 7.199028564820169e-05, "loss": 2.2467, "step": 14380 }, { "epoch": 1.25, "learning_rate": 7.198811726610385e-05, "loss": 2.2752, "step": 14390 }, { "epoch": 1.25, "learning_rate": 7.1985948884006e-05, "loss": 2.1997, "step": 14400 }, { "epoch": 1.25, "learning_rate": 7.198378050190817e-05, "loss": 2.1153, "step": 14410 }, { "epoch": 1.25, "learning_rate": 7.198161211981033e-05, "loss": 2.1675, "step": 14420 }, { "epoch": 1.25, "learning_rate": 7.197944373771249e-05, "loss": 2.1646, "step": 14430 }, { "epoch": 1.25, "learning_rate": 7.197727535561466e-05, "loss": 2.2753, "step": 14440 }, { "epoch": 1.25, "learning_rate": 7.197510697351682e-05, "loss": 2.5414, "step": 14450 }, { "epoch": 1.25, "learning_rate": 7.197293859141899e-05, "loss": 2.1798, "step": 14460 }, { "epoch": 1.25, "learning_rate": 7.197077020932114e-05, "loss": 2.2364, "step": 14470 }, { "epoch": 1.25, "learning_rate": 7.196860182722331e-05, "loss": 2.2549, "step": 14480 }, { "epoch": 1.25, "learning_rate": 7.196643344512546e-05, "loss": 2.4055, "step": 14490 }, { "epoch": 1.26, "learning_rate": 7.196426506302763e-05, "loss": 2.0944, "step": 14500 }, { "epoch": 1.26, "learning_rate": 7.19620966809298e-05, "loss": 2.2043, "step": 14510 }, { "epoch": 1.26, "learning_rate": 7.195992829883196e-05, "loss": 2.2621, "step": 14520 }, { "epoch": 1.26, "learning_rate": 7.195775991673413e-05, "loss": 2.4316, "step": 14530 }, { "epoch": 1.26, "learning_rate": 7.195559153463628e-05, "loss": 2.3355, "step": 14540 }, { "epoch": 1.26, "learning_rate": 7.195342315253845e-05, "loss": 2.2105, "step": 14550 }, { "epoch": 1.26, "learning_rate": 7.19512547704406e-05, "loss": 2.2675, "step": 14560 }, { "epoch": 1.26, "learning_rate": 7.194908638834277e-05, "loss": 2.1621, "step": 14570 }, { "epoch": 1.26, "learning_rate": 7.194691800624493e-05, "loss": 2.2814, "step": 14580 }, { "epoch": 1.26, "learning_rate": 7.19447496241471e-05, "loss": 2.0982, "step": 14590 }, { "epoch": 1.26, "learning_rate": 7.194258124204927e-05, "loss": 2.3916, "step": 14600 }, { "epoch": 1.27, "learning_rate": 7.194041285995142e-05, "loss": 2.2864, "step": 14610 }, { "epoch": 1.27, "learning_rate": 7.193824447785359e-05, "loss": 2.4214, "step": 14620 }, { "epoch": 1.27, "learning_rate": 7.193607609575574e-05, "loss": 2.2512, "step": 14630 }, { "epoch": 1.27, "learning_rate": 7.19339077136579e-05, "loss": 2.4205, "step": 14640 }, { "epoch": 1.27, "learning_rate": 7.193173933156007e-05, "loss": 2.1687, "step": 14650 }, { "epoch": 1.27, "learning_rate": 7.192957094946224e-05, "loss": 2.1733, "step": 14660 }, { "epoch": 1.27, "learning_rate": 7.19274025673644e-05, "loss": 2.1928, "step": 14670 }, { "epoch": 1.27, "learning_rate": 7.192523418526656e-05, "loss": 2.1713, "step": 14680 }, { "epoch": 1.27, "learning_rate": 7.192306580316872e-05, "loss": 2.3359, "step": 14690 }, { "epoch": 1.27, "learning_rate": 7.192089742107089e-05, "loss": 2.1576, "step": 14700 }, { "epoch": 1.27, "learning_rate": 7.191872903897304e-05, "loss": 2.4226, "step": 14710 }, { "epoch": 1.27, "learning_rate": 7.191656065687521e-05, "loss": 2.528, "step": 14720 }, { "epoch": 1.28, "learning_rate": 7.191439227477737e-05, "loss": 2.2304, "step": 14730 }, { "epoch": 1.28, "learning_rate": 7.191222389267954e-05, "loss": 2.1967, "step": 14740 }, { "epoch": 1.28, "learning_rate": 7.19100555105817e-05, "loss": 2.453, "step": 14750 }, { "epoch": 1.28, "learning_rate": 7.190788712848386e-05, "loss": 2.1559, "step": 14760 }, { "epoch": 1.28, "learning_rate": 7.190571874638603e-05, "loss": 2.3451, "step": 14770 }, { "epoch": 1.28, "learning_rate": 7.19035503642882e-05, "loss": 2.3705, "step": 14780 }, { "epoch": 1.28, "learning_rate": 7.190138198219035e-05, "loss": 2.2271, "step": 14790 }, { "epoch": 1.28, "learning_rate": 7.189921360009251e-05, "loss": 2.1683, "step": 14800 }, { "epoch": 1.28, "learning_rate": 7.189704521799468e-05, "loss": 2.2627, "step": 14810 }, { "epoch": 1.28, "learning_rate": 7.189487683589683e-05, "loss": 2.1259, "step": 14820 }, { "epoch": 1.28, "learning_rate": 7.1892708453799e-05, "loss": 2.2559, "step": 14830 }, { "epoch": 1.29, "learning_rate": 7.189054007170116e-05, "loss": 2.3999, "step": 14840 }, { "epoch": 1.29, "learning_rate": 7.188837168960333e-05, "loss": 2.3708, "step": 14850 }, { "epoch": 1.29, "learning_rate": 7.18862033075055e-05, "loss": 2.1288, "step": 14860 }, { "epoch": 1.29, "learning_rate": 7.188403492540765e-05, "loss": 2.2163, "step": 14870 }, { "epoch": 1.29, "learning_rate": 7.188186654330982e-05, "loss": 2.1236, "step": 14880 }, { "epoch": 1.29, "learning_rate": 7.187969816121197e-05, "loss": 2.3365, "step": 14890 }, { "epoch": 1.29, "learning_rate": 7.187752977911414e-05, "loss": 2.6058, "step": 14900 }, { "epoch": 1.29, "learning_rate": 7.18753613970163e-05, "loss": 2.195, "step": 14910 }, { "epoch": 1.29, "learning_rate": 7.187319301491847e-05, "loss": 2.285, "step": 14920 }, { "epoch": 1.29, "learning_rate": 7.187102463282063e-05, "loss": 2.1064, "step": 14930 }, { "epoch": 1.29, "learning_rate": 7.186885625072279e-05, "loss": 2.4765, "step": 14940 }, { "epoch": 1.29, "learning_rate": 7.186668786862495e-05, "loss": 2.404, "step": 14950 }, { "epoch": 1.3, "learning_rate": 7.186451948652711e-05, "loss": 2.4409, "step": 14960 }, { "epoch": 1.3, "learning_rate": 7.186235110442927e-05, "loss": 2.3208, "step": 14970 }, { "epoch": 1.3, "learning_rate": 7.186018272233144e-05, "loss": 2.2432, "step": 14980 }, { "epoch": 1.3, "learning_rate": 7.18580143402336e-05, "loss": 2.2191, "step": 14990 }, { "epoch": 1.3, "learning_rate": 7.185584595813577e-05, "loss": 2.3126, "step": 15000 }, { "epoch": 1.3, "learning_rate": 7.185367757603793e-05, "loss": 2.1208, "step": 15010 }, { "epoch": 1.3, "learning_rate": 7.185150919394008e-05, "loss": 2.3252, "step": 15020 }, { "epoch": 1.3, "learning_rate": 7.184934081184224e-05, "loss": 2.0438, "step": 15030 }, { "epoch": 1.3, "learning_rate": 7.184717242974441e-05, "loss": 2.3394, "step": 15040 }, { "epoch": 1.3, "learning_rate": 7.184500404764658e-05, "loss": 2.143, "step": 15050 }, { "epoch": 1.3, "learning_rate": 7.184283566554874e-05, "loss": 2.3355, "step": 15060 }, { "epoch": 1.31, "learning_rate": 7.18406672834509e-05, "loss": 2.4058, "step": 15070 }, { "epoch": 1.31, "learning_rate": 7.183849890135306e-05, "loss": 2.1831, "step": 15080 }, { "epoch": 1.31, "learning_rate": 7.183633051925523e-05, "loss": 2.2499, "step": 15090 }, { "epoch": 1.31, "learning_rate": 7.183416213715738e-05, "loss": 2.0964, "step": 15100 }, { "epoch": 1.31, "learning_rate": 7.183199375505955e-05, "loss": 2.3229, "step": 15110 }, { "epoch": 1.31, "learning_rate": 7.182982537296172e-05, "loss": 2.1256, "step": 15120 }, { "epoch": 1.31, "learning_rate": 7.182765699086388e-05, "loss": 2.2247, "step": 15130 }, { "epoch": 1.31, "learning_rate": 7.182548860876603e-05, "loss": 2.333, "step": 15140 }, { "epoch": 1.31, "learning_rate": 7.18233202266682e-05, "loss": 2.1324, "step": 15150 }, { "epoch": 1.31, "learning_rate": 7.182115184457037e-05, "loss": 2.222, "step": 15160 }, { "epoch": 1.31, "learning_rate": 7.181898346247252e-05, "loss": 2.3791, "step": 15170 }, { "epoch": 1.31, "learning_rate": 7.181681508037469e-05, "loss": 2.2861, "step": 15180 }, { "epoch": 1.32, "learning_rate": 7.181464669827685e-05, "loss": 2.2939, "step": 15190 }, { "epoch": 1.32, "learning_rate": 7.181247831617902e-05, "loss": 1.894, "step": 15200 }, { "epoch": 1.32, "learning_rate": 7.181030993408117e-05, "loss": 2.48, "step": 15210 }, { "epoch": 1.32, "learning_rate": 7.180814155198334e-05, "loss": 2.3461, "step": 15220 }, { "epoch": 1.32, "learning_rate": 7.18059731698855e-05, "loss": 2.1828, "step": 15230 }, { "epoch": 1.32, "learning_rate": 7.180380478778767e-05, "loss": 2.3364, "step": 15240 }, { "epoch": 1.32, "learning_rate": 7.180163640568982e-05, "loss": 2.4399, "step": 15250 }, { "epoch": 1.32, "learning_rate": 7.179946802359199e-05, "loss": 2.4059, "step": 15260 }, { "epoch": 1.32, "learning_rate": 7.179729964149416e-05, "loss": 2.587, "step": 15270 }, { "epoch": 1.32, "learning_rate": 7.179513125939631e-05, "loss": 2.039, "step": 15280 }, { "epoch": 1.32, "learning_rate": 7.179296287729848e-05, "loss": 2.0791, "step": 15290 }, { "epoch": 1.33, "learning_rate": 7.179079449520064e-05, "loss": 2.2058, "step": 15300 }, { "epoch": 1.33, "learning_rate": 7.178862611310281e-05, "loss": 2.2128, "step": 15310 }, { "epoch": 1.33, "learning_rate": 7.178645773100498e-05, "loss": 2.1639, "step": 15320 }, { "epoch": 1.33, "learning_rate": 7.178428934890713e-05, "loss": 2.1761, "step": 15330 }, { "epoch": 1.33, "learning_rate": 7.17821209668093e-05, "loss": 2.0679, "step": 15340 }, { "epoch": 1.33, "learning_rate": 7.177995258471145e-05, "loss": 2.2713, "step": 15350 }, { "epoch": 1.33, "learning_rate": 7.177778420261361e-05, "loss": 2.2869, "step": 15360 }, { "epoch": 1.33, "learning_rate": 7.177561582051578e-05, "loss": 2.0441, "step": 15370 }, { "epoch": 1.33, "learning_rate": 7.177344743841795e-05, "loss": 2.3183, "step": 15380 }, { "epoch": 1.33, "learning_rate": 7.177127905632011e-05, "loss": 2.1242, "step": 15390 }, { "epoch": 1.33, "learning_rate": 7.176911067422227e-05, "loss": 2.2091, "step": 15400 }, { "epoch": 1.33, "learning_rate": 7.176694229212443e-05, "loss": 2.3232, "step": 15410 }, { "epoch": 1.34, "learning_rate": 7.176477391002658e-05, "loss": 2.168, "step": 15420 }, { "epoch": 1.34, "learning_rate": 7.176260552792875e-05, "loss": 2.1841, "step": 15430 }, { "epoch": 1.34, "learning_rate": 7.176043714583092e-05, "loss": 2.1006, "step": 15440 }, { "epoch": 1.34, "learning_rate": 7.175826876373308e-05, "loss": 2.3084, "step": 15450 }, { "epoch": 1.34, "learning_rate": 7.175610038163525e-05, "loss": 2.2523, "step": 15460 }, { "epoch": 1.34, "learning_rate": 7.17539319995374e-05, "loss": 2.1739, "step": 15470 }, { "epoch": 1.34, "learning_rate": 7.175176361743957e-05, "loss": 2.1816, "step": 15480 }, { "epoch": 1.34, "learning_rate": 7.174959523534172e-05, "loss": 2.1984, "step": 15490 }, { "epoch": 1.34, "learning_rate": 7.174742685324389e-05, "loss": 2.1249, "step": 15500 }, { "epoch": 1.34, "learning_rate": 7.174525847114606e-05, "loss": 2.1791, "step": 15510 }, { "epoch": 1.34, "learning_rate": 7.174309008904822e-05, "loss": 2.0147, "step": 15520 }, { "epoch": 1.34, "learning_rate": 7.174092170695039e-05, "loss": 2.1276, "step": 15530 }, { "epoch": 1.35, "learning_rate": 7.173875332485254e-05, "loss": 2.1052, "step": 15540 }, { "epoch": 1.35, "learning_rate": 7.173658494275471e-05, "loss": 2.337, "step": 15550 }, { "epoch": 1.35, "learning_rate": 7.173441656065686e-05, "loss": 2.1563, "step": 15560 }, { "epoch": 1.35, "learning_rate": 7.173224817855903e-05, "loss": 2.1994, "step": 15570 }, { "epoch": 1.35, "learning_rate": 7.173007979646119e-05, "loss": 2.2578, "step": 15580 }, { "epoch": 1.35, "learning_rate": 7.172791141436336e-05, "loss": 2.2685, "step": 15590 }, { "epoch": 1.35, "learning_rate": 7.172574303226553e-05, "loss": 2.1607, "step": 15600 }, { "epoch": 1.35, "learning_rate": 7.172357465016768e-05, "loss": 1.977, "step": 15610 }, { "epoch": 1.35, "learning_rate": 7.172140626806984e-05, "loss": 2.3276, "step": 15620 }, { "epoch": 1.35, "learning_rate": 7.171923788597201e-05, "loss": 2.0913, "step": 15630 }, { "epoch": 1.35, "learning_rate": 7.171706950387416e-05, "loss": 2.3601, "step": 15640 }, { "epoch": 1.36, "learning_rate": 7.171490112177633e-05, "loss": 2.2572, "step": 15650 }, { "epoch": 1.36, "learning_rate": 7.17127327396785e-05, "loss": 2.4673, "step": 15660 }, { "epoch": 1.36, "learning_rate": 7.171056435758066e-05, "loss": 2.2583, "step": 15670 }, { "epoch": 1.36, "learning_rate": 7.170839597548282e-05, "loss": 2.3169, "step": 15680 }, { "epoch": 1.36, "learning_rate": 7.170622759338498e-05, "loss": 2.3613, "step": 15690 }, { "epoch": 1.36, "learning_rate": 7.170405921128715e-05, "loss": 2.2067, "step": 15700 }, { "epoch": 1.36, "learning_rate": 7.17018908291893e-05, "loss": 2.1297, "step": 15710 }, { "epoch": 1.36, "learning_rate": 7.169972244709147e-05, "loss": 2.1612, "step": 15720 }, { "epoch": 1.36, "learning_rate": 7.169755406499363e-05, "loss": 1.9952, "step": 15730 }, { "epoch": 1.36, "learning_rate": 7.16953856828958e-05, "loss": 2.1967, "step": 15740 }, { "epoch": 1.36, "learning_rate": 7.169321730079795e-05, "loss": 2.407, "step": 15750 }, { "epoch": 1.36, "learning_rate": 7.169104891870012e-05, "loss": 2.3869, "step": 15760 }, { "epoch": 1.37, "learning_rate": 7.168888053660229e-05, "loss": 2.094, "step": 15770 }, { "epoch": 1.37, "learning_rate": 7.168671215450445e-05, "loss": 2.1231, "step": 15780 }, { "epoch": 1.37, "learning_rate": 7.16845437724066e-05, "loss": 2.1456, "step": 15790 }, { "epoch": 1.37, "learning_rate": 7.168237539030877e-05, "loss": 2.254, "step": 15800 }, { "epoch": 1.37, "learning_rate": 7.168020700821094e-05, "loss": 2.1805, "step": 15810 }, { "epoch": 1.37, "learning_rate": 7.167803862611309e-05, "loss": 2.3633, "step": 15820 }, { "epoch": 1.37, "learning_rate": 7.167587024401526e-05, "loss": 2.2114, "step": 15830 }, { "epoch": 1.37, "learning_rate": 7.167370186191742e-05, "loss": 2.4401, "step": 15840 }, { "epoch": 1.37, "learning_rate": 7.167153347981959e-05, "loss": 2.0849, "step": 15850 }, { "epoch": 1.37, "learning_rate": 7.166936509772176e-05, "loss": 2.2227, "step": 15860 }, { "epoch": 1.37, "learning_rate": 7.166719671562391e-05, "loss": 2.3704, "step": 15870 }, { "epoch": 1.38, "learning_rate": 7.166502833352608e-05, "loss": 2.144, "step": 15880 }, { "epoch": 1.38, "learning_rate": 7.166285995142823e-05, "loss": 2.2913, "step": 15890 }, { "epoch": 1.38, "learning_rate": 7.16606915693304e-05, "loss": 2.2041, "step": 15900 }, { "epoch": 1.38, "learning_rate": 7.165852318723256e-05, "loss": 2.1012, "step": 15910 }, { "epoch": 1.38, "learning_rate": 7.165635480513473e-05, "loss": 2.179, "step": 15920 }, { "epoch": 1.38, "learning_rate": 7.16541864230369e-05, "loss": 1.9432, "step": 15930 }, { "epoch": 1.38, "learning_rate": 7.165201804093905e-05, "loss": 2.1521, "step": 15940 }, { "epoch": 1.38, "learning_rate": 7.164984965884121e-05, "loss": 2.0326, "step": 15950 }, { "epoch": 1.38, "learning_rate": 7.164768127674337e-05, "loss": 2.4794, "step": 15960 }, { "epoch": 1.38, "learning_rate": 7.164551289464553e-05, "loss": 2.21, "step": 15970 }, { "epoch": 1.38, "learning_rate": 7.16433445125477e-05, "loss": 2.1925, "step": 15980 }, { "epoch": 1.38, "learning_rate": 7.164117613044987e-05, "loss": 2.3233, "step": 15990 }, { "epoch": 1.39, "learning_rate": 7.163900774835203e-05, "loss": 2.0754, "step": 16000 }, { "epoch": 1.39, "learning_rate": 7.163683936625419e-05, "loss": 2.1648, "step": 16010 }, { "epoch": 1.39, "learning_rate": 7.163467098415635e-05, "loss": 2.1552, "step": 16020 }, { "epoch": 1.39, "learning_rate": 7.16325026020585e-05, "loss": 2.1181, "step": 16030 }, { "epoch": 1.39, "learning_rate": 7.163033421996067e-05, "loss": 2.2419, "step": 16040 }, { "epoch": 1.39, "learning_rate": 7.162816583786284e-05, "loss": 2.0521, "step": 16050 }, { "epoch": 1.39, "learning_rate": 7.1625997455765e-05, "loss": 2.1503, "step": 16060 }, { "epoch": 1.39, "learning_rate": 7.162382907366717e-05, "loss": 2.4531, "step": 16070 }, { "epoch": 1.39, "learning_rate": 7.162166069156932e-05, "loss": 2.0063, "step": 16080 }, { "epoch": 1.39, "learning_rate": 7.161949230947149e-05, "loss": 2.2506, "step": 16090 }, { "epoch": 1.39, "learning_rate": 7.161732392737364e-05, "loss": 2.3587, "step": 16100 }, { "epoch": 1.4, "learning_rate": 7.161515554527581e-05, "loss": 2.3882, "step": 16110 }, { "epoch": 1.4, "learning_rate": 7.161298716317797e-05, "loss": 2.2353, "step": 16120 }, { "epoch": 1.4, "learning_rate": 7.161081878108014e-05, "loss": 2.0815, "step": 16130 }, { "epoch": 1.4, "learning_rate": 7.16086503989823e-05, "loss": 2.2406, "step": 16140 }, { "epoch": 1.4, "learning_rate": 7.160648201688446e-05, "loss": 2.2284, "step": 16150 }, { "epoch": 1.4, "learning_rate": 7.160431363478663e-05, "loss": 2.1473, "step": 16160 }, { "epoch": 1.4, "learning_rate": 7.160214525268878e-05, "loss": 2.2159, "step": 16170 }, { "epoch": 1.4, "learning_rate": 7.159997687059095e-05, "loss": 2.1341, "step": 16180 }, { "epoch": 1.4, "learning_rate": 7.159780848849311e-05, "loss": 2.1386, "step": 16190 }, { "epoch": 1.4, "learning_rate": 7.159564010639528e-05, "loss": 2.0006, "step": 16200 }, { "epoch": 1.4, "learning_rate": 7.159347172429743e-05, "loss": 2.3805, "step": 16210 }, { "epoch": 1.4, "learning_rate": 7.15913033421996e-05, "loss": 2.24, "step": 16220 }, { "epoch": 1.41, "learning_rate": 7.158913496010176e-05, "loss": 2.1246, "step": 16230 }, { "epoch": 1.41, "learning_rate": 7.158696657800393e-05, "loss": 2.1951, "step": 16240 }, { "epoch": 1.41, "learning_rate": 7.158479819590608e-05, "loss": 2.2372, "step": 16250 }, { "epoch": 1.41, "learning_rate": 7.158262981380825e-05, "loss": 2.1762, "step": 16260 }, { "epoch": 1.41, "learning_rate": 7.158046143171042e-05, "loss": 2.4856, "step": 16270 }, { "epoch": 1.41, "learning_rate": 7.157829304961257e-05, "loss": 2.1876, "step": 16280 }, { "epoch": 1.41, "learning_rate": 7.157612466751474e-05, "loss": 2.1707, "step": 16290 }, { "epoch": 1.41, "learning_rate": 7.15739562854169e-05, "loss": 2.1019, "step": 16300 }, { "epoch": 1.41, "learning_rate": 7.157178790331907e-05, "loss": 2.096, "step": 16310 }, { "epoch": 1.41, "learning_rate": 7.156961952122123e-05, "loss": 2.193, "step": 16320 }, { "epoch": 1.41, "learning_rate": 7.156745113912339e-05, "loss": 2.1841, "step": 16330 }, { "epoch": 1.42, "learning_rate": 7.156528275702555e-05, "loss": 2.1115, "step": 16340 }, { "epoch": 1.42, "learning_rate": 7.156311437492771e-05, "loss": 2.2874, "step": 16350 }, { "epoch": 1.42, "learning_rate": 7.156094599282987e-05, "loss": 2.0789, "step": 16360 }, { "epoch": 1.42, "learning_rate": 7.155877761073204e-05, "loss": 2.3592, "step": 16370 }, { "epoch": 1.42, "learning_rate": 7.15566092286342e-05, "loss": 2.1021, "step": 16380 }, { "epoch": 1.42, "learning_rate": 7.155444084653637e-05, "loss": 2.3517, "step": 16390 }, { "epoch": 1.42, "learning_rate": 7.155227246443853e-05, "loss": 2.3382, "step": 16400 }, { "epoch": 1.42, "learning_rate": 7.155010408234069e-05, "loss": 2.142, "step": 16410 }, { "epoch": 1.42, "learning_rate": 7.154793570024284e-05, "loss": 2.1281, "step": 16420 }, { "epoch": 1.42, "learning_rate": 7.154576731814501e-05, "loss": 2.1464, "step": 16430 }, { "epoch": 1.42, "learning_rate": 7.154359893604718e-05, "loss": 2.2636, "step": 16440 }, { "epoch": 1.42, "learning_rate": 7.154143055394934e-05, "loss": 2.1274, "step": 16450 }, { "epoch": 1.43, "learning_rate": 7.153926217185151e-05, "loss": 2.3768, "step": 16460 }, { "epoch": 1.43, "learning_rate": 7.153709378975366e-05, "loss": 2.1634, "step": 16470 }, { "epoch": 1.43, "learning_rate": 7.153492540765583e-05, "loss": 2.2905, "step": 16480 }, { "epoch": 1.43, "learning_rate": 7.153275702555798e-05, "loss": 2.3877, "step": 16490 }, { "epoch": 1.43, "learning_rate": 7.153058864346015e-05, "loss": 2.157, "step": 16500 }, { "epoch": 1.43, "learning_rate": 7.152842026136231e-05, "loss": 2.1703, "step": 16510 }, { "epoch": 1.43, "learning_rate": 7.152625187926448e-05, "loss": 2.2166, "step": 16520 }, { "epoch": 1.43, "learning_rate": 7.152408349716665e-05, "loss": 2.3467, "step": 16530 }, { "epoch": 1.43, "learning_rate": 7.15219151150688e-05, "loss": 2.2546, "step": 16540 }, { "epoch": 1.43, "learning_rate": 7.151974673297097e-05, "loss": 2.2189, "step": 16550 }, { "epoch": 1.43, "learning_rate": 7.151757835087312e-05, "loss": 2.2932, "step": 16560 }, { "epoch": 1.44, "learning_rate": 7.151540996877529e-05, "loss": 1.9787, "step": 16570 }, { "epoch": 1.44, "learning_rate": 7.151324158667745e-05, "loss": 1.9562, "step": 16580 }, { "epoch": 1.44, "learning_rate": 7.151107320457962e-05, "loss": 1.9967, "step": 16590 }, { "epoch": 1.44, "learning_rate": 7.150890482248179e-05, "loss": 1.8321, "step": 16600 }, { "epoch": 1.44, "learning_rate": 7.150673644038394e-05, "loss": 2.3641, "step": 16610 }, { "epoch": 1.44, "learning_rate": 7.15045680582861e-05, "loss": 2.28, "step": 16620 }, { "epoch": 1.44, "learning_rate": 7.150239967618827e-05, "loss": 2.0258, "step": 16630 }, { "epoch": 1.44, "learning_rate": 7.150023129409042e-05, "loss": 2.1756, "step": 16640 }, { "epoch": 1.44, "learning_rate": 7.149806291199259e-05, "loss": 2.0928, "step": 16650 }, { "epoch": 1.44, "learning_rate": 7.149589452989476e-05, "loss": 2.2269, "step": 16660 }, { "epoch": 1.44, "learning_rate": 7.149372614779692e-05, "loss": 2.1824, "step": 16670 }, { "epoch": 1.44, "learning_rate": 7.149155776569908e-05, "loss": 2.1766, "step": 16680 }, { "epoch": 1.45, "learning_rate": 7.148938938360124e-05, "loss": 2.3873, "step": 16690 }, { "epoch": 1.45, "learning_rate": 7.148722100150341e-05, "loss": 2.1999, "step": 16700 }, { "epoch": 1.45, "learning_rate": 7.148505261940556e-05, "loss": 2.0848, "step": 16710 }, { "epoch": 1.45, "learning_rate": 7.148288423730773e-05, "loss": 2.2039, "step": 16720 }, { "epoch": 1.45, "learning_rate": 7.14807158552099e-05, "loss": 2.4148, "step": 16730 }, { "epoch": 1.45, "learning_rate": 7.147854747311206e-05, "loss": 2.1179, "step": 16740 }, { "epoch": 1.45, "learning_rate": 7.147637909101421e-05, "loss": 2.2462, "step": 16750 }, { "epoch": 1.45, "learning_rate": 7.147421070891638e-05, "loss": 2.103, "step": 16760 }, { "epoch": 1.45, "learning_rate": 7.147204232681855e-05, "loss": 2.3753, "step": 16770 }, { "epoch": 1.45, "learning_rate": 7.146987394472071e-05, "loss": 2.4718, "step": 16780 }, { "epoch": 1.45, "learning_rate": 7.146770556262287e-05, "loss": 2.294, "step": 16790 }, { "epoch": 1.45, "learning_rate": 7.146553718052503e-05, "loss": 2.0978, "step": 16800 }, { "epoch": 1.46, "learning_rate": 7.14633687984272e-05, "loss": 2.0392, "step": 16810 }, { "epoch": 1.46, "learning_rate": 7.146120041632935e-05, "loss": 2.0016, "step": 16820 }, { "epoch": 1.46, "learning_rate": 7.145903203423152e-05, "loss": 2.2875, "step": 16830 }, { "epoch": 1.46, "learning_rate": 7.145686365213368e-05, "loss": 2.3402, "step": 16840 }, { "epoch": 1.46, "learning_rate": 7.145469527003585e-05, "loss": 2.0698, "step": 16850 }, { "epoch": 1.46, "learning_rate": 7.145252688793802e-05, "loss": 2.1335, "step": 16860 }, { "epoch": 1.46, "learning_rate": 7.145035850584017e-05, "loss": 2.3041, "step": 16870 }, { "epoch": 1.46, "learning_rate": 7.144819012374234e-05, "loss": 2.3844, "step": 16880 }, { "epoch": 1.46, "learning_rate": 7.144602174164449e-05, "loss": 2.2783, "step": 16890 }, { "epoch": 1.46, "learning_rate": 7.144385335954666e-05, "loss": 2.258, "step": 16900 }, { "epoch": 1.46, "learning_rate": 7.144168497744882e-05, "loss": 2.154, "step": 16910 }, { "epoch": 1.47, "learning_rate": 7.143951659535099e-05, "loss": 1.9353, "step": 16920 }, { "epoch": 1.47, "learning_rate": 7.143734821325315e-05, "loss": 2.2101, "step": 16930 }, { "epoch": 1.47, "learning_rate": 7.143517983115531e-05, "loss": 2.1627, "step": 16940 }, { "epoch": 1.47, "learning_rate": 7.143301144905747e-05, "loss": 2.1558, "step": 16950 }, { "epoch": 1.47, "learning_rate": 7.143084306695963e-05, "loss": 2.1922, "step": 16960 }, { "epoch": 1.47, "learning_rate": 7.142867468486179e-05, "loss": 2.1482, "step": 16970 }, { "epoch": 1.47, "learning_rate": 7.142650630276396e-05, "loss": 2.2801, "step": 16980 }, { "epoch": 1.47, "learning_rate": 7.142433792066613e-05, "loss": 2.3146, "step": 16990 }, { "epoch": 1.47, "learning_rate": 7.142216953856829e-05, "loss": 2.1635, "step": 17000 }, { "epoch": 1.47, "learning_rate": 7.142000115647044e-05, "loss": 2.1469, "step": 17010 }, { "epoch": 1.47, "learning_rate": 7.141783277437261e-05, "loss": 1.9936, "step": 17020 }, { "epoch": 1.47, "learning_rate": 7.141566439227476e-05, "loss": 2.0618, "step": 17030 }, { "epoch": 1.48, "learning_rate": 7.141349601017693e-05, "loss": 2.4121, "step": 17040 }, { "epoch": 1.48, "learning_rate": 7.14113276280791e-05, "loss": 2.2246, "step": 17050 }, { "epoch": 1.48, "learning_rate": 7.140915924598126e-05, "loss": 2.158, "step": 17060 }, { "epoch": 1.48, "learning_rate": 7.140699086388343e-05, "loss": 2.2489, "step": 17070 }, { "epoch": 1.48, "learning_rate": 7.140482248178558e-05, "loss": 2.3136, "step": 17080 }, { "epoch": 1.48, "learning_rate": 7.140265409968775e-05, "loss": 2.2531, "step": 17090 }, { "epoch": 1.48, "learning_rate": 7.14004857175899e-05, "loss": 2.1928, "step": 17100 }, { "epoch": 1.48, "learning_rate": 7.139831733549207e-05, "loss": 1.9999, "step": 17110 }, { "epoch": 1.48, "learning_rate": 7.139614895339423e-05, "loss": 2.1915, "step": 17120 }, { "epoch": 1.48, "learning_rate": 7.13939805712964e-05, "loss": 2.0228, "step": 17130 }, { "epoch": 1.48, "learning_rate": 7.139181218919857e-05, "loss": 2.0608, "step": 17140 }, { "epoch": 1.49, "learning_rate": 7.138964380710072e-05, "loss": 2.1659, "step": 17150 }, { "epoch": 1.49, "learning_rate": 7.138747542500289e-05, "loss": 2.0054, "step": 17160 }, { "epoch": 1.49, "learning_rate": 7.138530704290505e-05, "loss": 2.2305, "step": 17170 }, { "epoch": 1.49, "learning_rate": 7.13831386608072e-05, "loss": 2.2416, "step": 17180 }, { "epoch": 1.49, "learning_rate": 7.138097027870937e-05, "loss": 2.1895, "step": 17190 }, { "epoch": 1.49, "learning_rate": 7.137880189661154e-05, "loss": 1.8728, "step": 17200 }, { "epoch": 1.49, "learning_rate": 7.137663351451369e-05, "loss": 2.2637, "step": 17210 }, { "epoch": 1.49, "learning_rate": 7.137446513241586e-05, "loss": 2.2594, "step": 17220 }, { "epoch": 1.49, "learning_rate": 7.137229675031802e-05, "loss": 2.205, "step": 17230 }, { "epoch": 1.49, "learning_rate": 7.137012836822019e-05, "loss": 2.1662, "step": 17240 }, { "epoch": 1.49, "learning_rate": 7.136795998612234e-05, "loss": 2.2657, "step": 17250 }, { "epoch": 1.49, "learning_rate": 7.136579160402451e-05, "loss": 2.136, "step": 17260 }, { "epoch": 1.5, "learning_rate": 7.136362322192668e-05, "loss": 2.1879, "step": 17270 }, { "epoch": 1.5, "learning_rate": 7.136145483982883e-05, "loss": 2.0314, "step": 17280 }, { "epoch": 1.5, "learning_rate": 7.1359286457731e-05, "loss": 2.2263, "step": 17290 }, { "epoch": 1.5, "learning_rate": 7.135711807563316e-05, "loss": 2.2228, "step": 17300 }, { "epoch": 1.5, "learning_rate": 7.135494969353533e-05, "loss": 2.356, "step": 17310 }, { "epoch": 1.5, "learning_rate": 7.13527813114375e-05, "loss": 2.1769, "step": 17320 }, { "epoch": 1.5, "learning_rate": 7.135061292933965e-05, "loss": 2.1246, "step": 17330 }, { "epoch": 1.5, "learning_rate": 7.134844454724181e-05, "loss": 2.1473, "step": 17340 }, { "epoch": 1.5, "learning_rate": 7.134627616514397e-05, "loss": 2.1222, "step": 17350 }, { "epoch": 1.5, "learning_rate": 7.134410778304613e-05, "loss": 2.2211, "step": 17360 }, { "epoch": 1.5, "learning_rate": 7.13419394009483e-05, "loss": 2.1729, "step": 17370 }, { "epoch": 1.51, "learning_rate": 7.133977101885047e-05, "loss": 2.3325, "step": 17380 }, { "epoch": 1.51, "learning_rate": 7.133760263675263e-05, "loss": 2.5692, "step": 17390 }, { "epoch": 1.51, "learning_rate": 7.133543425465478e-05, "loss": 2.3583, "step": 17400 }, { "epoch": 1.51, "learning_rate": 7.133326587255695e-05, "loss": 1.9934, "step": 17410 }, { "epoch": 1.51, "learning_rate": 7.13310974904591e-05, "loss": 2.2185, "step": 17420 }, { "epoch": 1.51, "learning_rate": 7.132892910836127e-05, "loss": 2.1945, "step": 17430 }, { "epoch": 1.51, "learning_rate": 7.132676072626344e-05, "loss": 2.1462, "step": 17440 }, { "epoch": 1.51, "learning_rate": 7.13245923441656e-05, "loss": 1.9606, "step": 17450 }, { "epoch": 1.51, "learning_rate": 7.132242396206777e-05, "loss": 2.4057, "step": 17460 }, { "epoch": 1.51, "learning_rate": 7.132025557996992e-05, "loss": 2.192, "step": 17470 }, { "epoch": 1.51, "learning_rate": 7.131808719787209e-05, "loss": 2.005, "step": 17480 }, { "epoch": 1.51, "learning_rate": 7.131591881577424e-05, "loss": 2.1283, "step": 17490 }, { "epoch": 1.52, "learning_rate": 7.131375043367641e-05, "loss": 2.222, "step": 17500 }, { "epoch": 1.52, "learning_rate": 7.131158205157857e-05, "loss": 2.3496, "step": 17510 }, { "epoch": 1.52, "learning_rate": 7.130941366948074e-05, "loss": 2.171, "step": 17520 }, { "epoch": 1.52, "learning_rate": 7.130724528738291e-05, "loss": 2.1675, "step": 17530 }, { "epoch": 1.52, "learning_rate": 7.130507690528506e-05, "loss": 2.1895, "step": 17540 }, { "epoch": 1.52, "learning_rate": 7.130290852318723e-05, "loss": 2.1196, "step": 17550 }, { "epoch": 1.52, "learning_rate": 7.130074014108938e-05, "loss": 1.9644, "step": 17560 }, { "epoch": 1.52, "learning_rate": 7.129857175899155e-05, "loss": 2.3527, "step": 17570 }, { "epoch": 1.52, "learning_rate": 7.129640337689371e-05, "loss": 2.1146, "step": 17580 }, { "epoch": 1.52, "learning_rate": 7.129423499479588e-05, "loss": 2.2618, "step": 17590 }, { "epoch": 1.52, "learning_rate": 7.129206661269805e-05, "loss": 2.1755, "step": 17600 }, { "epoch": 1.53, "learning_rate": 7.12898982306002e-05, "loss": 2.2948, "step": 17610 }, { "epoch": 1.53, "learning_rate": 7.128772984850236e-05, "loss": 1.9428, "step": 17620 }, { "epoch": 1.53, "learning_rate": 7.128556146640453e-05, "loss": 2.1265, "step": 17630 }, { "epoch": 1.53, "learning_rate": 7.128339308430668e-05, "loss": 2.2365, "step": 17640 }, { "epoch": 1.53, "learning_rate": 7.128122470220885e-05, "loss": 2.2138, "step": 17650 }, { "epoch": 1.53, "learning_rate": 7.127905632011102e-05, "loss": 2.2123, "step": 17660 }, { "epoch": 1.53, "learning_rate": 7.127688793801318e-05, "loss": 2.2368, "step": 17670 }, { "epoch": 1.53, "learning_rate": 7.127471955591534e-05, "loss": 2.2851, "step": 17680 }, { "epoch": 1.53, "learning_rate": 7.12725511738175e-05, "loss": 2.2595, "step": 17690 }, { "epoch": 1.53, "learning_rate": 7.127038279171967e-05, "loss": 2.263, "step": 17700 }, { "epoch": 1.53, "learning_rate": 7.126821440962182e-05, "loss": 1.96, "step": 17710 }, { "epoch": 1.53, "learning_rate": 7.126604602752399e-05, "loss": 2.2508, "step": 17720 }, { "epoch": 1.54, "learning_rate": 7.126387764542615e-05, "loss": 2.3117, "step": 17730 }, { "epoch": 1.54, "learning_rate": 7.126170926332832e-05, "loss": 2.2588, "step": 17740 }, { "epoch": 1.54, "learning_rate": 7.125954088123047e-05, "loss": 2.1594, "step": 17750 }, { "epoch": 1.54, "learning_rate": 7.125737249913264e-05, "loss": 2.2269, "step": 17760 }, { "epoch": 1.54, "learning_rate": 7.12552041170348e-05, "loss": 2.2138, "step": 17770 }, { "epoch": 1.54, "learning_rate": 7.125303573493697e-05, "loss": 2.2701, "step": 17780 }, { "epoch": 1.54, "learning_rate": 7.125086735283913e-05, "loss": 2.1671, "step": 17790 }, { "epoch": 1.54, "learning_rate": 7.124869897074129e-05, "loss": 2.2811, "step": 17800 }, { "epoch": 1.54, "learning_rate": 7.124653058864346e-05, "loss": 1.8438, "step": 17810 }, { "epoch": 1.54, "learning_rate": 7.124436220654561e-05, "loss": 2.1844, "step": 17820 }, { "epoch": 1.54, "learning_rate": 7.124219382444778e-05, "loss": 2.2885, "step": 17830 }, { "epoch": 1.55, "learning_rate": 7.124002544234994e-05, "loss": 2.2865, "step": 17840 }, { "epoch": 1.55, "learning_rate": 7.123785706025211e-05, "loss": 2.2736, "step": 17850 }, { "epoch": 1.55, "learning_rate": 7.123568867815428e-05, "loss": 1.9066, "step": 17860 }, { "epoch": 1.55, "learning_rate": 7.123352029605643e-05, "loss": 2.0926, "step": 17870 }, { "epoch": 1.55, "learning_rate": 7.12313519139586e-05, "loss": 2.0808, "step": 17880 }, { "epoch": 1.55, "learning_rate": 7.122918353186075e-05, "loss": 2.1383, "step": 17890 }, { "epoch": 1.55, "learning_rate": 7.122701514976291e-05, "loss": 1.9551, "step": 17900 }, { "epoch": 1.55, "learning_rate": 7.122484676766508e-05, "loss": 2.4119, "step": 17910 }, { "epoch": 1.55, "learning_rate": 7.122267838556725e-05, "loss": 2.0379, "step": 17920 }, { "epoch": 1.55, "learning_rate": 7.122051000346941e-05, "loss": 2.2976, "step": 17930 }, { "epoch": 1.55, "learning_rate": 7.121834162137157e-05, "loss": 2.1934, "step": 17940 }, { "epoch": 1.55, "learning_rate": 7.121617323927373e-05, "loss": 2.06, "step": 17950 }, { "epoch": 1.56, "learning_rate": 7.121400485717589e-05, "loss": 2.2311, "step": 17960 }, { "epoch": 1.56, "learning_rate": 7.121183647507805e-05, "loss": 2.2086, "step": 17970 }, { "epoch": 1.56, "learning_rate": 7.120966809298022e-05, "loss": 2.1563, "step": 17980 }, { "epoch": 1.56, "learning_rate": 7.120749971088239e-05, "loss": 1.9793, "step": 17990 }, { "epoch": 1.56, "learning_rate": 7.120533132878455e-05, "loss": 2.3525, "step": 18000 }, { "epoch": 1.56, "learning_rate": 7.12031629466867e-05, "loss": 2.0586, "step": 18010 }, { "epoch": 1.56, "learning_rate": 7.120099456458887e-05, "loss": 2.2079, "step": 18020 }, { "epoch": 1.56, "learning_rate": 7.119882618249102e-05, "loss": 1.9623, "step": 18030 }, { "epoch": 1.56, "learning_rate": 7.119665780039319e-05, "loss": 2.0392, "step": 18040 }, { "epoch": 1.56, "learning_rate": 7.119448941829536e-05, "loss": 2.247, "step": 18050 }, { "epoch": 1.56, "learning_rate": 7.119232103619752e-05, "loss": 2.0513, "step": 18060 }, { "epoch": 1.56, "learning_rate": 7.119015265409969e-05, "loss": 2.3646, "step": 18070 }, { "epoch": 1.57, "learning_rate": 7.118798427200184e-05, "loss": 2.3209, "step": 18080 }, { "epoch": 1.57, "learning_rate": 7.118581588990401e-05, "loss": 2.4755, "step": 18090 }, { "epoch": 1.57, "learning_rate": 7.118364750780616e-05, "loss": 2.2141, "step": 18100 }, { "epoch": 1.57, "learning_rate": 7.118147912570833e-05, "loss": 2.4704, "step": 18110 }, { "epoch": 1.57, "learning_rate": 7.11793107436105e-05, "loss": 2.1207, "step": 18120 }, { "epoch": 1.57, "learning_rate": 7.117714236151266e-05, "loss": 2.2505, "step": 18130 }, { "epoch": 1.57, "learning_rate": 7.117497397941483e-05, "loss": 2.2303, "step": 18140 }, { "epoch": 1.57, "learning_rate": 7.117280559731698e-05, "loss": 2.1875, "step": 18150 }, { "epoch": 1.57, "learning_rate": 7.117063721521915e-05, "loss": 2.2574, "step": 18160 }, { "epoch": 1.57, "learning_rate": 7.116846883312131e-05, "loss": 2.2142, "step": 18170 }, { "epoch": 1.57, "learning_rate": 7.116630045102347e-05, "loss": 2.104, "step": 18180 }, { "epoch": 1.58, "learning_rate": 7.116413206892563e-05, "loss": 2.284, "step": 18190 }, { "epoch": 1.58, "learning_rate": 7.11619636868278e-05, "loss": 2.1617, "step": 18200 }, { "epoch": 1.58, "learning_rate": 7.115979530472996e-05, "loss": 2.2782, "step": 18210 }, { "epoch": 1.58, "learning_rate": 7.115762692263212e-05, "loss": 2.0096, "step": 18220 }, { "epoch": 1.58, "learning_rate": 7.115545854053428e-05, "loss": 1.9752, "step": 18230 }, { "epoch": 1.58, "learning_rate": 7.115329015843645e-05, "loss": 2.2052, "step": 18240 }, { "epoch": 1.58, "learning_rate": 7.11511217763386e-05, "loss": 2.3047, "step": 18250 }, { "epoch": 1.58, "learning_rate": 7.114895339424077e-05, "loss": 2.1944, "step": 18260 }, { "epoch": 1.58, "learning_rate": 7.114678501214294e-05, "loss": 2.1459, "step": 18270 }, { "epoch": 1.58, "learning_rate": 7.114461663004509e-05, "loss": 2.0507, "step": 18280 }, { "epoch": 1.58, "learning_rate": 7.114244824794726e-05, "loss": 1.9991, "step": 18290 }, { "epoch": 1.58, "learning_rate": 7.114027986584942e-05, "loss": 2.2128, "step": 18300 }, { "epoch": 1.59, "learning_rate": 7.113811148375159e-05, "loss": 2.0301, "step": 18310 }, { "epoch": 1.59, "learning_rate": 7.113594310165375e-05, "loss": 2.0913, "step": 18320 }, { "epoch": 1.59, "learning_rate": 7.113377471955591e-05, "loss": 2.0215, "step": 18330 }, { "epoch": 1.59, "learning_rate": 7.113160633745807e-05, "loss": 2.11, "step": 18340 }, { "epoch": 1.59, "learning_rate": 7.112943795536023e-05, "loss": 2.2769, "step": 18350 }, { "epoch": 1.59, "learning_rate": 7.112726957326239e-05, "loss": 2.0403, "step": 18360 }, { "epoch": 1.59, "learning_rate": 7.112510119116456e-05, "loss": 2.2522, "step": 18370 }, { "epoch": 1.59, "learning_rate": 7.112293280906673e-05, "loss": 2.2577, "step": 18380 }, { "epoch": 1.59, "learning_rate": 7.112076442696889e-05, "loss": 2.286, "step": 18390 }, { "epoch": 1.59, "learning_rate": 7.111859604487104e-05, "loss": 2.0726, "step": 18400 }, { "epoch": 1.59, "learning_rate": 7.111642766277321e-05, "loss": 2.1563, "step": 18410 }, { "epoch": 1.6, "learning_rate": 7.111425928067536e-05, "loss": 1.9793, "step": 18420 }, { "epoch": 1.6, "learning_rate": 7.111209089857753e-05, "loss": 2.0212, "step": 18430 }, { "epoch": 1.6, "learning_rate": 7.11099225164797e-05, "loss": 2.17, "step": 18440 }, { "epoch": 1.6, "learning_rate": 7.110775413438186e-05, "loss": 2.1369, "step": 18450 }, { "epoch": 1.6, "learning_rate": 7.110558575228403e-05, "loss": 2.1253, "step": 18460 }, { "epoch": 1.6, "learning_rate": 7.110341737018618e-05, "loss": 2.1201, "step": 18470 }, { "epoch": 1.6, "learning_rate": 7.110124898808835e-05, "loss": 2.1616, "step": 18480 }, { "epoch": 1.6, "learning_rate": 7.10990806059905e-05, "loss": 2.1686, "step": 18490 }, { "epoch": 1.6, "learning_rate": 7.109691222389267e-05, "loss": 2.2006, "step": 18500 }, { "epoch": 1.6, "learning_rate": 7.109474384179483e-05, "loss": 2.1778, "step": 18510 }, { "epoch": 1.6, "learning_rate": 7.1092575459697e-05, "loss": 1.8788, "step": 18520 }, { "epoch": 1.6, "learning_rate": 7.109040707759917e-05, "loss": 2.1071, "step": 18530 }, { "epoch": 1.61, "learning_rate": 7.108823869550132e-05, "loss": 2.2515, "step": 18540 }, { "epoch": 1.61, "learning_rate": 7.108607031340349e-05, "loss": 2.2417, "step": 18550 }, { "epoch": 1.61, "learning_rate": 7.108390193130564e-05, "loss": 2.1406, "step": 18560 }, { "epoch": 1.61, "learning_rate": 7.10817335492078e-05, "loss": 2.2901, "step": 18570 }, { "epoch": 1.61, "learning_rate": 7.107956516710997e-05, "loss": 2.162, "step": 18580 }, { "epoch": 1.61, "learning_rate": 7.107739678501214e-05, "loss": 2.1738, "step": 18590 }, { "epoch": 1.61, "learning_rate": 7.10752284029143e-05, "loss": 2.1745, "step": 18600 }, { "epoch": 1.61, "learning_rate": 7.107306002081646e-05, "loss": 2.2145, "step": 18610 }, { "epoch": 1.61, "learning_rate": 7.107089163871862e-05, "loss": 2.0582, "step": 18620 }, { "epoch": 1.61, "learning_rate": 7.106872325662079e-05, "loss": 2.4334, "step": 18630 }, { "epoch": 1.61, "learning_rate": 7.106655487452294e-05, "loss": 2.2173, "step": 18640 }, { "epoch": 1.62, "learning_rate": 7.106438649242511e-05, "loss": 2.1953, "step": 18650 }, { "epoch": 1.62, "learning_rate": 7.106221811032728e-05, "loss": 2.2666, "step": 18660 }, { "epoch": 1.62, "learning_rate": 7.106004972822944e-05, "loss": 2.1483, "step": 18670 }, { "epoch": 1.62, "learning_rate": 7.10578813461316e-05, "loss": 2.0673, "step": 18680 }, { "epoch": 1.62, "learning_rate": 7.105571296403376e-05, "loss": 2.2046, "step": 18690 }, { "epoch": 1.62, "learning_rate": 7.105354458193593e-05, "loss": 2.2391, "step": 18700 }, { "epoch": 1.62, "learning_rate": 7.10513761998381e-05, "loss": 2.1086, "step": 18710 }, { "epoch": 1.62, "learning_rate": 7.104920781774025e-05, "loss": 2.0454, "step": 18720 }, { "epoch": 1.62, "learning_rate": 7.104703943564241e-05, "loss": 2.097, "step": 18730 }, { "epoch": 1.62, "learning_rate": 7.104487105354458e-05, "loss": 2.4419, "step": 18740 }, { "epoch": 1.62, "learning_rate": 7.104270267144673e-05, "loss": 2.1994, "step": 18750 }, { "epoch": 1.62, "learning_rate": 7.10405342893489e-05, "loss": 2.0361, "step": 18760 }, { "epoch": 1.63, "learning_rate": 7.103836590725107e-05, "loss": 2.2174, "step": 18770 }, { "epoch": 1.63, "learning_rate": 7.103619752515323e-05, "loss": 2.2066, "step": 18780 }, { "epoch": 1.63, "learning_rate": 7.103402914305538e-05, "loss": 2.2749, "step": 18790 }, { "epoch": 1.63, "learning_rate": 7.103186076095755e-05, "loss": 2.2617, "step": 18800 }, { "epoch": 1.63, "learning_rate": 7.102969237885972e-05, "loss": 2.02, "step": 18810 }, { "epoch": 1.63, "learning_rate": 7.102752399676187e-05, "loss": 2.004, "step": 18820 }, { "epoch": 1.63, "learning_rate": 7.102535561466404e-05, "loss": 2.3, "step": 18830 }, { "epoch": 1.63, "learning_rate": 7.10231872325662e-05, "loss": 2.3415, "step": 18840 }, { "epoch": 1.63, "learning_rate": 7.102101885046837e-05, "loss": 2.2678, "step": 18850 }, { "epoch": 1.63, "learning_rate": 7.101885046837054e-05, "loss": 1.9427, "step": 18860 }, { "epoch": 1.63, "learning_rate": 7.101668208627269e-05, "loss": 2.1082, "step": 18870 }, { "epoch": 1.64, "learning_rate": 7.101451370417486e-05, "loss": 2.1485, "step": 18880 }, { "epoch": 1.64, "learning_rate": 7.101234532207701e-05, "loss": 1.9675, "step": 18890 }, { "epoch": 1.64, "learning_rate": 7.101017693997917e-05, "loss": 1.9938, "step": 18900 }, { "epoch": 1.64, "learning_rate": 7.100800855788134e-05, "loss": 2.2418, "step": 18910 }, { "epoch": 1.64, "learning_rate": 7.100584017578351e-05, "loss": 2.2082, "step": 18920 }, { "epoch": 1.64, "learning_rate": 7.100367179368567e-05, "loss": 2.0894, "step": 18930 }, { "epoch": 1.64, "learning_rate": 7.100150341158783e-05, "loss": 2.2154, "step": 18940 }, { "epoch": 1.64, "learning_rate": 7.099933502948999e-05, "loss": 2.2459, "step": 18950 }, { "epoch": 1.64, "learning_rate": 7.099716664739215e-05, "loss": 1.9461, "step": 18960 }, { "epoch": 1.64, "learning_rate": 7.099499826529431e-05, "loss": 2.3245, "step": 18970 }, { "epoch": 1.64, "learning_rate": 7.099282988319648e-05, "loss": 2.2608, "step": 18980 }, { "epoch": 1.64, "learning_rate": 7.099066150109864e-05, "loss": 2.0754, "step": 18990 }, { "epoch": 1.65, "learning_rate": 7.098849311900081e-05, "loss": 2.1383, "step": 19000 }, { "epoch": 1.65, "learning_rate": 7.098632473690296e-05, "loss": 2.1925, "step": 19010 }, { "epoch": 1.65, "learning_rate": 7.098415635480513e-05, "loss": 2.2065, "step": 19020 }, { "epoch": 1.65, "learning_rate": 7.098198797270728e-05, "loss": 2.1054, "step": 19030 }, { "epoch": 1.65, "learning_rate": 7.097981959060945e-05, "loss": 2.2268, "step": 19040 }, { "epoch": 1.65, "learning_rate": 7.097765120851162e-05, "loss": 2.0643, "step": 19050 }, { "epoch": 1.65, "learning_rate": 7.097548282641378e-05, "loss": 1.9876, "step": 19060 }, { "epoch": 1.65, "learning_rate": 7.097331444431595e-05, "loss": 2.0336, "step": 19070 }, { "epoch": 1.65, "learning_rate": 7.09711460622181e-05, "loss": 2.3742, "step": 19080 }, { "epoch": 1.65, "learning_rate": 7.096897768012027e-05, "loss": 2.2195, "step": 19090 }, { "epoch": 1.65, "learning_rate": 7.096680929802242e-05, "loss": 2.1703, "step": 19100 }, { "epoch": 1.66, "learning_rate": 7.096464091592459e-05, "loss": 2.4401, "step": 19110 }, { "epoch": 1.66, "learning_rate": 7.096247253382675e-05, "loss": 2.2766, "step": 19120 }, { "epoch": 1.66, "learning_rate": 7.096030415172892e-05, "loss": 2.2017, "step": 19130 }, { "epoch": 1.66, "learning_rate": 7.095813576963109e-05, "loss": 2.2193, "step": 19140 }, { "epoch": 1.66, "learning_rate": 7.095596738753324e-05, "loss": 2.2139, "step": 19150 }, { "epoch": 1.66, "learning_rate": 7.09537990054354e-05, "loss": 1.9546, "step": 19160 }, { "epoch": 1.66, "learning_rate": 7.095163062333757e-05, "loss": 2.1595, "step": 19170 }, { "epoch": 1.66, "learning_rate": 7.094946224123973e-05, "loss": 2.1159, "step": 19180 }, { "epoch": 1.66, "learning_rate": 7.094729385914189e-05, "loss": 2.1691, "step": 19190 }, { "epoch": 1.66, "learning_rate": 7.094512547704406e-05, "loss": 2.1393, "step": 19200 }, { "epoch": 1.66, "learning_rate": 7.094295709494622e-05, "loss": 2.0722, "step": 19210 }, { "epoch": 1.66, "learning_rate": 7.094078871284838e-05, "loss": 1.9979, "step": 19220 }, { "epoch": 1.67, "learning_rate": 7.093862033075054e-05, "loss": 1.9579, "step": 19230 }, { "epoch": 1.67, "learning_rate": 7.093645194865271e-05, "loss": 1.9834, "step": 19240 }, { "epoch": 1.67, "learning_rate": 7.093428356655486e-05, "loss": 2.132, "step": 19250 }, { "epoch": 1.67, "learning_rate": 7.093211518445703e-05, "loss": 2.1102, "step": 19260 }, { "epoch": 1.67, "learning_rate": 7.09299468023592e-05, "loss": 2.0468, "step": 19270 }, { "epoch": 1.67, "learning_rate": 7.092777842026136e-05, "loss": 2.1771, "step": 19280 }, { "epoch": 1.67, "learning_rate": 7.092561003816351e-05, "loss": 2.0634, "step": 19290 }, { "epoch": 1.67, "learning_rate": 7.092344165606568e-05, "loss": 2.0303, "step": 19300 }, { "epoch": 1.67, "learning_rate": 7.092127327396785e-05, "loss": 2.2442, "step": 19310 }, { "epoch": 1.67, "learning_rate": 7.091910489187001e-05, "loss": 2.0265, "step": 19320 }, { "epoch": 1.67, "learning_rate": 7.091693650977217e-05, "loss": 2.1049, "step": 19330 }, { "epoch": 1.67, "learning_rate": 7.091476812767433e-05, "loss": 2.0784, "step": 19340 }, { "epoch": 1.68, "learning_rate": 7.091259974557649e-05, "loss": 2.0669, "step": 19350 }, { "epoch": 1.68, "learning_rate": 7.091043136347865e-05, "loss": 2.0096, "step": 19360 }, { "epoch": 1.68, "learning_rate": 7.090826298138082e-05, "loss": 2.0921, "step": 19370 }, { "epoch": 1.68, "learning_rate": 7.090609459928299e-05, "loss": 2.139, "step": 19380 }, { "epoch": 1.68, "learning_rate": 7.090392621718515e-05, "loss": 2.2288, "step": 19390 }, { "epoch": 1.68, "learning_rate": 7.09017578350873e-05, "loss": 2.1624, "step": 19400 }, { "epoch": 1.68, "learning_rate": 7.089958945298947e-05, "loss": 2.0674, "step": 19410 }, { "epoch": 1.68, "learning_rate": 7.089742107089162e-05, "loss": 1.9923, "step": 19420 }, { "epoch": 1.68, "learning_rate": 7.089525268879379e-05, "loss": 2.2025, "step": 19430 }, { "epoch": 1.68, "learning_rate": 7.089308430669596e-05, "loss": 2.0863, "step": 19440 }, { "epoch": 1.68, "learning_rate": 7.089091592459812e-05, "loss": 2.3277, "step": 19450 }, { "epoch": 1.69, "learning_rate": 7.088874754250029e-05, "loss": 2.2243, "step": 19460 }, { "epoch": 1.69, "learning_rate": 7.088657916040244e-05, "loss": 2.1131, "step": 19470 }, { "epoch": 1.69, "learning_rate": 7.088441077830461e-05, "loss": 2.1741, "step": 19480 }, { "epoch": 1.69, "learning_rate": 7.088224239620676e-05, "loss": 2.0592, "step": 19490 }, { "epoch": 1.69, "learning_rate": 7.088007401410893e-05, "loss": 2.0001, "step": 19500 }, { "epoch": 1.69, "learning_rate": 7.08779056320111e-05, "loss": 2.162, "step": 19510 }, { "epoch": 1.69, "learning_rate": 7.087573724991326e-05, "loss": 2.2884, "step": 19520 }, { "epoch": 1.69, "learning_rate": 7.087356886781543e-05, "loss": 2.2022, "step": 19530 }, { "epoch": 1.69, "learning_rate": 7.087140048571758e-05, "loss": 2.0483, "step": 19540 }, { "epoch": 1.69, "learning_rate": 7.086923210361975e-05, "loss": 2.2984, "step": 19550 }, { "epoch": 1.69, "learning_rate": 7.08670637215219e-05, "loss": 2.0855, "step": 19560 }, { "epoch": 1.69, "learning_rate": 7.086489533942407e-05, "loss": 2.2839, "step": 19570 }, { "epoch": 1.7, "learning_rate": 7.086272695732623e-05, "loss": 2.2147, "step": 19580 }, { "epoch": 1.7, "learning_rate": 7.08605585752284e-05, "loss": 2.3285, "step": 19590 }, { "epoch": 1.7, "learning_rate": 7.085839019313056e-05, "loss": 2.1541, "step": 19600 }, { "epoch": 1.7, "learning_rate": 7.085622181103272e-05, "loss": 2.2798, "step": 19610 }, { "epoch": 1.7, "learning_rate": 7.085405342893488e-05, "loss": 2.2691, "step": 19620 }, { "epoch": 1.7, "learning_rate": 7.085188504683705e-05, "loss": 2.2237, "step": 19630 }, { "epoch": 1.7, "learning_rate": 7.08497166647392e-05, "loss": 2.0401, "step": 19640 }, { "epoch": 1.7, "learning_rate": 7.084754828264137e-05, "loss": 2.5207, "step": 19650 }, { "epoch": 1.7, "learning_rate": 7.084537990054354e-05, "loss": 2.1437, "step": 19660 }, { "epoch": 1.7, "learning_rate": 7.08432115184457e-05, "loss": 2.2296, "step": 19670 }, { "epoch": 1.7, "learning_rate": 7.084104313634785e-05, "loss": 2.3511, "step": 19680 }, { "epoch": 1.71, "learning_rate": 7.083887475425002e-05, "loss": 1.9569, "step": 19690 }, { "epoch": 1.71, "learning_rate": 7.083670637215219e-05, "loss": 2.1769, "step": 19700 }, { "epoch": 1.71, "learning_rate": 7.083453799005435e-05, "loss": 2.148, "step": 19710 }, { "epoch": 1.71, "learning_rate": 7.083236960795651e-05, "loss": 2.0108, "step": 19720 }, { "epoch": 1.71, "learning_rate": 7.083020122585867e-05, "loss": 2.0278, "step": 19730 }, { "epoch": 1.71, "learning_rate": 7.082803284376084e-05, "loss": 2.1481, "step": 19740 }, { "epoch": 1.71, "learning_rate": 7.082586446166299e-05, "loss": 2.1379, "step": 19750 }, { "epoch": 1.71, "learning_rate": 7.082369607956516e-05, "loss": 2.1899, "step": 19760 }, { "epoch": 1.71, "learning_rate": 7.082152769746733e-05, "loss": 2.1228, "step": 19770 }, { "epoch": 1.71, "learning_rate": 7.081935931536949e-05, "loss": 2.139, "step": 19780 }, { "epoch": 1.71, "learning_rate": 7.081719093327164e-05, "loss": 2.2662, "step": 19790 }, { "epoch": 1.71, "learning_rate": 7.081502255117381e-05, "loss": 2.0052, "step": 19800 }, { "epoch": 1.72, "learning_rate": 7.081285416907598e-05, "loss": 2.2993, "step": 19810 }, { "epoch": 1.72, "learning_rate": 7.081068578697813e-05, "loss": 2.2449, "step": 19820 }, { "epoch": 1.72, "learning_rate": 7.08085174048803e-05, "loss": 2.0471, "step": 19830 }, { "epoch": 1.72, "learning_rate": 7.080634902278246e-05, "loss": 2.0351, "step": 19840 }, { "epoch": 1.72, "learning_rate": 7.080418064068463e-05, "loss": 2.2149, "step": 19850 }, { "epoch": 1.72, "learning_rate": 7.08020122585868e-05, "loss": 2.2083, "step": 19860 }, { "epoch": 1.72, "learning_rate": 7.079984387648895e-05, "loss": 1.9639, "step": 19870 }, { "epoch": 1.72, "learning_rate": 7.079767549439111e-05, "loss": 1.7758, "step": 19880 }, { "epoch": 1.72, "learning_rate": 7.079550711229327e-05, "loss": 2.3595, "step": 19890 }, { "epoch": 1.72, "learning_rate": 7.079333873019543e-05, "loss": 2.214, "step": 19900 }, { "epoch": 1.72, "learning_rate": 7.07911703480976e-05, "loss": 2.0643, "step": 19910 }, { "epoch": 1.73, "learning_rate": 7.078900196599977e-05, "loss": 2.2171, "step": 19920 }, { "epoch": 1.73, "learning_rate": 7.078683358390193e-05, "loss": 1.909, "step": 19930 }, { "epoch": 1.73, "learning_rate": 7.078466520180409e-05, "loss": 2.2241, "step": 19940 }, { "epoch": 1.73, "learning_rate": 7.078249681970625e-05, "loss": 2.2734, "step": 19950 }, { "epoch": 1.73, "learning_rate": 7.07803284376084e-05, "loss": 2.0496, "step": 19960 }, { "epoch": 1.73, "learning_rate": 7.077816005551057e-05, "loss": 2.1318, "step": 19970 }, { "epoch": 1.73, "learning_rate": 7.077599167341274e-05, "loss": 2.1632, "step": 19980 }, { "epoch": 1.73, "learning_rate": 7.07738232913149e-05, "loss": 2.1599, "step": 19990 }, { "epoch": 1.73, "learning_rate": 7.077165490921707e-05, "loss": 2.0805, "step": 20000 }, { "epoch": 1.73, "learning_rate": 7.076948652711922e-05, "loss": 1.9155, "step": 20010 }, { "epoch": 1.73, "learning_rate": 7.076731814502139e-05, "loss": 1.9879, "step": 20020 }, { "epoch": 1.73, "learning_rate": 7.076514976292354e-05, "loss": 2.0855, "step": 20030 }, { "epoch": 1.74, "learning_rate": 7.076298138082571e-05, "loss": 2.2468, "step": 20040 }, { "epoch": 1.74, "learning_rate": 7.076081299872788e-05, "loss": 2.0758, "step": 20050 }, { "epoch": 1.74, "learning_rate": 7.075864461663004e-05, "loss": 2.0616, "step": 20060 }, { "epoch": 1.74, "learning_rate": 7.075647623453221e-05, "loss": 2.3004, "step": 20070 }, { "epoch": 1.74, "learning_rate": 7.075430785243436e-05, "loss": 2.245, "step": 20080 }, { "epoch": 1.74, "learning_rate": 7.075213947033653e-05, "loss": 2.008, "step": 20090 }, { "epoch": 1.74, "learning_rate": 7.074997108823868e-05, "loss": 2.089, "step": 20100 }, { "epoch": 1.74, "learning_rate": 7.074780270614085e-05, "loss": 1.8186, "step": 20110 }, { "epoch": 1.74, "learning_rate": 7.074563432404301e-05, "loss": 2.2414, "step": 20120 }, { "epoch": 1.74, "learning_rate": 7.074346594194518e-05, "loss": 1.783, "step": 20130 }, { "epoch": 1.74, "learning_rate": 7.074129755984735e-05, "loss": 2.2905, "step": 20140 }, { "epoch": 1.75, "learning_rate": 7.07391291777495e-05, "loss": 2.1309, "step": 20150 }, { "epoch": 1.75, "learning_rate": 7.073696079565167e-05, "loss": 2.1381, "step": 20160 }, { "epoch": 1.75, "learning_rate": 7.073479241355383e-05, "loss": 2.2949, "step": 20170 }, { "epoch": 1.75, "learning_rate": 7.073262403145598e-05, "loss": 2.2968, "step": 20180 }, { "epoch": 1.75, "learning_rate": 7.073045564935815e-05, "loss": 2.0762, "step": 20190 }, { "epoch": 1.75, "learning_rate": 7.072828726726032e-05, "loss": 2.1677, "step": 20200 }, { "epoch": 1.75, "learning_rate": 7.072611888516248e-05, "loss": 2.1784, "step": 20210 }, { "epoch": 1.75, "learning_rate": 7.072395050306464e-05, "loss": 2.0331, "step": 20220 }, { "epoch": 1.75, "learning_rate": 7.07217821209668e-05, "loss": 2.0662, "step": 20230 }, { "epoch": 1.75, "learning_rate": 7.071961373886897e-05, "loss": 2.0726, "step": 20240 }, { "epoch": 1.75, "learning_rate": 7.071744535677114e-05, "loss": 1.9686, "step": 20250 }, { "epoch": 1.75, "learning_rate": 7.071527697467329e-05, "loss": 2.0041, "step": 20260 }, { "epoch": 1.76, "learning_rate": 7.071310859257546e-05, "loss": 2.2143, "step": 20270 }, { "epoch": 1.76, "learning_rate": 7.071094021047762e-05, "loss": 2.2368, "step": 20280 }, { "epoch": 1.76, "learning_rate": 7.070877182837977e-05, "loss": 2.4099, "step": 20290 }, { "epoch": 1.76, "learning_rate": 7.070660344628194e-05, "loss": 2.1669, "step": 20300 }, { "epoch": 1.76, "learning_rate": 7.070443506418411e-05, "loss": 2.2408, "step": 20310 }, { "epoch": 1.76, "learning_rate": 7.070226668208627e-05, "loss": 2.3553, "step": 20320 }, { "epoch": 1.76, "learning_rate": 7.070009829998843e-05, "loss": 2.0013, "step": 20330 }, { "epoch": 1.76, "learning_rate": 7.069792991789059e-05, "loss": 2.2839, "step": 20340 }, { "epoch": 1.76, "learning_rate": 7.069576153579276e-05, "loss": 2.2773, "step": 20350 }, { "epoch": 1.76, "learning_rate": 7.069359315369491e-05, "loss": 2.1805, "step": 20360 }, { "epoch": 1.76, "learning_rate": 7.069142477159708e-05, "loss": 2.1738, "step": 20370 }, { "epoch": 1.77, "learning_rate": 7.068925638949924e-05, "loss": 1.934, "step": 20380 }, { "epoch": 1.77, "learning_rate": 7.068708800740141e-05, "loss": 2.1229, "step": 20390 }, { "epoch": 1.77, "learning_rate": 7.068491962530358e-05, "loss": 2.0221, "step": 20400 }, { "epoch": 1.77, "learning_rate": 7.068275124320573e-05, "loss": 2.2151, "step": 20410 }, { "epoch": 1.77, "learning_rate": 7.068058286110788e-05, "loss": 2.2416, "step": 20420 }, { "epoch": 1.77, "learning_rate": 7.067841447901005e-05, "loss": 2.253, "step": 20430 }, { "epoch": 1.77, "learning_rate": 7.067624609691222e-05, "loss": 2.1303, "step": 20440 }, { "epoch": 1.77, "learning_rate": 7.067407771481438e-05, "loss": 2.2332, "step": 20450 }, { "epoch": 1.77, "learning_rate": 7.067190933271655e-05, "loss": 2.3492, "step": 20460 }, { "epoch": 1.77, "learning_rate": 7.06697409506187e-05, "loss": 2.1593, "step": 20470 }, { "epoch": 1.77, "learning_rate": 7.066757256852087e-05, "loss": 2.2112, "step": 20480 }, { "epoch": 1.77, "learning_rate": 7.066540418642302e-05, "loss": 2.1723, "step": 20490 }, { "epoch": 1.78, "learning_rate": 7.066323580432519e-05, "loss": 2.2553, "step": 20500 }, { "epoch": 1.78, "learning_rate": 7.066106742222735e-05, "loss": 2.3515, "step": 20510 }, { "epoch": 1.78, "learning_rate": 7.065889904012952e-05, "loss": 2.0289, "step": 20520 }, { "epoch": 1.78, "learning_rate": 7.065673065803169e-05, "loss": 2.2798, "step": 20530 }, { "epoch": 1.78, "learning_rate": 7.065456227593384e-05, "loss": 1.9657, "step": 20540 }, { "epoch": 1.78, "learning_rate": 7.0652393893836e-05, "loss": 1.9382, "step": 20550 }, { "epoch": 1.78, "learning_rate": 7.065022551173816e-05, "loss": 2.1893, "step": 20560 }, { "epoch": 1.78, "learning_rate": 7.064805712964032e-05, "loss": 2.1346, "step": 20570 }, { "epoch": 1.78, "learning_rate": 7.064588874754249e-05, "loss": 2.1539, "step": 20580 }, { "epoch": 1.78, "learning_rate": 7.064372036544466e-05, "loss": 2.2206, "step": 20590 }, { "epoch": 1.78, "learning_rate": 7.064155198334682e-05, "loss": 2.1331, "step": 20600 }, { "epoch": 1.78, "learning_rate": 7.063938360124898e-05, "loss": 2.0029, "step": 20610 }, { "epoch": 1.79, "learning_rate": 7.063721521915114e-05, "loss": 2.0597, "step": 20620 }, { "epoch": 1.79, "learning_rate": 7.063504683705331e-05, "loss": 2.016, "step": 20630 }, { "epoch": 1.79, "learning_rate": 7.063287845495546e-05, "loss": 2.1535, "step": 20640 }, { "epoch": 1.79, "learning_rate": 7.063071007285763e-05, "loss": 2.2152, "step": 20650 }, { "epoch": 1.79, "learning_rate": 7.06285416907598e-05, "loss": 2.0247, "step": 20660 }, { "epoch": 1.79, "learning_rate": 7.062637330866196e-05, "loss": 2.2557, "step": 20670 }, { "epoch": 1.79, "learning_rate": 7.062420492656411e-05, "loss": 1.99, "step": 20680 }, { "epoch": 1.79, "learning_rate": 7.062203654446628e-05, "loss": 2.1176, "step": 20690 }, { "epoch": 1.79, "learning_rate": 7.061986816236845e-05, "loss": 2.0479, "step": 20700 }, { "epoch": 1.79, "learning_rate": 7.061769978027061e-05, "loss": 2.0214, "step": 20710 }, { "epoch": 1.79, "learning_rate": 7.061553139817277e-05, "loss": 2.0623, "step": 20720 }, { "epoch": 1.8, "learning_rate": 7.061336301607493e-05, "loss": 2.0917, "step": 20730 }, { "epoch": 1.8, "learning_rate": 7.06111946339771e-05, "loss": 2.1525, "step": 20740 }, { "epoch": 1.8, "learning_rate": 7.060902625187925e-05, "loss": 2.4636, "step": 20750 }, { "epoch": 1.8, "learning_rate": 7.060685786978142e-05, "loss": 2.1089, "step": 20760 }, { "epoch": 1.8, "learning_rate": 7.060468948768359e-05, "loss": 2.0012, "step": 20770 }, { "epoch": 1.8, "learning_rate": 7.060252110558575e-05, "loss": 2.2335, "step": 20780 }, { "epoch": 1.8, "learning_rate": 7.06003527234879e-05, "loss": 1.9683, "step": 20790 }, { "epoch": 1.8, "learning_rate": 7.059818434139007e-05, "loss": 1.7978, "step": 20800 }, { "epoch": 1.8, "learning_rate": 7.059601595929224e-05, "loss": 2.1336, "step": 20810 }, { "epoch": 1.8, "learning_rate": 7.059384757719439e-05, "loss": 2.1396, "step": 20820 }, { "epoch": 1.8, "learning_rate": 7.059167919509656e-05, "loss": 2.2097, "step": 20830 }, { "epoch": 1.8, "learning_rate": 7.058951081299872e-05, "loss": 2.3197, "step": 20840 }, { "epoch": 1.81, "learning_rate": 7.058734243090089e-05, "loss": 2.0978, "step": 20850 }, { "epoch": 1.81, "learning_rate": 7.058517404880306e-05, "loss": 2.2165, "step": 20860 }, { "epoch": 1.81, "learning_rate": 7.058300566670521e-05, "loss": 1.9448, "step": 20870 }, { "epoch": 1.81, "learning_rate": 7.058083728460737e-05, "loss": 2.0291, "step": 20880 }, { "epoch": 1.81, "learning_rate": 7.057866890250953e-05, "loss": 2.0947, "step": 20890 }, { "epoch": 1.81, "learning_rate": 7.05765005204117e-05, "loss": 2.1431, "step": 20900 }, { "epoch": 1.81, "learning_rate": 7.057433213831386e-05, "loss": 2.0274, "step": 20910 }, { "epoch": 1.81, "learning_rate": 7.057216375621603e-05, "loss": 2.1121, "step": 20920 }, { "epoch": 1.81, "learning_rate": 7.056999537411819e-05, "loss": 2.3763, "step": 20930 }, { "epoch": 1.81, "learning_rate": 7.056782699202035e-05, "loss": 2.0719, "step": 20940 }, { "epoch": 1.81, "learning_rate": 7.056565860992251e-05, "loss": 2.2491, "step": 20950 }, { "epoch": 1.82, "learning_rate": 7.056349022782467e-05, "loss": 2.2492, "step": 20960 }, { "epoch": 1.82, "learning_rate": 7.056132184572683e-05, "loss": 2.1531, "step": 20970 }, { "epoch": 1.82, "learning_rate": 7.0559153463629e-05, "loss": 2.163, "step": 20980 }, { "epoch": 1.82, "learning_rate": 7.055698508153116e-05, "loss": 2.2735, "step": 20990 }, { "epoch": 1.82, "learning_rate": 7.055481669943333e-05, "loss": 2.1879, "step": 21000 }, { "epoch": 1.82, "learning_rate": 7.055264831733548e-05, "loss": 2.2415, "step": 21010 }, { "epoch": 1.82, "learning_rate": 7.055047993523765e-05, "loss": 2.0822, "step": 21020 }, { "epoch": 1.82, "learning_rate": 7.05483115531398e-05, "loss": 1.9155, "step": 21030 }, { "epoch": 1.82, "learning_rate": 7.054614317104197e-05, "loss": 2.3274, "step": 21040 }, { "epoch": 1.82, "learning_rate": 7.054397478894414e-05, "loss": 2.1626, "step": 21050 }, { "epoch": 1.82, "learning_rate": 7.05418064068463e-05, "loss": 2.1034, "step": 21060 }, { "epoch": 1.82, "learning_rate": 7.053963802474847e-05, "loss": 2.3274, "step": 21070 }, { "epoch": 1.83, "learning_rate": 7.053746964265062e-05, "loss": 2.0608, "step": 21080 }, { "epoch": 1.83, "learning_rate": 7.053530126055279e-05, "loss": 2.2155, "step": 21090 }, { "epoch": 1.83, "learning_rate": 7.053313287845494e-05, "loss": 2.1429, "step": 21100 }, { "epoch": 1.83, "learning_rate": 7.053096449635711e-05, "loss": 2.1231, "step": 21110 }, { "epoch": 1.83, "learning_rate": 7.052879611425927e-05, "loss": 2.3015, "step": 21120 }, { "epoch": 1.83, "learning_rate": 7.052662773216144e-05, "loss": 2.1815, "step": 21130 }, { "epoch": 1.83, "learning_rate": 7.05244593500636e-05, "loss": 2.2555, "step": 21140 }, { "epoch": 1.83, "learning_rate": 7.052229096796576e-05, "loss": 2.1926, "step": 21150 }, { "epoch": 1.83, "learning_rate": 7.052012258586793e-05, "loss": 2.0332, "step": 21160 }, { "epoch": 1.83, "learning_rate": 7.051795420377009e-05, "loss": 2.3213, "step": 21170 }, { "epoch": 1.83, "learning_rate": 7.051578582167224e-05, "loss": 2.1011, "step": 21180 }, { "epoch": 1.84, "learning_rate": 7.051361743957441e-05, "loss": 2.2278, "step": 21190 }, { "epoch": 1.84, "learning_rate": 7.051144905747658e-05, "loss": 2.1139, "step": 21200 }, { "epoch": 1.84, "learning_rate": 7.050928067537874e-05, "loss": 2.0435, "step": 21210 }, { "epoch": 1.84, "learning_rate": 7.05071122932809e-05, "loss": 2.3644, "step": 21220 }, { "epoch": 1.84, "learning_rate": 7.050494391118306e-05, "loss": 2.2439, "step": 21230 }, { "epoch": 1.84, "learning_rate": 7.050277552908523e-05, "loss": 1.9594, "step": 21240 }, { "epoch": 1.84, "learning_rate": 7.05006071469874e-05, "loss": 2.2479, "step": 21250 }, { "epoch": 1.84, "learning_rate": 7.049843876488955e-05, "loss": 2.143, "step": 21260 }, { "epoch": 1.84, "learning_rate": 7.049627038279171e-05, "loss": 2.1421, "step": 21270 }, { "epoch": 1.84, "learning_rate": 7.049410200069388e-05, "loss": 2.2295, "step": 21280 }, { "epoch": 1.84, "learning_rate": 7.049193361859603e-05, "loss": 1.853, "step": 21290 }, { "epoch": 1.84, "learning_rate": 7.04897652364982e-05, "loss": 2.1692, "step": 21300 }, { "epoch": 1.85, "learning_rate": 7.048759685440037e-05, "loss": 2.0169, "step": 21310 }, { "epoch": 1.85, "learning_rate": 7.048542847230253e-05, "loss": 2.0331, "step": 21320 }, { "epoch": 1.85, "learning_rate": 7.048326009020469e-05, "loss": 2.0992, "step": 21330 }, { "epoch": 1.85, "learning_rate": 7.048109170810685e-05, "loss": 2.0038, "step": 21340 }, { "epoch": 1.85, "learning_rate": 7.047892332600902e-05, "loss": 1.9956, "step": 21350 }, { "epoch": 1.85, "learning_rate": 7.047675494391117e-05, "loss": 2.3846, "step": 21360 }, { "epoch": 1.85, "learning_rate": 7.047458656181334e-05, "loss": 2.2947, "step": 21370 }, { "epoch": 1.85, "learning_rate": 7.04724181797155e-05, "loss": 2.027, "step": 21380 }, { "epoch": 1.85, "learning_rate": 7.047024979761767e-05, "loss": 2.0788, "step": 21390 }, { "epoch": 1.85, "learning_rate": 7.046808141551984e-05, "loss": 2.2649, "step": 21400 }, { "epoch": 1.85, "learning_rate": 7.046591303342199e-05, "loss": 1.9843, "step": 21410 }, { "epoch": 1.86, "learning_rate": 7.046374465132416e-05, "loss": 1.9694, "step": 21420 }, { "epoch": 1.86, "learning_rate": 7.046157626922631e-05, "loss": 2.0222, "step": 21430 }, { "epoch": 1.86, "learning_rate": 7.045940788712848e-05, "loss": 2.1797, "step": 21440 }, { "epoch": 1.86, "learning_rate": 7.045723950503064e-05, "loss": 2.1183, "step": 21450 }, { "epoch": 1.86, "learning_rate": 7.045507112293281e-05, "loss": 2.0228, "step": 21460 }, { "epoch": 1.86, "learning_rate": 7.045290274083497e-05, "loss": 2.0079, "step": 21470 }, { "epoch": 1.86, "learning_rate": 7.045073435873713e-05, "loss": 1.9825, "step": 21480 }, { "epoch": 1.86, "learning_rate": 7.044856597663928e-05, "loss": 1.935, "step": 21490 }, { "epoch": 1.86, "learning_rate": 7.044639759454145e-05, "loss": 1.9842, "step": 21500 }, { "epoch": 1.86, "learning_rate": 7.044422921244361e-05, "loss": 2.1161, "step": 21510 }, { "epoch": 1.86, "learning_rate": 7.044206083034578e-05, "loss": 1.9743, "step": 21520 }, { "epoch": 1.86, "learning_rate": 7.043989244824795e-05, "loss": 2.2489, "step": 21530 }, { "epoch": 1.87, "learning_rate": 7.04377240661501e-05, "loss": 2.044, "step": 21540 }, { "epoch": 1.87, "learning_rate": 7.043555568405227e-05, "loss": 2.0076, "step": 21550 }, { "epoch": 1.87, "learning_rate": 7.043338730195442e-05, "loss": 2.4026, "step": 21560 }, { "epoch": 1.87, "learning_rate": 7.043121891985658e-05, "loss": 2.1664, "step": 21570 }, { "epoch": 1.87, "learning_rate": 7.042905053775875e-05, "loss": 2.143, "step": 21580 }, { "epoch": 1.87, "learning_rate": 7.042688215566092e-05, "loss": 2.2467, "step": 21590 }, { "epoch": 1.87, "learning_rate": 7.042471377356308e-05, "loss": 2.1434, "step": 21600 }, { "epoch": 1.87, "learning_rate": 7.042254539146524e-05, "loss": 2.155, "step": 21610 }, { "epoch": 1.87, "learning_rate": 7.04203770093674e-05, "loss": 1.9645, "step": 21620 }, { "epoch": 1.87, "learning_rate": 7.041820862726957e-05, "loss": 2.2711, "step": 21630 }, { "epoch": 1.87, "learning_rate": 7.041604024517172e-05, "loss": 2.2341, "step": 21640 }, { "epoch": 1.88, "learning_rate": 7.041387186307389e-05, "loss": 2.0394, "step": 21650 }, { "epoch": 1.88, "learning_rate": 7.041170348097606e-05, "loss": 2.1053, "step": 21660 }, { "epoch": 1.88, "learning_rate": 7.040953509887822e-05, "loss": 2.1404, "step": 21670 }, { "epoch": 1.88, "learning_rate": 7.040736671678037e-05, "loss": 2.2714, "step": 21680 }, { "epoch": 1.88, "learning_rate": 7.040519833468254e-05, "loss": 2.1876, "step": 21690 }, { "epoch": 1.88, "learning_rate": 7.040302995258471e-05, "loss": 2.226, "step": 21700 }, { "epoch": 1.88, "learning_rate": 7.040086157048687e-05, "loss": 2.0765, "step": 21710 }, { "epoch": 1.88, "learning_rate": 7.039869318838903e-05, "loss": 2.2084, "step": 21720 }, { "epoch": 1.88, "learning_rate": 7.039652480629119e-05, "loss": 2.0635, "step": 21730 }, { "epoch": 1.88, "learning_rate": 7.039435642419336e-05, "loss": 2.1703, "step": 21740 }, { "epoch": 1.88, "learning_rate": 7.039218804209551e-05, "loss": 2.0279, "step": 21750 }, { "epoch": 1.88, "learning_rate": 7.039001965999768e-05, "loss": 2.005, "step": 21760 }, { "epoch": 1.89, "learning_rate": 7.038785127789984e-05, "loss": 2.1758, "step": 21770 }, { "epoch": 1.89, "learning_rate": 7.038568289580201e-05, "loss": 2.1857, "step": 21780 }, { "epoch": 1.89, "learning_rate": 7.038351451370418e-05, "loss": 2.2452, "step": 21790 }, { "epoch": 1.89, "learning_rate": 7.038134613160633e-05, "loss": 2.0316, "step": 21800 }, { "epoch": 1.89, "learning_rate": 7.03791777495085e-05, "loss": 1.957, "step": 21810 }, { "epoch": 1.89, "learning_rate": 7.037700936741065e-05, "loss": 2.3652, "step": 21820 }, { "epoch": 1.89, "learning_rate": 7.037484098531282e-05, "loss": 2.0021, "step": 21830 }, { "epoch": 1.89, "learning_rate": 7.037267260321498e-05, "loss": 2.1463, "step": 21840 }, { "epoch": 1.89, "learning_rate": 7.037050422111715e-05, "loss": 1.9374, "step": 21850 }, { "epoch": 1.89, "learning_rate": 7.036833583901932e-05, "loss": 2.1861, "step": 21860 }, { "epoch": 1.89, "learning_rate": 7.036616745692147e-05, "loss": 2.2244, "step": 21870 }, { "epoch": 1.89, "learning_rate": 7.036399907482363e-05, "loss": 2.3237, "step": 21880 }, { "epoch": 1.9, "learning_rate": 7.036183069272579e-05, "loss": 2.1816, "step": 21890 }, { "epoch": 1.9, "learning_rate": 7.035966231062795e-05, "loss": 2.0466, "step": 21900 }, { "epoch": 1.9, "learning_rate": 7.035749392853012e-05, "loss": 1.9431, "step": 21910 }, { "epoch": 1.9, "learning_rate": 7.035532554643229e-05, "loss": 2.361, "step": 21920 }, { "epoch": 1.9, "learning_rate": 7.035315716433445e-05, "loss": 2.1641, "step": 21930 }, { "epoch": 1.9, "learning_rate": 7.03509887822366e-05, "loss": 2.0024, "step": 21940 }, { "epoch": 1.9, "learning_rate": 7.034882040013877e-05, "loss": 2.1829, "step": 21950 }, { "epoch": 1.9, "learning_rate": 7.034665201804092e-05, "loss": 2.063, "step": 21960 }, { "epoch": 1.9, "learning_rate": 7.034448363594309e-05, "loss": 2.286, "step": 21970 }, { "epoch": 1.9, "learning_rate": 7.034231525384526e-05, "loss": 1.8716, "step": 21980 }, { "epoch": 1.9, "learning_rate": 7.034014687174742e-05, "loss": 2.0874, "step": 21990 }, { "epoch": 1.91, "learning_rate": 7.033797848964959e-05, "loss": 2.1685, "step": 22000 }, { "epoch": 1.91, "learning_rate": 7.033581010755174e-05, "loss": 2.1097, "step": 22010 }, { "epoch": 1.91, "learning_rate": 7.033364172545391e-05, "loss": 2.3242, "step": 22020 }, { "epoch": 1.91, "learning_rate": 7.033147334335606e-05, "loss": 2.0656, "step": 22030 }, { "epoch": 1.91, "learning_rate": 7.032930496125823e-05, "loss": 2.2962, "step": 22040 }, { "epoch": 1.91, "learning_rate": 7.03271365791604e-05, "loss": 2.189, "step": 22050 }, { "epoch": 1.91, "learning_rate": 7.032496819706256e-05, "loss": 1.9257, "step": 22060 }, { "epoch": 1.91, "learning_rate": 7.032279981496473e-05, "loss": 2.2283, "step": 22070 }, { "epoch": 1.91, "learning_rate": 7.032063143286688e-05, "loss": 2.1487, "step": 22080 }, { "epoch": 1.91, "learning_rate": 7.031846305076905e-05, "loss": 1.9631, "step": 22090 }, { "epoch": 1.91, "learning_rate": 7.03162946686712e-05, "loss": 2.1501, "step": 22100 }, { "epoch": 1.91, "learning_rate": 7.031412628657337e-05, "loss": 1.9911, "step": 22110 }, { "epoch": 1.92, "learning_rate": 7.031195790447553e-05, "loss": 2.2261, "step": 22120 }, { "epoch": 1.92, "learning_rate": 7.03097895223777e-05, "loss": 2.0498, "step": 22130 }, { "epoch": 1.92, "learning_rate": 7.030762114027987e-05, "loss": 2.1494, "step": 22140 }, { "epoch": 1.92, "learning_rate": 7.030545275818202e-05, "loss": 1.8927, "step": 22150 }, { "epoch": 1.92, "learning_rate": 7.030328437608418e-05, "loss": 1.9201, "step": 22160 }, { "epoch": 1.92, "learning_rate": 7.030111599398635e-05, "loss": 2.0585, "step": 22170 }, { "epoch": 1.92, "learning_rate": 7.02989476118885e-05, "loss": 2.0122, "step": 22180 }, { "epoch": 1.92, "learning_rate": 7.029677922979067e-05, "loss": 2.0821, "step": 22190 }, { "epoch": 1.92, "learning_rate": 7.029461084769284e-05, "loss": 1.9299, "step": 22200 }, { "epoch": 1.92, "learning_rate": 7.0292442465595e-05, "loss": 2.1663, "step": 22210 }, { "epoch": 1.92, "learning_rate": 7.029027408349716e-05, "loss": 2.0, "step": 22220 }, { "epoch": 1.93, "learning_rate": 7.028810570139932e-05, "loss": 2.1504, "step": 22230 }, { "epoch": 1.93, "learning_rate": 7.028593731930149e-05, "loss": 2.1415, "step": 22240 }, { "epoch": 1.93, "learning_rate": 7.028376893720366e-05, "loss": 2.0434, "step": 22250 }, { "epoch": 1.93, "learning_rate": 7.028160055510581e-05, "loss": 2.1169, "step": 22260 }, { "epoch": 1.93, "learning_rate": 7.027943217300797e-05, "loss": 2.0939, "step": 22270 }, { "epoch": 1.93, "learning_rate": 7.027726379091014e-05, "loss": 1.9464, "step": 22280 }, { "epoch": 1.93, "learning_rate": 7.02750954088123e-05, "loss": 2.2433, "step": 22290 }, { "epoch": 1.93, "learning_rate": 7.027292702671446e-05, "loss": 2.0212, "step": 22300 }, { "epoch": 1.93, "learning_rate": 7.027075864461663e-05, "loss": 2.1322, "step": 22310 }, { "epoch": 1.93, "learning_rate": 7.026859026251879e-05, "loss": 2.1151, "step": 22320 }, { "epoch": 1.93, "learning_rate": 7.026642188042095e-05, "loss": 2.0752, "step": 22330 }, { "epoch": 1.93, "learning_rate": 7.026425349832311e-05, "loss": 2.1492, "step": 22340 }, { "epoch": 1.94, "learning_rate": 7.026208511622528e-05, "loss": 2.1378, "step": 22350 }, { "epoch": 1.94, "learning_rate": 7.025991673412743e-05, "loss": 2.2331, "step": 22360 }, { "epoch": 1.94, "learning_rate": 7.02577483520296e-05, "loss": 2.1628, "step": 22370 }, { "epoch": 1.94, "learning_rate": 7.025557996993176e-05, "loss": 2.1267, "step": 22380 }, { "epoch": 1.94, "learning_rate": 7.025341158783393e-05, "loss": 2.1721, "step": 22390 }, { "epoch": 1.94, "learning_rate": 7.02512432057361e-05, "loss": 2.2078, "step": 22400 }, { "epoch": 1.94, "learning_rate": 7.024907482363825e-05, "loss": 2.2966, "step": 22410 }, { "epoch": 1.94, "learning_rate": 7.024690644154042e-05, "loss": 1.9681, "step": 22420 }, { "epoch": 1.94, "learning_rate": 7.024473805944257e-05, "loss": 2.1835, "step": 22430 }, { "epoch": 1.94, "learning_rate": 7.024256967734474e-05, "loss": 2.2379, "step": 22440 }, { "epoch": 1.94, "learning_rate": 7.02404012952469e-05, "loss": 2.0605, "step": 22450 }, { "epoch": 1.95, "learning_rate": 7.023823291314907e-05, "loss": 2.0981, "step": 22460 }, { "epoch": 1.95, "learning_rate": 7.023606453105123e-05, "loss": 2.0847, "step": 22470 }, { "epoch": 1.95, "learning_rate": 7.023389614895339e-05, "loss": 2.0382, "step": 22480 }, { "epoch": 1.95, "learning_rate": 7.023172776685555e-05, "loss": 2.0356, "step": 22490 }, { "epoch": 1.95, "learning_rate": 7.02295593847577e-05, "loss": 1.9929, "step": 22500 }, { "epoch": 1.95, "learning_rate": 7.022739100265987e-05, "loss": 2.0627, "step": 22510 }, { "epoch": 1.95, "learning_rate": 7.022522262056204e-05, "loss": 2.2198, "step": 22520 }, { "epoch": 1.95, "learning_rate": 7.02230542384642e-05, "loss": 2.319, "step": 22530 }, { "epoch": 1.95, "learning_rate": 7.022088585636637e-05, "loss": 2.0495, "step": 22540 }, { "epoch": 1.95, "learning_rate": 7.021871747426853e-05, "loss": 2.1146, "step": 22550 }, { "epoch": 1.95, "learning_rate": 7.021654909217069e-05, "loss": 2.0776, "step": 22560 }, { "epoch": 1.95, "learning_rate": 7.021438071007284e-05, "loss": 2.0658, "step": 22570 }, { "epoch": 1.96, "learning_rate": 7.021221232797501e-05, "loss": 2.076, "step": 22580 }, { "epoch": 1.96, "learning_rate": 7.021004394587718e-05, "loss": 2.0418, "step": 22590 }, { "epoch": 1.96, "learning_rate": 7.020787556377934e-05, "loss": 2.0512, "step": 22600 }, { "epoch": 1.96, "learning_rate": 7.02057071816815e-05, "loss": 2.2272, "step": 22610 }, { "epoch": 1.96, "learning_rate": 7.020353879958366e-05, "loss": 2.1097, "step": 22620 }, { "epoch": 1.96, "learning_rate": 7.020137041748583e-05, "loss": 2.2069, "step": 22630 }, { "epoch": 1.96, "learning_rate": 7.019920203538798e-05, "loss": 2.1277, "step": 22640 }, { "epoch": 1.96, "learning_rate": 7.019703365329015e-05, "loss": 1.9842, "step": 22650 }, { "epoch": 1.96, "learning_rate": 7.019486527119231e-05, "loss": 1.9006, "step": 22660 }, { "epoch": 1.96, "learning_rate": 7.019269688909448e-05, "loss": 2.1081, "step": 22670 }, { "epoch": 1.96, "learning_rate": 7.019052850699663e-05, "loss": 1.8751, "step": 22680 }, { "epoch": 1.97, "learning_rate": 7.01883601248988e-05, "loss": 2.1485, "step": 22690 }, { "epoch": 1.97, "learning_rate": 7.018619174280097e-05, "loss": 1.9871, "step": 22700 }, { "epoch": 1.97, "learning_rate": 7.018402336070313e-05, "loss": 2.0722, "step": 22710 }, { "epoch": 1.97, "learning_rate": 7.018185497860529e-05, "loss": 2.2934, "step": 22720 }, { "epoch": 1.97, "learning_rate": 7.017968659650745e-05, "loss": 2.2253, "step": 22730 }, { "epoch": 1.97, "learning_rate": 7.017751821440962e-05, "loss": 1.9165, "step": 22740 }, { "epoch": 1.97, "learning_rate": 7.017534983231177e-05, "loss": 2.1926, "step": 22750 }, { "epoch": 1.97, "learning_rate": 7.017318145021394e-05, "loss": 2.2907, "step": 22760 }, { "epoch": 1.97, "learning_rate": 7.01710130681161e-05, "loss": 1.9256, "step": 22770 }, { "epoch": 1.97, "learning_rate": 7.016884468601827e-05, "loss": 2.1584, "step": 22780 }, { "epoch": 1.97, "learning_rate": 7.016667630392044e-05, "loss": 2.2268, "step": 22790 }, { "epoch": 1.97, "learning_rate": 7.016450792182259e-05, "loss": 1.9554, "step": 22800 }, { "epoch": 1.98, "learning_rate": 7.016233953972476e-05, "loss": 2.0193, "step": 22810 }, { "epoch": 1.98, "learning_rate": 7.016017115762691e-05, "loss": 1.8512, "step": 22820 }, { "epoch": 1.98, "learning_rate": 7.015800277552908e-05, "loss": 2.0394, "step": 22830 }, { "epoch": 1.98, "learning_rate": 7.015583439343124e-05, "loss": 1.8992, "step": 22840 }, { "epoch": 1.98, "learning_rate": 7.015366601133341e-05, "loss": 2.2927, "step": 22850 }, { "epoch": 1.98, "learning_rate": 7.015149762923557e-05, "loss": 2.2343, "step": 22860 }, { "epoch": 1.98, "learning_rate": 7.014932924713773e-05, "loss": 2.1545, "step": 22870 }, { "epoch": 1.98, "learning_rate": 7.01471608650399e-05, "loss": 1.9432, "step": 22880 }, { "epoch": 1.98, "learning_rate": 7.014499248294205e-05, "loss": 2.0191, "step": 22890 }, { "epoch": 1.98, "learning_rate": 7.014282410084421e-05, "loss": 2.2071, "step": 22900 }, { "epoch": 1.98, "learning_rate": 7.014065571874638e-05, "loss": 2.1473, "step": 22910 }, { "epoch": 1.99, "learning_rate": 7.013848733664855e-05, "loss": 2.1335, "step": 22920 }, { "epoch": 1.99, "learning_rate": 7.013631895455071e-05, "loss": 2.1876, "step": 22930 }, { "epoch": 1.99, "learning_rate": 7.013415057245287e-05, "loss": 2.0848, "step": 22940 }, { "epoch": 1.99, "learning_rate": 7.013198219035503e-05, "loss": 2.2581, "step": 22950 }, { "epoch": 1.99, "learning_rate": 7.012981380825718e-05, "loss": 2.0479, "step": 22960 }, { "epoch": 1.99, "learning_rate": 7.012764542615935e-05, "loss": 2.2978, "step": 22970 }, { "epoch": 1.99, "learning_rate": 7.012547704406152e-05, "loss": 1.8528, "step": 22980 }, { "epoch": 1.99, "learning_rate": 7.012330866196368e-05, "loss": 1.8764, "step": 22990 }, { "epoch": 1.99, "learning_rate": 7.012114027986585e-05, "loss": 1.9132, "step": 23000 }, { "epoch": 1.99, "learning_rate": 7.0118971897768e-05, "loss": 2.0268, "step": 23010 }, { "epoch": 1.99, "learning_rate": 7.011680351567017e-05, "loss": 2.264, "step": 23020 }, { "epoch": 1.99, "learning_rate": 7.011463513357232e-05, "loss": 2.1576, "step": 23030 }, { "epoch": 2.0, "learning_rate": 7.011246675147449e-05, "loss": 2.0906, "step": 23040 }, { "epoch": 2.0, "learning_rate": 7.011029836937665e-05, "loss": 2.0918, "step": 23050 }, { "epoch": 2.0, "learning_rate": 7.010812998727882e-05, "loss": 2.0862, "step": 23060 }, { "epoch": 2.0, "learning_rate": 7.010596160518099e-05, "loss": 2.1817, "step": 23070 }, { "epoch": 2.0, "learning_rate": 7.010379322308314e-05, "loss": 1.9041, "step": 23080 }, { "epoch": 2.0, "learning_rate": 7.010162484098531e-05, "loss": 2.0626, "step": 23090 }, { "epoch": 2.0, "eval_cer": 0.553762116352821, "eval_em": 0.0025786487880350697, "eval_f1": 0.0025786487880350697, "eval_loss": 1.8428248167037964, "eval_runtime": 250.5476, "eval_samples_per_second": 7.739, "eval_steps_per_second": 0.97, "eval_wer": 0.997421351211965, "step": 23093 }, { "epoch": 2.0, "learning_rate": 7.009945645888746e-05, "loss": 2.1702, "step": 23100 }, { "epoch": 2.0, "learning_rate": 7.009728807678963e-05, "loss": 2.1595, "step": 23110 }, { "epoch": 2.0, "learning_rate": 7.009511969469179e-05, "loss": 2.0848, "step": 23120 }, { "epoch": 2.0, "learning_rate": 7.009295131259396e-05, "loss": 1.9526, "step": 23130 }, { "epoch": 2.0, "learning_rate": 7.009078293049613e-05, "loss": 2.0551, "step": 23140 }, { "epoch": 2.0, "learning_rate": 7.008861454839828e-05, "loss": 2.1178, "step": 23150 }, { "epoch": 2.01, "learning_rate": 7.008644616630044e-05, "loss": 2.1536, "step": 23160 }, { "epoch": 2.01, "learning_rate": 7.008427778420261e-05, "loss": 1.8861, "step": 23170 }, { "epoch": 2.01, "learning_rate": 7.008210940210476e-05, "loss": 2.1034, "step": 23180 }, { "epoch": 2.01, "learning_rate": 7.007994102000693e-05, "loss": 1.9273, "step": 23190 }, { "epoch": 2.01, "learning_rate": 7.00777726379091e-05, "loss": 1.8725, "step": 23200 }, { "epoch": 2.01, "learning_rate": 7.007560425581126e-05, "loss": 1.9605, "step": 23210 }, { "epoch": 2.01, "learning_rate": 7.007343587371342e-05, "loss": 1.8522, "step": 23220 }, { "epoch": 2.01, "learning_rate": 7.007126749161558e-05, "loss": 1.9342, "step": 23230 }, { "epoch": 2.01, "learning_rate": 7.006909910951775e-05, "loss": 2.0723, "step": 23240 }, { "epoch": 2.01, "learning_rate": 7.006693072741992e-05, "loss": 2.0888, "step": 23250 }, { "epoch": 2.01, "learning_rate": 7.006476234532207e-05, "loss": 2.1096, "step": 23260 }, { "epoch": 2.02, "learning_rate": 7.006259396322423e-05, "loss": 1.9254, "step": 23270 }, { "epoch": 2.02, "learning_rate": 7.00604255811264e-05, "loss": 2.0364, "step": 23280 }, { "epoch": 2.02, "learning_rate": 7.005825719902855e-05, "loss": 1.9598, "step": 23290 }, { "epoch": 2.02, "learning_rate": 7.005608881693072e-05, "loss": 1.7782, "step": 23300 }, { "epoch": 2.02, "learning_rate": 7.005392043483289e-05, "loss": 1.9433, "step": 23310 }, { "epoch": 2.02, "learning_rate": 7.005175205273505e-05, "loss": 1.9355, "step": 23320 }, { "epoch": 2.02, "learning_rate": 7.004958367063722e-05, "loss": 2.0413, "step": 23330 }, { "epoch": 2.02, "learning_rate": 7.004741528853937e-05, "loss": 2.1556, "step": 23340 }, { "epoch": 2.02, "learning_rate": 7.004524690644154e-05, "loss": 2.3032, "step": 23350 }, { "epoch": 2.02, "learning_rate": 7.004307852434369e-05, "loss": 1.9755, "step": 23360 }, { "epoch": 2.02, "learning_rate": 7.004091014224586e-05, "loss": 1.9299, "step": 23370 }, { "epoch": 2.02, "learning_rate": 7.003874176014802e-05, "loss": 1.9372, "step": 23380 }, { "epoch": 2.03, "learning_rate": 7.003657337805019e-05, "loss": 2.2002, "step": 23390 }, { "epoch": 2.03, "learning_rate": 7.003440499595236e-05, "loss": 2.2277, "step": 23400 }, { "epoch": 2.03, "learning_rate": 7.003223661385451e-05, "loss": 1.9565, "step": 23410 }, { "epoch": 2.03, "learning_rate": 7.003006823175668e-05, "loss": 2.2106, "step": 23420 }, { "epoch": 2.03, "learning_rate": 7.002789984965883e-05, "loss": 2.0695, "step": 23430 }, { "epoch": 2.03, "learning_rate": 7.0025731467561e-05, "loss": 1.8537, "step": 23440 }, { "epoch": 2.03, "learning_rate": 7.002356308546316e-05, "loss": 1.8499, "step": 23450 }, { "epoch": 2.03, "learning_rate": 7.002139470336533e-05, "loss": 2.0587, "step": 23460 }, { "epoch": 2.03, "learning_rate": 7.00192263212675e-05, "loss": 2.0456, "step": 23470 }, { "epoch": 2.03, "learning_rate": 7.001705793916965e-05, "loss": 1.8916, "step": 23480 }, { "epoch": 2.03, "learning_rate": 7.001488955707181e-05, "loss": 1.8942, "step": 23490 }, { "epoch": 2.04, "learning_rate": 7.001272117497397e-05, "loss": 1.9004, "step": 23500 }, { "epoch": 2.04, "learning_rate": 7.001055279287613e-05, "loss": 1.9945, "step": 23510 }, { "epoch": 2.04, "learning_rate": 7.00083844107783e-05, "loss": 2.0006, "step": 23520 }, { "epoch": 2.04, "learning_rate": 7.000621602868047e-05, "loss": 2.0241, "step": 23530 }, { "epoch": 2.04, "learning_rate": 7.000404764658263e-05, "loss": 2.0643, "step": 23540 }, { "epoch": 2.04, "learning_rate": 7.000187926448478e-05, "loss": 2.1434, "step": 23550 }, { "epoch": 2.04, "learning_rate": 6.999971088238695e-05, "loss": 2.0495, "step": 23560 }, { "epoch": 2.04, "learning_rate": 6.99975425002891e-05, "loss": 1.9451, "step": 23570 }, { "epoch": 2.04, "learning_rate": 6.999537411819127e-05, "loss": 1.8597, "step": 23580 }, { "epoch": 2.04, "learning_rate": 6.999320573609344e-05, "loss": 1.9413, "step": 23590 }, { "epoch": 2.04, "learning_rate": 6.99910373539956e-05, "loss": 1.9332, "step": 23600 }, { "epoch": 2.04, "learning_rate": 6.998886897189777e-05, "loss": 2.0273, "step": 23610 }, { "epoch": 2.05, "learning_rate": 6.998670058979992e-05, "loss": 1.979, "step": 23620 }, { "epoch": 2.05, "learning_rate": 6.998453220770209e-05, "loss": 1.9933, "step": 23630 }, { "epoch": 2.05, "learning_rate": 6.998236382560424e-05, "loss": 2.0474, "step": 23640 }, { "epoch": 2.05, "learning_rate": 6.998019544350641e-05, "loss": 1.9668, "step": 23650 }, { "epoch": 2.05, "learning_rate": 6.997802706140857e-05, "loss": 2.2367, "step": 23660 }, { "epoch": 2.05, "learning_rate": 6.997585867931074e-05, "loss": 2.0276, "step": 23670 }, { "epoch": 2.05, "learning_rate": 6.99736902972129e-05, "loss": 1.9367, "step": 23680 }, { "epoch": 2.05, "learning_rate": 6.997152191511506e-05, "loss": 2.0689, "step": 23690 }, { "epoch": 2.05, "learning_rate": 6.996935353301723e-05, "loss": 2.047, "step": 23700 }, { "epoch": 2.05, "learning_rate": 6.996718515091939e-05, "loss": 1.969, "step": 23710 }, { "epoch": 2.05, "learning_rate": 6.996501676882155e-05, "loss": 1.9232, "step": 23720 }, { "epoch": 2.06, "learning_rate": 6.996284838672371e-05, "loss": 1.8822, "step": 23730 }, { "epoch": 2.06, "learning_rate": 6.996068000462588e-05, "loss": 2.0883, "step": 23740 }, { "epoch": 2.06, "learning_rate": 6.995851162252803e-05, "loss": 2.0412, "step": 23750 }, { "epoch": 2.06, "learning_rate": 6.99563432404302e-05, "loss": 2.0235, "step": 23760 }, { "epoch": 2.06, "learning_rate": 6.995417485833236e-05, "loss": 1.9861, "step": 23770 }, { "epoch": 2.06, "learning_rate": 6.995200647623453e-05, "loss": 2.1372, "step": 23780 }, { "epoch": 2.06, "learning_rate": 6.99498380941367e-05, "loss": 1.8658, "step": 23790 }, { "epoch": 2.06, "learning_rate": 6.994766971203885e-05, "loss": 2.1079, "step": 23800 }, { "epoch": 2.06, "learning_rate": 6.994550132994102e-05, "loss": 1.969, "step": 23810 }, { "epoch": 2.06, "learning_rate": 6.994333294784317e-05, "loss": 1.9226, "step": 23820 }, { "epoch": 2.06, "learning_rate": 6.994116456574534e-05, "loss": 2.0255, "step": 23830 }, { "epoch": 2.06, "learning_rate": 6.99389961836475e-05, "loss": 1.9903, "step": 23840 }, { "epoch": 2.07, "learning_rate": 6.993682780154967e-05, "loss": 2.2101, "step": 23850 }, { "epoch": 2.07, "learning_rate": 6.993465941945183e-05, "loss": 2.0432, "step": 23860 }, { "epoch": 2.07, "learning_rate": 6.993249103735399e-05, "loss": 2.0571, "step": 23870 }, { "epoch": 2.07, "learning_rate": 6.993032265525615e-05, "loss": 1.8722, "step": 23880 }, { "epoch": 2.07, "learning_rate": 6.99281542731583e-05, "loss": 2.0817, "step": 23890 }, { "epoch": 2.07, "learning_rate": 6.992598589106047e-05, "loss": 2.0191, "step": 23900 }, { "epoch": 2.07, "learning_rate": 6.992381750896264e-05, "loss": 2.1615, "step": 23910 }, { "epoch": 2.07, "learning_rate": 6.99216491268648e-05, "loss": 1.9054, "step": 23920 }, { "epoch": 2.07, "learning_rate": 6.991948074476697e-05, "loss": 2.0023, "step": 23930 }, { "epoch": 2.07, "learning_rate": 6.991731236266912e-05, "loss": 2.0794, "step": 23940 }, { "epoch": 2.07, "learning_rate": 6.991514398057129e-05, "loss": 1.9383, "step": 23950 }, { "epoch": 2.08, "learning_rate": 6.991297559847344e-05, "loss": 2.0591, "step": 23960 }, { "epoch": 2.08, "learning_rate": 6.991080721637561e-05, "loss": 2.1082, "step": 23970 }, { "epoch": 2.08, "learning_rate": 6.990863883427778e-05, "loss": 2.0522, "step": 23980 }, { "epoch": 2.08, "learning_rate": 6.990647045217994e-05, "loss": 2.129, "step": 23990 }, { "epoch": 2.08, "learning_rate": 6.990430207008211e-05, "loss": 1.9527, "step": 24000 }, { "epoch": 2.08, "learning_rate": 6.990213368798426e-05, "loss": 1.993, "step": 24010 }, { "epoch": 2.08, "learning_rate": 6.989996530588643e-05, "loss": 1.9419, "step": 24020 }, { "epoch": 2.08, "learning_rate": 6.989779692378858e-05, "loss": 2.0897, "step": 24030 }, { "epoch": 2.08, "learning_rate": 6.989562854169075e-05, "loss": 2.201, "step": 24040 }, { "epoch": 2.08, "learning_rate": 6.989346015959291e-05, "loss": 2.133, "step": 24050 }, { "epoch": 2.08, "learning_rate": 6.989129177749508e-05, "loss": 1.9653, "step": 24060 }, { "epoch": 2.08, "learning_rate": 6.988912339539725e-05, "loss": 2.0875, "step": 24070 }, { "epoch": 2.09, "learning_rate": 6.98869550132994e-05, "loss": 2.0523, "step": 24080 }, { "epoch": 2.09, "learning_rate": 6.988478663120157e-05, "loss": 2.1971, "step": 24090 }, { "epoch": 2.09, "learning_rate": 6.988261824910373e-05, "loss": 2.0684, "step": 24100 }, { "epoch": 2.09, "learning_rate": 6.988044986700589e-05, "loss": 2.0331, "step": 24110 }, { "epoch": 2.09, "learning_rate": 6.987828148490805e-05, "loss": 2.2112, "step": 24120 }, { "epoch": 2.09, "learning_rate": 6.987611310281022e-05, "loss": 1.8641, "step": 24130 }, { "epoch": 2.09, "learning_rate": 6.987394472071239e-05, "loss": 1.9834, "step": 24140 }, { "epoch": 2.09, "learning_rate": 6.987177633861454e-05, "loss": 2.267, "step": 24150 }, { "epoch": 2.09, "learning_rate": 6.98696079565167e-05, "loss": 2.3372, "step": 24160 }, { "epoch": 2.09, "learning_rate": 6.986743957441887e-05, "loss": 1.7154, "step": 24170 }, { "epoch": 2.09, "learning_rate": 6.986527119232102e-05, "loss": 2.0396, "step": 24180 }, { "epoch": 2.1, "learning_rate": 6.986310281022319e-05, "loss": 2.0439, "step": 24190 }, { "epoch": 2.1, "learning_rate": 6.986093442812536e-05, "loss": 2.0451, "step": 24200 }, { "epoch": 2.1, "learning_rate": 6.985876604602752e-05, "loss": 1.9497, "step": 24210 }, { "epoch": 2.1, "learning_rate": 6.985659766392968e-05, "loss": 2.0818, "step": 24220 }, { "epoch": 2.1, "learning_rate": 6.985442928183184e-05, "loss": 2.0821, "step": 24230 }, { "epoch": 2.1, "learning_rate": 6.985226089973401e-05, "loss": 1.8711, "step": 24240 }, { "epoch": 2.1, "learning_rate": 6.985009251763617e-05, "loss": 2.0165, "step": 24250 }, { "epoch": 2.1, "learning_rate": 6.984792413553833e-05, "loss": 1.8351, "step": 24260 }, { "epoch": 2.1, "learning_rate": 6.98457557534405e-05, "loss": 1.9105, "step": 24270 }, { "epoch": 2.1, "learning_rate": 6.984358737134266e-05, "loss": 1.8437, "step": 24280 }, { "epoch": 2.1, "learning_rate": 6.984141898924481e-05, "loss": 1.7181, "step": 24290 }, { "epoch": 2.1, "learning_rate": 6.983925060714698e-05, "loss": 2.0614, "step": 24300 }, { "epoch": 2.11, "learning_rate": 6.983708222504915e-05, "loss": 1.9101, "step": 24310 }, { "epoch": 2.11, "learning_rate": 6.983491384295131e-05, "loss": 1.7061, "step": 24320 }, { "epoch": 2.11, "learning_rate": 6.983274546085348e-05, "loss": 2.1433, "step": 24330 }, { "epoch": 2.11, "learning_rate": 6.983057707875563e-05, "loss": 1.9405, "step": 24340 }, { "epoch": 2.11, "learning_rate": 6.98284086966578e-05, "loss": 1.9755, "step": 24350 }, { "epoch": 2.11, "learning_rate": 6.982624031455995e-05, "loss": 2.1947, "step": 24360 }, { "epoch": 2.11, "learning_rate": 6.982407193246212e-05, "loss": 2.1447, "step": 24370 }, { "epoch": 2.11, "learning_rate": 6.982190355036428e-05, "loss": 2.0544, "step": 24380 }, { "epoch": 2.11, "learning_rate": 6.981973516826645e-05, "loss": 2.0756, "step": 24390 }, { "epoch": 2.11, "learning_rate": 6.981756678616862e-05, "loss": 1.9353, "step": 24400 }, { "epoch": 2.11, "learning_rate": 6.981539840407077e-05, "loss": 2.0305, "step": 24410 }, { "epoch": 2.11, "learning_rate": 6.981323002197294e-05, "loss": 1.939, "step": 24420 }, { "epoch": 2.12, "learning_rate": 6.981106163987509e-05, "loss": 2.0648, "step": 24430 }, { "epoch": 2.12, "learning_rate": 6.980889325777725e-05, "loss": 1.9397, "step": 24440 }, { "epoch": 2.12, "learning_rate": 6.980672487567942e-05, "loss": 1.952, "step": 24450 }, { "epoch": 2.12, "learning_rate": 6.980455649358159e-05, "loss": 1.8452, "step": 24460 }, { "epoch": 2.12, "learning_rate": 6.980238811148375e-05, "loss": 1.7955, "step": 24470 }, { "epoch": 2.12, "learning_rate": 6.980021972938591e-05, "loss": 1.9858, "step": 24480 }, { "epoch": 2.12, "learning_rate": 6.979805134728807e-05, "loss": 1.9964, "step": 24490 }, { "epoch": 2.12, "learning_rate": 6.979588296519023e-05, "loss": 1.9152, "step": 24500 }, { "epoch": 2.12, "learning_rate": 6.979371458309239e-05, "loss": 2.0589, "step": 24510 }, { "epoch": 2.12, "learning_rate": 6.979154620099456e-05, "loss": 2.1392, "step": 24520 }, { "epoch": 2.12, "learning_rate": 6.978937781889673e-05, "loss": 2.1727, "step": 24530 }, { "epoch": 2.13, "learning_rate": 6.978720943679889e-05, "loss": 1.9776, "step": 24540 }, { "epoch": 2.13, "learning_rate": 6.978504105470104e-05, "loss": 2.1067, "step": 24550 }, { "epoch": 2.13, "learning_rate": 6.978287267260321e-05, "loss": 2.1081, "step": 24560 }, { "epoch": 2.13, "learning_rate": 6.978070429050536e-05, "loss": 1.7495, "step": 24570 }, { "epoch": 2.13, "learning_rate": 6.977853590840753e-05, "loss": 2.0716, "step": 24580 }, { "epoch": 2.13, "learning_rate": 6.97763675263097e-05, "loss": 2.0355, "step": 24590 }, { "epoch": 2.13, "learning_rate": 6.977419914421186e-05, "loss": 2.1691, "step": 24600 }, { "epoch": 2.13, "learning_rate": 6.977203076211403e-05, "loss": 1.9397, "step": 24610 }, { "epoch": 2.13, "learning_rate": 6.976986238001618e-05, "loss": 1.8967, "step": 24620 }, { "epoch": 2.13, "learning_rate": 6.976769399791835e-05, "loss": 2.1612, "step": 24630 }, { "epoch": 2.13, "learning_rate": 6.97655256158205e-05, "loss": 1.8077, "step": 24640 }, { "epoch": 2.13, "learning_rate": 6.976335723372267e-05, "loss": 1.7916, "step": 24650 }, { "epoch": 2.14, "learning_rate": 6.976118885162483e-05, "loss": 1.8379, "step": 24660 }, { "epoch": 2.14, "learning_rate": 6.9759020469527e-05, "loss": 2.1108, "step": 24670 }, { "epoch": 2.14, "learning_rate": 6.975685208742917e-05, "loss": 2.0989, "step": 24680 }, { "epoch": 2.14, "learning_rate": 6.975468370533132e-05, "loss": 1.9076, "step": 24690 }, { "epoch": 2.14, "learning_rate": 6.975251532323349e-05, "loss": 2.0993, "step": 24700 }, { "epoch": 2.14, "learning_rate": 6.975034694113565e-05, "loss": 2.0611, "step": 24710 }, { "epoch": 2.14, "learning_rate": 6.97481785590378e-05, "loss": 1.916, "step": 24720 }, { "epoch": 2.14, "learning_rate": 6.974601017693997e-05, "loss": 1.8382, "step": 24730 }, { "epoch": 2.14, "learning_rate": 6.974384179484214e-05, "loss": 2.025, "step": 24740 }, { "epoch": 2.14, "learning_rate": 6.974167341274429e-05, "loss": 1.9006, "step": 24750 }, { "epoch": 2.14, "learning_rate": 6.973950503064646e-05, "loss": 2.1068, "step": 24760 }, { "epoch": 2.15, "learning_rate": 6.973733664854862e-05, "loss": 2.0753, "step": 24770 }, { "epoch": 2.15, "learning_rate": 6.973516826645079e-05, "loss": 1.8702, "step": 24780 }, { "epoch": 2.15, "learning_rate": 6.973299988435296e-05, "loss": 2.0434, "step": 24790 }, { "epoch": 2.15, "learning_rate": 6.973083150225511e-05, "loss": 2.2069, "step": 24800 }, { "epoch": 2.15, "learning_rate": 6.972866312015728e-05, "loss": 2.0168, "step": 24810 }, { "epoch": 2.15, "learning_rate": 6.972649473805943e-05, "loss": 2.3323, "step": 24820 }, { "epoch": 2.15, "learning_rate": 6.97243263559616e-05, "loss": 2.0226, "step": 24830 }, { "epoch": 2.15, "learning_rate": 6.972215797386376e-05, "loss": 1.8627, "step": 24840 }, { "epoch": 2.15, "learning_rate": 6.971998959176593e-05, "loss": 2.0153, "step": 24850 }, { "epoch": 2.15, "learning_rate": 6.97178212096681e-05, "loss": 1.852, "step": 24860 }, { "epoch": 2.15, "learning_rate": 6.971565282757025e-05, "loss": 1.9901, "step": 24870 }, { "epoch": 2.15, "learning_rate": 6.971348444547241e-05, "loss": 1.9359, "step": 24880 }, { "epoch": 2.16, "learning_rate": 6.971131606337457e-05, "loss": 1.9799, "step": 24890 }, { "epoch": 2.16, "learning_rate": 6.970914768127673e-05, "loss": 2.0704, "step": 24900 }, { "epoch": 2.16, "learning_rate": 6.97069792991789e-05, "loss": 1.9341, "step": 24910 }, { "epoch": 2.16, "learning_rate": 6.970481091708107e-05, "loss": 2.0445, "step": 24920 }, { "epoch": 2.16, "learning_rate": 6.970264253498323e-05, "loss": 2.1424, "step": 24930 }, { "epoch": 2.16, "learning_rate": 6.970047415288538e-05, "loss": 2.0886, "step": 24940 }, { "epoch": 2.16, "learning_rate": 6.969830577078755e-05, "loss": 1.7996, "step": 24950 }, { "epoch": 2.16, "learning_rate": 6.96961373886897e-05, "loss": 1.9263, "step": 24960 }, { "epoch": 2.16, "learning_rate": 6.969396900659187e-05, "loss": 1.9154, "step": 24970 }, { "epoch": 2.16, "learning_rate": 6.969180062449404e-05, "loss": 1.9521, "step": 24980 }, { "epoch": 2.16, "learning_rate": 6.96896322423962e-05, "loss": 1.9648, "step": 24990 }, { "epoch": 2.17, "learning_rate": 6.968746386029837e-05, "loss": 1.8745, "step": 25000 }, { "epoch": 2.17, "learning_rate": 6.968529547820052e-05, "loss": 2.0556, "step": 25010 }, { "epoch": 2.17, "learning_rate": 6.968312709610269e-05, "loss": 1.9055, "step": 25020 }, { "epoch": 2.17, "learning_rate": 6.968095871400484e-05, "loss": 1.9798, "step": 25030 }, { "epoch": 2.17, "learning_rate": 6.967879033190701e-05, "loss": 1.7689, "step": 25040 }, { "epoch": 2.17, "learning_rate": 6.967662194980917e-05, "loss": 1.8999, "step": 25050 }, { "epoch": 2.17, "learning_rate": 6.967445356771134e-05, "loss": 1.9329, "step": 25060 }, { "epoch": 2.17, "learning_rate": 6.967228518561351e-05, "loss": 1.9733, "step": 25070 }, { "epoch": 2.17, "learning_rate": 6.967011680351566e-05, "loss": 2.2103, "step": 25080 }, { "epoch": 2.17, "learning_rate": 6.966794842141783e-05, "loss": 2.0356, "step": 25090 }, { "epoch": 2.17, "learning_rate": 6.966578003931999e-05, "loss": 2.0997, "step": 25100 }, { "epoch": 2.17, "learning_rate": 6.966361165722215e-05, "loss": 1.9506, "step": 25110 }, { "epoch": 2.18, "learning_rate": 6.966144327512431e-05, "loss": 2.0212, "step": 25120 }, { "epoch": 2.18, "learning_rate": 6.965927489302648e-05, "loss": 2.056, "step": 25130 }, { "epoch": 2.18, "learning_rate": 6.965710651092864e-05, "loss": 1.9654, "step": 25140 }, { "epoch": 2.18, "learning_rate": 6.96549381288308e-05, "loss": 1.8873, "step": 25150 }, { "epoch": 2.18, "learning_rate": 6.965276974673296e-05, "loss": 2.3188, "step": 25160 }, { "epoch": 2.18, "learning_rate": 6.965060136463513e-05, "loss": 1.7464, "step": 25170 }, { "epoch": 2.18, "learning_rate": 6.964843298253728e-05, "loss": 2.0539, "step": 25180 }, { "epoch": 2.18, "learning_rate": 6.964626460043945e-05, "loss": 2.0243, "step": 25190 }, { "epoch": 2.18, "learning_rate": 6.964409621834162e-05, "loss": 2.1256, "step": 25200 }, { "epoch": 2.18, "learning_rate": 6.964192783624378e-05, "loss": 2.0599, "step": 25210 }, { "epoch": 2.18, "learning_rate": 6.963975945414594e-05, "loss": 2.1103, "step": 25220 }, { "epoch": 2.19, "learning_rate": 6.96375910720481e-05, "loss": 1.8432, "step": 25230 }, { "epoch": 2.19, "learning_rate": 6.963542268995027e-05, "loss": 2.0058, "step": 25240 }, { "epoch": 2.19, "learning_rate": 6.963325430785243e-05, "loss": 2.1996, "step": 25250 }, { "epoch": 2.19, "learning_rate": 6.963108592575459e-05, "loss": 1.9189, "step": 25260 }, { "epoch": 2.19, "learning_rate": 6.962891754365675e-05, "loss": 2.0389, "step": 25270 }, { "epoch": 2.19, "learning_rate": 6.962674916155892e-05, "loss": 1.7316, "step": 25280 }, { "epoch": 2.19, "learning_rate": 6.962458077946107e-05, "loss": 2.3397, "step": 25290 }, { "epoch": 2.19, "learning_rate": 6.962241239736324e-05, "loss": 2.0761, "step": 25300 }, { "epoch": 2.19, "learning_rate": 6.96202440152654e-05, "loss": 1.8363, "step": 25310 }, { "epoch": 2.19, "learning_rate": 6.961807563316757e-05, "loss": 1.8812, "step": 25320 }, { "epoch": 2.19, "learning_rate": 6.961590725106974e-05, "loss": 2.106, "step": 25330 }, { "epoch": 2.19, "learning_rate": 6.961373886897189e-05, "loss": 1.8763, "step": 25340 }, { "epoch": 2.2, "learning_rate": 6.961157048687406e-05, "loss": 2.0205, "step": 25350 }, { "epoch": 2.2, "learning_rate": 6.960940210477621e-05, "loss": 1.931, "step": 25360 }, { "epoch": 2.2, "learning_rate": 6.960723372267838e-05, "loss": 2.2204, "step": 25370 }, { "epoch": 2.2, "learning_rate": 6.960506534058054e-05, "loss": 1.9643, "step": 25380 }, { "epoch": 2.2, "learning_rate": 6.960289695848271e-05, "loss": 1.8378, "step": 25390 }, { "epoch": 2.2, "learning_rate": 6.960072857638488e-05, "loss": 1.8819, "step": 25400 }, { "epoch": 2.2, "learning_rate": 6.959856019428703e-05, "loss": 1.9717, "step": 25410 }, { "epoch": 2.2, "learning_rate": 6.95963918121892e-05, "loss": 1.9717, "step": 25420 }, { "epoch": 2.2, "learning_rate": 6.959422343009135e-05, "loss": 2.0667, "step": 25430 }, { "epoch": 2.2, "learning_rate": 6.959205504799351e-05, "loss": 2.1457, "step": 25440 }, { "epoch": 2.2, "learning_rate": 6.958988666589568e-05, "loss": 1.8188, "step": 25450 }, { "epoch": 2.2, "learning_rate": 6.958771828379785e-05, "loss": 2.2575, "step": 25460 }, { "epoch": 2.21, "learning_rate": 6.958554990170001e-05, "loss": 2.036, "step": 25470 }, { "epoch": 2.21, "learning_rate": 6.958338151960217e-05, "loss": 2.0622, "step": 25480 }, { "epoch": 2.21, "learning_rate": 6.958121313750433e-05, "loss": 2.0289, "step": 25490 }, { "epoch": 2.21, "learning_rate": 6.957904475540649e-05, "loss": 1.8622, "step": 25500 }, { "epoch": 2.21, "learning_rate": 6.957687637330865e-05, "loss": 2.244, "step": 25510 }, { "epoch": 2.21, "learning_rate": 6.957470799121082e-05, "loss": 2.0073, "step": 25520 }, { "epoch": 2.21, "learning_rate": 6.957253960911298e-05, "loss": 2.1013, "step": 25530 }, { "epoch": 2.21, "learning_rate": 6.957037122701515e-05, "loss": 2.1, "step": 25540 }, { "epoch": 2.21, "learning_rate": 6.95682028449173e-05, "loss": 2.0102, "step": 25550 }, { "epoch": 2.21, "learning_rate": 6.956603446281947e-05, "loss": 1.8548, "step": 25560 }, { "epoch": 2.21, "learning_rate": 6.956386608072162e-05, "loss": 1.8358, "step": 25570 }, { "epoch": 2.22, "learning_rate": 6.956169769862379e-05, "loss": 1.8292, "step": 25580 }, { "epoch": 2.22, "learning_rate": 6.955952931652596e-05, "loss": 1.8944, "step": 25590 }, { "epoch": 2.22, "learning_rate": 6.955736093442812e-05, "loss": 1.9152, "step": 25600 }, { "epoch": 2.22, "learning_rate": 6.955519255233029e-05, "loss": 1.8522, "step": 25610 }, { "epoch": 2.22, "learning_rate": 6.955302417023244e-05, "loss": 2.3293, "step": 25620 }, { "epoch": 2.22, "learning_rate": 6.955085578813461e-05, "loss": 1.8503, "step": 25630 }, { "epoch": 2.22, "learning_rate": 6.954868740603677e-05, "loss": 2.0424, "step": 25640 }, { "epoch": 2.22, "learning_rate": 6.954651902393893e-05, "loss": 1.9121, "step": 25650 }, { "epoch": 2.22, "learning_rate": 6.95443506418411e-05, "loss": 2.0051, "step": 25660 }, { "epoch": 2.22, "learning_rate": 6.954218225974326e-05, "loss": 1.9192, "step": 25670 }, { "epoch": 2.22, "learning_rate": 6.954001387764543e-05, "loss": 2.0121, "step": 25680 }, { "epoch": 2.22, "learning_rate": 6.953784549554758e-05, "loss": 1.8751, "step": 25690 }, { "epoch": 2.23, "learning_rate": 6.953567711344975e-05, "loss": 1.9403, "step": 25700 }, { "epoch": 2.23, "learning_rate": 6.953350873135191e-05, "loss": 1.9136, "step": 25710 }, { "epoch": 2.23, "learning_rate": 6.953134034925407e-05, "loss": 2.1277, "step": 25720 }, { "epoch": 2.23, "learning_rate": 6.952917196715623e-05, "loss": 1.9105, "step": 25730 }, { "epoch": 2.23, "learning_rate": 6.95270035850584e-05, "loss": 1.9222, "step": 25740 }, { "epoch": 2.23, "learning_rate": 6.952483520296056e-05, "loss": 1.8898, "step": 25750 }, { "epoch": 2.23, "learning_rate": 6.952266682086272e-05, "loss": 2.1906, "step": 25760 }, { "epoch": 2.23, "learning_rate": 6.952049843876488e-05, "loss": 2.2092, "step": 25770 }, { "epoch": 2.23, "learning_rate": 6.951833005666705e-05, "loss": 1.9566, "step": 25780 }, { "epoch": 2.23, "learning_rate": 6.951616167456922e-05, "loss": 1.6663, "step": 25790 }, { "epoch": 2.23, "learning_rate": 6.951399329247137e-05, "loss": 2.1355, "step": 25800 }, { "epoch": 2.24, "learning_rate": 6.951182491037354e-05, "loss": 1.8975, "step": 25810 }, { "epoch": 2.24, "learning_rate": 6.950965652827569e-05, "loss": 2.0304, "step": 25820 }, { "epoch": 2.24, "learning_rate": 6.950748814617785e-05, "loss": 2.0598, "step": 25830 }, { "epoch": 2.24, "learning_rate": 6.950531976408002e-05, "loss": 2.0078, "step": 25840 }, { "epoch": 2.24, "learning_rate": 6.950315138198219e-05, "loss": 1.9388, "step": 25850 }, { "epoch": 2.24, "learning_rate": 6.950098299988435e-05, "loss": 1.8548, "step": 25860 }, { "epoch": 2.24, "learning_rate": 6.94988146177865e-05, "loss": 1.9578, "step": 25870 }, { "epoch": 2.24, "learning_rate": 6.949664623568867e-05, "loss": 2.1537, "step": 25880 }, { "epoch": 2.24, "learning_rate": 6.949447785359083e-05, "loss": 2.1454, "step": 25890 }, { "epoch": 2.24, "learning_rate": 6.949230947149299e-05, "loss": 1.9071, "step": 25900 }, { "epoch": 2.24, "learning_rate": 6.949014108939516e-05, "loss": 2.1381, "step": 25910 }, { "epoch": 2.24, "learning_rate": 6.948797270729733e-05, "loss": 1.8602, "step": 25920 }, { "epoch": 2.25, "learning_rate": 6.948580432519949e-05, "loss": 2.3904, "step": 25930 }, { "epoch": 2.25, "learning_rate": 6.948363594310164e-05, "loss": 1.8988, "step": 25940 }, { "epoch": 2.25, "learning_rate": 6.948146756100381e-05, "loss": 2.0112, "step": 25950 }, { "epoch": 2.25, "learning_rate": 6.947929917890596e-05, "loss": 2.0366, "step": 25960 }, { "epoch": 2.25, "learning_rate": 6.947713079680813e-05, "loss": 1.8071, "step": 25970 }, { "epoch": 2.25, "learning_rate": 6.94749624147103e-05, "loss": 1.9503, "step": 25980 }, { "epoch": 2.25, "learning_rate": 6.947279403261246e-05, "loss": 2.1006, "step": 25990 }, { "epoch": 2.25, "learning_rate": 6.947062565051463e-05, "loss": 2.2418, "step": 26000 }, { "epoch": 2.25, "learning_rate": 6.946845726841678e-05, "loss": 2.2689, "step": 26010 }, { "epoch": 2.25, "learning_rate": 6.946628888631895e-05, "loss": 2.0287, "step": 26020 }, { "epoch": 2.25, "learning_rate": 6.94641205042211e-05, "loss": 1.9437, "step": 26030 }, { "epoch": 2.26, "learning_rate": 6.946195212212327e-05, "loss": 2.0029, "step": 26040 }, { "epoch": 2.26, "learning_rate": 6.945978374002543e-05, "loss": 2.1157, "step": 26050 }, { "epoch": 2.26, "learning_rate": 6.94576153579276e-05, "loss": 2.0366, "step": 26060 }, { "epoch": 2.26, "learning_rate": 6.945544697582977e-05, "loss": 1.9657, "step": 26070 }, { "epoch": 2.26, "learning_rate": 6.945327859373192e-05, "loss": 2.03, "step": 26080 }, { "epoch": 2.26, "learning_rate": 6.945111021163409e-05, "loss": 2.0856, "step": 26090 }, { "epoch": 2.26, "learning_rate": 6.944894182953625e-05, "loss": 2.0653, "step": 26100 }, { "epoch": 2.26, "learning_rate": 6.94467734474384e-05, "loss": 1.9471, "step": 26110 }, { "epoch": 2.26, "learning_rate": 6.944460506534057e-05, "loss": 1.9044, "step": 26120 }, { "epoch": 2.26, "learning_rate": 6.944243668324274e-05, "loss": 2.0594, "step": 26130 }, { "epoch": 2.26, "learning_rate": 6.94402683011449e-05, "loss": 1.9479, "step": 26140 }, { "epoch": 2.26, "learning_rate": 6.943809991904706e-05, "loss": 2.0781, "step": 26150 }, { "epoch": 2.27, "learning_rate": 6.943593153694922e-05, "loss": 1.9935, "step": 26160 }, { "epoch": 2.27, "learning_rate": 6.943376315485139e-05, "loss": 2.2011, "step": 26170 }, { "epoch": 2.27, "learning_rate": 6.943159477275354e-05, "loss": 1.8664, "step": 26180 }, { "epoch": 2.27, "learning_rate": 6.942942639065571e-05, "loss": 2.013, "step": 26190 }, { "epoch": 2.27, "learning_rate": 6.942725800855788e-05, "loss": 1.9234, "step": 26200 }, { "epoch": 2.27, "learning_rate": 6.942508962646004e-05, "loss": 1.9639, "step": 26210 }, { "epoch": 2.27, "learning_rate": 6.94229212443622e-05, "loss": 2.0061, "step": 26220 }, { "epoch": 2.27, "learning_rate": 6.942075286226436e-05, "loss": 1.9749, "step": 26230 }, { "epoch": 2.27, "learning_rate": 6.941858448016653e-05, "loss": 2.1072, "step": 26240 }, { "epoch": 2.27, "learning_rate": 6.94164160980687e-05, "loss": 1.8914, "step": 26250 }, { "epoch": 2.27, "learning_rate": 6.941424771597085e-05, "loss": 2.0132, "step": 26260 }, { "epoch": 2.28, "learning_rate": 6.941207933387301e-05, "loss": 2.2375, "step": 26270 }, { "epoch": 2.28, "learning_rate": 6.940991095177518e-05, "loss": 1.9222, "step": 26280 }, { "epoch": 2.28, "learning_rate": 6.940774256967733e-05, "loss": 1.9075, "step": 26290 }, { "epoch": 2.28, "learning_rate": 6.94055741875795e-05, "loss": 2.0394, "step": 26300 }, { "epoch": 2.28, "learning_rate": 6.940340580548167e-05, "loss": 2.0653, "step": 26310 }, { "epoch": 2.28, "learning_rate": 6.940123742338383e-05, "loss": 1.9917, "step": 26320 }, { "epoch": 2.28, "learning_rate": 6.9399069041286e-05, "loss": 2.0142, "step": 26330 }, { "epoch": 2.28, "learning_rate": 6.939690065918815e-05, "loss": 1.9527, "step": 26340 }, { "epoch": 2.28, "learning_rate": 6.939473227709032e-05, "loss": 1.9655, "step": 26350 }, { "epoch": 2.28, "learning_rate": 6.939256389499247e-05, "loss": 2.1039, "step": 26360 }, { "epoch": 2.28, "learning_rate": 6.939039551289464e-05, "loss": 1.964, "step": 26370 }, { "epoch": 2.28, "learning_rate": 6.93882271307968e-05, "loss": 1.9219, "step": 26380 }, { "epoch": 2.29, "learning_rate": 6.938605874869897e-05, "loss": 2.0205, "step": 26390 }, { "epoch": 2.29, "learning_rate": 6.938389036660114e-05, "loss": 2.0336, "step": 26400 }, { "epoch": 2.29, "learning_rate": 6.938172198450329e-05, "loss": 1.9017, "step": 26410 }, { "epoch": 2.29, "learning_rate": 6.937955360240546e-05, "loss": 1.7984, "step": 26420 }, { "epoch": 2.29, "learning_rate": 6.937738522030761e-05, "loss": 1.916, "step": 26430 }, { "epoch": 2.29, "learning_rate": 6.937521683820977e-05, "loss": 1.8147, "step": 26440 }, { "epoch": 2.29, "learning_rate": 6.937304845611194e-05, "loss": 1.9687, "step": 26450 }, { "epoch": 2.29, "learning_rate": 6.937088007401411e-05, "loss": 2.0772, "step": 26460 }, { "epoch": 2.29, "learning_rate": 6.936871169191627e-05, "loss": 2.0094, "step": 26470 }, { "epoch": 2.29, "learning_rate": 6.936654330981843e-05, "loss": 1.9218, "step": 26480 }, { "epoch": 2.29, "learning_rate": 6.936437492772059e-05, "loss": 1.9077, "step": 26490 }, { "epoch": 2.3, "learning_rate": 6.936220654562275e-05, "loss": 2.0534, "step": 26500 }, { "epoch": 2.3, "learning_rate": 6.936003816352491e-05, "loss": 1.9315, "step": 26510 }, { "epoch": 2.3, "learning_rate": 6.935786978142708e-05, "loss": 2.1305, "step": 26520 }, { "epoch": 2.3, "learning_rate": 6.935570139932924e-05, "loss": 1.6963, "step": 26530 }, { "epoch": 2.3, "learning_rate": 6.935353301723141e-05, "loss": 2.0837, "step": 26540 }, { "epoch": 2.3, "learning_rate": 6.935136463513356e-05, "loss": 1.9795, "step": 26550 }, { "epoch": 2.3, "learning_rate": 6.934919625303573e-05, "loss": 1.914, "step": 26560 }, { "epoch": 2.3, "learning_rate": 6.934702787093788e-05, "loss": 1.9744, "step": 26570 }, { "epoch": 2.3, "learning_rate": 6.934485948884005e-05, "loss": 1.8685, "step": 26580 }, { "epoch": 2.3, "learning_rate": 6.934269110674222e-05, "loss": 2.0722, "step": 26590 }, { "epoch": 2.3, "learning_rate": 6.934052272464438e-05, "loss": 1.8327, "step": 26600 }, { "epoch": 2.3, "learning_rate": 6.933835434254655e-05, "loss": 1.9216, "step": 26610 }, { "epoch": 2.31, "learning_rate": 6.93361859604487e-05, "loss": 1.8717, "step": 26620 }, { "epoch": 2.31, "learning_rate": 6.933401757835087e-05, "loss": 2.0722, "step": 26630 }, { "epoch": 2.31, "learning_rate": 6.933184919625303e-05, "loss": 1.8954, "step": 26640 }, { "epoch": 2.31, "learning_rate": 6.932968081415519e-05, "loss": 2.0522, "step": 26650 }, { "epoch": 2.31, "learning_rate": 6.932751243205735e-05, "loss": 2.0339, "step": 26660 }, { "epoch": 2.31, "learning_rate": 6.932534404995952e-05, "loss": 1.6677, "step": 26670 }, { "epoch": 2.31, "learning_rate": 6.932317566786169e-05, "loss": 1.8215, "step": 26680 }, { "epoch": 2.31, "learning_rate": 6.932100728576384e-05, "loss": 2.0899, "step": 26690 }, { "epoch": 2.31, "learning_rate": 6.9318838903666e-05, "loss": 1.9236, "step": 26700 }, { "epoch": 2.31, "learning_rate": 6.931667052156817e-05, "loss": 1.8364, "step": 26710 }, { "epoch": 2.31, "learning_rate": 6.931450213947032e-05, "loss": 2.112, "step": 26720 }, { "epoch": 2.31, "learning_rate": 6.931233375737249e-05, "loss": 2.0069, "step": 26730 }, { "epoch": 2.32, "learning_rate": 6.931016537527466e-05, "loss": 1.9423, "step": 26740 }, { "epoch": 2.32, "learning_rate": 6.930799699317682e-05, "loss": 2.1404, "step": 26750 }, { "epoch": 2.32, "learning_rate": 6.930582861107898e-05, "loss": 1.9099, "step": 26760 }, { "epoch": 2.32, "learning_rate": 6.930366022898114e-05, "loss": 1.9087, "step": 26770 }, { "epoch": 2.32, "learning_rate": 6.930149184688331e-05, "loss": 1.8692, "step": 26780 }, { "epoch": 2.32, "learning_rate": 6.929932346478548e-05, "loss": 1.8668, "step": 26790 }, { "epoch": 2.32, "learning_rate": 6.929715508268763e-05, "loss": 2.0455, "step": 26800 }, { "epoch": 2.32, "learning_rate": 6.92949867005898e-05, "loss": 2.0303, "step": 26810 }, { "epoch": 2.32, "learning_rate": 6.929281831849196e-05, "loss": 2.0761, "step": 26820 }, { "epoch": 2.32, "learning_rate": 6.929064993639411e-05, "loss": 2.0803, "step": 26830 }, { "epoch": 2.32, "learning_rate": 6.928848155429628e-05, "loss": 2.098, "step": 26840 }, { "epoch": 2.33, "learning_rate": 6.928631317219845e-05, "loss": 2.0693, "step": 26850 }, { "epoch": 2.33, "learning_rate": 6.928414479010061e-05, "loss": 1.9374, "step": 26860 }, { "epoch": 2.33, "learning_rate": 6.928197640800278e-05, "loss": 1.935, "step": 26870 }, { "epoch": 2.33, "learning_rate": 6.927980802590493e-05, "loss": 2.0658, "step": 26880 }, { "epoch": 2.33, "learning_rate": 6.927763964380709e-05, "loss": 1.8768, "step": 26890 }, { "epoch": 2.33, "learning_rate": 6.927547126170925e-05, "loss": 2.1876, "step": 26900 }, { "epoch": 2.33, "learning_rate": 6.927330287961142e-05, "loss": 2.0077, "step": 26910 }, { "epoch": 2.33, "learning_rate": 6.927113449751358e-05, "loss": 1.9146, "step": 26920 }, { "epoch": 2.33, "learning_rate": 6.926896611541575e-05, "loss": 2.0443, "step": 26930 }, { "epoch": 2.33, "learning_rate": 6.92667977333179e-05, "loss": 2.0759, "step": 26940 }, { "epoch": 2.33, "learning_rate": 6.926462935122007e-05, "loss": 1.9476, "step": 26950 }, { "epoch": 2.33, "learning_rate": 6.926246096912222e-05, "loss": 2.2531, "step": 26960 }, { "epoch": 2.34, "learning_rate": 6.926029258702439e-05, "loss": 2.0909, "step": 26970 }, { "epoch": 2.34, "learning_rate": 6.925812420492656e-05, "loss": 1.9466, "step": 26980 }, { "epoch": 2.34, "learning_rate": 6.925595582282872e-05, "loss": 2.046, "step": 26990 }, { "epoch": 2.34, "learning_rate": 6.925378744073089e-05, "loss": 1.7967, "step": 27000 }, { "epoch": 2.34, "learning_rate": 6.925161905863304e-05, "loss": 2.0116, "step": 27010 }, { "epoch": 2.34, "learning_rate": 6.924945067653521e-05, "loss": 1.9318, "step": 27020 }, { "epoch": 2.34, "learning_rate": 6.924728229443736e-05, "loss": 1.8897, "step": 27030 }, { "epoch": 2.34, "learning_rate": 6.924511391233953e-05, "loss": 2.0535, "step": 27040 }, { "epoch": 2.34, "learning_rate": 6.92429455302417e-05, "loss": 1.8985, "step": 27050 }, { "epoch": 2.34, "learning_rate": 6.924077714814386e-05, "loss": 2.0175, "step": 27060 }, { "epoch": 2.34, "learning_rate": 6.923860876604603e-05, "loss": 2.0298, "step": 27070 }, { "epoch": 2.35, "learning_rate": 6.923644038394818e-05, "loss": 1.9345, "step": 27080 }, { "epoch": 2.35, "learning_rate": 6.923427200185035e-05, "loss": 2.0554, "step": 27090 }, { "epoch": 2.35, "learning_rate": 6.923210361975251e-05, "loss": 2.0368, "step": 27100 }, { "epoch": 2.35, "learning_rate": 6.922993523765466e-05, "loss": 1.9721, "step": 27110 }, { "epoch": 2.35, "learning_rate": 6.922776685555683e-05, "loss": 1.9029, "step": 27120 }, { "epoch": 2.35, "learning_rate": 6.9225598473459e-05, "loss": 1.88, "step": 27130 }, { "epoch": 2.35, "learning_rate": 6.922343009136116e-05, "loss": 1.9753, "step": 27140 }, { "epoch": 2.35, "learning_rate": 6.922126170926332e-05, "loss": 2.1575, "step": 27150 }, { "epoch": 2.35, "learning_rate": 6.921909332716548e-05, "loss": 2.0623, "step": 27160 }, { "epoch": 2.35, "learning_rate": 6.921692494506765e-05, "loss": 1.9428, "step": 27170 }, { "epoch": 2.35, "learning_rate": 6.921475656296982e-05, "loss": 1.8169, "step": 27180 }, { "epoch": 2.35, "learning_rate": 6.921258818087197e-05, "loss": 2.0967, "step": 27190 }, { "epoch": 2.36, "learning_rate": 6.921041979877414e-05, "loss": 1.8775, "step": 27200 }, { "epoch": 2.36, "learning_rate": 6.92082514166763e-05, "loss": 2.0629, "step": 27210 }, { "epoch": 2.36, "learning_rate": 6.920608303457845e-05, "loss": 1.92, "step": 27220 }, { "epoch": 2.36, "learning_rate": 6.920391465248062e-05, "loss": 1.948, "step": 27230 }, { "epoch": 2.36, "learning_rate": 6.920174627038279e-05, "loss": 1.9938, "step": 27240 }, { "epoch": 2.36, "learning_rate": 6.919957788828495e-05, "loss": 2.0723, "step": 27250 }, { "epoch": 2.36, "learning_rate": 6.91974095061871e-05, "loss": 2.1848, "step": 27260 }, { "epoch": 2.36, "learning_rate": 6.919524112408927e-05, "loss": 2.0351, "step": 27270 }, { "epoch": 2.36, "learning_rate": 6.919307274199144e-05, "loss": 1.9736, "step": 27280 }, { "epoch": 2.36, "learning_rate": 6.919090435989359e-05, "loss": 1.8604, "step": 27290 }, { "epoch": 2.36, "learning_rate": 6.918873597779576e-05, "loss": 2.1415, "step": 27300 }, { "epoch": 2.37, "learning_rate": 6.918656759569793e-05, "loss": 2.0682, "step": 27310 }, { "epoch": 2.37, "learning_rate": 6.918439921360009e-05, "loss": 1.785, "step": 27320 }, { "epoch": 2.37, "learning_rate": 6.918223083150226e-05, "loss": 1.8205, "step": 27330 }, { "epoch": 2.37, "learning_rate": 6.918006244940441e-05, "loss": 2.045, "step": 27340 }, { "epoch": 2.37, "learning_rate": 6.917789406730658e-05, "loss": 2.01, "step": 27350 }, { "epoch": 2.37, "learning_rate": 6.917572568520873e-05, "loss": 1.8458, "step": 27360 }, { "epoch": 2.37, "learning_rate": 6.91735573031109e-05, "loss": 2.0029, "step": 27370 }, { "epoch": 2.37, "learning_rate": 6.917138892101306e-05, "loss": 2.1653, "step": 27380 }, { "epoch": 2.37, "learning_rate": 6.916922053891523e-05, "loss": 2.3084, "step": 27390 }, { "epoch": 2.37, "learning_rate": 6.91670521568174e-05, "loss": 2.0526, "step": 27400 }, { "epoch": 2.37, "learning_rate": 6.916488377471955e-05, "loss": 1.8844, "step": 27410 }, { "epoch": 2.37, "learning_rate": 6.916271539262171e-05, "loss": 2.1629, "step": 27420 }, { "epoch": 2.38, "learning_rate": 6.916054701052387e-05, "loss": 2.1008, "step": 27430 }, { "epoch": 2.38, "learning_rate": 6.915837862842603e-05, "loss": 2.0379, "step": 27440 }, { "epoch": 2.38, "learning_rate": 6.91562102463282e-05, "loss": 2.1109, "step": 27450 }, { "epoch": 2.38, "learning_rate": 6.915404186423037e-05, "loss": 1.8547, "step": 27460 }, { "epoch": 2.38, "learning_rate": 6.915187348213253e-05, "loss": 2.2153, "step": 27470 }, { "epoch": 2.38, "learning_rate": 6.914970510003469e-05, "loss": 1.9822, "step": 27480 }, { "epoch": 2.38, "learning_rate": 6.914753671793685e-05, "loss": 1.7484, "step": 27490 }, { "epoch": 2.38, "learning_rate": 6.9145368335839e-05, "loss": 1.8616, "step": 27500 }, { "epoch": 2.38, "learning_rate": 6.914319995374117e-05, "loss": 2.0225, "step": 27510 }, { "epoch": 2.38, "learning_rate": 6.914103157164334e-05, "loss": 1.9518, "step": 27520 }, { "epoch": 2.38, "learning_rate": 6.91388631895455e-05, "loss": 2.032, "step": 27530 }, { "epoch": 2.39, "learning_rate": 6.913669480744767e-05, "loss": 1.9977, "step": 27540 }, { "epoch": 2.39, "learning_rate": 6.913452642534982e-05, "loss": 2.0742, "step": 27550 }, { "epoch": 2.39, "learning_rate": 6.913235804325199e-05, "loss": 1.9535, "step": 27560 }, { "epoch": 2.39, "learning_rate": 6.913018966115414e-05, "loss": 1.9688, "step": 27570 }, { "epoch": 2.39, "learning_rate": 6.912802127905631e-05, "loss": 1.987, "step": 27580 }, { "epoch": 2.39, "learning_rate": 6.912585289695848e-05, "loss": 2.0391, "step": 27590 }, { "epoch": 2.39, "learning_rate": 6.912368451486064e-05, "loss": 1.9638, "step": 27600 }, { "epoch": 2.39, "learning_rate": 6.912151613276281e-05, "loss": 2.1118, "step": 27610 }, { "epoch": 2.39, "learning_rate": 6.911934775066496e-05, "loss": 2.0121, "step": 27620 }, { "epoch": 2.39, "learning_rate": 6.911717936856713e-05, "loss": 2.1158, "step": 27630 }, { "epoch": 2.39, "learning_rate": 6.91150109864693e-05, "loss": 1.9321, "step": 27640 }, { "epoch": 2.39, "learning_rate": 6.911284260437145e-05, "loss": 1.9929, "step": 27650 }, { "epoch": 2.4, "learning_rate": 6.911067422227361e-05, "loss": 1.9131, "step": 27660 }, { "epoch": 2.4, "learning_rate": 6.910850584017578e-05, "loss": 2.1521, "step": 27670 }, { "epoch": 2.4, "learning_rate": 6.910633745807795e-05, "loss": 2.0038, "step": 27680 }, { "epoch": 2.4, "learning_rate": 6.91041690759801e-05, "loss": 1.8073, "step": 27690 }, { "epoch": 2.4, "learning_rate": 6.910200069388227e-05, "loss": 1.8971, "step": 27700 }, { "epoch": 2.4, "learning_rate": 6.909983231178443e-05, "loss": 1.9278, "step": 27710 }, { "epoch": 2.4, "learning_rate": 6.909766392968658e-05, "loss": 1.9574, "step": 27720 }, { "epoch": 2.4, "learning_rate": 6.909549554758875e-05, "loss": 1.9235, "step": 27730 }, { "epoch": 2.4, "learning_rate": 6.909332716549092e-05, "loss": 1.9375, "step": 27740 }, { "epoch": 2.4, "learning_rate": 6.909115878339308e-05, "loss": 1.8543, "step": 27750 }, { "epoch": 2.4, "learning_rate": 6.908899040129524e-05, "loss": 1.9574, "step": 27760 }, { "epoch": 2.41, "learning_rate": 6.90868220191974e-05, "loss": 1.9983, "step": 27770 }, { "epoch": 2.41, "learning_rate": 6.908465363709957e-05, "loss": 1.9619, "step": 27780 }, { "epoch": 2.41, "learning_rate": 6.908248525500174e-05, "loss": 2.0913, "step": 27790 }, { "epoch": 2.41, "learning_rate": 6.908031687290389e-05, "loss": 1.7769, "step": 27800 }, { "epoch": 2.41, "learning_rate": 6.907814849080605e-05, "loss": 2.045, "step": 27810 }, { "epoch": 2.41, "learning_rate": 6.907598010870822e-05, "loss": 2.0908, "step": 27820 }, { "epoch": 2.41, "learning_rate": 6.907381172661037e-05, "loss": 2.2151, "step": 27830 }, { "epoch": 2.41, "learning_rate": 6.907164334451254e-05, "loss": 1.7742, "step": 27840 }, { "epoch": 2.41, "learning_rate": 6.906947496241471e-05, "loss": 1.8668, "step": 27850 }, { "epoch": 2.41, "learning_rate": 6.906730658031687e-05, "loss": 2.0942, "step": 27860 }, { "epoch": 2.41, "learning_rate": 6.906513819821904e-05, "loss": 2.2434, "step": 27870 }, { "epoch": 2.41, "learning_rate": 6.906296981612119e-05, "loss": 2.0994, "step": 27880 }, { "epoch": 2.42, "learning_rate": 6.906080143402336e-05, "loss": 2.007, "step": 27890 }, { "epoch": 2.42, "learning_rate": 6.905863305192551e-05, "loss": 1.939, "step": 27900 }, { "epoch": 2.42, "learning_rate": 6.905646466982768e-05, "loss": 1.8224, "step": 27910 }, { "epoch": 2.42, "learning_rate": 6.905429628772984e-05, "loss": 2.1552, "step": 27920 }, { "epoch": 2.42, "learning_rate": 6.905212790563201e-05, "loss": 2.0323, "step": 27930 }, { "epoch": 2.42, "learning_rate": 6.904995952353418e-05, "loss": 2.0184, "step": 27940 }, { "epoch": 2.42, "learning_rate": 6.904779114143633e-05, "loss": 1.9346, "step": 27950 }, { "epoch": 2.42, "learning_rate": 6.904562275933848e-05, "loss": 1.8854, "step": 27960 }, { "epoch": 2.42, "learning_rate": 6.904345437724065e-05, "loss": 1.7993, "step": 27970 }, { "epoch": 2.42, "learning_rate": 6.904128599514282e-05, "loss": 1.9639, "step": 27980 }, { "epoch": 2.42, "learning_rate": 6.903911761304498e-05, "loss": 1.753, "step": 27990 }, { "epoch": 2.42, "learning_rate": 6.903694923094715e-05, "loss": 1.8513, "step": 28000 }, { "epoch": 2.43, "learning_rate": 6.90347808488493e-05, "loss": 1.8719, "step": 28010 }, { "epoch": 2.43, "learning_rate": 6.903261246675147e-05, "loss": 1.9998, "step": 28020 }, { "epoch": 2.43, "learning_rate": 6.903044408465362e-05, "loss": 1.9803, "step": 28030 }, { "epoch": 2.43, "learning_rate": 6.902827570255579e-05, "loss": 2.0183, "step": 28040 }, { "epoch": 2.43, "learning_rate": 6.902610732045795e-05, "loss": 2.0263, "step": 28050 }, { "epoch": 2.43, "learning_rate": 6.902393893836012e-05, "loss": 2.0823, "step": 28060 }, { "epoch": 2.43, "learning_rate": 6.902177055626229e-05, "loss": 1.88, "step": 28070 }, { "epoch": 2.43, "learning_rate": 6.901960217416444e-05, "loss": 1.9083, "step": 28080 }, { "epoch": 2.43, "learning_rate": 6.90174337920666e-05, "loss": 2.1419, "step": 28090 }, { "epoch": 2.43, "learning_rate": 6.901526540996877e-05, "loss": 1.9806, "step": 28100 }, { "epoch": 2.43, "learning_rate": 6.901309702787092e-05, "loss": 2.0389, "step": 28110 }, { "epoch": 2.44, "learning_rate": 6.901092864577309e-05, "loss": 1.9973, "step": 28120 }, { "epoch": 2.44, "learning_rate": 6.900876026367526e-05, "loss": 1.9742, "step": 28130 }, { "epoch": 2.44, "learning_rate": 6.900659188157742e-05, "loss": 1.9025, "step": 28140 }, { "epoch": 2.44, "learning_rate": 6.900442349947958e-05, "loss": 1.936, "step": 28150 }, { "epoch": 2.44, "learning_rate": 6.900225511738174e-05, "loss": 1.8576, "step": 28160 }, { "epoch": 2.44, "learning_rate": 6.900008673528391e-05, "loss": 2.0468, "step": 28170 }, { "epoch": 2.44, "learning_rate": 6.899791835318608e-05, "loss": 2.162, "step": 28180 }, { "epoch": 2.44, "learning_rate": 6.899574997108823e-05, "loss": 1.9291, "step": 28190 }, { "epoch": 2.44, "learning_rate": 6.89935815889904e-05, "loss": 2.0805, "step": 28200 }, { "epoch": 2.44, "learning_rate": 6.899141320689256e-05, "loss": 1.9473, "step": 28210 }, { "epoch": 2.44, "learning_rate": 6.898924482479471e-05, "loss": 1.9709, "step": 28220 }, { "epoch": 2.44, "learning_rate": 6.898707644269688e-05, "loss": 1.6925, "step": 28230 }, { "epoch": 2.45, "learning_rate": 6.898490806059905e-05, "loss": 2.1234, "step": 28240 }, { "epoch": 2.45, "learning_rate": 6.898273967850121e-05, "loss": 2.0391, "step": 28250 }, { "epoch": 2.45, "learning_rate": 6.898057129640337e-05, "loss": 1.96, "step": 28260 }, { "epoch": 2.45, "learning_rate": 6.897840291430553e-05, "loss": 2.0878, "step": 28270 }, { "epoch": 2.45, "learning_rate": 6.89762345322077e-05, "loss": 1.9334, "step": 28280 }, { "epoch": 2.45, "learning_rate": 6.897406615010985e-05, "loss": 1.9589, "step": 28290 }, { "epoch": 2.45, "learning_rate": 6.897189776801202e-05, "loss": 1.8289, "step": 28300 }, { "epoch": 2.45, "learning_rate": 6.896972938591418e-05, "loss": 1.9185, "step": 28310 }, { "epoch": 2.45, "learning_rate": 6.896756100381635e-05, "loss": 1.9546, "step": 28320 }, { "epoch": 2.45, "learning_rate": 6.896539262171852e-05, "loss": 1.9008, "step": 28330 }, { "epoch": 2.45, "learning_rate": 6.896322423962067e-05, "loss": 2.1744, "step": 28340 }, { "epoch": 2.46, "learning_rate": 6.896105585752284e-05, "loss": 1.9802, "step": 28350 }, { "epoch": 2.46, "learning_rate": 6.895888747542499e-05, "loss": 1.85, "step": 28360 }, { "epoch": 2.46, "learning_rate": 6.895671909332716e-05, "loss": 1.9838, "step": 28370 }, { "epoch": 2.46, "learning_rate": 6.895455071122932e-05, "loss": 2.0541, "step": 28380 }, { "epoch": 2.46, "learning_rate": 6.895238232913149e-05, "loss": 1.9005, "step": 28390 }, { "epoch": 2.46, "learning_rate": 6.895021394703366e-05, "loss": 2.206, "step": 28400 }, { "epoch": 2.46, "learning_rate": 6.894804556493581e-05, "loss": 2.0207, "step": 28410 }, { "epoch": 2.46, "learning_rate": 6.894587718283797e-05, "loss": 2.0309, "step": 28420 }, { "epoch": 2.46, "learning_rate": 6.894370880074013e-05, "loss": 1.8848, "step": 28430 }, { "epoch": 2.46, "learning_rate": 6.89415404186423e-05, "loss": 1.9723, "step": 28440 }, { "epoch": 2.46, "learning_rate": 6.893937203654446e-05, "loss": 1.8447, "step": 28450 }, { "epoch": 2.46, "learning_rate": 6.893720365444663e-05, "loss": 2.0803, "step": 28460 }, { "epoch": 2.47, "learning_rate": 6.893503527234879e-05, "loss": 1.8218, "step": 28470 }, { "epoch": 2.47, "learning_rate": 6.893286689025095e-05, "loss": 2.211, "step": 28480 }, { "epoch": 2.47, "learning_rate": 6.893069850815311e-05, "loss": 2.145, "step": 28490 }, { "epoch": 2.47, "learning_rate": 6.892853012605526e-05, "loss": 2.0258, "step": 28500 }, { "epoch": 2.47, "learning_rate": 6.892636174395743e-05, "loss": 2.0029, "step": 28510 }, { "epoch": 2.47, "learning_rate": 6.89241933618596e-05, "loss": 1.9221, "step": 28520 }, { "epoch": 2.47, "learning_rate": 6.892202497976176e-05, "loss": 1.7664, "step": 28530 }, { "epoch": 2.47, "learning_rate": 6.891985659766393e-05, "loss": 1.9986, "step": 28540 }, { "epoch": 2.47, "learning_rate": 6.891768821556608e-05, "loss": 1.9598, "step": 28550 }, { "epoch": 2.47, "learning_rate": 6.891551983346825e-05, "loss": 1.9129, "step": 28560 }, { "epoch": 2.47, "learning_rate": 6.89133514513704e-05, "loss": 1.9789, "step": 28570 }, { "epoch": 2.48, "learning_rate": 6.891118306927257e-05, "loss": 1.8304, "step": 28580 }, { "epoch": 2.48, "learning_rate": 6.890901468717474e-05, "loss": 2.0551, "step": 28590 }, { "epoch": 2.48, "learning_rate": 6.89068463050769e-05, "loss": 1.9406, "step": 28600 }, { "epoch": 2.48, "learning_rate": 6.890467792297907e-05, "loss": 2.0266, "step": 28610 }, { "epoch": 2.48, "learning_rate": 6.890250954088122e-05, "loss": 1.9586, "step": 28620 }, { "epoch": 2.48, "learning_rate": 6.890034115878339e-05, "loss": 1.8906, "step": 28630 }, { "epoch": 2.48, "learning_rate": 6.889817277668555e-05, "loss": 1.9021, "step": 28640 }, { "epoch": 2.48, "learning_rate": 6.88960043945877e-05, "loss": 1.9525, "step": 28650 }, { "epoch": 2.48, "learning_rate": 6.889383601248987e-05, "loss": 2.1003, "step": 28660 }, { "epoch": 2.48, "learning_rate": 6.889166763039204e-05, "loss": 1.8689, "step": 28670 }, { "epoch": 2.48, "learning_rate": 6.88894992482942e-05, "loss": 2.0301, "step": 28680 }, { "epoch": 2.48, "learning_rate": 6.888733086619636e-05, "loss": 2.0268, "step": 28690 }, { "epoch": 2.49, "learning_rate": 6.888516248409852e-05, "loss": 1.8978, "step": 28700 }, { "epoch": 2.49, "learning_rate": 6.888299410200069e-05, "loss": 1.9259, "step": 28710 }, { "epoch": 2.49, "learning_rate": 6.888082571990286e-05, "loss": 1.9471, "step": 28720 }, { "epoch": 2.49, "learning_rate": 6.887865733780501e-05, "loss": 2.0331, "step": 28730 }, { "epoch": 2.49, "learning_rate": 6.887648895570718e-05, "loss": 1.9313, "step": 28740 }, { "epoch": 2.49, "learning_rate": 6.887432057360934e-05, "loss": 2.0615, "step": 28750 }, { "epoch": 2.49, "learning_rate": 6.88721521915115e-05, "loss": 1.9468, "step": 28760 }, { "epoch": 2.49, "learning_rate": 6.886998380941366e-05, "loss": 1.9614, "step": 28770 }, { "epoch": 2.49, "learning_rate": 6.886781542731583e-05, "loss": 1.9025, "step": 28780 }, { "epoch": 2.49, "learning_rate": 6.8865647045218e-05, "loss": 2.1098, "step": 28790 }, { "epoch": 2.49, "learning_rate": 6.886347866312015e-05, "loss": 1.9962, "step": 28800 }, { "epoch": 2.5, "learning_rate": 6.886131028102231e-05, "loss": 1.9804, "step": 28810 }, { "epoch": 2.5, "learning_rate": 6.885914189892448e-05, "loss": 2.1086, "step": 28820 }, { "epoch": 2.5, "learning_rate": 6.885697351682663e-05, "loss": 1.8968, "step": 28830 }, { "epoch": 2.5, "learning_rate": 6.88548051347288e-05, "loss": 1.7924, "step": 28840 }, { "epoch": 2.5, "learning_rate": 6.885263675263097e-05, "loss": 1.9004, "step": 28850 }, { "epoch": 2.5, "learning_rate": 6.885046837053313e-05, "loss": 2.1241, "step": 28860 }, { "epoch": 2.5, "learning_rate": 6.88482999884353e-05, "loss": 1.9563, "step": 28870 }, { "epoch": 2.5, "learning_rate": 6.884613160633745e-05, "loss": 1.9563, "step": 28880 }, { "epoch": 2.5, "learning_rate": 6.884396322423962e-05, "loss": 2.0361, "step": 28890 }, { "epoch": 2.5, "learning_rate": 6.884179484214177e-05, "loss": 2.0698, "step": 28900 }, { "epoch": 2.5, "learning_rate": 6.883962646004394e-05, "loss": 1.9753, "step": 28910 }, { "epoch": 2.5, "learning_rate": 6.88374580779461e-05, "loss": 2.0199, "step": 28920 }, { "epoch": 2.51, "learning_rate": 6.883528969584827e-05, "loss": 2.0229, "step": 28930 }, { "epoch": 2.51, "learning_rate": 6.883312131375044e-05, "loss": 1.9412, "step": 28940 }, { "epoch": 2.51, "learning_rate": 6.883095293165259e-05, "loss": 2.0556, "step": 28950 }, { "epoch": 2.51, "learning_rate": 6.882878454955476e-05, "loss": 1.9612, "step": 28960 }, { "epoch": 2.51, "learning_rate": 6.882661616745691e-05, "loss": 2.0072, "step": 28970 }, { "epoch": 2.51, "learning_rate": 6.882444778535908e-05, "loss": 1.9423, "step": 28980 }, { "epoch": 2.51, "learning_rate": 6.882227940326124e-05, "loss": 2.0192, "step": 28990 }, { "epoch": 2.51, "learning_rate": 6.882011102116341e-05, "loss": 1.9014, "step": 29000 }, { "epoch": 2.51, "learning_rate": 6.881794263906557e-05, "loss": 2.0987, "step": 29010 }, { "epoch": 2.51, "learning_rate": 6.881577425696773e-05, "loss": 2.0217, "step": 29020 }, { "epoch": 2.51, "learning_rate": 6.881360587486988e-05, "loss": 2.0539, "step": 29030 }, { "epoch": 2.52, "learning_rate": 6.881143749277205e-05, "loss": 2.0032, "step": 29040 }, { "epoch": 2.52, "learning_rate": 6.880926911067421e-05, "loss": 1.9798, "step": 29050 }, { "epoch": 2.52, "learning_rate": 6.880710072857638e-05, "loss": 2.1006, "step": 29060 }, { "epoch": 2.52, "learning_rate": 6.880493234647855e-05, "loss": 1.884, "step": 29070 }, { "epoch": 2.52, "learning_rate": 6.88027639643807e-05, "loss": 2.052, "step": 29080 }, { "epoch": 2.52, "learning_rate": 6.880059558228287e-05, "loss": 2.088, "step": 29090 }, { "epoch": 2.52, "learning_rate": 6.879842720018503e-05, "loss": 1.9624, "step": 29100 }, { "epoch": 2.52, "learning_rate": 6.879625881808718e-05, "loss": 1.7482, "step": 29110 }, { "epoch": 2.52, "learning_rate": 6.879409043598935e-05, "loss": 1.9875, "step": 29120 }, { "epoch": 2.52, "learning_rate": 6.879192205389152e-05, "loss": 2.081, "step": 29130 }, { "epoch": 2.52, "learning_rate": 6.878975367179368e-05, "loss": 1.7667, "step": 29140 }, { "epoch": 2.52, "learning_rate": 6.878758528969584e-05, "loss": 1.7837, "step": 29150 }, { "epoch": 2.53, "learning_rate": 6.8785416907598e-05, "loss": 1.8137, "step": 29160 }, { "epoch": 2.53, "learning_rate": 6.878324852550017e-05, "loss": 2.038, "step": 29170 }, { "epoch": 2.53, "learning_rate": 6.878108014340234e-05, "loss": 2.0214, "step": 29180 }, { "epoch": 2.53, "learning_rate": 6.877891176130449e-05, "loss": 2.0296, "step": 29190 }, { "epoch": 2.53, "learning_rate": 6.877674337920665e-05, "loss": 1.8481, "step": 29200 }, { "epoch": 2.53, "learning_rate": 6.877457499710882e-05, "loss": 1.8122, "step": 29210 }, { "epoch": 2.53, "learning_rate": 6.877240661501097e-05, "loss": 1.9891, "step": 29220 }, { "epoch": 2.53, "learning_rate": 6.877023823291314e-05, "loss": 1.7269, "step": 29230 }, { "epoch": 2.53, "learning_rate": 6.876806985081531e-05, "loss": 1.8319, "step": 29240 }, { "epoch": 2.53, "learning_rate": 6.876590146871747e-05, "loss": 2.0688, "step": 29250 }, { "epoch": 2.53, "learning_rate": 6.876373308661963e-05, "loss": 2.0016, "step": 29260 }, { "epoch": 2.53, "learning_rate": 6.876156470452179e-05, "loss": 2.0904, "step": 29270 }, { "epoch": 2.54, "learning_rate": 6.875939632242396e-05, "loss": 1.9118, "step": 29280 }, { "epoch": 2.54, "learning_rate": 6.875722794032611e-05, "loss": 1.8853, "step": 29290 }, { "epoch": 2.54, "learning_rate": 6.875505955822828e-05, "loss": 2.0153, "step": 29300 }, { "epoch": 2.54, "learning_rate": 6.875289117613044e-05, "loss": 1.8283, "step": 29310 }, { "epoch": 2.54, "learning_rate": 6.875072279403261e-05, "loss": 1.9355, "step": 29320 }, { "epoch": 2.54, "learning_rate": 6.874855441193478e-05, "loss": 1.7888, "step": 29330 }, { "epoch": 2.54, "learning_rate": 6.874638602983693e-05, "loss": 1.823, "step": 29340 }, { "epoch": 2.54, "learning_rate": 6.87442176477391e-05, "loss": 1.8391, "step": 29350 }, { "epoch": 2.54, "learning_rate": 6.874204926564125e-05, "loss": 1.943, "step": 29360 }, { "epoch": 2.54, "learning_rate": 6.873988088354342e-05, "loss": 2.0698, "step": 29370 }, { "epoch": 2.54, "learning_rate": 6.873771250144558e-05, "loss": 2.0718, "step": 29380 }, { "epoch": 2.55, "learning_rate": 6.873554411934775e-05, "loss": 1.8645, "step": 29390 }, { "epoch": 2.55, "learning_rate": 6.873337573724991e-05, "loss": 2.0576, "step": 29400 }, { "epoch": 2.55, "learning_rate": 6.873120735515207e-05, "loss": 2.102, "step": 29410 }, { "epoch": 2.55, "learning_rate": 6.872903897305423e-05, "loss": 1.7408, "step": 29420 }, { "epoch": 2.55, "learning_rate": 6.872687059095639e-05, "loss": 2.2499, "step": 29430 }, { "epoch": 2.55, "learning_rate": 6.872470220885855e-05, "loss": 1.9876, "step": 29440 }, { "epoch": 2.55, "learning_rate": 6.872253382676072e-05, "loss": 1.9416, "step": 29450 }, { "epoch": 2.55, "learning_rate": 6.872036544466289e-05, "loss": 1.9385, "step": 29460 }, { "epoch": 2.55, "learning_rate": 6.871819706256505e-05, "loss": 1.9155, "step": 29470 }, { "epoch": 2.55, "learning_rate": 6.87160286804672e-05, "loss": 2.0727, "step": 29480 }, { "epoch": 2.55, "learning_rate": 6.871386029836937e-05, "loss": 1.932, "step": 29490 }, { "epoch": 2.55, "learning_rate": 6.871169191627152e-05, "loss": 1.753, "step": 29500 }, { "epoch": 2.56, "learning_rate": 6.870952353417369e-05, "loss": 1.962, "step": 29510 }, { "epoch": 2.56, "learning_rate": 6.870735515207586e-05, "loss": 1.9761, "step": 29520 }, { "epoch": 2.56, "learning_rate": 6.870518676997802e-05, "loss": 1.6535, "step": 29530 }, { "epoch": 2.56, "learning_rate": 6.870301838788019e-05, "loss": 1.9975, "step": 29540 }, { "epoch": 2.56, "learning_rate": 6.870085000578234e-05, "loss": 2.3183, "step": 29550 }, { "epoch": 2.56, "learning_rate": 6.869868162368451e-05, "loss": 1.981, "step": 29560 }, { "epoch": 2.56, "learning_rate": 6.869651324158666e-05, "loss": 1.7663, "step": 29570 }, { "epoch": 2.56, "learning_rate": 6.869434485948883e-05, "loss": 1.8371, "step": 29580 }, { "epoch": 2.56, "learning_rate": 6.8692176477391e-05, "loss": 2.1452, "step": 29590 }, { "epoch": 2.56, "learning_rate": 6.869000809529316e-05, "loss": 2.1517, "step": 29600 }, { "epoch": 2.56, "learning_rate": 6.868783971319533e-05, "loss": 1.9187, "step": 29610 }, { "epoch": 2.57, "learning_rate": 6.868567133109748e-05, "loss": 1.819, "step": 29620 }, { "epoch": 2.57, "learning_rate": 6.868350294899965e-05, "loss": 2.1017, "step": 29630 }, { "epoch": 2.57, "learning_rate": 6.868133456690181e-05, "loss": 2.0764, "step": 29640 }, { "epoch": 2.57, "learning_rate": 6.867916618480397e-05, "loss": 1.9813, "step": 29650 }, { "epoch": 2.57, "learning_rate": 6.867699780270613e-05, "loss": 2.0515, "step": 29660 }, { "epoch": 2.57, "learning_rate": 6.86748294206083e-05, "loss": 1.8886, "step": 29670 }, { "epoch": 2.57, "learning_rate": 6.867266103851047e-05, "loss": 2.104, "step": 29680 }, { "epoch": 2.57, "learning_rate": 6.867049265641262e-05, "loss": 1.8122, "step": 29690 }, { "epoch": 2.57, "learning_rate": 6.866832427431478e-05, "loss": 1.9628, "step": 29700 }, { "epoch": 2.57, "learning_rate": 6.866615589221695e-05, "loss": 1.994, "step": 29710 }, { "epoch": 2.57, "learning_rate": 6.866398751011912e-05, "loss": 2.167, "step": 29720 }, { "epoch": 2.57, "learning_rate": 6.866181912802127e-05, "loss": 1.8698, "step": 29730 }, { "epoch": 2.58, "learning_rate": 6.865965074592344e-05, "loss": 2.0457, "step": 29740 }, { "epoch": 2.58, "learning_rate": 6.86574823638256e-05, "loss": 1.9572, "step": 29750 }, { "epoch": 2.58, "learning_rate": 6.865531398172776e-05, "loss": 2.0685, "step": 29760 }, { "epoch": 2.58, "learning_rate": 6.865314559962992e-05, "loss": 2.0331, "step": 29770 }, { "epoch": 2.58, "learning_rate": 6.865097721753209e-05, "loss": 1.9169, "step": 29780 }, { "epoch": 2.58, "learning_rate": 6.864880883543426e-05, "loss": 1.7935, "step": 29790 }, { "epoch": 2.58, "learning_rate": 6.864664045333641e-05, "loss": 1.8626, "step": 29800 }, { "epoch": 2.58, "learning_rate": 6.864447207123857e-05, "loss": 2.2088, "step": 29810 }, { "epoch": 2.58, "learning_rate": 6.864230368914074e-05, "loss": 1.9675, "step": 29820 }, { "epoch": 2.58, "learning_rate": 6.86401353070429e-05, "loss": 2.1262, "step": 29830 }, { "epoch": 2.58, "learning_rate": 6.863796692494506e-05, "loss": 2.0855, "step": 29840 }, { "epoch": 2.59, "learning_rate": 6.863579854284723e-05, "loss": 2.0056, "step": 29850 }, { "epoch": 2.59, "learning_rate": 6.863363016074939e-05, "loss": 2.1163, "step": 29860 }, { "epoch": 2.59, "learning_rate": 6.863146177865156e-05, "loss": 1.9197, "step": 29870 }, { "epoch": 2.59, "learning_rate": 6.862929339655371e-05, "loss": 1.8993, "step": 29880 }, { "epoch": 2.59, "learning_rate": 6.862712501445588e-05, "loss": 1.9467, "step": 29890 }, { "epoch": 2.59, "learning_rate": 6.862495663235803e-05, "loss": 2.0757, "step": 29900 }, { "epoch": 2.59, "learning_rate": 6.86227882502602e-05, "loss": 1.9972, "step": 29910 }, { "epoch": 2.59, "learning_rate": 6.862061986816236e-05, "loss": 1.9491, "step": 29920 }, { "epoch": 2.59, "learning_rate": 6.861845148606453e-05, "loss": 1.9975, "step": 29930 }, { "epoch": 2.59, "learning_rate": 6.86162831039667e-05, "loss": 1.8504, "step": 29940 }, { "epoch": 2.59, "learning_rate": 6.861411472186885e-05, "loss": 2.2035, "step": 29950 }, { "epoch": 2.59, "learning_rate": 6.861194633977102e-05, "loss": 2.0512, "step": 29960 }, { "epoch": 2.6, "learning_rate": 6.860977795767317e-05, "loss": 1.8971, "step": 29970 }, { "epoch": 2.6, "learning_rate": 6.860760957557534e-05, "loss": 1.9691, "step": 29980 }, { "epoch": 2.6, "learning_rate": 6.86054411934775e-05, "loss": 1.9666, "step": 29990 }, { "epoch": 2.6, "learning_rate": 6.860327281137967e-05, "loss": 2.0008, "step": 30000 }, { "epoch": 2.6, "learning_rate": 6.860110442928183e-05, "loss": 2.0593, "step": 30010 }, { "epoch": 2.6, "learning_rate": 6.859893604718399e-05, "loss": 1.6908, "step": 30020 }, { "epoch": 2.6, "learning_rate": 6.859676766508615e-05, "loss": 1.7859, "step": 30030 }, { "epoch": 2.6, "learning_rate": 6.85945992829883e-05, "loss": 2.0729, "step": 30040 }, { "epoch": 2.6, "learning_rate": 6.859243090089047e-05, "loss": 2.1282, "step": 30050 }, { "epoch": 2.6, "learning_rate": 6.859026251879264e-05, "loss": 1.9467, "step": 30060 }, { "epoch": 2.6, "learning_rate": 6.85880941366948e-05, "loss": 1.9001, "step": 30070 }, { "epoch": 2.61, "learning_rate": 6.858592575459697e-05, "loss": 1.9258, "step": 30080 }, { "epoch": 2.61, "learning_rate": 6.858375737249912e-05, "loss": 1.9746, "step": 30090 }, { "epoch": 2.61, "learning_rate": 6.858158899040129e-05, "loss": 1.8409, "step": 30100 }, { "epoch": 2.61, "learning_rate": 6.857942060830344e-05, "loss": 2.1085, "step": 30110 }, { "epoch": 2.61, "learning_rate": 6.857725222620561e-05, "loss": 1.9262, "step": 30120 }, { "epoch": 2.61, "learning_rate": 6.857508384410778e-05, "loss": 1.8994, "step": 30130 }, { "epoch": 2.61, "learning_rate": 6.857291546200994e-05, "loss": 1.7505, "step": 30140 }, { "epoch": 2.61, "learning_rate": 6.85707470799121e-05, "loss": 1.7446, "step": 30150 }, { "epoch": 2.61, "learning_rate": 6.856857869781426e-05, "loss": 1.8325, "step": 30160 }, { "epoch": 2.61, "learning_rate": 6.856641031571643e-05, "loss": 1.9108, "step": 30170 }, { "epoch": 2.61, "learning_rate": 6.85642419336186e-05, "loss": 2.139, "step": 30180 }, { "epoch": 2.61, "learning_rate": 6.856207355152075e-05, "loss": 1.7832, "step": 30190 }, { "epoch": 2.62, "learning_rate": 6.855990516942291e-05, "loss": 1.9206, "step": 30200 }, { "epoch": 2.62, "learning_rate": 6.855773678732508e-05, "loss": 1.666, "step": 30210 }, { "epoch": 2.62, "learning_rate": 6.855556840522723e-05, "loss": 2.031, "step": 30220 }, { "epoch": 2.62, "learning_rate": 6.85534000231294e-05, "loss": 1.9936, "step": 30230 }, { "epoch": 2.62, "learning_rate": 6.855123164103157e-05, "loss": 2.0215, "step": 30240 }, { "epoch": 2.62, "learning_rate": 6.854906325893373e-05, "loss": 2.0343, "step": 30250 }, { "epoch": 2.62, "learning_rate": 6.85468948768359e-05, "loss": 1.896, "step": 30260 }, { "epoch": 2.62, "learning_rate": 6.854472649473805e-05, "loss": 1.9237, "step": 30270 }, { "epoch": 2.62, "learning_rate": 6.854255811264022e-05, "loss": 2.1365, "step": 30280 }, { "epoch": 2.62, "learning_rate": 6.854038973054237e-05, "loss": 1.9447, "step": 30290 }, { "epoch": 2.62, "learning_rate": 6.853822134844454e-05, "loss": 2.0236, "step": 30300 }, { "epoch": 2.63, "learning_rate": 6.85360529663467e-05, "loss": 1.7691, "step": 30310 }, { "epoch": 2.63, "learning_rate": 6.853388458424887e-05, "loss": 1.9355, "step": 30320 }, { "epoch": 2.63, "learning_rate": 6.853171620215104e-05, "loss": 1.9325, "step": 30330 }, { "epoch": 2.63, "learning_rate": 6.852954782005319e-05, "loss": 1.8347, "step": 30340 }, { "epoch": 2.63, "learning_rate": 6.852737943795536e-05, "loss": 1.9668, "step": 30350 }, { "epoch": 2.63, "learning_rate": 6.852521105585751e-05, "loss": 2.2739, "step": 30360 }, { "epoch": 2.63, "learning_rate": 6.852304267375968e-05, "loss": 2.0548, "step": 30370 }, { "epoch": 2.63, "learning_rate": 6.852087429166184e-05, "loss": 2.1149, "step": 30380 }, { "epoch": 2.63, "learning_rate": 6.851870590956401e-05, "loss": 1.9409, "step": 30390 }, { "epoch": 2.63, "learning_rate": 6.851653752746617e-05, "loss": 1.9062, "step": 30400 }, { "epoch": 2.63, "learning_rate": 6.851436914536833e-05, "loss": 1.8364, "step": 30410 }, { "epoch": 2.63, "learning_rate": 6.85122007632705e-05, "loss": 1.8759, "step": 30420 }, { "epoch": 2.64, "learning_rate": 6.851003238117265e-05, "loss": 1.8207, "step": 30430 }, { "epoch": 2.64, "learning_rate": 6.850786399907481e-05, "loss": 1.7398, "step": 30440 }, { "epoch": 2.64, "learning_rate": 6.850569561697698e-05, "loss": 2.0454, "step": 30450 }, { "epoch": 2.64, "learning_rate": 6.850352723487915e-05, "loss": 1.7674, "step": 30460 }, { "epoch": 2.64, "learning_rate": 6.850135885278131e-05, "loss": 2.0679, "step": 30470 }, { "epoch": 2.64, "learning_rate": 6.849919047068347e-05, "loss": 1.8768, "step": 30480 }, { "epoch": 2.64, "learning_rate": 6.849702208858563e-05, "loss": 1.9744, "step": 30490 }, { "epoch": 2.64, "learning_rate": 6.849485370648778e-05, "loss": 1.9298, "step": 30500 }, { "epoch": 2.64, "learning_rate": 6.849268532438995e-05, "loss": 1.8827, "step": 30510 }, { "epoch": 2.64, "learning_rate": 6.849051694229212e-05, "loss": 2.1178, "step": 30520 }, { "epoch": 2.64, "learning_rate": 6.848834856019428e-05, "loss": 1.8484, "step": 30530 }, { "epoch": 2.64, "learning_rate": 6.848618017809645e-05, "loss": 1.8606, "step": 30540 }, { "epoch": 2.65, "learning_rate": 6.84840117959986e-05, "loss": 2.0073, "step": 30550 }, { "epoch": 2.65, "learning_rate": 6.848184341390077e-05, "loss": 1.8874, "step": 30560 }, { "epoch": 2.65, "learning_rate": 6.847967503180292e-05, "loss": 1.9455, "step": 30570 }, { "epoch": 2.65, "learning_rate": 6.847750664970509e-05, "loss": 2.059, "step": 30580 }, { "epoch": 2.65, "learning_rate": 6.847533826760725e-05, "loss": 1.9547, "step": 30590 }, { "epoch": 2.65, "learning_rate": 6.847316988550942e-05, "loss": 1.9033, "step": 30600 }, { "epoch": 2.65, "learning_rate": 6.847100150341159e-05, "loss": 2.0762, "step": 30610 }, { "epoch": 2.65, "learning_rate": 6.846883312131374e-05, "loss": 1.9627, "step": 30620 }, { "epoch": 2.65, "learning_rate": 6.84666647392159e-05, "loss": 1.803, "step": 30630 }, { "epoch": 2.65, "learning_rate": 6.846449635711807e-05, "loss": 1.9525, "step": 30640 }, { "epoch": 2.65, "learning_rate": 6.846232797502023e-05, "loss": 1.9658, "step": 30650 }, { "epoch": 2.66, "learning_rate": 6.846015959292239e-05, "loss": 2.0638, "step": 30660 }, { "epoch": 2.66, "learning_rate": 6.845799121082456e-05, "loss": 1.8151, "step": 30670 }, { "epoch": 2.66, "learning_rate": 6.845582282872673e-05, "loss": 1.9936, "step": 30680 }, { "epoch": 2.66, "learning_rate": 6.845365444662888e-05, "loss": 1.9001, "step": 30690 }, { "epoch": 2.66, "learning_rate": 6.845148606453104e-05, "loss": 1.8621, "step": 30700 }, { "epoch": 2.66, "learning_rate": 6.844931768243321e-05, "loss": 2.1083, "step": 30710 }, { "epoch": 2.66, "learning_rate": 6.844714930033538e-05, "loss": 1.8891, "step": 30720 }, { "epoch": 2.66, "learning_rate": 6.844498091823753e-05, "loss": 1.9099, "step": 30730 }, { "epoch": 2.66, "learning_rate": 6.84428125361397e-05, "loss": 1.954, "step": 30740 }, { "epoch": 2.66, "learning_rate": 6.844064415404186e-05, "loss": 1.8623, "step": 30750 }, { "epoch": 2.66, "learning_rate": 6.843847577194402e-05, "loss": 1.914, "step": 30760 }, { "epoch": 2.66, "learning_rate": 6.843630738984618e-05, "loss": 2.1027, "step": 30770 }, { "epoch": 2.67, "learning_rate": 6.843413900774835e-05, "loss": 1.9291, "step": 30780 }, { "epoch": 2.67, "learning_rate": 6.843197062565051e-05, "loss": 1.6851, "step": 30790 }, { "epoch": 2.67, "learning_rate": 6.842980224355267e-05, "loss": 2.0549, "step": 30800 }, { "epoch": 2.67, "learning_rate": 6.842763386145483e-05, "loss": 2.0178, "step": 30810 }, { "epoch": 2.67, "learning_rate": 6.8425465479357e-05, "loss": 2.0866, "step": 30820 }, { "epoch": 2.67, "learning_rate": 6.842329709725915e-05, "loss": 1.703, "step": 30830 }, { "epoch": 2.67, "learning_rate": 6.842112871516132e-05, "loss": 1.9913, "step": 30840 }, { "epoch": 2.67, "learning_rate": 6.841896033306349e-05, "loss": 2.2272, "step": 30850 }, { "epoch": 2.67, "learning_rate": 6.841679195096565e-05, "loss": 2.1176, "step": 30860 }, { "epoch": 2.67, "learning_rate": 6.841462356886782e-05, "loss": 2.268, "step": 30870 }, { "epoch": 2.67, "learning_rate": 6.841245518676997e-05, "loss": 1.9966, "step": 30880 }, { "epoch": 2.68, "learning_rate": 6.841028680467214e-05, "loss": 1.8555, "step": 30890 }, { "epoch": 2.68, "learning_rate": 6.840811842257429e-05, "loss": 1.8004, "step": 30900 }, { "epoch": 2.68, "learning_rate": 6.840595004047646e-05, "loss": 2.04, "step": 30910 }, { "epoch": 2.68, "learning_rate": 6.840378165837862e-05, "loss": 2.1406, "step": 30920 }, { "epoch": 2.68, "learning_rate": 6.840161327628079e-05, "loss": 1.7419, "step": 30930 }, { "epoch": 2.68, "learning_rate": 6.839944489418296e-05, "loss": 1.9092, "step": 30940 }, { "epoch": 2.68, "learning_rate": 6.839727651208511e-05, "loss": 1.8978, "step": 30950 }, { "epoch": 2.68, "learning_rate": 6.839510812998728e-05, "loss": 1.9507, "step": 30960 }, { "epoch": 2.68, "learning_rate": 6.839293974788943e-05, "loss": 1.937, "step": 30970 }, { "epoch": 2.68, "learning_rate": 6.83907713657916e-05, "loss": 1.8801, "step": 30980 }, { "epoch": 2.68, "learning_rate": 6.838860298369376e-05, "loss": 1.995, "step": 30990 }, { "epoch": 2.68, "learning_rate": 6.838643460159593e-05, "loss": 2.0742, "step": 31000 }, { "epoch": 2.69, "learning_rate": 6.83842662194981e-05, "loss": 1.9876, "step": 31010 }, { "epoch": 2.69, "learning_rate": 6.838209783740025e-05, "loss": 1.9894, "step": 31020 }, { "epoch": 2.69, "learning_rate": 6.837992945530241e-05, "loss": 1.9601, "step": 31030 }, { "epoch": 2.69, "learning_rate": 6.837776107320457e-05, "loss": 1.9713, "step": 31040 }, { "epoch": 2.69, "learning_rate": 6.837559269110673e-05, "loss": 1.8201, "step": 31050 }, { "epoch": 2.69, "learning_rate": 6.83734243090089e-05, "loss": 1.8363, "step": 31060 }, { "epoch": 2.69, "learning_rate": 6.837125592691107e-05, "loss": 2.1223, "step": 31070 }, { "epoch": 2.69, "learning_rate": 6.836908754481323e-05, "loss": 1.9779, "step": 31080 }, { "epoch": 2.69, "learning_rate": 6.836691916271538e-05, "loss": 2.0984, "step": 31090 }, { "epoch": 2.69, "learning_rate": 6.836475078061755e-05, "loss": 1.8068, "step": 31100 }, { "epoch": 2.69, "learning_rate": 6.83625823985197e-05, "loss": 1.9016, "step": 31110 }, { "epoch": 2.7, "learning_rate": 6.836041401642187e-05, "loss": 1.9738, "step": 31120 }, { "epoch": 2.7, "learning_rate": 6.835824563432404e-05, "loss": 2.1473, "step": 31130 }, { "epoch": 2.7, "learning_rate": 6.83560772522262e-05, "loss": 1.9165, "step": 31140 }, { "epoch": 2.7, "learning_rate": 6.835390887012837e-05, "loss": 1.8147, "step": 31150 }, { "epoch": 2.7, "learning_rate": 6.835174048803052e-05, "loss": 1.998, "step": 31160 }, { "epoch": 2.7, "learning_rate": 6.834957210593269e-05, "loss": 2.1426, "step": 31170 }, { "epoch": 2.7, "learning_rate": 6.834740372383485e-05, "loss": 1.9155, "step": 31180 }, { "epoch": 2.7, "learning_rate": 6.834523534173701e-05, "loss": 2.0375, "step": 31190 }, { "epoch": 2.7, "learning_rate": 6.834306695963917e-05, "loss": 1.9948, "step": 31200 }, { "epoch": 2.7, "learning_rate": 6.834089857754134e-05, "loss": 2.0044, "step": 31210 }, { "epoch": 2.7, "learning_rate": 6.83387301954435e-05, "loss": 1.9866, "step": 31220 }, { "epoch": 2.7, "learning_rate": 6.833656181334566e-05, "loss": 2.0543, "step": 31230 }, { "epoch": 2.71, "learning_rate": 6.833439343124783e-05, "loss": 1.8389, "step": 31240 }, { "epoch": 2.71, "learning_rate": 6.833222504914999e-05, "loss": 1.9913, "step": 31250 }, { "epoch": 2.71, "learning_rate": 6.833005666705216e-05, "loss": 1.8565, "step": 31260 }, { "epoch": 2.71, "learning_rate": 6.832788828495431e-05, "loss": 2.0822, "step": 31270 }, { "epoch": 2.71, "learning_rate": 6.832571990285648e-05, "loss": 1.8583, "step": 31280 }, { "epoch": 2.71, "learning_rate": 6.832355152075863e-05, "loss": 1.9683, "step": 31290 }, { "epoch": 2.71, "learning_rate": 6.83213831386608e-05, "loss": 1.9543, "step": 31300 }, { "epoch": 2.71, "learning_rate": 6.831921475656296e-05, "loss": 2.0205, "step": 31310 }, { "epoch": 2.71, "learning_rate": 6.831704637446513e-05, "loss": 1.8773, "step": 31320 }, { "epoch": 2.71, "learning_rate": 6.83148779923673e-05, "loss": 1.9114, "step": 31330 }, { "epoch": 2.71, "learning_rate": 6.831270961026945e-05, "loss": 1.9905, "step": 31340 }, { "epoch": 2.72, "learning_rate": 6.831054122817162e-05, "loss": 1.854, "step": 31350 }, { "epoch": 2.72, "learning_rate": 6.830837284607377e-05, "loss": 1.9674, "step": 31360 }, { "epoch": 2.72, "learning_rate": 6.830620446397594e-05, "loss": 2.1436, "step": 31370 }, { "epoch": 2.72, "learning_rate": 6.83040360818781e-05, "loss": 2.0816, "step": 31380 }, { "epoch": 2.72, "learning_rate": 6.830186769978027e-05, "loss": 1.9069, "step": 31390 }, { "epoch": 2.72, "learning_rate": 6.829969931768243e-05, "loss": 1.9046, "step": 31400 }, { "epoch": 2.72, "learning_rate": 6.829753093558459e-05, "loss": 1.6728, "step": 31410 }, { "epoch": 2.72, "learning_rate": 6.829536255348675e-05, "loss": 2.0049, "step": 31420 }, { "epoch": 2.72, "learning_rate": 6.82931941713889e-05, "loss": 1.93, "step": 31430 }, { "epoch": 2.72, "learning_rate": 6.829102578929107e-05, "loss": 1.8094, "step": 31440 }, { "epoch": 2.72, "learning_rate": 6.828885740719324e-05, "loss": 1.7602, "step": 31450 }, { "epoch": 2.72, "learning_rate": 6.82866890250954e-05, "loss": 1.962, "step": 31460 }, { "epoch": 2.73, "learning_rate": 6.828452064299757e-05, "loss": 1.9743, "step": 31470 }, { "epoch": 2.73, "learning_rate": 6.828235226089972e-05, "loss": 2.0066, "step": 31480 }, { "epoch": 2.73, "learning_rate": 6.828018387880189e-05, "loss": 1.8618, "step": 31490 }, { "epoch": 2.73, "learning_rate": 6.827801549670404e-05, "loss": 1.8409, "step": 31500 }, { "epoch": 2.73, "learning_rate": 6.827584711460621e-05, "loss": 1.975, "step": 31510 }, { "epoch": 2.73, "learning_rate": 6.827367873250838e-05, "loss": 2.1176, "step": 31520 }, { "epoch": 2.73, "learning_rate": 6.827151035041054e-05, "loss": 1.7845, "step": 31530 }, { "epoch": 2.73, "learning_rate": 6.826934196831271e-05, "loss": 2.0466, "step": 31540 }, { "epoch": 2.73, "learning_rate": 6.826717358621486e-05, "loss": 1.9138, "step": 31550 }, { "epoch": 2.73, "learning_rate": 6.826500520411703e-05, "loss": 2.0576, "step": 31560 }, { "epoch": 2.73, "learning_rate": 6.826283682201918e-05, "loss": 2.0277, "step": 31570 }, { "epoch": 2.74, "learning_rate": 6.826066843992135e-05, "loss": 2.1417, "step": 31580 }, { "epoch": 2.74, "learning_rate": 6.825850005782351e-05, "loss": 1.893, "step": 31590 }, { "epoch": 2.74, "learning_rate": 6.825633167572568e-05, "loss": 1.8149, "step": 31600 }, { "epoch": 2.74, "learning_rate": 6.825416329362785e-05, "loss": 1.8736, "step": 31610 }, { "epoch": 2.74, "learning_rate": 6.825199491153e-05, "loss": 1.8553, "step": 31620 }, { "epoch": 2.74, "learning_rate": 6.824982652943217e-05, "loss": 1.9709, "step": 31630 }, { "epoch": 2.74, "learning_rate": 6.824765814733433e-05, "loss": 2.1881, "step": 31640 }, { "epoch": 2.74, "learning_rate": 6.824548976523649e-05, "loss": 1.785, "step": 31650 }, { "epoch": 2.74, "learning_rate": 6.824332138313865e-05, "loss": 1.9504, "step": 31660 }, { "epoch": 2.74, "learning_rate": 6.824115300104082e-05, "loss": 2.058, "step": 31670 }, { "epoch": 2.74, "learning_rate": 6.823898461894298e-05, "loss": 1.6379, "step": 31680 }, { "epoch": 2.74, "learning_rate": 6.823681623684514e-05, "loss": 2.0989, "step": 31690 }, { "epoch": 2.75, "learning_rate": 6.82346478547473e-05, "loss": 1.7531, "step": 31700 }, { "epoch": 2.75, "learning_rate": 6.823247947264947e-05, "loss": 1.8962, "step": 31710 }, { "epoch": 2.75, "learning_rate": 6.823031109055164e-05, "loss": 2.1233, "step": 31720 }, { "epoch": 2.75, "learning_rate": 6.822814270845379e-05, "loss": 1.8088, "step": 31730 }, { "epoch": 2.75, "learning_rate": 6.822597432635596e-05, "loss": 2.0428, "step": 31740 }, { "epoch": 2.75, "learning_rate": 6.822380594425812e-05, "loss": 1.9996, "step": 31750 }, { "epoch": 2.75, "learning_rate": 6.822163756216028e-05, "loss": 1.9526, "step": 31760 }, { "epoch": 2.75, "learning_rate": 6.821946918006244e-05, "loss": 1.711, "step": 31770 }, { "epoch": 2.75, "learning_rate": 6.821730079796461e-05, "loss": 1.9689, "step": 31780 }, { "epoch": 2.75, "learning_rate": 6.821513241586677e-05, "loss": 1.8657, "step": 31790 }, { "epoch": 2.75, "learning_rate": 6.821296403376894e-05, "loss": 2.0763, "step": 31800 }, { "epoch": 2.75, "learning_rate": 6.82107956516711e-05, "loss": 1.8417, "step": 31810 }, { "epoch": 2.76, "learning_rate": 6.820862726957326e-05, "loss": 1.8725, "step": 31820 }, { "epoch": 2.76, "learning_rate": 6.820645888747541e-05, "loss": 1.9788, "step": 31830 }, { "epoch": 2.76, "learning_rate": 6.820429050537758e-05, "loss": 1.8368, "step": 31840 }, { "epoch": 2.76, "learning_rate": 6.820212212327975e-05, "loss": 2.105, "step": 31850 }, { "epoch": 2.76, "learning_rate": 6.819995374118191e-05, "loss": 1.91, "step": 31860 }, { "epoch": 2.76, "learning_rate": 6.819778535908408e-05, "loss": 1.9821, "step": 31870 }, { "epoch": 2.76, "learning_rate": 6.819561697698623e-05, "loss": 2.1267, "step": 31880 }, { "epoch": 2.76, "learning_rate": 6.81934485948884e-05, "loss": 1.8699, "step": 31890 }, { "epoch": 2.76, "learning_rate": 6.819128021279055e-05, "loss": 1.9032, "step": 31900 }, { "epoch": 2.76, "learning_rate": 6.818911183069272e-05, "loss": 2.0955, "step": 31910 }, { "epoch": 2.76, "learning_rate": 6.818694344859488e-05, "loss": 1.9544, "step": 31920 }, { "epoch": 2.77, "learning_rate": 6.818477506649705e-05, "loss": 1.8349, "step": 31930 }, { "epoch": 2.77, "learning_rate": 6.818260668439922e-05, "loss": 2.0914, "step": 31940 }, { "epoch": 2.77, "learning_rate": 6.818043830230137e-05, "loss": 1.9522, "step": 31950 }, { "epoch": 2.77, "learning_rate": 6.817826992020354e-05, "loss": 2.0076, "step": 31960 }, { "epoch": 2.77, "learning_rate": 6.817610153810569e-05, "loss": 1.8528, "step": 31970 }, { "epoch": 2.77, "learning_rate": 6.817393315600785e-05, "loss": 1.8079, "step": 31980 }, { "epoch": 2.77, "learning_rate": 6.817176477391002e-05, "loss": 1.9892, "step": 31990 }, { "epoch": 2.77, "learning_rate": 6.816959639181219e-05, "loss": 1.8958, "step": 32000 }, { "epoch": 2.77, "learning_rate": 6.816742800971435e-05, "loss": 1.9811, "step": 32010 }, { "epoch": 2.77, "learning_rate": 6.81652596276165e-05, "loss": 1.8046, "step": 32020 }, { "epoch": 2.77, "learning_rate": 6.816309124551867e-05, "loss": 1.9361, "step": 32030 }, { "epoch": 2.77, "learning_rate": 6.816092286342083e-05, "loss": 1.8852, "step": 32040 }, { "epoch": 2.78, "learning_rate": 6.815875448132299e-05, "loss": 1.8835, "step": 32050 }, { "epoch": 2.78, "learning_rate": 6.815658609922516e-05, "loss": 1.8964, "step": 32060 }, { "epoch": 2.78, "learning_rate": 6.815441771712732e-05, "loss": 2.2055, "step": 32070 }, { "epoch": 2.78, "learning_rate": 6.815224933502949e-05, "loss": 1.8504, "step": 32080 }, { "epoch": 2.78, "learning_rate": 6.815008095293164e-05, "loss": 1.8454, "step": 32090 }, { "epoch": 2.78, "learning_rate": 6.814791257083381e-05, "loss": 1.9036, "step": 32100 }, { "epoch": 2.78, "learning_rate": 6.814574418873596e-05, "loss": 1.9592, "step": 32110 }, { "epoch": 2.78, "learning_rate": 6.814357580663813e-05, "loss": 1.73, "step": 32120 }, { "epoch": 2.78, "learning_rate": 6.81414074245403e-05, "loss": 1.9421, "step": 32130 }, { "epoch": 2.78, "learning_rate": 6.813923904244246e-05, "loss": 1.8602, "step": 32140 }, { "epoch": 2.78, "learning_rate": 6.813707066034463e-05, "loss": 2.048, "step": 32150 }, { "epoch": 2.79, "learning_rate": 6.813490227824678e-05, "loss": 2.0527, "step": 32160 }, { "epoch": 2.79, "learning_rate": 6.813273389614895e-05, "loss": 2.0139, "step": 32170 }, { "epoch": 2.79, "learning_rate": 6.813056551405111e-05, "loss": 1.9518, "step": 32180 }, { "epoch": 2.79, "learning_rate": 6.812839713195327e-05, "loss": 1.9393, "step": 32190 }, { "epoch": 2.79, "learning_rate": 6.812622874985543e-05, "loss": 1.8809, "step": 32200 }, { "epoch": 2.79, "learning_rate": 6.81240603677576e-05, "loss": 1.8932, "step": 32210 }, { "epoch": 2.79, "learning_rate": 6.812189198565977e-05, "loss": 2.0101, "step": 32220 }, { "epoch": 2.79, "learning_rate": 6.811972360356192e-05, "loss": 1.7374, "step": 32230 }, { "epoch": 2.79, "learning_rate": 6.811755522146409e-05, "loss": 2.017, "step": 32240 }, { "epoch": 2.79, "learning_rate": 6.811538683936625e-05, "loss": 2.0167, "step": 32250 }, { "epoch": 2.79, "learning_rate": 6.811321845726842e-05, "loss": 1.9238, "step": 32260 }, { "epoch": 2.79, "learning_rate": 6.811105007517057e-05, "loss": 1.7596, "step": 32270 }, { "epoch": 2.8, "learning_rate": 6.810888169307274e-05, "loss": 1.9317, "step": 32280 }, { "epoch": 2.8, "learning_rate": 6.810671331097489e-05, "loss": 1.8855, "step": 32290 }, { "epoch": 2.8, "learning_rate": 6.810454492887706e-05, "loss": 1.7769, "step": 32300 }, { "epoch": 2.8, "learning_rate": 6.810237654677922e-05, "loss": 2.1392, "step": 32310 }, { "epoch": 2.8, "learning_rate": 6.810020816468139e-05, "loss": 1.9761, "step": 32320 }, { "epoch": 2.8, "learning_rate": 6.809803978258356e-05, "loss": 1.7475, "step": 32330 }, { "epoch": 2.8, "learning_rate": 6.809587140048571e-05, "loss": 1.9028, "step": 32340 }, { "epoch": 2.8, "learning_rate": 6.809370301838788e-05, "loss": 1.8735, "step": 32350 }, { "epoch": 2.8, "learning_rate": 6.809153463629003e-05, "loss": 1.8854, "step": 32360 }, { "epoch": 2.8, "learning_rate": 6.80893662541922e-05, "loss": 1.8623, "step": 32370 }, { "epoch": 2.8, "learning_rate": 6.808719787209436e-05, "loss": 1.8079, "step": 32380 }, { "epoch": 2.81, "learning_rate": 6.808502948999653e-05, "loss": 1.6738, "step": 32390 }, { "epoch": 2.81, "learning_rate": 6.80828611078987e-05, "loss": 1.8094, "step": 32400 }, { "epoch": 2.81, "learning_rate": 6.808069272580085e-05, "loss": 2.1435, "step": 32410 }, { "epoch": 2.81, "learning_rate": 6.807852434370301e-05, "loss": 2.0433, "step": 32420 }, { "epoch": 2.81, "learning_rate": 6.807635596160517e-05, "loss": 1.7637, "step": 32430 }, { "epoch": 2.81, "learning_rate": 6.807418757950733e-05, "loss": 1.8048, "step": 32440 }, { "epoch": 2.81, "learning_rate": 6.80720191974095e-05, "loss": 1.8093, "step": 32450 }, { "epoch": 2.81, "learning_rate": 6.806985081531167e-05, "loss": 1.879, "step": 32460 }, { "epoch": 2.81, "learning_rate": 6.806768243321383e-05, "loss": 1.7522, "step": 32470 }, { "epoch": 2.81, "learning_rate": 6.806551405111598e-05, "loss": 2.0276, "step": 32480 }, { "epoch": 2.81, "learning_rate": 6.806334566901815e-05, "loss": 1.9978, "step": 32490 }, { "epoch": 2.81, "learning_rate": 6.80611772869203e-05, "loss": 2.0423, "step": 32500 }, { "epoch": 2.82, "learning_rate": 6.805900890482247e-05, "loss": 1.9653, "step": 32510 }, { "epoch": 2.82, "learning_rate": 6.805684052272464e-05, "loss": 1.6569, "step": 32520 }, { "epoch": 2.82, "learning_rate": 6.80546721406268e-05, "loss": 1.9145, "step": 32530 }, { "epoch": 2.82, "learning_rate": 6.805250375852897e-05, "loss": 1.8217, "step": 32540 }, { "epoch": 2.82, "learning_rate": 6.805033537643112e-05, "loss": 1.7848, "step": 32550 }, { "epoch": 2.82, "learning_rate": 6.804816699433329e-05, "loss": 1.8906, "step": 32560 }, { "epoch": 2.82, "learning_rate": 6.804599861223545e-05, "loss": 2.0739, "step": 32570 }, { "epoch": 2.82, "learning_rate": 6.804383023013761e-05, "loss": 1.964, "step": 32580 }, { "epoch": 2.82, "learning_rate": 6.804166184803977e-05, "loss": 1.8567, "step": 32590 }, { "epoch": 2.82, "learning_rate": 6.803949346594194e-05, "loss": 1.9685, "step": 32600 }, { "epoch": 2.82, "learning_rate": 6.803732508384411e-05, "loss": 1.9391, "step": 32610 }, { "epoch": 2.83, "learning_rate": 6.803515670174626e-05, "loss": 2.0563, "step": 32620 }, { "epoch": 2.83, "learning_rate": 6.803298831964843e-05, "loss": 1.8566, "step": 32630 }, { "epoch": 2.83, "learning_rate": 6.803081993755059e-05, "loss": 1.9461, "step": 32640 }, { "epoch": 2.83, "learning_rate": 6.802865155545275e-05, "loss": 1.8214, "step": 32650 }, { "epoch": 2.83, "learning_rate": 6.802648317335491e-05, "loss": 2.081, "step": 32660 }, { "epoch": 2.83, "learning_rate": 6.802431479125708e-05, "loss": 1.9483, "step": 32670 }, { "epoch": 2.83, "learning_rate": 6.802214640915924e-05, "loss": 1.8193, "step": 32680 }, { "epoch": 2.83, "learning_rate": 6.80199780270614e-05, "loss": 1.7624, "step": 32690 }, { "epoch": 2.83, "learning_rate": 6.801780964496356e-05, "loss": 1.9169, "step": 32700 }, { "epoch": 2.83, "learning_rate": 6.801564126286573e-05, "loss": 1.9977, "step": 32710 }, { "epoch": 2.83, "learning_rate": 6.80134728807679e-05, "loss": 1.9374, "step": 32720 }, { "epoch": 2.83, "learning_rate": 6.801130449867005e-05, "loss": 2.169, "step": 32730 }, { "epoch": 2.84, "learning_rate": 6.800913611657222e-05, "loss": 1.783, "step": 32740 }, { "epoch": 2.84, "learning_rate": 6.800696773447438e-05, "loss": 1.8088, "step": 32750 }, { "epoch": 2.84, "learning_rate": 6.800479935237653e-05, "loss": 1.8301, "step": 32760 }, { "epoch": 2.84, "learning_rate": 6.80026309702787e-05, "loss": 1.8725, "step": 32770 }, { "epoch": 2.84, "learning_rate": 6.800046258818087e-05, "loss": 2.0018, "step": 32780 }, { "epoch": 2.84, "learning_rate": 6.799829420608303e-05, "loss": 2.0133, "step": 32790 }, { "epoch": 2.84, "learning_rate": 6.79961258239852e-05, "loss": 1.9475, "step": 32800 }, { "epoch": 2.84, "learning_rate": 6.799395744188735e-05, "loss": 1.795, "step": 32810 }, { "epoch": 2.84, "learning_rate": 6.799178905978952e-05, "loss": 2.15, "step": 32820 }, { "epoch": 2.84, "learning_rate": 6.798962067769167e-05, "loss": 1.9925, "step": 32830 }, { "epoch": 2.84, "learning_rate": 6.798745229559384e-05, "loss": 2.1082, "step": 32840 }, { "epoch": 2.85, "learning_rate": 6.7985283913496e-05, "loss": 1.8628, "step": 32850 }, { "epoch": 2.85, "learning_rate": 6.798311553139817e-05, "loss": 1.8889, "step": 32860 }, { "epoch": 2.85, "learning_rate": 6.798094714930034e-05, "loss": 1.9308, "step": 32870 }, { "epoch": 2.85, "learning_rate": 6.797877876720249e-05, "loss": 1.7396, "step": 32880 }, { "epoch": 2.85, "learning_rate": 6.797661038510466e-05, "loss": 2.0507, "step": 32890 }, { "epoch": 2.85, "learning_rate": 6.797444200300681e-05, "loss": 1.973, "step": 32900 }, { "epoch": 2.85, "learning_rate": 6.797227362090898e-05, "loss": 1.615, "step": 32910 }, { "epoch": 2.85, "learning_rate": 6.797010523881114e-05, "loss": 2.027, "step": 32920 }, { "epoch": 2.85, "learning_rate": 6.796793685671331e-05, "loss": 1.8647, "step": 32930 }, { "epoch": 2.85, "learning_rate": 6.796576847461548e-05, "loss": 1.8502, "step": 32940 }, { "epoch": 2.85, "learning_rate": 6.796360009251763e-05, "loss": 1.9395, "step": 32950 }, { "epoch": 2.85, "learning_rate": 6.79614317104198e-05, "loss": 1.9216, "step": 32960 }, { "epoch": 2.86, "learning_rate": 6.795926332832195e-05, "loss": 1.9324, "step": 32970 }, { "epoch": 2.86, "learning_rate": 6.795709494622411e-05, "loss": 1.8381, "step": 32980 }, { "epoch": 2.86, "learning_rate": 6.795492656412628e-05, "loss": 1.9053, "step": 32990 }, { "epoch": 2.86, "learning_rate": 6.795275818202845e-05, "loss": 1.7176, "step": 33000 }, { "epoch": 2.86, "learning_rate": 6.795058979993061e-05, "loss": 1.7983, "step": 33010 }, { "epoch": 2.86, "learning_rate": 6.794842141783277e-05, "loss": 1.913, "step": 33020 }, { "epoch": 2.86, "learning_rate": 6.794625303573493e-05, "loss": 1.7932, "step": 33030 }, { "epoch": 2.86, "learning_rate": 6.794408465363709e-05, "loss": 1.8808, "step": 33040 }, { "epoch": 2.86, "learning_rate": 6.794191627153925e-05, "loss": 1.9121, "step": 33050 }, { "epoch": 2.86, "learning_rate": 6.793974788944142e-05, "loss": 1.9037, "step": 33060 }, { "epoch": 2.86, "learning_rate": 6.793757950734358e-05, "loss": 2.0925, "step": 33070 }, { "epoch": 2.86, "learning_rate": 6.793541112524575e-05, "loss": 1.9166, "step": 33080 }, { "epoch": 2.87, "learning_rate": 6.79332427431479e-05, "loss": 2.0727, "step": 33090 }, { "epoch": 2.87, "learning_rate": 6.793107436105007e-05, "loss": 1.9249, "step": 33100 }, { "epoch": 2.87, "learning_rate": 6.792890597895222e-05, "loss": 2.076, "step": 33110 }, { "epoch": 2.87, "learning_rate": 6.792673759685439e-05, "loss": 1.9134, "step": 33120 }, { "epoch": 2.87, "learning_rate": 6.792456921475656e-05, "loss": 2.0159, "step": 33130 }, { "epoch": 2.87, "learning_rate": 6.792240083265872e-05, "loss": 1.988, "step": 33140 }, { "epoch": 2.87, "learning_rate": 6.792023245056089e-05, "loss": 1.977, "step": 33150 }, { "epoch": 2.87, "learning_rate": 6.791806406846304e-05, "loss": 1.8724, "step": 33160 }, { "epoch": 2.87, "learning_rate": 6.791589568636521e-05, "loss": 1.9608, "step": 33170 }, { "epoch": 2.87, "learning_rate": 6.791372730426737e-05, "loss": 1.8959, "step": 33180 }, { "epoch": 2.87, "learning_rate": 6.791155892216953e-05, "loss": 1.7211, "step": 33190 }, { "epoch": 2.88, "learning_rate": 6.79093905400717e-05, "loss": 1.9704, "step": 33200 }, { "epoch": 2.88, "learning_rate": 6.790722215797386e-05, "loss": 1.9139, "step": 33210 }, { "epoch": 2.88, "learning_rate": 6.790505377587603e-05, "loss": 1.9271, "step": 33220 }, { "epoch": 2.88, "learning_rate": 6.790288539377818e-05, "loss": 2.0135, "step": 33230 }, { "epoch": 2.88, "learning_rate": 6.790071701168035e-05, "loss": 1.8427, "step": 33240 }, { "epoch": 2.88, "learning_rate": 6.789854862958251e-05, "loss": 1.8335, "step": 33250 }, { "epoch": 2.88, "learning_rate": 6.789638024748468e-05, "loss": 1.9758, "step": 33260 }, { "epoch": 2.88, "learning_rate": 6.789421186538683e-05, "loss": 1.8068, "step": 33270 }, { "epoch": 2.88, "learning_rate": 6.7892043483289e-05, "loss": 2.0705, "step": 33280 }, { "epoch": 2.88, "learning_rate": 6.788987510119116e-05, "loss": 1.8884, "step": 33290 }, { "epoch": 2.88, "learning_rate": 6.788770671909332e-05, "loss": 1.9461, "step": 33300 }, { "epoch": 2.88, "learning_rate": 6.788553833699548e-05, "loss": 1.8914, "step": 33310 }, { "epoch": 2.89, "learning_rate": 6.788336995489765e-05, "loss": 1.9848, "step": 33320 }, { "epoch": 2.89, "learning_rate": 6.788120157279982e-05, "loss": 1.9071, "step": 33330 }, { "epoch": 2.89, "learning_rate": 6.787903319070198e-05, "loss": 2.0962, "step": 33340 }, { "epoch": 2.89, "learning_rate": 6.787686480860414e-05, "loss": 1.6179, "step": 33350 }, { "epoch": 2.89, "learning_rate": 6.787469642650629e-05, "loss": 1.9737, "step": 33360 }, { "epoch": 2.89, "learning_rate": 6.787252804440845e-05, "loss": 2.0417, "step": 33370 }, { "epoch": 2.89, "learning_rate": 6.787035966231062e-05, "loss": 1.8834, "step": 33380 }, { "epoch": 2.89, "learning_rate": 6.786819128021279e-05, "loss": 1.9703, "step": 33390 }, { "epoch": 2.89, "learning_rate": 6.786602289811495e-05, "loss": 1.6361, "step": 33400 }, { "epoch": 2.89, "learning_rate": 6.78638545160171e-05, "loss": 2.0769, "step": 33410 }, { "epoch": 2.89, "learning_rate": 6.786168613391927e-05, "loss": 1.9155, "step": 33420 }, { "epoch": 2.9, "learning_rate": 6.785951775182143e-05, "loss": 1.7062, "step": 33430 }, { "epoch": 2.9, "learning_rate": 6.785734936972359e-05, "loss": 1.9488, "step": 33440 }, { "epoch": 2.9, "learning_rate": 6.785518098762576e-05, "loss": 1.8266, "step": 33450 }, { "epoch": 2.9, "learning_rate": 6.785301260552792e-05, "loss": 2.0534, "step": 33460 }, { "epoch": 2.9, "learning_rate": 6.785084422343009e-05, "loss": 1.8329, "step": 33470 }, { "epoch": 2.9, "learning_rate": 6.784867584133224e-05, "loss": 1.6146, "step": 33480 }, { "epoch": 2.9, "learning_rate": 6.784650745923441e-05, "loss": 1.8963, "step": 33490 }, { "epoch": 2.9, "learning_rate": 6.784433907713656e-05, "loss": 1.8304, "step": 33500 }, { "epoch": 2.9, "learning_rate": 6.784217069503873e-05, "loss": 1.8734, "step": 33510 }, { "epoch": 2.9, "learning_rate": 6.78400023129409e-05, "loss": 1.8992, "step": 33520 }, { "epoch": 2.9, "learning_rate": 6.783783393084306e-05, "loss": 2.2249, "step": 33530 }, { "epoch": 2.9, "learning_rate": 6.783566554874523e-05, "loss": 1.8362, "step": 33540 }, { "epoch": 2.91, "learning_rate": 6.783349716664738e-05, "loss": 1.7982, "step": 33550 }, { "epoch": 2.91, "learning_rate": 6.783132878454955e-05, "loss": 1.8891, "step": 33560 }, { "epoch": 2.91, "learning_rate": 6.782916040245171e-05, "loss": 1.8181, "step": 33570 }, { "epoch": 2.91, "learning_rate": 6.782699202035387e-05, "loss": 1.8769, "step": 33580 }, { "epoch": 2.91, "learning_rate": 6.782482363825603e-05, "loss": 1.7307, "step": 33590 }, { "epoch": 2.91, "learning_rate": 6.78226552561582e-05, "loss": 1.9425, "step": 33600 }, { "epoch": 2.91, "learning_rate": 6.782048687406037e-05, "loss": 1.8038, "step": 33610 }, { "epoch": 2.91, "learning_rate": 6.781831849196252e-05, "loss": 1.68, "step": 33620 }, { "epoch": 2.91, "learning_rate": 6.781615010986469e-05, "loss": 1.7445, "step": 33630 }, { "epoch": 2.91, "learning_rate": 6.781398172776685e-05, "loss": 1.9035, "step": 33640 }, { "epoch": 2.91, "learning_rate": 6.7811813345669e-05, "loss": 2.1215, "step": 33650 }, { "epoch": 2.92, "learning_rate": 6.780964496357117e-05, "loss": 2.0982, "step": 33660 }, { "epoch": 2.92, "learning_rate": 6.780747658147334e-05, "loss": 2.0346, "step": 33670 }, { "epoch": 2.92, "learning_rate": 6.78053081993755e-05, "loss": 1.9447, "step": 33680 }, { "epoch": 2.92, "learning_rate": 6.780313981727766e-05, "loss": 1.9411, "step": 33690 }, { "epoch": 2.92, "learning_rate": 6.780097143517982e-05, "loss": 1.8516, "step": 33700 }, { "epoch": 2.92, "learning_rate": 6.779880305308199e-05, "loss": 1.9888, "step": 33710 }, { "epoch": 2.92, "learning_rate": 6.779663467098416e-05, "loss": 1.9125, "step": 33720 }, { "epoch": 2.92, "learning_rate": 6.779446628888631e-05, "loss": 1.9006, "step": 33730 }, { "epoch": 2.92, "learning_rate": 6.779229790678848e-05, "loss": 1.8144, "step": 33740 }, { "epoch": 2.92, "learning_rate": 6.779012952469064e-05, "loss": 1.9629, "step": 33750 }, { "epoch": 2.92, "learning_rate": 6.77879611425928e-05, "loss": 2.0951, "step": 33760 }, { "epoch": 2.92, "learning_rate": 6.778579276049496e-05, "loss": 2.0815, "step": 33770 }, { "epoch": 2.93, "learning_rate": 6.778362437839713e-05, "loss": 1.8269, "step": 33780 }, { "epoch": 2.93, "learning_rate": 6.77814559962993e-05, "loss": 1.8687, "step": 33790 }, { "epoch": 2.93, "learning_rate": 6.777928761420146e-05, "loss": 1.9731, "step": 33800 }, { "epoch": 2.93, "learning_rate": 6.777711923210361e-05, "loss": 1.9243, "step": 33810 }, { "epoch": 2.93, "learning_rate": 6.777495085000578e-05, "loss": 2.0712, "step": 33820 }, { "epoch": 2.93, "learning_rate": 6.777278246790793e-05, "loss": 1.6575, "step": 33830 }, { "epoch": 2.93, "learning_rate": 6.77706140858101e-05, "loss": 1.8532, "step": 33840 }, { "epoch": 2.93, "learning_rate": 6.776844570371227e-05, "loss": 1.7171, "step": 33850 }, { "epoch": 2.93, "learning_rate": 6.776627732161443e-05, "loss": 1.7888, "step": 33860 }, { "epoch": 2.93, "learning_rate": 6.77641089395166e-05, "loss": 1.9542, "step": 33870 }, { "epoch": 2.93, "learning_rate": 6.776194055741875e-05, "loss": 1.9506, "step": 33880 }, { "epoch": 2.94, "learning_rate": 6.775977217532092e-05, "loss": 1.6316, "step": 33890 }, { "epoch": 2.94, "learning_rate": 6.775760379322307e-05, "loss": 1.7609, "step": 33900 }, { "epoch": 2.94, "learning_rate": 6.775543541112524e-05, "loss": 1.8154, "step": 33910 }, { "epoch": 2.94, "learning_rate": 6.77532670290274e-05, "loss": 1.9291, "step": 33920 }, { "epoch": 2.94, "learning_rate": 6.775109864692957e-05, "loss": 1.7922, "step": 33930 }, { "epoch": 2.94, "learning_rate": 6.774893026483174e-05, "loss": 1.9528, "step": 33940 }, { "epoch": 2.94, "learning_rate": 6.774676188273389e-05, "loss": 1.9781, "step": 33950 }, { "epoch": 2.94, "learning_rate": 6.774459350063605e-05, "loss": 1.8561, "step": 33960 }, { "epoch": 2.94, "learning_rate": 6.774242511853821e-05, "loss": 1.7059, "step": 33970 }, { "epoch": 2.94, "learning_rate": 6.774025673644037e-05, "loss": 1.8878, "step": 33980 }, { "epoch": 2.94, "learning_rate": 6.773808835434254e-05, "loss": 2.1723, "step": 33990 }, { "epoch": 2.94, "learning_rate": 6.77359199722447e-05, "loss": 1.9811, "step": 34000 }, { "epoch": 2.95, "learning_rate": 6.773375159014687e-05, "loss": 1.8743, "step": 34010 }, { "epoch": 2.95, "learning_rate": 6.773158320804903e-05, "loss": 2.0735, "step": 34020 }, { "epoch": 2.95, "learning_rate": 6.772941482595119e-05, "loss": 1.8281, "step": 34030 }, { "epoch": 2.95, "learning_rate": 6.772724644385335e-05, "loss": 1.8339, "step": 34040 }, { "epoch": 2.95, "learning_rate": 6.772507806175551e-05, "loss": 1.9979, "step": 34050 }, { "epoch": 2.95, "learning_rate": 6.772290967965768e-05, "loss": 1.7997, "step": 34060 }, { "epoch": 2.95, "learning_rate": 6.772074129755984e-05, "loss": 1.9188, "step": 34070 }, { "epoch": 2.95, "learning_rate": 6.771857291546201e-05, "loss": 1.9443, "step": 34080 }, { "epoch": 2.95, "learning_rate": 6.771640453336416e-05, "loss": 1.9774, "step": 34090 }, { "epoch": 2.95, "learning_rate": 6.771423615126633e-05, "loss": 1.8327, "step": 34100 }, { "epoch": 2.95, "learning_rate": 6.77120677691685e-05, "loss": 1.8603, "step": 34110 }, { "epoch": 2.96, "learning_rate": 6.770989938707065e-05, "loss": 1.705, "step": 34120 }, { "epoch": 2.96, "learning_rate": 6.770773100497282e-05, "loss": 2.1472, "step": 34130 }, { "epoch": 2.96, "learning_rate": 6.770556262287498e-05, "loss": 2.0195, "step": 34140 }, { "epoch": 2.96, "learning_rate": 6.770339424077715e-05, "loss": 1.9545, "step": 34150 }, { "epoch": 2.96, "learning_rate": 6.77012258586793e-05, "loss": 1.8456, "step": 34160 }, { "epoch": 2.96, "learning_rate": 6.769905747658147e-05, "loss": 1.9412, "step": 34170 }, { "epoch": 2.96, "learning_rate": 6.769688909448363e-05, "loss": 2.0465, "step": 34180 }, { "epoch": 2.96, "learning_rate": 6.769472071238579e-05, "loss": 1.9613, "step": 34190 }, { "epoch": 2.96, "learning_rate": 6.769255233028795e-05, "loss": 1.9567, "step": 34200 }, { "epoch": 2.96, "learning_rate": 6.769038394819012e-05, "loss": 2.0431, "step": 34210 }, { "epoch": 2.96, "learning_rate": 6.768821556609229e-05, "loss": 2.1287, "step": 34220 }, { "epoch": 2.96, "learning_rate": 6.768604718399444e-05, "loss": 1.8815, "step": 34230 }, { "epoch": 2.97, "learning_rate": 6.76838788018966e-05, "loss": 1.7964, "step": 34240 }, { "epoch": 2.97, "learning_rate": 6.768171041979877e-05, "loss": 2.1848, "step": 34250 }, { "epoch": 2.97, "learning_rate": 6.767954203770094e-05, "loss": 1.8256, "step": 34260 }, { "epoch": 2.97, "learning_rate": 6.767737365560309e-05, "loss": 1.9838, "step": 34270 }, { "epoch": 2.97, "learning_rate": 6.767520527350526e-05, "loss": 2.369, "step": 34280 }, { "epoch": 2.97, "learning_rate": 6.767303689140742e-05, "loss": 1.8915, "step": 34290 }, { "epoch": 2.97, "learning_rate": 6.767086850930958e-05, "loss": 2.0837, "step": 34300 }, { "epoch": 2.97, "learning_rate": 6.766870012721174e-05, "loss": 1.9126, "step": 34310 }, { "epoch": 2.97, "learning_rate": 6.766653174511391e-05, "loss": 1.7701, "step": 34320 }, { "epoch": 2.97, "learning_rate": 6.766436336301608e-05, "loss": 1.9845, "step": 34330 }, { "epoch": 2.97, "learning_rate": 6.766219498091824e-05, "loss": 2.0393, "step": 34340 }, { "epoch": 2.97, "learning_rate": 6.76600265988204e-05, "loss": 2.1584, "step": 34350 }, { "epoch": 2.98, "learning_rate": 6.765785821672256e-05, "loss": 1.7408, "step": 34360 }, { "epoch": 2.98, "learning_rate": 6.765568983462471e-05, "loss": 1.8955, "step": 34370 }, { "epoch": 2.98, "learning_rate": 6.765352145252688e-05, "loss": 1.9113, "step": 34380 }, { "epoch": 2.98, "learning_rate": 6.765135307042905e-05, "loss": 1.8005, "step": 34390 }, { "epoch": 2.98, "learning_rate": 6.764918468833121e-05, "loss": 1.8502, "step": 34400 }, { "epoch": 2.98, "learning_rate": 6.764701630623338e-05, "loss": 1.9689, "step": 34410 }, { "epoch": 2.98, "learning_rate": 6.764484792413553e-05, "loss": 1.7533, "step": 34420 }, { "epoch": 2.98, "learning_rate": 6.764267954203769e-05, "loss": 1.8029, "step": 34430 }, { "epoch": 2.98, "learning_rate": 6.764051115993985e-05, "loss": 1.6704, "step": 34440 }, { "epoch": 2.98, "learning_rate": 6.763834277784202e-05, "loss": 1.9642, "step": 34450 }, { "epoch": 2.98, "learning_rate": 6.763617439574418e-05, "loss": 1.664, "step": 34460 }, { "epoch": 2.99, "learning_rate": 6.763400601364635e-05, "loss": 2.0006, "step": 34470 }, { "epoch": 2.99, "learning_rate": 6.76318376315485e-05, "loss": 1.9012, "step": 34480 }, { "epoch": 2.99, "learning_rate": 6.762966924945067e-05, "loss": 1.8194, "step": 34490 }, { "epoch": 2.99, "learning_rate": 6.762750086735282e-05, "loss": 1.8724, "step": 34500 }, { "epoch": 2.99, "learning_rate": 6.762533248525499e-05, "loss": 2.0062, "step": 34510 }, { "epoch": 2.99, "learning_rate": 6.762316410315716e-05, "loss": 1.992, "step": 34520 }, { "epoch": 2.99, "learning_rate": 6.762099572105932e-05, "loss": 1.759, "step": 34530 }, { "epoch": 2.99, "learning_rate": 6.761882733896149e-05, "loss": 2.0305, "step": 34540 }, { "epoch": 2.99, "learning_rate": 6.761665895686364e-05, "loss": 2.018, "step": 34550 }, { "epoch": 2.99, "learning_rate": 6.761449057476581e-05, "loss": 2.0703, "step": 34560 }, { "epoch": 2.99, "learning_rate": 6.761232219266797e-05, "loss": 1.9093, "step": 34570 }, { "epoch": 2.99, "learning_rate": 6.761015381057013e-05, "loss": 1.8708, "step": 34580 }, { "epoch": 3.0, "learning_rate": 6.76079854284723e-05, "loss": 1.9557, "step": 34590 }, { "epoch": 3.0, "learning_rate": 6.760581704637446e-05, "loss": 2.1368, "step": 34600 }, { "epoch": 3.0, "learning_rate": 6.760364866427663e-05, "loss": 1.8934, "step": 34610 }, { "epoch": 3.0, "learning_rate": 6.760148028217878e-05, "loss": 1.9739, "step": 34620 }, { "epoch": 3.0, "learning_rate": 6.759931190008095e-05, "loss": 1.9072, "step": 34630 }, { "epoch": 3.0, "eval_cer": 0.5511462383315715, "eval_em": 0.0020629190304280558, "eval_f1": 0.0020629190304280558, "eval_loss": 1.7403308153152466, "eval_runtime": 240.3561, "eval_samples_per_second": 8.067, "eval_steps_per_second": 1.011, "eval_wer": 0.9979370809695719, "step": 34639 }, { "epoch": 3.0, "learning_rate": 6.759714351798311e-05, "loss": 1.8577, "step": 34640 }, { "epoch": 3.0, "learning_rate": 6.759497513588526e-05, "loss": 1.8454, "step": 34650 }, { "epoch": 3.0, "learning_rate": 6.759280675378743e-05, "loss": 1.7975, "step": 34660 }, { "epoch": 3.0, "learning_rate": 6.75906383716896e-05, "loss": 1.626, "step": 34670 }, { "epoch": 3.0, "learning_rate": 6.758846998959176e-05, "loss": 1.8414, "step": 34680 }, { "epoch": 3.0, "learning_rate": 6.758630160749392e-05, "loss": 1.7231, "step": 34690 }, { "epoch": 3.01, "learning_rate": 6.758413322539608e-05, "loss": 1.9313, "step": 34700 }, { "epoch": 3.01, "learning_rate": 6.758196484329825e-05, "loss": 1.9576, "step": 34710 }, { "epoch": 3.01, "learning_rate": 6.757979646120042e-05, "loss": 1.8071, "step": 34720 }, { "epoch": 3.01, "learning_rate": 6.757762807910257e-05, "loss": 1.7357, "step": 34730 }, { "epoch": 3.01, "learning_rate": 6.757545969700474e-05, "loss": 2.0024, "step": 34740 }, { "epoch": 3.01, "learning_rate": 6.75732913149069e-05, "loss": 1.6577, "step": 34750 }, { "epoch": 3.01, "learning_rate": 6.757112293280905e-05, "loss": 1.93, "step": 34760 }, { "epoch": 3.01, "learning_rate": 6.756895455071122e-05, "loss": 1.8075, "step": 34770 }, { "epoch": 3.01, "learning_rate": 6.756678616861339e-05, "loss": 1.8348, "step": 34780 }, { "epoch": 3.01, "learning_rate": 6.756461778651555e-05, "loss": 1.7727, "step": 34790 }, { "epoch": 3.01, "learning_rate": 6.756244940441772e-05, "loss": 1.7737, "step": 34800 }, { "epoch": 3.01, "learning_rate": 6.756028102231987e-05, "loss": 1.8638, "step": 34810 }, { "epoch": 3.02, "learning_rate": 6.755811264022204e-05, "loss": 1.6902, "step": 34820 }, { "epoch": 3.02, "learning_rate": 6.755594425812419e-05, "loss": 1.7643, "step": 34830 }, { "epoch": 3.02, "learning_rate": 6.755377587602636e-05, "loss": 1.7895, "step": 34840 }, { "epoch": 3.02, "learning_rate": 6.755160749392852e-05, "loss": 1.7701, "step": 34850 }, { "epoch": 3.02, "learning_rate": 6.754943911183069e-05, "loss": 1.7843, "step": 34860 }, { "epoch": 3.02, "learning_rate": 6.754727072973286e-05, "loss": 1.9272, "step": 34870 }, { "epoch": 3.02, "learning_rate": 6.754510234763501e-05, "loss": 2.0765, "step": 34880 }, { "epoch": 3.02, "learning_rate": 6.754293396553718e-05, "loss": 1.8424, "step": 34890 }, { "epoch": 3.02, "learning_rate": 6.754076558343933e-05, "loss": 1.9377, "step": 34900 }, { "epoch": 3.02, "learning_rate": 6.75385972013415e-05, "loss": 1.711, "step": 34910 }, { "epoch": 3.02, "learning_rate": 6.753642881924366e-05, "loss": 1.6671, "step": 34920 }, { "epoch": 3.03, "learning_rate": 6.753426043714583e-05, "loss": 1.797, "step": 34930 }, { "epoch": 3.03, "learning_rate": 6.7532092055048e-05, "loss": 1.8938, "step": 34940 }, { "epoch": 3.03, "learning_rate": 6.752992367295015e-05, "loss": 1.7474, "step": 34950 }, { "epoch": 3.03, "learning_rate": 6.752775529085231e-05, "loss": 1.728, "step": 34960 }, { "epoch": 3.03, "learning_rate": 6.752558690875447e-05, "loss": 1.7947, "step": 34970 }, { "epoch": 3.03, "learning_rate": 6.752341852665663e-05, "loss": 1.6905, "step": 34980 }, { "epoch": 3.03, "learning_rate": 6.75212501445588e-05, "loss": 1.7673, "step": 34990 }, { "epoch": 3.03, "learning_rate": 6.751908176246097e-05, "loss": 2.1311, "step": 35000 }, { "epoch": 3.03, "learning_rate": 6.751691338036313e-05, "loss": 1.8721, "step": 35010 }, { "epoch": 3.03, "learning_rate": 6.751474499826529e-05, "loss": 1.9263, "step": 35020 }, { "epoch": 3.03, "learning_rate": 6.751257661616745e-05, "loss": 1.7543, "step": 35030 }, { "epoch": 3.03, "learning_rate": 6.75104082340696e-05, "loss": 1.8126, "step": 35040 }, { "epoch": 3.04, "learning_rate": 6.750823985197177e-05, "loss": 1.8339, "step": 35050 }, { "epoch": 3.04, "learning_rate": 6.750607146987394e-05, "loss": 1.784, "step": 35060 }, { "epoch": 3.04, "learning_rate": 6.75039030877761e-05, "loss": 1.8965, "step": 35070 }, { "epoch": 3.04, "learning_rate": 6.750173470567827e-05, "loss": 1.8798, "step": 35080 }, { "epoch": 3.04, "learning_rate": 6.749956632358042e-05, "loss": 1.8888, "step": 35090 }, { "epoch": 3.04, "learning_rate": 6.749739794148259e-05, "loss": 1.8261, "step": 35100 }, { "epoch": 3.04, "learning_rate": 6.749522955938476e-05, "loss": 1.6118, "step": 35110 }, { "epoch": 3.04, "learning_rate": 6.749306117728691e-05, "loss": 1.8157, "step": 35120 }, { "epoch": 3.04, "learning_rate": 6.749089279518908e-05, "loss": 1.7038, "step": 35130 }, { "epoch": 3.04, "learning_rate": 6.748872441309124e-05, "loss": 1.6188, "step": 35140 }, { "epoch": 3.04, "learning_rate": 6.748655603099341e-05, "loss": 1.8806, "step": 35150 }, { "epoch": 3.05, "learning_rate": 6.748438764889556e-05, "loss": 1.7544, "step": 35160 }, { "epoch": 3.05, "learning_rate": 6.748221926679773e-05, "loss": 1.9428, "step": 35170 }, { "epoch": 3.05, "learning_rate": 6.74800508846999e-05, "loss": 1.7598, "step": 35180 }, { "epoch": 3.05, "learning_rate": 6.747788250260205e-05, "loss": 1.8454, "step": 35190 }, { "epoch": 3.05, "learning_rate": 6.747571412050421e-05, "loss": 1.7629, "step": 35200 }, { "epoch": 3.05, "learning_rate": 6.747354573840638e-05, "loss": 1.8194, "step": 35210 }, { "epoch": 3.05, "learning_rate": 6.747137735630855e-05, "loss": 1.7194, "step": 35220 }, { "epoch": 3.05, "learning_rate": 6.74692089742107e-05, "loss": 1.8079, "step": 35230 }, { "epoch": 3.05, "learning_rate": 6.746704059211286e-05, "loss": 1.9258, "step": 35240 }, { "epoch": 3.05, "learning_rate": 6.746487221001503e-05, "loss": 1.746, "step": 35250 }, { "epoch": 3.05, "learning_rate": 6.74627038279172e-05, "loss": 1.7631, "step": 35260 }, { "epoch": 3.05, "learning_rate": 6.746053544581935e-05, "loss": 1.7969, "step": 35270 }, { "epoch": 3.06, "learning_rate": 6.745836706372152e-05, "loss": 1.6737, "step": 35280 }, { "epoch": 3.06, "learning_rate": 6.745619868162368e-05, "loss": 1.6556, "step": 35290 }, { "epoch": 3.06, "learning_rate": 6.745403029952584e-05, "loss": 1.8229, "step": 35300 }, { "epoch": 3.06, "learning_rate": 6.7451861917428e-05, "loss": 1.8115, "step": 35310 }, { "epoch": 3.06, "learning_rate": 6.744969353533017e-05, "loss": 1.6857, "step": 35320 }, { "epoch": 3.06, "learning_rate": 6.744752515323234e-05, "loss": 1.7432, "step": 35330 }, { "epoch": 3.06, "learning_rate": 6.74453567711345e-05, "loss": 1.8032, "step": 35340 }, { "epoch": 3.06, "learning_rate": 6.744318838903665e-05, "loss": 1.877, "step": 35350 }, { "epoch": 3.06, "learning_rate": 6.744102000693882e-05, "loss": 1.8619, "step": 35360 }, { "epoch": 3.06, "learning_rate": 6.743885162484097e-05, "loss": 1.6632, "step": 35370 }, { "epoch": 3.06, "learning_rate": 6.743668324274314e-05, "loss": 1.8147, "step": 35380 }, { "epoch": 3.06, "learning_rate": 6.74345148606453e-05, "loss": 1.7452, "step": 35390 }, { "epoch": 3.07, "learning_rate": 6.743234647854747e-05, "loss": 1.9787, "step": 35400 }, { "epoch": 3.07, "learning_rate": 6.743017809644964e-05, "loss": 1.8363, "step": 35410 }, { "epoch": 3.07, "learning_rate": 6.742800971435179e-05, "loss": 1.8754, "step": 35420 }, { "epoch": 3.07, "learning_rate": 6.742584133225396e-05, "loss": 1.7867, "step": 35430 }, { "epoch": 3.07, "learning_rate": 6.742367295015611e-05, "loss": 1.7389, "step": 35440 }, { "epoch": 3.07, "learning_rate": 6.742150456805828e-05, "loss": 1.8057, "step": 35450 }, { "epoch": 3.07, "learning_rate": 6.741933618596044e-05, "loss": 1.7641, "step": 35460 }, { "epoch": 3.07, "learning_rate": 6.741716780386261e-05, "loss": 1.8402, "step": 35470 }, { "epoch": 3.07, "learning_rate": 6.741499942176478e-05, "loss": 1.8223, "step": 35480 }, { "epoch": 3.07, "learning_rate": 6.741283103966693e-05, "loss": 1.6456, "step": 35490 }, { "epoch": 3.07, "learning_rate": 6.741066265756908e-05, "loss": 1.9462, "step": 35500 }, { "epoch": 3.08, "learning_rate": 6.740849427547125e-05, "loss": 2.1034, "step": 35510 }, { "epoch": 3.08, "learning_rate": 6.740632589337342e-05, "loss": 1.8337, "step": 35520 }, { "epoch": 3.08, "learning_rate": 6.740415751127558e-05, "loss": 1.7221, "step": 35530 }, { "epoch": 3.08, "learning_rate": 6.740198912917775e-05, "loss": 2.0583, "step": 35540 }, { "epoch": 3.08, "learning_rate": 6.73998207470799e-05, "loss": 1.8458, "step": 35550 }, { "epoch": 3.08, "learning_rate": 6.739765236498207e-05, "loss": 1.6984, "step": 35560 }, { "epoch": 3.08, "learning_rate": 6.739548398288423e-05, "loss": 1.7372, "step": 35570 }, { "epoch": 3.08, "learning_rate": 6.739331560078639e-05, "loss": 1.9675, "step": 35580 }, { "epoch": 3.08, "learning_rate": 6.739114721868855e-05, "loss": 1.7484, "step": 35590 }, { "epoch": 3.08, "learning_rate": 6.738897883659072e-05, "loss": 1.8323, "step": 35600 }, { "epoch": 3.08, "learning_rate": 6.738681045449289e-05, "loss": 1.9693, "step": 35610 }, { "epoch": 3.08, "learning_rate": 6.738464207239504e-05, "loss": 1.8202, "step": 35620 }, { "epoch": 3.09, "learning_rate": 6.73824736902972e-05, "loss": 2.0424, "step": 35630 }, { "epoch": 3.09, "learning_rate": 6.738030530819937e-05, "loss": 1.8854, "step": 35640 }, { "epoch": 3.09, "learning_rate": 6.737813692610154e-05, "loss": 1.7419, "step": 35650 }, { "epoch": 3.09, "learning_rate": 6.737596854400369e-05, "loss": 1.8072, "step": 35660 }, { "epoch": 3.09, "learning_rate": 6.737380016190586e-05, "loss": 2.0335, "step": 35670 }, { "epoch": 3.09, "learning_rate": 6.737163177980802e-05, "loss": 1.7631, "step": 35680 }, { "epoch": 3.09, "learning_rate": 6.736946339771018e-05, "loss": 1.8303, "step": 35690 }, { "epoch": 3.09, "learning_rate": 6.736729501561234e-05, "loss": 1.8914, "step": 35700 }, { "epoch": 3.09, "learning_rate": 6.736512663351451e-05, "loss": 1.9506, "step": 35710 }, { "epoch": 3.09, "learning_rate": 6.736295825141668e-05, "loss": 1.5094, "step": 35720 }, { "epoch": 3.09, "learning_rate": 6.736078986931883e-05, "loss": 1.8642, "step": 35730 }, { "epoch": 3.1, "learning_rate": 6.7358621487221e-05, "loss": 1.7116, "step": 35740 }, { "epoch": 3.1, "learning_rate": 6.735645310512316e-05, "loss": 1.783, "step": 35750 }, { "epoch": 3.1, "learning_rate": 6.735428472302531e-05, "loss": 1.8314, "step": 35760 }, { "epoch": 3.1, "learning_rate": 6.735211634092748e-05, "loss": 1.7084, "step": 35770 }, { "epoch": 3.1, "learning_rate": 6.734994795882965e-05, "loss": 1.825, "step": 35780 }, { "epoch": 3.1, "learning_rate": 6.734777957673181e-05, "loss": 1.8795, "step": 35790 }, { "epoch": 3.1, "learning_rate": 6.734561119463398e-05, "loss": 1.8373, "step": 35800 }, { "epoch": 3.1, "learning_rate": 6.734344281253613e-05, "loss": 1.643, "step": 35810 }, { "epoch": 3.1, "learning_rate": 6.73412744304383e-05, "loss": 1.8758, "step": 35820 }, { "epoch": 3.1, "learning_rate": 6.733910604834045e-05, "loss": 1.7692, "step": 35830 }, { "epoch": 3.1, "learning_rate": 6.733693766624262e-05, "loss": 1.9863, "step": 35840 }, { "epoch": 3.1, "learning_rate": 6.733476928414478e-05, "loss": 1.9928, "step": 35850 }, { "epoch": 3.11, "learning_rate": 6.733260090204695e-05, "loss": 1.9967, "step": 35860 }, { "epoch": 3.11, "learning_rate": 6.733043251994912e-05, "loss": 1.9318, "step": 35870 }, { "epoch": 3.11, "learning_rate": 6.732826413785127e-05, "loss": 1.7066, "step": 35880 }, { "epoch": 3.11, "learning_rate": 6.732609575575344e-05, "loss": 1.9287, "step": 35890 }, { "epoch": 3.11, "learning_rate": 6.732392737365559e-05, "loss": 1.8668, "step": 35900 }, { "epoch": 3.11, "learning_rate": 6.732175899155776e-05, "loss": 1.7976, "step": 35910 }, { "epoch": 3.11, "learning_rate": 6.731959060945992e-05, "loss": 1.7541, "step": 35920 }, { "epoch": 3.11, "learning_rate": 6.731742222736209e-05, "loss": 2.0105, "step": 35930 }, { "epoch": 3.11, "learning_rate": 6.731525384526425e-05, "loss": 1.6828, "step": 35940 }, { "epoch": 3.11, "learning_rate": 6.731308546316641e-05, "loss": 1.9681, "step": 35950 }, { "epoch": 3.11, "learning_rate": 6.731091708106857e-05, "loss": 1.958, "step": 35960 }, { "epoch": 3.12, "learning_rate": 6.730874869897073e-05, "loss": 1.8493, "step": 35970 }, { "epoch": 3.12, "learning_rate": 6.73065803168729e-05, "loss": 1.8704, "step": 35980 }, { "epoch": 3.12, "learning_rate": 6.730441193477506e-05, "loss": 1.9831, "step": 35990 }, { "epoch": 3.12, "learning_rate": 6.730224355267723e-05, "loss": 1.7661, "step": 36000 }, { "epoch": 3.12, "learning_rate": 6.730007517057939e-05, "loss": 2.0181, "step": 36010 }, { "epoch": 3.12, "learning_rate": 6.729790678848155e-05, "loss": 1.8614, "step": 36020 }, { "epoch": 3.12, "learning_rate": 6.729573840638371e-05, "loss": 1.7132, "step": 36030 }, { "epoch": 3.12, "learning_rate": 6.729357002428586e-05, "loss": 1.6576, "step": 36040 }, { "epoch": 3.12, "learning_rate": 6.729140164218803e-05, "loss": 1.8178, "step": 36050 }, { "epoch": 3.12, "learning_rate": 6.72892332600902e-05, "loss": 1.8822, "step": 36060 }, { "epoch": 3.12, "learning_rate": 6.728706487799236e-05, "loss": 1.7964, "step": 36070 }, { "epoch": 3.12, "learning_rate": 6.728489649589453e-05, "loss": 1.7089, "step": 36080 }, { "epoch": 3.13, "learning_rate": 6.728272811379668e-05, "loss": 1.8008, "step": 36090 }, { "epoch": 3.13, "learning_rate": 6.728055973169885e-05, "loss": 1.8574, "step": 36100 }, { "epoch": 3.13, "learning_rate": 6.727839134960102e-05, "loss": 1.4958, "step": 36110 }, { "epoch": 3.13, "learning_rate": 6.727622296750317e-05, "loss": 1.8623, "step": 36120 }, { "epoch": 3.13, "learning_rate": 6.727405458540534e-05, "loss": 1.8798, "step": 36130 }, { "epoch": 3.13, "learning_rate": 6.72718862033075e-05, "loss": 1.9455, "step": 36140 }, { "epoch": 3.13, "learning_rate": 6.726971782120967e-05, "loss": 1.7726, "step": 36150 }, { "epoch": 3.13, "learning_rate": 6.726754943911182e-05, "loss": 2.1204, "step": 36160 }, { "epoch": 3.13, "learning_rate": 6.726538105701399e-05, "loss": 1.7845, "step": 36170 }, { "epoch": 3.13, "learning_rate": 6.726321267491615e-05, "loss": 1.8514, "step": 36180 }, { "epoch": 3.13, "learning_rate": 6.72610442928183e-05, "loss": 1.7198, "step": 36190 }, { "epoch": 3.14, "learning_rate": 6.725887591072047e-05, "loss": 1.8658, "step": 36200 }, { "epoch": 3.14, "learning_rate": 6.725670752862264e-05, "loss": 1.767, "step": 36210 }, { "epoch": 3.14, "learning_rate": 6.72545391465248e-05, "loss": 1.8381, "step": 36220 }, { "epoch": 3.14, "learning_rate": 6.725237076442696e-05, "loss": 1.8114, "step": 36230 }, { "epoch": 3.14, "learning_rate": 6.725020238232912e-05, "loss": 1.8873, "step": 36240 }, { "epoch": 3.14, "learning_rate": 6.724803400023129e-05, "loss": 1.723, "step": 36250 }, { "epoch": 3.14, "learning_rate": 6.724586561813346e-05, "loss": 1.9327, "step": 36260 }, { "epoch": 3.14, "learning_rate": 6.724369723603561e-05, "loss": 1.9288, "step": 36270 }, { "epoch": 3.14, "learning_rate": 6.724152885393778e-05, "loss": 1.8609, "step": 36280 }, { "epoch": 3.14, "learning_rate": 6.723936047183994e-05, "loss": 1.8437, "step": 36290 }, { "epoch": 3.14, "learning_rate": 6.72371920897421e-05, "loss": 1.9432, "step": 36300 }, { "epoch": 3.14, "learning_rate": 6.723502370764426e-05, "loss": 1.7088, "step": 36310 }, { "epoch": 3.15, "learning_rate": 6.723285532554643e-05, "loss": 1.8753, "step": 36320 }, { "epoch": 3.15, "learning_rate": 6.72306869434486e-05, "loss": 2.2045, "step": 36330 }, { "epoch": 3.15, "learning_rate": 6.722851856135076e-05, "loss": 1.7805, "step": 36340 }, { "epoch": 3.15, "learning_rate": 6.722635017925291e-05, "loss": 2.0924, "step": 36350 }, { "epoch": 3.15, "learning_rate": 6.722418179715508e-05, "loss": 1.8695, "step": 36360 }, { "epoch": 3.15, "learning_rate": 6.722201341505723e-05, "loss": 1.6462, "step": 36370 }, { "epoch": 3.15, "learning_rate": 6.72198450329594e-05, "loss": 1.8666, "step": 36380 }, { "epoch": 3.15, "learning_rate": 6.721767665086157e-05, "loss": 1.8324, "step": 36390 }, { "epoch": 3.15, "learning_rate": 6.721550826876373e-05, "loss": 1.7132, "step": 36400 }, { "epoch": 3.15, "learning_rate": 6.72133398866659e-05, "loss": 1.8045, "step": 36410 }, { "epoch": 3.15, "learning_rate": 6.721117150456805e-05, "loss": 1.9862, "step": 36420 }, { "epoch": 3.16, "learning_rate": 6.720900312247022e-05, "loss": 1.7953, "step": 36430 }, { "epoch": 3.16, "learning_rate": 6.720683474037237e-05, "loss": 1.8483, "step": 36440 }, { "epoch": 3.16, "learning_rate": 6.720466635827454e-05, "loss": 1.8001, "step": 36450 }, { "epoch": 3.16, "learning_rate": 6.72024979761767e-05, "loss": 1.79, "step": 36460 }, { "epoch": 3.16, "learning_rate": 6.720032959407887e-05, "loss": 1.9303, "step": 36470 }, { "epoch": 3.16, "learning_rate": 6.719816121198104e-05, "loss": 1.6917, "step": 36480 }, { "epoch": 3.16, "learning_rate": 6.719599282988319e-05, "loss": 1.9263, "step": 36490 }, { "epoch": 3.16, "learning_rate": 6.719382444778536e-05, "loss": 1.7769, "step": 36500 }, { "epoch": 3.16, "learning_rate": 6.719165606568751e-05, "loss": 1.7916, "step": 36510 }, { "epoch": 3.16, "learning_rate": 6.718948768358968e-05, "loss": 1.6873, "step": 36520 }, { "epoch": 3.16, "learning_rate": 6.718731930149184e-05, "loss": 1.9485, "step": 36530 }, { "epoch": 3.16, "learning_rate": 6.718515091939401e-05, "loss": 1.8498, "step": 36540 }, { "epoch": 3.17, "learning_rate": 6.718298253729617e-05, "loss": 1.7916, "step": 36550 }, { "epoch": 3.17, "learning_rate": 6.718081415519833e-05, "loss": 2.0576, "step": 36560 }, { "epoch": 3.17, "learning_rate": 6.71786457731005e-05, "loss": 1.8187, "step": 36570 }, { "epoch": 3.17, "learning_rate": 6.717647739100265e-05, "loss": 1.5816, "step": 36580 }, { "epoch": 3.17, "learning_rate": 6.717430900890481e-05, "loss": 1.7816, "step": 36590 }, { "epoch": 3.17, "learning_rate": 6.717214062680698e-05, "loss": 1.8354, "step": 36600 }, { "epoch": 3.17, "learning_rate": 6.716997224470915e-05, "loss": 1.856, "step": 36610 }, { "epoch": 3.17, "learning_rate": 6.71678038626113e-05, "loss": 1.7631, "step": 36620 }, { "epoch": 3.17, "learning_rate": 6.716563548051346e-05, "loss": 1.8465, "step": 36630 }, { "epoch": 3.17, "learning_rate": 6.716346709841563e-05, "loss": 1.7722, "step": 36640 }, { "epoch": 3.17, "learning_rate": 6.71612987163178e-05, "loss": 1.6979, "step": 36650 }, { "epoch": 3.17, "learning_rate": 6.715913033421995e-05, "loss": 2.0649, "step": 36660 }, { "epoch": 3.18, "learning_rate": 6.715696195212212e-05, "loss": 1.8285, "step": 36670 }, { "epoch": 3.18, "learning_rate": 6.715479357002428e-05, "loss": 1.8246, "step": 36680 }, { "epoch": 3.18, "learning_rate": 6.715262518792644e-05, "loss": 1.8262, "step": 36690 }, { "epoch": 3.18, "learning_rate": 6.71504568058286e-05, "loss": 1.6407, "step": 36700 }, { "epoch": 3.18, "learning_rate": 6.714828842373077e-05, "loss": 1.6951, "step": 36710 }, { "epoch": 3.18, "learning_rate": 6.714612004163294e-05, "loss": 1.8145, "step": 36720 }, { "epoch": 3.18, "learning_rate": 6.714395165953509e-05, "loss": 1.8308, "step": 36730 }, { "epoch": 3.18, "learning_rate": 6.714178327743725e-05, "loss": 1.798, "step": 36740 }, { "epoch": 3.18, "learning_rate": 6.713961489533942e-05, "loss": 1.8084, "step": 36750 }, { "epoch": 3.18, "learning_rate": 6.713744651324157e-05, "loss": 1.7501, "step": 36760 }, { "epoch": 3.18, "learning_rate": 6.713527813114374e-05, "loss": 1.8716, "step": 36770 }, { "epoch": 3.19, "learning_rate": 6.71331097490459e-05, "loss": 1.6893, "step": 36780 }, { "epoch": 3.19, "learning_rate": 6.713094136694807e-05, "loss": 1.9129, "step": 36790 }, { "epoch": 3.19, "learning_rate": 6.712877298485024e-05, "loss": 1.8874, "step": 36800 }, { "epoch": 3.19, "learning_rate": 6.712660460275239e-05, "loss": 1.8044, "step": 36810 }, { "epoch": 3.19, "learning_rate": 6.712443622065456e-05, "loss": 1.7263, "step": 36820 }, { "epoch": 3.19, "learning_rate": 6.712226783855671e-05, "loss": 1.8018, "step": 36830 }, { "epoch": 3.19, "learning_rate": 6.712009945645888e-05, "loss": 1.4175, "step": 36840 }, { "epoch": 3.19, "learning_rate": 6.711793107436104e-05, "loss": 1.9582, "step": 36850 }, { "epoch": 3.19, "learning_rate": 6.711576269226321e-05, "loss": 1.6993, "step": 36860 }, { "epoch": 3.19, "learning_rate": 6.711359431016538e-05, "loss": 1.7732, "step": 36870 }, { "epoch": 3.19, "learning_rate": 6.711142592806753e-05, "loss": 1.8229, "step": 36880 }, { "epoch": 3.19, "learning_rate": 6.71092575459697e-05, "loss": 1.8875, "step": 36890 }, { "epoch": 3.2, "learning_rate": 6.710708916387185e-05, "loss": 2.0139, "step": 36900 }, { "epoch": 3.2, "learning_rate": 6.710492078177402e-05, "loss": 1.8543, "step": 36910 }, { "epoch": 3.2, "learning_rate": 6.710275239967618e-05, "loss": 1.8426, "step": 36920 }, { "epoch": 3.2, "learning_rate": 6.710058401757835e-05, "loss": 1.7513, "step": 36930 }, { "epoch": 3.2, "learning_rate": 6.709841563548051e-05, "loss": 1.8233, "step": 36940 }, { "epoch": 3.2, "learning_rate": 6.709624725338267e-05, "loss": 2.0926, "step": 36950 }, { "epoch": 3.2, "learning_rate": 6.709407887128483e-05, "loss": 1.9535, "step": 36960 }, { "epoch": 3.2, "learning_rate": 6.709191048918699e-05, "loss": 1.7587, "step": 36970 }, { "epoch": 3.2, "learning_rate": 6.708974210708915e-05, "loss": 1.8919, "step": 36980 }, { "epoch": 3.2, "learning_rate": 6.708757372499132e-05, "loss": 1.9022, "step": 36990 }, { "epoch": 3.2, "learning_rate": 6.708540534289349e-05, "loss": 1.7369, "step": 37000 }, { "epoch": 3.21, "learning_rate": 6.708323696079565e-05, "loss": 2.042, "step": 37010 }, { "epoch": 3.21, "learning_rate": 6.70810685786978e-05, "loss": 1.8031, "step": 37020 }, { "epoch": 3.21, "learning_rate": 6.707890019659997e-05, "loss": 1.819, "step": 37030 }, { "epoch": 3.21, "learning_rate": 6.707673181450212e-05, "loss": 1.54, "step": 37040 }, { "epoch": 3.21, "learning_rate": 6.707456343240429e-05, "loss": 1.8063, "step": 37050 }, { "epoch": 3.21, "learning_rate": 6.707239505030646e-05, "loss": 1.9419, "step": 37060 }, { "epoch": 3.21, "learning_rate": 6.707022666820862e-05, "loss": 1.7825, "step": 37070 }, { "epoch": 3.21, "learning_rate": 6.706805828611079e-05, "loss": 2.1392, "step": 37080 }, { "epoch": 3.21, "learning_rate": 6.706588990401294e-05, "loss": 2.1167, "step": 37090 }, { "epoch": 3.21, "learning_rate": 6.706372152191511e-05, "loss": 1.8041, "step": 37100 }, { "epoch": 3.21, "learning_rate": 6.706155313981728e-05, "loss": 1.9085, "step": 37110 }, { "epoch": 3.21, "learning_rate": 6.705938475771943e-05, "loss": 1.7917, "step": 37120 }, { "epoch": 3.22, "learning_rate": 6.70572163756216e-05, "loss": 1.8205, "step": 37130 }, { "epoch": 3.22, "learning_rate": 6.705504799352376e-05, "loss": 1.8543, "step": 37140 }, { "epoch": 3.22, "learning_rate": 6.705287961142593e-05, "loss": 1.787, "step": 37150 }, { "epoch": 3.22, "learning_rate": 6.705071122932808e-05, "loss": 1.8503, "step": 37160 }, { "epoch": 3.22, "learning_rate": 6.704854284723025e-05, "loss": 1.7961, "step": 37170 }, { "epoch": 3.22, "learning_rate": 6.704637446513241e-05, "loss": 1.8275, "step": 37180 }, { "epoch": 3.22, "learning_rate": 6.704420608303458e-05, "loss": 1.9502, "step": 37190 }, { "epoch": 3.22, "learning_rate": 6.704203770093673e-05, "loss": 1.8733, "step": 37200 }, { "epoch": 3.22, "learning_rate": 6.70398693188389e-05, "loss": 1.9804, "step": 37210 }, { "epoch": 3.22, "learning_rate": 6.703770093674107e-05, "loss": 1.8079, "step": 37220 }, { "epoch": 3.22, "learning_rate": 6.703553255464322e-05, "loss": 1.8198, "step": 37230 }, { "epoch": 3.23, "learning_rate": 6.703336417254538e-05, "loss": 1.888, "step": 37240 }, { "epoch": 3.23, "learning_rate": 6.703119579044755e-05, "loss": 1.7822, "step": 37250 }, { "epoch": 3.23, "learning_rate": 6.702902740834972e-05, "loss": 1.7103, "step": 37260 }, { "epoch": 3.23, "learning_rate": 6.702685902625187e-05, "loss": 1.8025, "step": 37270 }, { "epoch": 3.23, "learning_rate": 6.702469064415404e-05, "loss": 1.8828, "step": 37280 }, { "epoch": 3.23, "learning_rate": 6.70225222620562e-05, "loss": 1.9018, "step": 37290 }, { "epoch": 3.23, "learning_rate": 6.702035387995836e-05, "loss": 1.9385, "step": 37300 }, { "epoch": 3.23, "learning_rate": 6.701818549786052e-05, "loss": 1.7926, "step": 37310 }, { "epoch": 3.23, "learning_rate": 6.701601711576269e-05, "loss": 1.7121, "step": 37320 }, { "epoch": 3.23, "learning_rate": 6.701384873366485e-05, "loss": 1.9078, "step": 37330 }, { "epoch": 3.23, "learning_rate": 6.701168035156702e-05, "loss": 1.6901, "step": 37340 }, { "epoch": 3.23, "learning_rate": 6.700951196946917e-05, "loss": 1.8091, "step": 37350 }, { "epoch": 3.24, "learning_rate": 6.700734358737134e-05, "loss": 1.9151, "step": 37360 }, { "epoch": 3.24, "learning_rate": 6.70051752052735e-05, "loss": 1.7211, "step": 37370 }, { "epoch": 3.24, "learning_rate": 6.700300682317566e-05, "loss": 1.7047, "step": 37380 }, { "epoch": 3.24, "learning_rate": 6.700083844107783e-05, "loss": 1.8549, "step": 37390 }, { "epoch": 3.24, "learning_rate": 6.699867005897999e-05, "loss": 1.9091, "step": 37400 }, { "epoch": 3.24, "learning_rate": 6.699650167688216e-05, "loss": 1.6583, "step": 37410 }, { "epoch": 3.24, "learning_rate": 6.699433329478431e-05, "loss": 1.6413, "step": 37420 }, { "epoch": 3.24, "learning_rate": 6.699216491268648e-05, "loss": 1.9955, "step": 37430 }, { "epoch": 3.24, "learning_rate": 6.698999653058863e-05, "loss": 1.7946, "step": 37440 }, { "epoch": 3.24, "learning_rate": 6.69878281484908e-05, "loss": 1.9088, "step": 37450 }, { "epoch": 3.24, "learning_rate": 6.698565976639296e-05, "loss": 1.8335, "step": 37460 }, { "epoch": 3.25, "learning_rate": 6.698349138429513e-05, "loss": 1.7749, "step": 37470 }, { "epoch": 3.25, "learning_rate": 6.69813230021973e-05, "loss": 1.857, "step": 37480 }, { "epoch": 3.25, "learning_rate": 6.697915462009945e-05, "loss": 1.7605, "step": 37490 }, { "epoch": 3.25, "learning_rate": 6.697698623800162e-05, "loss": 1.6956, "step": 37500 }, { "epoch": 3.25, "learning_rate": 6.697481785590377e-05, "loss": 1.8136, "step": 37510 }, { "epoch": 3.25, "learning_rate": 6.697264947380593e-05, "loss": 1.7968, "step": 37520 }, { "epoch": 3.25, "learning_rate": 6.69704810917081e-05, "loss": 1.6848, "step": 37530 }, { "epoch": 3.25, "learning_rate": 6.696831270961027e-05, "loss": 1.7786, "step": 37540 }, { "epoch": 3.25, "learning_rate": 6.696614432751243e-05, "loss": 1.8484, "step": 37550 }, { "epoch": 3.25, "learning_rate": 6.696397594541459e-05, "loss": 1.8026, "step": 37560 }, { "epoch": 3.25, "learning_rate": 6.696180756331675e-05, "loss": 1.9148, "step": 37570 }, { "epoch": 3.25, "learning_rate": 6.69596391812189e-05, "loss": 1.8003, "step": 37580 }, { "epoch": 3.26, "learning_rate": 6.695747079912107e-05, "loss": 1.8185, "step": 37590 }, { "epoch": 3.26, "learning_rate": 6.695530241702324e-05, "loss": 1.81, "step": 37600 }, { "epoch": 3.26, "learning_rate": 6.69531340349254e-05, "loss": 1.8711, "step": 37610 }, { "epoch": 3.26, "learning_rate": 6.695096565282757e-05, "loss": 1.9463, "step": 37620 }, { "epoch": 3.26, "learning_rate": 6.694879727072972e-05, "loss": 1.7542, "step": 37630 }, { "epoch": 3.26, "learning_rate": 6.694662888863189e-05, "loss": 1.7706, "step": 37640 }, { "epoch": 3.26, "learning_rate": 6.694446050653406e-05, "loss": 1.7738, "step": 37650 }, { "epoch": 3.26, "learning_rate": 6.694229212443621e-05, "loss": 1.6181, "step": 37660 }, { "epoch": 3.26, "learning_rate": 6.694012374233838e-05, "loss": 1.6963, "step": 37670 }, { "epoch": 3.26, "learning_rate": 6.693795536024054e-05, "loss": 1.8362, "step": 37680 }, { "epoch": 3.26, "learning_rate": 6.69357869781427e-05, "loss": 1.7669, "step": 37690 }, { "epoch": 3.27, "learning_rate": 6.693361859604486e-05, "loss": 1.8222, "step": 37700 }, { "epoch": 3.27, "learning_rate": 6.693145021394703e-05, "loss": 1.5991, "step": 37710 }, { "epoch": 3.27, "learning_rate": 6.69292818318492e-05, "loss": 1.937, "step": 37720 }, { "epoch": 3.27, "learning_rate": 6.692711344975135e-05, "loss": 2.053, "step": 37730 }, { "epoch": 3.27, "learning_rate": 6.692494506765351e-05, "loss": 1.8725, "step": 37740 }, { "epoch": 3.27, "learning_rate": 6.692277668555568e-05, "loss": 1.9369, "step": 37750 }, { "epoch": 3.27, "learning_rate": 6.692060830345783e-05, "loss": 2.0243, "step": 37760 }, { "epoch": 3.27, "learning_rate": 6.691843992136e-05, "loss": 1.808, "step": 37770 }, { "epoch": 3.27, "learning_rate": 6.691627153926217e-05, "loss": 1.7915, "step": 37780 }, { "epoch": 3.27, "learning_rate": 6.691410315716433e-05, "loss": 1.9098, "step": 37790 }, { "epoch": 3.27, "learning_rate": 6.69119347750665e-05, "loss": 1.8906, "step": 37800 }, { "epoch": 3.27, "learning_rate": 6.690976639296865e-05, "loss": 1.6489, "step": 37810 }, { "epoch": 3.28, "learning_rate": 6.690759801087082e-05, "loss": 1.8657, "step": 37820 }, { "epoch": 3.28, "learning_rate": 6.690542962877297e-05, "loss": 1.7552, "step": 37830 }, { "epoch": 3.28, "learning_rate": 6.690326124667514e-05, "loss": 1.9044, "step": 37840 }, { "epoch": 3.28, "learning_rate": 6.69010928645773e-05, "loss": 1.852, "step": 37850 }, { "epoch": 3.28, "learning_rate": 6.689892448247947e-05, "loss": 1.8012, "step": 37860 }, { "epoch": 3.28, "learning_rate": 6.689675610038164e-05, "loss": 1.7634, "step": 37870 }, { "epoch": 3.28, "learning_rate": 6.689458771828379e-05, "loss": 1.9867, "step": 37880 }, { "epoch": 3.28, "learning_rate": 6.689241933618596e-05, "loss": 1.6123, "step": 37890 }, { "epoch": 3.28, "learning_rate": 6.689025095408811e-05, "loss": 2.0446, "step": 37900 }, { "epoch": 3.28, "learning_rate": 6.688808257199028e-05, "loss": 1.9598, "step": 37910 }, { "epoch": 3.28, "learning_rate": 6.688591418989244e-05, "loss": 1.8536, "step": 37920 }, { "epoch": 3.28, "learning_rate": 6.688374580779461e-05, "loss": 1.946, "step": 37930 }, { "epoch": 3.29, "learning_rate": 6.688157742569677e-05, "loss": 1.873, "step": 37940 }, { "epoch": 3.29, "learning_rate": 6.687940904359893e-05, "loss": 1.7294, "step": 37950 }, { "epoch": 3.29, "learning_rate": 6.68772406615011e-05, "loss": 1.8108, "step": 37960 }, { "epoch": 3.29, "learning_rate": 6.687507227940325e-05, "loss": 1.9739, "step": 37970 }, { "epoch": 3.29, "learning_rate": 6.687290389730541e-05, "loss": 1.7495, "step": 37980 }, { "epoch": 3.29, "learning_rate": 6.687073551520758e-05, "loss": 1.7445, "step": 37990 }, { "epoch": 3.29, "learning_rate": 6.686856713310975e-05, "loss": 1.8754, "step": 38000 }, { "epoch": 3.29, "learning_rate": 6.686639875101191e-05, "loss": 2.0043, "step": 38010 }, { "epoch": 3.29, "learning_rate": 6.686423036891406e-05, "loss": 1.9392, "step": 38020 }, { "epoch": 3.29, "learning_rate": 6.686206198681623e-05, "loss": 1.7246, "step": 38030 }, { "epoch": 3.29, "learning_rate": 6.685989360471838e-05, "loss": 1.633, "step": 38040 }, { "epoch": 3.3, "learning_rate": 6.685772522262055e-05, "loss": 1.9293, "step": 38050 }, { "epoch": 3.3, "learning_rate": 6.685555684052272e-05, "loss": 1.7844, "step": 38060 }, { "epoch": 3.3, "learning_rate": 6.685338845842488e-05, "loss": 1.7821, "step": 38070 }, { "epoch": 3.3, "learning_rate": 6.685122007632705e-05, "loss": 1.8229, "step": 38080 }, { "epoch": 3.3, "learning_rate": 6.68490516942292e-05, "loss": 1.7283, "step": 38090 }, { "epoch": 3.3, "learning_rate": 6.684688331213137e-05, "loss": 1.7852, "step": 38100 }, { "epoch": 3.3, "learning_rate": 6.684471493003354e-05, "loss": 1.9859, "step": 38110 }, { "epoch": 3.3, "learning_rate": 6.684254654793569e-05, "loss": 1.9148, "step": 38120 }, { "epoch": 3.3, "learning_rate": 6.684037816583785e-05, "loss": 1.7264, "step": 38130 }, { "epoch": 3.3, "learning_rate": 6.683820978374002e-05, "loss": 1.7515, "step": 38140 }, { "epoch": 3.3, "learning_rate": 6.683604140164219e-05, "loss": 1.9221, "step": 38150 }, { "epoch": 3.3, "learning_rate": 6.683387301954434e-05, "loss": 2.0855, "step": 38160 }, { "epoch": 3.31, "learning_rate": 6.68317046374465e-05, "loss": 1.9527, "step": 38170 }, { "epoch": 3.31, "learning_rate": 6.682953625534867e-05, "loss": 1.7404, "step": 38180 }, { "epoch": 3.31, "learning_rate": 6.682736787325084e-05, "loss": 1.9033, "step": 38190 }, { "epoch": 3.31, "learning_rate": 6.682519949115299e-05, "loss": 1.6783, "step": 38200 }, { "epoch": 3.31, "learning_rate": 6.682303110905516e-05, "loss": 1.7686, "step": 38210 }, { "epoch": 3.31, "learning_rate": 6.682086272695732e-05, "loss": 1.8035, "step": 38220 }, { "epoch": 3.31, "learning_rate": 6.681869434485948e-05, "loss": 1.7046, "step": 38230 }, { "epoch": 3.31, "learning_rate": 6.681652596276164e-05, "loss": 1.9766, "step": 38240 }, { "epoch": 3.31, "learning_rate": 6.681435758066381e-05, "loss": 1.7282, "step": 38250 }, { "epoch": 3.31, "learning_rate": 6.681218919856598e-05, "loss": 1.8329, "step": 38260 }, { "epoch": 3.31, "learning_rate": 6.681002081646813e-05, "loss": 1.8715, "step": 38270 }, { "epoch": 3.32, "learning_rate": 6.68078524343703e-05, "loss": 1.6649, "step": 38280 }, { "epoch": 3.32, "learning_rate": 6.680568405227246e-05, "loss": 1.7869, "step": 38290 }, { "epoch": 3.32, "learning_rate": 6.680351567017462e-05, "loss": 1.7734, "step": 38300 }, { "epoch": 3.32, "learning_rate": 6.680134728807678e-05, "loss": 1.9198, "step": 38310 }, { "epoch": 3.32, "learning_rate": 6.679917890597895e-05, "loss": 1.8407, "step": 38320 }, { "epoch": 3.32, "learning_rate": 6.679701052388111e-05, "loss": 1.912, "step": 38330 }, { "epoch": 3.32, "learning_rate": 6.679484214178328e-05, "loss": 1.6794, "step": 38340 }, { "epoch": 3.32, "learning_rate": 6.679267375968543e-05, "loss": 1.821, "step": 38350 }, { "epoch": 3.32, "learning_rate": 6.67905053775876e-05, "loss": 1.8266, "step": 38360 }, { "epoch": 3.32, "learning_rate": 6.678833699548975e-05, "loss": 1.8329, "step": 38370 }, { "epoch": 3.32, "learning_rate": 6.678616861339192e-05, "loss": 1.8265, "step": 38380 }, { "epoch": 3.32, "learning_rate": 6.678400023129409e-05, "loss": 1.7165, "step": 38390 }, { "epoch": 3.33, "learning_rate": 6.678183184919625e-05, "loss": 2.1261, "step": 38400 }, { "epoch": 3.33, "learning_rate": 6.677966346709842e-05, "loss": 1.9469, "step": 38410 }, { "epoch": 3.33, "learning_rate": 6.677749508500057e-05, "loss": 1.7402, "step": 38420 }, { "epoch": 3.33, "learning_rate": 6.677532670290274e-05, "loss": 1.7103, "step": 38430 }, { "epoch": 3.33, "learning_rate": 6.677315832080489e-05, "loss": 1.7304, "step": 38440 }, { "epoch": 3.33, "learning_rate": 6.677098993870706e-05, "loss": 1.8562, "step": 38450 }, { "epoch": 3.33, "learning_rate": 6.676882155660922e-05, "loss": 1.6829, "step": 38460 }, { "epoch": 3.33, "learning_rate": 6.676665317451139e-05, "loss": 1.8294, "step": 38470 }, { "epoch": 3.33, "learning_rate": 6.676448479241356e-05, "loss": 1.9829, "step": 38480 }, { "epoch": 3.33, "learning_rate": 6.676231641031571e-05, "loss": 1.7837, "step": 38490 }, { "epoch": 3.33, "learning_rate": 6.676014802821788e-05, "loss": 1.7608, "step": 38500 }, { "epoch": 3.34, "learning_rate": 6.675797964612003e-05, "loss": 1.5681, "step": 38510 }, { "epoch": 3.34, "learning_rate": 6.67558112640222e-05, "loss": 2.0128, "step": 38520 }, { "epoch": 3.34, "learning_rate": 6.675364288192436e-05, "loss": 1.7796, "step": 38530 }, { "epoch": 3.34, "learning_rate": 6.675147449982653e-05, "loss": 1.9524, "step": 38540 }, { "epoch": 3.34, "learning_rate": 6.67493061177287e-05, "loss": 1.8972, "step": 38550 }, { "epoch": 3.34, "learning_rate": 6.674713773563085e-05, "loss": 1.5698, "step": 38560 }, { "epoch": 3.34, "learning_rate": 6.674496935353301e-05, "loss": 1.6147, "step": 38570 }, { "epoch": 3.34, "learning_rate": 6.674280097143517e-05, "loss": 1.685, "step": 38580 }, { "epoch": 3.34, "learning_rate": 6.674063258933733e-05, "loss": 1.7384, "step": 38590 }, { "epoch": 3.34, "learning_rate": 6.67384642072395e-05, "loss": 2.0152, "step": 38600 }, { "epoch": 3.34, "learning_rate": 6.673629582514167e-05, "loss": 1.732, "step": 38610 }, { "epoch": 3.34, "learning_rate": 6.673412744304383e-05, "loss": 1.7238, "step": 38620 }, { "epoch": 3.35, "learning_rate": 6.673195906094598e-05, "loss": 1.6291, "step": 38630 }, { "epoch": 3.35, "learning_rate": 6.672979067884815e-05, "loss": 1.8597, "step": 38640 }, { "epoch": 3.35, "learning_rate": 6.672762229675032e-05, "loss": 1.835, "step": 38650 }, { "epoch": 3.35, "learning_rate": 6.672545391465247e-05, "loss": 1.659, "step": 38660 }, { "epoch": 3.35, "learning_rate": 6.672328553255464e-05, "loss": 1.8486, "step": 38670 }, { "epoch": 3.35, "learning_rate": 6.67211171504568e-05, "loss": 1.7363, "step": 38680 }, { "epoch": 3.35, "learning_rate": 6.671894876835897e-05, "loss": 1.8041, "step": 38690 }, { "epoch": 3.35, "learning_rate": 6.671678038626112e-05, "loss": 1.7985, "step": 38700 }, { "epoch": 3.35, "learning_rate": 6.671461200416329e-05, "loss": 1.9503, "step": 38710 }, { "epoch": 3.35, "learning_rate": 6.671244362206545e-05, "loss": 1.9205, "step": 38720 }, { "epoch": 3.35, "learning_rate": 6.671027523996762e-05, "loss": 1.7458, "step": 38730 }, { "epoch": 3.36, "learning_rate": 6.670810685786977e-05, "loss": 1.8823, "step": 38740 }, { "epoch": 3.36, "learning_rate": 6.670593847577194e-05, "loss": 1.8654, "step": 38750 }, { "epoch": 3.36, "learning_rate": 6.670377009367409e-05, "loss": 2.0078, "step": 38760 }, { "epoch": 3.36, "learning_rate": 6.670160171157626e-05, "loss": 1.8021, "step": 38770 }, { "epoch": 3.36, "learning_rate": 6.669943332947843e-05, "loss": 1.9125, "step": 38780 }, { "epoch": 3.36, "learning_rate": 6.669726494738059e-05, "loss": 1.9016, "step": 38790 }, { "epoch": 3.36, "learning_rate": 6.669509656528276e-05, "loss": 1.9097, "step": 38800 }, { "epoch": 3.36, "learning_rate": 6.669292818318491e-05, "loss": 1.9256, "step": 38810 }, { "epoch": 3.36, "learning_rate": 6.669075980108708e-05, "loss": 1.7102, "step": 38820 }, { "epoch": 3.36, "learning_rate": 6.668859141898923e-05, "loss": 1.9532, "step": 38830 }, { "epoch": 3.36, "learning_rate": 6.66864230368914e-05, "loss": 1.7335, "step": 38840 }, { "epoch": 3.36, "learning_rate": 6.668425465479356e-05, "loss": 1.7888, "step": 38850 }, { "epoch": 3.37, "learning_rate": 6.668208627269573e-05, "loss": 1.8094, "step": 38860 }, { "epoch": 3.37, "learning_rate": 6.66799178905979e-05, "loss": 1.821, "step": 38870 }, { "epoch": 3.37, "learning_rate": 6.667774950850005e-05, "loss": 1.7999, "step": 38880 }, { "epoch": 3.37, "learning_rate": 6.667558112640222e-05, "loss": 1.5641, "step": 38890 }, { "epoch": 3.37, "learning_rate": 6.667341274430437e-05, "loss": 1.9579, "step": 38900 }, { "epoch": 3.37, "learning_rate": 6.667124436220653e-05, "loss": 1.8014, "step": 38910 }, { "epoch": 3.37, "learning_rate": 6.66690759801087e-05, "loss": 1.8946, "step": 38920 }, { "epoch": 3.37, "learning_rate": 6.666690759801087e-05, "loss": 1.827, "step": 38930 }, { "epoch": 3.37, "learning_rate": 6.666473921591303e-05, "loss": 1.7345, "step": 38940 }, { "epoch": 3.37, "learning_rate": 6.666257083381519e-05, "loss": 2.0348, "step": 38950 }, { "epoch": 3.37, "learning_rate": 6.666040245171735e-05, "loss": 1.9739, "step": 38960 }, { "epoch": 3.38, "learning_rate": 6.66582340696195e-05, "loss": 1.6616, "step": 38970 }, { "epoch": 3.38, "learning_rate": 6.665606568752167e-05, "loss": 1.7473, "step": 38980 }, { "epoch": 3.38, "learning_rate": 6.665389730542384e-05, "loss": 1.8519, "step": 38990 }, { "epoch": 3.38, "learning_rate": 6.6651728923326e-05, "loss": 1.8913, "step": 39000 }, { "epoch": 3.38, "learning_rate": 6.664956054122817e-05, "loss": 1.8257, "step": 39010 }, { "epoch": 3.38, "learning_rate": 6.664739215913032e-05, "loss": 1.8574, "step": 39020 }, { "epoch": 3.38, "learning_rate": 6.664522377703249e-05, "loss": 1.9153, "step": 39030 }, { "epoch": 3.38, "learning_rate": 6.664305539493464e-05, "loss": 1.7816, "step": 39040 }, { "epoch": 3.38, "learning_rate": 6.664088701283681e-05, "loss": 1.7972, "step": 39050 }, { "epoch": 3.38, "learning_rate": 6.663871863073898e-05, "loss": 1.6971, "step": 39060 }, { "epoch": 3.38, "learning_rate": 6.663655024864114e-05, "loss": 1.9088, "step": 39070 }, { "epoch": 3.38, "learning_rate": 6.663438186654331e-05, "loss": 1.8125, "step": 39080 }, { "epoch": 3.39, "learning_rate": 6.663221348444546e-05, "loss": 1.9587, "step": 39090 }, { "epoch": 3.39, "learning_rate": 6.663004510234763e-05, "loss": 1.817, "step": 39100 }, { "epoch": 3.39, "learning_rate": 6.66278767202498e-05, "loss": 2.0472, "step": 39110 }, { "epoch": 3.39, "learning_rate": 6.662570833815195e-05, "loss": 1.772, "step": 39120 }, { "epoch": 3.39, "learning_rate": 6.662353995605411e-05, "loss": 1.9203, "step": 39130 }, { "epoch": 3.39, "learning_rate": 6.662137157395628e-05, "loss": 1.8294, "step": 39140 }, { "epoch": 3.39, "learning_rate": 6.661920319185845e-05, "loss": 1.7634, "step": 39150 }, { "epoch": 3.39, "learning_rate": 6.66170348097606e-05, "loss": 1.9298, "step": 39160 }, { "epoch": 3.39, "learning_rate": 6.661486642766277e-05, "loss": 2.0076, "step": 39170 }, { "epoch": 3.39, "learning_rate": 6.661269804556493e-05, "loss": 1.9853, "step": 39180 }, { "epoch": 3.39, "learning_rate": 6.66105296634671e-05, "loss": 1.9505, "step": 39190 }, { "epoch": 3.39, "learning_rate": 6.660836128136925e-05, "loss": 1.8919, "step": 39200 }, { "epoch": 3.4, "learning_rate": 6.660619289927142e-05, "loss": 1.778, "step": 39210 }, { "epoch": 3.4, "learning_rate": 6.660402451717358e-05, "loss": 1.5876, "step": 39220 }, { "epoch": 3.4, "learning_rate": 6.660185613507574e-05, "loss": 1.7546, "step": 39230 }, { "epoch": 3.4, "learning_rate": 6.65996877529779e-05, "loss": 1.8625, "step": 39240 }, { "epoch": 3.4, "learning_rate": 6.659751937088007e-05, "loss": 2.0864, "step": 39250 }, { "epoch": 3.4, "learning_rate": 6.659535098878224e-05, "loss": 1.8618, "step": 39260 }, { "epoch": 3.4, "learning_rate": 6.659318260668439e-05, "loss": 1.8474, "step": 39270 }, { "epoch": 3.4, "learning_rate": 6.659101422458656e-05, "loss": 1.9273, "step": 39280 }, { "epoch": 3.4, "learning_rate": 6.658884584248872e-05, "loss": 1.7269, "step": 39290 }, { "epoch": 3.4, "learning_rate": 6.658667746039087e-05, "loss": 1.8899, "step": 39300 }, { "epoch": 3.4, "learning_rate": 6.658450907829304e-05, "loss": 1.8096, "step": 39310 }, { "epoch": 3.41, "learning_rate": 6.658234069619521e-05, "loss": 1.8904, "step": 39320 }, { "epoch": 3.41, "learning_rate": 6.658017231409737e-05, "loss": 1.7429, "step": 39330 }, { "epoch": 3.41, "learning_rate": 6.657800393199954e-05, "loss": 1.511, "step": 39340 }, { "epoch": 3.41, "learning_rate": 6.65758355499017e-05, "loss": 1.8019, "step": 39350 }, { "epoch": 3.41, "learning_rate": 6.657366716780386e-05, "loss": 1.7953, "step": 39360 }, { "epoch": 3.41, "learning_rate": 6.657149878570601e-05, "loss": 1.755, "step": 39370 }, { "epoch": 3.41, "learning_rate": 6.656933040360818e-05, "loss": 1.5382, "step": 39380 }, { "epoch": 3.41, "learning_rate": 6.656716202151035e-05, "loss": 1.798, "step": 39390 }, { "epoch": 3.41, "learning_rate": 6.656499363941251e-05, "loss": 1.863, "step": 39400 }, { "epoch": 3.41, "learning_rate": 6.656282525731468e-05, "loss": 1.9513, "step": 39410 }, { "epoch": 3.41, "learning_rate": 6.656065687521683e-05, "loss": 1.8573, "step": 39420 }, { "epoch": 3.41, "learning_rate": 6.6558488493119e-05, "loss": 1.8625, "step": 39430 }, { "epoch": 3.42, "learning_rate": 6.655632011102115e-05, "loss": 1.8324, "step": 39440 }, { "epoch": 3.42, "learning_rate": 6.655415172892332e-05, "loss": 1.7113, "step": 39450 }, { "epoch": 3.42, "learning_rate": 6.655198334682548e-05, "loss": 1.6266, "step": 39460 }, { "epoch": 3.42, "learning_rate": 6.654981496472765e-05, "loss": 1.7778, "step": 39470 }, { "epoch": 3.42, "learning_rate": 6.654764658262982e-05, "loss": 1.7591, "step": 39480 }, { "epoch": 3.42, "learning_rate": 6.654547820053197e-05, "loss": 1.8469, "step": 39490 }, { "epoch": 3.42, "learning_rate": 6.654330981843414e-05, "loss": 1.8817, "step": 39500 }, { "epoch": 3.42, "learning_rate": 6.654114143633629e-05, "loss": 1.8045, "step": 39510 }, { "epoch": 3.42, "learning_rate": 6.653897305423845e-05, "loss": 1.675, "step": 39520 }, { "epoch": 3.42, "learning_rate": 6.653680467214062e-05, "loss": 1.8611, "step": 39530 }, { "epoch": 3.42, "learning_rate": 6.653463629004279e-05, "loss": 1.756, "step": 39540 }, { "epoch": 3.43, "learning_rate": 6.653246790794495e-05, "loss": 1.9351, "step": 39550 }, { "epoch": 3.43, "learning_rate": 6.65302995258471e-05, "loss": 1.7336, "step": 39560 }, { "epoch": 3.43, "learning_rate": 6.652813114374927e-05, "loss": 1.9431, "step": 39570 }, { "epoch": 3.43, "learning_rate": 6.652596276165143e-05, "loss": 1.6206, "step": 39580 }, { "epoch": 3.43, "learning_rate": 6.652379437955359e-05, "loss": 1.7451, "step": 39590 }, { "epoch": 3.43, "learning_rate": 6.652162599745576e-05, "loss": 1.8403, "step": 39600 }, { "epoch": 3.43, "learning_rate": 6.651945761535792e-05, "loss": 1.5818, "step": 39610 }, { "epoch": 3.43, "learning_rate": 6.651728923326009e-05, "loss": 1.8792, "step": 39620 }, { "epoch": 3.43, "learning_rate": 6.651512085116224e-05, "loss": 1.7188, "step": 39630 }, { "epoch": 3.43, "learning_rate": 6.651295246906441e-05, "loss": 1.8466, "step": 39640 }, { "epoch": 3.43, "learning_rate": 6.651078408696658e-05, "loss": 1.7909, "step": 39650 }, { "epoch": 3.43, "learning_rate": 6.650861570486873e-05, "loss": 1.7561, "step": 39660 }, { "epoch": 3.44, "learning_rate": 6.65064473227709e-05, "loss": 1.8379, "step": 39670 }, { "epoch": 3.44, "learning_rate": 6.650427894067306e-05, "loss": 1.7834, "step": 39680 }, { "epoch": 3.44, "learning_rate": 6.650211055857523e-05, "loss": 1.6889, "step": 39690 }, { "epoch": 3.44, "learning_rate": 6.649994217647738e-05, "loss": 1.7206, "step": 39700 }, { "epoch": 3.44, "learning_rate": 6.649777379437955e-05, "loss": 1.8402, "step": 39710 }, { "epoch": 3.44, "learning_rate": 6.649560541228171e-05, "loss": 1.8228, "step": 39720 }, { "epoch": 3.44, "learning_rate": 6.649343703018388e-05, "loss": 1.8835, "step": 39730 }, { "epoch": 3.44, "learning_rate": 6.649126864808603e-05, "loss": 1.7648, "step": 39740 }, { "epoch": 3.44, "learning_rate": 6.64891002659882e-05, "loss": 1.7528, "step": 39750 }, { "epoch": 3.44, "learning_rate": 6.648693188389037e-05, "loss": 1.8283, "step": 39760 }, { "epoch": 3.44, "learning_rate": 6.648476350179252e-05, "loss": 1.5386, "step": 39770 }, { "epoch": 3.45, "learning_rate": 6.648259511969469e-05, "loss": 1.5563, "step": 39780 }, { "epoch": 3.45, "learning_rate": 6.648042673759685e-05, "loss": 1.9289, "step": 39790 }, { "epoch": 3.45, "learning_rate": 6.647825835549902e-05, "loss": 1.7432, "step": 39800 }, { "epoch": 3.45, "learning_rate": 6.647608997340117e-05, "loss": 1.6722, "step": 39810 }, { "epoch": 3.45, "learning_rate": 6.647392159130334e-05, "loss": 2.0097, "step": 39820 }, { "epoch": 3.45, "learning_rate": 6.647175320920549e-05, "loss": 1.8539, "step": 39830 }, { "epoch": 3.45, "learning_rate": 6.646958482710766e-05, "loss": 1.926, "step": 39840 }, { "epoch": 3.45, "learning_rate": 6.646741644500982e-05, "loss": 2.0767, "step": 39850 }, { "epoch": 3.45, "learning_rate": 6.646524806291199e-05, "loss": 1.8801, "step": 39860 }, { "epoch": 3.45, "learning_rate": 6.646307968081416e-05, "loss": 1.8003, "step": 39870 }, { "epoch": 3.45, "learning_rate": 6.646091129871631e-05, "loss": 1.7949, "step": 39880 }, { "epoch": 3.45, "learning_rate": 6.645874291661848e-05, "loss": 1.9011, "step": 39890 }, { "epoch": 3.46, "learning_rate": 6.645657453452063e-05, "loss": 1.6586, "step": 39900 }, { "epoch": 3.46, "learning_rate": 6.64544061524228e-05, "loss": 2.032, "step": 39910 }, { "epoch": 3.46, "learning_rate": 6.645223777032496e-05, "loss": 1.6162, "step": 39920 }, { "epoch": 3.46, "learning_rate": 6.645006938822713e-05, "loss": 1.6494, "step": 39930 }, { "epoch": 3.46, "learning_rate": 6.64479010061293e-05, "loss": 1.7803, "step": 39940 }, { "epoch": 3.46, "learning_rate": 6.644573262403145e-05, "loss": 1.7746, "step": 39950 }, { "epoch": 3.46, "learning_rate": 6.644356424193361e-05, "loss": 1.8839, "step": 39960 }, { "epoch": 3.46, "learning_rate": 6.644139585983577e-05, "loss": 1.6365, "step": 39970 }, { "epoch": 3.46, "learning_rate": 6.643922747773793e-05, "loss": 1.8232, "step": 39980 }, { "epoch": 3.46, "learning_rate": 6.64370590956401e-05, "loss": 2.061, "step": 39990 }, { "epoch": 3.46, "learning_rate": 6.643489071354226e-05, "loss": 1.6607, "step": 40000 }, { "epoch": 3.47, "learning_rate": 6.643272233144443e-05, "loss": 1.9167, "step": 40010 }, { "epoch": 3.47, "learning_rate": 6.643055394934658e-05, "loss": 1.7382, "step": 40020 }, { "epoch": 3.47, "learning_rate": 6.642838556724875e-05, "loss": 1.8327, "step": 40030 }, { "epoch": 3.47, "learning_rate": 6.64262171851509e-05, "loss": 1.6206, "step": 40040 }, { "epoch": 3.47, "learning_rate": 6.642404880305307e-05, "loss": 1.7958, "step": 40050 }, { "epoch": 3.47, "learning_rate": 6.642188042095524e-05, "loss": 1.7037, "step": 40060 }, { "epoch": 3.47, "learning_rate": 6.64197120388574e-05, "loss": 1.8416, "step": 40070 }, { "epoch": 3.47, "learning_rate": 6.641754365675957e-05, "loss": 1.6559, "step": 40080 }, { "epoch": 3.47, "learning_rate": 6.641537527466172e-05, "loss": 1.7391, "step": 40090 }, { "epoch": 3.47, "learning_rate": 6.641320689256389e-05, "loss": 1.7191, "step": 40100 }, { "epoch": 3.47, "learning_rate": 6.641103851046605e-05, "loss": 1.7289, "step": 40110 }, { "epoch": 3.47, "learning_rate": 6.640887012836821e-05, "loss": 1.7393, "step": 40120 }, { "epoch": 3.48, "learning_rate": 6.640670174627037e-05, "loss": 1.742, "step": 40130 }, { "epoch": 3.48, "learning_rate": 6.640453336417254e-05, "loss": 1.7431, "step": 40140 }, { "epoch": 3.48, "learning_rate": 6.64023649820747e-05, "loss": 1.7636, "step": 40150 }, { "epoch": 3.48, "learning_rate": 6.640019659997686e-05, "loss": 1.849, "step": 40160 }, { "epoch": 3.48, "learning_rate": 6.639802821787903e-05, "loss": 1.8059, "step": 40170 }, { "epoch": 3.48, "learning_rate": 6.639585983578119e-05, "loss": 1.8159, "step": 40180 }, { "epoch": 3.48, "learning_rate": 6.639369145368336e-05, "loss": 1.8918, "step": 40190 }, { "epoch": 3.48, "learning_rate": 6.639152307158551e-05, "loss": 1.6237, "step": 40200 }, { "epoch": 3.48, "learning_rate": 6.638935468948768e-05, "loss": 1.6225, "step": 40210 }, { "epoch": 3.48, "learning_rate": 6.638718630738984e-05, "loss": 1.9073, "step": 40220 }, { "epoch": 3.48, "learning_rate": 6.6385017925292e-05, "loss": 1.8879, "step": 40230 }, { "epoch": 3.49, "learning_rate": 6.638284954319416e-05, "loss": 1.8838, "step": 40240 }, { "epoch": 3.49, "learning_rate": 6.638068116109633e-05, "loss": 1.9273, "step": 40250 }, { "epoch": 3.49, "learning_rate": 6.63785127789985e-05, "loss": 2.0033, "step": 40260 }, { "epoch": 3.49, "learning_rate": 6.637634439690066e-05, "loss": 1.8219, "step": 40270 }, { "epoch": 3.49, "learning_rate": 6.637417601480282e-05, "loss": 1.7566, "step": 40280 }, { "epoch": 3.49, "learning_rate": 6.637200763270498e-05, "loss": 1.7675, "step": 40290 }, { "epoch": 3.49, "learning_rate": 6.636983925060713e-05, "loss": 1.8478, "step": 40300 }, { "epoch": 3.49, "learning_rate": 6.63676708685093e-05, "loss": 1.8458, "step": 40310 }, { "epoch": 3.49, "learning_rate": 6.636550248641147e-05, "loss": 1.704, "step": 40320 }, { "epoch": 3.49, "learning_rate": 6.636333410431363e-05, "loss": 1.8496, "step": 40330 }, { "epoch": 3.49, "learning_rate": 6.63611657222158e-05, "loss": 1.8707, "step": 40340 }, { "epoch": 3.49, "learning_rate": 6.635899734011795e-05, "loss": 1.7873, "step": 40350 }, { "epoch": 3.5, "learning_rate": 6.635682895802012e-05, "loss": 1.5335, "step": 40360 }, { "epoch": 3.5, "learning_rate": 6.635466057592227e-05, "loss": 1.8792, "step": 40370 }, { "epoch": 3.5, "learning_rate": 6.635249219382444e-05, "loss": 1.7828, "step": 40380 }, { "epoch": 3.5, "learning_rate": 6.63503238117266e-05, "loss": 1.8054, "step": 40390 }, { "epoch": 3.5, "learning_rate": 6.634815542962877e-05, "loss": 1.8593, "step": 40400 }, { "epoch": 3.5, "learning_rate": 6.634598704753094e-05, "loss": 1.688, "step": 40410 }, { "epoch": 3.5, "learning_rate": 6.634381866543309e-05, "loss": 2.0136, "step": 40420 }, { "epoch": 3.5, "learning_rate": 6.634165028333526e-05, "loss": 1.6253, "step": 40430 }, { "epoch": 3.5, "learning_rate": 6.633948190123741e-05, "loss": 1.8641, "step": 40440 }, { "epoch": 3.5, "learning_rate": 6.633731351913958e-05, "loss": 1.847, "step": 40450 }, { "epoch": 3.5, "learning_rate": 6.633514513704174e-05, "loss": 1.8123, "step": 40460 }, { "epoch": 3.5, "learning_rate": 6.633297675494391e-05, "loss": 2.0923, "step": 40470 }, { "epoch": 3.51, "learning_rate": 6.633080837284608e-05, "loss": 1.7078, "step": 40480 }, { "epoch": 3.51, "learning_rate": 6.632863999074823e-05, "loss": 1.7824, "step": 40490 }, { "epoch": 3.51, "learning_rate": 6.63264716086504e-05, "loss": 1.7165, "step": 40500 }, { "epoch": 3.51, "learning_rate": 6.632430322655255e-05, "loss": 1.7103, "step": 40510 }, { "epoch": 3.51, "learning_rate": 6.632213484445471e-05, "loss": 1.6649, "step": 40520 }, { "epoch": 3.51, "learning_rate": 6.631996646235688e-05, "loss": 1.5983, "step": 40530 }, { "epoch": 3.51, "learning_rate": 6.631779808025905e-05, "loss": 1.8113, "step": 40540 }, { "epoch": 3.51, "learning_rate": 6.631562969816121e-05, "loss": 1.7766, "step": 40550 }, { "epoch": 3.51, "learning_rate": 6.631346131606337e-05, "loss": 1.8803, "step": 40560 }, { "epoch": 3.51, "learning_rate": 6.631129293396553e-05, "loss": 1.7025, "step": 40570 }, { "epoch": 3.51, "learning_rate": 6.630912455186769e-05, "loss": 1.6784, "step": 40580 }, { "epoch": 3.52, "learning_rate": 6.630695616976985e-05, "loss": 1.7605, "step": 40590 }, { "epoch": 3.52, "learning_rate": 6.630478778767202e-05, "loss": 2.0852, "step": 40600 }, { "epoch": 3.52, "learning_rate": 6.630261940557418e-05, "loss": 1.8311, "step": 40610 }, { "epoch": 3.52, "learning_rate": 6.630045102347635e-05, "loss": 1.8917, "step": 40620 }, { "epoch": 3.52, "learning_rate": 6.62982826413785e-05, "loss": 1.7181, "step": 40630 }, { "epoch": 3.52, "learning_rate": 6.629611425928067e-05, "loss": 1.8582, "step": 40640 }, { "epoch": 3.52, "learning_rate": 6.629394587718284e-05, "loss": 1.8025, "step": 40650 }, { "epoch": 3.52, "learning_rate": 6.629177749508499e-05, "loss": 1.8485, "step": 40660 }, { "epoch": 3.52, "learning_rate": 6.628960911298716e-05, "loss": 1.8036, "step": 40670 }, { "epoch": 3.52, "learning_rate": 6.628744073088932e-05, "loss": 1.6723, "step": 40680 }, { "epoch": 3.52, "learning_rate": 6.628527234879149e-05, "loss": 1.9171, "step": 40690 }, { "epoch": 3.52, "learning_rate": 6.628310396669364e-05, "loss": 1.5662, "step": 40700 }, { "epoch": 3.53, "learning_rate": 6.628093558459581e-05, "loss": 1.8177, "step": 40710 }, { "epoch": 3.53, "learning_rate": 6.627876720249797e-05, "loss": 1.8329, "step": 40720 }, { "epoch": 3.53, "learning_rate": 6.627659882040014e-05, "loss": 1.7942, "step": 40730 }, { "epoch": 3.53, "learning_rate": 6.62744304383023e-05, "loss": 1.659, "step": 40740 }, { "epoch": 3.53, "learning_rate": 6.627226205620446e-05, "loss": 1.7787, "step": 40750 }, { "epoch": 3.53, "learning_rate": 6.627009367410663e-05, "loss": 1.932, "step": 40760 }, { "epoch": 3.53, "learning_rate": 6.626792529200878e-05, "loss": 1.8027, "step": 40770 }, { "epoch": 3.53, "learning_rate": 6.626575690991095e-05, "loss": 1.718, "step": 40780 }, { "epoch": 3.53, "learning_rate": 6.626358852781311e-05, "loss": 1.7301, "step": 40790 }, { "epoch": 3.53, "learning_rate": 6.626142014571528e-05, "loss": 1.6968, "step": 40800 }, { "epoch": 3.53, "learning_rate": 6.625925176361743e-05, "loss": 1.7148, "step": 40810 }, { "epoch": 3.54, "learning_rate": 6.62570833815196e-05, "loss": 1.6976, "step": 40820 }, { "epoch": 3.54, "learning_rate": 6.625491499942176e-05, "loss": 1.8351, "step": 40830 }, { "epoch": 3.54, "learning_rate": 6.625274661732392e-05, "loss": 1.7818, "step": 40840 }, { "epoch": 3.54, "learning_rate": 6.625057823522608e-05, "loss": 1.9155, "step": 40850 }, { "epoch": 3.54, "learning_rate": 6.624840985312825e-05, "loss": 1.7023, "step": 40860 }, { "epoch": 3.54, "learning_rate": 6.624624147103042e-05, "loss": 1.6566, "step": 40870 }, { "epoch": 3.54, "learning_rate": 6.624407308893258e-05, "loss": 1.9718, "step": 40880 }, { "epoch": 3.54, "learning_rate": 6.624190470683473e-05, "loss": 1.8251, "step": 40890 }, { "epoch": 3.54, "learning_rate": 6.623973632473689e-05, "loss": 1.8786, "step": 40900 }, { "epoch": 3.54, "learning_rate": 6.623756794263905e-05, "loss": 1.8129, "step": 40910 }, { "epoch": 3.54, "learning_rate": 6.623539956054122e-05, "loss": 1.8424, "step": 40920 }, { "epoch": 3.54, "learning_rate": 6.623323117844339e-05, "loss": 1.7492, "step": 40930 }, { "epoch": 3.55, "learning_rate": 6.623106279634555e-05, "loss": 1.8506, "step": 40940 }, { "epoch": 3.55, "learning_rate": 6.62288944142477e-05, "loss": 1.8959, "step": 40950 }, { "epoch": 3.55, "learning_rate": 6.622672603214987e-05, "loss": 1.9622, "step": 40960 }, { "epoch": 3.55, "learning_rate": 6.622455765005203e-05, "loss": 1.8411, "step": 40970 }, { "epoch": 3.55, "learning_rate": 6.622238926795419e-05, "loss": 1.7241, "step": 40980 }, { "epoch": 3.55, "learning_rate": 6.622022088585636e-05, "loss": 1.7529, "step": 40990 }, { "epoch": 3.55, "learning_rate": 6.621805250375852e-05, "loss": 1.9268, "step": 41000 }, { "epoch": 3.55, "learning_rate": 6.621588412166069e-05, "loss": 1.7933, "step": 41010 }, { "epoch": 3.55, "learning_rate": 6.621371573956284e-05, "loss": 1.9611, "step": 41020 }, { "epoch": 3.55, "learning_rate": 6.621154735746501e-05, "loss": 1.8299, "step": 41030 }, { "epoch": 3.55, "learning_rate": 6.620937897536718e-05, "loss": 2.0457, "step": 41040 }, { "epoch": 3.56, "learning_rate": 6.620721059326933e-05, "loss": 1.8675, "step": 41050 }, { "epoch": 3.56, "learning_rate": 6.62050422111715e-05, "loss": 1.983, "step": 41060 }, { "epoch": 3.56, "learning_rate": 6.620287382907366e-05, "loss": 1.8429, "step": 41070 }, { "epoch": 3.56, "learning_rate": 6.620070544697583e-05, "loss": 1.8008, "step": 41080 }, { "epoch": 3.56, "learning_rate": 6.619853706487798e-05, "loss": 1.6797, "step": 41090 }, { "epoch": 3.56, "learning_rate": 6.619636868278015e-05, "loss": 1.8124, "step": 41100 }, { "epoch": 3.56, "learning_rate": 6.619420030068231e-05, "loss": 1.6629, "step": 41110 }, { "epoch": 3.56, "learning_rate": 6.619203191858447e-05, "loss": 1.7739, "step": 41120 }, { "epoch": 3.56, "learning_rate": 6.618986353648663e-05, "loss": 1.6756, "step": 41130 }, { "epoch": 3.56, "learning_rate": 6.61876951543888e-05, "loss": 1.6863, "step": 41140 }, { "epoch": 3.56, "learning_rate": 6.618552677229097e-05, "loss": 1.6946, "step": 41150 }, { "epoch": 3.56, "learning_rate": 6.618335839019312e-05, "loss": 1.818, "step": 41160 }, { "epoch": 3.57, "learning_rate": 6.618119000809529e-05, "loss": 1.7841, "step": 41170 }, { "epoch": 3.57, "learning_rate": 6.617902162599745e-05, "loss": 1.7156, "step": 41180 }, { "epoch": 3.57, "learning_rate": 6.617685324389962e-05, "loss": 2.0074, "step": 41190 }, { "epoch": 3.57, "learning_rate": 6.617468486180177e-05, "loss": 1.7363, "step": 41200 }, { "epoch": 3.57, "learning_rate": 6.617251647970394e-05, "loss": 1.778, "step": 41210 }, { "epoch": 3.57, "learning_rate": 6.61703480976061e-05, "loss": 1.9804, "step": 41220 }, { "epoch": 3.57, "learning_rate": 6.616817971550826e-05, "loss": 1.8157, "step": 41230 }, { "epoch": 3.57, "learning_rate": 6.616601133341042e-05, "loss": 1.8027, "step": 41240 }, { "epoch": 3.57, "learning_rate": 6.616384295131259e-05, "loss": 1.6283, "step": 41250 }, { "epoch": 3.57, "learning_rate": 6.616167456921476e-05, "loss": 1.9035, "step": 41260 }, { "epoch": 3.57, "learning_rate": 6.615950618711692e-05, "loss": 1.632, "step": 41270 }, { "epoch": 3.58, "learning_rate": 6.615733780501908e-05, "loss": 1.7536, "step": 41280 }, { "epoch": 3.58, "learning_rate": 6.615516942292124e-05, "loss": 1.6177, "step": 41290 }, { "epoch": 3.58, "learning_rate": 6.61530010408234e-05, "loss": 1.7783, "step": 41300 }, { "epoch": 3.58, "learning_rate": 6.615083265872556e-05, "loss": 1.8097, "step": 41310 }, { "epoch": 3.58, "learning_rate": 6.614866427662773e-05, "loss": 1.8436, "step": 41320 }, { "epoch": 3.58, "learning_rate": 6.61464958945299e-05, "loss": 1.8356, "step": 41330 }, { "epoch": 3.58, "learning_rate": 6.614432751243206e-05, "loss": 1.8493, "step": 41340 }, { "epoch": 3.58, "learning_rate": 6.614215913033421e-05, "loss": 1.6779, "step": 41350 }, { "epoch": 3.58, "learning_rate": 6.613999074823638e-05, "loss": 1.5349, "step": 41360 }, { "epoch": 3.58, "learning_rate": 6.613782236613853e-05, "loss": 1.883, "step": 41370 }, { "epoch": 3.58, "learning_rate": 6.61356539840407e-05, "loss": 1.8073, "step": 41380 }, { "epoch": 3.58, "learning_rate": 6.613348560194286e-05, "loss": 1.9599, "step": 41390 }, { "epoch": 3.59, "learning_rate": 6.613131721984503e-05, "loss": 1.9538, "step": 41400 }, { "epoch": 3.59, "learning_rate": 6.61291488377472e-05, "loss": 1.7893, "step": 41410 }, { "epoch": 3.59, "learning_rate": 6.612698045564935e-05, "loss": 1.8948, "step": 41420 }, { "epoch": 3.59, "learning_rate": 6.612481207355152e-05, "loss": 2.0617, "step": 41430 }, { "epoch": 3.59, "learning_rate": 6.612264369145367e-05, "loss": 2.0107, "step": 41440 }, { "epoch": 3.59, "learning_rate": 6.612047530935584e-05, "loss": 1.7351, "step": 41450 }, { "epoch": 3.59, "learning_rate": 6.6118306927258e-05, "loss": 1.7111, "step": 41460 }, { "epoch": 3.59, "learning_rate": 6.611613854516017e-05, "loss": 1.5733, "step": 41470 }, { "epoch": 3.59, "learning_rate": 6.611397016306234e-05, "loss": 1.933, "step": 41480 }, { "epoch": 3.59, "learning_rate": 6.611180178096449e-05, "loss": 1.7992, "step": 41490 }, { "epoch": 3.59, "learning_rate": 6.610963339886665e-05, "loss": 1.706, "step": 41500 }, { "epoch": 3.6, "learning_rate": 6.610746501676881e-05, "loss": 1.7492, "step": 41510 }, { "epoch": 3.6, "learning_rate": 6.610529663467097e-05, "loss": 1.9564, "step": 41520 }, { "epoch": 3.6, "learning_rate": 6.610312825257314e-05, "loss": 1.9399, "step": 41530 }, { "epoch": 3.6, "learning_rate": 6.61009598704753e-05, "loss": 1.5264, "step": 41540 }, { "epoch": 3.6, "learning_rate": 6.609879148837747e-05, "loss": 1.7095, "step": 41550 }, { "epoch": 3.6, "learning_rate": 6.609662310627963e-05, "loss": 1.734, "step": 41560 }, { "epoch": 3.6, "learning_rate": 6.609445472418179e-05, "loss": 1.7348, "step": 41570 }, { "epoch": 3.6, "learning_rate": 6.609228634208394e-05, "loss": 1.8479, "step": 41580 }, { "epoch": 3.6, "learning_rate": 6.609011795998611e-05, "loss": 1.5857, "step": 41590 }, { "epoch": 3.6, "learning_rate": 6.608794957788828e-05, "loss": 1.7669, "step": 41600 }, { "epoch": 3.6, "learning_rate": 6.608578119579044e-05, "loss": 1.6637, "step": 41610 }, { "epoch": 3.6, "learning_rate": 6.608361281369261e-05, "loss": 1.7198, "step": 41620 }, { "epoch": 3.61, "learning_rate": 6.608144443159476e-05, "loss": 1.5109, "step": 41630 }, { "epoch": 3.61, "learning_rate": 6.607927604949693e-05, "loss": 1.7648, "step": 41640 }, { "epoch": 3.61, "learning_rate": 6.60771076673991e-05, "loss": 1.888, "step": 41650 }, { "epoch": 3.61, "learning_rate": 6.607493928530125e-05, "loss": 1.6883, "step": 41660 }, { "epoch": 3.61, "learning_rate": 6.607277090320342e-05, "loss": 1.8405, "step": 41670 }, { "epoch": 3.61, "learning_rate": 6.607060252110558e-05, "loss": 1.8074, "step": 41680 }, { "epoch": 3.61, "learning_rate": 6.606843413900775e-05, "loss": 1.922, "step": 41690 }, { "epoch": 3.61, "learning_rate": 6.60662657569099e-05, "loss": 1.6983, "step": 41700 }, { "epoch": 3.61, "learning_rate": 6.606409737481207e-05, "loss": 1.8577, "step": 41710 }, { "epoch": 3.61, "learning_rate": 6.606192899271423e-05, "loss": 1.6064, "step": 41720 }, { "epoch": 3.61, "learning_rate": 6.60597606106164e-05, "loss": 2.0764, "step": 41730 }, { "epoch": 3.61, "learning_rate": 6.605759222851855e-05, "loss": 1.736, "step": 41740 }, { "epoch": 3.62, "learning_rate": 6.605542384642072e-05, "loss": 1.8043, "step": 41750 }, { "epoch": 3.62, "learning_rate": 6.605325546432289e-05, "loss": 1.6314, "step": 41760 }, { "epoch": 3.62, "learning_rate": 6.605108708222504e-05, "loss": 1.8194, "step": 41770 }, { "epoch": 3.62, "learning_rate": 6.60489187001272e-05, "loss": 1.9442, "step": 41780 }, { "epoch": 3.62, "learning_rate": 6.604675031802937e-05, "loss": 1.9542, "step": 41790 }, { "epoch": 3.62, "learning_rate": 6.604458193593154e-05, "loss": 1.9671, "step": 41800 }, { "epoch": 3.62, "learning_rate": 6.60424135538337e-05, "loss": 1.993, "step": 41810 }, { "epoch": 3.62, "learning_rate": 6.604024517173586e-05, "loss": 2.0013, "step": 41820 }, { "epoch": 3.62, "learning_rate": 6.603807678963802e-05, "loss": 1.8384, "step": 41830 }, { "epoch": 3.62, "learning_rate": 6.603590840754018e-05, "loss": 1.8761, "step": 41840 }, { "epoch": 3.62, "learning_rate": 6.603374002544234e-05, "loss": 1.894, "step": 41850 }, { "epoch": 3.63, "learning_rate": 6.603157164334451e-05, "loss": 1.9996, "step": 41860 }, { "epoch": 3.63, "learning_rate": 6.602940326124668e-05, "loss": 2.0244, "step": 41870 }, { "epoch": 3.63, "learning_rate": 6.602723487914884e-05, "loss": 1.9054, "step": 41880 }, { "epoch": 3.63, "learning_rate": 6.6025066497051e-05, "loss": 1.874, "step": 41890 }, { "epoch": 3.63, "learning_rate": 6.602289811495316e-05, "loss": 1.8463, "step": 41900 }, { "epoch": 3.63, "learning_rate": 6.602072973285531e-05, "loss": 1.8873, "step": 41910 }, { "epoch": 3.63, "learning_rate": 6.601856135075748e-05, "loss": 1.7177, "step": 41920 }, { "epoch": 3.63, "learning_rate": 6.601639296865965e-05, "loss": 1.7983, "step": 41930 }, { "epoch": 3.63, "learning_rate": 6.601422458656181e-05, "loss": 1.8988, "step": 41940 }, { "epoch": 3.63, "learning_rate": 6.601205620446398e-05, "loss": 1.9088, "step": 41950 }, { "epoch": 3.63, "learning_rate": 6.600988782236613e-05, "loss": 1.6705, "step": 41960 }, { "epoch": 3.63, "learning_rate": 6.600771944026829e-05, "loss": 1.7056, "step": 41970 }, { "epoch": 3.64, "learning_rate": 6.600555105817045e-05, "loss": 1.896, "step": 41980 }, { "epoch": 3.64, "learning_rate": 6.600338267607262e-05, "loss": 1.9011, "step": 41990 }, { "epoch": 3.64, "learning_rate": 6.600121429397478e-05, "loss": 1.8398, "step": 42000 }, { "epoch": 3.64, "learning_rate": 6.599904591187695e-05, "loss": 1.6395, "step": 42010 }, { "epoch": 3.64, "learning_rate": 6.59968775297791e-05, "loss": 1.7519, "step": 42020 }, { "epoch": 3.64, "learning_rate": 6.599470914768127e-05, "loss": 1.9133, "step": 42030 }, { "epoch": 3.64, "learning_rate": 6.599254076558344e-05, "loss": 2.0074, "step": 42040 }, { "epoch": 3.64, "learning_rate": 6.599037238348559e-05, "loss": 2.0921, "step": 42050 }, { "epoch": 3.64, "learning_rate": 6.598820400138776e-05, "loss": 1.6536, "step": 42060 }, { "epoch": 3.64, "learning_rate": 6.598603561928992e-05, "loss": 1.6926, "step": 42070 }, { "epoch": 3.64, "learning_rate": 6.598386723719209e-05, "loss": 1.6906, "step": 42080 }, { "epoch": 3.65, "learning_rate": 6.598169885509424e-05, "loss": 1.7246, "step": 42090 }, { "epoch": 3.65, "learning_rate": 6.597953047299641e-05, "loss": 1.7791, "step": 42100 }, { "epoch": 3.65, "learning_rate": 6.597736209089857e-05, "loss": 1.8252, "step": 42110 }, { "epoch": 3.65, "learning_rate": 6.597519370880073e-05, "loss": 1.8463, "step": 42120 }, { "epoch": 3.65, "learning_rate": 6.597302532670289e-05, "loss": 1.9132, "step": 42130 }, { "epoch": 3.65, "learning_rate": 6.597085694460506e-05, "loss": 1.8093, "step": 42140 }, { "epoch": 3.65, "learning_rate": 6.596868856250723e-05, "loss": 1.7955, "step": 42150 }, { "epoch": 3.65, "learning_rate": 6.596652018040938e-05, "loss": 1.8579, "step": 42160 }, { "epoch": 3.65, "learning_rate": 6.596435179831155e-05, "loss": 1.8624, "step": 42170 }, { "epoch": 3.65, "learning_rate": 6.596218341621371e-05, "loss": 1.6903, "step": 42180 }, { "epoch": 3.65, "learning_rate": 6.596001503411588e-05, "loss": 1.7148, "step": 42190 }, { "epoch": 3.65, "learning_rate": 6.595784665201803e-05, "loss": 1.7976, "step": 42200 }, { "epoch": 3.66, "learning_rate": 6.59556782699202e-05, "loss": 1.9384, "step": 42210 }, { "epoch": 3.66, "learning_rate": 6.595350988782236e-05, "loss": 1.8177, "step": 42220 }, { "epoch": 3.66, "learning_rate": 6.595134150572452e-05, "loss": 1.911, "step": 42230 }, { "epoch": 3.66, "learning_rate": 6.594917312362668e-05, "loss": 1.6609, "step": 42240 }, { "epoch": 3.66, "learning_rate": 6.594700474152885e-05, "loss": 1.7688, "step": 42250 }, { "epoch": 3.66, "learning_rate": 6.594483635943102e-05, "loss": 1.648, "step": 42260 }, { "epoch": 3.66, "learning_rate": 6.594266797733318e-05, "loss": 1.8984, "step": 42270 }, { "epoch": 3.66, "learning_rate": 6.594049959523533e-05, "loss": 1.6373, "step": 42280 }, { "epoch": 3.66, "learning_rate": 6.59383312131375e-05, "loss": 1.7452, "step": 42290 }, { "epoch": 3.66, "learning_rate": 6.593616283103965e-05, "loss": 1.8156, "step": 42300 }, { "epoch": 3.66, "learning_rate": 6.593399444894182e-05, "loss": 1.6888, "step": 42310 }, { "epoch": 3.67, "learning_rate": 6.593182606684399e-05, "loss": 1.8878, "step": 42320 }, { "epoch": 3.67, "learning_rate": 6.592965768474615e-05, "loss": 1.7147, "step": 42330 }, { "epoch": 3.67, "learning_rate": 6.592748930264832e-05, "loss": 1.9726, "step": 42340 }, { "epoch": 3.67, "learning_rate": 6.592532092055047e-05, "loss": 1.9254, "step": 42350 }, { "epoch": 3.67, "learning_rate": 6.592315253845264e-05, "loss": 1.8564, "step": 42360 }, { "epoch": 3.67, "learning_rate": 6.592098415635479e-05, "loss": 1.8749, "step": 42370 }, { "epoch": 3.67, "learning_rate": 6.591881577425696e-05, "loss": 2.1286, "step": 42380 }, { "epoch": 3.67, "learning_rate": 6.591664739215912e-05, "loss": 1.7592, "step": 42390 }, { "epoch": 3.67, "learning_rate": 6.591447901006129e-05, "loss": 1.8177, "step": 42400 }, { "epoch": 3.67, "learning_rate": 6.591231062796346e-05, "loss": 1.9027, "step": 42410 }, { "epoch": 3.67, "learning_rate": 6.591014224586561e-05, "loss": 1.7332, "step": 42420 }, { "epoch": 3.67, "learning_rate": 6.590797386376778e-05, "loss": 1.8229, "step": 42430 }, { "epoch": 3.68, "learning_rate": 6.590580548166993e-05, "loss": 1.9434, "step": 42440 }, { "epoch": 3.68, "learning_rate": 6.59036370995721e-05, "loss": 1.6833, "step": 42450 }, { "epoch": 3.68, "learning_rate": 6.590146871747426e-05, "loss": 1.8733, "step": 42460 }, { "epoch": 3.68, "learning_rate": 6.589930033537643e-05, "loss": 1.8533, "step": 42470 }, { "epoch": 3.68, "learning_rate": 6.58971319532786e-05, "loss": 1.8624, "step": 42480 }, { "epoch": 3.68, "learning_rate": 6.589496357118075e-05, "loss": 1.8901, "step": 42490 }, { "epoch": 3.68, "learning_rate": 6.589279518908291e-05, "loss": 1.5798, "step": 42500 }, { "epoch": 3.68, "learning_rate": 6.589062680698507e-05, "loss": 1.6858, "step": 42510 }, { "epoch": 3.68, "learning_rate": 6.588845842488723e-05, "loss": 1.802, "step": 42520 }, { "epoch": 3.68, "learning_rate": 6.58862900427894e-05, "loss": 1.7931, "step": 42530 }, { "epoch": 3.68, "learning_rate": 6.588412166069157e-05, "loss": 1.747, "step": 42540 }, { "epoch": 3.69, "learning_rate": 6.588195327859373e-05, "loss": 1.8122, "step": 42550 }, { "epoch": 3.69, "learning_rate": 6.587978489649589e-05, "loss": 1.9494, "step": 42560 }, { "epoch": 3.69, "learning_rate": 6.587761651439805e-05, "loss": 1.8606, "step": 42570 }, { "epoch": 3.69, "learning_rate": 6.587544813230022e-05, "loss": 1.6134, "step": 42580 }, { "epoch": 3.69, "learning_rate": 6.587327975020237e-05, "loss": 1.6536, "step": 42590 }, { "epoch": 3.69, "learning_rate": 6.587111136810454e-05, "loss": 1.5538, "step": 42600 }, { "epoch": 3.69, "learning_rate": 6.58689429860067e-05, "loss": 1.6165, "step": 42610 }, { "epoch": 3.69, "learning_rate": 6.586677460390887e-05, "loss": 1.7198, "step": 42620 }, { "epoch": 3.69, "learning_rate": 6.586460622181102e-05, "loss": 1.8708, "step": 42630 }, { "epoch": 3.69, "learning_rate": 6.586243783971319e-05, "loss": 1.7673, "step": 42640 }, { "epoch": 3.69, "learning_rate": 6.586026945761536e-05, "loss": 1.8065, "step": 42650 }, { "epoch": 3.69, "learning_rate": 6.585810107551751e-05, "loss": 1.7888, "step": 42660 }, { "epoch": 3.7, "learning_rate": 6.585593269341968e-05, "loss": 1.823, "step": 42670 }, { "epoch": 3.7, "learning_rate": 6.585376431132184e-05, "loss": 2.0106, "step": 42680 }, { "epoch": 3.7, "learning_rate": 6.585159592922401e-05, "loss": 1.8457, "step": 42690 }, { "epoch": 3.7, "learning_rate": 6.584942754712616e-05, "loss": 1.9447, "step": 42700 }, { "epoch": 3.7, "learning_rate": 6.584725916502833e-05, "loss": 2.0716, "step": 42710 }, { "epoch": 3.7, "learning_rate": 6.58450907829305e-05, "loss": 1.9824, "step": 42720 }, { "epoch": 3.7, "learning_rate": 6.584292240083266e-05, "loss": 1.712, "step": 42730 }, { "epoch": 3.7, "learning_rate": 6.584075401873481e-05, "loss": 1.6828, "step": 42740 }, { "epoch": 3.7, "learning_rate": 6.583858563663698e-05, "loss": 1.7922, "step": 42750 }, { "epoch": 3.7, "learning_rate": 6.583641725453915e-05, "loss": 1.7058, "step": 42760 }, { "epoch": 3.7, "learning_rate": 6.58342488724413e-05, "loss": 1.8259, "step": 42770 }, { "epoch": 3.71, "learning_rate": 6.583208049034346e-05, "loss": 1.8733, "step": 42780 }, { "epoch": 3.71, "learning_rate": 6.582991210824563e-05, "loss": 1.731, "step": 42790 }, { "epoch": 3.71, "learning_rate": 6.58277437261478e-05, "loss": 1.7688, "step": 42800 }, { "epoch": 3.71, "learning_rate": 6.582557534404996e-05, "loss": 1.7979, "step": 42810 }, { "epoch": 3.71, "learning_rate": 6.582340696195212e-05, "loss": 1.8651, "step": 42820 }, { "epoch": 3.71, "learning_rate": 6.582123857985428e-05, "loss": 1.7044, "step": 42830 }, { "epoch": 3.71, "learning_rate": 6.581907019775644e-05, "loss": 1.737, "step": 42840 }, { "epoch": 3.71, "learning_rate": 6.58169018156586e-05, "loss": 1.7497, "step": 42850 }, { "epoch": 3.71, "learning_rate": 6.581473343356077e-05, "loss": 1.9478, "step": 42860 }, { "epoch": 3.71, "learning_rate": 6.581256505146294e-05, "loss": 1.9294, "step": 42870 }, { "epoch": 3.71, "learning_rate": 6.58103966693651e-05, "loss": 1.9109, "step": 42880 }, { "epoch": 3.71, "learning_rate": 6.580822828726725e-05, "loss": 1.8246, "step": 42890 }, { "epoch": 3.72, "learning_rate": 6.580605990516942e-05, "loss": 1.6368, "step": 42900 }, { "epoch": 3.72, "learning_rate": 6.580389152307157e-05, "loss": 1.6193, "step": 42910 }, { "epoch": 3.72, "learning_rate": 6.580172314097374e-05, "loss": 1.7871, "step": 42920 }, { "epoch": 3.72, "learning_rate": 6.57995547588759e-05, "loss": 1.78, "step": 42930 }, { "epoch": 3.72, "learning_rate": 6.579738637677807e-05, "loss": 1.6852, "step": 42940 }, { "epoch": 3.72, "learning_rate": 6.579521799468024e-05, "loss": 2.0027, "step": 42950 }, { "epoch": 3.72, "learning_rate": 6.579304961258239e-05, "loss": 1.7622, "step": 42960 }, { "epoch": 3.72, "learning_rate": 6.579088123048456e-05, "loss": 1.8564, "step": 42970 }, { "epoch": 3.72, "learning_rate": 6.578871284838671e-05, "loss": 1.7393, "step": 42980 }, { "epoch": 3.72, "learning_rate": 6.578654446628888e-05, "loss": 1.7311, "step": 42990 }, { "epoch": 3.72, "learning_rate": 6.578437608419104e-05, "loss": 1.7375, "step": 43000 }, { "epoch": 3.72, "learning_rate": 6.578220770209321e-05, "loss": 1.758, "step": 43010 }, { "epoch": 3.73, "learning_rate": 6.578003931999538e-05, "loss": 1.5915, "step": 43020 }, { "epoch": 3.73, "learning_rate": 6.577787093789753e-05, "loss": 1.7145, "step": 43030 }, { "epoch": 3.73, "learning_rate": 6.57757025557997e-05, "loss": 2.0171, "step": 43040 }, { "epoch": 3.73, "learning_rate": 6.577353417370185e-05, "loss": 1.6742, "step": 43050 }, { "epoch": 3.73, "learning_rate": 6.577136579160402e-05, "loss": 1.7319, "step": 43060 }, { "epoch": 3.73, "learning_rate": 6.576919740950618e-05, "loss": 1.7187, "step": 43070 }, { "epoch": 3.73, "learning_rate": 6.576702902740835e-05, "loss": 1.4878, "step": 43080 }, { "epoch": 3.73, "learning_rate": 6.57648606453105e-05, "loss": 2.0317, "step": 43090 }, { "epoch": 3.73, "learning_rate": 6.576269226321267e-05, "loss": 1.9563, "step": 43100 }, { "epoch": 3.73, "learning_rate": 6.576052388111483e-05, "loss": 1.9439, "step": 43110 }, { "epoch": 3.73, "learning_rate": 6.575835549901699e-05, "loss": 1.8491, "step": 43120 }, { "epoch": 3.74, "learning_rate": 6.575618711691915e-05, "loss": 1.6415, "step": 43130 }, { "epoch": 3.74, "learning_rate": 6.575401873482132e-05, "loss": 1.8275, "step": 43140 }, { "epoch": 3.74, "learning_rate": 6.575185035272349e-05, "loss": 1.8628, "step": 43150 }, { "epoch": 3.74, "learning_rate": 6.574968197062564e-05, "loss": 1.9116, "step": 43160 }, { "epoch": 3.74, "learning_rate": 6.57475135885278e-05, "loss": 1.6915, "step": 43170 }, { "epoch": 3.74, "learning_rate": 6.574534520642997e-05, "loss": 1.7585, "step": 43180 }, { "epoch": 3.74, "learning_rate": 6.574317682433214e-05, "loss": 1.6671, "step": 43190 }, { "epoch": 3.74, "learning_rate": 6.574100844223429e-05, "loss": 1.8829, "step": 43200 }, { "epoch": 3.74, "learning_rate": 6.573884006013646e-05, "loss": 1.9461, "step": 43210 }, { "epoch": 3.74, "learning_rate": 6.573667167803862e-05, "loss": 1.8899, "step": 43220 }, { "epoch": 3.74, "learning_rate": 6.573450329594078e-05, "loss": 1.7062, "step": 43230 }, { "epoch": 3.74, "learning_rate": 6.573233491384294e-05, "loss": 1.7192, "step": 43240 }, { "epoch": 3.75, "learning_rate": 6.573016653174511e-05, "loss": 1.6072, "step": 43250 }, { "epoch": 3.75, "learning_rate": 6.572799814964728e-05, "loss": 1.9758, "step": 43260 }, { "epoch": 3.75, "learning_rate": 6.572582976754944e-05, "loss": 1.8289, "step": 43270 }, { "epoch": 3.75, "learning_rate": 6.57236613854516e-05, "loss": 1.7902, "step": 43280 }, { "epoch": 3.75, "learning_rate": 6.572149300335376e-05, "loss": 1.8492, "step": 43290 }, { "epoch": 3.75, "learning_rate": 6.571932462125591e-05, "loss": 2.0307, "step": 43300 }, { "epoch": 3.75, "learning_rate": 6.571715623915808e-05, "loss": 1.6986, "step": 43310 }, { "epoch": 3.75, "learning_rate": 6.571498785706025e-05, "loss": 1.7429, "step": 43320 }, { "epoch": 3.75, "learning_rate": 6.571281947496241e-05, "loss": 1.7804, "step": 43330 }, { "epoch": 3.75, "learning_rate": 6.571065109286458e-05, "loss": 1.8738, "step": 43340 }, { "epoch": 3.75, "learning_rate": 6.570848271076673e-05, "loss": 1.6017, "step": 43350 }, { "epoch": 3.76, "learning_rate": 6.57063143286689e-05, "loss": 1.791, "step": 43360 }, { "epoch": 3.76, "learning_rate": 6.570414594657105e-05, "loss": 1.9182, "step": 43370 }, { "epoch": 3.76, "learning_rate": 6.570197756447322e-05, "loss": 1.6182, "step": 43380 }, { "epoch": 3.76, "learning_rate": 6.569980918237538e-05, "loss": 1.7636, "step": 43390 }, { "epoch": 3.76, "learning_rate": 6.569764080027755e-05, "loss": 1.8566, "step": 43400 }, { "epoch": 3.76, "learning_rate": 6.569547241817972e-05, "loss": 1.7178, "step": 43410 }, { "epoch": 3.76, "learning_rate": 6.569330403608187e-05, "loss": 1.6847, "step": 43420 }, { "epoch": 3.76, "learning_rate": 6.569113565398404e-05, "loss": 1.773, "step": 43430 }, { "epoch": 3.76, "learning_rate": 6.568896727188619e-05, "loss": 1.8017, "step": 43440 }, { "epoch": 3.76, "learning_rate": 6.568679888978836e-05, "loss": 1.7652, "step": 43450 }, { "epoch": 3.76, "learning_rate": 6.568463050769052e-05, "loss": 1.7718, "step": 43460 }, { "epoch": 3.76, "learning_rate": 6.568246212559269e-05, "loss": 1.6687, "step": 43470 }, { "epoch": 3.77, "learning_rate": 6.568029374349485e-05, "loss": 1.7471, "step": 43480 }, { "epoch": 3.77, "learning_rate": 6.567812536139701e-05, "loss": 2.13, "step": 43490 }, { "epoch": 3.77, "learning_rate": 6.567595697929917e-05, "loss": 1.9619, "step": 43500 }, { "epoch": 3.77, "learning_rate": 6.567378859720133e-05, "loss": 1.9016, "step": 43510 }, { "epoch": 3.77, "learning_rate": 6.567162021510349e-05, "loss": 1.8815, "step": 43520 }, { "epoch": 3.77, "learning_rate": 6.566945183300566e-05, "loss": 2.1158, "step": 43530 }, { "epoch": 3.77, "learning_rate": 6.566728345090783e-05, "loss": 1.6211, "step": 43540 }, { "epoch": 3.77, "learning_rate": 6.566511506880999e-05, "loss": 1.8842, "step": 43550 }, { "epoch": 3.77, "learning_rate": 6.566294668671215e-05, "loss": 1.914, "step": 43560 }, { "epoch": 3.77, "learning_rate": 6.566077830461431e-05, "loss": 1.6775, "step": 43570 }, { "epoch": 3.77, "learning_rate": 6.565860992251648e-05, "loss": 1.7579, "step": 43580 }, { "epoch": 3.78, "learning_rate": 6.565644154041863e-05, "loss": 1.8465, "step": 43590 }, { "epoch": 3.78, "learning_rate": 6.56542731583208e-05, "loss": 1.7548, "step": 43600 }, { "epoch": 3.78, "learning_rate": 6.565210477622296e-05, "loss": 1.8547, "step": 43610 }, { "epoch": 3.78, "learning_rate": 6.564993639412513e-05, "loss": 1.8606, "step": 43620 }, { "epoch": 3.78, "learning_rate": 6.564776801202728e-05, "loss": 2.0289, "step": 43630 }, { "epoch": 3.78, "learning_rate": 6.564559962992945e-05, "loss": 1.8317, "step": 43640 }, { "epoch": 3.78, "learning_rate": 6.564343124783162e-05, "loss": 1.8058, "step": 43650 }, { "epoch": 3.78, "learning_rate": 6.564126286573377e-05, "loss": 1.7937, "step": 43660 }, { "epoch": 3.78, "learning_rate": 6.563909448363593e-05, "loss": 1.7639, "step": 43670 }, { "epoch": 3.78, "learning_rate": 6.56369261015381e-05, "loss": 1.729, "step": 43680 }, { "epoch": 3.78, "learning_rate": 6.563475771944027e-05, "loss": 1.8573, "step": 43690 }, { "epoch": 3.78, "learning_rate": 6.563258933734242e-05, "loss": 1.8674, "step": 43700 }, { "epoch": 3.79, "learning_rate": 6.563042095524459e-05, "loss": 1.7123, "step": 43710 }, { "epoch": 3.79, "learning_rate": 6.562825257314675e-05, "loss": 1.9311, "step": 43720 }, { "epoch": 3.79, "learning_rate": 6.562608419104892e-05, "loss": 1.7159, "step": 43730 }, { "epoch": 3.79, "learning_rate": 6.562391580895107e-05, "loss": 1.8333, "step": 43740 }, { "epoch": 3.79, "learning_rate": 6.562174742685324e-05, "loss": 1.6631, "step": 43750 }, { "epoch": 3.79, "learning_rate": 6.56195790447554e-05, "loss": 1.5375, "step": 43760 }, { "epoch": 3.79, "learning_rate": 6.561741066265756e-05, "loss": 1.6458, "step": 43770 }, { "epoch": 3.79, "learning_rate": 6.561524228055972e-05, "loss": 1.638, "step": 43780 }, { "epoch": 3.79, "learning_rate": 6.561307389846189e-05, "loss": 1.7766, "step": 43790 }, { "epoch": 3.79, "learning_rate": 6.561090551636406e-05, "loss": 1.7709, "step": 43800 }, { "epoch": 3.79, "learning_rate": 6.560873713426622e-05, "loss": 2.1745, "step": 43810 }, { "epoch": 3.8, "learning_rate": 6.560656875216838e-05, "loss": 1.6233, "step": 43820 }, { "epoch": 3.8, "learning_rate": 6.560440037007054e-05, "loss": 1.6608, "step": 43830 }, { "epoch": 3.8, "learning_rate": 6.56022319879727e-05, "loss": 2.0368, "step": 43840 }, { "epoch": 3.8, "learning_rate": 6.560006360587486e-05, "loss": 1.6795, "step": 43850 }, { "epoch": 3.8, "learning_rate": 6.559789522377703e-05, "loss": 1.9066, "step": 43860 }, { "epoch": 3.8, "learning_rate": 6.55957268416792e-05, "loss": 1.757, "step": 43870 }, { "epoch": 3.8, "learning_rate": 6.559355845958136e-05, "loss": 1.8514, "step": 43880 }, { "epoch": 3.8, "learning_rate": 6.559139007748351e-05, "loss": 1.6855, "step": 43890 }, { "epoch": 3.8, "learning_rate": 6.558922169538568e-05, "loss": 1.8796, "step": 43900 }, { "epoch": 3.8, "learning_rate": 6.558705331328783e-05, "loss": 1.6531, "step": 43910 }, { "epoch": 3.8, "learning_rate": 6.558488493119e-05, "loss": 1.8947, "step": 43920 }, { "epoch": 3.8, "learning_rate": 6.558271654909217e-05, "loss": 1.848, "step": 43930 }, { "epoch": 3.81, "learning_rate": 6.558054816699433e-05, "loss": 1.7908, "step": 43940 }, { "epoch": 3.81, "learning_rate": 6.55783797848965e-05, "loss": 1.7081, "step": 43950 }, { "epoch": 3.81, "learning_rate": 6.557621140279865e-05, "loss": 1.8795, "step": 43960 }, { "epoch": 3.81, "learning_rate": 6.557404302070082e-05, "loss": 1.9102, "step": 43970 }, { "epoch": 3.81, "learning_rate": 6.557187463860297e-05, "loss": 1.7983, "step": 43980 }, { "epoch": 3.81, "learning_rate": 6.556970625650514e-05, "loss": 1.7588, "step": 43990 }, { "epoch": 3.81, "learning_rate": 6.55675378744073e-05, "loss": 1.5651, "step": 44000 }, { "epoch": 3.81, "learning_rate": 6.556536949230947e-05, "loss": 1.6682, "step": 44010 }, { "epoch": 3.81, "learning_rate": 6.556320111021164e-05, "loss": 1.6209, "step": 44020 }, { "epoch": 3.81, "learning_rate": 6.556103272811379e-05, "loss": 1.6623, "step": 44030 }, { "epoch": 3.81, "learning_rate": 6.555886434601596e-05, "loss": 1.5677, "step": 44040 }, { "epoch": 3.82, "learning_rate": 6.555669596391811e-05, "loss": 1.7083, "step": 44050 }, { "epoch": 3.82, "learning_rate": 6.555452758182027e-05, "loss": 1.8523, "step": 44060 }, { "epoch": 3.82, "learning_rate": 6.555235919972244e-05, "loss": 1.6463, "step": 44070 }, { "epoch": 3.82, "learning_rate": 6.555019081762461e-05, "loss": 2.125, "step": 44080 }, { "epoch": 3.82, "learning_rate": 6.554802243552677e-05, "loss": 1.7319, "step": 44090 }, { "epoch": 3.82, "learning_rate": 6.554585405342893e-05, "loss": 1.8771, "step": 44100 }, { "epoch": 3.82, "learning_rate": 6.55436856713311e-05, "loss": 1.9266, "step": 44110 }, { "epoch": 3.82, "learning_rate": 6.554151728923325e-05, "loss": 1.9871, "step": 44120 }, { "epoch": 3.82, "learning_rate": 6.553934890713541e-05, "loss": 1.5396, "step": 44130 }, { "epoch": 3.82, "learning_rate": 6.553718052503758e-05, "loss": 1.8001, "step": 44140 }, { "epoch": 3.82, "learning_rate": 6.553501214293975e-05, "loss": 1.8122, "step": 44150 }, { "epoch": 3.82, "learning_rate": 6.55328437608419e-05, "loss": 1.7635, "step": 44160 }, { "epoch": 3.83, "learning_rate": 6.553067537874406e-05, "loss": 1.7965, "step": 44170 }, { "epoch": 3.83, "learning_rate": 6.552850699664623e-05, "loss": 1.8723, "step": 44180 }, { "epoch": 3.83, "learning_rate": 6.55263386145484e-05, "loss": 1.861, "step": 44190 }, { "epoch": 3.83, "learning_rate": 6.552417023245055e-05, "loss": 1.7555, "step": 44200 }, { "epoch": 3.83, "learning_rate": 6.552200185035272e-05, "loss": 1.7346, "step": 44210 }, { "epoch": 3.83, "learning_rate": 6.551983346825488e-05, "loss": 1.7031, "step": 44220 }, { "epoch": 3.83, "learning_rate": 6.551766508615704e-05, "loss": 1.738, "step": 44230 }, { "epoch": 3.83, "learning_rate": 6.55154967040592e-05, "loss": 1.9792, "step": 44240 }, { "epoch": 3.83, "learning_rate": 6.551332832196137e-05, "loss": 1.7003, "step": 44250 }, { "epoch": 3.83, "learning_rate": 6.551115993986353e-05, "loss": 1.6737, "step": 44260 }, { "epoch": 3.83, "learning_rate": 6.55089915577657e-05, "loss": 1.8268, "step": 44270 }, { "epoch": 3.83, "learning_rate": 6.550682317566785e-05, "loss": 1.6234, "step": 44280 }, { "epoch": 3.84, "learning_rate": 6.550465479357002e-05, "loss": 1.7318, "step": 44290 }, { "epoch": 3.84, "learning_rate": 6.550248641147217e-05, "loss": 1.882, "step": 44300 }, { "epoch": 3.84, "learning_rate": 6.550031802937434e-05, "loss": 1.8969, "step": 44310 }, { "epoch": 3.84, "learning_rate": 6.54981496472765e-05, "loss": 1.9734, "step": 44320 }, { "epoch": 3.84, "learning_rate": 6.549598126517867e-05, "loss": 1.6339, "step": 44330 }, { "epoch": 3.84, "learning_rate": 6.549381288308084e-05, "loss": 1.6519, "step": 44340 }, { "epoch": 3.84, "learning_rate": 6.549164450098299e-05, "loss": 1.7386, "step": 44350 }, { "epoch": 3.84, "learning_rate": 6.548947611888516e-05, "loss": 1.7995, "step": 44360 }, { "epoch": 3.84, "learning_rate": 6.548730773678731e-05, "loss": 1.7608, "step": 44370 }, { "epoch": 3.84, "learning_rate": 6.548513935468948e-05, "loss": 1.772, "step": 44380 }, { "epoch": 3.84, "learning_rate": 6.548297097259164e-05, "loss": 2.144, "step": 44390 }, { "epoch": 3.85, "learning_rate": 6.548080259049381e-05, "loss": 1.8946, "step": 44400 }, { "epoch": 3.85, "learning_rate": 6.547863420839598e-05, "loss": 1.741, "step": 44410 }, { "epoch": 3.85, "learning_rate": 6.547646582629813e-05, "loss": 1.8603, "step": 44420 }, { "epoch": 3.85, "learning_rate": 6.54742974442003e-05, "loss": 1.8843, "step": 44430 }, { "epoch": 3.85, "learning_rate": 6.547212906210245e-05, "loss": 1.6296, "step": 44440 }, { "epoch": 3.85, "learning_rate": 6.546996068000462e-05, "loss": 1.9709, "step": 44450 }, { "epoch": 3.85, "learning_rate": 6.546779229790678e-05, "loss": 1.745, "step": 44460 }, { "epoch": 3.85, "learning_rate": 6.546562391580895e-05, "loss": 1.654, "step": 44470 }, { "epoch": 3.85, "learning_rate": 6.546345553371111e-05, "loss": 1.722, "step": 44480 }, { "epoch": 3.85, "learning_rate": 6.546128715161327e-05, "loss": 1.6481, "step": 44490 }, { "epoch": 3.85, "learning_rate": 6.545911876951543e-05, "loss": 1.7462, "step": 44500 }, { "epoch": 3.85, "learning_rate": 6.545695038741759e-05, "loss": 1.6775, "step": 44510 }, { "epoch": 3.86, "learning_rate": 6.545478200531975e-05, "loss": 1.9245, "step": 44520 }, { "epoch": 3.86, "learning_rate": 6.545261362322192e-05, "loss": 1.9275, "step": 44530 }, { "epoch": 3.86, "learning_rate": 6.545044524112409e-05, "loss": 1.8708, "step": 44540 }, { "epoch": 3.86, "learning_rate": 6.544827685902625e-05, "loss": 1.6981, "step": 44550 }, { "epoch": 3.86, "learning_rate": 6.54461084769284e-05, "loss": 1.7427, "step": 44560 }, { "epoch": 3.86, "learning_rate": 6.544394009483057e-05, "loss": 1.582, "step": 44570 }, { "epoch": 3.86, "learning_rate": 6.544177171273274e-05, "loss": 1.9244, "step": 44580 }, { "epoch": 3.86, "learning_rate": 6.543960333063489e-05, "loss": 1.7877, "step": 44590 }, { "epoch": 3.86, "learning_rate": 6.543743494853706e-05, "loss": 1.8277, "step": 44600 }, { "epoch": 3.86, "learning_rate": 6.543526656643922e-05, "loss": 1.6902, "step": 44610 }, { "epoch": 3.86, "learning_rate": 6.543309818434139e-05, "loss": 1.7762, "step": 44620 }, { "epoch": 3.87, "learning_rate": 6.543092980224354e-05, "loss": 1.9885, "step": 44630 }, { "epoch": 3.87, "learning_rate": 6.542876142014571e-05, "loss": 1.905, "step": 44640 }, { "epoch": 3.87, "learning_rate": 6.542659303804788e-05, "loss": 1.8861, "step": 44650 }, { "epoch": 3.87, "learning_rate": 6.542442465595003e-05, "loss": 1.8149, "step": 44660 }, { "epoch": 3.87, "learning_rate": 6.54222562738522e-05, "loss": 2.0298, "step": 44670 }, { "epoch": 3.87, "learning_rate": 6.542008789175436e-05, "loss": 1.9202, "step": 44680 }, { "epoch": 3.87, "learning_rate": 6.541791950965653e-05, "loss": 1.7858, "step": 44690 }, { "epoch": 3.87, "learning_rate": 6.541575112755868e-05, "loss": 1.7639, "step": 44700 }, { "epoch": 3.87, "learning_rate": 6.541358274546085e-05, "loss": 1.8081, "step": 44710 }, { "epoch": 3.87, "learning_rate": 6.541141436336301e-05, "loss": 1.7918, "step": 44720 }, { "epoch": 3.87, "learning_rate": 6.540924598126518e-05, "loss": 1.8734, "step": 44730 }, { "epoch": 3.87, "learning_rate": 6.540707759916733e-05, "loss": 1.8409, "step": 44740 }, { "epoch": 3.88, "learning_rate": 6.54049092170695e-05, "loss": 1.9869, "step": 44750 }, { "epoch": 3.88, "learning_rate": 6.540274083497166e-05, "loss": 1.9045, "step": 44760 }, { "epoch": 3.88, "learning_rate": 6.540057245287382e-05, "loss": 1.9511, "step": 44770 }, { "epoch": 3.88, "learning_rate": 6.539840407077598e-05, "loss": 1.8086, "step": 44780 }, { "epoch": 3.88, "learning_rate": 6.539623568867815e-05, "loss": 1.8261, "step": 44790 }, { "epoch": 3.88, "learning_rate": 6.539406730658032e-05, "loss": 2.0005, "step": 44800 }, { "epoch": 3.88, "learning_rate": 6.539189892448248e-05, "loss": 1.6598, "step": 44810 }, { "epoch": 3.88, "learning_rate": 6.538973054238464e-05, "loss": 1.7197, "step": 44820 }, { "epoch": 3.88, "learning_rate": 6.53875621602868e-05, "loss": 1.9109, "step": 44830 }, { "epoch": 3.88, "learning_rate": 6.538539377818896e-05, "loss": 1.894, "step": 44840 }, { "epoch": 3.88, "learning_rate": 6.538322539609112e-05, "loss": 1.7792, "step": 44850 }, { "epoch": 3.89, "learning_rate": 6.538105701399329e-05, "loss": 1.856, "step": 44860 }, { "epoch": 3.89, "learning_rate": 6.537888863189545e-05, "loss": 1.6283, "step": 44870 }, { "epoch": 3.89, "learning_rate": 6.537672024979762e-05, "loss": 1.8338, "step": 44880 }, { "epoch": 3.89, "learning_rate": 6.537455186769977e-05, "loss": 1.7382, "step": 44890 }, { "epoch": 3.89, "learning_rate": 6.537238348560194e-05, "loss": 1.697, "step": 44900 }, { "epoch": 3.89, "learning_rate": 6.537021510350409e-05, "loss": 1.8971, "step": 44910 }, { "epoch": 3.89, "learning_rate": 6.536804672140626e-05, "loss": 1.7669, "step": 44920 }, { "epoch": 3.89, "learning_rate": 6.536587833930843e-05, "loss": 1.6176, "step": 44930 }, { "epoch": 3.89, "learning_rate": 6.536370995721059e-05, "loss": 1.8541, "step": 44940 }, { "epoch": 3.89, "learning_rate": 6.536154157511276e-05, "loss": 1.6474, "step": 44950 }, { "epoch": 3.89, "learning_rate": 6.535937319301491e-05, "loss": 1.7862, "step": 44960 }, { "epoch": 3.89, "learning_rate": 6.535720481091708e-05, "loss": 1.7894, "step": 44970 }, { "epoch": 3.9, "learning_rate": 6.535503642881923e-05, "loss": 1.7451, "step": 44980 }, { "epoch": 3.9, "learning_rate": 6.53528680467214e-05, "loss": 1.838, "step": 44990 }, { "epoch": 3.9, "learning_rate": 6.535069966462356e-05, "loss": 1.8726, "step": 45000 }, { "epoch": 3.9, "learning_rate": 6.534853128252573e-05, "loss": 1.6713, "step": 45010 }, { "epoch": 3.9, "learning_rate": 6.53463629004279e-05, "loss": 1.8974, "step": 45020 }, { "epoch": 3.9, "learning_rate": 6.534419451833005e-05, "loss": 1.7992, "step": 45030 }, { "epoch": 3.9, "learning_rate": 6.534202613623222e-05, "loss": 1.9057, "step": 45040 }, { "epoch": 3.9, "learning_rate": 6.533985775413437e-05, "loss": 1.5403, "step": 45050 }, { "epoch": 3.9, "learning_rate": 6.533768937203653e-05, "loss": 1.7623, "step": 45060 }, { "epoch": 3.9, "learning_rate": 6.53355209899387e-05, "loss": 1.8562, "step": 45070 }, { "epoch": 3.9, "learning_rate": 6.533335260784087e-05, "loss": 1.7651, "step": 45080 }, { "epoch": 3.91, "learning_rate": 6.533118422574303e-05, "loss": 1.7442, "step": 45090 }, { "epoch": 3.91, "learning_rate": 6.532901584364519e-05, "loss": 1.9542, "step": 45100 }, { "epoch": 3.91, "learning_rate": 6.532684746154735e-05, "loss": 1.8656, "step": 45110 }, { "epoch": 3.91, "learning_rate": 6.532467907944952e-05, "loss": 1.5561, "step": 45120 }, { "epoch": 3.91, "learning_rate": 6.532251069735167e-05, "loss": 1.7161, "step": 45130 }, { "epoch": 3.91, "learning_rate": 6.532034231525384e-05, "loss": 1.9735, "step": 45140 }, { "epoch": 3.91, "learning_rate": 6.5318173933156e-05, "loss": 1.9204, "step": 45150 }, { "epoch": 3.91, "learning_rate": 6.531600555105817e-05, "loss": 1.6927, "step": 45160 }, { "epoch": 3.91, "learning_rate": 6.531383716896032e-05, "loss": 1.5518, "step": 45170 }, { "epoch": 3.91, "learning_rate": 6.531166878686249e-05, "loss": 1.7295, "step": 45180 }, { "epoch": 3.91, "learning_rate": 6.530950040476466e-05, "loss": 1.7884, "step": 45190 }, { "epoch": 3.91, "learning_rate": 6.530733202266681e-05, "loss": 1.5995, "step": 45200 }, { "epoch": 3.92, "learning_rate": 6.530516364056898e-05, "loss": 1.6615, "step": 45210 }, { "epoch": 3.92, "learning_rate": 6.530299525847114e-05, "loss": 1.9394, "step": 45220 }, { "epoch": 3.92, "learning_rate": 6.53008268763733e-05, "loss": 1.8651, "step": 45230 }, { "epoch": 3.92, "learning_rate": 6.529865849427546e-05, "loss": 2.0472, "step": 45240 }, { "epoch": 3.92, "learning_rate": 6.529649011217763e-05, "loss": 1.8215, "step": 45250 }, { "epoch": 3.92, "learning_rate": 6.52943217300798e-05, "loss": 1.7455, "step": 45260 }, { "epoch": 3.92, "learning_rate": 6.529215334798196e-05, "loss": 1.8858, "step": 45270 }, { "epoch": 3.92, "learning_rate": 6.528998496588411e-05, "loss": 1.6681, "step": 45280 }, { "epoch": 3.92, "learning_rate": 6.528781658378628e-05, "loss": 1.756, "step": 45290 }, { "epoch": 3.92, "learning_rate": 6.528564820168843e-05, "loss": 1.6073, "step": 45300 }, { "epoch": 3.92, "learning_rate": 6.52834798195906e-05, "loss": 1.6602, "step": 45310 }, { "epoch": 3.92, "learning_rate": 6.528131143749277e-05, "loss": 1.8086, "step": 45320 }, { "epoch": 3.93, "learning_rate": 6.527914305539493e-05, "loss": 1.7192, "step": 45330 }, { "epoch": 3.93, "learning_rate": 6.52769746732971e-05, "loss": 1.6758, "step": 45340 }, { "epoch": 3.93, "learning_rate": 6.527480629119925e-05, "loss": 1.7504, "step": 45350 }, { "epoch": 3.93, "learning_rate": 6.527263790910142e-05, "loss": 1.6106, "step": 45360 }, { "epoch": 3.93, "learning_rate": 6.527046952700357e-05, "loss": 1.822, "step": 45370 }, { "epoch": 3.93, "learning_rate": 6.526830114490574e-05, "loss": 1.9784, "step": 45380 }, { "epoch": 3.93, "learning_rate": 6.52661327628079e-05, "loss": 1.9265, "step": 45390 }, { "epoch": 3.93, "learning_rate": 6.526396438071007e-05, "loss": 1.7225, "step": 45400 }, { "epoch": 3.93, "learning_rate": 6.526179599861224e-05, "loss": 1.8368, "step": 45410 }, { "epoch": 3.93, "learning_rate": 6.525962761651439e-05, "loss": 1.8599, "step": 45420 }, { "epoch": 3.93, "learning_rate": 6.525745923441656e-05, "loss": 1.9779, "step": 45430 }, { "epoch": 3.94, "learning_rate": 6.525529085231871e-05, "loss": 1.8804, "step": 45440 }, { "epoch": 3.94, "learning_rate": 6.525312247022087e-05, "loss": 1.6925, "step": 45450 }, { "epoch": 3.94, "learning_rate": 6.525095408812304e-05, "loss": 1.8364, "step": 45460 }, { "epoch": 3.94, "learning_rate": 6.524878570602521e-05, "loss": 1.7929, "step": 45470 }, { "epoch": 3.94, "learning_rate": 6.524661732392737e-05, "loss": 1.8465, "step": 45480 }, { "epoch": 3.94, "learning_rate": 6.524444894182953e-05, "loss": 1.8048, "step": 45490 }, { "epoch": 3.94, "learning_rate": 6.524228055973169e-05, "loss": 1.9108, "step": 45500 }, { "epoch": 3.94, "learning_rate": 6.524011217763385e-05, "loss": 1.849, "step": 45510 }, { "epoch": 3.94, "learning_rate": 6.523794379553601e-05, "loss": 1.6798, "step": 45520 }, { "epoch": 3.94, "learning_rate": 6.523577541343818e-05, "loss": 1.9352, "step": 45530 }, { "epoch": 3.94, "learning_rate": 6.523360703134035e-05, "loss": 1.7068, "step": 45540 }, { "epoch": 3.94, "learning_rate": 6.523143864924251e-05, "loss": 1.7694, "step": 45550 }, { "epoch": 3.95, "learning_rate": 6.522927026714466e-05, "loss": 2.0668, "step": 45560 }, { "epoch": 3.95, "learning_rate": 6.522710188504683e-05, "loss": 1.9662, "step": 45570 }, { "epoch": 3.95, "learning_rate": 6.5224933502949e-05, "loss": 1.9655, "step": 45580 }, { "epoch": 3.95, "learning_rate": 6.522276512085115e-05, "loss": 1.6344, "step": 45590 }, { "epoch": 3.95, "learning_rate": 6.522059673875332e-05, "loss": 1.8677, "step": 45600 }, { "epoch": 3.95, "learning_rate": 6.521842835665548e-05, "loss": 1.997, "step": 45610 }, { "epoch": 3.95, "learning_rate": 6.521625997455765e-05, "loss": 1.6671, "step": 45620 }, { "epoch": 3.95, "learning_rate": 6.52140915924598e-05, "loss": 1.885, "step": 45630 }, { "epoch": 3.95, "learning_rate": 6.521192321036197e-05, "loss": 1.8174, "step": 45640 }, { "epoch": 3.95, "learning_rate": 6.520975482826413e-05, "loss": 1.8276, "step": 45650 }, { "epoch": 3.95, "learning_rate": 6.520758644616629e-05, "loss": 2.0661, "step": 45660 }, { "epoch": 3.96, "learning_rate": 6.520541806406845e-05, "loss": 1.8827, "step": 45670 }, { "epoch": 3.96, "learning_rate": 6.520324968197062e-05, "loss": 1.7521, "step": 45680 }, { "epoch": 3.96, "learning_rate": 6.520108129987279e-05, "loss": 1.6677, "step": 45690 }, { "epoch": 3.96, "learning_rate": 6.519891291777494e-05, "loss": 1.8136, "step": 45700 }, { "epoch": 3.96, "learning_rate": 6.51967445356771e-05, "loss": 2.0261, "step": 45710 }, { "epoch": 3.96, "learning_rate": 6.519457615357927e-05, "loss": 1.58, "step": 45720 }, { "epoch": 3.96, "learning_rate": 6.519240777148144e-05, "loss": 1.8225, "step": 45730 }, { "epoch": 3.96, "learning_rate": 6.519023938938359e-05, "loss": 1.7561, "step": 45740 }, { "epoch": 3.96, "learning_rate": 6.518807100728576e-05, "loss": 1.8131, "step": 45750 }, { "epoch": 3.96, "learning_rate": 6.518590262518792e-05, "loss": 1.8636, "step": 45760 }, { "epoch": 3.96, "learning_rate": 6.518373424309008e-05, "loss": 1.8363, "step": 45770 }, { "epoch": 3.96, "learning_rate": 6.518156586099224e-05, "loss": 1.4678, "step": 45780 }, { "epoch": 3.97, "learning_rate": 6.517939747889441e-05, "loss": 1.8317, "step": 45790 }, { "epoch": 3.97, "learning_rate": 6.517722909679658e-05, "loss": 1.5823, "step": 45800 }, { "epoch": 3.97, "learning_rate": 6.517506071469874e-05, "loss": 1.8747, "step": 45810 }, { "epoch": 3.97, "learning_rate": 6.51728923326009e-05, "loss": 1.84, "step": 45820 }, { "epoch": 3.97, "learning_rate": 6.517072395050306e-05, "loss": 1.8726, "step": 45830 }, { "epoch": 3.97, "learning_rate": 6.516855556840522e-05, "loss": 1.7264, "step": 45840 }, { "epoch": 3.97, "learning_rate": 6.516638718630738e-05, "loss": 1.6871, "step": 45850 }, { "epoch": 3.97, "learning_rate": 6.516421880420955e-05, "loss": 1.843, "step": 45860 }, { "epoch": 3.97, "learning_rate": 6.516205042211171e-05, "loss": 1.9542, "step": 45870 }, { "epoch": 3.97, "learning_rate": 6.515988204001388e-05, "loss": 1.7358, "step": 45880 }, { "epoch": 3.97, "learning_rate": 6.515771365791603e-05, "loss": 1.6572, "step": 45890 }, { "epoch": 3.98, "learning_rate": 6.51555452758182e-05, "loss": 1.5226, "step": 45900 }, { "epoch": 3.98, "learning_rate": 6.515337689372035e-05, "loss": 1.8503, "step": 45910 }, { "epoch": 3.98, "learning_rate": 6.515120851162252e-05, "loss": 2.1971, "step": 45920 }, { "epoch": 3.98, "learning_rate": 6.514904012952469e-05, "loss": 1.7126, "step": 45930 }, { "epoch": 3.98, "learning_rate": 6.514687174742685e-05, "loss": 1.5182, "step": 45940 }, { "epoch": 3.98, "learning_rate": 6.514470336532902e-05, "loss": 1.9308, "step": 45950 }, { "epoch": 3.98, "learning_rate": 6.514253498323117e-05, "loss": 2.208, "step": 45960 }, { "epoch": 3.98, "learning_rate": 6.514036660113334e-05, "loss": 1.5829, "step": 45970 }, { "epoch": 3.98, "learning_rate": 6.513819821903549e-05, "loss": 1.8242, "step": 45980 }, { "epoch": 3.98, "learning_rate": 6.513602983693766e-05, "loss": 1.8575, "step": 45990 }, { "epoch": 3.98, "learning_rate": 6.513386145483982e-05, "loss": 1.7627, "step": 46000 }, { "epoch": 3.98, "learning_rate": 6.513169307274199e-05, "loss": 1.9614, "step": 46010 }, { "epoch": 3.99, "learning_rate": 6.512952469064416e-05, "loss": 1.9186, "step": 46020 }, { "epoch": 3.99, "learning_rate": 6.512735630854631e-05, "loss": 1.8908, "step": 46030 }, { "epoch": 3.99, "learning_rate": 6.512518792644848e-05, "loss": 1.8272, "step": 46040 }, { "epoch": 3.99, "learning_rate": 6.512301954435063e-05, "loss": 1.7439, "step": 46050 }, { "epoch": 3.99, "learning_rate": 6.51208511622528e-05, "loss": 1.8216, "step": 46060 }, { "epoch": 3.99, "learning_rate": 6.511868278015496e-05, "loss": 1.9284, "step": 46070 }, { "epoch": 3.99, "learning_rate": 6.511651439805713e-05, "loss": 1.7058, "step": 46080 }, { "epoch": 3.99, "learning_rate": 6.51143460159593e-05, "loss": 1.7729, "step": 46090 }, { "epoch": 3.99, "learning_rate": 6.511217763386145e-05, "loss": 2.1528, "step": 46100 }, { "epoch": 3.99, "learning_rate": 6.511000925176361e-05, "loss": 1.6183, "step": 46110 }, { "epoch": 3.99, "learning_rate": 6.510784086966578e-05, "loss": 1.7304, "step": 46120 }, { "epoch": 4.0, "learning_rate": 6.510567248756793e-05, "loss": 1.7749, "step": 46130 }, { "epoch": 4.0, "learning_rate": 6.51035041054701e-05, "loss": 1.6611, "step": 46140 }, { "epoch": 4.0, "learning_rate": 6.510133572337226e-05, "loss": 1.8212, "step": 46150 }, { "epoch": 4.0, "learning_rate": 6.509916734127443e-05, "loss": 2.0482, "step": 46160 }, { "epoch": 4.0, "learning_rate": 6.509699895917658e-05, "loss": 1.9115, "step": 46170 }, { "epoch": 4.0, "learning_rate": 6.509483057707875e-05, "loss": 1.8071, "step": 46180 }, { "epoch": 4.0, "eval_cer": 0.5536389371651765, "eval_em": 0.0046415678184631255, "eval_f1": 0.0046415678184631255, "eval_loss": 1.6631836891174316, "eval_runtime": 272.1451, "eval_samples_per_second": 7.125, "eval_steps_per_second": 0.893, "eval_wer": 0.9953584321815369, "step": 46186 }, { "epoch": 4.0, "learning_rate": 6.509266219498092e-05, "loss": 1.5569, "step": 46190 }, { "epoch": 4.0, "learning_rate": 6.509049381288307e-05, "loss": 1.7436, "step": 46200 }, { "epoch": 4.0, "learning_rate": 6.508832543078524e-05, "loss": 1.677, "step": 46210 }, { "epoch": 4.0, "learning_rate": 6.50861570486874e-05, "loss": 1.7423, "step": 46220 }, { "epoch": 4.0, "learning_rate": 6.508398866658957e-05, "loss": 1.7105, "step": 46230 }, { "epoch": 4.0, "learning_rate": 6.508182028449172e-05, "loss": 1.687, "step": 46240 }, { "epoch": 4.01, "learning_rate": 6.507965190239389e-05, "loss": 1.6531, "step": 46250 }, { "epoch": 4.01, "learning_rate": 6.507748352029605e-05, "loss": 1.6894, "step": 46260 }, { "epoch": 4.01, "learning_rate": 6.507531513819822e-05, "loss": 1.678, "step": 46270 }, { "epoch": 4.01, "learning_rate": 6.507314675610037e-05, "loss": 1.758, "step": 46280 }, { "epoch": 4.01, "learning_rate": 6.507097837400254e-05, "loss": 1.6753, "step": 46290 }, { "epoch": 4.01, "learning_rate": 6.506880999190469e-05, "loss": 1.8565, "step": 46300 }, { "epoch": 4.01, "learning_rate": 6.506664160980686e-05, "loss": 1.7509, "step": 46310 }, { "epoch": 4.01, "learning_rate": 6.506447322770903e-05, "loss": 1.7329, "step": 46320 }, { "epoch": 4.01, "learning_rate": 6.506230484561119e-05, "loss": 1.6821, "step": 46330 }, { "epoch": 4.01, "learning_rate": 6.506013646351336e-05, "loss": 1.7366, "step": 46340 }, { "epoch": 4.01, "learning_rate": 6.505796808141551e-05, "loss": 1.6746, "step": 46350 }, { "epoch": 4.02, "learning_rate": 6.505579969931768e-05, "loss": 1.697, "step": 46360 }, { "epoch": 4.02, "learning_rate": 6.505363131721983e-05, "loss": 1.8418, "step": 46370 }, { "epoch": 4.02, "learning_rate": 6.5051462935122e-05, "loss": 1.4449, "step": 46380 }, { "epoch": 4.02, "learning_rate": 6.504929455302416e-05, "loss": 1.6838, "step": 46390 }, { "epoch": 4.02, "learning_rate": 6.504712617092633e-05, "loss": 1.6762, "step": 46400 }, { "epoch": 4.02, "learning_rate": 6.50449577888285e-05, "loss": 1.7622, "step": 46410 }, { "epoch": 4.02, "learning_rate": 6.504278940673065e-05, "loss": 1.6684, "step": 46420 }, { "epoch": 4.02, "learning_rate": 6.504062102463282e-05, "loss": 1.8579, "step": 46430 }, { "epoch": 4.02, "learning_rate": 6.503845264253497e-05, "loss": 1.6523, "step": 46440 }, { "epoch": 4.02, "learning_rate": 6.503628426043713e-05, "loss": 1.5823, "step": 46450 }, { "epoch": 4.02, "learning_rate": 6.50341158783393e-05, "loss": 1.4789, "step": 46460 }, { "epoch": 4.02, "learning_rate": 6.503194749624147e-05, "loss": 1.76, "step": 46470 }, { "epoch": 4.03, "learning_rate": 6.502977911414363e-05, "loss": 1.64, "step": 46480 }, { "epoch": 4.03, "learning_rate": 6.502761073204579e-05, "loss": 1.5479, "step": 46490 }, { "epoch": 4.03, "learning_rate": 6.502544234994795e-05, "loss": 1.6338, "step": 46500 }, { "epoch": 4.03, "learning_rate": 6.50232739678501e-05, "loss": 1.7635, "step": 46510 }, { "epoch": 4.03, "learning_rate": 6.502110558575227e-05, "loss": 1.6208, "step": 46520 }, { "epoch": 4.03, "learning_rate": 6.501893720365444e-05, "loss": 1.6252, "step": 46530 }, { "epoch": 4.03, "learning_rate": 6.50167688215566e-05, "loss": 1.8642, "step": 46540 }, { "epoch": 4.03, "learning_rate": 6.501460043945877e-05, "loss": 1.6315, "step": 46550 }, { "epoch": 4.03, "learning_rate": 6.501243205736092e-05, "loss": 1.6024, "step": 46560 }, { "epoch": 4.03, "learning_rate": 6.501026367526309e-05, "loss": 1.6642, "step": 46570 }, { "epoch": 4.03, "learning_rate": 6.500809529316526e-05, "loss": 1.6463, "step": 46580 }, { "epoch": 4.03, "learning_rate": 6.500592691106741e-05, "loss": 1.7496, "step": 46590 }, { "epoch": 4.04, "learning_rate": 6.500375852896958e-05, "loss": 1.7753, "step": 46600 }, { "epoch": 4.04, "learning_rate": 6.500159014687174e-05, "loss": 1.8348, "step": 46610 }, { "epoch": 4.04, "learning_rate": 6.499942176477391e-05, "loss": 1.5813, "step": 46620 }, { "epoch": 4.04, "learning_rate": 6.499725338267606e-05, "loss": 1.6522, "step": 46630 }, { "epoch": 4.04, "learning_rate": 6.499508500057823e-05, "loss": 1.8074, "step": 46640 }, { "epoch": 4.04, "learning_rate": 6.49929166184804e-05, "loss": 1.8425, "step": 46650 }, { "epoch": 4.04, "learning_rate": 6.499074823638256e-05, "loss": 1.8128, "step": 46660 }, { "epoch": 4.04, "learning_rate": 6.498857985428471e-05, "loss": 1.9399, "step": 46670 }, { "epoch": 4.04, "learning_rate": 6.498641147218688e-05, "loss": 1.6848, "step": 46680 }, { "epoch": 4.04, "learning_rate": 6.498424309008905e-05, "loss": 1.7073, "step": 46690 }, { "epoch": 4.04, "learning_rate": 6.49820747079912e-05, "loss": 1.7288, "step": 46700 }, { "epoch": 4.05, "learning_rate": 6.497990632589337e-05, "loss": 1.7272, "step": 46710 }, { "epoch": 4.05, "learning_rate": 6.497773794379553e-05, "loss": 1.5846, "step": 46720 }, { "epoch": 4.05, "learning_rate": 6.49755695616977e-05, "loss": 1.8262, "step": 46730 }, { "epoch": 4.05, "learning_rate": 6.497340117959985e-05, "loss": 1.8057, "step": 46740 }, { "epoch": 4.05, "learning_rate": 6.497123279750202e-05, "loss": 1.6784, "step": 46750 }, { "epoch": 4.05, "learning_rate": 6.496906441540418e-05, "loss": 1.6148, "step": 46760 }, { "epoch": 4.05, "learning_rate": 6.496689603330634e-05, "loss": 1.6237, "step": 46770 }, { "epoch": 4.05, "learning_rate": 6.49647276512085e-05, "loss": 1.5283, "step": 46780 }, { "epoch": 4.05, "learning_rate": 6.496255926911067e-05, "loss": 1.7187, "step": 46790 }, { "epoch": 4.05, "learning_rate": 6.496039088701284e-05, "loss": 1.8341, "step": 46800 }, { "epoch": 4.05, "learning_rate": 6.4958222504915e-05, "loss": 1.6213, "step": 46810 }, { "epoch": 4.05, "learning_rate": 6.495605412281716e-05, "loss": 1.9722, "step": 46820 }, { "epoch": 4.06, "learning_rate": 6.495388574071932e-05, "loss": 1.6241, "step": 46830 }, { "epoch": 4.06, "learning_rate": 6.495171735862147e-05, "loss": 1.7721, "step": 46840 }, { "epoch": 4.06, "learning_rate": 6.494954897652364e-05, "loss": 1.6335, "step": 46850 }, { "epoch": 4.06, "learning_rate": 6.494738059442581e-05, "loss": 1.7641, "step": 46860 }, { "epoch": 4.06, "learning_rate": 6.494521221232797e-05, "loss": 1.8621, "step": 46870 }, { "epoch": 4.06, "learning_rate": 6.494304383023014e-05, "loss": 1.788, "step": 46880 }, { "epoch": 4.06, "learning_rate": 6.494087544813229e-05, "loss": 1.9769, "step": 46890 }, { "epoch": 4.06, "learning_rate": 6.493870706603446e-05, "loss": 1.7511, "step": 46900 }, { "epoch": 4.06, "learning_rate": 6.493653868393661e-05, "loss": 1.8511, "step": 46910 }, { "epoch": 4.06, "learning_rate": 6.493437030183878e-05, "loss": 1.7708, "step": 46920 }, { "epoch": 4.06, "learning_rate": 6.493220191974095e-05, "loss": 1.6726, "step": 46930 }, { "epoch": 4.07, "learning_rate": 6.493003353764311e-05, "loss": 1.7336, "step": 46940 }, { "epoch": 4.07, "learning_rate": 6.492786515554528e-05, "loss": 1.7534, "step": 46950 }, { "epoch": 4.07, "learning_rate": 6.492569677344743e-05, "loss": 1.7714, "step": 46960 }, { "epoch": 4.07, "learning_rate": 6.49235283913496e-05, "loss": 1.8594, "step": 46970 }, { "epoch": 4.07, "learning_rate": 6.492136000925175e-05, "loss": 1.5249, "step": 46980 }, { "epoch": 4.07, "learning_rate": 6.491919162715392e-05, "loss": 1.6604, "step": 46990 }, { "epoch": 4.07, "learning_rate": 6.491702324505608e-05, "loss": 1.3909, "step": 47000 }, { "epoch": 4.07, "learning_rate": 6.491485486295825e-05, "loss": 1.2112, "step": 47010 }, { "epoch": 4.07, "learning_rate": 6.491268648086042e-05, "loss": 1.8431, "step": 47020 }, { "epoch": 4.07, "learning_rate": 6.491051809876257e-05, "loss": 1.5966, "step": 47030 }, { "epoch": 4.07, "learning_rate": 6.490834971666473e-05, "loss": 1.8027, "step": 47040 }, { "epoch": 4.07, "learning_rate": 6.490618133456689e-05, "loss": 1.7701, "step": 47050 }, { "epoch": 4.08, "learning_rate": 6.490401295246905e-05, "loss": 1.5118, "step": 47060 }, { "epoch": 4.08, "learning_rate": 6.490184457037122e-05, "loss": 1.6179, "step": 47070 }, { "epoch": 4.08, "learning_rate": 6.489967618827339e-05, "loss": 1.7648, "step": 47080 }, { "epoch": 4.08, "learning_rate": 6.489750780617555e-05, "loss": 1.6371, "step": 47090 }, { "epoch": 4.08, "learning_rate": 6.48953394240777e-05, "loss": 1.6751, "step": 47100 }, { "epoch": 4.08, "learning_rate": 6.489317104197987e-05, "loss": 1.8629, "step": 47110 }, { "epoch": 4.08, "learning_rate": 6.489100265988204e-05, "loss": 1.5111, "step": 47120 }, { "epoch": 4.08, "learning_rate": 6.488883427778419e-05, "loss": 1.6842, "step": 47130 }, { "epoch": 4.08, "learning_rate": 6.488666589568636e-05, "loss": 1.4874, "step": 47140 }, { "epoch": 4.08, "learning_rate": 6.488449751358852e-05, "loss": 1.457, "step": 47150 }, { "epoch": 4.08, "learning_rate": 6.488232913149069e-05, "loss": 1.7972, "step": 47160 }, { "epoch": 4.09, "learning_rate": 6.488016074939284e-05, "loss": 1.8935, "step": 47170 }, { "epoch": 4.09, "learning_rate": 6.487799236729501e-05, "loss": 1.8518, "step": 47180 }, { "epoch": 4.09, "learning_rate": 6.487582398519718e-05, "loss": 1.4932, "step": 47190 }, { "epoch": 4.09, "learning_rate": 6.487365560309933e-05, "loss": 1.7015, "step": 47200 }, { "epoch": 4.09, "learning_rate": 6.48714872210015e-05, "loss": 1.7554, "step": 47210 }, { "epoch": 4.09, "learning_rate": 6.486931883890366e-05, "loss": 1.8723, "step": 47220 }, { "epoch": 4.09, "learning_rate": 6.486715045680583e-05, "loss": 1.548, "step": 47230 }, { "epoch": 4.09, "learning_rate": 6.486498207470798e-05, "loss": 1.7926, "step": 47240 }, { "epoch": 4.09, "learning_rate": 6.486281369261015e-05, "loss": 1.5902, "step": 47250 }, { "epoch": 4.09, "learning_rate": 6.486064531051231e-05, "loss": 1.4953, "step": 47260 }, { "epoch": 4.09, "learning_rate": 6.485847692841448e-05, "loss": 1.722, "step": 47270 }, { "epoch": 4.09, "learning_rate": 6.485630854631663e-05, "loss": 1.677, "step": 47280 }, { "epoch": 4.1, "learning_rate": 6.48541401642188e-05, "loss": 1.7368, "step": 47290 }, { "epoch": 4.1, "learning_rate": 6.485197178212097e-05, "loss": 1.7152, "step": 47300 }, { "epoch": 4.1, "learning_rate": 6.484980340002312e-05, "loss": 1.7363, "step": 47310 }, { "epoch": 4.1, "learning_rate": 6.484763501792529e-05, "loss": 1.89, "step": 47320 }, { "epoch": 4.1, "learning_rate": 6.484546663582745e-05, "loss": 1.8138, "step": 47330 }, { "epoch": 4.1, "learning_rate": 6.484329825372962e-05, "loss": 1.6617, "step": 47340 }, { "epoch": 4.1, "learning_rate": 6.484112987163178e-05, "loss": 1.6626, "step": 47350 }, { "epoch": 4.1, "learning_rate": 6.483896148953394e-05, "loss": 1.7791, "step": 47360 }, { "epoch": 4.1, "learning_rate": 6.483679310743609e-05, "loss": 1.5642, "step": 47370 }, { "epoch": 4.1, "learning_rate": 6.483462472533826e-05, "loss": 1.6834, "step": 47380 }, { "epoch": 4.1, "learning_rate": 6.483245634324042e-05, "loss": 1.9385, "step": 47390 }, { "epoch": 4.11, "learning_rate": 6.483028796114259e-05, "loss": 1.8921, "step": 47400 }, { "epoch": 4.11, "learning_rate": 6.482811957904476e-05, "loss": 1.933, "step": 47410 }, { "epoch": 4.11, "learning_rate": 6.482595119694691e-05, "loss": 1.7962, "step": 47420 }, { "epoch": 4.11, "learning_rate": 6.482378281484907e-05, "loss": 1.6337, "step": 47430 }, { "epoch": 4.11, "learning_rate": 6.482161443275123e-05, "loss": 1.8577, "step": 47440 }, { "epoch": 4.11, "learning_rate": 6.48194460506534e-05, "loss": 1.6953, "step": 47450 }, { "epoch": 4.11, "learning_rate": 6.481727766855556e-05, "loss": 1.7199, "step": 47460 }, { "epoch": 4.11, "learning_rate": 6.481510928645773e-05, "loss": 1.6773, "step": 47470 }, { "epoch": 4.11, "learning_rate": 6.48129409043599e-05, "loss": 1.8103, "step": 47480 }, { "epoch": 4.11, "learning_rate": 6.481077252226205e-05, "loss": 1.8317, "step": 47490 }, { "epoch": 4.11, "learning_rate": 6.480860414016421e-05, "loss": 1.7821, "step": 47500 }, { "epoch": 4.11, "learning_rate": 6.480643575806637e-05, "loss": 1.8113, "step": 47510 }, { "epoch": 4.12, "learning_rate": 6.480426737596853e-05, "loss": 2.0115, "step": 47520 }, { "epoch": 4.12, "learning_rate": 6.48020989938707e-05, "loss": 1.6726, "step": 47530 }, { "epoch": 4.12, "learning_rate": 6.479993061177286e-05, "loss": 1.4158, "step": 47540 }, { "epoch": 4.12, "learning_rate": 6.479776222967503e-05, "loss": 1.7793, "step": 47550 }, { "epoch": 4.12, "learning_rate": 6.479559384757718e-05, "loss": 1.887, "step": 47560 }, { "epoch": 4.12, "learning_rate": 6.479342546547935e-05, "loss": 1.7333, "step": 47570 }, { "epoch": 4.12, "learning_rate": 6.479125708338152e-05, "loss": 1.7914, "step": 47580 }, { "epoch": 4.12, "learning_rate": 6.478908870128367e-05, "loss": 1.783, "step": 47590 }, { "epoch": 4.12, "learning_rate": 6.478692031918584e-05, "loss": 1.5297, "step": 47600 }, { "epoch": 4.12, "learning_rate": 6.4784751937088e-05, "loss": 1.3792, "step": 47610 }, { "epoch": 4.12, "learning_rate": 6.478258355499017e-05, "loss": 1.5717, "step": 47620 }, { "epoch": 4.13, "learning_rate": 6.478041517289232e-05, "loss": 1.9421, "step": 47630 }, { "epoch": 4.13, "learning_rate": 6.477824679079449e-05, "loss": 1.7434, "step": 47640 }, { "epoch": 4.13, "learning_rate": 6.477607840869665e-05, "loss": 1.7033, "step": 47650 }, { "epoch": 4.13, "learning_rate": 6.477391002659882e-05, "loss": 1.5745, "step": 47660 }, { "epoch": 4.13, "learning_rate": 6.477174164450097e-05, "loss": 1.6994, "step": 47670 }, { "epoch": 4.13, "learning_rate": 6.476957326240314e-05, "loss": 1.6853, "step": 47680 }, { "epoch": 4.13, "learning_rate": 6.47674048803053e-05, "loss": 1.5366, "step": 47690 }, { "epoch": 4.13, "learning_rate": 6.476523649820746e-05, "loss": 1.5118, "step": 47700 }, { "epoch": 4.13, "learning_rate": 6.476306811610963e-05, "loss": 1.6372, "step": 47710 }, { "epoch": 4.13, "learning_rate": 6.476089973401179e-05, "loss": 1.8013, "step": 47720 }, { "epoch": 4.13, "learning_rate": 6.475873135191396e-05, "loss": 1.8141, "step": 47730 }, { "epoch": 4.13, "learning_rate": 6.475656296981611e-05, "loss": 1.8507, "step": 47740 }, { "epoch": 4.14, "learning_rate": 6.475439458771828e-05, "loss": 1.6047, "step": 47750 }, { "epoch": 4.14, "learning_rate": 6.475222620562044e-05, "loss": 1.839, "step": 47760 }, { "epoch": 4.14, "learning_rate": 6.47500578235226e-05, "loss": 1.7439, "step": 47770 }, { "epoch": 4.14, "learning_rate": 6.474788944142476e-05, "loss": 1.5469, "step": 47780 }, { "epoch": 4.14, "learning_rate": 6.474572105932693e-05, "loss": 1.7598, "step": 47790 }, { "epoch": 4.14, "learning_rate": 6.47435526772291e-05, "loss": 1.7271, "step": 47800 }, { "epoch": 4.14, "learning_rate": 6.474138429513126e-05, "loss": 1.5528, "step": 47810 }, { "epoch": 4.14, "learning_rate": 6.473921591303342e-05, "loss": 1.6727, "step": 47820 }, { "epoch": 4.14, "learning_rate": 6.473704753093558e-05, "loss": 1.8928, "step": 47830 }, { "epoch": 4.14, "learning_rate": 6.473487914883773e-05, "loss": 1.6585, "step": 47840 }, { "epoch": 4.14, "learning_rate": 6.47327107667399e-05, "loss": 1.5773, "step": 47850 }, { "epoch": 4.14, "learning_rate": 6.473054238464207e-05, "loss": 1.7535, "step": 47860 }, { "epoch": 4.15, "learning_rate": 6.472837400254423e-05, "loss": 1.8111, "step": 47870 }, { "epoch": 4.15, "learning_rate": 6.47262056204464e-05, "loss": 1.903, "step": 47880 }, { "epoch": 4.15, "learning_rate": 6.472403723834855e-05, "loss": 1.6171, "step": 47890 }, { "epoch": 4.15, "learning_rate": 6.472186885625072e-05, "loss": 1.6122, "step": 47900 }, { "epoch": 4.15, "learning_rate": 6.471970047415287e-05, "loss": 1.707, "step": 47910 }, { "epoch": 4.15, "learning_rate": 6.471753209205504e-05, "loss": 1.6959, "step": 47920 }, { "epoch": 4.15, "learning_rate": 6.47153637099572e-05, "loss": 1.7343, "step": 47930 }, { "epoch": 4.15, "learning_rate": 6.471319532785937e-05, "loss": 1.7164, "step": 47940 }, { "epoch": 4.15, "learning_rate": 6.471102694576154e-05, "loss": 1.5685, "step": 47950 }, { "epoch": 4.15, "learning_rate": 6.470885856366369e-05, "loss": 1.6775, "step": 47960 }, { "epoch": 4.15, "learning_rate": 6.470669018156586e-05, "loss": 1.7098, "step": 47970 }, { "epoch": 4.16, "learning_rate": 6.470452179946801e-05, "loss": 1.8608, "step": 47980 }, { "epoch": 4.16, "learning_rate": 6.470235341737018e-05, "loss": 2.072, "step": 47990 }, { "epoch": 4.16, "learning_rate": 6.470018503527234e-05, "loss": 1.7213, "step": 48000 }, { "epoch": 4.16, "learning_rate": 6.469801665317451e-05, "loss": 1.5366, "step": 48010 }, { "epoch": 4.16, "learning_rate": 6.469584827107668e-05, "loss": 1.6194, "step": 48020 }, { "epoch": 4.16, "learning_rate": 6.469367988897883e-05, "loss": 1.6637, "step": 48030 }, { "epoch": 4.16, "learning_rate": 6.4691511506881e-05, "loss": 1.7342, "step": 48040 }, { "epoch": 4.16, "learning_rate": 6.468934312478315e-05, "loss": 1.9233, "step": 48050 }, { "epoch": 4.16, "learning_rate": 6.468717474268531e-05, "loss": 1.4581, "step": 48060 }, { "epoch": 4.16, "learning_rate": 6.468500636058748e-05, "loss": 1.674, "step": 48070 }, { "epoch": 4.16, "learning_rate": 6.468283797848965e-05, "loss": 1.5768, "step": 48080 }, { "epoch": 4.16, "learning_rate": 6.468066959639181e-05, "loss": 1.3424, "step": 48090 }, { "epoch": 4.17, "learning_rate": 6.467850121429397e-05, "loss": 1.6763, "step": 48100 }, { "epoch": 4.17, "learning_rate": 6.467633283219613e-05, "loss": 1.7389, "step": 48110 }, { "epoch": 4.17, "learning_rate": 6.46741644500983e-05, "loss": 1.7083, "step": 48120 }, { "epoch": 4.17, "learning_rate": 6.467199606800045e-05, "loss": 1.5244, "step": 48130 }, { "epoch": 4.17, "learning_rate": 6.466982768590262e-05, "loss": 1.6034, "step": 48140 }, { "epoch": 4.17, "learning_rate": 6.466765930380478e-05, "loss": 1.6566, "step": 48150 }, { "epoch": 4.17, "learning_rate": 6.466549092170695e-05, "loss": 1.7177, "step": 48160 }, { "epoch": 4.17, "learning_rate": 6.46633225396091e-05, "loss": 1.6357, "step": 48170 }, { "epoch": 4.17, "learning_rate": 6.466115415751127e-05, "loss": 1.7135, "step": 48180 }, { "epoch": 4.17, "learning_rate": 6.465898577541344e-05, "loss": 1.6845, "step": 48190 }, { "epoch": 4.17, "learning_rate": 6.46568173933156e-05, "loss": 1.7587, "step": 48200 }, { "epoch": 4.18, "learning_rate": 6.465464901121776e-05, "loss": 1.5967, "step": 48210 }, { "epoch": 4.18, "learning_rate": 6.465248062911992e-05, "loss": 2.0111, "step": 48220 }, { "epoch": 4.18, "learning_rate": 6.465031224702209e-05, "loss": 1.6399, "step": 48230 }, { "epoch": 4.18, "learning_rate": 6.464814386492424e-05, "loss": 1.6374, "step": 48240 }, { "epoch": 4.18, "learning_rate": 6.464597548282641e-05, "loss": 1.5482, "step": 48250 }, { "epoch": 4.18, "learning_rate": 6.464380710072857e-05, "loss": 1.5737, "step": 48260 }, { "epoch": 4.18, "learning_rate": 6.464163871863074e-05, "loss": 1.7284, "step": 48270 }, { "epoch": 4.18, "learning_rate": 6.463947033653289e-05, "loss": 1.6164, "step": 48280 }, { "epoch": 4.18, "learning_rate": 6.463730195443506e-05, "loss": 1.7878, "step": 48290 }, { "epoch": 4.18, "learning_rate": 6.463513357233723e-05, "loss": 1.7557, "step": 48300 }, { "epoch": 4.18, "learning_rate": 6.463296519023938e-05, "loss": 1.5833, "step": 48310 }, { "epoch": 4.18, "learning_rate": 6.463079680814155e-05, "loss": 1.7175, "step": 48320 }, { "epoch": 4.19, "learning_rate": 6.462862842604371e-05, "loss": 1.5762, "step": 48330 }, { "epoch": 4.19, "learning_rate": 6.462646004394588e-05, "loss": 1.6842, "step": 48340 }, { "epoch": 4.19, "learning_rate": 6.462429166184804e-05, "loss": 1.8439, "step": 48350 }, { "epoch": 4.19, "learning_rate": 6.46221232797502e-05, "loss": 1.6309, "step": 48360 }, { "epoch": 4.19, "learning_rate": 6.461995489765236e-05, "loss": 1.7072, "step": 48370 }, { "epoch": 4.19, "learning_rate": 6.461778651555452e-05, "loss": 1.6447, "step": 48380 }, { "epoch": 4.19, "learning_rate": 6.461561813345668e-05, "loss": 1.8015, "step": 48390 }, { "epoch": 4.19, "learning_rate": 6.461344975135885e-05, "loss": 1.4878, "step": 48400 }, { "epoch": 4.19, "learning_rate": 6.461128136926102e-05, "loss": 1.729, "step": 48410 }, { "epoch": 4.19, "learning_rate": 6.460911298716318e-05, "loss": 1.5581, "step": 48420 }, { "epoch": 4.19, "learning_rate": 6.460694460506533e-05, "loss": 1.7383, "step": 48430 }, { "epoch": 4.2, "learning_rate": 6.460477622296749e-05, "loss": 1.6627, "step": 48440 }, { "epoch": 4.2, "learning_rate": 6.460260784086965e-05, "loss": 1.4584, "step": 48450 }, { "epoch": 4.2, "learning_rate": 6.460043945877182e-05, "loss": 1.6931, "step": 48460 }, { "epoch": 4.2, "learning_rate": 6.459827107667399e-05, "loss": 1.7089, "step": 48470 }, { "epoch": 4.2, "learning_rate": 6.459610269457615e-05, "loss": 1.8005, "step": 48480 }, { "epoch": 4.2, "learning_rate": 6.45939343124783e-05, "loss": 1.7132, "step": 48490 }, { "epoch": 4.2, "learning_rate": 6.459176593038047e-05, "loss": 1.8143, "step": 48500 }, { "epoch": 4.2, "learning_rate": 6.458959754828263e-05, "loss": 1.6413, "step": 48510 }, { "epoch": 4.2, "learning_rate": 6.458742916618479e-05, "loss": 1.7776, "step": 48520 }, { "epoch": 4.2, "learning_rate": 6.458526078408696e-05, "loss": 1.4763, "step": 48530 }, { "epoch": 4.2, "learning_rate": 6.458309240198912e-05, "loss": 1.8122, "step": 48540 }, { "epoch": 4.2, "learning_rate": 6.458092401989129e-05, "loss": 1.5498, "step": 48550 }, { "epoch": 4.21, "learning_rate": 6.457875563779344e-05, "loss": 1.7378, "step": 48560 }, { "epoch": 4.21, "learning_rate": 6.457658725569561e-05, "loss": 1.7233, "step": 48570 }, { "epoch": 4.21, "learning_rate": 6.457441887359778e-05, "loss": 1.6535, "step": 48580 }, { "epoch": 4.21, "learning_rate": 6.457225049149993e-05, "loss": 1.6161, "step": 48590 }, { "epoch": 4.21, "learning_rate": 6.45700821094021e-05, "loss": 1.5996, "step": 48600 }, { "epoch": 4.21, "learning_rate": 6.456791372730426e-05, "loss": 1.7727, "step": 48610 }, { "epoch": 4.21, "learning_rate": 6.456574534520643e-05, "loss": 1.5743, "step": 48620 }, { "epoch": 4.21, "learning_rate": 6.456357696310858e-05, "loss": 1.8793, "step": 48630 }, { "epoch": 4.21, "learning_rate": 6.456140858101075e-05, "loss": 1.8392, "step": 48640 }, { "epoch": 4.21, "learning_rate": 6.455924019891291e-05, "loss": 1.5912, "step": 48650 }, { "epoch": 4.21, "learning_rate": 6.455707181681508e-05, "loss": 1.6932, "step": 48660 }, { "epoch": 4.22, "learning_rate": 6.455490343471723e-05, "loss": 1.8291, "step": 48670 }, { "epoch": 4.22, "learning_rate": 6.45527350526194e-05, "loss": 1.7523, "step": 48680 }, { "epoch": 4.22, "learning_rate": 6.455056667052157e-05, "loss": 1.6844, "step": 48690 }, { "epoch": 4.22, "learning_rate": 6.454839828842372e-05, "loss": 1.8883, "step": 48700 }, { "epoch": 4.22, "learning_rate": 6.454622990632589e-05, "loss": 1.7277, "step": 48710 }, { "epoch": 4.22, "learning_rate": 6.454406152422805e-05, "loss": 1.6442, "step": 48720 }, { "epoch": 4.22, "learning_rate": 6.454189314213022e-05, "loss": 1.7891, "step": 48730 }, { "epoch": 4.22, "learning_rate": 6.453972476003237e-05, "loss": 1.7433, "step": 48740 }, { "epoch": 4.22, "learning_rate": 6.453755637793454e-05, "loss": 1.7159, "step": 48750 }, { "epoch": 4.22, "learning_rate": 6.45353879958367e-05, "loss": 1.8497, "step": 48760 }, { "epoch": 4.22, "learning_rate": 6.453321961373886e-05, "loss": 1.8356, "step": 48770 }, { "epoch": 4.22, "learning_rate": 6.453105123164102e-05, "loss": 1.7481, "step": 48780 }, { "epoch": 4.23, "learning_rate": 6.452888284954319e-05, "loss": 1.74, "step": 48790 }, { "epoch": 4.23, "learning_rate": 6.452671446744536e-05, "loss": 1.8699, "step": 48800 }, { "epoch": 4.23, "learning_rate": 6.452454608534752e-05, "loss": 1.4069, "step": 48810 }, { "epoch": 4.23, "learning_rate": 6.452237770324967e-05, "loss": 1.5075, "step": 48820 }, { "epoch": 4.23, "learning_rate": 6.452020932115184e-05, "loss": 1.7512, "step": 48830 }, { "epoch": 4.23, "learning_rate": 6.4518040939054e-05, "loss": 1.7861, "step": 48840 }, { "epoch": 4.23, "learning_rate": 6.451587255695616e-05, "loss": 1.6671, "step": 48850 }, { "epoch": 4.23, "learning_rate": 6.451370417485833e-05, "loss": 1.8039, "step": 48860 }, { "epoch": 4.23, "learning_rate": 6.45115357927605e-05, "loss": 1.6843, "step": 48870 }, { "epoch": 4.23, "learning_rate": 6.450936741066266e-05, "loss": 1.5834, "step": 48880 }, { "epoch": 4.23, "learning_rate": 6.450719902856481e-05, "loss": 1.6469, "step": 48890 }, { "epoch": 4.24, "learning_rate": 6.450503064646698e-05, "loss": 1.882, "step": 48900 }, { "epoch": 4.24, "learning_rate": 6.450286226436913e-05, "loss": 1.7444, "step": 48910 }, { "epoch": 4.24, "learning_rate": 6.45006938822713e-05, "loss": 1.674, "step": 48920 }, { "epoch": 4.24, "learning_rate": 6.449852550017346e-05, "loss": 1.6221, "step": 48930 }, { "epoch": 4.24, "learning_rate": 6.449635711807563e-05, "loss": 1.5456, "step": 48940 }, { "epoch": 4.24, "learning_rate": 6.44941887359778e-05, "loss": 1.8153, "step": 48950 }, { "epoch": 4.24, "learning_rate": 6.449202035387995e-05, "loss": 1.7345, "step": 48960 }, { "epoch": 4.24, "learning_rate": 6.448985197178212e-05, "loss": 1.7958, "step": 48970 }, { "epoch": 4.24, "learning_rate": 6.448768358968427e-05, "loss": 1.6067, "step": 48980 }, { "epoch": 4.24, "learning_rate": 6.448551520758644e-05, "loss": 1.4905, "step": 48990 }, { "epoch": 4.24, "learning_rate": 6.44833468254886e-05, "loss": 1.7614, "step": 49000 }, { "epoch": 4.24, "learning_rate": 6.448117844339077e-05, "loss": 1.9184, "step": 49010 }, { "epoch": 4.25, "learning_rate": 6.447901006129293e-05, "loss": 1.8291, "step": 49020 }, { "epoch": 4.25, "learning_rate": 6.447684167919509e-05, "loss": 1.5483, "step": 49030 }, { "epoch": 4.25, "learning_rate": 6.447467329709725e-05, "loss": 1.7293, "step": 49040 }, { "epoch": 4.25, "learning_rate": 6.447250491499941e-05, "loss": 1.7356, "step": 49050 }, { "epoch": 4.25, "learning_rate": 6.447033653290157e-05, "loss": 1.944, "step": 49060 }, { "epoch": 4.25, "learning_rate": 6.446816815080374e-05, "loss": 1.5648, "step": 49070 }, { "epoch": 4.25, "learning_rate": 6.44659997687059e-05, "loss": 1.7782, "step": 49080 }, { "epoch": 4.25, "learning_rate": 6.446383138660807e-05, "loss": 1.7623, "step": 49090 }, { "epoch": 4.25, "learning_rate": 6.446166300451023e-05, "loss": 1.7741, "step": 49100 }, { "epoch": 4.25, "learning_rate": 6.445949462241239e-05, "loss": 1.808, "step": 49110 }, { "epoch": 4.25, "learning_rate": 6.445732624031456e-05, "loss": 1.8155, "step": 49120 }, { "epoch": 4.25, "learning_rate": 6.445515785821671e-05, "loss": 1.8337, "step": 49130 }, { "epoch": 4.26, "learning_rate": 6.445298947611888e-05, "loss": 1.4429, "step": 49140 }, { "epoch": 4.26, "learning_rate": 6.445082109402104e-05, "loss": 1.7742, "step": 49150 }, { "epoch": 4.26, "learning_rate": 6.444865271192321e-05, "loss": 1.7114, "step": 49160 }, { "epoch": 4.26, "learning_rate": 6.444648432982536e-05, "loss": 1.7808, "step": 49170 }, { "epoch": 4.26, "learning_rate": 6.444431594772753e-05, "loss": 1.8971, "step": 49180 }, { "epoch": 4.26, "learning_rate": 6.44421475656297e-05, "loss": 1.6747, "step": 49190 }, { "epoch": 4.26, "learning_rate": 6.443997918353186e-05, "loss": 1.7296, "step": 49200 }, { "epoch": 4.26, "learning_rate": 6.443781080143402e-05, "loss": 1.6376, "step": 49210 }, { "epoch": 4.26, "learning_rate": 6.443564241933618e-05, "loss": 1.6866, "step": 49220 }, { "epoch": 4.26, "learning_rate": 6.443347403723835e-05, "loss": 1.7197, "step": 49230 }, { "epoch": 4.26, "learning_rate": 6.44313056551405e-05, "loss": 1.8655, "step": 49240 }, { "epoch": 4.27, "learning_rate": 6.442913727304267e-05, "loss": 1.658, "step": 49250 }, { "epoch": 4.27, "learning_rate": 6.442696889094483e-05, "loss": 1.6346, "step": 49260 }, { "epoch": 4.27, "learning_rate": 6.4424800508847e-05, "loss": 1.7122, "step": 49270 }, { "epoch": 4.27, "learning_rate": 6.442263212674915e-05, "loss": 1.779, "step": 49280 }, { "epoch": 4.27, "learning_rate": 6.442046374465132e-05, "loss": 1.8117, "step": 49290 }, { "epoch": 4.27, "learning_rate": 6.441829536255349e-05, "loss": 1.6545, "step": 49300 }, { "epoch": 4.27, "learning_rate": 6.441612698045564e-05, "loss": 1.5474, "step": 49310 }, { "epoch": 4.27, "learning_rate": 6.44139585983578e-05, "loss": 1.6887, "step": 49320 }, { "epoch": 4.27, "learning_rate": 6.441179021625997e-05, "loss": 1.8189, "step": 49330 }, { "epoch": 4.27, "learning_rate": 6.440962183416214e-05, "loss": 1.5636, "step": 49340 }, { "epoch": 4.27, "learning_rate": 6.44074534520643e-05, "loss": 1.7707, "step": 49350 }, { "epoch": 4.27, "learning_rate": 6.440528506996646e-05, "loss": 1.7255, "step": 49360 }, { "epoch": 4.28, "learning_rate": 6.440311668786862e-05, "loss": 1.5121, "step": 49370 }, { "epoch": 4.28, "learning_rate": 6.440094830577078e-05, "loss": 1.6954, "step": 49380 }, { "epoch": 4.28, "learning_rate": 6.439877992367294e-05, "loss": 1.5033, "step": 49390 }, { "epoch": 4.28, "learning_rate": 6.439661154157511e-05, "loss": 1.5774, "step": 49400 }, { "epoch": 4.28, "learning_rate": 6.439444315947728e-05, "loss": 1.8322, "step": 49410 }, { "epoch": 4.28, "learning_rate": 6.439227477737944e-05, "loss": 1.7889, "step": 49420 }, { "epoch": 4.28, "learning_rate": 6.43901063952816e-05, "loss": 1.6854, "step": 49430 }, { "epoch": 4.28, "learning_rate": 6.438793801318376e-05, "loss": 1.7168, "step": 49440 }, { "epoch": 4.28, "learning_rate": 6.438576963108591e-05, "loss": 1.6451, "step": 49450 }, { "epoch": 4.28, "learning_rate": 6.438360124898808e-05, "loss": 1.7512, "step": 49460 }, { "epoch": 4.28, "learning_rate": 6.438143286689025e-05, "loss": 1.6941, "step": 49470 }, { "epoch": 4.29, "learning_rate": 6.437926448479241e-05, "loss": 1.7233, "step": 49480 }, { "epoch": 4.29, "learning_rate": 6.437709610269458e-05, "loss": 1.5407, "step": 49490 }, { "epoch": 4.29, "learning_rate": 6.437492772059673e-05, "loss": 1.7527, "step": 49500 }, { "epoch": 4.29, "learning_rate": 6.437275933849888e-05, "loss": 1.6291, "step": 49510 }, { "epoch": 4.29, "learning_rate": 6.437059095640105e-05, "loss": 1.4308, "step": 49520 }, { "epoch": 4.29, "learning_rate": 6.436842257430322e-05, "loss": 1.724, "step": 49530 }, { "epoch": 4.29, "learning_rate": 6.436625419220538e-05, "loss": 1.6751, "step": 49540 }, { "epoch": 4.29, "learning_rate": 6.436408581010755e-05, "loss": 1.6412, "step": 49550 }, { "epoch": 4.29, "learning_rate": 6.43619174280097e-05, "loss": 1.544, "step": 49560 }, { "epoch": 4.29, "learning_rate": 6.435974904591187e-05, "loss": 1.7655, "step": 49570 }, { "epoch": 4.29, "learning_rate": 6.435758066381404e-05, "loss": 1.7127, "step": 49580 }, { "epoch": 4.29, "learning_rate": 6.435541228171619e-05, "loss": 1.6439, "step": 49590 }, { "epoch": 4.3, "learning_rate": 6.435324389961836e-05, "loss": 1.8921, "step": 49600 }, { "epoch": 4.3, "learning_rate": 6.435107551752052e-05, "loss": 1.5115, "step": 49610 }, { "epoch": 4.3, "learning_rate": 6.434890713542269e-05, "loss": 1.9608, "step": 49620 }, { "epoch": 4.3, "learning_rate": 6.434673875332484e-05, "loss": 1.8758, "step": 49630 }, { "epoch": 4.3, "learning_rate": 6.434457037122701e-05, "loss": 1.9533, "step": 49640 }, { "epoch": 4.3, "learning_rate": 6.434240198912917e-05, "loss": 1.7319, "step": 49650 }, { "epoch": 4.3, "learning_rate": 6.434023360703134e-05, "loss": 1.5621, "step": 49660 }, { "epoch": 4.3, "learning_rate": 6.433806522493349e-05, "loss": 1.6052, "step": 49670 }, { "epoch": 4.3, "learning_rate": 6.433589684283566e-05, "loss": 1.6489, "step": 49680 }, { "epoch": 4.3, "learning_rate": 6.433372846073783e-05, "loss": 1.8283, "step": 49690 }, { "epoch": 4.3, "learning_rate": 6.433156007863998e-05, "loss": 1.485, "step": 49700 }, { "epoch": 4.31, "learning_rate": 6.432939169654214e-05, "loss": 1.7135, "step": 49710 }, { "epoch": 4.31, "learning_rate": 6.432722331444431e-05, "loss": 1.6848, "step": 49720 }, { "epoch": 4.31, "learning_rate": 6.432505493234648e-05, "loss": 1.8001, "step": 49730 }, { "epoch": 4.31, "learning_rate": 6.432288655024864e-05, "loss": 1.6874, "step": 49740 }, { "epoch": 4.31, "learning_rate": 6.43207181681508e-05, "loss": 1.4919, "step": 49750 }, { "epoch": 4.31, "learning_rate": 6.431854978605296e-05, "loss": 1.5185, "step": 49760 }, { "epoch": 4.31, "learning_rate": 6.431638140395512e-05, "loss": 1.6064, "step": 49770 }, { "epoch": 4.31, "learning_rate": 6.431421302185728e-05, "loss": 1.7201, "step": 49780 }, { "epoch": 4.31, "learning_rate": 6.431204463975945e-05, "loss": 1.9275, "step": 49790 }, { "epoch": 4.31, "learning_rate": 6.430987625766162e-05, "loss": 1.725, "step": 49800 }, { "epoch": 4.31, "learning_rate": 6.430770787556378e-05, "loss": 1.7649, "step": 49810 }, { "epoch": 4.31, "learning_rate": 6.430553949346593e-05, "loss": 1.7074, "step": 49820 }, { "epoch": 4.32, "learning_rate": 6.43033711113681e-05, "loss": 1.7293, "step": 49830 }, { "epoch": 4.32, "learning_rate": 6.430120272927025e-05, "loss": 1.5155, "step": 49840 }, { "epoch": 4.32, "learning_rate": 6.429903434717242e-05, "loss": 1.6651, "step": 49850 }, { "epoch": 4.32, "learning_rate": 6.429686596507459e-05, "loss": 1.5706, "step": 49860 }, { "epoch": 4.32, "learning_rate": 6.429469758297675e-05, "loss": 1.9265, "step": 49870 }, { "epoch": 4.32, "learning_rate": 6.429252920087892e-05, "loss": 1.7097, "step": 49880 }, { "epoch": 4.32, "learning_rate": 6.429036081878107e-05, "loss": 1.557, "step": 49890 }, { "epoch": 4.32, "learning_rate": 6.428819243668324e-05, "loss": 1.6778, "step": 49900 }, { "epoch": 4.32, "learning_rate": 6.428602405458539e-05, "loss": 1.7717, "step": 49910 }, { "epoch": 4.32, "learning_rate": 6.428385567248756e-05, "loss": 1.6878, "step": 49920 }, { "epoch": 4.32, "learning_rate": 6.428168729038972e-05, "loss": 1.7521, "step": 49930 }, { "epoch": 4.33, "learning_rate": 6.427951890829189e-05, "loss": 1.6586, "step": 49940 }, { "epoch": 4.33, "learning_rate": 6.427735052619406e-05, "loss": 1.5305, "step": 49950 }, { "epoch": 4.33, "learning_rate": 6.427518214409621e-05, "loss": 1.5703, "step": 49960 }, { "epoch": 4.33, "learning_rate": 6.427301376199838e-05, "loss": 1.7851, "step": 49970 }, { "epoch": 4.33, "learning_rate": 6.427084537990053e-05, "loss": 1.7372, "step": 49980 }, { "epoch": 4.33, "learning_rate": 6.42686769978027e-05, "loss": 1.8235, "step": 49990 }, { "epoch": 4.33, "learning_rate": 6.426650861570486e-05, "loss": 1.7241, "step": 50000 }, { "epoch": 4.33, "learning_rate": 6.426434023360703e-05, "loss": 1.8828, "step": 50010 }, { "epoch": 4.33, "learning_rate": 6.42621718515092e-05, "loss": 1.7567, "step": 50020 }, { "epoch": 4.33, "learning_rate": 6.426000346941135e-05, "loss": 1.7063, "step": 50030 }, { "epoch": 4.33, "learning_rate": 6.425783508731351e-05, "loss": 1.7139, "step": 50040 }, { "epoch": 4.33, "learning_rate": 6.425566670521567e-05, "loss": 1.7851, "step": 50050 }, { "epoch": 4.34, "learning_rate": 6.425349832311783e-05, "loss": 1.8837, "step": 50060 }, { "epoch": 4.34, "learning_rate": 6.425132994102e-05, "loss": 1.5853, "step": 50070 }, { "epoch": 4.34, "learning_rate": 6.424916155892217e-05, "loss": 1.7876, "step": 50080 }, { "epoch": 4.34, "learning_rate": 6.424699317682433e-05, "loss": 1.6967, "step": 50090 }, { "epoch": 4.34, "learning_rate": 6.424482479472649e-05, "loss": 1.7002, "step": 50100 }, { "epoch": 4.34, "learning_rate": 6.424265641262865e-05, "loss": 1.4995, "step": 50110 }, { "epoch": 4.34, "learning_rate": 6.424048803053082e-05, "loss": 1.7483, "step": 50120 }, { "epoch": 4.34, "learning_rate": 6.423831964843297e-05, "loss": 1.4766, "step": 50130 }, { "epoch": 4.34, "learning_rate": 6.423615126633514e-05, "loss": 1.8085, "step": 50140 }, { "epoch": 4.34, "learning_rate": 6.42339828842373e-05, "loss": 1.7211, "step": 50150 }, { "epoch": 4.34, "learning_rate": 6.423181450213947e-05, "loss": 1.7451, "step": 50160 }, { "epoch": 4.35, "learning_rate": 6.422964612004162e-05, "loss": 1.645, "step": 50170 }, { "epoch": 4.35, "learning_rate": 6.422747773794379e-05, "loss": 1.7146, "step": 50180 }, { "epoch": 4.35, "learning_rate": 6.422530935584596e-05, "loss": 1.61, "step": 50190 }, { "epoch": 4.35, "learning_rate": 6.422314097374812e-05, "loss": 1.5367, "step": 50200 }, { "epoch": 4.35, "learning_rate": 6.422097259165027e-05, "loss": 1.8528, "step": 50210 }, { "epoch": 4.35, "learning_rate": 6.421880420955244e-05, "loss": 1.5472, "step": 50220 }, { "epoch": 4.35, "learning_rate": 6.421663582745461e-05, "loss": 1.7969, "step": 50230 }, { "epoch": 4.35, "learning_rate": 6.421446744535676e-05, "loss": 1.7292, "step": 50240 }, { "epoch": 4.35, "learning_rate": 6.421229906325893e-05, "loss": 1.4417, "step": 50250 }, { "epoch": 4.35, "learning_rate": 6.421013068116109e-05, "loss": 1.6261, "step": 50260 }, { "epoch": 4.35, "learning_rate": 6.420796229906326e-05, "loss": 1.7451, "step": 50270 }, { "epoch": 4.35, "learning_rate": 6.420579391696541e-05, "loss": 1.7447, "step": 50280 }, { "epoch": 4.36, "learning_rate": 6.420362553486758e-05, "loss": 1.7729, "step": 50290 }, { "epoch": 4.36, "learning_rate": 6.420145715276975e-05, "loss": 1.5538, "step": 50300 }, { "epoch": 4.36, "learning_rate": 6.41992887706719e-05, "loss": 1.74, "step": 50310 }, { "epoch": 4.36, "learning_rate": 6.419712038857406e-05, "loss": 1.7611, "step": 50320 }, { "epoch": 4.36, "learning_rate": 6.419495200647623e-05, "loss": 1.7482, "step": 50330 }, { "epoch": 4.36, "learning_rate": 6.41927836243784e-05, "loss": 1.6326, "step": 50340 }, { "epoch": 4.36, "learning_rate": 6.419061524228056e-05, "loss": 1.7399, "step": 50350 }, { "epoch": 4.36, "learning_rate": 6.418844686018272e-05, "loss": 1.8116, "step": 50360 }, { "epoch": 4.36, "learning_rate": 6.418627847808488e-05, "loss": 1.7319, "step": 50370 }, { "epoch": 4.36, "learning_rate": 6.418411009598704e-05, "loss": 1.6399, "step": 50380 }, { "epoch": 4.36, "learning_rate": 6.41819417138892e-05, "loss": 1.6012, "step": 50390 }, { "epoch": 4.36, "learning_rate": 6.417977333179137e-05, "loss": 1.6162, "step": 50400 }, { "epoch": 4.37, "learning_rate": 6.417760494969353e-05, "loss": 1.5953, "step": 50410 }, { "epoch": 4.37, "learning_rate": 6.41754365675957e-05, "loss": 1.4889, "step": 50420 }, { "epoch": 4.37, "learning_rate": 6.417326818549785e-05, "loss": 1.8248, "step": 50430 }, { "epoch": 4.37, "learning_rate": 6.417109980340002e-05, "loss": 1.7499, "step": 50440 }, { "epoch": 4.37, "learning_rate": 6.416893142130217e-05, "loss": 1.4961, "step": 50450 }, { "epoch": 4.37, "learning_rate": 6.416676303920434e-05, "loss": 1.6688, "step": 50460 }, { "epoch": 4.37, "learning_rate": 6.41645946571065e-05, "loss": 1.7818, "step": 50470 }, { "epoch": 4.37, "learning_rate": 6.416242627500867e-05, "loss": 1.7719, "step": 50480 }, { "epoch": 4.37, "learning_rate": 6.416025789291084e-05, "loss": 1.7769, "step": 50490 }, { "epoch": 4.37, "learning_rate": 6.415808951081299e-05, "loss": 1.5034, "step": 50500 }, { "epoch": 4.37, "learning_rate": 6.415592112871516e-05, "loss": 1.6305, "step": 50510 }, { "epoch": 4.38, "learning_rate": 6.415375274661731e-05, "loss": 1.899, "step": 50520 }, { "epoch": 4.38, "learning_rate": 6.415158436451948e-05, "loss": 1.5772, "step": 50530 }, { "epoch": 4.38, "learning_rate": 6.414941598242164e-05, "loss": 1.9117, "step": 50540 }, { "epoch": 4.38, "learning_rate": 6.414724760032381e-05, "loss": 1.9044, "step": 50550 }, { "epoch": 4.38, "learning_rate": 6.414507921822598e-05, "loss": 1.855, "step": 50560 }, { "epoch": 4.38, "learning_rate": 6.414291083612813e-05, "loss": 1.8283, "step": 50570 }, { "epoch": 4.38, "learning_rate": 6.41407424540303e-05, "loss": 1.9706, "step": 50580 }, { "epoch": 4.38, "learning_rate": 6.413857407193245e-05, "loss": 1.7229, "step": 50590 }, { "epoch": 4.38, "learning_rate": 6.413640568983461e-05, "loss": 1.6602, "step": 50600 }, { "epoch": 4.38, "learning_rate": 6.413423730773678e-05, "loss": 1.6444, "step": 50610 }, { "epoch": 4.38, "learning_rate": 6.413206892563895e-05, "loss": 1.6075, "step": 50620 }, { "epoch": 4.38, "learning_rate": 6.41299005435411e-05, "loss": 1.4646, "step": 50630 }, { "epoch": 4.39, "learning_rate": 6.412773216144327e-05, "loss": 1.6938, "step": 50640 }, { "epoch": 4.39, "learning_rate": 6.412556377934543e-05, "loss": 1.6058, "step": 50650 }, { "epoch": 4.39, "learning_rate": 6.41233953972476e-05, "loss": 1.7363, "step": 50660 }, { "epoch": 4.39, "learning_rate": 6.412122701514975e-05, "loss": 1.6719, "step": 50670 }, { "epoch": 4.39, "learning_rate": 6.411905863305192e-05, "loss": 1.7656, "step": 50680 }, { "epoch": 4.39, "learning_rate": 6.411689025095409e-05, "loss": 1.6243, "step": 50690 }, { "epoch": 4.39, "learning_rate": 6.411472186885624e-05, "loss": 1.514, "step": 50700 }, { "epoch": 4.39, "learning_rate": 6.41125534867584e-05, "loss": 1.918, "step": 50710 }, { "epoch": 4.39, "learning_rate": 6.411038510466057e-05, "loss": 1.537, "step": 50720 }, { "epoch": 4.39, "learning_rate": 6.410821672256274e-05, "loss": 1.8083, "step": 50730 }, { "epoch": 4.39, "learning_rate": 6.41060483404649e-05, "loss": 1.7364, "step": 50740 }, { "epoch": 4.4, "learning_rate": 6.410387995836706e-05, "loss": 1.646, "step": 50750 }, { "epoch": 4.4, "learning_rate": 6.410171157626922e-05, "loss": 1.6659, "step": 50760 }, { "epoch": 4.4, "learning_rate": 6.409954319417138e-05, "loss": 1.6899, "step": 50770 }, { "epoch": 4.4, "learning_rate": 6.409737481207354e-05, "loss": 1.739, "step": 50780 }, { "epoch": 4.4, "learning_rate": 6.409520642997571e-05, "loss": 1.7301, "step": 50790 }, { "epoch": 4.4, "learning_rate": 6.409303804787788e-05, "loss": 1.4235, "step": 50800 }, { "epoch": 4.4, "learning_rate": 6.409086966578004e-05, "loss": 1.4667, "step": 50810 }, { "epoch": 4.4, "learning_rate": 6.40887012836822e-05, "loss": 1.8066, "step": 50820 }, { "epoch": 4.4, "learning_rate": 6.408653290158436e-05, "loss": 1.6078, "step": 50830 }, { "epoch": 4.4, "learning_rate": 6.408436451948651e-05, "loss": 1.6064, "step": 50840 }, { "epoch": 4.4, "learning_rate": 6.408219613738868e-05, "loss": 1.6957, "step": 50850 }, { "epoch": 4.4, "learning_rate": 6.408002775529085e-05, "loss": 1.5787, "step": 50860 }, { "epoch": 4.41, "learning_rate": 6.407785937319301e-05, "loss": 1.7158, "step": 50870 }, { "epoch": 4.41, "learning_rate": 6.407569099109518e-05, "loss": 1.5717, "step": 50880 }, { "epoch": 4.41, "learning_rate": 6.407352260899733e-05, "loss": 1.6324, "step": 50890 }, { "epoch": 4.41, "learning_rate": 6.40713542268995e-05, "loss": 1.4828, "step": 50900 }, { "epoch": 4.41, "learning_rate": 6.406918584480165e-05, "loss": 1.7911, "step": 50910 }, { "epoch": 4.41, "learning_rate": 6.406701746270382e-05, "loss": 1.7992, "step": 50920 }, { "epoch": 4.41, "learning_rate": 6.406484908060598e-05, "loss": 1.6877, "step": 50930 }, { "epoch": 4.41, "learning_rate": 6.406268069850815e-05, "loss": 1.5191, "step": 50940 }, { "epoch": 4.41, "learning_rate": 6.406051231641032e-05, "loss": 1.8321, "step": 50950 }, { "epoch": 4.41, "learning_rate": 6.405834393431247e-05, "loss": 1.6125, "step": 50960 }, { "epoch": 4.41, "learning_rate": 6.405617555221464e-05, "loss": 1.6082, "step": 50970 }, { "epoch": 4.42, "learning_rate": 6.405400717011679e-05, "loss": 1.6134, "step": 50980 }, { "epoch": 4.42, "learning_rate": 6.405183878801896e-05, "loss": 1.6634, "step": 50990 }, { "epoch": 4.42, "learning_rate": 6.404967040592112e-05, "loss": 1.655, "step": 51000 }, { "epoch": 4.42, "learning_rate": 6.404750202382329e-05, "loss": 1.787, "step": 51010 }, { "epoch": 4.42, "learning_rate": 6.404533364172545e-05, "loss": 1.7178, "step": 51020 }, { "epoch": 4.42, "learning_rate": 6.404316525962761e-05, "loss": 1.6855, "step": 51030 }, { "epoch": 4.42, "learning_rate": 6.404099687752977e-05, "loss": 1.5155, "step": 51040 }, { "epoch": 4.42, "learning_rate": 6.403882849543193e-05, "loss": 1.6648, "step": 51050 }, { "epoch": 4.42, "learning_rate": 6.403666011333409e-05, "loss": 1.7174, "step": 51060 }, { "epoch": 4.42, "learning_rate": 6.403449173123626e-05, "loss": 1.4349, "step": 51070 }, { "epoch": 4.42, "learning_rate": 6.403232334913843e-05, "loss": 1.5316, "step": 51080 }, { "epoch": 4.42, "learning_rate": 6.403015496704059e-05, "loss": 1.8591, "step": 51090 }, { "epoch": 4.43, "learning_rate": 6.402798658494274e-05, "loss": 1.5901, "step": 51100 }, { "epoch": 4.43, "learning_rate": 6.402581820284491e-05, "loss": 1.8209, "step": 51110 }, { "epoch": 4.43, "learning_rate": 6.402364982074708e-05, "loss": 1.605, "step": 51120 }, { "epoch": 4.43, "learning_rate": 6.402148143864923e-05, "loss": 1.7714, "step": 51130 }, { "epoch": 4.43, "learning_rate": 6.40193130565514e-05, "loss": 1.6551, "step": 51140 }, { "epoch": 4.43, "learning_rate": 6.401714467445356e-05, "loss": 1.6637, "step": 51150 }, { "epoch": 4.43, "learning_rate": 6.401497629235573e-05, "loss": 1.8207, "step": 51160 }, { "epoch": 4.43, "learning_rate": 6.401280791025788e-05, "loss": 1.9185, "step": 51170 }, { "epoch": 4.43, "learning_rate": 6.401063952816005e-05, "loss": 1.5809, "step": 51180 }, { "epoch": 4.43, "learning_rate": 6.400847114606222e-05, "loss": 1.7902, "step": 51190 }, { "epoch": 4.43, "learning_rate": 6.400630276396438e-05, "loss": 1.8389, "step": 51200 }, { "epoch": 4.44, "learning_rate": 6.400413438186653e-05, "loss": 1.7229, "step": 51210 }, { "epoch": 4.44, "learning_rate": 6.40019659997687e-05, "loss": 1.8262, "step": 51220 }, { "epoch": 4.44, "learning_rate": 6.399979761767087e-05, "loss": 1.7556, "step": 51230 }, { "epoch": 4.44, "learning_rate": 6.399762923557302e-05, "loss": 1.8542, "step": 51240 }, { "epoch": 4.44, "learning_rate": 6.399546085347519e-05, "loss": 1.6218, "step": 51250 }, { "epoch": 4.44, "learning_rate": 6.399329247137735e-05, "loss": 1.5413, "step": 51260 }, { "epoch": 4.44, "learning_rate": 6.399112408927952e-05, "loss": 1.7889, "step": 51270 }, { "epoch": 4.44, "learning_rate": 6.398895570718169e-05, "loss": 1.7139, "step": 51280 }, { "epoch": 4.44, "learning_rate": 6.398678732508384e-05, "loss": 1.5602, "step": 51290 }, { "epoch": 4.44, "learning_rate": 6.3984618942986e-05, "loss": 1.6165, "step": 51300 }, { "epoch": 4.44, "learning_rate": 6.398245056088816e-05, "loss": 1.6512, "step": 51310 }, { "epoch": 4.44, "learning_rate": 6.398028217879032e-05, "loss": 1.6492, "step": 51320 }, { "epoch": 4.45, "learning_rate": 6.397811379669249e-05, "loss": 1.7075, "step": 51330 }, { "epoch": 4.45, "learning_rate": 6.397594541459466e-05, "loss": 1.6298, "step": 51340 }, { "epoch": 4.45, "learning_rate": 6.397377703249682e-05, "loss": 1.9097, "step": 51350 }, { "epoch": 4.45, "learning_rate": 6.397160865039898e-05, "loss": 1.496, "step": 51360 }, { "epoch": 4.45, "learning_rate": 6.396944026830114e-05, "loss": 1.803, "step": 51370 }, { "epoch": 4.45, "learning_rate": 6.39672718862033e-05, "loss": 1.6375, "step": 51380 }, { "epoch": 4.45, "learning_rate": 6.396510350410546e-05, "loss": 1.7325, "step": 51390 }, { "epoch": 4.45, "learning_rate": 6.396293512200763e-05, "loss": 1.8171, "step": 51400 }, { "epoch": 4.45, "learning_rate": 6.39607667399098e-05, "loss": 1.9227, "step": 51410 }, { "epoch": 4.45, "learning_rate": 6.395859835781196e-05, "loss": 1.6591, "step": 51420 }, { "epoch": 4.45, "learning_rate": 6.395642997571411e-05, "loss": 1.5752, "step": 51430 }, { "epoch": 4.46, "learning_rate": 6.395426159361628e-05, "loss": 1.7193, "step": 51440 }, { "epoch": 4.46, "learning_rate": 6.395209321151843e-05, "loss": 1.6702, "step": 51450 }, { "epoch": 4.46, "learning_rate": 6.39499248294206e-05, "loss": 1.4285, "step": 51460 }, { "epoch": 4.46, "learning_rate": 6.394775644732277e-05, "loss": 1.6182, "step": 51470 }, { "epoch": 4.46, "learning_rate": 6.394558806522493e-05, "loss": 1.7275, "step": 51480 }, { "epoch": 4.46, "learning_rate": 6.39434196831271e-05, "loss": 1.6166, "step": 51490 }, { "epoch": 4.46, "learning_rate": 6.394125130102925e-05, "loss": 1.6639, "step": 51500 }, { "epoch": 4.46, "learning_rate": 6.393908291893142e-05, "loss": 1.7044, "step": 51510 }, { "epoch": 4.46, "learning_rate": 6.393691453683357e-05, "loss": 1.6249, "step": 51520 }, { "epoch": 4.46, "learning_rate": 6.393474615473574e-05, "loss": 1.6902, "step": 51530 }, { "epoch": 4.46, "learning_rate": 6.39325777726379e-05, "loss": 1.4305, "step": 51540 }, { "epoch": 4.46, "learning_rate": 6.393040939054007e-05, "loss": 1.489, "step": 51550 }, { "epoch": 4.47, "learning_rate": 6.392824100844224e-05, "loss": 1.8592, "step": 51560 }, { "epoch": 4.47, "learning_rate": 6.392607262634439e-05, "loss": 1.5702, "step": 51570 }, { "epoch": 4.47, "learning_rate": 6.392390424424656e-05, "loss": 1.5878, "step": 51580 }, { "epoch": 4.47, "learning_rate": 6.392173586214871e-05, "loss": 1.8613, "step": 51590 }, { "epoch": 4.47, "learning_rate": 6.391956748005087e-05, "loss": 1.6472, "step": 51600 }, { "epoch": 4.47, "learning_rate": 6.391739909795304e-05, "loss": 1.7325, "step": 51610 }, { "epoch": 4.47, "learning_rate": 6.391523071585521e-05, "loss": 1.8723, "step": 51620 }, { "epoch": 4.47, "learning_rate": 6.391306233375737e-05, "loss": 1.607, "step": 51630 }, { "epoch": 4.47, "learning_rate": 6.391089395165953e-05, "loss": 1.9247, "step": 51640 }, { "epoch": 4.47, "learning_rate": 6.390872556956169e-05, "loss": 1.7059, "step": 51650 }, { "epoch": 4.47, "learning_rate": 6.390655718746386e-05, "loss": 1.8056, "step": 51660 }, { "epoch": 4.47, "learning_rate": 6.390438880536601e-05, "loss": 1.7723, "step": 51670 }, { "epoch": 4.48, "learning_rate": 6.390222042326818e-05, "loss": 1.9412, "step": 51680 }, { "epoch": 4.48, "learning_rate": 6.390005204117035e-05, "loss": 1.4831, "step": 51690 }, { "epoch": 4.48, "learning_rate": 6.38978836590725e-05, "loss": 1.6916, "step": 51700 }, { "epoch": 4.48, "learning_rate": 6.389571527697466e-05, "loss": 1.63, "step": 51710 }, { "epoch": 4.48, "learning_rate": 6.389354689487683e-05, "loss": 1.5913, "step": 51720 }, { "epoch": 4.48, "learning_rate": 6.3891378512779e-05, "loss": 1.6809, "step": 51730 }, { "epoch": 4.48, "learning_rate": 6.388921013068116e-05, "loss": 1.803, "step": 51740 }, { "epoch": 4.48, "learning_rate": 6.388704174858332e-05, "loss": 1.6426, "step": 51750 }, { "epoch": 4.48, "learning_rate": 6.388487336648548e-05, "loss": 1.6507, "step": 51760 }, { "epoch": 4.48, "learning_rate": 6.388270498438764e-05, "loss": 1.5308, "step": 51770 }, { "epoch": 4.48, "learning_rate": 6.38805366022898e-05, "loss": 1.6561, "step": 51780 }, { "epoch": 4.49, "learning_rate": 6.387836822019197e-05, "loss": 1.6887, "step": 51790 }, { "epoch": 4.49, "learning_rate": 6.387619983809413e-05, "loss": 1.8687, "step": 51800 }, { "epoch": 4.49, "learning_rate": 6.38740314559963e-05, "loss": 1.7655, "step": 51810 }, { "epoch": 4.49, "learning_rate": 6.387186307389845e-05, "loss": 1.7575, "step": 51820 }, { "epoch": 4.49, "learning_rate": 6.386969469180062e-05, "loss": 1.731, "step": 51830 }, { "epoch": 4.49, "learning_rate": 6.386752630970277e-05, "loss": 1.5007, "step": 51840 }, { "epoch": 4.49, "learning_rate": 6.386535792760494e-05, "loss": 1.6341, "step": 51850 }, { "epoch": 4.49, "learning_rate": 6.38631895455071e-05, "loss": 1.7098, "step": 51860 }, { "epoch": 4.49, "learning_rate": 6.386102116340927e-05, "loss": 1.5783, "step": 51870 }, { "epoch": 4.49, "learning_rate": 6.385885278131144e-05, "loss": 1.7038, "step": 51880 }, { "epoch": 4.49, "learning_rate": 6.385668439921359e-05, "loss": 1.8019, "step": 51890 }, { "epoch": 4.49, "learning_rate": 6.385451601711576e-05, "loss": 1.7311, "step": 51900 }, { "epoch": 4.5, "learning_rate": 6.385234763501791e-05, "loss": 1.5791, "step": 51910 }, { "epoch": 4.5, "learning_rate": 6.385017925292008e-05, "loss": 1.7337, "step": 51920 }, { "epoch": 4.5, "learning_rate": 6.384801087082224e-05, "loss": 1.8093, "step": 51930 }, { "epoch": 4.5, "learning_rate": 6.384584248872441e-05, "loss": 2.0347, "step": 51940 }, { "epoch": 4.5, "learning_rate": 6.384367410662658e-05, "loss": 1.7416, "step": 51950 }, { "epoch": 4.5, "learning_rate": 6.384150572452873e-05, "loss": 1.5576, "step": 51960 }, { "epoch": 4.5, "learning_rate": 6.38393373424309e-05, "loss": 1.5949, "step": 51970 }, { "epoch": 4.5, "learning_rate": 6.383716896033305e-05, "loss": 1.8529, "step": 51980 }, { "epoch": 4.5, "learning_rate": 6.383500057823521e-05, "loss": 1.5571, "step": 51990 }, { "epoch": 4.5, "learning_rate": 6.383283219613738e-05, "loss": 1.8446, "step": 52000 }, { "epoch": 4.5, "learning_rate": 6.383066381403955e-05, "loss": 1.53, "step": 52010 }, { "epoch": 4.51, "learning_rate": 6.382849543194171e-05, "loss": 1.7616, "step": 52020 }, { "epoch": 4.51, "learning_rate": 6.382632704984387e-05, "loss": 1.615, "step": 52030 }, { "epoch": 4.51, "learning_rate": 6.382415866774603e-05, "loss": 1.7855, "step": 52040 }, { "epoch": 4.51, "learning_rate": 6.38219902856482e-05, "loss": 1.5123, "step": 52050 }, { "epoch": 4.51, "learning_rate": 6.381982190355035e-05, "loss": 1.5966, "step": 52060 }, { "epoch": 4.51, "learning_rate": 6.381765352145252e-05, "loss": 1.7571, "step": 52070 }, { "epoch": 4.51, "learning_rate": 6.381548513935469e-05, "loss": 1.6092, "step": 52080 }, { "epoch": 4.51, "learning_rate": 6.381331675725685e-05, "loss": 1.8081, "step": 52090 }, { "epoch": 4.51, "learning_rate": 6.3811148375159e-05, "loss": 1.778, "step": 52100 }, { "epoch": 4.51, "learning_rate": 6.380897999306117e-05, "loss": 1.5316, "step": 52110 }, { "epoch": 4.51, "learning_rate": 6.380681161096334e-05, "loss": 1.517, "step": 52120 }, { "epoch": 4.51, "learning_rate": 6.380464322886549e-05, "loss": 1.6844, "step": 52130 }, { "epoch": 4.52, "learning_rate": 6.380247484676766e-05, "loss": 1.653, "step": 52140 }, { "epoch": 4.52, "learning_rate": 6.380030646466982e-05, "loss": 1.7919, "step": 52150 }, { "epoch": 4.52, "learning_rate": 6.379813808257199e-05, "loss": 1.6644, "step": 52160 }, { "epoch": 4.52, "learning_rate": 6.379596970047414e-05, "loss": 1.9233, "step": 52170 }, { "epoch": 4.52, "learning_rate": 6.379380131837631e-05, "loss": 1.9073, "step": 52180 }, { "epoch": 4.52, "learning_rate": 6.379163293627847e-05, "loss": 1.417, "step": 52190 }, { "epoch": 4.52, "learning_rate": 6.378946455418064e-05, "loss": 1.5309, "step": 52200 }, { "epoch": 4.52, "learning_rate": 6.37872961720828e-05, "loss": 1.7557, "step": 52210 }, { "epoch": 4.52, "learning_rate": 6.378512778998496e-05, "loss": 1.7926, "step": 52220 }, { "epoch": 4.52, "learning_rate": 6.378295940788713e-05, "loss": 1.6114, "step": 52230 }, { "epoch": 4.52, "learning_rate": 6.378079102578928e-05, "loss": 1.6715, "step": 52240 }, { "epoch": 4.53, "learning_rate": 6.377862264369145e-05, "loss": 1.7386, "step": 52250 }, { "epoch": 4.53, "learning_rate": 6.377645426159361e-05, "loss": 1.6139, "step": 52260 }, { "epoch": 4.53, "learning_rate": 6.377428587949578e-05, "loss": 1.7129, "step": 52270 }, { "epoch": 4.53, "learning_rate": 6.377211749739795e-05, "loss": 1.7128, "step": 52280 }, { "epoch": 4.53, "learning_rate": 6.37699491153001e-05, "loss": 1.5982, "step": 52290 }, { "epoch": 4.53, "learning_rate": 6.376778073320226e-05, "loss": 1.7482, "step": 52300 }, { "epoch": 4.53, "learning_rate": 6.376561235110442e-05, "loss": 1.6398, "step": 52310 }, { "epoch": 4.53, "learning_rate": 6.376344396900658e-05, "loss": 1.5963, "step": 52320 }, { "epoch": 4.53, "learning_rate": 6.376127558690875e-05, "loss": 1.7509, "step": 52330 }, { "epoch": 4.53, "learning_rate": 6.375910720481092e-05, "loss": 1.8104, "step": 52340 }, { "epoch": 4.53, "learning_rate": 6.375693882271308e-05, "loss": 1.6826, "step": 52350 }, { "epoch": 4.53, "learning_rate": 6.375477044061524e-05, "loss": 1.7418, "step": 52360 }, { "epoch": 4.54, "learning_rate": 6.37526020585174e-05, "loss": 1.7509, "step": 52370 }, { "epoch": 4.54, "learning_rate": 6.375043367641956e-05, "loss": 1.8596, "step": 52380 }, { "epoch": 4.54, "learning_rate": 6.374826529432172e-05, "loss": 1.7341, "step": 52390 }, { "epoch": 4.54, "learning_rate": 6.374609691222389e-05, "loss": 1.7148, "step": 52400 }, { "epoch": 4.54, "learning_rate": 6.374392853012605e-05, "loss": 1.6028, "step": 52410 }, { "epoch": 4.54, "learning_rate": 6.374176014802822e-05, "loss": 1.8986, "step": 52420 }, { "epoch": 4.54, "learning_rate": 6.373959176593037e-05, "loss": 1.6644, "step": 52430 }, { "epoch": 4.54, "learning_rate": 6.373742338383254e-05, "loss": 1.8401, "step": 52440 }, { "epoch": 4.54, "learning_rate": 6.373525500173469e-05, "loss": 1.4931, "step": 52450 }, { "epoch": 4.54, "learning_rate": 6.373308661963686e-05, "loss": 1.5303, "step": 52460 }, { "epoch": 4.54, "learning_rate": 6.373091823753903e-05, "loss": 1.5307, "step": 52470 }, { "epoch": 4.55, "learning_rate": 6.372874985544119e-05, "loss": 1.6971, "step": 52480 }, { "epoch": 4.55, "learning_rate": 6.372658147334336e-05, "loss": 1.8229, "step": 52490 }, { "epoch": 4.55, "learning_rate": 6.372441309124551e-05, "loss": 1.6388, "step": 52500 }, { "epoch": 4.55, "learning_rate": 6.372224470914768e-05, "loss": 1.4415, "step": 52510 }, { "epoch": 4.55, "learning_rate": 6.372007632704983e-05, "loss": 1.8501, "step": 52520 }, { "epoch": 4.55, "learning_rate": 6.3717907944952e-05, "loss": 1.6578, "step": 52530 }, { "epoch": 4.55, "learning_rate": 6.371573956285416e-05, "loss": 1.702, "step": 52540 }, { "epoch": 4.55, "learning_rate": 6.371357118075633e-05, "loss": 1.6596, "step": 52550 }, { "epoch": 4.55, "learning_rate": 6.37114027986585e-05, "loss": 1.6511, "step": 52560 }, { "epoch": 4.55, "learning_rate": 6.370923441656065e-05, "loss": 1.5857, "step": 52570 }, { "epoch": 4.55, "learning_rate": 6.370706603446282e-05, "loss": 1.6225, "step": 52580 }, { "epoch": 4.55, "learning_rate": 6.370489765236497e-05, "loss": 1.4662, "step": 52590 }, { "epoch": 4.56, "learning_rate": 6.370272927026713e-05, "loss": 1.6824, "step": 52600 }, { "epoch": 4.56, "learning_rate": 6.37005608881693e-05, "loss": 1.4977, "step": 52610 }, { "epoch": 4.56, "learning_rate": 6.369839250607147e-05, "loss": 1.5387, "step": 52620 }, { "epoch": 4.56, "learning_rate": 6.369622412397363e-05, "loss": 1.7445, "step": 52630 }, { "epoch": 4.56, "learning_rate": 6.369405574187579e-05, "loss": 1.7026, "step": 52640 }, { "epoch": 4.56, "learning_rate": 6.369188735977795e-05, "loss": 1.6843, "step": 52650 }, { "epoch": 4.56, "learning_rate": 6.368971897768012e-05, "loss": 1.5926, "step": 52660 }, { "epoch": 4.56, "learning_rate": 6.368755059558227e-05, "loss": 1.7056, "step": 52670 }, { "epoch": 4.56, "learning_rate": 6.368538221348444e-05, "loss": 1.6462, "step": 52680 }, { "epoch": 4.56, "learning_rate": 6.36832138313866e-05, "loss": 1.6529, "step": 52690 }, { "epoch": 4.56, "learning_rate": 6.368104544928877e-05, "loss": 1.8539, "step": 52700 }, { "epoch": 4.57, "learning_rate": 6.367887706719092e-05, "loss": 1.4848, "step": 52710 }, { "epoch": 4.57, "learning_rate": 6.367670868509309e-05, "loss": 1.5619, "step": 52720 }, { "epoch": 4.57, "learning_rate": 6.367454030299526e-05, "loss": 1.7927, "step": 52730 }, { "epoch": 4.57, "learning_rate": 6.367237192089742e-05, "loss": 1.7453, "step": 52740 }, { "epoch": 4.57, "learning_rate": 6.367020353879958e-05, "loss": 1.694, "step": 52750 }, { "epoch": 4.57, "learning_rate": 6.366803515670174e-05, "loss": 1.6519, "step": 52760 }, { "epoch": 4.57, "learning_rate": 6.36658667746039e-05, "loss": 1.9366, "step": 52770 }, { "epoch": 4.57, "learning_rate": 6.366369839250606e-05, "loss": 1.7674, "step": 52780 }, { "epoch": 4.57, "learning_rate": 6.366153001040823e-05, "loss": 1.4343, "step": 52790 }, { "epoch": 4.57, "learning_rate": 6.36593616283104e-05, "loss": 1.9463, "step": 52800 }, { "epoch": 4.57, "learning_rate": 6.365719324621256e-05, "loss": 1.6354, "step": 52810 }, { "epoch": 4.57, "learning_rate": 6.365502486411471e-05, "loss": 1.6322, "step": 52820 }, { "epoch": 4.58, "learning_rate": 6.365285648201688e-05, "loss": 1.49, "step": 52830 }, { "epoch": 4.58, "learning_rate": 6.365068809991903e-05, "loss": 1.7942, "step": 52840 }, { "epoch": 4.58, "learning_rate": 6.36485197178212e-05, "loss": 1.6242, "step": 52850 }, { "epoch": 4.58, "learning_rate": 6.364635133572337e-05, "loss": 1.7138, "step": 52860 }, { "epoch": 4.58, "learning_rate": 6.364418295362553e-05, "loss": 1.4442, "step": 52870 }, { "epoch": 4.58, "learning_rate": 6.36420145715277e-05, "loss": 1.8166, "step": 52880 }, { "epoch": 4.58, "learning_rate": 6.363984618942985e-05, "loss": 1.3895, "step": 52890 }, { "epoch": 4.58, "learning_rate": 6.363767780733202e-05, "loss": 1.7584, "step": 52900 }, { "epoch": 4.58, "learning_rate": 6.363550942523417e-05, "loss": 1.8081, "step": 52910 }, { "epoch": 4.58, "learning_rate": 6.363334104313634e-05, "loss": 1.7295, "step": 52920 }, { "epoch": 4.58, "learning_rate": 6.36311726610385e-05, "loss": 1.7093, "step": 52930 }, { "epoch": 4.58, "learning_rate": 6.362900427894067e-05, "loss": 1.7998, "step": 52940 }, { "epoch": 4.59, "learning_rate": 6.362683589684284e-05, "loss": 1.8898, "step": 52950 }, { "epoch": 4.59, "learning_rate": 6.362466751474499e-05, "loss": 1.614, "step": 52960 }, { "epoch": 4.59, "learning_rate": 6.362249913264716e-05, "loss": 1.7079, "step": 52970 }, { "epoch": 4.59, "learning_rate": 6.362033075054931e-05, "loss": 1.774, "step": 52980 }, { "epoch": 4.59, "learning_rate": 6.361816236845147e-05, "loss": 1.7526, "step": 52990 }, { "epoch": 4.59, "learning_rate": 6.361599398635364e-05, "loss": 1.5946, "step": 53000 }, { "epoch": 4.59, "learning_rate": 6.361382560425581e-05, "loss": 1.7605, "step": 53010 }, { "epoch": 4.59, "learning_rate": 6.361165722215797e-05, "loss": 1.5021, "step": 53020 }, { "epoch": 4.59, "learning_rate": 6.360948884006013e-05, "loss": 1.6559, "step": 53030 }, { "epoch": 4.59, "learning_rate": 6.360732045796229e-05, "loss": 1.7921, "step": 53040 }, { "epoch": 4.59, "learning_rate": 6.360515207586446e-05, "loss": 1.8141, "step": 53050 }, { "epoch": 4.6, "learning_rate": 6.360298369376661e-05, "loss": 1.7029, "step": 53060 }, { "epoch": 4.6, "learning_rate": 6.360081531166878e-05, "loss": 1.7161, "step": 53070 }, { "epoch": 4.6, "learning_rate": 6.359864692957094e-05, "loss": 1.6441, "step": 53080 }, { "epoch": 4.6, "learning_rate": 6.359647854747311e-05, "loss": 1.8579, "step": 53090 }, { "epoch": 4.6, "learning_rate": 6.359431016537526e-05, "loss": 1.486, "step": 53100 }, { "epoch": 4.6, "learning_rate": 6.359214178327743e-05, "loss": 1.7037, "step": 53110 }, { "epoch": 4.6, "learning_rate": 6.35899734011796e-05, "loss": 1.7326, "step": 53120 }, { "epoch": 4.6, "learning_rate": 6.358780501908175e-05, "loss": 1.686, "step": 53130 }, { "epoch": 4.6, "learning_rate": 6.358563663698392e-05, "loss": 1.5685, "step": 53140 }, { "epoch": 4.6, "learning_rate": 6.358346825488608e-05, "loss": 1.5866, "step": 53150 }, { "epoch": 4.6, "learning_rate": 6.358129987278825e-05, "loss": 1.5871, "step": 53160 }, { "epoch": 4.6, "learning_rate": 6.35791314906904e-05, "loss": 1.7098, "step": 53170 }, { "epoch": 4.61, "learning_rate": 6.357696310859257e-05, "loss": 1.6275, "step": 53180 }, { "epoch": 4.61, "learning_rate": 6.357479472649473e-05, "loss": 1.5281, "step": 53190 }, { "epoch": 4.61, "learning_rate": 6.35726263443969e-05, "loss": 1.4845, "step": 53200 }, { "epoch": 4.61, "learning_rate": 6.357045796229905e-05, "loss": 1.7861, "step": 53210 }, { "epoch": 4.61, "learning_rate": 6.356828958020122e-05, "loss": 1.6383, "step": 53220 }, { "epoch": 4.61, "learning_rate": 6.356612119810339e-05, "loss": 1.8719, "step": 53230 }, { "epoch": 4.61, "learning_rate": 6.356395281600554e-05, "loss": 1.752, "step": 53240 }, { "epoch": 4.61, "learning_rate": 6.35617844339077e-05, "loss": 1.6477, "step": 53250 }, { "epoch": 4.61, "learning_rate": 6.355961605180987e-05, "loss": 1.7919, "step": 53260 }, { "epoch": 4.61, "learning_rate": 6.355744766971204e-05, "loss": 1.8798, "step": 53270 }, { "epoch": 4.61, "learning_rate": 6.35552792876142e-05, "loss": 1.829, "step": 53280 }, { "epoch": 4.62, "learning_rate": 6.355311090551636e-05, "loss": 1.6152, "step": 53290 }, { "epoch": 4.62, "learning_rate": 6.355094252341852e-05, "loss": 1.6861, "step": 53300 }, { "epoch": 4.62, "learning_rate": 6.354877414132068e-05, "loss": 1.4953, "step": 53310 }, { "epoch": 4.62, "learning_rate": 6.354660575922284e-05, "loss": 1.87, "step": 53320 }, { "epoch": 4.62, "learning_rate": 6.354443737712501e-05, "loss": 1.5906, "step": 53330 }, { "epoch": 4.62, "learning_rate": 6.354226899502718e-05, "loss": 1.6297, "step": 53340 }, { "epoch": 4.62, "learning_rate": 6.354010061292934e-05, "loss": 1.639, "step": 53350 }, { "epoch": 4.62, "learning_rate": 6.35379322308315e-05, "loss": 1.8174, "step": 53360 }, { "epoch": 4.62, "learning_rate": 6.353576384873366e-05, "loss": 1.6482, "step": 53370 }, { "epoch": 4.62, "learning_rate": 6.353359546663581e-05, "loss": 1.8161, "step": 53380 }, { "epoch": 4.62, "learning_rate": 6.353142708453798e-05, "loss": 1.6967, "step": 53390 }, { "epoch": 4.62, "learning_rate": 6.352925870244015e-05, "loss": 1.5732, "step": 53400 }, { "epoch": 4.63, "learning_rate": 6.352709032034231e-05, "loss": 1.7694, "step": 53410 }, { "epoch": 4.63, "learning_rate": 6.352492193824448e-05, "loss": 1.8247, "step": 53420 }, { "epoch": 4.63, "learning_rate": 6.352275355614663e-05, "loss": 1.7857, "step": 53430 }, { "epoch": 4.63, "learning_rate": 6.35205851740488e-05, "loss": 1.7257, "step": 53440 }, { "epoch": 4.63, "learning_rate": 6.351841679195095e-05, "loss": 1.6068, "step": 53450 }, { "epoch": 4.63, "learning_rate": 6.351624840985312e-05, "loss": 1.6717, "step": 53460 }, { "epoch": 4.63, "learning_rate": 6.351408002775529e-05, "loss": 1.6563, "step": 53470 }, { "epoch": 4.63, "learning_rate": 6.351191164565745e-05, "loss": 1.7115, "step": 53480 }, { "epoch": 4.63, "learning_rate": 6.350974326355962e-05, "loss": 1.7366, "step": 53490 }, { "epoch": 4.63, "learning_rate": 6.350757488146177e-05, "loss": 1.6557, "step": 53500 }, { "epoch": 4.63, "learning_rate": 6.350540649936394e-05, "loss": 1.6035, "step": 53510 }, { "epoch": 4.64, "learning_rate": 6.350323811726609e-05, "loss": 1.53, "step": 53520 }, { "epoch": 4.64, "learning_rate": 6.350106973516826e-05, "loss": 1.5965, "step": 53530 }, { "epoch": 4.64, "learning_rate": 6.349890135307042e-05, "loss": 1.714, "step": 53540 }, { "epoch": 4.64, "learning_rate": 6.349673297097259e-05, "loss": 1.7343, "step": 53550 }, { "epoch": 4.64, "learning_rate": 6.349456458887476e-05, "loss": 1.4978, "step": 53560 }, { "epoch": 4.64, "learning_rate": 6.349239620677691e-05, "loss": 1.5713, "step": 53570 }, { "epoch": 4.64, "learning_rate": 6.349022782467907e-05, "loss": 1.7412, "step": 53580 }, { "epoch": 4.64, "learning_rate": 6.348805944258124e-05, "loss": 1.5193, "step": 53590 }, { "epoch": 4.64, "learning_rate": 6.34858910604834e-05, "loss": 1.4792, "step": 53600 }, { "epoch": 4.64, "learning_rate": 6.348372267838556e-05, "loss": 1.7485, "step": 53610 }, { "epoch": 4.64, "learning_rate": 6.348155429628773e-05, "loss": 1.771, "step": 53620 }, { "epoch": 4.64, "learning_rate": 6.347938591418989e-05, "loss": 1.572, "step": 53630 }, { "epoch": 4.65, "learning_rate": 6.347721753209205e-05, "loss": 1.7566, "step": 53640 }, { "epoch": 4.65, "learning_rate": 6.347504914999421e-05, "loss": 1.5787, "step": 53650 }, { "epoch": 4.65, "learning_rate": 6.347288076789638e-05, "loss": 1.7993, "step": 53660 }, { "epoch": 4.65, "learning_rate": 6.347071238579853e-05, "loss": 1.636, "step": 53670 }, { "epoch": 4.65, "learning_rate": 6.34685440037007e-05, "loss": 1.635, "step": 53680 }, { "epoch": 4.65, "learning_rate": 6.346637562160286e-05, "loss": 1.6937, "step": 53690 }, { "epoch": 4.65, "learning_rate": 6.346420723950503e-05, "loss": 1.5699, "step": 53700 }, { "epoch": 4.65, "learning_rate": 6.346203885740718e-05, "loss": 1.5557, "step": 53710 }, { "epoch": 4.65, "learning_rate": 6.345987047530935e-05, "loss": 1.8502, "step": 53720 }, { "epoch": 4.65, "learning_rate": 6.345770209321152e-05, "loss": 1.5455, "step": 53730 }, { "epoch": 4.65, "learning_rate": 6.345553371111368e-05, "loss": 1.6732, "step": 53740 }, { "epoch": 4.66, "learning_rate": 6.345336532901584e-05, "loss": 1.4614, "step": 53750 }, { "epoch": 4.66, "learning_rate": 6.3451196946918e-05, "loss": 1.8221, "step": 53760 }, { "epoch": 4.66, "learning_rate": 6.344902856482017e-05, "loss": 1.7835, "step": 53770 }, { "epoch": 4.66, "learning_rate": 6.344686018272232e-05, "loss": 1.5556, "step": 53780 }, { "epoch": 4.66, "learning_rate": 6.344469180062449e-05, "loss": 1.5008, "step": 53790 }, { "epoch": 4.66, "learning_rate": 6.344252341852665e-05, "loss": 1.6182, "step": 53800 }, { "epoch": 4.66, "learning_rate": 6.344035503642882e-05, "loss": 1.408, "step": 53810 }, { "epoch": 4.66, "learning_rate": 6.343818665433099e-05, "loss": 1.8056, "step": 53820 }, { "epoch": 4.66, "learning_rate": 6.343601827223314e-05, "loss": 1.5052, "step": 53830 }, { "epoch": 4.66, "learning_rate": 6.343384989013529e-05, "loss": 1.8803, "step": 53840 }, { "epoch": 4.66, "learning_rate": 6.343168150803746e-05, "loss": 1.6992, "step": 53850 }, { "epoch": 4.66, "learning_rate": 6.342951312593963e-05, "loss": 1.5182, "step": 53860 }, { "epoch": 4.67, "learning_rate": 6.342734474384179e-05, "loss": 1.6584, "step": 53870 }, { "epoch": 4.67, "learning_rate": 6.342517636174396e-05, "loss": 1.661, "step": 53880 }, { "epoch": 4.67, "learning_rate": 6.342300797964611e-05, "loss": 1.6621, "step": 53890 }, { "epoch": 4.67, "learning_rate": 6.342083959754828e-05, "loss": 1.6333, "step": 53900 }, { "epoch": 4.67, "learning_rate": 6.341867121545043e-05, "loss": 1.5149, "step": 53910 }, { "epoch": 4.67, "learning_rate": 6.34165028333526e-05, "loss": 1.8133, "step": 53920 }, { "epoch": 4.67, "learning_rate": 6.341433445125476e-05, "loss": 1.6334, "step": 53930 }, { "epoch": 4.67, "learning_rate": 6.341216606915693e-05, "loss": 1.6539, "step": 53940 }, { "epoch": 4.67, "learning_rate": 6.34099976870591e-05, "loss": 1.6745, "step": 53950 }, { "epoch": 4.67, "learning_rate": 6.340782930496125e-05, "loss": 1.56, "step": 53960 }, { "epoch": 4.67, "learning_rate": 6.340566092286341e-05, "loss": 1.8201, "step": 53970 }, { "epoch": 4.68, "learning_rate": 6.340349254076557e-05, "loss": 1.5549, "step": 53980 }, { "epoch": 4.68, "learning_rate": 6.340132415866773e-05, "loss": 1.695, "step": 53990 }, { "epoch": 4.68, "learning_rate": 6.33991557765699e-05, "loss": 1.4995, "step": 54000 }, { "epoch": 4.68, "learning_rate": 6.339698739447207e-05, "loss": 1.674, "step": 54010 }, { "epoch": 4.68, "learning_rate": 6.339481901237423e-05, "loss": 1.9396, "step": 54020 }, { "epoch": 4.68, "learning_rate": 6.339265063027639e-05, "loss": 1.6383, "step": 54030 }, { "epoch": 4.68, "learning_rate": 6.339048224817855e-05, "loss": 1.7342, "step": 54040 }, { "epoch": 4.68, "learning_rate": 6.338831386608072e-05, "loss": 1.6843, "step": 54050 }, { "epoch": 4.68, "learning_rate": 6.338614548398287e-05, "loss": 1.5987, "step": 54060 }, { "epoch": 4.68, "learning_rate": 6.338397710188504e-05, "loss": 1.7521, "step": 54070 }, { "epoch": 4.68, "learning_rate": 6.33818087197872e-05, "loss": 1.6582, "step": 54080 }, { "epoch": 4.68, "learning_rate": 6.337964033768937e-05, "loss": 1.4456, "step": 54090 }, { "epoch": 4.69, "learning_rate": 6.337747195559152e-05, "loss": 1.6987, "step": 54100 }, { "epoch": 4.69, "learning_rate": 6.337530357349369e-05, "loss": 1.5539, "step": 54110 }, { "epoch": 4.69, "learning_rate": 6.337313519139586e-05, "loss": 1.4982, "step": 54120 }, { "epoch": 4.69, "learning_rate": 6.337096680929801e-05, "loss": 1.5267, "step": 54130 }, { "epoch": 4.69, "learning_rate": 6.336879842720018e-05, "loss": 1.6156, "step": 54140 }, { "epoch": 4.69, "learning_rate": 6.336663004510234e-05, "loss": 1.7054, "step": 54150 }, { "epoch": 4.69, "learning_rate": 6.336446166300451e-05, "loss": 1.8462, "step": 54160 }, { "epoch": 4.69, "learning_rate": 6.336229328090666e-05, "loss": 1.3332, "step": 54170 }, { "epoch": 4.69, "learning_rate": 6.336012489880883e-05, "loss": 1.77, "step": 54180 }, { "epoch": 4.69, "learning_rate": 6.3357956516711e-05, "loss": 1.5063, "step": 54190 }, { "epoch": 4.69, "learning_rate": 6.335578813461316e-05, "loss": 1.7703, "step": 54200 }, { "epoch": 4.69, "learning_rate": 6.335361975251531e-05, "loss": 1.3668, "step": 54210 }, { "epoch": 4.7, "learning_rate": 6.335145137041748e-05, "loss": 1.7134, "step": 54220 }, { "epoch": 4.7, "learning_rate": 6.334928298831965e-05, "loss": 1.5147, "step": 54230 }, { "epoch": 4.7, "learning_rate": 6.33471146062218e-05, "loss": 1.707, "step": 54240 }, { "epoch": 4.7, "learning_rate": 6.334494622412397e-05, "loss": 1.8073, "step": 54250 }, { "epoch": 4.7, "learning_rate": 6.334277784202613e-05, "loss": 1.687, "step": 54260 }, { "epoch": 4.7, "learning_rate": 6.33406094599283e-05, "loss": 1.8113, "step": 54270 }, { "epoch": 4.7, "learning_rate": 6.333844107783046e-05, "loss": 1.8511, "step": 54280 }, { "epoch": 4.7, "learning_rate": 6.333627269573262e-05, "loss": 1.5593, "step": 54290 }, { "epoch": 4.7, "learning_rate": 6.333410431363478e-05, "loss": 1.7688, "step": 54300 }, { "epoch": 4.7, "learning_rate": 6.333193593153694e-05, "loss": 1.7254, "step": 54310 }, { "epoch": 4.7, "learning_rate": 6.33297675494391e-05, "loss": 1.7103, "step": 54320 }, { "epoch": 4.71, "learning_rate": 6.332759916734127e-05, "loss": 1.6005, "step": 54330 }, { "epoch": 4.71, "learning_rate": 6.332543078524344e-05, "loss": 1.7729, "step": 54340 }, { "epoch": 4.71, "learning_rate": 6.33232624031456e-05, "loss": 1.6478, "step": 54350 }, { "epoch": 4.71, "learning_rate": 6.332109402104776e-05, "loss": 1.6302, "step": 54360 }, { "epoch": 4.71, "learning_rate": 6.331892563894992e-05, "loss": 1.6714, "step": 54370 }, { "epoch": 4.71, "learning_rate": 6.331675725685207e-05, "loss": 1.7031, "step": 54380 }, { "epoch": 4.71, "learning_rate": 6.331458887475424e-05, "loss": 1.5673, "step": 54390 }, { "epoch": 4.71, "learning_rate": 6.331242049265641e-05, "loss": 1.6126, "step": 54400 }, { "epoch": 4.71, "learning_rate": 6.331025211055857e-05, "loss": 1.452, "step": 54410 }, { "epoch": 4.71, "learning_rate": 6.330808372846074e-05, "loss": 1.7503, "step": 54420 }, { "epoch": 4.71, "learning_rate": 6.330591534636289e-05, "loss": 1.8798, "step": 54430 }, { "epoch": 4.71, "learning_rate": 6.330374696426506e-05, "loss": 1.6348, "step": 54440 }, { "epoch": 4.72, "learning_rate": 6.330157858216721e-05, "loss": 1.6835, "step": 54450 }, { "epoch": 4.72, "learning_rate": 6.329941020006938e-05, "loss": 1.5215, "step": 54460 }, { "epoch": 4.72, "learning_rate": 6.329724181797154e-05, "loss": 1.5965, "step": 54470 }, { "epoch": 4.72, "learning_rate": 6.329507343587371e-05, "loss": 1.7513, "step": 54480 }, { "epoch": 4.72, "learning_rate": 6.329290505377588e-05, "loss": 1.7119, "step": 54490 }, { "epoch": 4.72, "learning_rate": 6.329073667167803e-05, "loss": 1.6915, "step": 54500 }, { "epoch": 4.72, "learning_rate": 6.32885682895802e-05, "loss": 1.6536, "step": 54510 }, { "epoch": 4.72, "learning_rate": 6.328639990748235e-05, "loss": 1.6874, "step": 54520 }, { "epoch": 4.72, "learning_rate": 6.328423152538452e-05, "loss": 1.7328, "step": 54530 }, { "epoch": 4.72, "learning_rate": 6.328206314328668e-05, "loss": 1.4666, "step": 54540 }, { "epoch": 4.72, "learning_rate": 6.327989476118885e-05, "loss": 1.6201, "step": 54550 }, { "epoch": 4.73, "learning_rate": 6.327772637909102e-05, "loss": 1.7295, "step": 54560 }, { "epoch": 4.73, "learning_rate": 6.327555799699317e-05, "loss": 1.5755, "step": 54570 }, { "epoch": 4.73, "learning_rate": 6.327338961489533e-05, "loss": 1.7031, "step": 54580 }, { "epoch": 4.73, "learning_rate": 6.32712212327975e-05, "loss": 1.8252, "step": 54590 }, { "epoch": 4.73, "learning_rate": 6.326905285069965e-05, "loss": 1.5779, "step": 54600 }, { "epoch": 4.73, "learning_rate": 6.326688446860182e-05, "loss": 1.6733, "step": 54610 }, { "epoch": 4.73, "learning_rate": 6.326471608650399e-05, "loss": 1.9787, "step": 54620 }, { "epoch": 4.73, "learning_rate": 6.326254770440615e-05, "loss": 1.8754, "step": 54630 }, { "epoch": 4.73, "learning_rate": 6.32603793223083e-05, "loss": 1.4886, "step": 54640 }, { "epoch": 4.73, "learning_rate": 6.325821094021047e-05, "loss": 1.8619, "step": 54650 }, { "epoch": 4.73, "learning_rate": 6.325604255811264e-05, "loss": 1.458, "step": 54660 }, { "epoch": 4.73, "learning_rate": 6.325387417601479e-05, "loss": 1.7741, "step": 54670 }, { "epoch": 4.74, "learning_rate": 6.325170579391696e-05, "loss": 1.6609, "step": 54680 }, { "epoch": 4.74, "learning_rate": 6.324953741181912e-05, "loss": 1.474, "step": 54690 }, { "epoch": 4.74, "learning_rate": 6.324736902972129e-05, "loss": 1.6192, "step": 54700 }, { "epoch": 4.74, "learning_rate": 6.324520064762344e-05, "loss": 1.6174, "step": 54710 }, { "epoch": 4.74, "learning_rate": 6.324303226552561e-05, "loss": 1.604, "step": 54720 }, { "epoch": 4.74, "learning_rate": 6.324086388342778e-05, "loss": 1.7309, "step": 54730 }, { "epoch": 4.74, "learning_rate": 6.323869550132994e-05, "loss": 1.8702, "step": 54740 }, { "epoch": 4.74, "learning_rate": 6.32365271192321e-05, "loss": 1.4761, "step": 54750 }, { "epoch": 4.74, "learning_rate": 6.323435873713426e-05, "loss": 1.7506, "step": 54760 }, { "epoch": 4.74, "learning_rate": 6.323219035503643e-05, "loss": 1.6373, "step": 54770 }, { "epoch": 4.74, "learning_rate": 6.323002197293858e-05, "loss": 1.6239, "step": 54780 }, { "epoch": 4.75, "learning_rate": 6.322785359084075e-05, "loss": 1.6326, "step": 54790 }, { "epoch": 4.75, "learning_rate": 6.322568520874291e-05, "loss": 1.6936, "step": 54800 }, { "epoch": 4.75, "learning_rate": 6.322351682664508e-05, "loss": 1.5448, "step": 54810 }, { "epoch": 4.75, "learning_rate": 6.322134844454725e-05, "loss": 1.6872, "step": 54820 }, { "epoch": 4.75, "learning_rate": 6.32191800624494e-05, "loss": 1.589, "step": 54830 }, { "epoch": 4.75, "learning_rate": 6.321701168035157e-05, "loss": 1.6114, "step": 54840 }, { "epoch": 4.75, "learning_rate": 6.321484329825372e-05, "loss": 1.662, "step": 54850 }, { "epoch": 4.75, "learning_rate": 6.321267491615589e-05, "loss": 1.7631, "step": 54860 }, { "epoch": 4.75, "learning_rate": 6.321050653405805e-05, "loss": 1.4822, "step": 54870 }, { "epoch": 4.75, "learning_rate": 6.320833815196022e-05, "loss": 1.6204, "step": 54880 }, { "epoch": 4.75, "learning_rate": 6.320616976986238e-05, "loss": 1.5911, "step": 54890 }, { "epoch": 4.75, "learning_rate": 6.320400138776454e-05, "loss": 1.6101, "step": 54900 }, { "epoch": 4.76, "learning_rate": 6.320183300566669e-05, "loss": 1.873, "step": 54910 }, { "epoch": 4.76, "learning_rate": 6.319966462356886e-05, "loss": 1.6435, "step": 54920 }, { "epoch": 4.76, "learning_rate": 6.319749624147102e-05, "loss": 2.0423, "step": 54930 }, { "epoch": 4.76, "learning_rate": 6.319532785937319e-05, "loss": 1.7052, "step": 54940 }, { "epoch": 4.76, "learning_rate": 6.319315947727536e-05, "loss": 1.6429, "step": 54950 }, { "epoch": 4.76, "learning_rate": 6.319099109517751e-05, "loss": 1.802, "step": 54960 }, { "epoch": 4.76, "learning_rate": 6.318882271307967e-05, "loss": 1.8096, "step": 54970 }, { "epoch": 4.76, "learning_rate": 6.318665433098183e-05, "loss": 1.529, "step": 54980 }, { "epoch": 4.76, "learning_rate": 6.3184485948884e-05, "loss": 1.5526, "step": 54990 }, { "epoch": 4.76, "learning_rate": 6.318231756678616e-05, "loss": 1.5769, "step": 55000 }, { "epoch": 4.76, "learning_rate": 6.318014918468833e-05, "loss": 1.8451, "step": 55010 }, { "epoch": 4.77, "learning_rate": 6.317798080259049e-05, "loss": 1.5538, "step": 55020 }, { "epoch": 4.77, "learning_rate": 6.317581242049265e-05, "loss": 1.8058, "step": 55030 }, { "epoch": 4.77, "learning_rate": 6.317364403839481e-05, "loss": 1.8904, "step": 55040 }, { "epoch": 4.77, "learning_rate": 6.317147565629698e-05, "loss": 1.6963, "step": 55050 }, { "epoch": 4.77, "learning_rate": 6.316930727419913e-05, "loss": 1.7102, "step": 55060 }, { "epoch": 4.77, "learning_rate": 6.31671388921013e-05, "loss": 1.8629, "step": 55070 }, { "epoch": 4.77, "learning_rate": 6.316497051000346e-05, "loss": 1.768, "step": 55080 }, { "epoch": 4.77, "learning_rate": 6.316280212790563e-05, "loss": 1.5262, "step": 55090 }, { "epoch": 4.77, "learning_rate": 6.316063374580778e-05, "loss": 1.4529, "step": 55100 }, { "epoch": 4.77, "learning_rate": 6.315846536370995e-05, "loss": 1.6268, "step": 55110 }, { "epoch": 4.77, "learning_rate": 6.315629698161212e-05, "loss": 1.5619, "step": 55120 }, { "epoch": 4.77, "learning_rate": 6.315412859951428e-05, "loss": 1.5951, "step": 55130 }, { "epoch": 4.78, "learning_rate": 6.315196021741644e-05, "loss": 1.5887, "step": 55140 }, { "epoch": 4.78, "learning_rate": 6.31497918353186e-05, "loss": 1.5696, "step": 55150 }, { "epoch": 4.78, "learning_rate": 6.314762345322077e-05, "loss": 1.6995, "step": 55160 }, { "epoch": 4.78, "learning_rate": 6.314545507112292e-05, "loss": 1.7184, "step": 55170 }, { "epoch": 4.78, "learning_rate": 6.314328668902509e-05, "loss": 1.5904, "step": 55180 }, { "epoch": 4.78, "learning_rate": 6.314111830692725e-05, "loss": 1.396, "step": 55190 }, { "epoch": 4.78, "learning_rate": 6.313894992482942e-05, "loss": 1.6879, "step": 55200 }, { "epoch": 4.78, "learning_rate": 6.313678154273157e-05, "loss": 1.6526, "step": 55210 }, { "epoch": 4.78, "learning_rate": 6.313461316063374e-05, "loss": 1.4068, "step": 55220 }, { "epoch": 4.78, "learning_rate": 6.31324447785359e-05, "loss": 1.558, "step": 55230 }, { "epoch": 4.78, "learning_rate": 6.313027639643806e-05, "loss": 1.5878, "step": 55240 }, { "epoch": 4.78, "learning_rate": 6.312810801434023e-05, "loss": 1.672, "step": 55250 }, { "epoch": 4.79, "learning_rate": 6.312593963224239e-05, "loss": 1.7249, "step": 55260 }, { "epoch": 4.79, "learning_rate": 6.312377125014456e-05, "loss": 1.6996, "step": 55270 }, { "epoch": 4.79, "learning_rate": 6.312160286804672e-05, "loss": 1.787, "step": 55280 }, { "epoch": 4.79, "learning_rate": 6.311943448594888e-05, "loss": 1.588, "step": 55290 }, { "epoch": 4.79, "learning_rate": 6.311726610385104e-05, "loss": 1.6037, "step": 55300 }, { "epoch": 4.79, "learning_rate": 6.31150977217532e-05, "loss": 1.7595, "step": 55310 }, { "epoch": 4.79, "learning_rate": 6.311292933965536e-05, "loss": 1.7931, "step": 55320 }, { "epoch": 4.79, "learning_rate": 6.311076095755753e-05, "loss": 1.6466, "step": 55330 }, { "epoch": 4.79, "learning_rate": 6.31085925754597e-05, "loss": 1.6593, "step": 55340 }, { "epoch": 4.79, "learning_rate": 6.310642419336186e-05, "loss": 1.4636, "step": 55350 }, { "epoch": 4.79, "learning_rate": 6.310425581126401e-05, "loss": 1.7376, "step": 55360 }, { "epoch": 4.8, "learning_rate": 6.310208742916618e-05, "loss": 1.7827, "step": 55370 }, { "epoch": 4.8, "learning_rate": 6.309991904706833e-05, "loss": 1.4315, "step": 55380 }, { "epoch": 4.8, "learning_rate": 6.30977506649705e-05, "loss": 1.7772, "step": 55390 }, { "epoch": 4.8, "learning_rate": 6.309558228287267e-05, "loss": 1.7859, "step": 55400 }, { "epoch": 4.8, "learning_rate": 6.309341390077483e-05, "loss": 1.5412, "step": 55410 }, { "epoch": 4.8, "learning_rate": 6.3091245518677e-05, "loss": 1.5577, "step": 55420 }, { "epoch": 4.8, "learning_rate": 6.308907713657915e-05, "loss": 1.6164, "step": 55430 }, { "epoch": 4.8, "learning_rate": 6.308690875448132e-05, "loss": 1.8119, "step": 55440 }, { "epoch": 4.8, "learning_rate": 6.308474037238347e-05, "loss": 1.7432, "step": 55450 }, { "epoch": 4.8, "learning_rate": 6.308257199028564e-05, "loss": 1.5681, "step": 55460 }, { "epoch": 4.8, "learning_rate": 6.30804036081878e-05, "loss": 1.6006, "step": 55470 }, { "epoch": 4.8, "learning_rate": 6.307823522608997e-05, "loss": 1.6826, "step": 55480 }, { "epoch": 4.81, "learning_rate": 6.307606684399214e-05, "loss": 1.9152, "step": 55490 }, { "epoch": 4.81, "learning_rate": 6.307389846189429e-05, "loss": 1.677, "step": 55500 }, { "epoch": 4.81, "learning_rate": 6.307173007979646e-05, "loss": 1.9499, "step": 55510 }, { "epoch": 4.81, "learning_rate": 6.306956169769861e-05, "loss": 1.7901, "step": 55520 }, { "epoch": 4.81, "learning_rate": 6.306739331560078e-05, "loss": 1.9502, "step": 55530 }, { "epoch": 4.81, "learning_rate": 6.306522493350294e-05, "loss": 1.6629, "step": 55540 }, { "epoch": 4.81, "learning_rate": 6.306305655140511e-05, "loss": 1.4488, "step": 55550 }, { "epoch": 4.81, "learning_rate": 6.306088816930727e-05, "loss": 1.4796, "step": 55560 }, { "epoch": 4.81, "learning_rate": 6.305871978720943e-05, "loss": 1.8705, "step": 55570 }, { "epoch": 4.81, "learning_rate": 6.30565514051116e-05, "loss": 1.6474, "step": 55580 }, { "epoch": 4.81, "learning_rate": 6.305438302301376e-05, "loss": 1.6422, "step": 55590 }, { "epoch": 4.82, "learning_rate": 6.305221464091591e-05, "loss": 1.6293, "step": 55600 }, { "epoch": 4.82, "learning_rate": 6.305004625881808e-05, "loss": 1.5833, "step": 55610 }, { "epoch": 4.82, "learning_rate": 6.304787787672025e-05, "loss": 1.787, "step": 55620 }, { "epoch": 4.82, "learning_rate": 6.304570949462241e-05, "loss": 1.7806, "step": 55630 }, { "epoch": 4.82, "learning_rate": 6.304354111252457e-05, "loss": 1.555, "step": 55640 }, { "epoch": 4.82, "learning_rate": 6.304137273042673e-05, "loss": 1.5116, "step": 55650 }, { "epoch": 4.82, "learning_rate": 6.30392043483289e-05, "loss": 1.6255, "step": 55660 }, { "epoch": 4.82, "learning_rate": 6.303703596623105e-05, "loss": 1.4365, "step": 55670 }, { "epoch": 4.82, "learning_rate": 6.303486758413322e-05, "loss": 1.6228, "step": 55680 }, { "epoch": 4.82, "learning_rate": 6.303269920203538e-05, "loss": 1.7176, "step": 55690 }, { "epoch": 4.82, "learning_rate": 6.303053081993755e-05, "loss": 1.7837, "step": 55700 }, { "epoch": 4.82, "learning_rate": 6.30283624378397e-05, "loss": 1.6313, "step": 55710 }, { "epoch": 4.83, "learning_rate": 6.302619405574187e-05, "loss": 1.791, "step": 55720 }, { "epoch": 4.83, "learning_rate": 6.302402567364404e-05, "loss": 1.7347, "step": 55730 }, { "epoch": 4.83, "learning_rate": 6.30218572915462e-05, "loss": 1.5682, "step": 55740 }, { "epoch": 4.83, "learning_rate": 6.301968890944836e-05, "loss": 1.7975, "step": 55750 }, { "epoch": 4.83, "learning_rate": 6.301752052735052e-05, "loss": 1.6314, "step": 55760 }, { "epoch": 4.83, "learning_rate": 6.301535214525269e-05, "loss": 1.658, "step": 55770 }, { "epoch": 4.83, "learning_rate": 6.301318376315484e-05, "loss": 1.6841, "step": 55780 }, { "epoch": 4.83, "learning_rate": 6.301101538105701e-05, "loss": 1.494, "step": 55790 }, { "epoch": 4.83, "learning_rate": 6.300884699895917e-05, "loss": 1.5884, "step": 55800 }, { "epoch": 4.83, "learning_rate": 6.300667861686134e-05, "loss": 1.6194, "step": 55810 }, { "epoch": 4.83, "learning_rate": 6.30045102347635e-05, "loss": 1.773, "step": 55820 }, { "epoch": 4.84, "learning_rate": 6.300234185266566e-05, "loss": 1.7452, "step": 55830 }, { "epoch": 4.84, "learning_rate": 6.300017347056783e-05, "loss": 1.784, "step": 55840 }, { "epoch": 4.84, "learning_rate": 6.299800508846998e-05, "loss": 1.5429, "step": 55850 }, { "epoch": 4.84, "learning_rate": 6.299583670637214e-05, "loss": 1.9204, "step": 55860 }, { "epoch": 4.84, "learning_rate": 6.299366832427431e-05, "loss": 1.8198, "step": 55870 }, { "epoch": 4.84, "learning_rate": 6.299149994217648e-05, "loss": 1.6497, "step": 55880 }, { "epoch": 4.84, "learning_rate": 6.298933156007864e-05, "loss": 1.6422, "step": 55890 }, { "epoch": 4.84, "learning_rate": 6.29871631779808e-05, "loss": 2.0181, "step": 55900 }, { "epoch": 4.84, "learning_rate": 6.298499479588296e-05, "loss": 1.5748, "step": 55910 }, { "epoch": 4.84, "learning_rate": 6.298282641378512e-05, "loss": 1.7379, "step": 55920 }, { "epoch": 4.84, "learning_rate": 6.298065803168728e-05, "loss": 1.6074, "step": 55930 }, { "epoch": 4.84, "learning_rate": 6.297848964958945e-05, "loss": 1.7445, "step": 55940 }, { "epoch": 4.85, "learning_rate": 6.297632126749162e-05, "loss": 1.8545, "step": 55950 }, { "epoch": 4.85, "learning_rate": 6.297415288539378e-05, "loss": 1.6657, "step": 55960 }, { "epoch": 4.85, "learning_rate": 6.297198450329593e-05, "loss": 1.4725, "step": 55970 }, { "epoch": 4.85, "learning_rate": 6.296981612119809e-05, "loss": 1.6816, "step": 55980 }, { "epoch": 4.85, "learning_rate": 6.296764773910025e-05, "loss": 1.6702, "step": 55990 }, { "epoch": 4.85, "learning_rate": 6.296547935700242e-05, "loss": 1.517, "step": 56000 }, { "epoch": 4.85, "learning_rate": 6.296331097490459e-05, "loss": 1.5962, "step": 56010 }, { "epoch": 4.85, "learning_rate": 6.296114259280675e-05, "loss": 1.8145, "step": 56020 }, { "epoch": 4.85, "learning_rate": 6.29589742107089e-05, "loss": 1.7458, "step": 56030 }, { "epoch": 4.85, "learning_rate": 6.295680582861107e-05, "loss": 1.6436, "step": 56040 }, { "epoch": 4.85, "learning_rate": 6.295463744651324e-05, "loss": 1.9619, "step": 56050 }, { "epoch": 4.86, "learning_rate": 6.295246906441539e-05, "loss": 1.6996, "step": 56060 }, { "epoch": 4.86, "learning_rate": 6.295030068231756e-05, "loss": 1.5376, "step": 56070 }, { "epoch": 4.86, "learning_rate": 6.294813230021972e-05, "loss": 1.9148, "step": 56080 }, { "epoch": 4.86, "learning_rate": 6.294596391812189e-05, "loss": 1.7291, "step": 56090 }, { "epoch": 4.86, "learning_rate": 6.294379553602404e-05, "loss": 1.8076, "step": 56100 }, { "epoch": 4.86, "learning_rate": 6.294162715392621e-05, "loss": 1.7607, "step": 56110 }, { "epoch": 4.86, "learning_rate": 6.293945877182838e-05, "loss": 1.6367, "step": 56120 }, { "epoch": 4.86, "learning_rate": 6.293729038973054e-05, "loss": 1.6673, "step": 56130 }, { "epoch": 4.86, "learning_rate": 6.29351220076327e-05, "loss": 1.5451, "step": 56140 }, { "epoch": 4.86, "learning_rate": 6.293295362553486e-05, "loss": 1.5232, "step": 56150 }, { "epoch": 4.86, "learning_rate": 6.293078524343703e-05, "loss": 1.6596, "step": 56160 }, { "epoch": 4.86, "learning_rate": 6.292861686133918e-05, "loss": 1.7496, "step": 56170 }, { "epoch": 4.87, "learning_rate": 6.292644847924135e-05, "loss": 1.7994, "step": 56180 }, { "epoch": 4.87, "learning_rate": 6.292428009714351e-05, "loss": 1.6998, "step": 56190 }, { "epoch": 4.87, "learning_rate": 6.292211171504568e-05, "loss": 1.6447, "step": 56200 }, { "epoch": 4.87, "learning_rate": 6.291994333294783e-05, "loss": 1.6643, "step": 56210 }, { "epoch": 4.87, "learning_rate": 6.291777495085e-05, "loss": 2.0019, "step": 56220 }, { "epoch": 4.87, "learning_rate": 6.291560656875217e-05, "loss": 1.541, "step": 56230 }, { "epoch": 4.87, "learning_rate": 6.291343818665432e-05, "loss": 1.8105, "step": 56240 }, { "epoch": 4.87, "learning_rate": 6.291126980455648e-05, "loss": 1.6573, "step": 56250 }, { "epoch": 4.87, "learning_rate": 6.290910142245865e-05, "loss": 1.696, "step": 56260 }, { "epoch": 4.87, "learning_rate": 6.290693304036082e-05, "loss": 1.6077, "step": 56270 }, { "epoch": 4.87, "learning_rate": 6.290476465826298e-05, "loss": 1.6085, "step": 56280 }, { "epoch": 4.88, "learning_rate": 6.290259627616514e-05, "loss": 1.6582, "step": 56290 }, { "epoch": 4.88, "learning_rate": 6.29004278940673e-05, "loss": 1.6078, "step": 56300 }, { "epoch": 4.88, "learning_rate": 6.289825951196946e-05, "loss": 1.7135, "step": 56310 }, { "epoch": 4.88, "learning_rate": 6.289609112987162e-05, "loss": 1.675, "step": 56320 }, { "epoch": 4.88, "learning_rate": 6.289392274777379e-05, "loss": 1.5798, "step": 56330 }, { "epoch": 4.88, "learning_rate": 6.289175436567596e-05, "loss": 1.6707, "step": 56340 }, { "epoch": 4.88, "learning_rate": 6.288958598357812e-05, "loss": 1.5607, "step": 56350 }, { "epoch": 4.88, "learning_rate": 6.288741760148027e-05, "loss": 1.5556, "step": 56360 }, { "epoch": 4.88, "learning_rate": 6.288524921938244e-05, "loss": 1.53, "step": 56370 }, { "epoch": 4.88, "learning_rate": 6.28830808372846e-05, "loss": 1.635, "step": 56380 }, { "epoch": 4.88, "learning_rate": 6.288091245518676e-05, "loss": 1.6868, "step": 56390 }, { "epoch": 4.88, "learning_rate": 6.287874407308893e-05, "loss": 1.6664, "step": 56400 }, { "epoch": 4.89, "learning_rate": 6.287657569099109e-05, "loss": 1.7466, "step": 56410 }, { "epoch": 4.89, "learning_rate": 6.287440730889326e-05, "loss": 1.6007, "step": 56420 }, { "epoch": 4.89, "learning_rate": 6.287223892679541e-05, "loss": 1.7785, "step": 56430 }, { "epoch": 4.89, "learning_rate": 6.287007054469758e-05, "loss": 1.7091, "step": 56440 }, { "epoch": 4.89, "learning_rate": 6.286790216259973e-05, "loss": 1.7444, "step": 56450 }, { "epoch": 4.89, "learning_rate": 6.28657337805019e-05, "loss": 1.7339, "step": 56460 }, { "epoch": 4.89, "learning_rate": 6.286356539840406e-05, "loss": 1.5698, "step": 56470 }, { "epoch": 4.89, "learning_rate": 6.286139701630623e-05, "loss": 1.7574, "step": 56480 }, { "epoch": 4.89, "learning_rate": 6.28592286342084e-05, "loss": 1.6158, "step": 56490 }, { "epoch": 4.89, "learning_rate": 6.285706025211055e-05, "loss": 1.3954, "step": 56500 }, { "epoch": 4.89, "learning_rate": 6.285489187001272e-05, "loss": 1.6413, "step": 56510 }, { "epoch": 4.89, "learning_rate": 6.285272348791487e-05, "loss": 1.7015, "step": 56520 }, { "epoch": 4.9, "learning_rate": 6.285055510581704e-05, "loss": 1.7474, "step": 56530 }, { "epoch": 4.9, "learning_rate": 6.28483867237192e-05, "loss": 1.7442, "step": 56540 }, { "epoch": 4.9, "learning_rate": 6.284621834162137e-05, "loss": 1.5861, "step": 56550 }, { "epoch": 4.9, "learning_rate": 6.284404995952353e-05, "loss": 1.5498, "step": 56560 }, { "epoch": 4.9, "learning_rate": 6.284188157742569e-05, "loss": 1.6988, "step": 56570 }, { "epoch": 4.9, "learning_rate": 6.283971319532785e-05, "loss": 1.7944, "step": 56580 }, { "epoch": 4.9, "learning_rate": 6.283754481323002e-05, "loss": 1.8167, "step": 56590 }, { "epoch": 4.9, "learning_rate": 6.283537643113217e-05, "loss": 1.7253, "step": 56600 }, { "epoch": 4.9, "learning_rate": 6.283320804903434e-05, "loss": 1.6939, "step": 56610 }, { "epoch": 4.9, "learning_rate": 6.28310396669365e-05, "loss": 1.7203, "step": 56620 }, { "epoch": 4.9, "learning_rate": 6.282887128483867e-05, "loss": 1.6956, "step": 56630 }, { "epoch": 4.91, "learning_rate": 6.282670290274083e-05, "loss": 1.3055, "step": 56640 }, { "epoch": 4.91, "learning_rate": 6.282453452064299e-05, "loss": 1.6555, "step": 56650 }, { "epoch": 4.91, "learning_rate": 6.282236613854516e-05, "loss": 1.6167, "step": 56660 }, { "epoch": 4.91, "learning_rate": 6.282019775644732e-05, "loss": 1.6999, "step": 56670 }, { "epoch": 4.91, "learning_rate": 6.281802937434948e-05, "loss": 1.6131, "step": 56680 }, { "epoch": 4.91, "learning_rate": 6.281586099225164e-05, "loss": 1.6383, "step": 56690 }, { "epoch": 4.91, "learning_rate": 6.281369261015381e-05, "loss": 1.6116, "step": 56700 }, { "epoch": 4.91, "learning_rate": 6.281152422805596e-05, "loss": 1.6409, "step": 56710 }, { "epoch": 4.91, "learning_rate": 6.280935584595813e-05, "loss": 1.7692, "step": 56720 }, { "epoch": 4.91, "learning_rate": 6.28071874638603e-05, "loss": 1.8022, "step": 56730 }, { "epoch": 4.91, "learning_rate": 6.280501908176246e-05, "loss": 1.6154, "step": 56740 }, { "epoch": 4.91, "learning_rate": 6.280285069966461e-05, "loss": 1.6148, "step": 56750 }, { "epoch": 4.92, "learning_rate": 6.280068231756678e-05, "loss": 1.7784, "step": 56760 }, { "epoch": 4.92, "learning_rate": 6.279851393546895e-05, "loss": 1.5011, "step": 56770 }, { "epoch": 4.92, "learning_rate": 6.27963455533711e-05, "loss": 1.7646, "step": 56780 }, { "epoch": 4.92, "learning_rate": 6.279417717127327e-05, "loss": 1.5098, "step": 56790 }, { "epoch": 4.92, "learning_rate": 6.279200878917543e-05, "loss": 1.5705, "step": 56800 }, { "epoch": 4.92, "learning_rate": 6.27898404070776e-05, "loss": 1.5932, "step": 56810 }, { "epoch": 4.92, "learning_rate": 6.278767202497977e-05, "loss": 1.7849, "step": 56820 }, { "epoch": 4.92, "learning_rate": 6.278550364288192e-05, "loss": 1.7753, "step": 56830 }, { "epoch": 4.92, "learning_rate": 6.278333526078409e-05, "loss": 1.7696, "step": 56840 }, { "epoch": 4.92, "learning_rate": 6.278116687868624e-05, "loss": 1.6156, "step": 56850 }, { "epoch": 4.92, "learning_rate": 6.27789984965884e-05, "loss": 1.4099, "step": 56860 }, { "epoch": 4.93, "learning_rate": 6.277683011449057e-05, "loss": 1.5872, "step": 56870 }, { "epoch": 4.93, "learning_rate": 6.277466173239274e-05, "loss": 1.6099, "step": 56880 }, { "epoch": 4.93, "learning_rate": 6.27724933502949e-05, "loss": 1.7881, "step": 56890 }, { "epoch": 4.93, "learning_rate": 6.277032496819706e-05, "loss": 1.7389, "step": 56900 }, { "epoch": 4.93, "learning_rate": 6.276815658609922e-05, "loss": 1.6771, "step": 56910 }, { "epoch": 4.93, "learning_rate": 6.276598820400138e-05, "loss": 1.8776, "step": 56920 }, { "epoch": 4.93, "learning_rate": 6.276381982190354e-05, "loss": 1.6605, "step": 56930 }, { "epoch": 4.93, "learning_rate": 6.276165143980571e-05, "loss": 1.7772, "step": 56940 }, { "epoch": 4.93, "learning_rate": 6.275948305770787e-05, "loss": 1.8023, "step": 56950 }, { "epoch": 4.93, "learning_rate": 6.275731467561004e-05, "loss": 1.6299, "step": 56960 }, { "epoch": 4.93, "learning_rate": 6.27551462935122e-05, "loss": 1.8579, "step": 56970 }, { "epoch": 4.93, "learning_rate": 6.275297791141436e-05, "loss": 1.7942, "step": 56980 }, { "epoch": 4.94, "learning_rate": 6.275080952931651e-05, "loss": 1.5272, "step": 56990 }, { "epoch": 4.94, "learning_rate": 6.274864114721868e-05, "loss": 1.7773, "step": 57000 }, { "epoch": 4.94, "learning_rate": 6.274647276512085e-05, "loss": 1.6271, "step": 57010 }, { "epoch": 4.94, "learning_rate": 6.274430438302301e-05, "loss": 1.5947, "step": 57020 }, { "epoch": 4.94, "learning_rate": 6.274213600092518e-05, "loss": 1.481, "step": 57030 }, { "epoch": 4.94, "learning_rate": 6.273996761882733e-05, "loss": 1.5866, "step": 57040 }, { "epoch": 4.94, "learning_rate": 6.27377992367295e-05, "loss": 1.6817, "step": 57050 }, { "epoch": 4.94, "learning_rate": 6.273563085463165e-05, "loss": 1.727, "step": 57060 }, { "epoch": 4.94, "learning_rate": 6.273346247253382e-05, "loss": 1.7072, "step": 57070 }, { "epoch": 4.94, "learning_rate": 6.273129409043598e-05, "loss": 1.8239, "step": 57080 }, { "epoch": 4.94, "learning_rate": 6.272912570833815e-05, "loss": 1.848, "step": 57090 }, { "epoch": 4.95, "learning_rate": 6.27269573262403e-05, "loss": 1.7237, "step": 57100 }, { "epoch": 4.95, "learning_rate": 6.272478894414247e-05, "loss": 1.778, "step": 57110 }, { "epoch": 4.95, "learning_rate": 6.272262056204464e-05, "loss": 1.6948, "step": 57120 }, { "epoch": 4.95, "learning_rate": 6.27204521799468e-05, "loss": 1.7234, "step": 57130 }, { "epoch": 4.95, "learning_rate": 6.271828379784895e-05, "loss": 1.6559, "step": 57140 }, { "epoch": 4.95, "learning_rate": 6.271611541575112e-05, "loss": 1.7801, "step": 57150 }, { "epoch": 4.95, "learning_rate": 6.271394703365329e-05, "loss": 1.6188, "step": 57160 }, { "epoch": 4.95, "learning_rate": 6.271177865155544e-05, "loss": 1.5868, "step": 57170 }, { "epoch": 4.95, "learning_rate": 6.270961026945761e-05, "loss": 1.851, "step": 57180 }, { "epoch": 4.95, "learning_rate": 6.270744188735977e-05, "loss": 1.8619, "step": 57190 }, { "epoch": 4.95, "learning_rate": 6.270527350526194e-05, "loss": 1.7362, "step": 57200 }, { "epoch": 4.95, "learning_rate": 6.270310512316409e-05, "loss": 1.6601, "step": 57210 }, { "epoch": 4.96, "learning_rate": 6.270093674106626e-05, "loss": 1.6058, "step": 57220 }, { "epoch": 4.96, "learning_rate": 6.269876835896843e-05, "loss": 1.6396, "step": 57230 }, { "epoch": 4.96, "learning_rate": 6.269659997687058e-05, "loss": 1.5227, "step": 57240 }, { "epoch": 4.96, "learning_rate": 6.269443159477274e-05, "loss": 1.5066, "step": 57250 }, { "epoch": 4.96, "learning_rate": 6.269226321267491e-05, "loss": 1.7384, "step": 57260 }, { "epoch": 4.96, "learning_rate": 6.269009483057708e-05, "loss": 1.7794, "step": 57270 }, { "epoch": 4.96, "learning_rate": 6.268792644847924e-05, "loss": 1.9453, "step": 57280 }, { "epoch": 4.96, "learning_rate": 6.26857580663814e-05, "loss": 1.79, "step": 57290 }, { "epoch": 4.96, "learning_rate": 6.268358968428356e-05, "loss": 1.7649, "step": 57300 }, { "epoch": 4.96, "learning_rate": 6.268142130218572e-05, "loss": 1.728, "step": 57310 }, { "epoch": 4.96, "learning_rate": 6.267925292008788e-05, "loss": 1.5576, "step": 57320 }, { "epoch": 4.97, "learning_rate": 6.267708453799005e-05, "loss": 1.7684, "step": 57330 }, { "epoch": 4.97, "learning_rate": 6.267491615589222e-05, "loss": 1.6897, "step": 57340 }, { "epoch": 4.97, "learning_rate": 6.267274777379438e-05, "loss": 1.7031, "step": 57350 }, { "epoch": 4.97, "learning_rate": 6.267057939169653e-05, "loss": 1.8032, "step": 57360 }, { "epoch": 4.97, "learning_rate": 6.26684110095987e-05, "loss": 1.7613, "step": 57370 }, { "epoch": 4.97, "learning_rate": 6.266624262750085e-05, "loss": 1.5785, "step": 57380 }, { "epoch": 4.97, "learning_rate": 6.266407424540302e-05, "loss": 1.619, "step": 57390 }, { "epoch": 4.97, "learning_rate": 6.266190586330519e-05, "loss": 1.6856, "step": 57400 }, { "epoch": 4.97, "learning_rate": 6.265973748120735e-05, "loss": 1.5089, "step": 57410 }, { "epoch": 4.97, "learning_rate": 6.265756909910952e-05, "loss": 1.6545, "step": 57420 }, { "epoch": 4.97, "learning_rate": 6.265540071701167e-05, "loss": 1.7504, "step": 57430 }, { "epoch": 4.97, "learning_rate": 6.265323233491384e-05, "loss": 1.6591, "step": 57440 }, { "epoch": 4.98, "learning_rate": 6.265106395281599e-05, "loss": 1.6514, "step": 57450 }, { "epoch": 4.98, "learning_rate": 6.264889557071816e-05, "loss": 1.7138, "step": 57460 }, { "epoch": 4.98, "learning_rate": 6.264672718862032e-05, "loss": 1.6267, "step": 57470 }, { "epoch": 4.98, "learning_rate": 6.264455880652249e-05, "loss": 1.8342, "step": 57480 }, { "epoch": 4.98, "learning_rate": 6.264239042442466e-05, "loss": 1.7028, "step": 57490 }, { "epoch": 4.98, "learning_rate": 6.264022204232681e-05, "loss": 1.7379, "step": 57500 }, { "epoch": 4.98, "learning_rate": 6.263805366022898e-05, "loss": 1.8167, "step": 57510 }, { "epoch": 4.98, "learning_rate": 6.263588527813113e-05, "loss": 1.9594, "step": 57520 }, { "epoch": 4.98, "learning_rate": 6.26337168960333e-05, "loss": 1.7578, "step": 57530 }, { "epoch": 4.98, "learning_rate": 6.263154851393546e-05, "loss": 1.5727, "step": 57540 }, { "epoch": 4.98, "learning_rate": 6.262938013183763e-05, "loss": 1.4844, "step": 57550 }, { "epoch": 4.99, "learning_rate": 6.26272117497398e-05, "loss": 1.3889, "step": 57560 }, { "epoch": 4.99, "learning_rate": 6.262504336764195e-05, "loss": 1.5254, "step": 57570 }, { "epoch": 4.99, "learning_rate": 6.262287498554411e-05, "loss": 1.6436, "step": 57580 }, { "epoch": 4.99, "learning_rate": 6.262070660344628e-05, "loss": 1.8892, "step": 57590 }, { "epoch": 4.99, "learning_rate": 6.261853822134843e-05, "loss": 1.5635, "step": 57600 }, { "epoch": 4.99, "learning_rate": 6.26163698392506e-05, "loss": 1.5509, "step": 57610 }, { "epoch": 4.99, "learning_rate": 6.261420145715277e-05, "loss": 1.4119, "step": 57620 }, { "epoch": 4.99, "learning_rate": 6.261203307505493e-05, "loss": 1.5606, "step": 57630 }, { "epoch": 4.99, "learning_rate": 6.260986469295708e-05, "loss": 1.7337, "step": 57640 }, { "epoch": 4.99, "learning_rate": 6.260769631085925e-05, "loss": 1.5899, "step": 57650 }, { "epoch": 4.99, "learning_rate": 6.260552792876142e-05, "loss": 1.8244, "step": 57660 }, { "epoch": 4.99, "learning_rate": 6.260335954666358e-05, "loss": 1.7327, "step": 57670 }, { "epoch": 5.0, "learning_rate": 6.260119116456574e-05, "loss": 1.5976, "step": 57680 }, { "epoch": 5.0, "learning_rate": 6.25990227824679e-05, "loss": 1.663, "step": 57690 }, { "epoch": 5.0, "learning_rate": 6.259685440037007e-05, "loss": 1.6761, "step": 57700 }, { "epoch": 5.0, "learning_rate": 6.259468601827222e-05, "loss": 1.5893, "step": 57710 }, { "epoch": 5.0, "learning_rate": 6.259251763617439e-05, "loss": 1.6856, "step": 57720 }, { "epoch": 5.0, "learning_rate": 6.259034925407656e-05, "loss": 1.6153, "step": 57730 }, { "epoch": 5.0, "eval_cer": 0.551654632433304, "eval_em": 0.010314595152140279, "eval_f1": 0.010314595152140279, "eval_loss": 1.608575701713562, "eval_runtime": 252.2533, "eval_samples_per_second": 7.687, "eval_steps_per_second": 0.963, "eval_wer": 0.9896854048478597, "step": 57732 }, { "epoch": 5.0, "learning_rate": 6.258818087197872e-05, "loss": 1.5396, "step": 57740 }, { "epoch": 5.0, "learning_rate": 6.258601248988087e-05, "loss": 1.6289, "step": 57750 }, { "epoch": 5.0, "learning_rate": 6.258384410778304e-05, "loss": 1.6536, "step": 57760 }, { "epoch": 5.0, "learning_rate": 6.258167572568521e-05, "loss": 1.5363, "step": 57770 }, { "epoch": 5.0, "learning_rate": 6.257950734358736e-05, "loss": 1.5311, "step": 57780 }, { "epoch": 5.0, "learning_rate": 6.257733896148953e-05, "loss": 1.5033, "step": 57790 }, { "epoch": 5.01, "learning_rate": 6.257517057939169e-05, "loss": 1.6687, "step": 57800 }, { "epoch": 5.01, "learning_rate": 6.257300219729386e-05, "loss": 1.5329, "step": 57810 }, { "epoch": 5.01, "learning_rate": 6.257083381519603e-05, "loss": 1.5687, "step": 57820 }, { "epoch": 5.01, "learning_rate": 6.256866543309818e-05, "loss": 1.5235, "step": 57830 }, { "epoch": 5.01, "learning_rate": 6.256649705100034e-05, "loss": 1.6682, "step": 57840 }, { "epoch": 5.01, "learning_rate": 6.25643286689025e-05, "loss": 1.6843, "step": 57850 }, { "epoch": 5.01, "learning_rate": 6.256216028680466e-05, "loss": 1.5415, "step": 57860 }, { "epoch": 5.01, "learning_rate": 6.255999190470683e-05, "loss": 1.6142, "step": 57870 }, { "epoch": 5.01, "learning_rate": 6.2557823522609e-05, "loss": 1.4796, "step": 57880 }, { "epoch": 5.01, "learning_rate": 6.255565514051116e-05, "loss": 1.7948, "step": 57890 }, { "epoch": 5.01, "learning_rate": 6.255348675841332e-05, "loss": 1.4177, "step": 57900 }, { "epoch": 5.02, "learning_rate": 6.255131837631548e-05, "loss": 1.4446, "step": 57910 }, { "epoch": 5.02, "learning_rate": 6.254914999421764e-05, "loss": 1.6614, "step": 57920 }, { "epoch": 5.02, "learning_rate": 6.25469816121198e-05, "loss": 1.5766, "step": 57930 }, { "epoch": 5.02, "learning_rate": 6.254481323002197e-05, "loss": 1.599, "step": 57940 }, { "epoch": 5.02, "learning_rate": 6.254264484792413e-05, "loss": 1.5883, "step": 57950 }, { "epoch": 5.02, "learning_rate": 6.25404764658263e-05, "loss": 1.5718, "step": 57960 }, { "epoch": 5.02, "learning_rate": 6.253830808372845e-05, "loss": 1.5295, "step": 57970 }, { "epoch": 5.02, "learning_rate": 6.253613970163062e-05, "loss": 1.357, "step": 57980 }, { "epoch": 5.02, "learning_rate": 6.253397131953277e-05, "loss": 1.4922, "step": 57990 }, { "epoch": 5.02, "learning_rate": 6.253180293743494e-05, "loss": 1.4787, "step": 58000 }, { "epoch": 5.02, "learning_rate": 6.25296345553371e-05, "loss": 1.457, "step": 58010 }, { "epoch": 5.02, "learning_rate": 6.252746617323927e-05, "loss": 1.5995, "step": 58020 }, { "epoch": 5.03, "learning_rate": 6.252529779114144e-05, "loss": 1.5127, "step": 58030 }, { "epoch": 5.03, "learning_rate": 6.252312940904359e-05, "loss": 1.3271, "step": 58040 }, { "epoch": 5.03, "learning_rate": 6.252096102694576e-05, "loss": 1.5915, "step": 58050 }, { "epoch": 5.03, "learning_rate": 6.251879264484791e-05, "loss": 1.5379, "step": 58060 }, { "epoch": 5.03, "learning_rate": 6.251662426275008e-05, "loss": 1.5163, "step": 58070 }, { "epoch": 5.03, "learning_rate": 6.251445588065224e-05, "loss": 1.637, "step": 58080 }, { "epoch": 5.03, "learning_rate": 6.251228749855441e-05, "loss": 1.5662, "step": 58090 }, { "epoch": 5.03, "learning_rate": 6.251011911645658e-05, "loss": 1.3679, "step": 58100 }, { "epoch": 5.03, "learning_rate": 6.250795073435873e-05, "loss": 1.6227, "step": 58110 }, { "epoch": 5.03, "learning_rate": 6.25057823522609e-05, "loss": 1.6492, "step": 58120 }, { "epoch": 5.03, "learning_rate": 6.250361397016306e-05, "loss": 1.6217, "step": 58130 }, { "epoch": 5.04, "learning_rate": 6.250144558806521e-05, "loss": 1.6259, "step": 58140 }, { "epoch": 5.04, "learning_rate": 6.249927720596738e-05, "loss": 1.5213, "step": 58150 }, { "epoch": 5.04, "learning_rate": 6.249710882386955e-05, "loss": 1.4496, "step": 58160 }, { "epoch": 5.04, "learning_rate": 6.24949404417717e-05, "loss": 1.6723, "step": 58170 }, { "epoch": 5.04, "learning_rate": 6.249277205967387e-05, "loss": 1.5344, "step": 58180 }, { "epoch": 5.04, "learning_rate": 6.249060367757603e-05, "loss": 1.7138, "step": 58190 }, { "epoch": 5.04, "learning_rate": 6.24884352954782e-05, "loss": 1.5048, "step": 58200 }, { "epoch": 5.04, "learning_rate": 6.248626691338037e-05, "loss": 1.5437, "step": 58210 }, { "epoch": 5.04, "learning_rate": 6.248409853128252e-05, "loss": 1.4531, "step": 58220 }, { "epoch": 5.04, "learning_rate": 6.248193014918469e-05, "loss": 1.6834, "step": 58230 }, { "epoch": 5.04, "learning_rate": 6.247976176708684e-05, "loss": 1.5964, "step": 58240 }, { "epoch": 5.04, "learning_rate": 6.2477593384989e-05, "loss": 1.5313, "step": 58250 }, { "epoch": 5.05, "learning_rate": 6.247542500289117e-05, "loss": 1.5692, "step": 58260 }, { "epoch": 5.05, "learning_rate": 6.247325662079334e-05, "loss": 1.4686, "step": 58270 }, { "epoch": 5.05, "learning_rate": 6.24710882386955e-05, "loss": 1.5331, "step": 58280 }, { "epoch": 5.05, "learning_rate": 6.246891985659766e-05, "loss": 1.5816, "step": 58290 }, { "epoch": 5.05, "learning_rate": 6.246675147449982e-05, "loss": 1.6012, "step": 58300 }, { "epoch": 5.05, "learning_rate": 6.246458309240198e-05, "loss": 1.5166, "step": 58310 }, { "epoch": 5.05, "learning_rate": 6.246241471030414e-05, "loss": 1.5881, "step": 58320 }, { "epoch": 5.05, "learning_rate": 6.246024632820631e-05, "loss": 1.618, "step": 58330 }, { "epoch": 5.05, "learning_rate": 6.245807794610847e-05, "loss": 1.6834, "step": 58340 }, { "epoch": 5.05, "learning_rate": 6.245590956401064e-05, "loss": 1.5747, "step": 58350 }, { "epoch": 5.05, "learning_rate": 6.24537411819128e-05, "loss": 1.5613, "step": 58360 }, { "epoch": 5.06, "learning_rate": 6.245157279981496e-05, "loss": 1.5832, "step": 58370 }, { "epoch": 5.06, "learning_rate": 6.244940441771711e-05, "loss": 1.4866, "step": 58380 }, { "epoch": 5.06, "learning_rate": 6.244723603561928e-05, "loss": 1.3388, "step": 58390 }, { "epoch": 5.06, "learning_rate": 6.244506765352145e-05, "loss": 1.5996, "step": 58400 }, { "epoch": 5.06, "learning_rate": 6.244289927142361e-05, "loss": 1.5466, "step": 58410 }, { "epoch": 5.06, "learning_rate": 6.244073088932578e-05, "loss": 1.5786, "step": 58420 }, { "epoch": 5.06, "learning_rate": 6.243856250722793e-05, "loss": 1.4082, "step": 58430 }, { "epoch": 5.06, "learning_rate": 6.24363941251301e-05, "loss": 1.417, "step": 58440 }, { "epoch": 5.06, "learning_rate": 6.243422574303225e-05, "loss": 1.4247, "step": 58450 }, { "epoch": 5.06, "learning_rate": 6.243205736093442e-05, "loss": 1.3129, "step": 58460 }, { "epoch": 5.06, "learning_rate": 6.242988897883658e-05, "loss": 1.6241, "step": 58470 }, { "epoch": 5.06, "learning_rate": 6.242772059673875e-05, "loss": 1.6946, "step": 58480 }, { "epoch": 5.07, "learning_rate": 6.242555221464092e-05, "loss": 1.5069, "step": 58490 }, { "epoch": 5.07, "learning_rate": 6.242338383254307e-05, "loss": 1.5911, "step": 58500 }, { "epoch": 5.07, "learning_rate": 6.242121545044524e-05, "loss": 1.4691, "step": 58510 }, { "epoch": 5.07, "learning_rate": 6.241904706834739e-05, "loss": 1.6017, "step": 58520 }, { "epoch": 5.07, "learning_rate": 6.241687868624955e-05, "loss": 1.7286, "step": 58530 }, { "epoch": 5.07, "learning_rate": 6.241471030415172e-05, "loss": 1.5367, "step": 58540 }, { "epoch": 5.07, "learning_rate": 6.241254192205389e-05, "loss": 1.5992, "step": 58550 }, { "epoch": 5.07, "learning_rate": 6.241037353995605e-05, "loss": 1.6208, "step": 58560 }, { "epoch": 5.07, "learning_rate": 6.240820515785821e-05, "loss": 1.7286, "step": 58570 }, { "epoch": 5.07, "learning_rate": 6.240603677576037e-05, "loss": 1.607, "step": 58580 }, { "epoch": 5.07, "learning_rate": 6.240386839366254e-05, "loss": 1.418, "step": 58590 }, { "epoch": 5.08, "learning_rate": 6.240170001156469e-05, "loss": 1.5003, "step": 58600 }, { "epoch": 5.08, "learning_rate": 6.239953162946686e-05, "loss": 1.5041, "step": 58610 }, { "epoch": 5.08, "learning_rate": 6.239736324736903e-05, "loss": 1.3353, "step": 58620 }, { "epoch": 5.08, "learning_rate": 6.239519486527119e-05, "loss": 1.6875, "step": 58630 }, { "epoch": 5.08, "learning_rate": 6.239302648317334e-05, "loss": 1.4668, "step": 58640 }, { "epoch": 5.08, "learning_rate": 6.239085810107551e-05, "loss": 1.4079, "step": 58650 }, { "epoch": 5.08, "learning_rate": 6.238868971897768e-05, "loss": 1.675, "step": 58660 }, { "epoch": 5.08, "learning_rate": 6.238652133687984e-05, "loss": 1.5224, "step": 58670 }, { "epoch": 5.08, "learning_rate": 6.2384352954782e-05, "loss": 1.6974, "step": 58680 }, { "epoch": 5.08, "learning_rate": 6.238218457268416e-05, "loss": 1.5226, "step": 58690 }, { "epoch": 5.08, "learning_rate": 6.238001619058633e-05, "loss": 1.7286, "step": 58700 }, { "epoch": 5.08, "learning_rate": 6.237784780848848e-05, "loss": 1.5524, "step": 58710 }, { "epoch": 5.09, "learning_rate": 6.237567942639065e-05, "loss": 1.6275, "step": 58720 }, { "epoch": 5.09, "learning_rate": 6.237351104429281e-05, "loss": 1.6832, "step": 58730 }, { "epoch": 5.09, "learning_rate": 6.237134266219498e-05, "loss": 1.5632, "step": 58740 }, { "epoch": 5.09, "learning_rate": 6.236917428009713e-05, "loss": 1.6214, "step": 58750 }, { "epoch": 5.09, "learning_rate": 6.23670058979993e-05, "loss": 1.5853, "step": 58760 }, { "epoch": 5.09, "learning_rate": 6.236483751590147e-05, "loss": 1.6096, "step": 58770 }, { "epoch": 5.09, "learning_rate": 6.236266913380362e-05, "loss": 1.6178, "step": 58780 }, { "epoch": 5.09, "learning_rate": 6.236050075170579e-05, "loss": 1.372, "step": 58790 }, { "epoch": 5.09, "learning_rate": 6.235833236960795e-05, "loss": 1.6758, "step": 58800 }, { "epoch": 5.09, "learning_rate": 6.235616398751012e-05, "loss": 1.6054, "step": 58810 }, { "epoch": 5.09, "learning_rate": 6.235399560541229e-05, "loss": 1.4015, "step": 58820 }, { "epoch": 5.1, "learning_rate": 6.235182722331444e-05, "loss": 1.5141, "step": 58830 }, { "epoch": 5.1, "learning_rate": 6.23496588412166e-05, "loss": 1.5992, "step": 58840 }, { "epoch": 5.1, "learning_rate": 6.234749045911876e-05, "loss": 1.6888, "step": 58850 }, { "epoch": 5.1, "learning_rate": 6.234532207702092e-05, "loss": 1.5774, "step": 58860 }, { "epoch": 5.1, "learning_rate": 6.234315369492309e-05, "loss": 1.7226, "step": 58870 }, { "epoch": 5.1, "learning_rate": 6.234098531282526e-05, "loss": 1.5927, "step": 58880 }, { "epoch": 5.1, "learning_rate": 6.233881693072742e-05, "loss": 1.6397, "step": 58890 }, { "epoch": 5.1, "learning_rate": 6.233664854862958e-05, "loss": 1.5276, "step": 58900 }, { "epoch": 5.1, "learning_rate": 6.233448016653174e-05, "loss": 1.5348, "step": 58910 }, { "epoch": 5.1, "learning_rate": 6.23323117844339e-05, "loss": 1.4765, "step": 58920 }, { "epoch": 5.1, "learning_rate": 6.233014340233606e-05, "loss": 1.4603, "step": 58930 }, { "epoch": 5.1, "learning_rate": 6.232797502023823e-05, "loss": 1.5663, "step": 58940 }, { "epoch": 5.11, "learning_rate": 6.23258066381404e-05, "loss": 1.6398, "step": 58950 }, { "epoch": 5.11, "learning_rate": 6.232363825604256e-05, "loss": 1.5331, "step": 58960 }, { "epoch": 5.11, "learning_rate": 6.232146987394471e-05, "loss": 1.4132, "step": 58970 }, { "epoch": 5.11, "learning_rate": 6.231930149184688e-05, "loss": 1.642, "step": 58980 }, { "epoch": 5.11, "learning_rate": 6.231713310974903e-05, "loss": 1.5132, "step": 58990 }, { "epoch": 5.11, "learning_rate": 6.23149647276512e-05, "loss": 1.4947, "step": 59000 }, { "epoch": 5.11, "learning_rate": 6.231279634555337e-05, "loss": 1.622, "step": 59010 }, { "epoch": 5.11, "learning_rate": 6.231062796345553e-05, "loss": 1.3184, "step": 59020 }, { "epoch": 5.11, "learning_rate": 6.23084595813577e-05, "loss": 1.6452, "step": 59030 }, { "epoch": 5.11, "learning_rate": 6.230629119925985e-05, "loss": 1.5897, "step": 59040 }, { "epoch": 5.11, "learning_rate": 6.230412281716202e-05, "loss": 1.6271, "step": 59050 }, { "epoch": 5.11, "learning_rate": 6.230195443506417e-05, "loss": 1.6475, "step": 59060 }, { "epoch": 5.12, "learning_rate": 6.229978605296634e-05, "loss": 1.6902, "step": 59070 }, { "epoch": 5.12, "learning_rate": 6.22976176708685e-05, "loss": 1.5044, "step": 59080 }, { "epoch": 5.12, "learning_rate": 6.229544928877067e-05, "loss": 1.7165, "step": 59090 }, { "epoch": 5.12, "learning_rate": 6.229328090667284e-05, "loss": 1.5503, "step": 59100 }, { "epoch": 5.12, "learning_rate": 6.229111252457499e-05, "loss": 1.5549, "step": 59110 }, { "epoch": 5.12, "learning_rate": 6.228894414247716e-05, "loss": 1.5455, "step": 59120 }, { "epoch": 5.12, "learning_rate": 6.228677576037932e-05, "loss": 1.6849, "step": 59130 }, { "epoch": 5.12, "learning_rate": 6.228460737828147e-05, "loss": 1.5431, "step": 59140 }, { "epoch": 5.12, "learning_rate": 6.228243899618364e-05, "loss": 1.5154, "step": 59150 }, { "epoch": 5.12, "learning_rate": 6.228027061408581e-05, "loss": 1.5419, "step": 59160 }, { "epoch": 5.12, "learning_rate": 6.227810223198797e-05, "loss": 1.5716, "step": 59170 }, { "epoch": 5.13, "learning_rate": 6.227593384989013e-05, "loss": 1.7896, "step": 59180 }, { "epoch": 5.13, "learning_rate": 6.227376546779229e-05, "loss": 1.3989, "step": 59190 }, { "epoch": 5.13, "learning_rate": 6.227159708569446e-05, "loss": 1.401, "step": 59200 }, { "epoch": 5.13, "learning_rate": 6.226942870359663e-05, "loss": 1.5229, "step": 59210 }, { "epoch": 5.13, "learning_rate": 6.226726032149878e-05, "loss": 1.6339, "step": 59220 }, { "epoch": 5.13, "learning_rate": 6.226509193940094e-05, "loss": 1.8003, "step": 59230 }, { "epoch": 5.13, "learning_rate": 6.22629235573031e-05, "loss": 1.5749, "step": 59240 }, { "epoch": 5.13, "learning_rate": 6.226075517520526e-05, "loss": 1.5112, "step": 59250 }, { "epoch": 5.13, "learning_rate": 6.225858679310743e-05, "loss": 1.4646, "step": 59260 }, { "epoch": 5.13, "learning_rate": 6.22564184110096e-05, "loss": 1.6656, "step": 59270 }, { "epoch": 5.13, "learning_rate": 6.225425002891176e-05, "loss": 1.6545, "step": 59280 }, { "epoch": 5.13, "learning_rate": 6.225208164681392e-05, "loss": 1.6089, "step": 59290 }, { "epoch": 5.14, "learning_rate": 6.224991326471608e-05, "loss": 1.7373, "step": 59300 }, { "epoch": 5.14, "learning_rate": 6.224774488261824e-05, "loss": 1.63, "step": 59310 }, { "epoch": 5.14, "learning_rate": 6.22455765005204e-05, "loss": 1.5241, "step": 59320 }, { "epoch": 5.14, "learning_rate": 6.224340811842257e-05, "loss": 1.5146, "step": 59330 }, { "epoch": 5.14, "learning_rate": 6.224123973632473e-05, "loss": 1.4253, "step": 59340 }, { "epoch": 5.14, "learning_rate": 6.22390713542269e-05, "loss": 1.4961, "step": 59350 }, { "epoch": 5.14, "learning_rate": 6.223690297212905e-05, "loss": 1.4588, "step": 59360 }, { "epoch": 5.14, "learning_rate": 6.223473459003122e-05, "loss": 1.7916, "step": 59370 }, { "epoch": 5.14, "learning_rate": 6.223256620793337e-05, "loss": 1.5993, "step": 59380 }, { "epoch": 5.14, "learning_rate": 6.223039782583554e-05, "loss": 1.6291, "step": 59390 }, { "epoch": 5.14, "learning_rate": 6.22282294437377e-05, "loss": 1.5026, "step": 59400 }, { "epoch": 5.15, "learning_rate": 6.222606106163987e-05, "loss": 1.5216, "step": 59410 }, { "epoch": 5.15, "learning_rate": 6.222389267954204e-05, "loss": 1.4395, "step": 59420 }, { "epoch": 5.15, "learning_rate": 6.222172429744419e-05, "loss": 1.4883, "step": 59430 }, { "epoch": 5.15, "learning_rate": 6.221955591534636e-05, "loss": 1.6566, "step": 59440 }, { "epoch": 5.15, "learning_rate": 6.221738753324851e-05, "loss": 1.5178, "step": 59450 }, { "epoch": 5.15, "learning_rate": 6.221521915115068e-05, "loss": 1.6542, "step": 59460 }, { "epoch": 5.15, "learning_rate": 6.221305076905284e-05, "loss": 1.5098, "step": 59470 }, { "epoch": 5.15, "learning_rate": 6.221088238695501e-05, "loss": 1.3514, "step": 59480 }, { "epoch": 5.15, "learning_rate": 6.220871400485718e-05, "loss": 1.5732, "step": 59490 }, { "epoch": 5.15, "learning_rate": 6.220654562275933e-05, "loss": 1.4463, "step": 59500 }, { "epoch": 5.15, "learning_rate": 6.22043772406615e-05, "loss": 1.579, "step": 59510 }, { "epoch": 5.15, "learning_rate": 6.220220885856365e-05, "loss": 1.5436, "step": 59520 }, { "epoch": 5.16, "learning_rate": 6.220004047646581e-05, "loss": 1.5525, "step": 59530 }, { "epoch": 5.16, "learning_rate": 6.219787209436798e-05, "loss": 1.4679, "step": 59540 }, { "epoch": 5.16, "learning_rate": 6.219570371227015e-05, "loss": 1.5483, "step": 59550 }, { "epoch": 5.16, "learning_rate": 6.219353533017231e-05, "loss": 1.6864, "step": 59560 }, { "epoch": 5.16, "learning_rate": 6.219136694807447e-05, "loss": 1.3466, "step": 59570 }, { "epoch": 5.16, "learning_rate": 6.218919856597663e-05, "loss": 1.556, "step": 59580 }, { "epoch": 5.16, "learning_rate": 6.21870301838788e-05, "loss": 1.6974, "step": 59590 }, { "epoch": 5.16, "learning_rate": 6.218486180178095e-05, "loss": 1.5882, "step": 59600 }, { "epoch": 5.16, "learning_rate": 6.218269341968312e-05, "loss": 1.5626, "step": 59610 }, { "epoch": 5.16, "learning_rate": 6.218052503758528e-05, "loss": 1.431, "step": 59620 }, { "epoch": 5.16, "learning_rate": 6.217835665548745e-05, "loss": 1.6418, "step": 59630 }, { "epoch": 5.17, "learning_rate": 6.21761882733896e-05, "loss": 1.5592, "step": 59640 }, { "epoch": 5.17, "learning_rate": 6.217401989129177e-05, "loss": 1.782, "step": 59650 }, { "epoch": 5.17, "learning_rate": 6.217185150919394e-05, "loss": 1.6661, "step": 59660 }, { "epoch": 5.17, "learning_rate": 6.21696831270961e-05, "loss": 1.4845, "step": 59670 }, { "epoch": 5.17, "learning_rate": 6.216751474499826e-05, "loss": 1.618, "step": 59680 }, { "epoch": 5.17, "learning_rate": 6.216534636290042e-05, "loss": 1.6291, "step": 59690 }, { "epoch": 5.17, "learning_rate": 6.216317798080259e-05, "loss": 1.5573, "step": 59700 }, { "epoch": 5.17, "learning_rate": 6.216100959870474e-05, "loss": 1.4754, "step": 59710 }, { "epoch": 5.17, "learning_rate": 6.215884121660691e-05, "loss": 1.5518, "step": 59720 }, { "epoch": 5.17, "learning_rate": 6.215667283450907e-05, "loss": 1.7229, "step": 59730 }, { "epoch": 5.17, "learning_rate": 6.215450445241124e-05, "loss": 1.4686, "step": 59740 }, { "epoch": 5.17, "learning_rate": 6.215233607031341e-05, "loss": 1.6279, "step": 59750 }, { "epoch": 5.18, "learning_rate": 6.215016768821556e-05, "loss": 1.7505, "step": 59760 }, { "epoch": 5.18, "learning_rate": 6.214799930611773e-05, "loss": 1.3657, "step": 59770 }, { "epoch": 5.18, "learning_rate": 6.214583092401988e-05, "loss": 1.6948, "step": 59780 }, { "epoch": 5.18, "learning_rate": 6.214366254192205e-05, "loss": 1.4345, "step": 59790 }, { "epoch": 5.18, "learning_rate": 6.214149415982421e-05, "loss": 1.6136, "step": 59800 }, { "epoch": 5.18, "learning_rate": 6.213932577772638e-05, "loss": 1.7544, "step": 59810 }, { "epoch": 5.18, "learning_rate": 6.213715739562855e-05, "loss": 1.5467, "step": 59820 }, { "epoch": 5.18, "learning_rate": 6.21349890135307e-05, "loss": 1.5271, "step": 59830 }, { "epoch": 5.18, "learning_rate": 6.213282063143286e-05, "loss": 1.4565, "step": 59840 }, { "epoch": 5.18, "learning_rate": 6.213065224933502e-05, "loss": 1.5567, "step": 59850 }, { "epoch": 5.18, "learning_rate": 6.212848386723718e-05, "loss": 1.6177, "step": 59860 }, { "epoch": 5.19, "learning_rate": 6.212631548513935e-05, "loss": 1.5685, "step": 59870 }, { "epoch": 5.19, "learning_rate": 6.212414710304152e-05, "loss": 1.5835, "step": 59880 }, { "epoch": 5.19, "learning_rate": 6.212197872094368e-05, "loss": 1.6747, "step": 59890 }, { "epoch": 5.19, "learning_rate": 6.211981033884584e-05, "loss": 1.654, "step": 59900 }, { "epoch": 5.19, "learning_rate": 6.2117641956748e-05, "loss": 1.4786, "step": 59910 }, { "epoch": 5.19, "learning_rate": 6.211547357465015e-05, "loss": 1.6235, "step": 59920 }, { "epoch": 5.19, "learning_rate": 6.211330519255232e-05, "loss": 1.4691, "step": 59930 }, { "epoch": 5.19, "learning_rate": 6.211113681045449e-05, "loss": 1.528, "step": 59940 }, { "epoch": 5.19, "learning_rate": 6.210896842835665e-05, "loss": 1.7522, "step": 59950 }, { "epoch": 5.19, "learning_rate": 6.210680004625882e-05, "loss": 1.5966, "step": 59960 }, { "epoch": 5.19, "learning_rate": 6.210463166416097e-05, "loss": 1.6183, "step": 59970 }, { "epoch": 5.19, "learning_rate": 6.210246328206314e-05, "loss": 1.3765, "step": 59980 }, { "epoch": 5.2, "learning_rate": 6.210029489996529e-05, "loss": 1.4393, "step": 59990 }, { "epoch": 5.2, "learning_rate": 6.209812651786746e-05, "loss": 1.5373, "step": 60000 }, { "epoch": 5.2, "learning_rate": 6.209595813576963e-05, "loss": 1.5741, "step": 60010 }, { "epoch": 5.2, "learning_rate": 6.209378975367179e-05, "loss": 1.747, "step": 60020 }, { "epoch": 5.2, "learning_rate": 6.209162137157396e-05, "loss": 1.8665, "step": 60030 }, { "epoch": 5.2, "learning_rate": 6.208945298947611e-05, "loss": 1.6557, "step": 60040 }, { "epoch": 5.2, "learning_rate": 6.208728460737828e-05, "loss": 1.4013, "step": 60050 }, { "epoch": 5.2, "learning_rate": 6.208511622528043e-05, "loss": 1.5452, "step": 60060 }, { "epoch": 5.2, "learning_rate": 6.20829478431826e-05, "loss": 1.6643, "step": 60070 }, { "epoch": 5.2, "learning_rate": 6.208077946108476e-05, "loss": 1.4874, "step": 60080 }, { "epoch": 5.2, "learning_rate": 6.207861107898693e-05, "loss": 1.4871, "step": 60090 }, { "epoch": 5.21, "learning_rate": 6.20764426968891e-05, "loss": 1.4649, "step": 60100 }, { "epoch": 5.21, "learning_rate": 6.207427431479125e-05, "loss": 1.5529, "step": 60110 }, { "epoch": 5.21, "learning_rate": 6.207210593269341e-05, "loss": 1.4921, "step": 60120 }, { "epoch": 5.21, "learning_rate": 6.206993755059558e-05, "loss": 1.3598, "step": 60130 }, { "epoch": 5.21, "learning_rate": 6.206776916849773e-05, "loss": 1.4738, "step": 60140 }, { "epoch": 5.21, "learning_rate": 6.20656007863999e-05, "loss": 1.5813, "step": 60150 }, { "epoch": 5.21, "learning_rate": 6.206343240430207e-05, "loss": 1.7451, "step": 60160 }, { "epoch": 5.21, "learning_rate": 6.206126402220423e-05, "loss": 1.5587, "step": 60170 }, { "epoch": 5.21, "learning_rate": 6.205909564010639e-05, "loss": 1.4661, "step": 60180 }, { "epoch": 5.21, "learning_rate": 6.205692725800855e-05, "loss": 1.5964, "step": 60190 }, { "epoch": 5.21, "learning_rate": 6.205475887591072e-05, "loss": 1.6731, "step": 60200 }, { "epoch": 5.21, "learning_rate": 6.205259049381289e-05, "loss": 1.5105, "step": 60210 }, { "epoch": 5.22, "learning_rate": 6.205042211171504e-05, "loss": 1.6736, "step": 60220 }, { "epoch": 5.22, "learning_rate": 6.20482537296172e-05, "loss": 1.5048, "step": 60230 }, { "epoch": 5.22, "learning_rate": 6.204608534751937e-05, "loss": 1.635, "step": 60240 }, { "epoch": 5.22, "learning_rate": 6.204391696542152e-05, "loss": 1.5507, "step": 60250 }, { "epoch": 5.22, "learning_rate": 6.204174858332369e-05, "loss": 1.4474, "step": 60260 }, { "epoch": 5.22, "learning_rate": 6.203958020122586e-05, "loss": 1.5737, "step": 60270 }, { "epoch": 5.22, "learning_rate": 6.203741181912802e-05, "loss": 1.6654, "step": 60280 }, { "epoch": 5.22, "learning_rate": 6.203524343703018e-05, "loss": 1.8866, "step": 60290 }, { "epoch": 5.22, "learning_rate": 6.203307505493234e-05, "loss": 1.4552, "step": 60300 }, { "epoch": 5.22, "learning_rate": 6.20309066728345e-05, "loss": 1.5307, "step": 60310 }, { "epoch": 5.22, "learning_rate": 6.202873829073666e-05, "loss": 1.8621, "step": 60320 }, { "epoch": 5.22, "learning_rate": 6.202656990863883e-05, "loss": 1.6742, "step": 60330 }, { "epoch": 5.23, "learning_rate": 6.2024401526541e-05, "loss": 1.5824, "step": 60340 }, { "epoch": 5.23, "learning_rate": 6.202223314444316e-05, "loss": 1.394, "step": 60350 }, { "epoch": 5.23, "learning_rate": 6.202006476234531e-05, "loss": 1.4223, "step": 60360 }, { "epoch": 5.23, "learning_rate": 6.201789638024748e-05, "loss": 1.5073, "step": 60370 }, { "epoch": 5.23, "learning_rate": 6.201572799814963e-05, "loss": 1.5164, "step": 60380 }, { "epoch": 5.23, "learning_rate": 6.20135596160518e-05, "loss": 1.2624, "step": 60390 }, { "epoch": 5.23, "learning_rate": 6.201139123395397e-05, "loss": 1.4957, "step": 60400 }, { "epoch": 5.23, "learning_rate": 6.200922285185613e-05, "loss": 1.7122, "step": 60410 }, { "epoch": 5.23, "learning_rate": 6.20070544697583e-05, "loss": 1.5217, "step": 60420 }, { "epoch": 5.23, "learning_rate": 6.200488608766045e-05, "loss": 1.6088, "step": 60430 }, { "epoch": 5.23, "learning_rate": 6.200271770556262e-05, "loss": 1.6304, "step": 60440 }, { "epoch": 5.24, "learning_rate": 6.200054932346477e-05, "loss": 1.4873, "step": 60450 }, { "epoch": 5.24, "learning_rate": 6.199838094136694e-05, "loss": 1.6739, "step": 60460 }, { "epoch": 5.24, "learning_rate": 6.19962125592691e-05, "loss": 1.4386, "step": 60470 }, { "epoch": 5.24, "learning_rate": 6.199404417717127e-05, "loss": 1.5419, "step": 60480 }, { "epoch": 5.24, "learning_rate": 6.199187579507344e-05, "loss": 1.4872, "step": 60490 }, { "epoch": 5.24, "learning_rate": 6.198970741297559e-05, "loss": 1.7471, "step": 60500 }, { "epoch": 5.24, "learning_rate": 6.198753903087776e-05, "loss": 1.5092, "step": 60510 }, { "epoch": 5.24, "learning_rate": 6.198537064877992e-05, "loss": 1.5234, "step": 60520 }, { "epoch": 5.24, "learning_rate": 6.198320226668207e-05, "loss": 1.5672, "step": 60530 }, { "epoch": 5.24, "learning_rate": 6.198103388458424e-05, "loss": 1.6536, "step": 60540 }, { "epoch": 5.24, "learning_rate": 6.197886550248641e-05, "loss": 1.5284, "step": 60550 }, { "epoch": 5.24, "learning_rate": 6.197669712038857e-05, "loss": 1.5572, "step": 60560 }, { "epoch": 5.25, "learning_rate": 6.197452873829073e-05, "loss": 1.4621, "step": 60570 }, { "epoch": 5.25, "learning_rate": 6.197236035619289e-05, "loss": 1.6093, "step": 60580 }, { "epoch": 5.25, "learning_rate": 6.197019197409506e-05, "loss": 1.4835, "step": 60590 }, { "epoch": 5.25, "learning_rate": 6.196802359199721e-05, "loss": 1.5208, "step": 60600 }, { "epoch": 5.25, "learning_rate": 6.196585520989938e-05, "loss": 1.6131, "step": 60610 }, { "epoch": 5.25, "learning_rate": 6.196368682780154e-05, "loss": 1.4765, "step": 60620 }, { "epoch": 5.25, "learning_rate": 6.196151844570371e-05, "loss": 1.8413, "step": 60630 }, { "epoch": 5.25, "learning_rate": 6.195935006360586e-05, "loss": 1.7436, "step": 60640 }, { "epoch": 5.25, "learning_rate": 6.195718168150803e-05, "loss": 1.7267, "step": 60650 }, { "epoch": 5.25, "learning_rate": 6.19550132994102e-05, "loss": 1.6564, "step": 60660 }, { "epoch": 5.25, "learning_rate": 6.195284491731236e-05, "loss": 1.5953, "step": 60670 }, { "epoch": 5.26, "learning_rate": 6.195067653521452e-05, "loss": 1.5381, "step": 60680 }, { "epoch": 5.26, "learning_rate": 6.194850815311668e-05, "loss": 1.5205, "step": 60690 }, { "epoch": 5.26, "learning_rate": 6.194633977101885e-05, "loss": 1.4618, "step": 60700 }, { "epoch": 5.26, "learning_rate": 6.1944171388921e-05, "loss": 1.6117, "step": 60710 }, { "epoch": 5.26, "learning_rate": 6.194200300682317e-05, "loss": 1.5588, "step": 60720 }, { "epoch": 5.26, "learning_rate": 6.193983462472533e-05, "loss": 1.4302, "step": 60730 }, { "epoch": 5.26, "learning_rate": 6.19376662426275e-05, "loss": 1.6622, "step": 60740 }, { "epoch": 5.26, "learning_rate": 6.193549786052967e-05, "loss": 1.5812, "step": 60750 }, { "epoch": 5.26, "learning_rate": 6.193332947843182e-05, "loss": 1.5416, "step": 60760 }, { "epoch": 5.26, "learning_rate": 6.193116109633399e-05, "loss": 1.6133, "step": 60770 }, { "epoch": 5.26, "learning_rate": 6.192899271423614e-05, "loss": 1.702, "step": 60780 }, { "epoch": 5.26, "learning_rate": 6.19268243321383e-05, "loss": 1.4493, "step": 60790 }, { "epoch": 5.27, "learning_rate": 6.192465595004047e-05, "loss": 1.3107, "step": 60800 }, { "epoch": 5.27, "learning_rate": 6.192248756794264e-05, "loss": 1.6102, "step": 60810 }, { "epoch": 5.27, "learning_rate": 6.19203191858448e-05, "loss": 1.5938, "step": 60820 }, { "epoch": 5.27, "learning_rate": 6.191815080374696e-05, "loss": 1.6293, "step": 60830 }, { "epoch": 5.27, "learning_rate": 6.191598242164912e-05, "loss": 1.6328, "step": 60840 }, { "epoch": 5.27, "learning_rate": 6.191381403955128e-05, "loss": 1.5887, "step": 60850 }, { "epoch": 5.27, "learning_rate": 6.191164565745344e-05, "loss": 1.4429, "step": 60860 }, { "epoch": 5.27, "learning_rate": 6.190947727535561e-05, "loss": 1.5129, "step": 60870 }, { "epoch": 5.27, "learning_rate": 6.190730889325778e-05, "loss": 1.4923, "step": 60880 }, { "epoch": 5.27, "learning_rate": 6.190514051115994e-05, "loss": 1.6825, "step": 60890 }, { "epoch": 5.27, "learning_rate": 6.19029721290621e-05, "loss": 1.6585, "step": 60900 }, { "epoch": 5.28, "learning_rate": 6.190080374696426e-05, "loss": 1.4966, "step": 60910 }, { "epoch": 5.28, "learning_rate": 6.189863536486641e-05, "loss": 1.5561, "step": 60920 }, { "epoch": 5.28, "learning_rate": 6.189646698276858e-05, "loss": 1.5327, "step": 60930 }, { "epoch": 5.28, "learning_rate": 6.189429860067075e-05, "loss": 1.5596, "step": 60940 }, { "epoch": 5.28, "learning_rate": 6.189213021857291e-05, "loss": 1.7425, "step": 60950 }, { "epoch": 5.28, "learning_rate": 6.188996183647508e-05, "loss": 1.5696, "step": 60960 }, { "epoch": 5.28, "learning_rate": 6.188779345437723e-05, "loss": 1.5245, "step": 60970 }, { "epoch": 5.28, "learning_rate": 6.18856250722794e-05, "loss": 1.3009, "step": 60980 }, { "epoch": 5.28, "learning_rate": 6.188345669018155e-05, "loss": 1.6035, "step": 60990 }, { "epoch": 5.28, "learning_rate": 6.188128830808372e-05, "loss": 1.4445, "step": 61000 }, { "epoch": 5.28, "learning_rate": 6.187911992598588e-05, "loss": 1.4812, "step": 61010 }, { "epoch": 5.28, "learning_rate": 6.187695154388805e-05, "loss": 1.5343, "step": 61020 }, { "epoch": 5.29, "learning_rate": 6.187478316179022e-05, "loss": 1.5499, "step": 61030 }, { "epoch": 5.29, "learning_rate": 6.187261477969237e-05, "loss": 1.7133, "step": 61040 }, { "epoch": 5.29, "learning_rate": 6.187044639759454e-05, "loss": 1.5272, "step": 61050 }, { "epoch": 5.29, "learning_rate": 6.186827801549669e-05, "loss": 1.6203, "step": 61060 }, { "epoch": 5.29, "learning_rate": 6.186610963339886e-05, "loss": 1.6192, "step": 61070 }, { "epoch": 5.29, "learning_rate": 6.186394125130102e-05, "loss": 1.5631, "step": 61080 }, { "epoch": 5.29, "learning_rate": 6.186177286920319e-05, "loss": 1.4758, "step": 61090 }, { "epoch": 5.29, "learning_rate": 6.185960448710536e-05, "loss": 1.5336, "step": 61100 }, { "epoch": 5.29, "learning_rate": 6.185743610500751e-05, "loss": 1.6986, "step": 61110 }, { "epoch": 5.29, "learning_rate": 6.185526772290967e-05, "loss": 1.4525, "step": 61120 }, { "epoch": 5.29, "learning_rate": 6.185309934081184e-05, "loss": 1.4605, "step": 61130 }, { "epoch": 5.3, "learning_rate": 6.1850930958714e-05, "loss": 1.5123, "step": 61140 }, { "epoch": 5.3, "learning_rate": 6.184876257661616e-05, "loss": 1.6258, "step": 61150 }, { "epoch": 5.3, "learning_rate": 6.184659419451833e-05, "loss": 1.6338, "step": 61160 }, { "epoch": 5.3, "learning_rate": 6.184442581242049e-05, "loss": 1.535, "step": 61170 }, { "epoch": 5.3, "learning_rate": 6.184225743032265e-05, "loss": 1.7816, "step": 61180 }, { "epoch": 5.3, "learning_rate": 6.184008904822481e-05, "loss": 1.6198, "step": 61190 }, { "epoch": 5.3, "learning_rate": 6.183792066612698e-05, "loss": 1.7263, "step": 61200 }, { "epoch": 5.3, "learning_rate": 6.183575228402914e-05, "loss": 1.6452, "step": 61210 }, { "epoch": 5.3, "learning_rate": 6.18335839019313e-05, "loss": 1.6612, "step": 61220 }, { "epoch": 5.3, "learning_rate": 6.183141551983346e-05, "loss": 1.5765, "step": 61230 }, { "epoch": 5.3, "learning_rate": 6.182924713773563e-05, "loss": 1.7606, "step": 61240 }, { "epoch": 5.3, "learning_rate": 6.182707875563778e-05, "loss": 1.5215, "step": 61250 }, { "epoch": 5.31, "learning_rate": 6.182491037353995e-05, "loss": 1.5277, "step": 61260 }, { "epoch": 5.31, "learning_rate": 6.182274199144212e-05, "loss": 1.6913, "step": 61270 }, { "epoch": 5.31, "learning_rate": 6.182057360934428e-05, "loss": 1.7646, "step": 61280 }, { "epoch": 5.31, "learning_rate": 6.181840522724645e-05, "loss": 1.5816, "step": 61290 }, { "epoch": 5.31, "learning_rate": 6.18162368451486e-05, "loss": 1.7119, "step": 61300 }, { "epoch": 5.31, "learning_rate": 6.181406846305077e-05, "loss": 1.6954, "step": 61310 }, { "epoch": 5.31, "learning_rate": 6.181190008095292e-05, "loss": 1.3578, "step": 61320 }, { "epoch": 5.31, "learning_rate": 6.180973169885509e-05, "loss": 1.8273, "step": 61330 }, { "epoch": 5.31, "learning_rate": 6.180756331675725e-05, "loss": 1.4251, "step": 61340 }, { "epoch": 5.31, "learning_rate": 6.180539493465942e-05, "loss": 1.5541, "step": 61350 }, { "epoch": 5.31, "learning_rate": 6.180322655256159e-05, "loss": 1.3958, "step": 61360 }, { "epoch": 5.32, "learning_rate": 6.180105817046374e-05, "loss": 1.7555, "step": 61370 }, { "epoch": 5.32, "learning_rate": 6.179888978836589e-05, "loss": 1.4715, "step": 61380 }, { "epoch": 5.32, "learning_rate": 6.179672140626806e-05, "loss": 1.6013, "step": 61390 }, { "epoch": 5.32, "learning_rate": 6.179455302417023e-05, "loss": 1.6579, "step": 61400 }, { "epoch": 5.32, "learning_rate": 6.179238464207239e-05, "loss": 1.715, "step": 61410 }, { "epoch": 5.32, "learning_rate": 6.179021625997456e-05, "loss": 1.813, "step": 61420 }, { "epoch": 5.32, "learning_rate": 6.178804787787671e-05, "loss": 1.3806, "step": 61430 }, { "epoch": 5.32, "learning_rate": 6.178587949577888e-05, "loss": 1.5111, "step": 61440 }, { "epoch": 5.32, "learning_rate": 6.178371111368103e-05, "loss": 1.5273, "step": 61450 }, { "epoch": 5.32, "learning_rate": 6.17815427315832e-05, "loss": 1.67, "step": 61460 }, { "epoch": 5.32, "learning_rate": 6.177937434948536e-05, "loss": 1.6739, "step": 61470 }, { "epoch": 5.32, "learning_rate": 6.177720596738753e-05, "loss": 1.3925, "step": 61480 }, { "epoch": 5.33, "learning_rate": 6.17750375852897e-05, "loss": 1.6331, "step": 61490 }, { "epoch": 5.33, "learning_rate": 6.177286920319185e-05, "loss": 1.4483, "step": 61500 }, { "epoch": 5.33, "learning_rate": 6.177070082109401e-05, "loss": 1.6736, "step": 61510 }, { "epoch": 5.33, "learning_rate": 6.176853243899618e-05, "loss": 1.6767, "step": 61520 }, { "epoch": 5.33, "learning_rate": 6.176636405689833e-05, "loss": 1.6578, "step": 61530 }, { "epoch": 5.33, "learning_rate": 6.17641956748005e-05, "loss": 1.503, "step": 61540 }, { "epoch": 5.33, "learning_rate": 6.176202729270267e-05, "loss": 1.6524, "step": 61550 }, { "epoch": 5.33, "learning_rate": 6.175985891060483e-05, "loss": 1.7044, "step": 61560 }, { "epoch": 5.33, "learning_rate": 6.175769052850699e-05, "loss": 1.4346, "step": 61570 }, { "epoch": 5.33, "learning_rate": 6.175552214640915e-05, "loss": 1.785, "step": 61580 }, { "epoch": 5.33, "learning_rate": 6.175335376431132e-05, "loss": 1.4947, "step": 61590 }, { "epoch": 5.33, "learning_rate": 6.175118538221347e-05, "loss": 1.6713, "step": 61600 }, { "epoch": 5.34, "learning_rate": 6.174901700011564e-05, "loss": 1.5397, "step": 61610 }, { "epoch": 5.34, "learning_rate": 6.17468486180178e-05, "loss": 1.5904, "step": 61620 }, { "epoch": 5.34, "learning_rate": 6.174468023591997e-05, "loss": 1.5842, "step": 61630 }, { "epoch": 5.34, "learning_rate": 6.174251185382212e-05, "loss": 1.4326, "step": 61640 }, { "epoch": 5.34, "learning_rate": 6.174034347172429e-05, "loss": 1.6077, "step": 61650 }, { "epoch": 5.34, "learning_rate": 6.173817508962646e-05, "loss": 1.5849, "step": 61660 }, { "epoch": 5.34, "learning_rate": 6.173600670752862e-05, "loss": 1.6275, "step": 61670 }, { "epoch": 5.34, "learning_rate": 6.173383832543078e-05, "loss": 1.78, "step": 61680 }, { "epoch": 5.34, "learning_rate": 6.173166994333294e-05, "loss": 1.4667, "step": 61690 }, { "epoch": 5.34, "learning_rate": 6.172950156123511e-05, "loss": 1.5006, "step": 61700 }, { "epoch": 5.34, "learning_rate": 6.172733317913726e-05, "loss": 1.7618, "step": 61710 }, { "epoch": 5.35, "learning_rate": 6.172516479703943e-05, "loss": 1.9035, "step": 61720 }, { "epoch": 5.35, "learning_rate": 6.17229964149416e-05, "loss": 1.6386, "step": 61730 }, { "epoch": 5.35, "learning_rate": 6.172082803284376e-05, "loss": 1.6132, "step": 61740 }, { "epoch": 5.35, "learning_rate": 6.171865965074593e-05, "loss": 1.414, "step": 61750 }, { "epoch": 5.35, "learning_rate": 6.171649126864808e-05, "loss": 1.7136, "step": 61760 }, { "epoch": 5.35, "learning_rate": 6.171432288655025e-05, "loss": 1.66, "step": 61770 }, { "epoch": 5.35, "learning_rate": 6.17121545044524e-05, "loss": 1.7221, "step": 61780 }, { "epoch": 5.35, "learning_rate": 6.170998612235457e-05, "loss": 1.6424, "step": 61790 }, { "epoch": 5.35, "learning_rate": 6.170781774025673e-05, "loss": 1.4547, "step": 61800 }, { "epoch": 5.35, "learning_rate": 6.17056493581589e-05, "loss": 1.8844, "step": 61810 }, { "epoch": 5.35, "learning_rate": 6.170348097606106e-05, "loss": 1.4815, "step": 61820 }, { "epoch": 5.35, "learning_rate": 6.170131259396322e-05, "loss": 1.527, "step": 61830 }, { "epoch": 5.36, "learning_rate": 6.169914421186538e-05, "loss": 1.4627, "step": 61840 }, { "epoch": 5.36, "learning_rate": 6.169697582976754e-05, "loss": 1.7067, "step": 61850 }, { "epoch": 5.36, "learning_rate": 6.16948074476697e-05, "loss": 1.5808, "step": 61860 }, { "epoch": 5.36, "learning_rate": 6.169263906557187e-05, "loss": 1.5567, "step": 61870 }, { "epoch": 5.36, "learning_rate": 6.169047068347404e-05, "loss": 1.563, "step": 61880 }, { "epoch": 5.36, "learning_rate": 6.16883023013762e-05, "loss": 1.5918, "step": 61890 }, { "epoch": 5.36, "learning_rate": 6.168613391927835e-05, "loss": 1.2978, "step": 61900 }, { "epoch": 5.36, "learning_rate": 6.168396553718052e-05, "loss": 1.6396, "step": 61910 }, { "epoch": 5.36, "learning_rate": 6.168179715508267e-05, "loss": 1.5392, "step": 61920 }, { "epoch": 5.36, "learning_rate": 6.167962877298484e-05, "loss": 1.3434, "step": 61930 }, { "epoch": 5.36, "learning_rate": 6.167746039088701e-05, "loss": 1.2365, "step": 61940 }, { "epoch": 5.37, "learning_rate": 6.167529200878917e-05, "loss": 1.4721, "step": 61950 }, { "epoch": 5.37, "learning_rate": 6.167312362669134e-05, "loss": 1.7701, "step": 61960 }, { "epoch": 5.37, "learning_rate": 6.167095524459349e-05, "loss": 1.4625, "step": 61970 }, { "epoch": 5.37, "learning_rate": 6.166878686249566e-05, "loss": 1.6962, "step": 61980 }, { "epoch": 5.37, "learning_rate": 6.166661848039781e-05, "loss": 1.4891, "step": 61990 }, { "epoch": 5.37, "learning_rate": 6.166445009829998e-05, "loss": 1.6279, "step": 62000 }, { "epoch": 5.37, "learning_rate": 6.166228171620214e-05, "loss": 1.4957, "step": 62010 }, { "epoch": 5.37, "learning_rate": 6.166011333410431e-05, "loss": 1.4802, "step": 62020 }, { "epoch": 5.37, "learning_rate": 6.165794495200648e-05, "loss": 1.7952, "step": 62030 }, { "epoch": 5.37, "learning_rate": 6.165577656990863e-05, "loss": 1.7388, "step": 62040 }, { "epoch": 5.37, "learning_rate": 6.16536081878108e-05, "loss": 1.6124, "step": 62050 }, { "epoch": 5.37, "learning_rate": 6.165143980571296e-05, "loss": 1.7019, "step": 62060 }, { "epoch": 5.38, "learning_rate": 6.164927142361512e-05, "loss": 1.5984, "step": 62070 }, { "epoch": 5.38, "learning_rate": 6.164710304151728e-05, "loss": 1.4906, "step": 62080 }, { "epoch": 5.38, "learning_rate": 6.164493465941945e-05, "loss": 1.7257, "step": 62090 }, { "epoch": 5.38, "learning_rate": 6.164276627732161e-05, "loss": 1.5469, "step": 62100 }, { "epoch": 5.38, "learning_rate": 6.164059789522377e-05, "loss": 1.4111, "step": 62110 }, { "epoch": 5.38, "learning_rate": 6.163842951312593e-05, "loss": 1.395, "step": 62120 }, { "epoch": 5.38, "learning_rate": 6.16362611310281e-05, "loss": 1.7807, "step": 62130 }, { "epoch": 5.38, "learning_rate": 6.163409274893025e-05, "loss": 1.4756, "step": 62140 }, { "epoch": 5.38, "learning_rate": 6.163192436683242e-05, "loss": 1.4647, "step": 62150 }, { "epoch": 5.38, "learning_rate": 6.162975598473459e-05, "loss": 1.4551, "step": 62160 }, { "epoch": 5.38, "learning_rate": 6.162758760263675e-05, "loss": 1.4936, "step": 62170 }, { "epoch": 5.39, "learning_rate": 6.16254192205389e-05, "loss": 1.544, "step": 62180 }, { "epoch": 5.39, "learning_rate": 6.162325083844107e-05, "loss": 1.6564, "step": 62190 }, { "epoch": 5.39, "learning_rate": 6.162108245634324e-05, "loss": 1.6043, "step": 62200 }, { "epoch": 5.39, "learning_rate": 6.16189140742454e-05, "loss": 1.6319, "step": 62210 }, { "epoch": 5.39, "learning_rate": 6.161674569214756e-05, "loss": 1.5767, "step": 62220 }, { "epoch": 5.39, "learning_rate": 6.161457731004972e-05, "loss": 1.5224, "step": 62230 }, { "epoch": 5.39, "learning_rate": 6.161240892795189e-05, "loss": 1.5741, "step": 62240 }, { "epoch": 5.39, "learning_rate": 6.161024054585404e-05, "loss": 1.545, "step": 62250 }, { "epoch": 5.39, "learning_rate": 6.160807216375621e-05, "loss": 1.7753, "step": 62260 }, { "epoch": 5.39, "learning_rate": 6.160590378165838e-05, "loss": 1.4297, "step": 62270 }, { "epoch": 5.39, "learning_rate": 6.160373539956054e-05, "loss": 1.3766, "step": 62280 }, { "epoch": 5.39, "learning_rate": 6.160156701746271e-05, "loss": 1.5151, "step": 62290 }, { "epoch": 5.4, "learning_rate": 6.159939863536486e-05, "loss": 1.6432, "step": 62300 }, { "epoch": 5.4, "learning_rate": 6.159723025326703e-05, "loss": 1.4899, "step": 62310 }, { "epoch": 5.4, "learning_rate": 6.159506187116918e-05, "loss": 1.7194, "step": 62320 }, { "epoch": 5.4, "learning_rate": 6.159289348907135e-05, "loss": 1.6566, "step": 62330 }, { "epoch": 5.4, "learning_rate": 6.159072510697351e-05, "loss": 1.4771, "step": 62340 }, { "epoch": 5.4, "learning_rate": 6.158855672487568e-05, "loss": 1.4442, "step": 62350 }, { "epoch": 5.4, "learning_rate": 6.158638834277785e-05, "loss": 1.6461, "step": 62360 }, { "epoch": 5.4, "learning_rate": 6.158421996068e-05, "loss": 1.6693, "step": 62370 }, { "epoch": 5.4, "learning_rate": 6.158205157858217e-05, "loss": 1.7127, "step": 62380 }, { "epoch": 5.4, "learning_rate": 6.157988319648432e-05, "loss": 1.5456, "step": 62390 }, { "epoch": 5.4, "learning_rate": 6.157771481438648e-05, "loss": 1.7131, "step": 62400 }, { "epoch": 5.41, "learning_rate": 6.157554643228865e-05, "loss": 1.6753, "step": 62410 }, { "epoch": 5.41, "learning_rate": 6.157337805019082e-05, "loss": 1.6134, "step": 62420 }, { "epoch": 5.41, "learning_rate": 6.157120966809298e-05, "loss": 1.6984, "step": 62430 }, { "epoch": 5.41, "learning_rate": 6.156904128599514e-05, "loss": 1.5821, "step": 62440 }, { "epoch": 5.41, "learning_rate": 6.156687290389729e-05, "loss": 1.6555, "step": 62450 }, { "epoch": 5.41, "learning_rate": 6.156470452179946e-05, "loss": 1.5648, "step": 62460 }, { "epoch": 5.41, "learning_rate": 6.156253613970162e-05, "loss": 1.5123, "step": 62470 }, { "epoch": 5.41, "learning_rate": 6.156036775760379e-05, "loss": 1.5582, "step": 62480 }, { "epoch": 5.41, "learning_rate": 6.155819937550596e-05, "loss": 1.4889, "step": 62490 }, { "epoch": 5.41, "learning_rate": 6.155603099340811e-05, "loss": 1.3937, "step": 62500 }, { "epoch": 5.41, "learning_rate": 6.155386261131027e-05, "loss": 1.4032, "step": 62510 }, { "epoch": 5.41, "learning_rate": 6.155169422921244e-05, "loss": 1.4247, "step": 62520 }, { "epoch": 5.42, "learning_rate": 6.15495258471146e-05, "loss": 1.3954, "step": 62530 }, { "epoch": 5.42, "learning_rate": 6.154735746501676e-05, "loss": 1.5811, "step": 62540 }, { "epoch": 5.42, "learning_rate": 6.154518908291893e-05, "loss": 1.6715, "step": 62550 }, { "epoch": 5.42, "learning_rate": 6.154302070082109e-05, "loss": 1.6795, "step": 62560 }, { "epoch": 5.42, "learning_rate": 6.154085231872325e-05, "loss": 1.5481, "step": 62570 }, { "epoch": 5.42, "learning_rate": 6.153868393662541e-05, "loss": 1.7229, "step": 62580 }, { "epoch": 5.42, "learning_rate": 6.153651555452758e-05, "loss": 1.5709, "step": 62590 }, { "epoch": 5.42, "learning_rate": 6.153434717242973e-05, "loss": 1.7494, "step": 62600 }, { "epoch": 5.42, "learning_rate": 6.15321787903319e-05, "loss": 1.4655, "step": 62610 }, { "epoch": 5.42, "learning_rate": 6.153001040823406e-05, "loss": 1.5912, "step": 62620 }, { "epoch": 5.42, "learning_rate": 6.152784202613623e-05, "loss": 1.5428, "step": 62630 }, { "epoch": 5.43, "learning_rate": 6.152567364403838e-05, "loss": 1.5672, "step": 62640 }, { "epoch": 5.43, "learning_rate": 6.152350526194055e-05, "loss": 1.6344, "step": 62650 }, { "epoch": 5.43, "learning_rate": 6.152133687984272e-05, "loss": 1.7258, "step": 62660 }, { "epoch": 5.43, "learning_rate": 6.151916849774488e-05, "loss": 1.4953, "step": 62670 }, { "epoch": 5.43, "learning_rate": 6.151700011564704e-05, "loss": 1.629, "step": 62680 }, { "epoch": 5.43, "learning_rate": 6.15148317335492e-05, "loss": 1.6659, "step": 62690 }, { "epoch": 5.43, "learning_rate": 6.151266335145137e-05, "loss": 1.5056, "step": 62700 }, { "epoch": 5.43, "learning_rate": 6.151049496935352e-05, "loss": 1.6794, "step": 62710 }, { "epoch": 5.43, "learning_rate": 6.150832658725569e-05, "loss": 1.401, "step": 62720 }, { "epoch": 5.43, "learning_rate": 6.150615820515785e-05, "loss": 1.8153, "step": 62730 }, { "epoch": 5.43, "learning_rate": 6.150398982306002e-05, "loss": 1.4968, "step": 62740 }, { "epoch": 5.43, "learning_rate": 6.150182144096219e-05, "loss": 1.5491, "step": 62750 }, { "epoch": 5.44, "learning_rate": 6.149965305886434e-05, "loss": 1.5316, "step": 62760 }, { "epoch": 5.44, "learning_rate": 6.14974846767665e-05, "loss": 1.5375, "step": 62770 }, { "epoch": 5.44, "learning_rate": 6.149531629466866e-05, "loss": 1.719, "step": 62780 }, { "epoch": 5.44, "learning_rate": 6.149314791257082e-05, "loss": 1.6452, "step": 62790 }, { "epoch": 5.44, "learning_rate": 6.149097953047299e-05, "loss": 1.4717, "step": 62800 }, { "epoch": 5.44, "learning_rate": 6.148881114837516e-05, "loss": 1.7731, "step": 62810 }, { "epoch": 5.44, "learning_rate": 6.148664276627732e-05, "loss": 1.7818, "step": 62820 }, { "epoch": 5.44, "learning_rate": 6.148447438417948e-05, "loss": 1.3398, "step": 62830 }, { "epoch": 5.44, "learning_rate": 6.148230600208164e-05, "loss": 1.4661, "step": 62840 }, { "epoch": 5.44, "learning_rate": 6.14801376199838e-05, "loss": 1.4272, "step": 62850 }, { "epoch": 5.44, "learning_rate": 6.147796923788596e-05, "loss": 1.5303, "step": 62860 }, { "epoch": 5.44, "learning_rate": 6.147580085578813e-05, "loss": 1.4952, "step": 62870 }, { "epoch": 5.45, "learning_rate": 6.14736324736903e-05, "loss": 1.3758, "step": 62880 }, { "epoch": 5.45, "learning_rate": 6.147146409159246e-05, "loss": 1.6687, "step": 62890 }, { "epoch": 5.45, "learning_rate": 6.146929570949461e-05, "loss": 1.7038, "step": 62900 }, { "epoch": 5.45, "learning_rate": 6.146712732739678e-05, "loss": 1.7153, "step": 62910 }, { "epoch": 5.45, "learning_rate": 6.146495894529893e-05, "loss": 1.6152, "step": 62920 }, { "epoch": 5.45, "learning_rate": 6.14627905632011e-05, "loss": 1.5897, "step": 62930 }, { "epoch": 5.45, "learning_rate": 6.146062218110327e-05, "loss": 1.5824, "step": 62940 }, { "epoch": 5.45, "learning_rate": 6.145845379900543e-05, "loss": 1.4161, "step": 62950 }, { "epoch": 5.45, "learning_rate": 6.14562854169076e-05, "loss": 1.6165, "step": 62960 }, { "epoch": 5.45, "learning_rate": 6.145411703480975e-05, "loss": 1.6002, "step": 62970 }, { "epoch": 5.45, "learning_rate": 6.145194865271192e-05, "loss": 1.4786, "step": 62980 }, { "epoch": 5.46, "learning_rate": 6.144978027061407e-05, "loss": 1.5071, "step": 62990 }, { "epoch": 5.46, "learning_rate": 6.144761188851624e-05, "loss": 1.8117, "step": 63000 }, { "epoch": 5.46, "learning_rate": 6.14454435064184e-05, "loss": 1.8245, "step": 63010 }, { "epoch": 5.46, "learning_rate": 6.144327512432057e-05, "loss": 1.489, "step": 63020 }, { "epoch": 5.46, "learning_rate": 6.144110674222274e-05, "loss": 1.6399, "step": 63030 }, { "epoch": 5.46, "learning_rate": 6.143893836012489e-05, "loss": 1.5387, "step": 63040 }, { "epoch": 5.46, "learning_rate": 6.143676997802706e-05, "loss": 1.6733, "step": 63050 }, { "epoch": 5.46, "learning_rate": 6.143460159592922e-05, "loss": 1.5711, "step": 63060 }, { "epoch": 5.46, "learning_rate": 6.143243321383138e-05, "loss": 1.4515, "step": 63070 }, { "epoch": 5.46, "learning_rate": 6.143026483173354e-05, "loss": 1.7782, "step": 63080 }, { "epoch": 5.46, "learning_rate": 6.142809644963571e-05, "loss": 1.5713, "step": 63090 }, { "epoch": 5.46, "learning_rate": 6.142592806753787e-05, "loss": 1.5161, "step": 63100 }, { "epoch": 5.47, "learning_rate": 6.142375968544003e-05, "loss": 1.3659, "step": 63110 }, { "epoch": 5.47, "learning_rate": 6.14215913033422e-05, "loss": 1.6082, "step": 63120 }, { "epoch": 5.47, "learning_rate": 6.141942292124436e-05, "loss": 1.4693, "step": 63130 }, { "epoch": 5.47, "learning_rate": 6.141725453914651e-05, "loss": 1.4857, "step": 63140 }, { "epoch": 5.47, "learning_rate": 6.141508615704868e-05, "loss": 1.6269, "step": 63150 }, { "epoch": 5.47, "learning_rate": 6.141291777495085e-05, "loss": 1.5029, "step": 63160 }, { "epoch": 5.47, "learning_rate": 6.141074939285301e-05, "loss": 1.4415, "step": 63170 }, { "epoch": 5.47, "learning_rate": 6.140858101075517e-05, "loss": 1.5936, "step": 63180 }, { "epoch": 5.47, "learning_rate": 6.140641262865733e-05, "loss": 1.4365, "step": 63190 }, { "epoch": 5.47, "learning_rate": 6.14042442465595e-05, "loss": 1.5264, "step": 63200 }, { "epoch": 5.47, "learning_rate": 6.140207586446166e-05, "loss": 1.5407, "step": 63210 }, { "epoch": 5.48, "learning_rate": 6.139990748236382e-05, "loss": 1.6733, "step": 63220 }, { "epoch": 5.48, "learning_rate": 6.139773910026598e-05, "loss": 1.6792, "step": 63230 }, { "epoch": 5.48, "learning_rate": 6.139557071816815e-05, "loss": 1.5995, "step": 63240 }, { "epoch": 5.48, "learning_rate": 6.13934023360703e-05, "loss": 1.5239, "step": 63250 }, { "epoch": 5.48, "learning_rate": 6.139123395397247e-05, "loss": 1.5415, "step": 63260 }, { "epoch": 5.48, "learning_rate": 6.138906557187464e-05, "loss": 1.6598, "step": 63270 }, { "epoch": 5.48, "learning_rate": 6.13868971897768e-05, "loss": 1.8637, "step": 63280 }, { "epoch": 5.48, "learning_rate": 6.138472880767897e-05, "loss": 1.5869, "step": 63290 }, { "epoch": 5.48, "learning_rate": 6.138256042558112e-05, "loss": 1.4576, "step": 63300 }, { "epoch": 5.48, "learning_rate": 6.138039204348329e-05, "loss": 1.6105, "step": 63310 }, { "epoch": 5.48, "learning_rate": 6.137822366138544e-05, "loss": 1.737, "step": 63320 }, { "epoch": 5.48, "learning_rate": 6.137605527928761e-05, "loss": 1.5003, "step": 63330 }, { "epoch": 5.49, "learning_rate": 6.137388689718977e-05, "loss": 1.4277, "step": 63340 }, { "epoch": 5.49, "learning_rate": 6.137171851509194e-05, "loss": 1.5229, "step": 63350 }, { "epoch": 5.49, "learning_rate": 6.13695501329941e-05, "loss": 1.7158, "step": 63360 }, { "epoch": 5.49, "learning_rate": 6.136738175089626e-05, "loss": 1.6976, "step": 63370 }, { "epoch": 5.49, "learning_rate": 6.136521336879843e-05, "loss": 1.6851, "step": 63380 }, { "epoch": 5.49, "learning_rate": 6.136304498670058e-05, "loss": 1.6421, "step": 63390 }, { "epoch": 5.49, "learning_rate": 6.136087660460274e-05, "loss": 1.5793, "step": 63400 }, { "epoch": 5.49, "learning_rate": 6.135870822250491e-05, "loss": 1.6663, "step": 63410 }, { "epoch": 5.49, "learning_rate": 6.135653984040708e-05, "loss": 1.6075, "step": 63420 }, { "epoch": 5.49, "learning_rate": 6.135437145830924e-05, "loss": 1.3834, "step": 63430 }, { "epoch": 5.49, "learning_rate": 6.13522030762114e-05, "loss": 1.603, "step": 63440 }, { "epoch": 5.5, "learning_rate": 6.135003469411356e-05, "loss": 1.6546, "step": 63450 }, { "epoch": 5.5, "learning_rate": 6.134786631201572e-05, "loss": 1.8627, "step": 63460 }, { "epoch": 5.5, "learning_rate": 6.134569792991788e-05, "loss": 1.5763, "step": 63470 }, { "epoch": 5.5, "learning_rate": 6.134352954782005e-05, "loss": 1.7098, "step": 63480 }, { "epoch": 5.5, "learning_rate": 6.134136116572221e-05, "loss": 1.6232, "step": 63490 }, { "epoch": 5.5, "learning_rate": 6.133919278362438e-05, "loss": 1.5815, "step": 63500 }, { "epoch": 5.5, "learning_rate": 6.133702440152653e-05, "loss": 1.6682, "step": 63510 }, { "epoch": 5.5, "learning_rate": 6.13348560194287e-05, "loss": 1.5394, "step": 63520 }, { "epoch": 5.5, "learning_rate": 6.133268763733085e-05, "loss": 1.6624, "step": 63530 }, { "epoch": 5.5, "learning_rate": 6.133051925523302e-05, "loss": 1.3763, "step": 63540 }, { "epoch": 5.5, "learning_rate": 6.132835087313519e-05, "loss": 1.5265, "step": 63550 }, { "epoch": 5.5, "learning_rate": 6.132618249103735e-05, "loss": 1.6044, "step": 63560 }, { "epoch": 5.51, "learning_rate": 6.13240141089395e-05, "loss": 1.6107, "step": 63570 }, { "epoch": 5.51, "learning_rate": 6.132184572684167e-05, "loss": 1.6788, "step": 63580 }, { "epoch": 5.51, "learning_rate": 6.131967734474384e-05, "loss": 1.5576, "step": 63590 }, { "epoch": 5.51, "learning_rate": 6.1317508962646e-05, "loss": 1.4941, "step": 63600 }, { "epoch": 5.51, "learning_rate": 6.131534058054816e-05, "loss": 1.507, "step": 63610 }, { "epoch": 5.51, "learning_rate": 6.131317219845032e-05, "loss": 1.6306, "step": 63620 }, { "epoch": 5.51, "learning_rate": 6.131100381635249e-05, "loss": 1.4016, "step": 63630 }, { "epoch": 5.51, "learning_rate": 6.130883543425464e-05, "loss": 1.5686, "step": 63640 }, { "epoch": 5.51, "learning_rate": 6.130666705215681e-05, "loss": 1.6394, "step": 63650 }, { "epoch": 5.51, "learning_rate": 6.130449867005898e-05, "loss": 1.6436, "step": 63660 }, { "epoch": 5.51, "learning_rate": 6.130233028796114e-05, "loss": 1.7114, "step": 63670 }, { "epoch": 5.52, "learning_rate": 6.13001619058633e-05, "loss": 1.4631, "step": 63680 }, { "epoch": 5.52, "learning_rate": 6.129799352376546e-05, "loss": 1.6717, "step": 63690 }, { "epoch": 5.52, "learning_rate": 6.129582514166763e-05, "loss": 1.6782, "step": 63700 }, { "epoch": 5.52, "learning_rate": 6.129365675956978e-05, "loss": 1.6716, "step": 63710 }, { "epoch": 5.52, "learning_rate": 6.129148837747195e-05, "loss": 1.7252, "step": 63720 }, { "epoch": 5.52, "learning_rate": 6.128931999537411e-05, "loss": 1.5633, "step": 63730 }, { "epoch": 5.52, "learning_rate": 6.128715161327628e-05, "loss": 1.4743, "step": 63740 }, { "epoch": 5.52, "learning_rate": 6.128498323117845e-05, "loss": 1.3845, "step": 63750 }, { "epoch": 5.52, "learning_rate": 6.12828148490806e-05, "loss": 1.8658, "step": 63760 }, { "epoch": 5.52, "learning_rate": 6.128064646698277e-05, "loss": 1.4749, "step": 63770 }, { "epoch": 5.52, "learning_rate": 6.127847808488492e-05, "loss": 1.6077, "step": 63780 }, { "epoch": 5.52, "learning_rate": 6.127630970278708e-05, "loss": 1.4393, "step": 63790 }, { "epoch": 5.53, "learning_rate": 6.127414132068925e-05, "loss": 1.6283, "step": 63800 }, { "epoch": 5.53, "learning_rate": 6.127197293859142e-05, "loss": 1.5238, "step": 63810 }, { "epoch": 5.53, "learning_rate": 6.126980455649358e-05, "loss": 1.7618, "step": 63820 }, { "epoch": 5.53, "learning_rate": 6.126763617439574e-05, "loss": 1.6996, "step": 63830 }, { "epoch": 5.53, "learning_rate": 6.12654677922979e-05, "loss": 1.5797, "step": 63840 }, { "epoch": 5.53, "learning_rate": 6.126329941020006e-05, "loss": 1.7392, "step": 63850 }, { "epoch": 5.53, "learning_rate": 6.126113102810222e-05, "loss": 1.5393, "step": 63860 }, { "epoch": 5.53, "learning_rate": 6.125896264600439e-05, "loss": 1.5286, "step": 63870 }, { "epoch": 5.53, "learning_rate": 6.125679426390656e-05, "loss": 1.6219, "step": 63880 }, { "epoch": 5.53, "learning_rate": 6.125462588180872e-05, "loss": 1.6586, "step": 63890 }, { "epoch": 5.53, "learning_rate": 6.125245749971087e-05, "loss": 1.534, "step": 63900 }, { "epoch": 5.54, "learning_rate": 6.125028911761304e-05, "loss": 1.5992, "step": 63910 }, { "epoch": 5.54, "learning_rate": 6.12481207355152e-05, "loss": 1.444, "step": 63920 }, { "epoch": 5.54, "learning_rate": 6.124595235341736e-05, "loss": 1.4037, "step": 63930 }, { "epoch": 5.54, "learning_rate": 6.124378397131953e-05, "loss": 1.5032, "step": 63940 }, { "epoch": 5.54, "learning_rate": 6.124161558922169e-05, "loss": 1.6503, "step": 63950 }, { "epoch": 5.54, "learning_rate": 6.123944720712386e-05, "loss": 1.6735, "step": 63960 }, { "epoch": 5.54, "learning_rate": 6.123727882502601e-05, "loss": 1.6884, "step": 63970 }, { "epoch": 5.54, "learning_rate": 6.123511044292818e-05, "loss": 1.6795, "step": 63980 }, { "epoch": 5.54, "learning_rate": 6.123294206083033e-05, "loss": 1.5826, "step": 63990 }, { "epoch": 5.54, "learning_rate": 6.12307736787325e-05, "loss": 1.5467, "step": 64000 }, { "epoch": 5.54, "learning_rate": 6.122860529663466e-05, "loss": 1.5727, "step": 64010 }, { "epoch": 5.54, "learning_rate": 6.122643691453683e-05, "loss": 1.6089, "step": 64020 }, { "epoch": 5.55, "learning_rate": 6.1224268532439e-05, "loss": 1.6899, "step": 64030 }, { "epoch": 5.55, "learning_rate": 6.122210015034115e-05, "loss": 1.4974, "step": 64040 }, { "epoch": 5.55, "learning_rate": 6.121993176824332e-05, "loss": 1.7738, "step": 64050 }, { "epoch": 5.55, "learning_rate": 6.121776338614548e-05, "loss": 1.7062, "step": 64060 }, { "epoch": 5.55, "learning_rate": 6.121559500404764e-05, "loss": 1.788, "step": 64070 }, { "epoch": 5.55, "learning_rate": 6.12134266219498e-05, "loss": 1.5889, "step": 64080 }, { "epoch": 5.55, "learning_rate": 6.121125823985197e-05, "loss": 1.8375, "step": 64090 }, { "epoch": 5.55, "learning_rate": 6.120908985775413e-05, "loss": 1.7236, "step": 64100 }, { "epoch": 5.55, "learning_rate": 6.120692147565629e-05, "loss": 1.6003, "step": 64110 }, { "epoch": 5.55, "learning_rate": 6.120475309355845e-05, "loss": 1.6043, "step": 64120 }, { "epoch": 5.55, "learning_rate": 6.120258471146062e-05, "loss": 1.45, "step": 64130 }, { "epoch": 5.55, "learning_rate": 6.120041632936277e-05, "loss": 1.6164, "step": 64140 }, { "epoch": 5.56, "learning_rate": 6.119824794726494e-05, "loss": 1.8515, "step": 64150 }, { "epoch": 5.56, "learning_rate": 6.11960795651671e-05, "loss": 1.6966, "step": 64160 }, { "epoch": 5.56, "learning_rate": 6.119391118306927e-05, "loss": 1.7265, "step": 64170 }, { "epoch": 5.56, "learning_rate": 6.119174280097142e-05, "loss": 1.641, "step": 64180 }, { "epoch": 5.56, "learning_rate": 6.118957441887359e-05, "loss": 1.6696, "step": 64190 }, { "epoch": 5.56, "learning_rate": 6.118740603677576e-05, "loss": 1.6883, "step": 64200 }, { "epoch": 5.56, "learning_rate": 6.118523765467792e-05, "loss": 1.4592, "step": 64210 }, { "epoch": 5.56, "learning_rate": 6.118306927258008e-05, "loss": 1.5019, "step": 64220 }, { "epoch": 5.56, "learning_rate": 6.118090089048224e-05, "loss": 1.6492, "step": 64230 }, { "epoch": 5.56, "learning_rate": 6.117873250838441e-05, "loss": 1.3624, "step": 64240 }, { "epoch": 5.56, "learning_rate": 6.117656412628656e-05, "loss": 1.46, "step": 64250 }, { "epoch": 5.57, "learning_rate": 6.117439574418873e-05, "loss": 1.5695, "step": 64260 }, { "epoch": 5.57, "learning_rate": 6.11722273620909e-05, "loss": 1.7057, "step": 64270 }, { "epoch": 5.57, "learning_rate": 6.117005897999306e-05, "loss": 1.6606, "step": 64280 }, { "epoch": 5.57, "learning_rate": 6.116789059789523e-05, "loss": 1.7078, "step": 64290 }, { "epoch": 5.57, "learning_rate": 6.116572221579738e-05, "loss": 1.3767, "step": 64300 }, { "epoch": 5.57, "learning_rate": 6.116355383369955e-05, "loss": 1.8138, "step": 64310 }, { "epoch": 5.57, "learning_rate": 6.11613854516017e-05, "loss": 1.6022, "step": 64320 }, { "epoch": 5.57, "learning_rate": 6.115921706950387e-05, "loss": 1.6672, "step": 64330 }, { "epoch": 5.57, "learning_rate": 6.115704868740603e-05, "loss": 1.5325, "step": 64340 }, { "epoch": 5.57, "learning_rate": 6.11548803053082e-05, "loss": 1.578, "step": 64350 }, { "epoch": 5.57, "learning_rate": 6.115271192321037e-05, "loss": 1.5983, "step": 64360 }, { "epoch": 5.57, "learning_rate": 6.115054354111252e-05, "loss": 1.738, "step": 64370 }, { "epoch": 5.58, "learning_rate": 6.114837515901468e-05, "loss": 1.6569, "step": 64380 }, { "epoch": 5.58, "learning_rate": 6.114620677691684e-05, "loss": 1.6236, "step": 64390 }, { "epoch": 5.58, "learning_rate": 6.1144038394819e-05, "loss": 1.6321, "step": 64400 }, { "epoch": 5.58, "learning_rate": 6.114187001272117e-05, "loss": 1.673, "step": 64410 }, { "epoch": 5.58, "learning_rate": 6.113970163062334e-05, "loss": 1.6281, "step": 64420 }, { "epoch": 5.58, "learning_rate": 6.11375332485255e-05, "loss": 1.4161, "step": 64430 }, { "epoch": 5.58, "learning_rate": 6.113536486642766e-05, "loss": 1.6377, "step": 64440 }, { "epoch": 5.58, "learning_rate": 6.113319648432982e-05, "loss": 1.6072, "step": 64450 }, { "epoch": 5.58, "learning_rate": 6.113102810223198e-05, "loss": 1.5958, "step": 64460 }, { "epoch": 5.58, "learning_rate": 6.112885972013414e-05, "loss": 1.6344, "step": 64470 }, { "epoch": 5.58, "learning_rate": 6.112669133803631e-05, "loss": 1.5832, "step": 64480 }, { "epoch": 5.59, "learning_rate": 6.112452295593847e-05, "loss": 1.7361, "step": 64490 }, { "epoch": 5.59, "learning_rate": 6.112235457384064e-05, "loss": 1.66, "step": 64500 }, { "epoch": 5.59, "learning_rate": 6.11201861917428e-05, "loss": 1.438, "step": 64510 }, { "epoch": 5.59, "learning_rate": 6.111801780964496e-05, "loss": 1.4753, "step": 64520 }, { "epoch": 5.59, "learning_rate": 6.111584942754711e-05, "loss": 1.5555, "step": 64530 }, { "epoch": 5.59, "learning_rate": 6.111368104544928e-05, "loss": 1.4373, "step": 64540 }, { "epoch": 5.59, "learning_rate": 6.111151266335145e-05, "loss": 1.469, "step": 64550 }, { "epoch": 5.59, "learning_rate": 6.110934428125361e-05, "loss": 1.5452, "step": 64560 }, { "epoch": 5.59, "learning_rate": 6.110717589915578e-05, "loss": 1.7761, "step": 64570 }, { "epoch": 5.59, "learning_rate": 6.110500751705793e-05, "loss": 1.5352, "step": 64580 }, { "epoch": 5.59, "learning_rate": 6.11028391349601e-05, "loss": 1.6349, "step": 64590 }, { "epoch": 5.59, "learning_rate": 6.110067075286226e-05, "loss": 1.5636, "step": 64600 }, { "epoch": 5.6, "learning_rate": 6.109850237076442e-05, "loss": 1.5672, "step": 64610 }, { "epoch": 5.6, "learning_rate": 6.109633398866658e-05, "loss": 1.3851, "step": 64620 }, { "epoch": 5.6, "learning_rate": 6.109416560656875e-05, "loss": 1.7157, "step": 64630 }, { "epoch": 5.6, "learning_rate": 6.10919972244709e-05, "loss": 1.3884, "step": 64640 }, { "epoch": 5.6, "learning_rate": 6.108982884237307e-05, "loss": 1.7436, "step": 64650 }, { "epoch": 5.6, "learning_rate": 6.108766046027524e-05, "loss": 1.6493, "step": 64660 }, { "epoch": 5.6, "learning_rate": 6.10854920781774e-05, "loss": 1.4435, "step": 64670 }, { "epoch": 5.6, "learning_rate": 6.108332369607955e-05, "loss": 1.6582, "step": 64680 }, { "epoch": 5.6, "learning_rate": 6.108115531398172e-05, "loss": 1.5794, "step": 64690 }, { "epoch": 5.6, "learning_rate": 6.107898693188389e-05, "loss": 1.6218, "step": 64700 }, { "epoch": 5.6, "learning_rate": 6.107681854978604e-05, "loss": 1.653, "step": 64710 }, { "epoch": 5.61, "learning_rate": 6.10746501676882e-05, "loss": 1.5933, "step": 64720 }, { "epoch": 5.61, "learning_rate": 6.107248178559037e-05, "loss": 1.5234, "step": 64730 }, { "epoch": 5.61, "learning_rate": 6.107031340349254e-05, "loss": 1.5309, "step": 64740 }, { "epoch": 5.61, "learning_rate": 6.10681450213947e-05, "loss": 1.6357, "step": 64750 }, { "epoch": 5.61, "learning_rate": 6.106597663929686e-05, "loss": 1.5764, "step": 64760 }, { "epoch": 5.61, "learning_rate": 6.106380825719903e-05, "loss": 1.5374, "step": 64770 }, { "epoch": 5.61, "learning_rate": 6.106163987510118e-05, "loss": 1.4121, "step": 64780 }, { "epoch": 5.61, "learning_rate": 6.105947149300334e-05, "loss": 1.6888, "step": 64790 }, { "epoch": 5.61, "learning_rate": 6.105730311090551e-05, "loss": 1.7593, "step": 64800 }, { "epoch": 5.61, "learning_rate": 6.105513472880768e-05, "loss": 1.5655, "step": 64810 }, { "epoch": 5.61, "learning_rate": 6.105296634670984e-05, "loss": 1.4546, "step": 64820 }, { "epoch": 5.61, "learning_rate": 6.1050797964612e-05, "loss": 1.5325, "step": 64830 }, { "epoch": 5.62, "learning_rate": 6.104862958251416e-05, "loss": 1.6289, "step": 64840 }, { "epoch": 5.62, "learning_rate": 6.104646120041632e-05, "loss": 1.5447, "step": 64850 }, { "epoch": 5.62, "learning_rate": 6.104429281831848e-05, "loss": 1.621, "step": 64860 }, { "epoch": 5.62, "learning_rate": 6.104212443622065e-05, "loss": 1.8714, "step": 64870 }, { "epoch": 5.62, "learning_rate": 6.103995605412281e-05, "loss": 1.7845, "step": 64880 }, { "epoch": 5.62, "learning_rate": 6.103778767202498e-05, "loss": 1.4392, "step": 64890 }, { "epoch": 5.62, "learning_rate": 6.103561928992713e-05, "loss": 1.5759, "step": 64900 }, { "epoch": 5.62, "learning_rate": 6.1033450907829294e-05, "loss": 1.5325, "step": 64910 }, { "epoch": 5.62, "learning_rate": 6.103128252573146e-05, "loss": 1.6636, "step": 64920 }, { "epoch": 5.62, "learning_rate": 6.102911414363362e-05, "loss": 1.5111, "step": 64930 }, { "epoch": 5.62, "learning_rate": 6.1026945761535786e-05, "loss": 1.5809, "step": 64940 }, { "epoch": 5.63, "learning_rate": 6.102477737943795e-05, "loss": 1.5799, "step": 64950 }, { "epoch": 5.63, "learning_rate": 6.102260899734011e-05, "loss": 1.3761, "step": 64960 }, { "epoch": 5.63, "learning_rate": 6.102044061524228e-05, "loss": 1.6355, "step": 64970 }, { "epoch": 5.63, "learning_rate": 6.101827223314444e-05, "loss": 1.5779, "step": 64980 }, { "epoch": 5.63, "learning_rate": 6.10161038510466e-05, "loss": 1.5018, "step": 64990 }, { "epoch": 5.63, "learning_rate": 6.101393546894876e-05, "loss": 1.7149, "step": 65000 }, { "epoch": 5.63, "learning_rate": 6.1011767086850924e-05, "loss": 1.5768, "step": 65010 }, { "epoch": 5.63, "learning_rate": 6.100959870475309e-05, "loss": 1.4752, "step": 65020 }, { "epoch": 5.63, "learning_rate": 6.100743032265525e-05, "loss": 1.6044, "step": 65030 }, { "epoch": 5.63, "learning_rate": 6.1005261940557416e-05, "loss": 1.5538, "step": 65040 }, { "epoch": 5.63, "learning_rate": 6.1003093558459576e-05, "loss": 1.7931, "step": 65050 }, { "epoch": 5.63, "learning_rate": 6.100092517636174e-05, "loss": 1.5204, "step": 65060 }, { "epoch": 5.64, "learning_rate": 6.0998756794263895e-05, "loss": 1.6886, "step": 65070 }, { "epoch": 5.64, "learning_rate": 6.099658841216606e-05, "loss": 1.4954, "step": 65080 }, { "epoch": 5.64, "learning_rate": 6.099442003006823e-05, "loss": 1.9747, "step": 65090 }, { "epoch": 5.64, "learning_rate": 6.099225164797039e-05, "loss": 1.5776, "step": 65100 }, { "epoch": 5.64, "learning_rate": 6.0990083265872554e-05, "loss": 1.643, "step": 65110 }, { "epoch": 5.64, "learning_rate": 6.098791488377471e-05, "loss": 1.4885, "step": 65120 }, { "epoch": 5.64, "learning_rate": 6.098574650167688e-05, "loss": 1.3708, "step": 65130 }, { "epoch": 5.64, "learning_rate": 6.0983578119579046e-05, "loss": 1.437, "step": 65140 }, { "epoch": 5.64, "learning_rate": 6.09814097374812e-05, "loss": 1.6254, "step": 65150 }, { "epoch": 5.64, "learning_rate": 6.0979241355383365e-05, "loss": 1.6417, "step": 65160 }, { "epoch": 5.64, "learning_rate": 6.0977072973285525e-05, "loss": 1.8998, "step": 65170 }, { "epoch": 5.65, "learning_rate": 6.097490459118769e-05, "loss": 1.5487, "step": 65180 }, { "epoch": 5.65, "learning_rate": 6.097273620908985e-05, "loss": 1.3328, "step": 65190 }, { "epoch": 5.65, "learning_rate": 6.097056782699202e-05, "loss": 1.6163, "step": 65200 }, { "epoch": 5.65, "learning_rate": 6.0968399444894184e-05, "loss": 1.6817, "step": 65210 }, { "epoch": 5.65, "learning_rate": 6.0966231062796337e-05, "loss": 1.4995, "step": 65220 }, { "epoch": 5.65, "learning_rate": 6.09640626806985e-05, "loss": 1.5018, "step": 65230 }, { "epoch": 5.65, "learning_rate": 6.096189429860066e-05, "loss": 1.6703, "step": 65240 }, { "epoch": 5.65, "learning_rate": 6.095972591650283e-05, "loss": 1.6136, "step": 65250 }, { "epoch": 5.65, "learning_rate": 6.095755753440499e-05, "loss": 1.4381, "step": 65260 }, { "epoch": 5.65, "learning_rate": 6.0955389152307155e-05, "loss": 1.7816, "step": 65270 }, { "epoch": 5.65, "learning_rate": 6.095322077020932e-05, "loss": 1.6615, "step": 65280 }, { "epoch": 5.65, "learning_rate": 6.095105238811148e-05, "loss": 1.5153, "step": 65290 }, { "epoch": 5.66, "learning_rate": 6.094888400601364e-05, "loss": 1.6153, "step": 65300 }, { "epoch": 5.66, "learning_rate": 6.09467156239158e-05, "loss": 1.6044, "step": 65310 }, { "epoch": 5.66, "learning_rate": 6.0944547241817967e-05, "loss": 1.5985, "step": 65320 }, { "epoch": 5.66, "learning_rate": 6.0942378859720126e-05, "loss": 1.488, "step": 65330 }, { "epoch": 5.66, "learning_rate": 6.094021047762229e-05, "loss": 1.7322, "step": 65340 }, { "epoch": 5.66, "learning_rate": 6.093804209552446e-05, "loss": 1.747, "step": 65350 }, { "epoch": 5.66, "learning_rate": 6.093587371342662e-05, "loss": 1.6983, "step": 65360 }, { "epoch": 5.66, "learning_rate": 6.0933705331328785e-05, "loss": 1.5909, "step": 65370 }, { "epoch": 5.66, "learning_rate": 6.093153694923094e-05, "loss": 1.5066, "step": 65380 }, { "epoch": 5.66, "learning_rate": 6.0929368567133104e-05, "loss": 1.4757, "step": 65390 }, { "epoch": 5.66, "learning_rate": 6.0927200185035264e-05, "loss": 1.598, "step": 65400 }, { "epoch": 5.66, "learning_rate": 6.092503180293743e-05, "loss": 1.4868, "step": 65410 }, { "epoch": 5.67, "learning_rate": 6.09228634208396e-05, "loss": 1.3823, "step": 65420 }, { "epoch": 5.67, "learning_rate": 6.0920695038741756e-05, "loss": 1.568, "step": 65430 }, { "epoch": 5.67, "learning_rate": 6.091852665664392e-05, "loss": 1.8037, "step": 65440 }, { "epoch": 5.67, "learning_rate": 6.0916358274546075e-05, "loss": 1.5493, "step": 65450 }, { "epoch": 5.67, "learning_rate": 6.091418989244824e-05, "loss": 1.8053, "step": 65460 }, { "epoch": 5.67, "learning_rate": 6.09120215103504e-05, "loss": 1.6976, "step": 65470 }, { "epoch": 5.67, "learning_rate": 6.090985312825257e-05, "loss": 1.4882, "step": 65480 }, { "epoch": 5.67, "learning_rate": 6.090768474615473e-05, "loss": 1.5773, "step": 65490 }, { "epoch": 5.67, "learning_rate": 6.0905516364056894e-05, "loss": 1.4376, "step": 65500 }, { "epoch": 5.67, "learning_rate": 6.090334798195906e-05, "loss": 1.7015, "step": 65510 }, { "epoch": 5.67, "learning_rate": 6.090117959986122e-05, "loss": 1.5003, "step": 65520 }, { "epoch": 5.68, "learning_rate": 6.089901121776338e-05, "loss": 1.4727, "step": 65530 }, { "epoch": 5.68, "learning_rate": 6.089684283566554e-05, "loss": 1.5069, "step": 65540 }, { "epoch": 5.68, "learning_rate": 6.0894674453567706e-05, "loss": 1.2514, "step": 65550 }, { "epoch": 5.68, "learning_rate": 6.0892506071469865e-05, "loss": 1.692, "step": 65560 }, { "epoch": 5.68, "learning_rate": 6.089033768937203e-05, "loss": 1.6167, "step": 65570 }, { "epoch": 5.68, "learning_rate": 6.08881693072742e-05, "loss": 1.7476, "step": 65580 }, { "epoch": 5.68, "learning_rate": 6.088600092517636e-05, "loss": 1.5604, "step": 65590 }, { "epoch": 5.68, "learning_rate": 6.0883832543078524e-05, "loss": 1.4398, "step": 65600 }, { "epoch": 5.68, "learning_rate": 6.088166416098068e-05, "loss": 1.6107, "step": 65610 }, { "epoch": 5.68, "learning_rate": 6.087949577888284e-05, "loss": 1.2573, "step": 65620 }, { "epoch": 5.68, "learning_rate": 6.0877327396785e-05, "loss": 1.7464, "step": 65630 }, { "epoch": 5.68, "learning_rate": 6.087515901468717e-05, "loss": 1.7621, "step": 65640 }, { "epoch": 5.69, "learning_rate": 6.0872990632589336e-05, "loss": 1.6484, "step": 65650 }, { "epoch": 5.69, "learning_rate": 6.0870822250491495e-05, "loss": 1.4986, "step": 65660 }, { "epoch": 5.69, "learning_rate": 6.086865386839366e-05, "loss": 1.4717, "step": 65670 }, { "epoch": 5.69, "learning_rate": 6.0866485486295814e-05, "loss": 1.6319, "step": 65680 }, { "epoch": 5.69, "learning_rate": 6.086431710419798e-05, "loss": 1.5985, "step": 65690 }, { "epoch": 5.69, "learning_rate": 6.086214872210014e-05, "loss": 1.6249, "step": 65700 }, { "epoch": 5.69, "learning_rate": 6.085998034000231e-05, "loss": 1.6662, "step": 65710 }, { "epoch": 5.69, "learning_rate": 6.085781195790447e-05, "loss": 1.6657, "step": 65720 }, { "epoch": 5.69, "learning_rate": 6.085564357580663e-05, "loss": 1.4405, "step": 65730 }, { "epoch": 5.69, "learning_rate": 6.08534751937088e-05, "loss": 1.5639, "step": 65740 }, { "epoch": 5.69, "learning_rate": 6.085130681161096e-05, "loss": 1.5728, "step": 65750 }, { "epoch": 5.7, "learning_rate": 6.084913842951312e-05, "loss": 1.5977, "step": 65760 }, { "epoch": 5.7, "learning_rate": 6.084697004741528e-05, "loss": 1.5974, "step": 65770 }, { "epoch": 5.7, "learning_rate": 6.0844801665317444e-05, "loss": 1.7172, "step": 65780 }, { "epoch": 5.7, "learning_rate": 6.084263328321961e-05, "loss": 1.6163, "step": 65790 }, { "epoch": 5.7, "learning_rate": 6.084046490112177e-05, "loss": 1.558, "step": 65800 }, { "epoch": 5.7, "learning_rate": 6.083829651902394e-05, "loss": 1.5668, "step": 65810 }, { "epoch": 5.7, "learning_rate": 6.0836128136926096e-05, "loss": 1.5907, "step": 65820 }, { "epoch": 5.7, "learning_rate": 6.083395975482826e-05, "loss": 1.3857, "step": 65830 }, { "epoch": 5.7, "learning_rate": 6.0831791372730416e-05, "loss": 1.6715, "step": 65840 }, { "epoch": 5.7, "learning_rate": 6.082962299063258e-05, "loss": 1.485, "step": 65850 }, { "epoch": 5.7, "learning_rate": 6.082745460853475e-05, "loss": 1.597, "step": 65860 }, { "epoch": 5.7, "learning_rate": 6.082528622643691e-05, "loss": 1.5402, "step": 65870 }, { "epoch": 5.71, "learning_rate": 6.0823117844339074e-05, "loss": 1.622, "step": 65880 }, { "epoch": 5.71, "learning_rate": 6.0820949462241234e-05, "loss": 1.5288, "step": 65890 }, { "epoch": 5.71, "learning_rate": 6.08187810801434e-05, "loss": 1.5283, "step": 65900 }, { "epoch": 5.71, "learning_rate": 6.081661269804557e-05, "loss": 1.5859, "step": 65910 }, { "epoch": 5.71, "learning_rate": 6.081444431594772e-05, "loss": 1.5841, "step": 65920 }, { "epoch": 5.71, "learning_rate": 6.0812275933849886e-05, "loss": 1.7295, "step": 65930 }, { "epoch": 5.71, "learning_rate": 6.0810107551752046e-05, "loss": 1.7036, "step": 65940 }, { "epoch": 5.71, "learning_rate": 6.080793916965421e-05, "loss": 1.5494, "step": 65950 }, { "epoch": 5.71, "learning_rate": 6.080577078755637e-05, "loss": 1.6808, "step": 65960 }, { "epoch": 5.71, "learning_rate": 6.080360240545854e-05, "loss": 1.6693, "step": 65970 }, { "epoch": 5.71, "learning_rate": 6.0801434023360704e-05, "loss": 1.6931, "step": 65980 }, { "epoch": 5.72, "learning_rate": 6.079926564126286e-05, "loss": 1.7655, "step": 65990 }, { "epoch": 5.72, "learning_rate": 6.079709725916502e-05, "loss": 1.4742, "step": 66000 }, { "epoch": 5.72, "learning_rate": 6.079492887706718e-05, "loss": 1.2588, "step": 66010 }, { "epoch": 5.72, "learning_rate": 6.079276049496935e-05, "loss": 1.7426, "step": 66020 }, { "epoch": 5.72, "learning_rate": 6.079059211287151e-05, "loss": 1.4066, "step": 66030 }, { "epoch": 5.72, "learning_rate": 6.0788423730773676e-05, "loss": 1.8555, "step": 66040 }, { "epoch": 5.72, "learning_rate": 6.0786255348675835e-05, "loss": 1.4469, "step": 66050 }, { "epoch": 5.72, "learning_rate": 6.0784086966578e-05, "loss": 1.6919, "step": 66060 }, { "epoch": 5.72, "learning_rate": 6.0781918584480155e-05, "loss": 1.6949, "step": 66070 }, { "epoch": 5.72, "learning_rate": 6.077975020238232e-05, "loss": 1.7768, "step": 66080 }, { "epoch": 5.72, "learning_rate": 6.077758182028449e-05, "loss": 1.6547, "step": 66090 }, { "epoch": 5.72, "learning_rate": 6.077541343818665e-05, "loss": 1.5791, "step": 66100 }, { "epoch": 5.73, "learning_rate": 6.077324505608881e-05, "loss": 1.6388, "step": 66110 }, { "epoch": 5.73, "learning_rate": 6.077107667399097e-05, "loss": 1.3962, "step": 66120 }, { "epoch": 5.73, "learning_rate": 6.076890829189314e-05, "loss": 1.6064, "step": 66130 }, { "epoch": 5.73, "learning_rate": 6.0766739909795306e-05, "loss": 1.3783, "step": 66140 }, { "epoch": 5.73, "learning_rate": 6.076457152769746e-05, "loss": 1.4119, "step": 66150 }, { "epoch": 5.73, "learning_rate": 6.0762403145599625e-05, "loss": 1.4969, "step": 66160 }, { "epoch": 5.73, "learning_rate": 6.0760234763501785e-05, "loss": 1.6541, "step": 66170 }, { "epoch": 5.73, "learning_rate": 6.075806638140395e-05, "loss": 1.5332, "step": 66180 }, { "epoch": 5.73, "learning_rate": 6.075589799930611e-05, "loss": 1.4671, "step": 66190 }, { "epoch": 5.73, "learning_rate": 6.075372961720828e-05, "loss": 1.5276, "step": 66200 }, { "epoch": 5.73, "learning_rate": 6.0751561235110443e-05, "loss": 1.5932, "step": 66210 }, { "epoch": 5.74, "learning_rate": 6.0749392853012596e-05, "loss": 1.6226, "step": 66220 }, { "epoch": 5.74, "learning_rate": 6.074722447091476e-05, "loss": 1.636, "step": 66230 }, { "epoch": 5.74, "learning_rate": 6.074505608881692e-05, "loss": 1.5896, "step": 66240 }, { "epoch": 5.74, "learning_rate": 6.074288770671909e-05, "loss": 1.9094, "step": 66250 }, { "epoch": 5.74, "learning_rate": 6.074071932462125e-05, "loss": 1.6131, "step": 66260 }, { "epoch": 5.74, "learning_rate": 6.0738550942523415e-05, "loss": 1.6707, "step": 66270 }, { "epoch": 5.74, "learning_rate": 6.073638256042558e-05, "loss": 1.375, "step": 66280 }, { "epoch": 5.74, "learning_rate": 6.073421417832774e-05, "loss": 1.4064, "step": 66290 }, { "epoch": 5.74, "learning_rate": 6.07320457962299e-05, "loss": 1.613, "step": 66300 }, { "epoch": 5.74, "learning_rate": 6.072987741413206e-05, "loss": 1.8267, "step": 66310 }, { "epoch": 5.74, "learning_rate": 6.0727709032034226e-05, "loss": 1.6514, "step": 66320 }, { "epoch": 5.74, "learning_rate": 6.0725540649936386e-05, "loss": 1.65, "step": 66330 }, { "epoch": 5.75, "learning_rate": 6.072337226783855e-05, "loss": 1.6079, "step": 66340 }, { "epoch": 5.75, "learning_rate": 6.072120388574072e-05, "loss": 1.4847, "step": 66350 }, { "epoch": 5.75, "learning_rate": 6.071903550364288e-05, "loss": 1.8112, "step": 66360 }, { "epoch": 5.75, "learning_rate": 6.0716867121545045e-05, "loss": 1.43, "step": 66370 }, { "epoch": 5.75, "learning_rate": 6.07146987394472e-05, "loss": 1.6438, "step": 66380 }, { "epoch": 5.75, "learning_rate": 6.0712530357349364e-05, "loss": 1.4453, "step": 66390 }, { "epoch": 5.75, "learning_rate": 6.0710361975251524e-05, "loss": 1.6523, "step": 66400 }, { "epoch": 5.75, "learning_rate": 6.070819359315369e-05, "loss": 1.3528, "step": 66410 }, { "epoch": 5.75, "learning_rate": 6.0706025211055856e-05, "loss": 1.6756, "step": 66420 }, { "epoch": 5.75, "learning_rate": 6.0703856828958016e-05, "loss": 1.4747, "step": 66430 }, { "epoch": 5.75, "learning_rate": 6.070168844686018e-05, "loss": 1.4772, "step": 66440 }, { "epoch": 5.75, "learning_rate": 6.0699520064762335e-05, "loss": 1.7257, "step": 66450 }, { "epoch": 5.76, "learning_rate": 6.06973516826645e-05, "loss": 1.6044, "step": 66460 }, { "epoch": 5.76, "learning_rate": 6.069518330056666e-05, "loss": 1.5962, "step": 66470 }, { "epoch": 5.76, "learning_rate": 6.069301491846883e-05, "loss": 1.2757, "step": 66480 }, { "epoch": 5.76, "learning_rate": 6.0690846536370994e-05, "loss": 1.5533, "step": 66490 }, { "epoch": 5.76, "learning_rate": 6.0688678154273154e-05, "loss": 1.4415, "step": 66500 }, { "epoch": 5.76, "learning_rate": 6.068650977217532e-05, "loss": 1.7271, "step": 66510 }, { "epoch": 5.76, "learning_rate": 6.068434139007748e-05, "loss": 1.4548, "step": 66520 }, { "epoch": 5.76, "learning_rate": 6.068217300797964e-05, "loss": 1.5497, "step": 66530 }, { "epoch": 5.76, "learning_rate": 6.06800046258818e-05, "loss": 1.4889, "step": 66540 }, { "epoch": 5.76, "learning_rate": 6.0677836243783965e-05, "loss": 1.5825, "step": 66550 }, { "epoch": 5.76, "learning_rate": 6.0675667861686125e-05, "loss": 1.5501, "step": 66560 }, { "epoch": 5.77, "learning_rate": 6.067349947958829e-05, "loss": 1.5054, "step": 66570 }, { "epoch": 5.77, "learning_rate": 6.067133109749046e-05, "loss": 1.4933, "step": 66580 }, { "epoch": 5.77, "learning_rate": 6.066916271539262e-05, "loss": 1.7749, "step": 66590 }, { "epoch": 5.77, "learning_rate": 6.0666994333294784e-05, "loss": 1.4555, "step": 66600 }, { "epoch": 5.77, "learning_rate": 6.0664825951196936e-05, "loss": 1.5799, "step": 66610 }, { "epoch": 5.77, "learning_rate": 6.06626575690991e-05, "loss": 1.6668, "step": 66620 }, { "epoch": 5.77, "learning_rate": 6.066048918700126e-05, "loss": 1.6474, "step": 66630 }, { "epoch": 5.77, "learning_rate": 6.065832080490343e-05, "loss": 1.7071, "step": 66640 }, { "epoch": 5.77, "learning_rate": 6.0656152422805595e-05, "loss": 1.8022, "step": 66650 }, { "epoch": 5.77, "learning_rate": 6.0653984040707755e-05, "loss": 1.4157, "step": 66660 }, { "epoch": 5.77, "learning_rate": 6.065181565860992e-05, "loss": 1.5773, "step": 66670 }, { "epoch": 5.77, "learning_rate": 6.064964727651208e-05, "loss": 1.6449, "step": 66680 }, { "epoch": 5.78, "learning_rate": 6.064747889441424e-05, "loss": 1.5058, "step": 66690 }, { "epoch": 5.78, "learning_rate": 6.06453105123164e-05, "loss": 1.6709, "step": 66700 }, { "epoch": 5.78, "learning_rate": 6.0643142130218566e-05, "loss": 1.6105, "step": 66710 }, { "epoch": 5.78, "learning_rate": 6.064097374812073e-05, "loss": 1.5297, "step": 66720 }, { "epoch": 5.78, "learning_rate": 6.063880536602289e-05, "loss": 1.4956, "step": 66730 }, { "epoch": 5.78, "learning_rate": 6.063663698392506e-05, "loss": 1.7773, "step": 66740 }, { "epoch": 5.78, "learning_rate": 6.063446860182722e-05, "loss": 1.4289, "step": 66750 }, { "epoch": 5.78, "learning_rate": 6.063230021972938e-05, "loss": 1.6121, "step": 66760 }, { "epoch": 5.78, "learning_rate": 6.063013183763154e-05, "loss": 1.6052, "step": 66770 }, { "epoch": 5.78, "learning_rate": 6.0627963455533704e-05, "loss": 1.4691, "step": 66780 }, { "epoch": 5.78, "learning_rate": 6.062579507343587e-05, "loss": 1.4741, "step": 66790 }, { "epoch": 5.79, "learning_rate": 6.062362669133803e-05, "loss": 1.5807, "step": 66800 }, { "epoch": 5.79, "learning_rate": 6.0621458309240197e-05, "loss": 1.6124, "step": 66810 }, { "epoch": 5.79, "learning_rate": 6.0619289927142356e-05, "loss": 1.5276, "step": 66820 }, { "epoch": 5.79, "learning_rate": 6.061712154504452e-05, "loss": 1.4709, "step": 66830 }, { "epoch": 5.79, "learning_rate": 6.0614953162946675e-05, "loss": 1.5179, "step": 66840 }, { "epoch": 5.79, "learning_rate": 6.061278478084884e-05, "loss": 1.4187, "step": 66850 }, { "epoch": 5.79, "learning_rate": 6.061061639875101e-05, "loss": 1.6884, "step": 66860 }, { "epoch": 5.79, "learning_rate": 6.060844801665317e-05, "loss": 1.6589, "step": 66870 }, { "epoch": 5.79, "learning_rate": 6.0606279634555334e-05, "loss": 1.7559, "step": 66880 }, { "epoch": 5.79, "learning_rate": 6.0604111252457494e-05, "loss": 1.6367, "step": 66890 }, { "epoch": 5.79, "learning_rate": 6.060194287035966e-05, "loss": 1.5332, "step": 66900 }, { "epoch": 5.79, "learning_rate": 6.0599774488261827e-05, "loss": 1.5358, "step": 66910 }, { "epoch": 5.8, "learning_rate": 6.059760610616398e-05, "loss": 1.3819, "step": 66920 }, { "epoch": 5.8, "learning_rate": 6.0595437724066146e-05, "loss": 1.5846, "step": 66930 }, { "epoch": 5.8, "learning_rate": 6.0593269341968305e-05, "loss": 1.6653, "step": 66940 }, { "epoch": 5.8, "learning_rate": 6.059110095987047e-05, "loss": 1.5548, "step": 66950 }, { "epoch": 5.8, "learning_rate": 6.058893257777263e-05, "loss": 1.6473, "step": 66960 }, { "epoch": 5.8, "learning_rate": 6.05867641956748e-05, "loss": 1.483, "step": 66970 }, { "epoch": 5.8, "learning_rate": 6.0584595813576964e-05, "loss": 1.6293, "step": 66980 }, { "epoch": 5.8, "learning_rate": 6.058242743147912e-05, "loss": 1.6264, "step": 66990 }, { "epoch": 5.8, "learning_rate": 6.0580259049381283e-05, "loss": 1.4402, "step": 67000 }, { "epoch": 5.8, "learning_rate": 6.057809066728344e-05, "loss": 1.3376, "step": 67010 }, { "epoch": 5.8, "learning_rate": 6.057592228518561e-05, "loss": 1.6231, "step": 67020 }, { "epoch": 5.81, "learning_rate": 6.057375390308777e-05, "loss": 1.5918, "step": 67030 }, { "epoch": 5.81, "learning_rate": 6.0571585520989935e-05, "loss": 1.413, "step": 67040 }, { "epoch": 5.81, "learning_rate": 6.05694171388921e-05, "loss": 1.3464, "step": 67050 }, { "epoch": 5.81, "learning_rate": 6.056724875679426e-05, "loss": 1.4923, "step": 67060 }, { "epoch": 5.81, "learning_rate": 6.0565080374696414e-05, "loss": 1.7473, "step": 67070 }, { "epoch": 5.81, "learning_rate": 6.056291199259858e-05, "loss": 1.4447, "step": 67080 }, { "epoch": 5.81, "learning_rate": 6.056074361050075e-05, "loss": 1.4465, "step": 67090 }, { "epoch": 5.81, "learning_rate": 6.055857522840291e-05, "loss": 1.5981, "step": 67100 }, { "epoch": 5.81, "learning_rate": 6.055640684630507e-05, "loss": 1.4621, "step": 67110 }, { "epoch": 5.81, "learning_rate": 6.055423846420723e-05, "loss": 1.3783, "step": 67120 }, { "epoch": 5.81, "learning_rate": 6.05520700821094e-05, "loss": 1.6088, "step": 67130 }, { "epoch": 5.81, "learning_rate": 6.0549901700011565e-05, "loss": 1.7093, "step": 67140 }, { "epoch": 5.82, "learning_rate": 6.054773331791372e-05, "loss": 1.6337, "step": 67150 }, { "epoch": 5.82, "learning_rate": 6.0545564935815885e-05, "loss": 1.4195, "step": 67160 }, { "epoch": 5.82, "learning_rate": 6.0543396553718044e-05, "loss": 1.6545, "step": 67170 }, { "epoch": 5.82, "learning_rate": 6.054122817162021e-05, "loss": 1.6103, "step": 67180 }, { "epoch": 5.82, "learning_rate": 6.053905978952237e-05, "loss": 1.3235, "step": 67190 }, { "epoch": 5.82, "learning_rate": 6.053689140742454e-05, "loss": 1.3783, "step": 67200 }, { "epoch": 5.82, "learning_rate": 6.05347230253267e-05, "loss": 1.541, "step": 67210 }, { "epoch": 5.82, "learning_rate": 6.0532554643228856e-05, "loss": 1.62, "step": 67220 }, { "epoch": 5.82, "learning_rate": 6.053038626113102e-05, "loss": 1.5217, "step": 67230 }, { "epoch": 5.82, "learning_rate": 6.052821787903318e-05, "loss": 1.5026, "step": 67240 }, { "epoch": 5.82, "learning_rate": 6.052604949693535e-05, "loss": 1.7597, "step": 67250 }, { "epoch": 5.83, "learning_rate": 6.052388111483751e-05, "loss": 1.7293, "step": 67260 }, { "epoch": 5.83, "learning_rate": 6.0521712732739674e-05, "loss": 1.6654, "step": 67270 }, { "epoch": 5.83, "learning_rate": 6.051954435064184e-05, "loss": 1.4885, "step": 67280 }, { "epoch": 5.83, "learning_rate": 6.0517375968544e-05, "loss": 1.617, "step": 67290 }, { "epoch": 5.83, "learning_rate": 6.051520758644616e-05, "loss": 1.5969, "step": 67300 }, { "epoch": 5.83, "learning_rate": 6.051303920434832e-05, "loss": 1.7547, "step": 67310 }, { "epoch": 5.83, "learning_rate": 6.0510870822250486e-05, "loss": 1.4145, "step": 67320 }, { "epoch": 5.83, "learning_rate": 6.0508702440152646e-05, "loss": 1.4371, "step": 67330 }, { "epoch": 5.83, "learning_rate": 6.050653405805481e-05, "loss": 1.6208, "step": 67340 }, { "epoch": 5.83, "learning_rate": 6.050436567595698e-05, "loss": 1.4595, "step": 67350 }, { "epoch": 5.83, "learning_rate": 6.050219729385914e-05, "loss": 1.6944, "step": 67360 }, { "epoch": 5.83, "learning_rate": 6.0500028911761304e-05, "loss": 1.6377, "step": 67370 }, { "epoch": 5.84, "learning_rate": 6.049786052966346e-05, "loss": 1.294, "step": 67380 }, { "epoch": 5.84, "learning_rate": 6.0495692147565624e-05, "loss": 1.5226, "step": 67390 }, { "epoch": 5.84, "learning_rate": 6.049352376546778e-05, "loss": 1.4103, "step": 67400 }, { "epoch": 5.84, "learning_rate": 6.049135538336995e-05, "loss": 1.5991, "step": 67410 }, { "epoch": 5.84, "learning_rate": 6.0489187001272116e-05, "loss": 1.336, "step": 67420 }, { "epoch": 5.84, "learning_rate": 6.0487018619174276e-05, "loss": 1.6451, "step": 67430 }, { "epoch": 5.84, "learning_rate": 6.048485023707644e-05, "loss": 1.5862, "step": 67440 }, { "epoch": 5.84, "learning_rate": 6.04826818549786e-05, "loss": 1.5745, "step": 67450 }, { "epoch": 5.84, "learning_rate": 6.048051347288076e-05, "loss": 1.4738, "step": 67460 }, { "epoch": 5.84, "learning_rate": 6.047834509078292e-05, "loss": 1.8863, "step": 67470 }, { "epoch": 5.84, "learning_rate": 6.047617670868509e-05, "loss": 1.4529, "step": 67480 }, { "epoch": 5.85, "learning_rate": 6.0474008326587254e-05, "loss": 1.4433, "step": 67490 }, { "epoch": 5.85, "learning_rate": 6.047183994448941e-05, "loss": 1.6164, "step": 67500 }, { "epoch": 5.85, "learning_rate": 6.046967156239158e-05, "loss": 1.7645, "step": 67510 }, { "epoch": 5.85, "learning_rate": 6.046750318029374e-05, "loss": 1.3527, "step": 67520 }, { "epoch": 5.85, "learning_rate": 6.04653347981959e-05, "loss": 1.6697, "step": 67530 }, { "epoch": 5.85, "learning_rate": 6.046316641609806e-05, "loss": 1.7065, "step": 67540 }, { "epoch": 5.85, "learning_rate": 6.0460998034000225e-05, "loss": 1.5393, "step": 67550 }, { "epoch": 5.85, "learning_rate": 6.045882965190239e-05, "loss": 1.5187, "step": 67560 }, { "epoch": 5.85, "learning_rate": 6.045666126980455e-05, "loss": 1.4182, "step": 67570 }, { "epoch": 5.85, "learning_rate": 6.045449288770672e-05, "loss": 1.5589, "step": 67580 }, { "epoch": 5.85, "learning_rate": 6.045232450560888e-05, "loss": 1.5918, "step": 67590 }, { "epoch": 5.85, "learning_rate": 6.045015612351104e-05, "loss": 1.7316, "step": 67600 }, { "epoch": 5.86, "learning_rate": 6.0447987741413196e-05, "loss": 1.7558, "step": 67610 }, { "epoch": 5.86, "learning_rate": 6.044581935931536e-05, "loss": 1.4741, "step": 67620 }, { "epoch": 5.86, "learning_rate": 6.044365097721752e-05, "loss": 1.5479, "step": 67630 }, { "epoch": 5.86, "learning_rate": 6.044148259511969e-05, "loss": 1.7141, "step": 67640 }, { "epoch": 5.86, "learning_rate": 6.0439314213021855e-05, "loss": 1.5925, "step": 67650 }, { "epoch": 5.86, "learning_rate": 6.0437145830924015e-05, "loss": 1.3639, "step": 67660 }, { "epoch": 5.86, "learning_rate": 6.043497744882618e-05, "loss": 1.6292, "step": 67670 }, { "epoch": 5.86, "learning_rate": 6.043280906672834e-05, "loss": 1.5223, "step": 67680 }, { "epoch": 5.86, "learning_rate": 6.04306406846305e-05, "loss": 1.5767, "step": 67690 }, { "epoch": 5.86, "learning_rate": 6.042847230253266e-05, "loss": 1.5597, "step": 67700 }, { "epoch": 5.86, "learning_rate": 6.0426303920434826e-05, "loss": 1.5234, "step": 67710 }, { "epoch": 5.86, "learning_rate": 6.042413553833699e-05, "loss": 1.5991, "step": 67720 }, { "epoch": 5.87, "learning_rate": 6.042196715623915e-05, "loss": 1.6999, "step": 67730 }, { "epoch": 5.87, "learning_rate": 6.041979877414132e-05, "loss": 1.4678, "step": 67740 }, { "epoch": 5.87, "learning_rate": 6.041763039204348e-05, "loss": 1.5549, "step": 67750 }, { "epoch": 5.87, "learning_rate": 6.041546200994564e-05, "loss": 1.7022, "step": 67760 }, { "epoch": 5.87, "learning_rate": 6.04132936278478e-05, "loss": 1.6955, "step": 67770 }, { "epoch": 5.87, "learning_rate": 6.0411125245749964e-05, "loss": 1.4829, "step": 67780 }, { "epoch": 5.87, "learning_rate": 6.040895686365213e-05, "loss": 1.5363, "step": 67790 }, { "epoch": 5.87, "learning_rate": 6.040678848155429e-05, "loss": 1.4791, "step": 67800 }, { "epoch": 5.87, "learning_rate": 6.0404620099456456e-05, "loss": 1.6264, "step": 67810 }, { "epoch": 5.87, "learning_rate": 6.0402451717358616e-05, "loss": 1.6494, "step": 67820 }, { "epoch": 5.87, "learning_rate": 6.040028333526078e-05, "loss": 1.5784, "step": 67830 }, { "epoch": 5.88, "learning_rate": 6.0398114953162935e-05, "loss": 1.7004, "step": 67840 }, { "epoch": 5.88, "learning_rate": 6.03959465710651e-05, "loss": 1.6289, "step": 67850 }, { "epoch": 5.88, "learning_rate": 6.039377818896727e-05, "loss": 1.5106, "step": 67860 }, { "epoch": 5.88, "learning_rate": 6.039160980686943e-05, "loss": 1.7566, "step": 67870 }, { "epoch": 5.88, "learning_rate": 6.0389441424771594e-05, "loss": 1.8051, "step": 67880 }, { "epoch": 5.88, "learning_rate": 6.0387273042673753e-05, "loss": 1.5865, "step": 67890 }, { "epoch": 5.88, "learning_rate": 6.038510466057592e-05, "loss": 1.7857, "step": 67900 }, { "epoch": 5.88, "learning_rate": 6.0382936278478086e-05, "loss": 1.3926, "step": 67910 }, { "epoch": 5.88, "learning_rate": 6.038076789638024e-05, "loss": 1.5699, "step": 67920 }, { "epoch": 5.88, "learning_rate": 6.0378599514282405e-05, "loss": 1.5459, "step": 67930 }, { "epoch": 5.88, "learning_rate": 6.0376431132184565e-05, "loss": 1.6606, "step": 67940 }, { "epoch": 5.88, "learning_rate": 6.037426275008673e-05, "loss": 1.4744, "step": 67950 }, { "epoch": 5.89, "learning_rate": 6.037209436798889e-05, "loss": 1.5102, "step": 67960 }, { "epoch": 5.89, "learning_rate": 6.036992598589106e-05, "loss": 1.5775, "step": 67970 }, { "epoch": 5.89, "learning_rate": 6.0367757603793224e-05, "loss": 1.5063, "step": 67980 }, { "epoch": 5.89, "learning_rate": 6.036558922169538e-05, "loss": 1.3956, "step": 67990 }, { "epoch": 5.89, "learning_rate": 6.036342083959754e-05, "loss": 1.3744, "step": 68000 }, { "epoch": 5.89, "learning_rate": 6.03612524574997e-05, "loss": 1.3993, "step": 68010 }, { "epoch": 5.89, "learning_rate": 6.035908407540187e-05, "loss": 1.6885, "step": 68020 }, { "epoch": 5.89, "learning_rate": 6.035691569330403e-05, "loss": 1.5924, "step": 68030 }, { "epoch": 5.89, "learning_rate": 6.0354747311206195e-05, "loss": 1.5498, "step": 68040 }, { "epoch": 5.89, "learning_rate": 6.035257892910836e-05, "loss": 1.4996, "step": 68050 }, { "epoch": 5.89, "learning_rate": 6.035041054701052e-05, "loss": 1.6756, "step": 68060 }, { "epoch": 5.9, "learning_rate": 6.034824216491268e-05, "loss": 1.7305, "step": 68070 }, { "epoch": 5.9, "learning_rate": 6.034607378281484e-05, "loss": 1.4085, "step": 68080 }, { "epoch": 5.9, "learning_rate": 6.034390540071701e-05, "loss": 1.7607, "step": 68090 }, { "epoch": 5.9, "learning_rate": 6.0341737018619166e-05, "loss": 1.5961, "step": 68100 }, { "epoch": 5.9, "learning_rate": 6.033956863652133e-05, "loss": 1.6377, "step": 68110 }, { "epoch": 5.9, "learning_rate": 6.03374002544235e-05, "loss": 1.5806, "step": 68120 }, { "epoch": 5.9, "learning_rate": 6.033523187232566e-05, "loss": 1.4589, "step": 68130 }, { "epoch": 5.9, "learning_rate": 6.0333063490227825e-05, "loss": 1.4965, "step": 68140 }, { "epoch": 5.9, "learning_rate": 6.033089510812998e-05, "loss": 1.4359, "step": 68150 }, { "epoch": 5.9, "learning_rate": 6.0328726726032144e-05, "loss": 1.5603, "step": 68160 }, { "epoch": 5.9, "learning_rate": 6.0326558343934304e-05, "loss": 1.4402, "step": 68170 }, { "epoch": 5.9, "learning_rate": 6.032438996183647e-05, "loss": 1.5072, "step": 68180 }, { "epoch": 5.91, "learning_rate": 6.032222157973863e-05, "loss": 1.5718, "step": 68190 }, { "epoch": 5.91, "learning_rate": 6.0320053197640796e-05, "loss": 1.6533, "step": 68200 }, { "epoch": 5.91, "learning_rate": 6.031788481554296e-05, "loss": 1.5019, "step": 68210 }, { "epoch": 5.91, "learning_rate": 6.031571643344512e-05, "loss": 1.6604, "step": 68220 }, { "epoch": 5.91, "learning_rate": 6.031354805134728e-05, "loss": 1.722, "step": 68230 }, { "epoch": 5.91, "learning_rate": 6.031137966924944e-05, "loss": 1.45, "step": 68240 }, { "epoch": 5.91, "learning_rate": 6.030921128715161e-05, "loss": 1.8395, "step": 68250 }, { "epoch": 5.91, "learning_rate": 6.030704290505377e-05, "loss": 1.4169, "step": 68260 }, { "epoch": 5.91, "learning_rate": 6.0304874522955934e-05, "loss": 1.6933, "step": 68270 }, { "epoch": 5.91, "learning_rate": 6.03027061408581e-05, "loss": 1.601, "step": 68280 }, { "epoch": 5.91, "learning_rate": 6.030053775876026e-05, "loss": 1.5397, "step": 68290 }, { "epoch": 5.92, "learning_rate": 6.029836937666242e-05, "loss": 1.7369, "step": 68300 }, { "epoch": 5.92, "learning_rate": 6.029620099456458e-05, "loss": 1.448, "step": 68310 }, { "epoch": 5.92, "learning_rate": 6.0294032612466746e-05, "loss": 1.3759, "step": 68320 }, { "epoch": 5.92, "learning_rate": 6.0291864230368905e-05, "loss": 1.5982, "step": 68330 }, { "epoch": 5.92, "learning_rate": 6.028969584827107e-05, "loss": 1.506, "step": 68340 }, { "epoch": 5.92, "learning_rate": 6.028752746617324e-05, "loss": 1.5879, "step": 68350 }, { "epoch": 5.92, "learning_rate": 6.02853590840754e-05, "loss": 1.6346, "step": 68360 }, { "epoch": 5.92, "learning_rate": 6.0283190701977564e-05, "loss": 1.6196, "step": 68370 }, { "epoch": 5.92, "learning_rate": 6.028102231987972e-05, "loss": 1.7554, "step": 68380 }, { "epoch": 5.92, "learning_rate": 6.027885393778188e-05, "loss": 1.5283, "step": 68390 }, { "epoch": 5.92, "learning_rate": 6.027668555568404e-05, "loss": 1.6593, "step": 68400 }, { "epoch": 5.92, "learning_rate": 6.027451717358621e-05, "loss": 1.8614, "step": 68410 }, { "epoch": 5.93, "learning_rate": 6.0272348791488376e-05, "loss": 1.4363, "step": 68420 }, { "epoch": 5.93, "learning_rate": 6.0270180409390535e-05, "loss": 1.763, "step": 68430 }, { "epoch": 5.93, "learning_rate": 6.02680120272927e-05, "loss": 1.3886, "step": 68440 }, { "epoch": 5.93, "learning_rate": 6.026584364519486e-05, "loss": 1.717, "step": 68450 }, { "epoch": 5.93, "learning_rate": 6.026367526309702e-05, "loss": 1.7901, "step": 68460 }, { "epoch": 5.93, "learning_rate": 6.026150688099918e-05, "loss": 1.6941, "step": 68470 }, { "epoch": 5.93, "learning_rate": 6.025933849890135e-05, "loss": 1.6121, "step": 68480 }, { "epoch": 5.93, "learning_rate": 6.025717011680351e-05, "loss": 1.6978, "step": 68490 }, { "epoch": 5.93, "learning_rate": 6.025500173470567e-05, "loss": 1.5737, "step": 68500 }, { "epoch": 5.93, "learning_rate": 6.025283335260784e-05, "loss": 1.6077, "step": 68510 }, { "epoch": 5.93, "learning_rate": 6.025066497051e-05, "loss": 1.5711, "step": 68520 }, { "epoch": 5.94, "learning_rate": 6.024849658841216e-05, "loss": 1.404, "step": 68530 }, { "epoch": 5.94, "learning_rate": 6.024632820631432e-05, "loss": 1.3616, "step": 68540 }, { "epoch": 5.94, "learning_rate": 6.0244159824216485e-05, "loss": 1.7227, "step": 68550 }, { "epoch": 5.94, "learning_rate": 6.024199144211865e-05, "loss": 1.4908, "step": 68560 }, { "epoch": 5.94, "learning_rate": 6.023982306002081e-05, "loss": 1.4723, "step": 68570 }, { "epoch": 5.94, "learning_rate": 6.023765467792298e-05, "loss": 1.4739, "step": 68580 }, { "epoch": 5.94, "learning_rate": 6.0235486295825137e-05, "loss": 1.6227, "step": 68590 }, { "epoch": 5.94, "learning_rate": 6.02333179137273e-05, "loss": 1.6636, "step": 68600 }, { "epoch": 5.94, "learning_rate": 6.0231149531629456e-05, "loss": 1.6465, "step": 68610 }, { "epoch": 5.94, "learning_rate": 6.022898114953162e-05, "loss": 1.5766, "step": 68620 }, { "epoch": 5.94, "learning_rate": 6.022681276743379e-05, "loss": 1.6597, "step": 68630 }, { "epoch": 5.94, "learning_rate": 6.022464438533595e-05, "loss": 1.7339, "step": 68640 }, { "epoch": 5.95, "learning_rate": 6.0222476003238115e-05, "loss": 1.6637, "step": 68650 }, { "epoch": 5.95, "learning_rate": 6.0220307621140274e-05, "loss": 1.7354, "step": 68660 }, { "epoch": 5.95, "learning_rate": 6.021813923904244e-05, "loss": 1.5892, "step": 68670 }, { "epoch": 5.95, "learning_rate": 6.021597085694461e-05, "loss": 1.5609, "step": 68680 }, { "epoch": 5.95, "learning_rate": 6.021380247484676e-05, "loss": 1.452, "step": 68690 }, { "epoch": 5.95, "learning_rate": 6.021163409274892e-05, "loss": 1.6817, "step": 68700 }, { "epoch": 5.95, "learning_rate": 6.0209465710651086e-05, "loss": 1.7249, "step": 68710 }, { "epoch": 5.95, "learning_rate": 6.020729732855325e-05, "loss": 1.5749, "step": 68720 }, { "epoch": 5.95, "learning_rate": 6.020512894645541e-05, "loss": 1.6989, "step": 68730 }, { "epoch": 5.95, "learning_rate": 6.020296056435758e-05, "loss": 1.6408, "step": 68740 }, { "epoch": 5.95, "learning_rate": 6.020079218225974e-05, "loss": 1.5612, "step": 68750 }, { "epoch": 5.96, "learning_rate": 6.01986238001619e-05, "loss": 1.5509, "step": 68760 }, { "epoch": 5.96, "learning_rate": 6.019645541806406e-05, "loss": 1.4905, "step": 68770 }, { "epoch": 5.96, "learning_rate": 6.0194287035966224e-05, "loss": 1.4207, "step": 68780 }, { "epoch": 5.96, "learning_rate": 6.019211865386839e-05, "loss": 1.5123, "step": 68790 }, { "epoch": 5.96, "learning_rate": 6.018995027177055e-05, "loss": 1.6493, "step": 68800 }, { "epoch": 5.96, "learning_rate": 6.0187781889672716e-05, "loss": 1.5635, "step": 68810 }, { "epoch": 5.96, "learning_rate": 6.0185613507574876e-05, "loss": 1.6855, "step": 68820 }, { "epoch": 5.96, "learning_rate": 6.018344512547704e-05, "loss": 1.7468, "step": 68830 }, { "epoch": 5.96, "learning_rate": 6.0181276743379195e-05, "loss": 1.7756, "step": 68840 }, { "epoch": 5.96, "learning_rate": 6.017910836128136e-05, "loss": 1.5334, "step": 68850 }, { "epoch": 5.96, "learning_rate": 6.017693997918353e-05, "loss": 1.7669, "step": 68860 }, { "epoch": 5.96, "learning_rate": 6.017477159708569e-05, "loss": 1.419, "step": 68870 }, { "epoch": 5.97, "learning_rate": 6.0172603214987854e-05, "loss": 1.4443, "step": 68880 }, { "epoch": 5.97, "learning_rate": 6.017043483289001e-05, "loss": 1.5899, "step": 68890 }, { "epoch": 5.97, "learning_rate": 6.016826645079218e-05, "loss": 1.6755, "step": 68900 }, { "epoch": 5.97, "learning_rate": 6.0166098068694346e-05, "loss": 1.6682, "step": 68910 }, { "epoch": 5.97, "learning_rate": 6.01639296865965e-05, "loss": 1.4525, "step": 68920 }, { "epoch": 5.97, "learning_rate": 6.0161761304498665e-05, "loss": 1.8091, "step": 68930 }, { "epoch": 5.97, "learning_rate": 6.0159592922400825e-05, "loss": 1.3994, "step": 68940 }, { "epoch": 5.97, "learning_rate": 6.015742454030299e-05, "loss": 1.5973, "step": 68950 }, { "epoch": 5.97, "learning_rate": 6.015525615820515e-05, "loss": 1.6184, "step": 68960 }, { "epoch": 5.97, "learning_rate": 6.015308777610732e-05, "loss": 1.6055, "step": 68970 }, { "epoch": 5.97, "learning_rate": 6.0150919394009484e-05, "loss": 1.6793, "step": 68980 }, { "epoch": 5.97, "learning_rate": 6.014875101191164e-05, "loss": 1.4904, "step": 68990 }, { "epoch": 5.98, "learning_rate": 6.01465826298138e-05, "loss": 1.6618, "step": 69000 }, { "epoch": 5.98, "learning_rate": 6.014441424771596e-05, "loss": 1.6948, "step": 69010 }, { "epoch": 5.98, "learning_rate": 6.014224586561813e-05, "loss": 1.4846, "step": 69020 }, { "epoch": 5.98, "learning_rate": 6.014007748352029e-05, "loss": 1.5237, "step": 69030 }, { "epoch": 5.98, "learning_rate": 6.0137909101422455e-05, "loss": 1.4241, "step": 69040 }, { "epoch": 5.98, "learning_rate": 6.013574071932462e-05, "loss": 1.6533, "step": 69050 }, { "epoch": 5.98, "learning_rate": 6.013357233722678e-05, "loss": 1.5604, "step": 69060 }, { "epoch": 5.98, "learning_rate": 6.013140395512894e-05, "loss": 1.6868, "step": 69070 }, { "epoch": 5.98, "learning_rate": 6.01292355730311e-05, "loss": 1.6454, "step": 69080 }, { "epoch": 5.98, "learning_rate": 6.0127067190933266e-05, "loss": 1.3643, "step": 69090 }, { "epoch": 5.98, "learning_rate": 6.0124898808835426e-05, "loss": 1.6682, "step": 69100 }, { "epoch": 5.99, "learning_rate": 6.012273042673759e-05, "loss": 1.6023, "step": 69110 }, { "epoch": 5.99, "learning_rate": 6.012056204463976e-05, "loss": 1.8281, "step": 69120 }, { "epoch": 5.99, "learning_rate": 6.011839366254192e-05, "loss": 1.4638, "step": 69130 }, { "epoch": 5.99, "learning_rate": 6.0116225280444085e-05, "loss": 1.5742, "step": 69140 }, { "epoch": 5.99, "learning_rate": 6.011405689834624e-05, "loss": 1.608, "step": 69150 }, { "epoch": 5.99, "learning_rate": 6.0111888516248404e-05, "loss": 1.6149, "step": 69160 }, { "epoch": 5.99, "learning_rate": 6.0109720134150564e-05, "loss": 1.3575, "step": 69170 }, { "epoch": 5.99, "learning_rate": 6.010755175205273e-05, "loss": 1.4152, "step": 69180 }, { "epoch": 5.99, "learning_rate": 6.0105383369954896e-05, "loss": 1.6008, "step": 69190 }, { "epoch": 5.99, "learning_rate": 6.0103214987857056e-05, "loss": 1.4631, "step": 69200 }, { "epoch": 5.99, "learning_rate": 6.010104660575922e-05, "loss": 1.7433, "step": 69210 }, { "epoch": 5.99, "learning_rate": 6.009887822366138e-05, "loss": 1.4947, "step": 69220 }, { "epoch": 6.0, "learning_rate": 6.009670984156354e-05, "loss": 1.5421, "step": 69230 }, { "epoch": 6.0, "learning_rate": 6.00945414594657e-05, "loss": 1.7619, "step": 69240 }, { "epoch": 6.0, "learning_rate": 6.009237307736787e-05, "loss": 1.2871, "step": 69250 }, { "epoch": 6.0, "learning_rate": 6.009020469527003e-05, "loss": 1.5119, "step": 69260 }, { "epoch": 6.0, "learning_rate": 6.0088036313172194e-05, "loss": 1.6863, "step": 69270 }, { "epoch": 6.0, "eval_cer": 0.5496053786752191, "eval_em": 0.014956162970603403, "eval_f1": 0.014956162970603403, "eval_loss": 1.558223843574524, "eval_runtime": 255.1261, "eval_samples_per_second": 7.6, "eval_steps_per_second": 0.952, "eval_wer": 0.9850438370293966, "step": 69279 }, { "epoch": 6.0, "learning_rate": 6.008586793107436e-05, "loss": 1.6505, "step": 69280 }, { "epoch": 6.0, "learning_rate": 6.008369954897652e-05, "loss": 1.4626, "step": 69290 }, { "epoch": 6.0, "learning_rate": 6.008153116687868e-05, "loss": 1.3916, "step": 69300 }, { "epoch": 6.0, "learning_rate": 6.007936278478084e-05, "loss": 1.4609, "step": 69310 }, { "epoch": 6.0, "learning_rate": 6.0077194402683005e-05, "loss": 1.7264, "step": 69320 }, { "epoch": 6.0, "learning_rate": 6.0075026020585165e-05, "loss": 1.4855, "step": 69330 }, { "epoch": 6.01, "learning_rate": 6.007285763848733e-05, "loss": 1.5386, "step": 69340 }, { "epoch": 6.01, "learning_rate": 6.00706892563895e-05, "loss": 1.5573, "step": 69350 }, { "epoch": 6.01, "learning_rate": 6.006852087429166e-05, "loss": 1.5147, "step": 69360 }, { "epoch": 6.01, "learning_rate": 6.0066352492193824e-05, "loss": 1.6717, "step": 69370 }, { "epoch": 6.01, "learning_rate": 6.006418411009598e-05, "loss": 1.5068, "step": 69380 }, { "epoch": 6.01, "learning_rate": 6.006201572799814e-05, "loss": 1.474, "step": 69390 }, { "epoch": 6.01, "learning_rate": 6.00598473459003e-05, "loss": 1.5204, "step": 69400 }, { "epoch": 6.01, "learning_rate": 6.005767896380247e-05, "loss": 1.4498, "step": 69410 }, { "epoch": 6.01, "learning_rate": 6.0055510581704635e-05, "loss": 1.7223, "step": 69420 }, { "epoch": 6.01, "learning_rate": 6.0053342199606795e-05, "loss": 1.4269, "step": 69430 }, { "epoch": 6.01, "learning_rate": 6.005117381750896e-05, "loss": 1.4045, "step": 69440 }, { "epoch": 6.01, "learning_rate": 6.004900543541112e-05, "loss": 1.4959, "step": 69450 }, { "epoch": 6.02, "learning_rate": 6.004683705331328e-05, "loss": 1.5057, "step": 69460 }, { "epoch": 6.02, "learning_rate": 6.004466867121544e-05, "loss": 1.7626, "step": 69470 }, { "epoch": 6.02, "learning_rate": 6.004250028911761e-05, "loss": 1.3457, "step": 69480 }, { "epoch": 6.02, "learning_rate": 6.004033190701977e-05, "loss": 1.5765, "step": 69490 }, { "epoch": 6.02, "learning_rate": 6.003816352492193e-05, "loss": 1.4092, "step": 69500 }, { "epoch": 6.02, "learning_rate": 6.00359951428241e-05, "loss": 1.437, "step": 69510 }, { "epoch": 6.02, "learning_rate": 6.003382676072626e-05, "loss": 1.4731, "step": 69520 }, { "epoch": 6.02, "learning_rate": 6.003165837862842e-05, "loss": 1.4167, "step": 69530 }, { "epoch": 6.02, "learning_rate": 6.002948999653058e-05, "loss": 1.5082, "step": 69540 }, { "epoch": 6.02, "learning_rate": 6.0027321614432744e-05, "loss": 1.4321, "step": 69550 }, { "epoch": 6.02, "learning_rate": 6.002515323233491e-05, "loss": 1.4072, "step": 69560 }, { "epoch": 6.03, "learning_rate": 6.002298485023707e-05, "loss": 1.467, "step": 69570 }, { "epoch": 6.03, "learning_rate": 6.002081646813924e-05, "loss": 1.5296, "step": 69580 }, { "epoch": 6.03, "learning_rate": 6.0018648086041396e-05, "loss": 1.5135, "step": 69590 }, { "epoch": 6.03, "learning_rate": 6.001647970394356e-05, "loss": 1.5194, "step": 69600 }, { "epoch": 6.03, "learning_rate": 6.0014311321845716e-05, "loss": 1.4076, "step": 69610 }, { "epoch": 6.03, "learning_rate": 6.001214293974788e-05, "loss": 1.4321, "step": 69620 }, { "epoch": 6.03, "learning_rate": 6.000997455765005e-05, "loss": 1.4624, "step": 69630 }, { "epoch": 6.03, "learning_rate": 6.000780617555221e-05, "loss": 1.4536, "step": 69640 }, { "epoch": 6.03, "learning_rate": 6.0005637793454374e-05, "loss": 1.4638, "step": 69650 }, { "epoch": 6.03, "learning_rate": 6.0003469411356534e-05, "loss": 1.6022, "step": 69660 }, { "epoch": 6.03, "learning_rate": 6.00013010292587e-05, "loss": 1.459, "step": 69670 }, { "epoch": 6.03, "learning_rate": 5.999913264716087e-05, "loss": 1.405, "step": 69680 }, { "epoch": 6.04, "learning_rate": 5.999696426506302e-05, "loss": 1.6388, "step": 69690 }, { "epoch": 6.04, "learning_rate": 5.9994795882965186e-05, "loss": 1.353, "step": 69700 }, { "epoch": 6.04, "learning_rate": 5.9992627500867346e-05, "loss": 1.4947, "step": 69710 }, { "epoch": 6.04, "learning_rate": 5.999045911876951e-05, "loss": 1.4232, "step": 69720 }, { "epoch": 6.04, "learning_rate": 5.998829073667167e-05, "loss": 1.3975, "step": 69730 }, { "epoch": 6.04, "learning_rate": 5.998612235457384e-05, "loss": 1.4584, "step": 69740 }, { "epoch": 6.04, "learning_rate": 5.9983953972476004e-05, "loss": 1.3696, "step": 69750 }, { "epoch": 6.04, "learning_rate": 5.9981785590378164e-05, "loss": 1.4459, "step": 69760 }, { "epoch": 6.04, "learning_rate": 5.997961720828032e-05, "loss": 1.3554, "step": 69770 }, { "epoch": 6.04, "learning_rate": 5.997744882618248e-05, "loss": 1.4596, "step": 69780 }, { "epoch": 6.04, "learning_rate": 5.997528044408465e-05, "loss": 1.4933, "step": 69790 }, { "epoch": 6.05, "learning_rate": 5.997311206198681e-05, "loss": 1.3789, "step": 69800 }, { "epoch": 6.05, "learning_rate": 5.9970943679888976e-05, "loss": 1.6161, "step": 69810 }, { "epoch": 6.05, "learning_rate": 5.9968775297791135e-05, "loss": 1.4797, "step": 69820 }, { "epoch": 6.05, "learning_rate": 5.99666069156933e-05, "loss": 1.3375, "step": 69830 }, { "epoch": 6.05, "learning_rate": 5.9964438533595454e-05, "loss": 1.5992, "step": 69840 }, { "epoch": 6.05, "learning_rate": 5.996227015149762e-05, "loss": 1.5695, "step": 69850 }, { "epoch": 6.05, "learning_rate": 5.996010176939979e-05, "loss": 1.4787, "step": 69860 }, { "epoch": 6.05, "learning_rate": 5.995793338730195e-05, "loss": 1.5142, "step": 69870 }, { "epoch": 6.05, "learning_rate": 5.995576500520411e-05, "loss": 1.457, "step": 69880 }, { "epoch": 6.05, "learning_rate": 5.995359662310627e-05, "loss": 1.4738, "step": 69890 }, { "epoch": 6.05, "learning_rate": 5.995142824100844e-05, "loss": 1.4425, "step": 69900 }, { "epoch": 6.05, "learning_rate": 5.9949259858910606e-05, "loss": 1.5158, "step": 69910 }, { "epoch": 6.06, "learning_rate": 5.994709147681276e-05, "loss": 1.4817, "step": 69920 }, { "epoch": 6.06, "learning_rate": 5.9944923094714925e-05, "loss": 1.4267, "step": 69930 }, { "epoch": 6.06, "learning_rate": 5.9942754712617084e-05, "loss": 1.4142, "step": 69940 }, { "epoch": 6.06, "learning_rate": 5.994058633051925e-05, "loss": 1.5924, "step": 69950 }, { "epoch": 6.06, "learning_rate": 5.993841794842141e-05, "loss": 1.2467, "step": 69960 }, { "epoch": 6.06, "learning_rate": 5.993624956632358e-05, "loss": 1.2929, "step": 69970 }, { "epoch": 6.06, "learning_rate": 5.993408118422574e-05, "loss": 1.4628, "step": 69980 }, { "epoch": 6.06, "learning_rate": 5.99319128021279e-05, "loss": 1.5855, "step": 69990 }, { "epoch": 6.06, "learning_rate": 5.992974442003006e-05, "loss": 1.5151, "step": 70000 }, { "epoch": 6.06, "learning_rate": 5.992757603793222e-05, "loss": 1.6273, "step": 70010 }, { "epoch": 6.06, "learning_rate": 5.992540765583439e-05, "loss": 1.4529, "step": 70020 }, { "epoch": 6.07, "learning_rate": 5.992323927373655e-05, "loss": 1.411, "step": 70030 }, { "epoch": 6.07, "learning_rate": 5.9921070891638715e-05, "loss": 1.4184, "step": 70040 }, { "epoch": 6.07, "learning_rate": 5.991890250954088e-05, "loss": 1.4501, "step": 70050 }, { "epoch": 6.07, "learning_rate": 5.991673412744304e-05, "loss": 1.4191, "step": 70060 }, { "epoch": 6.07, "learning_rate": 5.99145657453452e-05, "loss": 1.4246, "step": 70070 }, { "epoch": 6.07, "learning_rate": 5.991239736324736e-05, "loss": 1.5325, "step": 70080 }, { "epoch": 6.07, "learning_rate": 5.9910228981149526e-05, "loss": 1.5393, "step": 70090 }, { "epoch": 6.07, "learning_rate": 5.9908060599051686e-05, "loss": 1.5397, "step": 70100 }, { "epoch": 6.07, "learning_rate": 5.990589221695385e-05, "loss": 1.5151, "step": 70110 }, { "epoch": 6.07, "learning_rate": 5.990372383485602e-05, "loss": 1.6692, "step": 70120 }, { "epoch": 6.07, "learning_rate": 5.990155545275818e-05, "loss": 1.4274, "step": 70130 }, { "epoch": 6.07, "learning_rate": 5.9899387070660345e-05, "loss": 1.5279, "step": 70140 }, { "epoch": 6.08, "learning_rate": 5.98972186885625e-05, "loss": 1.5129, "step": 70150 }, { "epoch": 6.08, "learning_rate": 5.9895050306464664e-05, "loss": 1.6567, "step": 70160 }, { "epoch": 6.08, "learning_rate": 5.9892881924366823e-05, "loss": 1.5373, "step": 70170 }, { "epoch": 6.08, "learning_rate": 5.989071354226899e-05, "loss": 1.3349, "step": 70180 }, { "epoch": 6.08, "learning_rate": 5.9888545160171156e-05, "loss": 1.5417, "step": 70190 }, { "epoch": 6.08, "learning_rate": 5.9886376778073316e-05, "loss": 1.4131, "step": 70200 }, { "epoch": 6.08, "learning_rate": 5.988420839597548e-05, "loss": 1.3017, "step": 70210 }, { "epoch": 6.08, "learning_rate": 5.988204001387764e-05, "loss": 1.5572, "step": 70220 }, { "epoch": 6.08, "learning_rate": 5.98798716317798e-05, "loss": 1.3641, "step": 70230 }, { "epoch": 6.08, "learning_rate": 5.987770324968196e-05, "loss": 1.3969, "step": 70240 }, { "epoch": 6.08, "learning_rate": 5.987553486758413e-05, "loss": 1.559, "step": 70250 }, { "epoch": 6.08, "learning_rate": 5.9873366485486294e-05, "loss": 1.6381, "step": 70260 }, { "epoch": 6.09, "learning_rate": 5.9871198103388453e-05, "loss": 1.4214, "step": 70270 }, { "epoch": 6.09, "learning_rate": 5.986902972129062e-05, "loss": 1.6851, "step": 70280 }, { "epoch": 6.09, "learning_rate": 5.986686133919278e-05, "loss": 1.4126, "step": 70290 }, { "epoch": 6.09, "learning_rate": 5.986469295709494e-05, "loss": 1.4859, "step": 70300 }, { "epoch": 6.09, "learning_rate": 5.98625245749971e-05, "loss": 1.3189, "step": 70310 }, { "epoch": 6.09, "learning_rate": 5.9860356192899265e-05, "loss": 1.5095, "step": 70320 }, { "epoch": 6.09, "learning_rate": 5.9858187810801425e-05, "loss": 1.3951, "step": 70330 }, { "epoch": 6.09, "learning_rate": 5.985601942870359e-05, "loss": 1.439, "step": 70340 }, { "epoch": 6.09, "learning_rate": 5.985385104660576e-05, "loss": 1.5511, "step": 70350 }, { "epoch": 6.09, "learning_rate": 5.985168266450792e-05, "loss": 1.5977, "step": 70360 }, { "epoch": 6.09, "learning_rate": 5.9849514282410083e-05, "loss": 1.2975, "step": 70370 }, { "epoch": 6.1, "learning_rate": 5.9847345900312236e-05, "loss": 1.3471, "step": 70380 }, { "epoch": 6.1, "learning_rate": 5.98451775182144e-05, "loss": 1.5506, "step": 70390 }, { "epoch": 6.1, "learning_rate": 5.984300913611656e-05, "loss": 1.4474, "step": 70400 }, { "epoch": 6.1, "learning_rate": 5.984084075401873e-05, "loss": 1.4825, "step": 70410 }, { "epoch": 6.1, "learning_rate": 5.9838672371920895e-05, "loss": 1.3842, "step": 70420 }, { "epoch": 6.1, "learning_rate": 5.9836503989823055e-05, "loss": 1.5232, "step": 70430 }, { "epoch": 6.1, "learning_rate": 5.983433560772522e-05, "loss": 1.4823, "step": 70440 }, { "epoch": 6.1, "learning_rate": 5.983216722562738e-05, "loss": 1.4982, "step": 70450 }, { "epoch": 6.1, "learning_rate": 5.982999884352954e-05, "loss": 1.4583, "step": 70460 }, { "epoch": 6.1, "learning_rate": 5.98278304614317e-05, "loss": 1.4927, "step": 70470 }, { "epoch": 6.1, "learning_rate": 5.9825662079333866e-05, "loss": 1.5447, "step": 70480 }, { "epoch": 6.1, "learning_rate": 5.982349369723603e-05, "loss": 1.5097, "step": 70490 }, { "epoch": 6.11, "learning_rate": 5.982132531513819e-05, "loss": 1.4241, "step": 70500 }, { "epoch": 6.11, "learning_rate": 5.981915693304036e-05, "loss": 1.5144, "step": 70510 }, { "epoch": 6.11, "learning_rate": 5.981698855094252e-05, "loss": 1.3394, "step": 70520 }, { "epoch": 6.11, "learning_rate": 5.9814820168844685e-05, "loss": 1.3445, "step": 70530 }, { "epoch": 6.11, "learning_rate": 5.981265178674684e-05, "loss": 1.5986, "step": 70540 }, { "epoch": 6.11, "learning_rate": 5.9810483404649004e-05, "loss": 1.5828, "step": 70550 }, { "epoch": 6.11, "learning_rate": 5.980831502255117e-05, "loss": 1.6146, "step": 70560 }, { "epoch": 6.11, "learning_rate": 5.980614664045333e-05, "loss": 1.3728, "step": 70570 }, { "epoch": 6.11, "learning_rate": 5.9803978258355496e-05, "loss": 1.2262, "step": 70580 }, { "epoch": 6.11, "learning_rate": 5.9801809876257656e-05, "loss": 1.5733, "step": 70590 }, { "epoch": 6.11, "learning_rate": 5.979964149415982e-05, "loss": 1.4205, "step": 70600 }, { "epoch": 6.12, "learning_rate": 5.9797473112061975e-05, "loss": 1.5324, "step": 70610 }, { "epoch": 6.12, "learning_rate": 5.979530472996414e-05, "loss": 1.4629, "step": 70620 }, { "epoch": 6.12, "learning_rate": 5.979313634786631e-05, "loss": 1.4719, "step": 70630 }, { "epoch": 6.12, "learning_rate": 5.979096796576847e-05, "loss": 1.4866, "step": 70640 }, { "epoch": 6.12, "learning_rate": 5.9788799583670634e-05, "loss": 1.5991, "step": 70650 }, { "epoch": 6.12, "learning_rate": 5.9786631201572794e-05, "loss": 1.4695, "step": 70660 }, { "epoch": 6.12, "learning_rate": 5.978446281947496e-05, "loss": 1.6132, "step": 70670 }, { "epoch": 6.12, "learning_rate": 5.9782294437377126e-05, "loss": 1.5292, "step": 70680 }, { "epoch": 6.12, "learning_rate": 5.978012605527928e-05, "loss": 1.3971, "step": 70690 }, { "epoch": 6.12, "learning_rate": 5.9777957673181446e-05, "loss": 1.62, "step": 70700 }, { "epoch": 6.12, "learning_rate": 5.9775789291083605e-05, "loss": 1.4544, "step": 70710 }, { "epoch": 6.12, "learning_rate": 5.977362090898577e-05, "loss": 1.6689, "step": 70720 }, { "epoch": 6.13, "learning_rate": 5.977145252688793e-05, "loss": 1.3783, "step": 70730 }, { "epoch": 6.13, "learning_rate": 5.97692841447901e-05, "loss": 1.4608, "step": 70740 }, { "epoch": 6.13, "learning_rate": 5.9767115762692264e-05, "loss": 1.4247, "step": 70750 }, { "epoch": 6.13, "learning_rate": 5.9764947380594424e-05, "loss": 1.5243, "step": 70760 }, { "epoch": 6.13, "learning_rate": 5.976277899849658e-05, "loss": 1.29, "step": 70770 }, { "epoch": 6.13, "learning_rate": 5.976061061639874e-05, "loss": 1.3905, "step": 70780 }, { "epoch": 6.13, "learning_rate": 5.975844223430091e-05, "loss": 1.4465, "step": 70790 }, { "epoch": 6.13, "learning_rate": 5.975627385220307e-05, "loss": 1.6522, "step": 70800 }, { "epoch": 6.13, "learning_rate": 5.9754105470105235e-05, "loss": 1.4569, "step": 70810 }, { "epoch": 6.13, "learning_rate": 5.97519370880074e-05, "loss": 1.2762, "step": 70820 }, { "epoch": 6.13, "learning_rate": 5.974976870590956e-05, "loss": 1.4338, "step": 70830 }, { "epoch": 6.14, "learning_rate": 5.9747600323811714e-05, "loss": 1.308, "step": 70840 }, { "epoch": 6.14, "learning_rate": 5.974543194171388e-05, "loss": 1.5911, "step": 70850 }, { "epoch": 6.14, "learning_rate": 5.974326355961605e-05, "loss": 1.6989, "step": 70860 }, { "epoch": 6.14, "learning_rate": 5.9741095177518207e-05, "loss": 1.5078, "step": 70870 }, { "epoch": 6.14, "learning_rate": 5.973892679542037e-05, "loss": 1.3246, "step": 70880 }, { "epoch": 6.14, "learning_rate": 5.973675841332253e-05, "loss": 1.3652, "step": 70890 }, { "epoch": 6.14, "learning_rate": 5.97345900312247e-05, "loss": 1.4523, "step": 70900 }, { "epoch": 6.14, "learning_rate": 5.9732421649126865e-05, "loss": 1.4698, "step": 70910 }, { "epoch": 6.14, "learning_rate": 5.973025326702902e-05, "loss": 1.4702, "step": 70920 }, { "epoch": 6.14, "learning_rate": 5.9728084884931185e-05, "loss": 1.628, "step": 70930 }, { "epoch": 6.14, "learning_rate": 5.9725916502833344e-05, "loss": 1.3528, "step": 70940 }, { "epoch": 6.14, "learning_rate": 5.972374812073551e-05, "loss": 1.6021, "step": 70950 }, { "epoch": 6.15, "learning_rate": 5.972157973863767e-05, "loss": 1.5571, "step": 70960 }, { "epoch": 6.15, "learning_rate": 5.9719411356539837e-05, "loss": 1.4027, "step": 70970 }, { "epoch": 6.15, "learning_rate": 5.9717242974442e-05, "loss": 1.3835, "step": 70980 }, { "epoch": 6.15, "learning_rate": 5.971507459234416e-05, "loss": 1.5364, "step": 70990 }, { "epoch": 6.15, "learning_rate": 5.971290621024632e-05, "loss": 1.3164, "step": 71000 }, { "epoch": 6.15, "learning_rate": 5.971073782814848e-05, "loss": 1.3586, "step": 71010 }, { "epoch": 6.15, "learning_rate": 5.970856944605065e-05, "loss": 1.5309, "step": 71020 }, { "epoch": 6.15, "learning_rate": 5.970640106395281e-05, "loss": 1.5104, "step": 71030 }, { "epoch": 6.15, "learning_rate": 5.9704232681854974e-05, "loss": 1.5661, "step": 71040 }, { "epoch": 6.15, "learning_rate": 5.970206429975714e-05, "loss": 1.4334, "step": 71050 }, { "epoch": 6.15, "learning_rate": 5.96998959176593e-05, "loss": 1.4911, "step": 71060 }, { "epoch": 6.16, "learning_rate": 5.969772753556146e-05, "loss": 1.5608, "step": 71070 }, { "epoch": 6.16, "learning_rate": 5.969555915346362e-05, "loss": 1.556, "step": 71080 }, { "epoch": 6.16, "learning_rate": 5.9693390771365786e-05, "loss": 1.5012, "step": 71090 }, { "epoch": 6.16, "learning_rate": 5.9691222389267945e-05, "loss": 1.4855, "step": 71100 }, { "epoch": 6.16, "learning_rate": 5.968905400717011e-05, "loss": 1.4313, "step": 71110 }, { "epoch": 6.16, "learning_rate": 5.968688562507228e-05, "loss": 1.35, "step": 71120 }, { "epoch": 6.16, "learning_rate": 5.968471724297444e-05, "loss": 1.5483, "step": 71130 }, { "epoch": 6.16, "learning_rate": 5.9682548860876604e-05, "loss": 1.5373, "step": 71140 }, { "epoch": 6.16, "learning_rate": 5.968038047877876e-05, "loss": 1.2856, "step": 71150 }, { "epoch": 6.16, "learning_rate": 5.9678212096680923e-05, "loss": 1.5474, "step": 71160 }, { "epoch": 6.16, "learning_rate": 5.967604371458308e-05, "loss": 1.5056, "step": 71170 }, { "epoch": 6.16, "learning_rate": 5.967387533248525e-05, "loss": 1.4735, "step": 71180 }, { "epoch": 6.17, "learning_rate": 5.9671706950387416e-05, "loss": 1.5542, "step": 71190 }, { "epoch": 6.17, "learning_rate": 5.9669538568289575e-05, "loss": 1.5712, "step": 71200 }, { "epoch": 6.17, "learning_rate": 5.966737018619174e-05, "loss": 1.6282, "step": 71210 }, { "epoch": 6.17, "learning_rate": 5.96652018040939e-05, "loss": 1.4934, "step": 71220 }, { "epoch": 6.17, "learning_rate": 5.966303342199606e-05, "loss": 1.2508, "step": 71230 }, { "epoch": 6.17, "learning_rate": 5.966086503989822e-05, "loss": 1.4644, "step": 71240 }, { "epoch": 6.17, "learning_rate": 5.965869665780039e-05, "loss": 1.2682, "step": 71250 }, { "epoch": 6.17, "learning_rate": 5.9656528275702554e-05, "loss": 1.3748, "step": 71260 }, { "epoch": 6.17, "learning_rate": 5.965435989360471e-05, "loss": 1.4353, "step": 71270 }, { "epoch": 6.17, "learning_rate": 5.965219151150688e-05, "loss": 1.3848, "step": 71280 }, { "epoch": 6.17, "learning_rate": 5.965002312940904e-05, "loss": 1.4003, "step": 71290 }, { "epoch": 6.18, "learning_rate": 5.9647854747311206e-05, "loss": 1.6002, "step": 71300 }, { "epoch": 6.18, "learning_rate": 5.964568636521336e-05, "loss": 1.7597, "step": 71310 }, { "epoch": 6.18, "learning_rate": 5.9643517983115525e-05, "loss": 1.5947, "step": 71320 }, { "epoch": 6.18, "learning_rate": 5.964134960101769e-05, "loss": 1.2762, "step": 71330 }, { "epoch": 6.18, "learning_rate": 5.963918121891985e-05, "loss": 1.4448, "step": 71340 }, { "epoch": 6.18, "learning_rate": 5.963701283682202e-05, "loss": 1.4943, "step": 71350 }, { "epoch": 6.18, "learning_rate": 5.963484445472418e-05, "loss": 1.5124, "step": 71360 }, { "epoch": 6.18, "learning_rate": 5.963267607262634e-05, "loss": 1.515, "step": 71370 }, { "epoch": 6.18, "learning_rate": 5.9630507690528496e-05, "loss": 1.5116, "step": 71380 }, { "epoch": 6.18, "learning_rate": 5.962833930843066e-05, "loss": 1.308, "step": 71390 }, { "epoch": 6.18, "learning_rate": 5.962617092633282e-05, "loss": 1.447, "step": 71400 }, { "epoch": 6.18, "learning_rate": 5.962400254423499e-05, "loss": 1.3992, "step": 71410 }, { "epoch": 6.19, "learning_rate": 5.9621834162137155e-05, "loss": 1.4728, "step": 71420 }, { "epoch": 6.19, "learning_rate": 5.9619665780039314e-05, "loss": 1.5192, "step": 71430 }, { "epoch": 6.19, "learning_rate": 5.961749739794148e-05, "loss": 1.4962, "step": 71440 }, { "epoch": 6.19, "learning_rate": 5.961532901584364e-05, "loss": 1.4471, "step": 71450 }, { "epoch": 6.19, "learning_rate": 5.96131606337458e-05, "loss": 1.4193, "step": 71460 }, { "epoch": 6.19, "learning_rate": 5.961099225164796e-05, "loss": 1.5213, "step": 71470 }, { "epoch": 6.19, "learning_rate": 5.9608823869550126e-05, "loss": 1.3935, "step": 71480 }, { "epoch": 6.19, "learning_rate": 5.960665548745229e-05, "loss": 1.5371, "step": 71490 }, { "epoch": 6.19, "learning_rate": 5.960448710535445e-05, "loss": 1.4915, "step": 71500 }, { "epoch": 6.19, "learning_rate": 5.960231872325662e-05, "loss": 1.5254, "step": 71510 }, { "epoch": 6.19, "learning_rate": 5.960015034115878e-05, "loss": 1.5141, "step": 71520 }, { "epoch": 6.19, "learning_rate": 5.9597981959060944e-05, "loss": 1.4378, "step": 71530 }, { "epoch": 6.2, "learning_rate": 5.95958135769631e-05, "loss": 1.4947, "step": 71540 }, { "epoch": 6.2, "learning_rate": 5.9593645194865264e-05, "loss": 1.2866, "step": 71550 }, { "epoch": 6.2, "learning_rate": 5.959147681276743e-05, "loss": 1.5132, "step": 71560 }, { "epoch": 6.2, "learning_rate": 5.958930843066959e-05, "loss": 1.6315, "step": 71570 }, { "epoch": 6.2, "learning_rate": 5.9587140048571756e-05, "loss": 1.4916, "step": 71580 }, { "epoch": 6.2, "learning_rate": 5.9584971666473916e-05, "loss": 1.5006, "step": 71590 }, { "epoch": 6.2, "learning_rate": 5.958280328437608e-05, "loss": 1.3953, "step": 71600 }, { "epoch": 6.2, "learning_rate": 5.9580634902278235e-05, "loss": 1.599, "step": 71610 }, { "epoch": 6.2, "learning_rate": 5.95784665201804e-05, "loss": 1.3278, "step": 71620 }, { "epoch": 6.2, "learning_rate": 5.957629813808257e-05, "loss": 1.5339, "step": 71630 }, { "epoch": 6.2, "learning_rate": 5.957412975598473e-05, "loss": 1.3822, "step": 71640 }, { "epoch": 6.21, "learning_rate": 5.9571961373886894e-05, "loss": 1.4811, "step": 71650 }, { "epoch": 6.21, "learning_rate": 5.956979299178905e-05, "loss": 1.2896, "step": 71660 }, { "epoch": 6.21, "learning_rate": 5.956762460969122e-05, "loss": 1.5965, "step": 71670 }, { "epoch": 6.21, "learning_rate": 5.9565456227593386e-05, "loss": 1.5323, "step": 71680 }, { "epoch": 6.21, "learning_rate": 5.956328784549554e-05, "loss": 1.394, "step": 71690 }, { "epoch": 6.21, "learning_rate": 5.9561119463397705e-05, "loss": 1.4315, "step": 71700 }, { "epoch": 6.21, "learning_rate": 5.9558951081299865e-05, "loss": 1.6546, "step": 71710 }, { "epoch": 6.21, "learning_rate": 5.955678269920203e-05, "loss": 1.5386, "step": 71720 }, { "epoch": 6.21, "learning_rate": 5.955461431710419e-05, "loss": 1.5144, "step": 71730 }, { "epoch": 6.21, "learning_rate": 5.955244593500636e-05, "loss": 1.6021, "step": 71740 }, { "epoch": 6.21, "learning_rate": 5.9550277552908524e-05, "loss": 1.2644, "step": 71750 }, { "epoch": 6.21, "learning_rate": 5.954810917081068e-05, "loss": 1.5757, "step": 71760 }, { "epoch": 6.22, "learning_rate": 5.954594078871284e-05, "loss": 1.5768, "step": 71770 }, { "epoch": 6.22, "learning_rate": 5.9543772406615e-05, "loss": 1.4881, "step": 71780 }, { "epoch": 6.22, "learning_rate": 5.954160402451717e-05, "loss": 1.6891, "step": 71790 }, { "epoch": 6.22, "learning_rate": 5.953943564241933e-05, "loss": 1.4232, "step": 71800 }, { "epoch": 6.22, "learning_rate": 5.9537267260321495e-05, "loss": 1.5057, "step": 71810 }, { "epoch": 6.22, "learning_rate": 5.953509887822366e-05, "loss": 1.7174, "step": 71820 }, { "epoch": 6.22, "learning_rate": 5.953293049612582e-05, "loss": 1.5091, "step": 71830 }, { "epoch": 6.22, "learning_rate": 5.953076211402798e-05, "loss": 1.3997, "step": 71840 }, { "epoch": 6.22, "learning_rate": 5.952859373193014e-05, "loss": 1.6214, "step": 71850 }, { "epoch": 6.22, "learning_rate": 5.952642534983231e-05, "loss": 1.5952, "step": 71860 }, { "epoch": 6.22, "learning_rate": 5.9524256967734466e-05, "loss": 1.5679, "step": 71870 }, { "epoch": 6.23, "learning_rate": 5.952208858563663e-05, "loss": 1.4674, "step": 71880 }, { "epoch": 6.23, "learning_rate": 5.95199202035388e-05, "loss": 1.6119, "step": 71890 }, { "epoch": 6.23, "learning_rate": 5.951775182144096e-05, "loss": 1.3742, "step": 71900 }, { "epoch": 6.23, "learning_rate": 5.9515583439343125e-05, "loss": 1.1921, "step": 71910 }, { "epoch": 6.23, "learning_rate": 5.951341505724528e-05, "loss": 1.5653, "step": 71920 }, { "epoch": 6.23, "learning_rate": 5.9511246675147444e-05, "loss": 1.4325, "step": 71930 }, { "epoch": 6.23, "learning_rate": 5.9509078293049604e-05, "loss": 1.4766, "step": 71940 }, { "epoch": 6.23, "learning_rate": 5.950690991095177e-05, "loss": 1.1554, "step": 71950 }, { "epoch": 6.23, "learning_rate": 5.950474152885393e-05, "loss": 1.507, "step": 71960 }, { "epoch": 6.23, "learning_rate": 5.9502573146756096e-05, "loss": 1.3292, "step": 71970 }, { "epoch": 6.23, "learning_rate": 5.950040476465826e-05, "loss": 1.4385, "step": 71980 }, { "epoch": 6.23, "learning_rate": 5.949823638256042e-05, "loss": 1.3841, "step": 71990 }, { "epoch": 6.24, "learning_rate": 5.949606800046258e-05, "loss": 1.4614, "step": 72000 }, { "epoch": 6.24, "learning_rate": 5.949389961836474e-05, "loss": 1.5356, "step": 72010 }, { "epoch": 6.24, "learning_rate": 5.949173123626691e-05, "loss": 1.569, "step": 72020 }, { "epoch": 6.24, "learning_rate": 5.948956285416907e-05, "loss": 1.4264, "step": 72030 }, { "epoch": 6.24, "learning_rate": 5.9487394472071234e-05, "loss": 1.5692, "step": 72040 }, { "epoch": 6.24, "learning_rate": 5.94852260899734e-05, "loss": 1.4917, "step": 72050 }, { "epoch": 6.24, "learning_rate": 5.948305770787556e-05, "loss": 1.3812, "step": 72060 }, { "epoch": 6.24, "learning_rate": 5.9480889325777726e-05, "loss": 1.4347, "step": 72070 }, { "epoch": 6.24, "learning_rate": 5.947872094367988e-05, "loss": 1.505, "step": 72080 }, { "epoch": 6.24, "learning_rate": 5.9476552561582046e-05, "loss": 1.4692, "step": 72090 }, { "epoch": 6.24, "learning_rate": 5.9474384179484205e-05, "loss": 1.5148, "step": 72100 }, { "epoch": 6.25, "learning_rate": 5.947221579738637e-05, "loss": 1.4287, "step": 72110 }, { "epoch": 6.25, "learning_rate": 5.947004741528854e-05, "loss": 1.629, "step": 72120 }, { "epoch": 6.25, "learning_rate": 5.94678790331907e-05, "loss": 1.4481, "step": 72130 }, { "epoch": 6.25, "learning_rate": 5.9465710651092864e-05, "loss": 1.5362, "step": 72140 }, { "epoch": 6.25, "learning_rate": 5.946354226899502e-05, "loss": 1.684, "step": 72150 }, { "epoch": 6.25, "learning_rate": 5.946137388689718e-05, "loss": 1.395, "step": 72160 }, { "epoch": 6.25, "learning_rate": 5.945920550479934e-05, "loss": 1.6641, "step": 72170 }, { "epoch": 6.25, "learning_rate": 5.945703712270151e-05, "loss": 1.6406, "step": 72180 }, { "epoch": 6.25, "learning_rate": 5.9454868740603676e-05, "loss": 1.5134, "step": 72190 }, { "epoch": 6.25, "learning_rate": 5.9452700358505835e-05, "loss": 1.3351, "step": 72200 }, { "epoch": 6.25, "learning_rate": 5.9450531976408e-05, "loss": 1.2097, "step": 72210 }, { "epoch": 6.25, "learning_rate": 5.944836359431016e-05, "loss": 1.4076, "step": 72220 }, { "epoch": 6.26, "learning_rate": 5.944619521221232e-05, "loss": 1.5118, "step": 72230 }, { "epoch": 6.26, "learning_rate": 5.944402683011448e-05, "loss": 1.3918, "step": 72240 }, { "epoch": 6.26, "learning_rate": 5.944185844801665e-05, "loss": 1.3875, "step": 72250 }, { "epoch": 6.26, "learning_rate": 5.943969006591881e-05, "loss": 1.4884, "step": 72260 }, { "epoch": 6.26, "learning_rate": 5.943752168382097e-05, "loss": 1.5644, "step": 72270 }, { "epoch": 6.26, "learning_rate": 5.943535330172314e-05, "loss": 1.2516, "step": 72280 }, { "epoch": 6.26, "learning_rate": 5.94331849196253e-05, "loss": 1.5386, "step": 72290 }, { "epoch": 6.26, "learning_rate": 5.9431016537527465e-05, "loss": 1.4175, "step": 72300 }, { "epoch": 6.26, "learning_rate": 5.942884815542962e-05, "loss": 1.4227, "step": 72310 }, { "epoch": 6.26, "learning_rate": 5.9426679773331784e-05, "loss": 1.5763, "step": 72320 }, { "epoch": 6.26, "learning_rate": 5.942451139123395e-05, "loss": 1.4683, "step": 72330 }, { "epoch": 6.27, "learning_rate": 5.942234300913611e-05, "loss": 1.5758, "step": 72340 }, { "epoch": 6.27, "learning_rate": 5.942017462703828e-05, "loss": 1.4046, "step": 72350 }, { "epoch": 6.27, "learning_rate": 5.9418006244940436e-05, "loss": 1.4905, "step": 72360 }, { "epoch": 6.27, "learning_rate": 5.94158378628426e-05, "loss": 1.4047, "step": 72370 }, { "epoch": 6.27, "learning_rate": 5.9413669480744756e-05, "loss": 1.4742, "step": 72380 }, { "epoch": 6.27, "learning_rate": 5.941150109864692e-05, "loss": 1.497, "step": 72390 }, { "epoch": 6.27, "learning_rate": 5.940933271654909e-05, "loss": 1.4876, "step": 72400 }, { "epoch": 6.27, "learning_rate": 5.940716433445125e-05, "loss": 1.5546, "step": 72410 }, { "epoch": 6.27, "learning_rate": 5.9404995952353414e-05, "loss": 1.6429, "step": 72420 }, { "epoch": 6.27, "learning_rate": 5.9402827570255574e-05, "loss": 1.7315, "step": 72430 }, { "epoch": 6.27, "learning_rate": 5.940065918815774e-05, "loss": 1.5637, "step": 72440 }, { "epoch": 6.27, "learning_rate": 5.939849080605991e-05, "loss": 1.4221, "step": 72450 }, { "epoch": 6.28, "learning_rate": 5.939632242396206e-05, "loss": 1.5239, "step": 72460 }, { "epoch": 6.28, "learning_rate": 5.939415404186422e-05, "loss": 1.5161, "step": 72470 }, { "epoch": 6.28, "learning_rate": 5.9391985659766386e-05, "loss": 1.5371, "step": 72480 }, { "epoch": 6.28, "learning_rate": 5.938981727766855e-05, "loss": 1.5241, "step": 72490 }, { "epoch": 6.28, "learning_rate": 5.938764889557071e-05, "loss": 1.4648, "step": 72500 }, { "epoch": 6.28, "learning_rate": 5.938548051347288e-05, "loss": 1.4474, "step": 72510 }, { "epoch": 6.28, "learning_rate": 5.938331213137504e-05, "loss": 1.4744, "step": 72520 }, { "epoch": 6.28, "learning_rate": 5.9381143749277204e-05, "loss": 1.4193, "step": 72530 }, { "epoch": 6.28, "learning_rate": 5.937897536717936e-05, "loss": 1.4741, "step": 72540 }, { "epoch": 6.28, "learning_rate": 5.937680698508152e-05, "loss": 1.3486, "step": 72550 }, { "epoch": 6.28, "learning_rate": 5.937463860298369e-05, "loss": 1.4173, "step": 72560 }, { "epoch": 6.29, "learning_rate": 5.937247022088585e-05, "loss": 1.3722, "step": 72570 }, { "epoch": 6.29, "learning_rate": 5.9370301838788016e-05, "loss": 1.5186, "step": 72580 }, { "epoch": 6.29, "learning_rate": 5.9368133456690175e-05, "loss": 1.4246, "step": 72590 }, { "epoch": 6.29, "learning_rate": 5.936596507459234e-05, "loss": 1.6354, "step": 72600 }, { "epoch": 6.29, "learning_rate": 5.9363796692494495e-05, "loss": 1.6799, "step": 72610 }, { "epoch": 6.29, "learning_rate": 5.936162831039666e-05, "loss": 1.459, "step": 72620 }, { "epoch": 6.29, "learning_rate": 5.935945992829883e-05, "loss": 1.5105, "step": 72630 }, { "epoch": 6.29, "learning_rate": 5.935729154620099e-05, "loss": 1.436, "step": 72640 }, { "epoch": 6.29, "learning_rate": 5.9355123164103153e-05, "loss": 1.5472, "step": 72650 }, { "epoch": 6.29, "learning_rate": 5.935295478200531e-05, "loss": 1.5945, "step": 72660 }, { "epoch": 6.29, "learning_rate": 5.935078639990748e-05, "loss": 1.6483, "step": 72670 }, { "epoch": 6.29, "learning_rate": 5.9348618017809646e-05, "loss": 1.7135, "step": 72680 }, { "epoch": 6.3, "learning_rate": 5.93464496357118e-05, "loss": 1.2892, "step": 72690 }, { "epoch": 6.3, "learning_rate": 5.9344281253613965e-05, "loss": 1.4859, "step": 72700 }, { "epoch": 6.3, "learning_rate": 5.9342112871516125e-05, "loss": 1.5621, "step": 72710 }, { "epoch": 6.3, "learning_rate": 5.933994448941829e-05, "loss": 1.3559, "step": 72720 }, { "epoch": 6.3, "learning_rate": 5.933777610732045e-05, "loss": 1.5361, "step": 72730 }, { "epoch": 6.3, "learning_rate": 5.933560772522262e-05, "loss": 1.4191, "step": 72740 }, { "epoch": 6.3, "learning_rate": 5.9333439343124783e-05, "loss": 1.4856, "step": 72750 }, { "epoch": 6.3, "learning_rate": 5.933127096102694e-05, "loss": 1.4573, "step": 72760 }, { "epoch": 6.3, "learning_rate": 5.93291025789291e-05, "loss": 1.4736, "step": 72770 }, { "epoch": 6.3, "learning_rate": 5.932693419683126e-05, "loss": 1.4726, "step": 72780 }, { "epoch": 6.3, "learning_rate": 5.932476581473343e-05, "loss": 1.6393, "step": 72790 }, { "epoch": 6.3, "learning_rate": 5.932259743263559e-05, "loss": 1.5303, "step": 72800 }, { "epoch": 6.31, "learning_rate": 5.9320429050537755e-05, "loss": 1.5202, "step": 72810 }, { "epoch": 6.31, "learning_rate": 5.931826066843992e-05, "loss": 1.3597, "step": 72820 }, { "epoch": 6.31, "learning_rate": 5.931609228634208e-05, "loss": 1.5184, "step": 72830 }, { "epoch": 6.31, "learning_rate": 5.931392390424425e-05, "loss": 1.4097, "step": 72840 }, { "epoch": 6.31, "learning_rate": 5.93117555221464e-05, "loss": 1.4815, "step": 72850 }, { "epoch": 6.31, "learning_rate": 5.9309587140048566e-05, "loss": 1.3403, "step": 72860 }, { "epoch": 6.31, "learning_rate": 5.9307418757950726e-05, "loss": 1.2879, "step": 72870 }, { "epoch": 6.31, "learning_rate": 5.930525037585289e-05, "loss": 1.4683, "step": 72880 }, { "epoch": 6.31, "learning_rate": 5.930308199375506e-05, "loss": 1.4737, "step": 72890 }, { "epoch": 6.31, "learning_rate": 5.930091361165722e-05, "loss": 1.492, "step": 72900 }, { "epoch": 6.31, "learning_rate": 5.9298745229559385e-05, "loss": 1.4697, "step": 72910 }, { "epoch": 6.32, "learning_rate": 5.929657684746154e-05, "loss": 1.4394, "step": 72920 }, { "epoch": 6.32, "learning_rate": 5.9294408465363704e-05, "loss": 1.3447, "step": 72930 }, { "epoch": 6.32, "learning_rate": 5.9292240083265864e-05, "loss": 1.4722, "step": 72940 }, { "epoch": 6.32, "learning_rate": 5.929007170116803e-05, "loss": 1.4816, "step": 72950 }, { "epoch": 6.32, "learning_rate": 5.9287903319070196e-05, "loss": 1.5633, "step": 72960 }, { "epoch": 6.32, "learning_rate": 5.9285734936972356e-05, "loss": 1.3655, "step": 72970 }, { "epoch": 6.32, "learning_rate": 5.928356655487452e-05, "loss": 1.4541, "step": 72980 }, { "epoch": 6.32, "learning_rate": 5.928139817277668e-05, "loss": 1.3939, "step": 72990 }, { "epoch": 6.32, "learning_rate": 5.927922979067884e-05, "loss": 1.6653, "step": 73000 }, { "epoch": 6.32, "learning_rate": 5.9277061408581e-05, "loss": 1.4007, "step": 73010 }, { "epoch": 6.32, "learning_rate": 5.927489302648317e-05, "loss": 1.4571, "step": 73020 }, { "epoch": 6.32, "learning_rate": 5.927272464438533e-05, "loss": 1.4252, "step": 73030 }, { "epoch": 6.33, "learning_rate": 5.9270556262287494e-05, "loss": 1.331, "step": 73040 }, { "epoch": 6.33, "learning_rate": 5.926838788018966e-05, "loss": 1.5681, "step": 73050 }, { "epoch": 6.33, "learning_rate": 5.926621949809182e-05, "loss": 1.454, "step": 73060 }, { "epoch": 6.33, "learning_rate": 5.9264051115993986e-05, "loss": 1.6052, "step": 73070 }, { "epoch": 6.33, "learning_rate": 5.926188273389614e-05, "loss": 1.5713, "step": 73080 }, { "epoch": 6.33, "learning_rate": 5.9259714351798305e-05, "loss": 1.4082, "step": 73090 }, { "epoch": 6.33, "learning_rate": 5.9257545969700465e-05, "loss": 1.5141, "step": 73100 }, { "epoch": 6.33, "learning_rate": 5.925537758760263e-05, "loss": 1.4483, "step": 73110 }, { "epoch": 6.33, "learning_rate": 5.92532092055048e-05, "loss": 1.5985, "step": 73120 }, { "epoch": 6.33, "learning_rate": 5.925104082340696e-05, "loss": 1.6043, "step": 73130 }, { "epoch": 6.33, "learning_rate": 5.9248872441309124e-05, "loss": 1.4651, "step": 73140 }, { "epoch": 6.34, "learning_rate": 5.9246704059211276e-05, "loss": 1.5918, "step": 73150 }, { "epoch": 6.34, "learning_rate": 5.924453567711344e-05, "loss": 1.5529, "step": 73160 }, { "epoch": 6.34, "learning_rate": 5.92423672950156e-05, "loss": 1.355, "step": 73170 }, { "epoch": 6.34, "learning_rate": 5.924019891291777e-05, "loss": 1.5065, "step": 73180 }, { "epoch": 6.34, "learning_rate": 5.9238030530819935e-05, "loss": 1.5489, "step": 73190 }, { "epoch": 6.34, "learning_rate": 5.9235862148722095e-05, "loss": 1.5161, "step": 73200 }, { "epoch": 6.34, "learning_rate": 5.923369376662426e-05, "loss": 1.5281, "step": 73210 }, { "epoch": 6.34, "learning_rate": 5.923152538452642e-05, "loss": 1.4097, "step": 73220 }, { "epoch": 6.34, "learning_rate": 5.922935700242858e-05, "loss": 1.5312, "step": 73230 }, { "epoch": 6.34, "learning_rate": 5.922718862033074e-05, "loss": 1.3648, "step": 73240 }, { "epoch": 6.34, "learning_rate": 5.9225020238232907e-05, "loss": 1.5103, "step": 73250 }, { "epoch": 6.34, "learning_rate": 5.922285185613507e-05, "loss": 1.3618, "step": 73260 }, { "epoch": 6.35, "learning_rate": 5.922068347403723e-05, "loss": 1.6724, "step": 73270 }, { "epoch": 6.35, "learning_rate": 5.92185150919394e-05, "loss": 1.4953, "step": 73280 }, { "epoch": 6.35, "learning_rate": 5.921634670984156e-05, "loss": 1.6498, "step": 73290 }, { "epoch": 6.35, "learning_rate": 5.9214178327743725e-05, "loss": 1.4644, "step": 73300 }, { "epoch": 6.35, "learning_rate": 5.921200994564588e-05, "loss": 1.4063, "step": 73310 }, { "epoch": 6.35, "learning_rate": 5.9209841563548044e-05, "loss": 1.5066, "step": 73320 }, { "epoch": 6.35, "learning_rate": 5.920767318145021e-05, "loss": 1.5283, "step": 73330 }, { "epoch": 6.35, "learning_rate": 5.920550479935237e-05, "loss": 1.3764, "step": 73340 }, { "epoch": 6.35, "learning_rate": 5.9203336417254537e-05, "loss": 1.4457, "step": 73350 }, { "epoch": 6.35, "learning_rate": 5.9201168035156696e-05, "loss": 1.3768, "step": 73360 }, { "epoch": 6.35, "learning_rate": 5.919899965305886e-05, "loss": 1.4232, "step": 73370 }, { "epoch": 6.36, "learning_rate": 5.9196831270961015e-05, "loss": 1.3775, "step": 73380 }, { "epoch": 6.36, "learning_rate": 5.919466288886318e-05, "loss": 1.5807, "step": 73390 }, { "epoch": 6.36, "learning_rate": 5.919249450676535e-05, "loss": 1.6706, "step": 73400 }, { "epoch": 6.36, "learning_rate": 5.919032612466751e-05, "loss": 1.4908, "step": 73410 }, { "epoch": 6.36, "learning_rate": 5.9188157742569674e-05, "loss": 1.3188, "step": 73420 }, { "epoch": 6.36, "learning_rate": 5.9185989360471834e-05, "loss": 1.4757, "step": 73430 }, { "epoch": 6.36, "learning_rate": 5.9183820978374e-05, "loss": 1.3072, "step": 73440 }, { "epoch": 6.36, "learning_rate": 5.9181652596276167e-05, "loss": 1.3874, "step": 73450 }, { "epoch": 6.36, "learning_rate": 5.917948421417832e-05, "loss": 1.4048, "step": 73460 }, { "epoch": 6.36, "learning_rate": 5.9177315832080486e-05, "loss": 1.4852, "step": 73470 }, { "epoch": 6.36, "learning_rate": 5.9175147449982645e-05, "loss": 1.5383, "step": 73480 }, { "epoch": 6.36, "learning_rate": 5.917297906788481e-05, "loss": 1.5712, "step": 73490 }, { "epoch": 6.37, "learning_rate": 5.917081068578697e-05, "loss": 1.5737, "step": 73500 }, { "epoch": 6.37, "learning_rate": 5.916864230368914e-05, "loss": 1.4824, "step": 73510 }, { "epoch": 6.37, "learning_rate": 5.9166473921591304e-05, "loss": 1.4912, "step": 73520 }, { "epoch": 6.37, "learning_rate": 5.9164305539493464e-05, "loss": 1.5474, "step": 73530 }, { "epoch": 6.37, "learning_rate": 5.916213715739562e-05, "loss": 1.5087, "step": 73540 }, { "epoch": 6.37, "learning_rate": 5.915996877529778e-05, "loss": 1.5407, "step": 73550 }, { "epoch": 6.37, "learning_rate": 5.915780039319995e-05, "loss": 1.5458, "step": 73560 }, { "epoch": 6.37, "learning_rate": 5.915563201110211e-05, "loss": 1.5304, "step": 73570 }, { "epoch": 6.37, "learning_rate": 5.9153463629004275e-05, "loss": 1.5444, "step": 73580 }, { "epoch": 6.37, "learning_rate": 5.9151295246906435e-05, "loss": 1.486, "step": 73590 }, { "epoch": 6.37, "learning_rate": 5.91491268648086e-05, "loss": 1.5885, "step": 73600 }, { "epoch": 6.38, "learning_rate": 5.914695848271077e-05, "loss": 1.5086, "step": 73610 }, { "epoch": 6.38, "learning_rate": 5.914479010061292e-05, "loss": 1.4269, "step": 73620 }, { "epoch": 6.38, "learning_rate": 5.914262171851509e-05, "loss": 1.4896, "step": 73630 }, { "epoch": 6.38, "learning_rate": 5.914045333641725e-05, "loss": 1.3866, "step": 73640 }, { "epoch": 6.38, "learning_rate": 5.913828495431941e-05, "loss": 1.3332, "step": 73650 }, { "epoch": 6.38, "learning_rate": 5.913611657222157e-05, "loss": 1.5667, "step": 73660 }, { "epoch": 6.38, "learning_rate": 5.913394819012374e-05, "loss": 1.6426, "step": 73670 }, { "epoch": 6.38, "learning_rate": 5.9131779808025905e-05, "loss": 1.6512, "step": 73680 }, { "epoch": 6.38, "learning_rate": 5.912961142592806e-05, "loss": 1.5527, "step": 73690 }, { "epoch": 6.38, "learning_rate": 5.9127443043830225e-05, "loss": 1.3929, "step": 73700 }, { "epoch": 6.38, "learning_rate": 5.9125274661732384e-05, "loss": 1.3906, "step": 73710 }, { "epoch": 6.38, "learning_rate": 5.912310627963455e-05, "loss": 1.5692, "step": 73720 }, { "epoch": 6.39, "learning_rate": 5.912093789753671e-05, "loss": 1.3775, "step": 73730 }, { "epoch": 6.39, "learning_rate": 5.911876951543888e-05, "loss": 1.4846, "step": 73740 }, { "epoch": 6.39, "learning_rate": 5.911660113334104e-05, "loss": 1.5433, "step": 73750 }, { "epoch": 6.39, "learning_rate": 5.91144327512432e-05, "loss": 1.6628, "step": 73760 }, { "epoch": 6.39, "learning_rate": 5.911226436914536e-05, "loss": 1.464, "step": 73770 }, { "epoch": 6.39, "learning_rate": 5.911009598704752e-05, "loss": 1.46, "step": 73780 }, { "epoch": 6.39, "learning_rate": 5.910792760494969e-05, "loss": 1.4518, "step": 73790 }, { "epoch": 6.39, "learning_rate": 5.910575922285185e-05, "loss": 1.3508, "step": 73800 }, { "epoch": 6.39, "learning_rate": 5.9103590840754014e-05, "loss": 1.5046, "step": 73810 }, { "epoch": 6.39, "learning_rate": 5.910142245865618e-05, "loss": 1.6873, "step": 73820 }, { "epoch": 6.39, "learning_rate": 5.909925407655834e-05, "loss": 1.7275, "step": 73830 }, { "epoch": 6.4, "learning_rate": 5.909708569446051e-05, "loss": 1.3127, "step": 73840 }, { "epoch": 6.4, "learning_rate": 5.909491731236266e-05, "loss": 1.358, "step": 73850 }, { "epoch": 6.4, "learning_rate": 5.9092748930264826e-05, "loss": 1.4917, "step": 73860 }, { "epoch": 6.4, "learning_rate": 5.9090580548166986e-05, "loss": 1.5357, "step": 73870 }, { "epoch": 6.4, "learning_rate": 5.908841216606915e-05, "loss": 1.5434, "step": 73880 }, { "epoch": 6.4, "learning_rate": 5.908624378397132e-05, "loss": 1.4742, "step": 73890 }, { "epoch": 6.4, "learning_rate": 5.908407540187348e-05, "loss": 1.5423, "step": 73900 }, { "epoch": 6.4, "learning_rate": 5.9081907019775644e-05, "loss": 1.5969, "step": 73910 }, { "epoch": 6.4, "learning_rate": 5.90797386376778e-05, "loss": 1.455, "step": 73920 }, { "epoch": 6.4, "learning_rate": 5.9077570255579964e-05, "loss": 1.3519, "step": 73930 }, { "epoch": 6.4, "learning_rate": 5.907540187348212e-05, "loss": 1.4216, "step": 73940 }, { "epoch": 6.4, "learning_rate": 5.907323349138429e-05, "loss": 1.2599, "step": 73950 }, { "epoch": 6.41, "learning_rate": 5.9071065109286456e-05, "loss": 1.5706, "step": 73960 }, { "epoch": 6.41, "learning_rate": 5.9068896727188616e-05, "loss": 1.3632, "step": 73970 }, { "epoch": 6.41, "learning_rate": 5.906672834509078e-05, "loss": 1.6018, "step": 73980 }, { "epoch": 6.41, "learning_rate": 5.906455996299294e-05, "loss": 1.5213, "step": 73990 }, { "epoch": 6.41, "learning_rate": 5.90623915808951e-05, "loss": 1.5281, "step": 74000 }, { "epoch": 6.41, "learning_rate": 5.906022319879726e-05, "loss": 1.6396, "step": 74010 }, { "epoch": 6.41, "learning_rate": 5.905805481669943e-05, "loss": 1.4471, "step": 74020 }, { "epoch": 6.41, "learning_rate": 5.9055886434601594e-05, "loss": 1.4013, "step": 74030 }, { "epoch": 6.41, "learning_rate": 5.905371805250375e-05, "loss": 1.4523, "step": 74040 }, { "epoch": 6.41, "learning_rate": 5.905154967040592e-05, "loss": 1.4844, "step": 74050 }, { "epoch": 6.41, "learning_rate": 5.904938128830808e-05, "loss": 1.4146, "step": 74060 }, { "epoch": 6.41, "learning_rate": 5.9047212906210246e-05, "loss": 1.3707, "step": 74070 }, { "epoch": 6.42, "learning_rate": 5.90450445241124e-05, "loss": 1.4213, "step": 74080 }, { "epoch": 6.42, "learning_rate": 5.9042876142014565e-05, "loss": 1.5832, "step": 74090 }, { "epoch": 6.42, "learning_rate": 5.9040707759916725e-05, "loss": 1.367, "step": 74100 }, { "epoch": 6.42, "learning_rate": 5.903853937781889e-05, "loss": 1.6137, "step": 74110 }, { "epoch": 6.42, "learning_rate": 5.903637099572106e-05, "loss": 1.506, "step": 74120 }, { "epoch": 6.42, "learning_rate": 5.903420261362322e-05, "loss": 1.5568, "step": 74130 }, { "epoch": 6.42, "learning_rate": 5.903203423152538e-05, "loss": 1.4995, "step": 74140 }, { "epoch": 6.42, "learning_rate": 5.9029865849427536e-05, "loss": 1.3395, "step": 74150 }, { "epoch": 6.42, "learning_rate": 5.90276974673297e-05, "loss": 1.4828, "step": 74160 }, { "epoch": 6.42, "learning_rate": 5.902552908523186e-05, "loss": 1.3446, "step": 74170 }, { "epoch": 6.42, "learning_rate": 5.902336070313403e-05, "loss": 1.4627, "step": 74180 }, { "epoch": 6.43, "learning_rate": 5.9021192321036195e-05, "loss": 1.4779, "step": 74190 }, { "epoch": 6.43, "learning_rate": 5.9019023938938355e-05, "loss": 1.4427, "step": 74200 }, { "epoch": 6.43, "learning_rate": 5.901685555684052e-05, "loss": 1.6946, "step": 74210 }, { "epoch": 6.43, "learning_rate": 5.901468717474268e-05, "loss": 1.3606, "step": 74220 }, { "epoch": 6.43, "learning_rate": 5.901251879264484e-05, "loss": 1.444, "step": 74230 }, { "epoch": 6.43, "learning_rate": 5.9010350410547e-05, "loss": 1.4723, "step": 74240 }, { "epoch": 6.43, "learning_rate": 5.9008182028449166e-05, "loss": 1.4324, "step": 74250 }, { "epoch": 6.43, "learning_rate": 5.900601364635133e-05, "loss": 1.7221, "step": 74260 }, { "epoch": 6.43, "learning_rate": 5.900384526425349e-05, "loss": 1.6292, "step": 74270 }, { "epoch": 6.43, "learning_rate": 5.900167688215566e-05, "loss": 1.3472, "step": 74280 }, { "epoch": 6.43, "learning_rate": 5.899950850005782e-05, "loss": 1.4705, "step": 74290 }, { "epoch": 6.43, "learning_rate": 5.8997340117959985e-05, "loss": 1.3618, "step": 74300 }, { "epoch": 6.44, "learning_rate": 5.899517173586214e-05, "loss": 1.5205, "step": 74310 }, { "epoch": 6.44, "learning_rate": 5.8993003353764304e-05, "loss": 1.4977, "step": 74320 }, { "epoch": 6.44, "learning_rate": 5.899083497166647e-05, "loss": 1.6217, "step": 74330 }, { "epoch": 6.44, "learning_rate": 5.898866658956863e-05, "loss": 1.4249, "step": 74340 }, { "epoch": 6.44, "learning_rate": 5.8986498207470796e-05, "loss": 1.5068, "step": 74350 }, { "epoch": 6.44, "learning_rate": 5.8984329825372956e-05, "loss": 1.3667, "step": 74360 }, { "epoch": 6.44, "learning_rate": 5.898216144327512e-05, "loss": 1.6783, "step": 74370 }, { "epoch": 6.44, "learning_rate": 5.897999306117729e-05, "loss": 1.4406, "step": 74380 }, { "epoch": 6.44, "learning_rate": 5.897782467907944e-05, "loss": 1.5751, "step": 74390 }, { "epoch": 6.44, "learning_rate": 5.897565629698161e-05, "loss": 1.7734, "step": 74400 }, { "epoch": 6.44, "learning_rate": 5.897348791488377e-05, "loss": 1.4645, "step": 74410 }, { "epoch": 6.45, "learning_rate": 5.8971319532785934e-05, "loss": 1.4389, "step": 74420 }, { "epoch": 6.45, "learning_rate": 5.8969151150688093e-05, "loss": 1.5177, "step": 74430 }, { "epoch": 6.45, "learning_rate": 5.896698276859026e-05, "loss": 1.5027, "step": 74440 }, { "epoch": 6.45, "learning_rate": 5.8964814386492426e-05, "loss": 1.3381, "step": 74450 }, { "epoch": 6.45, "learning_rate": 5.896264600439458e-05, "loss": 1.2191, "step": 74460 }, { "epoch": 6.45, "learning_rate": 5.8960477622296746e-05, "loss": 1.5314, "step": 74470 }, { "epoch": 6.45, "learning_rate": 5.8958309240198905e-05, "loss": 1.6664, "step": 74480 }, { "epoch": 6.45, "learning_rate": 5.895614085810107e-05, "loss": 1.3022, "step": 74490 }, { "epoch": 6.45, "learning_rate": 5.895397247600323e-05, "loss": 1.4995, "step": 74500 }, { "epoch": 6.45, "learning_rate": 5.89518040939054e-05, "loss": 1.2887, "step": 74510 }, { "epoch": 6.45, "learning_rate": 5.8949635711807564e-05, "loss": 1.7376, "step": 74520 }, { "epoch": 6.45, "learning_rate": 5.8947467329709724e-05, "loss": 1.3589, "step": 74530 }, { "epoch": 6.46, "learning_rate": 5.894529894761188e-05, "loss": 1.6779, "step": 74540 }, { "epoch": 6.46, "learning_rate": 5.894313056551404e-05, "loss": 1.3608, "step": 74550 }, { "epoch": 6.46, "learning_rate": 5.894096218341621e-05, "loss": 1.4199, "step": 74560 }, { "epoch": 6.46, "learning_rate": 5.893879380131837e-05, "loss": 1.4027, "step": 74570 }, { "epoch": 6.46, "learning_rate": 5.8936625419220535e-05, "loss": 1.6535, "step": 74580 }, { "epoch": 6.46, "learning_rate": 5.89344570371227e-05, "loss": 1.4215, "step": 74590 }, { "epoch": 6.46, "learning_rate": 5.893228865502486e-05, "loss": 1.4072, "step": 74600 }, { "epoch": 6.46, "learning_rate": 5.893012027292703e-05, "loss": 1.3289, "step": 74610 }, { "epoch": 6.46, "learning_rate": 5.892795189082918e-05, "loss": 1.6564, "step": 74620 }, { "epoch": 6.46, "learning_rate": 5.892578350873135e-05, "loss": 1.4669, "step": 74630 }, { "epoch": 6.46, "learning_rate": 5.8923615126633506e-05, "loss": 1.4902, "step": 74640 }, { "epoch": 6.47, "learning_rate": 5.892144674453567e-05, "loss": 1.5246, "step": 74650 }, { "epoch": 6.47, "learning_rate": 5.891927836243783e-05, "loss": 1.544, "step": 74660 }, { "epoch": 6.47, "learning_rate": 5.891710998034e-05, "loss": 1.58, "step": 74670 }, { "epoch": 6.47, "learning_rate": 5.8914941598242165e-05, "loss": 1.5167, "step": 74680 }, { "epoch": 6.47, "learning_rate": 5.891277321614432e-05, "loss": 1.4072, "step": 74690 }, { "epoch": 6.47, "learning_rate": 5.8910604834046484e-05, "loss": 1.4747, "step": 74700 }, { "epoch": 6.47, "learning_rate": 5.8908436451948644e-05, "loss": 1.5219, "step": 74710 }, { "epoch": 6.47, "learning_rate": 5.890626806985081e-05, "loss": 1.4434, "step": 74720 }, { "epoch": 6.47, "learning_rate": 5.890409968775297e-05, "loss": 1.6171, "step": 74730 }, { "epoch": 6.47, "learning_rate": 5.8901931305655136e-05, "loss": 1.2753, "step": 74740 }, { "epoch": 6.47, "learning_rate": 5.88997629235573e-05, "loss": 1.7078, "step": 74750 }, { "epoch": 6.47, "learning_rate": 5.889759454145946e-05, "loss": 1.2684, "step": 74760 }, { "epoch": 6.48, "learning_rate": 5.889542615936162e-05, "loss": 1.4952, "step": 74770 }, { "epoch": 6.48, "learning_rate": 5.889325777726378e-05, "loss": 1.4418, "step": 74780 }, { "epoch": 6.48, "learning_rate": 5.889108939516595e-05, "loss": 1.2968, "step": 74790 }, { "epoch": 6.48, "learning_rate": 5.888892101306811e-05, "loss": 1.7034, "step": 74800 }, { "epoch": 6.48, "learning_rate": 5.8886752630970274e-05, "loss": 1.473, "step": 74810 }, { "epoch": 6.48, "learning_rate": 5.888458424887244e-05, "loss": 1.435, "step": 74820 }, { "epoch": 6.48, "learning_rate": 5.88824158667746e-05, "loss": 1.5922, "step": 74830 }, { "epoch": 6.48, "learning_rate": 5.8880247484676766e-05, "loss": 1.4633, "step": 74840 }, { "epoch": 6.48, "learning_rate": 5.887807910257892e-05, "loss": 1.3476, "step": 74850 }, { "epoch": 6.48, "learning_rate": 5.8875910720481086e-05, "loss": 1.3767, "step": 74860 }, { "epoch": 6.48, "learning_rate": 5.8873742338383245e-05, "loss": 1.5175, "step": 74870 }, { "epoch": 6.49, "learning_rate": 5.887157395628541e-05, "loss": 1.4008, "step": 74880 }, { "epoch": 6.49, "learning_rate": 5.886940557418758e-05, "loss": 1.5245, "step": 74890 }, { "epoch": 6.49, "learning_rate": 5.886723719208974e-05, "loss": 1.6298, "step": 74900 }, { "epoch": 6.49, "learning_rate": 5.8865068809991904e-05, "loss": 1.4866, "step": 74910 }, { "epoch": 6.49, "learning_rate": 5.886290042789406e-05, "loss": 1.3725, "step": 74920 }, { "epoch": 6.49, "learning_rate": 5.886073204579622e-05, "loss": 1.5517, "step": 74930 }, { "epoch": 6.49, "learning_rate": 5.885856366369838e-05, "loss": 1.4992, "step": 74940 }, { "epoch": 6.49, "learning_rate": 5.885639528160055e-05, "loss": 1.7563, "step": 74950 }, { "epoch": 6.49, "learning_rate": 5.8854226899502716e-05, "loss": 1.4242, "step": 74960 }, { "epoch": 6.49, "learning_rate": 5.8852058517404875e-05, "loss": 1.4644, "step": 74970 }, { "epoch": 6.49, "learning_rate": 5.884989013530704e-05, "loss": 1.4913, "step": 74980 }, { "epoch": 6.49, "learning_rate": 5.88477217532092e-05, "loss": 1.349, "step": 74990 }, { "epoch": 6.5, "learning_rate": 5.884555337111136e-05, "loss": 1.414, "step": 75000 }, { "epoch": 6.5, "learning_rate": 5.884338498901352e-05, "loss": 1.5467, "step": 75010 }, { "epoch": 6.5, "learning_rate": 5.884121660691569e-05, "loss": 1.3948, "step": 75020 }, { "epoch": 6.5, "learning_rate": 5.883904822481785e-05, "loss": 1.5162, "step": 75030 }, { "epoch": 6.5, "learning_rate": 5.883687984272001e-05, "loss": 1.588, "step": 75040 }, { "epoch": 6.5, "learning_rate": 5.883471146062218e-05, "loss": 1.4798, "step": 75050 }, { "epoch": 6.5, "learning_rate": 5.883254307852434e-05, "loss": 1.4901, "step": 75060 }, { "epoch": 6.5, "learning_rate": 5.8830374696426505e-05, "loss": 1.5212, "step": 75070 }, { "epoch": 6.5, "learning_rate": 5.882820631432866e-05, "loss": 1.5305, "step": 75080 }, { "epoch": 6.5, "learning_rate": 5.8826037932230825e-05, "loss": 1.3041, "step": 75090 }, { "epoch": 6.5, "learning_rate": 5.882386955013299e-05, "loss": 1.619, "step": 75100 }, { "epoch": 6.51, "learning_rate": 5.882170116803515e-05, "loss": 1.2988, "step": 75110 }, { "epoch": 6.51, "learning_rate": 5.881953278593732e-05, "loss": 1.5581, "step": 75120 }, { "epoch": 6.51, "learning_rate": 5.881736440383948e-05, "loss": 1.5443, "step": 75130 }, { "epoch": 6.51, "learning_rate": 5.881519602174164e-05, "loss": 1.5663, "step": 75140 }, { "epoch": 6.51, "learning_rate": 5.881302763964381e-05, "loss": 1.4388, "step": 75150 }, { "epoch": 6.51, "learning_rate": 5.881085925754596e-05, "loss": 1.5167, "step": 75160 }, { "epoch": 6.51, "learning_rate": 5.880869087544812e-05, "loss": 1.3486, "step": 75170 }, { "epoch": 6.51, "learning_rate": 5.880652249335029e-05, "loss": 1.2974, "step": 75180 }, { "epoch": 6.51, "learning_rate": 5.8804354111252455e-05, "loss": 1.4463, "step": 75190 }, { "epoch": 6.51, "learning_rate": 5.8802185729154614e-05, "loss": 1.3229, "step": 75200 }, { "epoch": 6.51, "learning_rate": 5.880001734705678e-05, "loss": 1.4729, "step": 75210 }, { "epoch": 6.51, "learning_rate": 5.879784896495894e-05, "loss": 1.6005, "step": 75220 }, { "epoch": 6.52, "learning_rate": 5.87956805828611e-05, "loss": 1.4068, "step": 75230 }, { "epoch": 6.52, "learning_rate": 5.879351220076326e-05, "loss": 1.6131, "step": 75240 }, { "epoch": 6.52, "learning_rate": 5.8791343818665426e-05, "loss": 1.5158, "step": 75250 }, { "epoch": 6.52, "learning_rate": 5.878917543656759e-05, "loss": 1.4832, "step": 75260 }, { "epoch": 6.52, "learning_rate": 5.878700705446975e-05, "loss": 1.4541, "step": 75270 }, { "epoch": 6.52, "learning_rate": 5.878483867237192e-05, "loss": 1.6023, "step": 75280 }, { "epoch": 6.52, "learning_rate": 5.878267029027408e-05, "loss": 1.6707, "step": 75290 }, { "epoch": 6.52, "learning_rate": 5.8780501908176244e-05, "loss": 1.7977, "step": 75300 }, { "epoch": 6.52, "learning_rate": 5.87783335260784e-05, "loss": 1.2525, "step": 75310 }, { "epoch": 6.52, "learning_rate": 5.8776165143980564e-05, "loss": 1.4098, "step": 75320 }, { "epoch": 6.52, "learning_rate": 5.877399676188273e-05, "loss": 1.3288, "step": 75330 }, { "epoch": 6.52, "learning_rate": 5.877182837978489e-05, "loss": 1.4153, "step": 75340 }, { "epoch": 6.53, "learning_rate": 5.8769659997687056e-05, "loss": 1.3928, "step": 75350 }, { "epoch": 6.53, "learning_rate": 5.8767491615589216e-05, "loss": 1.5615, "step": 75360 }, { "epoch": 6.53, "learning_rate": 5.876532323349138e-05, "loss": 1.4811, "step": 75370 }, { "epoch": 6.53, "learning_rate": 5.876315485139355e-05, "loss": 1.3225, "step": 75380 }, { "epoch": 6.53, "learning_rate": 5.87609864692957e-05, "loss": 1.5688, "step": 75390 }, { "epoch": 6.53, "learning_rate": 5.875881808719787e-05, "loss": 1.4201, "step": 75400 }, { "epoch": 6.53, "learning_rate": 5.875664970510003e-05, "loss": 1.483, "step": 75410 }, { "epoch": 6.53, "learning_rate": 5.8754481323002194e-05, "loss": 1.4099, "step": 75420 }, { "epoch": 6.53, "learning_rate": 5.875231294090435e-05, "loss": 1.6481, "step": 75430 }, { "epoch": 6.53, "learning_rate": 5.875014455880652e-05, "loss": 1.6404, "step": 75440 }, { "epoch": 6.53, "learning_rate": 5.8747976176708686e-05, "loss": 1.5037, "step": 75450 }, { "epoch": 6.54, "learning_rate": 5.874580779461084e-05, "loss": 1.6931, "step": 75460 }, { "epoch": 6.54, "learning_rate": 5.8743639412513005e-05, "loss": 1.5126, "step": 75470 }, { "epoch": 6.54, "learning_rate": 5.8741471030415165e-05, "loss": 1.6499, "step": 75480 }, { "epoch": 6.54, "learning_rate": 5.873930264831733e-05, "loss": 1.6047, "step": 75490 }, { "epoch": 6.54, "learning_rate": 5.873713426621949e-05, "loss": 1.6904, "step": 75500 }, { "epoch": 6.54, "learning_rate": 5.873496588412166e-05, "loss": 1.362, "step": 75510 }, { "epoch": 6.54, "learning_rate": 5.8732797502023824e-05, "loss": 1.4602, "step": 75520 }, { "epoch": 6.54, "learning_rate": 5.873062911992598e-05, "loss": 1.552, "step": 75530 }, { "epoch": 6.54, "learning_rate": 5.872846073782814e-05, "loss": 1.1756, "step": 75540 }, { "epoch": 6.54, "learning_rate": 5.87262923557303e-05, "loss": 1.4937, "step": 75550 }, { "epoch": 6.54, "learning_rate": 5.872412397363247e-05, "loss": 1.4055, "step": 75560 }, { "epoch": 6.54, "learning_rate": 5.872195559153463e-05, "loss": 1.5497, "step": 75570 }, { "epoch": 6.55, "learning_rate": 5.8719787209436795e-05, "loss": 1.6964, "step": 75580 }, { "epoch": 6.55, "learning_rate": 5.871761882733896e-05, "loss": 1.8394, "step": 75590 }, { "epoch": 6.55, "learning_rate": 5.871545044524112e-05, "loss": 1.4406, "step": 75600 }, { "epoch": 6.55, "learning_rate": 5.871328206314329e-05, "loss": 1.4305, "step": 75610 }, { "epoch": 6.55, "learning_rate": 5.871111368104544e-05, "loss": 1.3764, "step": 75620 }, { "epoch": 6.55, "learning_rate": 5.8708945298947606e-05, "loss": 1.5238, "step": 75630 }, { "epoch": 6.55, "learning_rate": 5.8706776916849766e-05, "loss": 1.6489, "step": 75640 }, { "epoch": 6.55, "learning_rate": 5.870460853475193e-05, "loss": 1.5099, "step": 75650 }, { "epoch": 6.55, "learning_rate": 5.87024401526541e-05, "loss": 1.4241, "step": 75660 }, { "epoch": 6.55, "learning_rate": 5.870027177055626e-05, "loss": 1.7258, "step": 75670 }, { "epoch": 6.55, "learning_rate": 5.8698103388458425e-05, "loss": 1.6341, "step": 75680 }, { "epoch": 6.56, "learning_rate": 5.869593500636058e-05, "loss": 1.3491, "step": 75690 }, { "epoch": 6.56, "learning_rate": 5.8693766624262744e-05, "loss": 1.2936, "step": 75700 }, { "epoch": 6.56, "learning_rate": 5.8691598242164904e-05, "loss": 1.5458, "step": 75710 }, { "epoch": 6.56, "learning_rate": 5.868942986006707e-05, "loss": 1.4025, "step": 75720 }, { "epoch": 6.56, "learning_rate": 5.868726147796923e-05, "loss": 1.5373, "step": 75730 }, { "epoch": 6.56, "learning_rate": 5.8685093095871396e-05, "loss": 1.1757, "step": 75740 }, { "epoch": 6.56, "learning_rate": 5.868292471377356e-05, "loss": 1.4624, "step": 75750 }, { "epoch": 6.56, "learning_rate": 5.868075633167572e-05, "loss": 1.3865, "step": 75760 }, { "epoch": 6.56, "learning_rate": 5.867858794957788e-05, "loss": 1.5611, "step": 75770 }, { "epoch": 6.56, "learning_rate": 5.867641956748004e-05, "loss": 1.3852, "step": 75780 }, { "epoch": 6.56, "learning_rate": 5.867425118538221e-05, "loss": 1.365, "step": 75790 }, { "epoch": 6.56, "learning_rate": 5.867208280328437e-05, "loss": 1.4428, "step": 75800 }, { "epoch": 6.57, "learning_rate": 5.8669914421186534e-05, "loss": 1.6334, "step": 75810 }, { "epoch": 6.57, "learning_rate": 5.86677460390887e-05, "loss": 1.4586, "step": 75820 }, { "epoch": 6.57, "learning_rate": 5.866557765699086e-05, "loss": 1.7234, "step": 75830 }, { "epoch": 6.57, "learning_rate": 5.8663409274893026e-05, "loss": 1.475, "step": 75840 }, { "epoch": 6.57, "learning_rate": 5.866124089279518e-05, "loss": 1.4993, "step": 75850 }, { "epoch": 6.57, "learning_rate": 5.8659072510697345e-05, "loss": 1.5228, "step": 75860 }, { "epoch": 6.57, "learning_rate": 5.8656904128599505e-05, "loss": 1.433, "step": 75870 }, { "epoch": 6.57, "learning_rate": 5.865473574650167e-05, "loss": 1.49, "step": 75880 }, { "epoch": 6.57, "learning_rate": 5.865256736440384e-05, "loss": 1.7876, "step": 75890 }, { "epoch": 6.57, "learning_rate": 5.8650398982306e-05, "loss": 1.3024, "step": 75900 }, { "epoch": 6.57, "learning_rate": 5.8648230600208164e-05, "loss": 1.689, "step": 75910 }, { "epoch": 6.58, "learning_rate": 5.8646062218110323e-05, "loss": 1.4705, "step": 75920 }, { "epoch": 6.58, "learning_rate": 5.864389383601248e-05, "loss": 1.5328, "step": 75930 }, { "epoch": 6.58, "learning_rate": 5.864172545391464e-05, "loss": 1.4856, "step": 75940 }, { "epoch": 6.58, "learning_rate": 5.863955707181681e-05, "loss": 1.5442, "step": 75950 }, { "epoch": 6.58, "learning_rate": 5.8637388689718975e-05, "loss": 1.371, "step": 75960 }, { "epoch": 6.58, "learning_rate": 5.8635220307621135e-05, "loss": 1.5631, "step": 75970 }, { "epoch": 6.58, "learning_rate": 5.86330519255233e-05, "loss": 1.3585, "step": 75980 }, { "epoch": 6.58, "learning_rate": 5.863088354342546e-05, "loss": 1.5033, "step": 75990 }, { "epoch": 6.58, "learning_rate": 5.862871516132762e-05, "loss": 1.4005, "step": 76000 }, { "epoch": 6.58, "learning_rate": 5.862654677922978e-05, "loss": 1.3541, "step": 76010 }, { "epoch": 6.58, "learning_rate": 5.862437839713195e-05, "loss": 1.4763, "step": 76020 }, { "epoch": 6.58, "learning_rate": 5.862221001503411e-05, "loss": 1.6743, "step": 76030 }, { "epoch": 6.59, "learning_rate": 5.862004163293627e-05, "loss": 1.6424, "step": 76040 }, { "epoch": 6.59, "learning_rate": 5.861787325083844e-05, "loss": 1.4136, "step": 76050 }, { "epoch": 6.59, "learning_rate": 5.86157048687406e-05, "loss": 1.7141, "step": 76060 }, { "epoch": 6.59, "learning_rate": 5.8613536486642765e-05, "loss": 1.3266, "step": 76070 }, { "epoch": 6.59, "learning_rate": 5.861136810454492e-05, "loss": 1.5382, "step": 76080 }, { "epoch": 6.59, "learning_rate": 5.8609199722447084e-05, "loss": 1.5786, "step": 76090 }, { "epoch": 6.59, "learning_rate": 5.860703134034925e-05, "loss": 1.7006, "step": 76100 }, { "epoch": 6.59, "learning_rate": 5.860486295825141e-05, "loss": 1.5139, "step": 76110 }, { "epoch": 6.59, "learning_rate": 5.860269457615358e-05, "loss": 1.4244, "step": 76120 }, { "epoch": 6.59, "learning_rate": 5.8600526194055736e-05, "loss": 1.7443, "step": 76130 }, { "epoch": 6.59, "learning_rate": 5.85983578119579e-05, "loss": 1.6667, "step": 76140 }, { "epoch": 6.6, "learning_rate": 5.859618942986007e-05, "loss": 1.3768, "step": 76150 }, { "epoch": 6.6, "learning_rate": 5.859402104776222e-05, "loss": 1.5192, "step": 76160 }, { "epoch": 6.6, "learning_rate": 5.859185266566439e-05, "loss": 1.3035, "step": 76170 }, { "epoch": 6.6, "learning_rate": 5.858968428356655e-05, "loss": 1.6735, "step": 76180 }, { "epoch": 6.6, "learning_rate": 5.8587515901468714e-05, "loss": 1.384, "step": 76190 }, { "epoch": 6.6, "learning_rate": 5.8585347519370874e-05, "loss": 1.368, "step": 76200 }, { "epoch": 6.6, "learning_rate": 5.858317913727304e-05, "loss": 1.4323, "step": 76210 }, { "epoch": 6.6, "learning_rate": 5.858101075517521e-05, "loss": 1.6919, "step": 76220 }, { "epoch": 6.6, "learning_rate": 5.857884237307736e-05, "loss": 1.5481, "step": 76230 }, { "epoch": 6.6, "learning_rate": 5.857667399097952e-05, "loss": 1.5129, "step": 76240 }, { "epoch": 6.6, "learning_rate": 5.8574505608881686e-05, "loss": 1.3458, "step": 76250 }, { "epoch": 6.6, "learning_rate": 5.857233722678385e-05, "loss": 1.4089, "step": 76260 }, { "epoch": 6.61, "learning_rate": 5.857016884468601e-05, "loss": 1.6068, "step": 76270 }, { "epoch": 6.61, "learning_rate": 5.856800046258818e-05, "loss": 1.5165, "step": 76280 }, { "epoch": 6.61, "learning_rate": 5.856583208049034e-05, "loss": 1.4178, "step": 76290 }, { "epoch": 6.61, "learning_rate": 5.8563663698392504e-05, "loss": 1.3845, "step": 76300 }, { "epoch": 6.61, "learning_rate": 5.856149531629466e-05, "loss": 1.7984, "step": 76310 }, { "epoch": 6.61, "learning_rate": 5.855932693419682e-05, "loss": 1.8013, "step": 76320 }, { "epoch": 6.61, "learning_rate": 5.855715855209899e-05, "loss": 1.4943, "step": 76330 }, { "epoch": 6.61, "learning_rate": 5.855499017000115e-05, "loss": 1.5042, "step": 76340 }, { "epoch": 6.61, "learning_rate": 5.8552821787903316e-05, "loss": 1.5505, "step": 76350 }, { "epoch": 6.61, "learning_rate": 5.8550653405805475e-05, "loss": 1.2729, "step": 76360 }, { "epoch": 6.61, "learning_rate": 5.854848502370764e-05, "loss": 1.6014, "step": 76370 }, { "epoch": 6.61, "learning_rate": 5.854631664160981e-05, "loss": 1.4208, "step": 76380 }, { "epoch": 6.62, "learning_rate": 5.854414825951196e-05, "loss": 1.5515, "step": 76390 }, { "epoch": 6.62, "learning_rate": 5.854197987741413e-05, "loss": 1.4251, "step": 76400 }, { "epoch": 6.62, "learning_rate": 5.853981149531629e-05, "loss": 1.5031, "step": 76410 }, { "epoch": 6.62, "learning_rate": 5.853764311321845e-05, "loss": 1.6102, "step": 76420 }, { "epoch": 6.62, "learning_rate": 5.853547473112061e-05, "loss": 1.5641, "step": 76430 }, { "epoch": 6.62, "learning_rate": 5.853330634902278e-05, "loss": 1.4509, "step": 76440 }, { "epoch": 6.62, "learning_rate": 5.8531137966924946e-05, "loss": 1.4769, "step": 76450 }, { "epoch": 6.62, "learning_rate": 5.85289695848271e-05, "loss": 1.2758, "step": 76460 }, { "epoch": 6.62, "learning_rate": 5.8526801202729265e-05, "loss": 1.4248, "step": 76470 }, { "epoch": 6.62, "learning_rate": 5.8524632820631425e-05, "loss": 1.4739, "step": 76480 }, { "epoch": 6.62, "learning_rate": 5.852246443853359e-05, "loss": 1.3982, "step": 76490 }, { "epoch": 6.63, "learning_rate": 5.852029605643575e-05, "loss": 1.5819, "step": 76500 }, { "epoch": 6.63, "learning_rate": 5.851812767433792e-05, "loss": 1.5118, "step": 76510 }, { "epoch": 6.63, "learning_rate": 5.851595929224008e-05, "loss": 1.0504, "step": 76520 }, { "epoch": 6.63, "learning_rate": 5.851379091014224e-05, "loss": 1.4975, "step": 76530 }, { "epoch": 6.63, "learning_rate": 5.85116225280444e-05, "loss": 1.732, "step": 76540 }, { "epoch": 6.63, "learning_rate": 5.850945414594656e-05, "loss": 1.6172, "step": 76550 }, { "epoch": 6.63, "learning_rate": 5.850728576384873e-05, "loss": 1.2577, "step": 76560 }, { "epoch": 6.63, "learning_rate": 5.850511738175089e-05, "loss": 1.218, "step": 76570 }, { "epoch": 6.63, "learning_rate": 5.8502948999653055e-05, "loss": 1.5782, "step": 76580 }, { "epoch": 6.63, "learning_rate": 5.850078061755522e-05, "loss": 1.5034, "step": 76590 }, { "epoch": 6.63, "learning_rate": 5.849861223545738e-05, "loss": 1.4769, "step": 76600 }, { "epoch": 6.63, "learning_rate": 5.849644385335955e-05, "loss": 1.3562, "step": 76610 }, { "epoch": 6.64, "learning_rate": 5.84942754712617e-05, "loss": 1.2647, "step": 76620 }, { "epoch": 6.64, "learning_rate": 5.8492107089163866e-05, "loss": 1.7523, "step": 76630 }, { "epoch": 6.64, "learning_rate": 5.8489938707066026e-05, "loss": 1.3832, "step": 76640 }, { "epoch": 6.64, "learning_rate": 5.848777032496819e-05, "loss": 1.4529, "step": 76650 }, { "epoch": 6.64, "learning_rate": 5.848560194287036e-05, "loss": 1.2085, "step": 76660 }, { "epoch": 6.64, "learning_rate": 5.848343356077252e-05, "loss": 1.4107, "step": 76670 }, { "epoch": 6.64, "learning_rate": 5.8481265178674685e-05, "loss": 1.4846, "step": 76680 }, { "epoch": 6.64, "learning_rate": 5.8479096796576844e-05, "loss": 1.5402, "step": 76690 }, { "epoch": 6.64, "learning_rate": 5.8476928414479004e-05, "loss": 1.3503, "step": 76700 }, { "epoch": 6.64, "learning_rate": 5.8474760032381163e-05, "loss": 1.4517, "step": 76710 }, { "epoch": 6.64, "learning_rate": 5.847259165028333e-05, "loss": 1.4079, "step": 76720 }, { "epoch": 6.65, "learning_rate": 5.8470423268185496e-05, "loss": 1.4846, "step": 76730 }, { "epoch": 6.65, "learning_rate": 5.8468254886087656e-05, "loss": 1.4444, "step": 76740 }, { "epoch": 6.65, "learning_rate": 5.846608650398982e-05, "loss": 1.6612, "step": 76750 }, { "epoch": 6.65, "learning_rate": 5.846391812189198e-05, "loss": 1.4243, "step": 76760 }, { "epoch": 6.65, "learning_rate": 5.846174973979414e-05, "loss": 1.3907, "step": 76770 }, { "epoch": 6.65, "learning_rate": 5.84595813576963e-05, "loss": 1.3856, "step": 76780 }, { "epoch": 6.65, "learning_rate": 5.845741297559847e-05, "loss": 1.3994, "step": 76790 }, { "epoch": 6.65, "learning_rate": 5.845524459350063e-05, "loss": 1.5518, "step": 76800 }, { "epoch": 6.65, "learning_rate": 5.8453076211402793e-05, "loss": 1.533, "step": 76810 }, { "epoch": 6.65, "learning_rate": 5.845090782930496e-05, "loss": 1.3774, "step": 76820 }, { "epoch": 6.65, "learning_rate": 5.844873944720712e-05, "loss": 1.5465, "step": 76830 }, { "epoch": 6.65, "learning_rate": 5.8446571065109286e-05, "loss": 1.3943, "step": 76840 }, { "epoch": 6.66, "learning_rate": 5.844440268301144e-05, "loss": 1.5438, "step": 76850 }, { "epoch": 6.66, "learning_rate": 5.8442234300913605e-05, "loss": 1.5761, "step": 76860 }, { "epoch": 6.66, "learning_rate": 5.8440065918815765e-05, "loss": 1.4976, "step": 76870 }, { "epoch": 6.66, "learning_rate": 5.843789753671793e-05, "loss": 1.3985, "step": 76880 }, { "epoch": 6.66, "learning_rate": 5.84357291546201e-05, "loss": 1.4338, "step": 76890 }, { "epoch": 6.66, "learning_rate": 5.843356077252226e-05, "loss": 1.4384, "step": 76900 }, { "epoch": 6.66, "learning_rate": 5.8431392390424423e-05, "loss": 1.4228, "step": 76910 }, { "epoch": 6.66, "learning_rate": 5.842922400832658e-05, "loss": 1.8035, "step": 76920 }, { "epoch": 6.66, "learning_rate": 5.842705562622874e-05, "loss": 1.6253, "step": 76930 }, { "epoch": 6.66, "learning_rate": 5.84248872441309e-05, "loss": 1.4391, "step": 76940 }, { "epoch": 6.66, "learning_rate": 5.842271886203307e-05, "loss": 1.5097, "step": 76950 }, { "epoch": 6.67, "learning_rate": 5.8420550479935235e-05, "loss": 1.6569, "step": 76960 }, { "epoch": 6.67, "learning_rate": 5.8418382097837395e-05, "loss": 1.3445, "step": 76970 }, { "epoch": 6.67, "learning_rate": 5.841621371573956e-05, "loss": 1.5257, "step": 76980 }, { "epoch": 6.67, "learning_rate": 5.841404533364172e-05, "loss": 1.2418, "step": 76990 }, { "epoch": 6.67, "learning_rate": 5.841187695154388e-05, "loss": 1.4352, "step": 77000 }, { "epoch": 6.67, "learning_rate": 5.840970856944604e-05, "loss": 1.4744, "step": 77010 }, { "epoch": 6.67, "learning_rate": 5.8407540187348206e-05, "loss": 1.4353, "step": 77020 }, { "epoch": 6.67, "learning_rate": 5.840537180525037e-05, "loss": 1.4472, "step": 77030 }, { "epoch": 6.67, "learning_rate": 5.840320342315253e-05, "loss": 1.689, "step": 77040 }, { "epoch": 6.67, "learning_rate": 5.84010350410547e-05, "loss": 1.6611, "step": 77050 }, { "epoch": 6.67, "learning_rate": 5.839886665895686e-05, "loss": 1.4624, "step": 77060 }, { "epoch": 6.67, "learning_rate": 5.8396698276859025e-05, "loss": 1.2986, "step": 77070 }, { "epoch": 6.68, "learning_rate": 5.839452989476118e-05, "loss": 1.3087, "step": 77080 }, { "epoch": 6.68, "learning_rate": 5.8392361512663344e-05, "loss": 1.4628, "step": 77090 }, { "epoch": 6.68, "learning_rate": 5.839019313056551e-05, "loss": 1.5291, "step": 77100 }, { "epoch": 6.68, "learning_rate": 5.838802474846767e-05, "loss": 1.397, "step": 77110 }, { "epoch": 6.68, "learning_rate": 5.8385856366369836e-05, "loss": 1.6041, "step": 77120 }, { "epoch": 6.68, "learning_rate": 5.8383687984271996e-05, "loss": 1.7194, "step": 77130 }, { "epoch": 6.68, "learning_rate": 5.838151960217416e-05, "loss": 1.4437, "step": 77140 }, { "epoch": 6.68, "learning_rate": 5.837935122007633e-05, "loss": 1.6523, "step": 77150 }, { "epoch": 6.68, "learning_rate": 5.837718283797848e-05, "loss": 1.3922, "step": 77160 }, { "epoch": 6.68, "learning_rate": 5.837501445588065e-05, "loss": 1.5057, "step": 77170 }, { "epoch": 6.68, "learning_rate": 5.837284607378281e-05, "loss": 1.418, "step": 77180 }, { "epoch": 6.69, "learning_rate": 5.8370677691684974e-05, "loss": 1.6201, "step": 77190 }, { "epoch": 6.69, "learning_rate": 5.8368509309587134e-05, "loss": 1.4859, "step": 77200 }, { "epoch": 6.69, "learning_rate": 5.83663409274893e-05, "loss": 1.5599, "step": 77210 }, { "epoch": 6.69, "learning_rate": 5.8364172545391466e-05, "loss": 1.4048, "step": 77220 }, { "epoch": 6.69, "learning_rate": 5.836200416329362e-05, "loss": 1.2796, "step": 77230 }, { "epoch": 6.69, "learning_rate": 5.8359835781195786e-05, "loss": 1.3353, "step": 77240 }, { "epoch": 6.69, "learning_rate": 5.8357667399097945e-05, "loss": 1.621, "step": 77250 }, { "epoch": 6.69, "learning_rate": 5.835549901700011e-05, "loss": 1.5868, "step": 77260 }, { "epoch": 6.69, "learning_rate": 5.835333063490227e-05, "loss": 1.5943, "step": 77270 }, { "epoch": 6.69, "learning_rate": 5.835116225280444e-05, "loss": 1.6483, "step": 77280 }, { "epoch": 6.69, "learning_rate": 5.8348993870706604e-05, "loss": 1.3064, "step": 77290 }, { "epoch": 6.69, "learning_rate": 5.8346825488608764e-05, "loss": 1.4139, "step": 77300 }, { "epoch": 6.7, "learning_rate": 5.8344657106510917e-05, "loss": 1.5614, "step": 77310 }, { "epoch": 6.7, "learning_rate": 5.834248872441308e-05, "loss": 1.4968, "step": 77320 }, { "epoch": 6.7, "learning_rate": 5.834032034231525e-05, "loss": 1.2283, "step": 77330 }, { "epoch": 6.7, "learning_rate": 5.833815196021741e-05, "loss": 1.3231, "step": 77340 }, { "epoch": 6.7, "learning_rate": 5.8335983578119575e-05, "loss": 1.4517, "step": 77350 }, { "epoch": 6.7, "learning_rate": 5.8333815196021735e-05, "loss": 1.2032, "step": 77360 }, { "epoch": 6.7, "learning_rate": 5.83316468139239e-05, "loss": 1.3599, "step": 77370 }, { "epoch": 6.7, "learning_rate": 5.832947843182607e-05, "loss": 1.2502, "step": 77380 }, { "epoch": 6.7, "learning_rate": 5.832731004972822e-05, "loss": 1.2, "step": 77390 }, { "epoch": 6.7, "learning_rate": 5.832514166763039e-05, "loss": 1.6539, "step": 77400 }, { "epoch": 6.7, "learning_rate": 5.8322973285532547e-05, "loss": 1.6526, "step": 77410 }, { "epoch": 6.71, "learning_rate": 5.832080490343471e-05, "loss": 1.637, "step": 77420 }, { "epoch": 6.71, "learning_rate": 5.831863652133687e-05, "loss": 1.5353, "step": 77430 }, { "epoch": 6.71, "learning_rate": 5.831646813923904e-05, "loss": 1.4604, "step": 77440 }, { "epoch": 6.71, "learning_rate": 5.8314299757141205e-05, "loss": 1.5801, "step": 77450 }, { "epoch": 6.71, "learning_rate": 5.8312131375043365e-05, "loss": 1.4567, "step": 77460 }, { "epoch": 6.71, "learning_rate": 5.8309962992945525e-05, "loss": 1.4909, "step": 77470 }, { "epoch": 6.71, "learning_rate": 5.8307794610847684e-05, "loss": 1.3951, "step": 77480 }, { "epoch": 6.71, "learning_rate": 5.830562622874985e-05, "loss": 1.5053, "step": 77490 }, { "epoch": 6.71, "learning_rate": 5.830345784665201e-05, "loss": 1.6602, "step": 77500 }, { "epoch": 6.71, "learning_rate": 5.8301289464554177e-05, "loss": 1.3464, "step": 77510 }, { "epoch": 6.71, "learning_rate": 5.829912108245634e-05, "loss": 1.5013, "step": 77520 }, { "epoch": 6.71, "learning_rate": 5.82969527003585e-05, "loss": 1.6544, "step": 77530 }, { "epoch": 6.72, "learning_rate": 5.829478431826066e-05, "loss": 1.7521, "step": 77540 }, { "epoch": 6.72, "learning_rate": 5.829261593616282e-05, "loss": 1.373, "step": 77550 }, { "epoch": 6.72, "learning_rate": 5.829044755406499e-05, "loss": 1.5501, "step": 77560 }, { "epoch": 6.72, "learning_rate": 5.828827917196715e-05, "loss": 1.7489, "step": 77570 }, { "epoch": 6.72, "learning_rate": 5.8286110789869314e-05, "loss": 1.5183, "step": 77580 }, { "epoch": 6.72, "learning_rate": 5.828394240777148e-05, "loss": 1.5596, "step": 77590 }, { "epoch": 6.72, "learning_rate": 5.828177402567364e-05, "loss": 1.6488, "step": 77600 }, { "epoch": 6.72, "learning_rate": 5.827960564357581e-05, "loss": 1.351, "step": 77610 }, { "epoch": 6.72, "learning_rate": 5.827743726147796e-05, "loss": 1.5779, "step": 77620 }, { "epoch": 6.72, "learning_rate": 5.8275268879380126e-05, "loss": 1.2229, "step": 77630 }, { "epoch": 6.72, "learning_rate": 5.8273100497282285e-05, "loss": 1.4581, "step": 77640 }, { "epoch": 6.72, "learning_rate": 5.827093211518445e-05, "loss": 1.3626, "step": 77650 }, { "epoch": 6.73, "learning_rate": 5.826876373308662e-05, "loss": 1.3413, "step": 77660 }, { "epoch": 6.73, "learning_rate": 5.826659535098878e-05, "loss": 1.4233, "step": 77670 }, { "epoch": 6.73, "learning_rate": 5.8264426968890944e-05, "loss": 1.4951, "step": 77680 }, { "epoch": 6.73, "learning_rate": 5.8262258586793104e-05, "loss": 1.3938, "step": 77690 }, { "epoch": 6.73, "learning_rate": 5.8260090204695264e-05, "loss": 1.3577, "step": 77700 }, { "epoch": 6.73, "learning_rate": 5.825792182259742e-05, "loss": 1.4661, "step": 77710 }, { "epoch": 6.73, "learning_rate": 5.825575344049959e-05, "loss": 1.4391, "step": 77720 }, { "epoch": 6.73, "learning_rate": 5.8253585058401756e-05, "loss": 1.6134, "step": 77730 }, { "epoch": 6.73, "learning_rate": 5.8251416676303916e-05, "loss": 1.4175, "step": 77740 }, { "epoch": 6.73, "learning_rate": 5.824924829420608e-05, "loss": 1.5909, "step": 77750 }, { "epoch": 6.73, "learning_rate": 5.824707991210824e-05, "loss": 1.5236, "step": 77760 }, { "epoch": 6.74, "learning_rate": 5.82449115300104e-05, "loss": 1.5183, "step": 77770 }, { "epoch": 6.74, "learning_rate": 5.824274314791256e-05, "loss": 1.6101, "step": 77780 }, { "epoch": 6.74, "learning_rate": 5.824057476581473e-05, "loss": 1.5388, "step": 77790 }, { "epoch": 6.74, "learning_rate": 5.8238406383716894e-05, "loss": 1.6591, "step": 77800 }, { "epoch": 6.74, "learning_rate": 5.823623800161905e-05, "loss": 1.246, "step": 77810 }, { "epoch": 6.74, "learning_rate": 5.823406961952122e-05, "loss": 1.3205, "step": 77820 }, { "epoch": 6.74, "learning_rate": 5.823190123742338e-05, "loss": 1.5752, "step": 77830 }, { "epoch": 6.74, "learning_rate": 5.8229732855325546e-05, "loss": 1.6831, "step": 77840 }, { "epoch": 6.74, "learning_rate": 5.82275644732277e-05, "loss": 1.7005, "step": 77850 }, { "epoch": 6.74, "learning_rate": 5.8225396091129865e-05, "loss": 1.3168, "step": 77860 }, { "epoch": 6.74, "learning_rate": 5.8223227709032024e-05, "loss": 1.5122, "step": 77870 }, { "epoch": 6.74, "learning_rate": 5.822105932693419e-05, "loss": 1.6864, "step": 77880 }, { "epoch": 6.75, "learning_rate": 5.821889094483636e-05, "loss": 1.4925, "step": 77890 }, { "epoch": 6.75, "learning_rate": 5.821672256273852e-05, "loss": 1.5723, "step": 77900 }, { "epoch": 6.75, "learning_rate": 5.821455418064068e-05, "loss": 1.5384, "step": 77910 }, { "epoch": 6.75, "learning_rate": 5.821238579854284e-05, "loss": 1.4453, "step": 77920 }, { "epoch": 6.75, "learning_rate": 5.8210217416445e-05, "loss": 1.6324, "step": 77930 }, { "epoch": 6.75, "learning_rate": 5.820804903434716e-05, "loss": 1.4348, "step": 77940 }, { "epoch": 6.75, "learning_rate": 5.820588065224933e-05, "loss": 1.4038, "step": 77950 }, { "epoch": 6.75, "learning_rate": 5.8203712270151495e-05, "loss": 1.6749, "step": 77960 }, { "epoch": 6.75, "learning_rate": 5.8201543888053654e-05, "loss": 1.4454, "step": 77970 }, { "epoch": 6.75, "learning_rate": 5.819937550595582e-05, "loss": 1.5957, "step": 77980 }, { "epoch": 6.75, "learning_rate": 5.819720712385798e-05, "loss": 1.4977, "step": 77990 }, { "epoch": 6.76, "learning_rate": 5.819503874176014e-05, "loss": 1.6374, "step": 78000 }, { "epoch": 6.76, "learning_rate": 5.81928703596623e-05, "loss": 1.4853, "step": 78010 }, { "epoch": 6.76, "learning_rate": 5.8190701977564466e-05, "loss": 1.4442, "step": 78020 }, { "epoch": 6.76, "learning_rate": 5.818853359546663e-05, "loss": 1.3702, "step": 78030 }, { "epoch": 6.76, "learning_rate": 5.818636521336879e-05, "loss": 1.3491, "step": 78040 }, { "epoch": 6.76, "learning_rate": 5.818419683127096e-05, "loss": 1.4941, "step": 78050 }, { "epoch": 6.76, "learning_rate": 5.818202844917312e-05, "loss": 1.4546, "step": 78060 }, { "epoch": 6.76, "learning_rate": 5.8179860067075284e-05, "loss": 1.5792, "step": 78070 }, { "epoch": 6.76, "learning_rate": 5.817769168497744e-05, "loss": 1.5327, "step": 78080 }, { "epoch": 6.76, "learning_rate": 5.8175523302879604e-05, "loss": 1.6631, "step": 78090 }, { "epoch": 6.76, "learning_rate": 5.817335492078177e-05, "loss": 1.4098, "step": 78100 }, { "epoch": 6.76, "learning_rate": 5.817118653868393e-05, "loss": 1.5871, "step": 78110 }, { "epoch": 6.77, "learning_rate": 5.8169018156586096e-05, "loss": 1.6992, "step": 78120 }, { "epoch": 6.77, "learning_rate": 5.8166849774488256e-05, "loss": 1.5791, "step": 78130 }, { "epoch": 6.77, "learning_rate": 5.816468139239042e-05, "loss": 1.3117, "step": 78140 }, { "epoch": 6.77, "learning_rate": 5.816251301029259e-05, "loss": 1.4613, "step": 78150 }, { "epoch": 6.77, "learning_rate": 5.816034462819474e-05, "loss": 1.6346, "step": 78160 }, { "epoch": 6.77, "learning_rate": 5.815817624609691e-05, "loss": 1.5, "step": 78170 }, { "epoch": 6.77, "learning_rate": 5.815600786399907e-05, "loss": 1.4807, "step": 78180 }, { "epoch": 6.77, "learning_rate": 5.8153839481901234e-05, "loss": 1.5516, "step": 78190 }, { "epoch": 6.77, "learning_rate": 5.815167109980339e-05, "loss": 1.4601, "step": 78200 }, { "epoch": 6.77, "learning_rate": 5.814950271770556e-05, "loss": 1.5742, "step": 78210 }, { "epoch": 6.77, "learning_rate": 5.8147334335607726e-05, "loss": 1.6015, "step": 78220 }, { "epoch": 6.78, "learning_rate": 5.8145165953509886e-05, "loss": 1.5809, "step": 78230 }, { "epoch": 6.78, "learning_rate": 5.8142997571412045e-05, "loss": 1.5118, "step": 78240 }, { "epoch": 6.78, "learning_rate": 5.8140829189314205e-05, "loss": 1.4817, "step": 78250 }, { "epoch": 6.78, "learning_rate": 5.813866080721637e-05, "loss": 1.5044, "step": 78260 }, { "epoch": 6.78, "learning_rate": 5.813649242511853e-05, "loss": 1.4855, "step": 78270 }, { "epoch": 6.78, "learning_rate": 5.81343240430207e-05, "loss": 1.3851, "step": 78280 }, { "epoch": 6.78, "learning_rate": 5.8132155660922864e-05, "loss": 1.3918, "step": 78290 }, { "epoch": 6.78, "learning_rate": 5.812998727882502e-05, "loss": 1.47, "step": 78300 }, { "epoch": 6.78, "learning_rate": 5.812781889672718e-05, "loss": 1.3556, "step": 78310 }, { "epoch": 6.78, "learning_rate": 5.812565051462934e-05, "loss": 1.573, "step": 78320 }, { "epoch": 6.78, "learning_rate": 5.812348213253151e-05, "loss": 1.4642, "step": 78330 }, { "epoch": 6.78, "learning_rate": 5.812131375043367e-05, "loss": 1.3751, "step": 78340 }, { "epoch": 6.79, "learning_rate": 5.8119145368335835e-05, "loss": 1.5109, "step": 78350 }, { "epoch": 6.79, "learning_rate": 5.8116976986238e-05, "loss": 1.4498, "step": 78360 }, { "epoch": 6.79, "learning_rate": 5.811480860414016e-05, "loss": 1.4788, "step": 78370 }, { "epoch": 6.79, "learning_rate": 5.811264022204233e-05, "loss": 1.6295, "step": 78380 }, { "epoch": 6.79, "learning_rate": 5.811047183994448e-05, "loss": 1.5232, "step": 78390 }, { "epoch": 6.79, "learning_rate": 5.810830345784665e-05, "loss": 1.197, "step": 78400 }, { "epoch": 6.79, "learning_rate": 5.8106135075748806e-05, "loss": 1.7044, "step": 78410 }, { "epoch": 6.79, "learning_rate": 5.810396669365097e-05, "loss": 1.5205, "step": 78420 }, { "epoch": 6.79, "learning_rate": 5.810179831155313e-05, "loss": 1.3596, "step": 78430 }, { "epoch": 6.79, "learning_rate": 5.80996299294553e-05, "loss": 1.4981, "step": 78440 }, { "epoch": 6.79, "learning_rate": 5.8097461547357465e-05, "loss": 1.5693, "step": 78450 }, { "epoch": 6.8, "learning_rate": 5.8095293165259625e-05, "loss": 1.4426, "step": 78460 }, { "epoch": 6.8, "learning_rate": 5.8093124783161784e-05, "loss": 1.35, "step": 78470 }, { "epoch": 6.8, "learning_rate": 5.8090956401063944e-05, "loss": 1.2962, "step": 78480 }, { "epoch": 6.8, "learning_rate": 5.808878801896611e-05, "loss": 1.5952, "step": 78490 }, { "epoch": 6.8, "learning_rate": 5.808661963686827e-05, "loss": 1.4531, "step": 78500 }, { "epoch": 6.8, "learning_rate": 5.8084451254770436e-05, "loss": 1.4673, "step": 78510 }, { "epoch": 6.8, "learning_rate": 5.80822828726726e-05, "loss": 1.5017, "step": 78520 }, { "epoch": 6.8, "learning_rate": 5.808011449057476e-05, "loss": 1.4873, "step": 78530 }, { "epoch": 6.8, "learning_rate": 5.807794610847692e-05, "loss": 1.4739, "step": 78540 }, { "epoch": 6.8, "learning_rate": 5.807577772637908e-05, "loss": 1.5613, "step": 78550 }, { "epoch": 6.8, "learning_rate": 5.807360934428125e-05, "loss": 1.3794, "step": 78560 }, { "epoch": 6.8, "learning_rate": 5.807144096218341e-05, "loss": 1.5508, "step": 78570 }, { "epoch": 6.81, "learning_rate": 5.8069272580085574e-05, "loss": 1.4964, "step": 78580 }, { "epoch": 6.81, "learning_rate": 5.806710419798774e-05, "loss": 1.2538, "step": 78590 }, { "epoch": 6.81, "learning_rate": 5.80649358158899e-05, "loss": 1.5385, "step": 78600 }, { "epoch": 6.81, "learning_rate": 5.8062767433792066e-05, "loss": 1.4893, "step": 78610 }, { "epoch": 6.81, "learning_rate": 5.806059905169422e-05, "loss": 1.3432, "step": 78620 }, { "epoch": 6.81, "learning_rate": 5.8058430669596386e-05, "loss": 1.491, "step": 78630 }, { "epoch": 6.81, "learning_rate": 5.8056262287498545e-05, "loss": 1.5625, "step": 78640 }, { "epoch": 6.81, "learning_rate": 5.805409390540071e-05, "loss": 1.2359, "step": 78650 }, { "epoch": 6.81, "learning_rate": 5.805192552330288e-05, "loss": 1.6657, "step": 78660 }, { "epoch": 6.81, "learning_rate": 5.804975714120504e-05, "loss": 1.5243, "step": 78670 }, { "epoch": 6.81, "learning_rate": 5.8047588759107204e-05, "loss": 1.7334, "step": 78680 }, { "epoch": 6.82, "learning_rate": 5.8045420377009364e-05, "loss": 1.8374, "step": 78690 }, { "epoch": 6.82, "learning_rate": 5.804325199491152e-05, "loss": 1.5412, "step": 78700 }, { "epoch": 6.82, "learning_rate": 5.804108361281368e-05, "loss": 1.4199, "step": 78710 }, { "epoch": 6.82, "learning_rate": 5.803891523071585e-05, "loss": 1.585, "step": 78720 }, { "epoch": 6.82, "learning_rate": 5.8036746848618016e-05, "loss": 1.4369, "step": 78730 }, { "epoch": 6.82, "learning_rate": 5.8034578466520175e-05, "loss": 1.4305, "step": 78740 }, { "epoch": 6.82, "learning_rate": 5.803241008442234e-05, "loss": 1.4643, "step": 78750 }, { "epoch": 6.82, "learning_rate": 5.80302417023245e-05, "loss": 1.3966, "step": 78760 }, { "epoch": 6.82, "learning_rate": 5.802807332022666e-05, "loss": 1.4381, "step": 78770 }, { "epoch": 6.82, "learning_rate": 5.802590493812882e-05, "loss": 1.5412, "step": 78780 }, { "epoch": 6.82, "learning_rate": 5.802373655603099e-05, "loss": 1.3657, "step": 78790 }, { "epoch": 6.82, "learning_rate": 5.802156817393315e-05, "loss": 1.4185, "step": 78800 }, { "epoch": 6.83, "learning_rate": 5.801939979183531e-05, "loss": 1.4297, "step": 78810 }, { "epoch": 6.83, "learning_rate": 5.801723140973748e-05, "loss": 1.5454, "step": 78820 }, { "epoch": 6.83, "learning_rate": 5.801506302763964e-05, "loss": 1.5678, "step": 78830 }, { "epoch": 6.83, "learning_rate": 5.8012894645541805e-05, "loss": 1.5478, "step": 78840 }, { "epoch": 6.83, "learning_rate": 5.801072626344396e-05, "loss": 1.5576, "step": 78850 }, { "epoch": 6.83, "learning_rate": 5.8008557881346124e-05, "loss": 1.3137, "step": 78860 }, { "epoch": 6.83, "learning_rate": 5.800638949924829e-05, "loss": 1.3776, "step": 78870 }, { "epoch": 6.83, "learning_rate": 5.800422111715045e-05, "loss": 1.5151, "step": 78880 }, { "epoch": 6.83, "learning_rate": 5.800205273505262e-05, "loss": 1.5088, "step": 78890 }, { "epoch": 6.83, "learning_rate": 5.7999884352954776e-05, "loss": 1.3823, "step": 78900 }, { "epoch": 6.83, "learning_rate": 5.799771597085694e-05, "loss": 1.5671, "step": 78910 }, { "epoch": 6.83, "learning_rate": 5.799554758875911e-05, "loss": 1.5789, "step": 78920 }, { "epoch": 6.84, "learning_rate": 5.799337920666126e-05, "loss": 1.3697, "step": 78930 }, { "epoch": 6.84, "learning_rate": 5.799121082456342e-05, "loss": 1.3801, "step": 78940 }, { "epoch": 6.84, "learning_rate": 5.798904244246559e-05, "loss": 1.5606, "step": 78950 }, { "epoch": 6.84, "learning_rate": 5.7986874060367755e-05, "loss": 1.5122, "step": 78960 }, { "epoch": 6.84, "learning_rate": 5.7984705678269914e-05, "loss": 1.4177, "step": 78970 }, { "epoch": 6.84, "learning_rate": 5.798253729617208e-05, "loss": 1.5396, "step": 78980 }, { "epoch": 6.84, "learning_rate": 5.798036891407424e-05, "loss": 1.3338, "step": 78990 }, { "epoch": 6.84, "learning_rate": 5.7978200531976407e-05, "loss": 1.4334, "step": 79000 }, { "epoch": 6.84, "learning_rate": 5.797603214987856e-05, "loss": 1.3197, "step": 79010 }, { "epoch": 6.84, "learning_rate": 5.7973863767780726e-05, "loss": 1.524, "step": 79020 }, { "epoch": 6.84, "learning_rate": 5.797169538568289e-05, "loss": 1.4013, "step": 79030 }, { "epoch": 6.85, "learning_rate": 5.796952700358505e-05, "loss": 1.3426, "step": 79040 }, { "epoch": 6.85, "learning_rate": 5.796735862148722e-05, "loss": 1.4198, "step": 79050 }, { "epoch": 6.85, "learning_rate": 5.796519023938938e-05, "loss": 1.392, "step": 79060 }, { "epoch": 6.85, "learning_rate": 5.7963021857291544e-05, "loss": 1.4167, "step": 79070 }, { "epoch": 6.85, "learning_rate": 5.79608534751937e-05, "loss": 1.4879, "step": 79080 }, { "epoch": 6.85, "learning_rate": 5.7958685093095863e-05, "loss": 1.5599, "step": 79090 }, { "epoch": 6.85, "learning_rate": 5.795651671099803e-05, "loss": 1.463, "step": 79100 }, { "epoch": 6.85, "learning_rate": 5.795434832890019e-05, "loss": 1.536, "step": 79110 }, { "epoch": 6.85, "learning_rate": 5.7952179946802356e-05, "loss": 1.3539, "step": 79120 }, { "epoch": 6.85, "learning_rate": 5.7950011564704515e-05, "loss": 1.5338, "step": 79130 }, { "epoch": 6.85, "learning_rate": 5.794784318260668e-05, "loss": 1.4941, "step": 79140 }, { "epoch": 6.85, "learning_rate": 5.794567480050885e-05, "loss": 1.418, "step": 79150 }, { "epoch": 6.86, "learning_rate": 5.7943506418411e-05, "loss": 1.8125, "step": 79160 }, { "epoch": 6.86, "learning_rate": 5.794133803631317e-05, "loss": 1.424, "step": 79170 }, { "epoch": 6.86, "learning_rate": 5.793916965421533e-05, "loss": 1.3581, "step": 79180 }, { "epoch": 6.86, "learning_rate": 5.7937001272117493e-05, "loss": 1.4079, "step": 79190 }, { "epoch": 6.86, "learning_rate": 5.793483289001965e-05, "loss": 1.4322, "step": 79200 }, { "epoch": 6.86, "learning_rate": 5.793266450792182e-05, "loss": 1.4956, "step": 79210 }, { "epoch": 6.86, "learning_rate": 5.7930496125823986e-05, "loss": 1.3489, "step": 79220 }, { "epoch": 6.86, "learning_rate": 5.7928327743726145e-05, "loss": 1.4795, "step": 79230 }, { "epoch": 6.86, "learning_rate": 5.7926159361628305e-05, "loss": 1.4113, "step": 79240 }, { "epoch": 6.86, "learning_rate": 5.7923990979530465e-05, "loss": 1.4286, "step": 79250 }, { "epoch": 6.86, "learning_rate": 5.792182259743263e-05, "loss": 1.3827, "step": 79260 }, { "epoch": 6.87, "learning_rate": 5.791965421533479e-05, "loss": 1.457, "step": 79270 }, { "epoch": 6.87, "learning_rate": 5.791748583323696e-05, "loss": 1.5373, "step": 79280 }, { "epoch": 6.87, "learning_rate": 5.7915317451139123e-05, "loss": 1.4919, "step": 79290 }, { "epoch": 6.87, "learning_rate": 5.791314906904128e-05, "loss": 1.6605, "step": 79300 }, { "epoch": 6.87, "learning_rate": 5.791098068694344e-05, "loss": 1.3547, "step": 79310 }, { "epoch": 6.87, "learning_rate": 5.79088123048456e-05, "loss": 1.4657, "step": 79320 }, { "epoch": 6.87, "learning_rate": 5.790664392274777e-05, "loss": 1.5504, "step": 79330 }, { "epoch": 6.87, "learning_rate": 5.790447554064993e-05, "loss": 1.7021, "step": 79340 }, { "epoch": 6.87, "learning_rate": 5.7902307158552095e-05, "loss": 1.5004, "step": 79350 }, { "epoch": 6.87, "learning_rate": 5.790013877645426e-05, "loss": 1.4757, "step": 79360 }, { "epoch": 6.87, "learning_rate": 5.789797039435642e-05, "loss": 1.3119, "step": 79370 }, { "epoch": 6.87, "learning_rate": 5.789580201225859e-05, "loss": 1.5391, "step": 79380 }, { "epoch": 6.88, "learning_rate": 5.789363363016074e-05, "loss": 1.4677, "step": 79390 }, { "epoch": 6.88, "learning_rate": 5.7891465248062906e-05, "loss": 1.4979, "step": 79400 }, { "epoch": 6.88, "learning_rate": 5.7889296865965066e-05, "loss": 1.7555, "step": 79410 }, { "epoch": 6.88, "learning_rate": 5.788712848386723e-05, "loss": 1.4172, "step": 79420 }, { "epoch": 6.88, "learning_rate": 5.78849601017694e-05, "loss": 1.6284, "step": 79430 }, { "epoch": 6.88, "learning_rate": 5.788279171967156e-05, "loss": 1.4352, "step": 79440 }, { "epoch": 6.88, "learning_rate": 5.7880623337573725e-05, "loss": 1.5922, "step": 79450 }, { "epoch": 6.88, "learning_rate": 5.7878454955475884e-05, "loss": 1.4643, "step": 79460 }, { "epoch": 6.88, "learning_rate": 5.7876286573378044e-05, "loss": 1.5622, "step": 79470 }, { "epoch": 6.88, "learning_rate": 5.7874118191280204e-05, "loss": 1.5811, "step": 79480 }, { "epoch": 6.88, "learning_rate": 5.787194980918237e-05, "loss": 1.6005, "step": 79490 }, { "epoch": 6.89, "learning_rate": 5.786978142708453e-05, "loss": 1.4535, "step": 79500 }, { "epoch": 6.89, "learning_rate": 5.7867613044986696e-05, "loss": 1.3824, "step": 79510 }, { "epoch": 6.89, "learning_rate": 5.786544466288886e-05, "loss": 1.4447, "step": 79520 }, { "epoch": 6.89, "learning_rate": 5.786327628079102e-05, "loss": 1.3665, "step": 79530 }, { "epoch": 6.89, "learning_rate": 5.786110789869318e-05, "loss": 1.4773, "step": 79540 }, { "epoch": 6.89, "learning_rate": 5.785893951659534e-05, "loss": 1.4238, "step": 79550 }, { "epoch": 6.89, "learning_rate": 5.785677113449751e-05, "loss": 1.4442, "step": 79560 }, { "epoch": 6.89, "learning_rate": 5.785460275239967e-05, "loss": 1.3948, "step": 79570 }, { "epoch": 6.89, "learning_rate": 5.7852434370301834e-05, "loss": 1.548, "step": 79580 }, { "epoch": 6.89, "learning_rate": 5.7850265988204e-05, "loss": 1.6328, "step": 79590 }, { "epoch": 6.89, "learning_rate": 5.784809760610616e-05, "loss": 1.3467, "step": 79600 }, { "epoch": 6.89, "learning_rate": 5.7845929224008326e-05, "loss": 1.4371, "step": 79610 }, { "epoch": 6.9, "learning_rate": 5.784376084191048e-05, "loss": 1.4948, "step": 79620 }, { "epoch": 6.9, "learning_rate": 5.7841592459812645e-05, "loss": 1.475, "step": 79630 }, { "epoch": 6.9, "learning_rate": 5.7839424077714805e-05, "loss": 1.4493, "step": 79640 }, { "epoch": 6.9, "learning_rate": 5.783725569561697e-05, "loss": 1.5793, "step": 79650 }, { "epoch": 6.9, "learning_rate": 5.783508731351914e-05, "loss": 1.5633, "step": 79660 }, { "epoch": 6.9, "learning_rate": 5.78329189314213e-05, "loss": 1.468, "step": 79670 }, { "epoch": 6.9, "learning_rate": 5.7830750549323464e-05, "loss": 1.3439, "step": 79680 }, { "epoch": 6.9, "learning_rate": 5.782858216722562e-05, "loss": 1.2584, "step": 79690 }, { "epoch": 6.9, "learning_rate": 5.782641378512778e-05, "loss": 1.3405, "step": 79700 }, { "epoch": 6.9, "learning_rate": 5.782424540302994e-05, "loss": 1.477, "step": 79710 }, { "epoch": 6.9, "learning_rate": 5.782207702093211e-05, "loss": 1.4363, "step": 79720 }, { "epoch": 6.91, "learning_rate": 5.7819908638834275e-05, "loss": 1.4172, "step": 79730 }, { "epoch": 6.91, "learning_rate": 5.7817740256736435e-05, "loss": 1.6131, "step": 79740 }, { "epoch": 6.91, "learning_rate": 5.78155718746386e-05, "loss": 1.5165, "step": 79750 }, { "epoch": 6.91, "learning_rate": 5.781340349254076e-05, "loss": 1.2734, "step": 79760 }, { "epoch": 6.91, "learning_rate": 5.781123511044293e-05, "loss": 1.5254, "step": 79770 }, { "epoch": 6.91, "learning_rate": 5.780906672834508e-05, "loss": 1.3712, "step": 79780 }, { "epoch": 6.91, "learning_rate": 5.7806898346247247e-05, "loss": 1.5601, "step": 79790 }, { "epoch": 6.91, "learning_rate": 5.780472996414941e-05, "loss": 1.3853, "step": 79800 }, { "epoch": 6.91, "learning_rate": 5.780256158205157e-05, "loss": 1.4912, "step": 79810 }, { "epoch": 6.91, "learning_rate": 5.780039319995374e-05, "loss": 1.4802, "step": 79820 }, { "epoch": 6.91, "learning_rate": 5.77982248178559e-05, "loss": 1.2976, "step": 79830 }, { "epoch": 6.91, "learning_rate": 5.7796056435758065e-05, "loss": 1.3046, "step": 79840 }, { "epoch": 6.92, "learning_rate": 5.779388805366022e-05, "loss": 1.5561, "step": 79850 }, { "epoch": 6.92, "learning_rate": 5.7791719671562384e-05, "loss": 1.5209, "step": 79860 }, { "epoch": 6.92, "learning_rate": 5.778955128946455e-05, "loss": 1.6631, "step": 79870 }, { "epoch": 6.92, "learning_rate": 5.778738290736671e-05, "loss": 1.4053, "step": 79880 }, { "epoch": 6.92, "learning_rate": 5.7785214525268877e-05, "loss": 1.3422, "step": 79890 }, { "epoch": 6.92, "learning_rate": 5.7783046143171036e-05, "loss": 1.3587, "step": 79900 }, { "epoch": 6.92, "learning_rate": 5.77808777610732e-05, "loss": 1.4315, "step": 79910 }, { "epoch": 6.92, "learning_rate": 5.777870937897537e-05, "loss": 1.5618, "step": 79920 }, { "epoch": 6.92, "learning_rate": 5.777654099687752e-05, "loss": 1.6217, "step": 79930 }, { "epoch": 6.92, "learning_rate": 5.777437261477969e-05, "loss": 1.4247, "step": 79940 }, { "epoch": 6.92, "learning_rate": 5.777220423268185e-05, "loss": 1.4255, "step": 79950 }, { "epoch": 6.93, "learning_rate": 5.7770035850584014e-05, "loss": 1.7107, "step": 79960 }, { "epoch": 6.93, "learning_rate": 5.7767867468486174e-05, "loss": 1.5053, "step": 79970 }, { "epoch": 6.93, "learning_rate": 5.776569908638834e-05, "loss": 1.5163, "step": 79980 }, { "epoch": 6.93, "learning_rate": 5.7763530704290507e-05, "loss": 1.7031, "step": 79990 }, { "epoch": 6.93, "learning_rate": 5.7761362322192666e-05, "loss": 1.2692, "step": 80000 }, { "epoch": 6.93, "learning_rate": 5.775919394009482e-05, "loss": 1.5305, "step": 80010 }, { "epoch": 6.93, "learning_rate": 5.7757025557996985e-05, "loss": 1.3965, "step": 80020 }, { "epoch": 6.93, "learning_rate": 5.775485717589915e-05, "loss": 1.5588, "step": 80030 }, { "epoch": 6.93, "learning_rate": 5.775268879380131e-05, "loss": 1.3192, "step": 80040 }, { "epoch": 6.93, "learning_rate": 5.775052041170348e-05, "loss": 1.6132, "step": 80050 }, { "epoch": 6.93, "learning_rate": 5.774835202960564e-05, "loss": 1.301, "step": 80060 }, { "epoch": 6.93, "learning_rate": 5.7746183647507804e-05, "loss": 1.5746, "step": 80070 }, { "epoch": 6.94, "learning_rate": 5.774401526540996e-05, "loss": 1.6098, "step": 80080 }, { "epoch": 6.94, "learning_rate": 5.774184688331212e-05, "loss": 1.4914, "step": 80090 }, { "epoch": 6.94, "learning_rate": 5.773967850121429e-05, "loss": 1.3957, "step": 80100 }, { "epoch": 6.94, "learning_rate": 5.773751011911645e-05, "loss": 1.3825, "step": 80110 }, { "epoch": 6.94, "learning_rate": 5.7735341737018615e-05, "loss": 1.3259, "step": 80120 }, { "epoch": 6.94, "learning_rate": 5.7733173354920775e-05, "loss": 1.37, "step": 80130 }, { "epoch": 6.94, "learning_rate": 5.773100497282294e-05, "loss": 1.347, "step": 80140 }, { "epoch": 6.94, "learning_rate": 5.772883659072511e-05, "loss": 1.3919, "step": 80150 }, { "epoch": 6.94, "learning_rate": 5.772666820862726e-05, "loss": 1.3278, "step": 80160 }, { "epoch": 6.94, "learning_rate": 5.772449982652943e-05, "loss": 1.3993, "step": 80170 }, { "epoch": 6.94, "learning_rate": 5.772233144443159e-05, "loss": 1.3017, "step": 80180 }, { "epoch": 6.94, "learning_rate": 5.772016306233375e-05, "loss": 1.3102, "step": 80190 }, { "epoch": 6.95, "learning_rate": 5.771799468023591e-05, "loss": 1.4717, "step": 80200 }, { "epoch": 6.95, "learning_rate": 5.771582629813808e-05, "loss": 1.4233, "step": 80210 }, { "epoch": 6.95, "learning_rate": 5.7713657916040246e-05, "loss": 1.3977, "step": 80220 }, { "epoch": 6.95, "learning_rate": 5.7711489533942405e-05, "loss": 1.5804, "step": 80230 }, { "epoch": 6.95, "learning_rate": 5.7709321151844565e-05, "loss": 1.3086, "step": 80240 }, { "epoch": 6.95, "learning_rate": 5.7707152769746724e-05, "loss": 1.3496, "step": 80250 }, { "epoch": 6.95, "learning_rate": 5.770498438764889e-05, "loss": 1.4753, "step": 80260 }, { "epoch": 6.95, "learning_rate": 5.770281600555105e-05, "loss": 1.5582, "step": 80270 }, { "epoch": 6.95, "learning_rate": 5.770064762345322e-05, "loss": 1.476, "step": 80280 }, { "epoch": 6.95, "learning_rate": 5.769847924135538e-05, "loss": 1.2267, "step": 80290 }, { "epoch": 6.95, "learning_rate": 5.769631085925754e-05, "loss": 1.4072, "step": 80300 }, { "epoch": 6.96, "learning_rate": 5.76941424771597e-05, "loss": 1.2576, "step": 80310 }, { "epoch": 6.96, "learning_rate": 5.769197409506186e-05, "loss": 1.5114, "step": 80320 }, { "epoch": 6.96, "learning_rate": 5.768980571296403e-05, "loss": 1.5436, "step": 80330 }, { "epoch": 6.96, "learning_rate": 5.768763733086619e-05, "loss": 1.6082, "step": 80340 }, { "epoch": 6.96, "learning_rate": 5.7685468948768354e-05, "loss": 1.4544, "step": 80350 }, { "epoch": 6.96, "learning_rate": 5.768330056667052e-05, "loss": 1.533, "step": 80360 }, { "epoch": 6.96, "learning_rate": 5.768113218457268e-05, "loss": 1.5941, "step": 80370 }, { "epoch": 6.96, "learning_rate": 5.767896380247485e-05, "loss": 1.3607, "step": 80380 }, { "epoch": 6.96, "learning_rate": 5.7676795420377e-05, "loss": 1.4138, "step": 80390 }, { "epoch": 6.96, "learning_rate": 5.7674627038279166e-05, "loss": 1.4366, "step": 80400 }, { "epoch": 6.96, "learning_rate": 5.7672458656181326e-05, "loss": 1.2805, "step": 80410 }, { "epoch": 6.96, "learning_rate": 5.767029027408349e-05, "loss": 1.6298, "step": 80420 }, { "epoch": 6.97, "learning_rate": 5.766812189198566e-05, "loss": 1.3882, "step": 80430 }, { "epoch": 6.97, "learning_rate": 5.766595350988782e-05, "loss": 1.665, "step": 80440 }, { "epoch": 6.97, "learning_rate": 5.7663785127789984e-05, "loss": 1.4186, "step": 80450 }, { "epoch": 6.97, "learning_rate": 5.7661616745692144e-05, "loss": 1.4744, "step": 80460 }, { "epoch": 6.97, "learning_rate": 5.7659448363594304e-05, "loss": 1.425, "step": 80470 }, { "epoch": 6.97, "learning_rate": 5.765727998149646e-05, "loss": 1.402, "step": 80480 }, { "epoch": 6.97, "learning_rate": 5.765511159939863e-05, "loss": 1.7648, "step": 80490 }, { "epoch": 6.97, "learning_rate": 5.7652943217300796e-05, "loss": 1.5032, "step": 80500 }, { "epoch": 6.97, "learning_rate": 5.7650774835202956e-05, "loss": 1.419, "step": 80510 }, { "epoch": 6.97, "learning_rate": 5.764860645310512e-05, "loss": 1.3403, "step": 80520 }, { "epoch": 6.97, "learning_rate": 5.764643807100728e-05, "loss": 1.5783, "step": 80530 }, { "epoch": 6.98, "learning_rate": 5.764426968890945e-05, "loss": 1.3163, "step": 80540 }, { "epoch": 6.98, "learning_rate": 5.76421013068116e-05, "loss": 1.4351, "step": 80550 }, { "epoch": 6.98, "learning_rate": 5.763993292471377e-05, "loss": 1.5203, "step": 80560 }, { "epoch": 6.98, "learning_rate": 5.763776454261593e-05, "loss": 1.4007, "step": 80570 }, { "epoch": 6.98, "learning_rate": 5.763559616051809e-05, "loss": 1.3852, "step": 80580 }, { "epoch": 6.98, "learning_rate": 5.763342777842026e-05, "loss": 1.6274, "step": 80590 }, { "epoch": 6.98, "learning_rate": 5.763125939632242e-05, "loss": 1.3356, "step": 80600 }, { "epoch": 6.98, "learning_rate": 5.7629091014224586e-05, "loss": 1.6454, "step": 80610 }, { "epoch": 6.98, "learning_rate": 5.762692263212674e-05, "loss": 1.5455, "step": 80620 }, { "epoch": 6.98, "learning_rate": 5.7624754250028905e-05, "loss": 1.5188, "step": 80630 }, { "epoch": 6.98, "learning_rate": 5.7622585867931065e-05, "loss": 1.2803, "step": 80640 }, { "epoch": 6.98, "learning_rate": 5.762041748583323e-05, "loss": 1.5546, "step": 80650 }, { "epoch": 6.99, "learning_rate": 5.76182491037354e-05, "loss": 1.5565, "step": 80660 }, { "epoch": 6.99, "learning_rate": 5.761608072163756e-05, "loss": 1.457, "step": 80670 }, { "epoch": 6.99, "learning_rate": 5.761391233953972e-05, "loss": 1.365, "step": 80680 }, { "epoch": 6.99, "learning_rate": 5.761174395744188e-05, "loss": 1.541, "step": 80690 }, { "epoch": 6.99, "learning_rate": 5.760957557534404e-05, "loss": 1.4627, "step": 80700 }, { "epoch": 6.99, "learning_rate": 5.76074071932462e-05, "loss": 1.5933, "step": 80710 }, { "epoch": 6.99, "learning_rate": 5.760523881114837e-05, "loss": 1.3838, "step": 80720 }, { "epoch": 6.99, "learning_rate": 5.7603070429050535e-05, "loss": 1.3864, "step": 80730 }, { "epoch": 6.99, "learning_rate": 5.7600902046952695e-05, "loss": 1.5055, "step": 80740 }, { "epoch": 6.99, "learning_rate": 5.759873366485486e-05, "loss": 1.414, "step": 80750 }, { "epoch": 6.99, "learning_rate": 5.759656528275702e-05, "loss": 1.5662, "step": 80760 }, { "epoch": 7.0, "learning_rate": 5.759439690065919e-05, "loss": 1.4858, "step": 80770 }, { "epoch": 7.0, "learning_rate": 5.759222851856134e-05, "loss": 1.4814, "step": 80780 }, { "epoch": 7.0, "learning_rate": 5.7590060136463506e-05, "loss": 1.3684, "step": 80790 }, { "epoch": 7.0, "learning_rate": 5.758789175436567e-05, "loss": 1.5366, "step": 80800 }, { "epoch": 7.0, "learning_rate": 5.758572337226783e-05, "loss": 1.5321, "step": 80810 }, { "epoch": 7.0, "learning_rate": 5.758355499017e-05, "loss": 1.5403, "step": 80820 }, { "epoch": 7.0, "eval_cer": 0.5578091125723398, "eval_em": 0.02320783909231563, "eval_f1": 0.02320783909231563, "eval_loss": 1.532400369644165, "eval_runtime": 275.6685, "eval_samples_per_second": 7.034, "eval_steps_per_second": 0.881, "eval_wer": 0.9767921609076844, "step": 80825 }, { "epoch": 7.0, "learning_rate": 5.758138660807216e-05, "loss": 1.3355, "step": 80830 }, { "epoch": 7.0, "learning_rate": 5.7579218225974325e-05, "loss": 1.4409, "step": 80840 }, { "epoch": 7.0, "learning_rate": 5.757704984387648e-05, "loss": 1.3896, "step": 80850 }, { "epoch": 7.0, "learning_rate": 5.7574881461778644e-05, "loss": 1.4118, "step": 80860 }, { "epoch": 7.0, "learning_rate": 5.757271307968081e-05, "loss": 1.5068, "step": 80870 }, { "epoch": 7.0, "learning_rate": 5.757054469758297e-05, "loss": 1.3839, "step": 80880 }, { "epoch": 7.01, "learning_rate": 5.7568376315485136e-05, "loss": 1.3526, "step": 80890 }, { "epoch": 7.01, "learning_rate": 5.7566207933387296e-05, "loss": 1.2898, "step": 80900 }, { "epoch": 7.01, "learning_rate": 5.756403955128946e-05, "loss": 1.4846, "step": 80910 }, { "epoch": 7.01, "learning_rate": 5.756187116919163e-05, "loss": 1.5906, "step": 80920 }, { "epoch": 7.01, "learning_rate": 5.755970278709378e-05, "loss": 1.4228, "step": 80930 }, { "epoch": 7.01, "learning_rate": 5.755753440499595e-05, "loss": 1.2288, "step": 80940 }, { "epoch": 7.01, "learning_rate": 5.755536602289811e-05, "loss": 1.3234, "step": 80950 }, { "epoch": 7.01, "learning_rate": 5.7553197640800274e-05, "loss": 1.2911, "step": 80960 }, { "epoch": 7.01, "learning_rate": 5.7551029258702434e-05, "loss": 1.2704, "step": 80970 }, { "epoch": 7.01, "learning_rate": 5.75488608766046e-05, "loss": 1.373, "step": 80980 }, { "epoch": 7.01, "learning_rate": 5.7546692494506766e-05, "loss": 1.4396, "step": 80990 }, { "epoch": 7.02, "learning_rate": 5.7544524112408926e-05, "loss": 1.4476, "step": 81000 }, { "epoch": 7.02, "learning_rate": 5.7542355730311086e-05, "loss": 1.4187, "step": 81010 }, { "epoch": 7.02, "learning_rate": 5.7540187348213245e-05, "loss": 1.2864, "step": 81020 }, { "epoch": 7.02, "learning_rate": 5.753801896611541e-05, "loss": 1.2726, "step": 81030 }, { "epoch": 7.02, "learning_rate": 5.753585058401757e-05, "loss": 1.2898, "step": 81040 }, { "epoch": 7.02, "learning_rate": 5.753368220191974e-05, "loss": 1.238, "step": 81050 }, { "epoch": 7.02, "learning_rate": 5.7531513819821904e-05, "loss": 1.309, "step": 81060 }, { "epoch": 7.02, "learning_rate": 5.7529345437724064e-05, "loss": 1.2913, "step": 81070 }, { "epoch": 7.02, "learning_rate": 5.7527177055626216e-05, "loss": 1.1221, "step": 81080 }, { "epoch": 7.02, "learning_rate": 5.752500867352838e-05, "loss": 1.4614, "step": 81090 }, { "epoch": 7.02, "learning_rate": 5.752284029143055e-05, "loss": 1.386, "step": 81100 }, { "epoch": 7.02, "learning_rate": 5.752067190933271e-05, "loss": 1.5705, "step": 81110 }, { "epoch": 7.03, "learning_rate": 5.7518503527234875e-05, "loss": 1.2841, "step": 81120 }, { "epoch": 7.03, "learning_rate": 5.7516335145137035e-05, "loss": 1.4525, "step": 81130 }, { "epoch": 7.03, "learning_rate": 5.75141667630392e-05, "loss": 1.7474, "step": 81140 }, { "epoch": 7.03, "learning_rate": 5.751199838094137e-05, "loss": 1.2451, "step": 81150 }, { "epoch": 7.03, "learning_rate": 5.750982999884352e-05, "loss": 1.2942, "step": 81160 }, { "epoch": 7.03, "learning_rate": 5.750766161674569e-05, "loss": 1.4026, "step": 81170 }, { "epoch": 7.03, "learning_rate": 5.7505493234647846e-05, "loss": 1.3224, "step": 81180 }, { "epoch": 7.03, "learning_rate": 5.750332485255001e-05, "loss": 1.3938, "step": 81190 }, { "epoch": 7.03, "learning_rate": 5.750115647045217e-05, "loss": 1.2195, "step": 81200 }, { "epoch": 7.03, "learning_rate": 5.749898808835434e-05, "loss": 1.2714, "step": 81210 }, { "epoch": 7.03, "learning_rate": 5.7496819706256505e-05, "loss": 1.2688, "step": 81220 }, { "epoch": 7.04, "learning_rate": 5.7494651324158665e-05, "loss": 1.3781, "step": 81230 }, { "epoch": 7.04, "learning_rate": 5.7492482942060824e-05, "loss": 1.5377, "step": 81240 }, { "epoch": 7.04, "learning_rate": 5.7490314559962984e-05, "loss": 1.317, "step": 81250 }, { "epoch": 7.04, "learning_rate": 5.748814617786515e-05, "loss": 1.423, "step": 81260 }, { "epoch": 7.04, "learning_rate": 5.748597779576731e-05, "loss": 1.4265, "step": 81270 }, { "epoch": 7.04, "learning_rate": 5.7483809413669476e-05, "loss": 1.4199, "step": 81280 }, { "epoch": 7.04, "learning_rate": 5.748164103157164e-05, "loss": 1.3911, "step": 81290 }, { "epoch": 7.04, "learning_rate": 5.74794726494738e-05, "loss": 1.4414, "step": 81300 }, { "epoch": 7.04, "learning_rate": 5.747730426737597e-05, "loss": 1.4647, "step": 81310 }, { "epoch": 7.04, "learning_rate": 5.747513588527812e-05, "loss": 1.4208, "step": 81320 }, { "epoch": 7.04, "learning_rate": 5.747296750318029e-05, "loss": 1.3661, "step": 81330 }, { "epoch": 7.04, "learning_rate": 5.747079912108245e-05, "loss": 1.4688, "step": 81340 }, { "epoch": 7.05, "learning_rate": 5.7468630738984614e-05, "loss": 1.4437, "step": 81350 }, { "epoch": 7.05, "learning_rate": 5.746646235688678e-05, "loss": 1.3267, "step": 81360 }, { "epoch": 7.05, "learning_rate": 5.746429397478894e-05, "loss": 1.4698, "step": 81370 }, { "epoch": 7.05, "learning_rate": 5.7462125592691106e-05, "loss": 1.47, "step": 81380 }, { "epoch": 7.05, "learning_rate": 5.745995721059326e-05, "loss": 1.3786, "step": 81390 }, { "epoch": 7.05, "learning_rate": 5.7457788828495426e-05, "loss": 1.4205, "step": 81400 }, { "epoch": 7.05, "learning_rate": 5.7455620446397585e-05, "loss": 1.3223, "step": 81410 }, { "epoch": 7.05, "learning_rate": 5.745345206429975e-05, "loss": 1.2583, "step": 81420 }, { "epoch": 7.05, "learning_rate": 5.745128368220192e-05, "loss": 1.6239, "step": 81430 }, { "epoch": 7.05, "learning_rate": 5.744911530010408e-05, "loss": 1.4752, "step": 81440 }, { "epoch": 7.05, "learning_rate": 5.7446946918006244e-05, "loss": 1.6673, "step": 81450 }, { "epoch": 7.05, "learning_rate": 5.7444778535908404e-05, "loss": 1.5587, "step": 81460 }, { "epoch": 7.06, "learning_rate": 5.744261015381056e-05, "loss": 1.4767, "step": 81470 }, { "epoch": 7.06, "learning_rate": 5.744044177171272e-05, "loss": 1.3543, "step": 81480 }, { "epoch": 7.06, "learning_rate": 5.743827338961489e-05, "loss": 1.4309, "step": 81490 }, { "epoch": 7.06, "learning_rate": 5.7436105007517056e-05, "loss": 1.2944, "step": 81500 }, { "epoch": 7.06, "learning_rate": 5.7433936625419215e-05, "loss": 1.2593, "step": 81510 }, { "epoch": 7.06, "learning_rate": 5.743176824332138e-05, "loss": 1.3619, "step": 81520 }, { "epoch": 7.06, "learning_rate": 5.742959986122354e-05, "loss": 1.3946, "step": 81530 }, { "epoch": 7.06, "learning_rate": 5.742743147912571e-05, "loss": 1.4762, "step": 81540 }, { "epoch": 7.06, "learning_rate": 5.742526309702786e-05, "loss": 1.2738, "step": 81550 }, { "epoch": 7.06, "learning_rate": 5.742309471493003e-05, "loss": 1.3133, "step": 81560 }, { "epoch": 7.06, "learning_rate": 5.7420926332832193e-05, "loss": 1.4564, "step": 81570 }, { "epoch": 7.07, "learning_rate": 5.741875795073435e-05, "loss": 1.5146, "step": 81580 }, { "epoch": 7.07, "learning_rate": 5.741658956863652e-05, "loss": 1.3912, "step": 81590 }, { "epoch": 7.07, "learning_rate": 5.741442118653868e-05, "loss": 1.3971, "step": 81600 }, { "epoch": 7.07, "learning_rate": 5.7412252804440845e-05, "loss": 1.2273, "step": 81610 }, { "epoch": 7.07, "learning_rate": 5.7410084422343e-05, "loss": 1.2976, "step": 81620 }, { "epoch": 7.07, "learning_rate": 5.7407916040245165e-05, "loss": 1.4642, "step": 81630 }, { "epoch": 7.07, "learning_rate": 5.7405747658147324e-05, "loss": 1.4969, "step": 81640 }, { "epoch": 7.07, "learning_rate": 5.740357927604949e-05, "loss": 1.5137, "step": 81650 }, { "epoch": 7.07, "learning_rate": 5.740141089395166e-05, "loss": 1.447, "step": 81660 }, { "epoch": 7.07, "learning_rate": 5.739924251185382e-05, "loss": 1.2217, "step": 81670 }, { "epoch": 7.07, "learning_rate": 5.739707412975598e-05, "loss": 1.3072, "step": 81680 }, { "epoch": 7.07, "learning_rate": 5.739490574765814e-05, "loss": 1.4136, "step": 81690 }, { "epoch": 7.08, "learning_rate": 5.73927373655603e-05, "loss": 1.5946, "step": 81700 }, { "epoch": 7.08, "learning_rate": 5.739056898346246e-05, "loss": 1.5236, "step": 81710 }, { "epoch": 7.08, "learning_rate": 5.738840060136463e-05, "loss": 1.6155, "step": 81720 }, { "epoch": 7.08, "learning_rate": 5.7386232219266795e-05, "loss": 1.4865, "step": 81730 }, { "epoch": 7.08, "learning_rate": 5.7384063837168954e-05, "loss": 1.4335, "step": 81740 }, { "epoch": 7.08, "learning_rate": 5.738189545507112e-05, "loss": 1.2633, "step": 81750 }, { "epoch": 7.08, "learning_rate": 5.737972707297328e-05, "loss": 1.2929, "step": 81760 }, { "epoch": 7.08, "learning_rate": 5.737755869087545e-05, "loss": 1.2197, "step": 81770 }, { "epoch": 7.08, "learning_rate": 5.73753903087776e-05, "loss": 1.2812, "step": 81780 }, { "epoch": 7.08, "learning_rate": 5.7373221926679766e-05, "loss": 1.5524, "step": 81790 }, { "epoch": 7.08, "learning_rate": 5.737105354458193e-05, "loss": 1.5126, "step": 81800 }, { "epoch": 7.09, "learning_rate": 5.736888516248409e-05, "loss": 1.2307, "step": 81810 }, { "epoch": 7.09, "learning_rate": 5.736671678038626e-05, "loss": 1.3917, "step": 81820 }, { "epoch": 7.09, "learning_rate": 5.736454839828842e-05, "loss": 1.5654, "step": 81830 }, { "epoch": 7.09, "learning_rate": 5.7362380016190584e-05, "loss": 1.2483, "step": 81840 }, { "epoch": 7.09, "learning_rate": 5.736021163409274e-05, "loss": 1.4537, "step": 81850 }, { "epoch": 7.09, "learning_rate": 5.7358043251994904e-05, "loss": 1.4715, "step": 81860 }, { "epoch": 7.09, "learning_rate": 5.735587486989707e-05, "loss": 1.237, "step": 81870 }, { "epoch": 7.09, "learning_rate": 5.735370648779923e-05, "loss": 1.2523, "step": 81880 }, { "epoch": 7.09, "learning_rate": 5.7351538105701396e-05, "loss": 1.2767, "step": 81890 }, { "epoch": 7.09, "learning_rate": 5.7349369723603556e-05, "loss": 1.3836, "step": 81900 }, { "epoch": 7.09, "learning_rate": 5.734720134150572e-05, "loss": 1.333, "step": 81910 }, { "epoch": 7.09, "learning_rate": 5.734503295940789e-05, "loss": 1.1801, "step": 81920 }, { "epoch": 7.1, "learning_rate": 5.734286457731004e-05, "loss": 1.2749, "step": 81930 }, { "epoch": 7.1, "learning_rate": 5.734069619521221e-05, "loss": 1.3305, "step": 81940 }, { "epoch": 7.1, "learning_rate": 5.733852781311437e-05, "loss": 1.3835, "step": 81950 }, { "epoch": 7.1, "learning_rate": 5.7336359431016534e-05, "loss": 1.4211, "step": 81960 }, { "epoch": 7.1, "learning_rate": 5.733419104891869e-05, "loss": 1.4035, "step": 81970 }, { "epoch": 7.1, "learning_rate": 5.733202266682086e-05, "loss": 1.4203, "step": 81980 }, { "epoch": 7.1, "learning_rate": 5.7329854284723026e-05, "loss": 1.4926, "step": 81990 }, { "epoch": 7.1, "learning_rate": 5.7327685902625186e-05, "loss": 1.3857, "step": 82000 }, { "epoch": 7.1, "learning_rate": 5.7325517520527345e-05, "loss": 1.4024, "step": 82010 }, { "epoch": 7.1, "learning_rate": 5.7323349138429505e-05, "loss": 1.3703, "step": 82020 }, { "epoch": 7.1, "learning_rate": 5.732118075633167e-05, "loss": 1.304, "step": 82030 }, { "epoch": 7.11, "learning_rate": 5.731901237423383e-05, "loss": 1.5526, "step": 82040 }, { "epoch": 7.11, "learning_rate": 5.7316843992136e-05, "loss": 1.2584, "step": 82050 }, { "epoch": 7.11, "learning_rate": 5.7314675610038164e-05, "loss": 1.4953, "step": 82060 }, { "epoch": 7.11, "learning_rate": 5.731250722794032e-05, "loss": 1.2961, "step": 82070 }, { "epoch": 7.11, "learning_rate": 5.731033884584249e-05, "loss": 1.3617, "step": 82080 }, { "epoch": 7.11, "learning_rate": 5.730817046374464e-05, "loss": 1.2727, "step": 82090 }, { "epoch": 7.11, "learning_rate": 5.730600208164681e-05, "loss": 1.5396, "step": 82100 }, { "epoch": 7.11, "learning_rate": 5.730383369954897e-05, "loss": 1.2508, "step": 82110 }, { "epoch": 7.11, "learning_rate": 5.7301665317451135e-05, "loss": 1.1469, "step": 82120 }, { "epoch": 7.11, "learning_rate": 5.72994969353533e-05, "loss": 1.1877, "step": 82130 }, { "epoch": 7.11, "learning_rate": 5.729732855325546e-05, "loss": 1.5265, "step": 82140 }, { "epoch": 7.11, "learning_rate": 5.729516017115763e-05, "loss": 1.3389, "step": 82150 }, { "epoch": 7.12, "learning_rate": 5.729299178905978e-05, "loss": 1.3792, "step": 82160 }, { "epoch": 7.12, "learning_rate": 5.7290823406961947e-05, "loss": 1.4438, "step": 82170 }, { "epoch": 7.12, "learning_rate": 5.7288655024864106e-05, "loss": 1.4008, "step": 82180 }, { "epoch": 7.12, "learning_rate": 5.728648664276627e-05, "loss": 1.5251, "step": 82190 }, { "epoch": 7.12, "learning_rate": 5.728431826066843e-05, "loss": 1.3563, "step": 82200 }, { "epoch": 7.12, "learning_rate": 5.72821498785706e-05, "loss": 1.31, "step": 82210 }, { "epoch": 7.12, "learning_rate": 5.7279981496472765e-05, "loss": 1.3017, "step": 82220 }, { "epoch": 7.12, "learning_rate": 5.7277813114374925e-05, "loss": 1.5137, "step": 82230 }, { "epoch": 7.12, "learning_rate": 5.7275644732277084e-05, "loss": 1.3602, "step": 82240 }, { "epoch": 7.12, "learning_rate": 5.7273476350179244e-05, "loss": 1.4132, "step": 82250 }, { "epoch": 7.12, "learning_rate": 5.727130796808141e-05, "loss": 1.3506, "step": 82260 }, { "epoch": 7.13, "learning_rate": 5.726913958598357e-05, "loss": 1.4469, "step": 82270 }, { "epoch": 7.13, "learning_rate": 5.7266971203885736e-05, "loss": 1.3916, "step": 82280 }, { "epoch": 7.13, "learning_rate": 5.72648028217879e-05, "loss": 1.224, "step": 82290 }, { "epoch": 7.13, "learning_rate": 5.726263443969006e-05, "loss": 1.6037, "step": 82300 }, { "epoch": 7.13, "learning_rate": 5.726046605759223e-05, "loss": 1.5029, "step": 82310 }, { "epoch": 7.13, "learning_rate": 5.725829767549438e-05, "loss": 1.5523, "step": 82320 }, { "epoch": 7.13, "learning_rate": 5.725612929339655e-05, "loss": 1.5298, "step": 82330 }, { "epoch": 7.13, "learning_rate": 5.725396091129871e-05, "loss": 1.6295, "step": 82340 }, { "epoch": 7.13, "learning_rate": 5.7251792529200874e-05, "loss": 1.5662, "step": 82350 }, { "epoch": 7.13, "learning_rate": 5.724962414710304e-05, "loss": 1.4962, "step": 82360 }, { "epoch": 7.13, "learning_rate": 5.72474557650052e-05, "loss": 1.4223, "step": 82370 }, { "epoch": 7.13, "learning_rate": 5.7245287382907366e-05, "loss": 1.3556, "step": 82380 }, { "epoch": 7.14, "learning_rate": 5.724311900080952e-05, "loss": 1.364, "step": 82390 }, { "epoch": 7.14, "learning_rate": 5.7240950618711685e-05, "loss": 1.3981, "step": 82400 }, { "epoch": 7.14, "learning_rate": 5.7238782236613845e-05, "loss": 1.3531, "step": 82410 }, { "epoch": 7.14, "learning_rate": 5.723661385451601e-05, "loss": 1.2133, "step": 82420 }, { "epoch": 7.14, "learning_rate": 5.723444547241818e-05, "loss": 1.4004, "step": 82430 }, { "epoch": 7.14, "learning_rate": 5.723227709032034e-05, "loss": 1.5294, "step": 82440 }, { "epoch": 7.14, "learning_rate": 5.7230108708222504e-05, "loss": 1.4742, "step": 82450 }, { "epoch": 7.14, "learning_rate": 5.7227940326124663e-05, "loss": 1.4549, "step": 82460 }, { "epoch": 7.14, "learning_rate": 5.722577194402682e-05, "loss": 1.3945, "step": 82470 }, { "epoch": 7.14, "learning_rate": 5.722360356192898e-05, "loss": 1.3687, "step": 82480 }, { "epoch": 7.14, "learning_rate": 5.722143517983115e-05, "loss": 1.3627, "step": 82490 }, { "epoch": 7.15, "learning_rate": 5.7219266797733315e-05, "loss": 1.3811, "step": 82500 }, { "epoch": 7.15, "learning_rate": 5.7217098415635475e-05, "loss": 1.2133, "step": 82510 }, { "epoch": 7.15, "learning_rate": 5.721493003353764e-05, "loss": 1.606, "step": 82520 }, { "epoch": 7.15, "learning_rate": 5.72127616514398e-05, "loss": 1.1659, "step": 82530 }, { "epoch": 7.15, "learning_rate": 5.721059326934197e-05, "loss": 1.1659, "step": 82540 }, { "epoch": 7.15, "learning_rate": 5.720842488724412e-05, "loss": 1.5526, "step": 82550 }, { "epoch": 7.15, "learning_rate": 5.720625650514629e-05, "loss": 1.5008, "step": 82560 }, { "epoch": 7.15, "learning_rate": 5.720408812304845e-05, "loss": 1.2811, "step": 82570 }, { "epoch": 7.15, "learning_rate": 5.720191974095061e-05, "loss": 1.3934, "step": 82580 }, { "epoch": 7.15, "learning_rate": 5.719975135885278e-05, "loss": 1.4182, "step": 82590 }, { "epoch": 7.15, "learning_rate": 5.719758297675494e-05, "loss": 1.4798, "step": 82600 }, { "epoch": 7.15, "learning_rate": 5.7195414594657105e-05, "loss": 1.4472, "step": 82610 }, { "epoch": 7.16, "learning_rate": 5.719324621255926e-05, "loss": 1.4351, "step": 82620 }, { "epoch": 7.16, "learning_rate": 5.7191077830461424e-05, "loss": 1.4656, "step": 82630 }, { "epoch": 7.16, "learning_rate": 5.718890944836359e-05, "loss": 1.3131, "step": 82640 }, { "epoch": 7.16, "learning_rate": 5.718674106626575e-05, "loss": 1.7152, "step": 82650 }, { "epoch": 7.16, "learning_rate": 5.718457268416792e-05, "loss": 1.3022, "step": 82660 }, { "epoch": 7.16, "learning_rate": 5.7182404302070076e-05, "loss": 1.305, "step": 82670 }, { "epoch": 7.16, "learning_rate": 5.718023591997224e-05, "loss": 1.4252, "step": 82680 }, { "epoch": 7.16, "learning_rate": 5.717806753787441e-05, "loss": 1.5625, "step": 82690 }, { "epoch": 7.16, "learning_rate": 5.717589915577656e-05, "loss": 1.2916, "step": 82700 }, { "epoch": 7.16, "learning_rate": 5.717373077367872e-05, "loss": 1.3185, "step": 82710 }, { "epoch": 7.16, "learning_rate": 5.717156239158089e-05, "loss": 1.2615, "step": 82720 }, { "epoch": 7.16, "learning_rate": 5.7169394009483054e-05, "loss": 1.6063, "step": 82730 }, { "epoch": 7.17, "learning_rate": 5.7167225627385214e-05, "loss": 1.4258, "step": 82740 }, { "epoch": 7.17, "learning_rate": 5.716505724528738e-05, "loss": 1.4298, "step": 82750 }, { "epoch": 7.17, "learning_rate": 5.716288886318954e-05, "loss": 1.3414, "step": 82760 }, { "epoch": 7.17, "learning_rate": 5.7160720481091706e-05, "loss": 1.3869, "step": 82770 }, { "epoch": 7.17, "learning_rate": 5.715855209899386e-05, "loss": 1.2509, "step": 82780 }, { "epoch": 7.17, "learning_rate": 5.7156383716896026e-05, "loss": 1.584, "step": 82790 }, { "epoch": 7.17, "learning_rate": 5.715421533479819e-05, "loss": 1.3733, "step": 82800 }, { "epoch": 7.17, "learning_rate": 5.715204695270035e-05, "loss": 1.439, "step": 82810 }, { "epoch": 7.17, "learning_rate": 5.714987857060252e-05, "loss": 1.6074, "step": 82820 }, { "epoch": 7.17, "learning_rate": 5.714771018850468e-05, "loss": 1.3688, "step": 82830 }, { "epoch": 7.17, "learning_rate": 5.7145541806406844e-05, "loss": 1.4253, "step": 82840 }, { "epoch": 7.18, "learning_rate": 5.714337342430901e-05, "loss": 1.5577, "step": 82850 }, { "epoch": 7.18, "learning_rate": 5.714120504221116e-05, "loss": 1.2845, "step": 82860 }, { "epoch": 7.18, "learning_rate": 5.713903666011333e-05, "loss": 1.3735, "step": 82870 }, { "epoch": 7.18, "learning_rate": 5.713686827801549e-05, "loss": 1.3698, "step": 82880 }, { "epoch": 7.18, "learning_rate": 5.7134699895917656e-05, "loss": 1.335, "step": 82890 }, { "epoch": 7.18, "learning_rate": 5.7132531513819815e-05, "loss": 1.5644, "step": 82900 }, { "epoch": 7.18, "learning_rate": 5.713036313172198e-05, "loss": 1.486, "step": 82910 }, { "epoch": 7.18, "learning_rate": 5.712819474962415e-05, "loss": 1.3134, "step": 82920 }, { "epoch": 7.18, "learning_rate": 5.71260263675263e-05, "loss": 1.438, "step": 82930 }, { "epoch": 7.18, "learning_rate": 5.712385798542847e-05, "loss": 1.3262, "step": 82940 }, { "epoch": 7.18, "learning_rate": 5.712168960333063e-05, "loss": 1.3701, "step": 82950 }, { "epoch": 7.18, "learning_rate": 5.711952122123279e-05, "loss": 1.3196, "step": 82960 }, { "epoch": 7.19, "learning_rate": 5.711735283913495e-05, "loss": 1.3748, "step": 82970 }, { "epoch": 7.19, "learning_rate": 5.711518445703712e-05, "loss": 1.5781, "step": 82980 }, { "epoch": 7.19, "learning_rate": 5.7113016074939286e-05, "loss": 1.3813, "step": 82990 }, { "epoch": 7.19, "learning_rate": 5.7110847692841445e-05, "loss": 1.3905, "step": 83000 }, { "epoch": 7.19, "learning_rate": 5.7108679310743605e-05, "loss": 1.357, "step": 83010 }, { "epoch": 7.19, "learning_rate": 5.7106510928645765e-05, "loss": 1.5839, "step": 83020 }, { "epoch": 7.19, "learning_rate": 5.710434254654793e-05, "loss": 1.3845, "step": 83030 }, { "epoch": 7.19, "learning_rate": 5.710217416445009e-05, "loss": 1.2707, "step": 83040 }, { "epoch": 7.19, "learning_rate": 5.710000578235226e-05, "loss": 1.42, "step": 83050 }, { "epoch": 7.19, "learning_rate": 5.709783740025442e-05, "loss": 1.3209, "step": 83060 }, { "epoch": 7.19, "learning_rate": 5.709566901815658e-05, "loss": 1.4919, "step": 83070 }, { "epoch": 7.2, "learning_rate": 5.709350063605875e-05, "loss": 1.2633, "step": 83080 }, { "epoch": 7.2, "learning_rate": 5.70913322539609e-05, "loss": 1.5087, "step": 83090 }, { "epoch": 7.2, "learning_rate": 5.708916387186307e-05, "loss": 1.3688, "step": 83100 }, { "epoch": 7.2, "learning_rate": 5.708699548976523e-05, "loss": 1.3749, "step": 83110 }, { "epoch": 7.2, "learning_rate": 5.7084827107667395e-05, "loss": 1.3485, "step": 83120 }, { "epoch": 7.2, "learning_rate": 5.708265872556956e-05, "loss": 1.4252, "step": 83130 }, { "epoch": 7.2, "learning_rate": 5.708049034347172e-05, "loss": 1.4012, "step": 83140 }, { "epoch": 7.2, "learning_rate": 5.707832196137389e-05, "loss": 1.3324, "step": 83150 }, { "epoch": 7.2, "learning_rate": 5.707615357927604e-05, "loss": 1.3902, "step": 83160 }, { "epoch": 7.2, "learning_rate": 5.7073985197178206e-05, "loss": 1.223, "step": 83170 }, { "epoch": 7.2, "learning_rate": 5.7071816815080366e-05, "loss": 1.3226, "step": 83180 }, { "epoch": 7.2, "learning_rate": 5.706964843298253e-05, "loss": 1.3764, "step": 83190 }, { "epoch": 7.21, "learning_rate": 5.70674800508847e-05, "loss": 1.3014, "step": 83200 }, { "epoch": 7.21, "learning_rate": 5.706531166878686e-05, "loss": 1.5114, "step": 83210 }, { "epoch": 7.21, "learning_rate": 5.7063143286689025e-05, "loss": 1.3682, "step": 83220 }, { "epoch": 7.21, "learning_rate": 5.7060974904591184e-05, "loss": 1.379, "step": 83230 }, { "epoch": 7.21, "learning_rate": 5.7058806522493344e-05, "loss": 1.4987, "step": 83240 }, { "epoch": 7.21, "learning_rate": 5.7056638140395503e-05, "loss": 1.4211, "step": 83250 }, { "epoch": 7.21, "learning_rate": 5.705446975829767e-05, "loss": 1.555, "step": 83260 }, { "epoch": 7.21, "learning_rate": 5.705230137619983e-05, "loss": 1.2602, "step": 83270 }, { "epoch": 7.21, "learning_rate": 5.7050132994101996e-05, "loss": 1.5247, "step": 83280 }, { "epoch": 7.21, "learning_rate": 5.704796461200416e-05, "loss": 1.3288, "step": 83290 }, { "epoch": 7.21, "learning_rate": 5.704579622990632e-05, "loss": 1.6176, "step": 83300 }, { "epoch": 7.22, "learning_rate": 5.704362784780849e-05, "loss": 1.6004, "step": 83310 }, { "epoch": 7.22, "learning_rate": 5.704145946571064e-05, "loss": 1.2572, "step": 83320 }, { "epoch": 7.22, "learning_rate": 5.703929108361281e-05, "loss": 1.3628, "step": 83330 }, { "epoch": 7.22, "learning_rate": 5.703712270151497e-05, "loss": 1.5561, "step": 83340 }, { "epoch": 7.22, "learning_rate": 5.7034954319417133e-05, "loss": 1.4444, "step": 83350 }, { "epoch": 7.22, "learning_rate": 5.70327859373193e-05, "loss": 1.4807, "step": 83360 }, { "epoch": 7.22, "learning_rate": 5.703061755522146e-05, "loss": 1.3986, "step": 83370 }, { "epoch": 7.22, "learning_rate": 5.7028449173123626e-05, "loss": 1.3612, "step": 83380 }, { "epoch": 7.22, "learning_rate": 5.702628079102578e-05, "loss": 1.3034, "step": 83390 }, { "epoch": 7.22, "learning_rate": 5.7024112408927945e-05, "loss": 1.3433, "step": 83400 }, { "epoch": 7.22, "learning_rate": 5.7021944026830105e-05, "loss": 1.5471, "step": 83410 }, { "epoch": 7.22, "learning_rate": 5.701977564473227e-05, "loss": 1.6214, "step": 83420 }, { "epoch": 7.23, "learning_rate": 5.701760726263444e-05, "loss": 1.2735, "step": 83430 }, { "epoch": 7.23, "learning_rate": 5.70154388805366e-05, "loss": 1.2355, "step": 83440 }, { "epoch": 7.23, "learning_rate": 5.7013270498438764e-05, "loss": 1.4008, "step": 83450 }, { "epoch": 7.23, "learning_rate": 5.701110211634092e-05, "loss": 1.4179, "step": 83460 }, { "epoch": 7.23, "learning_rate": 5.700893373424308e-05, "loss": 1.308, "step": 83470 }, { "epoch": 7.23, "learning_rate": 5.700676535214524e-05, "loss": 1.3456, "step": 83480 }, { "epoch": 7.23, "learning_rate": 5.700459697004741e-05, "loss": 1.3485, "step": 83490 }, { "epoch": 7.23, "learning_rate": 5.7002428587949575e-05, "loss": 1.2773, "step": 83500 }, { "epoch": 7.23, "learning_rate": 5.7000260205851735e-05, "loss": 1.2842, "step": 83510 }, { "epoch": 7.23, "learning_rate": 5.69980918237539e-05, "loss": 1.4522, "step": 83520 }, { "epoch": 7.23, "learning_rate": 5.699592344165606e-05, "loss": 1.4467, "step": 83530 }, { "epoch": 7.24, "learning_rate": 5.699375505955823e-05, "loss": 1.4916, "step": 83540 }, { "epoch": 7.24, "learning_rate": 5.699158667746038e-05, "loss": 1.536, "step": 83550 }, { "epoch": 7.24, "learning_rate": 5.6989418295362546e-05, "loss": 1.282, "step": 83560 }, { "epoch": 7.24, "learning_rate": 5.698724991326471e-05, "loss": 1.1381, "step": 83570 }, { "epoch": 7.24, "learning_rate": 5.698508153116687e-05, "loss": 1.5743, "step": 83580 }, { "epoch": 7.24, "learning_rate": 5.698291314906904e-05, "loss": 1.3894, "step": 83590 }, { "epoch": 7.24, "learning_rate": 5.69807447669712e-05, "loss": 1.464, "step": 83600 }, { "epoch": 7.24, "learning_rate": 5.6978576384873365e-05, "loss": 1.3929, "step": 83610 }, { "epoch": 7.24, "learning_rate": 5.697640800277553e-05, "loss": 1.3288, "step": 83620 }, { "epoch": 7.24, "learning_rate": 5.6974239620677684e-05, "loss": 1.2394, "step": 83630 }, { "epoch": 7.24, "learning_rate": 5.697207123857985e-05, "loss": 1.6751, "step": 83640 }, { "epoch": 7.24, "learning_rate": 5.696990285648201e-05, "loss": 1.5188, "step": 83650 }, { "epoch": 7.25, "learning_rate": 5.6967734474384176e-05, "loss": 1.3018, "step": 83660 }, { "epoch": 7.25, "learning_rate": 5.6965566092286336e-05, "loss": 1.3999, "step": 83670 }, { "epoch": 7.25, "learning_rate": 5.69633977101885e-05, "loss": 1.4348, "step": 83680 }, { "epoch": 7.25, "learning_rate": 5.696122932809067e-05, "loss": 1.2888, "step": 83690 }, { "epoch": 7.25, "learning_rate": 5.695906094599282e-05, "loss": 1.5057, "step": 83700 }, { "epoch": 7.25, "learning_rate": 5.695689256389499e-05, "loss": 1.2649, "step": 83710 }, { "epoch": 7.25, "learning_rate": 5.695472418179715e-05, "loss": 1.4277, "step": 83720 }, { "epoch": 7.25, "learning_rate": 5.6952555799699314e-05, "loss": 1.5059, "step": 83730 }, { "epoch": 7.25, "learning_rate": 5.6950387417601474e-05, "loss": 1.5141, "step": 83740 }, { "epoch": 7.25, "learning_rate": 5.694821903550364e-05, "loss": 1.5044, "step": 83750 }, { "epoch": 7.25, "learning_rate": 5.6946050653405806e-05, "loss": 1.3968, "step": 83760 }, { "epoch": 7.26, "learning_rate": 5.6943882271307966e-05, "loss": 1.4543, "step": 83770 }, { "epoch": 7.26, "learning_rate": 5.694171388921012e-05, "loss": 1.4086, "step": 83780 }, { "epoch": 7.26, "learning_rate": 5.6939545507112285e-05, "loss": 1.4231, "step": 83790 }, { "epoch": 7.26, "learning_rate": 5.693737712501445e-05, "loss": 1.3123, "step": 83800 }, { "epoch": 7.26, "learning_rate": 5.693520874291661e-05, "loss": 1.3673, "step": 83810 }, { "epoch": 7.26, "learning_rate": 5.693304036081878e-05, "loss": 1.4308, "step": 83820 }, { "epoch": 7.26, "learning_rate": 5.693087197872094e-05, "loss": 1.3019, "step": 83830 }, { "epoch": 7.26, "learning_rate": 5.6928703596623104e-05, "loss": 1.2823, "step": 83840 }, { "epoch": 7.26, "learning_rate": 5.692653521452527e-05, "loss": 1.3418, "step": 83850 }, { "epoch": 7.26, "learning_rate": 5.692436683242742e-05, "loss": 1.3957, "step": 83860 }, { "epoch": 7.26, "learning_rate": 5.692219845032959e-05, "loss": 1.5064, "step": 83870 }, { "epoch": 7.26, "learning_rate": 5.692003006823175e-05, "loss": 1.3434, "step": 83880 }, { "epoch": 7.27, "learning_rate": 5.6917861686133915e-05, "loss": 1.3026, "step": 83890 }, { "epoch": 7.27, "learning_rate": 5.6915693304036075e-05, "loss": 1.403, "step": 83900 }, { "epoch": 7.27, "learning_rate": 5.691352492193824e-05, "loss": 1.4012, "step": 83910 }, { "epoch": 7.27, "learning_rate": 5.691135653984041e-05, "loss": 1.3683, "step": 83920 }, { "epoch": 7.27, "learning_rate": 5.690918815774256e-05, "loss": 1.357, "step": 83930 }, { "epoch": 7.27, "learning_rate": 5.690701977564473e-05, "loss": 1.5876, "step": 83940 }, { "epoch": 7.27, "learning_rate": 5.6904851393546887e-05, "loss": 1.233, "step": 83950 }, { "epoch": 7.27, "learning_rate": 5.690268301144905e-05, "loss": 1.3525, "step": 83960 }, { "epoch": 7.27, "learning_rate": 5.690051462935121e-05, "loss": 1.4249, "step": 83970 }, { "epoch": 7.27, "learning_rate": 5.689834624725338e-05, "loss": 1.3188, "step": 83980 }, { "epoch": 7.27, "learning_rate": 5.6896177865155545e-05, "loss": 1.2843, "step": 83990 }, { "epoch": 7.27, "learning_rate": 5.6894009483057705e-05, "loss": 1.4682, "step": 84000 }, { "epoch": 7.28, "learning_rate": 5.6891841100959865e-05, "loss": 1.4135, "step": 84010 }, { "epoch": 7.28, "learning_rate": 5.6889672718862024e-05, "loss": 1.6341, "step": 84020 }, { "epoch": 7.28, "learning_rate": 5.688750433676419e-05, "loss": 1.3419, "step": 84030 }, { "epoch": 7.28, "learning_rate": 5.688533595466635e-05, "loss": 1.2233, "step": 84040 }, { "epoch": 7.28, "learning_rate": 5.688316757256852e-05, "loss": 1.3742, "step": 84050 }, { "epoch": 7.28, "learning_rate": 5.688099919047068e-05, "loss": 1.4438, "step": 84060 }, { "epoch": 7.28, "learning_rate": 5.687883080837284e-05, "loss": 1.3691, "step": 84070 }, { "epoch": 7.28, "learning_rate": 5.687666242627501e-05, "loss": 1.5091, "step": 84080 }, { "epoch": 7.28, "learning_rate": 5.687449404417716e-05, "loss": 1.3485, "step": 84090 }, { "epoch": 7.28, "learning_rate": 5.687232566207933e-05, "loss": 1.4763, "step": 84100 }, { "epoch": 7.28, "learning_rate": 5.687015727998149e-05, "loss": 1.4529, "step": 84110 }, { "epoch": 7.29, "learning_rate": 5.6867988897883654e-05, "loss": 1.3824, "step": 84120 }, { "epoch": 7.29, "learning_rate": 5.686582051578582e-05, "loss": 1.3759, "step": 84130 }, { "epoch": 7.29, "learning_rate": 5.686365213368798e-05, "loss": 1.3282, "step": 84140 }, { "epoch": 7.29, "learning_rate": 5.686148375159015e-05, "loss": 1.2894, "step": 84150 }, { "epoch": 7.29, "learning_rate": 5.68593153694923e-05, "loss": 1.3997, "step": 84160 }, { "epoch": 7.29, "learning_rate": 5.6857146987394466e-05, "loss": 1.5466, "step": 84170 }, { "epoch": 7.29, "learning_rate": 5.6854978605296626e-05, "loss": 1.5101, "step": 84180 }, { "epoch": 7.29, "learning_rate": 5.685281022319879e-05, "loss": 1.3689, "step": 84190 }, { "epoch": 7.29, "learning_rate": 5.685064184110096e-05, "loss": 1.4094, "step": 84200 }, { "epoch": 7.29, "learning_rate": 5.684847345900312e-05, "loss": 1.5848, "step": 84210 }, { "epoch": 7.29, "learning_rate": 5.6846305076905284e-05, "loss": 1.4328, "step": 84220 }, { "epoch": 7.29, "learning_rate": 5.6844136694807444e-05, "loss": 1.4385, "step": 84230 }, { "epoch": 7.3, "learning_rate": 5.6841968312709604e-05, "loss": 1.4495, "step": 84240 }, { "epoch": 7.3, "learning_rate": 5.683979993061176e-05, "loss": 1.5424, "step": 84250 }, { "epoch": 7.3, "learning_rate": 5.683763154851393e-05, "loss": 1.594, "step": 84260 }, { "epoch": 7.3, "learning_rate": 5.6835463166416096e-05, "loss": 1.1637, "step": 84270 }, { "epoch": 7.3, "learning_rate": 5.6833294784318256e-05, "loss": 1.4196, "step": 84280 }, { "epoch": 7.3, "learning_rate": 5.683112640222042e-05, "loss": 1.3285, "step": 84290 }, { "epoch": 7.3, "learning_rate": 5.682895802012258e-05, "loss": 1.3663, "step": 84300 }, { "epoch": 7.3, "learning_rate": 5.682678963802475e-05, "loss": 1.4148, "step": 84310 }, { "epoch": 7.3, "learning_rate": 5.68246212559269e-05, "loss": 1.5669, "step": 84320 }, { "epoch": 7.3, "learning_rate": 5.682245287382907e-05, "loss": 1.5328, "step": 84330 }, { "epoch": 7.3, "learning_rate": 5.682028449173123e-05, "loss": 1.5157, "step": 84340 }, { "epoch": 7.31, "learning_rate": 5.681811610963339e-05, "loss": 1.4103, "step": 84350 }, { "epoch": 7.31, "learning_rate": 5.681594772753556e-05, "loss": 1.5175, "step": 84360 }, { "epoch": 7.31, "learning_rate": 5.681377934543772e-05, "loss": 1.33, "step": 84370 }, { "epoch": 7.31, "learning_rate": 5.6811610963339886e-05, "loss": 1.476, "step": 84380 }, { "epoch": 7.31, "learning_rate": 5.6809442581242045e-05, "loss": 1.4916, "step": 84390 }, { "epoch": 7.31, "learning_rate": 5.6807274199144205e-05, "loss": 1.497, "step": 84400 }, { "epoch": 7.31, "learning_rate": 5.6805105817046364e-05, "loss": 1.5185, "step": 84410 }, { "epoch": 7.31, "learning_rate": 5.680293743494853e-05, "loss": 1.6644, "step": 84420 }, { "epoch": 7.31, "learning_rate": 5.68007690528507e-05, "loss": 1.4319, "step": 84430 }, { "epoch": 7.31, "learning_rate": 5.679860067075286e-05, "loss": 1.4099, "step": 84440 }, { "epoch": 7.31, "learning_rate": 5.679643228865502e-05, "loss": 1.2996, "step": 84450 }, { "epoch": 7.31, "learning_rate": 5.679426390655718e-05, "loss": 1.4363, "step": 84460 }, { "epoch": 7.32, "learning_rate": 5.679209552445934e-05, "loss": 1.4905, "step": 84470 }, { "epoch": 7.32, "learning_rate": 5.67899271423615e-05, "loss": 1.397, "step": 84480 }, { "epoch": 7.32, "learning_rate": 5.678775876026367e-05, "loss": 1.3899, "step": 84490 }, { "epoch": 7.32, "learning_rate": 5.6785590378165835e-05, "loss": 1.3874, "step": 84500 }, { "epoch": 7.32, "learning_rate": 5.6783421996067994e-05, "loss": 1.6769, "step": 84510 }, { "epoch": 7.32, "learning_rate": 5.678125361397016e-05, "loss": 1.1784, "step": 84520 }, { "epoch": 7.32, "learning_rate": 5.677908523187232e-05, "loss": 1.3118, "step": 84530 }, { "epoch": 7.32, "learning_rate": 5.677691684977449e-05, "loss": 1.6042, "step": 84540 }, { "epoch": 7.32, "learning_rate": 5.677474846767664e-05, "loss": 1.4096, "step": 84550 }, { "epoch": 7.32, "learning_rate": 5.6772580085578806e-05, "loss": 1.2978, "step": 84560 }, { "epoch": 7.32, "learning_rate": 5.677041170348097e-05, "loss": 1.3612, "step": 84570 }, { "epoch": 7.33, "learning_rate": 5.676824332138313e-05, "loss": 1.3079, "step": 84580 }, { "epoch": 7.33, "learning_rate": 5.67660749392853e-05, "loss": 1.4004, "step": 84590 }, { "epoch": 7.33, "learning_rate": 5.676390655718746e-05, "loss": 1.4286, "step": 84600 }, { "epoch": 7.33, "learning_rate": 5.6761738175089624e-05, "loss": 1.3698, "step": 84610 }, { "epoch": 7.33, "learning_rate": 5.675956979299179e-05, "loss": 1.4476, "step": 84620 }, { "epoch": 7.33, "learning_rate": 5.6757401410893944e-05, "loss": 1.4538, "step": 84630 }, { "epoch": 7.33, "learning_rate": 5.675523302879611e-05, "loss": 1.4086, "step": 84640 }, { "epoch": 7.33, "learning_rate": 5.675306464669827e-05, "loss": 1.4593, "step": 84650 }, { "epoch": 7.33, "learning_rate": 5.6750896264600436e-05, "loss": 1.3433, "step": 84660 }, { "epoch": 7.33, "learning_rate": 5.6748727882502596e-05, "loss": 1.4732, "step": 84670 }, { "epoch": 7.33, "learning_rate": 5.674655950040476e-05, "loss": 1.3419, "step": 84680 }, { "epoch": 7.33, "learning_rate": 5.674439111830693e-05, "loss": 1.2812, "step": 84690 }, { "epoch": 7.34, "learning_rate": 5.674222273620908e-05, "loss": 1.5372, "step": 84700 }, { "epoch": 7.34, "learning_rate": 5.674005435411125e-05, "loss": 1.4192, "step": 84710 }, { "epoch": 7.34, "learning_rate": 5.673788597201341e-05, "loss": 1.2507, "step": 84720 }, { "epoch": 7.34, "learning_rate": 5.6735717589915574e-05, "loss": 1.588, "step": 84730 }, { "epoch": 7.34, "learning_rate": 5.673354920781773e-05, "loss": 1.5894, "step": 84740 }, { "epoch": 7.34, "learning_rate": 5.67313808257199e-05, "loss": 1.3888, "step": 84750 }, { "epoch": 7.34, "learning_rate": 5.6729212443622066e-05, "loss": 1.5162, "step": 84760 }, { "epoch": 7.34, "learning_rate": 5.6727044061524226e-05, "loss": 1.2268, "step": 84770 }, { "epoch": 7.34, "learning_rate": 5.6724875679426385e-05, "loss": 1.2862, "step": 84780 }, { "epoch": 7.34, "learning_rate": 5.6722707297328545e-05, "loss": 1.3856, "step": 84790 }, { "epoch": 7.34, "learning_rate": 5.672053891523071e-05, "loss": 1.3285, "step": 84800 }, { "epoch": 7.35, "learning_rate": 5.671837053313287e-05, "loss": 1.3058, "step": 84810 }, { "epoch": 7.35, "learning_rate": 5.671620215103504e-05, "loss": 1.3969, "step": 84820 }, { "epoch": 7.35, "learning_rate": 5.6714033768937204e-05, "loss": 1.3145, "step": 84830 }, { "epoch": 7.35, "learning_rate": 5.6711865386839363e-05, "loss": 1.3659, "step": 84840 }, { "epoch": 7.35, "learning_rate": 5.670969700474153e-05, "loss": 1.3295, "step": 84850 }, { "epoch": 7.35, "learning_rate": 5.670752862264368e-05, "loss": 1.4368, "step": 84860 }, { "epoch": 7.35, "learning_rate": 5.670536024054585e-05, "loss": 1.5207, "step": 84870 }, { "epoch": 7.35, "learning_rate": 5.670319185844801e-05, "loss": 1.3343, "step": 84880 }, { "epoch": 7.35, "learning_rate": 5.6701023476350175e-05, "loss": 1.3006, "step": 84890 }, { "epoch": 7.35, "learning_rate": 5.6698855094252335e-05, "loss": 1.3879, "step": 84900 }, { "epoch": 7.35, "learning_rate": 5.66966867121545e-05, "loss": 1.2848, "step": 84910 }, { "epoch": 7.35, "learning_rate": 5.669451833005667e-05, "loss": 1.3726, "step": 84920 }, { "epoch": 7.36, "learning_rate": 5.669234994795882e-05, "loss": 1.3918, "step": 84930 }, { "epoch": 7.36, "learning_rate": 5.669018156586099e-05, "loss": 1.2526, "step": 84940 }, { "epoch": 7.36, "learning_rate": 5.6688013183763146e-05, "loss": 1.521, "step": 84950 }, { "epoch": 7.36, "learning_rate": 5.668584480166531e-05, "loss": 1.4761, "step": 84960 }, { "epoch": 7.36, "learning_rate": 5.668367641956747e-05, "loss": 1.4064, "step": 84970 }, { "epoch": 7.36, "learning_rate": 5.668150803746964e-05, "loss": 1.3749, "step": 84980 }, { "epoch": 7.36, "learning_rate": 5.6679339655371805e-05, "loss": 1.4409, "step": 84990 }, { "epoch": 7.36, "learning_rate": 5.6677171273273965e-05, "loss": 1.4235, "step": 85000 }, { "epoch": 7.36, "learning_rate": 5.6675002891176124e-05, "loss": 1.4556, "step": 85010 }, { "epoch": 7.36, "learning_rate": 5.6672834509078284e-05, "loss": 1.5602, "step": 85020 }, { "epoch": 7.36, "learning_rate": 5.667066612698045e-05, "loss": 1.3374, "step": 85030 }, { "epoch": 7.37, "learning_rate": 5.666849774488261e-05, "loss": 1.4132, "step": 85040 }, { "epoch": 7.37, "learning_rate": 5.6666329362784776e-05, "loss": 1.3554, "step": 85050 }, { "epoch": 7.37, "learning_rate": 5.666416098068694e-05, "loss": 1.391, "step": 85060 }, { "epoch": 7.37, "learning_rate": 5.66619925985891e-05, "loss": 1.3444, "step": 85070 }, { "epoch": 7.37, "learning_rate": 5.665982421649127e-05, "loss": 1.3311, "step": 85080 }, { "epoch": 7.37, "learning_rate": 5.665765583439342e-05, "loss": 1.433, "step": 85090 }, { "epoch": 7.37, "learning_rate": 5.665548745229559e-05, "loss": 1.3812, "step": 85100 }, { "epoch": 7.37, "learning_rate": 5.665331907019775e-05, "loss": 1.3784, "step": 85110 }, { "epoch": 7.37, "learning_rate": 5.6651150688099914e-05, "loss": 1.5941, "step": 85120 }, { "epoch": 7.37, "learning_rate": 5.664898230600208e-05, "loss": 1.3658, "step": 85130 }, { "epoch": 7.37, "learning_rate": 5.664681392390424e-05, "loss": 1.516, "step": 85140 }, { "epoch": 7.37, "learning_rate": 5.6644645541806406e-05, "loss": 1.2024, "step": 85150 }, { "epoch": 7.38, "learning_rate": 5.6642477159708566e-05, "loss": 1.2717, "step": 85160 }, { "epoch": 7.38, "learning_rate": 5.6640308777610726e-05, "loss": 1.3798, "step": 85170 }, { "epoch": 7.38, "learning_rate": 5.6638140395512885e-05, "loss": 1.4197, "step": 85180 }, { "epoch": 7.38, "learning_rate": 5.663597201341505e-05, "loss": 1.2731, "step": 85190 }, { "epoch": 7.38, "learning_rate": 5.663380363131722e-05, "loss": 1.5277, "step": 85200 }, { "epoch": 7.38, "learning_rate": 5.663163524921938e-05, "loss": 1.4856, "step": 85210 }, { "epoch": 7.38, "learning_rate": 5.6629466867121544e-05, "loss": 1.5265, "step": 85220 }, { "epoch": 7.38, "learning_rate": 5.6627298485023704e-05, "loss": 1.5103, "step": 85230 }, { "epoch": 7.38, "learning_rate": 5.662513010292586e-05, "loss": 1.3349, "step": 85240 }, { "epoch": 7.38, "learning_rate": 5.662296172082802e-05, "loss": 1.4051, "step": 85250 }, { "epoch": 7.38, "learning_rate": 5.662079333873019e-05, "loss": 1.3177, "step": 85260 }, { "epoch": 7.38, "learning_rate": 5.6618624956632356e-05, "loss": 1.6218, "step": 85270 }, { "epoch": 7.39, "learning_rate": 5.6616456574534515e-05, "loss": 1.449, "step": 85280 }, { "epoch": 7.39, "learning_rate": 5.661428819243668e-05, "loss": 1.3724, "step": 85290 }, { "epoch": 7.39, "learning_rate": 5.661211981033884e-05, "loss": 1.3861, "step": 85300 }, { "epoch": 7.39, "learning_rate": 5.660995142824101e-05, "loss": 1.3988, "step": 85310 }, { "epoch": 7.39, "learning_rate": 5.660778304614316e-05, "loss": 1.4129, "step": 85320 }, { "epoch": 7.39, "learning_rate": 5.660561466404533e-05, "loss": 1.3768, "step": 85330 }, { "epoch": 7.39, "learning_rate": 5.660344628194749e-05, "loss": 1.2735, "step": 85340 }, { "epoch": 7.39, "learning_rate": 5.660127789984965e-05, "loss": 1.4166, "step": 85350 }, { "epoch": 7.39, "learning_rate": 5.659910951775182e-05, "loss": 1.4152, "step": 85360 }, { "epoch": 7.39, "learning_rate": 5.659694113565398e-05, "loss": 1.5221, "step": 85370 }, { "epoch": 7.39, "learning_rate": 5.6594772753556145e-05, "loss": 1.3538, "step": 85380 }, { "epoch": 7.4, "learning_rate": 5.659260437145831e-05, "loss": 1.5174, "step": 85390 }, { "epoch": 7.4, "learning_rate": 5.6590435989360465e-05, "loss": 1.5015, "step": 85400 }, { "epoch": 7.4, "learning_rate": 5.6588267607262624e-05, "loss": 1.4262, "step": 85410 }, { "epoch": 7.4, "learning_rate": 5.658609922516479e-05, "loss": 1.3251, "step": 85420 }, { "epoch": 7.4, "learning_rate": 5.658393084306696e-05, "loss": 1.2604, "step": 85430 }, { "epoch": 7.4, "learning_rate": 5.6581762460969117e-05, "loss": 1.4318, "step": 85440 }, { "epoch": 7.4, "learning_rate": 5.657959407887128e-05, "loss": 1.2924, "step": 85450 }, { "epoch": 7.4, "learning_rate": 5.657742569677344e-05, "loss": 1.3658, "step": 85460 }, { "epoch": 7.4, "learning_rate": 5.65752573146756e-05, "loss": 1.4272, "step": 85470 }, { "epoch": 7.4, "learning_rate": 5.657308893257776e-05, "loss": 1.5539, "step": 85480 }, { "epoch": 7.4, "learning_rate": 5.657092055047993e-05, "loss": 1.4933, "step": 85490 }, { "epoch": 7.4, "learning_rate": 5.6568752168382095e-05, "loss": 1.3179, "step": 85500 }, { "epoch": 7.41, "learning_rate": 5.6566583786284254e-05, "loss": 1.6147, "step": 85510 }, { "epoch": 7.41, "learning_rate": 5.656441540418642e-05, "loss": 1.4706, "step": 85520 }, { "epoch": 7.41, "learning_rate": 5.656224702208858e-05, "loss": 1.3003, "step": 85530 }, { "epoch": 7.41, "learning_rate": 5.6560078639990747e-05, "loss": 1.4482, "step": 85540 }, { "epoch": 7.41, "learning_rate": 5.65579102578929e-05, "loss": 1.431, "step": 85550 }, { "epoch": 7.41, "learning_rate": 5.6555741875795066e-05, "loss": 1.4283, "step": 85560 }, { "epoch": 7.41, "learning_rate": 5.655357349369723e-05, "loss": 1.2787, "step": 85570 }, { "epoch": 7.41, "learning_rate": 5.655140511159939e-05, "loss": 1.2384, "step": 85580 }, { "epoch": 7.41, "learning_rate": 5.654923672950156e-05, "loss": 1.406, "step": 85590 }, { "epoch": 7.41, "learning_rate": 5.654706834740372e-05, "loss": 1.3444, "step": 85600 }, { "epoch": 7.41, "learning_rate": 5.6544899965305884e-05, "loss": 1.539, "step": 85610 }, { "epoch": 7.42, "learning_rate": 5.654273158320805e-05, "loss": 1.3805, "step": 85620 }, { "epoch": 7.42, "learning_rate": 5.6540563201110203e-05, "loss": 1.4464, "step": 85630 }, { "epoch": 7.42, "learning_rate": 5.653839481901237e-05, "loss": 1.4312, "step": 85640 }, { "epoch": 7.42, "learning_rate": 5.653622643691453e-05, "loss": 1.3715, "step": 85650 }, { "epoch": 7.42, "learning_rate": 5.6534058054816696e-05, "loss": 1.4234, "step": 85660 }, { "epoch": 7.42, "learning_rate": 5.6531889672718855e-05, "loss": 1.495, "step": 85670 }, { "epoch": 7.42, "learning_rate": 5.652972129062102e-05, "loss": 1.2076, "step": 85680 }, { "epoch": 7.42, "learning_rate": 5.652755290852319e-05, "loss": 1.3785, "step": 85690 }, { "epoch": 7.42, "learning_rate": 5.652538452642534e-05, "loss": 1.2674, "step": 85700 }, { "epoch": 7.42, "learning_rate": 5.652321614432751e-05, "loss": 1.4076, "step": 85710 }, { "epoch": 7.42, "learning_rate": 5.652104776222967e-05, "loss": 1.2658, "step": 85720 }, { "epoch": 7.42, "learning_rate": 5.6518879380131833e-05, "loss": 1.2792, "step": 85730 }, { "epoch": 7.43, "learning_rate": 5.651671099803399e-05, "loss": 1.2989, "step": 85740 }, { "epoch": 7.43, "learning_rate": 5.651454261593616e-05, "loss": 1.4613, "step": 85750 }, { "epoch": 7.43, "learning_rate": 5.6512374233838326e-05, "loss": 1.3183, "step": 85760 }, { "epoch": 7.43, "learning_rate": 5.6510205851740485e-05, "loss": 1.2567, "step": 85770 }, { "epoch": 7.43, "learning_rate": 5.6508037469642645e-05, "loss": 1.3245, "step": 85780 }, { "epoch": 7.43, "learning_rate": 5.6505869087544805e-05, "loss": 1.4696, "step": 85790 }, { "epoch": 7.43, "learning_rate": 5.650370070544697e-05, "loss": 1.3413, "step": 85800 }, { "epoch": 7.43, "learning_rate": 5.650153232334913e-05, "loss": 1.4, "step": 85810 }, { "epoch": 7.43, "learning_rate": 5.64993639412513e-05, "loss": 1.5609, "step": 85820 }, { "epoch": 7.43, "learning_rate": 5.6497195559153463e-05, "loss": 1.3582, "step": 85830 }, { "epoch": 7.43, "learning_rate": 5.649502717705562e-05, "loss": 1.349, "step": 85840 }, { "epoch": 7.44, "learning_rate": 5.649285879495779e-05, "loss": 1.4026, "step": 85850 }, { "epoch": 7.44, "learning_rate": 5.649069041285994e-05, "loss": 1.3478, "step": 85860 }, { "epoch": 7.44, "learning_rate": 5.648852203076211e-05, "loss": 1.4058, "step": 85870 }, { "epoch": 7.44, "learning_rate": 5.648635364866427e-05, "loss": 1.3428, "step": 85880 }, { "epoch": 7.44, "learning_rate": 5.6484185266566435e-05, "loss": 1.4219, "step": 85890 }, { "epoch": 7.44, "learning_rate": 5.64820168844686e-05, "loss": 1.2417, "step": 85900 }, { "epoch": 7.44, "learning_rate": 5.647984850237076e-05, "loss": 1.4084, "step": 85910 }, { "epoch": 7.44, "learning_rate": 5.647768012027293e-05, "loss": 1.5264, "step": 85920 }, { "epoch": 7.44, "learning_rate": 5.647551173817509e-05, "loss": 1.3902, "step": 85930 }, { "epoch": 7.44, "learning_rate": 5.6473343356077246e-05, "loss": 1.414, "step": 85940 }, { "epoch": 7.44, "learning_rate": 5.6471174973979406e-05, "loss": 1.3712, "step": 85950 }, { "epoch": 7.44, "learning_rate": 5.646900659188157e-05, "loss": 1.5203, "step": 85960 }, { "epoch": 7.45, "learning_rate": 5.646683820978373e-05, "loss": 1.4712, "step": 85970 }, { "epoch": 7.45, "learning_rate": 5.64646698276859e-05, "loss": 1.5691, "step": 85980 }, { "epoch": 7.45, "learning_rate": 5.6462501445588065e-05, "loss": 1.224, "step": 85990 }, { "epoch": 7.45, "learning_rate": 5.6460333063490224e-05, "loss": 1.4549, "step": 86000 }, { "epoch": 7.45, "learning_rate": 5.6458164681392384e-05, "loss": 1.1675, "step": 86010 }, { "epoch": 7.45, "learning_rate": 5.6455996299294544e-05, "loss": 1.285, "step": 86020 }, { "epoch": 7.45, "learning_rate": 5.645382791719671e-05, "loss": 1.3775, "step": 86030 }, { "epoch": 7.45, "learning_rate": 5.645165953509887e-05, "loss": 1.5379, "step": 86040 }, { "epoch": 7.45, "learning_rate": 5.6449491153001036e-05, "loss": 1.375, "step": 86050 }, { "epoch": 7.45, "learning_rate": 5.64473227709032e-05, "loss": 1.2938, "step": 86060 }, { "epoch": 7.45, "learning_rate": 5.644515438880536e-05, "loss": 1.2631, "step": 86070 }, { "epoch": 7.46, "learning_rate": 5.644298600670753e-05, "loss": 1.5018, "step": 86080 }, { "epoch": 7.46, "learning_rate": 5.644081762460968e-05, "loss": 1.427, "step": 86090 }, { "epoch": 7.46, "learning_rate": 5.643864924251185e-05, "loss": 1.1042, "step": 86100 }, { "epoch": 7.46, "learning_rate": 5.643648086041401e-05, "loss": 1.4813, "step": 86110 }, { "epoch": 7.46, "learning_rate": 5.6434312478316174e-05, "loss": 1.3268, "step": 86120 }, { "epoch": 7.46, "learning_rate": 5.643214409621834e-05, "loss": 1.2161, "step": 86130 }, { "epoch": 7.46, "learning_rate": 5.64299757141205e-05, "loss": 1.3674, "step": 86140 }, { "epoch": 7.46, "learning_rate": 5.6427807332022666e-05, "loss": 1.4277, "step": 86150 }, { "epoch": 7.46, "learning_rate": 5.6425638949924826e-05, "loss": 1.2253, "step": 86160 }, { "epoch": 7.46, "learning_rate": 5.6423470567826985e-05, "loss": 1.4735, "step": 86170 }, { "epoch": 7.46, "learning_rate": 5.6421302185729145e-05, "loss": 1.5722, "step": 86180 }, { "epoch": 7.46, "learning_rate": 5.641913380363131e-05, "loss": 1.4246, "step": 86190 }, { "epoch": 7.47, "learning_rate": 5.641696542153348e-05, "loss": 1.3088, "step": 86200 }, { "epoch": 7.47, "learning_rate": 5.641479703943564e-05, "loss": 1.425, "step": 86210 }, { "epoch": 7.47, "learning_rate": 5.6412628657337804e-05, "loss": 1.4182, "step": 86220 }, { "epoch": 7.47, "learning_rate": 5.641046027523996e-05, "loss": 1.4184, "step": 86230 }, { "epoch": 7.47, "learning_rate": 5.640829189314212e-05, "loss": 1.4699, "step": 86240 }, { "epoch": 7.47, "learning_rate": 5.640612351104428e-05, "loss": 1.3671, "step": 86250 }, { "epoch": 7.47, "learning_rate": 5.640395512894645e-05, "loss": 1.4394, "step": 86260 }, { "epoch": 7.47, "learning_rate": 5.6401786746848615e-05, "loss": 1.3511, "step": 86270 }, { "epoch": 7.47, "learning_rate": 5.6399618364750775e-05, "loss": 1.295, "step": 86280 }, { "epoch": 7.47, "learning_rate": 5.639744998265294e-05, "loss": 1.4264, "step": 86290 }, { "epoch": 7.47, "learning_rate": 5.63952816005551e-05, "loss": 1.2854, "step": 86300 }, { "epoch": 7.47, "learning_rate": 5.639311321845727e-05, "loss": 1.3809, "step": 86310 }, { "epoch": 7.48, "learning_rate": 5.639094483635942e-05, "loss": 1.2633, "step": 86320 }, { "epoch": 7.48, "learning_rate": 5.6388776454261587e-05, "loss": 1.4392, "step": 86330 }, { "epoch": 7.48, "learning_rate": 5.638660807216375e-05, "loss": 1.2331, "step": 86340 }, { "epoch": 7.48, "learning_rate": 5.638443969006591e-05, "loss": 1.5493, "step": 86350 }, { "epoch": 7.48, "learning_rate": 5.638227130796808e-05, "loss": 1.5619, "step": 86360 }, { "epoch": 7.48, "learning_rate": 5.638010292587024e-05, "loss": 1.2328, "step": 86370 }, { "epoch": 7.48, "learning_rate": 5.6377934543772405e-05, "loss": 1.2783, "step": 86380 }, { "epoch": 7.48, "learning_rate": 5.637576616167457e-05, "loss": 1.552, "step": 86390 }, { "epoch": 7.48, "learning_rate": 5.6373597779576724e-05, "loss": 1.3028, "step": 86400 }, { "epoch": 7.48, "learning_rate": 5.637142939747889e-05, "loss": 1.6335, "step": 86410 }, { "epoch": 7.48, "learning_rate": 5.636926101538105e-05, "loss": 1.3851, "step": 86420 }, { "epoch": 7.49, "learning_rate": 5.6367092633283217e-05, "loss": 1.368, "step": 86430 }, { "epoch": 7.49, "learning_rate": 5.6364924251185376e-05, "loss": 1.3586, "step": 86440 }, { "epoch": 7.49, "learning_rate": 5.636275586908754e-05, "loss": 1.3154, "step": 86450 }, { "epoch": 7.49, "learning_rate": 5.636058748698971e-05, "loss": 1.4693, "step": 86460 }, { "epoch": 7.49, "learning_rate": 5.635841910489186e-05, "loss": 1.6649, "step": 86470 }, { "epoch": 7.49, "learning_rate": 5.635625072279402e-05, "loss": 1.2012, "step": 86480 }, { "epoch": 7.49, "learning_rate": 5.635408234069619e-05, "loss": 1.2729, "step": 86490 }, { "epoch": 7.49, "learning_rate": 5.6351913958598354e-05, "loss": 1.4385, "step": 86500 }, { "epoch": 7.49, "learning_rate": 5.6349745576500514e-05, "loss": 1.3619, "step": 86510 }, { "epoch": 7.49, "learning_rate": 5.634757719440268e-05, "loss": 1.5802, "step": 86520 }, { "epoch": 7.49, "learning_rate": 5.634540881230484e-05, "loss": 1.3068, "step": 86530 }, { "epoch": 7.49, "learning_rate": 5.6343240430207006e-05, "loss": 1.347, "step": 86540 }, { "epoch": 7.5, "learning_rate": 5.634107204810916e-05, "loss": 1.202, "step": 86550 }, { "epoch": 7.5, "learning_rate": 5.6338903666011325e-05, "loss": 1.2146, "step": 86560 }, { "epoch": 7.5, "learning_rate": 5.633673528391349e-05, "loss": 1.3738, "step": 86570 }, { "epoch": 7.5, "learning_rate": 5.633456690181565e-05, "loss": 1.3112, "step": 86580 }, { "epoch": 7.5, "learning_rate": 5.633239851971782e-05, "loss": 1.3544, "step": 86590 }, { "epoch": 7.5, "learning_rate": 5.633023013761998e-05, "loss": 1.2919, "step": 86600 }, { "epoch": 7.5, "learning_rate": 5.6328061755522144e-05, "loss": 1.5194, "step": 86610 }, { "epoch": 7.5, "learning_rate": 5.632589337342431e-05, "loss": 1.462, "step": 86620 }, { "epoch": 7.5, "learning_rate": 5.632372499132646e-05, "loss": 1.588, "step": 86630 }, { "epoch": 7.5, "learning_rate": 5.632155660922863e-05, "loss": 1.5121, "step": 86640 }, { "epoch": 7.5, "learning_rate": 5.631938822713079e-05, "loss": 1.4162, "step": 86650 }, { "epoch": 7.51, "learning_rate": 5.6317219845032956e-05, "loss": 1.3792, "step": 86660 }, { "epoch": 7.51, "learning_rate": 5.6315051462935115e-05, "loss": 1.5117, "step": 86670 }, { "epoch": 7.51, "learning_rate": 5.631288308083728e-05, "loss": 1.2642, "step": 86680 }, { "epoch": 7.51, "learning_rate": 5.631071469873945e-05, "loss": 1.287, "step": 86690 }, { "epoch": 7.51, "learning_rate": 5.630854631664161e-05, "loss": 1.2391, "step": 86700 }, { "epoch": 7.51, "learning_rate": 5.630637793454377e-05, "loss": 1.4633, "step": 86710 }, { "epoch": 7.51, "learning_rate": 5.630420955244593e-05, "loss": 1.378, "step": 86720 }, { "epoch": 7.51, "learning_rate": 5.630204117034809e-05, "loss": 1.3187, "step": 86730 }, { "epoch": 7.51, "learning_rate": 5.629987278825025e-05, "loss": 1.326, "step": 86740 }, { "epoch": 7.51, "learning_rate": 5.629770440615242e-05, "loss": 1.2961, "step": 86750 }, { "epoch": 7.51, "learning_rate": 5.6295536024054586e-05, "loss": 1.2233, "step": 86760 }, { "epoch": 7.51, "learning_rate": 5.6293367641956745e-05, "loss": 1.5007, "step": 86770 }, { "epoch": 7.52, "learning_rate": 5.6291199259858905e-05, "loss": 1.48, "step": 86780 }, { "epoch": 7.52, "learning_rate": 5.6289030877761064e-05, "loss": 1.6371, "step": 86790 }, { "epoch": 7.52, "learning_rate": 5.628686249566323e-05, "loss": 1.7271, "step": 86800 }, { "epoch": 7.52, "learning_rate": 5.628469411356539e-05, "loss": 1.4394, "step": 86810 }, { "epoch": 7.52, "learning_rate": 5.628252573146756e-05, "loss": 1.4947, "step": 86820 }, { "epoch": 7.52, "learning_rate": 5.628035734936972e-05, "loss": 1.2681, "step": 86830 }, { "epoch": 7.52, "learning_rate": 5.627818896727188e-05, "loss": 1.4308, "step": 86840 }, { "epoch": 7.52, "learning_rate": 5.627602058517405e-05, "loss": 1.6056, "step": 86850 }, { "epoch": 7.52, "learning_rate": 5.62738522030762e-05, "loss": 1.491, "step": 86860 }, { "epoch": 7.52, "learning_rate": 5.627168382097837e-05, "loss": 1.2045, "step": 86870 }, { "epoch": 7.52, "learning_rate": 5.626951543888053e-05, "loss": 1.441, "step": 86880 }, { "epoch": 7.53, "learning_rate": 5.6267347056782694e-05, "loss": 1.4028, "step": 86890 }, { "epoch": 7.53, "learning_rate": 5.626517867468486e-05, "loss": 1.3166, "step": 86900 }, { "epoch": 7.53, "learning_rate": 5.626301029258702e-05, "loss": 1.4637, "step": 86910 }, { "epoch": 7.53, "learning_rate": 5.626084191048919e-05, "loss": 1.2679, "step": 86920 }, { "epoch": 7.53, "learning_rate": 5.6258673528391346e-05, "loss": 1.6018, "step": 86930 }, { "epoch": 7.53, "learning_rate": 5.6256505146293506e-05, "loss": 1.4284, "step": 86940 }, { "epoch": 7.53, "learning_rate": 5.6254336764195666e-05, "loss": 1.6093, "step": 86950 }, { "epoch": 7.53, "learning_rate": 5.625216838209783e-05, "loss": 1.2964, "step": 86960 }, { "epoch": 7.53, "learning_rate": 5.625e-05, "loss": 1.484, "step": 86970 }, { "epoch": 7.53, "learning_rate": 5.624783161790216e-05, "loss": 1.3854, "step": 86980 }, { "epoch": 7.53, "learning_rate": 5.6245663235804324e-05, "loss": 1.2607, "step": 86990 }, { "epoch": 7.53, "learning_rate": 5.6243494853706484e-05, "loss": 1.6592, "step": 87000 }, { "epoch": 7.54, "learning_rate": 5.6241326471608644e-05, "loss": 1.1806, "step": 87010 }, { "epoch": 7.54, "learning_rate": 5.62391580895108e-05, "loss": 1.4463, "step": 87020 }, { "epoch": 7.54, "learning_rate": 5.623698970741297e-05, "loss": 1.592, "step": 87030 }, { "epoch": 7.54, "learning_rate": 5.623482132531513e-05, "loss": 1.2792, "step": 87040 }, { "epoch": 7.54, "learning_rate": 5.6232652943217296e-05, "loss": 1.2927, "step": 87050 }, { "epoch": 7.54, "learning_rate": 5.623048456111946e-05, "loss": 1.4437, "step": 87060 }, { "epoch": 7.54, "learning_rate": 5.622831617902162e-05, "loss": 1.3344, "step": 87070 }, { "epoch": 7.54, "learning_rate": 5.622614779692379e-05, "loss": 1.441, "step": 87080 }, { "epoch": 7.54, "learning_rate": 5.622397941482594e-05, "loss": 1.5951, "step": 87090 }, { "epoch": 7.54, "learning_rate": 5.622181103272811e-05, "loss": 1.5154, "step": 87100 }, { "epoch": 7.54, "learning_rate": 5.621964265063027e-05, "loss": 1.4138, "step": 87110 }, { "epoch": 7.55, "learning_rate": 5.621747426853243e-05, "loss": 1.3566, "step": 87120 }, { "epoch": 7.55, "learning_rate": 5.62153058864346e-05, "loss": 1.4631, "step": 87130 }, { "epoch": 7.55, "learning_rate": 5.621313750433676e-05, "loss": 1.4743, "step": 87140 }, { "epoch": 7.55, "learning_rate": 5.6210969122238926e-05, "loss": 1.3501, "step": 87150 }, { "epoch": 7.55, "learning_rate": 5.6208800740141085e-05, "loss": 1.3267, "step": 87160 }, { "epoch": 7.55, "learning_rate": 5.6206632358043245e-05, "loss": 1.3131, "step": 87170 }, { "epoch": 7.55, "learning_rate": 5.6204463975945405e-05, "loss": 1.4647, "step": 87180 }, { "epoch": 7.55, "learning_rate": 5.620229559384757e-05, "loss": 1.2649, "step": 87190 }, { "epoch": 7.55, "learning_rate": 5.620012721174974e-05, "loss": 1.4008, "step": 87200 }, { "epoch": 7.55, "learning_rate": 5.61979588296519e-05, "loss": 1.434, "step": 87210 }, { "epoch": 7.55, "learning_rate": 5.619579044755406e-05, "loss": 1.275, "step": 87220 }, { "epoch": 7.55, "learning_rate": 5.619362206545622e-05, "loss": 1.5779, "step": 87230 }, { "epoch": 7.56, "learning_rate": 5.619145368335838e-05, "loss": 1.5309, "step": 87240 }, { "epoch": 7.56, "learning_rate": 5.618928530126054e-05, "loss": 1.3372, "step": 87250 }, { "epoch": 7.56, "learning_rate": 5.618711691916271e-05, "loss": 1.562, "step": 87260 }, { "epoch": 7.56, "learning_rate": 5.6184948537064875e-05, "loss": 1.4642, "step": 87270 }, { "epoch": 7.56, "learning_rate": 5.6182780154967035e-05, "loss": 1.2211, "step": 87280 }, { "epoch": 7.56, "learning_rate": 5.61806117728692e-05, "loss": 1.5306, "step": 87290 }, { "epoch": 7.56, "learning_rate": 5.617844339077136e-05, "loss": 1.3495, "step": 87300 }, { "epoch": 7.56, "learning_rate": 5.617627500867353e-05, "loss": 1.2982, "step": 87310 }, { "epoch": 7.56, "learning_rate": 5.617410662657568e-05, "loss": 1.4695, "step": 87320 }, { "epoch": 7.56, "learning_rate": 5.6171938244477846e-05, "loss": 1.6774, "step": 87330 }, { "epoch": 7.56, "learning_rate": 5.616976986238001e-05, "loss": 1.4217, "step": 87340 }, { "epoch": 7.57, "learning_rate": 5.616760148028217e-05, "loss": 1.3298, "step": 87350 }, { "epoch": 7.57, "learning_rate": 5.616543309818434e-05, "loss": 1.2889, "step": 87360 }, { "epoch": 7.57, "learning_rate": 5.61632647160865e-05, "loss": 1.4323, "step": 87370 }, { "epoch": 7.57, "learning_rate": 5.6161096333988665e-05, "loss": 1.6003, "step": 87380 }, { "epoch": 7.57, "learning_rate": 5.615892795189083e-05, "loss": 1.3202, "step": 87390 }, { "epoch": 7.57, "learning_rate": 5.6156759569792984e-05, "loss": 1.6156, "step": 87400 }, { "epoch": 7.57, "learning_rate": 5.615459118769515e-05, "loss": 1.4274, "step": 87410 }, { "epoch": 7.57, "learning_rate": 5.615242280559731e-05, "loss": 1.4618, "step": 87420 }, { "epoch": 7.57, "learning_rate": 5.6150254423499476e-05, "loss": 1.4464, "step": 87430 }, { "epoch": 7.57, "learning_rate": 5.6148086041401636e-05, "loss": 1.3842, "step": 87440 }, { "epoch": 7.57, "learning_rate": 5.61459176593038e-05, "loss": 1.2245, "step": 87450 }, { "epoch": 7.57, "learning_rate": 5.614374927720597e-05, "loss": 1.7204, "step": 87460 }, { "epoch": 7.58, "learning_rate": 5.614158089510813e-05, "loss": 1.3006, "step": 87470 }, { "epoch": 7.58, "learning_rate": 5.613941251301028e-05, "loss": 1.2325, "step": 87480 }, { "epoch": 7.58, "learning_rate": 5.613724413091245e-05, "loss": 1.3119, "step": 87490 }, { "epoch": 7.58, "learning_rate": 5.6135075748814614e-05, "loss": 1.1515, "step": 87500 }, { "epoch": 7.58, "learning_rate": 5.6132907366716774e-05, "loss": 1.2569, "step": 87510 }, { "epoch": 7.58, "learning_rate": 5.613073898461894e-05, "loss": 1.3603, "step": 87520 }, { "epoch": 7.58, "learning_rate": 5.61285706025211e-05, "loss": 1.3221, "step": 87530 }, { "epoch": 7.58, "learning_rate": 5.6126402220423266e-05, "loss": 1.1849, "step": 87540 }, { "epoch": 7.58, "learning_rate": 5.612423383832542e-05, "loss": 1.328, "step": 87550 }, { "epoch": 7.58, "learning_rate": 5.6122065456227585e-05, "loss": 1.2245, "step": 87560 }, { "epoch": 7.58, "learning_rate": 5.611989707412975e-05, "loss": 1.6145, "step": 87570 }, { "epoch": 7.58, "learning_rate": 5.611772869203191e-05, "loss": 1.3935, "step": 87580 }, { "epoch": 7.59, "learning_rate": 5.611556030993408e-05, "loss": 1.1879, "step": 87590 }, { "epoch": 7.59, "learning_rate": 5.611339192783624e-05, "loss": 1.4127, "step": 87600 }, { "epoch": 7.59, "learning_rate": 5.6111223545738404e-05, "loss": 1.4925, "step": 87610 }, { "epoch": 7.59, "learning_rate": 5.610905516364057e-05, "loss": 1.3402, "step": 87620 }, { "epoch": 7.59, "learning_rate": 5.610688678154272e-05, "loss": 1.5292, "step": 87630 }, { "epoch": 7.59, "learning_rate": 5.610471839944489e-05, "loss": 1.4241, "step": 87640 }, { "epoch": 7.59, "learning_rate": 5.610255001734705e-05, "loss": 1.3967, "step": 87650 }, { "epoch": 7.59, "learning_rate": 5.6100381635249215e-05, "loss": 1.4495, "step": 87660 }, { "epoch": 7.59, "learning_rate": 5.6098213253151375e-05, "loss": 1.3826, "step": 87670 }, { "epoch": 7.59, "learning_rate": 5.609604487105354e-05, "loss": 1.5052, "step": 87680 }, { "epoch": 7.59, "learning_rate": 5.609387648895571e-05, "loss": 1.3544, "step": 87690 }, { "epoch": 7.6, "learning_rate": 5.609170810685787e-05, "loss": 1.4084, "step": 87700 }, { "epoch": 7.6, "learning_rate": 5.608953972476003e-05, "loss": 1.6361, "step": 87710 }, { "epoch": 7.6, "learning_rate": 5.6087371342662186e-05, "loss": 1.4663, "step": 87720 }, { "epoch": 7.6, "learning_rate": 5.608520296056435e-05, "loss": 1.3595, "step": 87730 }, { "epoch": 7.6, "learning_rate": 5.608303457846651e-05, "loss": 1.4231, "step": 87740 }, { "epoch": 7.6, "learning_rate": 5.608086619636868e-05, "loss": 1.1739, "step": 87750 }, { "epoch": 7.6, "learning_rate": 5.6078697814270845e-05, "loss": 1.4364, "step": 87760 }, { "epoch": 7.6, "learning_rate": 5.6076529432173005e-05, "loss": 1.5852, "step": 87770 }, { "epoch": 7.6, "learning_rate": 5.6074361050075164e-05, "loss": 1.5541, "step": 87780 }, { "epoch": 7.6, "learning_rate": 5.6072192667977324e-05, "loss": 1.2116, "step": 87790 }, { "epoch": 7.6, "learning_rate": 5.607002428587949e-05, "loss": 1.3284, "step": 87800 }, { "epoch": 7.6, "learning_rate": 5.606785590378165e-05, "loss": 1.2286, "step": 87810 }, { "epoch": 7.61, "learning_rate": 5.6065687521683816e-05, "loss": 1.5369, "step": 87820 }, { "epoch": 7.61, "learning_rate": 5.606351913958598e-05, "loss": 1.4325, "step": 87830 }, { "epoch": 7.61, "learning_rate": 5.606135075748814e-05, "loss": 1.3038, "step": 87840 }, { "epoch": 7.61, "learning_rate": 5.605918237539031e-05, "loss": 1.3476, "step": 87850 }, { "epoch": 7.61, "learning_rate": 5.605701399329246e-05, "loss": 1.2577, "step": 87860 }, { "epoch": 7.61, "learning_rate": 5.605484561119463e-05, "loss": 1.2425, "step": 87870 }, { "epoch": 7.61, "learning_rate": 5.605267722909679e-05, "loss": 1.2986, "step": 87880 }, { "epoch": 7.61, "learning_rate": 5.6050508846998954e-05, "loss": 1.5959, "step": 87890 }, { "epoch": 7.61, "learning_rate": 5.604834046490112e-05, "loss": 1.5094, "step": 87900 }, { "epoch": 7.61, "learning_rate": 5.604617208280328e-05, "loss": 1.3414, "step": 87910 }, { "epoch": 7.61, "learning_rate": 5.6044003700705447e-05, "loss": 1.2995, "step": 87920 }, { "epoch": 7.62, "learning_rate": 5.6041835318607606e-05, "loss": 1.2362, "step": 87930 }, { "epoch": 7.62, "learning_rate": 5.6039666936509766e-05, "loss": 1.4338, "step": 87940 }, { "epoch": 7.62, "learning_rate": 5.6037498554411925e-05, "loss": 1.5063, "step": 87950 }, { "epoch": 7.62, "learning_rate": 5.603533017231409e-05, "loss": 1.2937, "step": 87960 }, { "epoch": 7.62, "learning_rate": 5.603316179021626e-05, "loss": 1.2867, "step": 87970 }, { "epoch": 7.62, "learning_rate": 5.603099340811842e-05, "loss": 1.3426, "step": 87980 }, { "epoch": 7.62, "learning_rate": 5.6028825026020584e-05, "loss": 1.3925, "step": 87990 }, { "epoch": 7.62, "learning_rate": 5.6026656643922744e-05, "loss": 1.3894, "step": 88000 }, { "epoch": 7.62, "learning_rate": 5.6024488261824903e-05, "loss": 1.5118, "step": 88010 }, { "epoch": 7.62, "learning_rate": 5.602231987972706e-05, "loss": 1.4112, "step": 88020 }, { "epoch": 7.62, "learning_rate": 5.602015149762923e-05, "loss": 1.183, "step": 88030 }, { "epoch": 7.62, "learning_rate": 5.601798311553139e-05, "loss": 1.3884, "step": 88040 }, { "epoch": 7.63, "learning_rate": 5.6015814733433555e-05, "loss": 1.3533, "step": 88050 }, { "epoch": 7.63, "learning_rate": 5.601364635133572e-05, "loss": 1.4775, "step": 88060 }, { "epoch": 7.63, "learning_rate": 5.601147796923788e-05, "loss": 1.3567, "step": 88070 }, { "epoch": 7.63, "learning_rate": 5.600930958714005e-05, "loss": 1.4392, "step": 88080 }, { "epoch": 7.63, "learning_rate": 5.60071412050422e-05, "loss": 1.6242, "step": 88090 }, { "epoch": 7.63, "learning_rate": 5.600497282294437e-05, "loss": 1.2703, "step": 88100 }, { "epoch": 7.63, "learning_rate": 5.600280444084653e-05, "loss": 1.5378, "step": 88110 }, { "epoch": 7.63, "learning_rate": 5.600063605874869e-05, "loss": 1.3865, "step": 88120 }, { "epoch": 7.63, "learning_rate": 5.599846767665086e-05, "loss": 1.5901, "step": 88130 }, { "epoch": 7.63, "learning_rate": 5.599629929455302e-05, "loss": 1.349, "step": 88140 }, { "epoch": 7.63, "learning_rate": 5.5994130912455185e-05, "loss": 1.491, "step": 88150 }, { "epoch": 7.64, "learning_rate": 5.5991962530357345e-05, "loss": 1.4473, "step": 88160 }, { "epoch": 7.64, "learning_rate": 5.5989794148259505e-05, "loss": 1.4231, "step": 88170 }, { "epoch": 7.64, "learning_rate": 5.5987625766161664e-05, "loss": 1.3429, "step": 88180 }, { "epoch": 7.64, "learning_rate": 5.598545738406383e-05, "loss": 1.3827, "step": 88190 }, { "epoch": 7.64, "learning_rate": 5.5983289001966e-05, "loss": 1.6139, "step": 88200 }, { "epoch": 7.64, "learning_rate": 5.598112061986816e-05, "loss": 1.2576, "step": 88210 }, { "epoch": 7.64, "learning_rate": 5.597895223777032e-05, "loss": 1.4533, "step": 88220 }, { "epoch": 7.64, "learning_rate": 5.597678385567248e-05, "loss": 1.3087, "step": 88230 }, { "epoch": 7.64, "learning_rate": 5.597461547357465e-05, "loss": 1.4658, "step": 88240 }, { "epoch": 7.64, "learning_rate": 5.59724470914768e-05, "loss": 1.5064, "step": 88250 }, { "epoch": 7.64, "learning_rate": 5.597027870937897e-05, "loss": 1.3523, "step": 88260 }, { "epoch": 7.64, "learning_rate": 5.5968110327281135e-05, "loss": 1.2718, "step": 88270 }, { "epoch": 7.65, "learning_rate": 5.5965941945183294e-05, "loss": 1.3586, "step": 88280 }, { "epoch": 7.65, "learning_rate": 5.596377356308546e-05, "loss": 1.6396, "step": 88290 }, { "epoch": 7.65, "learning_rate": 5.596160518098762e-05, "loss": 1.3355, "step": 88300 }, { "epoch": 7.65, "learning_rate": 5.595943679888979e-05, "loss": 1.2318, "step": 88310 }, { "epoch": 7.65, "learning_rate": 5.595726841679194e-05, "loss": 1.4083, "step": 88320 }, { "epoch": 7.65, "learning_rate": 5.5955100034694106e-05, "loss": 1.5265, "step": 88330 }, { "epoch": 7.65, "learning_rate": 5.595293165259627e-05, "loss": 1.3454, "step": 88340 }, { "epoch": 7.65, "learning_rate": 5.595076327049843e-05, "loss": 1.43, "step": 88350 }, { "epoch": 7.65, "learning_rate": 5.59485948884006e-05, "loss": 1.6264, "step": 88360 }, { "epoch": 7.65, "learning_rate": 5.594642650630276e-05, "loss": 1.4093, "step": 88370 }, { "epoch": 7.65, "learning_rate": 5.5944258124204924e-05, "loss": 1.5342, "step": 88380 }, { "epoch": 7.66, "learning_rate": 5.594208974210709e-05, "loss": 1.3187, "step": 88390 }, { "epoch": 7.66, "learning_rate": 5.5939921360009244e-05, "loss": 1.2393, "step": 88400 }, { "epoch": 7.66, "learning_rate": 5.593775297791141e-05, "loss": 1.4099, "step": 88410 }, { "epoch": 7.66, "learning_rate": 5.593558459581357e-05, "loss": 1.2919, "step": 88420 }, { "epoch": 7.66, "learning_rate": 5.5933416213715736e-05, "loss": 1.464, "step": 88430 }, { "epoch": 7.66, "learning_rate": 5.5931247831617896e-05, "loss": 1.4763, "step": 88440 }, { "epoch": 7.66, "learning_rate": 5.592907944952006e-05, "loss": 1.4182, "step": 88450 }, { "epoch": 7.66, "learning_rate": 5.592691106742223e-05, "loss": 1.4656, "step": 88460 }, { "epoch": 7.66, "learning_rate": 5.592474268532439e-05, "loss": 1.3226, "step": 88470 }, { "epoch": 7.66, "learning_rate": 5.592257430322655e-05, "loss": 1.3529, "step": 88480 }, { "epoch": 7.66, "learning_rate": 5.592040592112871e-05, "loss": 1.3772, "step": 88490 }, { "epoch": 7.66, "learning_rate": 5.5918237539030874e-05, "loss": 1.4442, "step": 88500 }, { "epoch": 7.67, "learning_rate": 5.591606915693303e-05, "loss": 1.4055, "step": 88510 }, { "epoch": 7.67, "learning_rate": 5.59139007748352e-05, "loss": 1.3317, "step": 88520 }, { "epoch": 7.67, "learning_rate": 5.5911732392737366e-05, "loss": 1.4492, "step": 88530 }, { "epoch": 7.67, "learning_rate": 5.5909564010639526e-05, "loss": 1.0921, "step": 88540 }, { "epoch": 7.67, "learning_rate": 5.590739562854168e-05, "loss": 1.3309, "step": 88550 }, { "epoch": 7.67, "learning_rate": 5.5905227246443845e-05, "loss": 1.3096, "step": 88560 }, { "epoch": 7.67, "learning_rate": 5.590305886434601e-05, "loss": 1.4114, "step": 88570 }, { "epoch": 7.67, "learning_rate": 5.590089048224817e-05, "loss": 1.3109, "step": 88580 }, { "epoch": 7.67, "learning_rate": 5.589872210015034e-05, "loss": 1.3292, "step": 88590 }, { "epoch": 7.67, "learning_rate": 5.58965537180525e-05, "loss": 1.4374, "step": 88600 }, { "epoch": 7.67, "learning_rate": 5.589438533595466e-05, "loss": 1.2715, "step": 88610 }, { "epoch": 7.68, "learning_rate": 5.589221695385683e-05, "loss": 1.466, "step": 88620 }, { "epoch": 7.68, "learning_rate": 5.589004857175898e-05, "loss": 1.3406, "step": 88630 }, { "epoch": 7.68, "learning_rate": 5.588788018966115e-05, "loss": 1.3407, "step": 88640 }, { "epoch": 7.68, "learning_rate": 5.588571180756331e-05, "loss": 1.466, "step": 88650 }, { "epoch": 7.68, "learning_rate": 5.5883543425465475e-05, "loss": 1.4724, "step": 88660 }, { "epoch": 7.68, "learning_rate": 5.5881375043367635e-05, "loss": 1.3154, "step": 88670 }, { "epoch": 7.68, "learning_rate": 5.58792066612698e-05, "loss": 1.5294, "step": 88680 }, { "epoch": 7.68, "learning_rate": 5.587703827917197e-05, "loss": 1.5742, "step": 88690 }, { "epoch": 7.68, "learning_rate": 5.587486989707413e-05, "loss": 1.2968, "step": 88700 }, { "epoch": 7.68, "learning_rate": 5.5872701514976287e-05, "loss": 1.2442, "step": 88710 }, { "epoch": 7.68, "learning_rate": 5.5870533132878446e-05, "loss": 1.372, "step": 88720 }, { "epoch": 7.68, "learning_rate": 5.586836475078061e-05, "loss": 1.3144, "step": 88730 }, { "epoch": 7.69, "learning_rate": 5.586619636868277e-05, "loss": 1.4907, "step": 88740 }, { "epoch": 7.69, "learning_rate": 5.586402798658494e-05, "loss": 1.3235, "step": 88750 }, { "epoch": 7.69, "learning_rate": 5.5861859604487105e-05, "loss": 1.3151, "step": 88760 }, { "epoch": 7.69, "learning_rate": 5.5859691222389265e-05, "loss": 1.2905, "step": 88770 }, { "epoch": 7.69, "learning_rate": 5.5857522840291424e-05, "loss": 1.4288, "step": 88780 }, { "epoch": 7.69, "learning_rate": 5.5855354458193584e-05, "loss": 1.4006, "step": 88790 }, { "epoch": 7.69, "learning_rate": 5.585318607609575e-05, "loss": 1.4308, "step": 88800 }, { "epoch": 7.69, "learning_rate": 5.585101769399791e-05, "loss": 1.5654, "step": 88810 }, { "epoch": 7.69, "learning_rate": 5.5848849311900076e-05, "loss": 1.4316, "step": 88820 }, { "epoch": 7.69, "learning_rate": 5.584668092980224e-05, "loss": 1.3105, "step": 88830 }, { "epoch": 7.69, "learning_rate": 5.58445125477044e-05, "loss": 1.3459, "step": 88840 }, { "epoch": 7.69, "learning_rate": 5.584234416560657e-05, "loss": 1.1802, "step": 88850 }, { "epoch": 7.7, "learning_rate": 5.584017578350872e-05, "loss": 1.3646, "step": 88860 }, { "epoch": 7.7, "learning_rate": 5.583800740141089e-05, "loss": 1.5763, "step": 88870 }, { "epoch": 7.7, "learning_rate": 5.583583901931305e-05, "loss": 1.3356, "step": 88880 }, { "epoch": 7.7, "learning_rate": 5.5833670637215214e-05, "loss": 1.5238, "step": 88890 }, { "epoch": 7.7, "learning_rate": 5.583150225511738e-05, "loss": 1.4087, "step": 88900 }, { "epoch": 7.7, "learning_rate": 5.582933387301954e-05, "loss": 1.5263, "step": 88910 }, { "epoch": 7.7, "learning_rate": 5.5827165490921706e-05, "loss": 1.3511, "step": 88920 }, { "epoch": 7.7, "learning_rate": 5.5824997108823866e-05, "loss": 1.4442, "step": 88930 }, { "epoch": 7.7, "learning_rate": 5.5822828726726025e-05, "loss": 1.4037, "step": 88940 }, { "epoch": 7.7, "learning_rate": 5.5820660344628185e-05, "loss": 1.4154, "step": 88950 }, { "epoch": 7.7, "learning_rate": 5.581849196253035e-05, "loss": 1.4244, "step": 88960 }, { "epoch": 7.71, "learning_rate": 5.581632358043252e-05, "loss": 1.4901, "step": 88970 }, { "epoch": 7.71, "learning_rate": 5.581415519833468e-05, "loss": 1.532, "step": 88980 }, { "epoch": 7.71, "learning_rate": 5.5811986816236844e-05, "loss": 1.4558, "step": 88990 }, { "epoch": 7.71, "learning_rate": 5.5809818434139003e-05, "loss": 1.2355, "step": 89000 }, { "epoch": 7.71, "learning_rate": 5.580765005204117e-05, "loss": 1.4577, "step": 89010 }, { "epoch": 7.71, "learning_rate": 5.580548166994332e-05, "loss": 1.3284, "step": 89020 }, { "epoch": 7.71, "learning_rate": 5.580331328784549e-05, "loss": 1.3326, "step": 89030 }, { "epoch": 7.71, "learning_rate": 5.5801144905747655e-05, "loss": 1.4376, "step": 89040 }, { "epoch": 7.71, "learning_rate": 5.5798976523649815e-05, "loss": 1.4029, "step": 89050 }, { "epoch": 7.71, "learning_rate": 5.579680814155198e-05, "loss": 1.2994, "step": 89060 }, { "epoch": 7.71, "learning_rate": 5.579463975945414e-05, "loss": 1.1103, "step": 89070 }, { "epoch": 7.71, "learning_rate": 5.579247137735631e-05, "loss": 1.378, "step": 89080 }, { "epoch": 7.72, "learning_rate": 5.579030299525846e-05, "loss": 1.4431, "step": 89090 }, { "epoch": 7.72, "learning_rate": 5.578813461316063e-05, "loss": 1.3654, "step": 89100 }, { "epoch": 7.72, "learning_rate": 5.5785966231062786e-05, "loss": 1.5018, "step": 89110 }, { "epoch": 7.72, "learning_rate": 5.578379784896495e-05, "loss": 1.2494, "step": 89120 }, { "epoch": 7.72, "learning_rate": 5.578162946686712e-05, "loss": 1.3522, "step": 89130 }, { "epoch": 7.72, "learning_rate": 5.577946108476928e-05, "loss": 1.2245, "step": 89140 }, { "epoch": 7.72, "learning_rate": 5.5777292702671445e-05, "loss": 1.5412, "step": 89150 }, { "epoch": 7.72, "learning_rate": 5.5775124320573605e-05, "loss": 1.3198, "step": 89160 }, { "epoch": 7.72, "learning_rate": 5.5772955938475764e-05, "loss": 1.3535, "step": 89170 }, { "epoch": 7.72, "learning_rate": 5.5770787556377924e-05, "loss": 1.2252, "step": 89180 }, { "epoch": 7.72, "learning_rate": 5.576861917428009e-05, "loss": 1.2425, "step": 89190 }, { "epoch": 7.73, "learning_rate": 5.576645079218226e-05, "loss": 1.3282, "step": 89200 }, { "epoch": 7.73, "learning_rate": 5.5764282410084416e-05, "loss": 1.3998, "step": 89210 }, { "epoch": 7.73, "learning_rate": 5.576211402798658e-05, "loss": 1.5666, "step": 89220 }, { "epoch": 7.73, "learning_rate": 5.575994564588874e-05, "loss": 1.5286, "step": 89230 }, { "epoch": 7.73, "learning_rate": 5.575777726379091e-05, "loss": 1.5115, "step": 89240 }, { "epoch": 7.73, "learning_rate": 5.575560888169306e-05, "loss": 1.3728, "step": 89250 }, { "epoch": 7.73, "learning_rate": 5.575344049959523e-05, "loss": 1.4131, "step": 89260 }, { "epoch": 7.73, "learning_rate": 5.5751272117497394e-05, "loss": 1.4587, "step": 89270 }, { "epoch": 7.73, "learning_rate": 5.5749103735399554e-05, "loss": 1.3659, "step": 89280 }, { "epoch": 7.73, "learning_rate": 5.574693535330172e-05, "loss": 1.312, "step": 89290 }, { "epoch": 7.73, "learning_rate": 5.574476697120388e-05, "loss": 1.3418, "step": 89300 }, { "epoch": 7.73, "learning_rate": 5.5742598589106046e-05, "loss": 1.1723, "step": 89310 }, { "epoch": 7.74, "learning_rate": 5.57404302070082e-05, "loss": 1.3585, "step": 89320 }, { "epoch": 7.74, "learning_rate": 5.5738261824910366e-05, "loss": 1.3692, "step": 89330 }, { "epoch": 7.74, "learning_rate": 5.573609344281253e-05, "loss": 1.4497, "step": 89340 }, { "epoch": 7.74, "learning_rate": 5.573392506071469e-05, "loss": 1.3892, "step": 89350 }, { "epoch": 7.74, "learning_rate": 5.573175667861686e-05, "loss": 1.3772, "step": 89360 }, { "epoch": 7.74, "learning_rate": 5.572958829651902e-05, "loss": 1.5579, "step": 89370 }, { "epoch": 7.74, "learning_rate": 5.5727419914421184e-05, "loss": 1.3438, "step": 89380 }, { "epoch": 7.74, "learning_rate": 5.572525153232335e-05, "loss": 1.4013, "step": 89390 }, { "epoch": 7.74, "learning_rate": 5.57230831502255e-05, "loss": 1.3396, "step": 89400 }, { "epoch": 7.74, "learning_rate": 5.572091476812767e-05, "loss": 1.3712, "step": 89410 }, { "epoch": 7.74, "learning_rate": 5.571874638602983e-05, "loss": 1.2805, "step": 89420 }, { "epoch": 7.75, "learning_rate": 5.5716578003931996e-05, "loss": 1.4189, "step": 89430 }, { "epoch": 7.75, "learning_rate": 5.5714409621834155e-05, "loss": 1.2574, "step": 89440 }, { "epoch": 7.75, "learning_rate": 5.571224123973632e-05, "loss": 1.5295, "step": 89450 }, { "epoch": 7.75, "learning_rate": 5.571007285763849e-05, "loss": 1.4916, "step": 89460 }, { "epoch": 7.75, "learning_rate": 5.570790447554065e-05, "loss": 1.4309, "step": 89470 }, { "epoch": 7.75, "learning_rate": 5.570573609344281e-05, "loss": 1.4381, "step": 89480 }, { "epoch": 7.75, "learning_rate": 5.570356771134497e-05, "loss": 1.6287, "step": 89490 }, { "epoch": 7.75, "learning_rate": 5.570139932924713e-05, "loss": 1.3264, "step": 89500 }, { "epoch": 7.75, "learning_rate": 5.569923094714929e-05, "loss": 1.5035, "step": 89510 }, { "epoch": 7.75, "learning_rate": 5.569706256505146e-05, "loss": 1.3489, "step": 89520 }, { "epoch": 7.75, "learning_rate": 5.5694894182953626e-05, "loss": 1.5913, "step": 89530 }, { "epoch": 7.75, "learning_rate": 5.5692725800855785e-05, "loss": 1.4428, "step": 89540 }, { "epoch": 7.76, "learning_rate": 5.5690557418757945e-05, "loss": 1.3471, "step": 89550 }, { "epoch": 7.76, "learning_rate": 5.5688389036660105e-05, "loss": 1.4974, "step": 89560 }, { "epoch": 7.76, "learning_rate": 5.568622065456227e-05, "loss": 1.5147, "step": 89570 }, { "epoch": 7.76, "learning_rate": 5.568405227246443e-05, "loss": 1.1543, "step": 89580 }, { "epoch": 7.76, "learning_rate": 5.56818838903666e-05, "loss": 1.4323, "step": 89590 }, { "epoch": 7.76, "learning_rate": 5.567971550826876e-05, "loss": 1.6095, "step": 89600 }, { "epoch": 7.76, "learning_rate": 5.567754712617092e-05, "loss": 1.5854, "step": 89610 }, { "epoch": 7.76, "learning_rate": 5.567537874407309e-05, "loss": 1.4532, "step": 89620 }, { "epoch": 7.76, "learning_rate": 5.567321036197524e-05, "loss": 1.4614, "step": 89630 }, { "epoch": 7.76, "learning_rate": 5.567104197987741e-05, "loss": 1.5622, "step": 89640 }, { "epoch": 7.76, "learning_rate": 5.566887359777957e-05, "loss": 1.2959, "step": 89650 }, { "epoch": 7.77, "learning_rate": 5.5666705215681735e-05, "loss": 1.1965, "step": 89660 }, { "epoch": 7.77, "learning_rate": 5.5664536833583894e-05, "loss": 1.2373, "step": 89670 }, { "epoch": 7.77, "learning_rate": 5.566236845148606e-05, "loss": 1.3016, "step": 89680 }, { "epoch": 7.77, "learning_rate": 5.566020006938823e-05, "loss": 1.4773, "step": 89690 }, { "epoch": 7.77, "learning_rate": 5.565803168729039e-05, "loss": 1.2281, "step": 89700 }, { "epoch": 7.77, "learning_rate": 5.5655863305192546e-05, "loss": 1.2623, "step": 89710 }, { "epoch": 7.77, "learning_rate": 5.5653694923094706e-05, "loss": 1.2198, "step": 89720 }, { "epoch": 7.77, "learning_rate": 5.565152654099687e-05, "loss": 1.3586, "step": 89730 }, { "epoch": 7.77, "learning_rate": 5.564935815889903e-05, "loss": 1.4071, "step": 89740 }, { "epoch": 7.77, "learning_rate": 5.56471897768012e-05, "loss": 1.3672, "step": 89750 }, { "epoch": 7.77, "learning_rate": 5.5645021394703365e-05, "loss": 1.2807, "step": 89760 }, { "epoch": 7.77, "learning_rate": 5.5642853012605524e-05, "loss": 1.3561, "step": 89770 }, { "epoch": 7.78, "learning_rate": 5.564068463050769e-05, "loss": 1.3114, "step": 89780 }, { "epoch": 7.78, "learning_rate": 5.5638516248409843e-05, "loss": 1.5093, "step": 89790 }, { "epoch": 7.78, "learning_rate": 5.563634786631201e-05, "loss": 1.3368, "step": 89800 }, { "epoch": 7.78, "learning_rate": 5.563417948421417e-05, "loss": 1.3875, "step": 89810 }, { "epoch": 7.78, "learning_rate": 5.5632011102116336e-05, "loss": 1.2359, "step": 89820 }, { "epoch": 7.78, "learning_rate": 5.56298427200185e-05, "loss": 1.3492, "step": 89830 }, { "epoch": 7.78, "learning_rate": 5.562767433792066e-05, "loss": 1.5703, "step": 89840 }, { "epoch": 7.78, "learning_rate": 5.562550595582283e-05, "loss": 1.5672, "step": 89850 }, { "epoch": 7.78, "learning_rate": 5.562333757372498e-05, "loss": 1.1793, "step": 89860 }, { "epoch": 7.78, "learning_rate": 5.562116919162715e-05, "loss": 1.4048, "step": 89870 }, { "epoch": 7.78, "learning_rate": 5.561900080952931e-05, "loss": 1.4305, "step": 89880 }, { "epoch": 7.79, "learning_rate": 5.5616832427431474e-05, "loss": 1.4296, "step": 89890 }, { "epoch": 7.79, "learning_rate": 5.561466404533364e-05, "loss": 1.3041, "step": 89900 }, { "epoch": 7.79, "learning_rate": 5.56124956632358e-05, "loss": 1.6437, "step": 89910 }, { "epoch": 7.79, "learning_rate": 5.5610327281137966e-05, "loss": 1.3872, "step": 89920 }, { "epoch": 7.79, "learning_rate": 5.5608158899040126e-05, "loss": 1.2365, "step": 89930 }, { "epoch": 7.79, "learning_rate": 5.5605990516942285e-05, "loss": 1.6075, "step": 89940 }, { "epoch": 7.79, "learning_rate": 5.5603822134844445e-05, "loss": 1.4862, "step": 89950 }, { "epoch": 7.79, "learning_rate": 5.560165375274661e-05, "loss": 1.2868, "step": 89960 }, { "epoch": 7.79, "learning_rate": 5.559948537064878e-05, "loss": 1.4673, "step": 89970 }, { "epoch": 7.79, "learning_rate": 5.559731698855094e-05, "loss": 1.3557, "step": 89980 }, { "epoch": 7.79, "learning_rate": 5.5595148606453104e-05, "loss": 1.5038, "step": 89990 }, { "epoch": 7.79, "learning_rate": 5.559298022435526e-05, "loss": 1.4463, "step": 90000 }, { "epoch": 7.8, "learning_rate": 5.559081184225743e-05, "loss": 1.5479, "step": 90010 }, { "epoch": 7.8, "learning_rate": 5.558864346015958e-05, "loss": 1.2344, "step": 90020 }, { "epoch": 7.8, "learning_rate": 5.558647507806175e-05, "loss": 1.1306, "step": 90030 }, { "epoch": 7.8, "learning_rate": 5.5584306695963915e-05, "loss": 1.4141, "step": 90040 }, { "epoch": 7.8, "learning_rate": 5.5582138313866075e-05, "loss": 1.3587, "step": 90050 }, { "epoch": 7.8, "learning_rate": 5.557996993176824e-05, "loss": 1.235, "step": 90060 }, { "epoch": 7.8, "learning_rate": 5.55778015496704e-05, "loss": 1.2661, "step": 90070 }, { "epoch": 7.8, "learning_rate": 5.557563316757257e-05, "loss": 1.4384, "step": 90080 }, { "epoch": 7.8, "learning_rate": 5.557346478547472e-05, "loss": 1.4267, "step": 90090 }, { "epoch": 7.8, "learning_rate": 5.5571296403376886e-05, "loss": 1.5889, "step": 90100 }, { "epoch": 7.8, "learning_rate": 5.556912802127905e-05, "loss": 1.2933, "step": 90110 }, { "epoch": 7.8, "learning_rate": 5.556695963918121e-05, "loss": 1.1574, "step": 90120 }, { "epoch": 7.81, "learning_rate": 5.556479125708338e-05, "loss": 1.4798, "step": 90130 }, { "epoch": 7.81, "learning_rate": 5.556262287498554e-05, "loss": 1.6703, "step": 90140 }, { "epoch": 7.81, "learning_rate": 5.5560454492887705e-05, "loss": 1.4903, "step": 90150 }, { "epoch": 7.81, "learning_rate": 5.555828611078987e-05, "loss": 1.1404, "step": 90160 }, { "epoch": 7.81, "learning_rate": 5.5556117728692024e-05, "loss": 1.4058, "step": 90170 }, { "epoch": 7.81, "learning_rate": 5.5553949346594184e-05, "loss": 1.4232, "step": 90180 }, { "epoch": 7.81, "learning_rate": 5.555178096449635e-05, "loss": 1.3937, "step": 90190 }, { "epoch": 7.81, "learning_rate": 5.5549612582398516e-05, "loss": 1.3992, "step": 90200 }, { "epoch": 7.81, "learning_rate": 5.5547444200300676e-05, "loss": 1.4369, "step": 90210 }, { "epoch": 7.81, "learning_rate": 5.554527581820284e-05, "loss": 1.3442, "step": 90220 }, { "epoch": 7.81, "learning_rate": 5.5543107436105e-05, "loss": 1.3901, "step": 90230 }, { "epoch": 7.82, "learning_rate": 5.554093905400717e-05, "loss": 1.3435, "step": 90240 }, { "epoch": 7.82, "learning_rate": 5.553877067190932e-05, "loss": 1.6892, "step": 90250 }, { "epoch": 7.82, "learning_rate": 5.553660228981149e-05, "loss": 1.1892, "step": 90260 }, { "epoch": 7.82, "learning_rate": 5.5534433907713654e-05, "loss": 1.3777, "step": 90270 }, { "epoch": 7.82, "learning_rate": 5.5532265525615814e-05, "loss": 1.376, "step": 90280 }, { "epoch": 7.82, "learning_rate": 5.553009714351798e-05, "loss": 1.6456, "step": 90290 }, { "epoch": 7.82, "learning_rate": 5.552792876142014e-05, "loss": 1.3905, "step": 90300 }, { "epoch": 7.82, "learning_rate": 5.5525760379322306e-05, "loss": 1.4178, "step": 90310 }, { "epoch": 7.82, "learning_rate": 5.552359199722446e-05, "loss": 1.2894, "step": 90320 }, { "epoch": 7.82, "learning_rate": 5.5521423615126625e-05, "loss": 1.3647, "step": 90330 }, { "epoch": 7.82, "learning_rate": 5.551925523302879e-05, "loss": 1.384, "step": 90340 }, { "epoch": 7.82, "learning_rate": 5.551708685093095e-05, "loss": 1.4286, "step": 90350 }, { "epoch": 7.83, "learning_rate": 5.551491846883312e-05, "loss": 1.7373, "step": 90360 }, { "epoch": 7.83, "learning_rate": 5.551275008673528e-05, "loss": 1.4497, "step": 90370 }, { "epoch": 7.83, "learning_rate": 5.5510581704637444e-05, "loss": 1.2068, "step": 90380 }, { "epoch": 7.83, "learning_rate": 5.550841332253961e-05, "loss": 1.4869, "step": 90390 }, { "epoch": 7.83, "learning_rate": 5.550624494044176e-05, "loss": 1.2822, "step": 90400 }, { "epoch": 7.83, "learning_rate": 5.550407655834393e-05, "loss": 1.3087, "step": 90410 }, { "epoch": 7.83, "learning_rate": 5.550190817624609e-05, "loss": 1.3959, "step": 90420 }, { "epoch": 7.83, "learning_rate": 5.5499739794148255e-05, "loss": 1.4291, "step": 90430 }, { "epoch": 7.83, "learning_rate": 5.5497571412050415e-05, "loss": 1.4016, "step": 90440 }, { "epoch": 7.83, "learning_rate": 5.549540302995258e-05, "loss": 1.3696, "step": 90450 }, { "epoch": 7.83, "learning_rate": 5.549323464785475e-05, "loss": 1.2604, "step": 90460 }, { "epoch": 7.84, "learning_rate": 5.549106626575691e-05, "loss": 1.453, "step": 90470 }, { "epoch": 7.84, "learning_rate": 5.548889788365907e-05, "loss": 1.2739, "step": 90480 }, { "epoch": 7.84, "learning_rate": 5.548672950156123e-05, "loss": 1.6956, "step": 90490 }, { "epoch": 7.84, "learning_rate": 5.548456111946339e-05, "loss": 1.3755, "step": 90500 }, { "epoch": 7.84, "learning_rate": 5.548239273736555e-05, "loss": 1.3965, "step": 90510 }, { "epoch": 7.84, "learning_rate": 5.548022435526772e-05, "loss": 1.6792, "step": 90520 }, { "epoch": 7.84, "learning_rate": 5.5478055973169885e-05, "loss": 1.1344, "step": 90530 }, { "epoch": 7.84, "learning_rate": 5.5475887591072045e-05, "loss": 1.1716, "step": 90540 }, { "epoch": 7.84, "learning_rate": 5.547371920897421e-05, "loss": 1.402, "step": 90550 }, { "epoch": 7.84, "learning_rate": 5.5471550826876364e-05, "loss": 1.3474, "step": 90560 }, { "epoch": 7.84, "learning_rate": 5.546938244477853e-05, "loss": 1.4374, "step": 90570 }, { "epoch": 7.84, "learning_rate": 5.546721406268069e-05, "loss": 1.4972, "step": 90580 }, { "epoch": 7.85, "learning_rate": 5.546504568058286e-05, "loss": 1.5209, "step": 90590 }, { "epoch": 7.85, "learning_rate": 5.546287729848502e-05, "loss": 1.51, "step": 90600 }, { "epoch": 7.85, "learning_rate": 5.546070891638718e-05, "loss": 1.4249, "step": 90610 }, { "epoch": 7.85, "learning_rate": 5.545854053428935e-05, "loss": 1.3807, "step": 90620 }, { "epoch": 7.85, "learning_rate": 5.54563721521915e-05, "loss": 1.4469, "step": 90630 }, { "epoch": 7.85, "learning_rate": 5.545420377009367e-05, "loss": 1.3817, "step": 90640 }, { "epoch": 7.85, "learning_rate": 5.545203538799583e-05, "loss": 1.4104, "step": 90650 }, { "epoch": 7.85, "learning_rate": 5.5449867005897994e-05, "loss": 1.1344, "step": 90660 }, { "epoch": 7.85, "learning_rate": 5.544769862380016e-05, "loss": 1.4425, "step": 90670 }, { "epoch": 7.85, "learning_rate": 5.544553024170232e-05, "loss": 1.3646, "step": 90680 }, { "epoch": 7.85, "learning_rate": 5.544336185960449e-05, "loss": 1.314, "step": 90690 }, { "epoch": 7.86, "learning_rate": 5.5441193477506646e-05, "loss": 1.4625, "step": 90700 }, { "epoch": 7.86, "learning_rate": 5.5439025095408806e-05, "loss": 1.3702, "step": 90710 }, { "epoch": 7.86, "learning_rate": 5.5436856713310966e-05, "loss": 1.4185, "step": 90720 }, { "epoch": 7.86, "learning_rate": 5.543468833121313e-05, "loss": 1.3003, "step": 90730 }, { "epoch": 7.86, "learning_rate": 5.543251994911529e-05, "loss": 1.4434, "step": 90740 }, { "epoch": 7.86, "learning_rate": 5.543035156701746e-05, "loss": 1.4601, "step": 90750 }, { "epoch": 7.86, "learning_rate": 5.5428183184919624e-05, "loss": 1.5641, "step": 90760 }, { "epoch": 7.86, "learning_rate": 5.5426014802821784e-05, "loss": 1.3209, "step": 90770 }, { "epoch": 7.86, "learning_rate": 5.542384642072395e-05, "loss": 1.2743, "step": 90780 }, { "epoch": 7.86, "learning_rate": 5.54216780386261e-05, "loss": 1.1257, "step": 90790 }, { "epoch": 7.86, "learning_rate": 5.541950965652827e-05, "loss": 1.455, "step": 90800 }, { "epoch": 7.86, "learning_rate": 5.541734127443043e-05, "loss": 1.5082, "step": 90810 }, { "epoch": 7.87, "learning_rate": 5.5415172892332596e-05, "loss": 1.4102, "step": 90820 }, { "epoch": 7.87, "learning_rate": 5.541300451023476e-05, "loss": 1.4466, "step": 90830 }, { "epoch": 7.87, "learning_rate": 5.541083612813692e-05, "loss": 1.2426, "step": 90840 }, { "epoch": 7.87, "learning_rate": 5.540866774603909e-05, "loss": 1.3666, "step": 90850 }, { "epoch": 7.87, "learning_rate": 5.540649936394124e-05, "loss": 1.2796, "step": 90860 }, { "epoch": 7.87, "learning_rate": 5.540433098184341e-05, "loss": 1.2718, "step": 90870 }, { "epoch": 7.87, "learning_rate": 5.540216259974557e-05, "loss": 1.4012, "step": 90880 }, { "epoch": 7.87, "learning_rate": 5.539999421764773e-05, "loss": 1.1773, "step": 90890 }, { "epoch": 7.87, "learning_rate": 5.53978258355499e-05, "loss": 1.3184, "step": 90900 }, { "epoch": 7.87, "learning_rate": 5.539565745345206e-05, "loss": 1.3791, "step": 90910 }, { "epoch": 7.87, "learning_rate": 5.5393489071354226e-05, "loss": 1.3533, "step": 90920 }, { "epoch": 7.88, "learning_rate": 5.5391320689256385e-05, "loss": 1.3005, "step": 90930 }, { "epoch": 7.88, "learning_rate": 5.5389152307158545e-05, "loss": 1.3744, "step": 90940 }, { "epoch": 7.88, "learning_rate": 5.5386983925060704e-05, "loss": 1.2175, "step": 90950 }, { "epoch": 7.88, "learning_rate": 5.538481554296287e-05, "loss": 1.418, "step": 90960 }, { "epoch": 7.88, "learning_rate": 5.538264716086504e-05, "loss": 1.3443, "step": 90970 }, { "epoch": 7.88, "learning_rate": 5.53804787787672e-05, "loss": 1.3559, "step": 90980 }, { "epoch": 7.88, "learning_rate": 5.537831039666936e-05, "loss": 1.3357, "step": 90990 }, { "epoch": 7.88, "learning_rate": 5.537614201457152e-05, "loss": 1.4299, "step": 91000 }, { "epoch": 7.88, "learning_rate": 5.537397363247369e-05, "loss": 1.2963, "step": 91010 }, { "epoch": 7.88, "learning_rate": 5.537180525037584e-05, "loss": 1.4487, "step": 91020 }, { "epoch": 7.88, "learning_rate": 5.536963686827801e-05, "loss": 1.3033, "step": 91030 }, { "epoch": 7.88, "learning_rate": 5.5367468486180175e-05, "loss": 1.3399, "step": 91040 }, { "epoch": 7.89, "learning_rate": 5.5365300104082334e-05, "loss": 1.3997, "step": 91050 }, { "epoch": 7.89, "learning_rate": 5.53631317219845e-05, "loss": 1.4208, "step": 91060 }, { "epoch": 7.89, "learning_rate": 5.536096333988666e-05, "loss": 1.4465, "step": 91070 }, { "epoch": 7.89, "learning_rate": 5.535879495778883e-05, "loss": 1.4892, "step": 91080 }, { "epoch": 7.89, "learning_rate": 5.535662657569098e-05, "loss": 1.4224, "step": 91090 }, { "epoch": 7.89, "learning_rate": 5.5354458193593146e-05, "loss": 1.5136, "step": 91100 }, { "epoch": 7.89, "learning_rate": 5.535228981149531e-05, "loss": 1.5338, "step": 91110 }, { "epoch": 7.89, "learning_rate": 5.535012142939747e-05, "loss": 1.1753, "step": 91120 }, { "epoch": 7.89, "learning_rate": 5.534795304729964e-05, "loss": 1.3348, "step": 91130 }, { "epoch": 7.89, "learning_rate": 5.53457846652018e-05, "loss": 1.438, "step": 91140 }, { "epoch": 7.89, "learning_rate": 5.5343616283103965e-05, "loss": 1.4388, "step": 91150 }, { "epoch": 7.9, "learning_rate": 5.534144790100613e-05, "loss": 1.5529, "step": 91160 }, { "epoch": 7.9, "learning_rate": 5.5339279518908284e-05, "loss": 1.31, "step": 91170 }, { "epoch": 7.9, "learning_rate": 5.533711113681045e-05, "loss": 1.2971, "step": 91180 }, { "epoch": 7.9, "learning_rate": 5.533494275471261e-05, "loss": 1.5193, "step": 91190 }, { "epoch": 7.9, "learning_rate": 5.5332774372614776e-05, "loss": 1.3643, "step": 91200 }, { "epoch": 7.9, "learning_rate": 5.5330605990516936e-05, "loss": 1.3026, "step": 91210 }, { "epoch": 7.9, "learning_rate": 5.53284376084191e-05, "loss": 1.3412, "step": 91220 }, { "epoch": 7.9, "learning_rate": 5.532626922632127e-05, "loss": 1.5135, "step": 91230 }, { "epoch": 7.9, "learning_rate": 5.532410084422343e-05, "loss": 1.3212, "step": 91240 }, { "epoch": 7.9, "learning_rate": 5.532193246212558e-05, "loss": 1.2789, "step": 91250 }, { "epoch": 7.9, "learning_rate": 5.531976408002775e-05, "loss": 1.3665, "step": 91260 }, { "epoch": 7.9, "learning_rate": 5.5317595697929914e-05, "loss": 1.4414, "step": 91270 }, { "epoch": 7.91, "learning_rate": 5.5315427315832073e-05, "loss": 1.3379, "step": 91280 }, { "epoch": 7.91, "learning_rate": 5.531325893373424e-05, "loss": 1.3368, "step": 91290 }, { "epoch": 7.91, "learning_rate": 5.53110905516364e-05, "loss": 1.3917, "step": 91300 }, { "epoch": 7.91, "learning_rate": 5.5308922169538566e-05, "loss": 1.2331, "step": 91310 }, { "epoch": 7.91, "learning_rate": 5.530675378744073e-05, "loss": 1.3868, "step": 91320 }, { "epoch": 7.91, "learning_rate": 5.5304585405342885e-05, "loss": 1.3632, "step": 91330 }, { "epoch": 7.91, "learning_rate": 5.530241702324505e-05, "loss": 1.4966, "step": 91340 }, { "epoch": 7.91, "learning_rate": 5.530024864114721e-05, "loss": 1.3416, "step": 91350 }, { "epoch": 7.91, "learning_rate": 5.529808025904938e-05, "loss": 1.4339, "step": 91360 }, { "epoch": 7.91, "learning_rate": 5.529591187695154e-05, "loss": 1.5063, "step": 91370 }, { "epoch": 7.91, "learning_rate": 5.5293743494853703e-05, "loss": 1.4209, "step": 91380 }, { "epoch": 7.91, "learning_rate": 5.529157511275587e-05, "loss": 1.3167, "step": 91390 }, { "epoch": 7.92, "learning_rate": 5.528940673065802e-05, "loss": 1.3017, "step": 91400 }, { "epoch": 7.92, "learning_rate": 5.528723834856019e-05, "loss": 1.3565, "step": 91410 }, { "epoch": 7.92, "learning_rate": 5.528506996646235e-05, "loss": 1.4644, "step": 91420 }, { "epoch": 7.92, "learning_rate": 5.5282901584364515e-05, "loss": 1.4802, "step": 91430 }, { "epoch": 7.92, "learning_rate": 5.5280733202266675e-05, "loss": 1.4238, "step": 91440 }, { "epoch": 7.92, "learning_rate": 5.527856482016884e-05, "loss": 1.4238, "step": 91450 }, { "epoch": 7.92, "learning_rate": 5.527639643807101e-05, "loss": 1.4911, "step": 91460 }, { "epoch": 7.92, "learning_rate": 5.527422805597317e-05, "loss": 1.3773, "step": 91470 }, { "epoch": 7.92, "learning_rate": 5.527205967387533e-05, "loss": 1.6477, "step": 91480 }, { "epoch": 7.92, "learning_rate": 5.5269891291777486e-05, "loss": 1.2474, "step": 91490 }, { "epoch": 7.92, "learning_rate": 5.526772290967965e-05, "loss": 1.3662, "step": 91500 }, { "epoch": 7.93, "learning_rate": 5.526555452758181e-05, "loss": 1.1703, "step": 91510 }, { "epoch": 7.93, "learning_rate": 5.526338614548398e-05, "loss": 1.5658, "step": 91520 }, { "epoch": 7.93, "learning_rate": 5.5261217763386145e-05, "loss": 1.3855, "step": 91530 }, { "epoch": 7.93, "learning_rate": 5.5259049381288305e-05, "loss": 1.4973, "step": 91540 }, { "epoch": 7.93, "learning_rate": 5.525688099919047e-05, "loss": 1.4498, "step": 91550 }, { "epoch": 7.93, "learning_rate": 5.5254712617092624e-05, "loss": 1.3126, "step": 91560 }, { "epoch": 7.93, "learning_rate": 5.525254423499479e-05, "loss": 1.3391, "step": 91570 }, { "epoch": 7.93, "learning_rate": 5.525037585289695e-05, "loss": 1.1572, "step": 91580 }, { "epoch": 7.93, "learning_rate": 5.5248207470799116e-05, "loss": 1.3621, "step": 91590 }, { "epoch": 7.93, "learning_rate": 5.524603908870128e-05, "loss": 1.4392, "step": 91600 }, { "epoch": 7.93, "learning_rate": 5.524387070660344e-05, "loss": 1.3709, "step": 91610 }, { "epoch": 7.93, "learning_rate": 5.524170232450561e-05, "loss": 1.3457, "step": 91620 }, { "epoch": 7.94, "learning_rate": 5.523953394240776e-05, "loss": 1.3587, "step": 91630 }, { "epoch": 7.94, "learning_rate": 5.523736556030993e-05, "loss": 1.2769, "step": 91640 }, { "epoch": 7.94, "learning_rate": 5.523519717821209e-05, "loss": 1.4902, "step": 91650 }, { "epoch": 7.94, "learning_rate": 5.5233028796114254e-05, "loss": 1.3735, "step": 91660 }, { "epoch": 7.94, "learning_rate": 5.523086041401642e-05, "loss": 1.3121, "step": 91670 }, { "epoch": 7.94, "learning_rate": 5.522869203191858e-05, "loss": 1.3613, "step": 91680 }, { "epoch": 7.94, "learning_rate": 5.5226523649820746e-05, "loss": 1.4694, "step": 91690 }, { "epoch": 7.94, "learning_rate": 5.5224355267722906e-05, "loss": 1.2536, "step": 91700 }, { "epoch": 7.94, "learning_rate": 5.5222186885625066e-05, "loss": 1.2515, "step": 91710 }, { "epoch": 7.94, "learning_rate": 5.5220018503527225e-05, "loss": 1.2886, "step": 91720 }, { "epoch": 7.94, "learning_rate": 5.521785012142939e-05, "loss": 1.3985, "step": 91730 }, { "epoch": 7.95, "learning_rate": 5.521568173933156e-05, "loss": 1.2663, "step": 91740 }, { "epoch": 7.95, "learning_rate": 5.521351335723372e-05, "loss": 1.4282, "step": 91750 }, { "epoch": 7.95, "learning_rate": 5.5211344975135884e-05, "loss": 1.3717, "step": 91760 }, { "epoch": 7.95, "learning_rate": 5.5209176593038044e-05, "loss": 1.4752, "step": 91770 }, { "epoch": 7.95, "learning_rate": 5.520700821094021e-05, "loss": 1.1963, "step": 91780 }, { "epoch": 7.95, "learning_rate": 5.520483982884236e-05, "loss": 1.2088, "step": 91790 }, { "epoch": 7.95, "learning_rate": 5.520267144674453e-05, "loss": 1.4731, "step": 91800 }, { "epoch": 7.95, "learning_rate": 5.520050306464669e-05, "loss": 1.068, "step": 91810 }, { "epoch": 7.95, "learning_rate": 5.5198334682548855e-05, "loss": 1.381, "step": 91820 }, { "epoch": 7.95, "learning_rate": 5.519616630045102e-05, "loss": 1.3634, "step": 91830 }, { "epoch": 7.95, "learning_rate": 5.519399791835318e-05, "loss": 1.3296, "step": 91840 }, { "epoch": 7.95, "learning_rate": 5.519182953625535e-05, "loss": 1.3587, "step": 91850 }, { "epoch": 7.96, "learning_rate": 5.51896611541575e-05, "loss": 1.624, "step": 91860 }, { "epoch": 7.96, "learning_rate": 5.518749277205967e-05, "loss": 1.2887, "step": 91870 }, { "epoch": 7.96, "learning_rate": 5.5185324389961827e-05, "loss": 1.5531, "step": 91880 }, { "epoch": 7.96, "learning_rate": 5.518315600786399e-05, "loss": 1.4384, "step": 91890 }, { "epoch": 7.96, "learning_rate": 5.518098762576616e-05, "loss": 1.3884, "step": 91900 }, { "epoch": 7.96, "learning_rate": 5.517881924366832e-05, "loss": 1.3257, "step": 91910 }, { "epoch": 7.96, "learning_rate": 5.5176650861570485e-05, "loss": 1.4481, "step": 91920 }, { "epoch": 7.96, "learning_rate": 5.5174482479472645e-05, "loss": 1.4055, "step": 91930 }, { "epoch": 7.96, "learning_rate": 5.5172314097374805e-05, "loss": 1.2998, "step": 91940 }, { "epoch": 7.96, "learning_rate": 5.5170145715276964e-05, "loss": 1.4015, "step": 91950 }, { "epoch": 7.96, "learning_rate": 5.516797733317913e-05, "loss": 1.2749, "step": 91960 }, { "epoch": 7.97, "learning_rate": 5.51658089510813e-05, "loss": 1.4649, "step": 91970 }, { "epoch": 7.97, "learning_rate": 5.5163640568983457e-05, "loss": 1.4433, "step": 91980 }, { "epoch": 7.97, "learning_rate": 5.516147218688562e-05, "loss": 1.4984, "step": 91990 }, { "epoch": 7.97, "learning_rate": 5.515930380478778e-05, "loss": 1.109, "step": 92000 }, { "epoch": 7.97, "learning_rate": 5.515713542268995e-05, "loss": 1.189, "step": 92010 }, { "epoch": 7.97, "learning_rate": 5.51549670405921e-05, "loss": 1.525, "step": 92020 }, { "epoch": 7.97, "learning_rate": 5.515279865849427e-05, "loss": 1.2547, "step": 92030 }, { "epoch": 7.97, "learning_rate": 5.5150630276396435e-05, "loss": 1.3767, "step": 92040 }, { "epoch": 7.97, "learning_rate": 5.5148461894298594e-05, "loss": 1.6186, "step": 92050 }, { "epoch": 7.97, "learning_rate": 5.514629351220076e-05, "loss": 1.5217, "step": 92060 }, { "epoch": 7.97, "learning_rate": 5.514412513010292e-05, "loss": 1.7032, "step": 92070 }, { "epoch": 7.97, "learning_rate": 5.5141956748005087e-05, "loss": 1.396, "step": 92080 }, { "epoch": 7.98, "learning_rate": 5.513978836590725e-05, "loss": 1.1184, "step": 92090 }, { "epoch": 7.98, "learning_rate": 5.5137619983809406e-05, "loss": 1.6875, "step": 92100 }, { "epoch": 7.98, "learning_rate": 5.513545160171157e-05, "loss": 1.2532, "step": 92110 }, { "epoch": 7.98, "learning_rate": 5.513328321961373e-05, "loss": 1.3388, "step": 92120 }, { "epoch": 7.98, "learning_rate": 5.51311148375159e-05, "loss": 1.2081, "step": 92130 }, { "epoch": 7.98, "learning_rate": 5.512894645541806e-05, "loss": 1.3806, "step": 92140 }, { "epoch": 7.98, "learning_rate": 5.5126778073320224e-05, "loss": 1.5319, "step": 92150 }, { "epoch": 7.98, "learning_rate": 5.512460969122239e-05, "loss": 1.4165, "step": 92160 }, { "epoch": 7.98, "learning_rate": 5.5122441309124543e-05, "loss": 1.4036, "step": 92170 }, { "epoch": 7.98, "learning_rate": 5.512027292702671e-05, "loss": 1.2305, "step": 92180 }, { "epoch": 7.98, "learning_rate": 5.511810454492887e-05, "loss": 1.4251, "step": 92190 }, { "epoch": 7.99, "learning_rate": 5.5115936162831036e-05, "loss": 1.496, "step": 92200 }, { "epoch": 7.99, "learning_rate": 5.5113767780733195e-05, "loss": 1.4853, "step": 92210 }, { "epoch": 7.99, "learning_rate": 5.511159939863536e-05, "loss": 1.4133, "step": 92220 }, { "epoch": 7.99, "learning_rate": 5.510943101653753e-05, "loss": 1.4426, "step": 92230 }, { "epoch": 7.99, "learning_rate": 5.510726263443969e-05, "loss": 1.2326, "step": 92240 }, { "epoch": 7.99, "learning_rate": 5.510509425234185e-05, "loss": 1.3306, "step": 92250 }, { "epoch": 7.99, "learning_rate": 5.510292587024401e-05, "loss": 1.336, "step": 92260 }, { "epoch": 7.99, "learning_rate": 5.5100757488146173e-05, "loss": 1.3221, "step": 92270 }, { "epoch": 7.99, "learning_rate": 5.509858910604833e-05, "loss": 1.2368, "step": 92280 }, { "epoch": 7.99, "learning_rate": 5.50964207239505e-05, "loss": 1.3933, "step": 92290 }, { "epoch": 7.99, "learning_rate": 5.5094252341852666e-05, "loss": 1.3219, "step": 92300 }, { "epoch": 7.99, "learning_rate": 5.5092083959754825e-05, "loss": 1.1388, "step": 92310 }, { "epoch": 8.0, "learning_rate": 5.508991557765699e-05, "loss": 1.4218, "step": 92320 }, { "epoch": 8.0, "learning_rate": 5.5087747195559145e-05, "loss": 1.3195, "step": 92330 }, { "epoch": 8.0, "learning_rate": 5.508557881346131e-05, "loss": 1.4761, "step": 92340 }, { "epoch": 8.0, "learning_rate": 5.508341043136347e-05, "loss": 1.3618, "step": 92350 }, { "epoch": 8.0, "learning_rate": 5.508124204926564e-05, "loss": 1.6495, "step": 92360 }, { "epoch": 8.0, "learning_rate": 5.50790736671678e-05, "loss": 1.336, "step": 92370 }, { "epoch": 8.0, "eval_cer": 0.5487140092809919, "eval_em": 0.02475502836513667, "eval_f1": 0.02475502836513667, "eval_loss": 1.5162461996078491, "eval_runtime": 270.4403, "eval_samples_per_second": 7.17, "eval_steps_per_second": 0.899, "eval_wer": 0.9752449716348633, "step": 92372 }, { "epoch": 8.0, "learning_rate": 5.507690528506996e-05, "loss": 1.3605, "step": 92380 }, { "epoch": 8.0, "learning_rate": 5.507473690297213e-05, "loss": 1.5045, "step": 92390 }, { "epoch": 8.0, "learning_rate": 5.507256852087428e-05, "loss": 1.2913, "step": 92400 }, { "epoch": 8.0, "learning_rate": 5.507040013877645e-05, "loss": 1.209, "step": 92410 }, { "epoch": 8.0, "learning_rate": 5.506823175667861e-05, "loss": 1.1534, "step": 92420 }, { "epoch": 8.01, "learning_rate": 5.5066063374580775e-05, "loss": 1.4312, "step": 92430 }, { "epoch": 8.01, "learning_rate": 5.5063894992482934e-05, "loss": 1.3273, "step": 92440 }, { "epoch": 8.01, "learning_rate": 5.50617266103851e-05, "loss": 1.2055, "step": 92450 }, { "epoch": 8.01, "learning_rate": 5.505955822828727e-05, "loss": 1.2178, "step": 92460 }, { "epoch": 8.01, "learning_rate": 5.505738984618943e-05, "loss": 1.2204, "step": 92470 }, { "epoch": 8.01, "learning_rate": 5.5055221464091586e-05, "loss": 1.1837, "step": 92480 }, { "epoch": 8.01, "learning_rate": 5.5053053081993746e-05, "loss": 1.2093, "step": 92490 }, { "epoch": 8.01, "learning_rate": 5.505088469989591e-05, "loss": 1.2668, "step": 92500 }, { "epoch": 8.01, "learning_rate": 5.504871631779807e-05, "loss": 1.3058, "step": 92510 }, { "epoch": 8.01, "learning_rate": 5.504654793570024e-05, "loss": 1.3612, "step": 92520 }, { "epoch": 8.01, "learning_rate": 5.5044379553602405e-05, "loss": 1.0731, "step": 92530 }, { "epoch": 8.01, "learning_rate": 5.5042211171504564e-05, "loss": 1.2447, "step": 92540 }, { "epoch": 8.02, "learning_rate": 5.504004278940673e-05, "loss": 1.3457, "step": 92550 }, { "epoch": 8.02, "learning_rate": 5.5037874407308884e-05, "loss": 1.3822, "step": 92560 }, { "epoch": 8.02, "learning_rate": 5.503570602521105e-05, "loss": 1.4531, "step": 92570 }, { "epoch": 8.02, "learning_rate": 5.503353764311321e-05, "loss": 1.2181, "step": 92580 }, { "epoch": 8.02, "learning_rate": 5.5031369261015376e-05, "loss": 1.2174, "step": 92590 }, { "epoch": 8.02, "learning_rate": 5.502920087891754e-05, "loss": 1.3616, "step": 92600 }, { "epoch": 8.02, "learning_rate": 5.50270324968197e-05, "loss": 1.2841, "step": 92610 }, { "epoch": 8.02, "learning_rate": 5.502486411472187e-05, "loss": 1.4597, "step": 92620 }, { "epoch": 8.02, "learning_rate": 5.502269573262402e-05, "loss": 1.456, "step": 92630 }, { "epoch": 8.02, "learning_rate": 5.502052735052619e-05, "loss": 1.3189, "step": 92640 }, { "epoch": 8.02, "learning_rate": 5.501835896842835e-05, "loss": 1.264, "step": 92650 }, { "epoch": 8.02, "learning_rate": 5.5016190586330514e-05, "loss": 1.2294, "step": 92660 }, { "epoch": 8.03, "learning_rate": 5.501402220423268e-05, "loss": 1.4326, "step": 92670 }, { "epoch": 8.03, "learning_rate": 5.501185382213484e-05, "loss": 1.2361, "step": 92680 }, { "epoch": 8.03, "learning_rate": 5.5009685440037006e-05, "loss": 1.4642, "step": 92690 }, { "epoch": 8.03, "learning_rate": 5.5007517057939166e-05, "loss": 1.3099, "step": 92700 }, { "epoch": 8.03, "learning_rate": 5.5005348675841325e-05, "loss": 1.2631, "step": 92710 }, { "epoch": 8.03, "learning_rate": 5.5003180293743485e-05, "loss": 1.2084, "step": 92720 }, { "epoch": 8.03, "learning_rate": 5.500101191164565e-05, "loss": 1.3142, "step": 92730 }, { "epoch": 8.03, "learning_rate": 5.499884352954782e-05, "loss": 1.396, "step": 92740 }, { "epoch": 8.03, "learning_rate": 5.499667514744998e-05, "loss": 1.4717, "step": 92750 }, { "epoch": 8.03, "learning_rate": 5.4994506765352144e-05, "loss": 1.292, "step": 92760 }, { "epoch": 8.03, "learning_rate": 5.49923383832543e-05, "loss": 1.3589, "step": 92770 }, { "epoch": 8.04, "learning_rate": 5.499017000115647e-05, "loss": 1.2875, "step": 92780 }, { "epoch": 8.04, "learning_rate": 5.498800161905862e-05, "loss": 1.3063, "step": 92790 }, { "epoch": 8.04, "learning_rate": 5.498583323696079e-05, "loss": 1.6299, "step": 92800 }, { "epoch": 8.04, "learning_rate": 5.4983664854862955e-05, "loss": 1.2165, "step": 92810 }, { "epoch": 8.04, "learning_rate": 5.4981496472765115e-05, "loss": 1.2882, "step": 92820 }, { "epoch": 8.04, "learning_rate": 5.497932809066728e-05, "loss": 1.4171, "step": 92830 }, { "epoch": 8.04, "learning_rate": 5.497715970856944e-05, "loss": 1.245, "step": 92840 }, { "epoch": 8.04, "learning_rate": 5.497499132647161e-05, "loss": 1.2785, "step": 92850 }, { "epoch": 8.04, "learning_rate": 5.4972822944373774e-05, "loss": 1.2342, "step": 92860 }, { "epoch": 8.04, "learning_rate": 5.4970654562275927e-05, "loss": 1.2624, "step": 92870 }, { "epoch": 8.04, "learning_rate": 5.4968486180178086e-05, "loss": 1.2748, "step": 92880 }, { "epoch": 8.04, "learning_rate": 5.496631779808025e-05, "loss": 1.1168, "step": 92890 }, { "epoch": 8.05, "learning_rate": 5.496414941598242e-05, "loss": 1.3622, "step": 92900 }, { "epoch": 8.05, "learning_rate": 5.496198103388458e-05, "loss": 1.4041, "step": 92910 }, { "epoch": 8.05, "learning_rate": 5.4959812651786745e-05, "loss": 1.3944, "step": 92920 }, { "epoch": 8.05, "learning_rate": 5.4957644269688905e-05, "loss": 1.275, "step": 92930 }, { "epoch": 8.05, "learning_rate": 5.4955475887591064e-05, "loss": 1.5149, "step": 92940 }, { "epoch": 8.05, "learning_rate": 5.4953307505493224e-05, "loss": 1.435, "step": 92950 }, { "epoch": 8.05, "learning_rate": 5.495113912339539e-05, "loss": 1.279, "step": 92960 }, { "epoch": 8.05, "learning_rate": 5.494897074129756e-05, "loss": 1.3896, "step": 92970 }, { "epoch": 8.05, "learning_rate": 5.4946802359199716e-05, "loss": 1.4419, "step": 92980 }, { "epoch": 8.05, "learning_rate": 5.494463397710188e-05, "loss": 1.1035, "step": 92990 }, { "epoch": 8.05, "learning_rate": 5.494246559500404e-05, "loss": 1.1678, "step": 93000 }, { "epoch": 8.06, "learning_rate": 5.494029721290621e-05, "loss": 1.3462, "step": 93010 }, { "epoch": 8.06, "learning_rate": 5.493812883080836e-05, "loss": 1.1921, "step": 93020 }, { "epoch": 8.06, "learning_rate": 5.493596044871053e-05, "loss": 1.2169, "step": 93030 }, { "epoch": 8.06, "learning_rate": 5.4933792066612694e-05, "loss": 1.2005, "step": 93040 }, { "epoch": 8.06, "learning_rate": 5.4931623684514854e-05, "loss": 1.1214, "step": 93050 }, { "epoch": 8.06, "learning_rate": 5.492945530241702e-05, "loss": 1.3945, "step": 93060 }, { "epoch": 8.06, "learning_rate": 5.492728692031918e-05, "loss": 1.2851, "step": 93070 }, { "epoch": 8.06, "learning_rate": 5.4925118538221346e-05, "loss": 1.226, "step": 93080 }, { "epoch": 8.06, "learning_rate": 5.492295015612351e-05, "loss": 1.1767, "step": 93090 }, { "epoch": 8.06, "learning_rate": 5.4920781774025666e-05, "loss": 1.449, "step": 93100 }, { "epoch": 8.06, "learning_rate": 5.491861339192783e-05, "loss": 1.2631, "step": 93110 }, { "epoch": 8.06, "learning_rate": 5.491644500982999e-05, "loss": 1.3633, "step": 93120 }, { "epoch": 8.07, "learning_rate": 5.491427662773216e-05, "loss": 1.1969, "step": 93130 }, { "epoch": 8.07, "learning_rate": 5.491210824563432e-05, "loss": 1.3087, "step": 93140 }, { "epoch": 8.07, "learning_rate": 5.4909939863536484e-05, "loss": 1.4537, "step": 93150 }, { "epoch": 8.07, "learning_rate": 5.490777148143865e-05, "loss": 1.171, "step": 93160 }, { "epoch": 8.07, "learning_rate": 5.49056030993408e-05, "loss": 1.2425, "step": 93170 }, { "epoch": 8.07, "learning_rate": 5.490343471724297e-05, "loss": 1.3153, "step": 93180 }, { "epoch": 8.07, "learning_rate": 5.490126633514513e-05, "loss": 1.2058, "step": 93190 }, { "epoch": 8.07, "learning_rate": 5.4899097953047296e-05, "loss": 1.2408, "step": 93200 }, { "epoch": 8.07, "learning_rate": 5.4896929570949455e-05, "loss": 1.3381, "step": 93210 }, { "epoch": 8.07, "learning_rate": 5.489476118885162e-05, "loss": 1.1532, "step": 93220 }, { "epoch": 8.07, "learning_rate": 5.489259280675379e-05, "loss": 1.3525, "step": 93230 }, { "epoch": 8.08, "learning_rate": 5.489042442465595e-05, "loss": 1.4376, "step": 93240 }, { "epoch": 8.08, "learning_rate": 5.488825604255811e-05, "loss": 1.342, "step": 93250 }, { "epoch": 8.08, "learning_rate": 5.488608766046027e-05, "loss": 1.4057, "step": 93260 }, { "epoch": 8.08, "learning_rate": 5.488391927836243e-05, "loss": 1.245, "step": 93270 }, { "epoch": 8.08, "learning_rate": 5.488175089626459e-05, "loss": 1.3079, "step": 93280 }, { "epoch": 8.08, "learning_rate": 5.487958251416676e-05, "loss": 1.3177, "step": 93290 }, { "epoch": 8.08, "learning_rate": 5.4877414132068926e-05, "loss": 1.1438, "step": 93300 }, { "epoch": 8.08, "learning_rate": 5.4875245749971085e-05, "loss": 1.3042, "step": 93310 }, { "epoch": 8.08, "learning_rate": 5.487307736787325e-05, "loss": 1.1509, "step": 93320 }, { "epoch": 8.08, "learning_rate": 5.4870908985775404e-05, "loss": 1.295, "step": 93330 }, { "epoch": 8.08, "learning_rate": 5.486874060367757e-05, "loss": 1.3846, "step": 93340 }, { "epoch": 8.08, "learning_rate": 5.486657222157973e-05, "loss": 1.2347, "step": 93350 }, { "epoch": 8.09, "learning_rate": 5.48644038394819e-05, "loss": 1.2766, "step": 93360 }, { "epoch": 8.09, "learning_rate": 5.486223545738406e-05, "loss": 1.5337, "step": 93370 }, { "epoch": 8.09, "learning_rate": 5.486006707528622e-05, "loss": 1.3416, "step": 93380 }, { "epoch": 8.09, "learning_rate": 5.485789869318839e-05, "loss": 1.4436, "step": 93390 }, { "epoch": 8.09, "learning_rate": 5.485573031109054e-05, "loss": 1.2326, "step": 93400 }, { "epoch": 8.09, "learning_rate": 5.485356192899271e-05, "loss": 1.4048, "step": 93410 }, { "epoch": 8.09, "learning_rate": 5.485139354689487e-05, "loss": 1.3476, "step": 93420 }, { "epoch": 8.09, "learning_rate": 5.4849225164797034e-05, "loss": 1.6143, "step": 93430 }, { "epoch": 8.09, "learning_rate": 5.4847056782699194e-05, "loss": 1.3875, "step": 93440 }, { "epoch": 8.09, "learning_rate": 5.484488840060136e-05, "loss": 1.2764, "step": 93450 }, { "epoch": 8.09, "learning_rate": 5.484272001850353e-05, "loss": 1.2652, "step": 93460 }, { "epoch": 8.1, "learning_rate": 5.4840551636405686e-05, "loss": 1.4395, "step": 93470 }, { "epoch": 8.1, "learning_rate": 5.4838383254307846e-05, "loss": 1.1733, "step": 93480 }, { "epoch": 8.1, "learning_rate": 5.4836214872210006e-05, "loss": 1.2641, "step": 93490 }, { "epoch": 8.1, "learning_rate": 5.483404649011217e-05, "loss": 1.3536, "step": 93500 }, { "epoch": 8.1, "learning_rate": 5.483187810801433e-05, "loss": 1.1964, "step": 93510 }, { "epoch": 8.1, "learning_rate": 5.48297097259165e-05, "loss": 1.2609, "step": 93520 }, { "epoch": 8.1, "learning_rate": 5.4827541343818664e-05, "loss": 1.1429, "step": 93530 }, { "epoch": 8.1, "learning_rate": 5.4825372961720824e-05, "loss": 1.2785, "step": 93540 }, { "epoch": 8.1, "learning_rate": 5.482320457962299e-05, "loss": 1.1165, "step": 93550 }, { "epoch": 8.1, "learning_rate": 5.482103619752514e-05, "loss": 1.2377, "step": 93560 }, { "epoch": 8.1, "learning_rate": 5.481886781542731e-05, "loss": 1.1931, "step": 93570 }, { "epoch": 8.1, "learning_rate": 5.481669943332947e-05, "loss": 1.3675, "step": 93580 }, { "epoch": 8.11, "learning_rate": 5.4814531051231636e-05, "loss": 1.3603, "step": 93590 }, { "epoch": 8.11, "learning_rate": 5.48123626691338e-05, "loss": 1.2814, "step": 93600 }, { "epoch": 8.11, "learning_rate": 5.481019428703596e-05, "loss": 1.1459, "step": 93610 }, { "epoch": 8.11, "learning_rate": 5.480802590493813e-05, "loss": 1.4105, "step": 93620 }, { "epoch": 8.11, "learning_rate": 5.480585752284029e-05, "loss": 1.3014, "step": 93630 }, { "epoch": 8.11, "learning_rate": 5.480368914074245e-05, "loss": 1.2611, "step": 93640 }, { "epoch": 8.11, "learning_rate": 5.480152075864461e-05, "loss": 1.4513, "step": 93650 }, { "epoch": 8.11, "learning_rate": 5.479935237654677e-05, "loss": 1.2507, "step": 93660 }, { "epoch": 8.11, "learning_rate": 5.479718399444894e-05, "loss": 1.4311, "step": 93670 }, { "epoch": 8.11, "learning_rate": 5.47950156123511e-05, "loss": 1.4066, "step": 93680 }, { "epoch": 8.11, "learning_rate": 5.4792847230253266e-05, "loss": 1.5229, "step": 93690 }, { "epoch": 8.12, "learning_rate": 5.4790678848155425e-05, "loss": 1.2947, "step": 93700 }, { "epoch": 8.12, "learning_rate": 5.4788510466057585e-05, "loss": 1.2089, "step": 93710 }, { "epoch": 8.12, "learning_rate": 5.4786342083959745e-05, "loss": 1.2085, "step": 93720 }, { "epoch": 8.12, "learning_rate": 5.478417370186191e-05, "loss": 1.0925, "step": 93730 }, { "epoch": 8.12, "learning_rate": 5.478200531976408e-05, "loss": 1.4044, "step": 93740 }, { "epoch": 8.12, "learning_rate": 5.477983693766624e-05, "loss": 1.297, "step": 93750 }, { "epoch": 8.12, "learning_rate": 5.4777668555568403e-05, "loss": 1.1918, "step": 93760 }, { "epoch": 8.12, "learning_rate": 5.477550017347056e-05, "loss": 1.341, "step": 93770 }, { "epoch": 8.12, "learning_rate": 5.477333179137273e-05, "loss": 1.4168, "step": 93780 }, { "epoch": 8.12, "learning_rate": 5.477116340927488e-05, "loss": 1.139, "step": 93790 }, { "epoch": 8.12, "learning_rate": 5.476899502717705e-05, "loss": 1.2938, "step": 93800 }, { "epoch": 8.12, "learning_rate": 5.4766826645079215e-05, "loss": 1.1596, "step": 93810 }, { "epoch": 8.13, "learning_rate": 5.4764658262981375e-05, "loss": 1.1909, "step": 93820 }, { "epoch": 8.13, "learning_rate": 5.476248988088354e-05, "loss": 1.2422, "step": 93830 }, { "epoch": 8.13, "learning_rate": 5.47603214987857e-05, "loss": 1.2216, "step": 93840 }, { "epoch": 8.13, "learning_rate": 5.475815311668787e-05, "loss": 1.193, "step": 93850 }, { "epoch": 8.13, "learning_rate": 5.4755984734590033e-05, "loss": 1.3176, "step": 93860 }, { "epoch": 8.13, "learning_rate": 5.4753816352492186e-05, "loss": 1.313, "step": 93870 }, { "epoch": 8.13, "learning_rate": 5.475164797039435e-05, "loss": 1.3228, "step": 93880 }, { "epoch": 8.13, "learning_rate": 5.474947958829651e-05, "loss": 1.2522, "step": 93890 }, { "epoch": 8.13, "learning_rate": 5.474731120619868e-05, "loss": 1.3032, "step": 93900 }, { "epoch": 8.13, "learning_rate": 5.474514282410084e-05, "loss": 1.4757, "step": 93910 }, { "epoch": 8.13, "learning_rate": 5.4742974442003005e-05, "loss": 1.3283, "step": 93920 }, { "epoch": 8.13, "learning_rate": 5.474080605990517e-05, "loss": 1.0838, "step": 93930 }, { "epoch": 8.14, "learning_rate": 5.4738637677807324e-05, "loss": 1.484, "step": 93940 }, { "epoch": 8.14, "learning_rate": 5.4736469295709484e-05, "loss": 1.2789, "step": 93950 }, { "epoch": 8.14, "learning_rate": 5.473430091361165e-05, "loss": 1.3132, "step": 93960 }, { "epoch": 8.14, "learning_rate": 5.4732132531513816e-05, "loss": 1.3503, "step": 93970 }, { "epoch": 8.14, "learning_rate": 5.4729964149415976e-05, "loss": 1.0932, "step": 93980 }, { "epoch": 8.14, "learning_rate": 5.472779576731814e-05, "loss": 1.3117, "step": 93990 }, { "epoch": 8.14, "learning_rate": 5.47256273852203e-05, "loss": 1.4432, "step": 94000 }, { "epoch": 8.14, "learning_rate": 5.472345900312247e-05, "loss": 1.1748, "step": 94010 }, { "epoch": 8.14, "learning_rate": 5.472129062102462e-05, "loss": 1.2107, "step": 94020 }, { "epoch": 8.14, "learning_rate": 5.471912223892679e-05, "loss": 1.4047, "step": 94030 }, { "epoch": 8.14, "learning_rate": 5.4716953856828954e-05, "loss": 1.2252, "step": 94040 }, { "epoch": 8.15, "learning_rate": 5.4714785474731114e-05, "loss": 1.2448, "step": 94050 }, { "epoch": 8.15, "learning_rate": 5.471261709263328e-05, "loss": 1.4203, "step": 94060 }, { "epoch": 8.15, "learning_rate": 5.471044871053544e-05, "loss": 1.5404, "step": 94070 }, { "epoch": 8.15, "learning_rate": 5.4708280328437606e-05, "loss": 1.5031, "step": 94080 }, { "epoch": 8.15, "learning_rate": 5.470611194633977e-05, "loss": 1.406, "step": 94090 }, { "epoch": 8.15, "learning_rate": 5.4703943564241925e-05, "loss": 1.4459, "step": 94100 }, { "epoch": 8.15, "learning_rate": 5.470177518214409e-05, "loss": 1.4434, "step": 94110 }, { "epoch": 8.15, "learning_rate": 5.469960680004625e-05, "loss": 1.2867, "step": 94120 }, { "epoch": 8.15, "learning_rate": 5.469743841794842e-05, "loss": 1.2054, "step": 94130 }, { "epoch": 8.15, "learning_rate": 5.469527003585058e-05, "loss": 1.2034, "step": 94140 }, { "epoch": 8.15, "learning_rate": 5.4693101653752744e-05, "loss": 1.3718, "step": 94150 }, { "epoch": 8.15, "learning_rate": 5.469093327165491e-05, "loss": 1.3838, "step": 94160 }, { "epoch": 8.16, "learning_rate": 5.468876488955706e-05, "loss": 1.2763, "step": 94170 }, { "epoch": 8.16, "learning_rate": 5.468659650745923e-05, "loss": 1.1414, "step": 94180 }, { "epoch": 8.16, "learning_rate": 5.468442812536139e-05, "loss": 1.3645, "step": 94190 }, { "epoch": 8.16, "learning_rate": 5.4682259743263555e-05, "loss": 1.0457, "step": 94200 }, { "epoch": 8.16, "learning_rate": 5.4680091361165715e-05, "loss": 1.5296, "step": 94210 }, { "epoch": 8.16, "learning_rate": 5.467792297906788e-05, "loss": 1.3532, "step": 94220 }, { "epoch": 8.16, "learning_rate": 5.467575459697005e-05, "loss": 1.1353, "step": 94230 }, { "epoch": 8.16, "learning_rate": 5.467358621487221e-05, "loss": 1.2468, "step": 94240 }, { "epoch": 8.16, "learning_rate": 5.467141783277437e-05, "loss": 1.233, "step": 94250 }, { "epoch": 8.16, "learning_rate": 5.4669249450676526e-05, "loss": 1.2646, "step": 94260 }, { "epoch": 8.16, "learning_rate": 5.466708106857869e-05, "loss": 1.2625, "step": 94270 }, { "epoch": 8.17, "learning_rate": 5.466491268648085e-05, "loss": 1.2743, "step": 94280 }, { "epoch": 8.17, "learning_rate": 5.466274430438302e-05, "loss": 1.1093, "step": 94290 }, { "epoch": 8.17, "learning_rate": 5.4660575922285185e-05, "loss": 1.2945, "step": 94300 }, { "epoch": 8.17, "learning_rate": 5.4658407540187345e-05, "loss": 1.3654, "step": 94310 }, { "epoch": 8.17, "learning_rate": 5.465623915808951e-05, "loss": 1.386, "step": 94320 }, { "epoch": 8.17, "learning_rate": 5.4654070775991664e-05, "loss": 1.4044, "step": 94330 }, { "epoch": 8.17, "learning_rate": 5.465190239389383e-05, "loss": 1.2538, "step": 94340 }, { "epoch": 8.17, "learning_rate": 5.464973401179599e-05, "loss": 1.3227, "step": 94350 }, { "epoch": 8.17, "learning_rate": 5.4647565629698157e-05, "loss": 1.2695, "step": 94360 }, { "epoch": 8.17, "learning_rate": 5.464539724760032e-05, "loss": 1.3289, "step": 94370 }, { "epoch": 8.17, "learning_rate": 5.464322886550248e-05, "loss": 1.2028, "step": 94380 }, { "epoch": 8.17, "learning_rate": 5.464106048340465e-05, "loss": 1.6242, "step": 94390 }, { "epoch": 8.18, "learning_rate": 5.463889210130681e-05, "loss": 1.3068, "step": 94400 }, { "epoch": 8.18, "learning_rate": 5.463672371920897e-05, "loss": 1.2064, "step": 94410 }, { "epoch": 8.18, "learning_rate": 5.463455533711113e-05, "loss": 1.5729, "step": 94420 }, { "epoch": 8.18, "learning_rate": 5.4632386955013294e-05, "loss": 1.2384, "step": 94430 }, { "epoch": 8.18, "learning_rate": 5.463021857291546e-05, "loss": 1.3734, "step": 94440 }, { "epoch": 8.18, "learning_rate": 5.462805019081762e-05, "loss": 1.2838, "step": 94450 }, { "epoch": 8.18, "learning_rate": 5.4625881808719787e-05, "loss": 1.1651, "step": 94460 }, { "epoch": 8.18, "learning_rate": 5.4623713426621946e-05, "loss": 1.4036, "step": 94470 }, { "epoch": 8.18, "learning_rate": 5.4621545044524106e-05, "loss": 1.4447, "step": 94480 }, { "epoch": 8.18, "learning_rate": 5.4619376662426265e-05, "loss": 1.1928, "step": 94490 }, { "epoch": 8.18, "learning_rate": 5.461720828032843e-05, "loss": 1.3734, "step": 94500 }, { "epoch": 8.19, "learning_rate": 5.461503989823059e-05, "loss": 1.3511, "step": 94510 }, { "epoch": 8.19, "learning_rate": 5.461287151613276e-05, "loss": 1.3894, "step": 94520 }, { "epoch": 8.19, "learning_rate": 5.4610703134034924e-05, "loss": 1.2241, "step": 94530 }, { "epoch": 8.19, "learning_rate": 5.4608534751937084e-05, "loss": 1.3246, "step": 94540 }, { "epoch": 8.19, "learning_rate": 5.460636636983925e-05, "loss": 1.4367, "step": 94550 }, { "epoch": 8.19, "learning_rate": 5.46041979877414e-05, "loss": 1.3484, "step": 94560 }, { "epoch": 8.19, "learning_rate": 5.460202960564357e-05, "loss": 1.4017, "step": 94570 }, { "epoch": 8.19, "learning_rate": 5.459986122354573e-05, "loss": 1.2686, "step": 94580 }, { "epoch": 8.19, "learning_rate": 5.4597692841447895e-05, "loss": 1.1674, "step": 94590 }, { "epoch": 8.19, "learning_rate": 5.459552445935006e-05, "loss": 1.2984, "step": 94600 }, { "epoch": 8.19, "learning_rate": 5.459335607725222e-05, "loss": 1.2356, "step": 94610 }, { "epoch": 8.19, "learning_rate": 5.459118769515439e-05, "loss": 1.3295, "step": 94620 }, { "epoch": 8.2, "learning_rate": 5.458901931305655e-05, "loss": 1.3945, "step": 94630 }, { "epoch": 8.2, "learning_rate": 5.458685093095871e-05, "loss": 1.4134, "step": 94640 }, { "epoch": 8.2, "learning_rate": 5.458468254886087e-05, "loss": 1.223, "step": 94650 }, { "epoch": 8.2, "learning_rate": 5.458251416676303e-05, "loss": 1.428, "step": 94660 }, { "epoch": 8.2, "learning_rate": 5.45803457846652e-05, "loss": 1.3403, "step": 94670 }, { "epoch": 8.2, "learning_rate": 5.457817740256736e-05, "loss": 1.2543, "step": 94680 }, { "epoch": 8.2, "learning_rate": 5.4576009020469525e-05, "loss": 1.1582, "step": 94690 }, { "epoch": 8.2, "learning_rate": 5.4573840638371685e-05, "loss": 1.4164, "step": 94700 }, { "epoch": 8.2, "learning_rate": 5.4571672256273845e-05, "loss": 1.3283, "step": 94710 }, { "epoch": 8.2, "learning_rate": 5.4569503874176004e-05, "loss": 1.226, "step": 94720 }, { "epoch": 8.2, "learning_rate": 5.456733549207817e-05, "loss": 1.2507, "step": 94730 }, { "epoch": 8.21, "learning_rate": 5.456516710998034e-05, "loss": 1.4304, "step": 94740 }, { "epoch": 8.21, "learning_rate": 5.45629987278825e-05, "loss": 1.5617, "step": 94750 }, { "epoch": 8.21, "learning_rate": 5.456083034578466e-05, "loss": 1.2278, "step": 94760 }, { "epoch": 8.21, "learning_rate": 5.455866196368682e-05, "loss": 1.1865, "step": 94770 }, { "epoch": 8.21, "learning_rate": 5.455649358158899e-05, "loss": 1.4356, "step": 94780 }, { "epoch": 8.21, "learning_rate": 5.455432519949114e-05, "loss": 1.3908, "step": 94790 }, { "epoch": 8.21, "learning_rate": 5.455215681739331e-05, "loss": 1.3388, "step": 94800 }, { "epoch": 8.21, "learning_rate": 5.4549988435295475e-05, "loss": 1.333, "step": 94810 }, { "epoch": 8.21, "learning_rate": 5.4547820053197634e-05, "loss": 1.012, "step": 94820 }, { "epoch": 8.21, "learning_rate": 5.45456516710998e-05, "loss": 1.3961, "step": 94830 }, { "epoch": 8.21, "learning_rate": 5.454348328900196e-05, "loss": 1.3202, "step": 94840 }, { "epoch": 8.21, "learning_rate": 5.454131490690413e-05, "loss": 1.2145, "step": 94850 }, { "epoch": 8.22, "learning_rate": 5.453914652480629e-05, "loss": 1.3004, "step": 94860 }, { "epoch": 8.22, "learning_rate": 5.4536978142708446e-05, "loss": 1.4829, "step": 94870 }, { "epoch": 8.22, "learning_rate": 5.453480976061061e-05, "loss": 1.3255, "step": 94880 }, { "epoch": 8.22, "learning_rate": 5.453264137851277e-05, "loss": 1.1795, "step": 94890 }, { "epoch": 8.22, "learning_rate": 5.453047299641494e-05, "loss": 1.4067, "step": 94900 }, { "epoch": 8.22, "learning_rate": 5.45283046143171e-05, "loss": 1.272, "step": 94910 }, { "epoch": 8.22, "learning_rate": 5.4526136232219264e-05, "loss": 1.3324, "step": 94920 }, { "epoch": 8.22, "learning_rate": 5.452396785012143e-05, "loss": 1.3068, "step": 94930 }, { "epoch": 8.22, "learning_rate": 5.4521799468023584e-05, "loss": 1.0834, "step": 94940 }, { "epoch": 8.22, "learning_rate": 5.451963108592575e-05, "loss": 1.1308, "step": 94950 }, { "epoch": 8.22, "learning_rate": 5.451746270382791e-05, "loss": 1.1977, "step": 94960 }, { "epoch": 8.23, "learning_rate": 5.4515294321730076e-05, "loss": 1.4054, "step": 94970 }, { "epoch": 8.23, "learning_rate": 5.4513125939632236e-05, "loss": 1.2905, "step": 94980 }, { "epoch": 8.23, "learning_rate": 5.45109575575344e-05, "loss": 1.1742, "step": 94990 }, { "epoch": 8.23, "learning_rate": 5.450878917543657e-05, "loss": 1.2367, "step": 95000 }, { "epoch": 8.23, "learning_rate": 5.450662079333873e-05, "loss": 1.2743, "step": 95010 }, { "epoch": 8.23, "learning_rate": 5.450445241124088e-05, "loss": 1.3225, "step": 95020 }, { "epoch": 8.23, "learning_rate": 5.450228402914305e-05, "loss": 1.2068, "step": 95030 }, { "epoch": 8.23, "learning_rate": 5.4500115647045214e-05, "loss": 1.3758, "step": 95040 }, { "epoch": 8.23, "learning_rate": 5.449794726494737e-05, "loss": 1.3227, "step": 95050 }, { "epoch": 8.23, "learning_rate": 5.449577888284954e-05, "loss": 1.3568, "step": 95060 }, { "epoch": 8.23, "learning_rate": 5.44936105007517e-05, "loss": 1.3312, "step": 95070 }, { "epoch": 8.23, "learning_rate": 5.4491442118653866e-05, "loss": 1.3876, "step": 95080 }, { "epoch": 8.24, "learning_rate": 5.448927373655603e-05, "loss": 1.3812, "step": 95090 }, { "epoch": 8.24, "learning_rate": 5.4487105354458185e-05, "loss": 1.2564, "step": 95100 }, { "epoch": 8.24, "learning_rate": 5.448493697236035e-05, "loss": 1.3955, "step": 95110 }, { "epoch": 8.24, "learning_rate": 5.448276859026251e-05, "loss": 1.3106, "step": 95120 }, { "epoch": 8.24, "learning_rate": 5.448060020816468e-05, "loss": 1.2347, "step": 95130 }, { "epoch": 8.24, "learning_rate": 5.447843182606684e-05, "loss": 1.499, "step": 95140 }, { "epoch": 8.24, "learning_rate": 5.4476263443969e-05, "loss": 1.1584, "step": 95150 }, { "epoch": 8.24, "learning_rate": 5.447409506187117e-05, "loss": 1.22, "step": 95160 }, { "epoch": 8.24, "learning_rate": 5.447192667977333e-05, "loss": 1.2005, "step": 95170 }, { "epoch": 8.24, "learning_rate": 5.446975829767549e-05, "loss": 1.0512, "step": 95180 }, { "epoch": 8.24, "learning_rate": 5.446758991557765e-05, "loss": 1.3139, "step": 95190 }, { "epoch": 8.24, "learning_rate": 5.4465421533479815e-05, "loss": 1.1823, "step": 95200 }, { "epoch": 8.25, "learning_rate": 5.4463253151381975e-05, "loss": 1.1793, "step": 95210 }, { "epoch": 8.25, "learning_rate": 5.446108476928414e-05, "loss": 1.4461, "step": 95220 }, { "epoch": 8.25, "learning_rate": 5.445891638718631e-05, "loss": 1.3002, "step": 95230 }, { "epoch": 8.25, "learning_rate": 5.445674800508847e-05, "loss": 1.3766, "step": 95240 }, { "epoch": 8.25, "learning_rate": 5.4454579622990627e-05, "loss": 1.3731, "step": 95250 }, { "epoch": 8.25, "learning_rate": 5.4452411240892786e-05, "loss": 1.2052, "step": 95260 }, { "epoch": 8.25, "learning_rate": 5.445024285879495e-05, "loss": 1.5991, "step": 95270 }, { "epoch": 8.25, "learning_rate": 5.444807447669711e-05, "loss": 1.6172, "step": 95280 }, { "epoch": 8.25, "learning_rate": 5.444590609459928e-05, "loss": 1.3489, "step": 95290 }, { "epoch": 8.25, "learning_rate": 5.4443737712501445e-05, "loss": 1.2852, "step": 95300 }, { "epoch": 8.25, "learning_rate": 5.4441569330403605e-05, "loss": 1.3879, "step": 95310 }, { "epoch": 8.26, "learning_rate": 5.443940094830577e-05, "loss": 1.2179, "step": 95320 }, { "epoch": 8.26, "learning_rate": 5.4437232566207924e-05, "loss": 1.2707, "step": 95330 }, { "epoch": 8.26, "learning_rate": 5.443506418411009e-05, "loss": 1.2042, "step": 95340 }, { "epoch": 8.26, "learning_rate": 5.443289580201225e-05, "loss": 1.3622, "step": 95350 }, { "epoch": 8.26, "learning_rate": 5.4430727419914416e-05, "loss": 1.1726, "step": 95360 }, { "epoch": 8.26, "learning_rate": 5.442855903781658e-05, "loss": 1.3756, "step": 95370 }, { "epoch": 8.26, "learning_rate": 5.442639065571874e-05, "loss": 1.1757, "step": 95380 }, { "epoch": 8.26, "learning_rate": 5.442422227362091e-05, "loss": 1.3961, "step": 95390 }, { "epoch": 8.26, "learning_rate": 5.442205389152307e-05, "loss": 1.2718, "step": 95400 }, { "epoch": 8.26, "learning_rate": 5.441988550942523e-05, "loss": 1.3195, "step": 95410 }, { "epoch": 8.26, "learning_rate": 5.441771712732739e-05, "loss": 1.2827, "step": 95420 }, { "epoch": 8.26, "learning_rate": 5.4415548745229554e-05, "loss": 1.3161, "step": 95430 }, { "epoch": 8.27, "learning_rate": 5.441338036313172e-05, "loss": 1.2455, "step": 95440 }, { "epoch": 8.27, "learning_rate": 5.441121198103388e-05, "loss": 1.1833, "step": 95450 }, { "epoch": 8.27, "learning_rate": 5.4409043598936046e-05, "loss": 1.4176, "step": 95460 }, { "epoch": 8.27, "learning_rate": 5.4406875216838206e-05, "loss": 1.3364, "step": 95470 }, { "epoch": 8.27, "learning_rate": 5.4404706834740365e-05, "loss": 1.2793, "step": 95480 }, { "epoch": 8.27, "learning_rate": 5.4402538452642525e-05, "loss": 1.4377, "step": 95490 }, { "epoch": 8.27, "learning_rate": 5.440037007054469e-05, "loss": 1.3828, "step": 95500 }, { "epoch": 8.27, "learning_rate": 5.439820168844686e-05, "loss": 1.5036, "step": 95510 }, { "epoch": 8.27, "learning_rate": 5.439603330634902e-05, "loss": 1.2702, "step": 95520 }, { "epoch": 8.27, "learning_rate": 5.4393864924251184e-05, "loss": 1.4813, "step": 95530 }, { "epoch": 8.27, "learning_rate": 5.4391696542153343e-05, "loss": 1.3868, "step": 95540 }, { "epoch": 8.28, "learning_rate": 5.438952816005551e-05, "loss": 1.3094, "step": 95550 }, { "epoch": 8.28, "learning_rate": 5.438735977795766e-05, "loss": 1.3585, "step": 95560 }, { "epoch": 8.28, "learning_rate": 5.438519139585983e-05, "loss": 1.0841, "step": 95570 }, { "epoch": 8.28, "learning_rate": 5.438302301376199e-05, "loss": 1.2638, "step": 95580 }, { "epoch": 8.28, "learning_rate": 5.4380854631664155e-05, "loss": 1.3213, "step": 95590 }, { "epoch": 8.28, "learning_rate": 5.437868624956632e-05, "loss": 1.2386, "step": 95600 }, { "epoch": 8.28, "learning_rate": 5.437651786746848e-05, "loss": 1.0765, "step": 95610 }, { "epoch": 8.28, "learning_rate": 5.437434948537065e-05, "loss": 1.342, "step": 95620 }, { "epoch": 8.28, "learning_rate": 5.437218110327281e-05, "loss": 1.498, "step": 95630 }, { "epoch": 8.28, "learning_rate": 5.437001272117497e-05, "loss": 1.352, "step": 95640 }, { "epoch": 8.28, "learning_rate": 5.4367844339077126e-05, "loss": 1.2299, "step": 95650 }, { "epoch": 8.28, "learning_rate": 5.436567595697929e-05, "loss": 1.2521, "step": 95660 }, { "epoch": 8.29, "learning_rate": 5.436350757488146e-05, "loss": 1.3385, "step": 95670 }, { "epoch": 8.29, "learning_rate": 5.436133919278362e-05, "loss": 1.3556, "step": 95680 }, { "epoch": 8.29, "learning_rate": 5.4359170810685785e-05, "loss": 1.1453, "step": 95690 }, { "epoch": 8.29, "learning_rate": 5.4357002428587945e-05, "loss": 1.3057, "step": 95700 }, { "epoch": 8.29, "learning_rate": 5.4354834046490104e-05, "loss": 1.1141, "step": 95710 }, { "epoch": 8.29, "learning_rate": 5.4352665664392264e-05, "loss": 1.4946, "step": 95720 }, { "epoch": 8.29, "learning_rate": 5.435049728229443e-05, "loss": 1.5653, "step": 95730 }, { "epoch": 8.29, "learning_rate": 5.43483289001966e-05, "loss": 1.414, "step": 95740 }, { "epoch": 8.29, "learning_rate": 5.4346160518098756e-05, "loss": 1.2851, "step": 95750 }, { "epoch": 8.29, "learning_rate": 5.434399213600092e-05, "loss": 1.3027, "step": 95760 }, { "epoch": 8.29, "learning_rate": 5.434182375390308e-05, "loss": 1.2672, "step": 95770 }, { "epoch": 8.3, "learning_rate": 5.433965537180525e-05, "loss": 1.2678, "step": 95780 }, { "epoch": 8.3, "learning_rate": 5.43374869897074e-05, "loss": 1.2467, "step": 95790 }, { "epoch": 8.3, "learning_rate": 5.433531860760957e-05, "loss": 1.4026, "step": 95800 }, { "epoch": 8.3, "learning_rate": 5.4333150225511734e-05, "loss": 1.132, "step": 95810 }, { "epoch": 8.3, "learning_rate": 5.4330981843413894e-05, "loss": 1.4495, "step": 95820 }, { "epoch": 8.3, "learning_rate": 5.432881346131606e-05, "loss": 1.3677, "step": 95830 }, { "epoch": 8.3, "learning_rate": 5.432664507921822e-05, "loss": 1.3182, "step": 95840 }, { "epoch": 8.3, "learning_rate": 5.4324476697120386e-05, "loss": 1.3899, "step": 95850 }, { "epoch": 8.3, "learning_rate": 5.432230831502255e-05, "loss": 1.2775, "step": 95860 }, { "epoch": 8.3, "learning_rate": 5.4320139932924706e-05, "loss": 1.106, "step": 95870 }, { "epoch": 8.3, "learning_rate": 5.431797155082687e-05, "loss": 1.205, "step": 95880 }, { "epoch": 8.3, "learning_rate": 5.431580316872903e-05, "loss": 1.3637, "step": 95890 }, { "epoch": 8.31, "learning_rate": 5.43136347866312e-05, "loss": 1.2297, "step": 95900 }, { "epoch": 8.31, "learning_rate": 5.431146640453336e-05, "loss": 1.4373, "step": 95910 }, { "epoch": 8.31, "learning_rate": 5.4309298022435524e-05, "loss": 1.4293, "step": 95920 }, { "epoch": 8.31, "learning_rate": 5.430712964033769e-05, "loss": 1.3126, "step": 95930 }, { "epoch": 8.31, "learning_rate": 5.430496125823985e-05, "loss": 1.3959, "step": 95940 }, { "epoch": 8.31, "learning_rate": 5.430279287614201e-05, "loss": 1.2526, "step": 95950 }, { "epoch": 8.31, "learning_rate": 5.430062449404417e-05, "loss": 1.2823, "step": 95960 }, { "epoch": 8.31, "learning_rate": 5.4298456111946336e-05, "loss": 1.3345, "step": 95970 }, { "epoch": 8.31, "learning_rate": 5.4296287729848495e-05, "loss": 1.3453, "step": 95980 }, { "epoch": 8.31, "learning_rate": 5.429411934775066e-05, "loss": 1.2866, "step": 95990 }, { "epoch": 8.31, "learning_rate": 5.429195096565283e-05, "loss": 1.3674, "step": 96000 }, { "epoch": 8.32, "learning_rate": 5.428978258355499e-05, "loss": 1.1989, "step": 96010 }, { "epoch": 8.32, "learning_rate": 5.428761420145715e-05, "loss": 1.3322, "step": 96020 }, { "epoch": 8.32, "learning_rate": 5.428544581935931e-05, "loss": 1.4508, "step": 96030 }, { "epoch": 8.32, "learning_rate": 5.428327743726147e-05, "loss": 1.3153, "step": 96040 }, { "epoch": 8.32, "learning_rate": 5.428110905516363e-05, "loss": 1.137, "step": 96050 }, { "epoch": 8.32, "learning_rate": 5.42789406730658e-05, "loss": 1.2878, "step": 96060 }, { "epoch": 8.32, "learning_rate": 5.4276772290967966e-05, "loss": 1.359, "step": 96070 }, { "epoch": 8.32, "learning_rate": 5.4274603908870125e-05, "loss": 1.3065, "step": 96080 }, { "epoch": 8.32, "learning_rate": 5.427243552677229e-05, "loss": 1.2075, "step": 96090 }, { "epoch": 8.32, "learning_rate": 5.4270267144674445e-05, "loss": 1.3186, "step": 96100 }, { "epoch": 8.32, "learning_rate": 5.426809876257661e-05, "loss": 1.5276, "step": 96110 }, { "epoch": 8.32, "learning_rate": 5.426593038047877e-05, "loss": 1.4272, "step": 96120 }, { "epoch": 8.33, "learning_rate": 5.426376199838094e-05, "loss": 1.2419, "step": 96130 }, { "epoch": 8.33, "learning_rate": 5.4261593616283097e-05, "loss": 1.2698, "step": 96140 }, { "epoch": 8.33, "learning_rate": 5.425942523418526e-05, "loss": 1.28, "step": 96150 }, { "epoch": 8.33, "learning_rate": 5.425725685208743e-05, "loss": 1.3802, "step": 96160 }, { "epoch": 8.33, "learning_rate": 5.425508846998959e-05, "loss": 1.2229, "step": 96170 }, { "epoch": 8.33, "learning_rate": 5.425292008789175e-05, "loss": 1.3259, "step": 96180 }, { "epoch": 8.33, "learning_rate": 5.425075170579391e-05, "loss": 1.2785, "step": 96190 }, { "epoch": 8.33, "learning_rate": 5.4248583323696075e-05, "loss": 1.1668, "step": 96200 }, { "epoch": 8.33, "learning_rate": 5.4246414941598234e-05, "loss": 1.377, "step": 96210 }, { "epoch": 8.33, "learning_rate": 5.42442465595004e-05, "loss": 1.3009, "step": 96220 }, { "epoch": 8.33, "learning_rate": 5.424207817740257e-05, "loss": 1.3478, "step": 96230 }, { "epoch": 8.33, "learning_rate": 5.423990979530473e-05, "loss": 1.2502, "step": 96240 }, { "epoch": 8.34, "learning_rate": 5.4237741413206886e-05, "loss": 1.482, "step": 96250 }, { "epoch": 8.34, "learning_rate": 5.4235573031109046e-05, "loss": 1.3546, "step": 96260 }, { "epoch": 8.34, "learning_rate": 5.423340464901121e-05, "loss": 1.2934, "step": 96270 }, { "epoch": 8.34, "learning_rate": 5.423123626691337e-05, "loss": 1.1595, "step": 96280 }, { "epoch": 8.34, "learning_rate": 5.422906788481554e-05, "loss": 1.3794, "step": 96290 }, { "epoch": 8.34, "learning_rate": 5.4226899502717705e-05, "loss": 1.3114, "step": 96300 }, { "epoch": 8.34, "learning_rate": 5.4224731120619864e-05, "loss": 1.2868, "step": 96310 }, { "epoch": 8.34, "learning_rate": 5.422256273852203e-05, "loss": 1.3642, "step": 96320 }, { "epoch": 8.34, "learning_rate": 5.4220394356424184e-05, "loss": 1.2671, "step": 96330 }, { "epoch": 8.34, "learning_rate": 5.421822597432635e-05, "loss": 1.2764, "step": 96340 }, { "epoch": 8.34, "learning_rate": 5.421605759222851e-05, "loss": 1.3673, "step": 96350 }, { "epoch": 8.35, "learning_rate": 5.4213889210130676e-05, "loss": 1.4141, "step": 96360 }, { "epoch": 8.35, "learning_rate": 5.421172082803284e-05, "loss": 1.1345, "step": 96370 }, { "epoch": 8.35, "learning_rate": 5.4209552445935e-05, "loss": 1.5972, "step": 96380 }, { "epoch": 8.35, "learning_rate": 5.420738406383717e-05, "loss": 1.167, "step": 96390 }, { "epoch": 8.35, "learning_rate": 5.420521568173933e-05, "loss": 1.5486, "step": 96400 }, { "epoch": 8.35, "learning_rate": 5.420304729964149e-05, "loss": 1.1987, "step": 96410 }, { "epoch": 8.35, "learning_rate": 5.420087891754365e-05, "loss": 1.4725, "step": 96420 }, { "epoch": 8.35, "learning_rate": 5.4198710535445814e-05, "loss": 1.3619, "step": 96430 }, { "epoch": 8.35, "learning_rate": 5.419654215334798e-05, "loss": 1.4208, "step": 96440 }, { "epoch": 8.35, "learning_rate": 5.419437377125014e-05, "loss": 1.1214, "step": 96450 }, { "epoch": 8.35, "learning_rate": 5.4192205389152306e-05, "loss": 1.4077, "step": 96460 }, { "epoch": 8.35, "learning_rate": 5.4190037007054466e-05, "loss": 1.2569, "step": 96470 }, { "epoch": 8.36, "learning_rate": 5.4187868624956625e-05, "loss": 1.3359, "step": 96480 }, { "epoch": 8.36, "learning_rate": 5.4185700242858785e-05, "loss": 1.52, "step": 96490 }, { "epoch": 8.36, "learning_rate": 5.418353186076095e-05, "loss": 1.1423, "step": 96500 }, { "epoch": 8.36, "learning_rate": 5.418136347866312e-05, "loss": 1.4422, "step": 96510 }, { "epoch": 8.36, "learning_rate": 5.417919509656528e-05, "loss": 1.4649, "step": 96520 }, { "epoch": 8.36, "learning_rate": 5.4177026714467444e-05, "loss": 1.2738, "step": 96530 }, { "epoch": 8.36, "learning_rate": 5.41748583323696e-05, "loss": 1.224, "step": 96540 }, { "epoch": 8.36, "learning_rate": 5.417268995027177e-05, "loss": 1.2712, "step": 96550 }, { "epoch": 8.36, "learning_rate": 5.417052156817392e-05, "loss": 1.3818, "step": 96560 }, { "epoch": 8.36, "learning_rate": 5.416835318607609e-05, "loss": 1.2806, "step": 96570 }, { "epoch": 8.36, "learning_rate": 5.4166184803978255e-05, "loss": 1.2714, "step": 96580 }, { "epoch": 8.37, "learning_rate": 5.4164016421880415e-05, "loss": 1.203, "step": 96590 }, { "epoch": 8.37, "learning_rate": 5.416184803978258e-05, "loss": 1.3364, "step": 96600 }, { "epoch": 8.37, "learning_rate": 5.415967965768474e-05, "loss": 1.3362, "step": 96610 }, { "epoch": 8.37, "learning_rate": 5.415751127558691e-05, "loss": 1.3851, "step": 96620 }, { "epoch": 8.37, "learning_rate": 5.4155342893489074e-05, "loss": 1.2199, "step": 96630 }, { "epoch": 8.37, "learning_rate": 5.4153174511391226e-05, "loss": 1.3534, "step": 96640 }, { "epoch": 8.37, "learning_rate": 5.4151006129293386e-05, "loss": 1.2997, "step": 96650 }, { "epoch": 8.37, "learning_rate": 5.414883774719555e-05, "loss": 1.3598, "step": 96660 }, { "epoch": 8.37, "learning_rate": 5.414666936509772e-05, "loss": 1.2455, "step": 96670 }, { "epoch": 8.37, "learning_rate": 5.414450098299988e-05, "loss": 1.3444, "step": 96680 }, { "epoch": 8.37, "learning_rate": 5.4142332600902045e-05, "loss": 1.1207, "step": 96690 }, { "epoch": 8.37, "learning_rate": 5.4140164218804204e-05, "loss": 1.412, "step": 96700 }, { "epoch": 8.38, "learning_rate": 5.413799583670637e-05, "loss": 1.3101, "step": 96710 }, { "epoch": 8.38, "learning_rate": 5.4135827454608524e-05, "loss": 1.1281, "step": 96720 }, { "epoch": 8.38, "learning_rate": 5.413365907251069e-05, "loss": 1.1705, "step": 96730 }, { "epoch": 8.38, "learning_rate": 5.4131490690412856e-05, "loss": 1.125, "step": 96740 }, { "epoch": 8.38, "learning_rate": 5.4129322308315016e-05, "loss": 1.3939, "step": 96750 }, { "epoch": 8.38, "learning_rate": 5.412715392621718e-05, "loss": 1.1376, "step": 96760 }, { "epoch": 8.38, "learning_rate": 5.412498554411934e-05, "loss": 1.334, "step": 96770 }, { "epoch": 8.38, "learning_rate": 5.412281716202151e-05, "loss": 1.2181, "step": 96780 }, { "epoch": 8.38, "learning_rate": 5.412064877992366e-05, "loss": 1.4566, "step": 96790 }, { "epoch": 8.38, "learning_rate": 5.411848039782583e-05, "loss": 1.3144, "step": 96800 }, { "epoch": 8.38, "learning_rate": 5.4116312015727994e-05, "loss": 1.1197, "step": 96810 }, { "epoch": 8.39, "learning_rate": 5.4114143633630154e-05, "loss": 1.2195, "step": 96820 }, { "epoch": 8.39, "learning_rate": 5.411197525153232e-05, "loss": 1.252, "step": 96830 }, { "epoch": 8.39, "learning_rate": 5.410980686943448e-05, "loss": 1.3549, "step": 96840 }, { "epoch": 8.39, "learning_rate": 5.4107638487336646e-05, "loss": 1.4703, "step": 96850 }, { "epoch": 8.39, "learning_rate": 5.410547010523881e-05, "loss": 1.4929, "step": 96860 }, { "epoch": 8.39, "learning_rate": 5.4103301723140965e-05, "loss": 1.2263, "step": 96870 }, { "epoch": 8.39, "learning_rate": 5.410113334104313e-05, "loss": 1.1955, "step": 96880 }, { "epoch": 8.39, "learning_rate": 5.409896495894529e-05, "loss": 1.2412, "step": 96890 }, { "epoch": 8.39, "learning_rate": 5.409679657684746e-05, "loss": 1.2821, "step": 96900 }, { "epoch": 8.39, "learning_rate": 5.409462819474962e-05, "loss": 1.1768, "step": 96910 }, { "epoch": 8.39, "learning_rate": 5.4092459812651784e-05, "loss": 1.2621, "step": 96920 }, { "epoch": 8.39, "learning_rate": 5.409029143055395e-05, "loss": 1.327, "step": 96930 }, { "epoch": 8.4, "learning_rate": 5.408812304845611e-05, "loss": 1.3934, "step": 96940 }, { "epoch": 8.4, "learning_rate": 5.408595466635827e-05, "loss": 1.2535, "step": 96950 }, { "epoch": 8.4, "learning_rate": 5.408378628426043e-05, "loss": 1.5328, "step": 96960 }, { "epoch": 8.4, "learning_rate": 5.4081617902162595e-05, "loss": 1.3663, "step": 96970 }, { "epoch": 8.4, "learning_rate": 5.4079449520064755e-05, "loss": 1.4568, "step": 96980 }, { "epoch": 8.4, "learning_rate": 5.407728113796692e-05, "loss": 1.1775, "step": 96990 }, { "epoch": 8.4, "learning_rate": 5.407511275586909e-05, "loss": 1.1775, "step": 97000 }, { "epoch": 8.4, "learning_rate": 5.407294437377125e-05, "loss": 1.391, "step": 97010 }, { "epoch": 8.4, "learning_rate": 5.407077599167341e-05, "loss": 1.4569, "step": 97020 }, { "epoch": 8.4, "learning_rate": 5.406860760957557e-05, "loss": 1.2675, "step": 97030 }, { "epoch": 8.4, "learning_rate": 5.406643922747773e-05, "loss": 1.3879, "step": 97040 }, { "epoch": 8.41, "learning_rate": 5.406427084537989e-05, "loss": 1.3737, "step": 97050 }, { "epoch": 8.41, "learning_rate": 5.406210246328206e-05, "loss": 1.3178, "step": 97060 }, { "epoch": 8.41, "learning_rate": 5.4059934081184225e-05, "loss": 1.2551, "step": 97070 }, { "epoch": 8.41, "learning_rate": 5.4057765699086385e-05, "loss": 1.2701, "step": 97080 }, { "epoch": 8.41, "learning_rate": 5.405559731698855e-05, "loss": 1.2948, "step": 97090 }, { "epoch": 8.41, "learning_rate": 5.4053428934890704e-05, "loss": 1.5779, "step": 97100 }, { "epoch": 8.41, "learning_rate": 5.405126055279287e-05, "loss": 1.3164, "step": 97110 }, { "epoch": 8.41, "learning_rate": 5.404909217069503e-05, "loss": 1.4332, "step": 97120 }, { "epoch": 8.41, "learning_rate": 5.40469237885972e-05, "loss": 1.2946, "step": 97130 }, { "epoch": 8.41, "learning_rate": 5.404475540649936e-05, "loss": 1.3751, "step": 97140 }, { "epoch": 8.41, "learning_rate": 5.404258702440152e-05, "loss": 1.4315, "step": 97150 }, { "epoch": 8.41, "learning_rate": 5.404041864230369e-05, "loss": 1.4733, "step": 97160 }, { "epoch": 8.42, "learning_rate": 5.403825026020585e-05, "loss": 1.3093, "step": 97170 }, { "epoch": 8.42, "learning_rate": 5.403608187810801e-05, "loss": 1.1592, "step": 97180 }, { "epoch": 8.42, "learning_rate": 5.403391349601017e-05, "loss": 1.2736, "step": 97190 }, { "epoch": 8.42, "learning_rate": 5.4031745113912334e-05, "loss": 1.169, "step": 97200 }, { "epoch": 8.42, "learning_rate": 5.4029576731814494e-05, "loss": 1.4606, "step": 97210 }, { "epoch": 8.42, "learning_rate": 5.402740834971666e-05, "loss": 1.2148, "step": 97220 }, { "epoch": 8.42, "learning_rate": 5.402523996761883e-05, "loss": 1.3322, "step": 97230 }, { "epoch": 8.42, "learning_rate": 5.4023071585520986e-05, "loss": 1.2236, "step": 97240 }, { "epoch": 8.42, "learning_rate": 5.4020903203423146e-05, "loss": 1.3478, "step": 97250 }, { "epoch": 8.42, "learning_rate": 5.4018734821325306e-05, "loss": 1.2624, "step": 97260 }, { "epoch": 8.42, "learning_rate": 5.401656643922747e-05, "loss": 1.4968, "step": 97270 }, { "epoch": 8.43, "learning_rate": 5.401439805712963e-05, "loss": 1.3335, "step": 97280 }, { "epoch": 8.43, "learning_rate": 5.40122296750318e-05, "loss": 1.2869, "step": 97290 }, { "epoch": 8.43, "learning_rate": 5.4010061292933964e-05, "loss": 1.3381, "step": 97300 }, { "epoch": 8.43, "learning_rate": 5.4007892910836124e-05, "loss": 1.4418, "step": 97310 }, { "epoch": 8.43, "learning_rate": 5.400572452873829e-05, "loss": 1.3442, "step": 97320 }, { "epoch": 8.43, "learning_rate": 5.400355614664044e-05, "loss": 1.1252, "step": 97330 }, { "epoch": 8.43, "learning_rate": 5.400138776454261e-05, "loss": 1.2419, "step": 97340 }, { "epoch": 8.43, "learning_rate": 5.399921938244477e-05, "loss": 1.3338, "step": 97350 }, { "epoch": 8.43, "learning_rate": 5.3997051000346936e-05, "loss": 1.2137, "step": 97360 }, { "epoch": 8.43, "learning_rate": 5.39948826182491e-05, "loss": 1.3411, "step": 97370 }, { "epoch": 8.43, "learning_rate": 5.399271423615126e-05, "loss": 1.2783, "step": 97380 }, { "epoch": 8.43, "learning_rate": 5.399054585405343e-05, "loss": 1.4298, "step": 97390 }, { "epoch": 8.44, "learning_rate": 5.398837747195559e-05, "loss": 1.3151, "step": 97400 }, { "epoch": 8.44, "learning_rate": 5.398620908985775e-05, "loss": 1.3928, "step": 97410 }, { "epoch": 8.44, "learning_rate": 5.398404070775991e-05, "loss": 1.332, "step": 97420 }, { "epoch": 8.44, "learning_rate": 5.398187232566207e-05, "loss": 1.2912, "step": 97430 }, { "epoch": 8.44, "learning_rate": 5.397970394356424e-05, "loss": 1.3375, "step": 97440 }, { "epoch": 8.44, "learning_rate": 5.39775355614664e-05, "loss": 1.2851, "step": 97450 }, { "epoch": 8.44, "learning_rate": 5.3975367179368566e-05, "loss": 1.1287, "step": 97460 }, { "epoch": 8.44, "learning_rate": 5.3973198797270725e-05, "loss": 1.4041, "step": 97470 }, { "epoch": 8.44, "learning_rate": 5.397103041517289e-05, "loss": 1.213, "step": 97480 }, { "epoch": 8.44, "learning_rate": 5.3968862033075044e-05, "loss": 1.2655, "step": 97490 }, { "epoch": 8.44, "learning_rate": 5.396669365097721e-05, "loss": 0.9591, "step": 97500 }, { "epoch": 8.44, "learning_rate": 5.396452526887938e-05, "loss": 1.2048, "step": 97510 }, { "epoch": 8.45, "learning_rate": 5.396235688678154e-05, "loss": 1.499, "step": 97520 }, { "epoch": 8.45, "learning_rate": 5.39601885046837e-05, "loss": 1.4223, "step": 97530 }, { "epoch": 8.45, "learning_rate": 5.395802012258586e-05, "loss": 1.5186, "step": 97540 }, { "epoch": 8.45, "learning_rate": 5.395585174048803e-05, "loss": 1.2228, "step": 97550 }, { "epoch": 8.45, "learning_rate": 5.395368335839018e-05, "loss": 1.2347, "step": 97560 }, { "epoch": 8.45, "learning_rate": 5.395151497629235e-05, "loss": 1.1436, "step": 97570 }, { "epoch": 8.45, "learning_rate": 5.3949346594194515e-05, "loss": 1.3057, "step": 97580 }, { "epoch": 8.45, "learning_rate": 5.3947178212096675e-05, "loss": 1.3308, "step": 97590 }, { "epoch": 8.45, "learning_rate": 5.394500982999884e-05, "loss": 1.3525, "step": 97600 }, { "epoch": 8.45, "learning_rate": 5.3942841447901e-05, "loss": 1.0129, "step": 97610 }, { "epoch": 8.45, "learning_rate": 5.394067306580317e-05, "loss": 1.4524, "step": 97620 }, { "epoch": 8.46, "learning_rate": 5.393850468370533e-05, "loss": 1.2151, "step": 97630 }, { "epoch": 8.46, "learning_rate": 5.3936336301607486e-05, "loss": 1.3952, "step": 97640 }, { "epoch": 8.46, "learning_rate": 5.393416791950965e-05, "loss": 1.4489, "step": 97650 }, { "epoch": 8.46, "learning_rate": 5.393199953741181e-05, "loss": 1.21, "step": 97660 }, { "epoch": 8.46, "learning_rate": 5.392983115531398e-05, "loss": 1.4644, "step": 97670 }, { "epoch": 8.46, "learning_rate": 5.392766277321614e-05, "loss": 1.2138, "step": 97680 }, { "epoch": 8.46, "learning_rate": 5.3925494391118305e-05, "loss": 1.2593, "step": 97690 }, { "epoch": 8.46, "learning_rate": 5.392332600902047e-05, "loss": 1.1332, "step": 97700 }, { "epoch": 8.46, "learning_rate": 5.392115762692263e-05, "loss": 1.1851, "step": 97710 }, { "epoch": 8.46, "learning_rate": 5.3918989244824783e-05, "loss": 1.2729, "step": 97720 }, { "epoch": 8.46, "learning_rate": 5.391682086272695e-05, "loss": 1.1994, "step": 97730 }, { "epoch": 8.46, "learning_rate": 5.3914652480629116e-05, "loss": 1.4146, "step": 97740 }, { "epoch": 8.47, "learning_rate": 5.3912484098531276e-05, "loss": 1.3056, "step": 97750 }, { "epoch": 8.47, "learning_rate": 5.391031571643344e-05, "loss": 1.2168, "step": 97760 }, { "epoch": 8.47, "learning_rate": 5.39081473343356e-05, "loss": 1.3859, "step": 97770 }, { "epoch": 8.47, "learning_rate": 5.390597895223777e-05, "loss": 1.1683, "step": 97780 }, { "epoch": 8.47, "learning_rate": 5.390381057013992e-05, "loss": 1.3717, "step": 97790 }, { "epoch": 8.47, "learning_rate": 5.390164218804209e-05, "loss": 1.4081, "step": 97800 }, { "epoch": 8.47, "learning_rate": 5.3899473805944254e-05, "loss": 1.176, "step": 97810 }, { "epoch": 8.47, "learning_rate": 5.3897305423846413e-05, "loss": 1.294, "step": 97820 }, { "epoch": 8.47, "learning_rate": 5.389513704174858e-05, "loss": 1.3442, "step": 97830 }, { "epoch": 8.47, "learning_rate": 5.389296865965074e-05, "loss": 1.3781, "step": 97840 }, { "epoch": 8.47, "learning_rate": 5.3890800277552906e-05, "loss": 1.2645, "step": 97850 }, { "epoch": 8.48, "learning_rate": 5.388863189545507e-05, "loss": 1.1604, "step": 97860 }, { "epoch": 8.48, "learning_rate": 5.3886463513357225e-05, "loss": 1.425, "step": 97870 }, { "epoch": 8.48, "learning_rate": 5.388429513125939e-05, "loss": 1.2984, "step": 97880 }, { "epoch": 8.48, "learning_rate": 5.388212674916155e-05, "loss": 1.037, "step": 97890 }, { "epoch": 8.48, "learning_rate": 5.387995836706372e-05, "loss": 1.2952, "step": 97900 }, { "epoch": 8.48, "learning_rate": 5.387778998496588e-05, "loss": 1.395, "step": 97910 }, { "epoch": 8.48, "learning_rate": 5.3875621602868043e-05, "loss": 1.3415, "step": 97920 }, { "epoch": 8.48, "learning_rate": 5.387345322077021e-05, "loss": 1.5581, "step": 97930 }, { "epoch": 8.48, "learning_rate": 5.387128483867237e-05, "loss": 1.2601, "step": 97940 }, { "epoch": 8.48, "learning_rate": 5.386911645657453e-05, "loss": 1.3186, "step": 97950 }, { "epoch": 8.48, "learning_rate": 5.386694807447669e-05, "loss": 1.1991, "step": 97960 }, { "epoch": 8.48, "learning_rate": 5.3864779692378855e-05, "loss": 1.4142, "step": 97970 }, { "epoch": 8.49, "learning_rate": 5.3862611310281015e-05, "loss": 1.5095, "step": 97980 }, { "epoch": 8.49, "learning_rate": 5.386044292818318e-05, "loss": 1.3819, "step": 97990 }, { "epoch": 8.49, "learning_rate": 5.385827454608535e-05, "loss": 1.4002, "step": 98000 }, { "epoch": 8.49, "learning_rate": 5.385610616398751e-05, "loss": 1.4059, "step": 98010 }, { "epoch": 8.49, "learning_rate": 5.385393778188967e-05, "loss": 1.4154, "step": 98020 }, { "epoch": 8.49, "learning_rate": 5.3851769399791826e-05, "loss": 1.3034, "step": 98030 }, { "epoch": 8.49, "learning_rate": 5.384960101769399e-05, "loss": 1.2848, "step": 98040 }, { "epoch": 8.49, "learning_rate": 5.384743263559615e-05, "loss": 1.3187, "step": 98050 }, { "epoch": 8.49, "learning_rate": 5.384526425349832e-05, "loss": 1.4903, "step": 98060 }, { "epoch": 8.49, "learning_rate": 5.3843095871400485e-05, "loss": 1.0838, "step": 98070 }, { "epoch": 8.49, "learning_rate": 5.3840927489302645e-05, "loss": 1.0846, "step": 98080 }, { "epoch": 8.5, "learning_rate": 5.383875910720481e-05, "loss": 1.2388, "step": 98090 }, { "epoch": 8.5, "learning_rate": 5.3836590725106964e-05, "loss": 1.2433, "step": 98100 }, { "epoch": 8.5, "learning_rate": 5.383442234300913e-05, "loss": 1.1895, "step": 98110 }, { "epoch": 8.5, "learning_rate": 5.383225396091129e-05, "loss": 1.4264, "step": 98120 }, { "epoch": 8.5, "learning_rate": 5.3830085578813456e-05, "loss": 1.1645, "step": 98130 }, { "epoch": 8.5, "learning_rate": 5.382791719671562e-05, "loss": 1.4414, "step": 98140 }, { "epoch": 8.5, "learning_rate": 5.382574881461778e-05, "loss": 1.4067, "step": 98150 }, { "epoch": 8.5, "learning_rate": 5.382358043251995e-05, "loss": 1.42, "step": 98160 }, { "epoch": 8.5, "learning_rate": 5.382141205042211e-05, "loss": 1.3909, "step": 98170 }, { "epoch": 8.5, "learning_rate": 5.381924366832427e-05, "loss": 1.4505, "step": 98180 }, { "epoch": 8.5, "learning_rate": 5.381707528622643e-05, "loss": 1.1749, "step": 98190 }, { "epoch": 8.5, "learning_rate": 5.3814906904128594e-05, "loss": 1.459, "step": 98200 }, { "epoch": 8.51, "learning_rate": 5.381273852203076e-05, "loss": 1.4194, "step": 98210 }, { "epoch": 8.51, "learning_rate": 5.381057013993292e-05, "loss": 1.4648, "step": 98220 }, { "epoch": 8.51, "learning_rate": 5.3808401757835086e-05, "loss": 1.2224, "step": 98230 }, { "epoch": 8.51, "learning_rate": 5.3806233375737246e-05, "loss": 1.2893, "step": 98240 }, { "epoch": 8.51, "learning_rate": 5.380406499363941e-05, "loss": 1.2171, "step": 98250 }, { "epoch": 8.51, "learning_rate": 5.3801896611541565e-05, "loss": 1.2774, "step": 98260 }, { "epoch": 8.51, "learning_rate": 5.379972822944373e-05, "loss": 1.3055, "step": 98270 }, { "epoch": 8.51, "learning_rate": 5.379755984734589e-05, "loss": 1.3642, "step": 98280 }, { "epoch": 8.51, "learning_rate": 5.379539146524806e-05, "loss": 1.181, "step": 98290 }, { "epoch": 8.51, "learning_rate": 5.3793223083150224e-05, "loss": 1.2662, "step": 98300 }, { "epoch": 8.51, "learning_rate": 5.3791054701052384e-05, "loss": 1.2353, "step": 98310 }, { "epoch": 8.52, "learning_rate": 5.378888631895455e-05, "loss": 1.2477, "step": 98320 }, { "epoch": 8.52, "learning_rate": 5.37867179368567e-05, "loss": 1.2798, "step": 98330 }, { "epoch": 8.52, "learning_rate": 5.378454955475887e-05, "loss": 1.4054, "step": 98340 }, { "epoch": 8.52, "learning_rate": 5.378238117266103e-05, "loss": 1.0681, "step": 98350 }, { "epoch": 8.52, "learning_rate": 5.3780212790563195e-05, "loss": 1.3888, "step": 98360 }, { "epoch": 8.52, "learning_rate": 5.377804440846536e-05, "loss": 1.3312, "step": 98370 }, { "epoch": 8.52, "learning_rate": 5.377587602636752e-05, "loss": 1.2977, "step": 98380 }, { "epoch": 8.52, "learning_rate": 5.377370764426969e-05, "loss": 1.27, "step": 98390 }, { "epoch": 8.52, "learning_rate": 5.377153926217185e-05, "loss": 1.3423, "step": 98400 }, { "epoch": 8.52, "learning_rate": 5.376937088007401e-05, "loss": 1.2762, "step": 98410 }, { "epoch": 8.52, "learning_rate": 5.3767202497976167e-05, "loss": 1.353, "step": 98420 }, { "epoch": 8.52, "learning_rate": 5.376503411587833e-05, "loss": 1.2907, "step": 98430 }, { "epoch": 8.53, "learning_rate": 5.37628657337805e-05, "loss": 1.3354, "step": 98440 }, { "epoch": 8.53, "learning_rate": 5.376069735168266e-05, "loss": 1.2944, "step": 98450 }, { "epoch": 8.53, "learning_rate": 5.3758528969584825e-05, "loss": 1.2518, "step": 98460 }, { "epoch": 8.53, "learning_rate": 5.3756360587486985e-05, "loss": 1.1431, "step": 98470 }, { "epoch": 8.53, "learning_rate": 5.375419220538915e-05, "loss": 1.3773, "step": 98480 }, { "epoch": 8.53, "learning_rate": 5.3752023823291304e-05, "loss": 1.3109, "step": 98490 }, { "epoch": 8.53, "learning_rate": 5.374985544119347e-05, "loss": 1.4528, "step": 98500 }, { "epoch": 8.53, "learning_rate": 5.374768705909564e-05, "loss": 1.4404, "step": 98510 }, { "epoch": 8.53, "learning_rate": 5.3745518676997797e-05, "loss": 1.2569, "step": 98520 }, { "epoch": 8.53, "learning_rate": 5.374335029489996e-05, "loss": 1.2983, "step": 98530 }, { "epoch": 8.53, "learning_rate": 5.374118191280212e-05, "loss": 1.4516, "step": 98540 }, { "epoch": 8.54, "learning_rate": 5.373901353070429e-05, "loss": 1.2432, "step": 98550 }, { "epoch": 8.54, "learning_rate": 5.373684514860644e-05, "loss": 1.4409, "step": 98560 }, { "epoch": 8.54, "learning_rate": 5.373467676650861e-05, "loss": 1.325, "step": 98570 }, { "epoch": 8.54, "learning_rate": 5.3732508384410775e-05, "loss": 1.1656, "step": 98580 }, { "epoch": 8.54, "learning_rate": 5.3730340002312934e-05, "loss": 1.3677, "step": 98590 }, { "epoch": 8.54, "learning_rate": 5.37281716202151e-05, "loss": 1.3085, "step": 98600 }, { "epoch": 8.54, "learning_rate": 5.372600323811726e-05, "loss": 1.2022, "step": 98610 }, { "epoch": 8.54, "learning_rate": 5.3723834856019427e-05, "loss": 1.389, "step": 98620 }, { "epoch": 8.54, "learning_rate": 5.372166647392159e-05, "loss": 1.3024, "step": 98630 }, { "epoch": 8.54, "learning_rate": 5.3719498091823746e-05, "loss": 1.288, "step": 98640 }, { "epoch": 8.54, "learning_rate": 5.371732970972591e-05, "loss": 1.221, "step": 98650 }, { "epoch": 8.54, "learning_rate": 5.371516132762807e-05, "loss": 1.3144, "step": 98660 }, { "epoch": 8.55, "learning_rate": 5.371299294553024e-05, "loss": 1.0878, "step": 98670 }, { "epoch": 8.55, "learning_rate": 5.37108245634324e-05, "loss": 1.1885, "step": 98680 }, { "epoch": 8.55, "learning_rate": 5.3708656181334564e-05, "loss": 1.2429, "step": 98690 }, { "epoch": 8.55, "learning_rate": 5.370648779923673e-05, "loss": 1.2394, "step": 98700 }, { "epoch": 8.55, "learning_rate": 5.370431941713889e-05, "loss": 1.2609, "step": 98710 }, { "epoch": 8.55, "learning_rate": 5.370215103504105e-05, "loss": 1.1789, "step": 98720 }, { "epoch": 8.55, "learning_rate": 5.369998265294321e-05, "loss": 1.2891, "step": 98730 }, { "epoch": 8.55, "learning_rate": 5.3697814270845376e-05, "loss": 1.3478, "step": 98740 }, { "epoch": 8.55, "learning_rate": 5.3695645888747535e-05, "loss": 1.0674, "step": 98750 }, { "epoch": 8.55, "learning_rate": 5.36934775066497e-05, "loss": 1.5336, "step": 98760 }, { "epoch": 8.55, "learning_rate": 5.369130912455187e-05, "loss": 1.0661, "step": 98770 }, { "epoch": 8.55, "learning_rate": 5.368914074245403e-05, "loss": 1.5635, "step": 98780 }, { "epoch": 8.56, "learning_rate": 5.368697236035618e-05, "loss": 1.1531, "step": 98790 }, { "epoch": 8.56, "learning_rate": 5.368480397825835e-05, "loss": 1.3461, "step": 98800 }, { "epoch": 8.56, "learning_rate": 5.3682635596160514e-05, "loss": 1.1932, "step": 98810 }, { "epoch": 8.56, "learning_rate": 5.368046721406267e-05, "loss": 1.4528, "step": 98820 }, { "epoch": 8.56, "learning_rate": 5.367829883196484e-05, "loss": 1.4809, "step": 98830 }, { "epoch": 8.56, "learning_rate": 5.3676130449867e-05, "loss": 1.5019, "step": 98840 }, { "epoch": 8.56, "learning_rate": 5.3673962067769166e-05, "loss": 1.0609, "step": 98850 }, { "epoch": 8.56, "learning_rate": 5.367179368567133e-05, "loss": 1.2558, "step": 98860 }, { "epoch": 8.56, "learning_rate": 5.3669625303573485e-05, "loss": 1.4636, "step": 98870 }, { "epoch": 8.56, "learning_rate": 5.366745692147565e-05, "loss": 1.4979, "step": 98880 }, { "epoch": 8.56, "learning_rate": 5.366528853937781e-05, "loss": 1.4242, "step": 98890 }, { "epoch": 8.57, "learning_rate": 5.366312015727998e-05, "loss": 1.3169, "step": 98900 }, { "epoch": 8.57, "learning_rate": 5.366095177518214e-05, "loss": 1.2797, "step": 98910 }, { "epoch": 8.57, "learning_rate": 5.36587833930843e-05, "loss": 1.2144, "step": 98920 }, { "epoch": 8.57, "learning_rate": 5.365661501098647e-05, "loss": 1.4108, "step": 98930 }, { "epoch": 8.57, "learning_rate": 5.365444662888863e-05, "loss": 1.2565, "step": 98940 }, { "epoch": 8.57, "learning_rate": 5.365227824679079e-05, "loss": 1.2509, "step": 98950 }, { "epoch": 8.57, "learning_rate": 5.365010986469295e-05, "loss": 1.355, "step": 98960 }, { "epoch": 8.57, "learning_rate": 5.3647941482595115e-05, "loss": 1.2694, "step": 98970 }, { "epoch": 8.57, "learning_rate": 5.3645773100497274e-05, "loss": 1.2842, "step": 98980 }, { "epoch": 8.57, "learning_rate": 5.364360471839944e-05, "loss": 1.1754, "step": 98990 }, { "epoch": 8.57, "learning_rate": 5.364143633630161e-05, "loss": 1.1743, "step": 99000 }, { "epoch": 8.57, "learning_rate": 5.363926795420377e-05, "loss": 1.4213, "step": 99010 }, { "epoch": 8.58, "learning_rate": 5.363709957210593e-05, "loss": 1.2916, "step": 99020 }, { "epoch": 8.58, "learning_rate": 5.3634931190008086e-05, "loss": 1.2489, "step": 99030 }, { "epoch": 8.58, "learning_rate": 5.363276280791025e-05, "loss": 1.302, "step": 99040 }, { "epoch": 8.58, "learning_rate": 5.363059442581241e-05, "loss": 1.2963, "step": 99050 }, { "epoch": 8.58, "learning_rate": 5.362842604371458e-05, "loss": 1.2866, "step": 99060 }, { "epoch": 8.58, "learning_rate": 5.3626257661616745e-05, "loss": 1.3894, "step": 99070 }, { "epoch": 8.58, "learning_rate": 5.3624089279518904e-05, "loss": 1.383, "step": 99080 }, { "epoch": 8.58, "learning_rate": 5.362192089742107e-05, "loss": 1.3746, "step": 99090 }, { "epoch": 8.58, "learning_rate": 5.3619752515323224e-05, "loss": 1.3037, "step": 99100 }, { "epoch": 8.58, "learning_rate": 5.361758413322539e-05, "loss": 1.3838, "step": 99110 }, { "epoch": 8.58, "learning_rate": 5.361541575112755e-05, "loss": 1.2686, "step": 99120 }, { "epoch": 8.59, "learning_rate": 5.3613247369029716e-05, "loss": 1.4418, "step": 99130 }, { "epoch": 8.59, "learning_rate": 5.361107898693188e-05, "loss": 1.1538, "step": 99140 }, { "epoch": 8.59, "learning_rate": 5.360891060483404e-05, "loss": 1.3488, "step": 99150 }, { "epoch": 8.59, "learning_rate": 5.360674222273621e-05, "loss": 1.4033, "step": 99160 }, { "epoch": 8.59, "learning_rate": 5.360457384063837e-05, "loss": 1.4896, "step": 99170 }, { "epoch": 8.59, "learning_rate": 5.360240545854053e-05, "loss": 1.4306, "step": 99180 }, { "epoch": 8.59, "learning_rate": 5.360023707644269e-05, "loss": 1.1845, "step": 99190 }, { "epoch": 8.59, "learning_rate": 5.3598068694344854e-05, "loss": 1.1518, "step": 99200 }, { "epoch": 8.59, "learning_rate": 5.359590031224702e-05, "loss": 1.3095, "step": 99210 }, { "epoch": 8.59, "learning_rate": 5.359373193014918e-05, "loss": 1.3145, "step": 99220 }, { "epoch": 8.59, "learning_rate": 5.3591563548051346e-05, "loss": 1.2716, "step": 99230 }, { "epoch": 8.59, "learning_rate": 5.3589395165953506e-05, "loss": 1.1874, "step": 99240 }, { "epoch": 8.6, "learning_rate": 5.358722678385567e-05, "loss": 1.5338, "step": 99250 }, { "epoch": 8.6, "learning_rate": 5.3585058401757825e-05, "loss": 1.3423, "step": 99260 }, { "epoch": 8.6, "learning_rate": 5.358289001965999e-05, "loss": 1.316, "step": 99270 }, { "epoch": 8.6, "learning_rate": 5.358072163756216e-05, "loss": 1.2474, "step": 99280 }, { "epoch": 8.6, "learning_rate": 5.357855325546432e-05, "loss": 1.2358, "step": 99290 }, { "epoch": 8.6, "learning_rate": 5.3576384873366484e-05, "loss": 0.9655, "step": 99300 }, { "epoch": 8.6, "learning_rate": 5.357421649126864e-05, "loss": 1.3748, "step": 99310 }, { "epoch": 8.6, "learning_rate": 5.357204810917081e-05, "loss": 1.4039, "step": 99320 }, { "epoch": 8.6, "learning_rate": 5.356987972707296e-05, "loss": 1.4315, "step": 99330 }, { "epoch": 8.6, "learning_rate": 5.356771134497513e-05, "loss": 1.3242, "step": 99340 }, { "epoch": 8.6, "learning_rate": 5.356554296287729e-05, "loss": 1.1901, "step": 99350 }, { "epoch": 8.61, "learning_rate": 5.3563374580779455e-05, "loss": 1.2685, "step": 99360 }, { "epoch": 8.61, "learning_rate": 5.356120619868162e-05, "loss": 1.3196, "step": 99370 }, { "epoch": 8.61, "learning_rate": 5.355903781658378e-05, "loss": 1.4082, "step": 99380 }, { "epoch": 8.61, "learning_rate": 5.355686943448595e-05, "loss": 1.3057, "step": 99390 }, { "epoch": 8.61, "learning_rate": 5.355470105238811e-05, "loss": 1.3637, "step": 99400 }, { "epoch": 8.61, "learning_rate": 5.355253267029027e-05, "loss": 1.3015, "step": 99410 }, { "epoch": 8.61, "learning_rate": 5.3550364288192426e-05, "loss": 1.412, "step": 99420 }, { "epoch": 8.61, "learning_rate": 5.354819590609459e-05, "loss": 1.452, "step": 99430 }, { "epoch": 8.61, "learning_rate": 5.354602752399676e-05, "loss": 1.4574, "step": 99440 }, { "epoch": 8.61, "learning_rate": 5.354385914189892e-05, "loss": 1.3657, "step": 99450 }, { "epoch": 8.61, "learning_rate": 5.3541690759801085e-05, "loss": 1.0935, "step": 99460 }, { "epoch": 8.61, "learning_rate": 5.3539522377703245e-05, "loss": 1.4711, "step": 99470 }, { "epoch": 8.62, "learning_rate": 5.353735399560541e-05, "loss": 1.1991, "step": 99480 }, { "epoch": 8.62, "learning_rate": 5.3535185613507564e-05, "loss": 1.4286, "step": 99490 }, { "epoch": 8.62, "learning_rate": 5.353301723140973e-05, "loss": 1.1462, "step": 99500 }, { "epoch": 8.62, "learning_rate": 5.35308488493119e-05, "loss": 1.3355, "step": 99510 }, { "epoch": 8.62, "learning_rate": 5.3528680467214056e-05, "loss": 1.389, "step": 99520 }, { "epoch": 8.62, "learning_rate": 5.352651208511622e-05, "loss": 1.4205, "step": 99530 }, { "epoch": 8.62, "learning_rate": 5.352434370301838e-05, "loss": 1.367, "step": 99540 }, { "epoch": 8.62, "learning_rate": 5.352217532092055e-05, "loss": 1.1255, "step": 99550 }, { "epoch": 8.62, "learning_rate": 5.35200069388227e-05, "loss": 1.3982, "step": 99560 }, { "epoch": 8.62, "learning_rate": 5.351783855672487e-05, "loss": 1.3239, "step": 99570 }, { "epoch": 8.62, "learning_rate": 5.3515670174627034e-05, "loss": 1.3651, "step": 99580 }, { "epoch": 8.63, "learning_rate": 5.3513501792529194e-05, "loss": 1.4245, "step": 99590 }, { "epoch": 8.63, "learning_rate": 5.351133341043136e-05, "loss": 1.2572, "step": 99600 }, { "epoch": 8.63, "learning_rate": 5.350916502833352e-05, "loss": 1.391, "step": 99610 }, { "epoch": 8.63, "learning_rate": 5.3506996646235686e-05, "loss": 1.2289, "step": 99620 }, { "epoch": 8.63, "learning_rate": 5.350482826413785e-05, "loss": 1.2607, "step": 99630 }, { "epoch": 8.63, "learning_rate": 5.3502659882040006e-05, "loss": 1.3119, "step": 99640 }, { "epoch": 8.63, "learning_rate": 5.350049149994217e-05, "loss": 1.3584, "step": 99650 }, { "epoch": 8.63, "learning_rate": 5.349832311784433e-05, "loss": 1.4605, "step": 99660 }, { "epoch": 8.63, "learning_rate": 5.34961547357465e-05, "loss": 1.3085, "step": 99670 }, { "epoch": 8.63, "learning_rate": 5.349398635364866e-05, "loss": 1.1562, "step": 99680 }, { "epoch": 8.63, "learning_rate": 5.3491817971550824e-05, "loss": 1.2076, "step": 99690 }, { "epoch": 8.63, "learning_rate": 5.348964958945299e-05, "loss": 1.214, "step": 99700 }, { "epoch": 8.64, "learning_rate": 5.348748120735515e-05, "loss": 1.2111, "step": 99710 }, { "epoch": 8.64, "learning_rate": 5.348531282525731e-05, "loss": 1.3936, "step": 99720 }, { "epoch": 8.64, "learning_rate": 5.348314444315947e-05, "loss": 1.2467, "step": 99730 }, { "epoch": 8.64, "learning_rate": 5.3480976061061636e-05, "loss": 1.482, "step": 99740 }, { "epoch": 8.64, "learning_rate": 5.3478807678963795e-05, "loss": 1.4354, "step": 99750 }, { "epoch": 8.64, "learning_rate": 5.347663929686596e-05, "loss": 1.2768, "step": 99760 }, { "epoch": 8.64, "learning_rate": 5.347447091476813e-05, "loss": 1.3951, "step": 99770 }, { "epoch": 8.64, "learning_rate": 5.347230253267029e-05, "loss": 1.4166, "step": 99780 }, { "epoch": 8.64, "learning_rate": 5.3470134150572454e-05, "loss": 1.2896, "step": 99790 }, { "epoch": 8.64, "learning_rate": 5.346796576847461e-05, "loss": 1.2882, "step": 99800 }, { "epoch": 8.64, "learning_rate": 5.346579738637677e-05, "loss": 1.3226, "step": 99810 }, { "epoch": 8.65, "learning_rate": 5.346362900427893e-05, "loss": 1.1968, "step": 99820 }, { "epoch": 8.65, "learning_rate": 5.34614606221811e-05, "loss": 1.2606, "step": 99830 }, { "epoch": 8.65, "learning_rate": 5.3459292240083266e-05, "loss": 1.2328, "step": 99840 }, { "epoch": 8.65, "learning_rate": 5.3457123857985425e-05, "loss": 1.3538, "step": 99850 }, { "epoch": 8.65, "learning_rate": 5.345495547588759e-05, "loss": 1.2411, "step": 99860 }, { "epoch": 8.65, "learning_rate": 5.3452787093789744e-05, "loss": 1.2914, "step": 99870 }, { "epoch": 8.65, "learning_rate": 5.345061871169191e-05, "loss": 1.139, "step": 99880 }, { "epoch": 8.65, "learning_rate": 5.344845032959407e-05, "loss": 1.5267, "step": 99890 }, { "epoch": 8.65, "learning_rate": 5.344628194749624e-05, "loss": 1.3019, "step": 99900 }, { "epoch": 8.65, "learning_rate": 5.3444113565398396e-05, "loss": 1.4676, "step": 99910 }, { "epoch": 8.65, "learning_rate": 5.344194518330056e-05, "loss": 1.2205, "step": 99920 }, { "epoch": 8.65, "learning_rate": 5.343977680120273e-05, "loss": 1.3561, "step": 99930 }, { "epoch": 8.66, "learning_rate": 5.343760841910489e-05, "loss": 1.3268, "step": 99940 }, { "epoch": 8.66, "learning_rate": 5.343544003700705e-05, "loss": 1.5166, "step": 99950 }, { "epoch": 8.66, "learning_rate": 5.343327165490921e-05, "loss": 1.3282, "step": 99960 }, { "epoch": 8.66, "learning_rate": 5.3431103272811374e-05, "loss": 1.3506, "step": 99970 }, { "epoch": 8.66, "learning_rate": 5.3428934890713534e-05, "loss": 1.3971, "step": 99980 }, { "epoch": 8.66, "learning_rate": 5.34267665086157e-05, "loss": 1.3676, "step": 99990 }, { "epoch": 8.66, "learning_rate": 5.342459812651787e-05, "loss": 1.0854, "step": 100000 }, { "epoch": 8.66, "learning_rate": 5.3422429744420026e-05, "loss": 1.1793, "step": 100010 }, { "epoch": 8.66, "learning_rate": 5.342026136232219e-05, "loss": 1.2563, "step": 100020 }, { "epoch": 8.66, "learning_rate": 5.3418092980224346e-05, "loss": 1.1833, "step": 100030 }, { "epoch": 8.66, "learning_rate": 5.341592459812651e-05, "loss": 1.421, "step": 100040 }, { "epoch": 8.66, "learning_rate": 5.341375621602867e-05, "loss": 1.3054, "step": 100050 }, { "epoch": 8.67, "learning_rate": 5.341158783393084e-05, "loss": 1.2732, "step": 100060 }, { "epoch": 8.67, "learning_rate": 5.3409419451833005e-05, "loss": 1.2121, "step": 100070 }, { "epoch": 8.67, "learning_rate": 5.3407251069735164e-05, "loss": 1.2284, "step": 100080 }, { "epoch": 8.67, "learning_rate": 5.340508268763733e-05, "loss": 1.411, "step": 100090 }, { "epoch": 8.67, "learning_rate": 5.340291430553948e-05, "loss": 1.3782, "step": 100100 }, { "epoch": 8.67, "learning_rate": 5.340074592344165e-05, "loss": 1.2646, "step": 100110 }, { "epoch": 8.67, "learning_rate": 5.339857754134381e-05, "loss": 1.5378, "step": 100120 }, { "epoch": 8.67, "learning_rate": 5.3396409159245976e-05, "loss": 1.2043, "step": 100130 }, { "epoch": 8.67, "learning_rate": 5.339424077714814e-05, "loss": 1.4128, "step": 100140 }, { "epoch": 8.67, "learning_rate": 5.33920723950503e-05, "loss": 1.1008, "step": 100150 }, { "epoch": 8.67, "learning_rate": 5.338990401295247e-05, "loss": 1.2932, "step": 100160 }, { "epoch": 8.68, "learning_rate": 5.338773563085463e-05, "loss": 1.3456, "step": 100170 }, { "epoch": 8.68, "learning_rate": 5.338556724875679e-05, "loss": 1.3321, "step": 100180 }, { "epoch": 8.68, "learning_rate": 5.338339886665895e-05, "loss": 1.3364, "step": 100190 }, { "epoch": 8.68, "learning_rate": 5.3381230484561113e-05, "loss": 1.2869, "step": 100200 }, { "epoch": 8.68, "learning_rate": 5.337906210246328e-05, "loss": 1.3028, "step": 100210 }, { "epoch": 8.68, "learning_rate": 5.337689372036544e-05, "loss": 1.0199, "step": 100220 }, { "epoch": 8.68, "learning_rate": 5.3374725338267606e-05, "loss": 1.2608, "step": 100230 }, { "epoch": 8.68, "learning_rate": 5.3372556956169765e-05, "loss": 1.2283, "step": 100240 }, { "epoch": 8.68, "learning_rate": 5.337038857407193e-05, "loss": 1.3234, "step": 100250 }, { "epoch": 8.68, "learning_rate": 5.3368220191974085e-05, "loss": 1.518, "step": 100260 }, { "epoch": 8.68, "learning_rate": 5.336605180987625e-05, "loss": 1.4011, "step": 100270 }, { "epoch": 8.68, "learning_rate": 5.336388342777842e-05, "loss": 1.5304, "step": 100280 }, { "epoch": 8.69, "learning_rate": 5.336171504568058e-05, "loss": 1.368, "step": 100290 }, { "epoch": 8.69, "learning_rate": 5.3359546663582743e-05, "loss": 1.1874, "step": 100300 }, { "epoch": 8.69, "learning_rate": 5.33573782814849e-05, "loss": 1.2181, "step": 100310 }, { "epoch": 8.69, "learning_rate": 5.335520989938707e-05, "loss": 1.235, "step": 100320 }, { "epoch": 8.69, "learning_rate": 5.335304151728922e-05, "loss": 1.225, "step": 100330 }, { "epoch": 8.69, "learning_rate": 5.335087313519139e-05, "loss": 1.4746, "step": 100340 }, { "epoch": 8.69, "learning_rate": 5.3348704753093555e-05, "loss": 1.1449, "step": 100350 }, { "epoch": 8.69, "learning_rate": 5.3346536370995715e-05, "loss": 1.3281, "step": 100360 }, { "epoch": 8.69, "learning_rate": 5.334436798889788e-05, "loss": 1.443, "step": 100370 }, { "epoch": 8.69, "learning_rate": 5.334219960680004e-05, "loss": 1.2005, "step": 100380 }, { "epoch": 8.69, "learning_rate": 5.334003122470221e-05, "loss": 1.381, "step": 100390 }, { "epoch": 8.7, "learning_rate": 5.3337862842604373e-05, "loss": 1.1136, "step": 100400 }, { "epoch": 8.7, "learning_rate": 5.3335694460506526e-05, "loss": 1.1946, "step": 100410 }, { "epoch": 8.7, "learning_rate": 5.3333526078408686e-05, "loss": 1.3565, "step": 100420 }, { "epoch": 8.7, "learning_rate": 5.333135769631085e-05, "loss": 1.3523, "step": 100430 }, { "epoch": 8.7, "learning_rate": 5.332918931421302e-05, "loss": 1.3295, "step": 100440 }, { "epoch": 8.7, "learning_rate": 5.332702093211518e-05, "loss": 1.42, "step": 100450 }, { "epoch": 8.7, "learning_rate": 5.3324852550017345e-05, "loss": 1.3767, "step": 100460 }, { "epoch": 8.7, "learning_rate": 5.3322684167919504e-05, "loss": 1.3472, "step": 100470 }, { "epoch": 8.7, "learning_rate": 5.332051578582167e-05, "loss": 1.0375, "step": 100480 }, { "epoch": 8.7, "learning_rate": 5.3318347403723824e-05, "loss": 1.2115, "step": 100490 }, { "epoch": 8.7, "learning_rate": 5.331617902162599e-05, "loss": 1.2338, "step": 100500 }, { "epoch": 8.7, "learning_rate": 5.3314010639528156e-05, "loss": 1.4477, "step": 100510 }, { "epoch": 8.71, "learning_rate": 5.3311842257430316e-05, "loss": 1.3842, "step": 100520 }, { "epoch": 8.71, "learning_rate": 5.330967387533248e-05, "loss": 1.388, "step": 100530 }, { "epoch": 8.71, "learning_rate": 5.330750549323464e-05, "loss": 1.1383, "step": 100540 }, { "epoch": 8.71, "learning_rate": 5.330533711113681e-05, "loss": 1.405, "step": 100550 }, { "epoch": 8.71, "learning_rate": 5.3303168729038975e-05, "loss": 1.0747, "step": 100560 }, { "epoch": 8.71, "learning_rate": 5.330100034694113e-05, "loss": 1.2622, "step": 100570 }, { "epoch": 8.71, "learning_rate": 5.3298831964843294e-05, "loss": 1.3504, "step": 100580 }, { "epoch": 8.71, "learning_rate": 5.3296663582745454e-05, "loss": 1.2606, "step": 100590 }, { "epoch": 8.71, "learning_rate": 5.329449520064762e-05, "loss": 1.3094, "step": 100600 }, { "epoch": 8.71, "learning_rate": 5.329232681854978e-05, "loss": 1.1925, "step": 100610 }, { "epoch": 8.71, "learning_rate": 5.3290158436451946e-05, "loss": 1.3099, "step": 100620 }, { "epoch": 8.72, "learning_rate": 5.328799005435411e-05, "loss": 1.4592, "step": 100630 }, { "epoch": 8.72, "learning_rate": 5.3285821672256265e-05, "loss": 1.4517, "step": 100640 }, { "epoch": 8.72, "learning_rate": 5.328365329015843e-05, "loss": 1.4042, "step": 100650 }, { "epoch": 8.72, "learning_rate": 5.328148490806059e-05, "loss": 1.2603, "step": 100660 }, { "epoch": 8.72, "learning_rate": 5.327931652596276e-05, "loss": 1.3196, "step": 100670 }, { "epoch": 8.72, "learning_rate": 5.327714814386492e-05, "loss": 1.4201, "step": 100680 }, { "epoch": 8.72, "learning_rate": 5.3274979761767084e-05, "loss": 1.3262, "step": 100690 }, { "epoch": 8.72, "learning_rate": 5.327281137966925e-05, "loss": 1.3711, "step": 100700 }, { "epoch": 8.72, "learning_rate": 5.327064299757141e-05, "loss": 1.3374, "step": 100710 }, { "epoch": 8.72, "learning_rate": 5.326847461547357e-05, "loss": 1.0972, "step": 100720 }, { "epoch": 8.72, "learning_rate": 5.326630623337573e-05, "loss": 1.2839, "step": 100730 }, { "epoch": 8.72, "learning_rate": 5.3264137851277895e-05, "loss": 1.3887, "step": 100740 }, { "epoch": 8.73, "learning_rate": 5.3261969469180055e-05, "loss": 1.4114, "step": 100750 }, { "epoch": 8.73, "learning_rate": 5.325980108708222e-05, "loss": 1.3696, "step": 100760 }, { "epoch": 8.73, "learning_rate": 5.325763270498439e-05, "loss": 1.192, "step": 100770 }, { "epoch": 8.73, "learning_rate": 5.325546432288655e-05, "loss": 1.4007, "step": 100780 }, { "epoch": 8.73, "learning_rate": 5.3253295940788714e-05, "loss": 1.4492, "step": 100790 }, { "epoch": 8.73, "learning_rate": 5.3251127558690867e-05, "loss": 1.4473, "step": 100800 }, { "epoch": 8.73, "learning_rate": 5.324895917659303e-05, "loss": 1.2291, "step": 100810 }, { "epoch": 8.73, "learning_rate": 5.324679079449519e-05, "loss": 1.3083, "step": 100820 }, { "epoch": 8.73, "learning_rate": 5.324462241239736e-05, "loss": 1.4831, "step": 100830 }, { "epoch": 8.73, "learning_rate": 5.3242454030299525e-05, "loss": 1.275, "step": 100840 }, { "epoch": 8.73, "learning_rate": 5.3240285648201685e-05, "loss": 1.3377, "step": 100850 }, { "epoch": 8.74, "learning_rate": 5.323811726610385e-05, "loss": 1.3199, "step": 100860 }, { "epoch": 8.74, "learning_rate": 5.3235948884006004e-05, "loss": 1.3294, "step": 100870 }, { "epoch": 8.74, "learning_rate": 5.323378050190817e-05, "loss": 1.3245, "step": 100880 }, { "epoch": 8.74, "learning_rate": 5.323161211981033e-05, "loss": 1.4324, "step": 100890 }, { "epoch": 8.74, "learning_rate": 5.3229443737712497e-05, "loss": 1.3026, "step": 100900 }, { "epoch": 8.74, "learning_rate": 5.322727535561466e-05, "loss": 1.5193, "step": 100910 }, { "epoch": 8.74, "learning_rate": 5.322510697351682e-05, "loss": 1.1839, "step": 100920 }, { "epoch": 8.74, "learning_rate": 5.322293859141899e-05, "loss": 1.2268, "step": 100930 }, { "epoch": 8.74, "learning_rate": 5.322077020932115e-05, "loss": 1.4397, "step": 100940 }, { "epoch": 8.74, "learning_rate": 5.321860182722331e-05, "loss": 1.3387, "step": 100950 }, { "epoch": 8.74, "learning_rate": 5.321643344512547e-05, "loss": 1.3145, "step": 100960 }, { "epoch": 8.74, "learning_rate": 5.3214265063027634e-05, "loss": 1.229, "step": 100970 }, { "epoch": 8.75, "learning_rate": 5.3212096680929794e-05, "loss": 1.2837, "step": 100980 }, { "epoch": 8.75, "learning_rate": 5.320992829883196e-05, "loss": 1.5791, "step": 100990 }, { "epoch": 8.75, "learning_rate": 5.3207759916734127e-05, "loss": 1.2355, "step": 101000 }, { "epoch": 8.75, "learning_rate": 5.3205591534636286e-05, "loss": 1.0848, "step": 101010 }, { "epoch": 8.75, "learning_rate": 5.320342315253845e-05, "loss": 1.249, "step": 101020 }, { "epoch": 8.75, "learning_rate": 5.3201254770440605e-05, "loss": 1.4637, "step": 101030 }, { "epoch": 8.75, "learning_rate": 5.319908638834277e-05, "loss": 1.3079, "step": 101040 }, { "epoch": 8.75, "learning_rate": 5.319691800624493e-05, "loss": 1.2037, "step": 101050 }, { "epoch": 8.75, "learning_rate": 5.31947496241471e-05, "loss": 1.295, "step": 101060 }, { "epoch": 8.75, "learning_rate": 5.3192581242049264e-05, "loss": 1.3478, "step": 101070 }, { "epoch": 8.75, "learning_rate": 5.3190412859951424e-05, "loss": 1.3062, "step": 101080 }, { "epoch": 8.76, "learning_rate": 5.318824447785359e-05, "loss": 1.3308, "step": 101090 }, { "epoch": 8.76, "learning_rate": 5.318607609575574e-05, "loss": 1.1904, "step": 101100 }, { "epoch": 8.76, "learning_rate": 5.318390771365791e-05, "loss": 1.4332, "step": 101110 }, { "epoch": 8.76, "learning_rate": 5.318173933156007e-05, "loss": 1.4162, "step": 101120 }, { "epoch": 8.76, "learning_rate": 5.3179570949462235e-05, "loss": 1.416, "step": 101130 }, { "epoch": 8.76, "learning_rate": 5.31774025673644e-05, "loss": 1.1181, "step": 101140 }, { "epoch": 8.76, "learning_rate": 5.317523418526656e-05, "loss": 1.2637, "step": 101150 }, { "epoch": 8.76, "learning_rate": 5.317306580316873e-05, "loss": 1.1046, "step": 101160 }, { "epoch": 8.76, "learning_rate": 5.317089742107089e-05, "loss": 1.5374, "step": 101170 }, { "epoch": 8.76, "learning_rate": 5.316872903897305e-05, "loss": 1.412, "step": 101180 }, { "epoch": 8.76, "learning_rate": 5.316656065687521e-05, "loss": 1.3055, "step": 101190 }, { "epoch": 8.76, "learning_rate": 5.316439227477737e-05, "loss": 1.1522, "step": 101200 }, { "epoch": 8.77, "learning_rate": 5.316222389267954e-05, "loss": 1.2819, "step": 101210 }, { "epoch": 8.77, "learning_rate": 5.31600555105817e-05, "loss": 1.3207, "step": 101220 }, { "epoch": 8.77, "learning_rate": 5.3157887128483865e-05, "loss": 1.2826, "step": 101230 }, { "epoch": 8.77, "learning_rate": 5.3155718746386025e-05, "loss": 1.4782, "step": 101240 }, { "epoch": 8.77, "learning_rate": 5.315355036428819e-05, "loss": 1.195, "step": 101250 }, { "epoch": 8.77, "learning_rate": 5.3151381982190344e-05, "loss": 1.254, "step": 101260 }, { "epoch": 8.77, "learning_rate": 5.314921360009251e-05, "loss": 1.2227, "step": 101270 }, { "epoch": 8.77, "learning_rate": 5.314704521799468e-05, "loss": 1.3205, "step": 101280 }, { "epoch": 8.77, "learning_rate": 5.314487683589684e-05, "loss": 1.2078, "step": 101290 }, { "epoch": 8.77, "learning_rate": 5.3142708453799e-05, "loss": 1.5302, "step": 101300 }, { "epoch": 8.77, "learning_rate": 5.314054007170116e-05, "loss": 1.1929, "step": 101310 }, { "epoch": 8.77, "learning_rate": 5.313837168960333e-05, "loss": 1.133, "step": 101320 }, { "epoch": 8.78, "learning_rate": 5.3136203307505496e-05, "loss": 1.2394, "step": 101330 }, { "epoch": 8.78, "learning_rate": 5.313403492540765e-05, "loss": 1.4182, "step": 101340 }, { "epoch": 8.78, "learning_rate": 5.3131866543309815e-05, "loss": 1.149, "step": 101350 }, { "epoch": 8.78, "learning_rate": 5.3129698161211974e-05, "loss": 1.4144, "step": 101360 }, { "epoch": 8.78, "learning_rate": 5.312752977911414e-05, "loss": 1.2499, "step": 101370 }, { "epoch": 8.78, "learning_rate": 5.31253613970163e-05, "loss": 1.2463, "step": 101380 }, { "epoch": 8.78, "learning_rate": 5.312319301491847e-05, "loss": 1.3237, "step": 101390 }, { "epoch": 8.78, "learning_rate": 5.312102463282063e-05, "loss": 1.4907, "step": 101400 }, { "epoch": 8.78, "learning_rate": 5.3118856250722786e-05, "loss": 1.3249, "step": 101410 }, { "epoch": 8.78, "learning_rate": 5.311668786862495e-05, "loss": 0.9836, "step": 101420 }, { "epoch": 8.78, "learning_rate": 5.311451948652711e-05, "loss": 1.5449, "step": 101430 }, { "epoch": 8.79, "learning_rate": 5.311235110442928e-05, "loss": 1.3952, "step": 101440 }, { "epoch": 8.79, "learning_rate": 5.311018272233144e-05, "loss": 1.2024, "step": 101450 }, { "epoch": 8.79, "learning_rate": 5.3108014340233604e-05, "loss": 1.2854, "step": 101460 }, { "epoch": 8.79, "learning_rate": 5.310584595813577e-05, "loss": 1.2885, "step": 101470 }, { "epoch": 8.79, "learning_rate": 5.310367757603793e-05, "loss": 1.2626, "step": 101480 }, { "epoch": 8.79, "learning_rate": 5.310150919394008e-05, "loss": 1.4464, "step": 101490 }, { "epoch": 8.79, "learning_rate": 5.309934081184225e-05, "loss": 1.188, "step": 101500 }, { "epoch": 8.79, "learning_rate": 5.3097172429744416e-05, "loss": 1.2376, "step": 101510 }, { "epoch": 8.79, "learning_rate": 5.3095004047646576e-05, "loss": 1.3723, "step": 101520 }, { "epoch": 8.79, "learning_rate": 5.309283566554874e-05, "loss": 1.1938, "step": 101530 }, { "epoch": 8.79, "learning_rate": 5.30906672834509e-05, "loss": 1.2776, "step": 101540 }, { "epoch": 8.79, "learning_rate": 5.308849890135307e-05, "loss": 1.155, "step": 101550 }, { "epoch": 8.8, "learning_rate": 5.3086330519255234e-05, "loss": 1.476, "step": 101560 }, { "epoch": 8.8, "learning_rate": 5.308416213715739e-05, "loss": 1.2022, "step": 101570 }, { "epoch": 8.8, "learning_rate": 5.3081993755059554e-05, "loss": 1.2562, "step": 101580 }, { "epoch": 8.8, "learning_rate": 5.307982537296171e-05, "loss": 1.2163, "step": 101590 }, { "epoch": 8.8, "learning_rate": 5.307765699086388e-05, "loss": 1.3218, "step": 101600 }, { "epoch": 8.8, "learning_rate": 5.307548860876604e-05, "loss": 1.258, "step": 101610 }, { "epoch": 8.8, "learning_rate": 5.3073320226668206e-05, "loss": 1.254, "step": 101620 }, { "epoch": 8.8, "learning_rate": 5.307115184457037e-05, "loss": 1.3699, "step": 101630 }, { "epoch": 8.8, "learning_rate": 5.3068983462472525e-05, "loss": 1.2522, "step": 101640 }, { "epoch": 8.8, "learning_rate": 5.306681508037469e-05, "loss": 1.373, "step": 101650 }, { "epoch": 8.8, "learning_rate": 5.306464669827685e-05, "loss": 1.1879, "step": 101660 }, { "epoch": 8.81, "learning_rate": 5.306247831617902e-05, "loss": 1.3415, "step": 101670 }, { "epoch": 8.81, "learning_rate": 5.306030993408118e-05, "loss": 1.4145, "step": 101680 }, { "epoch": 8.81, "learning_rate": 5.305814155198334e-05, "loss": 1.3224, "step": 101690 }, { "epoch": 8.81, "learning_rate": 5.305597316988551e-05, "loss": 1.0888, "step": 101700 }, { "epoch": 8.81, "learning_rate": 5.305380478778767e-05, "loss": 1.3403, "step": 101710 }, { "epoch": 8.81, "learning_rate": 5.305163640568983e-05, "loss": 1.3782, "step": 101720 }, { "epoch": 8.81, "learning_rate": 5.304946802359199e-05, "loss": 1.439, "step": 101730 }, { "epoch": 8.81, "learning_rate": 5.3047299641494155e-05, "loss": 1.564, "step": 101740 }, { "epoch": 8.81, "learning_rate": 5.3045131259396315e-05, "loss": 1.5423, "step": 101750 }, { "epoch": 8.81, "learning_rate": 5.304296287729848e-05, "loss": 1.3395, "step": 101760 }, { "epoch": 8.81, "learning_rate": 5.304079449520065e-05, "loss": 1.2656, "step": 101770 }, { "epoch": 8.81, "learning_rate": 5.303862611310281e-05, "loss": 1.3248, "step": 101780 }, { "epoch": 8.82, "learning_rate": 5.303645773100497e-05, "loss": 1.3845, "step": 101790 }, { "epoch": 8.82, "learning_rate": 5.3034289348907126e-05, "loss": 1.4355, "step": 101800 }, { "epoch": 8.82, "learning_rate": 5.303212096680929e-05, "loss": 1.5117, "step": 101810 }, { "epoch": 8.82, "learning_rate": 5.302995258471145e-05, "loss": 1.4631, "step": 101820 }, { "epoch": 8.82, "learning_rate": 5.302778420261362e-05, "loss": 1.3484, "step": 101830 }, { "epoch": 8.82, "learning_rate": 5.3025615820515785e-05, "loss": 1.3842, "step": 101840 }, { "epoch": 8.82, "learning_rate": 5.3023447438417945e-05, "loss": 1.1759, "step": 101850 }, { "epoch": 8.82, "learning_rate": 5.302127905632011e-05, "loss": 1.2531, "step": 101860 }, { "epoch": 8.82, "learning_rate": 5.3019110674222264e-05, "loss": 1.2003, "step": 101870 }, { "epoch": 8.82, "learning_rate": 5.301694229212443e-05, "loss": 1.2472, "step": 101880 }, { "epoch": 8.82, "learning_rate": 5.301477391002659e-05, "loss": 1.3737, "step": 101890 }, { "epoch": 8.83, "learning_rate": 5.3012605527928756e-05, "loss": 1.5307, "step": 101900 }, { "epoch": 8.83, "learning_rate": 5.301043714583092e-05, "loss": 1.5392, "step": 101910 }, { "epoch": 8.83, "learning_rate": 5.300826876373308e-05, "loss": 1.4884, "step": 101920 }, { "epoch": 8.83, "learning_rate": 5.300610038163525e-05, "loss": 1.4559, "step": 101930 }, { "epoch": 8.83, "learning_rate": 5.300393199953741e-05, "loss": 1.3793, "step": 101940 }, { "epoch": 8.83, "learning_rate": 5.300176361743957e-05, "loss": 1.2468, "step": 101950 }, { "epoch": 8.83, "learning_rate": 5.299959523534173e-05, "loss": 1.151, "step": 101960 }, { "epoch": 8.83, "learning_rate": 5.2997426853243894e-05, "loss": 1.199, "step": 101970 }, { "epoch": 8.83, "learning_rate": 5.299525847114606e-05, "loss": 1.3649, "step": 101980 }, { "epoch": 8.83, "learning_rate": 5.299309008904822e-05, "loss": 1.2215, "step": 101990 }, { "epoch": 8.83, "learning_rate": 5.2990921706950386e-05, "loss": 1.3357, "step": 102000 }, { "epoch": 8.83, "learning_rate": 5.2988753324852546e-05, "loss": 1.2775, "step": 102010 }, { "epoch": 8.84, "learning_rate": 5.298658494275471e-05, "loss": 1.4552, "step": 102020 }, { "epoch": 8.84, "learning_rate": 5.2984416560656865e-05, "loss": 1.3526, "step": 102030 }, { "epoch": 8.84, "learning_rate": 5.298224817855903e-05, "loss": 1.3543, "step": 102040 }, { "epoch": 8.84, "learning_rate": 5.298007979646119e-05, "loss": 1.1218, "step": 102050 }, { "epoch": 8.84, "learning_rate": 5.297791141436336e-05, "loss": 1.3782, "step": 102060 }, { "epoch": 8.84, "learning_rate": 5.2975743032265524e-05, "loss": 1.3267, "step": 102070 }, { "epoch": 8.84, "learning_rate": 5.2973574650167684e-05, "loss": 1.4232, "step": 102080 }, { "epoch": 8.84, "learning_rate": 5.297140626806985e-05, "loss": 1.2679, "step": 102090 }, { "epoch": 8.84, "learning_rate": 5.296923788597201e-05, "loss": 1.2456, "step": 102100 }, { "epoch": 8.84, "learning_rate": 5.296706950387417e-05, "loss": 1.4041, "step": 102110 }, { "epoch": 8.84, "learning_rate": 5.296490112177633e-05, "loss": 1.3404, "step": 102120 }, { "epoch": 8.85, "learning_rate": 5.2962732739678495e-05, "loss": 1.3186, "step": 102130 }, { "epoch": 8.85, "learning_rate": 5.296056435758066e-05, "loss": 1.2691, "step": 102140 }, { "epoch": 8.85, "learning_rate": 5.295839597548282e-05, "loss": 1.3671, "step": 102150 }, { "epoch": 8.85, "learning_rate": 5.295622759338499e-05, "loss": 1.4885, "step": 102160 }, { "epoch": 8.85, "learning_rate": 5.295405921128715e-05, "loss": 1.3873, "step": 102170 }, { "epoch": 8.85, "learning_rate": 5.295189082918931e-05, "loss": 1.6036, "step": 102180 }, { "epoch": 8.85, "learning_rate": 5.2949722447091466e-05, "loss": 1.4972, "step": 102190 }, { "epoch": 8.85, "learning_rate": 5.294755406499363e-05, "loss": 1.2434, "step": 102200 }, { "epoch": 8.85, "learning_rate": 5.29453856828958e-05, "loss": 1.3326, "step": 102210 }, { "epoch": 8.85, "learning_rate": 5.294321730079796e-05, "loss": 1.4346, "step": 102220 }, { "epoch": 8.85, "learning_rate": 5.2941048918700125e-05, "loss": 1.4454, "step": 102230 }, { "epoch": 8.85, "learning_rate": 5.2938880536602285e-05, "loss": 1.3484, "step": 102240 }, { "epoch": 8.86, "learning_rate": 5.293671215450445e-05, "loss": 1.2618, "step": 102250 }, { "epoch": 8.86, "learning_rate": 5.2934543772406604e-05, "loss": 1.1833, "step": 102260 }, { "epoch": 8.86, "learning_rate": 5.293237539030877e-05, "loss": 1.4267, "step": 102270 }, { "epoch": 8.86, "learning_rate": 5.293020700821094e-05, "loss": 1.2648, "step": 102280 }, { "epoch": 8.86, "learning_rate": 5.2928038626113096e-05, "loss": 1.2754, "step": 102290 }, { "epoch": 8.86, "learning_rate": 5.292587024401526e-05, "loss": 1.5664, "step": 102300 }, { "epoch": 8.86, "learning_rate": 5.292370186191742e-05, "loss": 1.3296, "step": 102310 }, { "epoch": 8.86, "learning_rate": 5.292153347981959e-05, "loss": 1.1145, "step": 102320 }, { "epoch": 8.86, "learning_rate": 5.2919365097721755e-05, "loss": 1.3371, "step": 102330 }, { "epoch": 8.86, "learning_rate": 5.291719671562391e-05, "loss": 1.4105, "step": 102340 }, { "epoch": 8.86, "learning_rate": 5.2915028333526074e-05, "loss": 1.3373, "step": 102350 }, { "epoch": 8.87, "learning_rate": 5.2912859951428234e-05, "loss": 1.1379, "step": 102360 }, { "epoch": 8.87, "learning_rate": 5.29106915693304e-05, "loss": 1.0492, "step": 102370 }, { "epoch": 8.87, "learning_rate": 5.290852318723256e-05, "loss": 1.2761, "step": 102380 }, { "epoch": 8.87, "learning_rate": 5.2906354805134726e-05, "loss": 1.0517, "step": 102390 }, { "epoch": 8.87, "learning_rate": 5.290418642303689e-05, "loss": 1.1353, "step": 102400 }, { "epoch": 8.87, "learning_rate": 5.2902018040939046e-05, "loss": 1.4698, "step": 102410 }, { "epoch": 8.87, "learning_rate": 5.289984965884121e-05, "loss": 1.3432, "step": 102420 }, { "epoch": 8.87, "learning_rate": 5.289768127674337e-05, "loss": 1.2878, "step": 102430 }, { "epoch": 8.87, "learning_rate": 5.289551289464554e-05, "loss": 1.4237, "step": 102440 }, { "epoch": 8.87, "learning_rate": 5.28933445125477e-05, "loss": 1.2747, "step": 102450 }, { "epoch": 8.87, "learning_rate": 5.2891176130449864e-05, "loss": 1.5223, "step": 102460 }, { "epoch": 8.87, "learning_rate": 5.288900774835203e-05, "loss": 1.2957, "step": 102470 }, { "epoch": 8.88, "learning_rate": 5.288683936625419e-05, "loss": 1.3557, "step": 102480 }, { "epoch": 8.88, "learning_rate": 5.288467098415635e-05, "loss": 1.5068, "step": 102490 }, { "epoch": 8.88, "learning_rate": 5.288250260205851e-05, "loss": 1.4414, "step": 102500 }, { "epoch": 8.88, "learning_rate": 5.2880334219960676e-05, "loss": 1.3649, "step": 102510 }, { "epoch": 8.88, "learning_rate": 5.2878165837862835e-05, "loss": 1.1452, "step": 102520 }, { "epoch": 8.88, "learning_rate": 5.2875997455765e-05, "loss": 1.3271, "step": 102530 }, { "epoch": 8.88, "learning_rate": 5.287382907366717e-05, "loss": 1.2321, "step": 102540 }, { "epoch": 8.88, "learning_rate": 5.287166069156933e-05, "loss": 1.2138, "step": 102550 }, { "epoch": 8.88, "learning_rate": 5.2869492309471494e-05, "loss": 1.4377, "step": 102560 }, { "epoch": 8.88, "learning_rate": 5.286732392737365e-05, "loss": 1.3793, "step": 102570 }, { "epoch": 8.88, "learning_rate": 5.286515554527581e-05, "loss": 1.3554, "step": 102580 }, { "epoch": 8.88, "learning_rate": 5.286298716317797e-05, "loss": 1.3102, "step": 102590 }, { "epoch": 8.89, "learning_rate": 5.286081878108014e-05, "loss": 1.3086, "step": 102600 }, { "epoch": 8.89, "learning_rate": 5.28586503989823e-05, "loss": 1.4105, "step": 102610 }, { "epoch": 8.89, "learning_rate": 5.2856482016884465e-05, "loss": 1.3205, "step": 102620 }, { "epoch": 8.89, "learning_rate": 5.285431363478663e-05, "loss": 1.3356, "step": 102630 }, { "epoch": 8.89, "learning_rate": 5.2852145252688785e-05, "loss": 1.3774, "step": 102640 }, { "epoch": 8.89, "learning_rate": 5.284997687059095e-05, "loss": 1.1421, "step": 102650 }, { "epoch": 8.89, "learning_rate": 5.284780848849311e-05, "loss": 1.3096, "step": 102660 }, { "epoch": 8.89, "learning_rate": 5.284564010639528e-05, "loss": 1.3401, "step": 102670 }, { "epoch": 8.89, "learning_rate": 5.284347172429744e-05, "loss": 1.4529, "step": 102680 }, { "epoch": 8.89, "learning_rate": 5.28413033421996e-05, "loss": 1.2854, "step": 102690 }, { "epoch": 8.89, "learning_rate": 5.283913496010177e-05, "loss": 1.2221, "step": 102700 }, { "epoch": 8.9, "learning_rate": 5.283696657800393e-05, "loss": 1.4264, "step": 102710 }, { "epoch": 8.9, "learning_rate": 5.283479819590609e-05, "loss": 1.3631, "step": 102720 }, { "epoch": 8.9, "learning_rate": 5.283262981380825e-05, "loss": 1.5332, "step": 102730 }, { "epoch": 8.9, "learning_rate": 5.2830461431710415e-05, "loss": 1.2353, "step": 102740 }, { "epoch": 8.9, "learning_rate": 5.2828293049612574e-05, "loss": 1.6047, "step": 102750 }, { "epoch": 8.9, "learning_rate": 5.282612466751474e-05, "loss": 1.3294, "step": 102760 }, { "epoch": 8.9, "learning_rate": 5.282395628541691e-05, "loss": 1.3349, "step": 102770 }, { "epoch": 8.9, "learning_rate": 5.282178790331907e-05, "loss": 1.1984, "step": 102780 }, { "epoch": 8.9, "learning_rate": 5.281961952122123e-05, "loss": 1.4385, "step": 102790 }, { "epoch": 8.9, "learning_rate": 5.2817451139123386e-05, "loss": 1.5765, "step": 102800 }, { "epoch": 8.9, "learning_rate": 5.281528275702555e-05, "loss": 1.2633, "step": 102810 }, { "epoch": 8.9, "learning_rate": 5.281311437492771e-05, "loss": 1.4189, "step": 102820 }, { "epoch": 8.91, "learning_rate": 5.281094599282988e-05, "loss": 1.372, "step": 102830 }, { "epoch": 8.91, "learning_rate": 5.2808777610732045e-05, "loss": 1.4253, "step": 102840 }, { "epoch": 8.91, "learning_rate": 5.2806609228634204e-05, "loss": 1.4172, "step": 102850 }, { "epoch": 8.91, "learning_rate": 5.280444084653637e-05, "loss": 1.1166, "step": 102860 }, { "epoch": 8.91, "learning_rate": 5.280227246443853e-05, "loss": 1.2824, "step": 102870 }, { "epoch": 8.91, "learning_rate": 5.280010408234069e-05, "loss": 1.2356, "step": 102880 }, { "epoch": 8.91, "learning_rate": 5.279793570024285e-05, "loss": 1.4546, "step": 102890 }, { "epoch": 8.91, "learning_rate": 5.2795767318145016e-05, "loss": 1.2377, "step": 102900 }, { "epoch": 8.91, "learning_rate": 5.279359893604718e-05, "loss": 1.4584, "step": 102910 }, { "epoch": 8.91, "learning_rate": 5.279143055394934e-05, "loss": 1.1802, "step": 102920 }, { "epoch": 8.91, "learning_rate": 5.278926217185151e-05, "loss": 1.2395, "step": 102930 }, { "epoch": 8.92, "learning_rate": 5.278709378975367e-05, "loss": 1.4354, "step": 102940 }, { "epoch": 8.92, "learning_rate": 5.278492540765583e-05, "loss": 1.38, "step": 102950 }, { "epoch": 8.92, "learning_rate": 5.278275702555799e-05, "loss": 1.3705, "step": 102960 }, { "epoch": 8.92, "learning_rate": 5.2780588643460154e-05, "loss": 1.5211, "step": 102970 }, { "epoch": 8.92, "learning_rate": 5.277842026136232e-05, "loss": 1.2086, "step": 102980 }, { "epoch": 8.92, "learning_rate": 5.277625187926448e-05, "loss": 1.5137, "step": 102990 }, { "epoch": 8.92, "learning_rate": 5.2774083497166646e-05, "loss": 1.2765, "step": 103000 }, { "epoch": 8.92, "learning_rate": 5.2771915115068806e-05, "loss": 1.3592, "step": 103010 }, { "epoch": 8.92, "learning_rate": 5.276974673297097e-05, "loss": 1.3858, "step": 103020 }, { "epoch": 8.92, "learning_rate": 5.2767578350873125e-05, "loss": 1.5783, "step": 103030 }, { "epoch": 8.92, "learning_rate": 5.276540996877529e-05, "loss": 1.2893, "step": 103040 }, { "epoch": 8.92, "learning_rate": 5.276324158667746e-05, "loss": 1.2565, "step": 103050 }, { "epoch": 8.93, "learning_rate": 5.276107320457962e-05, "loss": 1.3978, "step": 103060 }, { "epoch": 8.93, "learning_rate": 5.2758904822481784e-05, "loss": 1.3174, "step": 103070 }, { "epoch": 8.93, "learning_rate": 5.275673644038394e-05, "loss": 1.2731, "step": 103080 }, { "epoch": 8.93, "learning_rate": 5.275456805828611e-05, "loss": 1.3314, "step": 103090 }, { "epoch": 8.93, "learning_rate": 5.2752399676188276e-05, "loss": 1.4665, "step": 103100 }, { "epoch": 8.93, "learning_rate": 5.275023129409043e-05, "loss": 1.4165, "step": 103110 }, { "epoch": 8.93, "learning_rate": 5.274806291199259e-05, "loss": 1.4316, "step": 103120 }, { "epoch": 8.93, "learning_rate": 5.2745894529894755e-05, "loss": 1.2536, "step": 103130 }, { "epoch": 8.93, "learning_rate": 5.274372614779692e-05, "loss": 1.4301, "step": 103140 }, { "epoch": 8.93, "learning_rate": 5.274155776569908e-05, "loss": 1.3987, "step": 103150 }, { "epoch": 8.93, "learning_rate": 5.273938938360125e-05, "loss": 1.4084, "step": 103160 }, { "epoch": 8.94, "learning_rate": 5.273722100150341e-05, "loss": 1.5512, "step": 103170 }, { "epoch": 8.94, "learning_rate": 5.2735052619405566e-05, "loss": 1.493, "step": 103180 }, { "epoch": 8.94, "learning_rate": 5.2732884237307726e-05, "loss": 1.4163, "step": 103190 }, { "epoch": 8.94, "learning_rate": 5.273071585520989e-05, "loss": 1.4344, "step": 103200 }, { "epoch": 8.94, "learning_rate": 5.272854747311206e-05, "loss": 1.2579, "step": 103210 }, { "epoch": 8.94, "learning_rate": 5.272637909101422e-05, "loss": 1.4203, "step": 103220 }, { "epoch": 8.94, "learning_rate": 5.2724210708916385e-05, "loss": 1.5373, "step": 103230 }, { "epoch": 8.94, "learning_rate": 5.2722042326818544e-05, "loss": 1.3319, "step": 103240 }, { "epoch": 8.94, "learning_rate": 5.271987394472071e-05, "loss": 1.2586, "step": 103250 }, { "epoch": 8.94, "learning_rate": 5.2717705562622864e-05, "loss": 1.3591, "step": 103260 }, { "epoch": 8.94, "learning_rate": 5.271553718052503e-05, "loss": 1.2016, "step": 103270 }, { "epoch": 8.94, "learning_rate": 5.2713368798427197e-05, "loss": 1.2042, "step": 103280 }, { "epoch": 8.95, "learning_rate": 5.2711200416329356e-05, "loss": 1.3914, "step": 103290 }, { "epoch": 8.95, "learning_rate": 5.270903203423152e-05, "loss": 1.3965, "step": 103300 }, { "epoch": 8.95, "learning_rate": 5.270686365213368e-05, "loss": 1.3209, "step": 103310 }, { "epoch": 8.95, "learning_rate": 5.270469527003585e-05, "loss": 1.2096, "step": 103320 }, { "epoch": 8.95, "learning_rate": 5.2702526887938015e-05, "loss": 1.1889, "step": 103330 }, { "epoch": 8.95, "learning_rate": 5.270035850584017e-05, "loss": 1.3121, "step": 103340 }, { "epoch": 8.95, "learning_rate": 5.2698190123742334e-05, "loss": 1.3588, "step": 103350 }, { "epoch": 8.95, "learning_rate": 5.2696021741644494e-05, "loss": 1.5009, "step": 103360 }, { "epoch": 8.95, "learning_rate": 5.269385335954666e-05, "loss": 1.237, "step": 103370 }, { "epoch": 8.95, "learning_rate": 5.269168497744882e-05, "loss": 1.2118, "step": 103380 }, { "epoch": 8.95, "learning_rate": 5.2689516595350986e-05, "loss": 1.4489, "step": 103390 }, { "epoch": 8.96, "learning_rate": 5.268734821325315e-05, "loss": 1.2764, "step": 103400 }, { "epoch": 8.96, "learning_rate": 5.2685179831155305e-05, "loss": 1.3652, "step": 103410 }, { "epoch": 8.96, "learning_rate": 5.268301144905747e-05, "loss": 1.2122, "step": 103420 }, { "epoch": 8.96, "learning_rate": 5.268084306695963e-05, "loss": 1.3505, "step": 103430 }, { "epoch": 8.96, "learning_rate": 5.26786746848618e-05, "loss": 1.3097, "step": 103440 }, { "epoch": 8.96, "learning_rate": 5.267650630276396e-05, "loss": 1.1628, "step": 103450 }, { "epoch": 8.96, "learning_rate": 5.2674337920666124e-05, "loss": 1.4096, "step": 103460 }, { "epoch": 8.96, "learning_rate": 5.267216953856829e-05, "loss": 1.2914, "step": 103470 }, { "epoch": 8.96, "learning_rate": 5.267000115647045e-05, "loss": 1.6053, "step": 103480 }, { "epoch": 8.96, "learning_rate": 5.266783277437261e-05, "loss": 1.3135, "step": 103490 }, { "epoch": 8.96, "learning_rate": 5.266566439227477e-05, "loss": 1.4274, "step": 103500 }, { "epoch": 8.96, "learning_rate": 5.2663496010176935e-05, "loss": 1.4162, "step": 103510 }, { "epoch": 8.97, "learning_rate": 5.2661327628079095e-05, "loss": 1.452, "step": 103520 }, { "epoch": 8.97, "learning_rate": 5.265915924598126e-05, "loss": 1.2912, "step": 103530 }, { "epoch": 8.97, "learning_rate": 5.265699086388343e-05, "loss": 1.3445, "step": 103540 }, { "epoch": 8.97, "learning_rate": 5.265482248178559e-05, "loss": 1.4945, "step": 103550 }, { "epoch": 8.97, "learning_rate": 5.2652654099687754e-05, "loss": 1.2697, "step": 103560 }, { "epoch": 8.97, "learning_rate": 5.265048571758991e-05, "loss": 1.3375, "step": 103570 }, { "epoch": 8.97, "learning_rate": 5.264831733549207e-05, "loss": 1.3371, "step": 103580 }, { "epoch": 8.97, "learning_rate": 5.264614895339423e-05, "loss": 1.4047, "step": 103590 }, { "epoch": 8.97, "learning_rate": 5.26439805712964e-05, "loss": 1.2152, "step": 103600 }, { "epoch": 8.97, "learning_rate": 5.2641812189198565e-05, "loss": 1.4944, "step": 103610 }, { "epoch": 8.97, "learning_rate": 5.2639643807100725e-05, "loss": 1.3416, "step": 103620 }, { "epoch": 8.98, "learning_rate": 5.263747542500289e-05, "loss": 1.3371, "step": 103630 }, { "epoch": 8.98, "learning_rate": 5.263530704290505e-05, "loss": 1.3362, "step": 103640 }, { "epoch": 8.98, "learning_rate": 5.263313866080721e-05, "loss": 1.5404, "step": 103650 }, { "epoch": 8.98, "learning_rate": 5.263097027870937e-05, "loss": 1.44, "step": 103660 }, { "epoch": 8.98, "learning_rate": 5.262880189661154e-05, "loss": 1.2637, "step": 103670 }, { "epoch": 8.98, "learning_rate": 5.2626633514513696e-05, "loss": 1.2854, "step": 103680 }, { "epoch": 8.98, "learning_rate": 5.262446513241586e-05, "loss": 1.3255, "step": 103690 }, { "epoch": 8.98, "learning_rate": 5.262229675031803e-05, "loss": 1.2177, "step": 103700 }, { "epoch": 8.98, "learning_rate": 5.262012836822019e-05, "loss": 1.2833, "step": 103710 }, { "epoch": 8.98, "learning_rate": 5.261795998612235e-05, "loss": 1.3153, "step": 103720 }, { "epoch": 8.98, "learning_rate": 5.261579160402451e-05, "loss": 1.4582, "step": 103730 }, { "epoch": 8.98, "learning_rate": 5.2613623221926674e-05, "loss": 1.3384, "step": 103740 }, { "epoch": 8.99, "learning_rate": 5.2611454839828834e-05, "loss": 1.286, "step": 103750 }, { "epoch": 8.99, "learning_rate": 5.2609286457731e-05, "loss": 1.3146, "step": 103760 }, { "epoch": 8.99, "learning_rate": 5.260711807563317e-05, "loss": 1.2683, "step": 103770 }, { "epoch": 8.99, "learning_rate": 5.2604949693535326e-05, "loss": 1.4724, "step": 103780 }, { "epoch": 8.99, "learning_rate": 5.260278131143749e-05, "loss": 1.317, "step": 103790 }, { "epoch": 8.99, "learning_rate": 5.2600612929339646e-05, "loss": 1.4495, "step": 103800 }, { "epoch": 8.99, "learning_rate": 5.259844454724181e-05, "loss": 1.4318, "step": 103810 }, { "epoch": 8.99, "learning_rate": 5.259627616514397e-05, "loss": 1.3738, "step": 103820 }, { "epoch": 8.99, "learning_rate": 5.259410778304614e-05, "loss": 1.4146, "step": 103830 }, { "epoch": 8.99, "learning_rate": 5.2591939400948304e-05, "loss": 1.2737, "step": 103840 }, { "epoch": 8.99, "learning_rate": 5.2589771018850464e-05, "loss": 1.44, "step": 103850 }, { "epoch": 8.99, "learning_rate": 5.258760263675263e-05, "loss": 1.3888, "step": 103860 }, { "epoch": 9.0, "learning_rate": 5.258543425465479e-05, "loss": 1.4713, "step": 103870 }, { "epoch": 9.0, "learning_rate": 5.258326587255695e-05, "loss": 1.3562, "step": 103880 }, { "epoch": 9.0, "learning_rate": 5.258109749045911e-05, "loss": 1.3434, "step": 103890 }, { "epoch": 9.0, "learning_rate": 5.2578929108361276e-05, "loss": 1.3832, "step": 103900 }, { "epoch": 9.0, "learning_rate": 5.257676072626344e-05, "loss": 1.2227, "step": 103910 }, { "epoch": 9.0, "eval_cer": 0.5401004246322542, "eval_em": 0.02784940691077875, "eval_f1": 0.02784940691077875, "eval_loss": 1.5033310651779175, "eval_runtime": 254.431, "eval_samples_per_second": 7.621, "eval_steps_per_second": 0.955, "eval_wer": 0.9721505930892212, "step": 103918 }, { "epoch": 9.0, "learning_rate": 5.25745923441656e-05, "loss": 1.3551, "step": 103920 }, { "epoch": 9.0, "learning_rate": 5.257242396206777e-05, "loss": 1.1245, "step": 103930 }, { "epoch": 9.0, "learning_rate": 5.257025557996993e-05, "loss": 1.3586, "step": 103940 }, { "epoch": 9.0, "learning_rate": 5.256808719787209e-05, "loss": 0.9175, "step": 103950 }, { "epoch": 9.0, "learning_rate": 5.256591881577425e-05, "loss": 1.2063, "step": 103960 }, { "epoch": 9.0, "learning_rate": 5.256375043367641e-05, "loss": 1.1014, "step": 103970 }, { "epoch": 9.01, "learning_rate": 5.256158205157858e-05, "loss": 1.3624, "step": 103980 }, { "epoch": 9.01, "learning_rate": 5.255941366948074e-05, "loss": 1.3406, "step": 103990 }, { "epoch": 9.01, "learning_rate": 5.2557245287382906e-05, "loss": 1.3508, "step": 104000 }, { "epoch": 9.01, "learning_rate": 5.2555076905285065e-05, "loss": 1.2522, "step": 104010 }, { "epoch": 9.01, "learning_rate": 5.255290852318723e-05, "loss": 1.1631, "step": 104020 }, { "epoch": 9.01, "learning_rate": 5.2550740141089385e-05, "loss": 1.1661, "step": 104030 }, { "epoch": 9.01, "learning_rate": 5.254857175899155e-05, "loss": 1.2209, "step": 104040 }, { "epoch": 9.01, "learning_rate": 5.254640337689372e-05, "loss": 1.1789, "step": 104050 }, { "epoch": 9.01, "learning_rate": 5.254423499479588e-05, "loss": 1.2537, "step": 104060 }, { "epoch": 9.01, "learning_rate": 5.254206661269804e-05, "loss": 1.1937, "step": 104070 }, { "epoch": 9.01, "learning_rate": 5.25398982306002e-05, "loss": 1.1736, "step": 104080 }, { "epoch": 9.01, "learning_rate": 5.253772984850237e-05, "loss": 1.2165, "step": 104090 }, { "epoch": 9.02, "learning_rate": 5.2535561466404536e-05, "loss": 1.1517, "step": 104100 }, { "epoch": 9.02, "learning_rate": 5.253339308430669e-05, "loss": 1.2209, "step": 104110 }, { "epoch": 9.02, "learning_rate": 5.2531224702208855e-05, "loss": 1.0292, "step": 104120 }, { "epoch": 9.02, "learning_rate": 5.2529056320111015e-05, "loss": 1.0741, "step": 104130 }, { "epoch": 9.02, "learning_rate": 5.252688793801318e-05, "loss": 1.2112, "step": 104140 }, { "epoch": 9.02, "learning_rate": 5.252471955591534e-05, "loss": 1.4372, "step": 104150 }, { "epoch": 9.02, "learning_rate": 5.252255117381751e-05, "loss": 1.1463, "step": 104160 }, { "epoch": 9.02, "learning_rate": 5.252038279171967e-05, "loss": 1.3011, "step": 104170 }, { "epoch": 9.02, "learning_rate": 5.2518214409621826e-05, "loss": 1.0689, "step": 104180 }, { "epoch": 9.02, "learning_rate": 5.2516046027523986e-05, "loss": 1.2624, "step": 104190 }, { "epoch": 9.02, "learning_rate": 5.251387764542615e-05, "loss": 1.285, "step": 104200 }, { "epoch": 9.03, "learning_rate": 5.251170926332832e-05, "loss": 1.2517, "step": 104210 }, { "epoch": 9.03, "learning_rate": 5.250954088123048e-05, "loss": 1.2449, "step": 104220 }, { "epoch": 9.03, "learning_rate": 5.2507372499132645e-05, "loss": 1.2367, "step": 104230 }, { "epoch": 9.03, "learning_rate": 5.2505204117034804e-05, "loss": 1.1525, "step": 104240 }, { "epoch": 9.03, "learning_rate": 5.250303573493697e-05, "loss": 1.0443, "step": 104250 }, { "epoch": 9.03, "learning_rate": 5.2500867352839123e-05, "loss": 1.1499, "step": 104260 }, { "epoch": 9.03, "learning_rate": 5.249869897074129e-05, "loss": 1.235, "step": 104270 }, { "epoch": 9.03, "learning_rate": 5.2496530588643456e-05, "loss": 1.1401, "step": 104280 }, { "epoch": 9.03, "learning_rate": 5.2494362206545616e-05, "loss": 1.1807, "step": 104290 }, { "epoch": 9.03, "learning_rate": 5.249219382444778e-05, "loss": 1.1496, "step": 104300 }, { "epoch": 9.03, "learning_rate": 5.249002544234994e-05, "loss": 1.2124, "step": 104310 }, { "epoch": 9.03, "learning_rate": 5.248785706025211e-05, "loss": 1.316, "step": 104320 }, { "epoch": 9.04, "learning_rate": 5.2485688678154275e-05, "loss": 1.257, "step": 104330 }, { "epoch": 9.04, "learning_rate": 5.248352029605643e-05, "loss": 1.1116, "step": 104340 }, { "epoch": 9.04, "learning_rate": 5.2481351913958594e-05, "loss": 1.2419, "step": 104350 }, { "epoch": 9.04, "learning_rate": 5.2479183531860753e-05, "loss": 1.3764, "step": 104360 }, { "epoch": 9.04, "learning_rate": 5.247701514976292e-05, "loss": 1.4532, "step": 104370 }, { "epoch": 9.04, "learning_rate": 5.247484676766508e-05, "loss": 1.1535, "step": 104380 }, { "epoch": 9.04, "learning_rate": 5.2472678385567246e-05, "loss": 1.1799, "step": 104390 }, { "epoch": 9.04, "learning_rate": 5.247051000346941e-05, "loss": 1.3756, "step": 104400 }, { "epoch": 9.04, "learning_rate": 5.246834162137157e-05, "loss": 1.35, "step": 104410 }, { "epoch": 9.04, "learning_rate": 5.246617323927373e-05, "loss": 1.1924, "step": 104420 }, { "epoch": 9.04, "learning_rate": 5.246400485717589e-05, "loss": 1.3861, "step": 104430 }, { "epoch": 9.05, "learning_rate": 5.246183647507806e-05, "loss": 1.3433, "step": 104440 }, { "epoch": 9.05, "learning_rate": 5.245966809298022e-05, "loss": 1.41, "step": 104450 }, { "epoch": 9.05, "learning_rate": 5.2457499710882383e-05, "loss": 1.4086, "step": 104460 }, { "epoch": 9.05, "learning_rate": 5.245533132878455e-05, "loss": 1.2255, "step": 104470 }, { "epoch": 9.05, "learning_rate": 5.245316294668671e-05, "loss": 1.0994, "step": 104480 }, { "epoch": 9.05, "learning_rate": 5.245099456458887e-05, "loss": 1.0967, "step": 104490 }, { "epoch": 9.05, "learning_rate": 5.244882618249103e-05, "loss": 1.1597, "step": 104500 }, { "epoch": 9.05, "learning_rate": 5.2446657800393195e-05, "loss": 1.1855, "step": 104510 }, { "epoch": 9.05, "learning_rate": 5.2444489418295355e-05, "loss": 1.3063, "step": 104520 }, { "epoch": 9.05, "learning_rate": 5.244232103619752e-05, "loss": 1.2724, "step": 104530 }, { "epoch": 9.05, "learning_rate": 5.244015265409969e-05, "loss": 1.1688, "step": 104540 }, { "epoch": 9.05, "learning_rate": 5.243798427200185e-05, "loss": 1.2306, "step": 104550 }, { "epoch": 9.06, "learning_rate": 5.2435815889904014e-05, "loss": 1.2838, "step": 104560 }, { "epoch": 9.06, "learning_rate": 5.2433647507806166e-05, "loss": 1.2277, "step": 104570 }, { "epoch": 9.06, "learning_rate": 5.243147912570833e-05, "loss": 1.2973, "step": 104580 }, { "epoch": 9.06, "learning_rate": 5.242931074361049e-05, "loss": 1.1407, "step": 104590 }, { "epoch": 9.06, "learning_rate": 5.242714236151266e-05, "loss": 1.2911, "step": 104600 }, { "epoch": 9.06, "learning_rate": 5.2424973979414825e-05, "loss": 1.4416, "step": 104610 }, { "epoch": 9.06, "learning_rate": 5.2422805597316985e-05, "loss": 1.1553, "step": 104620 }, { "epoch": 9.06, "learning_rate": 5.242063721521915e-05, "loss": 1.2647, "step": 104630 }, { "epoch": 9.06, "learning_rate": 5.241846883312131e-05, "loss": 1.3841, "step": 104640 }, { "epoch": 9.06, "learning_rate": 5.241630045102347e-05, "loss": 1.2181, "step": 104650 }, { "epoch": 9.06, "learning_rate": 5.241413206892563e-05, "loss": 1.237, "step": 104660 }, { "epoch": 9.07, "learning_rate": 5.2411963686827796e-05, "loss": 1.2255, "step": 104670 }, { "epoch": 9.07, "learning_rate": 5.240979530472996e-05, "loss": 1.0911, "step": 104680 }, { "epoch": 9.07, "learning_rate": 5.240762692263212e-05, "loss": 1.1863, "step": 104690 }, { "epoch": 9.07, "learning_rate": 5.240545854053429e-05, "loss": 1.1914, "step": 104700 }, { "epoch": 9.07, "learning_rate": 5.240329015843645e-05, "loss": 1.2372, "step": 104710 }, { "epoch": 9.07, "learning_rate": 5.240112177633861e-05, "loss": 1.1185, "step": 104720 }, { "epoch": 9.07, "learning_rate": 5.239895339424077e-05, "loss": 1.208, "step": 104730 }, { "epoch": 9.07, "learning_rate": 5.2396785012142934e-05, "loss": 1.3279, "step": 104740 }, { "epoch": 9.07, "learning_rate": 5.2394616630045094e-05, "loss": 1.2473, "step": 104750 }, { "epoch": 9.07, "learning_rate": 5.239244824794726e-05, "loss": 1.0565, "step": 104760 }, { "epoch": 9.07, "learning_rate": 5.2390279865849426e-05, "loss": 1.4235, "step": 104770 }, { "epoch": 9.07, "learning_rate": 5.2388111483751586e-05, "loss": 1.0651, "step": 104780 }, { "epoch": 9.08, "learning_rate": 5.238594310165375e-05, "loss": 1.0438, "step": 104790 }, { "epoch": 9.08, "learning_rate": 5.2383774719555905e-05, "loss": 1.4131, "step": 104800 }, { "epoch": 9.08, "learning_rate": 5.238160633745807e-05, "loss": 1.4612, "step": 104810 }, { "epoch": 9.08, "learning_rate": 5.237943795536023e-05, "loss": 1.1327, "step": 104820 }, { "epoch": 9.08, "learning_rate": 5.23772695732624e-05, "loss": 1.2265, "step": 104830 }, { "epoch": 9.08, "learning_rate": 5.2375101191164564e-05, "loss": 1.249, "step": 104840 }, { "epoch": 9.08, "learning_rate": 5.2372932809066724e-05, "loss": 1.0795, "step": 104850 }, { "epoch": 9.08, "learning_rate": 5.237076442696889e-05, "loss": 1.0922, "step": 104860 }, { "epoch": 9.08, "learning_rate": 5.236859604487105e-05, "loss": 1.2118, "step": 104870 }, { "epoch": 9.08, "learning_rate": 5.236642766277321e-05, "loss": 1.1263, "step": 104880 }, { "epoch": 9.08, "learning_rate": 5.236425928067537e-05, "loss": 1.3561, "step": 104890 }, { "epoch": 9.09, "learning_rate": 5.2362090898577535e-05, "loss": 1.3643, "step": 104900 }, { "epoch": 9.09, "learning_rate": 5.23599225164797e-05, "loss": 1.1385, "step": 104910 }, { "epoch": 9.09, "learning_rate": 5.235775413438186e-05, "loss": 1.3597, "step": 104920 }, { "epoch": 9.09, "learning_rate": 5.235558575228403e-05, "loss": 1.2261, "step": 104930 }, { "epoch": 9.09, "learning_rate": 5.235341737018619e-05, "loss": 1.2877, "step": 104940 }, { "epoch": 9.09, "learning_rate": 5.235124898808835e-05, "loss": 1.2205, "step": 104950 }, { "epoch": 9.09, "learning_rate": 5.2349080605990507e-05, "loss": 1.1279, "step": 104960 }, { "epoch": 9.09, "learning_rate": 5.234691222389267e-05, "loss": 1.2729, "step": 104970 }, { "epoch": 9.09, "learning_rate": 5.234474384179484e-05, "loss": 1.2861, "step": 104980 }, { "epoch": 9.09, "learning_rate": 5.2342575459697e-05, "loss": 1.2338, "step": 104990 }, { "epoch": 9.09, "learning_rate": 5.2340407077599165e-05, "loss": 1.3321, "step": 105000 }, { "epoch": 9.09, "learning_rate": 5.2338238695501325e-05, "loss": 1.3275, "step": 105010 }, { "epoch": 9.1, "learning_rate": 5.233607031340349e-05, "loss": 1.0694, "step": 105020 }, { "epoch": 9.1, "learning_rate": 5.2333901931305644e-05, "loss": 1.2247, "step": 105030 }, { "epoch": 9.1, "learning_rate": 5.233173354920781e-05, "loss": 1.1926, "step": 105040 }, { "epoch": 9.1, "learning_rate": 5.232956516710998e-05, "loss": 1.1529, "step": 105050 }, { "epoch": 9.1, "learning_rate": 5.2327396785012137e-05, "loss": 1.3673, "step": 105060 }, { "epoch": 9.1, "learning_rate": 5.23252284029143e-05, "loss": 1.1168, "step": 105070 }, { "epoch": 9.1, "learning_rate": 5.232306002081646e-05, "loss": 1.4133, "step": 105080 }, { "epoch": 9.1, "learning_rate": 5.232089163871863e-05, "loss": 1.2378, "step": 105090 }, { "epoch": 9.1, "learning_rate": 5.2318723256620795e-05, "loss": 1.0879, "step": 105100 }, { "epoch": 9.1, "learning_rate": 5.231655487452295e-05, "loss": 1.2174, "step": 105110 }, { "epoch": 9.1, "learning_rate": 5.2314386492425115e-05, "loss": 1.1321, "step": 105120 }, { "epoch": 9.1, "learning_rate": 5.2312218110327274e-05, "loss": 1.2696, "step": 105130 }, { "epoch": 9.11, "learning_rate": 5.231004972822944e-05, "loss": 1.407, "step": 105140 }, { "epoch": 9.11, "learning_rate": 5.23078813461316e-05, "loss": 1.3317, "step": 105150 }, { "epoch": 9.11, "learning_rate": 5.230571296403377e-05, "loss": 1.254, "step": 105160 }, { "epoch": 9.11, "learning_rate": 5.230354458193593e-05, "loss": 1.1486, "step": 105170 }, { "epoch": 9.11, "learning_rate": 5.230137619983809e-05, "loss": 1.3473, "step": 105180 }, { "epoch": 9.11, "learning_rate": 5.229920781774025e-05, "loss": 1.1453, "step": 105190 }, { "epoch": 9.11, "learning_rate": 5.229703943564241e-05, "loss": 1.2901, "step": 105200 }, { "epoch": 9.11, "learning_rate": 5.229487105354458e-05, "loss": 1.0457, "step": 105210 }, { "epoch": 9.11, "learning_rate": 5.229270267144674e-05, "loss": 1.2781, "step": 105220 }, { "epoch": 9.11, "learning_rate": 5.2290534289348904e-05, "loss": 1.3533, "step": 105230 }, { "epoch": 9.11, "learning_rate": 5.228836590725107e-05, "loss": 1.1724, "step": 105240 }, { "epoch": 9.12, "learning_rate": 5.228619752515323e-05, "loss": 1.245, "step": 105250 }, { "epoch": 9.12, "learning_rate": 5.228402914305538e-05, "loss": 1.1573, "step": 105260 }, { "epoch": 9.12, "learning_rate": 5.228186076095755e-05, "loss": 1.233, "step": 105270 }, { "epoch": 9.12, "learning_rate": 5.2279692378859716e-05, "loss": 1.2764, "step": 105280 }, { "epoch": 9.12, "learning_rate": 5.2277523996761876e-05, "loss": 1.1421, "step": 105290 }, { "epoch": 9.12, "learning_rate": 5.227535561466404e-05, "loss": 1.2395, "step": 105300 }, { "epoch": 9.12, "learning_rate": 5.22731872325662e-05, "loss": 1.2373, "step": 105310 }, { "epoch": 9.12, "learning_rate": 5.227101885046837e-05, "loss": 1.2033, "step": 105320 }, { "epoch": 9.12, "learning_rate": 5.2268850468370534e-05, "loss": 1.1748, "step": 105330 }, { "epoch": 9.12, "learning_rate": 5.226668208627269e-05, "loss": 1.335, "step": 105340 }, { "epoch": 9.12, "learning_rate": 5.2264513704174854e-05, "loss": 1.1932, "step": 105350 }, { "epoch": 9.12, "learning_rate": 5.226234532207701e-05, "loss": 1.0907, "step": 105360 }, { "epoch": 9.13, "learning_rate": 5.226017693997918e-05, "loss": 1.16, "step": 105370 }, { "epoch": 9.13, "learning_rate": 5.225800855788134e-05, "loss": 1.3525, "step": 105380 }, { "epoch": 9.13, "learning_rate": 5.2255840175783506e-05, "loss": 1.0876, "step": 105390 }, { "epoch": 9.13, "learning_rate": 5.225367179368567e-05, "loss": 1.397, "step": 105400 }, { "epoch": 9.13, "learning_rate": 5.225150341158783e-05, "loss": 1.1635, "step": 105410 }, { "epoch": 9.13, "learning_rate": 5.224933502948999e-05, "loss": 1.1387, "step": 105420 }, { "epoch": 9.13, "learning_rate": 5.224716664739215e-05, "loss": 1.2299, "step": 105430 }, { "epoch": 9.13, "learning_rate": 5.224499826529432e-05, "loss": 1.2807, "step": 105440 }, { "epoch": 9.13, "learning_rate": 5.224282988319648e-05, "loss": 1.3944, "step": 105450 }, { "epoch": 9.13, "learning_rate": 5.224066150109864e-05, "loss": 1.1928, "step": 105460 }, { "epoch": 9.13, "learning_rate": 5.223849311900081e-05, "loss": 1.206, "step": 105470 }, { "epoch": 9.14, "learning_rate": 5.223632473690297e-05, "loss": 1.265, "step": 105480 }, { "epoch": 9.14, "learning_rate": 5.223415635480513e-05, "loss": 1.2672, "step": 105490 }, { "epoch": 9.14, "learning_rate": 5.223198797270729e-05, "loss": 1.3172, "step": 105500 }, { "epoch": 9.14, "learning_rate": 5.2229819590609455e-05, "loss": 1.245, "step": 105510 }, { "epoch": 9.14, "learning_rate": 5.2227651208511614e-05, "loss": 1.3155, "step": 105520 }, { "epoch": 9.14, "learning_rate": 5.222548282641378e-05, "loss": 1.237, "step": 105530 }, { "epoch": 9.14, "learning_rate": 5.222331444431595e-05, "loss": 1.3636, "step": 105540 }, { "epoch": 9.14, "learning_rate": 5.222114606221811e-05, "loss": 1.1119, "step": 105550 }, { "epoch": 9.14, "learning_rate": 5.221897768012027e-05, "loss": 1.425, "step": 105560 }, { "epoch": 9.14, "learning_rate": 5.2216809298022426e-05, "loss": 1.3809, "step": 105570 }, { "epoch": 9.14, "learning_rate": 5.221464091592459e-05, "loss": 1.19, "step": 105580 }, { "epoch": 9.14, "learning_rate": 5.221247253382675e-05, "loss": 1.2604, "step": 105590 }, { "epoch": 9.15, "learning_rate": 5.221030415172892e-05, "loss": 1.1731, "step": 105600 }, { "epoch": 9.15, "learning_rate": 5.2208135769631085e-05, "loss": 1.1845, "step": 105610 }, { "epoch": 9.15, "learning_rate": 5.2205967387533244e-05, "loss": 1.3133, "step": 105620 }, { "epoch": 9.15, "learning_rate": 5.220379900543541e-05, "loss": 1.088, "step": 105630 }, { "epoch": 9.15, "learning_rate": 5.220163062333757e-05, "loss": 1.118, "step": 105640 }, { "epoch": 9.15, "learning_rate": 5.219946224123973e-05, "loss": 1.307, "step": 105650 }, { "epoch": 9.15, "learning_rate": 5.219729385914189e-05, "loss": 1.1361, "step": 105660 }, { "epoch": 9.15, "learning_rate": 5.2195125477044056e-05, "loss": 1.2187, "step": 105670 }, { "epoch": 9.15, "learning_rate": 5.219295709494622e-05, "loss": 1.1561, "step": 105680 }, { "epoch": 9.15, "learning_rate": 5.219078871284838e-05, "loss": 1.1033, "step": 105690 }, { "epoch": 9.15, "learning_rate": 5.218862033075055e-05, "loss": 1.4199, "step": 105700 }, { "epoch": 9.16, "learning_rate": 5.218645194865271e-05, "loss": 1.2424, "step": 105710 }, { "epoch": 9.16, "learning_rate": 5.218428356655487e-05, "loss": 1.2124, "step": 105720 }, { "epoch": 9.16, "learning_rate": 5.218211518445703e-05, "loss": 1.2056, "step": 105730 }, { "epoch": 9.16, "learning_rate": 5.2179946802359194e-05, "loss": 1.1927, "step": 105740 }, { "epoch": 9.16, "learning_rate": 5.217777842026136e-05, "loss": 1.1387, "step": 105750 }, { "epoch": 9.16, "learning_rate": 5.217561003816352e-05, "loss": 1.1093, "step": 105760 }, { "epoch": 9.16, "learning_rate": 5.2173441656065686e-05, "loss": 1.1544, "step": 105770 }, { "epoch": 9.16, "learning_rate": 5.2171273273967846e-05, "loss": 1.2979, "step": 105780 }, { "epoch": 9.16, "learning_rate": 5.216910489187001e-05, "loss": 1.1872, "step": 105790 }, { "epoch": 9.16, "learning_rate": 5.2166936509772165e-05, "loss": 1.2057, "step": 105800 }, { "epoch": 9.16, "learning_rate": 5.216476812767433e-05, "loss": 1.2528, "step": 105810 }, { "epoch": 9.16, "learning_rate": 5.216259974557649e-05, "loss": 1.1805, "step": 105820 }, { "epoch": 9.17, "learning_rate": 5.216043136347866e-05, "loss": 1.0042, "step": 105830 }, { "epoch": 9.17, "learning_rate": 5.2158262981380824e-05, "loss": 1.3768, "step": 105840 }, { "epoch": 9.17, "learning_rate": 5.215609459928298e-05, "loss": 1.1338, "step": 105850 }, { "epoch": 9.17, "learning_rate": 5.215392621718515e-05, "loss": 1.2843, "step": 105860 }, { "epoch": 9.17, "learning_rate": 5.215175783508731e-05, "loss": 0.9982, "step": 105870 }, { "epoch": 9.17, "learning_rate": 5.214958945298947e-05, "loss": 1.0867, "step": 105880 }, { "epoch": 9.17, "learning_rate": 5.214742107089163e-05, "loss": 1.3819, "step": 105890 }, { "epoch": 9.17, "learning_rate": 5.2145252688793795e-05, "loss": 1.2111, "step": 105900 }, { "epoch": 9.17, "learning_rate": 5.214308430669596e-05, "loss": 1.2279, "step": 105910 }, { "epoch": 9.17, "learning_rate": 5.214091592459812e-05, "loss": 1.177, "step": 105920 }, { "epoch": 9.17, "learning_rate": 5.213874754250029e-05, "loss": 1.2849, "step": 105930 }, { "epoch": 9.18, "learning_rate": 5.213657916040245e-05, "loss": 1.145, "step": 105940 }, { "epoch": 9.18, "learning_rate": 5.2134410778304613e-05, "loss": 1.2206, "step": 105950 }, { "epoch": 9.18, "learning_rate": 5.2132242396206766e-05, "loss": 1.2954, "step": 105960 }, { "epoch": 9.18, "learning_rate": 5.213007401410893e-05, "loss": 1.3544, "step": 105970 }, { "epoch": 9.18, "learning_rate": 5.21279056320111e-05, "loss": 1.2369, "step": 105980 }, { "epoch": 9.18, "learning_rate": 5.212573724991326e-05, "loss": 1.1033, "step": 105990 }, { "epoch": 9.18, "learning_rate": 5.2123568867815425e-05, "loss": 1.2101, "step": 106000 }, { "epoch": 9.18, "learning_rate": 5.2121400485717585e-05, "loss": 1.2502, "step": 106010 }, { "epoch": 9.18, "learning_rate": 5.211923210361975e-05, "loss": 1.2642, "step": 106020 }, { "epoch": 9.18, "learning_rate": 5.2117063721521904e-05, "loss": 1.1248, "step": 106030 }, { "epoch": 9.18, "learning_rate": 5.211489533942407e-05, "loss": 1.2954, "step": 106040 }, { "epoch": 9.18, "learning_rate": 5.211272695732624e-05, "loss": 1.2206, "step": 106050 }, { "epoch": 9.19, "learning_rate": 5.2110558575228396e-05, "loss": 1.275, "step": 106060 }, { "epoch": 9.19, "learning_rate": 5.210839019313056e-05, "loss": 1.0253, "step": 106070 }, { "epoch": 9.19, "learning_rate": 5.210622181103272e-05, "loss": 1.3062, "step": 106080 }, { "epoch": 9.19, "learning_rate": 5.210405342893489e-05, "loss": 1.2739, "step": 106090 }, { "epoch": 9.19, "learning_rate": 5.2101885046837055e-05, "loss": 1.3531, "step": 106100 }, { "epoch": 9.19, "learning_rate": 5.209971666473921e-05, "loss": 1.2848, "step": 106110 }, { "epoch": 9.19, "learning_rate": 5.2097548282641374e-05, "loss": 1.1705, "step": 106120 }, { "epoch": 9.19, "learning_rate": 5.2095379900543534e-05, "loss": 1.3412, "step": 106130 }, { "epoch": 9.19, "learning_rate": 5.20932115184457e-05, "loss": 1.3437, "step": 106140 }, { "epoch": 9.19, "learning_rate": 5.209104313634786e-05, "loss": 1.0662, "step": 106150 }, { "epoch": 9.19, "learning_rate": 5.2088874754250026e-05, "loss": 1.2648, "step": 106160 }, { "epoch": 9.19, "learning_rate": 5.208670637215219e-05, "loss": 0.9856, "step": 106170 }, { "epoch": 9.2, "learning_rate": 5.208453799005435e-05, "loss": 1.2827, "step": 106180 }, { "epoch": 9.2, "learning_rate": 5.208236960795651e-05, "loss": 1.1174, "step": 106190 }, { "epoch": 9.2, "learning_rate": 5.208020122585867e-05, "loss": 1.1598, "step": 106200 }, { "epoch": 9.2, "learning_rate": 5.207803284376084e-05, "loss": 1.2202, "step": 106210 }, { "epoch": 9.2, "learning_rate": 5.2075864461663e-05, "loss": 1.1119, "step": 106220 }, { "epoch": 9.2, "learning_rate": 5.2073696079565164e-05, "loss": 1.2087, "step": 106230 }, { "epoch": 9.2, "learning_rate": 5.207152769746733e-05, "loss": 1.2591, "step": 106240 }, { "epoch": 9.2, "learning_rate": 5.206935931536949e-05, "loss": 1.1202, "step": 106250 }, { "epoch": 9.2, "learning_rate": 5.206719093327165e-05, "loss": 1.4157, "step": 106260 }, { "epoch": 9.2, "learning_rate": 5.206502255117381e-05, "loss": 1.177, "step": 106270 }, { "epoch": 9.2, "learning_rate": 5.2062854169075976e-05, "loss": 1.3146, "step": 106280 }, { "epoch": 9.21, "learning_rate": 5.2060685786978135e-05, "loss": 1.2223, "step": 106290 }, { "epoch": 9.21, "learning_rate": 5.20585174048803e-05, "loss": 1.1815, "step": 106300 }, { "epoch": 9.21, "learning_rate": 5.205634902278247e-05, "loss": 1.2327, "step": 106310 }, { "epoch": 9.21, "learning_rate": 5.205418064068463e-05, "loss": 1.2731, "step": 106320 }, { "epoch": 9.21, "learning_rate": 5.2052012258586794e-05, "loss": 1.2097, "step": 106330 }, { "epoch": 9.21, "learning_rate": 5.204984387648895e-05, "loss": 1.2139, "step": 106340 }, { "epoch": 9.21, "learning_rate": 5.204767549439111e-05, "loss": 1.182, "step": 106350 }, { "epoch": 9.21, "learning_rate": 5.204550711229327e-05, "loss": 1.1233, "step": 106360 }, { "epoch": 9.21, "learning_rate": 5.204333873019544e-05, "loss": 1.3055, "step": 106370 }, { "epoch": 9.21, "learning_rate": 5.20411703480976e-05, "loss": 1.1756, "step": 106380 }, { "epoch": 9.21, "learning_rate": 5.2039001965999765e-05, "loss": 1.2417, "step": 106390 }, { "epoch": 9.21, "learning_rate": 5.203683358390193e-05, "loss": 1.1871, "step": 106400 }, { "epoch": 9.22, "learning_rate": 5.203466520180409e-05, "loss": 1.0637, "step": 106410 }, { "epoch": 9.22, "learning_rate": 5.203249681970625e-05, "loss": 1.1644, "step": 106420 }, { "epoch": 9.22, "learning_rate": 5.203032843760841e-05, "loss": 1.2457, "step": 106430 }, { "epoch": 9.22, "learning_rate": 5.202816005551058e-05, "loss": 1.4232, "step": 106440 }, { "epoch": 9.22, "learning_rate": 5.2025991673412736e-05, "loss": 1.4844, "step": 106450 }, { "epoch": 9.22, "learning_rate": 5.20238232913149e-05, "loss": 1.4273, "step": 106460 }, { "epoch": 9.22, "learning_rate": 5.202165490921707e-05, "loss": 1.3317, "step": 106470 }, { "epoch": 9.22, "learning_rate": 5.201948652711923e-05, "loss": 1.4259, "step": 106480 }, { "epoch": 9.22, "learning_rate": 5.201731814502139e-05, "loss": 1.2294, "step": 106490 }, { "epoch": 9.22, "learning_rate": 5.201514976292355e-05, "loss": 1.1364, "step": 106500 }, { "epoch": 9.22, "learning_rate": 5.2012981380825715e-05, "loss": 1.1812, "step": 106510 }, { "epoch": 9.23, "learning_rate": 5.2010812998727874e-05, "loss": 1.3687, "step": 106520 }, { "epoch": 9.23, "learning_rate": 5.200864461663004e-05, "loss": 1.1732, "step": 106530 }, { "epoch": 9.23, "learning_rate": 5.200647623453221e-05, "loss": 1.2296, "step": 106540 }, { "epoch": 9.23, "learning_rate": 5.2004307852434367e-05, "loss": 1.1636, "step": 106550 }, { "epoch": 9.23, "learning_rate": 5.200213947033653e-05, "loss": 1.4035, "step": 106560 }, { "epoch": 9.23, "learning_rate": 5.1999971088238686e-05, "loss": 1.3275, "step": 106570 }, { "epoch": 9.23, "learning_rate": 5.199780270614085e-05, "loss": 1.1123, "step": 106580 }, { "epoch": 9.23, "learning_rate": 5.199563432404301e-05, "loss": 1.1639, "step": 106590 }, { "epoch": 9.23, "learning_rate": 5.199346594194518e-05, "loss": 1.3376, "step": 106600 }, { "epoch": 9.23, "learning_rate": 5.1991297559847345e-05, "loss": 1.2818, "step": 106610 }, { "epoch": 9.23, "learning_rate": 5.1989129177749504e-05, "loss": 1.0896, "step": 106620 }, { "epoch": 9.23, "learning_rate": 5.198696079565167e-05, "loss": 1.2171, "step": 106630 }, { "epoch": 9.24, "learning_rate": 5.198479241355383e-05, "loss": 1.2439, "step": 106640 }, { "epoch": 9.24, "learning_rate": 5.198262403145599e-05, "loss": 1.2781, "step": 106650 }, { "epoch": 9.24, "learning_rate": 5.198045564935815e-05, "loss": 1.0169, "step": 106660 }, { "epoch": 9.24, "learning_rate": 5.1978287267260316e-05, "loss": 1.1457, "step": 106670 }, { "epoch": 9.24, "learning_rate": 5.197611888516248e-05, "loss": 1.0908, "step": 106680 }, { "epoch": 9.24, "learning_rate": 5.197395050306464e-05, "loss": 1.1575, "step": 106690 }, { "epoch": 9.24, "learning_rate": 5.197178212096681e-05, "loss": 1.4354, "step": 106700 }, { "epoch": 9.24, "learning_rate": 5.196961373886897e-05, "loss": 1.1476, "step": 106710 }, { "epoch": 9.24, "learning_rate": 5.1967445356771134e-05, "loss": 1.2123, "step": 106720 }, { "epoch": 9.24, "learning_rate": 5.196527697467329e-05, "loss": 1.2764, "step": 106730 }, { "epoch": 9.24, "learning_rate": 5.1963108592575453e-05, "loss": 1.1038, "step": 106740 }, { "epoch": 9.25, "learning_rate": 5.196094021047762e-05, "loss": 1.221, "step": 106750 }, { "epoch": 9.25, "learning_rate": 5.195877182837978e-05, "loss": 1.2268, "step": 106760 }, { "epoch": 9.25, "learning_rate": 5.1956603446281946e-05, "loss": 1.0435, "step": 106770 }, { "epoch": 9.25, "learning_rate": 5.1954435064184105e-05, "loss": 1.3104, "step": 106780 }, { "epoch": 9.25, "learning_rate": 5.195226668208627e-05, "loss": 1.2169, "step": 106790 }, { "epoch": 9.25, "learning_rate": 5.1950098299988425e-05, "loss": 1.0813, "step": 106800 }, { "epoch": 9.25, "learning_rate": 5.194792991789059e-05, "loss": 1.2593, "step": 106810 }, { "epoch": 9.25, "learning_rate": 5.194576153579276e-05, "loss": 1.2637, "step": 106820 }, { "epoch": 9.25, "learning_rate": 5.194359315369492e-05, "loss": 1.3018, "step": 106830 }, { "epoch": 9.25, "learning_rate": 5.1941424771597083e-05, "loss": 1.3545, "step": 106840 }, { "epoch": 9.25, "learning_rate": 5.193925638949924e-05, "loss": 0.997, "step": 106850 }, { "epoch": 9.25, "learning_rate": 5.193708800740141e-05, "loss": 1.0963, "step": 106860 }, { "epoch": 9.26, "learning_rate": 5.1934919625303576e-05, "loss": 1.2796, "step": 106870 }, { "epoch": 9.26, "learning_rate": 5.193275124320573e-05, "loss": 1.5589, "step": 106880 }, { "epoch": 9.26, "learning_rate": 5.193058286110789e-05, "loss": 1.2362, "step": 106890 }, { "epoch": 9.26, "learning_rate": 5.1928414479010055e-05, "loss": 1.3192, "step": 106900 }, { "epoch": 9.26, "learning_rate": 5.192624609691222e-05, "loss": 1.3274, "step": 106910 }, { "epoch": 9.26, "learning_rate": 5.192407771481438e-05, "loss": 1.1578, "step": 106920 }, { "epoch": 9.26, "learning_rate": 5.192190933271655e-05, "loss": 1.3222, "step": 106930 }, { "epoch": 9.26, "learning_rate": 5.191974095061871e-05, "loss": 1.1539, "step": 106940 }, { "epoch": 9.26, "learning_rate": 5.191757256852087e-05, "loss": 1.4374, "step": 106950 }, { "epoch": 9.26, "learning_rate": 5.1915404186423026e-05, "loss": 1.1481, "step": 106960 }, { "epoch": 9.26, "learning_rate": 5.191323580432519e-05, "loss": 1.3646, "step": 106970 }, { "epoch": 9.27, "learning_rate": 5.191106742222736e-05, "loss": 1.2444, "step": 106980 }, { "epoch": 9.27, "learning_rate": 5.190889904012952e-05, "loss": 1.2371, "step": 106990 }, { "epoch": 9.27, "learning_rate": 5.1906730658031685e-05, "loss": 1.1135, "step": 107000 }, { "epoch": 9.27, "learning_rate": 5.1904562275933844e-05, "loss": 1.286, "step": 107010 }, { "epoch": 9.27, "learning_rate": 5.190239389383601e-05, "loss": 1.5166, "step": 107020 }, { "epoch": 9.27, "learning_rate": 5.1900225511738164e-05, "loss": 1.3337, "step": 107030 }, { "epoch": 9.27, "learning_rate": 5.189805712964033e-05, "loss": 1.1413, "step": 107040 }, { "epoch": 9.27, "learning_rate": 5.1895888747542496e-05, "loss": 1.1654, "step": 107050 }, { "epoch": 9.27, "learning_rate": 5.1893720365444656e-05, "loss": 1.3838, "step": 107060 }, { "epoch": 9.27, "learning_rate": 5.189155198334682e-05, "loss": 1.1064, "step": 107070 }, { "epoch": 9.27, "learning_rate": 5.188938360124898e-05, "loss": 1.3906, "step": 107080 }, { "epoch": 9.27, "learning_rate": 5.188721521915115e-05, "loss": 1.1306, "step": 107090 }, { "epoch": 9.28, "learning_rate": 5.1885046837053315e-05, "loss": 1.1267, "step": 107100 }, { "epoch": 9.28, "learning_rate": 5.188287845495547e-05, "loss": 1.2116, "step": 107110 }, { "epoch": 9.28, "learning_rate": 5.1880710072857634e-05, "loss": 1.0445, "step": 107120 }, { "epoch": 9.28, "learning_rate": 5.1878541690759794e-05, "loss": 1.3193, "step": 107130 }, { "epoch": 9.28, "learning_rate": 5.187637330866196e-05, "loss": 1.237, "step": 107140 }, { "epoch": 9.28, "learning_rate": 5.187420492656412e-05, "loss": 1.3042, "step": 107150 }, { "epoch": 9.28, "learning_rate": 5.1872036544466286e-05, "loss": 1.295, "step": 107160 }, { "epoch": 9.28, "learning_rate": 5.186986816236845e-05, "loss": 1.2941, "step": 107170 }, { "epoch": 9.28, "learning_rate": 5.186769978027061e-05, "loss": 1.117, "step": 107180 }, { "epoch": 9.28, "learning_rate": 5.186553139817277e-05, "loss": 1.2398, "step": 107190 }, { "epoch": 9.28, "learning_rate": 5.186336301607493e-05, "loss": 1.2446, "step": 107200 }, { "epoch": 9.29, "learning_rate": 5.18611946339771e-05, "loss": 1.3335, "step": 107210 }, { "epoch": 9.29, "learning_rate": 5.185902625187926e-05, "loss": 1.3125, "step": 107220 }, { "epoch": 9.29, "learning_rate": 5.1856857869781424e-05, "loss": 1.2995, "step": 107230 }, { "epoch": 9.29, "learning_rate": 5.185468948768359e-05, "loss": 1.3615, "step": 107240 }, { "epoch": 9.29, "learning_rate": 5.185252110558575e-05, "loss": 1.2347, "step": 107250 }, { "epoch": 9.29, "learning_rate": 5.185035272348791e-05, "loss": 1.3689, "step": 107260 }, { "epoch": 9.29, "learning_rate": 5.184818434139007e-05, "loss": 1.3497, "step": 107270 }, { "epoch": 9.29, "learning_rate": 5.1846015959292235e-05, "loss": 1.1705, "step": 107280 }, { "epoch": 9.29, "learning_rate": 5.1843847577194395e-05, "loss": 1.0809, "step": 107290 }, { "epoch": 9.29, "learning_rate": 5.184167919509656e-05, "loss": 1.4005, "step": 107300 }, { "epoch": 9.29, "learning_rate": 5.183951081299873e-05, "loss": 1.3263, "step": 107310 }, { "epoch": 9.29, "learning_rate": 5.183734243090089e-05, "loss": 1.1091, "step": 107320 }, { "epoch": 9.3, "learning_rate": 5.1835174048803054e-05, "loss": 1.4291, "step": 107330 }, { "epoch": 9.3, "learning_rate": 5.1833005666705207e-05, "loss": 1.2242, "step": 107340 }, { "epoch": 9.3, "learning_rate": 5.183083728460737e-05, "loss": 1.0791, "step": 107350 }, { "epoch": 9.3, "learning_rate": 5.182866890250953e-05, "loss": 1.1666, "step": 107360 }, { "epoch": 9.3, "learning_rate": 5.18265005204117e-05, "loss": 1.2182, "step": 107370 }, { "epoch": 9.3, "learning_rate": 5.1824332138313865e-05, "loss": 1.2149, "step": 107380 }, { "epoch": 9.3, "learning_rate": 5.1822163756216025e-05, "loss": 1.0594, "step": 107390 }, { "epoch": 9.3, "learning_rate": 5.181999537411819e-05, "loss": 1.3091, "step": 107400 }, { "epoch": 9.3, "learning_rate": 5.181782699202035e-05, "loss": 1.1239, "step": 107410 }, { "epoch": 9.3, "learning_rate": 5.181565860992251e-05, "loss": 1.2913, "step": 107420 }, { "epoch": 9.3, "learning_rate": 5.181349022782467e-05, "loss": 1.1354, "step": 107430 }, { "epoch": 9.3, "learning_rate": 5.1811321845726837e-05, "loss": 1.275, "step": 107440 }, { "epoch": 9.31, "learning_rate": 5.1809153463628996e-05, "loss": 1.4134, "step": 107450 }, { "epoch": 9.31, "learning_rate": 5.180698508153116e-05, "loss": 1.2509, "step": 107460 }, { "epoch": 9.31, "learning_rate": 5.180481669943333e-05, "loss": 1.1231, "step": 107470 }, { "epoch": 9.31, "learning_rate": 5.180264831733549e-05, "loss": 1.2202, "step": 107480 }, { "epoch": 9.31, "learning_rate": 5.1800479935237655e-05, "loss": 1.287, "step": 107490 }, { "epoch": 9.31, "learning_rate": 5.179831155313981e-05, "loss": 1.3293, "step": 107500 }, { "epoch": 9.31, "learning_rate": 5.1796143171041974e-05, "loss": 1.3915, "step": 107510 }, { "epoch": 9.31, "learning_rate": 5.1793974788944134e-05, "loss": 1.4466, "step": 107520 }, { "epoch": 9.31, "learning_rate": 5.17918064068463e-05, "loss": 1.539, "step": 107530 }, { "epoch": 9.31, "learning_rate": 5.1789638024748467e-05, "loss": 1.1401, "step": 107540 }, { "epoch": 9.31, "learning_rate": 5.1787469642650626e-05, "loss": 1.3926, "step": 107550 }, { "epoch": 9.32, "learning_rate": 5.178530126055279e-05, "loss": 1.362, "step": 107560 }, { "epoch": 9.32, "learning_rate": 5.1783132878454945e-05, "loss": 1.2878, "step": 107570 }, { "epoch": 9.32, "learning_rate": 5.178096449635711e-05, "loss": 1.1673, "step": 107580 }, { "epoch": 9.32, "learning_rate": 5.177879611425927e-05, "loss": 1.1907, "step": 107590 }, { "epoch": 9.32, "learning_rate": 5.177662773216144e-05, "loss": 1.0558, "step": 107600 }, { "epoch": 9.32, "learning_rate": 5.1774459350063604e-05, "loss": 1.3059, "step": 107610 }, { "epoch": 9.32, "learning_rate": 5.1772290967965764e-05, "loss": 1.4279, "step": 107620 }, { "epoch": 9.32, "learning_rate": 5.177012258586793e-05, "loss": 1.3921, "step": 107630 }, { "epoch": 9.32, "learning_rate": 5.176795420377009e-05, "loss": 1.1361, "step": 107640 }, { "epoch": 9.32, "learning_rate": 5.176578582167225e-05, "loss": 1.1628, "step": 107650 }, { "epoch": 9.32, "learning_rate": 5.176361743957441e-05, "loss": 1.0913, "step": 107660 }, { "epoch": 9.32, "learning_rate": 5.1761449057476575e-05, "loss": 1.2256, "step": 107670 }, { "epoch": 9.33, "learning_rate": 5.175928067537874e-05, "loss": 1.4328, "step": 107680 }, { "epoch": 9.33, "learning_rate": 5.17571122932809e-05, "loss": 1.2998, "step": 107690 }, { "epoch": 9.33, "learning_rate": 5.175494391118307e-05, "loss": 1.1823, "step": 107700 }, { "epoch": 9.33, "learning_rate": 5.175277552908523e-05, "loss": 1.3781, "step": 107710 }, { "epoch": 9.33, "learning_rate": 5.1750607146987394e-05, "loss": 1.3638, "step": 107720 }, { "epoch": 9.33, "learning_rate": 5.174843876488955e-05, "loss": 1.2901, "step": 107730 }, { "epoch": 9.33, "learning_rate": 5.174627038279171e-05, "loss": 1.1503, "step": 107740 }, { "epoch": 9.33, "learning_rate": 5.174410200069388e-05, "loss": 1.1837, "step": 107750 }, { "epoch": 9.33, "learning_rate": 5.174193361859604e-05, "loss": 1.552, "step": 107760 }, { "epoch": 9.33, "learning_rate": 5.1739765236498206e-05, "loss": 1.1623, "step": 107770 }, { "epoch": 9.33, "learning_rate": 5.1737596854400365e-05, "loss": 1.315, "step": 107780 }, { "epoch": 9.34, "learning_rate": 5.173542847230253e-05, "loss": 1.3267, "step": 107790 }, { "epoch": 9.34, "learning_rate": 5.1733260090204684e-05, "loss": 1.0444, "step": 107800 }, { "epoch": 9.34, "learning_rate": 5.173109170810685e-05, "loss": 1.2187, "step": 107810 }, { "epoch": 9.34, "learning_rate": 5.172892332600902e-05, "loss": 1.3524, "step": 107820 }, { "epoch": 9.34, "learning_rate": 5.172675494391118e-05, "loss": 1.3027, "step": 107830 }, { "epoch": 9.34, "learning_rate": 5.172458656181334e-05, "loss": 1.3745, "step": 107840 }, { "epoch": 9.34, "learning_rate": 5.17224181797155e-05, "loss": 1.1316, "step": 107850 }, { "epoch": 9.34, "learning_rate": 5.172024979761767e-05, "loss": 1.2375, "step": 107860 }, { "epoch": 9.34, "learning_rate": 5.1718081415519836e-05, "loss": 1.3491, "step": 107870 }, { "epoch": 9.34, "learning_rate": 5.171591303342199e-05, "loss": 1.2933, "step": 107880 }, { "epoch": 9.34, "learning_rate": 5.1713744651324155e-05, "loss": 1.342, "step": 107890 }, { "epoch": 9.34, "learning_rate": 5.1711576269226314e-05, "loss": 1.3304, "step": 107900 }, { "epoch": 9.35, "learning_rate": 5.170940788712848e-05, "loss": 1.0228, "step": 107910 }, { "epoch": 9.35, "learning_rate": 5.170723950503064e-05, "loss": 1.368, "step": 107920 }, { "epoch": 9.35, "learning_rate": 5.170507112293281e-05, "loss": 1.2156, "step": 107930 }, { "epoch": 9.35, "learning_rate": 5.170290274083497e-05, "loss": 1.2137, "step": 107940 }, { "epoch": 9.35, "learning_rate": 5.170073435873713e-05, "loss": 1.1307, "step": 107950 }, { "epoch": 9.35, "learning_rate": 5.1698565976639286e-05, "loss": 1.3126, "step": 107960 }, { "epoch": 9.35, "learning_rate": 5.169639759454145e-05, "loss": 1.3329, "step": 107970 }, { "epoch": 9.35, "learning_rate": 5.169422921244362e-05, "loss": 1.3781, "step": 107980 }, { "epoch": 9.35, "learning_rate": 5.169206083034578e-05, "loss": 1.0772, "step": 107990 }, { "epoch": 9.35, "learning_rate": 5.1689892448247944e-05, "loss": 1.2358, "step": 108000 }, { "epoch": 9.35, "learning_rate": 5.1687724066150104e-05, "loss": 1.2049, "step": 108010 }, { "epoch": 9.36, "learning_rate": 5.168555568405227e-05, "loss": 1.355, "step": 108020 }, { "epoch": 9.36, "learning_rate": 5.168338730195442e-05, "loss": 1.2245, "step": 108030 }, { "epoch": 9.36, "learning_rate": 5.168121891985659e-05, "loss": 1.3821, "step": 108040 }, { "epoch": 9.36, "learning_rate": 5.1679050537758756e-05, "loss": 1.2535, "step": 108050 }, { "epoch": 9.36, "learning_rate": 5.1676882155660916e-05, "loss": 1.3278, "step": 108060 }, { "epoch": 9.36, "learning_rate": 5.167471377356308e-05, "loss": 1.2774, "step": 108070 }, { "epoch": 9.36, "learning_rate": 5.167254539146524e-05, "loss": 1.2774, "step": 108080 }, { "epoch": 9.36, "learning_rate": 5.167037700936741e-05, "loss": 1.2433, "step": 108090 }, { "epoch": 9.36, "learning_rate": 5.1668208627269574e-05, "loss": 1.2384, "step": 108100 }, { "epoch": 9.36, "learning_rate": 5.166604024517173e-05, "loss": 1.1133, "step": 108110 }, { "epoch": 9.36, "learning_rate": 5.1663871863073894e-05, "loss": 1.3714, "step": 108120 }, { "epoch": 9.36, "learning_rate": 5.166170348097605e-05, "loss": 1.1087, "step": 108130 }, { "epoch": 9.37, "learning_rate": 5.165953509887822e-05, "loss": 1.1623, "step": 108140 }, { "epoch": 9.37, "learning_rate": 5.165736671678038e-05, "loss": 1.3091, "step": 108150 }, { "epoch": 9.37, "learning_rate": 5.1655198334682546e-05, "loss": 1.2295, "step": 108160 }, { "epoch": 9.37, "learning_rate": 5.165302995258471e-05, "loss": 1.1825, "step": 108170 }, { "epoch": 9.37, "learning_rate": 5.165086157048687e-05, "loss": 1.222, "step": 108180 }, { "epoch": 9.37, "learning_rate": 5.164869318838903e-05, "loss": 1.1116, "step": 108190 }, { "epoch": 9.37, "learning_rate": 5.164652480629119e-05, "loss": 1.142, "step": 108200 }, { "epoch": 9.37, "learning_rate": 5.164435642419336e-05, "loss": 1.218, "step": 108210 }, { "epoch": 9.37, "learning_rate": 5.164218804209552e-05, "loss": 1.27, "step": 108220 }, { "epoch": 9.37, "learning_rate": 5.164001965999768e-05, "loss": 1.3416, "step": 108230 }, { "epoch": 9.37, "learning_rate": 5.163785127789985e-05, "loss": 1.2887, "step": 108240 }, { "epoch": 9.38, "learning_rate": 5.163568289580201e-05, "loss": 1.2514, "step": 108250 }, { "epoch": 9.38, "learning_rate": 5.1633514513704176e-05, "loss": 1.353, "step": 108260 }, { "epoch": 9.38, "learning_rate": 5.163134613160633e-05, "loss": 1.2644, "step": 108270 }, { "epoch": 9.38, "learning_rate": 5.1629177749508495e-05, "loss": 1.1814, "step": 108280 }, { "epoch": 9.38, "learning_rate": 5.1627009367410655e-05, "loss": 1.2812, "step": 108290 }, { "epoch": 9.38, "learning_rate": 5.162484098531282e-05, "loss": 1.322, "step": 108300 }, { "epoch": 9.38, "learning_rate": 5.162267260321499e-05, "loss": 1.1612, "step": 108310 }, { "epoch": 9.38, "learning_rate": 5.162050422111715e-05, "loss": 1.273, "step": 108320 }, { "epoch": 9.38, "learning_rate": 5.1618335839019313e-05, "loss": 1.2835, "step": 108330 }, { "epoch": 9.38, "learning_rate": 5.1616167456921466e-05, "loss": 1.1989, "step": 108340 }, { "epoch": 9.38, "learning_rate": 5.161399907482363e-05, "loss": 1.1831, "step": 108350 }, { "epoch": 9.38, "learning_rate": 5.161183069272579e-05, "loss": 1.0745, "step": 108360 }, { "epoch": 9.39, "learning_rate": 5.160966231062796e-05, "loss": 1.3583, "step": 108370 }, { "epoch": 9.39, "learning_rate": 5.1607493928530125e-05, "loss": 1.2958, "step": 108380 }, { "epoch": 9.39, "learning_rate": 5.1605325546432285e-05, "loss": 1.0698, "step": 108390 }, { "epoch": 9.39, "learning_rate": 5.160315716433445e-05, "loss": 1.101, "step": 108400 }, { "epoch": 9.39, "learning_rate": 5.160098878223661e-05, "loss": 1.1109, "step": 108410 }, { "epoch": 9.39, "learning_rate": 5.159882040013877e-05, "loss": 1.062, "step": 108420 }, { "epoch": 9.39, "learning_rate": 5.159665201804093e-05, "loss": 0.9739, "step": 108430 }, { "epoch": 9.39, "learning_rate": 5.1594483635943096e-05, "loss": 1.2293, "step": 108440 }, { "epoch": 9.39, "learning_rate": 5.159231525384526e-05, "loss": 1.1545, "step": 108450 }, { "epoch": 9.39, "learning_rate": 5.159014687174742e-05, "loss": 1.3101, "step": 108460 }, { "epoch": 9.39, "learning_rate": 5.158797848964959e-05, "loss": 1.4933, "step": 108470 }, { "epoch": 9.4, "learning_rate": 5.158581010755175e-05, "loss": 1.1556, "step": 108480 }, { "epoch": 9.4, "learning_rate": 5.1583641725453915e-05, "loss": 1.1361, "step": 108490 }, { "epoch": 9.4, "learning_rate": 5.158147334335607e-05, "loss": 1.3127, "step": 108500 }, { "epoch": 9.4, "learning_rate": 5.1579304961258234e-05, "loss": 1.2837, "step": 108510 }, { "epoch": 9.4, "learning_rate": 5.1577136579160394e-05, "loss": 1.3598, "step": 108520 }, { "epoch": 9.4, "learning_rate": 5.157496819706256e-05, "loss": 1.423, "step": 108530 }, { "epoch": 9.4, "learning_rate": 5.1572799814964726e-05, "loss": 1.3659, "step": 108540 }, { "epoch": 9.4, "learning_rate": 5.1570631432866886e-05, "loss": 1.1934, "step": 108550 }, { "epoch": 9.4, "learning_rate": 5.156846305076905e-05, "loss": 1.1276, "step": 108560 }, { "epoch": 9.4, "learning_rate": 5.1566294668671205e-05, "loss": 1.1997, "step": 108570 }, { "epoch": 9.4, "learning_rate": 5.156412628657337e-05, "loss": 1.2568, "step": 108580 }, { "epoch": 9.4, "learning_rate": 5.156195790447553e-05, "loss": 1.4487, "step": 108590 }, { "epoch": 9.41, "learning_rate": 5.15597895223777e-05, "loss": 1.1314, "step": 108600 }, { "epoch": 9.41, "learning_rate": 5.1557621140279864e-05, "loss": 1.3349, "step": 108610 }, { "epoch": 9.41, "learning_rate": 5.1555452758182024e-05, "loss": 1.2621, "step": 108620 }, { "epoch": 9.41, "learning_rate": 5.155328437608419e-05, "loss": 1.2125, "step": 108630 }, { "epoch": 9.41, "learning_rate": 5.155111599398635e-05, "loss": 1.1753, "step": 108640 }, { "epoch": 9.41, "learning_rate": 5.154894761188851e-05, "loss": 1.4315, "step": 108650 }, { "epoch": 9.41, "learning_rate": 5.154677922979067e-05, "loss": 1.0843, "step": 108660 }, { "epoch": 9.41, "learning_rate": 5.1544610847692835e-05, "loss": 1.0412, "step": 108670 }, { "epoch": 9.41, "learning_rate": 5.1542442465595e-05, "loss": 1.2086, "step": 108680 }, { "epoch": 9.41, "learning_rate": 5.154027408349716e-05, "loss": 1.2497, "step": 108690 }, { "epoch": 9.41, "learning_rate": 5.153810570139933e-05, "loss": 1.2488, "step": 108700 }, { "epoch": 9.41, "learning_rate": 5.153593731930149e-05, "loss": 1.1128, "step": 108710 }, { "epoch": 9.42, "learning_rate": 5.1533768937203654e-05, "loss": 1.0319, "step": 108720 }, { "epoch": 9.42, "learning_rate": 5.1531600555105806e-05, "loss": 1.224, "step": 108730 }, { "epoch": 9.42, "learning_rate": 5.152943217300797e-05, "loss": 1.3075, "step": 108740 }, { "epoch": 9.42, "learning_rate": 5.152726379091014e-05, "loss": 1.2714, "step": 108750 }, { "epoch": 9.42, "learning_rate": 5.15250954088123e-05, "loss": 1.067, "step": 108760 }, { "epoch": 9.42, "learning_rate": 5.1522927026714465e-05, "loss": 1.1386, "step": 108770 }, { "epoch": 9.42, "learning_rate": 5.1520758644616625e-05, "loss": 1.1318, "step": 108780 }, { "epoch": 9.42, "learning_rate": 5.151859026251879e-05, "loss": 1.2604, "step": 108790 }, { "epoch": 9.42, "learning_rate": 5.1516421880420944e-05, "loss": 1.1315, "step": 108800 }, { "epoch": 9.42, "learning_rate": 5.151425349832311e-05, "loss": 1.2322, "step": 108810 }, { "epoch": 9.42, "learning_rate": 5.151208511622528e-05, "loss": 1.3487, "step": 108820 }, { "epoch": 9.43, "learning_rate": 5.1509916734127436e-05, "loss": 1.4187, "step": 108830 }, { "epoch": 9.43, "learning_rate": 5.15077483520296e-05, "loss": 1.3206, "step": 108840 }, { "epoch": 9.43, "learning_rate": 5.150557996993176e-05, "loss": 1.1388, "step": 108850 }, { "epoch": 9.43, "learning_rate": 5.150341158783393e-05, "loss": 1.0162, "step": 108860 }, { "epoch": 9.43, "learning_rate": 5.1501243205736095e-05, "loss": 1.2303, "step": 108870 }, { "epoch": 9.43, "learning_rate": 5.149907482363825e-05, "loss": 1.3448, "step": 108880 }, { "epoch": 9.43, "learning_rate": 5.1496906441540414e-05, "loss": 1.3354, "step": 108890 }, { "epoch": 9.43, "learning_rate": 5.1494738059442574e-05, "loss": 1.2699, "step": 108900 }, { "epoch": 9.43, "learning_rate": 5.149256967734474e-05, "loss": 1.246, "step": 108910 }, { "epoch": 9.43, "learning_rate": 5.14904012952469e-05, "loss": 1.1947, "step": 108920 }, { "epoch": 9.43, "learning_rate": 5.1488232913149066e-05, "loss": 1.2088, "step": 108930 }, { "epoch": 9.43, "learning_rate": 5.148606453105123e-05, "loss": 0.9511, "step": 108940 }, { "epoch": 9.44, "learning_rate": 5.148389614895339e-05, "loss": 1.1667, "step": 108950 }, { "epoch": 9.44, "learning_rate": 5.148172776685555e-05, "loss": 1.3108, "step": 108960 }, { "epoch": 9.44, "learning_rate": 5.147955938475771e-05, "loss": 1.2303, "step": 108970 }, { "epoch": 9.44, "learning_rate": 5.147739100265988e-05, "loss": 1.2751, "step": 108980 }, { "epoch": 9.44, "learning_rate": 5.147522262056204e-05, "loss": 1.2234, "step": 108990 }, { "epoch": 9.44, "learning_rate": 5.1473054238464204e-05, "loss": 1.3546, "step": 109000 }, { "epoch": 9.44, "learning_rate": 5.147088585636637e-05, "loss": 1.0479, "step": 109010 }, { "epoch": 9.44, "learning_rate": 5.146871747426853e-05, "loss": 0.9886, "step": 109020 }, { "epoch": 9.44, "learning_rate": 5.1466549092170697e-05, "loss": 1.331, "step": 109030 }, { "epoch": 9.44, "learning_rate": 5.146438071007285e-05, "loss": 1.1411, "step": 109040 }, { "epoch": 9.44, "learning_rate": 5.1462212327975016e-05, "loss": 1.2691, "step": 109050 }, { "epoch": 9.45, "learning_rate": 5.1460043945877175e-05, "loss": 1.2037, "step": 109060 }, { "epoch": 9.45, "learning_rate": 5.145787556377934e-05, "loss": 1.2927, "step": 109070 }, { "epoch": 9.45, "learning_rate": 5.14557071816815e-05, "loss": 1.3755, "step": 109080 }, { "epoch": 9.45, "learning_rate": 5.145353879958367e-05, "loss": 1.1354, "step": 109090 }, { "epoch": 9.45, "learning_rate": 5.1451370417485834e-05, "loss": 1.3034, "step": 109100 }, { "epoch": 9.45, "learning_rate": 5.144920203538799e-05, "loss": 1.1352, "step": 109110 }, { "epoch": 9.45, "learning_rate": 5.1447033653290153e-05, "loss": 1.17, "step": 109120 }, { "epoch": 9.45, "learning_rate": 5.144486527119231e-05, "loss": 1.2464, "step": 109130 }, { "epoch": 9.45, "learning_rate": 5.144269688909448e-05, "loss": 1.1274, "step": 109140 }, { "epoch": 9.45, "learning_rate": 5.144052850699664e-05, "loss": 1.2452, "step": 109150 }, { "epoch": 9.45, "learning_rate": 5.1438360124898805e-05, "loss": 1.3479, "step": 109160 }, { "epoch": 9.45, "learning_rate": 5.143619174280097e-05, "loss": 1.4391, "step": 109170 }, { "epoch": 9.46, "learning_rate": 5.143402336070313e-05, "loss": 1.2496, "step": 109180 }, { "epoch": 9.46, "learning_rate": 5.143185497860529e-05, "loss": 1.2893, "step": 109190 }, { "epoch": 9.46, "learning_rate": 5.142968659650745e-05, "loss": 1.254, "step": 109200 }, { "epoch": 9.46, "learning_rate": 5.142751821440962e-05, "loss": 1.3966, "step": 109210 }, { "epoch": 9.46, "learning_rate": 5.142534983231178e-05, "loss": 1.0216, "step": 109220 }, { "epoch": 9.46, "learning_rate": 5.142318145021394e-05, "loss": 1.1469, "step": 109230 }, { "epoch": 9.46, "learning_rate": 5.142101306811611e-05, "loss": 1.3286, "step": 109240 }, { "epoch": 9.46, "learning_rate": 5.141884468601827e-05, "loss": 1.0791, "step": 109250 }, { "epoch": 9.46, "learning_rate": 5.1416676303920435e-05, "loss": 1.215, "step": 109260 }, { "epoch": 9.46, "learning_rate": 5.141450792182259e-05, "loss": 1.5335, "step": 109270 }, { "epoch": 9.46, "learning_rate": 5.1412339539724755e-05, "loss": 1.2999, "step": 109280 }, { "epoch": 9.47, "learning_rate": 5.1410171157626914e-05, "loss": 1.3248, "step": 109290 }, { "epoch": 9.47, "learning_rate": 5.140800277552908e-05, "loss": 1.2397, "step": 109300 }, { "epoch": 9.47, "learning_rate": 5.140583439343125e-05, "loss": 1.0169, "step": 109310 }, { "epoch": 9.47, "learning_rate": 5.140366601133341e-05, "loss": 1.1458, "step": 109320 }, { "epoch": 9.47, "learning_rate": 5.140149762923557e-05, "loss": 1.3199, "step": 109330 }, { "epoch": 9.47, "learning_rate": 5.1399329247137726e-05, "loss": 1.1824, "step": 109340 }, { "epoch": 9.47, "learning_rate": 5.139716086503989e-05, "loss": 1.2873, "step": 109350 }, { "epoch": 9.47, "learning_rate": 5.139499248294205e-05, "loss": 1.4309, "step": 109360 }, { "epoch": 9.47, "learning_rate": 5.139282410084422e-05, "loss": 1.1449, "step": 109370 }, { "epoch": 9.47, "learning_rate": 5.1390655718746385e-05, "loss": 1.3164, "step": 109380 }, { "epoch": 9.47, "learning_rate": 5.1388487336648544e-05, "loss": 1.3003, "step": 109390 }, { "epoch": 9.47, "learning_rate": 5.138631895455071e-05, "loss": 1.1544, "step": 109400 }, { "epoch": 9.48, "learning_rate": 5.138415057245287e-05, "loss": 1.4769, "step": 109410 }, { "epoch": 9.48, "learning_rate": 5.138198219035503e-05, "loss": 1.0843, "step": 109420 }, { "epoch": 9.48, "learning_rate": 5.137981380825719e-05, "loss": 1.3572, "step": 109430 }, { "epoch": 9.48, "learning_rate": 5.1377645426159356e-05, "loss": 1.1557, "step": 109440 }, { "epoch": 9.48, "learning_rate": 5.137547704406152e-05, "loss": 1.1325, "step": 109450 }, { "epoch": 9.48, "learning_rate": 5.137330866196368e-05, "loss": 1.2636, "step": 109460 }, { "epoch": 9.48, "learning_rate": 5.137114027986585e-05, "loss": 1.3266, "step": 109470 }, { "epoch": 9.48, "learning_rate": 5.136897189776801e-05, "loss": 1.1333, "step": 109480 }, { "epoch": 9.48, "learning_rate": 5.1366803515670174e-05, "loss": 1.3102, "step": 109490 }, { "epoch": 9.48, "learning_rate": 5.136463513357233e-05, "loss": 1.1999, "step": 109500 }, { "epoch": 9.48, "learning_rate": 5.1362466751474494e-05, "loss": 1.2001, "step": 109510 }, { "epoch": 9.49, "learning_rate": 5.136029836937666e-05, "loss": 1.1705, "step": 109520 }, { "epoch": 9.49, "learning_rate": 5.135812998727882e-05, "loss": 1.2701, "step": 109530 }, { "epoch": 9.49, "learning_rate": 5.1355961605180986e-05, "loss": 1.2018, "step": 109540 }, { "epoch": 9.49, "learning_rate": 5.1353793223083146e-05, "loss": 1.5108, "step": 109550 }, { "epoch": 9.49, "learning_rate": 5.135162484098531e-05, "loss": 1.245, "step": 109560 }, { "epoch": 9.49, "learning_rate": 5.1349456458887465e-05, "loss": 1.2799, "step": 109570 }, { "epoch": 9.49, "learning_rate": 5.134728807678963e-05, "loss": 1.2922, "step": 109580 }, { "epoch": 9.49, "learning_rate": 5.134511969469179e-05, "loss": 1.2785, "step": 109590 }, { "epoch": 9.49, "learning_rate": 5.134295131259396e-05, "loss": 1.2293, "step": 109600 }, { "epoch": 9.49, "learning_rate": 5.1340782930496124e-05, "loss": 1.2339, "step": 109610 }, { "epoch": 9.49, "learning_rate": 5.133861454839828e-05, "loss": 1.1946, "step": 109620 }, { "epoch": 9.49, "learning_rate": 5.133644616630045e-05, "loss": 1.224, "step": 109630 }, { "epoch": 9.5, "learning_rate": 5.133427778420261e-05, "loss": 1.1807, "step": 109640 }, { "epoch": 9.5, "learning_rate": 5.133210940210477e-05, "loss": 1.3463, "step": 109650 }, { "epoch": 9.5, "learning_rate": 5.132994102000693e-05, "loss": 0.968, "step": 109660 }, { "epoch": 9.5, "learning_rate": 5.1327772637909095e-05, "loss": 1.3712, "step": 109670 }, { "epoch": 9.5, "learning_rate": 5.132560425581126e-05, "loss": 1.122, "step": 109680 }, { "epoch": 9.5, "learning_rate": 5.132343587371342e-05, "loss": 1.0574, "step": 109690 }, { "epoch": 9.5, "learning_rate": 5.132126749161559e-05, "loss": 1.2583, "step": 109700 }, { "epoch": 9.5, "learning_rate": 5.131909910951775e-05, "loss": 1.3674, "step": 109710 }, { "epoch": 9.5, "learning_rate": 5.131693072741991e-05, "loss": 0.9879, "step": 109720 }, { "epoch": 9.5, "learning_rate": 5.1314762345322066e-05, "loss": 1.1059, "step": 109730 }, { "epoch": 9.5, "learning_rate": 5.131259396322423e-05, "loss": 1.0269, "step": 109740 }, { "epoch": 9.51, "learning_rate": 5.13104255811264e-05, "loss": 1.3452, "step": 109750 }, { "epoch": 9.51, "learning_rate": 5.130825719902856e-05, "loss": 1.1255, "step": 109760 }, { "epoch": 9.51, "learning_rate": 5.1306088816930725e-05, "loss": 1.275, "step": 109770 }, { "epoch": 9.51, "learning_rate": 5.1303920434832885e-05, "loss": 1.0763, "step": 109780 }, { "epoch": 9.51, "learning_rate": 5.130175205273505e-05, "loss": 1.1004, "step": 109790 }, { "epoch": 9.51, "learning_rate": 5.129958367063722e-05, "loss": 1.1571, "step": 109800 }, { "epoch": 9.51, "learning_rate": 5.129741528853937e-05, "loss": 1.3599, "step": 109810 }, { "epoch": 9.51, "learning_rate": 5.1295246906441537e-05, "loss": 1.361, "step": 109820 }, { "epoch": 9.51, "learning_rate": 5.1293078524343696e-05, "loss": 1.2903, "step": 109830 }, { "epoch": 9.51, "learning_rate": 5.129091014224586e-05, "loss": 1.1732, "step": 109840 }, { "epoch": 9.51, "learning_rate": 5.128874176014802e-05, "loss": 1.3096, "step": 109850 }, { "epoch": 9.51, "learning_rate": 5.128657337805019e-05, "loss": 1.4629, "step": 109860 }, { "epoch": 9.52, "learning_rate": 5.1284404995952355e-05, "loss": 1.2531, "step": 109870 }, { "epoch": 9.52, "learning_rate": 5.128223661385451e-05, "loss": 1.1069, "step": 109880 }, { "epoch": 9.52, "learning_rate": 5.1280068231756674e-05, "loss": 1.3834, "step": 109890 }, { "epoch": 9.52, "learning_rate": 5.1277899849658834e-05, "loss": 1.2666, "step": 109900 }, { "epoch": 9.52, "learning_rate": 5.1275731467561e-05, "loss": 1.3289, "step": 109910 }, { "epoch": 9.52, "learning_rate": 5.127356308546316e-05, "loss": 1.1278, "step": 109920 }, { "epoch": 9.52, "learning_rate": 5.1271394703365326e-05, "loss": 1.1602, "step": 109930 }, { "epoch": 9.52, "learning_rate": 5.126922632126749e-05, "loss": 1.202, "step": 109940 }, { "epoch": 9.52, "learning_rate": 5.126705793916965e-05, "loss": 1.273, "step": 109950 }, { "epoch": 9.52, "learning_rate": 5.126488955707181e-05, "loss": 1.2345, "step": 109960 }, { "epoch": 9.52, "learning_rate": 5.126272117497397e-05, "loss": 1.2065, "step": 109970 }, { "epoch": 9.52, "learning_rate": 5.126055279287614e-05, "loss": 1.1962, "step": 109980 }, { "epoch": 9.53, "learning_rate": 5.12583844107783e-05, "loss": 1.2842, "step": 109990 }, { "epoch": 9.53, "learning_rate": 5.1256216028680464e-05, "loss": 1.3342, "step": 110000 }, { "epoch": 9.53, "learning_rate": 5.125404764658263e-05, "loss": 1.1859, "step": 110010 }, { "epoch": 9.53, "learning_rate": 5.125187926448479e-05, "loss": 1.2999, "step": 110020 }, { "epoch": 9.53, "learning_rate": 5.1249710882386956e-05, "loss": 1.332, "step": 110030 }, { "epoch": 9.53, "learning_rate": 5.124754250028911e-05, "loss": 1.2366, "step": 110040 }, { "epoch": 9.53, "learning_rate": 5.1245374118191275e-05, "loss": 1.4167, "step": 110050 }, { "epoch": 9.53, "learning_rate": 5.1243205736093435e-05, "loss": 1.2947, "step": 110060 }, { "epoch": 9.53, "learning_rate": 5.12410373539956e-05, "loss": 1.1099, "step": 110070 }, { "epoch": 9.53, "learning_rate": 5.123886897189777e-05, "loss": 1.2515, "step": 110080 }, { "epoch": 9.53, "learning_rate": 5.123670058979993e-05, "loss": 1.1281, "step": 110090 }, { "epoch": 9.54, "learning_rate": 5.1234532207702094e-05, "loss": 1.3319, "step": 110100 }, { "epoch": 9.54, "learning_rate": 5.123236382560425e-05, "loss": 1.2919, "step": 110110 }, { "epoch": 9.54, "learning_rate": 5.123019544350641e-05, "loss": 1.3753, "step": 110120 }, { "epoch": 9.54, "learning_rate": 5.122802706140857e-05, "loss": 1.0636, "step": 110130 }, { "epoch": 9.54, "learning_rate": 5.122585867931074e-05, "loss": 1.3226, "step": 110140 }, { "epoch": 9.54, "learning_rate": 5.12236902972129e-05, "loss": 1.1625, "step": 110150 }, { "epoch": 9.54, "learning_rate": 5.1221521915115065e-05, "loss": 1.3387, "step": 110160 }, { "epoch": 9.54, "learning_rate": 5.121935353301723e-05, "loss": 1.3669, "step": 110170 }, { "epoch": 9.54, "learning_rate": 5.121718515091939e-05, "loss": 1.1058, "step": 110180 }, { "epoch": 9.54, "learning_rate": 5.121501676882155e-05, "loss": 1.2288, "step": 110190 }, { "epoch": 9.54, "learning_rate": 5.121284838672371e-05, "loss": 1.2149, "step": 110200 }, { "epoch": 9.54, "learning_rate": 5.121068000462588e-05, "loss": 1.2915, "step": 110210 }, { "epoch": 9.55, "learning_rate": 5.1208511622528036e-05, "loss": 1.2549, "step": 110220 }, { "epoch": 9.55, "learning_rate": 5.12063432404302e-05, "loss": 1.1048, "step": 110230 }, { "epoch": 9.55, "learning_rate": 5.120417485833237e-05, "loss": 1.1477, "step": 110240 }, { "epoch": 9.55, "learning_rate": 5.120200647623453e-05, "loss": 1.1936, "step": 110250 }, { "epoch": 9.55, "learning_rate": 5.1199838094136695e-05, "loss": 1.365, "step": 110260 }, { "epoch": 9.55, "learning_rate": 5.119766971203885e-05, "loss": 1.305, "step": 110270 }, { "epoch": 9.55, "learning_rate": 5.1195501329941014e-05, "loss": 1.2071, "step": 110280 }, { "epoch": 9.55, "learning_rate": 5.1193332947843174e-05, "loss": 1.2517, "step": 110290 }, { "epoch": 9.55, "learning_rate": 5.119116456574534e-05, "loss": 1.2523, "step": 110300 }, { "epoch": 9.55, "learning_rate": 5.118899618364751e-05, "loss": 1.348, "step": 110310 }, { "epoch": 9.55, "learning_rate": 5.1186827801549666e-05, "loss": 1.3613, "step": 110320 }, { "epoch": 9.56, "learning_rate": 5.118465941945183e-05, "loss": 1.1104, "step": 110330 }, { "epoch": 9.56, "learning_rate": 5.1182491037353986e-05, "loss": 1.2352, "step": 110340 }, { "epoch": 9.56, "learning_rate": 5.118032265525615e-05, "loss": 1.2982, "step": 110350 }, { "epoch": 9.56, "learning_rate": 5.117815427315831e-05, "loss": 1.2211, "step": 110360 }, { "epoch": 9.56, "learning_rate": 5.117598589106048e-05, "loss": 1.1582, "step": 110370 }, { "epoch": 9.56, "learning_rate": 5.1173817508962644e-05, "loss": 1.0944, "step": 110380 }, { "epoch": 9.56, "learning_rate": 5.1171649126864804e-05, "loss": 1.2388, "step": 110390 }, { "epoch": 9.56, "learning_rate": 5.116948074476697e-05, "loss": 1.2001, "step": 110400 }, { "epoch": 9.56, "learning_rate": 5.116731236266913e-05, "loss": 1.1764, "step": 110410 }, { "epoch": 9.56, "learning_rate": 5.116514398057129e-05, "loss": 1.0965, "step": 110420 }, { "epoch": 9.56, "learning_rate": 5.116297559847345e-05, "loss": 1.049, "step": 110430 }, { "epoch": 9.56, "learning_rate": 5.1160807216375616e-05, "loss": 1.0484, "step": 110440 }, { "epoch": 9.57, "learning_rate": 5.115863883427778e-05, "loss": 1.2632, "step": 110450 }, { "epoch": 9.57, "learning_rate": 5.115647045217994e-05, "loss": 1.21, "step": 110460 }, { "epoch": 9.57, "learning_rate": 5.115430207008211e-05, "loss": 1.2078, "step": 110470 }, { "epoch": 9.57, "learning_rate": 5.115213368798427e-05, "loss": 1.3473, "step": 110480 }, { "epoch": 9.57, "learning_rate": 5.1149965305886434e-05, "loss": 1.0937, "step": 110490 }, { "epoch": 9.57, "learning_rate": 5.114779692378859e-05, "loss": 1.0005, "step": 110500 }, { "epoch": 9.57, "learning_rate": 5.114562854169075e-05, "loss": 1.3326, "step": 110510 }, { "epoch": 9.57, "learning_rate": 5.114346015959292e-05, "loss": 1.2804, "step": 110520 }, { "epoch": 9.57, "learning_rate": 5.114129177749508e-05, "loss": 1.1128, "step": 110530 }, { "epoch": 9.57, "learning_rate": 5.1139123395397246e-05, "loss": 1.3196, "step": 110540 }, { "epoch": 9.57, "learning_rate": 5.1136955013299405e-05, "loss": 1.3669, "step": 110550 }, { "epoch": 9.58, "learning_rate": 5.113478663120157e-05, "loss": 1.2606, "step": 110560 }, { "epoch": 9.58, "learning_rate": 5.113261824910374e-05, "loss": 1.1166, "step": 110570 }, { "epoch": 9.58, "learning_rate": 5.113044986700589e-05, "loss": 1.2844, "step": 110580 }, { "epoch": 9.58, "learning_rate": 5.112828148490806e-05, "loss": 1.2119, "step": 110590 }, { "epoch": 9.58, "learning_rate": 5.112611310281022e-05, "loss": 1.2205, "step": 110600 }, { "epoch": 9.58, "learning_rate": 5.112394472071238e-05, "loss": 1.3091, "step": 110610 }, { "epoch": 9.58, "learning_rate": 5.112177633861454e-05, "loss": 1.2355, "step": 110620 }, { "epoch": 9.58, "learning_rate": 5.111960795651671e-05, "loss": 1.2981, "step": 110630 }, { "epoch": 9.58, "learning_rate": 5.1117439574418876e-05, "loss": 1.3682, "step": 110640 }, { "epoch": 9.58, "learning_rate": 5.111527119232103e-05, "loss": 1.4179, "step": 110650 }, { "epoch": 9.58, "learning_rate": 5.111310281022319e-05, "loss": 1.358, "step": 110660 }, { "epoch": 9.58, "learning_rate": 5.1110934428125355e-05, "loss": 1.2152, "step": 110670 }, { "epoch": 9.59, "learning_rate": 5.110876604602752e-05, "loss": 1.3319, "step": 110680 }, { "epoch": 9.59, "learning_rate": 5.110659766392968e-05, "loss": 1.2527, "step": 110690 }, { "epoch": 9.59, "learning_rate": 5.110442928183185e-05, "loss": 1.4513, "step": 110700 }, { "epoch": 9.59, "learning_rate": 5.1102260899734007e-05, "loss": 1.1118, "step": 110710 }, { "epoch": 9.59, "learning_rate": 5.110009251763617e-05, "loss": 1.2221, "step": 110720 }, { "epoch": 9.59, "learning_rate": 5.1097924135538326e-05, "loss": 1.291, "step": 110730 }, { "epoch": 9.59, "learning_rate": 5.109575575344049e-05, "loss": 1.015, "step": 110740 }, { "epoch": 9.59, "learning_rate": 5.109358737134266e-05, "loss": 1.2004, "step": 110750 }, { "epoch": 9.59, "learning_rate": 5.109141898924482e-05, "loss": 1.3863, "step": 110760 }, { "epoch": 9.59, "learning_rate": 5.1089250607146985e-05, "loss": 1.1141, "step": 110770 }, { "epoch": 9.59, "learning_rate": 5.1087082225049144e-05, "loss": 0.9922, "step": 110780 }, { "epoch": 9.6, "learning_rate": 5.108491384295131e-05, "loss": 1.2904, "step": 110790 }, { "epoch": 9.6, "learning_rate": 5.108274546085348e-05, "loss": 1.2556, "step": 110800 }, { "epoch": 9.6, "learning_rate": 5.108057707875563e-05, "loss": 1.2087, "step": 110810 }, { "epoch": 9.6, "learning_rate": 5.1078408696657796e-05, "loss": 1.3301, "step": 110820 }, { "epoch": 9.6, "learning_rate": 5.1076240314559956e-05, "loss": 1.2709, "step": 110830 }, { "epoch": 9.6, "learning_rate": 5.107407193246212e-05, "loss": 1.2738, "step": 110840 }, { "epoch": 9.6, "learning_rate": 5.107190355036428e-05, "loss": 1.2615, "step": 110850 }, { "epoch": 9.6, "learning_rate": 5.106973516826645e-05, "loss": 1.4752, "step": 110860 }, { "epoch": 9.6, "learning_rate": 5.1067566786168615e-05, "loss": 1.2936, "step": 110870 }, { "epoch": 9.6, "learning_rate": 5.106539840407077e-05, "loss": 1.2957, "step": 110880 }, { "epoch": 9.6, "learning_rate": 5.1063230021972934e-05, "loss": 1.0483, "step": 110890 }, { "epoch": 9.6, "learning_rate": 5.1061061639875093e-05, "loss": 1.2644, "step": 110900 }, { "epoch": 9.61, "learning_rate": 5.105889325777726e-05, "loss": 1.3318, "step": 110910 }, { "epoch": 9.61, "learning_rate": 5.105672487567942e-05, "loss": 1.3952, "step": 110920 }, { "epoch": 9.61, "learning_rate": 5.1054556493581586e-05, "loss": 1.0823, "step": 110930 }, { "epoch": 9.61, "learning_rate": 5.105238811148375e-05, "loss": 1.1031, "step": 110940 }, { "epoch": 9.61, "learning_rate": 5.105021972938591e-05, "loss": 1.3901, "step": 110950 }, { "epoch": 9.61, "learning_rate": 5.104805134728807e-05, "loss": 1.1561, "step": 110960 }, { "epoch": 9.61, "learning_rate": 5.104588296519023e-05, "loss": 0.9958, "step": 110970 }, { "epoch": 9.61, "learning_rate": 5.10437145830924e-05, "loss": 1.2329, "step": 110980 }, { "epoch": 9.61, "learning_rate": 5.104154620099456e-05, "loss": 1.07, "step": 110990 }, { "epoch": 9.61, "learning_rate": 5.1039377818896724e-05, "loss": 1.3642, "step": 111000 }, { "epoch": 9.61, "learning_rate": 5.103720943679889e-05, "loss": 0.952, "step": 111010 }, { "epoch": 9.62, "learning_rate": 5.103504105470105e-05, "loss": 1.472, "step": 111020 }, { "epoch": 9.62, "learning_rate": 5.1032872672603216e-05, "loss": 1.4072, "step": 111030 }, { "epoch": 9.62, "learning_rate": 5.103070429050537e-05, "loss": 1.4353, "step": 111040 }, { "epoch": 9.62, "learning_rate": 5.1028535908407535e-05, "loss": 1.0732, "step": 111050 }, { "epoch": 9.62, "learning_rate": 5.1026367526309695e-05, "loss": 1.3827, "step": 111060 }, { "epoch": 9.62, "learning_rate": 5.102419914421186e-05, "loss": 1.2653, "step": 111070 }, { "epoch": 9.62, "learning_rate": 5.102203076211403e-05, "loss": 1.2119, "step": 111080 }, { "epoch": 9.62, "learning_rate": 5.101986238001619e-05, "loss": 1.2115, "step": 111090 }, { "epoch": 9.62, "learning_rate": 5.1017693997918354e-05, "loss": 1.307, "step": 111100 }, { "epoch": 9.62, "learning_rate": 5.1015525615820506e-05, "loss": 1.3389, "step": 111110 }, { "epoch": 9.62, "learning_rate": 5.101335723372267e-05, "loss": 1.2229, "step": 111120 }, { "epoch": 9.62, "learning_rate": 5.101118885162483e-05, "loss": 1.1412, "step": 111130 }, { "epoch": 9.63, "learning_rate": 5.1009020469527e-05, "loss": 1.2462, "step": 111140 }, { "epoch": 9.63, "learning_rate": 5.1006852087429165e-05, "loss": 1.3668, "step": 111150 }, { "epoch": 9.63, "learning_rate": 5.1004683705331325e-05, "loss": 1.2455, "step": 111160 }, { "epoch": 9.63, "learning_rate": 5.100251532323349e-05, "loss": 1.1641, "step": 111170 }, { "epoch": 9.63, "learning_rate": 5.100034694113565e-05, "loss": 1.2471, "step": 111180 }, { "epoch": 9.63, "learning_rate": 5.099817855903781e-05, "loss": 1.1707, "step": 111190 }, { "epoch": 9.63, "learning_rate": 5.099601017693997e-05, "loss": 1.2856, "step": 111200 }, { "epoch": 9.63, "learning_rate": 5.0993841794842136e-05, "loss": 1.1764, "step": 111210 }, { "epoch": 9.63, "learning_rate": 5.0991673412744296e-05, "loss": 1.3945, "step": 111220 }, { "epoch": 9.63, "learning_rate": 5.098950503064646e-05, "loss": 1.1367, "step": 111230 }, { "epoch": 9.63, "learning_rate": 5.098733664854863e-05, "loss": 1.1499, "step": 111240 }, { "epoch": 9.63, "learning_rate": 5.098516826645079e-05, "loss": 1.2248, "step": 111250 }, { "epoch": 9.64, "learning_rate": 5.0982999884352955e-05, "loss": 1.2529, "step": 111260 }, { "epoch": 9.64, "learning_rate": 5.098083150225511e-05, "loss": 1.2636, "step": 111270 }, { "epoch": 9.64, "learning_rate": 5.0978663120157274e-05, "loss": 1.2212, "step": 111280 }, { "epoch": 9.64, "learning_rate": 5.0976494738059434e-05, "loss": 1.1831, "step": 111290 }, { "epoch": 9.64, "learning_rate": 5.09743263559616e-05, "loss": 1.1633, "step": 111300 }, { "epoch": 9.64, "learning_rate": 5.0972157973863766e-05, "loss": 1.4038, "step": 111310 }, { "epoch": 9.64, "learning_rate": 5.0969989591765926e-05, "loss": 1.2901, "step": 111320 }, { "epoch": 9.64, "learning_rate": 5.096782120966809e-05, "loss": 1.3897, "step": 111330 }, { "epoch": 9.64, "learning_rate": 5.096565282757025e-05, "loss": 1.147, "step": 111340 }, { "epoch": 9.64, "learning_rate": 5.096348444547241e-05, "loss": 1.3574, "step": 111350 }, { "epoch": 9.64, "learning_rate": 5.096131606337457e-05, "loss": 1.2604, "step": 111360 }, { "epoch": 9.65, "learning_rate": 5.095914768127674e-05, "loss": 1.3466, "step": 111370 }, { "epoch": 9.65, "learning_rate": 5.0956979299178904e-05, "loss": 1.258, "step": 111380 }, { "epoch": 9.65, "learning_rate": 5.0954810917081064e-05, "loss": 1.2163, "step": 111390 }, { "epoch": 9.65, "learning_rate": 5.095264253498323e-05, "loss": 1.4113, "step": 111400 }, { "epoch": 9.65, "learning_rate": 5.095047415288539e-05, "loss": 1.2104, "step": 111410 }, { "epoch": 9.65, "learning_rate": 5.094830577078755e-05, "loss": 1.1787, "step": 111420 }, { "epoch": 9.65, "learning_rate": 5.094613738868971e-05, "loss": 1.2519, "step": 111430 }, { "epoch": 9.65, "learning_rate": 5.0943969006591875e-05, "loss": 1.3746, "step": 111440 }, { "epoch": 9.65, "learning_rate": 5.094180062449404e-05, "loss": 1.4178, "step": 111450 }, { "epoch": 9.65, "learning_rate": 5.09396322423962e-05, "loss": 1.2573, "step": 111460 }, { "epoch": 9.65, "learning_rate": 5.093746386029837e-05, "loss": 1.182, "step": 111470 }, { "epoch": 9.65, "learning_rate": 5.093529547820053e-05, "loss": 1.1233, "step": 111480 }, { "epoch": 9.66, "learning_rate": 5.0933127096102694e-05, "loss": 1.2306, "step": 111490 }, { "epoch": 9.66, "learning_rate": 5.0930958714004847e-05, "loss": 1.1544, "step": 111500 }, { "epoch": 9.66, "learning_rate": 5.092879033190701e-05, "loss": 1.4131, "step": 111510 }, { "epoch": 9.66, "learning_rate": 5.092662194980918e-05, "loss": 1.0495, "step": 111520 }, { "epoch": 9.66, "learning_rate": 5.092445356771134e-05, "loss": 1.3138, "step": 111530 }, { "epoch": 9.66, "learning_rate": 5.0922285185613505e-05, "loss": 1.2951, "step": 111540 }, { "epoch": 9.66, "learning_rate": 5.0920116803515665e-05, "loss": 1.2715, "step": 111550 }, { "epoch": 9.66, "learning_rate": 5.091794842141783e-05, "loss": 1.3883, "step": 111560 }, { "epoch": 9.66, "learning_rate": 5.091578003932e-05, "loss": 1.0626, "step": 111570 }, { "epoch": 9.66, "learning_rate": 5.091361165722215e-05, "loss": 1.0897, "step": 111580 }, { "epoch": 9.66, "learning_rate": 5.091144327512432e-05, "loss": 1.3996, "step": 111590 }, { "epoch": 9.67, "learning_rate": 5.090927489302648e-05, "loss": 1.1409, "step": 111600 }, { "epoch": 9.67, "learning_rate": 5.090710651092864e-05, "loss": 1.1463, "step": 111610 }, { "epoch": 9.67, "learning_rate": 5.09049381288308e-05, "loss": 1.2727, "step": 111620 }, { "epoch": 9.67, "learning_rate": 5.090276974673297e-05, "loss": 1.3921, "step": 111630 }, { "epoch": 9.67, "learning_rate": 5.0900601364635135e-05, "loss": 1.1889, "step": 111640 }, { "epoch": 9.67, "learning_rate": 5.089843298253729e-05, "loss": 1.317, "step": 111650 }, { "epoch": 9.67, "learning_rate": 5.0896264600439455e-05, "loss": 1.2993, "step": 111660 }, { "epoch": 9.67, "learning_rate": 5.0894096218341614e-05, "loss": 1.3862, "step": 111670 }, { "epoch": 9.67, "learning_rate": 5.089192783624378e-05, "loss": 1.188, "step": 111680 }, { "epoch": 9.67, "learning_rate": 5.088975945414594e-05, "loss": 1.1512, "step": 111690 }, { "epoch": 9.67, "learning_rate": 5.088759107204811e-05, "loss": 1.1588, "step": 111700 }, { "epoch": 9.67, "learning_rate": 5.088542268995027e-05, "loss": 1.3489, "step": 111710 }, { "epoch": 9.68, "learning_rate": 5.088325430785243e-05, "loss": 1.2485, "step": 111720 }, { "epoch": 9.68, "learning_rate": 5.0881085925754586e-05, "loss": 1.1676, "step": 111730 }, { "epoch": 9.68, "learning_rate": 5.087891754365675e-05, "loss": 1.4431, "step": 111740 }, { "epoch": 9.68, "learning_rate": 5.087674916155892e-05, "loss": 1.3038, "step": 111750 }, { "epoch": 9.68, "learning_rate": 5.087458077946108e-05, "loss": 1.2507, "step": 111760 }, { "epoch": 9.68, "learning_rate": 5.0872412397363244e-05, "loss": 1.1813, "step": 111770 }, { "epoch": 9.68, "learning_rate": 5.0870244015265404e-05, "loss": 1.1392, "step": 111780 }, { "epoch": 9.68, "learning_rate": 5.086807563316757e-05, "loss": 1.1259, "step": 111790 }, { "epoch": 9.68, "learning_rate": 5.086590725106974e-05, "loss": 1.3417, "step": 111800 }, { "epoch": 9.68, "learning_rate": 5.086373886897189e-05, "loss": 1.0973, "step": 111810 }, { "epoch": 9.68, "learning_rate": 5.0861570486874056e-05, "loss": 1.1479, "step": 111820 }, { "epoch": 9.69, "learning_rate": 5.0859402104776216e-05, "loss": 1.3029, "step": 111830 }, { "epoch": 9.69, "learning_rate": 5.085723372267838e-05, "loss": 1.0549, "step": 111840 }, { "epoch": 9.69, "learning_rate": 5.085506534058054e-05, "loss": 1.1724, "step": 111850 }, { "epoch": 9.69, "learning_rate": 5.085289695848271e-05, "loss": 1.1659, "step": 111860 }, { "epoch": 9.69, "learning_rate": 5.0850728576384874e-05, "loss": 1.3813, "step": 111870 }, { "epoch": 9.69, "learning_rate": 5.084856019428703e-05, "loss": 1.296, "step": 111880 }, { "epoch": 9.69, "learning_rate": 5.0846391812189194e-05, "loss": 1.2616, "step": 111890 }, { "epoch": 9.69, "learning_rate": 5.084422343009135e-05, "loss": 1.3989, "step": 111900 }, { "epoch": 9.69, "learning_rate": 5.084205504799352e-05, "loss": 1.294, "step": 111910 }, { "epoch": 9.69, "learning_rate": 5.083988666589568e-05, "loss": 1.1501, "step": 111920 }, { "epoch": 9.69, "learning_rate": 5.0837718283797846e-05, "loss": 1.2777, "step": 111930 }, { "epoch": 9.69, "learning_rate": 5.083554990170001e-05, "loss": 1.3308, "step": 111940 }, { "epoch": 9.7, "learning_rate": 5.083338151960217e-05, "loss": 1.2646, "step": 111950 }, { "epoch": 9.7, "learning_rate": 5.083121313750433e-05, "loss": 1.0894, "step": 111960 }, { "epoch": 9.7, "learning_rate": 5.082904475540649e-05, "loss": 1.1026, "step": 111970 }, { "epoch": 9.7, "learning_rate": 5.082687637330866e-05, "loss": 1.0682, "step": 111980 }, { "epoch": 9.7, "learning_rate": 5.082470799121082e-05, "loss": 1.223, "step": 111990 }, { "epoch": 9.7, "learning_rate": 5.082253960911298e-05, "loss": 1.1998, "step": 112000 }, { "epoch": 9.7, "learning_rate": 5.082037122701515e-05, "loss": 1.2514, "step": 112010 }, { "epoch": 9.7, "learning_rate": 5.081820284491731e-05, "loss": 1.2991, "step": 112020 }, { "epoch": 9.7, "learning_rate": 5.0816034462819476e-05, "loss": 1.2708, "step": 112030 }, { "epoch": 9.7, "learning_rate": 5.081386608072163e-05, "loss": 1.373, "step": 112040 }, { "epoch": 9.7, "learning_rate": 5.0811697698623795e-05, "loss": 1.3484, "step": 112050 }, { "epoch": 9.71, "learning_rate": 5.0809529316525954e-05, "loss": 1.1966, "step": 112060 }, { "epoch": 9.71, "learning_rate": 5.080736093442812e-05, "loss": 1.2476, "step": 112070 }, { "epoch": 9.71, "learning_rate": 5.080519255233029e-05, "loss": 1.4469, "step": 112080 }, { "epoch": 9.71, "learning_rate": 5.080302417023245e-05, "loss": 1.2987, "step": 112090 }, { "epoch": 9.71, "learning_rate": 5.080085578813461e-05, "loss": 1.1373, "step": 112100 }, { "epoch": 9.71, "learning_rate": 5.079868740603677e-05, "loss": 1.1772, "step": 112110 }, { "epoch": 9.71, "learning_rate": 5.079651902393893e-05, "loss": 1.5058, "step": 112120 }, { "epoch": 9.71, "learning_rate": 5.079435064184109e-05, "loss": 1.4763, "step": 112130 }, { "epoch": 9.71, "learning_rate": 5.079218225974326e-05, "loss": 1.5036, "step": 112140 }, { "epoch": 9.71, "learning_rate": 5.0790013877645425e-05, "loss": 1.079, "step": 112150 }, { "epoch": 9.71, "learning_rate": 5.0787845495547584e-05, "loss": 1.2193, "step": 112160 }, { "epoch": 9.71, "learning_rate": 5.078567711344975e-05, "loss": 1.3699, "step": 112170 }, { "epoch": 9.72, "learning_rate": 5.078350873135191e-05, "loss": 1.3763, "step": 112180 }, { "epoch": 9.72, "learning_rate": 5.078134034925407e-05, "loss": 1.3686, "step": 112190 }, { "epoch": 9.72, "learning_rate": 5.077917196715623e-05, "loss": 1.329, "step": 112200 }, { "epoch": 9.72, "learning_rate": 5.0777003585058396e-05, "loss": 1.4115, "step": 112210 }, { "epoch": 9.72, "learning_rate": 5.077483520296056e-05, "loss": 1.3257, "step": 112220 }, { "epoch": 9.72, "learning_rate": 5.077266682086272e-05, "loss": 1.1553, "step": 112230 }, { "epoch": 9.72, "learning_rate": 5.077049843876489e-05, "loss": 1.3223, "step": 112240 }, { "epoch": 9.72, "learning_rate": 5.076833005666705e-05, "loss": 1.2516, "step": 112250 }, { "epoch": 9.72, "learning_rate": 5.0766161674569215e-05, "loss": 1.4263, "step": 112260 }, { "epoch": 9.72, "learning_rate": 5.076399329247137e-05, "loss": 1.1407, "step": 112270 }, { "epoch": 9.72, "learning_rate": 5.0761824910373534e-05, "loss": 1.1204, "step": 112280 }, { "epoch": 9.73, "learning_rate": 5.075965652827569e-05, "loss": 1.0549, "step": 112290 }, { "epoch": 9.73, "learning_rate": 5.075748814617786e-05, "loss": 1.2275, "step": 112300 }, { "epoch": 9.73, "learning_rate": 5.0755319764080026e-05, "loss": 1.2802, "step": 112310 }, { "epoch": 9.73, "learning_rate": 5.0753151381982186e-05, "loss": 1.1066, "step": 112320 }, { "epoch": 9.73, "learning_rate": 5.075098299988435e-05, "loss": 1.1027, "step": 112330 }, { "epoch": 9.73, "learning_rate": 5.074881461778651e-05, "loss": 1.3478, "step": 112340 }, { "epoch": 9.73, "learning_rate": 5.074664623568867e-05, "loss": 1.1333, "step": 112350 }, { "epoch": 9.73, "learning_rate": 5.074447785359083e-05, "loss": 1.4237, "step": 112360 }, { "epoch": 9.73, "learning_rate": 5.0742309471493e-05, "loss": 1.2326, "step": 112370 }, { "epoch": 9.73, "learning_rate": 5.0740141089395164e-05, "loss": 1.0901, "step": 112380 }, { "epoch": 9.73, "learning_rate": 5.0737972707297323e-05, "loss": 1.2259, "step": 112390 }, { "epoch": 9.73, "learning_rate": 5.073580432519949e-05, "loss": 1.1738, "step": 112400 }, { "epoch": 9.74, "learning_rate": 5.073363594310165e-05, "loss": 1.2127, "step": 112410 }, { "epoch": 9.74, "learning_rate": 5.073146756100381e-05, "loss": 1.1445, "step": 112420 }, { "epoch": 9.74, "learning_rate": 5.072929917890597e-05, "loss": 1.0204, "step": 112430 }, { "epoch": 9.74, "learning_rate": 5.0727130796808135e-05, "loss": 1.2557, "step": 112440 }, { "epoch": 9.74, "learning_rate": 5.07249624147103e-05, "loss": 1.1058, "step": 112450 }, { "epoch": 9.74, "learning_rate": 5.072279403261246e-05, "loss": 1.1386, "step": 112460 }, { "epoch": 9.74, "learning_rate": 5.072062565051463e-05, "loss": 1.2078, "step": 112470 }, { "epoch": 9.74, "learning_rate": 5.071845726841679e-05, "loss": 1.478, "step": 112480 }, { "epoch": 9.74, "learning_rate": 5.0716288886318953e-05, "loss": 1.2472, "step": 112490 }, { "epoch": 9.74, "learning_rate": 5.0714120504221106e-05, "loss": 1.2424, "step": 112500 }, { "epoch": 9.74, "learning_rate": 5.071195212212327e-05, "loss": 1.3207, "step": 112510 }, { "epoch": 9.74, "learning_rate": 5.070978374002544e-05, "loss": 1.1823, "step": 112520 }, { "epoch": 9.75, "learning_rate": 5.07076153579276e-05, "loss": 1.2213, "step": 112530 }, { "epoch": 9.75, "learning_rate": 5.0705446975829765e-05, "loss": 1.4199, "step": 112540 }, { "epoch": 9.75, "learning_rate": 5.0703278593731925e-05, "loss": 1.2894, "step": 112550 }, { "epoch": 9.75, "learning_rate": 5.070111021163409e-05, "loss": 1.2624, "step": 112560 }, { "epoch": 9.75, "learning_rate": 5.069894182953626e-05, "loss": 1.3628, "step": 112570 }, { "epoch": 9.75, "learning_rate": 5.069677344743841e-05, "loss": 1.1473, "step": 112580 }, { "epoch": 9.75, "learning_rate": 5.069460506534058e-05, "loss": 1.2326, "step": 112590 }, { "epoch": 9.75, "learning_rate": 5.0692436683242736e-05, "loss": 1.2927, "step": 112600 }, { "epoch": 9.75, "learning_rate": 5.06902683011449e-05, "loss": 1.4629, "step": 112610 }, { "epoch": 9.75, "learning_rate": 5.068809991904706e-05, "loss": 1.3057, "step": 112620 }, { "epoch": 9.75, "learning_rate": 5.068593153694923e-05, "loss": 1.0364, "step": 112630 }, { "epoch": 9.76, "learning_rate": 5.0683763154851395e-05, "loss": 1.5768, "step": 112640 }, { "epoch": 9.76, "learning_rate": 5.068159477275355e-05, "loss": 1.4142, "step": 112650 }, { "epoch": 9.76, "learning_rate": 5.0679426390655714e-05, "loss": 1.1828, "step": 112660 }, { "epoch": 9.76, "learning_rate": 5.0677258008557874e-05, "loss": 1.169, "step": 112670 }, { "epoch": 9.76, "learning_rate": 5.067508962646004e-05, "loss": 1.2079, "step": 112680 }, { "epoch": 9.76, "learning_rate": 5.06729212443622e-05, "loss": 1.2227, "step": 112690 }, { "epoch": 9.76, "learning_rate": 5.0670752862264366e-05, "loss": 1.1936, "step": 112700 }, { "epoch": 9.76, "learning_rate": 5.066858448016653e-05, "loss": 1.4726, "step": 112710 }, { "epoch": 9.76, "learning_rate": 5.066641609806869e-05, "loss": 1.2113, "step": 112720 }, { "epoch": 9.76, "learning_rate": 5.066424771597085e-05, "loss": 1.3692, "step": 112730 }, { "epoch": 9.76, "learning_rate": 5.066207933387301e-05, "loss": 1.2424, "step": 112740 }, { "epoch": 9.76, "learning_rate": 5.065991095177518e-05, "loss": 1.171, "step": 112750 }, { "epoch": 9.77, "learning_rate": 5.065774256967734e-05, "loss": 1.3798, "step": 112760 }, { "epoch": 9.77, "learning_rate": 5.0655574187579504e-05, "loss": 1.0492, "step": 112770 }, { "epoch": 9.77, "learning_rate": 5.065340580548167e-05, "loss": 1.3108, "step": 112780 }, { "epoch": 9.77, "learning_rate": 5.065123742338383e-05, "loss": 1.2152, "step": 112790 }, { "epoch": 9.77, "learning_rate": 5.0649069041285996e-05, "loss": 1.0729, "step": 112800 }, { "epoch": 9.77, "learning_rate": 5.064690065918815e-05, "loss": 1.1403, "step": 112810 }, { "epoch": 9.77, "learning_rate": 5.0644732277090316e-05, "loss": 1.1899, "step": 112820 }, { "epoch": 9.77, "learning_rate": 5.0642563894992475e-05, "loss": 1.2311, "step": 112830 }, { "epoch": 9.77, "learning_rate": 5.064039551289464e-05, "loss": 1.1795, "step": 112840 }, { "epoch": 9.77, "learning_rate": 5.06382271307968e-05, "loss": 1.1552, "step": 112850 }, { "epoch": 9.77, "learning_rate": 5.063605874869897e-05, "loss": 1.2596, "step": 112860 }, { "epoch": 9.78, "learning_rate": 5.0633890366601134e-05, "loss": 1.2387, "step": 112870 }, { "epoch": 9.78, "learning_rate": 5.0631721984503294e-05, "loss": 1.07, "step": 112880 }, { "epoch": 9.78, "learning_rate": 5.062955360240545e-05, "loss": 0.9958, "step": 112890 }, { "epoch": 9.78, "learning_rate": 5.062738522030761e-05, "loss": 1.229, "step": 112900 }, { "epoch": 9.78, "learning_rate": 5.062521683820978e-05, "loss": 1.2254, "step": 112910 }, { "epoch": 9.78, "learning_rate": 5.062304845611194e-05, "loss": 1.232, "step": 112920 }, { "epoch": 9.78, "learning_rate": 5.0620880074014105e-05, "loss": 1.2459, "step": 112930 }, { "epoch": 9.78, "learning_rate": 5.061871169191627e-05, "loss": 1.2023, "step": 112940 }, { "epoch": 9.78, "learning_rate": 5.061654330981843e-05, "loss": 1.2876, "step": 112950 }, { "epoch": 9.78, "learning_rate": 5.061437492772059e-05, "loss": 1.4171, "step": 112960 }, { "epoch": 9.78, "learning_rate": 5.061220654562275e-05, "loss": 1.2252, "step": 112970 }, { "epoch": 9.78, "learning_rate": 5.061003816352492e-05, "loss": 1.4125, "step": 112980 }, { "epoch": 9.79, "learning_rate": 5.0607869781427077e-05, "loss": 1.473, "step": 112990 }, { "epoch": 9.79, "learning_rate": 5.060570139932924e-05, "loss": 1.4355, "step": 113000 }, { "epoch": 9.79, "learning_rate": 5.060353301723141e-05, "loss": 1.2584, "step": 113010 }, { "epoch": 9.79, "learning_rate": 5.060136463513357e-05, "loss": 1.2753, "step": 113020 }, { "epoch": 9.79, "learning_rate": 5.0599196253035735e-05, "loss": 1.4946, "step": 113030 }, { "epoch": 9.79, "learning_rate": 5.059702787093789e-05, "loss": 1.3305, "step": 113040 }, { "epoch": 9.79, "learning_rate": 5.0594859488840055e-05, "loss": 1.1108, "step": 113050 }, { "epoch": 9.79, "learning_rate": 5.0592691106742214e-05, "loss": 1.2648, "step": 113060 }, { "epoch": 9.79, "learning_rate": 5.059052272464438e-05, "loss": 1.3462, "step": 113070 }, { "epoch": 9.79, "learning_rate": 5.058835434254655e-05, "loss": 1.2241, "step": 113080 }, { "epoch": 9.79, "learning_rate": 5.0586185960448707e-05, "loss": 1.3515, "step": 113090 }, { "epoch": 9.8, "learning_rate": 5.058401757835087e-05, "loss": 1.291, "step": 113100 }, { "epoch": 9.8, "learning_rate": 5.058184919625303e-05, "loss": 1.3415, "step": 113110 }, { "epoch": 9.8, "learning_rate": 5.057968081415519e-05, "loss": 1.4401, "step": 113120 }, { "epoch": 9.8, "learning_rate": 5.057751243205735e-05, "loss": 1.1216, "step": 113130 }, { "epoch": 9.8, "learning_rate": 5.057534404995952e-05, "loss": 1.5707, "step": 113140 }, { "epoch": 9.8, "learning_rate": 5.0573175667861685e-05, "loss": 1.3692, "step": 113150 }, { "epoch": 9.8, "learning_rate": 5.0571007285763844e-05, "loss": 1.1835, "step": 113160 }, { "epoch": 9.8, "learning_rate": 5.056883890366601e-05, "loss": 1.1909, "step": 113170 }, { "epoch": 9.8, "learning_rate": 5.056667052156817e-05, "loss": 1.1506, "step": 113180 }, { "epoch": 9.8, "learning_rate": 5.056450213947033e-05, "loss": 1.2212, "step": 113190 }, { "epoch": 9.8, "learning_rate": 5.056233375737249e-05, "loss": 1.1566, "step": 113200 }, { "epoch": 9.8, "learning_rate": 5.0560165375274656e-05, "loss": 1.0313, "step": 113210 }, { "epoch": 9.81, "learning_rate": 5.055799699317682e-05, "loss": 1.0749, "step": 113220 }, { "epoch": 9.81, "learning_rate": 5.055582861107898e-05, "loss": 1.2146, "step": 113230 }, { "epoch": 9.81, "learning_rate": 5.055366022898115e-05, "loss": 1.4021, "step": 113240 }, { "epoch": 9.81, "learning_rate": 5.055149184688331e-05, "loss": 1.372, "step": 113250 }, { "epoch": 9.81, "learning_rate": 5.0549323464785474e-05, "loss": 1.1414, "step": 113260 }, { "epoch": 9.81, "learning_rate": 5.054715508268763e-05, "loss": 1.1562, "step": 113270 }, { "epoch": 9.81, "learning_rate": 5.0544986700589793e-05, "loss": 1.2682, "step": 113280 }, { "epoch": 9.81, "learning_rate": 5.054281831849196e-05, "loss": 1.1498, "step": 113290 }, { "epoch": 9.81, "learning_rate": 5.054064993639412e-05, "loss": 1.2445, "step": 113300 }, { "epoch": 9.81, "learning_rate": 5.0538481554296286e-05, "loss": 1.144, "step": 113310 }, { "epoch": 9.81, "learning_rate": 5.0536313172198445e-05, "loss": 1.2998, "step": 113320 }, { "epoch": 9.82, "learning_rate": 5.053414479010061e-05, "loss": 1.2165, "step": 113330 }, { "epoch": 9.82, "learning_rate": 5.053197640800278e-05, "loss": 1.2992, "step": 113340 }, { "epoch": 9.82, "learning_rate": 5.052980802590493e-05, "loss": 1.3344, "step": 113350 }, { "epoch": 9.82, "learning_rate": 5.052763964380709e-05, "loss": 1.2893, "step": 113360 }, { "epoch": 9.82, "learning_rate": 5.052547126170926e-05, "loss": 1.1531, "step": 113370 }, { "epoch": 9.82, "learning_rate": 5.0523302879611423e-05, "loss": 1.3408, "step": 113380 }, { "epoch": 9.82, "learning_rate": 5.052113449751358e-05, "loss": 1.1861, "step": 113390 }, { "epoch": 9.82, "learning_rate": 5.051896611541575e-05, "loss": 1.4794, "step": 113400 }, { "epoch": 9.82, "learning_rate": 5.051679773331791e-05, "loss": 1.2752, "step": 113410 }, { "epoch": 9.82, "learning_rate": 5.051462935122007e-05, "loss": 1.1588, "step": 113420 }, { "epoch": 9.82, "learning_rate": 5.051246096912223e-05, "loss": 1.3778, "step": 113430 }, { "epoch": 9.82, "learning_rate": 5.0510292587024395e-05, "loss": 1.2053, "step": 113440 }, { "epoch": 9.83, "learning_rate": 5.050812420492656e-05, "loss": 1.4297, "step": 113450 }, { "epoch": 9.83, "learning_rate": 5.050595582282872e-05, "loss": 1.1999, "step": 113460 }, { "epoch": 9.83, "learning_rate": 5.050378744073089e-05, "loss": 1.3029, "step": 113470 }, { "epoch": 9.83, "learning_rate": 5.050161905863305e-05, "loss": 1.3503, "step": 113480 }, { "epoch": 9.83, "learning_rate": 5.049945067653521e-05, "loss": 1.3728, "step": 113490 }, { "epoch": 9.83, "learning_rate": 5.0497282294437366e-05, "loss": 1.1163, "step": 113500 }, { "epoch": 9.83, "learning_rate": 5.049511391233953e-05, "loss": 1.1611, "step": 113510 }, { "epoch": 9.83, "learning_rate": 5.04929455302417e-05, "loss": 1.1462, "step": 113520 }, { "epoch": 9.83, "learning_rate": 5.049077714814386e-05, "loss": 1.1658, "step": 113530 }, { "epoch": 9.83, "learning_rate": 5.0488608766046025e-05, "loss": 1.341, "step": 113540 }, { "epoch": 9.83, "learning_rate": 5.0486440383948184e-05, "loss": 0.9977, "step": 113550 }, { "epoch": 9.84, "learning_rate": 5.048427200185035e-05, "loss": 1.3463, "step": 113560 }, { "epoch": 9.84, "learning_rate": 5.048210361975252e-05, "loss": 1.3994, "step": 113570 }, { "epoch": 9.84, "learning_rate": 5.047993523765467e-05, "loss": 1.3606, "step": 113580 }, { "epoch": 9.84, "learning_rate": 5.0477766855556836e-05, "loss": 1.2781, "step": 113590 }, { "epoch": 9.84, "learning_rate": 5.0475598473458996e-05, "loss": 1.0582, "step": 113600 }, { "epoch": 9.84, "learning_rate": 5.047343009136116e-05, "loss": 1.2812, "step": 113610 }, { "epoch": 9.84, "learning_rate": 5.047126170926332e-05, "loss": 1.2509, "step": 113620 }, { "epoch": 9.84, "learning_rate": 5.046909332716549e-05, "loss": 1.2176, "step": 113630 }, { "epoch": 9.84, "learning_rate": 5.0466924945067655e-05, "loss": 1.2312, "step": 113640 }, { "epoch": 9.84, "learning_rate": 5.0464756562969814e-05, "loss": 1.5477, "step": 113650 }, { "epoch": 9.84, "learning_rate": 5.0462588180871974e-05, "loss": 1.1571, "step": 113660 }, { "epoch": 9.84, "learning_rate": 5.0460419798774134e-05, "loss": 1.5623, "step": 113670 }, { "epoch": 9.85, "learning_rate": 5.04582514166763e-05, "loss": 1.2911, "step": 113680 }, { "epoch": 9.85, "learning_rate": 5.045608303457846e-05, "loss": 1.177, "step": 113690 }, { "epoch": 9.85, "learning_rate": 5.0453914652480626e-05, "loss": 1.3514, "step": 113700 }, { "epoch": 9.85, "learning_rate": 5.045174627038279e-05, "loss": 1.1975, "step": 113710 }, { "epoch": 9.85, "learning_rate": 5.044957788828495e-05, "loss": 1.0627, "step": 113720 }, { "epoch": 9.85, "learning_rate": 5.044740950618711e-05, "loss": 1.2441, "step": 113730 }, { "epoch": 9.85, "learning_rate": 5.044524112408927e-05, "loss": 1.1932, "step": 113740 }, { "epoch": 9.85, "learning_rate": 5.044307274199144e-05, "loss": 1.2342, "step": 113750 }, { "epoch": 9.85, "learning_rate": 5.04409043598936e-05, "loss": 1.1415, "step": 113760 }, { "epoch": 9.85, "learning_rate": 5.0438735977795764e-05, "loss": 1.3703, "step": 113770 }, { "epoch": 9.85, "learning_rate": 5.043656759569793e-05, "loss": 1.1173, "step": 113780 }, { "epoch": 9.85, "learning_rate": 5.043439921360009e-05, "loss": 1.1793, "step": 113790 }, { "epoch": 9.86, "learning_rate": 5.0432230831502256e-05, "loss": 1.2742, "step": 113800 }, { "epoch": 9.86, "learning_rate": 5.043006244940441e-05, "loss": 1.2687, "step": 113810 }, { "epoch": 9.86, "learning_rate": 5.0427894067306575e-05, "loss": 1.3231, "step": 113820 }, { "epoch": 9.86, "learning_rate": 5.0425725685208735e-05, "loss": 1.3787, "step": 113830 }, { "epoch": 9.86, "learning_rate": 5.04235573031109e-05, "loss": 1.2425, "step": 113840 }, { "epoch": 9.86, "learning_rate": 5.042138892101307e-05, "loss": 1.1147, "step": 113850 }, { "epoch": 9.86, "learning_rate": 5.041922053891523e-05, "loss": 1.334, "step": 113860 }, { "epoch": 9.86, "learning_rate": 5.0417052156817394e-05, "loss": 1.3946, "step": 113870 }, { "epoch": 9.86, "learning_rate": 5.041488377471955e-05, "loss": 1.6192, "step": 113880 }, { "epoch": 9.86, "learning_rate": 5.041271539262171e-05, "loss": 1.2386, "step": 113890 }, { "epoch": 9.86, "learning_rate": 5.041054701052387e-05, "loss": 1.4376, "step": 113900 }, { "epoch": 9.87, "learning_rate": 5.040837862842604e-05, "loss": 1.3568, "step": 113910 }, { "epoch": 9.87, "learning_rate": 5.04062102463282e-05, "loss": 1.284, "step": 113920 }, { "epoch": 9.87, "learning_rate": 5.0404041864230365e-05, "loss": 1.2337, "step": 113930 }, { "epoch": 9.87, "learning_rate": 5.040187348213253e-05, "loss": 1.3255, "step": 113940 }, { "epoch": 9.87, "learning_rate": 5.039970510003469e-05, "loss": 1.3626, "step": 113950 }, { "epoch": 9.87, "learning_rate": 5.039753671793685e-05, "loss": 1.2732, "step": 113960 }, { "epoch": 9.87, "learning_rate": 5.039536833583901e-05, "loss": 1.1505, "step": 113970 }, { "epoch": 9.87, "learning_rate": 5.0393199953741177e-05, "loss": 1.2279, "step": 113980 }, { "epoch": 9.87, "learning_rate": 5.0391031571643336e-05, "loss": 1.2827, "step": 113990 }, { "epoch": 9.87, "learning_rate": 5.03888631895455e-05, "loss": 1.3054, "step": 114000 }, { "epoch": 9.87, "learning_rate": 5.038669480744767e-05, "loss": 1.3892, "step": 114010 }, { "epoch": 9.87, "learning_rate": 5.038452642534983e-05, "loss": 1.1286, "step": 114020 }, { "epoch": 9.88, "learning_rate": 5.0382358043251995e-05, "loss": 1.197, "step": 114030 }, { "epoch": 9.88, "learning_rate": 5.038018966115415e-05, "loss": 1.5463, "step": 114040 }, { "epoch": 9.88, "learning_rate": 5.0378021279056314e-05, "loss": 1.047, "step": 114050 }, { "epoch": 9.88, "learning_rate": 5.0375852896958474e-05, "loss": 1.1274, "step": 114060 }, { "epoch": 9.88, "learning_rate": 5.037368451486064e-05, "loss": 1.1648, "step": 114070 }, { "epoch": 9.88, "learning_rate": 5.037151613276281e-05, "loss": 1.2586, "step": 114080 }, { "epoch": 9.88, "learning_rate": 5.0369347750664966e-05, "loss": 1.1203, "step": 114090 }, { "epoch": 9.88, "learning_rate": 5.036717936856713e-05, "loss": 1.328, "step": 114100 }, { "epoch": 9.88, "learning_rate": 5.036501098646929e-05, "loss": 1.3981, "step": 114110 }, { "epoch": 9.88, "learning_rate": 5.036284260437145e-05, "loss": 1.2607, "step": 114120 }, { "epoch": 9.88, "learning_rate": 5.036067422227361e-05, "loss": 1.2576, "step": 114130 }, { "epoch": 9.89, "learning_rate": 5.035850584017578e-05, "loss": 1.2026, "step": 114140 }, { "epoch": 9.89, "learning_rate": 5.0356337458077944e-05, "loss": 1.3705, "step": 114150 }, { "epoch": 9.89, "learning_rate": 5.0354169075980104e-05, "loss": 1.1648, "step": 114160 }, { "epoch": 9.89, "learning_rate": 5.035200069388227e-05, "loss": 1.299, "step": 114170 }, { "epoch": 9.89, "learning_rate": 5.034983231178443e-05, "loss": 1.1639, "step": 114180 }, { "epoch": 9.89, "learning_rate": 5.034766392968659e-05, "loss": 1.1576, "step": 114190 }, { "epoch": 9.89, "learning_rate": 5.034549554758875e-05, "loss": 1.027, "step": 114200 }, { "epoch": 9.89, "learning_rate": 5.0343327165490916e-05, "loss": 1.2113, "step": 114210 }, { "epoch": 9.89, "learning_rate": 5.034115878339308e-05, "loss": 1.2769, "step": 114220 }, { "epoch": 9.89, "learning_rate": 5.033899040129524e-05, "loss": 1.1144, "step": 114230 }, { "epoch": 9.89, "learning_rate": 5.033682201919741e-05, "loss": 1.2167, "step": 114240 }, { "epoch": 9.89, "learning_rate": 5.033465363709957e-05, "loss": 1.3942, "step": 114250 }, { "epoch": 9.9, "learning_rate": 5.0332485255001734e-05, "loss": 1.3392, "step": 114260 }, { "epoch": 9.9, "learning_rate": 5.033031687290389e-05, "loss": 1.4147, "step": 114270 }, { "epoch": 9.9, "learning_rate": 5.032814849080605e-05, "loss": 1.1136, "step": 114280 }, { "epoch": 9.9, "learning_rate": 5.032598010870822e-05, "loss": 1.1426, "step": 114290 }, { "epoch": 9.9, "learning_rate": 5.032381172661038e-05, "loss": 1.174, "step": 114300 }, { "epoch": 9.9, "learning_rate": 5.0321643344512546e-05, "loss": 1.3409, "step": 114310 }, { "epoch": 9.9, "learning_rate": 5.0319474962414705e-05, "loss": 1.132, "step": 114320 }, { "epoch": 9.9, "learning_rate": 5.031730658031687e-05, "loss": 1.2245, "step": 114330 }, { "epoch": 9.9, "learning_rate": 5.031513819821904e-05, "loss": 1.4037, "step": 114340 }, { "epoch": 9.9, "learning_rate": 5.031296981612119e-05, "loss": 1.2927, "step": 114350 }, { "epoch": 9.9, "learning_rate": 5.031080143402336e-05, "loss": 1.3203, "step": 114360 }, { "epoch": 9.91, "learning_rate": 5.030863305192552e-05, "loss": 1.1225, "step": 114370 }, { "epoch": 9.91, "learning_rate": 5.030646466982768e-05, "loss": 1.2891, "step": 114380 }, { "epoch": 9.91, "learning_rate": 5.030429628772984e-05, "loss": 1.0638, "step": 114390 }, { "epoch": 9.91, "learning_rate": 5.030212790563201e-05, "loss": 1.4395, "step": 114400 }, { "epoch": 9.91, "learning_rate": 5.0299959523534176e-05, "loss": 1.2412, "step": 114410 }, { "epoch": 9.91, "learning_rate": 5.0297791141436335e-05, "loss": 1.4225, "step": 114420 }, { "epoch": 9.91, "learning_rate": 5.029562275933849e-05, "loss": 1.0839, "step": 114430 }, { "epoch": 9.91, "learning_rate": 5.0293454377240654e-05, "loss": 1.401, "step": 114440 }, { "epoch": 9.91, "learning_rate": 5.029128599514282e-05, "loss": 1.2317, "step": 114450 }, { "epoch": 9.91, "learning_rate": 5.028911761304498e-05, "loss": 1.219, "step": 114460 }, { "epoch": 9.91, "learning_rate": 5.028694923094715e-05, "loss": 1.2467, "step": 114470 }, { "epoch": 9.91, "learning_rate": 5.0284780848849306e-05, "loss": 1.3027, "step": 114480 }, { "epoch": 9.92, "learning_rate": 5.028261246675147e-05, "loss": 1.1989, "step": 114490 }, { "epoch": 9.92, "learning_rate": 5.0280444084653626e-05, "loss": 1.1315, "step": 114500 }, { "epoch": 9.92, "learning_rate": 5.027827570255579e-05, "loss": 1.2675, "step": 114510 }, { "epoch": 9.92, "learning_rate": 5.027610732045796e-05, "loss": 1.1729, "step": 114520 }, { "epoch": 9.92, "learning_rate": 5.027393893836012e-05, "loss": 1.2369, "step": 114530 }, { "epoch": 9.92, "learning_rate": 5.0271770556262284e-05, "loss": 1.4746, "step": 114540 }, { "epoch": 9.92, "learning_rate": 5.0269602174164444e-05, "loss": 1.2236, "step": 114550 }, { "epoch": 9.92, "learning_rate": 5.026743379206661e-05, "loss": 1.2482, "step": 114560 }, { "epoch": 9.92, "learning_rate": 5.026526540996878e-05, "loss": 1.1732, "step": 114570 }, { "epoch": 9.92, "learning_rate": 5.026309702787093e-05, "loss": 1.3423, "step": 114580 }, { "epoch": 9.92, "learning_rate": 5.0260928645773096e-05, "loss": 1.4146, "step": 114590 }, { "epoch": 9.93, "learning_rate": 5.0258760263675256e-05, "loss": 1.2703, "step": 114600 }, { "epoch": 9.93, "learning_rate": 5.025659188157742e-05, "loss": 1.0958, "step": 114610 }, { "epoch": 9.93, "learning_rate": 5.025442349947958e-05, "loss": 1.4227, "step": 114620 }, { "epoch": 9.93, "learning_rate": 5.025225511738175e-05, "loss": 1.2448, "step": 114630 }, { "epoch": 9.93, "learning_rate": 5.0250086735283914e-05, "loss": 1.261, "step": 114640 }, { "epoch": 9.93, "learning_rate": 5.0247918353186074e-05, "loss": 1.2918, "step": 114650 }, { "epoch": 9.93, "learning_rate": 5.0245749971088234e-05, "loss": 1.3715, "step": 114660 }, { "epoch": 9.93, "learning_rate": 5.024358158899039e-05, "loss": 1.3492, "step": 114670 }, { "epoch": 9.93, "learning_rate": 5.024141320689256e-05, "loss": 1.1711, "step": 114680 }, { "epoch": 9.93, "learning_rate": 5.023924482479472e-05, "loss": 1.1632, "step": 114690 }, { "epoch": 9.93, "learning_rate": 5.0237076442696886e-05, "loss": 1.321, "step": 114700 }, { "epoch": 9.93, "learning_rate": 5.023490806059905e-05, "loss": 1.4231, "step": 114710 }, { "epoch": 9.94, "learning_rate": 5.023273967850121e-05, "loss": 1.1632, "step": 114720 }, { "epoch": 9.94, "learning_rate": 5.023057129640337e-05, "loss": 1.278, "step": 114730 }, { "epoch": 9.94, "learning_rate": 5.022840291430553e-05, "loss": 1.1571, "step": 114740 }, { "epoch": 9.94, "learning_rate": 5.02262345322077e-05, "loss": 1.4, "step": 114750 }, { "epoch": 9.94, "learning_rate": 5.022406615010986e-05, "loss": 1.308, "step": 114760 }, { "epoch": 9.94, "learning_rate": 5.022189776801202e-05, "loss": 1.1611, "step": 114770 }, { "epoch": 9.94, "learning_rate": 5.021972938591419e-05, "loss": 1.1657, "step": 114780 }, { "epoch": 9.94, "learning_rate": 5.021756100381635e-05, "loss": 1.2958, "step": 114790 }, { "epoch": 9.94, "learning_rate": 5.0215392621718516e-05, "loss": 1.3889, "step": 114800 }, { "epoch": 9.94, "learning_rate": 5.021322423962067e-05, "loss": 1.2234, "step": 114810 }, { "epoch": 9.94, "learning_rate": 5.0211055857522835e-05, "loss": 1.126, "step": 114820 }, { "epoch": 9.95, "learning_rate": 5.0208887475424995e-05, "loss": 1.1958, "step": 114830 }, { "epoch": 9.95, "learning_rate": 5.020671909332716e-05, "loss": 1.2915, "step": 114840 }, { "epoch": 9.95, "learning_rate": 5.020455071122933e-05, "loss": 1.1051, "step": 114850 }, { "epoch": 9.95, "learning_rate": 5.020238232913149e-05, "loss": 1.2726, "step": 114860 }, { "epoch": 9.95, "learning_rate": 5.0200213947033653e-05, "loss": 1.1481, "step": 114870 }, { "epoch": 9.95, "learning_rate": 5.019804556493581e-05, "loss": 1.4266, "step": 114880 }, { "epoch": 9.95, "learning_rate": 5.019587718283797e-05, "loss": 1.2526, "step": 114890 }, { "epoch": 9.95, "learning_rate": 5.019370880074013e-05, "loss": 1.2644, "step": 114900 }, { "epoch": 9.95, "learning_rate": 5.01915404186423e-05, "loss": 1.2701, "step": 114910 }, { "epoch": 9.95, "learning_rate": 5.0189372036544465e-05, "loss": 1.2265, "step": 114920 }, { "epoch": 9.95, "learning_rate": 5.0187203654446625e-05, "loss": 1.3156, "step": 114930 }, { "epoch": 9.95, "learning_rate": 5.018503527234879e-05, "loss": 1.196, "step": 114940 }, { "epoch": 9.96, "learning_rate": 5.018286689025095e-05, "loss": 1.1916, "step": 114950 }, { "epoch": 9.96, "learning_rate": 5.018069850815311e-05, "loss": 1.2963, "step": 114960 }, { "epoch": 9.96, "learning_rate": 5.017853012605527e-05, "loss": 1.0518, "step": 114970 }, { "epoch": 9.96, "learning_rate": 5.0176361743957436e-05, "loss": 1.3123, "step": 114980 }, { "epoch": 9.96, "learning_rate": 5.0174193361859596e-05, "loss": 1.3488, "step": 114990 }, { "epoch": 9.96, "learning_rate": 5.017202497976176e-05, "loss": 1.2426, "step": 115000 }, { "epoch": 9.96, "learning_rate": 5.016985659766393e-05, "loss": 1.3946, "step": 115010 }, { "epoch": 9.96, "learning_rate": 5.016768821556609e-05, "loss": 1.1628, "step": 115020 }, { "epoch": 9.96, "learning_rate": 5.0165519833468255e-05, "loss": 1.3001, "step": 115030 }, { "epoch": 9.96, "learning_rate": 5.016335145137041e-05, "loss": 1.2155, "step": 115040 }, { "epoch": 9.96, "learning_rate": 5.0161183069272574e-05, "loss": 1.1013, "step": 115050 }, { "epoch": 9.96, "learning_rate": 5.0159014687174734e-05, "loss": 1.3567, "step": 115060 }, { "epoch": 9.97, "learning_rate": 5.01568463050769e-05, "loss": 1.4039, "step": 115070 }, { "epoch": 9.97, "learning_rate": 5.0154677922979066e-05, "loss": 1.4076, "step": 115080 }, { "epoch": 9.97, "learning_rate": 5.0152509540881226e-05, "loss": 1.1227, "step": 115090 }, { "epoch": 9.97, "learning_rate": 5.015034115878339e-05, "loss": 1.1336, "step": 115100 }, { "epoch": 9.97, "learning_rate": 5.014817277668555e-05, "loss": 1.3221, "step": 115110 }, { "epoch": 9.97, "learning_rate": 5.014600439458771e-05, "loss": 1.1677, "step": 115120 }, { "epoch": 9.97, "learning_rate": 5.014383601248987e-05, "loss": 1.2885, "step": 115130 }, { "epoch": 9.97, "learning_rate": 5.014166763039204e-05, "loss": 1.1128, "step": 115140 }, { "epoch": 9.97, "learning_rate": 5.0139499248294204e-05, "loss": 1.3222, "step": 115150 }, { "epoch": 9.97, "learning_rate": 5.0137330866196364e-05, "loss": 1.2745, "step": 115160 }, { "epoch": 9.97, "learning_rate": 5.013516248409853e-05, "loss": 1.3411, "step": 115170 }, { "epoch": 9.98, "learning_rate": 5.013299410200069e-05, "loss": 1.1946, "step": 115180 }, { "epoch": 9.98, "learning_rate": 5.0130825719902856e-05, "loss": 1.4788, "step": 115190 }, { "epoch": 9.98, "learning_rate": 5.012865733780501e-05, "loss": 1.044, "step": 115200 }, { "epoch": 9.98, "learning_rate": 5.0126488955707175e-05, "loss": 0.9906, "step": 115210 }, { "epoch": 9.98, "learning_rate": 5.012432057360934e-05, "loss": 1.1923, "step": 115220 }, { "epoch": 9.98, "learning_rate": 5.01221521915115e-05, "loss": 1.1331, "step": 115230 }, { "epoch": 9.98, "learning_rate": 5.011998380941367e-05, "loss": 1.2533, "step": 115240 }, { "epoch": 9.98, "learning_rate": 5.011781542731583e-05, "loss": 1.2615, "step": 115250 }, { "epoch": 9.98, "learning_rate": 5.0115647045217994e-05, "loss": 1.2559, "step": 115260 }, { "epoch": 9.98, "learning_rate": 5.0113478663120146e-05, "loss": 1.253, "step": 115270 }, { "epoch": 9.98, "learning_rate": 5.011131028102231e-05, "loss": 1.2738, "step": 115280 }, { "epoch": 9.98, "learning_rate": 5.010914189892448e-05, "loss": 1.1534, "step": 115290 }, { "epoch": 9.99, "learning_rate": 5.010697351682664e-05, "loss": 1.1364, "step": 115300 }, { "epoch": 9.99, "learning_rate": 5.0104805134728805e-05, "loss": 1.2509, "step": 115310 }, { "epoch": 9.99, "learning_rate": 5.0102636752630965e-05, "loss": 1.2725, "step": 115320 }, { "epoch": 9.99, "learning_rate": 5.010046837053313e-05, "loss": 1.369, "step": 115330 }, { "epoch": 9.99, "learning_rate": 5.00982999884353e-05, "loss": 1.4176, "step": 115340 }, { "epoch": 9.99, "learning_rate": 5.009613160633745e-05, "loss": 1.4214, "step": 115350 }, { "epoch": 9.99, "learning_rate": 5.009396322423962e-05, "loss": 1.1662, "step": 115360 }, { "epoch": 9.99, "learning_rate": 5.0091794842141776e-05, "loss": 1.0841, "step": 115370 }, { "epoch": 9.99, "learning_rate": 5.008962646004394e-05, "loss": 1.2275, "step": 115380 }, { "epoch": 9.99, "learning_rate": 5.00874580779461e-05, "loss": 1.3089, "step": 115390 }, { "epoch": 9.99, "learning_rate": 5.008528969584827e-05, "loss": 1.2168, "step": 115400 }, { "epoch": 10.0, "learning_rate": 5.0083121313750435e-05, "loss": 1.3239, "step": 115410 }, { "epoch": 10.0, "learning_rate": 5.0080952931652595e-05, "loss": 1.1658, "step": 115420 }, { "epoch": 10.0, "learning_rate": 5.0078784549554755e-05, "loss": 1.213, "step": 115430 }, { "epoch": 10.0, "learning_rate": 5.0076616167456914e-05, "loss": 1.4098, "step": 115440 }, { "epoch": 10.0, "learning_rate": 5.007444778535908e-05, "loss": 1.1237, "step": 115450 }, { "epoch": 10.0, "learning_rate": 5.007227940326124e-05, "loss": 1.3111, "step": 115460 }, { "epoch": 10.0, "eval_cer": 0.5453142637020049, "eval_em": 0.033522434244455904, "eval_f1": 0.033522434244455904, "eval_loss": 1.4759306907653809, "eval_runtime": 244.0348, "eval_samples_per_second": 7.946, "eval_steps_per_second": 0.996, "eval_wer": 0.9664775657555441, "step": 115465 }, { "epoch": 10.0, "learning_rate": 5.0070111021163407e-05, "loss": 1.0579, "step": 115470 }, { "epoch": 10.0, "learning_rate": 5.006794263906557e-05, "loss": 1.0509, "step": 115480 }, { "epoch": 10.0, "learning_rate": 5.006577425696773e-05, "loss": 1.0926, "step": 115490 }, { "epoch": 10.0, "learning_rate": 5.0063605874869885e-05, "loss": 1.2692, "step": 115500 }, { "epoch": 10.0, "learning_rate": 5.006143749277205e-05, "loss": 1.0351, "step": 115510 }, { "epoch": 10.0, "learning_rate": 5.005926911067422e-05, "loss": 1.0527, "step": 115520 }, { "epoch": 10.01, "learning_rate": 5.005710072857638e-05, "loss": 1.0039, "step": 115530 }, { "epoch": 10.01, "learning_rate": 5.0054932346478544e-05, "loss": 1.0652, "step": 115540 }, { "epoch": 10.01, "learning_rate": 5.0052763964380704e-05, "loss": 1.2395, "step": 115550 }, { "epoch": 10.01, "learning_rate": 5.005059558228287e-05, "loss": 1.3103, "step": 115560 }, { "epoch": 10.01, "learning_rate": 5.0048427200185037e-05, "loss": 1.162, "step": 115570 }, { "epoch": 10.01, "learning_rate": 5.004625881808719e-05, "loss": 1.0634, "step": 115580 }, { "epoch": 10.01, "learning_rate": 5.0044090435989356e-05, "loss": 1.0798, "step": 115590 }, { "epoch": 10.01, "learning_rate": 5.0041922053891515e-05, "loss": 1.1374, "step": 115600 }, { "epoch": 10.01, "learning_rate": 5.003975367179368e-05, "loss": 1.1577, "step": 115610 }, { "epoch": 10.01, "learning_rate": 5.003758528969584e-05, "loss": 1.3065, "step": 115620 }, { "epoch": 10.01, "learning_rate": 5.003541690759801e-05, "loss": 1.1486, "step": 115630 }, { "epoch": 10.02, "learning_rate": 5.0033248525500174e-05, "loss": 1.0176, "step": 115640 }, { "epoch": 10.02, "learning_rate": 5.0031080143402334e-05, "loss": 1.0216, "step": 115650 }, { "epoch": 10.02, "learning_rate": 5.0028911761304493e-05, "loss": 1.2802, "step": 115660 }, { "epoch": 10.02, "learning_rate": 5.002674337920665e-05, "loss": 1.3395, "step": 115670 }, { "epoch": 10.02, "learning_rate": 5.002457499710882e-05, "loss": 1.2278, "step": 115680 }, { "epoch": 10.02, "learning_rate": 5.002240661501098e-05, "loss": 1.143, "step": 115690 }, { "epoch": 10.02, "learning_rate": 5.0020238232913145e-05, "loss": 1.1921, "step": 115700 }, { "epoch": 10.02, "learning_rate": 5.001806985081531e-05, "loss": 1.2287, "step": 115710 }, { "epoch": 10.02, "learning_rate": 5.001590146871747e-05, "loss": 1.1449, "step": 115720 }, { "epoch": 10.02, "learning_rate": 5.001373308661963e-05, "loss": 1.1879, "step": 115730 }, { "epoch": 10.02, "learning_rate": 5.001156470452179e-05, "loss": 1.0562, "step": 115740 }, { "epoch": 10.02, "learning_rate": 5.000939632242396e-05, "loss": 0.9865, "step": 115750 }, { "epoch": 10.03, "learning_rate": 5.000722794032612e-05, "loss": 1.25, "step": 115760 }, { "epoch": 10.03, "learning_rate": 5.000505955822828e-05, "loss": 1.1912, "step": 115770 }, { "epoch": 10.03, "learning_rate": 5.000289117613045e-05, "loss": 1.1684, "step": 115780 }, { "epoch": 10.03, "learning_rate": 5.000072279403261e-05, "loss": 1.3026, "step": 115790 }, { "epoch": 10.03, "learning_rate": 4.9998554411934775e-05, "loss": 1.0588, "step": 115800 }, { "epoch": 10.03, "learning_rate": 4.999638602983693e-05, "loss": 1.2424, "step": 115810 }, { "epoch": 10.03, "learning_rate": 4.9994217647739095e-05, "loss": 1.0407, "step": 115820 }, { "epoch": 10.03, "learning_rate": 4.9992049265641254e-05, "loss": 1.1681, "step": 115830 }, { "epoch": 10.03, "learning_rate": 4.998988088354342e-05, "loss": 1.0789, "step": 115840 }, { "epoch": 10.03, "learning_rate": 4.998771250144559e-05, "loss": 1.2012, "step": 115850 }, { "epoch": 10.03, "learning_rate": 4.998554411934775e-05, "loss": 0.9957, "step": 115860 }, { "epoch": 10.04, "learning_rate": 4.998337573724991e-05, "loss": 1.167, "step": 115870 }, { "epoch": 10.04, "learning_rate": 4.998120735515207e-05, "loss": 1.0972, "step": 115880 }, { "epoch": 10.04, "learning_rate": 4.997903897305423e-05, "loss": 1.2026, "step": 115890 }, { "epoch": 10.04, "learning_rate": 4.997687059095639e-05, "loss": 0.9211, "step": 115900 }, { "epoch": 10.04, "learning_rate": 4.997470220885856e-05, "loss": 1.2573, "step": 115910 }, { "epoch": 10.04, "learning_rate": 4.9972533826760725e-05, "loss": 1.0308, "step": 115920 }, { "epoch": 10.04, "learning_rate": 4.9970365444662884e-05, "loss": 1.1767, "step": 115930 }, { "epoch": 10.04, "learning_rate": 4.996819706256505e-05, "loss": 0.961, "step": 115940 }, { "epoch": 10.04, "learning_rate": 4.996602868046721e-05, "loss": 1.1519, "step": 115950 }, { "epoch": 10.04, "learning_rate": 4.996386029836938e-05, "loss": 1.1716, "step": 115960 }, { "epoch": 10.04, "learning_rate": 4.996169191627153e-05, "loss": 1.1511, "step": 115970 }, { "epoch": 10.04, "learning_rate": 4.9959523534173696e-05, "loss": 1.1902, "step": 115980 }, { "epoch": 10.05, "learning_rate": 4.995735515207586e-05, "loss": 1.1596, "step": 115990 }, { "epoch": 10.05, "learning_rate": 4.995518676997802e-05, "loss": 1.1108, "step": 116000 }, { "epoch": 10.05, "learning_rate": 4.995301838788019e-05, "loss": 1.0646, "step": 116010 }, { "epoch": 10.05, "learning_rate": 4.995085000578235e-05, "loss": 1.2328, "step": 116020 }, { "epoch": 10.05, "learning_rate": 4.9948681623684514e-05, "loss": 1.0494, "step": 116030 }, { "epoch": 10.05, "learning_rate": 4.994651324158667e-05, "loss": 1.1218, "step": 116040 }, { "epoch": 10.05, "learning_rate": 4.9944344859488834e-05, "loss": 1.0645, "step": 116050 }, { "epoch": 10.05, "learning_rate": 4.994217647739099e-05, "loss": 1.0996, "step": 116060 }, { "epoch": 10.05, "learning_rate": 4.994000809529316e-05, "loss": 1.0629, "step": 116070 }, { "epoch": 10.05, "learning_rate": 4.9937839713195326e-05, "loss": 1.1356, "step": 116080 }, { "epoch": 10.05, "learning_rate": 4.9935671331097486e-05, "loss": 1.1772, "step": 116090 }, { "epoch": 10.05, "learning_rate": 4.993350294899965e-05, "loss": 1.0798, "step": 116100 }, { "epoch": 10.06, "learning_rate": 4.993133456690181e-05, "loss": 1.0674, "step": 116110 }, { "epoch": 10.06, "learning_rate": 4.992916618480397e-05, "loss": 1.0466, "step": 116120 }, { "epoch": 10.06, "learning_rate": 4.992699780270613e-05, "loss": 1.3449, "step": 116130 }, { "epoch": 10.06, "learning_rate": 4.99248294206083e-05, "loss": 1.0338, "step": 116140 }, { "epoch": 10.06, "learning_rate": 4.9922661038510464e-05, "loss": 1.2406, "step": 116150 }, { "epoch": 10.06, "learning_rate": 4.992049265641262e-05, "loss": 1.1231, "step": 116160 }, { "epoch": 10.06, "learning_rate": 4.991832427431479e-05, "loss": 0.9583, "step": 116170 }, { "epoch": 10.06, "learning_rate": 4.991615589221695e-05, "loss": 1.0715, "step": 116180 }, { "epoch": 10.06, "learning_rate": 4.9913987510119116e-05, "loss": 1.1261, "step": 116190 }, { "epoch": 10.06, "learning_rate": 4.991181912802127e-05, "loss": 1.1656, "step": 116200 }, { "epoch": 10.06, "learning_rate": 4.9909650745923435e-05, "loss": 1.1558, "step": 116210 }, { "epoch": 10.07, "learning_rate": 4.99074823638256e-05, "loss": 1.153, "step": 116220 }, { "epoch": 10.07, "learning_rate": 4.990531398172776e-05, "loss": 1.1935, "step": 116230 }, { "epoch": 10.07, "learning_rate": 4.990314559962993e-05, "loss": 1.155, "step": 116240 }, { "epoch": 10.07, "learning_rate": 4.990097721753209e-05, "loss": 1.1791, "step": 116250 }, { "epoch": 10.07, "learning_rate": 4.989880883543425e-05, "loss": 1.1435, "step": 116260 }, { "epoch": 10.07, "learning_rate": 4.9896640453336406e-05, "loss": 1.259, "step": 116270 }, { "epoch": 10.07, "learning_rate": 4.989447207123857e-05, "loss": 1.2294, "step": 116280 }, { "epoch": 10.07, "learning_rate": 4.989230368914074e-05, "loss": 1.2135, "step": 116290 }, { "epoch": 10.07, "learning_rate": 4.98901353070429e-05, "loss": 1.1212, "step": 116300 }, { "epoch": 10.07, "learning_rate": 4.9887966924945065e-05, "loss": 1.1116, "step": 116310 }, { "epoch": 10.07, "learning_rate": 4.9885798542847225e-05, "loss": 1.3379, "step": 116320 }, { "epoch": 10.07, "learning_rate": 4.988363016074939e-05, "loss": 1.2323, "step": 116330 }, { "epoch": 10.08, "learning_rate": 4.988146177865156e-05, "loss": 1.1266, "step": 116340 }, { "epoch": 10.08, "learning_rate": 4.987929339655371e-05, "loss": 1.1513, "step": 116350 }, { "epoch": 10.08, "learning_rate": 4.9877125014455877e-05, "loss": 1.0947, "step": 116360 }, { "epoch": 10.08, "learning_rate": 4.9874956632358036e-05, "loss": 1.2574, "step": 116370 }, { "epoch": 10.08, "learning_rate": 4.98727882502602e-05, "loss": 1.1464, "step": 116380 }, { "epoch": 10.08, "learning_rate": 4.987061986816236e-05, "loss": 1.2487, "step": 116390 }, { "epoch": 10.08, "learning_rate": 4.986845148606453e-05, "loss": 1.1093, "step": 116400 }, { "epoch": 10.08, "learning_rate": 4.9866283103966695e-05, "loss": 1.1155, "step": 116410 }, { "epoch": 10.08, "learning_rate": 4.9864114721868855e-05, "loss": 1.2924, "step": 116420 }, { "epoch": 10.08, "learning_rate": 4.9861946339771014e-05, "loss": 1.1039, "step": 116430 }, { "epoch": 10.08, "learning_rate": 4.9859777957673174e-05, "loss": 1.1813, "step": 116440 }, { "epoch": 10.09, "learning_rate": 4.985760957557534e-05, "loss": 1.2009, "step": 116450 }, { "epoch": 10.09, "learning_rate": 4.98554411934775e-05, "loss": 1.1609, "step": 116460 }, { "epoch": 10.09, "learning_rate": 4.9853272811379666e-05, "loss": 1.2501, "step": 116470 }, { "epoch": 10.09, "learning_rate": 4.985110442928183e-05, "loss": 0.9932, "step": 116480 }, { "epoch": 10.09, "learning_rate": 4.984893604718399e-05, "loss": 1.1797, "step": 116490 }, { "epoch": 10.09, "learning_rate": 4.984676766508615e-05, "loss": 1.2083, "step": 116500 }, { "epoch": 10.09, "learning_rate": 4.984459928298831e-05, "loss": 1.2681, "step": 116510 }, { "epoch": 10.09, "learning_rate": 4.984243090089048e-05, "loss": 1.0442, "step": 116520 }, { "epoch": 10.09, "learning_rate": 4.984026251879264e-05, "loss": 1.0959, "step": 116530 }, { "epoch": 10.09, "learning_rate": 4.9838094136694804e-05, "loss": 1.2577, "step": 116540 }, { "epoch": 10.09, "learning_rate": 4.983592575459697e-05, "loss": 0.9423, "step": 116550 }, { "epoch": 10.09, "learning_rate": 4.983375737249913e-05, "loss": 1.0384, "step": 116560 }, { "epoch": 10.1, "learning_rate": 4.9831588990401296e-05, "loss": 1.0334, "step": 116570 }, { "epoch": 10.1, "learning_rate": 4.982942060830345e-05, "loss": 0.9859, "step": 116580 }, { "epoch": 10.1, "learning_rate": 4.9827252226205615e-05, "loss": 1.31, "step": 116590 }, { "epoch": 10.1, "learning_rate": 4.9825083844107775e-05, "loss": 1.0691, "step": 116600 }, { "epoch": 10.1, "learning_rate": 4.982291546200994e-05, "loss": 1.2401, "step": 116610 }, { "epoch": 10.1, "learning_rate": 4.98207470799121e-05, "loss": 1.1721, "step": 116620 }, { "epoch": 10.1, "learning_rate": 4.981857869781427e-05, "loss": 1.0524, "step": 116630 }, { "epoch": 10.1, "learning_rate": 4.9816410315716434e-05, "loss": 1.2933, "step": 116640 }, { "epoch": 10.1, "learning_rate": 4.9814241933618594e-05, "loss": 0.9727, "step": 116650 }, { "epoch": 10.1, "learning_rate": 4.981207355152075e-05, "loss": 1.193, "step": 116660 }, { "epoch": 10.1, "learning_rate": 4.980990516942291e-05, "loss": 1.2569, "step": 116670 }, { "epoch": 10.11, "learning_rate": 4.980773678732508e-05, "loss": 1.1323, "step": 116680 }, { "epoch": 10.11, "learning_rate": 4.980556840522724e-05, "loss": 1.3124, "step": 116690 }, { "epoch": 10.11, "learning_rate": 4.9803400023129405e-05, "loss": 1.2121, "step": 116700 }, { "epoch": 10.11, "learning_rate": 4.980123164103157e-05, "loss": 1.1341, "step": 116710 }, { "epoch": 10.11, "learning_rate": 4.979906325893373e-05, "loss": 1.2292, "step": 116720 }, { "epoch": 10.11, "learning_rate": 4.97968948768359e-05, "loss": 1.0439, "step": 116730 }, { "epoch": 10.11, "learning_rate": 4.979472649473805e-05, "loss": 1.2141, "step": 116740 }, { "epoch": 10.11, "learning_rate": 4.979255811264022e-05, "loss": 1.1714, "step": 116750 }, { "epoch": 10.11, "learning_rate": 4.9790389730542376e-05, "loss": 1.1147, "step": 116760 }, { "epoch": 10.11, "learning_rate": 4.978822134844454e-05, "loss": 1.4014, "step": 116770 }, { "epoch": 10.11, "learning_rate": 4.978605296634671e-05, "loss": 1.0644, "step": 116780 }, { "epoch": 10.11, "learning_rate": 4.978388458424887e-05, "loss": 1.4218, "step": 116790 }, { "epoch": 10.12, "learning_rate": 4.9781716202151035e-05, "loss": 1.3039, "step": 116800 }, { "epoch": 10.12, "learning_rate": 4.977954782005319e-05, "loss": 1.1675, "step": 116810 }, { "epoch": 10.12, "learning_rate": 4.9777379437955354e-05, "loss": 1.2097, "step": 116820 }, { "epoch": 10.12, "learning_rate": 4.9775211055857514e-05, "loss": 1.0528, "step": 116830 }, { "epoch": 10.12, "learning_rate": 4.977304267375968e-05, "loss": 1.3482, "step": 116840 }, { "epoch": 10.12, "learning_rate": 4.977087429166185e-05, "loss": 1.1857, "step": 116850 }, { "epoch": 10.12, "learning_rate": 4.9768705909564006e-05, "loss": 1.135, "step": 116860 }, { "epoch": 10.12, "learning_rate": 4.976653752746617e-05, "loss": 1.0146, "step": 116870 }, { "epoch": 10.12, "learning_rate": 4.976436914536833e-05, "loss": 1.1259, "step": 116880 }, { "epoch": 10.12, "learning_rate": 4.976220076327049e-05, "loss": 1.2045, "step": 116890 }, { "epoch": 10.12, "learning_rate": 4.976003238117265e-05, "loss": 1.0011, "step": 116900 }, { "epoch": 10.13, "learning_rate": 4.975786399907482e-05, "loss": 1.1903, "step": 116910 }, { "epoch": 10.13, "learning_rate": 4.9755695616976984e-05, "loss": 1.2415, "step": 116920 }, { "epoch": 10.13, "learning_rate": 4.9753527234879144e-05, "loss": 1.457, "step": 116930 }, { "epoch": 10.13, "learning_rate": 4.975135885278131e-05, "loss": 1.2454, "step": 116940 }, { "epoch": 10.13, "learning_rate": 4.974919047068347e-05, "loss": 1.1604, "step": 116950 }, { "epoch": 10.13, "learning_rate": 4.9747022088585636e-05, "loss": 1.1025, "step": 116960 }, { "epoch": 10.13, "learning_rate": 4.974485370648779e-05, "loss": 1.1785, "step": 116970 }, { "epoch": 10.13, "learning_rate": 4.9742685324389956e-05, "loss": 1.0785, "step": 116980 }, { "epoch": 10.13, "learning_rate": 4.974051694229212e-05, "loss": 1.2615, "step": 116990 }, { "epoch": 10.13, "learning_rate": 4.973834856019428e-05, "loss": 1.0864, "step": 117000 }, { "epoch": 10.13, "learning_rate": 4.973618017809645e-05, "loss": 1.2844, "step": 117010 }, { "epoch": 10.13, "learning_rate": 4.973401179599861e-05, "loss": 1.2796, "step": 117020 }, { "epoch": 10.14, "learning_rate": 4.9731843413900774e-05, "loss": 1.3041, "step": 117030 }, { "epoch": 10.14, "learning_rate": 4.972967503180293e-05, "loss": 1.1475, "step": 117040 }, { "epoch": 10.14, "learning_rate": 4.972750664970509e-05, "loss": 1.008, "step": 117050 }, { "epoch": 10.14, "learning_rate": 4.972533826760726e-05, "loss": 1.1479, "step": 117060 }, { "epoch": 10.14, "learning_rate": 4.972316988550942e-05, "loss": 1.1551, "step": 117070 }, { "epoch": 10.14, "learning_rate": 4.9721001503411586e-05, "loss": 1.1217, "step": 117080 }, { "epoch": 10.14, "learning_rate": 4.9718833121313745e-05, "loss": 1.2376, "step": 117090 }, { "epoch": 10.14, "learning_rate": 4.971666473921591e-05, "loss": 1.1521, "step": 117100 }, { "epoch": 10.14, "learning_rate": 4.971449635711808e-05, "loss": 1.187, "step": 117110 }, { "epoch": 10.14, "learning_rate": 4.971232797502023e-05, "loss": 1.0258, "step": 117120 }, { "epoch": 10.14, "learning_rate": 4.971015959292239e-05, "loss": 1.0783, "step": 117130 }, { "epoch": 10.15, "learning_rate": 4.970799121082456e-05, "loss": 1.2797, "step": 117140 }, { "epoch": 10.15, "learning_rate": 4.970582282872672e-05, "loss": 1.2278, "step": 117150 }, { "epoch": 10.15, "learning_rate": 4.970365444662888e-05, "loss": 1.0919, "step": 117160 }, { "epoch": 10.15, "learning_rate": 4.970148606453105e-05, "loss": 0.8538, "step": 117170 }, { "epoch": 10.15, "learning_rate": 4.969931768243321e-05, "loss": 1.204, "step": 117180 }, { "epoch": 10.15, "learning_rate": 4.9697149300335375e-05, "loss": 1.1772, "step": 117190 }, { "epoch": 10.15, "learning_rate": 4.969498091823753e-05, "loss": 1.1444, "step": 117200 }, { "epoch": 10.15, "learning_rate": 4.9692812536139695e-05, "loss": 1.1563, "step": 117210 }, { "epoch": 10.15, "learning_rate": 4.969064415404186e-05, "loss": 1.0811, "step": 117220 }, { "epoch": 10.15, "learning_rate": 4.968847577194402e-05, "loss": 0.9815, "step": 117230 }, { "epoch": 10.15, "learning_rate": 4.968630738984619e-05, "loss": 1.1213, "step": 117240 }, { "epoch": 10.15, "learning_rate": 4.9684139007748347e-05, "loss": 1.0141, "step": 117250 }, { "epoch": 10.16, "learning_rate": 4.968197062565051e-05, "loss": 1.2506, "step": 117260 }, { "epoch": 10.16, "learning_rate": 4.9679802243552666e-05, "loss": 1.091, "step": 117270 }, { "epoch": 10.16, "learning_rate": 4.967763386145483e-05, "loss": 1.1313, "step": 117280 }, { "epoch": 10.16, "learning_rate": 4.9675465479357e-05, "loss": 1.2946, "step": 117290 }, { "epoch": 10.16, "learning_rate": 4.967329709725916e-05, "loss": 0.9092, "step": 117300 }, { "epoch": 10.16, "learning_rate": 4.9671128715161325e-05, "loss": 1.3233, "step": 117310 }, { "epoch": 10.16, "learning_rate": 4.9668960333063484e-05, "loss": 1.1965, "step": 117320 }, { "epoch": 10.16, "learning_rate": 4.966679195096565e-05, "loss": 1.1868, "step": 117330 }, { "epoch": 10.16, "learning_rate": 4.966462356886782e-05, "loss": 1.0714, "step": 117340 }, { "epoch": 10.16, "learning_rate": 4.966245518676997e-05, "loss": 1.2838, "step": 117350 }, { "epoch": 10.16, "learning_rate": 4.9660286804672136e-05, "loss": 1.1047, "step": 117360 }, { "epoch": 10.16, "learning_rate": 4.9658118422574296e-05, "loss": 1.1022, "step": 117370 }, { "epoch": 10.17, "learning_rate": 4.965595004047646e-05, "loss": 1.0956, "step": 117380 }, { "epoch": 10.17, "learning_rate": 4.965378165837862e-05, "loss": 1.0206, "step": 117390 }, { "epoch": 10.17, "learning_rate": 4.965161327628079e-05, "loss": 1.0773, "step": 117400 }, { "epoch": 10.17, "learning_rate": 4.9649444894182955e-05, "loss": 1.1286, "step": 117410 }, { "epoch": 10.17, "learning_rate": 4.9647276512085114e-05, "loss": 1.2011, "step": 117420 }, { "epoch": 10.17, "learning_rate": 4.9645108129987274e-05, "loss": 0.9579, "step": 117430 }, { "epoch": 10.17, "learning_rate": 4.9642939747889434e-05, "loss": 1.1985, "step": 117440 }, { "epoch": 10.17, "learning_rate": 4.96407713657916e-05, "loss": 1.2212, "step": 117450 }, { "epoch": 10.17, "learning_rate": 4.963860298369376e-05, "loss": 1.1513, "step": 117460 }, { "epoch": 10.17, "learning_rate": 4.9636434601595926e-05, "loss": 1.001, "step": 117470 }, { "epoch": 10.17, "learning_rate": 4.963426621949809e-05, "loss": 1.2382, "step": 117480 }, { "epoch": 10.18, "learning_rate": 4.963209783740025e-05, "loss": 1.3675, "step": 117490 }, { "epoch": 10.18, "learning_rate": 4.962992945530242e-05, "loss": 1.255, "step": 117500 }, { "epoch": 10.18, "learning_rate": 4.962776107320457e-05, "loss": 1.3096, "step": 117510 }, { "epoch": 10.18, "learning_rate": 4.962559269110674e-05, "loss": 1.18, "step": 117520 }, { "epoch": 10.18, "learning_rate": 4.96234243090089e-05, "loss": 1.0535, "step": 117530 }, { "epoch": 10.18, "learning_rate": 4.9621255926911064e-05, "loss": 1.2363, "step": 117540 }, { "epoch": 10.18, "learning_rate": 4.961908754481323e-05, "loss": 1.1838, "step": 117550 }, { "epoch": 10.18, "learning_rate": 4.961691916271539e-05, "loss": 1.1564, "step": 117560 }, { "epoch": 10.18, "learning_rate": 4.9614750780617556e-05, "loss": 1.2803, "step": 117570 }, { "epoch": 10.18, "learning_rate": 4.961258239851971e-05, "loss": 1.1708, "step": 117580 }, { "epoch": 10.18, "learning_rate": 4.9610414016421875e-05, "loss": 1.2001, "step": 117590 }, { "epoch": 10.18, "learning_rate": 4.9608245634324035e-05, "loss": 1.0913, "step": 117600 }, { "epoch": 10.19, "learning_rate": 4.96060772522262e-05, "loss": 1.2784, "step": 117610 }, { "epoch": 10.19, "learning_rate": 4.960390887012837e-05, "loss": 1.0723, "step": 117620 }, { "epoch": 10.19, "learning_rate": 4.960174048803053e-05, "loss": 1.2331, "step": 117630 }, { "epoch": 10.19, "learning_rate": 4.9599572105932694e-05, "loss": 1.3981, "step": 117640 }, { "epoch": 10.19, "learning_rate": 4.959740372383485e-05, "loss": 1.0763, "step": 117650 }, { "epoch": 10.19, "learning_rate": 4.959523534173701e-05, "loss": 1.2247, "step": 117660 }, { "epoch": 10.19, "learning_rate": 4.959306695963917e-05, "loss": 1.2279, "step": 117670 }, { "epoch": 10.19, "learning_rate": 4.959089857754134e-05, "loss": 1.2552, "step": 117680 }, { "epoch": 10.19, "learning_rate": 4.95887301954435e-05, "loss": 1.1716, "step": 117690 }, { "epoch": 10.19, "learning_rate": 4.9586561813345665e-05, "loss": 1.2318, "step": 117700 }, { "epoch": 10.19, "learning_rate": 4.958439343124783e-05, "loss": 1.0761, "step": 117710 }, { "epoch": 10.2, "learning_rate": 4.958222504914999e-05, "loss": 1.3009, "step": 117720 }, { "epoch": 10.2, "learning_rate": 4.958005666705216e-05, "loss": 1.0171, "step": 117730 }, { "epoch": 10.2, "learning_rate": 4.957788828495431e-05, "loss": 1.2479, "step": 117740 }, { "epoch": 10.2, "learning_rate": 4.9575719902856476e-05, "loss": 1.1528, "step": 117750 }, { "epoch": 10.2, "learning_rate": 4.9573551520758636e-05, "loss": 1.1871, "step": 117760 }, { "epoch": 10.2, "learning_rate": 4.95713831386608e-05, "loss": 1.4257, "step": 117770 }, { "epoch": 10.2, "learning_rate": 4.956921475656297e-05, "loss": 1.2484, "step": 117780 }, { "epoch": 10.2, "learning_rate": 4.956704637446513e-05, "loss": 1.0546, "step": 117790 }, { "epoch": 10.2, "learning_rate": 4.9564877992367295e-05, "loss": 1.1906, "step": 117800 }, { "epoch": 10.2, "learning_rate": 4.956270961026945e-05, "loss": 1.1221, "step": 117810 }, { "epoch": 10.2, "learning_rate": 4.9560541228171614e-05, "loss": 1.2024, "step": 117820 }, { "epoch": 10.2, "learning_rate": 4.9558372846073774e-05, "loss": 1.1344, "step": 117830 }, { "epoch": 10.21, "learning_rate": 4.955620446397594e-05, "loss": 1.191, "step": 117840 }, { "epoch": 10.21, "learning_rate": 4.9554036081878106e-05, "loss": 1.271, "step": 117850 }, { "epoch": 10.21, "learning_rate": 4.9551867699780266e-05, "loss": 1.0869, "step": 117860 }, { "epoch": 10.21, "learning_rate": 4.954969931768243e-05, "loss": 1.4254, "step": 117870 }, { "epoch": 10.21, "learning_rate": 4.954753093558459e-05, "loss": 1.2509, "step": 117880 }, { "epoch": 10.21, "learning_rate": 4.954536255348675e-05, "loss": 0.9769, "step": 117890 }, { "epoch": 10.21, "learning_rate": 4.954319417138891e-05, "loss": 1.1078, "step": 117900 }, { "epoch": 10.21, "learning_rate": 4.954102578929108e-05, "loss": 1.156, "step": 117910 }, { "epoch": 10.21, "learning_rate": 4.9538857407193244e-05, "loss": 1.1277, "step": 117920 }, { "epoch": 10.21, "learning_rate": 4.9536689025095404e-05, "loss": 1.2278, "step": 117930 }, { "epoch": 10.21, "learning_rate": 4.953452064299757e-05, "loss": 1.1835, "step": 117940 }, { "epoch": 10.22, "learning_rate": 4.953235226089973e-05, "loss": 1.3363, "step": 117950 }, { "epoch": 10.22, "learning_rate": 4.9530183878801896e-05, "loss": 1.1379, "step": 117960 }, { "epoch": 10.22, "learning_rate": 4.952801549670405e-05, "loss": 1.196, "step": 117970 }, { "epoch": 10.22, "learning_rate": 4.9525847114606215e-05, "loss": 1.1873, "step": 117980 }, { "epoch": 10.22, "learning_rate": 4.952367873250838e-05, "loss": 1.3707, "step": 117990 }, { "epoch": 10.22, "learning_rate": 4.952151035041054e-05, "loss": 1.0627, "step": 118000 }, { "epoch": 10.22, "learning_rate": 4.951934196831271e-05, "loss": 1.2612, "step": 118010 }, { "epoch": 10.22, "learning_rate": 4.951717358621487e-05, "loss": 1.0108, "step": 118020 }, { "epoch": 10.22, "learning_rate": 4.9515005204117034e-05, "loss": 1.1229, "step": 118030 }, { "epoch": 10.22, "learning_rate": 4.951283682201919e-05, "loss": 1.2602, "step": 118040 }, { "epoch": 10.22, "learning_rate": 4.951066843992135e-05, "loss": 1.271, "step": 118050 }, { "epoch": 10.22, "learning_rate": 4.950850005782352e-05, "loss": 0.9956, "step": 118060 }, { "epoch": 10.23, "learning_rate": 4.950633167572568e-05, "loss": 1.0393, "step": 118070 }, { "epoch": 10.23, "learning_rate": 4.9504163293627845e-05, "loss": 1.0156, "step": 118080 }, { "epoch": 10.23, "learning_rate": 4.9501994911530005e-05, "loss": 1.3367, "step": 118090 }, { "epoch": 10.23, "learning_rate": 4.949982652943217e-05, "loss": 1.2084, "step": 118100 }, { "epoch": 10.23, "learning_rate": 4.949765814733434e-05, "loss": 1.1757, "step": 118110 }, { "epoch": 10.23, "learning_rate": 4.949548976523649e-05, "loss": 1.05, "step": 118120 }, { "epoch": 10.23, "learning_rate": 4.949332138313866e-05, "loss": 1.2943, "step": 118130 }, { "epoch": 10.23, "learning_rate": 4.949115300104082e-05, "loss": 1.2652, "step": 118140 }, { "epoch": 10.23, "learning_rate": 4.948898461894298e-05, "loss": 1.0832, "step": 118150 }, { "epoch": 10.23, "learning_rate": 4.948681623684514e-05, "loss": 1.2534, "step": 118160 }, { "epoch": 10.23, "learning_rate": 4.948464785474731e-05, "loss": 1.3342, "step": 118170 }, { "epoch": 10.24, "learning_rate": 4.9482479472649475e-05, "loss": 1.0391, "step": 118180 }, { "epoch": 10.24, "learning_rate": 4.9480311090551635e-05, "loss": 1.2626, "step": 118190 }, { "epoch": 10.24, "learning_rate": 4.947814270845379e-05, "loss": 1.2123, "step": 118200 }, { "epoch": 10.24, "learning_rate": 4.9475974326355954e-05, "loss": 1.3082, "step": 118210 }, { "epoch": 10.24, "learning_rate": 4.947380594425812e-05, "loss": 1.4337, "step": 118220 }, { "epoch": 10.24, "learning_rate": 4.947163756216028e-05, "loss": 1.1758, "step": 118230 }, { "epoch": 10.24, "learning_rate": 4.946946918006245e-05, "loss": 1.2777, "step": 118240 }, { "epoch": 10.24, "learning_rate": 4.9467300797964606e-05, "loss": 1.2863, "step": 118250 }, { "epoch": 10.24, "learning_rate": 4.946513241586677e-05, "loss": 1.2195, "step": 118260 }, { "epoch": 10.24, "learning_rate": 4.946296403376894e-05, "loss": 1.21, "step": 118270 }, { "epoch": 10.24, "learning_rate": 4.946079565167109e-05, "loss": 1.2731, "step": 118280 }, { "epoch": 10.24, "learning_rate": 4.945862726957326e-05, "loss": 1.2284, "step": 118290 }, { "epoch": 10.25, "learning_rate": 4.945645888747542e-05, "loss": 1.0186, "step": 118300 }, { "epoch": 10.25, "learning_rate": 4.9454290505377584e-05, "loss": 1.1173, "step": 118310 }, { "epoch": 10.25, "learning_rate": 4.9452122123279744e-05, "loss": 1.1501, "step": 118320 }, { "epoch": 10.25, "learning_rate": 4.944995374118191e-05, "loss": 1.2021, "step": 118330 }, { "epoch": 10.25, "learning_rate": 4.944778535908408e-05, "loss": 1.1233, "step": 118340 }, { "epoch": 10.25, "learning_rate": 4.944561697698623e-05, "loss": 1.3647, "step": 118350 }, { "epoch": 10.25, "learning_rate": 4.9443448594888396e-05, "loss": 1.1404, "step": 118360 }, { "epoch": 10.25, "learning_rate": 4.9441280212790556e-05, "loss": 1.3476, "step": 118370 }, { "epoch": 10.25, "learning_rate": 4.943911183069272e-05, "loss": 1.0632, "step": 118380 }, { "epoch": 10.25, "learning_rate": 4.943694344859488e-05, "loss": 1.2819, "step": 118390 }, { "epoch": 10.25, "learning_rate": 4.943477506649705e-05, "loss": 1.0791, "step": 118400 }, { "epoch": 10.26, "learning_rate": 4.9432606684399214e-05, "loss": 1.2101, "step": 118410 }, { "epoch": 10.26, "learning_rate": 4.9430438302301374e-05, "loss": 1.1802, "step": 118420 }, { "epoch": 10.26, "learning_rate": 4.9428269920203534e-05, "loss": 1.0111, "step": 118430 }, { "epoch": 10.26, "learning_rate": 4.942610153810569e-05, "loss": 1.1927, "step": 118440 }, { "epoch": 10.26, "learning_rate": 4.942393315600786e-05, "loss": 1.0302, "step": 118450 }, { "epoch": 10.26, "learning_rate": 4.942176477391002e-05, "loss": 1.1545, "step": 118460 }, { "epoch": 10.26, "learning_rate": 4.9419596391812186e-05, "loss": 1.1909, "step": 118470 }, { "epoch": 10.26, "learning_rate": 4.941742800971435e-05, "loss": 0.9939, "step": 118480 }, { "epoch": 10.26, "learning_rate": 4.941525962761651e-05, "loss": 1.067, "step": 118490 }, { "epoch": 10.26, "learning_rate": 4.941309124551868e-05, "loss": 1.1731, "step": 118500 }, { "epoch": 10.26, "learning_rate": 4.941092286342083e-05, "loss": 1.1943, "step": 118510 }, { "epoch": 10.26, "learning_rate": 4.9408754481323e-05, "loss": 1.143, "step": 118520 }, { "epoch": 10.27, "learning_rate": 4.940658609922516e-05, "loss": 1.3464, "step": 118530 }, { "epoch": 10.27, "learning_rate": 4.940441771712732e-05, "loss": 1.0645, "step": 118540 }, { "epoch": 10.27, "learning_rate": 4.940224933502949e-05, "loss": 1.1584, "step": 118550 }, { "epoch": 10.27, "learning_rate": 4.940008095293165e-05, "loss": 1.0822, "step": 118560 }, { "epoch": 10.27, "learning_rate": 4.9397912570833816e-05, "loss": 1.1923, "step": 118570 }, { "epoch": 10.27, "learning_rate": 4.939574418873597e-05, "loss": 1.1387, "step": 118580 }, { "epoch": 10.27, "learning_rate": 4.9393575806638135e-05, "loss": 1.1612, "step": 118590 }, { "epoch": 10.27, "learning_rate": 4.9391407424540294e-05, "loss": 1.3814, "step": 118600 }, { "epoch": 10.27, "learning_rate": 4.938923904244246e-05, "loss": 1.1377, "step": 118610 }, { "epoch": 10.27, "learning_rate": 4.938707066034463e-05, "loss": 1.2478, "step": 118620 }, { "epoch": 10.27, "learning_rate": 4.938490227824679e-05, "loss": 1.1115, "step": 118630 }, { "epoch": 10.27, "learning_rate": 4.938273389614895e-05, "loss": 1.258, "step": 118640 }, { "epoch": 10.28, "learning_rate": 4.938056551405111e-05, "loss": 1.075, "step": 118650 }, { "epoch": 10.28, "learning_rate": 4.937839713195327e-05, "loss": 1.4254, "step": 118660 }, { "epoch": 10.28, "learning_rate": 4.937622874985543e-05, "loss": 1.2325, "step": 118670 }, { "epoch": 10.28, "learning_rate": 4.93740603677576e-05, "loss": 1.1858, "step": 118680 }, { "epoch": 10.28, "learning_rate": 4.9371891985659765e-05, "loss": 1.3141, "step": 118690 }, { "epoch": 10.28, "learning_rate": 4.9369723603561925e-05, "loss": 1.0387, "step": 118700 }, { "epoch": 10.28, "learning_rate": 4.936755522146409e-05, "loss": 1.1523, "step": 118710 }, { "epoch": 10.28, "learning_rate": 4.936538683936625e-05, "loss": 1.3659, "step": 118720 }, { "epoch": 10.28, "learning_rate": 4.936321845726842e-05, "loss": 1.0904, "step": 118730 }, { "epoch": 10.28, "learning_rate": 4.936105007517057e-05, "loss": 1.1751, "step": 118740 }, { "epoch": 10.28, "learning_rate": 4.9358881693072736e-05, "loss": 1.2168, "step": 118750 }, { "epoch": 10.29, "learning_rate": 4.9356713310974896e-05, "loss": 1.2201, "step": 118760 }, { "epoch": 10.29, "learning_rate": 4.935454492887706e-05, "loss": 1.1271, "step": 118770 }, { "epoch": 10.29, "learning_rate": 4.935237654677923e-05, "loss": 1.3902, "step": 118780 }, { "epoch": 10.29, "learning_rate": 4.935020816468139e-05, "loss": 1.1385, "step": 118790 }, { "epoch": 10.29, "learning_rate": 4.9348039782583555e-05, "loss": 1.2593, "step": 118800 }, { "epoch": 10.29, "learning_rate": 4.934587140048571e-05, "loss": 0.9791, "step": 118810 }, { "epoch": 10.29, "learning_rate": 4.9343703018387874e-05, "loss": 1.1749, "step": 118820 }, { "epoch": 10.29, "learning_rate": 4.9341534636290033e-05, "loss": 1.0366, "step": 118830 }, { "epoch": 10.29, "learning_rate": 4.93393662541922e-05, "loss": 1.3082, "step": 118840 }, { "epoch": 10.29, "learning_rate": 4.9337197872094366e-05, "loss": 1.0725, "step": 118850 }, { "epoch": 10.29, "learning_rate": 4.9335029489996526e-05, "loss": 1.0615, "step": 118860 }, { "epoch": 10.29, "learning_rate": 4.933286110789869e-05, "loss": 1.0511, "step": 118870 }, { "epoch": 10.3, "learning_rate": 4.933069272580085e-05, "loss": 1.1688, "step": 118880 }, { "epoch": 10.3, "learning_rate": 4.932852434370301e-05, "loss": 1.1983, "step": 118890 }, { "epoch": 10.3, "learning_rate": 4.932635596160517e-05, "loss": 1.0779, "step": 118900 }, { "epoch": 10.3, "learning_rate": 4.932418757950734e-05, "loss": 1.2251, "step": 118910 }, { "epoch": 10.3, "learning_rate": 4.9322019197409504e-05, "loss": 1.3034, "step": 118920 }, { "epoch": 10.3, "learning_rate": 4.9319850815311663e-05, "loss": 1.127, "step": 118930 }, { "epoch": 10.3, "learning_rate": 4.931768243321383e-05, "loss": 1.1096, "step": 118940 }, { "epoch": 10.3, "learning_rate": 4.931551405111599e-05, "loss": 1.2515, "step": 118950 }, { "epoch": 10.3, "learning_rate": 4.9313345669018156e-05, "loss": 1.1276, "step": 118960 }, { "epoch": 10.3, "learning_rate": 4.931117728692031e-05, "loss": 1.201, "step": 118970 }, { "epoch": 10.3, "learning_rate": 4.9309008904822475e-05, "loss": 1.2764, "step": 118980 }, { "epoch": 10.31, "learning_rate": 4.930684052272464e-05, "loss": 1.2308, "step": 118990 }, { "epoch": 10.31, "learning_rate": 4.93046721406268e-05, "loss": 1.3045, "step": 119000 }, { "epoch": 10.31, "learning_rate": 4.930250375852897e-05, "loss": 1.2918, "step": 119010 }, { "epoch": 10.31, "learning_rate": 4.930033537643113e-05, "loss": 1.0174, "step": 119020 }, { "epoch": 10.31, "learning_rate": 4.9298166994333293e-05, "loss": 1.277, "step": 119030 }, { "epoch": 10.31, "learning_rate": 4.929599861223546e-05, "loss": 1.123, "step": 119040 }, { "epoch": 10.31, "learning_rate": 4.929383023013761e-05, "loss": 1.107, "step": 119050 }, { "epoch": 10.31, "learning_rate": 4.929166184803978e-05, "loss": 1.3786, "step": 119060 }, { "epoch": 10.31, "learning_rate": 4.928949346594194e-05, "loss": 1.1369, "step": 119070 }, { "epoch": 10.31, "learning_rate": 4.9287325083844105e-05, "loss": 1.3262, "step": 119080 }, { "epoch": 10.31, "learning_rate": 4.9285156701746265e-05, "loss": 1.0739, "step": 119090 }, { "epoch": 10.31, "learning_rate": 4.928298831964843e-05, "loss": 1.2474, "step": 119100 }, { "epoch": 10.32, "learning_rate": 4.92808199375506e-05, "loss": 1.1659, "step": 119110 }, { "epoch": 10.32, "learning_rate": 4.927865155545275e-05, "loss": 1.1076, "step": 119120 }, { "epoch": 10.32, "learning_rate": 4.927648317335492e-05, "loss": 1.0683, "step": 119130 }, { "epoch": 10.32, "learning_rate": 4.9274314791257076e-05, "loss": 1.134, "step": 119140 }, { "epoch": 10.32, "learning_rate": 4.927214640915924e-05, "loss": 1.1239, "step": 119150 }, { "epoch": 10.32, "learning_rate": 4.92699780270614e-05, "loss": 1.1936, "step": 119160 }, { "epoch": 10.32, "learning_rate": 4.926780964496357e-05, "loss": 1.2179, "step": 119170 }, { "epoch": 10.32, "learning_rate": 4.9265641262865735e-05, "loss": 1.0234, "step": 119180 }, { "epoch": 10.32, "learning_rate": 4.9263472880767895e-05, "loss": 1.1665, "step": 119190 }, { "epoch": 10.32, "learning_rate": 4.9261304498670054e-05, "loss": 1.2033, "step": 119200 }, { "epoch": 10.32, "learning_rate": 4.9259136116572214e-05, "loss": 1.3383, "step": 119210 }, { "epoch": 10.33, "learning_rate": 4.925696773447438e-05, "loss": 1.2182, "step": 119220 }, { "epoch": 10.33, "learning_rate": 4.925479935237654e-05, "loss": 1.1921, "step": 119230 }, { "epoch": 10.33, "learning_rate": 4.9252630970278706e-05, "loss": 1.1522, "step": 119240 }, { "epoch": 10.33, "learning_rate": 4.925046258818087e-05, "loss": 1.3334, "step": 119250 }, { "epoch": 10.33, "learning_rate": 4.924829420608303e-05, "loss": 1.084, "step": 119260 }, { "epoch": 10.33, "learning_rate": 4.92461258239852e-05, "loss": 1.3345, "step": 119270 }, { "epoch": 10.33, "learning_rate": 4.924395744188735e-05, "loss": 1.1897, "step": 119280 }, { "epoch": 10.33, "learning_rate": 4.924178905978952e-05, "loss": 1.16, "step": 119290 }, { "epoch": 10.33, "learning_rate": 4.923962067769168e-05, "loss": 1.0426, "step": 119300 }, { "epoch": 10.33, "learning_rate": 4.9237452295593844e-05, "loss": 1.0069, "step": 119310 }, { "epoch": 10.33, "learning_rate": 4.9235283913496004e-05, "loss": 1.0754, "step": 119320 }, { "epoch": 10.33, "learning_rate": 4.923311553139817e-05, "loss": 1.1285, "step": 119330 }, { "epoch": 10.34, "learning_rate": 4.9230947149300336e-05, "loss": 1.2733, "step": 119340 }, { "epoch": 10.34, "learning_rate": 4.922877876720249e-05, "loss": 1.0238, "step": 119350 }, { "epoch": 10.34, "learning_rate": 4.9226610385104656e-05, "loss": 1.115, "step": 119360 }, { "epoch": 10.34, "learning_rate": 4.9224442003006815e-05, "loss": 1.2755, "step": 119370 }, { "epoch": 10.34, "learning_rate": 4.922227362090898e-05, "loss": 1.0848, "step": 119380 }, { "epoch": 10.34, "learning_rate": 4.922010523881114e-05, "loss": 1.3226, "step": 119390 }, { "epoch": 10.34, "learning_rate": 4.921793685671331e-05, "loss": 1.2375, "step": 119400 }, { "epoch": 10.34, "learning_rate": 4.9215768474615474e-05, "loss": 1.2992, "step": 119410 }, { "epoch": 10.34, "learning_rate": 4.9213600092517634e-05, "loss": 1.3441, "step": 119420 }, { "epoch": 10.34, "learning_rate": 4.921143171041979e-05, "loss": 1.1953, "step": 119430 }, { "epoch": 10.34, "learning_rate": 4.920926332832195e-05, "loss": 1.2785, "step": 119440 }, { "epoch": 10.35, "learning_rate": 4.920709494622412e-05, "loss": 1.3094, "step": 119450 }, { "epoch": 10.35, "learning_rate": 4.920492656412628e-05, "loss": 1.093, "step": 119460 }, { "epoch": 10.35, "learning_rate": 4.9202758182028445e-05, "loss": 1.0693, "step": 119470 }, { "epoch": 10.35, "learning_rate": 4.920058979993061e-05, "loss": 1.1167, "step": 119480 }, { "epoch": 10.35, "learning_rate": 4.919842141783277e-05, "loss": 1.3479, "step": 119490 }, { "epoch": 10.35, "learning_rate": 4.919625303573494e-05, "loss": 1.1306, "step": 119500 }, { "epoch": 10.35, "learning_rate": 4.919408465363709e-05, "loss": 1.1217, "step": 119510 }, { "epoch": 10.35, "learning_rate": 4.919191627153926e-05, "loss": 1.2907, "step": 119520 }, { "epoch": 10.35, "learning_rate": 4.9189747889441417e-05, "loss": 1.0837, "step": 119530 }, { "epoch": 10.35, "learning_rate": 4.918757950734358e-05, "loss": 1.2375, "step": 119540 }, { "epoch": 10.35, "learning_rate": 4.918541112524575e-05, "loss": 1.2144, "step": 119550 }, { "epoch": 10.35, "learning_rate": 4.918324274314791e-05, "loss": 1.0816, "step": 119560 }, { "epoch": 10.36, "learning_rate": 4.9181074361050075e-05, "loss": 1.2632, "step": 119570 }, { "epoch": 10.36, "learning_rate": 4.917890597895223e-05, "loss": 1.0837, "step": 119580 }, { "epoch": 10.36, "learning_rate": 4.9176737596854395e-05, "loss": 1.086, "step": 119590 }, { "epoch": 10.36, "learning_rate": 4.9174569214756554e-05, "loss": 1.106, "step": 119600 }, { "epoch": 10.36, "learning_rate": 4.917240083265872e-05, "loss": 1.075, "step": 119610 }, { "epoch": 10.36, "learning_rate": 4.917023245056089e-05, "loss": 1.2561, "step": 119620 }, { "epoch": 10.36, "learning_rate": 4.9168064068463047e-05, "loss": 1.0714, "step": 119630 }, { "epoch": 10.36, "learning_rate": 4.916589568636521e-05, "loss": 1.0624, "step": 119640 }, { "epoch": 10.36, "learning_rate": 4.916372730426737e-05, "loss": 1.0929, "step": 119650 }, { "epoch": 10.36, "learning_rate": 4.916155892216953e-05, "loss": 1.0418, "step": 119660 }, { "epoch": 10.36, "learning_rate": 4.915939054007169e-05, "loss": 1.3408, "step": 119670 }, { "epoch": 10.37, "learning_rate": 4.915722215797386e-05, "loss": 1.227, "step": 119680 }, { "epoch": 10.37, "learning_rate": 4.9155053775876025e-05, "loss": 1.2332, "step": 119690 }, { "epoch": 10.37, "learning_rate": 4.9152885393778184e-05, "loss": 1.2402, "step": 119700 }, { "epoch": 10.37, "learning_rate": 4.915071701168035e-05, "loss": 1.2445, "step": 119710 }, { "epoch": 10.37, "learning_rate": 4.914854862958251e-05, "loss": 1.1835, "step": 119720 }, { "epoch": 10.37, "learning_rate": 4.914638024748468e-05, "loss": 1.0209, "step": 119730 }, { "epoch": 10.37, "learning_rate": 4.914421186538683e-05, "loss": 1.2258, "step": 119740 }, { "epoch": 10.37, "learning_rate": 4.9142043483288996e-05, "loss": 1.1692, "step": 119750 }, { "epoch": 10.37, "learning_rate": 4.913987510119116e-05, "loss": 1.3297, "step": 119760 }, { "epoch": 10.37, "learning_rate": 4.913770671909332e-05, "loss": 1.1294, "step": 119770 }, { "epoch": 10.37, "learning_rate": 4.913553833699549e-05, "loss": 1.278, "step": 119780 }, { "epoch": 10.37, "learning_rate": 4.913336995489765e-05, "loss": 1.0308, "step": 119790 }, { "epoch": 10.38, "learning_rate": 4.9131201572799814e-05, "loss": 1.0479, "step": 119800 }, { "epoch": 10.38, "learning_rate": 4.912903319070198e-05, "loss": 1.1466, "step": 119810 }, { "epoch": 10.38, "learning_rate": 4.9126864808604133e-05, "loss": 1.1414, "step": 119820 }, { "epoch": 10.38, "learning_rate": 4.912469642650629e-05, "loss": 1.2698, "step": 119830 }, { "epoch": 10.38, "learning_rate": 4.912252804440846e-05, "loss": 1.0265, "step": 119840 }, { "epoch": 10.38, "learning_rate": 4.9120359662310626e-05, "loss": 1.339, "step": 119850 }, { "epoch": 10.38, "learning_rate": 4.9118191280212785e-05, "loss": 1.1766, "step": 119860 }, { "epoch": 10.38, "learning_rate": 4.911602289811495e-05, "loss": 1.1357, "step": 119870 }, { "epoch": 10.38, "learning_rate": 4.911385451601711e-05, "loss": 1.0292, "step": 119880 }, { "epoch": 10.38, "learning_rate": 4.911168613391927e-05, "loss": 1.1248, "step": 119890 }, { "epoch": 10.38, "learning_rate": 4.910951775182143e-05, "loss": 1.1786, "step": 119900 }, { "epoch": 10.38, "learning_rate": 4.91073493697236e-05, "loss": 1.2275, "step": 119910 }, { "epoch": 10.39, "learning_rate": 4.9105180987625764e-05, "loss": 1.1635, "step": 119920 }, { "epoch": 10.39, "learning_rate": 4.910301260552792e-05, "loss": 0.9195, "step": 119930 }, { "epoch": 10.39, "learning_rate": 4.910084422343009e-05, "loss": 1.2245, "step": 119940 }, { "epoch": 10.39, "learning_rate": 4.909867584133225e-05, "loss": 1.0988, "step": 119950 }, { "epoch": 10.39, "learning_rate": 4.9096507459234416e-05, "loss": 1.1441, "step": 119960 }, { "epoch": 10.39, "learning_rate": 4.909433907713657e-05, "loss": 1.1649, "step": 119970 }, { "epoch": 10.39, "learning_rate": 4.9092170695038735e-05, "loss": 1.179, "step": 119980 }, { "epoch": 10.39, "learning_rate": 4.90900023129409e-05, "loss": 1.1258, "step": 119990 }, { "epoch": 10.39, "learning_rate": 4.908783393084306e-05, "loss": 1.2304, "step": 120000 }, { "epoch": 10.39, "learning_rate": 4.908566554874523e-05, "loss": 1.2709, "step": 120010 }, { "epoch": 10.39, "learning_rate": 4.908349716664739e-05, "loss": 0.9731, "step": 120020 }, { "epoch": 10.4, "learning_rate": 4.908132878454955e-05, "loss": 1.1615, "step": 120030 }, { "epoch": 10.4, "learning_rate": 4.907916040245172e-05, "loss": 0.9344, "step": 120040 }, { "epoch": 10.4, "learning_rate": 4.907699202035387e-05, "loss": 1.1306, "step": 120050 }, { "epoch": 10.4, "learning_rate": 4.907482363825604e-05, "loss": 1.2458, "step": 120060 }, { "epoch": 10.4, "learning_rate": 4.90726552561582e-05, "loss": 1.1227, "step": 120070 }, { "epoch": 10.4, "learning_rate": 4.9070486874060365e-05, "loss": 1.1256, "step": 120080 }, { "epoch": 10.4, "learning_rate": 4.9068318491962524e-05, "loss": 1.2919, "step": 120090 }, { "epoch": 10.4, "learning_rate": 4.906615010986469e-05, "loss": 1.178, "step": 120100 }, { "epoch": 10.4, "learning_rate": 4.906398172776686e-05, "loss": 1.1152, "step": 120110 }, { "epoch": 10.4, "learning_rate": 4.906181334566901e-05, "loss": 1.1884, "step": 120120 }, { "epoch": 10.4, "learning_rate": 4.9059644963571176e-05, "loss": 1.127, "step": 120130 }, { "epoch": 10.4, "learning_rate": 4.9057476581473336e-05, "loss": 1.3144, "step": 120140 }, { "epoch": 10.41, "learning_rate": 4.90553081993755e-05, "loss": 1.2822, "step": 120150 }, { "epoch": 10.41, "learning_rate": 4.905313981727766e-05, "loss": 1.0103, "step": 120160 }, { "epoch": 10.41, "learning_rate": 4.905097143517983e-05, "loss": 1.3494, "step": 120170 }, { "epoch": 10.41, "learning_rate": 4.9048803053081995e-05, "loss": 1.1099, "step": 120180 }, { "epoch": 10.41, "learning_rate": 4.9046634670984154e-05, "loss": 1.1391, "step": 120190 }, { "epoch": 10.41, "learning_rate": 4.9044466288886314e-05, "loss": 1.3455, "step": 120200 }, { "epoch": 10.41, "learning_rate": 4.9042297906788474e-05, "loss": 1.256, "step": 120210 }, { "epoch": 10.41, "learning_rate": 4.904012952469064e-05, "loss": 1.2542, "step": 120220 }, { "epoch": 10.41, "learning_rate": 4.90379611425928e-05, "loss": 1.1904, "step": 120230 }, { "epoch": 10.41, "learning_rate": 4.9035792760494966e-05, "loss": 0.9952, "step": 120240 }, { "epoch": 10.41, "learning_rate": 4.903362437839713e-05, "loss": 1.1654, "step": 120250 }, { "epoch": 10.42, "learning_rate": 4.903145599629929e-05, "loss": 1.0678, "step": 120260 }, { "epoch": 10.42, "learning_rate": 4.902928761420146e-05, "loss": 1.2306, "step": 120270 }, { "epoch": 10.42, "learning_rate": 4.902711923210361e-05, "loss": 1.1087, "step": 120280 }, { "epoch": 10.42, "learning_rate": 4.902495085000578e-05, "loss": 1.133, "step": 120290 }, { "epoch": 10.42, "learning_rate": 4.902278246790794e-05, "loss": 1.1424, "step": 120300 }, { "epoch": 10.42, "learning_rate": 4.9020614085810104e-05, "loss": 1.1903, "step": 120310 }, { "epoch": 10.42, "learning_rate": 4.901844570371227e-05, "loss": 1.0716, "step": 120320 }, { "epoch": 10.42, "learning_rate": 4.901627732161443e-05, "loss": 1.0724, "step": 120330 }, { "epoch": 10.42, "learning_rate": 4.9014108939516596e-05, "loss": 1.1925, "step": 120340 }, { "epoch": 10.42, "learning_rate": 4.901194055741875e-05, "loss": 1.216, "step": 120350 }, { "epoch": 10.42, "learning_rate": 4.9009772175320915e-05, "loss": 1.1619, "step": 120360 }, { "epoch": 10.42, "learning_rate": 4.9007603793223075e-05, "loss": 0.9621, "step": 120370 }, { "epoch": 10.43, "learning_rate": 4.900543541112524e-05, "loss": 1.0822, "step": 120380 }, { "epoch": 10.43, "learning_rate": 4.90032670290274e-05, "loss": 1.2079, "step": 120390 }, { "epoch": 10.43, "learning_rate": 4.900109864692957e-05, "loss": 0.9987, "step": 120400 }, { "epoch": 10.43, "learning_rate": 4.8998930264831734e-05, "loss": 1.1956, "step": 120410 }, { "epoch": 10.43, "learning_rate": 4.899676188273389e-05, "loss": 1.2803, "step": 120420 }, { "epoch": 10.43, "learning_rate": 4.899459350063605e-05, "loss": 1.2106, "step": 120430 }, { "epoch": 10.43, "learning_rate": 4.899242511853821e-05, "loss": 1.3174, "step": 120440 }, { "epoch": 10.43, "learning_rate": 4.899025673644038e-05, "loss": 1.2379, "step": 120450 }, { "epoch": 10.43, "learning_rate": 4.898808835434254e-05, "loss": 1.2284, "step": 120460 }, { "epoch": 10.43, "learning_rate": 4.8985919972244705e-05, "loss": 1.1727, "step": 120470 }, { "epoch": 10.43, "learning_rate": 4.898375159014687e-05, "loss": 1.0066, "step": 120480 }, { "epoch": 10.44, "learning_rate": 4.898158320804903e-05, "loss": 1.1323, "step": 120490 }, { "epoch": 10.44, "learning_rate": 4.89794148259512e-05, "loss": 1.2177, "step": 120500 }, { "epoch": 10.44, "learning_rate": 4.897724644385335e-05, "loss": 1.2168, "step": 120510 }, { "epoch": 10.44, "learning_rate": 4.897507806175552e-05, "loss": 1.1905, "step": 120520 }, { "epoch": 10.44, "learning_rate": 4.8972909679657676e-05, "loss": 1.1944, "step": 120530 }, { "epoch": 10.44, "learning_rate": 4.897074129755984e-05, "loss": 1.2605, "step": 120540 }, { "epoch": 10.44, "learning_rate": 4.896857291546201e-05, "loss": 1.1685, "step": 120550 }, { "epoch": 10.44, "learning_rate": 4.896640453336417e-05, "loss": 1.0359, "step": 120560 }, { "epoch": 10.44, "learning_rate": 4.8964236151266335e-05, "loss": 1.2032, "step": 120570 }, { "epoch": 10.44, "learning_rate": 4.8962067769168495e-05, "loss": 1.2266, "step": 120580 }, { "epoch": 10.44, "learning_rate": 4.8959899387070654e-05, "loss": 1.348, "step": 120590 }, { "epoch": 10.44, "learning_rate": 4.8957731004972814e-05, "loss": 1.2311, "step": 120600 }, { "epoch": 10.45, "learning_rate": 4.895556262287498e-05, "loss": 1.1972, "step": 120610 }, { "epoch": 10.45, "learning_rate": 4.895339424077715e-05, "loss": 1.242, "step": 120620 }, { "epoch": 10.45, "learning_rate": 4.8951225858679306e-05, "loss": 1.1044, "step": 120630 }, { "epoch": 10.45, "learning_rate": 4.894905747658147e-05, "loss": 1.3062, "step": 120640 }, { "epoch": 10.45, "learning_rate": 4.894688909448363e-05, "loss": 1.1155, "step": 120650 }, { "epoch": 10.45, "learning_rate": 4.894472071238579e-05, "loss": 1.2392, "step": 120660 }, { "epoch": 10.45, "learning_rate": 4.894255233028795e-05, "loss": 1.2094, "step": 120670 }, { "epoch": 10.45, "learning_rate": 4.894038394819012e-05, "loss": 1.122, "step": 120680 }, { "epoch": 10.45, "learning_rate": 4.8938215566092284e-05, "loss": 1.2148, "step": 120690 }, { "epoch": 10.45, "learning_rate": 4.8936047183994444e-05, "loss": 1.1586, "step": 120700 }, { "epoch": 10.45, "learning_rate": 4.893387880189661e-05, "loss": 1.267, "step": 120710 }, { "epoch": 10.46, "learning_rate": 4.893171041979877e-05, "loss": 1.2809, "step": 120720 }, { "epoch": 10.46, "learning_rate": 4.8929542037700936e-05, "loss": 1.1423, "step": 120730 }, { "epoch": 10.46, "learning_rate": 4.892737365560309e-05, "loss": 1.0316, "step": 120740 }, { "epoch": 10.46, "learning_rate": 4.8925205273505256e-05, "loss": 1.4723, "step": 120750 }, { "epoch": 10.46, "learning_rate": 4.892303689140742e-05, "loss": 1.0969, "step": 120760 }, { "epoch": 10.46, "learning_rate": 4.892086850930958e-05, "loss": 1.1046, "step": 120770 }, { "epoch": 10.46, "learning_rate": 4.891870012721175e-05, "loss": 1.0797, "step": 120780 }, { "epoch": 10.46, "learning_rate": 4.891653174511391e-05, "loss": 1.1592, "step": 120790 }, { "epoch": 10.46, "learning_rate": 4.8914363363016074e-05, "loss": 1.251, "step": 120800 }, { "epoch": 10.46, "learning_rate": 4.891219498091824e-05, "loss": 1.362, "step": 120810 }, { "epoch": 10.46, "learning_rate": 4.891002659882039e-05, "loss": 1.025, "step": 120820 }, { "epoch": 10.46, "learning_rate": 4.890785821672256e-05, "loss": 1.2914, "step": 120830 }, { "epoch": 10.47, "learning_rate": 4.890568983462472e-05, "loss": 1.0572, "step": 120840 }, { "epoch": 10.47, "learning_rate": 4.8903521452526886e-05, "loss": 1.1581, "step": 120850 }, { "epoch": 10.47, "learning_rate": 4.8901353070429045e-05, "loss": 1.4162, "step": 120860 }, { "epoch": 10.47, "learning_rate": 4.889918468833121e-05, "loss": 1.1021, "step": 120870 }, { "epoch": 10.47, "learning_rate": 4.889701630623338e-05, "loss": 1.254, "step": 120880 }, { "epoch": 10.47, "learning_rate": 4.889484792413553e-05, "loss": 1.0447, "step": 120890 }, { "epoch": 10.47, "learning_rate": 4.889267954203769e-05, "loss": 1.2303, "step": 120900 }, { "epoch": 10.47, "learning_rate": 4.889051115993986e-05, "loss": 1.1602, "step": 120910 }, { "epoch": 10.47, "learning_rate": 4.888834277784202e-05, "loss": 1.0934, "step": 120920 }, { "epoch": 10.47, "learning_rate": 4.888617439574418e-05, "loss": 1.1453, "step": 120930 }, { "epoch": 10.47, "learning_rate": 4.888400601364635e-05, "loss": 1.0914, "step": 120940 }, { "epoch": 10.48, "learning_rate": 4.888183763154851e-05, "loss": 1.1901, "step": 120950 }, { "epoch": 10.48, "learning_rate": 4.8879669249450675e-05, "loss": 0.9315, "step": 120960 }, { "epoch": 10.48, "learning_rate": 4.887750086735283e-05, "loss": 1.3794, "step": 120970 }, { "epoch": 10.48, "learning_rate": 4.8875332485254994e-05, "loss": 1.3372, "step": 120980 }, { "epoch": 10.48, "learning_rate": 4.887316410315716e-05, "loss": 1.0848, "step": 120990 }, { "epoch": 10.48, "learning_rate": 4.887099572105932e-05, "loss": 1.1504, "step": 121000 }, { "epoch": 10.48, "learning_rate": 4.886882733896149e-05, "loss": 1.1378, "step": 121010 }, { "epoch": 10.48, "learning_rate": 4.8866658956863646e-05, "loss": 1.3767, "step": 121020 }, { "epoch": 10.48, "learning_rate": 4.886449057476581e-05, "loss": 1.2581, "step": 121030 }, { "epoch": 10.48, "learning_rate": 4.886232219266798e-05, "loss": 1.2648, "step": 121040 }, { "epoch": 10.48, "learning_rate": 4.886015381057013e-05, "loss": 1.1348, "step": 121050 }, { "epoch": 10.48, "learning_rate": 4.88579854284723e-05, "loss": 1.216, "step": 121060 }, { "epoch": 10.49, "learning_rate": 4.885581704637446e-05, "loss": 1.2355, "step": 121070 }, { "epoch": 10.49, "learning_rate": 4.8853648664276624e-05, "loss": 1.2517, "step": 121080 }, { "epoch": 10.49, "learning_rate": 4.8851480282178784e-05, "loss": 1.1703, "step": 121090 }, { "epoch": 10.49, "learning_rate": 4.884931190008095e-05, "loss": 1.2266, "step": 121100 }, { "epoch": 10.49, "learning_rate": 4.884714351798312e-05, "loss": 1.0597, "step": 121110 }, { "epoch": 10.49, "learning_rate": 4.884497513588527e-05, "loss": 1.261, "step": 121120 }, { "epoch": 10.49, "learning_rate": 4.8842806753787436e-05, "loss": 1.1549, "step": 121130 }, { "epoch": 10.49, "learning_rate": 4.8840638371689596e-05, "loss": 1.2345, "step": 121140 }, { "epoch": 10.49, "learning_rate": 4.883846998959176e-05, "loss": 1.1452, "step": 121150 }, { "epoch": 10.49, "learning_rate": 4.883630160749392e-05, "loss": 1.133, "step": 121160 }, { "epoch": 10.49, "learning_rate": 4.883413322539609e-05, "loss": 1.1452, "step": 121170 }, { "epoch": 10.49, "learning_rate": 4.8831964843298255e-05, "loss": 1.1313, "step": 121180 }, { "epoch": 10.5, "learning_rate": 4.8829796461200414e-05, "loss": 1.1681, "step": 121190 }, { "epoch": 10.5, "learning_rate": 4.8827628079102574e-05, "loss": 1.0719, "step": 121200 }, { "epoch": 10.5, "learning_rate": 4.882545969700473e-05, "loss": 1.1917, "step": 121210 }, { "epoch": 10.5, "learning_rate": 4.88232913149069e-05, "loss": 1.2157, "step": 121220 }, { "epoch": 10.5, "learning_rate": 4.882112293280906e-05, "loss": 1.1576, "step": 121230 }, { "epoch": 10.5, "learning_rate": 4.8818954550711226e-05, "loss": 1.2313, "step": 121240 }, { "epoch": 10.5, "learning_rate": 4.881678616861339e-05, "loss": 1.0072, "step": 121250 }, { "epoch": 10.5, "learning_rate": 4.881461778651555e-05, "loss": 1.098, "step": 121260 }, { "epoch": 10.5, "learning_rate": 4.881244940441772e-05, "loss": 1.1013, "step": 121270 }, { "epoch": 10.5, "learning_rate": 4.881028102231987e-05, "loss": 1.2307, "step": 121280 }, { "epoch": 10.5, "learning_rate": 4.880811264022204e-05, "loss": 1.2268, "step": 121290 }, { "epoch": 10.51, "learning_rate": 4.88059442581242e-05, "loss": 1.1489, "step": 121300 }, { "epoch": 10.51, "learning_rate": 4.8803775876026363e-05, "loss": 1.1493, "step": 121310 }, { "epoch": 10.51, "learning_rate": 4.880160749392853e-05, "loss": 1.1978, "step": 121320 }, { "epoch": 10.51, "learning_rate": 4.879943911183069e-05, "loss": 1.2092, "step": 121330 }, { "epoch": 10.51, "learning_rate": 4.8797270729732856e-05, "loss": 1.2888, "step": 121340 }, { "epoch": 10.51, "learning_rate": 4.8795102347635015e-05, "loss": 1.1867, "step": 121350 }, { "epoch": 10.51, "learning_rate": 4.8792933965537175e-05, "loss": 1.0665, "step": 121360 }, { "epoch": 10.51, "learning_rate": 4.8790765583439335e-05, "loss": 1.1206, "step": 121370 }, { "epoch": 10.51, "learning_rate": 4.87885972013415e-05, "loss": 1.117, "step": 121380 }, { "epoch": 10.51, "learning_rate": 4.878642881924367e-05, "loss": 1.1095, "step": 121390 }, { "epoch": 10.51, "learning_rate": 4.878426043714583e-05, "loss": 1.2352, "step": 121400 }, { "epoch": 10.51, "learning_rate": 4.8782092055047993e-05, "loss": 1.234, "step": 121410 }, { "epoch": 10.52, "learning_rate": 4.877992367295015e-05, "loss": 1.0498, "step": 121420 }, { "epoch": 10.52, "learning_rate": 4.877775529085231e-05, "loss": 1.3366, "step": 121430 }, { "epoch": 10.52, "learning_rate": 4.877558690875447e-05, "loss": 1.2936, "step": 121440 }, { "epoch": 10.52, "learning_rate": 4.877341852665664e-05, "loss": 1.3367, "step": 121450 }, { "epoch": 10.52, "learning_rate": 4.87712501445588e-05, "loss": 1.1695, "step": 121460 }, { "epoch": 10.52, "learning_rate": 4.8769081762460965e-05, "loss": 1.2969, "step": 121470 }, { "epoch": 10.52, "learning_rate": 4.876691338036313e-05, "loss": 1.3168, "step": 121480 }, { "epoch": 10.52, "learning_rate": 4.876474499826529e-05, "loss": 1.0866, "step": 121490 }, { "epoch": 10.52, "learning_rate": 4.876257661616746e-05, "loss": 1.1452, "step": 121500 }, { "epoch": 10.52, "learning_rate": 4.876040823406961e-05, "loss": 1.2496, "step": 121510 }, { "epoch": 10.52, "learning_rate": 4.8758239851971776e-05, "loss": 1.2708, "step": 121520 }, { "epoch": 10.53, "learning_rate": 4.8756071469873936e-05, "loss": 1.2339, "step": 121530 }, { "epoch": 10.53, "learning_rate": 4.87539030877761e-05, "loss": 1.1716, "step": 121540 }, { "epoch": 10.53, "learning_rate": 4.875173470567827e-05, "loss": 1.1285, "step": 121550 }, { "epoch": 10.53, "learning_rate": 4.874956632358043e-05, "loss": 1.4299, "step": 121560 }, { "epoch": 10.53, "learning_rate": 4.8747397941482595e-05, "loss": 1.2692, "step": 121570 }, { "epoch": 10.53, "learning_rate": 4.8745229559384754e-05, "loss": 0.8512, "step": 121580 }, { "epoch": 10.53, "learning_rate": 4.8743061177286914e-05, "loss": 1.1958, "step": 121590 }, { "epoch": 10.53, "learning_rate": 4.8740892795189074e-05, "loss": 1.2676, "step": 121600 }, { "epoch": 10.53, "learning_rate": 4.873872441309124e-05, "loss": 1.1198, "step": 121610 }, { "epoch": 10.53, "learning_rate": 4.8736556030993406e-05, "loss": 1.3677, "step": 121620 }, { "epoch": 10.53, "learning_rate": 4.8734387648895566e-05, "loss": 1.2766, "step": 121630 }, { "epoch": 10.53, "learning_rate": 4.873221926679773e-05, "loss": 1.1367, "step": 121640 }, { "epoch": 10.54, "learning_rate": 4.873005088469989e-05, "loss": 1.229, "step": 121650 }, { "epoch": 10.54, "learning_rate": 4.872788250260205e-05, "loss": 1.0601, "step": 121660 }, { "epoch": 10.54, "learning_rate": 4.872571412050421e-05, "loss": 1.1801, "step": 121670 }, { "epoch": 10.54, "learning_rate": 4.872354573840638e-05, "loss": 1.2407, "step": 121680 }, { "epoch": 10.54, "learning_rate": 4.8721377356308544e-05, "loss": 1.2533, "step": 121690 }, { "epoch": 10.54, "learning_rate": 4.8719208974210704e-05, "loss": 1.36, "step": 121700 }, { "epoch": 10.54, "learning_rate": 4.871704059211287e-05, "loss": 1.2691, "step": 121710 }, { "epoch": 10.54, "learning_rate": 4.871487221001503e-05, "loss": 1.1419, "step": 121720 }, { "epoch": 10.54, "learning_rate": 4.8712703827917196e-05, "loss": 1.1818, "step": 121730 }, { "epoch": 10.54, "learning_rate": 4.871053544581935e-05, "loss": 1.3237, "step": 121740 }, { "epoch": 10.54, "learning_rate": 4.8708367063721515e-05, "loss": 1.2676, "step": 121750 }, { "epoch": 10.55, "learning_rate": 4.870619868162368e-05, "loss": 0.9575, "step": 121760 }, { "epoch": 10.55, "learning_rate": 4.870403029952584e-05, "loss": 1.2315, "step": 121770 }, { "epoch": 10.55, "learning_rate": 4.870186191742801e-05, "loss": 1.3224, "step": 121780 }, { "epoch": 10.55, "learning_rate": 4.869969353533017e-05, "loss": 1.153, "step": 121790 }, { "epoch": 10.55, "learning_rate": 4.8697525153232334e-05, "loss": 1.3921, "step": 121800 }, { "epoch": 10.55, "learning_rate": 4.86953567711345e-05, "loss": 1.0199, "step": 121810 }, { "epoch": 10.55, "learning_rate": 4.869318838903665e-05, "loss": 1.3228, "step": 121820 }, { "epoch": 10.55, "learning_rate": 4.869102000693882e-05, "loss": 1.2452, "step": 121830 }, { "epoch": 10.55, "learning_rate": 4.868885162484098e-05, "loss": 1.1484, "step": 121840 }, { "epoch": 10.55, "learning_rate": 4.8686683242743145e-05, "loss": 1.1478, "step": 121850 }, { "epoch": 10.55, "learning_rate": 4.8684514860645305e-05, "loss": 1.0711, "step": 121860 }, { "epoch": 10.55, "learning_rate": 4.868234647854747e-05, "loss": 1.3076, "step": 121870 }, { "epoch": 10.56, "learning_rate": 4.868017809644964e-05, "loss": 1.3146, "step": 121880 }, { "epoch": 10.56, "learning_rate": 4.867800971435179e-05, "loss": 1.1908, "step": 121890 }, { "epoch": 10.56, "learning_rate": 4.867584133225396e-05, "loss": 1.2234, "step": 121900 }, { "epoch": 10.56, "learning_rate": 4.8673672950156117e-05, "loss": 1.0962, "step": 121910 }, { "epoch": 10.56, "learning_rate": 4.867150456805828e-05, "loss": 0.9638, "step": 121920 }, { "epoch": 10.56, "learning_rate": 4.866933618596044e-05, "loss": 1.1208, "step": 121930 }, { "epoch": 10.56, "learning_rate": 4.866716780386261e-05, "loss": 1.1779, "step": 121940 }, { "epoch": 10.56, "learning_rate": 4.8664999421764775e-05, "loss": 1.4296, "step": 121950 }, { "epoch": 10.56, "learning_rate": 4.8662831039666935e-05, "loss": 1.4284, "step": 121960 }, { "epoch": 10.56, "learning_rate": 4.866066265756909e-05, "loss": 1.2329, "step": 121970 }, { "epoch": 10.56, "learning_rate": 4.8658494275471254e-05, "loss": 1.2373, "step": 121980 }, { "epoch": 10.57, "learning_rate": 4.865632589337342e-05, "loss": 1.2658, "step": 121990 }, { "epoch": 10.57, "learning_rate": 4.865415751127558e-05, "loss": 1.0973, "step": 122000 }, { "epoch": 10.57, "learning_rate": 4.8651989129177747e-05, "loss": 1.0008, "step": 122010 }, { "epoch": 10.57, "learning_rate": 4.8649820747079906e-05, "loss": 1.271, "step": 122020 }, { "epoch": 10.57, "learning_rate": 4.864765236498207e-05, "loss": 1.2998, "step": 122030 }, { "epoch": 10.57, "learning_rate": 4.864548398288424e-05, "loss": 1.1466, "step": 122040 }, { "epoch": 10.57, "learning_rate": 4.864331560078639e-05, "loss": 1.1235, "step": 122050 }, { "epoch": 10.57, "learning_rate": 4.864114721868856e-05, "loss": 1.0921, "step": 122060 }, { "epoch": 10.57, "learning_rate": 4.863897883659072e-05, "loss": 1.1858, "step": 122070 }, { "epoch": 10.57, "learning_rate": 4.8636810454492884e-05, "loss": 1.1457, "step": 122080 }, { "epoch": 10.57, "learning_rate": 4.8634642072395044e-05, "loss": 1.1942, "step": 122090 }, { "epoch": 10.57, "learning_rate": 4.863247369029721e-05, "loss": 1.0089, "step": 122100 }, { "epoch": 10.58, "learning_rate": 4.8630305308199377e-05, "loss": 0.9243, "step": 122110 }, { "epoch": 10.58, "learning_rate": 4.8628136926101536e-05, "loss": 1.4314, "step": 122120 }, { "epoch": 10.58, "learning_rate": 4.8625968544003696e-05, "loss": 1.2785, "step": 122130 }, { "epoch": 10.58, "learning_rate": 4.8623800161905855e-05, "loss": 1.3041, "step": 122140 }, { "epoch": 10.58, "learning_rate": 4.862163177980802e-05, "loss": 1.3928, "step": 122150 }, { "epoch": 10.58, "learning_rate": 4.861946339771018e-05, "loss": 1.2081, "step": 122160 }, { "epoch": 10.58, "learning_rate": 4.861729501561235e-05, "loss": 1.3366, "step": 122170 }, { "epoch": 10.58, "learning_rate": 4.8615126633514514e-05, "loss": 1.1041, "step": 122180 }, { "epoch": 10.58, "learning_rate": 4.8612958251416674e-05, "loss": 0.9593, "step": 122190 }, { "epoch": 10.58, "learning_rate": 4.8610789869318833e-05, "loss": 1.1325, "step": 122200 }, { "epoch": 10.58, "learning_rate": 4.860862148722099e-05, "loss": 1.2998, "step": 122210 }, { "epoch": 10.59, "learning_rate": 4.860645310512316e-05, "loss": 1.245, "step": 122220 }, { "epoch": 10.59, "learning_rate": 4.860428472302532e-05, "loss": 1.4069, "step": 122230 }, { "epoch": 10.59, "learning_rate": 4.8602116340927485e-05, "loss": 1.2132, "step": 122240 }, { "epoch": 10.59, "learning_rate": 4.859994795882965e-05, "loss": 1.182, "step": 122250 }, { "epoch": 10.59, "learning_rate": 4.859777957673181e-05, "loss": 1.1735, "step": 122260 }, { "epoch": 10.59, "learning_rate": 4.859561119463398e-05, "loss": 1.2939, "step": 122270 }, { "epoch": 10.59, "learning_rate": 4.859344281253613e-05, "loss": 1.0734, "step": 122280 }, { "epoch": 10.59, "learning_rate": 4.85912744304383e-05, "loss": 1.1611, "step": 122290 }, { "epoch": 10.59, "learning_rate": 4.858910604834046e-05, "loss": 1.0729, "step": 122300 }, { "epoch": 10.59, "learning_rate": 4.858693766624262e-05, "loss": 1.3647, "step": 122310 }, { "epoch": 10.59, "learning_rate": 4.858476928414479e-05, "loss": 1.1259, "step": 122320 }, { "epoch": 10.59, "learning_rate": 4.858260090204695e-05, "loss": 1.1849, "step": 122330 }, { "epoch": 10.6, "learning_rate": 4.8580432519949115e-05, "loss": 1.1478, "step": 122340 }, { "epoch": 10.6, "learning_rate": 4.8578264137851275e-05, "loss": 1.2733, "step": 122350 }, { "epoch": 10.6, "learning_rate": 4.8576095755753435e-05, "loss": 1.141, "step": 122360 }, { "epoch": 10.6, "learning_rate": 4.8573927373655594e-05, "loss": 1.3204, "step": 122370 }, { "epoch": 10.6, "learning_rate": 4.857175899155776e-05, "loss": 1.0699, "step": 122380 }, { "epoch": 10.6, "learning_rate": 4.856959060945993e-05, "loss": 1.1912, "step": 122390 }, { "epoch": 10.6, "learning_rate": 4.856742222736209e-05, "loss": 1.1134, "step": 122400 }, { "epoch": 10.6, "learning_rate": 4.856525384526425e-05, "loss": 1.1177, "step": 122410 }, { "epoch": 10.6, "learning_rate": 4.856308546316641e-05, "loss": 1.2844, "step": 122420 }, { "epoch": 10.6, "learning_rate": 4.856091708106857e-05, "loss": 1.4715, "step": 122430 }, { "epoch": 10.6, "learning_rate": 4.855874869897073e-05, "loss": 1.1536, "step": 122440 }, { "epoch": 10.6, "learning_rate": 4.85565803168729e-05, "loss": 1.3659, "step": 122450 }, { "epoch": 10.61, "learning_rate": 4.8554411934775065e-05, "loss": 1.3457, "step": 122460 }, { "epoch": 10.61, "learning_rate": 4.8552243552677224e-05, "loss": 1.1722, "step": 122470 }, { "epoch": 10.61, "learning_rate": 4.855007517057939e-05, "loss": 1.0635, "step": 122480 }, { "epoch": 10.61, "learning_rate": 4.854790678848155e-05, "loss": 1.0617, "step": 122490 }, { "epoch": 10.61, "learning_rate": 4.854573840638372e-05, "loss": 1.0672, "step": 122500 }, { "epoch": 10.61, "learning_rate": 4.854357002428587e-05, "loss": 1.1982, "step": 122510 }, { "epoch": 10.61, "learning_rate": 4.8541401642188036e-05, "loss": 1.3221, "step": 122520 }, { "epoch": 10.61, "learning_rate": 4.8539233260090196e-05, "loss": 1.1433, "step": 122530 }, { "epoch": 10.61, "learning_rate": 4.853706487799236e-05, "loss": 1.1774, "step": 122540 }, { "epoch": 10.61, "learning_rate": 4.853489649589453e-05, "loss": 1.173, "step": 122550 }, { "epoch": 10.61, "learning_rate": 4.853272811379669e-05, "loss": 1.078, "step": 122560 }, { "epoch": 10.62, "learning_rate": 4.8530559731698854e-05, "loss": 1.3254, "step": 122570 }, { "epoch": 10.62, "learning_rate": 4.8528391349601014e-05, "loss": 0.9908, "step": 122580 }, { "epoch": 10.62, "learning_rate": 4.8526222967503174e-05, "loss": 1.1429, "step": 122590 }, { "epoch": 10.62, "learning_rate": 4.852405458540533e-05, "loss": 1.1389, "step": 122600 }, { "epoch": 10.62, "learning_rate": 4.85218862033075e-05, "loss": 1.0844, "step": 122610 }, { "epoch": 10.62, "learning_rate": 4.8519717821209666e-05, "loss": 1.0875, "step": 122620 }, { "epoch": 10.62, "learning_rate": 4.8517549439111826e-05, "loss": 1.0425, "step": 122630 }, { "epoch": 10.62, "learning_rate": 4.851538105701399e-05, "loss": 1.306, "step": 122640 }, { "epoch": 10.62, "learning_rate": 4.851321267491615e-05, "loss": 1.1822, "step": 122650 }, { "epoch": 10.62, "learning_rate": 4.851104429281831e-05, "loss": 1.315, "step": 122660 }, { "epoch": 10.62, "learning_rate": 4.850887591072047e-05, "loss": 1.4905, "step": 122670 }, { "epoch": 10.62, "learning_rate": 4.850670752862264e-05, "loss": 1.1717, "step": 122680 }, { "epoch": 10.63, "learning_rate": 4.8504539146524804e-05, "loss": 1.2746, "step": 122690 }, { "epoch": 10.63, "learning_rate": 4.850237076442696e-05, "loss": 1.0648, "step": 122700 }, { "epoch": 10.63, "learning_rate": 4.850020238232913e-05, "loss": 1.1917, "step": 122710 }, { "epoch": 10.63, "learning_rate": 4.849803400023129e-05, "loss": 1.1839, "step": 122720 }, { "epoch": 10.63, "learning_rate": 4.8495865618133456e-05, "loss": 1.1859, "step": 122730 }, { "epoch": 10.63, "learning_rate": 4.849369723603561e-05, "loss": 1.025, "step": 122740 }, { "epoch": 10.63, "learning_rate": 4.8491528853937775e-05, "loss": 1.2704, "step": 122750 }, { "epoch": 10.63, "learning_rate": 4.848936047183994e-05, "loss": 1.0634, "step": 122760 }, { "epoch": 10.63, "learning_rate": 4.84871920897421e-05, "loss": 1.2376, "step": 122770 }, { "epoch": 10.63, "learning_rate": 4.848502370764427e-05, "loss": 1.0972, "step": 122780 }, { "epoch": 10.63, "learning_rate": 4.848285532554643e-05, "loss": 1.1092, "step": 122790 }, { "epoch": 10.64, "learning_rate": 4.848068694344859e-05, "loss": 1.3449, "step": 122800 }, { "epoch": 10.64, "learning_rate": 4.847851856135076e-05, "loss": 1.0174, "step": 122810 }, { "epoch": 10.64, "learning_rate": 4.847635017925291e-05, "loss": 1.2905, "step": 122820 }, { "epoch": 10.64, "learning_rate": 4.847418179715508e-05, "loss": 1.1448, "step": 122830 }, { "epoch": 10.64, "learning_rate": 4.847201341505724e-05, "loss": 1.2245, "step": 122840 }, { "epoch": 10.64, "learning_rate": 4.8469845032959405e-05, "loss": 1.2258, "step": 122850 }, { "epoch": 10.64, "learning_rate": 4.8467676650861565e-05, "loss": 1.2032, "step": 122860 }, { "epoch": 10.64, "learning_rate": 4.846550826876373e-05, "loss": 1.3213, "step": 122870 }, { "epoch": 10.64, "learning_rate": 4.84633398866659e-05, "loss": 1.2167, "step": 122880 }, { "epoch": 10.64, "learning_rate": 4.846117150456806e-05, "loss": 1.203, "step": 122890 }, { "epoch": 10.64, "learning_rate": 4.8459003122470217e-05, "loss": 1.1694, "step": 122900 }, { "epoch": 10.64, "learning_rate": 4.8456834740372376e-05, "loss": 1.0296, "step": 122910 }, { "epoch": 10.65, "learning_rate": 4.845466635827454e-05, "loss": 1.0961, "step": 122920 }, { "epoch": 10.65, "learning_rate": 4.84524979761767e-05, "loss": 1.2725, "step": 122930 }, { "epoch": 10.65, "learning_rate": 4.845032959407887e-05, "loss": 1.2909, "step": 122940 }, { "epoch": 10.65, "learning_rate": 4.8448161211981035e-05, "loss": 1.1295, "step": 122950 }, { "epoch": 10.65, "learning_rate": 4.8445992829883195e-05, "loss": 1.2272, "step": 122960 }, { "epoch": 10.65, "learning_rate": 4.8443824447785354e-05, "loss": 1.2177, "step": 122970 }, { "epoch": 10.65, "learning_rate": 4.8441656065687514e-05, "loss": 1.2814, "step": 122980 }, { "epoch": 10.65, "learning_rate": 4.843948768358968e-05, "loss": 1.0859, "step": 122990 }, { "epoch": 10.65, "learning_rate": 4.843731930149184e-05, "loss": 1.1197, "step": 123000 }, { "epoch": 10.65, "learning_rate": 4.8435150919394006e-05, "loss": 1.0489, "step": 123010 }, { "epoch": 10.65, "learning_rate": 4.843298253729617e-05, "loss": 1.1589, "step": 123020 }, { "epoch": 10.66, "learning_rate": 4.843081415519833e-05, "loss": 1.2324, "step": 123030 }, { "epoch": 10.66, "learning_rate": 4.84286457731005e-05, "loss": 1.3422, "step": 123040 }, { "epoch": 10.66, "learning_rate": 4.842647739100265e-05, "loss": 1.0941, "step": 123050 }, { "epoch": 10.66, "learning_rate": 4.842430900890482e-05, "loss": 0.9948, "step": 123060 }, { "epoch": 10.66, "learning_rate": 4.842214062680698e-05, "loss": 1.1744, "step": 123070 }, { "epoch": 10.66, "learning_rate": 4.8419972244709144e-05, "loss": 1.1048, "step": 123080 }, { "epoch": 10.66, "learning_rate": 4.8417803862611303e-05, "loss": 1.1225, "step": 123090 }, { "epoch": 10.66, "learning_rate": 4.841563548051347e-05, "loss": 1.1166, "step": 123100 }, { "epoch": 10.66, "learning_rate": 4.8413467098415636e-05, "loss": 1.0722, "step": 123110 }, { "epoch": 10.66, "learning_rate": 4.8411298716317796e-05, "loss": 1.072, "step": 123120 }, { "epoch": 10.66, "learning_rate": 4.8409130334219956e-05, "loss": 1.2745, "step": 123130 }, { "epoch": 10.66, "learning_rate": 4.8406961952122115e-05, "loss": 1.283, "step": 123140 }, { "epoch": 10.67, "learning_rate": 4.840479357002428e-05, "loss": 1.0325, "step": 123150 }, { "epoch": 10.67, "learning_rate": 4.840262518792644e-05, "loss": 1.2384, "step": 123160 }, { "epoch": 10.67, "learning_rate": 4.840045680582861e-05, "loss": 1.1799, "step": 123170 }, { "epoch": 10.67, "learning_rate": 4.8398288423730774e-05, "loss": 1.2097, "step": 123180 }, { "epoch": 10.67, "learning_rate": 4.8396120041632934e-05, "loss": 1.2592, "step": 123190 }, { "epoch": 10.67, "learning_rate": 4.839395165953509e-05, "loss": 1.218, "step": 123200 }, { "epoch": 10.67, "learning_rate": 4.839178327743725e-05, "loss": 1.1144, "step": 123210 }, { "epoch": 10.67, "learning_rate": 4.838961489533942e-05, "loss": 1.0199, "step": 123220 }, { "epoch": 10.67, "learning_rate": 4.838744651324158e-05, "loss": 1.1356, "step": 123230 }, { "epoch": 10.67, "learning_rate": 4.8385278131143745e-05, "loss": 1.3936, "step": 123240 }, { "epoch": 10.67, "learning_rate": 4.838310974904591e-05, "loss": 1.1554, "step": 123250 }, { "epoch": 10.68, "learning_rate": 4.838094136694807e-05, "loss": 1.2701, "step": 123260 }, { "epoch": 10.68, "learning_rate": 4.837877298485024e-05, "loss": 1.1852, "step": 123270 }, { "epoch": 10.68, "learning_rate": 4.837660460275239e-05, "loss": 1.3006, "step": 123280 }, { "epoch": 10.68, "learning_rate": 4.837443622065456e-05, "loss": 1.1899, "step": 123290 }, { "epoch": 10.68, "learning_rate": 4.8372267838556716e-05, "loss": 0.8058, "step": 123300 }, { "epoch": 10.68, "learning_rate": 4.837009945645888e-05, "loss": 1.2334, "step": 123310 }, { "epoch": 10.68, "learning_rate": 4.836793107436105e-05, "loss": 1.1478, "step": 123320 }, { "epoch": 10.68, "learning_rate": 4.836576269226321e-05, "loss": 1.2138, "step": 123330 }, { "epoch": 10.68, "learning_rate": 4.8363594310165375e-05, "loss": 1.2254, "step": 123340 }, { "epoch": 10.68, "learning_rate": 4.8361425928067535e-05, "loss": 1.1378, "step": 123350 }, { "epoch": 10.68, "learning_rate": 4.8359257545969694e-05, "loss": 1.2765, "step": 123360 }, { "epoch": 10.68, "learning_rate": 4.8357089163871854e-05, "loss": 1.091, "step": 123370 }, { "epoch": 10.69, "learning_rate": 4.835492078177402e-05, "loss": 1.2047, "step": 123380 }, { "epoch": 10.69, "learning_rate": 4.835275239967619e-05, "loss": 1.1232, "step": 123390 }, { "epoch": 10.69, "learning_rate": 4.8350584017578346e-05, "loss": 1.0568, "step": 123400 }, { "epoch": 10.69, "learning_rate": 4.834841563548051e-05, "loss": 1.1952, "step": 123410 }, { "epoch": 10.69, "learning_rate": 4.834624725338267e-05, "loss": 1.1892, "step": 123420 }, { "epoch": 10.69, "learning_rate": 4.834407887128483e-05, "loss": 1.1285, "step": 123430 }, { "epoch": 10.69, "learning_rate": 4.834191048918699e-05, "loss": 1.1793, "step": 123440 }, { "epoch": 10.69, "learning_rate": 4.833974210708916e-05, "loss": 1.3183, "step": 123450 }, { "epoch": 10.69, "learning_rate": 4.8337573724991324e-05, "loss": 1.1675, "step": 123460 }, { "epoch": 10.69, "learning_rate": 4.8335405342893484e-05, "loss": 1.1924, "step": 123470 }, { "epoch": 10.69, "learning_rate": 4.833323696079565e-05, "loss": 1.1203, "step": 123480 }, { "epoch": 10.7, "learning_rate": 4.833106857869781e-05, "loss": 1.2772, "step": 123490 }, { "epoch": 10.7, "learning_rate": 4.8328900196599976e-05, "loss": 1.1591, "step": 123500 }, { "epoch": 10.7, "learning_rate": 4.832673181450213e-05, "loss": 1.1939, "step": 123510 }, { "epoch": 10.7, "learning_rate": 4.8324563432404296e-05, "loss": 1.1081, "step": 123520 }, { "epoch": 10.7, "learning_rate": 4.832239505030646e-05, "loss": 1.1913, "step": 123530 }, { "epoch": 10.7, "learning_rate": 4.832022666820862e-05, "loss": 1.1906, "step": 123540 }, { "epoch": 10.7, "learning_rate": 4.831805828611079e-05, "loss": 0.88, "step": 123550 }, { "epoch": 10.7, "learning_rate": 4.831588990401295e-05, "loss": 1.1745, "step": 123560 }, { "epoch": 10.7, "learning_rate": 4.8313721521915114e-05, "loss": 1.0282, "step": 123570 }, { "epoch": 10.7, "learning_rate": 4.831155313981728e-05, "loss": 1.3699, "step": 123580 }, { "epoch": 10.7, "learning_rate": 4.830938475771943e-05, "loss": 1.2485, "step": 123590 }, { "epoch": 10.7, "learning_rate": 4.830721637562159e-05, "loss": 1.2278, "step": 123600 }, { "epoch": 10.71, "learning_rate": 4.830504799352376e-05, "loss": 1.1896, "step": 123610 }, { "epoch": 10.71, "learning_rate": 4.8302879611425926e-05, "loss": 1.0976, "step": 123620 }, { "epoch": 10.71, "learning_rate": 4.8300711229328085e-05, "loss": 0.959, "step": 123630 }, { "epoch": 10.71, "learning_rate": 4.829854284723025e-05, "loss": 1.108, "step": 123640 }, { "epoch": 10.71, "learning_rate": 4.829637446513241e-05, "loss": 1.1323, "step": 123650 }, { "epoch": 10.71, "learning_rate": 4.829420608303458e-05, "loss": 1.0658, "step": 123660 }, { "epoch": 10.71, "learning_rate": 4.829203770093673e-05, "loss": 1.2748, "step": 123670 }, { "epoch": 10.71, "learning_rate": 4.82898693188389e-05, "loss": 1.2963, "step": 123680 }, { "epoch": 10.71, "learning_rate": 4.828770093674106e-05, "loss": 1.2332, "step": 123690 }, { "epoch": 10.71, "learning_rate": 4.828553255464322e-05, "loss": 1.0752, "step": 123700 }, { "epoch": 10.71, "learning_rate": 4.828336417254539e-05, "loss": 1.0743, "step": 123710 }, { "epoch": 10.71, "learning_rate": 4.828119579044755e-05, "loss": 1.3922, "step": 123720 }, { "epoch": 10.72, "learning_rate": 4.8279027408349715e-05, "loss": 1.2922, "step": 123730 }, { "epoch": 10.72, "learning_rate": 4.827685902625187e-05, "loss": 1.1468, "step": 123740 }, { "epoch": 10.72, "learning_rate": 4.8274690644154035e-05, "loss": 1.1341, "step": 123750 }, { "epoch": 10.72, "learning_rate": 4.82725222620562e-05, "loss": 1.1844, "step": 123760 }, { "epoch": 10.72, "learning_rate": 4.827035387995836e-05, "loss": 1.0847, "step": 123770 }, { "epoch": 10.72, "learning_rate": 4.826818549786053e-05, "loss": 1.0786, "step": 123780 }, { "epoch": 10.72, "learning_rate": 4.826601711576269e-05, "loss": 1.1412, "step": 123790 }, { "epoch": 10.72, "learning_rate": 4.826384873366485e-05, "loss": 1.2351, "step": 123800 }, { "epoch": 10.72, "learning_rate": 4.826168035156702e-05, "loss": 1.0912, "step": 123810 }, { "epoch": 10.72, "learning_rate": 4.825951196946917e-05, "loss": 1.394, "step": 123820 }, { "epoch": 10.72, "learning_rate": 4.825734358737134e-05, "loss": 1.3129, "step": 123830 }, { "epoch": 10.73, "learning_rate": 4.82551752052735e-05, "loss": 1.1724, "step": 123840 }, { "epoch": 10.73, "learning_rate": 4.8253006823175665e-05, "loss": 1.1592, "step": 123850 }, { "epoch": 10.73, "learning_rate": 4.8250838441077824e-05, "loss": 1.4087, "step": 123860 }, { "epoch": 10.73, "learning_rate": 4.824867005897999e-05, "loss": 1.1945, "step": 123870 }, { "epoch": 10.73, "learning_rate": 4.824650167688216e-05, "loss": 1.2206, "step": 123880 }, { "epoch": 10.73, "learning_rate": 4.824433329478432e-05, "loss": 1.4448, "step": 123890 }, { "epoch": 10.73, "learning_rate": 4.8242164912686476e-05, "loss": 1.2242, "step": 123900 }, { "epoch": 10.73, "learning_rate": 4.8239996530588636e-05, "loss": 1.1355, "step": 123910 }, { "epoch": 10.73, "learning_rate": 4.82378281484908e-05, "loss": 1.2086, "step": 123920 }, { "epoch": 10.73, "learning_rate": 4.823565976639296e-05, "loss": 1.1362, "step": 123930 }, { "epoch": 10.73, "learning_rate": 4.823349138429513e-05, "loss": 1.2628, "step": 123940 }, { "epoch": 10.73, "learning_rate": 4.8231323002197295e-05, "loss": 1.2741, "step": 123950 }, { "epoch": 10.74, "learning_rate": 4.8229154620099454e-05, "loss": 0.9879, "step": 123960 }, { "epoch": 10.74, "learning_rate": 4.8226986238001614e-05, "loss": 1.2776, "step": 123970 }, { "epoch": 10.74, "learning_rate": 4.8224817855903774e-05, "loss": 1.0748, "step": 123980 }, { "epoch": 10.74, "learning_rate": 4.822264947380594e-05, "loss": 1.1593, "step": 123990 }, { "epoch": 10.74, "learning_rate": 4.82204810917081e-05, "loss": 1.0766, "step": 124000 }, { "epoch": 10.74, "learning_rate": 4.8218312709610266e-05, "loss": 1.3434, "step": 124010 }, { "epoch": 10.74, "learning_rate": 4.821614432751243e-05, "loss": 1.2117, "step": 124020 }, { "epoch": 10.74, "learning_rate": 4.821397594541459e-05, "loss": 1.1589, "step": 124030 }, { "epoch": 10.74, "learning_rate": 4.821180756331676e-05, "loss": 1.3169, "step": 124040 }, { "epoch": 10.74, "learning_rate": 4.820963918121891e-05, "loss": 1.2549, "step": 124050 }, { "epoch": 10.74, "learning_rate": 4.820747079912108e-05, "loss": 1.3196, "step": 124060 }, { "epoch": 10.75, "learning_rate": 4.820530241702324e-05, "loss": 1.2522, "step": 124070 }, { "epoch": 10.75, "learning_rate": 4.8203134034925404e-05, "loss": 1.344, "step": 124080 }, { "epoch": 10.75, "learning_rate": 4.820096565282757e-05, "loss": 1.0012, "step": 124090 }, { "epoch": 10.75, "learning_rate": 4.819879727072973e-05, "loss": 1.1639, "step": 124100 }, { "epoch": 10.75, "learning_rate": 4.8196628888631896e-05, "loss": 1.157, "step": 124110 }, { "epoch": 10.75, "learning_rate": 4.8194460506534056e-05, "loss": 1.3212, "step": 124120 }, { "epoch": 10.75, "learning_rate": 4.8192292124436215e-05, "loss": 1.3025, "step": 124130 }, { "epoch": 10.75, "learning_rate": 4.8190123742338375e-05, "loss": 1.3268, "step": 124140 }, { "epoch": 10.75, "learning_rate": 4.818795536024054e-05, "loss": 1.0333, "step": 124150 }, { "epoch": 10.75, "learning_rate": 4.81857869781427e-05, "loss": 1.3089, "step": 124160 }, { "epoch": 10.75, "learning_rate": 4.818361859604487e-05, "loss": 1.1198, "step": 124170 }, { "epoch": 10.75, "learning_rate": 4.8181450213947034e-05, "loss": 1.2267, "step": 124180 }, { "epoch": 10.76, "learning_rate": 4.817928183184919e-05, "loss": 1.3473, "step": 124190 }, { "epoch": 10.76, "learning_rate": 4.817711344975135e-05, "loss": 1.0799, "step": 124200 }, { "epoch": 10.76, "learning_rate": 4.817494506765351e-05, "loss": 1.3692, "step": 124210 }, { "epoch": 10.76, "learning_rate": 4.817277668555568e-05, "loss": 1.197, "step": 124220 }, { "epoch": 10.76, "learning_rate": 4.817060830345784e-05, "loss": 1.2212, "step": 124230 }, { "epoch": 10.76, "learning_rate": 4.8168439921360005e-05, "loss": 1.2232, "step": 124240 }, { "epoch": 10.76, "learning_rate": 4.816627153926217e-05, "loss": 1.3093, "step": 124250 }, { "epoch": 10.76, "learning_rate": 4.816410315716433e-05, "loss": 1.2041, "step": 124260 }, { "epoch": 10.76, "learning_rate": 4.81619347750665e-05, "loss": 1.018, "step": 124270 }, { "epoch": 10.76, "learning_rate": 4.815976639296865e-05, "loss": 1.1289, "step": 124280 }, { "epoch": 10.76, "learning_rate": 4.8157598010870816e-05, "loss": 1.2542, "step": 124290 }, { "epoch": 10.77, "learning_rate": 4.8155429628772976e-05, "loss": 1.0864, "step": 124300 }, { "epoch": 10.77, "learning_rate": 4.815326124667514e-05, "loss": 1.239, "step": 124310 }, { "epoch": 10.77, "learning_rate": 4.815109286457731e-05, "loss": 1.1116, "step": 124320 }, { "epoch": 10.77, "learning_rate": 4.814892448247947e-05, "loss": 1.2059, "step": 124330 }, { "epoch": 10.77, "learning_rate": 4.8146756100381635e-05, "loss": 1.1361, "step": 124340 }, { "epoch": 10.77, "learning_rate": 4.8144587718283795e-05, "loss": 1.3402, "step": 124350 }, { "epoch": 10.77, "learning_rate": 4.8142419336185954e-05, "loss": 1.1905, "step": 124360 }, { "epoch": 10.77, "learning_rate": 4.8140250954088114e-05, "loss": 1.2395, "step": 124370 }, { "epoch": 10.77, "learning_rate": 4.813808257199028e-05, "loss": 1.0769, "step": 124380 }, { "epoch": 10.77, "learning_rate": 4.8135914189892447e-05, "loss": 1.1479, "step": 124390 }, { "epoch": 10.77, "learning_rate": 4.8133745807794606e-05, "loss": 1.004, "step": 124400 }, { "epoch": 10.77, "learning_rate": 4.813157742569677e-05, "loss": 1.4232, "step": 124410 }, { "epoch": 10.78, "learning_rate": 4.812940904359893e-05, "loss": 1.1026, "step": 124420 }, { "epoch": 10.78, "learning_rate": 4.81272406615011e-05, "loss": 1.4274, "step": 124430 }, { "epoch": 10.78, "learning_rate": 4.812507227940325e-05, "loss": 1.0462, "step": 124440 }, { "epoch": 10.78, "learning_rate": 4.812290389730542e-05, "loss": 1.0509, "step": 124450 }, { "epoch": 10.78, "learning_rate": 4.8120735515207584e-05, "loss": 1.1162, "step": 124460 }, { "epoch": 10.78, "learning_rate": 4.8118567133109744e-05, "loss": 1.198, "step": 124470 }, { "epoch": 10.78, "learning_rate": 4.811639875101191e-05, "loss": 1.4218, "step": 124480 }, { "epoch": 10.78, "learning_rate": 4.811423036891407e-05, "loss": 1.0938, "step": 124490 }, { "epoch": 10.78, "learning_rate": 4.8112061986816236e-05, "loss": 1.2114, "step": 124500 }, { "epoch": 10.78, "learning_rate": 4.810989360471839e-05, "loss": 1.2665, "step": 124510 }, { "epoch": 10.78, "learning_rate": 4.8107725222620555e-05, "loss": 0.9606, "step": 124520 }, { "epoch": 10.79, "learning_rate": 4.810555684052272e-05, "loss": 0.899, "step": 124530 }, { "epoch": 10.79, "learning_rate": 4.810338845842488e-05, "loss": 1.2141, "step": 124540 }, { "epoch": 10.79, "learning_rate": 4.810122007632705e-05, "loss": 1.3814, "step": 124550 }, { "epoch": 10.79, "learning_rate": 4.809905169422921e-05, "loss": 1.3067, "step": 124560 }, { "epoch": 10.79, "learning_rate": 4.8096883312131374e-05, "loss": 1.1737, "step": 124570 }, { "epoch": 10.79, "learning_rate": 4.809471493003354e-05, "loss": 1.3631, "step": 124580 }, { "epoch": 10.79, "learning_rate": 4.809254654793569e-05, "loss": 1.1557, "step": 124590 }, { "epoch": 10.79, "learning_rate": 4.809037816583786e-05, "loss": 1.14, "step": 124600 }, { "epoch": 10.79, "learning_rate": 4.808820978374002e-05, "loss": 1.214, "step": 124610 }, { "epoch": 10.79, "learning_rate": 4.8086041401642185e-05, "loss": 1.2532, "step": 124620 }, { "epoch": 10.79, "learning_rate": 4.8083873019544345e-05, "loss": 1.2141, "step": 124630 }, { "epoch": 10.79, "learning_rate": 4.808170463744651e-05, "loss": 1.0874, "step": 124640 }, { "epoch": 10.8, "learning_rate": 4.807953625534868e-05, "loss": 1.2663, "step": 124650 }, { "epoch": 10.8, "learning_rate": 4.807736787325084e-05, "loss": 1.1155, "step": 124660 }, { "epoch": 10.8, "learning_rate": 4.807519949115299e-05, "loss": 1.3649, "step": 124670 }, { "epoch": 10.8, "learning_rate": 4.807303110905516e-05, "loss": 1.1076, "step": 124680 }, { "epoch": 10.8, "learning_rate": 4.807086272695732e-05, "loss": 1.1856, "step": 124690 }, { "epoch": 10.8, "learning_rate": 4.806869434485948e-05, "loss": 0.9724, "step": 124700 }, { "epoch": 10.8, "learning_rate": 4.806652596276165e-05, "loss": 1.0712, "step": 124710 }, { "epoch": 10.8, "learning_rate": 4.806435758066381e-05, "loss": 1.269, "step": 124720 }, { "epoch": 10.8, "learning_rate": 4.8062189198565975e-05, "loss": 1.2157, "step": 124730 }, { "epoch": 10.8, "learning_rate": 4.806002081646813e-05, "loss": 1.0466, "step": 124740 }, { "epoch": 10.8, "learning_rate": 4.8057852434370294e-05, "loss": 1.2347, "step": 124750 }, { "epoch": 10.81, "learning_rate": 4.805568405227246e-05, "loss": 1.2665, "step": 124760 }, { "epoch": 10.81, "learning_rate": 4.805351567017462e-05, "loss": 1.1915, "step": 124770 }, { "epoch": 10.81, "learning_rate": 4.805134728807679e-05, "loss": 1.0505, "step": 124780 }, { "epoch": 10.81, "learning_rate": 4.8049178905978946e-05, "loss": 1.2042, "step": 124790 }, { "epoch": 10.81, "learning_rate": 4.804701052388111e-05, "loss": 1.2246, "step": 124800 }, { "epoch": 10.81, "learning_rate": 4.804484214178328e-05, "loss": 1.153, "step": 124810 }, { "epoch": 10.81, "learning_rate": 4.804267375968543e-05, "loss": 1.1637, "step": 124820 }, { "epoch": 10.81, "learning_rate": 4.80405053775876e-05, "loss": 1.0656, "step": 124830 }, { "epoch": 10.81, "learning_rate": 4.803833699548976e-05, "loss": 1.3517, "step": 124840 }, { "epoch": 10.81, "learning_rate": 4.8036168613391924e-05, "loss": 1.1403, "step": 124850 }, { "epoch": 10.81, "learning_rate": 4.8034000231294084e-05, "loss": 1.2337, "step": 124860 }, { "epoch": 10.81, "learning_rate": 4.803183184919625e-05, "loss": 1.0027, "step": 124870 }, { "epoch": 10.82, "learning_rate": 4.802966346709842e-05, "loss": 0.9976, "step": 124880 }, { "epoch": 10.82, "learning_rate": 4.8027495085000576e-05, "loss": 1.2701, "step": 124890 }, { "epoch": 10.82, "learning_rate": 4.8025326702902736e-05, "loss": 1.1561, "step": 124900 }, { "epoch": 10.82, "learning_rate": 4.8023158320804896e-05, "loss": 1.1826, "step": 124910 }, { "epoch": 10.82, "learning_rate": 4.802098993870706e-05, "loss": 1.2522, "step": 124920 }, { "epoch": 10.82, "learning_rate": 4.801882155660922e-05, "loss": 1.2245, "step": 124930 }, { "epoch": 10.82, "learning_rate": 4.801665317451139e-05, "loss": 1.191, "step": 124940 }, { "epoch": 10.82, "learning_rate": 4.8014484792413554e-05, "loss": 1.1556, "step": 124950 }, { "epoch": 10.82, "learning_rate": 4.8012316410315714e-05, "loss": 1.1816, "step": 124960 }, { "epoch": 10.82, "learning_rate": 4.8010148028217874e-05, "loss": 1.2043, "step": 124970 }, { "epoch": 10.82, "learning_rate": 4.800797964612003e-05, "loss": 1.352, "step": 124980 }, { "epoch": 10.82, "learning_rate": 4.80058112640222e-05, "loss": 1.3377, "step": 124990 }, { "epoch": 10.83, "learning_rate": 4.800364288192436e-05, "loss": 1.3064, "step": 125000 }, { "epoch": 10.83, "learning_rate": 4.8001474499826526e-05, "loss": 1.3171, "step": 125010 }, { "epoch": 10.83, "learning_rate": 4.799930611772869e-05, "loss": 1.2256, "step": 125020 }, { "epoch": 10.83, "learning_rate": 4.799713773563085e-05, "loss": 1.1556, "step": 125030 }, { "epoch": 10.83, "learning_rate": 4.799496935353302e-05, "loss": 1.254, "step": 125040 }, { "epoch": 10.83, "learning_rate": 4.799280097143517e-05, "loss": 0.9424, "step": 125050 }, { "epoch": 10.83, "learning_rate": 4.799063258933734e-05, "loss": 1.2107, "step": 125060 }, { "epoch": 10.83, "learning_rate": 4.79884642072395e-05, "loss": 1.0518, "step": 125070 }, { "epoch": 10.83, "learning_rate": 4.798629582514166e-05, "loss": 1.1493, "step": 125080 }, { "epoch": 10.83, "learning_rate": 4.798412744304383e-05, "loss": 0.9953, "step": 125090 }, { "epoch": 10.83, "learning_rate": 4.798195906094599e-05, "loss": 1.3224, "step": 125100 }, { "epoch": 10.84, "learning_rate": 4.7979790678848156e-05, "loss": 1.4191, "step": 125110 }, { "epoch": 10.84, "learning_rate": 4.7977622296750315e-05, "loss": 1.2895, "step": 125120 }, { "epoch": 10.84, "learning_rate": 4.7975453914652475e-05, "loss": 1.1138, "step": 125130 }, { "epoch": 10.84, "learning_rate": 4.7973285532554635e-05, "loss": 1.3587, "step": 125140 }, { "epoch": 10.84, "learning_rate": 4.79711171504568e-05, "loss": 1.202, "step": 125150 }, { "epoch": 10.84, "learning_rate": 4.796894876835897e-05, "loss": 1.2832, "step": 125160 }, { "epoch": 10.84, "learning_rate": 4.796678038626113e-05, "loss": 1.2096, "step": 125170 }, { "epoch": 10.84, "learning_rate": 4.796461200416329e-05, "loss": 1.0074, "step": 125180 }, { "epoch": 10.84, "learning_rate": 4.796244362206545e-05, "loss": 1.097, "step": 125190 }, { "epoch": 10.84, "learning_rate": 4.796027523996762e-05, "loss": 1.1439, "step": 125200 }, { "epoch": 10.84, "learning_rate": 4.795810685786977e-05, "loss": 1.3, "step": 125210 }, { "epoch": 10.84, "learning_rate": 4.795593847577194e-05, "loss": 1.0779, "step": 125220 }, { "epoch": 10.85, "learning_rate": 4.79537700936741e-05, "loss": 1.0476, "step": 125230 }, { "epoch": 10.85, "learning_rate": 4.7951601711576265e-05, "loss": 1.0899, "step": 125240 }, { "epoch": 10.85, "learning_rate": 4.794943332947843e-05, "loss": 1.1051, "step": 125250 }, { "epoch": 10.85, "learning_rate": 4.794726494738059e-05, "loss": 1.0278, "step": 125260 }, { "epoch": 10.85, "learning_rate": 4.794509656528276e-05, "loss": 1.4514, "step": 125270 }, { "epoch": 10.85, "learning_rate": 4.794292818318491e-05, "loss": 1.199, "step": 125280 }, { "epoch": 10.85, "learning_rate": 4.7940759801087076e-05, "loss": 1.3585, "step": 125290 }, { "epoch": 10.85, "learning_rate": 4.7938591418989236e-05, "loss": 0.8667, "step": 125300 }, { "epoch": 10.85, "learning_rate": 4.79364230368914e-05, "loss": 1.163, "step": 125310 }, { "epoch": 10.85, "learning_rate": 4.793425465479357e-05, "loss": 1.2807, "step": 125320 }, { "epoch": 10.85, "learning_rate": 4.793208627269573e-05, "loss": 1.0991, "step": 125330 }, { "epoch": 10.86, "learning_rate": 4.7929917890597895e-05, "loss": 1.2206, "step": 125340 }, { "epoch": 10.86, "learning_rate": 4.7927749508500054e-05, "loss": 1.191, "step": 125350 }, { "epoch": 10.86, "learning_rate": 4.7925581126402214e-05, "loss": 1.3151, "step": 125360 }, { "epoch": 10.86, "learning_rate": 4.7923412744304373e-05, "loss": 1.2424, "step": 125370 }, { "epoch": 10.86, "learning_rate": 4.792124436220654e-05, "loss": 1.1543, "step": 125380 }, { "epoch": 10.86, "learning_rate": 4.7919075980108706e-05, "loss": 1.127, "step": 125390 }, { "epoch": 10.86, "learning_rate": 4.7916907598010866e-05, "loss": 1.0438, "step": 125400 }, { "epoch": 10.86, "learning_rate": 4.791473921591303e-05, "loss": 1.0479, "step": 125410 }, { "epoch": 10.86, "learning_rate": 4.791257083381519e-05, "loss": 1.236, "step": 125420 }, { "epoch": 10.86, "learning_rate": 4.791040245171736e-05, "loss": 1.1773, "step": 125430 }, { "epoch": 10.86, "learning_rate": 4.790823406961951e-05, "loss": 1.1471, "step": 125440 }, { "epoch": 10.86, "learning_rate": 4.790606568752168e-05, "loss": 1.1022, "step": 125450 }, { "epoch": 10.87, "learning_rate": 4.7903897305423844e-05, "loss": 1.352, "step": 125460 }, { "epoch": 10.87, "learning_rate": 4.7901728923326003e-05, "loss": 1.1261, "step": 125470 }, { "epoch": 10.87, "learning_rate": 4.789956054122817e-05, "loss": 1.1321, "step": 125480 }, { "epoch": 10.87, "learning_rate": 4.789739215913033e-05, "loss": 1.0242, "step": 125490 }, { "epoch": 10.87, "learning_rate": 4.7895223777032496e-05, "loss": 1.1842, "step": 125500 }, { "epoch": 10.87, "learning_rate": 4.789305539493465e-05, "loss": 1.1182, "step": 125510 }, { "epoch": 10.87, "learning_rate": 4.7890887012836815e-05, "loss": 1.2814, "step": 125520 }, { "epoch": 10.87, "learning_rate": 4.788871863073898e-05, "loss": 1.1618, "step": 125530 }, { "epoch": 10.87, "learning_rate": 4.788655024864114e-05, "loss": 1.3639, "step": 125540 }, { "epoch": 10.87, "learning_rate": 4.788438186654331e-05, "loss": 1.3901, "step": 125550 }, { "epoch": 10.87, "learning_rate": 4.788221348444547e-05, "loss": 1.0707, "step": 125560 }, { "epoch": 10.88, "learning_rate": 4.7880045102347633e-05, "loss": 1.0718, "step": 125570 }, { "epoch": 10.88, "learning_rate": 4.78778767202498e-05, "loss": 1.3039, "step": 125580 }, { "epoch": 10.88, "learning_rate": 4.787570833815195e-05, "loss": 1.0439, "step": 125590 }, { "epoch": 10.88, "learning_rate": 4.787353995605412e-05, "loss": 1.1665, "step": 125600 }, { "epoch": 10.88, "learning_rate": 4.787137157395628e-05, "loss": 1.0944, "step": 125610 }, { "epoch": 10.88, "learning_rate": 4.7869203191858445e-05, "loss": 1.173, "step": 125620 }, { "epoch": 10.88, "learning_rate": 4.7867034809760605e-05, "loss": 1.2675, "step": 125630 }, { "epoch": 10.88, "learning_rate": 4.786486642766277e-05, "loss": 1.2027, "step": 125640 }, { "epoch": 10.88, "learning_rate": 4.786269804556494e-05, "loss": 1.2441, "step": 125650 }, { "epoch": 10.88, "learning_rate": 4.78605296634671e-05, "loss": 1.1197, "step": 125660 }, { "epoch": 10.88, "learning_rate": 4.785836128136926e-05, "loss": 1.1062, "step": 125670 }, { "epoch": 10.88, "learning_rate": 4.7856192899271416e-05, "loss": 1.2132, "step": 125680 }, { "epoch": 10.89, "learning_rate": 4.785402451717358e-05, "loss": 1.1264, "step": 125690 }, { "epoch": 10.89, "learning_rate": 4.785185613507574e-05, "loss": 1.2429, "step": 125700 }, { "epoch": 10.89, "learning_rate": 4.784968775297791e-05, "loss": 1.424, "step": 125710 }, { "epoch": 10.89, "learning_rate": 4.7847519370880075e-05, "loss": 1.2207, "step": 125720 }, { "epoch": 10.89, "learning_rate": 4.7845350988782235e-05, "loss": 1.0324, "step": 125730 }, { "epoch": 10.89, "learning_rate": 4.784318260668439e-05, "loss": 1.209, "step": 125740 }, { "epoch": 10.89, "learning_rate": 4.7841014224586554e-05, "loss": 1.3636, "step": 125750 }, { "epoch": 10.89, "learning_rate": 4.783884584248872e-05, "loss": 1.1349, "step": 125760 }, { "epoch": 10.89, "learning_rate": 4.783667746039088e-05, "loss": 1.0812, "step": 125770 }, { "epoch": 10.89, "learning_rate": 4.7834509078293046e-05, "loss": 1.1007, "step": 125780 }, { "epoch": 10.89, "learning_rate": 4.7832340696195206e-05, "loss": 1.0426, "step": 125790 }, { "epoch": 10.9, "learning_rate": 4.783017231409737e-05, "loss": 1.1296, "step": 125800 }, { "epoch": 10.9, "learning_rate": 4.782800393199954e-05, "loss": 0.9572, "step": 125810 }, { "epoch": 10.9, "learning_rate": 4.782583554990169e-05, "loss": 1.3587, "step": 125820 }, { "epoch": 10.9, "learning_rate": 4.782366716780386e-05, "loss": 1.2293, "step": 125830 }, { "epoch": 10.9, "learning_rate": 4.782149878570602e-05, "loss": 1.1598, "step": 125840 }, { "epoch": 10.9, "learning_rate": 4.7819330403608184e-05, "loss": 1.2964, "step": 125850 }, { "epoch": 10.9, "learning_rate": 4.7817162021510344e-05, "loss": 1.2659, "step": 125860 }, { "epoch": 10.9, "learning_rate": 4.781499363941251e-05, "loss": 1.209, "step": 125870 }, { "epoch": 10.9, "learning_rate": 4.7812825257314676e-05, "loss": 1.3256, "step": 125880 }, { "epoch": 10.9, "learning_rate": 4.7810656875216836e-05, "loss": 1.0842, "step": 125890 }, { "epoch": 10.9, "learning_rate": 4.7808488493118996e-05, "loss": 1.2017, "step": 125900 }, { "epoch": 10.9, "learning_rate": 4.7806320111021155e-05, "loss": 1.019, "step": 125910 }, { "epoch": 10.91, "learning_rate": 4.780415172892332e-05, "loss": 1.3115, "step": 125920 }, { "epoch": 10.91, "learning_rate": 4.780198334682548e-05, "loss": 1.096, "step": 125930 }, { "epoch": 10.91, "learning_rate": 4.779981496472765e-05, "loss": 1.237, "step": 125940 }, { "epoch": 10.91, "learning_rate": 4.7797646582629814e-05, "loss": 1.1895, "step": 125950 }, { "epoch": 10.91, "learning_rate": 4.7795478200531974e-05, "loss": 1.1151, "step": 125960 }, { "epoch": 10.91, "learning_rate": 4.779330981843414e-05, "loss": 1.2418, "step": 125970 }, { "epoch": 10.91, "learning_rate": 4.779114143633629e-05, "loss": 1.2076, "step": 125980 }, { "epoch": 10.91, "learning_rate": 4.778897305423846e-05, "loss": 1.2322, "step": 125990 }, { "epoch": 10.91, "learning_rate": 4.778680467214062e-05, "loss": 1.1881, "step": 126000 }, { "epoch": 10.91, "learning_rate": 4.7784636290042785e-05, "loss": 1.1687, "step": 126010 }, { "epoch": 10.91, "learning_rate": 4.778246790794495e-05, "loss": 1.2329, "step": 126020 }, { "epoch": 10.91, "learning_rate": 4.778029952584711e-05, "loss": 1.1484, "step": 126030 }, { "epoch": 10.92, "learning_rate": 4.777813114374928e-05, "loss": 1.347, "step": 126040 }, { "epoch": 10.92, "learning_rate": 4.777596276165143e-05, "loss": 1.0774, "step": 126050 }, { "epoch": 10.92, "learning_rate": 4.77737943795536e-05, "loss": 1.1248, "step": 126060 }, { "epoch": 10.92, "learning_rate": 4.7771625997455757e-05, "loss": 1.2184, "step": 126070 }, { "epoch": 10.92, "learning_rate": 4.776945761535792e-05, "loss": 1.2349, "step": 126080 }, { "epoch": 10.92, "learning_rate": 4.776728923326009e-05, "loss": 1.3002, "step": 126090 }, { "epoch": 10.92, "learning_rate": 4.776512085116225e-05, "loss": 1.0946, "step": 126100 }, { "epoch": 10.92, "learning_rate": 4.7762952469064415e-05, "loss": 1.0091, "step": 126110 }, { "epoch": 10.92, "learning_rate": 4.7760784086966575e-05, "loss": 1.2044, "step": 126120 }, { "epoch": 10.92, "learning_rate": 4.7758615704868735e-05, "loss": 1.1514, "step": 126130 }, { "epoch": 10.92, "learning_rate": 4.7756447322770894e-05, "loss": 1.0919, "step": 126140 }, { "epoch": 10.93, "learning_rate": 4.775427894067306e-05, "loss": 1.3128, "step": 126150 }, { "epoch": 10.93, "learning_rate": 4.775211055857523e-05, "loss": 1.2582, "step": 126160 }, { "epoch": 10.93, "learning_rate": 4.7749942176477387e-05, "loss": 1.2129, "step": 126170 }, { "epoch": 10.93, "learning_rate": 4.774777379437955e-05, "loss": 1.0829, "step": 126180 }, { "epoch": 10.93, "learning_rate": 4.774560541228171e-05, "loss": 1.124, "step": 126190 }, { "epoch": 10.93, "learning_rate": 4.774343703018388e-05, "loss": 1.0878, "step": 126200 }, { "epoch": 10.93, "learning_rate": 4.774126864808603e-05, "loss": 1.2369, "step": 126210 }, { "epoch": 10.93, "learning_rate": 4.77391002659882e-05, "loss": 1.1702, "step": 126220 }, { "epoch": 10.93, "learning_rate": 4.7736931883890365e-05, "loss": 1.1026, "step": 126230 }, { "epoch": 10.93, "learning_rate": 4.7734763501792524e-05, "loss": 1.2152, "step": 126240 }, { "epoch": 10.93, "learning_rate": 4.773259511969469e-05, "loss": 1.2209, "step": 126250 }, { "epoch": 10.93, "learning_rate": 4.773042673759685e-05, "loss": 1.4176, "step": 126260 }, { "epoch": 10.94, "learning_rate": 4.772825835549902e-05, "loss": 1.1733, "step": 126270 }, { "epoch": 10.94, "learning_rate": 4.772608997340117e-05, "loss": 1.2821, "step": 126280 }, { "epoch": 10.94, "learning_rate": 4.7723921591303336e-05, "loss": 1.2853, "step": 126290 }, { "epoch": 10.94, "learning_rate": 4.7721753209205495e-05, "loss": 1.0893, "step": 126300 }, { "epoch": 10.94, "learning_rate": 4.771958482710766e-05, "loss": 1.2175, "step": 126310 }, { "epoch": 10.94, "learning_rate": 4.771741644500983e-05, "loss": 1.2506, "step": 126320 }, { "epoch": 10.94, "learning_rate": 4.771524806291199e-05, "loss": 0.9941, "step": 126330 }, { "epoch": 10.94, "learning_rate": 4.7713079680814154e-05, "loss": 1.2996, "step": 126340 }, { "epoch": 10.94, "learning_rate": 4.7710911298716314e-05, "loss": 1.0229, "step": 126350 }, { "epoch": 10.94, "learning_rate": 4.7708742916618474e-05, "loss": 1.1253, "step": 126360 }, { "epoch": 10.94, "learning_rate": 4.770657453452063e-05, "loss": 0.8512, "step": 126370 }, { "epoch": 10.95, "learning_rate": 4.77044061524228e-05, "loss": 1.1308, "step": 126380 }, { "epoch": 10.95, "learning_rate": 4.7702237770324966e-05, "loss": 1.1285, "step": 126390 }, { "epoch": 10.95, "learning_rate": 4.7700069388227126e-05, "loss": 1.0827, "step": 126400 }, { "epoch": 10.95, "learning_rate": 4.769790100612929e-05, "loss": 1.2753, "step": 126410 }, { "epoch": 10.95, "learning_rate": 4.769573262403145e-05, "loss": 1.2724, "step": 126420 }, { "epoch": 10.95, "learning_rate": 4.769356424193362e-05, "loss": 1.1684, "step": 126430 }, { "epoch": 10.95, "learning_rate": 4.769139585983577e-05, "loss": 1.2541, "step": 126440 }, { "epoch": 10.95, "learning_rate": 4.768922747773794e-05, "loss": 1.1823, "step": 126450 }, { "epoch": 10.95, "learning_rate": 4.7687059095640104e-05, "loss": 1.2202, "step": 126460 }, { "epoch": 10.95, "learning_rate": 4.768489071354226e-05, "loss": 1.1424, "step": 126470 }, { "epoch": 10.95, "learning_rate": 4.768272233144443e-05, "loss": 1.204, "step": 126480 }, { "epoch": 10.95, "learning_rate": 4.768055394934659e-05, "loss": 1.1803, "step": 126490 }, { "epoch": 10.96, "learning_rate": 4.7678385567248756e-05, "loss": 1.3528, "step": 126500 }, { "epoch": 10.96, "learning_rate": 4.767621718515091e-05, "loss": 1.2102, "step": 126510 }, { "epoch": 10.96, "learning_rate": 4.7674048803053075e-05, "loss": 1.1764, "step": 126520 }, { "epoch": 10.96, "learning_rate": 4.767188042095524e-05, "loss": 1.3318, "step": 126530 }, { "epoch": 10.96, "learning_rate": 4.76697120388574e-05, "loss": 1.2643, "step": 126540 }, { "epoch": 10.96, "learning_rate": 4.766754365675957e-05, "loss": 1.1557, "step": 126550 }, { "epoch": 10.96, "learning_rate": 4.766537527466173e-05, "loss": 1.0423, "step": 126560 }, { "epoch": 10.96, "learning_rate": 4.766320689256389e-05, "loss": 1.1681, "step": 126570 }, { "epoch": 10.96, "learning_rate": 4.766103851046606e-05, "loss": 1.1283, "step": 126580 }, { "epoch": 10.96, "learning_rate": 4.765887012836821e-05, "loss": 1.1129, "step": 126590 }, { "epoch": 10.96, "learning_rate": 4.765670174627038e-05, "loss": 1.3628, "step": 126600 }, { "epoch": 10.97, "learning_rate": 4.765453336417254e-05, "loss": 1.1924, "step": 126610 }, { "epoch": 10.97, "learning_rate": 4.7652364982074705e-05, "loss": 1.1593, "step": 126620 }, { "epoch": 10.97, "learning_rate": 4.7650196599976864e-05, "loss": 1.0688, "step": 126630 }, { "epoch": 10.97, "learning_rate": 4.764802821787903e-05, "loss": 1.1917, "step": 126640 }, { "epoch": 10.97, "learning_rate": 4.76458598357812e-05, "loss": 1.2409, "step": 126650 }, { "epoch": 10.97, "learning_rate": 4.764369145368336e-05, "loss": 1.1993, "step": 126660 }, { "epoch": 10.97, "learning_rate": 4.7641523071585516e-05, "loss": 1.2245, "step": 126670 }, { "epoch": 10.97, "learning_rate": 4.7639354689487676e-05, "loss": 1.2008, "step": 126680 }, { "epoch": 10.97, "learning_rate": 4.763718630738984e-05, "loss": 1.0894, "step": 126690 }, { "epoch": 10.97, "learning_rate": 4.7635017925292e-05, "loss": 1.1294, "step": 126700 }, { "epoch": 10.97, "learning_rate": 4.763284954319417e-05, "loss": 1.1725, "step": 126710 }, { "epoch": 10.97, "learning_rate": 4.7630681161096335e-05, "loss": 1.1854, "step": 126720 }, { "epoch": 10.98, "learning_rate": 4.7628512778998494e-05, "loss": 1.0709, "step": 126730 }, { "epoch": 10.98, "learning_rate": 4.762634439690066e-05, "loss": 1.1882, "step": 126740 }, { "epoch": 10.98, "learning_rate": 4.7624176014802814e-05, "loss": 1.2652, "step": 126750 }, { "epoch": 10.98, "learning_rate": 4.762200763270498e-05, "loss": 1.1873, "step": 126760 }, { "epoch": 10.98, "learning_rate": 4.761983925060714e-05, "loss": 0.9049, "step": 126770 }, { "epoch": 10.98, "learning_rate": 4.7617670868509306e-05, "loss": 1.0775, "step": 126780 }, { "epoch": 10.98, "learning_rate": 4.761550248641147e-05, "loss": 1.3124, "step": 126790 }, { "epoch": 10.98, "learning_rate": 4.761333410431363e-05, "loss": 1.2593, "step": 126800 }, { "epoch": 10.98, "learning_rate": 4.76111657222158e-05, "loss": 1.221, "step": 126810 }, { "epoch": 10.98, "learning_rate": 4.760899734011795e-05, "loss": 1.2773, "step": 126820 }, { "epoch": 10.98, "learning_rate": 4.760682895802012e-05, "loss": 1.219, "step": 126830 }, { "epoch": 10.99, "learning_rate": 4.760466057592228e-05, "loss": 1.1613, "step": 126840 }, { "epoch": 10.99, "learning_rate": 4.7602492193824444e-05, "loss": 1.1388, "step": 126850 }, { "epoch": 10.99, "learning_rate": 4.76003238117266e-05, "loss": 1.041, "step": 126860 }, { "epoch": 10.99, "learning_rate": 4.759815542962877e-05, "loss": 1.0697, "step": 126870 }, { "epoch": 10.99, "learning_rate": 4.7595987047530936e-05, "loss": 1.1987, "step": 126880 }, { "epoch": 10.99, "learning_rate": 4.7593818665433096e-05, "loss": 1.3227, "step": 126890 }, { "epoch": 10.99, "learning_rate": 4.7591650283335255e-05, "loss": 1.1183, "step": 126900 }, { "epoch": 10.99, "learning_rate": 4.7589481901237415e-05, "loss": 1.2606, "step": 126910 }, { "epoch": 10.99, "learning_rate": 4.758731351913958e-05, "loss": 0.9542, "step": 126920 }, { "epoch": 10.99, "learning_rate": 4.758514513704174e-05, "loss": 1.1184, "step": 126930 }, { "epoch": 10.99, "learning_rate": 4.758297675494391e-05, "loss": 1.2555, "step": 126940 }, { "epoch": 10.99, "learning_rate": 4.7580808372846074e-05, "loss": 0.9468, "step": 126950 }, { "epoch": 11.0, "learning_rate": 4.7578639990748233e-05, "loss": 1.141, "step": 126960 }, { "epoch": 11.0, "learning_rate": 4.75764716086504e-05, "loss": 1.1206, "step": 126970 }, { "epoch": 11.0, "learning_rate": 4.757430322655255e-05, "loss": 1.4851, "step": 126980 }, { "epoch": 11.0, "learning_rate": 4.757213484445472e-05, "loss": 1.2166, "step": 126990 }, { "epoch": 11.0, "learning_rate": 4.756996646235688e-05, "loss": 1.29, "step": 127000 }, { "epoch": 11.0, "learning_rate": 4.7567798080259045e-05, "loss": 1.1027, "step": 127010 }, { "epoch": 11.0, "eval_cer": 0.5422527009836418, "eval_em": 0.03558535327488396, "eval_f1": 0.03558535327488396, "eval_loss": 1.4531387090682983, "eval_runtime": 257.1063, "eval_samples_per_second": 7.542, "eval_steps_per_second": 0.945, "eval_wer": 0.964414646725116, "step": 127011 }, { "epoch": 11.0, "learning_rate": 4.756562969816121e-05, "loss": 1.1198, "step": 127020 }, { "epoch": 11.0, "learning_rate": 4.756346131606337e-05, "loss": 1.1862, "step": 127030 }, { "epoch": 11.0, "learning_rate": 4.756129293396554e-05, "loss": 1.1614, "step": 127040 }, { "epoch": 11.0, "learning_rate": 4.755912455186769e-05, "loss": 1.1298, "step": 127050 }, { "epoch": 11.0, "learning_rate": 4.755695616976986e-05, "loss": 1.3459, "step": 127060 }, { "epoch": 11.01, "learning_rate": 4.7554787787672016e-05, "loss": 1.1455, "step": 127070 }, { "epoch": 11.01, "learning_rate": 4.755261940557418e-05, "loss": 1.2686, "step": 127080 }, { "epoch": 11.01, "learning_rate": 4.755045102347635e-05, "loss": 0.9653, "step": 127090 }, { "epoch": 11.01, "learning_rate": 4.754828264137851e-05, "loss": 1.2811, "step": 127100 }, { "epoch": 11.01, "learning_rate": 4.7546114259280675e-05, "loss": 1.2522, "step": 127110 }, { "epoch": 11.01, "learning_rate": 4.7543945877182835e-05, "loss": 1.025, "step": 127120 }, { "epoch": 11.01, "learning_rate": 4.7541777495084994e-05, "loss": 0.9896, "step": 127130 }, { "epoch": 11.01, "learning_rate": 4.7539609112987154e-05, "loss": 1.0929, "step": 127140 }, { "epoch": 11.01, "learning_rate": 4.753744073088932e-05, "loss": 1.0116, "step": 127150 }, { "epoch": 11.01, "learning_rate": 4.753527234879149e-05, "loss": 1.0885, "step": 127160 }, { "epoch": 11.01, "learning_rate": 4.7533103966693646e-05, "loss": 1.0884, "step": 127170 }, { "epoch": 11.01, "learning_rate": 4.753093558459581e-05, "loss": 1.1297, "step": 127180 }, { "epoch": 11.02, "learning_rate": 4.752876720249797e-05, "loss": 1.2248, "step": 127190 }, { "epoch": 11.02, "learning_rate": 4.752659882040014e-05, "loss": 1.027, "step": 127200 }, { "epoch": 11.02, "learning_rate": 4.752443043830229e-05, "loss": 0.9951, "step": 127210 }, { "epoch": 11.02, "learning_rate": 4.752226205620446e-05, "loss": 1.0854, "step": 127220 }, { "epoch": 11.02, "learning_rate": 4.7520093674106624e-05, "loss": 1.1308, "step": 127230 }, { "epoch": 11.02, "learning_rate": 4.7517925292008784e-05, "loss": 1.1051, "step": 127240 }, { "epoch": 11.02, "learning_rate": 4.751575690991095e-05, "loss": 0.9317, "step": 127250 }, { "epoch": 11.02, "learning_rate": 4.751358852781311e-05, "loss": 1.0618, "step": 127260 }, { "epoch": 11.02, "learning_rate": 4.7511420145715276e-05, "loss": 1.0844, "step": 127270 }, { "epoch": 11.02, "learning_rate": 4.750925176361743e-05, "loss": 1.1533, "step": 127280 }, { "epoch": 11.02, "learning_rate": 4.7507083381519596e-05, "loss": 1.1784, "step": 127290 }, { "epoch": 11.02, "learning_rate": 4.750491499942176e-05, "loss": 1.4086, "step": 127300 }, { "epoch": 11.03, "learning_rate": 4.750274661732392e-05, "loss": 1.1039, "step": 127310 }, { "epoch": 11.03, "learning_rate": 4.750057823522609e-05, "loss": 1.1233, "step": 127320 }, { "epoch": 11.03, "learning_rate": 4.749840985312825e-05, "loss": 1.0548, "step": 127330 }, { "epoch": 11.03, "learning_rate": 4.7496241471030414e-05, "loss": 0.9357, "step": 127340 }, { "epoch": 11.03, "learning_rate": 4.749407308893258e-05, "loss": 1.08, "step": 127350 }, { "epoch": 11.03, "learning_rate": 4.749190470683473e-05, "loss": 1.2186, "step": 127360 }, { "epoch": 11.03, "learning_rate": 4.748973632473689e-05, "loss": 1.1187, "step": 127370 }, { "epoch": 11.03, "learning_rate": 4.748756794263906e-05, "loss": 1.051, "step": 127380 }, { "epoch": 11.03, "learning_rate": 4.7485399560541226e-05, "loss": 1.1122, "step": 127390 }, { "epoch": 11.03, "learning_rate": 4.7483231178443385e-05, "loss": 1.0308, "step": 127400 }, { "epoch": 11.03, "learning_rate": 4.748106279634555e-05, "loss": 1.2002, "step": 127410 }, { "epoch": 11.04, "learning_rate": 4.747889441424771e-05, "loss": 1.115, "step": 127420 }, { "epoch": 11.04, "learning_rate": 4.747672603214988e-05, "loss": 1.1262, "step": 127430 }, { "epoch": 11.04, "learning_rate": 4.747455765005203e-05, "loss": 0.9308, "step": 127440 }, { "epoch": 11.04, "learning_rate": 4.74723892679542e-05, "loss": 1.1271, "step": 127450 }, { "epoch": 11.04, "learning_rate": 4.747022088585636e-05, "loss": 1.2538, "step": 127460 }, { "epoch": 11.04, "learning_rate": 4.746805250375852e-05, "loss": 1.0581, "step": 127470 }, { "epoch": 11.04, "learning_rate": 4.746588412166069e-05, "loss": 1.0565, "step": 127480 }, { "epoch": 11.04, "learning_rate": 4.746371573956285e-05, "loss": 1.181, "step": 127490 }, { "epoch": 11.04, "learning_rate": 4.7461547357465015e-05, "loss": 1.0921, "step": 127500 }, { "epoch": 11.04, "learning_rate": 4.745937897536718e-05, "loss": 1.1015, "step": 127510 }, { "epoch": 11.04, "learning_rate": 4.7457210593269334e-05, "loss": 1.1156, "step": 127520 }, { "epoch": 11.04, "learning_rate": 4.74550422111715e-05, "loss": 1.0825, "step": 127530 }, { "epoch": 11.05, "learning_rate": 4.745287382907366e-05, "loss": 1.0262, "step": 127540 }, { "epoch": 11.05, "learning_rate": 4.745070544697583e-05, "loss": 1.1468, "step": 127550 }, { "epoch": 11.05, "learning_rate": 4.7448537064877986e-05, "loss": 1.1267, "step": 127560 }, { "epoch": 11.05, "learning_rate": 4.744636868278015e-05, "loss": 0.9341, "step": 127570 }, { "epoch": 11.05, "learning_rate": 4.744420030068232e-05, "loss": 1.083, "step": 127580 }, { "epoch": 11.05, "learning_rate": 4.744203191858447e-05, "loss": 1.1048, "step": 127590 }, { "epoch": 11.05, "learning_rate": 4.743986353648664e-05, "loss": 1.2244, "step": 127600 }, { "epoch": 11.05, "learning_rate": 4.74376951543888e-05, "loss": 1.1423, "step": 127610 }, { "epoch": 11.05, "learning_rate": 4.7435526772290965e-05, "loss": 0.9609, "step": 127620 }, { "epoch": 11.05, "learning_rate": 4.7433358390193124e-05, "loss": 1.2294, "step": 127630 }, { "epoch": 11.05, "learning_rate": 4.743119000809529e-05, "loss": 1.1272, "step": 127640 }, { "epoch": 11.06, "learning_rate": 4.742902162599746e-05, "loss": 1.2423, "step": 127650 }, { "epoch": 11.06, "learning_rate": 4.7426853243899617e-05, "loss": 1.3648, "step": 127660 }, { "epoch": 11.06, "learning_rate": 4.7424684861801776e-05, "loss": 0.9979, "step": 127670 }, { "epoch": 11.06, "learning_rate": 4.7422516479703936e-05, "loss": 1.0574, "step": 127680 }, { "epoch": 11.06, "learning_rate": 4.74203480976061e-05, "loss": 0.9104, "step": 127690 }, { "epoch": 11.06, "learning_rate": 4.741817971550826e-05, "loss": 1.0867, "step": 127700 }, { "epoch": 11.06, "learning_rate": 4.741601133341043e-05, "loss": 1.0946, "step": 127710 }, { "epoch": 11.06, "learning_rate": 4.7413842951312595e-05, "loss": 1.1764, "step": 127720 }, { "epoch": 11.06, "learning_rate": 4.7411674569214754e-05, "loss": 0.9379, "step": 127730 }, { "epoch": 11.06, "learning_rate": 4.740950618711692e-05, "loss": 1.2207, "step": 127740 }, { "epoch": 11.06, "learning_rate": 4.7407337805019073e-05, "loss": 1.2019, "step": 127750 }, { "epoch": 11.06, "learning_rate": 4.740516942292124e-05, "loss": 1.0807, "step": 127760 }, { "epoch": 11.07, "learning_rate": 4.74030010408234e-05, "loss": 1.0917, "step": 127770 }, { "epoch": 11.07, "learning_rate": 4.7400832658725566e-05, "loss": 1.0601, "step": 127780 }, { "epoch": 11.07, "learning_rate": 4.739866427662773e-05, "loss": 1.1315, "step": 127790 }, { "epoch": 11.07, "learning_rate": 4.739649589452989e-05, "loss": 1.0178, "step": 127800 }, { "epoch": 11.07, "learning_rate": 4.739432751243206e-05, "loss": 1.1405, "step": 127810 }, { "epoch": 11.07, "learning_rate": 4.739215913033421e-05, "loss": 1.1892, "step": 127820 }, { "epoch": 11.07, "learning_rate": 4.738999074823638e-05, "loss": 1.0458, "step": 127830 }, { "epoch": 11.07, "learning_rate": 4.738782236613854e-05, "loss": 1.206, "step": 127840 }, { "epoch": 11.07, "learning_rate": 4.7385653984040703e-05, "loss": 1.02, "step": 127850 }, { "epoch": 11.07, "learning_rate": 4.738348560194287e-05, "loss": 1.0546, "step": 127860 }, { "epoch": 11.07, "learning_rate": 4.738131721984503e-05, "loss": 1.0791, "step": 127870 }, { "epoch": 11.08, "learning_rate": 4.7379148837747196e-05, "loss": 1.0491, "step": 127880 }, { "epoch": 11.08, "learning_rate": 4.7376980455649355e-05, "loss": 1.0152, "step": 127890 }, { "epoch": 11.08, "learning_rate": 4.7374812073551515e-05, "loss": 1.066, "step": 127900 }, { "epoch": 11.08, "learning_rate": 4.7372643691453675e-05, "loss": 1.1185, "step": 127910 }, { "epoch": 11.08, "learning_rate": 4.737047530935584e-05, "loss": 1.0535, "step": 127920 }, { "epoch": 11.08, "learning_rate": 4.7368306927258e-05, "loss": 1.3354, "step": 127930 }, { "epoch": 11.08, "learning_rate": 4.736613854516017e-05, "loss": 0.9709, "step": 127940 }, { "epoch": 11.08, "learning_rate": 4.7363970163062333e-05, "loss": 1.149, "step": 127950 }, { "epoch": 11.08, "learning_rate": 4.736180178096449e-05, "loss": 1.094, "step": 127960 }, { "epoch": 11.08, "learning_rate": 4.735963339886666e-05, "loss": 1.1866, "step": 127970 }, { "epoch": 11.08, "learning_rate": 4.735746501676881e-05, "loss": 1.1601, "step": 127980 }, { "epoch": 11.08, "learning_rate": 4.735529663467098e-05, "loss": 1.0249, "step": 127990 }, { "epoch": 11.09, "learning_rate": 4.735312825257314e-05, "loss": 1.2781, "step": 128000 }, { "epoch": 11.09, "learning_rate": 4.7350959870475305e-05, "loss": 1.4014, "step": 128010 }, { "epoch": 11.09, "learning_rate": 4.734879148837747e-05, "loss": 1.3339, "step": 128020 }, { "epoch": 11.09, "learning_rate": 4.734662310627963e-05, "loss": 1.0549, "step": 128030 }, { "epoch": 11.09, "learning_rate": 4.73444547241818e-05, "loss": 0.9763, "step": 128040 }, { "epoch": 11.09, "learning_rate": 4.734228634208395e-05, "loss": 1.0903, "step": 128050 }, { "epoch": 11.09, "learning_rate": 4.7340117959986116e-05, "loss": 1.0582, "step": 128060 }, { "epoch": 11.09, "learning_rate": 4.7337949577888276e-05, "loss": 1.047, "step": 128070 }, { "epoch": 11.09, "learning_rate": 4.733578119579044e-05, "loss": 1.0949, "step": 128080 }, { "epoch": 11.09, "learning_rate": 4.733361281369261e-05, "loss": 1.0825, "step": 128090 }, { "epoch": 11.09, "learning_rate": 4.733144443159477e-05, "loss": 1.1409, "step": 128100 }, { "epoch": 11.1, "learning_rate": 4.7329276049496935e-05, "loss": 0.9777, "step": 128110 }, { "epoch": 11.1, "learning_rate": 4.7327107667399094e-05, "loss": 1.0277, "step": 128120 }, { "epoch": 11.1, "learning_rate": 4.7324939285301254e-05, "loss": 1.2844, "step": 128130 }, { "epoch": 11.1, "learning_rate": 4.7322770903203414e-05, "loss": 1.0741, "step": 128140 }, { "epoch": 11.1, "learning_rate": 4.732060252110558e-05, "loss": 0.977, "step": 128150 }, { "epoch": 11.1, "learning_rate": 4.7318434139007746e-05, "loss": 1.0218, "step": 128160 }, { "epoch": 11.1, "learning_rate": 4.7316265756909906e-05, "loss": 1.1381, "step": 128170 }, { "epoch": 11.1, "learning_rate": 4.731409737481207e-05, "loss": 1.1934, "step": 128180 }, { "epoch": 11.1, "learning_rate": 4.731192899271423e-05, "loss": 1.0675, "step": 128190 }, { "epoch": 11.1, "learning_rate": 4.73097606106164e-05, "loss": 1.0905, "step": 128200 }, { "epoch": 11.1, "learning_rate": 4.730759222851855e-05, "loss": 1.058, "step": 128210 }, { "epoch": 11.1, "learning_rate": 4.730542384642072e-05, "loss": 1.1432, "step": 128220 }, { "epoch": 11.11, "learning_rate": 4.7303255464322884e-05, "loss": 0.9932, "step": 128230 }, { "epoch": 11.11, "learning_rate": 4.7301087082225044e-05, "loss": 1.2499, "step": 128240 }, { "epoch": 11.11, "learning_rate": 4.729891870012721e-05, "loss": 1.033, "step": 128250 }, { "epoch": 11.11, "learning_rate": 4.729675031802937e-05, "loss": 1.0939, "step": 128260 }, { "epoch": 11.11, "learning_rate": 4.7294581935931536e-05, "loss": 0.9863, "step": 128270 }, { "epoch": 11.11, "learning_rate": 4.72924135538337e-05, "loss": 1.0316, "step": 128280 }, { "epoch": 11.11, "learning_rate": 4.7290245171735855e-05, "loss": 1.2125, "step": 128290 }, { "epoch": 11.11, "learning_rate": 4.728807678963802e-05, "loss": 1.0206, "step": 128300 }, { "epoch": 11.11, "learning_rate": 4.728590840754018e-05, "loss": 0.931, "step": 128310 }, { "epoch": 11.11, "learning_rate": 4.728374002544235e-05, "loss": 1.0254, "step": 128320 }, { "epoch": 11.11, "learning_rate": 4.728157164334451e-05, "loss": 0.8744, "step": 128330 }, { "epoch": 11.12, "learning_rate": 4.7279403261246674e-05, "loss": 1.2154, "step": 128340 }, { "epoch": 11.12, "learning_rate": 4.727723487914884e-05, "loss": 1.1978, "step": 128350 }, { "epoch": 11.12, "learning_rate": 4.727506649705099e-05, "loss": 1.2298, "step": 128360 }, { "epoch": 11.12, "learning_rate": 4.727289811495316e-05, "loss": 1.1894, "step": 128370 }, { "epoch": 11.12, "learning_rate": 4.727072973285532e-05, "loss": 1.1675, "step": 128380 }, { "epoch": 11.12, "learning_rate": 4.7268561350757485e-05, "loss": 0.9883, "step": 128390 }, { "epoch": 11.12, "learning_rate": 4.7266392968659645e-05, "loss": 0.9861, "step": 128400 }, { "epoch": 11.12, "learning_rate": 4.726422458656181e-05, "loss": 1.1235, "step": 128410 }, { "epoch": 11.12, "learning_rate": 4.726205620446398e-05, "loss": 1.0992, "step": 128420 }, { "epoch": 11.12, "learning_rate": 4.725988782236614e-05, "loss": 1.1978, "step": 128430 }, { "epoch": 11.12, "learning_rate": 4.725771944026829e-05, "loss": 1.1722, "step": 128440 }, { "epoch": 11.12, "learning_rate": 4.7255551058170457e-05, "loss": 1.1536, "step": 128450 }, { "epoch": 11.13, "learning_rate": 4.725338267607262e-05, "loss": 0.9385, "step": 128460 }, { "epoch": 11.13, "learning_rate": 4.725121429397478e-05, "loss": 1.1602, "step": 128470 }, { "epoch": 11.13, "learning_rate": 4.724904591187695e-05, "loss": 1.0741, "step": 128480 }, { "epoch": 11.13, "learning_rate": 4.724687752977911e-05, "loss": 0.9811, "step": 128490 }, { "epoch": 11.13, "learning_rate": 4.7244709147681275e-05, "loss": 1.0189, "step": 128500 }, { "epoch": 11.13, "learning_rate": 4.724254076558344e-05, "loss": 1.0822, "step": 128510 }, { "epoch": 11.13, "learning_rate": 4.7240372383485594e-05, "loss": 0.9248, "step": 128520 }, { "epoch": 11.13, "learning_rate": 4.723820400138776e-05, "loss": 0.9911, "step": 128530 }, { "epoch": 11.13, "learning_rate": 4.723603561928992e-05, "loss": 1.1204, "step": 128540 }, { "epoch": 11.13, "learning_rate": 4.7233867237192087e-05, "loss": 1.1171, "step": 128550 }, { "epoch": 11.13, "learning_rate": 4.7231698855094246e-05, "loss": 1.0942, "step": 128560 }, { "epoch": 11.13, "learning_rate": 4.722953047299641e-05, "loss": 1.192, "step": 128570 }, { "epoch": 11.14, "learning_rate": 4.722736209089858e-05, "loss": 1.0626, "step": 128580 }, { "epoch": 11.14, "learning_rate": 4.722519370880073e-05, "loss": 0.8506, "step": 128590 }, { "epoch": 11.14, "learning_rate": 4.72230253267029e-05, "loss": 1.2153, "step": 128600 }, { "epoch": 11.14, "learning_rate": 4.722085694460506e-05, "loss": 1.122, "step": 128610 }, { "epoch": 11.14, "learning_rate": 4.7218688562507224e-05, "loss": 1.1445, "step": 128620 }, { "epoch": 11.14, "learning_rate": 4.7216520180409384e-05, "loss": 1.0906, "step": 128630 }, { "epoch": 11.14, "learning_rate": 4.721435179831155e-05, "loss": 1.2161, "step": 128640 }, { "epoch": 11.14, "learning_rate": 4.7212183416213717e-05, "loss": 1.0461, "step": 128650 }, { "epoch": 11.14, "learning_rate": 4.7210015034115876e-05, "loss": 1.1887, "step": 128660 }, { "epoch": 11.14, "learning_rate": 4.7207846652018036e-05, "loss": 0.974, "step": 128670 }, { "epoch": 11.14, "learning_rate": 4.7205678269920195e-05, "loss": 0.9804, "step": 128680 }, { "epoch": 11.15, "learning_rate": 4.720350988782236e-05, "loss": 1.0092, "step": 128690 }, { "epoch": 11.15, "learning_rate": 4.720134150572452e-05, "loss": 1.0464, "step": 128700 }, { "epoch": 11.15, "learning_rate": 4.719917312362669e-05, "loss": 1.1192, "step": 128710 }, { "epoch": 11.15, "learning_rate": 4.7197004741528854e-05, "loss": 1.0528, "step": 128720 }, { "epoch": 11.15, "learning_rate": 4.7194836359431014e-05, "loss": 0.9228, "step": 128730 }, { "epoch": 11.15, "learning_rate": 4.719266797733318e-05, "loss": 1.1167, "step": 128740 }, { "epoch": 11.15, "learning_rate": 4.719049959523533e-05, "loss": 1.0681, "step": 128750 }, { "epoch": 11.15, "learning_rate": 4.71883312131375e-05, "loss": 1.2483, "step": 128760 }, { "epoch": 11.15, "learning_rate": 4.718616283103966e-05, "loss": 1.0289, "step": 128770 }, { "epoch": 11.15, "learning_rate": 4.7183994448941825e-05, "loss": 1.1453, "step": 128780 }, { "epoch": 11.15, "learning_rate": 4.718182606684399e-05, "loss": 1.0025, "step": 128790 }, { "epoch": 11.15, "learning_rate": 4.717965768474615e-05, "loss": 1.0544, "step": 128800 }, { "epoch": 11.16, "learning_rate": 4.717748930264832e-05, "loss": 1.1071, "step": 128810 }, { "epoch": 11.16, "learning_rate": 4.717532092055047e-05, "loss": 1.0905, "step": 128820 }, { "epoch": 11.16, "learning_rate": 4.717315253845264e-05, "loss": 1.1969, "step": 128830 }, { "epoch": 11.16, "learning_rate": 4.71709841563548e-05, "loss": 0.8585, "step": 128840 }, { "epoch": 11.16, "learning_rate": 4.716881577425696e-05, "loss": 1.0896, "step": 128850 }, { "epoch": 11.16, "learning_rate": 4.716664739215913e-05, "loss": 1.2384, "step": 128860 }, { "epoch": 11.16, "learning_rate": 4.716447901006129e-05, "loss": 1.1086, "step": 128870 }, { "epoch": 11.16, "learning_rate": 4.7162310627963456e-05, "loss": 1.1438, "step": 128880 }, { "epoch": 11.16, "learning_rate": 4.7160142245865615e-05, "loss": 1.1726, "step": 128890 }, { "epoch": 11.16, "learning_rate": 4.7157973863767775e-05, "loss": 1.1261, "step": 128900 }, { "epoch": 11.16, "learning_rate": 4.7155805481669934e-05, "loss": 1.185, "step": 128910 }, { "epoch": 11.17, "learning_rate": 4.71536370995721e-05, "loss": 1.0564, "step": 128920 }, { "epoch": 11.17, "learning_rate": 4.715146871747427e-05, "loss": 1.0786, "step": 128930 }, { "epoch": 11.17, "learning_rate": 4.714930033537643e-05, "loss": 1.0593, "step": 128940 }, { "epoch": 11.17, "learning_rate": 4.714713195327859e-05, "loss": 1.1911, "step": 128950 }, { "epoch": 11.17, "learning_rate": 4.714496357118075e-05, "loss": 1.1639, "step": 128960 }, { "epoch": 11.17, "learning_rate": 4.714279518908292e-05, "loss": 0.9492, "step": 128970 }, { "epoch": 11.17, "learning_rate": 4.714062680698507e-05, "loss": 1.0967, "step": 128980 }, { "epoch": 11.17, "learning_rate": 4.713845842488724e-05, "loss": 1.045, "step": 128990 }, { "epoch": 11.17, "learning_rate": 4.71362900427894e-05, "loss": 1.1752, "step": 129000 }, { "epoch": 11.17, "learning_rate": 4.7134121660691564e-05, "loss": 1.322, "step": 129010 }, { "epoch": 11.17, "learning_rate": 4.713195327859373e-05, "loss": 1.116, "step": 129020 }, { "epoch": 11.17, "learning_rate": 4.712978489649589e-05, "loss": 1.177, "step": 129030 }, { "epoch": 11.18, "learning_rate": 4.712761651439806e-05, "loss": 1.178, "step": 129040 }, { "epoch": 11.18, "learning_rate": 4.7125448132300216e-05, "loss": 1.022, "step": 129050 }, { "epoch": 11.18, "learning_rate": 4.7123279750202376e-05, "loss": 1.2091, "step": 129060 }, { "epoch": 11.18, "learning_rate": 4.7121111368104536e-05, "loss": 1.0338, "step": 129070 }, { "epoch": 11.18, "learning_rate": 4.71189429860067e-05, "loss": 1.4382, "step": 129080 }, { "epoch": 11.18, "learning_rate": 4.711677460390887e-05, "loss": 1.0872, "step": 129090 }, { "epoch": 11.18, "learning_rate": 4.711460622181103e-05, "loss": 1.2018, "step": 129100 }, { "epoch": 11.18, "learning_rate": 4.7112437839713194e-05, "loss": 1.036, "step": 129110 }, { "epoch": 11.18, "learning_rate": 4.7110269457615354e-05, "loss": 0.9312, "step": 129120 }, { "epoch": 11.18, "learning_rate": 4.7108101075517514e-05, "loss": 0.981, "step": 129130 }, { "epoch": 11.18, "learning_rate": 4.710593269341967e-05, "loss": 0.9808, "step": 129140 }, { "epoch": 11.19, "learning_rate": 4.710376431132184e-05, "loss": 1.0618, "step": 129150 }, { "epoch": 11.19, "learning_rate": 4.7101595929224006e-05, "loss": 1.0019, "step": 129160 }, { "epoch": 11.19, "learning_rate": 4.7099427547126166e-05, "loss": 1.0697, "step": 129170 }, { "epoch": 11.19, "learning_rate": 4.709725916502833e-05, "loss": 1.0351, "step": 129180 }, { "epoch": 11.19, "learning_rate": 4.709509078293049e-05, "loss": 1.1583, "step": 129190 }, { "epoch": 11.19, "learning_rate": 4.709292240083266e-05, "loss": 1.1406, "step": 129200 }, { "epoch": 11.19, "learning_rate": 4.709075401873481e-05, "loss": 1.0637, "step": 129210 }, { "epoch": 11.19, "learning_rate": 4.708858563663698e-05, "loss": 0.9727, "step": 129220 }, { "epoch": 11.19, "learning_rate": 4.7086417254539144e-05, "loss": 1.0762, "step": 129230 }, { "epoch": 11.19, "learning_rate": 4.70842488724413e-05, "loss": 1.0881, "step": 129240 }, { "epoch": 11.19, "learning_rate": 4.708208049034347e-05, "loss": 1.0247, "step": 129250 }, { "epoch": 11.19, "learning_rate": 4.707991210824563e-05, "loss": 1.1264, "step": 129260 }, { "epoch": 11.2, "learning_rate": 4.7077743726147796e-05, "loss": 1.1902, "step": 129270 }, { "epoch": 11.2, "learning_rate": 4.707557534404996e-05, "loss": 1.0418, "step": 129280 }, { "epoch": 11.2, "learning_rate": 4.7073406961952115e-05, "loss": 1.2036, "step": 129290 }, { "epoch": 11.2, "learning_rate": 4.707123857985428e-05, "loss": 1.1208, "step": 129300 }, { "epoch": 11.2, "learning_rate": 4.706907019775644e-05, "loss": 1.2738, "step": 129310 }, { "epoch": 11.2, "learning_rate": 4.706690181565861e-05, "loss": 1.1163, "step": 129320 }, { "epoch": 11.2, "learning_rate": 4.706473343356077e-05, "loss": 1.0181, "step": 129330 }, { "epoch": 11.2, "learning_rate": 4.706256505146293e-05, "loss": 1.2699, "step": 129340 }, { "epoch": 11.2, "learning_rate": 4.70603966693651e-05, "loss": 1.1676, "step": 129350 }, { "epoch": 11.2, "learning_rate": 4.705822828726725e-05, "loss": 1.0601, "step": 129360 }, { "epoch": 11.2, "learning_rate": 4.705605990516942e-05, "loss": 1.1737, "step": 129370 }, { "epoch": 11.21, "learning_rate": 4.705389152307158e-05, "loss": 0.9036, "step": 129380 }, { "epoch": 11.21, "learning_rate": 4.7051723140973745e-05, "loss": 1.1989, "step": 129390 }, { "epoch": 11.21, "learning_rate": 4.7049554758875905e-05, "loss": 1.132, "step": 129400 }, { "epoch": 11.21, "learning_rate": 4.704738637677807e-05, "loss": 0.9239, "step": 129410 }, { "epoch": 11.21, "learning_rate": 4.704521799468024e-05, "loss": 1.1683, "step": 129420 }, { "epoch": 11.21, "learning_rate": 4.70430496125824e-05, "loss": 1.0979, "step": 129430 }, { "epoch": 11.21, "learning_rate": 4.704088123048456e-05, "loss": 1.1048, "step": 129440 }, { "epoch": 11.21, "learning_rate": 4.7038712848386716e-05, "loss": 1.0712, "step": 129450 }, { "epoch": 11.21, "learning_rate": 4.703654446628888e-05, "loss": 0.9863, "step": 129460 }, { "epoch": 11.21, "learning_rate": 4.703437608419104e-05, "loss": 0.9805, "step": 129470 }, { "epoch": 11.21, "learning_rate": 4.703220770209321e-05, "loss": 1.3317, "step": 129480 }, { "epoch": 11.21, "learning_rate": 4.7030039319995375e-05, "loss": 1.0211, "step": 129490 }, { "epoch": 11.22, "learning_rate": 4.7027870937897535e-05, "loss": 1.2545, "step": 129500 }, { "epoch": 11.22, "learning_rate": 4.70257025557997e-05, "loss": 1.2488, "step": 129510 }, { "epoch": 11.22, "learning_rate": 4.7023534173701854e-05, "loss": 1.067, "step": 129520 }, { "epoch": 11.22, "learning_rate": 4.702136579160402e-05, "loss": 1.1665, "step": 129530 }, { "epoch": 11.22, "learning_rate": 4.701919740950618e-05, "loss": 1.0127, "step": 129540 }, { "epoch": 11.22, "learning_rate": 4.7017029027408346e-05, "loss": 1.042, "step": 129550 }, { "epoch": 11.22, "learning_rate": 4.7014860645310506e-05, "loss": 1.0041, "step": 129560 }, { "epoch": 11.22, "learning_rate": 4.701269226321267e-05, "loss": 1.0092, "step": 129570 }, { "epoch": 11.22, "learning_rate": 4.701052388111484e-05, "loss": 1.1017, "step": 129580 }, { "epoch": 11.22, "learning_rate": 4.700835549901699e-05, "loss": 1.216, "step": 129590 }, { "epoch": 11.22, "learning_rate": 4.700618711691916e-05, "loss": 0.8732, "step": 129600 }, { "epoch": 11.23, "learning_rate": 4.700401873482132e-05, "loss": 0.9743, "step": 129610 }, { "epoch": 11.23, "learning_rate": 4.7001850352723484e-05, "loss": 1.099, "step": 129620 }, { "epoch": 11.23, "learning_rate": 4.6999681970625644e-05, "loss": 1.0946, "step": 129630 }, { "epoch": 11.23, "learning_rate": 4.699751358852781e-05, "loss": 1.1713, "step": 129640 }, { "epoch": 11.23, "learning_rate": 4.6995345206429976e-05, "loss": 0.9854, "step": 129650 }, { "epoch": 11.23, "learning_rate": 4.6993176824332136e-05, "loss": 1.2402, "step": 129660 }, { "epoch": 11.23, "learning_rate": 4.6991008442234296e-05, "loss": 1.14, "step": 129670 }, { "epoch": 11.23, "learning_rate": 4.6988840060136455e-05, "loss": 1.2409, "step": 129680 }, { "epoch": 11.23, "learning_rate": 4.698667167803862e-05, "loss": 1.0841, "step": 129690 }, { "epoch": 11.23, "learning_rate": 4.698450329594078e-05, "loss": 1.0236, "step": 129700 }, { "epoch": 11.23, "learning_rate": 4.698233491384295e-05, "loss": 1.067, "step": 129710 }, { "epoch": 11.23, "learning_rate": 4.6980166531745114e-05, "loss": 0.931, "step": 129720 }, { "epoch": 11.24, "learning_rate": 4.6977998149647274e-05, "loss": 1.1218, "step": 129730 }, { "epoch": 11.24, "learning_rate": 4.697582976754944e-05, "loss": 1.1969, "step": 129740 }, { "epoch": 11.24, "learning_rate": 4.697366138545159e-05, "loss": 1.134, "step": 129750 }, { "epoch": 11.24, "learning_rate": 4.697149300335376e-05, "loss": 1.1093, "step": 129760 }, { "epoch": 11.24, "learning_rate": 4.696932462125592e-05, "loss": 1.0759, "step": 129770 }, { "epoch": 11.24, "learning_rate": 4.6967156239158085e-05, "loss": 1.1042, "step": 129780 }, { "epoch": 11.24, "learning_rate": 4.696498785706025e-05, "loss": 1.2804, "step": 129790 }, { "epoch": 11.24, "learning_rate": 4.696281947496241e-05, "loss": 1.1594, "step": 129800 }, { "epoch": 11.24, "learning_rate": 4.696065109286458e-05, "loss": 1.0631, "step": 129810 }, { "epoch": 11.24, "learning_rate": 4.695848271076674e-05, "loss": 0.9411, "step": 129820 }, { "epoch": 11.24, "learning_rate": 4.69563143286689e-05, "loss": 1.0806, "step": 129830 }, { "epoch": 11.24, "learning_rate": 4.6954145946571056e-05, "loss": 0.9437, "step": 129840 }, { "epoch": 11.25, "learning_rate": 4.695197756447322e-05, "loss": 1.1644, "step": 129850 }, { "epoch": 11.25, "learning_rate": 4.694980918237539e-05, "loss": 0.9585, "step": 129860 }, { "epoch": 11.25, "learning_rate": 4.694764080027755e-05, "loss": 0.9831, "step": 129870 }, { "epoch": 11.25, "learning_rate": 4.6945472418179715e-05, "loss": 1.0378, "step": 129880 }, { "epoch": 11.25, "learning_rate": 4.6943304036081875e-05, "loss": 1.0784, "step": 129890 }, { "epoch": 11.25, "learning_rate": 4.6941135653984034e-05, "loss": 1.0781, "step": 129900 }, { "epoch": 11.25, "learning_rate": 4.6938967271886194e-05, "loss": 1.1703, "step": 129910 }, { "epoch": 11.25, "learning_rate": 4.693679888978836e-05, "loss": 1.3529, "step": 129920 }, { "epoch": 11.25, "learning_rate": 4.693463050769053e-05, "loss": 0.9315, "step": 129930 }, { "epoch": 11.25, "learning_rate": 4.6932462125592686e-05, "loss": 0.8844, "step": 129940 }, { "epoch": 11.25, "learning_rate": 4.693029374349485e-05, "loss": 1.0765, "step": 129950 }, { "epoch": 11.26, "learning_rate": 4.692812536139701e-05, "loss": 0.8876, "step": 129960 }, { "epoch": 11.26, "learning_rate": 4.692595697929918e-05, "loss": 1.2487, "step": 129970 }, { "epoch": 11.26, "learning_rate": 4.692378859720133e-05, "loss": 1.002, "step": 129980 }, { "epoch": 11.26, "learning_rate": 4.69216202151035e-05, "loss": 0.9629, "step": 129990 }, { "epoch": 11.26, "learning_rate": 4.6919451833005664e-05, "loss": 1.0101, "step": 130000 }, { "epoch": 11.26, "learning_rate": 4.6917283450907824e-05, "loss": 1.2702, "step": 130010 }, { "epoch": 11.26, "learning_rate": 4.691511506880999e-05, "loss": 1.1872, "step": 130020 }, { "epoch": 11.26, "learning_rate": 4.691294668671215e-05, "loss": 1.1586, "step": 130030 }, { "epoch": 11.26, "learning_rate": 4.6910778304614316e-05, "loss": 1.1654, "step": 130040 }, { "epoch": 11.26, "learning_rate": 4.690860992251648e-05, "loss": 1.0836, "step": 130050 }, { "epoch": 11.26, "learning_rate": 4.6906441540418636e-05, "loss": 0.9753, "step": 130060 }, { "epoch": 11.26, "learning_rate": 4.6904273158320795e-05, "loss": 1.0084, "step": 130070 }, { "epoch": 11.27, "learning_rate": 4.690210477622296e-05, "loss": 1.2017, "step": 130080 }, { "epoch": 11.27, "learning_rate": 4.689993639412513e-05, "loss": 1.1876, "step": 130090 }, { "epoch": 11.27, "learning_rate": 4.689776801202729e-05, "loss": 1.0623, "step": 130100 }, { "epoch": 11.27, "learning_rate": 4.6895599629929454e-05, "loss": 1.0406, "step": 130110 }, { "epoch": 11.27, "learning_rate": 4.6893431247831614e-05, "loss": 1.0085, "step": 130120 }, { "epoch": 11.27, "learning_rate": 4.689126286573377e-05, "loss": 0.972, "step": 130130 }, { "epoch": 11.27, "learning_rate": 4.688909448363593e-05, "loss": 1.2327, "step": 130140 }, { "epoch": 11.27, "learning_rate": 4.68869261015381e-05, "loss": 1.2286, "step": 130150 }, { "epoch": 11.27, "learning_rate": 4.6884757719440266e-05, "loss": 1.1571, "step": 130160 }, { "epoch": 11.27, "learning_rate": 4.6882589337342425e-05, "loss": 1.0553, "step": 130170 }, { "epoch": 11.27, "learning_rate": 4.688042095524459e-05, "loss": 1.0687, "step": 130180 }, { "epoch": 11.28, "learning_rate": 4.687825257314675e-05, "loss": 1.1563, "step": 130190 }, { "epoch": 11.28, "learning_rate": 4.687608419104892e-05, "loss": 1.1185, "step": 130200 }, { "epoch": 11.28, "learning_rate": 4.687391580895107e-05, "loss": 1.1737, "step": 130210 }, { "epoch": 11.28, "learning_rate": 4.687174742685324e-05, "loss": 1.2963, "step": 130220 }, { "epoch": 11.28, "learning_rate": 4.6869579044755403e-05, "loss": 1.1513, "step": 130230 }, { "epoch": 11.28, "learning_rate": 4.686741066265756e-05, "loss": 1.0726, "step": 130240 }, { "epoch": 11.28, "learning_rate": 4.686524228055973e-05, "loss": 1.0611, "step": 130250 }, { "epoch": 11.28, "learning_rate": 4.686307389846189e-05, "loss": 1.1046, "step": 130260 }, { "epoch": 11.28, "learning_rate": 4.6860905516364055e-05, "loss": 1.1411, "step": 130270 }, { "epoch": 11.28, "learning_rate": 4.685873713426622e-05, "loss": 1.026, "step": 130280 }, { "epoch": 11.28, "learning_rate": 4.6856568752168375e-05, "loss": 1.2528, "step": 130290 }, { "epoch": 11.28, "learning_rate": 4.685440037007054e-05, "loss": 1.2079, "step": 130300 }, { "epoch": 11.29, "learning_rate": 4.68522319879727e-05, "loss": 1.1358, "step": 130310 }, { "epoch": 11.29, "learning_rate": 4.685006360587487e-05, "loss": 1.1442, "step": 130320 }, { "epoch": 11.29, "learning_rate": 4.684789522377703e-05, "loss": 1.1878, "step": 130330 }, { "epoch": 11.29, "learning_rate": 4.684572684167919e-05, "loss": 1.0438, "step": 130340 }, { "epoch": 11.29, "learning_rate": 4.684355845958136e-05, "loss": 1.124, "step": 130350 }, { "epoch": 11.29, "learning_rate": 4.684139007748351e-05, "loss": 1.0893, "step": 130360 }, { "epoch": 11.29, "learning_rate": 4.683922169538568e-05, "loss": 1.1819, "step": 130370 }, { "epoch": 11.29, "learning_rate": 4.683705331328784e-05, "loss": 1.1678, "step": 130380 }, { "epoch": 11.29, "learning_rate": 4.6834884931190005e-05, "loss": 1.1282, "step": 130390 }, { "epoch": 11.29, "learning_rate": 4.6832716549092164e-05, "loss": 1.278, "step": 130400 }, { "epoch": 11.29, "learning_rate": 4.683054816699433e-05, "loss": 1.137, "step": 130410 }, { "epoch": 11.3, "learning_rate": 4.68283797848965e-05, "loss": 1.0313, "step": 130420 }, { "epoch": 11.3, "learning_rate": 4.682621140279866e-05, "loss": 1.034, "step": 130430 }, { "epoch": 11.3, "learning_rate": 4.6824043020700816e-05, "loss": 1.1707, "step": 130440 }, { "epoch": 11.3, "learning_rate": 4.6821874638602976e-05, "loss": 1.1351, "step": 130450 }, { "epoch": 11.3, "learning_rate": 4.681970625650514e-05, "loss": 1.0734, "step": 130460 }, { "epoch": 11.3, "learning_rate": 4.68175378744073e-05, "loss": 0.9223, "step": 130470 }, { "epoch": 11.3, "learning_rate": 4.681536949230947e-05, "loss": 1.1302, "step": 130480 }, { "epoch": 11.3, "learning_rate": 4.6813201110211635e-05, "loss": 1.0941, "step": 130490 }, { "epoch": 11.3, "learning_rate": 4.6811032728113794e-05, "loss": 1.2326, "step": 130500 }, { "epoch": 11.3, "learning_rate": 4.680886434601596e-05, "loss": 0.9062, "step": 130510 }, { "epoch": 11.3, "learning_rate": 4.6806695963918114e-05, "loss": 1.2447, "step": 130520 }, { "epoch": 11.3, "learning_rate": 4.680452758182028e-05, "loss": 1.2367, "step": 130530 }, { "epoch": 11.31, "learning_rate": 4.680235919972244e-05, "loss": 0.9439, "step": 130540 }, { "epoch": 11.31, "learning_rate": 4.6800190817624606e-05, "loss": 1.0662, "step": 130550 }, { "epoch": 11.31, "learning_rate": 4.679802243552677e-05, "loss": 1.0885, "step": 130560 }, { "epoch": 11.31, "learning_rate": 4.679585405342893e-05, "loss": 1.2348, "step": 130570 }, { "epoch": 11.31, "learning_rate": 4.67936856713311e-05, "loss": 0.9811, "step": 130580 }, { "epoch": 11.31, "learning_rate": 4.679151728923325e-05, "loss": 1.1828, "step": 130590 }, { "epoch": 11.31, "learning_rate": 4.678934890713542e-05, "loss": 1.4635, "step": 130600 }, { "epoch": 11.31, "learning_rate": 4.678718052503758e-05, "loss": 1.1865, "step": 130610 }, { "epoch": 11.31, "learning_rate": 4.6785012142939744e-05, "loss": 1.1171, "step": 130620 }, { "epoch": 11.31, "learning_rate": 4.67828437608419e-05, "loss": 1.3144, "step": 130630 }, { "epoch": 11.31, "learning_rate": 4.678067537874407e-05, "loss": 1.0281, "step": 130640 }, { "epoch": 11.32, "learning_rate": 4.6778506996646236e-05, "loss": 1.1829, "step": 130650 }, { "epoch": 11.32, "learning_rate": 4.6776338614548396e-05, "loss": 1.0984, "step": 130660 }, { "epoch": 11.32, "learning_rate": 4.6774170232450555e-05, "loss": 1.1878, "step": 130670 }, { "epoch": 11.32, "learning_rate": 4.6772001850352715e-05, "loss": 1.1144, "step": 130680 }, { "epoch": 11.32, "learning_rate": 4.676983346825488e-05, "loss": 1.0887, "step": 130690 }, { "epoch": 11.32, "learning_rate": 4.676766508615704e-05, "loss": 1.1231, "step": 130700 }, { "epoch": 11.32, "learning_rate": 4.676549670405921e-05, "loss": 1.0845, "step": 130710 }, { "epoch": 11.32, "learning_rate": 4.6763328321961374e-05, "loss": 1.2356, "step": 130720 }, { "epoch": 11.32, "learning_rate": 4.676115993986353e-05, "loss": 1.0792, "step": 130730 }, { "epoch": 11.32, "learning_rate": 4.67589915577657e-05, "loss": 1.0024, "step": 130740 }, { "epoch": 11.32, "learning_rate": 4.675682317566785e-05, "loss": 1.1426, "step": 130750 }, { "epoch": 11.32, "learning_rate": 4.675465479357002e-05, "loss": 1.2521, "step": 130760 }, { "epoch": 11.33, "learning_rate": 4.675248641147218e-05, "loss": 0.9509, "step": 130770 }, { "epoch": 11.33, "learning_rate": 4.6750318029374345e-05, "loss": 1.1102, "step": 130780 }, { "epoch": 11.33, "learning_rate": 4.674814964727651e-05, "loss": 1.0515, "step": 130790 }, { "epoch": 11.33, "learning_rate": 4.674598126517867e-05, "loss": 1.0826, "step": 130800 }, { "epoch": 11.33, "learning_rate": 4.674381288308084e-05, "loss": 1.2067, "step": 130810 }, { "epoch": 11.33, "learning_rate": 4.6741644500983e-05, "loss": 1.1883, "step": 130820 }, { "epoch": 11.33, "learning_rate": 4.6739476118885157e-05, "loss": 1.1381, "step": 130830 }, { "epoch": 11.33, "learning_rate": 4.6737307736787316e-05, "loss": 1.1839, "step": 130840 }, { "epoch": 11.33, "learning_rate": 4.673513935468948e-05, "loss": 1.0703, "step": 130850 }, { "epoch": 11.33, "learning_rate": 4.673297097259165e-05, "loss": 1.0064, "step": 130860 }, { "epoch": 11.33, "learning_rate": 4.673080259049381e-05, "loss": 1.0484, "step": 130870 }, { "epoch": 11.34, "learning_rate": 4.6728634208395975e-05, "loss": 0.9849, "step": 130880 }, { "epoch": 11.34, "learning_rate": 4.6726465826298135e-05, "loss": 1.1198, "step": 130890 }, { "epoch": 11.34, "learning_rate": 4.6724297444200294e-05, "loss": 1.0832, "step": 130900 }, { "epoch": 11.34, "learning_rate": 4.6722129062102454e-05, "loss": 1.0469, "step": 130910 }, { "epoch": 11.34, "learning_rate": 4.671996068000462e-05, "loss": 1.0313, "step": 130920 }, { "epoch": 11.34, "learning_rate": 4.6717792297906787e-05, "loss": 1.1732, "step": 130930 }, { "epoch": 11.34, "learning_rate": 4.6715623915808946e-05, "loss": 1.2582, "step": 130940 }, { "epoch": 11.34, "learning_rate": 4.671345553371111e-05, "loss": 1.0491, "step": 130950 }, { "epoch": 11.34, "learning_rate": 4.671128715161327e-05, "loss": 1.0649, "step": 130960 }, { "epoch": 11.34, "learning_rate": 4.670911876951544e-05, "loss": 1.1539, "step": 130970 }, { "epoch": 11.34, "learning_rate": 4.670695038741759e-05, "loss": 1.2637, "step": 130980 }, { "epoch": 11.34, "learning_rate": 4.670478200531976e-05, "loss": 1.1079, "step": 130990 }, { "epoch": 11.35, "learning_rate": 4.6702613623221924e-05, "loss": 1.1611, "step": 131000 }, { "epoch": 11.35, "learning_rate": 4.6700445241124084e-05, "loss": 1.2416, "step": 131010 }, { "epoch": 11.35, "learning_rate": 4.669827685902625e-05, "loss": 0.9739, "step": 131020 }, { "epoch": 11.35, "learning_rate": 4.669610847692841e-05, "loss": 1.071, "step": 131030 }, { "epoch": 11.35, "learning_rate": 4.6693940094830576e-05, "loss": 1.1181, "step": 131040 }, { "epoch": 11.35, "learning_rate": 4.669177171273274e-05, "loss": 1.1381, "step": 131050 }, { "epoch": 11.35, "learning_rate": 4.6689603330634895e-05, "loss": 1.2011, "step": 131060 }, { "epoch": 11.35, "learning_rate": 4.668743494853706e-05, "loss": 0.969, "step": 131070 }, { "epoch": 11.35, "learning_rate": 4.668526656643922e-05, "loss": 1.1125, "step": 131080 }, { "epoch": 11.35, "learning_rate": 4.668309818434139e-05, "loss": 1.1892, "step": 131090 }, { "epoch": 11.35, "learning_rate": 4.668092980224355e-05, "loss": 1.1917, "step": 131100 }, { "epoch": 11.35, "learning_rate": 4.6678761420145714e-05, "loss": 1.0169, "step": 131110 }, { "epoch": 11.36, "learning_rate": 4.667659303804788e-05, "loss": 1.0614, "step": 131120 }, { "epoch": 11.36, "learning_rate": 4.667442465595003e-05, "loss": 1.1699, "step": 131130 }, { "epoch": 11.36, "learning_rate": 4.667225627385219e-05, "loss": 1.134, "step": 131140 }, { "epoch": 11.36, "learning_rate": 4.667008789175436e-05, "loss": 1.0434, "step": 131150 }, { "epoch": 11.36, "learning_rate": 4.6667919509656525e-05, "loss": 1.3044, "step": 131160 }, { "epoch": 11.36, "learning_rate": 4.6665751127558685e-05, "loss": 1.0908, "step": 131170 }, { "epoch": 11.36, "learning_rate": 4.666358274546085e-05, "loss": 1.1291, "step": 131180 }, { "epoch": 11.36, "learning_rate": 4.666141436336301e-05, "loss": 1.1565, "step": 131190 }, { "epoch": 11.36, "learning_rate": 4.665924598126518e-05, "loss": 0.9109, "step": 131200 }, { "epoch": 11.36, "learning_rate": 4.665707759916733e-05, "loss": 1.1875, "step": 131210 }, { "epoch": 11.36, "learning_rate": 4.66549092170695e-05, "loss": 1.0337, "step": 131220 }, { "epoch": 11.37, "learning_rate": 4.665274083497166e-05, "loss": 1.2552, "step": 131230 }, { "epoch": 11.37, "learning_rate": 4.665057245287382e-05, "loss": 1.0562, "step": 131240 }, { "epoch": 11.37, "learning_rate": 4.664840407077599e-05, "loss": 1.1791, "step": 131250 }, { "epoch": 11.37, "learning_rate": 4.664623568867815e-05, "loss": 1.1364, "step": 131260 }, { "epoch": 11.37, "learning_rate": 4.6644067306580315e-05, "loss": 1.1386, "step": 131270 }, { "epoch": 11.37, "learning_rate": 4.664189892448248e-05, "loss": 1.1812, "step": 131280 }, { "epoch": 11.37, "learning_rate": 4.6639730542384634e-05, "loss": 1.235, "step": 131290 }, { "epoch": 11.37, "learning_rate": 4.66375621602868e-05, "loss": 1.1509, "step": 131300 }, { "epoch": 11.37, "learning_rate": 4.663539377818896e-05, "loss": 1.0495, "step": 131310 }, { "epoch": 11.37, "learning_rate": 4.663322539609113e-05, "loss": 1.1256, "step": 131320 }, { "epoch": 11.37, "learning_rate": 4.6631057013993286e-05, "loss": 1.2366, "step": 131330 }, { "epoch": 11.37, "learning_rate": 4.662888863189545e-05, "loss": 1.1177, "step": 131340 }, { "epoch": 11.38, "learning_rate": 4.662672024979762e-05, "loss": 1.1575, "step": 131350 }, { "epoch": 11.38, "learning_rate": 4.662455186769977e-05, "loss": 1.1885, "step": 131360 }, { "epoch": 11.38, "learning_rate": 4.662238348560194e-05, "loss": 1.1378, "step": 131370 }, { "epoch": 11.38, "learning_rate": 4.66202151035041e-05, "loss": 0.984, "step": 131380 }, { "epoch": 11.38, "learning_rate": 4.6618046721406264e-05, "loss": 1.1273, "step": 131390 }, { "epoch": 11.38, "learning_rate": 4.6615878339308424e-05, "loss": 1.1082, "step": 131400 }, { "epoch": 11.38, "learning_rate": 4.661370995721059e-05, "loss": 1.1924, "step": 131410 }, { "epoch": 11.38, "learning_rate": 4.661154157511276e-05, "loss": 1.0867, "step": 131420 }, { "epoch": 11.38, "learning_rate": 4.6609373193014916e-05, "loss": 1.1138, "step": 131430 }, { "epoch": 11.38, "learning_rate": 4.6607204810917076e-05, "loss": 1.2072, "step": 131440 }, { "epoch": 11.38, "learning_rate": 4.6605036428819236e-05, "loss": 1.0372, "step": 131450 }, { "epoch": 11.39, "learning_rate": 4.66028680467214e-05, "loss": 1.1317, "step": 131460 }, { "epoch": 11.39, "learning_rate": 4.660069966462356e-05, "loss": 1.1456, "step": 131470 }, { "epoch": 11.39, "learning_rate": 4.659853128252573e-05, "loss": 1.2349, "step": 131480 }, { "epoch": 11.39, "learning_rate": 4.6596362900427894e-05, "loss": 1.1424, "step": 131490 }, { "epoch": 11.39, "learning_rate": 4.6594194518330054e-05, "loss": 0.9685, "step": 131500 }, { "epoch": 11.39, "learning_rate": 4.659202613623222e-05, "loss": 0.9909, "step": 131510 }, { "epoch": 11.39, "learning_rate": 4.658985775413437e-05, "loss": 1.1695, "step": 131520 }, { "epoch": 11.39, "learning_rate": 4.658768937203654e-05, "loss": 0.9194, "step": 131530 }, { "epoch": 11.39, "learning_rate": 4.65855209899387e-05, "loss": 0.9552, "step": 131540 }, { "epoch": 11.39, "learning_rate": 4.6583352607840866e-05, "loss": 0.9506, "step": 131550 }, { "epoch": 11.39, "learning_rate": 4.658118422574303e-05, "loss": 1.1113, "step": 131560 }, { "epoch": 11.39, "learning_rate": 4.657901584364519e-05, "loss": 1.1625, "step": 131570 }, { "epoch": 11.4, "learning_rate": 4.657684746154736e-05, "loss": 0.9004, "step": 131580 }, { "epoch": 11.4, "learning_rate": 4.657467907944952e-05, "loss": 1.0415, "step": 131590 }, { "epoch": 11.4, "learning_rate": 4.657251069735168e-05, "loss": 1.0146, "step": 131600 }, { "epoch": 11.4, "learning_rate": 4.657034231525384e-05, "loss": 1.0341, "step": 131610 }, { "epoch": 11.4, "learning_rate": 4.6568173933156e-05, "loss": 1.2283, "step": 131620 }, { "epoch": 11.4, "learning_rate": 4.656600555105817e-05, "loss": 1.2275, "step": 131630 }, { "epoch": 11.4, "learning_rate": 4.656383716896033e-05, "loss": 1.179, "step": 131640 }, { "epoch": 11.4, "learning_rate": 4.6561668786862496e-05, "loss": 1.1043, "step": 131650 }, { "epoch": 11.4, "learning_rate": 4.6559500404764655e-05, "loss": 1.1576, "step": 131660 }, { "epoch": 11.4, "learning_rate": 4.6557332022666815e-05, "loss": 1.0645, "step": 131670 }, { "epoch": 11.4, "learning_rate": 4.6555163640568975e-05, "loss": 1.0011, "step": 131680 }, { "epoch": 11.41, "learning_rate": 4.655299525847114e-05, "loss": 1.1349, "step": 131690 }, { "epoch": 11.41, "learning_rate": 4.65508268763733e-05, "loss": 1.2374, "step": 131700 }, { "epoch": 11.41, "learning_rate": 4.654865849427547e-05, "loss": 1.0628, "step": 131710 }, { "epoch": 11.41, "learning_rate": 4.654649011217763e-05, "loss": 1.1812, "step": 131720 }, { "epoch": 11.41, "learning_rate": 4.654432173007979e-05, "loss": 1.0214, "step": 131730 }, { "epoch": 11.41, "learning_rate": 4.654215334798196e-05, "loss": 1.1907, "step": 131740 }, { "epoch": 11.41, "learning_rate": 4.653998496588411e-05, "loss": 0.9236, "step": 131750 }, { "epoch": 11.41, "learning_rate": 4.653781658378628e-05, "loss": 0.9802, "step": 131760 }, { "epoch": 11.41, "learning_rate": 4.653564820168844e-05, "loss": 0.9663, "step": 131770 }, { "epoch": 11.41, "learning_rate": 4.6533479819590605e-05, "loss": 1.1468, "step": 131780 }, { "epoch": 11.41, "learning_rate": 4.653131143749277e-05, "loss": 1.0002, "step": 131790 }, { "epoch": 11.41, "learning_rate": 4.652914305539493e-05, "loss": 1.2182, "step": 131800 }, { "epoch": 11.42, "learning_rate": 4.65269746732971e-05, "loss": 1.1963, "step": 131810 }, { "epoch": 11.42, "learning_rate": 4.6524806291199257e-05, "loss": 0.9884, "step": 131820 }, { "epoch": 11.42, "learning_rate": 4.6522637909101416e-05, "loss": 0.9093, "step": 131830 }, { "epoch": 11.42, "learning_rate": 4.6520469527003576e-05, "loss": 1.1101, "step": 131840 }, { "epoch": 11.42, "learning_rate": 4.651830114490574e-05, "loss": 1.1315, "step": 131850 }, { "epoch": 11.42, "learning_rate": 4.651613276280791e-05, "loss": 1.0416, "step": 131860 }, { "epoch": 11.42, "learning_rate": 4.651396438071007e-05, "loss": 0.9471, "step": 131870 }, { "epoch": 11.42, "learning_rate": 4.6511795998612235e-05, "loss": 1.1738, "step": 131880 }, { "epoch": 11.42, "learning_rate": 4.6509627616514394e-05, "loss": 1.0385, "step": 131890 }, { "epoch": 11.42, "learning_rate": 4.6507459234416554e-05, "loss": 1.0738, "step": 131900 }, { "epoch": 11.42, "learning_rate": 4.6505290852318713e-05, "loss": 1.0187, "step": 131910 }, { "epoch": 11.43, "learning_rate": 4.650312247022088e-05, "loss": 1.1944, "step": 131920 }, { "epoch": 11.43, "learning_rate": 4.6500954088123046e-05, "loss": 1.2044, "step": 131930 }, { "epoch": 11.43, "learning_rate": 4.6498785706025206e-05, "loss": 1.1257, "step": 131940 }, { "epoch": 11.43, "learning_rate": 4.649661732392737e-05, "loss": 1.0688, "step": 131950 }, { "epoch": 11.43, "learning_rate": 4.649444894182953e-05, "loss": 1.0536, "step": 131960 }, { "epoch": 11.43, "learning_rate": 4.64922805597317e-05, "loss": 1.1883, "step": 131970 }, { "epoch": 11.43, "learning_rate": 4.649011217763385e-05, "loss": 1.1287, "step": 131980 }, { "epoch": 11.43, "learning_rate": 4.648794379553602e-05, "loss": 1.1177, "step": 131990 }, { "epoch": 11.43, "learning_rate": 4.6485775413438184e-05, "loss": 0.9436, "step": 132000 }, { "epoch": 11.43, "learning_rate": 4.6483607031340343e-05, "loss": 1.1382, "step": 132010 }, { "epoch": 11.43, "learning_rate": 4.648143864924251e-05, "loss": 1.1505, "step": 132020 }, { "epoch": 11.43, "learning_rate": 4.647927026714467e-05, "loss": 1.0032, "step": 132030 }, { "epoch": 11.44, "learning_rate": 4.6477101885046836e-05, "loss": 0.987, "step": 132040 }, { "epoch": 11.44, "learning_rate": 4.6474933502949e-05, "loss": 1.0815, "step": 132050 }, { "epoch": 11.44, "learning_rate": 4.6472765120851155e-05, "loss": 1.1114, "step": 132060 }, { "epoch": 11.44, "learning_rate": 4.647059673875332e-05, "loss": 1.2116, "step": 132070 }, { "epoch": 11.44, "learning_rate": 4.646842835665548e-05, "loss": 0.9721, "step": 132080 }, { "epoch": 11.44, "learning_rate": 4.646625997455765e-05, "loss": 1.181, "step": 132090 }, { "epoch": 11.44, "learning_rate": 4.646409159245981e-05, "loss": 1.1267, "step": 132100 }, { "epoch": 11.44, "learning_rate": 4.6461923210361974e-05, "loss": 1.1548, "step": 132110 }, { "epoch": 11.44, "learning_rate": 4.645975482826414e-05, "loss": 0.9389, "step": 132120 }, { "epoch": 11.44, "learning_rate": 4.645758644616629e-05, "loss": 1.1201, "step": 132130 }, { "epoch": 11.44, "learning_rate": 4.645541806406846e-05, "loss": 0.9393, "step": 132140 }, { "epoch": 11.45, "learning_rate": 4.645324968197062e-05, "loss": 1.0953, "step": 132150 }, { "epoch": 11.45, "learning_rate": 4.6451081299872785e-05, "loss": 1.1559, "step": 132160 }, { "epoch": 11.45, "learning_rate": 4.6448912917774945e-05, "loss": 0.9382, "step": 132170 }, { "epoch": 11.45, "learning_rate": 4.644674453567711e-05, "loss": 1.2763, "step": 132180 }, { "epoch": 11.45, "learning_rate": 4.644457615357928e-05, "loss": 1.0468, "step": 132190 }, { "epoch": 11.45, "learning_rate": 4.644240777148144e-05, "loss": 1.0107, "step": 132200 }, { "epoch": 11.45, "learning_rate": 4.644023938938359e-05, "loss": 1.1706, "step": 132210 }, { "epoch": 11.45, "learning_rate": 4.6438071007285756e-05, "loss": 1.0612, "step": 132220 }, { "epoch": 11.45, "learning_rate": 4.643590262518792e-05, "loss": 1.1163, "step": 132230 }, { "epoch": 11.45, "learning_rate": 4.643373424309008e-05, "loss": 1.23, "step": 132240 }, { "epoch": 11.45, "learning_rate": 4.643156586099225e-05, "loss": 1.1499, "step": 132250 }, { "epoch": 11.45, "learning_rate": 4.642939747889441e-05, "loss": 1.1719, "step": 132260 }, { "epoch": 11.46, "learning_rate": 4.6427229096796575e-05, "loss": 1.2825, "step": 132270 }, { "epoch": 11.46, "learning_rate": 4.642506071469874e-05, "loss": 1.2614, "step": 132280 }, { "epoch": 11.46, "learning_rate": 4.6422892332600894e-05, "loss": 1.0805, "step": 132290 }, { "epoch": 11.46, "learning_rate": 4.642072395050306e-05, "loss": 1.0964, "step": 132300 }, { "epoch": 11.46, "learning_rate": 4.641855556840522e-05, "loss": 0.9515, "step": 132310 }, { "epoch": 11.46, "learning_rate": 4.6416387186307386e-05, "loss": 1.0229, "step": 132320 }, { "epoch": 11.46, "learning_rate": 4.6414218804209546e-05, "loss": 1.2261, "step": 132330 }, { "epoch": 11.46, "learning_rate": 4.641205042211171e-05, "loss": 1.1051, "step": 132340 }, { "epoch": 11.46, "learning_rate": 4.640988204001388e-05, "loss": 0.9047, "step": 132350 }, { "epoch": 11.46, "learning_rate": 4.640771365791604e-05, "loss": 1.0846, "step": 132360 }, { "epoch": 11.46, "learning_rate": 4.64055452758182e-05, "loss": 1.1063, "step": 132370 }, { "epoch": 11.46, "learning_rate": 4.640337689372036e-05, "loss": 1.1061, "step": 132380 }, { "epoch": 11.47, "learning_rate": 4.6401208511622524e-05, "loss": 1.0221, "step": 132390 }, { "epoch": 11.47, "learning_rate": 4.6399040129524684e-05, "loss": 1.2427, "step": 132400 }, { "epoch": 11.47, "learning_rate": 4.639687174742685e-05, "loss": 0.8669, "step": 132410 }, { "epoch": 11.47, "learning_rate": 4.6394703365329016e-05, "loss": 1.1856, "step": 132420 }, { "epoch": 11.47, "learning_rate": 4.6392534983231176e-05, "loss": 1.1396, "step": 132430 }, { "epoch": 11.47, "learning_rate": 4.6390366601133336e-05, "loss": 1.3084, "step": 132440 }, { "epoch": 11.47, "learning_rate": 4.6388198219035495e-05, "loss": 1.007, "step": 132450 }, { "epoch": 11.47, "learning_rate": 4.638602983693766e-05, "loss": 1.0661, "step": 132460 }, { "epoch": 11.47, "learning_rate": 4.638386145483982e-05, "loss": 0.9638, "step": 132470 }, { "epoch": 11.47, "learning_rate": 4.638169307274199e-05, "loss": 1.0335, "step": 132480 }, { "epoch": 11.47, "learning_rate": 4.6379524690644154e-05, "loss": 1.2367, "step": 132490 }, { "epoch": 11.48, "learning_rate": 4.6377356308546314e-05, "loss": 1.0589, "step": 132500 }, { "epoch": 11.48, "learning_rate": 4.637518792644848e-05, "loss": 1.1294, "step": 132510 }, { "epoch": 11.48, "learning_rate": 4.637301954435063e-05, "loss": 1.0741, "step": 132520 }, { "epoch": 11.48, "learning_rate": 4.63708511622528e-05, "loss": 1.0976, "step": 132530 }, { "epoch": 11.48, "learning_rate": 4.636868278015496e-05, "loss": 1.0141, "step": 132540 }, { "epoch": 11.48, "learning_rate": 4.6366514398057125e-05, "loss": 1.0203, "step": 132550 }, { "epoch": 11.48, "learning_rate": 4.636434601595929e-05, "loss": 0.8588, "step": 132560 }, { "epoch": 11.48, "learning_rate": 4.636217763386145e-05, "loss": 1.2193, "step": 132570 }, { "epoch": 11.48, "learning_rate": 4.636000925176362e-05, "loss": 1.1639, "step": 132580 }, { "epoch": 11.48, "learning_rate": 4.635784086966578e-05, "loss": 1.1279, "step": 132590 }, { "epoch": 11.48, "learning_rate": 4.635567248756794e-05, "loss": 1.1428, "step": 132600 }, { "epoch": 11.48, "learning_rate": 4.6353504105470097e-05, "loss": 1.0943, "step": 132610 }, { "epoch": 11.49, "learning_rate": 4.635133572337226e-05, "loss": 1.1482, "step": 132620 }, { "epoch": 11.49, "learning_rate": 4.634916734127443e-05, "loss": 1.232, "step": 132630 }, { "epoch": 11.49, "learning_rate": 4.634699895917659e-05, "loss": 1.2445, "step": 132640 }, { "epoch": 11.49, "learning_rate": 4.6344830577078755e-05, "loss": 1.1394, "step": 132650 }, { "epoch": 11.49, "learning_rate": 4.6342662194980915e-05, "loss": 1.1853, "step": 132660 }, { "epoch": 11.49, "learning_rate": 4.6340493812883075e-05, "loss": 1.0134, "step": 132670 }, { "epoch": 11.49, "learning_rate": 4.6338325430785234e-05, "loss": 1.1407, "step": 132680 }, { "epoch": 11.49, "learning_rate": 4.63361570486874e-05, "loss": 1.1992, "step": 132690 }, { "epoch": 11.49, "learning_rate": 4.633398866658957e-05, "loss": 1.0016, "step": 132700 }, { "epoch": 11.49, "learning_rate": 4.633182028449173e-05, "loss": 1.1725, "step": 132710 }, { "epoch": 11.49, "learning_rate": 4.632965190239389e-05, "loss": 1.0961, "step": 132720 }, { "epoch": 11.5, "learning_rate": 4.632748352029605e-05, "loss": 1.1512, "step": 132730 }, { "epoch": 11.5, "learning_rate": 4.632531513819822e-05, "loss": 1.0424, "step": 132740 }, { "epoch": 11.5, "learning_rate": 4.632314675610037e-05, "loss": 1.2077, "step": 132750 }, { "epoch": 11.5, "learning_rate": 4.632097837400254e-05, "loss": 1.0718, "step": 132760 }, { "epoch": 11.5, "learning_rate": 4.63188099919047e-05, "loss": 1.2855, "step": 132770 }, { "epoch": 11.5, "learning_rate": 4.6316641609806864e-05, "loss": 1.0417, "step": 132780 }, { "epoch": 11.5, "learning_rate": 4.631447322770903e-05, "loss": 1.0437, "step": 132790 }, { "epoch": 11.5, "learning_rate": 4.631230484561119e-05, "loss": 1.2797, "step": 132800 }, { "epoch": 11.5, "learning_rate": 4.631013646351336e-05, "loss": 1.1977, "step": 132810 }, { "epoch": 11.5, "learning_rate": 4.6307968081415516e-05, "loss": 1.1075, "step": 132820 }, { "epoch": 11.5, "learning_rate": 4.6305799699317676e-05, "loss": 1.2737, "step": 132830 }, { "epoch": 11.5, "learning_rate": 4.6303631317219836e-05, "loss": 1.1678, "step": 132840 }, { "epoch": 11.51, "learning_rate": 4.6301462935122e-05, "loss": 1.2679, "step": 132850 }, { "epoch": 11.51, "learning_rate": 4.629929455302417e-05, "loss": 1.2043, "step": 132860 }, { "epoch": 11.51, "learning_rate": 4.629712617092633e-05, "loss": 1.2253, "step": 132870 }, { "epoch": 11.51, "learning_rate": 4.6294957788828494e-05, "loss": 1.2122, "step": 132880 }, { "epoch": 11.51, "learning_rate": 4.6292789406730654e-05, "loss": 1.1248, "step": 132890 }, { "epoch": 11.51, "learning_rate": 4.6290621024632814e-05, "loss": 1.0416, "step": 132900 }, { "epoch": 11.51, "learning_rate": 4.628845264253497e-05, "loss": 1.0841, "step": 132910 }, { "epoch": 11.51, "learning_rate": 4.628628426043714e-05, "loss": 1.1456, "step": 132920 }, { "epoch": 11.51, "learning_rate": 4.6284115878339306e-05, "loss": 1.2005, "step": 132930 }, { "epoch": 11.51, "learning_rate": 4.6281947496241466e-05, "loss": 1.0325, "step": 132940 }, { "epoch": 11.51, "learning_rate": 4.627977911414363e-05, "loss": 1.1469, "step": 132950 }, { "epoch": 11.52, "learning_rate": 4.627761073204579e-05, "loss": 1.119, "step": 132960 }, { "epoch": 11.52, "learning_rate": 4.627544234994796e-05, "loss": 1.2106, "step": 132970 }, { "epoch": 11.52, "learning_rate": 4.627327396785011e-05, "loss": 1.0879, "step": 132980 }, { "epoch": 11.52, "learning_rate": 4.627110558575228e-05, "loss": 1.1622, "step": 132990 }, { "epoch": 11.52, "learning_rate": 4.6268937203654444e-05, "loss": 1.1776, "step": 133000 }, { "epoch": 11.52, "learning_rate": 4.62667688215566e-05, "loss": 1.0346, "step": 133010 }, { "epoch": 11.52, "learning_rate": 4.626460043945877e-05, "loss": 1.4322, "step": 133020 }, { "epoch": 11.52, "learning_rate": 4.626243205736093e-05, "loss": 1.177, "step": 133030 }, { "epoch": 11.52, "learning_rate": 4.6260263675263096e-05, "loss": 1.0743, "step": 133040 }, { "epoch": 11.52, "learning_rate": 4.625809529316526e-05, "loss": 1.0852, "step": 133050 }, { "epoch": 11.52, "learning_rate": 4.6255926911067415e-05, "loss": 1.1712, "step": 133060 }, { "epoch": 11.52, "learning_rate": 4.625375852896958e-05, "loss": 0.9573, "step": 133070 }, { "epoch": 11.53, "learning_rate": 4.625159014687174e-05, "loss": 1.0196, "step": 133080 }, { "epoch": 11.53, "learning_rate": 4.624942176477391e-05, "loss": 1.3316, "step": 133090 }, { "epoch": 11.53, "learning_rate": 4.624725338267607e-05, "loss": 1.2063, "step": 133100 }, { "epoch": 11.53, "learning_rate": 4.624508500057823e-05, "loss": 1.1312, "step": 133110 }, { "epoch": 11.53, "learning_rate": 4.62429166184804e-05, "loss": 1.2286, "step": 133120 }, { "epoch": 11.53, "learning_rate": 4.624074823638256e-05, "loss": 1.1974, "step": 133130 }, { "epoch": 11.53, "learning_rate": 4.623857985428472e-05, "loss": 1.081, "step": 133140 }, { "epoch": 11.53, "learning_rate": 4.623641147218688e-05, "loss": 1.2244, "step": 133150 }, { "epoch": 11.53, "learning_rate": 4.6234243090089045e-05, "loss": 1.0471, "step": 133160 }, { "epoch": 11.53, "learning_rate": 4.6232074707991204e-05, "loss": 1.3419, "step": 133170 }, { "epoch": 11.53, "learning_rate": 4.622990632589337e-05, "loss": 1.1568, "step": 133180 }, { "epoch": 11.54, "learning_rate": 4.622773794379554e-05, "loss": 1.2953, "step": 133190 }, { "epoch": 11.54, "learning_rate": 4.62255695616977e-05, "loss": 0.9643, "step": 133200 }, { "epoch": 11.54, "learning_rate": 4.6223401179599856e-05, "loss": 1.1297, "step": 133210 }, { "epoch": 11.54, "learning_rate": 4.6221232797502016e-05, "loss": 1.1991, "step": 133220 }, { "epoch": 11.54, "learning_rate": 4.621906441540418e-05, "loss": 1.1241, "step": 133230 }, { "epoch": 11.54, "learning_rate": 4.621689603330634e-05, "loss": 1.1869, "step": 133240 }, { "epoch": 11.54, "learning_rate": 4.621472765120851e-05, "loss": 1.095, "step": 133250 }, { "epoch": 11.54, "learning_rate": 4.6212559269110675e-05, "loss": 1.0595, "step": 133260 }, { "epoch": 11.54, "learning_rate": 4.6210390887012834e-05, "loss": 0.8526, "step": 133270 }, { "epoch": 11.54, "learning_rate": 4.6208222504915e-05, "loss": 1.1991, "step": 133280 }, { "epoch": 11.54, "learning_rate": 4.6206054122817154e-05, "loss": 1.1624, "step": 133290 }, { "epoch": 11.54, "learning_rate": 4.620388574071932e-05, "loss": 1.1251, "step": 133300 }, { "epoch": 11.55, "learning_rate": 4.620171735862148e-05, "loss": 0.9321, "step": 133310 }, { "epoch": 11.55, "learning_rate": 4.6199548976523646e-05, "loss": 1.115, "step": 133320 }, { "epoch": 11.55, "learning_rate": 4.6197380594425806e-05, "loss": 1.0619, "step": 133330 }, { "epoch": 11.55, "learning_rate": 4.619521221232797e-05, "loss": 0.9432, "step": 133340 }, { "epoch": 11.55, "learning_rate": 4.619304383023014e-05, "loss": 1.1273, "step": 133350 }, { "epoch": 11.55, "learning_rate": 4.61908754481323e-05, "loss": 1.141, "step": 133360 }, { "epoch": 11.55, "learning_rate": 4.618870706603446e-05, "loss": 1.0759, "step": 133370 }, { "epoch": 11.55, "learning_rate": 4.618653868393662e-05, "loss": 1.049, "step": 133380 }, { "epoch": 11.55, "learning_rate": 4.6184370301838784e-05, "loss": 1.036, "step": 133390 }, { "epoch": 11.55, "learning_rate": 4.618220191974094e-05, "loss": 1.0733, "step": 133400 }, { "epoch": 11.55, "learning_rate": 4.618003353764311e-05, "loss": 1.0836, "step": 133410 }, { "epoch": 11.56, "learning_rate": 4.6177865155545276e-05, "loss": 0.9742, "step": 133420 }, { "epoch": 11.56, "learning_rate": 4.6175696773447436e-05, "loss": 1.0802, "step": 133430 }, { "epoch": 11.56, "learning_rate": 4.6173528391349595e-05, "loss": 1.2082, "step": 133440 }, { "epoch": 11.56, "learning_rate": 4.6171360009251755e-05, "loss": 1.1041, "step": 133450 }, { "epoch": 11.56, "learning_rate": 4.616919162715392e-05, "loss": 1.0382, "step": 133460 }, { "epoch": 11.56, "learning_rate": 4.616702324505608e-05, "loss": 1.0344, "step": 133470 }, { "epoch": 11.56, "learning_rate": 4.616485486295825e-05, "loss": 1.2426, "step": 133480 }, { "epoch": 11.56, "learning_rate": 4.6162686480860414e-05, "loss": 1.1222, "step": 133490 }, { "epoch": 11.56, "learning_rate": 4.6160518098762573e-05, "loss": 0.9105, "step": 133500 }, { "epoch": 11.56, "learning_rate": 4.615834971666474e-05, "loss": 0.9725, "step": 133510 }, { "epoch": 11.56, "learning_rate": 4.615618133456689e-05, "loss": 0.9459, "step": 133520 }, { "epoch": 11.56, "learning_rate": 4.615401295246906e-05, "loss": 1.33, "step": 133530 }, { "epoch": 11.57, "learning_rate": 4.615184457037122e-05, "loss": 1.0481, "step": 133540 }, { "epoch": 11.57, "learning_rate": 4.6149676188273385e-05, "loss": 0.9828, "step": 133550 }, { "epoch": 11.57, "learning_rate": 4.614750780617555e-05, "loss": 0.9921, "step": 133560 }, { "epoch": 11.57, "learning_rate": 4.614533942407771e-05, "loss": 1.1515, "step": 133570 }, { "epoch": 11.57, "learning_rate": 4.614317104197988e-05, "loss": 1.1069, "step": 133580 }, { "epoch": 11.57, "learning_rate": 4.614100265988204e-05, "loss": 0.9024, "step": 133590 }, { "epoch": 11.57, "learning_rate": 4.61388342777842e-05, "loss": 1.117, "step": 133600 }, { "epoch": 11.57, "learning_rate": 4.6136665895686356e-05, "loss": 1.1222, "step": 133610 }, { "epoch": 11.57, "learning_rate": 4.613449751358852e-05, "loss": 1.2214, "step": 133620 }, { "epoch": 11.57, "learning_rate": 4.613232913149069e-05, "loss": 0.9357, "step": 133630 }, { "epoch": 11.57, "learning_rate": 4.613016074939285e-05, "loss": 1.2311, "step": 133640 }, { "epoch": 11.57, "learning_rate": 4.6127992367295015e-05, "loss": 1.1142, "step": 133650 }, { "epoch": 11.58, "learning_rate": 4.6125823985197175e-05, "loss": 1.348, "step": 133660 }, { "epoch": 11.58, "learning_rate": 4.6123655603099334e-05, "loss": 1.2561, "step": 133670 }, { "epoch": 11.58, "learning_rate": 4.6121487221001494e-05, "loss": 1.1884, "step": 133680 }, { "epoch": 11.58, "learning_rate": 4.611931883890366e-05, "loss": 1.1831, "step": 133690 }, { "epoch": 11.58, "learning_rate": 4.611715045680583e-05, "loss": 0.9994, "step": 133700 }, { "epoch": 11.58, "learning_rate": 4.6114982074707986e-05, "loss": 1.0331, "step": 133710 }, { "epoch": 11.58, "learning_rate": 4.611281369261015e-05, "loss": 1.0037, "step": 133720 }, { "epoch": 11.58, "learning_rate": 4.611064531051231e-05, "loss": 1.1098, "step": 133730 }, { "epoch": 11.58, "learning_rate": 4.610847692841448e-05, "loss": 1.0712, "step": 133740 }, { "epoch": 11.58, "learning_rate": 4.610630854631663e-05, "loss": 1.0427, "step": 133750 }, { "epoch": 11.58, "learning_rate": 4.61041401642188e-05, "loss": 1.1614, "step": 133760 }, { "epoch": 11.59, "learning_rate": 4.6101971782120964e-05, "loss": 1.2023, "step": 133770 }, { "epoch": 11.59, "learning_rate": 4.6099803400023124e-05, "loss": 1.0138, "step": 133780 }, { "epoch": 11.59, "learning_rate": 4.609763501792529e-05, "loss": 1.1154, "step": 133790 }, { "epoch": 11.59, "learning_rate": 4.609546663582745e-05, "loss": 1.1568, "step": 133800 }, { "epoch": 11.59, "learning_rate": 4.6093298253729616e-05, "loss": 1.2089, "step": 133810 }, { "epoch": 11.59, "learning_rate": 4.609112987163178e-05, "loss": 0.9896, "step": 133820 }, { "epoch": 11.59, "learning_rate": 4.6088961489533936e-05, "loss": 1.0519, "step": 133830 }, { "epoch": 11.59, "learning_rate": 4.6086793107436095e-05, "loss": 1.2452, "step": 133840 }, { "epoch": 11.59, "learning_rate": 4.608462472533826e-05, "loss": 0.9985, "step": 133850 }, { "epoch": 11.59, "learning_rate": 4.608245634324043e-05, "loss": 1.0408, "step": 133860 }, { "epoch": 11.59, "learning_rate": 4.608028796114259e-05, "loss": 1.1156, "step": 133870 }, { "epoch": 11.59, "learning_rate": 4.6078119579044754e-05, "loss": 1.0216, "step": 133880 }, { "epoch": 11.6, "learning_rate": 4.6075951196946914e-05, "loss": 1.0433, "step": 133890 }, { "epoch": 11.6, "learning_rate": 4.607378281484908e-05, "loss": 1.0601, "step": 133900 }, { "epoch": 11.6, "learning_rate": 4.607161443275123e-05, "loss": 1.0556, "step": 133910 }, { "epoch": 11.6, "learning_rate": 4.60694460506534e-05, "loss": 1.2495, "step": 133920 }, { "epoch": 11.6, "learning_rate": 4.6067277668555566e-05, "loss": 0.9489, "step": 133930 }, { "epoch": 11.6, "learning_rate": 4.6065109286457725e-05, "loss": 1.2023, "step": 133940 }, { "epoch": 11.6, "learning_rate": 4.606294090435989e-05, "loss": 1.2689, "step": 133950 }, { "epoch": 11.6, "learning_rate": 4.606077252226205e-05, "loss": 1.2016, "step": 133960 }, { "epoch": 11.6, "learning_rate": 4.605860414016422e-05, "loss": 1.3468, "step": 133970 }, { "epoch": 11.6, "learning_rate": 4.605643575806637e-05, "loss": 1.4474, "step": 133980 }, { "epoch": 11.6, "learning_rate": 4.605426737596854e-05, "loss": 1.0949, "step": 133990 }, { "epoch": 11.61, "learning_rate": 4.60520989938707e-05, "loss": 1.2088, "step": 134000 }, { "epoch": 11.61, "learning_rate": 4.604993061177286e-05, "loss": 1.0858, "step": 134010 }, { "epoch": 11.61, "learning_rate": 4.604776222967503e-05, "loss": 1.2384, "step": 134020 }, { "epoch": 11.61, "learning_rate": 4.604559384757719e-05, "loss": 1.216, "step": 134030 }, { "epoch": 11.61, "learning_rate": 4.6043425465479355e-05, "loss": 1.1458, "step": 134040 }, { "epoch": 11.61, "learning_rate": 4.604125708338152e-05, "loss": 1.1073, "step": 134050 }, { "epoch": 11.61, "learning_rate": 4.6039088701283675e-05, "loss": 1.2444, "step": 134060 }, { "epoch": 11.61, "learning_rate": 4.603692031918584e-05, "loss": 1.0333, "step": 134070 }, { "epoch": 11.61, "learning_rate": 4.6034751937088e-05, "loss": 1.001, "step": 134080 }, { "epoch": 11.61, "learning_rate": 4.603258355499017e-05, "loss": 1.0985, "step": 134090 }, { "epoch": 11.61, "learning_rate": 4.6030415172892327e-05, "loss": 1.0172, "step": 134100 }, { "epoch": 11.61, "learning_rate": 4.602824679079449e-05, "loss": 0.9891, "step": 134110 }, { "epoch": 11.62, "learning_rate": 4.602607840869666e-05, "loss": 1.1933, "step": 134120 }, { "epoch": 11.62, "learning_rate": 4.602391002659882e-05, "loss": 1.2397, "step": 134130 }, { "epoch": 11.62, "learning_rate": 4.602174164450098e-05, "loss": 1.1661, "step": 134140 }, { "epoch": 11.62, "learning_rate": 4.601957326240314e-05, "loss": 1.0887, "step": 134150 }, { "epoch": 11.62, "learning_rate": 4.6017404880305305e-05, "loss": 1.1932, "step": 134160 }, { "epoch": 11.62, "learning_rate": 4.6015236498207464e-05, "loss": 1.2105, "step": 134170 }, { "epoch": 11.62, "learning_rate": 4.601306811610963e-05, "loss": 1.0595, "step": 134180 }, { "epoch": 11.62, "learning_rate": 4.60108997340118e-05, "loss": 1.1542, "step": 134190 }, { "epoch": 11.62, "learning_rate": 4.6008731351913957e-05, "loss": 1.1965, "step": 134200 }, { "epoch": 11.62, "learning_rate": 4.6006562969816116e-05, "loss": 0.9822, "step": 134210 }, { "epoch": 11.62, "learning_rate": 4.6004394587718276e-05, "loss": 1.0225, "step": 134220 }, { "epoch": 11.63, "learning_rate": 4.600222620562044e-05, "loss": 0.9638, "step": 134230 }, { "epoch": 11.63, "learning_rate": 4.60000578235226e-05, "loss": 1.1466, "step": 134240 }, { "epoch": 11.63, "learning_rate": 4.599788944142477e-05, "loss": 1.0775, "step": 134250 }, { "epoch": 11.63, "learning_rate": 4.5995721059326935e-05, "loss": 1.2714, "step": 134260 }, { "epoch": 11.63, "learning_rate": 4.5993552677229094e-05, "loss": 1.0278, "step": 134270 }, { "epoch": 11.63, "learning_rate": 4.599138429513126e-05, "loss": 1.097, "step": 134280 }, { "epoch": 11.63, "learning_rate": 4.5989215913033413e-05, "loss": 1.1501, "step": 134290 }, { "epoch": 11.63, "learning_rate": 4.598704753093558e-05, "loss": 1.1687, "step": 134300 }, { "epoch": 11.63, "learning_rate": 4.598487914883774e-05, "loss": 1.0891, "step": 134310 }, { "epoch": 11.63, "learning_rate": 4.5982710766739906e-05, "loss": 1.0699, "step": 134320 }, { "epoch": 11.63, "learning_rate": 4.598054238464207e-05, "loss": 1.0713, "step": 134330 }, { "epoch": 11.63, "learning_rate": 4.597837400254423e-05, "loss": 1.2269, "step": 134340 }, { "epoch": 11.64, "learning_rate": 4.59762056204464e-05, "loss": 1.2429, "step": 134350 }, { "epoch": 11.64, "learning_rate": 4.597403723834856e-05, "loss": 1.188, "step": 134360 }, { "epoch": 11.64, "learning_rate": 4.597186885625072e-05, "loss": 1.2174, "step": 134370 }, { "epoch": 11.64, "learning_rate": 4.596970047415288e-05, "loss": 1.0493, "step": 134380 }, { "epoch": 11.64, "learning_rate": 4.5967532092055043e-05, "loss": 1.1, "step": 134390 }, { "epoch": 11.64, "learning_rate": 4.59653637099572e-05, "loss": 1.1322, "step": 134400 }, { "epoch": 11.64, "learning_rate": 4.596319532785937e-05, "loss": 1.1015, "step": 134410 }, { "epoch": 11.64, "learning_rate": 4.5961026945761536e-05, "loss": 1.2116, "step": 134420 }, { "epoch": 11.64, "learning_rate": 4.5958858563663695e-05, "loss": 1.101, "step": 134430 }, { "epoch": 11.64, "learning_rate": 4.5956690181565855e-05, "loss": 1.0786, "step": 134440 }, { "epoch": 11.64, "learning_rate": 4.5954521799468015e-05, "loss": 1.3201, "step": 134450 }, { "epoch": 11.65, "learning_rate": 4.595235341737018e-05, "loss": 1.0556, "step": 134460 }, { "epoch": 11.65, "learning_rate": 4.595018503527234e-05, "loss": 1.1331, "step": 134470 }, { "epoch": 11.65, "learning_rate": 4.594801665317451e-05, "loss": 0.9371, "step": 134480 }, { "epoch": 11.65, "learning_rate": 4.5945848271076673e-05, "loss": 1.1219, "step": 134490 }, { "epoch": 11.65, "learning_rate": 4.594367988897883e-05, "loss": 1.02, "step": 134500 }, { "epoch": 11.65, "learning_rate": 4.5941511506881e-05, "loss": 1.0922, "step": 134510 }, { "epoch": 11.65, "learning_rate": 4.593934312478315e-05, "loss": 0.9813, "step": 134520 }, { "epoch": 11.65, "learning_rate": 4.593717474268532e-05, "loss": 1.1504, "step": 134530 }, { "epoch": 11.65, "learning_rate": 4.593500636058748e-05, "loss": 1.0451, "step": 134540 }, { "epoch": 11.65, "learning_rate": 4.5932837978489645e-05, "loss": 1.1541, "step": 134550 }, { "epoch": 11.65, "learning_rate": 4.593066959639181e-05, "loss": 0.8906, "step": 134560 }, { "epoch": 11.65, "learning_rate": 4.592850121429397e-05, "loss": 1.0687, "step": 134570 }, { "epoch": 11.66, "learning_rate": 4.592633283219614e-05, "loss": 0.9553, "step": 134580 }, { "epoch": 11.66, "learning_rate": 4.59241644500983e-05, "loss": 1.21, "step": 134590 }, { "epoch": 11.66, "learning_rate": 4.5921996068000456e-05, "loss": 1.1964, "step": 134600 }, { "epoch": 11.66, "learning_rate": 4.5919827685902616e-05, "loss": 1.1228, "step": 134610 }, { "epoch": 11.66, "learning_rate": 4.591765930380478e-05, "loss": 1.1094, "step": 134620 }, { "epoch": 11.66, "learning_rate": 4.591549092170695e-05, "loss": 1.2403, "step": 134630 }, { "epoch": 11.66, "learning_rate": 4.591332253960911e-05, "loss": 1.0701, "step": 134640 }, { "epoch": 11.66, "learning_rate": 4.5911154157511275e-05, "loss": 1.0154, "step": 134650 }, { "epoch": 11.66, "learning_rate": 4.5908985775413434e-05, "loss": 1.2081, "step": 134660 }, { "epoch": 11.66, "learning_rate": 4.59068173933156e-05, "loss": 1.1071, "step": 134670 }, { "epoch": 11.66, "learning_rate": 4.5904649011217754e-05, "loss": 1.1719, "step": 134680 }, { "epoch": 11.67, "learning_rate": 4.590248062911992e-05, "loss": 1.1111, "step": 134690 }, { "epoch": 11.67, "learning_rate": 4.5900312247022086e-05, "loss": 0.9951, "step": 134700 }, { "epoch": 11.67, "learning_rate": 4.5898143864924246e-05, "loss": 1.1786, "step": 134710 }, { "epoch": 11.67, "learning_rate": 4.589597548282641e-05, "loss": 1.111, "step": 134720 }, { "epoch": 11.67, "learning_rate": 4.589380710072857e-05, "loss": 1.0376, "step": 134730 }, { "epoch": 11.67, "learning_rate": 4.589163871863074e-05, "loss": 1.4316, "step": 134740 }, { "epoch": 11.67, "learning_rate": 4.588947033653289e-05, "loss": 1.0628, "step": 134750 }, { "epoch": 11.67, "learning_rate": 4.588730195443506e-05, "loss": 1.1675, "step": 134760 }, { "epoch": 11.67, "learning_rate": 4.5885133572337224e-05, "loss": 1.1257, "step": 134770 }, { "epoch": 11.67, "learning_rate": 4.5882965190239384e-05, "loss": 1.0213, "step": 134780 }, { "epoch": 11.67, "learning_rate": 4.588079680814155e-05, "loss": 1.2235, "step": 134790 }, { "epoch": 11.67, "learning_rate": 4.587862842604371e-05, "loss": 1.3736, "step": 134800 }, { "epoch": 11.68, "learning_rate": 4.5876460043945876e-05, "loss": 0.9323, "step": 134810 }, { "epoch": 11.68, "learning_rate": 4.587429166184804e-05, "loss": 1.0428, "step": 134820 }, { "epoch": 11.68, "learning_rate": 4.5872123279750195e-05, "loss": 0.8993, "step": 134830 }, { "epoch": 11.68, "learning_rate": 4.586995489765236e-05, "loss": 1.1793, "step": 134840 }, { "epoch": 11.68, "learning_rate": 4.586778651555452e-05, "loss": 1.2352, "step": 134850 }, { "epoch": 11.68, "learning_rate": 4.586561813345669e-05, "loss": 1.0809, "step": 134860 }, { "epoch": 11.68, "learning_rate": 4.586344975135885e-05, "loss": 1.2305, "step": 134870 }, { "epoch": 11.68, "learning_rate": 4.5861281369261014e-05, "loss": 1.103, "step": 134880 }, { "epoch": 11.68, "learning_rate": 4.585911298716318e-05, "loss": 0.9985, "step": 134890 }, { "epoch": 11.68, "learning_rate": 4.585694460506534e-05, "loss": 1.0706, "step": 134900 }, { "epoch": 11.68, "learning_rate": 4.585477622296749e-05, "loss": 1.0953, "step": 134910 }, { "epoch": 11.68, "learning_rate": 4.585260784086966e-05, "loss": 1.3309, "step": 134920 }, { "epoch": 11.69, "learning_rate": 4.5850439458771825e-05, "loss": 1.0155, "step": 134930 }, { "epoch": 11.69, "learning_rate": 4.5848271076673985e-05, "loss": 1.0782, "step": 134940 }, { "epoch": 11.69, "learning_rate": 4.584610269457615e-05, "loss": 1.3606, "step": 134950 }, { "epoch": 11.69, "learning_rate": 4.584393431247831e-05, "loss": 1.351, "step": 134960 }, { "epoch": 11.69, "learning_rate": 4.584176593038048e-05, "loss": 1.1702, "step": 134970 }, { "epoch": 11.69, "learning_rate": 4.583959754828263e-05, "loss": 1.2454, "step": 134980 }, { "epoch": 11.69, "learning_rate": 4.5837429166184797e-05, "loss": 1.249, "step": 134990 }, { "epoch": 11.69, "learning_rate": 4.583526078408696e-05, "loss": 0.9369, "step": 135000 }, { "epoch": 11.69, "learning_rate": 4.583309240198912e-05, "loss": 1.2295, "step": 135010 }, { "epoch": 11.69, "learning_rate": 4.583092401989129e-05, "loss": 1.1578, "step": 135020 }, { "epoch": 11.69, "learning_rate": 4.582875563779345e-05, "loss": 0.9905, "step": 135030 }, { "epoch": 11.7, "learning_rate": 4.5826587255695615e-05, "loss": 0.926, "step": 135040 }, { "epoch": 11.7, "learning_rate": 4.582441887359778e-05, "loss": 1.167, "step": 135050 }, { "epoch": 11.7, "learning_rate": 4.5822250491499934e-05, "loss": 1.1174, "step": 135060 }, { "epoch": 11.7, "learning_rate": 4.58200821094021e-05, "loss": 1.0351, "step": 135070 }, { "epoch": 11.7, "learning_rate": 4.581791372730426e-05, "loss": 1.1072, "step": 135080 }, { "epoch": 11.7, "learning_rate": 4.5815745345206427e-05, "loss": 1.0064, "step": 135090 }, { "epoch": 11.7, "learning_rate": 4.5813576963108586e-05, "loss": 0.9902, "step": 135100 }, { "epoch": 11.7, "learning_rate": 4.581140858101075e-05, "loss": 1.0036, "step": 135110 }, { "epoch": 11.7, "learning_rate": 4.580924019891292e-05, "loss": 1.2877, "step": 135120 }, { "epoch": 11.7, "learning_rate": 4.580707181681508e-05, "loss": 1.1621, "step": 135130 }, { "epoch": 11.7, "learning_rate": 4.580490343471724e-05, "loss": 1.255, "step": 135140 }, { "epoch": 11.7, "learning_rate": 4.58027350526194e-05, "loss": 1.0823, "step": 135150 }, { "epoch": 11.71, "learning_rate": 4.5800566670521564e-05, "loss": 1.2977, "step": 135160 }, { "epoch": 11.71, "learning_rate": 4.5798398288423724e-05, "loss": 1.0518, "step": 135170 }, { "epoch": 11.71, "learning_rate": 4.579622990632589e-05, "loss": 1.1662, "step": 135180 }, { "epoch": 11.71, "learning_rate": 4.579406152422806e-05, "loss": 1.1921, "step": 135190 }, { "epoch": 11.71, "learning_rate": 4.5791893142130216e-05, "loss": 1.1298, "step": 135200 }, { "epoch": 11.71, "learning_rate": 4.5789724760032376e-05, "loss": 1.292, "step": 135210 }, { "epoch": 11.71, "learning_rate": 4.5787556377934535e-05, "loss": 0.8979, "step": 135220 }, { "epoch": 11.71, "learning_rate": 4.57853879958367e-05, "loss": 1.3711, "step": 135230 }, { "epoch": 11.71, "learning_rate": 4.578321961373886e-05, "loss": 1.2281, "step": 135240 }, { "epoch": 11.71, "learning_rate": 4.578105123164103e-05, "loss": 1.0005, "step": 135250 }, { "epoch": 11.71, "learning_rate": 4.5778882849543194e-05, "loss": 1.3209, "step": 135260 }, { "epoch": 11.72, "learning_rate": 4.5776714467445354e-05, "loss": 1.1479, "step": 135270 }, { "epoch": 11.72, "learning_rate": 4.577454608534752e-05, "loss": 1.3801, "step": 135280 }, { "epoch": 11.72, "learning_rate": 4.577237770324967e-05, "loss": 1.2675, "step": 135290 }, { "epoch": 11.72, "learning_rate": 4.577020932115184e-05, "loss": 1.2272, "step": 135300 }, { "epoch": 11.72, "learning_rate": 4.5768040939054e-05, "loss": 1.0881, "step": 135310 }, { "epoch": 11.72, "learning_rate": 4.5765872556956166e-05, "loss": 1.1512, "step": 135320 }, { "epoch": 11.72, "learning_rate": 4.576370417485833e-05, "loss": 1.0848, "step": 135330 }, { "epoch": 11.72, "learning_rate": 4.576153579276049e-05, "loss": 1.0724, "step": 135340 }, { "epoch": 11.72, "learning_rate": 4.575936741066266e-05, "loss": 1.1905, "step": 135350 }, { "epoch": 11.72, "learning_rate": 4.575719902856482e-05, "loss": 1.0029, "step": 135360 }, { "epoch": 11.72, "learning_rate": 4.575503064646698e-05, "loss": 1.2478, "step": 135370 }, { "epoch": 11.72, "learning_rate": 4.575286226436914e-05, "loss": 1.036, "step": 135380 }, { "epoch": 11.73, "learning_rate": 4.57506938822713e-05, "loss": 1.0561, "step": 135390 }, { "epoch": 11.73, "learning_rate": 4.574852550017347e-05, "loss": 1.3475, "step": 135400 }, { "epoch": 11.73, "learning_rate": 4.574635711807563e-05, "loss": 1.046, "step": 135410 }, { "epoch": 11.73, "learning_rate": 4.5744188735977796e-05, "loss": 1.0653, "step": 135420 }, { "epoch": 11.73, "learning_rate": 4.5742020353879955e-05, "loss": 1.1233, "step": 135430 }, { "epoch": 11.73, "learning_rate": 4.573985197178212e-05, "loss": 0.9348, "step": 135440 }, { "epoch": 11.73, "learning_rate": 4.5737683589684274e-05, "loss": 1.0723, "step": 135450 }, { "epoch": 11.73, "learning_rate": 4.573551520758644e-05, "loss": 1.0656, "step": 135460 }, { "epoch": 11.73, "learning_rate": 4.57333468254886e-05, "loss": 1.1778, "step": 135470 }, { "epoch": 11.73, "learning_rate": 4.573117844339077e-05, "loss": 0.9567, "step": 135480 }, { "epoch": 11.73, "learning_rate": 4.572901006129293e-05, "loss": 1.1814, "step": 135490 }, { "epoch": 11.74, "learning_rate": 4.572684167919509e-05, "loss": 1.1444, "step": 135500 }, { "epoch": 11.74, "learning_rate": 4.572467329709726e-05, "loss": 0.8276, "step": 135510 }, { "epoch": 11.74, "learning_rate": 4.572250491499941e-05, "loss": 1.1338, "step": 135520 }, { "epoch": 11.74, "learning_rate": 4.572033653290158e-05, "loss": 1.1624, "step": 135530 }, { "epoch": 11.74, "learning_rate": 4.571816815080374e-05, "loss": 1.1498, "step": 135540 }, { "epoch": 11.74, "learning_rate": 4.5715999768705904e-05, "loss": 1.0997, "step": 135550 }, { "epoch": 11.74, "learning_rate": 4.571383138660807e-05, "loss": 1.1568, "step": 135560 }, { "epoch": 11.74, "learning_rate": 4.571166300451023e-05, "loss": 1.2881, "step": 135570 }, { "epoch": 11.74, "learning_rate": 4.57094946224124e-05, "loss": 1.167, "step": 135580 }, { "epoch": 11.74, "learning_rate": 4.5707326240314556e-05, "loss": 1.1245, "step": 135590 }, { "epoch": 11.74, "learning_rate": 4.5705157858216716e-05, "loss": 1.1946, "step": 135600 }, { "epoch": 11.74, "learning_rate": 4.5702989476118876e-05, "loss": 1.2589, "step": 135610 }, { "epoch": 11.75, "learning_rate": 4.570082109402104e-05, "loss": 1.0054, "step": 135620 }, { "epoch": 11.75, "learning_rate": 4.569865271192321e-05, "loss": 1.3021, "step": 135630 }, { "epoch": 11.75, "learning_rate": 4.569648432982537e-05, "loss": 1.2356, "step": 135640 }, { "epoch": 11.75, "learning_rate": 4.5694315947727534e-05, "loss": 1.1021, "step": 135650 }, { "epoch": 11.75, "learning_rate": 4.5692147565629694e-05, "loss": 1.1292, "step": 135660 }, { "epoch": 11.75, "learning_rate": 4.568997918353186e-05, "loss": 1.2505, "step": 135670 }, { "epoch": 11.75, "learning_rate": 4.568781080143401e-05, "loss": 1.2012, "step": 135680 }, { "epoch": 11.75, "learning_rate": 4.568564241933618e-05, "loss": 1.2153, "step": 135690 }, { "epoch": 11.75, "learning_rate": 4.5683474037238346e-05, "loss": 1.1731, "step": 135700 }, { "epoch": 11.75, "learning_rate": 4.5681305655140506e-05, "loss": 1.1478, "step": 135710 }, { "epoch": 11.75, "learning_rate": 4.567913727304267e-05, "loss": 1.0853, "step": 135720 }, { "epoch": 11.76, "learning_rate": 4.567696889094483e-05, "loss": 0.9431, "step": 135730 }, { "epoch": 11.76, "learning_rate": 4.5674800508847e-05, "loss": 1.2593, "step": 135740 }, { "epoch": 11.76, "learning_rate": 4.567263212674915e-05, "loss": 1.2608, "step": 135750 }, { "epoch": 11.76, "learning_rate": 4.567046374465132e-05, "loss": 1.0991, "step": 135760 }, { "epoch": 11.76, "learning_rate": 4.5668295362553484e-05, "loss": 1.1253, "step": 135770 }, { "epoch": 11.76, "learning_rate": 4.566612698045564e-05, "loss": 1.0594, "step": 135780 }, { "epoch": 11.76, "learning_rate": 4.566395859835781e-05, "loss": 1.0044, "step": 135790 }, { "epoch": 11.76, "learning_rate": 4.566179021625997e-05, "loss": 1.1963, "step": 135800 }, { "epoch": 11.76, "learning_rate": 4.5659621834162136e-05, "loss": 1.0063, "step": 135810 }, { "epoch": 11.76, "learning_rate": 4.56574534520643e-05, "loss": 1.033, "step": 135820 }, { "epoch": 11.76, "learning_rate": 4.5655285069966455e-05, "loss": 1.1186, "step": 135830 }, { "epoch": 11.76, "learning_rate": 4.565311668786862e-05, "loss": 1.1095, "step": 135840 }, { "epoch": 11.77, "learning_rate": 4.565094830577078e-05, "loss": 0.9852, "step": 135850 }, { "epoch": 11.77, "learning_rate": 4.564877992367295e-05, "loss": 0.9879, "step": 135860 }, { "epoch": 11.77, "learning_rate": 4.564661154157511e-05, "loss": 1.1617, "step": 135870 }, { "epoch": 11.77, "learning_rate": 4.564444315947727e-05, "loss": 1.29, "step": 135880 }, { "epoch": 11.77, "learning_rate": 4.564227477737944e-05, "loss": 1.324, "step": 135890 }, { "epoch": 11.77, "learning_rate": 4.56401063952816e-05, "loss": 1.1324, "step": 135900 }, { "epoch": 11.77, "learning_rate": 4.563793801318376e-05, "loss": 1.1584, "step": 135910 }, { "epoch": 11.77, "learning_rate": 4.563576963108592e-05, "loss": 0.8538, "step": 135920 }, { "epoch": 11.77, "learning_rate": 4.5633601248988085e-05, "loss": 1.1593, "step": 135930 }, { "epoch": 11.77, "learning_rate": 4.5631432866890245e-05, "loss": 1.2057, "step": 135940 }, { "epoch": 11.77, "learning_rate": 4.562926448479241e-05, "loss": 1.0153, "step": 135950 }, { "epoch": 11.77, "learning_rate": 4.562709610269458e-05, "loss": 1.0567, "step": 135960 }, { "epoch": 11.78, "learning_rate": 4.562492772059674e-05, "loss": 1.3022, "step": 135970 }, { "epoch": 11.78, "learning_rate": 4.562275933849889e-05, "loss": 1.1631, "step": 135980 }, { "epoch": 11.78, "learning_rate": 4.5620590956401056e-05, "loss": 1.0366, "step": 135990 }, { "epoch": 11.78, "learning_rate": 4.561842257430322e-05, "loss": 1.0297, "step": 136000 }, { "epoch": 11.78, "learning_rate": 4.561625419220538e-05, "loss": 1.1693, "step": 136010 }, { "epoch": 11.78, "learning_rate": 4.561408581010755e-05, "loss": 0.948, "step": 136020 }, { "epoch": 11.78, "learning_rate": 4.561191742800971e-05, "loss": 1.3633, "step": 136030 }, { "epoch": 11.78, "learning_rate": 4.5609749045911875e-05, "loss": 1.0789, "step": 136040 }, { "epoch": 11.78, "learning_rate": 4.560758066381404e-05, "loss": 1.0517, "step": 136050 }, { "epoch": 11.78, "learning_rate": 4.5605412281716194e-05, "loss": 0.9551, "step": 136060 }, { "epoch": 11.78, "learning_rate": 4.560324389961836e-05, "loss": 1.057, "step": 136070 }, { "epoch": 11.79, "learning_rate": 4.560107551752052e-05, "loss": 1.2272, "step": 136080 }, { "epoch": 11.79, "learning_rate": 4.5598907135422686e-05, "loss": 1.1067, "step": 136090 }, { "epoch": 11.79, "learning_rate": 4.5596738753324846e-05, "loss": 1.0947, "step": 136100 }, { "epoch": 11.79, "learning_rate": 4.559457037122701e-05, "loss": 1.0824, "step": 136110 }, { "epoch": 11.79, "learning_rate": 4.559240198912918e-05, "loss": 1.1325, "step": 136120 }, { "epoch": 11.79, "learning_rate": 4.559023360703134e-05, "loss": 1.257, "step": 136130 }, { "epoch": 11.79, "learning_rate": 4.55880652249335e-05, "loss": 1.0862, "step": 136140 }, { "epoch": 11.79, "learning_rate": 4.558589684283566e-05, "loss": 1.1047, "step": 136150 }, { "epoch": 11.79, "learning_rate": 4.5583728460737824e-05, "loss": 1.1212, "step": 136160 }, { "epoch": 11.79, "learning_rate": 4.5581560078639984e-05, "loss": 1.0623, "step": 136170 }, { "epoch": 11.79, "learning_rate": 4.557939169654215e-05, "loss": 1.2265, "step": 136180 }, { "epoch": 11.79, "learning_rate": 4.5577223314444316e-05, "loss": 1.1676, "step": 136190 }, { "epoch": 11.8, "learning_rate": 4.5575054932346476e-05, "loss": 1.2377, "step": 136200 }, { "epoch": 11.8, "learning_rate": 4.557288655024864e-05, "loss": 1.0672, "step": 136210 }, { "epoch": 11.8, "learning_rate": 4.5570718168150795e-05, "loss": 1.2753, "step": 136220 }, { "epoch": 11.8, "learning_rate": 4.556854978605296e-05, "loss": 0.8925, "step": 136230 }, { "epoch": 11.8, "learning_rate": 4.556638140395512e-05, "loss": 1.0472, "step": 136240 }, { "epoch": 11.8, "learning_rate": 4.556421302185729e-05, "loss": 1.0552, "step": 136250 }, { "epoch": 11.8, "learning_rate": 4.5562044639759454e-05, "loss": 1.0975, "step": 136260 }, { "epoch": 11.8, "learning_rate": 4.5559876257661614e-05, "loss": 1.227, "step": 136270 }, { "epoch": 11.8, "learning_rate": 4.555770787556378e-05, "loss": 1.2277, "step": 136280 }, { "epoch": 11.8, "learning_rate": 4.555553949346593e-05, "loss": 1.1099, "step": 136290 }, { "epoch": 11.8, "learning_rate": 4.55533711113681e-05, "loss": 0.9336, "step": 136300 }, { "epoch": 11.81, "learning_rate": 4.555120272927026e-05, "loss": 1.0507, "step": 136310 }, { "epoch": 11.81, "learning_rate": 4.5549034347172425e-05, "loss": 1.1341, "step": 136320 }, { "epoch": 11.81, "learning_rate": 4.554686596507459e-05, "loss": 1.1776, "step": 136330 }, { "epoch": 11.81, "learning_rate": 4.554469758297675e-05, "loss": 1.1416, "step": 136340 }, { "epoch": 11.81, "learning_rate": 4.554252920087892e-05, "loss": 1.211, "step": 136350 }, { "epoch": 11.81, "learning_rate": 4.554036081878108e-05, "loss": 1.0494, "step": 136360 }, { "epoch": 11.81, "learning_rate": 4.553819243668324e-05, "loss": 1.1869, "step": 136370 }, { "epoch": 11.81, "learning_rate": 4.5536024054585396e-05, "loss": 1.292, "step": 136380 }, { "epoch": 11.81, "learning_rate": 4.553385567248756e-05, "loss": 1.1025, "step": 136390 }, { "epoch": 11.81, "learning_rate": 4.553168729038973e-05, "loss": 1.2729, "step": 136400 }, { "epoch": 11.81, "learning_rate": 4.552951890829189e-05, "loss": 1.224, "step": 136410 }, { "epoch": 11.81, "learning_rate": 4.5527350526194055e-05, "loss": 1.3898, "step": 136420 }, { "epoch": 11.82, "learning_rate": 4.5525182144096215e-05, "loss": 1.2277, "step": 136430 }, { "epoch": 11.82, "learning_rate": 4.552301376199838e-05, "loss": 1.0565, "step": 136440 }, { "epoch": 11.82, "learning_rate": 4.5520845379900534e-05, "loss": 0.9454, "step": 136450 }, { "epoch": 11.82, "learning_rate": 4.55186769978027e-05, "loss": 1.214, "step": 136460 }, { "epoch": 11.82, "learning_rate": 4.551650861570487e-05, "loss": 1.2252, "step": 136470 }, { "epoch": 11.82, "learning_rate": 4.5514340233607026e-05, "loss": 1.3144, "step": 136480 }, { "epoch": 11.82, "learning_rate": 4.551217185150919e-05, "loss": 1.0517, "step": 136490 }, { "epoch": 11.82, "learning_rate": 4.551000346941135e-05, "loss": 1.0656, "step": 136500 }, { "epoch": 11.82, "learning_rate": 4.550783508731352e-05, "loss": 1.1228, "step": 136510 }, { "epoch": 11.82, "learning_rate": 4.550566670521567e-05, "loss": 1.0916, "step": 136520 }, { "epoch": 11.82, "learning_rate": 4.550349832311784e-05, "loss": 1.1148, "step": 136530 }, { "epoch": 11.83, "learning_rate": 4.550132994102e-05, "loss": 1.0429, "step": 136540 }, { "epoch": 11.83, "learning_rate": 4.5499161558922164e-05, "loss": 1.1341, "step": 136550 }, { "epoch": 11.83, "learning_rate": 4.549699317682433e-05, "loss": 1.1134, "step": 136560 }, { "epoch": 11.83, "learning_rate": 4.549482479472649e-05, "loss": 1.0241, "step": 136570 }, { "epoch": 11.83, "learning_rate": 4.5492656412628657e-05, "loss": 1.011, "step": 136580 }, { "epoch": 11.83, "learning_rate": 4.5490488030530816e-05, "loss": 1.1882, "step": 136590 }, { "epoch": 11.83, "learning_rate": 4.5488319648432976e-05, "loss": 0.9703, "step": 136600 }, { "epoch": 11.83, "learning_rate": 4.5486151266335135e-05, "loss": 1.2258, "step": 136610 }, { "epoch": 11.83, "learning_rate": 4.54839828842373e-05, "loss": 1.1107, "step": 136620 }, { "epoch": 11.83, "learning_rate": 4.548181450213947e-05, "loss": 1.2129, "step": 136630 }, { "epoch": 11.83, "learning_rate": 4.547964612004163e-05, "loss": 1.2152, "step": 136640 }, { "epoch": 11.83, "learning_rate": 4.5477477737943794e-05, "loss": 1.2837, "step": 136650 }, { "epoch": 11.84, "learning_rate": 4.5475309355845954e-05, "loss": 1.1842, "step": 136660 }, { "epoch": 11.84, "learning_rate": 4.547314097374812e-05, "loss": 1.1945, "step": 136670 }, { "epoch": 11.84, "learning_rate": 4.547097259165027e-05, "loss": 1.3065, "step": 136680 }, { "epoch": 11.84, "learning_rate": 4.546880420955244e-05, "loss": 1.032, "step": 136690 }, { "epoch": 11.84, "learning_rate": 4.5466635827454606e-05, "loss": 1.2226, "step": 136700 }, { "epoch": 11.84, "learning_rate": 4.5464467445356765e-05, "loss": 1.0819, "step": 136710 }, { "epoch": 11.84, "learning_rate": 4.546229906325893e-05, "loss": 1.0925, "step": 136720 }, { "epoch": 11.84, "learning_rate": 4.546013068116109e-05, "loss": 1.1893, "step": 136730 }, { "epoch": 11.84, "learning_rate": 4.545796229906326e-05, "loss": 1.0523, "step": 136740 }, { "epoch": 11.84, "learning_rate": 4.545579391696541e-05, "loss": 1.2611, "step": 136750 }, { "epoch": 11.84, "learning_rate": 4.545362553486758e-05, "loss": 1.2784, "step": 136760 }, { "epoch": 11.85, "learning_rate": 4.5451457152769743e-05, "loss": 1.0643, "step": 136770 }, { "epoch": 11.85, "learning_rate": 4.54492887706719e-05, "loss": 1.0276, "step": 136780 }, { "epoch": 11.85, "learning_rate": 4.544712038857407e-05, "loss": 1.1223, "step": 136790 }, { "epoch": 11.85, "learning_rate": 4.544495200647623e-05, "loss": 1.1422, "step": 136800 }, { "epoch": 11.85, "learning_rate": 4.5442783624378395e-05, "loss": 1.0438, "step": 136810 }, { "epoch": 11.85, "learning_rate": 4.544061524228056e-05, "loss": 1.2585, "step": 136820 }, { "epoch": 11.85, "learning_rate": 4.5438446860182715e-05, "loss": 1.1327, "step": 136830 }, { "epoch": 11.85, "learning_rate": 4.543627847808488e-05, "loss": 1.2045, "step": 136840 }, { "epoch": 11.85, "learning_rate": 4.543411009598704e-05, "loss": 1.1364, "step": 136850 }, { "epoch": 11.85, "learning_rate": 4.543194171388921e-05, "loss": 1.272, "step": 136860 }, { "epoch": 11.85, "learning_rate": 4.542977333179137e-05, "loss": 1.1752, "step": 136870 }, { "epoch": 11.85, "learning_rate": 4.542760494969353e-05, "loss": 0.979, "step": 136880 }, { "epoch": 11.86, "learning_rate": 4.54254365675957e-05, "loss": 0.7895, "step": 136890 }, { "epoch": 11.86, "learning_rate": 4.542326818549786e-05, "loss": 1.0914, "step": 136900 }, { "epoch": 11.86, "learning_rate": 4.542109980340002e-05, "loss": 1.0949, "step": 136910 }, { "epoch": 11.86, "learning_rate": 4.541893142130218e-05, "loss": 1.0596, "step": 136920 }, { "epoch": 11.86, "learning_rate": 4.5416763039204345e-05, "loss": 0.9065, "step": 136930 }, { "epoch": 11.86, "learning_rate": 4.5414594657106504e-05, "loss": 1.1444, "step": 136940 }, { "epoch": 11.86, "learning_rate": 4.541242627500867e-05, "loss": 0.9459, "step": 136950 }, { "epoch": 11.86, "learning_rate": 4.541025789291084e-05, "loss": 1.2872, "step": 136960 }, { "epoch": 11.86, "learning_rate": 4.5408089510813e-05, "loss": 1.1054, "step": 136970 }, { "epoch": 11.86, "learning_rate": 4.540592112871516e-05, "loss": 1.2328, "step": 136980 }, { "epoch": 11.86, "learning_rate": 4.5403752746617316e-05, "loss": 1.1452, "step": 136990 }, { "epoch": 11.87, "learning_rate": 4.540158436451948e-05, "loss": 1.263, "step": 137000 }, { "epoch": 11.87, "learning_rate": 4.539941598242164e-05, "loss": 1.1284, "step": 137010 }, { "epoch": 11.87, "learning_rate": 4.539724760032381e-05, "loss": 1.1076, "step": 137020 }, { "epoch": 11.87, "learning_rate": 4.5395079218225975e-05, "loss": 1.2021, "step": 137030 }, { "epoch": 11.87, "learning_rate": 4.5392910836128134e-05, "loss": 1.1968, "step": 137040 }, { "epoch": 11.87, "learning_rate": 4.53907424540303e-05, "loss": 0.898, "step": 137050 }, { "epoch": 11.87, "learning_rate": 4.5388574071932454e-05, "loss": 1.1003, "step": 137060 }, { "epoch": 11.87, "learning_rate": 4.538640568983462e-05, "loss": 1.2841, "step": 137070 }, { "epoch": 11.87, "learning_rate": 4.538423730773678e-05, "loss": 1.2612, "step": 137080 }, { "epoch": 11.87, "learning_rate": 4.5382068925638946e-05, "loss": 1.2022, "step": 137090 }, { "epoch": 11.87, "learning_rate": 4.5379900543541106e-05, "loss": 1.1468, "step": 137100 }, { "epoch": 11.87, "learning_rate": 4.537773216144327e-05, "loss": 1.1608, "step": 137110 }, { "epoch": 11.88, "learning_rate": 4.537556377934544e-05, "loss": 1.1224, "step": 137120 }, { "epoch": 11.88, "learning_rate": 4.53733953972476e-05, "loss": 1.0524, "step": 137130 }, { "epoch": 11.88, "learning_rate": 4.537122701514976e-05, "loss": 1.1269, "step": 137140 }, { "epoch": 11.88, "learning_rate": 4.536905863305192e-05, "loss": 1.1232, "step": 137150 }, { "epoch": 11.88, "learning_rate": 4.5366890250954084e-05, "loss": 1.1006, "step": 137160 }, { "epoch": 11.88, "learning_rate": 4.536472186885624e-05, "loss": 0.9629, "step": 137170 }, { "epoch": 11.88, "learning_rate": 4.536255348675841e-05, "loss": 1.1733, "step": 137180 }, { "epoch": 11.88, "learning_rate": 4.5360385104660576e-05, "loss": 1.0738, "step": 137190 }, { "epoch": 11.88, "learning_rate": 4.5358216722562736e-05, "loss": 0.9345, "step": 137200 }, { "epoch": 11.88, "learning_rate": 4.53560483404649e-05, "loss": 1.145, "step": 137210 }, { "epoch": 11.88, "learning_rate": 4.5353879958367055e-05, "loss": 1.2195, "step": 137220 }, { "epoch": 11.88, "learning_rate": 4.535171157626922e-05, "loss": 1.0506, "step": 137230 }, { "epoch": 11.89, "learning_rate": 4.534954319417138e-05, "loss": 1.0641, "step": 137240 }, { "epoch": 11.89, "learning_rate": 4.534737481207355e-05, "loss": 1.0445, "step": 137250 }, { "epoch": 11.89, "learning_rate": 4.5345206429975714e-05, "loss": 1.0402, "step": 137260 }, { "epoch": 11.89, "learning_rate": 4.534303804787787e-05, "loss": 1.1092, "step": 137270 }, { "epoch": 11.89, "learning_rate": 4.534086966578004e-05, "loss": 1.1502, "step": 137280 }, { "epoch": 11.89, "learning_rate": 4.533870128368219e-05, "loss": 1.0983, "step": 137290 }, { "epoch": 11.89, "learning_rate": 4.533653290158436e-05, "loss": 0.9971, "step": 137300 }, { "epoch": 11.89, "learning_rate": 4.533436451948652e-05, "loss": 0.9398, "step": 137310 }, { "epoch": 11.89, "learning_rate": 4.5332196137388685e-05, "loss": 1.2959, "step": 137320 }, { "epoch": 11.89, "learning_rate": 4.533002775529085e-05, "loss": 1.004, "step": 137330 }, { "epoch": 11.89, "learning_rate": 4.532785937319301e-05, "loss": 1.1818, "step": 137340 }, { "epoch": 11.9, "learning_rate": 4.532569099109518e-05, "loss": 1.176, "step": 137350 }, { "epoch": 11.9, "learning_rate": 4.532352260899734e-05, "loss": 0.9336, "step": 137360 }, { "epoch": 11.9, "learning_rate": 4.5321354226899497e-05, "loss": 1.0976, "step": 137370 }, { "epoch": 11.9, "learning_rate": 4.5319185844801656e-05, "loss": 1.3695, "step": 137380 }, { "epoch": 11.9, "learning_rate": 4.531701746270382e-05, "loss": 1.0922, "step": 137390 }, { "epoch": 11.9, "learning_rate": 4.531484908060599e-05, "loss": 0.9746, "step": 137400 }, { "epoch": 11.9, "learning_rate": 4.531268069850815e-05, "loss": 1.167, "step": 137410 }, { "epoch": 11.9, "learning_rate": 4.5310512316410315e-05, "loss": 1.2652, "step": 137420 }, { "epoch": 11.9, "learning_rate": 4.5308343934312475e-05, "loss": 1.069, "step": 137430 }, { "epoch": 11.9, "learning_rate": 4.530617555221464e-05, "loss": 1.1559, "step": 137440 }, { "epoch": 11.9, "learning_rate": 4.5304007170116794e-05, "loss": 1.2485, "step": 137450 }, { "epoch": 11.9, "learning_rate": 4.530183878801896e-05, "loss": 1.2005, "step": 137460 }, { "epoch": 11.91, "learning_rate": 4.5299670405921127e-05, "loss": 1.1008, "step": 137470 }, { "epoch": 11.91, "learning_rate": 4.5297502023823286e-05, "loss": 1.1414, "step": 137480 }, { "epoch": 11.91, "learning_rate": 4.529533364172545e-05, "loss": 1.0997, "step": 137490 }, { "epoch": 11.91, "learning_rate": 4.529316525962761e-05, "loss": 1.0508, "step": 137500 }, { "epoch": 11.91, "learning_rate": 4.529099687752978e-05, "loss": 1.0105, "step": 137510 }, { "epoch": 11.91, "learning_rate": 4.528882849543193e-05, "loss": 1.0598, "step": 137520 }, { "epoch": 11.91, "learning_rate": 4.52866601133341e-05, "loss": 1.0969, "step": 137530 }, { "epoch": 11.91, "learning_rate": 4.5284491731236264e-05, "loss": 1.2124, "step": 137540 }, { "epoch": 11.91, "learning_rate": 4.5282323349138424e-05, "loss": 1.3352, "step": 137550 }, { "epoch": 11.91, "learning_rate": 4.528015496704059e-05, "loss": 1.3061, "step": 137560 }, { "epoch": 11.91, "learning_rate": 4.527798658494275e-05, "loss": 1.18, "step": 137570 }, { "epoch": 11.92, "learning_rate": 4.5275818202844916e-05, "loss": 1.1509, "step": 137580 }, { "epoch": 11.92, "learning_rate": 4.527364982074708e-05, "loss": 1.1398, "step": 137590 }, { "epoch": 11.92, "learning_rate": 4.5271481438649235e-05, "loss": 1.3208, "step": 137600 }, { "epoch": 11.92, "learning_rate": 4.5269313056551395e-05, "loss": 1.2093, "step": 137610 }, { "epoch": 11.92, "learning_rate": 4.526714467445356e-05, "loss": 1.082, "step": 137620 }, { "epoch": 11.92, "learning_rate": 4.526497629235573e-05, "loss": 1.0665, "step": 137630 }, { "epoch": 11.92, "learning_rate": 4.526280791025789e-05, "loss": 1.0128, "step": 137640 }, { "epoch": 11.92, "learning_rate": 4.5260639528160054e-05, "loss": 1.0633, "step": 137650 }, { "epoch": 11.92, "learning_rate": 4.5258471146062213e-05, "loss": 1.1614, "step": 137660 }, { "epoch": 11.92, "learning_rate": 4.525630276396438e-05, "loss": 0.9351, "step": 137670 }, { "epoch": 11.92, "learning_rate": 4.525413438186653e-05, "loss": 1.2425, "step": 137680 }, { "epoch": 11.92, "learning_rate": 4.52519659997687e-05, "loss": 1.2221, "step": 137690 }, { "epoch": 11.93, "learning_rate": 4.5249797617670865e-05, "loss": 1.2994, "step": 137700 }, { "epoch": 11.93, "learning_rate": 4.5247629235573025e-05, "loss": 1.403, "step": 137710 }, { "epoch": 11.93, "learning_rate": 4.524546085347519e-05, "loss": 1.3136, "step": 137720 }, { "epoch": 11.93, "learning_rate": 4.524329247137735e-05, "loss": 1.0793, "step": 137730 }, { "epoch": 11.93, "learning_rate": 4.524112408927952e-05, "loss": 1.1854, "step": 137740 }, { "epoch": 11.93, "learning_rate": 4.5238955707181684e-05, "loss": 1.1152, "step": 137750 }, { "epoch": 11.93, "learning_rate": 4.523678732508384e-05, "loss": 1.062, "step": 137760 }, { "epoch": 11.93, "learning_rate": 4.5234618942986e-05, "loss": 1.0511, "step": 137770 }, { "epoch": 11.93, "learning_rate": 4.523245056088816e-05, "loss": 1.1045, "step": 137780 }, { "epoch": 11.93, "learning_rate": 4.523028217879033e-05, "loss": 1.1321, "step": 137790 }, { "epoch": 11.93, "learning_rate": 4.522811379669249e-05, "loss": 1.1226, "step": 137800 }, { "epoch": 11.94, "learning_rate": 4.5225945414594655e-05, "loss": 1.0883, "step": 137810 }, { "epoch": 11.94, "learning_rate": 4.522377703249682e-05, "loss": 1.2176, "step": 137820 }, { "epoch": 11.94, "learning_rate": 4.5221608650398974e-05, "loss": 1.2362, "step": 137830 }, { "epoch": 11.94, "learning_rate": 4.521944026830114e-05, "loss": 1.0703, "step": 137840 }, { "epoch": 11.94, "learning_rate": 4.52172718862033e-05, "loss": 1.0186, "step": 137850 }, { "epoch": 11.94, "learning_rate": 4.521510350410547e-05, "loss": 1.3341, "step": 137860 }, { "epoch": 11.94, "learning_rate": 4.5212935122007626e-05, "loss": 1.0721, "step": 137870 }, { "epoch": 11.94, "learning_rate": 4.521076673990979e-05, "loss": 1.3032, "step": 137880 }, { "epoch": 11.94, "learning_rate": 4.520859835781196e-05, "loss": 1.1725, "step": 137890 }, { "epoch": 11.94, "learning_rate": 4.520642997571412e-05, "loss": 1.064, "step": 137900 }, { "epoch": 11.94, "learning_rate": 4.520426159361628e-05, "loss": 1.2399, "step": 137910 }, { "epoch": 11.94, "learning_rate": 4.520209321151844e-05, "loss": 1.1343, "step": 137920 }, { "epoch": 11.95, "learning_rate": 4.5199924829420604e-05, "loss": 0.8808, "step": 137930 }, { "epoch": 11.95, "learning_rate": 4.5197756447322764e-05, "loss": 1.0101, "step": 137940 }, { "epoch": 11.95, "learning_rate": 4.519558806522493e-05, "loss": 1.1893, "step": 137950 }, { "epoch": 11.95, "learning_rate": 4.51934196831271e-05, "loss": 0.9978, "step": 137960 }, { "epoch": 11.95, "learning_rate": 4.5191251301029256e-05, "loss": 1.1041, "step": 137970 }, { "epoch": 11.95, "learning_rate": 4.518908291893142e-05, "loss": 1.0596, "step": 137980 }, { "epoch": 11.95, "learning_rate": 4.5186914536833576e-05, "loss": 0.9797, "step": 137990 }, { "epoch": 11.95, "learning_rate": 4.518474615473574e-05, "loss": 1.2895, "step": 138000 }, { "epoch": 11.95, "learning_rate": 4.51825777726379e-05, "loss": 1.2755, "step": 138010 }, { "epoch": 11.95, "learning_rate": 4.518040939054007e-05, "loss": 1.1385, "step": 138020 }, { "epoch": 11.95, "learning_rate": 4.5178241008442234e-05, "loss": 1.1903, "step": 138030 }, { "epoch": 11.96, "learning_rate": 4.5176072626344394e-05, "loss": 1.1814, "step": 138040 }, { "epoch": 11.96, "learning_rate": 4.517390424424656e-05, "loss": 1.1697, "step": 138050 }, { "epoch": 11.96, "learning_rate": 4.517173586214871e-05, "loss": 1.2017, "step": 138060 }, { "epoch": 11.96, "learning_rate": 4.516956748005088e-05, "loss": 1.166, "step": 138070 }, { "epoch": 11.96, "learning_rate": 4.516739909795304e-05, "loss": 0.9144, "step": 138080 }, { "epoch": 11.96, "learning_rate": 4.5165230715855206e-05, "loss": 1.0793, "step": 138090 }, { "epoch": 11.96, "learning_rate": 4.516306233375737e-05, "loss": 1.122, "step": 138100 }, { "epoch": 11.96, "learning_rate": 4.516089395165953e-05, "loss": 1.0036, "step": 138110 }, { "epoch": 11.96, "learning_rate": 4.51587255695617e-05, "loss": 1.2242, "step": 138120 }, { "epoch": 11.96, "learning_rate": 4.515655718746386e-05, "loss": 1.1072, "step": 138130 }, { "epoch": 11.96, "learning_rate": 4.515438880536602e-05, "loss": 1.0819, "step": 138140 }, { "epoch": 11.96, "learning_rate": 4.515222042326818e-05, "loss": 1.3426, "step": 138150 }, { "epoch": 11.97, "learning_rate": 4.515005204117034e-05, "loss": 1.1849, "step": 138160 }, { "epoch": 11.97, "learning_rate": 4.51478836590725e-05, "loss": 1.1435, "step": 138170 }, { "epoch": 11.97, "learning_rate": 4.514571527697467e-05, "loss": 1.0748, "step": 138180 }, { "epoch": 11.97, "learning_rate": 4.5143546894876836e-05, "loss": 1.1897, "step": 138190 }, { "epoch": 11.97, "learning_rate": 4.5141378512778995e-05, "loss": 1.0851, "step": 138200 }, { "epoch": 11.97, "learning_rate": 4.513921013068116e-05, "loss": 1.066, "step": 138210 }, { "epoch": 11.97, "learning_rate": 4.5137041748583315e-05, "loss": 1.1853, "step": 138220 }, { "epoch": 11.97, "learning_rate": 4.513487336648548e-05, "loss": 1.0872, "step": 138230 }, { "epoch": 11.97, "learning_rate": 4.513270498438764e-05, "loss": 1.087, "step": 138240 }, { "epoch": 11.97, "learning_rate": 4.513053660228981e-05, "loss": 1.2364, "step": 138250 }, { "epoch": 11.97, "learning_rate": 4.512836822019197e-05, "loss": 1.2259, "step": 138260 }, { "epoch": 11.98, "learning_rate": 4.512619983809413e-05, "loss": 1.0505, "step": 138270 }, { "epoch": 11.98, "learning_rate": 4.51240314559963e-05, "loss": 1.0649, "step": 138280 }, { "epoch": 11.98, "learning_rate": 4.512186307389845e-05, "loss": 1.1331, "step": 138290 }, { "epoch": 11.98, "learning_rate": 4.511969469180062e-05, "loss": 1.0824, "step": 138300 }, { "epoch": 11.98, "learning_rate": 4.511752630970278e-05, "loss": 0.9668, "step": 138310 }, { "epoch": 11.98, "learning_rate": 4.5115357927604945e-05, "loss": 1.0793, "step": 138320 }, { "epoch": 11.98, "learning_rate": 4.511318954550711e-05, "loss": 0.9985, "step": 138330 }, { "epoch": 11.98, "learning_rate": 4.511102116340927e-05, "loss": 1.1042, "step": 138340 }, { "epoch": 11.98, "learning_rate": 4.510885278131144e-05, "loss": 1.0829, "step": 138350 }, { "epoch": 11.98, "learning_rate": 4.51066843992136e-05, "loss": 1.2408, "step": 138360 }, { "epoch": 11.98, "learning_rate": 4.5104516017115756e-05, "loss": 1.0139, "step": 138370 }, { "epoch": 11.98, "learning_rate": 4.5102347635017916e-05, "loss": 1.1765, "step": 138380 }, { "epoch": 11.99, "learning_rate": 4.510017925292008e-05, "loss": 1.1495, "step": 138390 }, { "epoch": 11.99, "learning_rate": 4.509801087082225e-05, "loss": 1.0066, "step": 138400 }, { "epoch": 11.99, "learning_rate": 4.509584248872441e-05, "loss": 1.0711, "step": 138410 }, { "epoch": 11.99, "learning_rate": 4.5093674106626575e-05, "loss": 1.3048, "step": 138420 }, { "epoch": 11.99, "learning_rate": 4.5091505724528734e-05, "loss": 1.2322, "step": 138430 }, { "epoch": 11.99, "learning_rate": 4.50893373424309e-05, "loss": 0.9925, "step": 138440 }, { "epoch": 11.99, "learning_rate": 4.5087168960333053e-05, "loss": 1.1468, "step": 138450 }, { "epoch": 11.99, "learning_rate": 4.508500057823522e-05, "loss": 1.0507, "step": 138460 }, { "epoch": 11.99, "learning_rate": 4.5082832196137386e-05, "loss": 1.2369, "step": 138470 }, { "epoch": 11.99, "learning_rate": 4.5080663814039546e-05, "loss": 0.9804, "step": 138480 }, { "epoch": 11.99, "learning_rate": 4.507849543194171e-05, "loss": 1.2554, "step": 138490 }, { "epoch": 11.99, "learning_rate": 4.507632704984387e-05, "loss": 1.0607, "step": 138500 }, { "epoch": 12.0, "learning_rate": 4.507415866774604e-05, "loss": 1.076, "step": 138510 }, { "epoch": 12.0, "learning_rate": 4.5071990285648205e-05, "loss": 1.0795, "step": 138520 }, { "epoch": 12.0, "learning_rate": 4.506982190355036e-05, "loss": 1.1015, "step": 138530 }, { "epoch": 12.0, "learning_rate": 4.5067653521452524e-05, "loss": 1.0151, "step": 138540 }, { "epoch": 12.0, "learning_rate": 4.5065485139354684e-05, "loss": 1.2042, "step": 138550 }, { "epoch": 12.0, "eval_cer": 0.5406648092738251, "eval_em": 0.039711191335740074, "eval_f1": 0.039711191335740074, "eval_loss": 1.4674012660980225, "eval_runtime": 254.4114, "eval_samples_per_second": 7.622, "eval_steps_per_second": 0.955, "eval_wer": 0.9602888086642599, "step": 138558 }, { "epoch": 12.0, "learning_rate": 4.506331675725685e-05, "loss": 1.0252, "step": 138560 }, { "epoch": 12.0, "learning_rate": 4.506114837515901e-05, "loss": 1.0558, "step": 138570 }, { "epoch": 12.0, "learning_rate": 4.5058979993061176e-05, "loss": 1.0327, "step": 138580 }, { "epoch": 12.0, "learning_rate": 4.505681161096334e-05, "loss": 1.1003, "step": 138590 }, { "epoch": 12.0, "learning_rate": 4.5054643228865495e-05, "loss": 1.0243, "step": 138600 }, { "epoch": 12.0, "learning_rate": 4.505247484676766e-05, "loss": 1.0464, "step": 138610 }, { "epoch": 12.01, "learning_rate": 4.505030646466982e-05, "loss": 1.0517, "step": 138620 }, { "epoch": 12.01, "learning_rate": 4.504813808257199e-05, "loss": 1.086, "step": 138630 }, { "epoch": 12.01, "learning_rate": 4.504596970047415e-05, "loss": 0.8667, "step": 138640 }, { "epoch": 12.01, "learning_rate": 4.5043801318376314e-05, "loss": 0.8928, "step": 138650 }, { "epoch": 12.01, "learning_rate": 4.504163293627848e-05, "loss": 1.0379, "step": 138660 }, { "epoch": 12.01, "learning_rate": 4.503946455418064e-05, "loss": 1.0107, "step": 138670 }, { "epoch": 12.01, "learning_rate": 4.503729617208279e-05, "loss": 0.9702, "step": 138680 }, { "epoch": 12.01, "learning_rate": 4.503512778998496e-05, "loss": 0.9687, "step": 138690 }, { "epoch": 12.01, "learning_rate": 4.5032959407887125e-05, "loss": 0.6792, "step": 138700 }, { "epoch": 12.01, "learning_rate": 4.5030791025789285e-05, "loss": 1.1459, "step": 138710 }, { "epoch": 12.01, "learning_rate": 4.502862264369145e-05, "loss": 0.8583, "step": 138720 }, { "epoch": 12.01, "learning_rate": 4.502645426159361e-05, "loss": 0.9931, "step": 138730 }, { "epoch": 12.02, "learning_rate": 4.502428587949578e-05, "loss": 1.0563, "step": 138740 }, { "epoch": 12.02, "learning_rate": 4.5022117497397944e-05, "loss": 0.9802, "step": 138750 }, { "epoch": 12.02, "learning_rate": 4.5019949115300096e-05, "loss": 0.9693, "step": 138760 }, { "epoch": 12.02, "learning_rate": 4.501778073320226e-05, "loss": 1.1265, "step": 138770 }, { "epoch": 12.02, "learning_rate": 4.501561235110442e-05, "loss": 1.1761, "step": 138780 }, { "epoch": 12.02, "learning_rate": 4.501344396900659e-05, "loss": 1.0123, "step": 138790 }, { "epoch": 12.02, "learning_rate": 4.501127558690875e-05, "loss": 1.0797, "step": 138800 }, { "epoch": 12.02, "learning_rate": 4.5009107204810915e-05, "loss": 1.2673, "step": 138810 }, { "epoch": 12.02, "learning_rate": 4.500693882271308e-05, "loss": 0.9889, "step": 138820 }, { "epoch": 12.02, "learning_rate": 4.5004770440615234e-05, "loss": 0.976, "step": 138830 }, { "epoch": 12.02, "learning_rate": 4.50026020585174e-05, "loss": 0.9827, "step": 138840 }, { "epoch": 12.03, "learning_rate": 4.500043367641956e-05, "loss": 1.1339, "step": 138850 }, { "epoch": 12.03, "learning_rate": 4.4998265294321726e-05, "loss": 0.9858, "step": 138860 }, { "epoch": 12.03, "learning_rate": 4.4996096912223886e-05, "loss": 1.0234, "step": 138870 }, { "epoch": 12.03, "learning_rate": 4.499392853012605e-05, "loss": 1.1236, "step": 138880 }, { "epoch": 12.03, "learning_rate": 4.499176014802822e-05, "loss": 0.7502, "step": 138890 }, { "epoch": 12.03, "learning_rate": 4.498959176593038e-05, "loss": 1.1602, "step": 138900 }, { "epoch": 12.03, "learning_rate": 4.498742338383254e-05, "loss": 0.869, "step": 138910 }, { "epoch": 12.03, "learning_rate": 4.49852550017347e-05, "loss": 1.0662, "step": 138920 }, { "epoch": 12.03, "learning_rate": 4.4983086619636864e-05, "loss": 1.0267, "step": 138930 }, { "epoch": 12.03, "learning_rate": 4.4980918237539024e-05, "loss": 0.9092, "step": 138940 }, { "epoch": 12.03, "learning_rate": 4.497874985544119e-05, "loss": 0.9428, "step": 138950 }, { "epoch": 12.03, "learning_rate": 4.4976581473343356e-05, "loss": 1.1283, "step": 138960 }, { "epoch": 12.04, "learning_rate": 4.4974413091245516e-05, "loss": 1.0758, "step": 138970 }, { "epoch": 12.04, "learning_rate": 4.497224470914768e-05, "loss": 0.9314, "step": 138980 }, { "epoch": 12.04, "learning_rate": 4.4970076327049835e-05, "loss": 1.0491, "step": 138990 }, { "epoch": 12.04, "learning_rate": 4.4967907944952e-05, "loss": 1.0396, "step": 139000 }, { "epoch": 12.04, "learning_rate": 4.496573956285416e-05, "loss": 0.9175, "step": 139010 }, { "epoch": 12.04, "learning_rate": 4.496357118075633e-05, "loss": 0.9998, "step": 139020 }, { "epoch": 12.04, "learning_rate": 4.4961402798658494e-05, "loss": 0.9123, "step": 139030 }, { "epoch": 12.04, "learning_rate": 4.4959234416560654e-05, "loss": 0.9722, "step": 139040 }, { "epoch": 12.04, "learning_rate": 4.495706603446282e-05, "loss": 0.7993, "step": 139050 }, { "epoch": 12.04, "learning_rate": 4.495489765236497e-05, "loss": 1.1293, "step": 139060 }, { "epoch": 12.04, "learning_rate": 4.495272927026714e-05, "loss": 0.96, "step": 139070 }, { "epoch": 12.05, "learning_rate": 4.49505608881693e-05, "loss": 1.0911, "step": 139080 }, { "epoch": 12.05, "learning_rate": 4.4948392506071465e-05, "loss": 1.1128, "step": 139090 }, { "epoch": 12.05, "learning_rate": 4.494622412397363e-05, "loss": 1.1621, "step": 139100 }, { "epoch": 12.05, "learning_rate": 4.494405574187579e-05, "loss": 1.0426, "step": 139110 }, { "epoch": 12.05, "learning_rate": 4.494188735977796e-05, "loss": 1.061, "step": 139120 }, { "epoch": 12.05, "learning_rate": 4.493971897768012e-05, "loss": 1.051, "step": 139130 }, { "epoch": 12.05, "learning_rate": 4.493755059558228e-05, "loss": 1.0939, "step": 139140 }, { "epoch": 12.05, "learning_rate": 4.493538221348444e-05, "loss": 1.0201, "step": 139150 }, { "epoch": 12.05, "learning_rate": 4.49332138313866e-05, "loss": 0.9716, "step": 139160 }, { "epoch": 12.05, "learning_rate": 4.493104544928877e-05, "loss": 1.0653, "step": 139170 }, { "epoch": 12.05, "learning_rate": 4.492887706719093e-05, "loss": 0.9825, "step": 139180 }, { "epoch": 12.05, "learning_rate": 4.4926708685093095e-05, "loss": 1.1776, "step": 139190 }, { "epoch": 12.06, "learning_rate": 4.4924540302995255e-05, "loss": 1.1034, "step": 139200 }, { "epoch": 12.06, "learning_rate": 4.492237192089742e-05, "loss": 1.0919, "step": 139210 }, { "epoch": 12.06, "learning_rate": 4.4920203538799574e-05, "loss": 1.0254, "step": 139220 }, { "epoch": 12.06, "learning_rate": 4.491803515670174e-05, "loss": 0.9206, "step": 139230 }, { "epoch": 12.06, "learning_rate": 4.49158667746039e-05, "loss": 0.9797, "step": 139240 }, { "epoch": 12.06, "learning_rate": 4.491369839250607e-05, "loss": 1.1587, "step": 139250 }, { "epoch": 12.06, "learning_rate": 4.491153001040823e-05, "loss": 1.1679, "step": 139260 }, { "epoch": 12.06, "learning_rate": 4.490936162831039e-05, "loss": 1.0514, "step": 139270 }, { "epoch": 12.06, "learning_rate": 4.490719324621256e-05, "loss": 0.9544, "step": 139280 }, { "epoch": 12.06, "learning_rate": 4.490502486411472e-05, "loss": 1.0305, "step": 139290 }, { "epoch": 12.06, "learning_rate": 4.490285648201688e-05, "loss": 1.0459, "step": 139300 }, { "epoch": 12.07, "learning_rate": 4.490068809991904e-05, "loss": 1.0689, "step": 139310 }, { "epoch": 12.07, "learning_rate": 4.4898519717821204e-05, "loss": 1.0672, "step": 139320 }, { "epoch": 12.07, "learning_rate": 4.489635133572337e-05, "loss": 1.0198, "step": 139330 }, { "epoch": 12.07, "learning_rate": 4.489418295362553e-05, "loss": 1.0434, "step": 139340 }, { "epoch": 12.07, "learning_rate": 4.48920145715277e-05, "loss": 1.0398, "step": 139350 }, { "epoch": 12.07, "learning_rate": 4.4889846189429856e-05, "loss": 1.1828, "step": 139360 }, { "epoch": 12.07, "learning_rate": 4.4887677807332016e-05, "loss": 1.0902, "step": 139370 }, { "epoch": 12.07, "learning_rate": 4.4885509425234176e-05, "loss": 1.0281, "step": 139380 }, { "epoch": 12.07, "learning_rate": 4.488334104313634e-05, "loss": 1.0018, "step": 139390 }, { "epoch": 12.07, "learning_rate": 4.488117266103851e-05, "loss": 1.0123, "step": 139400 }, { "epoch": 12.07, "learning_rate": 4.487900427894067e-05, "loss": 0.9456, "step": 139410 }, { "epoch": 12.07, "learning_rate": 4.4876835896842834e-05, "loss": 1.1498, "step": 139420 }, { "epoch": 12.08, "learning_rate": 4.4874667514744994e-05, "loss": 1.1722, "step": 139430 }, { "epoch": 12.08, "learning_rate": 4.487249913264716e-05, "loss": 1.0736, "step": 139440 }, { "epoch": 12.08, "learning_rate": 4.487033075054931e-05, "loss": 0.941, "step": 139450 }, { "epoch": 12.08, "learning_rate": 4.486816236845148e-05, "loss": 0.9349, "step": 139460 }, { "epoch": 12.08, "learning_rate": 4.4865993986353646e-05, "loss": 1.2838, "step": 139470 }, { "epoch": 12.08, "learning_rate": 4.4863825604255806e-05, "loss": 0.9577, "step": 139480 }, { "epoch": 12.08, "learning_rate": 4.486165722215797e-05, "loss": 0.8842, "step": 139490 }, { "epoch": 12.08, "learning_rate": 4.485948884006013e-05, "loss": 1.0953, "step": 139500 }, { "epoch": 12.08, "learning_rate": 4.48573204579623e-05, "loss": 0.9812, "step": 139510 }, { "epoch": 12.08, "learning_rate": 4.4855152075864464e-05, "loss": 1.1323, "step": 139520 }, { "epoch": 12.08, "learning_rate": 4.485298369376662e-05, "loss": 1.0899, "step": 139530 }, { "epoch": 12.09, "learning_rate": 4.4850815311668784e-05, "loss": 1.0458, "step": 139540 }, { "epoch": 12.09, "learning_rate": 4.484864692957094e-05, "loss": 1.141, "step": 139550 }, { "epoch": 12.09, "learning_rate": 4.484647854747311e-05, "loss": 1.032, "step": 139560 }, { "epoch": 12.09, "learning_rate": 4.484431016537527e-05, "loss": 1.0247, "step": 139570 }, { "epoch": 12.09, "learning_rate": 4.4842141783277436e-05, "loss": 1.0793, "step": 139580 }, { "epoch": 12.09, "learning_rate": 4.48399734011796e-05, "loss": 0.8128, "step": 139590 }, { "epoch": 12.09, "learning_rate": 4.4837805019081755e-05, "loss": 0.9503, "step": 139600 }, { "epoch": 12.09, "learning_rate": 4.483563663698392e-05, "loss": 1.1071, "step": 139610 }, { "epoch": 12.09, "learning_rate": 4.483346825488608e-05, "loss": 1.1212, "step": 139620 }, { "epoch": 12.09, "learning_rate": 4.483129987278825e-05, "loss": 1.0745, "step": 139630 }, { "epoch": 12.09, "learning_rate": 4.482913149069041e-05, "loss": 1.1817, "step": 139640 }, { "epoch": 12.09, "learning_rate": 4.482696310859257e-05, "loss": 1.0052, "step": 139650 }, { "epoch": 12.1, "learning_rate": 4.482479472649474e-05, "loss": 0.9565, "step": 139660 }, { "epoch": 12.1, "learning_rate": 4.48226263443969e-05, "loss": 1.1375, "step": 139670 }, { "epoch": 12.1, "learning_rate": 4.482045796229906e-05, "loss": 1.1566, "step": 139680 }, { "epoch": 12.1, "learning_rate": 4.481828958020122e-05, "loss": 0.9937, "step": 139690 }, { "epoch": 12.1, "learning_rate": 4.4816121198103385e-05, "loss": 1.0786, "step": 139700 }, { "epoch": 12.1, "learning_rate": 4.4813952816005544e-05, "loss": 1.1727, "step": 139710 }, { "epoch": 12.1, "learning_rate": 4.481178443390771e-05, "loss": 1.1419, "step": 139720 }, { "epoch": 12.1, "learning_rate": 4.480961605180988e-05, "loss": 0.9654, "step": 139730 }, { "epoch": 12.1, "learning_rate": 4.480744766971204e-05, "loss": 1.1041, "step": 139740 }, { "epoch": 12.1, "learning_rate": 4.48052792876142e-05, "loss": 1.0028, "step": 139750 }, { "epoch": 12.1, "learning_rate": 4.4803110905516356e-05, "loss": 0.9913, "step": 139760 }, { "epoch": 12.1, "learning_rate": 4.480094252341852e-05, "loss": 0.9139, "step": 139770 }, { "epoch": 12.11, "learning_rate": 4.479877414132068e-05, "loss": 1.0702, "step": 139780 }, { "epoch": 12.11, "learning_rate": 4.479660575922285e-05, "loss": 1.0525, "step": 139790 }, { "epoch": 12.11, "learning_rate": 4.479443737712501e-05, "loss": 1.0363, "step": 139800 }, { "epoch": 12.11, "learning_rate": 4.4792268995027175e-05, "loss": 1.0111, "step": 139810 }, { "epoch": 12.11, "learning_rate": 4.479010061292934e-05, "loss": 1.1197, "step": 139820 }, { "epoch": 12.11, "learning_rate": 4.4787932230831494e-05, "loss": 1.17, "step": 139830 }, { "epoch": 12.11, "learning_rate": 4.478576384873366e-05, "loss": 1.0621, "step": 139840 }, { "epoch": 12.11, "learning_rate": 4.478359546663582e-05, "loss": 1.2329, "step": 139850 }, { "epoch": 12.11, "learning_rate": 4.4781427084537986e-05, "loss": 1.0179, "step": 139860 }, { "epoch": 12.11, "learning_rate": 4.4779258702440146e-05, "loss": 1.129, "step": 139870 }, { "epoch": 12.11, "learning_rate": 4.477709032034231e-05, "loss": 1.0736, "step": 139880 }, { "epoch": 12.12, "learning_rate": 4.477492193824448e-05, "loss": 1.098, "step": 139890 }, { "epoch": 12.12, "learning_rate": 4.477275355614664e-05, "loss": 1.0744, "step": 139900 }, { "epoch": 12.12, "learning_rate": 4.47705851740488e-05, "loss": 1.0012, "step": 139910 }, { "epoch": 12.12, "learning_rate": 4.476841679195096e-05, "loss": 1.1304, "step": 139920 }, { "epoch": 12.12, "learning_rate": 4.4766248409853124e-05, "loss": 1.1202, "step": 139930 }, { "epoch": 12.12, "learning_rate": 4.4764080027755283e-05, "loss": 1.1568, "step": 139940 }, { "epoch": 12.12, "learning_rate": 4.476191164565745e-05, "loss": 1.0256, "step": 139950 }, { "epoch": 12.12, "learning_rate": 4.4759743263559616e-05, "loss": 0.9509, "step": 139960 }, { "epoch": 12.12, "learning_rate": 4.4757574881461776e-05, "loss": 1.0145, "step": 139970 }, { "epoch": 12.12, "learning_rate": 4.475540649936394e-05, "loss": 0.9557, "step": 139980 }, { "epoch": 12.12, "learning_rate": 4.4753238117266095e-05, "loss": 1.1275, "step": 139990 }, { "epoch": 12.12, "learning_rate": 4.475106973516826e-05, "loss": 1.0241, "step": 140000 }, { "epoch": 12.13, "learning_rate": 4.474890135307042e-05, "loss": 0.9964, "step": 140010 }, { "epoch": 12.13, "learning_rate": 4.474673297097259e-05, "loss": 0.9823, "step": 140020 }, { "epoch": 12.13, "learning_rate": 4.4744564588874754e-05, "loss": 1.1262, "step": 140030 }, { "epoch": 12.13, "learning_rate": 4.4742396206776913e-05, "loss": 0.9642, "step": 140040 }, { "epoch": 12.13, "learning_rate": 4.474022782467908e-05, "loss": 1.1142, "step": 140050 }, { "epoch": 12.13, "learning_rate": 4.473805944258124e-05, "loss": 1.1694, "step": 140060 }, { "epoch": 12.13, "learning_rate": 4.47358910604834e-05, "loss": 0.9979, "step": 140070 }, { "epoch": 12.13, "learning_rate": 4.473372267838556e-05, "loss": 0.7968, "step": 140080 }, { "epoch": 12.13, "learning_rate": 4.4731554296287725e-05, "loss": 0.8412, "step": 140090 }, { "epoch": 12.13, "learning_rate": 4.472938591418989e-05, "loss": 1.0549, "step": 140100 }, { "epoch": 12.13, "learning_rate": 4.472721753209205e-05, "loss": 0.9658, "step": 140110 }, { "epoch": 12.14, "learning_rate": 4.472504914999422e-05, "loss": 0.9754, "step": 140120 }, { "epoch": 12.14, "learning_rate": 4.472288076789638e-05, "loss": 1.0289, "step": 140130 }, { "epoch": 12.14, "learning_rate": 4.472071238579854e-05, "loss": 1.1822, "step": 140140 }, { "epoch": 12.14, "learning_rate": 4.4718544003700696e-05, "loss": 1.0155, "step": 140150 }, { "epoch": 12.14, "learning_rate": 4.471637562160286e-05, "loss": 1.1855, "step": 140160 }, { "epoch": 12.14, "learning_rate": 4.471420723950503e-05, "loss": 0.9969, "step": 140170 }, { "epoch": 12.14, "learning_rate": 4.471203885740719e-05, "loss": 0.8686, "step": 140180 }, { "epoch": 12.14, "learning_rate": 4.4709870475309355e-05, "loss": 1.0623, "step": 140190 }, { "epoch": 12.14, "learning_rate": 4.4707702093211515e-05, "loss": 1.082, "step": 140200 }, { "epoch": 12.14, "learning_rate": 4.470553371111368e-05, "loss": 0.9757, "step": 140210 }, { "epoch": 12.14, "learning_rate": 4.4703365329015834e-05, "loss": 1.0001, "step": 140220 }, { "epoch": 12.14, "learning_rate": 4.4701196946918e-05, "loss": 1.3943, "step": 140230 }, { "epoch": 12.15, "learning_rate": 4.469902856482017e-05, "loss": 1.1107, "step": 140240 }, { "epoch": 12.15, "learning_rate": 4.4696860182722326e-05, "loss": 0.9914, "step": 140250 }, { "epoch": 12.15, "learning_rate": 4.469469180062449e-05, "loss": 1.1461, "step": 140260 }, { "epoch": 12.15, "learning_rate": 4.469252341852665e-05, "loss": 1.0262, "step": 140270 }, { "epoch": 12.15, "learning_rate": 4.469035503642882e-05, "loss": 1.0497, "step": 140280 }, { "epoch": 12.15, "learning_rate": 4.4688186654330985e-05, "loss": 1.0999, "step": 140290 }, { "epoch": 12.15, "learning_rate": 4.468601827223314e-05, "loss": 0.9461, "step": 140300 }, { "epoch": 12.15, "learning_rate": 4.46838498901353e-05, "loss": 1.0757, "step": 140310 }, { "epoch": 12.15, "learning_rate": 4.4681681508037464e-05, "loss": 1.0355, "step": 140320 }, { "epoch": 12.15, "learning_rate": 4.467951312593963e-05, "loss": 1.0245, "step": 140330 }, { "epoch": 12.15, "learning_rate": 4.467734474384179e-05, "loss": 1.0108, "step": 140340 }, { "epoch": 12.16, "learning_rate": 4.4675176361743956e-05, "loss": 0.9816, "step": 140350 }, { "epoch": 12.16, "learning_rate": 4.4673007979646116e-05, "loss": 1.0577, "step": 140360 }, { "epoch": 12.16, "learning_rate": 4.4670839597548276e-05, "loss": 0.9499, "step": 140370 }, { "epoch": 12.16, "learning_rate": 4.4668671215450435e-05, "loss": 1.1992, "step": 140380 }, { "epoch": 12.16, "learning_rate": 4.46665028333526e-05, "loss": 1.1279, "step": 140390 }, { "epoch": 12.16, "learning_rate": 4.466433445125477e-05, "loss": 1.0196, "step": 140400 }, { "epoch": 12.16, "learning_rate": 4.466216606915693e-05, "loss": 1.0792, "step": 140410 }, { "epoch": 12.16, "learning_rate": 4.4659997687059094e-05, "loss": 1.0227, "step": 140420 }, { "epoch": 12.16, "learning_rate": 4.4657829304961254e-05, "loss": 1.0566, "step": 140430 }, { "epoch": 12.16, "learning_rate": 4.465566092286342e-05, "loss": 1.0495, "step": 140440 }, { "epoch": 12.16, "learning_rate": 4.465349254076557e-05, "loss": 1.2141, "step": 140450 }, { "epoch": 12.16, "learning_rate": 4.465132415866774e-05, "loss": 1.1037, "step": 140460 }, { "epoch": 12.17, "learning_rate": 4.4649155776569906e-05, "loss": 1.1541, "step": 140470 }, { "epoch": 12.17, "learning_rate": 4.4646987394472065e-05, "loss": 1.0805, "step": 140480 }, { "epoch": 12.17, "learning_rate": 4.464481901237423e-05, "loss": 0.9104, "step": 140490 }, { "epoch": 12.17, "learning_rate": 4.464265063027639e-05, "loss": 0.986, "step": 140500 }, { "epoch": 12.17, "learning_rate": 4.464048224817856e-05, "loss": 1.1751, "step": 140510 }, { "epoch": 12.17, "learning_rate": 4.4638313866080724e-05, "loss": 1.0811, "step": 140520 }, { "epoch": 12.17, "learning_rate": 4.463614548398288e-05, "loss": 1.0811, "step": 140530 }, { "epoch": 12.17, "learning_rate": 4.463397710188504e-05, "loss": 1.0282, "step": 140540 }, { "epoch": 12.17, "learning_rate": 4.46318087197872e-05, "loss": 1.0735, "step": 140550 }, { "epoch": 12.17, "learning_rate": 4.462964033768937e-05, "loss": 1.0639, "step": 140560 }, { "epoch": 12.17, "learning_rate": 4.462747195559153e-05, "loss": 0.7867, "step": 140570 }, { "epoch": 12.18, "learning_rate": 4.4625303573493695e-05, "loss": 0.8122, "step": 140580 }, { "epoch": 12.18, "learning_rate": 4.462313519139586e-05, "loss": 1.0044, "step": 140590 }, { "epoch": 12.18, "learning_rate": 4.4620966809298015e-05, "loss": 1.0668, "step": 140600 }, { "epoch": 12.18, "learning_rate": 4.461879842720018e-05, "loss": 1.0821, "step": 140610 }, { "epoch": 12.18, "learning_rate": 4.461663004510234e-05, "loss": 1.0757, "step": 140620 }, { "epoch": 12.18, "learning_rate": 4.461446166300451e-05, "loss": 1.0697, "step": 140630 }, { "epoch": 12.18, "learning_rate": 4.4612293280906667e-05, "loss": 0.8804, "step": 140640 }, { "epoch": 12.18, "learning_rate": 4.461012489880883e-05, "loss": 1.0149, "step": 140650 }, { "epoch": 12.18, "learning_rate": 4.4607956516711e-05, "loss": 1.0654, "step": 140660 }, { "epoch": 12.18, "learning_rate": 4.460578813461316e-05, "loss": 0.9988, "step": 140670 }, { "epoch": 12.18, "learning_rate": 4.460361975251532e-05, "loss": 1.1106, "step": 140680 }, { "epoch": 12.18, "learning_rate": 4.460145137041748e-05, "loss": 1.1417, "step": 140690 }, { "epoch": 12.19, "learning_rate": 4.4599282988319645e-05, "loss": 1.1584, "step": 140700 }, { "epoch": 12.19, "learning_rate": 4.4597114606221804e-05, "loss": 0.9786, "step": 140710 }, { "epoch": 12.19, "learning_rate": 4.459494622412397e-05, "loss": 1.0409, "step": 140720 }, { "epoch": 12.19, "learning_rate": 4.459277784202614e-05, "loss": 1.1664, "step": 140730 }, { "epoch": 12.19, "learning_rate": 4.4590609459928297e-05, "loss": 1.0892, "step": 140740 }, { "epoch": 12.19, "learning_rate": 4.458844107783046e-05, "loss": 1.1663, "step": 140750 }, { "epoch": 12.19, "learning_rate": 4.4586272695732616e-05, "loss": 1.1659, "step": 140760 }, { "epoch": 12.19, "learning_rate": 4.458410431363478e-05, "loss": 1.0617, "step": 140770 }, { "epoch": 12.19, "learning_rate": 4.458193593153694e-05, "loss": 1.1486, "step": 140780 }, { "epoch": 12.19, "learning_rate": 4.457976754943911e-05, "loss": 1.1954, "step": 140790 }, { "epoch": 12.19, "learning_rate": 4.4577599167341275e-05, "loss": 1.2298, "step": 140800 }, { "epoch": 12.2, "learning_rate": 4.4575430785243434e-05, "loss": 1.0008, "step": 140810 }, { "epoch": 12.2, "learning_rate": 4.45732624031456e-05, "loss": 1.003, "step": 140820 }, { "epoch": 12.2, "learning_rate": 4.457109402104776e-05, "loss": 1.0814, "step": 140830 }, { "epoch": 12.2, "learning_rate": 4.456892563894992e-05, "loss": 1.2062, "step": 140840 }, { "epoch": 12.2, "learning_rate": 4.456675725685208e-05, "loss": 1.0635, "step": 140850 }, { "epoch": 12.2, "learning_rate": 4.4564588874754246e-05, "loss": 1.0496, "step": 140860 }, { "epoch": 12.2, "learning_rate": 4.4562420492656405e-05, "loss": 1.1461, "step": 140870 }, { "epoch": 12.2, "learning_rate": 4.456025211055857e-05, "loss": 1.2452, "step": 140880 }, { "epoch": 12.2, "learning_rate": 4.455808372846074e-05, "loss": 1.0665, "step": 140890 }, { "epoch": 12.2, "learning_rate": 4.45559153463629e-05, "loss": 1.0471, "step": 140900 }, { "epoch": 12.2, "learning_rate": 4.455374696426506e-05, "loss": 1.1619, "step": 140910 }, { "epoch": 12.2, "learning_rate": 4.455157858216722e-05, "loss": 0.89, "step": 140920 }, { "epoch": 12.21, "learning_rate": 4.4549410200069383e-05, "loss": 1.158, "step": 140930 }, { "epoch": 12.21, "learning_rate": 4.454724181797154e-05, "loss": 1.0133, "step": 140940 }, { "epoch": 12.21, "learning_rate": 4.454507343587371e-05, "loss": 0.9751, "step": 140950 }, { "epoch": 12.21, "learning_rate": 4.4542905053775876e-05, "loss": 0.9243, "step": 140960 }, { "epoch": 12.21, "learning_rate": 4.4540736671678035e-05, "loss": 1.1411, "step": 140970 }, { "epoch": 12.21, "learning_rate": 4.45385682895802e-05, "loss": 1.0746, "step": 140980 }, { "epoch": 12.21, "learning_rate": 4.4536399907482355e-05, "loss": 0.8967, "step": 140990 }, { "epoch": 12.21, "learning_rate": 4.453423152538452e-05, "loss": 1.1528, "step": 141000 }, { "epoch": 12.21, "learning_rate": 4.453206314328668e-05, "loss": 1.0036, "step": 141010 }, { "epoch": 12.21, "learning_rate": 4.452989476118885e-05, "loss": 1.0267, "step": 141020 }, { "epoch": 12.21, "learning_rate": 4.4527726379091014e-05, "loss": 1.0968, "step": 141030 }, { "epoch": 12.21, "learning_rate": 4.452555799699317e-05, "loss": 1.0091, "step": 141040 }, { "epoch": 12.22, "learning_rate": 4.452338961489534e-05, "loss": 1.1014, "step": 141050 }, { "epoch": 12.22, "learning_rate": 4.45212212327975e-05, "loss": 1.0401, "step": 141060 }, { "epoch": 12.22, "learning_rate": 4.451905285069966e-05, "loss": 1.194, "step": 141070 }, { "epoch": 12.22, "learning_rate": 4.451688446860182e-05, "loss": 1.0439, "step": 141080 }, { "epoch": 12.22, "learning_rate": 4.4514716086503985e-05, "loss": 1.01, "step": 141090 }, { "epoch": 12.22, "learning_rate": 4.451254770440615e-05, "loss": 1.0137, "step": 141100 }, { "epoch": 12.22, "learning_rate": 4.451037932230831e-05, "loss": 1.0391, "step": 141110 }, { "epoch": 12.22, "learning_rate": 4.450821094021048e-05, "loss": 1.0517, "step": 141120 }, { "epoch": 12.22, "learning_rate": 4.450604255811264e-05, "loss": 1.0566, "step": 141130 }, { "epoch": 12.22, "learning_rate": 4.4503874176014796e-05, "loss": 0.9063, "step": 141140 }, { "epoch": 12.22, "learning_rate": 4.4501705793916956e-05, "loss": 1.1727, "step": 141150 }, { "epoch": 12.23, "learning_rate": 4.449953741181912e-05, "loss": 0.9778, "step": 141160 }, { "epoch": 12.23, "learning_rate": 4.449736902972129e-05, "loss": 1.0563, "step": 141170 }, { "epoch": 12.23, "learning_rate": 4.449520064762345e-05, "loss": 1.0527, "step": 141180 }, { "epoch": 12.23, "learning_rate": 4.4493032265525615e-05, "loss": 0.815, "step": 141190 }, { "epoch": 12.23, "learning_rate": 4.4490863883427774e-05, "loss": 1.1742, "step": 141200 }, { "epoch": 12.23, "learning_rate": 4.448869550132994e-05, "loss": 1.1457, "step": 141210 }, { "epoch": 12.23, "learning_rate": 4.4486527119232094e-05, "loss": 1.0294, "step": 141220 }, { "epoch": 12.23, "learning_rate": 4.448435873713426e-05, "loss": 1.0131, "step": 141230 }, { "epoch": 12.23, "learning_rate": 4.4482190355036426e-05, "loss": 1.0807, "step": 141240 }, { "epoch": 12.23, "learning_rate": 4.4480021972938586e-05, "loss": 1.0205, "step": 141250 }, { "epoch": 12.23, "learning_rate": 4.447785359084075e-05, "loss": 1.0397, "step": 141260 }, { "epoch": 12.23, "learning_rate": 4.447568520874291e-05, "loss": 0.9237, "step": 141270 }, { "epoch": 12.24, "learning_rate": 4.447351682664508e-05, "loss": 0.8522, "step": 141280 }, { "epoch": 12.24, "learning_rate": 4.4471348444547245e-05, "loss": 1.0703, "step": 141290 }, { "epoch": 12.24, "learning_rate": 4.44691800624494e-05, "loss": 1.0615, "step": 141300 }, { "epoch": 12.24, "learning_rate": 4.4467011680351564e-05, "loss": 0.8794, "step": 141310 }, { "epoch": 12.24, "learning_rate": 4.4464843298253724e-05, "loss": 1.1464, "step": 141320 }, { "epoch": 12.24, "learning_rate": 4.446267491615589e-05, "loss": 0.9509, "step": 141330 }, { "epoch": 12.24, "learning_rate": 4.446050653405805e-05, "loss": 1.0068, "step": 141340 }, { "epoch": 12.24, "learning_rate": 4.4458338151960216e-05, "loss": 1.0071, "step": 141350 }, { "epoch": 12.24, "learning_rate": 4.445616976986238e-05, "loss": 1.0975, "step": 141360 }, { "epoch": 12.24, "learning_rate": 4.4454001387764535e-05, "loss": 1.0302, "step": 141370 }, { "epoch": 12.24, "learning_rate": 4.4451833005666695e-05, "loss": 0.9531, "step": 141380 }, { "epoch": 12.25, "learning_rate": 4.444966462356886e-05, "loss": 1.156, "step": 141390 }, { "epoch": 12.25, "learning_rate": 4.444749624147103e-05, "loss": 1.0418, "step": 141400 }, { "epoch": 12.25, "learning_rate": 4.444532785937319e-05, "loss": 1.1362, "step": 141410 }, { "epoch": 12.25, "learning_rate": 4.4443159477275354e-05, "loss": 1.0033, "step": 141420 }, { "epoch": 12.25, "learning_rate": 4.444099109517751e-05, "loss": 1.0322, "step": 141430 }, { "epoch": 12.25, "learning_rate": 4.443882271307968e-05, "loss": 0.8946, "step": 141440 }, { "epoch": 12.25, "learning_rate": 4.443665433098183e-05, "loss": 1.1012, "step": 141450 }, { "epoch": 12.25, "learning_rate": 4.4434485948884e-05, "loss": 0.8837, "step": 141460 }, { "epoch": 12.25, "learning_rate": 4.4432317566786165e-05, "loss": 1.1694, "step": 141470 }, { "epoch": 12.25, "learning_rate": 4.4430149184688325e-05, "loss": 0.8968, "step": 141480 }, { "epoch": 12.25, "learning_rate": 4.442798080259049e-05, "loss": 0.9179, "step": 141490 }, { "epoch": 12.25, "learning_rate": 4.442581242049265e-05, "loss": 1.0554, "step": 141500 }, { "epoch": 12.26, "learning_rate": 4.442364403839482e-05, "loss": 0.9749, "step": 141510 }, { "epoch": 12.26, "learning_rate": 4.4421475656296984e-05, "loss": 0.867, "step": 141520 }, { "epoch": 12.26, "learning_rate": 4.4419307274199137e-05, "loss": 1.2143, "step": 141530 }, { "epoch": 12.26, "learning_rate": 4.44171388921013e-05, "loss": 0.9319, "step": 141540 }, { "epoch": 12.26, "learning_rate": 4.441497051000346e-05, "loss": 1.0686, "step": 141550 }, { "epoch": 12.26, "learning_rate": 4.441280212790563e-05, "loss": 0.8493, "step": 141560 }, { "epoch": 12.26, "learning_rate": 4.441063374580779e-05, "loss": 0.9667, "step": 141570 }, { "epoch": 12.26, "learning_rate": 4.4408465363709955e-05, "loss": 0.9201, "step": 141580 }, { "epoch": 12.26, "learning_rate": 4.440629698161212e-05, "loss": 0.8892, "step": 141590 }, { "epoch": 12.26, "learning_rate": 4.440412859951428e-05, "loss": 1.1002, "step": 141600 }, { "epoch": 12.26, "learning_rate": 4.440196021741644e-05, "loss": 1.2069, "step": 141610 }, { "epoch": 12.27, "learning_rate": 4.43997918353186e-05, "loss": 1.0194, "step": 141620 }, { "epoch": 12.27, "learning_rate": 4.439762345322077e-05, "loss": 0.9042, "step": 141630 }, { "epoch": 12.27, "learning_rate": 4.4395455071122926e-05, "loss": 1.0958, "step": 141640 }, { "epoch": 12.27, "learning_rate": 4.439328668902509e-05, "loss": 1.1548, "step": 141650 }, { "epoch": 12.27, "learning_rate": 4.439111830692726e-05, "loss": 0.9843, "step": 141660 }, { "epoch": 12.27, "learning_rate": 4.438894992482942e-05, "loss": 1.0859, "step": 141670 }, { "epoch": 12.27, "learning_rate": 4.438678154273158e-05, "loss": 1.0657, "step": 141680 }, { "epoch": 12.27, "learning_rate": 4.438461316063374e-05, "loss": 0.773, "step": 141690 }, { "epoch": 12.27, "learning_rate": 4.4382444778535904e-05, "loss": 1.0085, "step": 141700 }, { "epoch": 12.27, "learning_rate": 4.4380276396438064e-05, "loss": 0.9631, "step": 141710 }, { "epoch": 12.27, "learning_rate": 4.437810801434023e-05, "loss": 0.8812, "step": 141720 }, { "epoch": 12.27, "learning_rate": 4.43759396322424e-05, "loss": 1.1252, "step": 141730 }, { "epoch": 12.28, "learning_rate": 4.4373771250144556e-05, "loss": 1.1273, "step": 141740 }, { "epoch": 12.28, "learning_rate": 4.437160286804672e-05, "loss": 0.9919, "step": 141750 }, { "epoch": 12.28, "learning_rate": 4.4369434485948876e-05, "loss": 1.0131, "step": 141760 }, { "epoch": 12.28, "learning_rate": 4.436726610385104e-05, "loss": 0.9295, "step": 141770 }, { "epoch": 12.28, "learning_rate": 4.43650977217532e-05, "loss": 1.0197, "step": 141780 }, { "epoch": 12.28, "learning_rate": 4.436292933965537e-05, "loss": 0.9721, "step": 141790 }, { "epoch": 12.28, "learning_rate": 4.4360760957557534e-05, "loss": 1.0911, "step": 141800 }, { "epoch": 12.28, "learning_rate": 4.4358592575459694e-05, "loss": 1.1636, "step": 141810 }, { "epoch": 12.28, "learning_rate": 4.435642419336186e-05, "loss": 0.9775, "step": 141820 }, { "epoch": 12.28, "learning_rate": 4.435425581126402e-05, "loss": 0.9789, "step": 141830 }, { "epoch": 12.28, "learning_rate": 4.435208742916618e-05, "loss": 0.9969, "step": 141840 }, { "epoch": 12.29, "learning_rate": 4.434991904706834e-05, "loss": 0.9114, "step": 141850 }, { "epoch": 12.29, "learning_rate": 4.4347750664970506e-05, "loss": 0.9852, "step": 141860 }, { "epoch": 12.29, "learning_rate": 4.434558228287267e-05, "loss": 1.1056, "step": 141870 }, { "epoch": 12.29, "learning_rate": 4.434341390077483e-05, "loss": 1.0432, "step": 141880 }, { "epoch": 12.29, "learning_rate": 4.4341245518677e-05, "loss": 0.9364, "step": 141890 }, { "epoch": 12.29, "learning_rate": 4.433907713657916e-05, "loss": 1.0894, "step": 141900 }, { "epoch": 12.29, "learning_rate": 4.433690875448132e-05, "loss": 0.9425, "step": 141910 }, { "epoch": 12.29, "learning_rate": 4.433474037238348e-05, "loss": 1.1469, "step": 141920 }, { "epoch": 12.29, "learning_rate": 4.433257199028564e-05, "loss": 0.9538, "step": 141930 }, { "epoch": 12.29, "learning_rate": 4.43304036081878e-05, "loss": 0.982, "step": 141940 }, { "epoch": 12.29, "learning_rate": 4.432823522608997e-05, "loss": 1.1712, "step": 141950 }, { "epoch": 12.29, "learning_rate": 4.4326066843992136e-05, "loss": 0.9319, "step": 141960 }, { "epoch": 12.3, "learning_rate": 4.4323898461894295e-05, "loss": 1.141, "step": 141970 }, { "epoch": 12.3, "learning_rate": 4.432173007979646e-05, "loss": 1.0178, "step": 141980 }, { "epoch": 12.3, "learning_rate": 4.4319561697698614e-05, "loss": 1.0407, "step": 141990 }, { "epoch": 12.3, "learning_rate": 4.431739331560078e-05, "loss": 0.8242, "step": 142000 }, { "epoch": 12.3, "learning_rate": 4.431522493350294e-05, "loss": 1.0591, "step": 142010 }, { "epoch": 12.3, "learning_rate": 4.431305655140511e-05, "loss": 1.0759, "step": 142020 }, { "epoch": 12.3, "learning_rate": 4.431088816930727e-05, "loss": 1.1452, "step": 142030 }, { "epoch": 12.3, "learning_rate": 4.430871978720943e-05, "loss": 1.2003, "step": 142040 }, { "epoch": 12.3, "learning_rate": 4.43065514051116e-05, "loss": 1.0908, "step": 142050 }, { "epoch": 12.3, "learning_rate": 4.430438302301376e-05, "loss": 0.9366, "step": 142060 }, { "epoch": 12.3, "learning_rate": 4.430221464091592e-05, "loss": 0.8871, "step": 142070 }, { "epoch": 12.31, "learning_rate": 4.430004625881808e-05, "loss": 1.2039, "step": 142080 }, { "epoch": 12.31, "learning_rate": 4.4297877876720244e-05, "loss": 1.2356, "step": 142090 }, { "epoch": 12.31, "learning_rate": 4.429570949462241e-05, "loss": 1.1343, "step": 142100 }, { "epoch": 12.31, "learning_rate": 4.429354111252457e-05, "loss": 1.0805, "step": 142110 }, { "epoch": 12.31, "learning_rate": 4.429137273042674e-05, "loss": 1.2366, "step": 142120 }, { "epoch": 12.31, "learning_rate": 4.4289204348328896e-05, "loss": 1.1802, "step": 142130 }, { "epoch": 12.31, "learning_rate": 4.4287035966231056e-05, "loss": 1.0659, "step": 142140 }, { "epoch": 12.31, "learning_rate": 4.4284867584133216e-05, "loss": 1.0274, "step": 142150 }, { "epoch": 12.31, "learning_rate": 4.428269920203538e-05, "loss": 1.0344, "step": 142160 }, { "epoch": 12.31, "learning_rate": 4.428053081993755e-05, "loss": 0.9219, "step": 142170 }, { "epoch": 12.31, "learning_rate": 4.427836243783971e-05, "loss": 1.0806, "step": 142180 }, { "epoch": 12.31, "learning_rate": 4.4276194055741874e-05, "loss": 1.0674, "step": 142190 }, { "epoch": 12.32, "learning_rate": 4.4274025673644034e-05, "loss": 1.0551, "step": 142200 }, { "epoch": 12.32, "learning_rate": 4.42718572915462e-05, "loss": 1.1139, "step": 142210 }, { "epoch": 12.32, "learning_rate": 4.426968890944835e-05, "loss": 1.0644, "step": 142220 }, { "epoch": 12.32, "learning_rate": 4.426752052735052e-05, "loss": 1.0191, "step": 142230 }, { "epoch": 12.32, "learning_rate": 4.4265352145252686e-05, "loss": 1.1989, "step": 142240 }, { "epoch": 12.32, "learning_rate": 4.4263183763154846e-05, "loss": 1.0989, "step": 142250 }, { "epoch": 12.32, "learning_rate": 4.426101538105701e-05, "loss": 1.256, "step": 142260 }, { "epoch": 12.32, "learning_rate": 4.425884699895917e-05, "loss": 0.9405, "step": 142270 }, { "epoch": 12.32, "learning_rate": 4.425667861686134e-05, "loss": 1.0627, "step": 142280 }, { "epoch": 12.32, "learning_rate": 4.4254510234763505e-05, "loss": 1.0838, "step": 142290 }, { "epoch": 12.32, "learning_rate": 4.425234185266566e-05, "loss": 1.1025, "step": 142300 }, { "epoch": 12.32, "learning_rate": 4.4250173470567824e-05, "loss": 1.0358, "step": 142310 }, { "epoch": 12.33, "learning_rate": 4.424800508846998e-05, "loss": 1.2449, "step": 142320 }, { "epoch": 12.33, "learning_rate": 4.424583670637215e-05, "loss": 1.0606, "step": 142330 }, { "epoch": 12.33, "learning_rate": 4.424366832427431e-05, "loss": 0.9844, "step": 142340 }, { "epoch": 12.33, "learning_rate": 4.4241499942176476e-05, "loss": 1.1057, "step": 142350 }, { "epoch": 12.33, "learning_rate": 4.423933156007864e-05, "loss": 0.9808, "step": 142360 }, { "epoch": 12.33, "learning_rate": 4.42371631779808e-05, "loss": 0.9794, "step": 142370 }, { "epoch": 12.33, "learning_rate": 4.423499479588296e-05, "loss": 1.0752, "step": 142380 }, { "epoch": 12.33, "learning_rate": 4.423282641378512e-05, "loss": 0.9251, "step": 142390 }, { "epoch": 12.33, "learning_rate": 4.423065803168729e-05, "loss": 1.0156, "step": 142400 }, { "epoch": 12.33, "learning_rate": 4.422848964958945e-05, "loss": 1.1605, "step": 142410 }, { "epoch": 12.33, "learning_rate": 4.4226321267491613e-05, "loss": 1.0252, "step": 142420 }, { "epoch": 12.34, "learning_rate": 4.422415288539378e-05, "loss": 1.0087, "step": 142430 }, { "epoch": 12.34, "learning_rate": 4.422198450329594e-05, "loss": 1.0684, "step": 142440 }, { "epoch": 12.34, "learning_rate": 4.421981612119809e-05, "loss": 1.081, "step": 142450 }, { "epoch": 12.34, "learning_rate": 4.421764773910026e-05, "loss": 1.01, "step": 142460 }, { "epoch": 12.34, "learning_rate": 4.4215479357002425e-05, "loss": 1.125, "step": 142470 }, { "epoch": 12.34, "learning_rate": 4.4213310974904585e-05, "loss": 1.0668, "step": 142480 }, { "epoch": 12.34, "learning_rate": 4.421114259280675e-05, "loss": 1.1275, "step": 142490 }, { "epoch": 12.34, "learning_rate": 4.420897421070891e-05, "loss": 1.0491, "step": 142500 }, { "epoch": 12.34, "learning_rate": 4.420680582861108e-05, "loss": 1.1439, "step": 142510 }, { "epoch": 12.34, "learning_rate": 4.4204637446513243e-05, "loss": 1.1443, "step": 142520 }, { "epoch": 12.34, "learning_rate": 4.4202469064415396e-05, "loss": 1.1322, "step": 142530 }, { "epoch": 12.34, "learning_rate": 4.420030068231756e-05, "loss": 0.9051, "step": 142540 }, { "epoch": 12.35, "learning_rate": 4.419813230021972e-05, "loss": 0.9017, "step": 142550 }, { "epoch": 12.35, "learning_rate": 4.419596391812189e-05, "loss": 1.0289, "step": 142560 }, { "epoch": 12.35, "learning_rate": 4.419379553602405e-05, "loss": 1.0516, "step": 142570 }, { "epoch": 12.35, "learning_rate": 4.4191627153926215e-05, "loss": 0.9531, "step": 142580 }, { "epoch": 12.35, "learning_rate": 4.418945877182838e-05, "loss": 1.1055, "step": 142590 }, { "epoch": 12.35, "learning_rate": 4.418729038973054e-05, "loss": 0.9397, "step": 142600 }, { "epoch": 12.35, "learning_rate": 4.41851220076327e-05, "loss": 1.1005, "step": 142610 }, { "epoch": 12.35, "learning_rate": 4.418295362553486e-05, "loss": 1.1054, "step": 142620 }, { "epoch": 12.35, "learning_rate": 4.4180785243437026e-05, "loss": 0.8822, "step": 142630 }, { "epoch": 12.35, "learning_rate": 4.4178616861339186e-05, "loss": 0.9551, "step": 142640 }, { "epoch": 12.35, "learning_rate": 4.417644847924135e-05, "loss": 1.0051, "step": 142650 }, { "epoch": 12.36, "learning_rate": 4.417428009714352e-05, "loss": 1.0747, "step": 142660 }, { "epoch": 12.36, "learning_rate": 4.417211171504568e-05, "loss": 1.1093, "step": 142670 }, { "epoch": 12.36, "learning_rate": 4.416994333294784e-05, "loss": 0.9782, "step": 142680 }, { "epoch": 12.36, "learning_rate": 4.416777495085e-05, "loss": 1.176, "step": 142690 }, { "epoch": 12.36, "learning_rate": 4.4165606568752164e-05, "loss": 1.0764, "step": 142700 }, { "epoch": 12.36, "learning_rate": 4.4163438186654324e-05, "loss": 1.0021, "step": 142710 }, { "epoch": 12.36, "learning_rate": 4.416126980455649e-05, "loss": 1.0668, "step": 142720 }, { "epoch": 12.36, "learning_rate": 4.4159101422458656e-05, "loss": 0.8315, "step": 142730 }, { "epoch": 12.36, "learning_rate": 4.4156933040360816e-05, "loss": 1.0079, "step": 142740 }, { "epoch": 12.36, "learning_rate": 4.415476465826298e-05, "loss": 1.1836, "step": 142750 }, { "epoch": 12.36, "learning_rate": 4.4152596276165135e-05, "loss": 1.0043, "step": 142760 }, { "epoch": 12.36, "learning_rate": 4.41504278940673e-05, "loss": 1.0162, "step": 142770 }, { "epoch": 12.37, "learning_rate": 4.414825951196946e-05, "loss": 0.9962, "step": 142780 }, { "epoch": 12.37, "learning_rate": 4.414609112987163e-05, "loss": 0.9584, "step": 142790 }, { "epoch": 12.37, "learning_rate": 4.4143922747773794e-05, "loss": 1.0037, "step": 142800 }, { "epoch": 12.37, "learning_rate": 4.4141754365675954e-05, "loss": 1.0243, "step": 142810 }, { "epoch": 12.37, "learning_rate": 4.413958598357812e-05, "loss": 1.069, "step": 142820 }, { "epoch": 12.37, "learning_rate": 4.413741760148028e-05, "loss": 0.8647, "step": 142830 }, { "epoch": 12.37, "learning_rate": 4.413524921938244e-05, "loss": 1.1875, "step": 142840 }, { "epoch": 12.37, "learning_rate": 4.41330808372846e-05, "loss": 1.1719, "step": 142850 }, { "epoch": 12.37, "learning_rate": 4.4130912455186765e-05, "loss": 1.0343, "step": 142860 }, { "epoch": 12.37, "learning_rate": 4.412874407308893e-05, "loss": 0.9096, "step": 142870 }, { "epoch": 12.37, "learning_rate": 4.412657569099109e-05, "loss": 1.1144, "step": 142880 }, { "epoch": 12.38, "learning_rate": 4.412440730889326e-05, "loss": 0.9017, "step": 142890 }, { "epoch": 12.38, "learning_rate": 4.412223892679542e-05, "loss": 1.027, "step": 142900 }, { "epoch": 12.38, "learning_rate": 4.412007054469758e-05, "loss": 0.9415, "step": 142910 }, { "epoch": 12.38, "learning_rate": 4.4117902162599736e-05, "loss": 1.0766, "step": 142920 }, { "epoch": 12.38, "learning_rate": 4.41157337805019e-05, "loss": 1.064, "step": 142930 }, { "epoch": 12.38, "learning_rate": 4.411356539840407e-05, "loss": 1.0592, "step": 142940 }, { "epoch": 12.38, "learning_rate": 4.411139701630623e-05, "loss": 0.9941, "step": 142950 }, { "epoch": 12.38, "learning_rate": 4.4109228634208395e-05, "loss": 0.9901, "step": 142960 }, { "epoch": 12.38, "learning_rate": 4.4107060252110555e-05, "loss": 1.1098, "step": 142970 }, { "epoch": 12.38, "learning_rate": 4.410489187001272e-05, "loss": 1.0423, "step": 142980 }, { "epoch": 12.38, "learning_rate": 4.4102723487914874e-05, "loss": 0.8552, "step": 142990 }, { "epoch": 12.38, "learning_rate": 4.410055510581704e-05, "loss": 1.0667, "step": 143000 }, { "epoch": 12.39, "learning_rate": 4.40983867237192e-05, "loss": 1.0977, "step": 143010 }, { "epoch": 12.39, "learning_rate": 4.4096218341621367e-05, "loss": 1.0659, "step": 143020 }, { "epoch": 12.39, "learning_rate": 4.409404995952353e-05, "loss": 1.012, "step": 143030 }, { "epoch": 12.39, "learning_rate": 4.409188157742569e-05, "loss": 1.1055, "step": 143040 }, { "epoch": 12.39, "learning_rate": 4.408971319532786e-05, "loss": 1.1695, "step": 143050 }, { "epoch": 12.39, "learning_rate": 4.408754481323002e-05, "loss": 1.035, "step": 143060 }, { "epoch": 12.39, "learning_rate": 4.408537643113218e-05, "loss": 0.9781, "step": 143070 }, { "epoch": 12.39, "learning_rate": 4.408320804903434e-05, "loss": 1.1544, "step": 143080 }, { "epoch": 12.39, "learning_rate": 4.4081039666936504e-05, "loss": 1.0344, "step": 143090 }, { "epoch": 12.39, "learning_rate": 4.407887128483867e-05, "loss": 1.1945, "step": 143100 }, { "epoch": 12.39, "learning_rate": 4.407670290274083e-05, "loss": 1.1613, "step": 143110 }, { "epoch": 12.4, "learning_rate": 4.4074534520642997e-05, "loss": 1.1805, "step": 143120 }, { "epoch": 12.4, "learning_rate": 4.4072366138545156e-05, "loss": 1.1467, "step": 143130 }, { "epoch": 12.4, "learning_rate": 4.407019775644732e-05, "loss": 0.9171, "step": 143140 }, { "epoch": 12.4, "learning_rate": 4.4068029374349475e-05, "loss": 0.9552, "step": 143150 }, { "epoch": 12.4, "learning_rate": 4.406586099225164e-05, "loss": 0.9372, "step": 143160 }, { "epoch": 12.4, "learning_rate": 4.406369261015381e-05, "loss": 0.9627, "step": 143170 }, { "epoch": 12.4, "learning_rate": 4.406152422805597e-05, "loss": 1.0667, "step": 143180 }, { "epoch": 12.4, "learning_rate": 4.4059355845958134e-05, "loss": 0.9309, "step": 143190 }, { "epoch": 12.4, "learning_rate": 4.4057187463860294e-05, "loss": 1.0403, "step": 143200 }, { "epoch": 12.4, "learning_rate": 4.405501908176246e-05, "loss": 1.1019, "step": 143210 }, { "epoch": 12.4, "learning_rate": 4.405285069966461e-05, "loss": 0.9146, "step": 143220 }, { "epoch": 12.4, "learning_rate": 4.405068231756678e-05, "loss": 0.9813, "step": 143230 }, { "epoch": 12.41, "learning_rate": 4.4048513935468946e-05, "loss": 1.1831, "step": 143240 }, { "epoch": 12.41, "learning_rate": 4.4046345553371105e-05, "loss": 1.0231, "step": 143250 }, { "epoch": 12.41, "learning_rate": 4.404417717127327e-05, "loss": 1.2609, "step": 143260 }, { "epoch": 12.41, "learning_rate": 4.404200878917543e-05, "loss": 0.9789, "step": 143270 }, { "epoch": 12.41, "learning_rate": 4.40398404070776e-05, "loss": 1.0727, "step": 143280 }, { "epoch": 12.41, "learning_rate": 4.4037672024979764e-05, "loss": 0.9599, "step": 143290 }, { "epoch": 12.41, "learning_rate": 4.403550364288192e-05, "loss": 0.9983, "step": 143300 }, { "epoch": 12.41, "learning_rate": 4.4033335260784083e-05, "loss": 1.0826, "step": 143310 }, { "epoch": 12.41, "learning_rate": 4.403116687868624e-05, "loss": 1.2646, "step": 143320 }, { "epoch": 12.41, "learning_rate": 4.402899849658841e-05, "loss": 0.7731, "step": 143330 }, { "epoch": 12.41, "learning_rate": 4.402683011449057e-05, "loss": 0.9082, "step": 143340 }, { "epoch": 12.42, "learning_rate": 4.4024661732392735e-05, "loss": 1.0869, "step": 143350 }, { "epoch": 12.42, "learning_rate": 4.40224933502949e-05, "loss": 1.2144, "step": 143360 }, { "epoch": 12.42, "learning_rate": 4.402032496819706e-05, "loss": 0.9732, "step": 143370 }, { "epoch": 12.42, "learning_rate": 4.401815658609922e-05, "loss": 0.852, "step": 143380 }, { "epoch": 12.42, "learning_rate": 4.401598820400138e-05, "loss": 1.1271, "step": 143390 }, { "epoch": 12.42, "learning_rate": 4.401381982190355e-05, "loss": 1.0178, "step": 143400 }, { "epoch": 12.42, "learning_rate": 4.401165143980571e-05, "loss": 1.0474, "step": 143410 }, { "epoch": 12.42, "learning_rate": 4.400948305770787e-05, "loss": 1.1348, "step": 143420 }, { "epoch": 12.42, "learning_rate": 4.400731467561004e-05, "loss": 1.206, "step": 143430 }, { "epoch": 12.42, "learning_rate": 4.40051462935122e-05, "loss": 1.0824, "step": 143440 }, { "epoch": 12.42, "learning_rate": 4.400297791141436e-05, "loss": 1.1108, "step": 143450 }, { "epoch": 12.42, "learning_rate": 4.400080952931652e-05, "loss": 0.9265, "step": 143460 }, { "epoch": 12.43, "learning_rate": 4.3998641147218685e-05, "loss": 1.0939, "step": 143470 }, { "epoch": 12.43, "learning_rate": 4.3996472765120844e-05, "loss": 0.9761, "step": 143480 }, { "epoch": 12.43, "learning_rate": 4.399430438302301e-05, "loss": 1.091, "step": 143490 }, { "epoch": 12.43, "learning_rate": 4.399213600092518e-05, "loss": 1.0124, "step": 143500 }, { "epoch": 12.43, "learning_rate": 4.398996761882734e-05, "loss": 1.1258, "step": 143510 }, { "epoch": 12.43, "learning_rate": 4.39877992367295e-05, "loss": 1.1051, "step": 143520 }, { "epoch": 12.43, "learning_rate": 4.3985630854631656e-05, "loss": 0.9924, "step": 143530 }, { "epoch": 12.43, "learning_rate": 4.398346247253382e-05, "loss": 0.9694, "step": 143540 }, { "epoch": 12.43, "learning_rate": 4.398129409043598e-05, "loss": 1.0965, "step": 143550 }, { "epoch": 12.43, "learning_rate": 4.397912570833815e-05, "loss": 0.8911, "step": 143560 }, { "epoch": 12.43, "learning_rate": 4.397695732624031e-05, "loss": 0.9353, "step": 143570 }, { "epoch": 12.43, "learning_rate": 4.3974788944142474e-05, "loss": 1.1539, "step": 143580 }, { "epoch": 12.44, "learning_rate": 4.397262056204464e-05, "loss": 1.0101, "step": 143590 }, { "epoch": 12.44, "learning_rate": 4.39704521799468e-05, "loss": 1.026, "step": 143600 }, { "epoch": 12.44, "learning_rate": 4.396828379784896e-05, "loss": 1.1279, "step": 143610 }, { "epoch": 12.44, "learning_rate": 4.396611541575112e-05, "loss": 1.2791, "step": 143620 }, { "epoch": 12.44, "learning_rate": 4.3963947033653286e-05, "loss": 1.2085, "step": 143630 }, { "epoch": 12.44, "learning_rate": 4.3961778651555446e-05, "loss": 0.9029, "step": 143640 }, { "epoch": 12.44, "learning_rate": 4.395961026945761e-05, "loss": 1.2064, "step": 143650 }, { "epoch": 12.44, "learning_rate": 4.395744188735978e-05, "loss": 1.2049, "step": 143660 }, { "epoch": 12.44, "learning_rate": 4.395527350526194e-05, "loss": 1.0925, "step": 143670 }, { "epoch": 12.44, "learning_rate": 4.39531051231641e-05, "loss": 1.067, "step": 143680 }, { "epoch": 12.44, "learning_rate": 4.395093674106626e-05, "loss": 1.001, "step": 143690 }, { "epoch": 12.45, "learning_rate": 4.3948768358968424e-05, "loss": 1.0802, "step": 143700 }, { "epoch": 12.45, "learning_rate": 4.394659997687058e-05, "loss": 1.261, "step": 143710 }, { "epoch": 12.45, "learning_rate": 4.394443159477275e-05, "loss": 1.1994, "step": 143720 }, { "epoch": 12.45, "learning_rate": 4.3942263212674916e-05, "loss": 0.9538, "step": 143730 }, { "epoch": 12.45, "learning_rate": 4.3940094830577076e-05, "loss": 1.064, "step": 143740 }, { "epoch": 12.45, "learning_rate": 4.393792644847924e-05, "loss": 1.1105, "step": 143750 }, { "epoch": 12.45, "learning_rate": 4.3935758066381395e-05, "loss": 1.2856, "step": 143760 }, { "epoch": 12.45, "learning_rate": 4.393358968428356e-05, "loss": 1.1198, "step": 143770 }, { "epoch": 12.45, "learning_rate": 4.393142130218572e-05, "loss": 0.8673, "step": 143780 }, { "epoch": 12.45, "learning_rate": 4.392925292008789e-05, "loss": 1.1624, "step": 143790 }, { "epoch": 12.45, "learning_rate": 4.3927084537990054e-05, "loss": 1.1505, "step": 143800 }, { "epoch": 12.45, "learning_rate": 4.392491615589221e-05, "loss": 0.8999, "step": 143810 }, { "epoch": 12.46, "learning_rate": 4.392274777379438e-05, "loss": 1.1255, "step": 143820 }, { "epoch": 12.46, "learning_rate": 4.392057939169654e-05, "loss": 0.9684, "step": 143830 }, { "epoch": 12.46, "learning_rate": 4.39184110095987e-05, "loss": 1.3572, "step": 143840 }, { "epoch": 12.46, "learning_rate": 4.391624262750086e-05, "loss": 1.3151, "step": 143850 }, { "epoch": 12.46, "learning_rate": 4.3914074245403025e-05, "loss": 0.9552, "step": 143860 }, { "epoch": 12.46, "learning_rate": 4.391190586330519e-05, "loss": 1.0125, "step": 143870 }, { "epoch": 12.46, "learning_rate": 4.390973748120735e-05, "loss": 0.9983, "step": 143880 }, { "epoch": 12.46, "learning_rate": 4.390756909910952e-05, "loss": 0.8895, "step": 143890 }, { "epoch": 12.46, "learning_rate": 4.390540071701168e-05, "loss": 0.9302, "step": 143900 }, { "epoch": 12.46, "learning_rate": 4.390323233491384e-05, "loss": 1.1814, "step": 143910 }, { "epoch": 12.46, "learning_rate": 4.3901063952815996e-05, "loss": 0.9296, "step": 143920 }, { "epoch": 12.47, "learning_rate": 4.389889557071816e-05, "loss": 1.2035, "step": 143930 }, { "epoch": 12.47, "learning_rate": 4.389672718862033e-05, "loss": 0.9998, "step": 143940 }, { "epoch": 12.47, "learning_rate": 4.389455880652249e-05, "loss": 1.0308, "step": 143950 }, { "epoch": 12.47, "learning_rate": 4.3892390424424655e-05, "loss": 1.0123, "step": 143960 }, { "epoch": 12.47, "learning_rate": 4.3890222042326815e-05, "loss": 1.2087, "step": 143970 }, { "epoch": 12.47, "learning_rate": 4.388805366022898e-05, "loss": 1.2225, "step": 143980 }, { "epoch": 12.47, "learning_rate": 4.3885885278131134e-05, "loss": 0.8563, "step": 143990 }, { "epoch": 12.47, "learning_rate": 4.38837168960333e-05, "loss": 1.0788, "step": 144000 }, { "epoch": 12.47, "learning_rate": 4.3881548513935467e-05, "loss": 1.1405, "step": 144010 }, { "epoch": 12.47, "learning_rate": 4.3879380131837626e-05, "loss": 1.0771, "step": 144020 }, { "epoch": 12.47, "learning_rate": 4.387721174973979e-05, "loss": 1.1025, "step": 144030 }, { "epoch": 12.47, "learning_rate": 4.387504336764195e-05, "loss": 1.1534, "step": 144040 }, { "epoch": 12.48, "learning_rate": 4.387287498554412e-05, "loss": 0.8952, "step": 144050 }, { "epoch": 12.48, "learning_rate": 4.3870706603446285e-05, "loss": 1.1218, "step": 144060 }, { "epoch": 12.48, "learning_rate": 4.386853822134844e-05, "loss": 1.1317, "step": 144070 }, { "epoch": 12.48, "learning_rate": 4.38663698392506e-05, "loss": 0.9766, "step": 144080 }, { "epoch": 12.48, "learning_rate": 4.3864201457152764e-05, "loss": 1.0514, "step": 144090 }, { "epoch": 12.48, "learning_rate": 4.386203307505493e-05, "loss": 1.1563, "step": 144100 }, { "epoch": 12.48, "learning_rate": 4.385986469295709e-05, "loss": 0.9944, "step": 144110 }, { "epoch": 12.48, "learning_rate": 4.3857696310859256e-05, "loss": 0.9305, "step": 144120 }, { "epoch": 12.48, "learning_rate": 4.3855527928761416e-05, "loss": 1.179, "step": 144130 }, { "epoch": 12.48, "learning_rate": 4.385335954666358e-05, "loss": 0.9845, "step": 144140 }, { "epoch": 12.48, "learning_rate": 4.3851191164565735e-05, "loss": 0.959, "step": 144150 }, { "epoch": 12.49, "learning_rate": 4.38490227824679e-05, "loss": 1.1647, "step": 144160 }, { "epoch": 12.49, "learning_rate": 4.384685440037007e-05, "loss": 1.1608, "step": 144170 }, { "epoch": 12.49, "learning_rate": 4.384468601827223e-05, "loss": 1.233, "step": 144180 }, { "epoch": 12.49, "learning_rate": 4.3842517636174394e-05, "loss": 1.0631, "step": 144190 }, { "epoch": 12.49, "learning_rate": 4.3840349254076553e-05, "loss": 1.2475, "step": 144200 }, { "epoch": 12.49, "learning_rate": 4.383818087197872e-05, "loss": 1.0125, "step": 144210 }, { "epoch": 12.49, "learning_rate": 4.383601248988087e-05, "loss": 1.1251, "step": 144220 }, { "epoch": 12.49, "learning_rate": 4.383384410778304e-05, "loss": 1.1331, "step": 144230 }, { "epoch": 12.49, "learning_rate": 4.3831675725685206e-05, "loss": 1.0746, "step": 144240 }, { "epoch": 12.49, "learning_rate": 4.3829507343587365e-05, "loss": 0.9628, "step": 144250 }, { "epoch": 12.49, "learning_rate": 4.382733896148953e-05, "loss": 1.0549, "step": 144260 }, { "epoch": 12.49, "learning_rate": 4.382517057939169e-05, "loss": 1.0178, "step": 144270 }, { "epoch": 12.5, "learning_rate": 4.382300219729386e-05, "loss": 0.9424, "step": 144280 }, { "epoch": 12.5, "learning_rate": 4.3820833815196024e-05, "loss": 1.1402, "step": 144290 }, { "epoch": 12.5, "learning_rate": 4.381866543309818e-05, "loss": 1.0254, "step": 144300 }, { "epoch": 12.5, "learning_rate": 4.381649705100034e-05, "loss": 0.9548, "step": 144310 }, { "epoch": 12.5, "learning_rate": 4.38143286689025e-05, "loss": 1.0738, "step": 144320 }, { "epoch": 12.5, "learning_rate": 4.381216028680467e-05, "loss": 1.2726, "step": 144330 }, { "epoch": 12.5, "learning_rate": 4.380999190470683e-05, "loss": 1.0365, "step": 144340 }, { "epoch": 12.5, "learning_rate": 4.3807823522608995e-05, "loss": 1.1733, "step": 144350 }, { "epoch": 12.5, "learning_rate": 4.380565514051116e-05, "loss": 0.8925, "step": 144360 }, { "epoch": 12.5, "learning_rate": 4.380348675841332e-05, "loss": 1.0561, "step": 144370 }, { "epoch": 12.5, "learning_rate": 4.380131837631548e-05, "loss": 1.0319, "step": 144380 }, { "epoch": 12.51, "learning_rate": 4.379914999421764e-05, "loss": 1.0685, "step": 144390 }, { "epoch": 12.51, "learning_rate": 4.379698161211981e-05, "loss": 1.1933, "step": 144400 }, { "epoch": 12.51, "learning_rate": 4.3794813230021966e-05, "loss": 1.1601, "step": 144410 }, { "epoch": 12.51, "learning_rate": 4.379264484792413e-05, "loss": 0.9774, "step": 144420 }, { "epoch": 12.51, "learning_rate": 4.37904764658263e-05, "loss": 1.0881, "step": 144430 }, { "epoch": 12.51, "learning_rate": 4.378830808372846e-05, "loss": 1.1662, "step": 144440 }, { "epoch": 12.51, "learning_rate": 4.378613970163062e-05, "loss": 1.0902, "step": 144450 }, { "epoch": 12.51, "learning_rate": 4.378397131953278e-05, "loss": 1.0076, "step": 144460 }, { "epoch": 12.51, "learning_rate": 4.3781802937434944e-05, "loss": 0.9062, "step": 144470 }, { "epoch": 12.51, "learning_rate": 4.3779634555337104e-05, "loss": 1.0834, "step": 144480 }, { "epoch": 12.51, "learning_rate": 4.377746617323927e-05, "loss": 1.0455, "step": 144490 }, { "epoch": 12.51, "learning_rate": 4.377529779114144e-05, "loss": 0.9915, "step": 144500 }, { "epoch": 12.52, "learning_rate": 4.3773129409043596e-05, "loss": 0.9397, "step": 144510 }, { "epoch": 12.52, "learning_rate": 4.377096102694576e-05, "loss": 1.0801, "step": 144520 }, { "epoch": 12.52, "learning_rate": 4.3768792644847916e-05, "loss": 1.1437, "step": 144530 }, { "epoch": 12.52, "learning_rate": 4.376662426275008e-05, "loss": 1.0261, "step": 144540 }, { "epoch": 12.52, "learning_rate": 4.376445588065224e-05, "loss": 0.956, "step": 144550 }, { "epoch": 12.52, "learning_rate": 4.376228749855441e-05, "loss": 1.0296, "step": 144560 }, { "epoch": 12.52, "learning_rate": 4.3760119116456574e-05, "loss": 0.8824, "step": 144570 }, { "epoch": 12.52, "learning_rate": 4.3757950734358734e-05, "loss": 1.0208, "step": 144580 }, { "epoch": 12.52, "learning_rate": 4.37557823522609e-05, "loss": 1.1833, "step": 144590 }, { "epoch": 12.52, "learning_rate": 4.375361397016306e-05, "loss": 1.3245, "step": 144600 }, { "epoch": 12.52, "learning_rate": 4.375144558806522e-05, "loss": 1.1084, "step": 144610 }, { "epoch": 12.53, "learning_rate": 4.374927720596738e-05, "loss": 1.144, "step": 144620 }, { "epoch": 12.53, "learning_rate": 4.3747108823869546e-05, "loss": 0.9098, "step": 144630 }, { "epoch": 12.53, "learning_rate": 4.3744940441771705e-05, "loss": 1.0297, "step": 144640 }, { "epoch": 12.53, "learning_rate": 4.374277205967387e-05, "loss": 1.1604, "step": 144650 }, { "epoch": 12.53, "learning_rate": 4.374060367757604e-05, "loss": 1.0756, "step": 144660 }, { "epoch": 12.53, "learning_rate": 4.37384352954782e-05, "loss": 1.0734, "step": 144670 }, { "epoch": 12.53, "learning_rate": 4.3736266913380364e-05, "loss": 0.9879, "step": 144680 }, { "epoch": 12.53, "learning_rate": 4.373409853128252e-05, "loss": 1.0011, "step": 144690 }, { "epoch": 12.53, "learning_rate": 4.373193014918468e-05, "loss": 1.0218, "step": 144700 }, { "epoch": 12.53, "learning_rate": 4.372976176708684e-05, "loss": 0.9447, "step": 144710 }, { "epoch": 12.53, "learning_rate": 4.372759338498901e-05, "loss": 0.8351, "step": 144720 }, { "epoch": 12.53, "learning_rate": 4.3725425002891176e-05, "loss": 1.0755, "step": 144730 }, { "epoch": 12.54, "learning_rate": 4.3723256620793335e-05, "loss": 1.1978, "step": 144740 }, { "epoch": 12.54, "learning_rate": 4.37210882386955e-05, "loss": 1.0751, "step": 144750 }, { "epoch": 12.54, "learning_rate": 4.3718919856597655e-05, "loss": 1.1424, "step": 144760 }, { "epoch": 12.54, "learning_rate": 4.371675147449982e-05, "loss": 0.9165, "step": 144770 }, { "epoch": 12.54, "learning_rate": 4.371458309240198e-05, "loss": 1.1003, "step": 144780 }, { "epoch": 12.54, "learning_rate": 4.371241471030415e-05, "loss": 0.9107, "step": 144790 }, { "epoch": 12.54, "learning_rate": 4.371024632820631e-05, "loss": 1.046, "step": 144800 }, { "epoch": 12.54, "learning_rate": 4.370807794610847e-05, "loss": 1.1564, "step": 144810 }, { "epoch": 12.54, "learning_rate": 4.370590956401064e-05, "loss": 1.0626, "step": 144820 }, { "epoch": 12.54, "learning_rate": 4.37037411819128e-05, "loss": 0.9981, "step": 144830 }, { "epoch": 12.54, "learning_rate": 4.370157279981496e-05, "loss": 1.2407, "step": 144840 }, { "epoch": 12.54, "learning_rate": 4.369940441771712e-05, "loss": 1.0997, "step": 144850 }, { "epoch": 12.55, "learning_rate": 4.3697236035619285e-05, "loss": 0.9396, "step": 144860 }, { "epoch": 12.55, "learning_rate": 4.369506765352145e-05, "loss": 0.97, "step": 144870 }, { "epoch": 12.55, "learning_rate": 4.369289927142361e-05, "loss": 0.9749, "step": 144880 }, { "epoch": 12.55, "learning_rate": 4.369073088932578e-05, "loss": 1.2992, "step": 144890 }, { "epoch": 12.55, "learning_rate": 4.368856250722794e-05, "loss": 1.2906, "step": 144900 }, { "epoch": 12.55, "learning_rate": 4.36863941251301e-05, "loss": 1.0477, "step": 144910 }, { "epoch": 12.55, "learning_rate": 4.3684225743032256e-05, "loss": 0.8918, "step": 144920 }, { "epoch": 12.55, "learning_rate": 4.368205736093442e-05, "loss": 1.1145, "step": 144930 }, { "epoch": 12.55, "learning_rate": 4.367988897883659e-05, "loss": 1.0925, "step": 144940 }, { "epoch": 12.55, "learning_rate": 4.367772059673875e-05, "loss": 0.8519, "step": 144950 }, { "epoch": 12.55, "learning_rate": 4.3675552214640915e-05, "loss": 0.9723, "step": 144960 }, { "epoch": 12.56, "learning_rate": 4.3673383832543074e-05, "loss": 1.0808, "step": 144970 }, { "epoch": 12.56, "learning_rate": 4.367121545044524e-05, "loss": 1.1632, "step": 144980 }, { "epoch": 12.56, "learning_rate": 4.3669047068347394e-05, "loss": 1.0567, "step": 144990 }, { "epoch": 12.56, "learning_rate": 4.366687868624956e-05, "loss": 0.9851, "step": 145000 }, { "epoch": 12.56, "learning_rate": 4.3664710304151726e-05, "loss": 1.042, "step": 145010 }, { "epoch": 12.56, "learning_rate": 4.3662541922053886e-05, "loss": 0.9997, "step": 145020 }, { "epoch": 12.56, "learning_rate": 4.366037353995605e-05, "loss": 0.9423, "step": 145030 }, { "epoch": 12.56, "learning_rate": 4.365820515785821e-05, "loss": 1.1181, "step": 145040 }, { "epoch": 12.56, "learning_rate": 4.365603677576038e-05, "loss": 1.0202, "step": 145050 }, { "epoch": 12.56, "learning_rate": 4.3653868393662545e-05, "loss": 1.0804, "step": 145060 }, { "epoch": 12.56, "learning_rate": 4.36517000115647e-05, "loss": 1.11, "step": 145070 }, { "epoch": 12.56, "learning_rate": 4.3649531629466864e-05, "loss": 1.1692, "step": 145080 }, { "epoch": 12.57, "learning_rate": 4.3647363247369024e-05, "loss": 1.2792, "step": 145090 }, { "epoch": 12.57, "learning_rate": 4.364519486527119e-05, "loss": 0.9639, "step": 145100 }, { "epoch": 12.57, "learning_rate": 4.364302648317335e-05, "loss": 0.9537, "step": 145110 }, { "epoch": 12.57, "learning_rate": 4.3640858101075516e-05, "loss": 0.8174, "step": 145120 }, { "epoch": 12.57, "learning_rate": 4.363868971897768e-05, "loss": 1.1746, "step": 145130 }, { "epoch": 12.57, "learning_rate": 4.363652133687984e-05, "loss": 1.109, "step": 145140 }, { "epoch": 12.57, "learning_rate": 4.3634352954781995e-05, "loss": 1.0344, "step": 145150 }, { "epoch": 12.57, "learning_rate": 4.363218457268416e-05, "loss": 1.1131, "step": 145160 }, { "epoch": 12.57, "learning_rate": 4.363001619058633e-05, "loss": 1.2025, "step": 145170 }, { "epoch": 12.57, "learning_rate": 4.362784780848849e-05, "loss": 1.1469, "step": 145180 }, { "epoch": 12.57, "learning_rate": 4.3625679426390654e-05, "loss": 1.1163, "step": 145190 }, { "epoch": 12.58, "learning_rate": 4.362351104429281e-05, "loss": 1.1529, "step": 145200 }, { "epoch": 12.58, "learning_rate": 4.362134266219498e-05, "loss": 1.2305, "step": 145210 }, { "epoch": 12.58, "learning_rate": 4.361917428009713e-05, "loss": 0.889, "step": 145220 }, { "epoch": 12.58, "learning_rate": 4.36170058979993e-05, "loss": 1.1366, "step": 145230 }, { "epoch": 12.58, "learning_rate": 4.3614837515901465e-05, "loss": 1.1002, "step": 145240 }, { "epoch": 12.58, "learning_rate": 4.3612669133803625e-05, "loss": 0.906, "step": 145250 }, { "epoch": 12.58, "learning_rate": 4.361050075170579e-05, "loss": 1.0371, "step": 145260 }, { "epoch": 12.58, "learning_rate": 4.360833236960795e-05, "loss": 1.1565, "step": 145270 }, { "epoch": 12.58, "learning_rate": 4.360616398751012e-05, "loss": 0.9768, "step": 145280 }, { "epoch": 12.58, "learning_rate": 4.3603995605412284e-05, "loss": 1.1533, "step": 145290 }, { "epoch": 12.58, "learning_rate": 4.3601827223314436e-05, "loss": 1.2275, "step": 145300 }, { "epoch": 12.58, "learning_rate": 4.35996588412166e-05, "loss": 0.8774, "step": 145310 }, { "epoch": 12.59, "learning_rate": 4.359749045911876e-05, "loss": 1.2574, "step": 145320 }, { "epoch": 12.59, "learning_rate": 4.359532207702093e-05, "loss": 1.0243, "step": 145330 }, { "epoch": 12.59, "learning_rate": 4.359315369492309e-05, "loss": 0.9537, "step": 145340 }, { "epoch": 12.59, "learning_rate": 4.3590985312825255e-05, "loss": 1.1392, "step": 145350 }, { "epoch": 12.59, "learning_rate": 4.358881693072742e-05, "loss": 1.116, "step": 145360 }, { "epoch": 12.59, "learning_rate": 4.358664854862958e-05, "loss": 1.2288, "step": 145370 }, { "epoch": 12.59, "learning_rate": 4.358448016653174e-05, "loss": 1.1904, "step": 145380 }, { "epoch": 12.59, "learning_rate": 4.35823117844339e-05, "loss": 0.9801, "step": 145390 }, { "epoch": 12.59, "learning_rate": 4.3580143402336066e-05, "loss": 1.1237, "step": 145400 }, { "epoch": 12.59, "learning_rate": 4.3577975020238226e-05, "loss": 1.2703, "step": 145410 }, { "epoch": 12.59, "learning_rate": 4.357580663814039e-05, "loss": 1.1743, "step": 145420 }, { "epoch": 12.6, "learning_rate": 4.357363825604256e-05, "loss": 1.194, "step": 145430 }, { "epoch": 12.6, "learning_rate": 4.357146987394472e-05, "loss": 0.9949, "step": 145440 }, { "epoch": 12.6, "learning_rate": 4.3569301491846885e-05, "loss": 1.1105, "step": 145450 }, { "epoch": 12.6, "learning_rate": 4.356713310974904e-05, "loss": 0.9575, "step": 145460 }, { "epoch": 12.6, "learning_rate": 4.3564964727651204e-05, "loss": 1.0012, "step": 145470 }, { "epoch": 12.6, "learning_rate": 4.3562796345553364e-05, "loss": 0.9941, "step": 145480 }, { "epoch": 12.6, "learning_rate": 4.356062796345553e-05, "loss": 0.9846, "step": 145490 }, { "epoch": 12.6, "learning_rate": 4.3558459581357697e-05, "loss": 0.8935, "step": 145500 }, { "epoch": 12.6, "learning_rate": 4.3556291199259856e-05, "loss": 0.9967, "step": 145510 }, { "epoch": 12.6, "learning_rate": 4.355412281716202e-05, "loss": 0.8553, "step": 145520 }, { "epoch": 12.6, "learning_rate": 4.3551954435064175e-05, "loss": 1.0836, "step": 145530 }, { "epoch": 12.6, "learning_rate": 4.354978605296634e-05, "loss": 1.0199, "step": 145540 }, { "epoch": 12.61, "learning_rate": 4.35476176708685e-05, "loss": 1.2099, "step": 145550 }, { "epoch": 12.61, "learning_rate": 4.354544928877067e-05, "loss": 1.1473, "step": 145560 }, { "epoch": 12.61, "learning_rate": 4.3543280906672834e-05, "loss": 1.0725, "step": 145570 }, { "epoch": 12.61, "learning_rate": 4.3541112524574994e-05, "loss": 0.926, "step": 145580 }, { "epoch": 12.61, "learning_rate": 4.353894414247716e-05, "loss": 1.0255, "step": 145590 }, { "epoch": 12.61, "learning_rate": 4.353677576037932e-05, "loss": 1.0278, "step": 145600 }, { "epoch": 12.61, "learning_rate": 4.353460737828148e-05, "loss": 1.2575, "step": 145610 }, { "epoch": 12.61, "learning_rate": 4.353243899618364e-05, "loss": 0.9898, "step": 145620 }, { "epoch": 12.61, "learning_rate": 4.3530270614085805e-05, "loss": 0.9774, "step": 145630 }, { "epoch": 12.61, "learning_rate": 4.352810223198797e-05, "loss": 1.0547, "step": 145640 }, { "epoch": 12.61, "learning_rate": 4.352593384989013e-05, "loss": 1.0683, "step": 145650 }, { "epoch": 12.62, "learning_rate": 4.35237654677923e-05, "loss": 1.1978, "step": 145660 }, { "epoch": 12.62, "learning_rate": 4.352159708569446e-05, "loss": 1.0498, "step": 145670 }, { "epoch": 12.62, "learning_rate": 4.3519428703596624e-05, "loss": 1.1213, "step": 145680 }, { "epoch": 12.62, "learning_rate": 4.351726032149878e-05, "loss": 1.1053, "step": 145690 }, { "epoch": 12.62, "learning_rate": 4.351509193940094e-05, "loss": 1.0427, "step": 145700 }, { "epoch": 12.62, "learning_rate": 4.35129235573031e-05, "loss": 1.1442, "step": 145710 }, { "epoch": 12.62, "learning_rate": 4.351075517520527e-05, "loss": 0.9974, "step": 145720 }, { "epoch": 12.62, "learning_rate": 4.3508586793107435e-05, "loss": 1.1551, "step": 145730 }, { "epoch": 12.62, "learning_rate": 4.3506418411009595e-05, "loss": 1.0709, "step": 145740 }, { "epoch": 12.62, "learning_rate": 4.350425002891176e-05, "loss": 0.817, "step": 145750 }, { "epoch": 12.62, "learning_rate": 4.3502081646813914e-05, "loss": 1.0254, "step": 145760 }, { "epoch": 12.62, "learning_rate": 4.349991326471608e-05, "loss": 1.2073, "step": 145770 }, { "epoch": 12.63, "learning_rate": 4.349774488261824e-05, "loss": 1.2602, "step": 145780 }, { "epoch": 12.63, "learning_rate": 4.349557650052041e-05, "loss": 1.1571, "step": 145790 }, { "epoch": 12.63, "learning_rate": 4.349340811842257e-05, "loss": 1.0356, "step": 145800 }, { "epoch": 12.63, "learning_rate": 4.349123973632473e-05, "loss": 0.8207, "step": 145810 }, { "epoch": 12.63, "learning_rate": 4.34890713542269e-05, "loss": 1.1131, "step": 145820 }, { "epoch": 12.63, "learning_rate": 4.348690297212906e-05, "loss": 1.0843, "step": 145830 }, { "epoch": 12.63, "learning_rate": 4.348473459003122e-05, "loss": 1.0611, "step": 145840 }, { "epoch": 12.63, "learning_rate": 4.348256620793338e-05, "loss": 1.2602, "step": 145850 }, { "epoch": 12.63, "learning_rate": 4.3480397825835544e-05, "loss": 1.1439, "step": 145860 }, { "epoch": 12.63, "learning_rate": 4.347822944373771e-05, "loss": 1.1188, "step": 145870 }, { "epoch": 12.63, "learning_rate": 4.347606106163987e-05, "loss": 1.3168, "step": 145880 }, { "epoch": 12.63, "learning_rate": 4.347389267954204e-05, "loss": 1.2609, "step": 145890 }, { "epoch": 12.64, "learning_rate": 4.3471724297444196e-05, "loss": 1.2359, "step": 145900 }, { "epoch": 12.64, "learning_rate": 4.346955591534636e-05, "loss": 1.189, "step": 145910 }, { "epoch": 12.64, "learning_rate": 4.3467387533248516e-05, "loss": 1.1007, "step": 145920 }, { "epoch": 12.64, "learning_rate": 4.346521915115068e-05, "loss": 1.133, "step": 145930 }, { "epoch": 12.64, "learning_rate": 4.346305076905285e-05, "loss": 1.1073, "step": 145940 }, { "epoch": 12.64, "learning_rate": 4.346088238695501e-05, "loss": 1.1365, "step": 145950 }, { "epoch": 12.64, "learning_rate": 4.3458714004857174e-05, "loss": 1.0952, "step": 145960 }, { "epoch": 12.64, "learning_rate": 4.3456545622759334e-05, "loss": 1.1337, "step": 145970 }, { "epoch": 12.64, "learning_rate": 4.34543772406615e-05, "loss": 1.0358, "step": 145980 }, { "epoch": 12.64, "learning_rate": 4.345220885856365e-05, "loss": 1.3053, "step": 145990 }, { "epoch": 12.64, "learning_rate": 4.345004047646582e-05, "loss": 1.1551, "step": 146000 }, { "epoch": 12.65, "learning_rate": 4.3447872094367986e-05, "loss": 1.2035, "step": 146010 }, { "epoch": 12.65, "learning_rate": 4.3445703712270146e-05, "loss": 0.8958, "step": 146020 }, { "epoch": 12.65, "learning_rate": 4.344353533017231e-05, "loss": 1.0071, "step": 146030 }, { "epoch": 12.65, "learning_rate": 4.344136694807447e-05, "loss": 1.133, "step": 146040 }, { "epoch": 12.65, "learning_rate": 4.343919856597664e-05, "loss": 1.0667, "step": 146050 }, { "epoch": 12.65, "learning_rate": 4.3437030183878804e-05, "loss": 1.2219, "step": 146060 }, { "epoch": 12.65, "learning_rate": 4.343486180178096e-05, "loss": 1.1414, "step": 146070 }, { "epoch": 12.65, "learning_rate": 4.3432693419683124e-05, "loss": 0.9918, "step": 146080 }, { "epoch": 12.65, "learning_rate": 4.343052503758528e-05, "loss": 1.042, "step": 146090 }, { "epoch": 12.65, "learning_rate": 4.342835665548745e-05, "loss": 1.0822, "step": 146100 }, { "epoch": 12.65, "learning_rate": 4.342618827338961e-05, "loss": 1.1611, "step": 146110 }, { "epoch": 12.65, "learning_rate": 4.3424019891291776e-05, "loss": 0.9523, "step": 146120 }, { "epoch": 12.66, "learning_rate": 4.342185150919394e-05, "loss": 1.2076, "step": 146130 }, { "epoch": 12.66, "learning_rate": 4.34196831270961e-05, "loss": 1.0168, "step": 146140 }, { "epoch": 12.66, "learning_rate": 4.341751474499826e-05, "loss": 1.2862, "step": 146150 }, { "epoch": 12.66, "learning_rate": 4.341534636290042e-05, "loss": 1.1798, "step": 146160 }, { "epoch": 12.66, "learning_rate": 4.341317798080259e-05, "loss": 1.0515, "step": 146170 }, { "epoch": 12.66, "learning_rate": 4.341100959870475e-05, "loss": 0.8341, "step": 146180 }, { "epoch": 12.66, "learning_rate": 4.340884121660691e-05, "loss": 1.0526, "step": 146190 }, { "epoch": 12.66, "learning_rate": 4.340667283450908e-05, "loss": 1.0099, "step": 146200 }, { "epoch": 12.66, "learning_rate": 4.340450445241124e-05, "loss": 1.1011, "step": 146210 }, { "epoch": 12.66, "learning_rate": 4.3402336070313406e-05, "loss": 0.9591, "step": 146220 }, { "epoch": 12.66, "learning_rate": 4.340016768821556e-05, "loss": 1.0446, "step": 146230 }, { "epoch": 12.67, "learning_rate": 4.3397999306117725e-05, "loss": 1.1265, "step": 146240 }, { "epoch": 12.67, "learning_rate": 4.3395830924019885e-05, "loss": 0.8728, "step": 146250 }, { "epoch": 12.67, "learning_rate": 4.339366254192205e-05, "loss": 0.9478, "step": 146260 }, { "epoch": 12.67, "learning_rate": 4.339149415982421e-05, "loss": 1.2073, "step": 146270 }, { "epoch": 12.67, "learning_rate": 4.338932577772638e-05, "loss": 1.1894, "step": 146280 }, { "epoch": 12.67, "learning_rate": 4.338715739562854e-05, "loss": 1.033, "step": 146290 }, { "epoch": 12.67, "learning_rate": 4.3384989013530696e-05, "loss": 1.1223, "step": 146300 }, { "epoch": 12.67, "learning_rate": 4.338282063143286e-05, "loss": 1.0516, "step": 146310 }, { "epoch": 12.67, "learning_rate": 4.338065224933502e-05, "loss": 1.043, "step": 146320 }, { "epoch": 12.67, "learning_rate": 4.337848386723719e-05, "loss": 1.0023, "step": 146330 }, { "epoch": 12.67, "learning_rate": 4.337631548513935e-05, "loss": 0.9865, "step": 146340 }, { "epoch": 12.67, "learning_rate": 4.3374147103041515e-05, "loss": 1.2232, "step": 146350 }, { "epoch": 12.68, "learning_rate": 4.337197872094368e-05, "loss": 1.1981, "step": 146360 }, { "epoch": 12.68, "learning_rate": 4.336981033884584e-05, "loss": 1.0313, "step": 146370 }, { "epoch": 12.68, "learning_rate": 4.3367641956748e-05, "loss": 1.0509, "step": 146380 }, { "epoch": 12.68, "learning_rate": 4.336547357465016e-05, "loss": 0.9969, "step": 146390 }, { "epoch": 12.68, "learning_rate": 4.3363305192552326e-05, "loss": 1.0992, "step": 146400 }, { "epoch": 12.68, "learning_rate": 4.3361136810454486e-05, "loss": 0.7964, "step": 146410 }, { "epoch": 12.68, "learning_rate": 4.335896842835665e-05, "loss": 0.9994, "step": 146420 }, { "epoch": 12.68, "learning_rate": 4.335680004625882e-05, "loss": 1.1919, "step": 146430 }, { "epoch": 12.68, "learning_rate": 4.335463166416098e-05, "loss": 0.8853, "step": 146440 }, { "epoch": 12.68, "learning_rate": 4.3352463282063145e-05, "loss": 1.0472, "step": 146450 }, { "epoch": 12.68, "learning_rate": 4.33502948999653e-05, "loss": 1.0357, "step": 146460 }, { "epoch": 12.69, "learning_rate": 4.3348126517867464e-05, "loss": 1.0325, "step": 146470 }, { "epoch": 12.69, "learning_rate": 4.3345958135769623e-05, "loss": 1.1348, "step": 146480 }, { "epoch": 12.69, "learning_rate": 4.334378975367179e-05, "loss": 1.1448, "step": 146490 }, { "epoch": 12.69, "learning_rate": 4.3341621371573956e-05, "loss": 0.9813, "step": 146500 }, { "epoch": 12.69, "learning_rate": 4.3339452989476116e-05, "loss": 1.2802, "step": 146510 }, { "epoch": 12.69, "learning_rate": 4.333728460737828e-05, "loss": 1.1724, "step": 146520 }, { "epoch": 12.69, "learning_rate": 4.3335116225280435e-05, "loss": 1.2296, "step": 146530 }, { "epoch": 12.69, "learning_rate": 4.33329478431826e-05, "loss": 1.084, "step": 146540 }, { "epoch": 12.69, "learning_rate": 4.333077946108476e-05, "loss": 0.9894, "step": 146550 }, { "epoch": 12.69, "learning_rate": 4.332861107898693e-05, "loss": 1.3013, "step": 146560 }, { "epoch": 12.69, "learning_rate": 4.3326442696889094e-05, "loss": 1.0143, "step": 146570 }, { "epoch": 12.69, "learning_rate": 4.3324274314791253e-05, "loss": 1.0583, "step": 146580 }, { "epoch": 12.7, "learning_rate": 4.332210593269342e-05, "loss": 1.1471, "step": 146590 }, { "epoch": 12.7, "learning_rate": 4.331993755059558e-05, "loss": 1.0143, "step": 146600 }, { "epoch": 12.7, "learning_rate": 4.331776916849774e-05, "loss": 0.7598, "step": 146610 }, { "epoch": 12.7, "learning_rate": 4.33156007863999e-05, "loss": 1.2247, "step": 146620 }, { "epoch": 12.7, "learning_rate": 4.3313432404302065e-05, "loss": 1.0632, "step": 146630 }, { "epoch": 12.7, "learning_rate": 4.331126402220423e-05, "loss": 1.1092, "step": 146640 }, { "epoch": 12.7, "learning_rate": 4.330909564010639e-05, "loss": 0.964, "step": 146650 }, { "epoch": 12.7, "learning_rate": 4.330692725800856e-05, "loss": 0.9703, "step": 146660 }, { "epoch": 12.7, "learning_rate": 4.330475887591072e-05, "loss": 1.1158, "step": 146670 }, { "epoch": 12.7, "learning_rate": 4.3302590493812884e-05, "loss": 1.1078, "step": 146680 }, { "epoch": 12.7, "learning_rate": 4.3300422111715036e-05, "loss": 1.0691, "step": 146690 }, { "epoch": 12.71, "learning_rate": 4.32982537296172e-05, "loss": 1.1309, "step": 146700 }, { "epoch": 12.71, "learning_rate": 4.329608534751937e-05, "loss": 1.011, "step": 146710 }, { "epoch": 12.71, "learning_rate": 4.329391696542153e-05, "loss": 1.0088, "step": 146720 }, { "epoch": 12.71, "learning_rate": 4.3291748583323695e-05, "loss": 1.2346, "step": 146730 }, { "epoch": 12.71, "learning_rate": 4.3289580201225855e-05, "loss": 1.0961, "step": 146740 }, { "epoch": 12.71, "learning_rate": 4.328741181912802e-05, "loss": 1.2422, "step": 146750 }, { "epoch": 12.71, "learning_rate": 4.3285243437030174e-05, "loss": 1.2197, "step": 146760 }, { "epoch": 12.71, "learning_rate": 4.328307505493234e-05, "loss": 0.9053, "step": 146770 }, { "epoch": 12.71, "learning_rate": 4.32809066728345e-05, "loss": 1.0795, "step": 146780 }, { "epoch": 12.71, "learning_rate": 4.3278738290736666e-05, "loss": 1.1435, "step": 146790 }, { "epoch": 12.71, "learning_rate": 4.327656990863883e-05, "loss": 0.989, "step": 146800 }, { "epoch": 12.71, "learning_rate": 4.327440152654099e-05, "loss": 1.0173, "step": 146810 }, { "epoch": 12.72, "learning_rate": 4.327223314444316e-05, "loss": 1.0483, "step": 146820 }, { "epoch": 12.72, "learning_rate": 4.327006476234532e-05, "loss": 1.1946, "step": 146830 }, { "epoch": 12.72, "learning_rate": 4.326789638024748e-05, "loss": 0.8656, "step": 146840 }, { "epoch": 12.72, "learning_rate": 4.326572799814964e-05, "loss": 1.252, "step": 146850 }, { "epoch": 12.72, "learning_rate": 4.3263559616051804e-05, "loss": 0.9851, "step": 146860 }, { "epoch": 12.72, "learning_rate": 4.326139123395397e-05, "loss": 0.9155, "step": 146870 }, { "epoch": 12.72, "learning_rate": 4.325922285185613e-05, "loss": 0.9716, "step": 146880 }, { "epoch": 12.72, "learning_rate": 4.3257054469758296e-05, "loss": 0.8365, "step": 146890 }, { "epoch": 12.72, "learning_rate": 4.3254886087660456e-05, "loss": 1.0893, "step": 146900 }, { "epoch": 12.72, "learning_rate": 4.325271770556262e-05, "loss": 1.0803, "step": 146910 }, { "epoch": 12.72, "learning_rate": 4.3250549323464775e-05, "loss": 1.2684, "step": 146920 }, { "epoch": 12.73, "learning_rate": 4.324838094136694e-05, "loss": 1.1463, "step": 146930 }, { "epoch": 12.73, "learning_rate": 4.324621255926911e-05, "loss": 0.9148, "step": 146940 }, { "epoch": 12.73, "learning_rate": 4.324404417717127e-05, "loss": 0.9229, "step": 146950 }, { "epoch": 12.73, "learning_rate": 4.3241875795073434e-05, "loss": 0.9074, "step": 146960 }, { "epoch": 12.73, "learning_rate": 4.3239707412975594e-05, "loss": 1.0764, "step": 146970 }, { "epoch": 12.73, "learning_rate": 4.323753903087776e-05, "loss": 1.07, "step": 146980 }, { "epoch": 12.73, "learning_rate": 4.3235370648779926e-05, "loss": 1.0974, "step": 146990 }, { "epoch": 12.73, "learning_rate": 4.323320226668208e-05, "loss": 1.0269, "step": 147000 }, { "epoch": 12.73, "learning_rate": 4.3231033884584246e-05, "loss": 1.0322, "step": 147010 }, { "epoch": 12.73, "learning_rate": 4.3228865502486405e-05, "loss": 1.0292, "step": 147020 }, { "epoch": 12.73, "learning_rate": 4.322669712038857e-05, "loss": 1.111, "step": 147030 }, { "epoch": 12.73, "learning_rate": 4.322452873829073e-05, "loss": 1.1701, "step": 147040 }, { "epoch": 12.74, "learning_rate": 4.32223603561929e-05, "loss": 1.2289, "step": 147050 }, { "epoch": 12.74, "learning_rate": 4.3220191974095064e-05, "loss": 1.2884, "step": 147060 }, { "epoch": 12.74, "learning_rate": 4.321802359199722e-05, "loss": 1.1147, "step": 147070 }, { "epoch": 12.74, "learning_rate": 4.321585520989938e-05, "loss": 1.1409, "step": 147080 }, { "epoch": 12.74, "learning_rate": 4.321368682780154e-05, "loss": 1.0448, "step": 147090 }, { "epoch": 12.74, "learning_rate": 4.321151844570371e-05, "loss": 1.0004, "step": 147100 }, { "epoch": 12.74, "learning_rate": 4.320935006360587e-05, "loss": 1.0352, "step": 147110 }, { "epoch": 12.74, "learning_rate": 4.3207181681508035e-05, "loss": 1.0921, "step": 147120 }, { "epoch": 12.74, "learning_rate": 4.32050132994102e-05, "loss": 0.9754, "step": 147130 }, { "epoch": 12.74, "learning_rate": 4.320284491731236e-05, "loss": 1.0719, "step": 147140 }, { "epoch": 12.74, "learning_rate": 4.320067653521452e-05, "loss": 1.0285, "step": 147150 }, { "epoch": 12.74, "learning_rate": 4.319850815311668e-05, "loss": 0.9665, "step": 147160 }, { "epoch": 12.75, "learning_rate": 4.319633977101885e-05, "loss": 0.9802, "step": 147170 }, { "epoch": 12.75, "learning_rate": 4.3194171388921007e-05, "loss": 0.8045, "step": 147180 }, { "epoch": 12.75, "learning_rate": 4.319200300682317e-05, "loss": 1.0534, "step": 147190 }, { "epoch": 12.75, "learning_rate": 4.318983462472534e-05, "loss": 1.278, "step": 147200 }, { "epoch": 12.75, "learning_rate": 4.31876662426275e-05, "loss": 1.1544, "step": 147210 }, { "epoch": 12.75, "learning_rate": 4.3185497860529665e-05, "loss": 1.0501, "step": 147220 }, { "epoch": 12.75, "learning_rate": 4.318332947843182e-05, "loss": 1.2268, "step": 147230 }, { "epoch": 12.75, "learning_rate": 4.3181161096333985e-05, "loss": 1.1443, "step": 147240 }, { "epoch": 12.75, "learning_rate": 4.3178992714236144e-05, "loss": 0.9066, "step": 147250 }, { "epoch": 12.75, "learning_rate": 4.317682433213831e-05, "loss": 1.2699, "step": 147260 }, { "epoch": 12.75, "learning_rate": 4.317465595004048e-05, "loss": 0.9601, "step": 147270 }, { "epoch": 12.76, "learning_rate": 4.3172487567942637e-05, "loss": 0.9006, "step": 147280 }, { "epoch": 12.76, "learning_rate": 4.31703191858448e-05, "loss": 1.2495, "step": 147290 }, { "epoch": 12.76, "learning_rate": 4.3168150803746956e-05, "loss": 1.1322, "step": 147300 }, { "epoch": 12.76, "learning_rate": 4.316598242164912e-05, "loss": 0.9991, "step": 147310 }, { "epoch": 12.76, "learning_rate": 4.316381403955128e-05, "loss": 1.0635, "step": 147320 }, { "epoch": 12.76, "learning_rate": 4.316164565745345e-05, "loss": 1.067, "step": 147330 }, { "epoch": 12.76, "learning_rate": 4.315947727535561e-05, "loss": 1.1204, "step": 147340 }, { "epoch": 12.76, "learning_rate": 4.3157308893257774e-05, "loss": 1.202, "step": 147350 }, { "epoch": 12.76, "learning_rate": 4.315514051115994e-05, "loss": 1.1628, "step": 147360 }, { "epoch": 12.76, "learning_rate": 4.31529721290621e-05, "loss": 1.2821, "step": 147370 }, { "epoch": 12.76, "learning_rate": 4.315080374696426e-05, "loss": 1.0294, "step": 147380 }, { "epoch": 12.76, "learning_rate": 4.314863536486642e-05, "loss": 0.8739, "step": 147390 }, { "epoch": 12.77, "learning_rate": 4.3146466982768586e-05, "loss": 0.947, "step": 147400 }, { "epoch": 12.77, "learning_rate": 4.3144298600670745e-05, "loss": 1.1623, "step": 147410 }, { "epoch": 12.77, "learning_rate": 4.314213021857291e-05, "loss": 0.9749, "step": 147420 }, { "epoch": 12.77, "learning_rate": 4.313996183647508e-05, "loss": 1.2544, "step": 147430 }, { "epoch": 12.77, "learning_rate": 4.313779345437724e-05, "loss": 1.1227, "step": 147440 }, { "epoch": 12.77, "learning_rate": 4.3135625072279404e-05, "loss": 1.1733, "step": 147450 }, { "epoch": 12.77, "learning_rate": 4.313345669018156e-05, "loss": 1.0269, "step": 147460 }, { "epoch": 12.77, "learning_rate": 4.3131288308083724e-05, "loss": 1.0971, "step": 147470 }, { "epoch": 12.77, "learning_rate": 4.312911992598588e-05, "loss": 1.2281, "step": 147480 }, { "epoch": 12.77, "learning_rate": 4.312695154388805e-05, "loss": 1.1646, "step": 147490 }, { "epoch": 12.77, "learning_rate": 4.3124783161790216e-05, "loss": 0.9486, "step": 147500 }, { "epoch": 12.78, "learning_rate": 4.3122614779692376e-05, "loss": 1.192, "step": 147510 }, { "epoch": 12.78, "learning_rate": 4.312044639759454e-05, "loss": 1.093, "step": 147520 }, { "epoch": 12.78, "learning_rate": 4.3118278015496695e-05, "loss": 0.8989, "step": 147530 }, { "epoch": 12.78, "learning_rate": 4.311610963339886e-05, "loss": 1.0487, "step": 147540 }, { "epoch": 12.78, "learning_rate": 4.311394125130102e-05, "loss": 1.1653, "step": 147550 }, { "epoch": 12.78, "learning_rate": 4.311177286920319e-05, "loss": 1.0988, "step": 147560 }, { "epoch": 12.78, "learning_rate": 4.3109604487105354e-05, "loss": 0.9886, "step": 147570 }, { "epoch": 12.78, "learning_rate": 4.310743610500751e-05, "loss": 1.0911, "step": 147580 }, { "epoch": 12.78, "learning_rate": 4.310526772290968e-05, "loss": 0.9444, "step": 147590 }, { "epoch": 12.78, "learning_rate": 4.310309934081184e-05, "loss": 1.1653, "step": 147600 }, { "epoch": 12.78, "learning_rate": 4.3100930958714e-05, "loss": 0.8814, "step": 147610 }, { "epoch": 12.78, "learning_rate": 4.309876257661616e-05, "loss": 1.1157, "step": 147620 }, { "epoch": 12.79, "learning_rate": 4.3096594194518325e-05, "loss": 1.1643, "step": 147630 }, { "epoch": 12.79, "learning_rate": 4.309442581242049e-05, "loss": 1.0595, "step": 147640 }, { "epoch": 12.79, "learning_rate": 4.309225743032265e-05, "loss": 1.0876, "step": 147650 }, { "epoch": 12.79, "learning_rate": 4.309008904822482e-05, "loss": 1.2122, "step": 147660 }, { "epoch": 12.79, "learning_rate": 4.308792066612698e-05, "loss": 0.9771, "step": 147670 }, { "epoch": 12.79, "learning_rate": 4.308575228402914e-05, "loss": 1.1163, "step": 147680 }, { "epoch": 12.79, "learning_rate": 4.3083583901931296e-05, "loss": 1.1898, "step": 147690 }, { "epoch": 12.79, "learning_rate": 4.308141551983346e-05, "loss": 0.9384, "step": 147700 }, { "epoch": 12.79, "learning_rate": 4.307924713773563e-05, "loss": 0.9606, "step": 147710 }, { "epoch": 12.79, "learning_rate": 4.307707875563779e-05, "loss": 1.0858, "step": 147720 }, { "epoch": 12.79, "learning_rate": 4.3074910373539955e-05, "loss": 1.0856, "step": 147730 }, { "epoch": 12.8, "learning_rate": 4.3072741991442114e-05, "loss": 1.0553, "step": 147740 }, { "epoch": 12.8, "learning_rate": 4.307057360934428e-05, "loss": 0.9935, "step": 147750 }, { "epoch": 12.8, "learning_rate": 4.306840522724645e-05, "loss": 1.0392, "step": 147760 }, { "epoch": 12.8, "learning_rate": 4.30662368451486e-05, "loss": 1.1088, "step": 147770 }, { "epoch": 12.8, "learning_rate": 4.3064068463050766e-05, "loss": 1.1157, "step": 147780 }, { "epoch": 12.8, "learning_rate": 4.3061900080952926e-05, "loss": 1.1751, "step": 147790 }, { "epoch": 12.8, "learning_rate": 4.305973169885509e-05, "loss": 1.1848, "step": 147800 }, { "epoch": 12.8, "learning_rate": 4.305756331675725e-05, "loss": 0.7705, "step": 147810 }, { "epoch": 12.8, "learning_rate": 4.305539493465942e-05, "loss": 1.0599, "step": 147820 }, { "epoch": 12.8, "learning_rate": 4.3053226552561585e-05, "loss": 1.0859, "step": 147830 }, { "epoch": 12.8, "learning_rate": 4.305105817046374e-05, "loss": 0.9976, "step": 147840 }, { "epoch": 12.8, "learning_rate": 4.30488897883659e-05, "loss": 1.2008, "step": 147850 }, { "epoch": 12.81, "learning_rate": 4.3046721406268064e-05, "loss": 1.0392, "step": 147860 }, { "epoch": 12.81, "learning_rate": 4.304455302417023e-05, "loss": 1.088, "step": 147870 }, { "epoch": 12.81, "learning_rate": 4.304238464207239e-05, "loss": 1.07, "step": 147880 }, { "epoch": 12.81, "learning_rate": 4.3040216259974556e-05, "loss": 1.0779, "step": 147890 }, { "epoch": 12.81, "learning_rate": 4.3038047877876716e-05, "loss": 1.0276, "step": 147900 }, { "epoch": 12.81, "learning_rate": 4.303587949577888e-05, "loss": 1.0057, "step": 147910 }, { "epoch": 12.81, "learning_rate": 4.3033711113681035e-05, "loss": 1.0731, "step": 147920 }, { "epoch": 12.81, "learning_rate": 4.30315427315832e-05, "loss": 0.9395, "step": 147930 }, { "epoch": 12.81, "learning_rate": 4.302937434948537e-05, "loss": 1.033, "step": 147940 }, { "epoch": 12.81, "learning_rate": 4.302720596738753e-05, "loss": 0.944, "step": 147950 }, { "epoch": 12.81, "learning_rate": 4.3025037585289694e-05, "loss": 0.9751, "step": 147960 }, { "epoch": 12.82, "learning_rate": 4.302286920319185e-05, "loss": 1.0851, "step": 147970 }, { "epoch": 12.82, "learning_rate": 4.302070082109402e-05, "loss": 1.061, "step": 147980 }, { "epoch": 12.82, "learning_rate": 4.3018532438996186e-05, "loss": 1.1684, "step": 147990 }, { "epoch": 12.82, "learning_rate": 4.301636405689834e-05, "loss": 1.3722, "step": 148000 }, { "epoch": 12.82, "learning_rate": 4.3014195674800505e-05, "loss": 0.9808, "step": 148010 }, { "epoch": 12.82, "learning_rate": 4.3012027292702665e-05, "loss": 0.9248, "step": 148020 }, { "epoch": 12.82, "learning_rate": 4.300985891060483e-05, "loss": 1.1126, "step": 148030 }, { "epoch": 12.82, "learning_rate": 4.300769052850699e-05, "loss": 1.1114, "step": 148040 }, { "epoch": 12.82, "learning_rate": 4.300552214640916e-05, "loss": 0.8887, "step": 148050 }, { "epoch": 12.82, "learning_rate": 4.3003353764311324e-05, "loss": 1.2357, "step": 148060 }, { "epoch": 12.82, "learning_rate": 4.300118538221348e-05, "loss": 1.0114, "step": 148070 }, { "epoch": 12.82, "learning_rate": 4.299901700011564e-05, "loss": 0.9285, "step": 148080 }, { "epoch": 12.83, "learning_rate": 4.29968486180178e-05, "loss": 1.1487, "step": 148090 }, { "epoch": 12.83, "learning_rate": 4.299468023591997e-05, "loss": 1.1758, "step": 148100 }, { "epoch": 12.83, "learning_rate": 4.299251185382213e-05, "loss": 0.9247, "step": 148110 }, { "epoch": 12.83, "learning_rate": 4.2990343471724295e-05, "loss": 1.2165, "step": 148120 }, { "epoch": 12.83, "learning_rate": 4.298817508962646e-05, "loss": 1.1321, "step": 148130 }, { "epoch": 12.83, "learning_rate": 4.298600670752862e-05, "loss": 1.1936, "step": 148140 }, { "epoch": 12.83, "learning_rate": 4.298383832543078e-05, "loss": 0.909, "step": 148150 }, { "epoch": 12.83, "learning_rate": 4.298166994333294e-05, "loss": 0.9857, "step": 148160 }, { "epoch": 12.83, "learning_rate": 4.297950156123511e-05, "loss": 0.9275, "step": 148170 }, { "epoch": 12.83, "learning_rate": 4.2977333179137266e-05, "loss": 1.0931, "step": 148180 }, { "epoch": 12.83, "learning_rate": 4.297516479703943e-05, "loss": 1.0699, "step": 148190 }, { "epoch": 12.84, "learning_rate": 4.29729964149416e-05, "loss": 1.1101, "step": 148200 }, { "epoch": 12.84, "learning_rate": 4.297082803284376e-05, "loss": 1.109, "step": 148210 }, { "epoch": 12.84, "learning_rate": 4.2968659650745925e-05, "loss": 1.0962, "step": 148220 }, { "epoch": 12.84, "learning_rate": 4.296649126864808e-05, "loss": 1.0948, "step": 148230 }, { "epoch": 12.84, "learning_rate": 4.2964322886550244e-05, "loss": 0.9783, "step": 148240 }, { "epoch": 12.84, "learning_rate": 4.2962154504452404e-05, "loss": 0.9703, "step": 148250 }, { "epoch": 12.84, "learning_rate": 4.295998612235457e-05, "loss": 0.9957, "step": 148260 }, { "epoch": 12.84, "learning_rate": 4.295781774025674e-05, "loss": 0.9155, "step": 148270 }, { "epoch": 12.84, "learning_rate": 4.2955649358158896e-05, "loss": 1.0763, "step": 148280 }, { "epoch": 12.84, "learning_rate": 4.295348097606106e-05, "loss": 1.0711, "step": 148290 }, { "epoch": 12.84, "learning_rate": 4.2951312593963216e-05, "loss": 1.3367, "step": 148300 }, { "epoch": 12.84, "learning_rate": 4.294914421186538e-05, "loss": 1.1446, "step": 148310 }, { "epoch": 12.85, "learning_rate": 4.294697582976754e-05, "loss": 0.9614, "step": 148320 }, { "epoch": 12.85, "learning_rate": 4.294480744766971e-05, "loss": 1.236, "step": 148330 }, { "epoch": 12.85, "learning_rate": 4.2942639065571874e-05, "loss": 1.208, "step": 148340 }, { "epoch": 12.85, "learning_rate": 4.2940470683474034e-05, "loss": 1.008, "step": 148350 }, { "epoch": 12.85, "learning_rate": 4.29383023013762e-05, "loss": 1.2542, "step": 148360 }, { "epoch": 12.85, "learning_rate": 4.293613391927836e-05, "loss": 1.1925, "step": 148370 }, { "epoch": 12.85, "learning_rate": 4.293396553718052e-05, "loss": 1.0386, "step": 148380 }, { "epoch": 12.85, "learning_rate": 4.293179715508268e-05, "loss": 1.0515, "step": 148390 }, { "epoch": 12.85, "learning_rate": 4.2929628772984846e-05, "loss": 1.1526, "step": 148400 }, { "epoch": 12.85, "learning_rate": 4.2927460390887005e-05, "loss": 1.0109, "step": 148410 }, { "epoch": 12.85, "learning_rate": 4.292529200878917e-05, "loss": 0.8392, "step": 148420 }, { "epoch": 12.85, "learning_rate": 4.292312362669134e-05, "loss": 1.1816, "step": 148430 }, { "epoch": 12.86, "learning_rate": 4.29209552445935e-05, "loss": 1.0046, "step": 148440 }, { "epoch": 12.86, "learning_rate": 4.2918786862495664e-05, "loss": 1.1205, "step": 148450 }, { "epoch": 12.86, "learning_rate": 4.291661848039782e-05, "loss": 1.0054, "step": 148460 }, { "epoch": 12.86, "learning_rate": 4.291445009829998e-05, "loss": 1.117, "step": 148470 }, { "epoch": 12.86, "learning_rate": 4.291228171620214e-05, "loss": 1.1083, "step": 148480 }, { "epoch": 12.86, "learning_rate": 4.291011333410431e-05, "loss": 1.0274, "step": 148490 }, { "epoch": 12.86, "learning_rate": 4.2907944952006476e-05, "loss": 1.1109, "step": 148500 }, { "epoch": 12.86, "learning_rate": 4.2905776569908635e-05, "loss": 1.0618, "step": 148510 }, { "epoch": 12.86, "learning_rate": 4.29036081878108e-05, "loss": 0.9147, "step": 148520 }, { "epoch": 12.86, "learning_rate": 4.290143980571296e-05, "loss": 1.0409, "step": 148530 }, { "epoch": 12.86, "learning_rate": 4.289927142361512e-05, "loss": 1.1664, "step": 148540 }, { "epoch": 12.87, "learning_rate": 4.289710304151728e-05, "loss": 1.1145, "step": 148550 }, { "epoch": 12.87, "learning_rate": 4.289493465941945e-05, "loss": 1.0104, "step": 148560 }, { "epoch": 12.87, "learning_rate": 4.289276627732161e-05, "loss": 1.086, "step": 148570 }, { "epoch": 12.87, "learning_rate": 4.289059789522377e-05, "loss": 1.1364, "step": 148580 }, { "epoch": 12.87, "learning_rate": 4.288842951312594e-05, "loss": 1.0824, "step": 148590 }, { "epoch": 12.87, "learning_rate": 4.28862611310281e-05, "loss": 1.2575, "step": 148600 }, { "epoch": 12.87, "learning_rate": 4.288409274893026e-05, "loss": 1.0322, "step": 148610 }, { "epoch": 12.87, "learning_rate": 4.288192436683242e-05, "loss": 1.0257, "step": 148620 }, { "epoch": 12.87, "learning_rate": 4.2879755984734584e-05, "loss": 1.0758, "step": 148630 }, { "epoch": 12.87, "learning_rate": 4.287758760263675e-05, "loss": 1.0345, "step": 148640 }, { "epoch": 12.87, "learning_rate": 4.287541922053891e-05, "loss": 1.1233, "step": 148650 }, { "epoch": 12.87, "learning_rate": 4.287325083844108e-05, "loss": 1.1552, "step": 148660 }, { "epoch": 12.88, "learning_rate": 4.2871082456343237e-05, "loss": 0.9523, "step": 148670 }, { "epoch": 12.88, "learning_rate": 4.28689140742454e-05, "loss": 1.065, "step": 148680 }, { "epoch": 12.88, "learning_rate": 4.2866745692147556e-05, "loss": 1.0623, "step": 148690 }, { "epoch": 12.88, "learning_rate": 4.286457731004972e-05, "loss": 1.0998, "step": 148700 }, { "epoch": 12.88, "learning_rate": 4.286240892795189e-05, "loss": 0.8929, "step": 148710 }, { "epoch": 12.88, "learning_rate": 4.286024054585405e-05, "loss": 1.0601, "step": 148720 }, { "epoch": 12.88, "learning_rate": 4.2858072163756215e-05, "loss": 0.9352, "step": 148730 }, { "epoch": 12.88, "learning_rate": 4.2855903781658374e-05, "loss": 1.0414, "step": 148740 }, { "epoch": 12.88, "learning_rate": 4.285373539956054e-05, "loss": 1.258, "step": 148750 }, { "epoch": 12.88, "learning_rate": 4.285156701746271e-05, "loss": 0.94, "step": 148760 }, { "epoch": 12.88, "learning_rate": 4.284939863536486e-05, "loss": 1.0499, "step": 148770 }, { "epoch": 12.89, "learning_rate": 4.2847230253267026e-05, "loss": 0.8958, "step": 148780 }, { "epoch": 12.89, "learning_rate": 4.2845061871169186e-05, "loss": 1.0555, "step": 148790 }, { "epoch": 12.89, "learning_rate": 4.284289348907135e-05, "loss": 1.0462, "step": 148800 }, { "epoch": 12.89, "learning_rate": 4.284072510697351e-05, "loss": 1.1969, "step": 148810 }, { "epoch": 12.89, "learning_rate": 4.283855672487568e-05, "loss": 1.027, "step": 148820 }, { "epoch": 12.89, "learning_rate": 4.2836388342777845e-05, "loss": 1.0131, "step": 148830 }, { "epoch": 12.89, "learning_rate": 4.283421996068e-05, "loss": 1.0338, "step": 148840 }, { "epoch": 12.89, "learning_rate": 4.2832051578582164e-05, "loss": 1.0386, "step": 148850 }, { "epoch": 12.89, "learning_rate": 4.2829883196484323e-05, "loss": 1.1442, "step": 148860 }, { "epoch": 12.89, "learning_rate": 4.282771481438649e-05, "loss": 1.1433, "step": 148870 }, { "epoch": 12.89, "learning_rate": 4.282554643228865e-05, "loss": 1.1506, "step": 148880 }, { "epoch": 12.89, "learning_rate": 4.2823378050190816e-05, "loss": 1.0397, "step": 148890 }, { "epoch": 12.9, "learning_rate": 4.282120966809298e-05, "loss": 1.0553, "step": 148900 }, { "epoch": 12.9, "learning_rate": 4.281904128599514e-05, "loss": 1.1461, "step": 148910 }, { "epoch": 12.9, "learning_rate": 4.2816872903897295e-05, "loss": 1.3226, "step": 148920 }, { "epoch": 12.9, "learning_rate": 4.281470452179946e-05, "loss": 0.9671, "step": 148930 }, { "epoch": 12.9, "learning_rate": 4.281253613970163e-05, "loss": 1.0117, "step": 148940 }, { "epoch": 12.9, "learning_rate": 4.281036775760379e-05, "loss": 0.903, "step": 148950 }, { "epoch": 12.9, "learning_rate": 4.2808199375505953e-05, "loss": 1.3283, "step": 148960 }, { "epoch": 12.9, "learning_rate": 4.280603099340811e-05, "loss": 0.9431, "step": 148970 }, { "epoch": 12.9, "learning_rate": 4.280386261131028e-05, "loss": 0.9518, "step": 148980 }, { "epoch": 12.9, "learning_rate": 4.2801694229212446e-05, "loss": 1.1437, "step": 148990 }, { "epoch": 12.9, "learning_rate": 4.27995258471146e-05, "loss": 0.9929, "step": 149000 }, { "epoch": 12.91, "learning_rate": 4.2797357465016765e-05, "loss": 0.9834, "step": 149010 }, { "epoch": 12.91, "learning_rate": 4.2795189082918925e-05, "loss": 1.0926, "step": 149020 }, { "epoch": 12.91, "learning_rate": 4.279302070082109e-05, "loss": 0.9743, "step": 149030 }, { "epoch": 12.91, "learning_rate": 4.279085231872325e-05, "loss": 1.2181, "step": 149040 }, { "epoch": 12.91, "learning_rate": 4.278868393662542e-05, "loss": 1.1733, "step": 149050 }, { "epoch": 12.91, "learning_rate": 4.2786515554527583e-05, "loss": 0.9664, "step": 149060 }, { "epoch": 12.91, "learning_rate": 4.2784347172429736e-05, "loss": 1.0472, "step": 149070 }, { "epoch": 12.91, "learning_rate": 4.27821787903319e-05, "loss": 1.1393, "step": 149080 }, { "epoch": 12.91, "learning_rate": 4.278001040823406e-05, "loss": 1.0583, "step": 149090 }, { "epoch": 12.91, "learning_rate": 4.277784202613623e-05, "loss": 1.0213, "step": 149100 }, { "epoch": 12.91, "learning_rate": 4.277567364403839e-05, "loss": 1.23, "step": 149110 }, { "epoch": 12.91, "learning_rate": 4.2773505261940555e-05, "loss": 0.9853, "step": 149120 }, { "epoch": 12.92, "learning_rate": 4.277133687984272e-05, "loss": 0.9735, "step": 149130 }, { "epoch": 12.92, "learning_rate": 4.276916849774488e-05, "loss": 1.1727, "step": 149140 }, { "epoch": 12.92, "learning_rate": 4.276700011564704e-05, "loss": 1.0589, "step": 149150 }, { "epoch": 12.92, "learning_rate": 4.27648317335492e-05, "loss": 1.0159, "step": 149160 }, { "epoch": 12.92, "learning_rate": 4.2762663351451366e-05, "loss": 1.0785, "step": 149170 }, { "epoch": 12.92, "learning_rate": 4.2760494969353526e-05, "loss": 1.1017, "step": 149180 }, { "epoch": 12.92, "learning_rate": 4.275832658725569e-05, "loss": 1.1234, "step": 149190 }, { "epoch": 12.92, "learning_rate": 4.275615820515786e-05, "loss": 1.1094, "step": 149200 }, { "epoch": 12.92, "learning_rate": 4.275398982306002e-05, "loss": 0.8888, "step": 149210 }, { "epoch": 12.92, "learning_rate": 4.2751821440962185e-05, "loss": 1.0119, "step": 149220 }, { "epoch": 12.92, "learning_rate": 4.274965305886434e-05, "loss": 1.0745, "step": 149230 }, { "epoch": 12.93, "learning_rate": 4.2747484676766504e-05, "loss": 1.0966, "step": 149240 }, { "epoch": 12.93, "learning_rate": 4.2745316294668664e-05, "loss": 1.1934, "step": 149250 }, { "epoch": 12.93, "learning_rate": 4.274314791257083e-05, "loss": 0.9274, "step": 149260 }, { "epoch": 12.93, "learning_rate": 4.2740979530472996e-05, "loss": 1.0744, "step": 149270 }, { "epoch": 12.93, "learning_rate": 4.2738811148375156e-05, "loss": 1.1424, "step": 149280 }, { "epoch": 12.93, "learning_rate": 4.273664276627732e-05, "loss": 1.0438, "step": 149290 }, { "epoch": 12.93, "learning_rate": 4.273447438417948e-05, "loss": 1.0478, "step": 149300 }, { "epoch": 12.93, "learning_rate": 4.273230600208164e-05, "loss": 1.1604, "step": 149310 }, { "epoch": 12.93, "learning_rate": 4.27301376199838e-05, "loss": 1.0114, "step": 149320 }, { "epoch": 12.93, "learning_rate": 4.272796923788597e-05, "loss": 1.049, "step": 149330 }, { "epoch": 12.93, "learning_rate": 4.2725800855788134e-05, "loss": 1.0597, "step": 149340 }, { "epoch": 12.93, "learning_rate": 4.2723632473690294e-05, "loss": 0.927, "step": 149350 }, { "epoch": 12.94, "learning_rate": 4.272146409159246e-05, "loss": 1.1966, "step": 149360 }, { "epoch": 12.94, "learning_rate": 4.271929570949462e-05, "loss": 1.0818, "step": 149370 }, { "epoch": 12.94, "learning_rate": 4.271712732739678e-05, "loss": 1.0396, "step": 149380 }, { "epoch": 12.94, "learning_rate": 4.271495894529894e-05, "loss": 1.198, "step": 149390 }, { "epoch": 12.94, "learning_rate": 4.2712790563201105e-05, "loss": 1.047, "step": 149400 }, { "epoch": 12.94, "learning_rate": 4.271062218110327e-05, "loss": 1.1139, "step": 149410 }, { "epoch": 12.94, "learning_rate": 4.270845379900543e-05, "loss": 0.9904, "step": 149420 }, { "epoch": 12.94, "learning_rate": 4.27062854169076e-05, "loss": 1.3463, "step": 149430 }, { "epoch": 12.94, "learning_rate": 4.270411703480976e-05, "loss": 0.9974, "step": 149440 }, { "epoch": 12.94, "learning_rate": 4.2701948652711924e-05, "loss": 0.9744, "step": 149450 }, { "epoch": 12.94, "learning_rate": 4.2699780270614077e-05, "loss": 0.9754, "step": 149460 }, { "epoch": 12.95, "learning_rate": 4.269761188851624e-05, "loss": 1.0422, "step": 149470 }, { "epoch": 12.95, "learning_rate": 4.26954435064184e-05, "loss": 1.1018, "step": 149480 }, { "epoch": 12.95, "learning_rate": 4.269327512432057e-05, "loss": 1.0742, "step": 149490 }, { "epoch": 12.95, "learning_rate": 4.2691106742222735e-05, "loss": 0.9411, "step": 149500 }, { "epoch": 12.95, "learning_rate": 4.2688938360124895e-05, "loss": 1.2142, "step": 149510 }, { "epoch": 12.95, "learning_rate": 4.268676997802706e-05, "loss": 1.1582, "step": 149520 }, { "epoch": 12.95, "learning_rate": 4.268460159592922e-05, "loss": 0.9323, "step": 149530 }, { "epoch": 12.95, "learning_rate": 4.268243321383138e-05, "loss": 1.159, "step": 149540 }, { "epoch": 12.95, "learning_rate": 4.268026483173354e-05, "loss": 1.0997, "step": 149550 }, { "epoch": 12.95, "learning_rate": 4.2678096449635707e-05, "loss": 1.1073, "step": 149560 }, { "epoch": 12.95, "learning_rate": 4.267592806753787e-05, "loss": 0.9961, "step": 149570 }, { "epoch": 12.95, "learning_rate": 4.267375968544003e-05, "loss": 1.0931, "step": 149580 }, { "epoch": 12.96, "learning_rate": 4.26715913033422e-05, "loss": 1.1935, "step": 149590 }, { "epoch": 12.96, "learning_rate": 4.266942292124436e-05, "loss": 1.3053, "step": 149600 }, { "epoch": 12.96, "learning_rate": 4.266725453914652e-05, "loss": 1.2758, "step": 149610 }, { "epoch": 12.96, "learning_rate": 4.266508615704868e-05, "loss": 0.9651, "step": 149620 }, { "epoch": 12.96, "learning_rate": 4.2662917774950844e-05, "loss": 1.133, "step": 149630 }, { "epoch": 12.96, "learning_rate": 4.266074939285301e-05, "loss": 1.0568, "step": 149640 }, { "epoch": 12.96, "learning_rate": 4.265858101075517e-05, "loss": 1.1291, "step": 149650 }, { "epoch": 12.96, "learning_rate": 4.2656412628657337e-05, "loss": 1.115, "step": 149660 }, { "epoch": 12.96, "learning_rate": 4.2654244246559496e-05, "loss": 0.9622, "step": 149670 }, { "epoch": 12.96, "learning_rate": 4.265207586446166e-05, "loss": 0.9801, "step": 149680 }, { "epoch": 12.96, "learning_rate": 4.2649907482363815e-05, "loss": 1.2, "step": 149690 }, { "epoch": 12.96, "learning_rate": 4.264773910026598e-05, "loss": 0.9314, "step": 149700 }, { "epoch": 12.97, "learning_rate": 4.264557071816815e-05, "loss": 1.0417, "step": 149710 }, { "epoch": 12.97, "learning_rate": 4.264340233607031e-05, "loss": 1.168, "step": 149720 }, { "epoch": 12.97, "learning_rate": 4.2641233953972474e-05, "loss": 1.2988, "step": 149730 }, { "epoch": 12.97, "learning_rate": 4.2639065571874634e-05, "loss": 1.1834, "step": 149740 }, { "epoch": 12.97, "learning_rate": 4.26368971897768e-05, "loss": 1.08, "step": 149750 }, { "epoch": 12.97, "learning_rate": 4.263472880767897e-05, "loss": 1.0117, "step": 149760 }, { "epoch": 12.97, "learning_rate": 4.263256042558112e-05, "loss": 0.9912, "step": 149770 }, { "epoch": 12.97, "learning_rate": 4.2630392043483286e-05, "loss": 1.1322, "step": 149780 }, { "epoch": 12.97, "learning_rate": 4.2628223661385445e-05, "loss": 1.0413, "step": 149790 }, { "epoch": 12.97, "learning_rate": 4.262605527928761e-05, "loss": 1.0143, "step": 149800 }, { "epoch": 12.97, "learning_rate": 4.262388689718977e-05, "loss": 0.9209, "step": 149810 }, { "epoch": 12.98, "learning_rate": 4.262171851509194e-05, "loss": 1.2061, "step": 149820 }, { "epoch": 12.98, "learning_rate": 4.2619550132994104e-05, "loss": 1.036, "step": 149830 }, { "epoch": 12.98, "learning_rate": 4.261738175089626e-05, "loss": 1.1573, "step": 149840 }, { "epoch": 12.98, "learning_rate": 4.2615213368798423e-05, "loss": 1.1542, "step": 149850 }, { "epoch": 12.98, "learning_rate": 4.261304498670058e-05, "loss": 1.119, "step": 149860 }, { "epoch": 12.98, "learning_rate": 4.261087660460275e-05, "loss": 1.1638, "step": 149870 }, { "epoch": 12.98, "learning_rate": 4.260870822250491e-05, "loss": 1.2072, "step": 149880 }, { "epoch": 12.98, "learning_rate": 4.2606539840407075e-05, "loss": 1.1114, "step": 149890 }, { "epoch": 12.98, "learning_rate": 4.260437145830924e-05, "loss": 1.0507, "step": 149900 }, { "epoch": 12.98, "learning_rate": 4.26022030762114e-05, "loss": 0.8987, "step": 149910 }, { "epoch": 12.98, "learning_rate": 4.260003469411356e-05, "loss": 0.8927, "step": 149920 }, { "epoch": 12.98, "learning_rate": 4.259786631201572e-05, "loss": 1.1478, "step": 149930 }, { "epoch": 12.99, "learning_rate": 4.259569792991789e-05, "loss": 1.0396, "step": 149940 }, { "epoch": 12.99, "learning_rate": 4.259352954782005e-05, "loss": 1.2034, "step": 149950 }, { "epoch": 12.99, "learning_rate": 4.259136116572221e-05, "loss": 1.1213, "step": 149960 }, { "epoch": 12.99, "learning_rate": 4.258919278362438e-05, "loss": 1.0186, "step": 149970 }, { "epoch": 12.99, "learning_rate": 4.258702440152654e-05, "loss": 1.0094, "step": 149980 }, { "epoch": 12.99, "learning_rate": 4.2584856019428706e-05, "loss": 1.0222, "step": 149990 }, { "epoch": 12.99, "learning_rate": 4.258268763733086e-05, "loss": 0.9582, "step": 150000 }, { "epoch": 12.99, "learning_rate": 4.2580519255233025e-05, "loss": 1.2264, "step": 150010 }, { "epoch": 12.99, "learning_rate": 4.2578350873135184e-05, "loss": 0.7947, "step": 150020 }, { "epoch": 12.99, "learning_rate": 4.257618249103735e-05, "loss": 1.3536, "step": 150030 }, { "epoch": 12.99, "learning_rate": 4.257401410893951e-05, "loss": 1.1943, "step": 150040 }, { "epoch": 13.0, "learning_rate": 4.257184572684168e-05, "loss": 0.9962, "step": 150050 }, { "epoch": 13.0, "learning_rate": 4.256967734474384e-05, "loss": 1.0612, "step": 150060 }, { "epoch": 13.0, "learning_rate": 4.2567508962646e-05, "loss": 1.1234, "step": 150070 }, { "epoch": 13.0, "learning_rate": 4.256534058054816e-05, "loss": 1.0907, "step": 150080 }, { "epoch": 13.0, "learning_rate": 4.256317219845032e-05, "loss": 1.2319, "step": 150090 }, { "epoch": 13.0, "learning_rate": 4.256100381635249e-05, "loss": 1.0449, "step": 150100 }, { "epoch": 13.0, "eval_cer": 0.5356906097145826, "eval_em": 0.04177411036616813, "eval_f1": 0.04177411036616813, "eval_loss": 1.473405122756958, "eval_runtime": 255.4596, "eval_samples_per_second": 7.59, "eval_steps_per_second": 0.951, "eval_wer": 0.9582258896338318, "step": 150104 }, { "epoch": 13.0, "learning_rate": 4.255883543425465e-05, "loss": 1.0056, "step": 150110 }, { "epoch": 13.0, "learning_rate": 4.2556667052156814e-05, "loss": 1.163, "step": 150120 }, { "epoch": 13.0, "learning_rate": 4.255449867005898e-05, "loss": 1.0799, "step": 150130 }, { "epoch": 13.0, "learning_rate": 4.255233028796114e-05, "loss": 0.9899, "step": 150140 }, { "epoch": 13.0, "learning_rate": 4.25501619058633e-05, "loss": 0.8263, "step": 150150 }, { "epoch": 13.0, "learning_rate": 4.254799352376546e-05, "loss": 0.9845, "step": 150160 }, { "epoch": 13.01, "learning_rate": 4.2545825141667626e-05, "loss": 1.1173, "step": 150170 }, { "epoch": 13.01, "learning_rate": 4.2543656759569786e-05, "loss": 1.1728, "step": 150180 }, { "epoch": 13.01, "learning_rate": 4.254148837747195e-05, "loss": 1.3074, "step": 150190 }, { "epoch": 13.01, "learning_rate": 4.253931999537412e-05, "loss": 0.8032, "step": 150200 }, { "epoch": 13.01, "learning_rate": 4.253715161327628e-05, "loss": 1.0405, "step": 150210 }, { "epoch": 13.01, "learning_rate": 4.2534983231178444e-05, "loss": 1.036, "step": 150220 }, { "epoch": 13.01, "learning_rate": 4.25328148490806e-05, "loss": 1.0392, "step": 150230 }, { "epoch": 13.01, "learning_rate": 4.2530646466982764e-05, "loss": 1.0257, "step": 150240 }, { "epoch": 13.01, "learning_rate": 4.252847808488492e-05, "loss": 1.0981, "step": 150250 }, { "epoch": 13.01, "learning_rate": 4.252630970278709e-05, "loss": 1.0539, "step": 150260 }, { "epoch": 13.01, "learning_rate": 4.2524141320689256e-05, "loss": 0.9701, "step": 150270 }, { "epoch": 13.02, "learning_rate": 4.2521972938591416e-05, "loss": 0.9905, "step": 150280 }, { "epoch": 13.02, "learning_rate": 4.251980455649358e-05, "loss": 1.0556, "step": 150290 }, { "epoch": 13.02, "learning_rate": 4.251763617439574e-05, "loss": 1.0211, "step": 150300 }, { "epoch": 13.02, "learning_rate": 4.25154677922979e-05, "loss": 1.0435, "step": 150310 }, { "epoch": 13.02, "learning_rate": 4.251329941020006e-05, "loss": 1.048, "step": 150320 }, { "epoch": 13.02, "learning_rate": 4.251113102810223e-05, "loss": 0.9812, "step": 150330 }, { "epoch": 13.02, "learning_rate": 4.2508962646004394e-05, "loss": 1.0283, "step": 150340 }, { "epoch": 13.02, "learning_rate": 4.250679426390655e-05, "loss": 0.8862, "step": 150350 }, { "epoch": 13.02, "learning_rate": 4.250462588180872e-05, "loss": 0.9683, "step": 150360 }, { "epoch": 13.02, "learning_rate": 4.250245749971088e-05, "loss": 0.9947, "step": 150370 }, { "epoch": 13.02, "learning_rate": 4.250028911761304e-05, "loss": 0.9279, "step": 150380 }, { "epoch": 13.02, "learning_rate": 4.24981207355152e-05, "loss": 0.9963, "step": 150390 }, { "epoch": 13.03, "learning_rate": 4.2495952353417365e-05, "loss": 0.8973, "step": 150400 }, { "epoch": 13.03, "learning_rate": 4.249378397131953e-05, "loss": 1.0901, "step": 150410 }, { "epoch": 13.03, "learning_rate": 4.249161558922169e-05, "loss": 0.9609, "step": 150420 }, { "epoch": 13.03, "learning_rate": 4.248944720712386e-05, "loss": 0.9947, "step": 150430 }, { "epoch": 13.03, "learning_rate": 4.248727882502602e-05, "loss": 1.0102, "step": 150440 }, { "epoch": 13.03, "learning_rate": 4.248511044292818e-05, "loss": 1.0232, "step": 150450 }, { "epoch": 13.03, "learning_rate": 4.2482942060830336e-05, "loss": 1.0476, "step": 150460 }, { "epoch": 13.03, "learning_rate": 4.24807736787325e-05, "loss": 1.051, "step": 150470 }, { "epoch": 13.03, "learning_rate": 4.247860529663467e-05, "loss": 1.033, "step": 150480 }, { "epoch": 13.03, "learning_rate": 4.247643691453683e-05, "loss": 0.9493, "step": 150490 }, { "epoch": 13.03, "learning_rate": 4.2474268532438995e-05, "loss": 0.9899, "step": 150500 }, { "epoch": 13.04, "learning_rate": 4.2472100150341155e-05, "loss": 0.9885, "step": 150510 }, { "epoch": 13.04, "learning_rate": 4.246993176824332e-05, "loss": 0.9719, "step": 150520 }, { "epoch": 13.04, "learning_rate": 4.246776338614549e-05, "loss": 0.9399, "step": 150530 }, { "epoch": 13.04, "learning_rate": 4.246559500404764e-05, "loss": 0.926, "step": 150540 }, { "epoch": 13.04, "learning_rate": 4.24634266219498e-05, "loss": 1.04, "step": 150550 }, { "epoch": 13.04, "learning_rate": 4.2461258239851966e-05, "loss": 1.1294, "step": 150560 }, { "epoch": 13.04, "learning_rate": 4.245908985775413e-05, "loss": 1.0299, "step": 150570 }, { "epoch": 13.04, "learning_rate": 4.245692147565629e-05, "loss": 0.8159, "step": 150580 }, { "epoch": 13.04, "learning_rate": 4.245475309355846e-05, "loss": 1.0575, "step": 150590 }, { "epoch": 13.04, "learning_rate": 4.245258471146062e-05, "loss": 1.0442, "step": 150600 }, { "epoch": 13.04, "learning_rate": 4.245041632936278e-05, "loss": 1.075, "step": 150610 }, { "epoch": 13.04, "learning_rate": 4.244824794726494e-05, "loss": 1.0417, "step": 150620 }, { "epoch": 13.05, "learning_rate": 4.2446079565167104e-05, "loss": 1.0197, "step": 150630 }, { "epoch": 13.05, "learning_rate": 4.244391118306927e-05, "loss": 0.8143, "step": 150640 }, { "epoch": 13.05, "learning_rate": 4.244174280097143e-05, "loss": 0.9028, "step": 150650 }, { "epoch": 13.05, "learning_rate": 4.2439574418873596e-05, "loss": 0.9611, "step": 150660 }, { "epoch": 13.05, "learning_rate": 4.2437406036775756e-05, "loss": 0.973, "step": 150670 }, { "epoch": 13.05, "learning_rate": 4.243523765467792e-05, "loss": 0.9287, "step": 150680 }, { "epoch": 13.05, "learning_rate": 4.2433069272580075e-05, "loss": 1.1155, "step": 150690 }, { "epoch": 13.05, "learning_rate": 4.243090089048224e-05, "loss": 1.0621, "step": 150700 }, { "epoch": 13.05, "learning_rate": 4.242873250838441e-05, "loss": 1.0932, "step": 150710 }, { "epoch": 13.05, "learning_rate": 4.242656412628657e-05, "loss": 0.9699, "step": 150720 }, { "epoch": 13.05, "learning_rate": 4.2424395744188734e-05, "loss": 0.8259, "step": 150730 }, { "epoch": 13.06, "learning_rate": 4.2422227362090894e-05, "loss": 1.0341, "step": 150740 }, { "epoch": 13.06, "learning_rate": 4.242005897999306e-05, "loss": 1.0329, "step": 150750 }, { "epoch": 13.06, "learning_rate": 4.2417890597895226e-05, "loss": 0.8519, "step": 150760 }, { "epoch": 13.06, "learning_rate": 4.241572221579738e-05, "loss": 0.9161, "step": 150770 }, { "epoch": 13.06, "learning_rate": 4.2413553833699546e-05, "loss": 0.9258, "step": 150780 }, { "epoch": 13.06, "learning_rate": 4.2411385451601705e-05, "loss": 0.8996, "step": 150790 }, { "epoch": 13.06, "learning_rate": 4.240921706950387e-05, "loss": 1.0864, "step": 150800 }, { "epoch": 13.06, "learning_rate": 4.240704868740603e-05, "loss": 1.107, "step": 150810 }, { "epoch": 13.06, "learning_rate": 4.24048803053082e-05, "loss": 0.9347, "step": 150820 }, { "epoch": 13.06, "learning_rate": 4.2402711923210364e-05, "loss": 1.0096, "step": 150830 }, { "epoch": 13.06, "learning_rate": 4.2400543541112524e-05, "loss": 0.8254, "step": 150840 }, { "epoch": 13.06, "learning_rate": 4.239837515901468e-05, "loss": 1.1509, "step": 150850 }, { "epoch": 13.07, "learning_rate": 4.239620677691684e-05, "loss": 1.0778, "step": 150860 }, { "epoch": 13.07, "learning_rate": 4.239403839481901e-05, "loss": 0.9426, "step": 150870 }, { "epoch": 13.07, "learning_rate": 4.239187001272117e-05, "loss": 1.1064, "step": 150880 }, { "epoch": 13.07, "learning_rate": 4.2389701630623335e-05, "loss": 1.0563, "step": 150890 }, { "epoch": 13.07, "learning_rate": 4.23875332485255e-05, "loss": 1.0168, "step": 150900 }, { "epoch": 13.07, "learning_rate": 4.238536486642766e-05, "loss": 1.0147, "step": 150910 }, { "epoch": 13.07, "learning_rate": 4.238319648432982e-05, "loss": 0.9373, "step": 150920 }, { "epoch": 13.07, "learning_rate": 4.238102810223198e-05, "loss": 1.1028, "step": 150930 }, { "epoch": 13.07, "learning_rate": 4.237885972013415e-05, "loss": 0.8425, "step": 150940 }, { "epoch": 13.07, "learning_rate": 4.2376691338036306e-05, "loss": 0.9234, "step": 150950 }, { "epoch": 13.07, "learning_rate": 4.237452295593847e-05, "loss": 1.1571, "step": 150960 }, { "epoch": 13.07, "learning_rate": 4.237235457384064e-05, "loss": 0.9556, "step": 150970 }, { "epoch": 13.08, "learning_rate": 4.23701861917428e-05, "loss": 1.1294, "step": 150980 }, { "epoch": 13.08, "learning_rate": 4.2368017809644965e-05, "loss": 0.927, "step": 150990 }, { "epoch": 13.08, "learning_rate": 4.236584942754712e-05, "loss": 0.8562, "step": 151000 }, { "epoch": 13.08, "learning_rate": 4.2363681045449284e-05, "loss": 0.9186, "step": 151010 }, { "epoch": 13.08, "learning_rate": 4.2361512663351444e-05, "loss": 0.859, "step": 151020 }, { "epoch": 13.08, "learning_rate": 4.235934428125361e-05, "loss": 0.8445, "step": 151030 }, { "epoch": 13.08, "learning_rate": 4.235717589915578e-05, "loss": 0.8764, "step": 151040 }, { "epoch": 13.08, "learning_rate": 4.2355007517057936e-05, "loss": 0.946, "step": 151050 }, { "epoch": 13.08, "learning_rate": 4.23528391349601e-05, "loss": 1.0449, "step": 151060 }, { "epoch": 13.08, "learning_rate": 4.235067075286226e-05, "loss": 0.9684, "step": 151070 }, { "epoch": 13.08, "learning_rate": 4.234850237076442e-05, "loss": 0.991, "step": 151080 }, { "epoch": 13.09, "learning_rate": 4.234633398866658e-05, "loss": 0.9385, "step": 151090 }, { "epoch": 13.09, "learning_rate": 4.234416560656875e-05, "loss": 1.0508, "step": 151100 }, { "epoch": 13.09, "learning_rate": 4.234199722447091e-05, "loss": 0.825, "step": 151110 }, { "epoch": 13.09, "learning_rate": 4.2339828842373074e-05, "loss": 0.9071, "step": 151120 }, { "epoch": 13.09, "learning_rate": 4.233766046027524e-05, "loss": 0.8129, "step": 151130 }, { "epoch": 13.09, "learning_rate": 4.23354920781774e-05, "loss": 0.8852, "step": 151140 }, { "epoch": 13.09, "learning_rate": 4.233332369607956e-05, "loss": 1.048, "step": 151150 }, { "epoch": 13.09, "learning_rate": 4.233115531398172e-05, "loss": 0.8784, "step": 151160 }, { "epoch": 13.09, "learning_rate": 4.2328986931883886e-05, "loss": 1.0838, "step": 151170 }, { "epoch": 13.09, "learning_rate": 4.2326818549786045e-05, "loss": 0.9447, "step": 151180 }, { "epoch": 13.09, "learning_rate": 4.232465016768821e-05, "loss": 1.0249, "step": 151190 }, { "epoch": 13.09, "learning_rate": 4.232248178559038e-05, "loss": 1.2611, "step": 151200 }, { "epoch": 13.1, "learning_rate": 4.232031340349254e-05, "loss": 0.8531, "step": 151210 }, { "epoch": 13.1, "learning_rate": 4.2318145021394704e-05, "loss": 1.0673, "step": 151220 }, { "epoch": 13.1, "learning_rate": 4.231597663929686e-05, "loss": 1.0557, "step": 151230 }, { "epoch": 13.1, "learning_rate": 4.231380825719902e-05, "loss": 1.135, "step": 151240 }, { "epoch": 13.1, "learning_rate": 4.231163987510118e-05, "loss": 0.8787, "step": 151250 }, { "epoch": 13.1, "learning_rate": 4.230947149300335e-05, "loss": 1.0229, "step": 151260 }, { "epoch": 13.1, "learning_rate": 4.2307303110905516e-05, "loss": 0.905, "step": 151270 }, { "epoch": 13.1, "learning_rate": 4.2305134728807675e-05, "loss": 0.9487, "step": 151280 }, { "epoch": 13.1, "learning_rate": 4.230296634670984e-05, "loss": 0.945, "step": 151290 }, { "epoch": 13.1, "learning_rate": 4.2300797964612e-05, "loss": 1.0646, "step": 151300 }, { "epoch": 13.1, "learning_rate": 4.229862958251416e-05, "loss": 0.8959, "step": 151310 }, { "epoch": 13.11, "learning_rate": 4.229646120041632e-05, "loss": 0.9631, "step": 151320 }, { "epoch": 13.11, "learning_rate": 4.229429281831849e-05, "loss": 0.8581, "step": 151330 }, { "epoch": 13.11, "learning_rate": 4.2292124436220653e-05, "loss": 0.8671, "step": 151340 }, { "epoch": 13.11, "learning_rate": 4.228995605412281e-05, "loss": 1.0416, "step": 151350 }, { "epoch": 13.11, "learning_rate": 4.228778767202498e-05, "loss": 1.0123, "step": 151360 }, { "epoch": 13.11, "learning_rate": 4.228561928992714e-05, "loss": 0.8198, "step": 151370 }, { "epoch": 13.11, "learning_rate": 4.22834509078293e-05, "loss": 1.0127, "step": 151380 }, { "epoch": 13.11, "learning_rate": 4.228128252573146e-05, "loss": 1.0254, "step": 151390 }, { "epoch": 13.11, "learning_rate": 4.2279114143633625e-05, "loss": 0.9428, "step": 151400 }, { "epoch": 13.11, "learning_rate": 4.227694576153579e-05, "loss": 1.1272, "step": 151410 }, { "epoch": 13.11, "learning_rate": 4.227477737943795e-05, "loss": 0.9318, "step": 151420 }, { "epoch": 13.11, "learning_rate": 4.227260899734012e-05, "loss": 0.9657, "step": 151430 }, { "epoch": 13.12, "learning_rate": 4.227044061524228e-05, "loss": 1.0193, "step": 151440 }, { "epoch": 13.12, "learning_rate": 4.226827223314444e-05, "loss": 1.2105, "step": 151450 }, { "epoch": 13.12, "learning_rate": 4.2266103851046596e-05, "loss": 1.0815, "step": 151460 }, { "epoch": 13.12, "learning_rate": 4.226393546894876e-05, "loss": 1.001, "step": 151470 }, { "epoch": 13.12, "learning_rate": 4.226176708685093e-05, "loss": 0.9328, "step": 151480 }, { "epoch": 13.12, "learning_rate": 4.225959870475309e-05, "loss": 0.9579, "step": 151490 }, { "epoch": 13.12, "learning_rate": 4.2257430322655255e-05, "loss": 0.8824, "step": 151500 }, { "epoch": 13.12, "learning_rate": 4.2255261940557414e-05, "loss": 0.9412, "step": 151510 }, { "epoch": 13.12, "learning_rate": 4.225309355845958e-05, "loss": 1.124, "step": 151520 }, { "epoch": 13.12, "learning_rate": 4.225092517636175e-05, "loss": 0.9472, "step": 151530 }, { "epoch": 13.12, "learning_rate": 4.22487567942639e-05, "loss": 0.8938, "step": 151540 }, { "epoch": 13.13, "learning_rate": 4.2246588412166066e-05, "loss": 0.9918, "step": 151550 }, { "epoch": 13.13, "learning_rate": 4.2244420030068226e-05, "loss": 0.8892, "step": 151560 }, { "epoch": 13.13, "learning_rate": 4.224225164797039e-05, "loss": 1.0006, "step": 151570 }, { "epoch": 13.13, "learning_rate": 4.224008326587255e-05, "loss": 1.0282, "step": 151580 }, { "epoch": 13.13, "learning_rate": 4.223791488377472e-05, "loss": 0.9191, "step": 151590 }, { "epoch": 13.13, "learning_rate": 4.2235746501676885e-05, "loss": 0.9495, "step": 151600 }, { "epoch": 13.13, "learning_rate": 4.2233578119579044e-05, "loss": 0.963, "step": 151610 }, { "epoch": 13.13, "learning_rate": 4.22314097374812e-05, "loss": 0.955, "step": 151620 }, { "epoch": 13.13, "learning_rate": 4.2229241355383364e-05, "loss": 1.1568, "step": 151630 }, { "epoch": 13.13, "learning_rate": 4.222707297328553e-05, "loss": 0.9986, "step": 151640 }, { "epoch": 13.13, "learning_rate": 4.222490459118769e-05, "loss": 0.9974, "step": 151650 }, { "epoch": 13.13, "learning_rate": 4.2222736209089856e-05, "loss": 1.121, "step": 151660 }, { "epoch": 13.14, "learning_rate": 4.2220567826992016e-05, "loss": 0.986, "step": 151670 }, { "epoch": 13.14, "learning_rate": 4.221839944489418e-05, "loss": 1.1628, "step": 151680 }, { "epoch": 13.14, "learning_rate": 4.2216231062796335e-05, "loss": 0.7959, "step": 151690 }, { "epoch": 13.14, "learning_rate": 4.22140626806985e-05, "loss": 1.1498, "step": 151700 }, { "epoch": 13.14, "learning_rate": 4.221189429860067e-05, "loss": 0.8869, "step": 151710 }, { "epoch": 13.14, "learning_rate": 4.220972591650283e-05, "loss": 1.0957, "step": 151720 }, { "epoch": 13.14, "learning_rate": 4.2207557534404994e-05, "loss": 1.2773, "step": 151730 }, { "epoch": 13.14, "learning_rate": 4.220538915230715e-05, "loss": 0.778, "step": 151740 }, { "epoch": 13.14, "learning_rate": 4.220322077020932e-05, "loss": 0.873, "step": 151750 }, { "epoch": 13.14, "learning_rate": 4.2201052388111486e-05, "loss": 0.9768, "step": 151760 }, { "epoch": 13.14, "learning_rate": 4.219888400601364e-05, "loss": 0.8859, "step": 151770 }, { "epoch": 13.15, "learning_rate": 4.2196715623915805e-05, "loss": 1.09, "step": 151780 }, { "epoch": 13.15, "learning_rate": 4.2194547241817965e-05, "loss": 1.1268, "step": 151790 }, { "epoch": 13.15, "learning_rate": 4.219237885972013e-05, "loss": 0.9751, "step": 151800 }, { "epoch": 13.15, "learning_rate": 4.219021047762229e-05, "loss": 1.0298, "step": 151810 }, { "epoch": 13.15, "learning_rate": 4.218804209552446e-05, "loss": 1.0789, "step": 151820 }, { "epoch": 13.15, "learning_rate": 4.2185873713426624e-05, "loss": 1.0668, "step": 151830 }, { "epoch": 13.15, "learning_rate": 4.218370533132878e-05, "loss": 0.9506, "step": 151840 }, { "epoch": 13.15, "learning_rate": 4.218153694923094e-05, "loss": 1.0448, "step": 151850 }, { "epoch": 13.15, "learning_rate": 4.21793685671331e-05, "loss": 1.1736, "step": 151860 }, { "epoch": 13.15, "learning_rate": 4.217720018503527e-05, "loss": 1.2133, "step": 151870 }, { "epoch": 13.15, "learning_rate": 4.217503180293743e-05, "loss": 0.9943, "step": 151880 }, { "epoch": 13.15, "learning_rate": 4.2172863420839595e-05, "loss": 0.954, "step": 151890 }, { "epoch": 13.16, "learning_rate": 4.217069503874176e-05, "loss": 0.9807, "step": 151900 }, { "epoch": 13.16, "learning_rate": 4.216852665664392e-05, "loss": 1.1194, "step": 151910 }, { "epoch": 13.16, "learning_rate": 4.216635827454608e-05, "loss": 1.0675, "step": 151920 }, { "epoch": 13.16, "learning_rate": 4.216418989244824e-05, "loss": 0.8896, "step": 151930 }, { "epoch": 13.16, "learning_rate": 4.2162021510350407e-05, "loss": 1.0094, "step": 151940 }, { "epoch": 13.16, "learning_rate": 4.2159853128252566e-05, "loss": 0.8478, "step": 151950 }, { "epoch": 13.16, "learning_rate": 4.215768474615473e-05, "loss": 1.0198, "step": 151960 }, { "epoch": 13.16, "learning_rate": 4.21555163640569e-05, "loss": 0.8977, "step": 151970 }, { "epoch": 13.16, "learning_rate": 4.215334798195906e-05, "loss": 1.0202, "step": 151980 }, { "epoch": 13.16, "learning_rate": 4.2151179599861225e-05, "loss": 0.9361, "step": 151990 }, { "epoch": 13.16, "learning_rate": 4.214901121776338e-05, "loss": 0.8853, "step": 152000 }, { "epoch": 13.17, "learning_rate": 4.2146842835665544e-05, "loss": 1.0139, "step": 152010 }, { "epoch": 13.17, "learning_rate": 4.2144674453567704e-05, "loss": 1.0882, "step": 152020 }, { "epoch": 13.17, "learning_rate": 4.214250607146987e-05, "loss": 1.0897, "step": 152030 }, { "epoch": 13.17, "learning_rate": 4.2140337689372037e-05, "loss": 1.089, "step": 152040 }, { "epoch": 13.17, "learning_rate": 4.2138169307274196e-05, "loss": 1.0595, "step": 152050 }, { "epoch": 13.17, "learning_rate": 4.213600092517636e-05, "loss": 0.918, "step": 152060 }, { "epoch": 13.17, "learning_rate": 4.213383254307852e-05, "loss": 0.8495, "step": 152070 }, { "epoch": 13.17, "learning_rate": 4.213166416098068e-05, "loss": 0.9812, "step": 152080 }, { "epoch": 13.17, "learning_rate": 4.212949577888284e-05, "loss": 1.0013, "step": 152090 }, { "epoch": 13.17, "learning_rate": 4.212732739678501e-05, "loss": 1.0617, "step": 152100 }, { "epoch": 13.17, "learning_rate": 4.2125159014687174e-05, "loss": 0.9591, "step": 152110 }, { "epoch": 13.17, "learning_rate": 4.2122990632589334e-05, "loss": 1.0542, "step": 152120 }, { "epoch": 13.18, "learning_rate": 4.21208222504915e-05, "loss": 0.9617, "step": 152130 }, { "epoch": 13.18, "learning_rate": 4.211865386839366e-05, "loss": 1.0743, "step": 152140 }, { "epoch": 13.18, "learning_rate": 4.211648548629582e-05, "loss": 1.0758, "step": 152150 }, { "epoch": 13.18, "learning_rate": 4.211431710419798e-05, "loss": 1.0536, "step": 152160 }, { "epoch": 13.18, "learning_rate": 4.2112148722100145e-05, "loss": 0.9445, "step": 152170 }, { "epoch": 13.18, "learning_rate": 4.2109980340002305e-05, "loss": 1.0613, "step": 152180 }, { "epoch": 13.18, "learning_rate": 4.210781195790447e-05, "loss": 0.9748, "step": 152190 }, { "epoch": 13.18, "learning_rate": 4.210564357580664e-05, "loss": 0.9647, "step": 152200 }, { "epoch": 13.18, "learning_rate": 4.21034751937088e-05, "loss": 0.9995, "step": 152210 }, { "epoch": 13.18, "learning_rate": 4.2101306811610964e-05, "loss": 0.9658, "step": 152220 }, { "epoch": 13.18, "learning_rate": 4.209913842951312e-05, "loss": 1.116, "step": 152230 }, { "epoch": 13.18, "learning_rate": 4.209697004741528e-05, "loss": 0.9688, "step": 152240 }, { "epoch": 13.19, "learning_rate": 4.209480166531744e-05, "loss": 1.0311, "step": 152250 }, { "epoch": 13.19, "learning_rate": 4.209263328321961e-05, "loss": 1.0229, "step": 152260 }, { "epoch": 13.19, "learning_rate": 4.2090464901121775e-05, "loss": 0.9665, "step": 152270 }, { "epoch": 13.19, "learning_rate": 4.2088296519023935e-05, "loss": 0.9744, "step": 152280 }, { "epoch": 13.19, "learning_rate": 4.20861281369261e-05, "loss": 1.081, "step": 152290 }, { "epoch": 13.19, "learning_rate": 4.208395975482826e-05, "loss": 1.0705, "step": 152300 }, { "epoch": 13.19, "learning_rate": 4.208179137273042e-05, "loss": 1.1024, "step": 152310 }, { "epoch": 13.19, "learning_rate": 4.207962299063258e-05, "loss": 0.9828, "step": 152320 }, { "epoch": 13.19, "learning_rate": 4.207745460853475e-05, "loss": 1.0017, "step": 152330 }, { "epoch": 13.19, "learning_rate": 4.207528622643691e-05, "loss": 1.0623, "step": 152340 }, { "epoch": 13.19, "learning_rate": 4.207311784433907e-05, "loss": 1.0402, "step": 152350 }, { "epoch": 13.2, "learning_rate": 4.207094946224124e-05, "loss": 0.9541, "step": 152360 }, { "epoch": 13.2, "learning_rate": 4.20687810801434e-05, "loss": 0.9587, "step": 152370 }, { "epoch": 13.2, "learning_rate": 4.2066612698045565e-05, "loss": 0.988, "step": 152380 }, { "epoch": 13.2, "learning_rate": 4.206444431594772e-05, "loss": 1.0623, "step": 152390 }, { "epoch": 13.2, "learning_rate": 4.2062275933849884e-05, "loss": 0.9963, "step": 152400 }, { "epoch": 13.2, "learning_rate": 4.206010755175205e-05, "loss": 0.9469, "step": 152410 }, { "epoch": 13.2, "learning_rate": 4.205793916965421e-05, "loss": 1.0413, "step": 152420 }, { "epoch": 13.2, "learning_rate": 4.205577078755638e-05, "loss": 0.9499, "step": 152430 }, { "epoch": 13.2, "learning_rate": 4.2053602405458536e-05, "loss": 0.9757, "step": 152440 }, { "epoch": 13.2, "learning_rate": 4.20514340233607e-05, "loss": 1.0725, "step": 152450 }, { "epoch": 13.2, "learning_rate": 4.2049265641262856e-05, "loss": 1.0029, "step": 152460 }, { "epoch": 13.2, "learning_rate": 4.204709725916502e-05, "loss": 0.9995, "step": 152470 }, { "epoch": 13.21, "learning_rate": 4.204492887706719e-05, "loss": 0.9789, "step": 152480 }, { "epoch": 13.21, "learning_rate": 4.204276049496935e-05, "loss": 0.9483, "step": 152490 }, { "epoch": 13.21, "learning_rate": 4.2040592112871514e-05, "loss": 0.9286, "step": 152500 }, { "epoch": 13.21, "learning_rate": 4.2038423730773674e-05, "loss": 0.7783, "step": 152510 }, { "epoch": 13.21, "learning_rate": 4.203625534867584e-05, "loss": 1.1013, "step": 152520 }, { "epoch": 13.21, "learning_rate": 4.203408696657801e-05, "loss": 1.0377, "step": 152530 }, { "epoch": 13.21, "learning_rate": 4.203191858448016e-05, "loss": 0.9098, "step": 152540 }, { "epoch": 13.21, "learning_rate": 4.2029750202382326e-05, "loss": 1.0363, "step": 152550 }, { "epoch": 13.21, "learning_rate": 4.2027581820284486e-05, "loss": 1.0362, "step": 152560 }, { "epoch": 13.21, "learning_rate": 4.202541343818665e-05, "loss": 0.8627, "step": 152570 }, { "epoch": 13.21, "learning_rate": 4.202324505608881e-05, "loss": 0.9487, "step": 152580 }, { "epoch": 13.22, "learning_rate": 4.202107667399098e-05, "loss": 0.8983, "step": 152590 }, { "epoch": 13.22, "learning_rate": 4.2018908291893144e-05, "loss": 0.912, "step": 152600 }, { "epoch": 13.22, "learning_rate": 4.2016739909795304e-05, "loss": 0.9427, "step": 152610 }, { "epoch": 13.22, "learning_rate": 4.2014571527697464e-05, "loss": 0.9643, "step": 152620 }, { "epoch": 13.22, "learning_rate": 4.201240314559962e-05, "loss": 0.9763, "step": 152630 }, { "epoch": 13.22, "learning_rate": 4.201023476350179e-05, "loss": 0.9078, "step": 152640 }, { "epoch": 13.22, "learning_rate": 4.200806638140395e-05, "loss": 0.8733, "step": 152650 }, { "epoch": 13.22, "learning_rate": 4.2005897999306116e-05, "loss": 0.8991, "step": 152660 }, { "epoch": 13.22, "learning_rate": 4.200372961720828e-05, "loss": 1.0669, "step": 152670 }, { "epoch": 13.22, "learning_rate": 4.200156123511044e-05, "loss": 0.8975, "step": 152680 }, { "epoch": 13.22, "learning_rate": 4.1999392853012595e-05, "loss": 0.9466, "step": 152690 }, { "epoch": 13.22, "learning_rate": 4.199722447091476e-05, "loss": 1.0608, "step": 152700 }, { "epoch": 13.23, "learning_rate": 4.199505608881693e-05, "loss": 1.0697, "step": 152710 }, { "epoch": 13.23, "learning_rate": 4.199288770671909e-05, "loss": 1.0102, "step": 152720 }, { "epoch": 13.23, "learning_rate": 4.199071932462125e-05, "loss": 1.0178, "step": 152730 }, { "epoch": 13.23, "learning_rate": 4.198855094252341e-05, "loss": 1.0264, "step": 152740 }, { "epoch": 13.23, "learning_rate": 4.198638256042558e-05, "loss": 1.2696, "step": 152750 }, { "epoch": 13.23, "learning_rate": 4.1984214178327746e-05, "loss": 1.1572, "step": 152760 }, { "epoch": 13.23, "learning_rate": 4.19820457962299e-05, "loss": 0.9493, "step": 152770 }, { "epoch": 13.23, "learning_rate": 4.1979877414132065e-05, "loss": 0.9242, "step": 152780 }, { "epoch": 13.23, "learning_rate": 4.1977709032034225e-05, "loss": 1.1592, "step": 152790 }, { "epoch": 13.23, "learning_rate": 4.197554064993639e-05, "loss": 1.1328, "step": 152800 }, { "epoch": 13.23, "learning_rate": 4.197337226783855e-05, "loss": 1.138, "step": 152810 }, { "epoch": 13.24, "learning_rate": 4.197120388574072e-05, "loss": 0.9687, "step": 152820 }, { "epoch": 13.24, "learning_rate": 4.196903550364288e-05, "loss": 0.9983, "step": 152830 }, { "epoch": 13.24, "learning_rate": 4.196686712154504e-05, "loss": 0.9304, "step": 152840 }, { "epoch": 13.24, "learning_rate": 4.19646987394472e-05, "loss": 0.9839, "step": 152850 }, { "epoch": 13.24, "learning_rate": 4.196253035734936e-05, "loss": 0.8816, "step": 152860 }, { "epoch": 13.24, "learning_rate": 4.196036197525153e-05, "loss": 0.8902, "step": 152870 }, { "epoch": 13.24, "learning_rate": 4.195819359315369e-05, "loss": 1.039, "step": 152880 }, { "epoch": 13.24, "learning_rate": 4.1956025211055855e-05, "loss": 0.8747, "step": 152890 }, { "epoch": 13.24, "learning_rate": 4.195385682895802e-05, "loss": 0.8758, "step": 152900 }, { "epoch": 13.24, "learning_rate": 4.195168844686018e-05, "loss": 0.8848, "step": 152910 }, { "epoch": 13.24, "learning_rate": 4.194952006476234e-05, "loss": 1.0129, "step": 152920 }, { "epoch": 13.24, "learning_rate": 4.19473516826645e-05, "loss": 0.9934, "step": 152930 }, { "epoch": 13.25, "learning_rate": 4.1945183300566666e-05, "loss": 1.0163, "step": 152940 }, { "epoch": 13.25, "learning_rate": 4.1943014918468826e-05, "loss": 0.9674, "step": 152950 }, { "epoch": 13.25, "learning_rate": 4.194084653637099e-05, "loss": 0.8335, "step": 152960 }, { "epoch": 13.25, "learning_rate": 4.193867815427316e-05, "loss": 1.0536, "step": 152970 }, { "epoch": 13.25, "learning_rate": 4.193650977217532e-05, "loss": 0.9395, "step": 152980 }, { "epoch": 13.25, "learning_rate": 4.1934341390077485e-05, "loss": 1.0822, "step": 152990 }, { "epoch": 13.25, "learning_rate": 4.193217300797964e-05, "loss": 1.054, "step": 153000 }, { "epoch": 13.25, "learning_rate": 4.1930004625881804e-05, "loss": 0.9999, "step": 153010 }, { "epoch": 13.25, "learning_rate": 4.1927836243783963e-05, "loss": 0.8338, "step": 153020 }, { "epoch": 13.25, "learning_rate": 4.192566786168613e-05, "loss": 1.0418, "step": 153030 }, { "epoch": 13.25, "learning_rate": 4.1923499479588296e-05, "loss": 0.9017, "step": 153040 }, { "epoch": 13.26, "learning_rate": 4.1921331097490456e-05, "loss": 0.997, "step": 153050 }, { "epoch": 13.26, "learning_rate": 4.191916271539262e-05, "loss": 0.9285, "step": 153060 }, { "epoch": 13.26, "learning_rate": 4.191699433329478e-05, "loss": 1.0249, "step": 153070 }, { "epoch": 13.26, "learning_rate": 4.191482595119694e-05, "loss": 1.0698, "step": 153080 }, { "epoch": 13.26, "learning_rate": 4.19126575690991e-05, "loss": 0.9788, "step": 153090 }, { "epoch": 13.26, "learning_rate": 4.191048918700127e-05, "loss": 1.0273, "step": 153100 }, { "epoch": 13.26, "learning_rate": 4.1908320804903434e-05, "loss": 0.8802, "step": 153110 }, { "epoch": 13.26, "learning_rate": 4.1906152422805593e-05, "loss": 0.8392, "step": 153120 }, { "epoch": 13.26, "learning_rate": 4.190398404070776e-05, "loss": 0.9583, "step": 153130 }, { "epoch": 13.26, "learning_rate": 4.190181565860992e-05, "loss": 0.9255, "step": 153140 }, { "epoch": 13.26, "learning_rate": 4.1899647276512086e-05, "loss": 0.8322, "step": 153150 }, { "epoch": 13.26, "learning_rate": 4.189747889441424e-05, "loss": 1.0028, "step": 153160 }, { "epoch": 13.27, "learning_rate": 4.1895310512316405e-05, "loss": 1.2523, "step": 153170 }, { "epoch": 13.27, "learning_rate": 4.189314213021857e-05, "loss": 0.9935, "step": 153180 }, { "epoch": 13.27, "learning_rate": 4.189097374812073e-05, "loss": 0.994, "step": 153190 }, { "epoch": 13.27, "learning_rate": 4.18888053660229e-05, "loss": 0.9814, "step": 153200 }, { "epoch": 13.27, "learning_rate": 4.188663698392506e-05, "loss": 1.0033, "step": 153210 }, { "epoch": 13.27, "learning_rate": 4.1884468601827224e-05, "loss": 1.1048, "step": 153220 }, { "epoch": 13.27, "learning_rate": 4.1882300219729376e-05, "loss": 1.1136, "step": 153230 }, { "epoch": 13.27, "learning_rate": 4.188013183763154e-05, "loss": 0.978, "step": 153240 }, { "epoch": 13.27, "learning_rate": 4.18779634555337e-05, "loss": 0.9993, "step": 153250 }, { "epoch": 13.27, "learning_rate": 4.187579507343587e-05, "loss": 1.1658, "step": 153260 }, { "epoch": 13.27, "learning_rate": 4.1873626691338035e-05, "loss": 0.9676, "step": 153270 }, { "epoch": 13.28, "learning_rate": 4.1871458309240195e-05, "loss": 0.8885, "step": 153280 }, { "epoch": 13.28, "learning_rate": 4.186928992714236e-05, "loss": 0.9335, "step": 153290 }, { "epoch": 13.28, "learning_rate": 4.186712154504452e-05, "loss": 1.0998, "step": 153300 }, { "epoch": 13.28, "learning_rate": 4.186495316294668e-05, "loss": 1.0727, "step": 153310 }, { "epoch": 13.28, "learning_rate": 4.186278478084884e-05, "loss": 1.0775, "step": 153320 }, { "epoch": 13.28, "learning_rate": 4.1860616398751006e-05, "loss": 1.0564, "step": 153330 }, { "epoch": 13.28, "learning_rate": 4.185844801665317e-05, "loss": 0.898, "step": 153340 }, { "epoch": 13.28, "learning_rate": 4.185627963455533e-05, "loss": 1.1054, "step": 153350 }, { "epoch": 13.28, "learning_rate": 4.18541112524575e-05, "loss": 0.8846, "step": 153360 }, { "epoch": 13.28, "learning_rate": 4.185194287035966e-05, "loss": 1.0981, "step": 153370 }, { "epoch": 13.28, "learning_rate": 4.1849774488261825e-05, "loss": 0.8421, "step": 153380 }, { "epoch": 13.28, "learning_rate": 4.184760610616398e-05, "loss": 0.9825, "step": 153390 }, { "epoch": 13.29, "learning_rate": 4.1845437724066144e-05, "loss": 0.9459, "step": 153400 }, { "epoch": 13.29, "learning_rate": 4.184326934196831e-05, "loss": 1.0743, "step": 153410 }, { "epoch": 13.29, "learning_rate": 4.184110095987047e-05, "loss": 0.9649, "step": 153420 }, { "epoch": 13.29, "learning_rate": 4.1838932577772636e-05, "loss": 1.0314, "step": 153430 }, { "epoch": 13.29, "learning_rate": 4.1836764195674796e-05, "loss": 0.9425, "step": 153440 }, { "epoch": 13.29, "learning_rate": 4.183459581357696e-05, "loss": 1.0721, "step": 153450 }, { "epoch": 13.29, "learning_rate": 4.1832427431479115e-05, "loss": 1.0011, "step": 153460 }, { "epoch": 13.29, "learning_rate": 4.183025904938128e-05, "loss": 1.0221, "step": 153470 }, { "epoch": 13.29, "learning_rate": 4.182809066728345e-05, "loss": 0.9982, "step": 153480 }, { "epoch": 13.29, "learning_rate": 4.182592228518561e-05, "loss": 0.9547, "step": 153490 }, { "epoch": 13.29, "learning_rate": 4.1823753903087774e-05, "loss": 1.2089, "step": 153500 }, { "epoch": 13.29, "learning_rate": 4.1821585520989934e-05, "loss": 0.8615, "step": 153510 }, { "epoch": 13.3, "learning_rate": 4.18194171388921e-05, "loss": 1.0699, "step": 153520 }, { "epoch": 13.3, "learning_rate": 4.1817248756794266e-05, "loss": 1.0356, "step": 153530 }, { "epoch": 13.3, "learning_rate": 4.181508037469642e-05, "loss": 1.1065, "step": 153540 }, { "epoch": 13.3, "learning_rate": 4.1812911992598586e-05, "loss": 1.2263, "step": 153550 }, { "epoch": 13.3, "learning_rate": 4.1810743610500745e-05, "loss": 1.0628, "step": 153560 }, { "epoch": 13.3, "learning_rate": 4.180857522840291e-05, "loss": 0.8979, "step": 153570 }, { "epoch": 13.3, "learning_rate": 4.180640684630507e-05, "loss": 1.0256, "step": 153580 }, { "epoch": 13.3, "learning_rate": 4.180423846420724e-05, "loss": 1.1764, "step": 153590 }, { "epoch": 13.3, "learning_rate": 4.1802070082109404e-05, "loss": 0.9633, "step": 153600 }, { "epoch": 13.3, "learning_rate": 4.1799901700011564e-05, "loss": 1.097, "step": 153610 }, { "epoch": 13.3, "learning_rate": 4.179773331791372e-05, "loss": 0.8905, "step": 153620 }, { "epoch": 13.31, "learning_rate": 4.179556493581588e-05, "loss": 1.1469, "step": 153630 }, { "epoch": 13.31, "learning_rate": 4.179339655371805e-05, "loss": 1.0186, "step": 153640 }, { "epoch": 13.31, "learning_rate": 4.179122817162021e-05, "loss": 0.9042, "step": 153650 }, { "epoch": 13.31, "learning_rate": 4.1789059789522375e-05, "loss": 1.1285, "step": 153660 }, { "epoch": 13.31, "learning_rate": 4.178689140742454e-05, "loss": 0.9909, "step": 153670 }, { "epoch": 13.31, "learning_rate": 4.17847230253267e-05, "loss": 0.7888, "step": 153680 }, { "epoch": 13.31, "learning_rate": 4.178255464322886e-05, "loss": 1.1684, "step": 153690 }, { "epoch": 13.31, "learning_rate": 4.178038626113102e-05, "loss": 0.9482, "step": 153700 }, { "epoch": 13.31, "learning_rate": 4.177821787903319e-05, "loss": 0.9389, "step": 153710 }, { "epoch": 13.31, "learning_rate": 4.1776049496935347e-05, "loss": 0.8859, "step": 153720 }, { "epoch": 13.31, "learning_rate": 4.177388111483751e-05, "loss": 0.773, "step": 153730 }, { "epoch": 13.31, "learning_rate": 4.177171273273968e-05, "loss": 1.1886, "step": 153740 }, { "epoch": 13.32, "learning_rate": 4.176954435064184e-05, "loss": 0.9263, "step": 153750 }, { "epoch": 13.32, "learning_rate": 4.1767375968544005e-05, "loss": 0.9775, "step": 153760 }, { "epoch": 13.32, "learning_rate": 4.176520758644616e-05, "loss": 1.0555, "step": 153770 }, { "epoch": 13.32, "learning_rate": 4.1763039204348325e-05, "loss": 0.9607, "step": 153780 }, { "epoch": 13.32, "learning_rate": 4.1760870822250484e-05, "loss": 1.0971, "step": 153790 }, { "epoch": 13.32, "learning_rate": 4.175870244015265e-05, "loss": 0.9405, "step": 153800 }, { "epoch": 13.32, "learning_rate": 4.175653405805481e-05, "loss": 1.0245, "step": 153810 }, { "epoch": 13.32, "learning_rate": 4.175436567595698e-05, "loss": 1.1442, "step": 153820 }, { "epoch": 13.32, "learning_rate": 4.175219729385914e-05, "loss": 0.9576, "step": 153830 }, { "epoch": 13.32, "learning_rate": 4.17500289117613e-05, "loss": 0.9184, "step": 153840 }, { "epoch": 13.32, "learning_rate": 4.174786052966346e-05, "loss": 1.1066, "step": 153850 }, { "epoch": 13.33, "learning_rate": 4.174569214756562e-05, "loss": 1.077, "step": 153860 }, { "epoch": 13.33, "learning_rate": 4.174352376546779e-05, "loss": 0.8911, "step": 153870 }, { "epoch": 13.33, "learning_rate": 4.174135538336995e-05, "loss": 1.0274, "step": 153880 }, { "epoch": 13.33, "learning_rate": 4.1739187001272114e-05, "loss": 1.1371, "step": 153890 }, { "epoch": 13.33, "learning_rate": 4.173701861917428e-05, "loss": 1.0346, "step": 153900 }, { "epoch": 13.33, "learning_rate": 4.173485023707644e-05, "loss": 0.8744, "step": 153910 }, { "epoch": 13.33, "learning_rate": 4.173268185497861e-05, "loss": 0.9865, "step": 153920 }, { "epoch": 13.33, "learning_rate": 4.173051347288076e-05, "loss": 1.2161, "step": 153930 }, { "epoch": 13.33, "learning_rate": 4.1728345090782926e-05, "loss": 0.9772, "step": 153940 }, { "epoch": 13.33, "learning_rate": 4.1726176708685086e-05, "loss": 0.8201, "step": 153950 }, { "epoch": 13.33, "learning_rate": 4.172400832658725e-05, "loss": 0.9715, "step": 153960 }, { "epoch": 13.33, "learning_rate": 4.172183994448942e-05, "loss": 0.9244, "step": 153970 }, { "epoch": 13.34, "learning_rate": 4.171967156239158e-05, "loss": 0.8847, "step": 153980 }, { "epoch": 13.34, "learning_rate": 4.1717503180293744e-05, "loss": 1.0097, "step": 153990 }, { "epoch": 13.34, "learning_rate": 4.17153347981959e-05, "loss": 1.0403, "step": 154000 }, { "epoch": 13.34, "learning_rate": 4.1713166416098064e-05, "loss": 1.1322, "step": 154010 }, { "epoch": 13.34, "learning_rate": 4.171099803400022e-05, "loss": 0.9309, "step": 154020 }, { "epoch": 13.34, "learning_rate": 4.170882965190239e-05, "loss": 0.9914, "step": 154030 }, { "epoch": 13.34, "learning_rate": 4.1706661269804556e-05, "loss": 0.9052, "step": 154040 }, { "epoch": 13.34, "learning_rate": 4.1704492887706716e-05, "loss": 1.1416, "step": 154050 }, { "epoch": 13.34, "learning_rate": 4.170232450560888e-05, "loss": 0.9464, "step": 154060 }, { "epoch": 13.34, "learning_rate": 4.170015612351104e-05, "loss": 1.1493, "step": 154070 }, { "epoch": 13.34, "learning_rate": 4.16979877414132e-05, "loss": 1.0224, "step": 154080 }, { "epoch": 13.35, "learning_rate": 4.169581935931536e-05, "loss": 1.0429, "step": 154090 }, { "epoch": 13.35, "learning_rate": 4.169365097721753e-05, "loss": 1.1516, "step": 154100 }, { "epoch": 13.35, "learning_rate": 4.1691482595119694e-05, "loss": 0.8609, "step": 154110 }, { "epoch": 13.35, "learning_rate": 4.168931421302185e-05, "loss": 1.1464, "step": 154120 }, { "epoch": 13.35, "learning_rate": 4.168714583092402e-05, "loss": 0.9616, "step": 154130 }, { "epoch": 13.35, "learning_rate": 4.168497744882618e-05, "loss": 0.9072, "step": 154140 }, { "epoch": 13.35, "learning_rate": 4.1682809066728346e-05, "loss": 0.9899, "step": 154150 }, { "epoch": 13.35, "learning_rate": 4.16806406846305e-05, "loss": 0.7965, "step": 154160 }, { "epoch": 13.35, "learning_rate": 4.1678472302532665e-05, "loss": 1.0354, "step": 154170 }, { "epoch": 13.35, "learning_rate": 4.167630392043483e-05, "loss": 0.8554, "step": 154180 }, { "epoch": 13.35, "learning_rate": 4.167413553833699e-05, "loss": 1.0522, "step": 154190 }, { "epoch": 13.35, "learning_rate": 4.167196715623916e-05, "loss": 0.7951, "step": 154200 }, { "epoch": 13.36, "learning_rate": 4.166979877414132e-05, "loss": 0.9078, "step": 154210 }, { "epoch": 13.36, "learning_rate": 4.166763039204348e-05, "loss": 0.8709, "step": 154220 }, { "epoch": 13.36, "learning_rate": 4.1665462009945636e-05, "loss": 1.0005, "step": 154230 }, { "epoch": 13.36, "learning_rate": 4.16632936278478e-05, "loss": 1.1014, "step": 154240 }, { "epoch": 13.36, "learning_rate": 4.166112524574997e-05, "loss": 1.1365, "step": 154250 }, { "epoch": 13.36, "learning_rate": 4.165895686365213e-05, "loss": 0.9097, "step": 154260 }, { "epoch": 13.36, "learning_rate": 4.1656788481554295e-05, "loss": 1.0619, "step": 154270 }, { "epoch": 13.36, "learning_rate": 4.1654620099456454e-05, "loss": 0.8759, "step": 154280 }, { "epoch": 13.36, "learning_rate": 4.165245171735862e-05, "loss": 1.1084, "step": 154290 }, { "epoch": 13.36, "learning_rate": 4.165028333526079e-05, "loss": 0.9394, "step": 154300 }, { "epoch": 13.36, "learning_rate": 4.164811495316294e-05, "loss": 1.0124, "step": 154310 }, { "epoch": 13.37, "learning_rate": 4.16459465710651e-05, "loss": 1.0126, "step": 154320 }, { "epoch": 13.37, "learning_rate": 4.1643778188967266e-05, "loss": 1.0857, "step": 154330 }, { "epoch": 13.37, "learning_rate": 4.164160980686943e-05, "loss": 1.0553, "step": 154340 }, { "epoch": 13.37, "learning_rate": 4.163944142477159e-05, "loss": 0.8771, "step": 154350 }, { "epoch": 13.37, "learning_rate": 4.163727304267376e-05, "loss": 0.9349, "step": 154360 }, { "epoch": 13.37, "learning_rate": 4.163510466057592e-05, "loss": 0.9493, "step": 154370 }, { "epoch": 13.37, "learning_rate": 4.1632936278478085e-05, "loss": 0.8701, "step": 154380 }, { "epoch": 13.37, "learning_rate": 4.163076789638024e-05, "loss": 0.9565, "step": 154390 }, { "epoch": 13.37, "learning_rate": 4.1628599514282404e-05, "loss": 1.0843, "step": 154400 }, { "epoch": 13.37, "learning_rate": 4.162643113218457e-05, "loss": 1.0324, "step": 154410 }, { "epoch": 13.37, "learning_rate": 4.162426275008673e-05, "loss": 1.0372, "step": 154420 }, { "epoch": 13.37, "learning_rate": 4.1622094367988896e-05, "loss": 0.9808, "step": 154430 }, { "epoch": 13.38, "learning_rate": 4.1619925985891056e-05, "loss": 1.0793, "step": 154440 }, { "epoch": 13.38, "learning_rate": 4.161775760379322e-05, "loss": 1.1318, "step": 154450 }, { "epoch": 13.38, "learning_rate": 4.1615589221695375e-05, "loss": 1.1061, "step": 154460 }, { "epoch": 13.38, "learning_rate": 4.161342083959754e-05, "loss": 0.8629, "step": 154470 }, { "epoch": 13.38, "learning_rate": 4.161125245749971e-05, "loss": 0.8967, "step": 154480 }, { "epoch": 13.38, "learning_rate": 4.160908407540187e-05, "loss": 1.1566, "step": 154490 }, { "epoch": 13.38, "learning_rate": 4.1606915693304034e-05, "loss": 1.0274, "step": 154500 }, { "epoch": 13.38, "learning_rate": 4.160474731120619e-05, "loss": 0.981, "step": 154510 }, { "epoch": 13.38, "learning_rate": 4.160257892910836e-05, "loss": 0.9449, "step": 154520 }, { "epoch": 13.38, "learning_rate": 4.1600410547010526e-05, "loss": 1.0884, "step": 154530 }, { "epoch": 13.38, "learning_rate": 4.159824216491268e-05, "loss": 0.954, "step": 154540 }, { "epoch": 13.39, "learning_rate": 4.1596073782814845e-05, "loss": 0.8621, "step": 154550 }, { "epoch": 13.39, "learning_rate": 4.1593905400717005e-05, "loss": 0.9565, "step": 154560 }, { "epoch": 13.39, "learning_rate": 4.159173701861917e-05, "loss": 1.0001, "step": 154570 }, { "epoch": 13.39, "learning_rate": 4.158956863652133e-05, "loss": 1.0808, "step": 154580 }, { "epoch": 13.39, "learning_rate": 4.15874002544235e-05, "loss": 0.9874, "step": 154590 }, { "epoch": 13.39, "learning_rate": 4.1585231872325664e-05, "loss": 1.182, "step": 154600 }, { "epoch": 13.39, "learning_rate": 4.1583063490227823e-05, "loss": 0.9547, "step": 154610 }, { "epoch": 13.39, "learning_rate": 4.158089510812998e-05, "loss": 1.0013, "step": 154620 }, { "epoch": 13.39, "learning_rate": 4.157872672603214e-05, "loss": 0.8221, "step": 154630 }, { "epoch": 13.39, "learning_rate": 4.157655834393431e-05, "loss": 0.9841, "step": 154640 }, { "epoch": 13.39, "learning_rate": 4.157438996183647e-05, "loss": 1.0015, "step": 154650 }, { "epoch": 13.39, "learning_rate": 4.1572221579738635e-05, "loss": 0.9454, "step": 154660 }, { "epoch": 13.4, "learning_rate": 4.15700531976408e-05, "loss": 1.0173, "step": 154670 }, { "epoch": 13.4, "learning_rate": 4.156788481554296e-05, "loss": 0.9287, "step": 154680 }, { "epoch": 13.4, "learning_rate": 4.156571643344513e-05, "loss": 0.9927, "step": 154690 }, { "epoch": 13.4, "learning_rate": 4.156354805134728e-05, "loss": 0.9328, "step": 154700 }, { "epoch": 13.4, "learning_rate": 4.156137966924945e-05, "loss": 0.9433, "step": 154710 }, { "epoch": 13.4, "learning_rate": 4.1559211287151606e-05, "loss": 0.8074, "step": 154720 }, { "epoch": 13.4, "learning_rate": 4.155704290505377e-05, "loss": 0.9509, "step": 154730 }, { "epoch": 13.4, "learning_rate": 4.155487452295594e-05, "loss": 0.9163, "step": 154740 }, { "epoch": 13.4, "learning_rate": 4.15527061408581e-05, "loss": 0.9597, "step": 154750 }, { "epoch": 13.4, "learning_rate": 4.1550537758760265e-05, "loss": 1.142, "step": 154760 }, { "epoch": 13.4, "learning_rate": 4.154836937666242e-05, "loss": 0.9417, "step": 154770 }, { "epoch": 13.4, "learning_rate": 4.1546200994564584e-05, "loss": 0.9634, "step": 154780 }, { "epoch": 13.41, "learning_rate": 4.1544032612466744e-05, "loss": 1.022, "step": 154790 }, { "epoch": 13.41, "learning_rate": 4.154186423036891e-05, "loss": 0.9763, "step": 154800 }, { "epoch": 13.41, "learning_rate": 4.153969584827108e-05, "loss": 1.0536, "step": 154810 }, { "epoch": 13.41, "learning_rate": 4.1537527466173236e-05, "loss": 0.951, "step": 154820 }, { "epoch": 13.41, "learning_rate": 4.15353590840754e-05, "loss": 1.1134, "step": 154830 }, { "epoch": 13.41, "learning_rate": 4.153319070197756e-05, "loss": 1.1311, "step": 154840 }, { "epoch": 13.41, "learning_rate": 4.153102231987972e-05, "loss": 1.1007, "step": 154850 }, { "epoch": 13.41, "learning_rate": 4.152885393778188e-05, "loss": 0.9533, "step": 154860 }, { "epoch": 13.41, "learning_rate": 4.152668555568405e-05, "loss": 0.9089, "step": 154870 }, { "epoch": 13.41, "learning_rate": 4.152451717358621e-05, "loss": 0.9144, "step": 154880 }, { "epoch": 13.41, "learning_rate": 4.1522348791488374e-05, "loss": 1.1507, "step": 154890 }, { "epoch": 13.42, "learning_rate": 4.152018040939054e-05, "loss": 0.8409, "step": 154900 }, { "epoch": 13.42, "learning_rate": 4.15180120272927e-05, "loss": 1.1293, "step": 154910 }, { "epoch": 13.42, "learning_rate": 4.1515843645194866e-05, "loss": 1.0552, "step": 154920 }, { "epoch": 13.42, "learning_rate": 4.151367526309702e-05, "loss": 0.8967, "step": 154930 }, { "epoch": 13.42, "learning_rate": 4.1511506880999186e-05, "loss": 1.0115, "step": 154940 }, { "epoch": 13.42, "learning_rate": 4.1509338498901345e-05, "loss": 0.8879, "step": 154950 }, { "epoch": 13.42, "learning_rate": 4.150717011680351e-05, "loss": 1.0097, "step": 154960 }, { "epoch": 13.42, "learning_rate": 4.150500173470568e-05, "loss": 0.908, "step": 154970 }, { "epoch": 13.42, "learning_rate": 4.150283335260784e-05, "loss": 0.9804, "step": 154980 }, { "epoch": 13.42, "learning_rate": 4.1500664970510004e-05, "loss": 1.063, "step": 154990 }, { "epoch": 13.42, "learning_rate": 4.149849658841216e-05, "loss": 1.0966, "step": 155000 }, { "epoch": 13.42, "learning_rate": 4.149632820631432e-05, "loss": 1.1334, "step": 155010 }, { "epoch": 13.43, "learning_rate": 4.149415982421648e-05, "loss": 1.0293, "step": 155020 }, { "epoch": 13.43, "learning_rate": 4.149199144211865e-05, "loss": 0.9996, "step": 155030 }, { "epoch": 13.43, "learning_rate": 4.1489823060020816e-05, "loss": 1.0251, "step": 155040 }, { "epoch": 13.43, "learning_rate": 4.1487654677922975e-05, "loss": 1.0666, "step": 155050 }, { "epoch": 13.43, "learning_rate": 4.148548629582514e-05, "loss": 0.9793, "step": 155060 }, { "epoch": 13.43, "learning_rate": 4.14833179137273e-05, "loss": 1.0012, "step": 155070 }, { "epoch": 13.43, "learning_rate": 4.148114953162946e-05, "loss": 0.9681, "step": 155080 }, { "epoch": 13.43, "learning_rate": 4.147898114953162e-05, "loss": 1.0252, "step": 155090 }, { "epoch": 13.43, "learning_rate": 4.147681276743379e-05, "loss": 1.0856, "step": 155100 }, { "epoch": 13.43, "learning_rate": 4.147464438533595e-05, "loss": 1.0825, "step": 155110 }, { "epoch": 13.43, "learning_rate": 4.147247600323811e-05, "loss": 1.0118, "step": 155120 }, { "epoch": 13.44, "learning_rate": 4.147030762114028e-05, "loss": 0.9369, "step": 155130 }, { "epoch": 13.44, "learning_rate": 4.146813923904244e-05, "loss": 0.9111, "step": 155140 }, { "epoch": 13.44, "learning_rate": 4.1465970856944605e-05, "loss": 1.091, "step": 155150 }, { "epoch": 13.44, "learning_rate": 4.146380247484676e-05, "loss": 1.0143, "step": 155160 }, { "epoch": 13.44, "learning_rate": 4.1461634092748925e-05, "loss": 1.0311, "step": 155170 }, { "epoch": 13.44, "learning_rate": 4.145946571065109e-05, "loss": 1.02, "step": 155180 }, { "epoch": 13.44, "learning_rate": 4.145729732855325e-05, "loss": 0.9449, "step": 155190 }, { "epoch": 13.44, "learning_rate": 4.145512894645542e-05, "loss": 0.8484, "step": 155200 }, { "epoch": 13.44, "learning_rate": 4.1452960564357577e-05, "loss": 0.9599, "step": 155210 }, { "epoch": 13.44, "learning_rate": 4.145079218225974e-05, "loss": 1.0089, "step": 155220 }, { "epoch": 13.44, "learning_rate": 4.1448623800161896e-05, "loss": 0.9136, "step": 155230 }, { "epoch": 13.44, "learning_rate": 4.144645541806406e-05, "loss": 1.1483, "step": 155240 }, { "epoch": 13.45, "learning_rate": 4.144428703596623e-05, "loss": 1.1028, "step": 155250 }, { "epoch": 13.45, "learning_rate": 4.144211865386839e-05, "loss": 1.085, "step": 155260 }, { "epoch": 13.45, "learning_rate": 4.1439950271770555e-05, "loss": 1.0411, "step": 155270 }, { "epoch": 13.45, "learning_rate": 4.1437781889672714e-05, "loss": 0.9259, "step": 155280 }, { "epoch": 13.45, "learning_rate": 4.143561350757488e-05, "loss": 0.9695, "step": 155290 }, { "epoch": 13.45, "learning_rate": 4.143344512547705e-05, "loss": 1.1583, "step": 155300 }, { "epoch": 13.45, "learning_rate": 4.14312767433792e-05, "loss": 0.8324, "step": 155310 }, { "epoch": 13.45, "learning_rate": 4.1429108361281366e-05, "loss": 0.9793, "step": 155320 }, { "epoch": 13.45, "learning_rate": 4.1426939979183526e-05, "loss": 1.0505, "step": 155330 }, { "epoch": 13.45, "learning_rate": 4.142477159708569e-05, "loss": 1.0584, "step": 155340 }, { "epoch": 13.45, "learning_rate": 4.142260321498785e-05, "loss": 1.0401, "step": 155350 }, { "epoch": 13.46, "learning_rate": 4.142043483289002e-05, "loss": 1.0716, "step": 155360 }, { "epoch": 13.46, "learning_rate": 4.1418266450792185e-05, "loss": 0.8745, "step": 155370 }, { "epoch": 13.46, "learning_rate": 4.1416098068694344e-05, "loss": 1.1389, "step": 155380 }, { "epoch": 13.46, "learning_rate": 4.14139296865965e-05, "loss": 0.9786, "step": 155390 }, { "epoch": 13.46, "learning_rate": 4.1411761304498663e-05, "loss": 1.0479, "step": 155400 }, { "epoch": 13.46, "learning_rate": 4.140959292240083e-05, "loss": 0.9713, "step": 155410 }, { "epoch": 13.46, "learning_rate": 4.140742454030299e-05, "loss": 0.8918, "step": 155420 }, { "epoch": 13.46, "learning_rate": 4.1405256158205156e-05, "loss": 1.0404, "step": 155430 }, { "epoch": 13.46, "learning_rate": 4.1403087776107315e-05, "loss": 0.9514, "step": 155440 }, { "epoch": 13.46, "learning_rate": 4.140091939400948e-05, "loss": 0.9567, "step": 155450 }, { "epoch": 13.46, "learning_rate": 4.139875101191165e-05, "loss": 1.0062, "step": 155460 }, { "epoch": 13.46, "learning_rate": 4.13965826298138e-05, "loss": 0.9901, "step": 155470 }, { "epoch": 13.47, "learning_rate": 4.139441424771597e-05, "loss": 1.0451, "step": 155480 }, { "epoch": 13.47, "learning_rate": 4.139224586561813e-05, "loss": 0.9477, "step": 155490 }, { "epoch": 13.47, "learning_rate": 4.1390077483520293e-05, "loss": 0.8625, "step": 155500 }, { "epoch": 13.47, "learning_rate": 4.138790910142245e-05, "loss": 1.1299, "step": 155510 }, { "epoch": 13.47, "learning_rate": 4.138574071932462e-05, "loss": 1.0249, "step": 155520 }, { "epoch": 13.47, "learning_rate": 4.1383572337226786e-05, "loss": 1.0518, "step": 155530 }, { "epoch": 13.47, "learning_rate": 4.138140395512894e-05, "loss": 1.192, "step": 155540 }, { "epoch": 13.47, "learning_rate": 4.1379235573031105e-05, "loss": 0.9894, "step": 155550 }, { "epoch": 13.47, "learning_rate": 4.1377067190933265e-05, "loss": 0.889, "step": 155560 }, { "epoch": 13.47, "learning_rate": 4.137489880883543e-05, "loss": 1.0717, "step": 155570 }, { "epoch": 13.47, "learning_rate": 4.137273042673759e-05, "loss": 1.0098, "step": 155580 }, { "epoch": 13.48, "learning_rate": 4.137056204463976e-05, "loss": 1.2221, "step": 155590 }, { "epoch": 13.48, "learning_rate": 4.1368393662541923e-05, "loss": 0.9639, "step": 155600 }, { "epoch": 13.48, "learning_rate": 4.136622528044408e-05, "loss": 0.8202, "step": 155610 }, { "epoch": 13.48, "learning_rate": 4.136405689834624e-05, "loss": 1.0733, "step": 155620 }, { "epoch": 13.48, "learning_rate": 4.13618885162484e-05, "loss": 0.8121, "step": 155630 }, { "epoch": 13.48, "learning_rate": 4.135972013415057e-05, "loss": 0.9193, "step": 155640 }, { "epoch": 13.48, "learning_rate": 4.135755175205273e-05, "loss": 1.1515, "step": 155650 }, { "epoch": 13.48, "learning_rate": 4.1355383369954895e-05, "loss": 1.0553, "step": 155660 }, { "epoch": 13.48, "learning_rate": 4.135321498785706e-05, "loss": 1.1665, "step": 155670 }, { "epoch": 13.48, "learning_rate": 4.135104660575922e-05, "loss": 1.0571, "step": 155680 }, { "epoch": 13.48, "learning_rate": 4.134887822366139e-05, "loss": 1.039, "step": 155690 }, { "epoch": 13.48, "learning_rate": 4.134670984156354e-05, "loss": 1.0549, "step": 155700 }, { "epoch": 13.49, "learning_rate": 4.1344541459465706e-05, "loss": 0.9123, "step": 155710 }, { "epoch": 13.49, "learning_rate": 4.1342373077367866e-05, "loss": 0.978, "step": 155720 }, { "epoch": 13.49, "learning_rate": 4.134020469527003e-05, "loss": 0.9581, "step": 155730 }, { "epoch": 13.49, "learning_rate": 4.13380363131722e-05, "loss": 1.0487, "step": 155740 }, { "epoch": 13.49, "learning_rate": 4.133586793107436e-05, "loss": 1.0009, "step": 155750 }, { "epoch": 13.49, "learning_rate": 4.1333699548976525e-05, "loss": 0.955, "step": 155760 }, { "epoch": 13.49, "learning_rate": 4.133153116687868e-05, "loss": 1.1138, "step": 155770 }, { "epoch": 13.49, "learning_rate": 4.1329362784780844e-05, "loss": 1.0195, "step": 155780 }, { "epoch": 13.49, "learning_rate": 4.1327194402683004e-05, "loss": 0.8152, "step": 155790 }, { "epoch": 13.49, "learning_rate": 4.132502602058517e-05, "loss": 1.0151, "step": 155800 }, { "epoch": 13.49, "learning_rate": 4.1322857638487336e-05, "loss": 1.016, "step": 155810 }, { "epoch": 13.49, "learning_rate": 4.1320689256389496e-05, "loss": 1.097, "step": 155820 }, { "epoch": 13.5, "learning_rate": 4.131852087429166e-05, "loss": 1.0061, "step": 155830 }, { "epoch": 13.5, "learning_rate": 4.131635249219382e-05, "loss": 1.1128, "step": 155840 }, { "epoch": 13.5, "learning_rate": 4.131418411009598e-05, "loss": 0.9489, "step": 155850 }, { "epoch": 13.5, "learning_rate": 4.131201572799814e-05, "loss": 0.8996, "step": 155860 }, { "epoch": 13.5, "learning_rate": 4.130984734590031e-05, "loss": 1.1195, "step": 155870 }, { "epoch": 13.5, "learning_rate": 4.1307678963802474e-05, "loss": 1.0838, "step": 155880 }, { "epoch": 13.5, "learning_rate": 4.1305510581704634e-05, "loss": 1.0618, "step": 155890 }, { "epoch": 13.5, "learning_rate": 4.13033421996068e-05, "loss": 0.9418, "step": 155900 }, { "epoch": 13.5, "learning_rate": 4.130117381750896e-05, "loss": 1.064, "step": 155910 }, { "epoch": 13.5, "learning_rate": 4.1299005435411126e-05, "loss": 0.9339, "step": 155920 }, { "epoch": 13.5, "learning_rate": 4.129683705331328e-05, "loss": 0.9901, "step": 155930 }, { "epoch": 13.51, "learning_rate": 4.1294668671215445e-05, "loss": 1.1131, "step": 155940 }, { "epoch": 13.51, "learning_rate": 4.1292500289117605e-05, "loss": 1.0157, "step": 155950 }, { "epoch": 13.51, "learning_rate": 4.129033190701977e-05, "loss": 1.0054, "step": 155960 }, { "epoch": 13.51, "learning_rate": 4.128816352492194e-05, "loss": 0.9805, "step": 155970 }, { "epoch": 13.51, "learning_rate": 4.12859951428241e-05, "loss": 1.0646, "step": 155980 }, { "epoch": 13.51, "learning_rate": 4.1283826760726264e-05, "loss": 1.263, "step": 155990 }, { "epoch": 13.51, "learning_rate": 4.1281658378628417e-05, "loss": 1.0187, "step": 156000 }, { "epoch": 13.51, "learning_rate": 4.127948999653058e-05, "loss": 1.0046, "step": 156010 }, { "epoch": 13.51, "learning_rate": 4.127732161443274e-05, "loss": 0.9352, "step": 156020 }, { "epoch": 13.51, "learning_rate": 4.127515323233491e-05, "loss": 0.9902, "step": 156030 }, { "epoch": 13.51, "learning_rate": 4.1272984850237075e-05, "loss": 0.9619, "step": 156040 }, { "epoch": 13.51, "learning_rate": 4.1270816468139235e-05, "loss": 0.9038, "step": 156050 }, { "epoch": 13.52, "learning_rate": 4.12686480860414e-05, "loss": 0.9722, "step": 156060 }, { "epoch": 13.52, "learning_rate": 4.126647970394356e-05, "loss": 1.0885, "step": 156070 }, { "epoch": 13.52, "learning_rate": 4.126431132184572e-05, "loss": 1.2233, "step": 156080 }, { "epoch": 13.52, "learning_rate": 4.126214293974788e-05, "loss": 1.0288, "step": 156090 }, { "epoch": 13.52, "learning_rate": 4.1259974557650047e-05, "loss": 1.0122, "step": 156100 }, { "epoch": 13.52, "learning_rate": 4.125780617555221e-05, "loss": 0.8857, "step": 156110 }, { "epoch": 13.52, "learning_rate": 4.125563779345437e-05, "loss": 1.1049, "step": 156120 }, { "epoch": 13.52, "learning_rate": 4.125346941135654e-05, "loss": 0.9938, "step": 156130 }, { "epoch": 13.52, "learning_rate": 4.12513010292587e-05, "loss": 1.0878, "step": 156140 }, { "epoch": 13.52, "learning_rate": 4.1249132647160865e-05, "loss": 0.7469, "step": 156150 }, { "epoch": 13.52, "learning_rate": 4.124696426506302e-05, "loss": 1.0765, "step": 156160 }, { "epoch": 13.53, "learning_rate": 4.1244795882965184e-05, "loss": 0.8915, "step": 156170 }, { "epoch": 13.53, "learning_rate": 4.124262750086735e-05, "loss": 0.9693, "step": 156180 }, { "epoch": 13.53, "learning_rate": 4.124045911876951e-05, "loss": 0.9741, "step": 156190 }, { "epoch": 13.53, "learning_rate": 4.1238290736671677e-05, "loss": 1.0354, "step": 156200 }, { "epoch": 13.53, "learning_rate": 4.1236122354573836e-05, "loss": 1.033, "step": 156210 }, { "epoch": 13.53, "learning_rate": 4.1233953972476e-05, "loss": 1.1079, "step": 156220 }, { "epoch": 13.53, "learning_rate": 4.123178559037817e-05, "loss": 0.9293, "step": 156230 }, { "epoch": 13.53, "learning_rate": 4.122961720828032e-05, "loss": 1.0023, "step": 156240 }, { "epoch": 13.53, "learning_rate": 4.122744882618249e-05, "loss": 0.9992, "step": 156250 }, { "epoch": 13.53, "learning_rate": 4.122528044408465e-05, "loss": 1.0233, "step": 156260 }, { "epoch": 13.53, "learning_rate": 4.1223112061986814e-05, "loss": 0.917, "step": 156270 }, { "epoch": 13.53, "learning_rate": 4.1220943679888974e-05, "loss": 0.9544, "step": 156280 }, { "epoch": 13.54, "learning_rate": 4.121877529779114e-05, "loss": 1.11, "step": 156290 }, { "epoch": 13.54, "learning_rate": 4.121660691569331e-05, "loss": 0.9879, "step": 156300 }, { "epoch": 13.54, "learning_rate": 4.121443853359546e-05, "loss": 1.1215, "step": 156310 }, { "epoch": 13.54, "learning_rate": 4.1212270151497626e-05, "loss": 0.9441, "step": 156320 }, { "epoch": 13.54, "learning_rate": 4.1210101769399785e-05, "loss": 0.986, "step": 156330 }, { "epoch": 13.54, "learning_rate": 4.120793338730195e-05, "loss": 0.9912, "step": 156340 }, { "epoch": 13.54, "learning_rate": 4.120576500520411e-05, "loss": 0.8408, "step": 156350 }, { "epoch": 13.54, "learning_rate": 4.120359662310628e-05, "loss": 1.0873, "step": 156360 }, { "epoch": 13.54, "learning_rate": 4.1201428241008444e-05, "loss": 1.1547, "step": 156370 }, { "epoch": 13.54, "learning_rate": 4.1199259858910604e-05, "loss": 1.1376, "step": 156380 }, { "epoch": 13.54, "learning_rate": 4.1197091476812764e-05, "loss": 0.9744, "step": 156390 }, { "epoch": 13.55, "learning_rate": 4.119492309471492e-05, "loss": 1.0018, "step": 156400 }, { "epoch": 13.55, "learning_rate": 4.119275471261709e-05, "loss": 0.8271, "step": 156410 }, { "epoch": 13.55, "learning_rate": 4.119058633051925e-05, "loss": 0.9235, "step": 156420 }, { "epoch": 13.55, "learning_rate": 4.1188417948421416e-05, "loss": 1.1622, "step": 156430 }, { "epoch": 13.55, "learning_rate": 4.118624956632358e-05, "loss": 1.077, "step": 156440 }, { "epoch": 13.55, "learning_rate": 4.118408118422574e-05, "loss": 1.0005, "step": 156450 }, { "epoch": 13.55, "learning_rate": 4.118191280212791e-05, "loss": 1.078, "step": 156460 }, { "epoch": 13.55, "learning_rate": 4.117974442003006e-05, "loss": 0.9288, "step": 156470 }, { "epoch": 13.55, "learning_rate": 4.117757603793223e-05, "loss": 1.074, "step": 156480 }, { "epoch": 13.55, "learning_rate": 4.117540765583439e-05, "loss": 0.99, "step": 156490 }, { "epoch": 13.55, "learning_rate": 4.117323927373655e-05, "loss": 0.9738, "step": 156500 }, { "epoch": 13.55, "learning_rate": 4.117107089163871e-05, "loss": 1.17, "step": 156510 }, { "epoch": 13.56, "learning_rate": 4.116890250954088e-05, "loss": 1.0609, "step": 156520 }, { "epoch": 13.56, "learning_rate": 4.1166734127443046e-05, "loss": 1.1156, "step": 156530 }, { "epoch": 13.56, "learning_rate": 4.11645657453452e-05, "loss": 0.9073, "step": 156540 }, { "epoch": 13.56, "learning_rate": 4.1162397363247365e-05, "loss": 0.9946, "step": 156550 }, { "epoch": 13.56, "learning_rate": 4.1160228981149524e-05, "loss": 0.9367, "step": 156560 }, { "epoch": 13.56, "learning_rate": 4.115806059905169e-05, "loss": 0.9313, "step": 156570 }, { "epoch": 13.56, "learning_rate": 4.115589221695385e-05, "loss": 1.0486, "step": 156580 }, { "epoch": 13.56, "learning_rate": 4.115372383485602e-05, "loss": 1.0543, "step": 156590 }, { "epoch": 13.56, "learning_rate": 4.115155545275818e-05, "loss": 1.08, "step": 156600 }, { "epoch": 13.56, "learning_rate": 4.114938707066034e-05, "loss": 1.0378, "step": 156610 }, { "epoch": 13.56, "learning_rate": 4.11472186885625e-05, "loss": 1.1796, "step": 156620 }, { "epoch": 13.57, "learning_rate": 4.114505030646466e-05, "loss": 1.1721, "step": 156630 }, { "epoch": 13.57, "learning_rate": 4.114288192436683e-05, "loss": 1.056, "step": 156640 }, { "epoch": 13.57, "learning_rate": 4.114071354226899e-05, "loss": 0.9369, "step": 156650 }, { "epoch": 13.57, "learning_rate": 4.1138545160171154e-05, "loss": 1.029, "step": 156660 }, { "epoch": 13.57, "learning_rate": 4.113637677807332e-05, "loss": 0.9126, "step": 156670 }, { "epoch": 13.57, "learning_rate": 4.113420839597548e-05, "loss": 1.1325, "step": 156680 }, { "epoch": 13.57, "learning_rate": 4.113204001387765e-05, "loss": 1.1214, "step": 156690 }, { "epoch": 13.57, "learning_rate": 4.11298716317798e-05, "loss": 1.0182, "step": 156700 }, { "epoch": 13.57, "learning_rate": 4.1127703249681966e-05, "loss": 0.8393, "step": 156710 }, { "epoch": 13.57, "learning_rate": 4.1125534867584126e-05, "loss": 0.9628, "step": 156720 }, { "epoch": 13.57, "learning_rate": 4.112336648548629e-05, "loss": 0.9492, "step": 156730 }, { "epoch": 13.57, "learning_rate": 4.112119810338846e-05, "loss": 1.0296, "step": 156740 }, { "epoch": 13.58, "learning_rate": 4.111902972129062e-05, "loss": 0.9632, "step": 156750 }, { "epoch": 13.58, "learning_rate": 4.1116861339192784e-05, "loss": 1.0642, "step": 156760 }, { "epoch": 13.58, "learning_rate": 4.111469295709494e-05, "loss": 0.9729, "step": 156770 }, { "epoch": 13.58, "learning_rate": 4.1112524574997104e-05, "loss": 1.3305, "step": 156780 }, { "epoch": 13.58, "learning_rate": 4.111035619289926e-05, "loss": 0.8948, "step": 156790 }, { "epoch": 13.58, "learning_rate": 4.110818781080143e-05, "loss": 1.237, "step": 156800 }, { "epoch": 13.58, "learning_rate": 4.1106019428703596e-05, "loss": 1.0189, "step": 156810 }, { "epoch": 13.58, "learning_rate": 4.1103851046605756e-05, "loss": 0.808, "step": 156820 }, { "epoch": 13.58, "learning_rate": 4.110168266450792e-05, "loss": 1.0481, "step": 156830 }, { "epoch": 13.58, "learning_rate": 4.109951428241008e-05, "loss": 0.9969, "step": 156840 }, { "epoch": 13.58, "learning_rate": 4.109734590031224e-05, "loss": 1.0597, "step": 156850 }, { "epoch": 13.59, "learning_rate": 4.10951775182144e-05, "loss": 0.9417, "step": 156860 }, { "epoch": 13.59, "learning_rate": 4.109300913611657e-05, "loss": 0.9889, "step": 156870 }, { "epoch": 13.59, "learning_rate": 4.1090840754018734e-05, "loss": 0.8649, "step": 156880 }, { "epoch": 13.59, "learning_rate": 4.108867237192089e-05, "loss": 0.989, "step": 156890 }, { "epoch": 13.59, "learning_rate": 4.108650398982306e-05, "loss": 1.0977, "step": 156900 }, { "epoch": 13.59, "learning_rate": 4.108433560772522e-05, "loss": 1.1174, "step": 156910 }, { "epoch": 13.59, "learning_rate": 4.1082167225627386e-05, "loss": 0.9767, "step": 156920 }, { "epoch": 13.59, "learning_rate": 4.107999884352954e-05, "loss": 1.0861, "step": 156930 }, { "epoch": 13.59, "learning_rate": 4.1077830461431705e-05, "loss": 1.0462, "step": 156940 }, { "epoch": 13.59, "learning_rate": 4.107566207933387e-05, "loss": 0.9698, "step": 156950 }, { "epoch": 13.59, "learning_rate": 4.107349369723603e-05, "loss": 1.0463, "step": 156960 }, { "epoch": 13.59, "learning_rate": 4.10713253151382e-05, "loss": 1.009, "step": 156970 }, { "epoch": 13.6, "learning_rate": 4.106915693304036e-05, "loss": 0.987, "step": 156980 }, { "epoch": 13.6, "learning_rate": 4.1066988550942523e-05, "loss": 0.9884, "step": 156990 }, { "epoch": 13.6, "learning_rate": 4.106482016884469e-05, "loss": 0.9026, "step": 157000 }, { "epoch": 13.6, "learning_rate": 4.106265178674684e-05, "loss": 0.9881, "step": 157010 }, { "epoch": 13.6, "learning_rate": 4.1060483404649e-05, "loss": 0.7972, "step": 157020 }, { "epoch": 13.6, "learning_rate": 4.105831502255117e-05, "loss": 1.0467, "step": 157030 }, { "epoch": 13.6, "learning_rate": 4.1056146640453335e-05, "loss": 1.0383, "step": 157040 }, { "epoch": 13.6, "learning_rate": 4.1053978258355495e-05, "loss": 1.0087, "step": 157050 }, { "epoch": 13.6, "learning_rate": 4.105180987625766e-05, "loss": 1.1445, "step": 157060 }, { "epoch": 13.6, "learning_rate": 4.104964149415982e-05, "loss": 1.123, "step": 157070 }, { "epoch": 13.6, "learning_rate": 4.104747311206198e-05, "loss": 1.1294, "step": 157080 }, { "epoch": 13.6, "learning_rate": 4.104530472996414e-05, "loss": 0.87, "step": 157090 }, { "epoch": 13.61, "learning_rate": 4.1043136347866306e-05, "loss": 1.0789, "step": 157100 }, { "epoch": 13.61, "learning_rate": 4.104096796576847e-05, "loss": 0.9868, "step": 157110 }, { "epoch": 13.61, "learning_rate": 4.103879958367063e-05, "loss": 1.1202, "step": 157120 }, { "epoch": 13.61, "learning_rate": 4.10366312015728e-05, "loss": 0.9107, "step": 157130 }, { "epoch": 13.61, "learning_rate": 4.103446281947496e-05, "loss": 1.0786, "step": 157140 }, { "epoch": 13.61, "learning_rate": 4.1032294437377125e-05, "loss": 1.0099, "step": 157150 }, { "epoch": 13.61, "learning_rate": 4.103012605527928e-05, "loss": 0.9998, "step": 157160 }, { "epoch": 13.61, "learning_rate": 4.1027957673181444e-05, "loss": 0.8408, "step": 157170 }, { "epoch": 13.61, "learning_rate": 4.102578929108361e-05, "loss": 1.2229, "step": 157180 }, { "epoch": 13.61, "learning_rate": 4.102362090898577e-05, "loss": 0.9503, "step": 157190 }, { "epoch": 13.61, "learning_rate": 4.1021452526887936e-05, "loss": 1.0834, "step": 157200 }, { "epoch": 13.62, "learning_rate": 4.1019284144790096e-05, "loss": 1.0986, "step": 157210 }, { "epoch": 13.62, "learning_rate": 4.101711576269226e-05, "loss": 1.0593, "step": 157220 }, { "epoch": 13.62, "learning_rate": 4.101494738059443e-05, "loss": 1.1789, "step": 157230 }, { "epoch": 13.62, "learning_rate": 4.101277899849658e-05, "loss": 1.208, "step": 157240 }, { "epoch": 13.62, "learning_rate": 4.101061061639875e-05, "loss": 0.9238, "step": 157250 }, { "epoch": 13.62, "learning_rate": 4.100844223430091e-05, "loss": 1.0191, "step": 157260 }, { "epoch": 13.62, "learning_rate": 4.1006273852203074e-05, "loss": 1.1832, "step": 157270 }, { "epoch": 13.62, "learning_rate": 4.1004105470105234e-05, "loss": 1.0352, "step": 157280 }, { "epoch": 13.62, "learning_rate": 4.10019370880074e-05, "loss": 0.9497, "step": 157290 }, { "epoch": 13.62, "learning_rate": 4.0999768705909566e-05, "loss": 1.0053, "step": 157300 }, { "epoch": 13.62, "learning_rate": 4.099760032381172e-05, "loss": 1.0613, "step": 157310 }, { "epoch": 13.62, "learning_rate": 4.0995431941713886e-05, "loss": 0.7439, "step": 157320 }, { "epoch": 13.63, "learning_rate": 4.0993263559616045e-05, "loss": 0.903, "step": 157330 }, { "epoch": 13.63, "learning_rate": 4.099109517751821e-05, "loss": 1.0661, "step": 157340 }, { "epoch": 13.63, "learning_rate": 4.098892679542037e-05, "loss": 0.8469, "step": 157350 }, { "epoch": 13.63, "learning_rate": 4.098675841332254e-05, "loss": 1.0091, "step": 157360 }, { "epoch": 13.63, "learning_rate": 4.0984590031224704e-05, "loss": 0.8849, "step": 157370 }, { "epoch": 13.63, "learning_rate": 4.0982421649126864e-05, "loss": 1.1923, "step": 157380 }, { "epoch": 13.63, "learning_rate": 4.098025326702902e-05, "loss": 1.0277, "step": 157390 }, { "epoch": 13.63, "learning_rate": 4.097808488493118e-05, "loss": 1.0132, "step": 157400 }, { "epoch": 13.63, "learning_rate": 4.097591650283335e-05, "loss": 0.9883, "step": 157410 }, { "epoch": 13.63, "learning_rate": 4.097374812073551e-05, "loss": 0.7912, "step": 157420 }, { "epoch": 13.63, "learning_rate": 4.0971579738637675e-05, "loss": 0.9764, "step": 157430 }, { "epoch": 13.64, "learning_rate": 4.096941135653984e-05, "loss": 0.8558, "step": 157440 }, { "epoch": 13.64, "learning_rate": 4.0967242974442e-05, "loss": 1.0973, "step": 157450 }, { "epoch": 13.64, "learning_rate": 4.096507459234417e-05, "loss": 1.0164, "step": 157460 }, { "epoch": 13.64, "learning_rate": 4.096290621024632e-05, "loss": 0.9482, "step": 157470 }, { "epoch": 13.64, "learning_rate": 4.096073782814849e-05, "loss": 1.0994, "step": 157480 }, { "epoch": 13.64, "learning_rate": 4.0958569446050646e-05, "loss": 1.0273, "step": 157490 }, { "epoch": 13.64, "learning_rate": 4.095640106395281e-05, "loss": 0.9972, "step": 157500 }, { "epoch": 13.64, "learning_rate": 4.095423268185498e-05, "loss": 0.9037, "step": 157510 }, { "epoch": 13.64, "learning_rate": 4.095206429975714e-05, "loss": 0.9346, "step": 157520 }, { "epoch": 13.64, "learning_rate": 4.0949895917659305e-05, "loss": 0.9815, "step": 157530 }, { "epoch": 13.64, "learning_rate": 4.094772753556146e-05, "loss": 1.0287, "step": 157540 }, { "epoch": 13.64, "learning_rate": 4.0945559153463624e-05, "loss": 0.8766, "step": 157550 }, { "epoch": 13.65, "learning_rate": 4.0943390771365784e-05, "loss": 0.9179, "step": 157560 }, { "epoch": 13.65, "learning_rate": 4.094122238926795e-05, "loss": 1.0773, "step": 157570 }, { "epoch": 13.65, "learning_rate": 4.093905400717011e-05, "loss": 0.9232, "step": 157580 }, { "epoch": 13.65, "learning_rate": 4.0936885625072276e-05, "loss": 1.0738, "step": 157590 }, { "epoch": 13.65, "learning_rate": 4.093471724297444e-05, "loss": 1.0071, "step": 157600 }, { "epoch": 13.65, "learning_rate": 4.09325488608766e-05, "loss": 0.8106, "step": 157610 }, { "epoch": 13.65, "learning_rate": 4.093038047877876e-05, "loss": 1.0277, "step": 157620 }, { "epoch": 13.65, "learning_rate": 4.092821209668092e-05, "loss": 1.0117, "step": 157630 }, { "epoch": 13.65, "learning_rate": 4.092604371458309e-05, "loss": 1.1221, "step": 157640 }, { "epoch": 13.65, "learning_rate": 4.092387533248525e-05, "loss": 1.1377, "step": 157650 }, { "epoch": 13.65, "learning_rate": 4.0921706950387414e-05, "loss": 1.0441, "step": 157660 }, { "epoch": 13.66, "learning_rate": 4.091953856828958e-05, "loss": 0.9201, "step": 157670 }, { "epoch": 13.66, "learning_rate": 4.091737018619174e-05, "loss": 1.0368, "step": 157680 }, { "epoch": 13.66, "learning_rate": 4.0915201804093907e-05, "loss": 1.1204, "step": 157690 }, { "epoch": 13.66, "learning_rate": 4.091303342199606e-05, "loss": 1.2726, "step": 157700 }, { "epoch": 13.66, "learning_rate": 4.0910865039898226e-05, "loss": 0.976, "step": 157710 }, { "epoch": 13.66, "learning_rate": 4.0908696657800385e-05, "loss": 0.9633, "step": 157720 }, { "epoch": 13.66, "learning_rate": 4.090652827570255e-05, "loss": 1.0649, "step": 157730 }, { "epoch": 13.66, "learning_rate": 4.090435989360472e-05, "loss": 1.0477, "step": 157740 }, { "epoch": 13.66, "learning_rate": 4.090219151150688e-05, "loss": 1.0375, "step": 157750 }, { "epoch": 13.66, "learning_rate": 4.0900023129409044e-05, "loss": 1.2223, "step": 157760 }, { "epoch": 13.66, "learning_rate": 4.0897854747311204e-05, "loss": 0.9943, "step": 157770 }, { "epoch": 13.66, "learning_rate": 4.0895686365213363e-05, "loss": 1.0912, "step": 157780 }, { "epoch": 13.67, "learning_rate": 4.089351798311552e-05, "loss": 0.8378, "step": 157790 }, { "epoch": 13.67, "learning_rate": 4.089134960101769e-05, "loss": 1.0545, "step": 157800 }, { "epoch": 13.67, "learning_rate": 4.0889181218919856e-05, "loss": 0.9734, "step": 157810 }, { "epoch": 13.67, "learning_rate": 4.0887012836822015e-05, "loss": 1.1246, "step": 157820 }, { "epoch": 13.67, "learning_rate": 4.088484445472418e-05, "loss": 1.0089, "step": 157830 }, { "epoch": 13.67, "learning_rate": 4.088267607262634e-05, "loss": 0.9437, "step": 157840 }, { "epoch": 13.67, "learning_rate": 4.08805076905285e-05, "loss": 0.8867, "step": 157850 }, { "epoch": 13.67, "learning_rate": 4.087833930843066e-05, "loss": 0.7897, "step": 157860 }, { "epoch": 13.67, "learning_rate": 4.087617092633283e-05, "loss": 1.1348, "step": 157870 }, { "epoch": 13.67, "learning_rate": 4.0874002544234993e-05, "loss": 1.0832, "step": 157880 }, { "epoch": 13.67, "learning_rate": 4.087183416213715e-05, "loss": 0.964, "step": 157890 }, { "epoch": 13.68, "learning_rate": 4.086966578003932e-05, "loss": 0.9318, "step": 157900 }, { "epoch": 13.68, "learning_rate": 4.086749739794148e-05, "loss": 0.8929, "step": 157910 }, { "epoch": 13.68, "learning_rate": 4.0865329015843645e-05, "loss": 0.9616, "step": 157920 }, { "epoch": 13.68, "learning_rate": 4.08631606337458e-05, "loss": 0.9482, "step": 157930 }, { "epoch": 13.68, "learning_rate": 4.0860992251647965e-05, "loss": 0.9571, "step": 157940 }, { "epoch": 13.68, "learning_rate": 4.085882386955013e-05, "loss": 0.9109, "step": 157950 }, { "epoch": 13.68, "learning_rate": 4.085665548745229e-05, "loss": 1.0162, "step": 157960 }, { "epoch": 13.68, "learning_rate": 4.085448710535446e-05, "loss": 0.9002, "step": 157970 }, { "epoch": 13.68, "learning_rate": 4.085231872325662e-05, "loss": 0.9158, "step": 157980 }, { "epoch": 13.68, "learning_rate": 4.085015034115878e-05, "loss": 0.9739, "step": 157990 }, { "epoch": 13.68, "learning_rate": 4.084798195906095e-05, "loss": 1.0657, "step": 158000 }, { "epoch": 13.68, "learning_rate": 4.08458135769631e-05, "loss": 0.9704, "step": 158010 }, { "epoch": 13.69, "learning_rate": 4.084364519486527e-05, "loss": 1.2843, "step": 158020 }, { "epoch": 13.69, "learning_rate": 4.084147681276743e-05, "loss": 1.029, "step": 158030 }, { "epoch": 13.69, "learning_rate": 4.0839308430669595e-05, "loss": 0.9423, "step": 158040 }, { "epoch": 13.69, "learning_rate": 4.0837140048571754e-05, "loss": 1.1373, "step": 158050 }, { "epoch": 13.69, "learning_rate": 4.083497166647392e-05, "loss": 0.9796, "step": 158060 }, { "epoch": 13.69, "learning_rate": 4.083280328437609e-05, "loss": 0.9767, "step": 158070 }, { "epoch": 13.69, "learning_rate": 4.083063490227824e-05, "loss": 1.045, "step": 158080 }, { "epoch": 13.69, "learning_rate": 4.08284665201804e-05, "loss": 0.9801, "step": 158090 }, { "epoch": 13.69, "learning_rate": 4.0826298138082566e-05, "loss": 0.9962, "step": 158100 }, { "epoch": 13.69, "learning_rate": 4.082412975598473e-05, "loss": 1.1392, "step": 158110 }, { "epoch": 13.69, "learning_rate": 4.082196137388689e-05, "loss": 0.8179, "step": 158120 }, { "epoch": 13.7, "learning_rate": 4.081979299178906e-05, "loss": 0.8751, "step": 158130 }, { "epoch": 13.7, "learning_rate": 4.081762460969122e-05, "loss": 0.9639, "step": 158140 }, { "epoch": 13.7, "learning_rate": 4.0815456227593384e-05, "loss": 1.1309, "step": 158150 }, { "epoch": 13.7, "learning_rate": 4.081328784549554e-05, "loss": 0.8145, "step": 158160 }, { "epoch": 13.7, "learning_rate": 4.0811119463397704e-05, "loss": 1.0805, "step": 158170 }, { "epoch": 13.7, "learning_rate": 4.080895108129987e-05, "loss": 1.0009, "step": 158180 }, { "epoch": 13.7, "learning_rate": 4.080678269920203e-05, "loss": 0.9684, "step": 158190 }, { "epoch": 13.7, "learning_rate": 4.0804614317104196e-05, "loss": 0.9285, "step": 158200 }, { "epoch": 13.7, "learning_rate": 4.0802445935006356e-05, "loss": 1.1139, "step": 158210 }, { "epoch": 13.7, "learning_rate": 4.080027755290852e-05, "loss": 1.0711, "step": 158220 }, { "epoch": 13.7, "learning_rate": 4.079810917081069e-05, "loss": 1.1684, "step": 158230 }, { "epoch": 13.7, "learning_rate": 4.079594078871284e-05, "loss": 1.0589, "step": 158240 }, { "epoch": 13.71, "learning_rate": 4.079377240661501e-05, "loss": 1.0456, "step": 158250 }, { "epoch": 13.71, "learning_rate": 4.079160402451717e-05, "loss": 1.1105, "step": 158260 }, { "epoch": 13.71, "learning_rate": 4.0789435642419334e-05, "loss": 0.875, "step": 158270 }, { "epoch": 13.71, "learning_rate": 4.078726726032149e-05, "loss": 0.991, "step": 158280 }, { "epoch": 13.71, "learning_rate": 4.078509887822366e-05, "loss": 0.8346, "step": 158290 }, { "epoch": 13.71, "learning_rate": 4.0782930496125826e-05, "loss": 0.7489, "step": 158300 }, { "epoch": 13.71, "learning_rate": 4.078076211402798e-05, "loss": 0.9854, "step": 158310 }, { "epoch": 13.71, "learning_rate": 4.0778593731930145e-05, "loss": 0.9206, "step": 158320 }, { "epoch": 13.71, "learning_rate": 4.0776425349832305e-05, "loss": 1.0301, "step": 158330 }, { "epoch": 13.71, "learning_rate": 4.077425696773447e-05, "loss": 1.2647, "step": 158340 }, { "epoch": 13.71, "learning_rate": 4.077208858563663e-05, "loss": 1.1182, "step": 158350 }, { "epoch": 13.71, "learning_rate": 4.07699202035388e-05, "loss": 0.9712, "step": 158360 }, { "epoch": 13.72, "learning_rate": 4.0767751821440964e-05, "loss": 0.9214, "step": 158370 }, { "epoch": 13.72, "learning_rate": 4.076558343934312e-05, "loss": 1.0138, "step": 158380 }, { "epoch": 13.72, "learning_rate": 4.076341505724528e-05, "loss": 1.2339, "step": 158390 }, { "epoch": 13.72, "learning_rate": 4.076124667514744e-05, "loss": 1.0232, "step": 158400 }, { "epoch": 13.72, "learning_rate": 4.075907829304961e-05, "loss": 0.9615, "step": 158410 }, { "epoch": 13.72, "learning_rate": 4.075690991095177e-05, "loss": 1.0065, "step": 158420 }, { "epoch": 13.72, "learning_rate": 4.0754741528853935e-05, "loss": 0.9801, "step": 158430 }, { "epoch": 13.72, "learning_rate": 4.07525731467561e-05, "loss": 0.9442, "step": 158440 }, { "epoch": 13.72, "learning_rate": 4.075040476465826e-05, "loss": 1.0578, "step": 158450 }, { "epoch": 13.72, "learning_rate": 4.074823638256043e-05, "loss": 1.1407, "step": 158460 }, { "epoch": 13.72, "learning_rate": 4.074606800046258e-05, "loss": 0.8114, "step": 158470 }, { "epoch": 13.73, "learning_rate": 4.0743899618364747e-05, "loss": 1.0258, "step": 158480 }, { "epoch": 13.73, "learning_rate": 4.0741731236266906e-05, "loss": 0.9976, "step": 158490 }, { "epoch": 13.73, "learning_rate": 4.073956285416907e-05, "loss": 0.9086, "step": 158500 }, { "epoch": 13.73, "learning_rate": 4.073739447207124e-05, "loss": 0.9658, "step": 158510 }, { "epoch": 13.73, "learning_rate": 4.07352260899734e-05, "loss": 1.147, "step": 158520 }, { "epoch": 13.73, "learning_rate": 4.0733057707875565e-05, "loss": 1.2144, "step": 158530 }, { "epoch": 13.73, "learning_rate": 4.0730889325777725e-05, "loss": 1.2046, "step": 158540 }, { "epoch": 13.73, "learning_rate": 4.0728720943679884e-05, "loss": 0.911, "step": 158550 }, { "epoch": 13.73, "learning_rate": 4.0726552561582044e-05, "loss": 1.1777, "step": 158560 }, { "epoch": 13.73, "learning_rate": 4.072438417948421e-05, "loss": 0.9099, "step": 158570 }, { "epoch": 13.73, "learning_rate": 4.0722215797386377e-05, "loss": 0.7938, "step": 158580 }, { "epoch": 13.73, "learning_rate": 4.0720047415288536e-05, "loss": 0.8524, "step": 158590 }, { "epoch": 13.74, "learning_rate": 4.07178790331907e-05, "loss": 1.1102, "step": 158600 }, { "epoch": 13.74, "learning_rate": 4.071571065109286e-05, "loss": 1.0549, "step": 158610 }, { "epoch": 13.74, "learning_rate": 4.071354226899502e-05, "loss": 0.9598, "step": 158620 }, { "epoch": 13.74, "learning_rate": 4.071137388689718e-05, "loss": 1.1437, "step": 158630 }, { "epoch": 13.74, "learning_rate": 4.070920550479935e-05, "loss": 1.0111, "step": 158640 }, { "epoch": 13.74, "learning_rate": 4.070703712270151e-05, "loss": 1.0705, "step": 158650 }, { "epoch": 13.74, "learning_rate": 4.0704868740603674e-05, "loss": 1.0365, "step": 158660 }, { "epoch": 13.74, "learning_rate": 4.070270035850584e-05, "loss": 1.0263, "step": 158670 }, { "epoch": 13.74, "learning_rate": 4.0700531976408e-05, "loss": 1.0143, "step": 158680 }, { "epoch": 13.74, "learning_rate": 4.0698363594310166e-05, "loss": 1.2111, "step": 158690 }, { "epoch": 13.74, "learning_rate": 4.069619521221232e-05, "loss": 0.9502, "step": 158700 }, { "epoch": 13.75, "learning_rate": 4.0694026830114485e-05, "loss": 1.0064, "step": 158710 }, { "epoch": 13.75, "learning_rate": 4.0691858448016645e-05, "loss": 1.0534, "step": 158720 }, { "epoch": 13.75, "learning_rate": 4.068969006591881e-05, "loss": 0.839, "step": 158730 }, { "epoch": 13.75, "learning_rate": 4.068752168382098e-05, "loss": 0.9205, "step": 158740 }, { "epoch": 13.75, "learning_rate": 4.068535330172314e-05, "loss": 1.033, "step": 158750 }, { "epoch": 13.75, "learning_rate": 4.0683184919625304e-05, "loss": 0.9766, "step": 158760 }, { "epoch": 13.75, "learning_rate": 4.0681016537527463e-05, "loss": 1.1814, "step": 158770 }, { "epoch": 13.75, "learning_rate": 4.067884815542962e-05, "loss": 0.9307, "step": 158780 }, { "epoch": 13.75, "learning_rate": 4.067667977333178e-05, "loss": 0.9681, "step": 158790 }, { "epoch": 13.75, "learning_rate": 4.067451139123395e-05, "loss": 1.0911, "step": 158800 }, { "epoch": 13.75, "learning_rate": 4.0672343009136115e-05, "loss": 1.0022, "step": 158810 }, { "epoch": 13.75, "learning_rate": 4.0670174627038275e-05, "loss": 1.0532, "step": 158820 }, { "epoch": 13.76, "learning_rate": 4.066800624494044e-05, "loss": 1.1708, "step": 158830 }, { "epoch": 13.76, "learning_rate": 4.06658378628426e-05, "loss": 0.8674, "step": 158840 }, { "epoch": 13.76, "learning_rate": 4.066366948074476e-05, "loss": 0.8312, "step": 158850 }, { "epoch": 13.76, "learning_rate": 4.066150109864692e-05, "loss": 1.0693, "step": 158860 }, { "epoch": 13.76, "learning_rate": 4.065933271654909e-05, "loss": 0.8687, "step": 158870 }, { "epoch": 13.76, "learning_rate": 4.065716433445125e-05, "loss": 1.0632, "step": 158880 }, { "epoch": 13.76, "learning_rate": 4.065499595235341e-05, "loss": 1.2153, "step": 158890 }, { "epoch": 13.76, "learning_rate": 4.065282757025558e-05, "loss": 1.0902, "step": 158900 }, { "epoch": 13.76, "learning_rate": 4.065065918815774e-05, "loss": 0.8921, "step": 158910 }, { "epoch": 13.76, "learning_rate": 4.0648490806059905e-05, "loss": 1.0333, "step": 158920 }, { "epoch": 13.76, "learning_rate": 4.064632242396206e-05, "loss": 1.0046, "step": 158930 }, { "epoch": 13.77, "learning_rate": 4.0644154041864224e-05, "loss": 1.0068, "step": 158940 }, { "epoch": 13.77, "learning_rate": 4.064198565976639e-05, "loss": 1.0249, "step": 158950 }, { "epoch": 13.77, "learning_rate": 4.063981727766855e-05, "loss": 1.0182, "step": 158960 }, { "epoch": 13.77, "learning_rate": 4.063764889557072e-05, "loss": 0.9176, "step": 158970 }, { "epoch": 13.77, "learning_rate": 4.0635480513472876e-05, "loss": 0.9883, "step": 158980 }, { "epoch": 13.77, "learning_rate": 4.063331213137504e-05, "loss": 1.0217, "step": 158990 }, { "epoch": 13.77, "learning_rate": 4.063114374927721e-05, "loss": 1.0474, "step": 159000 }, { "epoch": 13.77, "learning_rate": 4.062897536717936e-05, "loss": 1.2093, "step": 159010 }, { "epoch": 13.77, "learning_rate": 4.062680698508153e-05, "loss": 1.0597, "step": 159020 }, { "epoch": 13.77, "learning_rate": 4.062463860298369e-05, "loss": 1.0801, "step": 159030 }, { "epoch": 13.77, "learning_rate": 4.0622470220885854e-05, "loss": 1.0626, "step": 159040 }, { "epoch": 13.77, "learning_rate": 4.0620301838788014e-05, "loss": 0.8985, "step": 159050 }, { "epoch": 13.78, "learning_rate": 4.061813345669018e-05, "loss": 0.9204, "step": 159060 }, { "epoch": 13.78, "learning_rate": 4.061596507459235e-05, "loss": 1.0836, "step": 159070 }, { "epoch": 13.78, "learning_rate": 4.06137966924945e-05, "loss": 1.1564, "step": 159080 }, { "epoch": 13.78, "learning_rate": 4.0611628310396666e-05, "loss": 1.0073, "step": 159090 }, { "epoch": 13.78, "learning_rate": 4.0609459928298826e-05, "loss": 1.1945, "step": 159100 }, { "epoch": 13.78, "learning_rate": 4.060729154620099e-05, "loss": 1.105, "step": 159110 }, { "epoch": 13.78, "learning_rate": 4.060512316410315e-05, "loss": 1.0343, "step": 159120 }, { "epoch": 13.78, "learning_rate": 4.060295478200532e-05, "loss": 1.1207, "step": 159130 }, { "epoch": 13.78, "learning_rate": 4.0600786399907484e-05, "loss": 1.0045, "step": 159140 }, { "epoch": 13.78, "learning_rate": 4.0598618017809644e-05, "loss": 1.1236, "step": 159150 }, { "epoch": 13.78, "learning_rate": 4.05964496357118e-05, "loss": 1.08, "step": 159160 }, { "epoch": 13.79, "learning_rate": 4.059428125361396e-05, "loss": 1.0524, "step": 159170 }, { "epoch": 13.79, "learning_rate": 4.059211287151613e-05, "loss": 0.9984, "step": 159180 }, { "epoch": 13.79, "learning_rate": 4.058994448941829e-05, "loss": 1.0187, "step": 159190 }, { "epoch": 13.79, "learning_rate": 4.0587776107320456e-05, "loss": 1.0176, "step": 159200 }, { "epoch": 13.79, "learning_rate": 4.0585607725222615e-05, "loss": 1.1078, "step": 159210 }, { "epoch": 13.79, "learning_rate": 4.058343934312478e-05, "loss": 0.9778, "step": 159220 }, { "epoch": 13.79, "learning_rate": 4.058127096102695e-05, "loss": 1.0556, "step": 159230 }, { "epoch": 13.79, "learning_rate": 4.05791025789291e-05, "loss": 0.9008, "step": 159240 }, { "epoch": 13.79, "learning_rate": 4.057693419683127e-05, "loss": 1.0153, "step": 159250 }, { "epoch": 13.79, "learning_rate": 4.057476581473343e-05, "loss": 0.8651, "step": 159260 }, { "epoch": 13.79, "learning_rate": 4.057259743263559e-05, "loss": 0.8656, "step": 159270 }, { "epoch": 13.79, "learning_rate": 4.057042905053775e-05, "loss": 1.1433, "step": 159280 }, { "epoch": 13.8, "learning_rate": 4.056826066843992e-05, "loss": 1.0982, "step": 159290 }, { "epoch": 13.8, "learning_rate": 4.0566092286342086e-05, "loss": 0.9472, "step": 159300 }, { "epoch": 13.8, "learning_rate": 4.0563923904244245e-05, "loss": 0.965, "step": 159310 }, { "epoch": 13.8, "learning_rate": 4.0561755522146405e-05, "loss": 0.9677, "step": 159320 }, { "epoch": 13.8, "learning_rate": 4.0559587140048565e-05, "loss": 1.0009, "step": 159330 }, { "epoch": 13.8, "learning_rate": 4.055741875795073e-05, "loss": 1.0579, "step": 159340 }, { "epoch": 13.8, "learning_rate": 4.055525037585289e-05, "loss": 1.0505, "step": 159350 }, { "epoch": 13.8, "learning_rate": 4.055308199375506e-05, "loss": 0.9621, "step": 159360 }, { "epoch": 13.8, "learning_rate": 4.055091361165722e-05, "loss": 0.7906, "step": 159370 }, { "epoch": 13.8, "learning_rate": 4.054874522955938e-05, "loss": 1.1001, "step": 159380 }, { "epoch": 13.8, "learning_rate": 4.054657684746154e-05, "loss": 1.0553, "step": 159390 }, { "epoch": 13.81, "learning_rate": 4.05444084653637e-05, "loss": 1.0611, "step": 159400 }, { "epoch": 13.81, "learning_rate": 4.054224008326587e-05, "loss": 1.0657, "step": 159410 }, { "epoch": 13.81, "learning_rate": 4.054007170116803e-05, "loss": 1.0519, "step": 159420 }, { "epoch": 13.81, "learning_rate": 4.0537903319070195e-05, "loss": 0.9408, "step": 159430 }, { "epoch": 13.81, "learning_rate": 4.053573493697236e-05, "loss": 1.2286, "step": 159440 }, { "epoch": 13.81, "learning_rate": 4.053356655487452e-05, "loss": 0.9313, "step": 159450 }, { "epoch": 13.81, "learning_rate": 4.053139817277669e-05, "loss": 0.9763, "step": 159460 }, { "epoch": 13.81, "learning_rate": 4.052922979067884e-05, "loss": 0.9242, "step": 159470 }, { "epoch": 13.81, "learning_rate": 4.0527061408581006e-05, "loss": 1.1211, "step": 159480 }, { "epoch": 13.81, "learning_rate": 4.0524893026483166e-05, "loss": 0.9048, "step": 159490 }, { "epoch": 13.81, "learning_rate": 4.052272464438533e-05, "loss": 1.1537, "step": 159500 }, { "epoch": 13.81, "learning_rate": 4.05205562622875e-05, "loss": 1.0922, "step": 159510 }, { "epoch": 13.82, "learning_rate": 4.051838788018966e-05, "loss": 0.9977, "step": 159520 }, { "epoch": 13.82, "learning_rate": 4.0516219498091825e-05, "loss": 0.9611, "step": 159530 }, { "epoch": 13.82, "learning_rate": 4.0514051115993984e-05, "loss": 1.0944, "step": 159540 }, { "epoch": 13.82, "learning_rate": 4.0511882733896144e-05, "loss": 0.8263, "step": 159550 }, { "epoch": 13.82, "learning_rate": 4.0509714351798303e-05, "loss": 1.0214, "step": 159560 }, { "epoch": 13.82, "learning_rate": 4.050754596970047e-05, "loss": 0.9755, "step": 159570 }, { "epoch": 13.82, "learning_rate": 4.0505377587602636e-05, "loss": 0.93, "step": 159580 }, { "epoch": 13.82, "learning_rate": 4.0503209205504796e-05, "loss": 1.1912, "step": 159590 }, { "epoch": 13.82, "learning_rate": 4.050104082340696e-05, "loss": 1.0114, "step": 159600 }, { "epoch": 13.82, "learning_rate": 4.049887244130912e-05, "loss": 1.0623, "step": 159610 }, { "epoch": 13.82, "learning_rate": 4.049670405921128e-05, "loss": 1.069, "step": 159620 }, { "epoch": 13.82, "learning_rate": 4.049453567711344e-05, "loss": 1.0071, "step": 159630 }, { "epoch": 13.83, "learning_rate": 4.049236729501561e-05, "loss": 1.231, "step": 159640 }, { "epoch": 13.83, "learning_rate": 4.0490198912917774e-05, "loss": 1.0244, "step": 159650 }, { "epoch": 13.83, "learning_rate": 4.0488030530819934e-05, "loss": 0.9479, "step": 159660 }, { "epoch": 13.83, "learning_rate": 4.04858621487221e-05, "loss": 0.9755, "step": 159670 }, { "epoch": 13.83, "learning_rate": 4.048369376662426e-05, "loss": 0.9563, "step": 159680 }, { "epoch": 13.83, "learning_rate": 4.0481525384526426e-05, "loss": 1.0217, "step": 159690 }, { "epoch": 13.83, "learning_rate": 4.047935700242858e-05, "loss": 0.9929, "step": 159700 }, { "epoch": 13.83, "learning_rate": 4.0477188620330745e-05, "loss": 0.9487, "step": 159710 }, { "epoch": 13.83, "learning_rate": 4.0475020238232905e-05, "loss": 1.0497, "step": 159720 }, { "epoch": 13.83, "learning_rate": 4.047285185613507e-05, "loss": 0.7667, "step": 159730 }, { "epoch": 13.83, "learning_rate": 4.047068347403724e-05, "loss": 1.0399, "step": 159740 }, { "epoch": 13.84, "learning_rate": 4.04685150919394e-05, "loss": 0.9193, "step": 159750 }, { "epoch": 13.84, "learning_rate": 4.0466346709841564e-05, "loss": 0.9763, "step": 159760 }, { "epoch": 13.84, "learning_rate": 4.046417832774372e-05, "loss": 1.1734, "step": 159770 }, { "epoch": 13.84, "learning_rate": 4.046200994564588e-05, "loss": 0.9512, "step": 159780 }, { "epoch": 13.84, "learning_rate": 4.045984156354804e-05, "loss": 1.1055, "step": 159790 }, { "epoch": 13.84, "learning_rate": 4.045767318145021e-05, "loss": 1.0468, "step": 159800 }, { "epoch": 13.84, "learning_rate": 4.0455504799352375e-05, "loss": 1.0808, "step": 159810 }, { "epoch": 13.84, "learning_rate": 4.0453336417254535e-05, "loss": 0.8861, "step": 159820 }, { "epoch": 13.84, "learning_rate": 4.04511680351567e-05, "loss": 0.9643, "step": 159830 }, { "epoch": 13.84, "learning_rate": 4.044899965305886e-05, "loss": 1.1979, "step": 159840 }, { "epoch": 13.84, "learning_rate": 4.044683127096102e-05, "loss": 0.8806, "step": 159850 }, { "epoch": 13.84, "learning_rate": 4.044466288886318e-05, "loss": 1.0823, "step": 159860 }, { "epoch": 13.85, "learning_rate": 4.0442494506765346e-05, "loss": 0.8849, "step": 159870 }, { "epoch": 13.85, "learning_rate": 4.044032612466751e-05, "loss": 0.963, "step": 159880 }, { "epoch": 13.85, "learning_rate": 4.043815774256967e-05, "loss": 1.1262, "step": 159890 }, { "epoch": 13.85, "learning_rate": 4.043598936047184e-05, "loss": 1.0547, "step": 159900 }, { "epoch": 13.85, "learning_rate": 4.0433820978374e-05, "loss": 0.9401, "step": 159910 }, { "epoch": 13.85, "learning_rate": 4.0431652596276165e-05, "loss": 0.9928, "step": 159920 }, { "epoch": 13.85, "learning_rate": 4.042948421417832e-05, "loss": 1.182, "step": 159930 }, { "epoch": 13.85, "learning_rate": 4.0427315832080484e-05, "loss": 1.2173, "step": 159940 }, { "epoch": 13.85, "learning_rate": 4.042514744998265e-05, "loss": 1.0281, "step": 159950 }, { "epoch": 13.85, "learning_rate": 4.042297906788481e-05, "loss": 0.9851, "step": 159960 }, { "epoch": 13.85, "learning_rate": 4.0420810685786976e-05, "loss": 1.0117, "step": 159970 }, { "epoch": 13.86, "learning_rate": 4.0418642303689136e-05, "loss": 0.9192, "step": 159980 }, { "epoch": 13.86, "learning_rate": 4.04164739215913e-05, "loss": 0.9738, "step": 159990 }, { "epoch": 13.86, "learning_rate": 4.041430553949347e-05, "loss": 0.8774, "step": 160000 }, { "epoch": 13.86, "learning_rate": 4.041213715739562e-05, "loss": 0.9472, "step": 160010 }, { "epoch": 13.86, "learning_rate": 4.040996877529779e-05, "loss": 1.0175, "step": 160020 }, { "epoch": 13.86, "learning_rate": 4.040780039319995e-05, "loss": 1.2111, "step": 160030 }, { "epoch": 13.86, "learning_rate": 4.0405632011102114e-05, "loss": 0.9226, "step": 160040 }, { "epoch": 13.86, "learning_rate": 4.0403463629004274e-05, "loss": 0.9103, "step": 160050 }, { "epoch": 13.86, "learning_rate": 4.040129524690644e-05, "loss": 1.0297, "step": 160060 }, { "epoch": 13.86, "learning_rate": 4.0399126864808607e-05, "loss": 0.9672, "step": 160070 }, { "epoch": 13.86, "learning_rate": 4.0396958482710766e-05, "loss": 0.9999, "step": 160080 }, { "epoch": 13.86, "learning_rate": 4.0394790100612926e-05, "loss": 0.9691, "step": 160090 }, { "epoch": 13.87, "learning_rate": 4.0392621718515085e-05, "loss": 1.0834, "step": 160100 }, { "epoch": 13.87, "learning_rate": 4.039045333641725e-05, "loss": 1.075, "step": 160110 }, { "epoch": 13.87, "learning_rate": 4.038828495431941e-05, "loss": 1.119, "step": 160120 }, { "epoch": 13.87, "learning_rate": 4.038611657222158e-05, "loss": 1.12, "step": 160130 }, { "epoch": 13.87, "learning_rate": 4.0383948190123744e-05, "loss": 1.1085, "step": 160140 }, { "epoch": 13.87, "learning_rate": 4.0381779808025904e-05, "loss": 1.0634, "step": 160150 }, { "epoch": 13.87, "learning_rate": 4.037961142592806e-05, "loss": 0.9485, "step": 160160 }, { "epoch": 13.87, "learning_rate": 4.037744304383022e-05, "loss": 1.1885, "step": 160170 }, { "epoch": 13.87, "learning_rate": 4.037527466173239e-05, "loss": 1.0746, "step": 160180 }, { "epoch": 13.87, "learning_rate": 4.037310627963455e-05, "loss": 1.0735, "step": 160190 }, { "epoch": 13.87, "learning_rate": 4.0370937897536715e-05, "loss": 0.9802, "step": 160200 }, { "epoch": 13.88, "learning_rate": 4.036876951543888e-05, "loss": 0.9691, "step": 160210 }, { "epoch": 13.88, "learning_rate": 4.036660113334104e-05, "loss": 0.9464, "step": 160220 }, { "epoch": 13.88, "learning_rate": 4.036443275124321e-05, "loss": 0.8822, "step": 160230 }, { "epoch": 13.88, "learning_rate": 4.036226436914536e-05, "loss": 0.8975, "step": 160240 }, { "epoch": 13.88, "learning_rate": 4.036009598704753e-05, "loss": 0.9971, "step": 160250 }, { "epoch": 13.88, "learning_rate": 4.035792760494969e-05, "loss": 0.9858, "step": 160260 }, { "epoch": 13.88, "learning_rate": 4.035575922285185e-05, "loss": 1.1728, "step": 160270 }, { "epoch": 13.88, "learning_rate": 4.035359084075401e-05, "loss": 0.9591, "step": 160280 }, { "epoch": 13.88, "learning_rate": 4.035142245865618e-05, "loss": 0.9546, "step": 160290 }, { "epoch": 13.88, "learning_rate": 4.0349254076558345e-05, "loss": 0.9663, "step": 160300 }, { "epoch": 13.88, "learning_rate": 4.0347085694460505e-05, "loss": 0.889, "step": 160310 }, { "epoch": 13.88, "learning_rate": 4.0344917312362665e-05, "loss": 0.9846, "step": 160320 }, { "epoch": 13.89, "learning_rate": 4.0342748930264824e-05, "loss": 1.194, "step": 160330 }, { "epoch": 13.89, "learning_rate": 4.034058054816699e-05, "loss": 0.995, "step": 160340 }, { "epoch": 13.89, "learning_rate": 4.033841216606915e-05, "loss": 1.0053, "step": 160350 }, { "epoch": 13.89, "learning_rate": 4.033624378397132e-05, "loss": 1.0478, "step": 160360 }, { "epoch": 13.89, "learning_rate": 4.033407540187348e-05, "loss": 0.9619, "step": 160370 }, { "epoch": 13.89, "learning_rate": 4.033190701977564e-05, "loss": 0.8873, "step": 160380 }, { "epoch": 13.89, "learning_rate": 4.03297386376778e-05, "loss": 1.0533, "step": 160390 }, { "epoch": 13.89, "learning_rate": 4.032757025557996e-05, "loss": 0.9242, "step": 160400 }, { "epoch": 13.89, "learning_rate": 4.032540187348213e-05, "loss": 0.9926, "step": 160410 }, { "epoch": 13.89, "learning_rate": 4.032323349138429e-05, "loss": 1.1036, "step": 160420 }, { "epoch": 13.89, "learning_rate": 4.0321065109286454e-05, "loss": 0.9881, "step": 160430 }, { "epoch": 13.9, "learning_rate": 4.031889672718862e-05, "loss": 1.0546, "step": 160440 }, { "epoch": 13.9, "learning_rate": 4.031672834509078e-05, "loss": 0.9104, "step": 160450 }, { "epoch": 13.9, "learning_rate": 4.031455996299295e-05, "loss": 1.1138, "step": 160460 }, { "epoch": 13.9, "learning_rate": 4.03123915808951e-05, "loss": 0.9029, "step": 160470 }, { "epoch": 13.9, "learning_rate": 4.0310223198797266e-05, "loss": 1.031, "step": 160480 }, { "epoch": 13.9, "learning_rate": 4.0308054816699426e-05, "loss": 1.0186, "step": 160490 }, { "epoch": 13.9, "learning_rate": 4.030588643460159e-05, "loss": 1.0331, "step": 160500 }, { "epoch": 13.9, "learning_rate": 4.030371805250376e-05, "loss": 0.9122, "step": 160510 }, { "epoch": 13.9, "learning_rate": 4.030154967040592e-05, "loss": 1.0622, "step": 160520 }, { "epoch": 13.9, "learning_rate": 4.0299381288308084e-05, "loss": 1.0705, "step": 160530 }, { "epoch": 13.9, "learning_rate": 4.0297212906210244e-05, "loss": 0.8006, "step": 160540 }, { "epoch": 13.9, "learning_rate": 4.0295044524112404e-05, "loss": 1.1079, "step": 160550 }, { "epoch": 13.91, "learning_rate": 4.029287614201456e-05, "loss": 1.0217, "step": 160560 }, { "epoch": 13.91, "learning_rate": 4.029070775991673e-05, "loss": 1.2057, "step": 160570 }, { "epoch": 13.91, "learning_rate": 4.0288539377818896e-05, "loss": 0.9846, "step": 160580 }, { "epoch": 13.91, "learning_rate": 4.0286370995721056e-05, "loss": 1.0885, "step": 160590 }, { "epoch": 13.91, "learning_rate": 4.028420261362322e-05, "loss": 1.2443, "step": 160600 }, { "epoch": 13.91, "learning_rate": 4.028203423152538e-05, "loss": 1.0892, "step": 160610 }, { "epoch": 13.91, "learning_rate": 4.027986584942754e-05, "loss": 0.9384, "step": 160620 }, { "epoch": 13.91, "learning_rate": 4.02776974673297e-05, "loss": 1.1062, "step": 160630 }, { "epoch": 13.91, "learning_rate": 4.027552908523187e-05, "loss": 1.0441, "step": 160640 }, { "epoch": 13.91, "learning_rate": 4.0273360703134034e-05, "loss": 1.0331, "step": 160650 }, { "epoch": 13.91, "learning_rate": 4.027119232103619e-05, "loss": 1.0557, "step": 160660 }, { "epoch": 13.92, "learning_rate": 4.026902393893836e-05, "loss": 1.0143, "step": 160670 }, { "epoch": 13.92, "learning_rate": 4.026685555684052e-05, "loss": 1.0785, "step": 160680 }, { "epoch": 13.92, "learning_rate": 4.0264687174742686e-05, "loss": 1.018, "step": 160690 }, { "epoch": 13.92, "learning_rate": 4.026251879264484e-05, "loss": 1.1969, "step": 160700 }, { "epoch": 13.92, "learning_rate": 4.0260350410547005e-05, "loss": 1.0607, "step": 160710 }, { "epoch": 13.92, "learning_rate": 4.025818202844917e-05, "loss": 0.944, "step": 160720 }, { "epoch": 13.92, "learning_rate": 4.025601364635133e-05, "loss": 0.8946, "step": 160730 }, { "epoch": 13.92, "learning_rate": 4.02538452642535e-05, "loss": 0.8881, "step": 160740 }, { "epoch": 13.92, "learning_rate": 4.025167688215566e-05, "loss": 1.1289, "step": 160750 }, { "epoch": 13.92, "learning_rate": 4.024950850005782e-05, "loss": 1.1646, "step": 160760 }, { "epoch": 13.92, "learning_rate": 4.024734011795999e-05, "loss": 1.0282, "step": 160770 }, { "epoch": 13.92, "learning_rate": 4.024517173586214e-05, "loss": 1.0104, "step": 160780 }, { "epoch": 13.93, "learning_rate": 4.02430033537643e-05, "loss": 0.8752, "step": 160790 }, { "epoch": 13.93, "learning_rate": 4.024083497166647e-05, "loss": 1.1423, "step": 160800 }, { "epoch": 13.93, "learning_rate": 4.0238666589568635e-05, "loss": 1.0155, "step": 160810 }, { "epoch": 13.93, "learning_rate": 4.0236498207470794e-05, "loss": 1.1273, "step": 160820 }, { "epoch": 13.93, "learning_rate": 4.023432982537296e-05, "loss": 0.9676, "step": 160830 }, { "epoch": 13.93, "learning_rate": 4.023216144327512e-05, "loss": 0.9543, "step": 160840 }, { "epoch": 13.93, "learning_rate": 4.022999306117729e-05, "loss": 0.8526, "step": 160850 }, { "epoch": 13.93, "learning_rate": 4.022782467907944e-05, "loss": 1.0665, "step": 160860 }, { "epoch": 13.93, "learning_rate": 4.0225656296981606e-05, "loss": 0.8978, "step": 160870 }, { "epoch": 13.93, "learning_rate": 4.022348791488377e-05, "loss": 0.8603, "step": 160880 }, { "epoch": 13.93, "learning_rate": 4.022131953278593e-05, "loss": 1.0688, "step": 160890 }, { "epoch": 13.93, "learning_rate": 4.02191511506881e-05, "loss": 1.0109, "step": 160900 }, { "epoch": 13.94, "learning_rate": 4.021698276859026e-05, "loss": 1.0391, "step": 160910 }, { "epoch": 13.94, "learning_rate": 4.0214814386492425e-05, "loss": 0.992, "step": 160920 }, { "epoch": 13.94, "learning_rate": 4.021264600439458e-05, "loss": 0.92, "step": 160930 }, { "epoch": 13.94, "learning_rate": 4.0210477622296744e-05, "loss": 0.976, "step": 160940 }, { "epoch": 13.94, "learning_rate": 4.020830924019891e-05, "loss": 0.8695, "step": 160950 }, { "epoch": 13.94, "learning_rate": 4.020614085810107e-05, "loss": 1.0606, "step": 160960 }, { "epoch": 13.94, "learning_rate": 4.0203972476003236e-05, "loss": 0.9511, "step": 160970 }, { "epoch": 13.94, "learning_rate": 4.0201804093905396e-05, "loss": 0.9939, "step": 160980 }, { "epoch": 13.94, "learning_rate": 4.019963571180756e-05, "loss": 1.0658, "step": 160990 }, { "epoch": 13.94, "learning_rate": 4.019746732970973e-05, "loss": 1.0701, "step": 161000 }, { "epoch": 13.94, "learning_rate": 4.019529894761188e-05, "loss": 1.0734, "step": 161010 }, { "epoch": 13.95, "learning_rate": 4.019313056551405e-05, "loss": 0.9859, "step": 161020 }, { "epoch": 13.95, "learning_rate": 4.019096218341621e-05, "loss": 0.8894, "step": 161030 }, { "epoch": 13.95, "learning_rate": 4.0188793801318374e-05, "loss": 1.0328, "step": 161040 }, { "epoch": 13.95, "learning_rate": 4.0186625419220533e-05, "loss": 1.1895, "step": 161050 }, { "epoch": 13.95, "learning_rate": 4.01844570371227e-05, "loss": 1.0251, "step": 161060 }, { "epoch": 13.95, "learning_rate": 4.0182288655024866e-05, "loss": 0.8213, "step": 161070 }, { "epoch": 13.95, "learning_rate": 4.0180120272927026e-05, "loss": 1.0905, "step": 161080 }, { "epoch": 13.95, "learning_rate": 4.0177951890829185e-05, "loss": 0.8432, "step": 161090 }, { "epoch": 13.95, "learning_rate": 4.0175783508731345e-05, "loss": 0.928, "step": 161100 }, { "epoch": 13.95, "learning_rate": 4.017361512663351e-05, "loss": 1.1526, "step": 161110 }, { "epoch": 13.95, "learning_rate": 4.017144674453567e-05, "loss": 1.0175, "step": 161120 }, { "epoch": 13.95, "learning_rate": 4.016927836243784e-05, "loss": 0.923, "step": 161130 }, { "epoch": 13.96, "learning_rate": 4.0167109980340004e-05, "loss": 0.8745, "step": 161140 }, { "epoch": 13.96, "learning_rate": 4.0164941598242163e-05, "loss": 1.1807, "step": 161150 }, { "epoch": 13.96, "learning_rate": 4.016277321614432e-05, "loss": 1.0014, "step": 161160 }, { "epoch": 13.96, "learning_rate": 4.016060483404648e-05, "loss": 0.8001, "step": 161170 }, { "epoch": 13.96, "learning_rate": 4.015843645194865e-05, "loss": 1.1555, "step": 161180 }, { "epoch": 13.96, "learning_rate": 4.015626806985081e-05, "loss": 0.9968, "step": 161190 }, { "epoch": 13.96, "learning_rate": 4.0154099687752975e-05, "loss": 1.077, "step": 161200 }, { "epoch": 13.96, "learning_rate": 4.015193130565514e-05, "loss": 1.1551, "step": 161210 }, { "epoch": 13.96, "learning_rate": 4.01497629235573e-05, "loss": 0.9864, "step": 161220 }, { "epoch": 13.96, "learning_rate": 4.014759454145947e-05, "loss": 1.0558, "step": 161230 }, { "epoch": 13.96, "learning_rate": 4.014542615936162e-05, "loss": 1.1745, "step": 161240 }, { "epoch": 13.97, "learning_rate": 4.014325777726379e-05, "loss": 1.0667, "step": 161250 }, { "epoch": 13.97, "learning_rate": 4.0141089395165946e-05, "loss": 0.8905, "step": 161260 }, { "epoch": 13.97, "learning_rate": 4.013892101306811e-05, "loss": 0.9136, "step": 161270 }, { "epoch": 13.97, "learning_rate": 4.013675263097028e-05, "loss": 1.0856, "step": 161280 }, { "epoch": 13.97, "learning_rate": 4.013458424887244e-05, "loss": 0.9532, "step": 161290 }, { "epoch": 13.97, "learning_rate": 4.0132415866774605e-05, "loss": 0.9335, "step": 161300 }, { "epoch": 13.97, "learning_rate": 4.0130247484676765e-05, "loss": 1.0955, "step": 161310 }, { "epoch": 13.97, "learning_rate": 4.0128079102578924e-05, "loss": 1.1857, "step": 161320 }, { "epoch": 13.97, "learning_rate": 4.0125910720481084e-05, "loss": 0.9827, "step": 161330 }, { "epoch": 13.97, "learning_rate": 4.012374233838325e-05, "loss": 0.9905, "step": 161340 }, { "epoch": 13.97, "learning_rate": 4.012157395628541e-05, "loss": 0.8844, "step": 161350 }, { "epoch": 13.97, "learning_rate": 4.0119405574187576e-05, "loss": 0.9415, "step": 161360 }, { "epoch": 13.98, "learning_rate": 4.011723719208974e-05, "loss": 1.0492, "step": 161370 }, { "epoch": 13.98, "learning_rate": 4.01150688099919e-05, "loss": 1.054, "step": 161380 }, { "epoch": 13.98, "learning_rate": 4.011290042789406e-05, "loss": 1.0147, "step": 161390 }, { "epoch": 13.98, "learning_rate": 4.011073204579622e-05, "loss": 0.9971, "step": 161400 }, { "epoch": 13.98, "learning_rate": 4.010856366369839e-05, "loss": 0.8972, "step": 161410 }, { "epoch": 13.98, "learning_rate": 4.010639528160055e-05, "loss": 0.8995, "step": 161420 }, { "epoch": 13.98, "learning_rate": 4.0104226899502714e-05, "loss": 0.8703, "step": 161430 }, { "epoch": 13.98, "learning_rate": 4.010205851740488e-05, "loss": 1.0865, "step": 161440 }, { "epoch": 13.98, "learning_rate": 4.009989013530704e-05, "loss": 1.1652, "step": 161450 }, { "epoch": 13.98, "learning_rate": 4.0097721753209206e-05, "loss": 1.088, "step": 161460 }, { "epoch": 13.98, "learning_rate": 4.009555337111136e-05, "loss": 0.947, "step": 161470 }, { "epoch": 13.99, "learning_rate": 4.0093384989013526e-05, "loss": 1.0966, "step": 161480 }, { "epoch": 13.99, "learning_rate": 4.0091216606915685e-05, "loss": 1.1462, "step": 161490 }, { "epoch": 13.99, "learning_rate": 4.008904822481785e-05, "loss": 0.8499, "step": 161500 }, { "epoch": 13.99, "learning_rate": 4.008687984272002e-05, "loss": 0.8344, "step": 161510 }, { "epoch": 13.99, "learning_rate": 4.008471146062218e-05, "loss": 1.124, "step": 161520 }, { "epoch": 13.99, "learning_rate": 4.0082543078524344e-05, "loss": 0.9953, "step": 161530 }, { "epoch": 13.99, "learning_rate": 4.0080374696426504e-05, "loss": 1.1216, "step": 161540 }, { "epoch": 13.99, "learning_rate": 4.007820631432866e-05, "loss": 0.9618, "step": 161550 }, { "epoch": 13.99, "learning_rate": 4.007603793223082e-05, "loss": 0.9863, "step": 161560 }, { "epoch": 13.99, "learning_rate": 4.007386955013299e-05, "loss": 0.9932, "step": 161570 }, { "epoch": 13.99, "learning_rate": 4.0071701168035156e-05, "loss": 0.9665, "step": 161580 }, { "epoch": 13.99, "learning_rate": 4.0069532785937315e-05, "loss": 1.0435, "step": 161590 }, { "epoch": 14.0, "learning_rate": 4.006736440383948e-05, "loss": 0.9458, "step": 161600 }, { "epoch": 14.0, "learning_rate": 4.006519602174164e-05, "loss": 0.8807, "step": 161610 }, { "epoch": 14.0, "learning_rate": 4.006302763964381e-05, "loss": 1.1482, "step": 161620 }, { "epoch": 14.0, "learning_rate": 4.006085925754596e-05, "loss": 1.0084, "step": 161630 }, { "epoch": 14.0, "learning_rate": 4.005869087544813e-05, "loss": 1.0485, "step": 161640 }, { "epoch": 14.0, "learning_rate": 4.005652249335029e-05, "loss": 1.0051, "step": 161650 }, { "epoch": 14.0, "eval_cer": 0.544846182788956, "eval_em": 0.04744713769984528, "eval_f1": 0.04744713769984528, "eval_loss": 1.454521894454956, "eval_runtime": 253.1624, "eval_samples_per_second": 7.659, "eval_steps_per_second": 0.96, "eval_wer": 0.9525528623001547, "step": 161651 }, { "epoch": 14.0, "learning_rate": 4.005435411125245e-05, "loss": 1.0386, "step": 161660 }, { "epoch": 14.0, "learning_rate": 4.005218572915462e-05, "loss": 0.9622, "step": 161670 }, { "epoch": 14.0, "learning_rate": 4.005001734705678e-05, "loss": 0.9775, "step": 161680 }, { "epoch": 14.0, "learning_rate": 4.0047848964958945e-05, "loss": 0.8307, "step": 161690 }, { "epoch": 14.0, "learning_rate": 4.00456805828611e-05, "loss": 1.009, "step": 161700 }, { "epoch": 14.01, "learning_rate": 4.0043512200763265e-05, "loss": 0.9455, "step": 161710 }, { "epoch": 14.01, "learning_rate": 4.004134381866543e-05, "loss": 0.9633, "step": 161720 }, { "epoch": 14.01, "learning_rate": 4.003917543656759e-05, "loss": 1.0346, "step": 161730 }, { "epoch": 14.01, "learning_rate": 4.003700705446976e-05, "loss": 0.9007, "step": 161740 }, { "epoch": 14.01, "learning_rate": 4.0034838672371917e-05, "loss": 1.0529, "step": 161750 }, { "epoch": 14.01, "learning_rate": 4.003267029027408e-05, "loss": 0.9101, "step": 161760 }, { "epoch": 14.01, "learning_rate": 4.003050190817625e-05, "loss": 0.9735, "step": 161770 }, { "epoch": 14.01, "learning_rate": 4.00283335260784e-05, "loss": 0.7922, "step": 161780 }, { "epoch": 14.01, "learning_rate": 4.002616514398057e-05, "loss": 0.8858, "step": 161790 }, { "epoch": 14.01, "learning_rate": 4.002399676188273e-05, "loss": 1.0804, "step": 161800 }, { "epoch": 14.01, "learning_rate": 4.0021828379784895e-05, "loss": 1.0411, "step": 161810 }, { "epoch": 14.01, "learning_rate": 4.0019659997687054e-05, "loss": 1.041, "step": 161820 }, { "epoch": 14.02, "learning_rate": 4.001749161558922e-05, "loss": 0.9439, "step": 161830 }, { "epoch": 14.02, "learning_rate": 4.001532323349139e-05, "loss": 0.8833, "step": 161840 }, { "epoch": 14.02, "learning_rate": 4.0013154851393547e-05, "loss": 0.9467, "step": 161850 }, { "epoch": 14.02, "learning_rate": 4.00109864692957e-05, "loss": 0.9472, "step": 161860 }, { "epoch": 14.02, "learning_rate": 4.0008818087197866e-05, "loss": 0.849, "step": 161870 }, { "epoch": 14.02, "learning_rate": 4.000664970510003e-05, "loss": 0.8052, "step": 161880 }, { "epoch": 14.02, "learning_rate": 4.000448132300219e-05, "loss": 1.0626, "step": 161890 }, { "epoch": 14.02, "learning_rate": 4.000231294090436e-05, "loss": 1.0318, "step": 161900 }, { "epoch": 14.02, "learning_rate": 4.000014455880652e-05, "loss": 1.0758, "step": 161910 }, { "epoch": 14.02, "learning_rate": 3.9997976176708684e-05, "loss": 0.976, "step": 161920 }, { "epoch": 14.02, "learning_rate": 3.999580779461084e-05, "loss": 0.8671, "step": 161930 }, { "epoch": 14.03, "learning_rate": 3.9993639412513003e-05, "loss": 0.7448, "step": 161940 }, { "epoch": 14.03, "learning_rate": 3.999147103041517e-05, "loss": 0.9256, "step": 161950 }, { "epoch": 14.03, "learning_rate": 3.998930264831733e-05, "loss": 0.8177, "step": 161960 }, { "epoch": 14.03, "learning_rate": 3.9987134266219496e-05, "loss": 1.0174, "step": 161970 }, { "epoch": 14.03, "learning_rate": 3.9984965884121655e-05, "loss": 1.0102, "step": 161980 }, { "epoch": 14.03, "learning_rate": 3.998279750202382e-05, "loss": 1.0403, "step": 161990 }, { "epoch": 14.03, "learning_rate": 3.998062911992599e-05, "loss": 0.8903, "step": 162000 }, { "epoch": 14.03, "learning_rate": 3.997846073782814e-05, "loss": 0.9984, "step": 162010 }, { "epoch": 14.03, "learning_rate": 3.997629235573031e-05, "loss": 0.9342, "step": 162020 }, { "epoch": 14.03, "learning_rate": 3.997412397363247e-05, "loss": 0.9307, "step": 162030 }, { "epoch": 14.03, "learning_rate": 3.9971955591534633e-05, "loss": 1.0187, "step": 162040 }, { "epoch": 14.03, "learning_rate": 3.996978720943679e-05, "loss": 0.94, "step": 162050 }, { "epoch": 14.04, "learning_rate": 3.996761882733896e-05, "loss": 0.9652, "step": 162060 }, { "epoch": 14.04, "learning_rate": 3.9965450445241126e-05, "loss": 0.9592, "step": 162070 }, { "epoch": 14.04, "learning_rate": 3.9963282063143286e-05, "loss": 1.1454, "step": 162080 }, { "epoch": 14.04, "learning_rate": 3.9961113681045445e-05, "loss": 1.0345, "step": 162090 }, { "epoch": 14.04, "learning_rate": 3.9958945298947605e-05, "loss": 0.9781, "step": 162100 }, { "epoch": 14.04, "learning_rate": 3.995677691684977e-05, "loss": 0.8909, "step": 162110 }, { "epoch": 14.04, "learning_rate": 3.995460853475193e-05, "loss": 0.9857, "step": 162120 }, { "epoch": 14.04, "learning_rate": 3.99524401526541e-05, "loss": 0.8784, "step": 162130 }, { "epoch": 14.04, "learning_rate": 3.9950271770556264e-05, "loss": 0.8406, "step": 162140 }, { "epoch": 14.04, "learning_rate": 3.994810338845842e-05, "loss": 0.9232, "step": 162150 }, { "epoch": 14.04, "learning_rate": 3.994593500636058e-05, "loss": 0.8186, "step": 162160 }, { "epoch": 14.04, "learning_rate": 3.994376662426274e-05, "loss": 1.0977, "step": 162170 }, { "epoch": 14.05, "learning_rate": 3.994159824216491e-05, "loss": 0.9254, "step": 162180 }, { "epoch": 14.05, "learning_rate": 3.993942986006707e-05, "loss": 0.9279, "step": 162190 }, { "epoch": 14.05, "learning_rate": 3.9937261477969235e-05, "loss": 1.0159, "step": 162200 }, { "epoch": 14.05, "learning_rate": 3.99350930958714e-05, "loss": 0.8779, "step": 162210 }, { "epoch": 14.05, "learning_rate": 3.993292471377356e-05, "loss": 0.8619, "step": 162220 }, { "epoch": 14.05, "learning_rate": 3.993075633167573e-05, "loss": 0.9664, "step": 162230 }, { "epoch": 14.05, "learning_rate": 3.992858794957788e-05, "loss": 0.8806, "step": 162240 }, { "epoch": 14.05, "learning_rate": 3.9926419567480046e-05, "loss": 1.024, "step": 162250 }, { "epoch": 14.05, "learning_rate": 3.9924251185382206e-05, "loss": 0.896, "step": 162260 }, { "epoch": 14.05, "learning_rate": 3.992208280328437e-05, "loss": 1.1176, "step": 162270 }, { "epoch": 14.05, "learning_rate": 3.991991442118654e-05, "loss": 0.8173, "step": 162280 }, { "epoch": 14.06, "learning_rate": 3.99177460390887e-05, "loss": 0.8966, "step": 162290 }, { "epoch": 14.06, "learning_rate": 3.9915577656990865e-05, "loss": 0.8606, "step": 162300 }, { "epoch": 14.06, "learning_rate": 3.9913409274893024e-05, "loss": 0.834, "step": 162310 }, { "epoch": 14.06, "learning_rate": 3.9911240892795184e-05, "loss": 1.0025, "step": 162320 }, { "epoch": 14.06, "learning_rate": 3.9909072510697344e-05, "loss": 1.0046, "step": 162330 }, { "epoch": 14.06, "learning_rate": 3.990690412859951e-05, "loss": 0.9867, "step": 162340 }, { "epoch": 14.06, "learning_rate": 3.9904735746501676e-05, "loss": 0.9069, "step": 162350 }, { "epoch": 14.06, "learning_rate": 3.9902567364403836e-05, "loss": 0.8739, "step": 162360 }, { "epoch": 14.06, "learning_rate": 3.9900398982306e-05, "loss": 0.9411, "step": 162370 }, { "epoch": 14.06, "learning_rate": 3.989823060020816e-05, "loss": 1.0681, "step": 162380 }, { "epoch": 14.06, "learning_rate": 3.989606221811033e-05, "loss": 0.8821, "step": 162390 }, { "epoch": 14.06, "learning_rate": 3.989389383601248e-05, "loss": 0.8453, "step": 162400 }, { "epoch": 14.07, "learning_rate": 3.989172545391465e-05, "loss": 1.0039, "step": 162410 }, { "epoch": 14.07, "learning_rate": 3.988955707181681e-05, "loss": 1.1056, "step": 162420 }, { "epoch": 14.07, "learning_rate": 3.9887388689718974e-05, "loss": 0.9875, "step": 162430 }, { "epoch": 14.07, "learning_rate": 3.988522030762114e-05, "loss": 0.9346, "step": 162440 }, { "epoch": 14.07, "learning_rate": 3.98830519255233e-05, "loss": 0.911, "step": 162450 }, { "epoch": 14.07, "learning_rate": 3.9880883543425466e-05, "loss": 0.9277, "step": 162460 }, { "epoch": 14.07, "learning_rate": 3.987871516132762e-05, "loss": 0.7823, "step": 162470 }, { "epoch": 14.07, "learning_rate": 3.9876546779229785e-05, "loss": 0.9679, "step": 162480 }, { "epoch": 14.07, "learning_rate": 3.9874378397131945e-05, "loss": 0.817, "step": 162490 }, { "epoch": 14.07, "learning_rate": 3.987221001503411e-05, "loss": 0.9406, "step": 162500 }, { "epoch": 14.07, "learning_rate": 3.987004163293628e-05, "loss": 0.8526, "step": 162510 }, { "epoch": 14.08, "learning_rate": 3.986787325083844e-05, "loss": 0.9751, "step": 162520 }, { "epoch": 14.08, "learning_rate": 3.9865704868740604e-05, "loss": 0.8865, "step": 162530 }, { "epoch": 14.08, "learning_rate": 3.986353648664276e-05, "loss": 1.0356, "step": 162540 }, { "epoch": 14.08, "learning_rate": 3.986136810454492e-05, "loss": 0.9965, "step": 162550 }, { "epoch": 14.08, "learning_rate": 3.985919972244708e-05, "loss": 0.8808, "step": 162560 }, { "epoch": 14.08, "learning_rate": 3.985703134034925e-05, "loss": 0.8802, "step": 162570 }, { "epoch": 14.08, "learning_rate": 3.9854862958251415e-05, "loss": 0.8948, "step": 162580 }, { "epoch": 14.08, "learning_rate": 3.9852694576153575e-05, "loss": 0.8515, "step": 162590 }, { "epoch": 14.08, "learning_rate": 3.985052619405574e-05, "loss": 0.9528, "step": 162600 }, { "epoch": 14.08, "learning_rate": 3.98483578119579e-05, "loss": 1.0925, "step": 162610 }, { "epoch": 14.08, "learning_rate": 3.984618942986007e-05, "loss": 0.9863, "step": 162620 }, { "epoch": 14.08, "learning_rate": 3.984402104776222e-05, "loss": 0.8322, "step": 162630 }, { "epoch": 14.09, "learning_rate": 3.9841852665664387e-05, "loss": 1.0151, "step": 162640 }, { "epoch": 14.09, "learning_rate": 3.983968428356655e-05, "loss": 0.9691, "step": 162650 }, { "epoch": 14.09, "learning_rate": 3.983751590146871e-05, "loss": 0.9244, "step": 162660 }, { "epoch": 14.09, "learning_rate": 3.983534751937088e-05, "loss": 0.8389, "step": 162670 }, { "epoch": 14.09, "learning_rate": 3.983317913727304e-05, "loss": 0.946, "step": 162680 }, { "epoch": 14.09, "learning_rate": 3.9831010755175205e-05, "loss": 0.8869, "step": 162690 }, { "epoch": 14.09, "learning_rate": 3.982884237307736e-05, "loss": 0.9831, "step": 162700 }, { "epoch": 14.09, "learning_rate": 3.9826673990979524e-05, "loss": 0.8882, "step": 162710 }, { "epoch": 14.09, "learning_rate": 3.982450560888169e-05, "loss": 0.9174, "step": 162720 }, { "epoch": 14.09, "learning_rate": 3.982233722678385e-05, "loss": 0.8972, "step": 162730 }, { "epoch": 14.09, "learning_rate": 3.982016884468602e-05, "loss": 1.0689, "step": 162740 }, { "epoch": 14.1, "learning_rate": 3.9818000462588176e-05, "loss": 1.0189, "step": 162750 }, { "epoch": 14.1, "learning_rate": 3.981583208049034e-05, "loss": 0.9014, "step": 162760 }, { "epoch": 14.1, "learning_rate": 3.981366369839251e-05, "loss": 0.8124, "step": 162770 }, { "epoch": 14.1, "learning_rate": 3.981149531629466e-05, "loss": 0.8628, "step": 162780 }, { "epoch": 14.1, "learning_rate": 3.980932693419683e-05, "loss": 1.0298, "step": 162790 }, { "epoch": 14.1, "learning_rate": 3.980715855209899e-05, "loss": 0.9456, "step": 162800 }, { "epoch": 14.1, "learning_rate": 3.9804990170001154e-05, "loss": 0.8748, "step": 162810 }, { "epoch": 14.1, "learning_rate": 3.9802821787903314e-05, "loss": 0.8831, "step": 162820 }, { "epoch": 14.1, "learning_rate": 3.980065340580548e-05, "loss": 0.7962, "step": 162830 }, { "epoch": 14.1, "learning_rate": 3.979848502370765e-05, "loss": 0.737, "step": 162840 }, { "epoch": 14.1, "learning_rate": 3.9796316641609806e-05, "loss": 1.0207, "step": 162850 }, { "epoch": 14.1, "learning_rate": 3.9794148259511966e-05, "loss": 0.9944, "step": 162860 }, { "epoch": 14.11, "learning_rate": 3.9791979877414126e-05, "loss": 0.9529, "step": 162870 }, { "epoch": 14.11, "learning_rate": 3.978981149531629e-05, "loss": 0.8463, "step": 162880 }, { "epoch": 14.11, "learning_rate": 3.978764311321845e-05, "loss": 0.9284, "step": 162890 }, { "epoch": 14.11, "learning_rate": 3.978547473112062e-05, "loss": 0.8483, "step": 162900 }, { "epoch": 14.11, "learning_rate": 3.9783306349022784e-05, "loss": 1.0203, "step": 162910 }, { "epoch": 14.11, "learning_rate": 3.9781137966924944e-05, "loss": 0.9619, "step": 162920 }, { "epoch": 14.11, "learning_rate": 3.97789695848271e-05, "loss": 0.9626, "step": 162930 }, { "epoch": 14.11, "learning_rate": 3.977680120272926e-05, "loss": 0.8778, "step": 162940 }, { "epoch": 14.11, "learning_rate": 3.977463282063143e-05, "loss": 0.8923, "step": 162950 }, { "epoch": 14.11, "learning_rate": 3.977246443853359e-05, "loss": 1.044, "step": 162960 }, { "epoch": 14.11, "learning_rate": 3.9770296056435756e-05, "loss": 0.8978, "step": 162970 }, { "epoch": 14.12, "learning_rate": 3.9768127674337915e-05, "loss": 1.0746, "step": 162980 }, { "epoch": 14.12, "learning_rate": 3.976595929224008e-05, "loss": 0.843, "step": 162990 }, { "epoch": 14.12, "learning_rate": 3.976379091014225e-05, "loss": 1.0131, "step": 163000 }, { "epoch": 14.12, "learning_rate": 3.97616225280444e-05, "loss": 1.0069, "step": 163010 }, { "epoch": 14.12, "learning_rate": 3.975945414594657e-05, "loss": 0.8338, "step": 163020 }, { "epoch": 14.12, "learning_rate": 3.975728576384873e-05, "loss": 1.1276, "step": 163030 }, { "epoch": 14.12, "learning_rate": 3.975511738175089e-05, "loss": 1.1511, "step": 163040 }, { "epoch": 14.12, "learning_rate": 3.975294899965305e-05, "loss": 1.152, "step": 163050 }, { "epoch": 14.12, "learning_rate": 3.975078061755522e-05, "loss": 0.9828, "step": 163060 }, { "epoch": 14.12, "learning_rate": 3.9748612235457386e-05, "loss": 0.8454, "step": 163070 }, { "epoch": 14.12, "learning_rate": 3.9746443853359545e-05, "loss": 0.9563, "step": 163080 }, { "epoch": 14.12, "learning_rate": 3.9744275471261705e-05, "loss": 1.069, "step": 163090 }, { "epoch": 14.13, "learning_rate": 3.9742107089163864e-05, "loss": 0.9057, "step": 163100 }, { "epoch": 14.13, "learning_rate": 3.973993870706603e-05, "loss": 1.0305, "step": 163110 }, { "epoch": 14.13, "learning_rate": 3.973777032496819e-05, "loss": 0.876, "step": 163120 }, { "epoch": 14.13, "learning_rate": 3.973560194287036e-05, "loss": 1.0299, "step": 163130 }, { "epoch": 14.13, "learning_rate": 3.973343356077252e-05, "loss": 1.0036, "step": 163140 }, { "epoch": 14.13, "learning_rate": 3.973126517867468e-05, "loss": 0.9293, "step": 163150 }, { "epoch": 14.13, "learning_rate": 3.972909679657685e-05, "loss": 0.9152, "step": 163160 }, { "epoch": 14.13, "learning_rate": 3.9726928414479e-05, "loss": 0.9553, "step": 163170 }, { "epoch": 14.13, "learning_rate": 3.972476003238117e-05, "loss": 1.0128, "step": 163180 }, { "epoch": 14.13, "learning_rate": 3.972259165028333e-05, "loss": 0.9864, "step": 163190 }, { "epoch": 14.13, "learning_rate": 3.9720423268185494e-05, "loss": 1.0029, "step": 163200 }, { "epoch": 14.14, "learning_rate": 3.971825488608766e-05, "loss": 0.9775, "step": 163210 }, { "epoch": 14.14, "learning_rate": 3.971608650398982e-05, "loss": 1.0448, "step": 163220 }, { "epoch": 14.14, "learning_rate": 3.971391812189199e-05, "loss": 0.9534, "step": 163230 }, { "epoch": 14.14, "learning_rate": 3.971174973979414e-05, "loss": 0.9894, "step": 163240 }, { "epoch": 14.14, "learning_rate": 3.9709581357696306e-05, "loss": 0.9254, "step": 163250 }, { "epoch": 14.14, "learning_rate": 3.9707412975598466e-05, "loss": 0.7472, "step": 163260 }, { "epoch": 14.14, "learning_rate": 3.970524459350063e-05, "loss": 1.0583, "step": 163270 }, { "epoch": 14.14, "learning_rate": 3.97030762114028e-05, "loss": 0.8868, "step": 163280 }, { "epoch": 14.14, "learning_rate": 3.970090782930496e-05, "loss": 1.011, "step": 163290 }, { "epoch": 14.14, "learning_rate": 3.9698739447207124e-05, "loss": 0.9737, "step": 163300 }, { "epoch": 14.14, "learning_rate": 3.9696571065109284e-05, "loss": 0.8213, "step": 163310 }, { "epoch": 14.14, "learning_rate": 3.9694402683011444e-05, "loss": 0.8412, "step": 163320 }, { "epoch": 14.15, "learning_rate": 3.96922343009136e-05, "loss": 0.9394, "step": 163330 }, { "epoch": 14.15, "learning_rate": 3.969006591881577e-05, "loss": 0.9535, "step": 163340 }, { "epoch": 14.15, "learning_rate": 3.9687897536717936e-05, "loss": 1.007, "step": 163350 }, { "epoch": 14.15, "learning_rate": 3.9685729154620096e-05, "loss": 1.1062, "step": 163360 }, { "epoch": 14.15, "learning_rate": 3.968356077252226e-05, "loss": 0.9409, "step": 163370 }, { "epoch": 14.15, "learning_rate": 3.968139239042442e-05, "loss": 0.9399, "step": 163380 }, { "epoch": 14.15, "learning_rate": 3.967922400832659e-05, "loss": 1.0554, "step": 163390 }, { "epoch": 14.15, "learning_rate": 3.967705562622874e-05, "loss": 0.803, "step": 163400 }, { "epoch": 14.15, "learning_rate": 3.967488724413091e-05, "loss": 0.8529, "step": 163410 }, { "epoch": 14.15, "learning_rate": 3.9672718862033074e-05, "loss": 1.038, "step": 163420 }, { "epoch": 14.15, "learning_rate": 3.967055047993523e-05, "loss": 0.9782, "step": 163430 }, { "epoch": 14.15, "learning_rate": 3.96683820978374e-05, "loss": 1.0723, "step": 163440 }, { "epoch": 14.16, "learning_rate": 3.966621371573956e-05, "loss": 1.0646, "step": 163450 }, { "epoch": 14.16, "learning_rate": 3.9664045333641726e-05, "loss": 0.9264, "step": 163460 }, { "epoch": 14.16, "learning_rate": 3.966187695154388e-05, "loss": 0.8634, "step": 163470 }, { "epoch": 14.16, "learning_rate": 3.9659708569446045e-05, "loss": 0.9408, "step": 163480 }, { "epoch": 14.16, "learning_rate": 3.9657540187348205e-05, "loss": 0.8859, "step": 163490 }, { "epoch": 14.16, "learning_rate": 3.965537180525037e-05, "loss": 0.8513, "step": 163500 }, { "epoch": 14.16, "learning_rate": 3.965320342315254e-05, "loss": 0.91, "step": 163510 }, { "epoch": 14.16, "learning_rate": 3.96510350410547e-05, "loss": 0.8469, "step": 163520 }, { "epoch": 14.16, "learning_rate": 3.9648866658956863e-05, "loss": 0.8842, "step": 163530 }, { "epoch": 14.16, "learning_rate": 3.964669827685902e-05, "loss": 0.9859, "step": 163540 }, { "epoch": 14.16, "learning_rate": 3.964452989476118e-05, "loss": 0.9115, "step": 163550 }, { "epoch": 14.17, "learning_rate": 3.964236151266334e-05, "loss": 0.8976, "step": 163560 }, { "epoch": 14.17, "learning_rate": 3.964019313056551e-05, "loss": 1.0908, "step": 163570 }, { "epoch": 14.17, "learning_rate": 3.9638024748467675e-05, "loss": 0.938, "step": 163580 }, { "epoch": 14.17, "learning_rate": 3.9635856366369835e-05, "loss": 0.9483, "step": 163590 }, { "epoch": 14.17, "learning_rate": 3.9633687984272e-05, "loss": 0.9421, "step": 163600 }, { "epoch": 14.17, "learning_rate": 3.963151960217416e-05, "loss": 1.1196, "step": 163610 }, { "epoch": 14.17, "learning_rate": 3.962935122007633e-05, "loss": 1.0966, "step": 163620 }, { "epoch": 14.17, "learning_rate": 3.962718283797848e-05, "loss": 0.9881, "step": 163630 }, { "epoch": 14.17, "learning_rate": 3.9625014455880646e-05, "loss": 0.7421, "step": 163640 }, { "epoch": 14.17, "learning_rate": 3.962284607378281e-05, "loss": 0.8939, "step": 163650 }, { "epoch": 14.17, "learning_rate": 3.962067769168497e-05, "loss": 0.8159, "step": 163660 }, { "epoch": 14.17, "learning_rate": 3.961850930958714e-05, "loss": 0.899, "step": 163670 }, { "epoch": 14.18, "learning_rate": 3.96163409274893e-05, "loss": 1.1256, "step": 163680 }, { "epoch": 14.18, "learning_rate": 3.9614172545391465e-05, "loss": 0.9215, "step": 163690 }, { "epoch": 14.18, "learning_rate": 3.961200416329362e-05, "loss": 0.9764, "step": 163700 }, { "epoch": 14.18, "learning_rate": 3.9609835781195784e-05, "loss": 1.0374, "step": 163710 }, { "epoch": 14.18, "learning_rate": 3.960766739909795e-05, "loss": 0.7821, "step": 163720 }, { "epoch": 14.18, "learning_rate": 3.960549901700011e-05, "loss": 1.0278, "step": 163730 }, { "epoch": 14.18, "learning_rate": 3.9603330634902276e-05, "loss": 0.8223, "step": 163740 }, { "epoch": 14.18, "learning_rate": 3.9601162252804436e-05, "loss": 0.9224, "step": 163750 }, { "epoch": 14.18, "learning_rate": 3.95989938707066e-05, "loss": 1.0292, "step": 163760 }, { "epoch": 14.18, "learning_rate": 3.959682548860877e-05, "loss": 1.0568, "step": 163770 }, { "epoch": 14.18, "learning_rate": 3.959465710651092e-05, "loss": 0.8388, "step": 163780 }, { "epoch": 14.19, "learning_rate": 3.959248872441309e-05, "loss": 0.9995, "step": 163790 }, { "epoch": 14.19, "learning_rate": 3.959032034231525e-05, "loss": 0.9493, "step": 163800 }, { "epoch": 14.19, "learning_rate": 3.9588151960217414e-05, "loss": 1.0234, "step": 163810 }, { "epoch": 14.19, "learning_rate": 3.9585983578119574e-05, "loss": 1.0326, "step": 163820 }, { "epoch": 14.19, "learning_rate": 3.958381519602174e-05, "loss": 0.9942, "step": 163830 }, { "epoch": 14.19, "learning_rate": 3.9581646813923906e-05, "loss": 1.019, "step": 163840 }, { "epoch": 14.19, "learning_rate": 3.9579478431826066e-05, "loss": 1.1493, "step": 163850 }, { "epoch": 14.19, "learning_rate": 3.9577310049728226e-05, "loss": 0.9151, "step": 163860 }, { "epoch": 14.19, "learning_rate": 3.9575141667630385e-05, "loss": 1.0822, "step": 163870 }, { "epoch": 14.19, "learning_rate": 3.957297328553255e-05, "loss": 0.8436, "step": 163880 }, { "epoch": 14.19, "learning_rate": 3.957080490343471e-05, "loss": 0.9731, "step": 163890 }, { "epoch": 14.19, "learning_rate": 3.956863652133688e-05, "loss": 1.0788, "step": 163900 }, { "epoch": 14.2, "learning_rate": 3.9566468139239044e-05, "loss": 0.8852, "step": 163910 }, { "epoch": 14.2, "learning_rate": 3.9564299757141204e-05, "loss": 0.8436, "step": 163920 }, { "epoch": 14.2, "learning_rate": 3.956213137504337e-05, "loss": 0.8624, "step": 163930 }, { "epoch": 14.2, "learning_rate": 3.955996299294552e-05, "loss": 0.9228, "step": 163940 }, { "epoch": 14.2, "learning_rate": 3.955779461084769e-05, "loss": 0.7721, "step": 163950 }, { "epoch": 14.2, "learning_rate": 3.955562622874985e-05, "loss": 0.8681, "step": 163960 }, { "epoch": 14.2, "learning_rate": 3.9553457846652015e-05, "loss": 0.9578, "step": 163970 }, { "epoch": 14.2, "learning_rate": 3.955128946455418e-05, "loss": 0.7988, "step": 163980 }, { "epoch": 14.2, "learning_rate": 3.954912108245634e-05, "loss": 1.0516, "step": 163990 }, { "epoch": 14.2, "learning_rate": 3.954695270035851e-05, "loss": 0.9909, "step": 164000 }, { "epoch": 14.2, "learning_rate": 3.954478431826066e-05, "loss": 0.8724, "step": 164010 }, { "epoch": 14.21, "learning_rate": 3.954261593616283e-05, "loss": 0.9811, "step": 164020 }, { "epoch": 14.21, "learning_rate": 3.9540447554064986e-05, "loss": 0.9196, "step": 164030 }, { "epoch": 14.21, "learning_rate": 3.953827917196715e-05, "loss": 0.8564, "step": 164040 }, { "epoch": 14.21, "learning_rate": 3.953611078986931e-05, "loss": 1.039, "step": 164050 }, { "epoch": 14.21, "learning_rate": 3.953394240777148e-05, "loss": 1.0035, "step": 164060 }, { "epoch": 14.21, "learning_rate": 3.9531774025673645e-05, "loss": 0.9688, "step": 164070 }, { "epoch": 14.21, "learning_rate": 3.9529605643575805e-05, "loss": 0.8213, "step": 164080 }, { "epoch": 14.21, "learning_rate": 3.9527437261477965e-05, "loss": 0.9931, "step": 164090 }, { "epoch": 14.21, "learning_rate": 3.9525268879380124e-05, "loss": 0.8084, "step": 164100 }, { "epoch": 14.21, "learning_rate": 3.952310049728229e-05, "loss": 1.1038, "step": 164110 }, { "epoch": 14.21, "learning_rate": 3.952093211518445e-05, "loss": 0.9493, "step": 164120 }, { "epoch": 14.21, "learning_rate": 3.9518763733086617e-05, "loss": 0.7873, "step": 164130 }, { "epoch": 14.22, "learning_rate": 3.951659535098878e-05, "loss": 1.1377, "step": 164140 }, { "epoch": 14.22, "learning_rate": 3.951442696889094e-05, "loss": 0.9324, "step": 164150 }, { "epoch": 14.22, "learning_rate": 3.951225858679311e-05, "loss": 0.8277, "step": 164160 }, { "epoch": 14.22, "learning_rate": 3.951009020469526e-05, "loss": 0.9014, "step": 164170 }, { "epoch": 14.22, "learning_rate": 3.950792182259743e-05, "loss": 1.074, "step": 164180 }, { "epoch": 14.22, "learning_rate": 3.950575344049959e-05, "loss": 0.9914, "step": 164190 }, { "epoch": 14.22, "learning_rate": 3.9503585058401754e-05, "loss": 0.9003, "step": 164200 }, { "epoch": 14.22, "learning_rate": 3.950141667630392e-05, "loss": 0.7624, "step": 164210 }, { "epoch": 14.22, "learning_rate": 3.949924829420608e-05, "loss": 1.0644, "step": 164220 }, { "epoch": 14.22, "learning_rate": 3.9497079912108247e-05, "loss": 0.8552, "step": 164230 }, { "epoch": 14.22, "learning_rate": 3.94949115300104e-05, "loss": 0.9162, "step": 164240 }, { "epoch": 14.23, "learning_rate": 3.9492743147912566e-05, "loss": 0.9643, "step": 164250 }, { "epoch": 14.23, "learning_rate": 3.9490574765814725e-05, "loss": 0.848, "step": 164260 }, { "epoch": 14.23, "learning_rate": 3.948840638371689e-05, "loss": 0.9299, "step": 164270 }, { "epoch": 14.23, "learning_rate": 3.948623800161906e-05, "loss": 0.9444, "step": 164280 }, { "epoch": 14.23, "learning_rate": 3.948406961952122e-05, "loss": 1.0053, "step": 164290 }, { "epoch": 14.23, "learning_rate": 3.9481901237423384e-05, "loss": 0.8777, "step": 164300 }, { "epoch": 14.23, "learning_rate": 3.9479732855325544e-05, "loss": 0.995, "step": 164310 }, { "epoch": 14.23, "learning_rate": 3.9477564473227703e-05, "loss": 1.0416, "step": 164320 }, { "epoch": 14.23, "learning_rate": 3.947539609112986e-05, "loss": 0.9119, "step": 164330 }, { "epoch": 14.23, "learning_rate": 3.947322770903203e-05, "loss": 0.9633, "step": 164340 }, { "epoch": 14.23, "learning_rate": 3.9471059326934196e-05, "loss": 0.991, "step": 164350 }, { "epoch": 14.23, "learning_rate": 3.9468890944836355e-05, "loss": 1.0123, "step": 164360 }, { "epoch": 14.24, "learning_rate": 3.946672256273852e-05, "loss": 0.9786, "step": 164370 }, { "epoch": 14.24, "learning_rate": 3.946455418064068e-05, "loss": 0.8807, "step": 164380 }, { "epoch": 14.24, "learning_rate": 3.946238579854285e-05, "loss": 0.8015, "step": 164390 }, { "epoch": 14.24, "learning_rate": 3.9460217416445e-05, "loss": 0.8777, "step": 164400 }, { "epoch": 14.24, "learning_rate": 3.945804903434717e-05, "loss": 1.0404, "step": 164410 }, { "epoch": 14.24, "learning_rate": 3.9455880652249333e-05, "loss": 1.0242, "step": 164420 }, { "epoch": 14.24, "learning_rate": 3.945371227015149e-05, "loss": 0.8951, "step": 164430 }, { "epoch": 14.24, "learning_rate": 3.945154388805366e-05, "loss": 0.8033, "step": 164440 }, { "epoch": 14.24, "learning_rate": 3.944937550595582e-05, "loss": 0.8965, "step": 164450 }, { "epoch": 14.24, "learning_rate": 3.9447207123857985e-05, "loss": 0.7629, "step": 164460 }, { "epoch": 14.24, "learning_rate": 3.944503874176014e-05, "loss": 1.0483, "step": 164470 }, { "epoch": 14.25, "learning_rate": 3.9442870359662305e-05, "loss": 0.8697, "step": 164480 }, { "epoch": 14.25, "learning_rate": 3.944070197756447e-05, "loss": 1.0454, "step": 164490 }, { "epoch": 14.25, "learning_rate": 3.943853359546663e-05, "loss": 0.8313, "step": 164500 }, { "epoch": 14.25, "learning_rate": 3.94363652133688e-05, "loss": 1.1674, "step": 164510 }, { "epoch": 14.25, "learning_rate": 3.943419683127096e-05, "loss": 0.713, "step": 164520 }, { "epoch": 14.25, "learning_rate": 3.943202844917312e-05, "loss": 1.1325, "step": 164530 }, { "epoch": 14.25, "learning_rate": 3.942986006707529e-05, "loss": 0.9804, "step": 164540 }, { "epoch": 14.25, "learning_rate": 3.942769168497744e-05, "loss": 0.7937, "step": 164550 }, { "epoch": 14.25, "learning_rate": 3.94255233028796e-05, "loss": 1.0898, "step": 164560 }, { "epoch": 14.25, "learning_rate": 3.942335492078177e-05, "loss": 0.9291, "step": 164570 }, { "epoch": 14.25, "learning_rate": 3.9421186538683935e-05, "loss": 0.9353, "step": 164580 }, { "epoch": 14.25, "learning_rate": 3.9419018156586094e-05, "loss": 1.104, "step": 164590 }, { "epoch": 14.26, "learning_rate": 3.941684977448826e-05, "loss": 0.9512, "step": 164600 }, { "epoch": 14.26, "learning_rate": 3.941468139239042e-05, "loss": 1.0003, "step": 164610 }, { "epoch": 14.26, "learning_rate": 3.941251301029259e-05, "loss": 1.0016, "step": 164620 }, { "epoch": 14.26, "learning_rate": 3.941034462819474e-05, "loss": 0.8794, "step": 164630 }, { "epoch": 14.26, "learning_rate": 3.9408176246096906e-05, "loss": 0.8411, "step": 164640 }, { "epoch": 14.26, "learning_rate": 3.940600786399907e-05, "loss": 0.9941, "step": 164650 }, { "epoch": 14.26, "learning_rate": 3.940383948190123e-05, "loss": 0.9737, "step": 164660 }, { "epoch": 14.26, "learning_rate": 3.94016710998034e-05, "loss": 1.0475, "step": 164670 }, { "epoch": 14.26, "learning_rate": 3.939950271770556e-05, "loss": 0.8295, "step": 164680 }, { "epoch": 14.26, "learning_rate": 3.9397334335607724e-05, "loss": 0.8136, "step": 164690 }, { "epoch": 14.26, "learning_rate": 3.939516595350989e-05, "loss": 0.8255, "step": 164700 }, { "epoch": 14.26, "learning_rate": 3.9392997571412044e-05, "loss": 0.9972, "step": 164710 }, { "epoch": 14.27, "learning_rate": 3.939082918931421e-05, "loss": 0.7981, "step": 164720 }, { "epoch": 14.27, "learning_rate": 3.938866080721637e-05, "loss": 0.9471, "step": 164730 }, { "epoch": 14.27, "learning_rate": 3.9386492425118536e-05, "loss": 1.0274, "step": 164740 }, { "epoch": 14.27, "learning_rate": 3.9384324043020696e-05, "loss": 0.8428, "step": 164750 }, { "epoch": 14.27, "learning_rate": 3.938215566092286e-05, "loss": 1.0968, "step": 164760 }, { "epoch": 14.27, "learning_rate": 3.937998727882503e-05, "loss": 0.9511, "step": 164770 }, { "epoch": 14.27, "learning_rate": 3.937781889672718e-05, "loss": 0.9118, "step": 164780 }, { "epoch": 14.27, "learning_rate": 3.937565051462935e-05, "loss": 1.0619, "step": 164790 }, { "epoch": 14.27, "learning_rate": 3.937348213253151e-05, "loss": 1.1815, "step": 164800 }, { "epoch": 14.27, "learning_rate": 3.9371313750433674e-05, "loss": 0.9422, "step": 164810 }, { "epoch": 14.27, "learning_rate": 3.936914536833583e-05, "loss": 0.9322, "step": 164820 }, { "epoch": 14.28, "learning_rate": 3.9366976986238e-05, "loss": 1.0788, "step": 164830 }, { "epoch": 14.28, "learning_rate": 3.9364808604140166e-05, "loss": 0.8171, "step": 164840 }, { "epoch": 14.28, "learning_rate": 3.9362640222042326e-05, "loss": 0.815, "step": 164850 }, { "epoch": 14.28, "learning_rate": 3.9360471839944485e-05, "loss": 0.8307, "step": 164860 }, { "epoch": 14.28, "learning_rate": 3.9358303457846645e-05, "loss": 0.838, "step": 164870 }, { "epoch": 14.28, "learning_rate": 3.935613507574881e-05, "loss": 0.9527, "step": 164880 }, { "epoch": 14.28, "learning_rate": 3.935396669365097e-05, "loss": 0.876, "step": 164890 }, { "epoch": 14.28, "learning_rate": 3.935179831155314e-05, "loss": 0.9886, "step": 164900 }, { "epoch": 14.28, "learning_rate": 3.9349629929455304e-05, "loss": 0.9416, "step": 164910 }, { "epoch": 14.28, "learning_rate": 3.934746154735746e-05, "loss": 1.0167, "step": 164920 }, { "epoch": 14.28, "learning_rate": 3.934529316525963e-05, "loss": 0.8672, "step": 164930 }, { "epoch": 14.28, "learning_rate": 3.934312478316178e-05, "loss": 0.882, "step": 164940 }, { "epoch": 14.29, "learning_rate": 3.934095640106395e-05, "loss": 0.9561, "step": 164950 }, { "epoch": 14.29, "learning_rate": 3.933878801896611e-05, "loss": 1.0072, "step": 164960 }, { "epoch": 14.29, "learning_rate": 3.9336619636868275e-05, "loss": 0.9606, "step": 164970 }, { "epoch": 14.29, "learning_rate": 3.933445125477044e-05, "loss": 0.9233, "step": 164980 }, { "epoch": 14.29, "learning_rate": 3.93322828726726e-05, "loss": 0.7454, "step": 164990 }, { "epoch": 14.29, "learning_rate": 3.933011449057477e-05, "loss": 0.9878, "step": 165000 }, { "epoch": 14.29, "learning_rate": 3.932794610847692e-05, "loss": 0.9023, "step": 165010 }, { "epoch": 14.29, "learning_rate": 3.9325777726379087e-05, "loss": 0.9322, "step": 165020 }, { "epoch": 14.29, "learning_rate": 3.9323609344281246e-05, "loss": 0.9411, "step": 165030 }, { "epoch": 14.29, "learning_rate": 3.932144096218341e-05, "loss": 0.8234, "step": 165040 }, { "epoch": 14.29, "learning_rate": 3.931927258008558e-05, "loss": 0.9472, "step": 165050 }, { "epoch": 14.3, "learning_rate": 3.931710419798774e-05, "loss": 1.0191, "step": 165060 }, { "epoch": 14.3, "learning_rate": 3.9314935815889905e-05, "loss": 0.9139, "step": 165070 }, { "epoch": 14.3, "learning_rate": 3.9312767433792065e-05, "loss": 0.8604, "step": 165080 }, { "epoch": 14.3, "learning_rate": 3.9310599051694224e-05, "loss": 0.9558, "step": 165090 }, { "epoch": 14.3, "learning_rate": 3.9308430669596384e-05, "loss": 0.9089, "step": 165100 }, { "epoch": 14.3, "learning_rate": 3.930626228749855e-05, "loss": 0.9008, "step": 165110 }, { "epoch": 14.3, "learning_rate": 3.930409390540071e-05, "loss": 1.0757, "step": 165120 }, { "epoch": 14.3, "learning_rate": 3.9301925523302876e-05, "loss": 0.9204, "step": 165130 }, { "epoch": 14.3, "learning_rate": 3.929975714120504e-05, "loss": 0.85, "step": 165140 }, { "epoch": 14.3, "learning_rate": 3.92975887591072e-05, "loss": 1.0913, "step": 165150 }, { "epoch": 14.3, "learning_rate": 3.929542037700937e-05, "loss": 0.863, "step": 165160 }, { "epoch": 14.3, "learning_rate": 3.929325199491152e-05, "loss": 1.0249, "step": 165170 }, { "epoch": 14.31, "learning_rate": 3.929108361281369e-05, "loss": 0.9868, "step": 165180 }, { "epoch": 14.31, "learning_rate": 3.928891523071585e-05, "loss": 0.859, "step": 165190 }, { "epoch": 14.31, "learning_rate": 3.9286746848618014e-05, "loss": 1.0746, "step": 165200 }, { "epoch": 14.31, "learning_rate": 3.928457846652018e-05, "loss": 0.7929, "step": 165210 }, { "epoch": 14.31, "learning_rate": 3.928241008442234e-05, "loss": 1.0116, "step": 165220 }, { "epoch": 14.31, "learning_rate": 3.9280241702324506e-05, "loss": 1.02, "step": 165230 }, { "epoch": 14.31, "learning_rate": 3.927807332022666e-05, "loss": 0.87, "step": 165240 }, { "epoch": 14.31, "learning_rate": 3.9275904938128825e-05, "loss": 1.0156, "step": 165250 }, { "epoch": 14.31, "learning_rate": 3.9273736556030985e-05, "loss": 0.8178, "step": 165260 }, { "epoch": 14.31, "learning_rate": 3.927156817393315e-05, "loss": 1.073, "step": 165270 }, { "epoch": 14.31, "learning_rate": 3.926939979183532e-05, "loss": 0.9952, "step": 165280 }, { "epoch": 14.32, "learning_rate": 3.926723140973748e-05, "loss": 1.1129, "step": 165290 }, { "epoch": 14.32, "learning_rate": 3.9265063027639644e-05, "loss": 0.9799, "step": 165300 }, { "epoch": 14.32, "learning_rate": 3.9262894645541804e-05, "loss": 0.8975, "step": 165310 }, { "epoch": 14.32, "learning_rate": 3.926072626344396e-05, "loss": 0.9293, "step": 165320 }, { "epoch": 14.32, "learning_rate": 3.925855788134612e-05, "loss": 0.7056, "step": 165330 }, { "epoch": 14.32, "learning_rate": 3.925638949924829e-05, "loss": 1.0189, "step": 165340 }, { "epoch": 14.32, "learning_rate": 3.9254221117150456e-05, "loss": 0.8989, "step": 165350 }, { "epoch": 14.32, "learning_rate": 3.9252052735052615e-05, "loss": 1.0461, "step": 165360 }, { "epoch": 14.32, "learning_rate": 3.924988435295478e-05, "loss": 0.9762, "step": 165370 }, { "epoch": 14.32, "learning_rate": 3.924771597085694e-05, "loss": 0.8259, "step": 165380 }, { "epoch": 14.32, "learning_rate": 3.924554758875911e-05, "loss": 0.9724, "step": 165390 }, { "epoch": 14.32, "learning_rate": 3.924337920666126e-05, "loss": 1.1618, "step": 165400 }, { "epoch": 14.33, "learning_rate": 3.924121082456343e-05, "loss": 1.028, "step": 165410 }, { "epoch": 14.33, "learning_rate": 3.923904244246559e-05, "loss": 1.1006, "step": 165420 }, { "epoch": 14.33, "learning_rate": 3.923687406036775e-05, "loss": 1.1609, "step": 165430 }, { "epoch": 14.33, "learning_rate": 3.923470567826992e-05, "loss": 0.8588, "step": 165440 }, { "epoch": 14.33, "learning_rate": 3.923253729617208e-05, "loss": 1.0471, "step": 165450 }, { "epoch": 14.33, "learning_rate": 3.9230368914074245e-05, "loss": 1.0523, "step": 165460 }, { "epoch": 14.33, "learning_rate": 3.922820053197641e-05, "loss": 0.846, "step": 165470 }, { "epoch": 14.33, "learning_rate": 3.9226032149878564e-05, "loss": 0.8636, "step": 165480 }, { "epoch": 14.33, "learning_rate": 3.922386376778073e-05, "loss": 1.0292, "step": 165490 }, { "epoch": 14.33, "learning_rate": 3.922169538568289e-05, "loss": 0.9757, "step": 165500 }, { "epoch": 14.33, "learning_rate": 3.921952700358506e-05, "loss": 1.028, "step": 165510 }, { "epoch": 14.34, "learning_rate": 3.9217358621487216e-05, "loss": 0.9439, "step": 165520 }, { "epoch": 14.34, "learning_rate": 3.921519023938938e-05, "loss": 0.8404, "step": 165530 }, { "epoch": 14.34, "learning_rate": 3.921302185729155e-05, "loss": 1.0291, "step": 165540 }, { "epoch": 14.34, "learning_rate": 3.92108534751937e-05, "loss": 0.9005, "step": 165550 }, { "epoch": 14.34, "learning_rate": 3.920868509309587e-05, "loss": 0.9126, "step": 165560 }, { "epoch": 14.34, "learning_rate": 3.920651671099803e-05, "loss": 0.7511, "step": 165570 }, { "epoch": 14.34, "learning_rate": 3.9204348328900194e-05, "loss": 1.006, "step": 165580 }, { "epoch": 14.34, "learning_rate": 3.9202179946802354e-05, "loss": 1.1622, "step": 165590 }, { "epoch": 14.34, "learning_rate": 3.920001156470452e-05, "loss": 1.1356, "step": 165600 }, { "epoch": 14.34, "learning_rate": 3.919784318260669e-05, "loss": 0.9366, "step": 165610 }, { "epoch": 14.34, "learning_rate": 3.9195674800508846e-05, "loss": 0.7931, "step": 165620 }, { "epoch": 14.34, "learning_rate": 3.9193506418411e-05, "loss": 0.9608, "step": 165630 }, { "epoch": 14.35, "learning_rate": 3.9191338036313166e-05, "loss": 1.0836, "step": 165640 }, { "epoch": 14.35, "learning_rate": 3.918916965421533e-05, "loss": 0.7709, "step": 165650 }, { "epoch": 14.35, "learning_rate": 3.918700127211749e-05, "loss": 1.0305, "step": 165660 }, { "epoch": 14.35, "learning_rate": 3.918483289001966e-05, "loss": 0.9999, "step": 165670 }, { "epoch": 14.35, "learning_rate": 3.918266450792182e-05, "loss": 0.9315, "step": 165680 }, { "epoch": 14.35, "learning_rate": 3.9180496125823984e-05, "loss": 0.9909, "step": 165690 }, { "epoch": 14.35, "learning_rate": 3.917832774372615e-05, "loss": 1.0099, "step": 165700 }, { "epoch": 14.35, "learning_rate": 3.91761593616283e-05, "loss": 1.0441, "step": 165710 }, { "epoch": 14.35, "learning_rate": 3.917399097953047e-05, "loss": 0.9944, "step": 165720 }, { "epoch": 14.35, "learning_rate": 3.917182259743263e-05, "loss": 0.9662, "step": 165730 }, { "epoch": 14.35, "learning_rate": 3.9169654215334796e-05, "loss": 1.0554, "step": 165740 }, { "epoch": 14.35, "learning_rate": 3.9167485833236955e-05, "loss": 1.0361, "step": 165750 }, { "epoch": 14.36, "learning_rate": 3.916531745113912e-05, "loss": 1.1816, "step": 165760 }, { "epoch": 14.36, "learning_rate": 3.916314906904129e-05, "loss": 0.7594, "step": 165770 }, { "epoch": 14.36, "learning_rate": 3.916098068694344e-05, "loss": 0.8855, "step": 165780 }, { "epoch": 14.36, "learning_rate": 3.915881230484561e-05, "loss": 0.9309, "step": 165790 }, { "epoch": 14.36, "learning_rate": 3.915664392274777e-05, "loss": 0.9999, "step": 165800 }, { "epoch": 14.36, "learning_rate": 3.915447554064993e-05, "loss": 0.9137, "step": 165810 }, { "epoch": 14.36, "learning_rate": 3.915230715855209e-05, "loss": 0.847, "step": 165820 }, { "epoch": 14.36, "learning_rate": 3.915013877645426e-05, "loss": 0.9506, "step": 165830 }, { "epoch": 14.36, "learning_rate": 3.9147970394356426e-05, "loss": 0.969, "step": 165840 }, { "epoch": 14.36, "learning_rate": 3.9145802012258585e-05, "loss": 0.8252, "step": 165850 }, { "epoch": 14.36, "learning_rate": 3.9143633630160745e-05, "loss": 0.8813, "step": 165860 }, { "epoch": 14.37, "learning_rate": 3.9141465248062905e-05, "loss": 0.7731, "step": 165870 }, { "epoch": 14.37, "learning_rate": 3.913929686596507e-05, "loss": 0.9377, "step": 165880 }, { "epoch": 14.37, "learning_rate": 3.913712848386723e-05, "loss": 0.859, "step": 165890 }, { "epoch": 14.37, "learning_rate": 3.91349601017694e-05, "loss": 1.0028, "step": 165900 }, { "epoch": 14.37, "learning_rate": 3.913279171967156e-05, "loss": 0.8565, "step": 165910 }, { "epoch": 14.37, "learning_rate": 3.913062333757372e-05, "loss": 1.0525, "step": 165920 }, { "epoch": 14.37, "learning_rate": 3.912845495547589e-05, "loss": 0.9643, "step": 165930 }, { "epoch": 14.37, "learning_rate": 3.912628657337804e-05, "loss": 1.0109, "step": 165940 }, { "epoch": 14.37, "learning_rate": 3.912411819128021e-05, "loss": 0.9297, "step": 165950 }, { "epoch": 14.37, "learning_rate": 3.912194980918237e-05, "loss": 0.9899, "step": 165960 }, { "epoch": 14.37, "learning_rate": 3.9119781427084535e-05, "loss": 1.0044, "step": 165970 }, { "epoch": 14.37, "learning_rate": 3.91176130449867e-05, "loss": 0.8755, "step": 165980 }, { "epoch": 14.38, "learning_rate": 3.911544466288886e-05, "loss": 1.2894, "step": 165990 }, { "epoch": 14.38, "learning_rate": 3.911327628079103e-05, "loss": 1.0073, "step": 166000 }, { "epoch": 14.38, "learning_rate": 3.911110789869318e-05, "loss": 1.0142, "step": 166010 }, { "epoch": 14.38, "learning_rate": 3.9108939516595346e-05, "loss": 0.9867, "step": 166020 }, { "epoch": 14.38, "learning_rate": 3.9106771134497506e-05, "loss": 0.9696, "step": 166030 }, { "epoch": 14.38, "learning_rate": 3.910460275239967e-05, "loss": 0.9858, "step": 166040 }, { "epoch": 14.38, "learning_rate": 3.910243437030184e-05, "loss": 0.9888, "step": 166050 }, { "epoch": 14.38, "learning_rate": 3.9100265988204e-05, "loss": 0.9869, "step": 166060 }, { "epoch": 14.38, "learning_rate": 3.9098097606106165e-05, "loss": 0.9809, "step": 166070 }, { "epoch": 14.38, "learning_rate": 3.9095929224008324e-05, "loss": 0.9851, "step": 166080 }, { "epoch": 14.38, "learning_rate": 3.9093760841910484e-05, "loss": 0.9115, "step": 166090 }, { "epoch": 14.39, "learning_rate": 3.9091592459812644e-05, "loss": 0.8717, "step": 166100 }, { "epoch": 14.39, "learning_rate": 3.908942407771481e-05, "loss": 0.8589, "step": 166110 }, { "epoch": 14.39, "learning_rate": 3.9087255695616976e-05, "loss": 0.9636, "step": 166120 }, { "epoch": 14.39, "learning_rate": 3.9085087313519136e-05, "loss": 1.1259, "step": 166130 }, { "epoch": 14.39, "learning_rate": 3.90829189314213e-05, "loss": 1.0434, "step": 166140 }, { "epoch": 14.39, "learning_rate": 3.908075054932346e-05, "loss": 0.9052, "step": 166150 }, { "epoch": 14.39, "learning_rate": 3.907858216722563e-05, "loss": 0.9939, "step": 166160 }, { "epoch": 14.39, "learning_rate": 3.907641378512778e-05, "loss": 0.9589, "step": 166170 }, { "epoch": 14.39, "learning_rate": 3.907424540302995e-05, "loss": 1.0597, "step": 166180 }, { "epoch": 14.39, "learning_rate": 3.907207702093211e-05, "loss": 0.9861, "step": 166190 }, { "epoch": 14.39, "learning_rate": 3.9069908638834274e-05, "loss": 0.7885, "step": 166200 }, { "epoch": 14.39, "learning_rate": 3.906774025673644e-05, "loss": 0.948, "step": 166210 }, { "epoch": 14.4, "learning_rate": 3.90655718746386e-05, "loss": 0.9554, "step": 166220 }, { "epoch": 14.4, "learning_rate": 3.9063403492540766e-05, "loss": 1.1067, "step": 166230 }, { "epoch": 14.4, "learning_rate": 3.9061235110442926e-05, "loss": 0.9725, "step": 166240 }, { "epoch": 14.4, "learning_rate": 3.9059066728345085e-05, "loss": 0.8622, "step": 166250 }, { "epoch": 14.4, "learning_rate": 3.9056898346247245e-05, "loss": 0.946, "step": 166260 }, { "epoch": 14.4, "learning_rate": 3.905472996414941e-05, "loss": 1.0448, "step": 166270 }, { "epoch": 14.4, "learning_rate": 3.905256158205158e-05, "loss": 0.8839, "step": 166280 }, { "epoch": 14.4, "learning_rate": 3.905039319995374e-05, "loss": 1.0949, "step": 166290 }, { "epoch": 14.4, "learning_rate": 3.9048224817855904e-05, "loss": 0.9838, "step": 166300 }, { "epoch": 14.4, "learning_rate": 3.904605643575806e-05, "loss": 1.0647, "step": 166310 }, { "epoch": 14.4, "learning_rate": 3.904388805366022e-05, "loss": 1.0549, "step": 166320 }, { "epoch": 14.41, "learning_rate": 3.904171967156238e-05, "loss": 0.9585, "step": 166330 }, { "epoch": 14.41, "learning_rate": 3.903955128946455e-05, "loss": 0.7865, "step": 166340 }, { "epoch": 14.41, "learning_rate": 3.9037382907366715e-05, "loss": 0.9551, "step": 166350 }, { "epoch": 14.41, "learning_rate": 3.9035214525268875e-05, "loss": 0.9983, "step": 166360 }, { "epoch": 14.41, "learning_rate": 3.903304614317104e-05, "loss": 1.0651, "step": 166370 }, { "epoch": 14.41, "learning_rate": 3.90308777610732e-05, "loss": 1.0677, "step": 166380 }, { "epoch": 14.41, "learning_rate": 3.902870937897537e-05, "loss": 0.9383, "step": 166390 }, { "epoch": 14.41, "learning_rate": 3.902654099687752e-05, "loss": 0.9895, "step": 166400 }, { "epoch": 14.41, "learning_rate": 3.9024372614779686e-05, "loss": 0.9368, "step": 166410 }, { "epoch": 14.41, "learning_rate": 3.902220423268185e-05, "loss": 1.1309, "step": 166420 }, { "epoch": 14.41, "learning_rate": 3.902003585058401e-05, "loss": 0.9591, "step": 166430 }, { "epoch": 14.41, "learning_rate": 3.901786746848618e-05, "loss": 0.887, "step": 166440 }, { "epoch": 14.42, "learning_rate": 3.901569908638834e-05, "loss": 1.2421, "step": 166450 }, { "epoch": 14.42, "learning_rate": 3.9013530704290505e-05, "loss": 0.9875, "step": 166460 }, { "epoch": 14.42, "learning_rate": 3.901136232219267e-05, "loss": 0.8142, "step": 166470 }, { "epoch": 14.42, "learning_rate": 3.9009193940094824e-05, "loss": 0.8909, "step": 166480 }, { "epoch": 14.42, "learning_rate": 3.900702555799699e-05, "loss": 0.9468, "step": 166490 }, { "epoch": 14.42, "learning_rate": 3.900485717589915e-05, "loss": 0.876, "step": 166500 }, { "epoch": 14.42, "learning_rate": 3.9002688793801316e-05, "loss": 0.9585, "step": 166510 }, { "epoch": 14.42, "learning_rate": 3.9000520411703476e-05, "loss": 1.0244, "step": 166520 }, { "epoch": 14.42, "learning_rate": 3.899835202960564e-05, "loss": 0.8171, "step": 166530 }, { "epoch": 14.42, "learning_rate": 3.899618364750781e-05, "loss": 0.8747, "step": 166540 }, { "epoch": 14.42, "learning_rate": 3.899401526540996e-05, "loss": 1.1095, "step": 166550 }, { "epoch": 14.43, "learning_rate": 3.899184688331213e-05, "loss": 1.0025, "step": 166560 }, { "epoch": 14.43, "learning_rate": 3.898967850121429e-05, "loss": 1.0024, "step": 166570 }, { "epoch": 14.43, "learning_rate": 3.8987510119116454e-05, "loss": 1.0044, "step": 166580 }, { "epoch": 14.43, "learning_rate": 3.8985341737018614e-05, "loss": 0.9839, "step": 166590 }, { "epoch": 14.43, "learning_rate": 3.898317335492078e-05, "loss": 0.9996, "step": 166600 }, { "epoch": 14.43, "learning_rate": 3.8981004972822947e-05, "loss": 0.8177, "step": 166610 }, { "epoch": 14.43, "learning_rate": 3.8978836590725106e-05, "loss": 0.8336, "step": 166620 }, { "epoch": 14.43, "learning_rate": 3.8976668208627266e-05, "loss": 0.9831, "step": 166630 }, { "epoch": 14.43, "learning_rate": 3.8974499826529425e-05, "loss": 0.9011, "step": 166640 }, { "epoch": 14.43, "learning_rate": 3.897233144443159e-05, "loss": 1.0108, "step": 166650 }, { "epoch": 14.43, "learning_rate": 3.897016306233375e-05, "loss": 0.9952, "step": 166660 }, { "epoch": 14.43, "learning_rate": 3.896799468023592e-05, "loss": 1.0499, "step": 166670 }, { "epoch": 14.44, "learning_rate": 3.8965826298138084e-05, "loss": 1.02, "step": 166680 }, { "epoch": 14.44, "learning_rate": 3.8963657916040244e-05, "loss": 0.8599, "step": 166690 }, { "epoch": 14.44, "learning_rate": 3.896148953394241e-05, "loss": 0.8249, "step": 166700 }, { "epoch": 14.44, "learning_rate": 3.895932115184456e-05, "loss": 1.0397, "step": 166710 }, { "epoch": 14.44, "learning_rate": 3.895715276974673e-05, "loss": 0.9074, "step": 166720 }, { "epoch": 14.44, "learning_rate": 3.895498438764889e-05, "loss": 1.0707, "step": 166730 }, { "epoch": 14.44, "learning_rate": 3.8952816005551055e-05, "loss": 0.9298, "step": 166740 }, { "epoch": 14.44, "learning_rate": 3.8950647623453215e-05, "loss": 0.9126, "step": 166750 }, { "epoch": 14.44, "learning_rate": 3.894847924135538e-05, "loss": 1.0143, "step": 166760 }, { "epoch": 14.44, "learning_rate": 3.894631085925755e-05, "loss": 0.8413, "step": 166770 }, { "epoch": 14.44, "learning_rate": 3.89441424771597e-05, "loss": 0.7814, "step": 166780 }, { "epoch": 14.45, "learning_rate": 3.894197409506187e-05, "loss": 1.0375, "step": 166790 }, { "epoch": 14.45, "learning_rate": 3.893980571296403e-05, "loss": 0.901, "step": 166800 }, { "epoch": 14.45, "learning_rate": 3.893763733086619e-05, "loss": 0.9125, "step": 166810 }, { "epoch": 14.45, "learning_rate": 3.893546894876835e-05, "loss": 0.9537, "step": 166820 }, { "epoch": 14.45, "learning_rate": 3.893330056667052e-05, "loss": 0.726, "step": 166830 }, { "epoch": 14.45, "learning_rate": 3.8931132184572685e-05, "loss": 0.8269, "step": 166840 }, { "epoch": 14.45, "learning_rate": 3.8928963802474845e-05, "loss": 0.907, "step": 166850 }, { "epoch": 14.45, "learning_rate": 3.8926795420377005e-05, "loss": 1.1785, "step": 166860 }, { "epoch": 14.45, "learning_rate": 3.8924627038279164e-05, "loss": 0.9617, "step": 166870 }, { "epoch": 14.45, "learning_rate": 3.892245865618133e-05, "loss": 0.8888, "step": 166880 }, { "epoch": 14.45, "learning_rate": 3.892029027408349e-05, "loss": 0.98, "step": 166890 }, { "epoch": 14.45, "learning_rate": 3.891812189198566e-05, "loss": 0.9269, "step": 166900 }, { "epoch": 14.46, "learning_rate": 3.891595350988782e-05, "loss": 1.1769, "step": 166910 }, { "epoch": 14.46, "learning_rate": 3.891378512778998e-05, "loss": 1.0111, "step": 166920 }, { "epoch": 14.46, "learning_rate": 3.891161674569215e-05, "loss": 1.1927, "step": 166930 }, { "epoch": 14.46, "learning_rate": 3.89094483635943e-05, "loss": 0.9544, "step": 166940 }, { "epoch": 14.46, "learning_rate": 3.890727998149647e-05, "loss": 0.9931, "step": 166950 }, { "epoch": 14.46, "learning_rate": 3.890511159939863e-05, "loss": 0.9167, "step": 166960 }, { "epoch": 14.46, "learning_rate": 3.8902943217300794e-05, "loss": 0.8903, "step": 166970 }, { "epoch": 14.46, "learning_rate": 3.890077483520296e-05, "loss": 0.8611, "step": 166980 }, { "epoch": 14.46, "learning_rate": 3.889860645310512e-05, "loss": 1.2083, "step": 166990 }, { "epoch": 14.46, "learning_rate": 3.889643807100729e-05, "loss": 0.9626, "step": 167000 }, { "epoch": 14.46, "learning_rate": 3.8894269688909446e-05, "loss": 0.963, "step": 167010 }, { "epoch": 14.46, "learning_rate": 3.8892101306811606e-05, "loss": 0.9524, "step": 167020 }, { "epoch": 14.47, "learning_rate": 3.8889932924713766e-05, "loss": 1.004, "step": 167030 }, { "epoch": 14.47, "learning_rate": 3.888776454261593e-05, "loss": 0.9067, "step": 167040 }, { "epoch": 14.47, "learning_rate": 3.88855961605181e-05, "loss": 0.9964, "step": 167050 }, { "epoch": 14.47, "learning_rate": 3.888342777842026e-05, "loss": 1.0501, "step": 167060 }, { "epoch": 14.47, "learning_rate": 3.8881259396322424e-05, "loss": 0.8247, "step": 167070 }, { "epoch": 14.47, "learning_rate": 3.8879091014224584e-05, "loss": 0.8683, "step": 167080 }, { "epoch": 14.47, "learning_rate": 3.8876922632126744e-05, "loss": 0.9247, "step": 167090 }, { "epoch": 14.47, "learning_rate": 3.88747542500289e-05, "loss": 1.0054, "step": 167100 }, { "epoch": 14.47, "learning_rate": 3.887258586793107e-05, "loss": 1.0768, "step": 167110 }, { "epoch": 14.47, "learning_rate": 3.8870417485833236e-05, "loss": 0.9835, "step": 167120 }, { "epoch": 14.47, "learning_rate": 3.8868249103735396e-05, "loss": 0.9604, "step": 167130 }, { "epoch": 14.48, "learning_rate": 3.886608072163756e-05, "loss": 0.9521, "step": 167140 }, { "epoch": 14.48, "learning_rate": 3.886391233953972e-05, "loss": 0.9559, "step": 167150 }, { "epoch": 14.48, "learning_rate": 3.886174395744189e-05, "loss": 0.8893, "step": 167160 }, { "epoch": 14.48, "learning_rate": 3.885957557534404e-05, "loss": 0.9105, "step": 167170 }, { "epoch": 14.48, "learning_rate": 3.885740719324621e-05, "loss": 0.9714, "step": 167180 }, { "epoch": 14.48, "learning_rate": 3.8855238811148374e-05, "loss": 0.9169, "step": 167190 }, { "epoch": 14.48, "learning_rate": 3.885307042905053e-05, "loss": 0.9601, "step": 167200 }, { "epoch": 14.48, "learning_rate": 3.88509020469527e-05, "loss": 0.9381, "step": 167210 }, { "epoch": 14.48, "learning_rate": 3.884873366485486e-05, "loss": 0.9137, "step": 167220 }, { "epoch": 14.48, "learning_rate": 3.8846565282757026e-05, "loss": 0.8595, "step": 167230 }, { "epoch": 14.48, "learning_rate": 3.884439690065919e-05, "loss": 0.994, "step": 167240 }, { "epoch": 14.48, "learning_rate": 3.8842228518561345e-05, "loss": 0.8721, "step": 167250 }, { "epoch": 14.49, "learning_rate": 3.8840060136463504e-05, "loss": 0.9898, "step": 167260 }, { "epoch": 14.49, "learning_rate": 3.883789175436567e-05, "loss": 0.8529, "step": 167270 }, { "epoch": 14.49, "learning_rate": 3.883572337226784e-05, "loss": 1.0526, "step": 167280 }, { "epoch": 14.49, "learning_rate": 3.883355499017e-05, "loss": 0.9652, "step": 167290 }, { "epoch": 14.49, "learning_rate": 3.883138660807216e-05, "loss": 0.908, "step": 167300 }, { "epoch": 14.49, "learning_rate": 3.882921822597432e-05, "loss": 0.8959, "step": 167310 }, { "epoch": 14.49, "learning_rate": 3.882704984387648e-05, "loss": 1.1258, "step": 167320 }, { "epoch": 14.49, "learning_rate": 3.882488146177864e-05, "loss": 1.0158, "step": 167330 }, { "epoch": 14.49, "learning_rate": 3.882271307968081e-05, "loss": 0.9585, "step": 167340 }, { "epoch": 14.49, "learning_rate": 3.8820544697582975e-05, "loss": 1.0893, "step": 167350 }, { "epoch": 14.49, "learning_rate": 3.8818376315485135e-05, "loss": 0.8444, "step": 167360 }, { "epoch": 14.5, "learning_rate": 3.88162079333873e-05, "loss": 0.9374, "step": 167370 }, { "epoch": 14.5, "learning_rate": 3.881403955128946e-05, "loss": 1.079, "step": 167380 }, { "epoch": 14.5, "learning_rate": 3.881187116919163e-05, "loss": 1.0739, "step": 167390 }, { "epoch": 14.5, "learning_rate": 3.880970278709378e-05, "loss": 0.8192, "step": 167400 }, { "epoch": 14.5, "learning_rate": 3.8807534404995946e-05, "loss": 1.0053, "step": 167410 }, { "epoch": 14.5, "learning_rate": 3.880536602289811e-05, "loss": 0.7358, "step": 167420 }, { "epoch": 14.5, "learning_rate": 3.880319764080027e-05, "loss": 1.0707, "step": 167430 }, { "epoch": 14.5, "learning_rate": 3.880102925870244e-05, "loss": 1.1771, "step": 167440 }, { "epoch": 14.5, "learning_rate": 3.87988608766046e-05, "loss": 0.9022, "step": 167450 }, { "epoch": 14.5, "learning_rate": 3.8796692494506765e-05, "loss": 1.0682, "step": 167460 }, { "epoch": 14.5, "learning_rate": 3.879452411240893e-05, "loss": 1.2146, "step": 167470 }, { "epoch": 14.5, "learning_rate": 3.8792355730311084e-05, "loss": 0.9369, "step": 167480 }, { "epoch": 14.51, "learning_rate": 3.879018734821325e-05, "loss": 0.8597, "step": 167490 }, { "epoch": 14.51, "learning_rate": 3.878801896611541e-05, "loss": 1.0782, "step": 167500 }, { "epoch": 14.51, "learning_rate": 3.8785850584017576e-05, "loss": 0.9864, "step": 167510 }, { "epoch": 14.51, "learning_rate": 3.8783682201919736e-05, "loss": 1.0456, "step": 167520 }, { "epoch": 14.51, "learning_rate": 3.87815138198219e-05, "loss": 1.0138, "step": 167530 }, { "epoch": 14.51, "learning_rate": 3.877934543772407e-05, "loss": 0.7985, "step": 167540 }, { "epoch": 14.51, "learning_rate": 3.877717705562622e-05, "loss": 0.8878, "step": 167550 }, { "epoch": 14.51, "learning_rate": 3.877500867352839e-05, "loss": 1.1329, "step": 167560 }, { "epoch": 14.51, "learning_rate": 3.877284029143055e-05, "loss": 1.0714, "step": 167570 }, { "epoch": 14.51, "learning_rate": 3.8770671909332714e-05, "loss": 0.8802, "step": 167580 }, { "epoch": 14.51, "learning_rate": 3.8768503527234873e-05, "loss": 0.9148, "step": 167590 }, { "epoch": 14.52, "learning_rate": 3.876633514513704e-05, "loss": 0.9639, "step": 167600 }, { "epoch": 14.52, "learning_rate": 3.8764166763039206e-05, "loss": 0.9633, "step": 167610 }, { "epoch": 14.52, "learning_rate": 3.8761998380941366e-05, "loss": 0.8739, "step": 167620 }, { "epoch": 14.52, "learning_rate": 3.8759829998843525e-05, "loss": 1.1203, "step": 167630 }, { "epoch": 14.52, "learning_rate": 3.8757661616745685e-05, "loss": 0.9039, "step": 167640 }, { "epoch": 14.52, "learning_rate": 3.875549323464785e-05, "loss": 0.944, "step": 167650 }, { "epoch": 14.52, "learning_rate": 3.875332485255001e-05, "loss": 0.8811, "step": 167660 }, { "epoch": 14.52, "learning_rate": 3.875115647045218e-05, "loss": 0.9101, "step": 167670 }, { "epoch": 14.52, "learning_rate": 3.8748988088354344e-05, "loss": 1.0754, "step": 167680 }, { "epoch": 14.52, "learning_rate": 3.8746819706256503e-05, "loss": 0.9294, "step": 167690 }, { "epoch": 14.52, "learning_rate": 3.874465132415867e-05, "loss": 0.8366, "step": 167700 }, { "epoch": 14.52, "learning_rate": 3.874248294206082e-05, "loss": 0.904, "step": 167710 }, { "epoch": 14.53, "learning_rate": 3.874031455996299e-05, "loss": 1.0546, "step": 167720 }, { "epoch": 14.53, "learning_rate": 3.873814617786515e-05, "loss": 0.8683, "step": 167730 }, { "epoch": 14.53, "learning_rate": 3.8735977795767315e-05, "loss": 1.1701, "step": 167740 }, { "epoch": 14.53, "learning_rate": 3.873380941366948e-05, "loss": 0.9536, "step": 167750 }, { "epoch": 14.53, "learning_rate": 3.873164103157164e-05, "loss": 0.8467, "step": 167760 }, { "epoch": 14.53, "learning_rate": 3.872947264947381e-05, "loss": 1.0438, "step": 167770 }, { "epoch": 14.53, "learning_rate": 3.872730426737597e-05, "loss": 0.9706, "step": 167780 }, { "epoch": 14.53, "learning_rate": 3.872513588527813e-05, "loss": 0.7356, "step": 167790 }, { "epoch": 14.53, "learning_rate": 3.8722967503180286e-05, "loss": 0.9473, "step": 167800 }, { "epoch": 14.53, "learning_rate": 3.872079912108245e-05, "loss": 0.919, "step": 167810 }, { "epoch": 14.53, "learning_rate": 3.871863073898461e-05, "loss": 1.1988, "step": 167820 }, { "epoch": 14.54, "learning_rate": 3.871646235688678e-05, "loss": 0.8885, "step": 167830 }, { "epoch": 14.54, "learning_rate": 3.8714293974788945e-05, "loss": 0.8915, "step": 167840 }, { "epoch": 14.54, "learning_rate": 3.8712125592691105e-05, "loss": 0.9304, "step": 167850 }, { "epoch": 14.54, "learning_rate": 3.8709957210593264e-05, "loss": 0.9837, "step": 167860 }, { "epoch": 14.54, "learning_rate": 3.8707788828495424e-05, "loss": 1.0742, "step": 167870 }, { "epoch": 14.54, "learning_rate": 3.870562044639759e-05, "loss": 0.8111, "step": 167880 }, { "epoch": 14.54, "learning_rate": 3.870345206429975e-05, "loss": 0.9734, "step": 167890 }, { "epoch": 14.54, "learning_rate": 3.8701283682201916e-05, "loss": 0.9532, "step": 167900 }, { "epoch": 14.54, "learning_rate": 3.869911530010408e-05, "loss": 1.0235, "step": 167910 }, { "epoch": 14.54, "learning_rate": 3.869694691800624e-05, "loss": 1.0135, "step": 167920 }, { "epoch": 14.54, "learning_rate": 3.869477853590841e-05, "loss": 0.9799, "step": 167930 }, { "epoch": 14.54, "learning_rate": 3.869261015381056e-05, "loss": 0.9747, "step": 167940 }, { "epoch": 14.55, "learning_rate": 3.869044177171273e-05, "loss": 0.9953, "step": 167950 }, { "epoch": 14.55, "learning_rate": 3.868827338961489e-05, "loss": 0.9127, "step": 167960 }, { "epoch": 14.55, "learning_rate": 3.8686105007517054e-05, "loss": 0.9801, "step": 167970 }, { "epoch": 14.55, "learning_rate": 3.868393662541922e-05, "loss": 0.9635, "step": 167980 }, { "epoch": 14.55, "learning_rate": 3.868176824332138e-05, "loss": 1.1232, "step": 167990 }, { "epoch": 14.55, "learning_rate": 3.8679599861223546e-05, "loss": 0.8715, "step": 168000 }, { "epoch": 14.55, "learning_rate": 3.8677431479125706e-05, "loss": 0.9838, "step": 168010 }, { "epoch": 14.55, "learning_rate": 3.8675263097027866e-05, "loss": 0.7708, "step": 168020 }, { "epoch": 14.55, "learning_rate": 3.8673094714930025e-05, "loss": 0.9029, "step": 168030 }, { "epoch": 14.55, "learning_rate": 3.867092633283219e-05, "loss": 0.8743, "step": 168040 }, { "epoch": 14.55, "learning_rate": 3.866875795073436e-05, "loss": 0.909, "step": 168050 }, { "epoch": 14.56, "learning_rate": 3.866658956863652e-05, "loss": 1.0201, "step": 168060 }, { "epoch": 14.56, "learning_rate": 3.8664421186538684e-05, "loss": 0.9299, "step": 168070 }, { "epoch": 14.56, "learning_rate": 3.8662252804440844e-05, "loss": 0.8117, "step": 168080 }, { "epoch": 14.56, "learning_rate": 3.8660084422343e-05, "loss": 1.0141, "step": 168090 }, { "epoch": 14.56, "learning_rate": 3.865791604024516e-05, "loss": 0.9965, "step": 168100 }, { "epoch": 14.56, "learning_rate": 3.865574765814733e-05, "loss": 0.8701, "step": 168110 }, { "epoch": 14.56, "learning_rate": 3.8653579276049496e-05, "loss": 1.1753, "step": 168120 }, { "epoch": 14.56, "learning_rate": 3.8651410893951655e-05, "loss": 0.9386, "step": 168130 }, { "epoch": 14.56, "learning_rate": 3.864924251185382e-05, "loss": 0.9537, "step": 168140 }, { "epoch": 14.56, "learning_rate": 3.864707412975598e-05, "loss": 0.8325, "step": 168150 }, { "epoch": 14.56, "learning_rate": 3.864490574765815e-05, "loss": 0.9201, "step": 168160 }, { "epoch": 14.56, "learning_rate": 3.86427373655603e-05, "loss": 0.8579, "step": 168170 }, { "epoch": 14.57, "learning_rate": 3.864056898346247e-05, "loss": 1.0103, "step": 168180 }, { "epoch": 14.57, "learning_rate": 3.863840060136463e-05, "loss": 0.9424, "step": 168190 }, { "epoch": 14.57, "learning_rate": 3.863623221926679e-05, "loss": 0.8517, "step": 168200 }, { "epoch": 14.57, "learning_rate": 3.863406383716896e-05, "loss": 0.9633, "step": 168210 }, { "epoch": 14.57, "learning_rate": 3.863189545507112e-05, "loss": 1.0345, "step": 168220 }, { "epoch": 14.57, "learning_rate": 3.8629727072973285e-05, "loss": 0.9815, "step": 168230 }, { "epoch": 14.57, "learning_rate": 3.862755869087545e-05, "loss": 1.0025, "step": 168240 }, { "epoch": 14.57, "learning_rate": 3.8625390308777605e-05, "loss": 0.9528, "step": 168250 }, { "epoch": 14.57, "learning_rate": 3.862322192667977e-05, "loss": 0.9202, "step": 168260 }, { "epoch": 14.57, "learning_rate": 3.862105354458193e-05, "loss": 0.9472, "step": 168270 }, { "epoch": 14.57, "learning_rate": 3.86188851624841e-05, "loss": 0.8724, "step": 168280 }, { "epoch": 14.57, "learning_rate": 3.8616716780386257e-05, "loss": 1.047, "step": 168290 }, { "epoch": 14.58, "learning_rate": 3.861454839828842e-05, "loss": 0.974, "step": 168300 }, { "epoch": 14.58, "learning_rate": 3.861238001619059e-05, "loss": 1.0211, "step": 168310 }, { "epoch": 14.58, "learning_rate": 3.861021163409274e-05, "loss": 1.1034, "step": 168320 }, { "epoch": 14.58, "learning_rate": 3.86080432519949e-05, "loss": 0.9059, "step": 168330 }, { "epoch": 14.58, "learning_rate": 3.860587486989707e-05, "loss": 0.9844, "step": 168340 }, { "epoch": 14.58, "learning_rate": 3.8603706487799235e-05, "loss": 0.8469, "step": 168350 }, { "epoch": 14.58, "learning_rate": 3.8601538105701394e-05, "loss": 0.9345, "step": 168360 }, { "epoch": 14.58, "learning_rate": 3.859936972360356e-05, "loss": 0.7887, "step": 168370 }, { "epoch": 14.58, "learning_rate": 3.859720134150572e-05, "loss": 1.0835, "step": 168380 }, { "epoch": 14.58, "learning_rate": 3.859503295940789e-05, "loss": 0.9588, "step": 168390 }, { "epoch": 14.58, "learning_rate": 3.859286457731004e-05, "loss": 0.8375, "step": 168400 }, { "epoch": 14.59, "learning_rate": 3.8590696195212206e-05, "loss": 1.1243, "step": 168410 }, { "epoch": 14.59, "learning_rate": 3.858852781311437e-05, "loss": 0.9309, "step": 168420 }, { "epoch": 14.59, "learning_rate": 3.858635943101653e-05, "loss": 0.9606, "step": 168430 }, { "epoch": 14.59, "learning_rate": 3.85841910489187e-05, "loss": 0.8637, "step": 168440 }, { "epoch": 14.59, "learning_rate": 3.858202266682086e-05, "loss": 1.0311, "step": 168450 }, { "epoch": 14.59, "learning_rate": 3.8579854284723024e-05, "loss": 0.8362, "step": 168460 }, { "epoch": 14.59, "learning_rate": 3.857768590262519e-05, "loss": 0.9517, "step": 168470 }, { "epoch": 14.59, "learning_rate": 3.8575517520527343e-05, "loss": 0.9578, "step": 168480 }, { "epoch": 14.59, "learning_rate": 3.857334913842951e-05, "loss": 1.0818, "step": 168490 }, { "epoch": 14.59, "learning_rate": 3.857118075633167e-05, "loss": 0.7623, "step": 168500 }, { "epoch": 14.59, "learning_rate": 3.8569012374233836e-05, "loss": 0.7294, "step": 168510 }, { "epoch": 14.59, "learning_rate": 3.8566843992135995e-05, "loss": 0.9197, "step": 168520 }, { "epoch": 14.6, "learning_rate": 3.856467561003816e-05, "loss": 0.7534, "step": 168530 }, { "epoch": 14.6, "learning_rate": 3.856250722794033e-05, "loss": 0.9418, "step": 168540 }, { "epoch": 14.6, "learning_rate": 3.856033884584249e-05, "loss": 0.723, "step": 168550 }, { "epoch": 14.6, "learning_rate": 3.855817046374465e-05, "loss": 0.9172, "step": 168560 }, { "epoch": 14.6, "learning_rate": 3.855600208164681e-05, "loss": 0.9637, "step": 168570 }, { "epoch": 14.6, "learning_rate": 3.8553833699548974e-05, "loss": 0.9794, "step": 168580 }, { "epoch": 14.6, "learning_rate": 3.855166531745113e-05, "loss": 0.8644, "step": 168590 }, { "epoch": 14.6, "learning_rate": 3.85494969353533e-05, "loss": 1.0193, "step": 168600 }, { "epoch": 14.6, "learning_rate": 3.8547328553255466e-05, "loss": 1.0832, "step": 168610 }, { "epoch": 14.6, "learning_rate": 3.8545160171157626e-05, "loss": 0.8658, "step": 168620 }, { "epoch": 14.6, "learning_rate": 3.8542991789059785e-05, "loss": 0.9334, "step": 168630 }, { "epoch": 14.61, "learning_rate": 3.8540823406961945e-05, "loss": 1.0065, "step": 168640 }, { "epoch": 14.61, "learning_rate": 3.853865502486411e-05, "loss": 0.8931, "step": 168650 }, { "epoch": 14.61, "learning_rate": 3.853648664276627e-05, "loss": 1.0578, "step": 168660 }, { "epoch": 14.61, "learning_rate": 3.853431826066844e-05, "loss": 0.9491, "step": 168670 }, { "epoch": 14.61, "learning_rate": 3.8532149878570604e-05, "loss": 1.1362, "step": 168680 }, { "epoch": 14.61, "learning_rate": 3.852998149647276e-05, "loss": 0.9987, "step": 168690 }, { "epoch": 14.61, "learning_rate": 3.852781311437493e-05, "loss": 0.8867, "step": 168700 }, { "epoch": 14.61, "learning_rate": 3.852564473227708e-05, "loss": 0.9045, "step": 168710 }, { "epoch": 14.61, "learning_rate": 3.852347635017925e-05, "loss": 0.8896, "step": 168720 }, { "epoch": 14.61, "learning_rate": 3.852130796808141e-05, "loss": 0.8874, "step": 168730 }, { "epoch": 14.61, "learning_rate": 3.8519139585983575e-05, "loss": 1.2349, "step": 168740 }, { "epoch": 14.61, "learning_rate": 3.851697120388574e-05, "loss": 1.0108, "step": 168750 }, { "epoch": 14.62, "learning_rate": 3.85148028217879e-05, "loss": 0.9719, "step": 168760 }, { "epoch": 14.62, "learning_rate": 3.851263443969007e-05, "loss": 1.0046, "step": 168770 }, { "epoch": 14.62, "learning_rate": 3.851046605759223e-05, "loss": 0.8781, "step": 168780 }, { "epoch": 14.62, "learning_rate": 3.8508297675494386e-05, "loss": 0.847, "step": 168790 }, { "epoch": 14.62, "learning_rate": 3.8506129293396546e-05, "loss": 0.9632, "step": 168800 }, { "epoch": 14.62, "learning_rate": 3.850396091129871e-05, "loss": 0.9741, "step": 168810 }, { "epoch": 14.62, "learning_rate": 3.850179252920088e-05, "loss": 0.9349, "step": 168820 }, { "epoch": 14.62, "learning_rate": 3.849962414710304e-05, "loss": 1.0697, "step": 168830 }, { "epoch": 14.62, "learning_rate": 3.8497455765005205e-05, "loss": 0.8319, "step": 168840 }, { "epoch": 14.62, "learning_rate": 3.8495287382907364e-05, "loss": 0.9649, "step": 168850 }, { "epoch": 14.62, "learning_rate": 3.8493119000809524e-05, "loss": 0.9371, "step": 168860 }, { "epoch": 14.63, "learning_rate": 3.8490950618711684e-05, "loss": 0.9793, "step": 168870 }, { "epoch": 14.63, "learning_rate": 3.848878223661385e-05, "loss": 0.9913, "step": 168880 }, { "epoch": 14.63, "learning_rate": 3.848661385451601e-05, "loss": 0.7973, "step": 168890 }, { "epoch": 14.63, "learning_rate": 3.8484445472418176e-05, "loss": 0.88, "step": 168900 }, { "epoch": 14.63, "learning_rate": 3.848227709032034e-05, "loss": 0.9237, "step": 168910 }, { "epoch": 14.63, "learning_rate": 3.84801087082225e-05, "loss": 0.9912, "step": 168920 }, { "epoch": 14.63, "learning_rate": 3.847794032612467e-05, "loss": 1.0729, "step": 168930 }, { "epoch": 14.63, "learning_rate": 3.847577194402682e-05, "loss": 0.8283, "step": 168940 }, { "epoch": 14.63, "learning_rate": 3.847360356192899e-05, "loss": 1.0191, "step": 168950 }, { "epoch": 14.63, "learning_rate": 3.847143517983115e-05, "loss": 0.9069, "step": 168960 }, { "epoch": 14.63, "learning_rate": 3.8469266797733314e-05, "loss": 1.0798, "step": 168970 }, { "epoch": 14.63, "learning_rate": 3.846709841563548e-05, "loss": 1.1084, "step": 168980 }, { "epoch": 14.64, "learning_rate": 3.846493003353764e-05, "loss": 0.8854, "step": 168990 }, { "epoch": 14.64, "learning_rate": 3.8462761651439806e-05, "loss": 0.9884, "step": 169000 }, { "epoch": 14.64, "learning_rate": 3.8460593269341966e-05, "loss": 0.9195, "step": 169010 }, { "epoch": 14.64, "learning_rate": 3.8458424887244125e-05, "loss": 1.0602, "step": 169020 }, { "epoch": 14.64, "learning_rate": 3.8456256505146285e-05, "loss": 0.9207, "step": 169030 }, { "epoch": 14.64, "learning_rate": 3.845408812304845e-05, "loss": 1.0811, "step": 169040 }, { "epoch": 14.64, "learning_rate": 3.845191974095062e-05, "loss": 1.0398, "step": 169050 }, { "epoch": 14.64, "learning_rate": 3.844975135885278e-05, "loss": 0.9267, "step": 169060 }, { "epoch": 14.64, "learning_rate": 3.8447582976754944e-05, "loss": 1.0072, "step": 169070 }, { "epoch": 14.64, "learning_rate": 3.84454145946571e-05, "loss": 1.0018, "step": 169080 }, { "epoch": 14.64, "learning_rate": 3.844324621255926e-05, "loss": 0.9598, "step": 169090 }, { "epoch": 14.65, "learning_rate": 3.844107783046142e-05, "loss": 1.0395, "step": 169100 }, { "epoch": 14.65, "learning_rate": 3.843890944836359e-05, "loss": 0.9679, "step": 169110 }, { "epoch": 14.65, "learning_rate": 3.8436741066265755e-05, "loss": 1.0529, "step": 169120 }, { "epoch": 14.65, "learning_rate": 3.8434572684167915e-05, "loss": 0.9145, "step": 169130 }, { "epoch": 14.65, "learning_rate": 3.843240430207008e-05, "loss": 1.0899, "step": 169140 }, { "epoch": 14.65, "learning_rate": 3.843023591997224e-05, "loss": 1.0444, "step": 169150 }, { "epoch": 14.65, "learning_rate": 3.842806753787441e-05, "loss": 0.8474, "step": 169160 }, { "epoch": 14.65, "learning_rate": 3.842589915577656e-05, "loss": 1.0135, "step": 169170 }, { "epoch": 14.65, "learning_rate": 3.842373077367873e-05, "loss": 0.8542, "step": 169180 }, { "epoch": 14.65, "learning_rate": 3.842156239158089e-05, "loss": 0.9089, "step": 169190 }, { "epoch": 14.65, "learning_rate": 3.841939400948305e-05, "loss": 0.8782, "step": 169200 }, { "epoch": 14.65, "learning_rate": 3.841722562738522e-05, "loss": 0.9655, "step": 169210 }, { "epoch": 14.66, "learning_rate": 3.841505724528738e-05, "loss": 0.9657, "step": 169220 }, { "epoch": 14.66, "learning_rate": 3.8412888863189545e-05, "loss": 0.8671, "step": 169230 }, { "epoch": 14.66, "learning_rate": 3.841072048109171e-05, "loss": 1.1052, "step": 169240 }, { "epoch": 14.66, "learning_rate": 3.8408552098993864e-05, "loss": 0.9277, "step": 169250 }, { "epoch": 14.66, "learning_rate": 3.840638371689603e-05, "loss": 0.9194, "step": 169260 }, { "epoch": 14.66, "learning_rate": 3.840421533479819e-05, "loss": 0.9258, "step": 169270 }, { "epoch": 14.66, "learning_rate": 3.840204695270036e-05, "loss": 0.9445, "step": 169280 }, { "epoch": 14.66, "learning_rate": 3.8399878570602516e-05, "loss": 1.023, "step": 169290 }, { "epoch": 14.66, "learning_rate": 3.839771018850468e-05, "loss": 0.9014, "step": 169300 }, { "epoch": 14.66, "learning_rate": 3.839554180640685e-05, "loss": 0.9767, "step": 169310 }, { "epoch": 14.66, "learning_rate": 3.839337342430901e-05, "loss": 1.0836, "step": 169320 }, { "epoch": 14.67, "learning_rate": 3.839120504221117e-05, "loss": 0.8767, "step": 169330 }, { "epoch": 14.67, "learning_rate": 3.838903666011333e-05, "loss": 1.0982, "step": 169340 }, { "epoch": 14.67, "learning_rate": 3.8386868278015494e-05, "loss": 0.9012, "step": 169350 }, { "epoch": 14.67, "learning_rate": 3.8384699895917654e-05, "loss": 0.8844, "step": 169360 }, { "epoch": 14.67, "learning_rate": 3.838253151381982e-05, "loss": 0.9016, "step": 169370 }, { "epoch": 14.67, "learning_rate": 3.838036313172199e-05, "loss": 0.8919, "step": 169380 }, { "epoch": 14.67, "learning_rate": 3.8378194749624146e-05, "loss": 0.7239, "step": 169390 }, { "epoch": 14.67, "learning_rate": 3.83760263675263e-05, "loss": 1.0085, "step": 169400 }, { "epoch": 14.67, "learning_rate": 3.8373857985428466e-05, "loss": 0.95, "step": 169410 }, { "epoch": 14.67, "learning_rate": 3.837168960333063e-05, "loss": 0.9392, "step": 169420 }, { "epoch": 14.67, "learning_rate": 3.836952122123279e-05, "loss": 0.9184, "step": 169430 }, { "epoch": 14.67, "learning_rate": 3.836735283913496e-05, "loss": 1.1796, "step": 169440 }, { "epoch": 14.68, "learning_rate": 3.836518445703712e-05, "loss": 1.0283, "step": 169450 }, { "epoch": 14.68, "learning_rate": 3.8363016074939284e-05, "loss": 0.9925, "step": 169460 }, { "epoch": 14.68, "learning_rate": 3.836084769284145e-05, "loss": 1.0235, "step": 169470 }, { "epoch": 14.68, "learning_rate": 3.83586793107436e-05, "loss": 1.1271, "step": 169480 }, { "epoch": 14.68, "learning_rate": 3.835651092864577e-05, "loss": 0.8107, "step": 169490 }, { "epoch": 14.68, "learning_rate": 3.835434254654793e-05, "loss": 0.9123, "step": 169500 }, { "epoch": 14.68, "learning_rate": 3.8352174164450096e-05, "loss": 0.8817, "step": 169510 }, { "epoch": 14.68, "learning_rate": 3.8350005782352255e-05, "loss": 0.9656, "step": 169520 }, { "epoch": 14.68, "learning_rate": 3.834783740025442e-05, "loss": 0.9127, "step": 169530 }, { "epoch": 14.68, "learning_rate": 3.834566901815659e-05, "loss": 0.9715, "step": 169540 }, { "epoch": 14.68, "learning_rate": 3.834350063605875e-05, "loss": 0.9679, "step": 169550 }, { "epoch": 14.68, "learning_rate": 3.834133225396091e-05, "loss": 0.9808, "step": 169560 }, { "epoch": 14.69, "learning_rate": 3.833916387186307e-05, "loss": 1.1223, "step": 169570 }, { "epoch": 14.69, "learning_rate": 3.833699548976523e-05, "loss": 0.9123, "step": 169580 }, { "epoch": 14.69, "learning_rate": 3.833482710766739e-05, "loss": 0.8899, "step": 169590 }, { "epoch": 14.69, "learning_rate": 3.833265872556956e-05, "loss": 0.9803, "step": 169600 }, { "epoch": 14.69, "learning_rate": 3.8330490343471726e-05, "loss": 0.92, "step": 169610 }, { "epoch": 14.69, "learning_rate": 3.8328321961373885e-05, "loss": 0.9005, "step": 169620 }, { "epoch": 14.69, "learning_rate": 3.8326153579276045e-05, "loss": 0.89, "step": 169630 }, { "epoch": 14.69, "learning_rate": 3.8323985197178204e-05, "loss": 1.0045, "step": 169640 }, { "epoch": 14.69, "learning_rate": 3.832181681508037e-05, "loss": 0.9771, "step": 169650 }, { "epoch": 14.69, "learning_rate": 3.831964843298253e-05, "loss": 0.9444, "step": 169660 }, { "epoch": 14.69, "learning_rate": 3.83174800508847e-05, "loss": 0.8918, "step": 169670 }, { "epoch": 14.7, "learning_rate": 3.831531166878686e-05, "loss": 0.8712, "step": 169680 }, { "epoch": 14.7, "learning_rate": 3.831314328668902e-05, "loss": 1.0525, "step": 169690 }, { "epoch": 14.7, "learning_rate": 3.831097490459119e-05, "loss": 0.9034, "step": 169700 }, { "epoch": 14.7, "learning_rate": 3.830880652249334e-05, "loss": 1.0336, "step": 169710 }, { "epoch": 14.7, "learning_rate": 3.830663814039551e-05, "loss": 0.8966, "step": 169720 }, { "epoch": 14.7, "learning_rate": 3.830446975829767e-05, "loss": 0.955, "step": 169730 }, { "epoch": 14.7, "learning_rate": 3.8302301376199834e-05, "loss": 1.0019, "step": 169740 }, { "epoch": 14.7, "learning_rate": 3.8300132994102e-05, "loss": 0.8117, "step": 169750 }, { "epoch": 14.7, "learning_rate": 3.829796461200416e-05, "loss": 0.9731, "step": 169760 }, { "epoch": 14.7, "learning_rate": 3.829579622990633e-05, "loss": 1.1092, "step": 169770 }, { "epoch": 14.7, "learning_rate": 3.8293627847808487e-05, "loss": 0.9769, "step": 169780 }, { "epoch": 14.7, "learning_rate": 3.8291459465710646e-05, "loss": 0.9053, "step": 169790 }, { "epoch": 14.71, "learning_rate": 3.8289291083612806e-05, "loss": 0.8728, "step": 169800 }, { "epoch": 14.71, "learning_rate": 3.828712270151497e-05, "loss": 0.8439, "step": 169810 }, { "epoch": 14.71, "learning_rate": 3.828495431941714e-05, "loss": 1.0656, "step": 169820 }, { "epoch": 14.71, "learning_rate": 3.82827859373193e-05, "loss": 1.0501, "step": 169830 }, { "epoch": 14.71, "learning_rate": 3.8280617555221465e-05, "loss": 0.8582, "step": 169840 }, { "epoch": 14.71, "learning_rate": 3.8278449173123624e-05, "loss": 1.0894, "step": 169850 }, { "epoch": 14.71, "learning_rate": 3.8276280791025784e-05, "loss": 0.9866, "step": 169860 }, { "epoch": 14.71, "learning_rate": 3.827411240892794e-05, "loss": 0.9769, "step": 169870 }, { "epoch": 14.71, "learning_rate": 3.827194402683011e-05, "loss": 0.9127, "step": 169880 }, { "epoch": 14.71, "learning_rate": 3.8269775644732276e-05, "loss": 0.9538, "step": 169890 }, { "epoch": 14.71, "learning_rate": 3.8267607262634436e-05, "loss": 1.0553, "step": 169900 }, { "epoch": 14.72, "learning_rate": 3.82654388805366e-05, "loss": 0.8305, "step": 169910 }, { "epoch": 14.72, "learning_rate": 3.826327049843876e-05, "loss": 1.1524, "step": 169920 }, { "epoch": 14.72, "learning_rate": 3.826110211634093e-05, "loss": 0.872, "step": 169930 }, { "epoch": 14.72, "learning_rate": 3.825893373424308e-05, "loss": 0.8837, "step": 169940 }, { "epoch": 14.72, "learning_rate": 3.825676535214525e-05, "loss": 1.0266, "step": 169950 }, { "epoch": 14.72, "learning_rate": 3.825459697004741e-05, "loss": 1.0364, "step": 169960 }, { "epoch": 14.72, "learning_rate": 3.8252428587949573e-05, "loss": 1.0204, "step": 169970 }, { "epoch": 14.72, "learning_rate": 3.825026020585174e-05, "loss": 0.8414, "step": 169980 }, { "epoch": 14.72, "learning_rate": 3.82480918237539e-05, "loss": 1.0325, "step": 169990 }, { "epoch": 14.72, "learning_rate": 3.8245923441656066e-05, "loss": 0.9962, "step": 170000 }, { "epoch": 14.72, "learning_rate": 3.8243755059558225e-05, "loss": 0.9966, "step": 170010 }, { "epoch": 14.72, "learning_rate": 3.8241586677460385e-05, "loss": 0.8575, "step": 170020 }, { "epoch": 14.73, "learning_rate": 3.8239418295362545e-05, "loss": 0.9374, "step": 170030 }, { "epoch": 14.73, "learning_rate": 3.823724991326471e-05, "loss": 0.865, "step": 170040 }, { "epoch": 14.73, "learning_rate": 3.823508153116688e-05, "loss": 1.1549, "step": 170050 }, { "epoch": 14.73, "learning_rate": 3.823291314906904e-05, "loss": 0.9938, "step": 170060 }, { "epoch": 14.73, "learning_rate": 3.8230744766971203e-05, "loss": 0.972, "step": 170070 }, { "epoch": 14.73, "learning_rate": 3.822857638487336e-05, "loss": 0.9227, "step": 170080 }, { "epoch": 14.73, "learning_rate": 3.822640800277553e-05, "loss": 1.0394, "step": 170090 }, { "epoch": 14.73, "learning_rate": 3.822423962067768e-05, "loss": 1.0449, "step": 170100 }, { "epoch": 14.73, "learning_rate": 3.822207123857985e-05, "loss": 0.9155, "step": 170110 }, { "epoch": 14.73, "learning_rate": 3.8219902856482015e-05, "loss": 1.0116, "step": 170120 }, { "epoch": 14.73, "learning_rate": 3.8217734474384175e-05, "loss": 0.9643, "step": 170130 }, { "epoch": 14.74, "learning_rate": 3.821556609228634e-05, "loss": 0.9712, "step": 170140 }, { "epoch": 14.74, "learning_rate": 3.82133977101885e-05, "loss": 1.0048, "step": 170150 }, { "epoch": 14.74, "learning_rate": 3.821122932809067e-05, "loss": 0.9515, "step": 170160 }, { "epoch": 14.74, "learning_rate": 3.820906094599282e-05, "loss": 0.9551, "step": 170170 }, { "epoch": 14.74, "learning_rate": 3.8206892563894986e-05, "loss": 0.9969, "step": 170180 }, { "epoch": 14.74, "learning_rate": 3.820472418179715e-05, "loss": 0.9294, "step": 170190 }, { "epoch": 14.74, "learning_rate": 3.820255579969931e-05, "loss": 1.0751, "step": 170200 }, { "epoch": 14.74, "learning_rate": 3.820038741760148e-05, "loss": 0.9851, "step": 170210 }, { "epoch": 14.74, "learning_rate": 3.819821903550364e-05, "loss": 1.0095, "step": 170220 }, { "epoch": 14.74, "learning_rate": 3.8196050653405805e-05, "loss": 0.9219, "step": 170230 }, { "epoch": 14.74, "learning_rate": 3.819388227130797e-05, "loss": 1.0076, "step": 170240 }, { "epoch": 14.74, "learning_rate": 3.8191713889210124e-05, "loss": 1.0769, "step": 170250 }, { "epoch": 14.75, "learning_rate": 3.818954550711229e-05, "loss": 1.0341, "step": 170260 }, { "epoch": 14.75, "learning_rate": 3.818737712501445e-05, "loss": 0.9159, "step": 170270 }, { "epoch": 14.75, "learning_rate": 3.8185208742916616e-05, "loss": 0.9078, "step": 170280 }, { "epoch": 14.75, "learning_rate": 3.8183040360818776e-05, "loss": 1.0896, "step": 170290 }, { "epoch": 14.75, "learning_rate": 3.818087197872094e-05, "loss": 0.7879, "step": 170300 }, { "epoch": 14.75, "learning_rate": 3.817870359662311e-05, "loss": 0.8831, "step": 170310 }, { "epoch": 14.75, "learning_rate": 3.817653521452527e-05, "loss": 1.0594, "step": 170320 }, { "epoch": 14.75, "learning_rate": 3.817436683242743e-05, "loss": 0.9829, "step": 170330 }, { "epoch": 14.75, "learning_rate": 3.817219845032959e-05, "loss": 1.0751, "step": 170340 }, { "epoch": 14.75, "learning_rate": 3.8170030068231754e-05, "loss": 1.0098, "step": 170350 }, { "epoch": 14.75, "learning_rate": 3.8167861686133914e-05, "loss": 0.9928, "step": 170360 }, { "epoch": 14.76, "learning_rate": 3.816569330403608e-05, "loss": 0.9627, "step": 170370 }, { "epoch": 14.76, "learning_rate": 3.8163524921938246e-05, "loss": 0.9286, "step": 170380 }, { "epoch": 14.76, "learning_rate": 3.8161356539840406e-05, "loss": 0.9902, "step": 170390 }, { "epoch": 14.76, "learning_rate": 3.8159188157742566e-05, "loss": 0.9932, "step": 170400 }, { "epoch": 14.76, "learning_rate": 3.8157019775644725e-05, "loss": 0.9731, "step": 170410 }, { "epoch": 14.76, "learning_rate": 3.815485139354689e-05, "loss": 1.107, "step": 170420 }, { "epoch": 14.76, "learning_rate": 3.815268301144905e-05, "loss": 0.9907, "step": 170430 }, { "epoch": 14.76, "learning_rate": 3.815051462935122e-05, "loss": 0.9976, "step": 170440 }, { "epoch": 14.76, "learning_rate": 3.8148346247253384e-05, "loss": 0.9625, "step": 170450 }, { "epoch": 14.76, "learning_rate": 3.8146177865155544e-05, "loss": 1.0182, "step": 170460 }, { "epoch": 14.76, "learning_rate": 3.814400948305771e-05, "loss": 0.8802, "step": 170470 }, { "epoch": 14.76, "learning_rate": 3.814184110095986e-05, "loss": 0.9731, "step": 170480 }, { "epoch": 14.77, "learning_rate": 3.813967271886203e-05, "loss": 0.9459, "step": 170490 }, { "epoch": 14.77, "learning_rate": 3.813750433676419e-05, "loss": 0.8219, "step": 170500 }, { "epoch": 14.77, "learning_rate": 3.8135335954666355e-05, "loss": 1.0614, "step": 170510 }, { "epoch": 14.77, "learning_rate": 3.8133167572568515e-05, "loss": 0.8711, "step": 170520 }, { "epoch": 14.77, "learning_rate": 3.813099919047068e-05, "loss": 1.0991, "step": 170530 }, { "epoch": 14.77, "learning_rate": 3.812883080837285e-05, "loss": 0.9019, "step": 170540 }, { "epoch": 14.77, "learning_rate": 3.812666242627501e-05, "loss": 1.1983, "step": 170550 }, { "epoch": 14.77, "learning_rate": 3.812449404417717e-05, "loss": 0.9764, "step": 170560 }, { "epoch": 14.77, "learning_rate": 3.8122325662079327e-05, "loss": 1.0261, "step": 170570 }, { "epoch": 14.77, "learning_rate": 3.812015727998149e-05, "loss": 0.9338, "step": 170580 }, { "epoch": 14.77, "learning_rate": 3.811798889788365e-05, "loss": 0.9112, "step": 170590 }, { "epoch": 14.78, "learning_rate": 3.811582051578582e-05, "loss": 1.0812, "step": 170600 }, { "epoch": 14.78, "learning_rate": 3.8113652133687985e-05, "loss": 1.031, "step": 170610 }, { "epoch": 14.78, "learning_rate": 3.8111483751590145e-05, "loss": 0.9713, "step": 170620 }, { "epoch": 14.78, "learning_rate": 3.8109315369492305e-05, "loss": 1.0398, "step": 170630 }, { "epoch": 14.78, "learning_rate": 3.8107146987394464e-05, "loss": 0.9102, "step": 170640 }, { "epoch": 14.78, "learning_rate": 3.810497860529663e-05, "loss": 0.7873, "step": 170650 }, { "epoch": 14.78, "learning_rate": 3.810281022319879e-05, "loss": 0.9497, "step": 170660 }, { "epoch": 14.78, "learning_rate": 3.8100641841100957e-05, "loss": 1.0971, "step": 170670 }, { "epoch": 14.78, "learning_rate": 3.809847345900312e-05, "loss": 0.949, "step": 170680 }, { "epoch": 14.78, "learning_rate": 3.809630507690528e-05, "loss": 1.0317, "step": 170690 }, { "epoch": 14.78, "learning_rate": 3.809413669480745e-05, "loss": 0.9291, "step": 170700 }, { "epoch": 14.78, "learning_rate": 3.80919683127096e-05, "loss": 0.8449, "step": 170710 }, { "epoch": 14.79, "learning_rate": 3.808979993061177e-05, "loss": 0.8781, "step": 170720 }, { "epoch": 14.79, "learning_rate": 3.808763154851393e-05, "loss": 1.0893, "step": 170730 }, { "epoch": 14.79, "learning_rate": 3.8085463166416094e-05, "loss": 1.0031, "step": 170740 }, { "epoch": 14.79, "learning_rate": 3.808329478431826e-05, "loss": 0.8355, "step": 170750 }, { "epoch": 14.79, "learning_rate": 3.808112640222042e-05, "loss": 0.9151, "step": 170760 }, { "epoch": 14.79, "learning_rate": 3.8078958020122587e-05, "loss": 0.8469, "step": 170770 }, { "epoch": 14.79, "learning_rate": 3.8076789638024746e-05, "loss": 1.0202, "step": 170780 }, { "epoch": 14.79, "learning_rate": 3.8074621255926906e-05, "loss": 0.9968, "step": 170790 }, { "epoch": 14.79, "learning_rate": 3.8072452873829065e-05, "loss": 0.9906, "step": 170800 }, { "epoch": 14.79, "learning_rate": 3.807028449173123e-05, "loss": 1.0195, "step": 170810 }, { "epoch": 14.79, "learning_rate": 3.80681161096334e-05, "loss": 0.9124, "step": 170820 }, { "epoch": 14.79, "learning_rate": 3.806594772753556e-05, "loss": 1.0339, "step": 170830 }, { "epoch": 14.8, "learning_rate": 3.8063779345437724e-05, "loss": 1.0992, "step": 170840 }, { "epoch": 14.8, "learning_rate": 3.8061610963339884e-05, "loss": 1.1096, "step": 170850 }, { "epoch": 14.8, "learning_rate": 3.805944258124205e-05, "loss": 1.0616, "step": 170860 }, { "epoch": 14.8, "learning_rate": 3.80572741991442e-05, "loss": 0.964, "step": 170870 }, { "epoch": 14.8, "learning_rate": 3.805510581704637e-05, "loss": 0.775, "step": 170880 }, { "epoch": 14.8, "learning_rate": 3.8052937434948536e-05, "loss": 1.0432, "step": 170890 }, { "epoch": 14.8, "learning_rate": 3.8050769052850695e-05, "loss": 0.9549, "step": 170900 }, { "epoch": 14.8, "learning_rate": 3.804860067075286e-05, "loss": 0.9947, "step": 170910 }, { "epoch": 14.8, "learning_rate": 3.804643228865502e-05, "loss": 0.9796, "step": 170920 }, { "epoch": 14.8, "learning_rate": 3.804426390655719e-05, "loss": 0.9566, "step": 170930 }, { "epoch": 14.8, "learning_rate": 3.804209552445934e-05, "loss": 0.9916, "step": 170940 }, { "epoch": 14.81, "learning_rate": 3.803992714236151e-05, "loss": 1.0489, "step": 170950 }, { "epoch": 14.81, "learning_rate": 3.8037758760263673e-05, "loss": 0.9899, "step": 170960 }, { "epoch": 14.81, "learning_rate": 3.803559037816583e-05, "loss": 0.9792, "step": 170970 }, { "epoch": 14.81, "learning_rate": 3.8033421996068e-05, "loss": 0.9615, "step": 170980 }, { "epoch": 14.81, "learning_rate": 3.803125361397016e-05, "loss": 0.9855, "step": 170990 }, { "epoch": 14.81, "learning_rate": 3.8029085231872326e-05, "loss": 1.0931, "step": 171000 }, { "epoch": 14.81, "learning_rate": 3.802691684977449e-05, "loss": 0.8788, "step": 171010 }, { "epoch": 14.81, "learning_rate": 3.8024748467676645e-05, "loss": 0.8692, "step": 171020 }, { "epoch": 14.81, "learning_rate": 3.8022580085578804e-05, "loss": 0.9491, "step": 171030 }, { "epoch": 14.81, "learning_rate": 3.802041170348097e-05, "loss": 0.867, "step": 171040 }, { "epoch": 14.81, "learning_rate": 3.801824332138314e-05, "loss": 0.743, "step": 171050 }, { "epoch": 14.81, "learning_rate": 3.80160749392853e-05, "loss": 0.9823, "step": 171060 }, { "epoch": 14.82, "learning_rate": 3.801390655718746e-05, "loss": 1.119, "step": 171070 }, { "epoch": 14.82, "learning_rate": 3.801173817508962e-05, "loss": 1.1547, "step": 171080 }, { "epoch": 14.82, "learning_rate": 3.800956979299179e-05, "loss": 1.0131, "step": 171090 }, { "epoch": 14.82, "learning_rate": 3.800740141089394e-05, "loss": 0.8216, "step": 171100 }, { "epoch": 14.82, "learning_rate": 3.800523302879611e-05, "loss": 0.9425, "step": 171110 }, { "epoch": 14.82, "learning_rate": 3.8003064646698275e-05, "loss": 1.0398, "step": 171120 }, { "epoch": 14.82, "learning_rate": 3.8000896264600434e-05, "loss": 0.8816, "step": 171130 }, { "epoch": 14.82, "learning_rate": 3.79987278825026e-05, "loss": 0.903, "step": 171140 }, { "epoch": 14.82, "learning_rate": 3.799655950040476e-05, "loss": 0.9341, "step": 171150 }, { "epoch": 14.82, "learning_rate": 3.799439111830693e-05, "loss": 1.2189, "step": 171160 }, { "epoch": 14.82, "learning_rate": 3.799222273620908e-05, "loss": 1.0707, "step": 171170 }, { "epoch": 14.83, "learning_rate": 3.7990054354111246e-05, "loss": 0.9724, "step": 171180 }, { "epoch": 14.83, "learning_rate": 3.798788597201341e-05, "loss": 0.7239, "step": 171190 }, { "epoch": 14.83, "learning_rate": 3.798571758991557e-05, "loss": 1.0053, "step": 171200 }, { "epoch": 14.83, "learning_rate": 3.798354920781774e-05, "loss": 0.9142, "step": 171210 }, { "epoch": 14.83, "learning_rate": 3.79813808257199e-05, "loss": 0.9285, "step": 171220 }, { "epoch": 14.83, "learning_rate": 3.7979212443622064e-05, "loss": 1.0442, "step": 171230 }, { "epoch": 14.83, "learning_rate": 3.797704406152423e-05, "loss": 0.9451, "step": 171240 }, { "epoch": 14.83, "learning_rate": 3.7974875679426384e-05, "loss": 0.9446, "step": 171250 }, { "epoch": 14.83, "learning_rate": 3.797270729732855e-05, "loss": 0.9648, "step": 171260 }, { "epoch": 14.83, "learning_rate": 3.797053891523071e-05, "loss": 1.0027, "step": 171270 }, { "epoch": 14.83, "learning_rate": 3.7968370533132876e-05, "loss": 1.0307, "step": 171280 }, { "epoch": 14.83, "learning_rate": 3.7966202151035036e-05, "loss": 0.9452, "step": 171290 }, { "epoch": 14.84, "learning_rate": 3.79640337689372e-05, "loss": 0.9556, "step": 171300 }, { "epoch": 14.84, "learning_rate": 3.796186538683937e-05, "loss": 1.0257, "step": 171310 }, { "epoch": 14.84, "learning_rate": 3.795969700474153e-05, "loss": 0.9457, "step": 171320 }, { "epoch": 14.84, "learning_rate": 3.795752862264369e-05, "loss": 0.8802, "step": 171330 }, { "epoch": 14.84, "learning_rate": 3.795536024054585e-05, "loss": 1.1064, "step": 171340 }, { "epoch": 14.84, "learning_rate": 3.7953191858448014e-05, "loss": 0.9139, "step": 171350 }, { "epoch": 14.84, "learning_rate": 3.795102347635017e-05, "loss": 0.8396, "step": 171360 }, { "epoch": 14.84, "learning_rate": 3.794885509425234e-05, "loss": 0.9636, "step": 171370 }, { "epoch": 14.84, "learning_rate": 3.7946686712154506e-05, "loss": 1.1157, "step": 171380 }, { "epoch": 14.84, "learning_rate": 3.7944518330056666e-05, "loss": 1.0578, "step": 171390 }, { "epoch": 14.84, "learning_rate": 3.7942349947958825e-05, "loss": 1.0356, "step": 171400 }, { "epoch": 14.85, "learning_rate": 3.7940181565860985e-05, "loss": 0.9062, "step": 171410 }, { "epoch": 14.85, "learning_rate": 3.793801318376315e-05, "loss": 0.9649, "step": 171420 }, { "epoch": 14.85, "learning_rate": 3.793584480166531e-05, "loss": 0.8641, "step": 171430 }, { "epoch": 14.85, "learning_rate": 3.793367641956748e-05, "loss": 1.0131, "step": 171440 }, { "epoch": 14.85, "learning_rate": 3.7931508037469644e-05, "loss": 0.8299, "step": 171450 }, { "epoch": 14.85, "learning_rate": 3.79293396553718e-05, "loss": 1.0438, "step": 171460 }, { "epoch": 14.85, "learning_rate": 3.792717127327397e-05, "loss": 1.1362, "step": 171470 }, { "epoch": 14.85, "learning_rate": 3.792500289117612e-05, "loss": 0.8748, "step": 171480 }, { "epoch": 14.85, "learning_rate": 3.792283450907829e-05, "loss": 0.8984, "step": 171490 }, { "epoch": 14.85, "learning_rate": 3.792066612698045e-05, "loss": 0.9109, "step": 171500 }, { "epoch": 14.85, "learning_rate": 3.7918497744882615e-05, "loss": 0.7529, "step": 171510 }, { "epoch": 14.85, "learning_rate": 3.791632936278478e-05, "loss": 0.9201, "step": 171520 }, { "epoch": 14.86, "learning_rate": 3.791416098068694e-05, "loss": 0.7535, "step": 171530 }, { "epoch": 14.86, "learning_rate": 3.791199259858911e-05, "loss": 1.1074, "step": 171540 }, { "epoch": 14.86, "learning_rate": 3.790982421649127e-05, "loss": 1.1233, "step": 171550 }, { "epoch": 14.86, "learning_rate": 3.7907655834393427e-05, "loss": 0.8614, "step": 171560 }, { "epoch": 14.86, "learning_rate": 3.7905487452295586e-05, "loss": 0.8724, "step": 171570 }, { "epoch": 14.86, "learning_rate": 3.790331907019775e-05, "loss": 0.8394, "step": 171580 }, { "epoch": 14.86, "learning_rate": 3.790115068809991e-05, "loss": 0.8851, "step": 171590 }, { "epoch": 14.86, "learning_rate": 3.789898230600208e-05, "loss": 0.9185, "step": 171600 }, { "epoch": 14.86, "learning_rate": 3.7896813923904245e-05, "loss": 1.0129, "step": 171610 }, { "epoch": 14.86, "learning_rate": 3.7894645541806405e-05, "loss": 1.1214, "step": 171620 }, { "epoch": 14.86, "learning_rate": 3.789247715970857e-05, "loss": 0.7625, "step": 171630 }, { "epoch": 14.87, "learning_rate": 3.7890308777610724e-05, "loss": 0.8921, "step": 171640 }, { "epoch": 14.87, "learning_rate": 3.788814039551289e-05, "loss": 0.9135, "step": 171650 }, { "epoch": 14.87, "learning_rate": 3.788597201341505e-05, "loss": 1.0335, "step": 171660 }, { "epoch": 14.87, "learning_rate": 3.7883803631317216e-05, "loss": 0.9959, "step": 171670 }, { "epoch": 14.87, "learning_rate": 3.788163524921938e-05, "loss": 0.9665, "step": 171680 }, { "epoch": 14.87, "learning_rate": 3.787946686712154e-05, "loss": 0.9946, "step": 171690 }, { "epoch": 14.87, "learning_rate": 3.787729848502371e-05, "loss": 0.9194, "step": 171700 }, { "epoch": 14.87, "learning_rate": 3.787513010292586e-05, "loss": 1.1496, "step": 171710 }, { "epoch": 14.87, "learning_rate": 3.787296172082803e-05, "loss": 1.0893, "step": 171720 }, { "epoch": 14.87, "learning_rate": 3.787079333873019e-05, "loss": 0.9667, "step": 171730 }, { "epoch": 14.87, "learning_rate": 3.7868624956632354e-05, "loss": 1.2572, "step": 171740 }, { "epoch": 14.87, "learning_rate": 3.786645657453452e-05, "loss": 1.18, "step": 171750 }, { "epoch": 14.88, "learning_rate": 3.786428819243668e-05, "loss": 0.9339, "step": 171760 }, { "epoch": 14.88, "learning_rate": 3.7862119810338846e-05, "loss": 1.1049, "step": 171770 }, { "epoch": 14.88, "learning_rate": 3.7859951428241006e-05, "loss": 1.1211, "step": 171780 }, { "epoch": 14.88, "learning_rate": 3.7857783046143166e-05, "loss": 0.8691, "step": 171790 }, { "epoch": 14.88, "learning_rate": 3.7855614664045325e-05, "loss": 0.9891, "step": 171800 }, { "epoch": 14.88, "learning_rate": 3.785344628194749e-05, "loss": 0.9033, "step": 171810 }, { "epoch": 14.88, "learning_rate": 3.785127789984966e-05, "loss": 0.8066, "step": 171820 }, { "epoch": 14.88, "learning_rate": 3.784910951775182e-05, "loss": 0.9066, "step": 171830 }, { "epoch": 14.88, "learning_rate": 3.7846941135653984e-05, "loss": 0.9495, "step": 171840 }, { "epoch": 14.88, "learning_rate": 3.7844772753556144e-05, "loss": 1.0067, "step": 171850 }, { "epoch": 14.88, "learning_rate": 3.784260437145831e-05, "loss": 0.8216, "step": 171860 }, { "epoch": 14.89, "learning_rate": 3.784043598936046e-05, "loss": 1.0353, "step": 171870 }, { "epoch": 14.89, "learning_rate": 3.783826760726263e-05, "loss": 0.8688, "step": 171880 }, { "epoch": 14.89, "learning_rate": 3.7836099225164796e-05, "loss": 0.9968, "step": 171890 }, { "epoch": 14.89, "learning_rate": 3.7833930843066955e-05, "loss": 0.9925, "step": 171900 }, { "epoch": 14.89, "learning_rate": 3.783176246096912e-05, "loss": 1.0279, "step": 171910 }, { "epoch": 14.89, "learning_rate": 3.782959407887128e-05, "loss": 0.997, "step": 171920 }, { "epoch": 14.89, "learning_rate": 3.782742569677345e-05, "loss": 1.1206, "step": 171930 }, { "epoch": 14.89, "learning_rate": 3.78252573146756e-05, "loss": 0.9898, "step": 171940 }, { "epoch": 14.89, "learning_rate": 3.782308893257777e-05, "loss": 0.9664, "step": 171950 }, { "epoch": 14.89, "learning_rate": 3.782092055047993e-05, "loss": 0.9487, "step": 171960 }, { "epoch": 14.89, "learning_rate": 3.781875216838209e-05, "loss": 0.9589, "step": 171970 }, { "epoch": 14.89, "learning_rate": 3.781658378628426e-05, "loss": 0.8848, "step": 171980 }, { "epoch": 14.9, "learning_rate": 3.781441540418642e-05, "loss": 1.1542, "step": 171990 }, { "epoch": 14.9, "learning_rate": 3.7812247022088585e-05, "loss": 0.9542, "step": 172000 }, { "epoch": 14.9, "learning_rate": 3.781007863999075e-05, "loss": 1.0381, "step": 172010 }, { "epoch": 14.9, "learning_rate": 3.7807910257892904e-05, "loss": 0.985, "step": 172020 }, { "epoch": 14.9, "learning_rate": 3.780574187579507e-05, "loss": 0.9525, "step": 172030 }, { "epoch": 14.9, "learning_rate": 3.780357349369723e-05, "loss": 1.1777, "step": 172040 }, { "epoch": 14.9, "learning_rate": 3.78014051115994e-05, "loss": 1.1162, "step": 172050 }, { "epoch": 14.9, "learning_rate": 3.7799236729501556e-05, "loss": 1.0736, "step": 172060 }, { "epoch": 14.9, "learning_rate": 3.779706834740372e-05, "loss": 1.0644, "step": 172070 }, { "epoch": 14.9, "learning_rate": 3.779489996530589e-05, "loss": 1.0663, "step": 172080 }, { "epoch": 14.9, "learning_rate": 3.779273158320805e-05, "loss": 1.0113, "step": 172090 }, { "epoch": 14.9, "learning_rate": 3.77905632011102e-05, "loss": 1.0601, "step": 172100 }, { "epoch": 14.91, "learning_rate": 3.778839481901237e-05, "loss": 1.1536, "step": 172110 }, { "epoch": 14.91, "learning_rate": 3.7786226436914534e-05, "loss": 0.8917, "step": 172120 }, { "epoch": 14.91, "learning_rate": 3.7784058054816694e-05, "loss": 0.9555, "step": 172130 }, { "epoch": 14.91, "learning_rate": 3.778188967271886e-05, "loss": 1.0366, "step": 172140 }, { "epoch": 14.91, "learning_rate": 3.777972129062102e-05, "loss": 1.0776, "step": 172150 }, { "epoch": 14.91, "learning_rate": 3.7777552908523186e-05, "loss": 1.1005, "step": 172160 }, { "epoch": 14.91, "learning_rate": 3.777538452642534e-05, "loss": 1.0568, "step": 172170 }, { "epoch": 14.91, "learning_rate": 3.7773216144327506e-05, "loss": 0.8834, "step": 172180 }, { "epoch": 14.91, "learning_rate": 3.777104776222967e-05, "loss": 0.7714, "step": 172190 }, { "epoch": 14.91, "learning_rate": 3.776887938013183e-05, "loss": 1.0689, "step": 172200 }, { "epoch": 14.91, "learning_rate": 3.7766710998034e-05, "loss": 1.0741, "step": 172210 }, { "epoch": 14.92, "learning_rate": 3.776454261593616e-05, "loss": 0.928, "step": 172220 }, { "epoch": 14.92, "learning_rate": 3.7762374233838324e-05, "loss": 1.0175, "step": 172230 }, { "epoch": 14.92, "learning_rate": 3.776020585174049e-05, "loss": 0.9496, "step": 172240 }, { "epoch": 14.92, "learning_rate": 3.775803746964264e-05, "loss": 1.0017, "step": 172250 }, { "epoch": 14.92, "learning_rate": 3.775586908754481e-05, "loss": 0.9001, "step": 172260 }, { "epoch": 14.92, "learning_rate": 3.775370070544697e-05, "loss": 0.9215, "step": 172270 }, { "epoch": 14.92, "learning_rate": 3.7751532323349136e-05, "loss": 0.9713, "step": 172280 }, { "epoch": 14.92, "learning_rate": 3.7749363941251295e-05, "loss": 0.9325, "step": 172290 }, { "epoch": 14.92, "learning_rate": 3.774719555915346e-05, "loss": 0.91, "step": 172300 }, { "epoch": 14.92, "learning_rate": 3.774502717705563e-05, "loss": 1.1149, "step": 172310 }, { "epoch": 14.92, "learning_rate": 3.774285879495779e-05, "loss": 0.9861, "step": 172320 }, { "epoch": 14.92, "learning_rate": 3.774069041285995e-05, "loss": 0.959, "step": 172330 }, { "epoch": 14.93, "learning_rate": 3.773852203076211e-05, "loss": 0.8753, "step": 172340 }, { "epoch": 14.93, "learning_rate": 3.773635364866427e-05, "loss": 1.1663, "step": 172350 }, { "epoch": 14.93, "learning_rate": 3.773418526656643e-05, "loss": 0.8859, "step": 172360 }, { "epoch": 14.93, "learning_rate": 3.77320168844686e-05, "loss": 1.0856, "step": 172370 }, { "epoch": 14.93, "learning_rate": 3.7729848502370766e-05, "loss": 0.9685, "step": 172380 }, { "epoch": 14.93, "learning_rate": 3.7727680120272925e-05, "loss": 0.7965, "step": 172390 }, { "epoch": 14.93, "learning_rate": 3.772551173817509e-05, "loss": 1.1821, "step": 172400 }, { "epoch": 14.93, "learning_rate": 3.7723343356077245e-05, "loss": 1.0321, "step": 172410 }, { "epoch": 14.93, "learning_rate": 3.772117497397941e-05, "loss": 0.9629, "step": 172420 }, { "epoch": 14.93, "learning_rate": 3.771900659188157e-05, "loss": 1.1708, "step": 172430 }, { "epoch": 14.93, "learning_rate": 3.771683820978374e-05, "loss": 0.9508, "step": 172440 }, { "epoch": 14.94, "learning_rate": 3.7714669827685903e-05, "loss": 1.0889, "step": 172450 }, { "epoch": 14.94, "learning_rate": 3.771250144558806e-05, "loss": 1.0995, "step": 172460 }, { "epoch": 14.94, "learning_rate": 3.771033306349023e-05, "loss": 0.9699, "step": 172470 }, { "epoch": 14.94, "learning_rate": 3.770816468139238e-05, "loss": 0.8187, "step": 172480 }, { "epoch": 14.94, "learning_rate": 3.770599629929455e-05, "loss": 1.0751, "step": 172490 }, { "epoch": 14.94, "learning_rate": 3.770382791719671e-05, "loss": 0.9253, "step": 172500 }, { "epoch": 14.94, "learning_rate": 3.7701659535098875e-05, "loss": 0.9587, "step": 172510 }, { "epoch": 14.94, "learning_rate": 3.769949115300104e-05, "loss": 0.88, "step": 172520 }, { "epoch": 14.94, "learning_rate": 3.76973227709032e-05, "loss": 0.9492, "step": 172530 }, { "epoch": 14.94, "learning_rate": 3.769515438880537e-05, "loss": 1.0115, "step": 172540 }, { "epoch": 14.94, "learning_rate": 3.769298600670753e-05, "loss": 0.9997, "step": 172550 }, { "epoch": 14.94, "learning_rate": 3.7690817624609686e-05, "loss": 0.9476, "step": 172560 }, { "epoch": 14.95, "learning_rate": 3.7688649242511846e-05, "loss": 0.9424, "step": 172570 }, { "epoch": 14.95, "learning_rate": 3.768648086041401e-05, "loss": 0.891, "step": 172580 }, { "epoch": 14.95, "learning_rate": 3.768431247831618e-05, "loss": 1.1563, "step": 172590 }, { "epoch": 14.95, "learning_rate": 3.768214409621834e-05, "loss": 0.8724, "step": 172600 }, { "epoch": 14.95, "learning_rate": 3.7679975714120505e-05, "loss": 0.8967, "step": 172610 }, { "epoch": 14.95, "learning_rate": 3.7677807332022664e-05, "loss": 0.7832, "step": 172620 }, { "epoch": 14.95, "learning_rate": 3.767563894992483e-05, "loss": 1.1053, "step": 172630 }, { "epoch": 14.95, "learning_rate": 3.7673470567826984e-05, "loss": 1.0206, "step": 172640 }, { "epoch": 14.95, "learning_rate": 3.767130218572915e-05, "loss": 0.9765, "step": 172650 }, { "epoch": 14.95, "learning_rate": 3.766913380363131e-05, "loss": 1.0481, "step": 172660 }, { "epoch": 14.95, "learning_rate": 3.7666965421533476e-05, "loss": 0.9803, "step": 172670 }, { "epoch": 14.96, "learning_rate": 3.766479703943564e-05, "loss": 0.8891, "step": 172680 }, { "epoch": 14.96, "learning_rate": 3.76626286573378e-05, "loss": 0.9867, "step": 172690 }, { "epoch": 14.96, "learning_rate": 3.766046027523997e-05, "loss": 0.7831, "step": 172700 }, { "epoch": 14.96, "learning_rate": 3.765829189314212e-05, "loss": 1.0351, "step": 172710 }, { "epoch": 14.96, "learning_rate": 3.765612351104429e-05, "loss": 0.8836, "step": 172720 }, { "epoch": 14.96, "learning_rate": 3.765395512894645e-05, "loss": 1.0676, "step": 172730 }, { "epoch": 14.96, "learning_rate": 3.7651786746848614e-05, "loss": 1.0147, "step": 172740 }, { "epoch": 14.96, "learning_rate": 3.764961836475078e-05, "loss": 0.9362, "step": 172750 }, { "epoch": 14.96, "learning_rate": 3.764744998265294e-05, "loss": 0.8568, "step": 172760 }, { "epoch": 14.96, "learning_rate": 3.7645281600555106e-05, "loss": 1.0202, "step": 172770 }, { "epoch": 14.96, "learning_rate": 3.7643113218457266e-05, "loss": 0.9666, "step": 172780 }, { "epoch": 14.96, "learning_rate": 3.7640944836359425e-05, "loss": 0.8536, "step": 172790 }, { "epoch": 14.97, "learning_rate": 3.7638776454261585e-05, "loss": 0.9131, "step": 172800 }, { "epoch": 14.97, "learning_rate": 3.763660807216375e-05, "loss": 0.8762, "step": 172810 }, { "epoch": 14.97, "learning_rate": 3.763443969006592e-05, "loss": 0.9221, "step": 172820 }, { "epoch": 14.97, "learning_rate": 3.763227130796808e-05, "loss": 0.9449, "step": 172830 }, { "epoch": 14.97, "learning_rate": 3.7630102925870244e-05, "loss": 1.0048, "step": 172840 }, { "epoch": 14.97, "learning_rate": 3.76279345437724e-05, "loss": 0.88, "step": 172850 }, { "epoch": 14.97, "learning_rate": 3.762576616167457e-05, "loss": 0.9281, "step": 172860 }, { "epoch": 14.97, "learning_rate": 3.762359777957672e-05, "loss": 0.9965, "step": 172870 }, { "epoch": 14.97, "learning_rate": 3.762142939747889e-05, "loss": 0.8348, "step": 172880 }, { "epoch": 14.97, "learning_rate": 3.7619261015381055e-05, "loss": 1.0684, "step": 172890 }, { "epoch": 14.97, "learning_rate": 3.7617092633283215e-05, "loss": 0.8469, "step": 172900 }, { "epoch": 14.98, "learning_rate": 3.761492425118538e-05, "loss": 0.9052, "step": 172910 }, { "epoch": 14.98, "learning_rate": 3.761275586908754e-05, "loss": 1.0856, "step": 172920 }, { "epoch": 14.98, "learning_rate": 3.761058748698971e-05, "loss": 0.8589, "step": 172930 }, { "epoch": 14.98, "learning_rate": 3.760841910489186e-05, "loss": 0.9419, "step": 172940 }, { "epoch": 14.98, "learning_rate": 3.7606250722794026e-05, "loss": 0.7333, "step": 172950 }, { "epoch": 14.98, "learning_rate": 3.760408234069619e-05, "loss": 0.9625, "step": 172960 }, { "epoch": 14.98, "learning_rate": 3.760191395859835e-05, "loss": 1.0437, "step": 172970 }, { "epoch": 14.98, "learning_rate": 3.759974557650052e-05, "loss": 1.0822, "step": 172980 }, { "epoch": 14.98, "learning_rate": 3.759757719440268e-05, "loss": 1.0668, "step": 172990 }, { "epoch": 14.98, "learning_rate": 3.7595408812304845e-05, "loss": 0.8684, "step": 173000 }, { "epoch": 14.98, "learning_rate": 3.759324043020701e-05, "loss": 1.0558, "step": 173010 }, { "epoch": 14.98, "learning_rate": 3.7591072048109164e-05, "loss": 0.9245, "step": 173020 }, { "epoch": 14.99, "learning_rate": 3.758890366601133e-05, "loss": 0.9407, "step": 173030 }, { "epoch": 14.99, "learning_rate": 3.758673528391349e-05, "loss": 0.853, "step": 173040 }, { "epoch": 14.99, "learning_rate": 3.7584566901815657e-05, "loss": 0.9622, "step": 173050 }, { "epoch": 14.99, "learning_rate": 3.7582398519717816e-05, "loss": 0.8688, "step": 173060 }, { "epoch": 14.99, "learning_rate": 3.758023013761998e-05, "loss": 1.0193, "step": 173070 }, { "epoch": 14.99, "learning_rate": 3.757806175552215e-05, "loss": 1.0854, "step": 173080 }, { "epoch": 14.99, "learning_rate": 3.757589337342431e-05, "loss": 1.0303, "step": 173090 }, { "epoch": 14.99, "learning_rate": 3.757372499132647e-05, "loss": 0.9762, "step": 173100 }, { "epoch": 14.99, "learning_rate": 3.757155660922863e-05, "loss": 1.0566, "step": 173110 }, { "epoch": 14.99, "learning_rate": 3.7569388227130794e-05, "loss": 0.9825, "step": 173120 }, { "epoch": 14.99, "learning_rate": 3.7567219845032954e-05, "loss": 0.8025, "step": 173130 }, { "epoch": 15.0, "learning_rate": 3.756505146293512e-05, "loss": 0.8006, "step": 173140 }, { "epoch": 15.0, "learning_rate": 3.7562883080837287e-05, "loss": 0.9599, "step": 173150 }, { "epoch": 15.0, "learning_rate": 3.7560714698739446e-05, "loss": 0.9668, "step": 173160 }, { "epoch": 15.0, "learning_rate": 3.755854631664161e-05, "loss": 0.9978, "step": 173170 }, { "epoch": 15.0, "learning_rate": 3.7556377934543765e-05, "loss": 0.9099, "step": 173180 }, { "epoch": 15.0, "learning_rate": 3.755420955244593e-05, "loss": 0.9076, "step": 173190 }, { "epoch": 15.0, "eval_cer": 0.5362124415458764, "eval_em": 0.04847859721505931, "eval_f1": 0.04847859721505931, "eval_loss": 1.4650990962982178, "eval_runtime": 235.7644, "eval_samples_per_second": 8.224, "eval_steps_per_second": 1.031, "eval_wer": 0.9515214027849407, "step": 173197 }, { "epoch": 15.0, "learning_rate": 3.755204117034809e-05, "loss": 0.9183, "step": 173200 }, { "epoch": 15.0, "learning_rate": 3.754987278825026e-05, "loss": 0.9116, "step": 173210 }, { "epoch": 15.0, "learning_rate": 3.754770440615242e-05, "loss": 0.912, "step": 173220 }, { "epoch": 15.0, "learning_rate": 3.7545536024054584e-05, "loss": 1.0424, "step": 173230 }, { "epoch": 15.0, "learning_rate": 3.754336764195675e-05, "loss": 0.9981, "step": 173240 }, { "epoch": 15.0, "learning_rate": 3.75411992598589e-05, "loss": 0.8214, "step": 173250 }, { "epoch": 15.01, "learning_rate": 3.753903087776107e-05, "loss": 0.8403, "step": 173260 }, { "epoch": 15.01, "learning_rate": 3.753686249566323e-05, "loss": 0.8308, "step": 173270 }, { "epoch": 15.01, "learning_rate": 3.7534694113565395e-05, "loss": 0.9475, "step": 173280 }, { "epoch": 15.01, "learning_rate": 3.7532525731467555e-05, "loss": 0.7198, "step": 173290 }, { "epoch": 15.01, "learning_rate": 3.753035734936972e-05, "loss": 0.9361, "step": 173300 }, { "epoch": 15.01, "learning_rate": 3.752818896727189e-05, "loss": 0.8967, "step": 173310 }, { "epoch": 15.01, "learning_rate": 3.752602058517405e-05, "loss": 0.796, "step": 173320 }, { "epoch": 15.01, "learning_rate": 3.752385220307621e-05, "loss": 0.8957, "step": 173330 }, { "epoch": 15.01, "learning_rate": 3.752168382097837e-05, "loss": 0.8377, "step": 173340 }, { "epoch": 15.01, "learning_rate": 3.751951543888053e-05, "loss": 0.92, "step": 173350 }, { "epoch": 15.01, "learning_rate": 3.751734705678269e-05, "loss": 0.8996, "step": 173360 }, { "epoch": 15.01, "learning_rate": 3.751517867468486e-05, "loss": 0.8728, "step": 173370 }, { "epoch": 15.02, "learning_rate": 3.7513010292587025e-05, "loss": 0.8372, "step": 173380 }, { "epoch": 15.02, "learning_rate": 3.7510841910489185e-05, "loss": 1.0507, "step": 173390 }, { "epoch": 15.02, "learning_rate": 3.750867352839135e-05, "loss": 0.8815, "step": 173400 }, { "epoch": 15.02, "learning_rate": 3.7506505146293504e-05, "loss": 0.9067, "step": 173410 }, { "epoch": 15.02, "learning_rate": 3.750433676419567e-05, "loss": 0.9258, "step": 173420 }, { "epoch": 15.02, "learning_rate": 3.750216838209783e-05, "loss": 0.8389, "step": 173430 }, { "epoch": 15.02, "learning_rate": 3.75e-05, "loss": 0.915, "step": 173440 }, { "epoch": 15.02, "learning_rate": 3.749783161790216e-05, "loss": 0.8294, "step": 173450 }, { "epoch": 15.02, "learning_rate": 3.749566323580432e-05, "loss": 1.0357, "step": 173460 }, { "epoch": 15.02, "learning_rate": 3.749349485370648e-05, "loss": 0.9498, "step": 173470 }, { "epoch": 15.02, "learning_rate": 3.749132647160865e-05, "loss": 0.8495, "step": 173480 }, { "epoch": 15.03, "learning_rate": 3.748915808951081e-05, "loss": 1.0128, "step": 173490 }, { "epoch": 15.03, "learning_rate": 3.748698970741297e-05, "loss": 1.1072, "step": 173500 }, { "epoch": 15.03, "learning_rate": 3.7484821325315134e-05, "loss": 0.8554, "step": 173510 }, { "epoch": 15.03, "learning_rate": 3.74826529432173e-05, "loss": 0.887, "step": 173520 }, { "epoch": 15.03, "learning_rate": 3.748048456111946e-05, "loss": 0.8027, "step": 173530 }, { "epoch": 15.03, "learning_rate": 3.747831617902162e-05, "loss": 1.0355, "step": 173540 }, { "epoch": 15.03, "learning_rate": 3.7476147796923786e-05, "loss": 0.9931, "step": 173550 }, { "epoch": 15.03, "learning_rate": 3.7473979414825946e-05, "loss": 0.8641, "step": 173560 }, { "epoch": 15.03, "learning_rate": 3.747181103272811e-05, "loss": 0.9639, "step": 173570 }, { "epoch": 15.03, "learning_rate": 3.746964265063027e-05, "loss": 0.8339, "step": 173580 }, { "epoch": 15.03, "learning_rate": 3.746747426853244e-05, "loss": 1.1279, "step": 173590 }, { "epoch": 15.03, "learning_rate": 3.74653058864346e-05, "loss": 0.8149, "step": 173600 }, { "epoch": 15.04, "learning_rate": 3.7463137504336764e-05, "loss": 0.8032, "step": 173610 }, { "epoch": 15.04, "learning_rate": 3.7460969122238924e-05, "loss": 0.6232, "step": 173620 }, { "epoch": 15.04, "learning_rate": 3.7458800740141084e-05, "loss": 0.8411, "step": 173630 }, { "epoch": 15.04, "learning_rate": 3.745663235804325e-05, "loss": 0.8684, "step": 173640 }, { "epoch": 15.04, "learning_rate": 3.7454463975945416e-05, "loss": 0.9317, "step": 173650 }, { "epoch": 15.04, "learning_rate": 3.7452295593847576e-05, "loss": 0.8814, "step": 173660 }, { "epoch": 15.04, "learning_rate": 3.7450127211749736e-05, "loss": 0.9989, "step": 173670 }, { "epoch": 15.04, "learning_rate": 3.74479588296519e-05, "loss": 0.7666, "step": 173680 }, { "epoch": 15.04, "learning_rate": 3.744579044755406e-05, "loss": 0.8612, "step": 173690 }, { "epoch": 15.04, "learning_rate": 3.744362206545622e-05, "loss": 0.7354, "step": 173700 }, { "epoch": 15.04, "learning_rate": 3.744145368335839e-05, "loss": 0.8959, "step": 173710 }, { "epoch": 15.05, "learning_rate": 3.743928530126055e-05, "loss": 0.9751, "step": 173720 }, { "epoch": 15.05, "learning_rate": 3.743711691916271e-05, "loss": 1.0051, "step": 173730 }, { "epoch": 15.05, "learning_rate": 3.743494853706487e-05, "loss": 0.8403, "step": 173740 }, { "epoch": 15.05, "learning_rate": 3.743278015496704e-05, "loss": 0.8745, "step": 173750 }, { "epoch": 15.05, "learning_rate": 3.74306117728692e-05, "loss": 1.0058, "step": 173760 }, { "epoch": 15.05, "learning_rate": 3.742844339077136e-05, "loss": 0.9807, "step": 173770 }, { "epoch": 15.05, "learning_rate": 3.7426275008673525e-05, "loss": 1.0066, "step": 173780 }, { "epoch": 15.05, "learning_rate": 3.7424106626575685e-05, "loss": 0.8566, "step": 173790 }, { "epoch": 15.05, "learning_rate": 3.742193824447785e-05, "loss": 0.9584, "step": 173800 }, { "epoch": 15.05, "learning_rate": 3.741976986238001e-05, "loss": 0.8905, "step": 173810 }, { "epoch": 15.05, "learning_rate": 3.741760148028218e-05, "loss": 0.7817, "step": 173820 }, { "epoch": 15.05, "learning_rate": 3.741543309818434e-05, "loss": 0.8124, "step": 173830 }, { "epoch": 15.06, "learning_rate": 3.74132647160865e-05, "loss": 0.9492, "step": 173840 }, { "epoch": 15.06, "learning_rate": 3.741109633398866e-05, "loss": 1.0359, "step": 173850 }, { "epoch": 15.06, "learning_rate": 3.740892795189082e-05, "loss": 0.7955, "step": 173860 }, { "epoch": 15.06, "learning_rate": 3.740675956979299e-05, "loss": 0.7685, "step": 173870 }, { "epoch": 15.06, "learning_rate": 3.7404591187695155e-05, "loss": 0.7686, "step": 173880 }, { "epoch": 15.06, "learning_rate": 3.7402422805597315e-05, "loss": 0.8726, "step": 173890 }, { "epoch": 15.06, "learning_rate": 3.7400254423499475e-05, "loss": 0.8121, "step": 173900 }, { "epoch": 15.06, "learning_rate": 3.739808604140164e-05, "loss": 1.0012, "step": 173910 }, { "epoch": 15.06, "learning_rate": 3.73959176593038e-05, "loss": 0.8816, "step": 173920 }, { "epoch": 15.06, "learning_rate": 3.739374927720596e-05, "loss": 0.8665, "step": 173930 }, { "epoch": 15.06, "learning_rate": 3.7391580895108127e-05, "loss": 0.9517, "step": 173940 }, { "epoch": 15.07, "learning_rate": 3.738941251301029e-05, "loss": 1.1571, "step": 173950 }, { "epoch": 15.07, "learning_rate": 3.738724413091245e-05, "loss": 0.826, "step": 173960 }, { "epoch": 15.07, "learning_rate": 3.738507574881461e-05, "loss": 0.9531, "step": 173970 }, { "epoch": 15.07, "learning_rate": 3.738290736671678e-05, "loss": 0.8164, "step": 173980 }, { "epoch": 15.07, "learning_rate": 3.738073898461894e-05, "loss": 0.8923, "step": 173990 }, { "epoch": 15.07, "learning_rate": 3.73785706025211e-05, "loss": 0.9278, "step": 174000 }, { "epoch": 15.07, "learning_rate": 3.7376402220423264e-05, "loss": 0.9828, "step": 174010 }, { "epoch": 15.07, "learning_rate": 3.737423383832543e-05, "loss": 0.9294, "step": 174020 }, { "epoch": 15.07, "learning_rate": 3.737206545622759e-05, "loss": 1.0694, "step": 174030 }, { "epoch": 15.07, "learning_rate": 3.736989707412975e-05, "loss": 0.9342, "step": 174040 }, { "epoch": 15.07, "learning_rate": 3.7367728692031916e-05, "loss": 0.8803, "step": 174050 }, { "epoch": 15.07, "learning_rate": 3.7365560309934076e-05, "loss": 0.8854, "step": 174060 }, { "epoch": 15.08, "learning_rate": 3.736339192783624e-05, "loss": 0.9852, "step": 174070 }, { "epoch": 15.08, "learning_rate": 3.73612235457384e-05, "loss": 1.0018, "step": 174080 }, { "epoch": 15.08, "learning_rate": 3.735905516364057e-05, "loss": 0.8958, "step": 174090 }, { "epoch": 15.08, "learning_rate": 3.735688678154273e-05, "loss": 0.9927, "step": 174100 }, { "epoch": 15.08, "learning_rate": 3.7354718399444894e-05, "loss": 1.0664, "step": 174110 }, { "epoch": 15.08, "learning_rate": 3.7352550017347054e-05, "loss": 0.871, "step": 174120 }, { "epoch": 15.08, "learning_rate": 3.7350381635249213e-05, "loss": 1.0296, "step": 174130 }, { "epoch": 15.08, "learning_rate": 3.734821325315138e-05, "loss": 0.9284, "step": 174140 }, { "epoch": 15.08, "learning_rate": 3.7346044871053546e-05, "loss": 0.7754, "step": 174150 }, { "epoch": 15.08, "learning_rate": 3.7343876488955706e-05, "loss": 0.9996, "step": 174160 }, { "epoch": 15.08, "learning_rate": 3.7341708106857865e-05, "loss": 0.8023, "step": 174170 }, { "epoch": 15.09, "learning_rate": 3.733953972476003e-05, "loss": 0.9025, "step": 174180 }, { "epoch": 15.09, "learning_rate": 3.733737134266219e-05, "loss": 0.9098, "step": 174190 }, { "epoch": 15.09, "learning_rate": 3.733520296056435e-05, "loss": 0.8835, "step": 174200 }, { "epoch": 15.09, "learning_rate": 3.733303457846652e-05, "loss": 0.8033, "step": 174210 }, { "epoch": 15.09, "learning_rate": 3.7330866196368684e-05, "loss": 0.819, "step": 174220 }, { "epoch": 15.09, "learning_rate": 3.732869781427084e-05, "loss": 0.9392, "step": 174230 }, { "epoch": 15.09, "learning_rate": 3.7326529432173e-05, "loss": 0.8248, "step": 174240 }, { "epoch": 15.09, "learning_rate": 3.732436105007517e-05, "loss": 0.8954, "step": 174250 }, { "epoch": 15.09, "learning_rate": 3.732219266797733e-05, "loss": 1.0496, "step": 174260 }, { "epoch": 15.09, "learning_rate": 3.732002428587949e-05, "loss": 0.825, "step": 174270 }, { "epoch": 15.09, "learning_rate": 3.7317855903781655e-05, "loss": 0.716, "step": 174280 }, { "epoch": 15.09, "learning_rate": 3.7315687521683815e-05, "loss": 0.7892, "step": 174290 }, { "epoch": 15.1, "learning_rate": 3.731351913958598e-05, "loss": 0.743, "step": 174300 }, { "epoch": 15.1, "learning_rate": 3.731135075748814e-05, "loss": 0.7755, "step": 174310 }, { "epoch": 15.1, "learning_rate": 3.730918237539031e-05, "loss": 0.8994, "step": 174320 }, { "epoch": 15.1, "learning_rate": 3.730701399329247e-05, "loss": 0.9111, "step": 174330 }, { "epoch": 15.1, "learning_rate": 3.730484561119463e-05, "loss": 0.8523, "step": 174340 }, { "epoch": 15.1, "learning_rate": 3.730267722909679e-05, "loss": 0.8619, "step": 174350 }, { "epoch": 15.1, "learning_rate": 3.730050884699895e-05, "loss": 0.8797, "step": 174360 }, { "epoch": 15.1, "learning_rate": 3.729834046490112e-05, "loss": 0.7811, "step": 174370 }, { "epoch": 15.1, "learning_rate": 3.7296172082803285e-05, "loss": 1.0698, "step": 174380 }, { "epoch": 15.1, "learning_rate": 3.7294003700705445e-05, "loss": 0.9819, "step": 174390 }, { "epoch": 15.1, "learning_rate": 3.7291835318607604e-05, "loss": 0.9608, "step": 174400 }, { "epoch": 15.11, "learning_rate": 3.728966693650977e-05, "loss": 1.2062, "step": 174410 }, { "epoch": 15.11, "learning_rate": 3.728749855441193e-05, "loss": 0.8725, "step": 174420 }, { "epoch": 15.11, "learning_rate": 3.728533017231409e-05, "loss": 0.8599, "step": 174430 }, { "epoch": 15.11, "learning_rate": 3.7283161790216256e-05, "loss": 0.8276, "step": 174440 }, { "epoch": 15.11, "learning_rate": 3.728099340811842e-05, "loss": 0.9717, "step": 174450 }, { "epoch": 15.11, "learning_rate": 3.727882502602058e-05, "loss": 0.748, "step": 174460 }, { "epoch": 15.11, "learning_rate": 3.727665664392274e-05, "loss": 0.8933, "step": 174470 }, { "epoch": 15.11, "learning_rate": 3.727448826182491e-05, "loss": 0.9367, "step": 174480 }, { "epoch": 15.11, "learning_rate": 3.727231987972707e-05, "loss": 0.9583, "step": 174490 }, { "epoch": 15.11, "learning_rate": 3.727015149762923e-05, "loss": 0.9735, "step": 174500 }, { "epoch": 15.11, "learning_rate": 3.7267983115531394e-05, "loss": 0.8299, "step": 174510 }, { "epoch": 15.11, "learning_rate": 3.726581473343356e-05, "loss": 0.8864, "step": 174520 }, { "epoch": 15.12, "learning_rate": 3.726364635133572e-05, "loss": 0.6483, "step": 174530 }, { "epoch": 15.12, "learning_rate": 3.726147796923788e-05, "loss": 0.9231, "step": 174540 }, { "epoch": 15.12, "learning_rate": 3.7259309587140046e-05, "loss": 0.7619, "step": 174550 }, { "epoch": 15.12, "learning_rate": 3.7257141205042206e-05, "loss": 0.8548, "step": 174560 }, { "epoch": 15.12, "learning_rate": 3.725497282294437e-05, "loss": 0.9095, "step": 174570 }, { "epoch": 15.12, "learning_rate": 3.725280444084653e-05, "loss": 0.8695, "step": 174580 }, { "epoch": 15.12, "learning_rate": 3.72506360587487e-05, "loss": 0.8402, "step": 174590 }, { "epoch": 15.12, "learning_rate": 3.724846767665086e-05, "loss": 0.8328, "step": 174600 }, { "epoch": 15.12, "learning_rate": 3.7246299294553024e-05, "loss": 1.0431, "step": 174610 }, { "epoch": 15.12, "learning_rate": 3.7244130912455184e-05, "loss": 0.9033, "step": 174620 }, { "epoch": 15.12, "learning_rate": 3.724196253035734e-05, "loss": 0.9678, "step": 174630 }, { "epoch": 15.12, "learning_rate": 3.723979414825951e-05, "loss": 0.8476, "step": 174640 }, { "epoch": 15.13, "learning_rate": 3.7237625766161676e-05, "loss": 0.9544, "step": 174650 }, { "epoch": 15.13, "learning_rate": 3.7235457384063836e-05, "loss": 1.0756, "step": 174660 }, { "epoch": 15.13, "learning_rate": 3.7233289001965995e-05, "loss": 0.8861, "step": 174670 }, { "epoch": 15.13, "learning_rate": 3.723112061986816e-05, "loss": 0.932, "step": 174680 }, { "epoch": 15.13, "learning_rate": 3.722895223777032e-05, "loss": 0.9715, "step": 174690 }, { "epoch": 15.13, "learning_rate": 3.722678385567248e-05, "loss": 0.682, "step": 174700 }, { "epoch": 15.13, "learning_rate": 3.722461547357465e-05, "loss": 0.9454, "step": 174710 }, { "epoch": 15.13, "learning_rate": 3.7222447091476814e-05, "loss": 0.932, "step": 174720 }, { "epoch": 15.13, "learning_rate": 3.722027870937897e-05, "loss": 0.9239, "step": 174730 }, { "epoch": 15.13, "learning_rate": 3.721811032728113e-05, "loss": 0.8736, "step": 174740 }, { "epoch": 15.13, "learning_rate": 3.72159419451833e-05, "loss": 0.8352, "step": 174750 }, { "epoch": 15.14, "learning_rate": 3.721377356308546e-05, "loss": 0.8861, "step": 174760 }, { "epoch": 15.14, "learning_rate": 3.721160518098762e-05, "loss": 0.8334, "step": 174770 }, { "epoch": 15.14, "learning_rate": 3.7209436798889785e-05, "loss": 1.049, "step": 174780 }, { "epoch": 15.14, "learning_rate": 3.7207268416791945e-05, "loss": 1.1, "step": 174790 }, { "epoch": 15.14, "learning_rate": 3.720510003469411e-05, "loss": 0.7833, "step": 174800 }, { "epoch": 15.14, "learning_rate": 3.720293165259627e-05, "loss": 0.9043, "step": 174810 }, { "epoch": 15.14, "learning_rate": 3.720076327049844e-05, "loss": 0.8614, "step": 174820 }, { "epoch": 15.14, "learning_rate": 3.7198594888400597e-05, "loss": 0.831, "step": 174830 }, { "epoch": 15.14, "learning_rate": 3.719642650630276e-05, "loss": 0.8661, "step": 174840 }, { "epoch": 15.14, "learning_rate": 3.719425812420492e-05, "loss": 0.8767, "step": 174850 }, { "epoch": 15.14, "learning_rate": 3.719208974210708e-05, "loss": 0.9716, "step": 174860 }, { "epoch": 15.14, "learning_rate": 3.718992136000925e-05, "loss": 0.8801, "step": 174870 }, { "epoch": 15.15, "learning_rate": 3.7187752977911415e-05, "loss": 0.8309, "step": 174880 }, { "epoch": 15.15, "learning_rate": 3.7185584595813575e-05, "loss": 0.7794, "step": 174890 }, { "epoch": 15.15, "learning_rate": 3.7183416213715734e-05, "loss": 1.0036, "step": 174900 }, { "epoch": 15.15, "learning_rate": 3.71812478316179e-05, "loss": 0.8949, "step": 174910 }, { "epoch": 15.15, "learning_rate": 3.717907944952006e-05, "loss": 1.0223, "step": 174920 }, { "epoch": 15.15, "learning_rate": 3.717691106742222e-05, "loss": 0.8779, "step": 174930 }, { "epoch": 15.15, "learning_rate": 3.7174742685324386e-05, "loss": 0.9264, "step": 174940 }, { "epoch": 15.15, "learning_rate": 3.717257430322655e-05, "loss": 0.8483, "step": 174950 }, { "epoch": 15.15, "learning_rate": 3.717040592112871e-05, "loss": 0.7469, "step": 174960 }, { "epoch": 15.15, "learning_rate": 3.716823753903087e-05, "loss": 0.9349, "step": 174970 }, { "epoch": 15.15, "learning_rate": 3.716606915693304e-05, "loss": 0.9935, "step": 174980 }, { "epoch": 15.16, "learning_rate": 3.71639007748352e-05, "loss": 0.928, "step": 174990 }, { "epoch": 15.16, "learning_rate": 3.716173239273736e-05, "loss": 0.9656, "step": 175000 }, { "epoch": 15.16, "learning_rate": 3.7159564010639524e-05, "loss": 0.8557, "step": 175010 }, { "epoch": 15.16, "learning_rate": 3.715739562854169e-05, "loss": 0.876, "step": 175020 }, { "epoch": 15.16, "learning_rate": 3.715522724644385e-05, "loss": 0.9988, "step": 175030 }, { "epoch": 15.16, "learning_rate": 3.715305886434601e-05, "loss": 0.8479, "step": 175040 }, { "epoch": 15.16, "learning_rate": 3.7150890482248176e-05, "loss": 1.0192, "step": 175050 }, { "epoch": 15.16, "learning_rate": 3.7148722100150336e-05, "loss": 0.8375, "step": 175060 }, { "epoch": 15.16, "learning_rate": 3.71465537180525e-05, "loss": 0.9982, "step": 175070 }, { "epoch": 15.16, "learning_rate": 3.714438533595466e-05, "loss": 0.8868, "step": 175080 }, { "epoch": 15.16, "learning_rate": 3.714221695385683e-05, "loss": 0.7875, "step": 175090 }, { "epoch": 15.16, "learning_rate": 3.714004857175899e-05, "loss": 0.9423, "step": 175100 }, { "epoch": 15.17, "learning_rate": 3.7137880189661154e-05, "loss": 0.8288, "step": 175110 }, { "epoch": 15.17, "learning_rate": 3.7135711807563314e-05, "loss": 0.8531, "step": 175120 }, { "epoch": 15.17, "learning_rate": 3.713354342546547e-05, "loss": 0.966, "step": 175130 }, { "epoch": 15.17, "learning_rate": 3.713137504336764e-05, "loss": 0.9821, "step": 175140 }, { "epoch": 15.17, "learning_rate": 3.7129206661269806e-05, "loss": 0.8814, "step": 175150 }, { "epoch": 15.17, "learning_rate": 3.7127038279171966e-05, "loss": 0.9572, "step": 175160 }, { "epoch": 15.17, "learning_rate": 3.7124869897074125e-05, "loss": 0.978, "step": 175170 }, { "epoch": 15.17, "learning_rate": 3.712270151497629e-05, "loss": 1.0201, "step": 175180 }, { "epoch": 15.17, "learning_rate": 3.712053313287845e-05, "loss": 0.8191, "step": 175190 }, { "epoch": 15.17, "learning_rate": 3.711836475078061e-05, "loss": 0.7085, "step": 175200 }, { "epoch": 15.17, "learning_rate": 3.711619636868278e-05, "loss": 0.9305, "step": 175210 }, { "epoch": 15.18, "learning_rate": 3.7114027986584944e-05, "loss": 0.7749, "step": 175220 }, { "epoch": 15.18, "learning_rate": 3.71118596044871e-05, "loss": 0.9511, "step": 175230 }, { "epoch": 15.18, "learning_rate": 3.710969122238926e-05, "loss": 0.8824, "step": 175240 }, { "epoch": 15.18, "learning_rate": 3.710752284029143e-05, "loss": 0.851, "step": 175250 }, { "epoch": 15.18, "learning_rate": 3.710535445819359e-05, "loss": 0.9152, "step": 175260 }, { "epoch": 15.18, "learning_rate": 3.710318607609575e-05, "loss": 0.9605, "step": 175270 }, { "epoch": 15.18, "learning_rate": 3.7101017693997915e-05, "loss": 0.9189, "step": 175280 }, { "epoch": 15.18, "learning_rate": 3.709884931190008e-05, "loss": 0.9085, "step": 175290 }, { "epoch": 15.18, "learning_rate": 3.709668092980224e-05, "loss": 0.7518, "step": 175300 }, { "epoch": 15.18, "learning_rate": 3.70945125477044e-05, "loss": 0.9328, "step": 175310 }, { "epoch": 15.18, "learning_rate": 3.709234416560657e-05, "loss": 0.9281, "step": 175320 }, { "epoch": 15.18, "learning_rate": 3.7090175783508726e-05, "loss": 1.0877, "step": 175330 }, { "epoch": 15.19, "learning_rate": 3.708800740141089e-05, "loss": 0.8842, "step": 175340 }, { "epoch": 15.19, "learning_rate": 3.708583901931305e-05, "loss": 0.9405, "step": 175350 }, { "epoch": 15.19, "learning_rate": 3.708367063721521e-05, "loss": 0.8179, "step": 175360 }, { "epoch": 15.19, "learning_rate": 3.708150225511738e-05, "loss": 0.8785, "step": 175370 }, { "epoch": 15.19, "learning_rate": 3.7079333873019545e-05, "loss": 0.8868, "step": 175380 }, { "epoch": 15.19, "learning_rate": 3.7077165490921704e-05, "loss": 0.8661, "step": 175390 }, { "epoch": 15.19, "learning_rate": 3.7074997108823864e-05, "loss": 0.8112, "step": 175400 }, { "epoch": 15.19, "learning_rate": 3.707282872672603e-05, "loss": 0.9102, "step": 175410 }, { "epoch": 15.19, "learning_rate": 3.707066034462819e-05, "loss": 1.0427, "step": 175420 }, { "epoch": 15.19, "learning_rate": 3.706849196253035e-05, "loss": 0.9202, "step": 175430 }, { "epoch": 15.19, "learning_rate": 3.7066323580432516e-05, "loss": 0.8563, "step": 175440 }, { "epoch": 15.2, "learning_rate": 3.706415519833468e-05, "loss": 0.9801, "step": 175450 }, { "epoch": 15.2, "learning_rate": 3.706198681623684e-05, "loss": 0.8772, "step": 175460 }, { "epoch": 15.2, "learning_rate": 3.7059818434139e-05, "loss": 0.9198, "step": 175470 }, { "epoch": 15.2, "learning_rate": 3.705765005204117e-05, "loss": 0.9524, "step": 175480 }, { "epoch": 15.2, "learning_rate": 3.705548166994333e-05, "loss": 1.0133, "step": 175490 }, { "epoch": 15.2, "learning_rate": 3.705331328784549e-05, "loss": 0.9606, "step": 175500 }, { "epoch": 15.2, "learning_rate": 3.7051144905747654e-05, "loss": 0.9271, "step": 175510 }, { "epoch": 15.2, "learning_rate": 3.704897652364982e-05, "loss": 0.9229, "step": 175520 }, { "epoch": 15.2, "learning_rate": 3.704680814155198e-05, "loss": 0.8327, "step": 175530 }, { "epoch": 15.2, "learning_rate": 3.704463975945414e-05, "loss": 0.8033, "step": 175540 }, { "epoch": 15.2, "learning_rate": 3.7042471377356306e-05, "loss": 0.7334, "step": 175550 }, { "epoch": 15.2, "learning_rate": 3.7040302995258465e-05, "loss": 0.9455, "step": 175560 }, { "epoch": 15.21, "learning_rate": 3.703813461316063e-05, "loss": 0.7638, "step": 175570 }, { "epoch": 15.21, "learning_rate": 3.703596623106279e-05, "loss": 0.9191, "step": 175580 }, { "epoch": 15.21, "learning_rate": 3.703379784896496e-05, "loss": 0.9077, "step": 175590 }, { "epoch": 15.21, "learning_rate": 3.703162946686712e-05, "loss": 0.9331, "step": 175600 }, { "epoch": 15.21, "learning_rate": 3.7029461084769284e-05, "loss": 0.9436, "step": 175610 }, { "epoch": 15.21, "learning_rate": 3.7027292702671443e-05, "loss": 1.1729, "step": 175620 }, { "epoch": 15.21, "learning_rate": 3.70251243205736e-05, "loss": 0.8858, "step": 175630 }, { "epoch": 15.21, "learning_rate": 3.702295593847577e-05, "loss": 0.9847, "step": 175640 }, { "epoch": 15.21, "learning_rate": 3.7020787556377936e-05, "loss": 0.8123, "step": 175650 }, { "epoch": 15.21, "learning_rate": 3.7018619174280095e-05, "loss": 0.8335, "step": 175660 }, { "epoch": 15.21, "learning_rate": 3.7016450792182255e-05, "loss": 0.7594, "step": 175670 }, { "epoch": 15.22, "learning_rate": 3.701428241008442e-05, "loss": 1.0231, "step": 175680 }, { "epoch": 15.22, "learning_rate": 3.701211402798658e-05, "loss": 0.9117, "step": 175690 }, { "epoch": 15.22, "learning_rate": 3.700994564588874e-05, "loss": 0.964, "step": 175700 }, { "epoch": 15.22, "learning_rate": 3.700777726379091e-05, "loss": 0.9413, "step": 175710 }, { "epoch": 15.22, "learning_rate": 3.7005608881693073e-05, "loss": 0.9053, "step": 175720 }, { "epoch": 15.22, "learning_rate": 3.700344049959523e-05, "loss": 0.8907, "step": 175730 }, { "epoch": 15.22, "learning_rate": 3.700127211749739e-05, "loss": 0.6611, "step": 175740 }, { "epoch": 15.22, "learning_rate": 3.699910373539956e-05, "loss": 0.9271, "step": 175750 }, { "epoch": 15.22, "learning_rate": 3.699693535330172e-05, "loss": 0.7732, "step": 175760 }, { "epoch": 15.22, "learning_rate": 3.699476697120388e-05, "loss": 0.8533, "step": 175770 }, { "epoch": 15.22, "learning_rate": 3.6992598589106045e-05, "loss": 0.903, "step": 175780 }, { "epoch": 15.22, "learning_rate": 3.699043020700821e-05, "loss": 0.8475, "step": 175790 }, { "epoch": 15.23, "learning_rate": 3.698826182491037e-05, "loss": 0.9615, "step": 175800 }, { "epoch": 15.23, "learning_rate": 3.698609344281253e-05, "loss": 0.8665, "step": 175810 }, { "epoch": 15.23, "learning_rate": 3.69839250607147e-05, "loss": 0.9814, "step": 175820 }, { "epoch": 15.23, "learning_rate": 3.6981756678616856e-05, "loss": 1.0907, "step": 175830 }, { "epoch": 15.23, "learning_rate": 3.697958829651902e-05, "loss": 0.9763, "step": 175840 }, { "epoch": 15.23, "learning_rate": 3.697741991442118e-05, "loss": 0.9057, "step": 175850 }, { "epoch": 15.23, "learning_rate": 3.697525153232334e-05, "loss": 0.848, "step": 175860 }, { "epoch": 15.23, "learning_rate": 3.697308315022551e-05, "loss": 0.694, "step": 175870 }, { "epoch": 15.23, "learning_rate": 3.6970914768127675e-05, "loss": 0.8826, "step": 175880 }, { "epoch": 15.23, "learning_rate": 3.6968746386029834e-05, "loss": 0.7674, "step": 175890 }, { "epoch": 15.23, "learning_rate": 3.6966578003931994e-05, "loss": 1.068, "step": 175900 }, { "epoch": 15.23, "learning_rate": 3.696440962183416e-05, "loss": 0.8297, "step": 175910 }, { "epoch": 15.24, "learning_rate": 3.696224123973632e-05, "loss": 0.8846, "step": 175920 }, { "epoch": 15.24, "learning_rate": 3.696007285763848e-05, "loss": 1.2676, "step": 175930 }, { "epoch": 15.24, "learning_rate": 3.6957904475540646e-05, "loss": 0.8354, "step": 175940 }, { "epoch": 15.24, "learning_rate": 3.695573609344281e-05, "loss": 0.7964, "step": 175950 }, { "epoch": 15.24, "learning_rate": 3.695356771134497e-05, "loss": 0.7241, "step": 175960 }, { "epoch": 15.24, "learning_rate": 3.695139932924713e-05, "loss": 0.8693, "step": 175970 }, { "epoch": 15.24, "learning_rate": 3.69492309471493e-05, "loss": 0.9485, "step": 175980 }, { "epoch": 15.24, "learning_rate": 3.694706256505146e-05, "loss": 0.9722, "step": 175990 }, { "epoch": 15.24, "learning_rate": 3.6944894182953624e-05, "loss": 0.8488, "step": 176000 }, { "epoch": 15.24, "learning_rate": 3.6942725800855784e-05, "loss": 0.8185, "step": 176010 }, { "epoch": 15.24, "learning_rate": 3.694055741875795e-05, "loss": 0.8984, "step": 176020 }, { "epoch": 15.25, "learning_rate": 3.693838903666011e-05, "loss": 0.7965, "step": 176030 }, { "epoch": 15.25, "learning_rate": 3.693622065456227e-05, "loss": 0.9247, "step": 176040 }, { "epoch": 15.25, "learning_rate": 3.6934052272464436e-05, "loss": 1.0194, "step": 176050 }, { "epoch": 15.25, "learning_rate": 3.6931883890366595e-05, "loss": 0.9171, "step": 176060 }, { "epoch": 15.25, "learning_rate": 3.692971550826876e-05, "loss": 0.9249, "step": 176070 }, { "epoch": 15.25, "learning_rate": 3.692754712617092e-05, "loss": 0.8926, "step": 176080 }, { "epoch": 15.25, "learning_rate": 3.692537874407309e-05, "loss": 1.0018, "step": 176090 }, { "epoch": 15.25, "learning_rate": 3.692321036197525e-05, "loss": 0.9436, "step": 176100 }, { "epoch": 15.25, "learning_rate": 3.6921041979877414e-05, "loss": 1.0927, "step": 176110 }, { "epoch": 15.25, "learning_rate": 3.691887359777957e-05, "loss": 0.7976, "step": 176120 }, { "epoch": 15.25, "learning_rate": 3.691670521568173e-05, "loss": 0.9007, "step": 176130 }, { "epoch": 15.25, "learning_rate": 3.69145368335839e-05, "loss": 0.8665, "step": 176140 }, { "epoch": 15.26, "learning_rate": 3.6912368451486066e-05, "loss": 0.9582, "step": 176150 }, { "epoch": 15.26, "learning_rate": 3.6910200069388225e-05, "loss": 0.798, "step": 176160 }, { "epoch": 15.26, "learning_rate": 3.6908031687290385e-05, "loss": 0.9291, "step": 176170 }, { "epoch": 15.26, "learning_rate": 3.690586330519255e-05, "loss": 1.0165, "step": 176180 }, { "epoch": 15.26, "learning_rate": 3.690369492309471e-05, "loss": 0.9146, "step": 176190 }, { "epoch": 15.26, "learning_rate": 3.690152654099687e-05, "loss": 0.8378, "step": 176200 }, { "epoch": 15.26, "learning_rate": 3.689935815889904e-05, "loss": 0.9888, "step": 176210 }, { "epoch": 15.26, "learning_rate": 3.68971897768012e-05, "loss": 1.0678, "step": 176220 }, { "epoch": 15.26, "learning_rate": 3.689502139470336e-05, "loss": 0.9447, "step": 176230 }, { "epoch": 15.26, "learning_rate": 3.689285301260552e-05, "loss": 0.9783, "step": 176240 }, { "epoch": 15.26, "learning_rate": 3.689068463050769e-05, "loss": 0.9576, "step": 176250 }, { "epoch": 15.27, "learning_rate": 3.688851624840985e-05, "loss": 0.8372, "step": 176260 }, { "epoch": 15.27, "learning_rate": 3.688634786631201e-05, "loss": 0.9653, "step": 176270 }, { "epoch": 15.27, "learning_rate": 3.6884179484214175e-05, "loss": 0.9415, "step": 176280 }, { "epoch": 15.27, "learning_rate": 3.688201110211634e-05, "loss": 0.9384, "step": 176290 }, { "epoch": 15.27, "learning_rate": 3.68798427200185e-05, "loss": 0.8827, "step": 176300 }, { "epoch": 15.27, "learning_rate": 3.687767433792066e-05, "loss": 0.8177, "step": 176310 }, { "epoch": 15.27, "learning_rate": 3.6875505955822827e-05, "loss": 0.9171, "step": 176320 }, { "epoch": 15.27, "learning_rate": 3.6873337573724986e-05, "loss": 0.6932, "step": 176330 }, { "epoch": 15.27, "learning_rate": 3.687116919162715e-05, "loss": 0.9373, "step": 176340 }, { "epoch": 15.27, "learning_rate": 3.686900080952931e-05, "loss": 0.855, "step": 176350 }, { "epoch": 15.27, "learning_rate": 3.686683242743148e-05, "loss": 0.9202, "step": 176360 }, { "epoch": 15.27, "learning_rate": 3.686466404533364e-05, "loss": 0.9856, "step": 176370 }, { "epoch": 15.28, "learning_rate": 3.6862495663235805e-05, "loss": 0.9274, "step": 176380 }, { "epoch": 15.28, "learning_rate": 3.6860327281137964e-05, "loss": 1.0782, "step": 176390 }, { "epoch": 15.28, "learning_rate": 3.6858158899040124e-05, "loss": 1.0, "step": 176400 }, { "epoch": 15.28, "learning_rate": 3.685599051694229e-05, "loss": 0.6638, "step": 176410 }, { "epoch": 15.28, "learning_rate": 3.685382213484445e-05, "loss": 0.897, "step": 176420 }, { "epoch": 15.28, "learning_rate": 3.685165375274661e-05, "loss": 0.973, "step": 176430 }, { "epoch": 15.28, "learning_rate": 3.6849485370648776e-05, "loss": 0.7964, "step": 176440 }, { "epoch": 15.28, "learning_rate": 3.684731698855094e-05, "loss": 1.0251, "step": 176450 }, { "epoch": 15.28, "learning_rate": 3.68451486064531e-05, "loss": 0.7082, "step": 176460 }, { "epoch": 15.28, "learning_rate": 3.684298022435526e-05, "loss": 1.0335, "step": 176470 }, { "epoch": 15.28, "learning_rate": 3.684081184225743e-05, "loss": 0.8301, "step": 176480 }, { "epoch": 15.29, "learning_rate": 3.683864346015959e-05, "loss": 0.8273, "step": 176490 }, { "epoch": 15.29, "learning_rate": 3.6836475078061754e-05, "loss": 1.0043, "step": 176500 }, { "epoch": 15.29, "learning_rate": 3.6834306695963913e-05, "loss": 0.9958, "step": 176510 }, { "epoch": 15.29, "learning_rate": 3.683213831386608e-05, "loss": 0.8468, "step": 176520 }, { "epoch": 15.29, "learning_rate": 3.682996993176824e-05, "loss": 0.8624, "step": 176530 }, { "epoch": 15.29, "learning_rate": 3.68278015496704e-05, "loss": 0.9011, "step": 176540 }, { "epoch": 15.29, "learning_rate": 3.6825633167572565e-05, "loss": 0.913, "step": 176550 }, { "epoch": 15.29, "learning_rate": 3.6823464785474725e-05, "loss": 0.9008, "step": 176560 }, { "epoch": 15.29, "learning_rate": 3.682129640337689e-05, "loss": 0.9098, "step": 176570 }, { "epoch": 15.29, "learning_rate": 3.681912802127905e-05, "loss": 0.774, "step": 176580 }, { "epoch": 15.29, "learning_rate": 3.681695963918122e-05, "loss": 1.0112, "step": 176590 }, { "epoch": 15.29, "learning_rate": 3.681479125708338e-05, "loss": 0.9202, "step": 176600 }, { "epoch": 15.3, "learning_rate": 3.6812622874985543e-05, "loss": 0.8629, "step": 176610 }, { "epoch": 15.3, "learning_rate": 3.68104544928877e-05, "loss": 0.7804, "step": 176620 }, { "epoch": 15.3, "learning_rate": 3.680828611078986e-05, "loss": 0.9037, "step": 176630 }, { "epoch": 15.3, "learning_rate": 3.680611772869203e-05, "loss": 0.7338, "step": 176640 }, { "epoch": 15.3, "learning_rate": 3.6803949346594195e-05, "loss": 0.7632, "step": 176650 }, { "epoch": 15.3, "learning_rate": 3.6801780964496355e-05, "loss": 0.8195, "step": 176660 }, { "epoch": 15.3, "learning_rate": 3.6799612582398515e-05, "loss": 0.9029, "step": 176670 }, { "epoch": 15.3, "learning_rate": 3.679744420030068e-05, "loss": 0.9482, "step": 176680 }, { "epoch": 15.3, "learning_rate": 3.679527581820284e-05, "loss": 0.8522, "step": 176690 }, { "epoch": 15.3, "learning_rate": 3.6793107436105e-05, "loss": 1.0562, "step": 176700 }, { "epoch": 15.3, "learning_rate": 3.679093905400717e-05, "loss": 0.9479, "step": 176710 }, { "epoch": 15.31, "learning_rate": 3.678877067190933e-05, "loss": 1.014, "step": 176720 }, { "epoch": 15.31, "learning_rate": 3.678660228981149e-05, "loss": 1.0413, "step": 176730 }, { "epoch": 15.31, "learning_rate": 3.678443390771365e-05, "loss": 0.781, "step": 176740 }, { "epoch": 15.31, "learning_rate": 3.678226552561582e-05, "loss": 0.9667, "step": 176750 }, { "epoch": 15.31, "learning_rate": 3.678009714351798e-05, "loss": 0.9391, "step": 176760 }, { "epoch": 15.31, "learning_rate": 3.6777928761420145e-05, "loss": 0.8908, "step": 176770 }, { "epoch": 15.31, "learning_rate": 3.6775760379322304e-05, "loss": 0.8455, "step": 176780 }, { "epoch": 15.31, "learning_rate": 3.677359199722447e-05, "loss": 0.9951, "step": 176790 }, { "epoch": 15.31, "learning_rate": 3.677142361512663e-05, "loss": 0.7987, "step": 176800 }, { "epoch": 15.31, "learning_rate": 3.676925523302879e-05, "loss": 0.85, "step": 176810 }, { "epoch": 15.31, "learning_rate": 3.6767086850930956e-05, "loss": 0.8341, "step": 176820 }, { "epoch": 15.31, "learning_rate": 3.6764918468833116e-05, "loss": 0.9921, "step": 176830 }, { "epoch": 15.32, "learning_rate": 3.676275008673528e-05, "loss": 1.0213, "step": 176840 }, { "epoch": 15.32, "learning_rate": 3.676058170463744e-05, "loss": 0.9414, "step": 176850 }, { "epoch": 15.32, "learning_rate": 3.675841332253961e-05, "loss": 0.8373, "step": 176860 }, { "epoch": 15.32, "learning_rate": 3.675624494044177e-05, "loss": 1.0811, "step": 176870 }, { "epoch": 15.32, "learning_rate": 3.6754076558343934e-05, "loss": 0.916, "step": 176880 }, { "epoch": 15.32, "learning_rate": 3.6751908176246094e-05, "loss": 0.8531, "step": 176890 }, { "epoch": 15.32, "learning_rate": 3.6749739794148254e-05, "loss": 0.874, "step": 176900 }, { "epoch": 15.32, "learning_rate": 3.674757141205042e-05, "loss": 1.1117, "step": 176910 }, { "epoch": 15.32, "learning_rate": 3.6745403029952586e-05, "loss": 0.8988, "step": 176920 }, { "epoch": 15.32, "learning_rate": 3.674323464785474e-05, "loss": 0.9427, "step": 176930 }, { "epoch": 15.32, "learning_rate": 3.6741066265756906e-05, "loss": 0.9603, "step": 176940 }, { "epoch": 15.32, "learning_rate": 3.673889788365907e-05, "loss": 0.9455, "step": 176950 }, { "epoch": 15.33, "learning_rate": 3.673672950156123e-05, "loss": 1.0826, "step": 176960 }, { "epoch": 15.33, "learning_rate": 3.673456111946339e-05, "loss": 0.9993, "step": 176970 }, { "epoch": 15.33, "learning_rate": 3.673239273736556e-05, "loss": 0.9326, "step": 176980 }, { "epoch": 15.33, "learning_rate": 3.673022435526772e-05, "loss": 0.9445, "step": 176990 }, { "epoch": 15.33, "learning_rate": 3.6728055973169884e-05, "loss": 0.9441, "step": 177000 }, { "epoch": 15.33, "learning_rate": 3.672588759107204e-05, "loss": 0.9777, "step": 177010 }, { "epoch": 15.33, "learning_rate": 3.672371920897421e-05, "loss": 0.9891, "step": 177020 }, { "epoch": 15.33, "learning_rate": 3.672155082687637e-05, "loss": 0.7955, "step": 177030 }, { "epoch": 15.33, "learning_rate": 3.671938244477853e-05, "loss": 0.8275, "step": 177040 }, { "epoch": 15.33, "learning_rate": 3.6717214062680695e-05, "loss": 0.9204, "step": 177050 }, { "epoch": 15.33, "learning_rate": 3.6715045680582855e-05, "loss": 0.8419, "step": 177060 }, { "epoch": 15.34, "learning_rate": 3.671287729848502e-05, "loss": 0.9113, "step": 177070 }, { "epoch": 15.34, "learning_rate": 3.671070891638718e-05, "loss": 0.7647, "step": 177080 }, { "epoch": 15.34, "learning_rate": 3.670854053428935e-05, "loss": 1.0223, "step": 177090 }, { "epoch": 15.34, "learning_rate": 3.670637215219151e-05, "loss": 0.8571, "step": 177100 }, { "epoch": 15.34, "learning_rate": 3.670420377009367e-05, "loss": 0.9534, "step": 177110 }, { "epoch": 15.34, "learning_rate": 3.670203538799583e-05, "loss": 0.9973, "step": 177120 }, { "epoch": 15.34, "learning_rate": 3.669986700589799e-05, "loss": 0.8916, "step": 177130 }, { "epoch": 15.34, "learning_rate": 3.669769862380016e-05, "loss": 0.8117, "step": 177140 }, { "epoch": 15.34, "learning_rate": 3.6695530241702325e-05, "loss": 0.9209, "step": 177150 }, { "epoch": 15.34, "learning_rate": 3.6693361859604485e-05, "loss": 0.9874, "step": 177160 }, { "epoch": 15.34, "learning_rate": 3.6691193477506645e-05, "loss": 0.888, "step": 177170 }, { "epoch": 15.34, "learning_rate": 3.668902509540881e-05, "loss": 0.8758, "step": 177180 }, { "epoch": 15.35, "learning_rate": 3.668685671331097e-05, "loss": 1.0097, "step": 177190 }, { "epoch": 15.35, "learning_rate": 3.668468833121313e-05, "loss": 0.9077, "step": 177200 }, { "epoch": 15.35, "learning_rate": 3.6682519949115297e-05, "loss": 1.0377, "step": 177210 }, { "epoch": 15.35, "learning_rate": 3.668035156701746e-05, "loss": 1.0158, "step": 177220 }, { "epoch": 15.35, "learning_rate": 3.667818318491962e-05, "loss": 0.9154, "step": 177230 }, { "epoch": 15.35, "learning_rate": 3.667601480282178e-05, "loss": 0.92, "step": 177240 }, { "epoch": 15.35, "learning_rate": 3.667384642072395e-05, "loss": 0.9293, "step": 177250 }, { "epoch": 15.35, "learning_rate": 3.667167803862611e-05, "loss": 0.9502, "step": 177260 }, { "epoch": 15.35, "learning_rate": 3.6669509656528275e-05, "loss": 0.7939, "step": 177270 }, { "epoch": 15.35, "learning_rate": 3.6667341274430434e-05, "loss": 0.9659, "step": 177280 }, { "epoch": 15.35, "learning_rate": 3.66651728923326e-05, "loss": 0.9711, "step": 177290 }, { "epoch": 15.36, "learning_rate": 3.666300451023476e-05, "loss": 0.9967, "step": 177300 }, { "epoch": 15.36, "learning_rate": 3.666083612813692e-05, "loss": 0.7666, "step": 177310 }, { "epoch": 15.36, "learning_rate": 3.6658667746039086e-05, "loss": 0.8828, "step": 177320 }, { "epoch": 15.36, "learning_rate": 3.6656499363941246e-05, "loss": 0.9504, "step": 177330 }, { "epoch": 15.36, "learning_rate": 3.665433098184341e-05, "loss": 0.8982, "step": 177340 }, { "epoch": 15.36, "learning_rate": 3.665216259974557e-05, "loss": 0.9751, "step": 177350 }, { "epoch": 15.36, "learning_rate": 3.664999421764774e-05, "loss": 0.8559, "step": 177360 }, { "epoch": 15.36, "learning_rate": 3.66478258355499e-05, "loss": 0.9288, "step": 177370 }, { "epoch": 15.36, "learning_rate": 3.6645657453452064e-05, "loss": 0.9846, "step": 177380 }, { "epoch": 15.36, "learning_rate": 3.6643489071354224e-05, "loss": 0.7568, "step": 177390 }, { "epoch": 15.36, "learning_rate": 3.6641320689256383e-05, "loss": 0.9296, "step": 177400 }, { "epoch": 15.36, "learning_rate": 3.663915230715855e-05, "loss": 0.743, "step": 177410 }, { "epoch": 15.37, "learning_rate": 3.6636983925060716e-05, "loss": 0.9039, "step": 177420 }, { "epoch": 15.37, "learning_rate": 3.6634815542962876e-05, "loss": 0.7817, "step": 177430 }, { "epoch": 15.37, "learning_rate": 3.6632647160865035e-05, "loss": 1.0337, "step": 177440 }, { "epoch": 15.37, "learning_rate": 3.66304787787672e-05, "loss": 0.8865, "step": 177450 }, { "epoch": 15.37, "learning_rate": 3.662831039666936e-05, "loss": 1.0975, "step": 177460 }, { "epoch": 15.37, "learning_rate": 3.662614201457152e-05, "loss": 0.8943, "step": 177470 }, { "epoch": 15.37, "learning_rate": 3.662397363247369e-05, "loss": 0.7826, "step": 177480 }, { "epoch": 15.37, "learning_rate": 3.662180525037585e-05, "loss": 0.8894, "step": 177490 }, { "epoch": 15.37, "learning_rate": 3.6619636868278014e-05, "loss": 0.9274, "step": 177500 }, { "epoch": 15.37, "learning_rate": 3.661746848618017e-05, "loss": 0.9609, "step": 177510 }, { "epoch": 15.37, "learning_rate": 3.661530010408234e-05, "loss": 0.9038, "step": 177520 }, { "epoch": 15.38, "learning_rate": 3.66131317219845e-05, "loss": 0.864, "step": 177530 }, { "epoch": 15.38, "learning_rate": 3.6610963339886666e-05, "loss": 0.8574, "step": 177540 }, { "epoch": 15.38, "learning_rate": 3.6608794957788825e-05, "loss": 1.0494, "step": 177550 }, { "epoch": 15.38, "learning_rate": 3.6606626575690985e-05, "loss": 0.8513, "step": 177560 }, { "epoch": 15.38, "learning_rate": 3.660445819359315e-05, "loss": 0.8543, "step": 177570 }, { "epoch": 15.38, "learning_rate": 3.660228981149531e-05, "loss": 1.0154, "step": 177580 }, { "epoch": 15.38, "learning_rate": 3.660012142939748e-05, "loss": 0.8976, "step": 177590 }, { "epoch": 15.38, "learning_rate": 3.659795304729964e-05, "loss": 1.0361, "step": 177600 }, { "epoch": 15.38, "learning_rate": 3.65957846652018e-05, "loss": 0.973, "step": 177610 }, { "epoch": 15.38, "learning_rate": 3.659361628310396e-05, "loss": 1.0651, "step": 177620 }, { "epoch": 15.38, "learning_rate": 3.659144790100612e-05, "loss": 0.8292, "step": 177630 }, { "epoch": 15.38, "learning_rate": 3.658927951890829e-05, "loss": 0.6147, "step": 177640 }, { "epoch": 15.39, "learning_rate": 3.6587111136810455e-05, "loss": 1.0627, "step": 177650 }, { "epoch": 15.39, "learning_rate": 3.6584942754712615e-05, "loss": 0.7865, "step": 177660 }, { "epoch": 15.39, "learning_rate": 3.6582774372614774e-05, "loss": 0.911, "step": 177670 }, { "epoch": 15.39, "learning_rate": 3.658060599051694e-05, "loss": 1.031, "step": 177680 }, { "epoch": 15.39, "learning_rate": 3.65784376084191e-05, "loss": 0.9048, "step": 177690 }, { "epoch": 15.39, "learning_rate": 3.657626922632126e-05, "loss": 0.9893, "step": 177700 }, { "epoch": 15.39, "learning_rate": 3.6574100844223426e-05, "loss": 0.8791, "step": 177710 }, { "epoch": 15.39, "learning_rate": 3.657193246212559e-05, "loss": 0.9865, "step": 177720 }, { "epoch": 15.39, "learning_rate": 3.656976408002775e-05, "loss": 0.9267, "step": 177730 }, { "epoch": 15.39, "learning_rate": 3.656759569792991e-05, "loss": 0.879, "step": 177740 }, { "epoch": 15.39, "learning_rate": 3.656542731583208e-05, "loss": 1.0006, "step": 177750 }, { "epoch": 15.4, "learning_rate": 3.656325893373424e-05, "loss": 0.9666, "step": 177760 }, { "epoch": 15.4, "learning_rate": 3.6561090551636404e-05, "loss": 0.7563, "step": 177770 }, { "epoch": 15.4, "learning_rate": 3.6558922169538564e-05, "loss": 0.9169, "step": 177780 }, { "epoch": 15.4, "learning_rate": 3.655675378744073e-05, "loss": 0.7534, "step": 177790 }, { "epoch": 15.4, "learning_rate": 3.655458540534289e-05, "loss": 0.9625, "step": 177800 }, { "epoch": 15.4, "learning_rate": 3.655241702324505e-05, "loss": 0.9423, "step": 177810 }, { "epoch": 15.4, "learning_rate": 3.6550248641147216e-05, "loss": 0.9801, "step": 177820 }, { "epoch": 15.4, "learning_rate": 3.6548080259049376e-05, "loss": 0.8567, "step": 177830 }, { "epoch": 15.4, "learning_rate": 3.654591187695154e-05, "loss": 0.9642, "step": 177840 }, { "epoch": 15.4, "learning_rate": 3.65437434948537e-05, "loss": 0.8686, "step": 177850 }, { "epoch": 15.4, "learning_rate": 3.654157511275587e-05, "loss": 0.9463, "step": 177860 }, { "epoch": 15.4, "learning_rate": 3.653940673065803e-05, "loss": 0.8759, "step": 177870 }, { "epoch": 15.41, "learning_rate": 3.6537238348560194e-05, "loss": 0.9103, "step": 177880 }, { "epoch": 15.41, "learning_rate": 3.6535069966462354e-05, "loss": 0.9815, "step": 177890 }, { "epoch": 15.41, "learning_rate": 3.653290158436451e-05, "loss": 0.9005, "step": 177900 }, { "epoch": 15.41, "learning_rate": 3.653073320226668e-05, "loss": 0.8742, "step": 177910 }, { "epoch": 15.41, "learning_rate": 3.6528564820168846e-05, "loss": 0.9604, "step": 177920 }, { "epoch": 15.41, "learning_rate": 3.6526396438071006e-05, "loss": 0.8617, "step": 177930 }, { "epoch": 15.41, "learning_rate": 3.6524228055973165e-05, "loss": 0.8937, "step": 177940 }, { "epoch": 15.41, "learning_rate": 3.652205967387533e-05, "loss": 0.9424, "step": 177950 }, { "epoch": 15.41, "learning_rate": 3.651989129177749e-05, "loss": 0.9913, "step": 177960 }, { "epoch": 15.41, "learning_rate": 3.651772290967965e-05, "loss": 0.7223, "step": 177970 }, { "epoch": 15.41, "learning_rate": 3.651555452758182e-05, "loss": 1.0463, "step": 177980 }, { "epoch": 15.42, "learning_rate": 3.6513386145483984e-05, "loss": 0.934, "step": 177990 }, { "epoch": 15.42, "learning_rate": 3.651121776338614e-05, "loss": 1.0545, "step": 178000 }, { "epoch": 15.42, "learning_rate": 3.65090493812883e-05, "loss": 0.8758, "step": 178010 }, { "epoch": 15.42, "learning_rate": 3.650688099919047e-05, "loss": 1.0325, "step": 178020 }, { "epoch": 15.42, "learning_rate": 3.650471261709263e-05, "loss": 0.9545, "step": 178030 }, { "epoch": 15.42, "learning_rate": 3.6502544234994795e-05, "loss": 0.8908, "step": 178040 }, { "epoch": 15.42, "learning_rate": 3.6500375852896955e-05, "loss": 0.9554, "step": 178050 }, { "epoch": 15.42, "learning_rate": 3.6498207470799115e-05, "loss": 0.8581, "step": 178060 }, { "epoch": 15.42, "learning_rate": 3.649603908870128e-05, "loss": 0.9432, "step": 178070 }, { "epoch": 15.42, "learning_rate": 3.649387070660344e-05, "loss": 0.9118, "step": 178080 }, { "epoch": 15.42, "learning_rate": 3.649170232450561e-05, "loss": 0.9122, "step": 178090 }, { "epoch": 15.42, "learning_rate": 3.648953394240777e-05, "loss": 0.8905, "step": 178100 }, { "epoch": 15.43, "learning_rate": 3.648736556030993e-05, "loss": 0.9409, "step": 178110 }, { "epoch": 15.43, "learning_rate": 3.648519717821209e-05, "loss": 0.9631, "step": 178120 }, { "epoch": 15.43, "learning_rate": 3.648302879611425e-05, "loss": 0.8733, "step": 178130 }, { "epoch": 15.43, "learning_rate": 3.648086041401642e-05, "loss": 1.05, "step": 178140 }, { "epoch": 15.43, "learning_rate": 3.6478692031918585e-05, "loss": 0.8507, "step": 178150 }, { "epoch": 15.43, "learning_rate": 3.6476523649820745e-05, "loss": 0.8196, "step": 178160 }, { "epoch": 15.43, "learning_rate": 3.6474355267722904e-05, "loss": 0.8301, "step": 178170 }, { "epoch": 15.43, "learning_rate": 3.647218688562507e-05, "loss": 0.9321, "step": 178180 }, { "epoch": 15.43, "learning_rate": 3.647001850352723e-05, "loss": 0.888, "step": 178190 }, { "epoch": 15.43, "learning_rate": 3.646785012142939e-05, "loss": 0.774, "step": 178200 }, { "epoch": 15.43, "learning_rate": 3.6465681739331556e-05, "loss": 0.9837, "step": 178210 }, { "epoch": 15.43, "learning_rate": 3.646351335723372e-05, "loss": 0.9506, "step": 178220 }, { "epoch": 15.44, "learning_rate": 3.646134497513588e-05, "loss": 0.9522, "step": 178230 }, { "epoch": 15.44, "learning_rate": 3.645917659303804e-05, "loss": 0.7734, "step": 178240 }, { "epoch": 15.44, "learning_rate": 3.645700821094021e-05, "loss": 0.8853, "step": 178250 }, { "epoch": 15.44, "learning_rate": 3.645483982884237e-05, "loss": 0.937, "step": 178260 }, { "epoch": 15.44, "learning_rate": 3.6452671446744534e-05, "loss": 1.0056, "step": 178270 }, { "epoch": 15.44, "learning_rate": 3.6450503064646694e-05, "loss": 1.0487, "step": 178280 }, { "epoch": 15.44, "learning_rate": 3.644833468254886e-05, "loss": 0.87, "step": 178290 }, { "epoch": 15.44, "learning_rate": 3.644616630045102e-05, "loss": 0.9989, "step": 178300 }, { "epoch": 15.44, "learning_rate": 3.6443997918353186e-05, "loss": 0.9236, "step": 178310 }, { "epoch": 15.44, "learning_rate": 3.6441829536255346e-05, "loss": 0.8722, "step": 178320 }, { "epoch": 15.44, "learning_rate": 3.6439661154157506e-05, "loss": 0.9484, "step": 178330 }, { "epoch": 15.45, "learning_rate": 3.643749277205967e-05, "loss": 0.9073, "step": 178340 }, { "epoch": 15.45, "learning_rate": 3.643532438996183e-05, "loss": 0.938, "step": 178350 }, { "epoch": 15.45, "learning_rate": 3.6433156007864e-05, "loss": 0.834, "step": 178360 }, { "epoch": 15.45, "learning_rate": 3.643098762576616e-05, "loss": 1.1612, "step": 178370 }, { "epoch": 15.45, "learning_rate": 3.6428819243668324e-05, "loss": 1.0599, "step": 178380 }, { "epoch": 15.45, "learning_rate": 3.6426650861570484e-05, "loss": 0.8723, "step": 178390 }, { "epoch": 15.45, "learning_rate": 3.642448247947264e-05, "loss": 0.9603, "step": 178400 }, { "epoch": 15.45, "learning_rate": 3.642231409737481e-05, "loss": 1.0759, "step": 178410 }, { "epoch": 15.45, "learning_rate": 3.6420145715276976e-05, "loss": 1.0284, "step": 178420 }, { "epoch": 15.45, "learning_rate": 3.6417977333179136e-05, "loss": 0.9449, "step": 178430 }, { "epoch": 15.45, "learning_rate": 3.6415808951081295e-05, "loss": 0.7497, "step": 178440 }, { "epoch": 15.45, "learning_rate": 3.641364056898346e-05, "loss": 0.8584, "step": 178450 }, { "epoch": 15.46, "learning_rate": 3.641147218688562e-05, "loss": 1.0087, "step": 178460 }, { "epoch": 15.46, "learning_rate": 3.640930380478778e-05, "loss": 0.9728, "step": 178470 }, { "epoch": 15.46, "learning_rate": 3.640713542268995e-05, "loss": 0.8612, "step": 178480 }, { "epoch": 15.46, "learning_rate": 3.6404967040592114e-05, "loss": 1.0396, "step": 178490 }, { "epoch": 15.46, "learning_rate": 3.640279865849427e-05, "loss": 0.7952, "step": 178500 }, { "epoch": 15.46, "learning_rate": 3.640063027639643e-05, "loss": 1.0818, "step": 178510 }, { "epoch": 15.46, "learning_rate": 3.63984618942986e-05, "loss": 0.986, "step": 178520 }, { "epoch": 15.46, "learning_rate": 3.639629351220076e-05, "loss": 0.929, "step": 178530 }, { "epoch": 15.46, "learning_rate": 3.6394125130102925e-05, "loss": 0.8278, "step": 178540 }, { "epoch": 15.46, "learning_rate": 3.6391956748005085e-05, "loss": 0.9586, "step": 178550 }, { "epoch": 15.46, "learning_rate": 3.6389788365907244e-05, "loss": 0.9567, "step": 178560 }, { "epoch": 15.47, "learning_rate": 3.638761998380941e-05, "loss": 0.8667, "step": 178570 }, { "epoch": 15.47, "learning_rate": 3.638545160171157e-05, "loss": 0.9314, "step": 178580 }, { "epoch": 15.47, "learning_rate": 3.638328321961374e-05, "loss": 0.941, "step": 178590 }, { "epoch": 15.47, "learning_rate": 3.6381114837515896e-05, "loss": 0.9229, "step": 178600 }, { "epoch": 15.47, "learning_rate": 3.637894645541806e-05, "loss": 0.7562, "step": 178610 }, { "epoch": 15.47, "learning_rate": 3.637677807332022e-05, "loss": 0.9155, "step": 178620 }, { "epoch": 15.47, "learning_rate": 3.637460969122238e-05, "loss": 1.1095, "step": 178630 }, { "epoch": 15.47, "learning_rate": 3.637244130912455e-05, "loss": 0.7329, "step": 178640 }, { "epoch": 15.47, "learning_rate": 3.6370272927026715e-05, "loss": 0.7882, "step": 178650 }, { "epoch": 15.47, "learning_rate": 3.6368104544928874e-05, "loss": 1.0101, "step": 178660 }, { "epoch": 15.47, "learning_rate": 3.6365936162831034e-05, "loss": 0.9933, "step": 178670 }, { "epoch": 15.47, "learning_rate": 3.63637677807332e-05, "loss": 0.7301, "step": 178680 }, { "epoch": 15.48, "learning_rate": 3.636159939863536e-05, "loss": 0.8747, "step": 178690 }, { "epoch": 15.48, "learning_rate": 3.635943101653752e-05, "loss": 0.9625, "step": 178700 }, { "epoch": 15.48, "learning_rate": 3.6357262634439686e-05, "loss": 1.0526, "step": 178710 }, { "epoch": 15.48, "learning_rate": 3.635509425234185e-05, "loss": 0.952, "step": 178720 }, { "epoch": 15.48, "learning_rate": 3.635292587024401e-05, "loss": 1.0276, "step": 178730 }, { "epoch": 15.48, "learning_rate": 3.635075748814617e-05, "loss": 1.0934, "step": 178740 }, { "epoch": 15.48, "learning_rate": 3.634858910604834e-05, "loss": 1.0167, "step": 178750 }, { "epoch": 15.48, "learning_rate": 3.63464207239505e-05, "loss": 0.7424, "step": 178760 }, { "epoch": 15.48, "learning_rate": 3.6344252341852664e-05, "loss": 0.7695, "step": 178770 }, { "epoch": 15.48, "learning_rate": 3.6342083959754824e-05, "loss": 0.9611, "step": 178780 }, { "epoch": 15.48, "learning_rate": 3.633991557765699e-05, "loss": 0.8846, "step": 178790 }, { "epoch": 15.49, "learning_rate": 3.633774719555915e-05, "loss": 0.8908, "step": 178800 }, { "epoch": 15.49, "learning_rate": 3.6335578813461316e-05, "loss": 1.0209, "step": 178810 }, { "epoch": 15.49, "learning_rate": 3.6333410431363476e-05, "loss": 0.8807, "step": 178820 }, { "epoch": 15.49, "learning_rate": 3.6331242049265635e-05, "loss": 0.8563, "step": 178830 }, { "epoch": 15.49, "learning_rate": 3.63290736671678e-05, "loss": 0.784, "step": 178840 }, { "epoch": 15.49, "learning_rate": 3.632690528506996e-05, "loss": 0.902, "step": 178850 }, { "epoch": 15.49, "learning_rate": 3.632473690297213e-05, "loss": 1.0125, "step": 178860 }, { "epoch": 15.49, "learning_rate": 3.632256852087429e-05, "loss": 0.8423, "step": 178870 }, { "epoch": 15.49, "learning_rate": 3.6320400138776454e-05, "loss": 0.8996, "step": 178880 }, { "epoch": 15.49, "learning_rate": 3.6318231756678613e-05, "loss": 1.0198, "step": 178890 }, { "epoch": 15.49, "learning_rate": 3.631606337458077e-05, "loss": 0.8236, "step": 178900 }, { "epoch": 15.49, "learning_rate": 3.631389499248294e-05, "loss": 0.9008, "step": 178910 }, { "epoch": 15.5, "learning_rate": 3.6311726610385106e-05, "loss": 0.9232, "step": 178920 }, { "epoch": 15.5, "learning_rate": 3.6309558228287265e-05, "loss": 1.0612, "step": 178930 }, { "epoch": 15.5, "learning_rate": 3.6307389846189425e-05, "loss": 0.9953, "step": 178940 }, { "epoch": 15.5, "learning_rate": 3.630522146409159e-05, "loss": 0.6921, "step": 178950 }, { "epoch": 15.5, "learning_rate": 3.630305308199375e-05, "loss": 1.1126, "step": 178960 }, { "epoch": 15.5, "learning_rate": 3.630088469989591e-05, "loss": 1.0136, "step": 178970 }, { "epoch": 15.5, "learning_rate": 3.629871631779808e-05, "loss": 0.866, "step": 178980 }, { "epoch": 15.5, "learning_rate": 3.6296547935700243e-05, "loss": 1.0312, "step": 178990 }, { "epoch": 15.5, "learning_rate": 3.62943795536024e-05, "loss": 0.8716, "step": 179000 }, { "epoch": 15.5, "learning_rate": 3.629221117150456e-05, "loss": 0.8806, "step": 179010 }, { "epoch": 15.5, "learning_rate": 3.629004278940673e-05, "loss": 0.7227, "step": 179020 }, { "epoch": 15.51, "learning_rate": 3.628787440730889e-05, "loss": 0.9464, "step": 179030 }, { "epoch": 15.51, "learning_rate": 3.6285706025211055e-05, "loss": 0.8904, "step": 179040 }, { "epoch": 15.51, "learning_rate": 3.6283537643113215e-05, "loss": 0.9266, "step": 179050 }, { "epoch": 15.51, "learning_rate": 3.628136926101538e-05, "loss": 0.9747, "step": 179060 }, { "epoch": 15.51, "learning_rate": 3.627920087891754e-05, "loss": 0.9739, "step": 179070 }, { "epoch": 15.51, "learning_rate": 3.627703249681971e-05, "loss": 0.9915, "step": 179080 }, { "epoch": 15.51, "learning_rate": 3.627486411472187e-05, "loss": 0.8563, "step": 179090 }, { "epoch": 15.51, "learning_rate": 3.6272695732624026e-05, "loss": 0.9104, "step": 179100 }, { "epoch": 15.51, "learning_rate": 3.627052735052619e-05, "loss": 0.9873, "step": 179110 }, { "epoch": 15.51, "learning_rate": 3.626835896842835e-05, "loss": 0.9243, "step": 179120 }, { "epoch": 15.51, "learning_rate": 3.626619058633051e-05, "loss": 0.9995, "step": 179130 }, { "epoch": 15.51, "learning_rate": 3.626402220423268e-05, "loss": 1.0442, "step": 179140 }, { "epoch": 15.52, "learning_rate": 3.6261853822134845e-05, "loss": 1.1083, "step": 179150 }, { "epoch": 15.52, "learning_rate": 3.6259685440037004e-05, "loss": 0.8933, "step": 179160 }, { "epoch": 15.52, "learning_rate": 3.6257517057939164e-05, "loss": 0.9813, "step": 179170 }, { "epoch": 15.52, "learning_rate": 3.625534867584133e-05, "loss": 0.8544, "step": 179180 }, { "epoch": 15.52, "learning_rate": 3.625318029374349e-05, "loss": 0.9981, "step": 179190 }, { "epoch": 15.52, "learning_rate": 3.625101191164565e-05, "loss": 1.0483, "step": 179200 }, { "epoch": 15.52, "learning_rate": 3.6248843529547816e-05, "loss": 0.8792, "step": 179210 }, { "epoch": 15.52, "learning_rate": 3.624667514744998e-05, "loss": 0.9875, "step": 179220 }, { "epoch": 15.52, "learning_rate": 3.624450676535214e-05, "loss": 1.0613, "step": 179230 }, { "epoch": 15.52, "learning_rate": 3.62423383832543e-05, "loss": 0.8807, "step": 179240 }, { "epoch": 15.52, "learning_rate": 3.624017000115647e-05, "loss": 0.8887, "step": 179250 }, { "epoch": 15.53, "learning_rate": 3.623800161905863e-05, "loss": 0.743, "step": 179260 }, { "epoch": 15.53, "learning_rate": 3.6235833236960794e-05, "loss": 1.025, "step": 179270 }, { "epoch": 15.53, "learning_rate": 3.6233664854862954e-05, "loss": 0.822, "step": 179280 }, { "epoch": 15.53, "learning_rate": 3.623149647276512e-05, "loss": 0.8778, "step": 179290 }, { "epoch": 15.53, "learning_rate": 3.622932809066728e-05, "loss": 0.8366, "step": 179300 }, { "epoch": 15.53, "learning_rate": 3.6227159708569446e-05, "loss": 1.0277, "step": 179310 }, { "epoch": 15.53, "learning_rate": 3.6224991326471606e-05, "loss": 0.8187, "step": 179320 }, { "epoch": 15.53, "learning_rate": 3.6222822944373765e-05, "loss": 0.9777, "step": 179330 }, { "epoch": 15.53, "learning_rate": 3.622065456227593e-05, "loss": 0.7191, "step": 179340 }, { "epoch": 15.53, "learning_rate": 3.621848618017809e-05, "loss": 0.8027, "step": 179350 }, { "epoch": 15.53, "learning_rate": 3.621631779808026e-05, "loss": 0.775, "step": 179360 }, { "epoch": 15.53, "learning_rate": 3.621414941598242e-05, "loss": 0.8073, "step": 179370 }, { "epoch": 15.54, "learning_rate": 3.6211981033884584e-05, "loss": 1.0134, "step": 179380 }, { "epoch": 15.54, "learning_rate": 3.620981265178674e-05, "loss": 0.8825, "step": 179390 }, { "epoch": 15.54, "learning_rate": 3.62076442696889e-05, "loss": 0.862, "step": 179400 }, { "epoch": 15.54, "learning_rate": 3.620547588759107e-05, "loss": 1.0021, "step": 179410 }, { "epoch": 15.54, "learning_rate": 3.6203307505493236e-05, "loss": 0.9364, "step": 179420 }, { "epoch": 15.54, "learning_rate": 3.6201139123395395e-05, "loss": 0.7155, "step": 179430 }, { "epoch": 15.54, "learning_rate": 3.6198970741297555e-05, "loss": 0.9033, "step": 179440 }, { "epoch": 15.54, "learning_rate": 3.619680235919972e-05, "loss": 1.0227, "step": 179450 }, { "epoch": 15.54, "learning_rate": 3.619463397710188e-05, "loss": 0.9973, "step": 179460 }, { "epoch": 15.54, "learning_rate": 3.619246559500404e-05, "loss": 0.9556, "step": 179470 }, { "epoch": 15.54, "learning_rate": 3.619029721290621e-05, "loss": 0.7382, "step": 179480 }, { "epoch": 15.54, "learning_rate": 3.618812883080837e-05, "loss": 0.9476, "step": 179490 }, { "epoch": 15.55, "learning_rate": 3.618596044871053e-05, "loss": 1.0767, "step": 179500 }, { "epoch": 15.55, "learning_rate": 3.618379206661269e-05, "loss": 0.7666, "step": 179510 }, { "epoch": 15.55, "learning_rate": 3.618162368451486e-05, "loss": 0.895, "step": 179520 }, { "epoch": 15.55, "learning_rate": 3.617945530241702e-05, "loss": 0.8346, "step": 179530 }, { "epoch": 15.55, "learning_rate": 3.6177286920319185e-05, "loss": 0.8654, "step": 179540 }, { "epoch": 15.55, "learning_rate": 3.6175118538221345e-05, "loss": 1.0744, "step": 179550 }, { "epoch": 15.55, "learning_rate": 3.617295015612351e-05, "loss": 0.946, "step": 179560 }, { "epoch": 15.55, "learning_rate": 3.617078177402567e-05, "loss": 0.9018, "step": 179570 }, { "epoch": 15.55, "learning_rate": 3.616861339192784e-05, "loss": 0.8331, "step": 179580 }, { "epoch": 15.55, "learning_rate": 3.6166445009829997e-05, "loss": 0.9845, "step": 179590 }, { "epoch": 15.55, "learning_rate": 3.6164276627732156e-05, "loss": 0.6966, "step": 179600 }, { "epoch": 15.56, "learning_rate": 3.616210824563432e-05, "loss": 1.0119, "step": 179610 }, { "epoch": 15.56, "learning_rate": 3.615993986353648e-05, "loss": 0.9266, "step": 179620 }, { "epoch": 15.56, "learning_rate": 3.615777148143864e-05, "loss": 0.9377, "step": 179630 }, { "epoch": 15.56, "learning_rate": 3.615560309934081e-05, "loss": 0.806, "step": 179640 }, { "epoch": 15.56, "learning_rate": 3.6153434717242975e-05, "loss": 0.8253, "step": 179650 }, { "epoch": 15.56, "learning_rate": 3.6151266335145134e-05, "loss": 0.8977, "step": 179660 }, { "epoch": 15.56, "learning_rate": 3.6149097953047294e-05, "loss": 1.1483, "step": 179670 }, { "epoch": 15.56, "learning_rate": 3.614692957094946e-05, "loss": 0.9163, "step": 179680 }, { "epoch": 15.56, "learning_rate": 3.614476118885162e-05, "loss": 0.8006, "step": 179690 }, { "epoch": 15.56, "learning_rate": 3.614259280675378e-05, "loss": 0.8951, "step": 179700 }, { "epoch": 15.56, "learning_rate": 3.6140424424655946e-05, "loss": 0.9709, "step": 179710 }, { "epoch": 15.56, "learning_rate": 3.613825604255811e-05, "loss": 1.0113, "step": 179720 }, { "epoch": 15.57, "learning_rate": 3.613608766046027e-05, "loss": 0.9465, "step": 179730 }, { "epoch": 15.57, "learning_rate": 3.613391927836243e-05, "loss": 0.7849, "step": 179740 }, { "epoch": 15.57, "learning_rate": 3.61317508962646e-05, "loss": 0.9657, "step": 179750 }, { "epoch": 15.57, "learning_rate": 3.612958251416676e-05, "loss": 0.954, "step": 179760 }, { "epoch": 15.57, "learning_rate": 3.6127414132068924e-05, "loss": 0.822, "step": 179770 }, { "epoch": 15.57, "learning_rate": 3.6125245749971083e-05, "loss": 1.0335, "step": 179780 }, { "epoch": 15.57, "learning_rate": 3.612307736787325e-05, "loss": 0.8928, "step": 179790 }, { "epoch": 15.57, "learning_rate": 3.612090898577541e-05, "loss": 0.8921, "step": 179800 }, { "epoch": 15.57, "learning_rate": 3.6118740603677576e-05, "loss": 0.88, "step": 179810 }, { "epoch": 15.57, "learning_rate": 3.6116572221579735e-05, "loss": 1.0934, "step": 179820 }, { "epoch": 15.57, "learning_rate": 3.6114403839481895e-05, "loss": 0.7948, "step": 179830 }, { "epoch": 15.58, "learning_rate": 3.611223545738406e-05, "loss": 1.0471, "step": 179840 }, { "epoch": 15.58, "learning_rate": 3.611006707528623e-05, "loss": 0.9364, "step": 179850 }, { "epoch": 15.58, "learning_rate": 3.610789869318839e-05, "loss": 0.8282, "step": 179860 }, { "epoch": 15.58, "learning_rate": 3.610573031109055e-05, "loss": 0.9382, "step": 179870 }, { "epoch": 15.58, "learning_rate": 3.6103561928992713e-05, "loss": 1.0182, "step": 179880 }, { "epoch": 15.58, "learning_rate": 3.610139354689487e-05, "loss": 0.9791, "step": 179890 }, { "epoch": 15.58, "learning_rate": 3.609922516479703e-05, "loss": 0.8136, "step": 179900 }, { "epoch": 15.58, "learning_rate": 3.60970567826992e-05, "loss": 0.9759, "step": 179910 }, { "epoch": 15.58, "learning_rate": 3.6094888400601365e-05, "loss": 0.9215, "step": 179920 }, { "epoch": 15.58, "learning_rate": 3.6092720018503525e-05, "loss": 0.8739, "step": 179930 }, { "epoch": 15.58, "learning_rate": 3.6090551636405685e-05, "loss": 0.8338, "step": 179940 }, { "epoch": 15.58, "learning_rate": 3.608838325430785e-05, "loss": 1.0376, "step": 179950 }, { "epoch": 15.59, "learning_rate": 3.608621487221001e-05, "loss": 0.9631, "step": 179960 }, { "epoch": 15.59, "learning_rate": 3.608404649011217e-05, "loss": 0.9167, "step": 179970 }, { "epoch": 15.59, "learning_rate": 3.608187810801434e-05, "loss": 1.0145, "step": 179980 }, { "epoch": 15.59, "learning_rate": 3.60797097259165e-05, "loss": 1.1815, "step": 179990 }, { "epoch": 15.59, "learning_rate": 3.607754134381866e-05, "loss": 0.8963, "step": 180000 }, { "epoch": 15.59, "learning_rate": 3.607537296172082e-05, "loss": 0.868, "step": 180010 }, { "epoch": 15.59, "learning_rate": 3.607320457962299e-05, "loss": 0.754, "step": 180020 }, { "epoch": 15.59, "learning_rate": 3.607103619752515e-05, "loss": 0.9126, "step": 180030 }, { "epoch": 15.59, "learning_rate": 3.6068867815427315e-05, "loss": 0.7503, "step": 180040 }, { "epoch": 15.59, "learning_rate": 3.6066699433329474e-05, "loss": 0.7721, "step": 180050 }, { "epoch": 15.59, "learning_rate": 3.606453105123164e-05, "loss": 0.9503, "step": 180060 }, { "epoch": 15.6, "learning_rate": 3.60623626691338e-05, "loss": 0.8388, "step": 180070 }, { "epoch": 15.6, "learning_rate": 3.606019428703597e-05, "loss": 0.9336, "step": 180080 }, { "epoch": 15.6, "learning_rate": 3.6058025904938126e-05, "loss": 1.0025, "step": 180090 }, { "epoch": 15.6, "learning_rate": 3.6055857522840286e-05, "loss": 0.8232, "step": 180100 }, { "epoch": 15.6, "learning_rate": 3.605368914074245e-05, "loss": 0.9716, "step": 180110 }, { "epoch": 15.6, "learning_rate": 3.605152075864461e-05, "loss": 0.9932, "step": 180120 }, { "epoch": 15.6, "learning_rate": 3.604935237654678e-05, "loss": 0.841, "step": 180130 }, { "epoch": 15.6, "learning_rate": 3.604718399444894e-05, "loss": 0.9505, "step": 180140 }, { "epoch": 15.6, "learning_rate": 3.6045015612351104e-05, "loss": 0.7402, "step": 180150 }, { "epoch": 15.6, "learning_rate": 3.6042847230253264e-05, "loss": 0.8228, "step": 180160 }, { "epoch": 15.6, "learning_rate": 3.6040678848155424e-05, "loss": 1.054, "step": 180170 }, { "epoch": 15.6, "learning_rate": 3.603851046605759e-05, "loss": 0.9425, "step": 180180 }, { "epoch": 15.61, "learning_rate": 3.603634208395975e-05, "loss": 0.8709, "step": 180190 }, { "epoch": 15.61, "learning_rate": 3.603417370186191e-05, "loss": 0.7827, "step": 180200 }, { "epoch": 15.61, "learning_rate": 3.6032005319764076e-05, "loss": 0.9054, "step": 180210 }, { "epoch": 15.61, "learning_rate": 3.602983693766624e-05, "loss": 0.6964, "step": 180220 }, { "epoch": 15.61, "learning_rate": 3.60276685555684e-05, "loss": 0.7388, "step": 180230 }, { "epoch": 15.61, "learning_rate": 3.602550017347056e-05, "loss": 0.8965, "step": 180240 }, { "epoch": 15.61, "learning_rate": 3.602333179137273e-05, "loss": 1.0328, "step": 180250 }, { "epoch": 15.61, "learning_rate": 3.602116340927489e-05, "loss": 0.944, "step": 180260 }, { "epoch": 15.61, "learning_rate": 3.6018995027177054e-05, "loss": 0.9605, "step": 180270 }, { "epoch": 15.61, "learning_rate": 3.601682664507921e-05, "loss": 0.9782, "step": 180280 }, { "epoch": 15.61, "learning_rate": 3.601465826298138e-05, "loss": 1.1575, "step": 180290 }, { "epoch": 15.62, "learning_rate": 3.601248988088354e-05, "loss": 0.9471, "step": 180300 }, { "epoch": 15.62, "learning_rate": 3.6010321498785706e-05, "loss": 0.8677, "step": 180310 }, { "epoch": 15.62, "learning_rate": 3.6008153116687865e-05, "loss": 0.9783, "step": 180320 }, { "epoch": 15.62, "learning_rate": 3.6005984734590025e-05, "loss": 0.8797, "step": 180330 }, { "epoch": 15.62, "learning_rate": 3.600381635249219e-05, "loss": 0.9906, "step": 180340 }, { "epoch": 15.62, "learning_rate": 3.600164797039436e-05, "loss": 0.925, "step": 180350 }, { "epoch": 15.62, "learning_rate": 3.599947958829652e-05, "loss": 1.0764, "step": 180360 }, { "epoch": 15.62, "learning_rate": 3.599731120619868e-05, "loss": 0.9643, "step": 180370 }, { "epoch": 15.62, "learning_rate": 3.599514282410084e-05, "loss": 0.9142, "step": 180380 }, { "epoch": 15.62, "learning_rate": 3.5992974442003e-05, "loss": 0.8938, "step": 180390 }, { "epoch": 15.62, "learning_rate": 3.599080605990516e-05, "loss": 0.911, "step": 180400 }, { "epoch": 15.62, "learning_rate": 3.598863767780733e-05, "loss": 0.883, "step": 180410 }, { "epoch": 15.63, "learning_rate": 3.5986469295709495e-05, "loss": 0.8693, "step": 180420 }, { "epoch": 15.63, "learning_rate": 3.5984300913611655e-05, "loss": 0.976, "step": 180430 }, { "epoch": 15.63, "learning_rate": 3.5982132531513815e-05, "loss": 0.7894, "step": 180440 }, { "epoch": 15.63, "learning_rate": 3.597996414941598e-05, "loss": 1.0332, "step": 180450 }, { "epoch": 15.63, "learning_rate": 3.597779576731814e-05, "loss": 0.9709, "step": 180460 }, { "epoch": 15.63, "learning_rate": 3.59756273852203e-05, "loss": 1.0516, "step": 180470 }, { "epoch": 15.63, "learning_rate": 3.5973459003122467e-05, "loss": 0.7444, "step": 180480 }, { "epoch": 15.63, "learning_rate": 3.597129062102463e-05, "loss": 0.8174, "step": 180490 }, { "epoch": 15.63, "learning_rate": 3.596912223892679e-05, "loss": 0.6841, "step": 180500 }, { "epoch": 15.63, "learning_rate": 3.596695385682895e-05, "loss": 1.0128, "step": 180510 }, { "epoch": 15.63, "learning_rate": 3.596478547473112e-05, "loss": 0.9405, "step": 180520 }, { "epoch": 15.64, "learning_rate": 3.596261709263328e-05, "loss": 0.8936, "step": 180530 }, { "epoch": 15.64, "learning_rate": 3.5960448710535445e-05, "loss": 0.9634, "step": 180540 }, { "epoch": 15.64, "learning_rate": 3.5958280328437604e-05, "loss": 0.9086, "step": 180550 }, { "epoch": 15.64, "learning_rate": 3.595611194633977e-05, "loss": 0.9572, "step": 180560 }, { "epoch": 15.64, "learning_rate": 3.595394356424193e-05, "loss": 0.8462, "step": 180570 }, { "epoch": 15.64, "learning_rate": 3.59517751821441e-05, "loss": 1.0614, "step": 180580 }, { "epoch": 15.64, "learning_rate": 3.5949606800046256e-05, "loss": 0.9851, "step": 180590 }, { "epoch": 15.64, "learning_rate": 3.5947438417948416e-05, "loss": 0.898, "step": 180600 }, { "epoch": 15.64, "learning_rate": 3.594527003585058e-05, "loss": 0.8594, "step": 180610 }, { "epoch": 15.64, "learning_rate": 3.594310165375275e-05, "loss": 0.981, "step": 180620 }, { "epoch": 15.64, "learning_rate": 3.594093327165491e-05, "loss": 1.0177, "step": 180630 }, { "epoch": 15.64, "learning_rate": 3.593876488955707e-05, "loss": 0.737, "step": 180640 }, { "epoch": 15.65, "learning_rate": 3.5936596507459234e-05, "loss": 1.0885, "step": 180650 }, { "epoch": 15.65, "learning_rate": 3.5934428125361394e-05, "loss": 0.7539, "step": 180660 }, { "epoch": 15.65, "learning_rate": 3.5932259743263553e-05, "loss": 0.9329, "step": 180670 }, { "epoch": 15.65, "learning_rate": 3.593009136116572e-05, "loss": 0.8869, "step": 180680 }, { "epoch": 15.65, "learning_rate": 3.5927922979067886e-05, "loss": 0.8426, "step": 180690 }, { "epoch": 15.65, "learning_rate": 3.592575459697004e-05, "loss": 0.7636, "step": 180700 }, { "epoch": 15.65, "learning_rate": 3.5923586214872206e-05, "loss": 0.8089, "step": 180710 }, { "epoch": 15.65, "learning_rate": 3.592141783277437e-05, "loss": 0.9185, "step": 180720 }, { "epoch": 15.65, "learning_rate": 3.591924945067653e-05, "loss": 0.9263, "step": 180730 }, { "epoch": 15.65, "learning_rate": 3.591708106857869e-05, "loss": 0.8527, "step": 180740 }, { "epoch": 15.65, "learning_rate": 3.591491268648086e-05, "loss": 1.1916, "step": 180750 }, { "epoch": 15.65, "learning_rate": 3.591274430438302e-05, "loss": 0.9967, "step": 180760 }, { "epoch": 15.66, "learning_rate": 3.5910575922285184e-05, "loss": 0.779, "step": 180770 }, { "epoch": 15.66, "learning_rate": 3.590840754018734e-05, "loss": 0.964, "step": 180780 }, { "epoch": 15.66, "learning_rate": 3.590623915808951e-05, "loss": 0.9289, "step": 180790 }, { "epoch": 15.66, "learning_rate": 3.590407077599167e-05, "loss": 0.8704, "step": 180800 }, { "epoch": 15.66, "learning_rate": 3.5901902393893836e-05, "loss": 1.0102, "step": 180810 }, { "epoch": 15.66, "learning_rate": 3.5899734011795995e-05, "loss": 1.0751, "step": 180820 }, { "epoch": 15.66, "learning_rate": 3.5897565629698155e-05, "loss": 0.918, "step": 180830 }, { "epoch": 15.66, "learning_rate": 3.589539724760032e-05, "loss": 1.0187, "step": 180840 }, { "epoch": 15.66, "learning_rate": 3.589322886550249e-05, "loss": 0.9066, "step": 180850 }, { "epoch": 15.66, "learning_rate": 3.589106048340465e-05, "loss": 0.8076, "step": 180860 }, { "epoch": 15.66, "learning_rate": 3.588889210130681e-05, "loss": 0.8369, "step": 180870 }, { "epoch": 15.67, "learning_rate": 3.588672371920897e-05, "loss": 0.9473, "step": 180880 }, { "epoch": 15.67, "learning_rate": 3.588455533711113e-05, "loss": 0.9349, "step": 180890 }, { "epoch": 15.67, "learning_rate": 3.588238695501329e-05, "loss": 1.0144, "step": 180900 }, { "epoch": 15.67, "learning_rate": 3.588021857291546e-05, "loss": 0.9339, "step": 180910 }, { "epoch": 15.67, "learning_rate": 3.5878050190817625e-05, "loss": 1.0475, "step": 180920 }, { "epoch": 15.67, "learning_rate": 3.5875881808719785e-05, "loss": 0.9997, "step": 180930 }, { "epoch": 15.67, "learning_rate": 3.5873713426621944e-05, "loss": 1.0202, "step": 180940 }, { "epoch": 15.67, "learning_rate": 3.587154504452411e-05, "loss": 1.0742, "step": 180950 }, { "epoch": 15.67, "learning_rate": 3.586937666242627e-05, "loss": 0.8611, "step": 180960 }, { "epoch": 15.67, "learning_rate": 3.586720828032843e-05, "loss": 0.9488, "step": 180970 }, { "epoch": 15.67, "learning_rate": 3.5865039898230596e-05, "loss": 0.8939, "step": 180980 }, { "epoch": 15.67, "learning_rate": 3.586287151613276e-05, "loss": 0.8759, "step": 180990 }, { "epoch": 15.68, "learning_rate": 3.586070313403492e-05, "loss": 0.8578, "step": 181000 }, { "epoch": 15.68, "learning_rate": 3.585853475193708e-05, "loss": 0.9624, "step": 181010 }, { "epoch": 15.68, "learning_rate": 3.585636636983925e-05, "loss": 0.8997, "step": 181020 }, { "epoch": 15.68, "learning_rate": 3.585419798774141e-05, "loss": 1.0704, "step": 181030 }, { "epoch": 15.68, "learning_rate": 3.5852029605643574e-05, "loss": 0.847, "step": 181040 }, { "epoch": 15.68, "learning_rate": 3.5849861223545734e-05, "loss": 0.8381, "step": 181050 }, { "epoch": 15.68, "learning_rate": 3.58476928414479e-05, "loss": 1.0046, "step": 181060 }, { "epoch": 15.68, "learning_rate": 3.584552445935006e-05, "loss": 0.9151, "step": 181070 }, { "epoch": 15.68, "learning_rate": 3.5843356077252226e-05, "loss": 0.8867, "step": 181080 }, { "epoch": 15.68, "learning_rate": 3.5841187695154386e-05, "loss": 0.918, "step": 181090 }, { "epoch": 15.68, "learning_rate": 3.5839019313056546e-05, "loss": 0.9023, "step": 181100 }, { "epoch": 15.69, "learning_rate": 3.583685093095871e-05, "loss": 0.875, "step": 181110 }, { "epoch": 15.69, "learning_rate": 3.583468254886088e-05, "loss": 0.9104, "step": 181120 }, { "epoch": 15.69, "learning_rate": 3.583251416676304e-05, "loss": 0.8805, "step": 181130 }, { "epoch": 15.69, "learning_rate": 3.58303457846652e-05, "loss": 0.7909, "step": 181140 }, { "epoch": 15.69, "learning_rate": 3.5828177402567364e-05, "loss": 0.917, "step": 181150 }, { "epoch": 15.69, "learning_rate": 3.5826009020469524e-05, "loss": 0.8818, "step": 181160 }, { "epoch": 15.69, "learning_rate": 3.582384063837168e-05, "loss": 0.9263, "step": 181170 }, { "epoch": 15.69, "learning_rate": 3.582167225627385e-05, "loss": 0.9191, "step": 181180 }, { "epoch": 15.69, "learning_rate": 3.5819503874176016e-05, "loss": 0.9508, "step": 181190 }, { "epoch": 15.69, "learning_rate": 3.5817335492078176e-05, "loss": 0.7855, "step": 181200 }, { "epoch": 15.69, "learning_rate": 3.5815167109980335e-05, "loss": 1.0075, "step": 181210 }, { "epoch": 15.69, "learning_rate": 3.58129987278825e-05, "loss": 0.8074, "step": 181220 }, { "epoch": 15.7, "learning_rate": 3.581083034578466e-05, "loss": 0.7583, "step": 181230 }, { "epoch": 15.7, "learning_rate": 3.580866196368682e-05, "loss": 0.8758, "step": 181240 }, { "epoch": 15.7, "learning_rate": 3.580649358158899e-05, "loss": 0.9089, "step": 181250 }, { "epoch": 15.7, "learning_rate": 3.580432519949115e-05, "loss": 0.795, "step": 181260 }, { "epoch": 15.7, "learning_rate": 3.580215681739331e-05, "loss": 0.9513, "step": 181270 }, { "epoch": 15.7, "learning_rate": 3.579998843529547e-05, "loss": 1.0194, "step": 181280 }, { "epoch": 15.7, "learning_rate": 3.579782005319764e-05, "loss": 0.9518, "step": 181290 }, { "epoch": 15.7, "learning_rate": 3.57956516710998e-05, "loss": 0.9644, "step": 181300 }, { "epoch": 15.7, "learning_rate": 3.5793483289001965e-05, "loss": 0.8784, "step": 181310 }, { "epoch": 15.7, "learning_rate": 3.5791314906904125e-05, "loss": 0.8913, "step": 181320 }, { "epoch": 15.7, "learning_rate": 3.5789146524806285e-05, "loss": 0.9162, "step": 181330 }, { "epoch": 15.71, "learning_rate": 3.578697814270845e-05, "loss": 0.9999, "step": 181340 }, { "epoch": 15.71, "learning_rate": 3.578480976061062e-05, "loss": 0.8453, "step": 181350 }, { "epoch": 15.71, "learning_rate": 3.578264137851278e-05, "loss": 0.8358, "step": 181360 }, { "epoch": 15.71, "learning_rate": 3.578047299641494e-05, "loss": 0.9795, "step": 181370 }, { "epoch": 15.71, "learning_rate": 3.57783046143171e-05, "loss": 0.9015, "step": 181380 }, { "epoch": 15.71, "learning_rate": 3.577613623221926e-05, "loss": 0.8472, "step": 181390 }, { "epoch": 15.71, "learning_rate": 3.577396785012142e-05, "loss": 0.7463, "step": 181400 }, { "epoch": 15.71, "learning_rate": 3.577179946802359e-05, "loss": 0.913, "step": 181410 }, { "epoch": 15.71, "learning_rate": 3.5769631085925755e-05, "loss": 0.8128, "step": 181420 }, { "epoch": 15.71, "learning_rate": 3.5767462703827915e-05, "loss": 0.8793, "step": 181430 }, { "epoch": 15.71, "learning_rate": 3.5765294321730074e-05, "loss": 1.011, "step": 181440 }, { "epoch": 15.71, "learning_rate": 3.576312593963224e-05, "loss": 0.8329, "step": 181450 }, { "epoch": 15.72, "learning_rate": 3.57609575575344e-05, "loss": 0.975, "step": 181460 }, { "epoch": 15.72, "learning_rate": 3.575878917543656e-05, "loss": 0.8836, "step": 181470 }, { "epoch": 15.72, "learning_rate": 3.5756620793338726e-05, "loss": 0.9661, "step": 181480 }, { "epoch": 15.72, "learning_rate": 3.575445241124089e-05, "loss": 0.8932, "step": 181490 }, { "epoch": 15.72, "learning_rate": 3.575228402914305e-05, "loss": 0.8462, "step": 181500 }, { "epoch": 15.72, "learning_rate": 3.575011564704521e-05, "loss": 0.9082, "step": 181510 }, { "epoch": 15.72, "learning_rate": 3.574794726494738e-05, "loss": 0.8175, "step": 181520 }, { "epoch": 15.72, "learning_rate": 3.574577888284954e-05, "loss": 0.8344, "step": 181530 }, { "epoch": 15.72, "learning_rate": 3.5743610500751704e-05, "loss": 1.0752, "step": 181540 }, { "epoch": 15.72, "learning_rate": 3.5741442118653864e-05, "loss": 0.8395, "step": 181550 }, { "epoch": 15.72, "learning_rate": 3.573927373655603e-05, "loss": 0.8307, "step": 181560 }, { "epoch": 15.73, "learning_rate": 3.573710535445819e-05, "loss": 0.9136, "step": 181570 }, { "epoch": 15.73, "learning_rate": 3.5734936972360356e-05, "loss": 1.038, "step": 181580 }, { "epoch": 15.73, "learning_rate": 3.5732768590262516e-05, "loss": 0.8937, "step": 181590 }, { "epoch": 15.73, "learning_rate": 3.5730600208164676e-05, "loss": 1.14, "step": 181600 }, { "epoch": 15.73, "learning_rate": 3.572843182606684e-05, "loss": 0.8538, "step": 181610 }, { "epoch": 15.73, "learning_rate": 3.572626344396901e-05, "loss": 0.8906, "step": 181620 }, { "epoch": 15.73, "learning_rate": 3.572409506187117e-05, "loss": 0.8028, "step": 181630 }, { "epoch": 15.73, "learning_rate": 3.572192667977333e-05, "loss": 0.83, "step": 181640 }, { "epoch": 15.73, "learning_rate": 3.5719758297675494e-05, "loss": 1.1271, "step": 181650 }, { "epoch": 15.73, "learning_rate": 3.5717589915577654e-05, "loss": 1.0602, "step": 181660 }, { "epoch": 15.73, "learning_rate": 3.571542153347981e-05, "loss": 0.8596, "step": 181670 }, { "epoch": 15.73, "learning_rate": 3.571325315138198e-05, "loss": 0.9679, "step": 181680 }, { "epoch": 15.74, "learning_rate": 3.5711084769284146e-05, "loss": 0.9986, "step": 181690 }, { "epoch": 15.74, "learning_rate": 3.5708916387186306e-05, "loss": 0.8868, "step": 181700 }, { "epoch": 15.74, "learning_rate": 3.5706748005088465e-05, "loss": 0.8736, "step": 181710 }, { "epoch": 15.74, "learning_rate": 3.570457962299063e-05, "loss": 0.9622, "step": 181720 }, { "epoch": 15.74, "learning_rate": 3.570241124089279e-05, "loss": 1.0085, "step": 181730 }, { "epoch": 15.74, "learning_rate": 3.570024285879495e-05, "loss": 0.8212, "step": 181740 }, { "epoch": 15.74, "learning_rate": 3.569807447669712e-05, "loss": 0.7964, "step": 181750 }, { "epoch": 15.74, "learning_rate": 3.5695906094599284e-05, "loss": 0.957, "step": 181760 }, { "epoch": 15.74, "learning_rate": 3.569373771250144e-05, "loss": 1.0009, "step": 181770 }, { "epoch": 15.74, "learning_rate": 3.56915693304036e-05, "loss": 0.7953, "step": 181780 }, { "epoch": 15.74, "learning_rate": 3.568940094830577e-05, "loss": 0.9673, "step": 181790 }, { "epoch": 15.75, "learning_rate": 3.568723256620793e-05, "loss": 0.7411, "step": 181800 }, { "epoch": 15.75, "learning_rate": 3.5685064184110095e-05, "loss": 0.9239, "step": 181810 }, { "epoch": 15.75, "learning_rate": 3.5682895802012255e-05, "loss": 0.8452, "step": 181820 }, { "epoch": 15.75, "learning_rate": 3.5680727419914414e-05, "loss": 0.9674, "step": 181830 }, { "epoch": 15.75, "learning_rate": 3.567855903781658e-05, "loss": 0.6986, "step": 181840 }, { "epoch": 15.75, "learning_rate": 3.567639065571875e-05, "loss": 0.9343, "step": 181850 }, { "epoch": 15.75, "learning_rate": 3.567422227362091e-05, "loss": 0.8717, "step": 181860 }, { "epoch": 15.75, "learning_rate": 3.5672053891523066e-05, "loss": 0.9955, "step": 181870 }, { "epoch": 15.75, "learning_rate": 3.566988550942523e-05, "loss": 0.9824, "step": 181880 }, { "epoch": 15.75, "learning_rate": 3.566771712732739e-05, "loss": 0.8601, "step": 181890 }, { "epoch": 15.75, "learning_rate": 3.566554874522955e-05, "loss": 1.0009, "step": 181900 }, { "epoch": 15.75, "learning_rate": 3.566338036313172e-05, "loss": 0.7798, "step": 181910 }, { "epoch": 15.76, "learning_rate": 3.5661211981033885e-05, "loss": 0.9599, "step": 181920 }, { "epoch": 15.76, "learning_rate": 3.5659043598936045e-05, "loss": 0.8592, "step": 181930 }, { "epoch": 15.76, "learning_rate": 3.5656875216838204e-05, "loss": 0.8381, "step": 181940 }, { "epoch": 15.76, "learning_rate": 3.565470683474037e-05, "loss": 0.9283, "step": 181950 }, { "epoch": 15.76, "learning_rate": 3.565253845264253e-05, "loss": 0.9855, "step": 181960 }, { "epoch": 15.76, "learning_rate": 3.565037007054469e-05, "loss": 1.0696, "step": 181970 }, { "epoch": 15.76, "learning_rate": 3.5648201688446856e-05, "loss": 0.7319, "step": 181980 }, { "epoch": 15.76, "learning_rate": 3.564603330634902e-05, "loss": 0.8281, "step": 181990 }, { "epoch": 15.76, "learning_rate": 3.564386492425118e-05, "loss": 0.8155, "step": 182000 }, { "epoch": 15.76, "learning_rate": 3.564169654215334e-05, "loss": 0.9637, "step": 182010 }, { "epoch": 15.76, "learning_rate": 3.563952816005551e-05, "loss": 0.9974, "step": 182020 }, { "epoch": 15.76, "learning_rate": 3.563735977795767e-05, "loss": 0.97, "step": 182030 }, { "epoch": 15.77, "learning_rate": 3.5635191395859834e-05, "loss": 0.9782, "step": 182040 }, { "epoch": 15.77, "learning_rate": 3.5633023013761994e-05, "loss": 0.863, "step": 182050 }, { "epoch": 15.77, "learning_rate": 3.563085463166416e-05, "loss": 1.0193, "step": 182060 }, { "epoch": 15.77, "learning_rate": 3.562868624956632e-05, "loss": 1.082, "step": 182070 }, { "epoch": 15.77, "learning_rate": 3.5626517867468486e-05, "loss": 0.8084, "step": 182080 }, { "epoch": 15.77, "learning_rate": 3.5624349485370646e-05, "loss": 0.7752, "step": 182090 }, { "epoch": 15.77, "learning_rate": 3.5622181103272805e-05, "loss": 0.9988, "step": 182100 }, { "epoch": 15.77, "learning_rate": 3.562001272117497e-05, "loss": 0.9198, "step": 182110 }, { "epoch": 15.77, "learning_rate": 3.561784433907714e-05, "loss": 1.0294, "step": 182120 }, { "epoch": 15.77, "learning_rate": 3.56156759569793e-05, "loss": 1.0087, "step": 182130 }, { "epoch": 15.77, "learning_rate": 3.561350757488146e-05, "loss": 0.9277, "step": 182140 }, { "epoch": 15.78, "learning_rate": 3.5611339192783624e-05, "loss": 1.0739, "step": 182150 }, { "epoch": 15.78, "learning_rate": 3.5609170810685783e-05, "loss": 1.04, "step": 182160 }, { "epoch": 15.78, "learning_rate": 3.560700242858794e-05, "loss": 0.7983, "step": 182170 }, { "epoch": 15.78, "learning_rate": 3.560483404649011e-05, "loss": 0.904, "step": 182180 }, { "epoch": 15.78, "learning_rate": 3.5602665664392276e-05, "loss": 0.9729, "step": 182190 }, { "epoch": 15.78, "learning_rate": 3.5600497282294435e-05, "loss": 0.7078, "step": 182200 }, { "epoch": 15.78, "learning_rate": 3.5598328900196595e-05, "loss": 0.9169, "step": 182210 }, { "epoch": 15.78, "learning_rate": 3.559616051809876e-05, "loss": 1.0183, "step": 182220 }, { "epoch": 15.78, "learning_rate": 3.559399213600092e-05, "loss": 0.9685, "step": 182230 }, { "epoch": 15.78, "learning_rate": 3.559182375390308e-05, "loss": 0.7949, "step": 182240 }, { "epoch": 15.78, "learning_rate": 3.558965537180525e-05, "loss": 0.7985, "step": 182250 }, { "epoch": 15.78, "learning_rate": 3.5587486989707413e-05, "loss": 0.9378, "step": 182260 }, { "epoch": 15.79, "learning_rate": 3.558531860760957e-05, "loss": 1.0417, "step": 182270 }, { "epoch": 15.79, "learning_rate": 3.558315022551173e-05, "loss": 0.9346, "step": 182280 }, { "epoch": 15.79, "learning_rate": 3.55809818434139e-05, "loss": 0.6088, "step": 182290 }, { "epoch": 15.79, "learning_rate": 3.557881346131606e-05, "loss": 1.0667, "step": 182300 }, { "epoch": 15.79, "learning_rate": 3.5576645079218225e-05, "loss": 0.8735, "step": 182310 }, { "epoch": 15.79, "learning_rate": 3.5574476697120385e-05, "loss": 1.0032, "step": 182320 }, { "epoch": 15.79, "learning_rate": 3.5572308315022544e-05, "loss": 1.1576, "step": 182330 }, { "epoch": 15.79, "learning_rate": 3.557013993292471e-05, "loss": 0.9691, "step": 182340 }, { "epoch": 15.79, "learning_rate": 3.556797155082688e-05, "loss": 1.0889, "step": 182350 }, { "epoch": 15.79, "learning_rate": 3.556580316872904e-05, "loss": 0.8362, "step": 182360 }, { "epoch": 15.79, "learning_rate": 3.5563634786631196e-05, "loss": 0.8958, "step": 182370 }, { "epoch": 15.8, "learning_rate": 3.556146640453336e-05, "loss": 1.0622, "step": 182380 }, { "epoch": 15.8, "learning_rate": 3.555929802243552e-05, "loss": 0.8727, "step": 182390 }, { "epoch": 15.8, "learning_rate": 3.555712964033768e-05, "loss": 0.9224, "step": 182400 }, { "epoch": 15.8, "learning_rate": 3.555496125823985e-05, "loss": 0.9122, "step": 182410 }, { "epoch": 15.8, "learning_rate": 3.5552792876142015e-05, "loss": 0.9174, "step": 182420 }, { "epoch": 15.8, "learning_rate": 3.5550624494044174e-05, "loss": 0.957, "step": 182430 }, { "epoch": 15.8, "learning_rate": 3.5548456111946334e-05, "loss": 1.0313, "step": 182440 }, { "epoch": 15.8, "learning_rate": 3.55462877298485e-05, "loss": 1.0381, "step": 182450 }, { "epoch": 15.8, "learning_rate": 3.554411934775066e-05, "loss": 0.8853, "step": 182460 }, { "epoch": 15.8, "learning_rate": 3.554195096565282e-05, "loss": 1.0308, "step": 182470 }, { "epoch": 15.8, "learning_rate": 3.5539782583554986e-05, "loss": 1.0616, "step": 182480 }, { "epoch": 15.8, "learning_rate": 3.553761420145715e-05, "loss": 1.0711, "step": 182490 }, { "epoch": 15.81, "learning_rate": 3.553544581935931e-05, "loss": 0.983, "step": 182500 }, { "epoch": 15.81, "learning_rate": 3.553327743726147e-05, "loss": 1.0279, "step": 182510 }, { "epoch": 15.81, "learning_rate": 3.553110905516364e-05, "loss": 0.8044, "step": 182520 }, { "epoch": 15.81, "learning_rate": 3.55289406730658e-05, "loss": 0.8154, "step": 182530 }, { "epoch": 15.81, "learning_rate": 3.5526772290967964e-05, "loss": 0.7142, "step": 182540 }, { "epoch": 15.81, "learning_rate": 3.5524603908870124e-05, "loss": 1.0973, "step": 182550 }, { "epoch": 15.81, "learning_rate": 3.552243552677229e-05, "loss": 0.9669, "step": 182560 }, { "epoch": 15.81, "learning_rate": 3.552026714467445e-05, "loss": 0.7838, "step": 182570 }, { "epoch": 15.81, "learning_rate": 3.5518098762576616e-05, "loss": 0.9446, "step": 182580 }, { "epoch": 15.81, "learning_rate": 3.5515930380478776e-05, "loss": 0.9753, "step": 182590 }, { "epoch": 15.81, "learning_rate": 3.5513761998380935e-05, "loss": 0.8615, "step": 182600 }, { "epoch": 15.82, "learning_rate": 3.55115936162831e-05, "loss": 0.8109, "step": 182610 }, { "epoch": 15.82, "learning_rate": 3.550942523418527e-05, "loss": 0.9333, "step": 182620 }, { "epoch": 15.82, "learning_rate": 3.550725685208743e-05, "loss": 0.906, "step": 182630 }, { "epoch": 15.82, "learning_rate": 3.550508846998959e-05, "loss": 0.8878, "step": 182640 }, { "epoch": 15.82, "learning_rate": 3.5502920087891754e-05, "loss": 0.9624, "step": 182650 }, { "epoch": 15.82, "learning_rate": 3.550075170579391e-05, "loss": 0.9671, "step": 182660 }, { "epoch": 15.82, "learning_rate": 3.549858332369607e-05, "loss": 0.9405, "step": 182670 }, { "epoch": 15.82, "learning_rate": 3.549641494159824e-05, "loss": 0.9162, "step": 182680 }, { "epoch": 15.82, "learning_rate": 3.5494246559500406e-05, "loss": 0.7139, "step": 182690 }, { "epoch": 15.82, "learning_rate": 3.5492078177402565e-05, "loss": 0.8008, "step": 182700 }, { "epoch": 15.82, "learning_rate": 3.5489909795304725e-05, "loss": 0.8456, "step": 182710 }, { "epoch": 15.82, "learning_rate": 3.548774141320689e-05, "loss": 0.9158, "step": 182720 }, { "epoch": 15.83, "learning_rate": 3.548557303110905e-05, "loss": 0.948, "step": 182730 }, { "epoch": 15.83, "learning_rate": 3.548340464901121e-05, "loss": 0.908, "step": 182740 }, { "epoch": 15.83, "learning_rate": 3.548123626691338e-05, "loss": 0.783, "step": 182750 }, { "epoch": 15.83, "learning_rate": 3.547906788481554e-05, "loss": 0.9145, "step": 182760 }, { "epoch": 15.83, "learning_rate": 3.54768995027177e-05, "loss": 0.8781, "step": 182770 }, { "epoch": 15.83, "learning_rate": 3.547473112061986e-05, "loss": 0.9755, "step": 182780 }, { "epoch": 15.83, "learning_rate": 3.547256273852203e-05, "loss": 1.0047, "step": 182790 }, { "epoch": 15.83, "learning_rate": 3.547039435642419e-05, "loss": 0.7876, "step": 182800 }, { "epoch": 15.83, "learning_rate": 3.5468225974326355e-05, "loss": 0.8548, "step": 182810 }, { "epoch": 15.83, "learning_rate": 3.5466057592228515e-05, "loss": 0.9151, "step": 182820 }, { "epoch": 15.83, "learning_rate": 3.546388921013068e-05, "loss": 0.8584, "step": 182830 }, { "epoch": 15.84, "learning_rate": 3.546172082803284e-05, "loss": 0.7575, "step": 182840 }, { "epoch": 15.84, "learning_rate": 3.545955244593501e-05, "loss": 0.9848, "step": 182850 }, { "epoch": 15.84, "learning_rate": 3.5457384063837167e-05, "loss": 1.0189, "step": 182860 }, { "epoch": 15.84, "learning_rate": 3.5455215681739326e-05, "loss": 0.8102, "step": 182870 }, { "epoch": 15.84, "learning_rate": 3.545304729964149e-05, "loss": 0.9102, "step": 182880 }, { "epoch": 15.84, "learning_rate": 3.545087891754365e-05, "loss": 1.0452, "step": 182890 }, { "epoch": 15.84, "learning_rate": 3.544871053544581e-05, "loss": 0.8439, "step": 182900 }, { "epoch": 15.84, "learning_rate": 3.544654215334798e-05, "loss": 1.0117, "step": 182910 }, { "epoch": 15.84, "learning_rate": 3.5444373771250145e-05, "loss": 0.9122, "step": 182920 }, { "epoch": 15.84, "learning_rate": 3.5442205389152304e-05, "loss": 0.8806, "step": 182930 }, { "epoch": 15.84, "learning_rate": 3.5440037007054464e-05, "loss": 1.1381, "step": 182940 }, { "epoch": 15.84, "learning_rate": 3.543786862495663e-05, "loss": 1.009, "step": 182950 }, { "epoch": 15.85, "learning_rate": 3.543570024285879e-05, "loss": 0.9908, "step": 182960 }, { "epoch": 15.85, "learning_rate": 3.543353186076095e-05, "loss": 0.9303, "step": 182970 }, { "epoch": 15.85, "learning_rate": 3.5431363478663116e-05, "loss": 0.868, "step": 182980 }, { "epoch": 15.85, "learning_rate": 3.542919509656528e-05, "loss": 0.9607, "step": 182990 }, { "epoch": 15.85, "learning_rate": 3.542702671446744e-05, "loss": 0.9895, "step": 183000 }, { "epoch": 15.85, "learning_rate": 3.54248583323696e-05, "loss": 0.8638, "step": 183010 }, { "epoch": 15.85, "learning_rate": 3.542268995027177e-05, "loss": 1.024, "step": 183020 }, { "epoch": 15.85, "learning_rate": 3.542052156817393e-05, "loss": 0.9796, "step": 183030 }, { "epoch": 15.85, "learning_rate": 3.5418353186076094e-05, "loss": 0.8866, "step": 183040 }, { "epoch": 15.85, "learning_rate": 3.5416184803978253e-05, "loss": 0.9113, "step": 183050 }, { "epoch": 15.85, "learning_rate": 3.541401642188042e-05, "loss": 0.9765, "step": 183060 }, { "epoch": 15.86, "learning_rate": 3.541184803978258e-05, "loss": 0.8526, "step": 183070 }, { "epoch": 15.86, "learning_rate": 3.5409679657684746e-05, "loss": 0.9533, "step": 183080 }, { "epoch": 15.86, "learning_rate": 3.5407511275586905e-05, "loss": 0.9479, "step": 183090 }, { "epoch": 15.86, "learning_rate": 3.5405342893489065e-05, "loss": 0.9114, "step": 183100 }, { "epoch": 15.86, "learning_rate": 3.540317451139123e-05, "loss": 0.8616, "step": 183110 }, { "epoch": 15.86, "learning_rate": 3.54010061292934e-05, "loss": 0.8719, "step": 183120 }, { "epoch": 15.86, "learning_rate": 3.539883774719556e-05, "loss": 0.9507, "step": 183130 }, { "epoch": 15.86, "learning_rate": 3.539666936509772e-05, "loss": 0.9177, "step": 183140 }, { "epoch": 15.86, "learning_rate": 3.5394500982999883e-05, "loss": 0.9497, "step": 183150 }, { "epoch": 15.86, "learning_rate": 3.539233260090204e-05, "loss": 1.0893, "step": 183160 }, { "epoch": 15.86, "learning_rate": 3.53901642188042e-05, "loss": 0.9077, "step": 183170 }, { "epoch": 15.86, "learning_rate": 3.538799583670637e-05, "loss": 0.9143, "step": 183180 }, { "epoch": 15.87, "learning_rate": 3.5385827454608536e-05, "loss": 0.8213, "step": 183190 }, { "epoch": 15.87, "learning_rate": 3.5383659072510695e-05, "loss": 0.9255, "step": 183200 }, { "epoch": 15.87, "learning_rate": 3.5381490690412855e-05, "loss": 0.8026, "step": 183210 }, { "epoch": 15.87, "learning_rate": 3.537932230831502e-05, "loss": 0.9529, "step": 183220 }, { "epoch": 15.87, "learning_rate": 3.537715392621718e-05, "loss": 0.9131, "step": 183230 }, { "epoch": 15.87, "learning_rate": 3.537498554411934e-05, "loss": 0.8717, "step": 183240 }, { "epoch": 15.87, "learning_rate": 3.537281716202151e-05, "loss": 1.0181, "step": 183250 }, { "epoch": 15.87, "learning_rate": 3.537064877992367e-05, "loss": 1.0175, "step": 183260 }, { "epoch": 15.87, "learning_rate": 3.536848039782583e-05, "loss": 0.9357, "step": 183270 }, { "epoch": 15.87, "learning_rate": 3.536631201572799e-05, "loss": 1.0031, "step": 183280 }, { "epoch": 15.87, "learning_rate": 3.536414363363016e-05, "loss": 0.9012, "step": 183290 }, { "epoch": 15.87, "learning_rate": 3.536197525153232e-05, "loss": 0.9276, "step": 183300 }, { "epoch": 15.88, "learning_rate": 3.5359806869434485e-05, "loss": 1.0277, "step": 183310 }, { "epoch": 15.88, "learning_rate": 3.5357638487336644e-05, "loss": 0.8793, "step": 183320 }, { "epoch": 15.88, "learning_rate": 3.535547010523881e-05, "loss": 0.9665, "step": 183330 }, { "epoch": 15.88, "learning_rate": 3.535330172314097e-05, "loss": 0.7649, "step": 183340 }, { "epoch": 15.88, "learning_rate": 3.535113334104314e-05, "loss": 0.8994, "step": 183350 }, { "epoch": 15.88, "learning_rate": 3.5348964958945296e-05, "loss": 0.9179, "step": 183360 }, { "epoch": 15.88, "learning_rate": 3.5346796576847456e-05, "loss": 0.7115, "step": 183370 }, { "epoch": 15.88, "learning_rate": 3.534462819474962e-05, "loss": 0.9532, "step": 183380 }, { "epoch": 15.88, "learning_rate": 3.534245981265179e-05, "loss": 0.8291, "step": 183390 }, { "epoch": 15.88, "learning_rate": 3.534029143055394e-05, "loss": 0.7445, "step": 183400 }, { "epoch": 15.88, "learning_rate": 3.533812304845611e-05, "loss": 0.886, "step": 183410 }, { "epoch": 15.89, "learning_rate": 3.5335954666358274e-05, "loss": 0.9235, "step": 183420 }, { "epoch": 15.89, "learning_rate": 3.5333786284260434e-05, "loss": 0.8328, "step": 183430 }, { "epoch": 15.89, "learning_rate": 3.5331617902162594e-05, "loss": 0.8811, "step": 183440 }, { "epoch": 15.89, "learning_rate": 3.532944952006476e-05, "loss": 0.8923, "step": 183450 }, { "epoch": 15.89, "learning_rate": 3.532728113796692e-05, "loss": 1.0537, "step": 183460 }, { "epoch": 15.89, "learning_rate": 3.532511275586908e-05, "loss": 1.123, "step": 183470 }, { "epoch": 15.89, "learning_rate": 3.5322944373771246e-05, "loss": 0.9158, "step": 183480 }, { "epoch": 15.89, "learning_rate": 3.532077599167341e-05, "loss": 0.9191, "step": 183490 }, { "epoch": 15.89, "learning_rate": 3.531860760957557e-05, "loss": 0.8429, "step": 183500 }, { "epoch": 15.89, "learning_rate": 3.531643922747773e-05, "loss": 0.9741, "step": 183510 }, { "epoch": 15.89, "learning_rate": 3.53142708453799e-05, "loss": 1.0307, "step": 183520 }, { "epoch": 15.89, "learning_rate": 3.531210246328206e-05, "loss": 0.887, "step": 183530 }, { "epoch": 15.9, "learning_rate": 3.5309934081184224e-05, "loss": 0.9151, "step": 183540 }, { "epoch": 15.9, "learning_rate": 3.530776569908638e-05, "loss": 1.0866, "step": 183550 }, { "epoch": 15.9, "learning_rate": 3.530559731698855e-05, "loss": 0.8767, "step": 183560 }, { "epoch": 15.9, "learning_rate": 3.530342893489071e-05, "loss": 0.8962, "step": 183570 }, { "epoch": 15.9, "learning_rate": 3.5301260552792876e-05, "loss": 0.937, "step": 183580 }, { "epoch": 15.9, "learning_rate": 3.5299092170695035e-05, "loss": 0.9683, "step": 183590 }, { "epoch": 15.9, "learning_rate": 3.5296923788597195e-05, "loss": 1.0581, "step": 183600 }, { "epoch": 15.9, "learning_rate": 3.529475540649936e-05, "loss": 0.8294, "step": 183610 }, { "epoch": 15.9, "learning_rate": 3.529258702440153e-05, "loss": 0.7915, "step": 183620 }, { "epoch": 15.9, "learning_rate": 3.529041864230369e-05, "loss": 0.8002, "step": 183630 }, { "epoch": 15.9, "learning_rate": 3.528825026020585e-05, "loss": 0.9021, "step": 183640 }, { "epoch": 15.91, "learning_rate": 3.528608187810801e-05, "loss": 0.855, "step": 183650 }, { "epoch": 15.91, "learning_rate": 3.528391349601017e-05, "loss": 0.8449, "step": 183660 }, { "epoch": 15.91, "learning_rate": 3.528174511391233e-05, "loss": 0.987, "step": 183670 }, { "epoch": 15.91, "learning_rate": 3.52795767318145e-05, "loss": 1.0191, "step": 183680 }, { "epoch": 15.91, "learning_rate": 3.5277408349716665e-05, "loss": 0.8726, "step": 183690 }, { "epoch": 15.91, "learning_rate": 3.5275239967618825e-05, "loss": 0.905, "step": 183700 }, { "epoch": 15.91, "learning_rate": 3.5273071585520985e-05, "loss": 0.9052, "step": 183710 }, { "epoch": 15.91, "learning_rate": 3.527090320342315e-05, "loss": 0.9353, "step": 183720 }, { "epoch": 15.91, "learning_rate": 3.526873482132531e-05, "loss": 1.0198, "step": 183730 }, { "epoch": 15.91, "learning_rate": 3.526656643922747e-05, "loss": 0.8476, "step": 183740 }, { "epoch": 15.91, "learning_rate": 3.5264398057129637e-05, "loss": 0.9034, "step": 183750 }, { "epoch": 15.91, "learning_rate": 3.52622296750318e-05, "loss": 0.7531, "step": 183760 }, { "epoch": 15.92, "learning_rate": 3.526006129293396e-05, "loss": 1.1179, "step": 183770 }, { "epoch": 15.92, "learning_rate": 3.525789291083612e-05, "loss": 0.9733, "step": 183780 }, { "epoch": 15.92, "learning_rate": 3.525572452873829e-05, "loss": 0.9322, "step": 183790 }, { "epoch": 15.92, "learning_rate": 3.525355614664045e-05, "loss": 0.8404, "step": 183800 }, { "epoch": 15.92, "learning_rate": 3.5251387764542615e-05, "loss": 0.863, "step": 183810 }, { "epoch": 15.92, "learning_rate": 3.5249219382444774e-05, "loss": 0.8492, "step": 183820 }, { "epoch": 15.92, "learning_rate": 3.524705100034694e-05, "loss": 0.98, "step": 183830 }, { "epoch": 15.92, "learning_rate": 3.52448826182491e-05, "loss": 0.8866, "step": 183840 }, { "epoch": 15.92, "learning_rate": 3.524271423615127e-05, "loss": 1.0639, "step": 183850 }, { "epoch": 15.92, "learning_rate": 3.5240545854053426e-05, "loss": 0.8157, "step": 183860 }, { "epoch": 15.92, "learning_rate": 3.5238377471955586e-05, "loss": 0.849, "step": 183870 }, { "epoch": 15.93, "learning_rate": 3.523620908985775e-05, "loss": 0.8859, "step": 183880 }, { "epoch": 15.93, "learning_rate": 3.523404070775992e-05, "loss": 1.05, "step": 183890 }, { "epoch": 15.93, "learning_rate": 3.523187232566208e-05, "loss": 0.9637, "step": 183900 }, { "epoch": 15.93, "learning_rate": 3.522970394356424e-05, "loss": 0.8895, "step": 183910 }, { "epoch": 15.93, "learning_rate": 3.5227535561466404e-05, "loss": 0.8032, "step": 183920 }, { "epoch": 15.93, "learning_rate": 3.5225367179368564e-05, "loss": 0.8578, "step": 183930 }, { "epoch": 15.93, "learning_rate": 3.5223198797270724e-05, "loss": 1.0266, "step": 183940 }, { "epoch": 15.93, "learning_rate": 3.522103041517289e-05, "loss": 0.9654, "step": 183950 }, { "epoch": 15.93, "learning_rate": 3.521886203307505e-05, "loss": 0.9462, "step": 183960 }, { "epoch": 15.93, "learning_rate": 3.521669365097721e-05, "loss": 0.9452, "step": 183970 }, { "epoch": 15.93, "learning_rate": 3.5214525268879376e-05, "loss": 0.8077, "step": 183980 }, { "epoch": 15.93, "learning_rate": 3.521235688678154e-05, "loss": 0.7479, "step": 183990 }, { "epoch": 15.94, "learning_rate": 3.52101885046837e-05, "loss": 0.9457, "step": 184000 }, { "epoch": 15.94, "learning_rate": 3.520802012258586e-05, "loss": 0.9551, "step": 184010 }, { "epoch": 15.94, "learning_rate": 3.520585174048803e-05, "loss": 0.9636, "step": 184020 }, { "epoch": 15.94, "learning_rate": 3.520368335839019e-05, "loss": 0.9617, "step": 184030 }, { "epoch": 15.94, "learning_rate": 3.5201514976292354e-05, "loss": 0.9846, "step": 184040 }, { "epoch": 15.94, "learning_rate": 3.519934659419451e-05, "loss": 0.9289, "step": 184050 }, { "epoch": 15.94, "learning_rate": 3.519717821209668e-05, "loss": 0.8749, "step": 184060 }, { "epoch": 15.94, "learning_rate": 3.519500982999884e-05, "loss": 1.0891, "step": 184070 }, { "epoch": 15.94, "learning_rate": 3.5192841447901006e-05, "loss": 0.9296, "step": 184080 }, { "epoch": 15.94, "learning_rate": 3.5190673065803165e-05, "loss": 0.9518, "step": 184090 }, { "epoch": 15.94, "learning_rate": 3.5188504683705325e-05, "loss": 0.7127, "step": 184100 }, { "epoch": 15.95, "learning_rate": 3.518633630160749e-05, "loss": 0.8137, "step": 184110 }, { "epoch": 15.95, "learning_rate": 3.518416791950966e-05, "loss": 1.055, "step": 184120 }, { "epoch": 15.95, "learning_rate": 3.518199953741182e-05, "loss": 1.0031, "step": 184130 }, { "epoch": 15.95, "learning_rate": 3.517983115531398e-05, "loss": 0.8717, "step": 184140 }, { "epoch": 15.95, "learning_rate": 3.517766277321614e-05, "loss": 0.9911, "step": 184150 }, { "epoch": 15.95, "learning_rate": 3.51754943911183e-05, "loss": 0.7577, "step": 184160 }, { "epoch": 15.95, "learning_rate": 3.517332600902046e-05, "loss": 0.8673, "step": 184170 }, { "epoch": 15.95, "learning_rate": 3.517115762692263e-05, "loss": 0.8637, "step": 184180 }, { "epoch": 15.95, "learning_rate": 3.5168989244824795e-05, "loss": 0.8639, "step": 184190 }, { "epoch": 15.95, "learning_rate": 3.5166820862726955e-05, "loss": 0.862, "step": 184200 }, { "epoch": 15.95, "learning_rate": 3.5164652480629114e-05, "loss": 0.8212, "step": 184210 }, { "epoch": 15.95, "learning_rate": 3.516248409853128e-05, "loss": 0.8512, "step": 184220 }, { "epoch": 15.96, "learning_rate": 3.516031571643344e-05, "loss": 0.9557, "step": 184230 }, { "epoch": 15.96, "learning_rate": 3.51581473343356e-05, "loss": 0.9815, "step": 184240 }, { "epoch": 15.96, "learning_rate": 3.5155978952237766e-05, "loss": 0.9259, "step": 184250 }, { "epoch": 15.96, "learning_rate": 3.515381057013993e-05, "loss": 0.8948, "step": 184260 }, { "epoch": 15.96, "learning_rate": 3.515164218804209e-05, "loss": 0.9215, "step": 184270 }, { "epoch": 15.96, "learning_rate": 3.514947380594425e-05, "loss": 0.8634, "step": 184280 }, { "epoch": 15.96, "learning_rate": 3.514730542384642e-05, "loss": 0.9948, "step": 184290 }, { "epoch": 15.96, "learning_rate": 3.514513704174858e-05, "loss": 0.9523, "step": 184300 }, { "epoch": 15.96, "learning_rate": 3.5142968659650744e-05, "loss": 0.9358, "step": 184310 }, { "epoch": 15.96, "learning_rate": 3.5140800277552904e-05, "loss": 1.0395, "step": 184320 }, { "epoch": 15.96, "learning_rate": 3.513863189545507e-05, "loss": 0.7694, "step": 184330 }, { "epoch": 15.97, "learning_rate": 3.513646351335723e-05, "loss": 0.8358, "step": 184340 }, { "epoch": 15.97, "learning_rate": 3.5134295131259396e-05, "loss": 0.8545, "step": 184350 }, { "epoch": 15.97, "learning_rate": 3.5132126749161556e-05, "loss": 0.7609, "step": 184360 }, { "epoch": 15.97, "learning_rate": 3.5129958367063716e-05, "loss": 0.9259, "step": 184370 }, { "epoch": 15.97, "learning_rate": 3.512778998496588e-05, "loss": 1.0314, "step": 184380 }, { "epoch": 15.97, "learning_rate": 3.512562160286805e-05, "loss": 0.8466, "step": 184390 }, { "epoch": 15.97, "learning_rate": 3.512345322077021e-05, "loss": 0.8353, "step": 184400 }, { "epoch": 15.97, "learning_rate": 3.512128483867237e-05, "loss": 1.0303, "step": 184410 }, { "epoch": 15.97, "learning_rate": 3.5119116456574534e-05, "loss": 0.8377, "step": 184420 }, { "epoch": 15.97, "learning_rate": 3.5116948074476694e-05, "loss": 0.7737, "step": 184430 }, { "epoch": 15.97, "learning_rate": 3.511477969237885e-05, "loss": 0.931, "step": 184440 }, { "epoch": 15.97, "learning_rate": 3.511261131028102e-05, "loss": 1.0225, "step": 184450 }, { "epoch": 15.98, "learning_rate": 3.5110442928183186e-05, "loss": 0.9003, "step": 184460 }, { "epoch": 15.98, "learning_rate": 3.5108274546085346e-05, "loss": 1.0872, "step": 184470 }, { "epoch": 15.98, "learning_rate": 3.5106106163987505e-05, "loss": 1.1161, "step": 184480 }, { "epoch": 15.98, "learning_rate": 3.510393778188967e-05, "loss": 0.9203, "step": 184490 }, { "epoch": 15.98, "learning_rate": 3.510176939979183e-05, "loss": 0.9231, "step": 184500 }, { "epoch": 15.98, "learning_rate": 3.509960101769399e-05, "loss": 0.8243, "step": 184510 }, { "epoch": 15.98, "learning_rate": 3.509743263559616e-05, "loss": 1.0381, "step": 184520 }, { "epoch": 15.98, "learning_rate": 3.509526425349832e-05, "loss": 1.0126, "step": 184530 }, { "epoch": 15.98, "learning_rate": 3.509309587140048e-05, "loss": 0.8642, "step": 184540 }, { "epoch": 15.98, "learning_rate": 3.509092748930264e-05, "loss": 0.9522, "step": 184550 }, { "epoch": 15.98, "learning_rate": 3.508875910720481e-05, "loss": 0.893, "step": 184560 }, { "epoch": 15.98, "learning_rate": 3.508659072510697e-05, "loss": 0.9206, "step": 184570 }, { "epoch": 15.99, "learning_rate": 3.5084422343009135e-05, "loss": 1.0213, "step": 184580 }, { "epoch": 15.99, "learning_rate": 3.5082253960911295e-05, "loss": 0.8623, "step": 184590 }, { "epoch": 15.99, "learning_rate": 3.5080085578813455e-05, "loss": 0.7761, "step": 184600 }, { "epoch": 15.99, "learning_rate": 3.507791719671562e-05, "loss": 0.8493, "step": 184610 }, { "epoch": 15.99, "learning_rate": 3.507574881461779e-05, "loss": 0.9324, "step": 184620 }, { "epoch": 15.99, "learning_rate": 3.507358043251995e-05, "loss": 0.8031, "step": 184630 }, { "epoch": 15.99, "learning_rate": 3.507141205042211e-05, "loss": 0.8067, "step": 184640 }, { "epoch": 15.99, "learning_rate": 3.506924366832427e-05, "loss": 0.8334, "step": 184650 }, { "epoch": 15.99, "learning_rate": 3.506707528622643e-05, "loss": 1.0277, "step": 184660 }, { "epoch": 15.99, "learning_rate": 3.506490690412859e-05, "loss": 0.8501, "step": 184670 }, { "epoch": 15.99, "learning_rate": 3.506273852203076e-05, "loss": 0.9276, "step": 184680 }, { "epoch": 16.0, "learning_rate": 3.5060570139932925e-05, "loss": 1.0541, "step": 184690 }, { "epoch": 16.0, "learning_rate": 3.5058401757835085e-05, "loss": 0.8905, "step": 184700 }, { "epoch": 16.0, "learning_rate": 3.5056233375737244e-05, "loss": 1.0245, "step": 184710 }, { "epoch": 16.0, "learning_rate": 3.505406499363941e-05, "loss": 0.8507, "step": 184720 }, { "epoch": 16.0, "learning_rate": 3.505189661154157e-05, "loss": 0.9466, "step": 184730 }, { "epoch": 16.0, "learning_rate": 3.504972822944373e-05, "loss": 0.9318, "step": 184740 }, { "epoch": 16.0, "eval_cer": 0.5406939243545411, "eval_em": 0.05105724600309438, "eval_f1": 0.05105724600309438, "eval_loss": 1.474674105644226, "eval_runtime": 262.2346, "eval_samples_per_second": 7.394, "eval_steps_per_second": 0.927, "eval_wer": 0.9489427539969056, "step": 184744 }, { "epoch": 16.0, "learning_rate": 3.5047559847345896e-05, "loss": 0.838, "step": 184750 }, { "epoch": 16.0, "learning_rate": 3.504539146524806e-05, "loss": 0.8495, "step": 184760 }, { "epoch": 16.0, "learning_rate": 3.504322308315022e-05, "loss": 0.894, "step": 184770 }, { "epoch": 16.0, "learning_rate": 3.504105470105238e-05, "loss": 0.9266, "step": 184780 }, { "epoch": 16.0, "learning_rate": 3.503888631895455e-05, "loss": 1.0, "step": 184790 }, { "epoch": 16.0, "learning_rate": 3.503671793685671e-05, "loss": 0.8534, "step": 184800 }, { "epoch": 16.01, "learning_rate": 3.5034549554758874e-05, "loss": 0.8282, "step": 184810 }, { "epoch": 16.01, "learning_rate": 3.5032381172661034e-05, "loss": 0.6687, "step": 184820 }, { "epoch": 16.01, "learning_rate": 3.50302127905632e-05, "loss": 0.7199, "step": 184830 }, { "epoch": 16.01, "learning_rate": 3.502804440846536e-05, "loss": 0.7838, "step": 184840 }, { "epoch": 16.01, "learning_rate": 3.5025876026367526e-05, "loss": 0.8446, "step": 184850 }, { "epoch": 16.01, "learning_rate": 3.5023707644269686e-05, "loss": 0.8378, "step": 184860 }, { "epoch": 16.01, "learning_rate": 3.5021539262171846e-05, "loss": 0.8524, "step": 184870 }, { "epoch": 16.01, "learning_rate": 3.501937088007401e-05, "loss": 0.899, "step": 184880 }, { "epoch": 16.01, "learning_rate": 3.501720249797618e-05, "loss": 0.8077, "step": 184890 }, { "epoch": 16.01, "learning_rate": 3.501503411587834e-05, "loss": 1.0064, "step": 184900 }, { "epoch": 16.01, "learning_rate": 3.50128657337805e-05, "loss": 0.6353, "step": 184910 }, { "epoch": 16.02, "learning_rate": 3.5010697351682664e-05, "loss": 0.8796, "step": 184920 }, { "epoch": 16.02, "learning_rate": 3.5008528969584824e-05, "loss": 0.7827, "step": 184930 }, { "epoch": 16.02, "learning_rate": 3.500636058748698e-05, "loss": 0.8555, "step": 184940 }, { "epoch": 16.02, "learning_rate": 3.500419220538915e-05, "loss": 0.8168, "step": 184950 }, { "epoch": 16.02, "learning_rate": 3.5002023823291316e-05, "loss": 0.7801, "step": 184960 }, { "epoch": 16.02, "learning_rate": 3.4999855441193476e-05, "loss": 0.9746, "step": 184970 }, { "epoch": 16.02, "learning_rate": 3.4997687059095635e-05, "loss": 0.7041, "step": 184980 }, { "epoch": 16.02, "learning_rate": 3.49955186769978e-05, "loss": 0.7824, "step": 184990 }, { "epoch": 16.02, "learning_rate": 3.499335029489996e-05, "loss": 0.8104, "step": 185000 }, { "epoch": 16.02, "learning_rate": 3.499118191280212e-05, "loss": 0.896, "step": 185010 }, { "epoch": 16.02, "learning_rate": 3.498901353070429e-05, "loss": 0.8792, "step": 185020 }, { "epoch": 16.02, "learning_rate": 3.498684514860645e-05, "loss": 0.8739, "step": 185030 }, { "epoch": 16.03, "learning_rate": 3.498467676650861e-05, "loss": 0.8107, "step": 185040 }, { "epoch": 16.03, "learning_rate": 3.498250838441077e-05, "loss": 0.9421, "step": 185050 }, { "epoch": 16.03, "learning_rate": 3.498034000231294e-05, "loss": 0.67, "step": 185060 }, { "epoch": 16.03, "learning_rate": 3.49781716202151e-05, "loss": 0.9239, "step": 185070 }, { "epoch": 16.03, "learning_rate": 3.4976003238117265e-05, "loss": 0.8405, "step": 185080 }, { "epoch": 16.03, "learning_rate": 3.4973834856019425e-05, "loss": 0.7087, "step": 185090 }, { "epoch": 16.03, "learning_rate": 3.4971666473921584e-05, "loss": 0.8442, "step": 185100 }, { "epoch": 16.03, "learning_rate": 3.496949809182375e-05, "loss": 0.8303, "step": 185110 }, { "epoch": 16.03, "learning_rate": 3.496732970972592e-05, "loss": 0.6439, "step": 185120 }, { "epoch": 16.03, "learning_rate": 3.496516132762808e-05, "loss": 0.8361, "step": 185130 }, { "epoch": 16.03, "learning_rate": 3.4962992945530236e-05, "loss": 0.9901, "step": 185140 }, { "epoch": 16.04, "learning_rate": 3.49608245634324e-05, "loss": 0.8063, "step": 185150 }, { "epoch": 16.04, "learning_rate": 3.495865618133456e-05, "loss": 0.9156, "step": 185160 }, { "epoch": 16.04, "learning_rate": 3.495648779923672e-05, "loss": 0.7379, "step": 185170 }, { "epoch": 16.04, "learning_rate": 3.495431941713889e-05, "loss": 0.7919, "step": 185180 }, { "epoch": 16.04, "learning_rate": 3.4952151035041055e-05, "loss": 0.8966, "step": 185190 }, { "epoch": 16.04, "learning_rate": 3.4949982652943215e-05, "loss": 0.8563, "step": 185200 }, { "epoch": 16.04, "learning_rate": 3.4947814270845374e-05, "loss": 0.8525, "step": 185210 }, { "epoch": 16.04, "learning_rate": 3.494564588874754e-05, "loss": 0.9141, "step": 185220 }, { "epoch": 16.04, "learning_rate": 3.49434775066497e-05, "loss": 0.7372, "step": 185230 }, { "epoch": 16.04, "learning_rate": 3.4941309124551867e-05, "loss": 0.8474, "step": 185240 }, { "epoch": 16.04, "learning_rate": 3.4939140742454026e-05, "loss": 0.8473, "step": 185250 }, { "epoch": 16.04, "learning_rate": 3.493697236035619e-05, "loss": 0.8956, "step": 185260 }, { "epoch": 16.05, "learning_rate": 3.493480397825835e-05, "loss": 0.9887, "step": 185270 }, { "epoch": 16.05, "learning_rate": 3.493263559616051e-05, "loss": 0.7481, "step": 185280 }, { "epoch": 16.05, "learning_rate": 3.493046721406268e-05, "loss": 0.8993, "step": 185290 }, { "epoch": 16.05, "learning_rate": 3.492829883196484e-05, "loss": 0.789, "step": 185300 }, { "epoch": 16.05, "learning_rate": 3.4926130449867004e-05, "loss": 0.9588, "step": 185310 }, { "epoch": 16.05, "learning_rate": 3.4923962067769164e-05, "loss": 0.7586, "step": 185320 }, { "epoch": 16.05, "learning_rate": 3.492179368567133e-05, "loss": 0.8411, "step": 185330 }, { "epoch": 16.05, "learning_rate": 3.491962530357349e-05, "loss": 0.8642, "step": 185340 }, { "epoch": 16.05, "learning_rate": 3.4917456921475656e-05, "loss": 0.8138, "step": 185350 }, { "epoch": 16.05, "learning_rate": 3.4915288539377816e-05, "loss": 1.0376, "step": 185360 }, { "epoch": 16.05, "learning_rate": 3.4913120157279975e-05, "loss": 0.8965, "step": 185370 }, { "epoch": 16.06, "learning_rate": 3.491095177518214e-05, "loss": 0.7927, "step": 185380 }, { "epoch": 16.06, "learning_rate": 3.490878339308431e-05, "loss": 0.955, "step": 185390 }, { "epoch": 16.06, "learning_rate": 3.490661501098647e-05, "loss": 0.7401, "step": 185400 }, { "epoch": 16.06, "learning_rate": 3.490444662888863e-05, "loss": 0.9234, "step": 185410 }, { "epoch": 16.06, "learning_rate": 3.4902278246790794e-05, "loss": 0.939, "step": 185420 }, { "epoch": 16.06, "learning_rate": 3.4900109864692953e-05, "loss": 0.7079, "step": 185430 }, { "epoch": 16.06, "learning_rate": 3.489794148259511e-05, "loss": 0.9792, "step": 185440 }, { "epoch": 16.06, "learning_rate": 3.489577310049728e-05, "loss": 0.7367, "step": 185450 }, { "epoch": 16.06, "learning_rate": 3.4893604718399446e-05, "loss": 0.9165, "step": 185460 }, { "epoch": 16.06, "learning_rate": 3.4891436336301605e-05, "loss": 0.7441, "step": 185470 }, { "epoch": 16.06, "learning_rate": 3.4889267954203765e-05, "loss": 0.7315, "step": 185480 }, { "epoch": 16.06, "learning_rate": 3.488709957210593e-05, "loss": 0.799, "step": 185490 }, { "epoch": 16.07, "learning_rate": 3.488493119000809e-05, "loss": 0.8391, "step": 185500 }, { "epoch": 16.07, "learning_rate": 3.488276280791025e-05, "loss": 0.9404, "step": 185510 }, { "epoch": 16.07, "learning_rate": 3.488059442581242e-05, "loss": 1.1195, "step": 185520 }, { "epoch": 16.07, "learning_rate": 3.4878426043714583e-05, "loss": 0.8573, "step": 185530 }, { "epoch": 16.07, "learning_rate": 3.487625766161674e-05, "loss": 0.703, "step": 185540 }, { "epoch": 16.07, "learning_rate": 3.48740892795189e-05, "loss": 0.8103, "step": 185550 }, { "epoch": 16.07, "learning_rate": 3.487192089742107e-05, "loss": 1.0063, "step": 185560 }, { "epoch": 16.07, "learning_rate": 3.486975251532323e-05, "loss": 0.8114, "step": 185570 }, { "epoch": 16.07, "learning_rate": 3.4867584133225395e-05, "loss": 1.0385, "step": 185580 }, { "epoch": 16.07, "learning_rate": 3.4865415751127555e-05, "loss": 0.9122, "step": 185590 }, { "epoch": 16.07, "learning_rate": 3.4863247369029714e-05, "loss": 0.8846, "step": 185600 }, { "epoch": 16.08, "learning_rate": 3.486107898693188e-05, "loss": 1.0704, "step": 185610 }, { "epoch": 16.08, "learning_rate": 3.485891060483405e-05, "loss": 0.8003, "step": 185620 }, { "epoch": 16.08, "learning_rate": 3.485674222273621e-05, "loss": 1.0377, "step": 185630 }, { "epoch": 16.08, "learning_rate": 3.4854573840638366e-05, "loss": 0.8782, "step": 185640 }, { "epoch": 16.08, "learning_rate": 3.485240545854053e-05, "loss": 0.8113, "step": 185650 }, { "epoch": 16.08, "learning_rate": 3.485023707644269e-05, "loss": 0.7541, "step": 185660 }, { "epoch": 16.08, "learning_rate": 3.484806869434485e-05, "loss": 0.8793, "step": 185670 }, { "epoch": 16.08, "learning_rate": 3.484590031224702e-05, "loss": 0.7983, "step": 185680 }, { "epoch": 16.08, "learning_rate": 3.4843731930149185e-05, "loss": 0.8778, "step": 185690 }, { "epoch": 16.08, "learning_rate": 3.4841563548051344e-05, "loss": 0.6765, "step": 185700 }, { "epoch": 16.08, "learning_rate": 3.4839395165953504e-05, "loss": 0.9267, "step": 185710 }, { "epoch": 16.08, "learning_rate": 3.483722678385567e-05, "loss": 0.9877, "step": 185720 }, { "epoch": 16.09, "learning_rate": 3.483505840175783e-05, "loss": 0.7652, "step": 185730 }, { "epoch": 16.09, "learning_rate": 3.4832890019659996e-05, "loss": 0.9743, "step": 185740 }, { "epoch": 16.09, "learning_rate": 3.4830721637562156e-05, "loss": 0.8642, "step": 185750 }, { "epoch": 16.09, "learning_rate": 3.482855325546432e-05, "loss": 0.9064, "step": 185760 }, { "epoch": 16.09, "learning_rate": 3.482638487336648e-05, "loss": 0.7852, "step": 185770 }, { "epoch": 16.09, "learning_rate": 3.482421649126864e-05, "loss": 0.7567, "step": 185780 }, { "epoch": 16.09, "learning_rate": 3.482204810917081e-05, "loss": 0.8668, "step": 185790 }, { "epoch": 16.09, "learning_rate": 3.481987972707297e-05, "loss": 0.8509, "step": 185800 }, { "epoch": 16.09, "learning_rate": 3.4817711344975134e-05, "loss": 0.9033, "step": 185810 }, { "epoch": 16.09, "learning_rate": 3.4815542962877294e-05, "loss": 0.8789, "step": 185820 }, { "epoch": 16.09, "learning_rate": 3.481337458077946e-05, "loss": 0.8218, "step": 185830 }, { "epoch": 16.09, "learning_rate": 3.481120619868162e-05, "loss": 0.7019, "step": 185840 }, { "epoch": 16.1, "learning_rate": 3.4809037816583786e-05, "loss": 0.7804, "step": 185850 }, { "epoch": 16.1, "learning_rate": 3.4806869434485946e-05, "loss": 0.8617, "step": 185860 }, { "epoch": 16.1, "learning_rate": 3.4804701052388105e-05, "loss": 0.8913, "step": 185870 }, { "epoch": 16.1, "learning_rate": 3.480253267029027e-05, "loss": 0.7911, "step": 185880 }, { "epoch": 16.1, "learning_rate": 3.480036428819244e-05, "loss": 0.7729, "step": 185890 }, { "epoch": 16.1, "learning_rate": 3.47981959060946e-05, "loss": 0.8177, "step": 185900 }, { "epoch": 16.1, "learning_rate": 3.479602752399676e-05, "loss": 0.9014, "step": 185910 }, { "epoch": 16.1, "learning_rate": 3.4793859141898924e-05, "loss": 0.8849, "step": 185920 }, { "epoch": 16.1, "learning_rate": 3.479169075980108e-05, "loss": 0.8132, "step": 185930 }, { "epoch": 16.1, "learning_rate": 3.478952237770324e-05, "loss": 0.9147, "step": 185940 }, { "epoch": 16.1, "learning_rate": 3.478735399560541e-05, "loss": 0.92, "step": 185950 }, { "epoch": 16.11, "learning_rate": 3.4785185613507576e-05, "loss": 1.0064, "step": 185960 }, { "epoch": 16.11, "learning_rate": 3.4783017231409735e-05, "loss": 0.9703, "step": 185970 }, { "epoch": 16.11, "learning_rate": 3.4780848849311895e-05, "loss": 0.8962, "step": 185980 }, { "epoch": 16.11, "learning_rate": 3.477868046721406e-05, "loss": 0.9034, "step": 185990 }, { "epoch": 16.11, "learning_rate": 3.477651208511622e-05, "loss": 0.9237, "step": 186000 }, { "epoch": 16.11, "learning_rate": 3.477434370301839e-05, "loss": 0.8138, "step": 186010 }, { "epoch": 16.11, "learning_rate": 3.477217532092055e-05, "loss": 0.8415, "step": 186020 }, { "epoch": 16.11, "learning_rate": 3.477000693882271e-05, "loss": 0.8716, "step": 186030 }, { "epoch": 16.11, "learning_rate": 3.476783855672487e-05, "loss": 0.9748, "step": 186040 }, { "epoch": 16.11, "learning_rate": 3.476567017462703e-05, "loss": 0.7028, "step": 186050 }, { "epoch": 16.11, "learning_rate": 3.47635017925292e-05, "loss": 1.021, "step": 186060 }, { "epoch": 16.11, "learning_rate": 3.476133341043136e-05, "loss": 0.8001, "step": 186070 }, { "epoch": 16.12, "learning_rate": 3.4759165028333525e-05, "loss": 0.8119, "step": 186080 }, { "epoch": 16.12, "learning_rate": 3.4756996646235685e-05, "loss": 0.7919, "step": 186090 }, { "epoch": 16.12, "learning_rate": 3.4754828264137844e-05, "loss": 0.73, "step": 186100 }, { "epoch": 16.12, "learning_rate": 3.475265988204001e-05, "loss": 0.9148, "step": 186110 }, { "epoch": 16.12, "learning_rate": 3.475049149994218e-05, "loss": 0.8876, "step": 186120 }, { "epoch": 16.12, "learning_rate": 3.4748323117844337e-05, "loss": 0.8158, "step": 186130 }, { "epoch": 16.12, "learning_rate": 3.4746154735746496e-05, "loss": 0.815, "step": 186140 }, { "epoch": 16.12, "learning_rate": 3.474398635364866e-05, "loss": 0.9431, "step": 186150 }, { "epoch": 16.12, "learning_rate": 3.474181797155082e-05, "loss": 0.8974, "step": 186160 }, { "epoch": 16.12, "learning_rate": 3.473964958945298e-05, "loss": 0.9452, "step": 186170 }, { "epoch": 16.12, "learning_rate": 3.473748120735515e-05, "loss": 0.7792, "step": 186180 }, { "epoch": 16.13, "learning_rate": 3.4735312825257315e-05, "loss": 0.9401, "step": 186190 }, { "epoch": 16.13, "learning_rate": 3.4733144443159474e-05, "loss": 0.8269, "step": 186200 }, { "epoch": 16.13, "learning_rate": 3.4730976061061634e-05, "loss": 0.8539, "step": 186210 }, { "epoch": 16.13, "learning_rate": 3.47288076789638e-05, "loss": 0.6936, "step": 186220 }, { "epoch": 16.13, "learning_rate": 3.472663929686596e-05, "loss": 0.7208, "step": 186230 }, { "epoch": 16.13, "learning_rate": 3.4724470914768126e-05, "loss": 0.994, "step": 186240 }, { "epoch": 16.13, "learning_rate": 3.4722302532670286e-05, "loss": 0.929, "step": 186250 }, { "epoch": 16.13, "learning_rate": 3.472013415057245e-05, "loss": 0.7786, "step": 186260 }, { "epoch": 16.13, "learning_rate": 3.471796576847461e-05, "loss": 0.835, "step": 186270 }, { "epoch": 16.13, "learning_rate": 3.471579738637677e-05, "loss": 0.8191, "step": 186280 }, { "epoch": 16.13, "learning_rate": 3.471362900427894e-05, "loss": 0.6961, "step": 186290 }, { "epoch": 16.13, "learning_rate": 3.47114606221811e-05, "loss": 0.8789, "step": 186300 }, { "epoch": 16.14, "learning_rate": 3.4709292240083264e-05, "loss": 0.8866, "step": 186310 }, { "epoch": 16.14, "learning_rate": 3.4707123857985423e-05, "loss": 0.9629, "step": 186320 }, { "epoch": 16.14, "learning_rate": 3.470495547588759e-05, "loss": 0.8914, "step": 186330 }, { "epoch": 16.14, "learning_rate": 3.470278709378975e-05, "loss": 0.8812, "step": 186340 }, { "epoch": 16.14, "learning_rate": 3.4700618711691916e-05, "loss": 0.9052, "step": 186350 }, { "epoch": 16.14, "learning_rate": 3.4698450329594075e-05, "loss": 0.6888, "step": 186360 }, { "epoch": 16.14, "learning_rate": 3.4696281947496235e-05, "loss": 0.9009, "step": 186370 }, { "epoch": 16.14, "learning_rate": 3.46941135653984e-05, "loss": 0.7129, "step": 186380 }, { "epoch": 16.14, "learning_rate": 3.469194518330057e-05, "loss": 0.6597, "step": 186390 }, { "epoch": 16.14, "learning_rate": 3.468977680120273e-05, "loss": 0.7954, "step": 186400 }, { "epoch": 16.14, "learning_rate": 3.468760841910489e-05, "loss": 0.6848, "step": 186410 }, { "epoch": 16.15, "learning_rate": 3.4685440037007054e-05, "loss": 0.9141, "step": 186420 }, { "epoch": 16.15, "learning_rate": 3.468327165490921e-05, "loss": 0.8482, "step": 186430 }, { "epoch": 16.15, "learning_rate": 3.468110327281137e-05, "loss": 0.9862, "step": 186440 }, { "epoch": 16.15, "learning_rate": 3.467893489071354e-05, "loss": 0.9174, "step": 186450 }, { "epoch": 16.15, "learning_rate": 3.4676766508615706e-05, "loss": 0.806, "step": 186460 }, { "epoch": 16.15, "learning_rate": 3.4674598126517865e-05, "loss": 0.728, "step": 186470 }, { "epoch": 16.15, "learning_rate": 3.4672429744420025e-05, "loss": 0.8184, "step": 186480 }, { "epoch": 16.15, "learning_rate": 3.467026136232219e-05, "loss": 0.9178, "step": 186490 }, { "epoch": 16.15, "learning_rate": 3.466809298022435e-05, "loss": 0.9709, "step": 186500 }, { "epoch": 16.15, "learning_rate": 3.466592459812652e-05, "loss": 0.9301, "step": 186510 }, { "epoch": 16.15, "learning_rate": 3.466375621602868e-05, "loss": 0.8728, "step": 186520 }, { "epoch": 16.15, "learning_rate": 3.466158783393084e-05, "loss": 0.756, "step": 186530 }, { "epoch": 16.16, "learning_rate": 3.4659419451833e-05, "loss": 0.8049, "step": 186540 }, { "epoch": 16.16, "learning_rate": 3.465725106973516e-05, "loss": 0.9562, "step": 186550 }, { "epoch": 16.16, "learning_rate": 3.465508268763733e-05, "loss": 0.8516, "step": 186560 }, { "epoch": 16.16, "learning_rate": 3.465291430553949e-05, "loss": 0.7651, "step": 186570 }, { "epoch": 16.16, "learning_rate": 3.4650745923441655e-05, "loss": 0.7548, "step": 186580 }, { "epoch": 16.16, "learning_rate": 3.4648577541343814e-05, "loss": 0.8579, "step": 186590 }, { "epoch": 16.16, "learning_rate": 3.464640915924598e-05, "loss": 0.7912, "step": 186600 }, { "epoch": 16.16, "learning_rate": 3.464424077714814e-05, "loss": 0.9455, "step": 186610 }, { "epoch": 16.16, "learning_rate": 3.464207239505031e-05, "loss": 0.7774, "step": 186620 }, { "epoch": 16.16, "learning_rate": 3.4639904012952466e-05, "loss": 0.8697, "step": 186630 }, { "epoch": 16.16, "learning_rate": 3.4637735630854626e-05, "loss": 0.8074, "step": 186640 }, { "epoch": 16.17, "learning_rate": 3.463556724875679e-05, "loss": 0.8181, "step": 186650 }, { "epoch": 16.17, "learning_rate": 3.463339886665895e-05, "loss": 0.8007, "step": 186660 }, { "epoch": 16.17, "learning_rate": 3.463123048456111e-05, "loss": 0.8871, "step": 186670 }, { "epoch": 16.17, "learning_rate": 3.462906210246328e-05, "loss": 0.8206, "step": 186680 }, { "epoch": 16.17, "learning_rate": 3.4626893720365444e-05, "loss": 0.8143, "step": 186690 }, { "epoch": 16.17, "learning_rate": 3.4624725338267604e-05, "loss": 0.9171, "step": 186700 }, { "epoch": 16.17, "learning_rate": 3.4622556956169764e-05, "loss": 0.8514, "step": 186710 }, { "epoch": 16.17, "learning_rate": 3.462038857407193e-05, "loss": 0.6754, "step": 186720 }, { "epoch": 16.17, "learning_rate": 3.461822019197409e-05, "loss": 0.7173, "step": 186730 }, { "epoch": 16.17, "learning_rate": 3.4616051809876256e-05, "loss": 0.8939, "step": 186740 }, { "epoch": 16.17, "learning_rate": 3.4613883427778416e-05, "loss": 0.8874, "step": 186750 }, { "epoch": 16.17, "learning_rate": 3.461171504568058e-05, "loss": 0.9442, "step": 186760 }, { "epoch": 16.18, "learning_rate": 3.460954666358274e-05, "loss": 0.866, "step": 186770 }, { "epoch": 16.18, "learning_rate": 3.460737828148491e-05, "loss": 0.8925, "step": 186780 }, { "epoch": 16.18, "learning_rate": 3.460520989938707e-05, "loss": 0.7973, "step": 186790 }, { "epoch": 16.18, "learning_rate": 3.460304151728923e-05, "loss": 0.6797, "step": 186800 }, { "epoch": 16.18, "learning_rate": 3.4600873135191394e-05, "loss": 0.8937, "step": 186810 }, { "epoch": 16.18, "learning_rate": 3.459870475309355e-05, "loss": 0.8688, "step": 186820 }, { "epoch": 16.18, "learning_rate": 3.459653637099572e-05, "loss": 0.879, "step": 186830 }, { "epoch": 16.18, "learning_rate": 3.459436798889788e-05, "loss": 0.8359, "step": 186840 }, { "epoch": 16.18, "learning_rate": 3.4592199606800046e-05, "loss": 0.7814, "step": 186850 }, { "epoch": 16.18, "learning_rate": 3.4590031224702205e-05, "loss": 0.9891, "step": 186860 }, { "epoch": 16.18, "learning_rate": 3.4587862842604365e-05, "loss": 0.7934, "step": 186870 }, { "epoch": 16.18, "learning_rate": 3.458569446050653e-05, "loss": 0.7236, "step": 186880 }, { "epoch": 16.19, "learning_rate": 3.45835260784087e-05, "loss": 0.9078, "step": 186890 }, { "epoch": 16.19, "learning_rate": 3.458135769631086e-05, "loss": 0.775, "step": 186900 }, { "epoch": 16.19, "learning_rate": 3.457918931421302e-05, "loss": 0.8633, "step": 186910 }, { "epoch": 16.19, "learning_rate": 3.457702093211518e-05, "loss": 0.828, "step": 186920 }, { "epoch": 16.19, "learning_rate": 3.457485255001734e-05, "loss": 0.9314, "step": 186930 }, { "epoch": 16.19, "learning_rate": 3.45726841679195e-05, "loss": 0.7756, "step": 186940 }, { "epoch": 16.19, "learning_rate": 3.457051578582167e-05, "loss": 0.9877, "step": 186950 }, { "epoch": 16.19, "learning_rate": 3.4568347403723835e-05, "loss": 0.819, "step": 186960 }, { "epoch": 16.19, "learning_rate": 3.4566179021625995e-05, "loss": 0.7654, "step": 186970 }, { "epoch": 16.19, "learning_rate": 3.4564010639528155e-05, "loss": 0.9514, "step": 186980 }, { "epoch": 16.19, "learning_rate": 3.456184225743032e-05, "loss": 0.9751, "step": 186990 }, { "epoch": 16.2, "learning_rate": 3.455967387533248e-05, "loss": 0.9559, "step": 187000 }, { "epoch": 16.2, "learning_rate": 3.455750549323465e-05, "loss": 0.7845, "step": 187010 }, { "epoch": 16.2, "learning_rate": 3.455533711113681e-05, "loss": 0.7762, "step": 187020 }, { "epoch": 16.2, "learning_rate": 3.455316872903897e-05, "loss": 0.6674, "step": 187030 }, { "epoch": 16.2, "learning_rate": 3.455100034694113e-05, "loss": 0.8163, "step": 187040 }, { "epoch": 16.2, "learning_rate": 3.454883196484329e-05, "loss": 0.9366, "step": 187050 }, { "epoch": 16.2, "learning_rate": 3.454666358274546e-05, "loss": 0.8541, "step": 187060 }, { "epoch": 16.2, "learning_rate": 3.454449520064762e-05, "loss": 0.9169, "step": 187070 }, { "epoch": 16.2, "learning_rate": 3.4542326818549785e-05, "loss": 0.795, "step": 187080 }, { "epoch": 16.2, "learning_rate": 3.4540158436451944e-05, "loss": 0.8025, "step": 187090 }, { "epoch": 16.2, "learning_rate": 3.453799005435411e-05, "loss": 0.8799, "step": 187100 }, { "epoch": 16.2, "learning_rate": 3.453582167225627e-05, "loss": 0.9946, "step": 187110 }, { "epoch": 16.21, "learning_rate": 3.453365329015844e-05, "loss": 0.8209, "step": 187120 }, { "epoch": 16.21, "learning_rate": 3.4531484908060596e-05, "loss": 0.7226, "step": 187130 }, { "epoch": 16.21, "learning_rate": 3.4529316525962756e-05, "loss": 1.0114, "step": 187140 }, { "epoch": 16.21, "learning_rate": 3.452714814386492e-05, "loss": 0.7348, "step": 187150 }, { "epoch": 16.21, "learning_rate": 3.452497976176709e-05, "loss": 0.8098, "step": 187160 }, { "epoch": 16.21, "learning_rate": 3.452281137966924e-05, "loss": 0.6694, "step": 187170 }, { "epoch": 16.21, "learning_rate": 3.452064299757141e-05, "loss": 0.8424, "step": 187180 }, { "epoch": 16.21, "learning_rate": 3.4518474615473574e-05, "loss": 0.7106, "step": 187190 }, { "epoch": 16.21, "learning_rate": 3.4516306233375734e-05, "loss": 0.8237, "step": 187200 }, { "epoch": 16.21, "learning_rate": 3.4514137851277894e-05, "loss": 0.8793, "step": 187210 }, { "epoch": 16.21, "learning_rate": 3.451196946918006e-05, "loss": 0.7679, "step": 187220 }, { "epoch": 16.22, "learning_rate": 3.450980108708222e-05, "loss": 0.8016, "step": 187230 }, { "epoch": 16.22, "learning_rate": 3.4507632704984386e-05, "loss": 0.8663, "step": 187240 }, { "epoch": 16.22, "learning_rate": 3.4505464322886546e-05, "loss": 0.8285, "step": 187250 }, { "epoch": 16.22, "learning_rate": 3.450329594078871e-05, "loss": 0.8799, "step": 187260 }, { "epoch": 16.22, "learning_rate": 3.450112755869087e-05, "loss": 0.8095, "step": 187270 }, { "epoch": 16.22, "learning_rate": 3.449895917659304e-05, "loss": 0.8472, "step": 187280 }, { "epoch": 16.22, "learning_rate": 3.44967907944952e-05, "loss": 0.8697, "step": 187290 }, { "epoch": 16.22, "learning_rate": 3.449462241239736e-05, "loss": 0.8528, "step": 187300 }, { "epoch": 16.22, "learning_rate": 3.4492454030299524e-05, "loss": 0.9597, "step": 187310 }, { "epoch": 16.22, "learning_rate": 3.449028564820168e-05, "loss": 0.7951, "step": 187320 }, { "epoch": 16.22, "learning_rate": 3.448811726610385e-05, "loss": 0.6846, "step": 187330 }, { "epoch": 16.22, "learning_rate": 3.448594888400601e-05, "loss": 0.8502, "step": 187340 }, { "epoch": 16.23, "learning_rate": 3.4483780501908176e-05, "loss": 0.9028, "step": 187350 }, { "epoch": 16.23, "learning_rate": 3.4481612119810335e-05, "loss": 0.8718, "step": 187360 }, { "epoch": 16.23, "learning_rate": 3.4479443737712495e-05, "loss": 0.8726, "step": 187370 }, { "epoch": 16.23, "learning_rate": 3.447727535561466e-05, "loss": 0.8539, "step": 187380 }, { "epoch": 16.23, "learning_rate": 3.447510697351683e-05, "loss": 0.8505, "step": 187390 }, { "epoch": 16.23, "learning_rate": 3.447293859141899e-05, "loss": 0.9482, "step": 187400 }, { "epoch": 16.23, "learning_rate": 3.447077020932115e-05, "loss": 0.9621, "step": 187410 }, { "epoch": 16.23, "learning_rate": 3.446860182722331e-05, "loss": 0.7244, "step": 187420 }, { "epoch": 16.23, "learning_rate": 3.446643344512547e-05, "loss": 0.9292, "step": 187430 }, { "epoch": 16.23, "learning_rate": 3.446426506302763e-05, "loss": 0.8173, "step": 187440 }, { "epoch": 16.23, "learning_rate": 3.44620966809298e-05, "loss": 0.8912, "step": 187450 }, { "epoch": 16.24, "learning_rate": 3.4459928298831965e-05, "loss": 0.8932, "step": 187460 }, { "epoch": 16.24, "learning_rate": 3.4457759916734125e-05, "loss": 0.9459, "step": 187470 }, { "epoch": 16.24, "learning_rate": 3.4455591534636284e-05, "loss": 0.9745, "step": 187480 }, { "epoch": 16.24, "learning_rate": 3.445342315253845e-05, "loss": 0.9336, "step": 187490 }, { "epoch": 16.24, "learning_rate": 3.445125477044061e-05, "loss": 0.7994, "step": 187500 }, { "epoch": 16.24, "learning_rate": 3.444908638834278e-05, "loss": 0.8355, "step": 187510 }, { "epoch": 16.24, "learning_rate": 3.4446918006244936e-05, "loss": 0.7932, "step": 187520 }, { "epoch": 16.24, "learning_rate": 3.44447496241471e-05, "loss": 0.9069, "step": 187530 }, { "epoch": 16.24, "learning_rate": 3.444258124204926e-05, "loss": 0.7868, "step": 187540 }, { "epoch": 16.24, "learning_rate": 3.444041285995143e-05, "loss": 1.0442, "step": 187550 }, { "epoch": 16.24, "learning_rate": 3.443824447785359e-05, "loss": 0.7503, "step": 187560 }, { "epoch": 16.24, "learning_rate": 3.443607609575575e-05, "loss": 0.8358, "step": 187570 }, { "epoch": 16.25, "learning_rate": 3.4433907713657914e-05, "loss": 0.7583, "step": 187580 }, { "epoch": 16.25, "learning_rate": 3.4431739331560074e-05, "loss": 0.919, "step": 187590 }, { "epoch": 16.25, "learning_rate": 3.442957094946224e-05, "loss": 0.804, "step": 187600 }, { "epoch": 16.25, "learning_rate": 3.44274025673644e-05, "loss": 0.8564, "step": 187610 }, { "epoch": 16.25, "learning_rate": 3.4425234185266566e-05, "loss": 0.879, "step": 187620 }, { "epoch": 16.25, "learning_rate": 3.4423065803168726e-05, "loss": 0.8877, "step": 187630 }, { "epoch": 16.25, "learning_rate": 3.4420897421070886e-05, "loss": 0.8022, "step": 187640 }, { "epoch": 16.25, "learning_rate": 3.441872903897305e-05, "loss": 0.6592, "step": 187650 }, { "epoch": 16.25, "learning_rate": 3.441656065687522e-05, "loss": 0.834, "step": 187660 }, { "epoch": 16.25, "learning_rate": 3.441439227477738e-05, "loss": 1.033, "step": 187670 }, { "epoch": 16.25, "learning_rate": 3.441222389267954e-05, "loss": 0.8394, "step": 187680 }, { "epoch": 16.26, "learning_rate": 3.4410055510581704e-05, "loss": 0.8613, "step": 187690 }, { "epoch": 16.26, "learning_rate": 3.4407887128483864e-05, "loss": 0.7997, "step": 187700 }, { "epoch": 16.26, "learning_rate": 3.440571874638602e-05, "loss": 0.9408, "step": 187710 }, { "epoch": 16.26, "learning_rate": 3.440355036428819e-05, "loss": 0.713, "step": 187720 }, { "epoch": 16.26, "learning_rate": 3.440138198219035e-05, "loss": 0.899, "step": 187730 }, { "epoch": 16.26, "learning_rate": 3.4399213600092516e-05, "loss": 0.7889, "step": 187740 }, { "epoch": 16.26, "learning_rate": 3.4397045217994675e-05, "loss": 1.0184, "step": 187750 }, { "epoch": 16.26, "learning_rate": 3.439487683589684e-05, "loss": 0.9426, "step": 187760 }, { "epoch": 16.26, "learning_rate": 3.4392708453799e-05, "loss": 0.9352, "step": 187770 }, { "epoch": 16.26, "learning_rate": 3.439054007170117e-05, "loss": 1.0068, "step": 187780 }, { "epoch": 16.26, "learning_rate": 3.438837168960333e-05, "loss": 0.8905, "step": 187790 }, { "epoch": 16.26, "learning_rate": 3.438620330750549e-05, "loss": 0.9498, "step": 187800 }, { "epoch": 16.27, "learning_rate": 3.4384034925407653e-05, "loss": 0.8759, "step": 187810 }, { "epoch": 16.27, "learning_rate": 3.438186654330981e-05, "loss": 0.8757, "step": 187820 }, { "epoch": 16.27, "learning_rate": 3.437969816121198e-05, "loss": 0.8934, "step": 187830 }, { "epoch": 16.27, "learning_rate": 3.437752977911414e-05, "loss": 0.8784, "step": 187840 }, { "epoch": 16.27, "learning_rate": 3.4375361397016305e-05, "loss": 0.8544, "step": 187850 }, { "epoch": 16.27, "learning_rate": 3.4373193014918465e-05, "loss": 0.8287, "step": 187860 }, { "epoch": 16.27, "learning_rate": 3.4371024632820625e-05, "loss": 0.8998, "step": 187870 }, { "epoch": 16.27, "learning_rate": 3.436885625072279e-05, "loss": 0.861, "step": 187880 }, { "epoch": 16.27, "learning_rate": 3.436668786862496e-05, "loss": 0.9396, "step": 187890 }, { "epoch": 16.27, "learning_rate": 3.436451948652712e-05, "loss": 0.8775, "step": 187900 }, { "epoch": 16.27, "learning_rate": 3.436235110442928e-05, "loss": 0.9111, "step": 187910 }, { "epoch": 16.28, "learning_rate": 3.436018272233144e-05, "loss": 0.8119, "step": 187920 }, { "epoch": 16.28, "learning_rate": 3.43580143402336e-05, "loss": 0.8579, "step": 187930 }, { "epoch": 16.28, "learning_rate": 3.435584595813576e-05, "loss": 0.7487, "step": 187940 }, { "epoch": 16.28, "learning_rate": 3.435367757603793e-05, "loss": 1.0778, "step": 187950 }, { "epoch": 16.28, "learning_rate": 3.4351509193940095e-05, "loss": 0.8819, "step": 187960 }, { "epoch": 16.28, "learning_rate": 3.4349340811842255e-05, "loss": 0.8581, "step": 187970 }, { "epoch": 16.28, "learning_rate": 3.4347172429744414e-05, "loss": 0.9054, "step": 187980 }, { "epoch": 16.28, "learning_rate": 3.434500404764658e-05, "loss": 0.7739, "step": 187990 }, { "epoch": 16.28, "learning_rate": 3.434283566554874e-05, "loss": 0.9046, "step": 188000 }, { "epoch": 16.28, "learning_rate": 3.434066728345091e-05, "loss": 0.9385, "step": 188010 }, { "epoch": 16.28, "learning_rate": 3.4338498901353066e-05, "loss": 0.6517, "step": 188020 }, { "epoch": 16.28, "learning_rate": 3.433633051925523e-05, "loss": 0.845, "step": 188030 }, { "epoch": 16.29, "learning_rate": 3.433416213715739e-05, "loss": 0.9693, "step": 188040 }, { "epoch": 16.29, "learning_rate": 3.433199375505956e-05, "loss": 0.867, "step": 188050 }, { "epoch": 16.29, "learning_rate": 3.432982537296172e-05, "loss": 0.9999, "step": 188060 }, { "epoch": 16.29, "learning_rate": 3.432765699086388e-05, "loss": 0.9053, "step": 188070 }, { "epoch": 16.29, "learning_rate": 3.4325488608766044e-05, "loss": 0.7302, "step": 188080 }, { "epoch": 16.29, "learning_rate": 3.4323320226668204e-05, "loss": 0.8929, "step": 188090 }, { "epoch": 16.29, "learning_rate": 3.432115184457037e-05, "loss": 0.7849, "step": 188100 }, { "epoch": 16.29, "learning_rate": 3.431898346247253e-05, "loss": 0.9238, "step": 188110 }, { "epoch": 16.29, "learning_rate": 3.4316815080374696e-05, "loss": 1.0236, "step": 188120 }, { "epoch": 16.29, "learning_rate": 3.4314646698276856e-05, "loss": 0.8318, "step": 188130 }, { "epoch": 16.29, "learning_rate": 3.4312478316179016e-05, "loss": 0.9687, "step": 188140 }, { "epoch": 16.29, "learning_rate": 3.431030993408118e-05, "loss": 0.8789, "step": 188150 }, { "epoch": 16.3, "learning_rate": 3.430814155198335e-05, "loss": 0.9431, "step": 188160 }, { "epoch": 16.3, "learning_rate": 3.430597316988551e-05, "loss": 0.9618, "step": 188170 }, { "epoch": 16.3, "learning_rate": 3.430380478778767e-05, "loss": 0.8137, "step": 188180 }, { "epoch": 16.3, "learning_rate": 3.4301636405689834e-05, "loss": 0.8486, "step": 188190 }, { "epoch": 16.3, "learning_rate": 3.4299468023591994e-05, "loss": 0.9897, "step": 188200 }, { "epoch": 16.3, "learning_rate": 3.429729964149415e-05, "loss": 0.8118, "step": 188210 }, { "epoch": 16.3, "learning_rate": 3.429513125939632e-05, "loss": 0.9268, "step": 188220 }, { "epoch": 16.3, "learning_rate": 3.4292962877298486e-05, "loss": 0.8303, "step": 188230 }, { "epoch": 16.3, "learning_rate": 3.4290794495200646e-05, "loss": 0.8469, "step": 188240 }, { "epoch": 16.3, "learning_rate": 3.4288626113102805e-05, "loss": 0.8978, "step": 188250 }, { "epoch": 16.3, "learning_rate": 3.428645773100497e-05, "loss": 0.8481, "step": 188260 }, { "epoch": 16.31, "learning_rate": 3.428428934890713e-05, "loss": 0.8052, "step": 188270 }, { "epoch": 16.31, "learning_rate": 3.42821209668093e-05, "loss": 0.8299, "step": 188280 }, { "epoch": 16.31, "learning_rate": 3.427995258471146e-05, "loss": 1.0965, "step": 188290 }, { "epoch": 16.31, "learning_rate": 3.427778420261362e-05, "loss": 0.6693, "step": 188300 }, { "epoch": 16.31, "learning_rate": 3.427561582051578e-05, "loss": 0.7453, "step": 188310 }, { "epoch": 16.31, "learning_rate": 3.427344743841795e-05, "loss": 0.9475, "step": 188320 }, { "epoch": 16.31, "learning_rate": 3.427127905632011e-05, "loss": 0.8252, "step": 188330 }, { "epoch": 16.31, "learning_rate": 3.426911067422227e-05, "loss": 0.8473, "step": 188340 }, { "epoch": 16.31, "learning_rate": 3.4266942292124435e-05, "loss": 1.0292, "step": 188350 }, { "epoch": 16.31, "learning_rate": 3.4264773910026595e-05, "loss": 0.9596, "step": 188360 }, { "epoch": 16.31, "learning_rate": 3.4262605527928754e-05, "loss": 0.8248, "step": 188370 }, { "epoch": 16.31, "learning_rate": 3.426043714583092e-05, "loss": 0.9057, "step": 188380 }, { "epoch": 16.32, "learning_rate": 3.425826876373309e-05, "loss": 0.8148, "step": 188390 }, { "epoch": 16.32, "learning_rate": 3.425610038163525e-05, "loss": 0.7552, "step": 188400 }, { "epoch": 16.32, "learning_rate": 3.4253931999537407e-05, "loss": 0.9397, "step": 188410 }, { "epoch": 16.32, "learning_rate": 3.425176361743957e-05, "loss": 0.7261, "step": 188420 }, { "epoch": 16.32, "learning_rate": 3.424959523534173e-05, "loss": 0.8868, "step": 188430 }, { "epoch": 16.32, "learning_rate": 3.424742685324389e-05, "loss": 0.9366, "step": 188440 }, { "epoch": 16.32, "learning_rate": 3.424525847114606e-05, "loss": 0.9223, "step": 188450 }, { "epoch": 16.32, "learning_rate": 3.4243090089048225e-05, "loss": 0.8706, "step": 188460 }, { "epoch": 16.32, "learning_rate": 3.4240921706950385e-05, "loss": 0.6989, "step": 188470 }, { "epoch": 16.32, "learning_rate": 3.4238753324852544e-05, "loss": 0.7231, "step": 188480 }, { "epoch": 16.32, "learning_rate": 3.423658494275471e-05, "loss": 0.9245, "step": 188490 }, { "epoch": 16.33, "learning_rate": 3.423441656065687e-05, "loss": 0.8431, "step": 188500 }, { "epoch": 16.33, "learning_rate": 3.4232248178559037e-05, "loss": 1.0339, "step": 188510 }, { "epoch": 16.33, "learning_rate": 3.4230079796461196e-05, "loss": 0.7198, "step": 188520 }, { "epoch": 16.33, "learning_rate": 3.422791141436336e-05, "loss": 0.9332, "step": 188530 }, { "epoch": 16.33, "learning_rate": 3.422574303226552e-05, "loss": 0.8095, "step": 188540 }, { "epoch": 16.33, "learning_rate": 3.422357465016769e-05, "loss": 0.9739, "step": 188550 }, { "epoch": 16.33, "learning_rate": 3.422140626806985e-05, "loss": 1.0242, "step": 188560 }, { "epoch": 16.33, "learning_rate": 3.421923788597201e-05, "loss": 0.831, "step": 188570 }, { "epoch": 16.33, "learning_rate": 3.4217069503874174e-05, "loss": 0.7606, "step": 188580 }, { "epoch": 16.33, "learning_rate": 3.4214901121776334e-05, "loss": 0.7748, "step": 188590 }, { "epoch": 16.33, "learning_rate": 3.42127327396785e-05, "loss": 0.8409, "step": 188600 }, { "epoch": 16.33, "learning_rate": 3.421056435758066e-05, "loss": 0.8837, "step": 188610 }, { "epoch": 16.34, "learning_rate": 3.4208395975482826e-05, "loss": 0.9031, "step": 188620 }, { "epoch": 16.34, "learning_rate": 3.4206227593384986e-05, "loss": 0.999, "step": 188630 }, { "epoch": 16.34, "learning_rate": 3.4204059211287145e-05, "loss": 0.9266, "step": 188640 }, { "epoch": 16.34, "learning_rate": 3.420189082918931e-05, "loss": 0.8672, "step": 188650 }, { "epoch": 16.34, "learning_rate": 3.419972244709148e-05, "loss": 0.8826, "step": 188660 }, { "epoch": 16.34, "learning_rate": 3.419755406499364e-05, "loss": 0.883, "step": 188670 }, { "epoch": 16.34, "learning_rate": 3.41953856828958e-05, "loss": 0.7269, "step": 188680 }, { "epoch": 16.34, "learning_rate": 3.4193217300797964e-05, "loss": 0.8478, "step": 188690 }, { "epoch": 16.34, "learning_rate": 3.4191048918700123e-05, "loss": 0.8241, "step": 188700 }, { "epoch": 16.34, "learning_rate": 3.418888053660228e-05, "loss": 0.8841, "step": 188710 }, { "epoch": 16.34, "learning_rate": 3.418671215450445e-05, "loss": 0.9366, "step": 188720 }, { "epoch": 16.35, "learning_rate": 3.4184543772406616e-05, "loss": 0.8772, "step": 188730 }, { "epoch": 16.35, "learning_rate": 3.4182375390308775e-05, "loss": 0.8163, "step": 188740 }, { "epoch": 16.35, "learning_rate": 3.4180207008210935e-05, "loss": 0.7519, "step": 188750 }, { "epoch": 16.35, "learning_rate": 3.41780386261131e-05, "loss": 0.8972, "step": 188760 }, { "epoch": 16.35, "learning_rate": 3.417587024401526e-05, "loss": 0.7554, "step": 188770 }, { "epoch": 16.35, "learning_rate": 3.417370186191743e-05, "loss": 0.9563, "step": 188780 }, { "epoch": 16.35, "learning_rate": 3.417153347981959e-05, "loss": 0.8059, "step": 188790 }, { "epoch": 16.35, "learning_rate": 3.416936509772175e-05, "loss": 0.8729, "step": 188800 }, { "epoch": 16.35, "learning_rate": 3.416719671562391e-05, "loss": 0.8132, "step": 188810 }, { "epoch": 16.35, "learning_rate": 3.416502833352608e-05, "loss": 0.9904, "step": 188820 }, { "epoch": 16.35, "learning_rate": 3.416285995142824e-05, "loss": 0.8904, "step": 188830 }, { "epoch": 16.35, "learning_rate": 3.41606915693304e-05, "loss": 0.9623, "step": 188840 }, { "epoch": 16.36, "learning_rate": 3.4158523187232565e-05, "loss": 0.8302, "step": 188850 }, { "epoch": 16.36, "learning_rate": 3.4156354805134725e-05, "loss": 0.8813, "step": 188860 }, { "epoch": 16.36, "learning_rate": 3.4154186423036884e-05, "loss": 0.8703, "step": 188870 }, { "epoch": 16.36, "learning_rate": 3.415201804093905e-05, "loss": 1.0011, "step": 188880 }, { "epoch": 16.36, "learning_rate": 3.414984965884122e-05, "loss": 0.8105, "step": 188890 }, { "epoch": 16.36, "learning_rate": 3.414768127674338e-05, "loss": 0.8101, "step": 188900 }, { "epoch": 16.36, "learning_rate": 3.4145512894645536e-05, "loss": 0.9722, "step": 188910 }, { "epoch": 16.36, "learning_rate": 3.41433445125477e-05, "loss": 0.8514, "step": 188920 }, { "epoch": 16.36, "learning_rate": 3.414117613044986e-05, "loss": 0.7564, "step": 188930 }, { "epoch": 16.36, "learning_rate": 3.413900774835202e-05, "loss": 0.8857, "step": 188940 }, { "epoch": 16.36, "learning_rate": 3.413683936625419e-05, "loss": 0.9336, "step": 188950 }, { "epoch": 16.37, "learning_rate": 3.4134670984156355e-05, "loss": 0.8354, "step": 188960 }, { "epoch": 16.37, "learning_rate": 3.4132502602058514e-05, "loss": 0.7855, "step": 188970 }, { "epoch": 16.37, "learning_rate": 3.4130334219960674e-05, "loss": 0.8363, "step": 188980 }, { "epoch": 16.37, "learning_rate": 3.412816583786284e-05, "loss": 0.9954, "step": 188990 }, { "epoch": 16.37, "learning_rate": 3.4125997455765e-05, "loss": 0.8056, "step": 189000 }, { "epoch": 16.37, "learning_rate": 3.4123829073667166e-05, "loss": 0.7405, "step": 189010 }, { "epoch": 16.37, "learning_rate": 3.4121660691569326e-05, "loss": 0.7263, "step": 189020 }, { "epoch": 16.37, "learning_rate": 3.411949230947149e-05, "loss": 0.7842, "step": 189030 }, { "epoch": 16.37, "learning_rate": 3.411732392737365e-05, "loss": 0.8842, "step": 189040 }, { "epoch": 16.37, "learning_rate": 3.411515554527582e-05, "loss": 0.7513, "step": 189050 }, { "epoch": 16.37, "learning_rate": 3.411298716317798e-05, "loss": 1.0235, "step": 189060 }, { "epoch": 16.37, "learning_rate": 3.411081878108014e-05, "loss": 0.8936, "step": 189070 }, { "epoch": 16.38, "learning_rate": 3.4108650398982304e-05, "loss": 0.8836, "step": 189080 }, { "epoch": 16.38, "learning_rate": 3.410648201688447e-05, "loss": 0.9036, "step": 189090 }, { "epoch": 16.38, "learning_rate": 3.410431363478663e-05, "loss": 0.7601, "step": 189100 }, { "epoch": 16.38, "learning_rate": 3.410214525268879e-05, "loss": 0.6881, "step": 189110 }, { "epoch": 16.38, "learning_rate": 3.4099976870590956e-05, "loss": 0.9882, "step": 189120 }, { "epoch": 16.38, "learning_rate": 3.4097808488493116e-05, "loss": 0.7256, "step": 189130 }, { "epoch": 16.38, "learning_rate": 3.4095640106395275e-05, "loss": 0.9851, "step": 189140 }, { "epoch": 16.38, "learning_rate": 3.409347172429744e-05, "loss": 0.9083, "step": 189150 }, { "epoch": 16.38, "learning_rate": 3.409130334219961e-05, "loss": 0.7914, "step": 189160 }, { "epoch": 16.38, "learning_rate": 3.408913496010177e-05, "loss": 1.1081, "step": 189170 }, { "epoch": 16.38, "learning_rate": 3.408696657800393e-05, "loss": 0.7852, "step": 189180 }, { "epoch": 16.39, "learning_rate": 3.4084798195906094e-05, "loss": 0.8545, "step": 189190 }, { "epoch": 16.39, "learning_rate": 3.408262981380825e-05, "loss": 0.9583, "step": 189200 }, { "epoch": 16.39, "learning_rate": 3.408046143171041e-05, "loss": 0.8444, "step": 189210 }, { "epoch": 16.39, "learning_rate": 3.407829304961258e-05, "loss": 0.7208, "step": 189220 }, { "epoch": 16.39, "learning_rate": 3.4076124667514746e-05, "loss": 0.8847, "step": 189230 }, { "epoch": 16.39, "learning_rate": 3.4073956285416905e-05, "loss": 0.6774, "step": 189240 }, { "epoch": 16.39, "learning_rate": 3.4071787903319065e-05, "loss": 0.9276, "step": 189250 }, { "epoch": 16.39, "learning_rate": 3.406961952122123e-05, "loss": 1.0041, "step": 189260 }, { "epoch": 16.39, "learning_rate": 3.406745113912339e-05, "loss": 0.9686, "step": 189270 }, { "epoch": 16.39, "learning_rate": 3.406528275702556e-05, "loss": 0.7281, "step": 189280 }, { "epoch": 16.39, "learning_rate": 3.406311437492772e-05, "loss": 0.9145, "step": 189290 }, { "epoch": 16.39, "learning_rate": 3.406094599282988e-05, "loss": 0.8769, "step": 189300 }, { "epoch": 16.4, "learning_rate": 3.405877761073204e-05, "loss": 0.7098, "step": 189310 }, { "epoch": 16.4, "learning_rate": 3.405660922863421e-05, "loss": 0.8964, "step": 189320 }, { "epoch": 16.4, "learning_rate": 3.405444084653637e-05, "loss": 0.8179, "step": 189330 }, { "epoch": 16.4, "learning_rate": 3.405227246443853e-05, "loss": 1.0365, "step": 189340 }, { "epoch": 16.4, "learning_rate": 3.4050104082340695e-05, "loss": 0.8422, "step": 189350 }, { "epoch": 16.4, "learning_rate": 3.4047935700242855e-05, "loss": 0.8628, "step": 189360 }, { "epoch": 16.4, "learning_rate": 3.4045767318145014e-05, "loss": 0.8083, "step": 189370 }, { "epoch": 16.4, "learning_rate": 3.404359893604718e-05, "loss": 0.8056, "step": 189380 }, { "epoch": 16.4, "learning_rate": 3.404143055394935e-05, "loss": 0.7028, "step": 189390 }, { "epoch": 16.4, "learning_rate": 3.4039262171851507e-05, "loss": 0.7599, "step": 189400 }, { "epoch": 16.4, "learning_rate": 3.4037093789753666e-05, "loss": 0.9701, "step": 189410 }, { "epoch": 16.4, "learning_rate": 3.403492540765583e-05, "loss": 0.8896, "step": 189420 }, { "epoch": 16.41, "learning_rate": 3.403275702555799e-05, "loss": 0.8724, "step": 189430 }, { "epoch": 16.41, "learning_rate": 3.403058864346015e-05, "loss": 0.9041, "step": 189440 }, { "epoch": 16.41, "learning_rate": 3.402842026136232e-05, "loss": 0.9783, "step": 189450 }, { "epoch": 16.41, "learning_rate": 3.4026251879264485e-05, "loss": 0.7968, "step": 189460 }, { "epoch": 16.41, "learning_rate": 3.4024083497166644e-05, "loss": 0.9791, "step": 189470 }, { "epoch": 16.41, "learning_rate": 3.4021915115068804e-05, "loss": 0.9461, "step": 189480 }, { "epoch": 16.41, "learning_rate": 3.401974673297097e-05, "loss": 0.7433, "step": 189490 }, { "epoch": 16.41, "learning_rate": 3.401757835087313e-05, "loss": 0.8044, "step": 189500 }, { "epoch": 16.41, "learning_rate": 3.4015409968775296e-05, "loss": 0.8243, "step": 189510 }, { "epoch": 16.41, "learning_rate": 3.4013241586677456e-05, "loss": 0.7795, "step": 189520 }, { "epoch": 16.41, "learning_rate": 3.401107320457962e-05, "loss": 0.9106, "step": 189530 }, { "epoch": 16.42, "learning_rate": 3.400890482248178e-05, "loss": 0.7274, "step": 189540 }, { "epoch": 16.42, "learning_rate": 3.400673644038395e-05, "loss": 0.96, "step": 189550 }, { "epoch": 16.42, "learning_rate": 3.400456805828611e-05, "loss": 1.0537, "step": 189560 }, { "epoch": 16.42, "learning_rate": 3.400239967618827e-05, "loss": 0.7274, "step": 189570 }, { "epoch": 16.42, "learning_rate": 3.4000231294090434e-05, "loss": 0.7524, "step": 189580 }, { "epoch": 16.42, "learning_rate": 3.39980629119926e-05, "loss": 0.7969, "step": 189590 }, { "epoch": 16.42, "learning_rate": 3.399589452989476e-05, "loss": 0.9589, "step": 189600 }, { "epoch": 16.42, "learning_rate": 3.399372614779692e-05, "loss": 0.8576, "step": 189610 }, { "epoch": 16.42, "learning_rate": 3.3991557765699086e-05, "loss": 0.8901, "step": 189620 }, { "epoch": 16.42, "learning_rate": 3.3989389383601246e-05, "loss": 0.7007, "step": 189630 }, { "epoch": 16.42, "learning_rate": 3.3987221001503405e-05, "loss": 0.9562, "step": 189640 }, { "epoch": 16.42, "learning_rate": 3.398505261940557e-05, "loss": 0.8592, "step": 189650 }, { "epoch": 16.43, "learning_rate": 3.398288423730774e-05, "loss": 0.7219, "step": 189660 }, { "epoch": 16.43, "learning_rate": 3.39807158552099e-05, "loss": 0.8749, "step": 189670 }, { "epoch": 16.43, "learning_rate": 3.397854747311206e-05, "loss": 0.8553, "step": 189680 }, { "epoch": 16.43, "learning_rate": 3.3976379091014224e-05, "loss": 0.8711, "step": 189690 }, { "epoch": 16.43, "learning_rate": 3.397421070891638e-05, "loss": 0.9529, "step": 189700 }, { "epoch": 16.43, "learning_rate": 3.397204232681854e-05, "loss": 0.9383, "step": 189710 }, { "epoch": 16.43, "learning_rate": 3.396987394472071e-05, "loss": 0.8117, "step": 189720 }, { "epoch": 16.43, "learning_rate": 3.3967705562622876e-05, "loss": 0.8476, "step": 189730 }, { "epoch": 16.43, "learning_rate": 3.3965537180525035e-05, "loss": 1.0039, "step": 189740 }, { "epoch": 16.43, "learning_rate": 3.3963368798427195e-05, "loss": 0.9989, "step": 189750 }, { "epoch": 16.43, "learning_rate": 3.396120041632936e-05, "loss": 0.8478, "step": 189760 }, { "epoch": 16.44, "learning_rate": 3.395903203423152e-05, "loss": 0.8959, "step": 189770 }, { "epoch": 16.44, "learning_rate": 3.395686365213369e-05, "loss": 0.982, "step": 189780 }, { "epoch": 16.44, "learning_rate": 3.395469527003585e-05, "loss": 1.0035, "step": 189790 }, { "epoch": 16.44, "learning_rate": 3.395252688793801e-05, "loss": 0.9012, "step": 189800 }, { "epoch": 16.44, "learning_rate": 3.395035850584017e-05, "loss": 0.7995, "step": 189810 }, { "epoch": 16.44, "learning_rate": 3.394819012374234e-05, "loss": 0.779, "step": 189820 }, { "epoch": 16.44, "learning_rate": 3.39460217416445e-05, "loss": 0.6512, "step": 189830 }, { "epoch": 16.44, "learning_rate": 3.394385335954666e-05, "loss": 0.9363, "step": 189840 }, { "epoch": 16.44, "learning_rate": 3.3941684977448825e-05, "loss": 0.77, "step": 189850 }, { "epoch": 16.44, "learning_rate": 3.393951659535099e-05, "loss": 0.9758, "step": 189860 }, { "epoch": 16.44, "learning_rate": 3.3937348213253144e-05, "loss": 0.8531, "step": 189870 }, { "epoch": 16.44, "learning_rate": 3.393517983115531e-05, "loss": 0.8726, "step": 189880 }, { "epoch": 16.45, "learning_rate": 3.393301144905748e-05, "loss": 0.8098, "step": 189890 }, { "epoch": 16.45, "learning_rate": 3.3930843066959636e-05, "loss": 0.9208, "step": 189900 }, { "epoch": 16.45, "learning_rate": 3.3928674684861796e-05, "loss": 0.8654, "step": 189910 }, { "epoch": 16.45, "learning_rate": 3.392650630276396e-05, "loss": 0.8707, "step": 189920 }, { "epoch": 16.45, "learning_rate": 3.392433792066612e-05, "loss": 0.8438, "step": 189930 }, { "epoch": 16.45, "learning_rate": 3.392216953856828e-05, "loss": 0.77, "step": 189940 }, { "epoch": 16.45, "learning_rate": 3.392000115647045e-05, "loss": 0.7388, "step": 189950 }, { "epoch": 16.45, "learning_rate": 3.3917832774372614e-05, "loss": 0.9328, "step": 189960 }, { "epoch": 16.45, "learning_rate": 3.3915664392274774e-05, "loss": 0.7405, "step": 189970 }, { "epoch": 16.45, "learning_rate": 3.3913496010176934e-05, "loss": 0.7852, "step": 189980 }, { "epoch": 16.45, "learning_rate": 3.39113276280791e-05, "loss": 0.9198, "step": 189990 }, { "epoch": 16.46, "learning_rate": 3.390915924598126e-05, "loss": 0.8293, "step": 190000 }, { "epoch": 16.46, "learning_rate": 3.3906990863883426e-05, "loss": 0.8016, "step": 190010 }, { "epoch": 16.46, "learning_rate": 3.3904822481785586e-05, "loss": 0.9144, "step": 190020 }, { "epoch": 16.46, "learning_rate": 3.390265409968775e-05, "loss": 1.0034, "step": 190030 }, { "epoch": 16.46, "learning_rate": 3.390048571758991e-05, "loss": 0.7336, "step": 190040 }, { "epoch": 16.46, "learning_rate": 3.389831733549208e-05, "loss": 1.0548, "step": 190050 }, { "epoch": 16.46, "learning_rate": 3.389614895339424e-05, "loss": 0.9071, "step": 190060 }, { "epoch": 16.46, "learning_rate": 3.38939805712964e-05, "loss": 0.9512, "step": 190070 }, { "epoch": 16.46, "learning_rate": 3.3891812189198564e-05, "loss": 0.9253, "step": 190080 }, { "epoch": 16.46, "learning_rate": 3.388964380710073e-05, "loss": 0.9655, "step": 190090 }, { "epoch": 16.46, "learning_rate": 3.388747542500289e-05, "loss": 0.7439, "step": 190100 }, { "epoch": 16.46, "learning_rate": 3.388530704290505e-05, "loss": 0.8194, "step": 190110 }, { "epoch": 16.47, "learning_rate": 3.3883138660807216e-05, "loss": 0.9693, "step": 190120 }, { "epoch": 16.47, "learning_rate": 3.3880970278709375e-05, "loss": 1.0213, "step": 190130 }, { "epoch": 16.47, "learning_rate": 3.3878801896611535e-05, "loss": 0.9038, "step": 190140 }, { "epoch": 16.47, "learning_rate": 3.38766335145137e-05, "loss": 0.8314, "step": 190150 }, { "epoch": 16.47, "learning_rate": 3.387446513241587e-05, "loss": 0.8338, "step": 190160 }, { "epoch": 16.47, "learning_rate": 3.387229675031803e-05, "loss": 0.8757, "step": 190170 }, { "epoch": 16.47, "learning_rate": 3.387012836822019e-05, "loss": 0.9266, "step": 190180 }, { "epoch": 16.47, "learning_rate": 3.386795998612235e-05, "loss": 0.8646, "step": 190190 }, { "epoch": 16.47, "learning_rate": 3.386579160402451e-05, "loss": 0.816, "step": 190200 }, { "epoch": 16.47, "learning_rate": 3.386362322192667e-05, "loss": 0.7679, "step": 190210 }, { "epoch": 16.47, "learning_rate": 3.386145483982884e-05, "loss": 1.0035, "step": 190220 }, { "epoch": 16.48, "learning_rate": 3.3859286457731005e-05, "loss": 0.8859, "step": 190230 }, { "epoch": 16.48, "learning_rate": 3.3857118075633165e-05, "loss": 0.7858, "step": 190240 }, { "epoch": 16.48, "learning_rate": 3.3854949693535325e-05, "loss": 0.8617, "step": 190250 }, { "epoch": 16.48, "learning_rate": 3.385278131143749e-05, "loss": 0.9016, "step": 190260 }, { "epoch": 16.48, "learning_rate": 3.385061292933965e-05, "loss": 0.6255, "step": 190270 }, { "epoch": 16.48, "learning_rate": 3.384844454724182e-05, "loss": 0.9099, "step": 190280 }, { "epoch": 16.48, "learning_rate": 3.384627616514398e-05, "loss": 0.8201, "step": 190290 }, { "epoch": 16.48, "learning_rate": 3.384410778304614e-05, "loss": 0.8893, "step": 190300 }, { "epoch": 16.48, "learning_rate": 3.38419394009483e-05, "loss": 0.926, "step": 190310 }, { "epoch": 16.48, "learning_rate": 3.383977101885047e-05, "loss": 0.8245, "step": 190320 }, { "epoch": 16.48, "learning_rate": 3.383760263675263e-05, "loss": 0.8559, "step": 190330 }, { "epoch": 16.48, "learning_rate": 3.383543425465479e-05, "loss": 0.955, "step": 190340 }, { "epoch": 16.49, "learning_rate": 3.3833265872556955e-05, "loss": 1.0309, "step": 190350 }, { "epoch": 16.49, "learning_rate": 3.383109749045912e-05, "loss": 0.7689, "step": 190360 }, { "epoch": 16.49, "learning_rate": 3.382892910836128e-05, "loss": 0.8859, "step": 190370 }, { "epoch": 16.49, "learning_rate": 3.382676072626344e-05, "loss": 0.8894, "step": 190380 }, { "epoch": 16.49, "learning_rate": 3.382459234416561e-05, "loss": 0.8845, "step": 190390 }, { "epoch": 16.49, "learning_rate": 3.3822423962067766e-05, "loss": 0.8854, "step": 190400 }, { "epoch": 16.49, "learning_rate": 3.3820255579969926e-05, "loss": 0.9631, "step": 190410 }, { "epoch": 16.49, "learning_rate": 3.381808719787209e-05, "loss": 0.8302, "step": 190420 }, { "epoch": 16.49, "learning_rate": 3.381591881577425e-05, "loss": 0.87, "step": 190430 }, { "epoch": 16.49, "learning_rate": 3.381375043367641e-05, "loss": 0.7824, "step": 190440 }, { "epoch": 16.49, "learning_rate": 3.381158205157858e-05, "loss": 0.9028, "step": 190450 }, { "epoch": 16.5, "learning_rate": 3.3809413669480744e-05, "loss": 0.7264, "step": 190460 }, { "epoch": 16.5, "learning_rate": 3.3807245287382904e-05, "loss": 0.878, "step": 190470 }, { "epoch": 16.5, "learning_rate": 3.3805076905285064e-05, "loss": 0.947, "step": 190480 }, { "epoch": 16.5, "learning_rate": 3.380290852318723e-05, "loss": 0.892, "step": 190490 }, { "epoch": 16.5, "learning_rate": 3.380074014108939e-05, "loss": 0.7447, "step": 190500 }, { "epoch": 16.5, "learning_rate": 3.3798571758991556e-05, "loss": 0.8371, "step": 190510 }, { "epoch": 16.5, "learning_rate": 3.3796403376893716e-05, "loss": 0.7858, "step": 190520 }, { "epoch": 16.5, "learning_rate": 3.379423499479588e-05, "loss": 0.905, "step": 190530 }, { "epoch": 16.5, "learning_rate": 3.379206661269804e-05, "loss": 1.057, "step": 190540 }, { "epoch": 16.5, "learning_rate": 3.378989823060021e-05, "loss": 0.879, "step": 190550 }, { "epoch": 16.5, "learning_rate": 3.378772984850237e-05, "loss": 0.8937, "step": 190560 }, { "epoch": 16.5, "learning_rate": 3.378556146640453e-05, "loss": 0.7573, "step": 190570 }, { "epoch": 16.51, "learning_rate": 3.3783393084306694e-05, "loss": 0.8525, "step": 190580 }, { "epoch": 16.51, "learning_rate": 3.378122470220886e-05, "loss": 0.7272, "step": 190590 }, { "epoch": 16.51, "learning_rate": 3.377905632011102e-05, "loss": 0.9093, "step": 190600 }, { "epoch": 16.51, "learning_rate": 3.377688793801318e-05, "loss": 0.8748, "step": 190610 }, { "epoch": 16.51, "learning_rate": 3.3774719555915346e-05, "loss": 0.8747, "step": 190620 }, { "epoch": 16.51, "learning_rate": 3.3772551173817505e-05, "loss": 0.7252, "step": 190630 }, { "epoch": 16.51, "learning_rate": 3.3770382791719665e-05, "loss": 0.9365, "step": 190640 }, { "epoch": 16.51, "learning_rate": 3.376821440962183e-05, "loss": 0.8011, "step": 190650 }, { "epoch": 16.51, "learning_rate": 3.3766046027524e-05, "loss": 0.9017, "step": 190660 }, { "epoch": 16.51, "learning_rate": 3.376387764542616e-05, "loss": 0.8381, "step": 190670 }, { "epoch": 16.51, "learning_rate": 3.376170926332832e-05, "loss": 0.764, "step": 190680 }, { "epoch": 16.51, "learning_rate": 3.375954088123048e-05, "loss": 0.7721, "step": 190690 }, { "epoch": 16.52, "learning_rate": 3.375737249913264e-05, "loss": 0.7621, "step": 190700 }, { "epoch": 16.52, "learning_rate": 3.37552041170348e-05, "loss": 0.8857, "step": 190710 }, { "epoch": 16.52, "learning_rate": 3.375303573493697e-05, "loss": 0.8938, "step": 190720 }, { "epoch": 16.52, "learning_rate": 3.3750867352839135e-05, "loss": 0.9795, "step": 190730 }, { "epoch": 16.52, "learning_rate": 3.3748698970741295e-05, "loss": 0.788, "step": 190740 }, { "epoch": 16.52, "learning_rate": 3.3746530588643454e-05, "loss": 0.8215, "step": 190750 }, { "epoch": 16.52, "learning_rate": 3.374436220654562e-05, "loss": 0.8504, "step": 190760 }, { "epoch": 16.52, "learning_rate": 3.374219382444778e-05, "loss": 0.779, "step": 190770 }, { "epoch": 16.52, "learning_rate": 3.374002544234995e-05, "loss": 1.0156, "step": 190780 }, { "epoch": 16.52, "learning_rate": 3.3737857060252106e-05, "loss": 0.7542, "step": 190790 }, { "epoch": 16.52, "learning_rate": 3.373568867815427e-05, "loss": 0.8841, "step": 190800 }, { "epoch": 16.53, "learning_rate": 3.373352029605643e-05, "loss": 0.7962, "step": 190810 }, { "epoch": 16.53, "learning_rate": 3.37313519139586e-05, "loss": 0.8404, "step": 190820 }, { "epoch": 16.53, "learning_rate": 3.372918353186076e-05, "loss": 0.9578, "step": 190830 }, { "epoch": 16.53, "learning_rate": 3.372701514976292e-05, "loss": 0.7628, "step": 190840 }, { "epoch": 16.53, "learning_rate": 3.3724846767665084e-05, "loss": 0.9845, "step": 190850 }, { "epoch": 16.53, "learning_rate": 3.372267838556725e-05, "loss": 0.87, "step": 190860 }, { "epoch": 16.53, "learning_rate": 3.372051000346941e-05, "loss": 0.7234, "step": 190870 }, { "epoch": 16.53, "learning_rate": 3.371834162137157e-05, "loss": 1.002, "step": 190880 }, { "epoch": 16.53, "learning_rate": 3.3716173239273737e-05, "loss": 0.8974, "step": 190890 }, { "epoch": 16.53, "learning_rate": 3.3714004857175896e-05, "loss": 0.6851, "step": 190900 }, { "epoch": 16.53, "learning_rate": 3.3711836475078056e-05, "loss": 0.9391, "step": 190910 }, { "epoch": 16.53, "learning_rate": 3.370966809298022e-05, "loss": 1.095, "step": 190920 }, { "epoch": 16.54, "learning_rate": 3.370749971088239e-05, "loss": 0.7815, "step": 190930 }, { "epoch": 16.54, "learning_rate": 3.370533132878454e-05, "loss": 0.9555, "step": 190940 }, { "epoch": 16.54, "learning_rate": 3.370316294668671e-05, "loss": 0.752, "step": 190950 }, { "epoch": 16.54, "learning_rate": 3.3700994564588874e-05, "loss": 0.8924, "step": 190960 }, { "epoch": 16.54, "learning_rate": 3.3698826182491034e-05, "loss": 0.7691, "step": 190970 }, { "epoch": 16.54, "learning_rate": 3.369665780039319e-05, "loss": 0.801, "step": 190980 }, { "epoch": 16.54, "learning_rate": 3.369448941829536e-05, "loss": 0.8636, "step": 190990 }, { "epoch": 16.54, "learning_rate": 3.369232103619752e-05, "loss": 0.8511, "step": 191000 }, { "epoch": 16.54, "learning_rate": 3.3690152654099686e-05, "loss": 0.8854, "step": 191010 }, { "epoch": 16.54, "learning_rate": 3.3687984272001845e-05, "loss": 0.8867, "step": 191020 }, { "epoch": 16.54, "learning_rate": 3.368581588990401e-05, "loss": 0.9587, "step": 191030 }, { "epoch": 16.55, "learning_rate": 3.368364750780617e-05, "loss": 0.6647, "step": 191040 }, { "epoch": 16.55, "learning_rate": 3.368147912570834e-05, "loss": 0.8431, "step": 191050 }, { "epoch": 16.55, "learning_rate": 3.36793107436105e-05, "loss": 0.8863, "step": 191060 }, { "epoch": 16.55, "learning_rate": 3.367714236151266e-05, "loss": 0.9464, "step": 191070 }, { "epoch": 16.55, "learning_rate": 3.3674973979414823e-05, "loss": 0.7687, "step": 191080 }, { "epoch": 16.55, "learning_rate": 3.367280559731699e-05, "loss": 0.8781, "step": 191090 }, { "epoch": 16.55, "learning_rate": 3.367063721521915e-05, "loss": 0.9148, "step": 191100 }, { "epoch": 16.55, "learning_rate": 3.366846883312131e-05, "loss": 0.9176, "step": 191110 }, { "epoch": 16.55, "learning_rate": 3.3666300451023475e-05, "loss": 0.8599, "step": 191120 }, { "epoch": 16.55, "learning_rate": 3.3664132068925635e-05, "loss": 0.8451, "step": 191130 }, { "epoch": 16.55, "learning_rate": 3.3661963686827795e-05, "loss": 1.0057, "step": 191140 }, { "epoch": 16.55, "learning_rate": 3.365979530472996e-05, "loss": 0.8525, "step": 191150 }, { "epoch": 16.56, "learning_rate": 3.365762692263213e-05, "loss": 0.789, "step": 191160 }, { "epoch": 16.56, "learning_rate": 3.365545854053429e-05, "loss": 0.7799, "step": 191170 }, { "epoch": 16.56, "learning_rate": 3.365329015843645e-05, "loss": 0.7137, "step": 191180 }, { "epoch": 16.56, "learning_rate": 3.365112177633861e-05, "loss": 0.7692, "step": 191190 }, { "epoch": 16.56, "learning_rate": 3.364895339424077e-05, "loss": 0.8524, "step": 191200 }, { "epoch": 16.56, "learning_rate": 3.364678501214293e-05, "loss": 0.8084, "step": 191210 }, { "epoch": 16.56, "learning_rate": 3.36446166300451e-05, "loss": 0.9248, "step": 191220 }, { "epoch": 16.56, "learning_rate": 3.3642448247947265e-05, "loss": 0.972, "step": 191230 }, { "epoch": 16.56, "learning_rate": 3.3640279865849425e-05, "loss": 0.9994, "step": 191240 }, { "epoch": 16.56, "learning_rate": 3.3638111483751584e-05, "loss": 0.8072, "step": 191250 }, { "epoch": 16.56, "learning_rate": 3.363594310165375e-05, "loss": 0.866, "step": 191260 }, { "epoch": 16.57, "learning_rate": 3.363377471955591e-05, "loss": 1.0099, "step": 191270 }, { "epoch": 16.57, "learning_rate": 3.363160633745808e-05, "loss": 0.8056, "step": 191280 }, { "epoch": 16.57, "learning_rate": 3.3629437955360236e-05, "loss": 0.9183, "step": 191290 }, { "epoch": 16.57, "learning_rate": 3.36272695732624e-05, "loss": 0.858, "step": 191300 }, { "epoch": 16.57, "learning_rate": 3.362510119116456e-05, "loss": 0.8673, "step": 191310 }, { "epoch": 16.57, "learning_rate": 3.362293280906673e-05, "loss": 0.7529, "step": 191320 }, { "epoch": 16.57, "learning_rate": 3.362076442696889e-05, "loss": 0.9406, "step": 191330 }, { "epoch": 16.57, "learning_rate": 3.361859604487105e-05, "loss": 0.9156, "step": 191340 }, { "epoch": 16.57, "learning_rate": 3.3616427662773214e-05, "loss": 0.8487, "step": 191350 }, { "epoch": 16.57, "learning_rate": 3.361425928067538e-05, "loss": 0.9891, "step": 191360 }, { "epoch": 16.57, "learning_rate": 3.361209089857754e-05, "loss": 0.8683, "step": 191370 }, { "epoch": 16.57, "learning_rate": 3.36099225164797e-05, "loss": 0.7919, "step": 191380 }, { "epoch": 16.58, "learning_rate": 3.3607754134381866e-05, "loss": 0.9987, "step": 191390 }, { "epoch": 16.58, "learning_rate": 3.3605585752284026e-05, "loss": 0.835, "step": 191400 }, { "epoch": 16.58, "learning_rate": 3.3603417370186186e-05, "loss": 0.9559, "step": 191410 }, { "epoch": 16.58, "learning_rate": 3.360124898808835e-05, "loss": 0.9015, "step": 191420 }, { "epoch": 16.58, "learning_rate": 3.359908060599052e-05, "loss": 0.8747, "step": 191430 }, { "epoch": 16.58, "learning_rate": 3.359691222389268e-05, "loss": 0.9445, "step": 191440 }, { "epoch": 16.58, "learning_rate": 3.359474384179484e-05, "loss": 1.1352, "step": 191450 }, { "epoch": 16.58, "learning_rate": 3.3592575459697004e-05, "loss": 1.1025, "step": 191460 }, { "epoch": 16.58, "learning_rate": 3.3590407077599164e-05, "loss": 0.859, "step": 191470 }, { "epoch": 16.58, "learning_rate": 3.358823869550132e-05, "loss": 0.861, "step": 191480 }, { "epoch": 16.58, "learning_rate": 3.358607031340349e-05, "loss": 0.7688, "step": 191490 }, { "epoch": 16.59, "learning_rate": 3.358390193130565e-05, "loss": 0.8126, "step": 191500 }, { "epoch": 16.59, "learning_rate": 3.3581733549207816e-05, "loss": 0.8096, "step": 191510 }, { "epoch": 16.59, "learning_rate": 3.3579565167109975e-05, "loss": 0.8632, "step": 191520 }, { "epoch": 16.59, "learning_rate": 3.357739678501214e-05, "loss": 0.9053, "step": 191530 }, { "epoch": 16.59, "learning_rate": 3.35752284029143e-05, "loss": 0.8521, "step": 191540 }, { "epoch": 16.59, "learning_rate": 3.357306002081647e-05, "loss": 0.8233, "step": 191550 }, { "epoch": 16.59, "learning_rate": 3.357089163871863e-05, "loss": 0.817, "step": 191560 }, { "epoch": 16.59, "learning_rate": 3.356872325662079e-05, "loss": 0.7799, "step": 191570 }, { "epoch": 16.59, "learning_rate": 3.356655487452295e-05, "loss": 0.8499, "step": 191580 }, { "epoch": 16.59, "learning_rate": 3.356438649242512e-05, "loss": 0.8273, "step": 191590 }, { "epoch": 16.59, "learning_rate": 3.356221811032728e-05, "loss": 0.7862, "step": 191600 }, { "epoch": 16.59, "learning_rate": 3.356004972822944e-05, "loss": 0.8391, "step": 191610 }, { "epoch": 16.6, "learning_rate": 3.3557881346131605e-05, "loss": 1.0312, "step": 191620 }, { "epoch": 16.6, "learning_rate": 3.3555712964033765e-05, "loss": 0.8251, "step": 191630 }, { "epoch": 16.6, "learning_rate": 3.3553544581935925e-05, "loss": 0.9462, "step": 191640 }, { "epoch": 16.6, "learning_rate": 3.355137619983809e-05, "loss": 0.9826, "step": 191650 }, { "epoch": 16.6, "learning_rate": 3.354920781774026e-05, "loss": 0.9911, "step": 191660 }, { "epoch": 16.6, "learning_rate": 3.354703943564242e-05, "loss": 0.9172, "step": 191670 }, { "epoch": 16.6, "learning_rate": 3.3544871053544577e-05, "loss": 0.8202, "step": 191680 }, { "epoch": 16.6, "learning_rate": 3.354270267144674e-05, "loss": 0.7847, "step": 191690 }, { "epoch": 16.6, "learning_rate": 3.35405342893489e-05, "loss": 0.8598, "step": 191700 }, { "epoch": 16.6, "learning_rate": 3.353836590725106e-05, "loss": 0.7485, "step": 191710 }, { "epoch": 16.6, "learning_rate": 3.353619752515323e-05, "loss": 0.7879, "step": 191720 }, { "epoch": 16.61, "learning_rate": 3.3534029143055395e-05, "loss": 0.9327, "step": 191730 }, { "epoch": 16.61, "learning_rate": 3.3531860760957555e-05, "loss": 1.0383, "step": 191740 }, { "epoch": 16.61, "learning_rate": 3.3529692378859714e-05, "loss": 0.9228, "step": 191750 }, { "epoch": 16.61, "learning_rate": 3.352752399676188e-05, "loss": 0.8303, "step": 191760 }, { "epoch": 16.61, "learning_rate": 3.352535561466404e-05, "loss": 0.8124, "step": 191770 }, { "epoch": 16.61, "learning_rate": 3.3523187232566207e-05, "loss": 0.8317, "step": 191780 }, { "epoch": 16.61, "learning_rate": 3.3521018850468366e-05, "loss": 0.8803, "step": 191790 }, { "epoch": 16.61, "learning_rate": 3.351885046837053e-05, "loss": 0.8813, "step": 191800 }, { "epoch": 16.61, "learning_rate": 3.351668208627269e-05, "loss": 0.7234, "step": 191810 }, { "epoch": 16.61, "learning_rate": 3.351451370417486e-05, "loss": 1.1094, "step": 191820 }, { "epoch": 16.61, "learning_rate": 3.351234532207702e-05, "loss": 0.8778, "step": 191830 }, { "epoch": 16.61, "learning_rate": 3.351017693997918e-05, "loss": 0.8649, "step": 191840 }, { "epoch": 16.62, "learning_rate": 3.3508008557881344e-05, "loss": 0.9791, "step": 191850 }, { "epoch": 16.62, "learning_rate": 3.350584017578351e-05, "loss": 0.9365, "step": 191860 }, { "epoch": 16.62, "learning_rate": 3.350367179368567e-05, "loss": 0.9693, "step": 191870 }, { "epoch": 16.62, "learning_rate": 3.350150341158783e-05, "loss": 0.8286, "step": 191880 }, { "epoch": 16.62, "learning_rate": 3.3499335029489996e-05, "loss": 0.7893, "step": 191890 }, { "epoch": 16.62, "learning_rate": 3.3497166647392156e-05, "loss": 1.0596, "step": 191900 }, { "epoch": 16.62, "learning_rate": 3.3494998265294315e-05, "loss": 0.9663, "step": 191910 }, { "epoch": 16.62, "learning_rate": 3.349282988319648e-05, "loss": 0.7693, "step": 191920 }, { "epoch": 16.62, "learning_rate": 3.349066150109865e-05, "loss": 0.9736, "step": 191930 }, { "epoch": 16.62, "learning_rate": 3.348849311900081e-05, "loss": 0.8524, "step": 191940 }, { "epoch": 16.62, "learning_rate": 3.348632473690297e-05, "loss": 0.9732, "step": 191950 }, { "epoch": 16.62, "learning_rate": 3.3484156354805134e-05, "loss": 0.8482, "step": 191960 }, { "epoch": 16.63, "learning_rate": 3.3481987972707293e-05, "loss": 1.0387, "step": 191970 }, { "epoch": 16.63, "learning_rate": 3.347981959060945e-05, "loss": 1.0557, "step": 191980 }, { "epoch": 16.63, "learning_rate": 3.347765120851162e-05, "loss": 0.8961, "step": 191990 }, { "epoch": 16.63, "learning_rate": 3.3475482826413786e-05, "loss": 0.8279, "step": 192000 }, { "epoch": 16.63, "learning_rate": 3.3473314444315945e-05, "loss": 1.0031, "step": 192010 }, { "epoch": 16.63, "learning_rate": 3.3471146062218105e-05, "loss": 0.8047, "step": 192020 }, { "epoch": 16.63, "learning_rate": 3.346897768012027e-05, "loss": 0.9733, "step": 192030 }, { "epoch": 16.63, "learning_rate": 3.346680929802243e-05, "loss": 0.9879, "step": 192040 }, { "epoch": 16.63, "learning_rate": 3.34646409159246e-05, "loss": 0.9087, "step": 192050 }, { "epoch": 16.63, "learning_rate": 3.346247253382676e-05, "loss": 1.2016, "step": 192060 }, { "epoch": 16.63, "learning_rate": 3.346030415172892e-05, "loss": 0.7791, "step": 192070 }, { "epoch": 16.64, "learning_rate": 3.345813576963108e-05, "loss": 0.7965, "step": 192080 }, { "epoch": 16.64, "learning_rate": 3.345596738753325e-05, "loss": 0.8902, "step": 192090 }, { "epoch": 16.64, "learning_rate": 3.345379900543541e-05, "loss": 0.7665, "step": 192100 }, { "epoch": 16.64, "learning_rate": 3.345163062333757e-05, "loss": 0.9569, "step": 192110 }, { "epoch": 16.64, "learning_rate": 3.3449462241239735e-05, "loss": 0.938, "step": 192120 }, { "epoch": 16.64, "learning_rate": 3.3447293859141895e-05, "loss": 0.7839, "step": 192130 }, { "epoch": 16.64, "learning_rate": 3.3445125477044054e-05, "loss": 0.7137, "step": 192140 }, { "epoch": 16.64, "learning_rate": 3.344295709494622e-05, "loss": 0.9351, "step": 192150 }, { "epoch": 16.64, "learning_rate": 3.344078871284839e-05, "loss": 1.1207, "step": 192160 }, { "epoch": 16.64, "learning_rate": 3.343862033075055e-05, "loss": 0.7968, "step": 192170 }, { "epoch": 16.64, "learning_rate": 3.3436451948652706e-05, "loss": 1.0923, "step": 192180 }, { "epoch": 16.64, "learning_rate": 3.343428356655487e-05, "loss": 0.8211, "step": 192190 }, { "epoch": 16.65, "learning_rate": 3.343211518445703e-05, "loss": 0.8931, "step": 192200 }, { "epoch": 16.65, "learning_rate": 3.342994680235919e-05, "loss": 0.8884, "step": 192210 }, { "epoch": 16.65, "learning_rate": 3.342777842026136e-05, "loss": 0.8561, "step": 192220 }, { "epoch": 16.65, "learning_rate": 3.3425610038163525e-05, "loss": 0.9639, "step": 192230 }, { "epoch": 16.65, "learning_rate": 3.3423441656065684e-05, "loss": 0.9063, "step": 192240 }, { "epoch": 16.65, "learning_rate": 3.3421273273967844e-05, "loss": 0.9499, "step": 192250 }, { "epoch": 16.65, "learning_rate": 3.341910489187001e-05, "loss": 1.059, "step": 192260 }, { "epoch": 16.65, "learning_rate": 3.341693650977217e-05, "loss": 0.8454, "step": 192270 }, { "epoch": 16.65, "learning_rate": 3.3414768127674336e-05, "loss": 0.8432, "step": 192280 }, { "epoch": 16.65, "learning_rate": 3.3412599745576496e-05, "loss": 0.9196, "step": 192290 }, { "epoch": 16.65, "learning_rate": 3.341043136347866e-05, "loss": 0.8431, "step": 192300 }, { "epoch": 16.66, "learning_rate": 3.340826298138082e-05, "loss": 0.7172, "step": 192310 }, { "epoch": 16.66, "learning_rate": 3.340609459928299e-05, "loss": 1.0428, "step": 192320 }, { "epoch": 16.66, "learning_rate": 3.340392621718515e-05, "loss": 0.9524, "step": 192330 }, { "epoch": 16.66, "learning_rate": 3.340175783508731e-05, "loss": 0.8247, "step": 192340 }, { "epoch": 16.66, "learning_rate": 3.3399589452989474e-05, "loss": 0.9465, "step": 192350 }, { "epoch": 16.66, "learning_rate": 3.339742107089164e-05, "loss": 0.7448, "step": 192360 }, { "epoch": 16.66, "learning_rate": 3.33952526887938e-05, "loss": 0.8815, "step": 192370 }, { "epoch": 16.66, "learning_rate": 3.339308430669596e-05, "loss": 0.7753, "step": 192380 }, { "epoch": 16.66, "learning_rate": 3.3390915924598126e-05, "loss": 0.9042, "step": 192390 }, { "epoch": 16.66, "learning_rate": 3.3388747542500286e-05, "loss": 0.8158, "step": 192400 }, { "epoch": 16.66, "learning_rate": 3.3386579160402445e-05, "loss": 0.7731, "step": 192410 }, { "epoch": 16.66, "learning_rate": 3.338441077830461e-05, "loss": 0.8713, "step": 192420 }, { "epoch": 16.67, "learning_rate": 3.338224239620678e-05, "loss": 0.8747, "step": 192430 }, { "epoch": 16.67, "learning_rate": 3.338007401410894e-05, "loss": 0.8096, "step": 192440 }, { "epoch": 16.67, "learning_rate": 3.33779056320111e-05, "loss": 1.0022, "step": 192450 }, { "epoch": 16.67, "learning_rate": 3.3375737249913264e-05, "loss": 0.8958, "step": 192460 }, { "epoch": 16.67, "learning_rate": 3.337356886781542e-05, "loss": 0.8999, "step": 192470 }, { "epoch": 16.67, "learning_rate": 3.337140048571758e-05, "loss": 0.7705, "step": 192480 }, { "epoch": 16.67, "learning_rate": 3.336923210361975e-05, "loss": 0.7284, "step": 192490 }, { "epoch": 16.67, "learning_rate": 3.3367063721521916e-05, "loss": 0.777, "step": 192500 }, { "epoch": 16.67, "learning_rate": 3.3364895339424075e-05, "loss": 0.8249, "step": 192510 }, { "epoch": 16.67, "learning_rate": 3.3362726957326235e-05, "loss": 0.7812, "step": 192520 }, { "epoch": 16.67, "learning_rate": 3.33605585752284e-05, "loss": 0.7914, "step": 192530 }, { "epoch": 16.68, "learning_rate": 3.335839019313056e-05, "loss": 0.9361, "step": 192540 }, { "epoch": 16.68, "learning_rate": 3.335622181103273e-05, "loss": 0.743, "step": 192550 }, { "epoch": 16.68, "learning_rate": 3.335405342893489e-05, "loss": 0.9665, "step": 192560 }, { "epoch": 16.68, "learning_rate": 3.3351885046837047e-05, "loss": 0.9639, "step": 192570 }, { "epoch": 16.68, "learning_rate": 3.334971666473921e-05, "loss": 0.9731, "step": 192580 }, { "epoch": 16.68, "learning_rate": 3.334754828264138e-05, "loss": 0.8292, "step": 192590 }, { "epoch": 16.68, "learning_rate": 3.334537990054354e-05, "loss": 0.8234, "step": 192600 }, { "epoch": 16.68, "learning_rate": 3.33432115184457e-05, "loss": 1.0276, "step": 192610 }, { "epoch": 16.68, "learning_rate": 3.3341043136347865e-05, "loss": 1.0696, "step": 192620 }, { "epoch": 16.68, "learning_rate": 3.3338874754250025e-05, "loss": 0.8036, "step": 192630 }, { "epoch": 16.68, "learning_rate": 3.3336706372152184e-05, "loss": 0.8887, "step": 192640 }, { "epoch": 16.68, "learning_rate": 3.333453799005435e-05, "loss": 0.9846, "step": 192650 }, { "epoch": 16.69, "learning_rate": 3.333236960795652e-05, "loss": 1.0582, "step": 192660 }, { "epoch": 16.69, "learning_rate": 3.3330201225858677e-05, "loss": 0.5601, "step": 192670 }, { "epoch": 16.69, "learning_rate": 3.3328032843760836e-05, "loss": 0.8289, "step": 192680 }, { "epoch": 16.69, "learning_rate": 3.3325864461663e-05, "loss": 0.7042, "step": 192690 }, { "epoch": 16.69, "learning_rate": 3.332369607956516e-05, "loss": 0.844, "step": 192700 }, { "epoch": 16.69, "learning_rate": 3.332152769746732e-05, "loss": 0.8925, "step": 192710 }, { "epoch": 16.69, "learning_rate": 3.331935931536949e-05, "loss": 0.8841, "step": 192720 }, { "epoch": 16.69, "learning_rate": 3.3317190933271655e-05, "loss": 0.9762, "step": 192730 }, { "epoch": 16.69, "learning_rate": 3.3315022551173814e-05, "loss": 0.9396, "step": 192740 }, { "epoch": 16.69, "learning_rate": 3.3312854169075974e-05, "loss": 0.7746, "step": 192750 }, { "epoch": 16.69, "learning_rate": 3.331068578697814e-05, "loss": 0.8292, "step": 192760 }, { "epoch": 16.7, "learning_rate": 3.33085174048803e-05, "loss": 1.0967, "step": 192770 }, { "epoch": 16.7, "learning_rate": 3.3306349022782466e-05, "loss": 0.8704, "step": 192780 }, { "epoch": 16.7, "learning_rate": 3.3304180640684626e-05, "loss": 0.9566, "step": 192790 }, { "epoch": 16.7, "learning_rate": 3.330201225858679e-05, "loss": 1.0594, "step": 192800 }, { "epoch": 16.7, "learning_rate": 3.329984387648895e-05, "loss": 0.7848, "step": 192810 }, { "epoch": 16.7, "learning_rate": 3.329767549439112e-05, "loss": 0.8922, "step": 192820 }, { "epoch": 16.7, "learning_rate": 3.329550711229328e-05, "loss": 0.8692, "step": 192830 }, { "epoch": 16.7, "learning_rate": 3.329333873019544e-05, "loss": 0.7235, "step": 192840 }, { "epoch": 16.7, "learning_rate": 3.3291170348097604e-05, "loss": 0.8077, "step": 192850 }, { "epoch": 16.7, "learning_rate": 3.328900196599977e-05, "loss": 0.9487, "step": 192860 }, { "epoch": 16.7, "learning_rate": 3.328683358390193e-05, "loss": 0.8832, "step": 192870 }, { "epoch": 16.7, "learning_rate": 3.328466520180409e-05, "loss": 0.8965, "step": 192880 }, { "epoch": 16.71, "learning_rate": 3.3282496819706256e-05, "loss": 1.0089, "step": 192890 }, { "epoch": 16.71, "learning_rate": 3.3280328437608416e-05, "loss": 0.8896, "step": 192900 }, { "epoch": 16.71, "learning_rate": 3.3278160055510575e-05, "loss": 0.8093, "step": 192910 }, { "epoch": 16.71, "learning_rate": 3.327599167341274e-05, "loss": 0.9235, "step": 192920 }, { "epoch": 16.71, "learning_rate": 3.327382329131491e-05, "loss": 0.9363, "step": 192930 }, { "epoch": 16.71, "learning_rate": 3.327165490921707e-05, "loss": 0.8896, "step": 192940 }, { "epoch": 16.71, "learning_rate": 3.326948652711923e-05, "loss": 0.8354, "step": 192950 }, { "epoch": 16.71, "learning_rate": 3.3267318145021394e-05, "loss": 0.7831, "step": 192960 }, { "epoch": 16.71, "learning_rate": 3.326514976292355e-05, "loss": 0.9781, "step": 192970 }, { "epoch": 16.71, "learning_rate": 3.326298138082571e-05, "loss": 1.055, "step": 192980 }, { "epoch": 16.71, "learning_rate": 3.326081299872788e-05, "loss": 0.8673, "step": 192990 }, { "epoch": 16.72, "learning_rate": 3.3258644616630046e-05, "loss": 0.9353, "step": 193000 }, { "epoch": 16.72, "learning_rate": 3.3256476234532205e-05, "loss": 0.9329, "step": 193010 }, { "epoch": 16.72, "learning_rate": 3.3254307852434365e-05, "loss": 0.737, "step": 193020 }, { "epoch": 16.72, "learning_rate": 3.325213947033653e-05, "loss": 0.819, "step": 193030 }, { "epoch": 16.72, "learning_rate": 3.324997108823869e-05, "loss": 0.8567, "step": 193040 }, { "epoch": 16.72, "learning_rate": 3.324780270614086e-05, "loss": 1.0725, "step": 193050 }, { "epoch": 16.72, "learning_rate": 3.324563432404302e-05, "loss": 0.8778, "step": 193060 }, { "epoch": 16.72, "learning_rate": 3.324346594194518e-05, "loss": 0.9564, "step": 193070 }, { "epoch": 16.72, "learning_rate": 3.324129755984734e-05, "loss": 0.8449, "step": 193080 }, { "epoch": 16.72, "learning_rate": 3.323912917774951e-05, "loss": 0.8363, "step": 193090 }, { "epoch": 16.72, "learning_rate": 3.323696079565167e-05, "loss": 0.9396, "step": 193100 }, { "epoch": 16.72, "learning_rate": 3.323479241355383e-05, "loss": 0.8474, "step": 193110 }, { "epoch": 16.73, "learning_rate": 3.3232624031455995e-05, "loss": 0.884, "step": 193120 }, { "epoch": 16.73, "learning_rate": 3.3230455649358154e-05, "loss": 0.8039, "step": 193130 }, { "epoch": 16.73, "learning_rate": 3.3228287267260314e-05, "loss": 0.9353, "step": 193140 }, { "epoch": 16.73, "learning_rate": 3.322611888516248e-05, "loss": 0.8854, "step": 193150 }, { "epoch": 16.73, "learning_rate": 3.322395050306465e-05, "loss": 0.9484, "step": 193160 }, { "epoch": 16.73, "learning_rate": 3.3221782120966806e-05, "loss": 0.9158, "step": 193170 }, { "epoch": 16.73, "learning_rate": 3.3219613738868966e-05, "loss": 1.0278, "step": 193180 }, { "epoch": 16.73, "learning_rate": 3.321744535677113e-05, "loss": 0.8879, "step": 193190 }, { "epoch": 16.73, "learning_rate": 3.321527697467329e-05, "loss": 1.0286, "step": 193200 }, { "epoch": 16.73, "learning_rate": 3.321310859257545e-05, "loss": 0.9653, "step": 193210 }, { "epoch": 16.73, "learning_rate": 3.321094021047762e-05, "loss": 0.8763, "step": 193220 }, { "epoch": 16.73, "learning_rate": 3.3208771828379784e-05, "loss": 0.9727, "step": 193230 }, { "epoch": 16.74, "learning_rate": 3.3206603446281944e-05, "loss": 0.9319, "step": 193240 }, { "epoch": 16.74, "learning_rate": 3.3204435064184104e-05, "loss": 0.6476, "step": 193250 }, { "epoch": 16.74, "learning_rate": 3.320226668208627e-05, "loss": 0.9234, "step": 193260 }, { "epoch": 16.74, "learning_rate": 3.320009829998843e-05, "loss": 0.7947, "step": 193270 }, { "epoch": 16.74, "learning_rate": 3.3197929917890596e-05, "loss": 0.7608, "step": 193280 }, { "epoch": 16.74, "learning_rate": 3.3195761535792756e-05, "loss": 0.9907, "step": 193290 }, { "epoch": 16.74, "learning_rate": 3.319359315369492e-05, "loss": 0.8929, "step": 193300 }, { "epoch": 16.74, "learning_rate": 3.319142477159708e-05, "loss": 0.826, "step": 193310 }, { "epoch": 16.74, "learning_rate": 3.318925638949925e-05, "loss": 0.906, "step": 193320 }, { "epoch": 16.74, "learning_rate": 3.318708800740141e-05, "loss": 0.9006, "step": 193330 }, { "epoch": 16.74, "learning_rate": 3.318491962530357e-05, "loss": 0.9922, "step": 193340 }, { "epoch": 16.75, "learning_rate": 3.3182751243205734e-05, "loss": 0.8347, "step": 193350 }, { "epoch": 16.75, "learning_rate": 3.31805828611079e-05, "loss": 0.9116, "step": 193360 }, { "epoch": 16.75, "learning_rate": 3.317841447901006e-05, "loss": 0.9012, "step": 193370 }, { "epoch": 16.75, "learning_rate": 3.317624609691222e-05, "loss": 0.847, "step": 193380 }, { "epoch": 16.75, "learning_rate": 3.3174077714814386e-05, "loss": 0.9974, "step": 193390 }, { "epoch": 16.75, "learning_rate": 3.3171909332716545e-05, "loss": 0.7722, "step": 193400 }, { "epoch": 16.75, "learning_rate": 3.3169740950618705e-05, "loss": 0.8193, "step": 193410 }, { "epoch": 16.75, "learning_rate": 3.316757256852087e-05, "loss": 0.8986, "step": 193420 }, { "epoch": 16.75, "learning_rate": 3.316540418642304e-05, "loss": 0.7029, "step": 193430 }, { "epoch": 16.75, "learning_rate": 3.31632358043252e-05, "loss": 0.8016, "step": 193440 }, { "epoch": 16.75, "learning_rate": 3.316106742222736e-05, "loss": 1.0243, "step": 193450 }, { "epoch": 16.75, "learning_rate": 3.315889904012952e-05, "loss": 1.0877, "step": 193460 }, { "epoch": 16.76, "learning_rate": 3.315673065803168e-05, "loss": 0.8499, "step": 193470 }, { "epoch": 16.76, "learning_rate": 3.315456227593384e-05, "loss": 0.7961, "step": 193480 }, { "epoch": 16.76, "learning_rate": 3.315239389383601e-05, "loss": 1.1378, "step": 193490 }, { "epoch": 16.76, "learning_rate": 3.3150225511738175e-05, "loss": 1.0618, "step": 193500 }, { "epoch": 16.76, "learning_rate": 3.3148057129640335e-05, "loss": 0.7526, "step": 193510 }, { "epoch": 16.76, "learning_rate": 3.3145888747542495e-05, "loss": 0.9173, "step": 193520 }, { "epoch": 16.76, "learning_rate": 3.314372036544466e-05, "loss": 0.8107, "step": 193530 }, { "epoch": 16.76, "learning_rate": 3.314155198334682e-05, "loss": 0.7565, "step": 193540 }, { "epoch": 16.76, "learning_rate": 3.313938360124899e-05, "loss": 1.0152, "step": 193550 }, { "epoch": 16.76, "learning_rate": 3.313721521915115e-05, "loss": 0.8235, "step": 193560 }, { "epoch": 16.76, "learning_rate": 3.313504683705331e-05, "loss": 0.9174, "step": 193570 }, { "epoch": 16.77, "learning_rate": 3.313287845495547e-05, "loss": 0.9292, "step": 193580 }, { "epoch": 16.77, "learning_rate": 3.313071007285764e-05, "loss": 0.8368, "step": 193590 }, { "epoch": 16.77, "learning_rate": 3.31285416907598e-05, "loss": 0.9303, "step": 193600 }, { "epoch": 16.77, "learning_rate": 3.312637330866196e-05, "loss": 0.6581, "step": 193610 }, { "epoch": 16.77, "learning_rate": 3.3124204926564125e-05, "loss": 0.9582, "step": 193620 }, { "epoch": 16.77, "learning_rate": 3.312203654446629e-05, "loss": 0.9745, "step": 193630 }, { "epoch": 16.77, "learning_rate": 3.3119868162368444e-05, "loss": 0.7821, "step": 193640 }, { "epoch": 16.77, "learning_rate": 3.311769978027061e-05, "loss": 0.8339, "step": 193650 }, { "epoch": 16.77, "learning_rate": 3.311553139817278e-05, "loss": 0.9358, "step": 193660 }, { "epoch": 16.77, "learning_rate": 3.3113363016074936e-05, "loss": 0.7988, "step": 193670 }, { "epoch": 16.77, "learning_rate": 3.3111194633977096e-05, "loss": 0.7191, "step": 193680 }, { "epoch": 16.77, "learning_rate": 3.310902625187926e-05, "loss": 0.8649, "step": 193690 }, { "epoch": 16.78, "learning_rate": 3.310685786978142e-05, "loss": 0.865, "step": 193700 }, { "epoch": 16.78, "learning_rate": 3.310468948768359e-05, "loss": 0.8495, "step": 193710 }, { "epoch": 16.78, "learning_rate": 3.310252110558575e-05, "loss": 0.873, "step": 193720 }, { "epoch": 16.78, "learning_rate": 3.3100352723487914e-05, "loss": 0.9776, "step": 193730 }, { "epoch": 16.78, "learning_rate": 3.3098184341390074e-05, "loss": 0.7923, "step": 193740 }, { "epoch": 16.78, "learning_rate": 3.3096015959292234e-05, "loss": 0.7017, "step": 193750 }, { "epoch": 16.78, "learning_rate": 3.30938475771944e-05, "loss": 0.8327, "step": 193760 }, { "epoch": 16.78, "learning_rate": 3.309167919509656e-05, "loss": 0.9443, "step": 193770 }, { "epoch": 16.78, "learning_rate": 3.3089510812998726e-05, "loss": 0.9547, "step": 193780 }, { "epoch": 16.78, "learning_rate": 3.3087342430900886e-05, "loss": 0.8895, "step": 193790 }, { "epoch": 16.78, "learning_rate": 3.308517404880305e-05, "loss": 1.0893, "step": 193800 }, { "epoch": 16.79, "learning_rate": 3.308300566670521e-05, "loss": 0.9773, "step": 193810 }, { "epoch": 16.79, "learning_rate": 3.308083728460738e-05, "loss": 1.0075, "step": 193820 }, { "epoch": 16.79, "learning_rate": 3.307866890250954e-05, "loss": 0.9211, "step": 193830 }, { "epoch": 16.79, "learning_rate": 3.30765005204117e-05, "loss": 0.8869, "step": 193840 }, { "epoch": 16.79, "learning_rate": 3.3074332138313864e-05, "loss": 0.9096, "step": 193850 }, { "epoch": 16.79, "learning_rate": 3.307216375621603e-05, "loss": 0.7814, "step": 193860 }, { "epoch": 16.79, "learning_rate": 3.306999537411819e-05, "loss": 0.8018, "step": 193870 }, { "epoch": 16.79, "learning_rate": 3.306782699202035e-05, "loss": 0.7012, "step": 193880 }, { "epoch": 16.79, "learning_rate": 3.3065658609922516e-05, "loss": 0.6749, "step": 193890 }, { "epoch": 16.79, "learning_rate": 3.3063490227824675e-05, "loss": 1.0513, "step": 193900 }, { "epoch": 16.79, "learning_rate": 3.3061321845726835e-05, "loss": 1.0026, "step": 193910 }, { "epoch": 16.79, "learning_rate": 3.3059153463629e-05, "loss": 0.8826, "step": 193920 }, { "epoch": 16.8, "learning_rate": 3.305698508153117e-05, "loss": 0.7193, "step": 193930 }, { "epoch": 16.8, "learning_rate": 3.305481669943333e-05, "loss": 0.7644, "step": 193940 }, { "epoch": 16.8, "learning_rate": 3.305264831733549e-05, "loss": 0.9725, "step": 193950 }, { "epoch": 16.8, "learning_rate": 3.305047993523765e-05, "loss": 0.8404, "step": 193960 }, { "epoch": 16.8, "learning_rate": 3.304831155313981e-05, "loss": 0.7891, "step": 193970 }, { "epoch": 16.8, "learning_rate": 3.304614317104197e-05, "loss": 0.9174, "step": 193980 }, { "epoch": 16.8, "learning_rate": 3.304397478894414e-05, "loss": 0.9784, "step": 193990 }, { "epoch": 16.8, "learning_rate": 3.3041806406846305e-05, "loss": 0.8734, "step": 194000 }, { "epoch": 16.8, "learning_rate": 3.3039638024748465e-05, "loss": 0.8562, "step": 194010 }, { "epoch": 16.8, "learning_rate": 3.3037469642650624e-05, "loss": 0.8367, "step": 194020 }, { "epoch": 16.8, "learning_rate": 3.303530126055279e-05, "loss": 0.9279, "step": 194030 }, { "epoch": 16.81, "learning_rate": 3.303313287845495e-05, "loss": 1.1317, "step": 194040 }, { "epoch": 16.81, "learning_rate": 3.303096449635712e-05, "loss": 0.964, "step": 194050 }, { "epoch": 16.81, "learning_rate": 3.3028796114259276e-05, "loss": 0.9973, "step": 194060 }, { "epoch": 16.81, "learning_rate": 3.302662773216144e-05, "loss": 0.7912, "step": 194070 }, { "epoch": 16.81, "learning_rate": 3.30244593500636e-05, "loss": 0.9523, "step": 194080 }, { "epoch": 16.81, "learning_rate": 3.302229096796577e-05, "loss": 0.929, "step": 194090 }, { "epoch": 16.81, "learning_rate": 3.302012258586793e-05, "loss": 0.9264, "step": 194100 }, { "epoch": 16.81, "learning_rate": 3.301795420377009e-05, "loss": 0.9192, "step": 194110 }, { "epoch": 16.81, "learning_rate": 3.3015785821672255e-05, "loss": 0.9543, "step": 194120 }, { "epoch": 16.81, "learning_rate": 3.301361743957442e-05, "loss": 0.8108, "step": 194130 }, { "epoch": 16.81, "learning_rate": 3.301144905747658e-05, "loss": 0.7837, "step": 194140 }, { "epoch": 16.81, "learning_rate": 3.300928067537874e-05, "loss": 0.9263, "step": 194150 }, { "epoch": 16.82, "learning_rate": 3.3007112293280907e-05, "loss": 0.9892, "step": 194160 }, { "epoch": 16.82, "learning_rate": 3.3004943911183066e-05, "loss": 0.9422, "step": 194170 }, { "epoch": 16.82, "learning_rate": 3.3002775529085226e-05, "loss": 0.8819, "step": 194180 }, { "epoch": 16.82, "learning_rate": 3.300060714698739e-05, "loss": 0.8995, "step": 194190 }, { "epoch": 16.82, "learning_rate": 3.299843876488955e-05, "loss": 0.8456, "step": 194200 }, { "epoch": 16.82, "learning_rate": 3.299627038279172e-05, "loss": 0.7451, "step": 194210 }, { "epoch": 16.82, "learning_rate": 3.299410200069388e-05, "loss": 0.9074, "step": 194220 }, { "epoch": 16.82, "learning_rate": 3.2991933618596044e-05, "loss": 0.8132, "step": 194230 }, { "epoch": 16.82, "learning_rate": 3.2989765236498204e-05, "loss": 0.9577, "step": 194240 }, { "epoch": 16.82, "learning_rate": 3.2987596854400363e-05, "loss": 0.8704, "step": 194250 }, { "epoch": 16.82, "learning_rate": 3.298542847230253e-05, "loss": 0.8733, "step": 194260 }, { "epoch": 16.83, "learning_rate": 3.298326009020469e-05, "loss": 0.8745, "step": 194270 }, { "epoch": 16.83, "learning_rate": 3.2981091708106856e-05, "loss": 0.9107, "step": 194280 }, { "epoch": 16.83, "learning_rate": 3.2978923326009015e-05, "loss": 1.0035, "step": 194290 }, { "epoch": 16.83, "learning_rate": 3.297675494391118e-05, "loss": 0.7547, "step": 194300 }, { "epoch": 16.83, "learning_rate": 3.297458656181334e-05, "loss": 0.9881, "step": 194310 }, { "epoch": 16.83, "learning_rate": 3.297241817971551e-05, "loss": 0.9098, "step": 194320 }, { "epoch": 16.83, "learning_rate": 3.297024979761767e-05, "loss": 0.8606, "step": 194330 }, { "epoch": 16.83, "learning_rate": 3.296808141551983e-05, "loss": 1.1278, "step": 194340 }, { "epoch": 16.83, "learning_rate": 3.2965913033421993e-05, "loss": 0.8878, "step": 194350 }, { "epoch": 16.83, "learning_rate": 3.296374465132416e-05, "loss": 0.8025, "step": 194360 }, { "epoch": 16.83, "learning_rate": 3.296157626922632e-05, "loss": 0.847, "step": 194370 }, { "epoch": 16.83, "learning_rate": 3.295940788712848e-05, "loss": 1.0621, "step": 194380 }, { "epoch": 16.84, "learning_rate": 3.2957239505030645e-05, "loss": 0.8451, "step": 194390 }, { "epoch": 16.84, "learning_rate": 3.2955071122932805e-05, "loss": 0.8595, "step": 194400 }, { "epoch": 16.84, "learning_rate": 3.2952902740834965e-05, "loss": 0.9481, "step": 194410 }, { "epoch": 16.84, "learning_rate": 3.295073435873713e-05, "loss": 0.7567, "step": 194420 }, { "epoch": 16.84, "learning_rate": 3.29485659766393e-05, "loss": 0.859, "step": 194430 }, { "epoch": 16.84, "learning_rate": 3.294639759454146e-05, "loss": 0.8882, "step": 194440 }, { "epoch": 16.84, "learning_rate": 3.294422921244362e-05, "loss": 0.7389, "step": 194450 }, { "epoch": 16.84, "learning_rate": 3.294206083034578e-05, "loss": 0.8482, "step": 194460 }, { "epoch": 16.84, "learning_rate": 3.293989244824794e-05, "loss": 0.9276, "step": 194470 }, { "epoch": 16.84, "learning_rate": 3.293772406615011e-05, "loss": 0.9271, "step": 194480 }, { "epoch": 16.84, "learning_rate": 3.293555568405227e-05, "loss": 0.8418, "step": 194490 }, { "epoch": 16.84, "learning_rate": 3.2933387301954435e-05, "loss": 0.8195, "step": 194500 }, { "epoch": 16.85, "learning_rate": 3.2931218919856595e-05, "loss": 0.9741, "step": 194510 }, { "epoch": 16.85, "learning_rate": 3.2929050537758754e-05, "loss": 0.849, "step": 194520 }, { "epoch": 16.85, "learning_rate": 3.292688215566092e-05, "loss": 0.8878, "step": 194530 }, { "epoch": 16.85, "learning_rate": 3.292471377356308e-05, "loss": 0.8473, "step": 194540 }, { "epoch": 16.85, "learning_rate": 3.292254539146525e-05, "loss": 0.9359, "step": 194550 }, { "epoch": 16.85, "learning_rate": 3.2920377009367406e-05, "loss": 1.0677, "step": 194560 }, { "epoch": 16.85, "learning_rate": 3.291820862726957e-05, "loss": 0.8898, "step": 194570 }, { "epoch": 16.85, "learning_rate": 3.291604024517173e-05, "loss": 0.8613, "step": 194580 }, { "epoch": 16.85, "learning_rate": 3.29138718630739e-05, "loss": 0.8589, "step": 194590 }, { "epoch": 16.85, "learning_rate": 3.291170348097606e-05, "loss": 0.971, "step": 194600 }, { "epoch": 16.85, "learning_rate": 3.290953509887822e-05, "loss": 0.7551, "step": 194610 }, { "epoch": 16.86, "learning_rate": 3.2907366716780384e-05, "loss": 0.7889, "step": 194620 }, { "epoch": 16.86, "learning_rate": 3.290519833468255e-05, "loss": 0.9461, "step": 194630 }, { "epoch": 16.86, "learning_rate": 3.290302995258471e-05, "loss": 0.8355, "step": 194640 }, { "epoch": 16.86, "learning_rate": 3.290086157048687e-05, "loss": 0.8163, "step": 194650 }, { "epoch": 16.86, "learning_rate": 3.2898693188389036e-05, "loss": 0.7996, "step": 194660 }, { "epoch": 16.86, "learning_rate": 3.2896524806291196e-05, "loss": 0.8772, "step": 194670 }, { "epoch": 16.86, "learning_rate": 3.2894356424193356e-05, "loss": 0.9406, "step": 194680 }, { "epoch": 16.86, "learning_rate": 3.289218804209552e-05, "loss": 0.9642, "step": 194690 }, { "epoch": 16.86, "learning_rate": 3.289001965999769e-05, "loss": 0.7864, "step": 194700 }, { "epoch": 16.86, "learning_rate": 3.288785127789985e-05, "loss": 0.7953, "step": 194710 }, { "epoch": 16.86, "learning_rate": 3.288568289580201e-05, "loss": 0.9615, "step": 194720 }, { "epoch": 16.86, "learning_rate": 3.2883514513704174e-05, "loss": 0.8931, "step": 194730 }, { "epoch": 16.87, "learning_rate": 3.2881346131606334e-05, "loss": 0.787, "step": 194740 }, { "epoch": 16.87, "learning_rate": 3.287917774950849e-05, "loss": 0.7482, "step": 194750 }, { "epoch": 16.87, "learning_rate": 3.287700936741066e-05, "loss": 0.7986, "step": 194760 }, { "epoch": 16.87, "learning_rate": 3.287484098531282e-05, "loss": 0.775, "step": 194770 }, { "epoch": 16.87, "learning_rate": 3.2872672603214986e-05, "loss": 0.705, "step": 194780 }, { "epoch": 16.87, "learning_rate": 3.2870504221117145e-05, "loss": 0.9611, "step": 194790 }, { "epoch": 16.87, "learning_rate": 3.286833583901931e-05, "loss": 0.9159, "step": 194800 }, { "epoch": 16.87, "learning_rate": 3.286616745692147e-05, "loss": 0.8688, "step": 194810 }, { "epoch": 16.87, "learning_rate": 3.286399907482364e-05, "loss": 0.9509, "step": 194820 }, { "epoch": 16.87, "learning_rate": 3.28618306927258e-05, "loss": 0.9398, "step": 194830 }, { "epoch": 16.87, "learning_rate": 3.285966231062796e-05, "loss": 0.7606, "step": 194840 }, { "epoch": 16.88, "learning_rate": 3.285749392853012e-05, "loss": 0.9955, "step": 194850 }, { "epoch": 16.88, "learning_rate": 3.285532554643229e-05, "loss": 0.8225, "step": 194860 }, { "epoch": 16.88, "learning_rate": 3.285315716433445e-05, "loss": 0.9444, "step": 194870 }, { "epoch": 16.88, "learning_rate": 3.285098878223661e-05, "loss": 1.013, "step": 194880 }, { "epoch": 16.88, "learning_rate": 3.2848820400138775e-05, "loss": 0.9163, "step": 194890 }, { "epoch": 16.88, "learning_rate": 3.2846652018040935e-05, "loss": 0.852, "step": 194900 }, { "epoch": 16.88, "learning_rate": 3.2844483635943095e-05, "loss": 0.9087, "step": 194910 }, { "epoch": 16.88, "learning_rate": 3.284231525384526e-05, "loss": 0.8832, "step": 194920 }, { "epoch": 16.88, "learning_rate": 3.284014687174743e-05, "loss": 0.9684, "step": 194930 }, { "epoch": 16.88, "learning_rate": 3.283797848964959e-05, "loss": 0.933, "step": 194940 }, { "epoch": 16.88, "learning_rate": 3.2835810107551747e-05, "loss": 0.7786, "step": 194950 }, { "epoch": 16.88, "learning_rate": 3.283364172545391e-05, "loss": 0.9244, "step": 194960 }, { "epoch": 16.89, "learning_rate": 3.283147334335607e-05, "loss": 0.9785, "step": 194970 }, { "epoch": 16.89, "learning_rate": 3.282930496125824e-05, "loss": 0.913, "step": 194980 }, { "epoch": 16.89, "learning_rate": 3.28271365791604e-05, "loss": 0.901, "step": 194990 }, { "epoch": 16.89, "learning_rate": 3.2824968197062565e-05, "loss": 0.8161, "step": 195000 }, { "epoch": 16.89, "learning_rate": 3.2822799814964725e-05, "loss": 0.7119, "step": 195010 }, { "epoch": 16.89, "learning_rate": 3.2820631432866884e-05, "loss": 0.9341, "step": 195020 }, { "epoch": 16.89, "learning_rate": 3.281846305076905e-05, "loss": 1.0583, "step": 195030 }, { "epoch": 16.89, "learning_rate": 3.281629466867121e-05, "loss": 0.9415, "step": 195040 }, { "epoch": 16.89, "learning_rate": 3.2814126286573377e-05, "loss": 0.8506, "step": 195050 }, { "epoch": 16.89, "learning_rate": 3.2811957904475536e-05, "loss": 1.0003, "step": 195060 }, { "epoch": 16.89, "learning_rate": 3.28097895223777e-05, "loss": 0.9404, "step": 195070 }, { "epoch": 16.9, "learning_rate": 3.280762114027986e-05, "loss": 0.8329, "step": 195080 }, { "epoch": 16.9, "learning_rate": 3.280545275818203e-05, "loss": 1.0505, "step": 195090 }, { "epoch": 16.9, "learning_rate": 3.280328437608419e-05, "loss": 1.001, "step": 195100 }, { "epoch": 16.9, "learning_rate": 3.280111599398635e-05, "loss": 0.8498, "step": 195110 }, { "epoch": 16.9, "learning_rate": 3.2798947611888514e-05, "loss": 0.9591, "step": 195120 }, { "epoch": 16.9, "learning_rate": 3.279677922979068e-05, "loss": 0.8577, "step": 195130 }, { "epoch": 16.9, "learning_rate": 3.279461084769284e-05, "loss": 0.8771, "step": 195140 }, { "epoch": 16.9, "learning_rate": 3.2792442465595e-05, "loss": 0.8326, "step": 195150 }, { "epoch": 16.9, "learning_rate": 3.2790274083497166e-05, "loss": 0.8087, "step": 195160 }, { "epoch": 16.9, "learning_rate": 3.2788105701399326e-05, "loss": 0.8693, "step": 195170 }, { "epoch": 16.9, "learning_rate": 3.2785937319301485e-05, "loss": 0.891, "step": 195180 }, { "epoch": 16.9, "learning_rate": 3.278376893720365e-05, "loss": 0.6936, "step": 195190 }, { "epoch": 16.91, "learning_rate": 3.278160055510582e-05, "loss": 1.0276, "step": 195200 }, { "epoch": 16.91, "learning_rate": 3.277943217300798e-05, "loss": 1.0281, "step": 195210 }, { "epoch": 16.91, "learning_rate": 3.277726379091014e-05, "loss": 0.9176, "step": 195220 }, { "epoch": 16.91, "learning_rate": 3.2775095408812304e-05, "loss": 0.8224, "step": 195230 }, { "epoch": 16.91, "learning_rate": 3.2772927026714463e-05, "loss": 0.9105, "step": 195240 }, { "epoch": 16.91, "learning_rate": 3.277075864461662e-05, "loss": 0.9948, "step": 195250 }, { "epoch": 16.91, "learning_rate": 3.276859026251879e-05, "loss": 0.8687, "step": 195260 }, { "epoch": 16.91, "learning_rate": 3.276642188042095e-05, "loss": 0.8425, "step": 195270 }, { "epoch": 16.91, "learning_rate": 3.2764253498323115e-05, "loss": 0.8769, "step": 195280 }, { "epoch": 16.91, "learning_rate": 3.2762085116225275e-05, "loss": 0.8619, "step": 195290 }, { "epoch": 16.91, "learning_rate": 3.275991673412744e-05, "loss": 0.8456, "step": 195300 }, { "epoch": 16.92, "learning_rate": 3.27577483520296e-05, "loss": 0.7115, "step": 195310 }, { "epoch": 16.92, "learning_rate": 3.275557996993177e-05, "loss": 1.002, "step": 195320 }, { "epoch": 16.92, "learning_rate": 3.275341158783393e-05, "loss": 0.9413, "step": 195330 }, { "epoch": 16.92, "learning_rate": 3.275124320573609e-05, "loss": 0.9651, "step": 195340 }, { "epoch": 16.92, "learning_rate": 3.274907482363825e-05, "loss": 0.8748, "step": 195350 }, { "epoch": 16.92, "learning_rate": 3.274690644154042e-05, "loss": 0.8107, "step": 195360 }, { "epoch": 16.92, "learning_rate": 3.274473805944258e-05, "loss": 0.817, "step": 195370 }, { "epoch": 16.92, "learning_rate": 3.274256967734474e-05, "loss": 1.0387, "step": 195380 }, { "epoch": 16.92, "learning_rate": 3.2740401295246905e-05, "loss": 0.9098, "step": 195390 }, { "epoch": 16.92, "learning_rate": 3.2738232913149065e-05, "loss": 0.7755, "step": 195400 }, { "epoch": 16.92, "learning_rate": 3.2736064531051224e-05, "loss": 0.846, "step": 195410 }, { "epoch": 16.92, "learning_rate": 3.273389614895339e-05, "loss": 0.9764, "step": 195420 }, { "epoch": 16.93, "learning_rate": 3.273172776685556e-05, "loss": 0.8351, "step": 195430 }, { "epoch": 16.93, "learning_rate": 3.272955938475772e-05, "loss": 1.1329, "step": 195440 }, { "epoch": 16.93, "learning_rate": 3.2727391002659876e-05, "loss": 0.8228, "step": 195450 }, { "epoch": 16.93, "learning_rate": 3.272522262056204e-05, "loss": 0.7652, "step": 195460 }, { "epoch": 16.93, "learning_rate": 3.27230542384642e-05, "loss": 1.0276, "step": 195470 }, { "epoch": 16.93, "learning_rate": 3.272088585636637e-05, "loss": 0.738, "step": 195480 }, { "epoch": 16.93, "learning_rate": 3.271871747426853e-05, "loss": 1.1034, "step": 195490 }, { "epoch": 16.93, "learning_rate": 3.2716549092170695e-05, "loss": 0.9058, "step": 195500 }, { "epoch": 16.93, "learning_rate": 3.2714380710072854e-05, "loss": 0.9107, "step": 195510 }, { "epoch": 16.93, "learning_rate": 3.2712212327975014e-05, "loss": 0.9645, "step": 195520 }, { "epoch": 16.93, "learning_rate": 3.271004394587718e-05, "loss": 1.0451, "step": 195530 }, { "epoch": 16.94, "learning_rate": 3.270787556377934e-05, "loss": 0.8664, "step": 195540 }, { "epoch": 16.94, "learning_rate": 3.2705707181681506e-05, "loss": 0.8926, "step": 195550 }, { "epoch": 16.94, "learning_rate": 3.2703538799583666e-05, "loss": 0.9602, "step": 195560 }, { "epoch": 16.94, "learning_rate": 3.270137041748583e-05, "loss": 0.7133, "step": 195570 }, { "epoch": 16.94, "learning_rate": 3.269920203538799e-05, "loss": 0.7413, "step": 195580 }, { "epoch": 16.94, "learning_rate": 3.269703365329016e-05, "loss": 0.8492, "step": 195590 }, { "epoch": 16.94, "learning_rate": 3.269486527119232e-05, "loss": 0.7392, "step": 195600 }, { "epoch": 16.94, "learning_rate": 3.269269688909448e-05, "loss": 1.0122, "step": 195610 }, { "epoch": 16.94, "learning_rate": 3.2690528506996644e-05, "loss": 0.7766, "step": 195620 }, { "epoch": 16.94, "learning_rate": 3.268836012489881e-05, "loss": 0.8012, "step": 195630 }, { "epoch": 16.94, "learning_rate": 3.268619174280097e-05, "loss": 0.8883, "step": 195640 }, { "epoch": 16.94, "learning_rate": 3.268402336070313e-05, "loss": 0.9134, "step": 195650 }, { "epoch": 16.95, "learning_rate": 3.2681854978605296e-05, "loss": 0.8752, "step": 195660 }, { "epoch": 16.95, "learning_rate": 3.2679686596507456e-05, "loss": 0.8742, "step": 195670 }, { "epoch": 16.95, "learning_rate": 3.2677518214409615e-05, "loss": 0.9052, "step": 195680 }, { "epoch": 16.95, "learning_rate": 3.267534983231178e-05, "loss": 1.0395, "step": 195690 }, { "epoch": 16.95, "learning_rate": 3.267318145021395e-05, "loss": 0.8337, "step": 195700 }, { "epoch": 16.95, "learning_rate": 3.267101306811611e-05, "loss": 0.8892, "step": 195710 }, { "epoch": 16.95, "learning_rate": 3.266884468601827e-05, "loss": 0.9047, "step": 195720 }, { "epoch": 16.95, "learning_rate": 3.2666676303920434e-05, "loss": 0.9878, "step": 195730 }, { "epoch": 16.95, "learning_rate": 3.266450792182259e-05, "loss": 0.9129, "step": 195740 }, { "epoch": 16.95, "learning_rate": 3.266233953972476e-05, "loss": 1.0073, "step": 195750 }, { "epoch": 16.95, "learning_rate": 3.266017115762692e-05, "loss": 1.0243, "step": 195760 }, { "epoch": 16.95, "learning_rate": 3.2658002775529086e-05, "loss": 0.9296, "step": 195770 }, { "epoch": 16.96, "learning_rate": 3.2655834393431245e-05, "loss": 0.7475, "step": 195780 }, { "epoch": 16.96, "learning_rate": 3.2653666011333405e-05, "loss": 0.9864, "step": 195790 }, { "epoch": 16.96, "learning_rate": 3.265149762923557e-05, "loss": 0.7756, "step": 195800 }, { "epoch": 16.96, "learning_rate": 3.264932924713773e-05, "loss": 0.7613, "step": 195810 }, { "epoch": 16.96, "learning_rate": 3.26471608650399e-05, "loss": 0.9178, "step": 195820 }, { "epoch": 16.96, "learning_rate": 3.264499248294206e-05, "loss": 0.8736, "step": 195830 }, { "epoch": 16.96, "learning_rate": 3.2642824100844217e-05, "loss": 0.8498, "step": 195840 }, { "epoch": 16.96, "learning_rate": 3.264065571874638e-05, "loss": 0.7687, "step": 195850 }, { "epoch": 16.96, "learning_rate": 3.263848733664855e-05, "loss": 0.9339, "step": 195860 }, { "epoch": 16.96, "learning_rate": 3.263631895455071e-05, "loss": 0.9834, "step": 195870 }, { "epoch": 16.96, "learning_rate": 3.263415057245287e-05, "loss": 0.7215, "step": 195880 }, { "epoch": 16.97, "learning_rate": 3.2631982190355035e-05, "loss": 0.8556, "step": 195890 }, { "epoch": 16.97, "learning_rate": 3.2629813808257195e-05, "loss": 0.8125, "step": 195900 }, { "epoch": 16.97, "learning_rate": 3.2627645426159354e-05, "loss": 1.1569, "step": 195910 }, { "epoch": 16.97, "learning_rate": 3.262547704406152e-05, "loss": 0.8389, "step": 195920 }, { "epoch": 16.97, "learning_rate": 3.262330866196369e-05, "loss": 0.8674, "step": 195930 }, { "epoch": 16.97, "learning_rate": 3.2621140279865847e-05, "loss": 0.9253, "step": 195940 }, { "epoch": 16.97, "learning_rate": 3.2618971897768006e-05, "loss": 0.6941, "step": 195950 }, { "epoch": 16.97, "learning_rate": 3.261680351567017e-05, "loss": 0.8754, "step": 195960 }, { "epoch": 16.97, "learning_rate": 3.261463513357233e-05, "loss": 0.8275, "step": 195970 }, { "epoch": 16.97, "learning_rate": 3.26124667514745e-05, "loss": 0.9971, "step": 195980 }, { "epoch": 16.97, "learning_rate": 3.261029836937666e-05, "loss": 0.8529, "step": 195990 }, { "epoch": 16.97, "learning_rate": 3.2608129987278825e-05, "loss": 0.8635, "step": 196000 }, { "epoch": 16.98, "learning_rate": 3.2605961605180984e-05, "loss": 0.8194, "step": 196010 }, { "epoch": 16.98, "learning_rate": 3.2603793223083144e-05, "loss": 0.9046, "step": 196020 }, { "epoch": 16.98, "learning_rate": 3.260162484098531e-05, "loss": 0.6444, "step": 196030 }, { "epoch": 16.98, "learning_rate": 3.259945645888747e-05, "loss": 0.9098, "step": 196040 }, { "epoch": 16.98, "learning_rate": 3.2597288076789636e-05, "loss": 0.7899, "step": 196050 }, { "epoch": 16.98, "learning_rate": 3.2595119694691796e-05, "loss": 0.9111, "step": 196060 }, { "epoch": 16.98, "learning_rate": 3.259295131259396e-05, "loss": 0.9457, "step": 196070 }, { "epoch": 16.98, "learning_rate": 3.259078293049612e-05, "loss": 0.9167, "step": 196080 }, { "epoch": 16.98, "learning_rate": 3.258861454839829e-05, "loss": 1.0379, "step": 196090 }, { "epoch": 16.98, "learning_rate": 3.258644616630045e-05, "loss": 0.8706, "step": 196100 }, { "epoch": 16.98, "learning_rate": 3.258427778420261e-05, "loss": 1.1085, "step": 196110 }, { "epoch": 16.99, "learning_rate": 3.2582109402104774e-05, "loss": 0.8337, "step": 196120 }, { "epoch": 16.99, "learning_rate": 3.257994102000694e-05, "loss": 0.8831, "step": 196130 }, { "epoch": 16.99, "learning_rate": 3.25777726379091e-05, "loss": 0.9624, "step": 196140 }, { "epoch": 16.99, "learning_rate": 3.257560425581126e-05, "loss": 0.8225, "step": 196150 }, { "epoch": 16.99, "learning_rate": 3.2573435873713426e-05, "loss": 1.0086, "step": 196160 }, { "epoch": 16.99, "learning_rate": 3.2571267491615586e-05, "loss": 0.8296, "step": 196170 }, { "epoch": 16.99, "learning_rate": 3.2569099109517745e-05, "loss": 0.9225, "step": 196180 }, { "epoch": 16.99, "learning_rate": 3.256693072741991e-05, "loss": 0.9175, "step": 196190 }, { "epoch": 16.99, "learning_rate": 3.256476234532208e-05, "loss": 0.7492, "step": 196200 }, { "epoch": 16.99, "learning_rate": 3.256259396322424e-05, "loss": 1.0064, "step": 196210 }, { "epoch": 16.99, "learning_rate": 3.25604255811264e-05, "loss": 1.0125, "step": 196220 }, { "epoch": 16.99, "learning_rate": 3.2558257199028564e-05, "loss": 0.9121, "step": 196230 }, { "epoch": 17.0, "learning_rate": 3.255608881693072e-05, "loss": 0.8991, "step": 196240 }, { "epoch": 17.0, "learning_rate": 3.255392043483289e-05, "loss": 0.9068, "step": 196250 }, { "epoch": 17.0, "learning_rate": 3.255175205273505e-05, "loss": 0.8631, "step": 196260 }, { "epoch": 17.0, "learning_rate": 3.2549583670637216e-05, "loss": 0.8755, "step": 196270 }, { "epoch": 17.0, "learning_rate": 3.2547415288539375e-05, "loss": 0.8822, "step": 196280 }, { "epoch": 17.0, "learning_rate": 3.2545246906441535e-05, "loss": 0.8936, "step": 196290 }, { "epoch": 17.0, "eval_cer": 0.5411418486732482, "eval_em": 0.04951005673027334, "eval_f1": 0.04951005673027334, "eval_loss": 1.4672828912734985, "eval_runtime": 264.4192, "eval_samples_per_second": 7.333, "eval_steps_per_second": 0.919, "eval_wer": 0.9504899432697267, "step": 196290 }, { "epoch": 17.0, "learning_rate": 3.25430785243437e-05, "loss": 0.8317, "step": 196300 }, { "epoch": 17.0, "learning_rate": 3.254091014224586e-05, "loss": 0.7981, "step": 196310 }, { "epoch": 17.0, "learning_rate": 3.253874176014803e-05, "loss": 0.6339, "step": 196320 }, { "epoch": 17.0, "learning_rate": 3.253657337805019e-05, "loss": 0.7874, "step": 196330 }, { "epoch": 17.0, "learning_rate": 3.2534404995952346e-05, "loss": 0.879, "step": 196340 }, { "epoch": 17.01, "learning_rate": 3.253223661385451e-05, "loss": 0.6469, "step": 196350 }, { "epoch": 17.01, "learning_rate": 3.253006823175668e-05, "loss": 0.9784, "step": 196360 }, { "epoch": 17.01, "learning_rate": 3.252789984965884e-05, "loss": 0.6905, "step": 196370 }, { "epoch": 17.01, "learning_rate": 3.2525731467561e-05, "loss": 0.8265, "step": 196380 }, { "epoch": 17.01, "learning_rate": 3.2523563085463165e-05, "loss": 0.6618, "step": 196390 }, { "epoch": 17.01, "learning_rate": 3.2521394703365324e-05, "loss": 1.0097, "step": 196400 }, { "epoch": 17.01, "learning_rate": 3.2519226321267484e-05, "loss": 0.7365, "step": 196410 }, { "epoch": 17.01, "learning_rate": 3.251705793916965e-05, "loss": 0.7963, "step": 196420 }, { "epoch": 17.01, "learning_rate": 3.251488955707182e-05, "loss": 0.7998, "step": 196430 }, { "epoch": 17.01, "learning_rate": 3.2512721174973976e-05, "loss": 0.9466, "step": 196440 }, { "epoch": 17.01, "learning_rate": 3.2510552792876136e-05, "loss": 0.8785, "step": 196450 }, { "epoch": 17.01, "learning_rate": 3.25083844107783e-05, "loss": 0.8209, "step": 196460 }, { "epoch": 17.02, "learning_rate": 3.250621602868046e-05, "loss": 0.8833, "step": 196470 }, { "epoch": 17.02, "learning_rate": 3.250404764658263e-05, "loss": 0.6931, "step": 196480 }, { "epoch": 17.02, "learning_rate": 3.250187926448479e-05, "loss": 0.7337, "step": 196490 }, { "epoch": 17.02, "learning_rate": 3.2499710882386954e-05, "loss": 0.7252, "step": 196500 }, { "epoch": 17.02, "learning_rate": 3.2497542500289114e-05, "loss": 0.7146, "step": 196510 }, { "epoch": 17.02, "learning_rate": 3.249537411819128e-05, "loss": 0.7906, "step": 196520 }, { "epoch": 17.02, "learning_rate": 3.249320573609344e-05, "loss": 0.9648, "step": 196530 }, { "epoch": 17.02, "learning_rate": 3.24910373539956e-05, "loss": 0.9297, "step": 196540 }, { "epoch": 17.02, "learning_rate": 3.2488868971897766e-05, "loss": 0.8928, "step": 196550 }, { "epoch": 17.02, "learning_rate": 3.2486700589799926e-05, "loss": 0.8252, "step": 196560 }, { "epoch": 17.02, "learning_rate": 3.248453220770209e-05, "loss": 0.8343, "step": 196570 }, { "epoch": 17.03, "learning_rate": 3.248236382560425e-05, "loss": 0.7691, "step": 196580 }, { "epoch": 17.03, "learning_rate": 3.248019544350642e-05, "loss": 0.8121, "step": 196590 }, { "epoch": 17.03, "learning_rate": 3.247802706140858e-05, "loss": 0.697, "step": 196600 }, { "epoch": 17.03, "learning_rate": 3.247585867931074e-05, "loss": 0.8856, "step": 196610 }, { "epoch": 17.03, "learning_rate": 3.2473690297212904e-05, "loss": 0.7695, "step": 196620 }, { "epoch": 17.03, "learning_rate": 3.247152191511507e-05, "loss": 0.7485, "step": 196630 }, { "epoch": 17.03, "learning_rate": 3.246935353301723e-05, "loss": 0.7798, "step": 196640 }, { "epoch": 17.03, "learning_rate": 3.246718515091939e-05, "loss": 0.9483, "step": 196650 }, { "epoch": 17.03, "learning_rate": 3.2465016768821556e-05, "loss": 0.7239, "step": 196660 }, { "epoch": 17.03, "learning_rate": 3.2462848386723715e-05, "loss": 0.8017, "step": 196670 }, { "epoch": 17.03, "learning_rate": 3.2460680004625875e-05, "loss": 0.7008, "step": 196680 }, { "epoch": 17.03, "learning_rate": 3.245851162252804e-05, "loss": 0.7539, "step": 196690 }, { "epoch": 17.04, "learning_rate": 3.245634324043021e-05, "loss": 0.6522, "step": 196700 }, { "epoch": 17.04, "learning_rate": 3.245417485833237e-05, "loss": 0.8953, "step": 196710 }, { "epoch": 17.04, "learning_rate": 3.245200647623453e-05, "loss": 0.8225, "step": 196720 }, { "epoch": 17.04, "learning_rate": 3.2449838094136693e-05, "loss": 0.7239, "step": 196730 }, { "epoch": 17.04, "learning_rate": 3.244766971203885e-05, "loss": 1.0135, "step": 196740 }, { "epoch": 17.04, "learning_rate": 3.244550132994102e-05, "loss": 0.7006, "step": 196750 }, { "epoch": 17.04, "learning_rate": 3.244333294784318e-05, "loss": 0.6689, "step": 196760 }, { "epoch": 17.04, "learning_rate": 3.2441164565745345e-05, "loss": 0.8478, "step": 196770 }, { "epoch": 17.04, "learning_rate": 3.2438996183647505e-05, "loss": 0.6731, "step": 196780 }, { "epoch": 17.04, "learning_rate": 3.2436827801549665e-05, "loss": 0.7242, "step": 196790 }, { "epoch": 17.04, "learning_rate": 3.243465941945183e-05, "loss": 0.9902, "step": 196800 }, { "epoch": 17.04, "learning_rate": 3.243249103735399e-05, "loss": 0.9713, "step": 196810 }, { "epoch": 17.05, "learning_rate": 3.243032265525616e-05, "loss": 0.6473, "step": 196820 }, { "epoch": 17.05, "learning_rate": 3.242815427315832e-05, "loss": 0.7296, "step": 196830 }, { "epoch": 17.05, "learning_rate": 3.242598589106048e-05, "loss": 0.7229, "step": 196840 }, { "epoch": 17.05, "learning_rate": 3.242381750896264e-05, "loss": 0.7538, "step": 196850 }, { "epoch": 17.05, "learning_rate": 3.242164912686481e-05, "loss": 0.7799, "step": 196860 }, { "epoch": 17.05, "learning_rate": 3.241948074476697e-05, "loss": 0.8961, "step": 196870 }, { "epoch": 17.05, "learning_rate": 3.241731236266913e-05, "loss": 1.0097, "step": 196880 }, { "epoch": 17.05, "learning_rate": 3.2415143980571295e-05, "loss": 0.8433, "step": 196890 }, { "epoch": 17.05, "learning_rate": 3.2412975598473454e-05, "loss": 0.7222, "step": 196900 }, { "epoch": 17.05, "learning_rate": 3.2410807216375614e-05, "loss": 0.9381, "step": 196910 }, { "epoch": 17.05, "learning_rate": 3.240863883427778e-05, "loss": 0.7509, "step": 196920 }, { "epoch": 17.06, "learning_rate": 3.240647045217995e-05, "loss": 0.7969, "step": 196930 }, { "epoch": 17.06, "learning_rate": 3.2404302070082106e-05, "loss": 0.9189, "step": 196940 }, { "epoch": 17.06, "learning_rate": 3.2402133687984266e-05, "loss": 0.8511, "step": 196950 }, { "epoch": 17.06, "learning_rate": 3.239996530588643e-05, "loss": 0.7976, "step": 196960 }, { "epoch": 17.06, "learning_rate": 3.239779692378859e-05, "loss": 0.9268, "step": 196970 }, { "epoch": 17.06, "learning_rate": 3.239562854169076e-05, "loss": 0.816, "step": 196980 }, { "epoch": 17.06, "learning_rate": 3.239346015959292e-05, "loss": 0.9085, "step": 196990 }, { "epoch": 17.06, "learning_rate": 3.2391291777495084e-05, "loss": 0.8034, "step": 197000 }, { "epoch": 17.06, "learning_rate": 3.2389123395397244e-05, "loss": 0.7663, "step": 197010 }, { "epoch": 17.06, "learning_rate": 3.238695501329941e-05, "loss": 0.8008, "step": 197020 }, { "epoch": 17.06, "learning_rate": 3.238478663120157e-05, "loss": 0.8059, "step": 197030 }, { "epoch": 17.06, "learning_rate": 3.238261824910373e-05, "loss": 0.7502, "step": 197040 }, { "epoch": 17.07, "learning_rate": 3.2380449867005896e-05, "loss": 0.9733, "step": 197050 }, { "epoch": 17.07, "learning_rate": 3.2378281484908056e-05, "loss": 0.855, "step": 197060 }, { "epoch": 17.07, "learning_rate": 3.237611310281022e-05, "loss": 0.8347, "step": 197070 }, { "epoch": 17.07, "learning_rate": 3.237394472071238e-05, "loss": 0.7872, "step": 197080 }, { "epoch": 17.07, "learning_rate": 3.237177633861455e-05, "loss": 0.7177, "step": 197090 }, { "epoch": 17.07, "learning_rate": 3.236960795651671e-05, "loss": 0.8347, "step": 197100 }, { "epoch": 17.07, "learning_rate": 3.236743957441887e-05, "loss": 0.7902, "step": 197110 }, { "epoch": 17.07, "learning_rate": 3.2365271192321034e-05, "loss": 0.6258, "step": 197120 }, { "epoch": 17.07, "learning_rate": 3.23631028102232e-05, "loss": 0.9173, "step": 197130 }, { "epoch": 17.07, "learning_rate": 3.236093442812536e-05, "loss": 0.7285, "step": 197140 }, { "epoch": 17.07, "learning_rate": 3.235876604602752e-05, "loss": 0.9579, "step": 197150 }, { "epoch": 17.08, "learning_rate": 3.2356597663929686e-05, "loss": 0.751, "step": 197160 }, { "epoch": 17.08, "learning_rate": 3.2354429281831845e-05, "loss": 0.9509, "step": 197170 }, { "epoch": 17.08, "learning_rate": 3.2352260899734005e-05, "loss": 0.8091, "step": 197180 }, { "epoch": 17.08, "learning_rate": 3.235009251763617e-05, "loss": 0.9468, "step": 197190 }, { "epoch": 17.08, "learning_rate": 3.234792413553834e-05, "loss": 0.9944, "step": 197200 }, { "epoch": 17.08, "learning_rate": 3.23457557534405e-05, "loss": 0.8189, "step": 197210 }, { "epoch": 17.08, "learning_rate": 3.234358737134266e-05, "loss": 0.802, "step": 197220 }, { "epoch": 17.08, "learning_rate": 3.234141898924482e-05, "loss": 0.7017, "step": 197230 }, { "epoch": 17.08, "learning_rate": 3.233925060714698e-05, "loss": 0.8903, "step": 197240 }, { "epoch": 17.08, "learning_rate": 3.233708222504915e-05, "loss": 0.775, "step": 197250 }, { "epoch": 17.08, "learning_rate": 3.233491384295131e-05, "loss": 0.7565, "step": 197260 }, { "epoch": 17.08, "learning_rate": 3.2332745460853475e-05, "loss": 0.8353, "step": 197270 }, { "epoch": 17.09, "learning_rate": 3.2330577078755635e-05, "loss": 0.8512, "step": 197280 }, { "epoch": 17.09, "learning_rate": 3.23284086966578e-05, "loss": 0.9225, "step": 197290 }, { "epoch": 17.09, "learning_rate": 3.232624031455996e-05, "loss": 1.0002, "step": 197300 }, { "epoch": 17.09, "learning_rate": 3.232407193246212e-05, "loss": 0.9323, "step": 197310 }, { "epoch": 17.09, "learning_rate": 3.232190355036429e-05, "loss": 0.8277, "step": 197320 }, { "epoch": 17.09, "learning_rate": 3.2319735168266447e-05, "loss": 0.8137, "step": 197330 }, { "epoch": 17.09, "learning_rate": 3.231756678616861e-05, "loss": 0.733, "step": 197340 }, { "epoch": 17.09, "learning_rate": 3.231539840407077e-05, "loss": 0.7971, "step": 197350 }, { "epoch": 17.09, "learning_rate": 3.231323002197294e-05, "loss": 0.9345, "step": 197360 }, { "epoch": 17.09, "learning_rate": 3.23110616398751e-05, "loss": 0.6592, "step": 197370 }, { "epoch": 17.09, "learning_rate": 3.230889325777726e-05, "loss": 0.7562, "step": 197380 }, { "epoch": 17.1, "learning_rate": 3.2306724875679425e-05, "loss": 0.7868, "step": 197390 }, { "epoch": 17.1, "learning_rate": 3.230455649358159e-05, "loss": 0.7691, "step": 197400 }, { "epoch": 17.1, "learning_rate": 3.2302388111483744e-05, "loss": 0.822, "step": 197410 }, { "epoch": 17.1, "learning_rate": 3.230021972938591e-05, "loss": 0.8219, "step": 197420 }, { "epoch": 17.1, "learning_rate": 3.2298051347288077e-05, "loss": 0.7869, "step": 197430 }, { "epoch": 17.1, "learning_rate": 3.2295882965190236e-05, "loss": 0.7583, "step": 197440 }, { "epoch": 17.1, "learning_rate": 3.2293714583092396e-05, "loss": 0.7107, "step": 197450 }, { "epoch": 17.1, "learning_rate": 3.229154620099456e-05, "loss": 0.691, "step": 197460 }, { "epoch": 17.1, "learning_rate": 3.228937781889672e-05, "loss": 0.6533, "step": 197470 }, { "epoch": 17.1, "learning_rate": 3.228720943679889e-05, "loss": 0.8177, "step": 197480 }, { "epoch": 17.1, "learning_rate": 3.228504105470105e-05, "loss": 0.9431, "step": 197490 }, { "epoch": 17.1, "learning_rate": 3.2282872672603214e-05, "loss": 0.8335, "step": 197500 }, { "epoch": 17.11, "learning_rate": 3.2280704290505374e-05, "loss": 0.8425, "step": 197510 }, { "epoch": 17.11, "learning_rate": 3.227853590840754e-05, "loss": 0.7972, "step": 197520 }, { "epoch": 17.11, "learning_rate": 3.22763675263097e-05, "loss": 0.9054, "step": 197530 }, { "epoch": 17.11, "learning_rate": 3.227419914421186e-05, "loss": 0.8486, "step": 197540 }, { "epoch": 17.11, "learning_rate": 3.2272030762114026e-05, "loss": 0.8577, "step": 197550 }, { "epoch": 17.11, "learning_rate": 3.2269862380016185e-05, "loss": 0.7362, "step": 197560 }, { "epoch": 17.11, "learning_rate": 3.226769399791835e-05, "loss": 0.7568, "step": 197570 }, { "epoch": 17.11, "learning_rate": 3.226552561582051e-05, "loss": 0.8885, "step": 197580 }, { "epoch": 17.11, "learning_rate": 3.226335723372268e-05, "loss": 0.691, "step": 197590 }, { "epoch": 17.11, "learning_rate": 3.226118885162484e-05, "loss": 0.7549, "step": 197600 }, { "epoch": 17.11, "learning_rate": 3.2259020469527e-05, "loss": 0.9464, "step": 197610 }, { "epoch": 17.12, "learning_rate": 3.2256852087429163e-05, "loss": 0.7371, "step": 197620 }, { "epoch": 17.12, "learning_rate": 3.225468370533133e-05, "loss": 0.876, "step": 197630 }, { "epoch": 17.12, "learning_rate": 3.225251532323349e-05, "loss": 0.8904, "step": 197640 }, { "epoch": 17.12, "learning_rate": 3.225034694113565e-05, "loss": 0.6259, "step": 197650 }, { "epoch": 17.12, "learning_rate": 3.2248178559037815e-05, "loss": 0.7226, "step": 197660 }, { "epoch": 17.12, "learning_rate": 3.2246010176939975e-05, "loss": 0.8392, "step": 197670 }, { "epoch": 17.12, "learning_rate": 3.2243841794842135e-05, "loss": 0.8693, "step": 197680 }, { "epoch": 17.12, "learning_rate": 3.22416734127443e-05, "loss": 0.9809, "step": 197690 }, { "epoch": 17.12, "learning_rate": 3.223950503064647e-05, "loss": 0.7789, "step": 197700 }, { "epoch": 17.12, "learning_rate": 3.223733664854863e-05, "loss": 0.7682, "step": 197710 }, { "epoch": 17.12, "learning_rate": 3.223516826645079e-05, "loss": 0.7757, "step": 197720 }, { "epoch": 17.12, "learning_rate": 3.223299988435295e-05, "loss": 0.8538, "step": 197730 }, { "epoch": 17.13, "learning_rate": 3.223083150225511e-05, "loss": 0.7219, "step": 197740 }, { "epoch": 17.13, "learning_rate": 3.222866312015728e-05, "loss": 0.6806, "step": 197750 }, { "epoch": 17.13, "learning_rate": 3.222649473805944e-05, "loss": 0.6273, "step": 197760 }, { "epoch": 17.13, "learning_rate": 3.2224326355961605e-05, "loss": 0.8932, "step": 197770 }, { "epoch": 17.13, "learning_rate": 3.2222157973863765e-05, "loss": 0.7561, "step": 197780 }, { "epoch": 17.13, "learning_rate": 3.221998959176593e-05, "loss": 0.7903, "step": 197790 }, { "epoch": 17.13, "learning_rate": 3.221782120966809e-05, "loss": 0.9228, "step": 197800 }, { "epoch": 17.13, "learning_rate": 3.221565282757025e-05, "loss": 0.8546, "step": 197810 }, { "epoch": 17.13, "learning_rate": 3.221348444547242e-05, "loss": 0.7609, "step": 197820 }, { "epoch": 17.13, "learning_rate": 3.2211316063374576e-05, "loss": 0.7194, "step": 197830 }, { "epoch": 17.13, "learning_rate": 3.220914768127674e-05, "loss": 0.7548, "step": 197840 }, { "epoch": 17.14, "learning_rate": 3.22069792991789e-05, "loss": 0.9025, "step": 197850 }, { "epoch": 17.14, "learning_rate": 3.220481091708107e-05, "loss": 0.8887, "step": 197860 }, { "epoch": 17.14, "learning_rate": 3.220264253498323e-05, "loss": 0.8019, "step": 197870 }, { "epoch": 17.14, "learning_rate": 3.220047415288539e-05, "loss": 0.83, "step": 197880 }, { "epoch": 17.14, "learning_rate": 3.2198305770787554e-05, "loss": 0.8082, "step": 197890 }, { "epoch": 17.14, "learning_rate": 3.219613738868972e-05, "loss": 0.8119, "step": 197900 }, { "epoch": 17.14, "learning_rate": 3.219396900659188e-05, "loss": 0.888, "step": 197910 }, { "epoch": 17.14, "learning_rate": 3.219180062449404e-05, "loss": 0.8579, "step": 197920 }, { "epoch": 17.14, "learning_rate": 3.2189632242396206e-05, "loss": 0.7277, "step": 197930 }, { "epoch": 17.14, "learning_rate": 3.2187463860298366e-05, "loss": 0.7949, "step": 197940 }, { "epoch": 17.14, "learning_rate": 3.2185295478200526e-05, "loss": 0.789, "step": 197950 }, { "epoch": 17.14, "learning_rate": 3.218312709610269e-05, "loss": 0.6519, "step": 197960 }, { "epoch": 17.15, "learning_rate": 3.218095871400485e-05, "loss": 0.7662, "step": 197970 }, { "epoch": 17.15, "learning_rate": 3.217879033190702e-05, "loss": 0.8252, "step": 197980 }, { "epoch": 17.15, "learning_rate": 3.217662194980918e-05, "loss": 0.915, "step": 197990 }, { "epoch": 17.15, "learning_rate": 3.2174453567711344e-05, "loss": 0.8421, "step": 198000 }, { "epoch": 17.15, "learning_rate": 3.2172285185613504e-05, "loss": 0.8018, "step": 198010 }, { "epoch": 17.15, "learning_rate": 3.217011680351567e-05, "loss": 0.8837, "step": 198020 }, { "epoch": 17.15, "learning_rate": 3.216794842141783e-05, "loss": 0.8067, "step": 198030 }, { "epoch": 17.15, "learning_rate": 3.216578003931999e-05, "loss": 0.7564, "step": 198040 }, { "epoch": 17.15, "learning_rate": 3.2163611657222156e-05, "loss": 0.9575, "step": 198050 }, { "epoch": 17.15, "learning_rate": 3.216144327512432e-05, "loss": 0.7738, "step": 198060 }, { "epoch": 17.15, "learning_rate": 3.215927489302648e-05, "loss": 0.7713, "step": 198070 }, { "epoch": 17.15, "learning_rate": 3.215710651092864e-05, "loss": 0.862, "step": 198080 }, { "epoch": 17.16, "learning_rate": 3.215493812883081e-05, "loss": 0.8324, "step": 198090 }, { "epoch": 17.16, "learning_rate": 3.215276974673297e-05, "loss": 0.8441, "step": 198100 }, { "epoch": 17.16, "learning_rate": 3.215060136463513e-05, "loss": 0.7764, "step": 198110 }, { "epoch": 17.16, "learning_rate": 3.214843298253729e-05, "loss": 0.8628, "step": 198120 }, { "epoch": 17.16, "learning_rate": 3.214626460043946e-05, "loss": 0.7538, "step": 198130 }, { "epoch": 17.16, "learning_rate": 3.214409621834162e-05, "loss": 0.774, "step": 198140 }, { "epoch": 17.16, "learning_rate": 3.214192783624378e-05, "loss": 0.7201, "step": 198150 }, { "epoch": 17.16, "learning_rate": 3.2139759454145945e-05, "loss": 0.8347, "step": 198160 }, { "epoch": 17.16, "learning_rate": 3.2137591072048105e-05, "loss": 0.8002, "step": 198170 }, { "epoch": 17.16, "learning_rate": 3.2135422689950265e-05, "loss": 0.6524, "step": 198180 }, { "epoch": 17.16, "learning_rate": 3.213325430785243e-05, "loss": 0.8874, "step": 198190 }, { "epoch": 17.17, "learning_rate": 3.21310859257546e-05, "loss": 0.8749, "step": 198200 }, { "epoch": 17.17, "learning_rate": 3.212891754365676e-05, "loss": 0.9768, "step": 198210 }, { "epoch": 17.17, "learning_rate": 3.2126749161558917e-05, "loss": 0.978, "step": 198220 }, { "epoch": 17.17, "learning_rate": 3.212458077946108e-05, "loss": 0.8143, "step": 198230 }, { "epoch": 17.17, "learning_rate": 3.212241239736324e-05, "loss": 0.8175, "step": 198240 }, { "epoch": 17.17, "learning_rate": 3.212024401526541e-05, "loss": 0.872, "step": 198250 }, { "epoch": 17.17, "learning_rate": 3.211807563316757e-05, "loss": 0.8875, "step": 198260 }, { "epoch": 17.17, "learning_rate": 3.2115907251069735e-05, "loss": 0.8642, "step": 198270 }, { "epoch": 17.17, "learning_rate": 3.2113738868971895e-05, "loss": 0.84, "step": 198280 }, { "epoch": 17.17, "learning_rate": 3.211157048687406e-05, "loss": 0.8314, "step": 198290 }, { "epoch": 17.17, "learning_rate": 3.210940210477622e-05, "loss": 0.7872, "step": 198300 }, { "epoch": 17.17, "learning_rate": 3.210723372267838e-05, "loss": 0.8335, "step": 198310 }, { "epoch": 17.18, "learning_rate": 3.2105065340580547e-05, "loss": 0.918, "step": 198320 }, { "epoch": 17.18, "learning_rate": 3.2102896958482706e-05, "loss": 0.9162, "step": 198330 }, { "epoch": 17.18, "learning_rate": 3.210072857638487e-05, "loss": 0.97, "step": 198340 }, { "epoch": 17.18, "learning_rate": 3.209856019428703e-05, "loss": 0.9548, "step": 198350 }, { "epoch": 17.18, "learning_rate": 3.20963918121892e-05, "loss": 0.8468, "step": 198360 }, { "epoch": 17.18, "learning_rate": 3.209422343009136e-05, "loss": 0.7441, "step": 198370 }, { "epoch": 17.18, "learning_rate": 3.209205504799352e-05, "loss": 0.7147, "step": 198380 }, { "epoch": 17.18, "learning_rate": 3.2089886665895684e-05, "loss": 0.8687, "step": 198390 }, { "epoch": 17.18, "learning_rate": 3.208771828379785e-05, "loss": 0.9167, "step": 198400 }, { "epoch": 17.18, "learning_rate": 3.208554990170001e-05, "loss": 0.7908, "step": 198410 }, { "epoch": 17.18, "learning_rate": 3.208338151960217e-05, "loss": 0.7721, "step": 198420 }, { "epoch": 17.19, "learning_rate": 3.2081213137504336e-05, "loss": 0.958, "step": 198430 }, { "epoch": 17.19, "learning_rate": 3.2079044755406496e-05, "loss": 0.7864, "step": 198440 }, { "epoch": 17.19, "learning_rate": 3.2076876373308655e-05, "loss": 0.7137, "step": 198450 }, { "epoch": 17.19, "learning_rate": 3.207470799121082e-05, "loss": 0.8583, "step": 198460 }, { "epoch": 17.19, "learning_rate": 3.207253960911299e-05, "loss": 0.912, "step": 198470 }, { "epoch": 17.19, "learning_rate": 3.207037122701515e-05, "loss": 1.0219, "step": 198480 }, { "epoch": 17.19, "learning_rate": 3.206820284491731e-05, "loss": 0.7861, "step": 198490 }, { "epoch": 17.19, "learning_rate": 3.2066034462819474e-05, "loss": 0.7838, "step": 198500 }, { "epoch": 17.19, "learning_rate": 3.2063866080721633e-05, "loss": 0.7448, "step": 198510 }, { "epoch": 17.19, "learning_rate": 3.20616976986238e-05, "loss": 0.9436, "step": 198520 }, { "epoch": 17.19, "learning_rate": 3.205952931652596e-05, "loss": 0.8282, "step": 198530 }, { "epoch": 17.19, "learning_rate": 3.205736093442812e-05, "loss": 0.9139, "step": 198540 }, { "epoch": 17.2, "learning_rate": 3.2055192552330285e-05, "loss": 0.9632, "step": 198550 }, { "epoch": 17.2, "learning_rate": 3.205302417023245e-05, "loss": 0.9256, "step": 198560 }, { "epoch": 17.2, "learning_rate": 3.205085578813461e-05, "loss": 0.7763, "step": 198570 }, { "epoch": 17.2, "learning_rate": 3.204868740603677e-05, "loss": 0.8311, "step": 198580 }, { "epoch": 17.2, "learning_rate": 3.204651902393894e-05, "loss": 0.7483, "step": 198590 }, { "epoch": 17.2, "learning_rate": 3.20443506418411e-05, "loss": 0.7814, "step": 198600 }, { "epoch": 17.2, "learning_rate": 3.204218225974326e-05, "loss": 0.6129, "step": 198610 }, { "epoch": 17.2, "learning_rate": 3.204001387764542e-05, "loss": 0.9395, "step": 198620 }, { "epoch": 17.2, "learning_rate": 3.203784549554759e-05, "loss": 0.8091, "step": 198630 }, { "epoch": 17.2, "learning_rate": 3.203567711344975e-05, "loss": 0.8839, "step": 198640 }, { "epoch": 17.2, "learning_rate": 3.203350873135191e-05, "loss": 0.9011, "step": 198650 }, { "epoch": 17.21, "learning_rate": 3.2031340349254075e-05, "loss": 0.9273, "step": 198660 }, { "epoch": 17.21, "learning_rate": 3.2029171967156235e-05, "loss": 0.7363, "step": 198670 }, { "epoch": 17.21, "learning_rate": 3.2027003585058394e-05, "loss": 0.7712, "step": 198680 }, { "epoch": 17.21, "learning_rate": 3.202483520296056e-05, "loss": 0.8529, "step": 198690 }, { "epoch": 17.21, "learning_rate": 3.202266682086273e-05, "loss": 0.6751, "step": 198700 }, { "epoch": 17.21, "learning_rate": 3.202049843876489e-05, "loss": 0.8335, "step": 198710 }, { "epoch": 17.21, "learning_rate": 3.2018330056667046e-05, "loss": 0.8539, "step": 198720 }, { "epoch": 17.21, "learning_rate": 3.201616167456921e-05, "loss": 1.0014, "step": 198730 }, { "epoch": 17.21, "learning_rate": 3.201399329247137e-05, "loss": 0.8366, "step": 198740 }, { "epoch": 17.21, "learning_rate": 3.201182491037354e-05, "loss": 0.9085, "step": 198750 }, { "epoch": 17.21, "learning_rate": 3.20096565282757e-05, "loss": 0.8506, "step": 198760 }, { "epoch": 17.21, "learning_rate": 3.2007488146177865e-05, "loss": 0.8751, "step": 198770 }, { "epoch": 17.22, "learning_rate": 3.2005319764080024e-05, "loss": 0.645, "step": 198780 }, { "epoch": 17.22, "learning_rate": 3.200315138198219e-05, "loss": 0.669, "step": 198790 }, { "epoch": 17.22, "learning_rate": 3.200098299988435e-05, "loss": 0.7216, "step": 198800 }, { "epoch": 17.22, "learning_rate": 3.199881461778651e-05, "loss": 0.7603, "step": 198810 }, { "epoch": 17.22, "learning_rate": 3.1996646235688676e-05, "loss": 0.644, "step": 198820 }, { "epoch": 17.22, "learning_rate": 3.199447785359084e-05, "loss": 0.8181, "step": 198830 }, { "epoch": 17.22, "learning_rate": 3.1992309471493e-05, "loss": 0.9408, "step": 198840 }, { "epoch": 17.22, "learning_rate": 3.199014108939516e-05, "loss": 0.6755, "step": 198850 }, { "epoch": 17.22, "learning_rate": 3.198797270729733e-05, "loss": 0.7618, "step": 198860 }, { "epoch": 17.22, "learning_rate": 3.198580432519949e-05, "loss": 0.7424, "step": 198870 }, { "epoch": 17.22, "learning_rate": 3.198363594310165e-05, "loss": 0.9985, "step": 198880 }, { "epoch": 17.23, "learning_rate": 3.1981467561003814e-05, "loss": 0.8232, "step": 198890 }, { "epoch": 17.23, "learning_rate": 3.197929917890598e-05, "loss": 0.6703, "step": 198900 }, { "epoch": 17.23, "learning_rate": 3.197713079680814e-05, "loss": 0.8603, "step": 198910 }, { "epoch": 17.23, "learning_rate": 3.19749624147103e-05, "loss": 0.7829, "step": 198920 }, { "epoch": 17.23, "learning_rate": 3.1972794032612466e-05, "loss": 0.9143, "step": 198930 }, { "epoch": 17.23, "learning_rate": 3.1970625650514626e-05, "loss": 0.9211, "step": 198940 }, { "epoch": 17.23, "learning_rate": 3.1968457268416785e-05, "loss": 0.7655, "step": 198950 }, { "epoch": 17.23, "learning_rate": 3.196628888631895e-05, "loss": 0.8293, "step": 198960 }, { "epoch": 17.23, "learning_rate": 3.196412050422112e-05, "loss": 0.7552, "step": 198970 }, { "epoch": 17.23, "learning_rate": 3.196195212212328e-05, "loss": 0.6668, "step": 198980 }, { "epoch": 17.23, "learning_rate": 3.195978374002544e-05, "loss": 0.7969, "step": 198990 }, { "epoch": 17.23, "learning_rate": 3.1957615357927604e-05, "loss": 0.7791, "step": 199000 }, { "epoch": 17.24, "learning_rate": 3.195544697582976e-05, "loss": 0.8673, "step": 199010 }, { "epoch": 17.24, "learning_rate": 3.195327859373193e-05, "loss": 0.9121, "step": 199020 }, { "epoch": 17.24, "learning_rate": 3.195111021163409e-05, "loss": 0.9848, "step": 199030 }, { "epoch": 17.24, "learning_rate": 3.194894182953625e-05, "loss": 0.9819, "step": 199040 }, { "epoch": 17.24, "learning_rate": 3.1946773447438415e-05, "loss": 0.8057, "step": 199050 }, { "epoch": 17.24, "learning_rate": 3.194460506534058e-05, "loss": 0.9078, "step": 199060 }, { "epoch": 17.24, "learning_rate": 3.194243668324274e-05, "loss": 0.8262, "step": 199070 }, { "epoch": 17.24, "learning_rate": 3.19402683011449e-05, "loss": 0.8514, "step": 199080 }, { "epoch": 17.24, "learning_rate": 3.193809991904707e-05, "loss": 0.8485, "step": 199090 }, { "epoch": 17.24, "learning_rate": 3.193593153694923e-05, "loss": 0.7453, "step": 199100 }, { "epoch": 17.24, "learning_rate": 3.1933763154851387e-05, "loss": 0.8845, "step": 199110 }, { "epoch": 17.25, "learning_rate": 3.193159477275355e-05, "loss": 0.6946, "step": 199120 }, { "epoch": 17.25, "learning_rate": 3.192942639065572e-05, "loss": 0.8464, "step": 199130 }, { "epoch": 17.25, "learning_rate": 3.192725800855788e-05, "loss": 0.7423, "step": 199140 }, { "epoch": 17.25, "learning_rate": 3.192508962646004e-05, "loss": 0.6677, "step": 199150 }, { "epoch": 17.25, "learning_rate": 3.1922921244362205e-05, "loss": 0.7005, "step": 199160 }, { "epoch": 17.25, "learning_rate": 3.1920752862264365e-05, "loss": 0.8953, "step": 199170 }, { "epoch": 17.25, "learning_rate": 3.1918584480166524e-05, "loss": 0.7576, "step": 199180 }, { "epoch": 17.25, "learning_rate": 3.191641609806869e-05, "loss": 0.9373, "step": 199190 }, { "epoch": 17.25, "learning_rate": 3.191424771597086e-05, "loss": 0.7547, "step": 199200 }, { "epoch": 17.25, "learning_rate": 3.191207933387302e-05, "loss": 0.8877, "step": 199210 }, { "epoch": 17.25, "learning_rate": 3.1909910951775176e-05, "loss": 0.9277, "step": 199220 }, { "epoch": 17.25, "learning_rate": 3.190774256967734e-05, "loss": 0.8849, "step": 199230 }, { "epoch": 17.26, "learning_rate": 3.19055741875795e-05, "loss": 0.8485, "step": 199240 }, { "epoch": 17.26, "learning_rate": 3.190340580548167e-05, "loss": 0.8373, "step": 199250 }, { "epoch": 17.26, "learning_rate": 3.190123742338383e-05, "loss": 0.7988, "step": 199260 }, { "epoch": 17.26, "learning_rate": 3.1899069041285995e-05, "loss": 0.9814, "step": 199270 }, { "epoch": 17.26, "learning_rate": 3.1896900659188154e-05, "loss": 0.7923, "step": 199280 }, { "epoch": 17.26, "learning_rate": 3.189473227709032e-05, "loss": 0.9608, "step": 199290 }, { "epoch": 17.26, "learning_rate": 3.189256389499248e-05, "loss": 0.8987, "step": 199300 }, { "epoch": 17.26, "learning_rate": 3.189039551289464e-05, "loss": 0.7267, "step": 199310 }, { "epoch": 17.26, "learning_rate": 3.1888227130796806e-05, "loss": 0.8551, "step": 199320 }, { "epoch": 17.26, "learning_rate": 3.188605874869897e-05, "loss": 0.849, "step": 199330 }, { "epoch": 17.26, "learning_rate": 3.188389036660113e-05, "loss": 0.8531, "step": 199340 }, { "epoch": 17.26, "learning_rate": 3.188172198450329e-05, "loss": 0.9422, "step": 199350 }, { "epoch": 17.27, "learning_rate": 3.187955360240546e-05, "loss": 0.8339, "step": 199360 }, { "epoch": 17.27, "learning_rate": 3.187738522030762e-05, "loss": 0.7877, "step": 199370 }, { "epoch": 17.27, "learning_rate": 3.187521683820978e-05, "loss": 0.9266, "step": 199380 }, { "epoch": 17.27, "learning_rate": 3.1873048456111944e-05, "loss": 0.9683, "step": 199390 }, { "epoch": 17.27, "learning_rate": 3.187088007401411e-05, "loss": 0.9737, "step": 199400 }, { "epoch": 17.27, "learning_rate": 3.186871169191627e-05, "loss": 0.7985, "step": 199410 }, { "epoch": 17.27, "learning_rate": 3.186654330981843e-05, "loss": 0.8346, "step": 199420 }, { "epoch": 17.27, "learning_rate": 3.1864374927720596e-05, "loss": 1.0763, "step": 199430 }, { "epoch": 17.27, "learning_rate": 3.1862206545622756e-05, "loss": 0.8585, "step": 199440 }, { "epoch": 17.27, "learning_rate": 3.1860038163524915e-05, "loss": 0.8861, "step": 199450 }, { "epoch": 17.27, "learning_rate": 3.185786978142708e-05, "loss": 0.8749, "step": 199460 }, { "epoch": 17.28, "learning_rate": 3.185570139932925e-05, "loss": 0.7786, "step": 199470 }, { "epoch": 17.28, "learning_rate": 3.185353301723141e-05, "loss": 0.8723, "step": 199480 }, { "epoch": 17.28, "learning_rate": 3.185136463513357e-05, "loss": 0.859, "step": 199490 }, { "epoch": 17.28, "learning_rate": 3.1849196253035734e-05, "loss": 0.8799, "step": 199500 }, { "epoch": 17.28, "learning_rate": 3.184702787093789e-05, "loss": 1.0301, "step": 199510 }, { "epoch": 17.28, "learning_rate": 3.184485948884006e-05, "loss": 0.7209, "step": 199520 }, { "epoch": 17.28, "learning_rate": 3.184269110674222e-05, "loss": 0.8815, "step": 199530 }, { "epoch": 17.28, "learning_rate": 3.1840522724644386e-05, "loss": 0.7729, "step": 199540 }, { "epoch": 17.28, "learning_rate": 3.1838354342546545e-05, "loss": 0.7597, "step": 199550 }, { "epoch": 17.28, "learning_rate": 3.183618596044871e-05, "loss": 0.8487, "step": 199560 }, { "epoch": 17.28, "learning_rate": 3.183401757835087e-05, "loss": 0.8373, "step": 199570 }, { "epoch": 17.28, "learning_rate": 3.183184919625303e-05, "loss": 0.7032, "step": 199580 }, { "epoch": 17.29, "learning_rate": 3.18296808141552e-05, "loss": 0.8307, "step": 199590 }, { "epoch": 17.29, "learning_rate": 3.182751243205736e-05, "loss": 0.9331, "step": 199600 }, { "epoch": 17.29, "learning_rate": 3.1825344049959516e-05, "loss": 0.9585, "step": 199610 }, { "epoch": 17.29, "learning_rate": 3.182317566786168e-05, "loss": 0.8587, "step": 199620 }, { "epoch": 17.29, "learning_rate": 3.182100728576385e-05, "loss": 0.8782, "step": 199630 }, { "epoch": 17.29, "learning_rate": 3.181883890366601e-05, "loss": 0.6384, "step": 199640 }, { "epoch": 17.29, "learning_rate": 3.181667052156817e-05, "loss": 0.9607, "step": 199650 }, { "epoch": 17.29, "learning_rate": 3.1814502139470335e-05, "loss": 0.841, "step": 199660 }, { "epoch": 17.29, "learning_rate": 3.1812333757372494e-05, "loss": 0.74, "step": 199670 }, { "epoch": 17.29, "learning_rate": 3.1810165375274654e-05, "loss": 0.8612, "step": 199680 }, { "epoch": 17.29, "learning_rate": 3.180799699317682e-05, "loss": 0.8049, "step": 199690 }, { "epoch": 17.3, "learning_rate": 3.180582861107899e-05, "loss": 0.7045, "step": 199700 }, { "epoch": 17.3, "learning_rate": 3.1803660228981146e-05, "loss": 0.9087, "step": 199710 }, { "epoch": 17.3, "learning_rate": 3.1801491846883306e-05, "loss": 0.7805, "step": 199720 }, { "epoch": 17.3, "learning_rate": 3.179932346478547e-05, "loss": 0.8402, "step": 199730 }, { "epoch": 17.3, "learning_rate": 3.179715508268763e-05, "loss": 0.7396, "step": 199740 }, { "epoch": 17.3, "learning_rate": 3.17949867005898e-05, "loss": 0.8339, "step": 199750 }, { "epoch": 17.3, "learning_rate": 3.179281831849196e-05, "loss": 0.8685, "step": 199760 }, { "epoch": 17.3, "learning_rate": 3.1790649936394124e-05, "loss": 0.8189, "step": 199770 }, { "epoch": 17.3, "learning_rate": 3.1788481554296284e-05, "loss": 0.9403, "step": 199780 }, { "epoch": 17.3, "learning_rate": 3.178631317219845e-05, "loss": 0.7671, "step": 199790 }, { "epoch": 17.3, "learning_rate": 3.178414479010061e-05, "loss": 0.7804, "step": 199800 }, { "epoch": 17.3, "learning_rate": 3.178197640800277e-05, "loss": 0.8534, "step": 199810 }, { "epoch": 17.31, "learning_rate": 3.1779808025904936e-05, "loss": 0.741, "step": 199820 }, { "epoch": 17.31, "learning_rate": 3.17776396438071e-05, "loss": 0.8598, "step": 199830 }, { "epoch": 17.31, "learning_rate": 3.177547126170926e-05, "loss": 0.8625, "step": 199840 }, { "epoch": 17.31, "learning_rate": 3.177330287961142e-05, "loss": 0.918, "step": 199850 }, { "epoch": 17.31, "learning_rate": 3.177113449751359e-05, "loss": 0.8758, "step": 199860 }, { "epoch": 17.31, "learning_rate": 3.176896611541575e-05, "loss": 0.8744, "step": 199870 }, { "epoch": 17.31, "learning_rate": 3.176679773331791e-05, "loss": 0.8571, "step": 199880 }, { "epoch": 17.31, "learning_rate": 3.1764629351220074e-05, "loss": 0.7757, "step": 199890 }, { "epoch": 17.31, "learning_rate": 3.176246096912224e-05, "loss": 0.8035, "step": 199900 }, { "epoch": 17.31, "learning_rate": 3.17602925870244e-05, "loss": 0.8121, "step": 199910 }, { "epoch": 17.31, "learning_rate": 3.175812420492656e-05, "loss": 0.8302, "step": 199920 }, { "epoch": 17.32, "learning_rate": 3.1755955822828726e-05, "loss": 0.7418, "step": 199930 }, { "epoch": 17.32, "learning_rate": 3.1753787440730885e-05, "loss": 0.829, "step": 199940 }, { "epoch": 17.32, "learning_rate": 3.1751619058633045e-05, "loss": 0.9467, "step": 199950 }, { "epoch": 17.32, "learning_rate": 3.174945067653521e-05, "loss": 0.7346, "step": 199960 }, { "epoch": 17.32, "learning_rate": 3.174728229443738e-05, "loss": 0.9377, "step": 199970 }, { "epoch": 17.32, "learning_rate": 3.174511391233954e-05, "loss": 1.0249, "step": 199980 }, { "epoch": 17.32, "learning_rate": 3.17429455302417e-05, "loss": 0.7582, "step": 199990 }, { "epoch": 17.32, "learning_rate": 3.1740777148143863e-05, "loss": 0.7006, "step": 200000 }, { "epoch": 17.32, "learning_rate": 3.173860876604602e-05, "loss": 0.7943, "step": 200010 }, { "epoch": 17.32, "learning_rate": 3.173644038394819e-05, "loss": 0.8865, "step": 200020 }, { "epoch": 17.32, "learning_rate": 3.173427200185035e-05, "loss": 0.9461, "step": 200030 }, { "epoch": 17.32, "learning_rate": 3.1732103619752515e-05, "loss": 0.8513, "step": 200040 }, { "epoch": 17.33, "learning_rate": 3.1729935237654675e-05, "loss": 0.6924, "step": 200050 }, { "epoch": 17.33, "learning_rate": 3.172776685555684e-05, "loss": 0.8198, "step": 200060 }, { "epoch": 17.33, "learning_rate": 3.1725598473459e-05, "loss": 0.8794, "step": 200070 }, { "epoch": 17.33, "learning_rate": 3.172343009136116e-05, "loss": 0.6952, "step": 200080 }, { "epoch": 17.33, "learning_rate": 3.172126170926333e-05, "loss": 0.8612, "step": 200090 }, { "epoch": 17.33, "learning_rate": 3.1719093327165493e-05, "loss": 0.921, "step": 200100 }, { "epoch": 17.33, "learning_rate": 3.1716924945067646e-05, "loss": 0.7304, "step": 200110 }, { "epoch": 17.33, "learning_rate": 3.171475656296981e-05, "loss": 0.8597, "step": 200120 }, { "epoch": 17.33, "learning_rate": 3.171258818087198e-05, "loss": 0.7818, "step": 200130 }, { "epoch": 17.33, "learning_rate": 3.171041979877414e-05, "loss": 0.7215, "step": 200140 }, { "epoch": 17.33, "learning_rate": 3.17082514166763e-05, "loss": 0.7397, "step": 200150 }, { "epoch": 17.34, "learning_rate": 3.1706083034578465e-05, "loss": 0.7875, "step": 200160 }, { "epoch": 17.34, "learning_rate": 3.1703914652480624e-05, "loss": 0.6522, "step": 200170 }, { "epoch": 17.34, "learning_rate": 3.1701746270382784e-05, "loss": 0.7937, "step": 200180 }, { "epoch": 17.34, "learning_rate": 3.169957788828495e-05, "loss": 0.8862, "step": 200190 }, { "epoch": 17.34, "learning_rate": 3.169740950618712e-05, "loss": 0.9365, "step": 200200 }, { "epoch": 17.34, "learning_rate": 3.1695241124089276e-05, "loss": 0.912, "step": 200210 }, { "epoch": 17.34, "learning_rate": 3.1693072741991436e-05, "loss": 0.7941, "step": 200220 }, { "epoch": 17.34, "learning_rate": 3.16909043598936e-05, "loss": 0.8725, "step": 200230 }, { "epoch": 17.34, "learning_rate": 3.168873597779576e-05, "loss": 0.911, "step": 200240 }, { "epoch": 17.34, "learning_rate": 3.168656759569793e-05, "loss": 0.8344, "step": 200250 }, { "epoch": 17.34, "learning_rate": 3.168439921360009e-05, "loss": 0.8074, "step": 200260 }, { "epoch": 17.34, "learning_rate": 3.1682230831502254e-05, "loss": 0.8445, "step": 200270 }, { "epoch": 17.35, "learning_rate": 3.1680062449404414e-05, "loss": 0.6372, "step": 200280 }, { "epoch": 17.35, "learning_rate": 3.167789406730658e-05, "loss": 0.8344, "step": 200290 }, { "epoch": 17.35, "learning_rate": 3.167572568520874e-05, "loss": 0.8296, "step": 200300 }, { "epoch": 17.35, "learning_rate": 3.16735573031109e-05, "loss": 0.8011, "step": 200310 }, { "epoch": 17.35, "learning_rate": 3.1671388921013066e-05, "loss": 0.8345, "step": 200320 }, { "epoch": 17.35, "learning_rate": 3.166922053891523e-05, "loss": 0.9745, "step": 200330 }, { "epoch": 17.35, "learning_rate": 3.166705215681739e-05, "loss": 0.7275, "step": 200340 }, { "epoch": 17.35, "learning_rate": 3.166488377471955e-05, "loss": 0.72, "step": 200350 }, { "epoch": 17.35, "learning_rate": 3.166271539262172e-05, "loss": 0.8167, "step": 200360 }, { "epoch": 17.35, "learning_rate": 3.166054701052388e-05, "loss": 0.8338, "step": 200370 }, { "epoch": 17.35, "learning_rate": 3.165837862842604e-05, "loss": 0.9578, "step": 200380 }, { "epoch": 17.36, "learning_rate": 3.1656210246328204e-05, "loss": 0.7894, "step": 200390 }, { "epoch": 17.36, "learning_rate": 3.165404186423037e-05, "loss": 0.8665, "step": 200400 }, { "epoch": 17.36, "learning_rate": 3.165187348213253e-05, "loss": 0.7827, "step": 200410 }, { "epoch": 17.36, "learning_rate": 3.164970510003469e-05, "loss": 0.9189, "step": 200420 }, { "epoch": 17.36, "learning_rate": 3.1647536717936856e-05, "loss": 0.7496, "step": 200430 }, { "epoch": 17.36, "learning_rate": 3.1645368335839015e-05, "loss": 0.8113, "step": 200440 }, { "epoch": 17.36, "learning_rate": 3.1643199953741175e-05, "loss": 0.9073, "step": 200450 }, { "epoch": 17.36, "learning_rate": 3.164103157164334e-05, "loss": 0.8137, "step": 200460 }, { "epoch": 17.36, "learning_rate": 3.163886318954551e-05, "loss": 0.9716, "step": 200470 }, { "epoch": 17.36, "learning_rate": 3.163669480744767e-05, "loss": 0.9085, "step": 200480 }, { "epoch": 17.36, "learning_rate": 3.163452642534983e-05, "loss": 0.7798, "step": 200490 }, { "epoch": 17.36, "learning_rate": 3.163235804325199e-05, "loss": 0.9301, "step": 200500 }, { "epoch": 17.37, "learning_rate": 3.163018966115415e-05, "loss": 0.8299, "step": 200510 }, { "epoch": 17.37, "learning_rate": 3.162802127905632e-05, "loss": 0.8615, "step": 200520 }, { "epoch": 17.37, "learning_rate": 3.162585289695848e-05, "loss": 0.8237, "step": 200530 }, { "epoch": 17.37, "learning_rate": 3.1623684514860645e-05, "loss": 0.6973, "step": 200540 }, { "epoch": 17.37, "learning_rate": 3.1621516132762805e-05, "loss": 0.639, "step": 200550 }, { "epoch": 17.37, "learning_rate": 3.161934775066497e-05, "loss": 0.7634, "step": 200560 }, { "epoch": 17.37, "learning_rate": 3.161717936856713e-05, "loss": 0.724, "step": 200570 }, { "epoch": 17.37, "learning_rate": 3.161501098646929e-05, "loss": 0.7617, "step": 200580 }, { "epoch": 17.37, "learning_rate": 3.161284260437146e-05, "loss": 0.9551, "step": 200590 }, { "epoch": 17.37, "learning_rate": 3.161067422227362e-05, "loss": 0.767, "step": 200600 }, { "epoch": 17.37, "learning_rate": 3.160850584017578e-05, "loss": 0.7843, "step": 200610 }, { "epoch": 17.37, "learning_rate": 3.160633745807794e-05, "loss": 0.8998, "step": 200620 }, { "epoch": 17.38, "learning_rate": 3.160416907598011e-05, "loss": 0.7978, "step": 200630 }, { "epoch": 17.38, "learning_rate": 3.160200069388227e-05, "loss": 0.8458, "step": 200640 }, { "epoch": 17.38, "learning_rate": 3.159983231178443e-05, "loss": 0.8279, "step": 200650 }, { "epoch": 17.38, "learning_rate": 3.1597663929686595e-05, "loss": 0.7022, "step": 200660 }, { "epoch": 17.38, "learning_rate": 3.1595495547588754e-05, "loss": 0.8396, "step": 200670 }, { "epoch": 17.38, "learning_rate": 3.1593327165490914e-05, "loss": 0.7076, "step": 200680 }, { "epoch": 17.38, "learning_rate": 3.159115878339308e-05, "loss": 0.7955, "step": 200690 }, { "epoch": 17.38, "learning_rate": 3.1588990401295247e-05, "loss": 0.6726, "step": 200700 }, { "epoch": 17.38, "learning_rate": 3.1586822019197406e-05, "loss": 0.7502, "step": 200710 }, { "epoch": 17.38, "learning_rate": 3.1584653637099566e-05, "loss": 0.8241, "step": 200720 }, { "epoch": 17.38, "learning_rate": 3.158248525500173e-05, "loss": 0.7424, "step": 200730 }, { "epoch": 17.39, "learning_rate": 3.158031687290389e-05, "loss": 0.8122, "step": 200740 }, { "epoch": 17.39, "learning_rate": 3.157814849080606e-05, "loss": 0.793, "step": 200750 }, { "epoch": 17.39, "learning_rate": 3.157598010870822e-05, "loss": 0.897, "step": 200760 }, { "epoch": 17.39, "learning_rate": 3.1573811726610384e-05, "loss": 0.8281, "step": 200770 }, { "epoch": 17.39, "learning_rate": 3.1571643344512544e-05, "loss": 0.9187, "step": 200780 }, { "epoch": 17.39, "learning_rate": 3.156947496241471e-05, "loss": 0.7773, "step": 200790 }, { "epoch": 17.39, "learning_rate": 3.156730658031687e-05, "loss": 0.7825, "step": 200800 }, { "epoch": 17.39, "learning_rate": 3.156513819821903e-05, "loss": 0.7945, "step": 200810 }, { "epoch": 17.39, "learning_rate": 3.1562969816121196e-05, "loss": 0.8703, "step": 200820 }, { "epoch": 17.39, "learning_rate": 3.156080143402336e-05, "loss": 0.8686, "step": 200830 }, { "epoch": 17.39, "learning_rate": 3.155863305192552e-05, "loss": 0.9202, "step": 200840 }, { "epoch": 17.39, "learning_rate": 3.155646466982768e-05, "loss": 0.9074, "step": 200850 }, { "epoch": 17.4, "learning_rate": 3.155429628772985e-05, "loss": 0.9066, "step": 200860 }, { "epoch": 17.4, "learning_rate": 3.155212790563201e-05, "loss": 0.8078, "step": 200870 }, { "epoch": 17.4, "learning_rate": 3.154995952353417e-05, "loss": 0.8129, "step": 200880 }, { "epoch": 17.4, "learning_rate": 3.1547791141436333e-05, "loss": 0.9002, "step": 200890 }, { "epoch": 17.4, "learning_rate": 3.15456227593385e-05, "loss": 0.7565, "step": 200900 }, { "epoch": 17.4, "learning_rate": 3.154345437724066e-05, "loss": 0.9274, "step": 200910 }, { "epoch": 17.4, "learning_rate": 3.154128599514282e-05, "loss": 0.629, "step": 200920 }, { "epoch": 17.4, "learning_rate": 3.1539117613044985e-05, "loss": 0.6904, "step": 200930 }, { "epoch": 17.4, "learning_rate": 3.1536949230947145e-05, "loss": 0.7022, "step": 200940 }, { "epoch": 17.4, "learning_rate": 3.1534780848849305e-05, "loss": 0.998, "step": 200950 }, { "epoch": 17.4, "learning_rate": 3.153261246675147e-05, "loss": 0.7748, "step": 200960 }, { "epoch": 17.41, "learning_rate": 3.153044408465364e-05, "loss": 0.7594, "step": 200970 }, { "epoch": 17.41, "learning_rate": 3.15282757025558e-05, "loss": 0.8508, "step": 200980 }, { "epoch": 17.41, "learning_rate": 3.152610732045796e-05, "loss": 0.7512, "step": 200990 }, { "epoch": 17.41, "learning_rate": 3.152393893836012e-05, "loss": 0.803, "step": 201000 }, { "epoch": 17.41, "learning_rate": 3.152177055626228e-05, "loss": 0.9411, "step": 201010 }, { "epoch": 17.41, "learning_rate": 3.151960217416445e-05, "loss": 0.9523, "step": 201020 }, { "epoch": 17.41, "learning_rate": 3.151743379206661e-05, "loss": 0.957, "step": 201030 }, { "epoch": 17.41, "learning_rate": 3.1515265409968775e-05, "loss": 0.816, "step": 201040 }, { "epoch": 17.41, "learning_rate": 3.1513097027870935e-05, "loss": 0.7802, "step": 201050 }, { "epoch": 17.41, "learning_rate": 3.15109286457731e-05, "loss": 0.9712, "step": 201060 }, { "epoch": 17.41, "learning_rate": 3.150876026367526e-05, "loss": 0.6276, "step": 201070 }, { "epoch": 17.41, "learning_rate": 3.150659188157742e-05, "loss": 0.7819, "step": 201080 }, { "epoch": 17.42, "learning_rate": 3.150442349947959e-05, "loss": 0.9006, "step": 201090 }, { "epoch": 17.42, "learning_rate": 3.150225511738175e-05, "loss": 0.8297, "step": 201100 }, { "epoch": 17.42, "learning_rate": 3.150008673528391e-05, "loss": 0.8878, "step": 201110 }, { "epoch": 17.42, "learning_rate": 3.149791835318607e-05, "loss": 0.9378, "step": 201120 }, { "epoch": 17.42, "learning_rate": 3.149574997108824e-05, "loss": 0.8773, "step": 201130 }, { "epoch": 17.42, "learning_rate": 3.14935815889904e-05, "loss": 0.8878, "step": 201140 }, { "epoch": 17.42, "learning_rate": 3.149141320689256e-05, "loss": 0.6716, "step": 201150 }, { "epoch": 17.42, "learning_rate": 3.1489244824794724e-05, "loss": 0.9336, "step": 201160 }, { "epoch": 17.42, "learning_rate": 3.148707644269689e-05, "loss": 0.9252, "step": 201170 }, { "epoch": 17.42, "learning_rate": 3.1484908060599044e-05, "loss": 0.7908, "step": 201180 }, { "epoch": 17.42, "learning_rate": 3.148273967850121e-05, "loss": 0.6787, "step": 201190 }, { "epoch": 17.43, "learning_rate": 3.1480571296403376e-05, "loss": 0.7514, "step": 201200 }, { "epoch": 17.43, "learning_rate": 3.1478402914305536e-05, "loss": 0.8191, "step": 201210 }, { "epoch": 17.43, "learning_rate": 3.1476234532207696e-05, "loss": 0.7449, "step": 201220 }, { "epoch": 17.43, "learning_rate": 3.147406615010986e-05, "loss": 0.9451, "step": 201230 }, { "epoch": 17.43, "learning_rate": 3.147189776801202e-05, "loss": 0.7681, "step": 201240 }, { "epoch": 17.43, "learning_rate": 3.146972938591419e-05, "loss": 0.7531, "step": 201250 }, { "epoch": 17.43, "learning_rate": 3.146756100381635e-05, "loss": 0.8719, "step": 201260 }, { "epoch": 17.43, "learning_rate": 3.1465392621718514e-05, "loss": 0.8788, "step": 201270 }, { "epoch": 17.43, "learning_rate": 3.1463224239620674e-05, "loss": 0.6377, "step": 201280 }, { "epoch": 17.43, "learning_rate": 3.146105585752284e-05, "loss": 0.9087, "step": 201290 }, { "epoch": 17.43, "learning_rate": 3.1458887475425e-05, "loss": 0.8843, "step": 201300 }, { "epoch": 17.43, "learning_rate": 3.145671909332716e-05, "loss": 0.8701, "step": 201310 }, { "epoch": 17.44, "learning_rate": 3.1454550711229326e-05, "loss": 0.9189, "step": 201320 }, { "epoch": 17.44, "learning_rate": 3.145238232913149e-05, "loss": 1.0472, "step": 201330 }, { "epoch": 17.44, "learning_rate": 3.145021394703365e-05, "loss": 0.7699, "step": 201340 }, { "epoch": 17.44, "learning_rate": 3.144804556493581e-05, "loss": 0.8266, "step": 201350 }, { "epoch": 17.44, "learning_rate": 3.144587718283798e-05, "loss": 0.6946, "step": 201360 }, { "epoch": 17.44, "learning_rate": 3.144370880074014e-05, "loss": 0.7541, "step": 201370 }, { "epoch": 17.44, "learning_rate": 3.14415404186423e-05, "loss": 0.9294, "step": 201380 }, { "epoch": 17.44, "learning_rate": 3.143937203654446e-05, "loss": 0.9167, "step": 201390 }, { "epoch": 17.44, "learning_rate": 3.143720365444663e-05, "loss": 0.7151, "step": 201400 }, { "epoch": 17.44, "learning_rate": 3.143503527234879e-05, "loss": 0.8941, "step": 201410 }, { "epoch": 17.44, "learning_rate": 3.143286689025095e-05, "loss": 0.7506, "step": 201420 }, { "epoch": 17.45, "learning_rate": 3.1430698508153115e-05, "loss": 0.7629, "step": 201430 }, { "epoch": 17.45, "learning_rate": 3.1428530126055275e-05, "loss": 0.9715, "step": 201440 }, { "epoch": 17.45, "learning_rate": 3.1426361743957435e-05, "loss": 0.6846, "step": 201450 }, { "epoch": 17.45, "learning_rate": 3.14241933618596e-05, "loss": 0.7835, "step": 201460 }, { "epoch": 17.45, "learning_rate": 3.142202497976177e-05, "loss": 0.8155, "step": 201470 }, { "epoch": 17.45, "learning_rate": 3.141985659766393e-05, "loss": 0.8082, "step": 201480 }, { "epoch": 17.45, "learning_rate": 3.1417688215566087e-05, "loss": 0.6794, "step": 201490 }, { "epoch": 17.45, "learning_rate": 3.141551983346825e-05, "loss": 0.8641, "step": 201500 }, { "epoch": 17.45, "learning_rate": 3.141335145137041e-05, "loss": 0.8807, "step": 201510 }, { "epoch": 17.45, "learning_rate": 3.141118306927258e-05, "loss": 0.7687, "step": 201520 }, { "epoch": 17.45, "learning_rate": 3.140901468717474e-05, "loss": 0.8724, "step": 201530 }, { "epoch": 17.45, "learning_rate": 3.1406846305076905e-05, "loss": 0.9722, "step": 201540 }, { "epoch": 17.46, "learning_rate": 3.1404677922979065e-05, "loss": 0.9295, "step": 201550 }, { "epoch": 17.46, "learning_rate": 3.140250954088123e-05, "loss": 0.7669, "step": 201560 }, { "epoch": 17.46, "learning_rate": 3.140034115878339e-05, "loss": 0.8332, "step": 201570 }, { "epoch": 17.46, "learning_rate": 3.139817277668555e-05, "loss": 0.7025, "step": 201580 }, { "epoch": 17.46, "learning_rate": 3.1396004394587717e-05, "loss": 0.9601, "step": 201590 }, { "epoch": 17.46, "learning_rate": 3.139383601248988e-05, "loss": 0.7681, "step": 201600 }, { "epoch": 17.46, "learning_rate": 3.139166763039204e-05, "loss": 0.9192, "step": 201610 }, { "epoch": 17.46, "learning_rate": 3.13894992482942e-05, "loss": 0.9268, "step": 201620 }, { "epoch": 17.46, "learning_rate": 3.138733086619637e-05, "loss": 0.9755, "step": 201630 }, { "epoch": 17.46, "learning_rate": 3.138516248409853e-05, "loss": 0.9894, "step": 201640 }, { "epoch": 17.46, "learning_rate": 3.138299410200069e-05, "loss": 0.7881, "step": 201650 }, { "epoch": 17.47, "learning_rate": 3.1380825719902854e-05, "loss": 0.7606, "step": 201660 }, { "epoch": 17.47, "learning_rate": 3.137865733780502e-05, "loss": 0.8989, "step": 201670 }, { "epoch": 17.47, "learning_rate": 3.137648895570718e-05, "loss": 0.7886, "step": 201680 }, { "epoch": 17.47, "learning_rate": 3.137432057360934e-05, "loss": 0.7984, "step": 201690 }, { "epoch": 17.47, "learning_rate": 3.1372152191511506e-05, "loss": 0.8773, "step": 201700 }, { "epoch": 17.47, "learning_rate": 3.1369983809413666e-05, "loss": 0.8464, "step": 201710 }, { "epoch": 17.47, "learning_rate": 3.1367815427315825e-05, "loss": 0.8709, "step": 201720 }, { "epoch": 17.47, "learning_rate": 3.136564704521799e-05, "loss": 0.7545, "step": 201730 }, { "epoch": 17.47, "learning_rate": 3.136347866312015e-05, "loss": 0.9172, "step": 201740 }, { "epoch": 17.47, "learning_rate": 3.136131028102232e-05, "loss": 0.9699, "step": 201750 }, { "epoch": 17.47, "learning_rate": 3.135914189892448e-05, "loss": 0.9484, "step": 201760 }, { "epoch": 17.47, "learning_rate": 3.1356973516826644e-05, "loss": 0.7709, "step": 201770 }, { "epoch": 17.48, "learning_rate": 3.1354805134728803e-05, "loss": 0.744, "step": 201780 }, { "epoch": 17.48, "learning_rate": 3.135263675263097e-05, "loss": 0.9209, "step": 201790 }, { "epoch": 17.48, "learning_rate": 3.135046837053313e-05, "loss": 0.9761, "step": 201800 }, { "epoch": 17.48, "learning_rate": 3.134829998843529e-05, "loss": 0.6703, "step": 201810 }, { "epoch": 17.48, "learning_rate": 3.1346131606337456e-05, "loss": 0.8886, "step": 201820 }, { "epoch": 17.48, "learning_rate": 3.134396322423962e-05, "loss": 0.919, "step": 201830 }, { "epoch": 17.48, "learning_rate": 3.134179484214178e-05, "loss": 0.9088, "step": 201840 }, { "epoch": 17.48, "learning_rate": 3.133962646004394e-05, "loss": 1.0906, "step": 201850 }, { "epoch": 17.48, "learning_rate": 3.133745807794611e-05, "loss": 0.849, "step": 201860 }, { "epoch": 17.48, "learning_rate": 3.133528969584827e-05, "loss": 0.8442, "step": 201870 }, { "epoch": 17.48, "learning_rate": 3.133312131375043e-05, "loss": 0.9284, "step": 201880 }, { "epoch": 17.48, "learning_rate": 3.133095293165259e-05, "loss": 0.8395, "step": 201890 }, { "epoch": 17.49, "learning_rate": 3.132878454955476e-05, "loss": 0.8556, "step": 201900 }, { "epoch": 17.49, "learning_rate": 3.132661616745692e-05, "loss": 0.8084, "step": 201910 }, { "epoch": 17.49, "learning_rate": 3.132444778535908e-05, "loss": 0.675, "step": 201920 }, { "epoch": 17.49, "learning_rate": 3.1322279403261245e-05, "loss": 0.9963, "step": 201930 }, { "epoch": 17.49, "learning_rate": 3.1320111021163405e-05, "loss": 0.9162, "step": 201940 }, { "epoch": 17.49, "learning_rate": 3.1317942639065564e-05, "loss": 0.8503, "step": 201950 }, { "epoch": 17.49, "learning_rate": 3.131577425696773e-05, "loss": 0.6419, "step": 201960 }, { "epoch": 17.49, "learning_rate": 3.13136058748699e-05, "loss": 0.6983, "step": 201970 }, { "epoch": 17.49, "learning_rate": 3.131143749277206e-05, "loss": 0.8218, "step": 201980 }, { "epoch": 17.49, "learning_rate": 3.1309269110674216e-05, "loss": 0.7353, "step": 201990 }, { "epoch": 17.49, "learning_rate": 3.130710072857638e-05, "loss": 0.8274, "step": 202000 }, { "epoch": 17.5, "learning_rate": 3.130493234647854e-05, "loss": 0.9637, "step": 202010 }, { "epoch": 17.5, "learning_rate": 3.130276396438071e-05, "loss": 0.7621, "step": 202020 }, { "epoch": 17.5, "learning_rate": 3.130059558228287e-05, "loss": 0.6742, "step": 202030 }, { "epoch": 17.5, "learning_rate": 3.1298427200185035e-05, "loss": 0.8756, "step": 202040 }, { "epoch": 17.5, "learning_rate": 3.1296258818087194e-05, "loss": 0.7817, "step": 202050 }, { "epoch": 17.5, "learning_rate": 3.129409043598936e-05, "loss": 0.8395, "step": 202060 }, { "epoch": 17.5, "learning_rate": 3.129192205389152e-05, "loss": 0.8825, "step": 202070 }, { "epoch": 17.5, "learning_rate": 3.128975367179368e-05, "loss": 0.7593, "step": 202080 }, { "epoch": 17.5, "learning_rate": 3.1287585289695846e-05, "loss": 0.8525, "step": 202090 }, { "epoch": 17.5, "learning_rate": 3.128541690759801e-05, "loss": 1.05, "step": 202100 }, { "epoch": 17.5, "learning_rate": 3.128324852550017e-05, "loss": 0.8876, "step": 202110 }, { "epoch": 17.5, "learning_rate": 3.128108014340233e-05, "loss": 0.8569, "step": 202120 }, { "epoch": 17.51, "learning_rate": 3.12789117613045e-05, "loss": 0.9317, "step": 202130 }, { "epoch": 17.51, "learning_rate": 3.127674337920666e-05, "loss": 0.8797, "step": 202140 }, { "epoch": 17.51, "learning_rate": 3.127457499710882e-05, "loss": 0.8386, "step": 202150 }, { "epoch": 17.51, "learning_rate": 3.1272406615010984e-05, "loss": 0.9044, "step": 202160 }, { "epoch": 17.51, "learning_rate": 3.127023823291315e-05, "loss": 0.9386, "step": 202170 }, { "epoch": 17.51, "learning_rate": 3.126806985081531e-05, "loss": 0.7903, "step": 202180 }, { "epoch": 17.51, "learning_rate": 3.126590146871747e-05, "loss": 0.8306, "step": 202190 }, { "epoch": 17.51, "learning_rate": 3.1263733086619636e-05, "loss": 0.8236, "step": 202200 }, { "epoch": 17.51, "learning_rate": 3.1261564704521796e-05, "loss": 0.8955, "step": 202210 }, { "epoch": 17.51, "learning_rate": 3.1259396322423955e-05, "loss": 0.9274, "step": 202220 }, { "epoch": 17.51, "learning_rate": 3.125722794032612e-05, "loss": 0.8194, "step": 202230 }, { "epoch": 17.52, "learning_rate": 3.125505955822829e-05, "loss": 0.7447, "step": 202240 }, { "epoch": 17.52, "learning_rate": 3.125289117613045e-05, "loss": 0.8145, "step": 202250 }, { "epoch": 17.52, "learning_rate": 3.125072279403261e-05, "loss": 0.7676, "step": 202260 }, { "epoch": 17.52, "learning_rate": 3.1248554411934774e-05, "loss": 0.9002, "step": 202270 }, { "epoch": 17.52, "learning_rate": 3.124638602983693e-05, "loss": 0.8499, "step": 202280 }, { "epoch": 17.52, "learning_rate": 3.12442176477391e-05, "loss": 0.8546, "step": 202290 }, { "epoch": 17.52, "learning_rate": 3.124204926564126e-05, "loss": 0.7953, "step": 202300 }, { "epoch": 17.52, "learning_rate": 3.123988088354342e-05, "loss": 0.8071, "step": 202310 }, { "epoch": 17.52, "learning_rate": 3.1237712501445585e-05, "loss": 0.9008, "step": 202320 }, { "epoch": 17.52, "learning_rate": 3.123554411934775e-05, "loss": 0.7542, "step": 202330 }, { "epoch": 17.52, "learning_rate": 3.123337573724991e-05, "loss": 0.9431, "step": 202340 }, { "epoch": 17.52, "learning_rate": 3.123120735515207e-05, "loss": 0.849, "step": 202350 }, { "epoch": 17.53, "learning_rate": 3.122903897305424e-05, "loss": 0.862, "step": 202360 }, { "epoch": 17.53, "learning_rate": 3.12268705909564e-05, "loss": 0.8085, "step": 202370 }, { "epoch": 17.53, "learning_rate": 3.1224702208858557e-05, "loss": 0.8184, "step": 202380 }, { "epoch": 17.53, "learning_rate": 3.122253382676072e-05, "loss": 0.7261, "step": 202390 }, { "epoch": 17.53, "learning_rate": 3.122036544466289e-05, "loss": 0.8034, "step": 202400 }, { "epoch": 17.53, "learning_rate": 3.121819706256505e-05, "loss": 0.9001, "step": 202410 }, { "epoch": 17.53, "learning_rate": 3.121602868046721e-05, "loss": 0.8406, "step": 202420 }, { "epoch": 17.53, "learning_rate": 3.1213860298369375e-05, "loss": 0.8622, "step": 202430 }, { "epoch": 17.53, "learning_rate": 3.1211691916271535e-05, "loss": 0.9721, "step": 202440 }, { "epoch": 17.53, "learning_rate": 3.1209523534173694e-05, "loss": 0.9295, "step": 202450 }, { "epoch": 17.53, "learning_rate": 3.120735515207586e-05, "loss": 0.84, "step": 202460 }, { "epoch": 17.54, "learning_rate": 3.120518676997803e-05, "loss": 0.7775, "step": 202470 }, { "epoch": 17.54, "learning_rate": 3.120301838788019e-05, "loss": 0.7915, "step": 202480 }, { "epoch": 17.54, "learning_rate": 3.1200850005782346e-05, "loss": 0.7843, "step": 202490 }, { "epoch": 17.54, "learning_rate": 3.119868162368451e-05, "loss": 0.8156, "step": 202500 }, { "epoch": 17.54, "learning_rate": 3.119651324158667e-05, "loss": 0.7715, "step": 202510 }, { "epoch": 17.54, "learning_rate": 3.119434485948884e-05, "loss": 0.8485, "step": 202520 }, { "epoch": 17.54, "learning_rate": 3.1192176477391e-05, "loss": 0.8984, "step": 202530 }, { "epoch": 17.54, "learning_rate": 3.1190008095293165e-05, "loss": 0.9278, "step": 202540 }, { "epoch": 17.54, "learning_rate": 3.1187839713195324e-05, "loss": 0.9086, "step": 202550 }, { "epoch": 17.54, "learning_rate": 3.118567133109749e-05, "loss": 0.9198, "step": 202560 }, { "epoch": 17.54, "learning_rate": 3.118350294899965e-05, "loss": 0.9375, "step": 202570 }, { "epoch": 17.54, "learning_rate": 3.118133456690181e-05, "loss": 0.9347, "step": 202580 }, { "epoch": 17.55, "learning_rate": 3.1179166184803976e-05, "loss": 1.0076, "step": 202590 }, { "epoch": 17.55, "learning_rate": 3.117699780270614e-05, "loss": 0.8836, "step": 202600 }, { "epoch": 17.55, "learning_rate": 3.11748294206083e-05, "loss": 0.9307, "step": 202610 }, { "epoch": 17.55, "learning_rate": 3.117266103851046e-05, "loss": 0.7532, "step": 202620 }, { "epoch": 17.55, "learning_rate": 3.117049265641263e-05, "loss": 1.0322, "step": 202630 }, { "epoch": 17.55, "learning_rate": 3.116832427431479e-05, "loss": 0.852, "step": 202640 }, { "epoch": 17.55, "learning_rate": 3.116615589221695e-05, "loss": 0.8887, "step": 202650 }, { "epoch": 17.55, "learning_rate": 3.1163987510119114e-05, "loss": 0.8541, "step": 202660 }, { "epoch": 17.55, "learning_rate": 3.116181912802128e-05, "loss": 0.8351, "step": 202670 }, { "epoch": 17.55, "learning_rate": 3.115965074592344e-05, "loss": 0.9382, "step": 202680 }, { "epoch": 17.55, "learning_rate": 3.11574823638256e-05, "loss": 0.7886, "step": 202690 }, { "epoch": 17.56, "learning_rate": 3.1155313981727766e-05, "loss": 0.8856, "step": 202700 }, { "epoch": 17.56, "learning_rate": 3.1153145599629926e-05, "loss": 0.9055, "step": 202710 }, { "epoch": 17.56, "learning_rate": 3.1150977217532085e-05, "loss": 0.9933, "step": 202720 }, { "epoch": 17.56, "learning_rate": 3.114880883543425e-05, "loss": 1.0243, "step": 202730 }, { "epoch": 17.56, "learning_rate": 3.114664045333642e-05, "loss": 0.7856, "step": 202740 }, { "epoch": 17.56, "learning_rate": 3.114447207123858e-05, "loss": 0.7809, "step": 202750 }, { "epoch": 17.56, "learning_rate": 3.114230368914074e-05, "loss": 0.7731, "step": 202760 }, { "epoch": 17.56, "learning_rate": 3.1140135307042904e-05, "loss": 0.8676, "step": 202770 }, { "epoch": 17.56, "learning_rate": 3.113796692494506e-05, "loss": 0.8434, "step": 202780 }, { "epoch": 17.56, "learning_rate": 3.113579854284723e-05, "loss": 0.9627, "step": 202790 }, { "epoch": 17.56, "learning_rate": 3.113363016074939e-05, "loss": 0.7554, "step": 202800 }, { "epoch": 17.56, "learning_rate": 3.113146177865155e-05, "loss": 0.7641, "step": 202810 }, { "epoch": 17.57, "learning_rate": 3.1129293396553715e-05, "loss": 0.8691, "step": 202820 }, { "epoch": 17.57, "learning_rate": 3.112712501445588e-05, "loss": 0.8442, "step": 202830 }, { "epoch": 17.57, "learning_rate": 3.112495663235804e-05, "loss": 0.8643, "step": 202840 }, { "epoch": 17.57, "learning_rate": 3.11227882502602e-05, "loss": 0.837, "step": 202850 }, { "epoch": 17.57, "learning_rate": 3.112061986816237e-05, "loss": 0.8469, "step": 202860 }, { "epoch": 17.57, "learning_rate": 3.111845148606453e-05, "loss": 0.794, "step": 202870 }, { "epoch": 17.57, "learning_rate": 3.1116283103966686e-05, "loss": 0.9318, "step": 202880 }, { "epoch": 17.57, "learning_rate": 3.111411472186885e-05, "loss": 0.7907, "step": 202890 }, { "epoch": 17.57, "learning_rate": 3.111194633977102e-05, "loss": 0.8345, "step": 202900 }, { "epoch": 17.57, "learning_rate": 3.110977795767318e-05, "loss": 0.7413, "step": 202910 }, { "epoch": 17.57, "learning_rate": 3.110760957557534e-05, "loss": 0.8169, "step": 202920 }, { "epoch": 17.58, "learning_rate": 3.1105441193477505e-05, "loss": 0.7942, "step": 202930 }, { "epoch": 17.58, "learning_rate": 3.1103272811379664e-05, "loss": 0.76, "step": 202940 }, { "epoch": 17.58, "learning_rate": 3.1101104429281824e-05, "loss": 0.7799, "step": 202950 }, { "epoch": 17.58, "learning_rate": 3.109893604718399e-05, "loss": 0.7954, "step": 202960 }, { "epoch": 17.58, "learning_rate": 3.109676766508616e-05, "loss": 0.8578, "step": 202970 }, { "epoch": 17.58, "learning_rate": 3.1094599282988316e-05, "loss": 0.8326, "step": 202980 }, { "epoch": 17.58, "learning_rate": 3.1092430900890476e-05, "loss": 0.705, "step": 202990 }, { "epoch": 17.58, "learning_rate": 3.109026251879264e-05, "loss": 0.94, "step": 203000 }, { "epoch": 17.58, "learning_rate": 3.10880941366948e-05, "loss": 0.7387, "step": 203010 }, { "epoch": 17.58, "learning_rate": 3.108592575459697e-05, "loss": 0.8387, "step": 203020 }, { "epoch": 17.58, "learning_rate": 3.108375737249913e-05, "loss": 0.6964, "step": 203030 }, { "epoch": 17.58, "learning_rate": 3.1081588990401295e-05, "loss": 0.7714, "step": 203040 }, { "epoch": 17.59, "learning_rate": 3.1079420608303454e-05, "loss": 0.8066, "step": 203050 }, { "epoch": 17.59, "learning_rate": 3.107725222620562e-05, "loss": 0.8907, "step": 203060 }, { "epoch": 17.59, "learning_rate": 3.107508384410778e-05, "loss": 0.8715, "step": 203070 }, { "epoch": 17.59, "learning_rate": 3.107291546200994e-05, "loss": 0.7511, "step": 203080 }, { "epoch": 17.59, "learning_rate": 3.1070747079912106e-05, "loss": 0.8825, "step": 203090 }, { "epoch": 17.59, "learning_rate": 3.106857869781427e-05, "loss": 0.7777, "step": 203100 }, { "epoch": 17.59, "learning_rate": 3.106641031571643e-05, "loss": 0.8927, "step": 203110 }, { "epoch": 17.59, "learning_rate": 3.106424193361859e-05, "loss": 0.6648, "step": 203120 }, { "epoch": 17.59, "learning_rate": 3.106207355152076e-05, "loss": 0.8396, "step": 203130 }, { "epoch": 17.59, "learning_rate": 3.105990516942292e-05, "loss": 0.8038, "step": 203140 }, { "epoch": 17.59, "learning_rate": 3.105773678732508e-05, "loss": 0.8117, "step": 203150 }, { "epoch": 17.59, "learning_rate": 3.1055568405227244e-05, "loss": 0.6993, "step": 203160 }, { "epoch": 17.6, "learning_rate": 3.105340002312941e-05, "loss": 0.8793, "step": 203170 }, { "epoch": 17.6, "learning_rate": 3.105123164103157e-05, "loss": 0.7538, "step": 203180 }, { "epoch": 17.6, "learning_rate": 3.104906325893373e-05, "loss": 0.8327, "step": 203190 }, { "epoch": 17.6, "learning_rate": 3.1046894876835896e-05, "loss": 0.8414, "step": 203200 }, { "epoch": 17.6, "learning_rate": 3.1044726494738055e-05, "loss": 0.8887, "step": 203210 }, { "epoch": 17.6, "learning_rate": 3.1042558112640215e-05, "loss": 0.817, "step": 203220 }, { "epoch": 17.6, "learning_rate": 3.104038973054238e-05, "loss": 0.7759, "step": 203230 }, { "epoch": 17.6, "learning_rate": 3.103822134844455e-05, "loss": 0.8826, "step": 203240 }, { "epoch": 17.6, "learning_rate": 3.103605296634671e-05, "loss": 1.0565, "step": 203250 }, { "epoch": 17.6, "learning_rate": 3.103388458424887e-05, "loss": 0.8036, "step": 203260 }, { "epoch": 17.6, "learning_rate": 3.1031716202151033e-05, "loss": 0.8977, "step": 203270 }, { "epoch": 17.61, "learning_rate": 3.102954782005319e-05, "loss": 0.9195, "step": 203280 }, { "epoch": 17.61, "learning_rate": 3.102737943795536e-05, "loss": 0.8241, "step": 203290 }, { "epoch": 17.61, "learning_rate": 3.102521105585752e-05, "loss": 0.9141, "step": 203300 }, { "epoch": 17.61, "learning_rate": 3.1023042673759685e-05, "loss": 0.7523, "step": 203310 }, { "epoch": 17.61, "learning_rate": 3.1020874291661845e-05, "loss": 0.9777, "step": 203320 }, { "epoch": 17.61, "learning_rate": 3.101870590956401e-05, "loss": 0.9148, "step": 203330 }, { "epoch": 17.61, "learning_rate": 3.101653752746617e-05, "loss": 0.8444, "step": 203340 }, { "epoch": 17.61, "learning_rate": 3.101436914536833e-05, "loss": 1.0086, "step": 203350 }, { "epoch": 17.61, "learning_rate": 3.10122007632705e-05, "loss": 0.8662, "step": 203360 }, { "epoch": 17.61, "learning_rate": 3.101003238117266e-05, "loss": 0.7576, "step": 203370 }, { "epoch": 17.61, "learning_rate": 3.1007863999074816e-05, "loss": 0.8009, "step": 203380 }, { "epoch": 17.61, "learning_rate": 3.100569561697698e-05, "loss": 0.7451, "step": 203390 }, { "epoch": 17.62, "learning_rate": 3.100352723487915e-05, "loss": 0.781, "step": 203400 }, { "epoch": 17.62, "learning_rate": 3.100135885278131e-05, "loss": 0.9771, "step": 203410 }, { "epoch": 17.62, "learning_rate": 3.099919047068347e-05, "loss": 0.9052, "step": 203420 }, { "epoch": 17.62, "learning_rate": 3.0997022088585635e-05, "loss": 0.8639, "step": 203430 }, { "epoch": 17.62, "learning_rate": 3.0994853706487794e-05, "loss": 1.0274, "step": 203440 }, { "epoch": 17.62, "learning_rate": 3.099268532438996e-05, "loss": 0.8808, "step": 203450 }, { "epoch": 17.62, "learning_rate": 3.099051694229212e-05, "loss": 0.8781, "step": 203460 }, { "epoch": 17.62, "learning_rate": 3.098834856019429e-05, "loss": 0.8052, "step": 203470 }, { "epoch": 17.62, "learning_rate": 3.0986180178096446e-05, "loss": 0.8083, "step": 203480 }, { "epoch": 17.62, "learning_rate": 3.0984011795998606e-05, "loss": 0.8607, "step": 203490 }, { "epoch": 17.62, "learning_rate": 3.098184341390077e-05, "loss": 0.8866, "step": 203500 }, { "epoch": 17.63, "learning_rate": 3.097967503180293e-05, "loss": 0.6883, "step": 203510 }, { "epoch": 17.63, "learning_rate": 3.09775066497051e-05, "loss": 0.6694, "step": 203520 }, { "epoch": 17.63, "learning_rate": 3.097533826760726e-05, "loss": 0.7204, "step": 203530 }, { "epoch": 17.63, "learning_rate": 3.0973169885509424e-05, "loss": 0.8353, "step": 203540 }, { "epoch": 17.63, "learning_rate": 3.0971001503411584e-05, "loss": 1.0332, "step": 203550 }, { "epoch": 17.63, "learning_rate": 3.096883312131375e-05, "loss": 0.8276, "step": 203560 }, { "epoch": 17.63, "learning_rate": 3.096666473921591e-05, "loss": 0.7592, "step": 203570 }, { "epoch": 17.63, "learning_rate": 3.096449635711807e-05, "loss": 0.9651, "step": 203580 }, { "epoch": 17.63, "learning_rate": 3.0962327975020236e-05, "loss": 0.8628, "step": 203590 }, { "epoch": 17.63, "learning_rate": 3.09601595929224e-05, "loss": 0.8466, "step": 203600 }, { "epoch": 17.63, "learning_rate": 3.095799121082456e-05, "loss": 1.1212, "step": 203610 }, { "epoch": 17.63, "learning_rate": 3.095582282872672e-05, "loss": 0.9575, "step": 203620 }, { "epoch": 17.64, "learning_rate": 3.095365444662889e-05, "loss": 0.7337, "step": 203630 }, { "epoch": 17.64, "learning_rate": 3.095148606453105e-05, "loss": 0.7483, "step": 203640 }, { "epoch": 17.64, "learning_rate": 3.094931768243321e-05, "loss": 0.7346, "step": 203650 }, { "epoch": 17.64, "learning_rate": 3.0947149300335374e-05, "loss": 0.7716, "step": 203660 }, { "epoch": 17.64, "learning_rate": 3.094498091823754e-05, "loss": 0.7103, "step": 203670 }, { "epoch": 17.64, "learning_rate": 3.09428125361397e-05, "loss": 0.8257, "step": 203680 }, { "epoch": 17.64, "learning_rate": 3.094064415404186e-05, "loss": 0.8223, "step": 203690 }, { "epoch": 17.64, "learning_rate": 3.0938475771944026e-05, "loss": 0.7616, "step": 203700 }, { "epoch": 17.64, "learning_rate": 3.0936307389846185e-05, "loss": 0.9367, "step": 203710 }, { "epoch": 17.64, "learning_rate": 3.0934139007748345e-05, "loss": 0.7283, "step": 203720 }, { "epoch": 17.64, "learning_rate": 3.093197062565051e-05, "loss": 0.6786, "step": 203730 }, { "epoch": 17.65, "learning_rate": 3.092980224355268e-05, "loss": 0.8529, "step": 203740 }, { "epoch": 17.65, "learning_rate": 3.092763386145484e-05, "loss": 0.8697, "step": 203750 }, { "epoch": 17.65, "learning_rate": 3.0925465479357e-05, "loss": 0.7459, "step": 203760 }, { "epoch": 17.65, "learning_rate": 3.092329709725916e-05, "loss": 0.8785, "step": 203770 }, { "epoch": 17.65, "learning_rate": 3.092112871516132e-05, "loss": 0.8468, "step": 203780 }, { "epoch": 17.65, "learning_rate": 3.091896033306349e-05, "loss": 0.7976, "step": 203790 }, { "epoch": 17.65, "learning_rate": 3.091679195096565e-05, "loss": 0.6786, "step": 203800 }, { "epoch": 17.65, "learning_rate": 3.0914623568867815e-05, "loss": 0.9047, "step": 203810 }, { "epoch": 17.65, "learning_rate": 3.0912455186769975e-05, "loss": 0.8833, "step": 203820 }, { "epoch": 17.65, "learning_rate": 3.091028680467214e-05, "loss": 0.7992, "step": 203830 }, { "epoch": 17.65, "learning_rate": 3.09081184225743e-05, "loss": 0.7997, "step": 203840 }, { "epoch": 17.65, "learning_rate": 3.090595004047646e-05, "loss": 0.8048, "step": 203850 }, { "epoch": 17.66, "learning_rate": 3.090378165837863e-05, "loss": 0.8902, "step": 203860 }, { "epoch": 17.66, "learning_rate": 3.090161327628079e-05, "loss": 0.9245, "step": 203870 }, { "epoch": 17.66, "learning_rate": 3.0899444894182946e-05, "loss": 0.926, "step": 203880 }, { "epoch": 17.66, "learning_rate": 3.089727651208511e-05, "loss": 0.6503, "step": 203890 }, { "epoch": 17.66, "learning_rate": 3.089510812998728e-05, "loss": 0.6839, "step": 203900 }, { "epoch": 17.66, "learning_rate": 3.089293974788944e-05, "loss": 0.8718, "step": 203910 }, { "epoch": 17.66, "learning_rate": 3.08907713657916e-05, "loss": 0.7164, "step": 203920 }, { "epoch": 17.66, "learning_rate": 3.0888602983693765e-05, "loss": 0.8385, "step": 203930 }, { "epoch": 17.66, "learning_rate": 3.0886434601595924e-05, "loss": 0.772, "step": 203940 }, { "epoch": 17.66, "learning_rate": 3.088426621949809e-05, "loss": 0.9571, "step": 203950 }, { "epoch": 17.66, "learning_rate": 3.088209783740025e-05, "loss": 0.8506, "step": 203960 }, { "epoch": 17.67, "learning_rate": 3.0879929455302417e-05, "loss": 0.9092, "step": 203970 }, { "epoch": 17.67, "learning_rate": 3.0877761073204576e-05, "loss": 0.726, "step": 203980 }, { "epoch": 17.67, "learning_rate": 3.0875592691106736e-05, "loss": 0.9299, "step": 203990 }, { "epoch": 17.67, "learning_rate": 3.08734243090089e-05, "loss": 0.7802, "step": 204000 }, { "epoch": 17.67, "learning_rate": 3.087125592691106e-05, "loss": 0.9342, "step": 204010 }, { "epoch": 17.67, "learning_rate": 3.086908754481323e-05, "loss": 0.8034, "step": 204020 }, { "epoch": 17.67, "learning_rate": 3.086691916271539e-05, "loss": 0.7112, "step": 204030 }, { "epoch": 17.67, "learning_rate": 3.0864750780617554e-05, "loss": 0.9301, "step": 204040 }, { "epoch": 17.67, "learning_rate": 3.0862582398519714e-05, "loss": 1.037, "step": 204050 }, { "epoch": 17.67, "learning_rate": 3.086041401642188e-05, "loss": 0.7902, "step": 204060 }, { "epoch": 17.67, "learning_rate": 3.085824563432404e-05, "loss": 0.8876, "step": 204070 }, { "epoch": 17.67, "learning_rate": 3.08560772522262e-05, "loss": 1.0223, "step": 204080 }, { "epoch": 17.68, "learning_rate": 3.0853908870128366e-05, "loss": 0.9066, "step": 204090 }, { "epoch": 17.68, "learning_rate": 3.085174048803053e-05, "loss": 0.8117, "step": 204100 }, { "epoch": 17.68, "learning_rate": 3.084957210593269e-05, "loss": 0.862, "step": 204110 }, { "epoch": 17.68, "learning_rate": 3.084740372383485e-05, "loss": 0.6583, "step": 204120 }, { "epoch": 17.68, "learning_rate": 3.084523534173702e-05, "loss": 0.6127, "step": 204130 }, { "epoch": 17.68, "learning_rate": 3.084306695963918e-05, "loss": 0.7832, "step": 204140 }, { "epoch": 17.68, "learning_rate": 3.084089857754134e-05, "loss": 0.8629, "step": 204150 }, { "epoch": 17.68, "learning_rate": 3.0838730195443503e-05, "loss": 1.0334, "step": 204160 }, { "epoch": 17.68, "learning_rate": 3.083656181334567e-05, "loss": 0.7777, "step": 204170 }, { "epoch": 17.68, "learning_rate": 3.083439343124783e-05, "loss": 0.7496, "step": 204180 }, { "epoch": 17.68, "learning_rate": 3.083222504914999e-05, "loss": 0.7337, "step": 204190 }, { "epoch": 17.69, "learning_rate": 3.0830056667052155e-05, "loss": 0.7936, "step": 204200 }, { "epoch": 17.69, "learning_rate": 3.0827888284954315e-05, "loss": 1.0295, "step": 204210 }, { "epoch": 17.69, "learning_rate": 3.082571990285648e-05, "loss": 0.9284, "step": 204220 }, { "epoch": 17.69, "learning_rate": 3.082355152075864e-05, "loss": 0.8939, "step": 204230 }, { "epoch": 17.69, "learning_rate": 3.082138313866081e-05, "loss": 0.9196, "step": 204240 }, { "epoch": 17.69, "learning_rate": 3.081921475656297e-05, "loss": 0.8905, "step": 204250 }, { "epoch": 17.69, "learning_rate": 3.081704637446513e-05, "loss": 0.8416, "step": 204260 }, { "epoch": 17.69, "learning_rate": 3.081487799236729e-05, "loss": 0.8535, "step": 204270 }, { "epoch": 17.69, "learning_rate": 3.081270961026945e-05, "loss": 0.7839, "step": 204280 }, { "epoch": 17.69, "learning_rate": 3.081054122817162e-05, "loss": 0.8785, "step": 204290 }, { "epoch": 17.69, "learning_rate": 3.080837284607378e-05, "loss": 0.9194, "step": 204300 }, { "epoch": 17.69, "learning_rate": 3.0806204463975945e-05, "loss": 0.8611, "step": 204310 }, { "epoch": 17.7, "learning_rate": 3.0804036081878105e-05, "loss": 0.923, "step": 204320 }, { "epoch": 17.7, "learning_rate": 3.080186769978027e-05, "loss": 0.7002, "step": 204330 }, { "epoch": 17.7, "learning_rate": 3.079969931768243e-05, "loss": 0.7466, "step": 204340 }, { "epoch": 17.7, "learning_rate": 3.079753093558459e-05, "loss": 0.9505, "step": 204350 }, { "epoch": 17.7, "learning_rate": 3.079536255348676e-05, "loss": 0.8387, "step": 204360 }, { "epoch": 17.7, "learning_rate": 3.079319417138892e-05, "loss": 0.9131, "step": 204370 }, { "epoch": 17.7, "learning_rate": 3.079102578929108e-05, "loss": 0.9112, "step": 204380 }, { "epoch": 17.7, "learning_rate": 3.078885740719324e-05, "loss": 0.8022, "step": 204390 }, { "epoch": 17.7, "learning_rate": 3.078668902509541e-05, "loss": 0.8131, "step": 204400 }, { "epoch": 17.7, "learning_rate": 3.078452064299757e-05, "loss": 0.6442, "step": 204410 }, { "epoch": 17.7, "learning_rate": 3.078235226089973e-05, "loss": 0.9998, "step": 204420 }, { "epoch": 17.7, "learning_rate": 3.0780183878801894e-05, "loss": 0.8847, "step": 204430 }, { "epoch": 17.71, "learning_rate": 3.0778015496704054e-05, "loss": 0.7018, "step": 204440 }, { "epoch": 17.71, "learning_rate": 3.077584711460622e-05, "loss": 0.8707, "step": 204450 }, { "epoch": 17.71, "learning_rate": 3.077367873250838e-05, "loss": 0.9655, "step": 204460 }, { "epoch": 17.71, "learning_rate": 3.0771510350410546e-05, "loss": 0.8196, "step": 204470 }, { "epoch": 17.71, "learning_rate": 3.0769341968312706e-05, "loss": 0.7983, "step": 204480 }, { "epoch": 17.71, "learning_rate": 3.0767173586214866e-05, "loss": 0.7781, "step": 204490 }, { "epoch": 17.71, "learning_rate": 3.076500520411703e-05, "loss": 0.8953, "step": 204500 }, { "epoch": 17.71, "learning_rate": 3.076283682201919e-05, "loss": 0.8449, "step": 204510 }, { "epoch": 17.71, "learning_rate": 3.076066843992136e-05, "loss": 0.7949, "step": 204520 }, { "epoch": 17.71, "learning_rate": 3.075850005782352e-05, "loss": 0.7793, "step": 204530 }, { "epoch": 17.71, "learning_rate": 3.0756331675725684e-05, "loss": 0.8625, "step": 204540 }, { "epoch": 17.72, "learning_rate": 3.0754163293627844e-05, "loss": 0.7694, "step": 204550 }, { "epoch": 17.72, "learning_rate": 3.075199491153001e-05, "loss": 0.9188, "step": 204560 }, { "epoch": 17.72, "learning_rate": 3.074982652943217e-05, "loss": 0.8165, "step": 204570 }, { "epoch": 17.72, "learning_rate": 3.074765814733433e-05, "loss": 0.7738, "step": 204580 }, { "epoch": 17.72, "learning_rate": 3.0745489765236496e-05, "loss": 0.7821, "step": 204590 }, { "epoch": 17.72, "learning_rate": 3.074332138313866e-05, "loss": 0.852, "step": 204600 }, { "epoch": 17.72, "learning_rate": 3.074115300104082e-05, "loss": 0.9029, "step": 204610 }, { "epoch": 17.72, "learning_rate": 3.073898461894298e-05, "loss": 0.9979, "step": 204620 }, { "epoch": 17.72, "learning_rate": 3.073681623684515e-05, "loss": 0.7717, "step": 204630 }, { "epoch": 17.72, "learning_rate": 3.073464785474731e-05, "loss": 0.9254, "step": 204640 }, { "epoch": 17.72, "learning_rate": 3.073247947264947e-05, "loss": 0.6751, "step": 204650 }, { "epoch": 17.72, "learning_rate": 3.073031109055163e-05, "loss": 0.9333, "step": 204660 }, { "epoch": 17.73, "learning_rate": 3.07281427084538e-05, "loss": 0.7719, "step": 204670 }, { "epoch": 17.73, "learning_rate": 3.072597432635596e-05, "loss": 0.7576, "step": 204680 }, { "epoch": 17.73, "learning_rate": 3.072380594425812e-05, "loss": 0.8096, "step": 204690 }, { "epoch": 17.73, "learning_rate": 3.0721637562160285e-05, "loss": 0.7807, "step": 204700 }, { "epoch": 17.73, "learning_rate": 3.0719469180062445e-05, "loss": 0.8619, "step": 204710 }, { "epoch": 17.73, "learning_rate": 3.071730079796461e-05, "loss": 0.9297, "step": 204720 }, { "epoch": 17.73, "learning_rate": 3.071513241586677e-05, "loss": 0.8843, "step": 204730 }, { "epoch": 17.73, "learning_rate": 3.071296403376894e-05, "loss": 0.7791, "step": 204740 }, { "epoch": 17.73, "learning_rate": 3.07107956516711e-05, "loss": 0.9096, "step": 204750 }, { "epoch": 17.73, "learning_rate": 3.0708627269573257e-05, "loss": 0.8005, "step": 204760 }, { "epoch": 17.73, "learning_rate": 3.070645888747542e-05, "loss": 0.6994, "step": 204770 }, { "epoch": 17.74, "learning_rate": 3.070429050537758e-05, "loss": 0.7248, "step": 204780 }, { "epoch": 17.74, "learning_rate": 3.070212212327975e-05, "loss": 0.9063, "step": 204790 }, { "epoch": 17.74, "learning_rate": 3.069995374118191e-05, "loss": 0.8861, "step": 204800 }, { "epoch": 17.74, "learning_rate": 3.0697785359084075e-05, "loss": 0.8242, "step": 204810 }, { "epoch": 17.74, "learning_rate": 3.0695616976986235e-05, "loss": 0.7739, "step": 204820 }, { "epoch": 17.74, "learning_rate": 3.06934485948884e-05, "loss": 0.8595, "step": 204830 }, { "epoch": 17.74, "learning_rate": 3.069128021279056e-05, "loss": 0.694, "step": 204840 }, { "epoch": 17.74, "learning_rate": 3.068911183069272e-05, "loss": 0.8933, "step": 204850 }, { "epoch": 17.74, "learning_rate": 3.0686943448594887e-05, "loss": 0.8312, "step": 204860 }, { "epoch": 17.74, "learning_rate": 3.068477506649705e-05, "loss": 0.7764, "step": 204870 }, { "epoch": 17.74, "learning_rate": 3.068260668439921e-05, "loss": 0.6735, "step": 204880 }, { "epoch": 17.74, "learning_rate": 3.068043830230137e-05, "loss": 0.8234, "step": 204890 }, { "epoch": 17.75, "learning_rate": 3.067826992020354e-05, "loss": 0.7507, "step": 204900 }, { "epoch": 17.75, "learning_rate": 3.06761015381057e-05, "loss": 0.8468, "step": 204910 }, { "epoch": 17.75, "learning_rate": 3.067393315600786e-05, "loss": 0.7612, "step": 204920 }, { "epoch": 17.75, "learning_rate": 3.0671764773910024e-05, "loss": 0.7514, "step": 204930 }, { "epoch": 17.75, "learning_rate": 3.066959639181219e-05, "loss": 0.7819, "step": 204940 }, { "epoch": 17.75, "learning_rate": 3.066742800971435e-05, "loss": 0.8876, "step": 204950 }, { "epoch": 17.75, "learning_rate": 3.066525962761651e-05, "loss": 1.0055, "step": 204960 }, { "epoch": 17.75, "learning_rate": 3.0663091245518676e-05, "loss": 0.846, "step": 204970 }, { "epoch": 17.75, "learning_rate": 3.0660922863420836e-05, "loss": 0.8661, "step": 204980 }, { "epoch": 17.75, "learning_rate": 3.0658754481323e-05, "loss": 0.8632, "step": 204990 }, { "epoch": 17.75, "learning_rate": 3.065658609922516e-05, "loss": 0.8418, "step": 205000 }, { "epoch": 17.76, "learning_rate": 3.065441771712732e-05, "loss": 0.8836, "step": 205010 }, { "epoch": 17.76, "learning_rate": 3.065224933502949e-05, "loss": 0.8034, "step": 205020 }, { "epoch": 17.76, "learning_rate": 3.065008095293165e-05, "loss": 0.7914, "step": 205030 }, { "epoch": 17.76, "learning_rate": 3.0647912570833814e-05, "loss": 1.0181, "step": 205040 }, { "epoch": 17.76, "learning_rate": 3.0645744188735974e-05, "loss": 0.8864, "step": 205050 }, { "epoch": 17.76, "learning_rate": 3.064357580663814e-05, "loss": 0.9878, "step": 205060 }, { "epoch": 17.76, "learning_rate": 3.06414074245403e-05, "loss": 0.8111, "step": 205070 }, { "epoch": 17.76, "learning_rate": 3.063923904244246e-05, "loss": 0.9834, "step": 205080 }, { "epoch": 17.76, "learning_rate": 3.0637070660344626e-05, "loss": 0.6946, "step": 205090 }, { "epoch": 17.76, "learning_rate": 3.063490227824679e-05, "loss": 0.8583, "step": 205100 }, { "epoch": 17.76, "learning_rate": 3.063273389614895e-05, "loss": 0.9198, "step": 205110 }, { "epoch": 17.76, "learning_rate": 3.063056551405111e-05, "loss": 0.8338, "step": 205120 }, { "epoch": 17.77, "learning_rate": 3.062839713195328e-05, "loss": 0.8564, "step": 205130 }, { "epoch": 17.77, "learning_rate": 3.062622874985544e-05, "loss": 0.7739, "step": 205140 }, { "epoch": 17.77, "learning_rate": 3.06240603677576e-05, "loss": 0.7684, "step": 205150 }, { "epoch": 17.77, "learning_rate": 3.062189198565976e-05, "loss": 0.8624, "step": 205160 }, { "epoch": 17.77, "learning_rate": 3.061972360356193e-05, "loss": 0.7073, "step": 205170 }, { "epoch": 17.77, "learning_rate": 3.061755522146409e-05, "loss": 0.8156, "step": 205180 }, { "epoch": 17.77, "learning_rate": 3.061538683936625e-05, "loss": 0.9241, "step": 205190 }, { "epoch": 17.77, "learning_rate": 3.0613218457268415e-05, "loss": 0.8952, "step": 205200 }, { "epoch": 17.77, "learning_rate": 3.0611050075170575e-05, "loss": 0.8752, "step": 205210 }, { "epoch": 17.77, "learning_rate": 3.060888169307274e-05, "loss": 0.6972, "step": 205220 }, { "epoch": 17.77, "learning_rate": 3.06067133109749e-05, "loss": 0.9084, "step": 205230 }, { "epoch": 17.78, "learning_rate": 3.060454492887707e-05, "loss": 0.7315, "step": 205240 }, { "epoch": 17.78, "learning_rate": 3.060237654677923e-05, "loss": 1.0946, "step": 205250 }, { "epoch": 17.78, "learning_rate": 3.0600208164681386e-05, "loss": 0.8739, "step": 205260 }, { "epoch": 17.78, "learning_rate": 3.059803978258355e-05, "loss": 0.9205, "step": 205270 }, { "epoch": 17.78, "learning_rate": 3.059587140048571e-05, "loss": 0.6761, "step": 205280 }, { "epoch": 17.78, "learning_rate": 3.059370301838788e-05, "loss": 0.753, "step": 205290 }, { "epoch": 17.78, "learning_rate": 3.059153463629004e-05, "loss": 0.8088, "step": 205300 }, { "epoch": 17.78, "learning_rate": 3.0589366254192205e-05, "loss": 0.892, "step": 205310 }, { "epoch": 17.78, "learning_rate": 3.0587197872094364e-05, "loss": 0.6712, "step": 205320 }, { "epoch": 17.78, "learning_rate": 3.058502948999653e-05, "loss": 0.956, "step": 205330 }, { "epoch": 17.78, "learning_rate": 3.058286110789869e-05, "loss": 0.9471, "step": 205340 }, { "epoch": 17.78, "learning_rate": 3.058069272580085e-05, "loss": 0.7511, "step": 205350 }, { "epoch": 17.79, "learning_rate": 3.0578524343703016e-05, "loss": 0.7797, "step": 205360 }, { "epoch": 17.79, "learning_rate": 3.057635596160518e-05, "loss": 0.8701, "step": 205370 }, { "epoch": 17.79, "learning_rate": 3.057418757950734e-05, "loss": 0.9546, "step": 205380 }, { "epoch": 17.79, "learning_rate": 3.05720191974095e-05, "loss": 0.9365, "step": 205390 }, { "epoch": 17.79, "learning_rate": 3.056985081531167e-05, "loss": 0.7303, "step": 205400 }, { "epoch": 17.79, "learning_rate": 3.056768243321383e-05, "loss": 0.8303, "step": 205410 }, { "epoch": 17.79, "learning_rate": 3.056551405111599e-05, "loss": 0.8324, "step": 205420 }, { "epoch": 17.79, "learning_rate": 3.0563345669018154e-05, "loss": 0.7562, "step": 205430 }, { "epoch": 17.79, "learning_rate": 3.056117728692032e-05, "loss": 0.9307, "step": 205440 }, { "epoch": 17.79, "learning_rate": 3.055900890482248e-05, "loss": 0.8764, "step": 205450 }, { "epoch": 17.79, "learning_rate": 3.055684052272464e-05, "loss": 0.9001, "step": 205460 }, { "epoch": 17.8, "learning_rate": 3.0554672140626806e-05, "loss": 0.7965, "step": 205470 }, { "epoch": 17.8, "learning_rate": 3.0552503758528966e-05, "loss": 0.815, "step": 205480 }, { "epoch": 17.8, "learning_rate": 3.055033537643113e-05, "loss": 0.7973, "step": 205490 }, { "epoch": 17.8, "learning_rate": 3.054816699433329e-05, "loss": 0.7282, "step": 205500 }, { "epoch": 17.8, "learning_rate": 3.054599861223545e-05, "loss": 0.8863, "step": 205510 }, { "epoch": 17.8, "learning_rate": 3.054383023013762e-05, "loss": 0.7953, "step": 205520 }, { "epoch": 17.8, "learning_rate": 3.054166184803978e-05, "loss": 0.7282, "step": 205530 }, { "epoch": 17.8, "learning_rate": 3.0539493465941944e-05, "loss": 0.7925, "step": 205540 }, { "epoch": 17.8, "learning_rate": 3.05373250838441e-05, "loss": 0.7859, "step": 205550 }, { "epoch": 17.8, "learning_rate": 3.053515670174627e-05, "loss": 0.8035, "step": 205560 }, { "epoch": 17.8, "learning_rate": 3.053298831964843e-05, "loss": 0.8664, "step": 205570 }, { "epoch": 17.8, "learning_rate": 3.053081993755059e-05, "loss": 0.7855, "step": 205580 }, { "epoch": 17.81, "learning_rate": 3.0528651555452755e-05, "loss": 0.96, "step": 205590 }, { "epoch": 17.81, "learning_rate": 3.052648317335492e-05, "loss": 0.9162, "step": 205600 }, { "epoch": 17.81, "learning_rate": 3.052431479125708e-05, "loss": 0.7874, "step": 205610 }, { "epoch": 17.81, "learning_rate": 3.052214640915924e-05, "loss": 0.8416, "step": 205620 }, { "epoch": 17.81, "learning_rate": 3.051997802706141e-05, "loss": 0.8386, "step": 205630 }, { "epoch": 17.81, "learning_rate": 3.051780964496357e-05, "loss": 0.8821, "step": 205640 }, { "epoch": 17.81, "learning_rate": 3.051564126286573e-05, "loss": 0.7896, "step": 205650 }, { "epoch": 17.81, "learning_rate": 3.0513472880767893e-05, "loss": 0.7342, "step": 205660 }, { "epoch": 17.81, "learning_rate": 3.0511304498670056e-05, "loss": 0.8663, "step": 205670 }, { "epoch": 17.81, "learning_rate": 3.050913611657222e-05, "loss": 0.7751, "step": 205680 }, { "epoch": 17.81, "learning_rate": 3.050696773447438e-05, "loss": 0.7413, "step": 205690 }, { "epoch": 17.81, "learning_rate": 3.0504799352376545e-05, "loss": 0.9851, "step": 205700 }, { "epoch": 17.82, "learning_rate": 3.0502630970278708e-05, "loss": 0.8631, "step": 205710 }, { "epoch": 17.82, "learning_rate": 3.050046258818087e-05, "loss": 0.9572, "step": 205720 }, { "epoch": 17.82, "learning_rate": 3.049829420608303e-05, "loss": 0.7403, "step": 205730 }, { "epoch": 17.82, "learning_rate": 3.0496125823985194e-05, "loss": 0.8754, "step": 205740 }, { "epoch": 17.82, "learning_rate": 3.0493957441887357e-05, "loss": 0.8938, "step": 205750 }, { "epoch": 17.82, "learning_rate": 3.0491789059789523e-05, "loss": 0.8273, "step": 205760 }, { "epoch": 17.82, "learning_rate": 3.0489620677691683e-05, "loss": 0.8528, "step": 205770 }, { "epoch": 17.82, "learning_rate": 3.0487452295593846e-05, "loss": 0.5896, "step": 205780 }, { "epoch": 17.82, "learning_rate": 3.048528391349601e-05, "loss": 0.784, "step": 205790 }, { "epoch": 17.82, "learning_rate": 3.0483115531398168e-05, "loss": 0.8066, "step": 205800 }, { "epoch": 17.82, "learning_rate": 3.048094714930033e-05, "loss": 0.9636, "step": 205810 }, { "epoch": 17.83, "learning_rate": 3.0478778767202494e-05, "loss": 0.7774, "step": 205820 }, { "epoch": 17.83, "learning_rate": 3.047661038510466e-05, "loss": 0.7746, "step": 205830 }, { "epoch": 17.83, "learning_rate": 3.047444200300682e-05, "loss": 0.7148, "step": 205840 }, { "epoch": 17.83, "learning_rate": 3.0472273620908983e-05, "loss": 0.91, "step": 205850 }, { "epoch": 17.83, "learning_rate": 3.0470105238811146e-05, "loss": 0.8277, "step": 205860 }, { "epoch": 17.83, "learning_rate": 3.046793685671331e-05, "loss": 0.7361, "step": 205870 }, { "epoch": 17.83, "learning_rate": 3.046576847461547e-05, "loss": 0.8104, "step": 205880 }, { "epoch": 17.83, "learning_rate": 3.0463600092517632e-05, "loss": 0.9367, "step": 205890 }, { "epoch": 17.83, "learning_rate": 3.04614317104198e-05, "loss": 0.7799, "step": 205900 }, { "epoch": 17.83, "learning_rate": 3.045926332832196e-05, "loss": 0.9998, "step": 205910 }, { "epoch": 17.83, "learning_rate": 3.045709494622412e-05, "loss": 0.8862, "step": 205920 }, { "epoch": 17.83, "learning_rate": 3.0454926564126284e-05, "loss": 0.6641, "step": 205930 }, { "epoch": 17.84, "learning_rate": 3.0452758182028447e-05, "loss": 0.9446, "step": 205940 }, { "epoch": 17.84, "learning_rate": 3.045058979993061e-05, "loss": 0.8035, "step": 205950 }, { "epoch": 17.84, "learning_rate": 3.044842141783277e-05, "loss": 0.792, "step": 205960 }, { "epoch": 17.84, "learning_rate": 3.0446253035734933e-05, "loss": 1.0235, "step": 205970 }, { "epoch": 17.84, "learning_rate": 3.04440846536371e-05, "loss": 0.7343, "step": 205980 }, { "epoch": 17.84, "learning_rate": 3.0441916271539262e-05, "loss": 0.7855, "step": 205990 }, { "epoch": 17.84, "learning_rate": 3.043974788944142e-05, "loss": 0.879, "step": 206000 }, { "epoch": 17.84, "learning_rate": 3.0437579507343585e-05, "loss": 0.8475, "step": 206010 }, { "epoch": 17.84, "learning_rate": 3.0435411125245748e-05, "loss": 0.8677, "step": 206020 }, { "epoch": 17.84, "learning_rate": 3.0433242743147907e-05, "loss": 0.821, "step": 206030 }, { "epoch": 17.84, "learning_rate": 3.043107436105007e-05, "loss": 0.9265, "step": 206040 }, { "epoch": 17.85, "learning_rate": 3.0428905978952237e-05, "loss": 0.8272, "step": 206050 }, { "epoch": 17.85, "learning_rate": 3.04267375968544e-05, "loss": 0.9011, "step": 206060 }, { "epoch": 17.85, "learning_rate": 3.042456921475656e-05, "loss": 0.8164, "step": 206070 }, { "epoch": 17.85, "learning_rate": 3.0422400832658722e-05, "loss": 0.9332, "step": 206080 }, { "epoch": 17.85, "learning_rate": 3.0420232450560885e-05, "loss": 0.7791, "step": 206090 }, { "epoch": 17.85, "learning_rate": 3.0418064068463048e-05, "loss": 0.7241, "step": 206100 }, { "epoch": 17.85, "learning_rate": 3.0415895686365208e-05, "loss": 0.7987, "step": 206110 }, { "epoch": 17.85, "learning_rate": 3.0413727304267374e-05, "loss": 0.9409, "step": 206120 }, { "epoch": 17.85, "learning_rate": 3.0411558922169537e-05, "loss": 0.9016, "step": 206130 }, { "epoch": 17.85, "learning_rate": 3.04093905400717e-05, "loss": 0.9747, "step": 206140 }, { "epoch": 17.85, "learning_rate": 3.040722215797386e-05, "loss": 0.884, "step": 206150 }, { "epoch": 17.85, "learning_rate": 3.0405053775876023e-05, "loss": 0.897, "step": 206160 }, { "epoch": 17.86, "learning_rate": 3.0402885393778186e-05, "loss": 0.9075, "step": 206170 }, { "epoch": 17.86, "learning_rate": 3.0400717011680352e-05, "loss": 0.8744, "step": 206180 }, { "epoch": 17.86, "learning_rate": 3.039854862958251e-05, "loss": 0.9492, "step": 206190 }, { "epoch": 17.86, "learning_rate": 3.0396380247484675e-05, "loss": 0.966, "step": 206200 }, { "epoch": 17.86, "learning_rate": 3.0394211865386838e-05, "loss": 0.6639, "step": 206210 }, { "epoch": 17.86, "learning_rate": 3.0392043483289e-05, "loss": 0.8442, "step": 206220 }, { "epoch": 17.86, "learning_rate": 3.038987510119116e-05, "loss": 0.7435, "step": 206230 }, { "epoch": 17.86, "learning_rate": 3.0387706719093323e-05, "loss": 0.8024, "step": 206240 }, { "epoch": 17.86, "learning_rate": 3.0385538336995486e-05, "loss": 0.8092, "step": 206250 }, { "epoch": 17.86, "learning_rate": 3.0383369954897653e-05, "loss": 0.79, "step": 206260 }, { "epoch": 17.86, "learning_rate": 3.0381201572799813e-05, "loss": 0.9755, "step": 206270 }, { "epoch": 17.87, "learning_rate": 3.0379033190701976e-05, "loss": 0.8786, "step": 206280 }, { "epoch": 17.87, "learning_rate": 3.037686480860414e-05, "loss": 0.9234, "step": 206290 }, { "epoch": 17.87, "learning_rate": 3.0374696426506298e-05, "loss": 0.8937, "step": 206300 }, { "epoch": 17.87, "learning_rate": 3.037252804440846e-05, "loss": 0.7305, "step": 206310 }, { "epoch": 17.87, "learning_rate": 3.0370359662310624e-05, "loss": 0.6919, "step": 206320 }, { "epoch": 17.87, "learning_rate": 3.036819128021279e-05, "loss": 0.8016, "step": 206330 }, { "epoch": 17.87, "learning_rate": 3.036602289811495e-05, "loss": 0.7388, "step": 206340 }, { "epoch": 17.87, "learning_rate": 3.0363854516017113e-05, "loss": 0.8749, "step": 206350 }, { "epoch": 17.87, "learning_rate": 3.0361686133919276e-05, "loss": 0.7759, "step": 206360 }, { "epoch": 17.87, "learning_rate": 3.035951775182144e-05, "loss": 0.8383, "step": 206370 }, { "epoch": 17.87, "learning_rate": 3.03573493697236e-05, "loss": 0.7877, "step": 206380 }, { "epoch": 17.87, "learning_rate": 3.0355180987625762e-05, "loss": 0.9029, "step": 206390 }, { "epoch": 17.88, "learning_rate": 3.0353012605527928e-05, "loss": 0.8604, "step": 206400 }, { "epoch": 17.88, "learning_rate": 3.035084422343009e-05, "loss": 0.8021, "step": 206410 }, { "epoch": 17.88, "learning_rate": 3.034867584133225e-05, "loss": 0.7987, "step": 206420 }, { "epoch": 17.88, "learning_rate": 3.0346507459234414e-05, "loss": 0.8164, "step": 206430 }, { "epoch": 17.88, "learning_rate": 3.0344339077136577e-05, "loss": 0.6908, "step": 206440 }, { "epoch": 17.88, "learning_rate": 3.034217069503874e-05, "loss": 0.9804, "step": 206450 }, { "epoch": 17.88, "learning_rate": 3.03400023129409e-05, "loss": 0.6813, "step": 206460 }, { "epoch": 17.88, "learning_rate": 3.0337833930843062e-05, "loss": 0.7681, "step": 206470 }, { "epoch": 17.88, "learning_rate": 3.033566554874523e-05, "loss": 0.9031, "step": 206480 }, { "epoch": 17.88, "learning_rate": 3.0333497166647392e-05, "loss": 0.8878, "step": 206490 }, { "epoch": 17.88, "learning_rate": 3.033132878454955e-05, "loss": 0.9065, "step": 206500 }, { "epoch": 17.89, "learning_rate": 3.0329160402451714e-05, "loss": 1.0649, "step": 206510 }, { "epoch": 17.89, "learning_rate": 3.0326992020353877e-05, "loss": 0.8843, "step": 206520 }, { "epoch": 17.89, "learning_rate": 3.032482363825604e-05, "loss": 0.8175, "step": 206530 }, { "epoch": 17.89, "learning_rate": 3.03226552561582e-05, "loss": 0.7625, "step": 206540 }, { "epoch": 17.89, "learning_rate": 3.0320486874060366e-05, "loss": 0.804, "step": 206550 }, { "epoch": 17.89, "learning_rate": 3.031831849196253e-05, "loss": 0.9241, "step": 206560 }, { "epoch": 17.89, "learning_rate": 3.031615010986469e-05, "loss": 0.8573, "step": 206570 }, { "epoch": 17.89, "learning_rate": 3.0313981727766852e-05, "loss": 0.8007, "step": 206580 }, { "epoch": 17.89, "learning_rate": 3.0311813345669015e-05, "loss": 0.9217, "step": 206590 }, { "epoch": 17.89, "learning_rate": 3.0309644963571178e-05, "loss": 0.8386, "step": 206600 }, { "epoch": 17.89, "learning_rate": 3.0307476581473338e-05, "loss": 0.9536, "step": 206610 }, { "epoch": 17.89, "learning_rate": 3.0305308199375504e-05, "loss": 0.7817, "step": 206620 }, { "epoch": 17.9, "learning_rate": 3.0303139817277667e-05, "loss": 0.9154, "step": 206630 }, { "epoch": 17.9, "learning_rate": 3.030097143517983e-05, "loss": 0.9821, "step": 206640 }, { "epoch": 17.9, "learning_rate": 3.029880305308199e-05, "loss": 0.8843, "step": 206650 }, { "epoch": 17.9, "learning_rate": 3.0296634670984153e-05, "loss": 0.8332, "step": 206660 }, { "epoch": 17.9, "learning_rate": 3.0294466288886316e-05, "loss": 0.8582, "step": 206670 }, { "epoch": 17.9, "learning_rate": 3.0292297906788482e-05, "loss": 0.8796, "step": 206680 }, { "epoch": 17.9, "learning_rate": 3.0290129524690642e-05, "loss": 0.9198, "step": 206690 }, { "epoch": 17.9, "learning_rate": 3.0287961142592805e-05, "loss": 0.8663, "step": 206700 }, { "epoch": 17.9, "learning_rate": 3.0285792760494968e-05, "loss": 0.7272, "step": 206710 }, { "epoch": 17.9, "learning_rate": 3.028362437839713e-05, "loss": 0.7354, "step": 206720 }, { "epoch": 17.9, "learning_rate": 3.028145599629929e-05, "loss": 0.9712, "step": 206730 }, { "epoch": 17.9, "learning_rate": 3.0279287614201453e-05, "loss": 0.7875, "step": 206740 }, { "epoch": 17.91, "learning_rate": 3.0277119232103616e-05, "loss": 0.7637, "step": 206750 }, { "epoch": 17.91, "learning_rate": 3.0274950850005783e-05, "loss": 0.8194, "step": 206760 }, { "epoch": 17.91, "learning_rate": 3.0272782467907942e-05, "loss": 0.7959, "step": 206770 }, { "epoch": 17.91, "learning_rate": 3.0270614085810105e-05, "loss": 0.8824, "step": 206780 }, { "epoch": 17.91, "learning_rate": 3.026844570371227e-05, "loss": 0.9148, "step": 206790 }, { "epoch": 17.91, "learning_rate": 3.0266277321614428e-05, "loss": 0.7517, "step": 206800 }, { "epoch": 17.91, "learning_rate": 3.026410893951659e-05, "loss": 0.9619, "step": 206810 }, { "epoch": 17.91, "learning_rate": 3.0261940557418754e-05, "loss": 0.7763, "step": 206820 }, { "epoch": 17.91, "learning_rate": 3.025977217532092e-05, "loss": 0.9017, "step": 206830 }, { "epoch": 17.91, "learning_rate": 3.025760379322308e-05, "loss": 0.9587, "step": 206840 }, { "epoch": 17.91, "learning_rate": 3.0255435411125243e-05, "loss": 0.7749, "step": 206850 }, { "epoch": 17.92, "learning_rate": 3.0253267029027406e-05, "loss": 0.7302, "step": 206860 }, { "epoch": 17.92, "learning_rate": 3.025109864692957e-05, "loss": 0.8562, "step": 206870 }, { "epoch": 17.92, "learning_rate": 3.024893026483173e-05, "loss": 0.9301, "step": 206880 }, { "epoch": 17.92, "learning_rate": 3.024676188273389e-05, "loss": 0.7647, "step": 206890 }, { "epoch": 17.92, "learning_rate": 3.0244593500636058e-05, "loss": 0.8466, "step": 206900 }, { "epoch": 17.92, "learning_rate": 3.024242511853822e-05, "loss": 0.8338, "step": 206910 }, { "epoch": 17.92, "learning_rate": 3.024025673644038e-05, "loss": 0.7212, "step": 206920 }, { "epoch": 17.92, "learning_rate": 3.0238088354342544e-05, "loss": 0.8306, "step": 206930 }, { "epoch": 17.92, "learning_rate": 3.0235919972244707e-05, "loss": 0.8437, "step": 206940 }, { "epoch": 17.92, "learning_rate": 3.023375159014687e-05, "loss": 0.8985, "step": 206950 }, { "epoch": 17.92, "learning_rate": 3.023158320804903e-05, "loss": 0.847, "step": 206960 }, { "epoch": 17.92, "learning_rate": 3.0229414825951196e-05, "loss": 0.752, "step": 206970 }, { "epoch": 17.93, "learning_rate": 3.022724644385336e-05, "loss": 0.7661, "step": 206980 }, { "epoch": 17.93, "learning_rate": 3.022507806175552e-05, "loss": 0.9584, "step": 206990 }, { "epoch": 17.93, "learning_rate": 3.022290967965768e-05, "loss": 0.8292, "step": 207000 }, { "epoch": 17.93, "learning_rate": 3.0220741297559844e-05, "loss": 0.8501, "step": 207010 }, { "epoch": 17.93, "learning_rate": 3.0218572915462007e-05, "loss": 0.6948, "step": 207020 }, { "epoch": 17.93, "learning_rate": 3.021640453336417e-05, "loss": 0.9098, "step": 207030 }, { "epoch": 17.93, "learning_rate": 3.021423615126633e-05, "loss": 0.8062, "step": 207040 }, { "epoch": 17.93, "learning_rate": 3.0212067769168496e-05, "loss": 1.0761, "step": 207050 }, { "epoch": 17.93, "learning_rate": 3.020989938707066e-05, "loss": 0.8088, "step": 207060 }, { "epoch": 17.93, "learning_rate": 3.020773100497282e-05, "loss": 0.7029, "step": 207070 }, { "epoch": 17.93, "learning_rate": 3.0205562622874982e-05, "loss": 0.8962, "step": 207080 }, { "epoch": 17.94, "learning_rate": 3.0203394240777145e-05, "loss": 0.6801, "step": 207090 }, { "epoch": 17.94, "learning_rate": 3.0201225858679308e-05, "loss": 0.8535, "step": 207100 }, { "epoch": 17.94, "learning_rate": 3.0199057476581468e-05, "loss": 0.7659, "step": 207110 }, { "epoch": 17.94, "learning_rate": 3.0196889094483634e-05, "loss": 0.8361, "step": 207120 }, { "epoch": 17.94, "learning_rate": 3.0194720712385797e-05, "loss": 0.7628, "step": 207130 }, { "epoch": 17.94, "learning_rate": 3.019255233028796e-05, "loss": 0.7275, "step": 207140 }, { "epoch": 17.94, "learning_rate": 3.019038394819012e-05, "loss": 0.9784, "step": 207150 }, { "epoch": 17.94, "learning_rate": 3.0188215566092283e-05, "loss": 0.8515, "step": 207160 }, { "epoch": 17.94, "learning_rate": 3.0186047183994446e-05, "loss": 0.6992, "step": 207170 }, { "epoch": 17.94, "learning_rate": 3.0183878801896612e-05, "loss": 0.6905, "step": 207180 }, { "epoch": 17.94, "learning_rate": 3.018171041979877e-05, "loss": 0.8455, "step": 207190 }, { "epoch": 17.94, "learning_rate": 3.0179542037700935e-05, "loss": 0.886, "step": 207200 }, { "epoch": 17.95, "learning_rate": 3.0177373655603098e-05, "loss": 0.8503, "step": 207210 }, { "epoch": 17.95, "learning_rate": 3.017520527350526e-05, "loss": 0.8595, "step": 207220 }, { "epoch": 17.95, "learning_rate": 3.017303689140742e-05, "loss": 0.8746, "step": 207230 }, { "epoch": 17.95, "learning_rate": 3.0170868509309583e-05, "loss": 0.8085, "step": 207240 }, { "epoch": 17.95, "learning_rate": 3.016870012721175e-05, "loss": 0.8064, "step": 207250 }, { "epoch": 17.95, "learning_rate": 3.0166531745113913e-05, "loss": 0.8146, "step": 207260 }, { "epoch": 17.95, "learning_rate": 3.0164363363016072e-05, "loss": 0.9182, "step": 207270 }, { "epoch": 17.95, "learning_rate": 3.0162194980918235e-05, "loss": 0.8726, "step": 207280 }, { "epoch": 17.95, "learning_rate": 3.0160026598820398e-05, "loss": 0.7673, "step": 207290 }, { "epoch": 17.95, "learning_rate": 3.015785821672256e-05, "loss": 0.8341, "step": 207300 }, { "epoch": 17.95, "learning_rate": 3.015568983462472e-05, "loss": 0.8069, "step": 207310 }, { "epoch": 17.96, "learning_rate": 3.0153521452526884e-05, "loss": 0.8129, "step": 207320 }, { "epoch": 17.96, "learning_rate": 3.015135307042905e-05, "loss": 0.9612, "step": 207330 }, { "epoch": 17.96, "learning_rate": 3.014918468833121e-05, "loss": 0.8384, "step": 207340 }, { "epoch": 17.96, "learning_rate": 3.0147016306233373e-05, "loss": 0.9373, "step": 207350 }, { "epoch": 17.96, "learning_rate": 3.0144847924135536e-05, "loss": 0.7735, "step": 207360 }, { "epoch": 17.96, "learning_rate": 3.01426795420377e-05, "loss": 0.8395, "step": 207370 }, { "epoch": 17.96, "learning_rate": 3.014051115993986e-05, "loss": 0.6446, "step": 207380 }, { "epoch": 17.96, "learning_rate": 3.013834277784202e-05, "loss": 0.8312, "step": 207390 }, { "epoch": 17.96, "learning_rate": 3.0136174395744188e-05, "loss": 0.8688, "step": 207400 }, { "epoch": 17.96, "learning_rate": 3.013400601364635e-05, "loss": 0.8722, "step": 207410 }, { "epoch": 17.96, "learning_rate": 3.013183763154851e-05, "loss": 0.8854, "step": 207420 }, { "epoch": 17.96, "learning_rate": 3.0129669249450673e-05, "loss": 0.7962, "step": 207430 }, { "epoch": 17.97, "learning_rate": 3.0127500867352836e-05, "loss": 0.6641, "step": 207440 }, { "epoch": 17.97, "learning_rate": 3.0125332485255e-05, "loss": 0.7449, "step": 207450 }, { "epoch": 17.97, "learning_rate": 3.012316410315716e-05, "loss": 0.7652, "step": 207460 }, { "epoch": 17.97, "learning_rate": 3.0120995721059325e-05, "loss": 0.8619, "step": 207470 }, { "epoch": 17.97, "learning_rate": 3.011882733896149e-05, "loss": 0.7445, "step": 207480 }, { "epoch": 17.97, "learning_rate": 3.011665895686365e-05, "loss": 0.8741, "step": 207490 }, { "epoch": 17.97, "learning_rate": 3.011449057476581e-05, "loss": 0.8821, "step": 207500 }, { "epoch": 17.97, "learning_rate": 3.0112322192667974e-05, "loss": 0.8133, "step": 207510 }, { "epoch": 17.97, "learning_rate": 3.0110153810570137e-05, "loss": 0.9542, "step": 207520 }, { "epoch": 17.97, "learning_rate": 3.0107985428472304e-05, "loss": 0.8099, "step": 207530 }, { "epoch": 17.97, "learning_rate": 3.010581704637446e-05, "loss": 0.9795, "step": 207540 }, { "epoch": 17.98, "learning_rate": 3.0103648664276626e-05, "loss": 0.6409, "step": 207550 }, { "epoch": 17.98, "learning_rate": 3.010148028217879e-05, "loss": 0.9481, "step": 207560 }, { "epoch": 17.98, "learning_rate": 3.009931190008095e-05, "loss": 0.7622, "step": 207570 }, { "epoch": 17.98, "learning_rate": 3.0097143517983112e-05, "loss": 0.837, "step": 207580 }, { "epoch": 17.98, "learning_rate": 3.0094975135885275e-05, "loss": 0.8279, "step": 207590 }, { "epoch": 17.98, "learning_rate": 3.0092806753787438e-05, "loss": 1.0215, "step": 207600 }, { "epoch": 17.98, "learning_rate": 3.0090638371689597e-05, "loss": 0.684, "step": 207610 }, { "epoch": 17.98, "learning_rate": 3.0088469989591764e-05, "loss": 0.751, "step": 207620 }, { "epoch": 17.98, "learning_rate": 3.0086301607493927e-05, "loss": 0.8817, "step": 207630 }, { "epoch": 17.98, "learning_rate": 3.008413322539609e-05, "loss": 0.683, "step": 207640 }, { "epoch": 17.98, "learning_rate": 3.008196484329825e-05, "loss": 0.8089, "step": 207650 }, { "epoch": 17.98, "learning_rate": 3.0079796461200412e-05, "loss": 0.871, "step": 207660 }, { "epoch": 17.99, "learning_rate": 3.0077628079102575e-05, "loss": 0.6796, "step": 207670 }, { "epoch": 17.99, "learning_rate": 3.0075459697004742e-05, "loss": 0.9432, "step": 207680 }, { "epoch": 17.99, "learning_rate": 3.00732913149069e-05, "loss": 1.0314, "step": 207690 }, { "epoch": 17.99, "learning_rate": 3.0071122932809064e-05, "loss": 0.7654, "step": 207700 }, { "epoch": 17.99, "learning_rate": 3.0068954550711227e-05, "loss": 0.8837, "step": 207710 }, { "epoch": 17.99, "learning_rate": 3.006678616861339e-05, "loss": 0.8474, "step": 207720 }, { "epoch": 17.99, "learning_rate": 3.006461778651555e-05, "loss": 0.9054, "step": 207730 }, { "epoch": 17.99, "learning_rate": 3.0062449404417713e-05, "loss": 0.7823, "step": 207740 }, { "epoch": 17.99, "learning_rate": 3.006028102231988e-05, "loss": 0.7656, "step": 207750 }, { "epoch": 17.99, "learning_rate": 3.0058112640222042e-05, "loss": 0.7494, "step": 207760 }, { "epoch": 17.99, "learning_rate": 3.0055944258124202e-05, "loss": 0.6902, "step": 207770 }, { "epoch": 18.0, "learning_rate": 3.0053775876026365e-05, "loss": 0.8015, "step": 207780 }, { "epoch": 18.0, "learning_rate": 3.0051607493928528e-05, "loss": 0.9033, "step": 207790 }, { "epoch": 18.0, "learning_rate": 3.004943911183069e-05, "loss": 0.7478, "step": 207800 }, { "epoch": 18.0, "learning_rate": 3.004727072973285e-05, "loss": 0.9313, "step": 207810 }, { "epoch": 18.0, "learning_rate": 3.0045102347635014e-05, "loss": 0.9113, "step": 207820 }, { "epoch": 18.0, "learning_rate": 3.004293396553718e-05, "loss": 0.8599, "step": 207830 }, { "epoch": 18.0, "eval_cer": 0.5350791930195474, "eval_em": 0.05157297576070139, "eval_f1": 0.05157297576070139, "eval_loss": 1.4710338115692139, "eval_runtime": 254.1277, "eval_samples_per_second": 7.63, "eval_steps_per_second": 0.956, "eval_wer": 0.9484270242392986, "step": 207837 }, { "epoch": 18.0, "learning_rate": 3.004076558343934e-05, "loss": 0.8566, "step": 207840 }, { "epoch": 18.0, "learning_rate": 3.0038597201341503e-05, "loss": 0.6228, "step": 207850 }, { "epoch": 18.0, "learning_rate": 3.0036428819243666e-05, "loss": 0.8227, "step": 207860 }, { "epoch": 18.0, "learning_rate": 3.003426043714583e-05, "loss": 0.9897, "step": 207870 }, { "epoch": 18.0, "learning_rate": 3.003209205504799e-05, "loss": 0.8371, "step": 207880 }, { "epoch": 18.0, "learning_rate": 3.002992367295015e-05, "loss": 0.9757, "step": 207890 }, { "epoch": 18.01, "learning_rate": 3.0027755290852318e-05, "loss": 0.9093, "step": 207900 }, { "epoch": 18.01, "learning_rate": 3.002558690875448e-05, "loss": 0.8167, "step": 207910 }, { "epoch": 18.01, "learning_rate": 3.002341852665664e-05, "loss": 0.7514, "step": 207920 }, { "epoch": 18.01, "learning_rate": 3.0021250144558803e-05, "loss": 0.8072, "step": 207930 }, { "epoch": 18.01, "learning_rate": 3.0019081762460966e-05, "loss": 0.6641, "step": 207940 }, { "epoch": 18.01, "learning_rate": 3.001691338036313e-05, "loss": 0.7553, "step": 207950 }, { "epoch": 18.01, "learning_rate": 3.001474499826529e-05, "loss": 0.8141, "step": 207960 }, { "epoch": 18.01, "learning_rate": 3.0012576616167455e-05, "loss": 0.8592, "step": 207970 }, { "epoch": 18.01, "learning_rate": 3.001040823406962e-05, "loss": 0.9511, "step": 207980 }, { "epoch": 18.01, "learning_rate": 3.000823985197178e-05, "loss": 0.7756, "step": 207990 }, { "epoch": 18.01, "learning_rate": 3.000607146987394e-05, "loss": 0.6947, "step": 208000 }, { "epoch": 18.01, "learning_rate": 3.0003903087776104e-05, "loss": 0.902, "step": 208010 }, { "epoch": 18.02, "learning_rate": 3.0001734705678267e-05, "loss": 0.7487, "step": 208020 }, { "epoch": 18.02, "learning_rate": 2.9999566323580433e-05, "loss": 0.7501, "step": 208030 }, { "epoch": 18.02, "learning_rate": 2.9997397941482593e-05, "loss": 0.8821, "step": 208040 }, { "epoch": 18.02, "learning_rate": 2.9995229559384756e-05, "loss": 0.7703, "step": 208050 }, { "epoch": 18.02, "learning_rate": 2.999306117728692e-05, "loss": 0.7601, "step": 208060 }, { "epoch": 18.02, "learning_rate": 2.9990892795189082e-05, "loss": 0.7426, "step": 208070 }, { "epoch": 18.02, "learning_rate": 2.998872441309124e-05, "loss": 0.8621, "step": 208080 }, { "epoch": 18.02, "learning_rate": 2.9986556030993405e-05, "loss": 0.6371, "step": 208090 }, { "epoch": 18.02, "learning_rate": 2.9984387648895568e-05, "loss": 0.7483, "step": 208100 }, { "epoch": 18.02, "learning_rate": 2.9982219266797727e-05, "loss": 0.7246, "step": 208110 }, { "epoch": 18.02, "learning_rate": 2.9980050884699894e-05, "loss": 0.8755, "step": 208120 }, { "epoch": 18.03, "learning_rate": 2.9977882502602057e-05, "loss": 0.7148, "step": 208130 }, { "epoch": 18.03, "learning_rate": 2.997571412050422e-05, "loss": 0.7181, "step": 208140 }, { "epoch": 18.03, "learning_rate": 2.997354573840638e-05, "loss": 0.8444, "step": 208150 }, { "epoch": 18.03, "learning_rate": 2.9971377356308542e-05, "loss": 0.833, "step": 208160 }, { "epoch": 18.03, "learning_rate": 2.9969208974210705e-05, "loss": 0.8771, "step": 208170 }, { "epoch": 18.03, "learning_rate": 2.996704059211287e-05, "loss": 0.7343, "step": 208180 }, { "epoch": 18.03, "learning_rate": 2.996487221001503e-05, "loss": 0.9009, "step": 208190 }, { "epoch": 18.03, "learning_rate": 2.9962703827917194e-05, "loss": 0.6925, "step": 208200 }, { "epoch": 18.03, "learning_rate": 2.9960535445819357e-05, "loss": 0.7603, "step": 208210 }, { "epoch": 18.03, "learning_rate": 2.995836706372152e-05, "loss": 0.7762, "step": 208220 }, { "epoch": 18.03, "learning_rate": 2.995619868162368e-05, "loss": 0.7884, "step": 208230 }, { "epoch": 18.03, "learning_rate": 2.9954030299525843e-05, "loss": 0.8002, "step": 208240 }, { "epoch": 18.04, "learning_rate": 2.995186191742801e-05, "loss": 0.9352, "step": 208250 }, { "epoch": 18.04, "learning_rate": 2.9949693535330172e-05, "loss": 0.8017, "step": 208260 }, { "epoch": 18.04, "learning_rate": 2.9947525153232332e-05, "loss": 0.9393, "step": 208270 }, { "epoch": 18.04, "learning_rate": 2.9945356771134495e-05, "loss": 0.857, "step": 208280 }, { "epoch": 18.04, "learning_rate": 2.9943188389036658e-05, "loss": 0.7598, "step": 208290 }, { "epoch": 18.04, "learning_rate": 2.994102000693882e-05, "loss": 0.6152, "step": 208300 }, { "epoch": 18.04, "learning_rate": 2.993885162484098e-05, "loss": 0.7835, "step": 208310 }, { "epoch": 18.04, "learning_rate": 2.9936683242743147e-05, "loss": 1.0185, "step": 208320 }, { "epoch": 18.04, "learning_rate": 2.993451486064531e-05, "loss": 0.7338, "step": 208330 }, { "epoch": 18.04, "learning_rate": 2.993234647854747e-05, "loss": 0.7576, "step": 208340 }, { "epoch": 18.04, "learning_rate": 2.9930178096449633e-05, "loss": 0.7856, "step": 208350 }, { "epoch": 18.05, "learning_rate": 2.9928009714351796e-05, "loss": 0.8488, "step": 208360 }, { "epoch": 18.05, "learning_rate": 2.992584133225396e-05, "loss": 0.817, "step": 208370 }, { "epoch": 18.05, "learning_rate": 2.9923672950156118e-05, "loss": 0.751, "step": 208380 }, { "epoch": 18.05, "learning_rate": 2.992150456805828e-05, "loss": 0.8846, "step": 208390 }, { "epoch": 18.05, "learning_rate": 2.9919336185960448e-05, "loss": 0.823, "step": 208400 }, { "epoch": 18.05, "learning_rate": 2.991716780386261e-05, "loss": 0.9013, "step": 208410 }, { "epoch": 18.05, "learning_rate": 2.991499942176477e-05, "loss": 0.7623, "step": 208420 }, { "epoch": 18.05, "learning_rate": 2.9912831039666933e-05, "loss": 0.6754, "step": 208430 }, { "epoch": 18.05, "learning_rate": 2.9910662657569096e-05, "loss": 0.7715, "step": 208440 }, { "epoch": 18.05, "learning_rate": 2.990849427547126e-05, "loss": 0.8222, "step": 208450 }, { "epoch": 18.05, "learning_rate": 2.990632589337342e-05, "loss": 0.8254, "step": 208460 }, { "epoch": 18.05, "learning_rate": 2.9904157511275585e-05, "loss": 0.8048, "step": 208470 }, { "epoch": 18.06, "learning_rate": 2.9901989129177748e-05, "loss": 0.885, "step": 208480 }, { "epoch": 18.06, "learning_rate": 2.989982074707991e-05, "loss": 0.716, "step": 208490 }, { "epoch": 18.06, "learning_rate": 2.989765236498207e-05, "loss": 0.7786, "step": 208500 }, { "epoch": 18.06, "learning_rate": 2.9895483982884234e-05, "loss": 0.7124, "step": 208510 }, { "epoch": 18.06, "learning_rate": 2.9893315600786397e-05, "loss": 0.8268, "step": 208520 }, { "epoch": 18.06, "learning_rate": 2.9891147218688563e-05, "loss": 0.6162, "step": 208530 }, { "epoch": 18.06, "learning_rate": 2.9888978836590723e-05, "loss": 0.6682, "step": 208540 }, { "epoch": 18.06, "learning_rate": 2.9886810454492886e-05, "loss": 0.7582, "step": 208550 }, { "epoch": 18.06, "learning_rate": 2.988464207239505e-05, "loss": 0.7405, "step": 208560 }, { "epoch": 18.06, "learning_rate": 2.9882473690297212e-05, "loss": 0.8267, "step": 208570 }, { "epoch": 18.06, "learning_rate": 2.988030530819937e-05, "loss": 0.6276, "step": 208580 }, { "epoch": 18.07, "learning_rate": 2.9878136926101534e-05, "loss": 0.8012, "step": 208590 }, { "epoch": 18.07, "learning_rate": 2.98759685440037e-05, "loss": 0.5931, "step": 208600 }, { "epoch": 18.07, "learning_rate": 2.9873800161905857e-05, "loss": 0.9345, "step": 208610 }, { "epoch": 18.07, "learning_rate": 2.9871631779808023e-05, "loss": 0.7227, "step": 208620 }, { "epoch": 18.07, "learning_rate": 2.9869463397710186e-05, "loss": 0.6969, "step": 208630 }, { "epoch": 18.07, "learning_rate": 2.986729501561235e-05, "loss": 0.7906, "step": 208640 }, { "epoch": 18.07, "learning_rate": 2.986512663351451e-05, "loss": 0.6969, "step": 208650 }, { "epoch": 18.07, "learning_rate": 2.9862958251416672e-05, "loss": 0.8178, "step": 208660 }, { "epoch": 18.07, "learning_rate": 2.9860789869318835e-05, "loss": 0.8074, "step": 208670 }, { "epoch": 18.07, "learning_rate": 2.9858621487221e-05, "loss": 0.8024, "step": 208680 }, { "epoch": 18.07, "learning_rate": 2.985645310512316e-05, "loss": 1.0525, "step": 208690 }, { "epoch": 18.07, "learning_rate": 2.9854284723025324e-05, "loss": 0.9151, "step": 208700 }, { "epoch": 18.08, "learning_rate": 2.9852116340927487e-05, "loss": 0.7676, "step": 208710 }, { "epoch": 18.08, "learning_rate": 2.984994795882965e-05, "loss": 0.8682, "step": 208720 }, { "epoch": 18.08, "learning_rate": 2.984777957673181e-05, "loss": 0.8687, "step": 208730 }, { "epoch": 18.08, "learning_rate": 2.9845611194633973e-05, "loss": 0.7774, "step": 208740 }, { "epoch": 18.08, "learning_rate": 2.984344281253614e-05, "loss": 0.7476, "step": 208750 }, { "epoch": 18.08, "learning_rate": 2.9841274430438302e-05, "loss": 0.645, "step": 208760 }, { "epoch": 18.08, "learning_rate": 2.9839106048340462e-05, "loss": 0.6462, "step": 208770 }, { "epoch": 18.08, "learning_rate": 2.9836937666242625e-05, "loss": 0.7791, "step": 208780 }, { "epoch": 18.08, "learning_rate": 2.9834769284144788e-05, "loss": 0.7242, "step": 208790 }, { "epoch": 18.08, "learning_rate": 2.983260090204695e-05, "loss": 0.8942, "step": 208800 }, { "epoch": 18.08, "learning_rate": 2.983043251994911e-05, "loss": 0.8099, "step": 208810 }, { "epoch": 18.09, "learning_rate": 2.9828264137851277e-05, "loss": 0.8341, "step": 208820 }, { "epoch": 18.09, "learning_rate": 2.982609575575344e-05, "loss": 0.7979, "step": 208830 }, { "epoch": 18.09, "learning_rate": 2.9823927373655603e-05, "loss": 0.7597, "step": 208840 }, { "epoch": 18.09, "learning_rate": 2.9821758991557762e-05, "loss": 0.976, "step": 208850 }, { "epoch": 18.09, "learning_rate": 2.9819590609459925e-05, "loss": 0.7744, "step": 208860 }, { "epoch": 18.09, "learning_rate": 2.981742222736209e-05, "loss": 0.8571, "step": 208870 }, { "epoch": 18.09, "learning_rate": 2.9815253845264248e-05, "loss": 0.6488, "step": 208880 }, { "epoch": 18.09, "learning_rate": 2.981308546316641e-05, "loss": 0.8302, "step": 208890 }, { "epoch": 18.09, "learning_rate": 2.9810917081068577e-05, "loss": 0.665, "step": 208900 }, { "epoch": 18.09, "learning_rate": 2.980874869897074e-05, "loss": 0.9212, "step": 208910 }, { "epoch": 18.09, "learning_rate": 2.98065803168729e-05, "loss": 0.7616, "step": 208920 }, { "epoch": 18.09, "learning_rate": 2.9804411934775063e-05, "loss": 0.6356, "step": 208930 }, { "epoch": 18.1, "learning_rate": 2.9802243552677226e-05, "loss": 0.824, "step": 208940 }, { "epoch": 18.1, "learning_rate": 2.980007517057939e-05, "loss": 0.7685, "step": 208950 }, { "epoch": 18.1, "learning_rate": 2.979790678848155e-05, "loss": 0.6573, "step": 208960 }, { "epoch": 18.1, "learning_rate": 2.9795738406383715e-05, "loss": 0.7658, "step": 208970 }, { "epoch": 18.1, "learning_rate": 2.9793570024285878e-05, "loss": 0.7968, "step": 208980 }, { "epoch": 18.1, "learning_rate": 2.979140164218804e-05, "loss": 0.7894, "step": 208990 }, { "epoch": 18.1, "learning_rate": 2.97892332600902e-05, "loss": 0.897, "step": 209000 }, { "epoch": 18.1, "learning_rate": 2.9787064877992364e-05, "loss": 0.7575, "step": 209010 }, { "epoch": 18.1, "learning_rate": 2.9784896495894527e-05, "loss": 0.8215, "step": 209020 }, { "epoch": 18.1, "learning_rate": 2.9782728113796693e-05, "loss": 0.8299, "step": 209030 }, { "epoch": 18.1, "learning_rate": 2.9780559731698853e-05, "loss": 0.753, "step": 209040 }, { "epoch": 18.11, "learning_rate": 2.9778391349601016e-05, "loss": 0.8327, "step": 209050 }, { "epoch": 18.11, "learning_rate": 2.977622296750318e-05, "loss": 0.8134, "step": 209060 }, { "epoch": 18.11, "learning_rate": 2.977405458540534e-05, "loss": 0.7497, "step": 209070 }, { "epoch": 18.11, "learning_rate": 2.97718862033075e-05, "loss": 0.7617, "step": 209080 }, { "epoch": 18.11, "learning_rate": 2.9769717821209664e-05, "loss": 0.7499, "step": 209090 }, { "epoch": 18.11, "learning_rate": 2.976754943911183e-05, "loss": 0.6901, "step": 209100 }, { "epoch": 18.11, "learning_rate": 2.976538105701399e-05, "loss": 0.8381, "step": 209110 }, { "epoch": 18.11, "learning_rate": 2.9763212674916153e-05, "loss": 0.8382, "step": 209120 }, { "epoch": 18.11, "learning_rate": 2.9761044292818316e-05, "loss": 0.7613, "step": 209130 }, { "epoch": 18.11, "learning_rate": 2.975887591072048e-05, "loss": 0.8238, "step": 209140 }, { "epoch": 18.11, "learning_rate": 2.975670752862264e-05, "loss": 0.8296, "step": 209150 }, { "epoch": 18.11, "learning_rate": 2.9754539146524802e-05, "loss": 1.0522, "step": 209160 }, { "epoch": 18.12, "learning_rate": 2.9752370764426965e-05, "loss": 0.6968, "step": 209170 }, { "epoch": 18.12, "learning_rate": 2.975020238232913e-05, "loss": 0.7261, "step": 209180 }, { "epoch": 18.12, "learning_rate": 2.974803400023129e-05, "loss": 0.8078, "step": 209190 }, { "epoch": 18.12, "learning_rate": 2.9745865618133454e-05, "loss": 0.8075, "step": 209200 }, { "epoch": 18.12, "learning_rate": 2.9743697236035617e-05, "loss": 0.7895, "step": 209210 }, { "epoch": 18.12, "learning_rate": 2.974152885393778e-05, "loss": 0.8176, "step": 209220 }, { "epoch": 18.12, "learning_rate": 2.973936047183994e-05, "loss": 0.8947, "step": 209230 }, { "epoch": 18.12, "learning_rate": 2.9737192089742103e-05, "loss": 0.7607, "step": 209240 }, { "epoch": 18.12, "learning_rate": 2.973502370764427e-05, "loss": 0.7845, "step": 209250 }, { "epoch": 18.12, "learning_rate": 2.9732855325546432e-05, "loss": 0.6876, "step": 209260 }, { "epoch": 18.12, "learning_rate": 2.973068694344859e-05, "loss": 0.784, "step": 209270 }, { "epoch": 18.12, "learning_rate": 2.9728518561350755e-05, "loss": 0.8425, "step": 209280 }, { "epoch": 18.13, "learning_rate": 2.9726350179252918e-05, "loss": 0.7681, "step": 209290 }, { "epoch": 18.13, "learning_rate": 2.972418179715508e-05, "loss": 0.6848, "step": 209300 }, { "epoch": 18.13, "learning_rate": 2.972201341505724e-05, "loss": 0.6717, "step": 209310 }, { "epoch": 18.13, "learning_rate": 2.9719845032959407e-05, "loss": 0.8326, "step": 209320 }, { "epoch": 18.13, "learning_rate": 2.971767665086157e-05, "loss": 0.8606, "step": 209330 }, { "epoch": 18.13, "learning_rate": 2.9715508268763733e-05, "loss": 0.7464, "step": 209340 }, { "epoch": 18.13, "learning_rate": 2.9713339886665892e-05, "loss": 0.9466, "step": 209350 }, { "epoch": 18.13, "learning_rate": 2.9711171504568055e-05, "loss": 0.7946, "step": 209360 }, { "epoch": 18.13, "learning_rate": 2.9709003122470218e-05, "loss": 0.8144, "step": 209370 }, { "epoch": 18.13, "learning_rate": 2.9706834740372378e-05, "loss": 0.8716, "step": 209380 }, { "epoch": 18.13, "learning_rate": 2.9704666358274544e-05, "loss": 0.6855, "step": 209390 }, { "epoch": 18.14, "learning_rate": 2.9702497976176707e-05, "loss": 0.8122, "step": 209400 }, { "epoch": 18.14, "learning_rate": 2.970032959407887e-05, "loss": 0.9289, "step": 209410 }, { "epoch": 18.14, "learning_rate": 2.969816121198103e-05, "loss": 0.8037, "step": 209420 }, { "epoch": 18.14, "learning_rate": 2.9695992829883193e-05, "loss": 0.7966, "step": 209430 }, { "epoch": 18.14, "learning_rate": 2.9693824447785356e-05, "loss": 0.8401, "step": 209440 }, { "epoch": 18.14, "learning_rate": 2.969165606568752e-05, "loss": 0.7769, "step": 209450 }, { "epoch": 18.14, "learning_rate": 2.968948768358968e-05, "loss": 0.8288, "step": 209460 }, { "epoch": 18.14, "learning_rate": 2.9687319301491845e-05, "loss": 0.8135, "step": 209470 }, { "epoch": 18.14, "learning_rate": 2.9685150919394008e-05, "loss": 0.6867, "step": 209480 }, { "epoch": 18.14, "learning_rate": 2.968298253729617e-05, "loss": 0.7361, "step": 209490 }, { "epoch": 18.14, "learning_rate": 2.968081415519833e-05, "loss": 0.828, "step": 209500 }, { "epoch": 18.14, "learning_rate": 2.9678645773100494e-05, "loss": 0.851, "step": 209510 }, { "epoch": 18.15, "learning_rate": 2.9676477391002657e-05, "loss": 0.7332, "step": 209520 }, { "epoch": 18.15, "learning_rate": 2.9674309008904823e-05, "loss": 0.7297, "step": 209530 }, { "epoch": 18.15, "learning_rate": 2.9672140626806983e-05, "loss": 0.6886, "step": 209540 }, { "epoch": 18.15, "learning_rate": 2.9669972244709146e-05, "loss": 0.8113, "step": 209550 }, { "epoch": 18.15, "learning_rate": 2.966780386261131e-05, "loss": 0.8533, "step": 209560 }, { "epoch": 18.15, "learning_rate": 2.966563548051347e-05, "loss": 0.7755, "step": 209570 }, { "epoch": 18.15, "learning_rate": 2.966346709841563e-05, "loss": 0.8563, "step": 209580 }, { "epoch": 18.15, "learning_rate": 2.9661298716317794e-05, "loss": 0.7732, "step": 209590 }, { "epoch": 18.15, "learning_rate": 2.965913033421996e-05, "loss": 0.8269, "step": 209600 }, { "epoch": 18.15, "learning_rate": 2.9656961952122124e-05, "loss": 0.8883, "step": 209610 }, { "epoch": 18.15, "learning_rate": 2.9654793570024283e-05, "loss": 0.6324, "step": 209620 }, { "epoch": 18.16, "learning_rate": 2.9652625187926446e-05, "loss": 0.9051, "step": 209630 }, { "epoch": 18.16, "learning_rate": 2.965045680582861e-05, "loss": 0.7753, "step": 209640 }, { "epoch": 18.16, "learning_rate": 2.964828842373077e-05, "loss": 0.9692, "step": 209650 }, { "epoch": 18.16, "learning_rate": 2.9646120041632932e-05, "loss": 0.8929, "step": 209660 }, { "epoch": 18.16, "learning_rate": 2.9643951659535098e-05, "loss": 0.6907, "step": 209670 }, { "epoch": 18.16, "learning_rate": 2.964178327743726e-05, "loss": 0.8717, "step": 209680 }, { "epoch": 18.16, "learning_rate": 2.963961489533942e-05, "loss": 0.8277, "step": 209690 }, { "epoch": 18.16, "learning_rate": 2.9637446513241584e-05, "loss": 0.8692, "step": 209700 }, { "epoch": 18.16, "learning_rate": 2.9635278131143747e-05, "loss": 0.678, "step": 209710 }, { "epoch": 18.16, "learning_rate": 2.963310974904591e-05, "loss": 0.6894, "step": 209720 }, { "epoch": 18.16, "learning_rate": 2.963094136694807e-05, "loss": 0.7707, "step": 209730 }, { "epoch": 18.16, "learning_rate": 2.9628772984850232e-05, "loss": 0.6949, "step": 209740 }, { "epoch": 18.17, "learning_rate": 2.96266046027524e-05, "loss": 0.8622, "step": 209750 }, { "epoch": 18.17, "learning_rate": 2.9624436220654562e-05, "loss": 0.9073, "step": 209760 }, { "epoch": 18.17, "learning_rate": 2.962226783855672e-05, "loss": 0.8755, "step": 209770 }, { "epoch": 18.17, "learning_rate": 2.9620099456458884e-05, "loss": 0.8369, "step": 209780 }, { "epoch": 18.17, "learning_rate": 2.9617931074361047e-05, "loss": 0.8141, "step": 209790 }, { "epoch": 18.17, "learning_rate": 2.961576269226321e-05, "loss": 0.8604, "step": 209800 }, { "epoch": 18.17, "learning_rate": 2.961359431016537e-05, "loss": 0.6399, "step": 209810 }, { "epoch": 18.17, "learning_rate": 2.9611425928067536e-05, "loss": 0.9763, "step": 209820 }, { "epoch": 18.17, "learning_rate": 2.96092575459697e-05, "loss": 0.8577, "step": 209830 }, { "epoch": 18.17, "learning_rate": 2.9607089163871862e-05, "loss": 0.6826, "step": 209840 }, { "epoch": 18.17, "learning_rate": 2.9604920781774022e-05, "loss": 0.7751, "step": 209850 }, { "epoch": 18.18, "learning_rate": 2.9602752399676185e-05, "loss": 0.9491, "step": 209860 }, { "epoch": 18.18, "learning_rate": 2.9600584017578348e-05, "loss": 0.808, "step": 209870 }, { "epoch": 18.18, "learning_rate": 2.9598415635480508e-05, "loss": 0.8103, "step": 209880 }, { "epoch": 18.18, "learning_rate": 2.9596247253382674e-05, "loss": 0.8333, "step": 209890 }, { "epoch": 18.18, "learning_rate": 2.9594078871284837e-05, "loss": 0.8864, "step": 209900 }, { "epoch": 18.18, "learning_rate": 2.9591910489187e-05, "loss": 0.7759, "step": 209910 }, { "epoch": 18.18, "learning_rate": 2.958974210708916e-05, "loss": 0.8487, "step": 209920 }, { "epoch": 18.18, "learning_rate": 2.9587573724991323e-05, "loss": 0.8096, "step": 209930 }, { "epoch": 18.18, "learning_rate": 2.9585405342893486e-05, "loss": 0.9468, "step": 209940 }, { "epoch": 18.18, "learning_rate": 2.9583236960795652e-05, "loss": 0.8373, "step": 209950 }, { "epoch": 18.18, "learning_rate": 2.958106857869781e-05, "loss": 0.6079, "step": 209960 }, { "epoch": 18.18, "learning_rate": 2.9578900196599975e-05, "loss": 0.7526, "step": 209970 }, { "epoch": 18.19, "learning_rate": 2.9576731814502138e-05, "loss": 0.7917, "step": 209980 }, { "epoch": 18.19, "learning_rate": 2.95745634324043e-05, "loss": 0.85, "step": 209990 }, { "epoch": 18.19, "learning_rate": 2.957239505030646e-05, "loss": 0.6863, "step": 210000 }, { "epoch": 18.19, "learning_rate": 2.9570226668208623e-05, "loss": 0.7516, "step": 210010 }, { "epoch": 18.19, "learning_rate": 2.9568058286110786e-05, "loss": 0.8126, "step": 210020 }, { "epoch": 18.19, "learning_rate": 2.9565889904012953e-05, "loss": 0.8454, "step": 210030 }, { "epoch": 18.19, "learning_rate": 2.9563721521915112e-05, "loss": 0.7946, "step": 210040 }, { "epoch": 18.19, "learning_rate": 2.9561553139817275e-05, "loss": 0.7466, "step": 210050 }, { "epoch": 18.19, "learning_rate": 2.955938475771944e-05, "loss": 0.7963, "step": 210060 }, { "epoch": 18.19, "learning_rate": 2.95572163756216e-05, "loss": 0.7082, "step": 210070 }, { "epoch": 18.19, "learning_rate": 2.955504799352376e-05, "loss": 0.8158, "step": 210080 }, { "epoch": 18.2, "learning_rate": 2.9552879611425924e-05, "loss": 0.78, "step": 210090 }, { "epoch": 18.2, "learning_rate": 2.955071122932809e-05, "loss": 0.8079, "step": 210100 }, { "epoch": 18.2, "learning_rate": 2.9548542847230253e-05, "loss": 0.6942, "step": 210110 }, { "epoch": 18.2, "learning_rate": 2.9546374465132413e-05, "loss": 0.9637, "step": 210120 }, { "epoch": 18.2, "learning_rate": 2.9544206083034576e-05, "loss": 0.6603, "step": 210130 }, { "epoch": 18.2, "learning_rate": 2.954203770093674e-05, "loss": 0.8011, "step": 210140 }, { "epoch": 18.2, "learning_rate": 2.95398693188389e-05, "loss": 0.8974, "step": 210150 }, { "epoch": 18.2, "learning_rate": 2.953770093674106e-05, "loss": 0.6209, "step": 210160 }, { "epoch": 18.2, "learning_rate": 2.9535532554643228e-05, "loss": 0.8917, "step": 210170 }, { "epoch": 18.2, "learning_rate": 2.953336417254539e-05, "loss": 0.6502, "step": 210180 }, { "epoch": 18.2, "learning_rate": 2.953119579044755e-05, "loss": 0.8607, "step": 210190 }, { "epoch": 18.2, "learning_rate": 2.9529027408349714e-05, "loss": 0.8084, "step": 210200 }, { "epoch": 18.21, "learning_rate": 2.9526859026251877e-05, "loss": 0.8013, "step": 210210 }, { "epoch": 18.21, "learning_rate": 2.952469064415404e-05, "loss": 0.7461, "step": 210220 }, { "epoch": 18.21, "learning_rate": 2.95225222620562e-05, "loss": 0.8671, "step": 210230 }, { "epoch": 18.21, "learning_rate": 2.9520353879958362e-05, "loss": 0.8551, "step": 210240 }, { "epoch": 18.21, "learning_rate": 2.951818549786053e-05, "loss": 0.798, "step": 210250 }, { "epoch": 18.21, "learning_rate": 2.951601711576269e-05, "loss": 0.8171, "step": 210260 }, { "epoch": 18.21, "learning_rate": 2.951384873366485e-05, "loss": 0.8832, "step": 210270 }, { "epoch": 18.21, "learning_rate": 2.9511680351567014e-05, "loss": 0.7132, "step": 210280 }, { "epoch": 18.21, "learning_rate": 2.9509511969469177e-05, "loss": 0.8336, "step": 210290 }, { "epoch": 18.21, "learning_rate": 2.950734358737134e-05, "loss": 0.6652, "step": 210300 }, { "epoch": 18.21, "learning_rate": 2.95051752052735e-05, "loss": 0.7393, "step": 210310 }, { "epoch": 18.22, "learning_rate": 2.9503006823175666e-05, "loss": 0.7314, "step": 210320 }, { "epoch": 18.22, "learning_rate": 2.950083844107783e-05, "loss": 0.9011, "step": 210330 }, { "epoch": 18.22, "learning_rate": 2.9498670058979992e-05, "loss": 0.6581, "step": 210340 }, { "epoch": 18.22, "learning_rate": 2.9496501676882152e-05, "loss": 0.8193, "step": 210350 }, { "epoch": 18.22, "learning_rate": 2.9494333294784315e-05, "loss": 0.7729, "step": 210360 }, { "epoch": 18.22, "learning_rate": 2.9492164912686478e-05, "loss": 0.8584, "step": 210370 }, { "epoch": 18.22, "learning_rate": 2.9489996530588644e-05, "loss": 0.7734, "step": 210380 }, { "epoch": 18.22, "learning_rate": 2.9487828148490804e-05, "loss": 0.5994, "step": 210390 }, { "epoch": 18.22, "learning_rate": 2.9485659766392967e-05, "loss": 0.7224, "step": 210400 }, { "epoch": 18.22, "learning_rate": 2.948349138429513e-05, "loss": 0.7678, "step": 210410 }, { "epoch": 18.22, "learning_rate": 2.948132300219729e-05, "loss": 0.7223, "step": 210420 }, { "epoch": 18.22, "learning_rate": 2.9479154620099453e-05, "loss": 0.7188, "step": 210430 }, { "epoch": 18.23, "learning_rate": 2.9476986238001616e-05, "loss": 0.8751, "step": 210440 }, { "epoch": 18.23, "learning_rate": 2.9474817855903782e-05, "loss": 0.9587, "step": 210450 }, { "epoch": 18.23, "learning_rate": 2.947264947380594e-05, "loss": 0.783, "step": 210460 }, { "epoch": 18.23, "learning_rate": 2.9470481091708105e-05, "loss": 0.7518, "step": 210470 }, { "epoch": 18.23, "learning_rate": 2.9468312709610268e-05, "loss": 0.8657, "step": 210480 }, { "epoch": 18.23, "learning_rate": 2.946614432751243e-05, "loss": 0.7369, "step": 210490 }, { "epoch": 18.23, "learning_rate": 2.946397594541459e-05, "loss": 0.7652, "step": 210500 }, { "epoch": 18.23, "learning_rate": 2.9461807563316753e-05, "loss": 0.7649, "step": 210510 }, { "epoch": 18.23, "learning_rate": 2.9459639181218916e-05, "loss": 0.706, "step": 210520 }, { "epoch": 18.23, "learning_rate": 2.9457470799121083e-05, "loss": 0.9925, "step": 210530 }, { "epoch": 18.23, "learning_rate": 2.9455302417023242e-05, "loss": 0.8464, "step": 210540 }, { "epoch": 18.23, "learning_rate": 2.9453134034925405e-05, "loss": 0.7945, "step": 210550 }, { "epoch": 18.24, "learning_rate": 2.9450965652827568e-05, "loss": 0.8282, "step": 210560 }, { "epoch": 18.24, "learning_rate": 2.944879727072973e-05, "loss": 0.7154, "step": 210570 }, { "epoch": 18.24, "learning_rate": 2.944662888863189e-05, "loss": 0.736, "step": 210580 }, { "epoch": 18.24, "learning_rate": 2.9444460506534054e-05, "loss": 0.8088, "step": 210590 }, { "epoch": 18.24, "learning_rate": 2.944229212443622e-05, "loss": 0.7997, "step": 210600 }, { "epoch": 18.24, "learning_rate": 2.9440123742338383e-05, "loss": 0.6126, "step": 210610 }, { "epoch": 18.24, "learning_rate": 2.9437955360240543e-05, "loss": 0.8893, "step": 210620 }, { "epoch": 18.24, "learning_rate": 2.9435786978142706e-05, "loss": 0.5466, "step": 210630 }, { "epoch": 18.24, "learning_rate": 2.943361859604487e-05, "loss": 0.9105, "step": 210640 }, { "epoch": 18.24, "learning_rate": 2.943145021394703e-05, "loss": 0.8648, "step": 210650 }, { "epoch": 18.24, "learning_rate": 2.942928183184919e-05, "loss": 0.6909, "step": 210660 }, { "epoch": 18.25, "learning_rate": 2.9427113449751358e-05, "loss": 0.8156, "step": 210670 }, { "epoch": 18.25, "learning_rate": 2.942494506765352e-05, "loss": 0.7193, "step": 210680 }, { "epoch": 18.25, "learning_rate": 2.942277668555568e-05, "loss": 0.7565, "step": 210690 }, { "epoch": 18.25, "learning_rate": 2.9420608303457843e-05, "loss": 0.8048, "step": 210700 }, { "epoch": 18.25, "learning_rate": 2.9418439921360006e-05, "loss": 0.7266, "step": 210710 }, { "epoch": 18.25, "learning_rate": 2.941627153926217e-05, "loss": 0.616, "step": 210720 }, { "epoch": 18.25, "learning_rate": 2.941410315716433e-05, "loss": 0.7322, "step": 210730 }, { "epoch": 18.25, "learning_rate": 2.9411934775066496e-05, "loss": 0.8079, "step": 210740 }, { "epoch": 18.25, "learning_rate": 2.940976639296866e-05, "loss": 0.8362, "step": 210750 }, { "epoch": 18.25, "learning_rate": 2.940759801087082e-05, "loss": 0.7995, "step": 210760 }, { "epoch": 18.25, "learning_rate": 2.940542962877298e-05, "loss": 0.7697, "step": 210770 }, { "epoch": 18.25, "learning_rate": 2.9403261246675144e-05, "loss": 0.705, "step": 210780 }, { "epoch": 18.26, "learning_rate": 2.9401092864577307e-05, "loss": 0.7781, "step": 210790 }, { "epoch": 18.26, "learning_rate": 2.939892448247947e-05, "loss": 0.7933, "step": 210800 }, { "epoch": 18.26, "learning_rate": 2.939675610038163e-05, "loss": 0.9554, "step": 210810 }, { "epoch": 18.26, "learning_rate": 2.9394587718283796e-05, "loss": 0.8413, "step": 210820 }, { "epoch": 18.26, "learning_rate": 2.939241933618596e-05, "loss": 0.8555, "step": 210830 }, { "epoch": 18.26, "learning_rate": 2.9390250954088122e-05, "loss": 0.7622, "step": 210840 }, { "epoch": 18.26, "learning_rate": 2.9388082571990282e-05, "loss": 0.7809, "step": 210850 }, { "epoch": 18.26, "learning_rate": 2.9385914189892445e-05, "loss": 0.7581, "step": 210860 }, { "epoch": 18.26, "learning_rate": 2.9383745807794608e-05, "loss": 0.8403, "step": 210870 }, { "epoch": 18.26, "learning_rate": 2.9381577425696774e-05, "loss": 1.0316, "step": 210880 }, { "epoch": 18.26, "learning_rate": 2.9379409043598934e-05, "loss": 0.7901, "step": 210890 }, { "epoch": 18.27, "learning_rate": 2.9377240661501097e-05, "loss": 0.8671, "step": 210900 }, { "epoch": 18.27, "learning_rate": 2.937507227940326e-05, "loss": 0.7656, "step": 210910 }, { "epoch": 18.27, "learning_rate": 2.937290389730542e-05, "loss": 0.7499, "step": 210920 }, { "epoch": 18.27, "learning_rate": 2.9370735515207582e-05, "loss": 0.6899, "step": 210930 }, { "epoch": 18.27, "learning_rate": 2.9368567133109745e-05, "loss": 0.7713, "step": 210940 }, { "epoch": 18.27, "learning_rate": 2.9366398751011912e-05, "loss": 0.7656, "step": 210950 }, { "epoch": 18.27, "learning_rate": 2.936423036891407e-05, "loss": 0.7841, "step": 210960 }, { "epoch": 18.27, "learning_rate": 2.9362061986816234e-05, "loss": 0.7104, "step": 210970 }, { "epoch": 18.27, "learning_rate": 2.9359893604718397e-05, "loss": 0.6752, "step": 210980 }, { "epoch": 18.27, "learning_rate": 2.935772522262056e-05, "loss": 0.8978, "step": 210990 }, { "epoch": 18.27, "learning_rate": 2.935555684052272e-05, "loss": 0.8664, "step": 211000 }, { "epoch": 18.27, "learning_rate": 2.9353388458424883e-05, "loss": 0.7921, "step": 211010 }, { "epoch": 18.28, "learning_rate": 2.935122007632705e-05, "loss": 0.7004, "step": 211020 }, { "epoch": 18.28, "learning_rate": 2.9349051694229212e-05, "loss": 0.8111, "step": 211030 }, { "epoch": 18.28, "learning_rate": 2.9346883312131372e-05, "loss": 0.6867, "step": 211040 }, { "epoch": 18.28, "learning_rate": 2.9344714930033535e-05, "loss": 0.7414, "step": 211050 }, { "epoch": 18.28, "learning_rate": 2.9342546547935698e-05, "loss": 0.8276, "step": 211060 }, { "epoch": 18.28, "learning_rate": 2.934037816583786e-05, "loss": 0.6533, "step": 211070 }, { "epoch": 18.28, "learning_rate": 2.933820978374002e-05, "loss": 0.8466, "step": 211080 }, { "epoch": 18.28, "learning_rate": 2.9336041401642184e-05, "loss": 0.6576, "step": 211090 }, { "epoch": 18.28, "learning_rate": 2.933387301954435e-05, "loss": 0.8656, "step": 211100 }, { "epoch": 18.28, "learning_rate": 2.9331704637446513e-05, "loss": 0.819, "step": 211110 }, { "epoch": 18.28, "learning_rate": 2.9329536255348673e-05, "loss": 0.8365, "step": 211120 }, { "epoch": 18.29, "learning_rate": 2.9327367873250836e-05, "loss": 0.8852, "step": 211130 }, { "epoch": 18.29, "learning_rate": 2.9325199491153e-05, "loss": 0.8911, "step": 211140 }, { "epoch": 18.29, "learning_rate": 2.9323031109055162e-05, "loss": 0.6652, "step": 211150 }, { "epoch": 18.29, "learning_rate": 2.932086272695732e-05, "loss": 0.6243, "step": 211160 }, { "epoch": 18.29, "learning_rate": 2.9318694344859488e-05, "loss": 0.9017, "step": 211170 }, { "epoch": 18.29, "learning_rate": 2.931652596276165e-05, "loss": 0.8902, "step": 211180 }, { "epoch": 18.29, "learning_rate": 2.931435758066381e-05, "loss": 0.8278, "step": 211190 }, { "epoch": 18.29, "learning_rate": 2.9312189198565973e-05, "loss": 0.803, "step": 211200 }, { "epoch": 18.29, "learning_rate": 2.9310020816468136e-05, "loss": 0.7934, "step": 211210 }, { "epoch": 18.29, "learning_rate": 2.93078524343703e-05, "loss": 0.7674, "step": 211220 }, { "epoch": 18.29, "learning_rate": 2.930568405227246e-05, "loss": 0.6648, "step": 211230 }, { "epoch": 18.29, "learning_rate": 2.9303515670174625e-05, "loss": 0.6364, "step": 211240 }, { "epoch": 18.3, "learning_rate": 2.930134728807679e-05, "loss": 0.783, "step": 211250 }, { "epoch": 18.3, "learning_rate": 2.929917890597895e-05, "loss": 0.7264, "step": 211260 }, { "epoch": 18.3, "learning_rate": 2.929701052388111e-05, "loss": 0.6563, "step": 211270 }, { "epoch": 18.3, "learning_rate": 2.9294842141783274e-05, "loss": 0.8114, "step": 211280 }, { "epoch": 18.3, "learning_rate": 2.9292673759685437e-05, "loss": 0.8459, "step": 211290 }, { "epoch": 18.3, "learning_rate": 2.9290505377587603e-05, "loss": 0.881, "step": 211300 }, { "epoch": 18.3, "learning_rate": 2.928833699548976e-05, "loss": 0.8817, "step": 211310 }, { "epoch": 18.3, "learning_rate": 2.9286168613391926e-05, "loss": 0.9928, "step": 211320 }, { "epoch": 18.3, "learning_rate": 2.928400023129409e-05, "loss": 0.7944, "step": 211330 }, { "epoch": 18.3, "learning_rate": 2.9281831849196252e-05, "loss": 0.7916, "step": 211340 }, { "epoch": 18.3, "learning_rate": 2.927966346709841e-05, "loss": 0.7791, "step": 211350 }, { "epoch": 18.31, "learning_rate": 2.9277495085000575e-05, "loss": 0.6849, "step": 211360 }, { "epoch": 18.31, "learning_rate": 2.9275326702902738e-05, "loss": 0.7197, "step": 211370 }, { "epoch": 18.31, "learning_rate": 2.9273158320804904e-05, "loss": 0.7904, "step": 211380 }, { "epoch": 18.31, "learning_rate": 2.9270989938707064e-05, "loss": 0.7506, "step": 211390 }, { "epoch": 18.31, "learning_rate": 2.9268821556609227e-05, "loss": 0.9204, "step": 211400 }, { "epoch": 18.31, "learning_rate": 2.926665317451139e-05, "loss": 0.6527, "step": 211410 }, { "epoch": 18.31, "learning_rate": 2.926448479241355e-05, "loss": 0.8291, "step": 211420 }, { "epoch": 18.31, "learning_rate": 2.9262316410315712e-05, "loss": 0.8786, "step": 211430 }, { "epoch": 18.31, "learning_rate": 2.9260148028217875e-05, "loss": 0.8283, "step": 211440 }, { "epoch": 18.31, "learning_rate": 2.925797964612004e-05, "loss": 0.9184, "step": 211450 }, { "epoch": 18.31, "learning_rate": 2.92558112640222e-05, "loss": 0.7831, "step": 211460 }, { "epoch": 18.31, "learning_rate": 2.9253642881924364e-05, "loss": 0.803, "step": 211470 }, { "epoch": 18.32, "learning_rate": 2.9251474499826527e-05, "loss": 0.8899, "step": 211480 }, { "epoch": 18.32, "learning_rate": 2.924930611772869e-05, "loss": 0.7123, "step": 211490 }, { "epoch": 18.32, "learning_rate": 2.924713773563085e-05, "loss": 0.7742, "step": 211500 }, { "epoch": 18.32, "learning_rate": 2.9244969353533013e-05, "loss": 0.656, "step": 211510 }, { "epoch": 18.32, "learning_rate": 2.924280097143518e-05, "loss": 0.6973, "step": 211520 }, { "epoch": 18.32, "learning_rate": 2.9240632589337342e-05, "loss": 0.8695, "step": 211530 }, { "epoch": 18.32, "learning_rate": 2.9238464207239502e-05, "loss": 0.6881, "step": 211540 }, { "epoch": 18.32, "learning_rate": 2.9236295825141665e-05, "loss": 0.7679, "step": 211550 }, { "epoch": 18.32, "learning_rate": 2.9234127443043828e-05, "loss": 0.7242, "step": 211560 }, { "epoch": 18.32, "learning_rate": 2.923195906094599e-05, "loss": 0.8006, "step": 211570 }, { "epoch": 18.32, "learning_rate": 2.922979067884815e-05, "loss": 0.7352, "step": 211580 }, { "epoch": 18.33, "learning_rate": 2.9227622296750314e-05, "loss": 0.7295, "step": 211590 }, { "epoch": 18.33, "learning_rate": 2.922545391465248e-05, "loss": 0.6842, "step": 211600 }, { "epoch": 18.33, "learning_rate": 2.9223285532554643e-05, "loss": 0.7875, "step": 211610 }, { "epoch": 18.33, "learning_rate": 2.9221117150456803e-05, "loss": 0.7198, "step": 211620 }, { "epoch": 18.33, "learning_rate": 2.9218948768358966e-05, "loss": 0.699, "step": 211630 }, { "epoch": 18.33, "learning_rate": 2.921678038626113e-05, "loss": 0.8036, "step": 211640 }, { "epoch": 18.33, "learning_rate": 2.921461200416329e-05, "loss": 0.7911, "step": 211650 }, { "epoch": 18.33, "learning_rate": 2.921244362206545e-05, "loss": 0.8302, "step": 211660 }, { "epoch": 18.33, "learning_rate": 2.9210275239967618e-05, "loss": 0.6243, "step": 211670 }, { "epoch": 18.33, "learning_rate": 2.920810685786978e-05, "loss": 0.7749, "step": 211680 }, { "epoch": 18.33, "learning_rate": 2.920593847577194e-05, "loss": 0.6202, "step": 211690 }, { "epoch": 18.33, "learning_rate": 2.9203770093674103e-05, "loss": 0.8459, "step": 211700 }, { "epoch": 18.34, "learning_rate": 2.9201601711576266e-05, "loss": 0.7827, "step": 211710 }, { "epoch": 18.34, "learning_rate": 2.919943332947843e-05, "loss": 0.7667, "step": 211720 }, { "epoch": 18.34, "learning_rate": 2.919726494738059e-05, "loss": 0.8467, "step": 211730 }, { "epoch": 18.34, "learning_rate": 2.9195096565282755e-05, "loss": 0.7888, "step": 211740 }, { "epoch": 18.34, "learning_rate": 2.9192928183184918e-05, "loss": 0.6961, "step": 211750 }, { "epoch": 18.34, "learning_rate": 2.919075980108708e-05, "loss": 0.7986, "step": 211760 }, { "epoch": 18.34, "learning_rate": 2.918859141898924e-05, "loss": 1.0105, "step": 211770 }, { "epoch": 18.34, "learning_rate": 2.9186423036891404e-05, "loss": 0.7338, "step": 211780 }, { "epoch": 18.34, "learning_rate": 2.9184254654793567e-05, "loss": 0.7027, "step": 211790 }, { "epoch": 18.34, "learning_rate": 2.9182086272695733e-05, "loss": 0.771, "step": 211800 }, { "epoch": 18.34, "learning_rate": 2.9179917890597893e-05, "loss": 0.7396, "step": 211810 }, { "epoch": 18.34, "learning_rate": 2.9177749508500056e-05, "loss": 0.8269, "step": 211820 }, { "epoch": 18.35, "learning_rate": 2.917558112640222e-05, "loss": 0.7008, "step": 211830 }, { "epoch": 18.35, "learning_rate": 2.9173412744304382e-05, "loss": 0.9339, "step": 211840 }, { "epoch": 18.35, "learning_rate": 2.917124436220654e-05, "loss": 0.7758, "step": 211850 }, { "epoch": 18.35, "learning_rate": 2.9169075980108704e-05, "loss": 0.7807, "step": 211860 }, { "epoch": 18.35, "learning_rate": 2.9166907598010867e-05, "loss": 0.9136, "step": 211870 }, { "epoch": 18.35, "learning_rate": 2.9164739215913034e-05, "loss": 0.8097, "step": 211880 }, { "epoch": 18.35, "learning_rate": 2.9162570833815193e-05, "loss": 0.7899, "step": 211890 }, { "epoch": 18.35, "learning_rate": 2.9160402451717356e-05, "loss": 0.9116, "step": 211900 }, { "epoch": 18.35, "learning_rate": 2.915823406961952e-05, "loss": 0.6873, "step": 211910 }, { "epoch": 18.35, "learning_rate": 2.9156065687521682e-05, "loss": 0.7533, "step": 211920 }, { "epoch": 18.35, "learning_rate": 2.9153897305423842e-05, "loss": 0.7964, "step": 211930 }, { "epoch": 18.36, "learning_rate": 2.9151728923326005e-05, "loss": 0.8232, "step": 211940 }, { "epoch": 18.36, "learning_rate": 2.914956054122817e-05, "loss": 0.8281, "step": 211950 }, { "epoch": 18.36, "learning_rate": 2.914739215913033e-05, "loss": 0.7873, "step": 211960 }, { "epoch": 18.36, "learning_rate": 2.9145223777032494e-05, "loss": 0.6379, "step": 211970 }, { "epoch": 18.36, "learning_rate": 2.9143055394934657e-05, "loss": 0.6469, "step": 211980 }, { "epoch": 18.36, "learning_rate": 2.914088701283682e-05, "loss": 0.8132, "step": 211990 }, { "epoch": 18.36, "learning_rate": 2.913871863073898e-05, "loss": 0.689, "step": 212000 }, { "epoch": 18.36, "learning_rate": 2.9136550248641143e-05, "loss": 0.7721, "step": 212010 }, { "epoch": 18.36, "learning_rate": 2.913438186654331e-05, "loss": 0.7108, "step": 212020 }, { "epoch": 18.36, "learning_rate": 2.9132213484445472e-05, "loss": 0.7764, "step": 212030 }, { "epoch": 18.36, "learning_rate": 2.9130045102347632e-05, "loss": 0.685, "step": 212040 }, { "epoch": 18.36, "learning_rate": 2.9127876720249795e-05, "loss": 0.825, "step": 212050 }, { "epoch": 18.37, "learning_rate": 2.9125708338151958e-05, "loss": 0.8419, "step": 212060 }, { "epoch": 18.37, "learning_rate": 2.912353995605412e-05, "loss": 0.7554, "step": 212070 }, { "epoch": 18.37, "learning_rate": 2.912137157395628e-05, "loss": 0.7858, "step": 212080 }, { "epoch": 18.37, "learning_rate": 2.9119203191858447e-05, "loss": 0.7769, "step": 212090 }, { "epoch": 18.37, "learning_rate": 2.911703480976061e-05, "loss": 0.7116, "step": 212100 }, { "epoch": 18.37, "learning_rate": 2.9114866427662773e-05, "loss": 0.7255, "step": 212110 }, { "epoch": 18.37, "learning_rate": 2.9112698045564932e-05, "loss": 0.8627, "step": 212120 }, { "epoch": 18.37, "learning_rate": 2.9110529663467095e-05, "loss": 0.8782, "step": 212130 }, { "epoch": 18.37, "learning_rate": 2.910836128136926e-05, "loss": 0.831, "step": 212140 }, { "epoch": 18.37, "learning_rate": 2.910619289927142e-05, "loss": 0.6462, "step": 212150 }, { "epoch": 18.37, "learning_rate": 2.910402451717358e-05, "loss": 0.8751, "step": 212160 }, { "epoch": 18.38, "learning_rate": 2.9101856135075747e-05, "loss": 0.7169, "step": 212170 }, { "epoch": 18.38, "learning_rate": 2.909968775297791e-05, "loss": 0.762, "step": 212180 }, { "epoch": 18.38, "learning_rate": 2.909751937088007e-05, "loss": 0.8078, "step": 212190 }, { "epoch": 18.38, "learning_rate": 2.9095350988782233e-05, "loss": 0.8527, "step": 212200 }, { "epoch": 18.38, "learning_rate": 2.9093182606684396e-05, "loss": 0.7613, "step": 212210 }, { "epoch": 18.38, "learning_rate": 2.909101422458656e-05, "loss": 0.8766, "step": 212220 }, { "epoch": 18.38, "learning_rate": 2.908884584248872e-05, "loss": 0.6496, "step": 212230 }, { "epoch": 18.38, "learning_rate": 2.9086677460390885e-05, "loss": 0.6044, "step": 212240 }, { "epoch": 18.38, "learning_rate": 2.9084509078293048e-05, "loss": 0.7357, "step": 212250 }, { "epoch": 18.38, "learning_rate": 2.908234069619521e-05, "loss": 0.728, "step": 212260 }, { "epoch": 18.38, "learning_rate": 2.908017231409737e-05, "loss": 0.8682, "step": 212270 }, { "epoch": 18.38, "learning_rate": 2.9078003931999534e-05, "loss": 0.6873, "step": 212280 }, { "epoch": 18.39, "learning_rate": 2.9075835549901697e-05, "loss": 0.8257, "step": 212290 }, { "epoch": 18.39, "learning_rate": 2.9073667167803863e-05, "loss": 0.7484, "step": 212300 }, { "epoch": 18.39, "learning_rate": 2.9071498785706023e-05, "loss": 0.8096, "step": 212310 }, { "epoch": 18.39, "learning_rate": 2.9069330403608186e-05, "loss": 0.7075, "step": 212320 }, { "epoch": 18.39, "learning_rate": 2.906716202151035e-05, "loss": 0.7732, "step": 212330 }, { "epoch": 18.39, "learning_rate": 2.906499363941251e-05, "loss": 0.7569, "step": 212340 }, { "epoch": 18.39, "learning_rate": 2.906282525731467e-05, "loss": 0.741, "step": 212350 }, { "epoch": 18.39, "learning_rate": 2.9060656875216834e-05, "loss": 0.7724, "step": 212360 }, { "epoch": 18.39, "learning_rate": 2.9058488493119e-05, "loss": 0.8757, "step": 212370 }, { "epoch": 18.39, "learning_rate": 2.9056320111021164e-05, "loss": 0.7595, "step": 212380 }, { "epoch": 18.39, "learning_rate": 2.9054151728923323e-05, "loss": 0.7645, "step": 212390 }, { "epoch": 18.4, "learning_rate": 2.9051983346825486e-05, "loss": 0.7979, "step": 212400 }, { "epoch": 18.4, "learning_rate": 2.904981496472765e-05, "loss": 0.8477, "step": 212410 }, { "epoch": 18.4, "learning_rate": 2.9047646582629812e-05, "loss": 0.8029, "step": 212420 }, { "epoch": 18.4, "learning_rate": 2.9045478200531972e-05, "loss": 0.7968, "step": 212430 }, { "epoch": 18.4, "learning_rate": 2.9043309818434135e-05, "loss": 0.7363, "step": 212440 }, { "epoch": 18.4, "learning_rate": 2.90411414363363e-05, "loss": 0.7424, "step": 212450 }, { "epoch": 18.4, "learning_rate": 2.903897305423846e-05, "loss": 0.9896, "step": 212460 }, { "epoch": 18.4, "learning_rate": 2.9036804672140624e-05, "loss": 0.7361, "step": 212470 }, { "epoch": 18.4, "learning_rate": 2.9034636290042787e-05, "loss": 0.9053, "step": 212480 }, { "epoch": 18.4, "learning_rate": 2.903246790794495e-05, "loss": 0.7733, "step": 212490 }, { "epoch": 18.4, "learning_rate": 2.903029952584711e-05, "loss": 0.7351, "step": 212500 }, { "epoch": 18.4, "learning_rate": 2.9028131143749273e-05, "loss": 0.8239, "step": 212510 }, { "epoch": 18.41, "learning_rate": 2.902596276165144e-05, "loss": 0.8739, "step": 212520 }, { "epoch": 18.41, "learning_rate": 2.9023794379553602e-05, "loss": 0.7809, "step": 212530 }, { "epoch": 18.41, "learning_rate": 2.902162599745576e-05, "loss": 0.8005, "step": 212540 }, { "epoch": 18.41, "learning_rate": 2.9019457615357925e-05, "loss": 0.7065, "step": 212550 }, { "epoch": 18.41, "learning_rate": 2.9017289233260088e-05, "loss": 0.8333, "step": 212560 }, { "epoch": 18.41, "learning_rate": 2.901512085116225e-05, "loss": 0.7944, "step": 212570 }, { "epoch": 18.41, "learning_rate": 2.901295246906441e-05, "loss": 0.8275, "step": 212580 }, { "epoch": 18.41, "learning_rate": 2.9010784086966577e-05, "loss": 0.6855, "step": 212590 }, { "epoch": 18.41, "learning_rate": 2.900861570486874e-05, "loss": 0.7401, "step": 212600 }, { "epoch": 18.41, "learning_rate": 2.9006447322770903e-05, "loss": 0.7845, "step": 212610 }, { "epoch": 18.41, "learning_rate": 2.9004278940673062e-05, "loss": 0.7806, "step": 212620 }, { "epoch": 18.42, "learning_rate": 2.9002110558575225e-05, "loss": 0.7405, "step": 212630 }, { "epoch": 18.42, "learning_rate": 2.8999942176477388e-05, "loss": 0.9039, "step": 212640 }, { "epoch": 18.42, "learning_rate": 2.8997773794379555e-05, "loss": 0.796, "step": 212650 }, { "epoch": 18.42, "learning_rate": 2.899560541228171e-05, "loss": 0.7988, "step": 212660 }, { "epoch": 18.42, "learning_rate": 2.8993437030183877e-05, "loss": 0.6932, "step": 212670 }, { "epoch": 18.42, "learning_rate": 2.899126864808604e-05, "loss": 0.6592, "step": 212680 }, { "epoch": 18.42, "learning_rate": 2.8989100265988203e-05, "loss": 1.1017, "step": 212690 }, { "epoch": 18.42, "learning_rate": 2.8986931883890363e-05, "loss": 0.7962, "step": 212700 }, { "epoch": 18.42, "learning_rate": 2.8984763501792526e-05, "loss": 0.797, "step": 212710 }, { "epoch": 18.42, "learning_rate": 2.898259511969469e-05, "loss": 0.7834, "step": 212720 }, { "epoch": 18.42, "learning_rate": 2.898042673759685e-05, "loss": 0.7695, "step": 212730 }, { "epoch": 18.42, "learning_rate": 2.8978258355499015e-05, "loss": 0.7611, "step": 212740 }, { "epoch": 18.43, "learning_rate": 2.8976089973401178e-05, "loss": 0.8368, "step": 212750 }, { "epoch": 18.43, "learning_rate": 2.897392159130334e-05, "loss": 0.8255, "step": 212760 }, { "epoch": 18.43, "learning_rate": 2.89717532092055e-05, "loss": 0.73, "step": 212770 }, { "epoch": 18.43, "learning_rate": 2.8969584827107664e-05, "loss": 0.7362, "step": 212780 }, { "epoch": 18.43, "learning_rate": 2.8967416445009827e-05, "loss": 0.6334, "step": 212790 }, { "epoch": 18.43, "learning_rate": 2.8965248062911993e-05, "loss": 0.7621, "step": 212800 }, { "epoch": 18.43, "learning_rate": 2.8963079680814153e-05, "loss": 0.693, "step": 212810 }, { "epoch": 18.43, "learning_rate": 2.8960911298716316e-05, "loss": 0.7831, "step": 212820 }, { "epoch": 18.43, "learning_rate": 2.895874291661848e-05, "loss": 0.7307, "step": 212830 }, { "epoch": 18.43, "learning_rate": 2.895657453452064e-05, "loss": 0.6823, "step": 212840 }, { "epoch": 18.43, "learning_rate": 2.89544061524228e-05, "loss": 0.9725, "step": 212850 }, { "epoch": 18.44, "learning_rate": 2.8952237770324964e-05, "loss": 0.7055, "step": 212860 }, { "epoch": 18.44, "learning_rate": 2.895006938822713e-05, "loss": 0.818, "step": 212870 }, { "epoch": 18.44, "learning_rate": 2.8947901006129294e-05, "loss": 0.9272, "step": 212880 }, { "epoch": 18.44, "learning_rate": 2.8945732624031453e-05, "loss": 0.7657, "step": 212890 }, { "epoch": 18.44, "learning_rate": 2.8943564241933616e-05, "loss": 0.8619, "step": 212900 }, { "epoch": 18.44, "learning_rate": 2.894139585983578e-05, "loss": 0.7452, "step": 212910 }, { "epoch": 18.44, "learning_rate": 2.8939227477737942e-05, "loss": 0.8807, "step": 212920 }, { "epoch": 18.44, "learning_rate": 2.8937059095640102e-05, "loss": 0.6978, "step": 212930 }, { "epoch": 18.44, "learning_rate": 2.8934890713542265e-05, "loss": 0.6777, "step": 212940 }, { "epoch": 18.44, "learning_rate": 2.893272233144443e-05, "loss": 0.7537, "step": 212950 }, { "epoch": 18.44, "learning_rate": 2.893055394934659e-05, "loss": 0.7753, "step": 212960 }, { "epoch": 18.44, "learning_rate": 2.8928385567248754e-05, "loss": 0.8739, "step": 212970 }, { "epoch": 18.45, "learning_rate": 2.8926217185150917e-05, "loss": 0.7635, "step": 212980 }, { "epoch": 18.45, "learning_rate": 2.892404880305308e-05, "loss": 0.8707, "step": 212990 }, { "epoch": 18.45, "learning_rate": 2.892188042095524e-05, "loss": 0.8748, "step": 213000 }, { "epoch": 18.45, "learning_rate": 2.8919712038857402e-05, "loss": 0.9052, "step": 213010 }, { "epoch": 18.45, "learning_rate": 2.891754365675957e-05, "loss": 0.6935, "step": 213020 }, { "epoch": 18.45, "learning_rate": 2.8915375274661732e-05, "loss": 0.7742, "step": 213030 }, { "epoch": 18.45, "learning_rate": 2.891320689256389e-05, "loss": 0.7678, "step": 213040 }, { "epoch": 18.45, "learning_rate": 2.8911038510466054e-05, "loss": 0.8485, "step": 213050 }, { "epoch": 18.45, "learning_rate": 2.8908870128368217e-05, "loss": 0.6791, "step": 213060 }, { "epoch": 18.45, "learning_rate": 2.890670174627038e-05, "loss": 0.5476, "step": 213070 }, { "epoch": 18.45, "learning_rate": 2.890453336417254e-05, "loss": 0.8652, "step": 213080 }, { "epoch": 18.45, "learning_rate": 2.8902364982074706e-05, "loss": 0.7993, "step": 213090 }, { "epoch": 18.46, "learning_rate": 2.890019659997687e-05, "loss": 0.7862, "step": 213100 }, { "epoch": 18.46, "learning_rate": 2.8898028217879032e-05, "loss": 0.8394, "step": 213110 }, { "epoch": 18.46, "learning_rate": 2.8895859835781192e-05, "loss": 0.931, "step": 213120 }, { "epoch": 18.46, "learning_rate": 2.8893691453683355e-05, "loss": 0.72, "step": 213130 }, { "epoch": 18.46, "learning_rate": 2.8891523071585518e-05, "loss": 0.8178, "step": 213140 }, { "epoch": 18.46, "learning_rate": 2.8889354689487684e-05, "loss": 0.7661, "step": 213150 }, { "epoch": 18.46, "learning_rate": 2.8887186307389844e-05, "loss": 0.7882, "step": 213160 }, { "epoch": 18.46, "learning_rate": 2.8885017925292007e-05, "loss": 0.6937, "step": 213170 }, { "epoch": 18.46, "learning_rate": 2.888284954319417e-05, "loss": 0.7661, "step": 213180 }, { "epoch": 18.46, "learning_rate": 2.8880681161096333e-05, "loss": 0.8219, "step": 213190 }, { "epoch": 18.46, "learning_rate": 2.8878512778998493e-05, "loss": 0.8309, "step": 213200 }, { "epoch": 18.47, "learning_rate": 2.8876344396900656e-05, "loss": 0.7797, "step": 213210 }, { "epoch": 18.47, "learning_rate": 2.887417601480282e-05, "loss": 0.7994, "step": 213220 }, { "epoch": 18.47, "learning_rate": 2.887200763270498e-05, "loss": 0.9474, "step": 213230 }, { "epoch": 18.47, "learning_rate": 2.8869839250607145e-05, "loss": 0.6759, "step": 213240 }, { "epoch": 18.47, "learning_rate": 2.8867670868509308e-05, "loss": 0.8192, "step": 213250 }, { "epoch": 18.47, "learning_rate": 2.886550248641147e-05, "loss": 0.7714, "step": 213260 }, { "epoch": 18.47, "learning_rate": 2.886333410431363e-05, "loss": 0.7589, "step": 213270 }, { "epoch": 18.47, "learning_rate": 2.8861165722215793e-05, "loss": 0.8015, "step": 213280 }, { "epoch": 18.47, "learning_rate": 2.8858997340117956e-05, "loss": 0.8313, "step": 213290 }, { "epoch": 18.47, "learning_rate": 2.8856828958020123e-05, "loss": 0.7728, "step": 213300 }, { "epoch": 18.47, "learning_rate": 2.8854660575922282e-05, "loss": 0.9733, "step": 213310 }, { "epoch": 18.47, "learning_rate": 2.8852492193824445e-05, "loss": 0.8187, "step": 213320 }, { "epoch": 18.48, "learning_rate": 2.885032381172661e-05, "loss": 0.72, "step": 213330 }, { "epoch": 18.48, "learning_rate": 2.884815542962877e-05, "loss": 0.6521, "step": 213340 }, { "epoch": 18.48, "learning_rate": 2.884598704753093e-05, "loss": 0.8824, "step": 213350 }, { "epoch": 18.48, "learning_rate": 2.8843818665433094e-05, "loss": 0.8069, "step": 213360 }, { "epoch": 18.48, "learning_rate": 2.884165028333526e-05, "loss": 0.7708, "step": 213370 }, { "epoch": 18.48, "learning_rate": 2.8839481901237423e-05, "loss": 0.6825, "step": 213380 }, { "epoch": 18.48, "learning_rate": 2.8837313519139583e-05, "loss": 0.8073, "step": 213390 }, { "epoch": 18.48, "learning_rate": 2.8835145137041746e-05, "loss": 0.6639, "step": 213400 }, { "epoch": 18.48, "learning_rate": 2.883297675494391e-05, "loss": 0.6506, "step": 213410 }, { "epoch": 18.48, "learning_rate": 2.8830808372846072e-05, "loss": 0.84, "step": 213420 }, { "epoch": 18.48, "learning_rate": 2.882863999074823e-05, "loss": 0.8481, "step": 213430 }, { "epoch": 18.49, "learning_rate": 2.8826471608650398e-05, "loss": 0.6941, "step": 213440 }, { "epoch": 18.49, "learning_rate": 2.882430322655256e-05, "loss": 0.7591, "step": 213450 }, { "epoch": 18.49, "learning_rate": 2.8822134844454724e-05, "loss": 0.8432, "step": 213460 }, { "epoch": 18.49, "learning_rate": 2.8819966462356884e-05, "loss": 0.6738, "step": 213470 }, { "epoch": 18.49, "learning_rate": 2.8817798080259047e-05, "loss": 0.7754, "step": 213480 }, { "epoch": 18.49, "learning_rate": 2.881562969816121e-05, "loss": 0.8911, "step": 213490 }, { "epoch": 18.49, "learning_rate": 2.881346131606337e-05, "loss": 0.8239, "step": 213500 }, { "epoch": 18.49, "learning_rate": 2.8811292933965532e-05, "loss": 0.8498, "step": 213510 }, { "epoch": 18.49, "learning_rate": 2.88091245518677e-05, "loss": 0.7721, "step": 213520 }, { "epoch": 18.49, "learning_rate": 2.880695616976986e-05, "loss": 0.8423, "step": 213530 }, { "epoch": 18.49, "learning_rate": 2.880478778767202e-05, "loss": 0.8532, "step": 213540 }, { "epoch": 18.49, "learning_rate": 2.8802619405574184e-05, "loss": 0.7178, "step": 213550 }, { "epoch": 18.5, "learning_rate": 2.8800451023476347e-05, "loss": 0.9059, "step": 213560 }, { "epoch": 18.5, "learning_rate": 2.879828264137851e-05, "loss": 0.7805, "step": 213570 }, { "epoch": 18.5, "learning_rate": 2.879611425928067e-05, "loss": 0.6985, "step": 213580 }, { "epoch": 18.5, "learning_rate": 2.8793945877182836e-05, "loss": 0.9692, "step": 213590 }, { "epoch": 18.5, "learning_rate": 2.8791777495085e-05, "loss": 0.856, "step": 213600 }, { "epoch": 18.5, "learning_rate": 2.8789609112987162e-05, "loss": 0.7668, "step": 213610 }, { "epoch": 18.5, "learning_rate": 2.8787440730889322e-05, "loss": 0.7366, "step": 213620 }, { "epoch": 18.5, "learning_rate": 2.8785272348791485e-05, "loss": 0.7604, "step": 213630 }, { "epoch": 18.5, "learning_rate": 2.8783103966693648e-05, "loss": 0.9549, "step": 213640 }, { "epoch": 18.5, "learning_rate": 2.8780935584595814e-05, "loss": 0.7164, "step": 213650 }, { "epoch": 18.5, "learning_rate": 2.8778767202497974e-05, "loss": 0.7575, "step": 213660 }, { "epoch": 18.51, "learning_rate": 2.8776598820400137e-05, "loss": 0.9199, "step": 213670 }, { "epoch": 18.51, "learning_rate": 2.87744304383023e-05, "loss": 0.7463, "step": 213680 }, { "epoch": 18.51, "learning_rate": 2.8772262056204463e-05, "loss": 0.7952, "step": 213690 }, { "epoch": 18.51, "learning_rate": 2.8770093674106623e-05, "loss": 0.867, "step": 213700 }, { "epoch": 18.51, "learning_rate": 2.8767925292008786e-05, "loss": 0.7502, "step": 213710 }, { "epoch": 18.51, "learning_rate": 2.8765756909910952e-05, "loss": 0.8154, "step": 213720 }, { "epoch": 18.51, "learning_rate": 2.8763588527813108e-05, "loss": 0.9622, "step": 213730 }, { "epoch": 18.51, "learning_rate": 2.8761420145715275e-05, "loss": 0.8203, "step": 213740 }, { "epoch": 18.51, "learning_rate": 2.8759251763617438e-05, "loss": 0.616, "step": 213750 }, { "epoch": 18.51, "learning_rate": 2.87570833815196e-05, "loss": 0.9198, "step": 213760 }, { "epoch": 18.51, "learning_rate": 2.875491499942176e-05, "loss": 0.6968, "step": 213770 }, { "epoch": 18.51, "learning_rate": 2.8752746617323923e-05, "loss": 0.934, "step": 213780 }, { "epoch": 18.52, "learning_rate": 2.8750578235226086e-05, "loss": 0.6939, "step": 213790 }, { "epoch": 18.52, "learning_rate": 2.8748409853128253e-05, "loss": 0.6829, "step": 213800 }, { "epoch": 18.52, "learning_rate": 2.8746241471030412e-05, "loss": 0.7547, "step": 213810 }, { "epoch": 18.52, "learning_rate": 2.8744073088932575e-05, "loss": 0.8657, "step": 213820 }, { "epoch": 18.52, "learning_rate": 2.8741904706834738e-05, "loss": 0.9615, "step": 213830 }, { "epoch": 18.52, "learning_rate": 2.87397363247369e-05, "loss": 0.8395, "step": 213840 }, { "epoch": 18.52, "learning_rate": 2.873756794263906e-05, "loss": 0.7367, "step": 213850 }, { "epoch": 18.52, "learning_rate": 2.8735399560541224e-05, "loss": 0.8639, "step": 213860 }, { "epoch": 18.52, "learning_rate": 2.873323117844339e-05, "loss": 0.9723, "step": 213870 }, { "epoch": 18.52, "learning_rate": 2.8731062796345553e-05, "loss": 0.8089, "step": 213880 }, { "epoch": 18.52, "learning_rate": 2.8728894414247713e-05, "loss": 0.879, "step": 213890 }, { "epoch": 18.53, "learning_rate": 2.8726726032149876e-05, "loss": 0.7718, "step": 213900 }, { "epoch": 18.53, "learning_rate": 2.872455765005204e-05, "loss": 0.7365, "step": 213910 }, { "epoch": 18.53, "learning_rate": 2.8722389267954202e-05, "loss": 0.8827, "step": 213920 }, { "epoch": 18.53, "learning_rate": 2.872022088585636e-05, "loss": 0.7492, "step": 213930 }, { "epoch": 18.53, "learning_rate": 2.8718052503758528e-05, "loss": 0.6578, "step": 213940 }, { "epoch": 18.53, "learning_rate": 2.871588412166069e-05, "loss": 0.8113, "step": 213950 }, { "epoch": 18.53, "learning_rate": 2.8713715739562854e-05, "loss": 0.8276, "step": 213960 }, { "epoch": 18.53, "learning_rate": 2.8711547357465014e-05, "loss": 0.7928, "step": 213970 }, { "epoch": 18.53, "learning_rate": 2.8709378975367177e-05, "loss": 0.7832, "step": 213980 }, { "epoch": 18.53, "learning_rate": 2.870721059326934e-05, "loss": 0.6876, "step": 213990 }, { "epoch": 18.53, "learning_rate": 2.87050422111715e-05, "loss": 0.9127, "step": 214000 }, { "epoch": 18.53, "learning_rate": 2.8702873829073662e-05, "loss": 0.9042, "step": 214010 }, { "epoch": 18.54, "learning_rate": 2.870070544697583e-05, "loss": 0.8299, "step": 214020 }, { "epoch": 18.54, "learning_rate": 2.869853706487799e-05, "loss": 0.7757, "step": 214030 }, { "epoch": 18.54, "learning_rate": 2.869636868278015e-05, "loss": 0.8007, "step": 214040 }, { "epoch": 18.54, "learning_rate": 2.8694200300682314e-05, "loss": 0.7589, "step": 214050 }, { "epoch": 18.54, "learning_rate": 2.8692031918584477e-05, "loss": 0.7297, "step": 214060 }, { "epoch": 18.54, "learning_rate": 2.868986353648664e-05, "loss": 0.7643, "step": 214070 }, { "epoch": 18.54, "learning_rate": 2.86876951543888e-05, "loss": 0.8982, "step": 214080 }, { "epoch": 18.54, "learning_rate": 2.8685526772290966e-05, "loss": 0.8239, "step": 214090 }, { "epoch": 18.54, "learning_rate": 2.868335839019313e-05, "loss": 0.7978, "step": 214100 }, { "epoch": 18.54, "learning_rate": 2.8681190008095292e-05, "loss": 0.6688, "step": 214110 }, { "epoch": 18.54, "learning_rate": 2.8679021625997452e-05, "loss": 0.7701, "step": 214120 }, { "epoch": 18.55, "learning_rate": 2.8676853243899615e-05, "loss": 0.772, "step": 214130 }, { "epoch": 18.55, "learning_rate": 2.8674684861801778e-05, "loss": 0.7531, "step": 214140 }, { "epoch": 18.55, "learning_rate": 2.8672516479703944e-05, "loss": 0.8804, "step": 214150 }, { "epoch": 18.55, "learning_rate": 2.8670348097606104e-05, "loss": 0.6863, "step": 214160 }, { "epoch": 18.55, "learning_rate": 2.8668179715508267e-05, "loss": 1.0076, "step": 214170 }, { "epoch": 18.55, "learning_rate": 2.866601133341043e-05, "loss": 0.7982, "step": 214180 }, { "epoch": 18.55, "learning_rate": 2.8663842951312593e-05, "loss": 0.8863, "step": 214190 }, { "epoch": 18.55, "learning_rate": 2.8661674569214752e-05, "loss": 0.913, "step": 214200 }, { "epoch": 18.55, "learning_rate": 2.8659506187116915e-05, "loss": 0.8855, "step": 214210 }, { "epoch": 18.55, "learning_rate": 2.8657337805019082e-05, "loss": 0.7882, "step": 214220 }, { "epoch": 18.55, "learning_rate": 2.8655169422921245e-05, "loss": 0.8472, "step": 214230 }, { "epoch": 18.55, "learning_rate": 2.8653001040823404e-05, "loss": 0.7115, "step": 214240 }, { "epoch": 18.56, "learning_rate": 2.8650832658725567e-05, "loss": 0.6733, "step": 214250 }, { "epoch": 18.56, "learning_rate": 2.864866427662773e-05, "loss": 0.7653, "step": 214260 }, { "epoch": 18.56, "learning_rate": 2.864649589452989e-05, "loss": 0.7501, "step": 214270 }, { "epoch": 18.56, "learning_rate": 2.8644327512432053e-05, "loss": 0.8114, "step": 214280 }, { "epoch": 18.56, "learning_rate": 2.8642159130334216e-05, "loss": 0.7689, "step": 214290 }, { "epoch": 18.56, "learning_rate": 2.8639990748236382e-05, "loss": 0.7977, "step": 214300 }, { "epoch": 18.56, "learning_rate": 2.8637822366138542e-05, "loss": 0.9048, "step": 214310 }, { "epoch": 18.56, "learning_rate": 2.8635653984040705e-05, "loss": 0.757, "step": 214320 }, { "epoch": 18.56, "learning_rate": 2.8633485601942868e-05, "loss": 0.9606, "step": 214330 }, { "epoch": 18.56, "learning_rate": 2.863131721984503e-05, "loss": 0.8159, "step": 214340 }, { "epoch": 18.56, "learning_rate": 2.862914883774719e-05, "loss": 0.8581, "step": 214350 }, { "epoch": 18.56, "learning_rate": 2.8626980455649354e-05, "loss": 0.7523, "step": 214360 }, { "epoch": 18.57, "learning_rate": 2.862481207355152e-05, "loss": 0.6969, "step": 214370 }, { "epoch": 18.57, "learning_rate": 2.8622643691453683e-05, "loss": 0.9739, "step": 214380 }, { "epoch": 18.57, "learning_rate": 2.8620475309355843e-05, "loss": 1.0126, "step": 214390 }, { "epoch": 18.57, "learning_rate": 2.8618306927258006e-05, "loss": 0.7009, "step": 214400 }, { "epoch": 18.57, "learning_rate": 2.861613854516017e-05, "loss": 0.769, "step": 214410 }, { "epoch": 18.57, "learning_rate": 2.8613970163062332e-05, "loss": 0.763, "step": 214420 }, { "epoch": 18.57, "learning_rate": 2.861180178096449e-05, "loss": 0.6962, "step": 214430 }, { "epoch": 18.57, "learning_rate": 2.8609633398866658e-05, "loss": 0.9278, "step": 214440 }, { "epoch": 18.57, "learning_rate": 2.860746501676882e-05, "loss": 0.7513, "step": 214450 }, { "epoch": 18.57, "learning_rate": 2.8605296634670984e-05, "loss": 0.7563, "step": 214460 }, { "epoch": 18.57, "learning_rate": 2.8603128252573143e-05, "loss": 0.829, "step": 214470 }, { "epoch": 18.58, "learning_rate": 2.8600959870475306e-05, "loss": 0.6647, "step": 214480 }, { "epoch": 18.58, "learning_rate": 2.859879148837747e-05, "loss": 0.8143, "step": 214490 }, { "epoch": 18.58, "learning_rate": 2.859662310627963e-05, "loss": 0.6782, "step": 214500 }, { "epoch": 18.58, "learning_rate": 2.8594454724181795e-05, "loss": 0.8213, "step": 214510 }, { "epoch": 18.58, "learning_rate": 2.859228634208396e-05, "loss": 0.7831, "step": 214520 }, { "epoch": 18.58, "learning_rate": 2.859011795998612e-05, "loss": 0.7155, "step": 214530 }, { "epoch": 18.58, "learning_rate": 2.858794957788828e-05, "loss": 0.903, "step": 214540 }, { "epoch": 18.58, "learning_rate": 2.8585781195790444e-05, "loss": 0.8675, "step": 214550 }, { "epoch": 18.58, "learning_rate": 2.8583612813692607e-05, "loss": 0.9556, "step": 214560 }, { "epoch": 18.58, "learning_rate": 2.858144443159477e-05, "loss": 0.7783, "step": 214570 }, { "epoch": 18.58, "learning_rate": 2.857927604949693e-05, "loss": 0.7294, "step": 214580 }, { "epoch": 18.58, "learning_rate": 2.8577107667399096e-05, "loss": 0.9021, "step": 214590 }, { "epoch": 18.59, "learning_rate": 2.857493928530126e-05, "loss": 0.8133, "step": 214600 }, { "epoch": 18.59, "learning_rate": 2.8572770903203422e-05, "loss": 0.7914, "step": 214610 }, { "epoch": 18.59, "learning_rate": 2.857060252110558e-05, "loss": 0.9078, "step": 214620 }, { "epoch": 18.59, "learning_rate": 2.8568434139007745e-05, "loss": 0.7289, "step": 214630 }, { "epoch": 18.59, "learning_rate": 2.8566265756909908e-05, "loss": 0.8431, "step": 214640 }, { "epoch": 18.59, "learning_rate": 2.8564097374812074e-05, "loss": 0.7083, "step": 214650 }, { "epoch": 18.59, "learning_rate": 2.8561928992714234e-05, "loss": 0.7635, "step": 214660 }, { "epoch": 18.59, "learning_rate": 2.8559760610616397e-05, "loss": 0.825, "step": 214670 }, { "epoch": 18.59, "learning_rate": 2.855759222851856e-05, "loss": 0.7455, "step": 214680 }, { "epoch": 18.59, "learning_rate": 2.8555423846420723e-05, "loss": 0.926, "step": 214690 }, { "epoch": 18.59, "learning_rate": 2.8553255464322882e-05, "loss": 0.8411, "step": 214700 }, { "epoch": 18.6, "learning_rate": 2.8551087082225045e-05, "loss": 0.7385, "step": 214710 }, { "epoch": 18.6, "learning_rate": 2.854891870012721e-05, "loss": 0.8152, "step": 214720 }, { "epoch": 18.6, "learning_rate": 2.8546750318029375e-05, "loss": 0.7913, "step": 214730 }, { "epoch": 18.6, "learning_rate": 2.8544581935931534e-05, "loss": 0.778, "step": 214740 }, { "epoch": 18.6, "learning_rate": 2.8542413553833697e-05, "loss": 0.722, "step": 214750 }, { "epoch": 18.6, "learning_rate": 2.854024517173586e-05, "loss": 0.8693, "step": 214760 }, { "epoch": 18.6, "learning_rate": 2.853807678963802e-05, "loss": 0.7086, "step": 214770 }, { "epoch": 18.6, "learning_rate": 2.8535908407540183e-05, "loss": 0.7771, "step": 214780 }, { "epoch": 18.6, "learning_rate": 2.853374002544235e-05, "loss": 0.7727, "step": 214790 }, { "epoch": 18.6, "learning_rate": 2.8531571643344512e-05, "loss": 0.7835, "step": 214800 }, { "epoch": 18.6, "learning_rate": 2.8529403261246672e-05, "loss": 0.8607, "step": 214810 }, { "epoch": 18.6, "learning_rate": 2.8527234879148835e-05, "loss": 0.9111, "step": 214820 }, { "epoch": 18.61, "learning_rate": 2.8525066497050998e-05, "loss": 0.6075, "step": 214830 }, { "epoch": 18.61, "learning_rate": 2.852289811495316e-05, "loss": 0.8937, "step": 214840 }, { "epoch": 18.61, "learning_rate": 2.852072973285532e-05, "loss": 0.8776, "step": 214850 }, { "epoch": 18.61, "learning_rate": 2.8518561350757484e-05, "loss": 0.7301, "step": 214860 }, { "epoch": 18.61, "learning_rate": 2.851639296865965e-05, "loss": 0.8888, "step": 214870 }, { "epoch": 18.61, "learning_rate": 2.8514224586561813e-05, "loss": 0.7906, "step": 214880 }, { "epoch": 18.61, "learning_rate": 2.8512056204463973e-05, "loss": 0.7759, "step": 214890 }, { "epoch": 18.61, "learning_rate": 2.8509887822366136e-05, "loss": 0.8356, "step": 214900 }, { "epoch": 18.61, "learning_rate": 2.85077194402683e-05, "loss": 0.8345, "step": 214910 }, { "epoch": 18.61, "learning_rate": 2.850555105817046e-05, "loss": 0.8463, "step": 214920 }, { "epoch": 18.61, "learning_rate": 2.850338267607262e-05, "loss": 0.681, "step": 214930 }, { "epoch": 18.62, "learning_rate": 2.8501214293974788e-05, "loss": 0.6829, "step": 214940 }, { "epoch": 18.62, "learning_rate": 2.849904591187695e-05, "loss": 0.8487, "step": 214950 }, { "epoch": 18.62, "learning_rate": 2.8496877529779114e-05, "loss": 0.8776, "step": 214960 }, { "epoch": 18.62, "learning_rate": 2.8494709147681273e-05, "loss": 0.812, "step": 214970 }, { "epoch": 18.62, "learning_rate": 2.8492540765583436e-05, "loss": 0.8664, "step": 214980 }, { "epoch": 18.62, "learning_rate": 2.84903723834856e-05, "loss": 0.842, "step": 214990 }, { "epoch": 18.62, "learning_rate": 2.8488204001387766e-05, "loss": 0.9734, "step": 215000 }, { "epoch": 18.62, "learning_rate": 2.8486035619289925e-05, "loss": 0.8405, "step": 215010 }, { "epoch": 18.62, "learning_rate": 2.8483867237192088e-05, "loss": 0.8867, "step": 215020 }, { "epoch": 18.62, "learning_rate": 2.848169885509425e-05, "loss": 0.8627, "step": 215030 }, { "epoch": 18.62, "learning_rate": 2.847953047299641e-05, "loss": 0.7172, "step": 215040 }, { "epoch": 18.62, "learning_rate": 2.8477362090898574e-05, "loss": 0.7956, "step": 215050 }, { "epoch": 18.63, "learning_rate": 2.8475193708800737e-05, "loss": 0.9446, "step": 215060 }, { "epoch": 18.63, "learning_rate": 2.8473025326702903e-05, "loss": 0.85, "step": 215070 }, { "epoch": 18.63, "learning_rate": 2.847085694460506e-05, "loss": 0.7562, "step": 215080 }, { "epoch": 18.63, "learning_rate": 2.8468688562507226e-05, "loss": 0.8479, "step": 215090 }, { "epoch": 18.63, "learning_rate": 2.846652018040939e-05, "loss": 0.7543, "step": 215100 }, { "epoch": 18.63, "learning_rate": 2.8464351798311552e-05, "loss": 0.8295, "step": 215110 }, { "epoch": 18.63, "learning_rate": 2.846218341621371e-05, "loss": 0.7723, "step": 215120 }, { "epoch": 18.63, "learning_rate": 2.8460015034115874e-05, "loss": 0.7257, "step": 215130 }, { "epoch": 18.63, "learning_rate": 2.8457846652018037e-05, "loss": 0.7836, "step": 215140 }, { "epoch": 18.63, "learning_rate": 2.8455678269920204e-05, "loss": 0.9482, "step": 215150 }, { "epoch": 18.63, "learning_rate": 2.8453509887822363e-05, "loss": 0.9726, "step": 215160 }, { "epoch": 18.64, "learning_rate": 2.8451341505724526e-05, "loss": 0.8807, "step": 215170 }, { "epoch": 18.64, "learning_rate": 2.844917312362669e-05, "loss": 0.7997, "step": 215180 }, { "epoch": 18.64, "learning_rate": 2.8447004741528853e-05, "loss": 0.6677, "step": 215190 }, { "epoch": 18.64, "learning_rate": 2.8444836359431012e-05, "loss": 0.6745, "step": 215200 }, { "epoch": 18.64, "learning_rate": 2.8442667977333175e-05, "loss": 0.8038, "step": 215210 }, { "epoch": 18.64, "learning_rate": 2.844049959523534e-05, "loss": 0.8587, "step": 215220 }, { "epoch": 18.64, "learning_rate": 2.8438331213137505e-05, "loss": 0.7099, "step": 215230 }, { "epoch": 18.64, "learning_rate": 2.8436162831039664e-05, "loss": 0.8837, "step": 215240 }, { "epoch": 18.64, "learning_rate": 2.8433994448941827e-05, "loss": 0.8018, "step": 215250 }, { "epoch": 18.64, "learning_rate": 2.843182606684399e-05, "loss": 0.8199, "step": 215260 }, { "epoch": 18.64, "learning_rate": 2.842965768474615e-05, "loss": 0.6005, "step": 215270 }, { "epoch": 18.64, "learning_rate": 2.8427489302648313e-05, "loss": 0.9475, "step": 215280 }, { "epoch": 18.65, "learning_rate": 2.842532092055048e-05, "loss": 0.739, "step": 215290 }, { "epoch": 18.65, "learning_rate": 2.8423152538452642e-05, "loss": 0.9909, "step": 215300 }, { "epoch": 18.65, "learning_rate": 2.8420984156354802e-05, "loss": 0.9087, "step": 215310 }, { "epoch": 18.65, "learning_rate": 2.8418815774256965e-05, "loss": 0.7496, "step": 215320 }, { "epoch": 18.65, "learning_rate": 2.8416647392159128e-05, "loss": 0.7642, "step": 215330 }, { "epoch": 18.65, "learning_rate": 2.841447901006129e-05, "loss": 0.7868, "step": 215340 }, { "epoch": 18.65, "learning_rate": 2.841231062796345e-05, "loss": 0.6988, "step": 215350 }, { "epoch": 18.65, "learning_rate": 2.8410142245865613e-05, "loss": 0.7581, "step": 215360 }, { "epoch": 18.65, "learning_rate": 2.840797386376778e-05, "loss": 0.8307, "step": 215370 }, { "epoch": 18.65, "learning_rate": 2.8405805481669943e-05, "loss": 0.7323, "step": 215380 }, { "epoch": 18.65, "learning_rate": 2.8403637099572102e-05, "loss": 0.8176, "step": 215390 }, { "epoch": 18.66, "learning_rate": 2.8401468717474265e-05, "loss": 0.8311, "step": 215400 }, { "epoch": 18.66, "learning_rate": 2.839930033537643e-05, "loss": 0.8621, "step": 215410 }, { "epoch": 18.66, "learning_rate": 2.839713195327859e-05, "loss": 0.7492, "step": 215420 }, { "epoch": 18.66, "learning_rate": 2.839496357118075e-05, "loss": 0.922, "step": 215430 }, { "epoch": 18.66, "learning_rate": 2.8392795189082917e-05, "loss": 0.7192, "step": 215440 }, { "epoch": 18.66, "learning_rate": 2.839062680698508e-05, "loss": 0.8103, "step": 215450 }, { "epoch": 18.66, "learning_rate": 2.8388458424887243e-05, "loss": 0.8531, "step": 215460 }, { "epoch": 18.66, "learning_rate": 2.8386290042789403e-05, "loss": 0.7711, "step": 215470 }, { "epoch": 18.66, "learning_rate": 2.8384121660691566e-05, "loss": 0.7461, "step": 215480 }, { "epoch": 18.66, "learning_rate": 2.838195327859373e-05, "loss": 0.799, "step": 215490 }, { "epoch": 18.66, "learning_rate": 2.8379784896495895e-05, "loss": 0.8566, "step": 215500 }, { "epoch": 18.66, "learning_rate": 2.8377616514398055e-05, "loss": 0.8159, "step": 215510 }, { "epoch": 18.67, "learning_rate": 2.8375448132300218e-05, "loss": 0.7573, "step": 215520 }, { "epoch": 18.67, "learning_rate": 2.837327975020238e-05, "loss": 0.8197, "step": 215530 }, { "epoch": 18.67, "learning_rate": 2.837111136810454e-05, "loss": 0.7269, "step": 215540 }, { "epoch": 18.67, "learning_rate": 2.8368942986006704e-05, "loss": 0.7509, "step": 215550 }, { "epoch": 18.67, "learning_rate": 2.8366774603908867e-05, "loss": 0.7134, "step": 215560 }, { "epoch": 18.67, "learning_rate": 2.8364606221811033e-05, "loss": 0.7756, "step": 215570 }, { "epoch": 18.67, "learning_rate": 2.8362437839713193e-05, "loss": 0.8605, "step": 215580 }, { "epoch": 18.67, "learning_rate": 2.8360269457615356e-05, "loss": 0.8664, "step": 215590 }, { "epoch": 18.67, "learning_rate": 2.835810107551752e-05, "loss": 0.7636, "step": 215600 }, { "epoch": 18.67, "learning_rate": 2.8355932693419682e-05, "loss": 0.8062, "step": 215610 }, { "epoch": 18.67, "learning_rate": 2.835376431132184e-05, "loss": 0.7017, "step": 215620 }, { "epoch": 18.67, "learning_rate": 2.8351595929224004e-05, "loss": 0.6556, "step": 215630 }, { "epoch": 18.68, "learning_rate": 2.8349427547126167e-05, "loss": 0.782, "step": 215640 }, { "epoch": 18.68, "learning_rate": 2.8347259165028334e-05, "loss": 0.7718, "step": 215650 }, { "epoch": 18.68, "learning_rate": 2.8345090782930493e-05, "loss": 0.9218, "step": 215660 }, { "epoch": 18.68, "learning_rate": 2.8342922400832656e-05, "loss": 0.797, "step": 215670 }, { "epoch": 18.68, "learning_rate": 2.834075401873482e-05, "loss": 0.932, "step": 215680 }, { "epoch": 18.68, "learning_rate": 2.8338585636636982e-05, "loss": 0.8577, "step": 215690 }, { "epoch": 18.68, "learning_rate": 2.8336417254539142e-05, "loss": 0.767, "step": 215700 }, { "epoch": 18.68, "learning_rate": 2.8334248872441305e-05, "loss": 0.8116, "step": 215710 }, { "epoch": 18.68, "learning_rate": 2.833208049034347e-05, "loss": 0.8937, "step": 215720 }, { "epoch": 18.68, "learning_rate": 2.8329912108245634e-05, "loss": 0.6612, "step": 215730 }, { "epoch": 18.68, "learning_rate": 2.8327743726147794e-05, "loss": 0.8212, "step": 215740 }, { "epoch": 18.69, "learning_rate": 2.8325575344049957e-05, "loss": 0.6885, "step": 215750 }, { "epoch": 18.69, "learning_rate": 2.832340696195212e-05, "loss": 0.7259, "step": 215760 }, { "epoch": 18.69, "learning_rate": 2.8321238579854283e-05, "loss": 0.8742, "step": 215770 }, { "epoch": 18.69, "learning_rate": 2.8319070197756443e-05, "loss": 0.9276, "step": 215780 }, { "epoch": 18.69, "learning_rate": 2.831690181565861e-05, "loss": 0.8209, "step": 215790 }, { "epoch": 18.69, "learning_rate": 2.8314733433560772e-05, "loss": 0.7767, "step": 215800 }, { "epoch": 18.69, "learning_rate": 2.831256505146293e-05, "loss": 0.779, "step": 215810 }, { "epoch": 18.69, "learning_rate": 2.8310396669365095e-05, "loss": 0.7822, "step": 215820 }, { "epoch": 18.69, "learning_rate": 2.8308228287267258e-05, "loss": 0.7567, "step": 215830 }, { "epoch": 18.69, "learning_rate": 2.830605990516942e-05, "loss": 0.8, "step": 215840 }, { "epoch": 18.69, "learning_rate": 2.830389152307158e-05, "loss": 0.7454, "step": 215850 }, { "epoch": 18.69, "learning_rate": 2.8301723140973747e-05, "loss": 0.8826, "step": 215860 }, { "epoch": 18.7, "learning_rate": 2.829955475887591e-05, "loss": 0.955, "step": 215870 }, { "epoch": 18.7, "learning_rate": 2.8297386376778073e-05, "loss": 0.921, "step": 215880 }, { "epoch": 18.7, "learning_rate": 2.8295217994680232e-05, "loss": 0.8905, "step": 215890 }, { "epoch": 18.7, "learning_rate": 2.8293049612582395e-05, "loss": 0.8644, "step": 215900 }, { "epoch": 18.7, "learning_rate": 2.8290881230484558e-05, "loss": 0.8812, "step": 215910 }, { "epoch": 18.7, "learning_rate": 2.828871284838672e-05, "loss": 0.9346, "step": 215920 }, { "epoch": 18.7, "learning_rate": 2.828654446628888e-05, "loss": 0.5934, "step": 215930 }, { "epoch": 18.7, "learning_rate": 2.8284376084191047e-05, "loss": 0.6762, "step": 215940 }, { "epoch": 18.7, "learning_rate": 2.828220770209321e-05, "loss": 0.9656, "step": 215950 }, { "epoch": 18.7, "learning_rate": 2.8280039319995373e-05, "loss": 0.7562, "step": 215960 }, { "epoch": 18.7, "learning_rate": 2.8277870937897533e-05, "loss": 0.7946, "step": 215970 }, { "epoch": 18.71, "learning_rate": 2.8275702555799696e-05, "loss": 0.6494, "step": 215980 }, { "epoch": 18.71, "learning_rate": 2.827353417370186e-05, "loss": 0.7363, "step": 215990 }, { "epoch": 18.71, "learning_rate": 2.8271365791604025e-05, "loss": 0.7789, "step": 216000 }, { "epoch": 18.71, "learning_rate": 2.8269197409506185e-05, "loss": 0.7888, "step": 216010 }, { "epoch": 18.71, "learning_rate": 2.8267029027408348e-05, "loss": 0.8462, "step": 216020 }, { "epoch": 18.71, "learning_rate": 2.826486064531051e-05, "loss": 0.7388, "step": 216030 }, { "epoch": 18.71, "learning_rate": 2.826269226321267e-05, "loss": 0.9125, "step": 216040 }, { "epoch": 18.71, "learning_rate": 2.8260523881114834e-05, "loss": 0.8076, "step": 216050 }, { "epoch": 18.71, "learning_rate": 2.8258355499016997e-05, "loss": 0.6624, "step": 216060 }, { "epoch": 18.71, "learning_rate": 2.8256187116919163e-05, "loss": 0.8569, "step": 216070 }, { "epoch": 18.71, "learning_rate": 2.8254018734821323e-05, "loss": 0.6946, "step": 216080 }, { "epoch": 18.71, "learning_rate": 2.8251850352723486e-05, "loss": 1.0441, "step": 216090 }, { "epoch": 18.72, "learning_rate": 2.824968197062565e-05, "loss": 0.6851, "step": 216100 }, { "epoch": 18.72, "learning_rate": 2.824751358852781e-05, "loss": 0.8719, "step": 216110 }, { "epoch": 18.72, "learning_rate": 2.824534520642997e-05, "loss": 0.8203, "step": 216120 }, { "epoch": 18.72, "learning_rate": 2.8243176824332134e-05, "loss": 0.8374, "step": 216130 }, { "epoch": 18.72, "learning_rate": 2.82410084422343e-05, "loss": 0.7946, "step": 216140 }, { "epoch": 18.72, "learning_rate": 2.8238840060136464e-05, "loss": 0.8984, "step": 216150 }, { "epoch": 18.72, "learning_rate": 2.8236671678038623e-05, "loss": 0.8266, "step": 216160 }, { "epoch": 18.72, "learning_rate": 2.8234503295940786e-05, "loss": 0.6319, "step": 216170 }, { "epoch": 18.72, "learning_rate": 2.823233491384295e-05, "loss": 0.8947, "step": 216180 }, { "epoch": 18.72, "learning_rate": 2.8230166531745112e-05, "loss": 0.8188, "step": 216190 }, { "epoch": 18.72, "learning_rate": 2.8227998149647272e-05, "loss": 0.6448, "step": 216200 }, { "epoch": 18.73, "learning_rate": 2.8225829767549435e-05, "loss": 0.952, "step": 216210 }, { "epoch": 18.73, "learning_rate": 2.82236613854516e-05, "loss": 0.7399, "step": 216220 }, { "epoch": 18.73, "learning_rate": 2.8221493003353764e-05, "loss": 0.7994, "step": 216230 }, { "epoch": 18.73, "learning_rate": 2.8219324621255924e-05, "loss": 0.875, "step": 216240 }, { "epoch": 18.73, "learning_rate": 2.8217156239158087e-05, "loss": 0.8392, "step": 216250 }, { "epoch": 18.73, "learning_rate": 2.821498785706025e-05, "loss": 0.8866, "step": 216260 }, { "epoch": 18.73, "learning_rate": 2.8212819474962413e-05, "loss": 0.7877, "step": 216270 }, { "epoch": 18.73, "learning_rate": 2.8210651092864572e-05, "loss": 0.6103, "step": 216280 }, { "epoch": 18.73, "learning_rate": 2.820848271076674e-05, "loss": 0.7381, "step": 216290 }, { "epoch": 18.73, "learning_rate": 2.8206314328668902e-05, "loss": 0.8632, "step": 216300 }, { "epoch": 18.73, "learning_rate": 2.820414594657106e-05, "loss": 0.7493, "step": 216310 }, { "epoch": 18.73, "learning_rate": 2.8201977564473224e-05, "loss": 0.8761, "step": 216320 }, { "epoch": 18.74, "learning_rate": 2.8199809182375387e-05, "loss": 0.7898, "step": 216330 }, { "epoch": 18.74, "learning_rate": 2.819764080027755e-05, "loss": 0.8075, "step": 216340 }, { "epoch": 18.74, "learning_rate": 2.819547241817971e-05, "loss": 0.8618, "step": 216350 }, { "epoch": 18.74, "learning_rate": 2.8193304036081876e-05, "loss": 0.739, "step": 216360 }, { "epoch": 18.74, "learning_rate": 2.819113565398404e-05, "loss": 0.7867, "step": 216370 }, { "epoch": 18.74, "learning_rate": 2.8188967271886202e-05, "loss": 0.6986, "step": 216380 }, { "epoch": 18.74, "learning_rate": 2.8186798889788362e-05, "loss": 0.7018, "step": 216390 }, { "epoch": 18.74, "learning_rate": 2.8184630507690525e-05, "loss": 0.6902, "step": 216400 }, { "epoch": 18.74, "learning_rate": 2.8182462125592688e-05, "loss": 0.6689, "step": 216410 }, { "epoch": 18.74, "learning_rate": 2.8180293743494854e-05, "loss": 0.714, "step": 216420 }, { "epoch": 18.74, "learning_rate": 2.817812536139701e-05, "loss": 0.897, "step": 216430 }, { "epoch": 18.75, "learning_rate": 2.8175956979299177e-05, "loss": 0.8793, "step": 216440 }, { "epoch": 18.75, "learning_rate": 2.817378859720134e-05, "loss": 0.8815, "step": 216450 }, { "epoch": 18.75, "learning_rate": 2.8171620215103503e-05, "loss": 0.8044, "step": 216460 }, { "epoch": 18.75, "learning_rate": 2.8169451833005663e-05, "loss": 0.6397, "step": 216470 }, { "epoch": 18.75, "learning_rate": 2.8167283450907826e-05, "loss": 0.9108, "step": 216480 }, { "epoch": 18.75, "learning_rate": 2.816511506880999e-05, "loss": 0.9359, "step": 216490 }, { "epoch": 18.75, "learning_rate": 2.8162946686712155e-05, "loss": 0.6968, "step": 216500 }, { "epoch": 18.75, "learning_rate": 2.8160778304614315e-05, "loss": 0.9037, "step": 216510 }, { "epoch": 18.75, "learning_rate": 2.8158609922516478e-05, "loss": 0.7476, "step": 216520 }, { "epoch": 18.75, "learning_rate": 2.815644154041864e-05, "loss": 0.7954, "step": 216530 }, { "epoch": 18.75, "learning_rate": 2.8154273158320804e-05, "loss": 0.8658, "step": 216540 }, { "epoch": 18.75, "learning_rate": 2.8152104776222963e-05, "loss": 0.9888, "step": 216550 }, { "epoch": 18.76, "learning_rate": 2.8149936394125126e-05, "loss": 0.837, "step": 216560 }, { "epoch": 18.76, "learning_rate": 2.8147768012027293e-05, "loss": 0.71, "step": 216570 }, { "epoch": 18.76, "learning_rate": 2.8145599629929452e-05, "loss": 0.7533, "step": 216580 }, { "epoch": 18.76, "learning_rate": 2.8143431247831615e-05, "loss": 0.7655, "step": 216590 }, { "epoch": 18.76, "learning_rate": 2.814126286573378e-05, "loss": 0.6644, "step": 216600 }, { "epoch": 18.76, "learning_rate": 2.813909448363594e-05, "loss": 0.9306, "step": 216610 }, { "epoch": 18.76, "learning_rate": 2.81369261015381e-05, "loss": 0.6849, "step": 216620 }, { "epoch": 18.76, "learning_rate": 2.8134757719440264e-05, "loss": 0.7053, "step": 216630 }, { "epoch": 18.76, "learning_rate": 2.813258933734243e-05, "loss": 0.8492, "step": 216640 }, { "epoch": 18.76, "learning_rate": 2.8130420955244593e-05, "loss": 0.7742, "step": 216650 }, { "epoch": 18.76, "learning_rate": 2.8128252573146753e-05, "loss": 0.8419, "step": 216660 }, { "epoch": 18.76, "learning_rate": 2.8126084191048916e-05, "loss": 0.8335, "step": 216670 }, { "epoch": 18.77, "learning_rate": 2.812391580895108e-05, "loss": 0.8169, "step": 216680 }, { "epoch": 18.77, "learning_rate": 2.8121747426853242e-05, "loss": 0.8374, "step": 216690 }, { "epoch": 18.77, "learning_rate": 2.81195790447554e-05, "loss": 0.8746, "step": 216700 }, { "epoch": 18.77, "learning_rate": 2.8117410662657565e-05, "loss": 0.7906, "step": 216710 }, { "epoch": 18.77, "learning_rate": 2.811524228055973e-05, "loss": 0.7541, "step": 216720 }, { "epoch": 18.77, "learning_rate": 2.8113073898461894e-05, "loss": 0.7071, "step": 216730 }, { "epoch": 18.77, "learning_rate": 2.8110905516364054e-05, "loss": 0.7887, "step": 216740 }, { "epoch": 18.77, "learning_rate": 2.8108737134266217e-05, "loss": 0.7299, "step": 216750 }, { "epoch": 18.77, "learning_rate": 2.810656875216838e-05, "loss": 0.7726, "step": 216760 }, { "epoch": 18.77, "learning_rate": 2.8104400370070543e-05, "loss": 0.8358, "step": 216770 }, { "epoch": 18.77, "learning_rate": 2.8102231987972702e-05, "loss": 0.8289, "step": 216780 }, { "epoch": 18.78, "learning_rate": 2.810006360587487e-05, "loss": 0.7434, "step": 216790 }, { "epoch": 18.78, "learning_rate": 2.809789522377703e-05, "loss": 0.7658, "step": 216800 }, { "epoch": 18.78, "learning_rate": 2.809572684167919e-05, "loss": 0.7987, "step": 216810 }, { "epoch": 18.78, "learning_rate": 2.8093558459581354e-05, "loss": 0.7712, "step": 216820 }, { "epoch": 18.78, "learning_rate": 2.8091390077483517e-05, "loss": 0.811, "step": 216830 }, { "epoch": 18.78, "learning_rate": 2.808922169538568e-05, "loss": 0.8073, "step": 216840 }, { "epoch": 18.78, "learning_rate": 2.808705331328784e-05, "loss": 0.7369, "step": 216850 }, { "epoch": 18.78, "learning_rate": 2.8084884931190006e-05, "loss": 0.9054, "step": 216860 }, { "epoch": 18.78, "learning_rate": 2.808271654909217e-05, "loss": 0.8802, "step": 216870 }, { "epoch": 18.78, "learning_rate": 2.8080548166994332e-05, "loss": 0.7959, "step": 216880 }, { "epoch": 18.78, "learning_rate": 2.8078379784896492e-05, "loss": 0.7491, "step": 216890 }, { "epoch": 18.78, "learning_rate": 2.8076211402798655e-05, "loss": 0.799, "step": 216900 }, { "epoch": 18.79, "learning_rate": 2.8074043020700818e-05, "loss": 0.8692, "step": 216910 }, { "epoch": 18.79, "learning_rate": 2.8071874638602984e-05, "loss": 0.8432, "step": 216920 }, { "epoch": 18.79, "learning_rate": 2.806970625650514e-05, "loss": 0.7086, "step": 216930 }, { "epoch": 18.79, "learning_rate": 2.8067537874407307e-05, "loss": 0.8063, "step": 216940 }, { "epoch": 18.79, "learning_rate": 2.806536949230947e-05, "loss": 1.0163, "step": 216950 }, { "epoch": 18.79, "learning_rate": 2.8063201110211633e-05, "loss": 0.9242, "step": 216960 }, { "epoch": 18.79, "learning_rate": 2.8061032728113793e-05, "loss": 0.7975, "step": 216970 }, { "epoch": 18.79, "learning_rate": 2.8058864346015956e-05, "loss": 0.8103, "step": 216980 }, { "epoch": 18.79, "learning_rate": 2.805669596391812e-05, "loss": 0.8537, "step": 216990 }, { "epoch": 18.79, "learning_rate": 2.8054527581820285e-05, "loss": 0.7082, "step": 217000 }, { "epoch": 18.79, "learning_rate": 2.8052359199722445e-05, "loss": 0.828, "step": 217010 }, { "epoch": 18.8, "learning_rate": 2.8050190817624608e-05, "loss": 0.8093, "step": 217020 }, { "epoch": 18.8, "learning_rate": 2.804802243552677e-05, "loss": 0.8406, "step": 217030 }, { "epoch": 18.8, "learning_rate": 2.8045854053428934e-05, "loss": 0.65, "step": 217040 }, { "epoch": 18.8, "learning_rate": 2.8043685671331093e-05, "loss": 0.9341, "step": 217050 }, { "epoch": 18.8, "learning_rate": 2.8041517289233256e-05, "loss": 0.7376, "step": 217060 }, { "epoch": 18.8, "learning_rate": 2.8039348907135423e-05, "loss": 0.7939, "step": 217070 }, { "epoch": 18.8, "learning_rate": 2.8037180525037582e-05, "loss": 0.7964, "step": 217080 }, { "epoch": 18.8, "learning_rate": 2.8035012142939745e-05, "loss": 0.6869, "step": 217090 }, { "epoch": 18.8, "learning_rate": 2.8032843760841908e-05, "loss": 0.8715, "step": 217100 }, { "epoch": 18.8, "learning_rate": 2.803067537874407e-05, "loss": 0.8157, "step": 217110 }, { "epoch": 18.8, "learning_rate": 2.802850699664623e-05, "loss": 0.7481, "step": 217120 }, { "epoch": 18.8, "learning_rate": 2.8026338614548394e-05, "loss": 0.9336, "step": 217130 }, { "epoch": 18.81, "learning_rate": 2.802417023245056e-05, "loss": 0.8126, "step": 217140 }, { "epoch": 18.81, "learning_rate": 2.8022001850352723e-05, "loss": 0.8221, "step": 217150 }, { "epoch": 18.81, "learning_rate": 2.8019833468254883e-05, "loss": 0.7976, "step": 217160 }, { "epoch": 18.81, "learning_rate": 2.8017665086157046e-05, "loss": 0.7873, "step": 217170 }, { "epoch": 18.81, "learning_rate": 2.801549670405921e-05, "loss": 0.7545, "step": 217180 }, { "epoch": 18.81, "learning_rate": 2.8013328321961372e-05, "loss": 1.0278, "step": 217190 }, { "epoch": 18.81, "learning_rate": 2.801115993986353e-05, "loss": 0.9018, "step": 217200 }, { "epoch": 18.81, "learning_rate": 2.8008991557765695e-05, "loss": 0.7859, "step": 217210 }, { "epoch": 18.81, "learning_rate": 2.800682317566786e-05, "loss": 0.7485, "step": 217220 }, { "epoch": 18.81, "learning_rate": 2.8004654793570024e-05, "loss": 0.8775, "step": 217230 }, { "epoch": 18.81, "learning_rate": 2.8002486411472184e-05, "loss": 0.7371, "step": 217240 }, { "epoch": 18.82, "learning_rate": 2.8000318029374347e-05, "loss": 0.8619, "step": 217250 }, { "epoch": 18.82, "learning_rate": 2.799814964727651e-05, "loss": 0.9946, "step": 217260 }, { "epoch": 18.82, "learning_rate": 2.7995981265178673e-05, "loss": 0.6719, "step": 217270 }, { "epoch": 18.82, "learning_rate": 2.7993812883080832e-05, "loss": 0.7213, "step": 217280 }, { "epoch": 18.82, "learning_rate": 2.7991644500983e-05, "loss": 0.7871, "step": 217290 }, { "epoch": 18.82, "learning_rate": 2.798947611888516e-05, "loss": 0.8768, "step": 217300 }, { "epoch": 18.82, "learning_rate": 2.7987307736787325e-05, "loss": 0.7671, "step": 217310 }, { "epoch": 18.82, "learning_rate": 2.7985139354689484e-05, "loss": 0.7201, "step": 217320 }, { "epoch": 18.82, "learning_rate": 2.7982970972591647e-05, "loss": 0.8011, "step": 217330 }, { "epoch": 18.82, "learning_rate": 2.798080259049381e-05, "loss": 0.7602, "step": 217340 }, { "epoch": 18.82, "learning_rate": 2.797863420839597e-05, "loss": 0.8138, "step": 217350 }, { "epoch": 18.82, "learning_rate": 2.7976465826298136e-05, "loss": 0.8903, "step": 217360 }, { "epoch": 18.83, "learning_rate": 2.79742974442003e-05, "loss": 0.8033, "step": 217370 }, { "epoch": 18.83, "learning_rate": 2.7972129062102462e-05, "loss": 0.784, "step": 217380 }, { "epoch": 18.83, "learning_rate": 2.7969960680004622e-05, "loss": 0.6745, "step": 217390 }, { "epoch": 18.83, "learning_rate": 2.7967792297906785e-05, "loss": 0.8451, "step": 217400 }, { "epoch": 18.83, "learning_rate": 2.7965623915808948e-05, "loss": 0.7004, "step": 217410 }, { "epoch": 18.83, "learning_rate": 2.7963455533711114e-05, "loss": 0.8022, "step": 217420 }, { "epoch": 18.83, "learning_rate": 2.7961287151613274e-05, "loss": 0.9803, "step": 217430 }, { "epoch": 18.83, "learning_rate": 2.7959118769515437e-05, "loss": 0.8353, "step": 217440 }, { "epoch": 18.83, "learning_rate": 2.79569503874176e-05, "loss": 0.6968, "step": 217450 }, { "epoch": 18.83, "learning_rate": 2.7954782005319763e-05, "loss": 0.953, "step": 217460 }, { "epoch": 18.83, "learning_rate": 2.7952613623221922e-05, "loss": 0.9398, "step": 217470 }, { "epoch": 18.84, "learning_rate": 2.7950445241124085e-05, "loss": 0.8443, "step": 217480 }, { "epoch": 18.84, "learning_rate": 2.794827685902625e-05, "loss": 0.7751, "step": 217490 }, { "epoch": 18.84, "learning_rate": 2.7946108476928415e-05, "loss": 0.8345, "step": 217500 }, { "epoch": 18.84, "learning_rate": 2.7943940094830574e-05, "loss": 0.787, "step": 217510 }, { "epoch": 18.84, "learning_rate": 2.7941771712732737e-05, "loss": 0.7203, "step": 217520 }, { "epoch": 18.84, "learning_rate": 2.79396033306349e-05, "loss": 0.7506, "step": 217530 }, { "epoch": 18.84, "learning_rate": 2.7937434948537063e-05, "loss": 0.9705, "step": 217540 }, { "epoch": 18.84, "learning_rate": 2.7935266566439223e-05, "loss": 0.7168, "step": 217550 }, { "epoch": 18.84, "learning_rate": 2.7933098184341386e-05, "loss": 0.8144, "step": 217560 }, { "epoch": 18.84, "learning_rate": 2.7930929802243552e-05, "loss": 0.7532, "step": 217570 }, { "epoch": 18.84, "learning_rate": 2.7928761420145712e-05, "loss": 0.6116, "step": 217580 }, { "epoch": 18.84, "learning_rate": 2.7926593038047875e-05, "loss": 0.8945, "step": 217590 }, { "epoch": 18.85, "learning_rate": 2.7924424655950038e-05, "loss": 0.9007, "step": 217600 }, { "epoch": 18.85, "learning_rate": 2.79222562738522e-05, "loss": 0.8738, "step": 217610 }, { "epoch": 18.85, "learning_rate": 2.792008789175436e-05, "loss": 0.7148, "step": 217620 }, { "epoch": 18.85, "learning_rate": 2.7917919509656524e-05, "loss": 0.6854, "step": 217630 }, { "epoch": 18.85, "learning_rate": 2.791575112755869e-05, "loss": 0.729, "step": 217640 }, { "epoch": 18.85, "learning_rate": 2.7913582745460853e-05, "loss": 0.8324, "step": 217650 }, { "epoch": 18.85, "learning_rate": 2.7911414363363013e-05, "loss": 0.7942, "step": 217660 }, { "epoch": 18.85, "learning_rate": 2.7909245981265176e-05, "loss": 0.8938, "step": 217670 }, { "epoch": 18.85, "learning_rate": 2.790707759916734e-05, "loss": 0.9841, "step": 217680 }, { "epoch": 18.85, "learning_rate": 2.7904909217069502e-05, "loss": 0.7795, "step": 217690 }, { "epoch": 18.85, "learning_rate": 2.790274083497166e-05, "loss": 0.632, "step": 217700 }, { "epoch": 18.86, "learning_rate": 2.7900572452873828e-05, "loss": 0.7236, "step": 217710 }, { "epoch": 18.86, "learning_rate": 2.789840407077599e-05, "loss": 0.7977, "step": 217720 }, { "epoch": 18.86, "learning_rate": 2.7896235688678154e-05, "loss": 0.5879, "step": 217730 }, { "epoch": 18.86, "learning_rate": 2.7894067306580313e-05, "loss": 0.6463, "step": 217740 }, { "epoch": 18.86, "learning_rate": 2.7891898924482476e-05, "loss": 0.7717, "step": 217750 }, { "epoch": 18.86, "learning_rate": 2.788973054238464e-05, "loss": 0.6816, "step": 217760 }, { "epoch": 18.86, "learning_rate": 2.7887562160286802e-05, "loss": 0.7309, "step": 217770 }, { "epoch": 18.86, "learning_rate": 2.7885393778188962e-05, "loss": 0.7815, "step": 217780 }, { "epoch": 18.86, "learning_rate": 2.788322539609113e-05, "loss": 0.971, "step": 217790 }, { "epoch": 18.86, "learning_rate": 2.788105701399329e-05, "loss": 0.8007, "step": 217800 }, { "epoch": 18.86, "learning_rate": 2.7878888631895454e-05, "loss": 0.8447, "step": 217810 }, { "epoch": 18.86, "learning_rate": 2.7876720249797614e-05, "loss": 0.7638, "step": 217820 }, { "epoch": 18.87, "learning_rate": 2.7874551867699777e-05, "loss": 0.9826, "step": 217830 }, { "epoch": 18.87, "learning_rate": 2.787238348560194e-05, "loss": 0.7941, "step": 217840 }, { "epoch": 18.87, "learning_rate": 2.78702151035041e-05, "loss": 0.7117, "step": 217850 }, { "epoch": 18.87, "learning_rate": 2.7868046721406266e-05, "loss": 0.8363, "step": 217860 }, { "epoch": 18.87, "learning_rate": 2.786587833930843e-05, "loss": 0.9121, "step": 217870 }, { "epoch": 18.87, "learning_rate": 2.7863709957210592e-05, "loss": 0.7214, "step": 217880 }, { "epoch": 18.87, "learning_rate": 2.786154157511275e-05, "loss": 0.7722, "step": 217890 }, { "epoch": 18.87, "learning_rate": 2.7859373193014915e-05, "loss": 0.9297, "step": 217900 }, { "epoch": 18.87, "learning_rate": 2.7857204810917078e-05, "loss": 0.6655, "step": 217910 }, { "epoch": 18.87, "learning_rate": 2.7855036428819244e-05, "loss": 0.8012, "step": 217920 }, { "epoch": 18.87, "learning_rate": 2.7852868046721404e-05, "loss": 0.7531, "step": 217930 }, { "epoch": 18.87, "learning_rate": 2.7850699664623567e-05, "loss": 0.9047, "step": 217940 }, { "epoch": 18.88, "learning_rate": 2.784853128252573e-05, "loss": 0.7621, "step": 217950 }, { "epoch": 18.88, "learning_rate": 2.7846362900427893e-05, "loss": 0.7788, "step": 217960 }, { "epoch": 18.88, "learning_rate": 2.7844194518330052e-05, "loss": 0.8578, "step": 217970 }, { "epoch": 18.88, "learning_rate": 2.7842026136232215e-05, "loss": 0.8214, "step": 217980 }, { "epoch": 18.88, "learning_rate": 2.783985775413438e-05, "loss": 0.7172, "step": 217990 }, { "epoch": 18.88, "learning_rate": 2.7837689372036545e-05, "loss": 0.8954, "step": 218000 }, { "epoch": 18.88, "learning_rate": 2.7835520989938704e-05, "loss": 0.61, "step": 218010 }, { "epoch": 18.88, "learning_rate": 2.7833352607840867e-05, "loss": 0.8082, "step": 218020 }, { "epoch": 18.88, "learning_rate": 2.783118422574303e-05, "loss": 0.9104, "step": 218030 }, { "epoch": 18.88, "learning_rate": 2.7829015843645193e-05, "loss": 0.7664, "step": 218040 }, { "epoch": 18.88, "learning_rate": 2.7826847461547353e-05, "loss": 0.9158, "step": 218050 }, { "epoch": 18.89, "learning_rate": 2.7824679079449516e-05, "loss": 0.7313, "step": 218060 }, { "epoch": 18.89, "learning_rate": 2.7822510697351682e-05, "loss": 0.876, "step": 218070 }, { "epoch": 18.89, "learning_rate": 2.7820342315253845e-05, "loss": 0.8932, "step": 218080 }, { "epoch": 18.89, "learning_rate": 2.7818173933156005e-05, "loss": 0.6931, "step": 218090 }, { "epoch": 18.89, "learning_rate": 2.7816005551058168e-05, "loss": 0.9476, "step": 218100 }, { "epoch": 18.89, "learning_rate": 2.781383716896033e-05, "loss": 0.8162, "step": 218110 }, { "epoch": 18.89, "learning_rate": 2.781166878686249e-05, "loss": 0.8199, "step": 218120 }, { "epoch": 18.89, "learning_rate": 2.7809500404764654e-05, "loss": 0.669, "step": 218130 }, { "epoch": 18.89, "learning_rate": 2.780733202266682e-05, "loss": 0.7111, "step": 218140 }, { "epoch": 18.89, "learning_rate": 2.7805163640568983e-05, "loss": 0.9034, "step": 218150 }, { "epoch": 18.89, "learning_rate": 2.7802995258471143e-05, "loss": 0.8706, "step": 218160 }, { "epoch": 18.89, "learning_rate": 2.7800826876373306e-05, "loss": 0.8404, "step": 218170 }, { "epoch": 18.9, "learning_rate": 2.779865849427547e-05, "loss": 0.6983, "step": 218180 }, { "epoch": 18.9, "learning_rate": 2.779649011217763e-05, "loss": 0.979, "step": 218190 }, { "epoch": 18.9, "learning_rate": 2.779432173007979e-05, "loss": 0.7264, "step": 218200 }, { "epoch": 18.9, "learning_rate": 2.7792153347981958e-05, "loss": 0.8073, "step": 218210 }, { "epoch": 18.9, "learning_rate": 2.778998496588412e-05, "loss": 0.884, "step": 218220 }, { "epoch": 18.9, "learning_rate": 2.7787816583786284e-05, "loss": 0.9572, "step": 218230 }, { "epoch": 18.9, "learning_rate": 2.7785648201688443e-05, "loss": 0.7794, "step": 218240 }, { "epoch": 18.9, "learning_rate": 2.7783479819590606e-05, "loss": 0.8816, "step": 218250 }, { "epoch": 18.9, "learning_rate": 2.778131143749277e-05, "loss": 0.8005, "step": 218260 }, { "epoch": 18.9, "learning_rate": 2.7779143055394936e-05, "loss": 0.9576, "step": 218270 }, { "epoch": 18.9, "learning_rate": 2.7776974673297092e-05, "loss": 0.7067, "step": 218280 }, { "epoch": 18.91, "learning_rate": 2.7774806291199258e-05, "loss": 0.7611, "step": 218290 }, { "epoch": 18.91, "learning_rate": 2.777263790910142e-05, "loss": 0.6866, "step": 218300 }, { "epoch": 18.91, "learning_rate": 2.7770469527003584e-05, "loss": 0.8113, "step": 218310 }, { "epoch": 18.91, "learning_rate": 2.7768301144905744e-05, "loss": 0.7566, "step": 218320 }, { "epoch": 18.91, "learning_rate": 2.7766132762807907e-05, "loss": 0.8853, "step": 218330 }, { "epoch": 18.91, "learning_rate": 2.776396438071007e-05, "loss": 0.9265, "step": 218340 }, { "epoch": 18.91, "learning_rate": 2.776179599861223e-05, "loss": 0.9197, "step": 218350 }, { "epoch": 18.91, "learning_rate": 2.7759627616514396e-05, "loss": 0.7979, "step": 218360 }, { "epoch": 18.91, "learning_rate": 2.775745923441656e-05, "loss": 0.8359, "step": 218370 }, { "epoch": 18.91, "learning_rate": 2.7755290852318722e-05, "loss": 0.7834, "step": 218380 }, { "epoch": 18.91, "learning_rate": 2.775312247022088e-05, "loss": 0.7255, "step": 218390 }, { "epoch": 18.91, "learning_rate": 2.7750954088123044e-05, "loss": 0.7815, "step": 218400 }, { "epoch": 18.92, "learning_rate": 2.7748785706025207e-05, "loss": 0.7693, "step": 218410 }, { "epoch": 18.92, "learning_rate": 2.7746617323927374e-05, "loss": 0.786, "step": 218420 }, { "epoch": 18.92, "learning_rate": 2.7744448941829534e-05, "loss": 0.8572, "step": 218430 }, { "epoch": 18.92, "learning_rate": 2.7742280559731697e-05, "loss": 0.8453, "step": 218440 }, { "epoch": 18.92, "learning_rate": 2.774011217763386e-05, "loss": 0.8106, "step": 218450 }, { "epoch": 18.92, "learning_rate": 2.7737943795536023e-05, "loss": 0.8515, "step": 218460 }, { "epoch": 18.92, "learning_rate": 2.7735775413438182e-05, "loss": 0.8253, "step": 218470 }, { "epoch": 18.92, "learning_rate": 2.7733607031340345e-05, "loss": 0.6487, "step": 218480 }, { "epoch": 18.92, "learning_rate": 2.773143864924251e-05, "loss": 0.7679, "step": 218490 }, { "epoch": 18.92, "learning_rate": 2.7729270267144675e-05, "loss": 0.8608, "step": 218500 }, { "epoch": 18.92, "learning_rate": 2.7727101885046834e-05, "loss": 0.8682, "step": 218510 }, { "epoch": 18.93, "learning_rate": 2.7724933502948997e-05, "loss": 0.768, "step": 218520 }, { "epoch": 18.93, "learning_rate": 2.772276512085116e-05, "loss": 0.6765, "step": 218530 }, { "epoch": 18.93, "learning_rate": 2.7720596738753323e-05, "loss": 0.8175, "step": 218540 }, { "epoch": 18.93, "learning_rate": 2.7718428356655483e-05, "loss": 0.7732, "step": 218550 }, { "epoch": 18.93, "learning_rate": 2.7716259974557646e-05, "loss": 0.7855, "step": 218560 }, { "epoch": 18.93, "learning_rate": 2.7714091592459812e-05, "loss": 0.7869, "step": 218570 }, { "epoch": 18.93, "learning_rate": 2.7711923210361975e-05, "loss": 0.8866, "step": 218580 }, { "epoch": 18.93, "learning_rate": 2.7709754828264135e-05, "loss": 0.8972, "step": 218590 }, { "epoch": 18.93, "learning_rate": 2.7707586446166298e-05, "loss": 0.8098, "step": 218600 }, { "epoch": 18.93, "learning_rate": 2.770541806406846e-05, "loss": 0.68, "step": 218610 }, { "epoch": 18.93, "learning_rate": 2.770324968197062e-05, "loss": 0.8978, "step": 218620 }, { "epoch": 18.93, "learning_rate": 2.7701081299872783e-05, "loss": 0.8529, "step": 218630 }, { "epoch": 18.94, "learning_rate": 2.769891291777495e-05, "loss": 0.8769, "step": 218640 }, { "epoch": 18.94, "learning_rate": 2.7696744535677113e-05, "loss": 0.8616, "step": 218650 }, { "epoch": 18.94, "learning_rate": 2.7694576153579272e-05, "loss": 0.7574, "step": 218660 }, { "epoch": 18.94, "learning_rate": 2.7692407771481435e-05, "loss": 0.7922, "step": 218670 }, { "epoch": 18.94, "learning_rate": 2.76902393893836e-05, "loss": 0.8378, "step": 218680 }, { "epoch": 18.94, "learning_rate": 2.768807100728576e-05, "loss": 0.8257, "step": 218690 }, { "epoch": 18.94, "learning_rate": 2.768590262518792e-05, "loss": 0.7773, "step": 218700 }, { "epoch": 18.94, "learning_rate": 2.7683734243090087e-05, "loss": 0.8033, "step": 218710 }, { "epoch": 18.94, "learning_rate": 2.768156586099225e-05, "loss": 1.0185, "step": 218720 }, { "epoch": 18.94, "learning_rate": 2.7679397478894413e-05, "loss": 0.814, "step": 218730 }, { "epoch": 18.94, "learning_rate": 2.7677229096796573e-05, "loss": 0.6698, "step": 218740 }, { "epoch": 18.95, "learning_rate": 2.7675060714698736e-05, "loss": 0.6597, "step": 218750 }, { "epoch": 18.95, "learning_rate": 2.76728923326009e-05, "loss": 0.7525, "step": 218760 }, { "epoch": 18.95, "learning_rate": 2.7670723950503065e-05, "loss": 0.8802, "step": 218770 }, { "epoch": 18.95, "learning_rate": 2.7668555568405225e-05, "loss": 0.7103, "step": 218780 }, { "epoch": 18.95, "learning_rate": 2.7666387186307388e-05, "loss": 0.8371, "step": 218790 }, { "epoch": 18.95, "learning_rate": 2.766421880420955e-05, "loss": 0.9038, "step": 218800 }, { "epoch": 18.95, "learning_rate": 2.7662050422111714e-05, "loss": 0.8681, "step": 218810 }, { "epoch": 18.95, "learning_rate": 2.7659882040013874e-05, "loss": 0.7585, "step": 218820 }, { "epoch": 18.95, "learning_rate": 2.7657713657916037e-05, "loss": 0.8268, "step": 218830 }, { "epoch": 18.95, "learning_rate": 2.76555452758182e-05, "loss": 0.8145, "step": 218840 }, { "epoch": 18.95, "learning_rate": 2.7653376893720366e-05, "loss": 0.9162, "step": 218850 }, { "epoch": 18.95, "learning_rate": 2.7651208511622526e-05, "loss": 0.9566, "step": 218860 }, { "epoch": 18.96, "learning_rate": 2.764904012952469e-05, "loss": 0.7481, "step": 218870 }, { "epoch": 18.96, "learning_rate": 2.7646871747426852e-05, "loss": 0.8067, "step": 218880 }, { "epoch": 18.96, "learning_rate": 2.764470336532901e-05, "loss": 0.8632, "step": 218890 }, { "epoch": 18.96, "learning_rate": 2.7642534983231174e-05, "loss": 0.9657, "step": 218900 }, { "epoch": 18.96, "learning_rate": 2.7640366601133337e-05, "loss": 0.8918, "step": 218910 }, { "epoch": 18.96, "learning_rate": 2.7638198219035504e-05, "loss": 0.8538, "step": 218920 }, { "epoch": 18.96, "learning_rate": 2.7636029836937663e-05, "loss": 0.665, "step": 218930 }, { "epoch": 18.96, "learning_rate": 2.7633861454839826e-05, "loss": 0.8172, "step": 218940 }, { "epoch": 18.96, "learning_rate": 2.763169307274199e-05, "loss": 0.8172, "step": 218950 }, { "epoch": 18.96, "learning_rate": 2.7629524690644152e-05, "loss": 0.7027, "step": 218960 }, { "epoch": 18.96, "learning_rate": 2.7627356308546312e-05, "loss": 0.8208, "step": 218970 }, { "epoch": 18.97, "learning_rate": 2.7625187926448475e-05, "loss": 0.6438, "step": 218980 }, { "epoch": 18.97, "learning_rate": 2.762301954435064e-05, "loss": 0.8115, "step": 218990 }, { "epoch": 18.97, "learning_rate": 2.7620851162252804e-05, "loss": 0.7768, "step": 219000 }, { "epoch": 18.97, "learning_rate": 2.7618682780154964e-05, "loss": 0.6722, "step": 219010 }, { "epoch": 18.97, "learning_rate": 2.7616514398057127e-05, "loss": 0.6486, "step": 219020 }, { "epoch": 18.97, "learning_rate": 2.761434601595929e-05, "loss": 0.8369, "step": 219030 }, { "epoch": 18.97, "learning_rate": 2.7612177633861453e-05, "loss": 0.8727, "step": 219040 }, { "epoch": 18.97, "learning_rate": 2.7610009251763613e-05, "loss": 0.8164, "step": 219050 }, { "epoch": 18.97, "learning_rate": 2.760784086966578e-05, "loss": 0.8591, "step": 219060 }, { "epoch": 18.97, "learning_rate": 2.7605672487567942e-05, "loss": 0.7103, "step": 219070 }, { "epoch": 18.97, "learning_rate": 2.7603504105470105e-05, "loss": 0.6391, "step": 219080 }, { "epoch": 18.97, "learning_rate": 2.7601335723372265e-05, "loss": 0.6617, "step": 219090 }, { "epoch": 18.98, "learning_rate": 2.7599167341274428e-05, "loss": 0.8008, "step": 219100 }, { "epoch": 18.98, "learning_rate": 2.759699895917659e-05, "loss": 0.7313, "step": 219110 }, { "epoch": 18.98, "learning_rate": 2.759483057707875e-05, "loss": 0.8195, "step": 219120 }, { "epoch": 18.98, "learning_rate": 2.7592662194980913e-05, "loss": 0.8005, "step": 219130 }, { "epoch": 18.98, "learning_rate": 2.759049381288308e-05, "loss": 0.7695, "step": 219140 }, { "epoch": 18.98, "learning_rate": 2.7588325430785243e-05, "loss": 0.665, "step": 219150 }, { "epoch": 18.98, "learning_rate": 2.7586157048687402e-05, "loss": 0.8189, "step": 219160 }, { "epoch": 18.98, "learning_rate": 2.7583988666589565e-05, "loss": 0.8842, "step": 219170 }, { "epoch": 18.98, "learning_rate": 2.7581820284491728e-05, "loss": 0.6466, "step": 219180 }, { "epoch": 18.98, "learning_rate": 2.757965190239389e-05, "loss": 0.8912, "step": 219190 }, { "epoch": 18.98, "learning_rate": 2.757748352029605e-05, "loss": 0.7319, "step": 219200 }, { "epoch": 18.98, "learning_rate": 2.7575315138198217e-05, "loss": 0.6429, "step": 219210 }, { "epoch": 18.99, "learning_rate": 2.757314675610038e-05, "loss": 0.7797, "step": 219220 }, { "epoch": 18.99, "learning_rate": 2.7570978374002543e-05, "loss": 0.8326, "step": 219230 }, { "epoch": 18.99, "learning_rate": 2.7568809991904703e-05, "loss": 0.759, "step": 219240 }, { "epoch": 18.99, "learning_rate": 2.7566641609806866e-05, "loss": 0.9339, "step": 219250 }, { "epoch": 18.99, "learning_rate": 2.756447322770903e-05, "loss": 0.7843, "step": 219260 }, { "epoch": 18.99, "learning_rate": 2.7562304845611195e-05, "loss": 0.7865, "step": 219270 }, { "epoch": 18.99, "learning_rate": 2.7560136463513355e-05, "loss": 0.8789, "step": 219280 }, { "epoch": 18.99, "learning_rate": 2.7557968081415518e-05, "loss": 0.8598, "step": 219290 }, { "epoch": 18.99, "learning_rate": 2.755579969931768e-05, "loss": 0.8939, "step": 219300 }, { "epoch": 18.99, "learning_rate": 2.7553631317219844e-05, "loss": 0.779, "step": 219310 }, { "epoch": 18.99, "learning_rate": 2.7551462935122004e-05, "loss": 0.9998, "step": 219320 }, { "epoch": 19.0, "learning_rate": 2.7549294553024167e-05, "loss": 0.7823, "step": 219330 }, { "epoch": 19.0, "learning_rate": 2.7547126170926333e-05, "loss": 0.6947, "step": 219340 }, { "epoch": 19.0, "learning_rate": 2.7544957788828496e-05, "loss": 0.8079, "step": 219350 }, { "epoch": 19.0, "learning_rate": 2.7542789406730656e-05, "loss": 0.9623, "step": 219360 }, { "epoch": 19.0, "learning_rate": 2.754062102463282e-05, "loss": 0.8787, "step": 219370 }, { "epoch": 19.0, "learning_rate": 2.753845264253498e-05, "loss": 0.7138, "step": 219380 }, { "epoch": 19.0, "eval_cer": 0.542362442441725, "eval_em": 0.05466735430634347, "eval_f1": 0.05466735430634347, "eval_loss": 1.5116215944290161, "eval_runtime": 254.6804, "eval_samples_per_second": 7.613, "eval_steps_per_second": 0.954, "eval_wer": 0.9453326456936565, "step": 219383 }, { "epoch": 19.0, "learning_rate": 2.753628426043714e-05, "loss": 0.884, "step": 219390 }, { "epoch": 19.0, "learning_rate": 2.7534115878339304e-05, "loss": 0.6448, "step": 219400 }, { "epoch": 19.0, "learning_rate": 2.7531947496241467e-05, "loss": 0.6723, "step": 219410 }, { "epoch": 19.0, "learning_rate": 2.7529779114143634e-05, "loss": 0.7461, "step": 219420 }, { "epoch": 19.0, "learning_rate": 2.7527610732045793e-05, "loss": 0.7063, "step": 219430 }, { "epoch": 19.0, "learning_rate": 2.7525442349947956e-05, "loss": 0.7623, "step": 219440 }, { "epoch": 19.01, "learning_rate": 2.752327396785012e-05, "loss": 0.7958, "step": 219450 }, { "epoch": 19.01, "learning_rate": 2.7521105585752282e-05, "loss": 0.7832, "step": 219460 }, { "epoch": 19.01, "learning_rate": 2.7518937203654442e-05, "loss": 0.778, "step": 219470 }, { "epoch": 19.01, "learning_rate": 2.7516768821556605e-05, "loss": 0.6821, "step": 219480 }, { "epoch": 19.01, "learning_rate": 2.751460043945877e-05, "loss": 0.7586, "step": 219490 }, { "epoch": 19.01, "learning_rate": 2.7512432057360934e-05, "loss": 0.6462, "step": 219500 }, { "epoch": 19.01, "learning_rate": 2.7510263675263094e-05, "loss": 0.7548, "step": 219510 }, { "epoch": 19.01, "learning_rate": 2.7508095293165257e-05, "loss": 0.6538, "step": 219520 }, { "epoch": 19.01, "learning_rate": 2.750592691106742e-05, "loss": 0.7209, "step": 219530 }, { "epoch": 19.01, "learning_rate": 2.7503758528969583e-05, "loss": 0.8646, "step": 219540 }, { "epoch": 19.01, "learning_rate": 2.7501590146871742e-05, "loss": 0.6852, "step": 219550 }, { "epoch": 19.02, "learning_rate": 2.749942176477391e-05, "loss": 0.7452, "step": 219560 }, { "epoch": 19.02, "learning_rate": 2.7497253382676072e-05, "loss": 0.7171, "step": 219570 }, { "epoch": 19.02, "learning_rate": 2.7495085000578235e-05, "loss": 0.7236, "step": 219580 }, { "epoch": 19.02, "learning_rate": 2.7492916618480394e-05, "loss": 0.846, "step": 219590 }, { "epoch": 19.02, "learning_rate": 2.7490748236382557e-05, "loss": 0.6372, "step": 219600 }, { "epoch": 19.02, "learning_rate": 2.748857985428472e-05, "loss": 0.7899, "step": 219610 }, { "epoch": 19.02, "learning_rate": 2.7486411472186887e-05, "loss": 0.7495, "step": 219620 }, { "epoch": 19.02, "learning_rate": 2.7484243090089043e-05, "loss": 0.8825, "step": 219630 }, { "epoch": 19.02, "learning_rate": 2.748207470799121e-05, "loss": 0.5913, "step": 219640 }, { "epoch": 19.02, "learning_rate": 2.7479906325893372e-05, "loss": 0.87, "step": 219650 }, { "epoch": 19.02, "learning_rate": 2.7477737943795532e-05, "loss": 0.7317, "step": 219660 }, { "epoch": 19.02, "learning_rate": 2.7475569561697695e-05, "loss": 0.7878, "step": 219670 }, { "epoch": 19.03, "learning_rate": 2.7473401179599858e-05, "loss": 0.8059, "step": 219680 }, { "epoch": 19.03, "learning_rate": 2.747123279750202e-05, "loss": 0.6171, "step": 219690 }, { "epoch": 19.03, "learning_rate": 2.746906441540418e-05, "loss": 0.6099, "step": 219700 }, { "epoch": 19.03, "learning_rate": 2.7466896033306347e-05, "loss": 0.7973, "step": 219710 }, { "epoch": 19.03, "learning_rate": 2.746472765120851e-05, "loss": 0.7674, "step": 219720 }, { "epoch": 19.03, "learning_rate": 2.7462559269110673e-05, "loss": 0.748, "step": 219730 }, { "epoch": 19.03, "learning_rate": 2.7460390887012833e-05, "loss": 0.7391, "step": 219740 }, { "epoch": 19.03, "learning_rate": 2.7458222504914996e-05, "loss": 0.7002, "step": 219750 }, { "epoch": 19.03, "learning_rate": 2.745605412281716e-05, "loss": 0.6392, "step": 219760 }, { "epoch": 19.03, "learning_rate": 2.7453885740719325e-05, "loss": 0.7805, "step": 219770 }, { "epoch": 19.03, "learning_rate": 2.7451717358621485e-05, "loss": 0.779, "step": 219780 }, { "epoch": 19.04, "learning_rate": 2.7449548976523648e-05, "loss": 0.6109, "step": 219790 }, { "epoch": 19.04, "learning_rate": 2.744738059442581e-05, "loss": 0.6581, "step": 219800 }, { "epoch": 19.04, "learning_rate": 2.7445212212327974e-05, "loss": 0.6958, "step": 219810 }, { "epoch": 19.04, "learning_rate": 2.7443043830230133e-05, "loss": 0.7987, "step": 219820 }, { "epoch": 19.04, "learning_rate": 2.7440875448132296e-05, "loss": 0.8087, "step": 219830 }, { "epoch": 19.04, "learning_rate": 2.7438707066034463e-05, "loss": 0.7985, "step": 219840 }, { "epoch": 19.04, "learning_rate": 2.7436538683936626e-05, "loss": 0.6979, "step": 219850 }, { "epoch": 19.04, "learning_rate": 2.7434370301838785e-05, "loss": 0.678, "step": 219860 }, { "epoch": 19.04, "learning_rate": 2.743220191974095e-05, "loss": 0.7421, "step": 219870 }, { "epoch": 19.04, "learning_rate": 2.743003353764311e-05, "loss": 0.9031, "step": 219880 }, { "epoch": 19.04, "learning_rate": 2.742786515554527e-05, "loss": 0.7093, "step": 219890 }, { "epoch": 19.04, "learning_rate": 2.7425696773447434e-05, "loss": 0.6202, "step": 219900 }, { "epoch": 19.05, "learning_rate": 2.7423528391349597e-05, "loss": 0.6608, "step": 219910 }, { "epoch": 19.05, "learning_rate": 2.7421360009251763e-05, "loss": 0.7819, "step": 219920 }, { "epoch": 19.05, "learning_rate": 2.7419191627153923e-05, "loss": 0.6975, "step": 219930 }, { "epoch": 19.05, "learning_rate": 2.7417023245056086e-05, "loss": 0.8261, "step": 219940 }, { "epoch": 19.05, "learning_rate": 2.741485486295825e-05, "loss": 0.853, "step": 219950 }, { "epoch": 19.05, "learning_rate": 2.7412686480860412e-05, "loss": 0.7054, "step": 219960 }, { "epoch": 19.05, "learning_rate": 2.741051809876257e-05, "loss": 0.7827, "step": 219970 }, { "epoch": 19.05, "learning_rate": 2.7408349716664735e-05, "loss": 0.9466, "step": 219980 }, { "epoch": 19.05, "learning_rate": 2.74061813345669e-05, "loss": 0.7086, "step": 219990 }, { "epoch": 19.05, "learning_rate": 2.7404012952469064e-05, "loss": 0.803, "step": 220000 }, { "epoch": 19.05, "learning_rate": 2.7401844570371224e-05, "loss": 0.8166, "step": 220010 }, { "epoch": 19.06, "learning_rate": 2.7399676188273387e-05, "loss": 0.6041, "step": 220020 }, { "epoch": 19.06, "learning_rate": 2.739750780617555e-05, "loss": 0.7957, "step": 220030 }, { "epoch": 19.06, "learning_rate": 2.7395339424077713e-05, "loss": 0.6516, "step": 220040 }, { "epoch": 19.06, "learning_rate": 2.7393171041979872e-05, "loss": 0.6842, "step": 220050 }, { "epoch": 19.06, "learning_rate": 2.739100265988204e-05, "loss": 0.6423, "step": 220060 }, { "epoch": 19.06, "learning_rate": 2.7388834277784202e-05, "loss": 0.7254, "step": 220070 }, { "epoch": 19.06, "learning_rate": 2.7386665895686365e-05, "loss": 0.7854, "step": 220080 }, { "epoch": 19.06, "learning_rate": 2.7384497513588524e-05, "loss": 0.6568, "step": 220090 }, { "epoch": 19.06, "learning_rate": 2.7382329131490687e-05, "loss": 0.6968, "step": 220100 }, { "epoch": 19.06, "learning_rate": 2.738016074939285e-05, "loss": 0.7733, "step": 220110 }, { "epoch": 19.06, "learning_rate": 2.7377992367295017e-05, "loss": 0.6888, "step": 220120 }, { "epoch": 19.06, "learning_rate": 2.7375823985197176e-05, "loss": 0.7041, "step": 220130 }, { "epoch": 19.07, "learning_rate": 2.737365560309934e-05, "loss": 0.795, "step": 220140 }, { "epoch": 19.07, "learning_rate": 2.7371487221001502e-05, "loss": 0.8179, "step": 220150 }, { "epoch": 19.07, "learning_rate": 2.7369318838903662e-05, "loss": 0.648, "step": 220160 }, { "epoch": 19.07, "learning_rate": 2.7367150456805825e-05, "loss": 0.7339, "step": 220170 }, { "epoch": 19.07, "learning_rate": 2.7364982074707988e-05, "loss": 0.8047, "step": 220180 }, { "epoch": 19.07, "learning_rate": 2.736281369261015e-05, "loss": 0.7629, "step": 220190 }, { "epoch": 19.07, "learning_rate": 2.736064531051231e-05, "loss": 0.6505, "step": 220200 }, { "epoch": 19.07, "learning_rate": 2.7358476928414477e-05, "loss": 0.665, "step": 220210 }, { "epoch": 19.07, "learning_rate": 2.735630854631664e-05, "loss": 0.6401, "step": 220220 }, { "epoch": 19.07, "learning_rate": 2.7354140164218803e-05, "loss": 0.6515, "step": 220230 }, { "epoch": 19.07, "learning_rate": 2.7351971782120963e-05, "loss": 0.7875, "step": 220240 }, { "epoch": 19.08, "learning_rate": 2.7349803400023126e-05, "loss": 0.7676, "step": 220250 }, { "epoch": 19.08, "learning_rate": 2.734763501792529e-05, "loss": 0.63, "step": 220260 }, { "epoch": 19.08, "learning_rate": 2.7345466635827455e-05, "loss": 0.9473, "step": 220270 }, { "epoch": 19.08, "learning_rate": 2.7343298253729615e-05, "loss": 0.7398, "step": 220280 }, { "epoch": 19.08, "learning_rate": 2.7341129871631778e-05, "loss": 0.8249, "step": 220290 }, { "epoch": 19.08, "learning_rate": 2.733896148953394e-05, "loss": 0.7618, "step": 220300 }, { "epoch": 19.08, "learning_rate": 2.7336793107436104e-05, "loss": 0.7661, "step": 220310 }, { "epoch": 19.08, "learning_rate": 2.7334624725338263e-05, "loss": 0.8802, "step": 220320 }, { "epoch": 19.08, "learning_rate": 2.7332456343240426e-05, "loss": 0.9163, "step": 220330 }, { "epoch": 19.08, "learning_rate": 2.7330287961142593e-05, "loss": 0.7508, "step": 220340 }, { "epoch": 19.08, "learning_rate": 2.7328119579044756e-05, "loss": 0.7486, "step": 220350 }, { "epoch": 19.08, "learning_rate": 2.7325951196946915e-05, "loss": 0.7271, "step": 220360 }, { "epoch": 19.09, "learning_rate": 2.7323782814849078e-05, "loss": 0.8004, "step": 220370 }, { "epoch": 19.09, "learning_rate": 2.732161443275124e-05, "loss": 0.7385, "step": 220380 }, { "epoch": 19.09, "learning_rate": 2.7319446050653404e-05, "loss": 0.6934, "step": 220390 }, { "epoch": 19.09, "learning_rate": 2.7317277668555564e-05, "loss": 0.8858, "step": 220400 }, { "epoch": 19.09, "learning_rate": 2.731510928645773e-05, "loss": 0.6911, "step": 220410 }, { "epoch": 19.09, "learning_rate": 2.7312940904359893e-05, "loss": 0.8505, "step": 220420 }, { "epoch": 19.09, "learning_rate": 2.7310772522262053e-05, "loss": 0.771, "step": 220430 }, { "epoch": 19.09, "learning_rate": 2.7308604140164216e-05, "loss": 0.8859, "step": 220440 }, { "epoch": 19.09, "learning_rate": 2.730643575806638e-05, "loss": 0.7103, "step": 220450 }, { "epoch": 19.09, "learning_rate": 2.7304267375968542e-05, "loss": 0.6822, "step": 220460 }, { "epoch": 19.09, "learning_rate": 2.73020989938707e-05, "loss": 0.9773, "step": 220470 }, { "epoch": 19.09, "learning_rate": 2.7299930611772865e-05, "loss": 0.7088, "step": 220480 }, { "epoch": 19.1, "learning_rate": 2.729776222967503e-05, "loss": 0.8619, "step": 220490 }, { "epoch": 19.1, "learning_rate": 2.7295593847577194e-05, "loss": 0.7869, "step": 220500 }, { "epoch": 19.1, "learning_rate": 2.7293425465479354e-05, "loss": 0.7631, "step": 220510 }, { "epoch": 19.1, "learning_rate": 2.7291257083381517e-05, "loss": 0.6762, "step": 220520 }, { "epoch": 19.1, "learning_rate": 2.728908870128368e-05, "loss": 0.5959, "step": 220530 }, { "epoch": 19.1, "learning_rate": 2.7286920319185843e-05, "loss": 0.6949, "step": 220540 }, { "epoch": 19.1, "learning_rate": 2.7284751937088002e-05, "loss": 0.7267, "step": 220550 }, { "epoch": 19.1, "learning_rate": 2.728258355499017e-05, "loss": 0.7553, "step": 220560 }, { "epoch": 19.1, "learning_rate": 2.728041517289233e-05, "loss": 0.8402, "step": 220570 }, { "epoch": 19.1, "learning_rate": 2.7278246790794495e-05, "loss": 0.6905, "step": 220580 }, { "epoch": 19.1, "learning_rate": 2.7276078408696654e-05, "loss": 0.7554, "step": 220590 }, { "epoch": 19.11, "learning_rate": 2.7273910026598817e-05, "loss": 0.7489, "step": 220600 }, { "epoch": 19.11, "learning_rate": 2.727174164450098e-05, "loss": 0.7282, "step": 220610 }, { "epoch": 19.11, "learning_rate": 2.7269573262403147e-05, "loss": 0.7565, "step": 220620 }, { "epoch": 19.11, "learning_rate": 2.7267404880305306e-05, "loss": 0.6897, "step": 220630 }, { "epoch": 19.11, "learning_rate": 2.726523649820747e-05, "loss": 0.8371, "step": 220640 }, { "epoch": 19.11, "learning_rate": 2.7263068116109632e-05, "loss": 0.7232, "step": 220650 }, { "epoch": 19.11, "learning_rate": 2.7260899734011792e-05, "loss": 0.6797, "step": 220660 }, { "epoch": 19.11, "learning_rate": 2.7258731351913955e-05, "loss": 0.7268, "step": 220670 }, { "epoch": 19.11, "learning_rate": 2.7256562969816118e-05, "loss": 0.8465, "step": 220680 }, { "epoch": 19.11, "learning_rate": 2.7254394587718284e-05, "loss": 0.7753, "step": 220690 }, { "epoch": 19.11, "learning_rate": 2.725222620562044e-05, "loss": 0.7837, "step": 220700 }, { "epoch": 19.11, "learning_rate": 2.7250057823522607e-05, "loss": 0.6093, "step": 220710 }, { "epoch": 19.12, "learning_rate": 2.724788944142477e-05, "loss": 0.7141, "step": 220720 }, { "epoch": 19.12, "learning_rate": 2.7245721059326933e-05, "loss": 0.734, "step": 220730 }, { "epoch": 19.12, "learning_rate": 2.7243552677229092e-05, "loss": 0.8544, "step": 220740 }, { "epoch": 19.12, "learning_rate": 2.7241384295131255e-05, "loss": 0.8262, "step": 220750 }, { "epoch": 19.12, "learning_rate": 2.723921591303342e-05, "loss": 0.8583, "step": 220760 }, { "epoch": 19.12, "learning_rate": 2.7237047530935585e-05, "loss": 0.7916, "step": 220770 }, { "epoch": 19.12, "learning_rate": 2.7234879148837744e-05, "loss": 0.6287, "step": 220780 }, { "epoch": 19.12, "learning_rate": 2.7232710766739907e-05, "loss": 0.7372, "step": 220790 }, { "epoch": 19.12, "learning_rate": 2.723054238464207e-05, "loss": 0.7401, "step": 220800 }, { "epoch": 19.12, "learning_rate": 2.7228374002544233e-05, "loss": 0.7831, "step": 220810 }, { "epoch": 19.12, "learning_rate": 2.7226205620446393e-05, "loss": 0.7804, "step": 220820 }, { "epoch": 19.13, "learning_rate": 2.7224037238348556e-05, "loss": 0.8229, "step": 220830 }, { "epoch": 19.13, "learning_rate": 2.7221868856250722e-05, "loss": 0.7547, "step": 220840 }, { "epoch": 19.13, "learning_rate": 2.7219700474152885e-05, "loss": 0.6442, "step": 220850 }, { "epoch": 19.13, "learning_rate": 2.7217532092055045e-05, "loss": 0.7976, "step": 220860 }, { "epoch": 19.13, "learning_rate": 2.7215363709957208e-05, "loss": 0.7042, "step": 220870 }, { "epoch": 19.13, "learning_rate": 2.721319532785937e-05, "loss": 0.6207, "step": 220880 }, { "epoch": 19.13, "learning_rate": 2.7211026945761534e-05, "loss": 0.7423, "step": 220890 }, { "epoch": 19.13, "learning_rate": 2.7208858563663694e-05, "loss": 0.8518, "step": 220900 }, { "epoch": 19.13, "learning_rate": 2.720669018156586e-05, "loss": 0.7799, "step": 220910 }, { "epoch": 19.13, "learning_rate": 2.7204521799468023e-05, "loss": 0.5842, "step": 220920 }, { "epoch": 19.13, "learning_rate": 2.7202353417370183e-05, "loss": 0.891, "step": 220930 }, { "epoch": 19.13, "learning_rate": 2.7200185035272346e-05, "loss": 0.7589, "step": 220940 }, { "epoch": 19.14, "learning_rate": 2.719801665317451e-05, "loss": 0.8221, "step": 220950 }, { "epoch": 19.14, "learning_rate": 2.7195848271076672e-05, "loss": 0.6682, "step": 220960 }, { "epoch": 19.14, "learning_rate": 2.719367988897883e-05, "loss": 0.7723, "step": 220970 }, { "epoch": 19.14, "learning_rate": 2.7191511506880994e-05, "loss": 0.7945, "step": 220980 }, { "epoch": 19.14, "learning_rate": 2.718934312478316e-05, "loss": 0.7301, "step": 220990 }, { "epoch": 19.14, "learning_rate": 2.7187174742685324e-05, "loss": 0.7807, "step": 221000 }, { "epoch": 19.14, "learning_rate": 2.7185006360587483e-05, "loss": 0.623, "step": 221010 }, { "epoch": 19.14, "learning_rate": 2.7182837978489646e-05, "loss": 0.836, "step": 221020 }, { "epoch": 19.14, "learning_rate": 2.718066959639181e-05, "loss": 0.6873, "step": 221030 }, { "epoch": 19.14, "learning_rate": 2.7178501214293972e-05, "loss": 0.7246, "step": 221040 }, { "epoch": 19.14, "learning_rate": 2.7176332832196132e-05, "loss": 0.7348, "step": 221050 }, { "epoch": 19.15, "learning_rate": 2.71741644500983e-05, "loss": 0.7565, "step": 221060 }, { "epoch": 19.15, "learning_rate": 2.717199606800046e-05, "loss": 0.6938, "step": 221070 }, { "epoch": 19.15, "learning_rate": 2.7169827685902624e-05, "loss": 0.6112, "step": 221080 }, { "epoch": 19.15, "learning_rate": 2.7167659303804784e-05, "loss": 0.7464, "step": 221090 }, { "epoch": 19.15, "learning_rate": 2.7165490921706947e-05, "loss": 0.8567, "step": 221100 }, { "epoch": 19.15, "learning_rate": 2.716332253960911e-05, "loss": 0.7279, "step": 221110 }, { "epoch": 19.15, "learning_rate": 2.7161154157511276e-05, "loss": 0.779, "step": 221120 }, { "epoch": 19.15, "learning_rate": 2.7158985775413436e-05, "loss": 0.7543, "step": 221130 }, { "epoch": 19.15, "learning_rate": 2.71568173933156e-05, "loss": 0.8929, "step": 221140 }, { "epoch": 19.15, "learning_rate": 2.7154649011217762e-05, "loss": 0.7903, "step": 221150 }, { "epoch": 19.15, "learning_rate": 2.7152480629119925e-05, "loss": 0.6779, "step": 221160 }, { "epoch": 19.15, "learning_rate": 2.7150312247022085e-05, "loss": 0.7889, "step": 221170 }, { "epoch": 19.16, "learning_rate": 2.7148143864924248e-05, "loss": 0.7981, "step": 221180 }, { "epoch": 19.16, "learning_rate": 2.7145975482826414e-05, "loss": 0.7895, "step": 221190 }, { "epoch": 19.16, "learning_rate": 2.7143807100728574e-05, "loss": 0.8075, "step": 221200 }, { "epoch": 19.16, "learning_rate": 2.7141638718630737e-05, "loss": 0.6416, "step": 221210 }, { "epoch": 19.16, "learning_rate": 2.71394703365329e-05, "loss": 0.8097, "step": 221220 }, { "epoch": 19.16, "learning_rate": 2.7137301954435063e-05, "loss": 0.7276, "step": 221230 }, { "epoch": 19.16, "learning_rate": 2.7135133572337222e-05, "loss": 0.7593, "step": 221240 }, { "epoch": 19.16, "learning_rate": 2.7132965190239385e-05, "loss": 0.715, "step": 221250 }, { "epoch": 19.16, "learning_rate": 2.7130796808141548e-05, "loss": 0.9045, "step": 221260 }, { "epoch": 19.16, "learning_rate": 2.7128628426043715e-05, "loss": 0.6729, "step": 221270 }, { "epoch": 19.16, "learning_rate": 2.7126460043945874e-05, "loss": 0.7631, "step": 221280 }, { "epoch": 19.17, "learning_rate": 2.7124291661848037e-05, "loss": 0.8412, "step": 221290 }, { "epoch": 19.17, "learning_rate": 2.71221232797502e-05, "loss": 0.7658, "step": 221300 }, { "epoch": 19.17, "learning_rate": 2.7119954897652363e-05, "loss": 0.8508, "step": 221310 }, { "epoch": 19.17, "learning_rate": 2.7117786515554523e-05, "loss": 0.7246, "step": 221320 }, { "epoch": 19.17, "learning_rate": 2.7115618133456686e-05, "loss": 0.782, "step": 221330 }, { "epoch": 19.17, "learning_rate": 2.7113449751358852e-05, "loss": 0.7653, "step": 221340 }, { "epoch": 19.17, "learning_rate": 2.7111281369261015e-05, "loss": 0.7644, "step": 221350 }, { "epoch": 19.17, "learning_rate": 2.7109112987163175e-05, "loss": 0.8765, "step": 221360 }, { "epoch": 19.17, "learning_rate": 2.7106944605065338e-05, "loss": 0.9021, "step": 221370 }, { "epoch": 19.17, "learning_rate": 2.71047762229675e-05, "loss": 0.7404, "step": 221380 }, { "epoch": 19.17, "learning_rate": 2.7102607840869664e-05, "loss": 0.7374, "step": 221390 }, { "epoch": 19.17, "learning_rate": 2.7100439458771824e-05, "loss": 0.7114, "step": 221400 }, { "epoch": 19.18, "learning_rate": 2.709827107667399e-05, "loss": 0.7743, "step": 221410 }, { "epoch": 19.18, "learning_rate": 2.7096102694576153e-05, "loss": 0.6619, "step": 221420 }, { "epoch": 19.18, "learning_rate": 2.7093934312478313e-05, "loss": 0.7152, "step": 221430 }, { "epoch": 19.18, "learning_rate": 2.7091765930380476e-05, "loss": 0.7367, "step": 221440 }, { "epoch": 19.18, "learning_rate": 2.708959754828264e-05, "loss": 0.8484, "step": 221450 }, { "epoch": 19.18, "learning_rate": 2.70874291661848e-05, "loss": 1.0065, "step": 221460 }, { "epoch": 19.18, "learning_rate": 2.708526078408696e-05, "loss": 0.73, "step": 221470 }, { "epoch": 19.18, "learning_rate": 2.7083092401989128e-05, "loss": 0.7808, "step": 221480 }, { "epoch": 19.18, "learning_rate": 2.708092401989129e-05, "loss": 0.7667, "step": 221490 }, { "epoch": 19.18, "learning_rate": 2.7078755637793454e-05, "loss": 0.6282, "step": 221500 }, { "epoch": 19.18, "learning_rate": 2.7076587255695613e-05, "loss": 0.792, "step": 221510 }, { "epoch": 19.19, "learning_rate": 2.7074418873597776e-05, "loss": 0.8265, "step": 221520 }, { "epoch": 19.19, "learning_rate": 2.707225049149994e-05, "loss": 0.7896, "step": 221530 }, { "epoch": 19.19, "learning_rate": 2.7070082109402102e-05, "loss": 0.8367, "step": 221540 }, { "epoch": 19.19, "learning_rate": 2.7067913727304262e-05, "loss": 0.8193, "step": 221550 }, { "epoch": 19.19, "learning_rate": 2.7065745345206428e-05, "loss": 0.8943, "step": 221560 }, { "epoch": 19.19, "learning_rate": 2.706357696310859e-05, "loss": 0.7023, "step": 221570 }, { "epoch": 19.19, "learning_rate": 2.7061408581010754e-05, "loss": 0.615, "step": 221580 }, { "epoch": 19.19, "learning_rate": 2.7059240198912914e-05, "loss": 0.7743, "step": 221590 }, { "epoch": 19.19, "learning_rate": 2.7057071816815077e-05, "loss": 0.753, "step": 221600 }, { "epoch": 19.19, "learning_rate": 2.705490343471724e-05, "loss": 0.7524, "step": 221610 }, { "epoch": 19.19, "learning_rate": 2.7052735052619406e-05, "loss": 0.8009, "step": 221620 }, { "epoch": 19.19, "learning_rate": 2.7050566670521566e-05, "loss": 0.8755, "step": 221630 }, { "epoch": 19.2, "learning_rate": 2.704839828842373e-05, "loss": 0.7298, "step": 221640 }, { "epoch": 19.2, "learning_rate": 2.7046229906325892e-05, "loss": 0.7478, "step": 221650 }, { "epoch": 19.2, "learning_rate": 2.7044061524228055e-05, "loss": 0.9109, "step": 221660 }, { "epoch": 19.2, "learning_rate": 2.7041893142130215e-05, "loss": 0.6825, "step": 221670 }, { "epoch": 19.2, "learning_rate": 2.7039724760032378e-05, "loss": 0.8196, "step": 221680 }, { "epoch": 19.2, "learning_rate": 2.7037556377934544e-05, "loss": 0.759, "step": 221690 }, { "epoch": 19.2, "learning_rate": 2.7035387995836704e-05, "loss": 0.7879, "step": 221700 }, { "epoch": 19.2, "learning_rate": 2.7033219613738867e-05, "loss": 0.6232, "step": 221710 }, { "epoch": 19.2, "learning_rate": 2.703105123164103e-05, "loss": 0.7346, "step": 221720 }, { "epoch": 19.2, "learning_rate": 2.7028882849543193e-05, "loss": 0.6959, "step": 221730 }, { "epoch": 19.2, "learning_rate": 2.7026714467445352e-05, "loss": 0.7778, "step": 221740 }, { "epoch": 19.2, "learning_rate": 2.7024546085347515e-05, "loss": 0.8742, "step": 221750 }, { "epoch": 19.21, "learning_rate": 2.702237770324968e-05, "loss": 0.8897, "step": 221760 }, { "epoch": 19.21, "learning_rate": 2.7020209321151845e-05, "loss": 0.7167, "step": 221770 }, { "epoch": 19.21, "learning_rate": 2.7018040939054004e-05, "loss": 0.756, "step": 221780 }, { "epoch": 19.21, "learning_rate": 2.7015872556956167e-05, "loss": 0.6591, "step": 221790 }, { "epoch": 19.21, "learning_rate": 2.701370417485833e-05, "loss": 0.7058, "step": 221800 }, { "epoch": 19.21, "learning_rate": 2.7011535792760493e-05, "loss": 0.8279, "step": 221810 }, { "epoch": 19.21, "learning_rate": 2.7009367410662653e-05, "loss": 0.7526, "step": 221820 }, { "epoch": 19.21, "learning_rate": 2.7007199028564816e-05, "loss": 0.8295, "step": 221830 }, { "epoch": 19.21, "learning_rate": 2.7005030646466982e-05, "loss": 0.8326, "step": 221840 }, { "epoch": 19.21, "learning_rate": 2.7002862264369145e-05, "loss": 0.851, "step": 221850 }, { "epoch": 19.21, "learning_rate": 2.7000693882271305e-05, "loss": 0.7717, "step": 221860 }, { "epoch": 19.22, "learning_rate": 2.6998525500173468e-05, "loss": 0.7584, "step": 221870 }, { "epoch": 19.22, "learning_rate": 2.699635711807563e-05, "loss": 0.7677, "step": 221880 }, { "epoch": 19.22, "learning_rate": 2.6994188735977794e-05, "loss": 1.0478, "step": 221890 }, { "epoch": 19.22, "learning_rate": 2.6992020353879953e-05, "loss": 0.8625, "step": 221900 }, { "epoch": 19.22, "learning_rate": 2.698985197178212e-05, "loss": 0.8689, "step": 221910 }, { "epoch": 19.22, "learning_rate": 2.6987683589684283e-05, "loss": 0.7618, "step": 221920 }, { "epoch": 19.22, "learning_rate": 2.6985515207586446e-05, "loss": 0.6503, "step": 221930 }, { "epoch": 19.22, "learning_rate": 2.6983346825488605e-05, "loss": 0.5811, "step": 221940 }, { "epoch": 19.22, "learning_rate": 2.698117844339077e-05, "loss": 0.7405, "step": 221950 }, { "epoch": 19.22, "learning_rate": 2.697901006129293e-05, "loss": 0.642, "step": 221960 }, { "epoch": 19.22, "learning_rate": 2.697684167919509e-05, "loss": 0.7212, "step": 221970 }, { "epoch": 19.22, "learning_rate": 2.6974673297097257e-05, "loss": 0.8103, "step": 221980 }, { "epoch": 19.23, "learning_rate": 2.697250491499942e-05, "loss": 0.7299, "step": 221990 }, { "epoch": 19.23, "learning_rate": 2.6970336532901583e-05, "loss": 0.8466, "step": 222000 }, { "epoch": 19.23, "learning_rate": 2.6968168150803743e-05, "loss": 0.8133, "step": 222010 }, { "epoch": 19.23, "learning_rate": 2.6965999768705906e-05, "loss": 0.6927, "step": 222020 }, { "epoch": 19.23, "learning_rate": 2.696383138660807e-05, "loss": 0.7447, "step": 222030 }, { "epoch": 19.23, "learning_rate": 2.6961663004510235e-05, "loss": 0.689, "step": 222040 }, { "epoch": 19.23, "learning_rate": 2.6959494622412392e-05, "loss": 0.6145, "step": 222050 }, { "epoch": 19.23, "learning_rate": 2.6957326240314558e-05, "loss": 0.7086, "step": 222060 }, { "epoch": 19.23, "learning_rate": 2.695515785821672e-05, "loss": 0.7749, "step": 222070 }, { "epoch": 19.23, "learning_rate": 2.6952989476118884e-05, "loss": 0.7608, "step": 222080 }, { "epoch": 19.23, "learning_rate": 2.6950821094021044e-05, "loss": 0.6502, "step": 222090 }, { "epoch": 19.24, "learning_rate": 2.6948652711923207e-05, "loss": 0.839, "step": 222100 }, { "epoch": 19.24, "learning_rate": 2.694648432982537e-05, "loss": 0.7731, "step": 222110 }, { "epoch": 19.24, "learning_rate": 2.6944315947727536e-05, "loss": 0.8794, "step": 222120 }, { "epoch": 19.24, "learning_rate": 2.6942147565629696e-05, "loss": 0.7743, "step": 222130 }, { "epoch": 19.24, "learning_rate": 2.693997918353186e-05, "loss": 0.8835, "step": 222140 }, { "epoch": 19.24, "learning_rate": 2.6937810801434022e-05, "loss": 0.8609, "step": 222150 }, { "epoch": 19.24, "learning_rate": 2.6935642419336185e-05, "loss": 0.7922, "step": 222160 }, { "epoch": 19.24, "learning_rate": 2.6933474037238344e-05, "loss": 0.7486, "step": 222170 }, { "epoch": 19.24, "learning_rate": 2.6931305655140507e-05, "loss": 0.6621, "step": 222180 }, { "epoch": 19.24, "learning_rate": 2.6929137273042674e-05, "loss": 0.7967, "step": 222190 }, { "epoch": 19.24, "learning_rate": 2.6926968890944833e-05, "loss": 0.7972, "step": 222200 }, { "epoch": 19.24, "learning_rate": 2.6924800508846996e-05, "loss": 0.7151, "step": 222210 }, { "epoch": 19.25, "learning_rate": 2.692263212674916e-05, "loss": 0.6063, "step": 222220 }, { "epoch": 19.25, "learning_rate": 2.6920463744651322e-05, "loss": 0.7444, "step": 222230 }, { "epoch": 19.25, "learning_rate": 2.6918295362553482e-05, "loss": 0.8896, "step": 222240 }, { "epoch": 19.25, "learning_rate": 2.6916126980455645e-05, "loss": 0.8175, "step": 222250 }, { "epoch": 19.25, "learning_rate": 2.691395859835781e-05, "loss": 0.8255, "step": 222260 }, { "epoch": 19.25, "learning_rate": 2.6911790216259974e-05, "loss": 0.8881, "step": 222270 }, { "epoch": 19.25, "learning_rate": 2.6909621834162134e-05, "loss": 0.7418, "step": 222280 }, { "epoch": 19.25, "learning_rate": 2.6907453452064297e-05, "loss": 0.8857, "step": 222290 }, { "epoch": 19.25, "learning_rate": 2.690528506996646e-05, "loss": 0.8796, "step": 222300 }, { "epoch": 19.25, "learning_rate": 2.6903116687868623e-05, "loss": 0.6386, "step": 222310 }, { "epoch": 19.25, "learning_rate": 2.6900948305770783e-05, "loss": 0.6785, "step": 222320 }, { "epoch": 19.26, "learning_rate": 2.6898779923672946e-05, "loss": 0.8728, "step": 222330 }, { "epoch": 19.26, "learning_rate": 2.6896611541575112e-05, "loss": 0.858, "step": 222340 }, { "epoch": 19.26, "learning_rate": 2.6894443159477275e-05, "loss": 0.7105, "step": 222350 }, { "epoch": 19.26, "learning_rate": 2.6892274777379435e-05, "loss": 0.7719, "step": 222360 }, { "epoch": 19.26, "learning_rate": 2.6890106395281598e-05, "loss": 0.7605, "step": 222370 }, { "epoch": 19.26, "learning_rate": 2.688793801318376e-05, "loss": 0.8115, "step": 222380 }, { "epoch": 19.26, "learning_rate": 2.6885769631085924e-05, "loss": 0.8621, "step": 222390 }, { "epoch": 19.26, "learning_rate": 2.6883601248988083e-05, "loss": 0.6452, "step": 222400 }, { "epoch": 19.26, "learning_rate": 2.688143286689025e-05, "loss": 0.7825, "step": 222410 }, { "epoch": 19.26, "learning_rate": 2.6879264484792413e-05, "loss": 0.6776, "step": 222420 }, { "epoch": 19.26, "learning_rate": 2.6877096102694576e-05, "loss": 0.8854, "step": 222430 }, { "epoch": 19.26, "learning_rate": 2.6874927720596735e-05, "loss": 0.8687, "step": 222440 }, { "epoch": 19.27, "learning_rate": 2.6872759338498898e-05, "loss": 0.8171, "step": 222450 }, { "epoch": 19.27, "learning_rate": 2.687059095640106e-05, "loss": 0.7462, "step": 222460 }, { "epoch": 19.27, "learning_rate": 2.686842257430322e-05, "loss": 0.7854, "step": 222470 }, { "epoch": 19.27, "learning_rate": 2.6866254192205387e-05, "loss": 0.893, "step": 222480 }, { "epoch": 19.27, "learning_rate": 2.686408581010755e-05, "loss": 0.7484, "step": 222490 }, { "epoch": 19.27, "learning_rate": 2.6861917428009713e-05, "loss": 0.8612, "step": 222500 }, { "epoch": 19.27, "learning_rate": 2.6859749045911873e-05, "loss": 0.6235, "step": 222510 }, { "epoch": 19.27, "learning_rate": 2.6857580663814036e-05, "loss": 0.7877, "step": 222520 }, { "epoch": 19.27, "learning_rate": 2.68554122817162e-05, "loss": 0.594, "step": 222530 }, { "epoch": 19.27, "learning_rate": 2.6853243899618365e-05, "loss": 0.8207, "step": 222540 }, { "epoch": 19.27, "learning_rate": 2.6851075517520525e-05, "loss": 0.7176, "step": 222550 }, { "epoch": 19.28, "learning_rate": 2.6848907135422688e-05, "loss": 0.638, "step": 222560 }, { "epoch": 19.28, "learning_rate": 2.684673875332485e-05, "loss": 0.7219, "step": 222570 }, { "epoch": 19.28, "learning_rate": 2.6844570371227014e-05, "loss": 0.824, "step": 222580 }, { "epoch": 19.28, "learning_rate": 2.6842401989129174e-05, "loss": 0.7275, "step": 222590 }, { "epoch": 19.28, "learning_rate": 2.6840233607031337e-05, "loss": 0.8193, "step": 222600 }, { "epoch": 19.28, "learning_rate": 2.68380652249335e-05, "loss": 0.6528, "step": 222610 }, { "epoch": 19.28, "learning_rate": 2.6835896842835666e-05, "loss": 0.6598, "step": 222620 }, { "epoch": 19.28, "learning_rate": 2.6833728460737826e-05, "loss": 0.6751, "step": 222630 }, { "epoch": 19.28, "learning_rate": 2.683156007863999e-05, "loss": 0.7521, "step": 222640 }, { "epoch": 19.28, "learning_rate": 2.682939169654215e-05, "loss": 0.7135, "step": 222650 }, { "epoch": 19.28, "learning_rate": 2.6827223314444315e-05, "loss": 0.6818, "step": 222660 }, { "epoch": 19.28, "learning_rate": 2.6825054932346474e-05, "loss": 0.7472, "step": 222670 }, { "epoch": 19.29, "learning_rate": 2.6822886550248637e-05, "loss": 0.8471, "step": 222680 }, { "epoch": 19.29, "learning_rate": 2.6820718168150804e-05, "loss": 0.7279, "step": 222690 }, { "epoch": 19.29, "learning_rate": 2.6818549786052967e-05, "loss": 0.7991, "step": 222700 }, { "epoch": 19.29, "learning_rate": 2.6816381403955126e-05, "loss": 0.7363, "step": 222710 }, { "epoch": 19.29, "learning_rate": 2.681421302185729e-05, "loss": 0.7432, "step": 222720 }, { "epoch": 19.29, "learning_rate": 2.6812044639759452e-05, "loss": 0.7713, "step": 222730 }, { "epoch": 19.29, "learning_rate": 2.6809876257661612e-05, "loss": 0.9684, "step": 222740 }, { "epoch": 19.29, "learning_rate": 2.6807707875563775e-05, "loss": 0.6851, "step": 222750 }, { "epoch": 19.29, "learning_rate": 2.680553949346594e-05, "loss": 0.6216, "step": 222760 }, { "epoch": 19.29, "learning_rate": 2.6803371111368104e-05, "loss": 0.8819, "step": 222770 }, { "epoch": 19.29, "learning_rate": 2.6801202729270264e-05, "loss": 0.9065, "step": 222780 }, { "epoch": 19.3, "learning_rate": 2.6799034347172427e-05, "loss": 0.6885, "step": 222790 }, { "epoch": 19.3, "learning_rate": 2.679686596507459e-05, "loss": 0.7083, "step": 222800 }, { "epoch": 19.3, "learning_rate": 2.6794697582976753e-05, "loss": 0.8288, "step": 222810 }, { "epoch": 19.3, "learning_rate": 2.6792529200878912e-05, "loss": 0.6211, "step": 222820 }, { "epoch": 19.3, "learning_rate": 2.679036081878108e-05, "loss": 0.6914, "step": 222830 }, { "epoch": 19.3, "learning_rate": 2.6788192436683242e-05, "loss": 0.8003, "step": 222840 }, { "epoch": 19.3, "learning_rate": 2.6786024054585405e-05, "loss": 0.6921, "step": 222850 }, { "epoch": 19.3, "learning_rate": 2.6783855672487564e-05, "loss": 0.6479, "step": 222860 }, { "epoch": 19.3, "learning_rate": 2.6781687290389727e-05, "loss": 0.7851, "step": 222870 }, { "epoch": 19.3, "learning_rate": 2.677951890829189e-05, "loss": 0.8867, "step": 222880 }, { "epoch": 19.3, "learning_rate": 2.6777350526194054e-05, "loss": 0.7734, "step": 222890 }, { "epoch": 19.3, "learning_rate": 2.6775182144096213e-05, "loss": 0.7718, "step": 222900 }, { "epoch": 19.31, "learning_rate": 2.677301376199838e-05, "loss": 0.7316, "step": 222910 }, { "epoch": 19.31, "learning_rate": 2.6770845379900543e-05, "loss": 0.6749, "step": 222920 }, { "epoch": 19.31, "learning_rate": 2.6768676997802706e-05, "loss": 0.8061, "step": 222930 }, { "epoch": 19.31, "learning_rate": 2.6766508615704865e-05, "loss": 0.7421, "step": 222940 }, { "epoch": 19.31, "learning_rate": 2.6764340233607028e-05, "loss": 0.8193, "step": 222950 }, { "epoch": 19.31, "learning_rate": 2.676217185150919e-05, "loss": 0.7542, "step": 222960 }, { "epoch": 19.31, "learning_rate": 2.676000346941135e-05, "loss": 0.6111, "step": 222970 }, { "epoch": 19.31, "learning_rate": 2.6757835087313517e-05, "loss": 0.6281, "step": 222980 }, { "epoch": 19.31, "learning_rate": 2.675566670521568e-05, "loss": 0.777, "step": 222990 }, { "epoch": 19.31, "learning_rate": 2.6753498323117843e-05, "loss": 0.8679, "step": 223000 }, { "epoch": 19.31, "learning_rate": 2.6751329941020003e-05, "loss": 0.7756, "step": 223010 }, { "epoch": 19.31, "learning_rate": 2.6749161558922166e-05, "loss": 0.8418, "step": 223020 }, { "epoch": 19.32, "learning_rate": 2.674699317682433e-05, "loss": 0.7011, "step": 223030 }, { "epoch": 19.32, "learning_rate": 2.6744824794726495e-05, "loss": 0.8432, "step": 223040 }, { "epoch": 19.32, "learning_rate": 2.6742656412628655e-05, "loss": 0.6681, "step": 223050 }, { "epoch": 19.32, "learning_rate": 2.6740488030530818e-05, "loss": 0.6361, "step": 223060 }, { "epoch": 19.32, "learning_rate": 2.673831964843298e-05, "loss": 0.7296, "step": 223070 }, { "epoch": 19.32, "learning_rate": 2.6736151266335144e-05, "loss": 0.7884, "step": 223080 }, { "epoch": 19.32, "learning_rate": 2.6733982884237303e-05, "loss": 0.8593, "step": 223090 }, { "epoch": 19.32, "learning_rate": 2.6731814502139466e-05, "loss": 0.7107, "step": 223100 }, { "epoch": 19.32, "learning_rate": 2.6729646120041633e-05, "loss": 0.7706, "step": 223110 }, { "epoch": 19.32, "learning_rate": 2.6727477737943796e-05, "loss": 0.7984, "step": 223120 }, { "epoch": 19.32, "learning_rate": 2.6725309355845955e-05, "loss": 0.6906, "step": 223130 }, { "epoch": 19.33, "learning_rate": 2.672314097374812e-05, "loss": 0.6826, "step": 223140 }, { "epoch": 19.33, "learning_rate": 2.672097259165028e-05, "loss": 0.6364, "step": 223150 }, { "epoch": 19.33, "learning_rate": 2.6718804209552444e-05, "loss": 0.782, "step": 223160 }, { "epoch": 19.33, "learning_rate": 2.6716635827454604e-05, "loss": 0.7924, "step": 223170 }, { "epoch": 19.33, "learning_rate": 2.6714467445356767e-05, "loss": 0.7948, "step": 223180 }, { "epoch": 19.33, "learning_rate": 2.6712299063258933e-05, "loss": 0.7398, "step": 223190 }, { "epoch": 19.33, "learning_rate": 2.6710130681161096e-05, "loss": 0.9397, "step": 223200 }, { "epoch": 19.33, "learning_rate": 2.6707962299063256e-05, "loss": 0.7588, "step": 223210 }, { "epoch": 19.33, "learning_rate": 2.670579391696542e-05, "loss": 0.9368, "step": 223220 }, { "epoch": 19.33, "learning_rate": 2.6703625534867582e-05, "loss": 0.7635, "step": 223230 }, { "epoch": 19.33, "learning_rate": 2.670145715276974e-05, "loss": 0.554, "step": 223240 }, { "epoch": 19.33, "learning_rate": 2.6699288770671905e-05, "loss": 0.7843, "step": 223250 }, { "epoch": 19.34, "learning_rate": 2.669712038857407e-05, "loss": 0.6356, "step": 223260 }, { "epoch": 19.34, "learning_rate": 2.6694952006476234e-05, "loss": 0.6194, "step": 223270 }, { "epoch": 19.34, "learning_rate": 2.6692783624378394e-05, "loss": 0.7295, "step": 223280 }, { "epoch": 19.34, "learning_rate": 2.6690615242280557e-05, "loss": 0.6925, "step": 223290 }, { "epoch": 19.34, "learning_rate": 2.668844686018272e-05, "loss": 0.81, "step": 223300 }, { "epoch": 19.34, "learning_rate": 2.6686278478084883e-05, "loss": 0.7691, "step": 223310 }, { "epoch": 19.34, "learning_rate": 2.6684110095987042e-05, "loss": 0.7228, "step": 223320 }, { "epoch": 19.34, "learning_rate": 2.668194171388921e-05, "loss": 0.6932, "step": 223330 }, { "epoch": 19.34, "learning_rate": 2.6679773331791372e-05, "loss": 0.7408, "step": 223340 }, { "epoch": 19.34, "learning_rate": 2.6677604949693535e-05, "loss": 0.7329, "step": 223350 }, { "epoch": 19.34, "learning_rate": 2.6675436567595694e-05, "loss": 0.821, "step": 223360 }, { "epoch": 19.35, "learning_rate": 2.6673268185497857e-05, "loss": 0.6405, "step": 223370 }, { "epoch": 19.35, "learning_rate": 2.667109980340002e-05, "loss": 0.7511, "step": 223380 }, { "epoch": 19.35, "learning_rate": 2.6668931421302187e-05, "loss": 0.6424, "step": 223390 }, { "epoch": 19.35, "learning_rate": 2.6666763039204343e-05, "loss": 0.6042, "step": 223400 }, { "epoch": 19.35, "learning_rate": 2.666459465710651e-05, "loss": 0.9632, "step": 223410 }, { "epoch": 19.35, "learning_rate": 2.6662426275008672e-05, "loss": 0.8282, "step": 223420 }, { "epoch": 19.35, "learning_rate": 2.6660257892910835e-05, "loss": 0.7851, "step": 223430 }, { "epoch": 19.35, "learning_rate": 2.6658089510812995e-05, "loss": 0.6431, "step": 223440 }, { "epoch": 19.35, "learning_rate": 2.6655921128715158e-05, "loss": 0.7512, "step": 223450 }, { "epoch": 19.35, "learning_rate": 2.665375274661732e-05, "loss": 0.7976, "step": 223460 }, { "epoch": 19.35, "learning_rate": 2.6651584364519487e-05, "loss": 0.7006, "step": 223470 }, { "epoch": 19.35, "learning_rate": 2.6649415982421647e-05, "loss": 0.6158, "step": 223480 }, { "epoch": 19.36, "learning_rate": 2.664724760032381e-05, "loss": 0.7656, "step": 223490 }, { "epoch": 19.36, "learning_rate": 2.6645079218225973e-05, "loss": 0.7086, "step": 223500 }, { "epoch": 19.36, "learning_rate": 2.6642910836128133e-05, "loss": 0.8957, "step": 223510 }, { "epoch": 19.36, "learning_rate": 2.6640742454030296e-05, "loss": 0.8391, "step": 223520 }, { "epoch": 19.36, "learning_rate": 2.663857407193246e-05, "loss": 0.6839, "step": 223530 }, { "epoch": 19.36, "learning_rate": 2.6636405689834625e-05, "loss": 0.9534, "step": 223540 }, { "epoch": 19.36, "learning_rate": 2.6634237307736785e-05, "loss": 0.8145, "step": 223550 }, { "epoch": 19.36, "learning_rate": 2.6632068925638948e-05, "loss": 0.979, "step": 223560 }, { "epoch": 19.36, "learning_rate": 2.662990054354111e-05, "loss": 0.7486, "step": 223570 }, { "epoch": 19.36, "learning_rate": 2.6627732161443274e-05, "loss": 0.8348, "step": 223580 }, { "epoch": 19.36, "learning_rate": 2.6625563779345433e-05, "loss": 0.7972, "step": 223590 }, { "epoch": 19.37, "learning_rate": 2.6623395397247596e-05, "loss": 0.8373, "step": 223600 }, { "epoch": 19.37, "learning_rate": 2.6621227015149763e-05, "loss": 0.7665, "step": 223610 }, { "epoch": 19.37, "learning_rate": 2.6619058633051926e-05, "loss": 0.6995, "step": 223620 }, { "epoch": 19.37, "learning_rate": 2.6616890250954085e-05, "loss": 0.7119, "step": 223630 }, { "epoch": 19.37, "learning_rate": 2.6614721868856248e-05, "loss": 0.7989, "step": 223640 }, { "epoch": 19.37, "learning_rate": 2.661255348675841e-05, "loss": 0.6677, "step": 223650 }, { "epoch": 19.37, "learning_rate": 2.6610385104660574e-05, "loss": 0.7423, "step": 223660 }, { "epoch": 19.37, "learning_rate": 2.6608216722562734e-05, "loss": 0.7871, "step": 223670 }, { "epoch": 19.37, "learning_rate": 2.6606048340464897e-05, "loss": 0.5603, "step": 223680 }, { "epoch": 19.37, "learning_rate": 2.6603879958367063e-05, "loss": 0.7242, "step": 223690 }, { "epoch": 19.37, "learning_rate": 2.6601711576269226e-05, "loss": 0.8087, "step": 223700 }, { "epoch": 19.37, "learning_rate": 2.6599543194171386e-05, "loss": 0.7202, "step": 223710 }, { "epoch": 19.38, "learning_rate": 2.659737481207355e-05, "loss": 0.8698, "step": 223720 }, { "epoch": 19.38, "learning_rate": 2.6595206429975712e-05, "loss": 0.7675, "step": 223730 }, { "epoch": 19.38, "learning_rate": 2.659303804787787e-05, "loss": 0.7582, "step": 223740 }, { "epoch": 19.38, "learning_rate": 2.6590869665780035e-05, "loss": 0.7971, "step": 223750 }, { "epoch": 19.38, "learning_rate": 2.65887012836822e-05, "loss": 0.8301, "step": 223760 }, { "epoch": 19.38, "learning_rate": 2.6586532901584364e-05, "loss": 0.7195, "step": 223770 }, { "epoch": 19.38, "learning_rate": 2.6584364519486524e-05, "loss": 0.603, "step": 223780 }, { "epoch": 19.38, "learning_rate": 2.6582196137388687e-05, "loss": 0.7341, "step": 223790 }, { "epoch": 19.38, "learning_rate": 2.658002775529085e-05, "loss": 0.7966, "step": 223800 }, { "epoch": 19.38, "learning_rate": 2.6577859373193013e-05, "loss": 0.9244, "step": 223810 }, { "epoch": 19.38, "learning_rate": 2.6575690991095172e-05, "loss": 0.7088, "step": 223820 }, { "epoch": 19.39, "learning_rate": 2.657352260899734e-05, "loss": 0.8018, "step": 223830 }, { "epoch": 19.39, "learning_rate": 2.65713542268995e-05, "loss": 0.7825, "step": 223840 }, { "epoch": 19.39, "learning_rate": 2.6569185844801665e-05, "loss": 0.8003, "step": 223850 }, { "epoch": 19.39, "learning_rate": 2.6567017462703824e-05, "loss": 0.7673, "step": 223860 }, { "epoch": 19.39, "learning_rate": 2.6564849080605987e-05, "loss": 0.8252, "step": 223870 }, { "epoch": 19.39, "learning_rate": 2.656268069850815e-05, "loss": 0.7417, "step": 223880 }, { "epoch": 19.39, "learning_rate": 2.6560512316410317e-05, "loss": 0.6968, "step": 223890 }, { "epoch": 19.39, "learning_rate": 2.6558343934312476e-05, "loss": 0.7621, "step": 223900 }, { "epoch": 19.39, "learning_rate": 2.655617555221464e-05, "loss": 0.6914, "step": 223910 }, { "epoch": 19.39, "learning_rate": 2.6554007170116802e-05, "loss": 0.7253, "step": 223920 }, { "epoch": 19.39, "learning_rate": 2.6551838788018965e-05, "loss": 0.5877, "step": 223930 }, { "epoch": 19.39, "learning_rate": 2.6549670405921125e-05, "loss": 0.7537, "step": 223940 }, { "epoch": 19.4, "learning_rate": 2.6547502023823288e-05, "loss": 0.8582, "step": 223950 }, { "epoch": 19.4, "learning_rate": 2.654533364172545e-05, "loss": 0.598, "step": 223960 }, { "epoch": 19.4, "learning_rate": 2.6543165259627617e-05, "loss": 0.9123, "step": 223970 }, { "epoch": 19.4, "learning_rate": 2.6540996877529777e-05, "loss": 0.8092, "step": 223980 }, { "epoch": 19.4, "learning_rate": 2.653882849543194e-05, "loss": 0.7789, "step": 223990 }, { "epoch": 19.4, "learning_rate": 2.6536660113334103e-05, "loss": 0.715, "step": 224000 }, { "epoch": 19.4, "learning_rate": 2.6534491731236262e-05, "loss": 0.6799, "step": 224010 }, { "epoch": 19.4, "learning_rate": 2.6532323349138425e-05, "loss": 0.7873, "step": 224020 }, { "epoch": 19.4, "learning_rate": 2.653015496704059e-05, "loss": 0.9732, "step": 224030 }, { "epoch": 19.4, "learning_rate": 2.6527986584942755e-05, "loss": 0.7077, "step": 224040 }, { "epoch": 19.4, "learning_rate": 2.6525818202844914e-05, "loss": 0.8609, "step": 224050 }, { "epoch": 19.41, "learning_rate": 2.6523649820747077e-05, "loss": 0.6613, "step": 224060 }, { "epoch": 19.41, "learning_rate": 2.652148143864924e-05, "loss": 0.9393, "step": 224070 }, { "epoch": 19.41, "learning_rate": 2.6519313056551403e-05, "loss": 0.8956, "step": 224080 }, { "epoch": 19.41, "learning_rate": 2.6517144674453563e-05, "loss": 0.6557, "step": 224090 }, { "epoch": 19.41, "learning_rate": 2.6514976292355726e-05, "loss": 0.5974, "step": 224100 }, { "epoch": 19.41, "learning_rate": 2.6512807910257892e-05, "loss": 0.8272, "step": 224110 }, { "epoch": 19.41, "learning_rate": 2.6510639528160055e-05, "loss": 0.6877, "step": 224120 }, { "epoch": 19.41, "learning_rate": 2.6508471146062215e-05, "loss": 0.7409, "step": 224130 }, { "epoch": 19.41, "learning_rate": 2.6506302763964378e-05, "loss": 0.7757, "step": 224140 }, { "epoch": 19.41, "learning_rate": 2.650413438186654e-05, "loss": 0.7806, "step": 224150 }, { "epoch": 19.41, "learning_rate": 2.6501965999768704e-05, "loss": 0.7721, "step": 224160 }, { "epoch": 19.41, "learning_rate": 2.6499797617670864e-05, "loss": 0.6874, "step": 224170 }, { "epoch": 19.42, "learning_rate": 2.649762923557303e-05, "loss": 0.8159, "step": 224180 }, { "epoch": 19.42, "learning_rate": 2.6495460853475193e-05, "loss": 0.724, "step": 224190 }, { "epoch": 19.42, "learning_rate": 2.6493292471377356e-05, "loss": 0.7332, "step": 224200 }, { "epoch": 19.42, "learning_rate": 2.6491124089279516e-05, "loss": 0.7162, "step": 224210 }, { "epoch": 19.42, "learning_rate": 2.648895570718168e-05, "loss": 0.8769, "step": 224220 }, { "epoch": 19.42, "learning_rate": 2.6486787325083842e-05, "loss": 0.7809, "step": 224230 }, { "epoch": 19.42, "learning_rate": 2.6484618942986005e-05, "loss": 0.7084, "step": 224240 }, { "epoch": 19.42, "learning_rate": 2.6482450560888164e-05, "loss": 0.6664, "step": 224250 }, { "epoch": 19.42, "learning_rate": 2.648028217879033e-05, "loss": 0.6175, "step": 224260 }, { "epoch": 19.42, "learning_rate": 2.6478113796692494e-05, "loss": 0.8459, "step": 224270 }, { "epoch": 19.42, "learning_rate": 2.6475945414594653e-05, "loss": 0.7843, "step": 224280 }, { "epoch": 19.42, "learning_rate": 2.6473777032496816e-05, "loss": 0.9281, "step": 224290 }, { "epoch": 19.43, "learning_rate": 2.647160865039898e-05, "loss": 0.7949, "step": 224300 }, { "epoch": 19.43, "learning_rate": 2.6469440268301142e-05, "loss": 0.5665, "step": 224310 }, { "epoch": 19.43, "learning_rate": 2.6467271886203302e-05, "loss": 0.8282, "step": 224320 }, { "epoch": 19.43, "learning_rate": 2.646510350410547e-05, "loss": 0.743, "step": 224330 }, { "epoch": 19.43, "learning_rate": 2.646293512200763e-05, "loss": 0.727, "step": 224340 }, { "epoch": 19.43, "learning_rate": 2.6460766739909794e-05, "loss": 0.5729, "step": 224350 }, { "epoch": 19.43, "learning_rate": 2.6458598357811954e-05, "loss": 0.7691, "step": 224360 }, { "epoch": 19.43, "learning_rate": 2.6456429975714117e-05, "loss": 0.7998, "step": 224370 }, { "epoch": 19.43, "learning_rate": 2.645426159361628e-05, "loss": 0.8615, "step": 224380 }, { "epoch": 19.43, "learning_rate": 2.6452093211518446e-05, "loss": 0.8082, "step": 224390 }, { "epoch": 19.43, "learning_rate": 2.6449924829420606e-05, "loss": 0.908, "step": 224400 }, { "epoch": 19.44, "learning_rate": 2.644775644732277e-05, "loss": 0.9091, "step": 224410 }, { "epoch": 19.44, "learning_rate": 2.6445588065224932e-05, "loss": 0.8145, "step": 224420 }, { "epoch": 19.44, "learning_rate": 2.6443419683127095e-05, "loss": 0.6503, "step": 224430 }, { "epoch": 19.44, "learning_rate": 2.6441251301029255e-05, "loss": 0.758, "step": 224440 }, { "epoch": 19.44, "learning_rate": 2.6439082918931418e-05, "loss": 0.7991, "step": 224450 }, { "epoch": 19.44, "learning_rate": 2.6436914536833584e-05, "loss": 0.8883, "step": 224460 }, { "epoch": 19.44, "learning_rate": 2.6434746154735747e-05, "loss": 0.7649, "step": 224470 }, { "epoch": 19.44, "learning_rate": 2.6432577772637907e-05, "loss": 0.5473, "step": 224480 }, { "epoch": 19.44, "learning_rate": 2.643040939054007e-05, "loss": 0.6422, "step": 224490 }, { "epoch": 19.44, "learning_rate": 2.6428241008442233e-05, "loss": 0.6566, "step": 224500 }, { "epoch": 19.44, "learning_rate": 2.6426072626344392e-05, "loss": 0.8203, "step": 224510 }, { "epoch": 19.44, "learning_rate": 2.6423904244246555e-05, "loss": 0.7664, "step": 224520 }, { "epoch": 19.45, "learning_rate": 2.642173586214872e-05, "loss": 0.8277, "step": 224530 }, { "epoch": 19.45, "learning_rate": 2.6419567480050885e-05, "loss": 0.7095, "step": 224540 }, { "epoch": 19.45, "learning_rate": 2.6417399097953044e-05, "loss": 0.7056, "step": 224550 }, { "epoch": 19.45, "learning_rate": 2.6415230715855207e-05, "loss": 0.8522, "step": 224560 }, { "epoch": 19.45, "learning_rate": 2.641306233375737e-05, "loss": 0.7311, "step": 224570 }, { "epoch": 19.45, "learning_rate": 2.6410893951659533e-05, "loss": 0.8749, "step": 224580 }, { "epoch": 19.45, "learning_rate": 2.6408725569561693e-05, "loss": 0.7887, "step": 224590 }, { "epoch": 19.45, "learning_rate": 2.6406557187463856e-05, "loss": 0.8266, "step": 224600 }, { "epoch": 19.45, "learning_rate": 2.6404388805366022e-05, "loss": 0.6428, "step": 224610 }, { "epoch": 19.45, "learning_rate": 2.6402220423268185e-05, "loss": 0.7613, "step": 224620 }, { "epoch": 19.45, "learning_rate": 2.6400052041170345e-05, "loss": 0.7676, "step": 224630 }, { "epoch": 19.46, "learning_rate": 2.6397883659072508e-05, "loss": 0.8602, "step": 224640 }, { "epoch": 19.46, "learning_rate": 2.639571527697467e-05, "loss": 0.8486, "step": 224650 }, { "epoch": 19.46, "learning_rate": 2.6393546894876834e-05, "loss": 0.7474, "step": 224660 }, { "epoch": 19.46, "learning_rate": 2.6391378512778994e-05, "loss": 0.832, "step": 224670 }, { "epoch": 19.46, "learning_rate": 2.638921013068116e-05, "loss": 0.8025, "step": 224680 }, { "epoch": 19.46, "learning_rate": 2.6387041748583323e-05, "loss": 0.8328, "step": 224690 }, { "epoch": 19.46, "learning_rate": 2.6384873366485486e-05, "loss": 0.7498, "step": 224700 }, { "epoch": 19.46, "learning_rate": 2.6382704984387646e-05, "loss": 0.7584, "step": 224710 }, { "epoch": 19.46, "learning_rate": 2.638053660228981e-05, "loss": 0.7511, "step": 224720 }, { "epoch": 19.46, "learning_rate": 2.637836822019197e-05, "loss": 0.8052, "step": 224730 }, { "epoch": 19.46, "learning_rate": 2.6376199838094138e-05, "loss": 0.7784, "step": 224740 }, { "epoch": 19.46, "learning_rate": 2.6374031455996294e-05, "loss": 0.6549, "step": 224750 }, { "epoch": 19.47, "learning_rate": 2.637186307389846e-05, "loss": 0.6864, "step": 224760 }, { "epoch": 19.47, "learning_rate": 2.6369694691800624e-05, "loss": 0.7768, "step": 224770 }, { "epoch": 19.47, "learning_rate": 2.6367526309702783e-05, "loss": 0.812, "step": 224780 }, { "epoch": 19.47, "learning_rate": 2.6365357927604946e-05, "loss": 0.8304, "step": 224790 }, { "epoch": 19.47, "learning_rate": 2.636318954550711e-05, "loss": 0.6102, "step": 224800 }, { "epoch": 19.47, "learning_rate": 2.6361021163409272e-05, "loss": 0.7558, "step": 224810 }, { "epoch": 19.47, "learning_rate": 2.6358852781311432e-05, "loss": 0.6437, "step": 224820 }, { "epoch": 19.47, "learning_rate": 2.6356684399213598e-05, "loss": 0.6364, "step": 224830 }, { "epoch": 19.47, "learning_rate": 2.635451601711576e-05, "loss": 0.8178, "step": 224840 }, { "epoch": 19.47, "learning_rate": 2.6352347635017924e-05, "loss": 0.7064, "step": 224850 }, { "epoch": 19.47, "learning_rate": 2.6350179252920084e-05, "loss": 0.6963, "step": 224860 }, { "epoch": 19.48, "learning_rate": 2.6348010870822247e-05, "loss": 0.9507, "step": 224870 }, { "epoch": 19.48, "learning_rate": 2.634584248872441e-05, "loss": 0.6503, "step": 224880 }, { "epoch": 19.48, "learning_rate": 2.6343674106626576e-05, "loss": 0.842, "step": 224890 }, { "epoch": 19.48, "learning_rate": 2.6341505724528736e-05, "loss": 0.6477, "step": 224900 }, { "epoch": 19.48, "learning_rate": 2.63393373424309e-05, "loss": 0.9281, "step": 224910 }, { "epoch": 19.48, "learning_rate": 2.6337168960333062e-05, "loss": 0.8972, "step": 224920 }, { "epoch": 19.48, "learning_rate": 2.6335000578235225e-05, "loss": 0.7149, "step": 224930 }, { "epoch": 19.48, "learning_rate": 2.6332832196137385e-05, "loss": 0.7363, "step": 224940 }, { "epoch": 19.48, "learning_rate": 2.6330663814039548e-05, "loss": 0.737, "step": 224950 }, { "epoch": 19.48, "learning_rate": 2.6328495431941714e-05, "loss": 0.6285, "step": 224960 }, { "epoch": 19.48, "learning_rate": 2.6326327049843877e-05, "loss": 0.8507, "step": 224970 }, { "epoch": 19.48, "learning_rate": 2.6324158667746037e-05, "loss": 0.6925, "step": 224980 }, { "epoch": 19.49, "learning_rate": 2.63219902856482e-05, "loss": 0.8259, "step": 224990 }, { "epoch": 19.49, "learning_rate": 2.6319821903550363e-05, "loss": 0.7193, "step": 225000 }, { "epoch": 19.49, "learning_rate": 2.6317653521452526e-05, "loss": 0.7673, "step": 225010 }, { "epoch": 19.49, "learning_rate": 2.6315485139354685e-05, "loss": 0.7442, "step": 225020 }, { "epoch": 19.49, "learning_rate": 2.6313316757256848e-05, "loss": 0.8334, "step": 225030 }, { "epoch": 19.49, "learning_rate": 2.6311148375159015e-05, "loss": 0.7525, "step": 225040 }, { "epoch": 19.49, "learning_rate": 2.6308979993061174e-05, "loss": 0.7756, "step": 225050 }, { "epoch": 19.49, "learning_rate": 2.6306811610963337e-05, "loss": 0.5648, "step": 225060 }, { "epoch": 19.49, "learning_rate": 2.63046432288655e-05, "loss": 0.6996, "step": 225070 }, { "epoch": 19.49, "learning_rate": 2.6302474846767663e-05, "loss": 0.8155, "step": 225080 }, { "epoch": 19.49, "learning_rate": 2.6300306464669823e-05, "loss": 0.7523, "step": 225090 }, { "epoch": 19.5, "learning_rate": 2.6298138082571986e-05, "loss": 0.822, "step": 225100 }, { "epoch": 19.5, "learning_rate": 2.6295969700474152e-05, "loss": 0.7332, "step": 225110 }, { "epoch": 19.5, "learning_rate": 2.6293801318376315e-05, "loss": 0.9027, "step": 225120 }, { "epoch": 19.5, "learning_rate": 2.6291632936278475e-05, "loss": 0.7366, "step": 225130 }, { "epoch": 19.5, "learning_rate": 2.6289464554180638e-05, "loss": 0.7663, "step": 225140 }, { "epoch": 19.5, "learning_rate": 2.62872961720828e-05, "loss": 0.8305, "step": 225150 }, { "epoch": 19.5, "learning_rate": 2.6285127789984964e-05, "loss": 0.7557, "step": 225160 }, { "epoch": 19.5, "learning_rate": 2.6282959407887123e-05, "loss": 0.5932, "step": 225170 }, { "epoch": 19.5, "learning_rate": 2.628079102578929e-05, "loss": 0.7535, "step": 225180 }, { "epoch": 19.5, "learning_rate": 2.6278622643691453e-05, "loss": 0.7052, "step": 225190 }, { "epoch": 19.5, "learning_rate": 2.6276454261593616e-05, "loss": 0.8174, "step": 225200 }, { "epoch": 19.5, "learning_rate": 2.6274285879495775e-05, "loss": 0.8386, "step": 225210 }, { "epoch": 19.51, "learning_rate": 2.627211749739794e-05, "loss": 0.6719, "step": 225220 }, { "epoch": 19.51, "learning_rate": 2.62699491153001e-05, "loss": 0.8466, "step": 225230 }, { "epoch": 19.51, "learning_rate": 2.6267780733202268e-05, "loss": 0.7832, "step": 225240 }, { "epoch": 19.51, "learning_rate": 2.6265612351104427e-05, "loss": 0.8143, "step": 225250 }, { "epoch": 19.51, "learning_rate": 2.626344396900659e-05, "loss": 0.7857, "step": 225260 }, { "epoch": 19.51, "learning_rate": 2.6261275586908753e-05, "loss": 0.7655, "step": 225270 }, { "epoch": 19.51, "learning_rate": 2.6259107204810913e-05, "loss": 0.6679, "step": 225280 }, { "epoch": 19.51, "learning_rate": 2.6256938822713076e-05, "loss": 0.6813, "step": 225290 }, { "epoch": 19.51, "learning_rate": 2.625477044061524e-05, "loss": 0.6989, "step": 225300 }, { "epoch": 19.51, "learning_rate": 2.6252602058517402e-05, "loss": 0.6703, "step": 225310 }, { "epoch": 19.51, "learning_rate": 2.6250433676419562e-05, "loss": 0.7137, "step": 225320 }, { "epoch": 19.52, "learning_rate": 2.6248265294321728e-05, "loss": 0.7631, "step": 225330 }, { "epoch": 19.52, "learning_rate": 2.624609691222389e-05, "loss": 0.6496, "step": 225340 }, { "epoch": 19.52, "learning_rate": 2.6243928530126054e-05, "loss": 0.8819, "step": 225350 }, { "epoch": 19.52, "learning_rate": 2.6241760148028214e-05, "loss": 0.857, "step": 225360 }, { "epoch": 19.52, "learning_rate": 2.6239591765930377e-05, "loss": 0.6831, "step": 225370 }, { "epoch": 19.52, "learning_rate": 2.623742338383254e-05, "loss": 0.7056, "step": 225380 }, { "epoch": 19.52, "learning_rate": 2.6235255001734706e-05, "loss": 0.7504, "step": 225390 }, { "epoch": 19.52, "learning_rate": 2.6233086619636866e-05, "loss": 0.6023, "step": 225400 }, { "epoch": 19.52, "learning_rate": 2.623091823753903e-05, "loss": 0.929, "step": 225410 }, { "epoch": 19.52, "learning_rate": 2.6228749855441192e-05, "loss": 0.6934, "step": 225420 }, { "epoch": 19.52, "learning_rate": 2.6226581473343355e-05, "loss": 0.6735, "step": 225430 }, { "epoch": 19.52, "learning_rate": 2.6224413091245514e-05, "loss": 0.6967, "step": 225440 }, { "epoch": 19.53, "learning_rate": 2.6222244709147677e-05, "loss": 0.8467, "step": 225450 }, { "epoch": 19.53, "learning_rate": 2.6220076327049844e-05, "loss": 0.7226, "step": 225460 }, { "epoch": 19.53, "learning_rate": 2.6217907944952007e-05, "loss": 0.7719, "step": 225470 }, { "epoch": 19.53, "learning_rate": 2.6215739562854166e-05, "loss": 0.7041, "step": 225480 }, { "epoch": 19.53, "learning_rate": 2.621357118075633e-05, "loss": 0.6722, "step": 225490 }, { "epoch": 19.53, "learning_rate": 2.6211402798658492e-05, "loss": 0.7209, "step": 225500 }, { "epoch": 19.53, "learning_rate": 2.6209234416560655e-05, "loss": 0.8603, "step": 225510 }, { "epoch": 19.53, "learning_rate": 2.6207066034462815e-05, "loss": 0.7442, "step": 225520 }, { "epoch": 19.53, "learning_rate": 2.620489765236498e-05, "loss": 0.6531, "step": 225530 }, { "epoch": 19.53, "learning_rate": 2.6202729270267144e-05, "loss": 0.7735, "step": 225540 }, { "epoch": 19.53, "learning_rate": 2.6200560888169304e-05, "loss": 0.7743, "step": 225550 }, { "epoch": 19.53, "learning_rate": 2.6198392506071467e-05, "loss": 0.7394, "step": 225560 }, { "epoch": 19.54, "learning_rate": 2.619622412397363e-05, "loss": 0.8911, "step": 225570 }, { "epoch": 19.54, "learning_rate": 2.6194055741875793e-05, "loss": 0.8608, "step": 225580 }, { "epoch": 19.54, "learning_rate": 2.6191887359777953e-05, "loss": 0.9921, "step": 225590 }, { "epoch": 19.54, "learning_rate": 2.6189718977680116e-05, "loss": 0.7545, "step": 225600 }, { "epoch": 19.54, "learning_rate": 2.6187550595582282e-05, "loss": 0.873, "step": 225610 }, { "epoch": 19.54, "learning_rate": 2.6185382213484445e-05, "loss": 0.8727, "step": 225620 }, { "epoch": 19.54, "learning_rate": 2.6183213831386605e-05, "loss": 0.8246, "step": 225630 }, { "epoch": 19.54, "learning_rate": 2.6181045449288768e-05, "loss": 0.9003, "step": 225640 }, { "epoch": 19.54, "learning_rate": 2.617887706719093e-05, "loss": 0.7518, "step": 225650 }, { "epoch": 19.54, "learning_rate": 2.6176708685093094e-05, "loss": 0.8377, "step": 225660 }, { "epoch": 19.54, "learning_rate": 2.6174540302995253e-05, "loss": 0.7425, "step": 225670 }, { "epoch": 19.55, "learning_rate": 2.617237192089742e-05, "loss": 0.9244, "step": 225680 }, { "epoch": 19.55, "learning_rate": 2.6170203538799583e-05, "loss": 0.7355, "step": 225690 }, { "epoch": 19.55, "learning_rate": 2.6168035156701746e-05, "loss": 0.6677, "step": 225700 }, { "epoch": 19.55, "learning_rate": 2.6165866774603905e-05, "loss": 0.5779, "step": 225710 }, { "epoch": 19.55, "learning_rate": 2.6163698392506068e-05, "loss": 0.7118, "step": 225720 }, { "epoch": 19.55, "learning_rate": 2.616153001040823e-05, "loss": 0.8079, "step": 225730 }, { "epoch": 19.55, "learning_rate": 2.6159361628310398e-05, "loss": 0.7865, "step": 225740 }, { "epoch": 19.55, "learning_rate": 2.6157193246212557e-05, "loss": 0.7868, "step": 225750 }, { "epoch": 19.55, "learning_rate": 2.615502486411472e-05, "loss": 0.7393, "step": 225760 }, { "epoch": 19.55, "learning_rate": 2.6152856482016883e-05, "loss": 0.698, "step": 225770 }, { "epoch": 19.55, "learning_rate": 2.6150688099919046e-05, "loss": 0.6644, "step": 225780 }, { "epoch": 19.55, "learning_rate": 2.6148519717821206e-05, "loss": 0.6808, "step": 225790 }, { "epoch": 19.56, "learning_rate": 2.614635133572337e-05, "loss": 0.6492, "step": 225800 }, { "epoch": 19.56, "learning_rate": 2.6144182953625535e-05, "loss": 0.7145, "step": 225810 }, { "epoch": 19.56, "learning_rate": 2.614201457152769e-05, "loss": 0.726, "step": 225820 }, { "epoch": 19.56, "learning_rate": 2.6139846189429858e-05, "loss": 0.7209, "step": 225830 }, { "epoch": 19.56, "learning_rate": 2.613767780733202e-05, "loss": 0.515, "step": 225840 }, { "epoch": 19.56, "learning_rate": 2.6135509425234184e-05, "loss": 0.8325, "step": 225850 }, { "epoch": 19.56, "learning_rate": 2.6133341043136344e-05, "loss": 0.7445, "step": 225860 }, { "epoch": 19.56, "learning_rate": 2.6131172661038507e-05, "loss": 0.739, "step": 225870 }, { "epoch": 19.56, "learning_rate": 2.612900427894067e-05, "loss": 0.7942, "step": 225880 }, { "epoch": 19.56, "learning_rate": 2.6126835896842836e-05, "loss": 0.7579, "step": 225890 }, { "epoch": 19.56, "learning_rate": 2.6124667514744996e-05, "loss": 0.801, "step": 225900 }, { "epoch": 19.57, "learning_rate": 2.612249913264716e-05, "loss": 0.826, "step": 225910 }, { "epoch": 19.57, "learning_rate": 2.612033075054932e-05, "loss": 0.8709, "step": 225920 }, { "epoch": 19.57, "learning_rate": 2.6118162368451485e-05, "loss": 0.6645, "step": 225930 }, { "epoch": 19.57, "learning_rate": 2.6115993986353644e-05, "loss": 0.7767, "step": 225940 }, { "epoch": 19.57, "learning_rate": 2.6113825604255807e-05, "loss": 0.6816, "step": 225950 }, { "epoch": 19.57, "learning_rate": 2.6111657222157974e-05, "loss": 0.783, "step": 225960 }, { "epoch": 19.57, "learning_rate": 2.6109488840060137e-05, "loss": 0.6974, "step": 225970 }, { "epoch": 19.57, "learning_rate": 2.6107320457962296e-05, "loss": 0.7222, "step": 225980 }, { "epoch": 19.57, "learning_rate": 2.610515207586446e-05, "loss": 0.7921, "step": 225990 }, { "epoch": 19.57, "learning_rate": 2.6102983693766622e-05, "loss": 0.8104, "step": 226000 }, { "epoch": 19.57, "learning_rate": 2.6100815311668785e-05, "loss": 0.7675, "step": 226010 }, { "epoch": 19.57, "learning_rate": 2.6098646929570945e-05, "loss": 0.7255, "step": 226020 }, { "epoch": 19.58, "learning_rate": 2.609647854747311e-05, "loss": 0.7207, "step": 226030 }, { "epoch": 19.58, "learning_rate": 2.6094310165375274e-05, "loss": 0.8615, "step": 226040 }, { "epoch": 19.58, "learning_rate": 2.6092141783277434e-05, "loss": 0.7361, "step": 226050 }, { "epoch": 19.58, "learning_rate": 2.6089973401179597e-05, "loss": 0.6422, "step": 226060 }, { "epoch": 19.58, "learning_rate": 2.608780501908176e-05, "loss": 0.7351, "step": 226070 }, { "epoch": 19.58, "learning_rate": 2.6085636636983923e-05, "loss": 0.7078, "step": 226080 }, { "epoch": 19.58, "learning_rate": 2.6083468254886082e-05, "loss": 0.7008, "step": 226090 }, { "epoch": 19.58, "learning_rate": 2.6081299872788245e-05, "loss": 0.8765, "step": 226100 }, { "epoch": 19.58, "learning_rate": 2.6079131490690412e-05, "loss": 0.8139, "step": 226110 }, { "epoch": 19.58, "learning_rate": 2.6076963108592575e-05, "loss": 0.7078, "step": 226120 }, { "epoch": 19.58, "learning_rate": 2.6074794726494735e-05, "loss": 0.7394, "step": 226130 }, { "epoch": 19.59, "learning_rate": 2.6072626344396898e-05, "loss": 0.85, "step": 226140 }, { "epoch": 19.59, "learning_rate": 2.607045796229906e-05, "loss": 0.8132, "step": 226150 }, { "epoch": 19.59, "learning_rate": 2.6068289580201224e-05, "loss": 0.8746, "step": 226160 }, { "epoch": 19.59, "learning_rate": 2.6066121198103383e-05, "loss": 0.8428, "step": 226170 }, { "epoch": 19.59, "learning_rate": 2.606395281600555e-05, "loss": 0.7274, "step": 226180 }, { "epoch": 19.59, "learning_rate": 2.6061784433907713e-05, "loss": 0.6521, "step": 226190 }, { "epoch": 19.59, "learning_rate": 2.6059616051809876e-05, "loss": 0.8366, "step": 226200 }, { "epoch": 19.59, "learning_rate": 2.6057447669712035e-05, "loss": 0.7177, "step": 226210 }, { "epoch": 19.59, "learning_rate": 2.6055279287614198e-05, "loss": 0.663, "step": 226220 }, { "epoch": 19.59, "learning_rate": 2.605311090551636e-05, "loss": 0.7691, "step": 226230 }, { "epoch": 19.59, "learning_rate": 2.6050942523418528e-05, "loss": 0.8906, "step": 226240 }, { "epoch": 19.59, "learning_rate": 2.6048774141320687e-05, "loss": 0.8592, "step": 226250 }, { "epoch": 19.6, "learning_rate": 2.604660575922285e-05, "loss": 0.7641, "step": 226260 }, { "epoch": 19.6, "learning_rate": 2.6044437377125013e-05, "loss": 0.6021, "step": 226270 }, { "epoch": 19.6, "learning_rate": 2.6042268995027176e-05, "loss": 0.9159, "step": 226280 }, { "epoch": 19.6, "learning_rate": 2.6040100612929336e-05, "loss": 0.8219, "step": 226290 }, { "epoch": 19.6, "learning_rate": 2.60379322308315e-05, "loss": 0.8517, "step": 226300 }, { "epoch": 19.6, "learning_rate": 2.6035763848733665e-05, "loss": 0.8427, "step": 226310 }, { "epoch": 19.6, "learning_rate": 2.6033595466635825e-05, "loss": 0.8634, "step": 226320 }, { "epoch": 19.6, "learning_rate": 2.6031427084537988e-05, "loss": 0.7917, "step": 226330 }, { "epoch": 19.6, "learning_rate": 2.602925870244015e-05, "loss": 0.9575, "step": 226340 }, { "epoch": 19.6, "learning_rate": 2.6027090320342314e-05, "loss": 0.8722, "step": 226350 }, { "epoch": 19.6, "learning_rate": 2.6024921938244473e-05, "loss": 0.7203, "step": 226360 }, { "epoch": 19.61, "learning_rate": 2.6022753556146636e-05, "loss": 0.7487, "step": 226370 }, { "epoch": 19.61, "learning_rate": 2.60205851740488e-05, "loss": 0.745, "step": 226380 }, { "epoch": 19.61, "learning_rate": 2.6018416791950966e-05, "loss": 0.7769, "step": 226390 }, { "epoch": 19.61, "learning_rate": 2.6016248409853125e-05, "loss": 0.7619, "step": 226400 }, { "epoch": 19.61, "learning_rate": 2.601408002775529e-05, "loss": 0.8006, "step": 226410 }, { "epoch": 19.61, "learning_rate": 2.601191164565745e-05, "loss": 0.7345, "step": 226420 }, { "epoch": 19.61, "learning_rate": 2.6009743263559614e-05, "loss": 0.6999, "step": 226430 }, { "epoch": 19.61, "learning_rate": 2.6007574881461774e-05, "loss": 0.7787, "step": 226440 }, { "epoch": 19.61, "learning_rate": 2.6005406499363937e-05, "loss": 0.6898, "step": 226450 }, { "epoch": 19.61, "learning_rate": 2.6003238117266103e-05, "loss": 0.7008, "step": 226460 }, { "epoch": 19.61, "learning_rate": 2.6001069735168266e-05, "loss": 0.8115, "step": 226470 }, { "epoch": 19.61, "learning_rate": 2.5998901353070426e-05, "loss": 0.6931, "step": 226480 }, { "epoch": 19.62, "learning_rate": 2.599673297097259e-05, "loss": 0.6919, "step": 226490 }, { "epoch": 19.62, "learning_rate": 2.5994564588874752e-05, "loss": 0.6448, "step": 226500 }, { "epoch": 19.62, "learning_rate": 2.5992396206776915e-05, "loss": 0.7784, "step": 226510 }, { "epoch": 19.62, "learning_rate": 2.5990227824679075e-05, "loss": 0.7061, "step": 226520 }, { "epoch": 19.62, "learning_rate": 2.598805944258124e-05, "loss": 0.693, "step": 226530 }, { "epoch": 19.62, "learning_rate": 2.5985891060483404e-05, "loss": 0.6996, "step": 226540 }, { "epoch": 19.62, "learning_rate": 2.5983722678385567e-05, "loss": 0.7359, "step": 226550 }, { "epoch": 19.62, "learning_rate": 2.5981554296287727e-05, "loss": 0.8523, "step": 226560 }, { "epoch": 19.62, "learning_rate": 2.597938591418989e-05, "loss": 0.7493, "step": 226570 }, { "epoch": 19.62, "learning_rate": 2.5977217532092053e-05, "loss": 0.7315, "step": 226580 }, { "epoch": 19.62, "learning_rate": 2.5975049149994212e-05, "loss": 0.8479, "step": 226590 }, { "epoch": 19.62, "learning_rate": 2.597288076789638e-05, "loss": 0.7426, "step": 226600 }, { "epoch": 19.63, "learning_rate": 2.5970712385798542e-05, "loss": 0.5961, "step": 226610 }, { "epoch": 19.63, "learning_rate": 2.5968544003700705e-05, "loss": 0.7182, "step": 226620 }, { "epoch": 19.63, "learning_rate": 2.5966375621602864e-05, "loss": 0.7324, "step": 226630 }, { "epoch": 19.63, "learning_rate": 2.5964207239505027e-05, "loss": 0.779, "step": 226640 }, { "epoch": 19.63, "learning_rate": 2.596203885740719e-05, "loss": 0.8515, "step": 226650 }, { "epoch": 19.63, "learning_rate": 2.5959870475309353e-05, "loss": 0.8886, "step": 226660 }, { "epoch": 19.63, "learning_rate": 2.5957702093211513e-05, "loss": 0.8217, "step": 226670 }, { "epoch": 19.63, "learning_rate": 2.595553371111368e-05, "loss": 0.6729, "step": 226680 }, { "epoch": 19.63, "learning_rate": 2.5953365329015842e-05, "loss": 0.8451, "step": 226690 }, { "epoch": 19.63, "learning_rate": 2.5951196946918005e-05, "loss": 0.7468, "step": 226700 }, { "epoch": 19.63, "learning_rate": 2.5949028564820165e-05, "loss": 0.6234, "step": 226710 }, { "epoch": 19.64, "learning_rate": 2.5946860182722328e-05, "loss": 0.6204, "step": 226720 }, { "epoch": 19.64, "learning_rate": 2.594469180062449e-05, "loss": 0.7873, "step": 226730 }, { "epoch": 19.64, "learning_rate": 2.5942523418526657e-05, "loss": 0.7606, "step": 226740 }, { "epoch": 19.64, "learning_rate": 2.5940355036428817e-05, "loss": 0.744, "step": 226750 }, { "epoch": 19.64, "learning_rate": 2.593818665433098e-05, "loss": 0.8053, "step": 226760 }, { "epoch": 19.64, "learning_rate": 2.5936018272233143e-05, "loss": 0.7736, "step": 226770 }, { "epoch": 19.64, "learning_rate": 2.5933849890135306e-05, "loss": 0.6717, "step": 226780 }, { "epoch": 19.64, "learning_rate": 2.5931681508037466e-05, "loss": 0.6161, "step": 226790 }, { "epoch": 19.64, "learning_rate": 2.592951312593963e-05, "loss": 0.833, "step": 226800 }, { "epoch": 19.64, "learning_rate": 2.5927344743841795e-05, "loss": 0.8431, "step": 226810 }, { "epoch": 19.64, "learning_rate": 2.5925176361743955e-05, "loss": 0.7969, "step": 226820 }, { "epoch": 19.64, "learning_rate": 2.5923007979646118e-05, "loss": 0.9389, "step": 226830 }, { "epoch": 19.65, "learning_rate": 2.592083959754828e-05, "loss": 0.7177, "step": 226840 }, { "epoch": 19.65, "learning_rate": 2.5918671215450444e-05, "loss": 0.754, "step": 226850 }, { "epoch": 19.65, "learning_rate": 2.5916502833352603e-05, "loss": 0.8332, "step": 226860 }, { "epoch": 19.65, "learning_rate": 2.5914334451254766e-05, "loss": 0.72, "step": 226870 }, { "epoch": 19.65, "learning_rate": 2.5912166069156933e-05, "loss": 0.8502, "step": 226880 }, { "epoch": 19.65, "learning_rate": 2.5909997687059096e-05, "loss": 0.747, "step": 226890 }, { "epoch": 19.65, "learning_rate": 2.5907829304961255e-05, "loss": 0.8715, "step": 226900 }, { "epoch": 19.65, "learning_rate": 2.5905660922863418e-05, "loss": 0.7542, "step": 226910 }, { "epoch": 19.65, "learning_rate": 2.590349254076558e-05, "loss": 0.7863, "step": 226920 }, { "epoch": 19.65, "learning_rate": 2.5901324158667744e-05, "loss": 0.7012, "step": 226930 }, { "epoch": 19.65, "learning_rate": 2.5899155776569904e-05, "loss": 0.8207, "step": 226940 }, { "epoch": 19.66, "learning_rate": 2.5896987394472067e-05, "loss": 0.8344, "step": 226950 }, { "epoch": 19.66, "learning_rate": 2.5894819012374233e-05, "loss": 0.7624, "step": 226960 }, { "epoch": 19.66, "learning_rate": 2.5892650630276396e-05, "loss": 0.905, "step": 226970 }, { "epoch": 19.66, "learning_rate": 2.5890482248178556e-05, "loss": 0.6453, "step": 226980 }, { "epoch": 19.66, "learning_rate": 2.588831386608072e-05, "loss": 0.8353, "step": 226990 }, { "epoch": 19.66, "learning_rate": 2.5886145483982882e-05, "loss": 0.6698, "step": 227000 }, { "epoch": 19.66, "learning_rate": 2.5883977101885045e-05, "loss": 0.756, "step": 227010 }, { "epoch": 19.66, "learning_rate": 2.5881808719787205e-05, "loss": 0.716, "step": 227020 }, { "epoch": 19.66, "learning_rate": 2.587964033768937e-05, "loss": 0.8379, "step": 227030 }, { "epoch": 19.66, "learning_rate": 2.5877471955591534e-05, "loss": 0.6595, "step": 227040 }, { "epoch": 19.66, "learning_rate": 2.5875303573493697e-05, "loss": 0.7493, "step": 227050 }, { "epoch": 19.66, "learning_rate": 2.5873135191395857e-05, "loss": 0.7105, "step": 227060 }, { "epoch": 19.67, "learning_rate": 2.587096680929802e-05, "loss": 0.7905, "step": 227070 }, { "epoch": 19.67, "learning_rate": 2.5868798427200183e-05, "loss": 0.6831, "step": 227080 }, { "epoch": 19.67, "learning_rate": 2.5866630045102342e-05, "loss": 0.8145, "step": 227090 }, { "epoch": 19.67, "learning_rate": 2.586446166300451e-05, "loss": 0.8606, "step": 227100 }, { "epoch": 19.67, "learning_rate": 2.586229328090667e-05, "loss": 0.7555, "step": 227110 }, { "epoch": 19.67, "learning_rate": 2.5860124898808835e-05, "loss": 0.8122, "step": 227120 }, { "epoch": 19.67, "learning_rate": 2.5857956516710994e-05, "loss": 0.7442, "step": 227130 }, { "epoch": 19.67, "learning_rate": 2.5855788134613157e-05, "loss": 0.8452, "step": 227140 }, { "epoch": 19.67, "learning_rate": 2.585361975251532e-05, "loss": 0.8207, "step": 227150 }, { "epoch": 19.67, "learning_rate": 2.5851451370417487e-05, "loss": 0.6525, "step": 227160 }, { "epoch": 19.67, "learning_rate": 2.5849282988319643e-05, "loss": 0.6577, "step": 227170 }, { "epoch": 19.68, "learning_rate": 2.584711460622181e-05, "loss": 0.9838, "step": 227180 }, { "epoch": 19.68, "learning_rate": 2.5844946224123972e-05, "loss": 0.6793, "step": 227190 }, { "epoch": 19.68, "learning_rate": 2.5842777842026135e-05, "loss": 0.6487, "step": 227200 }, { "epoch": 19.68, "learning_rate": 2.5840609459928295e-05, "loss": 0.8575, "step": 227210 }, { "epoch": 19.68, "learning_rate": 2.5838441077830458e-05, "loss": 0.763, "step": 227220 }, { "epoch": 19.68, "learning_rate": 2.583627269573262e-05, "loss": 0.7526, "step": 227230 }, { "epoch": 19.68, "learning_rate": 2.5834104313634787e-05, "loss": 0.6467, "step": 227240 }, { "epoch": 19.68, "learning_rate": 2.5831935931536947e-05, "loss": 0.7505, "step": 227250 }, { "epoch": 19.68, "learning_rate": 2.582976754943911e-05, "loss": 0.6344, "step": 227260 }, { "epoch": 19.68, "learning_rate": 2.5827599167341273e-05, "loss": 0.8778, "step": 227270 }, { "epoch": 19.68, "learning_rate": 2.5825430785243436e-05, "loss": 0.6693, "step": 227280 }, { "epoch": 19.68, "learning_rate": 2.5823262403145595e-05, "loss": 0.7434, "step": 227290 }, { "epoch": 19.69, "learning_rate": 2.582109402104776e-05, "loss": 0.8289, "step": 227300 }, { "epoch": 19.69, "learning_rate": 2.5818925638949925e-05, "loss": 0.8, "step": 227310 }, { "epoch": 19.69, "learning_rate": 2.5816757256852088e-05, "loss": 0.9175, "step": 227320 }, { "epoch": 19.69, "learning_rate": 2.5814588874754247e-05, "loss": 0.7515, "step": 227330 }, { "epoch": 19.69, "learning_rate": 2.581242049265641e-05, "loss": 0.7895, "step": 227340 }, { "epoch": 19.69, "learning_rate": 2.5810252110558573e-05, "loss": 0.7521, "step": 227350 }, { "epoch": 19.69, "learning_rate": 2.5808083728460733e-05, "loss": 0.7301, "step": 227360 }, { "epoch": 19.69, "learning_rate": 2.5805915346362896e-05, "loss": 0.7579, "step": 227370 }, { "epoch": 19.69, "learning_rate": 2.5803746964265063e-05, "loss": 0.6653, "step": 227380 }, { "epoch": 19.69, "learning_rate": 2.5801578582167226e-05, "loss": 0.6468, "step": 227390 }, { "epoch": 19.69, "learning_rate": 2.5799410200069385e-05, "loss": 0.7183, "step": 227400 }, { "epoch": 19.7, "learning_rate": 2.5797241817971548e-05, "loss": 0.8547, "step": 227410 }, { "epoch": 19.7, "learning_rate": 2.579507343587371e-05, "loss": 0.6485, "step": 227420 }, { "epoch": 19.7, "learning_rate": 2.5792905053775874e-05, "loss": 0.752, "step": 227430 }, { "epoch": 19.7, "learning_rate": 2.5790736671678034e-05, "loss": 0.6723, "step": 227440 }, { "epoch": 19.7, "learning_rate": 2.5788568289580197e-05, "loss": 0.9072, "step": 227450 }, { "epoch": 19.7, "learning_rate": 2.5786399907482363e-05, "loss": 0.7996, "step": 227460 }, { "epoch": 19.7, "learning_rate": 2.5784231525384526e-05, "loss": 0.8107, "step": 227470 }, { "epoch": 19.7, "learning_rate": 2.5782063143286686e-05, "loss": 0.8456, "step": 227480 }, { "epoch": 19.7, "learning_rate": 2.577989476118885e-05, "loss": 0.8623, "step": 227490 }, { "epoch": 19.7, "learning_rate": 2.5777726379091012e-05, "loss": 0.756, "step": 227500 }, { "epoch": 19.7, "learning_rate": 2.5775557996993175e-05, "loss": 0.8639, "step": 227510 }, { "epoch": 19.7, "learning_rate": 2.5773389614895334e-05, "loss": 0.6567, "step": 227520 }, { "epoch": 19.71, "learning_rate": 2.57712212327975e-05, "loss": 0.8301, "step": 227530 }, { "epoch": 19.71, "learning_rate": 2.5769052850699664e-05, "loss": 0.8133, "step": 227540 }, { "epoch": 19.71, "learning_rate": 2.5766884468601827e-05, "loss": 0.8139, "step": 227550 }, { "epoch": 19.71, "learning_rate": 2.5764716086503986e-05, "loss": 0.8306, "step": 227560 }, { "epoch": 19.71, "learning_rate": 2.576254770440615e-05, "loss": 0.7603, "step": 227570 }, { "epoch": 19.71, "learning_rate": 2.5760379322308312e-05, "loss": 0.7101, "step": 227580 }, { "epoch": 19.71, "learning_rate": 2.5758210940210472e-05, "loss": 0.8649, "step": 227590 }, { "epoch": 19.71, "learning_rate": 2.575604255811264e-05, "loss": 0.7592, "step": 227600 }, { "epoch": 19.71, "learning_rate": 2.57538741760148e-05, "loss": 0.6512, "step": 227610 }, { "epoch": 19.71, "learning_rate": 2.5751705793916964e-05, "loss": 0.8503, "step": 227620 }, { "epoch": 19.71, "learning_rate": 2.5749537411819124e-05, "loss": 0.728, "step": 227630 }, { "epoch": 19.72, "learning_rate": 2.5747369029721287e-05, "loss": 0.8077, "step": 227640 }, { "epoch": 19.72, "learning_rate": 2.574520064762345e-05, "loss": 0.7195, "step": 227650 }, { "epoch": 19.72, "learning_rate": 2.5743032265525616e-05, "loss": 0.7566, "step": 227660 }, { "epoch": 19.72, "learning_rate": 2.5740863883427776e-05, "loss": 0.6935, "step": 227670 }, { "epoch": 19.72, "learning_rate": 2.573869550132994e-05, "loss": 0.8778, "step": 227680 }, { "epoch": 19.72, "learning_rate": 2.5736527119232102e-05, "loss": 0.7554, "step": 227690 }, { "epoch": 19.72, "learning_rate": 2.5734358737134265e-05, "loss": 0.7588, "step": 227700 }, { "epoch": 19.72, "learning_rate": 2.5732190355036425e-05, "loss": 0.8868, "step": 227710 }, { "epoch": 19.72, "learning_rate": 2.5730021972938588e-05, "loss": 0.8721, "step": 227720 }, { "epoch": 19.72, "learning_rate": 2.572785359084075e-05, "loss": 0.6577, "step": 227730 }, { "epoch": 19.72, "learning_rate": 2.5725685208742917e-05, "loss": 0.7087, "step": 227740 }, { "epoch": 19.72, "learning_rate": 2.5723516826645077e-05, "loss": 0.7586, "step": 227750 }, { "epoch": 19.73, "learning_rate": 2.572134844454724e-05, "loss": 0.844, "step": 227760 }, { "epoch": 19.73, "learning_rate": 2.5719180062449403e-05, "loss": 1.0464, "step": 227770 }, { "epoch": 19.73, "learning_rate": 2.5717011680351566e-05, "loss": 0.6929, "step": 227780 }, { "epoch": 19.73, "learning_rate": 2.5714843298253725e-05, "loss": 0.7112, "step": 227790 }, { "epoch": 19.73, "learning_rate": 2.571267491615589e-05, "loss": 0.7117, "step": 227800 }, { "epoch": 19.73, "learning_rate": 2.5710506534058055e-05, "loss": 0.6601, "step": 227810 }, { "epoch": 19.73, "learning_rate": 2.5708338151960218e-05, "loss": 0.7766, "step": 227820 }, { "epoch": 19.73, "learning_rate": 2.5706169769862377e-05, "loss": 0.8688, "step": 227830 }, { "epoch": 19.73, "learning_rate": 2.570400138776454e-05, "loss": 0.7797, "step": 227840 }, { "epoch": 19.73, "learning_rate": 2.5701833005666703e-05, "loss": 0.8057, "step": 227850 }, { "epoch": 19.73, "learning_rate": 2.5699664623568863e-05, "loss": 0.7543, "step": 227860 }, { "epoch": 19.73, "learning_rate": 2.5697496241471026e-05, "loss": 0.6864, "step": 227870 }, { "epoch": 19.74, "learning_rate": 2.5695327859373192e-05, "loss": 0.8516, "step": 227880 }, { "epoch": 19.74, "learning_rate": 2.5693159477275355e-05, "loss": 0.8035, "step": 227890 }, { "epoch": 19.74, "learning_rate": 2.5690991095177515e-05, "loss": 0.8178, "step": 227900 }, { "epoch": 19.74, "learning_rate": 2.5688822713079678e-05, "loss": 0.7153, "step": 227910 }, { "epoch": 19.74, "learning_rate": 2.568665433098184e-05, "loss": 0.8219, "step": 227920 }, { "epoch": 19.74, "learning_rate": 2.5684485948884004e-05, "loss": 0.9746, "step": 227930 }, { "epoch": 19.74, "learning_rate": 2.5682317566786164e-05, "loss": 0.8241, "step": 227940 }, { "epoch": 19.74, "learning_rate": 2.568014918468833e-05, "loss": 0.7886, "step": 227950 }, { "epoch": 19.74, "learning_rate": 2.5677980802590493e-05, "loss": 0.8525, "step": 227960 }, { "epoch": 19.74, "learning_rate": 2.5675812420492656e-05, "loss": 0.7506, "step": 227970 }, { "epoch": 19.74, "learning_rate": 2.5673644038394816e-05, "loss": 0.7419, "step": 227980 }, { "epoch": 19.75, "learning_rate": 2.567147565629698e-05, "loss": 0.8179, "step": 227990 }, { "epoch": 19.75, "learning_rate": 2.566930727419914e-05, "loss": 0.7629, "step": 228000 }, { "epoch": 19.75, "learning_rate": 2.5667138892101305e-05, "loss": 0.6086, "step": 228010 }, { "epoch": 19.75, "learning_rate": 2.5664970510003464e-05, "loss": 0.6432, "step": 228020 }, { "epoch": 19.75, "learning_rate": 2.566280212790563e-05, "loss": 0.9589, "step": 228030 }, { "epoch": 19.75, "learning_rate": 2.5660633745807794e-05, "loss": 0.8542, "step": 228040 }, { "epoch": 19.75, "learning_rate": 2.5658465363709957e-05, "loss": 0.8933, "step": 228050 }, { "epoch": 19.75, "learning_rate": 2.5656296981612116e-05, "loss": 0.6928, "step": 228060 }, { "epoch": 19.75, "learning_rate": 2.565412859951428e-05, "loss": 0.834, "step": 228070 }, { "epoch": 19.75, "learning_rate": 2.5651960217416442e-05, "loss": 0.7305, "step": 228080 }, { "epoch": 19.75, "learning_rate": 2.564979183531861e-05, "loss": 0.8925, "step": 228090 }, { "epoch": 19.75, "learning_rate": 2.5647623453220768e-05, "loss": 0.7341, "step": 228100 }, { "epoch": 19.76, "learning_rate": 2.564545507112293e-05, "loss": 0.6811, "step": 228110 }, { "epoch": 19.76, "learning_rate": 2.5643286689025094e-05, "loss": 0.7435, "step": 228120 }, { "epoch": 19.76, "learning_rate": 2.5641118306927254e-05, "loss": 0.7432, "step": 228130 }, { "epoch": 19.76, "learning_rate": 2.5638949924829417e-05, "loss": 0.7566, "step": 228140 }, { "epoch": 19.76, "learning_rate": 2.563678154273158e-05, "loss": 0.7841, "step": 228150 }, { "epoch": 19.76, "learning_rate": 2.5634613160633746e-05, "loss": 0.7806, "step": 228160 }, { "epoch": 19.76, "learning_rate": 2.5632444778535906e-05, "loss": 0.7208, "step": 228170 }, { "epoch": 19.76, "learning_rate": 2.563027639643807e-05, "loss": 0.7523, "step": 228180 }, { "epoch": 19.76, "learning_rate": 2.5628108014340232e-05, "loss": 0.9192, "step": 228190 }, { "epoch": 19.76, "learning_rate": 2.5625939632242395e-05, "loss": 0.7927, "step": 228200 }, { "epoch": 19.76, "learning_rate": 2.5623771250144555e-05, "loss": 0.6787, "step": 228210 }, { "epoch": 19.77, "learning_rate": 2.5621602868046718e-05, "loss": 0.7346, "step": 228220 }, { "epoch": 19.77, "learning_rate": 2.5619434485948884e-05, "loss": 0.7613, "step": 228230 }, { "epoch": 19.77, "learning_rate": 2.5617266103851047e-05, "loss": 0.754, "step": 228240 }, { "epoch": 19.77, "learning_rate": 2.5615097721753207e-05, "loss": 0.8621, "step": 228250 }, { "epoch": 19.77, "learning_rate": 2.561292933965537e-05, "loss": 0.6524, "step": 228260 }, { "epoch": 19.77, "learning_rate": 2.5610760957557533e-05, "loss": 0.8243, "step": 228270 }, { "epoch": 19.77, "learning_rate": 2.5608592575459696e-05, "loss": 0.7693, "step": 228280 }, { "epoch": 19.77, "learning_rate": 2.5606424193361855e-05, "loss": 0.7391, "step": 228290 }, { "epoch": 19.77, "learning_rate": 2.5604255811264018e-05, "loss": 0.633, "step": 228300 }, { "epoch": 19.77, "learning_rate": 2.5602087429166185e-05, "loss": 0.7621, "step": 228310 }, { "epoch": 19.77, "learning_rate": 2.5599919047068348e-05, "loss": 0.6391, "step": 228320 }, { "epoch": 19.77, "learning_rate": 2.5597750664970507e-05, "loss": 0.7845, "step": 228330 }, { "epoch": 19.78, "learning_rate": 2.559558228287267e-05, "loss": 0.8639, "step": 228340 }, { "epoch": 19.78, "learning_rate": 2.5593413900774833e-05, "loss": 0.7401, "step": 228350 }, { "epoch": 19.78, "learning_rate": 2.5591245518676993e-05, "loss": 0.9018, "step": 228360 }, { "epoch": 19.78, "learning_rate": 2.5589077136579156e-05, "loss": 0.7585, "step": 228370 }, { "epoch": 19.78, "learning_rate": 2.5586908754481322e-05, "loss": 0.6745, "step": 228380 }, { "epoch": 19.78, "learning_rate": 2.5584740372383485e-05, "loss": 0.8126, "step": 228390 }, { "epoch": 19.78, "learning_rate": 2.5582571990285645e-05, "loss": 0.8165, "step": 228400 }, { "epoch": 19.78, "learning_rate": 2.5580403608187808e-05, "loss": 0.8662, "step": 228410 }, { "epoch": 19.78, "learning_rate": 2.557823522608997e-05, "loss": 0.8194, "step": 228420 }, { "epoch": 19.78, "learning_rate": 2.5576066843992134e-05, "loss": 0.9146, "step": 228430 }, { "epoch": 19.78, "learning_rate": 2.5573898461894293e-05, "loss": 0.7896, "step": 228440 }, { "epoch": 19.79, "learning_rate": 2.557173007979646e-05, "loss": 0.7992, "step": 228450 }, { "epoch": 19.79, "learning_rate": 2.5569561697698623e-05, "loss": 0.8795, "step": 228460 }, { "epoch": 19.79, "learning_rate": 2.5567393315600786e-05, "loss": 0.8681, "step": 228470 }, { "epoch": 19.79, "learning_rate": 2.5565224933502945e-05, "loss": 0.76, "step": 228480 }, { "epoch": 19.79, "learning_rate": 2.556305655140511e-05, "loss": 0.7857, "step": 228490 }, { "epoch": 19.79, "learning_rate": 2.556088816930727e-05, "loss": 0.8031, "step": 228500 }, { "epoch": 19.79, "learning_rate": 2.5558719787209438e-05, "loss": 0.5856, "step": 228510 }, { "epoch": 19.79, "learning_rate": 2.5556551405111594e-05, "loss": 0.8678, "step": 228520 }, { "epoch": 19.79, "learning_rate": 2.555438302301376e-05, "loss": 0.8458, "step": 228530 }, { "epoch": 19.79, "learning_rate": 2.5552214640915923e-05, "loss": 0.6355, "step": 228540 }, { "epoch": 19.79, "learning_rate": 2.5550046258818086e-05, "loss": 0.6901, "step": 228550 }, { "epoch": 19.79, "learning_rate": 2.5547877876720246e-05, "loss": 0.6726, "step": 228560 }, { "epoch": 19.8, "learning_rate": 2.554570949462241e-05, "loss": 0.7807, "step": 228570 }, { "epoch": 19.8, "learning_rate": 2.5543541112524572e-05, "loss": 0.7, "step": 228580 }, { "epoch": 19.8, "learning_rate": 2.554137273042674e-05, "loss": 0.8399, "step": 228590 }, { "epoch": 19.8, "learning_rate": 2.5539204348328898e-05, "loss": 0.5311, "step": 228600 }, { "epoch": 19.8, "learning_rate": 2.553703596623106e-05, "loss": 0.8782, "step": 228610 }, { "epoch": 19.8, "learning_rate": 2.5534867584133224e-05, "loss": 0.6587, "step": 228620 }, { "epoch": 19.8, "learning_rate": 2.5532699202035384e-05, "loss": 0.6327, "step": 228630 }, { "epoch": 19.8, "learning_rate": 2.5530530819937547e-05, "loss": 0.6879, "step": 228640 }, { "epoch": 19.8, "learning_rate": 2.552836243783971e-05, "loss": 0.8153, "step": 228650 }, { "epoch": 19.8, "learning_rate": 2.5526194055741876e-05, "loss": 0.8975, "step": 228660 }, { "epoch": 19.8, "learning_rate": 2.5524025673644036e-05, "loss": 0.7864, "step": 228670 }, { "epoch": 19.81, "learning_rate": 2.55218572915462e-05, "loss": 0.7556, "step": 228680 }, { "epoch": 19.81, "learning_rate": 2.5519688909448362e-05, "loss": 0.8026, "step": 228690 }, { "epoch": 19.81, "learning_rate": 2.5517520527350525e-05, "loss": 0.9306, "step": 228700 }, { "epoch": 19.81, "learning_rate": 2.5515352145252684e-05, "loss": 0.8483, "step": 228710 }, { "epoch": 19.81, "learning_rate": 2.5513183763154847e-05, "loss": 0.6715, "step": 228720 }, { "epoch": 19.81, "learning_rate": 2.5511015381057014e-05, "loss": 0.7086, "step": 228730 }, { "epoch": 19.81, "learning_rate": 2.5508846998959177e-05, "loss": 0.6795, "step": 228740 }, { "epoch": 19.81, "learning_rate": 2.5506678616861336e-05, "loss": 0.6535, "step": 228750 }, { "epoch": 19.81, "learning_rate": 2.55045102347635e-05, "loss": 0.7277, "step": 228760 }, { "epoch": 19.81, "learning_rate": 2.5502341852665662e-05, "loss": 0.5595, "step": 228770 }, { "epoch": 19.81, "learning_rate": 2.5500173470567825e-05, "loss": 0.8361, "step": 228780 }, { "epoch": 19.81, "learning_rate": 2.5498005088469985e-05, "loss": 0.7049, "step": 228790 }, { "epoch": 19.82, "learning_rate": 2.5495836706372148e-05, "loss": 0.8327, "step": 228800 }, { "epoch": 19.82, "learning_rate": 2.5493668324274314e-05, "loss": 0.6324, "step": 228810 }, { "epoch": 19.82, "learning_rate": 2.5491499942176477e-05, "loss": 0.9962, "step": 228820 }, { "epoch": 19.82, "learning_rate": 2.5489331560078637e-05, "loss": 0.7749, "step": 228830 }, { "epoch": 19.82, "learning_rate": 2.54871631779808e-05, "loss": 0.6527, "step": 228840 }, { "epoch": 19.82, "learning_rate": 2.5484994795882963e-05, "loss": 0.7348, "step": 228850 }, { "epoch": 19.82, "learning_rate": 2.5482826413785126e-05, "loss": 0.7985, "step": 228860 }, { "epoch": 19.82, "learning_rate": 2.5480658031687286e-05, "loss": 0.7726, "step": 228870 }, { "epoch": 19.82, "learning_rate": 2.5478489649589452e-05, "loss": 0.6588, "step": 228880 }, { "epoch": 19.82, "learning_rate": 2.5476321267491615e-05, "loss": 0.6757, "step": 228890 }, { "epoch": 19.82, "learning_rate": 2.5474152885393775e-05, "loss": 0.9233, "step": 228900 }, { "epoch": 19.83, "learning_rate": 2.5471984503295938e-05, "loss": 0.7711, "step": 228910 }, { "epoch": 19.83, "learning_rate": 2.54698161211981e-05, "loss": 0.6767, "step": 228920 }, { "epoch": 19.83, "learning_rate": 2.5467647739100264e-05, "loss": 0.746, "step": 228930 }, { "epoch": 19.83, "learning_rate": 2.5465479357002423e-05, "loss": 0.8844, "step": 228940 }, { "epoch": 19.83, "learning_rate": 2.546331097490459e-05, "loss": 0.7872, "step": 228950 }, { "epoch": 19.83, "learning_rate": 2.5461142592806753e-05, "loss": 0.7926, "step": 228960 }, { "epoch": 19.83, "learning_rate": 2.5458974210708916e-05, "loss": 0.8359, "step": 228970 }, { "epoch": 19.83, "learning_rate": 2.5456805828611075e-05, "loss": 0.7192, "step": 228980 }, { "epoch": 19.83, "learning_rate": 2.545463744651324e-05, "loss": 0.8984, "step": 228990 }, { "epoch": 19.83, "learning_rate": 2.54524690644154e-05, "loss": 0.6811, "step": 229000 }, { "epoch": 19.83, "learning_rate": 2.5450300682317568e-05, "loss": 0.8034, "step": 229010 }, { "epoch": 19.83, "learning_rate": 2.5448132300219727e-05, "loss": 0.7423, "step": 229020 }, { "epoch": 19.84, "learning_rate": 2.544596391812189e-05, "loss": 0.8168, "step": 229030 }, { "epoch": 19.84, "learning_rate": 2.5443795536024053e-05, "loss": 0.7234, "step": 229040 }, { "epoch": 19.84, "learning_rate": 2.5441627153926216e-05, "loss": 0.6864, "step": 229050 }, { "epoch": 19.84, "learning_rate": 2.5439458771828376e-05, "loss": 0.7, "step": 229060 }, { "epoch": 19.84, "learning_rate": 2.543729038973054e-05, "loss": 0.8133, "step": 229070 }, { "epoch": 19.84, "learning_rate": 2.5435122007632702e-05, "loss": 0.8633, "step": 229080 }, { "epoch": 19.84, "learning_rate": 2.543295362553487e-05, "loss": 0.6184, "step": 229090 }, { "epoch": 19.84, "learning_rate": 2.5430785243437028e-05, "loss": 0.7607, "step": 229100 }, { "epoch": 19.84, "learning_rate": 2.542861686133919e-05, "loss": 0.8802, "step": 229110 }, { "epoch": 19.84, "learning_rate": 2.5426448479241354e-05, "loss": 0.8255, "step": 229120 }, { "epoch": 19.84, "learning_rate": 2.5424280097143514e-05, "loss": 0.813, "step": 229130 }, { "epoch": 19.84, "learning_rate": 2.5422111715045677e-05, "loss": 0.7892, "step": 229140 }, { "epoch": 19.85, "learning_rate": 2.541994333294784e-05, "loss": 0.6182, "step": 229150 }, { "epoch": 19.85, "learning_rate": 2.5417774950850006e-05, "loss": 0.711, "step": 229160 }, { "epoch": 19.85, "learning_rate": 2.5415606568752166e-05, "loss": 0.8131, "step": 229170 }, { "epoch": 19.85, "learning_rate": 2.541343818665433e-05, "loss": 0.5419, "step": 229180 }, { "epoch": 19.85, "learning_rate": 2.541126980455649e-05, "loss": 0.5565, "step": 229190 }, { "epoch": 19.85, "learning_rate": 2.5409101422458655e-05, "loss": 0.7043, "step": 229200 }, { "epoch": 19.85, "learning_rate": 2.5406933040360814e-05, "loss": 0.8602, "step": 229210 }, { "epoch": 19.85, "learning_rate": 2.5404764658262977e-05, "loss": 0.7564, "step": 229220 }, { "epoch": 19.85, "learning_rate": 2.5402596276165144e-05, "loss": 0.878, "step": 229230 }, { "epoch": 19.85, "learning_rate": 2.5400427894067307e-05, "loss": 0.707, "step": 229240 }, { "epoch": 19.85, "learning_rate": 2.5398259511969466e-05, "loss": 0.8192, "step": 229250 }, { "epoch": 19.86, "learning_rate": 2.539609112987163e-05, "loss": 0.5943, "step": 229260 }, { "epoch": 19.86, "learning_rate": 2.5393922747773792e-05, "loss": 0.762, "step": 229270 }, { "epoch": 19.86, "learning_rate": 2.5391754365675955e-05, "loss": 0.7437, "step": 229280 }, { "epoch": 19.86, "learning_rate": 2.5389585983578115e-05, "loss": 0.8658, "step": 229290 }, { "epoch": 19.86, "learning_rate": 2.538741760148028e-05, "loss": 0.7805, "step": 229300 }, { "epoch": 19.86, "learning_rate": 2.5385249219382444e-05, "loss": 0.8081, "step": 229310 }, { "epoch": 19.86, "learning_rate": 2.5383080837284607e-05, "loss": 0.708, "step": 229320 }, { "epoch": 19.86, "learning_rate": 2.5380912455186767e-05, "loss": 0.7528, "step": 229330 }, { "epoch": 19.86, "learning_rate": 2.537874407308893e-05, "loss": 0.7176, "step": 229340 }, { "epoch": 19.86, "learning_rate": 2.5376575690991093e-05, "loss": 0.7068, "step": 229350 }, { "epoch": 19.86, "learning_rate": 2.5374407308893256e-05, "loss": 0.8378, "step": 229360 }, { "epoch": 19.86, "learning_rate": 2.5372238926795416e-05, "loss": 0.7465, "step": 229370 }, { "epoch": 19.87, "learning_rate": 2.5370070544697582e-05, "loss": 0.7105, "step": 229380 }, { "epoch": 19.87, "learning_rate": 2.5367902162599745e-05, "loss": 0.6571, "step": 229390 }, { "epoch": 19.87, "learning_rate": 2.5365733780501905e-05, "loss": 0.8632, "step": 229400 }, { "epoch": 19.87, "learning_rate": 2.5363565398404068e-05, "loss": 0.8602, "step": 229410 }, { "epoch": 19.87, "learning_rate": 2.536139701630623e-05, "loss": 0.7302, "step": 229420 }, { "epoch": 19.87, "learning_rate": 2.5359228634208394e-05, "loss": 0.7179, "step": 229430 }, { "epoch": 19.87, "learning_rate": 2.5357060252110553e-05, "loss": 0.844, "step": 229440 }, { "epoch": 19.87, "learning_rate": 2.535489187001272e-05, "loss": 0.7768, "step": 229450 }, { "epoch": 19.87, "learning_rate": 2.5352723487914883e-05, "loss": 0.8823, "step": 229460 }, { "epoch": 19.87, "learning_rate": 2.5350555105817046e-05, "loss": 0.8389, "step": 229470 }, { "epoch": 19.87, "learning_rate": 2.5348386723719205e-05, "loss": 0.7225, "step": 229480 }, { "epoch": 19.88, "learning_rate": 2.5346218341621368e-05, "loss": 0.753, "step": 229490 }, { "epoch": 19.88, "learning_rate": 2.534404995952353e-05, "loss": 0.7639, "step": 229500 }, { "epoch": 19.88, "learning_rate": 2.5341881577425698e-05, "loss": 0.8911, "step": 229510 }, { "epoch": 19.88, "learning_rate": 2.5339713195327857e-05, "loss": 0.8425, "step": 229520 }, { "epoch": 19.88, "learning_rate": 2.533754481323002e-05, "loss": 0.8145, "step": 229530 }, { "epoch": 19.88, "learning_rate": 2.5335376431132183e-05, "loss": 0.7159, "step": 229540 }, { "epoch": 19.88, "learning_rate": 2.5333208049034346e-05, "loss": 0.8599, "step": 229550 }, { "epoch": 19.88, "learning_rate": 2.5331039666936506e-05, "loss": 0.7396, "step": 229560 }, { "epoch": 19.88, "learning_rate": 2.532887128483867e-05, "loss": 0.7574, "step": 229570 }, { "epoch": 19.88, "learning_rate": 2.5326702902740835e-05, "loss": 0.7699, "step": 229580 }, { "epoch": 19.88, "learning_rate": 2.5324534520642998e-05, "loss": 0.6413, "step": 229590 }, { "epoch": 19.88, "learning_rate": 2.5322366138545158e-05, "loss": 0.8366, "step": 229600 }, { "epoch": 19.89, "learning_rate": 2.532019775644732e-05, "loss": 0.7865, "step": 229610 }, { "epoch": 19.89, "learning_rate": 2.5318029374349484e-05, "loss": 0.6866, "step": 229620 }, { "epoch": 19.89, "learning_rate": 2.5315860992251647e-05, "loss": 0.7774, "step": 229630 }, { "epoch": 19.89, "learning_rate": 2.5313692610153806e-05, "loss": 0.824, "step": 229640 }, { "epoch": 19.89, "learning_rate": 2.531152422805597e-05, "loss": 0.8849, "step": 229650 }, { "epoch": 19.89, "learning_rate": 2.5309355845958136e-05, "loss": 0.7085, "step": 229660 }, { "epoch": 19.89, "learning_rate": 2.5307187463860295e-05, "loss": 0.9189, "step": 229670 }, { "epoch": 19.89, "learning_rate": 2.530501908176246e-05, "loss": 0.8738, "step": 229680 }, { "epoch": 19.89, "learning_rate": 2.530285069966462e-05, "loss": 0.8852, "step": 229690 }, { "epoch": 19.89, "learning_rate": 2.5300682317566784e-05, "loss": 0.7445, "step": 229700 }, { "epoch": 19.89, "learning_rate": 2.5298513935468944e-05, "loss": 0.7397, "step": 229710 }, { "epoch": 19.9, "learning_rate": 2.5296345553371107e-05, "loss": 0.8087, "step": 229720 }, { "epoch": 19.9, "learning_rate": 2.5294177171273273e-05, "loss": 0.6574, "step": 229730 }, { "epoch": 19.9, "learning_rate": 2.5292008789175436e-05, "loss": 0.7898, "step": 229740 }, { "epoch": 19.9, "learning_rate": 2.5289840407077596e-05, "loss": 0.6603, "step": 229750 }, { "epoch": 19.9, "learning_rate": 2.528767202497976e-05, "loss": 0.6004, "step": 229760 }, { "epoch": 19.9, "learning_rate": 2.5285503642881922e-05, "loss": 0.8955, "step": 229770 }, { "epoch": 19.9, "learning_rate": 2.5283335260784085e-05, "loss": 0.703, "step": 229780 }, { "epoch": 19.9, "learning_rate": 2.5281166878686245e-05, "loss": 0.5504, "step": 229790 }, { "epoch": 19.9, "learning_rate": 2.527899849658841e-05, "loss": 0.7501, "step": 229800 }, { "epoch": 19.9, "learning_rate": 2.5276830114490574e-05, "loss": 0.7845, "step": 229810 }, { "epoch": 19.9, "learning_rate": 2.5274661732392737e-05, "loss": 0.7732, "step": 229820 }, { "epoch": 19.9, "learning_rate": 2.5272493350294897e-05, "loss": 0.8352, "step": 229830 }, { "epoch": 19.91, "learning_rate": 2.527032496819706e-05, "loss": 0.5833, "step": 229840 }, { "epoch": 19.91, "learning_rate": 2.5268156586099223e-05, "loss": 0.703, "step": 229850 }, { "epoch": 19.91, "learning_rate": 2.526598820400139e-05, "loss": 0.64, "step": 229860 }, { "epoch": 19.91, "learning_rate": 2.5263819821903545e-05, "loss": 0.7141, "step": 229870 }, { "epoch": 19.91, "learning_rate": 2.5261651439805712e-05, "loss": 0.8054, "step": 229880 }, { "epoch": 19.91, "learning_rate": 2.5259483057707875e-05, "loss": 0.8386, "step": 229890 }, { "epoch": 19.91, "learning_rate": 2.5257314675610034e-05, "loss": 0.635, "step": 229900 }, { "epoch": 19.91, "learning_rate": 2.5255146293512197e-05, "loss": 0.7924, "step": 229910 }, { "epoch": 19.91, "learning_rate": 2.525297791141436e-05, "loss": 0.8533, "step": 229920 }, { "epoch": 19.91, "learning_rate": 2.5250809529316523e-05, "loss": 0.6466, "step": 229930 }, { "epoch": 19.91, "learning_rate": 2.5248641147218683e-05, "loss": 0.6122, "step": 229940 }, { "epoch": 19.92, "learning_rate": 2.524647276512085e-05, "loss": 0.6925, "step": 229950 }, { "epoch": 19.92, "learning_rate": 2.5244304383023012e-05, "loss": 0.7289, "step": 229960 }, { "epoch": 19.92, "learning_rate": 2.5242136000925175e-05, "loss": 0.6559, "step": 229970 }, { "epoch": 19.92, "learning_rate": 2.5239967618827335e-05, "loss": 0.8171, "step": 229980 }, { "epoch": 19.92, "learning_rate": 2.5237799236729498e-05, "loss": 0.6684, "step": 229990 }, { "epoch": 19.92, "learning_rate": 2.523563085463166e-05, "loss": 0.7899, "step": 230000 }, { "epoch": 19.92, "learning_rate": 2.5233462472533827e-05, "loss": 0.6559, "step": 230010 }, { "epoch": 19.92, "learning_rate": 2.5231294090435987e-05, "loss": 0.8598, "step": 230020 }, { "epoch": 19.92, "learning_rate": 2.522912570833815e-05, "loss": 0.7339, "step": 230030 }, { "epoch": 19.92, "learning_rate": 2.5226957326240313e-05, "loss": 0.6561, "step": 230040 }, { "epoch": 19.92, "learning_rate": 2.5224788944142476e-05, "loss": 0.7687, "step": 230050 }, { "epoch": 19.92, "learning_rate": 2.5222620562044636e-05, "loss": 0.6964, "step": 230060 }, { "epoch": 19.93, "learning_rate": 2.52204521799468e-05, "loss": 0.7157, "step": 230070 }, { "epoch": 19.93, "learning_rate": 2.5218283797848965e-05, "loss": 0.7932, "step": 230080 }, { "epoch": 19.93, "learning_rate": 2.5216115415751128e-05, "loss": 0.5992, "step": 230090 }, { "epoch": 19.93, "learning_rate": 2.5213947033653288e-05, "loss": 0.6641, "step": 230100 }, { "epoch": 19.93, "learning_rate": 2.521177865155545e-05, "loss": 0.8056, "step": 230110 }, { "epoch": 19.93, "learning_rate": 2.5209610269457614e-05, "loss": 0.7958, "step": 230120 }, { "epoch": 19.93, "learning_rate": 2.5207441887359777e-05, "loss": 0.8427, "step": 230130 }, { "epoch": 19.93, "learning_rate": 2.5205273505261936e-05, "loss": 0.8104, "step": 230140 }, { "epoch": 19.93, "learning_rate": 2.52031051231641e-05, "loss": 0.6803, "step": 230150 }, { "epoch": 19.93, "learning_rate": 2.5200936741066266e-05, "loss": 0.6711, "step": 230160 }, { "epoch": 19.93, "learning_rate": 2.5198768358968425e-05, "loss": 0.9166, "step": 230170 }, { "epoch": 19.94, "learning_rate": 2.5196599976870588e-05, "loss": 0.7987, "step": 230180 }, { "epoch": 19.94, "learning_rate": 2.519443159477275e-05, "loss": 0.7604, "step": 230190 }, { "epoch": 19.94, "learning_rate": 2.5192263212674914e-05, "loss": 1.0702, "step": 230200 }, { "epoch": 19.94, "learning_rate": 2.5190094830577074e-05, "loss": 0.8176, "step": 230210 }, { "epoch": 19.94, "learning_rate": 2.5187926448479237e-05, "loss": 0.7213, "step": 230220 }, { "epoch": 19.94, "learning_rate": 2.5185758066381403e-05, "loss": 0.6054, "step": 230230 }, { "epoch": 19.94, "learning_rate": 2.5183589684283566e-05, "loss": 0.8984, "step": 230240 }, { "epoch": 19.94, "learning_rate": 2.5181421302185726e-05, "loss": 0.7562, "step": 230250 }, { "epoch": 19.94, "learning_rate": 2.517925292008789e-05, "loss": 0.8938, "step": 230260 }, { "epoch": 19.94, "learning_rate": 2.5177084537990052e-05, "loss": 0.8083, "step": 230270 }, { "epoch": 19.94, "learning_rate": 2.5174916155892215e-05, "loss": 0.795, "step": 230280 }, { "epoch": 19.94, "learning_rate": 2.5172747773794375e-05, "loss": 0.8423, "step": 230290 }, { "epoch": 19.95, "learning_rate": 2.517057939169654e-05, "loss": 0.8203, "step": 230300 }, { "epoch": 19.95, "learning_rate": 2.5168411009598704e-05, "loss": 0.7345, "step": 230310 }, { "epoch": 19.95, "learning_rate": 2.5166242627500867e-05, "loss": 0.9558, "step": 230320 }, { "epoch": 19.95, "learning_rate": 2.5164074245403027e-05, "loss": 0.7276, "step": 230330 }, { "epoch": 19.95, "learning_rate": 2.516190586330519e-05, "loss": 0.7746, "step": 230340 }, { "epoch": 19.95, "learning_rate": 2.5159737481207353e-05, "loss": 0.8727, "step": 230350 }, { "epoch": 19.95, "learning_rate": 2.515756909910952e-05, "loss": 0.8662, "step": 230360 }, { "epoch": 19.95, "learning_rate": 2.515540071701168e-05, "loss": 0.5989, "step": 230370 }, { "epoch": 19.95, "learning_rate": 2.515323233491384e-05, "loss": 0.712, "step": 230380 }, { "epoch": 19.95, "learning_rate": 2.5151063952816005e-05, "loss": 0.867, "step": 230390 }, { "epoch": 19.95, "learning_rate": 2.5148895570718168e-05, "loss": 0.7902, "step": 230400 }, { "epoch": 19.95, "learning_rate": 2.5146727188620327e-05, "loss": 0.878, "step": 230410 }, { "epoch": 19.96, "learning_rate": 2.514455880652249e-05, "loss": 0.7713, "step": 230420 }, { "epoch": 19.96, "learning_rate": 2.5142390424424653e-05, "loss": 0.8045, "step": 230430 }, { "epoch": 19.96, "learning_rate": 2.5140222042326813e-05, "loss": 0.7751, "step": 230440 }, { "epoch": 19.96, "learning_rate": 2.513805366022898e-05, "loss": 0.6494, "step": 230450 }, { "epoch": 19.96, "learning_rate": 2.5135885278131142e-05, "loss": 0.7553, "step": 230460 }, { "epoch": 19.96, "learning_rate": 2.5133716896033305e-05, "loss": 0.9615, "step": 230470 }, { "epoch": 19.96, "learning_rate": 2.5131548513935465e-05, "loss": 0.7672, "step": 230480 }, { "epoch": 19.96, "learning_rate": 2.5129380131837628e-05, "loss": 0.7537, "step": 230490 }, { "epoch": 19.96, "learning_rate": 2.512721174973979e-05, "loss": 0.7864, "step": 230500 }, { "epoch": 19.96, "learning_rate": 2.5125043367641957e-05, "loss": 0.6585, "step": 230510 }, { "epoch": 19.96, "learning_rate": 2.5122874985544117e-05, "loss": 0.7627, "step": 230520 }, { "epoch": 19.97, "learning_rate": 2.512070660344628e-05, "loss": 0.7643, "step": 230530 }, { "epoch": 19.97, "learning_rate": 2.5118538221348443e-05, "loss": 0.7315, "step": 230540 }, { "epoch": 19.97, "learning_rate": 2.5116369839250606e-05, "loss": 0.8457, "step": 230550 }, { "epoch": 19.97, "learning_rate": 2.5114201457152765e-05, "loss": 0.7004, "step": 230560 }, { "epoch": 19.97, "learning_rate": 2.511203307505493e-05, "loss": 0.8746, "step": 230570 }, { "epoch": 19.97, "learning_rate": 2.5109864692957095e-05, "loss": 0.9976, "step": 230580 }, { "epoch": 19.97, "learning_rate": 2.5107696310859258e-05, "loss": 0.7756, "step": 230590 }, { "epoch": 19.97, "learning_rate": 2.5105527928761418e-05, "loss": 0.8129, "step": 230600 }, { "epoch": 19.97, "learning_rate": 2.510335954666358e-05, "loss": 0.8082, "step": 230610 }, { "epoch": 19.97, "learning_rate": 2.5101191164565744e-05, "loss": 0.8954, "step": 230620 }, { "epoch": 19.97, "learning_rate": 2.5099022782467907e-05, "loss": 0.7447, "step": 230630 }, { "epoch": 19.97, "learning_rate": 2.5096854400370066e-05, "loss": 0.8715, "step": 230640 }, { "epoch": 19.98, "learning_rate": 2.5094686018272233e-05, "loss": 0.8791, "step": 230650 }, { "epoch": 19.98, "learning_rate": 2.5092517636174396e-05, "loss": 0.7556, "step": 230660 }, { "epoch": 19.98, "learning_rate": 2.5090349254076555e-05, "loss": 0.8462, "step": 230670 }, { "epoch": 19.98, "learning_rate": 2.5088180871978718e-05, "loss": 0.8061, "step": 230680 }, { "epoch": 19.98, "learning_rate": 2.508601248988088e-05, "loss": 0.834, "step": 230690 }, { "epoch": 19.98, "learning_rate": 2.5083844107783044e-05, "loss": 0.9195, "step": 230700 }, { "epoch": 19.98, "learning_rate": 2.5081675725685204e-05, "loss": 0.738, "step": 230710 }, { "epoch": 19.98, "learning_rate": 2.5079507343587367e-05, "loss": 0.8226, "step": 230720 }, { "epoch": 19.98, "learning_rate": 2.5077338961489533e-05, "loss": 0.6706, "step": 230730 }, { "epoch": 19.98, "learning_rate": 2.5075170579391696e-05, "loss": 0.7243, "step": 230740 }, { "epoch": 19.98, "learning_rate": 2.5073002197293856e-05, "loss": 0.7333, "step": 230750 }, { "epoch": 19.99, "learning_rate": 2.507083381519602e-05, "loss": 0.7112, "step": 230760 }, { "epoch": 19.99, "learning_rate": 2.5068665433098182e-05, "loss": 0.6818, "step": 230770 }, { "epoch": 19.99, "learning_rate": 2.5066497051000345e-05, "loss": 0.8288, "step": 230780 }, { "epoch": 19.99, "learning_rate": 2.5064328668902504e-05, "loss": 0.8093, "step": 230790 }, { "epoch": 19.99, "learning_rate": 2.506216028680467e-05, "loss": 0.5969, "step": 230800 }, { "epoch": 19.99, "learning_rate": 2.5059991904706834e-05, "loss": 0.7409, "step": 230810 }, { "epoch": 19.99, "learning_rate": 2.5057823522608997e-05, "loss": 0.7867, "step": 230820 }, { "epoch": 19.99, "learning_rate": 2.5055655140511156e-05, "loss": 0.7806, "step": 230830 }, { "epoch": 19.99, "learning_rate": 2.505348675841332e-05, "loss": 0.9483, "step": 230840 }, { "epoch": 19.99, "learning_rate": 2.5051318376315482e-05, "loss": 0.8489, "step": 230850 }, { "epoch": 19.99, "learning_rate": 2.504914999421765e-05, "loss": 0.831, "step": 230860 }, { "epoch": 19.99, "learning_rate": 2.504698161211981e-05, "loss": 0.6575, "step": 230870 }, { "epoch": 20.0, "learning_rate": 2.504481323002197e-05, "loss": 0.6906, "step": 230880 }, { "epoch": 20.0, "learning_rate": 2.5042644847924134e-05, "loss": 0.5817, "step": 230890 }, { "epoch": 20.0, "learning_rate": 2.5040476465826297e-05, "loss": 0.7453, "step": 230900 }, { "epoch": 20.0, "learning_rate": 2.5038308083728457e-05, "loss": 0.8233, "step": 230910 }, { "epoch": 20.0, "learning_rate": 2.503613970163062e-05, "loss": 0.702, "step": 230920 }, { "epoch": 20.0, "learning_rate": 2.5033971319532786e-05, "loss": 0.8649, "step": 230930 }, { "epoch": 20.0, "eval_cer": 0.537699550283984, "eval_em": 0.05518308406395049, "eval_f1": 0.05518308406395049, "eval_loss": 1.5010526180267334, "eval_runtime": 255.2137, "eval_samples_per_second": 7.598, "eval_steps_per_second": 0.952, "eval_wer": 0.9448169159360496, "step": 230930 }, { "epoch": 20.0, "learning_rate": 2.5031802937434943e-05, "loss": 0.7612, "step": 230940 }, { "epoch": 20.0, "learning_rate": 2.502963455533711e-05, "loss": 0.7441, "step": 230950 }, { "epoch": 20.0, "learning_rate": 2.5027466173239272e-05, "loss": 0.6305, "step": 230960 }, { "epoch": 20.0, "learning_rate": 2.5025297791141435e-05, "loss": 0.7333, "step": 230970 }, { "epoch": 20.0, "learning_rate": 2.5023129409043595e-05, "loss": 0.8024, "step": 230980 }, { "epoch": 20.01, "learning_rate": 2.5020961026945758e-05, "loss": 0.6008, "step": 230990 }, { "epoch": 20.01, "learning_rate": 2.501879264484792e-05, "loss": 0.7648, "step": 231000 }, { "epoch": 20.01, "learning_rate": 2.5016624262750087e-05, "loss": 0.8458, "step": 231010 }, { "epoch": 20.01, "learning_rate": 2.5014455880652247e-05, "loss": 0.7621, "step": 231020 }, { "epoch": 20.01, "learning_rate": 2.501228749855441e-05, "loss": 0.6676, "step": 231030 }, { "epoch": 20.01, "learning_rate": 2.5010119116456573e-05, "loss": 0.5679, "step": 231040 }, { "epoch": 20.01, "learning_rate": 2.5007950734358736e-05, "loss": 0.6692, "step": 231050 }, { "epoch": 20.01, "learning_rate": 2.5005782352260895e-05, "loss": 0.7483, "step": 231060 }, { "epoch": 20.01, "learning_rate": 2.500361397016306e-05, "loss": 0.9751, "step": 231070 }, { "epoch": 20.01, "learning_rate": 2.5001445588065225e-05, "loss": 0.6507, "step": 231080 }, { "epoch": 20.01, "learning_rate": 2.4999277205967388e-05, "loss": 0.7648, "step": 231090 }, { "epoch": 20.01, "learning_rate": 2.4997108823869547e-05, "loss": 0.6937, "step": 231100 }, { "epoch": 20.02, "learning_rate": 2.499494044177171e-05, "loss": 0.734, "step": 231110 }, { "epoch": 20.02, "learning_rate": 2.4992772059673873e-05, "loss": 0.7255, "step": 231120 }, { "epoch": 20.02, "learning_rate": 2.4990603677576036e-05, "loss": 0.7272, "step": 231130 }, { "epoch": 20.02, "learning_rate": 2.4988435295478196e-05, "loss": 0.7755, "step": 231140 }, { "epoch": 20.02, "learning_rate": 2.4986266913380362e-05, "loss": 0.7143, "step": 231150 }, { "epoch": 20.02, "learning_rate": 2.4984098531282525e-05, "loss": 0.7317, "step": 231160 }, { "epoch": 20.02, "learning_rate": 2.498193014918469e-05, "loss": 0.6355, "step": 231170 }, { "epoch": 20.02, "learning_rate": 2.4979761767086848e-05, "loss": 0.6383, "step": 231180 }, { "epoch": 20.02, "learning_rate": 2.497759338498901e-05, "loss": 0.7762, "step": 231190 }, { "epoch": 20.02, "learning_rate": 2.4975425002891174e-05, "loss": 0.8084, "step": 231200 }, { "epoch": 20.02, "learning_rate": 2.4973256620793334e-05, "loss": 0.9437, "step": 231210 }, { "epoch": 20.03, "learning_rate": 2.4971088238695497e-05, "loss": 0.6828, "step": 231220 }, { "epoch": 20.03, "learning_rate": 2.4968919856597663e-05, "loss": 0.6593, "step": 231230 }, { "epoch": 20.03, "learning_rate": 2.4966751474499826e-05, "loss": 0.7339, "step": 231240 }, { "epoch": 20.03, "learning_rate": 2.4964583092401986e-05, "loss": 0.8567, "step": 231250 }, { "epoch": 20.03, "learning_rate": 2.496241471030415e-05, "loss": 0.75, "step": 231260 }, { "epoch": 20.03, "learning_rate": 2.496024632820631e-05, "loss": 0.6301, "step": 231270 }, { "epoch": 20.03, "learning_rate": 2.4958077946108475e-05, "loss": 0.6212, "step": 231280 }, { "epoch": 20.03, "learning_rate": 2.4955909564010634e-05, "loss": 0.5797, "step": 231290 }, { "epoch": 20.03, "learning_rate": 2.49537411819128e-05, "loss": 0.6923, "step": 231300 }, { "epoch": 20.03, "learning_rate": 2.4951572799814964e-05, "loss": 0.605, "step": 231310 }, { "epoch": 20.03, "learning_rate": 2.4949404417717127e-05, "loss": 0.7812, "step": 231320 }, { "epoch": 20.03, "learning_rate": 2.4947236035619286e-05, "loss": 0.805, "step": 231330 }, { "epoch": 20.04, "learning_rate": 2.494506765352145e-05, "loss": 0.765, "step": 231340 }, { "epoch": 20.04, "learning_rate": 2.4942899271423612e-05, "loss": 0.692, "step": 231350 }, { "epoch": 20.04, "learning_rate": 2.494073088932578e-05, "loss": 0.7949, "step": 231360 }, { "epoch": 20.04, "learning_rate": 2.4938562507227938e-05, "loss": 0.6678, "step": 231370 }, { "epoch": 20.04, "learning_rate": 2.49363941251301e-05, "loss": 0.8583, "step": 231380 }, { "epoch": 20.04, "learning_rate": 2.4934225743032264e-05, "loss": 0.7225, "step": 231390 }, { "epoch": 20.04, "learning_rate": 2.4932057360934427e-05, "loss": 0.7288, "step": 231400 }, { "epoch": 20.04, "learning_rate": 2.4929888978836587e-05, "loss": 0.6287, "step": 231410 }, { "epoch": 20.04, "learning_rate": 2.492772059673875e-05, "loss": 0.6581, "step": 231420 }, { "epoch": 20.04, "learning_rate": 2.4925552214640916e-05, "loss": 0.7008, "step": 231430 }, { "epoch": 20.04, "learning_rate": 2.4923383832543076e-05, "loss": 0.6983, "step": 231440 }, { "epoch": 20.05, "learning_rate": 2.492121545044524e-05, "loss": 0.7209, "step": 231450 }, { "epoch": 20.05, "learning_rate": 2.4919047068347402e-05, "loss": 0.6365, "step": 231460 }, { "epoch": 20.05, "learning_rate": 2.4916878686249565e-05, "loss": 0.8207, "step": 231470 }, { "epoch": 20.05, "learning_rate": 2.4914710304151725e-05, "loss": 0.8628, "step": 231480 }, { "epoch": 20.05, "learning_rate": 2.4912541922053888e-05, "loss": 0.6622, "step": 231490 }, { "epoch": 20.05, "learning_rate": 2.491037353995605e-05, "loss": 0.5851, "step": 231500 }, { "epoch": 20.05, "learning_rate": 2.4908205157858217e-05, "loss": 0.6648, "step": 231510 }, { "epoch": 20.05, "learning_rate": 2.4906036775760377e-05, "loss": 0.8762, "step": 231520 }, { "epoch": 20.05, "learning_rate": 2.490386839366254e-05, "loss": 0.7288, "step": 231530 }, { "epoch": 20.05, "learning_rate": 2.4901700011564703e-05, "loss": 0.7156, "step": 231540 }, { "epoch": 20.05, "learning_rate": 2.4899531629466866e-05, "loss": 0.8029, "step": 231550 }, { "epoch": 20.05, "learning_rate": 2.4897363247369025e-05, "loss": 0.7482, "step": 231560 }, { "epoch": 20.06, "learning_rate": 2.4895194865271188e-05, "loss": 0.6084, "step": 231570 }, { "epoch": 20.06, "learning_rate": 2.4893026483173355e-05, "loss": 0.899, "step": 231580 }, { "epoch": 20.06, "learning_rate": 2.4890858101075518e-05, "loss": 0.7299, "step": 231590 }, { "epoch": 20.06, "learning_rate": 2.4888689718977677e-05, "loss": 0.7542, "step": 231600 }, { "epoch": 20.06, "learning_rate": 2.488652133687984e-05, "loss": 0.6874, "step": 231610 }, { "epoch": 20.06, "learning_rate": 2.4884352954782003e-05, "loss": 0.8176, "step": 231620 }, { "epoch": 20.06, "learning_rate": 2.4882184572684166e-05, "loss": 0.632, "step": 231630 }, { "epoch": 20.06, "learning_rate": 2.4880016190586326e-05, "loss": 0.7988, "step": 231640 }, { "epoch": 20.06, "learning_rate": 2.4877847808488492e-05, "loss": 0.6287, "step": 231650 }, { "epoch": 20.06, "learning_rate": 2.4875679426390655e-05, "loss": 0.7103, "step": 231660 }, { "epoch": 20.06, "learning_rate": 2.4873511044292818e-05, "loss": 0.7622, "step": 231670 }, { "epoch": 20.06, "learning_rate": 2.4871342662194978e-05, "loss": 0.6619, "step": 231680 }, { "epoch": 20.07, "learning_rate": 2.486917428009714e-05, "loss": 0.6031, "step": 231690 }, { "epoch": 20.07, "learning_rate": 2.4867005897999304e-05, "loss": 0.8938, "step": 231700 }, { "epoch": 20.07, "learning_rate": 2.4864837515901463e-05, "loss": 0.703, "step": 231710 }, { "epoch": 20.07, "learning_rate": 2.486266913380363e-05, "loss": 0.6913, "step": 231720 }, { "epoch": 20.07, "learning_rate": 2.4860500751705793e-05, "loss": 0.7932, "step": 231730 }, { "epoch": 20.07, "learning_rate": 2.4858332369607956e-05, "loss": 0.6891, "step": 231740 }, { "epoch": 20.07, "learning_rate": 2.4856163987510115e-05, "loss": 0.8323, "step": 231750 }, { "epoch": 20.07, "learning_rate": 2.485399560541228e-05, "loss": 0.7313, "step": 231760 }, { "epoch": 20.07, "learning_rate": 2.485182722331444e-05, "loss": 0.7618, "step": 231770 }, { "epoch": 20.07, "learning_rate": 2.4849658841216604e-05, "loss": 0.7087, "step": 231780 }, { "epoch": 20.07, "learning_rate": 2.4847490459118764e-05, "loss": 0.723, "step": 231790 }, { "epoch": 20.08, "learning_rate": 2.484532207702093e-05, "loss": 0.739, "step": 231800 }, { "epoch": 20.08, "learning_rate": 2.4843153694923093e-05, "loss": 0.6982, "step": 231810 }, { "epoch": 20.08, "learning_rate": 2.4840985312825257e-05, "loss": 0.7081, "step": 231820 }, { "epoch": 20.08, "learning_rate": 2.4838816930727416e-05, "loss": 0.7303, "step": 231830 }, { "epoch": 20.08, "learning_rate": 2.483664854862958e-05, "loss": 0.8005, "step": 231840 }, { "epoch": 20.08, "learning_rate": 2.4834480166531742e-05, "loss": 0.6836, "step": 231850 }, { "epoch": 20.08, "learning_rate": 2.483231178443391e-05, "loss": 0.6887, "step": 231860 }, { "epoch": 20.08, "learning_rate": 2.4830143402336068e-05, "loss": 0.7044, "step": 231870 }, { "epoch": 20.08, "learning_rate": 2.482797502023823e-05, "loss": 0.5955, "step": 231880 }, { "epoch": 20.08, "learning_rate": 2.4825806638140394e-05, "loss": 0.6768, "step": 231890 }, { "epoch": 20.08, "learning_rate": 2.4823638256042557e-05, "loss": 0.589, "step": 231900 }, { "epoch": 20.08, "learning_rate": 2.4821469873944717e-05, "loss": 0.6036, "step": 231910 }, { "epoch": 20.09, "learning_rate": 2.481930149184688e-05, "loss": 0.8494, "step": 231920 }, { "epoch": 20.09, "learning_rate": 2.4817133109749046e-05, "loss": 0.8207, "step": 231930 }, { "epoch": 20.09, "learning_rate": 2.481496472765121e-05, "loss": 0.6145, "step": 231940 }, { "epoch": 20.09, "learning_rate": 2.481279634555337e-05, "loss": 0.6878, "step": 231950 }, { "epoch": 20.09, "learning_rate": 2.4810627963455532e-05, "loss": 0.7483, "step": 231960 }, { "epoch": 20.09, "learning_rate": 2.4808459581357695e-05, "loss": 0.74, "step": 231970 }, { "epoch": 20.09, "learning_rate": 2.4806291199259854e-05, "loss": 0.6794, "step": 231980 }, { "epoch": 20.09, "learning_rate": 2.4804122817162017e-05, "loss": 0.8636, "step": 231990 }, { "epoch": 20.09, "learning_rate": 2.4801954435064184e-05, "loss": 0.7257, "step": 232000 }, { "epoch": 20.09, "learning_rate": 2.4799786052966347e-05, "loss": 0.5998, "step": 232010 }, { "epoch": 20.09, "learning_rate": 2.4797617670868506e-05, "loss": 0.8173, "step": 232020 }, { "epoch": 20.1, "learning_rate": 2.479544928877067e-05, "loss": 0.5392, "step": 232030 }, { "epoch": 20.1, "learning_rate": 2.4793280906672832e-05, "loss": 0.6342, "step": 232040 }, { "epoch": 20.1, "learning_rate": 2.4791112524574995e-05, "loss": 0.7065, "step": 232050 }, { "epoch": 20.1, "learning_rate": 2.4788944142477155e-05, "loss": 0.789, "step": 232060 }, { "epoch": 20.1, "learning_rate": 2.4786775760379318e-05, "loss": 0.7683, "step": 232070 }, { "epoch": 20.1, "learning_rate": 2.4784607378281484e-05, "loss": 0.7104, "step": 232080 }, { "epoch": 20.1, "learning_rate": 2.4782438996183647e-05, "loss": 0.6867, "step": 232090 }, { "epoch": 20.1, "learning_rate": 2.4780270614085807e-05, "loss": 0.662, "step": 232100 }, { "epoch": 20.1, "learning_rate": 2.477810223198797e-05, "loss": 0.801, "step": 232110 }, { "epoch": 20.1, "learning_rate": 2.4775933849890133e-05, "loss": 0.6678, "step": 232120 }, { "epoch": 20.1, "learning_rate": 2.4773765467792296e-05, "loss": 0.7465, "step": 232130 }, { "epoch": 20.1, "learning_rate": 2.4771597085694456e-05, "loss": 0.6404, "step": 232140 }, { "epoch": 20.11, "learning_rate": 2.4769428703596622e-05, "loss": 0.6565, "step": 232150 }, { "epoch": 20.11, "learning_rate": 2.4767260321498785e-05, "loss": 0.7374, "step": 232160 }, { "epoch": 20.11, "learning_rate": 2.4765091939400948e-05, "loss": 0.6883, "step": 232170 }, { "epoch": 20.11, "learning_rate": 2.4762923557303108e-05, "loss": 0.5611, "step": 232180 }, { "epoch": 20.11, "learning_rate": 2.476075517520527e-05, "loss": 0.7739, "step": 232190 }, { "epoch": 20.11, "learning_rate": 2.4758586793107434e-05, "loss": 0.7078, "step": 232200 }, { "epoch": 20.11, "learning_rate": 2.4756418411009593e-05, "loss": 0.603, "step": 232210 }, { "epoch": 20.11, "learning_rate": 2.475425002891176e-05, "loss": 0.7984, "step": 232220 }, { "epoch": 20.11, "learning_rate": 2.4752081646813923e-05, "loss": 0.716, "step": 232230 }, { "epoch": 20.11, "learning_rate": 2.4749913264716086e-05, "loss": 0.6911, "step": 232240 }, { "epoch": 20.11, "learning_rate": 2.4747744882618245e-05, "loss": 0.6829, "step": 232250 }, { "epoch": 20.12, "learning_rate": 2.474557650052041e-05, "loss": 0.6822, "step": 232260 }, { "epoch": 20.12, "learning_rate": 2.474340811842257e-05, "loss": 0.6776, "step": 232270 }, { "epoch": 20.12, "learning_rate": 2.4741239736324738e-05, "loss": 0.6726, "step": 232280 }, { "epoch": 20.12, "learning_rate": 2.4739071354226894e-05, "loss": 0.6259, "step": 232290 }, { "epoch": 20.12, "learning_rate": 2.473690297212906e-05, "loss": 0.7446, "step": 232300 }, { "epoch": 20.12, "learning_rate": 2.4734734590031223e-05, "loss": 0.8983, "step": 232310 }, { "epoch": 20.12, "learning_rate": 2.4732566207933386e-05, "loss": 0.8284, "step": 232320 }, { "epoch": 20.12, "learning_rate": 2.4730397825835546e-05, "loss": 0.5574, "step": 232330 }, { "epoch": 20.12, "learning_rate": 2.472822944373771e-05, "loss": 0.8642, "step": 232340 }, { "epoch": 20.12, "learning_rate": 2.4726061061639872e-05, "loss": 0.7364, "step": 232350 }, { "epoch": 20.12, "learning_rate": 2.472389267954204e-05, "loss": 0.783, "step": 232360 }, { "epoch": 20.12, "learning_rate": 2.4721724297444198e-05, "loss": 0.6405, "step": 232370 }, { "epoch": 20.13, "learning_rate": 2.471955591534636e-05, "loss": 0.6534, "step": 232380 }, { "epoch": 20.13, "learning_rate": 2.4717387533248524e-05, "loss": 0.7281, "step": 232390 }, { "epoch": 20.13, "learning_rate": 2.4715219151150687e-05, "loss": 0.8775, "step": 232400 }, { "epoch": 20.13, "learning_rate": 2.4713050769052847e-05, "loss": 0.8004, "step": 232410 }, { "epoch": 20.13, "learning_rate": 2.471088238695501e-05, "loss": 0.7543, "step": 232420 }, { "epoch": 20.13, "learning_rate": 2.4708714004857176e-05, "loss": 0.856, "step": 232430 }, { "epoch": 20.13, "learning_rate": 2.470654562275934e-05, "loss": 0.6944, "step": 232440 }, { "epoch": 20.13, "learning_rate": 2.47043772406615e-05, "loss": 0.7455, "step": 232450 }, { "epoch": 20.13, "learning_rate": 2.470220885856366e-05, "loss": 0.728, "step": 232460 }, { "epoch": 20.13, "learning_rate": 2.4700040476465825e-05, "loss": 0.7097, "step": 232470 }, { "epoch": 20.13, "learning_rate": 2.4697872094367984e-05, "loss": 0.7924, "step": 232480 }, { "epoch": 20.14, "learning_rate": 2.4695703712270147e-05, "loss": 0.7156, "step": 232490 }, { "epoch": 20.14, "learning_rate": 2.4693535330172314e-05, "loss": 0.6854, "step": 232500 }, { "epoch": 20.14, "learning_rate": 2.4691366948074477e-05, "loss": 0.9344, "step": 232510 }, { "epoch": 20.14, "learning_rate": 2.4689198565976636e-05, "loss": 0.6815, "step": 232520 }, { "epoch": 20.14, "learning_rate": 2.46870301838788e-05, "loss": 0.8233, "step": 232530 }, { "epoch": 20.14, "learning_rate": 2.4684861801780962e-05, "loss": 0.7354, "step": 232540 }, { "epoch": 20.14, "learning_rate": 2.4682693419683125e-05, "loss": 0.7647, "step": 232550 }, { "epoch": 20.14, "learning_rate": 2.4680525037585285e-05, "loss": 0.7288, "step": 232560 }, { "epoch": 20.14, "learning_rate": 2.4678356655487448e-05, "loss": 0.639, "step": 232570 }, { "epoch": 20.14, "learning_rate": 2.4676188273389614e-05, "loss": 0.6812, "step": 232580 }, { "epoch": 20.14, "learning_rate": 2.4674019891291777e-05, "loss": 0.7703, "step": 232590 }, { "epoch": 20.14, "learning_rate": 2.4671851509193937e-05, "loss": 0.5728, "step": 232600 }, { "epoch": 20.15, "learning_rate": 2.46696831270961e-05, "loss": 0.8654, "step": 232610 }, { "epoch": 20.15, "learning_rate": 2.4667514744998263e-05, "loss": 0.743, "step": 232620 }, { "epoch": 20.15, "learning_rate": 2.4665346362900426e-05, "loss": 0.7226, "step": 232630 }, { "epoch": 20.15, "learning_rate": 2.4663177980802586e-05, "loss": 0.6616, "step": 232640 }, { "epoch": 20.15, "learning_rate": 2.4661009598704752e-05, "loss": 0.821, "step": 232650 }, { "epoch": 20.15, "learning_rate": 2.4658841216606915e-05, "loss": 0.5949, "step": 232660 }, { "epoch": 20.15, "learning_rate": 2.4656672834509078e-05, "loss": 0.6988, "step": 232670 }, { "epoch": 20.15, "learning_rate": 2.4654504452411238e-05, "loss": 0.8633, "step": 232680 }, { "epoch": 20.15, "learning_rate": 2.46523360703134e-05, "loss": 0.5386, "step": 232690 }, { "epoch": 20.15, "learning_rate": 2.4650167688215564e-05, "loss": 0.6779, "step": 232700 }, { "epoch": 20.15, "learning_rate": 2.464799930611773e-05, "loss": 0.7335, "step": 232710 }, { "epoch": 20.16, "learning_rate": 2.464583092401989e-05, "loss": 0.7708, "step": 232720 }, { "epoch": 20.16, "learning_rate": 2.4643662541922053e-05, "loss": 0.6154, "step": 232730 }, { "epoch": 20.16, "learning_rate": 2.4641494159824216e-05, "loss": 0.7379, "step": 232740 }, { "epoch": 20.16, "learning_rate": 2.4639325777726375e-05, "loss": 0.6356, "step": 232750 }, { "epoch": 20.16, "learning_rate": 2.4637157395628538e-05, "loss": 0.8041, "step": 232760 }, { "epoch": 20.16, "learning_rate": 2.46349890135307e-05, "loss": 0.6818, "step": 232770 }, { "epoch": 20.16, "learning_rate": 2.4632820631432868e-05, "loss": 0.7096, "step": 232780 }, { "epoch": 20.16, "learning_rate": 2.4630652249335027e-05, "loss": 0.6447, "step": 232790 }, { "epoch": 20.16, "learning_rate": 2.462848386723719e-05, "loss": 0.6152, "step": 232800 }, { "epoch": 20.16, "learning_rate": 2.4626315485139353e-05, "loss": 0.6716, "step": 232810 }, { "epoch": 20.16, "learning_rate": 2.4624147103041516e-05, "loss": 0.6245, "step": 232820 }, { "epoch": 20.16, "learning_rate": 2.4621978720943676e-05, "loss": 0.7166, "step": 232830 }, { "epoch": 20.17, "learning_rate": 2.461981033884584e-05, "loss": 0.7562, "step": 232840 }, { "epoch": 20.17, "learning_rate": 2.4617641956748002e-05, "loss": 0.7348, "step": 232850 }, { "epoch": 20.17, "learning_rate": 2.4615473574650168e-05, "loss": 0.6998, "step": 232860 }, { "epoch": 20.17, "learning_rate": 2.4613305192552328e-05, "loss": 0.6418, "step": 232870 }, { "epoch": 20.17, "learning_rate": 2.461113681045449e-05, "loss": 0.7107, "step": 232880 }, { "epoch": 20.17, "learning_rate": 2.4608968428356654e-05, "loss": 0.7446, "step": 232890 }, { "epoch": 20.17, "learning_rate": 2.4606800046258817e-05, "loss": 0.7053, "step": 232900 }, { "epoch": 20.17, "learning_rate": 2.4604631664160976e-05, "loss": 0.7424, "step": 232910 }, { "epoch": 20.17, "learning_rate": 2.460246328206314e-05, "loss": 0.7042, "step": 232920 }, { "epoch": 20.17, "learning_rate": 2.4600294899965306e-05, "loss": 0.7525, "step": 232930 }, { "epoch": 20.17, "learning_rate": 2.459812651786747e-05, "loss": 0.8167, "step": 232940 }, { "epoch": 20.17, "learning_rate": 2.459595813576963e-05, "loss": 0.6328, "step": 232950 }, { "epoch": 20.18, "learning_rate": 2.459378975367179e-05, "loss": 0.6512, "step": 232960 }, { "epoch": 20.18, "learning_rate": 2.4591621371573954e-05, "loss": 0.6145, "step": 232970 }, { "epoch": 20.18, "learning_rate": 2.4589452989476114e-05, "loss": 0.7964, "step": 232980 }, { "epoch": 20.18, "learning_rate": 2.4587284607378277e-05, "loss": 0.7453, "step": 232990 }, { "epoch": 20.18, "learning_rate": 2.4585116225280443e-05, "loss": 0.6813, "step": 233000 }, { "epoch": 20.18, "learning_rate": 2.4582947843182606e-05, "loss": 0.7265, "step": 233010 }, { "epoch": 20.18, "learning_rate": 2.4580779461084766e-05, "loss": 0.7722, "step": 233020 }, { "epoch": 20.18, "learning_rate": 2.457861107898693e-05, "loss": 0.7418, "step": 233030 }, { "epoch": 20.18, "learning_rate": 2.4576442696889092e-05, "loss": 0.7491, "step": 233040 }, { "epoch": 20.18, "learning_rate": 2.4574274314791255e-05, "loss": 0.77, "step": 233050 }, { "epoch": 20.18, "learning_rate": 2.4572105932693415e-05, "loss": 0.6477, "step": 233060 }, { "epoch": 20.19, "learning_rate": 2.456993755059558e-05, "loss": 0.7181, "step": 233070 }, { "epoch": 20.19, "learning_rate": 2.4567769168497744e-05, "loss": 0.6135, "step": 233080 }, { "epoch": 20.19, "learning_rate": 2.4565600786399907e-05, "loss": 0.7233, "step": 233090 }, { "epoch": 20.19, "learning_rate": 2.4563432404302067e-05, "loss": 0.7424, "step": 233100 }, { "epoch": 20.19, "learning_rate": 2.456126402220423e-05, "loss": 0.8201, "step": 233110 }, { "epoch": 20.19, "learning_rate": 2.4559095640106393e-05, "loss": 0.7437, "step": 233120 }, { "epoch": 20.19, "learning_rate": 2.4556927258008556e-05, "loss": 0.7901, "step": 233130 }, { "epoch": 20.19, "learning_rate": 2.4554758875910715e-05, "loss": 0.5375, "step": 233140 }, { "epoch": 20.19, "learning_rate": 2.4552590493812882e-05, "loss": 0.7274, "step": 233150 }, { "epoch": 20.19, "learning_rate": 2.4550422111715045e-05, "loss": 0.6537, "step": 233160 }, { "epoch": 20.19, "learning_rate": 2.4548253729617208e-05, "loss": 0.6146, "step": 233170 }, { "epoch": 20.19, "learning_rate": 2.4546085347519367e-05, "loss": 0.8229, "step": 233180 }, { "epoch": 20.2, "learning_rate": 2.454391696542153e-05, "loss": 0.7875, "step": 233190 }, { "epoch": 20.2, "learning_rate": 2.4541748583323693e-05, "loss": 0.7387, "step": 233200 }, { "epoch": 20.2, "learning_rate": 2.453958020122586e-05, "loss": 0.9966, "step": 233210 }, { "epoch": 20.2, "learning_rate": 2.453741181912802e-05, "loss": 0.7504, "step": 233220 }, { "epoch": 20.2, "learning_rate": 2.4535243437030182e-05, "loss": 0.6605, "step": 233230 }, { "epoch": 20.2, "learning_rate": 2.4533075054932345e-05, "loss": 0.8025, "step": 233240 }, { "epoch": 20.2, "learning_rate": 2.4530906672834505e-05, "loss": 0.6128, "step": 233250 }, { "epoch": 20.2, "learning_rate": 2.4528738290736668e-05, "loss": 0.7022, "step": 233260 }, { "epoch": 20.2, "learning_rate": 2.452656990863883e-05, "loss": 0.6395, "step": 233270 }, { "epoch": 20.2, "learning_rate": 2.4524401526540997e-05, "loss": 0.7388, "step": 233280 }, { "epoch": 20.2, "learning_rate": 2.4522233144443157e-05, "loss": 0.6363, "step": 233290 }, { "epoch": 20.21, "learning_rate": 2.452006476234532e-05, "loss": 0.7083, "step": 233300 }, { "epoch": 20.21, "learning_rate": 2.4517896380247483e-05, "loss": 0.5825, "step": 233310 }, { "epoch": 20.21, "learning_rate": 2.4515727998149646e-05, "loss": 0.6981, "step": 233320 }, { "epoch": 20.21, "learning_rate": 2.4513559616051806e-05, "loss": 0.7034, "step": 233330 }, { "epoch": 20.21, "learning_rate": 2.451139123395397e-05, "loss": 0.8539, "step": 233340 }, { "epoch": 20.21, "learning_rate": 2.4509222851856135e-05, "loss": 0.6794, "step": 233350 }, { "epoch": 20.21, "learning_rate": 2.4507054469758298e-05, "loss": 0.8503, "step": 233360 }, { "epoch": 20.21, "learning_rate": 2.4504886087660458e-05, "loss": 0.653, "step": 233370 }, { "epoch": 20.21, "learning_rate": 2.450271770556262e-05, "loss": 0.6859, "step": 233380 }, { "epoch": 20.21, "learning_rate": 2.4500549323464784e-05, "loss": 0.7183, "step": 233390 }, { "epoch": 20.21, "learning_rate": 2.4498380941366947e-05, "loss": 0.8585, "step": 233400 }, { "epoch": 20.21, "learning_rate": 2.4496212559269106e-05, "loss": 0.8959, "step": 233410 }, { "epoch": 20.22, "learning_rate": 2.449404417717127e-05, "loss": 0.7792, "step": 233420 }, { "epoch": 20.22, "learning_rate": 2.4491875795073436e-05, "loss": 0.761, "step": 233430 }, { "epoch": 20.22, "learning_rate": 2.44897074129756e-05, "loss": 0.7518, "step": 233440 }, { "epoch": 20.22, "learning_rate": 2.448753903087776e-05, "loss": 0.7469, "step": 233450 }, { "epoch": 20.22, "learning_rate": 2.448537064877992e-05, "loss": 0.6854, "step": 233460 }, { "epoch": 20.22, "learning_rate": 2.4483202266682084e-05, "loss": 0.6573, "step": 233470 }, { "epoch": 20.22, "learning_rate": 2.4481033884584247e-05, "loss": 0.6549, "step": 233480 }, { "epoch": 20.22, "learning_rate": 2.4478865502486407e-05, "loss": 0.5552, "step": 233490 }, { "epoch": 20.22, "learning_rate": 2.4476697120388573e-05, "loss": 0.6479, "step": 233500 }, { "epoch": 20.22, "learning_rate": 2.4474528738290736e-05, "loss": 0.8265, "step": 233510 }, { "epoch": 20.22, "learning_rate": 2.4472360356192896e-05, "loss": 0.7294, "step": 233520 }, { "epoch": 20.23, "learning_rate": 2.447019197409506e-05, "loss": 0.6918, "step": 233530 }, { "epoch": 20.23, "learning_rate": 2.4468023591997222e-05, "loss": 0.6968, "step": 233540 }, { "epoch": 20.23, "learning_rate": 2.4465855209899385e-05, "loss": 0.6592, "step": 233550 }, { "epoch": 20.23, "learning_rate": 2.4463686827801545e-05, "loss": 0.7412, "step": 233560 }, { "epoch": 20.23, "learning_rate": 2.446151844570371e-05, "loss": 0.6812, "step": 233570 }, { "epoch": 20.23, "learning_rate": 2.4459350063605874e-05, "loss": 0.9086, "step": 233580 }, { "epoch": 20.23, "learning_rate": 2.4457181681508037e-05, "loss": 0.8212, "step": 233590 }, { "epoch": 20.23, "learning_rate": 2.4455013299410197e-05, "loss": 0.6694, "step": 233600 }, { "epoch": 20.23, "learning_rate": 2.445284491731236e-05, "loss": 0.6768, "step": 233610 }, { "epoch": 20.23, "learning_rate": 2.4450676535214523e-05, "loss": 0.7102, "step": 233620 }, { "epoch": 20.23, "learning_rate": 2.444850815311669e-05, "loss": 0.7708, "step": 233630 }, { "epoch": 20.23, "learning_rate": 2.4446339771018845e-05, "loss": 0.7728, "step": 233640 }, { "epoch": 20.24, "learning_rate": 2.444417138892101e-05, "loss": 0.8136, "step": 233650 }, { "epoch": 20.24, "learning_rate": 2.4442003006823175e-05, "loss": 0.7849, "step": 233660 }, { "epoch": 20.24, "learning_rate": 2.4439834624725338e-05, "loss": 0.6325, "step": 233670 }, { "epoch": 20.24, "learning_rate": 2.4437666242627497e-05, "loss": 0.8404, "step": 233680 }, { "epoch": 20.24, "learning_rate": 2.443549786052966e-05, "loss": 0.7295, "step": 233690 }, { "epoch": 20.24, "learning_rate": 2.4433329478431823e-05, "loss": 0.7707, "step": 233700 }, { "epoch": 20.24, "learning_rate": 2.443116109633399e-05, "loss": 0.7225, "step": 233710 }, { "epoch": 20.24, "learning_rate": 2.442899271423615e-05, "loss": 0.7873, "step": 233720 }, { "epoch": 20.24, "learning_rate": 2.4426824332138312e-05, "loss": 0.654, "step": 233730 }, { "epoch": 20.24, "learning_rate": 2.4424655950040475e-05, "loss": 0.7902, "step": 233740 }, { "epoch": 20.24, "learning_rate": 2.4422487567942635e-05, "loss": 0.6176, "step": 233750 }, { "epoch": 20.25, "learning_rate": 2.4420319185844798e-05, "loss": 0.6434, "step": 233760 }, { "epoch": 20.25, "learning_rate": 2.441815080374696e-05, "loss": 0.7701, "step": 233770 }, { "epoch": 20.25, "learning_rate": 2.4415982421649127e-05, "loss": 0.6946, "step": 233780 }, { "epoch": 20.25, "learning_rate": 2.4413814039551287e-05, "loss": 0.7792, "step": 233790 }, { "epoch": 20.25, "learning_rate": 2.441164565745345e-05, "loss": 0.7172, "step": 233800 }, { "epoch": 20.25, "learning_rate": 2.4409477275355613e-05, "loss": 0.8052, "step": 233810 }, { "epoch": 20.25, "learning_rate": 2.4407308893257776e-05, "loss": 0.7974, "step": 233820 }, { "epoch": 20.25, "learning_rate": 2.4405140511159936e-05, "loss": 0.8528, "step": 233830 }, { "epoch": 20.25, "learning_rate": 2.44029721290621e-05, "loss": 0.7156, "step": 233840 }, { "epoch": 20.25, "learning_rate": 2.4400803746964265e-05, "loss": 0.8198, "step": 233850 }, { "epoch": 20.25, "learning_rate": 2.4398635364866428e-05, "loss": 0.6928, "step": 233860 }, { "epoch": 20.25, "learning_rate": 2.4396466982768588e-05, "loss": 0.7383, "step": 233870 }, { "epoch": 20.26, "learning_rate": 2.439429860067075e-05, "loss": 0.7376, "step": 233880 }, { "epoch": 20.26, "learning_rate": 2.4392130218572914e-05, "loss": 0.8308, "step": 233890 }, { "epoch": 20.26, "learning_rate": 2.4389961836475077e-05, "loss": 0.8955, "step": 233900 }, { "epoch": 20.26, "learning_rate": 2.4387793454377236e-05, "loss": 0.7773, "step": 233910 }, { "epoch": 20.26, "learning_rate": 2.43856250722794e-05, "loss": 0.7552, "step": 233920 }, { "epoch": 20.26, "learning_rate": 2.4383456690181566e-05, "loss": 0.6024, "step": 233930 }, { "epoch": 20.26, "learning_rate": 2.438128830808373e-05, "loss": 0.6631, "step": 233940 }, { "epoch": 20.26, "learning_rate": 2.4379119925985888e-05, "loss": 0.6367, "step": 233950 }, { "epoch": 20.26, "learning_rate": 2.437695154388805e-05, "loss": 0.583, "step": 233960 }, { "epoch": 20.26, "learning_rate": 2.4374783161790214e-05, "loss": 0.5389, "step": 233970 }, { "epoch": 20.26, "learning_rate": 2.4372614779692377e-05, "loss": 0.781, "step": 233980 }, { "epoch": 20.27, "learning_rate": 2.4370446397594537e-05, "loss": 0.8211, "step": 233990 }, { "epoch": 20.27, "learning_rate": 2.4368278015496703e-05, "loss": 0.6522, "step": 234000 }, { "epoch": 20.27, "learning_rate": 2.4366109633398866e-05, "loss": 0.6827, "step": 234010 }, { "epoch": 20.27, "learning_rate": 2.4363941251301026e-05, "loss": 0.6675, "step": 234020 }, { "epoch": 20.27, "learning_rate": 2.436177286920319e-05, "loss": 0.7409, "step": 234030 }, { "epoch": 20.27, "learning_rate": 2.4359604487105352e-05, "loss": 0.6793, "step": 234040 }, { "epoch": 20.27, "learning_rate": 2.4357436105007515e-05, "loss": 0.7391, "step": 234050 }, { "epoch": 20.27, "learning_rate": 2.4355267722909674e-05, "loss": 0.6543, "step": 234060 }, { "epoch": 20.27, "learning_rate": 2.435309934081184e-05, "loss": 0.6132, "step": 234070 }, { "epoch": 20.27, "learning_rate": 2.4350930958714004e-05, "loss": 0.7193, "step": 234080 }, { "epoch": 20.27, "learning_rate": 2.4348762576616167e-05, "loss": 0.754, "step": 234090 }, { "epoch": 20.27, "learning_rate": 2.4346594194518326e-05, "loss": 0.8201, "step": 234100 }, { "epoch": 20.28, "learning_rate": 2.434442581242049e-05, "loss": 0.7444, "step": 234110 }, { "epoch": 20.28, "learning_rate": 2.4342257430322652e-05, "loss": 0.6603, "step": 234120 }, { "epoch": 20.28, "learning_rate": 2.434008904822482e-05, "loss": 0.7395, "step": 234130 }, { "epoch": 20.28, "learning_rate": 2.433792066612698e-05, "loss": 0.7268, "step": 234140 }, { "epoch": 20.28, "learning_rate": 2.433575228402914e-05, "loss": 0.6869, "step": 234150 }, { "epoch": 20.28, "learning_rate": 2.4333583901931304e-05, "loss": 0.8862, "step": 234160 }, { "epoch": 20.28, "learning_rate": 2.4331415519833467e-05, "loss": 0.7061, "step": 234170 }, { "epoch": 20.28, "learning_rate": 2.4329247137735627e-05, "loss": 0.7379, "step": 234180 }, { "epoch": 20.28, "learning_rate": 2.432707875563779e-05, "loss": 0.5804, "step": 234190 }, { "epoch": 20.28, "learning_rate": 2.4324910373539953e-05, "loss": 0.8639, "step": 234200 }, { "epoch": 20.28, "learning_rate": 2.432274199144212e-05, "loss": 0.7263, "step": 234210 }, { "epoch": 20.28, "learning_rate": 2.432057360934428e-05, "loss": 0.7875, "step": 234220 }, { "epoch": 20.29, "learning_rate": 2.4318405227246442e-05, "loss": 0.8457, "step": 234230 }, { "epoch": 20.29, "learning_rate": 2.4316236845148605e-05, "loss": 0.6879, "step": 234240 }, { "epoch": 20.29, "learning_rate": 2.4314068463050768e-05, "loss": 0.6376, "step": 234250 }, { "epoch": 20.29, "learning_rate": 2.4311900080952928e-05, "loss": 0.6376, "step": 234260 }, { "epoch": 20.29, "learning_rate": 2.430973169885509e-05, "loss": 0.7801, "step": 234270 }, { "epoch": 20.29, "learning_rate": 2.4307563316757257e-05, "loss": 0.6751, "step": 234280 }, { "epoch": 20.29, "learning_rate": 2.4305394934659417e-05, "loss": 0.7445, "step": 234290 }, { "epoch": 20.29, "learning_rate": 2.430322655256158e-05, "loss": 0.661, "step": 234300 }, { "epoch": 20.29, "learning_rate": 2.4301058170463743e-05, "loss": 0.8942, "step": 234310 }, { "epoch": 20.29, "learning_rate": 2.4298889788365906e-05, "loss": 0.7331, "step": 234320 }, { "epoch": 20.29, "learning_rate": 2.4296721406268065e-05, "loss": 0.752, "step": 234330 }, { "epoch": 20.3, "learning_rate": 2.429455302417023e-05, "loss": 0.6167, "step": 234340 }, { "epoch": 20.3, "learning_rate": 2.4292384642072395e-05, "loss": 0.6891, "step": 234350 }, { "epoch": 20.3, "learning_rate": 2.4290216259974558e-05, "loss": 0.6636, "step": 234360 }, { "epoch": 20.3, "learning_rate": 2.4288047877876717e-05, "loss": 0.7481, "step": 234370 }, { "epoch": 20.3, "learning_rate": 2.428587949577888e-05, "loss": 0.5871, "step": 234380 }, { "epoch": 20.3, "learning_rate": 2.4283711113681043e-05, "loss": 0.6822, "step": 234390 }, { "epoch": 20.3, "learning_rate": 2.4281542731583206e-05, "loss": 0.7152, "step": 234400 }, { "epoch": 20.3, "learning_rate": 2.4279374349485366e-05, "loss": 0.7863, "step": 234410 }, { "epoch": 20.3, "learning_rate": 2.4277205967387532e-05, "loss": 0.8474, "step": 234420 }, { "epoch": 20.3, "learning_rate": 2.4275037585289695e-05, "loss": 0.7536, "step": 234430 }, { "epoch": 20.3, "learning_rate": 2.427286920319186e-05, "loss": 0.9178, "step": 234440 }, { "epoch": 20.3, "learning_rate": 2.4270700821094018e-05, "loss": 0.5799, "step": 234450 }, { "epoch": 20.31, "learning_rate": 2.426853243899618e-05, "loss": 0.6753, "step": 234460 }, { "epoch": 20.31, "learning_rate": 2.4266364056898344e-05, "loss": 0.6572, "step": 234470 }, { "epoch": 20.31, "learning_rate": 2.4264195674800507e-05, "loss": 0.7496, "step": 234480 }, { "epoch": 20.31, "learning_rate": 2.4262027292702667e-05, "loss": 0.6781, "step": 234490 }, { "epoch": 20.31, "learning_rate": 2.4259858910604833e-05, "loss": 0.6412, "step": 234500 }, { "epoch": 20.31, "learning_rate": 2.4257690528506996e-05, "loss": 0.6739, "step": 234510 }, { "epoch": 20.31, "learning_rate": 2.4255522146409156e-05, "loss": 0.8295, "step": 234520 }, { "epoch": 20.31, "learning_rate": 2.425335376431132e-05, "loss": 0.7309, "step": 234530 }, { "epoch": 20.31, "learning_rate": 2.425118538221348e-05, "loss": 0.7697, "step": 234540 }, { "epoch": 20.31, "learning_rate": 2.4249017000115645e-05, "loss": 0.7582, "step": 234550 }, { "epoch": 20.31, "learning_rate": 2.4246848618017804e-05, "loss": 0.719, "step": 234560 }, { "epoch": 20.32, "learning_rate": 2.424468023591997e-05, "loss": 0.7246, "step": 234570 }, { "epoch": 20.32, "learning_rate": 2.4242511853822134e-05, "loss": 0.6436, "step": 234580 }, { "epoch": 20.32, "learning_rate": 2.4240343471724297e-05, "loss": 0.5969, "step": 234590 }, { "epoch": 20.32, "learning_rate": 2.4238175089626456e-05, "loss": 0.7405, "step": 234600 }, { "epoch": 20.32, "learning_rate": 2.423600670752862e-05, "loss": 0.7682, "step": 234610 }, { "epoch": 20.32, "learning_rate": 2.4233838325430782e-05, "loss": 0.7258, "step": 234620 }, { "epoch": 20.32, "learning_rate": 2.423166994333295e-05, "loss": 0.754, "step": 234630 }, { "epoch": 20.32, "learning_rate": 2.4229501561235108e-05, "loss": 0.775, "step": 234640 }, { "epoch": 20.32, "learning_rate": 2.422733317913727e-05, "loss": 0.796, "step": 234650 }, { "epoch": 20.32, "learning_rate": 2.4225164797039434e-05, "loss": 0.6666, "step": 234660 }, { "epoch": 20.32, "learning_rate": 2.4222996414941597e-05, "loss": 0.6365, "step": 234670 }, { "epoch": 20.32, "learning_rate": 2.4220828032843757e-05, "loss": 0.6737, "step": 234680 }, { "epoch": 20.33, "learning_rate": 2.421865965074592e-05, "loss": 0.7561, "step": 234690 }, { "epoch": 20.33, "learning_rate": 2.4216491268648086e-05, "loss": 0.6586, "step": 234700 }, { "epoch": 20.33, "learning_rate": 2.421432288655025e-05, "loss": 0.7762, "step": 234710 }, { "epoch": 20.33, "learning_rate": 2.421215450445241e-05, "loss": 0.7958, "step": 234720 }, { "epoch": 20.33, "learning_rate": 2.4209986122354572e-05, "loss": 0.7756, "step": 234730 }, { "epoch": 20.33, "learning_rate": 2.4207817740256735e-05, "loss": 0.8966, "step": 234740 }, { "epoch": 20.33, "learning_rate": 2.4205649358158898e-05, "loss": 0.7561, "step": 234750 }, { "epoch": 20.33, "learning_rate": 2.4203480976061058e-05, "loss": 0.7045, "step": 234760 }, { "epoch": 20.33, "learning_rate": 2.420131259396322e-05, "loss": 0.7128, "step": 234770 }, { "epoch": 20.33, "learning_rate": 2.4199144211865387e-05, "loss": 0.774, "step": 234780 }, { "epoch": 20.33, "learning_rate": 2.4196975829767547e-05, "loss": 0.584, "step": 234790 }, { "epoch": 20.34, "learning_rate": 2.419480744766971e-05, "loss": 0.7407, "step": 234800 }, { "epoch": 20.34, "learning_rate": 2.4192639065571873e-05, "loss": 0.7836, "step": 234810 }, { "epoch": 20.34, "learning_rate": 2.4190470683474036e-05, "loss": 0.7289, "step": 234820 }, { "epoch": 20.34, "learning_rate": 2.4188302301376195e-05, "loss": 0.7941, "step": 234830 }, { "epoch": 20.34, "learning_rate": 2.4186133919278358e-05, "loss": 0.7212, "step": 234840 }, { "epoch": 20.34, "learning_rate": 2.4183965537180525e-05, "loss": 0.7395, "step": 234850 }, { "epoch": 20.34, "learning_rate": 2.4181797155082688e-05, "loss": 0.8071, "step": 234860 }, { "epoch": 20.34, "learning_rate": 2.4179628772984847e-05, "loss": 0.6936, "step": 234870 }, { "epoch": 20.34, "learning_rate": 2.417746039088701e-05, "loss": 0.7543, "step": 234880 }, { "epoch": 20.34, "learning_rate": 2.4175292008789173e-05, "loss": 0.663, "step": 234890 }, { "epoch": 20.34, "learning_rate": 2.4173123626691336e-05, "loss": 0.6418, "step": 234900 }, { "epoch": 20.34, "learning_rate": 2.4170955244593496e-05, "loss": 0.6642, "step": 234910 }, { "epoch": 20.35, "learning_rate": 2.4168786862495662e-05, "loss": 0.8019, "step": 234920 }, { "epoch": 20.35, "learning_rate": 2.4166618480397825e-05, "loss": 0.6962, "step": 234930 }, { "epoch": 20.35, "learning_rate": 2.4164450098299988e-05, "loss": 0.6932, "step": 234940 }, { "epoch": 20.35, "learning_rate": 2.4162281716202148e-05, "loss": 0.6791, "step": 234950 }, { "epoch": 20.35, "learning_rate": 2.416011333410431e-05, "loss": 0.7265, "step": 234960 }, { "epoch": 20.35, "learning_rate": 2.4157944952006474e-05, "loss": 0.7389, "step": 234970 }, { "epoch": 20.35, "learning_rate": 2.415577656990864e-05, "loss": 0.8208, "step": 234980 }, { "epoch": 20.35, "learning_rate": 2.4153608187810796e-05, "loss": 0.8281, "step": 234990 }, { "epoch": 20.35, "learning_rate": 2.4151439805712963e-05, "loss": 0.7767, "step": 235000 }, { "epoch": 20.35, "learning_rate": 2.4149271423615126e-05, "loss": 0.8199, "step": 235010 }, { "epoch": 20.35, "learning_rate": 2.414710304151729e-05, "loss": 0.7517, "step": 235020 }, { "epoch": 20.36, "learning_rate": 2.414493465941945e-05, "loss": 0.6595, "step": 235030 }, { "epoch": 20.36, "learning_rate": 2.414276627732161e-05, "loss": 0.691, "step": 235040 }, { "epoch": 20.36, "learning_rate": 2.4140597895223774e-05, "loss": 0.7428, "step": 235050 }, { "epoch": 20.36, "learning_rate": 2.4138429513125934e-05, "loss": 0.7359, "step": 235060 }, { "epoch": 20.36, "learning_rate": 2.41362611310281e-05, "loss": 0.7461, "step": 235070 }, { "epoch": 20.36, "learning_rate": 2.4134092748930264e-05, "loss": 0.6978, "step": 235080 }, { "epoch": 20.36, "learning_rate": 2.4131924366832427e-05, "loss": 0.6318, "step": 235090 }, { "epoch": 20.36, "learning_rate": 2.4129755984734586e-05, "loss": 0.6766, "step": 235100 }, { "epoch": 20.36, "learning_rate": 2.412758760263675e-05, "loss": 0.7461, "step": 235110 }, { "epoch": 20.36, "learning_rate": 2.4125419220538912e-05, "loss": 0.6913, "step": 235120 }, { "epoch": 20.36, "learning_rate": 2.412325083844108e-05, "loss": 0.7471, "step": 235130 }, { "epoch": 20.36, "learning_rate": 2.4121082456343238e-05, "loss": 0.6468, "step": 235140 }, { "epoch": 20.37, "learning_rate": 2.41189140742454e-05, "loss": 0.7169, "step": 235150 }, { "epoch": 20.37, "learning_rate": 2.4116745692147564e-05, "loss": 0.6993, "step": 235160 }, { "epoch": 20.37, "learning_rate": 2.4114577310049727e-05, "loss": 0.821, "step": 235170 }, { "epoch": 20.37, "learning_rate": 2.4112408927951887e-05, "loss": 0.7973, "step": 235180 }, { "epoch": 20.37, "learning_rate": 2.411024054585405e-05, "loss": 0.8595, "step": 235190 }, { "epoch": 20.37, "learning_rate": 2.4108072163756216e-05, "loss": 0.7289, "step": 235200 }, { "epoch": 20.37, "learning_rate": 2.410590378165838e-05, "loss": 0.699, "step": 235210 }, { "epoch": 20.37, "learning_rate": 2.410373539956054e-05, "loss": 0.9162, "step": 235220 }, { "epoch": 20.37, "learning_rate": 2.4101567017462702e-05, "loss": 0.65, "step": 235230 }, { "epoch": 20.37, "learning_rate": 2.4099398635364865e-05, "loss": 0.7005, "step": 235240 }, { "epoch": 20.37, "learning_rate": 2.4097230253267028e-05, "loss": 0.7353, "step": 235250 }, { "epoch": 20.38, "learning_rate": 2.4095061871169187e-05, "loss": 0.7931, "step": 235260 }, { "epoch": 20.38, "learning_rate": 2.409289348907135e-05, "loss": 0.6962, "step": 235270 }, { "epoch": 20.38, "learning_rate": 2.4090725106973517e-05, "loss": 0.7733, "step": 235280 }, { "epoch": 20.38, "learning_rate": 2.4088556724875676e-05, "loss": 0.7496, "step": 235290 }, { "epoch": 20.38, "learning_rate": 2.408638834277784e-05, "loss": 0.7714, "step": 235300 }, { "epoch": 20.38, "learning_rate": 2.4084219960680002e-05, "loss": 0.6864, "step": 235310 }, { "epoch": 20.38, "learning_rate": 2.4082051578582165e-05, "loss": 0.8202, "step": 235320 }, { "epoch": 20.38, "learning_rate": 2.4079883196484325e-05, "loss": 0.8071, "step": 235330 }, { "epoch": 20.38, "learning_rate": 2.4077714814386488e-05, "loss": 0.6782, "step": 235340 }, { "epoch": 20.38, "learning_rate": 2.4075546432288654e-05, "loss": 0.603, "step": 235350 }, { "epoch": 20.38, "learning_rate": 2.4073378050190817e-05, "loss": 0.7824, "step": 235360 }, { "epoch": 20.38, "learning_rate": 2.4071209668092977e-05, "loss": 0.7327, "step": 235370 }, { "epoch": 20.39, "learning_rate": 2.406904128599514e-05, "loss": 0.7947, "step": 235380 }, { "epoch": 20.39, "learning_rate": 2.4066872903897303e-05, "loss": 0.7013, "step": 235390 }, { "epoch": 20.39, "learning_rate": 2.4064704521799466e-05, "loss": 0.7487, "step": 235400 }, { "epoch": 20.39, "learning_rate": 2.4062536139701626e-05, "loss": 0.6923, "step": 235410 }, { "epoch": 20.39, "learning_rate": 2.4060367757603792e-05, "loss": 0.6611, "step": 235420 }, { "epoch": 20.39, "learning_rate": 2.4058199375505955e-05, "loss": 0.7655, "step": 235430 }, { "epoch": 20.39, "learning_rate": 2.4056030993408118e-05, "loss": 0.6679, "step": 235440 }, { "epoch": 20.39, "learning_rate": 2.4053862611310278e-05, "loss": 0.8, "step": 235450 }, { "epoch": 20.39, "learning_rate": 2.405169422921244e-05, "loss": 0.6914, "step": 235460 }, { "epoch": 20.39, "learning_rate": 2.4049525847114604e-05, "loss": 0.6874, "step": 235470 }, { "epoch": 20.39, "learning_rate": 2.404735746501677e-05, "loss": 0.7852, "step": 235480 }, { "epoch": 20.39, "learning_rate": 2.404518908291893e-05, "loss": 0.8616, "step": 235490 }, { "epoch": 20.4, "learning_rate": 2.4043020700821093e-05, "loss": 0.8158, "step": 235500 }, { "epoch": 20.4, "learning_rate": 2.4040852318723256e-05, "loss": 0.8021, "step": 235510 }, { "epoch": 20.4, "learning_rate": 2.403868393662542e-05, "loss": 0.6532, "step": 235520 }, { "epoch": 20.4, "learning_rate": 2.403651555452758e-05, "loss": 0.8109, "step": 235530 }, { "epoch": 20.4, "learning_rate": 2.403434717242974e-05, "loss": 0.755, "step": 235540 }, { "epoch": 20.4, "learning_rate": 2.4032178790331904e-05, "loss": 0.8705, "step": 235550 }, { "epoch": 20.4, "learning_rate": 2.4030010408234064e-05, "loss": 0.6923, "step": 235560 }, { "epoch": 20.4, "learning_rate": 2.402784202613623e-05, "loss": 0.6854, "step": 235570 }, { "epoch": 20.4, "learning_rate": 2.4025673644038393e-05, "loss": 0.8083, "step": 235580 }, { "epoch": 20.4, "learning_rate": 2.4023505261940556e-05, "loss": 0.7162, "step": 235590 }, { "epoch": 20.4, "learning_rate": 2.4021336879842716e-05, "loss": 0.7267, "step": 235600 }, { "epoch": 20.41, "learning_rate": 2.401916849774488e-05, "loss": 0.8289, "step": 235610 }, { "epoch": 20.41, "learning_rate": 2.4017000115647042e-05, "loss": 0.7583, "step": 235620 }, { "epoch": 20.41, "learning_rate": 2.401483173354921e-05, "loss": 0.7144, "step": 235630 }, { "epoch": 20.41, "learning_rate": 2.4012663351451368e-05, "loss": 0.6315, "step": 235640 }, { "epoch": 20.41, "learning_rate": 2.401049496935353e-05, "loss": 0.718, "step": 235650 }, { "epoch": 20.41, "learning_rate": 2.4008326587255694e-05, "loss": 0.8544, "step": 235660 }, { "epoch": 20.41, "learning_rate": 2.4006158205157857e-05, "loss": 0.729, "step": 235670 }, { "epoch": 20.41, "learning_rate": 2.4003989823060017e-05, "loss": 0.7614, "step": 235680 }, { "epoch": 20.41, "learning_rate": 2.400182144096218e-05, "loss": 0.5419, "step": 235690 }, { "epoch": 20.41, "learning_rate": 2.3999653058864346e-05, "loss": 0.6517, "step": 235700 }, { "epoch": 20.41, "learning_rate": 2.399748467676651e-05, "loss": 0.7704, "step": 235710 }, { "epoch": 20.41, "learning_rate": 2.399531629466867e-05, "loss": 0.8608, "step": 235720 }, { "epoch": 20.42, "learning_rate": 2.399314791257083e-05, "loss": 0.7826, "step": 235730 }, { "epoch": 20.42, "learning_rate": 2.3990979530472995e-05, "loss": 0.7, "step": 235740 }, { "epoch": 20.42, "learning_rate": 2.3988811148375158e-05, "loss": 0.7234, "step": 235750 }, { "epoch": 20.42, "learning_rate": 2.3986642766277317e-05, "loss": 0.7383, "step": 235760 }, { "epoch": 20.42, "learning_rate": 2.3984474384179484e-05, "loss": 0.6297, "step": 235770 }, { "epoch": 20.42, "learning_rate": 2.3982306002081647e-05, "loss": 0.979, "step": 235780 }, { "epoch": 20.42, "learning_rate": 2.398013761998381e-05, "loss": 0.7565, "step": 235790 }, { "epoch": 20.42, "learning_rate": 2.397796923788597e-05, "loss": 0.7136, "step": 235800 }, { "epoch": 20.42, "learning_rate": 2.3975800855788132e-05, "loss": 0.6951, "step": 235810 }, { "epoch": 20.42, "learning_rate": 2.3973632473690295e-05, "loss": 0.6603, "step": 235820 }, { "epoch": 20.42, "learning_rate": 2.3971464091592455e-05, "loss": 0.7616, "step": 235830 }, { "epoch": 20.43, "learning_rate": 2.3969295709494618e-05, "loss": 0.6549, "step": 235840 }, { "epoch": 20.43, "learning_rate": 2.3967127327396784e-05, "loss": 0.7069, "step": 235850 }, { "epoch": 20.43, "learning_rate": 2.3964958945298947e-05, "loss": 0.6764, "step": 235860 }, { "epoch": 20.43, "learning_rate": 2.3962790563201107e-05, "loss": 0.5416, "step": 235870 }, { "epoch": 20.43, "learning_rate": 2.396062218110327e-05, "loss": 0.688, "step": 235880 }, { "epoch": 20.43, "learning_rate": 2.3958453799005433e-05, "loss": 0.7303, "step": 235890 }, { "epoch": 20.43, "learning_rate": 2.3956285416907596e-05, "loss": 0.7116, "step": 235900 }, { "epoch": 20.43, "learning_rate": 2.3954117034809756e-05, "loss": 0.7887, "step": 235910 }, { "epoch": 20.43, "learning_rate": 2.3951948652711922e-05, "loss": 0.8657, "step": 235920 }, { "epoch": 20.43, "learning_rate": 2.3949780270614085e-05, "loss": 0.6071, "step": 235930 }, { "epoch": 20.43, "learning_rate": 2.3947611888516248e-05, "loss": 0.7244, "step": 235940 }, { "epoch": 20.43, "learning_rate": 2.3945443506418408e-05, "loss": 0.8543, "step": 235950 }, { "epoch": 20.44, "learning_rate": 2.394327512432057e-05, "loss": 0.5944, "step": 235960 }, { "epoch": 20.44, "learning_rate": 2.3941106742222734e-05, "loss": 0.7298, "step": 235970 }, { "epoch": 20.44, "learning_rate": 2.39389383601249e-05, "loss": 0.7244, "step": 235980 }, { "epoch": 20.44, "learning_rate": 2.393676997802706e-05, "loss": 0.8802, "step": 235990 }, { "epoch": 20.44, "learning_rate": 2.3934601595929223e-05, "loss": 0.6863, "step": 236000 }, { "epoch": 20.44, "learning_rate": 2.3932433213831386e-05, "loss": 0.8041, "step": 236010 }, { "epoch": 20.44, "learning_rate": 2.393026483173355e-05, "loss": 0.5936, "step": 236020 }, { "epoch": 20.44, "learning_rate": 2.3928096449635708e-05, "loss": 0.7497, "step": 236030 }, { "epoch": 20.44, "learning_rate": 2.392592806753787e-05, "loss": 0.7424, "step": 236040 }, { "epoch": 20.44, "learning_rate": 2.3923759685440038e-05, "loss": 0.7097, "step": 236050 }, { "epoch": 20.44, "learning_rate": 2.3921591303342194e-05, "loss": 0.8936, "step": 236060 }, { "epoch": 20.45, "learning_rate": 2.391942292124436e-05, "loss": 0.8431, "step": 236070 }, { "epoch": 20.45, "learning_rate": 2.3917254539146523e-05, "loss": 0.743, "step": 236080 }, { "epoch": 20.45, "learning_rate": 2.3915086157048686e-05, "loss": 0.7206, "step": 236090 }, { "epoch": 20.45, "learning_rate": 2.3912917774950846e-05, "loss": 0.7455, "step": 236100 }, { "epoch": 20.45, "learning_rate": 2.391074939285301e-05, "loss": 0.621, "step": 236110 }, { "epoch": 20.45, "learning_rate": 2.3908581010755172e-05, "loss": 0.7266, "step": 236120 }, { "epoch": 20.45, "learning_rate": 2.3906412628657338e-05, "loss": 0.6093, "step": 236130 }, { "epoch": 20.45, "learning_rate": 2.3904244246559498e-05, "loss": 0.8926, "step": 236140 }, { "epoch": 20.45, "learning_rate": 2.390207586446166e-05, "loss": 0.8462, "step": 236150 }, { "epoch": 20.45, "learning_rate": 2.3899907482363824e-05, "loss": 0.7269, "step": 236160 }, { "epoch": 20.45, "learning_rate": 2.3897739100265987e-05, "loss": 0.824, "step": 236170 }, { "epoch": 20.45, "learning_rate": 2.3895570718168146e-05, "loss": 0.7771, "step": 236180 }, { "epoch": 20.46, "learning_rate": 2.389340233607031e-05, "loss": 0.6972, "step": 236190 }, { "epoch": 20.46, "learning_rate": 2.3891233953972476e-05, "loss": 0.8686, "step": 236200 }, { "epoch": 20.46, "learning_rate": 2.388906557187464e-05, "loss": 0.8334, "step": 236210 }, { "epoch": 20.46, "learning_rate": 2.38868971897768e-05, "loss": 0.7307, "step": 236220 }, { "epoch": 20.46, "learning_rate": 2.388472880767896e-05, "loss": 0.7796, "step": 236230 }, { "epoch": 20.46, "learning_rate": 2.3882560425581124e-05, "loss": 0.7322, "step": 236240 }, { "epoch": 20.46, "learning_rate": 2.3880392043483287e-05, "loss": 0.6904, "step": 236250 }, { "epoch": 20.46, "learning_rate": 2.3878223661385447e-05, "loss": 0.7417, "step": 236260 }, { "epoch": 20.46, "learning_rate": 2.3876055279287613e-05, "loss": 0.6537, "step": 236270 }, { "epoch": 20.46, "learning_rate": 2.3873886897189776e-05, "loss": 0.6297, "step": 236280 }, { "epoch": 20.46, "learning_rate": 2.387171851509194e-05, "loss": 0.8155, "step": 236290 }, { "epoch": 20.47, "learning_rate": 2.38695501329941e-05, "loss": 0.6041, "step": 236300 }, { "epoch": 20.47, "learning_rate": 2.3867381750896262e-05, "loss": 0.7041, "step": 236310 }, { "epoch": 20.47, "learning_rate": 2.3865213368798425e-05, "loss": 0.6621, "step": 236320 }, { "epoch": 20.47, "learning_rate": 2.3863044986700585e-05, "loss": 0.6564, "step": 236330 }, { "epoch": 20.47, "learning_rate": 2.3860876604602748e-05, "loss": 0.7221, "step": 236340 }, { "epoch": 20.47, "learning_rate": 2.3858708222504914e-05, "loss": 0.7097, "step": 236350 }, { "epoch": 20.47, "learning_rate": 2.3856539840407077e-05, "loss": 0.7922, "step": 236360 }, { "epoch": 20.47, "learning_rate": 2.3854371458309237e-05, "loss": 0.7662, "step": 236370 }, { "epoch": 20.47, "learning_rate": 2.38522030762114e-05, "loss": 0.8838, "step": 236380 }, { "epoch": 20.47, "learning_rate": 2.3850034694113563e-05, "loss": 0.6253, "step": 236390 }, { "epoch": 20.47, "learning_rate": 2.3847866312015726e-05, "loss": 0.6135, "step": 236400 }, { "epoch": 20.47, "learning_rate": 2.3845697929917885e-05, "loss": 0.7655, "step": 236410 }, { "epoch": 20.48, "learning_rate": 2.3843529547820052e-05, "loss": 0.6587, "step": 236420 }, { "epoch": 20.48, "learning_rate": 2.3841361165722215e-05, "loss": 0.6523, "step": 236430 }, { "epoch": 20.48, "learning_rate": 2.3839192783624378e-05, "loss": 0.6453, "step": 236440 }, { "epoch": 20.48, "learning_rate": 2.3837024401526537e-05, "loss": 0.7105, "step": 236450 }, { "epoch": 20.48, "learning_rate": 2.38348560194287e-05, "loss": 0.7531, "step": 236460 }, { "epoch": 20.48, "learning_rate": 2.3832687637330863e-05, "loss": 0.899, "step": 236470 }, { "epoch": 20.48, "learning_rate": 2.383051925523303e-05, "loss": 0.6991, "step": 236480 }, { "epoch": 20.48, "learning_rate": 2.382835087313519e-05, "loss": 0.7535, "step": 236490 }, { "epoch": 20.48, "learning_rate": 2.3826182491037352e-05, "loss": 0.6669, "step": 236500 }, { "epoch": 20.48, "learning_rate": 2.3824014108939515e-05, "loss": 0.7813, "step": 236510 }, { "epoch": 20.48, "learning_rate": 2.382184572684168e-05, "loss": 0.7347, "step": 236520 }, { "epoch": 20.48, "learning_rate": 2.3819677344743838e-05, "loss": 0.7616, "step": 236530 }, { "epoch": 20.49, "learning_rate": 2.3817508962646e-05, "loss": 0.7096, "step": 236540 }, { "epoch": 20.49, "learning_rate": 2.3815340580548167e-05, "loss": 0.7705, "step": 236550 }, { "epoch": 20.49, "learning_rate": 2.381317219845033e-05, "loss": 0.8062, "step": 236560 }, { "epoch": 20.49, "learning_rate": 2.381100381635249e-05, "loss": 0.7735, "step": 236570 }, { "epoch": 20.49, "learning_rate": 2.3808835434254653e-05, "loss": 0.6847, "step": 236580 }, { "epoch": 20.49, "learning_rate": 2.3806667052156816e-05, "loss": 0.5391, "step": 236590 }, { "epoch": 20.49, "learning_rate": 2.3804498670058976e-05, "loss": 0.6606, "step": 236600 }, { "epoch": 20.49, "learning_rate": 2.380233028796114e-05, "loss": 0.63, "step": 236610 }, { "epoch": 20.49, "learning_rate": 2.38001619058633e-05, "loss": 0.6635, "step": 236620 }, { "epoch": 20.49, "learning_rate": 2.3797993523765468e-05, "loss": 0.7533, "step": 236630 }, { "epoch": 20.49, "learning_rate": 2.3795825141667628e-05, "loss": 0.6335, "step": 236640 }, { "epoch": 20.5, "learning_rate": 2.379365675956979e-05, "loss": 0.9025, "step": 236650 }, { "epoch": 20.5, "learning_rate": 2.3791488377471954e-05, "loss": 0.6477, "step": 236660 }, { "epoch": 20.5, "learning_rate": 2.3789319995374117e-05, "loss": 0.711, "step": 236670 }, { "epoch": 20.5, "learning_rate": 2.3787151613276276e-05, "loss": 0.7481, "step": 236680 }, { "epoch": 20.5, "learning_rate": 2.378498323117844e-05, "loss": 0.672, "step": 236690 }, { "epoch": 20.5, "learning_rate": 2.3782814849080606e-05, "loss": 0.6427, "step": 236700 }, { "epoch": 20.5, "learning_rate": 2.378064646698277e-05, "loss": 0.7985, "step": 236710 }, { "epoch": 20.5, "learning_rate": 2.377847808488493e-05, "loss": 0.6749, "step": 236720 }, { "epoch": 20.5, "learning_rate": 2.377630970278709e-05, "loss": 0.8258, "step": 236730 }, { "epoch": 20.5, "learning_rate": 2.3774141320689254e-05, "loss": 0.7729, "step": 236740 }, { "epoch": 20.5, "learning_rate": 2.3771972938591417e-05, "loss": 0.564, "step": 236750 }, { "epoch": 20.5, "learning_rate": 2.3769804556493577e-05, "loss": 0.7362, "step": 236760 }, { "epoch": 20.51, "learning_rate": 2.3767636174395743e-05, "loss": 0.7241, "step": 236770 }, { "epoch": 20.51, "learning_rate": 2.3765467792297906e-05, "loss": 0.7612, "step": 236780 }, { "epoch": 20.51, "learning_rate": 2.376329941020007e-05, "loss": 0.7459, "step": 236790 }, { "epoch": 20.51, "learning_rate": 2.376113102810223e-05, "loss": 0.8844, "step": 236800 }, { "epoch": 20.51, "learning_rate": 2.3758962646004392e-05, "loss": 0.6102, "step": 236810 }, { "epoch": 20.51, "learning_rate": 2.3756794263906555e-05, "loss": 0.8476, "step": 236820 }, { "epoch": 20.51, "learning_rate": 2.3754625881808715e-05, "loss": 0.7752, "step": 236830 }, { "epoch": 20.51, "learning_rate": 2.375245749971088e-05, "loss": 0.8025, "step": 236840 }, { "epoch": 20.51, "learning_rate": 2.3750289117613044e-05, "loss": 0.6467, "step": 236850 }, { "epoch": 20.51, "learning_rate": 2.3748120735515207e-05, "loss": 0.5927, "step": 236860 }, { "epoch": 20.51, "learning_rate": 2.3745952353417367e-05, "loss": 0.6555, "step": 236870 }, { "epoch": 20.52, "learning_rate": 2.374378397131953e-05, "loss": 0.6961, "step": 236880 }, { "epoch": 20.52, "learning_rate": 2.3741615589221693e-05, "loss": 0.6289, "step": 236890 }, { "epoch": 20.52, "learning_rate": 2.3739447207123856e-05, "loss": 0.7342, "step": 236900 }, { "epoch": 20.52, "learning_rate": 2.3737278825026015e-05, "loss": 0.6497, "step": 236910 }, { "epoch": 20.52, "learning_rate": 2.373511044292818e-05, "loss": 0.6152, "step": 236920 }, { "epoch": 20.52, "learning_rate": 2.3732942060830345e-05, "loss": 0.7203, "step": 236930 }, { "epoch": 20.52, "learning_rate": 2.3730773678732508e-05, "loss": 0.7922, "step": 236940 }, { "epoch": 20.52, "learning_rate": 2.3728605296634667e-05, "loss": 0.7251, "step": 236950 }, { "epoch": 20.52, "learning_rate": 2.372643691453683e-05, "loss": 0.7306, "step": 236960 }, { "epoch": 20.52, "learning_rate": 2.3724268532438993e-05, "loss": 0.7108, "step": 236970 }, { "epoch": 20.52, "learning_rate": 2.372210015034116e-05, "loss": 0.7388, "step": 236980 }, { "epoch": 20.52, "learning_rate": 2.371993176824332e-05, "loss": 0.6169, "step": 236990 }, { "epoch": 20.53, "learning_rate": 2.3717763386145482e-05, "loss": 0.7771, "step": 237000 }, { "epoch": 20.53, "learning_rate": 2.3715595004047645e-05, "loss": 0.6966, "step": 237010 }, { "epoch": 20.53, "learning_rate": 2.3713426621949808e-05, "loss": 0.7599, "step": 237020 }, { "epoch": 20.53, "learning_rate": 2.3711258239851968e-05, "loss": 0.7365, "step": 237030 }, { "epoch": 20.53, "learning_rate": 2.370908985775413e-05, "loss": 0.8645, "step": 237040 }, { "epoch": 20.53, "learning_rate": 2.3706921475656297e-05, "loss": 0.7177, "step": 237050 }, { "epoch": 20.53, "learning_rate": 2.370475309355846e-05, "loss": 0.63, "step": 237060 }, { "epoch": 20.53, "learning_rate": 2.370258471146062e-05, "loss": 0.7308, "step": 237070 }, { "epoch": 20.53, "learning_rate": 2.3700416329362783e-05, "loss": 0.7181, "step": 237080 }, { "epoch": 20.53, "learning_rate": 2.3698247947264946e-05, "loss": 0.7452, "step": 237090 }, { "epoch": 20.53, "learning_rate": 2.3696079565167106e-05, "loss": 0.6832, "step": 237100 }, { "epoch": 20.54, "learning_rate": 2.369391118306927e-05, "loss": 0.6388, "step": 237110 }, { "epoch": 20.54, "learning_rate": 2.3691742800971435e-05, "loss": 0.7042, "step": 237120 }, { "epoch": 20.54, "learning_rate": 2.3689574418873598e-05, "loss": 0.8259, "step": 237130 }, { "epoch": 20.54, "learning_rate": 2.3687406036775758e-05, "loss": 0.7289, "step": 237140 }, { "epoch": 20.54, "learning_rate": 2.368523765467792e-05, "loss": 0.7767, "step": 237150 }, { "epoch": 20.54, "learning_rate": 2.3683069272580084e-05, "loss": 0.7321, "step": 237160 }, { "epoch": 20.54, "learning_rate": 2.3680900890482247e-05, "loss": 0.8605, "step": 237170 }, { "epoch": 20.54, "learning_rate": 2.3678732508384406e-05, "loss": 0.7497, "step": 237180 }, { "epoch": 20.54, "learning_rate": 2.367656412628657e-05, "loss": 0.7347, "step": 237190 }, { "epoch": 20.54, "learning_rate": 2.3674395744188736e-05, "loss": 0.784, "step": 237200 }, { "epoch": 20.54, "learning_rate": 2.36722273620909e-05, "loss": 0.819, "step": 237210 }, { "epoch": 20.54, "learning_rate": 2.3670058979993058e-05, "loss": 0.8618, "step": 237220 }, { "epoch": 20.55, "learning_rate": 2.366789059789522e-05, "loss": 0.8453, "step": 237230 }, { "epoch": 20.55, "learning_rate": 2.3665722215797384e-05, "loss": 0.7111, "step": 237240 }, { "epoch": 20.55, "learning_rate": 2.3663553833699547e-05, "loss": 0.7308, "step": 237250 }, { "epoch": 20.55, "learning_rate": 2.3661385451601707e-05, "loss": 0.7238, "step": 237260 }, { "epoch": 20.55, "learning_rate": 2.3659217069503873e-05, "loss": 0.7101, "step": 237270 }, { "epoch": 20.55, "learning_rate": 2.3657048687406036e-05, "loss": 0.6859, "step": 237280 }, { "epoch": 20.55, "learning_rate": 2.36548803053082e-05, "loss": 0.8427, "step": 237290 }, { "epoch": 20.55, "learning_rate": 2.365271192321036e-05, "loss": 0.8048, "step": 237300 }, { "epoch": 20.55, "learning_rate": 2.3650543541112522e-05, "loss": 0.6506, "step": 237310 }, { "epoch": 20.55, "learning_rate": 2.3648375159014685e-05, "loss": 0.6732, "step": 237320 }, { "epoch": 20.55, "learning_rate": 2.364620677691685e-05, "loss": 0.7932, "step": 237330 }, { "epoch": 20.56, "learning_rate": 2.364403839481901e-05, "loss": 0.7149, "step": 237340 }, { "epoch": 20.56, "learning_rate": 2.3641870012721174e-05, "loss": 0.6082, "step": 237350 }, { "epoch": 20.56, "learning_rate": 2.3639701630623337e-05, "loss": 0.6633, "step": 237360 }, { "epoch": 20.56, "learning_rate": 2.3637533248525496e-05, "loss": 0.6441, "step": 237370 }, { "epoch": 20.56, "learning_rate": 2.363536486642766e-05, "loss": 0.6498, "step": 237380 }, { "epoch": 20.56, "learning_rate": 2.3633196484329822e-05, "loss": 0.7611, "step": 237390 }, { "epoch": 20.56, "learning_rate": 2.363102810223199e-05, "loss": 0.8664, "step": 237400 }, { "epoch": 20.56, "learning_rate": 2.3628859720134145e-05, "loss": 0.8858, "step": 237410 }, { "epoch": 20.56, "learning_rate": 2.362669133803631e-05, "loss": 0.7213, "step": 237420 }, { "epoch": 20.56, "learning_rate": 2.3624522955938474e-05, "loss": 0.7896, "step": 237430 }, { "epoch": 20.56, "learning_rate": 2.3622354573840637e-05, "loss": 0.738, "step": 237440 }, { "epoch": 20.56, "learning_rate": 2.3620186191742797e-05, "loss": 0.7945, "step": 237450 }, { "epoch": 20.57, "learning_rate": 2.361801780964496e-05, "loss": 0.8238, "step": 237460 }, { "epoch": 20.57, "learning_rate": 2.3615849427547123e-05, "loss": 0.6735, "step": 237470 }, { "epoch": 20.57, "learning_rate": 2.361368104544929e-05, "loss": 0.6951, "step": 237480 }, { "epoch": 20.57, "learning_rate": 2.361151266335145e-05, "loss": 0.6235, "step": 237490 }, { "epoch": 20.57, "learning_rate": 2.3609344281253612e-05, "loss": 0.7205, "step": 237500 }, { "epoch": 20.57, "learning_rate": 2.3607175899155775e-05, "loss": 0.7012, "step": 237510 }, { "epoch": 20.57, "learning_rate": 2.3605007517057938e-05, "loss": 0.6694, "step": 237520 }, { "epoch": 20.57, "learning_rate": 2.3602839134960098e-05, "loss": 0.6608, "step": 237530 }, { "epoch": 20.57, "learning_rate": 2.360067075286226e-05, "loss": 0.6699, "step": 237540 }, { "epoch": 20.57, "learning_rate": 2.3598502370764427e-05, "loss": 0.7323, "step": 237550 }, { "epoch": 20.57, "learning_rate": 2.359633398866659e-05, "loss": 0.6565, "step": 237560 }, { "epoch": 20.58, "learning_rate": 2.359416560656875e-05, "loss": 0.739, "step": 237570 }, { "epoch": 20.58, "learning_rate": 2.3591997224470913e-05, "loss": 0.7046, "step": 237580 }, { "epoch": 20.58, "learning_rate": 2.3589828842373076e-05, "loss": 0.8204, "step": 237590 }, { "epoch": 20.58, "learning_rate": 2.3587660460275235e-05, "loss": 0.8, "step": 237600 }, { "epoch": 20.58, "learning_rate": 2.35854920781774e-05, "loss": 0.8096, "step": 237610 }, { "epoch": 20.58, "learning_rate": 2.3583323696079565e-05, "loss": 0.7901, "step": 237620 }, { "epoch": 20.58, "learning_rate": 2.3581155313981728e-05, "loss": 0.7715, "step": 237630 }, { "epoch": 20.58, "learning_rate": 2.3578986931883887e-05, "loss": 0.7002, "step": 237640 }, { "epoch": 20.58, "learning_rate": 2.357681854978605e-05, "loss": 0.7362, "step": 237650 }, { "epoch": 20.58, "learning_rate": 2.3574650167688213e-05, "loss": 0.7327, "step": 237660 }, { "epoch": 20.58, "learning_rate": 2.3572481785590376e-05, "loss": 0.7486, "step": 237670 }, { "epoch": 20.58, "learning_rate": 2.3570313403492536e-05, "loss": 0.5712, "step": 237680 }, { "epoch": 20.59, "learning_rate": 2.35681450213947e-05, "loss": 0.8095, "step": 237690 }, { "epoch": 20.59, "learning_rate": 2.3565976639296865e-05, "loss": 0.5847, "step": 237700 }, { "epoch": 20.59, "learning_rate": 2.356380825719903e-05, "loss": 0.7758, "step": 237710 }, { "epoch": 20.59, "learning_rate": 2.3561639875101188e-05, "loss": 0.8342, "step": 237720 }, { "epoch": 20.59, "learning_rate": 2.355947149300335e-05, "loss": 0.8029, "step": 237730 }, { "epoch": 20.59, "learning_rate": 2.3557303110905514e-05, "loss": 0.7112, "step": 237740 }, { "epoch": 20.59, "learning_rate": 2.3555134728807677e-05, "loss": 0.7969, "step": 237750 }, { "epoch": 20.59, "learning_rate": 2.3552966346709837e-05, "loss": 0.7423, "step": 237760 }, { "epoch": 20.59, "learning_rate": 2.3550797964612003e-05, "loss": 0.7066, "step": 237770 }, { "epoch": 20.59, "learning_rate": 2.3548629582514166e-05, "loss": 0.7867, "step": 237780 }, { "epoch": 20.59, "learning_rate": 2.354646120041633e-05, "loss": 0.6591, "step": 237790 }, { "epoch": 20.59, "learning_rate": 2.354429281831849e-05, "loss": 0.6251, "step": 237800 }, { "epoch": 20.6, "learning_rate": 2.354212443622065e-05, "loss": 0.7389, "step": 237810 }, { "epoch": 20.6, "learning_rate": 2.3539956054122815e-05, "loss": 0.7921, "step": 237820 }, { "epoch": 20.6, "learning_rate": 2.353778767202498e-05, "loss": 0.7058, "step": 237830 }, { "epoch": 20.6, "learning_rate": 2.353561928992714e-05, "loss": 0.7545, "step": 237840 }, { "epoch": 20.6, "learning_rate": 2.3533450907829304e-05, "loss": 0.9085, "step": 237850 }, { "epoch": 20.6, "learning_rate": 2.3531282525731467e-05, "loss": 0.6855, "step": 237860 }, { "epoch": 20.6, "learning_rate": 2.3529114143633626e-05, "loss": 0.8117, "step": 237870 }, { "epoch": 20.6, "learning_rate": 2.352694576153579e-05, "loss": 0.8345, "step": 237880 }, { "epoch": 20.6, "learning_rate": 2.3524777379437952e-05, "loss": 0.7273, "step": 237890 }, { "epoch": 20.6, "learning_rate": 2.352260899734012e-05, "loss": 0.8286, "step": 237900 }, { "epoch": 20.6, "learning_rate": 2.352044061524228e-05, "loss": 0.72, "step": 237910 }, { "epoch": 20.61, "learning_rate": 2.351827223314444e-05, "loss": 0.7978, "step": 237920 }, { "epoch": 20.61, "learning_rate": 2.3516103851046604e-05, "loss": 0.8281, "step": 237930 }, { "epoch": 20.61, "learning_rate": 2.3513935468948767e-05, "loss": 0.6882, "step": 237940 }, { "epoch": 20.61, "learning_rate": 2.3511767086850927e-05, "loss": 0.6184, "step": 237950 }, { "epoch": 20.61, "learning_rate": 2.350959870475309e-05, "loss": 0.7133, "step": 237960 }, { "epoch": 20.61, "learning_rate": 2.3507430322655253e-05, "loss": 0.6682, "step": 237970 }, { "epoch": 20.61, "learning_rate": 2.350526194055742e-05, "loss": 0.8252, "step": 237980 }, { "epoch": 20.61, "learning_rate": 2.350309355845958e-05, "loss": 0.7602, "step": 237990 }, { "epoch": 20.61, "learning_rate": 2.3500925176361742e-05, "loss": 0.7044, "step": 238000 }, { "epoch": 20.61, "learning_rate": 2.3498756794263905e-05, "loss": 0.7138, "step": 238010 }, { "epoch": 20.61, "learning_rate": 2.3496588412166068e-05, "loss": 0.5996, "step": 238020 }, { "epoch": 20.61, "learning_rate": 2.3494420030068228e-05, "loss": 0.7422, "step": 238030 }, { "epoch": 20.62, "learning_rate": 2.349225164797039e-05, "loss": 0.7708, "step": 238040 }, { "epoch": 20.62, "learning_rate": 2.3490083265872557e-05, "loss": 0.7215, "step": 238050 }, { "epoch": 20.62, "learning_rate": 2.348791488377472e-05, "loss": 0.7713, "step": 238060 }, { "epoch": 20.62, "learning_rate": 2.348574650167688e-05, "loss": 0.8254, "step": 238070 }, { "epoch": 20.62, "learning_rate": 2.3483578119579043e-05, "loss": 0.774, "step": 238080 }, { "epoch": 20.62, "learning_rate": 2.3481409737481206e-05, "loss": 0.8891, "step": 238090 }, { "epoch": 20.62, "learning_rate": 2.347924135538337e-05, "loss": 0.6348, "step": 238100 }, { "epoch": 20.62, "learning_rate": 2.3477072973285528e-05, "loss": 0.7878, "step": 238110 }, { "epoch": 20.62, "learning_rate": 2.3474904591187695e-05, "loss": 0.7643, "step": 238120 }, { "epoch": 20.62, "learning_rate": 2.3472736209089858e-05, "loss": 0.6271, "step": 238130 }, { "epoch": 20.62, "learning_rate": 2.3470567826992017e-05, "loss": 0.8071, "step": 238140 }, { "epoch": 20.63, "learning_rate": 2.346839944489418e-05, "loss": 0.8168, "step": 238150 }, { "epoch": 20.63, "learning_rate": 2.3466231062796343e-05, "loss": 0.5591, "step": 238160 }, { "epoch": 20.63, "learning_rate": 2.3464062680698506e-05, "loss": 0.7535, "step": 238170 }, { "epoch": 20.63, "learning_rate": 2.3461894298600666e-05, "loss": 0.7735, "step": 238180 }, { "epoch": 20.63, "learning_rate": 2.3459725916502832e-05, "loss": 0.7727, "step": 238190 }, { "epoch": 20.63, "learning_rate": 2.3457557534404995e-05, "loss": 0.7011, "step": 238200 }, { "epoch": 20.63, "learning_rate": 2.3455389152307158e-05, "loss": 0.6931, "step": 238210 }, { "epoch": 20.63, "learning_rate": 2.3453220770209318e-05, "loss": 0.8636, "step": 238220 }, { "epoch": 20.63, "learning_rate": 2.345105238811148e-05, "loss": 0.658, "step": 238230 }, { "epoch": 20.63, "learning_rate": 2.3448884006013644e-05, "loss": 0.8138, "step": 238240 }, { "epoch": 20.63, "learning_rate": 2.3446715623915807e-05, "loss": 0.8795, "step": 238250 }, { "epoch": 20.63, "learning_rate": 2.3444547241817966e-05, "loss": 0.6913, "step": 238260 }, { "epoch": 20.64, "learning_rate": 2.3442378859720133e-05, "loss": 0.7137, "step": 238270 }, { "epoch": 20.64, "learning_rate": 2.3440210477622296e-05, "loss": 0.6364, "step": 238280 }, { "epoch": 20.64, "learning_rate": 2.343804209552446e-05, "loss": 0.7425, "step": 238290 }, { "epoch": 20.64, "learning_rate": 2.343587371342662e-05, "loss": 0.6295, "step": 238300 }, { "epoch": 20.64, "learning_rate": 2.343370533132878e-05, "loss": 0.6415, "step": 238310 }, { "epoch": 20.64, "learning_rate": 2.3431536949230945e-05, "loss": 0.7315, "step": 238320 }, { "epoch": 20.64, "learning_rate": 2.342936856713311e-05, "loss": 0.9048, "step": 238330 }, { "epoch": 20.64, "learning_rate": 2.342720018503527e-05, "loss": 0.7467, "step": 238340 }, { "epoch": 20.64, "learning_rate": 2.3425031802937434e-05, "loss": 0.7359, "step": 238350 }, { "epoch": 20.64, "learning_rate": 2.3422863420839597e-05, "loss": 0.774, "step": 238360 }, { "epoch": 20.64, "learning_rate": 2.3420695038741756e-05, "loss": 0.7719, "step": 238370 }, { "epoch": 20.65, "learning_rate": 2.341852665664392e-05, "loss": 0.8896, "step": 238380 }, { "epoch": 20.65, "learning_rate": 2.3416358274546082e-05, "loss": 0.7218, "step": 238390 }, { "epoch": 20.65, "learning_rate": 2.341418989244825e-05, "loss": 0.7863, "step": 238400 }, { "epoch": 20.65, "learning_rate": 2.3412021510350408e-05, "loss": 0.714, "step": 238410 }, { "epoch": 20.65, "learning_rate": 2.340985312825257e-05, "loss": 0.8126, "step": 238420 }, { "epoch": 20.65, "learning_rate": 2.3407684746154734e-05, "loss": 0.7136, "step": 238430 }, { "epoch": 20.65, "learning_rate": 2.3405516364056897e-05, "loss": 0.8313, "step": 238440 }, { "epoch": 20.65, "learning_rate": 2.3403347981959057e-05, "loss": 0.5658, "step": 238450 }, { "epoch": 20.65, "learning_rate": 2.340117959986122e-05, "loss": 0.8521, "step": 238460 }, { "epoch": 20.65, "learning_rate": 2.3399011217763386e-05, "loss": 0.6184, "step": 238470 }, { "epoch": 20.65, "learning_rate": 2.339684283566555e-05, "loss": 0.6551, "step": 238480 }, { "epoch": 20.65, "learning_rate": 2.339467445356771e-05, "loss": 0.7296, "step": 238490 }, { "epoch": 20.66, "learning_rate": 2.3392506071469872e-05, "loss": 0.724, "step": 238500 }, { "epoch": 20.66, "learning_rate": 2.3390337689372035e-05, "loss": 0.7499, "step": 238510 }, { "epoch": 20.66, "learning_rate": 2.3388169307274198e-05, "loss": 0.7094, "step": 238520 }, { "epoch": 20.66, "learning_rate": 2.3386000925176357e-05, "loss": 0.6208, "step": 238530 }, { "epoch": 20.66, "learning_rate": 2.338383254307852e-05, "loss": 0.6236, "step": 238540 }, { "epoch": 20.66, "learning_rate": 2.3381664160980687e-05, "loss": 0.9375, "step": 238550 }, { "epoch": 20.66, "learning_rate": 2.337949577888285e-05, "loss": 0.7975, "step": 238560 }, { "epoch": 20.66, "learning_rate": 2.337732739678501e-05, "loss": 0.7927, "step": 238570 }, { "epoch": 20.66, "learning_rate": 2.3375159014687172e-05, "loss": 0.8714, "step": 238580 }, { "epoch": 20.66, "learning_rate": 2.3372990632589335e-05, "loss": 0.7477, "step": 238590 }, { "epoch": 20.66, "learning_rate": 2.33708222504915e-05, "loss": 0.7402, "step": 238600 }, { "epoch": 20.67, "learning_rate": 2.3368653868393658e-05, "loss": 0.7443, "step": 238610 }, { "epoch": 20.67, "learning_rate": 2.3366485486295824e-05, "loss": 0.7249, "step": 238620 }, { "epoch": 20.67, "learning_rate": 2.3364317104197987e-05, "loss": 0.7587, "step": 238630 }, { "epoch": 20.67, "learning_rate": 2.3362148722100147e-05, "loss": 0.6757, "step": 238640 }, { "epoch": 20.67, "learning_rate": 2.335998034000231e-05, "loss": 0.9011, "step": 238650 }, { "epoch": 20.67, "learning_rate": 2.3357811957904473e-05, "loss": 0.6488, "step": 238660 }, { "epoch": 20.67, "learning_rate": 2.3355643575806636e-05, "loss": 0.6912, "step": 238670 }, { "epoch": 20.67, "learning_rate": 2.3353475193708796e-05, "loss": 0.565, "step": 238680 }, { "epoch": 20.67, "learning_rate": 2.3351306811610962e-05, "loss": 0.8728, "step": 238690 }, { "epoch": 20.67, "learning_rate": 2.3349138429513125e-05, "loss": 0.7078, "step": 238700 }, { "epoch": 20.67, "learning_rate": 2.3346970047415288e-05, "loss": 0.5674, "step": 238710 }, { "epoch": 20.67, "learning_rate": 2.3344801665317448e-05, "loss": 0.6293, "step": 238720 }, { "epoch": 20.68, "learning_rate": 2.334263328321961e-05, "loss": 0.6319, "step": 238730 }, { "epoch": 20.68, "learning_rate": 2.3340464901121774e-05, "loss": 0.7792, "step": 238740 }, { "epoch": 20.68, "learning_rate": 2.333829651902394e-05, "loss": 0.7664, "step": 238750 }, { "epoch": 20.68, "learning_rate": 2.3336128136926096e-05, "loss": 0.8622, "step": 238760 }, { "epoch": 20.68, "learning_rate": 2.3333959754828263e-05, "loss": 0.8466, "step": 238770 }, { "epoch": 20.68, "learning_rate": 2.3331791372730426e-05, "loss": 0.8273, "step": 238780 }, { "epoch": 20.68, "learning_rate": 2.332962299063259e-05, "loss": 0.7722, "step": 238790 }, { "epoch": 20.68, "learning_rate": 2.332745460853475e-05, "loss": 0.8568, "step": 238800 }, { "epoch": 20.68, "learning_rate": 2.332528622643691e-05, "loss": 0.7466, "step": 238810 }, { "epoch": 20.68, "learning_rate": 2.3323117844339074e-05, "loss": 1.0037, "step": 238820 }, { "epoch": 20.68, "learning_rate": 2.332094946224124e-05, "loss": 0.7591, "step": 238830 }, { "epoch": 20.69, "learning_rate": 2.33187810801434e-05, "loss": 0.7399, "step": 238840 }, { "epoch": 20.69, "learning_rate": 2.3316612698045563e-05, "loss": 0.8092, "step": 238850 }, { "epoch": 20.69, "learning_rate": 2.3314444315947726e-05, "loss": 0.7837, "step": 238860 }, { "epoch": 20.69, "learning_rate": 2.3312275933849886e-05, "loss": 0.5995, "step": 238870 }, { "epoch": 20.69, "learning_rate": 2.331010755175205e-05, "loss": 0.6738, "step": 238880 }, { "epoch": 20.69, "learning_rate": 2.3307939169654212e-05, "loss": 0.645, "step": 238890 }, { "epoch": 20.69, "learning_rate": 2.330577078755638e-05, "loss": 0.8416, "step": 238900 }, { "epoch": 20.69, "learning_rate": 2.3303602405458538e-05, "loss": 0.745, "step": 238910 }, { "epoch": 20.69, "learning_rate": 2.33014340233607e-05, "loss": 0.8606, "step": 238920 }, { "epoch": 20.69, "learning_rate": 2.3299265641262864e-05, "loss": 0.7679, "step": 238930 }, { "epoch": 20.69, "learning_rate": 2.3297097259165027e-05, "loss": 0.8118, "step": 238940 }, { "epoch": 20.69, "learning_rate": 2.3294928877067187e-05, "loss": 0.7494, "step": 238950 }, { "epoch": 20.7, "learning_rate": 2.329276049496935e-05, "loss": 0.6384, "step": 238960 }, { "epoch": 20.7, "learning_rate": 2.3290592112871516e-05, "loss": 0.6113, "step": 238970 }, { "epoch": 20.7, "learning_rate": 2.328842373077368e-05, "loss": 0.8647, "step": 238980 }, { "epoch": 20.7, "learning_rate": 2.328625534867584e-05, "loss": 0.7204, "step": 238990 }, { "epoch": 20.7, "learning_rate": 2.3284086966578e-05, "loss": 0.871, "step": 239000 }, { "epoch": 20.7, "learning_rate": 2.3281918584480165e-05, "loss": 0.6905, "step": 239010 }, { "epoch": 20.7, "learning_rate": 2.3279750202382328e-05, "loss": 0.8558, "step": 239020 }, { "epoch": 20.7, "learning_rate": 2.3277581820284487e-05, "loss": 0.7129, "step": 239030 }, { "epoch": 20.7, "learning_rate": 2.327541343818665e-05, "loss": 0.674, "step": 239040 }, { "epoch": 20.7, "learning_rate": 2.3273245056088817e-05, "loss": 0.844, "step": 239050 }, { "epoch": 20.7, "learning_rate": 2.327107667399098e-05, "loss": 0.7032, "step": 239060 }, { "epoch": 20.7, "learning_rate": 2.326890829189314e-05, "loss": 0.5389, "step": 239070 }, { "epoch": 20.71, "learning_rate": 2.3266739909795302e-05, "loss": 0.6191, "step": 239080 }, { "epoch": 20.71, "learning_rate": 2.3264571527697465e-05, "loss": 0.5772, "step": 239090 }, { "epoch": 20.71, "learning_rate": 2.3262403145599628e-05, "loss": 0.8423, "step": 239100 }, { "epoch": 20.71, "learning_rate": 2.3260234763501788e-05, "loss": 0.754, "step": 239110 }, { "epoch": 20.71, "learning_rate": 2.3258066381403954e-05, "loss": 0.753, "step": 239120 }, { "epoch": 20.71, "learning_rate": 2.3255897999306117e-05, "loss": 0.654, "step": 239130 }, { "epoch": 20.71, "learning_rate": 2.3253729617208277e-05, "loss": 0.5608, "step": 239140 }, { "epoch": 20.71, "learning_rate": 2.325156123511044e-05, "loss": 0.9144, "step": 239150 }, { "epoch": 20.71, "learning_rate": 2.3249392853012603e-05, "loss": 0.5451, "step": 239160 }, { "epoch": 20.71, "learning_rate": 2.3247224470914766e-05, "loss": 0.7617, "step": 239170 }, { "epoch": 20.71, "learning_rate": 2.3245056088816926e-05, "loss": 0.8019, "step": 239180 }, { "epoch": 20.72, "learning_rate": 2.3242887706719092e-05, "loss": 0.7196, "step": 239190 }, { "epoch": 20.72, "learning_rate": 2.3240719324621255e-05, "loss": 0.6607, "step": 239200 }, { "epoch": 20.72, "learning_rate": 2.3238550942523418e-05, "loss": 0.8202, "step": 239210 }, { "epoch": 20.72, "learning_rate": 2.3236382560425578e-05, "loss": 0.5228, "step": 239220 }, { "epoch": 20.72, "learning_rate": 2.323421417832774e-05, "loss": 0.8182, "step": 239230 }, { "epoch": 20.72, "learning_rate": 2.3232045796229904e-05, "loss": 0.7648, "step": 239240 }, { "epoch": 20.72, "learning_rate": 2.322987741413207e-05, "loss": 0.8141, "step": 239250 }, { "epoch": 20.72, "learning_rate": 2.322770903203423e-05, "loss": 0.7604, "step": 239260 }, { "epoch": 20.72, "learning_rate": 2.3225540649936393e-05, "loss": 0.6958, "step": 239270 }, { "epoch": 20.72, "learning_rate": 2.3223372267838556e-05, "loss": 0.7175, "step": 239280 }, { "epoch": 20.72, "learning_rate": 2.322120388574072e-05, "loss": 0.7666, "step": 239290 }, { "epoch": 20.72, "learning_rate": 2.3219035503642878e-05, "loss": 0.721, "step": 239300 }, { "epoch": 20.73, "learning_rate": 2.321686712154504e-05, "loss": 0.7148, "step": 239310 }, { "epoch": 20.73, "learning_rate": 2.3214698739447204e-05, "loss": 0.7323, "step": 239320 }, { "epoch": 20.73, "learning_rate": 2.321253035734937e-05, "loss": 0.5652, "step": 239330 }, { "epoch": 20.73, "learning_rate": 2.321036197525153e-05, "loss": 0.9327, "step": 239340 }, { "epoch": 20.73, "learning_rate": 2.3208193593153693e-05, "loss": 0.6249, "step": 239350 }, { "epoch": 20.73, "learning_rate": 2.3206025211055856e-05, "loss": 0.7737, "step": 239360 }, { "epoch": 20.73, "learning_rate": 2.320385682895802e-05, "loss": 0.9823, "step": 239370 }, { "epoch": 20.73, "learning_rate": 2.320168844686018e-05, "loss": 0.639, "step": 239380 }, { "epoch": 20.73, "learning_rate": 2.3199520064762342e-05, "loss": 0.6735, "step": 239390 }, { "epoch": 20.73, "learning_rate": 2.3197351682664508e-05, "loss": 0.7985, "step": 239400 }, { "epoch": 20.73, "learning_rate": 2.3195183300566668e-05, "loss": 0.6717, "step": 239410 }, { "epoch": 20.74, "learning_rate": 2.319301491846883e-05, "loss": 0.7005, "step": 239420 }, { "epoch": 20.74, "learning_rate": 2.3190846536370994e-05, "loss": 0.7666, "step": 239430 }, { "epoch": 20.74, "learning_rate": 2.3188678154273157e-05, "loss": 0.6987, "step": 239440 }, { "epoch": 20.74, "learning_rate": 2.3186509772175316e-05, "loss": 0.5967, "step": 239450 }, { "epoch": 20.74, "learning_rate": 2.318434139007748e-05, "loss": 0.773, "step": 239460 }, { "epoch": 20.74, "learning_rate": 2.3182173007979646e-05, "loss": 0.6736, "step": 239470 }, { "epoch": 20.74, "learning_rate": 2.318000462588181e-05, "loss": 0.8134, "step": 239480 }, { "epoch": 20.74, "learning_rate": 2.317783624378397e-05, "loss": 0.5988, "step": 239490 }, { "epoch": 20.74, "learning_rate": 2.317566786168613e-05, "loss": 0.7088, "step": 239500 }, { "epoch": 20.74, "learning_rate": 2.3173499479588294e-05, "loss": 0.919, "step": 239510 }, { "epoch": 20.74, "learning_rate": 2.3171331097490458e-05, "loss": 0.7778, "step": 239520 }, { "epoch": 20.74, "learning_rate": 2.3169162715392617e-05, "loss": 0.7614, "step": 239530 }, { "epoch": 20.75, "learning_rate": 2.3166994333294784e-05, "loss": 0.8169, "step": 239540 }, { "epoch": 20.75, "learning_rate": 2.3164825951196947e-05, "loss": 0.6802, "step": 239550 }, { "epoch": 20.75, "learning_rate": 2.316265756909911e-05, "loss": 0.8608, "step": 239560 }, { "epoch": 20.75, "learning_rate": 2.316048918700127e-05, "loss": 0.6304, "step": 239570 }, { "epoch": 20.75, "learning_rate": 2.3158320804903432e-05, "loss": 0.8685, "step": 239580 }, { "epoch": 20.75, "learning_rate": 2.3156152422805595e-05, "loss": 0.7101, "step": 239590 }, { "epoch": 20.75, "learning_rate": 2.3153984040707758e-05, "loss": 0.7739, "step": 239600 }, { "epoch": 20.75, "learning_rate": 2.3151815658609918e-05, "loss": 0.7713, "step": 239610 }, { "epoch": 20.75, "learning_rate": 2.3149647276512084e-05, "loss": 0.7558, "step": 239620 }, { "epoch": 20.75, "learning_rate": 2.3147478894414247e-05, "loss": 0.8196, "step": 239630 }, { "epoch": 20.75, "learning_rate": 2.3145310512316407e-05, "loss": 0.7249, "step": 239640 }, { "epoch": 20.76, "learning_rate": 2.314314213021857e-05, "loss": 0.5852, "step": 239650 }, { "epoch": 20.76, "learning_rate": 2.3140973748120733e-05, "loss": 0.6545, "step": 239660 }, { "epoch": 20.76, "learning_rate": 2.3138805366022896e-05, "loss": 0.7282, "step": 239670 }, { "epoch": 20.76, "learning_rate": 2.3136636983925055e-05, "loss": 0.6081, "step": 239680 }, { "epoch": 20.76, "learning_rate": 2.3134468601827222e-05, "loss": 0.6105, "step": 239690 }, { "epoch": 20.76, "learning_rate": 2.3132300219729385e-05, "loss": 0.8273, "step": 239700 }, { "epoch": 20.76, "learning_rate": 2.3130131837631548e-05, "loss": 0.6379, "step": 239710 }, { "epoch": 20.76, "learning_rate": 2.3127963455533707e-05, "loss": 0.5967, "step": 239720 }, { "epoch": 20.76, "learning_rate": 2.312579507343587e-05, "loss": 0.6696, "step": 239730 }, { "epoch": 20.76, "learning_rate": 2.3123626691338033e-05, "loss": 0.7169, "step": 239740 }, { "epoch": 20.76, "learning_rate": 2.31214583092402e-05, "loss": 0.8244, "step": 239750 }, { "epoch": 20.76, "learning_rate": 2.311928992714236e-05, "loss": 0.6943, "step": 239760 }, { "epoch": 20.77, "learning_rate": 2.3117121545044522e-05, "loss": 0.8182, "step": 239770 }, { "epoch": 20.77, "learning_rate": 2.3114953162946685e-05, "loss": 0.7223, "step": 239780 }, { "epoch": 20.77, "learning_rate": 2.311278478084885e-05, "loss": 0.7452, "step": 239790 }, { "epoch": 20.77, "learning_rate": 2.3110616398751008e-05, "loss": 0.749, "step": 239800 }, { "epoch": 20.77, "learning_rate": 2.310844801665317e-05, "loss": 0.7135, "step": 239810 }, { "epoch": 20.77, "learning_rate": 2.3106279634555337e-05, "loss": 0.7713, "step": 239820 }, { "epoch": 20.77, "learning_rate": 2.31041112524575e-05, "loss": 0.7008, "step": 239830 }, { "epoch": 20.77, "learning_rate": 2.310194287035966e-05, "loss": 0.6491, "step": 239840 }, { "epoch": 20.77, "learning_rate": 2.3099774488261823e-05, "loss": 0.7035, "step": 239850 }, { "epoch": 20.77, "learning_rate": 2.3097606106163986e-05, "loss": 0.6548, "step": 239860 }, { "epoch": 20.77, "learning_rate": 2.309543772406615e-05, "loss": 0.6708, "step": 239870 }, { "epoch": 20.78, "learning_rate": 2.309326934196831e-05, "loss": 0.7398, "step": 239880 }, { "epoch": 20.78, "learning_rate": 2.309110095987047e-05, "loss": 0.7588, "step": 239890 }, { "epoch": 20.78, "learning_rate": 2.3088932577772638e-05, "loss": 0.6232, "step": 239900 }, { "epoch": 20.78, "learning_rate": 2.3086764195674798e-05, "loss": 0.7373, "step": 239910 }, { "epoch": 20.78, "learning_rate": 2.308459581357696e-05, "loss": 0.6942, "step": 239920 }, { "epoch": 20.78, "learning_rate": 2.3082427431479124e-05, "loss": 0.7979, "step": 239930 }, { "epoch": 20.78, "learning_rate": 2.3080259049381287e-05, "loss": 0.7527, "step": 239940 }, { "epoch": 20.78, "learning_rate": 2.3078090667283446e-05, "loss": 0.7268, "step": 239950 }, { "epoch": 20.78, "learning_rate": 2.307592228518561e-05, "loss": 0.8456, "step": 239960 }, { "epoch": 20.78, "learning_rate": 2.3073753903087776e-05, "loss": 0.9628, "step": 239970 }, { "epoch": 20.78, "learning_rate": 2.307158552098994e-05, "loss": 0.7372, "step": 239980 }, { "epoch": 20.78, "learning_rate": 2.30694171388921e-05, "loss": 0.5994, "step": 239990 }, { "epoch": 20.79, "learning_rate": 2.306724875679426e-05, "loss": 0.7585, "step": 240000 }, { "epoch": 20.79, "learning_rate": 2.3065080374696424e-05, "loss": 0.7859, "step": 240010 }, { "epoch": 20.79, "learning_rate": 2.3062911992598587e-05, "loss": 0.8863, "step": 240020 }, { "epoch": 20.79, "learning_rate": 2.3060743610500747e-05, "loss": 0.8748, "step": 240030 }, { "epoch": 20.79, "learning_rate": 2.3058575228402913e-05, "loss": 0.8185, "step": 240040 }, { "epoch": 20.79, "learning_rate": 2.3056406846305076e-05, "loss": 0.6065, "step": 240050 }, { "epoch": 20.79, "learning_rate": 2.305423846420724e-05, "loss": 0.8638, "step": 240060 }, { "epoch": 20.79, "learning_rate": 2.30520700821094e-05, "loss": 0.6548, "step": 240070 }, { "epoch": 20.79, "learning_rate": 2.3049901700011562e-05, "loss": 0.7452, "step": 240080 }, { "epoch": 20.79, "learning_rate": 2.3047733317913725e-05, "loss": 0.7037, "step": 240090 }, { "epoch": 20.79, "learning_rate": 2.304556493581589e-05, "loss": 0.6768, "step": 240100 }, { "epoch": 20.8, "learning_rate": 2.3043396553718048e-05, "loss": 0.6494, "step": 240110 }, { "epoch": 20.8, "learning_rate": 2.3041228171620214e-05, "loss": 0.7515, "step": 240120 }, { "epoch": 20.8, "learning_rate": 2.3039059789522377e-05, "loss": 0.8456, "step": 240130 }, { "epoch": 20.8, "learning_rate": 2.303689140742454e-05, "loss": 0.7253, "step": 240140 }, { "epoch": 20.8, "learning_rate": 2.30347230253267e-05, "loss": 0.7284, "step": 240150 }, { "epoch": 20.8, "learning_rate": 2.3032554643228863e-05, "loss": 0.7552, "step": 240160 }, { "epoch": 20.8, "learning_rate": 2.3030386261131026e-05, "loss": 0.6973, "step": 240170 }, { "epoch": 20.8, "learning_rate": 2.3028217879033185e-05, "loss": 0.757, "step": 240180 }, { "epoch": 20.8, "learning_rate": 2.302604949693535e-05, "loss": 0.701, "step": 240190 }, { "epoch": 20.8, "learning_rate": 2.3023881114837515e-05, "loss": 0.7234, "step": 240200 }, { "epoch": 20.8, "learning_rate": 2.3021712732739678e-05, "loss": 0.6779, "step": 240210 }, { "epoch": 20.8, "learning_rate": 2.3019544350641837e-05, "loss": 0.676, "step": 240220 }, { "epoch": 20.81, "learning_rate": 2.3017375968544e-05, "loss": 0.8402, "step": 240230 }, { "epoch": 20.81, "learning_rate": 2.3015207586446163e-05, "loss": 0.7162, "step": 240240 }, { "epoch": 20.81, "learning_rate": 2.301303920434833e-05, "loss": 0.6488, "step": 240250 }, { "epoch": 20.81, "learning_rate": 2.301087082225049e-05, "loss": 0.6711, "step": 240260 }, { "epoch": 20.81, "learning_rate": 2.3008702440152652e-05, "loss": 0.7225, "step": 240270 }, { "epoch": 20.81, "learning_rate": 2.3006534058054815e-05, "loss": 0.9919, "step": 240280 }, { "epoch": 20.81, "learning_rate": 2.3004365675956978e-05, "loss": 0.5604, "step": 240290 }, { "epoch": 20.81, "learning_rate": 2.3002197293859138e-05, "loss": 0.754, "step": 240300 }, { "epoch": 20.81, "learning_rate": 2.30000289117613e-05, "loss": 0.8699, "step": 240310 }, { "epoch": 20.81, "learning_rate": 2.2997860529663467e-05, "loss": 0.734, "step": 240320 }, { "epoch": 20.81, "learning_rate": 2.299569214756563e-05, "loss": 0.6883, "step": 240330 }, { "epoch": 20.81, "learning_rate": 2.299352376546779e-05, "loss": 0.8291, "step": 240340 }, { "epoch": 20.82, "learning_rate": 2.2991355383369953e-05, "loss": 0.7636, "step": 240350 }, { "epoch": 20.82, "learning_rate": 2.2989187001272116e-05, "loss": 0.751, "step": 240360 }, { "epoch": 20.82, "learning_rate": 2.298701861917428e-05, "loss": 0.7807, "step": 240370 }, { "epoch": 20.82, "learning_rate": 2.298485023707644e-05, "loss": 0.6602, "step": 240380 }, { "epoch": 20.82, "learning_rate": 2.29826818549786e-05, "loss": 0.8061, "step": 240390 }, { "epoch": 20.82, "learning_rate": 2.2980513472880768e-05, "loss": 0.8244, "step": 240400 }, { "epoch": 20.82, "learning_rate": 2.2978345090782928e-05, "loss": 0.8339, "step": 240410 }, { "epoch": 20.82, "learning_rate": 2.297617670868509e-05, "loss": 0.8865, "step": 240420 }, { "epoch": 20.82, "learning_rate": 2.2974008326587254e-05, "loss": 0.8647, "step": 240430 }, { "epoch": 20.82, "learning_rate": 2.2971839944489417e-05, "loss": 0.7452, "step": 240440 }, { "epoch": 20.82, "learning_rate": 2.2969671562391576e-05, "loss": 0.7934, "step": 240450 }, { "epoch": 20.83, "learning_rate": 2.296750318029374e-05, "loss": 0.6485, "step": 240460 }, { "epoch": 20.83, "learning_rate": 2.2965334798195906e-05, "loss": 0.8467, "step": 240470 }, { "epoch": 20.83, "learning_rate": 2.296316641609807e-05, "loss": 0.7218, "step": 240480 }, { "epoch": 20.83, "learning_rate": 2.2960998034000228e-05, "loss": 0.8599, "step": 240490 }, { "epoch": 20.83, "learning_rate": 2.295882965190239e-05, "loss": 0.6563, "step": 240500 }, { "epoch": 20.83, "learning_rate": 2.2956661269804554e-05, "loss": 0.6415, "step": 240510 }, { "epoch": 20.83, "learning_rate": 2.2954492887706717e-05, "loss": 0.6709, "step": 240520 }, { "epoch": 20.83, "learning_rate": 2.2952324505608877e-05, "loss": 0.6644, "step": 240530 }, { "epoch": 20.83, "learning_rate": 2.2950156123511043e-05, "loss": 0.7187, "step": 240540 }, { "epoch": 20.83, "learning_rate": 2.2947987741413206e-05, "loss": 0.941, "step": 240550 }, { "epoch": 20.83, "learning_rate": 2.294581935931537e-05, "loss": 0.7776, "step": 240560 }, { "epoch": 20.83, "learning_rate": 2.294365097721753e-05, "loss": 0.7264, "step": 240570 }, { "epoch": 20.84, "learning_rate": 2.2941482595119692e-05, "loss": 0.5566, "step": 240580 }, { "epoch": 20.84, "learning_rate": 2.2939314213021855e-05, "loss": 0.7718, "step": 240590 }, { "epoch": 20.84, "learning_rate": 2.293714583092402e-05, "loss": 0.88, "step": 240600 }, { "epoch": 20.84, "learning_rate": 2.293497744882618e-05, "loss": 0.7947, "step": 240610 }, { "epoch": 20.84, "learning_rate": 2.2932809066728344e-05, "loss": 0.7389, "step": 240620 }, { "epoch": 20.84, "learning_rate": 2.2930640684630507e-05, "loss": 0.7206, "step": 240630 }, { "epoch": 20.84, "learning_rate": 2.292847230253267e-05, "loss": 0.7183, "step": 240640 }, { "epoch": 20.84, "learning_rate": 2.292630392043483e-05, "loss": 0.7401, "step": 240650 }, { "epoch": 20.84, "learning_rate": 2.2924135538336992e-05, "loss": 0.8208, "step": 240660 }, { "epoch": 20.84, "learning_rate": 2.2921967156239155e-05, "loss": 0.7577, "step": 240670 }, { "epoch": 20.84, "learning_rate": 2.2919798774141315e-05, "loss": 0.6756, "step": 240680 }, { "epoch": 20.85, "learning_rate": 2.291763039204348e-05, "loss": 0.6224, "step": 240690 }, { "epoch": 20.85, "learning_rate": 2.2915462009945644e-05, "loss": 0.7882, "step": 240700 }, { "epoch": 20.85, "learning_rate": 2.2913293627847807e-05, "loss": 0.7986, "step": 240710 }, { "epoch": 20.85, "learning_rate": 2.2911125245749967e-05, "loss": 0.6824, "step": 240720 }, { "epoch": 20.85, "learning_rate": 2.290895686365213e-05, "loss": 0.7746, "step": 240730 }, { "epoch": 20.85, "learning_rate": 2.2906788481554293e-05, "loss": 0.6888, "step": 240740 }, { "epoch": 20.85, "learning_rate": 2.290462009945646e-05, "loss": 0.795, "step": 240750 }, { "epoch": 20.85, "learning_rate": 2.290245171735862e-05, "loss": 0.7293, "step": 240760 }, { "epoch": 20.85, "learning_rate": 2.2900283335260782e-05, "loss": 0.8192, "step": 240770 }, { "epoch": 20.85, "learning_rate": 2.2898114953162945e-05, "loss": 0.8567, "step": 240780 }, { "epoch": 20.85, "learning_rate": 2.2895946571065108e-05, "loss": 0.6203, "step": 240790 }, { "epoch": 20.85, "learning_rate": 2.2893778188967268e-05, "loss": 0.8084, "step": 240800 }, { "epoch": 20.86, "learning_rate": 2.289160980686943e-05, "loss": 0.7252, "step": 240810 }, { "epoch": 20.86, "learning_rate": 2.2889441424771597e-05, "loss": 0.7174, "step": 240820 }, { "epoch": 20.86, "learning_rate": 2.288727304267376e-05, "loss": 0.7994, "step": 240830 }, { "epoch": 20.86, "learning_rate": 2.288510466057592e-05, "loss": 0.8512, "step": 240840 }, { "epoch": 20.86, "learning_rate": 2.2882936278478083e-05, "loss": 0.7299, "step": 240850 }, { "epoch": 20.86, "learning_rate": 2.2880767896380246e-05, "loss": 0.7872, "step": 240860 }, { "epoch": 20.86, "learning_rate": 2.287859951428241e-05, "loss": 0.7547, "step": 240870 }, { "epoch": 20.86, "learning_rate": 2.287643113218457e-05, "loss": 0.6529, "step": 240880 }, { "epoch": 20.86, "learning_rate": 2.2874262750086735e-05, "loss": 0.7821, "step": 240890 }, { "epoch": 20.86, "learning_rate": 2.2872094367988898e-05, "loss": 0.6547, "step": 240900 }, { "epoch": 20.86, "learning_rate": 2.286992598589106e-05, "loss": 0.7228, "step": 240910 }, { "epoch": 20.87, "learning_rate": 2.286775760379322e-05, "loss": 0.785, "step": 240920 }, { "epoch": 20.87, "learning_rate": 2.2865589221695383e-05, "loss": 0.7293, "step": 240930 }, { "epoch": 20.87, "learning_rate": 2.2863420839597546e-05, "loss": 0.7147, "step": 240940 }, { "epoch": 20.87, "learning_rate": 2.2861252457499706e-05, "loss": 0.7654, "step": 240950 }, { "epoch": 20.87, "learning_rate": 2.285908407540187e-05, "loss": 0.6759, "step": 240960 }, { "epoch": 20.87, "learning_rate": 2.2856915693304035e-05, "loss": 0.8726, "step": 240970 }, { "epoch": 20.87, "learning_rate": 2.28547473112062e-05, "loss": 0.7629, "step": 240980 }, { "epoch": 20.87, "learning_rate": 2.2852578929108358e-05, "loss": 0.7151, "step": 240990 }, { "epoch": 20.87, "learning_rate": 2.285041054701052e-05, "loss": 0.65, "step": 241000 }, { "epoch": 20.87, "learning_rate": 2.2848242164912684e-05, "loss": 0.753, "step": 241010 }, { "epoch": 20.87, "learning_rate": 2.2846073782814847e-05, "loss": 0.6893, "step": 241020 }, { "epoch": 20.87, "learning_rate": 2.2843905400717007e-05, "loss": 0.7122, "step": 241030 }, { "epoch": 20.88, "learning_rate": 2.2841737018619173e-05, "loss": 0.7364, "step": 241040 }, { "epoch": 20.88, "learning_rate": 2.2839568636521336e-05, "loss": 0.7423, "step": 241050 }, { "epoch": 20.88, "learning_rate": 2.28374002544235e-05, "loss": 0.6677, "step": 241060 }, { "epoch": 20.88, "learning_rate": 2.283523187232566e-05, "loss": 0.6711, "step": 241070 }, { "epoch": 20.88, "learning_rate": 2.283306349022782e-05, "loss": 0.6263, "step": 241080 }, { "epoch": 20.88, "learning_rate": 2.2830895108129985e-05, "loss": 0.7397, "step": 241090 }, { "epoch": 20.88, "learning_rate": 2.282872672603215e-05, "loss": 0.9287, "step": 241100 }, { "epoch": 20.88, "learning_rate": 2.282655834393431e-05, "loss": 0.6215, "step": 241110 }, { "epoch": 20.88, "learning_rate": 2.2824389961836474e-05, "loss": 0.6776, "step": 241120 }, { "epoch": 20.88, "learning_rate": 2.2822221579738637e-05, "loss": 0.7304, "step": 241130 }, { "epoch": 20.88, "learning_rate": 2.28200531976408e-05, "loss": 0.698, "step": 241140 }, { "epoch": 20.89, "learning_rate": 2.281788481554296e-05, "loss": 0.7942, "step": 241150 }, { "epoch": 20.89, "learning_rate": 2.2815716433445122e-05, "loss": 0.7738, "step": 241160 }, { "epoch": 20.89, "learning_rate": 2.281354805134729e-05, "loss": 0.7809, "step": 241170 }, { "epoch": 20.89, "learning_rate": 2.2811379669249445e-05, "loss": 0.6166, "step": 241180 }, { "epoch": 20.89, "learning_rate": 2.280921128715161e-05, "loss": 0.8793, "step": 241190 }, { "epoch": 20.89, "learning_rate": 2.2807042905053774e-05, "loss": 0.6576, "step": 241200 }, { "epoch": 20.89, "learning_rate": 2.2804874522955937e-05, "loss": 0.504, "step": 241210 }, { "epoch": 20.89, "learning_rate": 2.2802706140858097e-05, "loss": 0.8827, "step": 241220 }, { "epoch": 20.89, "learning_rate": 2.280053775876026e-05, "loss": 0.6356, "step": 241230 }, { "epoch": 20.89, "learning_rate": 2.2798369376662423e-05, "loss": 0.8547, "step": 241240 }, { "epoch": 20.89, "learning_rate": 2.279620099456459e-05, "loss": 0.6419, "step": 241250 }, { "epoch": 20.89, "learning_rate": 2.279403261246675e-05, "loss": 0.7705, "step": 241260 }, { "epoch": 20.9, "learning_rate": 2.2791864230368912e-05, "loss": 0.7344, "step": 241270 }, { "epoch": 20.9, "learning_rate": 2.2789695848271075e-05, "loss": 0.7639, "step": 241280 }, { "epoch": 20.9, "learning_rate": 2.2787527466173238e-05, "loss": 0.848, "step": 241290 }, { "epoch": 20.9, "learning_rate": 2.2785359084075398e-05, "loss": 0.7386, "step": 241300 }, { "epoch": 20.9, "learning_rate": 2.278319070197756e-05, "loss": 0.7208, "step": 241310 }, { "epoch": 20.9, "learning_rate": 2.2781022319879727e-05, "loss": 0.7856, "step": 241320 }, { "epoch": 20.9, "learning_rate": 2.277885393778189e-05, "loss": 0.7127, "step": 241330 }, { "epoch": 20.9, "learning_rate": 2.277668555568405e-05, "loss": 0.6778, "step": 241340 }, { "epoch": 20.9, "learning_rate": 2.2774517173586213e-05, "loss": 0.7602, "step": 241350 }, { "epoch": 20.9, "learning_rate": 2.2772348791488376e-05, "loss": 0.7447, "step": 241360 }, { "epoch": 20.9, "learning_rate": 2.277018040939054e-05, "loss": 0.6949, "step": 241370 }, { "epoch": 20.91, "learning_rate": 2.2768012027292698e-05, "loss": 0.7515, "step": 241380 }, { "epoch": 20.91, "learning_rate": 2.2765843645194865e-05, "loss": 0.6066, "step": 241390 }, { "epoch": 20.91, "learning_rate": 2.2763675263097028e-05, "loss": 0.7465, "step": 241400 }, { "epoch": 20.91, "learning_rate": 2.276150688099919e-05, "loss": 0.8008, "step": 241410 }, { "epoch": 20.91, "learning_rate": 2.275933849890135e-05, "loss": 0.8237, "step": 241420 }, { "epoch": 20.91, "learning_rate": 2.2757170116803513e-05, "loss": 0.5797, "step": 241430 }, { "epoch": 20.91, "learning_rate": 2.2755001734705676e-05, "loss": 0.6442, "step": 241440 }, { "epoch": 20.91, "learning_rate": 2.2752833352607836e-05, "loss": 0.8955, "step": 241450 }, { "epoch": 20.91, "learning_rate": 2.275066497051e-05, "loss": 0.7966, "step": 241460 }, { "epoch": 20.91, "learning_rate": 2.2748496588412165e-05, "loss": 0.5778, "step": 241470 }, { "epoch": 20.91, "learning_rate": 2.2746328206314328e-05, "loss": 0.6709, "step": 241480 }, { "epoch": 20.91, "learning_rate": 2.2744159824216488e-05, "loss": 0.9511, "step": 241490 }, { "epoch": 20.92, "learning_rate": 2.274199144211865e-05, "loss": 0.7411, "step": 241500 }, { "epoch": 20.92, "learning_rate": 2.2739823060020814e-05, "loss": 0.7144, "step": 241510 }, { "epoch": 20.92, "learning_rate": 2.2737654677922977e-05, "loss": 0.7271, "step": 241520 }, { "epoch": 20.92, "learning_rate": 2.2735486295825137e-05, "loss": 0.6526, "step": 241530 }, { "epoch": 20.92, "learning_rate": 2.2733317913727303e-05, "loss": 0.7738, "step": 241540 }, { "epoch": 20.92, "learning_rate": 2.2731149531629466e-05, "loss": 0.7496, "step": 241550 }, { "epoch": 20.92, "learning_rate": 2.272898114953163e-05, "loss": 0.6807, "step": 241560 }, { "epoch": 20.92, "learning_rate": 2.272681276743379e-05, "loss": 0.6011, "step": 241570 }, { "epoch": 20.92, "learning_rate": 2.272464438533595e-05, "loss": 0.6157, "step": 241580 }, { "epoch": 20.92, "learning_rate": 2.2722476003238115e-05, "loss": 0.7596, "step": 241590 }, { "epoch": 20.92, "learning_rate": 2.272030762114028e-05, "loss": 0.8252, "step": 241600 }, { "epoch": 20.92, "learning_rate": 2.271813923904244e-05, "loss": 0.7146, "step": 241610 }, { "epoch": 20.93, "learning_rate": 2.2715970856944604e-05, "loss": 0.868, "step": 241620 }, { "epoch": 20.93, "learning_rate": 2.2713802474846767e-05, "loss": 0.6599, "step": 241630 }, { "epoch": 20.93, "learning_rate": 2.271163409274893e-05, "loss": 0.7471, "step": 241640 }, { "epoch": 20.93, "learning_rate": 2.270946571065109e-05, "loss": 0.7642, "step": 241650 }, { "epoch": 20.93, "learning_rate": 2.2707297328553252e-05, "loss": 0.7612, "step": 241660 }, { "epoch": 20.93, "learning_rate": 2.270512894645542e-05, "loss": 0.7415, "step": 241670 }, { "epoch": 20.93, "learning_rate": 2.270296056435758e-05, "loss": 0.8049, "step": 241680 }, { "epoch": 20.93, "learning_rate": 2.270079218225974e-05, "loss": 0.6694, "step": 241690 }, { "epoch": 20.93, "learning_rate": 2.2698623800161904e-05, "loss": 0.6302, "step": 241700 }, { "epoch": 20.93, "learning_rate": 2.2696455418064067e-05, "loss": 0.7535, "step": 241710 }, { "epoch": 20.93, "learning_rate": 2.2694287035966227e-05, "loss": 0.7265, "step": 241720 }, { "epoch": 20.94, "learning_rate": 2.269211865386839e-05, "loss": 0.8184, "step": 241730 }, { "epoch": 20.94, "learning_rate": 2.2689950271770553e-05, "loss": 0.7562, "step": 241740 }, { "epoch": 20.94, "learning_rate": 2.268778188967272e-05, "loss": 0.7739, "step": 241750 }, { "epoch": 20.94, "learning_rate": 2.268561350757488e-05, "loss": 0.8639, "step": 241760 }, { "epoch": 20.94, "learning_rate": 2.2683445125477042e-05, "loss": 0.6302, "step": 241770 }, { "epoch": 20.94, "learning_rate": 2.2681276743379205e-05, "loss": 0.765, "step": 241780 }, { "epoch": 20.94, "learning_rate": 2.2679108361281368e-05, "loss": 0.5752, "step": 241790 }, { "epoch": 20.94, "learning_rate": 2.2676939979183527e-05, "loss": 0.8392, "step": 241800 }, { "epoch": 20.94, "learning_rate": 2.267477159708569e-05, "loss": 0.7437, "step": 241810 }, { "epoch": 20.94, "learning_rate": 2.2672603214987857e-05, "loss": 0.7105, "step": 241820 }, { "epoch": 20.94, "learning_rate": 2.267043483289002e-05, "loss": 0.7573, "step": 241830 }, { "epoch": 20.94, "learning_rate": 2.266826645079218e-05, "loss": 0.7856, "step": 241840 }, { "epoch": 20.95, "learning_rate": 2.2666098068694342e-05, "loss": 0.7578, "step": 241850 }, { "epoch": 20.95, "learning_rate": 2.2663929686596505e-05, "loss": 0.5904, "step": 241860 }, { "epoch": 20.95, "learning_rate": 2.266176130449867e-05, "loss": 0.5865, "step": 241870 }, { "epoch": 20.95, "learning_rate": 2.2659592922400828e-05, "loss": 0.7628, "step": 241880 }, { "epoch": 20.95, "learning_rate": 2.2657424540302994e-05, "loss": 0.906, "step": 241890 }, { "epoch": 20.95, "learning_rate": 2.2655256158205157e-05, "loss": 0.636, "step": 241900 }, { "epoch": 20.95, "learning_rate": 2.265308777610732e-05, "loss": 0.7943, "step": 241910 }, { "epoch": 20.95, "learning_rate": 2.265091939400948e-05, "loss": 0.648, "step": 241920 }, { "epoch": 20.95, "learning_rate": 2.2648751011911643e-05, "loss": 0.8438, "step": 241930 }, { "epoch": 20.95, "learning_rate": 2.2646582629813806e-05, "loss": 0.6318, "step": 241940 }, { "epoch": 20.95, "learning_rate": 2.2644414247715966e-05, "loss": 0.7859, "step": 241950 }, { "epoch": 20.96, "learning_rate": 2.2642245865618132e-05, "loss": 0.8746, "step": 241960 }, { "epoch": 20.96, "learning_rate": 2.2640077483520295e-05, "loss": 0.7068, "step": 241970 }, { "epoch": 20.96, "learning_rate": 2.2637909101422458e-05, "loss": 0.7557, "step": 241980 }, { "epoch": 20.96, "learning_rate": 2.2635740719324618e-05, "loss": 0.7609, "step": 241990 }, { "epoch": 20.96, "learning_rate": 2.263357233722678e-05, "loss": 0.9089, "step": 242000 }, { "epoch": 20.96, "learning_rate": 2.2631403955128944e-05, "loss": 0.6586, "step": 242010 }, { "epoch": 20.96, "learning_rate": 2.2629235573031107e-05, "loss": 0.7234, "step": 242020 }, { "epoch": 20.96, "learning_rate": 2.2627067190933266e-05, "loss": 0.571, "step": 242030 }, { "epoch": 20.96, "learning_rate": 2.2624898808835433e-05, "loss": 0.8614, "step": 242040 }, { "epoch": 20.96, "learning_rate": 2.2622730426737596e-05, "loss": 0.8008, "step": 242050 }, { "epoch": 20.96, "learning_rate": 2.262056204463976e-05, "loss": 0.7988, "step": 242060 }, { "epoch": 20.96, "learning_rate": 2.261839366254192e-05, "loss": 0.6829, "step": 242070 }, { "epoch": 20.97, "learning_rate": 2.261622528044408e-05, "loss": 0.7366, "step": 242080 }, { "epoch": 20.97, "learning_rate": 2.2614056898346244e-05, "loss": 0.6923, "step": 242090 }, { "epoch": 20.97, "learning_rate": 2.261188851624841e-05, "loss": 0.7286, "step": 242100 }, { "epoch": 20.97, "learning_rate": 2.260972013415057e-05, "loss": 0.7378, "step": 242110 }, { "epoch": 20.97, "learning_rate": 2.2607551752052733e-05, "loss": 0.8169, "step": 242120 }, { "epoch": 20.97, "learning_rate": 2.2605383369954896e-05, "loss": 0.7431, "step": 242130 }, { "epoch": 20.97, "learning_rate": 2.260321498785706e-05, "loss": 0.7882, "step": 242140 }, { "epoch": 20.97, "learning_rate": 2.260104660575922e-05, "loss": 0.8647, "step": 242150 }, { "epoch": 20.97, "learning_rate": 2.2598878223661382e-05, "loss": 0.8579, "step": 242160 }, { "epoch": 20.97, "learning_rate": 2.259670984156355e-05, "loss": 0.6915, "step": 242170 }, { "epoch": 20.97, "learning_rate": 2.259454145946571e-05, "loss": 0.703, "step": 242180 }, { "epoch": 20.98, "learning_rate": 2.259237307736787e-05, "loss": 0.5799, "step": 242190 }, { "epoch": 20.98, "learning_rate": 2.2590204695270034e-05, "loss": 0.6001, "step": 242200 }, { "epoch": 20.98, "learning_rate": 2.2588036313172197e-05, "loss": 0.7297, "step": 242210 }, { "epoch": 20.98, "learning_rate": 2.2585867931074357e-05, "loss": 0.5409, "step": 242220 }, { "epoch": 20.98, "learning_rate": 2.258369954897652e-05, "loss": 0.7966, "step": 242230 }, { "epoch": 20.98, "learning_rate": 2.2581531166878686e-05, "loss": 0.6981, "step": 242240 }, { "epoch": 20.98, "learning_rate": 2.257936278478085e-05, "loss": 0.7506, "step": 242250 }, { "epoch": 20.98, "learning_rate": 2.257719440268301e-05, "loss": 0.7036, "step": 242260 }, { "epoch": 20.98, "learning_rate": 2.257502602058517e-05, "loss": 0.8766, "step": 242270 }, { "epoch": 20.98, "learning_rate": 2.2572857638487335e-05, "loss": 0.6469, "step": 242280 }, { "epoch": 20.98, "learning_rate": 2.2570689256389498e-05, "loss": 0.7514, "step": 242290 }, { "epoch": 20.98, "learning_rate": 2.2568520874291657e-05, "loss": 0.717, "step": 242300 }, { "epoch": 20.99, "learning_rate": 2.256635249219382e-05, "loss": 0.9184, "step": 242310 }, { "epoch": 20.99, "learning_rate": 2.2564184110095987e-05, "loss": 0.6552, "step": 242320 }, { "epoch": 20.99, "learning_rate": 2.256201572799815e-05, "loss": 0.7791, "step": 242330 }, { "epoch": 20.99, "learning_rate": 2.255984734590031e-05, "loss": 0.7818, "step": 242340 }, { "epoch": 20.99, "learning_rate": 2.2557678963802472e-05, "loss": 0.8125, "step": 242350 }, { "epoch": 20.99, "learning_rate": 2.2555510581704635e-05, "loss": 0.7711, "step": 242360 }, { "epoch": 20.99, "learning_rate": 2.25533421996068e-05, "loss": 0.7375, "step": 242370 }, { "epoch": 20.99, "learning_rate": 2.2551173817508958e-05, "loss": 0.772, "step": 242380 }, { "epoch": 20.99, "learning_rate": 2.2549005435411124e-05, "loss": 0.8032, "step": 242390 }, { "epoch": 20.99, "learning_rate": 2.2546837053313287e-05, "loss": 0.7382, "step": 242400 }, { "epoch": 20.99, "learning_rate": 2.254466867121545e-05, "loss": 0.7293, "step": 242410 }, { "epoch": 21.0, "learning_rate": 2.254250028911761e-05, "loss": 0.6852, "step": 242420 }, { "epoch": 21.0, "learning_rate": 2.2540331907019773e-05, "loss": 0.5987, "step": 242430 }, { "epoch": 21.0, "learning_rate": 2.2538163524921936e-05, "loss": 0.5812, "step": 242440 }, { "epoch": 21.0, "learning_rate": 2.2535995142824102e-05, "loss": 0.6934, "step": 242450 }, { "epoch": 21.0, "learning_rate": 2.2533826760726262e-05, "loss": 0.5976, "step": 242460 }, { "epoch": 21.0, "learning_rate": 2.2531658378628425e-05, "loss": 0.7535, "step": 242470 }, { "epoch": 21.0, "eval_cer": 0.5388753516205902, "eval_em": 0.05518308406395049, "eval_f1": 0.05518308406395049, "eval_loss": 1.5174046754837036, "eval_runtime": 412.4055, "eval_samples_per_second": 4.702, "eval_steps_per_second": 0.589, "eval_wer": 0.9448169159360496, "step": 242476 }, { "epoch": 21.0, "learning_rate": 2.2529489996530588e-05, "loss": 0.813, "step": 242480 }, { "epoch": 21.0, "learning_rate": 2.2527321614432748e-05, "loss": 0.6706, "step": 242490 }, { "epoch": 21.0, "learning_rate": 2.252515323233491e-05, "loss": 0.7337, "step": 242500 }, { "epoch": 21.0, "learning_rate": 2.2522984850237074e-05, "loss": 0.6653, "step": 242510 }, { "epoch": 21.0, "learning_rate": 2.252081646813924e-05, "loss": 0.7015, "step": 242520 }, { "epoch": 21.0, "learning_rate": 2.2518648086041396e-05, "loss": 0.6121, "step": 242530 }, { "epoch": 21.01, "learning_rate": 2.2516479703943563e-05, "loss": 0.6919, "step": 242540 }, { "epoch": 21.01, "learning_rate": 2.2514311321845726e-05, "loss": 0.5923, "step": 242550 }, { "epoch": 21.01, "learning_rate": 2.251214293974789e-05, "loss": 0.6548, "step": 242560 }, { "epoch": 21.01, "learning_rate": 2.2509974557650048e-05, "loss": 0.7495, "step": 242570 }, { "epoch": 21.01, "learning_rate": 2.250780617555221e-05, "loss": 0.7307, "step": 242580 }, { "epoch": 21.01, "learning_rate": 2.2505637793454374e-05, "loss": 0.7203, "step": 242590 }, { "epoch": 21.01, "learning_rate": 2.250346941135654e-05, "loss": 0.6415, "step": 242600 }, { "epoch": 21.01, "learning_rate": 2.25013010292587e-05, "loss": 0.6487, "step": 242610 }, { "epoch": 21.01, "learning_rate": 2.2499132647160863e-05, "loss": 0.7048, "step": 242620 }, { "epoch": 21.01, "learning_rate": 2.2496964265063026e-05, "loss": 0.7384, "step": 242630 }, { "epoch": 21.01, "learning_rate": 2.249479588296519e-05, "loss": 0.7148, "step": 242640 }, { "epoch": 21.02, "learning_rate": 2.249262750086735e-05, "loss": 0.7805, "step": 242650 }, { "epoch": 21.02, "learning_rate": 2.2490459118769512e-05, "loss": 0.6591, "step": 242660 }, { "epoch": 21.02, "learning_rate": 2.2488290736671678e-05, "loss": 0.6554, "step": 242670 }, { "epoch": 21.02, "learning_rate": 2.248612235457384e-05, "loss": 0.6555, "step": 242680 }, { "epoch": 21.02, "learning_rate": 2.2483953972476e-05, "loss": 0.6701, "step": 242690 }, { "epoch": 21.02, "learning_rate": 2.2481785590378164e-05, "loss": 0.7938, "step": 242700 }, { "epoch": 21.02, "learning_rate": 2.2479617208280327e-05, "loss": 0.6464, "step": 242710 }, { "epoch": 21.02, "learning_rate": 2.2477448826182486e-05, "loss": 0.6766, "step": 242720 }, { "epoch": 21.02, "learning_rate": 2.247528044408465e-05, "loss": 0.8242, "step": 242730 }, { "epoch": 21.02, "learning_rate": 2.2473112061986816e-05, "loss": 0.8078, "step": 242740 }, { "epoch": 21.02, "learning_rate": 2.247094367988898e-05, "loss": 0.6999, "step": 242750 }, { "epoch": 21.02, "learning_rate": 2.246877529779114e-05, "loss": 0.6845, "step": 242760 }, { "epoch": 21.03, "learning_rate": 2.24666069156933e-05, "loss": 0.7056, "step": 242770 }, { "epoch": 21.03, "learning_rate": 2.2464438533595465e-05, "loss": 0.5861, "step": 242780 }, { "epoch": 21.03, "learning_rate": 2.2462270151497628e-05, "loss": 0.6873, "step": 242790 }, { "epoch": 21.03, "learning_rate": 2.2460101769399787e-05, "loss": 0.7579, "step": 242800 }, { "epoch": 21.03, "learning_rate": 2.245793338730195e-05, "loss": 0.6725, "step": 242810 }, { "epoch": 21.03, "learning_rate": 2.2455765005204117e-05, "loss": 0.7303, "step": 242820 }, { "epoch": 21.03, "learning_rate": 2.245359662310628e-05, "loss": 0.7856, "step": 242830 }, { "epoch": 21.03, "learning_rate": 2.245142824100844e-05, "loss": 0.6795, "step": 242840 }, { "epoch": 21.03, "learning_rate": 2.2449259858910602e-05, "loss": 0.7895, "step": 242850 }, { "epoch": 21.03, "learning_rate": 2.2447091476812765e-05, "loss": 0.7681, "step": 242860 }, { "epoch": 21.03, "learning_rate": 2.2444923094714928e-05, "loss": 0.6669, "step": 242870 }, { "epoch": 21.03, "learning_rate": 2.2442754712617088e-05, "loss": 0.7017, "step": 242880 }, { "epoch": 21.04, "learning_rate": 2.2440586330519254e-05, "loss": 0.5779, "step": 242890 }, { "epoch": 21.04, "learning_rate": 2.2438417948421417e-05, "loss": 0.7103, "step": 242900 }, { "epoch": 21.04, "learning_rate": 2.243624956632358e-05, "loss": 0.6362, "step": 242910 }, { "epoch": 21.04, "learning_rate": 2.243408118422574e-05, "loss": 0.6836, "step": 242920 }, { "epoch": 21.04, "learning_rate": 2.2431912802127903e-05, "loss": 0.8169, "step": 242930 }, { "epoch": 21.04, "learning_rate": 2.2429744420030066e-05, "loss": 0.7796, "step": 242940 }, { "epoch": 21.04, "learning_rate": 2.2427576037932232e-05, "loss": 0.8014, "step": 242950 }, { "epoch": 21.04, "learning_rate": 2.2425407655834392e-05, "loss": 0.8169, "step": 242960 }, { "epoch": 21.04, "learning_rate": 2.2423239273736555e-05, "loss": 0.7018, "step": 242970 }, { "epoch": 21.04, "learning_rate": 2.2421070891638718e-05, "loss": 0.6001, "step": 242980 }, { "epoch": 21.04, "learning_rate": 2.2418902509540877e-05, "loss": 0.8118, "step": 242990 }, { "epoch": 21.05, "learning_rate": 2.241673412744304e-05, "loss": 0.6163, "step": 243000 }, { "epoch": 21.05, "learning_rate": 2.2414565745345203e-05, "loss": 0.7662, "step": 243010 }, { "epoch": 21.05, "learning_rate": 2.241239736324737e-05, "loss": 0.725, "step": 243020 }, { "epoch": 21.05, "learning_rate": 2.241022898114953e-05, "loss": 0.595, "step": 243030 }, { "epoch": 21.05, "learning_rate": 2.2408060599051692e-05, "loss": 0.6884, "step": 243040 }, { "epoch": 21.05, "learning_rate": 2.2405892216953855e-05, "loss": 0.6504, "step": 243050 }, { "epoch": 21.05, "learning_rate": 2.240372383485602e-05, "loss": 0.7303, "step": 243060 }, { "epoch": 21.05, "learning_rate": 2.2401555452758178e-05, "loss": 0.6668, "step": 243070 }, { "epoch": 21.05, "learning_rate": 2.239938707066034e-05, "loss": 0.7661, "step": 243080 }, { "epoch": 21.05, "learning_rate": 2.2397218688562504e-05, "loss": 0.7462, "step": 243090 }, { "epoch": 21.05, "learning_rate": 2.239505030646467e-05, "loss": 0.6514, "step": 243100 }, { "epoch": 21.05, "learning_rate": 2.239288192436683e-05, "loss": 0.6702, "step": 243110 }, { "epoch": 21.06, "learning_rate": 2.2390713542268993e-05, "loss": 0.7774, "step": 243120 }, { "epoch": 21.06, "learning_rate": 2.2388545160171156e-05, "loss": 0.6746, "step": 243130 }, { "epoch": 21.06, "learning_rate": 2.238637677807332e-05, "loss": 0.8173, "step": 243140 }, { "epoch": 21.06, "learning_rate": 2.238420839597548e-05, "loss": 0.7512, "step": 243150 }, { "epoch": 21.06, "learning_rate": 2.2382040013877642e-05, "loss": 0.6341, "step": 243160 }, { "epoch": 21.06, "learning_rate": 2.2379871631779808e-05, "loss": 0.7317, "step": 243170 }, { "epoch": 21.06, "learning_rate": 2.237770324968197e-05, "loss": 0.8401, "step": 243180 }, { "epoch": 21.06, "learning_rate": 2.237553486758413e-05, "loss": 0.8023, "step": 243190 }, { "epoch": 21.06, "learning_rate": 2.2373366485486294e-05, "loss": 0.6755, "step": 243200 }, { "epoch": 21.06, "learning_rate": 2.2371198103388457e-05, "loss": 0.6053, "step": 243210 }, { "epoch": 21.06, "learning_rate": 2.236902972129062e-05, "loss": 0.7598, "step": 243220 }, { "epoch": 21.07, "learning_rate": 2.236686133919278e-05, "loss": 0.7053, "step": 243230 }, { "epoch": 21.07, "learning_rate": 2.2364692957094946e-05, "loss": 0.6099, "step": 243240 }, { "epoch": 21.07, "learning_rate": 2.236252457499711e-05, "loss": 0.6621, "step": 243250 }, { "epoch": 21.07, "learning_rate": 2.236035619289927e-05, "loss": 0.5767, "step": 243260 }, { "epoch": 21.07, "learning_rate": 2.235818781080143e-05, "loss": 0.7486, "step": 243270 }, { "epoch": 21.07, "learning_rate": 2.2356019428703594e-05, "loss": 0.7398, "step": 243280 }, { "epoch": 21.07, "learning_rate": 2.2353851046605757e-05, "loss": 0.6398, "step": 243290 }, { "epoch": 21.07, "learning_rate": 2.2351682664507917e-05, "loss": 0.7911, "step": 243300 }, { "epoch": 21.07, "learning_rate": 2.2349514282410083e-05, "loss": 0.7274, "step": 243310 }, { "epoch": 21.07, "learning_rate": 2.2347345900312246e-05, "loss": 0.5631, "step": 243320 }, { "epoch": 21.07, "learning_rate": 2.234517751821441e-05, "loss": 0.6782, "step": 243330 }, { "epoch": 21.07, "learning_rate": 2.234300913611657e-05, "loss": 0.6422, "step": 243340 }, { "epoch": 21.08, "learning_rate": 2.2340840754018732e-05, "loss": 0.6692, "step": 243350 }, { "epoch": 21.08, "learning_rate": 2.2338672371920895e-05, "loss": 0.7988, "step": 243360 }, { "epoch": 21.08, "learning_rate": 2.2336503989823058e-05, "loss": 0.7578, "step": 243370 }, { "epoch": 21.08, "learning_rate": 2.2334335607725218e-05, "loss": 0.6789, "step": 243380 }, { "epoch": 21.08, "learning_rate": 2.2332167225627384e-05, "loss": 0.8077, "step": 243390 }, { "epoch": 21.08, "learning_rate": 2.2329998843529547e-05, "loss": 0.6629, "step": 243400 }, { "epoch": 21.08, "learning_rate": 2.232783046143171e-05, "loss": 0.5156, "step": 243410 }, { "epoch": 21.08, "learning_rate": 2.232566207933387e-05, "loss": 0.6751, "step": 243420 }, { "epoch": 21.08, "learning_rate": 2.2323493697236033e-05, "loss": 0.789, "step": 243430 }, { "epoch": 21.08, "learning_rate": 2.2321325315138196e-05, "loss": 0.697, "step": 243440 }, { "epoch": 21.08, "learning_rate": 2.2319156933040362e-05, "loss": 0.7666, "step": 243450 }, { "epoch": 21.09, "learning_rate": 2.231698855094252e-05, "loss": 0.7189, "step": 243460 }, { "epoch": 21.09, "learning_rate": 2.2314820168844685e-05, "loss": 0.6011, "step": 243470 }, { "epoch": 21.09, "learning_rate": 2.2312651786746848e-05, "loss": 0.7617, "step": 243480 }, { "epoch": 21.09, "learning_rate": 2.2310483404649007e-05, "loss": 0.6593, "step": 243490 }, { "epoch": 21.09, "learning_rate": 2.230831502255117e-05, "loss": 0.6734, "step": 243500 }, { "epoch": 21.09, "learning_rate": 2.2306146640453333e-05, "loss": 0.6729, "step": 243510 }, { "epoch": 21.09, "learning_rate": 2.23039782583555e-05, "loss": 0.6103, "step": 243520 }, { "epoch": 21.09, "learning_rate": 2.230180987625766e-05, "loss": 0.7201, "step": 243530 }, { "epoch": 21.09, "learning_rate": 2.2299641494159822e-05, "loss": 0.589, "step": 243540 }, { "epoch": 21.09, "learning_rate": 2.2297473112061985e-05, "loss": 0.7381, "step": 243550 }, { "epoch": 21.09, "learning_rate": 2.2295304729964148e-05, "loss": 0.6625, "step": 243560 }, { "epoch": 21.09, "learning_rate": 2.2293136347866308e-05, "loss": 0.6699, "step": 243570 }, { "epoch": 21.1, "learning_rate": 2.229096796576847e-05, "loss": 0.6917, "step": 243580 }, { "epoch": 21.1, "learning_rate": 2.2288799583670637e-05, "loss": 0.6979, "step": 243590 }, { "epoch": 21.1, "learning_rate": 2.22866312015728e-05, "loss": 0.6522, "step": 243600 }, { "epoch": 21.1, "learning_rate": 2.228446281947496e-05, "loss": 0.6375, "step": 243610 }, { "epoch": 21.1, "learning_rate": 2.2282294437377123e-05, "loss": 0.6929, "step": 243620 }, { "epoch": 21.1, "learning_rate": 2.2280126055279286e-05, "loss": 0.5249, "step": 243630 }, { "epoch": 21.1, "learning_rate": 2.227795767318145e-05, "loss": 0.6, "step": 243640 }, { "epoch": 21.1, "learning_rate": 2.227578929108361e-05, "loss": 0.7473, "step": 243650 }, { "epoch": 21.1, "learning_rate": 2.227362090898577e-05, "loss": 0.6214, "step": 243660 }, { "epoch": 21.1, "learning_rate": 2.2271452526887938e-05, "loss": 0.6625, "step": 243670 }, { "epoch": 21.1, "learning_rate": 2.22692841447901e-05, "loss": 0.6525, "step": 243680 }, { "epoch": 21.11, "learning_rate": 2.226711576269226e-05, "loss": 0.5265, "step": 243690 }, { "epoch": 21.11, "learning_rate": 2.2264947380594424e-05, "loss": 0.6433, "step": 243700 }, { "epoch": 21.11, "learning_rate": 2.2262778998496587e-05, "loss": 0.7303, "step": 243710 }, { "epoch": 21.11, "learning_rate": 2.226061061639875e-05, "loss": 0.6854, "step": 243720 }, { "epoch": 21.11, "learning_rate": 2.225844223430091e-05, "loss": 0.7038, "step": 243730 }, { "epoch": 21.11, "learning_rate": 2.2256273852203076e-05, "loss": 0.6039, "step": 243740 }, { "epoch": 21.11, "learning_rate": 2.225410547010524e-05, "loss": 0.7956, "step": 243750 }, { "epoch": 21.11, "learning_rate": 2.2251937088007398e-05, "loss": 0.6035, "step": 243760 }, { "epoch": 21.11, "learning_rate": 2.224976870590956e-05, "loss": 0.8234, "step": 243770 }, { "epoch": 21.11, "learning_rate": 2.2247600323811724e-05, "loss": 0.8083, "step": 243780 }, { "epoch": 21.11, "learning_rate": 2.2245431941713887e-05, "loss": 0.6189, "step": 243790 }, { "epoch": 21.11, "learning_rate": 2.2243263559616047e-05, "loss": 0.6433, "step": 243800 }, { "epoch": 21.12, "learning_rate": 2.2241095177518213e-05, "loss": 0.714, "step": 243810 }, { "epoch": 21.12, "learning_rate": 2.2238926795420376e-05, "loss": 0.8164, "step": 243820 }, { "epoch": 21.12, "learning_rate": 2.223675841332254e-05, "loss": 0.7489, "step": 243830 }, { "epoch": 21.12, "learning_rate": 2.22345900312247e-05, "loss": 0.7713, "step": 243840 }, { "epoch": 21.12, "learning_rate": 2.2232421649126862e-05, "loss": 0.7636, "step": 243850 }, { "epoch": 21.12, "learning_rate": 2.2230253267029025e-05, "loss": 0.5994, "step": 243860 }, { "epoch": 21.12, "learning_rate": 2.222808488493119e-05, "loss": 0.5261, "step": 243870 }, { "epoch": 21.12, "learning_rate": 2.2225916502833347e-05, "loss": 0.6757, "step": 243880 }, { "epoch": 21.12, "learning_rate": 2.2223748120735514e-05, "loss": 0.6131, "step": 243890 }, { "epoch": 21.12, "learning_rate": 2.2221579738637677e-05, "loss": 0.5478, "step": 243900 }, { "epoch": 21.12, "learning_rate": 2.221941135653984e-05, "loss": 0.6408, "step": 243910 }, { "epoch": 21.13, "learning_rate": 2.2217242974442e-05, "loss": 0.5506, "step": 243920 }, { "epoch": 21.13, "learning_rate": 2.2215074592344162e-05, "loss": 0.5366, "step": 243930 }, { "epoch": 21.13, "learning_rate": 2.2212906210246325e-05, "loss": 0.6518, "step": 243940 }, { "epoch": 21.13, "learning_rate": 2.2210737828148492e-05, "loss": 0.7691, "step": 243950 }, { "epoch": 21.13, "learning_rate": 2.220856944605065e-05, "loss": 0.642, "step": 243960 }, { "epoch": 21.13, "learning_rate": 2.2206401063952814e-05, "loss": 0.6965, "step": 243970 }, { "epoch": 21.13, "learning_rate": 2.2204232681854977e-05, "loss": 0.7493, "step": 243980 }, { "epoch": 21.13, "learning_rate": 2.220206429975714e-05, "loss": 0.6263, "step": 243990 }, { "epoch": 21.13, "learning_rate": 2.21998959176593e-05, "loss": 0.7359, "step": 244000 }, { "epoch": 21.13, "learning_rate": 2.2197727535561463e-05, "loss": 0.602, "step": 244010 }, { "epoch": 21.13, "learning_rate": 2.219555915346363e-05, "loss": 0.8443, "step": 244020 }, { "epoch": 21.13, "learning_rate": 2.219339077136579e-05, "loss": 0.732, "step": 244030 }, { "epoch": 21.14, "learning_rate": 2.2191222389267952e-05, "loss": 0.7544, "step": 244040 }, { "epoch": 21.14, "learning_rate": 2.2189054007170115e-05, "loss": 0.6904, "step": 244050 }, { "epoch": 21.14, "learning_rate": 2.2186885625072278e-05, "loss": 0.7546, "step": 244060 }, { "epoch": 21.14, "learning_rate": 2.2184717242974438e-05, "loss": 0.7188, "step": 244070 }, { "epoch": 21.14, "learning_rate": 2.21825488608766e-05, "loss": 0.6986, "step": 244080 }, { "epoch": 21.14, "learning_rate": 2.2180380478778767e-05, "loss": 0.7945, "step": 244090 }, { "epoch": 21.14, "learning_rate": 2.217821209668093e-05, "loss": 0.6439, "step": 244100 }, { "epoch": 21.14, "learning_rate": 2.217604371458309e-05, "loss": 0.65, "step": 244110 }, { "epoch": 21.14, "learning_rate": 2.2173875332485253e-05, "loss": 0.7295, "step": 244120 }, { "epoch": 21.14, "learning_rate": 2.2171706950387416e-05, "loss": 0.6605, "step": 244130 }, { "epoch": 21.14, "learning_rate": 2.216953856828958e-05, "loss": 0.6111, "step": 244140 }, { "epoch": 21.14, "learning_rate": 2.216737018619174e-05, "loss": 0.8348, "step": 244150 }, { "epoch": 21.15, "learning_rate": 2.21652018040939e-05, "loss": 0.674, "step": 244160 }, { "epoch": 21.15, "learning_rate": 2.2163033421996068e-05, "loss": 0.6838, "step": 244170 }, { "epoch": 21.15, "learning_rate": 2.216086503989823e-05, "loss": 0.8103, "step": 244180 }, { "epoch": 21.15, "learning_rate": 2.215869665780039e-05, "loss": 0.7084, "step": 244190 }, { "epoch": 21.15, "learning_rate": 2.2156528275702553e-05, "loss": 0.7254, "step": 244200 }, { "epoch": 21.15, "learning_rate": 2.2154359893604716e-05, "loss": 0.7181, "step": 244210 }, { "epoch": 21.15, "learning_rate": 2.215219151150688e-05, "loss": 0.7438, "step": 244220 }, { "epoch": 21.15, "learning_rate": 2.215002312940904e-05, "loss": 0.7345, "step": 244230 }, { "epoch": 21.15, "learning_rate": 2.2147854747311205e-05, "loss": 0.6416, "step": 244240 }, { "epoch": 21.15, "learning_rate": 2.214568636521337e-05, "loss": 0.7011, "step": 244250 }, { "epoch": 21.15, "learning_rate": 2.2143517983115528e-05, "loss": 0.6818, "step": 244260 }, { "epoch": 21.16, "learning_rate": 2.214134960101769e-05, "loss": 0.6375, "step": 244270 }, { "epoch": 21.16, "learning_rate": 2.2139181218919854e-05, "loss": 0.5891, "step": 244280 }, { "epoch": 21.16, "learning_rate": 2.2137012836822017e-05, "loss": 0.6345, "step": 244290 }, { "epoch": 21.16, "learning_rate": 2.2134844454724177e-05, "loss": 0.6734, "step": 244300 }, { "epoch": 21.16, "learning_rate": 2.2132676072626343e-05, "loss": 0.6915, "step": 244310 }, { "epoch": 21.16, "learning_rate": 2.2130507690528506e-05, "loss": 0.7872, "step": 244320 }, { "epoch": 21.16, "learning_rate": 2.212833930843067e-05, "loss": 0.7021, "step": 244330 }, { "epoch": 21.16, "learning_rate": 2.212617092633283e-05, "loss": 0.711, "step": 244340 }, { "epoch": 21.16, "learning_rate": 2.212400254423499e-05, "loss": 0.6586, "step": 244350 }, { "epoch": 21.16, "learning_rate": 2.2121834162137155e-05, "loss": 0.7379, "step": 244360 }, { "epoch": 21.16, "learning_rate": 2.211966578003932e-05, "loss": 0.8363, "step": 244370 }, { "epoch": 21.16, "learning_rate": 2.211749739794148e-05, "loss": 0.7277, "step": 244380 }, { "epoch": 21.17, "learning_rate": 2.2115329015843644e-05, "loss": 0.5956, "step": 244390 }, { "epoch": 21.17, "learning_rate": 2.2113160633745807e-05, "loss": 0.6419, "step": 244400 }, { "epoch": 21.17, "learning_rate": 2.211099225164797e-05, "loss": 0.6673, "step": 244410 }, { "epoch": 21.17, "learning_rate": 2.210882386955013e-05, "loss": 0.6552, "step": 244420 }, { "epoch": 21.17, "learning_rate": 2.2106655487452292e-05, "loss": 0.7012, "step": 244430 }, { "epoch": 21.17, "learning_rate": 2.2104487105354455e-05, "loss": 0.8179, "step": 244440 }, { "epoch": 21.17, "learning_rate": 2.2102318723256622e-05, "loss": 0.6164, "step": 244450 }, { "epoch": 21.17, "learning_rate": 2.210015034115878e-05, "loss": 0.633, "step": 244460 }, { "epoch": 21.17, "learning_rate": 2.2097981959060944e-05, "loss": 0.6411, "step": 244470 }, { "epoch": 21.17, "learning_rate": 2.2095813576963107e-05, "loss": 0.7846, "step": 244480 }, { "epoch": 21.17, "learning_rate": 2.209364519486527e-05, "loss": 0.6799, "step": 244490 }, { "epoch": 21.18, "learning_rate": 2.209147681276743e-05, "loss": 0.7233, "step": 244500 }, { "epoch": 21.18, "learning_rate": 2.2089308430669593e-05, "loss": 0.5784, "step": 244510 }, { "epoch": 21.18, "learning_rate": 2.208714004857176e-05, "loss": 0.5328, "step": 244520 }, { "epoch": 21.18, "learning_rate": 2.208497166647392e-05, "loss": 0.6201, "step": 244530 }, { "epoch": 21.18, "learning_rate": 2.2082803284376082e-05, "loss": 0.7103, "step": 244540 }, { "epoch": 21.18, "learning_rate": 2.2080634902278245e-05, "loss": 0.7126, "step": 244550 }, { "epoch": 21.18, "learning_rate": 2.2078466520180408e-05, "loss": 0.5351, "step": 244560 }, { "epoch": 21.18, "learning_rate": 2.2076298138082568e-05, "loss": 0.6767, "step": 244570 }, { "epoch": 21.18, "learning_rate": 2.207412975598473e-05, "loss": 0.6581, "step": 244580 }, { "epoch": 21.18, "learning_rate": 2.2071961373886897e-05, "loss": 0.8042, "step": 244590 }, { "epoch": 21.18, "learning_rate": 2.206979299178906e-05, "loss": 0.5948, "step": 244600 }, { "epoch": 21.18, "learning_rate": 2.206762460969122e-05, "loss": 0.7624, "step": 244610 }, { "epoch": 21.19, "learning_rate": 2.2065456227593383e-05, "loss": 0.6991, "step": 244620 }, { "epoch": 21.19, "learning_rate": 2.2063287845495546e-05, "loss": 0.7319, "step": 244630 }, { "epoch": 21.19, "learning_rate": 2.206111946339771e-05, "loss": 0.6, "step": 244640 }, { "epoch": 21.19, "learning_rate": 2.2058951081299868e-05, "loss": 0.8655, "step": 244650 }, { "epoch": 21.19, "learning_rate": 2.2056782699202035e-05, "loss": 0.6564, "step": 244660 }, { "epoch": 21.19, "learning_rate": 2.2054614317104198e-05, "loss": 0.6656, "step": 244670 }, { "epoch": 21.19, "learning_rate": 2.205244593500636e-05, "loss": 0.653, "step": 244680 }, { "epoch": 21.19, "learning_rate": 2.205027755290852e-05, "loss": 0.7454, "step": 244690 }, { "epoch": 21.19, "learning_rate": 2.2048109170810683e-05, "loss": 0.6574, "step": 244700 }, { "epoch": 21.19, "learning_rate": 2.2045940788712846e-05, "loss": 0.7247, "step": 244710 }, { "epoch": 21.19, "learning_rate": 2.204377240661501e-05, "loss": 0.7011, "step": 244720 }, { "epoch": 21.2, "learning_rate": 2.204160402451717e-05, "loss": 0.5948, "step": 244730 }, { "epoch": 21.2, "learning_rate": 2.2039435642419335e-05, "loss": 0.7172, "step": 244740 }, { "epoch": 21.2, "learning_rate": 2.2037267260321498e-05, "loss": 0.8352, "step": 244750 }, { "epoch": 21.2, "learning_rate": 2.203509887822366e-05, "loss": 0.6999, "step": 244760 }, { "epoch": 21.2, "learning_rate": 2.203293049612582e-05, "loss": 0.6633, "step": 244770 }, { "epoch": 21.2, "learning_rate": 2.2030762114027984e-05, "loss": 0.6166, "step": 244780 }, { "epoch": 21.2, "learning_rate": 2.2028593731930147e-05, "loss": 0.6534, "step": 244790 }, { "epoch": 21.2, "learning_rate": 2.2026425349832307e-05, "loss": 0.7285, "step": 244800 }, { "epoch": 21.2, "learning_rate": 2.2024256967734473e-05, "loss": 0.7256, "step": 244810 }, { "epoch": 21.2, "learning_rate": 2.2022088585636636e-05, "loss": 0.8114, "step": 244820 }, { "epoch": 21.2, "learning_rate": 2.20199202035388e-05, "loss": 0.654, "step": 244830 }, { "epoch": 21.2, "learning_rate": 2.201775182144096e-05, "loss": 0.6791, "step": 244840 }, { "epoch": 21.21, "learning_rate": 2.201558343934312e-05, "loss": 0.7768, "step": 244850 }, { "epoch": 21.21, "learning_rate": 2.2013415057245285e-05, "loss": 0.6326, "step": 244860 }, { "epoch": 21.21, "learning_rate": 2.201124667514745e-05, "loss": 0.6697, "step": 244870 }, { "epoch": 21.21, "learning_rate": 2.200907829304961e-05, "loss": 0.6819, "step": 244880 }, { "epoch": 21.21, "learning_rate": 2.2006909910951774e-05, "loss": 0.6721, "step": 244890 }, { "epoch": 21.21, "learning_rate": 2.2004741528853937e-05, "loss": 0.7852, "step": 244900 }, { "epoch": 21.21, "learning_rate": 2.20025731467561e-05, "loss": 0.6388, "step": 244910 }, { "epoch": 21.21, "learning_rate": 2.200040476465826e-05, "loss": 0.6729, "step": 244920 }, { "epoch": 21.21, "learning_rate": 2.1998236382560422e-05, "loss": 0.7936, "step": 244930 }, { "epoch": 21.21, "learning_rate": 2.199606800046259e-05, "loss": 0.7372, "step": 244940 }, { "epoch": 21.21, "learning_rate": 2.199389961836475e-05, "loss": 0.5999, "step": 244950 }, { "epoch": 21.22, "learning_rate": 2.199173123626691e-05, "loss": 0.7508, "step": 244960 }, { "epoch": 21.22, "learning_rate": 2.1989562854169074e-05, "loss": 0.6556, "step": 244970 }, { "epoch": 21.22, "learning_rate": 2.1987394472071237e-05, "loss": 0.7171, "step": 244980 }, { "epoch": 21.22, "learning_rate": 2.19852260899734e-05, "loss": 0.864, "step": 244990 }, { "epoch": 21.22, "learning_rate": 2.198305770787556e-05, "loss": 0.5762, "step": 245000 }, { "epoch": 21.22, "learning_rate": 2.1980889325777723e-05, "loss": 0.7503, "step": 245010 }, { "epoch": 21.22, "learning_rate": 2.197872094367989e-05, "loss": 0.6279, "step": 245020 }, { "epoch": 21.22, "learning_rate": 2.197655256158205e-05, "loss": 0.7241, "step": 245030 }, { "epoch": 21.22, "learning_rate": 2.1974384179484212e-05, "loss": 0.6481, "step": 245040 }, { "epoch": 21.22, "learning_rate": 2.1972215797386375e-05, "loss": 0.714, "step": 245050 }, { "epoch": 21.22, "learning_rate": 2.1970047415288538e-05, "loss": 0.7586, "step": 245060 }, { "epoch": 21.22, "learning_rate": 2.1967879033190697e-05, "loss": 0.806, "step": 245070 }, { "epoch": 21.23, "learning_rate": 2.196571065109286e-05, "loss": 0.7262, "step": 245080 }, { "epoch": 21.23, "learning_rate": 2.1963542268995027e-05, "loss": 0.7603, "step": 245090 }, { "epoch": 21.23, "learning_rate": 2.196137388689719e-05, "loss": 0.755, "step": 245100 }, { "epoch": 21.23, "learning_rate": 2.195920550479935e-05, "loss": 0.7021, "step": 245110 }, { "epoch": 21.23, "learning_rate": 2.1957037122701512e-05, "loss": 0.6896, "step": 245120 }, { "epoch": 21.23, "learning_rate": 2.1954868740603675e-05, "loss": 0.7238, "step": 245130 }, { "epoch": 21.23, "learning_rate": 2.195270035850584e-05, "loss": 0.722, "step": 245140 }, { "epoch": 21.23, "learning_rate": 2.1950531976407998e-05, "loss": 0.7109, "step": 245150 }, { "epoch": 21.23, "learning_rate": 2.1948363594310164e-05, "loss": 0.8133, "step": 245160 }, { "epoch": 21.23, "learning_rate": 2.1946195212212327e-05, "loss": 0.6378, "step": 245170 }, { "epoch": 21.23, "learning_rate": 2.194402683011449e-05, "loss": 0.7539, "step": 245180 }, { "epoch": 21.24, "learning_rate": 2.194185844801665e-05, "loss": 0.6527, "step": 245190 }, { "epoch": 21.24, "learning_rate": 2.1939690065918813e-05, "loss": 0.8067, "step": 245200 }, { "epoch": 21.24, "learning_rate": 2.1937521683820976e-05, "loss": 0.698, "step": 245210 }, { "epoch": 21.24, "learning_rate": 2.1935353301723143e-05, "loss": 0.7624, "step": 245220 }, { "epoch": 21.24, "learning_rate": 2.19331849196253e-05, "loss": 0.6163, "step": 245230 }, { "epoch": 21.24, "learning_rate": 2.1931016537527465e-05, "loss": 0.7172, "step": 245240 }, { "epoch": 21.24, "learning_rate": 2.1928848155429628e-05, "loss": 0.614, "step": 245250 }, { "epoch": 21.24, "learning_rate": 2.192667977333179e-05, "loss": 0.659, "step": 245260 }, { "epoch": 21.24, "learning_rate": 2.192451139123395e-05, "loss": 0.6728, "step": 245270 }, { "epoch": 21.24, "learning_rate": 2.1922343009136114e-05, "loss": 0.5892, "step": 245280 }, { "epoch": 21.24, "learning_rate": 2.1920174627038277e-05, "loss": 0.6652, "step": 245290 }, { "epoch": 21.24, "learning_rate": 2.1918006244940436e-05, "loss": 0.7451, "step": 245300 }, { "epoch": 21.25, "learning_rate": 2.1915837862842603e-05, "loss": 0.6341, "step": 245310 }, { "epoch": 21.25, "learning_rate": 2.1913669480744766e-05, "loss": 0.7444, "step": 245320 }, { "epoch": 21.25, "learning_rate": 2.191150109864693e-05, "loss": 0.7421, "step": 245330 }, { "epoch": 21.25, "learning_rate": 2.190933271654909e-05, "loss": 0.8347, "step": 245340 }, { "epoch": 21.25, "learning_rate": 2.190716433445125e-05, "loss": 0.7951, "step": 245350 }, { "epoch": 21.25, "learning_rate": 2.1904995952353414e-05, "loss": 0.7663, "step": 245360 }, { "epoch": 21.25, "learning_rate": 2.190282757025558e-05, "loss": 0.7942, "step": 245370 }, { "epoch": 21.25, "learning_rate": 2.190065918815774e-05, "loss": 0.7395, "step": 245380 }, { "epoch": 21.25, "learning_rate": 2.1898490806059903e-05, "loss": 0.6431, "step": 245390 }, { "epoch": 21.25, "learning_rate": 2.1896322423962066e-05, "loss": 0.7572, "step": 245400 }, { "epoch": 21.25, "learning_rate": 2.189415404186423e-05, "loss": 0.6682, "step": 245410 }, { "epoch": 21.25, "learning_rate": 2.189198565976639e-05, "loss": 0.7856, "step": 245420 }, { "epoch": 21.26, "learning_rate": 2.1889817277668552e-05, "loss": 0.8254, "step": 245430 }, { "epoch": 21.26, "learning_rate": 2.188764889557072e-05, "loss": 0.731, "step": 245440 }, { "epoch": 21.26, "learning_rate": 2.188548051347288e-05, "loss": 0.691, "step": 245450 }, { "epoch": 21.26, "learning_rate": 2.188331213137504e-05, "loss": 0.6374, "step": 245460 }, { "epoch": 21.26, "learning_rate": 2.1881143749277204e-05, "loss": 0.8453, "step": 245470 }, { "epoch": 21.26, "learning_rate": 2.1878975367179367e-05, "loss": 0.7242, "step": 245480 }, { "epoch": 21.26, "learning_rate": 2.187680698508153e-05, "loss": 0.4823, "step": 245490 }, { "epoch": 21.26, "learning_rate": 2.187463860298369e-05, "loss": 0.8237, "step": 245500 }, { "epoch": 21.26, "learning_rate": 2.1872470220885853e-05, "loss": 0.5832, "step": 245510 }, { "epoch": 21.26, "learning_rate": 2.187030183878802e-05, "loss": 0.6505, "step": 245520 }, { "epoch": 21.26, "learning_rate": 2.1868133456690182e-05, "loss": 0.693, "step": 245530 }, { "epoch": 21.27, "learning_rate": 2.186596507459234e-05, "loss": 0.6209, "step": 245540 }, { "epoch": 21.27, "learning_rate": 2.1863796692494505e-05, "loss": 0.7326, "step": 245550 }, { "epoch": 21.27, "learning_rate": 2.1861628310396668e-05, "loss": 0.6552, "step": 245560 }, { "epoch": 21.27, "learning_rate": 2.1859459928298827e-05, "loss": 0.8978, "step": 245570 }, { "epoch": 21.27, "learning_rate": 2.185729154620099e-05, "loss": 0.6795, "step": 245580 }, { "epoch": 21.27, "learning_rate": 2.1855123164103157e-05, "loss": 0.6833, "step": 245590 }, { "epoch": 21.27, "learning_rate": 2.185295478200532e-05, "loss": 0.6899, "step": 245600 }, { "epoch": 21.27, "learning_rate": 2.185078639990748e-05, "loss": 0.6273, "step": 245610 }, { "epoch": 21.27, "learning_rate": 2.1848618017809642e-05, "loss": 0.8107, "step": 245620 }, { "epoch": 21.27, "learning_rate": 2.1846449635711805e-05, "loss": 0.5788, "step": 245630 }, { "epoch": 21.27, "learning_rate": 2.184428125361397e-05, "loss": 0.6062, "step": 245640 }, { "epoch": 21.27, "learning_rate": 2.1842112871516128e-05, "loss": 0.6781, "step": 245650 }, { "epoch": 21.28, "learning_rate": 2.1839944489418294e-05, "loss": 0.6209, "step": 245660 }, { "epoch": 21.28, "learning_rate": 2.1837776107320457e-05, "loss": 0.7618, "step": 245670 }, { "epoch": 21.28, "learning_rate": 2.183560772522262e-05, "loss": 0.5766, "step": 245680 }, { "epoch": 21.28, "learning_rate": 2.183343934312478e-05, "loss": 0.6762, "step": 245690 }, { "epoch": 21.28, "learning_rate": 2.1831270961026943e-05, "loss": 0.7238, "step": 245700 }, { "epoch": 21.28, "learning_rate": 2.1829102578929106e-05, "loss": 0.6986, "step": 245710 }, { "epoch": 21.28, "learning_rate": 2.1826934196831272e-05, "loss": 0.6507, "step": 245720 }, { "epoch": 21.28, "learning_rate": 2.1824765814733432e-05, "loss": 0.6708, "step": 245730 }, { "epoch": 21.28, "learning_rate": 2.1822597432635595e-05, "loss": 0.7475, "step": 245740 }, { "epoch": 21.28, "learning_rate": 2.1820429050537758e-05, "loss": 0.6482, "step": 245750 }, { "epoch": 21.28, "learning_rate": 2.181826066843992e-05, "loss": 0.6473, "step": 245760 }, { "epoch": 21.29, "learning_rate": 2.181609228634208e-05, "loss": 0.6946, "step": 245770 }, { "epoch": 21.29, "learning_rate": 2.1813923904244244e-05, "loss": 0.5648, "step": 245780 }, { "epoch": 21.29, "learning_rate": 2.1811755522146407e-05, "loss": 0.6686, "step": 245790 }, { "epoch": 21.29, "learning_rate": 2.1809587140048566e-05, "loss": 0.7981, "step": 245800 }, { "epoch": 21.29, "learning_rate": 2.1807418757950733e-05, "loss": 0.6607, "step": 245810 }, { "epoch": 21.29, "learning_rate": 2.1805250375852896e-05, "loss": 0.7168, "step": 245820 }, { "epoch": 21.29, "learning_rate": 2.180308199375506e-05, "loss": 0.6175, "step": 245830 }, { "epoch": 21.29, "learning_rate": 2.1800913611657218e-05, "loss": 0.6551, "step": 245840 }, { "epoch": 21.29, "learning_rate": 2.179874522955938e-05, "loss": 0.6553, "step": 245850 }, { "epoch": 21.29, "learning_rate": 2.1796576847461544e-05, "loss": 0.7104, "step": 245860 }, { "epoch": 21.29, "learning_rate": 2.179440846536371e-05, "loss": 0.6504, "step": 245870 }, { "epoch": 21.29, "learning_rate": 2.179224008326587e-05, "loss": 0.72, "step": 245880 }, { "epoch": 21.3, "learning_rate": 2.1790071701168033e-05, "loss": 0.5549, "step": 245890 }, { "epoch": 21.3, "learning_rate": 2.1787903319070196e-05, "loss": 0.5808, "step": 245900 }, { "epoch": 21.3, "learning_rate": 2.178573493697236e-05, "loss": 0.7476, "step": 245910 }, { "epoch": 21.3, "learning_rate": 2.178356655487452e-05, "loss": 0.8024, "step": 245920 }, { "epoch": 21.3, "learning_rate": 2.1781398172776682e-05, "loss": 0.7678, "step": 245930 }, { "epoch": 21.3, "learning_rate": 2.1779229790678848e-05, "loss": 0.8162, "step": 245940 }, { "epoch": 21.3, "learning_rate": 2.177706140858101e-05, "loss": 0.9419, "step": 245950 }, { "epoch": 21.3, "learning_rate": 2.177489302648317e-05, "loss": 0.7941, "step": 245960 }, { "epoch": 21.3, "learning_rate": 2.1772724644385334e-05, "loss": 0.6387, "step": 245970 }, { "epoch": 21.3, "learning_rate": 2.1770556262287497e-05, "loss": 0.7736, "step": 245980 }, { "epoch": 21.3, "learning_rate": 2.176838788018966e-05, "loss": 0.7578, "step": 245990 }, { "epoch": 21.31, "learning_rate": 2.176621949809182e-05, "loss": 0.6644, "step": 246000 }, { "epoch": 21.31, "learning_rate": 2.1764051115993986e-05, "loss": 0.5883, "step": 246010 }, { "epoch": 21.31, "learning_rate": 2.176188273389615e-05, "loss": 0.7766, "step": 246020 }, { "epoch": 21.31, "learning_rate": 2.1759714351798312e-05, "loss": 0.9118, "step": 246030 }, { "epoch": 21.31, "learning_rate": 2.175754596970047e-05, "loss": 0.6839, "step": 246040 }, { "epoch": 21.31, "learning_rate": 2.1755377587602635e-05, "loss": 0.7001, "step": 246050 }, { "epoch": 21.31, "learning_rate": 2.1753209205504798e-05, "loss": 0.7057, "step": 246060 }, { "epoch": 21.31, "learning_rate": 2.1751040823406957e-05, "loss": 0.642, "step": 246070 }, { "epoch": 21.31, "learning_rate": 2.174887244130912e-05, "loss": 0.8445, "step": 246080 }, { "epoch": 21.31, "learning_rate": 2.1746704059211287e-05, "loss": 0.7402, "step": 246090 }, { "epoch": 21.31, "learning_rate": 2.174453567711345e-05, "loss": 0.7144, "step": 246100 }, { "epoch": 21.31, "learning_rate": 2.174236729501561e-05, "loss": 0.7403, "step": 246110 }, { "epoch": 21.32, "learning_rate": 2.1740198912917772e-05, "loss": 0.6269, "step": 246120 }, { "epoch": 21.32, "learning_rate": 2.1738030530819935e-05, "loss": 0.6921, "step": 246130 }, { "epoch": 21.32, "learning_rate": 2.1735862148722098e-05, "loss": 0.7734, "step": 246140 }, { "epoch": 21.32, "learning_rate": 2.1733693766624258e-05, "loss": 0.6125, "step": 246150 }, { "epoch": 21.32, "learning_rate": 2.1731525384526424e-05, "loss": 0.6485, "step": 246160 }, { "epoch": 21.32, "learning_rate": 2.1729357002428587e-05, "loss": 0.6034, "step": 246170 }, { "epoch": 21.32, "learning_rate": 2.172718862033075e-05, "loss": 0.6693, "step": 246180 }, { "epoch": 21.32, "learning_rate": 2.172502023823291e-05, "loss": 0.5132, "step": 246190 }, { "epoch": 21.32, "learning_rate": 2.1722851856135073e-05, "loss": 0.7777, "step": 246200 }, { "epoch": 21.32, "learning_rate": 2.1720683474037236e-05, "loss": 0.6451, "step": 246210 }, { "epoch": 21.32, "learning_rate": 2.1718515091939402e-05, "loss": 0.5808, "step": 246220 }, { "epoch": 21.33, "learning_rate": 2.1716346709841562e-05, "loss": 0.73, "step": 246230 }, { "epoch": 21.33, "learning_rate": 2.1714178327743725e-05, "loss": 0.7173, "step": 246240 }, { "epoch": 21.33, "learning_rate": 2.1712009945645888e-05, "loss": 0.5045, "step": 246250 }, { "epoch": 21.33, "learning_rate": 2.170984156354805e-05, "loss": 0.7597, "step": 246260 }, { "epoch": 21.33, "learning_rate": 2.170767318145021e-05, "loss": 0.5533, "step": 246270 }, { "epoch": 21.33, "learning_rate": 2.1705504799352373e-05, "loss": 0.6338, "step": 246280 }, { "epoch": 21.33, "learning_rate": 2.170333641725454e-05, "loss": 0.671, "step": 246290 }, { "epoch": 21.33, "learning_rate": 2.1701168035156703e-05, "loss": 0.6494, "step": 246300 }, { "epoch": 21.33, "learning_rate": 2.1698999653058862e-05, "loss": 0.6158, "step": 246310 }, { "epoch": 21.33, "learning_rate": 2.1696831270961025e-05, "loss": 0.5234, "step": 246320 }, { "epoch": 21.33, "learning_rate": 2.169466288886319e-05, "loss": 0.6741, "step": 246330 }, { "epoch": 21.33, "learning_rate": 2.1692494506765348e-05, "loss": 0.7738, "step": 246340 }, { "epoch": 21.34, "learning_rate": 2.169032612466751e-05, "loss": 0.6767, "step": 246350 }, { "epoch": 21.34, "learning_rate": 2.1688157742569674e-05, "loss": 0.6211, "step": 246360 }, { "epoch": 21.34, "learning_rate": 2.168598936047184e-05, "loss": 0.8798, "step": 246370 }, { "epoch": 21.34, "learning_rate": 2.1683820978374e-05, "loss": 0.5803, "step": 246380 }, { "epoch": 21.34, "learning_rate": 2.1681652596276163e-05, "loss": 0.6609, "step": 246390 }, { "epoch": 21.34, "learning_rate": 2.1679484214178326e-05, "loss": 0.7719, "step": 246400 }, { "epoch": 21.34, "learning_rate": 2.167731583208049e-05, "loss": 0.6315, "step": 246410 }, { "epoch": 21.34, "learning_rate": 2.167514744998265e-05, "loss": 0.7442, "step": 246420 }, { "epoch": 21.34, "learning_rate": 2.1672979067884812e-05, "loss": 0.7919, "step": 246430 }, { "epoch": 21.34, "learning_rate": 2.1670810685786978e-05, "loss": 0.6465, "step": 246440 }, { "epoch": 21.34, "learning_rate": 2.166864230368914e-05, "loss": 0.5735, "step": 246450 }, { "epoch": 21.34, "learning_rate": 2.16664739215913e-05, "loss": 0.6445, "step": 246460 }, { "epoch": 21.35, "learning_rate": 2.1664305539493464e-05, "loss": 0.7795, "step": 246470 }, { "epoch": 21.35, "learning_rate": 2.1662137157395627e-05, "loss": 0.621, "step": 246480 }, { "epoch": 21.35, "learning_rate": 2.165996877529779e-05, "loss": 0.7449, "step": 246490 }, { "epoch": 21.35, "learning_rate": 2.165780039319995e-05, "loss": 0.6578, "step": 246500 }, { "epoch": 21.35, "learning_rate": 2.1655632011102116e-05, "loss": 0.7671, "step": 246510 }, { "epoch": 21.35, "learning_rate": 2.165346362900428e-05, "loss": 0.6986, "step": 246520 }, { "epoch": 21.35, "learning_rate": 2.1651295246906442e-05, "loss": 0.6875, "step": 246530 }, { "epoch": 21.35, "learning_rate": 2.16491268648086e-05, "loss": 0.7382, "step": 246540 }, { "epoch": 21.35, "learning_rate": 2.1646958482710764e-05, "loss": 0.728, "step": 246550 }, { "epoch": 21.35, "learning_rate": 2.1644790100612927e-05, "loss": 0.7409, "step": 246560 }, { "epoch": 21.35, "learning_rate": 2.1642621718515087e-05, "loss": 0.684, "step": 246570 }, { "epoch": 21.36, "learning_rate": 2.164045333641725e-05, "loss": 0.7355, "step": 246580 }, { "epoch": 21.36, "learning_rate": 2.1638284954319416e-05, "loss": 0.6232, "step": 246590 }, { "epoch": 21.36, "learning_rate": 2.163611657222158e-05, "loss": 0.8324, "step": 246600 }, { "epoch": 21.36, "learning_rate": 2.163394819012374e-05, "loss": 0.6642, "step": 246610 }, { "epoch": 21.36, "learning_rate": 2.1631779808025902e-05, "loss": 0.6906, "step": 246620 }, { "epoch": 21.36, "learning_rate": 2.1629611425928065e-05, "loss": 0.5636, "step": 246630 }, { "epoch": 21.36, "learning_rate": 2.1627443043830228e-05, "loss": 0.7746, "step": 246640 }, { "epoch": 21.36, "learning_rate": 2.1625274661732388e-05, "loss": 0.8671, "step": 246650 }, { "epoch": 21.36, "learning_rate": 2.1623106279634554e-05, "loss": 0.586, "step": 246660 }, { "epoch": 21.36, "learning_rate": 2.1620937897536717e-05, "loss": 0.6732, "step": 246670 }, { "epoch": 21.36, "learning_rate": 2.161876951543888e-05, "loss": 0.813, "step": 246680 }, { "epoch": 21.36, "learning_rate": 2.161660113334104e-05, "loss": 0.7487, "step": 246690 }, { "epoch": 21.37, "learning_rate": 2.1614432751243203e-05, "loss": 0.7035, "step": 246700 }, { "epoch": 21.37, "learning_rate": 2.1612264369145366e-05, "loss": 0.6569, "step": 246710 }, { "epoch": 21.37, "learning_rate": 2.1610095987047532e-05, "loss": 0.6642, "step": 246720 }, { "epoch": 21.37, "learning_rate": 2.160792760494969e-05, "loss": 0.6227, "step": 246730 }, { "epoch": 21.37, "learning_rate": 2.1605759222851855e-05, "loss": 0.5593, "step": 246740 }, { "epoch": 21.37, "learning_rate": 2.1603590840754018e-05, "loss": 0.7338, "step": 246750 }, { "epoch": 21.37, "learning_rate": 2.160142245865618e-05, "loss": 0.6413, "step": 246760 }, { "epoch": 21.37, "learning_rate": 2.159925407655834e-05, "loss": 0.6409, "step": 246770 }, { "epoch": 21.37, "learning_rate": 2.1597085694460503e-05, "loss": 0.7507, "step": 246780 }, { "epoch": 21.37, "learning_rate": 2.159491731236267e-05, "loss": 0.7632, "step": 246790 }, { "epoch": 21.37, "learning_rate": 2.1592748930264833e-05, "loss": 0.8102, "step": 246800 }, { "epoch": 21.38, "learning_rate": 2.1590580548166992e-05, "loss": 0.8058, "step": 246810 }, { "epoch": 21.38, "learning_rate": 2.1588412166069155e-05, "loss": 0.6439, "step": 246820 }, { "epoch": 21.38, "learning_rate": 2.1586243783971318e-05, "loss": 0.634, "step": 246830 }, { "epoch": 21.38, "learning_rate": 2.1584075401873478e-05, "loss": 0.6736, "step": 246840 }, { "epoch": 21.38, "learning_rate": 2.158190701977564e-05, "loss": 0.6469, "step": 246850 }, { "epoch": 21.38, "learning_rate": 2.1579738637677804e-05, "loss": 0.7138, "step": 246860 }, { "epoch": 21.38, "learning_rate": 2.157757025557997e-05, "loss": 0.6779, "step": 246870 }, { "epoch": 21.38, "learning_rate": 2.157540187348213e-05, "loss": 0.7303, "step": 246880 }, { "epoch": 21.38, "learning_rate": 2.1573233491384293e-05, "loss": 0.6839, "step": 246890 }, { "epoch": 21.38, "learning_rate": 2.1571065109286456e-05, "loss": 0.6252, "step": 246900 }, { "epoch": 21.38, "learning_rate": 2.156889672718862e-05, "loss": 0.7509, "step": 246910 }, { "epoch": 21.38, "learning_rate": 2.156672834509078e-05, "loss": 0.6393, "step": 246920 }, { "epoch": 21.39, "learning_rate": 2.156455996299294e-05, "loss": 0.6613, "step": 246930 }, { "epoch": 21.39, "learning_rate": 2.1562391580895108e-05, "loss": 0.6402, "step": 246940 }, { "epoch": 21.39, "learning_rate": 2.156022319879727e-05, "loss": 0.7506, "step": 246950 }, { "epoch": 21.39, "learning_rate": 2.155805481669943e-05, "loss": 0.539, "step": 246960 }, { "epoch": 21.39, "learning_rate": 2.1555886434601594e-05, "loss": 0.6136, "step": 246970 }, { "epoch": 21.39, "learning_rate": 2.1553718052503757e-05, "loss": 0.734, "step": 246980 }, { "epoch": 21.39, "learning_rate": 2.155154967040592e-05, "loss": 0.6616, "step": 246990 }, { "epoch": 21.39, "learning_rate": 2.154938128830808e-05, "loss": 0.7029, "step": 247000 }, { "epoch": 21.39, "learning_rate": 2.1547212906210246e-05, "loss": 0.7986, "step": 247010 }, { "epoch": 21.39, "learning_rate": 2.154504452411241e-05, "loss": 0.653, "step": 247020 }, { "epoch": 21.39, "learning_rate": 2.154287614201457e-05, "loss": 0.806, "step": 247030 }, { "epoch": 21.4, "learning_rate": 2.154070775991673e-05, "loss": 0.78, "step": 247040 }, { "epoch": 21.4, "learning_rate": 2.1538539377818894e-05, "loss": 0.7385, "step": 247050 }, { "epoch": 21.4, "learning_rate": 2.1536370995721057e-05, "loss": 0.7352, "step": 247060 }, { "epoch": 21.4, "learning_rate": 2.1534202613623224e-05, "loss": 0.601, "step": 247070 }, { "epoch": 21.4, "learning_rate": 2.1532034231525383e-05, "loss": 0.7462, "step": 247080 }, { "epoch": 21.4, "learning_rate": 2.1529865849427546e-05, "loss": 0.6835, "step": 247090 }, { "epoch": 21.4, "learning_rate": 2.152769746732971e-05, "loss": 0.8982, "step": 247100 }, { "epoch": 21.4, "learning_rate": 2.152552908523187e-05, "loss": 0.6687, "step": 247110 }, { "epoch": 21.4, "learning_rate": 2.1523360703134032e-05, "loss": 0.7332, "step": 247120 }, { "epoch": 21.4, "learning_rate": 2.1521192321036195e-05, "loss": 0.7224, "step": 247130 }, { "epoch": 21.4, "learning_rate": 2.1519023938938358e-05, "loss": 0.7315, "step": 247140 }, { "epoch": 21.4, "learning_rate": 2.1516855556840517e-05, "loss": 0.6491, "step": 247150 }, { "epoch": 21.41, "learning_rate": 2.1514687174742684e-05, "loss": 0.6195, "step": 247160 }, { "epoch": 21.41, "learning_rate": 2.1512518792644847e-05, "loss": 0.7315, "step": 247170 }, { "epoch": 21.41, "learning_rate": 2.151035041054701e-05, "loss": 0.6409, "step": 247180 }, { "epoch": 21.41, "learning_rate": 2.150818202844917e-05, "loss": 0.6835, "step": 247190 }, { "epoch": 21.41, "learning_rate": 2.1506013646351332e-05, "loss": 0.7614, "step": 247200 }, { "epoch": 21.41, "learning_rate": 2.1503845264253495e-05, "loss": 0.5931, "step": 247210 }, { "epoch": 21.41, "learning_rate": 2.1501676882155662e-05, "loss": 0.5518, "step": 247220 }, { "epoch": 21.41, "learning_rate": 2.149950850005782e-05, "loss": 0.7734, "step": 247230 }, { "epoch": 21.41, "learning_rate": 2.1497340117959985e-05, "loss": 0.6829, "step": 247240 }, { "epoch": 21.41, "learning_rate": 2.1495171735862148e-05, "loss": 0.6992, "step": 247250 }, { "epoch": 21.41, "learning_rate": 2.149300335376431e-05, "loss": 0.6986, "step": 247260 }, { "epoch": 21.42, "learning_rate": 2.149083497166647e-05, "loss": 0.7453, "step": 247270 }, { "epoch": 21.42, "learning_rate": 2.1488666589568633e-05, "loss": 0.6914, "step": 247280 }, { "epoch": 21.42, "learning_rate": 2.14864982074708e-05, "loss": 0.6294, "step": 247290 }, { "epoch": 21.42, "learning_rate": 2.1484329825372963e-05, "loss": 0.8436, "step": 247300 }, { "epoch": 21.42, "learning_rate": 2.1482161443275122e-05, "loss": 0.7548, "step": 247310 }, { "epoch": 21.42, "learning_rate": 2.1479993061177285e-05, "loss": 0.6861, "step": 247320 }, { "epoch": 21.42, "learning_rate": 2.1477824679079448e-05, "loss": 0.7494, "step": 247330 }, { "epoch": 21.42, "learning_rate": 2.1475656296981608e-05, "loss": 0.6133, "step": 247340 }, { "epoch": 21.42, "learning_rate": 2.147348791488377e-05, "loss": 0.7201, "step": 247350 }, { "epoch": 21.42, "learning_rate": 2.1471319532785937e-05, "loss": 0.7744, "step": 247360 }, { "epoch": 21.42, "learning_rate": 2.14691511506881e-05, "loss": 0.7115, "step": 247370 }, { "epoch": 21.42, "learning_rate": 2.146698276859026e-05, "loss": 0.6614, "step": 247380 }, { "epoch": 21.43, "learning_rate": 2.1464814386492423e-05, "loss": 0.7761, "step": 247390 }, { "epoch": 21.43, "learning_rate": 2.1462646004394586e-05, "loss": 0.6298, "step": 247400 }, { "epoch": 21.43, "learning_rate": 2.146047762229675e-05, "loss": 0.7639, "step": 247410 }, { "epoch": 21.43, "learning_rate": 2.145830924019891e-05, "loss": 0.6945, "step": 247420 }, { "epoch": 21.43, "learning_rate": 2.145614085810107e-05, "loss": 0.8897, "step": 247430 }, { "epoch": 21.43, "learning_rate": 2.1453972476003238e-05, "loss": 0.6556, "step": 247440 }, { "epoch": 21.43, "learning_rate": 2.14518040939054e-05, "loss": 0.8069, "step": 247450 }, { "epoch": 21.43, "learning_rate": 2.144963571180756e-05, "loss": 0.7738, "step": 247460 }, { "epoch": 21.43, "learning_rate": 2.1447467329709723e-05, "loss": 0.7334, "step": 247470 }, { "epoch": 21.43, "learning_rate": 2.1445298947611886e-05, "loss": 0.6702, "step": 247480 }, { "epoch": 21.43, "learning_rate": 2.144313056551405e-05, "loss": 0.6564, "step": 247490 }, { "epoch": 21.44, "learning_rate": 2.144096218341621e-05, "loss": 0.7925, "step": 247500 }, { "epoch": 21.44, "learning_rate": 2.1438793801318375e-05, "loss": 0.5623, "step": 247510 }, { "epoch": 21.44, "learning_rate": 2.143662541922054e-05, "loss": 0.6965, "step": 247520 }, { "epoch": 21.44, "learning_rate": 2.14344570371227e-05, "loss": 0.6326, "step": 247530 }, { "epoch": 21.44, "learning_rate": 2.143228865502486e-05, "loss": 0.7308, "step": 247540 }, { "epoch": 21.44, "learning_rate": 2.1430120272927024e-05, "loss": 0.7264, "step": 247550 }, { "epoch": 21.44, "learning_rate": 2.1427951890829187e-05, "loss": 0.815, "step": 247560 }, { "epoch": 21.44, "learning_rate": 2.1425783508731353e-05, "loss": 0.6739, "step": 247570 }, { "epoch": 21.44, "learning_rate": 2.1423615126633513e-05, "loss": 0.587, "step": 247580 }, { "epoch": 21.44, "learning_rate": 2.1421446744535676e-05, "loss": 0.5909, "step": 247590 }, { "epoch": 21.44, "learning_rate": 2.141927836243784e-05, "loss": 0.6118, "step": 247600 }, { "epoch": 21.44, "learning_rate": 2.141710998034e-05, "loss": 0.565, "step": 247610 }, { "epoch": 21.45, "learning_rate": 2.1414941598242162e-05, "loss": 0.8028, "step": 247620 }, { "epoch": 21.45, "learning_rate": 2.1412773216144325e-05, "loss": 0.6755, "step": 247630 }, { "epoch": 21.45, "learning_rate": 2.141060483404649e-05, "loss": 0.7725, "step": 247640 }, { "epoch": 21.45, "learning_rate": 2.1408436451948647e-05, "loss": 0.6122, "step": 247650 }, { "epoch": 21.45, "learning_rate": 2.1406268069850814e-05, "loss": 0.7686, "step": 247660 }, { "epoch": 21.45, "learning_rate": 2.1404099687752977e-05, "loss": 0.7563, "step": 247670 }, { "epoch": 21.45, "learning_rate": 2.140193130565514e-05, "loss": 0.6723, "step": 247680 }, { "epoch": 21.45, "learning_rate": 2.13997629235573e-05, "loss": 0.6377, "step": 247690 }, { "epoch": 21.45, "learning_rate": 2.1397594541459462e-05, "loss": 0.6489, "step": 247700 }, { "epoch": 21.45, "learning_rate": 2.1395426159361625e-05, "loss": 0.5981, "step": 247710 }, { "epoch": 21.45, "learning_rate": 2.1393257777263792e-05, "loss": 0.6538, "step": 247720 }, { "epoch": 21.45, "learning_rate": 2.139108939516595e-05, "loss": 0.5915, "step": 247730 }, { "epoch": 21.46, "learning_rate": 2.1388921013068114e-05, "loss": 0.7227, "step": 247740 }, { "epoch": 21.46, "learning_rate": 2.1386752630970277e-05, "loss": 0.6518, "step": 247750 }, { "epoch": 21.46, "learning_rate": 2.138458424887244e-05, "loss": 0.7161, "step": 247760 }, { "epoch": 21.46, "learning_rate": 2.13824158667746e-05, "loss": 0.6787, "step": 247770 }, { "epoch": 21.46, "learning_rate": 2.1380247484676763e-05, "loss": 0.6503, "step": 247780 }, { "epoch": 21.46, "learning_rate": 2.137807910257893e-05, "loss": 0.7159, "step": 247790 }, { "epoch": 21.46, "learning_rate": 2.1375910720481092e-05, "loss": 0.7377, "step": 247800 }, { "epoch": 21.46, "learning_rate": 2.1373742338383252e-05, "loss": 0.6909, "step": 247810 }, { "epoch": 21.46, "learning_rate": 2.1371573956285415e-05, "loss": 0.7737, "step": 247820 }, { "epoch": 21.46, "learning_rate": 2.1369405574187578e-05, "loss": 0.7911, "step": 247830 }, { "epoch": 21.46, "learning_rate": 2.136723719208974e-05, "loss": 0.6375, "step": 247840 }, { "epoch": 21.47, "learning_rate": 2.13650688099919e-05, "loss": 0.7207, "step": 247850 }, { "epoch": 21.47, "learning_rate": 2.1362900427894067e-05, "loss": 0.667, "step": 247860 }, { "epoch": 21.47, "learning_rate": 2.136073204579623e-05, "loss": 0.6852, "step": 247870 }, { "epoch": 21.47, "learning_rate": 2.135856366369839e-05, "loss": 0.5982, "step": 247880 }, { "epoch": 21.47, "learning_rate": 2.1356395281600553e-05, "loss": 0.7341, "step": 247890 }, { "epoch": 21.47, "learning_rate": 2.1354226899502716e-05, "loss": 0.7195, "step": 247900 }, { "epoch": 21.47, "learning_rate": 2.135205851740488e-05, "loss": 0.7556, "step": 247910 }, { "epoch": 21.47, "learning_rate": 2.1349890135307038e-05, "loss": 0.7583, "step": 247920 }, { "epoch": 21.47, "learning_rate": 2.13477217532092e-05, "loss": 0.5908, "step": 247930 }, { "epoch": 21.47, "learning_rate": 2.1345553371111368e-05, "loss": 0.6938, "step": 247940 }, { "epoch": 21.47, "learning_rate": 2.134338498901353e-05, "loss": 0.7733, "step": 247950 }, { "epoch": 21.47, "learning_rate": 2.134121660691569e-05, "loss": 0.728, "step": 247960 }, { "epoch": 21.48, "learning_rate": 2.1339048224817853e-05, "loss": 0.718, "step": 247970 }, { "epoch": 21.48, "learning_rate": 2.1336879842720016e-05, "loss": 0.6445, "step": 247980 }, { "epoch": 21.48, "learning_rate": 2.133471146062218e-05, "loss": 0.8003, "step": 247990 }, { "epoch": 21.48, "learning_rate": 2.133254307852434e-05, "loss": 0.6856, "step": 248000 }, { "epoch": 21.48, "learning_rate": 2.1330374696426505e-05, "loss": 0.6227, "step": 248010 }, { "epoch": 21.48, "learning_rate": 2.1328206314328668e-05, "loss": 0.7914, "step": 248020 }, { "epoch": 21.48, "learning_rate": 2.132603793223083e-05, "loss": 0.7088, "step": 248030 }, { "epoch": 21.48, "learning_rate": 2.132386955013299e-05, "loss": 0.671, "step": 248040 }, { "epoch": 21.48, "learning_rate": 2.1321701168035154e-05, "loss": 0.6882, "step": 248050 }, { "epoch": 21.48, "learning_rate": 2.1319532785937317e-05, "loss": 0.6896, "step": 248060 }, { "epoch": 21.48, "learning_rate": 2.1317364403839483e-05, "loss": 0.6877, "step": 248070 }, { "epoch": 21.49, "learning_rate": 2.1315196021741643e-05, "loss": 0.7282, "step": 248080 }, { "epoch": 21.49, "learning_rate": 2.1313027639643806e-05, "loss": 0.6808, "step": 248090 }, { "epoch": 21.49, "learning_rate": 2.131085925754597e-05, "loss": 0.7369, "step": 248100 }, { "epoch": 21.49, "learning_rate": 2.130869087544813e-05, "loss": 0.8034, "step": 248110 }, { "epoch": 21.49, "learning_rate": 2.130652249335029e-05, "loss": 0.775, "step": 248120 }, { "epoch": 21.49, "learning_rate": 2.1304354111252455e-05, "loss": 0.7273, "step": 248130 }, { "epoch": 21.49, "learning_rate": 2.130218572915462e-05, "loss": 0.6709, "step": 248140 }, { "epoch": 21.49, "learning_rate": 2.130001734705678e-05, "loss": 0.6826, "step": 248150 }, { "epoch": 21.49, "learning_rate": 2.1297848964958944e-05, "loss": 0.501, "step": 248160 }, { "epoch": 21.49, "learning_rate": 2.1295680582861107e-05, "loss": 0.6883, "step": 248170 }, { "epoch": 21.49, "learning_rate": 2.129351220076327e-05, "loss": 0.8665, "step": 248180 }, { "epoch": 21.49, "learning_rate": 2.129134381866543e-05, "loss": 0.717, "step": 248190 }, { "epoch": 21.5, "learning_rate": 2.1289175436567592e-05, "loss": 0.6745, "step": 248200 }, { "epoch": 21.5, "learning_rate": 2.1287007054469755e-05, "loss": 0.799, "step": 248210 }, { "epoch": 21.5, "learning_rate": 2.128483867237192e-05, "loss": 0.8373, "step": 248220 }, { "epoch": 21.5, "learning_rate": 2.128267029027408e-05, "loss": 0.6762, "step": 248230 }, { "epoch": 21.5, "learning_rate": 2.1280501908176244e-05, "loss": 0.5606, "step": 248240 }, { "epoch": 21.5, "learning_rate": 2.1278333526078407e-05, "loss": 0.8108, "step": 248250 }, { "epoch": 21.5, "learning_rate": 2.127616514398057e-05, "loss": 0.5399, "step": 248260 }, { "epoch": 21.5, "learning_rate": 2.127399676188273e-05, "loss": 0.8143, "step": 248270 }, { "epoch": 21.5, "learning_rate": 2.1271828379784893e-05, "loss": 0.6888, "step": 248280 }, { "epoch": 21.5, "learning_rate": 2.126965999768706e-05, "loss": 0.6906, "step": 248290 }, { "epoch": 21.5, "learning_rate": 2.1267491615589222e-05, "loss": 0.6887, "step": 248300 }, { "epoch": 21.51, "learning_rate": 2.1265323233491382e-05, "loss": 0.7129, "step": 248310 }, { "epoch": 21.51, "learning_rate": 2.1263154851393545e-05, "loss": 0.7349, "step": 248320 }, { "epoch": 21.51, "learning_rate": 2.1260986469295708e-05, "loss": 0.6934, "step": 248330 }, { "epoch": 21.51, "learning_rate": 2.125881808719787e-05, "loss": 0.8265, "step": 248340 }, { "epoch": 21.51, "learning_rate": 2.125664970510003e-05, "loss": 0.6339, "step": 248350 }, { "epoch": 21.51, "learning_rate": 2.1254481323002197e-05, "loss": 0.7475, "step": 248360 }, { "epoch": 21.51, "learning_rate": 2.125231294090436e-05, "loss": 0.6615, "step": 248370 }, { "epoch": 21.51, "learning_rate": 2.125014455880652e-05, "loss": 0.7274, "step": 248380 }, { "epoch": 21.51, "learning_rate": 2.1247976176708682e-05, "loss": 0.778, "step": 248390 }, { "epoch": 21.51, "learning_rate": 2.1245807794610845e-05, "loss": 0.7284, "step": 248400 }, { "epoch": 21.51, "learning_rate": 2.124363941251301e-05, "loss": 0.6662, "step": 248410 }, { "epoch": 21.51, "learning_rate": 2.1241471030415168e-05, "loss": 0.773, "step": 248420 }, { "epoch": 21.52, "learning_rate": 2.1239302648317334e-05, "loss": 0.8211, "step": 248430 }, { "epoch": 21.52, "learning_rate": 2.1237134266219497e-05, "loss": 0.6447, "step": 248440 }, { "epoch": 21.52, "learning_rate": 2.123496588412166e-05, "loss": 0.8337, "step": 248450 }, { "epoch": 21.52, "learning_rate": 2.123279750202382e-05, "loss": 0.6574, "step": 248460 }, { "epoch": 21.52, "learning_rate": 2.1230629119925983e-05, "loss": 0.781, "step": 248470 }, { "epoch": 21.52, "learning_rate": 2.1228460737828146e-05, "loss": 0.7887, "step": 248480 }, { "epoch": 21.52, "learning_rate": 2.122629235573031e-05, "loss": 0.7563, "step": 248490 }, { "epoch": 21.52, "learning_rate": 2.122412397363247e-05, "loss": 0.7703, "step": 248500 }, { "epoch": 21.52, "learning_rate": 2.1221955591534635e-05, "loss": 0.6286, "step": 248510 }, { "epoch": 21.52, "learning_rate": 2.1219787209436798e-05, "loss": 0.7604, "step": 248520 }, { "epoch": 21.52, "learning_rate": 2.121761882733896e-05, "loss": 0.8415, "step": 248530 }, { "epoch": 21.53, "learning_rate": 2.121545044524112e-05, "loss": 0.6534, "step": 248540 }, { "epoch": 21.53, "learning_rate": 2.1213282063143284e-05, "loss": 0.5889, "step": 248550 }, { "epoch": 21.53, "learning_rate": 2.1211113681045447e-05, "loss": 0.7846, "step": 248560 }, { "epoch": 21.53, "learning_rate": 2.1208945298947613e-05, "loss": 0.6046, "step": 248570 }, { "epoch": 21.53, "learning_rate": 2.1206776916849773e-05, "loss": 0.7353, "step": 248580 }, { "epoch": 21.53, "learning_rate": 2.1204608534751936e-05, "loss": 0.6275, "step": 248590 }, { "epoch": 21.53, "learning_rate": 2.12024401526541e-05, "loss": 0.7575, "step": 248600 }, { "epoch": 21.53, "learning_rate": 2.1200271770556262e-05, "loss": 0.6561, "step": 248610 }, { "epoch": 21.53, "learning_rate": 2.119810338845842e-05, "loss": 0.6661, "step": 248620 }, { "epoch": 21.53, "learning_rate": 2.1195935006360584e-05, "loss": 0.8144, "step": 248630 }, { "epoch": 21.53, "learning_rate": 2.119376662426275e-05, "loss": 0.7451, "step": 248640 }, { "epoch": 21.53, "learning_rate": 2.119159824216491e-05, "loss": 0.7191, "step": 248650 }, { "epoch": 21.54, "learning_rate": 2.1189429860067073e-05, "loss": 0.6807, "step": 248660 }, { "epoch": 21.54, "learning_rate": 2.1187261477969236e-05, "loss": 0.6714, "step": 248670 }, { "epoch": 21.54, "learning_rate": 2.11850930958714e-05, "loss": 0.7423, "step": 248680 }, { "epoch": 21.54, "learning_rate": 2.118292471377356e-05, "loss": 0.5493, "step": 248690 }, { "epoch": 21.54, "learning_rate": 2.1180756331675722e-05, "loss": 0.7069, "step": 248700 }, { "epoch": 21.54, "learning_rate": 2.117858794957789e-05, "loss": 0.6395, "step": 248710 }, { "epoch": 21.54, "learning_rate": 2.117641956748005e-05, "loss": 0.694, "step": 248720 }, { "epoch": 21.54, "learning_rate": 2.117425118538221e-05, "loss": 0.6509, "step": 248730 }, { "epoch": 21.54, "learning_rate": 2.1172082803284374e-05, "loss": 0.5897, "step": 248740 }, { "epoch": 21.54, "learning_rate": 2.1169914421186537e-05, "loss": 0.715, "step": 248750 }, { "epoch": 21.54, "learning_rate": 2.11677460390887e-05, "loss": 0.7388, "step": 248760 }, { "epoch": 21.55, "learning_rate": 2.116557765699086e-05, "loss": 0.6853, "step": 248770 }, { "epoch": 21.55, "learning_rate": 2.1163409274893023e-05, "loss": 0.6681, "step": 248780 }, { "epoch": 21.55, "learning_rate": 2.116124089279519e-05, "loss": 0.6865, "step": 248790 }, { "epoch": 21.55, "learning_rate": 2.1159072510697352e-05, "loss": 0.763, "step": 248800 }, { "epoch": 21.55, "learning_rate": 2.115690412859951e-05, "loss": 0.5542, "step": 248810 }, { "epoch": 21.55, "learning_rate": 2.1154735746501675e-05, "loss": 0.7885, "step": 248820 }, { "epoch": 21.55, "learning_rate": 2.1152567364403838e-05, "loss": 0.6513, "step": 248830 }, { "epoch": 21.55, "learning_rate": 2.1150398982306e-05, "loss": 0.851, "step": 248840 }, { "epoch": 21.55, "learning_rate": 2.114823060020816e-05, "loss": 0.8678, "step": 248850 }, { "epoch": 21.55, "learning_rate": 2.1146062218110327e-05, "loss": 0.676, "step": 248860 }, { "epoch": 21.55, "learning_rate": 2.114389383601249e-05, "loss": 0.8002, "step": 248870 }, { "epoch": 21.55, "learning_rate": 2.114172545391465e-05, "loss": 0.6445, "step": 248880 }, { "epoch": 21.56, "learning_rate": 2.1139557071816812e-05, "loss": 0.6595, "step": 248890 }, { "epoch": 21.56, "learning_rate": 2.1137388689718975e-05, "loss": 0.7174, "step": 248900 }, { "epoch": 21.56, "learning_rate": 2.113522030762114e-05, "loss": 0.5114, "step": 248910 }, { "epoch": 21.56, "learning_rate": 2.1133051925523298e-05, "loss": 0.8187, "step": 248920 }, { "epoch": 21.56, "learning_rate": 2.1130883543425464e-05, "loss": 0.7114, "step": 248930 }, { "epoch": 21.56, "learning_rate": 2.1128715161327627e-05, "loss": 0.8355, "step": 248940 }, { "epoch": 21.56, "learning_rate": 2.112654677922979e-05, "loss": 0.6707, "step": 248950 }, { "epoch": 21.56, "learning_rate": 2.112437839713195e-05, "loss": 0.7092, "step": 248960 }, { "epoch": 21.56, "learning_rate": 2.1122210015034113e-05, "loss": 0.7914, "step": 248970 }, { "epoch": 21.56, "learning_rate": 2.1120041632936276e-05, "loss": 0.5717, "step": 248980 }, { "epoch": 21.56, "learning_rate": 2.1117873250838442e-05, "loss": 0.7343, "step": 248990 }, { "epoch": 21.56, "learning_rate": 2.11157048687406e-05, "loss": 0.6964, "step": 249000 }, { "epoch": 21.57, "learning_rate": 2.1113536486642765e-05, "loss": 0.5763, "step": 249010 }, { "epoch": 21.57, "learning_rate": 2.1111368104544928e-05, "loss": 0.6475, "step": 249020 }, { "epoch": 21.57, "learning_rate": 2.110919972244709e-05, "loss": 0.6567, "step": 249030 }, { "epoch": 21.57, "learning_rate": 2.110703134034925e-05, "loss": 0.7524, "step": 249040 }, { "epoch": 21.57, "learning_rate": 2.1104862958251414e-05, "loss": 0.6757, "step": 249050 }, { "epoch": 21.57, "learning_rate": 2.1102694576153577e-05, "loss": 0.7068, "step": 249060 }, { "epoch": 21.57, "learning_rate": 2.1100526194055743e-05, "loss": 0.536, "step": 249070 }, { "epoch": 21.57, "learning_rate": 2.1098357811957903e-05, "loss": 0.707, "step": 249080 }, { "epoch": 21.57, "learning_rate": 2.1096189429860066e-05, "loss": 0.6631, "step": 249090 }, { "epoch": 21.57, "learning_rate": 2.109402104776223e-05, "loss": 0.7112, "step": 249100 }, { "epoch": 21.57, "learning_rate": 2.109185266566439e-05, "loss": 0.7188, "step": 249110 }, { "epoch": 21.58, "learning_rate": 2.108968428356655e-05, "loss": 0.9264, "step": 249120 }, { "epoch": 21.58, "learning_rate": 2.1087515901468714e-05, "loss": 0.7966, "step": 249130 }, { "epoch": 21.58, "learning_rate": 2.108534751937088e-05, "loss": 0.648, "step": 249140 }, { "epoch": 21.58, "learning_rate": 2.108317913727304e-05, "loss": 0.712, "step": 249150 }, { "epoch": 21.58, "learning_rate": 2.1081010755175203e-05, "loss": 0.8828, "step": 249160 }, { "epoch": 21.58, "learning_rate": 2.1078842373077366e-05, "loss": 0.7196, "step": 249170 }, { "epoch": 21.58, "learning_rate": 2.107667399097953e-05, "loss": 0.7177, "step": 249180 }, { "epoch": 21.58, "learning_rate": 2.107450560888169e-05, "loss": 0.7316, "step": 249190 }, { "epoch": 21.58, "learning_rate": 2.1072337226783852e-05, "loss": 0.7247, "step": 249200 }, { "epoch": 21.58, "learning_rate": 2.1070168844686018e-05, "loss": 0.7387, "step": 249210 }, { "epoch": 21.58, "learning_rate": 2.106800046258818e-05, "loss": 0.7687, "step": 249220 }, { "epoch": 21.58, "learning_rate": 2.106583208049034e-05, "loss": 0.6357, "step": 249230 }, { "epoch": 21.59, "learning_rate": 2.1063663698392504e-05, "loss": 0.6588, "step": 249240 }, { "epoch": 21.59, "learning_rate": 2.1061495316294667e-05, "loss": 0.7758, "step": 249250 }, { "epoch": 21.59, "learning_rate": 2.105932693419683e-05, "loss": 0.643, "step": 249260 }, { "epoch": 21.59, "learning_rate": 2.105715855209899e-05, "loss": 0.741, "step": 249270 }, { "epoch": 21.59, "learning_rate": 2.1054990170001153e-05, "loss": 0.5175, "step": 249280 }, { "epoch": 21.59, "learning_rate": 2.105282178790332e-05, "loss": 0.5191, "step": 249290 }, { "epoch": 21.59, "learning_rate": 2.1050653405805482e-05, "loss": 0.6457, "step": 249300 }, { "epoch": 21.59, "learning_rate": 2.104848502370764e-05, "loss": 0.6862, "step": 249310 }, { "epoch": 21.59, "learning_rate": 2.1046316641609805e-05, "loss": 0.6999, "step": 249320 }, { "epoch": 21.59, "learning_rate": 2.1044148259511968e-05, "loss": 0.598, "step": 249330 }, { "epoch": 21.59, "learning_rate": 2.104197987741413e-05, "loss": 0.7402, "step": 249340 }, { "epoch": 21.6, "learning_rate": 2.103981149531629e-05, "loss": 0.7152, "step": 249350 }, { "epoch": 21.6, "learning_rate": 2.1037643113218457e-05, "loss": 0.6647, "step": 249360 }, { "epoch": 21.6, "learning_rate": 2.103547473112062e-05, "loss": 0.7536, "step": 249370 }, { "epoch": 21.6, "learning_rate": 2.1033306349022783e-05, "loss": 0.6972, "step": 249380 }, { "epoch": 21.6, "learning_rate": 2.1031137966924942e-05, "loss": 0.6895, "step": 249390 }, { "epoch": 21.6, "learning_rate": 2.1028969584827105e-05, "loss": 0.6044, "step": 249400 }, { "epoch": 21.6, "learning_rate": 2.1026801202729268e-05, "loss": 0.6891, "step": 249410 }, { "epoch": 21.6, "learning_rate": 2.1024632820631428e-05, "loss": 0.7881, "step": 249420 }, { "epoch": 21.6, "learning_rate": 2.1022464438533594e-05, "loss": 0.7432, "step": 249430 }, { "epoch": 21.6, "learning_rate": 2.1020296056435757e-05, "loss": 0.6223, "step": 249440 }, { "epoch": 21.6, "learning_rate": 2.101812767433792e-05, "loss": 0.639, "step": 249450 }, { "epoch": 21.6, "learning_rate": 2.101595929224008e-05, "loss": 0.7438, "step": 249460 }, { "epoch": 21.61, "learning_rate": 2.1013790910142243e-05, "loss": 0.753, "step": 249470 }, { "epoch": 21.61, "learning_rate": 2.1011622528044406e-05, "loss": 0.707, "step": 249480 }, { "epoch": 21.61, "learning_rate": 2.1009454145946572e-05, "loss": 0.7074, "step": 249490 }, { "epoch": 21.61, "learning_rate": 2.1007285763848732e-05, "loss": 0.6368, "step": 249500 }, { "epoch": 21.61, "learning_rate": 2.1005117381750895e-05, "loss": 0.6347, "step": 249510 }, { "epoch": 21.61, "learning_rate": 2.1002948999653058e-05, "loss": 0.6861, "step": 249520 }, { "epoch": 21.61, "learning_rate": 2.100078061755522e-05, "loss": 0.7614, "step": 249530 }, { "epoch": 21.61, "learning_rate": 2.099861223545738e-05, "loss": 0.7524, "step": 249540 }, { "epoch": 21.61, "learning_rate": 2.0996443853359543e-05, "loss": 0.7415, "step": 249550 }, { "epoch": 21.61, "learning_rate": 2.0994275471261706e-05, "loss": 0.6314, "step": 249560 }, { "epoch": 21.61, "learning_rate": 2.0992107089163873e-05, "loss": 0.79, "step": 249570 }, { "epoch": 21.62, "learning_rate": 2.0989938707066032e-05, "loss": 0.6753, "step": 249580 }, { "epoch": 21.62, "learning_rate": 2.0987770324968195e-05, "loss": 0.7562, "step": 249590 }, { "epoch": 21.62, "learning_rate": 2.098560194287036e-05, "loss": 0.5677, "step": 249600 }, { "epoch": 21.62, "learning_rate": 2.098343356077252e-05, "loss": 0.7772, "step": 249610 }, { "epoch": 21.62, "learning_rate": 2.098126517867468e-05, "loss": 0.7058, "step": 249620 }, { "epoch": 21.62, "learning_rate": 2.0979096796576844e-05, "loss": 0.6577, "step": 249630 }, { "epoch": 21.62, "learning_rate": 2.097692841447901e-05, "loss": 0.7927, "step": 249640 }, { "epoch": 21.62, "learning_rate": 2.097476003238117e-05, "loss": 0.7799, "step": 249650 }, { "epoch": 21.62, "learning_rate": 2.0972591650283333e-05, "loss": 0.787, "step": 249660 }, { "epoch": 21.62, "learning_rate": 2.0970423268185496e-05, "loss": 0.6443, "step": 249670 }, { "epoch": 21.62, "learning_rate": 2.096825488608766e-05, "loss": 0.5985, "step": 249680 }, { "epoch": 21.62, "learning_rate": 2.096608650398982e-05, "loss": 0.6611, "step": 249690 }, { "epoch": 21.63, "learning_rate": 2.0963918121891982e-05, "loss": 0.6643, "step": 249700 }, { "epoch": 21.63, "learning_rate": 2.0961749739794148e-05, "loss": 0.6403, "step": 249710 }, { "epoch": 21.63, "learning_rate": 2.095958135769631e-05, "loss": 0.7808, "step": 249720 }, { "epoch": 21.63, "learning_rate": 2.095741297559847e-05, "loss": 0.7526, "step": 249730 }, { "epoch": 21.63, "learning_rate": 2.0955244593500634e-05, "loss": 0.6728, "step": 249740 }, { "epoch": 21.63, "learning_rate": 2.0953076211402797e-05, "loss": 0.6927, "step": 249750 }, { "epoch": 21.63, "learning_rate": 2.095090782930496e-05, "loss": 0.6408, "step": 249760 }, { "epoch": 21.63, "learning_rate": 2.094873944720712e-05, "loss": 0.648, "step": 249770 }, { "epoch": 21.63, "learning_rate": 2.0946571065109286e-05, "loss": 0.8619, "step": 249780 }, { "epoch": 21.63, "learning_rate": 2.094440268301145e-05, "loss": 0.6388, "step": 249790 }, { "epoch": 21.63, "learning_rate": 2.0942234300913612e-05, "loss": 0.6791, "step": 249800 }, { "epoch": 21.64, "learning_rate": 2.094006591881577e-05, "loss": 0.5982, "step": 249810 }, { "epoch": 21.64, "learning_rate": 2.0937897536717934e-05, "loss": 0.7078, "step": 249820 }, { "epoch": 21.64, "learning_rate": 2.0935729154620097e-05, "loss": 0.6987, "step": 249830 }, { "epoch": 21.64, "learning_rate": 2.093356077252226e-05, "loss": 0.7666, "step": 249840 }, { "epoch": 21.64, "learning_rate": 2.093139239042442e-05, "loss": 0.7572, "step": 249850 }, { "epoch": 21.64, "learning_rate": 2.0929224008326586e-05, "loss": 0.8374, "step": 249860 }, { "epoch": 21.64, "learning_rate": 2.092705562622875e-05, "loss": 0.6671, "step": 249870 }, { "epoch": 21.64, "learning_rate": 2.0924887244130912e-05, "loss": 0.7236, "step": 249880 }, { "epoch": 21.64, "learning_rate": 2.0922718862033072e-05, "loss": 0.7782, "step": 249890 }, { "epoch": 21.64, "learning_rate": 2.0920550479935235e-05, "loss": 0.7842, "step": 249900 }, { "epoch": 21.64, "learning_rate": 2.0918382097837398e-05, "loss": 0.7613, "step": 249910 }, { "epoch": 21.64, "learning_rate": 2.0916213715739558e-05, "loss": 0.8232, "step": 249920 }, { "epoch": 21.65, "learning_rate": 2.0914045333641724e-05, "loss": 0.8659, "step": 249930 }, { "epoch": 21.65, "learning_rate": 2.0911876951543887e-05, "loss": 0.7652, "step": 249940 }, { "epoch": 21.65, "learning_rate": 2.090970856944605e-05, "loss": 0.6356, "step": 249950 }, { "epoch": 21.65, "learning_rate": 2.090754018734821e-05, "loss": 0.6834, "step": 249960 }, { "epoch": 21.65, "learning_rate": 2.0905371805250373e-05, "loss": 0.697, "step": 249970 }, { "epoch": 21.65, "learning_rate": 2.0903203423152536e-05, "loss": 0.6619, "step": 249980 }, { "epoch": 21.65, "learning_rate": 2.0901035041054702e-05, "loss": 0.7023, "step": 249990 }, { "epoch": 21.65, "learning_rate": 2.089886665895686e-05, "loss": 0.7779, "step": 250000 }, { "epoch": 21.65, "learning_rate": 2.0896698276859025e-05, "loss": 0.8157, "step": 250010 }, { "epoch": 21.65, "learning_rate": 2.0894529894761188e-05, "loss": 0.8576, "step": 250020 }, { "epoch": 21.65, "learning_rate": 2.089236151266335e-05, "loss": 0.6532, "step": 250030 }, { "epoch": 21.66, "learning_rate": 2.089019313056551e-05, "loss": 0.7738, "step": 250040 }, { "epoch": 21.66, "learning_rate": 2.0888024748467673e-05, "loss": 0.7573, "step": 250050 }, { "epoch": 21.66, "learning_rate": 2.088585636636984e-05, "loss": 0.5618, "step": 250060 }, { "epoch": 21.66, "learning_rate": 2.0883687984272003e-05, "loss": 0.6346, "step": 250070 }, { "epoch": 21.66, "learning_rate": 2.0881519602174162e-05, "loss": 0.6808, "step": 250080 }, { "epoch": 21.66, "learning_rate": 2.0879351220076325e-05, "loss": 0.6954, "step": 250090 }, { "epoch": 21.66, "learning_rate": 2.087718283797849e-05, "loss": 0.701, "step": 250100 }, { "epoch": 21.66, "learning_rate": 2.087501445588065e-05, "loss": 0.8628, "step": 250110 }, { "epoch": 21.66, "learning_rate": 2.087284607378281e-05, "loss": 0.7094, "step": 250120 }, { "epoch": 21.66, "learning_rate": 2.0870677691684974e-05, "loss": 0.7463, "step": 250130 }, { "epoch": 21.66, "learning_rate": 2.086850930958714e-05, "loss": 0.6567, "step": 250140 }, { "epoch": 21.66, "learning_rate": 2.0866340927489303e-05, "loss": 0.7563, "step": 250150 }, { "epoch": 21.67, "learning_rate": 2.0864172545391463e-05, "loss": 0.5931, "step": 250160 }, { "epoch": 21.67, "learning_rate": 2.0862004163293626e-05, "loss": 0.6268, "step": 250170 }, { "epoch": 21.67, "learning_rate": 2.085983578119579e-05, "loss": 0.6519, "step": 250180 }, { "epoch": 21.67, "learning_rate": 2.085766739909795e-05, "loss": 0.5684, "step": 250190 }, { "epoch": 21.67, "learning_rate": 2.085549901700011e-05, "loss": 0.6841, "step": 250200 }, { "epoch": 21.67, "learning_rate": 2.0853330634902278e-05, "loss": 0.6606, "step": 250210 }, { "epoch": 21.67, "learning_rate": 2.085116225280444e-05, "loss": 0.7103, "step": 250220 }, { "epoch": 21.67, "learning_rate": 2.08489938707066e-05, "loss": 0.6521, "step": 250230 }, { "epoch": 21.67, "learning_rate": 2.0846825488608764e-05, "loss": 0.7131, "step": 250240 }, { "epoch": 21.67, "learning_rate": 2.0844657106510927e-05, "loss": 0.7998, "step": 250250 }, { "epoch": 21.67, "learning_rate": 2.084248872441309e-05, "loss": 0.7765, "step": 250260 }, { "epoch": 21.67, "learning_rate": 2.084032034231525e-05, "loss": 0.899, "step": 250270 }, { "epoch": 21.68, "learning_rate": 2.0838151960217416e-05, "loss": 0.8123, "step": 250280 }, { "epoch": 21.68, "learning_rate": 2.083598357811958e-05, "loss": 0.6952, "step": 250290 }, { "epoch": 21.68, "learning_rate": 2.083381519602174e-05, "loss": 0.7266, "step": 250300 }, { "epoch": 21.68, "learning_rate": 2.08316468139239e-05, "loss": 0.7358, "step": 250310 }, { "epoch": 21.68, "learning_rate": 2.0829478431826064e-05, "loss": 0.7796, "step": 250320 }, { "epoch": 21.68, "learning_rate": 2.0827310049728227e-05, "loss": 0.7475, "step": 250330 }, { "epoch": 21.68, "learning_rate": 2.0825141667630394e-05, "loss": 0.6869, "step": 250340 }, { "epoch": 21.68, "learning_rate": 2.082297328553255e-05, "loss": 0.6976, "step": 250350 }, { "epoch": 21.68, "learning_rate": 2.0820804903434716e-05, "loss": 0.6922, "step": 250360 }, { "epoch": 21.68, "learning_rate": 2.081863652133688e-05, "loss": 0.7447, "step": 250370 }, { "epoch": 21.68, "learning_rate": 2.0816468139239042e-05, "loss": 0.7262, "step": 250380 }, { "epoch": 21.69, "learning_rate": 2.0814299757141202e-05, "loss": 0.7753, "step": 250390 }, { "epoch": 21.69, "learning_rate": 2.0812131375043365e-05, "loss": 0.5672, "step": 250400 }, { "epoch": 21.69, "learning_rate": 2.0809962992945528e-05, "loss": 0.6985, "step": 250410 }, { "epoch": 21.69, "learning_rate": 2.0807794610847687e-05, "loss": 0.6835, "step": 250420 }, { "epoch": 21.69, "learning_rate": 2.0805626228749854e-05, "loss": 0.8088, "step": 250430 }, { "epoch": 21.69, "learning_rate": 2.0803457846652017e-05, "loss": 0.6383, "step": 250440 }, { "epoch": 21.69, "learning_rate": 2.080128946455418e-05, "loss": 0.7399, "step": 250450 }, { "epoch": 21.69, "learning_rate": 2.079912108245634e-05, "loss": 0.8284, "step": 250460 }, { "epoch": 21.69, "learning_rate": 2.0796952700358503e-05, "loss": 0.7196, "step": 250470 }, { "epoch": 21.69, "learning_rate": 2.0794784318260666e-05, "loss": 0.5248, "step": 250480 }, { "epoch": 21.69, "learning_rate": 2.0792615936162832e-05, "loss": 0.643, "step": 250490 }, { "epoch": 21.69, "learning_rate": 2.079044755406499e-05, "loss": 0.8093, "step": 250500 }, { "epoch": 21.7, "learning_rate": 2.0788279171967155e-05, "loss": 0.7166, "step": 250510 }, { "epoch": 21.7, "learning_rate": 2.0786110789869318e-05, "loss": 0.7529, "step": 250520 }, { "epoch": 21.7, "learning_rate": 2.078394240777148e-05, "loss": 0.7931, "step": 250530 }, { "epoch": 21.7, "learning_rate": 2.078177402567364e-05, "loss": 0.7109, "step": 250540 }, { "epoch": 21.7, "learning_rate": 2.0779605643575803e-05, "loss": 0.7295, "step": 250550 }, { "epoch": 21.7, "learning_rate": 2.077743726147797e-05, "loss": 0.7627, "step": 250560 }, { "epoch": 21.7, "learning_rate": 2.0775268879380133e-05, "loss": 0.712, "step": 250570 }, { "epoch": 21.7, "learning_rate": 2.0773100497282292e-05, "loss": 0.8177, "step": 250580 }, { "epoch": 21.7, "learning_rate": 2.0770932115184455e-05, "loss": 0.6394, "step": 250590 }, { "epoch": 21.7, "learning_rate": 2.0768763733086618e-05, "loss": 0.7679, "step": 250600 }, { "epoch": 21.7, "learning_rate": 2.076659535098878e-05, "loss": 0.7666, "step": 250610 }, { "epoch": 21.71, "learning_rate": 2.076442696889094e-05, "loss": 0.8104, "step": 250620 }, { "epoch": 21.71, "learning_rate": 2.0762258586793104e-05, "loss": 0.6534, "step": 250630 }, { "epoch": 21.71, "learning_rate": 2.076009020469527e-05, "loss": 0.6537, "step": 250640 }, { "epoch": 21.71, "learning_rate": 2.0757921822597433e-05, "loss": 0.7492, "step": 250650 }, { "epoch": 21.71, "learning_rate": 2.0755753440499593e-05, "loss": 0.6913, "step": 250660 }, { "epoch": 21.71, "learning_rate": 2.0753585058401756e-05, "loss": 0.7194, "step": 250670 }, { "epoch": 21.71, "learning_rate": 2.075141667630392e-05, "loss": 0.6916, "step": 250680 }, { "epoch": 21.71, "learning_rate": 2.074924829420608e-05, "loss": 0.5627, "step": 250690 }, { "epoch": 21.71, "learning_rate": 2.074707991210824e-05, "loss": 0.7538, "step": 250700 }, { "epoch": 21.71, "learning_rate": 2.0744911530010408e-05, "loss": 0.7206, "step": 250710 }, { "epoch": 21.71, "learning_rate": 2.074274314791257e-05, "loss": 0.7849, "step": 250720 }, { "epoch": 21.71, "learning_rate": 2.074057476581473e-05, "loss": 0.7262, "step": 250730 }, { "epoch": 21.72, "learning_rate": 2.0738406383716893e-05, "loss": 0.6288, "step": 250740 }, { "epoch": 21.72, "learning_rate": 2.0736238001619056e-05, "loss": 0.6075, "step": 250750 }, { "epoch": 21.72, "learning_rate": 2.073406961952122e-05, "loss": 0.8991, "step": 250760 }, { "epoch": 21.72, "learning_rate": 2.073190123742338e-05, "loss": 0.7163, "step": 250770 }, { "epoch": 21.72, "learning_rate": 2.0729732855325545e-05, "loss": 0.7196, "step": 250780 }, { "epoch": 21.72, "learning_rate": 2.072756447322771e-05, "loss": 0.6217, "step": 250790 }, { "epoch": 21.72, "learning_rate": 2.072539609112987e-05, "loss": 0.6404, "step": 250800 }, { "epoch": 21.72, "learning_rate": 2.072322770903203e-05, "loss": 0.5955, "step": 250810 }, { "epoch": 21.72, "learning_rate": 2.0721059326934194e-05, "loss": 0.6108, "step": 250820 }, { "epoch": 21.72, "learning_rate": 2.0718890944836357e-05, "loss": 0.5467, "step": 250830 }, { "epoch": 21.72, "learning_rate": 2.0716722562738523e-05, "loss": 0.6812, "step": 250840 }, { "epoch": 21.73, "learning_rate": 2.0714554180640683e-05, "loss": 0.7097, "step": 250850 }, { "epoch": 21.73, "learning_rate": 2.0712385798542846e-05, "loss": 0.6519, "step": 250860 }, { "epoch": 21.73, "learning_rate": 2.071021741644501e-05, "loss": 0.5751, "step": 250870 }, { "epoch": 21.73, "learning_rate": 2.0708049034347172e-05, "loss": 0.7189, "step": 250880 }, { "epoch": 21.73, "learning_rate": 2.0705880652249332e-05, "loss": 0.665, "step": 250890 }, { "epoch": 21.73, "learning_rate": 2.0703712270151495e-05, "loss": 0.6742, "step": 250900 }, { "epoch": 21.73, "learning_rate": 2.0701543888053658e-05, "loss": 0.6212, "step": 250910 }, { "epoch": 21.73, "learning_rate": 2.0699375505955824e-05, "loss": 0.7309, "step": 250920 }, { "epoch": 21.73, "learning_rate": 2.0697207123857984e-05, "loss": 0.6869, "step": 250930 }, { "epoch": 21.73, "learning_rate": 2.0695038741760147e-05, "loss": 0.7947, "step": 250940 }, { "epoch": 21.73, "learning_rate": 2.069287035966231e-05, "loss": 0.8028, "step": 250950 }, { "epoch": 21.73, "learning_rate": 2.069070197756447e-05, "loss": 0.6625, "step": 250960 }, { "epoch": 21.74, "learning_rate": 2.0688533595466632e-05, "loss": 0.7176, "step": 250970 }, { "epoch": 21.74, "learning_rate": 2.0686365213368795e-05, "loss": 0.7287, "step": 250980 }, { "epoch": 21.74, "learning_rate": 2.0684196831270962e-05, "loss": 0.6695, "step": 250990 }, { "epoch": 21.74, "learning_rate": 2.068202844917312e-05, "loss": 0.7501, "step": 251000 }, { "epoch": 21.74, "learning_rate": 2.0679860067075284e-05, "loss": 0.7002, "step": 251010 }, { "epoch": 21.74, "learning_rate": 2.0677691684977447e-05, "loss": 0.6615, "step": 251020 }, { "epoch": 21.74, "learning_rate": 2.067552330287961e-05, "loss": 0.7198, "step": 251030 }, { "epoch": 21.74, "learning_rate": 2.067335492078177e-05, "loss": 0.7368, "step": 251040 }, { "epoch": 21.74, "learning_rate": 2.0671186538683933e-05, "loss": 0.6298, "step": 251050 }, { "epoch": 21.74, "learning_rate": 2.06690181565861e-05, "loss": 0.8262, "step": 251060 }, { "epoch": 21.74, "learning_rate": 2.0666849774488262e-05, "loss": 0.8463, "step": 251070 }, { "epoch": 21.75, "learning_rate": 2.0664681392390422e-05, "loss": 0.9584, "step": 251080 }, { "epoch": 21.75, "learning_rate": 2.0662513010292585e-05, "loss": 0.6941, "step": 251090 }, { "epoch": 21.75, "learning_rate": 2.0660344628194748e-05, "loss": 0.6586, "step": 251100 }, { "epoch": 21.75, "learning_rate": 2.065817624609691e-05, "loss": 0.6473, "step": 251110 }, { "epoch": 21.75, "learning_rate": 2.065600786399907e-05, "loss": 0.6001, "step": 251120 }, { "epoch": 21.75, "learning_rate": 2.0653839481901237e-05, "loss": 0.9152, "step": 251130 }, { "epoch": 21.75, "learning_rate": 2.06516710998034e-05, "loss": 0.6514, "step": 251140 }, { "epoch": 21.75, "learning_rate": 2.0649502717705563e-05, "loss": 0.7235, "step": 251150 }, { "epoch": 21.75, "learning_rate": 2.0647334335607723e-05, "loss": 0.7346, "step": 251160 }, { "epoch": 21.75, "learning_rate": 2.0645165953509886e-05, "loss": 0.8248, "step": 251170 }, { "epoch": 21.75, "learning_rate": 2.064299757141205e-05, "loss": 0.7308, "step": 251180 }, { "epoch": 21.75, "learning_rate": 2.0640829189314208e-05, "loss": 0.6281, "step": 251190 }, { "epoch": 21.76, "learning_rate": 2.063866080721637e-05, "loss": 0.6234, "step": 251200 }, { "epoch": 21.76, "learning_rate": 2.0636492425118538e-05, "loss": 0.7906, "step": 251210 }, { "epoch": 21.76, "learning_rate": 2.06343240430207e-05, "loss": 0.796, "step": 251220 }, { "epoch": 21.76, "learning_rate": 2.063215566092286e-05, "loss": 0.7016, "step": 251230 }, { "epoch": 21.76, "learning_rate": 2.0629987278825023e-05, "loss": 0.6677, "step": 251240 }, { "epoch": 21.76, "learning_rate": 2.0627818896727186e-05, "loss": 0.9247, "step": 251250 }, { "epoch": 21.76, "learning_rate": 2.062565051462935e-05, "loss": 0.8276, "step": 251260 }, { "epoch": 21.76, "learning_rate": 2.062348213253151e-05, "loss": 0.5466, "step": 251270 }, { "epoch": 21.76, "learning_rate": 2.0621313750433675e-05, "loss": 0.6894, "step": 251280 }, { "epoch": 21.76, "learning_rate": 2.0619145368335838e-05, "loss": 0.6533, "step": 251290 }, { "epoch": 21.76, "learning_rate": 2.0616976986238e-05, "loss": 0.7818, "step": 251300 }, { "epoch": 21.77, "learning_rate": 2.061480860414016e-05, "loss": 0.682, "step": 251310 }, { "epoch": 21.77, "learning_rate": 2.0612640222042324e-05, "loss": 0.8047, "step": 251320 }, { "epoch": 21.77, "learning_rate": 2.0610471839944487e-05, "loss": 0.684, "step": 251330 }, { "epoch": 21.77, "learning_rate": 2.0608303457846653e-05, "loss": 0.6597, "step": 251340 }, { "epoch": 21.77, "learning_rate": 2.0606135075748813e-05, "loss": 0.7288, "step": 251350 }, { "epoch": 21.77, "learning_rate": 2.0603966693650976e-05, "loss": 0.8153, "step": 251360 }, { "epoch": 21.77, "learning_rate": 2.060179831155314e-05, "loss": 0.6416, "step": 251370 }, { "epoch": 21.77, "learning_rate": 2.0599629929455302e-05, "loss": 0.6447, "step": 251380 }, { "epoch": 21.77, "learning_rate": 2.059746154735746e-05, "loss": 0.7385, "step": 251390 }, { "epoch": 21.77, "learning_rate": 2.0595293165259625e-05, "loss": 0.6231, "step": 251400 }, { "epoch": 21.77, "learning_rate": 2.059312478316179e-05, "loss": 0.6555, "step": 251410 }, { "epoch": 21.77, "learning_rate": 2.0590956401063954e-05, "loss": 0.574, "step": 251420 }, { "epoch": 21.78, "learning_rate": 2.0588788018966114e-05, "loss": 0.6322, "step": 251430 }, { "epoch": 21.78, "learning_rate": 2.0586619636868277e-05, "loss": 0.7258, "step": 251440 }, { "epoch": 21.78, "learning_rate": 2.058445125477044e-05, "loss": 0.8067, "step": 251450 }, { "epoch": 21.78, "learning_rate": 2.05822828726726e-05, "loss": 0.7383, "step": 251460 }, { "epoch": 21.78, "learning_rate": 2.0580114490574762e-05, "loss": 0.6975, "step": 251470 }, { "epoch": 21.78, "learning_rate": 2.0577946108476925e-05, "loss": 0.6763, "step": 251480 }, { "epoch": 21.78, "learning_rate": 2.057577772637909e-05, "loss": 0.7481, "step": 251490 }, { "epoch": 21.78, "learning_rate": 2.057360934428125e-05, "loss": 0.6364, "step": 251500 }, { "epoch": 21.78, "learning_rate": 2.0571440962183414e-05, "loss": 0.6363, "step": 251510 }, { "epoch": 21.78, "learning_rate": 2.0569272580085577e-05, "loss": 0.781, "step": 251520 }, { "epoch": 21.78, "learning_rate": 2.056710419798774e-05, "loss": 0.7399, "step": 251530 }, { "epoch": 21.78, "learning_rate": 2.05649358158899e-05, "loss": 0.7101, "step": 251540 }, { "epoch": 21.79, "learning_rate": 2.0562767433792063e-05, "loss": 0.6494, "step": 251550 }, { "epoch": 21.79, "learning_rate": 2.056059905169423e-05, "loss": 0.7536, "step": 251560 }, { "epoch": 21.79, "learning_rate": 2.0558430669596392e-05, "loss": 0.6788, "step": 251570 }, { "epoch": 21.79, "learning_rate": 2.0556262287498552e-05, "loss": 0.6647, "step": 251580 }, { "epoch": 21.79, "learning_rate": 2.0554093905400715e-05, "loss": 0.5708, "step": 251590 }, { "epoch": 21.79, "learning_rate": 2.0551925523302878e-05, "loss": 0.7681, "step": 251600 }, { "epoch": 21.79, "learning_rate": 2.054975714120504e-05, "loss": 0.8005, "step": 251610 }, { "epoch": 21.79, "learning_rate": 2.05475887591072e-05, "loss": 0.6485, "step": 251620 }, { "epoch": 21.79, "learning_rate": 2.0545420377009367e-05, "loss": 0.721, "step": 251630 }, { "epoch": 21.79, "learning_rate": 2.054325199491153e-05, "loss": 0.7331, "step": 251640 }, { "epoch": 21.79, "learning_rate": 2.0541083612813693e-05, "loss": 0.639, "step": 251650 }, { "epoch": 21.8, "learning_rate": 2.0538915230715852e-05, "loss": 0.883, "step": 251660 }, { "epoch": 21.8, "learning_rate": 2.0536746848618015e-05, "loss": 0.7597, "step": 251670 }, { "epoch": 21.8, "learning_rate": 2.053457846652018e-05, "loss": 0.6604, "step": 251680 }, { "epoch": 21.8, "learning_rate": 2.0532410084422345e-05, "loss": 0.7265, "step": 251690 }, { "epoch": 21.8, "learning_rate": 2.05302417023245e-05, "loss": 0.7869, "step": 251700 }, { "epoch": 21.8, "learning_rate": 2.0528073320226668e-05, "loss": 0.8379, "step": 251710 }, { "epoch": 21.8, "learning_rate": 2.052590493812883e-05, "loss": 0.6709, "step": 251720 }, { "epoch": 21.8, "learning_rate": 2.052373655603099e-05, "loss": 0.8566, "step": 251730 }, { "epoch": 21.8, "learning_rate": 2.0521568173933153e-05, "loss": 0.6982, "step": 251740 }, { "epoch": 21.8, "learning_rate": 2.0519399791835316e-05, "loss": 0.6478, "step": 251750 }, { "epoch": 21.8, "learning_rate": 2.051723140973748e-05, "loss": 0.5783, "step": 251760 }, { "epoch": 21.8, "learning_rate": 2.051506302763964e-05, "loss": 0.7912, "step": 251770 }, { "epoch": 21.81, "learning_rate": 2.0512894645541805e-05, "loss": 0.7258, "step": 251780 }, { "epoch": 21.81, "learning_rate": 2.0510726263443968e-05, "loss": 0.6537, "step": 251790 }, { "epoch": 21.81, "learning_rate": 2.050855788134613e-05, "loss": 0.6104, "step": 251800 }, { "epoch": 21.81, "learning_rate": 2.050638949924829e-05, "loss": 0.7579, "step": 251810 }, { "epoch": 21.81, "learning_rate": 2.0504221117150454e-05, "loss": 0.6593, "step": 251820 }, { "epoch": 21.81, "learning_rate": 2.0502052735052617e-05, "loss": 0.6622, "step": 251830 }, { "epoch": 21.81, "learning_rate": 2.0499884352954783e-05, "loss": 0.6223, "step": 251840 }, { "epoch": 21.81, "learning_rate": 2.0497715970856943e-05, "loss": 0.8166, "step": 251850 }, { "epoch": 21.81, "learning_rate": 2.0495547588759106e-05, "loss": 0.7391, "step": 251860 }, { "epoch": 21.81, "learning_rate": 2.049337920666127e-05, "loss": 0.6784, "step": 251870 }, { "epoch": 21.81, "learning_rate": 2.0491210824563432e-05, "loss": 0.7126, "step": 251880 }, { "epoch": 21.82, "learning_rate": 2.048904244246559e-05, "loss": 0.7197, "step": 251890 }, { "epoch": 21.82, "learning_rate": 2.0486874060367754e-05, "loss": 0.8133, "step": 251900 }, { "epoch": 21.82, "learning_rate": 2.048470567826992e-05, "loss": 0.5797, "step": 251910 }, { "epoch": 21.82, "learning_rate": 2.0482537296172084e-05, "loss": 0.9094, "step": 251920 }, { "epoch": 21.82, "learning_rate": 2.0480368914074243e-05, "loss": 0.6546, "step": 251930 }, { "epoch": 21.82, "learning_rate": 2.0478200531976406e-05, "loss": 0.6028, "step": 251940 }, { "epoch": 21.82, "learning_rate": 2.047603214987857e-05, "loss": 0.7302, "step": 251950 }, { "epoch": 21.82, "learning_rate": 2.047386376778073e-05, "loss": 0.8103, "step": 251960 }, { "epoch": 21.82, "learning_rate": 2.0471695385682892e-05, "loss": 0.8247, "step": 251970 }, { "epoch": 21.82, "learning_rate": 2.0469527003585055e-05, "loss": 0.8124, "step": 251980 }, { "epoch": 21.82, "learning_rate": 2.046735862148722e-05, "loss": 0.8714, "step": 251990 }, { "epoch": 21.82, "learning_rate": 2.046519023938938e-05, "loss": 0.6294, "step": 252000 }, { "epoch": 21.83, "learning_rate": 2.0463021857291544e-05, "loss": 0.6672, "step": 252010 }, { "epoch": 21.83, "learning_rate": 2.0460853475193707e-05, "loss": 0.6002, "step": 252020 }, { "epoch": 21.83, "learning_rate": 2.045868509309587e-05, "loss": 0.6903, "step": 252030 }, { "epoch": 21.83, "learning_rate": 2.045651671099803e-05, "loss": 0.6176, "step": 252040 }, { "epoch": 21.83, "learning_rate": 2.0454348328900193e-05, "loss": 0.6779, "step": 252050 }, { "epoch": 21.83, "learning_rate": 2.045217994680236e-05, "loss": 0.6573, "step": 252060 }, { "epoch": 21.83, "learning_rate": 2.0450011564704522e-05, "loss": 0.772, "step": 252070 }, { "epoch": 21.83, "learning_rate": 2.0447843182606682e-05, "loss": 0.8024, "step": 252080 }, { "epoch": 21.83, "learning_rate": 2.0445674800508845e-05, "loss": 0.6631, "step": 252090 }, { "epoch": 21.83, "learning_rate": 2.0443506418411008e-05, "loss": 0.8544, "step": 252100 }, { "epoch": 21.83, "learning_rate": 2.044133803631317e-05, "loss": 0.7212, "step": 252110 }, { "epoch": 21.84, "learning_rate": 2.043916965421533e-05, "loss": 0.7584, "step": 252120 }, { "epoch": 21.84, "learning_rate": 2.0437001272117497e-05, "loss": 0.7896, "step": 252130 }, { "epoch": 21.84, "learning_rate": 2.043483289001966e-05, "loss": 0.6686, "step": 252140 }, { "epoch": 21.84, "learning_rate": 2.0432664507921823e-05, "loss": 0.7212, "step": 252150 }, { "epoch": 21.84, "learning_rate": 2.0430496125823982e-05, "loss": 0.7359, "step": 252160 }, { "epoch": 21.84, "learning_rate": 2.0428327743726145e-05, "loss": 0.6836, "step": 252170 }, { "epoch": 21.84, "learning_rate": 2.042615936162831e-05, "loss": 0.599, "step": 252180 }, { "epoch": 21.84, "learning_rate": 2.0423990979530475e-05, "loss": 0.9306, "step": 252190 }, { "epoch": 21.84, "learning_rate": 2.0421822597432634e-05, "loss": 0.6336, "step": 252200 }, { "epoch": 21.84, "learning_rate": 2.0419654215334797e-05, "loss": 0.7658, "step": 252210 }, { "epoch": 21.84, "learning_rate": 2.041748583323696e-05, "loss": 0.7744, "step": 252220 }, { "epoch": 21.84, "learning_rate": 2.041531745113912e-05, "loss": 0.9396, "step": 252230 }, { "epoch": 21.85, "learning_rate": 2.0413149069041283e-05, "loss": 0.7592, "step": 252240 }, { "epoch": 21.85, "learning_rate": 2.0410980686943446e-05, "loss": 0.6415, "step": 252250 }, { "epoch": 21.85, "learning_rate": 2.040881230484561e-05, "loss": 0.7243, "step": 252260 }, { "epoch": 21.85, "learning_rate": 2.040664392274777e-05, "loss": 0.5664, "step": 252270 }, { "epoch": 21.85, "learning_rate": 2.0404475540649935e-05, "loss": 0.9081, "step": 252280 }, { "epoch": 21.85, "learning_rate": 2.0402307158552098e-05, "loss": 0.673, "step": 252290 }, { "epoch": 21.85, "learning_rate": 2.040013877645426e-05, "loss": 0.7517, "step": 252300 }, { "epoch": 21.85, "learning_rate": 2.039797039435642e-05, "loss": 0.7103, "step": 252310 }, { "epoch": 21.85, "learning_rate": 2.0395802012258584e-05, "loss": 0.6311, "step": 252320 }, { "epoch": 21.85, "learning_rate": 2.0393633630160747e-05, "loss": 0.741, "step": 252330 }, { "epoch": 21.85, "learning_rate": 2.0391465248062913e-05, "loss": 0.7726, "step": 252340 }, { "epoch": 21.86, "learning_rate": 2.0389296865965073e-05, "loss": 0.6756, "step": 252350 }, { "epoch": 21.86, "learning_rate": 2.0387128483867236e-05, "loss": 0.6627, "step": 252360 }, { "epoch": 21.86, "learning_rate": 2.03849601017694e-05, "loss": 0.6683, "step": 252370 }, { "epoch": 21.86, "learning_rate": 2.038279171967156e-05, "loss": 0.6949, "step": 252380 }, { "epoch": 21.86, "learning_rate": 2.038062333757372e-05, "loss": 0.8082, "step": 252390 }, { "epoch": 21.86, "learning_rate": 2.0378454955475884e-05, "loss": 0.7785, "step": 252400 }, { "epoch": 21.86, "learning_rate": 2.037628657337805e-05, "loss": 0.7206, "step": 252410 }, { "epoch": 21.86, "learning_rate": 2.0374118191280214e-05, "loss": 0.681, "step": 252420 }, { "epoch": 21.86, "learning_rate": 2.0371949809182373e-05, "loss": 0.6295, "step": 252430 }, { "epoch": 21.86, "learning_rate": 2.0369781427084536e-05, "loss": 0.7583, "step": 252440 }, { "epoch": 21.86, "learning_rate": 2.03676130449867e-05, "loss": 0.7483, "step": 252450 }, { "epoch": 21.86, "learning_rate": 2.0365444662888862e-05, "loss": 0.6664, "step": 252460 }, { "epoch": 21.87, "learning_rate": 2.0363276280791022e-05, "loss": 0.6422, "step": 252470 }, { "epoch": 21.87, "learning_rate": 2.0361107898693188e-05, "loss": 0.7036, "step": 252480 }, { "epoch": 21.87, "learning_rate": 2.035893951659535e-05, "loss": 0.7014, "step": 252490 }, { "epoch": 21.87, "learning_rate": 2.035677113449751e-05, "loss": 0.6648, "step": 252500 }, { "epoch": 21.87, "learning_rate": 2.0354602752399674e-05, "loss": 0.479, "step": 252510 }, { "epoch": 21.87, "learning_rate": 2.0352434370301837e-05, "loss": 0.6782, "step": 252520 }, { "epoch": 21.87, "learning_rate": 2.0350265988204e-05, "loss": 0.6449, "step": 252530 }, { "epoch": 21.87, "learning_rate": 2.034809760610616e-05, "loss": 0.7415, "step": 252540 }, { "epoch": 21.87, "learning_rate": 2.0345929224008323e-05, "loss": 0.6633, "step": 252550 }, { "epoch": 21.87, "learning_rate": 2.034376084191049e-05, "loss": 0.7754, "step": 252560 }, { "epoch": 21.87, "learning_rate": 2.0341592459812652e-05, "loss": 0.6809, "step": 252570 }, { "epoch": 21.88, "learning_rate": 2.033942407771481e-05, "loss": 0.7169, "step": 252580 }, { "epoch": 21.88, "learning_rate": 2.0337255695616975e-05, "loss": 0.6046, "step": 252590 }, { "epoch": 21.88, "learning_rate": 2.0335087313519138e-05, "loss": 0.8334, "step": 252600 }, { "epoch": 21.88, "learning_rate": 2.03329189314213e-05, "loss": 0.6209, "step": 252610 }, { "epoch": 21.88, "learning_rate": 2.033075054932346e-05, "loss": 0.8247, "step": 252620 }, { "epoch": 21.88, "learning_rate": 2.0328582167225627e-05, "loss": 0.789, "step": 252630 }, { "epoch": 21.88, "learning_rate": 2.032641378512779e-05, "loss": 0.7263, "step": 252640 }, { "epoch": 21.88, "learning_rate": 2.0324245403029953e-05, "loss": 0.7188, "step": 252650 }, { "epoch": 21.88, "learning_rate": 2.0322077020932112e-05, "loss": 0.8303, "step": 252660 }, { "epoch": 21.88, "learning_rate": 2.0319908638834275e-05, "loss": 0.9132, "step": 252670 }, { "epoch": 21.88, "learning_rate": 2.0317740256736438e-05, "loss": 0.7818, "step": 252680 }, { "epoch": 21.88, "learning_rate": 2.0315571874638605e-05, "loss": 0.5982, "step": 252690 }, { "epoch": 21.89, "learning_rate": 2.0313403492540764e-05, "loss": 0.8307, "step": 252700 }, { "epoch": 21.89, "learning_rate": 2.0311235110442927e-05, "loss": 0.6556, "step": 252710 }, { "epoch": 21.89, "learning_rate": 2.030906672834509e-05, "loss": 0.7301, "step": 252720 }, { "epoch": 21.89, "learning_rate": 2.030689834624725e-05, "loss": 0.6584, "step": 252730 }, { "epoch": 21.89, "learning_rate": 2.0304729964149413e-05, "loss": 0.7353, "step": 252740 }, { "epoch": 21.89, "learning_rate": 2.0302561582051576e-05, "loss": 0.7258, "step": 252750 }, { "epoch": 21.89, "learning_rate": 2.0300393199953742e-05, "loss": 0.7157, "step": 252760 }, { "epoch": 21.89, "learning_rate": 2.02982248178559e-05, "loss": 0.6167, "step": 252770 }, { "epoch": 21.89, "learning_rate": 2.0296056435758065e-05, "loss": 0.6943, "step": 252780 }, { "epoch": 21.89, "learning_rate": 2.0293888053660228e-05, "loss": 0.8172, "step": 252790 }, { "epoch": 21.89, "learning_rate": 2.029171967156239e-05, "loss": 0.7135, "step": 252800 }, { "epoch": 21.89, "learning_rate": 2.028955128946455e-05, "loss": 0.7012, "step": 252810 }, { "epoch": 21.9, "learning_rate": 2.0287382907366713e-05, "loss": 0.6722, "step": 252820 }, { "epoch": 21.9, "learning_rate": 2.0285214525268876e-05, "loss": 0.586, "step": 252830 }, { "epoch": 21.9, "learning_rate": 2.0283046143171043e-05, "loss": 0.7772, "step": 252840 }, { "epoch": 21.9, "learning_rate": 2.0280877761073202e-05, "loss": 0.6573, "step": 252850 }, { "epoch": 21.9, "learning_rate": 2.0278709378975365e-05, "loss": 0.7484, "step": 252860 }, { "epoch": 21.9, "learning_rate": 2.027654099687753e-05, "loss": 0.7405, "step": 252870 }, { "epoch": 21.9, "learning_rate": 2.027437261477969e-05, "loss": 0.6555, "step": 252880 }, { "epoch": 21.9, "learning_rate": 2.027220423268185e-05, "loss": 0.6894, "step": 252890 }, { "epoch": 21.9, "learning_rate": 2.0270035850584014e-05, "loss": 0.8207, "step": 252900 }, { "epoch": 21.9, "learning_rate": 2.026786746848618e-05, "loss": 0.6095, "step": 252910 }, { "epoch": 21.9, "learning_rate": 2.0265699086388344e-05, "loss": 0.7298, "step": 252920 }, { "epoch": 21.91, "learning_rate": 2.0263530704290503e-05, "loss": 0.7326, "step": 252930 }, { "epoch": 21.91, "learning_rate": 2.0261362322192666e-05, "loss": 0.6981, "step": 252940 }, { "epoch": 21.91, "learning_rate": 2.025919394009483e-05, "loss": 0.5631, "step": 252950 }, { "epoch": 21.91, "learning_rate": 2.0257025557996992e-05, "loss": 0.772, "step": 252960 }, { "epoch": 21.91, "learning_rate": 2.0254857175899152e-05, "loss": 0.7824, "step": 252970 }, { "epoch": 21.91, "learning_rate": 2.0252688793801318e-05, "loss": 0.6585, "step": 252980 }, { "epoch": 21.91, "learning_rate": 2.025052041170348e-05, "loss": 0.6758, "step": 252990 }, { "epoch": 21.91, "learning_rate": 2.024835202960564e-05, "loss": 0.7956, "step": 253000 }, { "epoch": 21.91, "learning_rate": 2.0246183647507804e-05, "loss": 0.8181, "step": 253010 }, { "epoch": 21.91, "learning_rate": 2.0244015265409967e-05, "loss": 0.723, "step": 253020 }, { "epoch": 21.91, "learning_rate": 2.024184688331213e-05, "loss": 0.7151, "step": 253030 }, { "epoch": 21.91, "learning_rate": 2.023967850121429e-05, "loss": 0.635, "step": 253040 }, { "epoch": 21.92, "learning_rate": 2.0237510119116452e-05, "loss": 0.6047, "step": 253050 }, { "epoch": 21.92, "learning_rate": 2.023534173701862e-05, "loss": 0.6535, "step": 253060 }, { "epoch": 21.92, "learning_rate": 2.0233173354920782e-05, "loss": 0.7415, "step": 253070 }, { "epoch": 21.92, "learning_rate": 2.023100497282294e-05, "loss": 0.5374, "step": 253080 }, { "epoch": 21.92, "learning_rate": 2.0228836590725104e-05, "loss": 0.7337, "step": 253090 }, { "epoch": 21.92, "learning_rate": 2.0226668208627267e-05, "loss": 0.6824, "step": 253100 }, { "epoch": 21.92, "learning_rate": 2.022449982652943e-05, "loss": 0.6677, "step": 253110 }, { "epoch": 21.92, "learning_rate": 2.022233144443159e-05, "loss": 0.7389, "step": 253120 }, { "epoch": 21.92, "learning_rate": 2.0220163062333756e-05, "loss": 0.6751, "step": 253130 }, { "epoch": 21.92, "learning_rate": 2.021799468023592e-05, "loss": 0.7161, "step": 253140 }, { "epoch": 21.92, "learning_rate": 2.0215826298138082e-05, "loss": 0.7795, "step": 253150 }, { "epoch": 21.93, "learning_rate": 2.0213657916040242e-05, "loss": 0.6943, "step": 253160 }, { "epoch": 21.93, "learning_rate": 2.0211489533942405e-05, "loss": 0.6923, "step": 253170 }, { "epoch": 21.93, "learning_rate": 2.0209321151844568e-05, "loss": 0.5197, "step": 253180 }, { "epoch": 21.93, "learning_rate": 2.0207152769746734e-05, "loss": 0.7243, "step": 253190 }, { "epoch": 21.93, "learning_rate": 2.0204984387648894e-05, "loss": 0.507, "step": 253200 }, { "epoch": 21.93, "learning_rate": 2.0202816005551057e-05, "loss": 0.6789, "step": 253210 }, { "epoch": 21.93, "learning_rate": 2.020064762345322e-05, "loss": 0.7215, "step": 253220 }, { "epoch": 21.93, "learning_rate": 2.0198479241355383e-05, "loss": 0.6926, "step": 253230 }, { "epoch": 21.93, "learning_rate": 2.0196310859257543e-05, "loss": 0.5411, "step": 253240 }, { "epoch": 21.93, "learning_rate": 2.0194142477159706e-05, "loss": 0.6433, "step": 253250 }, { "epoch": 21.93, "learning_rate": 2.0191974095061872e-05, "loss": 0.7776, "step": 253260 }, { "epoch": 21.93, "learning_rate": 2.018980571296403e-05, "loss": 0.7251, "step": 253270 }, { "epoch": 21.94, "learning_rate": 2.0187637330866195e-05, "loss": 0.7418, "step": 253280 }, { "epoch": 21.94, "learning_rate": 2.0185468948768358e-05, "loss": 0.7519, "step": 253290 }, { "epoch": 21.94, "learning_rate": 2.018330056667052e-05, "loss": 0.6461, "step": 253300 }, { "epoch": 21.94, "learning_rate": 2.018113218457268e-05, "loss": 0.7909, "step": 253310 }, { "epoch": 21.94, "learning_rate": 2.0178963802474843e-05, "loss": 0.767, "step": 253320 }, { "epoch": 21.94, "learning_rate": 2.0176795420377006e-05, "loss": 0.6518, "step": 253330 }, { "epoch": 21.94, "learning_rate": 2.0174627038279173e-05, "loss": 0.6146, "step": 253340 }, { "epoch": 21.94, "learning_rate": 2.0172458656181332e-05, "loss": 0.6505, "step": 253350 }, { "epoch": 21.94, "learning_rate": 2.0170290274083495e-05, "loss": 0.5805, "step": 253360 }, { "epoch": 21.94, "learning_rate": 2.016812189198566e-05, "loss": 0.7765, "step": 253370 }, { "epoch": 21.94, "learning_rate": 2.016595350988782e-05, "loss": 0.792, "step": 253380 }, { "epoch": 21.95, "learning_rate": 2.016378512778998e-05, "loss": 0.7885, "step": 253390 }, { "epoch": 21.95, "learning_rate": 2.0161616745692144e-05, "loss": 0.6574, "step": 253400 }, { "epoch": 21.95, "learning_rate": 2.015944836359431e-05, "loss": 0.7606, "step": 253410 }, { "epoch": 21.95, "learning_rate": 2.0157279981496473e-05, "loss": 0.5998, "step": 253420 }, { "epoch": 21.95, "learning_rate": 2.0155111599398633e-05, "loss": 0.7222, "step": 253430 }, { "epoch": 21.95, "learning_rate": 2.0152943217300796e-05, "loss": 0.7063, "step": 253440 }, { "epoch": 21.95, "learning_rate": 2.015077483520296e-05, "loss": 0.6842, "step": 253450 }, { "epoch": 21.95, "learning_rate": 2.0148606453105122e-05, "loss": 0.7594, "step": 253460 }, { "epoch": 21.95, "learning_rate": 2.014643807100728e-05, "loss": 0.6428, "step": 253470 }, { "epoch": 21.95, "learning_rate": 2.0144269688909448e-05, "loss": 0.6695, "step": 253480 }, { "epoch": 21.95, "learning_rate": 2.014210130681161e-05, "loss": 0.6717, "step": 253490 }, { "epoch": 21.95, "learning_rate": 2.013993292471377e-05, "loss": 0.654, "step": 253500 }, { "epoch": 21.96, "learning_rate": 2.0137764542615934e-05, "loss": 0.763, "step": 253510 }, { "epoch": 21.96, "learning_rate": 2.0135596160518097e-05, "loss": 0.7275, "step": 253520 }, { "epoch": 21.96, "learning_rate": 2.013342777842026e-05, "loss": 0.7693, "step": 253530 }, { "epoch": 21.96, "learning_rate": 2.013125939632242e-05, "loss": 0.6118, "step": 253540 }, { "epoch": 21.96, "learning_rate": 2.0129091014224586e-05, "loss": 0.7345, "step": 253550 }, { "epoch": 21.96, "learning_rate": 2.012692263212675e-05, "loss": 0.6544, "step": 253560 }, { "epoch": 21.96, "learning_rate": 2.012475425002891e-05, "loss": 0.7085, "step": 253570 }, { "epoch": 21.96, "learning_rate": 2.012258586793107e-05, "loss": 0.8512, "step": 253580 }, { "epoch": 21.96, "learning_rate": 2.0120417485833234e-05, "loss": 0.7164, "step": 253590 }, { "epoch": 21.96, "learning_rate": 2.0118249103735397e-05, "loss": 0.6543, "step": 253600 }, { "epoch": 21.96, "learning_rate": 2.011608072163756e-05, "loss": 0.7351, "step": 253610 }, { "epoch": 21.97, "learning_rate": 2.011391233953972e-05, "loss": 0.6156, "step": 253620 }, { "epoch": 21.97, "learning_rate": 2.0111743957441886e-05, "loss": 0.8062, "step": 253630 }, { "epoch": 21.97, "learning_rate": 2.010957557534405e-05, "loss": 0.9821, "step": 253640 }, { "epoch": 21.97, "learning_rate": 2.0107407193246212e-05, "loss": 0.7637, "step": 253650 }, { "epoch": 21.97, "learning_rate": 2.0105238811148372e-05, "loss": 0.7214, "step": 253660 }, { "epoch": 21.97, "learning_rate": 2.0103070429050535e-05, "loss": 0.7969, "step": 253670 }, { "epoch": 21.97, "learning_rate": 2.0100902046952698e-05, "loss": 0.5706, "step": 253680 }, { "epoch": 21.97, "learning_rate": 2.0098733664854864e-05, "loss": 0.6563, "step": 253690 }, { "epoch": 21.97, "learning_rate": 2.0096565282757024e-05, "loss": 0.7383, "step": 253700 }, { "epoch": 21.97, "learning_rate": 2.0094396900659187e-05, "loss": 0.6085, "step": 253710 }, { "epoch": 21.97, "learning_rate": 2.009222851856135e-05, "loss": 0.8331, "step": 253720 }, { "epoch": 21.97, "learning_rate": 2.0090060136463513e-05, "loss": 0.7168, "step": 253730 }, { "epoch": 21.98, "learning_rate": 2.0087891754365673e-05, "loss": 0.6463, "step": 253740 }, { "epoch": 21.98, "learning_rate": 2.0085723372267836e-05, "loss": 0.9301, "step": 253750 }, { "epoch": 21.98, "learning_rate": 2.0083554990170002e-05, "loss": 0.6778, "step": 253760 }, { "epoch": 21.98, "learning_rate": 2.008138660807216e-05, "loss": 0.721, "step": 253770 }, { "epoch": 21.98, "learning_rate": 2.0079218225974325e-05, "loss": 0.6318, "step": 253780 }, { "epoch": 21.98, "learning_rate": 2.0077049843876488e-05, "loss": 0.8154, "step": 253790 }, { "epoch": 21.98, "learning_rate": 2.007488146177865e-05, "loss": 0.872, "step": 253800 }, { "epoch": 21.98, "learning_rate": 2.007271307968081e-05, "loss": 0.7411, "step": 253810 }, { "epoch": 21.98, "learning_rate": 2.0070544697582973e-05, "loss": 0.7076, "step": 253820 }, { "epoch": 21.98, "learning_rate": 2.006837631548514e-05, "loss": 0.7432, "step": 253830 }, { "epoch": 21.98, "learning_rate": 2.0066207933387303e-05, "loss": 0.7259, "step": 253840 }, { "epoch": 21.99, "learning_rate": 2.0064039551289462e-05, "loss": 0.6949, "step": 253850 }, { "epoch": 21.99, "learning_rate": 2.0061871169191625e-05, "loss": 0.8301, "step": 253860 }, { "epoch": 21.99, "learning_rate": 2.0059702787093788e-05, "loss": 0.7239, "step": 253870 }, { "epoch": 21.99, "learning_rate": 2.005753440499595e-05, "loss": 0.7124, "step": 253880 }, { "epoch": 21.99, "learning_rate": 2.005536602289811e-05, "loss": 0.7356, "step": 253890 }, { "epoch": 21.99, "learning_rate": 2.0053197640800274e-05, "loss": 0.6121, "step": 253900 }, { "epoch": 21.99, "learning_rate": 2.005102925870244e-05, "loss": 0.6254, "step": 253910 }, { "epoch": 21.99, "learning_rate": 2.0048860876604603e-05, "loss": 0.7734, "step": 253920 }, { "epoch": 21.99, "learning_rate": 2.0046692494506763e-05, "loss": 0.7949, "step": 253930 }, { "epoch": 21.99, "learning_rate": 2.0044524112408926e-05, "loss": 0.798, "step": 253940 }, { "epoch": 21.99, "learning_rate": 2.004235573031109e-05, "loss": 0.7353, "step": 253950 }, { "epoch": 21.99, "learning_rate": 2.0040187348213252e-05, "loss": 0.6512, "step": 253960 }, { "epoch": 22.0, "learning_rate": 2.003801896611541e-05, "loss": 0.7262, "step": 253970 }, { "epoch": 22.0, "learning_rate": 2.0035850584017578e-05, "loss": 0.6731, "step": 253980 }, { "epoch": 22.0, "learning_rate": 2.003368220191974e-05, "loss": 0.7988, "step": 253990 }, { "epoch": 22.0, "learning_rate": 2.0031513819821904e-05, "loss": 0.7628, "step": 254000 }, { "epoch": 22.0, "learning_rate": 2.0029345437724063e-05, "loss": 0.7286, "step": 254010 }, { "epoch": 22.0, "learning_rate": 2.0027177055626226e-05, "loss": 0.7385, "step": 254020 }, { "epoch": 22.0, "eval_cer": 0.5369380789421819, "eval_em": 0.05621454357916452, "eval_f1": 0.05621454357916452, "eval_loss": 1.5344864130020142, "eval_runtime": 253.4084, "eval_samples_per_second": 7.652, "eval_steps_per_second": 0.959, "eval_wer": 0.9437854564208354, "step": 254023 }, { "epoch": 22.0, "learning_rate": 2.002500867352839e-05, "loss": 0.7221, "step": 254030 }, { "epoch": 22.0, "learning_rate": 2.002284029143055e-05, "loss": 0.7291, "step": 254040 }, { "epoch": 22.0, "learning_rate": 2.0020671909332715e-05, "loss": 0.6581, "step": 254050 }, { "epoch": 22.0, "learning_rate": 2.001850352723488e-05, "loss": 0.6871, "step": 254060 }, { "epoch": 22.0, "learning_rate": 2.001633514513704e-05, "loss": 0.5856, "step": 254070 }, { "epoch": 22.0, "learning_rate": 2.00141667630392e-05, "loss": 0.6103, "step": 254080 }, { "epoch": 22.01, "learning_rate": 2.0011998380941364e-05, "loss": 0.7821, "step": 254090 }, { "epoch": 22.01, "learning_rate": 2.0009829998843527e-05, "loss": 0.5929, "step": 254100 }, { "epoch": 22.01, "learning_rate": 2.0007661616745693e-05, "loss": 0.5901, "step": 254110 }, { "epoch": 22.01, "learning_rate": 2.000549323464785e-05, "loss": 0.5956, "step": 254120 }, { "epoch": 22.01, "learning_rate": 2.0003324852550016e-05, "loss": 0.5968, "step": 254130 }, { "epoch": 22.01, "learning_rate": 2.000115647045218e-05, "loss": 0.5344, "step": 254140 }, { "epoch": 22.01, "learning_rate": 1.9998988088354342e-05, "loss": 0.6891, "step": 254150 }, { "epoch": 22.01, "learning_rate": 1.9996819706256502e-05, "loss": 0.7529, "step": 254160 }, { "epoch": 22.01, "learning_rate": 1.9994651324158665e-05, "loss": 0.767, "step": 254170 }, { "epoch": 22.01, "learning_rate": 1.9992482942060828e-05, "loss": 0.5784, "step": 254180 }, { "epoch": 22.01, "learning_rate": 1.9990314559962994e-05, "loss": 0.6671, "step": 254190 }, { "epoch": 22.02, "learning_rate": 1.9988146177865154e-05, "loss": 0.7653, "step": 254200 }, { "epoch": 22.02, "learning_rate": 1.9985977795767317e-05, "loss": 0.7224, "step": 254210 }, { "epoch": 22.02, "learning_rate": 1.998380941366948e-05, "loss": 0.6559, "step": 254220 }, { "epoch": 22.02, "learning_rate": 1.9981641031571643e-05, "loss": 0.5602, "step": 254230 }, { "epoch": 22.02, "learning_rate": 1.9979472649473802e-05, "loss": 0.6219, "step": 254240 }, { "epoch": 22.02, "learning_rate": 1.9977304267375965e-05, "loss": 0.6571, "step": 254250 }, { "epoch": 22.02, "learning_rate": 1.9975135885278132e-05, "loss": 0.6112, "step": 254260 }, { "epoch": 22.02, "learning_rate": 1.997296750318029e-05, "loss": 0.7507, "step": 254270 }, { "epoch": 22.02, "learning_rate": 1.9970799121082454e-05, "loss": 0.5414, "step": 254280 }, { "epoch": 22.02, "learning_rate": 1.9968630738984617e-05, "loss": 0.5598, "step": 254290 }, { "epoch": 22.02, "learning_rate": 1.996646235688678e-05, "loss": 0.7924, "step": 254300 }, { "epoch": 22.02, "learning_rate": 1.996429397478894e-05, "loss": 0.6879, "step": 254310 }, { "epoch": 22.03, "learning_rate": 1.9962125592691103e-05, "loss": 0.6092, "step": 254320 }, { "epoch": 22.03, "learning_rate": 1.995995721059327e-05, "loss": 0.7661, "step": 254330 }, { "epoch": 22.03, "learning_rate": 1.9957788828495432e-05, "loss": 0.8064, "step": 254340 }, { "epoch": 22.03, "learning_rate": 1.9955620446397592e-05, "loss": 0.5965, "step": 254350 }, { "epoch": 22.03, "learning_rate": 1.9953452064299755e-05, "loss": 0.5833, "step": 254360 }, { "epoch": 22.03, "learning_rate": 1.9951283682201918e-05, "loss": 0.6362, "step": 254370 }, { "epoch": 22.03, "learning_rate": 1.994911530010408e-05, "loss": 0.6132, "step": 254380 }, { "epoch": 22.03, "learning_rate": 1.994694691800624e-05, "loss": 0.6049, "step": 254390 }, { "epoch": 22.03, "learning_rate": 1.9944778535908404e-05, "loss": 0.6496, "step": 254400 }, { "epoch": 22.03, "learning_rate": 1.994261015381057e-05, "loss": 0.6703, "step": 254410 }, { "epoch": 22.03, "learning_rate": 1.9940441771712733e-05, "loss": 0.6993, "step": 254420 }, { "epoch": 22.04, "learning_rate": 1.9938273389614893e-05, "loss": 0.7372, "step": 254430 }, { "epoch": 22.04, "learning_rate": 1.9936105007517056e-05, "loss": 0.7132, "step": 254440 }, { "epoch": 22.04, "learning_rate": 1.993393662541922e-05, "loss": 0.6911, "step": 254450 }, { "epoch": 22.04, "learning_rate": 1.993176824332138e-05, "loss": 0.6685, "step": 254460 }, { "epoch": 22.04, "learning_rate": 1.992959986122354e-05, "loss": 0.7143, "step": 254470 }, { "epoch": 22.04, "learning_rate": 1.9927431479125708e-05, "loss": 0.6803, "step": 254480 }, { "epoch": 22.04, "learning_rate": 1.992526309702787e-05, "loss": 0.6149, "step": 254490 }, { "epoch": 22.04, "learning_rate": 1.9923094714930034e-05, "loss": 0.7372, "step": 254500 }, { "epoch": 22.04, "learning_rate": 1.9920926332832193e-05, "loss": 0.5797, "step": 254510 }, { "epoch": 22.04, "learning_rate": 1.9918757950734356e-05, "loss": 0.6613, "step": 254520 }, { "epoch": 22.04, "learning_rate": 1.991658956863652e-05, "loss": 0.7001, "step": 254530 }, { "epoch": 22.04, "learning_rate": 1.991442118653868e-05, "loss": 0.5926, "step": 254540 }, { "epoch": 22.05, "learning_rate": 1.9912252804440845e-05, "loss": 0.7059, "step": 254550 }, { "epoch": 22.05, "learning_rate": 1.991008442234301e-05, "loss": 0.8152, "step": 254560 }, { "epoch": 22.05, "learning_rate": 1.990791604024517e-05, "loss": 0.6217, "step": 254570 }, { "epoch": 22.05, "learning_rate": 1.990574765814733e-05, "loss": 0.7296, "step": 254580 }, { "epoch": 22.05, "learning_rate": 1.9903579276049494e-05, "loss": 0.753, "step": 254590 }, { "epoch": 22.05, "learning_rate": 1.9901410893951657e-05, "loss": 0.7434, "step": 254600 }, { "epoch": 22.05, "learning_rate": 1.9899242511853823e-05, "loss": 0.5812, "step": 254610 }, { "epoch": 22.05, "learning_rate": 1.9897074129755983e-05, "loss": 0.6228, "step": 254620 }, { "epoch": 22.05, "learning_rate": 1.9894905747658146e-05, "loss": 0.7094, "step": 254630 }, { "epoch": 22.05, "learning_rate": 1.989273736556031e-05, "loss": 0.5843, "step": 254640 }, { "epoch": 22.05, "learning_rate": 1.9890568983462472e-05, "loss": 0.6512, "step": 254650 }, { "epoch": 22.06, "learning_rate": 1.988840060136463e-05, "loss": 0.5851, "step": 254660 }, { "epoch": 22.06, "learning_rate": 1.9886232219266795e-05, "loss": 0.6921, "step": 254670 }, { "epoch": 22.06, "learning_rate": 1.9884063837168958e-05, "loss": 0.9094, "step": 254680 }, { "epoch": 22.06, "learning_rate": 1.9881895455071124e-05, "loss": 0.6884, "step": 254690 }, { "epoch": 22.06, "learning_rate": 1.9879727072973284e-05, "loss": 0.6087, "step": 254700 }, { "epoch": 22.06, "learning_rate": 1.9877558690875447e-05, "loss": 0.7094, "step": 254710 }, { "epoch": 22.06, "learning_rate": 1.987539030877761e-05, "loss": 0.7101, "step": 254720 }, { "epoch": 22.06, "learning_rate": 1.9873221926679773e-05, "loss": 0.6588, "step": 254730 }, { "epoch": 22.06, "learning_rate": 1.9871053544581932e-05, "loss": 0.7104, "step": 254740 }, { "epoch": 22.06, "learning_rate": 1.9868885162484095e-05, "loss": 0.7724, "step": 254750 }, { "epoch": 22.06, "learning_rate": 1.986671678038626e-05, "loss": 0.642, "step": 254760 }, { "epoch": 22.06, "learning_rate": 1.9864548398288425e-05, "loss": 0.6912, "step": 254770 }, { "epoch": 22.07, "learning_rate": 1.9862380016190584e-05, "loss": 0.6934, "step": 254780 }, { "epoch": 22.07, "learning_rate": 1.9860211634092747e-05, "loss": 0.7909, "step": 254790 }, { "epoch": 22.07, "learning_rate": 1.985804325199491e-05, "loss": 0.8169, "step": 254800 }, { "epoch": 22.07, "learning_rate": 1.985587486989707e-05, "loss": 0.6415, "step": 254810 }, { "epoch": 22.07, "learning_rate": 1.9853706487799233e-05, "loss": 0.5923, "step": 254820 }, { "epoch": 22.07, "learning_rate": 1.98515381057014e-05, "loss": 0.726, "step": 254830 }, { "epoch": 22.07, "learning_rate": 1.9849369723603562e-05, "loss": 0.7098, "step": 254840 }, { "epoch": 22.07, "learning_rate": 1.9847201341505722e-05, "loss": 0.616, "step": 254850 }, { "epoch": 22.07, "learning_rate": 1.9845032959407885e-05, "loss": 0.5718, "step": 254860 }, { "epoch": 22.07, "learning_rate": 1.9842864577310048e-05, "loss": 0.7668, "step": 254870 }, { "epoch": 22.07, "learning_rate": 1.984069619521221e-05, "loss": 0.6906, "step": 254880 }, { "epoch": 22.08, "learning_rate": 1.983852781311437e-05, "loss": 0.6259, "step": 254890 }, { "epoch": 22.08, "learning_rate": 1.9836359431016537e-05, "loss": 0.7103, "step": 254900 }, { "epoch": 22.08, "learning_rate": 1.98341910489187e-05, "loss": 0.6792, "step": 254910 }, { "epoch": 22.08, "learning_rate": 1.9832022666820863e-05, "loss": 0.6933, "step": 254920 }, { "epoch": 22.08, "learning_rate": 1.9829854284723023e-05, "loss": 0.7855, "step": 254930 }, { "epoch": 22.08, "learning_rate": 1.9827685902625186e-05, "loss": 0.7356, "step": 254940 }, { "epoch": 22.08, "learning_rate": 1.982551752052735e-05, "loss": 0.6212, "step": 254950 }, { "epoch": 22.08, "learning_rate": 1.982334913842951e-05, "loss": 0.738, "step": 254960 }, { "epoch": 22.08, "learning_rate": 1.982118075633167e-05, "loss": 0.7693, "step": 254970 }, { "epoch": 22.08, "learning_rate": 1.9819012374233838e-05, "loss": 0.6555, "step": 254980 }, { "epoch": 22.08, "learning_rate": 1.9816843992136e-05, "loss": 0.5916, "step": 254990 }, { "epoch": 22.08, "learning_rate": 1.9814675610038164e-05, "loss": 0.7048, "step": 255000 }, { "epoch": 22.09, "learning_rate": 1.9812507227940323e-05, "loss": 0.6738, "step": 255010 }, { "epoch": 22.09, "learning_rate": 1.9810338845842486e-05, "loss": 0.6648, "step": 255020 }, { "epoch": 22.09, "learning_rate": 1.980817046374465e-05, "loss": 0.5754, "step": 255030 }, { "epoch": 22.09, "learning_rate": 1.980600208164681e-05, "loss": 0.7092, "step": 255040 }, { "epoch": 22.09, "learning_rate": 1.9803833699548975e-05, "loss": 0.6618, "step": 255050 }, { "epoch": 22.09, "learning_rate": 1.9801665317451138e-05, "loss": 0.5972, "step": 255060 }, { "epoch": 22.09, "learning_rate": 1.97994969353533e-05, "loss": 0.7525, "step": 255070 }, { "epoch": 22.09, "learning_rate": 1.979732855325546e-05, "loss": 0.6102, "step": 255080 }, { "epoch": 22.09, "learning_rate": 1.9795160171157624e-05, "loss": 0.6359, "step": 255090 }, { "epoch": 22.09, "learning_rate": 1.9792991789059787e-05, "loss": 0.546, "step": 255100 }, { "epoch": 22.09, "learning_rate": 1.9790823406961953e-05, "loss": 0.6126, "step": 255110 }, { "epoch": 22.1, "learning_rate": 1.9788655024864113e-05, "loss": 0.5794, "step": 255120 }, { "epoch": 22.1, "learning_rate": 1.9786486642766276e-05, "loss": 0.7037, "step": 255130 }, { "epoch": 22.1, "learning_rate": 1.978431826066844e-05, "loss": 0.7228, "step": 255140 }, { "epoch": 22.1, "learning_rate": 1.9782149878570602e-05, "loss": 0.7171, "step": 255150 }, { "epoch": 22.1, "learning_rate": 1.977998149647276e-05, "loss": 0.6427, "step": 255160 }, { "epoch": 22.1, "learning_rate": 1.9777813114374924e-05, "loss": 0.5535, "step": 255170 }, { "epoch": 22.1, "learning_rate": 1.977564473227709e-05, "loss": 0.595, "step": 255180 }, { "epoch": 22.1, "learning_rate": 1.9773476350179254e-05, "loss": 0.6713, "step": 255190 }, { "epoch": 22.1, "learning_rate": 1.9771307968081413e-05, "loss": 0.6294, "step": 255200 }, { "epoch": 22.1, "learning_rate": 1.9769139585983576e-05, "loss": 0.6685, "step": 255210 }, { "epoch": 22.1, "learning_rate": 1.976697120388574e-05, "loss": 0.5954, "step": 255220 }, { "epoch": 22.1, "learning_rate": 1.9764802821787902e-05, "loss": 0.7326, "step": 255230 }, { "epoch": 22.11, "learning_rate": 1.9762634439690062e-05, "loss": 0.6104, "step": 255240 }, { "epoch": 22.11, "learning_rate": 1.9760466057592225e-05, "loss": 0.7288, "step": 255250 }, { "epoch": 22.11, "learning_rate": 1.975829767549439e-05, "loss": 0.7271, "step": 255260 }, { "epoch": 22.11, "learning_rate": 1.9756129293396554e-05, "loss": 0.6268, "step": 255270 }, { "epoch": 22.11, "learning_rate": 1.9753960911298714e-05, "loss": 0.6038, "step": 255280 }, { "epoch": 22.11, "learning_rate": 1.9751792529200877e-05, "loss": 0.6033, "step": 255290 }, { "epoch": 22.11, "learning_rate": 1.974962414710304e-05, "loss": 0.7183, "step": 255300 }, { "epoch": 22.11, "learning_rate": 1.97474557650052e-05, "loss": 0.5092, "step": 255310 }, { "epoch": 22.11, "learning_rate": 1.9745287382907363e-05, "loss": 0.6091, "step": 255320 }, { "epoch": 22.11, "learning_rate": 1.974311900080953e-05, "loss": 0.7059, "step": 255330 }, { "epoch": 22.11, "learning_rate": 1.9740950618711692e-05, "loss": 0.7422, "step": 255340 }, { "epoch": 22.11, "learning_rate": 1.9738782236613852e-05, "loss": 0.5238, "step": 255350 }, { "epoch": 22.12, "learning_rate": 1.9736613854516015e-05, "loss": 0.6407, "step": 255360 }, { "epoch": 22.12, "learning_rate": 1.9734445472418178e-05, "loss": 0.6031, "step": 255370 }, { "epoch": 22.12, "learning_rate": 1.973227709032034e-05, "loss": 0.775, "step": 255380 }, { "epoch": 22.12, "learning_rate": 1.97301087082225e-05, "loss": 0.5962, "step": 255390 }, { "epoch": 22.12, "learning_rate": 1.9727940326124667e-05, "loss": 0.5885, "step": 255400 }, { "epoch": 22.12, "learning_rate": 1.972577194402683e-05, "loss": 0.6341, "step": 255410 }, { "epoch": 22.12, "learning_rate": 1.9723603561928993e-05, "loss": 0.7465, "step": 255420 }, { "epoch": 22.12, "learning_rate": 1.9721435179831152e-05, "loss": 0.5216, "step": 255430 }, { "epoch": 22.12, "learning_rate": 1.9719266797733315e-05, "loss": 0.6644, "step": 255440 }, { "epoch": 22.12, "learning_rate": 1.971709841563548e-05, "loss": 0.7245, "step": 255450 }, { "epoch": 22.12, "learning_rate": 1.9714930033537645e-05, "loss": 0.6272, "step": 255460 }, { "epoch": 22.13, "learning_rate": 1.97127616514398e-05, "loss": 0.6623, "step": 255470 }, { "epoch": 22.13, "learning_rate": 1.9710593269341967e-05, "loss": 0.8863, "step": 255480 }, { "epoch": 22.13, "learning_rate": 1.970842488724413e-05, "loss": 0.7326, "step": 255490 }, { "epoch": 22.13, "learning_rate": 1.9706256505146293e-05, "loss": 0.6419, "step": 255500 }, { "epoch": 22.13, "learning_rate": 1.9704088123048453e-05, "loss": 0.6938, "step": 255510 }, { "epoch": 22.13, "learning_rate": 1.9701919740950616e-05, "loss": 0.7858, "step": 255520 }, { "epoch": 22.13, "learning_rate": 1.969975135885278e-05, "loss": 0.7776, "step": 255530 }, { "epoch": 22.13, "learning_rate": 1.9697582976754945e-05, "loss": 0.6485, "step": 255540 }, { "epoch": 22.13, "learning_rate": 1.9695414594657105e-05, "loss": 0.6491, "step": 255550 }, { "epoch": 22.13, "learning_rate": 1.9693246212559268e-05, "loss": 0.6639, "step": 255560 }, { "epoch": 22.13, "learning_rate": 1.969107783046143e-05, "loss": 0.5201, "step": 255570 }, { "epoch": 22.13, "learning_rate": 1.968890944836359e-05, "loss": 0.6818, "step": 255580 }, { "epoch": 22.14, "learning_rate": 1.9686741066265754e-05, "loss": 0.6857, "step": 255590 }, { "epoch": 22.14, "learning_rate": 1.9684572684167917e-05, "loss": 0.6259, "step": 255600 }, { "epoch": 22.14, "learning_rate": 1.9682404302070083e-05, "loss": 0.7108, "step": 255610 }, { "epoch": 22.14, "learning_rate": 1.9680235919972243e-05, "loss": 0.7235, "step": 255620 }, { "epoch": 22.14, "learning_rate": 1.9678067537874406e-05, "loss": 0.5335, "step": 255630 }, { "epoch": 22.14, "learning_rate": 1.967589915577657e-05, "loss": 0.6968, "step": 255640 }, { "epoch": 22.14, "learning_rate": 1.967373077367873e-05, "loss": 0.6664, "step": 255650 }, { "epoch": 22.14, "learning_rate": 1.967156239158089e-05, "loss": 0.7569, "step": 255660 }, { "epoch": 22.14, "learning_rate": 1.9669394009483054e-05, "loss": 0.6799, "step": 255670 }, { "epoch": 22.14, "learning_rate": 1.966722562738522e-05, "loss": 0.6817, "step": 255680 }, { "epoch": 22.14, "learning_rate": 1.9665057245287384e-05, "loss": 0.6714, "step": 255690 }, { "epoch": 22.15, "learning_rate": 1.9662888863189543e-05, "loss": 0.7226, "step": 255700 }, { "epoch": 22.15, "learning_rate": 1.9660720481091706e-05, "loss": 0.8567, "step": 255710 }, { "epoch": 22.15, "learning_rate": 1.965855209899387e-05, "loss": 0.66, "step": 255720 }, { "epoch": 22.15, "learning_rate": 1.9656383716896032e-05, "loss": 0.5778, "step": 255730 }, { "epoch": 22.15, "learning_rate": 1.9654215334798192e-05, "loss": 0.5607, "step": 255740 }, { "epoch": 22.15, "learning_rate": 1.9652046952700355e-05, "loss": 0.7907, "step": 255750 }, { "epoch": 22.15, "learning_rate": 1.964987857060252e-05, "loss": 0.6606, "step": 255760 }, { "epoch": 22.15, "learning_rate": 1.9647710188504684e-05, "loss": 0.7025, "step": 255770 }, { "epoch": 22.15, "learning_rate": 1.9645541806406844e-05, "loss": 0.6586, "step": 255780 }, { "epoch": 22.15, "learning_rate": 1.9643373424309007e-05, "loss": 0.5396, "step": 255790 }, { "epoch": 22.15, "learning_rate": 1.964120504221117e-05, "loss": 0.7164, "step": 255800 }, { "epoch": 22.15, "learning_rate": 1.963903666011333e-05, "loss": 0.6707, "step": 255810 }, { "epoch": 22.16, "learning_rate": 1.9636868278015493e-05, "loss": 0.7109, "step": 255820 }, { "epoch": 22.16, "learning_rate": 1.963469989591766e-05, "loss": 0.6153, "step": 255830 }, { "epoch": 22.16, "learning_rate": 1.9632531513819822e-05, "loss": 0.5905, "step": 255840 }, { "epoch": 22.16, "learning_rate": 1.963036313172198e-05, "loss": 0.5985, "step": 255850 }, { "epoch": 22.16, "learning_rate": 1.9628194749624145e-05, "loss": 0.6415, "step": 255860 }, { "epoch": 22.16, "learning_rate": 1.9626026367526308e-05, "loss": 0.5921, "step": 255870 }, { "epoch": 22.16, "learning_rate": 1.962385798542847e-05, "loss": 0.6978, "step": 255880 }, { "epoch": 22.16, "learning_rate": 1.962168960333063e-05, "loss": 0.7596, "step": 255890 }, { "epoch": 22.16, "learning_rate": 1.9619521221232797e-05, "loss": 0.6572, "step": 255900 }, { "epoch": 22.16, "learning_rate": 1.961735283913496e-05, "loss": 0.5296, "step": 255910 }, { "epoch": 22.16, "learning_rate": 1.9615184457037123e-05, "loss": 0.6444, "step": 255920 }, { "epoch": 22.17, "learning_rate": 1.9613016074939282e-05, "loss": 0.5563, "step": 255930 }, { "epoch": 22.17, "learning_rate": 1.9610847692841445e-05, "loss": 0.6934, "step": 255940 }, { "epoch": 22.17, "learning_rate": 1.9608679310743608e-05, "loss": 0.6817, "step": 255950 }, { "epoch": 22.17, "learning_rate": 1.9606510928645775e-05, "loss": 0.6451, "step": 255960 }, { "epoch": 22.17, "learning_rate": 1.9604342546547934e-05, "loss": 0.7493, "step": 255970 }, { "epoch": 22.17, "learning_rate": 1.9602174164450097e-05, "loss": 0.5367, "step": 255980 }, { "epoch": 22.17, "learning_rate": 1.960000578235226e-05, "loss": 0.6893, "step": 255990 }, { "epoch": 22.17, "learning_rate": 1.9597837400254423e-05, "loss": 0.6662, "step": 256000 }, { "epoch": 22.17, "learning_rate": 1.9595669018156583e-05, "loss": 0.665, "step": 256010 }, { "epoch": 22.17, "learning_rate": 1.9593500636058746e-05, "loss": 0.6818, "step": 256020 }, { "epoch": 22.17, "learning_rate": 1.959133225396091e-05, "loss": 0.6837, "step": 256030 }, { "epoch": 22.17, "learning_rate": 1.9589163871863075e-05, "loss": 0.7917, "step": 256040 }, { "epoch": 22.18, "learning_rate": 1.9586995489765235e-05, "loss": 0.7022, "step": 256050 }, { "epoch": 22.18, "learning_rate": 1.9584827107667398e-05, "loss": 0.6809, "step": 256060 }, { "epoch": 22.18, "learning_rate": 1.958265872556956e-05, "loss": 0.7116, "step": 256070 }, { "epoch": 22.18, "learning_rate": 1.958049034347172e-05, "loss": 0.6457, "step": 256080 }, { "epoch": 22.18, "learning_rate": 1.9578321961373883e-05, "loss": 0.7752, "step": 256090 }, { "epoch": 22.18, "learning_rate": 1.9576153579276046e-05, "loss": 0.8684, "step": 256100 }, { "epoch": 22.18, "learning_rate": 1.9573985197178213e-05, "loss": 0.6236, "step": 256110 }, { "epoch": 22.18, "learning_rate": 1.9571816815080372e-05, "loss": 0.6207, "step": 256120 }, { "epoch": 22.18, "learning_rate": 1.9569648432982535e-05, "loss": 0.8131, "step": 256130 }, { "epoch": 22.18, "learning_rate": 1.95674800508847e-05, "loss": 0.6377, "step": 256140 }, { "epoch": 22.18, "learning_rate": 1.956531166878686e-05, "loss": 0.6599, "step": 256150 }, { "epoch": 22.19, "learning_rate": 1.956314328668902e-05, "loss": 0.6427, "step": 256160 }, { "epoch": 22.19, "learning_rate": 1.9560974904591184e-05, "loss": 0.7443, "step": 256170 }, { "epoch": 22.19, "learning_rate": 1.955880652249335e-05, "loss": 0.6222, "step": 256180 }, { "epoch": 22.19, "learning_rate": 1.9556638140395514e-05, "loss": 0.6978, "step": 256190 }, { "epoch": 22.19, "learning_rate": 1.9554469758297673e-05, "loss": 0.6838, "step": 256200 }, { "epoch": 22.19, "learning_rate": 1.9552301376199836e-05, "loss": 0.6492, "step": 256210 }, { "epoch": 22.19, "learning_rate": 1.9550132994102e-05, "loss": 0.6689, "step": 256220 }, { "epoch": 22.19, "learning_rate": 1.9547964612004162e-05, "loss": 0.6816, "step": 256230 }, { "epoch": 22.19, "learning_rate": 1.9545796229906322e-05, "loss": 0.637, "step": 256240 }, { "epoch": 22.19, "learning_rate": 1.9543627847808488e-05, "loss": 0.5771, "step": 256250 }, { "epoch": 22.19, "learning_rate": 1.954145946571065e-05, "loss": 0.6848, "step": 256260 }, { "epoch": 22.19, "learning_rate": 1.9539291083612814e-05, "loss": 0.5512, "step": 256270 }, { "epoch": 22.2, "learning_rate": 1.9537122701514974e-05, "loss": 0.6815, "step": 256280 }, { "epoch": 22.2, "learning_rate": 1.9534954319417137e-05, "loss": 0.8389, "step": 256290 }, { "epoch": 22.2, "learning_rate": 1.95327859373193e-05, "loss": 0.7026, "step": 256300 }, { "epoch": 22.2, "learning_rate": 1.9530617555221463e-05, "loss": 0.5182, "step": 256310 }, { "epoch": 22.2, "learning_rate": 1.9528449173123622e-05, "loss": 0.6916, "step": 256320 }, { "epoch": 22.2, "learning_rate": 1.952628079102579e-05, "loss": 0.695, "step": 256330 }, { "epoch": 22.2, "learning_rate": 1.9524112408927952e-05, "loss": 0.7135, "step": 256340 }, { "epoch": 22.2, "learning_rate": 1.952194402683011e-05, "loss": 0.725, "step": 256350 }, { "epoch": 22.2, "learning_rate": 1.9519775644732274e-05, "loss": 0.619, "step": 256360 }, { "epoch": 22.2, "learning_rate": 1.9517607262634437e-05, "loss": 0.6582, "step": 256370 }, { "epoch": 22.2, "learning_rate": 1.95154388805366e-05, "loss": 0.6694, "step": 256380 }, { "epoch": 22.2, "learning_rate": 1.951327049843876e-05, "loss": 0.7638, "step": 256390 }, { "epoch": 22.21, "learning_rate": 1.9511102116340926e-05, "loss": 0.5344, "step": 256400 }, { "epoch": 22.21, "learning_rate": 1.950893373424309e-05, "loss": 0.6722, "step": 256410 }, { "epoch": 22.21, "learning_rate": 1.9506765352145252e-05, "loss": 0.6706, "step": 256420 }, { "epoch": 22.21, "learning_rate": 1.9504596970047412e-05, "loss": 0.67, "step": 256430 }, { "epoch": 22.21, "learning_rate": 1.9502428587949575e-05, "loss": 0.6986, "step": 256440 }, { "epoch": 22.21, "learning_rate": 1.9500260205851738e-05, "loss": 0.7313, "step": 256450 }, { "epoch": 22.21, "learning_rate": 1.9498091823753904e-05, "loss": 0.5127, "step": 256460 }, { "epoch": 22.21, "learning_rate": 1.9495923441656064e-05, "loss": 0.6173, "step": 256470 }, { "epoch": 22.21, "learning_rate": 1.9493755059558227e-05, "loss": 0.6353, "step": 256480 }, { "epoch": 22.21, "learning_rate": 1.949158667746039e-05, "loss": 0.6527, "step": 256490 }, { "epoch": 22.21, "learning_rate": 1.9489418295362553e-05, "loss": 0.6124, "step": 256500 }, { "epoch": 22.22, "learning_rate": 1.9487249913264713e-05, "loss": 0.7606, "step": 256510 }, { "epoch": 22.22, "learning_rate": 1.9485081531166876e-05, "loss": 0.6192, "step": 256520 }, { "epoch": 22.22, "learning_rate": 1.9482913149069042e-05, "loss": 0.6368, "step": 256530 }, { "epoch": 22.22, "learning_rate": 1.9480744766971205e-05, "loss": 0.7586, "step": 256540 }, { "epoch": 22.22, "learning_rate": 1.9478576384873365e-05, "loss": 0.7523, "step": 256550 }, { "epoch": 22.22, "learning_rate": 1.9476408002775528e-05, "loss": 0.6655, "step": 256560 }, { "epoch": 22.22, "learning_rate": 1.947423962067769e-05, "loss": 0.5519, "step": 256570 }, { "epoch": 22.22, "learning_rate": 1.947207123857985e-05, "loss": 0.7028, "step": 256580 }, { "epoch": 22.22, "learning_rate": 1.9469902856482013e-05, "loss": 0.7293, "step": 256590 }, { "epoch": 22.22, "learning_rate": 1.9467734474384176e-05, "loss": 0.6801, "step": 256600 }, { "epoch": 22.22, "learning_rate": 1.9465566092286343e-05, "loss": 0.5578, "step": 256610 }, { "epoch": 22.22, "learning_rate": 1.9463397710188502e-05, "loss": 0.6829, "step": 256620 }, { "epoch": 22.23, "learning_rate": 1.9461229328090665e-05, "loss": 0.7493, "step": 256630 }, { "epoch": 22.23, "learning_rate": 1.945906094599283e-05, "loss": 0.6959, "step": 256640 }, { "epoch": 22.23, "learning_rate": 1.945689256389499e-05, "loss": 0.5695, "step": 256650 }, { "epoch": 22.23, "learning_rate": 1.945472418179715e-05, "loss": 0.5641, "step": 256660 }, { "epoch": 22.23, "learning_rate": 1.9452555799699314e-05, "loss": 0.6217, "step": 256670 }, { "epoch": 22.23, "learning_rate": 1.945038741760148e-05, "loss": 0.588, "step": 256680 }, { "epoch": 22.23, "learning_rate": 1.9448219035503643e-05, "loss": 0.7949, "step": 256690 }, { "epoch": 22.23, "learning_rate": 1.9446050653405803e-05, "loss": 0.7245, "step": 256700 }, { "epoch": 22.23, "learning_rate": 1.9443882271307966e-05, "loss": 0.6238, "step": 256710 }, { "epoch": 22.23, "learning_rate": 1.944171388921013e-05, "loss": 0.7566, "step": 256720 }, { "epoch": 22.23, "learning_rate": 1.9439545507112292e-05, "loss": 0.5618, "step": 256730 }, { "epoch": 22.24, "learning_rate": 1.943737712501445e-05, "loss": 0.7655, "step": 256740 }, { "epoch": 22.24, "learning_rate": 1.9435208742916618e-05, "loss": 0.7542, "step": 256750 }, { "epoch": 22.24, "learning_rate": 1.943304036081878e-05, "loss": 0.6567, "step": 256760 }, { "epoch": 22.24, "learning_rate": 1.9430871978720944e-05, "loss": 0.6249, "step": 256770 }, { "epoch": 22.24, "learning_rate": 1.9428703596623104e-05, "loss": 0.7829, "step": 256780 }, { "epoch": 22.24, "learning_rate": 1.9426535214525267e-05, "loss": 0.6096, "step": 256790 }, { "epoch": 22.24, "learning_rate": 1.942436683242743e-05, "loss": 0.6905, "step": 256800 }, { "epoch": 22.24, "learning_rate": 1.9422198450329596e-05, "loss": 0.5942, "step": 256810 }, { "epoch": 22.24, "learning_rate": 1.9420030068231752e-05, "loss": 0.5649, "step": 256820 }, { "epoch": 22.24, "learning_rate": 1.941786168613392e-05, "loss": 0.6431, "step": 256830 }, { "epoch": 22.24, "learning_rate": 1.941569330403608e-05, "loss": 0.6183, "step": 256840 }, { "epoch": 22.24, "learning_rate": 1.941352492193824e-05, "loss": 0.8471, "step": 256850 }, { "epoch": 22.25, "learning_rate": 1.9411356539840404e-05, "loss": 0.6439, "step": 256860 }, { "epoch": 22.25, "learning_rate": 1.9409188157742567e-05, "loss": 0.7199, "step": 256870 }, { "epoch": 22.25, "learning_rate": 1.940701977564473e-05, "loss": 0.8234, "step": 256880 }, { "epoch": 22.25, "learning_rate": 1.940485139354689e-05, "loss": 0.7316, "step": 256890 }, { "epoch": 22.25, "learning_rate": 1.9402683011449056e-05, "loss": 0.6195, "step": 256900 }, { "epoch": 22.25, "learning_rate": 1.940051462935122e-05, "loss": 0.7268, "step": 256910 }, { "epoch": 22.25, "learning_rate": 1.9398346247253382e-05, "loss": 0.52, "step": 256920 }, { "epoch": 22.25, "learning_rate": 1.9396177865155542e-05, "loss": 0.7021, "step": 256930 }, { "epoch": 22.25, "learning_rate": 1.9394009483057705e-05, "loss": 0.7344, "step": 256940 }, { "epoch": 22.25, "learning_rate": 1.9391841100959868e-05, "loss": 0.688, "step": 256950 }, { "epoch": 22.25, "learning_rate": 1.9389672718862034e-05, "loss": 0.6258, "step": 256960 }, { "epoch": 22.26, "learning_rate": 1.9387504336764194e-05, "loss": 0.6325, "step": 256970 }, { "epoch": 22.26, "learning_rate": 1.9385335954666357e-05, "loss": 0.6806, "step": 256980 }, { "epoch": 22.26, "learning_rate": 1.938316757256852e-05, "loss": 0.5908, "step": 256990 }, { "epoch": 22.26, "learning_rate": 1.9380999190470683e-05, "loss": 0.6647, "step": 257000 }, { "epoch": 22.26, "learning_rate": 1.9378830808372843e-05, "loss": 0.6662, "step": 257010 }, { "epoch": 22.26, "learning_rate": 1.9376662426275006e-05, "loss": 0.6472, "step": 257020 }, { "epoch": 22.26, "learning_rate": 1.9374494044177172e-05, "loss": 0.6342, "step": 257030 }, { "epoch": 22.26, "learning_rate": 1.9372325662079335e-05, "loss": 0.8673, "step": 257040 }, { "epoch": 22.26, "learning_rate": 1.9370157279981495e-05, "loss": 0.8861, "step": 257050 }, { "epoch": 22.26, "learning_rate": 1.9367988897883658e-05, "loss": 0.7204, "step": 257060 }, { "epoch": 22.26, "learning_rate": 1.936582051578582e-05, "loss": 0.7506, "step": 257070 }, { "epoch": 22.26, "learning_rate": 1.9363652133687984e-05, "loss": 0.6628, "step": 257080 }, { "epoch": 22.27, "learning_rate": 1.9361483751590143e-05, "loss": 0.6765, "step": 257090 }, { "epoch": 22.27, "learning_rate": 1.9359315369492306e-05, "loss": 0.6449, "step": 257100 }, { "epoch": 22.27, "learning_rate": 1.9357146987394473e-05, "loss": 0.7521, "step": 257110 }, { "epoch": 22.27, "learning_rate": 1.9354978605296632e-05, "loss": 0.6542, "step": 257120 }, { "epoch": 22.27, "learning_rate": 1.9352810223198795e-05, "loss": 0.8383, "step": 257130 }, { "epoch": 22.27, "learning_rate": 1.9350641841100958e-05, "loss": 0.6342, "step": 257140 }, { "epoch": 22.27, "learning_rate": 1.934847345900312e-05, "loss": 0.5501, "step": 257150 }, { "epoch": 22.27, "learning_rate": 1.934630507690528e-05, "loss": 0.6996, "step": 257160 }, { "epoch": 22.27, "learning_rate": 1.9344136694807444e-05, "loss": 0.6804, "step": 257170 }, { "epoch": 22.27, "learning_rate": 1.934196831270961e-05, "loss": 0.6634, "step": 257180 }, { "epoch": 22.27, "learning_rate": 1.9339799930611773e-05, "loss": 0.5993, "step": 257190 }, { "epoch": 22.28, "learning_rate": 1.9337631548513933e-05, "loss": 0.7115, "step": 257200 }, { "epoch": 22.28, "learning_rate": 1.9335463166416096e-05, "loss": 0.79, "step": 257210 }, { "epoch": 22.28, "learning_rate": 1.933329478431826e-05, "loss": 0.6449, "step": 257220 }, { "epoch": 22.28, "learning_rate": 1.9331126402220422e-05, "loss": 0.673, "step": 257230 }, { "epoch": 22.28, "learning_rate": 1.932895802012258e-05, "loss": 0.6829, "step": 257240 }, { "epoch": 22.28, "learning_rate": 1.9326789638024748e-05, "loss": 0.6174, "step": 257250 }, { "epoch": 22.28, "learning_rate": 1.932462125592691e-05, "loss": 0.7305, "step": 257260 }, { "epoch": 22.28, "learning_rate": 1.9322452873829074e-05, "loss": 0.5967, "step": 257270 }, { "epoch": 22.28, "learning_rate": 1.9320284491731233e-05, "loss": 0.5799, "step": 257280 }, { "epoch": 22.28, "learning_rate": 1.9318116109633396e-05, "loss": 0.6223, "step": 257290 }, { "epoch": 22.28, "learning_rate": 1.931594772753556e-05, "loss": 0.5497, "step": 257300 }, { "epoch": 22.28, "learning_rate": 1.9313779345437726e-05, "loss": 0.7371, "step": 257310 }, { "epoch": 22.29, "learning_rate": 1.9311610963339885e-05, "loss": 0.6432, "step": 257320 }, { "epoch": 22.29, "learning_rate": 1.930944258124205e-05, "loss": 0.4291, "step": 257330 }, { "epoch": 22.29, "learning_rate": 1.930727419914421e-05, "loss": 0.6849, "step": 257340 }, { "epoch": 22.29, "learning_rate": 1.930510581704637e-05, "loss": 0.7616, "step": 257350 }, { "epoch": 22.29, "learning_rate": 1.9302937434948534e-05, "loss": 0.7388, "step": 257360 }, { "epoch": 22.29, "learning_rate": 1.9300769052850697e-05, "loss": 0.6581, "step": 257370 }, { "epoch": 22.29, "learning_rate": 1.929860067075286e-05, "loss": 0.6868, "step": 257380 }, { "epoch": 22.29, "learning_rate": 1.929643228865502e-05, "loss": 0.763, "step": 257390 }, { "epoch": 22.29, "learning_rate": 1.9294263906557186e-05, "loss": 0.5511, "step": 257400 }, { "epoch": 22.29, "learning_rate": 1.929209552445935e-05, "loss": 0.4979, "step": 257410 }, { "epoch": 22.29, "learning_rate": 1.9289927142361512e-05, "loss": 0.7167, "step": 257420 }, { "epoch": 22.3, "learning_rate": 1.9287758760263672e-05, "loss": 0.8446, "step": 257430 }, { "epoch": 22.3, "learning_rate": 1.9285590378165835e-05, "loss": 0.8237, "step": 257440 }, { "epoch": 22.3, "learning_rate": 1.9283421996067998e-05, "loss": 0.8235, "step": 257450 }, { "epoch": 22.3, "learning_rate": 1.9281253613970164e-05, "loss": 0.7385, "step": 257460 }, { "epoch": 22.3, "learning_rate": 1.9279085231872324e-05, "loss": 0.6855, "step": 257470 }, { "epoch": 22.3, "learning_rate": 1.9276916849774487e-05, "loss": 0.6216, "step": 257480 }, { "epoch": 22.3, "learning_rate": 1.927474846767665e-05, "loss": 0.7187, "step": 257490 }, { "epoch": 22.3, "learning_rate": 1.9272580085578813e-05, "loss": 0.6901, "step": 257500 }, { "epoch": 22.3, "learning_rate": 1.9270411703480972e-05, "loss": 0.5974, "step": 257510 }, { "epoch": 22.3, "learning_rate": 1.9268243321383135e-05, "loss": 0.8335, "step": 257520 }, { "epoch": 22.3, "learning_rate": 1.9266074939285302e-05, "loss": 0.6545, "step": 257530 }, { "epoch": 22.3, "learning_rate": 1.9263906557187465e-05, "loss": 0.6319, "step": 257540 }, { "epoch": 22.31, "learning_rate": 1.9261738175089624e-05, "loss": 0.7331, "step": 257550 }, { "epoch": 22.31, "learning_rate": 1.9259569792991787e-05, "loss": 0.7563, "step": 257560 }, { "epoch": 22.31, "learning_rate": 1.925740141089395e-05, "loss": 0.7541, "step": 257570 }, { "epoch": 22.31, "learning_rate": 1.9255233028796113e-05, "loss": 0.6182, "step": 257580 }, { "epoch": 22.31, "learning_rate": 1.9253064646698273e-05, "loss": 0.6174, "step": 257590 }, { "epoch": 22.31, "learning_rate": 1.925089626460044e-05, "loss": 0.6881, "step": 257600 }, { "epoch": 22.31, "learning_rate": 1.9248727882502602e-05, "loss": 0.7109, "step": 257610 }, { "epoch": 22.31, "learning_rate": 1.9246559500404762e-05, "loss": 0.5654, "step": 257620 }, { "epoch": 22.31, "learning_rate": 1.9244391118306925e-05, "loss": 0.6652, "step": 257630 }, { "epoch": 22.31, "learning_rate": 1.9242222736209088e-05, "loss": 0.6743, "step": 257640 }, { "epoch": 22.31, "learning_rate": 1.924005435411125e-05, "loss": 0.712, "step": 257650 }, { "epoch": 22.31, "learning_rate": 1.923788597201341e-05, "loss": 0.7018, "step": 257660 }, { "epoch": 22.32, "learning_rate": 1.9235717589915574e-05, "loss": 0.8239, "step": 257670 }, { "epoch": 22.32, "learning_rate": 1.923354920781774e-05, "loss": 0.6508, "step": 257680 }, { "epoch": 22.32, "learning_rate": 1.9231380825719903e-05, "loss": 0.6942, "step": 257690 }, { "epoch": 22.32, "learning_rate": 1.9229212443622063e-05, "loss": 0.6792, "step": 257700 }, { "epoch": 22.32, "learning_rate": 1.9227044061524226e-05, "loss": 0.7189, "step": 257710 }, { "epoch": 22.32, "learning_rate": 1.922487567942639e-05, "loss": 0.7316, "step": 257720 }, { "epoch": 22.32, "learning_rate": 1.922270729732855e-05, "loss": 0.4746, "step": 257730 }, { "epoch": 22.32, "learning_rate": 1.922053891523071e-05, "loss": 0.736, "step": 257740 }, { "epoch": 22.32, "learning_rate": 1.9218370533132878e-05, "loss": 0.5829, "step": 257750 }, { "epoch": 22.32, "learning_rate": 1.921620215103504e-05, "loss": 0.6108, "step": 257760 }, { "epoch": 22.32, "learning_rate": 1.9214033768937204e-05, "loss": 0.7852, "step": 257770 }, { "epoch": 22.33, "learning_rate": 1.9211865386839363e-05, "loss": 0.9413, "step": 257780 }, { "epoch": 22.33, "learning_rate": 1.9209697004741526e-05, "loss": 0.6288, "step": 257790 }, { "epoch": 22.33, "learning_rate": 1.920752862264369e-05, "loss": 0.7299, "step": 257800 }, { "epoch": 22.33, "learning_rate": 1.9205360240545856e-05, "loss": 0.7105, "step": 257810 }, { "epoch": 22.33, "learning_rate": 1.9203191858448015e-05, "loss": 0.6187, "step": 257820 }, { "epoch": 22.33, "learning_rate": 1.920102347635018e-05, "loss": 0.8023, "step": 257830 }, { "epoch": 22.33, "learning_rate": 1.919885509425234e-05, "loss": 0.5757, "step": 257840 }, { "epoch": 22.33, "learning_rate": 1.9196686712154504e-05, "loss": 0.5669, "step": 257850 }, { "epoch": 22.33, "learning_rate": 1.9194518330056664e-05, "loss": 0.6429, "step": 257860 }, { "epoch": 22.33, "learning_rate": 1.9192349947958827e-05, "loss": 0.679, "step": 257870 }, { "epoch": 22.33, "learning_rate": 1.9190181565860993e-05, "loss": 0.7279, "step": 257880 }, { "epoch": 22.33, "learning_rate": 1.918801318376315e-05, "loss": 0.6266, "step": 257890 }, { "epoch": 22.34, "learning_rate": 1.9185844801665316e-05, "loss": 0.7408, "step": 257900 }, { "epoch": 22.34, "learning_rate": 1.918367641956748e-05, "loss": 0.7748, "step": 257910 }, { "epoch": 22.34, "learning_rate": 1.9181508037469642e-05, "loss": 0.7121, "step": 257920 }, { "epoch": 22.34, "learning_rate": 1.91793396553718e-05, "loss": 0.7587, "step": 257930 }, { "epoch": 22.34, "learning_rate": 1.9177171273273965e-05, "loss": 0.8266, "step": 257940 }, { "epoch": 22.34, "learning_rate": 1.9175002891176128e-05, "loss": 0.6493, "step": 257950 }, { "epoch": 22.34, "learning_rate": 1.9172834509078294e-05, "loss": 0.6686, "step": 257960 }, { "epoch": 22.34, "learning_rate": 1.9170666126980454e-05, "loss": 0.6981, "step": 257970 }, { "epoch": 22.34, "learning_rate": 1.9168497744882617e-05, "loss": 0.6881, "step": 257980 }, { "epoch": 22.34, "learning_rate": 1.916632936278478e-05, "loss": 0.6185, "step": 257990 }, { "epoch": 22.34, "learning_rate": 1.9164160980686943e-05, "loss": 0.7112, "step": 258000 }, { "epoch": 22.35, "learning_rate": 1.9161992598589102e-05, "loss": 0.8482, "step": 258010 }, { "epoch": 22.35, "learning_rate": 1.9159824216491265e-05, "loss": 0.7046, "step": 258020 }, { "epoch": 22.35, "learning_rate": 1.915765583439343e-05, "loss": 0.6047, "step": 258030 }, { "epoch": 22.35, "learning_rate": 1.9155487452295595e-05, "loss": 0.7936, "step": 258040 }, { "epoch": 22.35, "learning_rate": 1.9153319070197754e-05, "loss": 0.5608, "step": 258050 }, { "epoch": 22.35, "learning_rate": 1.9151150688099917e-05, "loss": 0.6367, "step": 258060 }, { "epoch": 22.35, "learning_rate": 1.914898230600208e-05, "loss": 0.6463, "step": 258070 }, { "epoch": 22.35, "learning_rate": 1.9146813923904243e-05, "loss": 0.6957, "step": 258080 }, { "epoch": 22.35, "learning_rate": 1.9144645541806403e-05, "loss": 0.5534, "step": 258090 }, { "epoch": 22.35, "learning_rate": 1.914247715970857e-05, "loss": 0.6616, "step": 258100 }, { "epoch": 22.35, "learning_rate": 1.9140308777610732e-05, "loss": 0.5642, "step": 258110 }, { "epoch": 22.35, "learning_rate": 1.9138140395512892e-05, "loss": 0.6555, "step": 258120 }, { "epoch": 22.36, "learning_rate": 1.9135972013415055e-05, "loss": 0.7695, "step": 258130 }, { "epoch": 22.36, "learning_rate": 1.9133803631317218e-05, "loss": 0.7301, "step": 258140 }, { "epoch": 22.36, "learning_rate": 1.913163524921938e-05, "loss": 0.5933, "step": 258150 }, { "epoch": 22.36, "learning_rate": 1.912946686712154e-05, "loss": 0.5543, "step": 258160 }, { "epoch": 22.36, "learning_rate": 1.9127298485023704e-05, "loss": 0.6544, "step": 258170 }, { "epoch": 22.36, "learning_rate": 1.912513010292587e-05, "loss": 0.6461, "step": 258180 }, { "epoch": 22.36, "learning_rate": 1.9122961720828033e-05, "loss": 0.644, "step": 258190 }, { "epoch": 22.36, "learning_rate": 1.9120793338730193e-05, "loss": 0.7078, "step": 258200 }, { "epoch": 22.36, "learning_rate": 1.9118624956632356e-05, "loss": 0.7205, "step": 258210 }, { "epoch": 22.36, "learning_rate": 1.911645657453452e-05, "loss": 0.72, "step": 258220 }, { "epoch": 22.36, "learning_rate": 1.911428819243668e-05, "loss": 0.5907, "step": 258230 }, { "epoch": 22.37, "learning_rate": 1.911211981033884e-05, "loss": 0.7117, "step": 258240 }, { "epoch": 22.37, "learning_rate": 1.9109951428241008e-05, "loss": 0.6732, "step": 258250 }, { "epoch": 22.37, "learning_rate": 1.910778304614317e-05, "loss": 0.6227, "step": 258260 }, { "epoch": 22.37, "learning_rate": 1.9105614664045334e-05, "loss": 0.6736, "step": 258270 }, { "epoch": 22.37, "learning_rate": 1.9103446281947493e-05, "loss": 0.5913, "step": 258280 }, { "epoch": 22.37, "learning_rate": 1.9101277899849656e-05, "loss": 0.6739, "step": 258290 }, { "epoch": 22.37, "learning_rate": 1.909910951775182e-05, "loss": 0.7094, "step": 258300 }, { "epoch": 22.37, "learning_rate": 1.9096941135653986e-05, "loss": 0.6651, "step": 258310 }, { "epoch": 22.37, "learning_rate": 1.9094772753556145e-05, "loss": 0.5989, "step": 258320 }, { "epoch": 22.37, "learning_rate": 1.9092604371458308e-05, "loss": 0.652, "step": 258330 }, { "epoch": 22.37, "learning_rate": 1.909043598936047e-05, "loss": 0.6481, "step": 258340 }, { "epoch": 22.37, "learning_rate": 1.9088267607262634e-05, "loss": 0.5969, "step": 258350 }, { "epoch": 22.38, "learning_rate": 1.9086099225164794e-05, "loss": 0.5954, "step": 258360 }, { "epoch": 22.38, "learning_rate": 1.9083930843066957e-05, "loss": 0.6882, "step": 258370 }, { "epoch": 22.38, "learning_rate": 1.9081762460969123e-05, "loss": 0.7465, "step": 258380 }, { "epoch": 22.38, "learning_rate": 1.9079594078871283e-05, "loss": 0.6032, "step": 258390 }, { "epoch": 22.38, "learning_rate": 1.9077425696773446e-05, "loss": 0.8186, "step": 258400 }, { "epoch": 22.38, "learning_rate": 1.907525731467561e-05, "loss": 0.7044, "step": 258410 }, { "epoch": 22.38, "learning_rate": 1.9073088932577772e-05, "loss": 0.7654, "step": 258420 }, { "epoch": 22.38, "learning_rate": 1.907092055047993e-05, "loss": 0.7107, "step": 258430 }, { "epoch": 22.38, "learning_rate": 1.9068752168382094e-05, "loss": 0.6266, "step": 258440 }, { "epoch": 22.38, "learning_rate": 1.9066583786284257e-05, "loss": 0.8061, "step": 258450 }, { "epoch": 22.38, "learning_rate": 1.9064415404186424e-05, "loss": 0.6089, "step": 258460 }, { "epoch": 22.39, "learning_rate": 1.9062247022088583e-05, "loss": 0.592, "step": 258470 }, { "epoch": 22.39, "learning_rate": 1.9060078639990746e-05, "loss": 0.6733, "step": 258480 }, { "epoch": 22.39, "learning_rate": 1.905791025789291e-05, "loss": 0.6714, "step": 258490 }, { "epoch": 22.39, "learning_rate": 1.9055741875795072e-05, "loss": 0.6139, "step": 258500 }, { "epoch": 22.39, "learning_rate": 1.9053573493697232e-05, "loss": 0.5313, "step": 258510 }, { "epoch": 22.39, "learning_rate": 1.9051405111599395e-05, "loss": 0.7365, "step": 258520 }, { "epoch": 22.39, "learning_rate": 1.904923672950156e-05, "loss": 0.6334, "step": 258530 }, { "epoch": 22.39, "learning_rate": 1.9047068347403724e-05, "loss": 0.6798, "step": 258540 }, { "epoch": 22.39, "learning_rate": 1.9044899965305884e-05, "loss": 0.7388, "step": 258550 }, { "epoch": 22.39, "learning_rate": 1.9042731583208047e-05, "loss": 0.6392, "step": 258560 }, { "epoch": 22.39, "learning_rate": 1.904056320111021e-05, "loss": 0.7629, "step": 258570 }, { "epoch": 22.39, "learning_rate": 1.9038394819012373e-05, "loss": 0.6649, "step": 258580 }, { "epoch": 22.4, "learning_rate": 1.9036226436914533e-05, "loss": 0.5638, "step": 258590 }, { "epoch": 22.4, "learning_rate": 1.90340580548167e-05, "loss": 0.5563, "step": 258600 }, { "epoch": 22.4, "learning_rate": 1.9031889672718862e-05, "loss": 0.6637, "step": 258610 }, { "epoch": 22.4, "learning_rate": 1.9029721290621025e-05, "loss": 0.6966, "step": 258620 }, { "epoch": 22.4, "learning_rate": 1.9027552908523185e-05, "loss": 0.6796, "step": 258630 }, { "epoch": 22.4, "learning_rate": 1.9025384526425348e-05, "loss": 0.6883, "step": 258640 }, { "epoch": 22.4, "learning_rate": 1.902321614432751e-05, "loss": 0.7059, "step": 258650 }, { "epoch": 22.4, "learning_rate": 1.902104776222967e-05, "loss": 0.5731, "step": 258660 }, { "epoch": 22.4, "learning_rate": 1.9018879380131837e-05, "loss": 0.6543, "step": 258670 }, { "epoch": 22.4, "learning_rate": 1.9016710998034e-05, "loss": 0.7592, "step": 258680 }, { "epoch": 22.4, "learning_rate": 1.9014542615936163e-05, "loss": 0.6988, "step": 258690 }, { "epoch": 22.41, "learning_rate": 1.9012374233838322e-05, "loss": 0.6206, "step": 258700 }, { "epoch": 22.41, "learning_rate": 1.9010205851740485e-05, "loss": 0.5908, "step": 258710 }, { "epoch": 22.41, "learning_rate": 1.900803746964265e-05, "loss": 0.7209, "step": 258720 }, { "epoch": 22.41, "learning_rate": 1.900586908754481e-05, "loss": 0.6131, "step": 258730 }, { "epoch": 22.41, "learning_rate": 1.900370070544697e-05, "loss": 0.6304, "step": 258740 }, { "epoch": 22.41, "learning_rate": 1.9001532323349137e-05, "loss": 0.7472, "step": 258750 }, { "epoch": 22.41, "learning_rate": 1.89993639412513e-05, "loss": 0.6379, "step": 258760 }, { "epoch": 22.41, "learning_rate": 1.8997195559153463e-05, "loss": 0.7265, "step": 258770 }, { "epoch": 22.41, "learning_rate": 1.8995027177055623e-05, "loss": 0.608, "step": 258780 }, { "epoch": 22.41, "learning_rate": 1.8992858794957786e-05, "loss": 0.5845, "step": 258790 }, { "epoch": 22.41, "learning_rate": 1.899069041285995e-05, "loss": 0.6874, "step": 258800 }, { "epoch": 22.41, "learning_rate": 1.8988522030762115e-05, "loss": 0.6385, "step": 258810 }, { "epoch": 22.42, "learning_rate": 1.8986353648664275e-05, "loss": 0.7728, "step": 258820 }, { "epoch": 22.42, "learning_rate": 1.8984185266566438e-05, "loss": 0.7034, "step": 258830 }, { "epoch": 22.42, "learning_rate": 1.89820168844686e-05, "loss": 0.7177, "step": 258840 }, { "epoch": 22.42, "learning_rate": 1.8979848502370764e-05, "loss": 0.7564, "step": 258850 }, { "epoch": 22.42, "learning_rate": 1.8977680120272924e-05, "loss": 0.6868, "step": 258860 }, { "epoch": 22.42, "learning_rate": 1.8975511738175087e-05, "loss": 0.6384, "step": 258870 }, { "epoch": 22.42, "learning_rate": 1.8973343356077253e-05, "loss": 0.6365, "step": 258880 }, { "epoch": 22.42, "learning_rate": 1.8971174973979413e-05, "loss": 0.7309, "step": 258890 }, { "epoch": 22.42, "learning_rate": 1.8969006591881576e-05, "loss": 0.6723, "step": 258900 }, { "epoch": 22.42, "learning_rate": 1.896683820978374e-05, "loss": 0.7482, "step": 258910 }, { "epoch": 22.42, "learning_rate": 1.89646698276859e-05, "loss": 0.8146, "step": 258920 }, { "epoch": 22.42, "learning_rate": 1.896250144558806e-05, "loss": 0.6956, "step": 258930 }, { "epoch": 22.43, "learning_rate": 1.8960333063490224e-05, "loss": 0.6328, "step": 258940 }, { "epoch": 22.43, "learning_rate": 1.895816468139239e-05, "loss": 0.7498, "step": 258950 }, { "epoch": 22.43, "learning_rate": 1.8955996299294554e-05, "loss": 0.7532, "step": 258960 }, { "epoch": 22.43, "learning_rate": 1.8953827917196713e-05, "loss": 0.7268, "step": 258970 }, { "epoch": 22.43, "learning_rate": 1.8951659535098876e-05, "loss": 0.6673, "step": 258980 }, { "epoch": 22.43, "learning_rate": 1.894949115300104e-05, "loss": 0.6162, "step": 258990 }, { "epoch": 22.43, "learning_rate": 1.8947322770903202e-05, "loss": 0.6872, "step": 259000 }, { "epoch": 22.43, "learning_rate": 1.8945154388805362e-05, "loss": 0.6391, "step": 259010 }, { "epoch": 22.43, "learning_rate": 1.8942986006707525e-05, "loss": 0.5508, "step": 259020 }, { "epoch": 22.43, "learning_rate": 1.894081762460969e-05, "loss": 0.6821, "step": 259030 }, { "epoch": 22.43, "learning_rate": 1.8938649242511854e-05, "loss": 0.7124, "step": 259040 }, { "epoch": 22.44, "learning_rate": 1.8936480860414014e-05, "loss": 0.7563, "step": 259050 }, { "epoch": 22.44, "learning_rate": 1.8934312478316177e-05, "loss": 0.6225, "step": 259060 }, { "epoch": 22.44, "learning_rate": 1.893214409621834e-05, "loss": 0.7237, "step": 259070 }, { "epoch": 22.44, "learning_rate": 1.8929975714120503e-05, "loss": 0.8192, "step": 259080 }, { "epoch": 22.44, "learning_rate": 1.8927807332022663e-05, "loss": 0.6061, "step": 259090 }, { "epoch": 22.44, "learning_rate": 1.892563894992483e-05, "loss": 0.6239, "step": 259100 }, { "epoch": 22.44, "learning_rate": 1.8923470567826992e-05, "loss": 0.7849, "step": 259110 }, { "epoch": 22.44, "learning_rate": 1.8921302185729155e-05, "loss": 0.7691, "step": 259120 }, { "epoch": 22.44, "learning_rate": 1.8919133803631315e-05, "loss": 0.6208, "step": 259130 }, { "epoch": 22.44, "learning_rate": 1.8916965421533478e-05, "loss": 0.7091, "step": 259140 }, { "epoch": 22.44, "learning_rate": 1.891479703943564e-05, "loss": 0.7855, "step": 259150 }, { "epoch": 22.44, "learning_rate": 1.89126286573378e-05, "loss": 0.6241, "step": 259160 }, { "epoch": 22.45, "learning_rate": 1.8910460275239967e-05, "loss": 0.7407, "step": 259170 }, { "epoch": 22.45, "learning_rate": 1.890829189314213e-05, "loss": 0.7591, "step": 259180 }, { "epoch": 22.45, "learning_rate": 1.8906123511044293e-05, "loss": 0.6882, "step": 259190 }, { "epoch": 22.45, "learning_rate": 1.8903955128946452e-05, "loss": 0.7036, "step": 259200 }, { "epoch": 22.45, "learning_rate": 1.8901786746848615e-05, "loss": 0.715, "step": 259210 }, { "epoch": 22.45, "learning_rate": 1.8899618364750778e-05, "loss": 0.6538, "step": 259220 }, { "epoch": 22.45, "learning_rate": 1.8897449982652945e-05, "loss": 0.7366, "step": 259230 }, { "epoch": 22.45, "learning_rate": 1.88952816005551e-05, "loss": 0.6637, "step": 259240 }, { "epoch": 22.45, "learning_rate": 1.8893113218457267e-05, "loss": 0.671, "step": 259250 }, { "epoch": 22.45, "learning_rate": 1.889094483635943e-05, "loss": 0.7621, "step": 259260 }, { "epoch": 22.45, "learning_rate": 1.8888776454261593e-05, "loss": 0.7179, "step": 259270 }, { "epoch": 22.46, "learning_rate": 1.8886608072163753e-05, "loss": 0.6293, "step": 259280 }, { "epoch": 22.46, "learning_rate": 1.8884439690065916e-05, "loss": 0.6704, "step": 259290 }, { "epoch": 22.46, "learning_rate": 1.888227130796808e-05, "loss": 0.5971, "step": 259300 }, { "epoch": 22.46, "learning_rate": 1.8880102925870245e-05, "loss": 0.6164, "step": 259310 }, { "epoch": 22.46, "learning_rate": 1.8877934543772405e-05, "loss": 0.76, "step": 259320 }, { "epoch": 22.46, "learning_rate": 1.8875766161674568e-05, "loss": 0.5513, "step": 259330 }, { "epoch": 22.46, "learning_rate": 1.887359777957673e-05, "loss": 0.7647, "step": 259340 }, { "epoch": 22.46, "learning_rate": 1.8871429397478894e-05, "loss": 0.582, "step": 259350 }, { "epoch": 22.46, "learning_rate": 1.8869261015381053e-05, "loss": 0.5088, "step": 259360 }, { "epoch": 22.46, "learning_rate": 1.8867092633283216e-05, "loss": 0.6276, "step": 259370 }, { "epoch": 22.46, "learning_rate": 1.8864924251185383e-05, "loss": 0.5483, "step": 259380 }, { "epoch": 22.46, "learning_rate": 1.8862755869087546e-05, "loss": 0.6455, "step": 259390 }, { "epoch": 22.47, "learning_rate": 1.8860587486989706e-05, "loss": 0.8228, "step": 259400 }, { "epoch": 22.47, "learning_rate": 1.885841910489187e-05, "loss": 0.6987, "step": 259410 }, { "epoch": 22.47, "learning_rate": 1.885625072279403e-05, "loss": 0.5708, "step": 259420 }, { "epoch": 22.47, "learning_rate": 1.885408234069619e-05, "loss": 0.5834, "step": 259430 }, { "epoch": 22.47, "learning_rate": 1.8851913958598354e-05, "loss": 0.7524, "step": 259440 }, { "epoch": 22.47, "learning_rate": 1.884974557650052e-05, "loss": 0.6505, "step": 259450 }, { "epoch": 22.47, "learning_rate": 1.8847577194402684e-05, "loss": 0.4698, "step": 259460 }, { "epoch": 22.47, "learning_rate": 1.8845408812304843e-05, "loss": 0.6662, "step": 259470 }, { "epoch": 22.47, "learning_rate": 1.8843240430207006e-05, "loss": 0.5692, "step": 259480 }, { "epoch": 22.47, "learning_rate": 1.884107204810917e-05, "loss": 0.6375, "step": 259490 }, { "epoch": 22.47, "learning_rate": 1.8838903666011332e-05, "loss": 0.6036, "step": 259500 }, { "epoch": 22.48, "learning_rate": 1.8836735283913492e-05, "loss": 0.6684, "step": 259510 }, { "epoch": 22.48, "learning_rate": 1.8834566901815655e-05, "loss": 0.6745, "step": 259520 }, { "epoch": 22.48, "learning_rate": 1.883239851971782e-05, "loss": 0.6357, "step": 259530 }, { "epoch": 22.48, "learning_rate": 1.8830230137619984e-05, "loss": 0.8315, "step": 259540 }, { "epoch": 22.48, "learning_rate": 1.8828061755522144e-05, "loss": 0.7838, "step": 259550 }, { "epoch": 22.48, "learning_rate": 1.8825893373424307e-05, "loss": 0.6505, "step": 259560 }, { "epoch": 22.48, "learning_rate": 1.882372499132647e-05, "loss": 0.6163, "step": 259570 }, { "epoch": 22.48, "learning_rate": 1.8821556609228633e-05, "loss": 0.715, "step": 259580 }, { "epoch": 22.48, "learning_rate": 1.8819388227130792e-05, "loss": 0.5749, "step": 259590 }, { "epoch": 22.48, "learning_rate": 1.881721984503296e-05, "loss": 0.8008, "step": 259600 }, { "epoch": 22.48, "learning_rate": 1.8815051462935122e-05, "loss": 0.7377, "step": 259610 }, { "epoch": 22.48, "learning_rate": 1.8812883080837285e-05, "loss": 0.8316, "step": 259620 }, { "epoch": 22.49, "learning_rate": 1.8810714698739444e-05, "loss": 0.697, "step": 259630 }, { "epoch": 22.49, "learning_rate": 1.8808546316641607e-05, "loss": 0.7674, "step": 259640 }, { "epoch": 22.49, "learning_rate": 1.880637793454377e-05, "loss": 0.5711, "step": 259650 }, { "epoch": 22.49, "learning_rate": 1.880420955244593e-05, "loss": 0.7745, "step": 259660 }, { "epoch": 22.49, "learning_rate": 1.8802041170348096e-05, "loss": 0.6905, "step": 259670 }, { "epoch": 22.49, "learning_rate": 1.879987278825026e-05, "loss": 0.5883, "step": 259680 }, { "epoch": 22.49, "learning_rate": 1.8797704406152422e-05, "loss": 0.7996, "step": 259690 }, { "epoch": 22.49, "learning_rate": 1.8795536024054582e-05, "loss": 0.6796, "step": 259700 }, { "epoch": 22.49, "learning_rate": 1.8793367641956745e-05, "loss": 0.4895, "step": 259710 }, { "epoch": 22.49, "learning_rate": 1.8791199259858908e-05, "loss": 0.6394, "step": 259720 }, { "epoch": 22.49, "learning_rate": 1.8789030877761074e-05, "loss": 0.7061, "step": 259730 }, { "epoch": 22.5, "learning_rate": 1.8786862495663234e-05, "loss": 0.7615, "step": 259740 }, { "epoch": 22.5, "learning_rate": 1.8784694113565397e-05, "loss": 0.6118, "step": 259750 }, { "epoch": 22.5, "learning_rate": 1.878252573146756e-05, "loss": 0.6445, "step": 259760 }, { "epoch": 22.5, "learning_rate": 1.8780357349369723e-05, "loss": 0.6313, "step": 259770 }, { "epoch": 22.5, "learning_rate": 1.8778188967271883e-05, "loss": 0.746, "step": 259780 }, { "epoch": 22.5, "learning_rate": 1.8776020585174046e-05, "loss": 0.6136, "step": 259790 }, { "epoch": 22.5, "learning_rate": 1.877385220307621e-05, "loss": 0.6346, "step": 259800 }, { "epoch": 22.5, "learning_rate": 1.8771683820978375e-05, "loss": 0.7033, "step": 259810 }, { "epoch": 22.5, "learning_rate": 1.8769515438880535e-05, "loss": 0.5966, "step": 259820 }, { "epoch": 22.5, "learning_rate": 1.8767347056782698e-05, "loss": 0.4358, "step": 259830 }, { "epoch": 22.5, "learning_rate": 1.876517867468486e-05, "loss": 0.7674, "step": 259840 }, { "epoch": 22.5, "learning_rate": 1.8763010292587024e-05, "loss": 0.808, "step": 259850 }, { "epoch": 22.51, "learning_rate": 1.8760841910489183e-05, "loss": 0.6282, "step": 259860 }, { "epoch": 22.51, "learning_rate": 1.8758673528391346e-05, "loss": 0.6767, "step": 259870 }, { "epoch": 22.51, "learning_rate": 1.8756505146293513e-05, "loss": 0.6897, "step": 259880 }, { "epoch": 22.51, "learning_rate": 1.8754336764195676e-05, "loss": 0.5548, "step": 259890 }, { "epoch": 22.51, "learning_rate": 1.8752168382097835e-05, "loss": 0.6064, "step": 259900 }, { "epoch": 22.51, "learning_rate": 1.875e-05, "loss": 0.6266, "step": 259910 }, { "epoch": 22.51, "learning_rate": 1.874783161790216e-05, "loss": 0.699, "step": 259920 }, { "epoch": 22.51, "learning_rate": 1.8745663235804324e-05, "loss": 0.5567, "step": 259930 }, { "epoch": 22.51, "learning_rate": 1.8743494853706484e-05, "loss": 0.5922, "step": 259940 }, { "epoch": 22.51, "learning_rate": 1.874132647160865e-05, "loss": 0.4634, "step": 259950 }, { "epoch": 22.51, "learning_rate": 1.873915808951081e-05, "loss": 0.5981, "step": 259960 }, { "epoch": 22.52, "learning_rate": 1.8736989707412973e-05, "loss": 0.576, "step": 259970 }, { "epoch": 22.52, "learning_rate": 1.8734821325315136e-05, "loss": 0.6742, "step": 259980 }, { "epoch": 22.52, "learning_rate": 1.87326529432173e-05, "loss": 0.7297, "step": 259990 }, { "epoch": 22.52, "learning_rate": 1.8730484561119462e-05, "loss": 0.5644, "step": 260000 }, { "epoch": 22.52, "learning_rate": 1.8728316179021625e-05, "loss": 0.6938, "step": 260010 }, { "epoch": 22.52, "learning_rate": 1.8726147796923788e-05, "loss": 0.5511, "step": 260020 }, { "epoch": 22.52, "learning_rate": 1.872397941482595e-05, "loss": 0.6684, "step": 260030 }, { "epoch": 22.52, "learning_rate": 1.872181103272811e-05, "loss": 0.6264, "step": 260040 }, { "epoch": 22.52, "learning_rate": 1.8719642650630274e-05, "loss": 0.5774, "step": 260050 }, { "epoch": 22.52, "learning_rate": 1.8717474268532437e-05, "loss": 0.668, "step": 260060 }, { "epoch": 22.52, "learning_rate": 1.87153058864346e-05, "loss": 0.7868, "step": 260070 }, { "epoch": 22.52, "learning_rate": 1.8713137504336763e-05, "loss": 0.7446, "step": 260080 }, { "epoch": 22.53, "learning_rate": 1.8710969122238926e-05, "loss": 0.6637, "step": 260090 }, { "epoch": 22.53, "learning_rate": 1.870880074014109e-05, "loss": 0.5122, "step": 260100 }, { "epoch": 22.53, "learning_rate": 1.870663235804325e-05, "loss": 0.7434, "step": 260110 }, { "epoch": 22.53, "learning_rate": 1.870446397594541e-05, "loss": 0.7197, "step": 260120 }, { "epoch": 22.53, "learning_rate": 1.8702295593847578e-05, "loss": 0.8536, "step": 260130 }, { "epoch": 22.53, "learning_rate": 1.8700127211749737e-05, "loss": 0.5635, "step": 260140 }, { "epoch": 22.53, "learning_rate": 1.86979588296519e-05, "loss": 0.6647, "step": 260150 }, { "epoch": 22.53, "learning_rate": 1.8695790447554063e-05, "loss": 0.666, "step": 260160 }, { "epoch": 22.53, "learning_rate": 1.8693622065456226e-05, "loss": 0.6371, "step": 260170 }, { "epoch": 22.53, "learning_rate": 1.869145368335839e-05, "loss": 0.7341, "step": 260180 }, { "epoch": 22.53, "learning_rate": 1.868928530126055e-05, "loss": 0.6472, "step": 260190 }, { "epoch": 22.53, "learning_rate": 1.8687116919162715e-05, "loss": 0.7499, "step": 260200 }, { "epoch": 22.54, "learning_rate": 1.8684948537064875e-05, "loss": 0.6048, "step": 260210 }, { "epoch": 22.54, "learning_rate": 1.8682780154967038e-05, "loss": 0.6563, "step": 260220 }, { "epoch": 22.54, "learning_rate": 1.86806117728692e-05, "loss": 0.7789, "step": 260230 }, { "epoch": 22.54, "learning_rate": 1.8678443390771364e-05, "loss": 0.6135, "step": 260240 }, { "epoch": 22.54, "learning_rate": 1.8676275008673527e-05, "loss": 0.7452, "step": 260250 }, { "epoch": 22.54, "learning_rate": 1.867410662657569e-05, "loss": 0.6658, "step": 260260 }, { "epoch": 22.54, "learning_rate": 1.8671938244477853e-05, "loss": 0.8246, "step": 260270 }, { "epoch": 22.54, "learning_rate": 1.8669769862380016e-05, "loss": 0.624, "step": 260280 }, { "epoch": 22.54, "learning_rate": 1.8667601480282176e-05, "loss": 0.6436, "step": 260290 }, { "epoch": 22.54, "learning_rate": 1.8665433098184342e-05, "loss": 0.8168, "step": 260300 }, { "epoch": 22.54, "learning_rate": 1.86632647160865e-05, "loss": 0.6933, "step": 260310 }, { "epoch": 22.55, "learning_rate": 1.8661096333988665e-05, "loss": 0.6692, "step": 260320 }, { "epoch": 22.55, "learning_rate": 1.8658927951890828e-05, "loss": 0.5867, "step": 260330 }, { "epoch": 22.55, "learning_rate": 1.865675956979299e-05, "loss": 0.7242, "step": 260340 }, { "epoch": 22.55, "learning_rate": 1.8654591187695154e-05, "loss": 0.6833, "step": 260350 }, { "epoch": 22.55, "learning_rate": 1.8652422805597317e-05, "loss": 0.6897, "step": 260360 }, { "epoch": 22.55, "learning_rate": 1.8650254423499476e-05, "loss": 0.8378, "step": 260370 }, { "epoch": 22.55, "learning_rate": 1.8648086041401643e-05, "loss": 0.6449, "step": 260380 }, { "epoch": 22.55, "learning_rate": 1.8645917659303802e-05, "loss": 0.7453, "step": 260390 }, { "epoch": 22.55, "learning_rate": 1.8643749277205965e-05, "loss": 0.6898, "step": 260400 }, { "epoch": 22.55, "learning_rate": 1.8641580895108128e-05, "loss": 0.6218, "step": 260410 }, { "epoch": 22.55, "learning_rate": 1.863941251301029e-05, "loss": 0.7825, "step": 260420 }, { "epoch": 22.55, "learning_rate": 1.8637244130912454e-05, "loss": 0.6955, "step": 260430 }, { "epoch": 22.56, "learning_rate": 1.8635075748814614e-05, "loss": 0.8046, "step": 260440 }, { "epoch": 22.56, "learning_rate": 1.863290736671678e-05, "loss": 0.6138, "step": 260450 }, { "epoch": 22.56, "learning_rate": 1.863073898461894e-05, "loss": 0.588, "step": 260460 }, { "epoch": 22.56, "learning_rate": 1.8628570602521103e-05, "loss": 0.7063, "step": 260470 }, { "epoch": 22.56, "learning_rate": 1.8626402220423266e-05, "loss": 0.7193, "step": 260480 }, { "epoch": 22.56, "learning_rate": 1.862423383832543e-05, "loss": 0.7135, "step": 260490 }, { "epoch": 22.56, "learning_rate": 1.8622065456227592e-05, "loss": 0.6674, "step": 260500 }, { "epoch": 22.56, "learning_rate": 1.8619897074129755e-05, "loss": 0.7953, "step": 260510 }, { "epoch": 22.56, "learning_rate": 1.8617728692031918e-05, "loss": 0.6351, "step": 260520 }, { "epoch": 22.56, "learning_rate": 1.861556030993408e-05, "loss": 0.7122, "step": 260530 }, { "epoch": 22.56, "learning_rate": 1.861339192783624e-05, "loss": 0.5489, "step": 260540 }, { "epoch": 22.57, "learning_rate": 1.8611223545738407e-05, "loss": 0.765, "step": 260550 }, { "epoch": 22.57, "learning_rate": 1.8609055163640566e-05, "loss": 0.6533, "step": 260560 }, { "epoch": 22.57, "learning_rate": 1.860688678154273e-05, "loss": 0.6755, "step": 260570 }, { "epoch": 22.57, "learning_rate": 1.8604718399444892e-05, "loss": 0.6014, "step": 260580 }, { "epoch": 22.57, "learning_rate": 1.8602550017347055e-05, "loss": 0.7958, "step": 260590 }, { "epoch": 22.57, "learning_rate": 1.860038163524922e-05, "loss": 0.7404, "step": 260600 }, { "epoch": 22.57, "learning_rate": 1.859821325315138e-05, "loss": 0.6529, "step": 260610 }, { "epoch": 22.57, "learning_rate": 1.859604487105354e-05, "loss": 0.695, "step": 260620 }, { "epoch": 22.57, "learning_rate": 1.8593876488955708e-05, "loss": 0.765, "step": 260630 }, { "epoch": 22.57, "learning_rate": 1.8591708106857867e-05, "loss": 0.7386, "step": 260640 }, { "epoch": 22.57, "learning_rate": 1.858953972476003e-05, "loss": 0.7413, "step": 260650 }, { "epoch": 22.57, "learning_rate": 1.8587371342662193e-05, "loss": 0.6797, "step": 260660 }, { "epoch": 22.58, "learning_rate": 1.8585202960564356e-05, "loss": 0.7737, "step": 260670 }, { "epoch": 22.58, "learning_rate": 1.858303457846652e-05, "loss": 0.5688, "step": 260680 }, { "epoch": 22.58, "learning_rate": 1.858086619636868e-05, "loss": 0.5711, "step": 260690 }, { "epoch": 22.58, "learning_rate": 1.8578697814270845e-05, "loss": 0.7265, "step": 260700 }, { "epoch": 22.58, "learning_rate": 1.8576529432173005e-05, "loss": 0.6152, "step": 260710 }, { "epoch": 22.58, "learning_rate": 1.8574361050075168e-05, "loss": 0.7277, "step": 260720 }, { "epoch": 22.58, "learning_rate": 1.857219266797733e-05, "loss": 0.7088, "step": 260730 }, { "epoch": 22.58, "learning_rate": 1.8570024285879494e-05, "loss": 0.6793, "step": 260740 }, { "epoch": 22.58, "learning_rate": 1.8567855903781657e-05, "loss": 0.6838, "step": 260750 }, { "epoch": 22.58, "learning_rate": 1.856568752168382e-05, "loss": 0.5943, "step": 260760 }, { "epoch": 22.58, "learning_rate": 1.8563519139585983e-05, "loss": 0.7126, "step": 260770 }, { "epoch": 22.59, "learning_rate": 1.8561350757488146e-05, "loss": 0.5875, "step": 260780 }, { "epoch": 22.59, "learning_rate": 1.8559182375390305e-05, "loss": 0.8085, "step": 260790 }, { "epoch": 22.59, "learning_rate": 1.8557013993292472e-05, "loss": 0.6485, "step": 260800 }, { "epoch": 22.59, "learning_rate": 1.855484561119463e-05, "loss": 0.7157, "step": 260810 }, { "epoch": 22.59, "learning_rate": 1.8552677229096794e-05, "loss": 0.6795, "step": 260820 }, { "epoch": 22.59, "learning_rate": 1.8550508846998957e-05, "loss": 0.7058, "step": 260830 }, { "epoch": 22.59, "learning_rate": 1.854834046490112e-05, "loss": 0.6986, "step": 260840 }, { "epoch": 22.59, "learning_rate": 1.8546172082803283e-05, "loss": 0.6303, "step": 260850 }, { "epoch": 22.59, "learning_rate": 1.8544003700705446e-05, "loss": 0.5742, "step": 260860 }, { "epoch": 22.59, "learning_rate": 1.8541835318607606e-05, "loss": 0.741, "step": 260870 }, { "epoch": 22.59, "learning_rate": 1.8539666936509772e-05, "loss": 0.5534, "step": 260880 }, { "epoch": 22.59, "learning_rate": 1.8537498554411932e-05, "loss": 0.6576, "step": 260890 }, { "epoch": 22.6, "learning_rate": 1.8535330172314095e-05, "loss": 0.6178, "step": 260900 }, { "epoch": 22.6, "learning_rate": 1.8533161790216258e-05, "loss": 0.6029, "step": 260910 }, { "epoch": 22.6, "learning_rate": 1.853099340811842e-05, "loss": 0.677, "step": 260920 }, { "epoch": 22.6, "learning_rate": 1.8528825026020584e-05, "loss": 0.5726, "step": 260930 }, { "epoch": 22.6, "learning_rate": 1.8526656643922744e-05, "loss": 0.7248, "step": 260940 }, { "epoch": 22.6, "learning_rate": 1.852448826182491e-05, "loss": 0.6475, "step": 260950 }, { "epoch": 22.6, "learning_rate": 1.852231987972707e-05, "loss": 0.7299, "step": 260960 }, { "epoch": 22.6, "learning_rate": 1.8520151497629233e-05, "loss": 0.7158, "step": 260970 }, { "epoch": 22.6, "learning_rate": 1.8517983115531396e-05, "loss": 0.6231, "step": 260980 }, { "epoch": 22.6, "learning_rate": 1.851581473343356e-05, "loss": 0.8703, "step": 260990 }, { "epoch": 22.6, "learning_rate": 1.8513646351335722e-05, "loss": 0.8166, "step": 261000 }, { "epoch": 22.61, "learning_rate": 1.8511477969237885e-05, "loss": 0.5501, "step": 261010 }, { "epoch": 22.61, "learning_rate": 1.8509309587140048e-05, "loss": 0.66, "step": 261020 }, { "epoch": 22.61, "learning_rate": 1.850714120504221e-05, "loss": 0.6828, "step": 261030 }, { "epoch": 22.61, "learning_rate": 1.850497282294437e-05, "loss": 0.6297, "step": 261040 }, { "epoch": 22.61, "learning_rate": 1.8502804440846537e-05, "loss": 0.7279, "step": 261050 }, { "epoch": 22.61, "learning_rate": 1.8500636058748696e-05, "loss": 0.7413, "step": 261060 }, { "epoch": 22.61, "learning_rate": 1.849846767665086e-05, "loss": 0.7522, "step": 261070 }, { "epoch": 22.61, "learning_rate": 1.8496299294553022e-05, "loss": 0.7388, "step": 261080 }, { "epoch": 22.61, "learning_rate": 1.8494130912455185e-05, "loss": 0.6377, "step": 261090 }, { "epoch": 22.61, "learning_rate": 1.849196253035735e-05, "loss": 0.6327, "step": 261100 }, { "epoch": 22.61, "learning_rate": 1.848979414825951e-05, "loss": 0.7712, "step": 261110 }, { "epoch": 22.61, "learning_rate": 1.848762576616167e-05, "loss": 0.6232, "step": 261120 }, { "epoch": 22.62, "learning_rate": 1.8485457384063837e-05, "loss": 0.6589, "step": 261130 }, { "epoch": 22.62, "learning_rate": 1.8483289001965997e-05, "loss": 0.6139, "step": 261140 }, { "epoch": 22.62, "learning_rate": 1.848112061986816e-05, "loss": 0.7108, "step": 261150 }, { "epoch": 22.62, "learning_rate": 1.8478952237770323e-05, "loss": 0.7125, "step": 261160 }, { "epoch": 22.62, "learning_rate": 1.8476783855672486e-05, "loss": 0.7191, "step": 261170 }, { "epoch": 22.62, "learning_rate": 1.847461547357465e-05, "loss": 0.5626, "step": 261180 }, { "epoch": 22.62, "learning_rate": 1.8472447091476812e-05, "loss": 0.7188, "step": 261190 }, { "epoch": 22.62, "learning_rate": 1.8470278709378975e-05, "loss": 0.7268, "step": 261200 }, { "epoch": 22.62, "learning_rate": 1.8468110327281135e-05, "loss": 0.7598, "step": 261210 }, { "epoch": 22.62, "learning_rate": 1.8465941945183298e-05, "loss": 0.6426, "step": 261220 }, { "epoch": 22.62, "learning_rate": 1.846377356308546e-05, "loss": 0.7389, "step": 261230 }, { "epoch": 22.63, "learning_rate": 1.8461605180987624e-05, "loss": 0.7, "step": 261240 }, { "epoch": 22.63, "learning_rate": 1.8459436798889787e-05, "loss": 0.6874, "step": 261250 }, { "epoch": 22.63, "learning_rate": 1.845726841679195e-05, "loss": 0.7038, "step": 261260 }, { "epoch": 22.63, "learning_rate": 1.8455100034694113e-05, "loss": 0.755, "step": 261270 }, { "epoch": 22.63, "learning_rate": 1.8452931652596276e-05, "loss": 0.7229, "step": 261280 }, { "epoch": 22.63, "learning_rate": 1.8450763270498435e-05, "loss": 0.7801, "step": 261290 }, { "epoch": 22.63, "learning_rate": 1.84485948884006e-05, "loss": 0.6971, "step": 261300 }, { "epoch": 22.63, "learning_rate": 1.844642650630276e-05, "loss": 0.6171, "step": 261310 }, { "epoch": 22.63, "learning_rate": 1.8444258124204924e-05, "loss": 0.6107, "step": 261320 }, { "epoch": 22.63, "learning_rate": 1.8442089742107087e-05, "loss": 0.6678, "step": 261330 }, { "epoch": 22.63, "learning_rate": 1.843992136000925e-05, "loss": 0.6399, "step": 261340 }, { "epoch": 22.63, "learning_rate": 1.8437752977911413e-05, "loss": 0.7409, "step": 261350 }, { "epoch": 22.64, "learning_rate": 1.8435584595813576e-05, "loss": 0.7041, "step": 261360 }, { "epoch": 22.64, "learning_rate": 1.843341621371574e-05, "loss": 0.6528, "step": 261370 }, { "epoch": 22.64, "learning_rate": 1.8431247831617902e-05, "loss": 0.7169, "step": 261380 }, { "epoch": 22.64, "learning_rate": 1.8429079449520062e-05, "loss": 0.5624, "step": 261390 }, { "epoch": 22.64, "learning_rate": 1.8426911067422225e-05, "loss": 0.6503, "step": 261400 }, { "epoch": 22.64, "learning_rate": 1.8424742685324388e-05, "loss": 0.7708, "step": 261410 }, { "epoch": 22.64, "learning_rate": 1.842257430322655e-05, "loss": 0.635, "step": 261420 }, { "epoch": 22.64, "learning_rate": 1.8420405921128714e-05, "loss": 0.7027, "step": 261430 }, { "epoch": 22.64, "learning_rate": 1.8418237539030877e-05, "loss": 0.6336, "step": 261440 }, { "epoch": 22.64, "learning_rate": 1.841606915693304e-05, "loss": 0.6223, "step": 261450 }, { "epoch": 22.64, "learning_rate": 1.84139007748352e-05, "loss": 0.7212, "step": 261460 }, { "epoch": 22.64, "learning_rate": 1.8411732392737363e-05, "loss": 0.6828, "step": 261470 }, { "epoch": 22.65, "learning_rate": 1.8409564010639526e-05, "loss": 0.5732, "step": 261480 }, { "epoch": 22.65, "learning_rate": 1.840739562854169e-05, "loss": 0.6243, "step": 261490 }, { "epoch": 22.65, "learning_rate": 1.840522724644385e-05, "loss": 0.6279, "step": 261500 }, { "epoch": 22.65, "learning_rate": 1.8403058864346015e-05, "loss": 0.7553, "step": 261510 }, { "epoch": 22.65, "learning_rate": 1.8400890482248178e-05, "loss": 0.5911, "step": 261520 }, { "epoch": 22.65, "learning_rate": 1.839872210015034e-05, "loss": 0.6943, "step": 261530 }, { "epoch": 22.65, "learning_rate": 1.83965537180525e-05, "loss": 0.6787, "step": 261540 }, { "epoch": 22.65, "learning_rate": 1.8394385335954667e-05, "loss": 0.6477, "step": 261550 }, { "epoch": 22.65, "learning_rate": 1.8392216953856826e-05, "loss": 0.6641, "step": 261560 }, { "epoch": 22.65, "learning_rate": 1.839004857175899e-05, "loss": 0.5798, "step": 261570 }, { "epoch": 22.65, "learning_rate": 1.8387880189661152e-05, "loss": 0.724, "step": 261580 }, { "epoch": 22.66, "learning_rate": 1.8385711807563315e-05, "loss": 0.5649, "step": 261590 }, { "epoch": 22.66, "learning_rate": 1.8383543425465478e-05, "loss": 0.6852, "step": 261600 }, { "epoch": 22.66, "learning_rate": 1.838137504336764e-05, "loss": 0.6269, "step": 261610 }, { "epoch": 22.66, "learning_rate": 1.8379206661269804e-05, "loss": 0.7389, "step": 261620 }, { "epoch": 22.66, "learning_rate": 1.8377038279171967e-05, "loss": 0.6931, "step": 261630 }, { "epoch": 22.66, "learning_rate": 1.8374869897074127e-05, "loss": 0.6396, "step": 261640 }, { "epoch": 22.66, "learning_rate": 1.8372701514976293e-05, "loss": 0.5421, "step": 261650 }, { "epoch": 22.66, "learning_rate": 1.8370533132878453e-05, "loss": 0.6268, "step": 261660 }, { "epoch": 22.66, "learning_rate": 1.8368364750780616e-05, "loss": 0.6662, "step": 261670 }, { "epoch": 22.66, "learning_rate": 1.836619636868278e-05, "loss": 0.8048, "step": 261680 }, { "epoch": 22.66, "learning_rate": 1.8364027986584942e-05, "loss": 0.6349, "step": 261690 }, { "epoch": 22.66, "learning_rate": 1.8361859604487105e-05, "loss": 0.6859, "step": 261700 }, { "epoch": 22.67, "learning_rate": 1.8359691222389264e-05, "loss": 0.6826, "step": 261710 }, { "epoch": 22.67, "learning_rate": 1.8357522840291427e-05, "loss": 0.7751, "step": 261720 }, { "epoch": 22.67, "learning_rate": 1.835535445819359e-05, "loss": 0.6594, "step": 261730 }, { "epoch": 22.67, "learning_rate": 1.8353186076095753e-05, "loss": 0.7259, "step": 261740 }, { "epoch": 22.67, "learning_rate": 1.8351017693997916e-05, "loss": 0.7073, "step": 261750 }, { "epoch": 22.67, "learning_rate": 1.834884931190008e-05, "loss": 0.7672, "step": 261760 }, { "epoch": 22.67, "learning_rate": 1.8346680929802242e-05, "loss": 0.7021, "step": 261770 }, { "epoch": 22.67, "learning_rate": 1.8344512547704405e-05, "loss": 0.9114, "step": 261780 }, { "epoch": 22.67, "learning_rate": 1.8342344165606565e-05, "loss": 0.7458, "step": 261790 }, { "epoch": 22.67, "learning_rate": 1.834017578350873e-05, "loss": 0.8211, "step": 261800 }, { "epoch": 22.67, "learning_rate": 1.833800740141089e-05, "loss": 0.6448, "step": 261810 }, { "epoch": 22.68, "learning_rate": 1.8335839019313054e-05, "loss": 0.6668, "step": 261820 }, { "epoch": 22.68, "learning_rate": 1.8333670637215217e-05, "loss": 0.7455, "step": 261830 }, { "epoch": 22.68, "learning_rate": 1.833150225511738e-05, "loss": 0.7732, "step": 261840 }, { "epoch": 22.68, "learning_rate": 1.8329333873019543e-05, "loss": 0.6226, "step": 261850 }, { "epoch": 22.68, "learning_rate": 1.8327165490921706e-05, "loss": 0.8102, "step": 261860 }, { "epoch": 22.68, "learning_rate": 1.832499710882387e-05, "loss": 0.6606, "step": 261870 }, { "epoch": 22.68, "learning_rate": 1.8322828726726032e-05, "loss": 0.7036, "step": 261880 }, { "epoch": 22.68, "learning_rate": 1.8320660344628192e-05, "loss": 0.6367, "step": 261890 }, { "epoch": 22.68, "learning_rate": 1.8318491962530358e-05, "loss": 0.74, "step": 261900 }, { "epoch": 22.68, "learning_rate": 1.8316323580432518e-05, "loss": 0.7228, "step": 261910 }, { "epoch": 22.68, "learning_rate": 1.831415519833468e-05, "loss": 0.5986, "step": 261920 }, { "epoch": 22.68, "learning_rate": 1.8311986816236844e-05, "loss": 0.655, "step": 261930 }, { "epoch": 22.69, "learning_rate": 1.8309818434139007e-05, "loss": 0.753, "step": 261940 }, { "epoch": 22.69, "learning_rate": 1.830765005204117e-05, "loss": 0.7756, "step": 261950 }, { "epoch": 22.69, "learning_rate": 1.8305481669943333e-05, "loss": 0.525, "step": 261960 }, { "epoch": 22.69, "learning_rate": 1.8303313287845492e-05, "loss": 0.7296, "step": 261970 }, { "epoch": 22.69, "learning_rate": 1.8301144905747655e-05, "loss": 0.6395, "step": 261980 }, { "epoch": 22.69, "learning_rate": 1.829897652364982e-05, "loss": 0.6873, "step": 261990 }, { "epoch": 22.69, "learning_rate": 1.829680814155198e-05, "loss": 0.6944, "step": 262000 }, { "epoch": 22.69, "learning_rate": 1.8294639759454144e-05, "loss": 0.5904, "step": 262010 }, { "epoch": 22.69, "learning_rate": 1.8292471377356307e-05, "loss": 0.8099, "step": 262020 }, { "epoch": 22.69, "learning_rate": 1.829030299525847e-05, "loss": 0.5811, "step": 262030 }, { "epoch": 22.69, "learning_rate": 1.828813461316063e-05, "loss": 0.7515, "step": 262040 }, { "epoch": 22.7, "learning_rate": 1.8285966231062796e-05, "loss": 0.6162, "step": 262050 }, { "epoch": 22.7, "learning_rate": 1.8283797848964956e-05, "loss": 0.8411, "step": 262060 }, { "epoch": 22.7, "learning_rate": 1.828162946686712e-05, "loss": 0.7359, "step": 262070 }, { "epoch": 22.7, "learning_rate": 1.8279461084769282e-05, "loss": 0.7824, "step": 262080 }, { "epoch": 22.7, "learning_rate": 1.8277292702671445e-05, "loss": 0.6766, "step": 262090 }, { "epoch": 22.7, "learning_rate": 1.8275124320573608e-05, "loss": 0.7238, "step": 262100 }, { "epoch": 22.7, "learning_rate": 1.827295593847577e-05, "loss": 0.6843, "step": 262110 }, { "epoch": 22.7, "learning_rate": 1.8270787556377934e-05, "loss": 0.6586, "step": 262120 }, { "epoch": 22.7, "learning_rate": 1.8268619174280097e-05, "loss": 0.7158, "step": 262130 }, { "epoch": 22.7, "learning_rate": 1.8266450792182257e-05, "loss": 0.7399, "step": 262140 }, { "epoch": 22.7, "learning_rate": 1.8264282410084423e-05, "loss": 0.6707, "step": 262150 }, { "epoch": 22.7, "learning_rate": 1.8262114027986583e-05, "loss": 0.6461, "step": 262160 }, { "epoch": 22.71, "learning_rate": 1.8259945645888746e-05, "loss": 0.7781, "step": 262170 }, { "epoch": 22.71, "learning_rate": 1.825777726379091e-05, "loss": 0.6718, "step": 262180 }, { "epoch": 22.71, "learning_rate": 1.825560888169307e-05, "loss": 0.7314, "step": 262190 }, { "epoch": 22.71, "learning_rate": 1.8253440499595235e-05, "loss": 0.6092, "step": 262200 }, { "epoch": 22.71, "learning_rate": 1.8251272117497398e-05, "loss": 0.6868, "step": 262210 }, { "epoch": 22.71, "learning_rate": 1.8249103735399557e-05, "loss": 0.6842, "step": 262220 }, { "epoch": 22.71, "learning_rate": 1.824693535330172e-05, "loss": 0.7396, "step": 262230 }, { "epoch": 22.71, "learning_rate": 1.8244766971203883e-05, "loss": 0.6569, "step": 262240 }, { "epoch": 22.71, "learning_rate": 1.8242598589106046e-05, "loss": 0.6249, "step": 262250 }, { "epoch": 22.71, "learning_rate": 1.824043020700821e-05, "loss": 0.7016, "step": 262260 }, { "epoch": 22.71, "learning_rate": 1.8238261824910372e-05, "loss": 0.6336, "step": 262270 }, { "epoch": 22.72, "learning_rate": 1.8236093442812535e-05, "loss": 0.7883, "step": 262280 }, { "epoch": 22.72, "learning_rate": 1.8233925060714695e-05, "loss": 0.7806, "step": 262290 }, { "epoch": 22.72, "learning_rate": 1.823175667861686e-05, "loss": 0.7108, "step": 262300 }, { "epoch": 22.72, "learning_rate": 1.822958829651902e-05, "loss": 0.7539, "step": 262310 }, { "epoch": 22.72, "learning_rate": 1.8227419914421184e-05, "loss": 0.5833, "step": 262320 }, { "epoch": 22.72, "learning_rate": 1.8225251532323347e-05, "loss": 0.8243, "step": 262330 }, { "epoch": 22.72, "learning_rate": 1.822308315022551e-05, "loss": 0.684, "step": 262340 }, { "epoch": 22.72, "learning_rate": 1.8220914768127673e-05, "loss": 0.6535, "step": 262350 }, { "epoch": 22.72, "learning_rate": 1.8218746386029836e-05, "loss": 0.8497, "step": 262360 }, { "epoch": 22.72, "learning_rate": 1.8216578003932e-05, "loss": 0.8365, "step": 262370 }, { "epoch": 22.72, "learning_rate": 1.8214409621834162e-05, "loss": 0.6561, "step": 262380 }, { "epoch": 22.72, "learning_rate": 1.821224123973632e-05, "loss": 0.6393, "step": 262390 }, { "epoch": 22.73, "learning_rate": 1.8210072857638488e-05, "loss": 0.8048, "step": 262400 }, { "epoch": 22.73, "learning_rate": 1.8207904475540648e-05, "loss": 0.6833, "step": 262410 }, { "epoch": 22.73, "learning_rate": 1.820573609344281e-05, "loss": 0.7235, "step": 262420 }, { "epoch": 22.73, "learning_rate": 1.8203567711344974e-05, "loss": 0.8663, "step": 262430 }, { "epoch": 22.73, "learning_rate": 1.8201399329247137e-05, "loss": 0.7454, "step": 262440 }, { "epoch": 22.73, "learning_rate": 1.81992309471493e-05, "loss": 0.7195, "step": 262450 }, { "epoch": 22.73, "learning_rate": 1.8197062565051463e-05, "loss": 0.6483, "step": 262460 }, { "epoch": 22.73, "learning_rate": 1.8194894182953622e-05, "loss": 0.8133, "step": 262470 }, { "epoch": 22.73, "learning_rate": 1.8192725800855785e-05, "loss": 0.6551, "step": 262480 }, { "epoch": 22.73, "learning_rate": 1.8190557418757948e-05, "loss": 0.6635, "step": 262490 }, { "epoch": 22.73, "learning_rate": 1.818838903666011e-05, "loss": 0.6356, "step": 262500 }, { "epoch": 22.74, "learning_rate": 1.8186220654562274e-05, "loss": 0.7627, "step": 262510 }, { "epoch": 22.74, "learning_rate": 1.8184052272464437e-05, "loss": 0.506, "step": 262520 }, { "epoch": 22.74, "learning_rate": 1.81818838903666e-05, "loss": 0.5926, "step": 262530 }, { "epoch": 22.74, "learning_rate": 1.817971550826876e-05, "loss": 0.7653, "step": 262540 }, { "epoch": 22.74, "learning_rate": 1.8177547126170926e-05, "loss": 0.6787, "step": 262550 }, { "epoch": 22.74, "learning_rate": 1.8175378744073086e-05, "loss": 0.8516, "step": 262560 }, { "epoch": 22.74, "learning_rate": 1.817321036197525e-05, "loss": 0.69, "step": 262570 }, { "epoch": 22.74, "learning_rate": 1.8171041979877412e-05, "loss": 0.6765, "step": 262580 }, { "epoch": 22.74, "learning_rate": 1.8168873597779575e-05, "loss": 0.5975, "step": 262590 }, { "epoch": 22.74, "learning_rate": 1.8166705215681738e-05, "loss": 0.6453, "step": 262600 }, { "epoch": 22.74, "learning_rate": 1.81645368335839e-05, "loss": 0.4828, "step": 262610 }, { "epoch": 22.74, "learning_rate": 1.8162368451486064e-05, "loss": 0.69, "step": 262620 }, { "epoch": 22.75, "learning_rate": 1.8160200069388227e-05, "loss": 0.5685, "step": 262630 }, { "epoch": 22.75, "learning_rate": 1.8158031687290387e-05, "loss": 0.8361, "step": 262640 }, { "epoch": 22.75, "learning_rate": 1.8155863305192553e-05, "loss": 0.7211, "step": 262650 }, { "epoch": 22.75, "learning_rate": 1.8153694923094713e-05, "loss": 0.6619, "step": 262660 }, { "epoch": 22.75, "learning_rate": 1.8151526540996876e-05, "loss": 0.6459, "step": 262670 }, { "epoch": 22.75, "learning_rate": 1.814935815889904e-05, "loss": 0.6606, "step": 262680 }, { "epoch": 22.75, "learning_rate": 1.81471897768012e-05, "loss": 0.5916, "step": 262690 }, { "epoch": 22.75, "learning_rate": 1.8145021394703365e-05, "loss": 0.8292, "step": 262700 }, { "epoch": 22.75, "learning_rate": 1.8142853012605528e-05, "loss": 0.632, "step": 262710 }, { "epoch": 22.75, "learning_rate": 1.814068463050769e-05, "loss": 0.7789, "step": 262720 }, { "epoch": 22.75, "learning_rate": 1.8138516248409854e-05, "loss": 0.7013, "step": 262730 }, { "epoch": 22.75, "learning_rate": 1.8136347866312013e-05, "loss": 0.5947, "step": 262740 }, { "epoch": 22.76, "learning_rate": 1.8134179484214176e-05, "loss": 0.7112, "step": 262750 }, { "epoch": 22.76, "learning_rate": 1.813201110211634e-05, "loss": 0.6513, "step": 262760 }, { "epoch": 22.76, "learning_rate": 1.8129842720018502e-05, "loss": 0.7828, "step": 262770 }, { "epoch": 22.76, "learning_rate": 1.8127674337920665e-05, "loss": 0.6369, "step": 262780 }, { "epoch": 22.76, "learning_rate": 1.8125505955822825e-05, "loss": 0.6258, "step": 262790 }, { "epoch": 22.76, "learning_rate": 1.812333757372499e-05, "loss": 0.75, "step": 262800 }, { "epoch": 22.76, "learning_rate": 1.812116919162715e-05, "loss": 0.758, "step": 262810 }, { "epoch": 22.76, "learning_rate": 1.8119000809529314e-05, "loss": 0.695, "step": 262820 }, { "epoch": 22.76, "learning_rate": 1.8116832427431477e-05, "loss": 0.4854, "step": 262830 }, { "epoch": 22.76, "learning_rate": 1.811466404533364e-05, "loss": 0.5745, "step": 262840 }, { "epoch": 22.76, "learning_rate": 1.8112495663235803e-05, "loss": 0.5992, "step": 262850 }, { "epoch": 22.77, "learning_rate": 1.8110327281137966e-05, "loss": 0.7756, "step": 262860 }, { "epoch": 22.77, "learning_rate": 1.810815889904013e-05, "loss": 0.5904, "step": 262870 }, { "epoch": 22.77, "learning_rate": 1.8105990516942292e-05, "loss": 0.6661, "step": 262880 }, { "epoch": 22.77, "learning_rate": 1.810382213484445e-05, "loss": 0.7468, "step": 262890 }, { "epoch": 22.77, "learning_rate": 1.8101653752746618e-05, "loss": 0.7736, "step": 262900 }, { "epoch": 22.77, "learning_rate": 1.8099485370648777e-05, "loss": 0.6857, "step": 262910 }, { "epoch": 22.77, "learning_rate": 1.809731698855094e-05, "loss": 0.6081, "step": 262920 }, { "epoch": 22.77, "learning_rate": 1.8095148606453103e-05, "loss": 0.7233, "step": 262930 }, { "epoch": 22.77, "learning_rate": 1.8092980224355266e-05, "loss": 0.6636, "step": 262940 }, { "epoch": 22.77, "learning_rate": 1.809081184225743e-05, "loss": 0.6327, "step": 262950 }, { "epoch": 22.77, "learning_rate": 1.8088643460159592e-05, "loss": 0.6529, "step": 262960 }, { "epoch": 22.77, "learning_rate": 1.8086475078061755e-05, "loss": 0.5783, "step": 262970 }, { "epoch": 22.78, "learning_rate": 1.808430669596392e-05, "loss": 0.6778, "step": 262980 }, { "epoch": 22.78, "learning_rate": 1.8082138313866078e-05, "loss": 0.6746, "step": 262990 }, { "epoch": 22.78, "learning_rate": 1.807996993176824e-05, "loss": 0.8526, "step": 263000 }, { "epoch": 22.78, "learning_rate": 1.8077801549670404e-05, "loss": 0.8026, "step": 263010 }, { "epoch": 22.78, "learning_rate": 1.8075633167572567e-05, "loss": 0.6398, "step": 263020 }, { "epoch": 22.78, "learning_rate": 1.807346478547473e-05, "loss": 0.7584, "step": 263030 }, { "epoch": 22.78, "learning_rate": 1.807129640337689e-05, "loss": 0.6164, "step": 263040 }, { "epoch": 22.78, "learning_rate": 1.8069128021279056e-05, "loss": 0.5768, "step": 263050 }, { "epoch": 22.78, "learning_rate": 1.8066959639181216e-05, "loss": 0.6128, "step": 263060 }, { "epoch": 22.78, "learning_rate": 1.806479125708338e-05, "loss": 0.6494, "step": 263070 }, { "epoch": 22.78, "learning_rate": 1.8062622874985542e-05, "loss": 0.8015, "step": 263080 }, { "epoch": 22.79, "learning_rate": 1.8060454492887705e-05, "loss": 0.6657, "step": 263090 }, { "epoch": 22.79, "learning_rate": 1.8058286110789868e-05, "loss": 0.7094, "step": 263100 }, { "epoch": 22.79, "learning_rate": 1.805611772869203e-05, "loss": 0.4996, "step": 263110 }, { "epoch": 22.79, "learning_rate": 1.8053949346594194e-05, "loss": 0.5754, "step": 263120 }, { "epoch": 22.79, "learning_rate": 1.8051780964496357e-05, "loss": 0.755, "step": 263130 }, { "epoch": 22.79, "learning_rate": 1.8049612582398516e-05, "loss": 0.6609, "step": 263140 }, { "epoch": 22.79, "learning_rate": 1.8047444200300683e-05, "loss": 0.7882, "step": 263150 }, { "epoch": 22.79, "learning_rate": 1.8045275818202842e-05, "loss": 0.7321, "step": 263160 }, { "epoch": 22.79, "learning_rate": 1.8043107436105005e-05, "loss": 0.6668, "step": 263170 }, { "epoch": 22.79, "learning_rate": 1.804093905400717e-05, "loss": 0.5962, "step": 263180 }, { "epoch": 22.79, "learning_rate": 1.803877067190933e-05, "loss": 0.5682, "step": 263190 }, { "epoch": 22.79, "learning_rate": 1.8036602289811494e-05, "loss": 0.6355, "step": 263200 }, { "epoch": 22.8, "learning_rate": 1.8034433907713657e-05, "loss": 0.5335, "step": 263210 }, { "epoch": 22.8, "learning_rate": 1.803226552561582e-05, "loss": 0.61, "step": 263220 }, { "epoch": 22.8, "learning_rate": 1.8030097143517983e-05, "loss": 0.7611, "step": 263230 }, { "epoch": 22.8, "learning_rate": 1.8027928761420143e-05, "loss": 0.5355, "step": 263240 }, { "epoch": 22.8, "learning_rate": 1.8025760379322306e-05, "loss": 0.8208, "step": 263250 }, { "epoch": 22.8, "learning_rate": 1.802359199722447e-05, "loss": 0.7197, "step": 263260 }, { "epoch": 22.8, "learning_rate": 1.8021423615126632e-05, "loss": 0.7021, "step": 263270 }, { "epoch": 22.8, "learning_rate": 1.8019255233028795e-05, "loss": 0.7312, "step": 263280 }, { "epoch": 22.8, "learning_rate": 1.8017086850930955e-05, "loss": 0.6205, "step": 263290 }, { "epoch": 22.8, "learning_rate": 1.801491846883312e-05, "loss": 0.7103, "step": 263300 }, { "epoch": 22.8, "learning_rate": 1.801275008673528e-05, "loss": 0.5238, "step": 263310 }, { "epoch": 22.81, "learning_rate": 1.8010581704637444e-05, "loss": 0.6332, "step": 263320 }, { "epoch": 22.81, "learning_rate": 1.8008413322539607e-05, "loss": 0.6164, "step": 263330 }, { "epoch": 22.81, "learning_rate": 1.800624494044177e-05, "loss": 0.8475, "step": 263340 }, { "epoch": 22.81, "learning_rate": 1.8004076558343933e-05, "loss": 0.5702, "step": 263350 }, { "epoch": 22.81, "learning_rate": 1.8001908176246096e-05, "loss": 0.6615, "step": 263360 }, { "epoch": 22.81, "learning_rate": 1.799973979414826e-05, "loss": 0.5522, "step": 263370 }, { "epoch": 22.81, "learning_rate": 1.799757141205042e-05, "loss": 0.7329, "step": 263380 }, { "epoch": 22.81, "learning_rate": 1.799540302995258e-05, "loss": 0.9102, "step": 263390 }, { "epoch": 22.81, "learning_rate": 1.7993234647854748e-05, "loss": 0.6824, "step": 263400 }, { "epoch": 22.81, "learning_rate": 1.7991066265756907e-05, "loss": 0.6456, "step": 263410 }, { "epoch": 22.81, "learning_rate": 1.798889788365907e-05, "loss": 0.5585, "step": 263420 }, { "epoch": 22.81, "learning_rate": 1.7986729501561233e-05, "loss": 0.7457, "step": 263430 }, { "epoch": 22.82, "learning_rate": 1.7984561119463396e-05, "loss": 0.6696, "step": 263440 }, { "epoch": 22.82, "learning_rate": 1.798239273736556e-05, "loss": 0.7081, "step": 263450 }, { "epoch": 22.82, "learning_rate": 1.7980224355267722e-05, "loss": 0.5743, "step": 263460 }, { "epoch": 22.82, "learning_rate": 1.7978055973169885e-05, "loss": 0.5845, "step": 263470 }, { "epoch": 22.82, "learning_rate": 1.797588759107205e-05, "loss": 0.7024, "step": 263480 }, { "epoch": 22.82, "learning_rate": 1.7973719208974208e-05, "loss": 0.7075, "step": 263490 }, { "epoch": 22.82, "learning_rate": 1.7971550826876374e-05, "loss": 0.6348, "step": 263500 }, { "epoch": 22.82, "learning_rate": 1.7969382444778534e-05, "loss": 0.7115, "step": 263510 }, { "epoch": 22.82, "learning_rate": 1.7967214062680697e-05, "loss": 0.6313, "step": 263520 }, { "epoch": 22.82, "learning_rate": 1.796504568058286e-05, "loss": 0.5959, "step": 263530 }, { "epoch": 22.82, "learning_rate": 1.796287729848502e-05, "loss": 0.7856, "step": 263540 }, { "epoch": 22.83, "learning_rate": 1.7960708916387186e-05, "loss": 0.7382, "step": 263550 }, { "epoch": 22.83, "learning_rate": 1.7958540534289346e-05, "loss": 0.746, "step": 263560 }, { "epoch": 22.83, "learning_rate": 1.795637215219151e-05, "loss": 0.7655, "step": 263570 }, { "epoch": 22.83, "learning_rate": 1.795420377009367e-05, "loss": 0.7479, "step": 263580 }, { "epoch": 22.83, "learning_rate": 1.7952035387995835e-05, "loss": 0.7121, "step": 263590 }, { "epoch": 22.83, "learning_rate": 1.7949867005897998e-05, "loss": 0.5804, "step": 263600 }, { "epoch": 22.83, "learning_rate": 1.794769862380016e-05, "loss": 0.5752, "step": 263610 }, { "epoch": 22.83, "learning_rate": 1.7945530241702324e-05, "loss": 0.5644, "step": 263620 }, { "epoch": 22.83, "learning_rate": 1.7943361859604487e-05, "loss": 0.6942, "step": 263630 }, { "epoch": 22.83, "learning_rate": 1.7941193477506646e-05, "loss": 0.7746, "step": 263640 }, { "epoch": 22.83, "learning_rate": 1.7939025095408813e-05, "loss": 0.6046, "step": 263650 }, { "epoch": 22.83, "learning_rate": 1.7936856713310972e-05, "loss": 0.7328, "step": 263660 }, { "epoch": 22.84, "learning_rate": 1.7934688331213135e-05, "loss": 0.7015, "step": 263670 }, { "epoch": 22.84, "learning_rate": 1.7932519949115298e-05, "loss": 0.8002, "step": 263680 }, { "epoch": 22.84, "learning_rate": 1.793035156701746e-05, "loss": 0.6688, "step": 263690 }, { "epoch": 22.84, "learning_rate": 1.7928183184919624e-05, "loss": 0.5627, "step": 263700 }, { "epoch": 22.84, "learning_rate": 1.7926014802821787e-05, "loss": 0.7991, "step": 263710 }, { "epoch": 22.84, "learning_rate": 1.792384642072395e-05, "loss": 0.797, "step": 263720 }, { "epoch": 22.84, "learning_rate": 1.7921678038626113e-05, "loss": 0.741, "step": 263730 }, { "epoch": 22.84, "learning_rate": 1.7919509656528273e-05, "loss": 0.6528, "step": 263740 }, { "epoch": 22.84, "learning_rate": 1.791734127443044e-05, "loss": 0.763, "step": 263750 }, { "epoch": 22.84, "learning_rate": 1.79151728923326e-05, "loss": 0.6963, "step": 263760 }, { "epoch": 22.84, "learning_rate": 1.7913004510234762e-05, "loss": 0.6956, "step": 263770 }, { "epoch": 22.85, "learning_rate": 1.7910836128136925e-05, "loss": 0.6527, "step": 263780 }, { "epoch": 22.85, "learning_rate": 1.7908667746039088e-05, "loss": 0.6731, "step": 263790 }, { "epoch": 22.85, "learning_rate": 1.790649936394125e-05, "loss": 0.7672, "step": 263800 }, { "epoch": 22.85, "learning_rate": 1.790433098184341e-05, "loss": 0.5862, "step": 263810 }, { "epoch": 22.85, "learning_rate": 1.7902162599745573e-05, "loss": 0.6193, "step": 263820 }, { "epoch": 22.85, "learning_rate": 1.7899994217647736e-05, "loss": 0.8326, "step": 263830 }, { "epoch": 22.85, "learning_rate": 1.78978258355499e-05, "loss": 0.6634, "step": 263840 }, { "epoch": 22.85, "learning_rate": 1.7895657453452063e-05, "loss": 0.7229, "step": 263850 }, { "epoch": 22.85, "learning_rate": 1.7893489071354226e-05, "loss": 0.6148, "step": 263860 }, { "epoch": 22.85, "learning_rate": 1.789132068925639e-05, "loss": 0.7356, "step": 263870 }, { "epoch": 22.85, "learning_rate": 1.788915230715855e-05, "loss": 0.6782, "step": 263880 }, { "epoch": 22.85, "learning_rate": 1.788698392506071e-05, "loss": 0.7164, "step": 263890 }, { "epoch": 22.86, "learning_rate": 1.7884815542962878e-05, "loss": 0.6667, "step": 263900 }, { "epoch": 22.86, "learning_rate": 1.7882647160865037e-05, "loss": 0.5614, "step": 263910 }, { "epoch": 22.86, "learning_rate": 1.78804787787672e-05, "loss": 0.7356, "step": 263920 }, { "epoch": 22.86, "learning_rate": 1.7878310396669363e-05, "loss": 0.6192, "step": 263930 }, { "epoch": 22.86, "learning_rate": 1.7876142014571526e-05, "loss": 0.7015, "step": 263940 }, { "epoch": 22.86, "learning_rate": 1.787397363247369e-05, "loss": 0.6932, "step": 263950 }, { "epoch": 22.86, "learning_rate": 1.7871805250375852e-05, "loss": 0.8223, "step": 263960 }, { "epoch": 22.86, "learning_rate": 1.7869636868278015e-05, "loss": 0.8344, "step": 263970 }, { "epoch": 22.86, "learning_rate": 1.7867468486180178e-05, "loss": 0.6274, "step": 263980 }, { "epoch": 22.86, "learning_rate": 1.7865300104082338e-05, "loss": 0.8786, "step": 263990 }, { "epoch": 22.86, "learning_rate": 1.7863131721984504e-05, "loss": 0.6683, "step": 264000 }, { "epoch": 22.86, "learning_rate": 1.7860963339886664e-05, "loss": 0.5584, "step": 264010 }, { "epoch": 22.87, "learning_rate": 1.7858794957788827e-05, "loss": 0.6528, "step": 264020 }, { "epoch": 22.87, "learning_rate": 1.785662657569099e-05, "loss": 0.783, "step": 264030 }, { "epoch": 22.87, "learning_rate": 1.7854458193593153e-05, "loss": 0.6585, "step": 264040 }, { "epoch": 22.87, "learning_rate": 1.7852289811495316e-05, "loss": 0.5958, "step": 264050 }, { "epoch": 22.87, "learning_rate": 1.7850121429397475e-05, "loss": 0.8566, "step": 264060 }, { "epoch": 22.87, "learning_rate": 1.7847953047299642e-05, "loss": 0.8163, "step": 264070 }, { "epoch": 22.87, "learning_rate": 1.78457846652018e-05, "loss": 0.8085, "step": 264080 }, { "epoch": 22.87, "learning_rate": 1.7843616283103964e-05, "loss": 0.7334, "step": 264090 }, { "epoch": 22.87, "learning_rate": 1.7841447901006127e-05, "loss": 0.6842, "step": 264100 }, { "epoch": 22.87, "learning_rate": 1.783927951890829e-05, "loss": 0.569, "step": 264110 }, { "epoch": 22.87, "learning_rate": 1.7837111136810453e-05, "loss": 0.7719, "step": 264120 }, { "epoch": 22.88, "learning_rate": 1.7834942754712616e-05, "loss": 0.6537, "step": 264130 }, { "epoch": 22.88, "learning_rate": 1.7832774372614776e-05, "loss": 0.576, "step": 264140 }, { "epoch": 22.88, "learning_rate": 1.7830605990516942e-05, "loss": 0.6692, "step": 264150 }, { "epoch": 22.88, "learning_rate": 1.7828437608419102e-05, "loss": 0.716, "step": 264160 }, { "epoch": 22.88, "learning_rate": 1.7826269226321265e-05, "loss": 0.7274, "step": 264170 }, { "epoch": 22.88, "learning_rate": 1.7824100844223428e-05, "loss": 0.7737, "step": 264180 }, { "epoch": 22.88, "learning_rate": 1.782193246212559e-05, "loss": 0.6638, "step": 264190 }, { "epoch": 22.88, "learning_rate": 1.7819764080027754e-05, "loss": 0.6121, "step": 264200 }, { "epoch": 22.88, "learning_rate": 1.7817595697929917e-05, "loss": 0.718, "step": 264210 }, { "epoch": 22.88, "learning_rate": 1.781542731583208e-05, "loss": 0.732, "step": 264220 }, { "epoch": 22.88, "learning_rate": 1.7813258933734243e-05, "loss": 0.6604, "step": 264230 }, { "epoch": 22.88, "learning_rate": 1.7811090551636403e-05, "loss": 0.7331, "step": 264240 }, { "epoch": 22.89, "learning_rate": 1.780892216953857e-05, "loss": 0.6605, "step": 264250 }, { "epoch": 22.89, "learning_rate": 1.780675378744073e-05, "loss": 0.6342, "step": 264260 }, { "epoch": 22.89, "learning_rate": 1.7804585405342892e-05, "loss": 0.6882, "step": 264270 }, { "epoch": 22.89, "learning_rate": 1.7802417023245055e-05, "loss": 0.7462, "step": 264280 }, { "epoch": 22.89, "learning_rate": 1.7800248641147218e-05, "loss": 0.7543, "step": 264290 }, { "epoch": 22.89, "learning_rate": 1.779808025904938e-05, "loss": 0.6308, "step": 264300 }, { "epoch": 22.89, "learning_rate": 1.779591187695154e-05, "loss": 0.82, "step": 264310 }, { "epoch": 22.89, "learning_rate": 1.7793743494853707e-05, "loss": 0.7336, "step": 264320 }, { "epoch": 22.89, "learning_rate": 1.7791575112755866e-05, "loss": 0.7499, "step": 264330 }, { "epoch": 22.89, "learning_rate": 1.778940673065803e-05, "loss": 0.7837, "step": 264340 }, { "epoch": 22.89, "learning_rate": 1.7787238348560192e-05, "loss": 0.5798, "step": 264350 }, { "epoch": 22.9, "learning_rate": 1.7785069966462355e-05, "loss": 0.6832, "step": 264360 }, { "epoch": 22.9, "learning_rate": 1.778290158436452e-05, "loss": 0.7338, "step": 264370 }, { "epoch": 22.9, "learning_rate": 1.778073320226668e-05, "loss": 0.6063, "step": 264380 }, { "epoch": 22.9, "learning_rate": 1.777856482016884e-05, "loss": 0.648, "step": 264390 }, { "epoch": 22.9, "learning_rate": 1.7776396438071007e-05, "loss": 0.6122, "step": 264400 }, { "epoch": 22.9, "learning_rate": 1.7774228055973167e-05, "loss": 0.7028, "step": 264410 }, { "epoch": 22.9, "learning_rate": 1.777205967387533e-05, "loss": 0.6679, "step": 264420 }, { "epoch": 22.9, "learning_rate": 1.7769891291777493e-05, "loss": 0.7303, "step": 264430 }, { "epoch": 22.9, "learning_rate": 1.7767722909679656e-05, "loss": 0.5965, "step": 264440 }, { "epoch": 22.9, "learning_rate": 1.776555452758182e-05, "loss": 0.7565, "step": 264450 }, { "epoch": 22.9, "learning_rate": 1.7763386145483982e-05, "loss": 0.672, "step": 264460 }, { "epoch": 22.9, "learning_rate": 1.7761217763386145e-05, "loss": 0.821, "step": 264470 }, { "epoch": 22.91, "learning_rate": 1.7759049381288308e-05, "loss": 0.6974, "step": 264480 }, { "epoch": 22.91, "learning_rate": 1.7756880999190468e-05, "loss": 0.7142, "step": 264490 }, { "epoch": 22.91, "learning_rate": 1.7754712617092634e-05, "loss": 0.7512, "step": 264500 }, { "epoch": 22.91, "learning_rate": 1.7752544234994794e-05, "loss": 0.821, "step": 264510 }, { "epoch": 22.91, "learning_rate": 1.7750375852896957e-05, "loss": 0.7192, "step": 264520 }, { "epoch": 22.91, "learning_rate": 1.774820747079912e-05, "loss": 0.7678, "step": 264530 }, { "epoch": 22.91, "learning_rate": 1.7746039088701283e-05, "loss": 0.693, "step": 264540 }, { "epoch": 22.91, "learning_rate": 1.7743870706603446e-05, "loss": 0.6401, "step": 264550 }, { "epoch": 22.91, "learning_rate": 1.7741702324505605e-05, "loss": 0.703, "step": 264560 }, { "epoch": 22.91, "learning_rate": 1.773953394240777e-05, "loss": 0.6905, "step": 264570 }, { "epoch": 22.91, "learning_rate": 1.773736556030993e-05, "loss": 0.6448, "step": 264580 }, { "epoch": 22.92, "learning_rate": 1.7735197178212094e-05, "loss": 0.7626, "step": 264590 }, { "epoch": 22.92, "learning_rate": 1.7733028796114257e-05, "loss": 0.6501, "step": 264600 }, { "epoch": 22.92, "learning_rate": 1.773086041401642e-05, "loss": 0.529, "step": 264610 }, { "epoch": 22.92, "learning_rate": 1.7728692031918583e-05, "loss": 0.6712, "step": 264620 }, { "epoch": 22.92, "learning_rate": 1.7726523649820746e-05, "loss": 0.6012, "step": 264630 }, { "epoch": 22.92, "learning_rate": 1.7724355267722906e-05, "loss": 0.6201, "step": 264640 }, { "epoch": 22.92, "learning_rate": 1.7722186885625072e-05, "loss": 0.7376, "step": 264650 }, { "epoch": 22.92, "learning_rate": 1.7720018503527232e-05, "loss": 0.7383, "step": 264660 }, { "epoch": 22.92, "learning_rate": 1.7717850121429395e-05, "loss": 0.6487, "step": 264670 }, { "epoch": 22.92, "learning_rate": 1.7715681739331558e-05, "loss": 0.8312, "step": 264680 }, { "epoch": 22.92, "learning_rate": 1.771351335723372e-05, "loss": 0.7928, "step": 264690 }, { "epoch": 22.92, "learning_rate": 1.7711344975135884e-05, "loss": 0.6433, "step": 264700 }, { "epoch": 22.93, "learning_rate": 1.7709176593038047e-05, "loss": 0.5747, "step": 264710 }, { "epoch": 22.93, "learning_rate": 1.770700821094021e-05, "loss": 0.8255, "step": 264720 }, { "epoch": 22.93, "learning_rate": 1.7704839828842373e-05, "loss": 0.7875, "step": 264730 }, { "epoch": 22.93, "learning_rate": 1.7702671446744533e-05, "loss": 0.736, "step": 264740 }, { "epoch": 22.93, "learning_rate": 1.77005030646467e-05, "loss": 0.7008, "step": 264750 }, { "epoch": 22.93, "learning_rate": 1.769833468254886e-05, "loss": 0.74, "step": 264760 }, { "epoch": 22.93, "learning_rate": 1.769616630045102e-05, "loss": 0.6587, "step": 264770 }, { "epoch": 22.93, "learning_rate": 1.7693997918353185e-05, "loss": 0.6591, "step": 264780 }, { "epoch": 22.93, "learning_rate": 1.7691829536255348e-05, "loss": 0.6077, "step": 264790 }, { "epoch": 22.93, "learning_rate": 1.768966115415751e-05, "loss": 0.7776, "step": 264800 }, { "epoch": 22.93, "learning_rate": 1.768749277205967e-05, "loss": 0.6863, "step": 264810 }, { "epoch": 22.94, "learning_rate": 1.7685324389961837e-05, "loss": 0.5851, "step": 264820 }, { "epoch": 22.94, "learning_rate": 1.7683156007863996e-05, "loss": 0.5731, "step": 264830 }, { "epoch": 22.94, "learning_rate": 1.768098762576616e-05, "loss": 0.7627, "step": 264840 }, { "epoch": 22.94, "learning_rate": 1.7678819243668322e-05, "loss": 0.6474, "step": 264850 }, { "epoch": 22.94, "learning_rate": 1.7676650861570485e-05, "loss": 0.6459, "step": 264860 }, { "epoch": 22.94, "learning_rate": 1.7674482479472648e-05, "loss": 0.7546, "step": 264870 }, { "epoch": 22.94, "learning_rate": 1.767231409737481e-05, "loss": 0.6091, "step": 264880 }, { "epoch": 22.94, "learning_rate": 1.767014571527697e-05, "loss": 0.7648, "step": 264890 }, { "epoch": 22.94, "learning_rate": 1.7667977333179137e-05, "loss": 0.6707, "step": 264900 }, { "epoch": 22.94, "learning_rate": 1.7665808951081297e-05, "loss": 0.7743, "step": 264910 }, { "epoch": 22.94, "learning_rate": 1.766364056898346e-05, "loss": 0.6069, "step": 264920 }, { "epoch": 22.94, "learning_rate": 1.7661472186885623e-05, "loss": 0.733, "step": 264930 }, { "epoch": 22.95, "learning_rate": 1.7659303804787786e-05, "loss": 0.7246, "step": 264940 }, { "epoch": 22.95, "learning_rate": 1.765713542268995e-05, "loss": 0.6997, "step": 264950 }, { "epoch": 22.95, "learning_rate": 1.7654967040592112e-05, "loss": 0.5905, "step": 264960 }, { "epoch": 22.95, "learning_rate": 1.7652798658494275e-05, "loss": 0.7354, "step": 264970 }, { "epoch": 22.95, "learning_rate": 1.7650630276396438e-05, "loss": 0.5145, "step": 264980 }, { "epoch": 22.95, "learning_rate": 1.7648461894298597e-05, "loss": 0.4978, "step": 264990 }, { "epoch": 22.95, "learning_rate": 1.7646293512200764e-05, "loss": 0.6339, "step": 265000 }, { "epoch": 22.95, "learning_rate": 1.7644125130102923e-05, "loss": 0.6583, "step": 265010 }, { "epoch": 22.95, "learning_rate": 1.7641956748005086e-05, "loss": 0.6399, "step": 265020 }, { "epoch": 22.95, "learning_rate": 1.763978836590725e-05, "loss": 0.7261, "step": 265030 }, { "epoch": 22.95, "learning_rate": 1.7637619983809412e-05, "loss": 0.6278, "step": 265040 }, { "epoch": 22.96, "learning_rate": 1.7635451601711575e-05, "loss": 0.8136, "step": 265050 }, { "epoch": 22.96, "learning_rate": 1.7633283219613735e-05, "loss": 0.5135, "step": 265060 }, { "epoch": 22.96, "learning_rate": 1.76311148375159e-05, "loss": 0.7499, "step": 265070 }, { "epoch": 22.96, "learning_rate": 1.762894645541806e-05, "loss": 0.7089, "step": 265080 }, { "epoch": 22.96, "learning_rate": 1.7626778073320224e-05, "loss": 0.6485, "step": 265090 }, { "epoch": 22.96, "learning_rate": 1.7624609691222387e-05, "loss": 0.7064, "step": 265100 }, { "epoch": 22.96, "learning_rate": 1.762244130912455e-05, "loss": 0.5653, "step": 265110 }, { "epoch": 22.96, "learning_rate": 1.7620272927026713e-05, "loss": 0.6223, "step": 265120 }, { "epoch": 22.96, "learning_rate": 1.7618104544928876e-05, "loss": 0.7907, "step": 265130 }, { "epoch": 22.96, "learning_rate": 1.761593616283104e-05, "loss": 0.7351, "step": 265140 }, { "epoch": 22.96, "learning_rate": 1.7613767780733202e-05, "loss": 0.7627, "step": 265150 }, { "epoch": 22.96, "learning_rate": 1.7611599398635362e-05, "loss": 0.6321, "step": 265160 }, { "epoch": 22.97, "learning_rate": 1.7609431016537525e-05, "loss": 0.6252, "step": 265170 }, { "epoch": 22.97, "learning_rate": 1.7607262634439688e-05, "loss": 0.7201, "step": 265180 }, { "epoch": 22.97, "learning_rate": 1.760509425234185e-05, "loss": 0.7273, "step": 265190 }, { "epoch": 22.97, "learning_rate": 1.7602925870244014e-05, "loss": 0.8102, "step": 265200 }, { "epoch": 22.97, "learning_rate": 1.7600757488146177e-05, "loss": 0.7173, "step": 265210 }, { "epoch": 22.97, "learning_rate": 1.759858910604834e-05, "loss": 0.7917, "step": 265220 }, { "epoch": 22.97, "learning_rate": 1.7596420723950503e-05, "loss": 0.7332, "step": 265230 }, { "epoch": 22.97, "learning_rate": 1.7594252341852662e-05, "loss": 0.801, "step": 265240 }, { "epoch": 22.97, "learning_rate": 1.759208395975483e-05, "loss": 0.815, "step": 265250 }, { "epoch": 22.97, "learning_rate": 1.758991557765699e-05, "loss": 0.7587, "step": 265260 }, { "epoch": 22.97, "learning_rate": 1.758774719555915e-05, "loss": 0.5667, "step": 265270 }, { "epoch": 22.97, "learning_rate": 1.7585578813461314e-05, "loss": 0.8674, "step": 265280 }, { "epoch": 22.98, "learning_rate": 1.7583410431363477e-05, "loss": 0.6683, "step": 265290 }, { "epoch": 22.98, "learning_rate": 1.758124204926564e-05, "loss": 0.6234, "step": 265300 }, { "epoch": 22.98, "learning_rate": 1.75790736671678e-05, "loss": 0.6952, "step": 265310 }, { "epoch": 22.98, "learning_rate": 1.7576905285069966e-05, "loss": 0.6506, "step": 265320 }, { "epoch": 22.98, "learning_rate": 1.7574736902972126e-05, "loss": 0.6625, "step": 265330 }, { "epoch": 22.98, "learning_rate": 1.757256852087429e-05, "loss": 0.667, "step": 265340 }, { "epoch": 22.98, "learning_rate": 1.7570400138776452e-05, "loss": 0.5978, "step": 265350 }, { "epoch": 22.98, "learning_rate": 1.7568231756678615e-05, "loss": 0.5873, "step": 265360 }, { "epoch": 22.98, "learning_rate": 1.7566063374580778e-05, "loss": 0.6672, "step": 265370 }, { "epoch": 22.98, "learning_rate": 1.756389499248294e-05, "loss": 0.6536, "step": 265380 }, { "epoch": 22.98, "learning_rate": 1.7561726610385104e-05, "loss": 0.7237, "step": 265390 }, { "epoch": 22.99, "learning_rate": 1.7559558228287267e-05, "loss": 0.5439, "step": 265400 }, { "epoch": 22.99, "learning_rate": 1.7557389846189427e-05, "loss": 0.7904, "step": 265410 }, { "epoch": 22.99, "learning_rate": 1.7555221464091593e-05, "loss": 0.6766, "step": 265420 }, { "epoch": 22.99, "learning_rate": 1.7553053081993753e-05, "loss": 0.5748, "step": 265430 }, { "epoch": 22.99, "learning_rate": 1.7550884699895916e-05, "loss": 0.744, "step": 265440 }, { "epoch": 22.99, "learning_rate": 1.754871631779808e-05, "loss": 0.7253, "step": 265450 }, { "epoch": 22.99, "learning_rate": 1.754654793570024e-05, "loss": 0.6831, "step": 265460 }, { "epoch": 22.99, "learning_rate": 1.7544379553602405e-05, "loss": 0.7318, "step": 265470 }, { "epoch": 22.99, "learning_rate": 1.7542211171504568e-05, "loss": 0.7019, "step": 265480 }, { "epoch": 22.99, "learning_rate": 1.7540042789406727e-05, "loss": 0.5492, "step": 265490 }, { "epoch": 22.99, "learning_rate": 1.7537874407308894e-05, "loss": 0.5424, "step": 265500 }, { "epoch": 22.99, "learning_rate": 1.7535706025211053e-05, "loss": 0.7829, "step": 265510 }, { "epoch": 23.0, "learning_rate": 1.7533537643113216e-05, "loss": 0.6593, "step": 265520 }, { "epoch": 23.0, "learning_rate": 1.753136926101538e-05, "loss": 0.811, "step": 265530 }, { "epoch": 23.0, "learning_rate": 1.7529200878917542e-05, "loss": 0.5764, "step": 265540 }, { "epoch": 23.0, "learning_rate": 1.7527032496819705e-05, "loss": 0.8159, "step": 265550 }, { "epoch": 23.0, "learning_rate": 1.7524864114721865e-05, "loss": 0.6839, "step": 265560 }, { "epoch": 23.0, "eval_cer": 0.5334868220665436, "eval_em": 0.05827746260959257, "eval_f1": 0.05827746260959257, "eval_loss": 1.5337226390838623, "eval_runtime": 257.0169, "eval_samples_per_second": 7.544, "eval_steps_per_second": 0.945, "eval_wer": 0.9417225373904075, "step": 265569 }, { "epoch": 23.0, "learning_rate": 1.752269573262403e-05, "loss": 0.6217, "step": 265570 }, { "epoch": 23.0, "learning_rate": 1.752052735052619e-05, "loss": 0.6429, "step": 265580 }, { "epoch": 23.0, "learning_rate": 1.7518358968428354e-05, "loss": 0.6158, "step": 265590 }, { "epoch": 23.0, "learning_rate": 1.7516190586330517e-05, "loss": 0.5574, "step": 265600 }, { "epoch": 23.0, "learning_rate": 1.751402220423268e-05, "loss": 0.6777, "step": 265610 }, { "epoch": 23.0, "learning_rate": 1.7511853822134843e-05, "loss": 0.774, "step": 265620 }, { "epoch": 23.01, "learning_rate": 1.7509685440037006e-05, "loss": 0.6897, "step": 265630 }, { "epoch": 23.01, "learning_rate": 1.750751705793917e-05, "loss": 0.5954, "step": 265640 }, { "epoch": 23.01, "learning_rate": 1.7505348675841332e-05, "loss": 0.4692, "step": 265650 }, { "epoch": 23.01, "learning_rate": 1.750318029374349e-05, "loss": 0.5302, "step": 265660 }, { "epoch": 23.01, "learning_rate": 1.7501011911645658e-05, "loss": 0.5768, "step": 265670 }, { "epoch": 23.01, "learning_rate": 1.7498843529547818e-05, "loss": 0.552, "step": 265680 }, { "epoch": 23.01, "learning_rate": 1.749667514744998e-05, "loss": 0.6763, "step": 265690 }, { "epoch": 23.01, "learning_rate": 1.7494506765352144e-05, "loss": 0.6125, "step": 265700 }, { "epoch": 23.01, "learning_rate": 1.7492338383254307e-05, "loss": 0.6773, "step": 265710 }, { "epoch": 23.01, "learning_rate": 1.749017000115647e-05, "loss": 0.5759, "step": 265720 }, { "epoch": 23.01, "learning_rate": 1.7488001619058633e-05, "loss": 0.5552, "step": 265730 }, { "epoch": 23.01, "learning_rate": 1.7485833236960792e-05, "loss": 0.6876, "step": 265740 }, { "epoch": 23.02, "learning_rate": 1.748366485486296e-05, "loss": 0.722, "step": 265750 }, { "epoch": 23.02, "learning_rate": 1.7481496472765118e-05, "loss": 0.7823, "step": 265760 }, { "epoch": 23.02, "learning_rate": 1.747932809066728e-05, "loss": 0.6471, "step": 265770 }, { "epoch": 23.02, "learning_rate": 1.7477159708569444e-05, "loss": 0.6591, "step": 265780 }, { "epoch": 23.02, "learning_rate": 1.7474991326471607e-05, "loss": 0.6051, "step": 265790 }, { "epoch": 23.02, "learning_rate": 1.747282294437377e-05, "loss": 0.6365, "step": 265800 }, { "epoch": 23.02, "learning_rate": 1.7470654562275933e-05, "loss": 0.681, "step": 265810 }, { "epoch": 23.02, "learning_rate": 1.7468486180178096e-05, "loss": 0.7845, "step": 265820 }, { "epoch": 23.02, "learning_rate": 1.7466317798080256e-05, "loss": 0.6032, "step": 265830 }, { "epoch": 23.02, "learning_rate": 1.746414941598242e-05, "loss": 0.5757, "step": 265840 }, { "epoch": 23.02, "learning_rate": 1.7461981033884582e-05, "loss": 0.6054, "step": 265850 }, { "epoch": 23.03, "learning_rate": 1.7459812651786745e-05, "loss": 0.6843, "step": 265860 }, { "epoch": 23.03, "learning_rate": 1.7457644269688908e-05, "loss": 0.6727, "step": 265870 }, { "epoch": 23.03, "learning_rate": 1.745547588759107e-05, "loss": 0.5907, "step": 265880 }, { "epoch": 23.03, "learning_rate": 1.7453307505493234e-05, "loss": 0.7838, "step": 265890 }, { "epoch": 23.03, "learning_rate": 1.7451139123395397e-05, "loss": 0.78, "step": 265900 }, { "epoch": 23.03, "learning_rate": 1.7448970741297557e-05, "loss": 0.7938, "step": 265910 }, { "epoch": 23.03, "learning_rate": 1.7446802359199723e-05, "loss": 0.6552, "step": 265920 }, { "epoch": 23.03, "learning_rate": 1.7444633977101883e-05, "loss": 0.5871, "step": 265930 }, { "epoch": 23.03, "learning_rate": 1.7442465595004046e-05, "loss": 0.6407, "step": 265940 }, { "epoch": 23.03, "learning_rate": 1.744029721290621e-05, "loss": 0.5929, "step": 265950 }, { "epoch": 23.03, "learning_rate": 1.743812883080837e-05, "loss": 0.6945, "step": 265960 }, { "epoch": 23.03, "learning_rate": 1.7435960448710535e-05, "loss": 0.5964, "step": 265970 }, { "epoch": 23.04, "learning_rate": 1.7433792066612698e-05, "loss": 0.6197, "step": 265980 }, { "epoch": 23.04, "learning_rate": 1.7431623684514857e-05, "loss": 0.6313, "step": 265990 }, { "epoch": 23.04, "learning_rate": 1.7429455302417024e-05, "loss": 0.6082, "step": 266000 }, { "epoch": 23.04, "learning_rate": 1.7427286920319183e-05, "loss": 0.6472, "step": 266010 }, { "epoch": 23.04, "learning_rate": 1.7425118538221346e-05, "loss": 0.8911, "step": 266020 }, { "epoch": 23.04, "learning_rate": 1.742295015612351e-05, "loss": 0.6017, "step": 266030 }, { "epoch": 23.04, "learning_rate": 1.7420781774025672e-05, "loss": 0.5284, "step": 266040 }, { "epoch": 23.04, "learning_rate": 1.7418613391927835e-05, "loss": 0.625, "step": 266050 }, { "epoch": 23.04, "learning_rate": 1.7416445009829998e-05, "loss": 0.5761, "step": 266060 }, { "epoch": 23.04, "learning_rate": 1.741427662773216e-05, "loss": 0.7237, "step": 266070 }, { "epoch": 23.04, "learning_rate": 1.741210824563432e-05, "loss": 0.5312, "step": 266080 }, { "epoch": 23.05, "learning_rate": 1.7409939863536484e-05, "loss": 0.6357, "step": 266090 }, { "epoch": 23.05, "learning_rate": 1.7407771481438647e-05, "loss": 0.6504, "step": 266100 }, { "epoch": 23.05, "learning_rate": 1.740560309934081e-05, "loss": 0.6402, "step": 266110 }, { "epoch": 23.05, "learning_rate": 1.7403434717242973e-05, "loss": 0.6504, "step": 266120 }, { "epoch": 23.05, "learning_rate": 1.7401266335145136e-05, "loss": 0.6899, "step": 266130 }, { "epoch": 23.05, "learning_rate": 1.73990979530473e-05, "loss": 0.6636, "step": 266140 }, { "epoch": 23.05, "learning_rate": 1.7396929570949462e-05, "loss": 0.6688, "step": 266150 }, { "epoch": 23.05, "learning_rate": 1.739476118885162e-05, "loss": 0.6865, "step": 266160 }, { "epoch": 23.05, "learning_rate": 1.7392592806753788e-05, "loss": 0.7531, "step": 266170 }, { "epoch": 23.05, "learning_rate": 1.7390424424655947e-05, "loss": 0.6346, "step": 266180 }, { "epoch": 23.05, "learning_rate": 1.738825604255811e-05, "loss": 0.7028, "step": 266190 }, { "epoch": 23.05, "learning_rate": 1.7386087660460273e-05, "loss": 0.684, "step": 266200 }, { "epoch": 23.06, "learning_rate": 1.7383919278362436e-05, "loss": 0.5899, "step": 266210 }, { "epoch": 23.06, "learning_rate": 1.73817508962646e-05, "loss": 0.5534, "step": 266220 }, { "epoch": 23.06, "learning_rate": 1.7379582514166762e-05, "loss": 0.6959, "step": 266230 }, { "epoch": 23.06, "learning_rate": 1.7377414132068922e-05, "loss": 0.6099, "step": 266240 }, { "epoch": 23.06, "learning_rate": 1.737524574997109e-05, "loss": 0.5779, "step": 266250 }, { "epoch": 23.06, "learning_rate": 1.7373077367873248e-05, "loss": 0.5619, "step": 266260 }, { "epoch": 23.06, "learning_rate": 1.737090898577541e-05, "loss": 0.6715, "step": 266270 }, { "epoch": 23.06, "learning_rate": 1.7368740603677574e-05, "loss": 0.5171, "step": 266280 }, { "epoch": 23.06, "learning_rate": 1.7366572221579737e-05, "loss": 0.588, "step": 266290 }, { "epoch": 23.06, "learning_rate": 1.73644038394819e-05, "loss": 0.514, "step": 266300 }, { "epoch": 23.06, "learning_rate": 1.7362235457384063e-05, "loss": 0.5759, "step": 266310 }, { "epoch": 23.06, "learning_rate": 1.7360067075286226e-05, "loss": 0.5615, "step": 266320 }, { "epoch": 23.07, "learning_rate": 1.7357898693188386e-05, "loss": 0.7103, "step": 266330 }, { "epoch": 23.07, "learning_rate": 1.735573031109055e-05, "loss": 0.6249, "step": 266340 }, { "epoch": 23.07, "learning_rate": 1.7353561928992712e-05, "loss": 0.6595, "step": 266350 }, { "epoch": 23.07, "learning_rate": 1.7351393546894875e-05, "loss": 0.6735, "step": 266360 }, { "epoch": 23.07, "learning_rate": 1.7349225164797038e-05, "loss": 0.5165, "step": 266370 }, { "epoch": 23.07, "learning_rate": 1.73470567826992e-05, "loss": 0.7314, "step": 266380 }, { "epoch": 23.07, "learning_rate": 1.7344888400601364e-05, "loss": 0.6137, "step": 266390 }, { "epoch": 23.07, "learning_rate": 1.7342720018503527e-05, "loss": 0.5836, "step": 266400 }, { "epoch": 23.07, "learning_rate": 1.7340551636405686e-05, "loss": 0.6018, "step": 266410 }, { "epoch": 23.07, "learning_rate": 1.7338383254307853e-05, "loss": 0.57, "step": 266420 }, { "epoch": 23.07, "learning_rate": 1.7336214872210012e-05, "loss": 0.7723, "step": 266430 }, { "epoch": 23.08, "learning_rate": 1.7334046490112175e-05, "loss": 0.5454, "step": 266440 }, { "epoch": 23.08, "learning_rate": 1.733187810801434e-05, "loss": 0.6331, "step": 266450 }, { "epoch": 23.08, "learning_rate": 1.73297097259165e-05, "loss": 0.5144, "step": 266460 }, { "epoch": 23.08, "learning_rate": 1.7327541343818664e-05, "loss": 0.5865, "step": 266470 }, { "epoch": 23.08, "learning_rate": 1.7325372961720827e-05, "loss": 0.6603, "step": 266480 }, { "epoch": 23.08, "learning_rate": 1.732320457962299e-05, "loss": 0.6558, "step": 266490 }, { "epoch": 23.08, "learning_rate": 1.7321036197525153e-05, "loss": 0.6546, "step": 266500 }, { "epoch": 23.08, "learning_rate": 1.7318867815427313e-05, "loss": 0.6237, "step": 266510 }, { "epoch": 23.08, "learning_rate": 1.7316699433329476e-05, "loss": 0.6845, "step": 266520 }, { "epoch": 23.08, "learning_rate": 1.731453105123164e-05, "loss": 0.5367, "step": 266530 }, { "epoch": 23.08, "learning_rate": 1.7312362669133802e-05, "loss": 0.6288, "step": 266540 }, { "epoch": 23.08, "learning_rate": 1.7310194287035965e-05, "loss": 0.6719, "step": 266550 }, { "epoch": 23.09, "learning_rate": 1.7308025904938128e-05, "loss": 0.7202, "step": 266560 }, { "epoch": 23.09, "learning_rate": 1.730585752284029e-05, "loss": 0.5514, "step": 266570 }, { "epoch": 23.09, "learning_rate": 1.7303689140742454e-05, "loss": 0.6644, "step": 266580 }, { "epoch": 23.09, "learning_rate": 1.7301520758644614e-05, "loss": 0.6824, "step": 266590 }, { "epoch": 23.09, "learning_rate": 1.7299352376546777e-05, "loss": 0.6153, "step": 266600 }, { "epoch": 23.09, "learning_rate": 1.729718399444894e-05, "loss": 0.6207, "step": 266610 }, { "epoch": 23.09, "learning_rate": 1.7295015612351103e-05, "loss": 0.5885, "step": 266620 }, { "epoch": 23.09, "learning_rate": 1.7292847230253266e-05, "loss": 0.6472, "step": 266630 }, { "epoch": 23.09, "learning_rate": 1.729067884815543e-05, "loss": 0.6782, "step": 266640 }, { "epoch": 23.09, "learning_rate": 1.728851046605759e-05, "loss": 0.6438, "step": 266650 }, { "epoch": 23.09, "learning_rate": 1.728634208395975e-05, "loss": 0.5325, "step": 266660 }, { "epoch": 23.1, "learning_rate": 1.7284173701861918e-05, "loss": 0.6895, "step": 266670 }, { "epoch": 23.1, "learning_rate": 1.7282005319764077e-05, "loss": 0.6416, "step": 266680 }, { "epoch": 23.1, "learning_rate": 1.727983693766624e-05, "loss": 0.6468, "step": 266690 }, { "epoch": 23.1, "learning_rate": 1.7277668555568403e-05, "loss": 0.6266, "step": 266700 }, { "epoch": 23.1, "learning_rate": 1.7275500173470566e-05, "loss": 0.6343, "step": 266710 }, { "epoch": 23.1, "learning_rate": 1.727333179137273e-05, "loss": 0.5588, "step": 266720 }, { "epoch": 23.1, "learning_rate": 1.7271163409274892e-05, "loss": 0.6774, "step": 266730 }, { "epoch": 23.1, "learning_rate": 1.7268995027177055e-05, "loss": 0.6275, "step": 266740 }, { "epoch": 23.1, "learning_rate": 1.726682664507922e-05, "loss": 0.6319, "step": 266750 }, { "epoch": 23.1, "learning_rate": 1.7264658262981378e-05, "loss": 0.5569, "step": 266760 }, { "epoch": 23.1, "learning_rate": 1.7262489880883544e-05, "loss": 0.6408, "step": 266770 }, { "epoch": 23.1, "learning_rate": 1.7260321498785704e-05, "loss": 0.5637, "step": 266780 }, { "epoch": 23.11, "learning_rate": 1.7258153116687867e-05, "loss": 0.6605, "step": 266790 }, { "epoch": 23.11, "learning_rate": 1.725598473459003e-05, "loss": 0.5877, "step": 266800 }, { "epoch": 23.11, "learning_rate": 1.7253816352492193e-05, "loss": 0.7025, "step": 266810 }, { "epoch": 23.11, "learning_rate": 1.7251647970394356e-05, "loss": 0.7034, "step": 266820 }, { "epoch": 23.11, "learning_rate": 1.724947958829652e-05, "loss": 0.6606, "step": 266830 }, { "epoch": 23.11, "learning_rate": 1.724731120619868e-05, "loss": 0.5825, "step": 266840 }, { "epoch": 23.11, "learning_rate": 1.724514282410084e-05, "loss": 0.6092, "step": 266850 }, { "epoch": 23.11, "learning_rate": 1.7242974442003005e-05, "loss": 0.6666, "step": 266860 }, { "epoch": 23.11, "learning_rate": 1.7240806059905168e-05, "loss": 0.6656, "step": 266870 }, { "epoch": 23.11, "learning_rate": 1.723863767780733e-05, "loss": 0.6747, "step": 266880 }, { "epoch": 23.11, "learning_rate": 1.7236469295709494e-05, "loss": 0.6518, "step": 266890 }, { "epoch": 23.12, "learning_rate": 1.7234300913611657e-05, "loss": 0.714, "step": 266900 }, { "epoch": 23.12, "learning_rate": 1.7232132531513816e-05, "loss": 0.5978, "step": 266910 }, { "epoch": 23.12, "learning_rate": 1.7229964149415983e-05, "loss": 0.6891, "step": 266920 }, { "epoch": 23.12, "learning_rate": 1.7227795767318142e-05, "loss": 0.6902, "step": 266930 }, { "epoch": 23.12, "learning_rate": 1.7225627385220305e-05, "loss": 0.6209, "step": 266940 }, { "epoch": 23.12, "learning_rate": 1.7223459003122468e-05, "loss": 0.72, "step": 266950 }, { "epoch": 23.12, "learning_rate": 1.722129062102463e-05, "loss": 0.6033, "step": 266960 }, { "epoch": 23.12, "learning_rate": 1.7219122238926794e-05, "loss": 0.7623, "step": 266970 }, { "epoch": 23.12, "learning_rate": 1.7216953856828957e-05, "loss": 0.7626, "step": 266980 }, { "epoch": 23.12, "learning_rate": 1.721478547473112e-05, "loss": 0.5844, "step": 266990 }, { "epoch": 23.12, "learning_rate": 1.7212617092633283e-05, "loss": 0.6563, "step": 267000 }, { "epoch": 23.12, "learning_rate": 1.7210448710535443e-05, "loss": 0.5276, "step": 267010 }, { "epoch": 23.13, "learning_rate": 1.720828032843761e-05, "loss": 0.7173, "step": 267020 }, { "epoch": 23.13, "learning_rate": 1.720611194633977e-05, "loss": 0.6172, "step": 267030 }, { "epoch": 23.13, "learning_rate": 1.7203943564241932e-05, "loss": 0.833, "step": 267040 }, { "epoch": 23.13, "learning_rate": 1.7201775182144095e-05, "loss": 0.7654, "step": 267050 }, { "epoch": 23.13, "learning_rate": 1.7199606800046258e-05, "loss": 0.6888, "step": 267060 }, { "epoch": 23.13, "learning_rate": 1.719743841794842e-05, "loss": 0.7203, "step": 267070 }, { "epoch": 23.13, "learning_rate": 1.7195270035850584e-05, "loss": 0.7235, "step": 267080 }, { "epoch": 23.13, "learning_rate": 1.7193101653752744e-05, "loss": 0.5842, "step": 267090 }, { "epoch": 23.13, "learning_rate": 1.7190933271654907e-05, "loss": 0.5694, "step": 267100 }, { "epoch": 23.13, "learning_rate": 1.718876488955707e-05, "loss": 0.7202, "step": 267110 }, { "epoch": 23.13, "learning_rate": 1.7186596507459233e-05, "loss": 0.5735, "step": 267120 }, { "epoch": 23.14, "learning_rate": 1.7184428125361396e-05, "loss": 0.7008, "step": 267130 }, { "epoch": 23.14, "learning_rate": 1.718225974326356e-05, "loss": 0.7194, "step": 267140 }, { "epoch": 23.14, "learning_rate": 1.718009136116572e-05, "loss": 0.6328, "step": 267150 }, { "epoch": 23.14, "learning_rate": 1.717792297906788e-05, "loss": 0.6141, "step": 267160 }, { "epoch": 23.14, "learning_rate": 1.7175754596970048e-05, "loss": 0.7635, "step": 267170 }, { "epoch": 23.14, "learning_rate": 1.7173586214872207e-05, "loss": 0.747, "step": 267180 }, { "epoch": 23.14, "learning_rate": 1.717141783277437e-05, "loss": 0.6145, "step": 267190 }, { "epoch": 23.14, "learning_rate": 1.7169249450676533e-05, "loss": 0.6969, "step": 267200 }, { "epoch": 23.14, "learning_rate": 1.7167081068578696e-05, "loss": 0.717, "step": 267210 }, { "epoch": 23.14, "learning_rate": 1.716491268648086e-05, "loss": 0.5644, "step": 267220 }, { "epoch": 23.14, "learning_rate": 1.7162744304383022e-05, "loss": 0.7301, "step": 267230 }, { "epoch": 23.14, "learning_rate": 1.7160575922285185e-05, "loss": 0.7387, "step": 267240 }, { "epoch": 23.15, "learning_rate": 1.7158407540187348e-05, "loss": 0.7098, "step": 267250 }, { "epoch": 23.15, "learning_rate": 1.7156239158089508e-05, "loss": 0.6761, "step": 267260 }, { "epoch": 23.15, "learning_rate": 1.7154070775991674e-05, "loss": 0.7268, "step": 267270 }, { "epoch": 23.15, "learning_rate": 1.7151902393893834e-05, "loss": 0.5003, "step": 267280 }, { "epoch": 23.15, "learning_rate": 1.7149734011795997e-05, "loss": 0.6012, "step": 267290 }, { "epoch": 23.15, "learning_rate": 1.714756562969816e-05, "loss": 0.7239, "step": 267300 }, { "epoch": 23.15, "learning_rate": 1.7145397247600323e-05, "loss": 0.616, "step": 267310 }, { "epoch": 23.15, "learning_rate": 1.7143228865502486e-05, "loss": 0.6708, "step": 267320 }, { "epoch": 23.15, "learning_rate": 1.714106048340465e-05, "loss": 0.6069, "step": 267330 }, { "epoch": 23.15, "learning_rate": 1.713889210130681e-05, "loss": 0.646, "step": 267340 }, { "epoch": 23.15, "learning_rate": 1.7136723719208975e-05, "loss": 0.5701, "step": 267350 }, { "epoch": 23.16, "learning_rate": 1.7134555337111134e-05, "loss": 0.7189, "step": 267360 }, { "epoch": 23.16, "learning_rate": 1.7132386955013297e-05, "loss": 0.5804, "step": 267370 }, { "epoch": 23.16, "learning_rate": 1.713021857291546e-05, "loss": 0.7556, "step": 267380 }, { "epoch": 23.16, "learning_rate": 1.7128050190817623e-05, "loss": 0.6714, "step": 267390 }, { "epoch": 23.16, "learning_rate": 1.7125881808719786e-05, "loss": 0.8342, "step": 267400 }, { "epoch": 23.16, "learning_rate": 1.7123713426621946e-05, "loss": 0.835, "step": 267410 }, { "epoch": 23.16, "learning_rate": 1.7121545044524112e-05, "loss": 0.5112, "step": 267420 }, { "epoch": 23.16, "learning_rate": 1.7119376662426272e-05, "loss": 0.6085, "step": 267430 }, { "epoch": 23.16, "learning_rate": 1.7117208280328435e-05, "loss": 0.6852, "step": 267440 }, { "epoch": 23.16, "learning_rate": 1.7115039898230598e-05, "loss": 0.5861, "step": 267450 }, { "epoch": 23.16, "learning_rate": 1.711287151613276e-05, "loss": 0.6009, "step": 267460 }, { "epoch": 23.16, "learning_rate": 1.7110703134034924e-05, "loss": 0.7469, "step": 267470 }, { "epoch": 23.17, "learning_rate": 1.7108534751937087e-05, "loss": 0.6713, "step": 267480 }, { "epoch": 23.17, "learning_rate": 1.710636636983925e-05, "loss": 0.4942, "step": 267490 }, { "epoch": 23.17, "learning_rate": 1.7104197987741413e-05, "loss": 0.6491, "step": 267500 }, { "epoch": 23.17, "learning_rate": 1.7102029605643573e-05, "loss": 0.6414, "step": 267510 }, { "epoch": 23.17, "learning_rate": 1.709986122354574e-05, "loss": 0.6447, "step": 267520 }, { "epoch": 23.17, "learning_rate": 1.70976928414479e-05, "loss": 0.8119, "step": 267530 }, { "epoch": 23.17, "learning_rate": 1.7095524459350062e-05, "loss": 0.607, "step": 267540 }, { "epoch": 23.17, "learning_rate": 1.7093356077252225e-05, "loss": 0.6759, "step": 267550 }, { "epoch": 23.17, "learning_rate": 1.7091187695154388e-05, "loss": 0.646, "step": 267560 }, { "epoch": 23.17, "learning_rate": 1.708901931305655e-05, "loss": 0.4591, "step": 267570 }, { "epoch": 23.17, "learning_rate": 1.7086850930958714e-05, "loss": 0.5394, "step": 267580 }, { "epoch": 23.17, "learning_rate": 1.7084682548860873e-05, "loss": 0.6697, "step": 267590 }, { "epoch": 23.18, "learning_rate": 1.708251416676304e-05, "loss": 0.6006, "step": 267600 }, { "epoch": 23.18, "learning_rate": 1.70803457846652e-05, "loss": 0.6886, "step": 267610 }, { "epoch": 23.18, "learning_rate": 1.7078177402567362e-05, "loss": 0.7141, "step": 267620 }, { "epoch": 23.18, "learning_rate": 1.7076009020469525e-05, "loss": 0.5886, "step": 267630 }, { "epoch": 23.18, "learning_rate": 1.707384063837169e-05, "loss": 0.64, "step": 267640 }, { "epoch": 23.18, "learning_rate": 1.707167225627385e-05, "loss": 0.7075, "step": 267650 }, { "epoch": 23.18, "learning_rate": 1.706950387417601e-05, "loss": 0.587, "step": 267660 }, { "epoch": 23.18, "learning_rate": 1.7067335492078177e-05, "loss": 0.6323, "step": 267670 }, { "epoch": 23.18, "learning_rate": 1.7065167109980337e-05, "loss": 0.5604, "step": 267680 }, { "epoch": 23.18, "learning_rate": 1.70629987278825e-05, "loss": 0.8491, "step": 267690 }, { "epoch": 23.18, "learning_rate": 1.7060830345784663e-05, "loss": 0.653, "step": 267700 }, { "epoch": 23.19, "learning_rate": 1.7058661963686826e-05, "loss": 0.6401, "step": 267710 }, { "epoch": 23.19, "learning_rate": 1.705649358158899e-05, "loss": 0.6525, "step": 267720 }, { "epoch": 23.19, "learning_rate": 1.7054325199491152e-05, "loss": 0.7173, "step": 267730 }, { "epoch": 23.19, "learning_rate": 1.7052156817393315e-05, "loss": 0.6012, "step": 267740 }, { "epoch": 23.19, "learning_rate": 1.7049988435295478e-05, "loss": 0.5884, "step": 267750 }, { "epoch": 23.19, "learning_rate": 1.7047820053197638e-05, "loss": 0.6646, "step": 267760 }, { "epoch": 23.19, "learning_rate": 1.7045651671099804e-05, "loss": 0.717, "step": 267770 }, { "epoch": 23.19, "learning_rate": 1.7043483289001964e-05, "loss": 0.642, "step": 267780 }, { "epoch": 23.19, "learning_rate": 1.7041314906904127e-05, "loss": 0.5603, "step": 267790 }, { "epoch": 23.19, "learning_rate": 1.703914652480629e-05, "loss": 0.6469, "step": 267800 }, { "epoch": 23.19, "learning_rate": 1.7036978142708453e-05, "loss": 0.7866, "step": 267810 }, { "epoch": 23.19, "learning_rate": 1.7034809760610616e-05, "loss": 0.6647, "step": 267820 }, { "epoch": 23.2, "learning_rate": 1.703264137851278e-05, "loss": 0.6082, "step": 267830 }, { "epoch": 23.2, "learning_rate": 1.703047299641494e-05, "loss": 0.6053, "step": 267840 }, { "epoch": 23.2, "learning_rate": 1.7028304614317105e-05, "loss": 0.7388, "step": 267850 }, { "epoch": 23.2, "learning_rate": 1.7026136232219264e-05, "loss": 0.7, "step": 267860 }, { "epoch": 23.2, "learning_rate": 1.7023967850121427e-05, "loss": 0.6262, "step": 267870 }, { "epoch": 23.2, "learning_rate": 1.702179946802359e-05, "loss": 0.7588, "step": 267880 }, { "epoch": 23.2, "learning_rate": 1.7019631085925753e-05, "loss": 0.5775, "step": 267890 }, { "epoch": 23.2, "learning_rate": 1.7017462703827916e-05, "loss": 0.6153, "step": 267900 }, { "epoch": 23.2, "learning_rate": 1.7015294321730076e-05, "loss": 0.6361, "step": 267910 }, { "epoch": 23.2, "learning_rate": 1.7013125939632242e-05, "loss": 0.6778, "step": 267920 }, { "epoch": 23.2, "learning_rate": 1.7010957557534402e-05, "loss": 0.6232, "step": 267930 }, { "epoch": 23.21, "learning_rate": 1.7008789175436565e-05, "loss": 0.7488, "step": 267940 }, { "epoch": 23.21, "learning_rate": 1.7006620793338728e-05, "loss": 0.65, "step": 267950 }, { "epoch": 23.21, "learning_rate": 1.700445241124089e-05, "loss": 0.6817, "step": 267960 }, { "epoch": 23.21, "learning_rate": 1.7002284029143054e-05, "loss": 0.7296, "step": 267970 }, { "epoch": 23.21, "learning_rate": 1.7000115647045217e-05, "loss": 0.5658, "step": 267980 }, { "epoch": 23.21, "learning_rate": 1.699794726494738e-05, "loss": 0.5424, "step": 267990 }, { "epoch": 23.21, "learning_rate": 1.6995778882849543e-05, "loss": 0.7543, "step": 268000 }, { "epoch": 23.21, "learning_rate": 1.6993610500751703e-05, "loss": 0.6219, "step": 268010 }, { "epoch": 23.21, "learning_rate": 1.699144211865387e-05, "loss": 0.5319, "step": 268020 }, { "epoch": 23.21, "learning_rate": 1.698927373655603e-05, "loss": 0.7177, "step": 268030 }, { "epoch": 23.21, "learning_rate": 1.698710535445819e-05, "loss": 0.5775, "step": 268040 }, { "epoch": 23.21, "learning_rate": 1.6984936972360355e-05, "loss": 0.7219, "step": 268050 }, { "epoch": 23.22, "learning_rate": 1.6982768590262518e-05, "loss": 0.6912, "step": 268060 }, { "epoch": 23.22, "learning_rate": 1.698060020816468e-05, "loss": 0.634, "step": 268070 }, { "epoch": 23.22, "learning_rate": 1.6978431826066844e-05, "loss": 0.6528, "step": 268080 }, { "epoch": 23.22, "learning_rate": 1.6976263443969007e-05, "loss": 0.6714, "step": 268090 }, { "epoch": 23.22, "learning_rate": 1.697409506187117e-05, "loss": 0.6787, "step": 268100 }, { "epoch": 23.22, "learning_rate": 1.697192667977333e-05, "loss": 0.6528, "step": 268110 }, { "epoch": 23.22, "learning_rate": 1.6969758297675496e-05, "loss": 0.7672, "step": 268120 }, { "epoch": 23.22, "learning_rate": 1.6967589915577655e-05, "loss": 0.7042, "step": 268130 }, { "epoch": 23.22, "learning_rate": 1.6965421533479818e-05, "loss": 0.7322, "step": 268140 }, { "epoch": 23.22, "learning_rate": 1.696325315138198e-05, "loss": 0.5544, "step": 268150 }, { "epoch": 23.22, "learning_rate": 1.696108476928414e-05, "loss": 0.8077, "step": 268160 }, { "epoch": 23.23, "learning_rate": 1.6958916387186307e-05, "loss": 0.6085, "step": 268170 }, { "epoch": 23.23, "learning_rate": 1.6956748005088467e-05, "loss": 0.5501, "step": 268180 }, { "epoch": 23.23, "learning_rate": 1.695457962299063e-05, "loss": 0.6324, "step": 268190 }, { "epoch": 23.23, "learning_rate": 1.6952411240892793e-05, "loss": 0.603, "step": 268200 }, { "epoch": 23.23, "learning_rate": 1.6950242858794956e-05, "loss": 0.5945, "step": 268210 }, { "epoch": 23.23, "learning_rate": 1.694807447669712e-05, "loss": 0.4647, "step": 268220 }, { "epoch": 23.23, "learning_rate": 1.6945906094599282e-05, "loss": 0.5453, "step": 268230 }, { "epoch": 23.23, "learning_rate": 1.6943737712501445e-05, "loss": 0.6103, "step": 268240 }, { "epoch": 23.23, "learning_rate": 1.6941569330403608e-05, "loss": 0.5451, "step": 268250 }, { "epoch": 23.23, "learning_rate": 1.6939400948305767e-05, "loss": 0.6776, "step": 268260 }, { "epoch": 23.23, "learning_rate": 1.6937232566207934e-05, "loss": 0.6546, "step": 268270 }, { "epoch": 23.23, "learning_rate": 1.6935064184110093e-05, "loss": 0.7457, "step": 268280 }, { "epoch": 23.24, "learning_rate": 1.6932895802012256e-05, "loss": 0.5544, "step": 268290 }, { "epoch": 23.24, "learning_rate": 1.693072741991442e-05, "loss": 0.4872, "step": 268300 }, { "epoch": 23.24, "learning_rate": 1.6928559037816582e-05, "loss": 0.6471, "step": 268310 }, { "epoch": 23.24, "learning_rate": 1.6926390655718746e-05, "loss": 0.6155, "step": 268320 }, { "epoch": 23.24, "learning_rate": 1.692422227362091e-05, "loss": 0.5863, "step": 268330 }, { "epoch": 23.24, "learning_rate": 1.692205389152307e-05, "loss": 0.632, "step": 268340 }, { "epoch": 23.24, "learning_rate": 1.6919885509425235e-05, "loss": 0.687, "step": 268350 }, { "epoch": 23.24, "learning_rate": 1.6917717127327394e-05, "loss": 0.6656, "step": 268360 }, { "epoch": 23.24, "learning_rate": 1.691554874522956e-05, "loss": 0.6837, "step": 268370 }, { "epoch": 23.24, "learning_rate": 1.691338036313172e-05, "loss": 0.651, "step": 268380 }, { "epoch": 23.24, "learning_rate": 1.6911211981033883e-05, "loss": 0.6793, "step": 268390 }, { "epoch": 23.25, "learning_rate": 1.6909043598936046e-05, "loss": 0.6809, "step": 268400 }, { "epoch": 23.25, "learning_rate": 1.6906875216838206e-05, "loss": 0.5467, "step": 268410 }, { "epoch": 23.25, "learning_rate": 1.6904706834740372e-05, "loss": 0.6975, "step": 268420 }, { "epoch": 23.25, "learning_rate": 1.6902538452642532e-05, "loss": 0.6682, "step": 268430 }, { "epoch": 23.25, "learning_rate": 1.6900370070544695e-05, "loss": 0.6895, "step": 268440 }, { "epoch": 23.25, "learning_rate": 1.6898201688446858e-05, "loss": 0.6415, "step": 268450 }, { "epoch": 23.25, "learning_rate": 1.689603330634902e-05, "loss": 0.6011, "step": 268460 }, { "epoch": 23.25, "learning_rate": 1.6893864924251184e-05, "loss": 0.647, "step": 268470 }, { "epoch": 23.25, "learning_rate": 1.6891696542153347e-05, "loss": 0.8054, "step": 268480 }, { "epoch": 23.25, "learning_rate": 1.688952816005551e-05, "loss": 0.5833, "step": 268490 }, { "epoch": 23.25, "learning_rate": 1.6887359777957673e-05, "loss": 0.634, "step": 268500 }, { "epoch": 23.25, "learning_rate": 1.6885191395859832e-05, "loss": 0.6503, "step": 268510 }, { "epoch": 23.26, "learning_rate": 1.6883023013762e-05, "loss": 0.6892, "step": 268520 }, { "epoch": 23.26, "learning_rate": 1.688085463166416e-05, "loss": 0.6794, "step": 268530 }, { "epoch": 23.26, "learning_rate": 1.687868624956632e-05, "loss": 0.5075, "step": 268540 }, { "epoch": 23.26, "learning_rate": 1.6876517867468484e-05, "loss": 0.6996, "step": 268550 }, { "epoch": 23.26, "learning_rate": 1.6874349485370647e-05, "loss": 0.6664, "step": 268560 }, { "epoch": 23.26, "learning_rate": 1.687218110327281e-05, "loss": 0.6911, "step": 268570 }, { "epoch": 23.26, "learning_rate": 1.6870012721174973e-05, "loss": 0.7547, "step": 268580 }, { "epoch": 23.26, "learning_rate": 1.6867844339077136e-05, "loss": 0.5976, "step": 268590 }, { "epoch": 23.26, "learning_rate": 1.68656759569793e-05, "loss": 0.6666, "step": 268600 }, { "epoch": 23.26, "learning_rate": 1.686350757488146e-05, "loss": 0.6879, "step": 268610 }, { "epoch": 23.26, "learning_rate": 1.6861339192783625e-05, "loss": 0.6463, "step": 268620 }, { "epoch": 23.27, "learning_rate": 1.6859170810685785e-05, "loss": 0.6511, "step": 268630 }, { "epoch": 23.27, "learning_rate": 1.6857002428587948e-05, "loss": 0.6992, "step": 268640 }, { "epoch": 23.27, "learning_rate": 1.685483404649011e-05, "loss": 0.6006, "step": 268650 }, { "epoch": 23.27, "learning_rate": 1.685266566439227e-05, "loss": 0.7377, "step": 268660 }, { "epoch": 23.27, "learning_rate": 1.6850497282294437e-05, "loss": 0.826, "step": 268670 }, { "epoch": 23.27, "learning_rate": 1.6848328900196597e-05, "loss": 0.6683, "step": 268680 }, { "epoch": 23.27, "learning_rate": 1.684616051809876e-05, "loss": 0.6648, "step": 268690 }, { "epoch": 23.27, "learning_rate": 1.6843992136000923e-05, "loss": 0.6265, "step": 268700 }, { "epoch": 23.27, "learning_rate": 1.6841823753903086e-05, "loss": 0.6415, "step": 268710 }, { "epoch": 23.27, "learning_rate": 1.683965537180525e-05, "loss": 0.6375, "step": 268720 }, { "epoch": 23.27, "learning_rate": 1.6837486989707412e-05, "loss": 0.5856, "step": 268730 }, { "epoch": 23.27, "learning_rate": 1.6835318607609575e-05, "loss": 0.6165, "step": 268740 }, { "epoch": 23.28, "learning_rate": 1.6833150225511738e-05, "loss": 0.6673, "step": 268750 }, { "epoch": 23.28, "learning_rate": 1.6830981843413897e-05, "loss": 0.7037, "step": 268760 }, { "epoch": 23.28, "learning_rate": 1.6828813461316064e-05, "loss": 0.645, "step": 268770 }, { "epoch": 23.28, "learning_rate": 1.6826645079218223e-05, "loss": 0.5588, "step": 268780 }, { "epoch": 23.28, "learning_rate": 1.6824476697120386e-05, "loss": 0.5825, "step": 268790 }, { "epoch": 23.28, "learning_rate": 1.682230831502255e-05, "loss": 0.704, "step": 268800 }, { "epoch": 23.28, "learning_rate": 1.6820139932924712e-05, "loss": 0.6172, "step": 268810 }, { "epoch": 23.28, "learning_rate": 1.6817971550826875e-05, "loss": 0.6608, "step": 268820 }, { "epoch": 23.28, "learning_rate": 1.681580316872904e-05, "loss": 0.7013, "step": 268830 }, { "epoch": 23.28, "learning_rate": 1.68136347866312e-05, "loss": 0.6076, "step": 268840 }, { "epoch": 23.28, "learning_rate": 1.6811466404533364e-05, "loss": 0.6422, "step": 268850 }, { "epoch": 23.28, "learning_rate": 1.6809298022435524e-05, "loss": 0.7638, "step": 268860 }, { "epoch": 23.29, "learning_rate": 1.680712964033769e-05, "loss": 0.6384, "step": 268870 }, { "epoch": 23.29, "learning_rate": 1.680496125823985e-05, "loss": 0.7482, "step": 268880 }, { "epoch": 23.29, "learning_rate": 1.6802792876142013e-05, "loss": 0.685, "step": 268890 }, { "epoch": 23.29, "learning_rate": 1.6800624494044176e-05, "loss": 0.7817, "step": 268900 }, { "epoch": 23.29, "learning_rate": 1.679845611194634e-05, "loss": 0.7042, "step": 268910 }, { "epoch": 23.29, "learning_rate": 1.6796287729848502e-05, "loss": 0.649, "step": 268920 }, { "epoch": 23.29, "learning_rate": 1.679411934775066e-05, "loss": 0.7773, "step": 268930 }, { "epoch": 23.29, "learning_rate": 1.6791950965652825e-05, "loss": 0.5364, "step": 268940 }, { "epoch": 23.29, "learning_rate": 1.6789782583554988e-05, "loss": 0.599, "step": 268950 }, { "epoch": 23.29, "learning_rate": 1.678761420145715e-05, "loss": 0.5178, "step": 268960 }, { "epoch": 23.29, "learning_rate": 1.6785445819359314e-05, "loss": 0.745, "step": 268970 }, { "epoch": 23.3, "learning_rate": 1.6783277437261477e-05, "loss": 0.6283, "step": 268980 }, { "epoch": 23.3, "learning_rate": 1.678110905516364e-05, "loss": 0.6894, "step": 268990 }, { "epoch": 23.3, "learning_rate": 1.6778940673065803e-05, "loss": 0.6111, "step": 269000 }, { "epoch": 23.3, "learning_rate": 1.6776772290967962e-05, "loss": 0.6684, "step": 269010 }, { "epoch": 23.3, "learning_rate": 1.677460390887013e-05, "loss": 0.6785, "step": 269020 }, { "epoch": 23.3, "learning_rate": 1.6772435526772288e-05, "loss": 0.6311, "step": 269030 }, { "epoch": 23.3, "learning_rate": 1.677026714467445e-05, "loss": 0.7277, "step": 269040 }, { "epoch": 23.3, "learning_rate": 1.6768098762576614e-05, "loss": 0.683, "step": 269050 }, { "epoch": 23.3, "learning_rate": 1.6765930380478777e-05, "loss": 0.6017, "step": 269060 }, { "epoch": 23.3, "learning_rate": 1.676376199838094e-05, "loss": 0.7131, "step": 269070 }, { "epoch": 23.3, "learning_rate": 1.6761593616283103e-05, "loss": 0.7358, "step": 269080 }, { "epoch": 23.3, "learning_rate": 1.6759425234185266e-05, "loss": 0.6473, "step": 269090 }, { "epoch": 23.31, "learning_rate": 1.675725685208743e-05, "loss": 0.6215, "step": 269100 }, { "epoch": 23.31, "learning_rate": 1.675508846998959e-05, "loss": 0.6854, "step": 269110 }, { "epoch": 23.31, "learning_rate": 1.6752920087891755e-05, "loss": 0.8061, "step": 269120 }, { "epoch": 23.31, "learning_rate": 1.6750751705793915e-05, "loss": 0.6268, "step": 269130 }, { "epoch": 23.31, "learning_rate": 1.6748583323696078e-05, "loss": 0.6332, "step": 269140 }, { "epoch": 23.31, "learning_rate": 1.674641494159824e-05, "loss": 0.6948, "step": 269150 }, { "epoch": 23.31, "learning_rate": 1.6744246559500404e-05, "loss": 0.6596, "step": 269160 }, { "epoch": 23.31, "learning_rate": 1.6742078177402567e-05, "loss": 0.473, "step": 269170 }, { "epoch": 23.31, "learning_rate": 1.6739909795304727e-05, "loss": 0.6242, "step": 269180 }, { "epoch": 23.31, "learning_rate": 1.6737741413206893e-05, "loss": 0.5602, "step": 269190 }, { "epoch": 23.31, "learning_rate": 1.6735573031109053e-05, "loss": 0.6618, "step": 269200 }, { "epoch": 23.32, "learning_rate": 1.6733404649011216e-05, "loss": 0.866, "step": 269210 }, { "epoch": 23.32, "learning_rate": 1.673123626691338e-05, "loss": 0.6162, "step": 269220 }, { "epoch": 23.32, "learning_rate": 1.672906788481554e-05, "loss": 0.6973, "step": 269230 }, { "epoch": 23.32, "learning_rate": 1.6726899502717705e-05, "loss": 0.6565, "step": 269240 }, { "epoch": 23.32, "learning_rate": 1.6724731120619868e-05, "loss": 0.5842, "step": 269250 }, { "epoch": 23.32, "learning_rate": 1.6722562738522027e-05, "loss": 0.7294, "step": 269260 }, { "epoch": 23.32, "learning_rate": 1.6720394356424194e-05, "loss": 0.7587, "step": 269270 }, { "epoch": 23.32, "learning_rate": 1.6718225974326353e-05, "loss": 0.6067, "step": 269280 }, { "epoch": 23.32, "learning_rate": 1.6716057592228516e-05, "loss": 0.7245, "step": 269290 }, { "epoch": 23.32, "learning_rate": 1.671388921013068e-05, "loss": 0.6811, "step": 269300 }, { "epoch": 23.32, "learning_rate": 1.6711720828032842e-05, "loss": 0.634, "step": 269310 }, { "epoch": 23.32, "learning_rate": 1.6709552445935005e-05, "loss": 0.7395, "step": 269320 }, { "epoch": 23.33, "learning_rate": 1.6707384063837168e-05, "loss": 0.6013, "step": 269330 }, { "epoch": 23.33, "learning_rate": 1.670521568173933e-05, "loss": 0.7132, "step": 269340 }, { "epoch": 23.33, "learning_rate": 1.6703047299641494e-05, "loss": 0.7553, "step": 269350 }, { "epoch": 23.33, "learning_rate": 1.6700878917543654e-05, "loss": 0.6813, "step": 269360 }, { "epoch": 23.33, "learning_rate": 1.669871053544582e-05, "loss": 0.7746, "step": 269370 }, { "epoch": 23.33, "learning_rate": 1.669654215334798e-05, "loss": 0.7077, "step": 269380 }, { "epoch": 23.33, "learning_rate": 1.6694373771250143e-05, "loss": 0.6765, "step": 269390 }, { "epoch": 23.33, "learning_rate": 1.6692205389152306e-05, "loss": 0.673, "step": 269400 }, { "epoch": 23.33, "learning_rate": 1.669003700705447e-05, "loss": 0.7189, "step": 269410 }, { "epoch": 23.33, "learning_rate": 1.6687868624956632e-05, "loss": 0.6012, "step": 269420 }, { "epoch": 23.33, "learning_rate": 1.668570024285879e-05, "loss": 0.6798, "step": 269430 }, { "epoch": 23.34, "learning_rate": 1.6683531860760958e-05, "loss": 0.6697, "step": 269440 }, { "epoch": 23.34, "learning_rate": 1.6681363478663117e-05, "loss": 0.6608, "step": 269450 }, { "epoch": 23.34, "learning_rate": 1.667919509656528e-05, "loss": 0.676, "step": 269460 }, { "epoch": 23.34, "learning_rate": 1.6677026714467443e-05, "loss": 0.5974, "step": 269470 }, { "epoch": 23.34, "learning_rate": 1.6674858332369606e-05, "loss": 0.7136, "step": 269480 }, { "epoch": 23.34, "learning_rate": 1.667268995027177e-05, "loss": 0.7316, "step": 269490 }, { "epoch": 23.34, "learning_rate": 1.6670521568173932e-05, "loss": 0.5789, "step": 269500 }, { "epoch": 23.34, "learning_rate": 1.6668353186076092e-05, "loss": 0.6773, "step": 269510 }, { "epoch": 23.34, "learning_rate": 1.666618480397826e-05, "loss": 0.592, "step": 269520 }, { "epoch": 23.34, "learning_rate": 1.6664016421880418e-05, "loss": 0.6193, "step": 269530 }, { "epoch": 23.34, "learning_rate": 1.666184803978258e-05, "loss": 0.7945, "step": 269540 }, { "epoch": 23.34, "learning_rate": 1.6659679657684744e-05, "loss": 0.7379, "step": 269550 }, { "epoch": 23.35, "learning_rate": 1.6657511275586907e-05, "loss": 0.7812, "step": 269560 }, { "epoch": 23.35, "learning_rate": 1.665534289348907e-05, "loss": 0.5286, "step": 269570 }, { "epoch": 23.35, "learning_rate": 1.6653174511391233e-05, "loss": 0.6573, "step": 269580 }, { "epoch": 23.35, "learning_rate": 1.6651006129293396e-05, "loss": 0.701, "step": 269590 }, { "epoch": 23.35, "learning_rate": 1.664883774719556e-05, "loss": 0.6185, "step": 269600 }, { "epoch": 23.35, "learning_rate": 1.664666936509772e-05, "loss": 0.5898, "step": 269610 }, { "epoch": 23.35, "learning_rate": 1.6644500982999885e-05, "loss": 0.6605, "step": 269620 }, { "epoch": 23.35, "learning_rate": 1.6642332600902045e-05, "loss": 0.6383, "step": 269630 }, { "epoch": 23.35, "learning_rate": 1.6640164218804208e-05, "loss": 0.6708, "step": 269640 }, { "epoch": 23.35, "learning_rate": 1.663799583670637e-05, "loss": 0.6949, "step": 269650 }, { "epoch": 23.35, "learning_rate": 1.6635827454608534e-05, "loss": 0.7736, "step": 269660 }, { "epoch": 23.36, "learning_rate": 1.6633659072510697e-05, "loss": 0.7598, "step": 269670 }, { "epoch": 23.36, "learning_rate": 1.6631490690412856e-05, "loss": 0.6985, "step": 269680 }, { "epoch": 23.36, "learning_rate": 1.6629322308315023e-05, "loss": 0.7514, "step": 269690 }, { "epoch": 23.36, "learning_rate": 1.6627153926217182e-05, "loss": 0.6886, "step": 269700 }, { "epoch": 23.36, "learning_rate": 1.6624985544119345e-05, "loss": 0.6729, "step": 269710 }, { "epoch": 23.36, "learning_rate": 1.662281716202151e-05, "loss": 0.6207, "step": 269720 }, { "epoch": 23.36, "learning_rate": 1.662064877992367e-05, "loss": 0.762, "step": 269730 }, { "epoch": 23.36, "learning_rate": 1.6618480397825834e-05, "loss": 0.7161, "step": 269740 }, { "epoch": 23.36, "learning_rate": 1.6616312015727997e-05, "loss": 0.6245, "step": 269750 }, { "epoch": 23.36, "learning_rate": 1.6614143633630157e-05, "loss": 0.7363, "step": 269760 }, { "epoch": 23.36, "learning_rate": 1.6611975251532323e-05, "loss": 0.6806, "step": 269770 }, { "epoch": 23.36, "learning_rate": 1.6609806869434483e-05, "loss": 0.613, "step": 269780 }, { "epoch": 23.37, "learning_rate": 1.6607638487336646e-05, "loss": 0.5434, "step": 269790 }, { "epoch": 23.37, "learning_rate": 1.660547010523881e-05, "loss": 0.658, "step": 269800 }, { "epoch": 23.37, "learning_rate": 1.6603301723140972e-05, "loss": 0.5975, "step": 269810 }, { "epoch": 23.37, "learning_rate": 1.6601133341043135e-05, "loss": 0.7683, "step": 269820 }, { "epoch": 23.37, "learning_rate": 1.6598964958945298e-05, "loss": 0.6128, "step": 269830 }, { "epoch": 23.37, "learning_rate": 1.659679657684746e-05, "loss": 0.7596, "step": 269840 }, { "epoch": 23.37, "learning_rate": 1.6594628194749624e-05, "loss": 0.6384, "step": 269850 }, { "epoch": 23.37, "learning_rate": 1.6592459812651784e-05, "loss": 0.7136, "step": 269860 }, { "epoch": 23.37, "learning_rate": 1.659029143055395e-05, "loss": 0.4917, "step": 269870 }, { "epoch": 23.37, "learning_rate": 1.658812304845611e-05, "loss": 0.71, "step": 269880 }, { "epoch": 23.37, "learning_rate": 1.6585954666358273e-05, "loss": 0.552, "step": 269890 }, { "epoch": 23.38, "learning_rate": 1.6583786284260436e-05, "loss": 0.5507, "step": 269900 }, { "epoch": 23.38, "learning_rate": 1.65816179021626e-05, "loss": 0.6631, "step": 269910 }, { "epoch": 23.38, "learning_rate": 1.657944952006476e-05, "loss": 0.4958, "step": 269920 }, { "epoch": 23.38, "learning_rate": 1.657728113796692e-05, "loss": 0.7252, "step": 269930 }, { "epoch": 23.38, "learning_rate": 1.6575112755869088e-05, "loss": 0.6256, "step": 269940 }, { "epoch": 23.38, "learning_rate": 1.6572944373771247e-05, "loss": 0.6632, "step": 269950 }, { "epoch": 23.38, "learning_rate": 1.657077599167341e-05, "loss": 0.6831, "step": 269960 }, { "epoch": 23.38, "learning_rate": 1.6568607609575573e-05, "loss": 0.6265, "step": 269970 }, { "epoch": 23.38, "learning_rate": 1.6566439227477736e-05, "loss": 0.696, "step": 269980 }, { "epoch": 23.38, "learning_rate": 1.65642708453799e-05, "loss": 0.6648, "step": 269990 }, { "epoch": 23.38, "learning_rate": 1.6562102463282062e-05, "loss": 0.5647, "step": 270000 }, { "epoch": 23.38, "learning_rate": 1.6559934081184222e-05, "loss": 0.7819, "step": 270010 }, { "epoch": 23.39, "learning_rate": 1.655776569908639e-05, "loss": 0.6304, "step": 270020 }, { "epoch": 23.39, "learning_rate": 1.6555597316988548e-05, "loss": 0.8299, "step": 270030 }, { "epoch": 23.39, "learning_rate": 1.655342893489071e-05, "loss": 0.7293, "step": 270040 }, { "epoch": 23.39, "learning_rate": 1.6551260552792874e-05, "loss": 0.8259, "step": 270050 }, { "epoch": 23.39, "learning_rate": 1.6549092170695037e-05, "loss": 0.6683, "step": 270060 }, { "epoch": 23.39, "learning_rate": 1.65469237885972e-05, "loss": 0.7669, "step": 270070 }, { "epoch": 23.39, "learning_rate": 1.6544755406499363e-05, "loss": 0.6076, "step": 270080 }, { "epoch": 23.39, "learning_rate": 1.6542587024401526e-05, "loss": 0.6502, "step": 270090 }, { "epoch": 23.39, "learning_rate": 1.654041864230369e-05, "loss": 0.5959, "step": 270100 }, { "epoch": 23.39, "learning_rate": 1.653825026020585e-05, "loss": 0.679, "step": 270110 }, { "epoch": 23.39, "learning_rate": 1.6536081878108015e-05, "loss": 0.725, "step": 270120 }, { "epoch": 23.39, "learning_rate": 1.6533913496010175e-05, "loss": 0.6158, "step": 270130 }, { "epoch": 23.4, "learning_rate": 1.6531745113912338e-05, "loss": 0.6697, "step": 270140 }, { "epoch": 23.4, "learning_rate": 1.65295767318145e-05, "loss": 0.6994, "step": 270150 }, { "epoch": 23.4, "learning_rate": 1.6527408349716664e-05, "loss": 0.709, "step": 270160 }, { "epoch": 23.4, "learning_rate": 1.6525239967618827e-05, "loss": 0.7627, "step": 270170 }, { "epoch": 23.4, "learning_rate": 1.6523071585520986e-05, "loss": 0.6848, "step": 270180 }, { "epoch": 23.4, "learning_rate": 1.6520903203423153e-05, "loss": 0.6253, "step": 270190 }, { "epoch": 23.4, "learning_rate": 1.6518734821325312e-05, "loss": 0.484, "step": 270200 }, { "epoch": 23.4, "learning_rate": 1.6516566439227475e-05, "loss": 0.6263, "step": 270210 }, { "epoch": 23.4, "learning_rate": 1.6514398057129638e-05, "loss": 0.8145, "step": 270220 }, { "epoch": 23.4, "learning_rate": 1.65122296750318e-05, "loss": 0.6211, "step": 270230 }, { "epoch": 23.4, "learning_rate": 1.6510061292933964e-05, "loss": 0.6717, "step": 270240 }, { "epoch": 23.41, "learning_rate": 1.6507892910836127e-05, "loss": 0.6202, "step": 270250 }, { "epoch": 23.41, "learning_rate": 1.650572452873829e-05, "loss": 0.7083, "step": 270260 }, { "epoch": 23.41, "learning_rate": 1.6503556146640453e-05, "loss": 0.7196, "step": 270270 }, { "epoch": 23.41, "learning_rate": 1.6501387764542613e-05, "loss": 0.6506, "step": 270280 }, { "epoch": 23.41, "learning_rate": 1.6499219382444776e-05, "loss": 0.8373, "step": 270290 }, { "epoch": 23.41, "learning_rate": 1.649705100034694e-05, "loss": 0.6876, "step": 270300 }, { "epoch": 23.41, "learning_rate": 1.6494882618249102e-05, "loss": 0.7796, "step": 270310 }, { "epoch": 23.41, "learning_rate": 1.6492714236151265e-05, "loss": 0.4968, "step": 270320 }, { "epoch": 23.41, "learning_rate": 1.6490545854053428e-05, "loss": 0.659, "step": 270330 }, { "epoch": 23.41, "learning_rate": 1.648837747195559e-05, "loss": 0.6631, "step": 270340 }, { "epoch": 23.41, "learning_rate": 1.6486209089857754e-05, "loss": 0.6587, "step": 270350 }, { "epoch": 23.41, "learning_rate": 1.6484040707759914e-05, "loss": 0.5667, "step": 270360 }, { "epoch": 23.42, "learning_rate": 1.648187232566208e-05, "loss": 0.6592, "step": 270370 }, { "epoch": 23.42, "learning_rate": 1.647970394356424e-05, "loss": 0.6979, "step": 270380 }, { "epoch": 23.42, "learning_rate": 1.6477535561466403e-05, "loss": 0.6387, "step": 270390 }, { "epoch": 23.42, "learning_rate": 1.6475367179368566e-05, "loss": 0.643, "step": 270400 }, { "epoch": 23.42, "learning_rate": 1.647319879727073e-05, "loss": 0.6897, "step": 270410 }, { "epoch": 23.42, "learning_rate": 1.647103041517289e-05, "loss": 0.6037, "step": 270420 }, { "epoch": 23.42, "learning_rate": 1.6468862033075055e-05, "loss": 0.677, "step": 270430 }, { "epoch": 23.42, "learning_rate": 1.6466693650977218e-05, "loss": 0.6255, "step": 270440 }, { "epoch": 23.42, "learning_rate": 1.6464525268879377e-05, "loss": 0.7329, "step": 270450 }, { "epoch": 23.42, "learning_rate": 1.646235688678154e-05, "loss": 0.5852, "step": 270460 }, { "epoch": 23.42, "learning_rate": 1.6460188504683703e-05, "loss": 0.6534, "step": 270470 }, { "epoch": 23.43, "learning_rate": 1.6458020122585866e-05, "loss": 0.6543, "step": 270480 }, { "epoch": 23.43, "learning_rate": 1.645585174048803e-05, "loss": 0.5393, "step": 270490 }, { "epoch": 23.43, "learning_rate": 1.6453683358390192e-05, "loss": 0.5982, "step": 270500 }, { "epoch": 23.43, "learning_rate": 1.6451514976292355e-05, "loss": 0.6034, "step": 270510 }, { "epoch": 23.43, "learning_rate": 1.6449346594194518e-05, "loss": 0.718, "step": 270520 }, { "epoch": 23.43, "learning_rate": 1.6447178212096678e-05, "loss": 0.5678, "step": 270530 }, { "epoch": 23.43, "learning_rate": 1.6445009829998844e-05, "loss": 0.5387, "step": 270540 }, { "epoch": 23.43, "learning_rate": 1.6442841447901004e-05, "loss": 0.5098, "step": 270550 }, { "epoch": 23.43, "learning_rate": 1.6440673065803167e-05, "loss": 0.6907, "step": 270560 }, { "epoch": 23.43, "learning_rate": 1.643850468370533e-05, "loss": 0.6095, "step": 270570 }, { "epoch": 23.43, "learning_rate": 1.6436336301607493e-05, "loss": 0.6878, "step": 270580 }, { "epoch": 23.43, "learning_rate": 1.6434167919509656e-05, "loss": 0.6094, "step": 270590 }, { "epoch": 23.44, "learning_rate": 1.643199953741182e-05, "loss": 0.5266, "step": 270600 }, { "epoch": 23.44, "learning_rate": 1.642983115531398e-05, "loss": 0.55, "step": 270610 }, { "epoch": 23.44, "learning_rate": 1.6427662773216145e-05, "loss": 0.6113, "step": 270620 }, { "epoch": 23.44, "learning_rate": 1.6425494391118304e-05, "loss": 0.5944, "step": 270630 }, { "epoch": 23.44, "learning_rate": 1.6423326009020467e-05, "loss": 0.6055, "step": 270640 }, { "epoch": 23.44, "learning_rate": 1.642115762692263e-05, "loss": 0.6398, "step": 270650 }, { "epoch": 23.44, "learning_rate": 1.6418989244824793e-05, "loss": 0.8263, "step": 270660 }, { "epoch": 23.44, "learning_rate": 1.6416820862726956e-05, "loss": 0.7174, "step": 270670 }, { "epoch": 23.44, "learning_rate": 1.641465248062912e-05, "loss": 0.7558, "step": 270680 }, { "epoch": 23.44, "learning_rate": 1.6412484098531282e-05, "loss": 0.7402, "step": 270690 }, { "epoch": 23.44, "learning_rate": 1.6410315716433442e-05, "loss": 0.6671, "step": 270700 }, { "epoch": 23.45, "learning_rate": 1.6408147334335605e-05, "loss": 0.6056, "step": 270710 }, { "epoch": 23.45, "learning_rate": 1.6405978952237768e-05, "loss": 0.6084, "step": 270720 }, { "epoch": 23.45, "learning_rate": 1.640381057013993e-05, "loss": 0.5672, "step": 270730 }, { "epoch": 23.45, "learning_rate": 1.6401642188042094e-05, "loss": 0.6709, "step": 270740 }, { "epoch": 23.45, "learning_rate": 1.6399473805944257e-05, "loss": 0.6509, "step": 270750 }, { "epoch": 23.45, "learning_rate": 1.639730542384642e-05, "loss": 0.6839, "step": 270760 }, { "epoch": 23.45, "learning_rate": 1.6395137041748583e-05, "loss": 0.626, "step": 270770 }, { "epoch": 23.45, "learning_rate": 1.6392968659650743e-05, "loss": 0.5609, "step": 270780 }, { "epoch": 23.45, "learning_rate": 1.639080027755291e-05, "loss": 0.7448, "step": 270790 }, { "epoch": 23.45, "learning_rate": 1.638863189545507e-05, "loss": 0.5723, "step": 270800 }, { "epoch": 23.45, "learning_rate": 1.6386463513357232e-05, "loss": 0.6116, "step": 270810 }, { "epoch": 23.45, "learning_rate": 1.6384295131259395e-05, "loss": 0.7563, "step": 270820 }, { "epoch": 23.46, "learning_rate": 1.6382126749161558e-05, "loss": 0.6277, "step": 270830 }, { "epoch": 23.46, "learning_rate": 1.637995836706372e-05, "loss": 0.6298, "step": 270840 }, { "epoch": 23.46, "learning_rate": 1.6377789984965884e-05, "loss": 0.7835, "step": 270850 }, { "epoch": 23.46, "learning_rate": 1.6375621602868043e-05, "loss": 0.7206, "step": 270860 }, { "epoch": 23.46, "learning_rate": 1.637345322077021e-05, "loss": 0.6478, "step": 270870 }, { "epoch": 23.46, "learning_rate": 1.637128483867237e-05, "loss": 0.8, "step": 270880 }, { "epoch": 23.46, "learning_rate": 1.6369116456574532e-05, "loss": 0.6917, "step": 270890 }, { "epoch": 23.46, "learning_rate": 1.6366948074476695e-05, "loss": 0.7892, "step": 270900 }, { "epoch": 23.46, "learning_rate": 1.636477969237886e-05, "loss": 0.6185, "step": 270910 }, { "epoch": 23.46, "learning_rate": 1.636261131028102e-05, "loss": 0.5995, "step": 270920 }, { "epoch": 23.46, "learning_rate": 1.6360442928183184e-05, "loss": 0.6033, "step": 270930 }, { "epoch": 23.47, "learning_rate": 1.6358274546085347e-05, "loss": 0.6176, "step": 270940 }, { "epoch": 23.47, "learning_rate": 1.6356106163987507e-05, "loss": 0.6213, "step": 270950 }, { "epoch": 23.47, "learning_rate": 1.635393778188967e-05, "loss": 0.8615, "step": 270960 }, { "epoch": 23.47, "learning_rate": 1.6351769399791833e-05, "loss": 0.6481, "step": 270970 }, { "epoch": 23.47, "learning_rate": 1.6349601017693996e-05, "loss": 0.5367, "step": 270980 }, { "epoch": 23.47, "learning_rate": 1.634743263559616e-05, "loss": 0.6632, "step": 270990 }, { "epoch": 23.47, "learning_rate": 1.6345264253498322e-05, "loss": 0.5789, "step": 271000 }, { "epoch": 23.47, "learning_rate": 1.6343095871400485e-05, "loss": 0.6172, "step": 271010 }, { "epoch": 23.47, "learning_rate": 1.6340927489302648e-05, "loss": 0.6969, "step": 271020 }, { "epoch": 23.47, "learning_rate": 1.6338759107204808e-05, "loss": 0.7015, "step": 271030 }, { "epoch": 23.47, "learning_rate": 1.6336590725106974e-05, "loss": 0.7228, "step": 271040 }, { "epoch": 23.47, "learning_rate": 1.6334422343009134e-05, "loss": 0.6465, "step": 271050 }, { "epoch": 23.48, "learning_rate": 1.6332253960911297e-05, "loss": 0.5271, "step": 271060 }, { "epoch": 23.48, "learning_rate": 1.633008557881346e-05, "loss": 0.6928, "step": 271070 }, { "epoch": 23.48, "learning_rate": 1.6327917196715623e-05, "loss": 0.8374, "step": 271080 }, { "epoch": 23.48, "learning_rate": 1.6325748814617786e-05, "loss": 0.5347, "step": 271090 }, { "epoch": 23.48, "learning_rate": 1.632358043251995e-05, "loss": 0.611, "step": 271100 }, { "epoch": 23.48, "learning_rate": 1.6321412050422108e-05, "loss": 0.576, "step": 271110 }, { "epoch": 23.48, "learning_rate": 1.6319243668324275e-05, "loss": 0.6566, "step": 271120 }, { "epoch": 23.48, "learning_rate": 1.6317075286226434e-05, "loss": 0.62, "step": 271130 }, { "epoch": 23.48, "learning_rate": 1.6314906904128597e-05, "loss": 0.6816, "step": 271140 }, { "epoch": 23.48, "learning_rate": 1.631273852203076e-05, "loss": 0.6143, "step": 271150 }, { "epoch": 23.48, "learning_rate": 1.6310570139932923e-05, "loss": 0.6314, "step": 271160 }, { "epoch": 23.49, "learning_rate": 1.6308401757835086e-05, "loss": 0.7357, "step": 271170 }, { "epoch": 23.49, "learning_rate": 1.630623337573725e-05, "loss": 0.7269, "step": 271180 }, { "epoch": 23.49, "learning_rate": 1.6304064993639412e-05, "loss": 0.7631, "step": 271190 }, { "epoch": 23.49, "learning_rate": 1.6301896611541572e-05, "loss": 0.6078, "step": 271200 }, { "epoch": 23.49, "learning_rate": 1.6299728229443735e-05, "loss": 0.6795, "step": 271210 }, { "epoch": 23.49, "learning_rate": 1.6297559847345898e-05, "loss": 0.5891, "step": 271220 }, { "epoch": 23.49, "learning_rate": 1.629539146524806e-05, "loss": 0.7126, "step": 271230 }, { "epoch": 23.49, "learning_rate": 1.6293223083150224e-05, "loss": 0.6605, "step": 271240 }, { "epoch": 23.49, "learning_rate": 1.6291054701052387e-05, "loss": 0.6353, "step": 271250 }, { "epoch": 23.49, "learning_rate": 1.628888631895455e-05, "loss": 0.592, "step": 271260 }, { "epoch": 23.49, "learning_rate": 1.6286717936856713e-05, "loss": 0.7108, "step": 271270 }, { "epoch": 23.49, "learning_rate": 1.6284549554758873e-05, "loss": 0.8345, "step": 271280 }, { "epoch": 23.5, "learning_rate": 1.628238117266104e-05, "loss": 0.6137, "step": 271290 }, { "epoch": 23.5, "learning_rate": 1.62802127905632e-05, "loss": 0.6483, "step": 271300 }, { "epoch": 23.5, "learning_rate": 1.627804440846536e-05, "loss": 0.67, "step": 271310 }, { "epoch": 23.5, "learning_rate": 1.6275876026367525e-05, "loss": 0.5336, "step": 271320 }, { "epoch": 23.5, "learning_rate": 1.6273707644269688e-05, "loss": 0.657, "step": 271330 }, { "epoch": 23.5, "learning_rate": 1.627153926217185e-05, "loss": 0.6965, "step": 271340 }, { "epoch": 23.5, "learning_rate": 1.6269370880074014e-05, "loss": 0.7699, "step": 271350 }, { "epoch": 23.5, "learning_rate": 1.6267202497976173e-05, "loss": 0.8251, "step": 271360 }, { "epoch": 23.5, "learning_rate": 1.626503411587834e-05, "loss": 0.7263, "step": 271370 }, { "epoch": 23.5, "learning_rate": 1.62628657337805e-05, "loss": 0.6694, "step": 271380 }, { "epoch": 23.5, "learning_rate": 1.6260697351682662e-05, "loss": 0.7233, "step": 271390 }, { "epoch": 23.5, "learning_rate": 1.6258528969584825e-05, "loss": 0.6626, "step": 271400 }, { "epoch": 23.51, "learning_rate": 1.6256360587486988e-05, "loss": 0.6424, "step": 271410 }, { "epoch": 23.51, "learning_rate": 1.625419220538915e-05, "loss": 0.6956, "step": 271420 }, { "epoch": 23.51, "learning_rate": 1.6252023823291314e-05, "loss": 0.7046, "step": 271430 }, { "epoch": 23.51, "learning_rate": 1.6249855441193477e-05, "loss": 0.7483, "step": 271440 }, { "epoch": 23.51, "learning_rate": 1.624768705909564e-05, "loss": 0.7232, "step": 271450 }, { "epoch": 23.51, "learning_rate": 1.62455186769978e-05, "loss": 0.6599, "step": 271460 }, { "epoch": 23.51, "learning_rate": 1.6243350294899963e-05, "loss": 0.8223, "step": 271470 }, { "epoch": 23.51, "learning_rate": 1.6241181912802126e-05, "loss": 0.5031, "step": 271480 }, { "epoch": 23.51, "learning_rate": 1.623901353070429e-05, "loss": 0.6074, "step": 271490 }, { "epoch": 23.51, "learning_rate": 1.6236845148606452e-05, "loss": 0.779, "step": 271500 }, { "epoch": 23.51, "learning_rate": 1.6234676766508615e-05, "loss": 0.6067, "step": 271510 }, { "epoch": 23.52, "learning_rate": 1.6232508384410778e-05, "loss": 0.7266, "step": 271520 }, { "epoch": 23.52, "learning_rate": 1.6230340002312937e-05, "loss": 0.6876, "step": 271530 }, { "epoch": 23.52, "learning_rate": 1.6228171620215104e-05, "loss": 0.6178, "step": 271540 }, { "epoch": 23.52, "learning_rate": 1.6226003238117264e-05, "loss": 0.6687, "step": 271550 }, { "epoch": 23.52, "learning_rate": 1.6223834856019427e-05, "loss": 0.717, "step": 271560 }, { "epoch": 23.52, "learning_rate": 1.622166647392159e-05, "loss": 0.7181, "step": 271570 }, { "epoch": 23.52, "learning_rate": 1.6219498091823753e-05, "loss": 0.5732, "step": 271580 }, { "epoch": 23.52, "learning_rate": 1.6217329709725916e-05, "loss": 0.854, "step": 271590 }, { "epoch": 23.52, "learning_rate": 1.621516132762808e-05, "loss": 0.5535, "step": 271600 }, { "epoch": 23.52, "learning_rate": 1.621299294553024e-05, "loss": 0.5564, "step": 271610 }, { "epoch": 23.52, "learning_rate": 1.6210824563432405e-05, "loss": 0.614, "step": 271620 }, { "epoch": 23.52, "learning_rate": 1.6208656181334564e-05, "loss": 0.6174, "step": 271630 }, { "epoch": 23.53, "learning_rate": 1.6206487799236727e-05, "loss": 0.6268, "step": 271640 }, { "epoch": 23.53, "learning_rate": 1.620431941713889e-05, "loss": 0.7384, "step": 271650 }, { "epoch": 23.53, "learning_rate": 1.6202151035041053e-05, "loss": 0.7062, "step": 271660 }, { "epoch": 23.53, "learning_rate": 1.6199982652943216e-05, "loss": 0.6734, "step": 271670 }, { "epoch": 23.53, "learning_rate": 1.619781427084538e-05, "loss": 0.6462, "step": 271680 }, { "epoch": 23.53, "learning_rate": 1.6195645888747542e-05, "loss": 0.6775, "step": 271690 }, { "epoch": 23.53, "learning_rate": 1.6193477506649705e-05, "loss": 0.6193, "step": 271700 }, { "epoch": 23.53, "learning_rate": 1.6191309124551865e-05, "loss": 0.6727, "step": 271710 }, { "epoch": 23.53, "learning_rate": 1.6189140742454028e-05, "loss": 0.6616, "step": 271720 }, { "epoch": 23.53, "learning_rate": 1.618697236035619e-05, "loss": 0.6426, "step": 271730 }, { "epoch": 23.53, "learning_rate": 1.6184803978258354e-05, "loss": 0.6551, "step": 271740 }, { "epoch": 23.54, "learning_rate": 1.6182635596160517e-05, "loss": 0.7185, "step": 271750 }, { "epoch": 23.54, "learning_rate": 1.618046721406268e-05, "loss": 0.5888, "step": 271760 }, { "epoch": 23.54, "learning_rate": 1.6178298831964843e-05, "loss": 0.5038, "step": 271770 }, { "epoch": 23.54, "learning_rate": 1.6176130449867002e-05, "loss": 0.5264, "step": 271780 }, { "epoch": 23.54, "learning_rate": 1.617396206776917e-05, "loss": 0.7001, "step": 271790 }, { "epoch": 23.54, "learning_rate": 1.617179368567133e-05, "loss": 0.6605, "step": 271800 }, { "epoch": 23.54, "learning_rate": 1.616962530357349e-05, "loss": 0.5993, "step": 271810 }, { "epoch": 23.54, "learning_rate": 1.6167456921475654e-05, "loss": 0.5794, "step": 271820 }, { "epoch": 23.54, "learning_rate": 1.6165288539377817e-05, "loss": 0.8402, "step": 271830 }, { "epoch": 23.54, "learning_rate": 1.616312015727998e-05, "loss": 0.6404, "step": 271840 }, { "epoch": 23.54, "learning_rate": 1.6160951775182143e-05, "loss": 0.676, "step": 271850 }, { "epoch": 23.54, "learning_rate": 1.6158783393084306e-05, "loss": 0.6875, "step": 271860 }, { "epoch": 23.55, "learning_rate": 1.615661501098647e-05, "loss": 0.6357, "step": 271870 }, { "epoch": 23.55, "learning_rate": 1.615444662888863e-05, "loss": 0.7218, "step": 271880 }, { "epoch": 23.55, "learning_rate": 1.6152278246790795e-05, "loss": 0.6989, "step": 271890 }, { "epoch": 23.55, "learning_rate": 1.6150109864692955e-05, "loss": 0.5782, "step": 271900 }, { "epoch": 23.55, "learning_rate": 1.6147941482595118e-05, "loss": 0.523, "step": 271910 }, { "epoch": 23.55, "learning_rate": 1.614577310049728e-05, "loss": 0.6731, "step": 271920 }, { "epoch": 23.55, "learning_rate": 1.6143604718399444e-05, "loss": 0.5324, "step": 271930 }, { "epoch": 23.55, "learning_rate": 1.6141436336301607e-05, "loss": 0.5914, "step": 271940 }, { "epoch": 23.55, "learning_rate": 1.613926795420377e-05, "loss": 0.6949, "step": 271950 }, { "epoch": 23.55, "learning_rate": 1.613709957210593e-05, "loss": 0.5001, "step": 271960 }, { "epoch": 23.55, "learning_rate": 1.6134931190008093e-05, "loss": 0.7598, "step": 271970 }, { "epoch": 23.56, "learning_rate": 1.6132762807910256e-05, "loss": 0.6362, "step": 271980 }, { "epoch": 23.56, "learning_rate": 1.613059442581242e-05, "loss": 0.7505, "step": 271990 }, { "epoch": 23.56, "learning_rate": 1.6128426043714582e-05, "loss": 0.5716, "step": 272000 }, { "epoch": 23.56, "learning_rate": 1.6126257661616745e-05, "loss": 0.6022, "step": 272010 }, { "epoch": 23.56, "learning_rate": 1.6124089279518908e-05, "loss": 0.7244, "step": 272020 }, { "epoch": 23.56, "learning_rate": 1.6121920897421067e-05, "loss": 0.6966, "step": 272030 }, { "epoch": 23.56, "learning_rate": 1.6119752515323234e-05, "loss": 0.504, "step": 272040 }, { "epoch": 23.56, "learning_rate": 1.6117584133225393e-05, "loss": 0.6581, "step": 272050 }, { "epoch": 23.56, "learning_rate": 1.6115415751127556e-05, "loss": 0.626, "step": 272060 }, { "epoch": 23.56, "learning_rate": 1.611324736902972e-05, "loss": 0.7017, "step": 272070 }, { "epoch": 23.56, "learning_rate": 1.6111078986931882e-05, "loss": 0.6109, "step": 272080 }, { "epoch": 23.56, "learning_rate": 1.6108910604834045e-05, "loss": 0.7058, "step": 272090 }, { "epoch": 23.57, "learning_rate": 1.610674222273621e-05, "loss": 0.7711, "step": 272100 }, { "epoch": 23.57, "learning_rate": 1.610457384063837e-05, "loss": 0.7313, "step": 272110 }, { "epoch": 23.57, "learning_rate": 1.6102405458540534e-05, "loss": 0.6857, "step": 272120 }, { "epoch": 23.57, "learning_rate": 1.6100237076442694e-05, "loss": 0.6218, "step": 272130 }, { "epoch": 23.57, "learning_rate": 1.609806869434486e-05, "loss": 0.557, "step": 272140 }, { "epoch": 23.57, "learning_rate": 1.609590031224702e-05, "loss": 0.6479, "step": 272150 }, { "epoch": 23.57, "learning_rate": 1.6093731930149183e-05, "loss": 0.7108, "step": 272160 }, { "epoch": 23.57, "learning_rate": 1.6091563548051346e-05, "loss": 0.7199, "step": 272170 }, { "epoch": 23.57, "learning_rate": 1.608939516595351e-05, "loss": 0.6189, "step": 272180 }, { "epoch": 23.57, "learning_rate": 1.6087226783855672e-05, "loss": 0.6844, "step": 272190 }, { "epoch": 23.57, "learning_rate": 1.6085058401757835e-05, "loss": 0.6577, "step": 272200 }, { "epoch": 23.58, "learning_rate": 1.6082890019659995e-05, "loss": 0.6361, "step": 272210 }, { "epoch": 23.58, "learning_rate": 1.608072163756216e-05, "loss": 0.7381, "step": 272220 }, { "epoch": 23.58, "learning_rate": 1.607855325546432e-05, "loss": 0.6246, "step": 272230 }, { "epoch": 23.58, "learning_rate": 1.6076384873366484e-05, "loss": 0.4953, "step": 272240 }, { "epoch": 23.58, "learning_rate": 1.6074216491268647e-05, "loss": 0.5988, "step": 272250 }, { "epoch": 23.58, "learning_rate": 1.607204810917081e-05, "loss": 0.6508, "step": 272260 }, { "epoch": 23.58, "learning_rate": 1.6069879727072973e-05, "loss": 0.5801, "step": 272270 }, { "epoch": 23.58, "learning_rate": 1.6067711344975132e-05, "loss": 0.7299, "step": 272280 }, { "epoch": 23.58, "learning_rate": 1.60655429628773e-05, "loss": 0.6477, "step": 272290 }, { "epoch": 23.58, "learning_rate": 1.6063374580779458e-05, "loss": 0.6752, "step": 272300 }, { "epoch": 23.58, "learning_rate": 1.606120619868162e-05, "loss": 0.6468, "step": 272310 }, { "epoch": 23.58, "learning_rate": 1.6059037816583784e-05, "loss": 0.6223, "step": 272320 }, { "epoch": 23.59, "learning_rate": 1.6056869434485947e-05, "loss": 0.7396, "step": 272330 }, { "epoch": 23.59, "learning_rate": 1.605470105238811e-05, "loss": 0.7208, "step": 272340 }, { "epoch": 23.59, "learning_rate": 1.6052532670290273e-05, "loss": 0.7065, "step": 272350 }, { "epoch": 23.59, "learning_rate": 1.6050364288192436e-05, "loss": 0.6517, "step": 272360 }, { "epoch": 23.59, "learning_rate": 1.60481959060946e-05, "loss": 0.6961, "step": 272370 }, { "epoch": 23.59, "learning_rate": 1.604602752399676e-05, "loss": 0.665, "step": 272380 }, { "epoch": 23.59, "learning_rate": 1.6043859141898925e-05, "loss": 0.5401, "step": 272390 }, { "epoch": 23.59, "learning_rate": 1.6041690759801085e-05, "loss": 0.6194, "step": 272400 }, { "epoch": 23.59, "learning_rate": 1.6039522377703248e-05, "loss": 0.6212, "step": 272410 }, { "epoch": 23.59, "learning_rate": 1.603735399560541e-05, "loss": 0.721, "step": 272420 }, { "epoch": 23.59, "learning_rate": 1.6035185613507574e-05, "loss": 0.7376, "step": 272430 }, { "epoch": 23.6, "learning_rate": 1.6033017231409737e-05, "loss": 0.6576, "step": 272440 }, { "epoch": 23.6, "learning_rate": 1.60308488493119e-05, "loss": 0.784, "step": 272450 }, { "epoch": 23.6, "learning_rate": 1.602868046721406e-05, "loss": 0.581, "step": 272460 }, { "epoch": 23.6, "learning_rate": 1.6026512085116226e-05, "loss": 0.5831, "step": 272470 }, { "epoch": 23.6, "learning_rate": 1.6024343703018386e-05, "loss": 0.627, "step": 272480 }, { "epoch": 23.6, "learning_rate": 1.602217532092055e-05, "loss": 0.6104, "step": 272490 }, { "epoch": 23.6, "learning_rate": 1.602000693882271e-05, "loss": 0.6545, "step": 272500 }, { "epoch": 23.6, "learning_rate": 1.6017838556724875e-05, "loss": 0.7218, "step": 272510 }, { "epoch": 23.6, "learning_rate": 1.6015670174627038e-05, "loss": 0.5589, "step": 272520 }, { "epoch": 23.6, "learning_rate": 1.6013501792529197e-05, "loss": 0.7172, "step": 272530 }, { "epoch": 23.6, "learning_rate": 1.6011333410431364e-05, "loss": 0.5797, "step": 272540 }, { "epoch": 23.6, "learning_rate": 1.6009165028333523e-05, "loss": 0.7012, "step": 272550 }, { "epoch": 23.61, "learning_rate": 1.6006996646235686e-05, "loss": 0.6569, "step": 272560 }, { "epoch": 23.61, "learning_rate": 1.600482826413785e-05, "loss": 0.704, "step": 272570 }, { "epoch": 23.61, "learning_rate": 1.6002659882040012e-05, "loss": 0.5773, "step": 272580 }, { "epoch": 23.61, "learning_rate": 1.6000491499942175e-05, "loss": 0.6033, "step": 272590 }, { "epoch": 23.61, "learning_rate": 1.5998323117844338e-05, "loss": 0.6286, "step": 272600 }, { "epoch": 23.61, "learning_rate": 1.59961547357465e-05, "loss": 0.7149, "step": 272610 }, { "epoch": 23.61, "learning_rate": 1.5993986353648664e-05, "loss": 0.746, "step": 272620 }, { "epoch": 23.61, "learning_rate": 1.5991817971550824e-05, "loss": 0.7153, "step": 272630 }, { "epoch": 23.61, "learning_rate": 1.598964958945299e-05, "loss": 0.6427, "step": 272640 }, { "epoch": 23.61, "learning_rate": 1.598748120735515e-05, "loss": 0.4261, "step": 272650 }, { "epoch": 23.61, "learning_rate": 1.5985312825257313e-05, "loss": 0.8618, "step": 272660 }, { "epoch": 23.61, "learning_rate": 1.5983144443159476e-05, "loss": 0.6159, "step": 272670 }, { "epoch": 23.62, "learning_rate": 1.598097606106164e-05, "loss": 0.7362, "step": 272680 }, { "epoch": 23.62, "learning_rate": 1.5978807678963802e-05, "loss": 0.6914, "step": 272690 }, { "epoch": 23.62, "learning_rate": 1.5976639296865965e-05, "loss": 0.8389, "step": 272700 }, { "epoch": 23.62, "learning_rate": 1.5974470914768124e-05, "loss": 0.7083, "step": 272710 }, { "epoch": 23.62, "learning_rate": 1.597230253267029e-05, "loss": 0.6666, "step": 272720 }, { "epoch": 23.62, "learning_rate": 1.597013415057245e-05, "loss": 0.5829, "step": 272730 }, { "epoch": 23.62, "learning_rate": 1.5967965768474613e-05, "loss": 0.5401, "step": 272740 }, { "epoch": 23.62, "learning_rate": 1.5965797386376776e-05, "loss": 0.6839, "step": 272750 }, { "epoch": 23.62, "learning_rate": 1.596362900427894e-05, "loss": 0.6414, "step": 272760 }, { "epoch": 23.62, "learning_rate": 1.5961460622181102e-05, "loss": 0.7277, "step": 272770 }, { "epoch": 23.62, "learning_rate": 1.5959292240083262e-05, "loss": 0.705, "step": 272780 }, { "epoch": 23.63, "learning_rate": 1.595712385798543e-05, "loss": 0.6962, "step": 272790 }, { "epoch": 23.63, "learning_rate": 1.5954955475887588e-05, "loss": 0.5758, "step": 272800 }, { "epoch": 23.63, "learning_rate": 1.595278709378975e-05, "loss": 0.8514, "step": 272810 }, { "epoch": 23.63, "learning_rate": 1.5950618711691914e-05, "loss": 0.659, "step": 272820 }, { "epoch": 23.63, "learning_rate": 1.5948450329594077e-05, "loss": 0.6598, "step": 272830 }, { "epoch": 23.63, "learning_rate": 1.594628194749624e-05, "loss": 0.6517, "step": 272840 }, { "epoch": 23.63, "learning_rate": 1.5944113565398403e-05, "loss": 0.6789, "step": 272850 }, { "epoch": 23.63, "learning_rate": 1.5941945183300566e-05, "loss": 0.7561, "step": 272860 }, { "epoch": 23.63, "learning_rate": 1.593977680120273e-05, "loss": 0.5644, "step": 272870 }, { "epoch": 23.63, "learning_rate": 1.593760841910489e-05, "loss": 0.6134, "step": 272880 }, { "epoch": 23.63, "learning_rate": 1.5935440037007055e-05, "loss": 0.5736, "step": 272890 }, { "epoch": 23.63, "learning_rate": 1.5933271654909215e-05, "loss": 0.7821, "step": 272900 }, { "epoch": 23.64, "learning_rate": 1.5931103272811378e-05, "loss": 0.7627, "step": 272910 }, { "epoch": 23.64, "learning_rate": 1.592893489071354e-05, "loss": 0.7009, "step": 272920 }, { "epoch": 23.64, "learning_rate": 1.5926766508615704e-05, "loss": 0.7065, "step": 272930 }, { "epoch": 23.64, "learning_rate": 1.5924598126517867e-05, "loss": 0.585, "step": 272940 }, { "epoch": 23.64, "learning_rate": 1.592242974442003e-05, "loss": 0.658, "step": 272950 }, { "epoch": 23.64, "learning_rate": 1.5920261362322193e-05, "loss": 0.6186, "step": 272960 }, { "epoch": 23.64, "learning_rate": 1.5918092980224356e-05, "loss": 0.6782, "step": 272970 }, { "epoch": 23.64, "learning_rate": 1.5915924598126515e-05, "loss": 0.7673, "step": 272980 }, { "epoch": 23.64, "learning_rate": 1.591375621602868e-05, "loss": 0.5777, "step": 272990 }, { "epoch": 23.64, "learning_rate": 1.591158783393084e-05, "loss": 0.6469, "step": 273000 }, { "epoch": 23.64, "learning_rate": 1.5909419451833004e-05, "loss": 0.7054, "step": 273010 }, { "epoch": 23.65, "learning_rate": 1.5907251069735167e-05, "loss": 0.6209, "step": 273020 }, { "epoch": 23.65, "learning_rate": 1.5905082687637327e-05, "loss": 0.575, "step": 273030 }, { "epoch": 23.65, "learning_rate": 1.5902914305539493e-05, "loss": 0.6012, "step": 273040 }, { "epoch": 23.65, "learning_rate": 1.5900745923441653e-05, "loss": 0.6742, "step": 273050 }, { "epoch": 23.65, "learning_rate": 1.5898577541343816e-05, "loss": 0.648, "step": 273060 }, { "epoch": 23.65, "learning_rate": 1.589640915924598e-05, "loss": 0.7001, "step": 273070 }, { "epoch": 23.65, "learning_rate": 1.5894240777148142e-05, "loss": 0.6252, "step": 273080 }, { "epoch": 23.65, "learning_rate": 1.5892072395050305e-05, "loss": 0.6649, "step": 273090 }, { "epoch": 23.65, "learning_rate": 1.5889904012952468e-05, "loss": 0.6767, "step": 273100 }, { "epoch": 23.65, "learning_rate": 1.588773563085463e-05, "loss": 0.6358, "step": 273110 }, { "epoch": 23.65, "learning_rate": 1.5885567248756794e-05, "loss": 0.5339, "step": 273120 }, { "epoch": 23.65, "learning_rate": 1.5883398866658954e-05, "loss": 0.6287, "step": 273130 }, { "epoch": 23.66, "learning_rate": 1.588123048456112e-05, "loss": 0.6827, "step": 273140 }, { "epoch": 23.66, "learning_rate": 1.587906210246328e-05, "loss": 0.6646, "step": 273150 }, { "epoch": 23.66, "learning_rate": 1.5876893720365443e-05, "loss": 0.7136, "step": 273160 }, { "epoch": 23.66, "learning_rate": 1.5874725338267606e-05, "loss": 0.7094, "step": 273170 }, { "epoch": 23.66, "learning_rate": 1.587255695616977e-05, "loss": 0.6368, "step": 273180 }, { "epoch": 23.66, "learning_rate": 1.5870388574071932e-05, "loss": 0.5706, "step": 273190 }, { "epoch": 23.66, "learning_rate": 1.5868220191974095e-05, "loss": 0.6476, "step": 273200 }, { "epoch": 23.66, "learning_rate": 1.5866051809876258e-05, "loss": 0.6004, "step": 273210 }, { "epoch": 23.66, "learning_rate": 1.586388342777842e-05, "loss": 0.6263, "step": 273220 }, { "epoch": 23.66, "learning_rate": 1.586171504568058e-05, "loss": 0.6733, "step": 273230 }, { "epoch": 23.66, "learning_rate": 1.5859546663582747e-05, "loss": 0.6557, "step": 273240 }, { "epoch": 23.67, "learning_rate": 1.5857378281484906e-05, "loss": 0.7779, "step": 273250 }, { "epoch": 23.67, "learning_rate": 1.585520989938707e-05, "loss": 0.714, "step": 273260 }, { "epoch": 23.67, "learning_rate": 1.5853041517289232e-05, "loss": 0.7276, "step": 273270 }, { "epoch": 23.67, "learning_rate": 1.5850873135191392e-05, "loss": 0.7297, "step": 273280 }, { "epoch": 23.67, "learning_rate": 1.584870475309356e-05, "loss": 0.6267, "step": 273290 }, { "epoch": 23.67, "learning_rate": 1.5846536370995718e-05, "loss": 0.6737, "step": 273300 }, { "epoch": 23.67, "learning_rate": 1.584436798889788e-05, "loss": 0.7362, "step": 273310 }, { "epoch": 23.67, "learning_rate": 1.5842199606800044e-05, "loss": 0.5368, "step": 273320 }, { "epoch": 23.67, "learning_rate": 1.5840031224702207e-05, "loss": 0.5744, "step": 273330 }, { "epoch": 23.67, "learning_rate": 1.583786284260437e-05, "loss": 0.5767, "step": 273340 }, { "epoch": 23.67, "learning_rate": 1.5835694460506533e-05, "loss": 0.5698, "step": 273350 }, { "epoch": 23.67, "learning_rate": 1.5833526078408696e-05, "loss": 0.5755, "step": 273360 }, { "epoch": 23.68, "learning_rate": 1.583135769631086e-05, "loss": 0.7833, "step": 273370 }, { "epoch": 23.68, "learning_rate": 1.582918931421302e-05, "loss": 0.5957, "step": 273380 }, { "epoch": 23.68, "learning_rate": 1.5827020932115185e-05, "loss": 0.5884, "step": 273390 }, { "epoch": 23.68, "learning_rate": 1.5824852550017345e-05, "loss": 0.5942, "step": 273400 }, { "epoch": 23.68, "learning_rate": 1.5822684167919508e-05, "loss": 0.7308, "step": 273410 }, { "epoch": 23.68, "learning_rate": 1.582051578582167e-05, "loss": 0.6647, "step": 273420 }, { "epoch": 23.68, "learning_rate": 1.5818347403723834e-05, "loss": 0.594, "step": 273430 }, { "epoch": 23.68, "learning_rate": 1.5816179021625997e-05, "loss": 0.5584, "step": 273440 }, { "epoch": 23.68, "learning_rate": 1.581401063952816e-05, "loss": 0.6338, "step": 273450 }, { "epoch": 23.68, "learning_rate": 1.5811842257430323e-05, "loss": 0.6457, "step": 273460 }, { "epoch": 23.68, "learning_rate": 1.5809673875332486e-05, "loss": 0.6773, "step": 273470 }, { "epoch": 23.69, "learning_rate": 1.5807505493234645e-05, "loss": 0.5514, "step": 273480 }, { "epoch": 23.69, "learning_rate": 1.580533711113681e-05, "loss": 0.7538, "step": 273490 }, { "epoch": 23.69, "learning_rate": 1.580316872903897e-05, "loss": 0.555, "step": 273500 }, { "epoch": 23.69, "learning_rate": 1.5801000346941134e-05, "loss": 0.7892, "step": 273510 }, { "epoch": 23.69, "learning_rate": 1.5798831964843297e-05, "loss": 0.6862, "step": 273520 }, { "epoch": 23.69, "learning_rate": 1.5796663582745457e-05, "loss": 0.6701, "step": 273530 }, { "epoch": 23.69, "learning_rate": 1.5794495200647623e-05, "loss": 0.5782, "step": 273540 }, { "epoch": 23.69, "learning_rate": 1.5792326818549783e-05, "loss": 0.6397, "step": 273550 }, { "epoch": 23.69, "learning_rate": 1.5790158436451946e-05, "loss": 0.6161, "step": 273560 }, { "epoch": 23.69, "learning_rate": 1.578799005435411e-05, "loss": 0.5721, "step": 273570 }, { "epoch": 23.69, "learning_rate": 1.5785821672256272e-05, "loss": 0.6549, "step": 273580 }, { "epoch": 23.69, "learning_rate": 1.5783653290158435e-05, "loss": 0.7355, "step": 273590 }, { "epoch": 23.7, "learning_rate": 1.5781484908060598e-05, "loss": 0.512, "step": 273600 }, { "epoch": 23.7, "learning_rate": 1.577931652596276e-05, "loss": 0.5796, "step": 273610 }, { "epoch": 23.7, "learning_rate": 1.5777148143864924e-05, "loss": 0.638, "step": 273620 }, { "epoch": 23.7, "learning_rate": 1.5774979761767084e-05, "loss": 0.6518, "step": 273630 }, { "epoch": 23.7, "learning_rate": 1.577281137966925e-05, "loss": 0.6032, "step": 273640 }, { "epoch": 23.7, "learning_rate": 1.577064299757141e-05, "loss": 0.7563, "step": 273650 }, { "epoch": 23.7, "learning_rate": 1.5768474615473573e-05, "loss": 0.7751, "step": 273660 }, { "epoch": 23.7, "learning_rate": 1.5766306233375736e-05, "loss": 0.6291, "step": 273670 }, { "epoch": 23.7, "learning_rate": 1.57641378512779e-05, "loss": 0.7101, "step": 273680 }, { "epoch": 23.7, "learning_rate": 1.576196946918006e-05, "loss": 0.6937, "step": 273690 }, { "epoch": 23.7, "learning_rate": 1.5759801087082225e-05, "loss": 0.6803, "step": 273700 }, { "epoch": 23.71, "learning_rate": 1.5757632704984388e-05, "loss": 0.6931, "step": 273710 }, { "epoch": 23.71, "learning_rate": 1.575546432288655e-05, "loss": 0.5096, "step": 273720 }, { "epoch": 23.71, "learning_rate": 1.575329594078871e-05, "loss": 0.81, "step": 273730 }, { "epoch": 23.71, "learning_rate": 1.5751127558690877e-05, "loss": 0.7242, "step": 273740 }, { "epoch": 23.71, "learning_rate": 1.5748959176593036e-05, "loss": 0.5681, "step": 273750 }, { "epoch": 23.71, "learning_rate": 1.57467907944952e-05, "loss": 0.5638, "step": 273760 }, { "epoch": 23.71, "learning_rate": 1.5744622412397362e-05, "loss": 0.635, "step": 273770 }, { "epoch": 23.71, "learning_rate": 1.5742454030299522e-05, "loss": 0.6161, "step": 273780 }, { "epoch": 23.71, "learning_rate": 1.5740285648201688e-05, "loss": 0.6288, "step": 273790 }, { "epoch": 23.71, "learning_rate": 1.5738117266103848e-05, "loss": 0.6574, "step": 273800 }, { "epoch": 23.71, "learning_rate": 1.573594888400601e-05, "loss": 0.6251, "step": 273810 }, { "epoch": 23.71, "learning_rate": 1.5733780501908174e-05, "loss": 0.7008, "step": 273820 }, { "epoch": 23.72, "learning_rate": 1.5731612119810337e-05, "loss": 0.7816, "step": 273830 }, { "epoch": 23.72, "learning_rate": 1.57294437377125e-05, "loss": 0.499, "step": 273840 }, { "epoch": 23.72, "learning_rate": 1.5727275355614663e-05, "loss": 0.637, "step": 273850 }, { "epoch": 23.72, "learning_rate": 1.5725106973516826e-05, "loss": 0.4975, "step": 273860 }, { "epoch": 23.72, "learning_rate": 1.572293859141899e-05, "loss": 0.6276, "step": 273870 }, { "epoch": 23.72, "learning_rate": 1.572077020932115e-05, "loss": 0.8881, "step": 273880 }, { "epoch": 23.72, "learning_rate": 1.5718601827223315e-05, "loss": 0.6316, "step": 273890 }, { "epoch": 23.72, "learning_rate": 1.5716433445125474e-05, "loss": 0.7169, "step": 273900 }, { "epoch": 23.72, "learning_rate": 1.5714265063027637e-05, "loss": 0.5346, "step": 273910 }, { "epoch": 23.72, "learning_rate": 1.57120966809298e-05, "loss": 0.6579, "step": 273920 }, { "epoch": 23.72, "learning_rate": 1.5709928298831963e-05, "loss": 0.5468, "step": 273930 }, { "epoch": 23.72, "learning_rate": 1.5707759916734126e-05, "loss": 0.8308, "step": 273940 }, { "epoch": 23.73, "learning_rate": 1.570559153463629e-05, "loss": 0.5643, "step": 273950 }, { "epoch": 23.73, "learning_rate": 1.5703423152538452e-05, "loss": 0.5699, "step": 273960 }, { "epoch": 23.73, "learning_rate": 1.5701254770440615e-05, "loss": 0.693, "step": 273970 }, { "epoch": 23.73, "learning_rate": 1.5699086388342775e-05, "loss": 0.7355, "step": 273980 }, { "epoch": 23.73, "learning_rate": 1.569691800624494e-05, "loss": 0.544, "step": 273990 }, { "epoch": 23.73, "learning_rate": 1.56947496241471e-05, "loss": 0.6236, "step": 274000 }, { "epoch": 23.73, "learning_rate": 1.5692581242049264e-05, "loss": 0.5431, "step": 274010 }, { "epoch": 23.73, "learning_rate": 1.5690412859951427e-05, "loss": 0.6461, "step": 274020 }, { "epoch": 23.73, "learning_rate": 1.568824447785359e-05, "loss": 0.6244, "step": 274030 }, { "epoch": 23.73, "learning_rate": 1.5686076095755753e-05, "loss": 0.7121, "step": 274040 }, { "epoch": 23.73, "learning_rate": 1.5683907713657913e-05, "loss": 0.7103, "step": 274050 }, { "epoch": 23.74, "learning_rate": 1.5681739331560076e-05, "loss": 0.7605, "step": 274060 }, { "epoch": 23.74, "learning_rate": 1.567957094946224e-05, "loss": 0.6248, "step": 274070 }, { "epoch": 23.74, "learning_rate": 1.5677402567364402e-05, "loss": 0.6147, "step": 274080 }, { "epoch": 23.74, "learning_rate": 1.5675234185266565e-05, "loss": 0.6689, "step": 274090 }, { "epoch": 23.74, "learning_rate": 1.5673065803168728e-05, "loss": 0.7565, "step": 274100 }, { "epoch": 23.74, "learning_rate": 1.567089742107089e-05, "loss": 0.6706, "step": 274110 }, { "epoch": 23.74, "learning_rate": 1.5668729038973054e-05, "loss": 0.5589, "step": 274120 }, { "epoch": 23.74, "learning_rate": 1.5666560656875213e-05, "loss": 0.6292, "step": 274130 }, { "epoch": 23.74, "learning_rate": 1.566439227477738e-05, "loss": 0.5214, "step": 274140 }, { "epoch": 23.74, "learning_rate": 1.566222389267954e-05, "loss": 0.4724, "step": 274150 }, { "epoch": 23.74, "learning_rate": 1.5660055510581702e-05, "loss": 0.7211, "step": 274160 }, { "epoch": 23.74, "learning_rate": 1.5657887128483865e-05, "loss": 0.6178, "step": 274170 }, { "epoch": 23.75, "learning_rate": 1.565571874638603e-05, "loss": 0.6476, "step": 274180 }, { "epoch": 23.75, "learning_rate": 1.565355036428819e-05, "loss": 0.6554, "step": 274190 }, { "epoch": 23.75, "learning_rate": 1.5651381982190354e-05, "loss": 0.599, "step": 274200 }, { "epoch": 23.75, "learning_rate": 1.5649213600092517e-05, "loss": 0.6748, "step": 274210 }, { "epoch": 23.75, "learning_rate": 1.564704521799468e-05, "loss": 0.7481, "step": 274220 }, { "epoch": 23.75, "learning_rate": 1.564487683589684e-05, "loss": 0.7398, "step": 274230 }, { "epoch": 23.75, "learning_rate": 1.5642708453799006e-05, "loss": 0.6109, "step": 274240 }, { "epoch": 23.75, "learning_rate": 1.5640540071701166e-05, "loss": 0.6557, "step": 274250 }, { "epoch": 23.75, "learning_rate": 1.563837168960333e-05, "loss": 0.6676, "step": 274260 }, { "epoch": 23.75, "learning_rate": 1.5636203307505492e-05, "loss": 0.5912, "step": 274270 }, { "epoch": 23.75, "learning_rate": 1.5634034925407655e-05, "loss": 0.5347, "step": 274280 }, { "epoch": 23.76, "learning_rate": 1.5631866543309818e-05, "loss": 0.7341, "step": 274290 }, { "epoch": 23.76, "learning_rate": 1.5629698161211978e-05, "loss": 0.6579, "step": 274300 }, { "epoch": 23.76, "learning_rate": 1.5627529779114144e-05, "loss": 0.6511, "step": 274310 }, { "epoch": 23.76, "learning_rate": 1.5625361397016304e-05, "loss": 0.6094, "step": 274320 }, { "epoch": 23.76, "learning_rate": 1.5623193014918467e-05, "loss": 0.5735, "step": 274330 }, { "epoch": 23.76, "learning_rate": 1.562102463282063e-05, "loss": 0.6544, "step": 274340 }, { "epoch": 23.76, "learning_rate": 1.5618856250722793e-05, "loss": 0.6483, "step": 274350 }, { "epoch": 23.76, "learning_rate": 1.5616687868624956e-05, "loss": 0.5287, "step": 274360 }, { "epoch": 23.76, "learning_rate": 1.561451948652712e-05, "loss": 0.7301, "step": 274370 }, { "epoch": 23.76, "learning_rate": 1.5612351104429278e-05, "loss": 0.6128, "step": 274380 }, { "epoch": 23.76, "learning_rate": 1.5610182722331445e-05, "loss": 0.6104, "step": 274390 }, { "epoch": 23.76, "learning_rate": 1.5608014340233604e-05, "loss": 0.7831, "step": 274400 }, { "epoch": 23.77, "learning_rate": 1.5605845958135767e-05, "loss": 0.7087, "step": 274410 }, { "epoch": 23.77, "learning_rate": 1.560367757603793e-05, "loss": 0.6658, "step": 274420 }, { "epoch": 23.77, "learning_rate": 1.5601509193940093e-05, "loss": 0.7483, "step": 274430 }, { "epoch": 23.77, "learning_rate": 1.5599340811842256e-05, "loss": 0.7351, "step": 274440 }, { "epoch": 23.77, "learning_rate": 1.559717242974442e-05, "loss": 0.6704, "step": 274450 }, { "epoch": 23.77, "learning_rate": 1.5595004047646582e-05, "loss": 0.7374, "step": 274460 }, { "epoch": 23.77, "learning_rate": 1.5592835665548745e-05, "loss": 0.7283, "step": 274470 }, { "epoch": 23.77, "learning_rate": 1.5590667283450905e-05, "loss": 0.5608, "step": 274480 }, { "epoch": 23.77, "learning_rate": 1.558849890135307e-05, "loss": 0.5776, "step": 274490 }, { "epoch": 23.77, "learning_rate": 1.558633051925523e-05, "loss": 0.7125, "step": 274500 }, { "epoch": 23.77, "learning_rate": 1.5584162137157394e-05, "loss": 0.7161, "step": 274510 }, { "epoch": 23.78, "learning_rate": 1.5581993755059557e-05, "loss": 0.6474, "step": 274520 }, { "epoch": 23.78, "learning_rate": 1.557982537296172e-05, "loss": 0.5941, "step": 274530 }, { "epoch": 23.78, "learning_rate": 1.5577656990863883e-05, "loss": 0.7071, "step": 274540 }, { "epoch": 23.78, "learning_rate": 1.5575488608766043e-05, "loss": 0.557, "step": 274550 }, { "epoch": 23.78, "learning_rate": 1.557332022666821e-05, "loss": 0.6423, "step": 274560 }, { "epoch": 23.78, "learning_rate": 1.557115184457037e-05, "loss": 0.707, "step": 274570 }, { "epoch": 23.78, "learning_rate": 1.556898346247253e-05, "loss": 0.652, "step": 274580 }, { "epoch": 23.78, "learning_rate": 1.5566815080374695e-05, "loss": 0.4824, "step": 274590 }, { "epoch": 23.78, "learning_rate": 1.5564646698276858e-05, "loss": 0.6196, "step": 274600 }, { "epoch": 23.78, "learning_rate": 1.556247831617902e-05, "loss": 0.54, "step": 274610 }, { "epoch": 23.78, "learning_rate": 1.5560309934081184e-05, "loss": 0.7777, "step": 274620 }, { "epoch": 23.78, "learning_rate": 1.5558141551983343e-05, "loss": 0.5259, "step": 274630 }, { "epoch": 23.79, "learning_rate": 1.555597316988551e-05, "loss": 0.6859, "step": 274640 }, { "epoch": 23.79, "learning_rate": 1.555380478778767e-05, "loss": 0.6077, "step": 274650 }, { "epoch": 23.79, "learning_rate": 1.5551636405689832e-05, "loss": 0.6293, "step": 274660 }, { "epoch": 23.79, "learning_rate": 1.5549468023591995e-05, "loss": 0.8311, "step": 274670 }, { "epoch": 23.79, "learning_rate": 1.5547299641494158e-05, "loss": 0.6115, "step": 274680 }, { "epoch": 23.79, "learning_rate": 1.554513125939632e-05, "loss": 0.7504, "step": 274690 }, { "epoch": 23.79, "learning_rate": 1.5542962877298484e-05, "loss": 0.7467, "step": 274700 }, { "epoch": 23.79, "learning_rate": 1.5540794495200647e-05, "loss": 0.7231, "step": 274710 }, { "epoch": 23.79, "learning_rate": 1.553862611310281e-05, "loss": 0.5931, "step": 274720 }, { "epoch": 23.79, "learning_rate": 1.553645773100497e-05, "loss": 0.8534, "step": 274730 }, { "epoch": 23.79, "learning_rate": 1.5534289348907136e-05, "loss": 0.5834, "step": 274740 }, { "epoch": 23.8, "learning_rate": 1.5532120966809296e-05, "loss": 0.6956, "step": 274750 }, { "epoch": 23.8, "learning_rate": 1.552995258471146e-05, "loss": 0.822, "step": 274760 }, { "epoch": 23.8, "learning_rate": 1.5527784202613622e-05, "loss": 0.5583, "step": 274770 }, { "epoch": 23.8, "learning_rate": 1.5525615820515785e-05, "loss": 0.6846, "step": 274780 }, { "epoch": 23.8, "learning_rate": 1.5523447438417948e-05, "loss": 0.7634, "step": 274790 }, { "epoch": 23.8, "learning_rate": 1.5521279056320108e-05, "loss": 0.5176, "step": 274800 }, { "epoch": 23.8, "learning_rate": 1.5519110674222274e-05, "loss": 0.5425, "step": 274810 }, { "epoch": 23.8, "learning_rate": 1.5516942292124434e-05, "loss": 0.6216, "step": 274820 }, { "epoch": 23.8, "learning_rate": 1.5514773910026597e-05, "loss": 0.7422, "step": 274830 }, { "epoch": 23.8, "learning_rate": 1.551260552792876e-05, "loss": 0.7872, "step": 274840 }, { "epoch": 23.8, "learning_rate": 1.5510437145830923e-05, "loss": 0.7702, "step": 274850 }, { "epoch": 23.8, "learning_rate": 1.5508268763733086e-05, "loss": 0.7475, "step": 274860 }, { "epoch": 23.81, "learning_rate": 1.550610038163525e-05, "loss": 0.5304, "step": 274870 }, { "epoch": 23.81, "learning_rate": 1.5503931999537408e-05, "loss": 0.7015, "step": 274880 }, { "epoch": 23.81, "learning_rate": 1.5501763617439575e-05, "loss": 0.6077, "step": 274890 }, { "epoch": 23.81, "learning_rate": 1.5499595235341734e-05, "loss": 0.5161, "step": 274900 }, { "epoch": 23.81, "learning_rate": 1.5497426853243897e-05, "loss": 0.6748, "step": 274910 }, { "epoch": 23.81, "learning_rate": 1.549525847114606e-05, "loss": 0.7105, "step": 274920 }, { "epoch": 23.81, "learning_rate": 1.5493090089048223e-05, "loss": 0.692, "step": 274930 }, { "epoch": 23.81, "learning_rate": 1.5490921706950386e-05, "loss": 0.6772, "step": 274940 }, { "epoch": 23.81, "learning_rate": 1.548875332485255e-05, "loss": 0.7445, "step": 274950 }, { "epoch": 23.81, "learning_rate": 1.5486584942754712e-05, "loss": 0.5716, "step": 274960 }, { "epoch": 23.81, "learning_rate": 1.5484416560656875e-05, "loss": 0.4428, "step": 274970 }, { "epoch": 23.82, "learning_rate": 1.5482248178559035e-05, "loss": 0.6721, "step": 274980 }, { "epoch": 23.82, "learning_rate": 1.54800797964612e-05, "loss": 0.7408, "step": 274990 }, { "epoch": 23.82, "learning_rate": 1.547791141436336e-05, "loss": 0.736, "step": 275000 }, { "epoch": 23.82, "learning_rate": 1.5475743032265524e-05, "loss": 0.669, "step": 275010 }, { "epoch": 23.82, "learning_rate": 1.5473574650167687e-05, "loss": 0.7312, "step": 275020 }, { "epoch": 23.82, "learning_rate": 1.547140626806985e-05, "loss": 0.6365, "step": 275030 }, { "epoch": 23.82, "learning_rate": 1.5469237885972013e-05, "loss": 0.5951, "step": 275040 }, { "epoch": 23.82, "learning_rate": 1.5467069503874172e-05, "loss": 0.6858, "step": 275050 }, { "epoch": 23.82, "learning_rate": 1.546490112177634e-05, "loss": 0.6689, "step": 275060 }, { "epoch": 23.82, "learning_rate": 1.54627327396785e-05, "loss": 0.6146, "step": 275070 }, { "epoch": 23.82, "learning_rate": 1.546056435758066e-05, "loss": 0.5073, "step": 275080 }, { "epoch": 23.82, "learning_rate": 1.5458395975482824e-05, "loss": 0.7136, "step": 275090 }, { "epoch": 23.83, "learning_rate": 1.5456227593384987e-05, "loss": 0.6482, "step": 275100 }, { "epoch": 23.83, "learning_rate": 1.545405921128715e-05, "loss": 0.6623, "step": 275110 }, { "epoch": 23.83, "learning_rate": 1.5451890829189313e-05, "loss": 0.5463, "step": 275120 }, { "epoch": 23.83, "learning_rate": 1.5449722447091473e-05, "loss": 0.7434, "step": 275130 }, { "epoch": 23.83, "learning_rate": 1.544755406499364e-05, "loss": 0.8377, "step": 275140 }, { "epoch": 23.83, "learning_rate": 1.54453856828958e-05, "loss": 0.5832, "step": 275150 }, { "epoch": 23.83, "learning_rate": 1.5443217300797962e-05, "loss": 0.7262, "step": 275160 }, { "epoch": 23.83, "learning_rate": 1.5441048918700125e-05, "loss": 0.7509, "step": 275170 }, { "epoch": 23.83, "learning_rate": 1.5438880536602288e-05, "loss": 0.7189, "step": 275180 }, { "epoch": 23.83, "learning_rate": 1.543671215450445e-05, "loss": 0.6932, "step": 275190 }, { "epoch": 23.83, "learning_rate": 1.5434543772406614e-05, "loss": 0.6396, "step": 275200 }, { "epoch": 23.83, "learning_rate": 1.5432375390308777e-05, "loss": 0.6938, "step": 275210 }, { "epoch": 23.84, "learning_rate": 1.543020700821094e-05, "loss": 0.576, "step": 275220 }, { "epoch": 23.84, "learning_rate": 1.54280386261131e-05, "loss": 0.792, "step": 275230 }, { "epoch": 23.84, "learning_rate": 1.5425870244015266e-05, "loss": 0.6958, "step": 275240 }, { "epoch": 23.84, "learning_rate": 1.5423701861917426e-05, "loss": 0.6826, "step": 275250 }, { "epoch": 23.84, "learning_rate": 1.542153347981959e-05, "loss": 0.6909, "step": 275260 }, { "epoch": 23.84, "learning_rate": 1.5419365097721752e-05, "loss": 0.7573, "step": 275270 }, { "epoch": 23.84, "learning_rate": 1.5417196715623915e-05, "loss": 0.6241, "step": 275280 }, { "epoch": 23.84, "learning_rate": 1.5415028333526078e-05, "loss": 0.5915, "step": 275290 }, { "epoch": 23.84, "learning_rate": 1.541285995142824e-05, "loss": 0.7047, "step": 275300 }, { "epoch": 23.84, "learning_rate": 1.5410691569330404e-05, "loss": 0.7711, "step": 275310 }, { "epoch": 23.84, "learning_rate": 1.5408523187232563e-05, "loss": 0.676, "step": 275320 }, { "epoch": 23.85, "learning_rate": 1.5406354805134726e-05, "loss": 0.5385, "step": 275330 }, { "epoch": 23.85, "learning_rate": 1.540418642303689e-05, "loss": 0.6359, "step": 275340 }, { "epoch": 23.85, "learning_rate": 1.5402018040939052e-05, "loss": 0.6155, "step": 275350 }, { "epoch": 23.85, "learning_rate": 1.5399849658841215e-05, "loss": 0.5221, "step": 275360 }, { "epoch": 23.85, "learning_rate": 1.539768127674338e-05, "loss": 0.6586, "step": 275370 }, { "epoch": 23.85, "learning_rate": 1.539551289464554e-05, "loss": 0.6685, "step": 275380 }, { "epoch": 23.85, "learning_rate": 1.5393344512547704e-05, "loss": 0.6735, "step": 275390 }, { "epoch": 23.85, "learning_rate": 1.5391176130449864e-05, "loss": 0.8, "step": 275400 }, { "epoch": 23.85, "learning_rate": 1.5389007748352027e-05, "loss": 0.6307, "step": 275410 }, { "epoch": 23.85, "learning_rate": 1.538683936625419e-05, "loss": 0.6638, "step": 275420 }, { "epoch": 23.85, "learning_rate": 1.5384670984156353e-05, "loss": 0.6782, "step": 275430 }, { "epoch": 23.85, "learning_rate": 1.5382502602058516e-05, "loss": 0.7372, "step": 275440 }, { "epoch": 23.86, "learning_rate": 1.538033421996068e-05, "loss": 0.5862, "step": 275450 }, { "epoch": 23.86, "learning_rate": 1.5378165837862842e-05, "loss": 0.5688, "step": 275460 }, { "epoch": 23.86, "learning_rate": 1.5375997455765005e-05, "loss": 0.5557, "step": 275470 }, { "epoch": 23.86, "learning_rate": 1.5373829073667165e-05, "loss": 0.6956, "step": 275480 }, { "epoch": 23.86, "learning_rate": 1.537166069156933e-05, "loss": 0.7409, "step": 275490 }, { "epoch": 23.86, "learning_rate": 1.536949230947149e-05, "loss": 0.6375, "step": 275500 }, { "epoch": 23.86, "learning_rate": 1.5367323927373654e-05, "loss": 0.8192, "step": 275510 }, { "epoch": 23.86, "learning_rate": 1.5365155545275817e-05, "loss": 0.685, "step": 275520 }, { "epoch": 23.86, "learning_rate": 1.536298716317798e-05, "loss": 0.7368, "step": 275530 }, { "epoch": 23.86, "learning_rate": 1.5360818781080143e-05, "loss": 0.696, "step": 275540 }, { "epoch": 23.86, "learning_rate": 1.5358650398982306e-05, "loss": 0.8315, "step": 275550 }, { "epoch": 23.87, "learning_rate": 1.535648201688447e-05, "loss": 0.6228, "step": 275560 }, { "epoch": 23.87, "learning_rate": 1.5354313634786628e-05, "loss": 0.691, "step": 275570 }, { "epoch": 23.87, "learning_rate": 1.535214525268879e-05, "loss": 0.7112, "step": 275580 }, { "epoch": 23.87, "learning_rate": 1.5349976870590954e-05, "loss": 0.7203, "step": 275590 }, { "epoch": 23.87, "learning_rate": 1.5347808488493117e-05, "loss": 0.7297, "step": 275600 }, { "epoch": 23.87, "learning_rate": 1.534564010639528e-05, "loss": 0.7394, "step": 275610 }, { "epoch": 23.87, "learning_rate": 1.5343471724297443e-05, "loss": 0.6415, "step": 275620 }, { "epoch": 23.87, "learning_rate": 1.5341303342199606e-05, "loss": 0.6989, "step": 275630 }, { "epoch": 23.87, "learning_rate": 1.533913496010177e-05, "loss": 0.5843, "step": 275640 }, { "epoch": 23.87, "learning_rate": 1.533696657800393e-05, "loss": 0.6191, "step": 275650 }, { "epoch": 23.87, "learning_rate": 1.5334798195906095e-05, "loss": 0.6573, "step": 275660 }, { "epoch": 23.87, "learning_rate": 1.5332629813808255e-05, "loss": 0.6779, "step": 275670 }, { "epoch": 23.88, "learning_rate": 1.5330461431710418e-05, "loss": 0.4974, "step": 275680 }, { "epoch": 23.88, "learning_rate": 1.532829304961258e-05, "loss": 0.7214, "step": 275690 }, { "epoch": 23.88, "learning_rate": 1.5326124667514744e-05, "loss": 0.8075, "step": 275700 }, { "epoch": 23.88, "learning_rate": 1.5323956285416907e-05, "loss": 0.7082, "step": 275710 }, { "epoch": 23.88, "learning_rate": 1.532178790331907e-05, "loss": 0.5606, "step": 275720 }, { "epoch": 23.88, "learning_rate": 1.531961952122123e-05, "loss": 0.5992, "step": 275730 }, { "epoch": 23.88, "learning_rate": 1.5317451139123396e-05, "loss": 0.7363, "step": 275740 }, { "epoch": 23.88, "learning_rate": 1.5315282757025556e-05, "loss": 0.7475, "step": 275750 }, { "epoch": 23.88, "learning_rate": 1.531311437492772e-05, "loss": 0.5469, "step": 275760 }, { "epoch": 23.88, "learning_rate": 1.531094599282988e-05, "loss": 0.7615, "step": 275770 }, { "epoch": 23.88, "learning_rate": 1.5308777610732045e-05, "loss": 0.612, "step": 275780 }, { "epoch": 23.89, "learning_rate": 1.5306609228634208e-05, "loss": 0.5454, "step": 275790 }, { "epoch": 23.89, "learning_rate": 1.530444084653637e-05, "loss": 0.6472, "step": 275800 }, { "epoch": 23.89, "learning_rate": 1.5302272464438534e-05, "loss": 0.6398, "step": 275810 }, { "epoch": 23.89, "learning_rate": 1.5300104082340693e-05, "loss": 0.7539, "step": 275820 }, { "epoch": 23.89, "learning_rate": 1.5297935700242856e-05, "loss": 0.6674, "step": 275830 }, { "epoch": 23.89, "learning_rate": 1.529576731814502e-05, "loss": 0.9789, "step": 275840 }, { "epoch": 23.89, "learning_rate": 1.5293598936047182e-05, "loss": 0.5712, "step": 275850 }, { "epoch": 23.89, "learning_rate": 1.5291430553949345e-05, "loss": 0.6215, "step": 275860 }, { "epoch": 23.89, "learning_rate": 1.5289262171851508e-05, "loss": 0.7018, "step": 275870 }, { "epoch": 23.89, "learning_rate": 1.528709378975367e-05, "loss": 0.6569, "step": 275880 }, { "epoch": 23.89, "learning_rate": 1.5284925407655834e-05, "loss": 0.7378, "step": 275890 }, { "epoch": 23.89, "learning_rate": 1.5282757025557994e-05, "loss": 0.5579, "step": 275900 }, { "epoch": 23.9, "learning_rate": 1.528058864346016e-05, "loss": 0.6459, "step": 275910 }, { "epoch": 23.9, "learning_rate": 1.527842026136232e-05, "loss": 0.5928, "step": 275920 }, { "epoch": 23.9, "learning_rate": 1.5276251879264483e-05, "loss": 0.6115, "step": 275930 }, { "epoch": 23.9, "learning_rate": 1.5274083497166646e-05, "loss": 0.7081, "step": 275940 }, { "epoch": 23.9, "learning_rate": 1.527191511506881e-05, "loss": 0.6557, "step": 275950 }, { "epoch": 23.9, "learning_rate": 1.5269746732970972e-05, "loss": 0.5475, "step": 275960 }, { "epoch": 23.9, "learning_rate": 1.5267578350873135e-05, "loss": 0.5639, "step": 275970 }, { "epoch": 23.9, "learning_rate": 1.5265409968775294e-05, "loss": 0.506, "step": 275980 }, { "epoch": 23.9, "learning_rate": 1.526324158667746e-05, "loss": 0.6125, "step": 275990 }, { "epoch": 23.9, "learning_rate": 1.526107320457962e-05, "loss": 0.6231, "step": 276000 }, { "epoch": 23.9, "learning_rate": 1.5258904822481783e-05, "loss": 0.5943, "step": 276010 }, { "epoch": 23.91, "learning_rate": 1.5256736440383947e-05, "loss": 0.6578, "step": 276020 }, { "epoch": 23.91, "learning_rate": 1.525456805828611e-05, "loss": 0.707, "step": 276030 }, { "epoch": 23.91, "learning_rate": 1.5252399676188273e-05, "loss": 0.7181, "step": 276040 }, { "epoch": 23.91, "learning_rate": 1.5250231294090436e-05, "loss": 0.6247, "step": 276050 }, { "epoch": 23.91, "learning_rate": 1.5248062911992597e-05, "loss": 0.659, "step": 276060 }, { "epoch": 23.91, "learning_rate": 1.5245894529894762e-05, "loss": 0.6492, "step": 276070 }, { "epoch": 23.91, "learning_rate": 1.5243726147796923e-05, "loss": 0.6354, "step": 276080 }, { "epoch": 23.91, "learning_rate": 1.5241557765699084e-05, "loss": 0.7045, "step": 276090 }, { "epoch": 23.91, "learning_rate": 1.5239389383601247e-05, "loss": 0.7021, "step": 276100 }, { "epoch": 23.91, "learning_rate": 1.523722100150341e-05, "loss": 0.6812, "step": 276110 }, { "epoch": 23.91, "learning_rate": 1.5235052619405573e-05, "loss": 0.6854, "step": 276120 }, { "epoch": 23.91, "learning_rate": 1.5232884237307734e-05, "loss": 0.6877, "step": 276130 }, { "epoch": 23.92, "learning_rate": 1.52307158552099e-05, "loss": 0.5944, "step": 276140 }, { "epoch": 23.92, "learning_rate": 1.522854747311206e-05, "loss": 0.6936, "step": 276150 }, { "epoch": 23.92, "learning_rate": 1.5226379091014223e-05, "loss": 0.5675, "step": 276160 }, { "epoch": 23.92, "learning_rate": 1.5224210708916385e-05, "loss": 0.6462, "step": 276170 }, { "epoch": 23.92, "learning_rate": 1.522204232681855e-05, "loss": 0.7325, "step": 276180 }, { "epoch": 23.92, "learning_rate": 1.521987394472071e-05, "loss": 0.6176, "step": 276190 }, { "epoch": 23.92, "learning_rate": 1.5217705562622874e-05, "loss": 0.6011, "step": 276200 }, { "epoch": 23.92, "learning_rate": 1.5215537180525035e-05, "loss": 0.5697, "step": 276210 }, { "epoch": 23.92, "learning_rate": 1.52133687984272e-05, "loss": 0.6073, "step": 276220 }, { "epoch": 23.92, "learning_rate": 1.5211200416329361e-05, "loss": 0.7248, "step": 276230 }, { "epoch": 23.92, "learning_rate": 1.5209032034231524e-05, "loss": 0.596, "step": 276240 }, { "epoch": 23.92, "learning_rate": 1.5206863652133687e-05, "loss": 0.6149, "step": 276250 }, { "epoch": 23.93, "learning_rate": 1.520469527003585e-05, "loss": 0.6417, "step": 276260 }, { "epoch": 23.93, "learning_rate": 1.5202526887938011e-05, "loss": 0.6276, "step": 276270 }, { "epoch": 23.93, "learning_rate": 1.5200358505840176e-05, "loss": 0.5201, "step": 276280 }, { "epoch": 23.93, "learning_rate": 1.5198190123742337e-05, "loss": 0.6334, "step": 276290 }, { "epoch": 23.93, "learning_rate": 1.51960217416445e-05, "loss": 0.796, "step": 276300 }, { "epoch": 23.93, "learning_rate": 1.5193853359546662e-05, "loss": 0.7158, "step": 276310 }, { "epoch": 23.93, "learning_rate": 1.5191684977448826e-05, "loss": 0.6204, "step": 276320 }, { "epoch": 23.93, "learning_rate": 1.5189516595350988e-05, "loss": 0.6385, "step": 276330 }, { "epoch": 23.93, "learning_rate": 1.5187348213253149e-05, "loss": 0.6988, "step": 276340 }, { "epoch": 23.93, "learning_rate": 1.5185179831155312e-05, "loss": 0.6481, "step": 276350 }, { "epoch": 23.93, "learning_rate": 1.5183011449057475e-05, "loss": 0.7983, "step": 276360 }, { "epoch": 23.94, "learning_rate": 1.5180843066959638e-05, "loss": 0.7613, "step": 276370 }, { "epoch": 23.94, "learning_rate": 1.51786746848618e-05, "loss": 0.7096, "step": 276380 }, { "epoch": 23.94, "learning_rate": 1.5176506302763964e-05, "loss": 0.63, "step": 276390 }, { "epoch": 23.94, "learning_rate": 1.5174337920666125e-05, "loss": 0.7671, "step": 276400 }, { "epoch": 23.94, "learning_rate": 1.5172169538568288e-05, "loss": 0.791, "step": 276410 }, { "epoch": 23.94, "learning_rate": 1.517000115647045e-05, "loss": 0.5419, "step": 276420 }, { "epoch": 23.94, "learning_rate": 1.5167832774372614e-05, "loss": 0.7333, "step": 276430 }, { "epoch": 23.94, "learning_rate": 1.5165664392274776e-05, "loss": 0.8761, "step": 276440 }, { "epoch": 23.94, "learning_rate": 1.5163496010176939e-05, "loss": 0.732, "step": 276450 }, { "epoch": 23.94, "learning_rate": 1.51613276280791e-05, "loss": 0.6297, "step": 276460 }, { "epoch": 23.94, "learning_rate": 1.5159159245981265e-05, "loss": 0.7425, "step": 276470 }, { "epoch": 23.94, "learning_rate": 1.5156990863883426e-05, "loss": 0.7099, "step": 276480 }, { "epoch": 23.95, "learning_rate": 1.5154822481785589e-05, "loss": 0.6506, "step": 276490 }, { "epoch": 23.95, "learning_rate": 1.5152654099687752e-05, "loss": 0.5842, "step": 276500 }, { "epoch": 23.95, "learning_rate": 1.5150485717589915e-05, "loss": 0.7137, "step": 276510 }, { "epoch": 23.95, "learning_rate": 1.5148317335492076e-05, "loss": 0.6721, "step": 276520 }, { "epoch": 23.95, "learning_rate": 1.5146148953394241e-05, "loss": 0.6186, "step": 276530 }, { "epoch": 23.95, "learning_rate": 1.5143980571296402e-05, "loss": 0.7065, "step": 276540 }, { "epoch": 23.95, "learning_rate": 1.5141812189198565e-05, "loss": 0.779, "step": 276550 }, { "epoch": 23.95, "learning_rate": 1.5139643807100727e-05, "loss": 0.6793, "step": 276560 }, { "epoch": 23.95, "learning_rate": 1.5137475425002891e-05, "loss": 0.6885, "step": 276570 }, { "epoch": 23.95, "learning_rate": 1.5135307042905053e-05, "loss": 0.6403, "step": 276580 }, { "epoch": 23.95, "learning_rate": 1.5133138660807214e-05, "loss": 0.6659, "step": 276590 }, { "epoch": 23.96, "learning_rate": 1.5130970278709377e-05, "loss": 0.5681, "step": 276600 }, { "epoch": 23.96, "learning_rate": 1.512880189661154e-05, "loss": 0.7493, "step": 276610 }, { "epoch": 23.96, "learning_rate": 1.5126633514513703e-05, "loss": 0.6408, "step": 276620 }, { "epoch": 23.96, "learning_rate": 1.5124465132415864e-05, "loss": 0.6303, "step": 276630 }, { "epoch": 23.96, "learning_rate": 1.5122296750318029e-05, "loss": 0.6797, "step": 276640 }, { "epoch": 23.96, "learning_rate": 1.512012836822019e-05, "loss": 0.6501, "step": 276650 }, { "epoch": 23.96, "learning_rate": 1.5117959986122353e-05, "loss": 0.6487, "step": 276660 }, { "epoch": 23.96, "learning_rate": 1.5115791604024515e-05, "loss": 0.5936, "step": 276670 }, { "epoch": 23.96, "learning_rate": 1.511362322192668e-05, "loss": 0.7298, "step": 276680 }, { "epoch": 23.96, "learning_rate": 1.511145483982884e-05, "loss": 0.696, "step": 276690 }, { "epoch": 23.96, "learning_rate": 1.5109286457731004e-05, "loss": 0.6645, "step": 276700 }, { "epoch": 23.96, "learning_rate": 1.5107118075633165e-05, "loss": 0.5897, "step": 276710 }, { "epoch": 23.97, "learning_rate": 1.510494969353533e-05, "loss": 0.6825, "step": 276720 }, { "epoch": 23.97, "learning_rate": 1.5102781311437491e-05, "loss": 0.6156, "step": 276730 }, { "epoch": 23.97, "learning_rate": 1.5100612929339654e-05, "loss": 0.4914, "step": 276740 }, { "epoch": 23.97, "learning_rate": 1.5098444547241817e-05, "loss": 0.6561, "step": 276750 }, { "epoch": 23.97, "learning_rate": 1.509627616514398e-05, "loss": 0.7835, "step": 276760 }, { "epoch": 23.97, "learning_rate": 1.5094107783046141e-05, "loss": 0.7313, "step": 276770 }, { "epoch": 23.97, "learning_rate": 1.5091939400948306e-05, "loss": 0.7196, "step": 276780 }, { "epoch": 23.97, "learning_rate": 1.5089771018850467e-05, "loss": 0.6642, "step": 276790 }, { "epoch": 23.97, "learning_rate": 1.508760263675263e-05, "loss": 0.6493, "step": 276800 }, { "epoch": 23.97, "learning_rate": 1.5085434254654792e-05, "loss": 0.6212, "step": 276810 }, { "epoch": 23.97, "learning_rate": 1.5083265872556956e-05, "loss": 0.6119, "step": 276820 }, { "epoch": 23.98, "learning_rate": 1.5081097490459118e-05, "loss": 0.7035, "step": 276830 }, { "epoch": 23.98, "learning_rate": 1.507892910836128e-05, "loss": 0.6232, "step": 276840 }, { "epoch": 23.98, "learning_rate": 1.5076760726263442e-05, "loss": 0.6403, "step": 276850 }, { "epoch": 23.98, "learning_rate": 1.5074592344165605e-05, "loss": 0.6682, "step": 276860 }, { "epoch": 23.98, "learning_rate": 1.5072423962067768e-05, "loss": 0.651, "step": 276870 }, { "epoch": 23.98, "learning_rate": 1.507025557996993e-05, "loss": 0.6243, "step": 276880 }, { "epoch": 23.98, "learning_rate": 1.5068087197872094e-05, "loss": 0.6414, "step": 276890 }, { "epoch": 23.98, "learning_rate": 1.5065918815774255e-05, "loss": 0.7428, "step": 276900 }, { "epoch": 23.98, "learning_rate": 1.5063750433676418e-05, "loss": 0.5798, "step": 276910 }, { "epoch": 23.98, "learning_rate": 1.506158205157858e-05, "loss": 0.7102, "step": 276920 }, { "epoch": 23.98, "learning_rate": 1.5059413669480744e-05, "loss": 0.7225, "step": 276930 }, { "epoch": 23.98, "learning_rate": 1.5057245287382906e-05, "loss": 0.7264, "step": 276940 }, { "epoch": 23.99, "learning_rate": 1.5055076905285069e-05, "loss": 0.5677, "step": 276950 }, { "epoch": 23.99, "learning_rate": 1.505290852318723e-05, "loss": 0.7514, "step": 276960 }, { "epoch": 23.99, "learning_rate": 1.5050740141089395e-05, "loss": 0.5934, "step": 276970 }, { "epoch": 23.99, "learning_rate": 1.5048571758991556e-05, "loss": 0.6496, "step": 276980 }, { "epoch": 23.99, "learning_rate": 1.5046403376893719e-05, "loss": 0.6325, "step": 276990 }, { "epoch": 23.99, "learning_rate": 1.5044234994795882e-05, "loss": 0.5797, "step": 277000 }, { "epoch": 23.99, "learning_rate": 1.5042066612698045e-05, "loss": 0.5333, "step": 277010 }, { "epoch": 23.99, "learning_rate": 1.5039898230600206e-05, "loss": 0.6039, "step": 277020 }, { "epoch": 23.99, "learning_rate": 1.5037729848502371e-05, "loss": 0.849, "step": 277030 }, { "epoch": 23.99, "learning_rate": 1.5035561466404532e-05, "loss": 0.5719, "step": 277040 }, { "epoch": 23.99, "learning_rate": 1.5033393084306695e-05, "loss": 0.6877, "step": 277050 }, { "epoch": 24.0, "learning_rate": 1.5031224702208857e-05, "loss": 0.6974, "step": 277060 }, { "epoch": 24.0, "learning_rate": 1.5029056320111021e-05, "loss": 0.5706, "step": 277070 }, { "epoch": 24.0, "learning_rate": 1.5026887938013183e-05, "loss": 0.7163, "step": 277080 }, { "epoch": 24.0, "learning_rate": 1.5024719555915346e-05, "loss": 0.4918, "step": 277090 }, { "epoch": 24.0, "learning_rate": 1.5022551173817507e-05, "loss": 0.7792, "step": 277100 }, { "epoch": 24.0, "learning_rate": 1.502038279171967e-05, "loss": 0.69, "step": 277110 }, { "epoch": 24.0, "eval_cer": 0.5338026087112322, "eval_em": 0.060856111397627644, "eval_f1": 0.060856111397627644, "eval_loss": 1.552466630935669, "eval_runtime": 253.9595, "eval_samples_per_second": 7.635, "eval_steps_per_second": 0.957, "eval_wer": 0.9391438886023724, "step": 277116 }, { "epoch": 24.0, "learning_rate": 1.5018214409621833e-05, "loss": 0.6225, "step": 277120 }, { "epoch": 24.0, "learning_rate": 1.5016046027523994e-05, "loss": 0.5276, "step": 277130 }, { "epoch": 24.0, "learning_rate": 1.5013877645426159e-05, "loss": 0.5929, "step": 277140 }, { "epoch": 24.0, "learning_rate": 1.501170926332832e-05, "loss": 0.644, "step": 277150 }, { "epoch": 24.0, "learning_rate": 1.5009540881230483e-05, "loss": 0.5902, "step": 277160 }, { "epoch": 24.0, "learning_rate": 1.5007372499132644e-05, "loss": 0.5272, "step": 277170 }, { "epoch": 24.01, "learning_rate": 1.500520411703481e-05, "loss": 0.5781, "step": 277180 }, { "epoch": 24.01, "learning_rate": 1.500303573493697e-05, "loss": 0.6219, "step": 277190 }, { "epoch": 24.01, "learning_rate": 1.5000867352839133e-05, "loss": 0.5409, "step": 277200 }, { "epoch": 24.01, "learning_rate": 1.4998698970741296e-05, "loss": 0.7808, "step": 277210 }, { "epoch": 24.01, "learning_rate": 1.499653058864346e-05, "loss": 0.5601, "step": 277220 }, { "epoch": 24.01, "learning_rate": 1.499436220654562e-05, "loss": 0.5919, "step": 277230 }, { "epoch": 24.01, "learning_rate": 1.4992193824447784e-05, "loss": 0.5884, "step": 277240 }, { "epoch": 24.01, "learning_rate": 1.4990025442349947e-05, "loss": 0.5707, "step": 277250 }, { "epoch": 24.01, "learning_rate": 1.498785706025211e-05, "loss": 0.6144, "step": 277260 }, { "epoch": 24.01, "learning_rate": 1.4985688678154271e-05, "loss": 0.6105, "step": 277270 }, { "epoch": 24.01, "learning_rate": 1.4983520296056436e-05, "loss": 0.6738, "step": 277280 }, { "epoch": 24.02, "learning_rate": 1.4981351913958597e-05, "loss": 0.7761, "step": 277290 }, { "epoch": 24.02, "learning_rate": 1.497918353186076e-05, "loss": 0.6992, "step": 277300 }, { "epoch": 24.02, "learning_rate": 1.4977015149762921e-05, "loss": 0.6123, "step": 277310 }, { "epoch": 24.02, "learning_rate": 1.4974846767665086e-05, "loss": 0.4654, "step": 277320 }, { "epoch": 24.02, "learning_rate": 1.4972678385567247e-05, "loss": 0.5605, "step": 277330 }, { "epoch": 24.02, "learning_rate": 1.497051000346941e-05, "loss": 0.6085, "step": 277340 }, { "epoch": 24.02, "learning_rate": 1.4968341621371573e-05, "loss": 0.5592, "step": 277350 }, { "epoch": 24.02, "learning_rate": 1.4966173239273735e-05, "loss": 0.6438, "step": 277360 }, { "epoch": 24.02, "learning_rate": 1.4964004857175898e-05, "loss": 0.6283, "step": 277370 }, { "epoch": 24.02, "learning_rate": 1.4961836475078059e-05, "loss": 0.6751, "step": 277380 }, { "epoch": 24.02, "learning_rate": 1.4959668092980224e-05, "loss": 0.7894, "step": 277390 }, { "epoch": 24.02, "learning_rate": 1.4957499710882385e-05, "loss": 0.7222, "step": 277400 }, { "epoch": 24.03, "learning_rate": 1.4955331328784548e-05, "loss": 0.6015, "step": 277410 }, { "epoch": 24.03, "learning_rate": 1.495316294668671e-05, "loss": 0.6553, "step": 277420 }, { "epoch": 24.03, "learning_rate": 1.4950994564588874e-05, "loss": 0.5337, "step": 277430 }, { "epoch": 24.03, "learning_rate": 1.4948826182491035e-05, "loss": 0.6262, "step": 277440 }, { "epoch": 24.03, "learning_rate": 1.4946657800393198e-05, "loss": 0.5598, "step": 277450 }, { "epoch": 24.03, "learning_rate": 1.4944489418295361e-05, "loss": 0.4655, "step": 277460 }, { "epoch": 24.03, "learning_rate": 1.4942321036197524e-05, "loss": 0.6629, "step": 277470 }, { "epoch": 24.03, "learning_rate": 1.4940152654099686e-05, "loss": 0.6328, "step": 277480 }, { "epoch": 24.03, "learning_rate": 1.493798427200185e-05, "loss": 0.7153, "step": 277490 }, { "epoch": 24.03, "learning_rate": 1.4935815889904012e-05, "loss": 0.5611, "step": 277500 }, { "epoch": 24.03, "learning_rate": 1.4933647507806175e-05, "loss": 0.5831, "step": 277510 }, { "epoch": 24.03, "learning_rate": 1.4931479125708336e-05, "loss": 0.7889, "step": 277520 }, { "epoch": 24.04, "learning_rate": 1.49293107436105e-05, "loss": 0.6077, "step": 277530 }, { "epoch": 24.04, "learning_rate": 1.4927142361512662e-05, "loss": 0.5519, "step": 277540 }, { "epoch": 24.04, "learning_rate": 1.4924973979414825e-05, "loss": 0.6882, "step": 277550 }, { "epoch": 24.04, "learning_rate": 1.4922805597316986e-05, "loss": 0.6487, "step": 277560 }, { "epoch": 24.04, "learning_rate": 1.4920637215219151e-05, "loss": 0.5666, "step": 277570 }, { "epoch": 24.04, "learning_rate": 1.4918468833121312e-05, "loss": 0.5796, "step": 277580 }, { "epoch": 24.04, "learning_rate": 1.4916300451023475e-05, "loss": 0.6676, "step": 277590 }, { "epoch": 24.04, "learning_rate": 1.4914132068925638e-05, "loss": 0.6619, "step": 277600 }, { "epoch": 24.04, "learning_rate": 1.4911963686827801e-05, "loss": 0.4878, "step": 277610 }, { "epoch": 24.04, "learning_rate": 1.4909795304729963e-05, "loss": 0.4651, "step": 277620 }, { "epoch": 24.04, "learning_rate": 1.4907626922632124e-05, "loss": 0.6004, "step": 277630 }, { "epoch": 24.05, "learning_rate": 1.4905458540534289e-05, "loss": 0.6575, "step": 277640 }, { "epoch": 24.05, "learning_rate": 1.490329015843645e-05, "loss": 0.6498, "step": 277650 }, { "epoch": 24.05, "learning_rate": 1.4901121776338613e-05, "loss": 0.4716, "step": 277660 }, { "epoch": 24.05, "learning_rate": 1.4898953394240774e-05, "loss": 0.6669, "step": 277670 }, { "epoch": 24.05, "learning_rate": 1.4896785012142939e-05, "loss": 0.6617, "step": 277680 }, { "epoch": 24.05, "learning_rate": 1.48946166300451e-05, "loss": 0.6277, "step": 277690 }, { "epoch": 24.05, "learning_rate": 1.4892448247947263e-05, "loss": 0.6491, "step": 277700 }, { "epoch": 24.05, "learning_rate": 1.4890279865849426e-05, "loss": 0.7078, "step": 277710 }, { "epoch": 24.05, "learning_rate": 1.488811148375159e-05, "loss": 0.6496, "step": 277720 }, { "epoch": 24.05, "learning_rate": 1.488594310165375e-05, "loss": 0.5608, "step": 277730 }, { "epoch": 24.05, "learning_rate": 1.4883774719555915e-05, "loss": 0.6041, "step": 277740 }, { "epoch": 24.05, "learning_rate": 1.4881606337458077e-05, "loss": 0.654, "step": 277750 }, { "epoch": 24.06, "learning_rate": 1.487943795536024e-05, "loss": 0.5127, "step": 277760 }, { "epoch": 24.06, "learning_rate": 1.4877269573262401e-05, "loss": 0.5444, "step": 277770 }, { "epoch": 24.06, "learning_rate": 1.4875101191164566e-05, "loss": 0.6503, "step": 277780 }, { "epoch": 24.06, "learning_rate": 1.4872932809066727e-05, "loss": 0.6203, "step": 277790 }, { "epoch": 24.06, "learning_rate": 1.487076442696889e-05, "loss": 0.7483, "step": 277800 }, { "epoch": 24.06, "learning_rate": 1.4868596044871051e-05, "loss": 0.535, "step": 277810 }, { "epoch": 24.06, "learning_rate": 1.4866427662773216e-05, "loss": 0.7031, "step": 277820 }, { "epoch": 24.06, "learning_rate": 1.4864259280675377e-05, "loss": 0.5613, "step": 277830 }, { "epoch": 24.06, "learning_rate": 1.486209089857754e-05, "loss": 0.7295, "step": 277840 }, { "epoch": 24.06, "learning_rate": 1.4859922516479703e-05, "loss": 0.5465, "step": 277850 }, { "epoch": 24.06, "learning_rate": 1.4857754134381866e-05, "loss": 0.597, "step": 277860 }, { "epoch": 24.07, "learning_rate": 1.4855585752284028e-05, "loss": 0.7679, "step": 277870 }, { "epoch": 24.07, "learning_rate": 1.4853417370186189e-05, "loss": 0.545, "step": 277880 }, { "epoch": 24.07, "learning_rate": 1.4851248988088354e-05, "loss": 0.6233, "step": 277890 }, { "epoch": 24.07, "learning_rate": 1.4849080605990515e-05, "loss": 0.6194, "step": 277900 }, { "epoch": 24.07, "learning_rate": 1.4846912223892678e-05, "loss": 0.7283, "step": 277910 }, { "epoch": 24.07, "learning_rate": 1.484474384179484e-05, "loss": 0.564, "step": 277920 }, { "epoch": 24.07, "learning_rate": 1.4842575459697004e-05, "loss": 0.6148, "step": 277930 }, { "epoch": 24.07, "learning_rate": 1.4840407077599165e-05, "loss": 0.7614, "step": 277940 }, { "epoch": 24.07, "learning_rate": 1.4838238695501328e-05, "loss": 0.6974, "step": 277950 }, { "epoch": 24.07, "learning_rate": 1.4836070313403491e-05, "loss": 0.5315, "step": 277960 }, { "epoch": 24.07, "learning_rate": 1.4833901931305654e-05, "loss": 0.593, "step": 277970 }, { "epoch": 24.07, "learning_rate": 1.4831733549207816e-05, "loss": 0.636, "step": 277980 }, { "epoch": 24.08, "learning_rate": 1.482956516710998e-05, "loss": 0.6784, "step": 277990 }, { "epoch": 24.08, "learning_rate": 1.4827396785012142e-05, "loss": 0.6135, "step": 278000 }, { "epoch": 24.08, "learning_rate": 1.4825228402914305e-05, "loss": 0.6149, "step": 278010 }, { "epoch": 24.08, "learning_rate": 1.4823060020816466e-05, "loss": 0.531, "step": 278020 }, { "epoch": 24.08, "learning_rate": 1.482089163871863e-05, "loss": 0.5788, "step": 278030 }, { "epoch": 24.08, "learning_rate": 1.4818723256620792e-05, "loss": 0.5236, "step": 278040 }, { "epoch": 24.08, "learning_rate": 1.4816554874522955e-05, "loss": 0.449, "step": 278050 }, { "epoch": 24.08, "learning_rate": 1.4814386492425116e-05, "loss": 0.6859, "step": 278060 }, { "epoch": 24.08, "learning_rate": 1.4812218110327281e-05, "loss": 0.7044, "step": 278070 }, { "epoch": 24.08, "learning_rate": 1.4810049728229442e-05, "loss": 0.5324, "step": 278080 }, { "epoch": 24.08, "learning_rate": 1.4807881346131605e-05, "loss": 0.6031, "step": 278090 }, { "epoch": 24.09, "learning_rate": 1.4805712964033768e-05, "loss": 0.6528, "step": 278100 }, { "epoch": 24.09, "learning_rate": 1.4803544581935931e-05, "loss": 0.6512, "step": 278110 }, { "epoch": 24.09, "learning_rate": 1.4801376199838093e-05, "loss": 0.5575, "step": 278120 }, { "epoch": 24.09, "learning_rate": 1.4799207817740254e-05, "loss": 0.6585, "step": 278130 }, { "epoch": 24.09, "learning_rate": 1.4797039435642419e-05, "loss": 0.6096, "step": 278140 }, { "epoch": 24.09, "learning_rate": 1.479487105354458e-05, "loss": 0.5799, "step": 278150 }, { "epoch": 24.09, "learning_rate": 1.4792702671446743e-05, "loss": 0.5995, "step": 278160 }, { "epoch": 24.09, "learning_rate": 1.4790534289348904e-05, "loss": 0.5593, "step": 278170 }, { "epoch": 24.09, "learning_rate": 1.4788365907251069e-05, "loss": 0.6605, "step": 278180 }, { "epoch": 24.09, "learning_rate": 1.478619752515323e-05, "loss": 0.5984, "step": 278190 }, { "epoch": 24.09, "learning_rate": 1.4784029143055393e-05, "loss": 0.6678, "step": 278200 }, { "epoch": 24.09, "learning_rate": 1.4781860760957556e-05, "loss": 0.6782, "step": 278210 }, { "epoch": 24.1, "learning_rate": 1.477969237885972e-05, "loss": 0.6996, "step": 278220 }, { "epoch": 24.1, "learning_rate": 1.477752399676188e-05, "loss": 0.5847, "step": 278230 }, { "epoch": 24.1, "learning_rate": 1.4775355614664045e-05, "loss": 0.5557, "step": 278240 }, { "epoch": 24.1, "learning_rate": 1.4773187232566207e-05, "loss": 0.5989, "step": 278250 }, { "epoch": 24.1, "learning_rate": 1.477101885046837e-05, "loss": 0.623, "step": 278260 }, { "epoch": 24.1, "learning_rate": 1.476885046837053e-05, "loss": 0.7385, "step": 278270 }, { "epoch": 24.1, "learning_rate": 1.4766682086272696e-05, "loss": 0.7871, "step": 278280 }, { "epoch": 24.1, "learning_rate": 1.4764513704174857e-05, "loss": 0.5828, "step": 278290 }, { "epoch": 24.1, "learning_rate": 1.476234532207702e-05, "loss": 0.7376, "step": 278300 }, { "epoch": 24.1, "learning_rate": 1.4760176939979181e-05, "loss": 0.6289, "step": 278310 }, { "epoch": 24.1, "learning_rate": 1.4758008557881346e-05, "loss": 0.6153, "step": 278320 }, { "epoch": 24.11, "learning_rate": 1.4755840175783507e-05, "loss": 0.5162, "step": 278330 }, { "epoch": 24.11, "learning_rate": 1.475367179368567e-05, "loss": 0.7463, "step": 278340 }, { "epoch": 24.11, "learning_rate": 1.4751503411587833e-05, "loss": 0.625, "step": 278350 }, { "epoch": 24.11, "learning_rate": 1.4749335029489996e-05, "loss": 0.708, "step": 278360 }, { "epoch": 24.11, "learning_rate": 1.4747166647392157e-05, "loss": 0.6629, "step": 278370 }, { "epoch": 24.11, "learning_rate": 1.4744998265294322e-05, "loss": 0.791, "step": 278380 }, { "epoch": 24.11, "learning_rate": 1.4742829883196483e-05, "loss": 0.5088, "step": 278390 }, { "epoch": 24.11, "learning_rate": 1.4740661501098645e-05, "loss": 0.565, "step": 278400 }, { "epoch": 24.11, "learning_rate": 1.4738493119000808e-05, "loss": 0.5194, "step": 278410 }, { "epoch": 24.11, "learning_rate": 1.473632473690297e-05, "loss": 0.6444, "step": 278420 }, { "epoch": 24.11, "learning_rate": 1.4734156354805134e-05, "loss": 0.5307, "step": 278430 }, { "epoch": 24.11, "learning_rate": 1.4731987972707295e-05, "loss": 0.5879, "step": 278440 }, { "epoch": 24.12, "learning_rate": 1.4729819590609458e-05, "loss": 0.5489, "step": 278450 }, { "epoch": 24.12, "learning_rate": 1.4727651208511621e-05, "loss": 0.6458, "step": 278460 }, { "epoch": 24.12, "learning_rate": 1.4725482826413784e-05, "loss": 0.5652, "step": 278470 }, { "epoch": 24.12, "learning_rate": 1.4723314444315945e-05, "loss": 0.7061, "step": 278480 }, { "epoch": 24.12, "learning_rate": 1.472114606221811e-05, "loss": 0.614, "step": 278490 }, { "epoch": 24.12, "learning_rate": 1.4718977680120271e-05, "loss": 0.6758, "step": 278500 }, { "epoch": 24.12, "learning_rate": 1.4716809298022434e-05, "loss": 0.6166, "step": 278510 }, { "epoch": 24.12, "learning_rate": 1.4714640915924596e-05, "loss": 0.638, "step": 278520 }, { "epoch": 24.12, "learning_rate": 1.471247253382676e-05, "loss": 0.6376, "step": 278530 }, { "epoch": 24.12, "learning_rate": 1.4710304151728922e-05, "loss": 0.556, "step": 278540 }, { "epoch": 24.12, "learning_rate": 1.4708135769631085e-05, "loss": 0.6121, "step": 278550 }, { "epoch": 24.13, "learning_rate": 1.4705967387533248e-05, "loss": 0.7085, "step": 278560 }, { "epoch": 24.13, "learning_rate": 1.470379900543541e-05, "loss": 0.6393, "step": 278570 }, { "epoch": 24.13, "learning_rate": 1.4701630623337572e-05, "loss": 0.6694, "step": 278580 }, { "epoch": 24.13, "learning_rate": 1.4699462241239735e-05, "loss": 0.6717, "step": 278590 }, { "epoch": 24.13, "learning_rate": 1.4697293859141898e-05, "loss": 0.7072, "step": 278600 }, { "epoch": 24.13, "learning_rate": 1.4695125477044061e-05, "loss": 0.5889, "step": 278610 }, { "epoch": 24.13, "learning_rate": 1.4692957094946222e-05, "loss": 0.6035, "step": 278620 }, { "epoch": 24.13, "learning_rate": 1.4690788712848387e-05, "loss": 0.747, "step": 278630 }, { "epoch": 24.13, "learning_rate": 1.4688620330750548e-05, "loss": 0.6795, "step": 278640 }, { "epoch": 24.13, "learning_rate": 1.468645194865271e-05, "loss": 0.5029, "step": 278650 }, { "epoch": 24.13, "learning_rate": 1.4684283566554873e-05, "loss": 0.6257, "step": 278660 }, { "epoch": 24.13, "learning_rate": 1.4682115184457036e-05, "loss": 0.7742, "step": 278670 }, { "epoch": 24.14, "learning_rate": 1.4679946802359199e-05, "loss": 0.5643, "step": 278680 }, { "epoch": 24.14, "learning_rate": 1.467777842026136e-05, "loss": 0.6299, "step": 278690 }, { "epoch": 24.14, "learning_rate": 1.4675610038163525e-05, "loss": 0.4664, "step": 278700 }, { "epoch": 24.14, "learning_rate": 1.4673441656065686e-05, "loss": 0.7223, "step": 278710 }, { "epoch": 24.14, "learning_rate": 1.4671273273967849e-05, "loss": 0.6227, "step": 278720 }, { "epoch": 24.14, "learning_rate": 1.466910489187001e-05, "loss": 0.6359, "step": 278730 }, { "epoch": 24.14, "learning_rate": 1.4666936509772175e-05, "loss": 0.5321, "step": 278740 }, { "epoch": 24.14, "learning_rate": 1.4664768127674336e-05, "loss": 0.5797, "step": 278750 }, { "epoch": 24.14, "learning_rate": 1.46625997455765e-05, "loss": 0.6319, "step": 278760 }, { "epoch": 24.14, "learning_rate": 1.466043136347866e-05, "loss": 0.7114, "step": 278770 }, { "epoch": 24.14, "learning_rate": 1.4658262981380825e-05, "loss": 0.6519, "step": 278780 }, { "epoch": 24.14, "learning_rate": 1.4656094599282987e-05, "loss": 0.7875, "step": 278790 }, { "epoch": 24.15, "learning_rate": 1.465392621718515e-05, "loss": 0.6848, "step": 278800 }, { "epoch": 24.15, "learning_rate": 1.4651757835087313e-05, "loss": 0.6943, "step": 278810 }, { "epoch": 24.15, "learning_rate": 1.4649589452989476e-05, "loss": 0.5907, "step": 278820 }, { "epoch": 24.15, "learning_rate": 1.4647421070891637e-05, "loss": 0.711, "step": 278830 }, { "epoch": 24.15, "learning_rate": 1.4645252688793802e-05, "loss": 0.5424, "step": 278840 }, { "epoch": 24.15, "learning_rate": 1.4643084306695963e-05, "loss": 0.6307, "step": 278850 }, { "epoch": 24.15, "learning_rate": 1.4640915924598126e-05, "loss": 0.7068, "step": 278860 }, { "epoch": 24.15, "learning_rate": 1.4638747542500287e-05, "loss": 0.653, "step": 278870 }, { "epoch": 24.15, "learning_rate": 1.4636579160402452e-05, "loss": 0.644, "step": 278880 }, { "epoch": 24.15, "learning_rate": 1.4634410778304613e-05, "loss": 0.6854, "step": 278890 }, { "epoch": 24.15, "learning_rate": 1.4632242396206775e-05, "loss": 0.5919, "step": 278900 }, { "epoch": 24.16, "learning_rate": 1.4630074014108938e-05, "loss": 0.7372, "step": 278910 }, { "epoch": 24.16, "learning_rate": 1.46279056320111e-05, "loss": 0.553, "step": 278920 }, { "epoch": 24.16, "learning_rate": 1.4625737249913264e-05, "loss": 0.5666, "step": 278930 }, { "epoch": 24.16, "learning_rate": 1.4623568867815425e-05, "loss": 0.5995, "step": 278940 }, { "epoch": 24.16, "learning_rate": 1.462140048571759e-05, "loss": 0.5811, "step": 278950 }, { "epoch": 24.16, "learning_rate": 1.4619232103619751e-05, "loss": 0.6071, "step": 278960 }, { "epoch": 24.16, "learning_rate": 1.4617063721521914e-05, "loss": 0.7083, "step": 278970 }, { "epoch": 24.16, "learning_rate": 1.4614895339424075e-05, "loss": 0.5714, "step": 278980 }, { "epoch": 24.16, "learning_rate": 1.461272695732624e-05, "loss": 0.6406, "step": 278990 }, { "epoch": 24.16, "learning_rate": 1.4610558575228401e-05, "loss": 0.644, "step": 279000 }, { "epoch": 24.16, "learning_rate": 1.4608390193130564e-05, "loss": 0.6321, "step": 279010 }, { "epoch": 24.16, "learning_rate": 1.4606221811032726e-05, "loss": 0.5786, "step": 279020 }, { "epoch": 24.17, "learning_rate": 1.460405342893489e-05, "loss": 0.6716, "step": 279030 }, { "epoch": 24.17, "learning_rate": 1.4601885046837052e-05, "loss": 0.6869, "step": 279040 }, { "epoch": 24.17, "learning_rate": 1.4599716664739215e-05, "loss": 0.749, "step": 279050 }, { "epoch": 24.17, "learning_rate": 1.4597548282641378e-05, "loss": 0.6797, "step": 279060 }, { "epoch": 24.17, "learning_rate": 1.459537990054354e-05, "loss": 0.7218, "step": 279070 }, { "epoch": 24.17, "learning_rate": 1.4593211518445702e-05, "loss": 0.697, "step": 279080 }, { "epoch": 24.17, "learning_rate": 1.4591043136347867e-05, "loss": 0.5502, "step": 279090 }, { "epoch": 24.17, "learning_rate": 1.4588874754250028e-05, "loss": 0.6967, "step": 279100 }, { "epoch": 24.17, "learning_rate": 1.4586706372152191e-05, "loss": 0.5956, "step": 279110 }, { "epoch": 24.17, "learning_rate": 1.4584537990054352e-05, "loss": 0.538, "step": 279120 }, { "epoch": 24.17, "learning_rate": 1.4582369607956517e-05, "loss": 0.5235, "step": 279130 }, { "epoch": 24.18, "learning_rate": 1.4580201225858678e-05, "loss": 0.61, "step": 279140 }, { "epoch": 24.18, "learning_rate": 1.4578032843760841e-05, "loss": 0.6759, "step": 279150 }, { "epoch": 24.18, "learning_rate": 1.4575864461663003e-05, "loss": 0.5187, "step": 279160 }, { "epoch": 24.18, "learning_rate": 1.4573696079565166e-05, "loss": 0.553, "step": 279170 }, { "epoch": 24.18, "learning_rate": 1.4571527697467329e-05, "loss": 0.4985, "step": 279180 }, { "epoch": 24.18, "learning_rate": 1.456935931536949e-05, "loss": 0.6961, "step": 279190 }, { "epoch": 24.18, "learning_rate": 1.4567190933271655e-05, "loss": 0.5295, "step": 279200 }, { "epoch": 24.18, "learning_rate": 1.4565022551173816e-05, "loss": 0.7461, "step": 279210 }, { "epoch": 24.18, "learning_rate": 1.4562854169075979e-05, "loss": 0.6399, "step": 279220 }, { "epoch": 24.18, "learning_rate": 1.456068578697814e-05, "loss": 0.6788, "step": 279230 }, { "epoch": 24.18, "learning_rate": 1.4558517404880305e-05, "loss": 0.7137, "step": 279240 }, { "epoch": 24.18, "learning_rate": 1.4556349022782466e-05, "loss": 0.6388, "step": 279250 }, { "epoch": 24.19, "learning_rate": 1.455418064068463e-05, "loss": 0.6225, "step": 279260 }, { "epoch": 24.19, "learning_rate": 1.455201225858679e-05, "loss": 0.6284, "step": 279270 }, { "epoch": 24.19, "learning_rate": 1.4549843876488955e-05, "loss": 0.4148, "step": 279280 }, { "epoch": 24.19, "learning_rate": 1.4547675494391117e-05, "loss": 0.6296, "step": 279290 }, { "epoch": 24.19, "learning_rate": 1.454550711229328e-05, "loss": 0.7229, "step": 279300 }, { "epoch": 24.19, "learning_rate": 1.4543338730195443e-05, "loss": 0.6045, "step": 279310 }, { "epoch": 24.19, "learning_rate": 1.4541170348097606e-05, "loss": 0.4537, "step": 279320 }, { "epoch": 24.19, "learning_rate": 1.4539001965999767e-05, "loss": 0.6306, "step": 279330 }, { "epoch": 24.19, "learning_rate": 1.4536833583901932e-05, "loss": 0.7662, "step": 279340 }, { "epoch": 24.19, "learning_rate": 1.4534665201804093e-05, "loss": 0.6912, "step": 279350 }, { "epoch": 24.19, "learning_rate": 1.4532496819706256e-05, "loss": 0.6358, "step": 279360 }, { "epoch": 24.2, "learning_rate": 1.4530328437608417e-05, "loss": 0.618, "step": 279370 }, { "epoch": 24.2, "learning_rate": 1.4528160055510582e-05, "loss": 0.6252, "step": 279380 }, { "epoch": 24.2, "learning_rate": 1.4525991673412743e-05, "loss": 0.625, "step": 279390 }, { "epoch": 24.2, "learning_rate": 1.4523823291314906e-05, "loss": 0.5793, "step": 279400 }, { "epoch": 24.2, "learning_rate": 1.4521654909217067e-05, "loss": 0.6524, "step": 279410 }, { "epoch": 24.2, "learning_rate": 1.451948652711923e-05, "loss": 0.6068, "step": 279420 }, { "epoch": 24.2, "learning_rate": 1.4517318145021393e-05, "loss": 0.7121, "step": 279430 }, { "epoch": 24.2, "learning_rate": 1.4515149762923555e-05, "loss": 0.6195, "step": 279440 }, { "epoch": 24.2, "learning_rate": 1.451298138082572e-05, "loss": 0.4537, "step": 279450 }, { "epoch": 24.2, "learning_rate": 1.451081299872788e-05, "loss": 0.6461, "step": 279460 }, { "epoch": 24.2, "learning_rate": 1.4508644616630044e-05, "loss": 0.6352, "step": 279470 }, { "epoch": 24.2, "learning_rate": 1.4506476234532205e-05, "loss": 0.7037, "step": 279480 }, { "epoch": 24.21, "learning_rate": 1.450430785243437e-05, "loss": 0.5695, "step": 279490 }, { "epoch": 24.21, "learning_rate": 1.4502139470336531e-05, "loss": 0.6014, "step": 279500 }, { "epoch": 24.21, "learning_rate": 1.4499971088238694e-05, "loss": 0.6918, "step": 279510 }, { "epoch": 24.21, "learning_rate": 1.4497802706140855e-05, "loss": 0.6838, "step": 279520 }, { "epoch": 24.21, "learning_rate": 1.449563432404302e-05, "loss": 0.585, "step": 279530 }, { "epoch": 24.21, "learning_rate": 1.4493465941945181e-05, "loss": 0.635, "step": 279540 }, { "epoch": 24.21, "learning_rate": 1.4491297559847344e-05, "loss": 0.5868, "step": 279550 }, { "epoch": 24.21, "learning_rate": 1.4489129177749507e-05, "loss": 0.581, "step": 279560 }, { "epoch": 24.21, "learning_rate": 1.448696079565167e-05, "loss": 0.6307, "step": 279570 }, { "epoch": 24.21, "learning_rate": 1.4484792413553832e-05, "loss": 0.5763, "step": 279580 }, { "epoch": 24.21, "learning_rate": 1.4482624031455996e-05, "loss": 0.5322, "step": 279590 }, { "epoch": 24.22, "learning_rate": 1.4480455649358158e-05, "loss": 0.6361, "step": 279600 }, { "epoch": 24.22, "learning_rate": 1.447828726726032e-05, "loss": 0.7004, "step": 279610 }, { "epoch": 24.22, "learning_rate": 1.4476118885162482e-05, "loss": 0.8, "step": 279620 }, { "epoch": 24.22, "learning_rate": 1.4473950503064647e-05, "loss": 0.5851, "step": 279630 }, { "epoch": 24.22, "learning_rate": 1.4471782120966808e-05, "loss": 0.5621, "step": 279640 }, { "epoch": 24.22, "learning_rate": 1.4469613738868971e-05, "loss": 0.6969, "step": 279650 }, { "epoch": 24.22, "learning_rate": 1.4467445356771132e-05, "loss": 0.7825, "step": 279660 }, { "epoch": 24.22, "learning_rate": 1.4465276974673295e-05, "loss": 0.672, "step": 279670 }, { "epoch": 24.22, "learning_rate": 1.4463108592575458e-05, "loss": 0.5008, "step": 279680 }, { "epoch": 24.22, "learning_rate": 1.446094021047762e-05, "loss": 0.6603, "step": 279690 }, { "epoch": 24.22, "learning_rate": 1.4458771828379784e-05, "loss": 0.7059, "step": 279700 }, { "epoch": 24.22, "learning_rate": 1.4456603446281946e-05, "loss": 0.612, "step": 279710 }, { "epoch": 24.23, "learning_rate": 1.4454435064184109e-05, "loss": 0.6341, "step": 279720 }, { "epoch": 24.23, "learning_rate": 1.445226668208627e-05, "loss": 0.6715, "step": 279730 }, { "epoch": 24.23, "learning_rate": 1.4450098299988435e-05, "loss": 0.6989, "step": 279740 }, { "epoch": 24.23, "learning_rate": 1.4447929917890596e-05, "loss": 0.6708, "step": 279750 }, { "epoch": 24.23, "learning_rate": 1.4445761535792759e-05, "loss": 0.6825, "step": 279760 }, { "epoch": 24.23, "learning_rate": 1.4443593153694922e-05, "loss": 0.6225, "step": 279770 }, { "epoch": 24.23, "learning_rate": 1.4441424771597085e-05, "loss": 0.5634, "step": 279780 }, { "epoch": 24.23, "learning_rate": 1.4439256389499246e-05, "loss": 0.6522, "step": 279790 }, { "epoch": 24.23, "learning_rate": 1.443708800740141e-05, "loss": 0.5081, "step": 279800 }, { "epoch": 24.23, "learning_rate": 1.4434919625303572e-05, "loss": 0.5399, "step": 279810 }, { "epoch": 24.23, "learning_rate": 1.4432751243205735e-05, "loss": 0.6429, "step": 279820 }, { "epoch": 24.24, "learning_rate": 1.4430582861107897e-05, "loss": 0.6177, "step": 279830 }, { "epoch": 24.24, "learning_rate": 1.4428414479010061e-05, "loss": 0.642, "step": 279840 }, { "epoch": 24.24, "learning_rate": 1.4426246096912223e-05, "loss": 0.7455, "step": 279850 }, { "epoch": 24.24, "learning_rate": 1.4424077714814386e-05, "loss": 0.4681, "step": 279860 }, { "epoch": 24.24, "learning_rate": 1.4421909332716547e-05, "loss": 0.6061, "step": 279870 }, { "epoch": 24.24, "learning_rate": 1.4419740950618712e-05, "loss": 0.6517, "step": 279880 }, { "epoch": 24.24, "learning_rate": 1.4417572568520873e-05, "loss": 0.6548, "step": 279890 }, { "epoch": 24.24, "learning_rate": 1.4415404186423036e-05, "loss": 0.5162, "step": 279900 }, { "epoch": 24.24, "learning_rate": 1.4413235804325199e-05, "loss": 0.6132, "step": 279910 }, { "epoch": 24.24, "learning_rate": 1.4411067422227362e-05, "loss": 0.5384, "step": 279920 }, { "epoch": 24.24, "learning_rate": 1.4408899040129523e-05, "loss": 0.7087, "step": 279930 }, { "epoch": 24.24, "learning_rate": 1.4406730658031685e-05, "loss": 0.6727, "step": 279940 }, { "epoch": 24.25, "learning_rate": 1.440456227593385e-05, "loss": 0.6612, "step": 279950 }, { "epoch": 24.25, "learning_rate": 1.440239389383601e-05, "loss": 0.6874, "step": 279960 }, { "epoch": 24.25, "learning_rate": 1.4400225511738174e-05, "loss": 0.6036, "step": 279970 }, { "epoch": 24.25, "learning_rate": 1.4398057129640335e-05, "loss": 0.6238, "step": 279980 }, { "epoch": 24.25, "learning_rate": 1.43958887475425e-05, "loss": 0.6271, "step": 279990 }, { "epoch": 24.25, "learning_rate": 1.4393720365444661e-05, "loss": 0.5762, "step": 280000 }, { "epoch": 24.25, "learning_rate": 1.4391551983346824e-05, "loss": 0.8501, "step": 280010 }, { "epoch": 24.25, "learning_rate": 1.4389383601248987e-05, "loss": 0.5516, "step": 280020 }, { "epoch": 24.25, "learning_rate": 1.438721521915115e-05, "loss": 0.5708, "step": 280030 }, { "epoch": 24.25, "learning_rate": 1.4385046837053311e-05, "loss": 0.4585, "step": 280040 }, { "epoch": 24.25, "learning_rate": 1.4382878454955476e-05, "loss": 0.6517, "step": 280050 }, { "epoch": 24.25, "learning_rate": 1.4380710072857637e-05, "loss": 0.6242, "step": 280060 }, { "epoch": 24.26, "learning_rate": 1.43785416907598e-05, "loss": 0.6082, "step": 280070 }, { "epoch": 24.26, "learning_rate": 1.4376373308661962e-05, "loss": 0.4644, "step": 280080 }, { "epoch": 24.26, "learning_rate": 1.4374204926564126e-05, "loss": 0.7668, "step": 280090 }, { "epoch": 24.26, "learning_rate": 1.4372036544466288e-05, "loss": 0.7095, "step": 280100 }, { "epoch": 24.26, "learning_rate": 1.436986816236845e-05, "loss": 0.7056, "step": 280110 }, { "epoch": 24.26, "learning_rate": 1.4367699780270612e-05, "loss": 0.5837, "step": 280120 }, { "epoch": 24.26, "learning_rate": 1.4365531398172777e-05, "loss": 0.648, "step": 280130 }, { "epoch": 24.26, "learning_rate": 1.4363363016074938e-05, "loss": 0.5949, "step": 280140 }, { "epoch": 24.26, "learning_rate": 1.4361194633977101e-05, "loss": 0.6626, "step": 280150 }, { "epoch": 24.26, "learning_rate": 1.4359026251879264e-05, "loss": 0.653, "step": 280160 }, { "epoch": 24.26, "learning_rate": 1.4356857869781427e-05, "loss": 0.5563, "step": 280170 }, { "epoch": 24.27, "learning_rate": 1.4354689487683588e-05, "loss": 0.6172, "step": 280180 }, { "epoch": 24.27, "learning_rate": 1.435252110558575e-05, "loss": 0.6862, "step": 280190 }, { "epoch": 24.27, "learning_rate": 1.4350352723487914e-05, "loss": 0.5732, "step": 280200 }, { "epoch": 24.27, "learning_rate": 1.4348184341390076e-05, "loss": 0.5422, "step": 280210 }, { "epoch": 24.27, "learning_rate": 1.4346015959292239e-05, "loss": 0.4905, "step": 280220 }, { "epoch": 24.27, "learning_rate": 1.43438475771944e-05, "loss": 0.622, "step": 280230 }, { "epoch": 24.27, "learning_rate": 1.4341679195096565e-05, "loss": 0.7268, "step": 280240 }, { "epoch": 24.27, "learning_rate": 1.4339510812998726e-05, "loss": 0.5347, "step": 280250 }, { "epoch": 24.27, "learning_rate": 1.4337342430900889e-05, "loss": 0.7086, "step": 280260 }, { "epoch": 24.27, "learning_rate": 1.4335174048803052e-05, "loss": 0.6224, "step": 280270 }, { "epoch": 24.27, "learning_rate": 1.4333005666705215e-05, "loss": 0.6035, "step": 280280 }, { "epoch": 24.27, "learning_rate": 1.4330837284607376e-05, "loss": 0.6798, "step": 280290 }, { "epoch": 24.28, "learning_rate": 1.4328668902509541e-05, "loss": 0.5834, "step": 280300 }, { "epoch": 24.28, "learning_rate": 1.4326500520411702e-05, "loss": 0.6972, "step": 280310 }, { "epoch": 24.28, "learning_rate": 1.4324332138313865e-05, "loss": 0.5667, "step": 280320 }, { "epoch": 24.28, "learning_rate": 1.4322163756216027e-05, "loss": 0.6813, "step": 280330 }, { "epoch": 24.28, "learning_rate": 1.4319995374118191e-05, "loss": 0.6217, "step": 280340 }, { "epoch": 24.28, "learning_rate": 1.4317826992020353e-05, "loss": 0.7641, "step": 280350 }, { "epoch": 24.28, "learning_rate": 1.4315658609922516e-05, "loss": 0.7132, "step": 280360 }, { "epoch": 24.28, "learning_rate": 1.4313490227824677e-05, "loss": 0.5453, "step": 280370 }, { "epoch": 24.28, "learning_rate": 1.4311321845726842e-05, "loss": 0.6334, "step": 280380 }, { "epoch": 24.28, "learning_rate": 1.4309153463629003e-05, "loss": 0.5715, "step": 280390 }, { "epoch": 24.28, "learning_rate": 1.4306985081531166e-05, "loss": 0.7438, "step": 280400 }, { "epoch": 24.29, "learning_rate": 1.4304816699433329e-05, "loss": 0.5486, "step": 280410 }, { "epoch": 24.29, "learning_rate": 1.4302648317335492e-05, "loss": 0.667, "step": 280420 }, { "epoch": 24.29, "learning_rate": 1.4300479935237653e-05, "loss": 0.6053, "step": 280430 }, { "epoch": 24.29, "learning_rate": 1.4298311553139814e-05, "loss": 0.5826, "step": 280440 }, { "epoch": 24.29, "learning_rate": 1.429614317104198e-05, "loss": 0.6921, "step": 280450 }, { "epoch": 24.29, "learning_rate": 1.429397478894414e-05, "loss": 0.5493, "step": 280460 }, { "epoch": 24.29, "learning_rate": 1.4291806406846303e-05, "loss": 0.6615, "step": 280470 }, { "epoch": 24.29, "learning_rate": 1.4289638024748465e-05, "loss": 0.5432, "step": 280480 }, { "epoch": 24.29, "learning_rate": 1.428746964265063e-05, "loss": 0.7303, "step": 280490 }, { "epoch": 24.29, "learning_rate": 1.428530126055279e-05, "loss": 0.5578, "step": 280500 }, { "epoch": 24.29, "learning_rate": 1.4283132878454954e-05, "loss": 0.6058, "step": 280510 }, { "epoch": 24.29, "learning_rate": 1.4280964496357117e-05, "loss": 0.751, "step": 280520 }, { "epoch": 24.3, "learning_rate": 1.427879611425928e-05, "loss": 0.7303, "step": 280530 }, { "epoch": 24.3, "learning_rate": 1.4276627732161441e-05, "loss": 0.5571, "step": 280540 }, { "epoch": 24.3, "learning_rate": 1.4274459350063606e-05, "loss": 0.6156, "step": 280550 }, { "epoch": 24.3, "learning_rate": 1.4272290967965767e-05, "loss": 0.5964, "step": 280560 }, { "epoch": 24.3, "learning_rate": 1.427012258586793e-05, "loss": 0.6095, "step": 280570 }, { "epoch": 24.3, "learning_rate": 1.4267954203770091e-05, "loss": 0.7234, "step": 280580 }, { "epoch": 24.3, "learning_rate": 1.4265785821672256e-05, "loss": 0.6273, "step": 280590 }, { "epoch": 24.3, "learning_rate": 1.4263617439574417e-05, "loss": 0.5689, "step": 280600 }, { "epoch": 24.3, "learning_rate": 1.426144905747658e-05, "loss": 0.5682, "step": 280610 }, { "epoch": 24.3, "learning_rate": 1.4259280675378742e-05, "loss": 0.4729, "step": 280620 }, { "epoch": 24.3, "learning_rate": 1.4257112293280906e-05, "loss": 0.6225, "step": 280630 }, { "epoch": 24.31, "learning_rate": 1.4254943911183068e-05, "loss": 0.6321, "step": 280640 }, { "epoch": 24.31, "learning_rate": 1.425277552908523e-05, "loss": 0.6809, "step": 280650 }, { "epoch": 24.31, "learning_rate": 1.4250607146987394e-05, "loss": 0.6138, "step": 280660 }, { "epoch": 24.31, "learning_rate": 1.4248438764889557e-05, "loss": 0.5875, "step": 280670 }, { "epoch": 24.31, "learning_rate": 1.4246270382791718e-05, "loss": 0.6308, "step": 280680 }, { "epoch": 24.31, "learning_rate": 1.4244102000693883e-05, "loss": 0.5732, "step": 280690 }, { "epoch": 24.31, "learning_rate": 1.4241933618596044e-05, "loss": 0.7383, "step": 280700 }, { "epoch": 24.31, "learning_rate": 1.4239765236498205e-05, "loss": 0.6828, "step": 280710 }, { "epoch": 24.31, "learning_rate": 1.4237596854400368e-05, "loss": 0.6451, "step": 280720 }, { "epoch": 24.31, "learning_rate": 1.423542847230253e-05, "loss": 0.5698, "step": 280730 }, { "epoch": 24.31, "learning_rate": 1.4233260090204694e-05, "loss": 0.6849, "step": 280740 }, { "epoch": 24.31, "learning_rate": 1.4231091708106856e-05, "loss": 0.6022, "step": 280750 }, { "epoch": 24.32, "learning_rate": 1.4228923326009019e-05, "loss": 0.6758, "step": 280760 }, { "epoch": 24.32, "learning_rate": 1.4226754943911182e-05, "loss": 0.7074, "step": 280770 }, { "epoch": 24.32, "learning_rate": 1.4224586561813345e-05, "loss": 0.5678, "step": 280780 }, { "epoch": 24.32, "learning_rate": 1.4222418179715506e-05, "loss": 0.5933, "step": 280790 }, { "epoch": 24.32, "learning_rate": 1.422024979761767e-05, "loss": 0.4949, "step": 280800 }, { "epoch": 24.32, "learning_rate": 1.4218081415519832e-05, "loss": 0.6966, "step": 280810 }, { "epoch": 24.32, "learning_rate": 1.4215913033421995e-05, "loss": 0.6654, "step": 280820 }, { "epoch": 24.32, "learning_rate": 1.4213744651324156e-05, "loss": 0.6393, "step": 280830 }, { "epoch": 24.32, "learning_rate": 1.4211576269226321e-05, "loss": 0.6773, "step": 280840 }, { "epoch": 24.32, "learning_rate": 1.4209407887128482e-05, "loss": 0.6954, "step": 280850 }, { "epoch": 24.32, "learning_rate": 1.4207239505030645e-05, "loss": 0.6046, "step": 280860 }, { "epoch": 24.33, "learning_rate": 1.4205071122932807e-05, "loss": 0.5125, "step": 280870 }, { "epoch": 24.33, "learning_rate": 1.4202902740834971e-05, "loss": 0.6079, "step": 280880 }, { "epoch": 24.33, "learning_rate": 1.4200734358737133e-05, "loss": 0.5543, "step": 280890 }, { "epoch": 24.33, "learning_rate": 1.4198565976639296e-05, "loss": 0.585, "step": 280900 }, { "epoch": 24.33, "learning_rate": 1.4196397594541459e-05, "loss": 0.7564, "step": 280910 }, { "epoch": 24.33, "learning_rate": 1.4194229212443622e-05, "loss": 0.6827, "step": 280920 }, { "epoch": 24.33, "learning_rate": 1.4192060830345783e-05, "loss": 0.6604, "step": 280930 }, { "epoch": 24.33, "learning_rate": 1.4189892448247948e-05, "loss": 0.596, "step": 280940 }, { "epoch": 24.33, "learning_rate": 1.4187724066150109e-05, "loss": 0.6279, "step": 280950 }, { "epoch": 24.33, "learning_rate": 1.418555568405227e-05, "loss": 0.6677, "step": 280960 }, { "epoch": 24.33, "learning_rate": 1.4183387301954433e-05, "loss": 0.6774, "step": 280970 }, { "epoch": 24.33, "learning_rate": 1.4181218919856596e-05, "loss": 0.7303, "step": 280980 }, { "epoch": 24.34, "learning_rate": 1.417905053775876e-05, "loss": 0.6468, "step": 280990 }, { "epoch": 24.34, "learning_rate": 1.417688215566092e-05, "loss": 0.5263, "step": 281000 }, { "epoch": 24.34, "learning_rate": 1.4174713773563084e-05, "loss": 0.646, "step": 281010 }, { "epoch": 24.34, "learning_rate": 1.4172545391465247e-05, "loss": 0.7051, "step": 281020 }, { "epoch": 24.34, "learning_rate": 1.417037700936741e-05, "loss": 0.6163, "step": 281030 }, { "epoch": 24.34, "learning_rate": 1.4168208627269571e-05, "loss": 0.4457, "step": 281040 }, { "epoch": 24.34, "learning_rate": 1.4166040245171736e-05, "loss": 0.6825, "step": 281050 }, { "epoch": 24.34, "learning_rate": 1.4163871863073897e-05, "loss": 0.7219, "step": 281060 }, { "epoch": 24.34, "learning_rate": 1.416170348097606e-05, "loss": 0.4856, "step": 281070 }, { "epoch": 24.34, "learning_rate": 1.4159535098878221e-05, "loss": 0.5268, "step": 281080 }, { "epoch": 24.34, "learning_rate": 1.4157366716780386e-05, "loss": 0.78, "step": 281090 }, { "epoch": 24.35, "learning_rate": 1.4155198334682547e-05, "loss": 0.8114, "step": 281100 }, { "epoch": 24.35, "learning_rate": 1.415302995258471e-05, "loss": 0.6143, "step": 281110 }, { "epoch": 24.35, "learning_rate": 1.4150861570486873e-05, "loss": 0.6635, "step": 281120 }, { "epoch": 24.35, "learning_rate": 1.4148693188389036e-05, "loss": 0.651, "step": 281130 }, { "epoch": 24.35, "learning_rate": 1.4146524806291198e-05, "loss": 0.6968, "step": 281140 }, { "epoch": 24.35, "learning_rate": 1.414435642419336e-05, "loss": 0.6427, "step": 281150 }, { "epoch": 24.35, "learning_rate": 1.4142188042095524e-05, "loss": 0.5955, "step": 281160 }, { "epoch": 24.35, "learning_rate": 1.4140019659997687e-05, "loss": 0.6404, "step": 281170 }, { "epoch": 24.35, "learning_rate": 1.4137851277899848e-05, "loss": 0.6784, "step": 281180 }, { "epoch": 24.35, "learning_rate": 1.4135682895802013e-05, "loss": 0.5116, "step": 281190 }, { "epoch": 24.35, "learning_rate": 1.4133514513704174e-05, "loss": 0.6989, "step": 281200 }, { "epoch": 24.35, "learning_rate": 1.4131346131606335e-05, "loss": 0.6079, "step": 281210 }, { "epoch": 24.36, "learning_rate": 1.4129177749508498e-05, "loss": 0.6407, "step": 281220 }, { "epoch": 24.36, "learning_rate": 1.4127009367410661e-05, "loss": 0.6871, "step": 281230 }, { "epoch": 24.36, "learning_rate": 1.4124840985312824e-05, "loss": 0.5452, "step": 281240 }, { "epoch": 24.36, "learning_rate": 1.4122672603214986e-05, "loss": 0.7006, "step": 281250 }, { "epoch": 24.36, "learning_rate": 1.412050422111715e-05, "loss": 0.5202, "step": 281260 }, { "epoch": 24.36, "learning_rate": 1.4118335839019312e-05, "loss": 0.5736, "step": 281270 }, { "epoch": 24.36, "learning_rate": 1.4116167456921475e-05, "loss": 0.6052, "step": 281280 }, { "epoch": 24.36, "learning_rate": 1.4113999074823636e-05, "loss": 0.6921, "step": 281290 }, { "epoch": 24.36, "learning_rate": 1.41118306927258e-05, "loss": 0.6124, "step": 281300 }, { "epoch": 24.36, "learning_rate": 1.4109662310627962e-05, "loss": 0.6344, "step": 281310 }, { "epoch": 24.36, "learning_rate": 1.4107493928530125e-05, "loss": 0.7567, "step": 281320 }, { "epoch": 24.36, "learning_rate": 1.4105325546432286e-05, "loss": 0.6192, "step": 281330 }, { "epoch": 24.37, "learning_rate": 1.4103157164334451e-05, "loss": 0.6789, "step": 281340 }, { "epoch": 24.37, "learning_rate": 1.4100988782236612e-05, "loss": 0.6909, "step": 281350 }, { "epoch": 24.37, "learning_rate": 1.4098820400138775e-05, "loss": 0.6512, "step": 281360 }, { "epoch": 24.37, "learning_rate": 1.4096652018040938e-05, "loss": 0.7248, "step": 281370 }, { "epoch": 24.37, "learning_rate": 1.4094483635943101e-05, "loss": 0.6015, "step": 281380 }, { "epoch": 24.37, "learning_rate": 1.4092315253845263e-05, "loss": 0.5922, "step": 281390 }, { "epoch": 24.37, "learning_rate": 1.4090146871747427e-05, "loss": 0.6908, "step": 281400 }, { "epoch": 24.37, "learning_rate": 1.4087978489649589e-05, "loss": 0.6083, "step": 281410 }, { "epoch": 24.37, "learning_rate": 1.4085810107551752e-05, "loss": 0.5874, "step": 281420 }, { "epoch": 24.37, "learning_rate": 1.4083641725453913e-05, "loss": 0.5896, "step": 281430 }, { "epoch": 24.37, "learning_rate": 1.4081473343356078e-05, "loss": 0.616, "step": 281440 }, { "epoch": 24.38, "learning_rate": 1.4079304961258239e-05, "loss": 0.6363, "step": 281450 }, { "epoch": 24.38, "learning_rate": 1.4077136579160402e-05, "loss": 0.6227, "step": 281460 }, { "epoch": 24.38, "learning_rate": 1.4074968197062563e-05, "loss": 0.6249, "step": 281470 }, { "epoch": 24.38, "learning_rate": 1.4072799814964726e-05, "loss": 0.5662, "step": 281480 }, { "epoch": 24.38, "learning_rate": 1.407063143286689e-05, "loss": 0.703, "step": 281490 }, { "epoch": 24.38, "learning_rate": 1.406846305076905e-05, "loss": 0.5832, "step": 281500 }, { "epoch": 24.38, "learning_rate": 1.4066294668671215e-05, "loss": 0.6796, "step": 281510 }, { "epoch": 24.38, "learning_rate": 1.4064126286573377e-05, "loss": 0.6649, "step": 281520 }, { "epoch": 24.38, "learning_rate": 1.406195790447554e-05, "loss": 0.5776, "step": 281530 }, { "epoch": 24.38, "learning_rate": 1.40597895223777e-05, "loss": 0.6858, "step": 281540 }, { "epoch": 24.38, "learning_rate": 1.4057621140279866e-05, "loss": 0.8198, "step": 281550 }, { "epoch": 24.38, "learning_rate": 1.4055452758182027e-05, "loss": 0.6404, "step": 281560 }, { "epoch": 24.39, "learning_rate": 1.405328437608419e-05, "loss": 0.8113, "step": 281570 }, { "epoch": 24.39, "learning_rate": 1.4051115993986351e-05, "loss": 0.6355, "step": 281580 }, { "epoch": 24.39, "learning_rate": 1.4048947611888516e-05, "loss": 0.5278, "step": 281590 }, { "epoch": 24.39, "learning_rate": 1.4046779229790677e-05, "loss": 0.5507, "step": 281600 }, { "epoch": 24.39, "learning_rate": 1.404461084769284e-05, "loss": 0.5289, "step": 281610 }, { "epoch": 24.39, "learning_rate": 1.4042442465595003e-05, "loss": 0.6298, "step": 281620 }, { "epoch": 24.39, "learning_rate": 1.4040274083497166e-05, "loss": 0.715, "step": 281630 }, { "epoch": 24.39, "learning_rate": 1.4038105701399327e-05, "loss": 0.679, "step": 281640 }, { "epoch": 24.39, "learning_rate": 1.4035937319301492e-05, "loss": 0.5117, "step": 281650 }, { "epoch": 24.39, "learning_rate": 1.4033768937203653e-05, "loss": 0.7072, "step": 281660 }, { "epoch": 24.39, "learning_rate": 1.4031600555105816e-05, "loss": 0.6547, "step": 281670 }, { "epoch": 24.4, "learning_rate": 1.4029432173007978e-05, "loss": 0.6768, "step": 281680 }, { "epoch": 24.4, "learning_rate": 1.4027263790910142e-05, "loss": 0.6682, "step": 281690 }, { "epoch": 24.4, "learning_rate": 1.4025095408812304e-05, "loss": 0.5251, "step": 281700 }, { "epoch": 24.4, "learning_rate": 1.4022927026714467e-05, "loss": 0.6782, "step": 281710 }, { "epoch": 24.4, "learning_rate": 1.4020758644616628e-05, "loss": 0.6204, "step": 281720 }, { "epoch": 24.4, "learning_rate": 1.4018590262518791e-05, "loss": 0.5505, "step": 281730 }, { "epoch": 24.4, "learning_rate": 1.4016421880420954e-05, "loss": 0.5256, "step": 281740 }, { "epoch": 24.4, "learning_rate": 1.4014253498323115e-05, "loss": 0.6682, "step": 281750 }, { "epoch": 24.4, "learning_rate": 1.401208511622528e-05, "loss": 0.6631, "step": 281760 }, { "epoch": 24.4, "learning_rate": 1.4009916734127441e-05, "loss": 0.6469, "step": 281770 }, { "epoch": 24.4, "learning_rate": 1.4007748352029604e-05, "loss": 0.7173, "step": 281780 }, { "epoch": 24.4, "learning_rate": 1.4005579969931766e-05, "loss": 0.5045, "step": 281790 }, { "epoch": 24.41, "learning_rate": 1.400341158783393e-05, "loss": 0.6115, "step": 281800 }, { "epoch": 24.41, "learning_rate": 1.4001243205736092e-05, "loss": 0.5716, "step": 281810 }, { "epoch": 24.41, "learning_rate": 1.3999074823638255e-05, "loss": 0.6774, "step": 281820 }, { "epoch": 24.41, "learning_rate": 1.3996906441540416e-05, "loss": 0.7905, "step": 281830 }, { "epoch": 24.41, "learning_rate": 1.399473805944258e-05, "loss": 0.5064, "step": 281840 }, { "epoch": 24.41, "learning_rate": 1.3992569677344742e-05, "loss": 0.5533, "step": 281850 }, { "epoch": 24.41, "learning_rate": 1.3990401295246905e-05, "loss": 0.5738, "step": 281860 }, { "epoch": 24.41, "learning_rate": 1.3988232913149068e-05, "loss": 0.6149, "step": 281870 }, { "epoch": 24.41, "learning_rate": 1.3986064531051231e-05, "loss": 0.6979, "step": 281880 }, { "epoch": 24.41, "learning_rate": 1.3983896148953392e-05, "loss": 0.5599, "step": 281890 }, { "epoch": 24.41, "learning_rate": 1.3981727766855557e-05, "loss": 0.5508, "step": 281900 }, { "epoch": 24.42, "learning_rate": 1.3979559384757718e-05, "loss": 0.7407, "step": 281910 }, { "epoch": 24.42, "learning_rate": 1.3977391002659881e-05, "loss": 0.6631, "step": 281920 }, { "epoch": 24.42, "learning_rate": 1.3975222620562043e-05, "loss": 0.7169, "step": 281930 }, { "epoch": 24.42, "learning_rate": 1.3973054238464207e-05, "loss": 0.644, "step": 281940 }, { "epoch": 24.42, "learning_rate": 1.3970885856366369e-05, "loss": 0.5393, "step": 281950 }, { "epoch": 24.42, "learning_rate": 1.3968717474268532e-05, "loss": 0.6958, "step": 281960 }, { "epoch": 24.42, "learning_rate": 1.3966549092170693e-05, "loss": 0.6509, "step": 281970 }, { "epoch": 24.42, "learning_rate": 1.3964380710072856e-05, "loss": 0.6819, "step": 281980 }, { "epoch": 24.42, "learning_rate": 1.3962212327975019e-05, "loss": 0.6752, "step": 281990 }, { "epoch": 24.42, "learning_rate": 1.396004394587718e-05, "loss": 0.6368, "step": 282000 }, { "epoch": 24.42, "learning_rate": 1.3957875563779345e-05, "loss": 0.5899, "step": 282010 }, { "epoch": 24.42, "learning_rate": 1.3955707181681506e-05, "loss": 0.6479, "step": 282020 }, { "epoch": 24.43, "learning_rate": 1.395353879958367e-05, "loss": 0.7582, "step": 282030 }, { "epoch": 24.43, "learning_rate": 1.395137041748583e-05, "loss": 0.6793, "step": 282040 }, { "epoch": 24.43, "learning_rate": 1.3949202035387995e-05, "loss": 0.6556, "step": 282050 }, { "epoch": 24.43, "learning_rate": 1.3947033653290157e-05, "loss": 0.5988, "step": 282060 }, { "epoch": 24.43, "learning_rate": 1.394486527119232e-05, "loss": 0.6039, "step": 282070 }, { "epoch": 24.43, "learning_rate": 1.3942696889094481e-05, "loss": 0.674, "step": 282080 }, { "epoch": 24.43, "learning_rate": 1.3940528506996646e-05, "loss": 0.6352, "step": 282090 }, { "epoch": 24.43, "learning_rate": 1.3938360124898807e-05, "loss": 0.572, "step": 282100 }, { "epoch": 24.43, "learning_rate": 1.393619174280097e-05, "loss": 0.5663, "step": 282110 }, { "epoch": 24.43, "learning_rate": 1.3934023360703133e-05, "loss": 0.6273, "step": 282120 }, { "epoch": 24.43, "learning_rate": 1.3931854978605296e-05, "loss": 0.5823, "step": 282130 }, { "epoch": 24.44, "learning_rate": 1.3929686596507457e-05, "loss": 0.7892, "step": 282140 }, { "epoch": 24.44, "learning_rate": 1.3927518214409622e-05, "loss": 0.5225, "step": 282150 }, { "epoch": 24.44, "learning_rate": 1.3925349832311783e-05, "loss": 0.5835, "step": 282160 }, { "epoch": 24.44, "learning_rate": 1.3923181450213946e-05, "loss": 0.6804, "step": 282170 }, { "epoch": 24.44, "learning_rate": 1.3921013068116108e-05, "loss": 0.7107, "step": 282180 }, { "epoch": 24.44, "learning_rate": 1.3918844686018272e-05, "loss": 0.6744, "step": 282190 }, { "epoch": 24.44, "learning_rate": 1.3916676303920434e-05, "loss": 0.4791, "step": 282200 }, { "epoch": 24.44, "learning_rate": 1.3914507921822597e-05, "loss": 0.5683, "step": 282210 }, { "epoch": 24.44, "learning_rate": 1.3912339539724758e-05, "loss": 0.733, "step": 282220 }, { "epoch": 24.44, "learning_rate": 1.3910171157626923e-05, "loss": 0.6253, "step": 282230 }, { "epoch": 24.44, "learning_rate": 1.3908002775529084e-05, "loss": 0.5866, "step": 282240 }, { "epoch": 24.44, "learning_rate": 1.3905834393431245e-05, "loss": 0.5905, "step": 282250 }, { "epoch": 24.45, "learning_rate": 1.390366601133341e-05, "loss": 0.3932, "step": 282260 }, { "epoch": 24.45, "learning_rate": 1.3901497629235571e-05, "loss": 0.6034, "step": 282270 }, { "epoch": 24.45, "learning_rate": 1.3899329247137734e-05, "loss": 0.6399, "step": 282280 }, { "epoch": 24.45, "learning_rate": 1.3897160865039896e-05, "loss": 0.6178, "step": 282290 }, { "epoch": 24.45, "learning_rate": 1.389499248294206e-05, "loss": 0.7673, "step": 282300 }, { "epoch": 24.45, "learning_rate": 1.3892824100844222e-05, "loss": 0.5312, "step": 282310 }, { "epoch": 24.45, "learning_rate": 1.3890655718746385e-05, "loss": 0.6536, "step": 282320 }, { "epoch": 24.45, "learning_rate": 1.3888487336648546e-05, "loss": 0.6438, "step": 282330 }, { "epoch": 24.45, "learning_rate": 1.388631895455071e-05, "loss": 0.6222, "step": 282340 }, { "epoch": 24.45, "learning_rate": 1.3884150572452872e-05, "loss": 0.5576, "step": 282350 }, { "epoch": 24.45, "learning_rate": 1.3881982190355035e-05, "loss": 0.7027, "step": 282360 }, { "epoch": 24.46, "learning_rate": 1.3879813808257198e-05, "loss": 0.5566, "step": 282370 }, { "epoch": 24.46, "learning_rate": 1.3877645426159361e-05, "loss": 0.6613, "step": 282380 }, { "epoch": 24.46, "learning_rate": 1.3875477044061522e-05, "loss": 0.609, "step": 282390 }, { "epoch": 24.46, "learning_rate": 1.3873308661963687e-05, "loss": 0.7082, "step": 282400 }, { "epoch": 24.46, "learning_rate": 1.3871140279865848e-05, "loss": 0.5297, "step": 282410 }, { "epoch": 24.46, "learning_rate": 1.3868971897768011e-05, "loss": 0.5691, "step": 282420 }, { "epoch": 24.46, "learning_rate": 1.3866803515670173e-05, "loss": 0.6153, "step": 282430 }, { "epoch": 24.46, "learning_rate": 1.3864635133572337e-05, "loss": 0.6603, "step": 282440 }, { "epoch": 24.46, "learning_rate": 1.3862466751474499e-05, "loss": 0.6808, "step": 282450 }, { "epoch": 24.46, "learning_rate": 1.3860298369376662e-05, "loss": 0.461, "step": 282460 }, { "epoch": 24.46, "learning_rate": 1.3858129987278823e-05, "loss": 0.5695, "step": 282470 }, { "epoch": 24.46, "learning_rate": 1.3855961605180988e-05, "loss": 0.7353, "step": 282480 }, { "epoch": 24.47, "learning_rate": 1.3853793223083149e-05, "loss": 0.6537, "step": 282490 }, { "epoch": 24.47, "learning_rate": 1.385162484098531e-05, "loss": 0.6168, "step": 282500 }, { "epoch": 24.47, "learning_rate": 1.3849456458887475e-05, "loss": 0.5586, "step": 282510 }, { "epoch": 24.47, "learning_rate": 1.3847288076789636e-05, "loss": 0.8225, "step": 282520 }, { "epoch": 24.47, "learning_rate": 1.38451196946918e-05, "loss": 0.5841, "step": 282530 }, { "epoch": 24.47, "learning_rate": 1.384295131259396e-05, "loss": 0.7272, "step": 282540 }, { "epoch": 24.47, "learning_rate": 1.3840782930496125e-05, "loss": 0.6548, "step": 282550 }, { "epoch": 24.47, "learning_rate": 1.3838614548398287e-05, "loss": 0.6461, "step": 282560 }, { "epoch": 24.47, "learning_rate": 1.383644616630045e-05, "loss": 0.5997, "step": 282570 }, { "epoch": 24.47, "learning_rate": 1.3834277784202613e-05, "loss": 0.5278, "step": 282580 }, { "epoch": 24.47, "learning_rate": 1.3832109402104776e-05, "loss": 0.5803, "step": 282590 }, { "epoch": 24.47, "learning_rate": 1.3829941020006937e-05, "loss": 0.6543, "step": 282600 }, { "epoch": 24.48, "learning_rate": 1.38277726379091e-05, "loss": 0.6317, "step": 282610 }, { "epoch": 24.48, "learning_rate": 1.3825604255811263e-05, "loss": 0.7185, "step": 282620 }, { "epoch": 24.48, "learning_rate": 1.3823435873713426e-05, "loss": 0.6101, "step": 282630 }, { "epoch": 24.48, "learning_rate": 1.3821267491615587e-05, "loss": 0.6393, "step": 282640 }, { "epoch": 24.48, "learning_rate": 1.3819099109517752e-05, "loss": 0.6518, "step": 282650 }, { "epoch": 24.48, "learning_rate": 1.3816930727419913e-05, "loss": 0.5614, "step": 282660 }, { "epoch": 24.48, "learning_rate": 1.3814762345322076e-05, "loss": 0.6305, "step": 282670 }, { "epoch": 24.48, "learning_rate": 1.3812593963224237e-05, "loss": 0.6366, "step": 282680 }, { "epoch": 24.48, "learning_rate": 1.3810425581126402e-05, "loss": 0.6135, "step": 282690 }, { "epoch": 24.48, "learning_rate": 1.3808257199028563e-05, "loss": 0.7309, "step": 282700 }, { "epoch": 24.48, "learning_rate": 1.3806088816930727e-05, "loss": 0.6402, "step": 282710 }, { "epoch": 24.49, "learning_rate": 1.380392043483289e-05, "loss": 0.6828, "step": 282720 }, { "epoch": 24.49, "learning_rate": 1.3801752052735053e-05, "loss": 0.6946, "step": 282730 }, { "epoch": 24.49, "learning_rate": 1.3799583670637214e-05, "loss": 0.6521, "step": 282740 }, { "epoch": 24.49, "learning_rate": 1.3797415288539375e-05, "loss": 0.4704, "step": 282750 }, { "epoch": 24.49, "learning_rate": 1.379524690644154e-05, "loss": 0.6915, "step": 282760 }, { "epoch": 24.49, "learning_rate": 1.3793078524343701e-05, "loss": 0.6747, "step": 282770 }, { "epoch": 24.49, "learning_rate": 1.3790910142245864e-05, "loss": 0.5373, "step": 282780 }, { "epoch": 24.49, "learning_rate": 1.3788741760148025e-05, "loss": 0.7257, "step": 282790 }, { "epoch": 24.49, "learning_rate": 1.378657337805019e-05, "loss": 0.658, "step": 282800 }, { "epoch": 24.49, "learning_rate": 1.3784404995952351e-05, "loss": 0.6701, "step": 282810 }, { "epoch": 24.49, "learning_rate": 1.3782236613854514e-05, "loss": 0.6458, "step": 282820 }, { "epoch": 24.49, "learning_rate": 1.3780068231756677e-05, "loss": 0.592, "step": 282830 }, { "epoch": 24.5, "learning_rate": 1.377789984965884e-05, "loss": 0.6439, "step": 282840 }, { "epoch": 24.5, "learning_rate": 1.3775731467561002e-05, "loss": 0.5888, "step": 282850 }, { "epoch": 24.5, "learning_rate": 1.3773563085463166e-05, "loss": 0.5992, "step": 282860 }, { "epoch": 24.5, "learning_rate": 1.3771394703365328e-05, "loss": 0.5966, "step": 282870 }, { "epoch": 24.5, "learning_rate": 1.376922632126749e-05, "loss": 0.5231, "step": 282880 }, { "epoch": 24.5, "learning_rate": 1.3767057939169652e-05, "loss": 0.6152, "step": 282890 }, { "epoch": 24.5, "learning_rate": 1.3764889557071817e-05, "loss": 0.7058, "step": 282900 }, { "epoch": 24.5, "learning_rate": 1.3762721174973978e-05, "loss": 0.5992, "step": 282910 }, { "epoch": 24.5, "learning_rate": 1.3760552792876141e-05, "loss": 0.5638, "step": 282920 }, { "epoch": 24.5, "learning_rate": 1.3758384410778302e-05, "loss": 0.5767, "step": 282930 }, { "epoch": 24.5, "learning_rate": 1.3756216028680467e-05, "loss": 0.6016, "step": 282940 }, { "epoch": 24.51, "learning_rate": 1.3754047646582628e-05, "loss": 0.6541, "step": 282950 }, { "epoch": 24.51, "learning_rate": 1.3751879264484791e-05, "loss": 0.6299, "step": 282960 }, { "epoch": 24.51, "learning_rate": 1.3749710882386954e-05, "loss": 0.6426, "step": 282970 }, { "epoch": 24.51, "learning_rate": 1.3747542500289117e-05, "loss": 0.6631, "step": 282980 }, { "epoch": 24.51, "learning_rate": 1.3745374118191279e-05, "loss": 0.6575, "step": 282990 }, { "epoch": 24.51, "learning_rate": 1.3743205736093443e-05, "loss": 0.574, "step": 283000 }, { "epoch": 24.51, "learning_rate": 1.3741037353995605e-05, "loss": 0.5947, "step": 283010 }, { "epoch": 24.51, "learning_rate": 1.3738868971897766e-05, "loss": 0.5961, "step": 283020 }, { "epoch": 24.51, "learning_rate": 1.3736700589799929e-05, "loss": 0.5545, "step": 283030 }, { "epoch": 24.51, "learning_rate": 1.373453220770209e-05, "loss": 0.6243, "step": 283040 }, { "epoch": 24.51, "learning_rate": 1.3732363825604255e-05, "loss": 0.6801, "step": 283050 }, { "epoch": 24.51, "learning_rate": 1.3730195443506416e-05, "loss": 0.5564, "step": 283060 }, { "epoch": 24.52, "learning_rate": 1.372802706140858e-05, "loss": 0.5949, "step": 283070 }, { "epoch": 24.52, "learning_rate": 1.3725858679310742e-05, "loss": 0.5857, "step": 283080 }, { "epoch": 24.52, "learning_rate": 1.3723690297212905e-05, "loss": 0.8042, "step": 283090 }, { "epoch": 24.52, "learning_rate": 1.3721521915115067e-05, "loss": 0.7231, "step": 283100 }, { "epoch": 24.52, "learning_rate": 1.3719353533017231e-05, "loss": 0.5462, "step": 283110 }, { "epoch": 24.52, "learning_rate": 1.3717185150919393e-05, "loss": 0.6132, "step": 283120 }, { "epoch": 24.52, "learning_rate": 1.3715016768821556e-05, "loss": 0.6072, "step": 283130 }, { "epoch": 24.52, "learning_rate": 1.3712848386723717e-05, "loss": 0.6147, "step": 283140 }, { "epoch": 24.52, "learning_rate": 1.3710680004625882e-05, "loss": 0.6208, "step": 283150 }, { "epoch": 24.52, "learning_rate": 1.3708511622528043e-05, "loss": 0.7145, "step": 283160 }, { "epoch": 24.52, "learning_rate": 1.3706343240430206e-05, "loss": 0.7245, "step": 283170 }, { "epoch": 24.53, "learning_rate": 1.3704174858332367e-05, "loss": 0.7264, "step": 283180 }, { "epoch": 24.53, "learning_rate": 1.3702006476234532e-05, "loss": 0.7461, "step": 283190 }, { "epoch": 24.53, "learning_rate": 1.3699838094136693e-05, "loss": 0.6722, "step": 283200 }, { "epoch": 24.53, "learning_rate": 1.3697669712038856e-05, "loss": 0.5553, "step": 283210 }, { "epoch": 24.53, "learning_rate": 1.369550132994102e-05, "loss": 0.6193, "step": 283220 }, { "epoch": 24.53, "learning_rate": 1.3693332947843182e-05, "loss": 0.5279, "step": 283230 }, { "epoch": 24.53, "learning_rate": 1.3691164565745344e-05, "loss": 0.6772, "step": 283240 }, { "epoch": 24.53, "learning_rate": 1.3688996183647508e-05, "loss": 0.7564, "step": 283250 }, { "epoch": 24.53, "learning_rate": 1.368682780154967e-05, "loss": 0.7266, "step": 283260 }, { "epoch": 24.53, "learning_rate": 1.3684659419451831e-05, "loss": 0.7452, "step": 283270 }, { "epoch": 24.53, "learning_rate": 1.3682491037353994e-05, "loss": 0.6347, "step": 283280 }, { "epoch": 24.53, "learning_rate": 1.3680322655256155e-05, "loss": 0.7091, "step": 283290 }, { "epoch": 24.54, "learning_rate": 1.367815427315832e-05, "loss": 0.6213, "step": 283300 }, { "epoch": 24.54, "learning_rate": 1.3675985891060481e-05, "loss": 0.5396, "step": 283310 }, { "epoch": 24.54, "learning_rate": 1.3673817508962644e-05, "loss": 0.7024, "step": 283320 }, { "epoch": 24.54, "learning_rate": 1.3671649126864807e-05, "loss": 0.6219, "step": 283330 }, { "epoch": 24.54, "learning_rate": 1.366948074476697e-05, "loss": 0.5411, "step": 283340 }, { "epoch": 24.54, "learning_rate": 1.3667312362669132e-05, "loss": 0.7533, "step": 283350 }, { "epoch": 24.54, "learning_rate": 1.3665143980571296e-05, "loss": 0.6681, "step": 283360 }, { "epoch": 24.54, "learning_rate": 1.3662975598473458e-05, "loss": 0.6674, "step": 283370 }, { "epoch": 24.54, "learning_rate": 1.366080721637562e-05, "loss": 0.7099, "step": 283380 }, { "epoch": 24.54, "learning_rate": 1.3658638834277782e-05, "loss": 0.6109, "step": 283390 }, { "epoch": 24.54, "learning_rate": 1.3656470452179947e-05, "loss": 0.6817, "step": 283400 }, { "epoch": 24.55, "learning_rate": 1.3654302070082108e-05, "loss": 0.6838, "step": 283410 }, { "epoch": 24.55, "learning_rate": 1.3652133687984271e-05, "loss": 0.7127, "step": 283420 }, { "epoch": 24.55, "learning_rate": 1.3649965305886432e-05, "loss": 0.5636, "step": 283430 }, { "epoch": 24.55, "learning_rate": 1.3647796923788597e-05, "loss": 0.6719, "step": 283440 }, { "epoch": 24.55, "learning_rate": 1.3645628541690758e-05, "loss": 0.6008, "step": 283450 }, { "epoch": 24.55, "learning_rate": 1.3643460159592921e-05, "loss": 0.653, "step": 283460 }, { "epoch": 24.55, "learning_rate": 1.3641291777495084e-05, "loss": 0.5187, "step": 283470 }, { "epoch": 24.55, "learning_rate": 1.3639123395397247e-05, "loss": 0.596, "step": 283480 }, { "epoch": 24.55, "learning_rate": 1.3636955013299409e-05, "loss": 0.7868, "step": 283490 }, { "epoch": 24.55, "learning_rate": 1.3634786631201573e-05, "loss": 0.6043, "step": 283500 }, { "epoch": 24.55, "learning_rate": 1.3632618249103735e-05, "loss": 0.782, "step": 283510 }, { "epoch": 24.55, "learning_rate": 1.3630449867005896e-05, "loss": 0.6964, "step": 283520 }, { "epoch": 24.56, "learning_rate": 1.3628281484908059e-05, "loss": 0.4973, "step": 283530 }, { "epoch": 24.56, "learning_rate": 1.362611310281022e-05, "loss": 0.6247, "step": 283540 }, { "epoch": 24.56, "learning_rate": 1.3623944720712385e-05, "loss": 0.5749, "step": 283550 }, { "epoch": 24.56, "learning_rate": 1.3621776338614546e-05, "loss": 0.7382, "step": 283560 }, { "epoch": 24.56, "learning_rate": 1.361960795651671e-05, "loss": 0.6232, "step": 283570 }, { "epoch": 24.56, "learning_rate": 1.3617439574418872e-05, "loss": 0.6742, "step": 283580 }, { "epoch": 24.56, "learning_rate": 1.3615271192321035e-05, "loss": 0.7012, "step": 283590 }, { "epoch": 24.56, "learning_rate": 1.3613102810223197e-05, "loss": 0.5514, "step": 283600 }, { "epoch": 24.56, "learning_rate": 1.3610934428125361e-05, "loss": 0.6478, "step": 283610 }, { "epoch": 24.56, "learning_rate": 1.3608766046027523e-05, "loss": 0.5785, "step": 283620 }, { "epoch": 24.56, "learning_rate": 1.3606597663929686e-05, "loss": 0.7807, "step": 283630 }, { "epoch": 24.57, "learning_rate": 1.3604429281831847e-05, "loss": 0.7341, "step": 283640 }, { "epoch": 24.57, "learning_rate": 1.3602260899734012e-05, "loss": 0.6731, "step": 283650 }, { "epoch": 24.57, "learning_rate": 1.3600092517636173e-05, "loss": 0.7279, "step": 283660 }, { "epoch": 24.57, "learning_rate": 1.3597924135538336e-05, "loss": 0.601, "step": 283670 }, { "epoch": 24.57, "learning_rate": 1.3595755753440497e-05, "loss": 0.8018, "step": 283680 }, { "epoch": 24.57, "learning_rate": 1.3593587371342662e-05, "loss": 0.6272, "step": 283690 }, { "epoch": 24.57, "learning_rate": 1.3591418989244823e-05, "loss": 0.6264, "step": 283700 }, { "epoch": 24.57, "learning_rate": 1.3589250607146986e-05, "loss": 0.6668, "step": 283710 }, { "epoch": 24.57, "learning_rate": 1.358708222504915e-05, "loss": 0.5385, "step": 283720 }, { "epoch": 24.57, "learning_rate": 1.3584913842951312e-05, "loss": 0.6076, "step": 283730 }, { "epoch": 24.57, "learning_rate": 1.3582745460853474e-05, "loss": 0.5983, "step": 283740 }, { "epoch": 24.57, "learning_rate": 1.3580577078755638e-05, "loss": 0.6373, "step": 283750 }, { "epoch": 24.58, "learning_rate": 1.35784086966578e-05, "loss": 0.7825, "step": 283760 }, { "epoch": 24.58, "learning_rate": 1.3576240314559963e-05, "loss": 0.5798, "step": 283770 }, { "epoch": 24.58, "learning_rate": 1.3574071932462124e-05, "loss": 0.6383, "step": 283780 }, { "epoch": 24.58, "learning_rate": 1.3571903550364287e-05, "loss": 0.6938, "step": 283790 }, { "epoch": 24.58, "learning_rate": 1.356973516826645e-05, "loss": 0.6908, "step": 283800 }, { "epoch": 24.58, "learning_rate": 1.3567566786168611e-05, "loss": 0.5426, "step": 283810 }, { "epoch": 24.58, "learning_rate": 1.3565398404070774e-05, "loss": 0.715, "step": 283820 }, { "epoch": 24.58, "learning_rate": 1.3563230021972937e-05, "loss": 0.6912, "step": 283830 }, { "epoch": 24.58, "learning_rate": 1.35610616398751e-05, "loss": 0.5359, "step": 283840 }, { "epoch": 24.58, "learning_rate": 1.3558893257777261e-05, "loss": 0.5749, "step": 283850 }, { "epoch": 24.58, "learning_rate": 1.3556724875679426e-05, "loss": 0.5587, "step": 283860 }, { "epoch": 24.58, "learning_rate": 1.3554556493581587e-05, "loss": 0.681, "step": 283870 }, { "epoch": 24.59, "learning_rate": 1.355238811148375e-05, "loss": 0.7247, "step": 283880 }, { "epoch": 24.59, "learning_rate": 1.3550219729385912e-05, "loss": 0.7013, "step": 283890 }, { "epoch": 24.59, "learning_rate": 1.3548051347288076e-05, "loss": 0.6303, "step": 283900 }, { "epoch": 24.59, "learning_rate": 1.3545882965190238e-05, "loss": 0.6577, "step": 283910 }, { "epoch": 24.59, "learning_rate": 1.35437145830924e-05, "loss": 0.5686, "step": 283920 }, { "epoch": 24.59, "learning_rate": 1.3541546200994564e-05, "loss": 0.689, "step": 283930 }, { "epoch": 24.59, "learning_rate": 1.3539377818896727e-05, "loss": 0.5661, "step": 283940 }, { "epoch": 24.59, "learning_rate": 1.3537209436798888e-05, "loss": 0.6239, "step": 283950 }, { "epoch": 24.59, "learning_rate": 1.3535041054701051e-05, "loss": 0.6515, "step": 283960 }, { "epoch": 24.59, "learning_rate": 1.3532872672603214e-05, "loss": 0.5732, "step": 283970 }, { "epoch": 24.59, "learning_rate": 1.3530704290505377e-05, "loss": 0.5555, "step": 283980 }, { "epoch": 24.6, "learning_rate": 1.3528535908407538e-05, "loss": 0.7503, "step": 283990 }, { "epoch": 24.6, "learning_rate": 1.3526367526309703e-05, "loss": 0.679, "step": 284000 }, { "epoch": 24.6, "learning_rate": 1.3524199144211864e-05, "loss": 0.679, "step": 284010 }, { "epoch": 24.6, "learning_rate": 1.3522030762114027e-05, "loss": 0.5131, "step": 284020 }, { "epoch": 24.6, "learning_rate": 1.3519862380016189e-05, "loss": 0.7111, "step": 284030 }, { "epoch": 24.6, "learning_rate": 1.3517693997918352e-05, "loss": 0.6575, "step": 284040 }, { "epoch": 24.6, "learning_rate": 1.3515525615820515e-05, "loss": 0.7088, "step": 284050 }, { "epoch": 24.6, "learning_rate": 1.3513357233722676e-05, "loss": 0.6141, "step": 284060 }, { "epoch": 24.6, "learning_rate": 1.351118885162484e-05, "loss": 0.7106, "step": 284070 }, { "epoch": 24.6, "learning_rate": 1.3509020469527002e-05, "loss": 0.5779, "step": 284080 }, { "epoch": 24.6, "learning_rate": 1.3506852087429165e-05, "loss": 0.5588, "step": 284090 }, { "epoch": 24.6, "learning_rate": 1.3504683705331326e-05, "loss": 0.6015, "step": 284100 }, { "epoch": 24.61, "learning_rate": 1.3502515323233491e-05, "loss": 0.5168, "step": 284110 }, { "epoch": 24.61, "learning_rate": 1.3500346941135652e-05, "loss": 0.6005, "step": 284120 }, { "epoch": 24.61, "learning_rate": 1.3498178559037815e-05, "loss": 0.6805, "step": 284130 }, { "epoch": 24.61, "learning_rate": 1.3496010176939977e-05, "loss": 0.6253, "step": 284140 }, { "epoch": 24.61, "learning_rate": 1.3493841794842141e-05, "loss": 0.6827, "step": 284150 }, { "epoch": 24.61, "learning_rate": 1.3491673412744303e-05, "loss": 0.5927, "step": 284160 }, { "epoch": 24.61, "learning_rate": 1.3489505030646466e-05, "loss": 0.6412, "step": 284170 }, { "epoch": 24.61, "learning_rate": 1.3487336648548629e-05, "loss": 0.6352, "step": 284180 }, { "epoch": 24.61, "learning_rate": 1.3485168266450792e-05, "loss": 0.6735, "step": 284190 }, { "epoch": 24.61, "learning_rate": 1.3482999884352953e-05, "loss": 0.5068, "step": 284200 }, { "epoch": 24.61, "learning_rate": 1.3480831502255118e-05, "loss": 0.6568, "step": 284210 }, { "epoch": 24.62, "learning_rate": 1.3478663120157279e-05, "loss": 0.7063, "step": 284220 }, { "epoch": 24.62, "learning_rate": 1.3476494738059442e-05, "loss": 0.6151, "step": 284230 }, { "epoch": 24.62, "learning_rate": 1.3474326355961603e-05, "loss": 0.6855, "step": 284240 }, { "epoch": 24.62, "learning_rate": 1.3472157973863768e-05, "loss": 0.6837, "step": 284250 }, { "epoch": 24.62, "learning_rate": 1.346998959176593e-05, "loss": 0.6689, "step": 284260 }, { "epoch": 24.62, "learning_rate": 1.3467821209668092e-05, "loss": 0.6265, "step": 284270 }, { "epoch": 24.62, "learning_rate": 1.3465652827570254e-05, "loss": 0.5815, "step": 284280 }, { "epoch": 24.62, "learning_rate": 1.3463484445472417e-05, "loss": 0.5967, "step": 284290 }, { "epoch": 24.62, "learning_rate": 1.346131606337458e-05, "loss": 0.5889, "step": 284300 }, { "epoch": 24.62, "learning_rate": 1.3459147681276741e-05, "loss": 0.5575, "step": 284310 }, { "epoch": 24.62, "learning_rate": 1.3456979299178906e-05, "loss": 0.8278, "step": 284320 }, { "epoch": 24.62, "learning_rate": 1.3454810917081067e-05, "loss": 0.676, "step": 284330 }, { "epoch": 24.63, "learning_rate": 1.345264253498323e-05, "loss": 0.6015, "step": 284340 }, { "epoch": 24.63, "learning_rate": 1.3450474152885391e-05, "loss": 0.6432, "step": 284350 }, { "epoch": 24.63, "learning_rate": 1.3448305770787556e-05, "loss": 0.7565, "step": 284360 }, { "epoch": 24.63, "learning_rate": 1.3446137388689717e-05, "loss": 0.5507, "step": 284370 }, { "epoch": 24.63, "learning_rate": 1.344396900659188e-05, "loss": 0.5424, "step": 284380 }, { "epoch": 24.63, "learning_rate": 1.3441800624494042e-05, "loss": 0.6676, "step": 284390 }, { "epoch": 24.63, "learning_rate": 1.3439632242396206e-05, "loss": 0.6925, "step": 284400 }, { "epoch": 24.63, "learning_rate": 1.3437463860298368e-05, "loss": 0.7241, "step": 284410 }, { "epoch": 24.63, "learning_rate": 1.343529547820053e-05, "loss": 0.6322, "step": 284420 }, { "epoch": 24.63, "learning_rate": 1.3433127096102694e-05, "loss": 0.6454, "step": 284430 }, { "epoch": 24.63, "learning_rate": 1.3430958714004857e-05, "loss": 0.6307, "step": 284440 }, { "epoch": 24.64, "learning_rate": 1.3428790331907018e-05, "loss": 0.5592, "step": 284450 }, { "epoch": 24.64, "learning_rate": 1.3426621949809183e-05, "loss": 0.5676, "step": 284460 }, { "epoch": 24.64, "learning_rate": 1.3424453567711344e-05, "loss": 0.7972, "step": 284470 }, { "epoch": 24.64, "learning_rate": 1.3422285185613507e-05, "loss": 0.7385, "step": 284480 }, { "epoch": 24.64, "learning_rate": 1.3420116803515668e-05, "loss": 0.7142, "step": 284490 }, { "epoch": 24.64, "learning_rate": 1.3417948421417833e-05, "loss": 0.6833, "step": 284500 }, { "epoch": 24.64, "learning_rate": 1.3415780039319994e-05, "loss": 0.632, "step": 284510 }, { "epoch": 24.64, "learning_rate": 1.3413611657222157e-05, "loss": 0.5625, "step": 284520 }, { "epoch": 24.64, "learning_rate": 1.3411443275124319e-05, "loss": 0.6054, "step": 284530 }, { "epoch": 24.64, "learning_rate": 1.3409274893026483e-05, "loss": 0.5763, "step": 284540 }, { "epoch": 24.64, "learning_rate": 1.3407106510928645e-05, "loss": 0.5803, "step": 284550 }, { "epoch": 24.64, "learning_rate": 1.3404938128830806e-05, "loss": 0.7384, "step": 284560 }, { "epoch": 24.65, "learning_rate": 1.340276974673297e-05, "loss": 0.7303, "step": 284570 }, { "epoch": 24.65, "learning_rate": 1.3400601364635132e-05, "loss": 0.6648, "step": 284580 }, { "epoch": 24.65, "learning_rate": 1.3398432982537295e-05, "loss": 0.6765, "step": 284590 }, { "epoch": 24.65, "learning_rate": 1.3396264600439456e-05, "loss": 0.7063, "step": 284600 }, { "epoch": 24.65, "learning_rate": 1.3394096218341621e-05, "loss": 0.63, "step": 284610 }, { "epoch": 24.65, "learning_rate": 1.3391927836243782e-05, "loss": 0.5992, "step": 284620 }, { "epoch": 24.65, "learning_rate": 1.3389759454145945e-05, "loss": 0.7079, "step": 284630 }, { "epoch": 24.65, "learning_rate": 1.3387591072048107e-05, "loss": 0.6826, "step": 284640 }, { "epoch": 24.65, "learning_rate": 1.3385422689950271e-05, "loss": 0.6377, "step": 284650 }, { "epoch": 24.65, "learning_rate": 1.3383254307852433e-05, "loss": 0.7046, "step": 284660 }, { "epoch": 24.65, "learning_rate": 1.3381085925754596e-05, "loss": 0.582, "step": 284670 }, { "epoch": 24.66, "learning_rate": 1.3378917543656759e-05, "loss": 0.7547, "step": 284680 }, { "epoch": 24.66, "learning_rate": 1.3376749161558922e-05, "loss": 0.5997, "step": 284690 }, { "epoch": 24.66, "learning_rate": 1.3374580779461083e-05, "loss": 0.6382, "step": 284700 }, { "epoch": 24.66, "learning_rate": 1.3372412397363248e-05, "loss": 0.6864, "step": 284710 }, { "epoch": 24.66, "learning_rate": 1.3370244015265409e-05, "loss": 0.6788, "step": 284720 }, { "epoch": 24.66, "learning_rate": 1.3368075633167572e-05, "loss": 0.526, "step": 284730 }, { "epoch": 24.66, "learning_rate": 1.3365907251069733e-05, "loss": 0.6745, "step": 284740 }, { "epoch": 24.66, "learning_rate": 1.3363738868971898e-05, "loss": 0.582, "step": 284750 }, { "epoch": 24.66, "learning_rate": 1.336157048687406e-05, "loss": 0.6116, "step": 284760 }, { "epoch": 24.66, "learning_rate": 1.3359402104776222e-05, "loss": 0.6647, "step": 284770 }, { "epoch": 24.66, "learning_rate": 1.3357233722678384e-05, "loss": 0.7105, "step": 284780 }, { "epoch": 24.66, "learning_rate": 1.3355065340580548e-05, "loss": 0.7464, "step": 284790 }, { "epoch": 24.67, "learning_rate": 1.335289695848271e-05, "loss": 0.6867, "step": 284800 }, { "epoch": 24.67, "learning_rate": 1.335072857638487e-05, "loss": 0.6707, "step": 284810 }, { "epoch": 24.67, "learning_rate": 1.3348560194287036e-05, "loss": 0.6071, "step": 284820 }, { "epoch": 24.67, "learning_rate": 1.3346391812189197e-05, "loss": 0.6385, "step": 284830 }, { "epoch": 24.67, "learning_rate": 1.334422343009136e-05, "loss": 0.6827, "step": 284840 }, { "epoch": 24.67, "learning_rate": 1.3342055047993521e-05, "loss": 0.6078, "step": 284850 }, { "epoch": 24.67, "learning_rate": 1.3339886665895686e-05, "loss": 0.6215, "step": 284860 }, { "epoch": 24.67, "learning_rate": 1.3337718283797847e-05, "loss": 0.5914, "step": 284870 }, { "epoch": 24.67, "learning_rate": 1.333554990170001e-05, "loss": 0.6182, "step": 284880 }, { "epoch": 24.67, "learning_rate": 1.3333381519602171e-05, "loss": 0.6075, "step": 284890 }, { "epoch": 24.67, "learning_rate": 1.3331213137504336e-05, "loss": 0.5223, "step": 284900 }, { "epoch": 24.68, "learning_rate": 1.3329044755406497e-05, "loss": 0.703, "step": 284910 }, { "epoch": 24.68, "learning_rate": 1.332687637330866e-05, "loss": 0.6834, "step": 284920 }, { "epoch": 24.68, "learning_rate": 1.3324707991210823e-05, "loss": 0.629, "step": 284930 }, { "epoch": 24.68, "learning_rate": 1.3322539609112986e-05, "loss": 0.624, "step": 284940 }, { "epoch": 24.68, "learning_rate": 1.3320371227015148e-05, "loss": 0.5869, "step": 284950 }, { "epoch": 24.68, "learning_rate": 1.3318202844917313e-05, "loss": 0.6037, "step": 284960 }, { "epoch": 24.68, "learning_rate": 1.3316034462819474e-05, "loss": 0.781, "step": 284970 }, { "epoch": 24.68, "learning_rate": 1.3313866080721637e-05, "loss": 0.6468, "step": 284980 }, { "epoch": 24.68, "learning_rate": 1.3311697698623798e-05, "loss": 0.7038, "step": 284990 }, { "epoch": 24.68, "learning_rate": 1.3309529316525963e-05, "loss": 0.5204, "step": 285000 }, { "epoch": 24.68, "learning_rate": 1.3307360934428124e-05, "loss": 0.7995, "step": 285010 }, { "epoch": 24.68, "learning_rate": 1.3305192552330287e-05, "loss": 0.6097, "step": 285020 }, { "epoch": 24.69, "learning_rate": 1.3303024170232448e-05, "loss": 0.6408, "step": 285030 }, { "epoch": 24.69, "learning_rate": 1.3300855788134613e-05, "loss": 0.6283, "step": 285040 }, { "epoch": 24.69, "learning_rate": 1.3298687406036774e-05, "loss": 0.6088, "step": 285050 }, { "epoch": 24.69, "learning_rate": 1.3296519023938936e-05, "loss": 0.6531, "step": 285060 }, { "epoch": 24.69, "learning_rate": 1.32943506418411e-05, "loss": 0.7615, "step": 285070 }, { "epoch": 24.69, "learning_rate": 1.3292182259743262e-05, "loss": 0.6018, "step": 285080 }, { "epoch": 24.69, "learning_rate": 1.3290013877645425e-05, "loss": 0.6047, "step": 285090 }, { "epoch": 24.69, "learning_rate": 1.3287845495547586e-05, "loss": 0.7041, "step": 285100 }, { "epoch": 24.69, "learning_rate": 1.328567711344975e-05, "loss": 0.6076, "step": 285110 }, { "epoch": 24.69, "learning_rate": 1.3283508731351912e-05, "loss": 0.706, "step": 285120 }, { "epoch": 24.69, "learning_rate": 1.3281340349254075e-05, "loss": 0.4866, "step": 285130 }, { "epoch": 24.69, "learning_rate": 1.3279171967156238e-05, "loss": 0.7074, "step": 285140 }, { "epoch": 24.7, "learning_rate": 1.3277003585058401e-05, "loss": 0.5249, "step": 285150 }, { "epoch": 24.7, "learning_rate": 1.3274835202960562e-05, "loss": 0.6516, "step": 285160 }, { "epoch": 24.7, "learning_rate": 1.3272666820862725e-05, "loss": 0.6482, "step": 285170 }, { "epoch": 24.7, "learning_rate": 1.3270498438764888e-05, "loss": 0.66, "step": 285180 }, { "epoch": 24.7, "learning_rate": 1.3268330056667051e-05, "loss": 0.6944, "step": 285190 }, { "epoch": 24.7, "learning_rate": 1.3266161674569213e-05, "loss": 0.6199, "step": 285200 }, { "epoch": 24.7, "learning_rate": 1.3263993292471377e-05, "loss": 0.6822, "step": 285210 }, { "epoch": 24.7, "learning_rate": 1.3261824910373539e-05, "loss": 0.6905, "step": 285220 }, { "epoch": 24.7, "learning_rate": 1.3259656528275702e-05, "loss": 0.6578, "step": 285230 }, { "epoch": 24.7, "learning_rate": 1.3257488146177863e-05, "loss": 0.582, "step": 285240 }, { "epoch": 24.7, "learning_rate": 1.3255319764080028e-05, "loss": 0.5592, "step": 285250 }, { "epoch": 24.71, "learning_rate": 1.3253151381982189e-05, "loss": 0.6146, "step": 285260 }, { "epoch": 24.71, "learning_rate": 1.3250982999884352e-05, "loss": 0.5616, "step": 285270 }, { "epoch": 24.71, "learning_rate": 1.3248814617786515e-05, "loss": 0.5776, "step": 285280 }, { "epoch": 24.71, "learning_rate": 1.3246646235688678e-05, "loss": 0.586, "step": 285290 }, { "epoch": 24.71, "learning_rate": 1.324447785359084e-05, "loss": 0.4676, "step": 285300 }, { "epoch": 24.71, "learning_rate": 1.3242309471493002e-05, "loss": 0.6609, "step": 285310 }, { "epoch": 24.71, "learning_rate": 1.3240141089395165e-05, "loss": 0.6812, "step": 285320 }, { "epoch": 24.71, "learning_rate": 1.3237972707297327e-05, "loss": 0.5241, "step": 285330 }, { "epoch": 24.71, "learning_rate": 1.323580432519949e-05, "loss": 0.6927, "step": 285340 }, { "epoch": 24.71, "learning_rate": 1.3233635943101651e-05, "loss": 0.7686, "step": 285350 }, { "epoch": 24.71, "learning_rate": 1.3231467561003816e-05, "loss": 0.5591, "step": 285360 }, { "epoch": 24.71, "learning_rate": 1.3229299178905977e-05, "loss": 0.6397, "step": 285370 }, { "epoch": 24.72, "learning_rate": 1.322713079680814e-05, "loss": 0.5754, "step": 285380 }, { "epoch": 24.72, "learning_rate": 1.3224962414710303e-05, "loss": 0.6998, "step": 285390 }, { "epoch": 24.72, "learning_rate": 1.3222794032612466e-05, "loss": 0.7748, "step": 285400 }, { "epoch": 24.72, "learning_rate": 1.3220625650514627e-05, "loss": 0.71, "step": 285410 }, { "epoch": 24.72, "learning_rate": 1.3218457268416792e-05, "loss": 0.6312, "step": 285420 }, { "epoch": 24.72, "learning_rate": 1.3216288886318953e-05, "loss": 0.6502, "step": 285430 }, { "epoch": 24.72, "learning_rate": 1.3214120504221116e-05, "loss": 0.6396, "step": 285440 }, { "epoch": 24.72, "learning_rate": 1.3211952122123278e-05, "loss": 0.6268, "step": 285450 }, { "epoch": 24.72, "learning_rate": 1.3209783740025442e-05, "loss": 0.5412, "step": 285460 }, { "epoch": 24.72, "learning_rate": 1.3207615357927604e-05, "loss": 0.6623, "step": 285470 }, { "epoch": 24.72, "learning_rate": 1.3205446975829767e-05, "loss": 0.6656, "step": 285480 }, { "epoch": 24.73, "learning_rate": 1.3203278593731928e-05, "loss": 0.7178, "step": 285490 }, { "epoch": 24.73, "learning_rate": 1.3201110211634093e-05, "loss": 0.581, "step": 285500 }, { "epoch": 24.73, "learning_rate": 1.3198941829536254e-05, "loss": 0.6703, "step": 285510 }, { "epoch": 24.73, "learning_rate": 1.3196773447438417e-05, "loss": 0.6941, "step": 285520 }, { "epoch": 24.73, "learning_rate": 1.319460506534058e-05, "loss": 0.611, "step": 285530 }, { "epoch": 24.73, "learning_rate": 1.3192436683242743e-05, "loss": 0.6791, "step": 285540 }, { "epoch": 24.73, "learning_rate": 1.3190268301144904e-05, "loss": 0.7231, "step": 285550 }, { "epoch": 24.73, "learning_rate": 1.3188099919047069e-05, "loss": 0.6798, "step": 285560 }, { "epoch": 24.73, "learning_rate": 1.318593153694923e-05, "loss": 0.6465, "step": 285570 }, { "epoch": 24.73, "learning_rate": 1.3183763154851392e-05, "loss": 0.5113, "step": 285580 }, { "epoch": 24.73, "learning_rate": 1.3181594772753555e-05, "loss": 0.5605, "step": 285590 }, { "epoch": 24.73, "learning_rate": 1.3179426390655716e-05, "loss": 0.6283, "step": 285600 }, { "epoch": 24.74, "learning_rate": 1.317725800855788e-05, "loss": 0.6731, "step": 285610 }, { "epoch": 24.74, "learning_rate": 1.3175089626460042e-05, "loss": 0.5308, "step": 285620 }, { "epoch": 24.74, "learning_rate": 1.3172921244362205e-05, "loss": 0.6589, "step": 285630 }, { "epoch": 24.74, "learning_rate": 1.3170752862264368e-05, "loss": 0.6342, "step": 285640 }, { "epoch": 24.74, "learning_rate": 1.3168584480166531e-05, "loss": 0.6538, "step": 285650 }, { "epoch": 24.74, "learning_rate": 1.3166416098068692e-05, "loss": 0.6044, "step": 285660 }, { "epoch": 24.74, "learning_rate": 1.3164247715970857e-05, "loss": 0.6004, "step": 285670 }, { "epoch": 24.74, "learning_rate": 1.3162079333873018e-05, "loss": 0.6262, "step": 285680 }, { "epoch": 24.74, "learning_rate": 1.3159910951775181e-05, "loss": 0.737, "step": 285690 }, { "epoch": 24.74, "learning_rate": 1.3157742569677343e-05, "loss": 0.6308, "step": 285700 }, { "epoch": 24.74, "learning_rate": 1.3155574187579507e-05, "loss": 0.6138, "step": 285710 }, { "epoch": 24.75, "learning_rate": 1.3153405805481669e-05, "loss": 0.5868, "step": 285720 }, { "epoch": 24.75, "learning_rate": 1.3151237423383832e-05, "loss": 0.631, "step": 285730 }, { "epoch": 24.75, "learning_rate": 1.3149069041285993e-05, "loss": 0.704, "step": 285740 }, { "epoch": 24.75, "learning_rate": 1.3146900659188158e-05, "loss": 0.7699, "step": 285750 }, { "epoch": 24.75, "learning_rate": 1.3144732277090319e-05, "loss": 0.6695, "step": 285760 }, { "epoch": 24.75, "learning_rate": 1.3142563894992482e-05, "loss": 0.4863, "step": 285770 }, { "epoch": 24.75, "learning_rate": 1.3140395512894645e-05, "loss": 0.5672, "step": 285780 }, { "epoch": 24.75, "learning_rate": 1.3138227130796808e-05, "loss": 0.6555, "step": 285790 }, { "epoch": 24.75, "learning_rate": 1.313605874869897e-05, "loss": 0.6408, "step": 285800 }, { "epoch": 24.75, "learning_rate": 1.3133890366601134e-05, "loss": 0.633, "step": 285810 }, { "epoch": 24.75, "learning_rate": 1.3131721984503295e-05, "loss": 0.5794, "step": 285820 }, { "epoch": 24.75, "learning_rate": 1.3129553602405457e-05, "loss": 0.7404, "step": 285830 }, { "epoch": 24.76, "learning_rate": 1.312738522030762e-05, "loss": 0.5624, "step": 285840 }, { "epoch": 24.76, "learning_rate": 1.3125216838209781e-05, "loss": 0.4708, "step": 285850 }, { "epoch": 24.76, "learning_rate": 1.3123048456111946e-05, "loss": 0.6009, "step": 285860 }, { "epoch": 24.76, "learning_rate": 1.3120880074014107e-05, "loss": 0.756, "step": 285870 }, { "epoch": 24.76, "learning_rate": 1.311871169191627e-05, "loss": 0.6654, "step": 285880 }, { "epoch": 24.76, "learning_rate": 1.3116543309818433e-05, "loss": 0.5581, "step": 285890 }, { "epoch": 24.76, "learning_rate": 1.3114374927720596e-05, "loss": 0.6203, "step": 285900 }, { "epoch": 24.76, "learning_rate": 1.3112206545622757e-05, "loss": 0.6851, "step": 285910 }, { "epoch": 24.76, "learning_rate": 1.3110038163524922e-05, "loss": 0.8706, "step": 285920 }, { "epoch": 24.76, "learning_rate": 1.3107869781427083e-05, "loss": 0.7148, "step": 285930 }, { "epoch": 24.76, "learning_rate": 1.3105701399329246e-05, "loss": 0.6315, "step": 285940 }, { "epoch": 24.77, "learning_rate": 1.3103533017231408e-05, "loss": 0.6049, "step": 285950 }, { "epoch": 24.77, "learning_rate": 1.3101364635133572e-05, "loss": 0.5115, "step": 285960 }, { "epoch": 24.77, "learning_rate": 1.3099196253035734e-05, "loss": 0.6381, "step": 285970 }, { "epoch": 24.77, "learning_rate": 1.3097027870937897e-05, "loss": 0.6592, "step": 285980 }, { "epoch": 24.77, "learning_rate": 1.3094859488840058e-05, "loss": 0.6867, "step": 285990 }, { "epoch": 24.77, "learning_rate": 1.3092691106742223e-05, "loss": 0.7816, "step": 286000 }, { "epoch": 24.77, "learning_rate": 1.3090522724644384e-05, "loss": 0.73, "step": 286010 }, { "epoch": 24.77, "learning_rate": 1.3088354342546547e-05, "loss": 0.7198, "step": 286020 }, { "epoch": 24.77, "learning_rate": 1.308618596044871e-05, "loss": 0.6486, "step": 286030 }, { "epoch": 24.77, "learning_rate": 1.3084017578350873e-05, "loss": 0.661, "step": 286040 }, { "epoch": 24.77, "learning_rate": 1.3081849196253034e-05, "loss": 0.5643, "step": 286050 }, { "epoch": 24.77, "learning_rate": 1.3079680814155199e-05, "loss": 0.7643, "step": 286060 }, { "epoch": 24.78, "learning_rate": 1.307751243205736e-05, "loss": 0.7369, "step": 286070 }, { "epoch": 24.78, "learning_rate": 1.3075344049959523e-05, "loss": 0.6098, "step": 286080 }, { "epoch": 24.78, "learning_rate": 1.3073175667861684e-05, "loss": 0.7424, "step": 286090 }, { "epoch": 24.78, "learning_rate": 1.3071007285763846e-05, "loss": 0.5805, "step": 286100 }, { "epoch": 24.78, "learning_rate": 1.306883890366601e-05, "loss": 0.6697, "step": 286110 }, { "epoch": 24.78, "learning_rate": 1.3066670521568172e-05, "loss": 0.5252, "step": 286120 }, { "epoch": 24.78, "learning_rate": 1.3064502139470335e-05, "loss": 0.6567, "step": 286130 }, { "epoch": 24.78, "learning_rate": 1.3062333757372498e-05, "loss": 0.6854, "step": 286140 }, { "epoch": 24.78, "learning_rate": 1.306016537527466e-05, "loss": 0.6503, "step": 286150 }, { "epoch": 24.78, "learning_rate": 1.3057996993176822e-05, "loss": 0.6164, "step": 286160 }, { "epoch": 24.78, "learning_rate": 1.3055828611078987e-05, "loss": 0.6751, "step": 286170 }, { "epoch": 24.78, "learning_rate": 1.3053660228981148e-05, "loss": 0.5723, "step": 286180 }, { "epoch": 24.79, "learning_rate": 1.3051491846883311e-05, "loss": 0.6089, "step": 286190 }, { "epoch": 24.79, "learning_rate": 1.3049323464785472e-05, "loss": 0.6156, "step": 286200 }, { "epoch": 24.79, "learning_rate": 1.3047155082687637e-05, "loss": 0.6349, "step": 286210 }, { "epoch": 24.79, "learning_rate": 1.3044986700589798e-05, "loss": 0.5983, "step": 286220 }, { "epoch": 24.79, "learning_rate": 1.3042818318491961e-05, "loss": 0.654, "step": 286230 }, { "epoch": 24.79, "learning_rate": 1.3040649936394123e-05, "loss": 0.5514, "step": 286240 }, { "epoch": 24.79, "learning_rate": 1.3038481554296287e-05, "loss": 0.6534, "step": 286250 }, { "epoch": 24.79, "learning_rate": 1.3036313172198449e-05, "loss": 0.5583, "step": 286260 }, { "epoch": 24.79, "learning_rate": 1.3034144790100612e-05, "loss": 0.5812, "step": 286270 }, { "epoch": 24.79, "learning_rate": 1.3031976408002775e-05, "loss": 0.6045, "step": 286280 }, { "epoch": 24.79, "learning_rate": 1.3029808025904938e-05, "loss": 0.647, "step": 286290 }, { "epoch": 24.8, "learning_rate": 1.3027639643807099e-05, "loss": 0.5767, "step": 286300 }, { "epoch": 24.8, "learning_rate": 1.3025471261709264e-05, "loss": 0.6972, "step": 286310 }, { "epoch": 24.8, "learning_rate": 1.3023302879611425e-05, "loss": 0.6267, "step": 286320 }, { "epoch": 24.8, "learning_rate": 1.3021134497513588e-05, "loss": 0.7289, "step": 286330 }, { "epoch": 24.8, "learning_rate": 1.301896611541575e-05, "loss": 0.6088, "step": 286340 }, { "epoch": 24.8, "learning_rate": 1.3016797733317912e-05, "loss": 0.6864, "step": 286350 }, { "epoch": 24.8, "learning_rate": 1.3014629351220075e-05, "loss": 0.5627, "step": 286360 }, { "epoch": 24.8, "learning_rate": 1.3012460969122237e-05, "loss": 0.6051, "step": 286370 }, { "epoch": 24.8, "learning_rate": 1.30102925870244e-05, "loss": 0.6177, "step": 286380 }, { "epoch": 24.8, "learning_rate": 1.3008124204926563e-05, "loss": 0.6349, "step": 286390 }, { "epoch": 24.8, "learning_rate": 1.3005955822828726e-05, "loss": 0.694, "step": 286400 }, { "epoch": 24.8, "learning_rate": 1.3003787440730887e-05, "loss": 0.6173, "step": 286410 }, { "epoch": 24.81, "learning_rate": 1.3001619058633052e-05, "loss": 0.5068, "step": 286420 }, { "epoch": 24.81, "learning_rate": 1.2999450676535213e-05, "loss": 0.7318, "step": 286430 }, { "epoch": 24.81, "learning_rate": 1.2997282294437376e-05, "loss": 0.6874, "step": 286440 }, { "epoch": 24.81, "learning_rate": 1.2995113912339537e-05, "loss": 0.6178, "step": 286450 }, { "epoch": 24.81, "learning_rate": 1.2992945530241702e-05, "loss": 0.5692, "step": 286460 }, { "epoch": 24.81, "learning_rate": 1.2990777148143863e-05, "loss": 0.5716, "step": 286470 }, { "epoch": 24.81, "learning_rate": 1.2988608766046026e-05, "loss": 0.6045, "step": 286480 }, { "epoch": 24.81, "learning_rate": 1.298644038394819e-05, "loss": 0.6136, "step": 286490 }, { "epoch": 24.81, "learning_rate": 1.2984272001850352e-05, "loss": 0.639, "step": 286500 }, { "epoch": 24.81, "learning_rate": 1.2982103619752514e-05, "loss": 0.5908, "step": 286510 }, { "epoch": 24.81, "learning_rate": 1.2979935237654677e-05, "loss": 0.7179, "step": 286520 }, { "epoch": 24.82, "learning_rate": 1.297776685555684e-05, "loss": 0.5036, "step": 286530 }, { "epoch": 24.82, "learning_rate": 1.2975598473459003e-05, "loss": 0.6369, "step": 286540 }, { "epoch": 24.82, "learning_rate": 1.2973430091361164e-05, "loss": 0.6283, "step": 286550 }, { "epoch": 24.82, "learning_rate": 1.2971261709263329e-05, "loss": 0.6608, "step": 286560 }, { "epoch": 24.82, "learning_rate": 1.296909332716549e-05, "loss": 0.7582, "step": 286570 }, { "epoch": 24.82, "learning_rate": 1.2966924945067653e-05, "loss": 0.6436, "step": 286580 }, { "epoch": 24.82, "learning_rate": 1.2964756562969814e-05, "loss": 0.6387, "step": 286590 }, { "epoch": 24.82, "learning_rate": 1.2962588180871977e-05, "loss": 0.5287, "step": 286600 }, { "epoch": 24.82, "learning_rate": 1.296041979877414e-05, "loss": 0.7045, "step": 286610 }, { "epoch": 24.82, "learning_rate": 1.2958251416676302e-05, "loss": 0.6247, "step": 286620 }, { "epoch": 24.82, "learning_rate": 1.2956083034578466e-05, "loss": 0.7019, "step": 286630 }, { "epoch": 24.82, "learning_rate": 1.2953914652480628e-05, "loss": 0.6956, "step": 286640 }, { "epoch": 24.83, "learning_rate": 1.295174627038279e-05, "loss": 0.4949, "step": 286650 }, { "epoch": 24.83, "learning_rate": 1.2949577888284952e-05, "loss": 0.6077, "step": 286660 }, { "epoch": 24.83, "learning_rate": 1.2947409506187117e-05, "loss": 0.6925, "step": 286670 }, { "epoch": 24.83, "learning_rate": 1.2945241124089278e-05, "loss": 0.7251, "step": 286680 }, { "epoch": 24.83, "learning_rate": 1.2943072741991441e-05, "loss": 0.5577, "step": 286690 }, { "epoch": 24.83, "learning_rate": 1.2940904359893602e-05, "loss": 0.6123, "step": 286700 }, { "epoch": 24.83, "learning_rate": 1.2938735977795767e-05, "loss": 0.6826, "step": 286710 }, { "epoch": 24.83, "learning_rate": 1.2936567595697928e-05, "loss": 0.6849, "step": 286720 }, { "epoch": 24.83, "learning_rate": 1.2934399213600091e-05, "loss": 0.7584, "step": 286730 }, { "epoch": 24.83, "learning_rate": 1.2932230831502254e-05, "loss": 0.6827, "step": 286740 }, { "epoch": 24.83, "learning_rate": 1.2930062449404417e-05, "loss": 0.5257, "step": 286750 }, { "epoch": 24.84, "learning_rate": 1.2927894067306579e-05, "loss": 0.6016, "step": 286760 }, { "epoch": 24.84, "learning_rate": 1.2925725685208743e-05, "loss": 0.5942, "step": 286770 }, { "epoch": 24.84, "learning_rate": 1.2923557303110905e-05, "loss": 0.6667, "step": 286780 }, { "epoch": 24.84, "learning_rate": 1.2921388921013068e-05, "loss": 0.7559, "step": 286790 }, { "epoch": 24.84, "learning_rate": 1.2919220538915229e-05, "loss": 0.8077, "step": 286800 }, { "epoch": 24.84, "learning_rate": 1.2917052156817394e-05, "loss": 0.6815, "step": 286810 }, { "epoch": 24.84, "learning_rate": 1.2914883774719555e-05, "loss": 0.7022, "step": 286820 }, { "epoch": 24.84, "learning_rate": 1.2912715392621718e-05, "loss": 0.6741, "step": 286830 }, { "epoch": 24.84, "learning_rate": 1.291054701052388e-05, "loss": 0.6697, "step": 286840 }, { "epoch": 24.84, "learning_rate": 1.2908378628426044e-05, "loss": 0.6549, "step": 286850 }, { "epoch": 24.84, "learning_rate": 1.2906210246328205e-05, "loss": 0.6824, "step": 286860 }, { "epoch": 24.84, "learning_rate": 1.2904041864230367e-05, "loss": 0.757, "step": 286870 }, { "epoch": 24.85, "learning_rate": 1.2901873482132531e-05, "loss": 0.692, "step": 286880 }, { "epoch": 24.85, "learning_rate": 1.2899705100034693e-05, "loss": 0.6024, "step": 286890 }, { "epoch": 24.85, "learning_rate": 1.2897536717936856e-05, "loss": 0.5722, "step": 286900 }, { "epoch": 24.85, "learning_rate": 1.2895368335839017e-05, "loss": 0.5661, "step": 286910 }, { "epoch": 24.85, "learning_rate": 1.2893199953741182e-05, "loss": 0.6434, "step": 286920 }, { "epoch": 24.85, "learning_rate": 1.2891031571643343e-05, "loss": 0.5869, "step": 286930 }, { "epoch": 24.85, "learning_rate": 1.2888863189545506e-05, "loss": 0.6448, "step": 286940 }, { "epoch": 24.85, "learning_rate": 1.2886694807447667e-05, "loss": 0.5318, "step": 286950 }, { "epoch": 24.85, "learning_rate": 1.2884526425349832e-05, "loss": 0.6844, "step": 286960 }, { "epoch": 24.85, "learning_rate": 1.2882358043251993e-05, "loss": 0.627, "step": 286970 }, { "epoch": 24.85, "learning_rate": 1.2880189661154156e-05, "loss": 0.5649, "step": 286980 }, { "epoch": 24.86, "learning_rate": 1.287802127905632e-05, "loss": 0.4783, "step": 286990 }, { "epoch": 24.86, "learning_rate": 1.2875852896958482e-05, "loss": 0.6489, "step": 287000 }, { "epoch": 24.86, "learning_rate": 1.2873684514860644e-05, "loss": 0.8484, "step": 287010 }, { "epoch": 24.86, "learning_rate": 1.2871516132762808e-05, "loss": 0.645, "step": 287020 }, { "epoch": 24.86, "learning_rate": 1.286934775066497e-05, "loss": 0.6673, "step": 287030 }, { "epoch": 24.86, "learning_rate": 1.2867179368567133e-05, "loss": 0.6246, "step": 287040 }, { "epoch": 24.86, "learning_rate": 1.2865010986469294e-05, "loss": 0.5648, "step": 287050 }, { "epoch": 24.86, "learning_rate": 1.2862842604371459e-05, "loss": 0.5253, "step": 287060 }, { "epoch": 24.86, "learning_rate": 1.286067422227362e-05, "loss": 0.6638, "step": 287070 }, { "epoch": 24.86, "learning_rate": 1.2858505840175783e-05, "loss": 0.5271, "step": 287080 }, { "epoch": 24.86, "learning_rate": 1.2856337458077944e-05, "loss": 0.5907, "step": 287090 }, { "epoch": 24.86, "learning_rate": 1.2854169075980109e-05, "loss": 0.7064, "step": 287100 }, { "epoch": 24.87, "learning_rate": 1.285200069388227e-05, "loss": 0.5923, "step": 287110 }, { "epoch": 24.87, "learning_rate": 1.2849832311784431e-05, "loss": 0.6891, "step": 287120 }, { "epoch": 24.87, "learning_rate": 1.2847663929686596e-05, "loss": 0.6603, "step": 287130 }, { "epoch": 24.87, "learning_rate": 1.2845495547588757e-05, "loss": 0.5138, "step": 287140 }, { "epoch": 24.87, "learning_rate": 1.284332716549092e-05, "loss": 0.6467, "step": 287150 }, { "epoch": 24.87, "learning_rate": 1.2841158783393082e-05, "loss": 0.6043, "step": 287160 }, { "epoch": 24.87, "learning_rate": 1.2838990401295246e-05, "loss": 0.7178, "step": 287170 }, { "epoch": 24.87, "learning_rate": 1.2836822019197408e-05, "loss": 0.6035, "step": 287180 }, { "epoch": 24.87, "learning_rate": 1.283465363709957e-05, "loss": 0.497, "step": 287190 }, { "epoch": 24.87, "learning_rate": 1.2832485255001732e-05, "loss": 0.6423, "step": 287200 }, { "epoch": 24.87, "learning_rate": 1.2830316872903897e-05, "loss": 0.6639, "step": 287210 }, { "epoch": 24.88, "learning_rate": 1.2828148490806058e-05, "loss": 0.6449, "step": 287220 }, { "epoch": 24.88, "learning_rate": 1.2825980108708221e-05, "loss": 0.7064, "step": 287230 }, { "epoch": 24.88, "learning_rate": 1.2823811726610384e-05, "loss": 0.5834, "step": 287240 }, { "epoch": 24.88, "learning_rate": 1.2821643344512547e-05, "loss": 0.6335, "step": 287250 }, { "epoch": 24.88, "learning_rate": 1.2819474962414708e-05, "loss": 0.6438, "step": 287260 }, { "epoch": 24.88, "learning_rate": 1.2817306580316873e-05, "loss": 0.6889, "step": 287270 }, { "epoch": 24.88, "learning_rate": 1.2815138198219034e-05, "loss": 0.581, "step": 287280 }, { "epoch": 24.88, "learning_rate": 1.2812969816121197e-05, "loss": 0.6433, "step": 287290 }, { "epoch": 24.88, "learning_rate": 1.2810801434023359e-05, "loss": 0.5762, "step": 287300 }, { "epoch": 24.88, "learning_rate": 1.2808633051925523e-05, "loss": 0.6776, "step": 287310 }, { "epoch": 24.88, "learning_rate": 1.2806464669827685e-05, "loss": 0.7035, "step": 287320 }, { "epoch": 24.88, "learning_rate": 1.2804296287729848e-05, "loss": 0.6458, "step": 287330 }, { "epoch": 24.89, "learning_rate": 1.2802127905632009e-05, "loss": 0.5673, "step": 287340 }, { "epoch": 24.89, "learning_rate": 1.2799959523534174e-05, "loss": 0.5571, "step": 287350 }, { "epoch": 24.89, "learning_rate": 1.2797791141436335e-05, "loss": 0.702, "step": 287360 }, { "epoch": 24.89, "learning_rate": 1.2795622759338496e-05, "loss": 0.6513, "step": 287370 }, { "epoch": 24.89, "learning_rate": 1.2793454377240661e-05, "loss": 0.5765, "step": 287380 }, { "epoch": 24.89, "learning_rate": 1.2791285995142822e-05, "loss": 0.8138, "step": 287390 }, { "epoch": 24.89, "learning_rate": 1.2789117613044985e-05, "loss": 0.6109, "step": 287400 }, { "epoch": 24.89, "learning_rate": 1.2786949230947147e-05, "loss": 0.6341, "step": 287410 }, { "epoch": 24.89, "learning_rate": 1.2784780848849311e-05, "loss": 0.6988, "step": 287420 }, { "epoch": 24.89, "learning_rate": 1.2782612466751473e-05, "loss": 0.7751, "step": 287430 }, { "epoch": 24.89, "learning_rate": 1.2780444084653636e-05, "loss": 0.6125, "step": 287440 }, { "epoch": 24.89, "learning_rate": 1.2778275702555797e-05, "loss": 0.6857, "step": 287450 }, { "epoch": 24.9, "learning_rate": 1.2776107320457962e-05, "loss": 0.5379, "step": 287460 }, { "epoch": 24.9, "learning_rate": 1.2773938938360123e-05, "loss": 0.6803, "step": 287470 }, { "epoch": 24.9, "learning_rate": 1.2771770556262286e-05, "loss": 0.7075, "step": 287480 }, { "epoch": 24.9, "learning_rate": 1.2769602174164449e-05, "loss": 0.5341, "step": 287490 }, { "epoch": 24.9, "learning_rate": 1.2767433792066612e-05, "loss": 0.7971, "step": 287500 }, { "epoch": 24.9, "learning_rate": 1.2765265409968773e-05, "loss": 0.5107, "step": 287510 }, { "epoch": 24.9, "learning_rate": 1.2763097027870938e-05, "loss": 0.7253, "step": 287520 }, { "epoch": 24.9, "learning_rate": 1.27609286457731e-05, "loss": 0.7137, "step": 287530 }, { "epoch": 24.9, "learning_rate": 1.2758760263675262e-05, "loss": 0.6883, "step": 287540 }, { "epoch": 24.9, "learning_rate": 1.2756591881577424e-05, "loss": 0.6857, "step": 287550 }, { "epoch": 24.9, "learning_rate": 1.2754423499479588e-05, "loss": 0.7291, "step": 287560 }, { "epoch": 24.91, "learning_rate": 1.275225511738175e-05, "loss": 0.6101, "step": 287570 }, { "epoch": 24.91, "learning_rate": 1.2750086735283913e-05, "loss": 0.6345, "step": 287580 }, { "epoch": 24.91, "learning_rate": 1.2747918353186074e-05, "loss": 0.5499, "step": 287590 }, { "epoch": 24.91, "learning_rate": 1.2745749971088239e-05, "loss": 0.6394, "step": 287600 }, { "epoch": 24.91, "learning_rate": 1.27435815889904e-05, "loss": 0.7766, "step": 287610 }, { "epoch": 24.91, "learning_rate": 1.2741413206892563e-05, "loss": 0.7321, "step": 287620 }, { "epoch": 24.91, "learning_rate": 1.2739244824794726e-05, "loss": 0.4634, "step": 287630 }, { "epoch": 24.91, "learning_rate": 1.2737076442696887e-05, "loss": 0.5561, "step": 287640 }, { "epoch": 24.91, "learning_rate": 1.273490806059905e-05, "loss": 0.5077, "step": 287650 }, { "epoch": 24.91, "learning_rate": 1.2732739678501212e-05, "loss": 0.579, "step": 287660 }, { "epoch": 24.91, "learning_rate": 1.2730571296403376e-05, "loss": 0.6517, "step": 287670 }, { "epoch": 24.91, "learning_rate": 1.2728402914305538e-05, "loss": 0.5606, "step": 287680 }, { "epoch": 24.92, "learning_rate": 1.27262345322077e-05, "loss": 0.5583, "step": 287690 }, { "epoch": 24.92, "learning_rate": 1.2724066150109864e-05, "loss": 0.7295, "step": 287700 }, { "epoch": 24.92, "learning_rate": 1.2721897768012027e-05, "loss": 0.6085, "step": 287710 }, { "epoch": 24.92, "learning_rate": 1.2719729385914188e-05, "loss": 0.5368, "step": 287720 }, { "epoch": 24.92, "learning_rate": 1.2717561003816351e-05, "loss": 0.5773, "step": 287730 }, { "epoch": 24.92, "learning_rate": 1.2715392621718514e-05, "loss": 0.5641, "step": 287740 }, { "epoch": 24.92, "learning_rate": 1.2713224239620677e-05, "loss": 0.7119, "step": 287750 }, { "epoch": 24.92, "learning_rate": 1.2711055857522838e-05, "loss": 0.6794, "step": 287760 }, { "epoch": 24.92, "learning_rate": 1.2708887475425003e-05, "loss": 0.499, "step": 287770 }, { "epoch": 24.92, "learning_rate": 1.2706719093327164e-05, "loss": 0.6901, "step": 287780 }, { "epoch": 24.92, "learning_rate": 1.2704550711229327e-05, "loss": 0.6178, "step": 287790 }, { "epoch": 24.93, "learning_rate": 1.2702382329131489e-05, "loss": 0.7133, "step": 287800 }, { "epoch": 24.93, "learning_rate": 1.2700213947033653e-05, "loss": 0.4774, "step": 287810 }, { "epoch": 24.93, "learning_rate": 1.2698045564935815e-05, "loss": 0.493, "step": 287820 }, { "epoch": 24.93, "learning_rate": 1.2695877182837978e-05, "loss": 0.5641, "step": 287830 }, { "epoch": 24.93, "learning_rate": 1.269370880074014e-05, "loss": 0.6871, "step": 287840 }, { "epoch": 24.93, "learning_rate": 1.2691540418642304e-05, "loss": 0.5987, "step": 287850 }, { "epoch": 24.93, "learning_rate": 1.2689372036544465e-05, "loss": 0.7017, "step": 287860 }, { "epoch": 24.93, "learning_rate": 1.2687203654446628e-05, "loss": 0.6411, "step": 287870 }, { "epoch": 24.93, "learning_rate": 1.2685035272348791e-05, "loss": 0.5701, "step": 287880 }, { "epoch": 24.93, "learning_rate": 1.2682866890250952e-05, "loss": 0.5812, "step": 287890 }, { "epoch": 24.93, "learning_rate": 1.2680698508153115e-05, "loss": 0.6858, "step": 287900 }, { "epoch": 24.93, "learning_rate": 1.2678530126055277e-05, "loss": 0.5691, "step": 287910 }, { "epoch": 24.94, "learning_rate": 1.2676361743957441e-05, "loss": 0.5348, "step": 287920 }, { "epoch": 24.94, "learning_rate": 1.2674193361859603e-05, "loss": 0.6449, "step": 287930 }, { "epoch": 24.94, "learning_rate": 1.2672024979761766e-05, "loss": 0.5951, "step": 287940 }, { "epoch": 24.94, "learning_rate": 1.2669856597663929e-05, "loss": 0.5752, "step": 287950 }, { "epoch": 24.94, "learning_rate": 1.2667688215566092e-05, "loss": 0.6538, "step": 287960 }, { "epoch": 24.94, "learning_rate": 1.2665519833468253e-05, "loss": 0.5856, "step": 287970 }, { "epoch": 24.94, "learning_rate": 1.2663351451370418e-05, "loss": 0.6148, "step": 287980 }, { "epoch": 24.94, "learning_rate": 1.2661183069272579e-05, "loss": 0.6198, "step": 287990 }, { "epoch": 24.94, "learning_rate": 1.2659014687174742e-05, "loss": 0.6242, "step": 288000 }, { "epoch": 24.94, "learning_rate": 1.2656846305076903e-05, "loss": 0.7909, "step": 288010 }, { "epoch": 24.94, "learning_rate": 1.2654677922979068e-05, "loss": 0.6146, "step": 288020 }, { "epoch": 24.95, "learning_rate": 1.265250954088123e-05, "loss": 0.6907, "step": 288030 }, { "epoch": 24.95, "learning_rate": 1.2650341158783392e-05, "loss": 0.6539, "step": 288040 }, { "epoch": 24.95, "learning_rate": 1.2648172776685554e-05, "loss": 0.5617, "step": 288050 }, { "epoch": 24.95, "learning_rate": 1.2646004394587718e-05, "loss": 0.5115, "step": 288060 }, { "epoch": 24.95, "learning_rate": 1.264383601248988e-05, "loss": 0.614, "step": 288070 }, { "epoch": 24.95, "learning_rate": 1.2641667630392043e-05, "loss": 0.5489, "step": 288080 }, { "epoch": 24.95, "learning_rate": 1.2639499248294206e-05, "loss": 0.6447, "step": 288090 }, { "epoch": 24.95, "learning_rate": 1.2637330866196369e-05, "loss": 0.6997, "step": 288100 }, { "epoch": 24.95, "learning_rate": 1.263516248409853e-05, "loss": 0.677, "step": 288110 }, { "epoch": 24.95, "learning_rate": 1.2632994102000695e-05, "loss": 0.6822, "step": 288120 }, { "epoch": 24.95, "learning_rate": 1.2630825719902856e-05, "loss": 0.6777, "step": 288130 }, { "epoch": 24.95, "learning_rate": 1.2628657337805017e-05, "loss": 0.5426, "step": 288140 }, { "epoch": 24.96, "learning_rate": 1.262648895570718e-05, "loss": 0.6861, "step": 288150 }, { "epoch": 24.96, "learning_rate": 1.2624320573609341e-05, "loss": 0.6476, "step": 288160 }, { "epoch": 24.96, "learning_rate": 1.2622152191511506e-05, "loss": 0.7155, "step": 288170 }, { "epoch": 24.96, "learning_rate": 1.2619983809413668e-05, "loss": 0.6592, "step": 288180 }, { "epoch": 24.96, "learning_rate": 1.261781542731583e-05, "loss": 0.5917, "step": 288190 }, { "epoch": 24.96, "learning_rate": 1.2615647045217994e-05, "loss": 0.637, "step": 288200 }, { "epoch": 24.96, "learning_rate": 1.2613478663120157e-05, "loss": 0.6781, "step": 288210 }, { "epoch": 24.96, "learning_rate": 1.2611310281022318e-05, "loss": 0.7025, "step": 288220 }, { "epoch": 24.96, "learning_rate": 1.2609141898924483e-05, "loss": 0.607, "step": 288230 }, { "epoch": 24.96, "learning_rate": 1.2606973516826644e-05, "loss": 0.7594, "step": 288240 }, { "epoch": 24.96, "learning_rate": 1.2604805134728807e-05, "loss": 0.6798, "step": 288250 }, { "epoch": 24.97, "learning_rate": 1.2602636752630968e-05, "loss": 0.629, "step": 288260 }, { "epoch": 24.97, "learning_rate": 1.2600468370533133e-05, "loss": 0.6066, "step": 288270 }, { "epoch": 24.97, "learning_rate": 1.2598299988435294e-05, "loss": 0.6329, "step": 288280 }, { "epoch": 24.97, "learning_rate": 1.2596131606337457e-05, "loss": 0.7196, "step": 288290 }, { "epoch": 24.97, "learning_rate": 1.2593963224239618e-05, "loss": 0.6095, "step": 288300 }, { "epoch": 24.97, "learning_rate": 1.2591794842141783e-05, "loss": 0.6931, "step": 288310 }, { "epoch": 24.97, "learning_rate": 1.2589626460043944e-05, "loss": 0.6173, "step": 288320 }, { "epoch": 24.97, "learning_rate": 1.2587458077946107e-05, "loss": 0.7238, "step": 288330 }, { "epoch": 24.97, "learning_rate": 1.258528969584827e-05, "loss": 0.6795, "step": 288340 }, { "epoch": 24.97, "learning_rate": 1.2583121313750433e-05, "loss": 0.6925, "step": 288350 }, { "epoch": 24.97, "learning_rate": 1.2580952931652595e-05, "loss": 0.6168, "step": 288360 }, { "epoch": 24.97, "learning_rate": 1.257878454955476e-05, "loss": 0.6368, "step": 288370 }, { "epoch": 24.98, "learning_rate": 1.257661616745692e-05, "loss": 0.5755, "step": 288380 }, { "epoch": 24.98, "learning_rate": 1.2574447785359084e-05, "loss": 0.6066, "step": 288390 }, { "epoch": 24.98, "learning_rate": 1.2572279403261245e-05, "loss": 0.6163, "step": 288400 }, { "epoch": 24.98, "learning_rate": 1.2570111021163406e-05, "loss": 0.6631, "step": 288410 }, { "epoch": 24.98, "learning_rate": 1.2567942639065571e-05, "loss": 0.7062, "step": 288420 }, { "epoch": 24.98, "learning_rate": 1.2565774256967732e-05, "loss": 0.5249, "step": 288430 }, { "epoch": 24.98, "learning_rate": 1.2563605874869895e-05, "loss": 0.604, "step": 288440 }, { "epoch": 24.98, "learning_rate": 1.2561437492772058e-05, "loss": 0.8133, "step": 288450 }, { "epoch": 24.98, "learning_rate": 1.2559269110674221e-05, "loss": 0.7697, "step": 288460 }, { "epoch": 24.98, "learning_rate": 1.2557100728576383e-05, "loss": 0.5789, "step": 288470 }, { "epoch": 24.98, "learning_rate": 1.2554932346478547e-05, "loss": 0.6707, "step": 288480 }, { "epoch": 24.99, "learning_rate": 1.2552763964380709e-05, "loss": 0.6706, "step": 288490 }, { "epoch": 24.99, "learning_rate": 1.2550595582282872e-05, "loss": 0.6771, "step": 288500 }, { "epoch": 24.99, "learning_rate": 1.2548427200185033e-05, "loss": 0.7307, "step": 288510 }, { "epoch": 24.99, "learning_rate": 1.2546258818087198e-05, "loss": 0.7019, "step": 288520 }, { "epoch": 24.99, "learning_rate": 1.2544090435989359e-05, "loss": 0.5195, "step": 288530 }, { "epoch": 24.99, "learning_rate": 1.2541922053891522e-05, "loss": 0.6905, "step": 288540 }, { "epoch": 24.99, "learning_rate": 1.2539753671793683e-05, "loss": 0.7729, "step": 288550 }, { "epoch": 24.99, "learning_rate": 1.2537585289695848e-05, "loss": 0.6705, "step": 288560 }, { "epoch": 24.99, "learning_rate": 1.253541690759801e-05, "loss": 0.7858, "step": 288570 }, { "epoch": 24.99, "learning_rate": 1.2533248525500172e-05, "loss": 0.7456, "step": 288580 }, { "epoch": 24.99, "learning_rate": 1.2531080143402335e-05, "loss": 0.5893, "step": 288590 }, { "epoch": 24.99, "learning_rate": 1.2528911761304498e-05, "loss": 0.5713, "step": 288600 }, { "epoch": 25.0, "learning_rate": 1.252674337920666e-05, "loss": 0.5858, "step": 288610 }, { "epoch": 25.0, "learning_rate": 1.2524574997108824e-05, "loss": 0.665, "step": 288620 }, { "epoch": 25.0, "learning_rate": 1.2522406615010986e-05, "loss": 0.7248, "step": 288630 }, { "epoch": 25.0, "learning_rate": 1.2520238232913149e-05, "loss": 0.6679, "step": 288640 }, { "epoch": 25.0, "learning_rate": 1.251806985081531e-05, "loss": 0.6182, "step": 288650 }, { "epoch": 25.0, "learning_rate": 1.2515901468717471e-05, "loss": 0.6538, "step": 288660 }, { "epoch": 25.0, "eval_cer": 0.5310030817193127, "eval_em": 0.060856111397627644, "eval_f1": 0.060856111397627644, "eval_loss": 1.5604323148727417, "eval_runtime": 254.9099, "eval_samples_per_second": 7.607, "eval_steps_per_second": 0.953, "eval_wer": 0.9391438886023724, "step": 288662 }, { "epoch": 25.0, "learning_rate": 1.2513733086619636e-05, "loss": 0.6569, "step": 288670 }, { "epoch": 25.0, "learning_rate": 1.2511564704521797e-05, "loss": 0.6544, "step": 288680 }, { "epoch": 25.0, "learning_rate": 1.250939632242396e-05, "loss": 0.5507, "step": 288690 }, { "epoch": 25.0, "learning_rate": 1.2507227940326123e-05, "loss": 0.4996, "step": 288700 }, { "epoch": 25.0, "learning_rate": 1.2505059558228286e-05, "loss": 0.5766, "step": 288710 }, { "epoch": 25.0, "learning_rate": 1.2502891176130448e-05, "loss": 0.5962, "step": 288720 }, { "epoch": 25.01, "learning_rate": 1.2500722794032612e-05, "loss": 0.4458, "step": 288730 }, { "epoch": 25.01, "learning_rate": 1.2498554411934774e-05, "loss": 0.6457, "step": 288740 }, { "epoch": 25.01, "learning_rate": 1.2496386029836937e-05, "loss": 0.6089, "step": 288750 }, { "epoch": 25.01, "learning_rate": 1.2494217647739098e-05, "loss": 0.6009, "step": 288760 }, { "epoch": 25.01, "learning_rate": 1.2492049265641263e-05, "loss": 0.6001, "step": 288770 }, { "epoch": 25.01, "learning_rate": 1.2489880883543424e-05, "loss": 0.6033, "step": 288780 }, { "epoch": 25.01, "learning_rate": 1.2487712501445587e-05, "loss": 0.6271, "step": 288790 }, { "epoch": 25.01, "learning_rate": 1.2485544119347748e-05, "loss": 0.7275, "step": 288800 }, { "epoch": 25.01, "learning_rate": 1.2483375737249913e-05, "loss": 0.8829, "step": 288810 }, { "epoch": 25.01, "learning_rate": 1.2481207355152074e-05, "loss": 0.6197, "step": 288820 }, { "epoch": 25.01, "learning_rate": 1.2479038973054237e-05, "loss": 0.6172, "step": 288830 }, { "epoch": 25.02, "learning_rate": 1.24768705909564e-05, "loss": 0.5855, "step": 288840 }, { "epoch": 25.02, "learning_rate": 1.2474702208858563e-05, "loss": 0.6897, "step": 288850 }, { "epoch": 25.02, "learning_rate": 1.2472533826760725e-05, "loss": 0.5347, "step": 288860 }, { "epoch": 25.02, "learning_rate": 1.247036544466289e-05, "loss": 0.5442, "step": 288870 }, { "epoch": 25.02, "learning_rate": 1.246819706256505e-05, "loss": 0.6097, "step": 288880 }, { "epoch": 25.02, "learning_rate": 1.2466028680467214e-05, "loss": 0.5078, "step": 288890 }, { "epoch": 25.02, "learning_rate": 1.2463860298369375e-05, "loss": 0.7202, "step": 288900 }, { "epoch": 25.02, "learning_rate": 1.2461691916271538e-05, "loss": 0.5488, "step": 288910 }, { "epoch": 25.02, "learning_rate": 1.2459523534173701e-05, "loss": 0.6389, "step": 288920 }, { "epoch": 25.02, "learning_rate": 1.2457355152075862e-05, "loss": 0.6101, "step": 288930 }, { "epoch": 25.02, "learning_rate": 1.2455186769978025e-05, "loss": 0.7048, "step": 288940 }, { "epoch": 25.02, "learning_rate": 1.2453018387880188e-05, "loss": 0.584, "step": 288950 }, { "epoch": 25.03, "learning_rate": 1.2450850005782351e-05, "loss": 0.5851, "step": 288960 }, { "epoch": 25.03, "learning_rate": 1.2448681623684513e-05, "loss": 0.5743, "step": 288970 }, { "epoch": 25.03, "learning_rate": 1.2446513241586677e-05, "loss": 0.5311, "step": 288980 }, { "epoch": 25.03, "learning_rate": 1.2444344859488839e-05, "loss": 0.5313, "step": 288990 }, { "epoch": 25.03, "learning_rate": 1.2442176477391002e-05, "loss": 0.5966, "step": 289000 }, { "epoch": 25.03, "learning_rate": 1.2440008095293163e-05, "loss": 0.5655, "step": 289010 }, { "epoch": 25.03, "learning_rate": 1.2437839713195328e-05, "loss": 0.5358, "step": 289020 }, { "epoch": 25.03, "learning_rate": 1.2435671331097489e-05, "loss": 0.6349, "step": 289030 }, { "epoch": 25.03, "learning_rate": 1.2433502948999652e-05, "loss": 0.6071, "step": 289040 }, { "epoch": 25.03, "learning_rate": 1.2431334566901815e-05, "loss": 0.6139, "step": 289050 }, { "epoch": 25.03, "learning_rate": 1.2429166184803978e-05, "loss": 0.6403, "step": 289060 }, { "epoch": 25.04, "learning_rate": 1.242699780270614e-05, "loss": 0.6283, "step": 289070 }, { "epoch": 25.04, "learning_rate": 1.2424829420608302e-05, "loss": 0.6417, "step": 289080 }, { "epoch": 25.04, "learning_rate": 1.2422661038510465e-05, "loss": 0.6755, "step": 289090 }, { "epoch": 25.04, "learning_rate": 1.2420492656412628e-05, "loss": 0.5585, "step": 289100 }, { "epoch": 25.04, "learning_rate": 1.241832427431479e-05, "loss": 0.6527, "step": 289110 }, { "epoch": 25.04, "learning_rate": 1.2416155892216954e-05, "loss": 0.6455, "step": 289120 }, { "epoch": 25.04, "learning_rate": 1.2413987510119116e-05, "loss": 0.5134, "step": 289130 }, { "epoch": 25.04, "learning_rate": 1.2411819128021279e-05, "loss": 0.6815, "step": 289140 }, { "epoch": 25.04, "learning_rate": 1.240965074592344e-05, "loss": 0.7224, "step": 289150 }, { "epoch": 25.04, "learning_rate": 1.2407482363825605e-05, "loss": 0.5241, "step": 289160 }, { "epoch": 25.04, "learning_rate": 1.2405313981727766e-05, "loss": 0.5042, "step": 289170 }, { "epoch": 25.04, "learning_rate": 1.2403145599629927e-05, "loss": 0.7042, "step": 289180 }, { "epoch": 25.05, "learning_rate": 1.2400977217532092e-05, "loss": 0.6091, "step": 289190 }, { "epoch": 25.05, "learning_rate": 1.2398808835434253e-05, "loss": 0.653, "step": 289200 }, { "epoch": 25.05, "learning_rate": 1.2396640453336416e-05, "loss": 0.6094, "step": 289210 }, { "epoch": 25.05, "learning_rate": 1.2394472071238578e-05, "loss": 0.6446, "step": 289220 }, { "epoch": 25.05, "learning_rate": 1.2392303689140742e-05, "loss": 0.4628, "step": 289230 }, { "epoch": 25.05, "learning_rate": 1.2390135307042904e-05, "loss": 0.6946, "step": 289240 }, { "epoch": 25.05, "learning_rate": 1.2387966924945067e-05, "loss": 0.5206, "step": 289250 }, { "epoch": 25.05, "learning_rate": 1.2385798542847228e-05, "loss": 0.5887, "step": 289260 }, { "epoch": 25.05, "learning_rate": 1.2383630160749393e-05, "loss": 0.7099, "step": 289270 }, { "epoch": 25.05, "learning_rate": 1.2381461778651554e-05, "loss": 0.508, "step": 289280 }, { "epoch": 25.05, "learning_rate": 1.2379293396553717e-05, "loss": 0.6242, "step": 289290 }, { "epoch": 25.06, "learning_rate": 1.237712501445588e-05, "loss": 0.6032, "step": 289300 }, { "epoch": 25.06, "learning_rate": 1.2374956632358043e-05, "loss": 0.6591, "step": 289310 }, { "epoch": 25.06, "learning_rate": 1.2372788250260204e-05, "loss": 0.5827, "step": 289320 }, { "epoch": 25.06, "learning_rate": 1.2370619868162369e-05, "loss": 0.5022, "step": 289330 }, { "epoch": 25.06, "learning_rate": 1.236845148606453e-05, "loss": 0.6174, "step": 289340 }, { "epoch": 25.06, "learning_rate": 1.2366283103966693e-05, "loss": 0.5347, "step": 289350 }, { "epoch": 25.06, "learning_rate": 1.2364114721868854e-05, "loss": 0.5535, "step": 289360 }, { "epoch": 25.06, "learning_rate": 1.236194633977102e-05, "loss": 0.5732, "step": 289370 }, { "epoch": 25.06, "learning_rate": 1.235977795767318e-05, "loss": 0.496, "step": 289380 }, { "epoch": 25.06, "learning_rate": 1.2357609575575343e-05, "loss": 0.4796, "step": 289390 }, { "epoch": 25.06, "learning_rate": 1.2355441193477505e-05, "loss": 0.6667, "step": 289400 }, { "epoch": 25.06, "learning_rate": 1.235327281137967e-05, "loss": 0.7414, "step": 289410 }, { "epoch": 25.07, "learning_rate": 1.235110442928183e-05, "loss": 0.5799, "step": 289420 }, { "epoch": 25.07, "learning_rate": 1.2348936047183992e-05, "loss": 0.5757, "step": 289430 }, { "epoch": 25.07, "learning_rate": 1.2346767665086157e-05, "loss": 0.7449, "step": 289440 }, { "epoch": 25.07, "learning_rate": 1.2344599282988318e-05, "loss": 0.5134, "step": 289450 }, { "epoch": 25.07, "learning_rate": 1.2342430900890481e-05, "loss": 0.5485, "step": 289460 }, { "epoch": 25.07, "learning_rate": 1.2340262518792642e-05, "loss": 0.6169, "step": 289470 }, { "epoch": 25.07, "learning_rate": 1.2338094136694807e-05, "loss": 0.5755, "step": 289480 }, { "epoch": 25.07, "learning_rate": 1.2335925754596968e-05, "loss": 0.6382, "step": 289490 }, { "epoch": 25.07, "learning_rate": 1.2333757372499131e-05, "loss": 0.5805, "step": 289500 }, { "epoch": 25.07, "learning_rate": 1.2331588990401293e-05, "loss": 0.6384, "step": 289510 }, { "epoch": 25.07, "learning_rate": 1.2329420608303457e-05, "loss": 0.5557, "step": 289520 }, { "epoch": 25.08, "learning_rate": 1.2327252226205619e-05, "loss": 0.6597, "step": 289530 }, { "epoch": 25.08, "learning_rate": 1.2325083844107782e-05, "loss": 0.6883, "step": 289540 }, { "epoch": 25.08, "learning_rate": 1.2322915462009945e-05, "loss": 0.4907, "step": 289550 }, { "epoch": 25.08, "learning_rate": 1.2320747079912108e-05, "loss": 0.7986, "step": 289560 }, { "epoch": 25.08, "learning_rate": 1.2318578697814269e-05, "loss": 0.7243, "step": 289570 }, { "epoch": 25.08, "learning_rate": 1.2316410315716434e-05, "loss": 0.7786, "step": 289580 }, { "epoch": 25.08, "learning_rate": 1.2314241933618595e-05, "loss": 0.6528, "step": 289590 }, { "epoch": 25.08, "learning_rate": 1.2312073551520758e-05, "loss": 0.5391, "step": 289600 }, { "epoch": 25.08, "learning_rate": 1.230990516942292e-05, "loss": 0.5745, "step": 289610 }, { "epoch": 25.08, "learning_rate": 1.2307736787325084e-05, "loss": 0.69, "step": 289620 }, { "epoch": 25.08, "learning_rate": 1.2305568405227245e-05, "loss": 0.5459, "step": 289630 }, { "epoch": 25.08, "learning_rate": 1.2303400023129408e-05, "loss": 0.6024, "step": 289640 }, { "epoch": 25.09, "learning_rate": 1.230123164103157e-05, "loss": 0.534, "step": 289650 }, { "epoch": 25.09, "learning_rate": 1.2299063258933734e-05, "loss": 0.5548, "step": 289660 }, { "epoch": 25.09, "learning_rate": 1.2296894876835896e-05, "loss": 0.4553, "step": 289670 }, { "epoch": 25.09, "learning_rate": 1.2294726494738057e-05, "loss": 0.6263, "step": 289680 }, { "epoch": 25.09, "learning_rate": 1.2292558112640222e-05, "loss": 0.6659, "step": 289690 }, { "epoch": 25.09, "learning_rate": 1.2290389730542383e-05, "loss": 0.5407, "step": 289700 }, { "epoch": 25.09, "learning_rate": 1.2288221348444546e-05, "loss": 0.6379, "step": 289710 }, { "epoch": 25.09, "learning_rate": 1.2286052966346707e-05, "loss": 0.5675, "step": 289720 }, { "epoch": 25.09, "learning_rate": 1.2283884584248872e-05, "loss": 0.6105, "step": 289730 }, { "epoch": 25.09, "learning_rate": 1.2281716202151033e-05, "loss": 0.6197, "step": 289740 }, { "epoch": 25.09, "learning_rate": 1.2279547820053196e-05, "loss": 0.6603, "step": 289750 }, { "epoch": 25.1, "learning_rate": 1.2277379437955358e-05, "loss": 0.6873, "step": 289760 }, { "epoch": 25.1, "learning_rate": 1.2275211055857522e-05, "loss": 0.634, "step": 289770 }, { "epoch": 25.1, "learning_rate": 1.2273042673759684e-05, "loss": 0.703, "step": 289780 }, { "epoch": 25.1, "learning_rate": 1.2270874291661847e-05, "loss": 0.5794, "step": 289790 }, { "epoch": 25.1, "learning_rate": 1.226870590956401e-05, "loss": 0.6887, "step": 289800 }, { "epoch": 25.1, "learning_rate": 1.2266537527466173e-05, "loss": 0.6219, "step": 289810 }, { "epoch": 25.1, "learning_rate": 1.2264369145368334e-05, "loss": 0.5554, "step": 289820 }, { "epoch": 25.1, "learning_rate": 1.2262200763270499e-05, "loss": 0.6015, "step": 289830 }, { "epoch": 25.1, "learning_rate": 1.226003238117266e-05, "loss": 0.6576, "step": 289840 }, { "epoch": 25.1, "learning_rate": 1.2257863999074823e-05, "loss": 0.561, "step": 289850 }, { "epoch": 25.1, "learning_rate": 1.2255695616976984e-05, "loss": 0.5153, "step": 289860 }, { "epoch": 25.1, "learning_rate": 1.2253527234879149e-05, "loss": 0.7041, "step": 289870 }, { "epoch": 25.11, "learning_rate": 1.225135885278131e-05, "loss": 0.6051, "step": 289880 }, { "epoch": 25.11, "learning_rate": 1.2249190470683473e-05, "loss": 0.4902, "step": 289890 }, { "epoch": 25.11, "learning_rate": 1.2247022088585635e-05, "loss": 0.5666, "step": 289900 }, { "epoch": 25.11, "learning_rate": 1.22448537064878e-05, "loss": 0.5884, "step": 289910 }, { "epoch": 25.11, "learning_rate": 1.224268532438996e-05, "loss": 0.5369, "step": 289920 }, { "epoch": 25.11, "learning_rate": 1.2240516942292124e-05, "loss": 0.7327, "step": 289930 }, { "epoch": 25.11, "learning_rate": 1.2238348560194287e-05, "loss": 0.486, "step": 289940 }, { "epoch": 25.11, "learning_rate": 1.2236180178096448e-05, "loss": 0.5797, "step": 289950 }, { "epoch": 25.11, "learning_rate": 1.2234011795998611e-05, "loss": 0.5714, "step": 289960 }, { "epoch": 25.11, "learning_rate": 1.2231843413900772e-05, "loss": 0.7397, "step": 289970 }, { "epoch": 25.11, "learning_rate": 1.2229675031802937e-05, "loss": 0.5865, "step": 289980 }, { "epoch": 25.11, "learning_rate": 1.2227506649705098e-05, "loss": 0.522, "step": 289990 }, { "epoch": 25.12, "learning_rate": 1.2225338267607261e-05, "loss": 0.6769, "step": 290000 }, { "epoch": 25.12, "learning_rate": 1.2223169885509423e-05, "loss": 0.696, "step": 290010 }, { "epoch": 25.12, "learning_rate": 1.2221001503411587e-05, "loss": 0.5879, "step": 290020 }, { "epoch": 25.12, "learning_rate": 1.2218833121313749e-05, "loss": 0.5723, "step": 290030 }, { "epoch": 25.12, "learning_rate": 1.2216664739215912e-05, "loss": 0.5869, "step": 290040 }, { "epoch": 25.12, "learning_rate": 1.2214496357118075e-05, "loss": 0.6649, "step": 290050 }, { "epoch": 25.12, "learning_rate": 1.2212327975020238e-05, "loss": 0.5963, "step": 290060 }, { "epoch": 25.12, "learning_rate": 1.2210159592922399e-05, "loss": 0.6233, "step": 290070 }, { "epoch": 25.12, "learning_rate": 1.2207991210824564e-05, "loss": 0.5837, "step": 290080 }, { "epoch": 25.12, "learning_rate": 1.2205822828726725e-05, "loss": 0.5532, "step": 290090 }, { "epoch": 25.12, "learning_rate": 1.2203654446628888e-05, "loss": 0.5285, "step": 290100 }, { "epoch": 25.13, "learning_rate": 1.220148606453105e-05, "loss": 0.5588, "step": 290110 }, { "epoch": 25.13, "learning_rate": 1.2199317682433214e-05, "loss": 0.6377, "step": 290120 }, { "epoch": 25.13, "learning_rate": 1.2197149300335375e-05, "loss": 0.6338, "step": 290130 }, { "epoch": 25.13, "learning_rate": 1.2194980918237538e-05, "loss": 0.6758, "step": 290140 }, { "epoch": 25.13, "learning_rate": 1.21928125361397e-05, "loss": 0.55, "step": 290150 }, { "epoch": 25.13, "learning_rate": 1.2190644154041864e-05, "loss": 0.4382, "step": 290160 }, { "epoch": 25.13, "learning_rate": 1.2188475771944026e-05, "loss": 0.668, "step": 290170 }, { "epoch": 25.13, "learning_rate": 1.2186307389846189e-05, "loss": 0.713, "step": 290180 }, { "epoch": 25.13, "learning_rate": 1.2184139007748352e-05, "loss": 0.6722, "step": 290190 }, { "epoch": 25.13, "learning_rate": 1.2181970625650513e-05, "loss": 0.5691, "step": 290200 }, { "epoch": 25.13, "learning_rate": 1.2179802243552676e-05, "loss": 0.7214, "step": 290210 }, { "epoch": 25.13, "learning_rate": 1.2177633861454837e-05, "loss": 0.6348, "step": 290220 }, { "epoch": 25.14, "learning_rate": 1.2175465479357002e-05, "loss": 0.5982, "step": 290230 }, { "epoch": 25.14, "learning_rate": 1.2173297097259163e-05, "loss": 0.6758, "step": 290240 }, { "epoch": 25.14, "learning_rate": 1.2171128715161326e-05, "loss": 0.5319, "step": 290250 }, { "epoch": 25.14, "learning_rate": 1.216896033306349e-05, "loss": 0.562, "step": 290260 }, { "epoch": 25.14, "learning_rate": 1.2166791950965652e-05, "loss": 0.6403, "step": 290270 }, { "epoch": 25.14, "learning_rate": 1.2164623568867814e-05, "loss": 0.6275, "step": 290280 }, { "epoch": 25.14, "learning_rate": 1.2162455186769977e-05, "loss": 0.6455, "step": 290290 }, { "epoch": 25.14, "learning_rate": 1.216028680467214e-05, "loss": 0.5981, "step": 290300 }, { "epoch": 25.14, "learning_rate": 1.2158118422574303e-05, "loss": 0.5434, "step": 290310 }, { "epoch": 25.14, "learning_rate": 1.2155950040476464e-05, "loss": 0.6666, "step": 290320 }, { "epoch": 25.14, "learning_rate": 1.2153781658378629e-05, "loss": 0.7499, "step": 290330 }, { "epoch": 25.15, "learning_rate": 1.215161327628079e-05, "loss": 0.6131, "step": 290340 }, { "epoch": 25.15, "learning_rate": 1.2149444894182953e-05, "loss": 0.6087, "step": 290350 }, { "epoch": 25.15, "learning_rate": 1.2147276512085114e-05, "loss": 0.6051, "step": 290360 }, { "epoch": 25.15, "learning_rate": 1.2145108129987279e-05, "loss": 0.7337, "step": 290370 }, { "epoch": 25.15, "learning_rate": 1.214293974788944e-05, "loss": 0.5047, "step": 290380 }, { "epoch": 25.15, "learning_rate": 1.2140771365791603e-05, "loss": 0.56, "step": 290390 }, { "epoch": 25.15, "learning_rate": 1.2138602983693766e-05, "loss": 0.5779, "step": 290400 }, { "epoch": 25.15, "learning_rate": 1.213643460159593e-05, "loss": 0.597, "step": 290410 }, { "epoch": 25.15, "learning_rate": 1.213426621949809e-05, "loss": 0.5938, "step": 290420 }, { "epoch": 25.15, "learning_rate": 1.2132097837400254e-05, "loss": 0.6127, "step": 290430 }, { "epoch": 25.15, "learning_rate": 1.2129929455302417e-05, "loss": 0.6171, "step": 290440 }, { "epoch": 25.15, "learning_rate": 1.2127761073204578e-05, "loss": 0.6683, "step": 290450 }, { "epoch": 25.16, "learning_rate": 1.212559269110674e-05, "loss": 0.6618, "step": 290460 }, { "epoch": 25.16, "learning_rate": 1.2123424309008902e-05, "loss": 0.5545, "step": 290470 }, { "epoch": 25.16, "learning_rate": 1.2121255926911067e-05, "loss": 0.5196, "step": 290480 }, { "epoch": 25.16, "learning_rate": 1.2119087544813228e-05, "loss": 0.7091, "step": 290490 }, { "epoch": 25.16, "learning_rate": 1.2116919162715391e-05, "loss": 0.6404, "step": 290500 }, { "epoch": 25.16, "learning_rate": 1.2114750780617554e-05, "loss": 0.5781, "step": 290510 }, { "epoch": 25.16, "learning_rate": 1.2112582398519717e-05, "loss": 0.6241, "step": 290520 }, { "epoch": 25.16, "learning_rate": 1.2110414016421878e-05, "loss": 0.5052, "step": 290530 }, { "epoch": 25.16, "learning_rate": 1.2108245634324043e-05, "loss": 0.6248, "step": 290540 }, { "epoch": 25.16, "learning_rate": 1.2106077252226204e-05, "loss": 0.681, "step": 290550 }, { "epoch": 25.16, "learning_rate": 1.2103908870128367e-05, "loss": 0.5606, "step": 290560 }, { "epoch": 25.17, "learning_rate": 1.2101740488030529e-05, "loss": 0.739, "step": 290570 }, { "epoch": 25.17, "learning_rate": 1.2099572105932693e-05, "loss": 0.605, "step": 290580 }, { "epoch": 25.17, "learning_rate": 1.2097403723834855e-05, "loss": 0.6442, "step": 290590 }, { "epoch": 25.17, "learning_rate": 1.2095235341737018e-05, "loss": 0.5534, "step": 290600 }, { "epoch": 25.17, "learning_rate": 1.2093066959639179e-05, "loss": 0.6344, "step": 290610 }, { "epoch": 25.17, "learning_rate": 1.2090898577541344e-05, "loss": 0.6164, "step": 290620 }, { "epoch": 25.17, "learning_rate": 1.2088730195443505e-05, "loss": 0.4898, "step": 290630 }, { "epoch": 25.17, "learning_rate": 1.2086561813345668e-05, "loss": 0.6982, "step": 290640 }, { "epoch": 25.17, "learning_rate": 1.2084393431247831e-05, "loss": 0.6881, "step": 290650 }, { "epoch": 25.17, "learning_rate": 1.2082225049149994e-05, "loss": 0.7327, "step": 290660 }, { "epoch": 25.17, "learning_rate": 1.2080056667052155e-05, "loss": 0.6379, "step": 290670 }, { "epoch": 25.17, "learning_rate": 1.207788828495432e-05, "loss": 0.5484, "step": 290680 }, { "epoch": 25.18, "learning_rate": 1.2075719902856481e-05, "loss": 0.6526, "step": 290690 }, { "epoch": 25.18, "learning_rate": 1.2073551520758644e-05, "loss": 0.6146, "step": 290700 }, { "epoch": 25.18, "learning_rate": 1.2071383138660806e-05, "loss": 0.6804, "step": 290710 }, { "epoch": 25.18, "learning_rate": 1.2069214756562967e-05, "loss": 0.5814, "step": 290720 }, { "epoch": 25.18, "learning_rate": 1.2067046374465132e-05, "loss": 0.6933, "step": 290730 }, { "epoch": 25.18, "learning_rate": 1.2064877992367293e-05, "loss": 0.8056, "step": 290740 }, { "epoch": 25.18, "learning_rate": 1.2062709610269456e-05, "loss": 0.7701, "step": 290750 }, { "epoch": 25.18, "learning_rate": 1.2060541228171619e-05, "loss": 0.5764, "step": 290760 }, { "epoch": 25.18, "learning_rate": 1.2058372846073782e-05, "loss": 0.6652, "step": 290770 }, { "epoch": 25.18, "learning_rate": 1.2056204463975943e-05, "loss": 0.633, "step": 290780 }, { "epoch": 25.18, "learning_rate": 1.2054036081878108e-05, "loss": 0.6083, "step": 290790 }, { "epoch": 25.19, "learning_rate": 1.205186769978027e-05, "loss": 0.5805, "step": 290800 }, { "epoch": 25.19, "learning_rate": 1.2049699317682432e-05, "loss": 0.6292, "step": 290810 }, { "epoch": 25.19, "learning_rate": 1.2047530935584594e-05, "loss": 0.6238, "step": 290820 }, { "epoch": 25.19, "learning_rate": 1.2045362553486758e-05, "loss": 0.5707, "step": 290830 }, { "epoch": 25.19, "learning_rate": 1.204319417138892e-05, "loss": 0.5824, "step": 290840 }, { "epoch": 25.19, "learning_rate": 1.2041025789291083e-05, "loss": 0.6638, "step": 290850 }, { "epoch": 25.19, "learning_rate": 1.2038857407193244e-05, "loss": 0.5832, "step": 290860 }, { "epoch": 25.19, "learning_rate": 1.2036689025095409e-05, "loss": 0.6568, "step": 290870 }, { "epoch": 25.19, "learning_rate": 1.203452064299757e-05, "loss": 0.4609, "step": 290880 }, { "epoch": 25.19, "learning_rate": 1.2032352260899733e-05, "loss": 0.504, "step": 290890 }, { "epoch": 25.19, "learning_rate": 1.2030183878801896e-05, "loss": 0.6222, "step": 290900 }, { "epoch": 25.19, "learning_rate": 1.2028015496704059e-05, "loss": 0.6215, "step": 290910 }, { "epoch": 25.2, "learning_rate": 1.202584711460622e-05, "loss": 0.6509, "step": 290920 }, { "epoch": 25.2, "learning_rate": 1.2023678732508385e-05, "loss": 0.6254, "step": 290930 }, { "epoch": 25.2, "learning_rate": 1.2021510350410546e-05, "loss": 0.525, "step": 290940 }, { "epoch": 25.2, "learning_rate": 1.201934196831271e-05, "loss": 0.5678, "step": 290950 }, { "epoch": 25.2, "learning_rate": 1.201717358621487e-05, "loss": 0.5566, "step": 290960 }, { "epoch": 25.2, "learning_rate": 1.2015005204117032e-05, "loss": 0.5691, "step": 290970 }, { "epoch": 25.2, "learning_rate": 1.2012836822019197e-05, "loss": 0.6177, "step": 290980 }, { "epoch": 25.2, "learning_rate": 1.2010668439921358e-05, "loss": 0.6599, "step": 290990 }, { "epoch": 25.2, "learning_rate": 1.2008500057823521e-05, "loss": 0.6909, "step": 291000 }, { "epoch": 25.2, "learning_rate": 1.2006331675725684e-05, "loss": 0.6881, "step": 291010 }, { "epoch": 25.2, "learning_rate": 1.2004163293627847e-05, "loss": 0.5983, "step": 291020 }, { "epoch": 25.21, "learning_rate": 1.2001994911530008e-05, "loss": 0.6206, "step": 291030 }, { "epoch": 25.21, "learning_rate": 1.1999826529432173e-05, "loss": 0.5906, "step": 291040 }, { "epoch": 25.21, "learning_rate": 1.1997658147334334e-05, "loss": 0.5218, "step": 291050 }, { "epoch": 25.21, "learning_rate": 1.1995489765236497e-05, "loss": 0.5992, "step": 291060 }, { "epoch": 25.21, "learning_rate": 1.1993321383138659e-05, "loss": 0.6933, "step": 291070 }, { "epoch": 25.21, "learning_rate": 1.1991153001040823e-05, "loss": 0.5688, "step": 291080 }, { "epoch": 25.21, "learning_rate": 1.1988984618942985e-05, "loss": 0.618, "step": 291090 }, { "epoch": 25.21, "learning_rate": 1.1986816236845148e-05, "loss": 0.7233, "step": 291100 }, { "epoch": 25.21, "learning_rate": 1.1984647854747309e-05, "loss": 0.5982, "step": 291110 }, { "epoch": 25.21, "learning_rate": 1.1982479472649474e-05, "loss": 0.5936, "step": 291120 }, { "epoch": 25.21, "learning_rate": 1.1980311090551635e-05, "loss": 0.5372, "step": 291130 }, { "epoch": 25.21, "learning_rate": 1.1978142708453798e-05, "loss": 0.4889, "step": 291140 }, { "epoch": 25.22, "learning_rate": 1.1975974326355961e-05, "loss": 0.5997, "step": 291150 }, { "epoch": 25.22, "learning_rate": 1.1973805944258124e-05, "loss": 0.7535, "step": 291160 }, { "epoch": 25.22, "learning_rate": 1.1971637562160285e-05, "loss": 0.684, "step": 291170 }, { "epoch": 25.22, "learning_rate": 1.196946918006245e-05, "loss": 0.4337, "step": 291180 }, { "epoch": 25.22, "learning_rate": 1.1967300797964611e-05, "loss": 0.5693, "step": 291190 }, { "epoch": 25.22, "learning_rate": 1.1965132415866774e-05, "loss": 0.6234, "step": 291200 }, { "epoch": 25.22, "learning_rate": 1.1962964033768936e-05, "loss": 0.5845, "step": 291210 }, { "epoch": 25.22, "learning_rate": 1.1960795651671097e-05, "loss": 0.6594, "step": 291220 }, { "epoch": 25.22, "learning_rate": 1.1958627269573262e-05, "loss": 0.6286, "step": 291230 }, { "epoch": 25.22, "learning_rate": 1.1956458887475423e-05, "loss": 0.6925, "step": 291240 }, { "epoch": 25.22, "learning_rate": 1.1954290505377586e-05, "loss": 0.6093, "step": 291250 }, { "epoch": 25.22, "learning_rate": 1.1952122123279749e-05, "loss": 0.7327, "step": 291260 }, { "epoch": 25.23, "learning_rate": 1.1949953741181912e-05, "loss": 0.5798, "step": 291270 }, { "epoch": 25.23, "learning_rate": 1.1947785359084073e-05, "loss": 0.588, "step": 291280 }, { "epoch": 25.23, "learning_rate": 1.1945616976986238e-05, "loss": 0.4273, "step": 291290 }, { "epoch": 25.23, "learning_rate": 1.19434485948884e-05, "loss": 0.526, "step": 291300 }, { "epoch": 25.23, "learning_rate": 1.1941280212790562e-05, "loss": 0.6567, "step": 291310 }, { "epoch": 25.23, "learning_rate": 1.1939111830692724e-05, "loss": 0.5728, "step": 291320 }, { "epoch": 25.23, "learning_rate": 1.1936943448594888e-05, "loss": 0.7022, "step": 291330 }, { "epoch": 25.23, "learning_rate": 1.193477506649705e-05, "loss": 0.5188, "step": 291340 }, { "epoch": 25.23, "learning_rate": 1.1932606684399213e-05, "loss": 0.6038, "step": 291350 }, { "epoch": 25.23, "learning_rate": 1.1930438302301374e-05, "loss": 0.5433, "step": 291360 }, { "epoch": 25.23, "learning_rate": 1.1928269920203539e-05, "loss": 0.5444, "step": 291370 }, { "epoch": 25.24, "learning_rate": 1.19261015381057e-05, "loss": 0.6342, "step": 291380 }, { "epoch": 25.24, "learning_rate": 1.1923933156007863e-05, "loss": 0.6821, "step": 291390 }, { "epoch": 25.24, "learning_rate": 1.1921764773910026e-05, "loss": 0.7034, "step": 291400 }, { "epoch": 25.24, "learning_rate": 1.1919596391812189e-05, "loss": 0.7088, "step": 291410 }, { "epoch": 25.24, "learning_rate": 1.191742800971435e-05, "loss": 0.6944, "step": 291420 }, { "epoch": 25.24, "learning_rate": 1.1915259627616515e-05, "loss": 0.5561, "step": 291430 }, { "epoch": 25.24, "learning_rate": 1.1913091245518676e-05, "loss": 0.538, "step": 291440 }, { "epoch": 25.24, "learning_rate": 1.191092286342084e-05, "loss": 0.7452, "step": 291450 }, { "epoch": 25.24, "learning_rate": 1.1908754481323e-05, "loss": 0.5999, "step": 291460 }, { "epoch": 25.24, "learning_rate": 1.1906586099225165e-05, "loss": 0.6411, "step": 291470 }, { "epoch": 25.24, "learning_rate": 1.1904417717127327e-05, "loss": 0.5985, "step": 291480 }, { "epoch": 25.24, "learning_rate": 1.1902249335029488e-05, "loss": 0.8095, "step": 291490 }, { "epoch": 25.25, "learning_rate": 1.190008095293165e-05, "loss": 0.6957, "step": 291500 }, { "epoch": 25.25, "learning_rate": 1.1897912570833814e-05, "loss": 0.6472, "step": 291510 }, { "epoch": 25.25, "learning_rate": 1.1895744188735977e-05, "loss": 0.6153, "step": 291520 }, { "epoch": 25.25, "learning_rate": 1.1893575806638138e-05, "loss": 0.6918, "step": 291530 }, { "epoch": 25.25, "learning_rate": 1.1891407424540303e-05, "loss": 0.5634, "step": 291540 }, { "epoch": 25.25, "learning_rate": 1.1889239042442464e-05, "loss": 0.6511, "step": 291550 }, { "epoch": 25.25, "learning_rate": 1.1887070660344627e-05, "loss": 0.7297, "step": 291560 }, { "epoch": 25.25, "learning_rate": 1.1884902278246788e-05, "loss": 0.5297, "step": 291570 }, { "epoch": 25.25, "learning_rate": 1.1882733896148953e-05, "loss": 0.5278, "step": 291580 }, { "epoch": 25.25, "learning_rate": 1.1880565514051114e-05, "loss": 0.5084, "step": 291590 }, { "epoch": 25.25, "learning_rate": 1.1878397131953277e-05, "loss": 0.6966, "step": 291600 }, { "epoch": 25.26, "learning_rate": 1.187622874985544e-05, "loss": 0.6814, "step": 291610 }, { "epoch": 25.26, "learning_rate": 1.1874060367757603e-05, "loss": 0.5182, "step": 291620 }, { "epoch": 25.26, "learning_rate": 1.1871891985659765e-05, "loss": 0.6469, "step": 291630 }, { "epoch": 25.26, "learning_rate": 1.1869723603561928e-05, "loss": 0.6201, "step": 291640 }, { "epoch": 25.26, "learning_rate": 1.186755522146409e-05, "loss": 0.6581, "step": 291650 }, { "epoch": 25.26, "learning_rate": 1.1865386839366254e-05, "loss": 0.466, "step": 291660 }, { "epoch": 25.26, "learning_rate": 1.1863218457268415e-05, "loss": 0.7088, "step": 291670 }, { "epoch": 25.26, "learning_rate": 1.186105007517058e-05, "loss": 0.6257, "step": 291680 }, { "epoch": 25.26, "learning_rate": 1.1858881693072741e-05, "loss": 0.5362, "step": 291690 }, { "epoch": 25.26, "learning_rate": 1.1856713310974904e-05, "loss": 0.6795, "step": 291700 }, { "epoch": 25.26, "learning_rate": 1.1854544928877065e-05, "loss": 0.6358, "step": 291710 }, { "epoch": 25.26, "learning_rate": 1.185237654677923e-05, "loss": 0.6106, "step": 291720 }, { "epoch": 25.27, "learning_rate": 1.1850208164681391e-05, "loss": 0.7055, "step": 291730 }, { "epoch": 25.27, "learning_rate": 1.1848039782583553e-05, "loss": 0.5966, "step": 291740 }, { "epoch": 25.27, "learning_rate": 1.1845871400485717e-05, "loss": 0.5505, "step": 291750 }, { "epoch": 25.27, "learning_rate": 1.1843703018387879e-05, "loss": 0.489, "step": 291760 }, { "epoch": 25.27, "learning_rate": 1.1841534636290042e-05, "loss": 0.6774, "step": 291770 }, { "epoch": 25.27, "learning_rate": 1.1839366254192203e-05, "loss": 0.6697, "step": 291780 }, { "epoch": 25.27, "learning_rate": 1.1837197872094368e-05, "loss": 0.6966, "step": 291790 }, { "epoch": 25.27, "learning_rate": 1.1835029489996529e-05, "loss": 0.6941, "step": 291800 }, { "epoch": 25.27, "learning_rate": 1.1832861107898692e-05, "loss": 0.6775, "step": 291810 }, { "epoch": 25.27, "learning_rate": 1.1830692725800853e-05, "loss": 0.6043, "step": 291820 }, { "epoch": 25.27, "learning_rate": 1.1828524343703018e-05, "loss": 0.5013, "step": 291830 }, { "epoch": 25.28, "learning_rate": 1.182635596160518e-05, "loss": 0.7056, "step": 291840 }, { "epoch": 25.28, "learning_rate": 1.1824187579507342e-05, "loss": 0.441, "step": 291850 }, { "epoch": 25.28, "learning_rate": 1.1822019197409505e-05, "loss": 0.6104, "step": 291860 }, { "epoch": 25.28, "learning_rate": 1.1819850815311668e-05, "loss": 0.6904, "step": 291870 }, { "epoch": 25.28, "learning_rate": 1.181768243321383e-05, "loss": 0.7238, "step": 291880 }, { "epoch": 25.28, "learning_rate": 1.1815514051115994e-05, "loss": 0.6955, "step": 291890 }, { "epoch": 25.28, "learning_rate": 1.1813345669018156e-05, "loss": 0.5606, "step": 291900 }, { "epoch": 25.28, "learning_rate": 1.1811177286920319e-05, "loss": 0.6076, "step": 291910 }, { "epoch": 25.28, "learning_rate": 1.180900890482248e-05, "loss": 0.586, "step": 291920 }, { "epoch": 25.28, "learning_rate": 1.1806840522724645e-05, "loss": 0.3998, "step": 291930 }, { "epoch": 25.28, "learning_rate": 1.1804672140626806e-05, "loss": 0.6381, "step": 291940 }, { "epoch": 25.28, "learning_rate": 1.1802503758528969e-05, "loss": 0.5387, "step": 291950 }, { "epoch": 25.29, "learning_rate": 1.180033537643113e-05, "loss": 0.5682, "step": 291960 }, { "epoch": 25.29, "learning_rate": 1.1798166994333295e-05, "loss": 0.7367, "step": 291970 }, { "epoch": 25.29, "learning_rate": 1.1795998612235456e-05, "loss": 0.594, "step": 291980 }, { "epoch": 25.29, "learning_rate": 1.1793830230137618e-05, "loss": 0.4492, "step": 291990 }, { "epoch": 25.29, "learning_rate": 1.1791661848039782e-05, "loss": 0.5272, "step": 292000 }, { "epoch": 25.29, "learning_rate": 1.1789493465941944e-05, "loss": 0.5451, "step": 292010 }, { "epoch": 25.29, "learning_rate": 1.1787325083844107e-05, "loss": 0.588, "step": 292020 }, { "epoch": 25.29, "learning_rate": 1.1785156701746268e-05, "loss": 0.5762, "step": 292030 }, { "epoch": 25.29, "learning_rate": 1.1782988319648433e-05, "loss": 0.7243, "step": 292040 }, { "epoch": 25.29, "learning_rate": 1.1780819937550594e-05, "loss": 0.5915, "step": 292050 }, { "epoch": 25.29, "learning_rate": 1.1778651555452757e-05, "loss": 0.6801, "step": 292060 }, { "epoch": 25.3, "learning_rate": 1.1776483173354918e-05, "loss": 0.5202, "step": 292070 }, { "epoch": 25.3, "learning_rate": 1.1774314791257083e-05, "loss": 0.6874, "step": 292080 }, { "epoch": 25.3, "learning_rate": 1.1772146409159244e-05, "loss": 0.4174, "step": 292090 }, { "epoch": 25.3, "learning_rate": 1.1769978027061407e-05, "loss": 0.6434, "step": 292100 }, { "epoch": 25.3, "learning_rate": 1.176780964496357e-05, "loss": 0.6465, "step": 292110 }, { "epoch": 25.3, "learning_rate": 1.1765641262865733e-05, "loss": 0.7166, "step": 292120 }, { "epoch": 25.3, "learning_rate": 1.1763472880767895e-05, "loss": 0.6936, "step": 292130 }, { "epoch": 25.3, "learning_rate": 1.176130449867006e-05, "loss": 0.57, "step": 292140 }, { "epoch": 25.3, "learning_rate": 1.175913611657222e-05, "loss": 0.7371, "step": 292150 }, { "epoch": 25.3, "learning_rate": 1.1756967734474384e-05, "loss": 0.5288, "step": 292160 }, { "epoch": 25.3, "learning_rate": 1.1754799352376545e-05, "loss": 0.4837, "step": 292170 }, { "epoch": 25.3, "learning_rate": 1.175263097027871e-05, "loss": 0.4942, "step": 292180 }, { "epoch": 25.31, "learning_rate": 1.1750462588180871e-05, "loss": 0.549, "step": 292190 }, { "epoch": 25.31, "learning_rate": 1.1748294206083034e-05, "loss": 0.6978, "step": 292200 }, { "epoch": 25.31, "learning_rate": 1.1746125823985195e-05, "loss": 0.6317, "step": 292210 }, { "epoch": 25.31, "learning_rate": 1.174395744188736e-05, "loss": 0.6319, "step": 292220 }, { "epoch": 25.31, "learning_rate": 1.1741789059789521e-05, "loss": 0.6599, "step": 292230 }, { "epoch": 25.31, "learning_rate": 1.1739620677691684e-05, "loss": 0.4681, "step": 292240 }, { "epoch": 25.31, "learning_rate": 1.1737452295593847e-05, "loss": 0.5413, "step": 292250 }, { "epoch": 25.31, "learning_rate": 1.1735283913496009e-05, "loss": 0.7311, "step": 292260 }, { "epoch": 25.31, "learning_rate": 1.1733115531398172e-05, "loss": 0.6464, "step": 292270 }, { "epoch": 25.31, "learning_rate": 1.1730947149300333e-05, "loss": 0.5499, "step": 292280 }, { "epoch": 25.31, "learning_rate": 1.1728778767202498e-05, "loss": 0.5179, "step": 292290 }, { "epoch": 25.32, "learning_rate": 1.1726610385104659e-05, "loss": 0.6936, "step": 292300 }, { "epoch": 25.32, "learning_rate": 1.1724442003006822e-05, "loss": 0.6235, "step": 292310 }, { "epoch": 25.32, "learning_rate": 1.1722273620908983e-05, "loss": 0.4708, "step": 292320 }, { "epoch": 25.32, "learning_rate": 1.1720105238811148e-05, "loss": 0.6677, "step": 292330 }, { "epoch": 25.32, "learning_rate": 1.171793685671331e-05, "loss": 0.5562, "step": 292340 }, { "epoch": 25.32, "learning_rate": 1.1715768474615472e-05, "loss": 0.5658, "step": 292350 }, { "epoch": 25.32, "learning_rate": 1.1713600092517635e-05, "loss": 0.6141, "step": 292360 }, { "epoch": 25.32, "learning_rate": 1.1711431710419798e-05, "loss": 0.7303, "step": 292370 }, { "epoch": 25.32, "learning_rate": 1.170926332832196e-05, "loss": 0.721, "step": 292380 }, { "epoch": 25.32, "learning_rate": 1.1707094946224124e-05, "loss": 0.6418, "step": 292390 }, { "epoch": 25.32, "learning_rate": 1.1704926564126286e-05, "loss": 0.6288, "step": 292400 }, { "epoch": 25.32, "learning_rate": 1.1702758182028449e-05, "loss": 0.7179, "step": 292410 }, { "epoch": 25.33, "learning_rate": 1.170058979993061e-05, "loss": 0.436, "step": 292420 }, { "epoch": 25.33, "learning_rate": 1.1698421417832775e-05, "loss": 0.6481, "step": 292430 }, { "epoch": 25.33, "learning_rate": 1.1696253035734936e-05, "loss": 0.532, "step": 292440 }, { "epoch": 25.33, "learning_rate": 1.1694084653637099e-05, "loss": 0.5135, "step": 292450 }, { "epoch": 25.33, "learning_rate": 1.169191627153926e-05, "loss": 0.5273, "step": 292460 }, { "epoch": 25.33, "learning_rate": 1.1689747889441425e-05, "loss": 0.6476, "step": 292470 }, { "epoch": 25.33, "learning_rate": 1.1687579507343586e-05, "loss": 0.6888, "step": 292480 }, { "epoch": 25.33, "learning_rate": 1.168541112524575e-05, "loss": 0.7313, "step": 292490 }, { "epoch": 25.33, "learning_rate": 1.1683242743147912e-05, "loss": 0.5368, "step": 292500 }, { "epoch": 25.33, "learning_rate": 1.1681074361050074e-05, "loss": 0.7708, "step": 292510 }, { "epoch": 25.33, "learning_rate": 1.1678905978952237e-05, "loss": 0.5944, "step": 292520 }, { "epoch": 25.33, "learning_rate": 1.1676737596854398e-05, "loss": 0.6461, "step": 292530 }, { "epoch": 25.34, "learning_rate": 1.1674569214756563e-05, "loss": 0.6765, "step": 292540 }, { "epoch": 25.34, "learning_rate": 1.1672400832658724e-05, "loss": 0.5107, "step": 292550 }, { "epoch": 25.34, "learning_rate": 1.1670232450560887e-05, "loss": 0.6075, "step": 292560 }, { "epoch": 25.34, "learning_rate": 1.1668064068463048e-05, "loss": 0.7341, "step": 292570 }, { "epoch": 25.34, "learning_rate": 1.1665895686365213e-05, "loss": 0.53, "step": 292580 }, { "epoch": 25.34, "learning_rate": 1.1663727304267374e-05, "loss": 0.5959, "step": 292590 }, { "epoch": 25.34, "learning_rate": 1.1661558922169537e-05, "loss": 0.6133, "step": 292600 }, { "epoch": 25.34, "learning_rate": 1.16593905400717e-05, "loss": 0.5216, "step": 292610 }, { "epoch": 25.34, "learning_rate": 1.1657222157973863e-05, "loss": 0.6253, "step": 292620 }, { "epoch": 25.34, "learning_rate": 1.1655053775876024e-05, "loss": 0.5522, "step": 292630 }, { "epoch": 25.34, "learning_rate": 1.165288539377819e-05, "loss": 0.6399, "step": 292640 }, { "epoch": 25.35, "learning_rate": 1.165071701168035e-05, "loss": 0.5793, "step": 292650 }, { "epoch": 25.35, "learning_rate": 1.1648548629582514e-05, "loss": 0.625, "step": 292660 }, { "epoch": 25.35, "learning_rate": 1.1646380247484675e-05, "loss": 0.6258, "step": 292670 }, { "epoch": 25.35, "learning_rate": 1.164421186538684e-05, "loss": 0.5166, "step": 292680 }, { "epoch": 25.35, "learning_rate": 1.1642043483289e-05, "loss": 0.5831, "step": 292690 }, { "epoch": 25.35, "learning_rate": 1.1639875101191164e-05, "loss": 0.5662, "step": 292700 }, { "epoch": 25.35, "learning_rate": 1.1637706719093325e-05, "loss": 0.563, "step": 292710 }, { "epoch": 25.35, "learning_rate": 1.163553833699549e-05, "loss": 0.6052, "step": 292720 }, { "epoch": 25.35, "learning_rate": 1.1633369954897651e-05, "loss": 0.6499, "step": 292730 }, { "epoch": 25.35, "learning_rate": 1.1631201572799814e-05, "loss": 0.7323, "step": 292740 }, { "epoch": 25.35, "learning_rate": 1.1629033190701977e-05, "loss": 0.6395, "step": 292750 }, { "epoch": 25.35, "learning_rate": 1.1626864808604138e-05, "loss": 0.7274, "step": 292760 }, { "epoch": 25.36, "learning_rate": 1.1624696426506301e-05, "loss": 0.678, "step": 292770 }, { "epoch": 25.36, "learning_rate": 1.1622528044408463e-05, "loss": 0.5294, "step": 292780 }, { "epoch": 25.36, "learning_rate": 1.1620359662310627e-05, "loss": 0.601, "step": 292790 }, { "epoch": 25.36, "learning_rate": 1.1618191280212789e-05, "loss": 0.6404, "step": 292800 }, { "epoch": 25.36, "learning_rate": 1.1616022898114952e-05, "loss": 0.603, "step": 292810 }, { "epoch": 25.36, "learning_rate": 1.1613854516017115e-05, "loss": 0.6893, "step": 292820 }, { "epoch": 25.36, "learning_rate": 1.1611686133919278e-05, "loss": 0.5516, "step": 292830 }, { "epoch": 25.36, "learning_rate": 1.1609517751821439e-05, "loss": 0.5646, "step": 292840 }, { "epoch": 25.36, "learning_rate": 1.1607349369723602e-05, "loss": 0.6997, "step": 292850 }, { "epoch": 25.36, "learning_rate": 1.1605180987625765e-05, "loss": 0.5554, "step": 292860 }, { "epoch": 25.36, "learning_rate": 1.1603012605527928e-05, "loss": 0.6558, "step": 292870 }, { "epoch": 25.37, "learning_rate": 1.160084422343009e-05, "loss": 0.6643, "step": 292880 }, { "epoch": 25.37, "learning_rate": 1.1598675841332254e-05, "loss": 0.579, "step": 292890 }, { "epoch": 25.37, "learning_rate": 1.1596507459234415e-05, "loss": 0.5816, "step": 292900 }, { "epoch": 25.37, "learning_rate": 1.1594339077136578e-05, "loss": 0.785, "step": 292910 }, { "epoch": 25.37, "learning_rate": 1.159217069503874e-05, "loss": 0.4539, "step": 292920 }, { "epoch": 25.37, "learning_rate": 1.1590002312940904e-05, "loss": 0.7666, "step": 292930 }, { "epoch": 25.37, "learning_rate": 1.1587833930843066e-05, "loss": 0.5156, "step": 292940 }, { "epoch": 25.37, "learning_rate": 1.1585665548745229e-05, "loss": 0.5585, "step": 292950 }, { "epoch": 25.37, "learning_rate": 1.1583497166647392e-05, "loss": 0.6537, "step": 292960 }, { "epoch": 25.37, "learning_rate": 1.1581328784549555e-05, "loss": 0.6404, "step": 292970 }, { "epoch": 25.37, "learning_rate": 1.1579160402451716e-05, "loss": 0.6446, "step": 292980 }, { "epoch": 25.37, "learning_rate": 1.1576992020353879e-05, "loss": 0.5882, "step": 292990 }, { "epoch": 25.38, "learning_rate": 1.1574823638256042e-05, "loss": 0.6595, "step": 293000 }, { "epoch": 25.38, "learning_rate": 1.1572655256158203e-05, "loss": 0.5307, "step": 293010 }, { "epoch": 25.38, "learning_rate": 1.1570486874060366e-05, "loss": 0.6408, "step": 293020 }, { "epoch": 25.38, "learning_rate": 1.1568318491962528e-05, "loss": 0.6428, "step": 293030 }, { "epoch": 25.38, "learning_rate": 1.1566150109864692e-05, "loss": 0.625, "step": 293040 }, { "epoch": 25.38, "learning_rate": 1.1563981727766854e-05, "loss": 0.591, "step": 293050 }, { "epoch": 25.38, "learning_rate": 1.1561813345669017e-05, "loss": 0.6249, "step": 293060 }, { "epoch": 25.38, "learning_rate": 1.155964496357118e-05, "loss": 0.6134, "step": 293070 }, { "epoch": 25.38, "learning_rate": 1.1557476581473343e-05, "loss": 0.5537, "step": 293080 }, { "epoch": 25.38, "learning_rate": 1.1555308199375504e-05, "loss": 0.6336, "step": 293090 }, { "epoch": 25.38, "learning_rate": 1.1553139817277669e-05, "loss": 0.5997, "step": 293100 }, { "epoch": 25.39, "learning_rate": 1.155097143517983e-05, "loss": 0.5601, "step": 293110 }, { "epoch": 25.39, "learning_rate": 1.1548803053081993e-05, "loss": 0.5818, "step": 293120 }, { "epoch": 25.39, "learning_rate": 1.1546634670984154e-05, "loss": 0.5667, "step": 293130 }, { "epoch": 25.39, "learning_rate": 1.1544466288886319e-05, "loss": 0.673, "step": 293140 }, { "epoch": 25.39, "learning_rate": 1.154229790678848e-05, "loss": 0.5728, "step": 293150 }, { "epoch": 25.39, "learning_rate": 1.1540129524690643e-05, "loss": 0.6391, "step": 293160 }, { "epoch": 25.39, "learning_rate": 1.1537961142592805e-05, "loss": 0.5673, "step": 293170 }, { "epoch": 25.39, "learning_rate": 1.153579276049497e-05, "loss": 0.4693, "step": 293180 }, { "epoch": 25.39, "learning_rate": 1.153362437839713e-05, "loss": 0.6725, "step": 293190 }, { "epoch": 25.39, "learning_rate": 1.1531455996299294e-05, "loss": 0.5599, "step": 293200 }, { "epoch": 25.39, "learning_rate": 1.1529287614201457e-05, "loss": 0.62, "step": 293210 }, { "epoch": 25.39, "learning_rate": 1.152711923210362e-05, "loss": 0.6036, "step": 293220 }, { "epoch": 25.4, "learning_rate": 1.1524950850005781e-05, "loss": 0.5626, "step": 293230 }, { "epoch": 25.4, "learning_rate": 1.1522782467907946e-05, "loss": 0.6729, "step": 293240 }, { "epoch": 25.4, "learning_rate": 1.1520614085810107e-05, "loss": 0.676, "step": 293250 }, { "epoch": 25.4, "learning_rate": 1.151844570371227e-05, "loss": 0.5449, "step": 293260 }, { "epoch": 25.4, "learning_rate": 1.1516277321614431e-05, "loss": 0.5903, "step": 293270 }, { "epoch": 25.4, "learning_rate": 1.1514108939516593e-05, "loss": 0.5518, "step": 293280 }, { "epoch": 25.4, "learning_rate": 1.1511940557418757e-05, "loss": 0.6248, "step": 293290 }, { "epoch": 25.4, "learning_rate": 1.1509772175320919e-05, "loss": 0.6548, "step": 293300 }, { "epoch": 25.4, "learning_rate": 1.1507603793223082e-05, "loss": 0.5688, "step": 293310 }, { "epoch": 25.4, "learning_rate": 1.1505435411125245e-05, "loss": 0.6169, "step": 293320 }, { "epoch": 25.4, "learning_rate": 1.1503267029027408e-05, "loss": 0.8041, "step": 293330 }, { "epoch": 25.41, "learning_rate": 1.1501098646929569e-05, "loss": 0.5029, "step": 293340 }, { "epoch": 25.41, "learning_rate": 1.1498930264831734e-05, "loss": 0.6606, "step": 293350 }, { "epoch": 25.41, "learning_rate": 1.1496761882733895e-05, "loss": 0.5872, "step": 293360 }, { "epoch": 25.41, "learning_rate": 1.1494593500636058e-05, "loss": 0.665, "step": 293370 }, { "epoch": 25.41, "learning_rate": 1.149242511853822e-05, "loss": 0.6501, "step": 293380 }, { "epoch": 25.41, "learning_rate": 1.1490256736440384e-05, "loss": 0.6305, "step": 293390 }, { "epoch": 25.41, "learning_rate": 1.1488088354342545e-05, "loss": 0.5515, "step": 293400 }, { "epoch": 25.41, "learning_rate": 1.1485919972244708e-05, "loss": 0.5783, "step": 293410 }, { "epoch": 25.41, "learning_rate": 1.148375159014687e-05, "loss": 0.5131, "step": 293420 }, { "epoch": 25.41, "learning_rate": 1.1481583208049034e-05, "loss": 0.5703, "step": 293430 }, { "epoch": 25.41, "learning_rate": 1.1479414825951196e-05, "loss": 0.5997, "step": 293440 }, { "epoch": 25.41, "learning_rate": 1.1477246443853359e-05, "loss": 0.5628, "step": 293450 }, { "epoch": 25.42, "learning_rate": 1.1475078061755522e-05, "loss": 0.6243, "step": 293460 }, { "epoch": 25.42, "learning_rate": 1.1472909679657685e-05, "loss": 0.5174, "step": 293470 }, { "epoch": 25.42, "learning_rate": 1.1470741297559846e-05, "loss": 0.5774, "step": 293480 }, { "epoch": 25.42, "learning_rate": 1.146857291546201e-05, "loss": 0.5912, "step": 293490 }, { "epoch": 25.42, "learning_rate": 1.1466404533364172e-05, "loss": 0.6721, "step": 293500 }, { "epoch": 25.42, "learning_rate": 1.1464236151266335e-05, "loss": 0.5812, "step": 293510 }, { "epoch": 25.42, "learning_rate": 1.1462067769168496e-05, "loss": 0.6234, "step": 293520 }, { "epoch": 25.42, "learning_rate": 1.1459899387070658e-05, "loss": 0.6516, "step": 293530 }, { "epoch": 25.42, "learning_rate": 1.1457731004972822e-05, "loss": 0.6662, "step": 293540 }, { "epoch": 25.42, "learning_rate": 1.1455562622874984e-05, "loss": 0.5168, "step": 293550 }, { "epoch": 25.42, "learning_rate": 1.1453394240777147e-05, "loss": 0.6073, "step": 293560 }, { "epoch": 25.43, "learning_rate": 1.145122585867931e-05, "loss": 0.7883, "step": 293570 }, { "epoch": 25.43, "learning_rate": 1.1449057476581473e-05, "loss": 0.6752, "step": 293580 }, { "epoch": 25.43, "learning_rate": 1.1446889094483634e-05, "loss": 0.5887, "step": 293590 }, { "epoch": 25.43, "learning_rate": 1.1444720712385799e-05, "loss": 0.6404, "step": 293600 }, { "epoch": 25.43, "learning_rate": 1.144255233028796e-05, "loss": 0.6658, "step": 293610 }, { "epoch": 25.43, "learning_rate": 1.1440383948190123e-05, "loss": 0.6027, "step": 293620 }, { "epoch": 25.43, "learning_rate": 1.1438215566092284e-05, "loss": 0.6027, "step": 293630 }, { "epoch": 25.43, "learning_rate": 1.1436047183994449e-05, "loss": 0.5706, "step": 293640 }, { "epoch": 25.43, "learning_rate": 1.143387880189661e-05, "loss": 0.7004, "step": 293650 }, { "epoch": 25.43, "learning_rate": 1.1431710419798773e-05, "loss": 0.6728, "step": 293660 }, { "epoch": 25.43, "learning_rate": 1.1429542037700935e-05, "loss": 0.7155, "step": 293670 }, { "epoch": 25.43, "learning_rate": 1.14273736556031e-05, "loss": 0.7115, "step": 293680 }, { "epoch": 25.44, "learning_rate": 1.142520527350526e-05, "loss": 0.6588, "step": 293690 }, { "epoch": 25.44, "learning_rate": 1.1423036891407424e-05, "loss": 0.5863, "step": 293700 }, { "epoch": 25.44, "learning_rate": 1.1420868509309587e-05, "loss": 0.6227, "step": 293710 }, { "epoch": 25.44, "learning_rate": 1.141870012721175e-05, "loss": 0.457, "step": 293720 }, { "epoch": 25.44, "learning_rate": 1.141653174511391e-05, "loss": 0.5535, "step": 293730 }, { "epoch": 25.44, "learning_rate": 1.1414363363016076e-05, "loss": 0.5602, "step": 293740 }, { "epoch": 25.44, "learning_rate": 1.1412194980918237e-05, "loss": 0.6561, "step": 293750 }, { "epoch": 25.44, "learning_rate": 1.14100265988204e-05, "loss": 0.5869, "step": 293760 }, { "epoch": 25.44, "learning_rate": 1.1407858216722561e-05, "loss": 0.4136, "step": 293770 }, { "epoch": 25.44, "learning_rate": 1.1405689834624722e-05, "loss": 0.614, "step": 293780 }, { "epoch": 25.44, "learning_rate": 1.1403521452526887e-05, "loss": 0.6016, "step": 293790 }, { "epoch": 25.44, "learning_rate": 1.1401353070429048e-05, "loss": 0.6739, "step": 293800 }, { "epoch": 25.45, "learning_rate": 1.1399184688331211e-05, "loss": 0.4816, "step": 293810 }, { "epoch": 25.45, "learning_rate": 1.1397016306233374e-05, "loss": 0.6881, "step": 293820 }, { "epoch": 25.45, "learning_rate": 1.1394847924135537e-05, "loss": 0.5874, "step": 293830 }, { "epoch": 25.45, "learning_rate": 1.1392679542037699e-05, "loss": 0.6214, "step": 293840 }, { "epoch": 25.45, "learning_rate": 1.1390511159939863e-05, "loss": 0.5688, "step": 293850 }, { "epoch": 25.45, "learning_rate": 1.1388342777842025e-05, "loss": 0.6495, "step": 293860 }, { "epoch": 25.45, "learning_rate": 1.1386174395744188e-05, "loss": 0.7737, "step": 293870 }, { "epoch": 25.45, "learning_rate": 1.1384006013646349e-05, "loss": 0.6746, "step": 293880 }, { "epoch": 25.45, "learning_rate": 1.1381837631548514e-05, "loss": 0.5589, "step": 293890 }, { "epoch": 25.45, "learning_rate": 1.1379669249450675e-05, "loss": 0.7725, "step": 293900 }, { "epoch": 25.45, "learning_rate": 1.1377500867352838e-05, "loss": 0.582, "step": 293910 }, { "epoch": 25.46, "learning_rate": 1.1375332485255e-05, "loss": 0.5083, "step": 293920 }, { "epoch": 25.46, "learning_rate": 1.1373164103157164e-05, "loss": 0.4956, "step": 293930 }, { "epoch": 25.46, "learning_rate": 1.1370995721059325e-05, "loss": 0.6298, "step": 293940 }, { "epoch": 25.46, "learning_rate": 1.1368827338961488e-05, "loss": 0.6726, "step": 293950 }, { "epoch": 25.46, "learning_rate": 1.1366658956863651e-05, "loss": 0.5231, "step": 293960 }, { "epoch": 25.46, "learning_rate": 1.1364490574765814e-05, "loss": 0.5785, "step": 293970 }, { "epoch": 25.46, "learning_rate": 1.1362322192667976e-05, "loss": 0.5377, "step": 293980 }, { "epoch": 25.46, "learning_rate": 1.136015381057014e-05, "loss": 0.5794, "step": 293990 }, { "epoch": 25.46, "learning_rate": 1.1357985428472302e-05, "loss": 0.6402, "step": 294000 }, { "epoch": 25.46, "learning_rate": 1.1355817046374465e-05, "loss": 0.5647, "step": 294010 }, { "epoch": 25.46, "learning_rate": 1.1353648664276626e-05, "loss": 0.4991, "step": 294020 }, { "epoch": 25.46, "learning_rate": 1.135148028217879e-05, "loss": 0.5861, "step": 294030 }, { "epoch": 25.47, "learning_rate": 1.1349311900080952e-05, "loss": 0.5755, "step": 294040 }, { "epoch": 25.47, "learning_rate": 1.1347143517983113e-05, "loss": 0.4887, "step": 294050 }, { "epoch": 25.47, "learning_rate": 1.1344975135885276e-05, "loss": 0.5226, "step": 294060 }, { "epoch": 25.47, "learning_rate": 1.134280675378744e-05, "loss": 0.7275, "step": 294070 }, { "epoch": 25.47, "learning_rate": 1.1340638371689602e-05, "loss": 0.6633, "step": 294080 }, { "epoch": 25.47, "learning_rate": 1.1338469989591764e-05, "loss": 0.6414, "step": 294090 }, { "epoch": 25.47, "learning_rate": 1.1336301607493928e-05, "loss": 0.7129, "step": 294100 }, { "epoch": 25.47, "learning_rate": 1.133413322539609e-05, "loss": 0.5388, "step": 294110 }, { "epoch": 25.47, "learning_rate": 1.1331964843298253e-05, "loss": 0.5226, "step": 294120 }, { "epoch": 25.47, "learning_rate": 1.1329796461200414e-05, "loss": 0.5729, "step": 294130 }, { "epoch": 25.47, "learning_rate": 1.1327628079102579e-05, "loss": 0.5589, "step": 294140 }, { "epoch": 25.48, "learning_rate": 1.132545969700474e-05, "loss": 0.6541, "step": 294150 }, { "epoch": 25.48, "learning_rate": 1.1323291314906903e-05, "loss": 0.6248, "step": 294160 }, { "epoch": 25.48, "learning_rate": 1.1321122932809066e-05, "loss": 0.6562, "step": 294170 }, { "epoch": 25.48, "learning_rate": 1.1318954550711229e-05, "loss": 0.594, "step": 294180 }, { "epoch": 25.48, "learning_rate": 1.131678616861339e-05, "loss": 0.577, "step": 294190 }, { "epoch": 25.48, "learning_rate": 1.1314617786515553e-05, "loss": 0.5266, "step": 294200 }, { "epoch": 25.48, "learning_rate": 1.1312449404417716e-05, "loss": 0.5689, "step": 294210 }, { "epoch": 25.48, "learning_rate": 1.131028102231988e-05, "loss": 0.7481, "step": 294220 }, { "epoch": 25.48, "learning_rate": 1.130811264022204e-05, "loss": 0.4978, "step": 294230 }, { "epoch": 25.48, "learning_rate": 1.1305944258124205e-05, "loss": 0.5632, "step": 294240 }, { "epoch": 25.48, "learning_rate": 1.1303775876026367e-05, "loss": 0.5892, "step": 294250 }, { "epoch": 25.48, "learning_rate": 1.130160749392853e-05, "loss": 0.5231, "step": 294260 }, { "epoch": 25.49, "learning_rate": 1.1299439111830691e-05, "loss": 0.6355, "step": 294270 }, { "epoch": 25.49, "learning_rate": 1.1297270729732856e-05, "loss": 0.6077, "step": 294280 }, { "epoch": 25.49, "learning_rate": 1.1295102347635017e-05, "loss": 0.6128, "step": 294290 }, { "epoch": 25.49, "learning_rate": 1.1292933965537178e-05, "loss": 0.7008, "step": 294300 }, { "epoch": 25.49, "learning_rate": 1.1290765583439343e-05, "loss": 0.6439, "step": 294310 }, { "epoch": 25.49, "learning_rate": 1.1288597201341504e-05, "loss": 0.6028, "step": 294320 }, { "epoch": 25.49, "learning_rate": 1.1286428819243667e-05, "loss": 0.7041, "step": 294330 }, { "epoch": 25.49, "learning_rate": 1.1284260437145829e-05, "loss": 0.7169, "step": 294340 }, { "epoch": 25.49, "learning_rate": 1.1282092055047993e-05, "loss": 0.5473, "step": 294350 }, { "epoch": 25.49, "learning_rate": 1.1279923672950155e-05, "loss": 0.6339, "step": 294360 }, { "epoch": 25.49, "learning_rate": 1.1277755290852318e-05, "loss": 0.6192, "step": 294370 }, { "epoch": 25.5, "learning_rate": 1.1275586908754479e-05, "loss": 0.8246, "step": 294380 }, { "epoch": 25.5, "learning_rate": 1.1273418526656644e-05, "loss": 0.644, "step": 294390 }, { "epoch": 25.5, "learning_rate": 1.1271250144558805e-05, "loss": 0.6153, "step": 294400 }, { "epoch": 25.5, "learning_rate": 1.1269081762460968e-05, "loss": 0.5599, "step": 294410 }, { "epoch": 25.5, "learning_rate": 1.1266913380363131e-05, "loss": 0.6572, "step": 294420 }, { "epoch": 25.5, "learning_rate": 1.1264744998265294e-05, "loss": 0.57, "step": 294430 }, { "epoch": 25.5, "learning_rate": 1.1262576616167455e-05, "loss": 0.6546, "step": 294440 }, { "epoch": 25.5, "learning_rate": 1.126040823406962e-05, "loss": 0.6303, "step": 294450 }, { "epoch": 25.5, "learning_rate": 1.1258239851971781e-05, "loss": 0.6153, "step": 294460 }, { "epoch": 25.5, "learning_rate": 1.1256071469873944e-05, "loss": 0.6322, "step": 294470 }, { "epoch": 25.5, "learning_rate": 1.1253903087776106e-05, "loss": 0.602, "step": 294480 }, { "epoch": 25.5, "learning_rate": 1.125173470567827e-05, "loss": 0.6738, "step": 294490 }, { "epoch": 25.51, "learning_rate": 1.1249566323580432e-05, "loss": 0.7376, "step": 294500 }, { "epoch": 25.51, "learning_rate": 1.1247397941482595e-05, "loss": 0.6262, "step": 294510 }, { "epoch": 25.51, "learning_rate": 1.1245229559384756e-05, "loss": 0.7305, "step": 294520 }, { "epoch": 25.51, "learning_rate": 1.124306117728692e-05, "loss": 0.7019, "step": 294530 }, { "epoch": 25.51, "learning_rate": 1.1240892795189082e-05, "loss": 0.6431, "step": 294540 }, { "epoch": 25.51, "learning_rate": 1.1238724413091243e-05, "loss": 0.6544, "step": 294550 }, { "epoch": 25.51, "learning_rate": 1.1236556030993408e-05, "loss": 0.6359, "step": 294560 }, { "epoch": 25.51, "learning_rate": 1.123438764889557e-05, "loss": 0.5664, "step": 294570 }, { "epoch": 25.51, "learning_rate": 1.1232219266797732e-05, "loss": 0.5494, "step": 294580 }, { "epoch": 25.51, "learning_rate": 1.1230050884699894e-05, "loss": 0.7064, "step": 294590 }, { "epoch": 25.51, "learning_rate": 1.1227882502602058e-05, "loss": 0.6004, "step": 294600 }, { "epoch": 25.52, "learning_rate": 1.122571412050422e-05, "loss": 0.5231, "step": 294610 }, { "epoch": 25.52, "learning_rate": 1.1223545738406383e-05, "loss": 0.4799, "step": 294620 }, { "epoch": 25.52, "learning_rate": 1.1221377356308544e-05, "loss": 0.617, "step": 294630 }, { "epoch": 25.52, "learning_rate": 1.1219208974210709e-05, "loss": 0.5413, "step": 294640 }, { "epoch": 25.52, "learning_rate": 1.121704059211287e-05, "loss": 0.5546, "step": 294650 }, { "epoch": 25.52, "learning_rate": 1.1214872210015033e-05, "loss": 0.567, "step": 294660 }, { "epoch": 25.52, "learning_rate": 1.1212703827917196e-05, "loss": 0.5729, "step": 294670 }, { "epoch": 25.52, "learning_rate": 1.1210535445819359e-05, "loss": 0.5598, "step": 294680 }, { "epoch": 25.52, "learning_rate": 1.120836706372152e-05, "loss": 0.692, "step": 294690 }, { "epoch": 25.52, "learning_rate": 1.1206198681623685e-05, "loss": 0.7012, "step": 294700 }, { "epoch": 25.52, "learning_rate": 1.1204030299525846e-05, "loss": 0.5941, "step": 294710 }, { "epoch": 25.52, "learning_rate": 1.120186191742801e-05, "loss": 0.6165, "step": 294720 }, { "epoch": 25.53, "learning_rate": 1.119969353533017e-05, "loss": 0.6855, "step": 294730 }, { "epoch": 25.53, "learning_rate": 1.1197525153232335e-05, "loss": 0.5551, "step": 294740 }, { "epoch": 25.53, "learning_rate": 1.1195356771134497e-05, "loss": 0.5639, "step": 294750 }, { "epoch": 25.53, "learning_rate": 1.119318838903666e-05, "loss": 0.7568, "step": 294760 }, { "epoch": 25.53, "learning_rate": 1.1191020006938821e-05, "loss": 0.609, "step": 294770 }, { "epoch": 25.53, "learning_rate": 1.1188851624840986e-05, "loss": 0.5851, "step": 294780 }, { "epoch": 25.53, "learning_rate": 1.1186683242743147e-05, "loss": 0.5942, "step": 294790 }, { "epoch": 25.53, "learning_rate": 1.118451486064531e-05, "loss": 0.632, "step": 294800 }, { "epoch": 25.53, "learning_rate": 1.1182346478547473e-05, "loss": 0.465, "step": 294810 }, { "epoch": 25.53, "learning_rate": 1.1180178096449634e-05, "loss": 0.719, "step": 294820 }, { "epoch": 25.53, "learning_rate": 1.1178009714351797e-05, "loss": 0.617, "step": 294830 }, { "epoch": 25.54, "learning_rate": 1.1175841332253958e-05, "loss": 0.5252, "step": 294840 }, { "epoch": 25.54, "learning_rate": 1.1173672950156123e-05, "loss": 0.5855, "step": 294850 }, { "epoch": 25.54, "learning_rate": 1.1171504568058284e-05, "loss": 0.6955, "step": 294860 }, { "epoch": 25.54, "learning_rate": 1.1169336185960447e-05, "loss": 0.5779, "step": 294870 }, { "epoch": 25.54, "learning_rate": 1.1167167803862609e-05, "loss": 0.5873, "step": 294880 }, { "epoch": 25.54, "learning_rate": 1.1164999421764774e-05, "loss": 0.7257, "step": 294890 }, { "epoch": 25.54, "learning_rate": 1.1162831039666935e-05, "loss": 0.6655, "step": 294900 }, { "epoch": 25.54, "learning_rate": 1.1160662657569098e-05, "loss": 0.6243, "step": 294910 }, { "epoch": 25.54, "learning_rate": 1.115849427547126e-05, "loss": 0.7061, "step": 294920 }, { "epoch": 25.54, "learning_rate": 1.1156325893373424e-05, "loss": 0.5795, "step": 294930 }, { "epoch": 25.54, "learning_rate": 1.1154157511275585e-05, "loss": 0.6031, "step": 294940 }, { "epoch": 25.54, "learning_rate": 1.115198912917775e-05, "loss": 0.5781, "step": 294950 }, { "epoch": 25.55, "learning_rate": 1.1149820747079911e-05, "loss": 0.6144, "step": 294960 }, { "epoch": 25.55, "learning_rate": 1.1147652364982074e-05, "loss": 0.6615, "step": 294970 }, { "epoch": 25.55, "learning_rate": 1.1145483982884235e-05, "loss": 0.6139, "step": 294980 }, { "epoch": 25.55, "learning_rate": 1.11433156007864e-05, "loss": 0.6299, "step": 294990 }, { "epoch": 25.55, "learning_rate": 1.1141147218688561e-05, "loss": 0.7936, "step": 295000 }, { "epoch": 25.55, "learning_rate": 1.1138978836590724e-05, "loss": 0.6151, "step": 295010 }, { "epoch": 25.55, "learning_rate": 1.1136810454492886e-05, "loss": 0.5918, "step": 295020 }, { "epoch": 25.55, "learning_rate": 1.113464207239505e-05, "loss": 0.5759, "step": 295030 }, { "epoch": 25.55, "learning_rate": 1.1132473690297212e-05, "loss": 0.5206, "step": 295040 }, { "epoch": 25.55, "learning_rate": 1.1130305308199375e-05, "loss": 0.655, "step": 295050 }, { "epoch": 25.55, "learning_rate": 1.1128136926101538e-05, "loss": 0.6117, "step": 295060 }, { "epoch": 25.55, "learning_rate": 1.1125968544003699e-05, "loss": 0.6328, "step": 295070 }, { "epoch": 25.56, "learning_rate": 1.1123800161905862e-05, "loss": 0.6249, "step": 295080 }, { "epoch": 25.56, "learning_rate": 1.1121631779808023e-05, "loss": 0.6581, "step": 295090 }, { "epoch": 25.56, "learning_rate": 1.1119463397710188e-05, "loss": 0.6132, "step": 295100 }, { "epoch": 25.56, "learning_rate": 1.111729501561235e-05, "loss": 0.6997, "step": 295110 }, { "epoch": 25.56, "learning_rate": 1.1115126633514512e-05, "loss": 0.5237, "step": 295120 }, { "epoch": 25.56, "learning_rate": 1.1112958251416674e-05, "loss": 0.6427, "step": 295130 }, { "epoch": 25.56, "learning_rate": 1.1110789869318838e-05, "loss": 0.6804, "step": 295140 }, { "epoch": 25.56, "learning_rate": 1.1108621487221e-05, "loss": 0.6077, "step": 295150 }, { "epoch": 25.56, "learning_rate": 1.1106453105123163e-05, "loss": 0.6076, "step": 295160 }, { "epoch": 25.56, "learning_rate": 1.1104284723025326e-05, "loss": 0.6666, "step": 295170 }, { "epoch": 25.56, "learning_rate": 1.1102116340927489e-05, "loss": 0.5846, "step": 295180 }, { "epoch": 25.57, "learning_rate": 1.109994795882965e-05, "loss": 0.588, "step": 295190 }, { "epoch": 25.57, "learning_rate": 1.1097779576731815e-05, "loss": 0.6918, "step": 295200 }, { "epoch": 25.57, "learning_rate": 1.1095611194633976e-05, "loss": 0.4591, "step": 295210 }, { "epoch": 25.57, "learning_rate": 1.1093442812536139e-05, "loss": 0.6747, "step": 295220 }, { "epoch": 25.57, "learning_rate": 1.10912744304383e-05, "loss": 0.6575, "step": 295230 }, { "epoch": 25.57, "learning_rate": 1.1089106048340465e-05, "loss": 0.5476, "step": 295240 }, { "epoch": 25.57, "learning_rate": 1.1086937666242626e-05, "loss": 0.6694, "step": 295250 }, { "epoch": 25.57, "learning_rate": 1.108476928414479e-05, "loss": 0.7989, "step": 295260 }, { "epoch": 25.57, "learning_rate": 1.108260090204695e-05, "loss": 0.6684, "step": 295270 }, { "epoch": 25.57, "learning_rate": 1.1080432519949115e-05, "loss": 0.5059, "step": 295280 }, { "epoch": 25.57, "learning_rate": 1.1078264137851277e-05, "loss": 0.6596, "step": 295290 }, { "epoch": 25.57, "learning_rate": 1.107609575575344e-05, "loss": 0.565, "step": 295300 }, { "epoch": 25.58, "learning_rate": 1.1073927373655603e-05, "loss": 0.7573, "step": 295310 }, { "epoch": 25.58, "learning_rate": 1.1071758991557764e-05, "loss": 0.6097, "step": 295320 }, { "epoch": 25.58, "learning_rate": 1.1069590609459927e-05, "loss": 0.5653, "step": 295330 }, { "epoch": 25.58, "learning_rate": 1.1067422227362088e-05, "loss": 0.581, "step": 295340 }, { "epoch": 25.58, "learning_rate": 1.1065253845264253e-05, "loss": 0.6781, "step": 295350 }, { "epoch": 25.58, "learning_rate": 1.1063085463166414e-05, "loss": 0.5936, "step": 295360 }, { "epoch": 25.58, "learning_rate": 1.1060917081068577e-05, "loss": 0.6725, "step": 295370 }, { "epoch": 25.58, "learning_rate": 1.105874869897074e-05, "loss": 0.6429, "step": 295380 }, { "epoch": 25.58, "learning_rate": 1.1056580316872903e-05, "loss": 0.6466, "step": 295390 }, { "epoch": 25.58, "learning_rate": 1.1054411934775065e-05, "loss": 0.6152, "step": 295400 }, { "epoch": 25.58, "learning_rate": 1.1052243552677228e-05, "loss": 0.5445, "step": 295410 }, { "epoch": 25.59, "learning_rate": 1.105007517057939e-05, "loss": 0.5395, "step": 295420 }, { "epoch": 25.59, "learning_rate": 1.1047906788481554e-05, "loss": 0.5795, "step": 295430 }, { "epoch": 25.59, "learning_rate": 1.1045738406383715e-05, "loss": 0.5348, "step": 295440 }, { "epoch": 25.59, "learning_rate": 1.104357002428588e-05, "loss": 0.5222, "step": 295450 }, { "epoch": 25.59, "learning_rate": 1.1041401642188041e-05, "loss": 0.5971, "step": 295460 }, { "epoch": 25.59, "learning_rate": 1.1039233260090204e-05, "loss": 0.5875, "step": 295470 }, { "epoch": 25.59, "learning_rate": 1.1037064877992365e-05, "loss": 0.5413, "step": 295480 }, { "epoch": 25.59, "learning_rate": 1.103489649589453e-05, "loss": 0.4963, "step": 295490 }, { "epoch": 25.59, "learning_rate": 1.1032728113796691e-05, "loss": 0.6142, "step": 295500 }, { "epoch": 25.59, "learning_rate": 1.1030559731698854e-05, "loss": 0.7021, "step": 295510 }, { "epoch": 25.59, "learning_rate": 1.1028391349601017e-05, "loss": 0.5585, "step": 295520 }, { "epoch": 25.59, "learning_rate": 1.102622296750318e-05, "loss": 0.5622, "step": 295530 }, { "epoch": 25.6, "learning_rate": 1.1024054585405342e-05, "loss": 0.57, "step": 295540 }, { "epoch": 25.6, "learning_rate": 1.1021886203307505e-05, "loss": 0.5238, "step": 295550 }, { "epoch": 25.6, "learning_rate": 1.1019717821209668e-05, "loss": 0.6677, "step": 295560 }, { "epoch": 25.6, "learning_rate": 1.101754943911183e-05, "loss": 0.6226, "step": 295570 }, { "epoch": 25.6, "learning_rate": 1.1015381057013992e-05, "loss": 0.7462, "step": 295580 }, { "epoch": 25.6, "learning_rate": 1.1013212674916153e-05, "loss": 0.5032, "step": 295590 }, { "epoch": 25.6, "learning_rate": 1.1011044292818318e-05, "loss": 0.712, "step": 295600 }, { "epoch": 25.6, "learning_rate": 1.100887591072048e-05, "loss": 0.6557, "step": 295610 }, { "epoch": 25.6, "learning_rate": 1.1006707528622642e-05, "loss": 0.6634, "step": 295620 }, { "epoch": 25.6, "learning_rate": 1.1004539146524805e-05, "loss": 0.6002, "step": 295630 }, { "epoch": 25.6, "learning_rate": 1.1002370764426968e-05, "loss": 0.5428, "step": 295640 }, { "epoch": 25.61, "learning_rate": 1.100020238232913e-05, "loss": 0.6804, "step": 295650 }, { "epoch": 25.61, "learning_rate": 1.0998034000231294e-05, "loss": 0.6647, "step": 295660 }, { "epoch": 25.61, "learning_rate": 1.0995865618133456e-05, "loss": 0.7192, "step": 295670 }, { "epoch": 25.61, "learning_rate": 1.0993697236035619e-05, "loss": 0.66, "step": 295680 }, { "epoch": 25.61, "learning_rate": 1.099152885393778e-05, "loss": 0.6451, "step": 295690 }, { "epoch": 25.61, "learning_rate": 1.0989360471839945e-05, "loss": 0.5151, "step": 295700 }, { "epoch": 25.61, "learning_rate": 1.0987192089742106e-05, "loss": 0.6192, "step": 295710 }, { "epoch": 25.61, "learning_rate": 1.0985023707644269e-05, "loss": 0.6631, "step": 295720 }, { "epoch": 25.61, "learning_rate": 1.098285532554643e-05, "loss": 0.6126, "step": 295730 }, { "epoch": 25.61, "learning_rate": 1.0980686943448595e-05, "loss": 0.546, "step": 295740 }, { "epoch": 25.61, "learning_rate": 1.0978518561350756e-05, "loss": 0.5424, "step": 295750 }, { "epoch": 25.61, "learning_rate": 1.097635017925292e-05, "loss": 0.5262, "step": 295760 }, { "epoch": 25.62, "learning_rate": 1.0974181797155082e-05, "loss": 0.5278, "step": 295770 }, { "epoch": 25.62, "learning_rate": 1.0972013415057245e-05, "loss": 0.6486, "step": 295780 }, { "epoch": 25.62, "learning_rate": 1.0969845032959407e-05, "loss": 0.6434, "step": 295790 }, { "epoch": 25.62, "learning_rate": 1.0967676650861571e-05, "loss": 0.6119, "step": 295800 }, { "epoch": 25.62, "learning_rate": 1.0965508268763733e-05, "loss": 0.6063, "step": 295810 }, { "epoch": 25.62, "learning_rate": 1.0963339886665896e-05, "loss": 0.6772, "step": 295820 }, { "epoch": 25.62, "learning_rate": 1.0961171504568057e-05, "loss": 0.549, "step": 295830 }, { "epoch": 25.62, "learning_rate": 1.0959003122470218e-05, "loss": 0.5454, "step": 295840 }, { "epoch": 25.62, "learning_rate": 1.0956834740372383e-05, "loss": 0.6893, "step": 295850 }, { "epoch": 25.62, "learning_rate": 1.0954666358274544e-05, "loss": 0.6584, "step": 295860 }, { "epoch": 25.62, "learning_rate": 1.0952497976176707e-05, "loss": 0.7465, "step": 295870 }, { "epoch": 25.63, "learning_rate": 1.095032959407887e-05, "loss": 0.5626, "step": 295880 }, { "epoch": 25.63, "learning_rate": 1.0948161211981033e-05, "loss": 0.7442, "step": 295890 }, { "epoch": 25.63, "learning_rate": 1.0945992829883195e-05, "loss": 0.6139, "step": 295900 }, { "epoch": 25.63, "learning_rate": 1.094382444778536e-05, "loss": 0.481, "step": 295910 }, { "epoch": 25.63, "learning_rate": 1.094165606568752e-05, "loss": 0.5838, "step": 295920 }, { "epoch": 25.63, "learning_rate": 1.0939487683589684e-05, "loss": 0.8408, "step": 295930 }, { "epoch": 25.63, "learning_rate": 1.0937319301491845e-05, "loss": 0.579, "step": 295940 }, { "epoch": 25.63, "learning_rate": 1.093515091939401e-05, "loss": 0.6166, "step": 295950 }, { "epoch": 25.63, "learning_rate": 1.093298253729617e-05, "loss": 0.618, "step": 295960 }, { "epoch": 25.63, "learning_rate": 1.0930814155198334e-05, "loss": 0.5098, "step": 295970 }, { "epoch": 25.63, "learning_rate": 1.0928645773100495e-05, "loss": 0.5902, "step": 295980 }, { "epoch": 25.63, "learning_rate": 1.092647739100266e-05, "loss": 0.687, "step": 295990 }, { "epoch": 25.64, "learning_rate": 1.0924309008904821e-05, "loss": 0.5086, "step": 296000 }, { "epoch": 25.64, "learning_rate": 1.0922140626806984e-05, "loss": 0.6524, "step": 296010 }, { "epoch": 25.64, "learning_rate": 1.0919972244709147e-05, "loss": 0.5727, "step": 296020 }, { "epoch": 25.64, "learning_rate": 1.091780386261131e-05, "loss": 0.6757, "step": 296030 }, { "epoch": 25.64, "learning_rate": 1.0915635480513471e-05, "loss": 0.9439, "step": 296040 }, { "epoch": 25.64, "learning_rate": 1.0913467098415636e-05, "loss": 0.554, "step": 296050 }, { "epoch": 25.64, "learning_rate": 1.0911298716317797e-05, "loss": 0.6163, "step": 296060 }, { "epoch": 25.64, "learning_rate": 1.090913033421996e-05, "loss": 0.758, "step": 296070 }, { "epoch": 25.64, "learning_rate": 1.0906961952122122e-05, "loss": 0.6709, "step": 296080 }, { "epoch": 25.64, "learning_rate": 1.0904793570024283e-05, "loss": 0.6117, "step": 296090 }, { "epoch": 25.64, "learning_rate": 1.0902625187926448e-05, "loss": 0.7007, "step": 296100 }, { "epoch": 25.65, "learning_rate": 1.0900456805828609e-05, "loss": 0.4449, "step": 296110 }, { "epoch": 25.65, "learning_rate": 1.0898288423730772e-05, "loss": 0.6883, "step": 296120 }, { "epoch": 25.65, "learning_rate": 1.0896120041632935e-05, "loss": 0.5992, "step": 296130 }, { "epoch": 25.65, "learning_rate": 1.0893951659535098e-05, "loss": 0.5959, "step": 296140 }, { "epoch": 25.65, "learning_rate": 1.089178327743726e-05, "loss": 0.5763, "step": 296150 }, { "epoch": 25.65, "learning_rate": 1.0889614895339424e-05, "loss": 0.5474, "step": 296160 }, { "epoch": 25.65, "learning_rate": 1.0887446513241585e-05, "loss": 0.6629, "step": 296170 }, { "epoch": 25.65, "learning_rate": 1.0885278131143748e-05, "loss": 0.4753, "step": 296180 }, { "epoch": 25.65, "learning_rate": 1.088310974904591e-05, "loss": 0.5718, "step": 296190 }, { "epoch": 25.65, "learning_rate": 1.0880941366948074e-05, "loss": 0.4507, "step": 296200 }, { "epoch": 25.65, "learning_rate": 1.0878772984850236e-05, "loss": 0.7552, "step": 296210 }, { "epoch": 25.65, "learning_rate": 1.0876604602752399e-05, "loss": 0.7853, "step": 296220 }, { "epoch": 25.66, "learning_rate": 1.087443622065456e-05, "loss": 0.6887, "step": 296230 }, { "epoch": 25.66, "learning_rate": 1.0872267838556725e-05, "loss": 0.4486, "step": 296240 }, { "epoch": 25.66, "learning_rate": 1.0870099456458886e-05, "loss": 0.5298, "step": 296250 }, { "epoch": 25.66, "learning_rate": 1.0867931074361049e-05, "loss": 0.4618, "step": 296260 }, { "epoch": 25.66, "learning_rate": 1.0865762692263212e-05, "loss": 0.6924, "step": 296270 }, { "epoch": 25.66, "learning_rate": 1.0863594310165375e-05, "loss": 0.6768, "step": 296280 }, { "epoch": 25.66, "learning_rate": 1.0861425928067536e-05, "loss": 0.5047, "step": 296290 }, { "epoch": 25.66, "learning_rate": 1.0859257545969701e-05, "loss": 0.6273, "step": 296300 }, { "epoch": 25.66, "learning_rate": 1.0857089163871862e-05, "loss": 0.6287, "step": 296310 }, { "epoch": 25.66, "learning_rate": 1.0854920781774025e-05, "loss": 0.6096, "step": 296320 }, { "epoch": 25.66, "learning_rate": 1.0852752399676187e-05, "loss": 0.4857, "step": 296330 }, { "epoch": 25.66, "learning_rate": 1.0850584017578351e-05, "loss": 0.766, "step": 296340 }, { "epoch": 25.67, "learning_rate": 1.0848415635480513e-05, "loss": 0.5114, "step": 296350 }, { "epoch": 25.67, "learning_rate": 1.0846247253382674e-05, "loss": 0.6224, "step": 296360 }, { "epoch": 25.67, "learning_rate": 1.0844078871284837e-05, "loss": 0.7265, "step": 296370 }, { "epoch": 25.67, "learning_rate": 1.0841910489187e-05, "loss": 0.6377, "step": 296380 }, { "epoch": 25.67, "learning_rate": 1.0839742107089163e-05, "loss": 0.6845, "step": 296390 }, { "epoch": 25.67, "learning_rate": 1.0837573724991324e-05, "loss": 0.5707, "step": 296400 }, { "epoch": 25.67, "learning_rate": 1.0835405342893489e-05, "loss": 0.6812, "step": 296410 }, { "epoch": 25.67, "learning_rate": 1.083323696079565e-05, "loss": 0.6176, "step": 296420 }, { "epoch": 25.67, "learning_rate": 1.0831068578697813e-05, "loss": 0.629, "step": 296430 }, { "epoch": 25.67, "learning_rate": 1.0828900196599975e-05, "loss": 0.557, "step": 296440 }, { "epoch": 25.67, "learning_rate": 1.082673181450214e-05, "loss": 0.6052, "step": 296450 }, { "epoch": 25.68, "learning_rate": 1.08245634324043e-05, "loss": 0.6236, "step": 296460 }, { "epoch": 25.68, "learning_rate": 1.0822395050306464e-05, "loss": 0.6426, "step": 296470 }, { "epoch": 25.68, "learning_rate": 1.0820226668208625e-05, "loss": 0.7601, "step": 296480 }, { "epoch": 25.68, "learning_rate": 1.081805828611079e-05, "loss": 0.6491, "step": 296490 }, { "epoch": 25.68, "learning_rate": 1.0815889904012951e-05, "loss": 0.6115, "step": 296500 }, { "epoch": 25.68, "learning_rate": 1.0813721521915114e-05, "loss": 0.5448, "step": 296510 }, { "epoch": 25.68, "learning_rate": 1.0811553139817277e-05, "loss": 0.6014, "step": 296520 }, { "epoch": 25.68, "learning_rate": 1.080938475771944e-05, "loss": 0.6373, "step": 296530 }, { "epoch": 25.68, "learning_rate": 1.0807216375621601e-05, "loss": 0.6153, "step": 296540 }, { "epoch": 25.68, "learning_rate": 1.0805047993523766e-05, "loss": 0.6213, "step": 296550 }, { "epoch": 25.68, "learning_rate": 1.0802879611425927e-05, "loss": 0.6318, "step": 296560 }, { "epoch": 25.68, "learning_rate": 1.080071122932809e-05, "loss": 0.7277, "step": 296570 }, { "epoch": 25.69, "learning_rate": 1.0798542847230252e-05, "loss": 0.6511, "step": 296580 }, { "epoch": 25.69, "learning_rate": 1.0796374465132416e-05, "loss": 0.597, "step": 296590 }, { "epoch": 25.69, "learning_rate": 1.0794206083034578e-05, "loss": 0.7356, "step": 296600 }, { "epoch": 25.69, "learning_rate": 1.0792037700936739e-05, "loss": 0.4971, "step": 296610 }, { "epoch": 25.69, "learning_rate": 1.0789869318838902e-05, "loss": 0.7045, "step": 296620 }, { "epoch": 25.69, "learning_rate": 1.0787700936741065e-05, "loss": 0.4498, "step": 296630 }, { "epoch": 25.69, "learning_rate": 1.0785532554643228e-05, "loss": 0.5947, "step": 296640 }, { "epoch": 25.69, "learning_rate": 1.078336417254539e-05, "loss": 0.5745, "step": 296650 }, { "epoch": 25.69, "learning_rate": 1.0781195790447554e-05, "loss": 0.5415, "step": 296660 }, { "epoch": 25.69, "learning_rate": 1.0779027408349715e-05, "loss": 0.479, "step": 296670 }, { "epoch": 25.69, "learning_rate": 1.0776859026251878e-05, "loss": 0.7563, "step": 296680 }, { "epoch": 25.7, "learning_rate": 1.077469064415404e-05, "loss": 0.5749, "step": 296690 }, { "epoch": 25.7, "learning_rate": 1.0772522262056204e-05, "loss": 0.6261, "step": 296700 }, { "epoch": 25.7, "learning_rate": 1.0770353879958366e-05, "loss": 0.5654, "step": 296710 }, { "epoch": 25.7, "learning_rate": 1.0768185497860529e-05, "loss": 0.7038, "step": 296720 }, { "epoch": 25.7, "learning_rate": 1.0766017115762692e-05, "loss": 0.7753, "step": 296730 }, { "epoch": 25.7, "learning_rate": 1.0763848733664855e-05, "loss": 0.5908, "step": 296740 }, { "epoch": 25.7, "learning_rate": 1.0761680351567016e-05, "loss": 0.5156, "step": 296750 }, { "epoch": 25.7, "learning_rate": 1.0759511969469179e-05, "loss": 0.6102, "step": 296760 }, { "epoch": 25.7, "learning_rate": 1.0757343587371342e-05, "loss": 0.5666, "step": 296770 }, { "epoch": 25.7, "learning_rate": 1.0755175205273505e-05, "loss": 0.7342, "step": 296780 }, { "epoch": 25.7, "learning_rate": 1.0753006823175666e-05, "loss": 0.6063, "step": 296790 }, { "epoch": 25.7, "learning_rate": 1.0750838441077831e-05, "loss": 0.6531, "step": 296800 }, { "epoch": 25.71, "learning_rate": 1.0748670058979992e-05, "loss": 0.607, "step": 296810 }, { "epoch": 25.71, "learning_rate": 1.0746501676882155e-05, "loss": 0.4457, "step": 296820 }, { "epoch": 25.71, "learning_rate": 1.0744333294784317e-05, "loss": 0.5159, "step": 296830 }, { "epoch": 25.71, "learning_rate": 1.0742164912686481e-05, "loss": 0.6977, "step": 296840 }, { "epoch": 25.71, "learning_rate": 1.0739996530588643e-05, "loss": 0.608, "step": 296850 }, { "epoch": 25.71, "learning_rate": 1.0737828148490804e-05, "loss": 0.6112, "step": 296860 }, { "epoch": 25.71, "learning_rate": 1.0735659766392969e-05, "loss": 0.6289, "step": 296870 }, { "epoch": 25.71, "learning_rate": 1.073349138429513e-05, "loss": 0.5125, "step": 296880 }, { "epoch": 25.71, "learning_rate": 1.0731323002197293e-05, "loss": 0.5088, "step": 296890 }, { "epoch": 25.71, "learning_rate": 1.0729154620099454e-05, "loss": 0.6318, "step": 296900 }, { "epoch": 25.71, "learning_rate": 1.0726986238001619e-05, "loss": 0.6557, "step": 296910 }, { "epoch": 25.72, "learning_rate": 1.072481785590378e-05, "loss": 0.6578, "step": 296920 }, { "epoch": 25.72, "learning_rate": 1.0722649473805943e-05, "loss": 0.6174, "step": 296930 }, { "epoch": 25.72, "learning_rate": 1.0720481091708105e-05, "loss": 0.6297, "step": 296940 }, { "epoch": 25.72, "learning_rate": 1.071831270961027e-05, "loss": 0.6161, "step": 296950 }, { "epoch": 25.72, "learning_rate": 1.071614432751243e-05, "loss": 0.5797, "step": 296960 }, { "epoch": 25.72, "learning_rate": 1.0713975945414594e-05, "loss": 0.5726, "step": 296970 }, { "epoch": 25.72, "learning_rate": 1.0711807563316757e-05, "loss": 0.717, "step": 296980 }, { "epoch": 25.72, "learning_rate": 1.070963918121892e-05, "loss": 0.688, "step": 296990 }, { "epoch": 25.72, "learning_rate": 1.0707470799121081e-05, "loss": 0.6789, "step": 297000 }, { "epoch": 25.72, "learning_rate": 1.0705302417023246e-05, "loss": 0.6728, "step": 297010 }, { "epoch": 25.72, "learning_rate": 1.0703134034925407e-05, "loss": 0.7249, "step": 297020 }, { "epoch": 25.72, "learning_rate": 1.070096565282757e-05, "loss": 0.5545, "step": 297030 }, { "epoch": 25.73, "learning_rate": 1.0698797270729731e-05, "loss": 0.6629, "step": 297040 }, { "epoch": 25.73, "learning_rate": 1.0696628888631896e-05, "loss": 0.6312, "step": 297050 }, { "epoch": 25.73, "learning_rate": 1.0694460506534057e-05, "loss": 0.5098, "step": 297060 }, { "epoch": 25.73, "learning_rate": 1.069229212443622e-05, "loss": 0.5713, "step": 297070 }, { "epoch": 25.73, "learning_rate": 1.0690123742338381e-05, "loss": 0.7387, "step": 297080 }, { "epoch": 25.73, "learning_rate": 1.0687955360240546e-05, "loss": 0.5761, "step": 297090 }, { "epoch": 25.73, "learning_rate": 1.0685786978142707e-05, "loss": 0.6559, "step": 297100 }, { "epoch": 25.73, "learning_rate": 1.068361859604487e-05, "loss": 0.5337, "step": 297110 }, { "epoch": 25.73, "learning_rate": 1.0681450213947034e-05, "loss": 0.458, "step": 297120 }, { "epoch": 25.73, "learning_rate": 1.0679281831849195e-05, "loss": 0.6332, "step": 297130 }, { "epoch": 25.73, "learning_rate": 1.0677113449751358e-05, "loss": 0.6975, "step": 297140 }, { "epoch": 25.74, "learning_rate": 1.0674945067653519e-05, "loss": 0.5915, "step": 297150 }, { "epoch": 25.74, "learning_rate": 1.0672776685555684e-05, "loss": 0.5492, "step": 297160 }, { "epoch": 25.74, "learning_rate": 1.0670608303457845e-05, "loss": 0.6267, "step": 297170 }, { "epoch": 25.74, "learning_rate": 1.0668439921360008e-05, "loss": 0.6123, "step": 297180 }, { "epoch": 25.74, "learning_rate": 1.066627153926217e-05, "loss": 0.6639, "step": 297190 }, { "epoch": 25.74, "learning_rate": 1.0664103157164334e-05, "loss": 0.5959, "step": 297200 }, { "epoch": 25.74, "learning_rate": 1.0661934775066495e-05, "loss": 0.5975, "step": 297210 }, { "epoch": 25.74, "learning_rate": 1.0659766392968658e-05, "loss": 0.661, "step": 297220 }, { "epoch": 25.74, "learning_rate": 1.0657598010870821e-05, "loss": 0.5532, "step": 297230 }, { "epoch": 25.74, "learning_rate": 1.0655429628772984e-05, "loss": 0.6051, "step": 297240 }, { "epoch": 25.74, "learning_rate": 1.0653261246675146e-05, "loss": 0.6637, "step": 297250 }, { "epoch": 25.74, "learning_rate": 1.065109286457731e-05, "loss": 0.6085, "step": 297260 }, { "epoch": 25.75, "learning_rate": 1.0648924482479472e-05, "loss": 0.5087, "step": 297270 }, { "epoch": 25.75, "learning_rate": 1.0646756100381635e-05, "loss": 0.599, "step": 297280 }, { "epoch": 25.75, "learning_rate": 1.0644587718283796e-05, "loss": 0.5638, "step": 297290 }, { "epoch": 25.75, "learning_rate": 1.064241933618596e-05, "loss": 0.5603, "step": 297300 }, { "epoch": 25.75, "learning_rate": 1.0640250954088122e-05, "loss": 0.503, "step": 297310 }, { "epoch": 25.75, "learning_rate": 1.0638082571990285e-05, "loss": 0.5997, "step": 297320 }, { "epoch": 25.75, "learning_rate": 1.0635914189892446e-05, "loss": 0.6409, "step": 297330 }, { "epoch": 25.75, "learning_rate": 1.0633745807794611e-05, "loss": 0.6659, "step": 297340 }, { "epoch": 25.75, "learning_rate": 1.0631577425696772e-05, "loss": 0.7108, "step": 297350 }, { "epoch": 25.75, "learning_rate": 1.0629409043598935e-05, "loss": 0.6582, "step": 297360 }, { "epoch": 25.75, "learning_rate": 1.0627240661501098e-05, "loss": 0.6851, "step": 297370 }, { "epoch": 25.75, "learning_rate": 1.062507227940326e-05, "loss": 0.6847, "step": 297380 }, { "epoch": 25.76, "learning_rate": 1.0622903897305423e-05, "loss": 0.7006, "step": 297390 }, { "epoch": 25.76, "learning_rate": 1.0620735515207584e-05, "loss": 0.7504, "step": 297400 }, { "epoch": 25.76, "learning_rate": 1.0618567133109749e-05, "loss": 0.6451, "step": 297410 }, { "epoch": 25.76, "learning_rate": 1.061639875101191e-05, "loss": 0.6888, "step": 297420 }, { "epoch": 25.76, "learning_rate": 1.0614230368914073e-05, "loss": 0.6069, "step": 297430 }, { "epoch": 25.76, "learning_rate": 1.0612061986816234e-05, "loss": 0.5649, "step": 297440 }, { "epoch": 25.76, "learning_rate": 1.0609893604718399e-05, "loss": 0.6159, "step": 297450 }, { "epoch": 25.76, "learning_rate": 1.060772522262056e-05, "loss": 0.6639, "step": 297460 }, { "epoch": 25.76, "learning_rate": 1.0605556840522723e-05, "loss": 0.6084, "step": 297470 }, { "epoch": 25.76, "learning_rate": 1.0603388458424886e-05, "loss": 0.7124, "step": 297480 }, { "epoch": 25.76, "learning_rate": 1.060122007632705e-05, "loss": 0.6733, "step": 297490 }, { "epoch": 25.77, "learning_rate": 1.059905169422921e-05, "loss": 0.6709, "step": 297500 }, { "epoch": 25.77, "learning_rate": 1.0596883312131375e-05, "loss": 0.6052, "step": 297510 }, { "epoch": 25.77, "learning_rate": 1.0594714930033537e-05, "loss": 0.5878, "step": 297520 }, { "epoch": 25.77, "learning_rate": 1.05925465479357e-05, "loss": 0.6801, "step": 297530 }, { "epoch": 25.77, "learning_rate": 1.0590378165837861e-05, "loss": 0.7106, "step": 297540 }, { "epoch": 25.77, "learning_rate": 1.0588209783740026e-05, "loss": 0.5415, "step": 297550 }, { "epoch": 25.77, "learning_rate": 1.0586041401642187e-05, "loss": 0.7257, "step": 297560 }, { "epoch": 25.77, "learning_rate": 1.058387301954435e-05, "loss": 0.7114, "step": 297570 }, { "epoch": 25.77, "learning_rate": 1.0581704637446511e-05, "loss": 0.5416, "step": 297580 }, { "epoch": 25.77, "learning_rate": 1.0579536255348676e-05, "loss": 0.5535, "step": 297590 }, { "epoch": 25.77, "learning_rate": 1.0577367873250837e-05, "loss": 0.5412, "step": 297600 }, { "epoch": 25.77, "learning_rate": 1.0575199491153e-05, "loss": 0.6236, "step": 297610 }, { "epoch": 25.78, "learning_rate": 1.0573031109055163e-05, "loss": 0.6065, "step": 297620 }, { "epoch": 25.78, "learning_rate": 1.0570862726957325e-05, "loss": 0.6029, "step": 297630 }, { "epoch": 25.78, "learning_rate": 1.0568694344859488e-05, "loss": 0.7676, "step": 297640 }, { "epoch": 25.78, "learning_rate": 1.0566525962761649e-05, "loss": 0.6342, "step": 297650 }, { "epoch": 25.78, "learning_rate": 1.0564357580663814e-05, "loss": 0.5847, "step": 297660 }, { "epoch": 25.78, "learning_rate": 1.0562189198565975e-05, "loss": 0.6448, "step": 297670 }, { "epoch": 25.78, "learning_rate": 1.0560020816468138e-05, "loss": 0.5115, "step": 297680 }, { "epoch": 25.78, "learning_rate": 1.05578524343703e-05, "loss": 0.6456, "step": 297690 }, { "epoch": 25.78, "learning_rate": 1.0555684052272464e-05, "loss": 0.5901, "step": 297700 }, { "epoch": 25.78, "learning_rate": 1.0553515670174625e-05, "loss": 0.6284, "step": 297710 }, { "epoch": 25.78, "learning_rate": 1.0551347288076788e-05, "loss": 0.6967, "step": 297720 }, { "epoch": 25.79, "learning_rate": 1.0549178905978951e-05, "loss": 0.6602, "step": 297730 }, { "epoch": 25.79, "learning_rate": 1.0547010523881114e-05, "loss": 0.7358, "step": 297740 }, { "epoch": 25.79, "learning_rate": 1.0544842141783276e-05, "loss": 0.52, "step": 297750 }, { "epoch": 25.79, "learning_rate": 1.054267375968544e-05, "loss": 0.6821, "step": 297760 }, { "epoch": 25.79, "learning_rate": 1.0540505377587602e-05, "loss": 0.7553, "step": 297770 }, { "epoch": 25.79, "learning_rate": 1.0538336995489765e-05, "loss": 0.5884, "step": 297780 }, { "epoch": 25.79, "learning_rate": 1.0536168613391926e-05, "loss": 0.5632, "step": 297790 }, { "epoch": 25.79, "learning_rate": 1.053400023129409e-05, "loss": 0.5623, "step": 297800 }, { "epoch": 25.79, "learning_rate": 1.0531831849196252e-05, "loss": 0.7077, "step": 297810 }, { "epoch": 25.79, "learning_rate": 1.0529663467098415e-05, "loss": 0.5542, "step": 297820 }, { "epoch": 25.79, "learning_rate": 1.0527495085000576e-05, "loss": 0.607, "step": 297830 }, { "epoch": 25.79, "learning_rate": 1.0525326702902741e-05, "loss": 0.5967, "step": 297840 }, { "epoch": 25.8, "learning_rate": 1.0523158320804902e-05, "loss": 0.5975, "step": 297850 }, { "epoch": 25.8, "learning_rate": 1.0520989938707065e-05, "loss": 0.4101, "step": 297860 }, { "epoch": 25.8, "learning_rate": 1.0518821556609228e-05, "loss": 0.6137, "step": 297870 }, { "epoch": 25.8, "learning_rate": 1.0516653174511391e-05, "loss": 0.5221, "step": 297880 }, { "epoch": 25.8, "learning_rate": 1.0514484792413553e-05, "loss": 0.5861, "step": 297890 }, { "epoch": 25.8, "learning_rate": 1.0512316410315714e-05, "loss": 0.653, "step": 297900 }, { "epoch": 25.8, "learning_rate": 1.0510148028217879e-05, "loss": 0.6036, "step": 297910 }, { "epoch": 25.8, "learning_rate": 1.050797964612004e-05, "loss": 0.4506, "step": 297920 }, { "epoch": 25.8, "learning_rate": 1.0505811264022203e-05, "loss": 0.6906, "step": 297930 }, { "epoch": 25.8, "learning_rate": 1.0503642881924366e-05, "loss": 0.4776, "step": 297940 }, { "epoch": 25.8, "learning_rate": 1.0501474499826529e-05, "loss": 0.8004, "step": 297950 }, { "epoch": 25.81, "learning_rate": 1.049930611772869e-05, "loss": 0.5458, "step": 297960 }, { "epoch": 25.81, "learning_rate": 1.0497137735630853e-05, "loss": 0.631, "step": 297970 }, { "epoch": 25.81, "learning_rate": 1.0494969353533016e-05, "loss": 0.4952, "step": 297980 }, { "epoch": 25.81, "learning_rate": 1.049280097143518e-05, "loss": 0.5921, "step": 297990 }, { "epoch": 25.81, "learning_rate": 1.049063258933734e-05, "loss": 0.6361, "step": 298000 }, { "epoch": 25.81, "learning_rate": 1.0488464207239505e-05, "loss": 0.6619, "step": 298010 }, { "epoch": 25.81, "learning_rate": 1.0486295825141667e-05, "loss": 0.7086, "step": 298020 }, { "epoch": 25.81, "learning_rate": 1.048412744304383e-05, "loss": 0.6881, "step": 298030 }, { "epoch": 25.81, "learning_rate": 1.0481959060945991e-05, "loss": 0.6236, "step": 298040 }, { "epoch": 25.81, "learning_rate": 1.0479790678848156e-05, "loss": 0.5642, "step": 298050 }, { "epoch": 25.81, "learning_rate": 1.0477622296750317e-05, "loss": 0.5666, "step": 298060 }, { "epoch": 25.81, "learning_rate": 1.047545391465248e-05, "loss": 0.6251, "step": 298070 }, { "epoch": 25.82, "learning_rate": 1.0473285532554643e-05, "loss": 0.6935, "step": 298080 }, { "epoch": 25.82, "learning_rate": 1.0471117150456806e-05, "loss": 0.4971, "step": 298090 }, { "epoch": 25.82, "learning_rate": 1.0468948768358967e-05, "loss": 0.6496, "step": 298100 }, { "epoch": 25.82, "learning_rate": 1.046678038626113e-05, "loss": 0.5433, "step": 298110 }, { "epoch": 25.82, "learning_rate": 1.0464612004163293e-05, "loss": 0.6199, "step": 298120 }, { "epoch": 25.82, "learning_rate": 1.0462443622065456e-05, "loss": 0.5049, "step": 298130 }, { "epoch": 25.82, "learning_rate": 1.0460275239967618e-05, "loss": 0.7192, "step": 298140 }, { "epoch": 25.82, "learning_rate": 1.0458106857869779e-05, "loss": 0.6391, "step": 298150 }, { "epoch": 25.82, "learning_rate": 1.0455938475771944e-05, "loss": 0.4521, "step": 298160 }, { "epoch": 25.82, "learning_rate": 1.0453770093674105e-05, "loss": 0.6842, "step": 298170 }, { "epoch": 25.82, "learning_rate": 1.0451601711576268e-05, "loss": 0.5658, "step": 298180 }, { "epoch": 25.83, "learning_rate": 1.044943332947843e-05, "loss": 0.6861, "step": 298190 }, { "epoch": 25.83, "learning_rate": 1.0447264947380594e-05, "loss": 0.8089, "step": 298200 }, { "epoch": 25.83, "learning_rate": 1.0445096565282755e-05, "loss": 0.6978, "step": 298210 }, { "epoch": 25.83, "learning_rate": 1.044292818318492e-05, "loss": 0.6834, "step": 298220 }, { "epoch": 25.83, "learning_rate": 1.0440759801087081e-05, "loss": 0.5182, "step": 298230 }, { "epoch": 25.83, "learning_rate": 1.0438591418989244e-05, "loss": 0.5175, "step": 298240 }, { "epoch": 25.83, "learning_rate": 1.0436423036891405e-05, "loss": 0.6042, "step": 298250 }, { "epoch": 25.83, "learning_rate": 1.043425465479357e-05, "loss": 0.6746, "step": 298260 }, { "epoch": 25.83, "learning_rate": 1.0432086272695731e-05, "loss": 0.516, "step": 298270 }, { "epoch": 25.83, "learning_rate": 1.0429917890597894e-05, "loss": 0.6592, "step": 298280 }, { "epoch": 25.83, "learning_rate": 1.0427749508500056e-05, "loss": 0.6952, "step": 298290 }, { "epoch": 25.83, "learning_rate": 1.042558112640222e-05, "loss": 0.6473, "step": 298300 }, { "epoch": 25.84, "learning_rate": 1.0423412744304382e-05, "loss": 0.5622, "step": 298310 }, { "epoch": 25.84, "learning_rate": 1.0421244362206545e-05, "loss": 0.5385, "step": 298320 }, { "epoch": 25.84, "learning_rate": 1.0419075980108708e-05, "loss": 0.5024, "step": 298330 }, { "epoch": 25.84, "learning_rate": 1.041690759801087e-05, "loss": 0.5513, "step": 298340 }, { "epoch": 25.84, "learning_rate": 1.0414739215913032e-05, "loss": 0.5951, "step": 298350 }, { "epoch": 25.84, "learning_rate": 1.0412570833815197e-05, "loss": 0.5898, "step": 298360 }, { "epoch": 25.84, "learning_rate": 1.0410402451717358e-05, "loss": 0.6302, "step": 298370 }, { "epoch": 25.84, "learning_rate": 1.0408234069619521e-05, "loss": 0.6029, "step": 298380 }, { "epoch": 25.84, "learning_rate": 1.0406065687521682e-05, "loss": 0.7101, "step": 298390 }, { "epoch": 25.84, "learning_rate": 1.0403897305423844e-05, "loss": 0.6394, "step": 298400 }, { "epoch": 25.84, "learning_rate": 1.0401728923326008e-05, "loss": 0.7467, "step": 298410 }, { "epoch": 25.85, "learning_rate": 1.039956054122817e-05, "loss": 0.7905, "step": 298420 }, { "epoch": 25.85, "learning_rate": 1.0397392159130333e-05, "loss": 0.7716, "step": 298430 }, { "epoch": 25.85, "learning_rate": 1.0395223777032496e-05, "loss": 0.559, "step": 298440 }, { "epoch": 25.85, "learning_rate": 1.0393055394934659e-05, "loss": 0.5565, "step": 298450 }, { "epoch": 25.85, "learning_rate": 1.039088701283682e-05, "loss": 0.6157, "step": 298460 }, { "epoch": 25.85, "learning_rate": 1.0388718630738985e-05, "loss": 0.6857, "step": 298470 }, { "epoch": 25.85, "learning_rate": 1.0386550248641146e-05, "loss": 0.8483, "step": 298480 }, { "epoch": 25.85, "learning_rate": 1.0384381866543309e-05, "loss": 0.6969, "step": 298490 }, { "epoch": 25.85, "learning_rate": 1.038221348444547e-05, "loss": 0.5806, "step": 298500 }, { "epoch": 25.85, "learning_rate": 1.0380045102347635e-05, "loss": 0.5587, "step": 298510 }, { "epoch": 25.85, "learning_rate": 1.0377876720249796e-05, "loss": 0.5487, "step": 298520 }, { "epoch": 25.85, "learning_rate": 1.037570833815196e-05, "loss": 0.6003, "step": 298530 }, { "epoch": 25.86, "learning_rate": 1.037353995605412e-05, "loss": 0.6138, "step": 298540 }, { "epoch": 25.86, "learning_rate": 1.0371371573956285e-05, "loss": 0.7325, "step": 298550 }, { "epoch": 25.86, "learning_rate": 1.0369203191858447e-05, "loss": 0.633, "step": 298560 }, { "epoch": 25.86, "learning_rate": 1.036703480976061e-05, "loss": 0.5683, "step": 298570 }, { "epoch": 25.86, "learning_rate": 1.0364866427662773e-05, "loss": 0.6145, "step": 298580 }, { "epoch": 25.86, "learning_rate": 1.0362698045564936e-05, "loss": 0.6188, "step": 298590 }, { "epoch": 25.86, "learning_rate": 1.0360529663467097e-05, "loss": 0.6109, "step": 298600 }, { "epoch": 25.86, "learning_rate": 1.0358361281369262e-05, "loss": 0.6523, "step": 298610 }, { "epoch": 25.86, "learning_rate": 1.0356192899271423e-05, "loss": 0.697, "step": 298620 }, { "epoch": 25.86, "learning_rate": 1.0354024517173586e-05, "loss": 0.6872, "step": 298630 }, { "epoch": 25.86, "learning_rate": 1.0351856135075747e-05, "loss": 0.5169, "step": 298640 }, { "epoch": 25.86, "learning_rate": 1.0349687752977912e-05, "loss": 0.5843, "step": 298650 }, { "epoch": 25.87, "learning_rate": 1.0347519370880073e-05, "loss": 0.5546, "step": 298660 }, { "epoch": 25.87, "learning_rate": 1.0345350988782235e-05, "loss": 0.4914, "step": 298670 }, { "epoch": 25.87, "learning_rate": 1.0343182606684398e-05, "loss": 0.6537, "step": 298680 }, { "epoch": 25.87, "learning_rate": 1.034101422458656e-05, "loss": 0.6909, "step": 298690 }, { "epoch": 25.87, "learning_rate": 1.0338845842488724e-05, "loss": 0.5182, "step": 298700 }, { "epoch": 25.87, "learning_rate": 1.0336677460390885e-05, "loss": 0.7141, "step": 298710 }, { "epoch": 25.87, "learning_rate": 1.033450907829305e-05, "loss": 0.6568, "step": 298720 }, { "epoch": 25.87, "learning_rate": 1.0332340696195211e-05, "loss": 0.7789, "step": 298730 }, { "epoch": 25.87, "learning_rate": 1.0330172314097374e-05, "loss": 0.5413, "step": 298740 }, { "epoch": 25.87, "learning_rate": 1.0328003931999535e-05, "loss": 0.591, "step": 298750 }, { "epoch": 25.87, "learning_rate": 1.03258355499017e-05, "loss": 0.6086, "step": 298760 }, { "epoch": 25.88, "learning_rate": 1.0323667167803861e-05, "loss": 0.6749, "step": 298770 }, { "epoch": 25.88, "learning_rate": 1.0321498785706024e-05, "loss": 0.6761, "step": 298780 }, { "epoch": 25.88, "learning_rate": 1.0319330403608186e-05, "loss": 0.6219, "step": 298790 }, { "epoch": 25.88, "learning_rate": 1.031716202151035e-05, "loss": 0.6463, "step": 298800 }, { "epoch": 25.88, "learning_rate": 1.0314993639412512e-05, "loss": 0.7199, "step": 298810 }, { "epoch": 25.88, "learning_rate": 1.0312825257314675e-05, "loss": 0.6141, "step": 298820 }, { "epoch": 25.88, "learning_rate": 1.0310656875216838e-05, "loss": 0.5853, "step": 298830 }, { "epoch": 25.88, "learning_rate": 1.0308488493119e-05, "loss": 0.5572, "step": 298840 }, { "epoch": 25.88, "learning_rate": 1.0306320111021162e-05, "loss": 0.6232, "step": 298850 }, { "epoch": 25.88, "learning_rate": 1.0304151728923327e-05, "loss": 0.6486, "step": 298860 }, { "epoch": 25.88, "learning_rate": 1.0301983346825488e-05, "loss": 0.5702, "step": 298870 }, { "epoch": 25.88, "learning_rate": 1.0299814964727651e-05, "loss": 0.4464, "step": 298880 }, { "epoch": 25.89, "learning_rate": 1.0297646582629812e-05, "loss": 0.4589, "step": 298890 }, { "epoch": 25.89, "learning_rate": 1.0295478200531977e-05, "loss": 0.6024, "step": 298900 }, { "epoch": 25.89, "learning_rate": 1.0293309818434138e-05, "loss": 0.4951, "step": 298910 }, { "epoch": 25.89, "learning_rate": 1.02911414363363e-05, "loss": 0.5632, "step": 298920 }, { "epoch": 25.89, "learning_rate": 1.0288973054238463e-05, "loss": 0.6943, "step": 298930 }, { "epoch": 25.89, "learning_rate": 1.0286804672140626e-05, "loss": 0.6828, "step": 298940 }, { "epoch": 25.89, "learning_rate": 1.0284636290042789e-05, "loss": 0.612, "step": 298950 }, { "epoch": 25.89, "learning_rate": 1.028246790794495e-05, "loss": 0.6148, "step": 298960 }, { "epoch": 25.89, "learning_rate": 1.0280299525847115e-05, "loss": 0.5239, "step": 298970 }, { "epoch": 25.89, "learning_rate": 1.0278131143749276e-05, "loss": 0.7922, "step": 298980 }, { "epoch": 25.89, "learning_rate": 1.0275962761651439e-05, "loss": 0.5385, "step": 298990 }, { "epoch": 25.9, "learning_rate": 1.02737943795536e-05, "loss": 0.6504, "step": 299000 }, { "epoch": 25.9, "learning_rate": 1.0271625997455765e-05, "loss": 0.6248, "step": 299010 }, { "epoch": 25.9, "learning_rate": 1.0269457615357926e-05, "loss": 0.7149, "step": 299020 }, { "epoch": 25.9, "learning_rate": 1.026728923326009e-05, "loss": 0.4738, "step": 299030 }, { "epoch": 25.9, "learning_rate": 1.026512085116225e-05, "loss": 0.5674, "step": 299040 }, { "epoch": 25.9, "learning_rate": 1.0262952469064415e-05, "loss": 0.6147, "step": 299050 }, { "epoch": 25.9, "learning_rate": 1.0260784086966577e-05, "loss": 0.7013, "step": 299060 }, { "epoch": 25.9, "learning_rate": 1.025861570486874e-05, "loss": 0.6713, "step": 299070 }, { "epoch": 25.9, "learning_rate": 1.0256447322770903e-05, "loss": 0.6048, "step": 299080 }, { "epoch": 25.9, "learning_rate": 1.0254278940673066e-05, "loss": 0.692, "step": 299090 }, { "epoch": 25.9, "learning_rate": 1.0252110558575227e-05, "loss": 0.716, "step": 299100 }, { "epoch": 25.9, "learning_rate": 1.0249942176477392e-05, "loss": 0.481, "step": 299110 }, { "epoch": 25.91, "learning_rate": 1.0247773794379553e-05, "loss": 0.6048, "step": 299120 }, { "epoch": 25.91, "learning_rate": 1.0245605412281716e-05, "loss": 0.6586, "step": 299130 }, { "epoch": 25.91, "learning_rate": 1.0243437030183877e-05, "loss": 0.6032, "step": 299140 }, { "epoch": 25.91, "learning_rate": 1.0241268648086042e-05, "loss": 0.8055, "step": 299150 }, { "epoch": 25.91, "learning_rate": 1.0239100265988203e-05, "loss": 0.5492, "step": 299160 }, { "epoch": 25.91, "learning_rate": 1.0236931883890365e-05, "loss": 0.4705, "step": 299170 }, { "epoch": 25.91, "learning_rate": 1.0234763501792528e-05, "loss": 0.5552, "step": 299180 }, { "epoch": 25.91, "learning_rate": 1.023259511969469e-05, "loss": 0.6274, "step": 299190 }, { "epoch": 25.91, "learning_rate": 1.0230426737596854e-05, "loss": 0.6793, "step": 299200 }, { "epoch": 25.91, "learning_rate": 1.0228258355499015e-05, "loss": 0.6523, "step": 299210 }, { "epoch": 25.91, "learning_rate": 1.022608997340118e-05, "loss": 0.775, "step": 299220 }, { "epoch": 25.92, "learning_rate": 1.0223921591303341e-05, "loss": 0.4711, "step": 299230 }, { "epoch": 25.92, "learning_rate": 1.0221753209205504e-05, "loss": 0.6933, "step": 299240 }, { "epoch": 25.92, "learning_rate": 1.0219584827107665e-05, "loss": 0.6872, "step": 299250 }, { "epoch": 25.92, "learning_rate": 1.021741644500983e-05, "loss": 0.7351, "step": 299260 }, { "epoch": 25.92, "learning_rate": 1.0215248062911991e-05, "loss": 0.6273, "step": 299270 }, { "epoch": 25.92, "learning_rate": 1.0213079680814154e-05, "loss": 0.5603, "step": 299280 }, { "epoch": 25.92, "learning_rate": 1.0210911298716317e-05, "loss": 0.6864, "step": 299290 }, { "epoch": 25.92, "learning_rate": 1.020874291661848e-05, "loss": 0.7004, "step": 299300 }, { "epoch": 25.92, "learning_rate": 1.0206574534520641e-05, "loss": 0.6833, "step": 299310 }, { "epoch": 25.92, "learning_rate": 1.0204406152422804e-05, "loss": 0.5904, "step": 299320 }, { "epoch": 25.92, "learning_rate": 1.0202237770324967e-05, "loss": 0.5837, "step": 299330 }, { "epoch": 25.92, "learning_rate": 1.020006938822713e-05, "loss": 0.5568, "step": 299340 }, { "epoch": 25.93, "learning_rate": 1.0197901006129292e-05, "loss": 0.7491, "step": 299350 }, { "epoch": 25.93, "learning_rate": 1.0195732624031457e-05, "loss": 0.6293, "step": 299360 }, { "epoch": 25.93, "learning_rate": 1.0193564241933618e-05, "loss": 0.5746, "step": 299370 }, { "epoch": 25.93, "learning_rate": 1.019139585983578e-05, "loss": 0.7335, "step": 299380 }, { "epoch": 25.93, "learning_rate": 1.0189227477737942e-05, "loss": 0.7583, "step": 299390 }, { "epoch": 25.93, "learning_rate": 1.0187059095640107e-05, "loss": 0.4252, "step": 299400 }, { "epoch": 25.93, "learning_rate": 1.0184890713542268e-05, "loss": 0.663, "step": 299410 }, { "epoch": 25.93, "learning_rate": 1.0182722331444431e-05, "loss": 0.6581, "step": 299420 }, { "epoch": 25.93, "learning_rate": 1.0180553949346594e-05, "loss": 0.6273, "step": 299430 }, { "epoch": 25.93, "learning_rate": 1.0178385567248755e-05, "loss": 0.5451, "step": 299440 }, { "epoch": 25.93, "learning_rate": 1.0176217185150918e-05, "loss": 0.6011, "step": 299450 }, { "epoch": 25.94, "learning_rate": 1.017404880305308e-05, "loss": 0.611, "step": 299460 }, { "epoch": 25.94, "learning_rate": 1.0171880420955244e-05, "loss": 0.6922, "step": 299470 }, { "epoch": 25.94, "learning_rate": 1.0169712038857406e-05, "loss": 0.6229, "step": 299480 }, { "epoch": 25.94, "learning_rate": 1.0167543656759569e-05, "loss": 0.5815, "step": 299490 }, { "epoch": 25.94, "learning_rate": 1.016537527466173e-05, "loss": 0.5594, "step": 299500 }, { "epoch": 25.94, "learning_rate": 1.0163206892563895e-05, "loss": 0.6658, "step": 299510 }, { "epoch": 25.94, "learning_rate": 1.0161038510466056e-05, "loss": 0.7159, "step": 299520 }, { "epoch": 25.94, "learning_rate": 1.0158870128368219e-05, "loss": 0.5951, "step": 299530 }, { "epoch": 25.94, "learning_rate": 1.0156701746270382e-05, "loss": 0.7346, "step": 299540 }, { "epoch": 25.94, "learning_rate": 1.0154533364172545e-05, "loss": 0.5601, "step": 299550 }, { "epoch": 25.94, "learning_rate": 1.0152364982074706e-05, "loss": 0.78, "step": 299560 }, { "epoch": 25.94, "learning_rate": 1.0150196599976871e-05, "loss": 0.5567, "step": 299570 }, { "epoch": 25.95, "learning_rate": 1.0148028217879032e-05, "loss": 0.6329, "step": 299580 }, { "epoch": 25.95, "learning_rate": 1.0145859835781195e-05, "loss": 0.6604, "step": 299590 }, { "epoch": 25.95, "learning_rate": 1.0143691453683357e-05, "loss": 0.6525, "step": 299600 }, { "epoch": 25.95, "learning_rate": 1.0141523071585521e-05, "loss": 0.7151, "step": 299610 }, { "epoch": 25.95, "learning_rate": 1.0139354689487683e-05, "loss": 0.5956, "step": 299620 }, { "epoch": 25.95, "learning_rate": 1.0137186307389846e-05, "loss": 0.5532, "step": 299630 }, { "epoch": 25.95, "learning_rate": 1.0135017925292007e-05, "loss": 0.6698, "step": 299640 }, { "epoch": 25.95, "learning_rate": 1.0132849543194172e-05, "loss": 0.6123, "step": 299650 }, { "epoch": 25.95, "learning_rate": 1.0130681161096333e-05, "loss": 0.6988, "step": 299660 }, { "epoch": 25.95, "learning_rate": 1.0128512778998496e-05, "loss": 0.6062, "step": 299670 }, { "epoch": 25.95, "learning_rate": 1.0126344396900659e-05, "loss": 0.7191, "step": 299680 }, { "epoch": 25.96, "learning_rate": 1.012417601480282e-05, "loss": 0.7237, "step": 299690 }, { "epoch": 25.96, "learning_rate": 1.0122007632704983e-05, "loss": 0.6506, "step": 299700 }, { "epoch": 25.96, "learning_rate": 1.0119839250607145e-05, "loss": 0.6873, "step": 299710 }, { "epoch": 25.96, "learning_rate": 1.011767086850931e-05, "loss": 0.5877, "step": 299720 }, { "epoch": 25.96, "learning_rate": 1.011550248641147e-05, "loss": 0.6239, "step": 299730 }, { "epoch": 25.96, "learning_rate": 1.0113334104313634e-05, "loss": 0.5601, "step": 299740 }, { "epoch": 25.96, "learning_rate": 1.0111165722215795e-05, "loss": 0.5283, "step": 299750 }, { "epoch": 25.96, "learning_rate": 1.010899734011796e-05, "loss": 0.5332, "step": 299760 }, { "epoch": 25.96, "learning_rate": 1.0106828958020121e-05, "loss": 0.602, "step": 299770 }, { "epoch": 25.96, "learning_rate": 1.0104660575922284e-05, "loss": 0.6981, "step": 299780 }, { "epoch": 25.96, "learning_rate": 1.0102492193824447e-05, "loss": 0.6718, "step": 299790 }, { "epoch": 25.96, "learning_rate": 1.010032381172661e-05, "loss": 0.6586, "step": 299800 }, { "epoch": 25.97, "learning_rate": 1.0098155429628771e-05, "loss": 0.7629, "step": 299810 }, { "epoch": 25.97, "learning_rate": 1.0095987047530936e-05, "loss": 0.7565, "step": 299820 }, { "epoch": 25.97, "learning_rate": 1.0093818665433097e-05, "loss": 0.5921, "step": 299830 }, { "epoch": 25.97, "learning_rate": 1.009165028333526e-05, "loss": 0.5204, "step": 299840 }, { "epoch": 25.97, "learning_rate": 1.0089481901237422e-05, "loss": 0.7338, "step": 299850 }, { "epoch": 25.97, "learning_rate": 1.0087313519139586e-05, "loss": 0.6124, "step": 299860 }, { "epoch": 25.97, "learning_rate": 1.0085145137041748e-05, "loss": 0.482, "step": 299870 }, { "epoch": 25.97, "learning_rate": 1.008297675494391e-05, "loss": 0.4709, "step": 299880 }, { "epoch": 25.97, "learning_rate": 1.0080808372846072e-05, "loss": 0.5235, "step": 299890 }, { "epoch": 25.97, "learning_rate": 1.0078639990748237e-05, "loss": 0.5873, "step": 299900 }, { "epoch": 25.97, "learning_rate": 1.0076471608650398e-05, "loss": 0.8, "step": 299910 }, { "epoch": 25.97, "learning_rate": 1.0074303226552561e-05, "loss": 0.5716, "step": 299920 }, { "epoch": 25.98, "learning_rate": 1.0072134844454724e-05, "loss": 0.6247, "step": 299930 }, { "epoch": 25.98, "learning_rate": 1.0069966462356885e-05, "loss": 0.8853, "step": 299940 }, { "epoch": 25.98, "learning_rate": 1.0067798080259048e-05, "loss": 0.6591, "step": 299950 }, { "epoch": 25.98, "learning_rate": 1.006562969816121e-05, "loss": 0.6862, "step": 299960 }, { "epoch": 25.98, "learning_rate": 1.0063461316063374e-05, "loss": 0.7357, "step": 299970 }, { "epoch": 25.98, "learning_rate": 1.0061292933965536e-05, "loss": 0.5691, "step": 299980 }, { "epoch": 25.98, "learning_rate": 1.0059124551867699e-05, "loss": 0.6177, "step": 299990 }, { "epoch": 25.98, "learning_rate": 1.005695616976986e-05, "loss": 0.5592, "step": 300000 }, { "epoch": 25.98, "learning_rate": 1.0054787787672025e-05, "loss": 0.6911, "step": 300010 }, { "epoch": 25.98, "learning_rate": 1.0052619405574186e-05, "loss": 0.6171, "step": 300020 }, { "epoch": 25.98, "learning_rate": 1.0050451023476349e-05, "loss": 0.5061, "step": 300030 }, { "epoch": 25.99, "learning_rate": 1.0048282641378512e-05, "loss": 0.6545, "step": 300040 }, { "epoch": 25.99, "learning_rate": 1.0046114259280675e-05, "loss": 0.7018, "step": 300050 }, { "epoch": 25.99, "learning_rate": 1.0043945877182836e-05, "loss": 0.6894, "step": 300060 }, { "epoch": 25.99, "learning_rate": 1.0041777495085001e-05, "loss": 0.6, "step": 300070 }, { "epoch": 25.99, "learning_rate": 1.0039609112987162e-05, "loss": 0.6984, "step": 300080 }, { "epoch": 25.99, "learning_rate": 1.0037440730889325e-05, "loss": 0.6742, "step": 300090 }, { "epoch": 25.99, "learning_rate": 1.0035272348791487e-05, "loss": 0.6255, "step": 300100 }, { "epoch": 25.99, "learning_rate": 1.0033103966693651e-05, "loss": 0.5141, "step": 300110 }, { "epoch": 25.99, "learning_rate": 1.0030935584595813e-05, "loss": 0.6584, "step": 300120 }, { "epoch": 25.99, "learning_rate": 1.0028767202497976e-05, "loss": 0.5525, "step": 300130 }, { "epoch": 25.99, "learning_rate": 1.0026598820400137e-05, "loss": 0.5347, "step": 300140 }, { "epoch": 25.99, "learning_rate": 1.0024430438302302e-05, "loss": 0.6014, "step": 300150 }, { "epoch": 26.0, "learning_rate": 1.0022262056204463e-05, "loss": 0.6382, "step": 300160 }, { "epoch": 26.0, "learning_rate": 1.0020093674106626e-05, "loss": 0.5629, "step": 300170 }, { "epoch": 26.0, "learning_rate": 1.0017925292008789e-05, "loss": 0.8132, "step": 300180 }, { "epoch": 26.0, "learning_rate": 1.0015756909910952e-05, "loss": 0.6817, "step": 300190 }, { "epoch": 26.0, "learning_rate": 1.0013588527813113e-05, "loss": 0.553, "step": 300200 }, { "epoch": 26.0, "eval_cer": 0.5352426853958755, "eval_em": 0.06188757091284167, "eval_f1": 0.06188757091284167, "eval_loss": 1.5685616731643677, "eval_runtime": 254.0159, "eval_samples_per_second": 7.633, "eval_steps_per_second": 0.957, "eval_wer": 0.9381124290871583, "step": 300209 }, { "epoch": 26.0, "learning_rate": 1.0011420145715275e-05, "loss": 0.5341, "step": 300210 }, { "epoch": 26.0, "learning_rate": 1.000925176361744e-05, "loss": 0.6769, "step": 300220 }, { "epoch": 26.0, "learning_rate": 1.00070833815196e-05, "loss": 0.5761, "step": 300230 }, { "epoch": 26.0, "learning_rate": 1.0004914999421764e-05, "loss": 0.4883, "step": 300240 }, { "epoch": 26.0, "learning_rate": 1.0002746617323925e-05, "loss": 0.6151, "step": 300250 }, { "epoch": 26.0, "learning_rate": 1.000057823522609e-05, "loss": 0.5355, "step": 300260 }, { "epoch": 26.01, "learning_rate": 9.998409853128251e-06, "loss": 0.6883, "step": 300270 }, { "epoch": 26.01, "learning_rate": 9.996241471030414e-06, "loss": 0.5782, "step": 300280 }, { "epoch": 26.01, "learning_rate": 9.994073088932577e-06, "loss": 0.6556, "step": 300290 }, { "epoch": 26.01, "learning_rate": 9.99190470683474e-06, "loss": 0.5547, "step": 300300 }, { "epoch": 26.01, "learning_rate": 9.989736324736901e-06, "loss": 0.5506, "step": 300310 }, { "epoch": 26.01, "learning_rate": 9.987567942639066e-06, "loss": 0.5825, "step": 300320 }, { "epoch": 26.01, "learning_rate": 9.985399560541227e-06, "loss": 0.5884, "step": 300330 }, { "epoch": 26.01, "learning_rate": 9.98323117844339e-06, "loss": 0.5815, "step": 300340 }, { "epoch": 26.01, "learning_rate": 9.981062796345552e-06, "loss": 0.6236, "step": 300350 }, { "epoch": 26.01, "learning_rate": 9.978894414247716e-06, "loss": 0.5334, "step": 300360 }, { "epoch": 26.01, "learning_rate": 9.976726032149878e-06, "loss": 0.6596, "step": 300370 }, { "epoch": 26.01, "learning_rate": 9.97455765005204e-06, "loss": 0.5426, "step": 300380 }, { "epoch": 26.02, "learning_rate": 9.972389267954202e-06, "loss": 0.6045, "step": 300390 }, { "epoch": 26.02, "learning_rate": 9.970220885856367e-06, "loss": 0.6635, "step": 300400 }, { "epoch": 26.02, "learning_rate": 9.968052503758528e-06, "loss": 0.6299, "step": 300410 }, { "epoch": 26.02, "learning_rate": 9.96588412166069e-06, "loss": 0.6252, "step": 300420 }, { "epoch": 26.02, "learning_rate": 9.963715739562854e-06, "loss": 0.6584, "step": 300430 }, { "epoch": 26.02, "learning_rate": 9.961547357465017e-06, "loss": 0.5563, "step": 300440 }, { "epoch": 26.02, "learning_rate": 9.959378975367178e-06, "loss": 0.6309, "step": 300450 }, { "epoch": 26.02, "learning_rate": 9.95721059326934e-06, "loss": 0.6889, "step": 300460 }, { "epoch": 26.02, "learning_rate": 9.955042211171504e-06, "loss": 0.6629, "step": 300470 }, { "epoch": 26.02, "learning_rate": 9.952873829073665e-06, "loss": 0.5869, "step": 300480 }, { "epoch": 26.02, "learning_rate": 9.950705446975828e-06, "loss": 0.4883, "step": 300490 }, { "epoch": 26.03, "learning_rate": 9.948537064877991e-06, "loss": 0.5806, "step": 300500 }, { "epoch": 26.03, "learning_rate": 9.946368682780154e-06, "loss": 0.6358, "step": 300510 }, { "epoch": 26.03, "learning_rate": 9.944200300682316e-06, "loss": 0.5819, "step": 300520 }, { "epoch": 26.03, "learning_rate": 9.942031918584479e-06, "loss": 0.6384, "step": 300530 }, { "epoch": 26.03, "learning_rate": 9.939863536486642e-06, "loss": 0.7036, "step": 300540 }, { "epoch": 26.03, "learning_rate": 9.937695154388805e-06, "loss": 0.6655, "step": 300550 }, { "epoch": 26.03, "learning_rate": 9.935526772290966e-06, "loss": 0.6179, "step": 300560 }, { "epoch": 26.03, "learning_rate": 9.93335839019313e-06, "loss": 0.4708, "step": 300570 }, { "epoch": 26.03, "learning_rate": 9.931190008095292e-06, "loss": 0.5419, "step": 300580 }, { "epoch": 26.03, "learning_rate": 9.929021625997455e-06, "loss": 0.5926, "step": 300590 }, { "epoch": 26.03, "learning_rate": 9.926853243899616e-06, "loss": 0.5948, "step": 300600 }, { "epoch": 26.03, "learning_rate": 9.924684861801781e-06, "loss": 0.5258, "step": 300610 }, { "epoch": 26.04, "learning_rate": 9.922516479703942e-06, "loss": 0.6904, "step": 300620 }, { "epoch": 26.04, "learning_rate": 9.920348097606105e-06, "loss": 0.5168, "step": 300630 }, { "epoch": 26.04, "learning_rate": 9.918179715508268e-06, "loss": 0.4714, "step": 300640 }, { "epoch": 26.04, "learning_rate": 9.916011333410431e-06, "loss": 0.5423, "step": 300650 }, { "epoch": 26.04, "learning_rate": 9.913842951312593e-06, "loss": 0.5469, "step": 300660 }, { "epoch": 26.04, "learning_rate": 9.911674569214756e-06, "loss": 0.6, "step": 300670 }, { "epoch": 26.04, "learning_rate": 9.909506187116919e-06, "loss": 0.6176, "step": 300680 }, { "epoch": 26.04, "learning_rate": 9.907337805019082e-06, "loss": 0.6146, "step": 300690 }, { "epoch": 26.04, "learning_rate": 9.905169422921243e-06, "loss": 0.5675, "step": 300700 }, { "epoch": 26.04, "learning_rate": 9.903001040823404e-06, "loss": 0.5966, "step": 300710 }, { "epoch": 26.04, "learning_rate": 9.900832658725569e-06, "loss": 0.5714, "step": 300720 }, { "epoch": 26.05, "learning_rate": 9.89866427662773e-06, "loss": 0.795, "step": 300730 }, { "epoch": 26.05, "learning_rate": 9.896495894529893e-06, "loss": 0.6206, "step": 300740 }, { "epoch": 26.05, "learning_rate": 9.894327512432056e-06, "loss": 0.5649, "step": 300750 }, { "epoch": 26.05, "learning_rate": 9.89215913033422e-06, "loss": 0.5588, "step": 300760 }, { "epoch": 26.05, "learning_rate": 9.88999074823638e-06, "loss": 0.6742, "step": 300770 }, { "epoch": 26.05, "learning_rate": 9.887822366138545e-06, "loss": 0.5623, "step": 300780 }, { "epoch": 26.05, "learning_rate": 9.885653984040707e-06, "loss": 0.6443, "step": 300790 }, { "epoch": 26.05, "learning_rate": 9.88348560194287e-06, "loss": 0.6454, "step": 300800 }, { "epoch": 26.05, "learning_rate": 9.881317219845031e-06, "loss": 0.6511, "step": 300810 }, { "epoch": 26.05, "learning_rate": 9.879148837747196e-06, "loss": 0.4645, "step": 300820 }, { "epoch": 26.05, "learning_rate": 9.876980455649357e-06, "loss": 0.6528, "step": 300830 }, { "epoch": 26.05, "learning_rate": 9.87481207355152e-06, "loss": 0.5813, "step": 300840 }, { "epoch": 26.06, "learning_rate": 9.872643691453681e-06, "loss": 0.5885, "step": 300850 }, { "epoch": 26.06, "learning_rate": 9.870475309355846e-06, "loss": 0.5428, "step": 300860 }, { "epoch": 26.06, "learning_rate": 9.868306927258007e-06, "loss": 0.6361, "step": 300870 }, { "epoch": 26.06, "learning_rate": 9.86613854516017e-06, "loss": 0.5569, "step": 300880 }, { "epoch": 26.06, "learning_rate": 9.863970163062333e-06, "loss": 0.6067, "step": 300890 }, { "epoch": 26.06, "learning_rate": 9.861801780964496e-06, "loss": 0.4815, "step": 300900 }, { "epoch": 26.06, "learning_rate": 9.859633398866658e-06, "loss": 0.7766, "step": 300910 }, { "epoch": 26.06, "learning_rate": 9.857465016768822e-06, "loss": 0.3796, "step": 300920 }, { "epoch": 26.06, "learning_rate": 9.855296634670984e-06, "loss": 0.6052, "step": 300930 }, { "epoch": 26.06, "learning_rate": 9.853128252573147e-06, "loss": 0.559, "step": 300940 }, { "epoch": 26.06, "learning_rate": 9.850959870475308e-06, "loss": 0.6841, "step": 300950 }, { "epoch": 26.07, "learning_rate": 9.848791488377473e-06, "loss": 0.6219, "step": 300960 }, { "epoch": 26.07, "learning_rate": 9.846623106279634e-06, "loss": 0.5335, "step": 300970 }, { "epoch": 26.07, "learning_rate": 9.844454724181795e-06, "loss": 0.486, "step": 300980 }, { "epoch": 26.07, "learning_rate": 9.842286342083958e-06, "loss": 0.6276, "step": 300990 }, { "epoch": 26.07, "learning_rate": 9.840117959986121e-06, "loss": 0.5268, "step": 301000 }, { "epoch": 26.07, "learning_rate": 9.837949577888284e-06, "loss": 0.5825, "step": 301010 }, { "epoch": 26.07, "learning_rate": 9.835781195790446e-06, "loss": 0.4832, "step": 301020 }, { "epoch": 26.07, "learning_rate": 9.83361281369261e-06, "loss": 0.6729, "step": 301030 }, { "epoch": 26.07, "learning_rate": 9.831444431594772e-06, "loss": 0.6582, "step": 301040 }, { "epoch": 26.07, "learning_rate": 9.829276049496935e-06, "loss": 0.5973, "step": 301050 }, { "epoch": 26.07, "learning_rate": 9.827107667399096e-06, "loss": 0.7592, "step": 301060 }, { "epoch": 26.07, "learning_rate": 9.82493928530126e-06, "loss": 0.7237, "step": 301070 }, { "epoch": 26.08, "learning_rate": 9.822770903203422e-06, "loss": 0.5488, "step": 301080 }, { "epoch": 26.08, "learning_rate": 9.820602521105585e-06, "loss": 0.5483, "step": 301090 }, { "epoch": 26.08, "learning_rate": 9.818434139007746e-06, "loss": 0.5, "step": 301100 }, { "epoch": 26.08, "learning_rate": 9.816265756909911e-06, "loss": 0.5835, "step": 301110 }, { "epoch": 26.08, "learning_rate": 9.814097374812072e-06, "loss": 0.6593, "step": 301120 }, { "epoch": 26.08, "learning_rate": 9.811928992714235e-06, "loss": 0.5633, "step": 301130 }, { "epoch": 26.08, "learning_rate": 9.809760610616398e-06, "loss": 0.636, "step": 301140 }, { "epoch": 26.08, "learning_rate": 9.807592228518561e-06, "loss": 0.5498, "step": 301150 }, { "epoch": 26.08, "learning_rate": 9.805423846420723e-06, "loss": 0.6285, "step": 301160 }, { "epoch": 26.08, "learning_rate": 9.803255464322887e-06, "loss": 0.6571, "step": 301170 }, { "epoch": 26.08, "learning_rate": 9.801087082225049e-06, "loss": 0.633, "step": 301180 }, { "epoch": 26.08, "learning_rate": 9.798918700127212e-06, "loss": 0.5945, "step": 301190 }, { "epoch": 26.09, "learning_rate": 9.796750318029373e-06, "loss": 0.6345, "step": 301200 }, { "epoch": 26.09, "learning_rate": 9.794581935931538e-06, "loss": 0.6091, "step": 301210 }, { "epoch": 26.09, "learning_rate": 9.792413553833699e-06, "loss": 0.5359, "step": 301220 }, { "epoch": 26.09, "learning_rate": 9.79024517173586e-06, "loss": 0.4496, "step": 301230 }, { "epoch": 26.09, "learning_rate": 9.788076789638023e-06, "loss": 0.6025, "step": 301240 }, { "epoch": 26.09, "learning_rate": 9.785908407540186e-06, "loss": 0.5332, "step": 301250 }, { "epoch": 26.09, "learning_rate": 9.78374002544235e-06, "loss": 0.5892, "step": 301260 }, { "epoch": 26.09, "learning_rate": 9.78157164334451e-06, "loss": 0.7248, "step": 301270 }, { "epoch": 26.09, "learning_rate": 9.779403261246675e-06, "loss": 0.5967, "step": 301280 }, { "epoch": 26.09, "learning_rate": 9.777234879148837e-06, "loss": 0.5042, "step": 301290 }, { "epoch": 26.09, "learning_rate": 9.775066497051e-06, "loss": 0.5949, "step": 301300 }, { "epoch": 26.1, "learning_rate": 9.772898114953161e-06, "loss": 0.5704, "step": 301310 }, { "epoch": 26.1, "learning_rate": 9.770729732855326e-06, "loss": 0.5498, "step": 301320 }, { "epoch": 26.1, "learning_rate": 9.768561350757487e-06, "loss": 0.7014, "step": 301330 }, { "epoch": 26.1, "learning_rate": 9.76639296865965e-06, "loss": 0.5385, "step": 301340 }, { "epoch": 26.1, "learning_rate": 9.764224586561811e-06, "loss": 0.5677, "step": 301350 }, { "epoch": 26.1, "learning_rate": 9.762056204463976e-06, "loss": 0.5269, "step": 301360 }, { "epoch": 26.1, "learning_rate": 9.759887822366137e-06, "loss": 0.581, "step": 301370 }, { "epoch": 26.1, "learning_rate": 9.7577194402683e-06, "loss": 0.582, "step": 301380 }, { "epoch": 26.1, "learning_rate": 9.755551058170463e-06, "loss": 0.5229, "step": 301390 }, { "epoch": 26.1, "learning_rate": 9.753382676072626e-06, "loss": 0.6975, "step": 301400 }, { "epoch": 26.1, "learning_rate": 9.751214293974788e-06, "loss": 0.5165, "step": 301410 }, { "epoch": 26.1, "learning_rate": 9.749045911876952e-06, "loss": 0.5169, "step": 301420 }, { "epoch": 26.11, "learning_rate": 9.746877529779114e-06, "loss": 0.5058, "step": 301430 }, { "epoch": 26.11, "learning_rate": 9.744709147681277e-06, "loss": 0.7097, "step": 301440 }, { "epoch": 26.11, "learning_rate": 9.742540765583438e-06, "loss": 0.6525, "step": 301450 }, { "epoch": 26.11, "learning_rate": 9.740372383485603e-06, "loss": 0.6551, "step": 301460 }, { "epoch": 26.11, "learning_rate": 9.738204001387764e-06, "loss": 0.6226, "step": 301470 }, { "epoch": 26.11, "learning_rate": 9.736035619289925e-06, "loss": 0.5414, "step": 301480 }, { "epoch": 26.11, "learning_rate": 9.733867237192088e-06, "loss": 0.6299, "step": 301490 }, { "epoch": 26.11, "learning_rate": 9.731698855094251e-06, "loss": 0.6111, "step": 301500 }, { "epoch": 26.11, "learning_rate": 9.729530472996414e-06, "loss": 0.608, "step": 301510 }, { "epoch": 26.11, "learning_rate": 9.727362090898575e-06, "loss": 0.5408, "step": 301520 }, { "epoch": 26.11, "learning_rate": 9.72519370880074e-06, "loss": 0.5613, "step": 301530 }, { "epoch": 26.12, "learning_rate": 9.723025326702901e-06, "loss": 0.5971, "step": 301540 }, { "epoch": 26.12, "learning_rate": 9.720856944605064e-06, "loss": 0.5278, "step": 301550 }, { "epoch": 26.12, "learning_rate": 9.718688562507226e-06, "loss": 0.4754, "step": 301560 }, { "epoch": 26.12, "learning_rate": 9.71652018040939e-06, "loss": 0.6292, "step": 301570 }, { "epoch": 26.12, "learning_rate": 9.714351798311552e-06, "loss": 0.5868, "step": 301580 }, { "epoch": 26.12, "learning_rate": 9.712183416213715e-06, "loss": 0.5795, "step": 301590 }, { "epoch": 26.12, "learning_rate": 9.710015034115876e-06, "loss": 0.6827, "step": 301600 }, { "epoch": 26.12, "learning_rate": 9.70784665201804e-06, "loss": 0.5541, "step": 301610 }, { "epoch": 26.12, "learning_rate": 9.705678269920202e-06, "loss": 0.5792, "step": 301620 }, { "epoch": 26.12, "learning_rate": 9.703509887822365e-06, "loss": 0.5824, "step": 301630 }, { "epoch": 26.12, "learning_rate": 9.701341505724528e-06, "loss": 0.6395, "step": 301640 }, { "epoch": 26.12, "learning_rate": 9.699173123626691e-06, "loss": 0.6159, "step": 301650 }, { "epoch": 26.13, "learning_rate": 9.697004741528852e-06, "loss": 0.5786, "step": 301660 }, { "epoch": 26.13, "learning_rate": 9.694836359431017e-06, "loss": 0.6073, "step": 301670 }, { "epoch": 26.13, "learning_rate": 9.692667977333178e-06, "loss": 0.6292, "step": 301680 }, { "epoch": 26.13, "learning_rate": 9.690499595235341e-06, "loss": 0.5663, "step": 301690 }, { "epoch": 26.13, "learning_rate": 9.688331213137503e-06, "loss": 0.5406, "step": 301700 }, { "epoch": 26.13, "learning_rate": 9.686162831039667e-06, "loss": 0.6429, "step": 301710 }, { "epoch": 26.13, "learning_rate": 9.683994448941829e-06, "loss": 0.6364, "step": 301720 }, { "epoch": 26.13, "learning_rate": 9.681826066843992e-06, "loss": 0.5845, "step": 301730 }, { "epoch": 26.13, "learning_rate": 9.679657684746153e-06, "loss": 0.6875, "step": 301740 }, { "epoch": 26.13, "learning_rate": 9.677489302648316e-06, "loss": 0.6129, "step": 301750 }, { "epoch": 26.13, "learning_rate": 9.675320920550479e-06, "loss": 0.5923, "step": 301760 }, { "epoch": 26.14, "learning_rate": 9.67315253845264e-06, "loss": 0.6158, "step": 301770 }, { "epoch": 26.14, "learning_rate": 9.670984156354805e-06, "loss": 0.5519, "step": 301780 }, { "epoch": 26.14, "learning_rate": 9.668815774256966e-06, "loss": 0.6728, "step": 301790 }, { "epoch": 26.14, "learning_rate": 9.66664739215913e-06, "loss": 0.7228, "step": 301800 }, { "epoch": 26.14, "learning_rate": 9.66447901006129e-06, "loss": 0.5475, "step": 301810 }, { "epoch": 26.14, "learning_rate": 9.662310627963455e-06, "loss": 0.4761, "step": 301820 }, { "epoch": 26.14, "learning_rate": 9.660142245865617e-06, "loss": 0.5739, "step": 301830 }, { "epoch": 26.14, "learning_rate": 9.65797386376778e-06, "loss": 0.4893, "step": 301840 }, { "epoch": 26.14, "learning_rate": 9.655805481669943e-06, "loss": 0.7311, "step": 301850 }, { "epoch": 26.14, "learning_rate": 9.653637099572106e-06, "loss": 0.6225, "step": 301860 }, { "epoch": 26.14, "learning_rate": 9.651468717474267e-06, "loss": 0.5297, "step": 301870 }, { "epoch": 26.14, "learning_rate": 9.64930033537643e-06, "loss": 0.5179, "step": 301880 }, { "epoch": 26.15, "learning_rate": 9.647131953278593e-06, "loss": 0.5906, "step": 301890 }, { "epoch": 26.15, "learning_rate": 9.644963571180756e-06, "loss": 0.5895, "step": 301900 }, { "epoch": 26.15, "learning_rate": 9.642795189082917e-06, "loss": 0.6188, "step": 301910 }, { "epoch": 26.15, "learning_rate": 9.640626806985082e-06, "loss": 0.5692, "step": 301920 }, { "epoch": 26.15, "learning_rate": 9.638458424887243e-06, "loss": 0.6956, "step": 301930 }, { "epoch": 26.15, "learning_rate": 9.636290042789406e-06, "loss": 0.6322, "step": 301940 }, { "epoch": 26.15, "learning_rate": 9.634121660691568e-06, "loss": 0.5759, "step": 301950 }, { "epoch": 26.15, "learning_rate": 9.631953278593732e-06, "loss": 0.6012, "step": 301960 }, { "epoch": 26.15, "learning_rate": 9.629784896495894e-06, "loss": 0.6873, "step": 301970 }, { "epoch": 26.15, "learning_rate": 9.627616514398057e-06, "loss": 0.548, "step": 301980 }, { "epoch": 26.15, "learning_rate": 9.62544813230022e-06, "loss": 0.5037, "step": 301990 }, { "epoch": 26.16, "learning_rate": 9.623279750202381e-06, "loss": 0.5627, "step": 302000 }, { "epoch": 26.16, "learning_rate": 9.621111368104544e-06, "loss": 0.5529, "step": 302010 }, { "epoch": 26.16, "learning_rate": 9.618942986006705e-06, "loss": 0.6255, "step": 302020 }, { "epoch": 26.16, "learning_rate": 9.61677460390887e-06, "loss": 0.5623, "step": 302030 }, { "epoch": 26.16, "learning_rate": 9.614606221811031e-06, "loss": 0.5856, "step": 302040 }, { "epoch": 26.16, "learning_rate": 9.612437839713194e-06, "loss": 0.676, "step": 302050 }, { "epoch": 26.16, "learning_rate": 9.610269457615356e-06, "loss": 0.465, "step": 302060 }, { "epoch": 26.16, "learning_rate": 9.60810107551752e-06, "loss": 0.536, "step": 302070 }, { "epoch": 26.16, "learning_rate": 9.605932693419682e-06, "loss": 0.5794, "step": 302080 }, { "epoch": 26.16, "learning_rate": 9.603764311321845e-06, "loss": 0.5444, "step": 302090 }, { "epoch": 26.16, "learning_rate": 9.601595929224008e-06, "loss": 0.7123, "step": 302100 }, { "epoch": 26.16, "learning_rate": 9.59942754712617e-06, "loss": 0.5902, "step": 302110 }, { "epoch": 26.17, "learning_rate": 9.597259165028332e-06, "loss": 0.5907, "step": 302120 }, { "epoch": 26.17, "learning_rate": 9.595090782930497e-06, "loss": 0.6597, "step": 302130 }, { "epoch": 26.17, "learning_rate": 9.592922400832658e-06, "loss": 0.6027, "step": 302140 }, { "epoch": 26.17, "learning_rate": 9.590754018734821e-06, "loss": 0.5381, "step": 302150 }, { "epoch": 26.17, "learning_rate": 9.588585636636982e-06, "loss": 0.5714, "step": 302160 }, { "epoch": 26.17, "learning_rate": 9.586417254539147e-06, "loss": 0.7, "step": 302170 }, { "epoch": 26.17, "learning_rate": 9.584248872441308e-06, "loss": 0.4745, "step": 302180 }, { "epoch": 26.17, "learning_rate": 9.582080490343471e-06, "loss": 0.6174, "step": 302190 }, { "epoch": 26.17, "learning_rate": 9.579912108245633e-06, "loss": 0.5523, "step": 302200 }, { "epoch": 26.17, "learning_rate": 9.577743726147797e-06, "loss": 0.623, "step": 302210 }, { "epoch": 26.17, "learning_rate": 9.575575344049959e-06, "loss": 0.4791, "step": 302220 }, { "epoch": 26.18, "learning_rate": 9.573406961952122e-06, "loss": 0.5989, "step": 302230 }, { "epoch": 26.18, "learning_rate": 9.571238579854285e-06, "loss": 0.5838, "step": 302240 }, { "epoch": 26.18, "learning_rate": 9.569070197756446e-06, "loss": 0.5586, "step": 302250 }, { "epoch": 26.18, "learning_rate": 9.566901815658609e-06, "loss": 0.6101, "step": 302260 }, { "epoch": 26.18, "learning_rate": 9.56473343356077e-06, "loss": 0.5434, "step": 302270 }, { "epoch": 26.18, "learning_rate": 9.562565051462935e-06, "loss": 0.5365, "step": 302280 }, { "epoch": 26.18, "learning_rate": 9.560396669365096e-06, "loss": 0.6443, "step": 302290 }, { "epoch": 26.18, "learning_rate": 9.55822828726726e-06, "loss": 0.7436, "step": 302300 }, { "epoch": 26.18, "learning_rate": 9.55605990516942e-06, "loss": 0.5867, "step": 302310 }, { "epoch": 26.18, "learning_rate": 9.553891523071585e-06, "loss": 0.6522, "step": 302320 }, { "epoch": 26.18, "learning_rate": 9.551723140973747e-06, "loss": 0.5843, "step": 302330 }, { "epoch": 26.18, "learning_rate": 9.54955475887591e-06, "loss": 0.568, "step": 302340 }, { "epoch": 26.19, "learning_rate": 9.547386376778073e-06, "loss": 0.7561, "step": 302350 }, { "epoch": 26.19, "learning_rate": 9.545217994680236e-06, "loss": 0.5145, "step": 302360 }, { "epoch": 26.19, "learning_rate": 9.543049612582397e-06, "loss": 0.4813, "step": 302370 }, { "epoch": 26.19, "learning_rate": 9.540881230484562e-06, "loss": 0.5016, "step": 302380 }, { "epoch": 26.19, "learning_rate": 9.538712848386723e-06, "loss": 0.5656, "step": 302390 }, { "epoch": 26.19, "learning_rate": 9.536544466288886e-06, "loss": 0.684, "step": 302400 }, { "epoch": 26.19, "learning_rate": 9.534376084191047e-06, "loss": 0.5146, "step": 302410 }, { "epoch": 26.19, "learning_rate": 9.532207702093212e-06, "loss": 0.5177, "step": 302420 }, { "epoch": 26.19, "learning_rate": 9.530039319995373e-06, "loss": 0.4845, "step": 302430 }, { "epoch": 26.19, "learning_rate": 9.527870937897536e-06, "loss": 0.558, "step": 302440 }, { "epoch": 26.19, "learning_rate": 9.525702555799698e-06, "loss": 0.6559, "step": 302450 }, { "epoch": 26.19, "learning_rate": 9.523534173701862e-06, "loss": 0.5728, "step": 302460 }, { "epoch": 26.2, "learning_rate": 9.521365791604024e-06, "loss": 0.6002, "step": 302470 }, { "epoch": 26.2, "learning_rate": 9.519197409506187e-06, "loss": 0.511, "step": 302480 }, { "epoch": 26.2, "learning_rate": 9.51702902740835e-06, "loss": 0.5366, "step": 302490 }, { "epoch": 26.2, "learning_rate": 9.514860645310513e-06, "loss": 0.7241, "step": 302500 }, { "epoch": 26.2, "learning_rate": 9.512692263212674e-06, "loss": 0.4964, "step": 302510 }, { "epoch": 26.2, "learning_rate": 9.510523881114835e-06, "loss": 0.5228, "step": 302520 }, { "epoch": 26.2, "learning_rate": 9.508355499017e-06, "loss": 0.5896, "step": 302530 }, { "epoch": 26.2, "learning_rate": 9.506187116919161e-06, "loss": 0.5286, "step": 302540 }, { "epoch": 26.2, "learning_rate": 9.504018734821324e-06, "loss": 0.5895, "step": 302550 }, { "epoch": 26.2, "learning_rate": 9.501850352723485e-06, "loss": 0.5503, "step": 302560 }, { "epoch": 26.2, "learning_rate": 9.49968197062565e-06, "loss": 0.6616, "step": 302570 }, { "epoch": 26.21, "learning_rate": 9.497513588527812e-06, "loss": 0.626, "step": 302580 }, { "epoch": 26.21, "learning_rate": 9.495345206429975e-06, "loss": 0.58, "step": 302590 }, { "epoch": 26.21, "learning_rate": 9.493176824332138e-06, "loss": 0.6547, "step": 302600 }, { "epoch": 26.21, "learning_rate": 9.4910084422343e-06, "loss": 0.6176, "step": 302610 }, { "epoch": 26.21, "learning_rate": 9.488840060136462e-06, "loss": 0.5973, "step": 302620 }, { "epoch": 26.21, "learning_rate": 9.486671678038627e-06, "loss": 0.6436, "step": 302630 }, { "epoch": 26.21, "learning_rate": 9.484503295940788e-06, "loss": 0.7079, "step": 302640 }, { "epoch": 26.21, "learning_rate": 9.48233491384295e-06, "loss": 0.566, "step": 302650 }, { "epoch": 26.21, "learning_rate": 9.480166531745112e-06, "loss": 0.6293, "step": 302660 }, { "epoch": 26.21, "learning_rate": 9.477998149647277e-06, "loss": 0.5569, "step": 302670 }, { "epoch": 26.21, "learning_rate": 9.475829767549438e-06, "loss": 0.6795, "step": 302680 }, { "epoch": 26.21, "learning_rate": 9.473661385451601e-06, "loss": 0.5076, "step": 302690 }, { "epoch": 26.22, "learning_rate": 9.471493003353762e-06, "loss": 0.6198, "step": 302700 }, { "epoch": 26.22, "learning_rate": 9.469324621255927e-06, "loss": 0.6902, "step": 302710 }, { "epoch": 26.22, "learning_rate": 9.467156239158088e-06, "loss": 0.7022, "step": 302720 }, { "epoch": 26.22, "learning_rate": 9.464987857060251e-06, "loss": 0.6031, "step": 302730 }, { "epoch": 26.22, "learning_rate": 9.462819474962414e-06, "loss": 0.4999, "step": 302740 }, { "epoch": 26.22, "learning_rate": 9.460651092864577e-06, "loss": 0.6615, "step": 302750 }, { "epoch": 26.22, "learning_rate": 9.458482710766739e-06, "loss": 0.6577, "step": 302760 }, { "epoch": 26.22, "learning_rate": 9.4563143286689e-06, "loss": 0.7297, "step": 302770 }, { "epoch": 26.22, "learning_rate": 9.454145946571065e-06, "loss": 0.537, "step": 302780 }, { "epoch": 26.22, "learning_rate": 9.451977564473226e-06, "loss": 0.5964, "step": 302790 }, { "epoch": 26.22, "learning_rate": 9.449809182375389e-06, "loss": 0.5839, "step": 302800 }, { "epoch": 26.23, "learning_rate": 9.44764080027755e-06, "loss": 0.5166, "step": 302810 }, { "epoch": 26.23, "learning_rate": 9.445472418179715e-06, "loss": 0.6201, "step": 302820 }, { "epoch": 26.23, "learning_rate": 9.443304036081876e-06, "loss": 0.5552, "step": 302830 }, { "epoch": 26.23, "learning_rate": 9.44113565398404e-06, "loss": 0.6427, "step": 302840 }, { "epoch": 26.23, "learning_rate": 9.438967271886202e-06, "loss": 0.5145, "step": 302850 }, { "epoch": 26.23, "learning_rate": 9.436798889788365e-06, "loss": 0.5454, "step": 302860 }, { "epoch": 26.23, "learning_rate": 9.434630507690527e-06, "loss": 0.5127, "step": 302870 }, { "epoch": 26.23, "learning_rate": 9.432462125592691e-06, "loss": 0.6919, "step": 302880 }, { "epoch": 26.23, "learning_rate": 9.430293743494853e-06, "loss": 0.545, "step": 302890 }, { "epoch": 26.23, "learning_rate": 9.428125361397016e-06, "loss": 0.6673, "step": 302900 }, { "epoch": 26.23, "learning_rate": 9.425956979299177e-06, "loss": 0.4285, "step": 302910 }, { "epoch": 26.23, "learning_rate": 9.423788597201342e-06, "loss": 0.5215, "step": 302920 }, { "epoch": 26.24, "learning_rate": 9.421620215103503e-06, "loss": 0.6789, "step": 302930 }, { "epoch": 26.24, "learning_rate": 9.419451833005666e-06, "loss": 0.586, "step": 302940 }, { "epoch": 26.24, "learning_rate": 9.417283450907827e-06, "loss": 0.5526, "step": 302950 }, { "epoch": 26.24, "learning_rate": 9.415115068809992e-06, "loss": 0.5099, "step": 302960 }, { "epoch": 26.24, "learning_rate": 9.412946686712153e-06, "loss": 0.6065, "step": 302970 }, { "epoch": 26.24, "learning_rate": 9.410778304614316e-06, "loss": 0.549, "step": 302980 }, { "epoch": 26.24, "learning_rate": 9.40860992251648e-06, "loss": 0.4988, "step": 302990 }, { "epoch": 26.24, "learning_rate": 9.406441540418642e-06, "loss": 0.6673, "step": 303000 }, { "epoch": 26.24, "learning_rate": 9.404273158320804e-06, "loss": 0.5284, "step": 303010 }, { "epoch": 26.24, "learning_rate": 9.402104776222965e-06, "loss": 0.5358, "step": 303020 }, { "epoch": 26.24, "learning_rate": 9.39993639412513e-06, "loss": 0.7045, "step": 303030 }, { "epoch": 26.25, "learning_rate": 9.397768012027291e-06, "loss": 0.6123, "step": 303040 }, { "epoch": 26.25, "learning_rate": 9.395599629929454e-06, "loss": 0.7383, "step": 303050 }, { "epoch": 26.25, "learning_rate": 9.393431247831617e-06, "loss": 0.6399, "step": 303060 }, { "epoch": 26.25, "learning_rate": 9.39126286573378e-06, "loss": 0.4691, "step": 303070 }, { "epoch": 26.25, "learning_rate": 9.389094483635941e-06, "loss": 0.5606, "step": 303080 }, { "epoch": 26.25, "learning_rate": 9.386926101538104e-06, "loss": 0.6143, "step": 303090 }, { "epoch": 26.25, "learning_rate": 9.384757719440267e-06, "loss": 0.6667, "step": 303100 }, { "epoch": 26.25, "learning_rate": 9.38258933734243e-06, "loss": 0.5685, "step": 303110 }, { "epoch": 26.25, "learning_rate": 9.380420955244592e-06, "loss": 0.5183, "step": 303120 }, { "epoch": 26.25, "learning_rate": 9.378252573146756e-06, "loss": 0.6255, "step": 303130 }, { "epoch": 26.25, "learning_rate": 9.376084191048918e-06, "loss": 0.6143, "step": 303140 }, { "epoch": 26.25, "learning_rate": 9.37391580895108e-06, "loss": 0.5569, "step": 303150 }, { "epoch": 26.26, "learning_rate": 9.371747426853242e-06, "loss": 0.6041, "step": 303160 }, { "epoch": 26.26, "learning_rate": 9.369579044755405e-06, "loss": 0.601, "step": 303170 }, { "epoch": 26.26, "learning_rate": 9.367410662657568e-06, "loss": 0.5979, "step": 303180 }, { "epoch": 26.26, "learning_rate": 9.365242280559731e-06, "loss": 0.6167, "step": 303190 }, { "epoch": 26.26, "learning_rate": 9.363073898461894e-06, "loss": 0.6675, "step": 303200 }, { "epoch": 26.26, "learning_rate": 9.360905516364055e-06, "loss": 0.6532, "step": 303210 }, { "epoch": 26.26, "learning_rate": 9.358737134266218e-06, "loss": 0.6321, "step": 303220 }, { "epoch": 26.26, "learning_rate": 9.356568752168381e-06, "loss": 0.6592, "step": 303230 }, { "epoch": 26.26, "learning_rate": 9.354400370070544e-06, "loss": 0.5888, "step": 303240 }, { "epoch": 26.26, "learning_rate": 9.352231987972706e-06, "loss": 0.4899, "step": 303250 }, { "epoch": 26.26, "learning_rate": 9.350063605874869e-06, "loss": 0.6534, "step": 303260 }, { "epoch": 26.27, "learning_rate": 9.347895223777032e-06, "loss": 0.6836, "step": 303270 }, { "epoch": 26.27, "learning_rate": 9.345726841679195e-06, "loss": 0.5965, "step": 303280 }, { "epoch": 26.27, "learning_rate": 9.343558459581358e-06, "loss": 0.7056, "step": 303290 }, { "epoch": 26.27, "learning_rate": 9.341390077483519e-06, "loss": 0.5198, "step": 303300 }, { "epoch": 26.27, "learning_rate": 9.339221695385682e-06, "loss": 0.6716, "step": 303310 }, { "epoch": 26.27, "learning_rate": 9.337053313287845e-06, "loss": 0.6505, "step": 303320 }, { "epoch": 26.27, "learning_rate": 9.334884931190008e-06, "loss": 0.4683, "step": 303330 }, { "epoch": 26.27, "learning_rate": 9.332716549092171e-06, "loss": 0.5338, "step": 303340 }, { "epoch": 26.27, "learning_rate": 9.330548166994332e-06, "loss": 0.495, "step": 303350 }, { "epoch": 26.27, "learning_rate": 9.328379784896495e-06, "loss": 0.6207, "step": 303360 }, { "epoch": 26.27, "learning_rate": 9.326211402798658e-06, "loss": 0.5723, "step": 303370 }, { "epoch": 26.27, "learning_rate": 9.324043020700821e-06, "loss": 0.612, "step": 303380 }, { "epoch": 26.28, "learning_rate": 9.321874638602983e-06, "loss": 0.4982, "step": 303390 }, { "epoch": 26.28, "learning_rate": 9.319706256505146e-06, "loss": 0.6007, "step": 303400 }, { "epoch": 26.28, "learning_rate": 9.317537874407307e-06, "loss": 0.655, "step": 303410 }, { "epoch": 26.28, "learning_rate": 9.31536949230947e-06, "loss": 0.5897, "step": 303420 }, { "epoch": 26.28, "learning_rate": 9.313201110211633e-06, "loss": 0.6322, "step": 303430 }, { "epoch": 26.28, "learning_rate": 9.311032728113796e-06, "loss": 0.5848, "step": 303440 }, { "epoch": 26.28, "learning_rate": 9.308864346015959e-06, "loss": 0.5873, "step": 303450 }, { "epoch": 26.28, "learning_rate": 9.30669596391812e-06, "loss": 0.6831, "step": 303460 }, { "epoch": 26.28, "learning_rate": 9.304527581820283e-06, "loss": 0.6257, "step": 303470 }, { "epoch": 26.28, "learning_rate": 9.302359199722446e-06, "loss": 0.5552, "step": 303480 }, { "epoch": 26.28, "learning_rate": 9.30019081762461e-06, "loss": 0.6527, "step": 303490 }, { "epoch": 26.29, "learning_rate": 9.29802243552677e-06, "loss": 0.6123, "step": 303500 }, { "epoch": 26.29, "learning_rate": 9.295854053428934e-06, "loss": 0.6172, "step": 303510 }, { "epoch": 26.29, "learning_rate": 9.293685671331097e-06, "loss": 0.6278, "step": 303520 }, { "epoch": 26.29, "learning_rate": 9.29151728923326e-06, "loss": 0.6275, "step": 303530 }, { "epoch": 26.29, "learning_rate": 9.289348907135423e-06, "loss": 0.6058, "step": 303540 }, { "epoch": 26.29, "learning_rate": 9.287180525037584e-06, "loss": 0.6256, "step": 303550 }, { "epoch": 26.29, "learning_rate": 9.285012142939747e-06, "loss": 0.5893, "step": 303560 }, { "epoch": 26.29, "learning_rate": 9.28284376084191e-06, "loss": 0.7084, "step": 303570 }, { "epoch": 26.29, "learning_rate": 9.280675378744073e-06, "loss": 0.5415, "step": 303580 }, { "epoch": 26.29, "learning_rate": 9.278506996646236e-06, "loss": 0.7225, "step": 303590 }, { "epoch": 26.29, "learning_rate": 9.276338614548397e-06, "loss": 0.6334, "step": 303600 }, { "epoch": 26.29, "learning_rate": 9.27417023245056e-06, "loss": 0.5688, "step": 303610 }, { "epoch": 26.3, "learning_rate": 9.272001850352723e-06, "loss": 0.4851, "step": 303620 }, { "epoch": 26.3, "learning_rate": 9.269833468254886e-06, "loss": 0.7522, "step": 303630 }, { "epoch": 26.3, "learning_rate": 9.267665086157048e-06, "loss": 0.7023, "step": 303640 }, { "epoch": 26.3, "learning_rate": 9.26549670405921e-06, "loss": 0.7151, "step": 303650 }, { "epoch": 26.3, "learning_rate": 9.263328321961372e-06, "loss": 0.6007, "step": 303660 }, { "epoch": 26.3, "learning_rate": 9.261159939863535e-06, "loss": 0.7904, "step": 303670 }, { "epoch": 26.3, "learning_rate": 9.258991557765698e-06, "loss": 0.5645, "step": 303680 }, { "epoch": 26.3, "learning_rate": 9.256823175667861e-06, "loss": 0.6049, "step": 303690 }, { "epoch": 26.3, "learning_rate": 9.254654793570024e-06, "loss": 0.5496, "step": 303700 }, { "epoch": 26.3, "learning_rate": 9.252486411472185e-06, "loss": 0.7036, "step": 303710 }, { "epoch": 26.3, "learning_rate": 9.250318029374348e-06, "loss": 0.5188, "step": 303720 }, { "epoch": 26.3, "learning_rate": 9.248149647276511e-06, "loss": 0.5593, "step": 303730 }, { "epoch": 26.31, "learning_rate": 9.245981265178674e-06, "loss": 0.7098, "step": 303740 }, { "epoch": 26.31, "learning_rate": 9.243812883080835e-06, "loss": 0.5145, "step": 303750 }, { "epoch": 26.31, "learning_rate": 9.241644500982998e-06, "loss": 0.5447, "step": 303760 }, { "epoch": 26.31, "learning_rate": 9.239476118885161e-06, "loss": 0.5664, "step": 303770 }, { "epoch": 26.31, "learning_rate": 9.237307736787324e-06, "loss": 0.5456, "step": 303780 }, { "epoch": 26.31, "learning_rate": 9.235139354689487e-06, "loss": 0.5892, "step": 303790 }, { "epoch": 26.31, "learning_rate": 9.232970972591649e-06, "loss": 0.66, "step": 303800 }, { "epoch": 26.31, "learning_rate": 9.230802590493812e-06, "loss": 0.5788, "step": 303810 }, { "epoch": 26.31, "learning_rate": 9.228634208395975e-06, "loss": 0.4948, "step": 303820 }, { "epoch": 26.31, "learning_rate": 9.226465826298138e-06, "loss": 0.4453, "step": 303830 }, { "epoch": 26.31, "learning_rate": 9.2242974442003e-06, "loss": 0.5929, "step": 303840 }, { "epoch": 26.32, "learning_rate": 9.222129062102462e-06, "loss": 0.6236, "step": 303850 }, { "epoch": 26.32, "learning_rate": 9.219960680004625e-06, "loss": 0.7288, "step": 303860 }, { "epoch": 26.32, "learning_rate": 9.217792297906788e-06, "loss": 0.612, "step": 303870 }, { "epoch": 26.32, "learning_rate": 9.215623915808951e-06, "loss": 0.7264, "step": 303880 }, { "epoch": 26.32, "learning_rate": 9.213455533711112e-06, "loss": 0.5214, "step": 303890 }, { "epoch": 26.32, "learning_rate": 9.211287151613275e-06, "loss": 0.6567, "step": 303900 }, { "epoch": 26.32, "learning_rate": 9.209118769515438e-06, "loss": 0.5108, "step": 303910 }, { "epoch": 26.32, "learning_rate": 9.2069503874176e-06, "loss": 0.6024, "step": 303920 }, { "epoch": 26.32, "learning_rate": 9.204782005319763e-06, "loss": 0.7427, "step": 303930 }, { "epoch": 26.32, "learning_rate": 9.202613623221926e-06, "loss": 0.5853, "step": 303940 }, { "epoch": 26.32, "learning_rate": 9.200445241124089e-06, "loss": 0.5085, "step": 303950 }, { "epoch": 26.32, "learning_rate": 9.19827685902625e-06, "loss": 0.5052, "step": 303960 }, { "epoch": 26.33, "learning_rate": 9.196108476928413e-06, "loss": 0.5727, "step": 303970 }, { "epoch": 26.33, "learning_rate": 9.193940094830576e-06, "loss": 0.5328, "step": 303980 }, { "epoch": 26.33, "learning_rate": 9.191771712732739e-06, "loss": 0.5216, "step": 303990 }, { "epoch": 26.33, "learning_rate": 9.189603330634902e-06, "loss": 0.6374, "step": 304000 }, { "epoch": 26.33, "learning_rate": 9.187434948537063e-06, "loss": 0.6409, "step": 304010 }, { "epoch": 26.33, "learning_rate": 9.185266566439226e-06, "loss": 0.5757, "step": 304020 }, { "epoch": 26.33, "learning_rate": 9.18309818434139e-06, "loss": 0.5529, "step": 304030 }, { "epoch": 26.33, "learning_rate": 9.180929802243552e-06, "loss": 0.4995, "step": 304040 }, { "epoch": 26.33, "learning_rate": 9.178761420145714e-06, "loss": 0.6011, "step": 304050 }, { "epoch": 26.33, "learning_rate": 9.176593038047877e-06, "loss": 0.5933, "step": 304060 }, { "epoch": 26.33, "learning_rate": 9.17442465595004e-06, "loss": 0.5338, "step": 304070 }, { "epoch": 26.34, "learning_rate": 9.172256273852203e-06, "loss": 0.5905, "step": 304080 }, { "epoch": 26.34, "learning_rate": 9.170087891754366e-06, "loss": 0.4877, "step": 304090 }, { "epoch": 26.34, "learning_rate": 9.167919509656527e-06, "loss": 0.5954, "step": 304100 }, { "epoch": 26.34, "learning_rate": 9.16575112755869e-06, "loss": 0.5878, "step": 304110 }, { "epoch": 26.34, "learning_rate": 9.163582745460853e-06, "loss": 0.63, "step": 304120 }, { "epoch": 26.34, "learning_rate": 9.161414363363016e-06, "loss": 0.6489, "step": 304130 }, { "epoch": 26.34, "learning_rate": 9.159245981265179e-06, "loss": 0.5938, "step": 304140 }, { "epoch": 26.34, "learning_rate": 9.15707759916734e-06, "loss": 0.5407, "step": 304150 }, { "epoch": 26.34, "learning_rate": 9.154909217069503e-06, "loss": 0.6474, "step": 304160 }, { "epoch": 26.34, "learning_rate": 9.152740834971666e-06, "loss": 0.6374, "step": 304170 }, { "epoch": 26.34, "learning_rate": 9.150572452873828e-06, "loss": 0.6367, "step": 304180 }, { "epoch": 26.34, "learning_rate": 9.14840407077599e-06, "loss": 0.5818, "step": 304190 }, { "epoch": 26.35, "learning_rate": 9.146235688678154e-06, "loss": 0.5258, "step": 304200 }, { "epoch": 26.35, "learning_rate": 9.144067306580315e-06, "loss": 0.6876, "step": 304210 }, { "epoch": 26.35, "learning_rate": 9.141898924482478e-06, "loss": 0.5385, "step": 304220 }, { "epoch": 26.35, "learning_rate": 9.139730542384641e-06, "loss": 0.4868, "step": 304230 }, { "epoch": 26.35, "learning_rate": 9.137562160286804e-06, "loss": 0.6045, "step": 304240 }, { "epoch": 26.35, "learning_rate": 9.135393778188967e-06, "loss": 0.5551, "step": 304250 }, { "epoch": 26.35, "learning_rate": 9.133225396091128e-06, "loss": 0.659, "step": 304260 }, { "epoch": 26.35, "learning_rate": 9.131057013993291e-06, "loss": 0.5963, "step": 304270 }, { "epoch": 26.35, "learning_rate": 9.128888631895454e-06, "loss": 0.5792, "step": 304280 }, { "epoch": 26.35, "learning_rate": 9.126720249797617e-06, "loss": 0.8153, "step": 304290 }, { "epoch": 26.35, "learning_rate": 9.124551867699779e-06, "loss": 0.6577, "step": 304300 }, { "epoch": 26.36, "learning_rate": 9.122383485601942e-06, "loss": 0.5215, "step": 304310 }, { "epoch": 26.36, "learning_rate": 9.120215103504105e-06, "loss": 0.4711, "step": 304320 }, { "epoch": 26.36, "learning_rate": 9.118046721406268e-06, "loss": 0.66, "step": 304330 }, { "epoch": 26.36, "learning_rate": 9.11587833930843e-06, "loss": 0.531, "step": 304340 }, { "epoch": 26.36, "learning_rate": 9.113709957210592e-06, "loss": 0.7424, "step": 304350 }, { "epoch": 26.36, "learning_rate": 9.111541575112755e-06, "loss": 0.5907, "step": 304360 }, { "epoch": 26.36, "learning_rate": 9.109373193014918e-06, "loss": 0.5988, "step": 304370 }, { "epoch": 26.36, "learning_rate": 9.107204810917081e-06, "loss": 0.5871, "step": 304380 }, { "epoch": 26.36, "learning_rate": 9.105036428819244e-06, "loss": 0.4506, "step": 304390 }, { "epoch": 26.36, "learning_rate": 9.102868046721405e-06, "loss": 0.6979, "step": 304400 }, { "epoch": 26.36, "learning_rate": 9.100699664623568e-06, "loss": 0.6725, "step": 304410 }, { "epoch": 26.36, "learning_rate": 9.098531282525731e-06, "loss": 0.7006, "step": 304420 }, { "epoch": 26.37, "learning_rate": 9.096362900427893e-06, "loss": 0.6965, "step": 304430 }, { "epoch": 26.37, "learning_rate": 9.094194518330056e-06, "loss": 0.6373, "step": 304440 }, { "epoch": 26.37, "learning_rate": 9.092026136232219e-06, "loss": 0.5349, "step": 304450 }, { "epoch": 26.37, "learning_rate": 9.08985775413438e-06, "loss": 0.589, "step": 304460 }, { "epoch": 26.37, "learning_rate": 9.087689372036543e-06, "loss": 0.5613, "step": 304470 }, { "epoch": 26.37, "learning_rate": 9.085520989938706e-06, "loss": 0.6309, "step": 304480 }, { "epoch": 26.37, "learning_rate": 9.083352607840869e-06, "loss": 0.6446, "step": 304490 }, { "epoch": 26.37, "learning_rate": 9.081184225743032e-06, "loss": 0.508, "step": 304500 }, { "epoch": 26.37, "learning_rate": 9.079015843645193e-06, "loss": 0.6338, "step": 304510 }, { "epoch": 26.37, "learning_rate": 9.076847461547356e-06, "loss": 0.5375, "step": 304520 }, { "epoch": 26.37, "learning_rate": 9.07467907944952e-06, "loss": 0.6769, "step": 304530 }, { "epoch": 26.38, "learning_rate": 9.072510697351682e-06, "loss": 0.5747, "step": 304540 }, { "epoch": 26.38, "learning_rate": 9.070342315253845e-06, "loss": 0.6056, "step": 304550 }, { "epoch": 26.38, "learning_rate": 9.068173933156007e-06, "loss": 0.5708, "step": 304560 }, { "epoch": 26.38, "learning_rate": 9.06600555105817e-06, "loss": 0.6679, "step": 304570 }, { "epoch": 26.38, "learning_rate": 9.063837168960333e-06, "loss": 0.5302, "step": 304580 }, { "epoch": 26.38, "learning_rate": 9.061668786862496e-06, "loss": 0.5881, "step": 304590 }, { "epoch": 26.38, "learning_rate": 9.059500404764657e-06, "loss": 0.6873, "step": 304600 }, { "epoch": 26.38, "learning_rate": 9.05733202266682e-06, "loss": 0.7087, "step": 304610 }, { "epoch": 26.38, "learning_rate": 9.055163640568983e-06, "loss": 0.6149, "step": 304620 }, { "epoch": 26.38, "learning_rate": 9.052995258471146e-06, "loss": 0.6154, "step": 304630 }, { "epoch": 26.38, "learning_rate": 9.050826876373309e-06, "loss": 0.5502, "step": 304640 }, { "epoch": 26.38, "learning_rate": 9.04865849427547e-06, "loss": 0.6083, "step": 304650 }, { "epoch": 26.39, "learning_rate": 9.046490112177633e-06, "loss": 0.6348, "step": 304660 }, { "epoch": 26.39, "learning_rate": 9.044321730079796e-06, "loss": 0.641, "step": 304670 }, { "epoch": 26.39, "learning_rate": 9.04215334798196e-06, "loss": 0.5177, "step": 304680 }, { "epoch": 26.39, "learning_rate": 9.03998496588412e-06, "loss": 0.7261, "step": 304690 }, { "epoch": 26.39, "learning_rate": 9.037816583786284e-06, "loss": 0.5694, "step": 304700 }, { "epoch": 26.39, "learning_rate": 9.035648201688445e-06, "loss": 0.6452, "step": 304710 }, { "epoch": 26.39, "learning_rate": 9.033479819590608e-06, "loss": 0.527, "step": 304720 }, { "epoch": 26.39, "learning_rate": 9.031311437492771e-06, "loss": 0.676, "step": 304730 }, { "epoch": 26.39, "learning_rate": 9.029143055394934e-06, "loss": 0.6789, "step": 304740 }, { "epoch": 26.39, "learning_rate": 9.026974673297097e-06, "loss": 0.553, "step": 304750 }, { "epoch": 26.39, "learning_rate": 9.024806291199258e-06, "loss": 0.5488, "step": 304760 }, { "epoch": 26.4, "learning_rate": 9.022637909101421e-06, "loss": 0.5918, "step": 304770 }, { "epoch": 26.4, "learning_rate": 9.020469527003584e-06, "loss": 0.535, "step": 304780 }, { "epoch": 26.4, "learning_rate": 9.018301144905747e-06, "loss": 0.4493, "step": 304790 }, { "epoch": 26.4, "learning_rate": 9.01613276280791e-06, "loss": 0.6111, "step": 304800 }, { "epoch": 26.4, "learning_rate": 9.013964380710072e-06, "loss": 0.7, "step": 304810 }, { "epoch": 26.4, "learning_rate": 9.011795998612235e-06, "loss": 0.656, "step": 304820 }, { "epoch": 26.4, "learning_rate": 9.009627616514398e-06, "loss": 0.5611, "step": 304830 }, { "epoch": 26.4, "learning_rate": 9.00745923441656e-06, "loss": 0.521, "step": 304840 }, { "epoch": 26.4, "learning_rate": 9.005290852318722e-06, "loss": 0.6963, "step": 304850 }, { "epoch": 26.4, "learning_rate": 9.003122470220885e-06, "loss": 0.5683, "step": 304860 }, { "epoch": 26.4, "learning_rate": 9.000954088123048e-06, "loss": 0.6941, "step": 304870 }, { "epoch": 26.4, "learning_rate": 8.99878570602521e-06, "loss": 0.5252, "step": 304880 }, { "epoch": 26.41, "learning_rate": 8.996617323927374e-06, "loss": 0.6916, "step": 304890 }, { "epoch": 26.41, "learning_rate": 8.994448941829535e-06, "loss": 0.5544, "step": 304900 }, { "epoch": 26.41, "learning_rate": 8.992280559731698e-06, "loss": 0.6852, "step": 304910 }, { "epoch": 26.41, "learning_rate": 8.990112177633861e-06, "loss": 0.6181, "step": 304920 }, { "epoch": 26.41, "learning_rate": 8.987943795536024e-06, "loss": 0.6951, "step": 304930 }, { "epoch": 26.41, "learning_rate": 8.985775413438187e-06, "loss": 0.5902, "step": 304940 }, { "epoch": 26.41, "learning_rate": 8.983607031340348e-06, "loss": 0.5316, "step": 304950 }, { "epoch": 26.41, "learning_rate": 8.98143864924251e-06, "loss": 0.4604, "step": 304960 }, { "epoch": 26.41, "learning_rate": 8.979270267144673e-06, "loss": 0.6261, "step": 304970 }, { "epoch": 26.41, "learning_rate": 8.977101885046836e-06, "loss": 0.5658, "step": 304980 }, { "epoch": 26.41, "learning_rate": 8.974933502948999e-06, "loss": 0.5801, "step": 304990 }, { "epoch": 26.41, "learning_rate": 8.972765120851162e-06, "loss": 0.59, "step": 305000 }, { "epoch": 26.42, "learning_rate": 8.970596738753323e-06, "loss": 0.6084, "step": 305010 }, { "epoch": 26.42, "learning_rate": 8.968428356655486e-06, "loss": 0.6719, "step": 305020 }, { "epoch": 26.42, "learning_rate": 8.966259974557649e-06, "loss": 0.6388, "step": 305030 }, { "epoch": 26.42, "learning_rate": 8.964091592459812e-06, "loss": 0.5876, "step": 305040 }, { "epoch": 26.42, "learning_rate": 8.961923210361975e-06, "loss": 0.6498, "step": 305050 }, { "epoch": 26.42, "learning_rate": 8.959754828264136e-06, "loss": 0.5768, "step": 305060 }, { "epoch": 26.42, "learning_rate": 8.9575864461663e-06, "loss": 0.586, "step": 305070 }, { "epoch": 26.42, "learning_rate": 8.955418064068462e-06, "loss": 0.5699, "step": 305080 }, { "epoch": 26.42, "learning_rate": 8.953249681970625e-06, "loss": 0.6476, "step": 305090 }, { "epoch": 26.42, "learning_rate": 8.951081299872787e-06, "loss": 0.6323, "step": 305100 }, { "epoch": 26.42, "learning_rate": 8.94891291777495e-06, "loss": 0.7305, "step": 305110 }, { "epoch": 26.43, "learning_rate": 8.946744535677113e-06, "loss": 0.6147, "step": 305120 }, { "epoch": 26.43, "learning_rate": 8.944576153579276e-06, "loss": 0.6658, "step": 305130 }, { "epoch": 26.43, "learning_rate": 8.942407771481439e-06, "loss": 0.6652, "step": 305140 }, { "epoch": 26.43, "learning_rate": 8.9402393893836e-06, "loss": 0.6632, "step": 305150 }, { "epoch": 26.43, "learning_rate": 8.938071007285763e-06, "loss": 0.8249, "step": 305160 }, { "epoch": 26.43, "learning_rate": 8.935902625187926e-06, "loss": 0.6923, "step": 305170 }, { "epoch": 26.43, "learning_rate": 8.933734243090089e-06, "loss": 0.5881, "step": 305180 }, { "epoch": 26.43, "learning_rate": 8.931565860992252e-06, "loss": 0.6547, "step": 305190 }, { "epoch": 26.43, "learning_rate": 8.929397478894413e-06, "loss": 0.6413, "step": 305200 }, { "epoch": 26.43, "learning_rate": 8.927229096796576e-06, "loss": 0.739, "step": 305210 }, { "epoch": 26.43, "learning_rate": 8.925060714698738e-06, "loss": 0.5928, "step": 305220 }, { "epoch": 26.43, "learning_rate": 8.9228923326009e-06, "loss": 0.5269, "step": 305230 }, { "epoch": 26.44, "learning_rate": 8.920723950503064e-06, "loss": 0.5467, "step": 305240 }, { "epoch": 26.44, "learning_rate": 8.918555568405227e-06, "loss": 0.6777, "step": 305250 }, { "epoch": 26.44, "learning_rate": 8.916387186307388e-06, "loss": 0.6315, "step": 305260 }, { "epoch": 26.44, "learning_rate": 8.914218804209551e-06, "loss": 0.6037, "step": 305270 }, { "epoch": 26.44, "learning_rate": 8.912050422111714e-06, "loss": 0.4838, "step": 305280 }, { "epoch": 26.44, "learning_rate": 8.909882040013877e-06, "loss": 0.5045, "step": 305290 }, { "epoch": 26.44, "learning_rate": 8.90771365791604e-06, "loss": 0.5857, "step": 305300 }, { "epoch": 26.44, "learning_rate": 8.905545275818201e-06, "loss": 0.62, "step": 305310 }, { "epoch": 26.44, "learning_rate": 8.903376893720364e-06, "loss": 0.6267, "step": 305320 }, { "epoch": 26.44, "learning_rate": 8.901208511622527e-06, "loss": 0.4651, "step": 305330 }, { "epoch": 26.44, "learning_rate": 8.89904012952469e-06, "loss": 0.6012, "step": 305340 }, { "epoch": 26.45, "learning_rate": 8.896871747426853e-06, "loss": 0.5294, "step": 305350 }, { "epoch": 26.45, "learning_rate": 8.894703365329015e-06, "loss": 0.6209, "step": 305360 }, { "epoch": 26.45, "learning_rate": 8.892534983231178e-06, "loss": 0.7269, "step": 305370 }, { "epoch": 26.45, "learning_rate": 8.89036660113334e-06, "loss": 0.5937, "step": 305380 }, { "epoch": 26.45, "learning_rate": 8.888198219035504e-06, "loss": 0.6536, "step": 305390 }, { "epoch": 26.45, "learning_rate": 8.886029836937665e-06, "loss": 0.5128, "step": 305400 }, { "epoch": 26.45, "learning_rate": 8.883861454839828e-06, "loss": 0.5776, "step": 305410 }, { "epoch": 26.45, "learning_rate": 8.881693072741991e-06, "loss": 0.6596, "step": 305420 }, { "epoch": 26.45, "learning_rate": 8.879524690644154e-06, "loss": 0.6212, "step": 305430 }, { "epoch": 26.45, "learning_rate": 8.877356308546317e-06, "loss": 0.7264, "step": 305440 }, { "epoch": 26.45, "learning_rate": 8.875187926448478e-06, "loss": 0.5884, "step": 305450 }, { "epoch": 26.45, "learning_rate": 8.873019544350641e-06, "loss": 0.6061, "step": 305460 }, { "epoch": 26.46, "learning_rate": 8.870851162252803e-06, "loss": 0.624, "step": 305470 }, { "epoch": 26.46, "learning_rate": 8.868682780154966e-06, "loss": 0.6339, "step": 305480 }, { "epoch": 26.46, "learning_rate": 8.866514398057129e-06, "loss": 0.5114, "step": 305490 }, { "epoch": 26.46, "learning_rate": 8.864346015959292e-06, "loss": 0.6975, "step": 305500 }, { "epoch": 26.46, "learning_rate": 8.862177633861453e-06, "loss": 0.6554, "step": 305510 }, { "epoch": 26.46, "learning_rate": 8.860009251763616e-06, "loss": 0.7171, "step": 305520 }, { "epoch": 26.46, "learning_rate": 8.857840869665779e-06, "loss": 0.6069, "step": 305530 }, { "epoch": 26.46, "learning_rate": 8.855672487567942e-06, "loss": 0.5809, "step": 305540 }, { "epoch": 26.46, "learning_rate": 8.853504105470105e-06, "loss": 0.6659, "step": 305550 }, { "epoch": 26.46, "learning_rate": 8.851335723372266e-06, "loss": 0.5214, "step": 305560 }, { "epoch": 26.46, "learning_rate": 8.84916734127443e-06, "loss": 0.6138, "step": 305570 }, { "epoch": 26.47, "learning_rate": 8.846998959176592e-06, "loss": 0.5913, "step": 305580 }, { "epoch": 26.47, "learning_rate": 8.844830577078755e-06, "loss": 0.6311, "step": 305590 }, { "epoch": 26.47, "learning_rate": 8.842662194980918e-06, "loss": 0.6201, "step": 305600 }, { "epoch": 26.47, "learning_rate": 8.84049381288308e-06, "loss": 0.5361, "step": 305610 }, { "epoch": 26.47, "learning_rate": 8.838325430785243e-06, "loss": 0.5524, "step": 305620 }, { "epoch": 26.47, "learning_rate": 8.836157048687406e-06, "loss": 0.6275, "step": 305630 }, { "epoch": 26.47, "learning_rate": 8.833988666589569e-06, "loss": 0.5914, "step": 305640 }, { "epoch": 26.47, "learning_rate": 8.83182028449173e-06, "loss": 0.6475, "step": 305650 }, { "epoch": 26.47, "learning_rate": 8.829651902393893e-06, "loss": 0.5601, "step": 305660 }, { "epoch": 26.47, "learning_rate": 8.827483520296056e-06, "loss": 0.6601, "step": 305670 }, { "epoch": 26.47, "learning_rate": 8.825315138198219e-06, "loss": 0.5774, "step": 305680 }, { "epoch": 26.47, "learning_rate": 8.823146756100382e-06, "loss": 0.6679, "step": 305690 }, { "epoch": 26.48, "learning_rate": 8.820978374002543e-06, "loss": 0.5307, "step": 305700 }, { "epoch": 26.48, "learning_rate": 8.818809991904706e-06, "loss": 0.6325, "step": 305710 }, { "epoch": 26.48, "learning_rate": 8.816641609806868e-06, "loss": 0.726, "step": 305720 }, { "epoch": 26.48, "learning_rate": 8.81447322770903e-06, "loss": 0.7303, "step": 305730 }, { "epoch": 26.48, "learning_rate": 8.812304845611194e-06, "loss": 0.5488, "step": 305740 }, { "epoch": 26.48, "learning_rate": 8.810136463513357e-06, "loss": 0.5723, "step": 305750 }, { "epoch": 26.48, "learning_rate": 8.80796808141552e-06, "loss": 0.7142, "step": 305760 }, { "epoch": 26.48, "learning_rate": 8.805799699317681e-06, "loss": 0.6518, "step": 305770 }, { "epoch": 26.48, "learning_rate": 8.803631317219844e-06, "loss": 0.6094, "step": 305780 }, { "epoch": 26.48, "learning_rate": 8.801462935122007e-06, "loss": 0.7531, "step": 305790 }, { "epoch": 26.48, "learning_rate": 8.79929455302417e-06, "loss": 0.5991, "step": 305800 }, { "epoch": 26.49, "learning_rate": 8.797126170926331e-06, "loss": 0.3904, "step": 305810 }, { "epoch": 26.49, "learning_rate": 8.794957788828494e-06, "loss": 0.5819, "step": 305820 }, { "epoch": 26.49, "learning_rate": 8.792789406730657e-06, "loss": 0.5292, "step": 305830 }, { "epoch": 26.49, "learning_rate": 8.79062102463282e-06, "loss": 0.7167, "step": 305840 }, { "epoch": 26.49, "learning_rate": 8.788452642534983e-06, "loss": 0.5589, "step": 305850 }, { "epoch": 26.49, "learning_rate": 8.786284260437145e-06, "loss": 0.6647, "step": 305860 }, { "epoch": 26.49, "learning_rate": 8.784115878339308e-06, "loss": 0.568, "step": 305870 }, { "epoch": 26.49, "learning_rate": 8.78194749624147e-06, "loss": 0.4877, "step": 305880 }, { "epoch": 26.49, "learning_rate": 8.779779114143634e-06, "loss": 0.5768, "step": 305890 }, { "epoch": 26.49, "learning_rate": 8.777610732045797e-06, "loss": 0.5246, "step": 305900 }, { "epoch": 26.49, "learning_rate": 8.775442349947958e-06, "loss": 0.4547, "step": 305910 }, { "epoch": 26.49, "learning_rate": 8.77327396785012e-06, "loss": 0.5401, "step": 305920 }, { "epoch": 26.5, "learning_rate": 8.771105585752284e-06, "loss": 0.6317, "step": 305930 }, { "epoch": 26.5, "learning_rate": 8.768937203654447e-06, "loss": 0.6112, "step": 305940 }, { "epoch": 26.5, "learning_rate": 8.766768821556608e-06, "loss": 0.5431, "step": 305950 }, { "epoch": 26.5, "learning_rate": 8.764600439458771e-06, "loss": 0.7046, "step": 305960 }, { "epoch": 26.5, "learning_rate": 8.762432057360932e-06, "loss": 0.6462, "step": 305970 }, { "epoch": 26.5, "learning_rate": 8.760263675263095e-06, "loss": 0.6327, "step": 305980 }, { "epoch": 26.5, "learning_rate": 8.758095293165258e-06, "loss": 0.6134, "step": 305990 }, { "epoch": 26.5, "learning_rate": 8.755926911067421e-06, "loss": 0.6306, "step": 306000 }, { "epoch": 26.5, "learning_rate": 8.753758528969584e-06, "loss": 0.6407, "step": 306010 }, { "epoch": 26.5, "learning_rate": 8.751590146871746e-06, "loss": 0.6286, "step": 306020 }, { "epoch": 26.5, "learning_rate": 8.749421764773909e-06, "loss": 0.5735, "step": 306030 }, { "epoch": 26.51, "learning_rate": 8.747253382676072e-06, "loss": 0.5123, "step": 306040 }, { "epoch": 26.51, "learning_rate": 8.745085000578235e-06, "loss": 0.5153, "step": 306050 }, { "epoch": 26.51, "learning_rate": 8.742916618480396e-06, "loss": 0.5479, "step": 306060 }, { "epoch": 26.51, "learning_rate": 8.740748236382559e-06, "loss": 0.4674, "step": 306070 }, { "epoch": 26.51, "learning_rate": 8.738579854284722e-06, "loss": 0.6662, "step": 306080 }, { "epoch": 26.51, "learning_rate": 8.736411472186885e-06, "loss": 0.5692, "step": 306090 }, { "epoch": 26.51, "learning_rate": 8.734243090089048e-06, "loss": 0.6825, "step": 306100 }, { "epoch": 26.51, "learning_rate": 8.73207470799121e-06, "loss": 0.6843, "step": 306110 }, { "epoch": 26.51, "learning_rate": 8.729906325893372e-06, "loss": 0.5308, "step": 306120 }, { "epoch": 26.51, "learning_rate": 8.727737943795535e-06, "loss": 0.6939, "step": 306130 }, { "epoch": 26.51, "learning_rate": 8.725569561697698e-06, "loss": 0.5966, "step": 306140 }, { "epoch": 26.51, "learning_rate": 8.723401179599861e-06, "loss": 0.59, "step": 306150 }, { "epoch": 26.52, "learning_rate": 8.721232797502023e-06, "loss": 0.538, "step": 306160 }, { "epoch": 26.52, "learning_rate": 8.719064415404186e-06, "loss": 0.5549, "step": 306170 }, { "epoch": 26.52, "learning_rate": 8.716896033306349e-06, "loss": 0.6774, "step": 306180 }, { "epoch": 26.52, "learning_rate": 8.714727651208512e-06, "loss": 0.5755, "step": 306190 }, { "epoch": 26.52, "learning_rate": 8.712559269110673e-06, "loss": 0.6203, "step": 306200 }, { "epoch": 26.52, "learning_rate": 8.710390887012836e-06, "loss": 0.6342, "step": 306210 }, { "epoch": 26.52, "learning_rate": 8.708222504914999e-06, "loss": 0.6513, "step": 306220 }, { "epoch": 26.52, "learning_rate": 8.70605412281716e-06, "loss": 0.5462, "step": 306230 }, { "epoch": 26.52, "learning_rate": 8.703885740719323e-06, "loss": 0.6251, "step": 306240 }, { "epoch": 26.52, "learning_rate": 8.701717358621486e-06, "loss": 0.5638, "step": 306250 }, { "epoch": 26.52, "learning_rate": 8.69954897652365e-06, "loss": 0.6367, "step": 306260 }, { "epoch": 26.52, "learning_rate": 8.69738059442581e-06, "loss": 0.5798, "step": 306270 }, { "epoch": 26.53, "learning_rate": 8.695212212327974e-06, "loss": 0.5247, "step": 306280 }, { "epoch": 26.53, "learning_rate": 8.693043830230137e-06, "loss": 0.5421, "step": 306290 }, { "epoch": 26.53, "learning_rate": 8.6908754481323e-06, "loss": 0.4744, "step": 306300 }, { "epoch": 26.53, "learning_rate": 8.688707066034461e-06, "loss": 0.6428, "step": 306310 }, { "epoch": 26.53, "learning_rate": 8.686538683936624e-06, "loss": 0.6296, "step": 306320 }, { "epoch": 26.53, "learning_rate": 8.684370301838787e-06, "loss": 0.7557, "step": 306330 }, { "epoch": 26.53, "learning_rate": 8.68220191974095e-06, "loss": 0.5569, "step": 306340 }, { "epoch": 26.53, "learning_rate": 8.680033537643113e-06, "loss": 0.6223, "step": 306350 }, { "epoch": 26.53, "learning_rate": 8.677865155545274e-06, "loss": 0.5849, "step": 306360 }, { "epoch": 26.53, "learning_rate": 8.675696773447437e-06, "loss": 0.6551, "step": 306370 }, { "epoch": 26.53, "learning_rate": 8.6735283913496e-06, "loss": 0.5755, "step": 306380 }, { "epoch": 26.54, "learning_rate": 8.671360009251763e-06, "loss": 0.6238, "step": 306390 }, { "epoch": 26.54, "learning_rate": 8.669191627153926e-06, "loss": 0.5612, "step": 306400 }, { "epoch": 26.54, "learning_rate": 8.667023245056088e-06, "loss": 0.5554, "step": 306410 }, { "epoch": 26.54, "learning_rate": 8.66485486295825e-06, "loss": 0.6258, "step": 306420 }, { "epoch": 26.54, "learning_rate": 8.662686480860414e-06, "loss": 0.635, "step": 306430 }, { "epoch": 26.54, "learning_rate": 8.660518098762577e-06, "loss": 0.6427, "step": 306440 }, { "epoch": 26.54, "learning_rate": 8.658349716664738e-06, "loss": 0.7093, "step": 306450 }, { "epoch": 26.54, "learning_rate": 8.656181334566901e-06, "loss": 0.625, "step": 306460 }, { "epoch": 26.54, "learning_rate": 8.654012952469064e-06, "loss": 0.5681, "step": 306470 }, { "epoch": 26.54, "learning_rate": 8.651844570371227e-06, "loss": 0.636, "step": 306480 }, { "epoch": 26.54, "learning_rate": 8.649676188273388e-06, "loss": 0.628, "step": 306490 }, { "epoch": 26.54, "learning_rate": 8.647507806175551e-06, "loss": 0.6102, "step": 306500 }, { "epoch": 26.55, "learning_rate": 8.645339424077714e-06, "loss": 0.5733, "step": 306510 }, { "epoch": 26.55, "learning_rate": 8.643171041979876e-06, "loss": 0.5488, "step": 306520 }, { "epoch": 26.55, "learning_rate": 8.641002659882039e-06, "loss": 0.5261, "step": 306530 }, { "epoch": 26.55, "learning_rate": 8.638834277784202e-06, "loss": 0.548, "step": 306540 }, { "epoch": 26.55, "learning_rate": 8.636665895686365e-06, "loss": 0.5223, "step": 306550 }, { "epoch": 26.55, "learning_rate": 8.634497513588528e-06, "loss": 0.5193, "step": 306560 }, { "epoch": 26.55, "learning_rate": 8.632329131490689e-06, "loss": 0.5644, "step": 306570 }, { "epoch": 26.55, "learning_rate": 8.630160749392852e-06, "loss": 0.693, "step": 306580 }, { "epoch": 26.55, "learning_rate": 8.627992367295015e-06, "loss": 0.5484, "step": 306590 }, { "epoch": 26.55, "learning_rate": 8.625823985197178e-06, "loss": 0.5478, "step": 306600 }, { "epoch": 26.55, "learning_rate": 8.62365560309934e-06, "loss": 0.6319, "step": 306610 }, { "epoch": 26.56, "learning_rate": 8.621487221001502e-06, "loss": 0.6877, "step": 306620 }, { "epoch": 26.56, "learning_rate": 8.619318838903665e-06, "loss": 0.6561, "step": 306630 }, { "epoch": 26.56, "learning_rate": 8.617150456805828e-06, "loss": 0.4751, "step": 306640 }, { "epoch": 26.56, "learning_rate": 8.614982074707991e-06, "loss": 0.6648, "step": 306650 }, { "epoch": 26.56, "learning_rate": 8.612813692610153e-06, "loss": 0.5128, "step": 306660 }, { "epoch": 26.56, "learning_rate": 8.610645310512316e-06, "loss": 0.6415, "step": 306670 }, { "epoch": 26.56, "learning_rate": 8.608476928414479e-06, "loss": 0.6159, "step": 306680 }, { "epoch": 26.56, "learning_rate": 8.606308546316642e-06, "loss": 0.5155, "step": 306690 }, { "epoch": 26.56, "learning_rate": 8.604140164218805e-06, "loss": 0.5991, "step": 306700 }, { "epoch": 26.56, "learning_rate": 8.601971782120966e-06, "loss": 0.5917, "step": 306710 }, { "epoch": 26.56, "learning_rate": 8.599803400023129e-06, "loss": 0.4744, "step": 306720 }, { "epoch": 26.56, "learning_rate": 8.597635017925292e-06, "loss": 0.6299, "step": 306730 }, { "epoch": 26.57, "learning_rate": 8.595466635827453e-06, "loss": 0.6374, "step": 306740 }, { "epoch": 26.57, "learning_rate": 8.593298253729616e-06, "loss": 0.4908, "step": 306750 }, { "epoch": 26.57, "learning_rate": 8.59112987163178e-06, "loss": 0.4571, "step": 306760 }, { "epoch": 26.57, "learning_rate": 8.58896148953394e-06, "loss": 0.5869, "step": 306770 }, { "epoch": 26.57, "learning_rate": 8.586793107436104e-06, "loss": 0.6788, "step": 306780 }, { "epoch": 26.57, "learning_rate": 8.584624725338267e-06, "loss": 0.7163, "step": 306790 }, { "epoch": 26.57, "learning_rate": 8.58245634324043e-06, "loss": 0.6065, "step": 306800 }, { "epoch": 26.57, "learning_rate": 8.580287961142593e-06, "loss": 0.5855, "step": 306810 }, { "epoch": 26.57, "learning_rate": 8.578119579044754e-06, "loss": 0.5893, "step": 306820 }, { "epoch": 26.57, "learning_rate": 8.575951196946917e-06, "loss": 0.6492, "step": 306830 }, { "epoch": 26.57, "learning_rate": 8.57378281484908e-06, "loss": 0.6416, "step": 306840 }, { "epoch": 26.58, "learning_rate": 8.571614432751243e-06, "loss": 0.5869, "step": 306850 }, { "epoch": 26.58, "learning_rate": 8.569446050653404e-06, "loss": 0.572, "step": 306860 }, { "epoch": 26.58, "learning_rate": 8.567277668555567e-06, "loss": 0.4708, "step": 306870 }, { "epoch": 26.58, "learning_rate": 8.56510928645773e-06, "loss": 0.5851, "step": 306880 }, { "epoch": 26.58, "learning_rate": 8.562940904359893e-06, "loss": 0.5754, "step": 306890 }, { "epoch": 26.58, "learning_rate": 8.560772522262056e-06, "loss": 0.6036, "step": 306900 }, { "epoch": 26.58, "learning_rate": 8.558604140164218e-06, "loss": 0.5991, "step": 306910 }, { "epoch": 26.58, "learning_rate": 8.55643575806638e-06, "loss": 0.599, "step": 306920 }, { "epoch": 26.58, "learning_rate": 8.554267375968544e-06, "loss": 0.5368, "step": 306930 }, { "epoch": 26.58, "learning_rate": 8.552098993870707e-06, "loss": 0.5832, "step": 306940 }, { "epoch": 26.58, "learning_rate": 8.54993061177287e-06, "loss": 0.6388, "step": 306950 }, { "epoch": 26.58, "learning_rate": 8.547762229675031e-06, "loss": 0.6778, "step": 306960 }, { "epoch": 26.59, "learning_rate": 8.545593847577194e-06, "loss": 0.6452, "step": 306970 }, { "epoch": 26.59, "learning_rate": 8.543425465479357e-06, "loss": 0.5025, "step": 306980 }, { "epoch": 26.59, "learning_rate": 8.54125708338152e-06, "loss": 0.5727, "step": 306990 }, { "epoch": 26.59, "learning_rate": 8.539088701283681e-06, "loss": 0.6391, "step": 307000 }, { "epoch": 26.59, "learning_rate": 8.536920319185844e-06, "loss": 0.5865, "step": 307010 }, { "epoch": 26.59, "learning_rate": 8.534751937088005e-06, "loss": 0.5821, "step": 307020 }, { "epoch": 26.59, "learning_rate": 8.532583554990168e-06, "loss": 0.6895, "step": 307030 }, { "epoch": 26.59, "learning_rate": 8.530415172892332e-06, "loss": 0.5977, "step": 307040 }, { "epoch": 26.59, "learning_rate": 8.528246790794495e-06, "loss": 0.6065, "step": 307050 }, { "epoch": 26.59, "learning_rate": 8.526078408696658e-06, "loss": 0.5334, "step": 307060 }, { "epoch": 26.59, "learning_rate": 8.523910026598819e-06, "loss": 0.6492, "step": 307070 }, { "epoch": 26.6, "learning_rate": 8.521741644500982e-06, "loss": 0.62, "step": 307080 }, { "epoch": 26.6, "learning_rate": 8.519573262403145e-06, "loss": 0.691, "step": 307090 }, { "epoch": 26.6, "learning_rate": 8.517404880305308e-06, "loss": 0.5698, "step": 307100 }, { "epoch": 26.6, "learning_rate": 8.51523649820747e-06, "loss": 0.5558, "step": 307110 }, { "epoch": 26.6, "learning_rate": 8.513068116109632e-06, "loss": 0.5998, "step": 307120 }, { "epoch": 26.6, "learning_rate": 8.510899734011795e-06, "loss": 0.7053, "step": 307130 }, { "epoch": 26.6, "learning_rate": 8.508731351913958e-06, "loss": 0.5362, "step": 307140 }, { "epoch": 26.6, "learning_rate": 8.506562969816121e-06, "loss": 0.5665, "step": 307150 }, { "epoch": 26.6, "learning_rate": 8.504394587718282e-06, "loss": 0.5863, "step": 307160 }, { "epoch": 26.6, "learning_rate": 8.502226205620445e-06, "loss": 0.5444, "step": 307170 }, { "epoch": 26.6, "learning_rate": 8.500057823522608e-06, "loss": 0.5108, "step": 307180 }, { "epoch": 26.6, "learning_rate": 8.497889441424771e-06, "loss": 0.7106, "step": 307190 }, { "epoch": 26.61, "learning_rate": 8.495721059326934e-06, "loss": 0.5091, "step": 307200 }, { "epoch": 26.61, "learning_rate": 8.493552677229096e-06, "loss": 0.5492, "step": 307210 }, { "epoch": 26.61, "learning_rate": 8.491384295131259e-06, "loss": 0.4984, "step": 307220 }, { "epoch": 26.61, "learning_rate": 8.489215913033422e-06, "loss": 0.7501, "step": 307230 }, { "epoch": 26.61, "learning_rate": 8.487047530935585e-06, "loss": 0.7169, "step": 307240 }, { "epoch": 26.61, "learning_rate": 8.484879148837748e-06, "loss": 0.6169, "step": 307250 }, { "epoch": 26.61, "learning_rate": 8.482710766739909e-06, "loss": 0.5565, "step": 307260 }, { "epoch": 26.61, "learning_rate": 8.48054238464207e-06, "loss": 0.6465, "step": 307270 }, { "epoch": 26.61, "learning_rate": 8.478374002544233e-06, "loss": 0.5619, "step": 307280 }, { "epoch": 26.61, "learning_rate": 8.476205620446396e-06, "loss": 0.6451, "step": 307290 }, { "epoch": 26.61, "learning_rate": 8.47403723834856e-06, "loss": 0.7714, "step": 307300 }, { "epoch": 26.61, "learning_rate": 8.471868856250722e-06, "loss": 0.6816, "step": 307310 }, { "epoch": 26.62, "learning_rate": 8.469700474152884e-06, "loss": 0.6528, "step": 307320 }, { "epoch": 26.62, "learning_rate": 8.467532092055047e-06, "loss": 0.7031, "step": 307330 }, { "epoch": 26.62, "learning_rate": 8.46536370995721e-06, "loss": 0.6514, "step": 307340 }, { "epoch": 26.62, "learning_rate": 8.463195327859373e-06, "loss": 0.5256, "step": 307350 }, { "epoch": 26.62, "learning_rate": 8.461026945761536e-06, "loss": 0.5096, "step": 307360 }, { "epoch": 26.62, "learning_rate": 8.458858563663697e-06, "loss": 0.571, "step": 307370 }, { "epoch": 26.62, "learning_rate": 8.45669018156586e-06, "loss": 0.5721, "step": 307380 }, { "epoch": 26.62, "learning_rate": 8.454521799468023e-06, "loss": 0.6364, "step": 307390 }, { "epoch": 26.62, "learning_rate": 8.452353417370186e-06, "loss": 0.6005, "step": 307400 }, { "epoch": 26.62, "learning_rate": 8.450185035272347e-06, "loss": 0.4593, "step": 307410 }, { "epoch": 26.62, "learning_rate": 8.44801665317451e-06, "loss": 0.7154, "step": 307420 }, { "epoch": 26.63, "learning_rate": 8.445848271076673e-06, "loss": 0.4864, "step": 307430 }, { "epoch": 26.63, "learning_rate": 8.443679888978836e-06, "loss": 0.6184, "step": 307440 }, { "epoch": 26.63, "learning_rate": 8.441511506881e-06, "loss": 0.6388, "step": 307450 }, { "epoch": 26.63, "learning_rate": 8.43934312478316e-06, "loss": 0.6068, "step": 307460 }, { "epoch": 26.63, "learning_rate": 8.437174742685324e-06, "loss": 0.7269, "step": 307470 }, { "epoch": 26.63, "learning_rate": 8.435006360587487e-06, "loss": 0.625, "step": 307480 }, { "epoch": 26.63, "learning_rate": 8.43283797848965e-06, "loss": 0.6408, "step": 307490 }, { "epoch": 26.63, "learning_rate": 8.430669596391813e-06, "loss": 0.5256, "step": 307500 }, { "epoch": 26.63, "learning_rate": 8.428501214293974e-06, "loss": 0.5633, "step": 307510 }, { "epoch": 26.63, "learning_rate": 8.426332832196135e-06, "loss": 0.5148, "step": 307520 }, { "epoch": 26.63, "learning_rate": 8.424164450098298e-06, "loss": 0.6418, "step": 307530 }, { "epoch": 26.63, "learning_rate": 8.421996068000461e-06, "loss": 0.6202, "step": 307540 }, { "epoch": 26.64, "learning_rate": 8.419827685902624e-06, "loss": 0.595, "step": 307550 }, { "epoch": 26.64, "learning_rate": 8.417659303804787e-06, "loss": 0.5889, "step": 307560 }, { "epoch": 26.64, "learning_rate": 8.415490921706949e-06, "loss": 0.6342, "step": 307570 }, { "epoch": 26.64, "learning_rate": 8.413322539609112e-06, "loss": 0.6544, "step": 307580 }, { "epoch": 26.64, "learning_rate": 8.411154157511275e-06, "loss": 0.6022, "step": 307590 }, { "epoch": 26.64, "learning_rate": 8.408985775413438e-06, "loss": 0.5011, "step": 307600 }, { "epoch": 26.64, "learning_rate": 8.4068173933156e-06, "loss": 0.5552, "step": 307610 }, { "epoch": 26.64, "learning_rate": 8.404649011217762e-06, "loss": 0.5717, "step": 307620 }, { "epoch": 26.64, "learning_rate": 8.402480629119925e-06, "loss": 0.5579, "step": 307630 }, { "epoch": 26.64, "learning_rate": 8.400312247022088e-06, "loss": 0.4884, "step": 307640 }, { "epoch": 26.64, "learning_rate": 8.398143864924251e-06, "loss": 0.6095, "step": 307650 }, { "epoch": 26.65, "learning_rate": 8.395975482826412e-06, "loss": 0.5329, "step": 307660 }, { "epoch": 26.65, "learning_rate": 8.393807100728575e-06, "loss": 0.5002, "step": 307670 }, { "epoch": 26.65, "learning_rate": 8.391638718630738e-06, "loss": 0.5609, "step": 307680 }, { "epoch": 26.65, "learning_rate": 8.389470336532901e-06, "loss": 0.5561, "step": 307690 }, { "epoch": 26.65, "learning_rate": 8.387301954435064e-06, "loss": 0.5286, "step": 307700 }, { "epoch": 26.65, "learning_rate": 8.385133572337226e-06, "loss": 0.5947, "step": 307710 }, { "epoch": 26.65, "learning_rate": 8.382965190239389e-06, "loss": 0.6235, "step": 307720 }, { "epoch": 26.65, "learning_rate": 8.380796808141552e-06, "loss": 0.7437, "step": 307730 }, { "epoch": 26.65, "learning_rate": 8.378628426043715e-06, "loss": 0.5854, "step": 307740 }, { "epoch": 26.65, "learning_rate": 8.376460043945878e-06, "loss": 0.489, "step": 307750 }, { "epoch": 26.65, "learning_rate": 8.374291661848039e-06, "loss": 0.6157, "step": 307760 }, { "epoch": 26.65, "learning_rate": 8.372123279750202e-06, "loss": 0.6158, "step": 307770 }, { "epoch": 26.66, "learning_rate": 8.369954897652363e-06, "loss": 0.6594, "step": 307780 }, { "epoch": 26.66, "learning_rate": 8.367786515554526e-06, "loss": 0.724, "step": 307790 }, { "epoch": 26.66, "learning_rate": 8.36561813345669e-06, "loss": 0.547, "step": 307800 }, { "epoch": 26.66, "learning_rate": 8.363449751358852e-06, "loss": 0.5559, "step": 307810 }, { "epoch": 26.66, "learning_rate": 8.361281369261014e-06, "loss": 0.5604, "step": 307820 }, { "epoch": 26.66, "learning_rate": 8.359112987163177e-06, "loss": 0.7661, "step": 307830 }, { "epoch": 26.66, "learning_rate": 8.35694460506534e-06, "loss": 0.6341, "step": 307840 }, { "epoch": 26.66, "learning_rate": 8.354776222967503e-06, "loss": 0.5407, "step": 307850 }, { "epoch": 26.66, "learning_rate": 8.352607840869666e-06, "loss": 0.5588, "step": 307860 }, { "epoch": 26.66, "learning_rate": 8.350439458771827e-06, "loss": 0.5573, "step": 307870 }, { "epoch": 26.66, "learning_rate": 8.34827107667399e-06, "loss": 0.6534, "step": 307880 }, { "epoch": 26.67, "learning_rate": 8.346102694576153e-06, "loss": 0.4266, "step": 307890 }, { "epoch": 26.67, "learning_rate": 8.343934312478316e-06, "loss": 0.6606, "step": 307900 }, { "epoch": 26.67, "learning_rate": 8.341765930380479e-06, "loss": 0.6207, "step": 307910 }, { "epoch": 26.67, "learning_rate": 8.33959754828264e-06, "loss": 0.5958, "step": 307920 }, { "epoch": 26.67, "learning_rate": 8.337429166184803e-06, "loss": 0.6906, "step": 307930 }, { "epoch": 26.67, "learning_rate": 8.335260784086966e-06, "loss": 0.5748, "step": 307940 }, { "epoch": 26.67, "learning_rate": 8.33309240198913e-06, "loss": 0.6722, "step": 307950 }, { "epoch": 26.67, "learning_rate": 8.33092401989129e-06, "loss": 0.4904, "step": 307960 }, { "epoch": 26.67, "learning_rate": 8.328755637793454e-06, "loss": 0.5217, "step": 307970 }, { "epoch": 26.67, "learning_rate": 8.326587255695617e-06, "loss": 0.4829, "step": 307980 }, { "epoch": 26.67, "learning_rate": 8.32441887359778e-06, "loss": 0.6159, "step": 307990 }, { "epoch": 26.67, "learning_rate": 8.322250491499943e-06, "loss": 0.716, "step": 308000 }, { "epoch": 26.68, "learning_rate": 8.320082109402104e-06, "loss": 0.5963, "step": 308010 }, { "epoch": 26.68, "learning_rate": 8.317913727304267e-06, "loss": 0.5687, "step": 308020 }, { "epoch": 26.68, "learning_rate": 8.315745345206428e-06, "loss": 0.6057, "step": 308030 }, { "epoch": 26.68, "learning_rate": 8.313576963108591e-06, "loss": 0.5202, "step": 308040 }, { "epoch": 26.68, "learning_rate": 8.311408581010754e-06, "loss": 0.5443, "step": 308050 }, { "epoch": 26.68, "learning_rate": 8.309240198912917e-06, "loss": 0.6205, "step": 308060 }, { "epoch": 26.68, "learning_rate": 8.307071816815079e-06, "loss": 0.6441, "step": 308070 }, { "epoch": 26.68, "learning_rate": 8.304903434717242e-06, "loss": 0.6732, "step": 308080 }, { "epoch": 26.68, "learning_rate": 8.302735052619405e-06, "loss": 0.6054, "step": 308090 }, { "epoch": 26.68, "learning_rate": 8.300566670521568e-06, "loss": 0.6525, "step": 308100 }, { "epoch": 26.68, "learning_rate": 8.29839828842373e-06, "loss": 0.5988, "step": 308110 }, { "epoch": 26.69, "learning_rate": 8.296229906325892e-06, "loss": 0.5016, "step": 308120 }, { "epoch": 26.69, "learning_rate": 8.294061524228055e-06, "loss": 0.6096, "step": 308130 }, { "epoch": 26.69, "learning_rate": 8.291893142130218e-06, "loss": 0.4832, "step": 308140 }, { "epoch": 26.69, "learning_rate": 8.28972476003238e-06, "loss": 0.6954, "step": 308150 }, { "epoch": 26.69, "learning_rate": 8.287556377934544e-06, "loss": 0.6222, "step": 308160 }, { "epoch": 26.69, "learning_rate": 8.285387995836705e-06, "loss": 0.7181, "step": 308170 }, { "epoch": 26.69, "learning_rate": 8.283219613738868e-06, "loss": 0.6992, "step": 308180 }, { "epoch": 26.69, "learning_rate": 8.281051231641031e-06, "loss": 0.5449, "step": 308190 }, { "epoch": 26.69, "learning_rate": 8.278882849543194e-06, "loss": 0.6198, "step": 308200 }, { "epoch": 26.69, "learning_rate": 8.276714467445355e-06, "loss": 0.6037, "step": 308210 }, { "epoch": 26.69, "learning_rate": 8.274546085347518e-06, "loss": 0.6487, "step": 308220 }, { "epoch": 26.69, "learning_rate": 8.272377703249681e-06, "loss": 0.4752, "step": 308230 }, { "epoch": 26.7, "learning_rate": 8.270209321151844e-06, "loss": 0.5569, "step": 308240 }, { "epoch": 26.7, "learning_rate": 8.268040939054007e-06, "loss": 0.7079, "step": 308250 }, { "epoch": 26.7, "learning_rate": 8.265872556956169e-06, "loss": 0.6913, "step": 308260 }, { "epoch": 26.7, "learning_rate": 8.263704174858332e-06, "loss": 0.5393, "step": 308270 }, { "epoch": 26.7, "learning_rate": 8.261535792760493e-06, "loss": 0.5799, "step": 308280 }, { "epoch": 26.7, "learning_rate": 8.259367410662656e-06, "loss": 0.669, "step": 308290 }, { "epoch": 26.7, "learning_rate": 8.257199028564819e-06, "loss": 0.5946, "step": 308300 }, { "epoch": 26.7, "learning_rate": 8.255030646466982e-06, "loss": 0.6858, "step": 308310 }, { "epoch": 26.7, "learning_rate": 8.252862264369145e-06, "loss": 0.59, "step": 308320 }, { "epoch": 26.7, "learning_rate": 8.250693882271306e-06, "loss": 0.5538, "step": 308330 }, { "epoch": 26.7, "learning_rate": 8.24852550017347e-06, "loss": 0.6222, "step": 308340 }, { "epoch": 26.71, "learning_rate": 8.246357118075632e-06, "loss": 0.5879, "step": 308350 }, { "epoch": 26.71, "learning_rate": 8.244188735977795e-06, "loss": 0.5848, "step": 308360 }, { "epoch": 26.71, "learning_rate": 8.242020353879957e-06, "loss": 0.6626, "step": 308370 }, { "epoch": 26.71, "learning_rate": 8.23985197178212e-06, "loss": 0.7762, "step": 308380 }, { "epoch": 26.71, "learning_rate": 8.237683589684283e-06, "loss": 0.556, "step": 308390 }, { "epoch": 26.71, "learning_rate": 8.235515207586446e-06, "loss": 0.6749, "step": 308400 }, { "epoch": 26.71, "learning_rate": 8.233346825488609e-06, "loss": 0.4481, "step": 308410 }, { "epoch": 26.71, "learning_rate": 8.23117844339077e-06, "loss": 0.5544, "step": 308420 }, { "epoch": 26.71, "learning_rate": 8.229010061292933e-06, "loss": 0.6471, "step": 308430 }, { "epoch": 26.71, "learning_rate": 8.226841679195096e-06, "loss": 0.6056, "step": 308440 }, { "epoch": 26.71, "learning_rate": 8.224673297097259e-06, "loss": 0.6138, "step": 308450 }, { "epoch": 26.71, "learning_rate": 8.222504914999422e-06, "loss": 0.5619, "step": 308460 }, { "epoch": 26.72, "learning_rate": 8.220336532901583e-06, "loss": 0.6182, "step": 308470 }, { "epoch": 26.72, "learning_rate": 8.218168150803746e-06, "loss": 0.5278, "step": 308480 }, { "epoch": 26.72, "learning_rate": 8.21599976870591e-06, "loss": 0.6737, "step": 308490 }, { "epoch": 26.72, "learning_rate": 8.213831386608072e-06, "loss": 0.4483, "step": 308500 }, { "epoch": 26.72, "learning_rate": 8.211663004510234e-06, "loss": 0.4635, "step": 308510 }, { "epoch": 26.72, "learning_rate": 8.209494622412397e-06, "loss": 0.5719, "step": 308520 }, { "epoch": 26.72, "learning_rate": 8.20732624031456e-06, "loss": 0.6824, "step": 308530 }, { "epoch": 26.72, "learning_rate": 8.205157858216721e-06, "loss": 0.6737, "step": 308540 }, { "epoch": 26.72, "learning_rate": 8.202989476118884e-06, "loss": 0.5035, "step": 308550 }, { "epoch": 26.72, "learning_rate": 8.200821094021047e-06, "loss": 0.5475, "step": 308560 }, { "epoch": 26.72, "learning_rate": 8.19865271192321e-06, "loss": 0.5919, "step": 308570 }, { "epoch": 26.72, "learning_rate": 8.196484329825371e-06, "loss": 0.6285, "step": 308580 }, { "epoch": 26.73, "learning_rate": 8.194315947727534e-06, "loss": 0.5564, "step": 308590 }, { "epoch": 26.73, "learning_rate": 8.192147565629697e-06, "loss": 0.6261, "step": 308600 }, { "epoch": 26.73, "learning_rate": 8.18997918353186e-06, "loss": 0.6314, "step": 308610 }, { "epoch": 26.73, "learning_rate": 8.187810801434022e-06, "loss": 0.8207, "step": 308620 }, { "epoch": 26.73, "learning_rate": 8.185642419336185e-06, "loss": 0.6654, "step": 308630 }, { "epoch": 26.73, "learning_rate": 8.183474037238348e-06, "loss": 0.6501, "step": 308640 }, { "epoch": 26.73, "learning_rate": 8.18130565514051e-06, "loss": 0.5008, "step": 308650 }, { "epoch": 26.73, "learning_rate": 8.179137273042674e-06, "loss": 0.5658, "step": 308660 }, { "epoch": 26.73, "learning_rate": 8.176968890944835e-06, "loss": 0.5734, "step": 308670 }, { "epoch": 26.73, "learning_rate": 8.174800508846998e-06, "loss": 0.5858, "step": 308680 }, { "epoch": 26.73, "learning_rate": 8.172632126749161e-06, "loss": 0.8121, "step": 308690 }, { "epoch": 26.74, "learning_rate": 8.170463744651324e-06, "loss": 0.5334, "step": 308700 }, { "epoch": 26.74, "learning_rate": 8.168295362553487e-06, "loss": 0.4706, "step": 308710 }, { "epoch": 26.74, "learning_rate": 8.166126980455648e-06, "loss": 0.6595, "step": 308720 }, { "epoch": 26.74, "learning_rate": 8.163958598357811e-06, "loss": 0.5596, "step": 308730 }, { "epoch": 26.74, "learning_rate": 8.161790216259974e-06, "loss": 0.5212, "step": 308740 }, { "epoch": 26.74, "learning_rate": 8.159621834162137e-06, "loss": 0.6706, "step": 308750 }, { "epoch": 26.74, "learning_rate": 8.157453452064299e-06, "loss": 0.7001, "step": 308760 }, { "epoch": 26.74, "learning_rate": 8.155285069966462e-06, "loss": 0.5255, "step": 308770 }, { "epoch": 26.74, "learning_rate": 8.153116687868625e-06, "loss": 0.6438, "step": 308780 }, { "epoch": 26.74, "learning_rate": 8.150948305770786e-06, "loss": 0.4743, "step": 308790 }, { "epoch": 26.74, "learning_rate": 8.148779923672949e-06, "loss": 0.6196, "step": 308800 }, { "epoch": 26.74, "learning_rate": 8.146611541575112e-06, "loss": 0.5748, "step": 308810 }, { "epoch": 26.75, "learning_rate": 8.144443159477275e-06, "loss": 0.5081, "step": 308820 }, { "epoch": 26.75, "learning_rate": 8.142274777379436e-06, "loss": 0.6268, "step": 308830 }, { "epoch": 26.75, "learning_rate": 8.1401063952816e-06, "loss": 0.6109, "step": 308840 }, { "epoch": 26.75, "learning_rate": 8.137938013183762e-06, "loss": 0.5595, "step": 308850 }, { "epoch": 26.75, "learning_rate": 8.135769631085925e-06, "loss": 0.5982, "step": 308860 }, { "epoch": 26.75, "learning_rate": 8.133601248988087e-06, "loss": 0.6101, "step": 308870 }, { "epoch": 26.75, "learning_rate": 8.13143286689025e-06, "loss": 0.6281, "step": 308880 }, { "epoch": 26.75, "learning_rate": 8.129264484792413e-06, "loss": 0.6678, "step": 308890 }, { "epoch": 26.75, "learning_rate": 8.127096102694576e-06, "loss": 0.6282, "step": 308900 }, { "epoch": 26.75, "learning_rate": 8.124927720596739e-06, "loss": 0.5277, "step": 308910 }, { "epoch": 26.75, "learning_rate": 8.1227593384989e-06, "loss": 0.4268, "step": 308920 }, { "epoch": 26.76, "learning_rate": 8.120590956401063e-06, "loss": 0.5858, "step": 308930 }, { "epoch": 26.76, "learning_rate": 8.118422574303226e-06, "loss": 0.629, "step": 308940 }, { "epoch": 26.76, "learning_rate": 8.116254192205389e-06, "loss": 0.4646, "step": 308950 }, { "epoch": 26.76, "learning_rate": 8.114085810107552e-06, "loss": 0.6144, "step": 308960 }, { "epoch": 26.76, "learning_rate": 8.111917428009713e-06, "loss": 0.6228, "step": 308970 }, { "epoch": 26.76, "learning_rate": 8.109749045911876e-06, "loss": 0.5319, "step": 308980 }, { "epoch": 26.76, "learning_rate": 8.10758066381404e-06, "loss": 0.6814, "step": 308990 }, { "epoch": 26.76, "learning_rate": 8.105412281716202e-06, "loss": 0.5987, "step": 309000 }, { "epoch": 26.76, "learning_rate": 8.103243899618364e-06, "loss": 0.7447, "step": 309010 }, { "epoch": 26.76, "learning_rate": 8.101075517520527e-06, "loss": 0.5713, "step": 309020 }, { "epoch": 26.76, "learning_rate": 8.09890713542269e-06, "loss": 0.521, "step": 309030 }, { "epoch": 26.76, "learning_rate": 8.096738753324853e-06, "loss": 0.5802, "step": 309040 }, { "epoch": 26.77, "learning_rate": 8.094570371227014e-06, "loss": 0.525, "step": 309050 }, { "epoch": 26.77, "learning_rate": 8.092401989129177e-06, "loss": 0.6542, "step": 309060 }, { "epoch": 26.77, "learning_rate": 8.09023360703134e-06, "loss": 0.6129, "step": 309070 }, { "epoch": 26.77, "learning_rate": 8.088065224933501e-06, "loss": 0.5785, "step": 309080 }, { "epoch": 26.77, "learning_rate": 8.085896842835664e-06, "loss": 0.6181, "step": 309090 }, { "epoch": 26.77, "learning_rate": 8.083728460737827e-06, "loss": 0.4918, "step": 309100 }, { "epoch": 26.77, "learning_rate": 8.08156007863999e-06, "loss": 0.6276, "step": 309110 }, { "epoch": 26.77, "learning_rate": 8.079391696542153e-06, "loss": 0.6812, "step": 309120 }, { "epoch": 26.77, "learning_rate": 8.077223314444315e-06, "loss": 0.4361, "step": 309130 }, { "epoch": 26.77, "learning_rate": 8.075054932346478e-06, "loss": 0.6215, "step": 309140 }, { "epoch": 26.77, "learning_rate": 8.07288655024864e-06, "loss": 0.7257, "step": 309150 }, { "epoch": 26.78, "learning_rate": 8.070718168150804e-06, "loss": 0.626, "step": 309160 }, { "epoch": 26.78, "learning_rate": 8.068549786052965e-06, "loss": 0.4934, "step": 309170 }, { "epoch": 26.78, "learning_rate": 8.066381403955128e-06, "loss": 0.5835, "step": 309180 }, { "epoch": 26.78, "learning_rate": 8.064213021857291e-06, "loss": 0.4661, "step": 309190 }, { "epoch": 26.78, "learning_rate": 8.062044639759454e-06, "loss": 0.7078, "step": 309200 }, { "epoch": 26.78, "learning_rate": 8.059876257661617e-06, "loss": 0.6432, "step": 309210 }, { "epoch": 26.78, "learning_rate": 8.057707875563778e-06, "loss": 0.6381, "step": 309220 }, { "epoch": 26.78, "learning_rate": 8.055539493465941e-06, "loss": 0.6335, "step": 309230 }, { "epoch": 26.78, "learning_rate": 8.053371111368104e-06, "loss": 0.5941, "step": 309240 }, { "epoch": 26.78, "learning_rate": 8.051202729270267e-06, "loss": 0.6735, "step": 309250 }, { "epoch": 26.78, "learning_rate": 8.04903434717243e-06, "loss": 0.7028, "step": 309260 }, { "epoch": 26.78, "learning_rate": 8.046865965074591e-06, "loss": 0.4812, "step": 309270 }, { "epoch": 26.79, "learning_rate": 8.044697582976755e-06, "loss": 0.6153, "step": 309280 }, { "epoch": 26.79, "learning_rate": 8.042529200878918e-06, "loss": 0.5497, "step": 309290 }, { "epoch": 26.79, "learning_rate": 8.04036081878108e-06, "loss": 0.7105, "step": 309300 }, { "epoch": 26.79, "learning_rate": 8.038192436683242e-06, "loss": 0.6188, "step": 309310 }, { "epoch": 26.79, "learning_rate": 8.036024054585405e-06, "loss": 0.6044, "step": 309320 }, { "epoch": 26.79, "learning_rate": 8.033855672487566e-06, "loss": 0.5579, "step": 309330 }, { "epoch": 26.79, "learning_rate": 8.031687290389729e-06, "loss": 0.5646, "step": 309340 }, { "epoch": 26.79, "learning_rate": 8.029518908291892e-06, "loss": 0.5751, "step": 309350 }, { "epoch": 26.79, "learning_rate": 8.027350526194055e-06, "loss": 0.5344, "step": 309360 }, { "epoch": 26.79, "learning_rate": 8.025182144096218e-06, "loss": 0.6783, "step": 309370 }, { "epoch": 26.79, "learning_rate": 8.02301376199838e-06, "loss": 0.6434, "step": 309380 }, { "epoch": 26.8, "learning_rate": 8.020845379900542e-06, "loss": 0.5304, "step": 309390 }, { "epoch": 26.8, "learning_rate": 8.018676997802705e-06, "loss": 0.5653, "step": 309400 }, { "epoch": 26.8, "learning_rate": 8.016508615704868e-06, "loss": 0.6271, "step": 309410 }, { "epoch": 26.8, "learning_rate": 8.01434023360703e-06, "loss": 0.5288, "step": 309420 }, { "epoch": 26.8, "learning_rate": 8.012171851509193e-06, "loss": 0.6812, "step": 309430 }, { "epoch": 26.8, "learning_rate": 8.010003469411356e-06, "loss": 0.6024, "step": 309440 }, { "epoch": 26.8, "learning_rate": 8.007835087313519e-06, "loss": 0.5354, "step": 309450 }, { "epoch": 26.8, "learning_rate": 8.005666705215682e-06, "loss": 0.5766, "step": 309460 }, { "epoch": 26.8, "learning_rate": 8.003498323117843e-06, "loss": 0.5512, "step": 309470 }, { "epoch": 26.8, "learning_rate": 8.001329941020006e-06, "loss": 0.4912, "step": 309480 }, { "epoch": 26.8, "learning_rate": 7.999161558922169e-06, "loss": 0.5214, "step": 309490 }, { "epoch": 26.8, "learning_rate": 7.996993176824332e-06, "loss": 0.5667, "step": 309500 }, { "epoch": 26.81, "learning_rate": 7.994824794726495e-06, "loss": 0.6482, "step": 309510 }, { "epoch": 26.81, "learning_rate": 7.992656412628656e-06, "loss": 0.713, "step": 309520 }, { "epoch": 26.81, "learning_rate": 7.99048803053082e-06, "loss": 0.6336, "step": 309530 }, { "epoch": 26.81, "learning_rate": 7.988319648432982e-06, "loss": 0.5133, "step": 309540 }, { "epoch": 26.81, "learning_rate": 7.986151266335145e-06, "loss": 0.5661, "step": 309550 }, { "epoch": 26.81, "learning_rate": 7.983982884237307e-06, "loss": 0.6962, "step": 309560 }, { "epoch": 26.81, "learning_rate": 7.98181450213947e-06, "loss": 0.6273, "step": 309570 }, { "epoch": 26.81, "learning_rate": 7.979646120041631e-06, "loss": 0.5673, "step": 309580 }, { "epoch": 26.81, "learning_rate": 7.977477737943794e-06, "loss": 0.5097, "step": 309590 }, { "epoch": 26.81, "learning_rate": 7.975309355845957e-06, "loss": 0.5763, "step": 309600 }, { "epoch": 26.81, "learning_rate": 7.97314097374812e-06, "loss": 0.5498, "step": 309610 }, { "epoch": 26.82, "learning_rate": 7.970972591650283e-06, "loss": 0.4942, "step": 309620 }, { "epoch": 26.82, "learning_rate": 7.968804209552444e-06, "loss": 0.527, "step": 309630 }, { "epoch": 26.82, "learning_rate": 7.966635827454607e-06, "loss": 0.5978, "step": 309640 }, { "epoch": 26.82, "learning_rate": 7.96446744535677e-06, "loss": 0.6376, "step": 309650 }, { "epoch": 26.82, "learning_rate": 7.962299063258933e-06, "loss": 0.5819, "step": 309660 }, { "epoch": 26.82, "learning_rate": 7.960130681161096e-06, "loss": 0.5614, "step": 309670 }, { "epoch": 26.82, "learning_rate": 7.957962299063258e-06, "loss": 0.608, "step": 309680 }, { "epoch": 26.82, "learning_rate": 7.95579391696542e-06, "loss": 0.6806, "step": 309690 }, { "epoch": 26.82, "learning_rate": 7.953625534867584e-06, "loss": 0.5472, "step": 309700 }, { "epoch": 26.82, "learning_rate": 7.951457152769747e-06, "loss": 0.5252, "step": 309710 }, { "epoch": 26.82, "learning_rate": 7.949288770671908e-06, "loss": 0.6956, "step": 309720 }, { "epoch": 26.82, "learning_rate": 7.947120388574071e-06, "loss": 0.5523, "step": 309730 }, { "epoch": 26.83, "learning_rate": 7.944952006476234e-06, "loss": 0.517, "step": 309740 }, { "epoch": 26.83, "learning_rate": 7.942783624378397e-06, "loss": 0.5121, "step": 309750 }, { "epoch": 26.83, "learning_rate": 7.94061524228056e-06, "loss": 0.5877, "step": 309760 }, { "epoch": 26.83, "learning_rate": 7.938446860182721e-06, "loss": 0.43, "step": 309770 }, { "epoch": 26.83, "learning_rate": 7.936278478084884e-06, "loss": 0.6015, "step": 309780 }, { "epoch": 26.83, "learning_rate": 7.934110095987047e-06, "loss": 0.6452, "step": 309790 }, { "epoch": 26.83, "learning_rate": 7.93194171388921e-06, "loss": 0.737, "step": 309800 }, { "epoch": 26.83, "learning_rate": 7.929773331791373e-06, "loss": 0.5759, "step": 309810 }, { "epoch": 26.83, "learning_rate": 7.927604949693535e-06, "loss": 0.516, "step": 309820 }, { "epoch": 26.83, "learning_rate": 7.925436567595696e-06, "loss": 0.6253, "step": 309830 }, { "epoch": 26.83, "learning_rate": 7.923268185497859e-06, "loss": 0.4951, "step": 309840 }, { "epoch": 26.83, "learning_rate": 7.921099803400022e-06, "loss": 0.5604, "step": 309850 }, { "epoch": 26.84, "learning_rate": 7.918931421302185e-06, "loss": 0.7918, "step": 309860 }, { "epoch": 26.84, "learning_rate": 7.916763039204348e-06, "loss": 0.5892, "step": 309870 }, { "epoch": 26.84, "learning_rate": 7.91459465710651e-06, "loss": 0.7132, "step": 309880 }, { "epoch": 26.84, "learning_rate": 7.912426275008672e-06, "loss": 0.6274, "step": 309890 }, { "epoch": 26.84, "learning_rate": 7.910257892910835e-06, "loss": 0.5771, "step": 309900 }, { "epoch": 26.84, "learning_rate": 7.908089510812998e-06, "loss": 0.5782, "step": 309910 }, { "epoch": 26.84, "learning_rate": 7.905921128715161e-06, "loss": 0.6858, "step": 309920 }, { "epoch": 26.84, "learning_rate": 7.903752746617323e-06, "loss": 0.6254, "step": 309930 }, { "epoch": 26.84, "learning_rate": 7.901584364519486e-06, "loss": 0.5469, "step": 309940 }, { "epoch": 26.84, "learning_rate": 7.899415982421649e-06, "loss": 0.6702, "step": 309950 }, { "epoch": 26.84, "learning_rate": 7.897247600323812e-06, "loss": 0.5354, "step": 309960 }, { "epoch": 26.85, "learning_rate": 7.895079218225973e-06, "loss": 0.472, "step": 309970 }, { "epoch": 26.85, "learning_rate": 7.892910836128136e-06, "loss": 0.6238, "step": 309980 }, { "epoch": 26.85, "learning_rate": 7.890742454030299e-06, "loss": 0.6342, "step": 309990 }, { "epoch": 26.85, "learning_rate": 7.888574071932462e-06, "loss": 0.5909, "step": 310000 }, { "epoch": 26.85, "learning_rate": 7.886405689834625e-06, "loss": 0.5801, "step": 310010 }, { "epoch": 26.85, "learning_rate": 7.884237307736786e-06, "loss": 0.6052, "step": 310020 }, { "epoch": 26.85, "learning_rate": 7.88206892563895e-06, "loss": 0.451, "step": 310030 }, { "epoch": 26.85, "learning_rate": 7.879900543541112e-06, "loss": 0.6857, "step": 310040 }, { "epoch": 26.85, "learning_rate": 7.877732161443275e-06, "loss": 0.584, "step": 310050 }, { "epoch": 26.85, "learning_rate": 7.875563779345438e-06, "loss": 0.7011, "step": 310060 }, { "epoch": 26.85, "learning_rate": 7.8733953972476e-06, "loss": 0.5946, "step": 310070 }, { "epoch": 26.85, "learning_rate": 7.871227015149761e-06, "loss": 0.6475, "step": 310080 }, { "epoch": 26.86, "learning_rate": 7.869058633051924e-06, "loss": 0.6293, "step": 310090 }, { "epoch": 26.86, "learning_rate": 7.866890250954087e-06, "loss": 0.6811, "step": 310100 }, { "epoch": 26.86, "learning_rate": 7.86472186885625e-06, "loss": 0.5107, "step": 310110 }, { "epoch": 26.86, "learning_rate": 7.862553486758413e-06, "loss": 0.5852, "step": 310120 }, { "epoch": 26.86, "learning_rate": 7.860385104660574e-06, "loss": 0.4293, "step": 310130 }, { "epoch": 26.86, "learning_rate": 7.858216722562737e-06, "loss": 0.6426, "step": 310140 }, { "epoch": 26.86, "learning_rate": 7.8560483404649e-06, "loss": 0.6796, "step": 310150 }, { "epoch": 26.86, "learning_rate": 7.853879958367063e-06, "loss": 0.4892, "step": 310160 }, { "epoch": 26.86, "learning_rate": 7.851711576269226e-06, "loss": 0.4837, "step": 310170 }, { "epoch": 26.86, "learning_rate": 7.849543194171388e-06, "loss": 0.5393, "step": 310180 }, { "epoch": 26.86, "learning_rate": 7.84737481207355e-06, "loss": 0.6911, "step": 310190 }, { "epoch": 26.87, "learning_rate": 7.845206429975714e-06, "loss": 0.6183, "step": 310200 }, { "epoch": 26.87, "learning_rate": 7.843038047877877e-06, "loss": 0.5678, "step": 310210 }, { "epoch": 26.87, "learning_rate": 7.840869665780038e-06, "loss": 0.5464, "step": 310220 }, { "epoch": 26.87, "learning_rate": 7.838701283682201e-06, "loss": 0.7394, "step": 310230 }, { "epoch": 26.87, "learning_rate": 7.836532901584364e-06, "loss": 0.503, "step": 310240 }, { "epoch": 26.87, "learning_rate": 7.834364519486527e-06, "loss": 0.5405, "step": 310250 }, { "epoch": 26.87, "learning_rate": 7.83219613738869e-06, "loss": 0.5735, "step": 310260 }, { "epoch": 26.87, "learning_rate": 7.830027755290851e-06, "loss": 0.5599, "step": 310270 }, { "epoch": 26.87, "learning_rate": 7.827859373193014e-06, "loss": 0.5015, "step": 310280 }, { "epoch": 26.87, "learning_rate": 7.825690991095177e-06, "loss": 0.6343, "step": 310290 }, { "epoch": 26.87, "learning_rate": 7.82352260899734e-06, "loss": 0.6813, "step": 310300 }, { "epoch": 26.87, "learning_rate": 7.821354226899503e-06, "loss": 0.5914, "step": 310310 }, { "epoch": 26.88, "learning_rate": 7.819185844801665e-06, "loss": 0.552, "step": 310320 }, { "epoch": 26.88, "learning_rate": 7.817017462703828e-06, "loss": 0.668, "step": 310330 }, { "epoch": 26.88, "learning_rate": 7.814849080605989e-06, "loss": 0.627, "step": 310340 }, { "epoch": 26.88, "learning_rate": 7.812680698508152e-06, "loss": 0.5635, "step": 310350 }, { "epoch": 26.88, "learning_rate": 7.810512316410315e-06, "loss": 0.5654, "step": 310360 }, { "epoch": 26.88, "learning_rate": 7.808343934312478e-06, "loss": 0.672, "step": 310370 }, { "epoch": 26.88, "learning_rate": 7.806175552214639e-06, "loss": 0.5359, "step": 310380 }, { "epoch": 26.88, "learning_rate": 7.804007170116802e-06, "loss": 0.5321, "step": 310390 }, { "epoch": 26.88, "learning_rate": 7.801838788018965e-06, "loss": 0.6376, "step": 310400 }, { "epoch": 26.88, "learning_rate": 7.799670405921128e-06, "loss": 0.7338, "step": 310410 }, { "epoch": 26.88, "learning_rate": 7.797502023823291e-06, "loss": 0.6633, "step": 310420 }, { "epoch": 26.89, "learning_rate": 7.795333641725452e-06, "loss": 0.5426, "step": 310430 }, { "epoch": 26.89, "learning_rate": 7.793165259627615e-06, "loss": 0.5408, "step": 310440 }, { "epoch": 26.89, "learning_rate": 7.790996877529778e-06, "loss": 0.5869, "step": 310450 }, { "epoch": 26.89, "learning_rate": 7.788828495431941e-06, "loss": 0.6653, "step": 310460 }, { "epoch": 26.89, "learning_rate": 7.786660113334104e-06, "loss": 0.607, "step": 310470 }, { "epoch": 26.89, "learning_rate": 7.784491731236266e-06, "loss": 0.5948, "step": 310480 }, { "epoch": 26.89, "learning_rate": 7.782323349138429e-06, "loss": 0.6576, "step": 310490 }, { "epoch": 26.89, "learning_rate": 7.780154967040592e-06, "loss": 0.6357, "step": 310500 }, { "epoch": 26.89, "learning_rate": 7.777986584942755e-06, "loss": 0.4866, "step": 310510 }, { "epoch": 26.89, "learning_rate": 7.775818202844916e-06, "loss": 0.6763, "step": 310520 }, { "epoch": 26.89, "learning_rate": 7.773649820747079e-06, "loss": 0.6832, "step": 310530 }, { "epoch": 26.89, "learning_rate": 7.771481438649242e-06, "loss": 0.4958, "step": 310540 }, { "epoch": 26.9, "learning_rate": 7.769313056551405e-06, "loss": 0.6842, "step": 310550 }, { "epoch": 26.9, "learning_rate": 7.767144674453568e-06, "loss": 0.5982, "step": 310560 }, { "epoch": 26.9, "learning_rate": 7.76497629235573e-06, "loss": 0.5728, "step": 310570 }, { "epoch": 26.9, "learning_rate": 7.762807910257892e-06, "loss": 0.5874, "step": 310580 }, { "epoch": 26.9, "learning_rate": 7.760639528160054e-06, "loss": 0.5856, "step": 310590 }, { "epoch": 26.9, "learning_rate": 7.758471146062217e-06, "loss": 0.5218, "step": 310600 }, { "epoch": 26.9, "learning_rate": 7.75630276396438e-06, "loss": 0.5965, "step": 310610 }, { "epoch": 26.9, "learning_rate": 7.754134381866543e-06, "loss": 0.6306, "step": 310620 }, { "epoch": 26.9, "learning_rate": 7.751965999768704e-06, "loss": 0.7717, "step": 310630 }, { "epoch": 26.9, "learning_rate": 7.749797617670867e-06, "loss": 0.5546, "step": 310640 }, { "epoch": 26.9, "learning_rate": 7.74762923557303e-06, "loss": 0.6488, "step": 310650 }, { "epoch": 26.91, "learning_rate": 7.745460853475193e-06, "loss": 0.6313, "step": 310660 }, { "epoch": 26.91, "learning_rate": 7.743292471377356e-06, "loss": 0.5662, "step": 310670 }, { "epoch": 26.91, "learning_rate": 7.741124089279517e-06, "loss": 0.5603, "step": 310680 }, { "epoch": 26.91, "learning_rate": 7.73895570718168e-06, "loss": 0.6813, "step": 310690 }, { "epoch": 26.91, "learning_rate": 7.736787325083843e-06, "loss": 0.6002, "step": 310700 }, { "epoch": 26.91, "learning_rate": 7.734618942986006e-06, "loss": 0.4952, "step": 310710 }, { "epoch": 26.91, "learning_rate": 7.73245056088817e-06, "loss": 0.5433, "step": 310720 }, { "epoch": 26.91, "learning_rate": 7.73028217879033e-06, "loss": 0.6486, "step": 310730 }, { "epoch": 26.91, "learning_rate": 7.728113796692494e-06, "loss": 0.6163, "step": 310740 }, { "epoch": 26.91, "learning_rate": 7.725945414594657e-06, "loss": 0.5894, "step": 310750 }, { "epoch": 26.91, "learning_rate": 7.72377703249682e-06, "loss": 0.5686, "step": 310760 }, { "epoch": 26.91, "learning_rate": 7.721608650398981e-06, "loss": 0.4502, "step": 310770 }, { "epoch": 26.92, "learning_rate": 7.719440268301144e-06, "loss": 0.6177, "step": 310780 }, { "epoch": 26.92, "learning_rate": 7.717271886203307e-06, "loss": 0.5533, "step": 310790 }, { "epoch": 26.92, "learning_rate": 7.71510350410547e-06, "loss": 0.5261, "step": 310800 }, { "epoch": 26.92, "learning_rate": 7.712935122007633e-06, "loss": 0.5439, "step": 310810 }, { "epoch": 26.92, "learning_rate": 7.710766739909794e-06, "loss": 0.6117, "step": 310820 }, { "epoch": 26.92, "learning_rate": 7.708598357811957e-06, "loss": 0.7809, "step": 310830 }, { "epoch": 26.92, "learning_rate": 7.70642997571412e-06, "loss": 0.4973, "step": 310840 }, { "epoch": 26.92, "learning_rate": 7.704261593616282e-06, "loss": 0.5874, "step": 310850 }, { "epoch": 26.92, "learning_rate": 7.702093211518445e-06, "loss": 0.6385, "step": 310860 }, { "epoch": 26.92, "learning_rate": 7.699924829420608e-06, "loss": 0.7481, "step": 310870 }, { "epoch": 26.92, "learning_rate": 7.69775644732277e-06, "loss": 0.5608, "step": 310880 }, { "epoch": 26.93, "learning_rate": 7.695588065224932e-06, "loss": 0.653, "step": 310890 }, { "epoch": 26.93, "learning_rate": 7.693419683127095e-06, "loss": 0.5709, "step": 310900 }, { "epoch": 26.93, "learning_rate": 7.691251301029258e-06, "loss": 0.6013, "step": 310910 }, { "epoch": 26.93, "learning_rate": 7.689082918931421e-06, "loss": 0.5976, "step": 310920 }, { "epoch": 26.93, "learning_rate": 7.686914536833582e-06, "loss": 0.6153, "step": 310930 }, { "epoch": 26.93, "learning_rate": 7.684746154735745e-06, "loss": 0.5335, "step": 310940 }, { "epoch": 26.93, "learning_rate": 7.682577772637908e-06, "loss": 0.6714, "step": 310950 }, { "epoch": 26.93, "learning_rate": 7.680409390540071e-06, "loss": 0.6553, "step": 310960 }, { "epoch": 26.93, "learning_rate": 7.678241008442234e-06, "loss": 0.6064, "step": 310970 }, { "epoch": 26.93, "learning_rate": 7.676072626344396e-06, "loss": 0.548, "step": 310980 }, { "epoch": 26.93, "learning_rate": 7.673904244246559e-06, "loss": 0.5173, "step": 310990 }, { "epoch": 26.93, "learning_rate": 7.671735862148722e-06, "loss": 0.685, "step": 311000 }, { "epoch": 26.94, "learning_rate": 7.669567480050885e-06, "loss": 0.5833, "step": 311010 }, { "epoch": 26.94, "learning_rate": 7.667399097953048e-06, "loss": 0.4846, "step": 311020 }, { "epoch": 26.94, "learning_rate": 7.665230715855209e-06, "loss": 0.5069, "step": 311030 }, { "epoch": 26.94, "learning_rate": 7.663062333757372e-06, "loss": 0.6118, "step": 311040 }, { "epoch": 26.94, "learning_rate": 7.660893951659535e-06, "loss": 0.5844, "step": 311050 }, { "epoch": 26.94, "learning_rate": 7.658725569561698e-06, "loss": 0.6371, "step": 311060 }, { "epoch": 26.94, "learning_rate": 7.65655718746386e-06, "loss": 0.7008, "step": 311070 }, { "epoch": 26.94, "learning_rate": 7.654388805366022e-06, "loss": 0.6293, "step": 311080 }, { "epoch": 26.94, "learning_rate": 7.652220423268185e-06, "loss": 0.4875, "step": 311090 }, { "epoch": 26.94, "learning_rate": 7.650052041170347e-06, "loss": 0.5656, "step": 311100 }, { "epoch": 26.94, "learning_rate": 7.64788365907251e-06, "loss": 0.5994, "step": 311110 }, { "epoch": 26.94, "learning_rate": 7.645715276974673e-06, "loss": 0.6067, "step": 311120 }, { "epoch": 26.95, "learning_rate": 7.643546894876836e-06, "loss": 0.699, "step": 311130 }, { "epoch": 26.95, "learning_rate": 7.641378512778997e-06, "loss": 0.4956, "step": 311140 }, { "epoch": 26.95, "learning_rate": 7.63921013068116e-06, "loss": 0.7522, "step": 311150 }, { "epoch": 26.95, "learning_rate": 7.637041748583323e-06, "loss": 0.5093, "step": 311160 }, { "epoch": 26.95, "learning_rate": 7.634873366485486e-06, "loss": 0.6245, "step": 311170 }, { "epoch": 26.95, "learning_rate": 7.632704984387647e-06, "loss": 0.4885, "step": 311180 }, { "epoch": 26.95, "learning_rate": 7.63053660228981e-06, "loss": 0.532, "step": 311190 }, { "epoch": 26.95, "learning_rate": 7.628368220191973e-06, "loss": 0.7601, "step": 311200 }, { "epoch": 26.95, "learning_rate": 7.626199838094136e-06, "loss": 0.5545, "step": 311210 }, { "epoch": 26.95, "learning_rate": 7.624031455996298e-06, "loss": 0.5226, "step": 311220 }, { "epoch": 26.95, "learning_rate": 7.621863073898461e-06, "loss": 0.5866, "step": 311230 }, { "epoch": 26.96, "learning_rate": 7.6196946918006236e-06, "loss": 0.5815, "step": 311240 }, { "epoch": 26.96, "learning_rate": 7.6175263097027866e-06, "loss": 0.652, "step": 311250 }, { "epoch": 26.96, "learning_rate": 7.61535792760495e-06, "loss": 0.5909, "step": 311260 }, { "epoch": 26.96, "learning_rate": 7.613189545507112e-06, "loss": 0.6492, "step": 311270 }, { "epoch": 26.96, "learning_rate": 7.611021163409275e-06, "loss": 0.6039, "step": 311280 }, { "epoch": 26.96, "learning_rate": 7.608852781311437e-06, "loss": 0.5963, "step": 311290 }, { "epoch": 26.96, "learning_rate": 7.6066843992136e-06, "loss": 0.646, "step": 311300 }, { "epoch": 26.96, "learning_rate": 7.604516017115762e-06, "loss": 0.6927, "step": 311310 }, { "epoch": 26.96, "learning_rate": 7.602347635017925e-06, "loss": 0.6664, "step": 311320 }, { "epoch": 26.96, "learning_rate": 7.600179252920088e-06, "loss": 0.562, "step": 311330 }, { "epoch": 26.96, "learning_rate": 7.59801087082225e-06, "loss": 0.5952, "step": 311340 }, { "epoch": 26.96, "learning_rate": 7.595842488724413e-06, "loss": 0.6985, "step": 311350 }, { "epoch": 26.97, "learning_rate": 7.5936741066265745e-06, "loss": 0.5958, "step": 311360 }, { "epoch": 26.97, "learning_rate": 7.5915057245287375e-06, "loss": 0.5315, "step": 311370 }, { "epoch": 26.97, "learning_rate": 7.5893373424309e-06, "loss": 0.5744, "step": 311380 }, { "epoch": 26.97, "learning_rate": 7.587168960333063e-06, "loss": 0.6516, "step": 311390 }, { "epoch": 26.97, "learning_rate": 7.585000578235225e-06, "loss": 0.6302, "step": 311400 }, { "epoch": 26.97, "learning_rate": 7.582832196137388e-06, "loss": 0.546, "step": 311410 }, { "epoch": 26.97, "learning_rate": 7.58066381403955e-06, "loss": 0.5752, "step": 311420 }, { "epoch": 26.97, "learning_rate": 7.578495431941713e-06, "loss": 0.5911, "step": 311430 }, { "epoch": 26.97, "learning_rate": 7.576327049843876e-06, "loss": 0.6887, "step": 311440 }, { "epoch": 26.97, "learning_rate": 7.574158667746038e-06, "loss": 0.7864, "step": 311450 }, { "epoch": 26.97, "learning_rate": 7.571990285648201e-06, "loss": 0.7928, "step": 311460 }, { "epoch": 26.98, "learning_rate": 7.569821903550363e-06, "loss": 0.5625, "step": 311470 }, { "epoch": 26.98, "learning_rate": 7.567653521452526e-06, "loss": 0.7342, "step": 311480 }, { "epoch": 26.98, "learning_rate": 7.5654851393546885e-06, "loss": 0.6759, "step": 311490 }, { "epoch": 26.98, "learning_rate": 7.5633167572568515e-06, "loss": 0.573, "step": 311500 }, { "epoch": 26.98, "learning_rate": 7.5611483751590145e-06, "loss": 0.7044, "step": 311510 }, { "epoch": 26.98, "learning_rate": 7.558979993061177e-06, "loss": 0.5771, "step": 311520 }, { "epoch": 26.98, "learning_rate": 7.55681161096334e-06, "loss": 0.5001, "step": 311530 }, { "epoch": 26.98, "learning_rate": 7.554643228865502e-06, "loss": 0.6248, "step": 311540 }, { "epoch": 26.98, "learning_rate": 7.552474846767665e-06, "loss": 0.6133, "step": 311550 }, { "epoch": 26.98, "learning_rate": 7.550306464669827e-06, "loss": 0.7832, "step": 311560 }, { "epoch": 26.98, "learning_rate": 7.54813808257199e-06, "loss": 0.4737, "step": 311570 }, { "epoch": 26.98, "learning_rate": 7.545969700474153e-06, "loss": 0.4835, "step": 311580 }, { "epoch": 26.99, "learning_rate": 7.543801318376315e-06, "loss": 0.5772, "step": 311590 }, { "epoch": 26.99, "learning_rate": 7.541632936278478e-06, "loss": 0.5004, "step": 311600 }, { "epoch": 26.99, "learning_rate": 7.53946455418064e-06, "loss": 0.554, "step": 311610 }, { "epoch": 26.99, "learning_rate": 7.5372961720828025e-06, "loss": 0.7041, "step": 311620 }, { "epoch": 26.99, "learning_rate": 7.535127789984965e-06, "loss": 0.6982, "step": 311630 }, { "epoch": 26.99, "learning_rate": 7.532959407887128e-06, "loss": 0.5172, "step": 311640 }, { "epoch": 26.99, "learning_rate": 7.53079102578929e-06, "loss": 0.5708, "step": 311650 }, { "epoch": 26.99, "learning_rate": 7.528622643691453e-06, "loss": 0.5114, "step": 311660 }, { "epoch": 26.99, "learning_rate": 7.526454261593615e-06, "loss": 0.4114, "step": 311670 }, { "epoch": 26.99, "learning_rate": 7.524285879495778e-06, "loss": 0.7425, "step": 311680 }, { "epoch": 26.99, "learning_rate": 7.522117497397941e-06, "loss": 0.4612, "step": 311690 }, { "epoch": 27.0, "learning_rate": 7.519949115300103e-06, "loss": 0.597, "step": 311700 }, { "epoch": 27.0, "learning_rate": 7.517780733202266e-06, "loss": 0.6113, "step": 311710 }, { "epoch": 27.0, "learning_rate": 7.515612351104428e-06, "loss": 0.6557, "step": 311720 }, { "epoch": 27.0, "learning_rate": 7.513443969006591e-06, "loss": 0.537, "step": 311730 }, { "epoch": 27.0, "learning_rate": 7.511275586908753e-06, "loss": 0.6031, "step": 311740 }, { "epoch": 27.0, "learning_rate": 7.509107204810916e-06, "loss": 0.5752, "step": 311750 }, { "epoch": 27.0, "eval_cer": 0.5336144804973751, "eval_em": 0.06549767921609077, "eval_f1": 0.06549767921609077, "eval_loss": 1.5778391361236572, "eval_runtime": 241.7515, "eval_samples_per_second": 8.021, "eval_steps_per_second": 1.005, "eval_wer": 0.9345023207839093, "step": 311755 }, { "epoch": 27.0, "learning_rate": 7.5069388227130794e-06, "loss": 0.708, "step": 311760 }, { "epoch": 27.0, "learning_rate": 7.504770440615242e-06, "loss": 0.6726, "step": 311770 }, { "epoch": 27.0, "learning_rate": 7.502602058517405e-06, "loss": 0.6148, "step": 311780 }, { "epoch": 27.0, "learning_rate": 7.500433676419567e-06, "loss": 0.728, "step": 311790 }, { "epoch": 27.0, "learning_rate": 7.49826529432173e-06, "loss": 0.5688, "step": 311800 }, { "epoch": 27.0, "learning_rate": 7.496096912223892e-06, "loss": 0.5124, "step": 311810 }, { "epoch": 27.01, "learning_rate": 7.493928530126055e-06, "loss": 0.5916, "step": 311820 }, { "epoch": 27.01, "learning_rate": 7.491760148028218e-06, "loss": 0.5696, "step": 311830 }, { "epoch": 27.01, "learning_rate": 7.48959176593038e-06, "loss": 0.5264, "step": 311840 }, { "epoch": 27.01, "learning_rate": 7.487423383832543e-06, "loss": 0.6179, "step": 311850 }, { "epoch": 27.01, "learning_rate": 7.485255001734705e-06, "loss": 0.5242, "step": 311860 }, { "epoch": 27.01, "learning_rate": 7.483086619636867e-06, "loss": 0.5605, "step": 311870 }, { "epoch": 27.01, "learning_rate": 7.4809182375390295e-06, "loss": 0.5424, "step": 311880 }, { "epoch": 27.01, "learning_rate": 7.4787498554411925e-06, "loss": 0.5265, "step": 311890 }, { "epoch": 27.01, "learning_rate": 7.476581473343355e-06, "loss": 0.6233, "step": 311900 }, { "epoch": 27.01, "learning_rate": 7.474413091245518e-06, "loss": 0.5639, "step": 311910 }, { "epoch": 27.01, "learning_rate": 7.472244709147681e-06, "loss": 0.643, "step": 311920 }, { "epoch": 27.02, "learning_rate": 7.470076327049843e-06, "loss": 0.4324, "step": 311930 }, { "epoch": 27.02, "learning_rate": 7.467907944952006e-06, "loss": 0.649, "step": 311940 }, { "epoch": 27.02, "learning_rate": 7.465739562854168e-06, "loss": 0.5378, "step": 311950 }, { "epoch": 27.02, "learning_rate": 7.463571180756331e-06, "loss": 0.5574, "step": 311960 }, { "epoch": 27.02, "learning_rate": 7.461402798658493e-06, "loss": 0.4938, "step": 311970 }, { "epoch": 27.02, "learning_rate": 7.459234416560656e-06, "loss": 0.5732, "step": 311980 }, { "epoch": 27.02, "learning_rate": 7.457066034462819e-06, "loss": 0.6967, "step": 311990 }, { "epoch": 27.02, "learning_rate": 7.454897652364981e-06, "loss": 0.5009, "step": 312000 }, { "epoch": 27.02, "learning_rate": 7.452729270267144e-06, "loss": 0.5692, "step": 312010 }, { "epoch": 27.02, "learning_rate": 7.4505608881693065e-06, "loss": 0.7214, "step": 312020 }, { "epoch": 27.02, "learning_rate": 7.4483925060714695e-06, "loss": 0.5978, "step": 312030 }, { "epoch": 27.02, "learning_rate": 7.446224123973632e-06, "loss": 0.4443, "step": 312040 }, { "epoch": 27.03, "learning_rate": 7.444055741875795e-06, "loss": 0.4889, "step": 312050 }, { "epoch": 27.03, "learning_rate": 7.441887359777958e-06, "loss": 0.5226, "step": 312060 }, { "epoch": 27.03, "learning_rate": 7.43971897768012e-06, "loss": 0.7066, "step": 312070 }, { "epoch": 27.03, "learning_rate": 7.437550595582283e-06, "loss": 0.5272, "step": 312080 }, { "epoch": 27.03, "learning_rate": 7.435382213484445e-06, "loss": 0.5543, "step": 312090 }, { "epoch": 27.03, "learning_rate": 7.433213831386608e-06, "loss": 0.6153, "step": 312100 }, { "epoch": 27.03, "learning_rate": 7.43104544928877e-06, "loss": 0.4672, "step": 312110 }, { "epoch": 27.03, "learning_rate": 7.428877067190933e-06, "loss": 0.6304, "step": 312120 }, { "epoch": 27.03, "learning_rate": 7.4267086850930945e-06, "loss": 0.5528, "step": 312130 }, { "epoch": 27.03, "learning_rate": 7.4245403029952575e-06, "loss": 0.5892, "step": 312140 }, { "epoch": 27.03, "learning_rate": 7.42237192089742e-06, "loss": 0.6273, "step": 312150 }, { "epoch": 27.04, "learning_rate": 7.420203538799583e-06, "loss": 0.6451, "step": 312160 }, { "epoch": 27.04, "learning_rate": 7.418035156701746e-06, "loss": 0.5295, "step": 312170 }, { "epoch": 27.04, "learning_rate": 7.415866774603908e-06, "loss": 0.6764, "step": 312180 }, { "epoch": 27.04, "learning_rate": 7.413698392506071e-06, "loss": 0.5109, "step": 312190 }, { "epoch": 27.04, "learning_rate": 7.411530010408233e-06, "loss": 0.5859, "step": 312200 }, { "epoch": 27.04, "learning_rate": 7.409361628310396e-06, "loss": 0.6173, "step": 312210 }, { "epoch": 27.04, "learning_rate": 7.407193246212558e-06, "loss": 0.526, "step": 312220 }, { "epoch": 27.04, "learning_rate": 7.405024864114721e-06, "loss": 0.4377, "step": 312230 }, { "epoch": 27.04, "learning_rate": 7.402856482016884e-06, "loss": 0.4463, "step": 312240 }, { "epoch": 27.04, "learning_rate": 7.400688099919046e-06, "loss": 0.5485, "step": 312250 }, { "epoch": 27.04, "learning_rate": 7.398519717821209e-06, "loss": 0.4571, "step": 312260 }, { "epoch": 27.04, "learning_rate": 7.3963513357233714e-06, "loss": 0.6047, "step": 312270 }, { "epoch": 27.05, "learning_rate": 7.3941829536255344e-06, "loss": 0.6176, "step": 312280 }, { "epoch": 27.05, "learning_rate": 7.392014571527697e-06, "loss": 0.5017, "step": 312290 }, { "epoch": 27.05, "learning_rate": 7.38984618942986e-06, "loss": 0.561, "step": 312300 }, { "epoch": 27.05, "learning_rate": 7.387677807332023e-06, "loss": 0.7007, "step": 312310 }, { "epoch": 27.05, "learning_rate": 7.385509425234185e-06, "loss": 0.6685, "step": 312320 }, { "epoch": 27.05, "learning_rate": 7.383341043136348e-06, "loss": 0.5684, "step": 312330 }, { "epoch": 27.05, "learning_rate": 7.38117266103851e-06, "loss": 0.5994, "step": 312340 }, { "epoch": 27.05, "learning_rate": 7.379004278940673e-06, "loss": 0.6203, "step": 312350 }, { "epoch": 27.05, "learning_rate": 7.376835896842835e-06, "loss": 0.6128, "step": 312360 }, { "epoch": 27.05, "learning_rate": 7.374667514744998e-06, "loss": 0.4971, "step": 312370 }, { "epoch": 27.05, "learning_rate": 7.372499132647161e-06, "loss": 0.5904, "step": 312380 }, { "epoch": 27.05, "learning_rate": 7.370330750549322e-06, "loss": 0.5533, "step": 312390 }, { "epoch": 27.06, "learning_rate": 7.368162368451485e-06, "loss": 0.5953, "step": 312400 }, { "epoch": 27.06, "learning_rate": 7.3659939863536475e-06, "loss": 0.6137, "step": 312410 }, { "epoch": 27.06, "learning_rate": 7.3638256042558106e-06, "loss": 0.565, "step": 312420 }, { "epoch": 27.06, "learning_rate": 7.361657222157973e-06, "loss": 0.5986, "step": 312430 }, { "epoch": 27.06, "learning_rate": 7.359488840060136e-06, "loss": 0.553, "step": 312440 }, { "epoch": 27.06, "learning_rate": 7.357320457962298e-06, "loss": 0.6444, "step": 312450 }, { "epoch": 27.06, "learning_rate": 7.355152075864461e-06, "loss": 0.5266, "step": 312460 }, { "epoch": 27.06, "learning_rate": 7.352983693766624e-06, "loss": 0.6233, "step": 312470 }, { "epoch": 27.06, "learning_rate": 7.350815311668786e-06, "loss": 0.5818, "step": 312480 }, { "epoch": 27.06, "learning_rate": 7.348646929570949e-06, "loss": 0.6707, "step": 312490 }, { "epoch": 27.06, "learning_rate": 7.346478547473111e-06, "loss": 0.5765, "step": 312500 }, { "epoch": 27.07, "learning_rate": 7.344310165375274e-06, "loss": 0.5322, "step": 312510 }, { "epoch": 27.07, "learning_rate": 7.342141783277436e-06, "loss": 0.5334, "step": 312520 }, { "epoch": 27.07, "learning_rate": 7.339973401179599e-06, "loss": 0.6177, "step": 312530 }, { "epoch": 27.07, "learning_rate": 7.337805019081762e-06, "loss": 0.6265, "step": 312540 }, { "epoch": 27.07, "learning_rate": 7.3356366369839245e-06, "loss": 0.536, "step": 312550 }, { "epoch": 27.07, "learning_rate": 7.3334682548860875e-06, "loss": 0.6812, "step": 312560 }, { "epoch": 27.07, "learning_rate": 7.33129987278825e-06, "loss": 0.6334, "step": 312570 }, { "epoch": 27.07, "learning_rate": 7.329131490690413e-06, "loss": 0.6625, "step": 312580 }, { "epoch": 27.07, "learning_rate": 7.326963108592575e-06, "loss": 0.5365, "step": 312590 }, { "epoch": 27.07, "learning_rate": 7.324794726494738e-06, "loss": 0.6754, "step": 312600 }, { "epoch": 27.07, "learning_rate": 7.322626344396901e-06, "loss": 0.5637, "step": 312610 }, { "epoch": 27.07, "learning_rate": 7.320457962299063e-06, "loss": 0.4078, "step": 312620 }, { "epoch": 27.08, "learning_rate": 7.318289580201226e-06, "loss": 0.531, "step": 312630 }, { "epoch": 27.08, "learning_rate": 7.316121198103387e-06, "loss": 0.5147, "step": 312640 }, { "epoch": 27.08, "learning_rate": 7.31395281600555e-06, "loss": 0.5148, "step": 312650 }, { "epoch": 27.08, "learning_rate": 7.3117844339077125e-06, "loss": 0.5987, "step": 312660 }, { "epoch": 27.08, "learning_rate": 7.3096160518098755e-06, "loss": 0.6333, "step": 312670 }, { "epoch": 27.08, "learning_rate": 7.307447669712038e-06, "loss": 0.5921, "step": 312680 }, { "epoch": 27.08, "learning_rate": 7.305279287614201e-06, "loss": 0.5851, "step": 312690 }, { "epoch": 27.08, "learning_rate": 7.303110905516363e-06, "loss": 0.5042, "step": 312700 }, { "epoch": 27.08, "learning_rate": 7.300942523418526e-06, "loss": 0.521, "step": 312710 }, { "epoch": 27.08, "learning_rate": 7.298774141320689e-06, "loss": 0.4486, "step": 312720 }, { "epoch": 27.08, "learning_rate": 7.296605759222851e-06, "loss": 0.5231, "step": 312730 }, { "epoch": 27.09, "learning_rate": 7.294437377125014e-06, "loss": 0.626, "step": 312740 }, { "epoch": 27.09, "learning_rate": 7.292268995027176e-06, "loss": 0.5086, "step": 312750 }, { "epoch": 27.09, "learning_rate": 7.290100612929339e-06, "loss": 0.5896, "step": 312760 }, { "epoch": 27.09, "learning_rate": 7.287932230831501e-06, "loss": 0.4741, "step": 312770 }, { "epoch": 27.09, "learning_rate": 7.285763848733664e-06, "loss": 0.59, "step": 312780 }, { "epoch": 27.09, "learning_rate": 7.283595466635827e-06, "loss": 0.5557, "step": 312790 }, { "epoch": 27.09, "learning_rate": 7.2814270845379894e-06, "loss": 0.6971, "step": 312800 }, { "epoch": 27.09, "learning_rate": 7.2792587024401524e-06, "loss": 0.6992, "step": 312810 }, { "epoch": 27.09, "learning_rate": 7.277090320342315e-06, "loss": 0.5759, "step": 312820 }, { "epoch": 27.09, "learning_rate": 7.274921938244478e-06, "loss": 0.6028, "step": 312830 }, { "epoch": 27.09, "learning_rate": 7.27275355614664e-06, "loss": 0.5919, "step": 312840 }, { "epoch": 27.09, "learning_rate": 7.270585174048803e-06, "loss": 0.5227, "step": 312850 }, { "epoch": 27.1, "learning_rate": 7.268416791950966e-06, "loss": 0.561, "step": 312860 }, { "epoch": 27.1, "learning_rate": 7.266248409853128e-06, "loss": 0.4812, "step": 312870 }, { "epoch": 27.1, "learning_rate": 7.264080027755291e-06, "loss": 0.6754, "step": 312880 }, { "epoch": 27.1, "learning_rate": 7.261911645657453e-06, "loss": 0.6458, "step": 312890 }, { "epoch": 27.1, "learning_rate": 7.259743263559615e-06, "loss": 0.703, "step": 312900 }, { "epoch": 27.1, "learning_rate": 7.257574881461777e-06, "loss": 0.5796, "step": 312910 }, { "epoch": 27.1, "learning_rate": 7.25540649936394e-06, "loss": 0.5763, "step": 312920 }, { "epoch": 27.1, "learning_rate": 7.2532381172661026e-06, "loss": 0.5452, "step": 312930 }, { "epoch": 27.1, "learning_rate": 7.2510697351682656e-06, "loss": 0.6188, "step": 312940 }, { "epoch": 27.1, "learning_rate": 7.248901353070428e-06, "loss": 0.6959, "step": 312950 }, { "epoch": 27.1, "learning_rate": 7.246732970972591e-06, "loss": 0.4894, "step": 312960 }, { "epoch": 27.11, "learning_rate": 7.244564588874754e-06, "loss": 0.6193, "step": 312970 }, { "epoch": 27.11, "learning_rate": 7.242396206776916e-06, "loss": 0.5921, "step": 312980 }, { "epoch": 27.11, "learning_rate": 7.240227824679079e-06, "loss": 0.5978, "step": 312990 }, { "epoch": 27.11, "learning_rate": 7.238059442581241e-06, "loss": 0.5546, "step": 313000 }, { "epoch": 27.11, "learning_rate": 7.235891060483404e-06, "loss": 0.5463, "step": 313010 }, { "epoch": 27.11, "learning_rate": 7.233722678385566e-06, "loss": 0.4949, "step": 313020 }, { "epoch": 27.11, "learning_rate": 7.231554296287729e-06, "loss": 0.4785, "step": 313030 }, { "epoch": 27.11, "learning_rate": 7.229385914189892e-06, "loss": 0.5579, "step": 313040 }, { "epoch": 27.11, "learning_rate": 7.227217532092054e-06, "loss": 0.6213, "step": 313050 }, { "epoch": 27.11, "learning_rate": 7.225049149994217e-06, "loss": 0.5995, "step": 313060 }, { "epoch": 27.11, "learning_rate": 7.2228807678963795e-06, "loss": 0.5014, "step": 313070 }, { "epoch": 27.11, "learning_rate": 7.2207123857985425e-06, "loss": 0.6104, "step": 313080 }, { "epoch": 27.12, "learning_rate": 7.218544003700705e-06, "loss": 0.7368, "step": 313090 }, { "epoch": 27.12, "learning_rate": 7.216375621602868e-06, "loss": 0.6231, "step": 313100 }, { "epoch": 27.12, "learning_rate": 7.214207239505031e-06, "loss": 0.5721, "step": 313110 }, { "epoch": 27.12, "learning_rate": 7.212038857407193e-06, "loss": 0.5482, "step": 313120 }, { "epoch": 27.12, "learning_rate": 7.209870475309356e-06, "loss": 0.5817, "step": 313130 }, { "epoch": 27.12, "learning_rate": 7.207702093211518e-06, "loss": 0.4815, "step": 313140 }, { "epoch": 27.12, "learning_rate": 7.205533711113681e-06, "loss": 0.586, "step": 313150 }, { "epoch": 27.12, "learning_rate": 7.203365329015842e-06, "loss": 0.5048, "step": 313160 }, { "epoch": 27.12, "learning_rate": 7.201196946918005e-06, "loss": 0.4964, "step": 313170 }, { "epoch": 27.12, "learning_rate": 7.1990285648201675e-06, "loss": 0.6044, "step": 313180 }, { "epoch": 27.12, "learning_rate": 7.1968601827223305e-06, "loss": 0.6486, "step": 313190 }, { "epoch": 27.13, "learning_rate": 7.1946918006244935e-06, "loss": 0.5717, "step": 313200 }, { "epoch": 27.13, "learning_rate": 7.192523418526656e-06, "loss": 0.5599, "step": 313210 }, { "epoch": 27.13, "learning_rate": 7.190355036428819e-06, "loss": 0.5694, "step": 313220 }, { "epoch": 27.13, "learning_rate": 7.188186654330981e-06, "loss": 0.5531, "step": 313230 }, { "epoch": 27.13, "learning_rate": 7.186018272233144e-06, "loss": 0.5428, "step": 313240 }, { "epoch": 27.13, "learning_rate": 7.183849890135306e-06, "loss": 0.6522, "step": 313250 }, { "epoch": 27.13, "learning_rate": 7.181681508037469e-06, "loss": 0.6644, "step": 313260 }, { "epoch": 27.13, "learning_rate": 7.179513125939632e-06, "loss": 0.5513, "step": 313270 }, { "epoch": 27.13, "learning_rate": 7.177344743841794e-06, "loss": 0.6726, "step": 313280 }, { "epoch": 27.13, "learning_rate": 7.175176361743957e-06, "loss": 0.5742, "step": 313290 }, { "epoch": 27.13, "learning_rate": 7.173007979646119e-06, "loss": 0.6584, "step": 313300 }, { "epoch": 27.13, "learning_rate": 7.170839597548282e-06, "loss": 0.5867, "step": 313310 }, { "epoch": 27.14, "learning_rate": 7.1686712154504444e-06, "loss": 0.5862, "step": 313320 }, { "epoch": 27.14, "learning_rate": 7.1665028333526075e-06, "loss": 0.5643, "step": 313330 }, { "epoch": 27.14, "learning_rate": 7.1643344512547705e-06, "loss": 0.5734, "step": 313340 }, { "epoch": 27.14, "learning_rate": 7.162166069156933e-06, "loss": 0.5893, "step": 313350 }, { "epoch": 27.14, "learning_rate": 7.159997687059096e-06, "loss": 0.5828, "step": 313360 }, { "epoch": 27.14, "learning_rate": 7.157829304961258e-06, "loss": 0.6246, "step": 313370 }, { "epoch": 27.14, "learning_rate": 7.155660922863421e-06, "loss": 0.7682, "step": 313380 }, { "epoch": 27.14, "learning_rate": 7.153492540765583e-06, "loss": 0.6247, "step": 313390 }, { "epoch": 27.14, "learning_rate": 7.151324158667746e-06, "loss": 0.5152, "step": 313400 }, { "epoch": 27.14, "learning_rate": 7.149155776569907e-06, "loss": 0.6229, "step": 313410 }, { "epoch": 27.14, "learning_rate": 7.14698739447207e-06, "loss": 0.4731, "step": 313420 }, { "epoch": 27.15, "learning_rate": 7.144819012374232e-06, "loss": 0.581, "step": 313430 }, { "epoch": 27.15, "learning_rate": 7.142650630276395e-06, "loss": 0.6493, "step": 313440 }, { "epoch": 27.15, "learning_rate": 7.140482248178558e-06, "loss": 0.5959, "step": 313450 }, { "epoch": 27.15, "learning_rate": 7.1383138660807206e-06, "loss": 0.5593, "step": 313460 }, { "epoch": 27.15, "learning_rate": 7.1361454839828836e-06, "loss": 0.5346, "step": 313470 }, { "epoch": 27.15, "learning_rate": 7.133977101885046e-06, "loss": 0.5348, "step": 313480 }, { "epoch": 27.15, "learning_rate": 7.131808719787209e-06, "loss": 0.5325, "step": 313490 }, { "epoch": 27.15, "learning_rate": 7.129640337689371e-06, "loss": 0.6222, "step": 313500 }, { "epoch": 27.15, "learning_rate": 7.127471955591534e-06, "loss": 0.5743, "step": 313510 }, { "epoch": 27.15, "learning_rate": 7.125303573493697e-06, "loss": 0.5732, "step": 313520 }, { "epoch": 27.15, "learning_rate": 7.123135191395859e-06, "loss": 0.6158, "step": 313530 }, { "epoch": 27.15, "learning_rate": 7.120966809298022e-06, "loss": 0.5447, "step": 313540 }, { "epoch": 27.16, "learning_rate": 7.118798427200184e-06, "loss": 0.4977, "step": 313550 }, { "epoch": 27.16, "learning_rate": 7.116630045102347e-06, "loss": 0.6419, "step": 313560 }, { "epoch": 27.16, "learning_rate": 7.114461663004509e-06, "loss": 0.4819, "step": 313570 }, { "epoch": 27.16, "learning_rate": 7.112293280906672e-06, "loss": 0.5846, "step": 313580 }, { "epoch": 27.16, "learning_rate": 7.110124898808835e-06, "loss": 0.7063, "step": 313590 }, { "epoch": 27.16, "learning_rate": 7.1079565167109975e-06, "loss": 0.5014, "step": 313600 }, { "epoch": 27.16, "learning_rate": 7.1057881346131605e-06, "loss": 0.5723, "step": 313610 }, { "epoch": 27.16, "learning_rate": 7.103619752515323e-06, "loss": 0.5075, "step": 313620 }, { "epoch": 27.16, "learning_rate": 7.101451370417486e-06, "loss": 0.507, "step": 313630 }, { "epoch": 27.16, "learning_rate": 7.099282988319648e-06, "loss": 0.5825, "step": 313640 }, { "epoch": 27.16, "learning_rate": 7.097114606221811e-06, "loss": 0.6404, "step": 313650 }, { "epoch": 27.16, "learning_rate": 7.094946224123974e-06, "loss": 0.5776, "step": 313660 }, { "epoch": 27.17, "learning_rate": 7.092777842026135e-06, "loss": 0.6709, "step": 313670 }, { "epoch": 27.17, "learning_rate": 7.090609459928298e-06, "loss": 0.4858, "step": 313680 }, { "epoch": 27.17, "learning_rate": 7.08844107783046e-06, "loss": 0.6289, "step": 313690 }, { "epoch": 27.17, "learning_rate": 7.086272695732623e-06, "loss": 0.6446, "step": 313700 }, { "epoch": 27.17, "learning_rate": 7.0841043136347855e-06, "loss": 0.6159, "step": 313710 }, { "epoch": 27.17, "learning_rate": 7.0819359315369485e-06, "loss": 0.5474, "step": 313720 }, { "epoch": 27.17, "learning_rate": 7.079767549439111e-06, "loss": 0.6396, "step": 313730 }, { "epoch": 27.17, "learning_rate": 7.077599167341274e-06, "loss": 0.6206, "step": 313740 }, { "epoch": 27.17, "learning_rate": 7.075430785243437e-06, "loss": 0.5612, "step": 313750 }, { "epoch": 27.17, "learning_rate": 7.073262403145599e-06, "loss": 0.5625, "step": 313760 }, { "epoch": 27.17, "learning_rate": 7.071094021047762e-06, "loss": 0.4781, "step": 313770 }, { "epoch": 27.18, "learning_rate": 7.068925638949924e-06, "loss": 0.5562, "step": 313780 }, { "epoch": 27.18, "learning_rate": 7.066757256852087e-06, "loss": 0.5963, "step": 313790 }, { "epoch": 27.18, "learning_rate": 7.064588874754249e-06, "loss": 0.5462, "step": 313800 }, { "epoch": 27.18, "learning_rate": 7.062420492656412e-06, "loss": 0.7264, "step": 313810 }, { "epoch": 27.18, "learning_rate": 7.060252110558575e-06, "loss": 0.5382, "step": 313820 }, { "epoch": 27.18, "learning_rate": 7.058083728460737e-06, "loss": 0.6867, "step": 313830 }, { "epoch": 27.18, "learning_rate": 7.0559153463629e-06, "loss": 0.5528, "step": 313840 }, { "epoch": 27.18, "learning_rate": 7.0537469642650625e-06, "loss": 0.6107, "step": 313850 }, { "epoch": 27.18, "learning_rate": 7.0515785821672255e-06, "loss": 0.5747, "step": 313860 }, { "epoch": 27.18, "learning_rate": 7.049410200069388e-06, "loss": 0.6255, "step": 313870 }, { "epoch": 27.18, "learning_rate": 7.047241817971551e-06, "loss": 0.6515, "step": 313880 }, { "epoch": 27.18, "learning_rate": 7.045073435873714e-06, "loss": 0.587, "step": 313890 }, { "epoch": 27.19, "learning_rate": 7.042905053775876e-06, "loss": 0.4817, "step": 313900 }, { "epoch": 27.19, "learning_rate": 7.040736671678039e-06, "loss": 0.6225, "step": 313910 }, { "epoch": 27.19, "learning_rate": 7.038568289580201e-06, "loss": 0.7459, "step": 313920 }, { "epoch": 27.19, "learning_rate": 7.036399907482363e-06, "loss": 0.617, "step": 313930 }, { "epoch": 27.19, "learning_rate": 7.034231525384525e-06, "loss": 0.4782, "step": 313940 }, { "epoch": 27.19, "learning_rate": 7.032063143286688e-06, "loss": 0.5657, "step": 313950 }, { "epoch": 27.19, "learning_rate": 7.02989476118885e-06, "loss": 0.7087, "step": 313960 }, { "epoch": 27.19, "learning_rate": 7.027726379091013e-06, "loss": 0.4242, "step": 313970 }, { "epoch": 27.19, "learning_rate": 7.0255579969931756e-06, "loss": 0.6215, "step": 313980 }, { "epoch": 27.19, "learning_rate": 7.023389614895339e-06, "loss": 0.5415, "step": 313990 }, { "epoch": 27.19, "learning_rate": 7.021221232797502e-06, "loss": 0.6407, "step": 314000 }, { "epoch": 27.2, "learning_rate": 7.019052850699664e-06, "loss": 0.6172, "step": 314010 }, { "epoch": 27.2, "learning_rate": 7.016884468601827e-06, "loss": 0.7345, "step": 314020 }, { "epoch": 27.2, "learning_rate": 7.014716086503989e-06, "loss": 0.5097, "step": 314030 }, { "epoch": 27.2, "learning_rate": 7.012547704406152e-06, "loss": 0.6043, "step": 314040 }, { "epoch": 27.2, "learning_rate": 7.010379322308314e-06, "loss": 0.5095, "step": 314050 }, { "epoch": 27.2, "learning_rate": 7.008210940210477e-06, "loss": 0.5091, "step": 314060 }, { "epoch": 27.2, "learning_rate": 7.00604255811264e-06, "loss": 0.6846, "step": 314070 }, { "epoch": 27.2, "learning_rate": 7.003874176014802e-06, "loss": 0.5423, "step": 314080 }, { "epoch": 27.2, "learning_rate": 7.001705793916965e-06, "loss": 0.5548, "step": 314090 }, { "epoch": 27.2, "learning_rate": 6.999537411819127e-06, "loss": 0.5693, "step": 314100 }, { "epoch": 27.2, "learning_rate": 6.99736902972129e-06, "loss": 0.6241, "step": 314110 }, { "epoch": 27.2, "learning_rate": 6.9952006476234525e-06, "loss": 0.6747, "step": 314120 }, { "epoch": 27.21, "learning_rate": 6.9930322655256155e-06, "loss": 0.4859, "step": 314130 }, { "epoch": 27.21, "learning_rate": 6.9908638834277785e-06, "loss": 0.4266, "step": 314140 }, { "epoch": 27.21, "learning_rate": 6.988695501329941e-06, "loss": 0.6624, "step": 314150 }, { "epoch": 27.21, "learning_rate": 6.986527119232104e-06, "loss": 0.6274, "step": 314160 }, { "epoch": 27.21, "learning_rate": 6.984358737134266e-06, "loss": 0.497, "step": 314170 }, { "epoch": 27.21, "learning_rate": 6.982190355036428e-06, "loss": 0.4825, "step": 314180 }, { "epoch": 27.21, "learning_rate": 6.98002197293859e-06, "loss": 0.5365, "step": 314190 }, { "epoch": 27.21, "learning_rate": 6.977853590840753e-06, "loss": 0.6845, "step": 314200 }, { "epoch": 27.21, "learning_rate": 6.975685208742915e-06, "loss": 0.5616, "step": 314210 }, { "epoch": 27.21, "learning_rate": 6.973516826645078e-06, "loss": 0.5895, "step": 314220 }, { "epoch": 27.21, "learning_rate": 6.9713484445472405e-06, "loss": 0.6879, "step": 314230 }, { "epoch": 27.22, "learning_rate": 6.9691800624494035e-06, "loss": 0.5891, "step": 314240 }, { "epoch": 27.22, "learning_rate": 6.9670116803515665e-06, "loss": 0.6608, "step": 314250 }, { "epoch": 27.22, "learning_rate": 6.964843298253729e-06, "loss": 0.625, "step": 314260 }, { "epoch": 27.22, "learning_rate": 6.962674916155892e-06, "loss": 0.5758, "step": 314270 }, { "epoch": 27.22, "learning_rate": 6.960506534058054e-06, "loss": 0.538, "step": 314280 }, { "epoch": 27.22, "learning_rate": 6.958338151960217e-06, "loss": 0.5096, "step": 314290 }, { "epoch": 27.22, "learning_rate": 6.956169769862379e-06, "loss": 0.503, "step": 314300 }, { "epoch": 27.22, "learning_rate": 6.954001387764542e-06, "loss": 0.5551, "step": 314310 }, { "epoch": 27.22, "learning_rate": 6.951833005666705e-06, "loss": 0.6524, "step": 314320 }, { "epoch": 27.22, "learning_rate": 6.949664623568867e-06, "loss": 0.4937, "step": 314330 }, { "epoch": 27.22, "learning_rate": 6.94749624147103e-06, "loss": 0.5486, "step": 314340 }, { "epoch": 27.22, "learning_rate": 6.945327859373192e-06, "loss": 0.7478, "step": 314350 }, { "epoch": 27.23, "learning_rate": 6.943159477275355e-06, "loss": 0.4983, "step": 314360 }, { "epoch": 27.23, "learning_rate": 6.9409910951775175e-06, "loss": 0.5942, "step": 314370 }, { "epoch": 27.23, "learning_rate": 6.9388227130796805e-06, "loss": 0.5775, "step": 314380 }, { "epoch": 27.23, "learning_rate": 6.9366543309818435e-06, "loss": 0.6389, "step": 314390 }, { "epoch": 27.23, "learning_rate": 6.934485948884006e-06, "loss": 0.5902, "step": 314400 }, { "epoch": 27.23, "learning_rate": 6.932317566786169e-06, "loss": 0.512, "step": 314410 }, { "epoch": 27.23, "learning_rate": 6.930149184688331e-06, "loss": 0.5874, "step": 314420 }, { "epoch": 27.23, "learning_rate": 6.927980802590494e-06, "loss": 0.4856, "step": 314430 }, { "epoch": 27.23, "learning_rate": 6.925812420492655e-06, "loss": 0.7367, "step": 314440 }, { "epoch": 27.23, "learning_rate": 6.923644038394818e-06, "loss": 0.5313, "step": 314450 }, { "epoch": 27.23, "learning_rate": 6.92147565629698e-06, "loss": 0.5438, "step": 314460 }, { "epoch": 27.24, "learning_rate": 6.919307274199143e-06, "loss": 0.7105, "step": 314470 }, { "epoch": 27.24, "learning_rate": 6.917138892101306e-06, "loss": 0.6235, "step": 314480 }, { "epoch": 27.24, "learning_rate": 6.9149705100034684e-06, "loss": 0.6027, "step": 314490 }, { "epoch": 27.24, "learning_rate": 6.9128021279056314e-06, "loss": 0.543, "step": 314500 }, { "epoch": 27.24, "learning_rate": 6.910633745807794e-06, "loss": 0.5563, "step": 314510 }, { "epoch": 27.24, "learning_rate": 6.908465363709957e-06, "loss": 0.589, "step": 314520 }, { "epoch": 27.24, "learning_rate": 6.906296981612119e-06, "loss": 0.471, "step": 314530 }, { "epoch": 27.24, "learning_rate": 6.904128599514282e-06, "loss": 0.7022, "step": 314540 }, { "epoch": 27.24, "learning_rate": 6.901960217416445e-06, "loss": 0.5219, "step": 314550 }, { "epoch": 27.24, "learning_rate": 6.899791835318607e-06, "loss": 0.631, "step": 314560 }, { "epoch": 27.24, "learning_rate": 6.89762345322077e-06, "loss": 0.6926, "step": 314570 }, { "epoch": 27.24, "learning_rate": 6.895455071122932e-06, "loss": 0.5334, "step": 314580 }, { "epoch": 27.25, "learning_rate": 6.893286689025095e-06, "loss": 0.6162, "step": 314590 }, { "epoch": 27.25, "learning_rate": 6.891118306927257e-06, "loss": 0.591, "step": 314600 }, { "epoch": 27.25, "learning_rate": 6.88894992482942e-06, "loss": 0.745, "step": 314610 }, { "epoch": 27.25, "learning_rate": 6.886781542731583e-06, "loss": 0.543, "step": 314620 }, { "epoch": 27.25, "learning_rate": 6.884613160633745e-06, "loss": 0.4949, "step": 314630 }, { "epoch": 27.25, "learning_rate": 6.882444778535908e-06, "loss": 0.5407, "step": 314640 }, { "epoch": 27.25, "learning_rate": 6.8802763964380706e-06, "loss": 0.6206, "step": 314650 }, { "epoch": 27.25, "learning_rate": 6.8781080143402336e-06, "loss": 0.5826, "step": 314660 }, { "epoch": 27.25, "learning_rate": 6.875939632242396e-06, "loss": 0.5644, "step": 314670 }, { "epoch": 27.25, "learning_rate": 6.873771250144559e-06, "loss": 0.5742, "step": 314680 }, { "epoch": 27.25, "learning_rate": 6.871602868046722e-06, "loss": 0.6072, "step": 314690 }, { "epoch": 27.26, "learning_rate": 6.869434485948883e-06, "loss": 0.6642, "step": 314700 }, { "epoch": 27.26, "learning_rate": 6.867266103851045e-06, "loss": 0.5496, "step": 314710 }, { "epoch": 27.26, "learning_rate": 6.865097721753208e-06, "loss": 0.7637, "step": 314720 }, { "epoch": 27.26, "learning_rate": 6.862929339655371e-06, "loss": 0.5899, "step": 314730 }, { "epoch": 27.26, "learning_rate": 6.860760957557533e-06, "loss": 0.6119, "step": 314740 }, { "epoch": 27.26, "learning_rate": 6.858592575459696e-06, "loss": 0.4182, "step": 314750 }, { "epoch": 27.26, "learning_rate": 6.8564241933618585e-06, "loss": 0.6352, "step": 314760 }, { "epoch": 27.26, "learning_rate": 6.8542558112640215e-06, "loss": 0.5252, "step": 314770 }, { "epoch": 27.26, "learning_rate": 6.852087429166184e-06, "loss": 0.5933, "step": 314780 }, { "epoch": 27.26, "learning_rate": 6.849919047068347e-06, "loss": 0.5162, "step": 314790 }, { "epoch": 27.26, "learning_rate": 6.84775066497051e-06, "loss": 0.5446, "step": 314800 }, { "epoch": 27.26, "learning_rate": 6.845582282872672e-06, "loss": 0.5044, "step": 314810 }, { "epoch": 27.27, "learning_rate": 6.843413900774835e-06, "loss": 0.5666, "step": 314820 }, { "epoch": 27.27, "learning_rate": 6.841245518676997e-06, "loss": 0.6531, "step": 314830 }, { "epoch": 27.27, "learning_rate": 6.83907713657916e-06, "loss": 0.5957, "step": 314840 }, { "epoch": 27.27, "learning_rate": 6.836908754481322e-06, "loss": 0.6425, "step": 314850 }, { "epoch": 27.27, "learning_rate": 6.834740372383485e-06, "loss": 0.604, "step": 314860 }, { "epoch": 27.27, "learning_rate": 6.832571990285648e-06, "loss": 0.6344, "step": 314870 }, { "epoch": 27.27, "learning_rate": 6.83040360818781e-06, "loss": 0.5266, "step": 314880 }, { "epoch": 27.27, "learning_rate": 6.828235226089973e-06, "loss": 0.5805, "step": 314890 }, { "epoch": 27.27, "learning_rate": 6.8260668439921355e-06, "loss": 0.4633, "step": 314900 }, { "epoch": 27.27, "learning_rate": 6.8238984618942985e-06, "loss": 0.6586, "step": 314910 }, { "epoch": 27.27, "learning_rate": 6.821730079796461e-06, "loss": 0.5654, "step": 314920 }, { "epoch": 27.27, "learning_rate": 6.819561697698624e-06, "loss": 0.5959, "step": 314930 }, { "epoch": 27.28, "learning_rate": 6.817393315600787e-06, "loss": 0.6147, "step": 314940 }, { "epoch": 27.28, "learning_rate": 6.815224933502948e-06, "loss": 0.5566, "step": 314950 }, { "epoch": 27.28, "learning_rate": 6.81305655140511e-06, "loss": 0.599, "step": 314960 }, { "epoch": 27.28, "learning_rate": 6.810888169307273e-06, "loss": 0.4679, "step": 314970 }, { "epoch": 27.28, "learning_rate": 6.808719787209436e-06, "loss": 0.5042, "step": 314980 }, { "epoch": 27.28, "learning_rate": 6.806551405111598e-06, "loss": 0.5901, "step": 314990 }, { "epoch": 27.28, "learning_rate": 6.804383023013761e-06, "loss": 0.502, "step": 315000 }, { "epoch": 27.28, "learning_rate": 6.8022146409159234e-06, "loss": 0.5409, "step": 315010 }, { "epoch": 27.28, "learning_rate": 6.8000462588180864e-06, "loss": 0.6264, "step": 315020 }, { "epoch": 27.28, "learning_rate": 6.797877876720249e-06, "loss": 0.7144, "step": 315030 }, { "epoch": 27.28, "learning_rate": 6.795709494622412e-06, "loss": 0.6022, "step": 315040 }, { "epoch": 27.29, "learning_rate": 6.793541112524575e-06, "loss": 0.6343, "step": 315050 }, { "epoch": 27.29, "learning_rate": 6.791372730426737e-06, "loss": 0.5346, "step": 315060 }, { "epoch": 27.29, "learning_rate": 6.7892043483289e-06, "loss": 0.5108, "step": 315070 }, { "epoch": 27.29, "learning_rate": 6.787035966231062e-06, "loss": 0.5609, "step": 315080 }, { "epoch": 27.29, "learning_rate": 6.784867584133225e-06, "loss": 0.5757, "step": 315090 }, { "epoch": 27.29, "learning_rate": 6.782699202035387e-06, "loss": 0.5622, "step": 315100 }, { "epoch": 27.29, "learning_rate": 6.78053081993755e-06, "loss": 0.5401, "step": 315110 }, { "epoch": 27.29, "learning_rate": 6.778362437839713e-06, "loss": 0.5072, "step": 315120 }, { "epoch": 27.29, "learning_rate": 6.776194055741875e-06, "loss": 0.5178, "step": 315130 }, { "epoch": 27.29, "learning_rate": 6.774025673644038e-06, "loss": 0.7362, "step": 315140 }, { "epoch": 27.29, "learning_rate": 6.7718572915462e-06, "loss": 0.5389, "step": 315150 }, { "epoch": 27.29, "learning_rate": 6.769688909448363e-06, "loss": 0.5835, "step": 315160 }, { "epoch": 27.3, "learning_rate": 6.7675205273505256e-06, "loss": 0.5434, "step": 315170 }, { "epoch": 27.3, "learning_rate": 6.7653521452526886e-06, "loss": 0.659, "step": 315180 }, { "epoch": 27.3, "learning_rate": 6.7631837631548516e-06, "loss": 0.573, "step": 315190 }, { "epoch": 27.3, "learning_rate": 6.761015381057014e-06, "loss": 0.6128, "step": 315200 }, { "epoch": 27.3, "learning_rate": 6.758846998959176e-06, "loss": 0.5523, "step": 315210 }, { "epoch": 27.3, "learning_rate": 6.756678616861338e-06, "loss": 0.4033, "step": 315220 }, { "epoch": 27.3, "learning_rate": 6.754510234763501e-06, "loss": 0.4535, "step": 315230 }, { "epoch": 27.3, "learning_rate": 6.752341852665663e-06, "loss": 0.5725, "step": 315240 }, { "epoch": 27.3, "learning_rate": 6.750173470567826e-06, "loss": 0.5698, "step": 315250 }, { "epoch": 27.3, "learning_rate": 6.748005088469988e-06, "loss": 0.4569, "step": 315260 }, { "epoch": 27.3, "learning_rate": 6.745836706372151e-06, "loss": 0.687, "step": 315270 }, { "epoch": 27.31, "learning_rate": 6.743668324274314e-06, "loss": 0.5838, "step": 315280 }, { "epoch": 27.31, "learning_rate": 6.7414999421764765e-06, "loss": 0.6433, "step": 315290 }, { "epoch": 27.31, "learning_rate": 6.7393315600786395e-06, "loss": 0.6187, "step": 315300 }, { "epoch": 27.31, "learning_rate": 6.737163177980802e-06, "loss": 0.5326, "step": 315310 }, { "epoch": 27.31, "learning_rate": 6.734994795882965e-06, "loss": 0.6018, "step": 315320 }, { "epoch": 27.31, "learning_rate": 6.732826413785127e-06, "loss": 0.5852, "step": 315330 }, { "epoch": 27.31, "learning_rate": 6.73065803168729e-06, "loss": 0.6663, "step": 315340 }, { "epoch": 27.31, "learning_rate": 6.728489649589453e-06, "loss": 0.504, "step": 315350 }, { "epoch": 27.31, "learning_rate": 6.726321267491615e-06, "loss": 0.5877, "step": 315360 }, { "epoch": 27.31, "learning_rate": 6.724152885393778e-06, "loss": 0.5209, "step": 315370 }, { "epoch": 27.31, "learning_rate": 6.72198450329594e-06, "loss": 0.5817, "step": 315380 }, { "epoch": 27.31, "learning_rate": 6.719816121198103e-06, "loss": 0.6473, "step": 315390 }, { "epoch": 27.32, "learning_rate": 6.717647739100265e-06, "loss": 0.5355, "step": 315400 }, { "epoch": 27.32, "learning_rate": 6.715479357002428e-06, "loss": 0.6273, "step": 315410 }, { "epoch": 27.32, "learning_rate": 6.713310974904591e-06, "loss": 0.6154, "step": 315420 }, { "epoch": 27.32, "learning_rate": 6.7111425928067535e-06, "loss": 0.4859, "step": 315430 }, { "epoch": 27.32, "learning_rate": 6.7089742107089165e-06, "loss": 0.5889, "step": 315440 }, { "epoch": 27.32, "learning_rate": 6.706805828611079e-06, "loss": 0.6749, "step": 315450 }, { "epoch": 27.32, "learning_rate": 6.704637446513242e-06, "loss": 0.6308, "step": 315460 }, { "epoch": 27.32, "learning_rate": 6.702469064415403e-06, "loss": 0.5733, "step": 315470 }, { "epoch": 27.32, "learning_rate": 6.700300682317566e-06, "loss": 0.5514, "step": 315480 }, { "epoch": 27.32, "learning_rate": 6.698132300219728e-06, "loss": 0.5751, "step": 315490 }, { "epoch": 27.32, "learning_rate": 6.695963918121891e-06, "loss": 0.5763, "step": 315500 }, { "epoch": 27.33, "learning_rate": 6.693795536024053e-06, "loss": 0.464, "step": 315510 }, { "epoch": 27.33, "learning_rate": 6.691627153926216e-06, "loss": 0.5105, "step": 315520 }, { "epoch": 27.33, "learning_rate": 6.689458771828379e-06, "loss": 0.455, "step": 315530 }, { "epoch": 27.33, "learning_rate": 6.6872903897305414e-06, "loss": 0.7001, "step": 315540 }, { "epoch": 27.33, "learning_rate": 6.6851220076327044e-06, "loss": 0.5194, "step": 315550 }, { "epoch": 27.33, "learning_rate": 6.682953625534867e-06, "loss": 0.66, "step": 315560 }, { "epoch": 27.33, "learning_rate": 6.68078524343703e-06, "loss": 0.5021, "step": 315570 }, { "epoch": 27.33, "learning_rate": 6.678616861339192e-06, "loss": 0.6147, "step": 315580 }, { "epoch": 27.33, "learning_rate": 6.676448479241355e-06, "loss": 0.6032, "step": 315590 }, { "epoch": 27.33, "learning_rate": 6.674280097143518e-06, "loss": 0.6349, "step": 315600 }, { "epoch": 27.33, "learning_rate": 6.67211171504568e-06, "loss": 0.5025, "step": 315610 }, { "epoch": 27.33, "learning_rate": 6.669943332947843e-06, "loss": 0.6432, "step": 315620 }, { "epoch": 27.34, "learning_rate": 6.667774950850005e-06, "loss": 0.6526, "step": 315630 }, { "epoch": 27.34, "learning_rate": 6.665606568752168e-06, "loss": 0.5732, "step": 315640 }, { "epoch": 27.34, "learning_rate": 6.66343818665433e-06, "loss": 0.5695, "step": 315650 }, { "epoch": 27.34, "learning_rate": 6.661269804556493e-06, "loss": 0.6251, "step": 315660 }, { "epoch": 27.34, "learning_rate": 6.659101422458656e-06, "loss": 0.5043, "step": 315670 }, { "epoch": 27.34, "learning_rate": 6.656933040360818e-06, "loss": 0.6108, "step": 315680 }, { "epoch": 27.34, "learning_rate": 6.654764658262981e-06, "loss": 0.648, "step": 315690 }, { "epoch": 27.34, "learning_rate": 6.6525962761651436e-06, "loss": 0.5397, "step": 315700 }, { "epoch": 27.34, "learning_rate": 6.6504278940673066e-06, "loss": 0.568, "step": 315710 }, { "epoch": 27.34, "learning_rate": 6.648259511969468e-06, "loss": 0.6153, "step": 315720 }, { "epoch": 27.34, "learning_rate": 6.646091129871631e-06, "loss": 0.5066, "step": 315730 }, { "epoch": 27.35, "learning_rate": 6.643922747773793e-06, "loss": 0.5914, "step": 315740 }, { "epoch": 27.35, "learning_rate": 6.641754365675956e-06, "loss": 0.5648, "step": 315750 }, { "epoch": 27.35, "learning_rate": 6.639585983578119e-06, "loss": 0.6169, "step": 315760 }, { "epoch": 27.35, "learning_rate": 6.637417601480281e-06, "loss": 0.6563, "step": 315770 }, { "epoch": 27.35, "learning_rate": 6.635249219382444e-06, "loss": 0.5158, "step": 315780 }, { "epoch": 27.35, "learning_rate": 6.633080837284606e-06, "loss": 0.5603, "step": 315790 }, { "epoch": 27.35, "learning_rate": 6.630912455186769e-06, "loss": 0.6079, "step": 315800 }, { "epoch": 27.35, "learning_rate": 6.6287440730889315e-06, "loss": 0.5426, "step": 315810 }, { "epoch": 27.35, "learning_rate": 6.6265756909910945e-06, "loss": 0.6238, "step": 315820 }, { "epoch": 27.35, "learning_rate": 6.6244073088932575e-06, "loss": 0.5736, "step": 315830 }, { "epoch": 27.35, "learning_rate": 6.62223892679542e-06, "loss": 0.4961, "step": 315840 }, { "epoch": 27.35, "learning_rate": 6.620070544697583e-06, "loss": 0.6645, "step": 315850 }, { "epoch": 27.36, "learning_rate": 6.617902162599745e-06, "loss": 0.6297, "step": 315860 }, { "epoch": 27.36, "learning_rate": 6.615733780501908e-06, "loss": 0.5674, "step": 315870 }, { "epoch": 27.36, "learning_rate": 6.61356539840407e-06, "loss": 0.5665, "step": 315880 }, { "epoch": 27.36, "learning_rate": 6.611397016306233e-06, "loss": 0.5566, "step": 315890 }, { "epoch": 27.36, "learning_rate": 6.609228634208396e-06, "loss": 0.7123, "step": 315900 }, { "epoch": 27.36, "learning_rate": 6.607060252110558e-06, "loss": 0.5912, "step": 315910 }, { "epoch": 27.36, "learning_rate": 6.604891870012721e-06, "loss": 0.6042, "step": 315920 }, { "epoch": 27.36, "learning_rate": 6.602723487914883e-06, "loss": 0.6495, "step": 315930 }, { "epoch": 27.36, "learning_rate": 6.600555105817046e-06, "loss": 0.5283, "step": 315940 }, { "epoch": 27.36, "learning_rate": 6.5983867237192085e-06, "loss": 0.5035, "step": 315950 }, { "epoch": 27.36, "learning_rate": 6.5962183416213715e-06, "loss": 0.5637, "step": 315960 }, { "epoch": 27.37, "learning_rate": 6.5940499595235345e-06, "loss": 0.4855, "step": 315970 }, { "epoch": 27.37, "learning_rate": 6.591881577425696e-06, "loss": 0.6584, "step": 315980 }, { "epoch": 27.37, "learning_rate": 6.589713195327858e-06, "loss": 0.5359, "step": 315990 }, { "epoch": 27.37, "learning_rate": 6.587544813230021e-06, "loss": 0.5398, "step": 316000 }, { "epoch": 27.37, "learning_rate": 6.585376431132184e-06, "loss": 0.7595, "step": 316010 }, { "epoch": 27.37, "learning_rate": 6.583208049034346e-06, "loss": 0.7165, "step": 316020 }, { "epoch": 27.37, "learning_rate": 6.581039666936509e-06, "loss": 0.496, "step": 316030 }, { "epoch": 27.37, "learning_rate": 6.578871284838671e-06, "loss": 0.6869, "step": 316040 }, { "epoch": 27.37, "learning_rate": 6.576702902740834e-06, "loss": 0.5944, "step": 316050 }, { "epoch": 27.37, "learning_rate": 6.5745345206429964e-06, "loss": 0.7167, "step": 316060 }, { "epoch": 27.37, "learning_rate": 6.5723661385451595e-06, "loss": 0.6212, "step": 316070 }, { "epoch": 27.37, "learning_rate": 6.5701977564473225e-06, "loss": 0.5691, "step": 316080 }, { "epoch": 27.38, "learning_rate": 6.568029374349485e-06, "loss": 0.5738, "step": 316090 }, { "epoch": 27.38, "learning_rate": 6.565860992251648e-06, "loss": 0.5725, "step": 316100 }, { "epoch": 27.38, "learning_rate": 6.56369261015381e-06, "loss": 0.6793, "step": 316110 }, { "epoch": 27.38, "learning_rate": 6.561524228055973e-06, "loss": 0.6512, "step": 316120 }, { "epoch": 27.38, "learning_rate": 6.559355845958135e-06, "loss": 0.6339, "step": 316130 }, { "epoch": 27.38, "learning_rate": 6.557187463860298e-06, "loss": 0.4207, "step": 316140 }, { "epoch": 27.38, "learning_rate": 6.555019081762461e-06, "loss": 0.5512, "step": 316150 }, { "epoch": 27.38, "learning_rate": 6.552850699664623e-06, "loss": 0.5639, "step": 316160 }, { "epoch": 27.38, "learning_rate": 6.550682317566786e-06, "loss": 0.7063, "step": 316170 }, { "epoch": 27.38, "learning_rate": 6.548513935468948e-06, "loss": 0.6282, "step": 316180 }, { "epoch": 27.38, "learning_rate": 6.546345553371111e-06, "loss": 0.7244, "step": 316190 }, { "epoch": 27.38, "learning_rate": 6.544177171273273e-06, "loss": 0.6494, "step": 316200 }, { "epoch": 27.39, "learning_rate": 6.542008789175436e-06, "loss": 0.5315, "step": 316210 }, { "epoch": 27.39, "learning_rate": 6.5398404070775994e-06, "loss": 0.5262, "step": 316220 }, { "epoch": 27.39, "learning_rate": 6.537672024979762e-06, "loss": 0.6644, "step": 316230 }, { "epoch": 27.39, "learning_rate": 6.535503642881923e-06, "loss": 0.6754, "step": 316240 }, { "epoch": 27.39, "learning_rate": 6.533335260784086e-06, "loss": 0.6062, "step": 316250 }, { "epoch": 27.39, "learning_rate": 6.531166878686249e-06, "loss": 0.6035, "step": 316260 }, { "epoch": 27.39, "learning_rate": 6.528998496588411e-06, "loss": 0.5573, "step": 316270 }, { "epoch": 27.39, "learning_rate": 6.526830114490574e-06, "loss": 0.5491, "step": 316280 }, { "epoch": 27.39, "learning_rate": 6.524661732392736e-06, "loss": 0.5104, "step": 316290 }, { "epoch": 27.39, "learning_rate": 6.522493350294899e-06, "loss": 0.5609, "step": 316300 }, { "epoch": 27.39, "learning_rate": 6.520324968197061e-06, "loss": 0.6428, "step": 316310 }, { "epoch": 27.4, "learning_rate": 6.518156586099224e-06, "loss": 0.5144, "step": 316320 }, { "epoch": 27.4, "learning_rate": 6.515988204001387e-06, "loss": 0.6333, "step": 316330 }, { "epoch": 27.4, "learning_rate": 6.5138198219035495e-06, "loss": 0.6343, "step": 316340 }, { "epoch": 27.4, "learning_rate": 6.5116514398057125e-06, "loss": 0.617, "step": 316350 }, { "epoch": 27.4, "learning_rate": 6.509483057707875e-06, "loss": 0.559, "step": 316360 }, { "epoch": 27.4, "learning_rate": 6.507314675610038e-06, "loss": 0.5487, "step": 316370 }, { "epoch": 27.4, "learning_rate": 6.5051462935122e-06, "loss": 0.4786, "step": 316380 }, { "epoch": 27.4, "learning_rate": 6.502977911414363e-06, "loss": 0.6363, "step": 316390 }, { "epoch": 27.4, "learning_rate": 6.500809529316526e-06, "loss": 0.4684, "step": 316400 }, { "epoch": 27.4, "learning_rate": 6.498641147218688e-06, "loss": 0.4637, "step": 316410 }, { "epoch": 27.4, "learning_rate": 6.496472765120851e-06, "loss": 0.5669, "step": 316420 }, { "epoch": 27.4, "learning_rate": 6.494304383023013e-06, "loss": 0.666, "step": 316430 }, { "epoch": 27.41, "learning_rate": 6.492136000925176e-06, "loss": 0.6219, "step": 316440 }, { "epoch": 27.41, "learning_rate": 6.489967618827338e-06, "loss": 0.766, "step": 316450 }, { "epoch": 27.41, "learning_rate": 6.487799236729501e-06, "loss": 0.6338, "step": 316460 }, { "epoch": 27.41, "learning_rate": 6.485630854631664e-06, "loss": 0.5862, "step": 316470 }, { "epoch": 27.41, "learning_rate": 6.4834624725338265e-06, "loss": 0.5576, "step": 316480 }, { "epoch": 27.41, "learning_rate": 6.481294090435989e-06, "loss": 0.7297, "step": 316490 }, { "epoch": 27.41, "learning_rate": 6.479125708338151e-06, "loss": 0.5601, "step": 316500 }, { "epoch": 27.41, "learning_rate": 6.476957326240314e-06, "loss": 0.6214, "step": 316510 }, { "epoch": 27.41, "learning_rate": 6.474788944142476e-06, "loss": 0.686, "step": 316520 }, { "epoch": 27.41, "learning_rate": 6.472620562044639e-06, "loss": 0.6402, "step": 316530 }, { "epoch": 27.41, "learning_rate": 6.470452179946801e-06, "loss": 0.7014, "step": 316540 }, { "epoch": 27.42, "learning_rate": 6.468283797848964e-06, "loss": 0.5172, "step": 316550 }, { "epoch": 27.42, "learning_rate": 6.466115415751127e-06, "loss": 0.6112, "step": 316560 }, { "epoch": 27.42, "learning_rate": 6.463947033653289e-06, "loss": 0.5736, "step": 316570 }, { "epoch": 27.42, "learning_rate": 6.461778651555452e-06, "loss": 0.6231, "step": 316580 }, { "epoch": 27.42, "learning_rate": 6.4596102694576145e-06, "loss": 0.6593, "step": 316590 }, { "epoch": 27.42, "learning_rate": 6.4574418873597775e-06, "loss": 0.4924, "step": 316600 }, { "epoch": 27.42, "learning_rate": 6.45527350526194e-06, "loss": 0.5945, "step": 316610 }, { "epoch": 27.42, "learning_rate": 6.453105123164103e-06, "loss": 0.6202, "step": 316620 }, { "epoch": 27.42, "learning_rate": 6.450936741066266e-06, "loss": 0.4913, "step": 316630 }, { "epoch": 27.42, "learning_rate": 6.448768358968428e-06, "loss": 0.5506, "step": 316640 }, { "epoch": 27.42, "learning_rate": 6.446599976870591e-06, "loss": 0.5414, "step": 316650 }, { "epoch": 27.42, "learning_rate": 6.444431594772753e-06, "loss": 0.5547, "step": 316660 }, { "epoch": 27.43, "learning_rate": 6.442263212674916e-06, "loss": 0.6484, "step": 316670 }, { "epoch": 27.43, "learning_rate": 6.440094830577078e-06, "loss": 0.5646, "step": 316680 }, { "epoch": 27.43, "learning_rate": 6.437926448479241e-06, "loss": 0.6031, "step": 316690 }, { "epoch": 27.43, "learning_rate": 6.435758066381404e-06, "loss": 0.6295, "step": 316700 }, { "epoch": 27.43, "learning_rate": 6.433589684283566e-06, "loss": 0.6347, "step": 316710 }, { "epoch": 27.43, "learning_rate": 6.431421302185729e-06, "loss": 0.5972, "step": 316720 }, { "epoch": 27.43, "learning_rate": 6.4292529200878914e-06, "loss": 0.781, "step": 316730 }, { "epoch": 27.43, "learning_rate": 6.4270845379900544e-06, "loss": 0.4821, "step": 316740 }, { "epoch": 27.43, "learning_rate": 6.424916155892216e-06, "loss": 0.4398, "step": 316750 }, { "epoch": 27.43, "learning_rate": 6.422747773794379e-06, "loss": 0.5314, "step": 316760 }, { "epoch": 27.43, "learning_rate": 6.420579391696541e-06, "loss": 0.5831, "step": 316770 }, { "epoch": 27.44, "learning_rate": 6.418411009598704e-06, "loss": 0.5417, "step": 316780 }, { "epoch": 27.44, "learning_rate": 6.416242627500866e-06, "loss": 0.5812, "step": 316790 }, { "epoch": 27.44, "learning_rate": 6.414074245403029e-06, "loss": 0.6235, "step": 316800 }, { "epoch": 27.44, "learning_rate": 6.411905863305192e-06, "loss": 0.5636, "step": 316810 }, { "epoch": 27.44, "learning_rate": 6.409737481207354e-06, "loss": 0.6888, "step": 316820 }, { "epoch": 27.44, "learning_rate": 6.407569099109517e-06, "loss": 0.5585, "step": 316830 }, { "epoch": 27.44, "learning_rate": 6.405400717011679e-06, "loss": 0.5686, "step": 316840 }, { "epoch": 27.44, "learning_rate": 6.403232334913842e-06, "loss": 0.635, "step": 316850 }, { "epoch": 27.44, "learning_rate": 6.4010639528160045e-06, "loss": 0.6427, "step": 316860 }, { "epoch": 27.44, "learning_rate": 6.3988955707181675e-06, "loss": 0.5896, "step": 316870 }, { "epoch": 27.44, "learning_rate": 6.3967271886203306e-06, "loss": 0.623, "step": 316880 }, { "epoch": 27.44, "learning_rate": 6.394558806522493e-06, "loss": 0.6884, "step": 316890 }, { "epoch": 27.45, "learning_rate": 6.392390424424656e-06, "loss": 0.6597, "step": 316900 }, { "epoch": 27.45, "learning_rate": 6.390222042326818e-06, "loss": 0.55, "step": 316910 }, { "epoch": 27.45, "learning_rate": 6.388053660228981e-06, "loss": 0.5041, "step": 316920 }, { "epoch": 27.45, "learning_rate": 6.385885278131143e-06, "loss": 0.5076, "step": 316930 }, { "epoch": 27.45, "learning_rate": 6.383716896033306e-06, "loss": 0.6131, "step": 316940 }, { "epoch": 27.45, "learning_rate": 6.381548513935469e-06, "loss": 0.5481, "step": 316950 }, { "epoch": 27.45, "learning_rate": 6.379380131837631e-06, "loss": 0.6456, "step": 316960 }, { "epoch": 27.45, "learning_rate": 6.377211749739794e-06, "loss": 0.5905, "step": 316970 }, { "epoch": 27.45, "learning_rate": 6.375043367641956e-06, "loss": 0.6092, "step": 316980 }, { "epoch": 27.45, "learning_rate": 6.372874985544119e-06, "loss": 0.5294, "step": 316990 }, { "epoch": 27.45, "learning_rate": 6.3707066034462815e-06, "loss": 0.5223, "step": 317000 }, { "epoch": 27.46, "learning_rate": 6.368538221348444e-06, "loss": 0.5275, "step": 317010 }, { "epoch": 27.46, "learning_rate": 6.366369839250606e-06, "loss": 0.6838, "step": 317020 }, { "epoch": 27.46, "learning_rate": 6.364201457152769e-06, "loss": 0.6226, "step": 317030 }, { "epoch": 27.46, "learning_rate": 6.362033075054932e-06, "loss": 0.6249, "step": 317040 }, { "epoch": 27.46, "learning_rate": 6.359864692957094e-06, "loss": 0.6705, "step": 317050 }, { "epoch": 27.46, "learning_rate": 6.357696310859257e-06, "loss": 0.4712, "step": 317060 }, { "epoch": 27.46, "learning_rate": 6.355527928761419e-06, "loss": 0.6148, "step": 317070 }, { "epoch": 27.46, "learning_rate": 6.353359546663582e-06, "loss": 0.6586, "step": 317080 }, { "epoch": 27.46, "learning_rate": 6.351191164565744e-06, "loss": 0.4642, "step": 317090 }, { "epoch": 27.46, "learning_rate": 6.349022782467907e-06, "loss": 0.5575, "step": 317100 }, { "epoch": 27.46, "learning_rate": 6.34685440037007e-06, "loss": 0.5968, "step": 317110 }, { "epoch": 27.46, "learning_rate": 6.3446860182722325e-06, "loss": 0.571, "step": 317120 }, { "epoch": 27.47, "learning_rate": 6.3425176361743955e-06, "loss": 0.6333, "step": 317130 }, { "epoch": 27.47, "learning_rate": 6.340349254076558e-06, "loss": 0.5692, "step": 317140 }, { "epoch": 27.47, "learning_rate": 6.338180871978721e-06, "loss": 0.6431, "step": 317150 }, { "epoch": 27.47, "learning_rate": 6.336012489880883e-06, "loss": 0.5453, "step": 317160 }, { "epoch": 27.47, "learning_rate": 6.333844107783046e-06, "loss": 0.6334, "step": 317170 }, { "epoch": 27.47, "learning_rate": 6.331675725685209e-06, "loss": 0.4856, "step": 317180 }, { "epoch": 27.47, "learning_rate": 6.329507343587371e-06, "loss": 0.4695, "step": 317190 }, { "epoch": 27.47, "learning_rate": 6.327338961489534e-06, "loss": 0.5792, "step": 317200 }, { "epoch": 27.47, "learning_rate": 6.325170579391696e-06, "loss": 0.6391, "step": 317210 }, { "epoch": 27.47, "learning_rate": 6.323002197293859e-06, "loss": 0.5427, "step": 317220 }, { "epoch": 27.47, "learning_rate": 6.320833815196021e-06, "loss": 0.5207, "step": 317230 }, { "epoch": 27.47, "learning_rate": 6.318665433098184e-06, "loss": 0.5566, "step": 317240 }, { "epoch": 27.48, "learning_rate": 6.316497051000347e-06, "loss": 0.5901, "step": 317250 }, { "epoch": 27.48, "learning_rate": 6.314328668902509e-06, "loss": 0.5782, "step": 317260 }, { "epoch": 27.48, "learning_rate": 6.312160286804671e-06, "loss": 0.5191, "step": 317270 }, { "epoch": 27.48, "learning_rate": 6.309991904706834e-06, "loss": 0.5443, "step": 317280 }, { "epoch": 27.48, "learning_rate": 6.307823522608997e-06, "loss": 0.593, "step": 317290 }, { "epoch": 27.48, "learning_rate": 6.305655140511159e-06, "loss": 0.4562, "step": 317300 }, { "epoch": 27.48, "learning_rate": 6.303486758413322e-06, "loss": 0.5355, "step": 317310 }, { "epoch": 27.48, "learning_rate": 6.301318376315484e-06, "loss": 0.6166, "step": 317320 }, { "epoch": 27.48, "learning_rate": 6.299149994217647e-06, "loss": 0.6388, "step": 317330 }, { "epoch": 27.48, "learning_rate": 6.296981612119809e-06, "loss": 0.6733, "step": 317340 }, { "epoch": 27.48, "learning_rate": 6.294813230021972e-06, "loss": 0.5737, "step": 317350 }, { "epoch": 27.49, "learning_rate": 6.292644847924135e-06, "loss": 0.5307, "step": 317360 }, { "epoch": 27.49, "learning_rate": 6.290476465826297e-06, "loss": 0.5303, "step": 317370 }, { "epoch": 27.49, "learning_rate": 6.28830808372846e-06, "loss": 0.6141, "step": 317380 }, { "epoch": 27.49, "learning_rate": 6.2861397016306226e-06, "loss": 0.4766, "step": 317390 }, { "epoch": 27.49, "learning_rate": 6.2839713195327856e-06, "loss": 0.6463, "step": 317400 }, { "epoch": 27.49, "learning_rate": 6.281802937434948e-06, "loss": 0.6275, "step": 317410 }, { "epoch": 27.49, "learning_rate": 6.279634555337111e-06, "loss": 0.5179, "step": 317420 }, { "epoch": 27.49, "learning_rate": 6.277466173239274e-06, "loss": 0.6138, "step": 317430 }, { "epoch": 27.49, "learning_rate": 6.275297791141436e-06, "loss": 0.5998, "step": 317440 }, { "epoch": 27.49, "learning_rate": 6.273129409043599e-06, "loss": 0.5341, "step": 317450 }, { "epoch": 27.49, "learning_rate": 6.270961026945761e-06, "loss": 0.6854, "step": 317460 }, { "epoch": 27.49, "learning_rate": 6.268792644847924e-06, "loss": 0.6415, "step": 317470 }, { "epoch": 27.5, "learning_rate": 6.266624262750086e-06, "loss": 0.6151, "step": 317480 }, { "epoch": 27.5, "learning_rate": 6.264455880652249e-06, "loss": 0.6335, "step": 317490 }, { "epoch": 27.5, "learning_rate": 6.262287498554412e-06, "loss": 0.6025, "step": 317500 }, { "epoch": 27.5, "learning_rate": 6.260119116456574e-06, "loss": 0.6517, "step": 317510 }, { "epoch": 27.5, "learning_rate": 6.257950734358736e-06, "loss": 0.4856, "step": 317520 }, { "epoch": 27.5, "learning_rate": 6.255782352260899e-06, "loss": 0.5896, "step": 317530 }, { "epoch": 27.5, "learning_rate": 6.253613970163062e-06, "loss": 0.6678, "step": 317540 }, { "epoch": 27.5, "learning_rate": 6.251445588065224e-06, "loss": 0.7848, "step": 317550 }, { "epoch": 27.5, "learning_rate": 6.249277205967387e-06, "loss": 0.6911, "step": 317560 }, { "epoch": 27.5, "learning_rate": 6.247108823869549e-06, "loss": 0.6294, "step": 317570 }, { "epoch": 27.5, "learning_rate": 6.244940441771712e-06, "loss": 0.5373, "step": 317580 }, { "epoch": 27.51, "learning_rate": 6.242772059673874e-06, "loss": 0.592, "step": 317590 }, { "epoch": 27.51, "learning_rate": 6.240603677576037e-06, "loss": 0.6911, "step": 317600 }, { "epoch": 27.51, "learning_rate": 6.2384352954782e-06, "loss": 0.6439, "step": 317610 }, { "epoch": 27.51, "learning_rate": 6.236266913380362e-06, "loss": 0.5449, "step": 317620 }, { "epoch": 27.51, "learning_rate": 6.234098531282525e-06, "loss": 0.6478, "step": 317630 }, { "epoch": 27.51, "learning_rate": 6.2319301491846875e-06, "loss": 0.6341, "step": 317640 }, { "epoch": 27.51, "learning_rate": 6.2297617670868505e-06, "loss": 0.5022, "step": 317650 }, { "epoch": 27.51, "learning_rate": 6.227593384989013e-06, "loss": 0.5146, "step": 317660 }, { "epoch": 27.51, "learning_rate": 6.225425002891176e-06, "loss": 0.83, "step": 317670 }, { "epoch": 27.51, "learning_rate": 6.223256620793339e-06, "loss": 0.5542, "step": 317680 }, { "epoch": 27.51, "learning_rate": 6.221088238695501e-06, "loss": 0.7229, "step": 317690 }, { "epoch": 27.51, "learning_rate": 6.218919856597664e-06, "loss": 0.6149, "step": 317700 }, { "epoch": 27.52, "learning_rate": 6.216751474499826e-06, "loss": 0.533, "step": 317710 }, { "epoch": 27.52, "learning_rate": 6.214583092401989e-06, "loss": 0.6475, "step": 317720 }, { "epoch": 27.52, "learning_rate": 6.212414710304151e-06, "loss": 0.6306, "step": 317730 }, { "epoch": 27.52, "learning_rate": 6.210246328206314e-06, "loss": 0.6019, "step": 317740 }, { "epoch": 27.52, "learning_rate": 6.208077946108477e-06, "loss": 0.5736, "step": 317750 }, { "epoch": 27.52, "learning_rate": 6.205909564010639e-06, "loss": 0.6114, "step": 317760 }, { "epoch": 27.52, "learning_rate": 6.203741181912802e-06, "loss": 0.5413, "step": 317770 }, { "epoch": 27.52, "learning_rate": 6.201572799814964e-06, "loss": 0.6676, "step": 317780 }, { "epoch": 27.52, "learning_rate": 6.199404417717127e-06, "loss": 0.5393, "step": 317790 }, { "epoch": 27.52, "learning_rate": 6.197236035619289e-06, "loss": 0.6878, "step": 317800 }, { "epoch": 27.52, "learning_rate": 6.195067653521452e-06, "loss": 0.5263, "step": 317810 }, { "epoch": 27.53, "learning_rate": 6.192899271423614e-06, "loss": 0.5454, "step": 317820 }, { "epoch": 27.53, "learning_rate": 6.190730889325777e-06, "loss": 0.5254, "step": 317830 }, { "epoch": 27.53, "learning_rate": 6.18856250722794e-06, "loss": 0.6872, "step": 317840 }, { "epoch": 27.53, "learning_rate": 6.186394125130102e-06, "loss": 0.5694, "step": 317850 }, { "epoch": 27.53, "learning_rate": 6.184225743032265e-06, "loss": 0.5314, "step": 317860 }, { "epoch": 27.53, "learning_rate": 6.182057360934427e-06, "loss": 0.521, "step": 317870 }, { "epoch": 27.53, "learning_rate": 6.17988897883659e-06, "loss": 0.7166, "step": 317880 }, { "epoch": 27.53, "learning_rate": 6.177720596738752e-06, "loss": 0.5968, "step": 317890 }, { "epoch": 27.53, "learning_rate": 6.175552214640915e-06, "loss": 0.5921, "step": 317900 }, { "epoch": 27.53, "learning_rate": 6.173383832543078e-06, "loss": 0.6669, "step": 317910 }, { "epoch": 27.53, "learning_rate": 6.1712154504452406e-06, "loss": 0.5718, "step": 317920 }, { "epoch": 27.53, "learning_rate": 6.1690470683474036e-06, "loss": 0.5712, "step": 317930 }, { "epoch": 27.54, "learning_rate": 6.166878686249566e-06, "loss": 0.5858, "step": 317940 }, { "epoch": 27.54, "learning_rate": 6.164710304151729e-06, "loss": 0.5242, "step": 317950 }, { "epoch": 27.54, "learning_rate": 6.162541922053891e-06, "loss": 0.5688, "step": 317960 }, { "epoch": 27.54, "learning_rate": 6.160373539956054e-06, "loss": 0.691, "step": 317970 }, { "epoch": 27.54, "learning_rate": 6.158205157858217e-06, "loss": 0.6424, "step": 317980 }, { "epoch": 27.54, "learning_rate": 6.156036775760379e-06, "loss": 0.7459, "step": 317990 }, { "epoch": 27.54, "learning_rate": 6.153868393662542e-06, "loss": 0.5801, "step": 318000 }, { "epoch": 27.54, "learning_rate": 6.151700011564704e-06, "loss": 0.5332, "step": 318010 }, { "epoch": 27.54, "learning_rate": 6.149531629466867e-06, "loss": 0.5971, "step": 318020 }, { "epoch": 27.54, "learning_rate": 6.1473632473690285e-06, "loss": 0.528, "step": 318030 }, { "epoch": 27.54, "learning_rate": 6.1451948652711915e-06, "loss": 0.6522, "step": 318040 }, { "epoch": 27.55, "learning_rate": 6.143026483173354e-06, "loss": 0.5688, "step": 318050 }, { "epoch": 27.55, "learning_rate": 6.140858101075517e-06, "loss": 0.6612, "step": 318060 }, { "epoch": 27.55, "learning_rate": 6.138689718977679e-06, "loss": 0.5473, "step": 318070 }, { "epoch": 27.55, "learning_rate": 6.136521336879842e-06, "loss": 0.6025, "step": 318080 }, { "epoch": 27.55, "learning_rate": 6.134352954782005e-06, "loss": 0.4778, "step": 318090 }, { "epoch": 27.55, "learning_rate": 6.132184572684167e-06, "loss": 0.482, "step": 318100 }, { "epoch": 27.55, "learning_rate": 6.13001619058633e-06, "loss": 0.5429, "step": 318110 }, { "epoch": 27.55, "learning_rate": 6.127847808488492e-06, "loss": 0.5872, "step": 318120 }, { "epoch": 27.55, "learning_rate": 6.125679426390655e-06, "loss": 0.4444, "step": 318130 }, { "epoch": 27.55, "learning_rate": 6.123511044292817e-06, "loss": 0.5699, "step": 318140 }, { "epoch": 27.55, "learning_rate": 6.12134266219498e-06, "loss": 0.6054, "step": 318150 }, { "epoch": 27.55, "learning_rate": 6.119174280097143e-06, "loss": 0.4976, "step": 318160 }, { "epoch": 27.56, "learning_rate": 6.1170058979993055e-06, "loss": 0.5863, "step": 318170 }, { "epoch": 27.56, "learning_rate": 6.1148375159014685e-06, "loss": 0.6142, "step": 318180 }, { "epoch": 27.56, "learning_rate": 6.112669133803631e-06, "loss": 0.6239, "step": 318190 }, { "epoch": 27.56, "learning_rate": 6.110500751705794e-06, "loss": 0.5895, "step": 318200 }, { "epoch": 27.56, "learning_rate": 6.108332369607956e-06, "loss": 0.5195, "step": 318210 }, { "epoch": 27.56, "learning_rate": 6.106163987510119e-06, "loss": 0.5226, "step": 318220 }, { "epoch": 27.56, "learning_rate": 6.103995605412282e-06, "loss": 0.6332, "step": 318230 }, { "epoch": 27.56, "learning_rate": 6.101827223314444e-06, "loss": 0.5644, "step": 318240 }, { "epoch": 27.56, "learning_rate": 6.099658841216607e-06, "loss": 0.4891, "step": 318250 }, { "epoch": 27.56, "learning_rate": 6.097490459118769e-06, "loss": 0.6108, "step": 318260 }, { "epoch": 27.56, "learning_rate": 6.095322077020932e-06, "loss": 0.5962, "step": 318270 }, { "epoch": 27.57, "learning_rate": 6.093153694923094e-06, "loss": 0.7165, "step": 318280 }, { "epoch": 27.57, "learning_rate": 6.0909853128252564e-06, "loss": 0.5581, "step": 318290 }, { "epoch": 27.57, "learning_rate": 6.088816930727419e-06, "loss": 0.6033, "step": 318300 }, { "epoch": 27.57, "learning_rate": 6.086648548629582e-06, "loss": 0.4864, "step": 318310 }, { "epoch": 27.57, "learning_rate": 6.084480166531745e-06, "loss": 0.5121, "step": 318320 }, { "epoch": 27.57, "learning_rate": 6.082311784433907e-06, "loss": 0.728, "step": 318330 }, { "epoch": 27.57, "learning_rate": 6.08014340233607e-06, "loss": 0.6121, "step": 318340 }, { "epoch": 27.57, "learning_rate": 6.077975020238232e-06, "loss": 0.6308, "step": 318350 }, { "epoch": 27.57, "learning_rate": 6.075806638140395e-06, "loss": 0.5044, "step": 318360 }, { "epoch": 27.57, "learning_rate": 6.073638256042557e-06, "loss": 0.5799, "step": 318370 }, { "epoch": 27.57, "learning_rate": 6.07146987394472e-06, "loss": 0.609, "step": 318380 }, { "epoch": 27.57, "learning_rate": 6.069301491846883e-06, "loss": 0.628, "step": 318390 }, { "epoch": 27.58, "learning_rate": 6.067133109749045e-06, "loss": 0.6669, "step": 318400 }, { "epoch": 27.58, "learning_rate": 6.064964727651208e-06, "loss": 0.6031, "step": 318410 }, { "epoch": 27.58, "learning_rate": 6.06279634555337e-06, "loss": 0.6125, "step": 318420 }, { "epoch": 27.58, "learning_rate": 6.060627963455533e-06, "loss": 0.417, "step": 318430 }, { "epoch": 27.58, "learning_rate": 6.0584595813576956e-06, "loss": 0.5864, "step": 318440 }, { "epoch": 27.58, "learning_rate": 6.0562911992598586e-06, "loss": 0.691, "step": 318450 }, { "epoch": 27.58, "learning_rate": 6.054122817162022e-06, "loss": 0.6861, "step": 318460 }, { "epoch": 27.58, "learning_rate": 6.051954435064184e-06, "loss": 0.567, "step": 318470 }, { "epoch": 27.58, "learning_rate": 6.049786052966347e-06, "loss": 0.5886, "step": 318480 }, { "epoch": 27.58, "learning_rate": 6.047617670868509e-06, "loss": 0.5799, "step": 318490 }, { "epoch": 27.58, "learning_rate": 6.045449288770672e-06, "loss": 0.6691, "step": 318500 }, { "epoch": 27.58, "learning_rate": 6.043280906672834e-06, "loss": 0.5221, "step": 318510 }, { "epoch": 27.59, "learning_rate": 6.041112524574997e-06, "loss": 0.6135, "step": 318520 }, { "epoch": 27.59, "learning_rate": 6.03894414247716e-06, "loss": 0.6064, "step": 318530 }, { "epoch": 27.59, "learning_rate": 6.036775760379322e-06, "loss": 0.5137, "step": 318540 }, { "epoch": 27.59, "learning_rate": 6.0346073782814835e-06, "loss": 0.4672, "step": 318550 }, { "epoch": 27.59, "learning_rate": 6.0324389961836465e-06, "loss": 0.6055, "step": 318560 }, { "epoch": 27.59, "learning_rate": 6.0302706140858095e-06, "loss": 0.5381, "step": 318570 }, { "epoch": 27.59, "learning_rate": 6.028102231987972e-06, "loss": 0.6066, "step": 318580 }, { "epoch": 27.59, "learning_rate": 6.025933849890135e-06, "loss": 0.6784, "step": 318590 }, { "epoch": 27.59, "learning_rate": 6.023765467792297e-06, "loss": 0.5798, "step": 318600 }, { "epoch": 27.59, "learning_rate": 6.02159708569446e-06, "loss": 0.645, "step": 318610 }, { "epoch": 27.59, "learning_rate": 6.019428703596622e-06, "loss": 0.5631, "step": 318620 }, { "epoch": 27.6, "learning_rate": 6.017260321498785e-06, "loss": 0.6167, "step": 318630 }, { "epoch": 27.6, "learning_rate": 6.015091939400948e-06, "loss": 0.5034, "step": 318640 }, { "epoch": 27.6, "learning_rate": 6.01292355730311e-06, "loss": 0.5961, "step": 318650 }, { "epoch": 27.6, "learning_rate": 6.010755175205273e-06, "loss": 0.6615, "step": 318660 }, { "epoch": 27.6, "learning_rate": 6.008586793107435e-06, "loss": 0.5785, "step": 318670 }, { "epoch": 27.6, "learning_rate": 6.006418411009598e-06, "loss": 0.6125, "step": 318680 }, { "epoch": 27.6, "learning_rate": 6.0042500289117605e-06, "loss": 0.5766, "step": 318690 }, { "epoch": 27.6, "learning_rate": 6.0020816468139235e-06, "loss": 0.7514, "step": 318700 }, { "epoch": 27.6, "learning_rate": 5.9999132647160865e-06, "loss": 0.5575, "step": 318710 }, { "epoch": 27.6, "learning_rate": 5.997744882618249e-06, "loss": 0.6675, "step": 318720 }, { "epoch": 27.6, "learning_rate": 5.995576500520412e-06, "loss": 0.5334, "step": 318730 }, { "epoch": 27.6, "learning_rate": 5.993408118422574e-06, "loss": 0.6952, "step": 318740 }, { "epoch": 27.61, "learning_rate": 5.991239736324737e-06, "loss": 0.4564, "step": 318750 }, { "epoch": 27.61, "learning_rate": 5.989071354226899e-06, "loss": 0.5585, "step": 318760 }, { "epoch": 27.61, "learning_rate": 5.986902972129062e-06, "loss": 0.5698, "step": 318770 }, { "epoch": 27.61, "learning_rate": 5.984734590031225e-06, "loss": 0.6918, "step": 318780 }, { "epoch": 27.61, "learning_rate": 5.982566207933387e-06, "loss": 0.5651, "step": 318790 }, { "epoch": 27.61, "learning_rate": 5.9803978258355485e-06, "loss": 0.6414, "step": 318800 }, { "epoch": 27.61, "learning_rate": 5.9782294437377115e-06, "loss": 0.6223, "step": 318810 }, { "epoch": 27.61, "learning_rate": 5.9760610616398745e-06, "loss": 0.4443, "step": 318820 }, { "epoch": 27.61, "learning_rate": 5.973892679542037e-06, "loss": 0.5453, "step": 318830 }, { "epoch": 27.61, "learning_rate": 5.9717242974442e-06, "loss": 0.6195, "step": 318840 }, { "epoch": 27.61, "learning_rate": 5.969555915346362e-06, "loss": 0.5901, "step": 318850 }, { "epoch": 27.62, "learning_rate": 5.967387533248525e-06, "loss": 0.7008, "step": 318860 }, { "epoch": 27.62, "learning_rate": 5.965219151150687e-06, "loss": 0.5411, "step": 318870 }, { "epoch": 27.62, "learning_rate": 5.96305076905285e-06, "loss": 0.5981, "step": 318880 }, { "epoch": 27.62, "learning_rate": 5.960882386955013e-06, "loss": 0.4926, "step": 318890 }, { "epoch": 27.62, "learning_rate": 5.958714004857175e-06, "loss": 0.606, "step": 318900 }, { "epoch": 27.62, "learning_rate": 5.956545622759338e-06, "loss": 0.6378, "step": 318910 }, { "epoch": 27.62, "learning_rate": 5.9543772406615e-06, "loss": 0.6292, "step": 318920 }, { "epoch": 27.62, "learning_rate": 5.952208858563663e-06, "loss": 0.4979, "step": 318930 }, { "epoch": 27.62, "learning_rate": 5.950040476465825e-06, "loss": 0.6196, "step": 318940 }, { "epoch": 27.62, "learning_rate": 5.947872094367988e-06, "loss": 0.6103, "step": 318950 }, { "epoch": 27.62, "learning_rate": 5.9457037122701514e-06, "loss": 0.6552, "step": 318960 }, { "epoch": 27.62, "learning_rate": 5.943535330172314e-06, "loss": 0.5635, "step": 318970 }, { "epoch": 27.63, "learning_rate": 5.941366948074477e-06, "loss": 0.6465, "step": 318980 }, { "epoch": 27.63, "learning_rate": 5.939198565976639e-06, "loss": 0.4804, "step": 318990 }, { "epoch": 27.63, "learning_rate": 5.937030183878802e-06, "loss": 0.6063, "step": 319000 }, { "epoch": 27.63, "learning_rate": 5.934861801780964e-06, "loss": 0.6362, "step": 319010 }, { "epoch": 27.63, "learning_rate": 5.932693419683127e-06, "loss": 0.684, "step": 319020 }, { "epoch": 27.63, "learning_rate": 5.93052503758529e-06, "loss": 0.4721, "step": 319030 }, { "epoch": 27.63, "learning_rate": 5.928356655487452e-06, "loss": 0.6292, "step": 319040 }, { "epoch": 27.63, "learning_rate": 5.926188273389615e-06, "loss": 0.4334, "step": 319050 }, { "epoch": 27.63, "learning_rate": 5.924019891291776e-06, "loss": 0.6442, "step": 319060 }, { "epoch": 27.63, "learning_rate": 5.921851509193939e-06, "loss": 0.5859, "step": 319070 }, { "epoch": 27.63, "learning_rate": 5.9196831270961015e-06, "loss": 0.4854, "step": 319080 }, { "epoch": 27.64, "learning_rate": 5.9175147449982645e-06, "loss": 0.5701, "step": 319090 }, { "epoch": 27.64, "learning_rate": 5.915346362900427e-06, "loss": 0.5623, "step": 319100 }, { "epoch": 27.64, "learning_rate": 5.91317798080259e-06, "loss": 0.6341, "step": 319110 }, { "epoch": 27.64, "learning_rate": 5.911009598704753e-06, "loss": 0.5627, "step": 319120 }, { "epoch": 27.64, "learning_rate": 5.908841216606915e-06, "loss": 0.4805, "step": 319130 }, { "epoch": 27.64, "learning_rate": 5.906672834509078e-06, "loss": 0.5706, "step": 319140 }, { "epoch": 27.64, "learning_rate": 5.90450445241124e-06, "loss": 0.5868, "step": 319150 }, { "epoch": 27.64, "learning_rate": 5.902336070313403e-06, "loss": 0.7542, "step": 319160 }, { "epoch": 27.64, "learning_rate": 5.900167688215565e-06, "loss": 0.5764, "step": 319170 }, { "epoch": 27.64, "learning_rate": 5.897999306117728e-06, "loss": 0.5904, "step": 319180 }, { "epoch": 27.64, "learning_rate": 5.895830924019891e-06, "loss": 0.6023, "step": 319190 }, { "epoch": 27.64, "learning_rate": 5.893662541922053e-06, "loss": 0.6708, "step": 319200 }, { "epoch": 27.65, "learning_rate": 5.891494159824216e-06, "loss": 0.4799, "step": 319210 }, { "epoch": 27.65, "learning_rate": 5.8893257777263785e-06, "loss": 0.5923, "step": 319220 }, { "epoch": 27.65, "learning_rate": 5.8871573956285415e-06, "loss": 0.6567, "step": 319230 }, { "epoch": 27.65, "learning_rate": 5.884989013530704e-06, "loss": 0.5185, "step": 319240 }, { "epoch": 27.65, "learning_rate": 5.882820631432867e-06, "loss": 0.5756, "step": 319250 }, { "epoch": 27.65, "learning_rate": 5.88065224933503e-06, "loss": 0.5206, "step": 319260 }, { "epoch": 27.65, "learning_rate": 5.878483867237192e-06, "loss": 0.5751, "step": 319270 }, { "epoch": 27.65, "learning_rate": 5.876315485139355e-06, "loss": 0.5448, "step": 319280 }, { "epoch": 27.65, "learning_rate": 5.874147103041517e-06, "loss": 0.4439, "step": 319290 }, { "epoch": 27.65, "learning_rate": 5.87197872094368e-06, "loss": 0.646, "step": 319300 }, { "epoch": 27.65, "learning_rate": 5.869810338845842e-06, "loss": 0.6542, "step": 319310 }, { "epoch": 27.66, "learning_rate": 5.867641956748004e-06, "loss": 0.5923, "step": 319320 }, { "epoch": 27.66, "learning_rate": 5.8654735746501665e-06, "loss": 0.4089, "step": 319330 }, { "epoch": 27.66, "learning_rate": 5.8633051925523295e-06, "loss": 0.6215, "step": 319340 }, { "epoch": 27.66, "learning_rate": 5.861136810454492e-06, "loss": 0.6375, "step": 319350 }, { "epoch": 27.66, "learning_rate": 5.858968428356655e-06, "loss": 0.6518, "step": 319360 }, { "epoch": 27.66, "learning_rate": 5.856800046258818e-06, "loss": 0.646, "step": 319370 }, { "epoch": 27.66, "learning_rate": 5.85463166416098e-06, "loss": 0.5334, "step": 319380 }, { "epoch": 27.66, "learning_rate": 5.852463282063143e-06, "loss": 0.5584, "step": 319390 }, { "epoch": 27.66, "learning_rate": 5.850294899965305e-06, "loss": 0.5925, "step": 319400 }, { "epoch": 27.66, "learning_rate": 5.848126517867468e-06, "loss": 0.5473, "step": 319410 }, { "epoch": 27.66, "learning_rate": 5.84595813576963e-06, "loss": 0.4805, "step": 319420 }, { "epoch": 27.66, "learning_rate": 5.843789753671793e-06, "loss": 0.6536, "step": 319430 }, { "epoch": 27.67, "learning_rate": 5.841621371573956e-06, "loss": 0.6095, "step": 319440 }, { "epoch": 27.67, "learning_rate": 5.839452989476118e-06, "loss": 0.5403, "step": 319450 }, { "epoch": 27.67, "learning_rate": 5.837284607378281e-06, "loss": 0.5162, "step": 319460 }, { "epoch": 27.67, "learning_rate": 5.8351162252804434e-06, "loss": 0.5024, "step": 319470 }, { "epoch": 27.67, "learning_rate": 5.8329478431826064e-06, "loss": 0.5995, "step": 319480 }, { "epoch": 27.67, "learning_rate": 5.830779461084769e-06, "loss": 0.6151, "step": 319490 }, { "epoch": 27.67, "learning_rate": 5.828611078986932e-06, "loss": 0.5307, "step": 319500 }, { "epoch": 27.67, "learning_rate": 5.826442696889095e-06, "loss": 0.6066, "step": 319510 }, { "epoch": 27.67, "learning_rate": 5.824274314791257e-06, "loss": 0.6493, "step": 319520 }, { "epoch": 27.67, "learning_rate": 5.82210593269342e-06, "loss": 0.5848, "step": 319530 }, { "epoch": 27.67, "learning_rate": 5.819937550595582e-06, "loss": 0.5794, "step": 319540 }, { "epoch": 27.68, "learning_rate": 5.817769168497745e-06, "loss": 0.6674, "step": 319550 }, { "epoch": 27.68, "learning_rate": 5.815600786399907e-06, "loss": 0.5816, "step": 319560 }, { "epoch": 27.68, "learning_rate": 5.813432404302069e-06, "loss": 0.6313, "step": 319570 }, { "epoch": 27.68, "learning_rate": 5.811264022204231e-06, "loss": 0.4904, "step": 319580 }, { "epoch": 27.68, "learning_rate": 5.809095640106394e-06, "loss": 0.5002, "step": 319590 }, { "epoch": 27.68, "learning_rate": 5.806927258008557e-06, "loss": 0.5271, "step": 319600 }, { "epoch": 27.68, "learning_rate": 5.8047588759107195e-06, "loss": 0.6784, "step": 319610 }, { "epoch": 27.68, "learning_rate": 5.8025904938128826e-06, "loss": 0.6087, "step": 319620 }, { "epoch": 27.68, "learning_rate": 5.800422111715045e-06, "loss": 0.5014, "step": 319630 }, { "epoch": 27.68, "learning_rate": 5.798253729617208e-06, "loss": 0.5386, "step": 319640 }, { "epoch": 27.68, "learning_rate": 5.79608534751937e-06, "loss": 0.7189, "step": 319650 }, { "epoch": 27.68, "learning_rate": 5.793916965421533e-06, "loss": 0.5842, "step": 319660 }, { "epoch": 27.69, "learning_rate": 5.791748583323696e-06, "loss": 0.6851, "step": 319670 }, { "epoch": 27.69, "learning_rate": 5.789580201225858e-06, "loss": 0.6042, "step": 319680 }, { "epoch": 27.69, "learning_rate": 5.787411819128021e-06, "loss": 0.6054, "step": 319690 }, { "epoch": 27.69, "learning_rate": 5.785243437030183e-06, "loss": 0.5584, "step": 319700 }, { "epoch": 27.69, "learning_rate": 5.783075054932346e-06, "loss": 0.5999, "step": 319710 }, { "epoch": 27.69, "learning_rate": 5.780906672834508e-06, "loss": 0.5499, "step": 319720 }, { "epoch": 27.69, "learning_rate": 5.778738290736671e-06, "loss": 0.5406, "step": 319730 }, { "epoch": 27.69, "learning_rate": 5.776569908638834e-06, "loss": 0.6234, "step": 319740 }, { "epoch": 27.69, "learning_rate": 5.7744015265409965e-06, "loss": 0.6407, "step": 319750 }, { "epoch": 27.69, "learning_rate": 5.7722331444431595e-06, "loss": 0.6032, "step": 319760 }, { "epoch": 27.69, "learning_rate": 5.770064762345322e-06, "loss": 0.6499, "step": 319770 }, { "epoch": 27.69, "learning_rate": 5.767896380247485e-06, "loss": 0.6216, "step": 319780 }, { "epoch": 27.7, "learning_rate": 5.765727998149647e-06, "loss": 0.5956, "step": 319790 }, { "epoch": 27.7, "learning_rate": 5.76355961605181e-06, "loss": 0.471, "step": 319800 }, { "epoch": 27.7, "learning_rate": 5.761391233953973e-06, "loss": 0.5845, "step": 319810 }, { "epoch": 27.7, "learning_rate": 5.759222851856135e-06, "loss": 0.5914, "step": 319820 }, { "epoch": 27.7, "learning_rate": 5.757054469758296e-06, "loss": 0.5903, "step": 319830 }, { "epoch": 27.7, "learning_rate": 5.754886087660459e-06, "loss": 0.5712, "step": 319840 }, { "epoch": 27.7, "learning_rate": 5.752717705562622e-06, "loss": 0.717, "step": 319850 }, { "epoch": 27.7, "learning_rate": 5.7505493234647845e-06, "loss": 0.6823, "step": 319860 }, { "epoch": 27.7, "learning_rate": 5.7483809413669475e-06, "loss": 0.56, "step": 319870 }, { "epoch": 27.7, "learning_rate": 5.74621255926911e-06, "loss": 0.5369, "step": 319880 }, { "epoch": 27.7, "learning_rate": 5.744044177171273e-06, "loss": 0.5587, "step": 319890 }, { "epoch": 27.71, "learning_rate": 5.741875795073435e-06, "loss": 0.5717, "step": 319900 }, { "epoch": 27.71, "learning_rate": 5.739707412975598e-06, "loss": 0.5127, "step": 319910 }, { "epoch": 27.71, "learning_rate": 5.737539030877761e-06, "loss": 0.577, "step": 319920 }, { "epoch": 27.71, "learning_rate": 5.735370648779923e-06, "loss": 0.4798, "step": 319930 }, { "epoch": 27.71, "learning_rate": 5.733202266682086e-06, "loss": 0.675, "step": 319940 }, { "epoch": 27.71, "learning_rate": 5.731033884584248e-06, "loss": 0.4982, "step": 319950 }, { "epoch": 27.71, "learning_rate": 5.728865502486411e-06, "loss": 0.5715, "step": 319960 }, { "epoch": 27.71, "learning_rate": 5.726697120388573e-06, "loss": 0.5878, "step": 319970 }, { "epoch": 27.71, "learning_rate": 5.724528738290736e-06, "loss": 0.5711, "step": 319980 }, { "epoch": 27.71, "learning_rate": 5.722360356192899e-06, "loss": 0.5776, "step": 319990 }, { "epoch": 27.71, "learning_rate": 5.7201919740950614e-06, "loss": 0.5095, "step": 320000 }, { "epoch": 27.71, "learning_rate": 5.7180235919972244e-06, "loss": 0.6352, "step": 320010 }, { "epoch": 27.72, "learning_rate": 5.715855209899387e-06, "loss": 0.5025, "step": 320020 }, { "epoch": 27.72, "learning_rate": 5.71368682780155e-06, "loss": 0.5947, "step": 320030 }, { "epoch": 27.72, "learning_rate": 5.711518445703712e-06, "loss": 0.4909, "step": 320040 }, { "epoch": 27.72, "learning_rate": 5.709350063605875e-06, "loss": 0.6152, "step": 320050 }, { "epoch": 27.72, "learning_rate": 5.707181681508038e-06, "loss": 0.6719, "step": 320060 }, { "epoch": 27.72, "learning_rate": 5.7050132994102e-06, "loss": 0.5861, "step": 320070 }, { "epoch": 27.72, "learning_rate": 5.702844917312361e-06, "loss": 0.6271, "step": 320080 }, { "epoch": 27.72, "learning_rate": 5.700676535214524e-06, "loss": 0.501, "step": 320090 }, { "epoch": 27.72, "learning_rate": 5.698508153116687e-06, "loss": 0.7797, "step": 320100 }, { "epoch": 27.72, "learning_rate": 5.696339771018849e-06, "loss": 0.669, "step": 320110 }, { "epoch": 27.72, "learning_rate": 5.694171388921012e-06, "loss": 0.4878, "step": 320120 }, { "epoch": 27.73, "learning_rate": 5.6920030068231746e-06, "loss": 0.6059, "step": 320130 }, { "epoch": 27.73, "learning_rate": 5.6898346247253376e-06, "loss": 0.5603, "step": 320140 }, { "epoch": 27.73, "learning_rate": 5.6876662426275e-06, "loss": 0.6123, "step": 320150 }, { "epoch": 27.73, "learning_rate": 5.685497860529663e-06, "loss": 0.5525, "step": 320160 }, { "epoch": 27.73, "learning_rate": 5.683329478431826e-06, "loss": 0.5588, "step": 320170 }, { "epoch": 27.73, "learning_rate": 5.681161096333988e-06, "loss": 0.5814, "step": 320180 }, { "epoch": 27.73, "learning_rate": 5.678992714236151e-06, "loss": 0.6272, "step": 320190 }, { "epoch": 27.73, "learning_rate": 5.676824332138313e-06, "loss": 0.6317, "step": 320200 }, { "epoch": 27.73, "learning_rate": 5.674655950040476e-06, "loss": 0.4623, "step": 320210 }, { "epoch": 27.73, "learning_rate": 5.672487567942638e-06, "loss": 0.4435, "step": 320220 }, { "epoch": 27.73, "learning_rate": 5.670319185844801e-06, "loss": 0.5419, "step": 320230 }, { "epoch": 27.73, "learning_rate": 5.668150803746964e-06, "loss": 0.6451, "step": 320240 }, { "epoch": 27.74, "learning_rate": 5.665982421649126e-06, "loss": 0.7325, "step": 320250 }, { "epoch": 27.74, "learning_rate": 5.663814039551289e-06, "loss": 0.6194, "step": 320260 }, { "epoch": 27.74, "learning_rate": 5.6616456574534515e-06, "loss": 0.5569, "step": 320270 }, { "epoch": 27.74, "learning_rate": 5.6594772753556145e-06, "loss": 0.53, "step": 320280 }, { "epoch": 27.74, "learning_rate": 5.657308893257777e-06, "loss": 0.7512, "step": 320290 }, { "epoch": 27.74, "learning_rate": 5.65514051115994e-06, "loss": 0.5347, "step": 320300 }, { "epoch": 27.74, "learning_rate": 5.652972129062103e-06, "loss": 0.6562, "step": 320310 }, { "epoch": 27.74, "learning_rate": 5.650803746964265e-06, "loss": 0.4751, "step": 320320 }, { "epoch": 27.74, "learning_rate": 5.648635364866428e-06, "loss": 0.6153, "step": 320330 }, { "epoch": 27.74, "learning_rate": 5.646466982768589e-06, "loss": 0.6267, "step": 320340 }, { "epoch": 27.74, "learning_rate": 5.644298600670752e-06, "loss": 0.5639, "step": 320350 }, { "epoch": 27.75, "learning_rate": 5.642130218572914e-06, "loss": 0.6617, "step": 320360 }, { "epoch": 27.75, "learning_rate": 5.639961836475077e-06, "loss": 0.6267, "step": 320370 }, { "epoch": 27.75, "learning_rate": 5.6377934543772395e-06, "loss": 0.4953, "step": 320380 }, { "epoch": 27.75, "learning_rate": 5.6356250722794025e-06, "loss": 0.5939, "step": 320390 }, { "epoch": 27.75, "learning_rate": 5.6334566901815655e-06, "loss": 0.5713, "step": 320400 }, { "epoch": 27.75, "learning_rate": 5.631288308083728e-06, "loss": 0.619, "step": 320410 }, { "epoch": 27.75, "learning_rate": 5.629119925985891e-06, "loss": 0.6234, "step": 320420 }, { "epoch": 27.75, "learning_rate": 5.626951543888053e-06, "loss": 0.5273, "step": 320430 }, { "epoch": 27.75, "learning_rate": 5.624783161790216e-06, "loss": 0.5348, "step": 320440 }, { "epoch": 27.75, "learning_rate": 5.622614779692378e-06, "loss": 0.6259, "step": 320450 }, { "epoch": 27.75, "learning_rate": 5.620446397594541e-06, "loss": 0.6147, "step": 320460 }, { "epoch": 27.75, "learning_rate": 5.618278015496704e-06, "loss": 0.6733, "step": 320470 }, { "epoch": 27.76, "learning_rate": 5.616109633398866e-06, "loss": 0.5907, "step": 320480 }, { "epoch": 27.76, "learning_rate": 5.613941251301029e-06, "loss": 0.5722, "step": 320490 }, { "epoch": 27.76, "learning_rate": 5.611772869203191e-06, "loss": 0.6343, "step": 320500 }, { "epoch": 27.76, "learning_rate": 5.609604487105354e-06, "loss": 0.5182, "step": 320510 }, { "epoch": 27.76, "learning_rate": 5.6074361050075164e-06, "loss": 0.5604, "step": 320520 }, { "epoch": 27.76, "learning_rate": 5.6052677229096795e-06, "loss": 0.6553, "step": 320530 }, { "epoch": 27.76, "learning_rate": 5.6030993408118425e-06, "loss": 0.5361, "step": 320540 }, { "epoch": 27.76, "learning_rate": 5.600930958714005e-06, "loss": 0.5856, "step": 320550 }, { "epoch": 27.76, "learning_rate": 5.598762576616168e-06, "loss": 0.492, "step": 320560 }, { "epoch": 27.76, "learning_rate": 5.59659419451833e-06, "loss": 0.5959, "step": 320570 }, { "epoch": 27.76, "learning_rate": 5.594425812420493e-06, "loss": 0.6296, "step": 320580 }, { "epoch": 27.77, "learning_rate": 5.592257430322655e-06, "loss": 0.5674, "step": 320590 }, { "epoch": 27.77, "learning_rate": 5.590089048224817e-06, "loss": 0.5512, "step": 320600 }, { "epoch": 27.77, "learning_rate": 5.587920666126979e-06, "loss": 0.5223, "step": 320610 }, { "epoch": 27.77, "learning_rate": 5.585752284029142e-06, "loss": 0.5506, "step": 320620 }, { "epoch": 27.77, "learning_rate": 5.583583901931304e-06, "loss": 0.6583, "step": 320630 }, { "epoch": 27.77, "learning_rate": 5.581415519833467e-06, "loss": 0.5125, "step": 320640 }, { "epoch": 27.77, "learning_rate": 5.57924713773563e-06, "loss": 0.6804, "step": 320650 }, { "epoch": 27.77, "learning_rate": 5.5770787556377926e-06, "loss": 0.5697, "step": 320660 }, { "epoch": 27.77, "learning_rate": 5.5749103735399556e-06, "loss": 0.6988, "step": 320670 }, { "epoch": 27.77, "learning_rate": 5.572741991442118e-06, "loss": 0.5705, "step": 320680 }, { "epoch": 27.77, "learning_rate": 5.570573609344281e-06, "loss": 0.5023, "step": 320690 }, { "epoch": 27.77, "learning_rate": 5.568405227246443e-06, "loss": 0.6785, "step": 320700 }, { "epoch": 27.78, "learning_rate": 5.566236845148606e-06, "loss": 0.5669, "step": 320710 }, { "epoch": 27.78, "learning_rate": 5.564068463050769e-06, "loss": 0.582, "step": 320720 }, { "epoch": 27.78, "learning_rate": 5.561900080952931e-06, "loss": 0.5316, "step": 320730 }, { "epoch": 27.78, "learning_rate": 5.559731698855094e-06, "loss": 0.5544, "step": 320740 }, { "epoch": 27.78, "learning_rate": 5.557563316757256e-06, "loss": 0.5168, "step": 320750 }, { "epoch": 27.78, "learning_rate": 5.555394934659419e-06, "loss": 0.5648, "step": 320760 }, { "epoch": 27.78, "learning_rate": 5.553226552561581e-06, "loss": 0.4362, "step": 320770 }, { "epoch": 27.78, "learning_rate": 5.551058170463744e-06, "loss": 0.5814, "step": 320780 }, { "epoch": 27.78, "learning_rate": 5.548889788365907e-06, "loss": 0.6416, "step": 320790 }, { "epoch": 27.78, "learning_rate": 5.5467214062680695e-06, "loss": 0.5604, "step": 320800 }, { "epoch": 27.78, "learning_rate": 5.5445530241702325e-06, "loss": 0.6001, "step": 320810 }, { "epoch": 27.79, "learning_rate": 5.542384642072395e-06, "loss": 0.5617, "step": 320820 }, { "epoch": 27.79, "learning_rate": 5.540216259974558e-06, "loss": 0.6218, "step": 320830 }, { "epoch": 27.79, "learning_rate": 5.53804787787672e-06, "loss": 0.5788, "step": 320840 }, { "epoch": 27.79, "learning_rate": 5.535879495778882e-06, "loss": 0.5014, "step": 320850 }, { "epoch": 27.79, "learning_rate": 5.533711113681044e-06, "loss": 0.6121, "step": 320860 }, { "epoch": 27.79, "learning_rate": 5.531542731583207e-06, "loss": 0.4998, "step": 320870 }, { "epoch": 27.79, "learning_rate": 5.52937434948537e-06, "loss": 0.5744, "step": 320880 }, { "epoch": 27.79, "learning_rate": 5.527205967387532e-06, "loss": 0.6568, "step": 320890 }, { "epoch": 27.79, "learning_rate": 5.525037585289695e-06, "loss": 0.616, "step": 320900 }, { "epoch": 27.79, "learning_rate": 5.5228692031918575e-06, "loss": 0.6482, "step": 320910 }, { "epoch": 27.79, "learning_rate": 5.5207008210940205e-06, "loss": 0.5812, "step": 320920 }, { "epoch": 27.79, "learning_rate": 5.518532438996183e-06, "loss": 0.485, "step": 320930 }, { "epoch": 27.8, "learning_rate": 5.516364056898346e-06, "loss": 0.6553, "step": 320940 }, { "epoch": 27.8, "learning_rate": 5.514195674800509e-06, "loss": 0.6126, "step": 320950 }, { "epoch": 27.8, "learning_rate": 5.512027292702671e-06, "loss": 0.5167, "step": 320960 }, { "epoch": 27.8, "learning_rate": 5.509858910604834e-06, "loss": 0.5765, "step": 320970 }, { "epoch": 27.8, "learning_rate": 5.507690528506996e-06, "loss": 0.5221, "step": 320980 }, { "epoch": 27.8, "learning_rate": 5.505522146409159e-06, "loss": 0.7235, "step": 320990 }, { "epoch": 27.8, "learning_rate": 5.503353764311321e-06, "loss": 0.5687, "step": 321000 }, { "epoch": 27.8, "learning_rate": 5.501185382213484e-06, "loss": 0.5779, "step": 321010 }, { "epoch": 27.8, "learning_rate": 5.499017000115647e-06, "loss": 0.6996, "step": 321020 }, { "epoch": 27.8, "learning_rate": 5.496848618017809e-06, "loss": 0.655, "step": 321030 }, { "epoch": 27.8, "learning_rate": 5.494680235919972e-06, "loss": 0.6721, "step": 321040 }, { "epoch": 27.8, "learning_rate": 5.4925118538221345e-06, "loss": 0.5892, "step": 321050 }, { "epoch": 27.81, "learning_rate": 5.4903434717242975e-06, "loss": 0.6607, "step": 321060 }, { "epoch": 27.81, "learning_rate": 5.48817508962646e-06, "loss": 0.6143, "step": 321070 }, { "epoch": 27.81, "learning_rate": 5.486006707528623e-06, "loss": 0.5248, "step": 321080 }, { "epoch": 27.81, "learning_rate": 5.483838325430786e-06, "loss": 0.4926, "step": 321090 }, { "epoch": 27.81, "learning_rate": 5.481669943332948e-06, "loss": 0.575, "step": 321100 }, { "epoch": 27.81, "learning_rate": 5.479501561235109e-06, "loss": 0.5936, "step": 321110 }, { "epoch": 27.81, "learning_rate": 5.477333179137272e-06, "loss": 0.7487, "step": 321120 }, { "epoch": 27.81, "learning_rate": 5.475164797039435e-06, "loss": 0.5978, "step": 321130 }, { "epoch": 27.81, "learning_rate": 5.472996414941597e-06, "loss": 0.542, "step": 321140 }, { "epoch": 27.81, "learning_rate": 5.47082803284376e-06, "loss": 0.6957, "step": 321150 }, { "epoch": 27.81, "learning_rate": 5.468659650745922e-06, "loss": 0.6508, "step": 321160 }, { "epoch": 27.82, "learning_rate": 5.466491268648085e-06, "loss": 0.4845, "step": 321170 }, { "epoch": 27.82, "learning_rate": 5.4643228865502476e-06, "loss": 0.7836, "step": 321180 }, { "epoch": 27.82, "learning_rate": 5.462154504452411e-06, "loss": 0.4918, "step": 321190 }, { "epoch": 27.82, "learning_rate": 5.459986122354574e-06, "loss": 0.6269, "step": 321200 }, { "epoch": 27.82, "learning_rate": 5.457817740256736e-06, "loss": 0.552, "step": 321210 }, { "epoch": 27.82, "learning_rate": 5.455649358158899e-06, "loss": 0.6421, "step": 321220 }, { "epoch": 27.82, "learning_rate": 5.453480976061061e-06, "loss": 0.5028, "step": 321230 }, { "epoch": 27.82, "learning_rate": 5.451312593963224e-06, "loss": 0.5004, "step": 321240 }, { "epoch": 27.82, "learning_rate": 5.449144211865386e-06, "loss": 0.616, "step": 321250 }, { "epoch": 27.82, "learning_rate": 5.446975829767549e-06, "loss": 0.5091, "step": 321260 }, { "epoch": 27.82, "learning_rate": 5.444807447669712e-06, "loss": 0.6105, "step": 321270 }, { "epoch": 27.82, "learning_rate": 5.442639065571874e-06, "loss": 0.5469, "step": 321280 }, { "epoch": 27.83, "learning_rate": 5.440470683474037e-06, "loss": 0.5169, "step": 321290 }, { "epoch": 27.83, "learning_rate": 5.438302301376199e-06, "loss": 0.5366, "step": 321300 }, { "epoch": 27.83, "learning_rate": 5.436133919278362e-06, "loss": 0.421, "step": 321310 }, { "epoch": 27.83, "learning_rate": 5.4339655371805245e-06, "loss": 0.5439, "step": 321320 }, { "epoch": 27.83, "learning_rate": 5.4317971550826875e-06, "loss": 0.5285, "step": 321330 }, { "epoch": 27.83, "learning_rate": 5.4296287729848505e-06, "loss": 0.5643, "step": 321340 }, { "epoch": 27.83, "learning_rate": 5.427460390887013e-06, "loss": 0.5604, "step": 321350 }, { "epoch": 27.83, "learning_rate": 5.425292008789176e-06, "loss": 0.6767, "step": 321360 }, { "epoch": 27.83, "learning_rate": 5.423123626691337e-06, "loss": 0.6119, "step": 321370 }, { "epoch": 27.83, "learning_rate": 5.4209552445935e-06, "loss": 0.5691, "step": 321380 }, { "epoch": 27.83, "learning_rate": 5.418786862495662e-06, "loss": 0.6856, "step": 321390 }, { "epoch": 27.84, "learning_rate": 5.416618480397825e-06, "loss": 0.5818, "step": 321400 }, { "epoch": 27.84, "learning_rate": 5.414450098299987e-06, "loss": 0.7934, "step": 321410 }, { "epoch": 27.84, "learning_rate": 5.41228171620215e-06, "loss": 0.6727, "step": 321420 }, { "epoch": 27.84, "learning_rate": 5.4101133341043125e-06, "loss": 0.5548, "step": 321430 }, { "epoch": 27.84, "learning_rate": 5.4079449520064755e-06, "loss": 0.6576, "step": 321440 }, { "epoch": 27.84, "learning_rate": 5.4057765699086385e-06, "loss": 0.6076, "step": 321450 }, { "epoch": 27.84, "learning_rate": 5.403608187810801e-06, "loss": 0.5315, "step": 321460 }, { "epoch": 27.84, "learning_rate": 5.401439805712964e-06, "loss": 0.6286, "step": 321470 }, { "epoch": 27.84, "learning_rate": 5.399271423615126e-06, "loss": 0.4104, "step": 321480 }, { "epoch": 27.84, "learning_rate": 5.397103041517289e-06, "loss": 0.5898, "step": 321490 }, { "epoch": 27.84, "learning_rate": 5.394934659419451e-06, "loss": 0.567, "step": 321500 }, { "epoch": 27.84, "learning_rate": 5.392766277321614e-06, "loss": 0.5631, "step": 321510 }, { "epoch": 27.85, "learning_rate": 5.390597895223777e-06, "loss": 0.5185, "step": 321520 }, { "epoch": 27.85, "learning_rate": 5.388429513125939e-06, "loss": 0.5012, "step": 321530 }, { "epoch": 27.85, "learning_rate": 5.386261131028102e-06, "loss": 0.5066, "step": 321540 }, { "epoch": 27.85, "learning_rate": 5.384092748930264e-06, "loss": 0.4155, "step": 321550 }, { "epoch": 27.85, "learning_rate": 5.381924366832427e-06, "loss": 0.6434, "step": 321560 }, { "epoch": 27.85, "learning_rate": 5.3797559847345895e-06, "loss": 0.6503, "step": 321570 }, { "epoch": 27.85, "learning_rate": 5.3775876026367525e-06, "loss": 0.5995, "step": 321580 }, { "epoch": 27.85, "learning_rate": 5.3754192205389155e-06, "loss": 0.6079, "step": 321590 }, { "epoch": 27.85, "learning_rate": 5.373250838441078e-06, "loss": 0.693, "step": 321600 }, { "epoch": 27.85, "learning_rate": 5.371082456343241e-06, "loss": 0.5544, "step": 321610 }, { "epoch": 27.85, "learning_rate": 5.368914074245402e-06, "loss": 0.5202, "step": 321620 }, { "epoch": 27.86, "learning_rate": 5.366745692147565e-06, "loss": 0.6552, "step": 321630 }, { "epoch": 27.86, "learning_rate": 5.364577310049727e-06, "loss": 0.5605, "step": 321640 }, { "epoch": 27.86, "learning_rate": 5.36240892795189e-06, "loss": 0.5996, "step": 321650 }, { "epoch": 27.86, "learning_rate": 5.360240545854052e-06, "loss": 0.5024, "step": 321660 }, { "epoch": 27.86, "learning_rate": 5.358072163756215e-06, "loss": 0.6708, "step": 321670 }, { "epoch": 27.86, "learning_rate": 5.355903781658378e-06, "loss": 0.4928, "step": 321680 }, { "epoch": 27.86, "learning_rate": 5.3537353995605404e-06, "loss": 0.5054, "step": 321690 }, { "epoch": 27.86, "learning_rate": 5.3515670174627034e-06, "loss": 0.6301, "step": 321700 }, { "epoch": 27.86, "learning_rate": 5.349398635364866e-06, "loss": 0.5919, "step": 321710 }, { "epoch": 27.86, "learning_rate": 5.347230253267029e-06, "loss": 0.5363, "step": 321720 }, { "epoch": 27.86, "learning_rate": 5.345061871169191e-06, "loss": 0.5914, "step": 321730 }, { "epoch": 27.86, "learning_rate": 5.342893489071354e-06, "loss": 0.5921, "step": 321740 }, { "epoch": 27.87, "learning_rate": 5.340725106973517e-06, "loss": 0.6367, "step": 321750 }, { "epoch": 27.87, "learning_rate": 5.338556724875679e-06, "loss": 0.6315, "step": 321760 }, { "epoch": 27.87, "learning_rate": 5.336388342777842e-06, "loss": 0.5592, "step": 321770 }, { "epoch": 27.87, "learning_rate": 5.334219960680004e-06, "loss": 0.7108, "step": 321780 }, { "epoch": 27.87, "learning_rate": 5.332051578582167e-06, "loss": 0.4786, "step": 321790 }, { "epoch": 27.87, "learning_rate": 5.329883196484329e-06, "loss": 0.5821, "step": 321800 }, { "epoch": 27.87, "learning_rate": 5.327714814386492e-06, "loss": 0.593, "step": 321810 }, { "epoch": 27.87, "learning_rate": 5.325546432288655e-06, "loss": 0.567, "step": 321820 }, { "epoch": 27.87, "learning_rate": 5.323378050190817e-06, "loss": 0.5518, "step": 321830 }, { "epoch": 27.87, "learning_rate": 5.32120966809298e-06, "loss": 0.5718, "step": 321840 }, { "epoch": 27.87, "learning_rate": 5.3190412859951426e-06, "loss": 0.6096, "step": 321850 }, { "epoch": 27.88, "learning_rate": 5.3168729038973056e-06, "loss": 0.5655, "step": 321860 }, { "epoch": 27.88, "learning_rate": 5.314704521799468e-06, "loss": 0.5098, "step": 321870 }, { "epoch": 27.88, "learning_rate": 5.31253613970163e-06, "loss": 0.5276, "step": 321880 }, { "epoch": 27.88, "learning_rate": 5.310367757603792e-06, "loss": 0.677, "step": 321890 }, { "epoch": 27.88, "learning_rate": 5.308199375505955e-06, "loss": 0.7181, "step": 321900 }, { "epoch": 27.88, "learning_rate": 5.306030993408117e-06, "loss": 0.5403, "step": 321910 }, { "epoch": 27.88, "learning_rate": 5.30386261131028e-06, "loss": 0.5439, "step": 321920 }, { "epoch": 27.88, "learning_rate": 5.301694229212443e-06, "loss": 0.6831, "step": 321930 }, { "epoch": 27.88, "learning_rate": 5.299525847114605e-06, "loss": 0.5836, "step": 321940 }, { "epoch": 27.88, "learning_rate": 5.297357465016768e-06, "loss": 0.5683, "step": 321950 }, { "epoch": 27.88, "learning_rate": 5.2951890829189305e-06, "loss": 0.5971, "step": 321960 }, { "epoch": 27.88, "learning_rate": 5.2930207008210935e-06, "loss": 0.6998, "step": 321970 }, { "epoch": 27.89, "learning_rate": 5.290852318723256e-06, "loss": 0.5237, "step": 321980 }, { "epoch": 27.89, "learning_rate": 5.288683936625419e-06, "loss": 0.5152, "step": 321990 }, { "epoch": 27.89, "learning_rate": 5.286515554527582e-06, "loss": 0.6891, "step": 322000 }, { "epoch": 27.89, "learning_rate": 5.284347172429744e-06, "loss": 0.6417, "step": 322010 }, { "epoch": 27.89, "learning_rate": 5.282178790331907e-06, "loss": 0.643, "step": 322020 }, { "epoch": 27.89, "learning_rate": 5.280010408234069e-06, "loss": 0.5908, "step": 322030 }, { "epoch": 27.89, "learning_rate": 5.277842026136232e-06, "loss": 0.644, "step": 322040 }, { "epoch": 27.89, "learning_rate": 5.275673644038394e-06, "loss": 0.5782, "step": 322050 }, { "epoch": 27.89, "learning_rate": 5.273505261940557e-06, "loss": 0.5876, "step": 322060 }, { "epoch": 27.89, "learning_rate": 5.27133687984272e-06, "loss": 0.7329, "step": 322070 }, { "epoch": 27.89, "learning_rate": 5.269168497744882e-06, "loss": 0.599, "step": 322080 }, { "epoch": 27.9, "learning_rate": 5.267000115647045e-06, "loss": 0.5677, "step": 322090 }, { "epoch": 27.9, "learning_rate": 5.2648317335492075e-06, "loss": 0.5519, "step": 322100 }, { "epoch": 27.9, "learning_rate": 5.2626633514513705e-06, "loss": 0.4996, "step": 322110 }, { "epoch": 27.9, "learning_rate": 5.260494969353533e-06, "loss": 0.6427, "step": 322120 }, { "epoch": 27.9, "learning_rate": 5.258326587255696e-06, "loss": 0.6456, "step": 322130 }, { "epoch": 27.9, "learning_rate": 5.256158205157857e-06, "loss": 0.644, "step": 322140 }, { "epoch": 27.9, "learning_rate": 5.25398982306002e-06, "loss": 0.5563, "step": 322150 }, { "epoch": 27.9, "learning_rate": 5.251821440962183e-06, "loss": 0.5253, "step": 322160 }, { "epoch": 27.9, "learning_rate": 5.249653058864345e-06, "loss": 0.6925, "step": 322170 }, { "epoch": 27.9, "learning_rate": 5.247484676766508e-06, "loss": 0.5753, "step": 322180 }, { "epoch": 27.9, "learning_rate": 5.24531629466867e-06, "loss": 0.5584, "step": 322190 }, { "epoch": 27.9, "learning_rate": 5.243147912570833e-06, "loss": 0.4799, "step": 322200 }, { "epoch": 27.91, "learning_rate": 5.2409795304729954e-06, "loss": 0.417, "step": 322210 }, { "epoch": 27.91, "learning_rate": 5.2388111483751584e-06, "loss": 0.5366, "step": 322220 }, { "epoch": 27.91, "learning_rate": 5.2366427662773214e-06, "loss": 0.5558, "step": 322230 }, { "epoch": 27.91, "learning_rate": 5.234474384179484e-06, "loss": 0.5732, "step": 322240 }, { "epoch": 27.91, "learning_rate": 5.232306002081647e-06, "loss": 0.633, "step": 322250 }, { "epoch": 27.91, "learning_rate": 5.230137619983809e-06, "loss": 0.6088, "step": 322260 }, { "epoch": 27.91, "learning_rate": 5.227969237885972e-06, "loss": 0.5706, "step": 322270 }, { "epoch": 27.91, "learning_rate": 5.225800855788134e-06, "loss": 0.5752, "step": 322280 }, { "epoch": 27.91, "learning_rate": 5.223632473690297e-06, "loss": 0.5645, "step": 322290 }, { "epoch": 27.91, "learning_rate": 5.22146409159246e-06, "loss": 0.6915, "step": 322300 }, { "epoch": 27.91, "learning_rate": 5.219295709494622e-06, "loss": 0.6664, "step": 322310 }, { "epoch": 27.91, "learning_rate": 5.217127327396785e-06, "loss": 0.4876, "step": 322320 }, { "epoch": 27.92, "learning_rate": 5.214958945298947e-06, "loss": 0.5366, "step": 322330 }, { "epoch": 27.92, "learning_rate": 5.21279056320111e-06, "loss": 0.5221, "step": 322340 }, { "epoch": 27.92, "learning_rate": 5.210622181103272e-06, "loss": 0.5043, "step": 322350 }, { "epoch": 27.92, "learning_rate": 5.208453799005435e-06, "loss": 0.6294, "step": 322360 }, { "epoch": 27.92, "learning_rate": 5.206285416907598e-06, "loss": 0.7165, "step": 322370 }, { "epoch": 27.92, "learning_rate": 5.2041170348097606e-06, "loss": 0.629, "step": 322380 }, { "epoch": 27.92, "learning_rate": 5.201948652711922e-06, "loss": 0.5086, "step": 322390 }, { "epoch": 27.92, "learning_rate": 5.199780270614085e-06, "loss": 0.553, "step": 322400 }, { "epoch": 27.92, "learning_rate": 5.197611888516248e-06, "loss": 0.6243, "step": 322410 }, { "epoch": 27.92, "learning_rate": 5.19544350641841e-06, "loss": 0.5684, "step": 322420 }, { "epoch": 27.92, "learning_rate": 5.193275124320573e-06, "loss": 0.5405, "step": 322430 }, { "epoch": 27.93, "learning_rate": 5.191106742222735e-06, "loss": 0.5997, "step": 322440 }, { "epoch": 27.93, "learning_rate": 5.188938360124898e-06, "loss": 0.5845, "step": 322450 }, { "epoch": 27.93, "learning_rate": 5.18676997802706e-06, "loss": 0.6476, "step": 322460 }, { "epoch": 27.93, "learning_rate": 5.184601595929223e-06, "loss": 0.6438, "step": 322470 }, { "epoch": 27.93, "learning_rate": 5.182433213831386e-06, "loss": 0.616, "step": 322480 }, { "epoch": 27.93, "learning_rate": 5.1802648317335485e-06, "loss": 0.6343, "step": 322490 }, { "epoch": 27.93, "learning_rate": 5.1780964496357115e-06, "loss": 0.6591, "step": 322500 }, { "epoch": 27.93, "learning_rate": 5.175928067537874e-06, "loss": 0.5631, "step": 322510 }, { "epoch": 27.93, "learning_rate": 5.173759685440037e-06, "loss": 0.5804, "step": 322520 }, { "epoch": 27.93, "learning_rate": 5.171591303342199e-06, "loss": 0.6563, "step": 322530 }, { "epoch": 27.93, "learning_rate": 5.169422921244362e-06, "loss": 0.5284, "step": 322540 }, { "epoch": 27.93, "learning_rate": 5.167254539146525e-06, "loss": 0.3973, "step": 322550 }, { "epoch": 27.94, "learning_rate": 5.165086157048687e-06, "loss": 0.6433, "step": 322560 }, { "epoch": 27.94, "learning_rate": 5.16291777495085e-06, "loss": 0.6016, "step": 322570 }, { "epoch": 27.94, "learning_rate": 5.160749392853012e-06, "loss": 0.4436, "step": 322580 }, { "epoch": 27.94, "learning_rate": 5.158581010755175e-06, "loss": 0.7006, "step": 322590 }, { "epoch": 27.94, "learning_rate": 5.156412628657337e-06, "loss": 0.4458, "step": 322600 }, { "epoch": 27.94, "learning_rate": 5.1542442465595e-06, "loss": 0.6399, "step": 322610 }, { "epoch": 27.94, "learning_rate": 5.152075864461663e-06, "loss": 0.5739, "step": 322620 }, { "epoch": 27.94, "learning_rate": 5.1499074823638255e-06, "loss": 0.624, "step": 322630 }, { "epoch": 27.94, "learning_rate": 5.1477391002659885e-06, "loss": 0.5591, "step": 322640 }, { "epoch": 27.94, "learning_rate": 5.14557071816815e-06, "loss": 0.634, "step": 322650 }, { "epoch": 27.94, "learning_rate": 5.143402336070313e-06, "loss": 0.5904, "step": 322660 }, { "epoch": 27.95, "learning_rate": 5.141233953972475e-06, "loss": 0.5322, "step": 322670 }, { "epoch": 27.95, "learning_rate": 5.139065571874638e-06, "loss": 0.6314, "step": 322680 }, { "epoch": 27.95, "learning_rate": 5.1368971897768e-06, "loss": 0.5715, "step": 322690 }, { "epoch": 27.95, "learning_rate": 5.134728807678963e-06, "loss": 0.53, "step": 322700 }, { "epoch": 27.95, "learning_rate": 5.132560425581125e-06, "loss": 0.549, "step": 322710 }, { "epoch": 27.95, "learning_rate": 5.130392043483288e-06, "loss": 0.7639, "step": 322720 }, { "epoch": 27.95, "learning_rate": 5.128223661385451e-06, "loss": 0.6629, "step": 322730 }, { "epoch": 27.95, "learning_rate": 5.1260552792876134e-06, "loss": 0.5298, "step": 322740 }, { "epoch": 27.95, "learning_rate": 5.1238868971897764e-06, "loss": 0.6314, "step": 322750 }, { "epoch": 27.95, "learning_rate": 5.121718515091939e-06, "loss": 0.6912, "step": 322760 }, { "epoch": 27.95, "learning_rate": 5.119550132994102e-06, "loss": 0.6378, "step": 322770 }, { "epoch": 27.95, "learning_rate": 5.117381750896264e-06, "loss": 0.5087, "step": 322780 }, { "epoch": 27.96, "learning_rate": 5.115213368798427e-06, "loss": 0.6215, "step": 322790 }, { "epoch": 27.96, "learning_rate": 5.11304498670059e-06, "loss": 0.5986, "step": 322800 }, { "epoch": 27.96, "learning_rate": 5.110876604602752e-06, "loss": 0.6466, "step": 322810 }, { "epoch": 27.96, "learning_rate": 5.108708222504915e-06, "loss": 0.5546, "step": 322820 }, { "epoch": 27.96, "learning_rate": 5.106539840407077e-06, "loss": 0.6128, "step": 322830 }, { "epoch": 27.96, "learning_rate": 5.10437145830924e-06, "loss": 0.5453, "step": 322840 }, { "epoch": 27.96, "learning_rate": 5.102203076211402e-06, "loss": 0.7357, "step": 322850 }, { "epoch": 27.96, "learning_rate": 5.100034694113565e-06, "loss": 0.5605, "step": 322860 }, { "epoch": 27.96, "learning_rate": 5.097866312015728e-06, "loss": 0.5546, "step": 322870 }, { "epoch": 27.96, "learning_rate": 5.09569792991789e-06, "loss": 0.6911, "step": 322880 }, { "epoch": 27.96, "learning_rate": 5.093529547820053e-06, "loss": 0.5234, "step": 322890 }, { "epoch": 27.97, "learning_rate": 5.0913611657222156e-06, "loss": 0.5788, "step": 322900 }, { "epoch": 27.97, "learning_rate": 5.089192783624378e-06, "loss": 0.5697, "step": 322910 }, { "epoch": 27.97, "learning_rate": 5.08702440152654e-06, "loss": 0.527, "step": 322920 }, { "epoch": 27.97, "learning_rate": 5.084856019428703e-06, "loss": 0.5291, "step": 322930 }, { "epoch": 27.97, "learning_rate": 5.082687637330865e-06, "loss": 0.5026, "step": 322940 }, { "epoch": 27.97, "learning_rate": 5.080519255233028e-06, "loss": 0.6759, "step": 322950 }, { "epoch": 27.97, "learning_rate": 5.078350873135191e-06, "loss": 0.6252, "step": 322960 }, { "epoch": 27.97, "learning_rate": 5.076182491037353e-06, "loss": 0.5593, "step": 322970 }, { "epoch": 27.97, "learning_rate": 5.074014108939516e-06, "loss": 0.7354, "step": 322980 }, { "epoch": 27.97, "learning_rate": 5.071845726841678e-06, "loss": 0.6397, "step": 322990 }, { "epoch": 27.97, "learning_rate": 5.069677344743841e-06, "loss": 0.6643, "step": 323000 }, { "epoch": 27.97, "learning_rate": 5.0675089626460035e-06, "loss": 0.5359, "step": 323010 }, { "epoch": 27.98, "learning_rate": 5.0653405805481665e-06, "loss": 0.6307, "step": 323020 }, { "epoch": 27.98, "learning_rate": 5.0631721984503295e-06, "loss": 0.5262, "step": 323030 }, { "epoch": 27.98, "learning_rate": 5.061003816352492e-06, "loss": 0.6419, "step": 323040 }, { "epoch": 27.98, "learning_rate": 5.058835434254655e-06, "loss": 0.5727, "step": 323050 }, { "epoch": 27.98, "learning_rate": 5.056667052156817e-06, "loss": 0.6032, "step": 323060 }, { "epoch": 27.98, "learning_rate": 5.05449867005898e-06, "loss": 0.5981, "step": 323070 }, { "epoch": 27.98, "learning_rate": 5.052330287961142e-06, "loss": 0.6635, "step": 323080 }, { "epoch": 27.98, "learning_rate": 5.050161905863305e-06, "loss": 0.508, "step": 323090 }, { "epoch": 27.98, "learning_rate": 5.047993523765468e-06, "loss": 0.6063, "step": 323100 }, { "epoch": 27.98, "learning_rate": 5.04582514166763e-06, "loss": 0.5331, "step": 323110 }, { "epoch": 27.98, "learning_rate": 5.043656759569793e-06, "loss": 0.5898, "step": 323120 }, { "epoch": 27.99, "learning_rate": 5.041488377471955e-06, "loss": 0.6658, "step": 323130 }, { "epoch": 27.99, "learning_rate": 5.039319995374118e-06, "loss": 0.6733, "step": 323140 }, { "epoch": 27.99, "learning_rate": 5.0371516132762805e-06, "loss": 0.5642, "step": 323150 }, { "epoch": 27.99, "learning_rate": 5.034983231178443e-06, "loss": 0.6083, "step": 323160 }, { "epoch": 27.99, "learning_rate": 5.032814849080605e-06, "loss": 0.6144, "step": 323170 }, { "epoch": 27.99, "learning_rate": 5.030646466982768e-06, "loss": 0.7332, "step": 323180 }, { "epoch": 27.99, "learning_rate": 5.02847808488493e-06, "loss": 0.5752, "step": 323190 }, { "epoch": 27.99, "learning_rate": 5.026309702787093e-06, "loss": 0.6224, "step": 323200 }, { "epoch": 27.99, "learning_rate": 5.024141320689256e-06, "loss": 0.531, "step": 323210 }, { "epoch": 27.99, "learning_rate": 5.021972938591418e-06, "loss": 0.5556, "step": 323220 }, { "epoch": 27.99, "learning_rate": 5.019804556493581e-06, "loss": 0.5428, "step": 323230 }, { "epoch": 27.99, "learning_rate": 5.017636174395743e-06, "loss": 0.6843, "step": 323240 }, { "epoch": 28.0, "learning_rate": 5.015467792297906e-06, "loss": 0.5837, "step": 323250 }, { "epoch": 28.0, "learning_rate": 5.0132994102000685e-06, "loss": 0.5148, "step": 323260 }, { "epoch": 28.0, "learning_rate": 5.0111310281022315e-06, "loss": 0.5376, "step": 323270 }, { "epoch": 28.0, "learning_rate": 5.0089626460043945e-06, "loss": 0.5862, "step": 323280 }, { "epoch": 28.0, "learning_rate": 5.006794263906557e-06, "loss": 0.5869, "step": 323290 }, { "epoch": 28.0, "learning_rate": 5.00462588180872e-06, "loss": 0.6181, "step": 323300 }, { "epoch": 28.0, "eval_cer": 0.5321049755433322, "eval_em": 0.06498194945848375, "eval_f1": 0.06498194945848375, "eval_loss": 1.5841025114059448, "eval_runtime": 222.0377, "eval_samples_per_second": 8.733, "eval_steps_per_second": 1.094, "eval_wer": 0.9350180505415162, "step": 323302 }, { "epoch": 28.0, "learning_rate": 5.002457499710882e-06, "loss": 0.5852, "step": 323310 }, { "epoch": 28.0, "learning_rate": 5.000289117613045e-06, "loss": 0.574, "step": 323320 }, { "epoch": 28.0, "learning_rate": 4.998120735515207e-06, "loss": 0.6541, "step": 323330 }, { "epoch": 28.0, "learning_rate": 4.99595235341737e-06, "loss": 0.554, "step": 323340 }, { "epoch": 28.0, "learning_rate": 4.993783971319533e-06, "loss": 0.6459, "step": 323350 }, { "epoch": 28.01, "learning_rate": 4.991615589221695e-06, "loss": 0.5427, "step": 323360 }, { "epoch": 28.01, "learning_rate": 4.989447207123858e-06, "loss": 0.4746, "step": 323370 }, { "epoch": 28.01, "learning_rate": 4.98727882502602e-06, "loss": 0.5411, "step": 323380 }, { "epoch": 28.01, "learning_rate": 4.985110442928183e-06, "loss": 0.5487, "step": 323390 }, { "epoch": 28.01, "learning_rate": 4.982942060830345e-06, "loss": 0.5491, "step": 323400 }, { "epoch": 28.01, "learning_rate": 4.980773678732508e-06, "loss": 0.6882, "step": 323410 }, { "epoch": 28.01, "learning_rate": 4.97860529663467e-06, "loss": 0.7457, "step": 323420 }, { "epoch": 28.01, "learning_rate": 4.976436914536833e-06, "loss": 0.5551, "step": 323430 }, { "epoch": 28.01, "learning_rate": 4.974268532438996e-06, "loss": 0.5041, "step": 323440 }, { "epoch": 28.01, "learning_rate": 4.972100150341158e-06, "loss": 0.655, "step": 323450 }, { "epoch": 28.01, "learning_rate": 4.969931768243321e-06, "loss": 0.534, "step": 323460 }, { "epoch": 28.01, "learning_rate": 4.967763386145483e-06, "loss": 0.5176, "step": 323470 }, { "epoch": 28.02, "learning_rate": 4.965595004047646e-06, "loss": 0.541, "step": 323480 }, { "epoch": 28.02, "learning_rate": 4.963426621949808e-06, "loss": 0.7328, "step": 323490 }, { "epoch": 28.02, "learning_rate": 4.961258239851971e-06, "loss": 0.5261, "step": 323500 }, { "epoch": 28.02, "learning_rate": 4.959089857754134e-06, "loss": 0.6651, "step": 323510 }, { "epoch": 28.02, "learning_rate": 4.956921475656296e-06, "loss": 0.4994, "step": 323520 }, { "epoch": 28.02, "learning_rate": 4.954753093558459e-06, "loss": 0.5713, "step": 323530 }, { "epoch": 28.02, "learning_rate": 4.9525847114606215e-06, "loss": 0.5039, "step": 323540 }, { "epoch": 28.02, "learning_rate": 4.9504163293627845e-06, "loss": 0.5488, "step": 323550 }, { "epoch": 28.02, "learning_rate": 4.948247947264947e-06, "loss": 0.5424, "step": 323560 }, { "epoch": 28.02, "learning_rate": 4.94607956516711e-06, "loss": 0.585, "step": 323570 }, { "epoch": 28.02, "learning_rate": 4.943911183069273e-06, "loss": 0.6592, "step": 323580 }, { "epoch": 28.02, "learning_rate": 4.941742800971435e-06, "loss": 0.6536, "step": 323590 }, { "epoch": 28.03, "learning_rate": 4.939574418873598e-06, "loss": 0.6524, "step": 323600 }, { "epoch": 28.03, "learning_rate": 4.93740603677576e-06, "loss": 0.5406, "step": 323610 }, { "epoch": 28.03, "learning_rate": 4.935237654677923e-06, "loss": 0.6172, "step": 323620 }, { "epoch": 28.03, "learning_rate": 4.933069272580085e-06, "loss": 0.5368, "step": 323630 }, { "epoch": 28.03, "learning_rate": 4.930900890482248e-06, "loss": 0.6035, "step": 323640 }, { "epoch": 28.03, "learning_rate": 4.928732508384411e-06, "loss": 0.4438, "step": 323650 }, { "epoch": 28.03, "learning_rate": 4.926564126286573e-06, "loss": 0.6829, "step": 323660 }, { "epoch": 28.03, "learning_rate": 4.924395744188736e-06, "loss": 0.5049, "step": 323670 }, { "epoch": 28.03, "learning_rate": 4.922227362090898e-06, "loss": 0.5447, "step": 323680 }, { "epoch": 28.03, "learning_rate": 4.920058979993061e-06, "loss": 0.6368, "step": 323690 }, { "epoch": 28.03, "learning_rate": 4.917890597895223e-06, "loss": 0.5085, "step": 323700 }, { "epoch": 28.04, "learning_rate": 4.915722215797386e-06, "loss": 0.5573, "step": 323710 }, { "epoch": 28.04, "learning_rate": 4.913553833699548e-06, "loss": 0.5337, "step": 323720 }, { "epoch": 28.04, "learning_rate": 4.911385451601711e-06, "loss": 0.5784, "step": 323730 }, { "epoch": 28.04, "learning_rate": 4.909217069503873e-06, "loss": 0.5368, "step": 323740 }, { "epoch": 28.04, "learning_rate": 4.907048687406036e-06, "loss": 0.5552, "step": 323750 }, { "epoch": 28.04, "learning_rate": 4.904880305308199e-06, "loss": 0.594, "step": 323760 }, { "epoch": 28.04, "learning_rate": 4.902711923210361e-06, "loss": 0.6273, "step": 323770 }, { "epoch": 28.04, "learning_rate": 4.900543541112524e-06, "loss": 0.5153, "step": 323780 }, { "epoch": 28.04, "learning_rate": 4.8983751590146865e-06, "loss": 0.7497, "step": 323790 }, { "epoch": 28.04, "learning_rate": 4.8962067769168495e-06, "loss": 0.5864, "step": 323800 }, { "epoch": 28.04, "learning_rate": 4.894038394819012e-06, "loss": 0.5625, "step": 323810 }, { "epoch": 28.04, "learning_rate": 4.891870012721175e-06, "loss": 0.5978, "step": 323820 }, { "epoch": 28.05, "learning_rate": 4.889701630623338e-06, "loss": 0.6404, "step": 323830 }, { "epoch": 28.05, "learning_rate": 4.8875332485255e-06, "loss": 0.7593, "step": 323840 }, { "epoch": 28.05, "learning_rate": 4.885364866427663e-06, "loss": 0.5204, "step": 323850 }, { "epoch": 28.05, "learning_rate": 4.883196484329825e-06, "loss": 0.665, "step": 323860 }, { "epoch": 28.05, "learning_rate": 4.881028102231988e-06, "loss": 0.529, "step": 323870 }, { "epoch": 28.05, "learning_rate": 4.87885972013415e-06, "loss": 0.6941, "step": 323880 }, { "epoch": 28.05, "learning_rate": 4.876691338036313e-06, "loss": 0.5966, "step": 323890 }, { "epoch": 28.05, "learning_rate": 4.874522955938476e-06, "loss": 0.5643, "step": 323900 }, { "epoch": 28.05, "learning_rate": 4.872354573840638e-06, "loss": 0.561, "step": 323910 }, { "epoch": 28.05, "learning_rate": 4.870186191742801e-06, "loss": 0.5498, "step": 323920 }, { "epoch": 28.05, "learning_rate": 4.868017809644963e-06, "loss": 0.5089, "step": 323930 }, { "epoch": 28.06, "learning_rate": 4.865849427547126e-06, "loss": 0.6051, "step": 323940 }, { "epoch": 28.06, "learning_rate": 4.863681045449288e-06, "loss": 0.5609, "step": 323950 }, { "epoch": 28.06, "learning_rate": 4.861512663351451e-06, "loss": 0.6257, "step": 323960 }, { "epoch": 28.06, "learning_rate": 4.859344281253613e-06, "loss": 0.5753, "step": 323970 }, { "epoch": 28.06, "learning_rate": 4.857175899155776e-06, "loss": 0.6289, "step": 323980 }, { "epoch": 28.06, "learning_rate": 4.855007517057938e-06, "loss": 0.6367, "step": 323990 }, { "epoch": 28.06, "learning_rate": 4.852839134960101e-06, "loss": 0.5403, "step": 324000 }, { "epoch": 28.06, "learning_rate": 4.850670752862264e-06, "loss": 0.5678, "step": 324010 }, { "epoch": 28.06, "learning_rate": 4.848502370764426e-06, "loss": 0.4346, "step": 324020 }, { "epoch": 28.06, "learning_rate": 4.846333988666589e-06, "loss": 0.6616, "step": 324030 }, { "epoch": 28.06, "learning_rate": 4.844165606568751e-06, "loss": 0.4838, "step": 324040 }, { "epoch": 28.06, "learning_rate": 4.841997224470914e-06, "loss": 0.4913, "step": 324050 }, { "epoch": 28.07, "learning_rate": 4.8398288423730765e-06, "loss": 0.5803, "step": 324060 }, { "epoch": 28.07, "learning_rate": 4.8376604602752395e-06, "loss": 0.5738, "step": 324070 }, { "epoch": 28.07, "learning_rate": 4.8354920781774026e-06, "loss": 0.5622, "step": 324080 }, { "epoch": 28.07, "learning_rate": 4.833323696079565e-06, "loss": 0.7397, "step": 324090 }, { "epoch": 28.07, "learning_rate": 4.831155313981728e-06, "loss": 0.5094, "step": 324100 }, { "epoch": 28.07, "learning_rate": 4.82898693188389e-06, "loss": 0.5146, "step": 324110 }, { "epoch": 28.07, "learning_rate": 4.826818549786053e-06, "loss": 0.5795, "step": 324120 }, { "epoch": 28.07, "learning_rate": 4.824650167688215e-06, "loss": 0.6633, "step": 324130 }, { "epoch": 28.07, "learning_rate": 4.822481785590378e-06, "loss": 0.5042, "step": 324140 }, { "epoch": 28.07, "learning_rate": 4.820313403492541e-06, "loss": 0.5876, "step": 324150 }, { "epoch": 28.07, "learning_rate": 4.818145021394703e-06, "loss": 0.5947, "step": 324160 }, { "epoch": 28.08, "learning_rate": 4.815976639296866e-06, "loss": 0.6238, "step": 324170 }, { "epoch": 28.08, "learning_rate": 4.813808257199028e-06, "loss": 0.5019, "step": 324180 }, { "epoch": 28.08, "learning_rate": 4.8116398751011905e-06, "loss": 0.6492, "step": 324190 }, { "epoch": 28.08, "learning_rate": 4.809471493003353e-06, "loss": 0.6619, "step": 324200 }, { "epoch": 28.08, "learning_rate": 4.807303110905516e-06, "loss": 0.5586, "step": 324210 }, { "epoch": 28.08, "learning_rate": 4.805134728807678e-06, "loss": 0.5433, "step": 324220 }, { "epoch": 28.08, "learning_rate": 4.802966346709841e-06, "loss": 0.4675, "step": 324230 }, { "epoch": 28.08, "learning_rate": 4.800797964612004e-06, "loss": 0.5492, "step": 324240 }, { "epoch": 28.08, "learning_rate": 4.798629582514166e-06, "loss": 0.6308, "step": 324250 }, { "epoch": 28.08, "learning_rate": 4.796461200416329e-06, "loss": 0.5915, "step": 324260 }, { "epoch": 28.08, "learning_rate": 4.794292818318491e-06, "loss": 0.7548, "step": 324270 }, { "epoch": 28.08, "learning_rate": 4.792124436220654e-06, "loss": 0.5114, "step": 324280 }, { "epoch": 28.09, "learning_rate": 4.789956054122816e-06, "loss": 0.5693, "step": 324290 }, { "epoch": 28.09, "learning_rate": 4.787787672024979e-06, "loss": 0.5272, "step": 324300 }, { "epoch": 28.09, "learning_rate": 4.785619289927142e-06, "loss": 0.565, "step": 324310 }, { "epoch": 28.09, "learning_rate": 4.7834509078293045e-06, "loss": 0.6223, "step": 324320 }, { "epoch": 28.09, "learning_rate": 4.7812825257314675e-06, "loss": 0.5427, "step": 324330 }, { "epoch": 28.09, "learning_rate": 4.77911414363363e-06, "loss": 0.605, "step": 324340 }, { "epoch": 28.09, "learning_rate": 4.776945761535793e-06, "loss": 0.6309, "step": 324350 }, { "epoch": 28.09, "learning_rate": 4.774777379437955e-06, "loss": 0.6931, "step": 324360 }, { "epoch": 28.09, "learning_rate": 4.772608997340118e-06, "loss": 0.4808, "step": 324370 }, { "epoch": 28.09, "learning_rate": 4.770440615242281e-06, "loss": 0.7171, "step": 324380 }, { "epoch": 28.09, "learning_rate": 4.768272233144443e-06, "loss": 0.6144, "step": 324390 }, { "epoch": 28.1, "learning_rate": 4.766103851046606e-06, "loss": 0.6245, "step": 324400 }, { "epoch": 28.1, "learning_rate": 4.763935468948768e-06, "loss": 0.5014, "step": 324410 }, { "epoch": 28.1, "learning_rate": 4.761767086850931e-06, "loss": 0.6274, "step": 324420 }, { "epoch": 28.1, "learning_rate": 4.759598704753093e-06, "loss": 0.5343, "step": 324430 }, { "epoch": 28.1, "learning_rate": 4.757430322655256e-06, "loss": 0.5552, "step": 324440 }, { "epoch": 28.1, "learning_rate": 4.755261940557418e-06, "loss": 0.5209, "step": 324450 }, { "epoch": 28.1, "learning_rate": 4.753093558459581e-06, "loss": 0.539, "step": 324460 }, { "epoch": 28.1, "learning_rate": 4.750925176361743e-06, "loss": 0.4701, "step": 324470 }, { "epoch": 28.1, "learning_rate": 4.748756794263906e-06, "loss": 0.6784, "step": 324480 }, { "epoch": 28.1, "learning_rate": 4.746588412166069e-06, "loss": 0.5329, "step": 324490 }, { "epoch": 28.1, "learning_rate": 4.744420030068231e-06, "loss": 0.5564, "step": 324500 }, { "epoch": 28.1, "learning_rate": 4.742251647970394e-06, "loss": 0.6503, "step": 324510 }, { "epoch": 28.11, "learning_rate": 4.740083265872556e-06, "loss": 0.5744, "step": 324520 }, { "epoch": 28.11, "learning_rate": 4.737914883774719e-06, "loss": 0.6689, "step": 324530 }, { "epoch": 28.11, "learning_rate": 4.735746501676881e-06, "loss": 0.5369, "step": 324540 }, { "epoch": 28.11, "learning_rate": 4.733578119579044e-06, "loss": 0.4755, "step": 324550 }, { "epoch": 28.11, "learning_rate": 4.731409737481207e-06, "loss": 0.5579, "step": 324560 }, { "epoch": 28.11, "learning_rate": 4.729241355383369e-06, "loss": 0.5033, "step": 324570 }, { "epoch": 28.11, "learning_rate": 4.727072973285532e-06, "loss": 0.5761, "step": 324580 }, { "epoch": 28.11, "learning_rate": 4.7249045911876946e-06, "loss": 0.514, "step": 324590 }, { "epoch": 28.11, "learning_rate": 4.7227362090898576e-06, "loss": 0.5198, "step": 324600 }, { "epoch": 28.11, "learning_rate": 4.72056782699202e-06, "loss": 0.5795, "step": 324610 }, { "epoch": 28.11, "learning_rate": 4.718399444894183e-06, "loss": 0.6042, "step": 324620 }, { "epoch": 28.12, "learning_rate": 4.716231062796346e-06, "loss": 0.6879, "step": 324630 }, { "epoch": 28.12, "learning_rate": 4.714062680698508e-06, "loss": 0.498, "step": 324640 }, { "epoch": 28.12, "learning_rate": 4.711894298600671e-06, "loss": 0.6747, "step": 324650 }, { "epoch": 28.12, "learning_rate": 4.709725916502833e-06, "loss": 0.5385, "step": 324660 }, { "epoch": 28.12, "learning_rate": 4.707557534404996e-06, "loss": 0.5794, "step": 324670 }, { "epoch": 28.12, "learning_rate": 4.705389152307158e-06, "loss": 0.6115, "step": 324680 }, { "epoch": 28.12, "learning_rate": 4.703220770209321e-06, "loss": 0.6148, "step": 324690 }, { "epoch": 28.12, "learning_rate": 4.7010523881114825e-06, "loss": 0.5479, "step": 324700 }, { "epoch": 28.12, "learning_rate": 4.6988840060136455e-06, "loss": 0.5356, "step": 324710 }, { "epoch": 28.12, "learning_rate": 4.6967156239158085e-06, "loss": 0.4716, "step": 324720 }, { "epoch": 28.12, "learning_rate": 4.694547241817971e-06, "loss": 0.5859, "step": 324730 }, { "epoch": 28.12, "learning_rate": 4.692378859720134e-06, "loss": 0.5516, "step": 324740 }, { "epoch": 28.13, "learning_rate": 4.690210477622296e-06, "loss": 0.6359, "step": 324750 }, { "epoch": 28.13, "learning_rate": 4.688042095524459e-06, "loss": 0.6637, "step": 324760 }, { "epoch": 28.13, "learning_rate": 4.685873713426621e-06, "loss": 0.6215, "step": 324770 }, { "epoch": 28.13, "learning_rate": 4.683705331328784e-06, "loss": 0.5971, "step": 324780 }, { "epoch": 28.13, "learning_rate": 4.681536949230947e-06, "loss": 0.5308, "step": 324790 }, { "epoch": 28.13, "learning_rate": 4.679368567133109e-06, "loss": 0.5172, "step": 324800 }, { "epoch": 28.13, "learning_rate": 4.677200185035272e-06, "loss": 0.5785, "step": 324810 }, { "epoch": 28.13, "learning_rate": 4.675031802937434e-06, "loss": 0.5499, "step": 324820 }, { "epoch": 28.13, "learning_rate": 4.672863420839597e-06, "loss": 0.4791, "step": 324830 }, { "epoch": 28.13, "learning_rate": 4.6706950387417595e-06, "loss": 0.6212, "step": 324840 }, { "epoch": 28.13, "learning_rate": 4.6685266566439225e-06, "loss": 0.4251, "step": 324850 }, { "epoch": 28.13, "learning_rate": 4.6663582745460855e-06, "loss": 0.621, "step": 324860 }, { "epoch": 28.14, "learning_rate": 4.664189892448248e-06, "loss": 0.5522, "step": 324870 }, { "epoch": 28.14, "learning_rate": 4.662021510350411e-06, "loss": 0.6167, "step": 324880 }, { "epoch": 28.14, "learning_rate": 4.659853128252573e-06, "loss": 0.6266, "step": 324890 }, { "epoch": 28.14, "learning_rate": 4.657684746154735e-06, "loss": 0.5284, "step": 324900 }, { "epoch": 28.14, "learning_rate": 4.655516364056898e-06, "loss": 0.5824, "step": 324910 }, { "epoch": 28.14, "learning_rate": 4.65334798195906e-06, "loss": 0.5671, "step": 324920 }, { "epoch": 28.14, "learning_rate": 4.651179599861223e-06, "loss": 0.3791, "step": 324930 }, { "epoch": 28.14, "learning_rate": 4.649011217763385e-06, "loss": 0.6292, "step": 324940 }, { "epoch": 28.14, "learning_rate": 4.646842835665548e-06, "loss": 0.5203, "step": 324950 }, { "epoch": 28.14, "learning_rate": 4.644674453567711e-06, "loss": 0.6153, "step": 324960 }, { "epoch": 28.14, "learning_rate": 4.6425060714698734e-06, "loss": 0.5642, "step": 324970 }, { "epoch": 28.15, "learning_rate": 4.6403376893720364e-06, "loss": 0.6225, "step": 324980 }, { "epoch": 28.15, "learning_rate": 4.638169307274199e-06, "loss": 0.5837, "step": 324990 }, { "epoch": 28.15, "learning_rate": 4.636000925176362e-06, "loss": 0.5663, "step": 325000 }, { "epoch": 28.15, "learning_rate": 4.633832543078524e-06, "loss": 0.5194, "step": 325010 }, { "epoch": 28.15, "learning_rate": 4.631664160980686e-06, "loss": 0.5138, "step": 325020 }, { "epoch": 28.15, "learning_rate": 4.629495778882849e-06, "loss": 0.7768, "step": 325030 }, { "epoch": 28.15, "learning_rate": 4.627327396785012e-06, "loss": 0.5706, "step": 325040 }, { "epoch": 28.15, "learning_rate": 4.625159014687174e-06, "loss": 0.6516, "step": 325050 }, { "epoch": 28.15, "learning_rate": 4.622990632589337e-06, "loss": 0.5665, "step": 325060 }, { "epoch": 28.15, "learning_rate": 4.620822250491499e-06, "loss": 0.434, "step": 325070 }, { "epoch": 28.15, "learning_rate": 4.618653868393662e-06, "loss": 0.6341, "step": 325080 }, { "epoch": 28.15, "learning_rate": 4.616485486295824e-06, "loss": 0.4884, "step": 325090 }, { "epoch": 28.16, "learning_rate": 4.614317104197987e-06, "loss": 0.5456, "step": 325100 }, { "epoch": 28.16, "learning_rate": 4.61214872210015e-06, "loss": 0.6186, "step": 325110 }, { "epoch": 28.16, "learning_rate": 4.6099803400023126e-06, "loss": 0.5725, "step": 325120 }, { "epoch": 28.16, "learning_rate": 4.6078119579044756e-06, "loss": 0.53, "step": 325130 }, { "epoch": 28.16, "learning_rate": 4.605643575806638e-06, "loss": 0.5974, "step": 325140 }, { "epoch": 28.16, "learning_rate": 4.6034751937088e-06, "loss": 0.5076, "step": 325150 }, { "epoch": 28.16, "learning_rate": 4.601306811610963e-06, "loss": 0.6026, "step": 325160 }, { "epoch": 28.16, "learning_rate": 4.599138429513125e-06, "loss": 0.6078, "step": 325170 }, { "epoch": 28.16, "learning_rate": 4.596970047415288e-06, "loss": 0.5977, "step": 325180 }, { "epoch": 28.16, "learning_rate": 4.594801665317451e-06, "loss": 0.6028, "step": 325190 }, { "epoch": 28.16, "learning_rate": 4.592633283219613e-06, "loss": 0.6143, "step": 325200 }, { "epoch": 28.17, "learning_rate": 4.590464901121776e-06, "loss": 0.7497, "step": 325210 }, { "epoch": 28.17, "learning_rate": 4.588296519023938e-06, "loss": 0.5332, "step": 325220 }, { "epoch": 28.17, "learning_rate": 4.586128136926101e-06, "loss": 0.6838, "step": 325230 }, { "epoch": 28.17, "learning_rate": 4.5839597548282635e-06, "loss": 0.6238, "step": 325240 }, { "epoch": 28.17, "learning_rate": 4.5817913727304265e-06, "loss": 0.4233, "step": 325250 }, { "epoch": 28.17, "learning_rate": 4.5796229906325895e-06, "loss": 0.5814, "step": 325260 }, { "epoch": 28.17, "learning_rate": 4.577454608534752e-06, "loss": 0.5279, "step": 325270 }, { "epoch": 28.17, "learning_rate": 4.575286226436914e-06, "loss": 0.646, "step": 325280 }, { "epoch": 28.17, "learning_rate": 4.573117844339077e-06, "loss": 0.5331, "step": 325290 }, { "epoch": 28.17, "learning_rate": 4.570949462241239e-06, "loss": 0.6593, "step": 325300 }, { "epoch": 28.17, "learning_rate": 4.568781080143402e-06, "loss": 0.6207, "step": 325310 }, { "epoch": 28.17, "learning_rate": 4.566612698045564e-06, "loss": 0.6412, "step": 325320 }, { "epoch": 28.18, "learning_rate": 4.564444315947727e-06, "loss": 0.5766, "step": 325330 }, { "epoch": 28.18, "learning_rate": 4.562275933849889e-06, "loss": 0.5832, "step": 325340 }, { "epoch": 28.18, "learning_rate": 4.560107551752052e-06, "loss": 0.5635, "step": 325350 }, { "epoch": 28.18, "learning_rate": 4.557939169654215e-06, "loss": 0.6496, "step": 325360 }, { "epoch": 28.18, "learning_rate": 4.5557707875563775e-06, "loss": 0.6925, "step": 325370 }, { "epoch": 28.18, "learning_rate": 4.5536024054585405e-06, "loss": 0.699, "step": 325380 }, { "epoch": 28.18, "learning_rate": 4.551434023360703e-06, "loss": 0.5861, "step": 325390 }, { "epoch": 28.18, "learning_rate": 4.549265641262866e-06, "loss": 0.5704, "step": 325400 }, { "epoch": 28.18, "learning_rate": 4.547097259165028e-06, "loss": 0.5335, "step": 325410 }, { "epoch": 28.18, "learning_rate": 4.54492887706719e-06, "loss": 0.4739, "step": 325420 }, { "epoch": 28.18, "learning_rate": 4.542760494969353e-06, "loss": 0.5212, "step": 325430 }, { "epoch": 28.19, "learning_rate": 4.540592112871516e-06, "loss": 0.5383, "step": 325440 }, { "epoch": 28.19, "learning_rate": 4.538423730773678e-06, "loss": 0.6203, "step": 325450 }, { "epoch": 28.19, "learning_rate": 4.536255348675841e-06, "loss": 0.5889, "step": 325460 }, { "epoch": 28.19, "learning_rate": 4.534086966578003e-06, "loss": 0.6929, "step": 325470 }, { "epoch": 28.19, "learning_rate": 4.531918584480166e-06, "loss": 0.4826, "step": 325480 }, { "epoch": 28.19, "learning_rate": 4.5297502023823284e-06, "loss": 0.6257, "step": 325490 }, { "epoch": 28.19, "learning_rate": 4.5275818202844915e-06, "loss": 0.4997, "step": 325500 }, { "epoch": 28.19, "learning_rate": 4.5254134381866545e-06, "loss": 0.6212, "step": 325510 }, { "epoch": 28.19, "learning_rate": 4.523245056088817e-06, "loss": 0.601, "step": 325520 }, { "epoch": 28.19, "learning_rate": 4.52107667399098e-06, "loss": 0.5305, "step": 325530 }, { "epoch": 28.19, "learning_rate": 4.518908291893142e-06, "loss": 0.5761, "step": 325540 }, { "epoch": 28.19, "learning_rate": 4.516739909795304e-06, "loss": 0.5797, "step": 325550 }, { "epoch": 28.2, "learning_rate": 4.514571527697467e-06, "loss": 0.4971, "step": 325560 }, { "epoch": 28.2, "learning_rate": 4.512403145599629e-06, "loss": 0.5226, "step": 325570 }, { "epoch": 28.2, "learning_rate": 4.510234763501792e-06, "loss": 0.6886, "step": 325580 }, { "epoch": 28.2, "learning_rate": 4.508066381403955e-06, "loss": 0.6048, "step": 325590 }, { "epoch": 28.2, "learning_rate": 4.505897999306117e-06, "loss": 0.5973, "step": 325600 }, { "epoch": 28.2, "learning_rate": 4.50372961720828e-06, "loss": 0.6004, "step": 325610 }, { "epoch": 28.2, "learning_rate": 4.501561235110442e-06, "loss": 0.6779, "step": 325620 }, { "epoch": 28.2, "learning_rate": 4.499392853012605e-06, "loss": 0.5399, "step": 325630 }, { "epoch": 28.2, "learning_rate": 4.4972244709147676e-06, "loss": 0.5932, "step": 325640 }, { "epoch": 28.2, "learning_rate": 4.4950560888169306e-06, "loss": 0.6224, "step": 325650 }, { "epoch": 28.2, "learning_rate": 4.492887706719094e-06, "loss": 0.5564, "step": 325660 }, { "epoch": 28.21, "learning_rate": 4.490719324621255e-06, "loss": 0.4593, "step": 325670 }, { "epoch": 28.21, "learning_rate": 4.488550942523418e-06, "loss": 0.6519, "step": 325680 }, { "epoch": 28.21, "learning_rate": 4.486382560425581e-06, "loss": 0.6169, "step": 325690 }, { "epoch": 28.21, "learning_rate": 4.484214178327743e-06, "loss": 0.5915, "step": 325700 }, { "epoch": 28.21, "learning_rate": 4.482045796229906e-06, "loss": 0.6253, "step": 325710 }, { "epoch": 28.21, "learning_rate": 4.479877414132068e-06, "loss": 0.5266, "step": 325720 }, { "epoch": 28.21, "learning_rate": 4.477709032034231e-06, "loss": 0.5102, "step": 325730 }, { "epoch": 28.21, "learning_rate": 4.475540649936393e-06, "loss": 0.5648, "step": 325740 }, { "epoch": 28.21, "learning_rate": 4.473372267838556e-06, "loss": 0.6252, "step": 325750 }, { "epoch": 28.21, "learning_rate": 4.471203885740719e-06, "loss": 0.435, "step": 325760 }, { "epoch": 28.21, "learning_rate": 4.4690355036428815e-06, "loss": 0.6732, "step": 325770 }, { "epoch": 28.21, "learning_rate": 4.4668671215450445e-06, "loss": 0.5948, "step": 325780 }, { "epoch": 28.22, "learning_rate": 4.464698739447207e-06, "loss": 0.6233, "step": 325790 }, { "epoch": 28.22, "learning_rate": 4.462530357349369e-06, "loss": 0.5596, "step": 325800 }, { "epoch": 28.22, "learning_rate": 4.460361975251532e-06, "loss": 0.5895, "step": 325810 }, { "epoch": 28.22, "learning_rate": 4.458193593153694e-06, "loss": 0.6195, "step": 325820 }, { "epoch": 28.22, "learning_rate": 4.456025211055857e-06, "loss": 0.661, "step": 325830 }, { "epoch": 28.22, "learning_rate": 4.45385682895802e-06, "loss": 0.6121, "step": 325840 }, { "epoch": 28.22, "learning_rate": 4.451688446860182e-06, "loss": 0.5062, "step": 325850 }, { "epoch": 28.22, "learning_rate": 4.449520064762345e-06, "loss": 0.52, "step": 325860 }, { "epoch": 28.22, "learning_rate": 4.447351682664507e-06, "loss": 0.5581, "step": 325870 }, { "epoch": 28.22, "learning_rate": 4.44518330056667e-06, "loss": 0.4961, "step": 325880 }, { "epoch": 28.22, "learning_rate": 4.4430149184688325e-06, "loss": 0.5655, "step": 325890 }, { "epoch": 28.23, "learning_rate": 4.4408465363709955e-06, "loss": 0.5195, "step": 325900 }, { "epoch": 28.23, "learning_rate": 4.4386781542731585e-06, "loss": 0.6024, "step": 325910 }, { "epoch": 28.23, "learning_rate": 4.436509772175321e-06, "loss": 0.6142, "step": 325920 }, { "epoch": 28.23, "learning_rate": 4.434341390077483e-06, "loss": 0.5152, "step": 325930 }, { "epoch": 28.23, "learning_rate": 4.432173007979646e-06, "loss": 0.5934, "step": 325940 }, { "epoch": 28.23, "learning_rate": 4.430004625881808e-06, "loss": 0.5768, "step": 325950 }, { "epoch": 28.23, "learning_rate": 4.427836243783971e-06, "loss": 0.5568, "step": 325960 }, { "epoch": 28.23, "learning_rate": 4.425667861686133e-06, "loss": 0.554, "step": 325970 }, { "epoch": 28.23, "learning_rate": 4.423499479588296e-06, "loss": 0.4828, "step": 325980 }, { "epoch": 28.23, "learning_rate": 4.421331097490459e-06, "loss": 0.6816, "step": 325990 }, { "epoch": 28.23, "learning_rate": 4.419162715392621e-06, "loss": 0.6075, "step": 326000 }, { "epoch": 28.23, "learning_rate": 4.416994333294784e-06, "loss": 0.4297, "step": 326010 }, { "epoch": 28.24, "learning_rate": 4.4148259511969465e-06, "loss": 0.5463, "step": 326020 }, { "epoch": 28.24, "learning_rate": 4.4126575690991095e-06, "loss": 0.6322, "step": 326030 }, { "epoch": 28.24, "learning_rate": 4.410489187001272e-06, "loss": 0.4836, "step": 326040 }, { "epoch": 28.24, "learning_rate": 4.408320804903434e-06, "loss": 0.4647, "step": 326050 }, { "epoch": 28.24, "learning_rate": 4.406152422805597e-06, "loss": 0.5694, "step": 326060 }, { "epoch": 28.24, "learning_rate": 4.40398404070776e-06, "loss": 0.5267, "step": 326070 }, { "epoch": 28.24, "learning_rate": 4.401815658609922e-06, "loss": 0.7463, "step": 326080 }, { "epoch": 28.24, "learning_rate": 4.399647276512085e-06, "loss": 0.6166, "step": 326090 }, { "epoch": 28.24, "learning_rate": 4.397478894414247e-06, "loss": 0.6138, "step": 326100 }, { "epoch": 28.24, "learning_rate": 4.39531051231641e-06, "loss": 0.6433, "step": 326110 }, { "epoch": 28.24, "learning_rate": 4.393142130218572e-06, "loss": 0.7063, "step": 326120 }, { "epoch": 28.24, "learning_rate": 4.390973748120735e-06, "loss": 0.5199, "step": 326130 }, { "epoch": 28.25, "learning_rate": 4.388805366022898e-06, "loss": 0.5647, "step": 326140 }, { "epoch": 28.25, "learning_rate": 4.38663698392506e-06, "loss": 0.5249, "step": 326150 }, { "epoch": 28.25, "learning_rate": 4.3844686018272234e-06, "loss": 0.6801, "step": 326160 }, { "epoch": 28.25, "learning_rate": 4.382300219729386e-06, "loss": 0.5105, "step": 326170 }, { "epoch": 28.25, "learning_rate": 4.380131837631548e-06, "loss": 0.5764, "step": 326180 }, { "epoch": 28.25, "learning_rate": 4.377963455533711e-06, "loss": 0.5455, "step": 326190 }, { "epoch": 28.25, "learning_rate": 4.375795073435873e-06, "loss": 0.6208, "step": 326200 }, { "epoch": 28.25, "learning_rate": 4.373626691338036e-06, "loss": 0.6011, "step": 326210 }, { "epoch": 28.25, "learning_rate": 4.371458309240198e-06, "loss": 0.5705, "step": 326220 }, { "epoch": 28.25, "learning_rate": 4.369289927142361e-06, "loss": 0.5668, "step": 326230 }, { "epoch": 28.25, "learning_rate": 4.367121545044524e-06, "loss": 0.6336, "step": 326240 }, { "epoch": 28.26, "learning_rate": 4.364953162946686e-06, "loss": 0.5913, "step": 326250 }, { "epoch": 28.26, "learning_rate": 4.362784780848849e-06, "loss": 0.5345, "step": 326260 }, { "epoch": 28.26, "learning_rate": 4.360616398751011e-06, "loss": 0.5911, "step": 326270 }, { "epoch": 28.26, "learning_rate": 4.358448016653174e-06, "loss": 0.5174, "step": 326280 }, { "epoch": 28.26, "learning_rate": 4.3562796345553365e-06, "loss": 0.6289, "step": 326290 }, { "epoch": 28.26, "learning_rate": 4.3541112524574995e-06, "loss": 0.5828, "step": 326300 }, { "epoch": 28.26, "learning_rate": 4.351942870359662e-06, "loss": 0.5589, "step": 326310 }, { "epoch": 28.26, "learning_rate": 4.349774488261825e-06, "loss": 0.5938, "step": 326320 }, { "epoch": 28.26, "learning_rate": 4.347606106163987e-06, "loss": 0.4933, "step": 326330 }, { "epoch": 28.26, "learning_rate": 4.34543772406615e-06, "loss": 0.63, "step": 326340 }, { "epoch": 28.26, "learning_rate": 4.343269341968312e-06, "loss": 0.5971, "step": 326350 }, { "epoch": 28.26, "learning_rate": 4.341100959870475e-06, "loss": 0.5185, "step": 326360 }, { "epoch": 28.27, "learning_rate": 4.338932577772637e-06, "loss": 0.5882, "step": 326370 }, { "epoch": 28.27, "learning_rate": 4.3367641956748e-06, "loss": 0.6934, "step": 326380 }, { "epoch": 28.27, "learning_rate": 4.334595813576963e-06, "loss": 0.5199, "step": 326390 }, { "epoch": 28.27, "learning_rate": 4.332427431479125e-06, "loss": 0.5233, "step": 326400 }, { "epoch": 28.27, "learning_rate": 4.330259049381288e-06, "loss": 0.5059, "step": 326410 }, { "epoch": 28.27, "learning_rate": 4.3280906672834505e-06, "loss": 0.5434, "step": 326420 }, { "epoch": 28.27, "learning_rate": 4.3259222851856135e-06, "loss": 0.6965, "step": 326430 }, { "epoch": 28.27, "learning_rate": 4.323753903087776e-06, "loss": 0.6959, "step": 326440 }, { "epoch": 28.27, "learning_rate": 4.321585520989938e-06, "loss": 0.4606, "step": 326450 }, { "epoch": 28.27, "learning_rate": 4.319417138892101e-06, "loss": 0.5133, "step": 326460 }, { "epoch": 28.27, "learning_rate": 4.317248756794264e-06, "loss": 0.4798, "step": 326470 }, { "epoch": 28.28, "learning_rate": 4.315080374696426e-06, "loss": 0.4782, "step": 326480 }, { "epoch": 28.28, "learning_rate": 4.312911992598589e-06, "loss": 0.5411, "step": 326490 }, { "epoch": 28.28, "learning_rate": 4.310743610500751e-06, "loss": 0.5944, "step": 326500 }, { "epoch": 28.28, "learning_rate": 4.308575228402914e-06, "loss": 0.4218, "step": 326510 }, { "epoch": 28.28, "learning_rate": 4.306406846305076e-06, "loss": 0.4947, "step": 326520 }, { "epoch": 28.28, "learning_rate": 4.304238464207239e-06, "loss": 0.5623, "step": 326530 }, { "epoch": 28.28, "learning_rate": 4.302070082109402e-06, "loss": 0.5503, "step": 326540 }, { "epoch": 28.28, "learning_rate": 4.2999017000115645e-06, "loss": 0.5071, "step": 326550 }, { "epoch": 28.28, "learning_rate": 4.297733317913727e-06, "loss": 0.4489, "step": 326560 }, { "epoch": 28.28, "learning_rate": 4.29556493581589e-06, "loss": 0.5452, "step": 326570 }, { "epoch": 28.28, "learning_rate": 4.293396553718052e-06, "loss": 0.5392, "step": 326580 }, { "epoch": 28.28, "learning_rate": 4.291228171620215e-06, "loss": 0.4985, "step": 326590 }, { "epoch": 28.29, "learning_rate": 4.289059789522377e-06, "loss": 0.5179, "step": 326600 }, { "epoch": 28.29, "learning_rate": 4.28689140742454e-06, "loss": 0.5129, "step": 326610 }, { "epoch": 28.29, "learning_rate": 4.284723025326702e-06, "loss": 0.6662, "step": 326620 }, { "epoch": 28.29, "learning_rate": 4.282554643228865e-06, "loss": 0.4446, "step": 326630 }, { "epoch": 28.29, "learning_rate": 4.280386261131028e-06, "loss": 0.6974, "step": 326640 }, { "epoch": 28.29, "learning_rate": 4.27821787903319e-06, "loss": 0.5654, "step": 326650 }, { "epoch": 28.29, "learning_rate": 4.276049496935353e-06, "loss": 0.5048, "step": 326660 }, { "epoch": 28.29, "learning_rate": 4.2738811148375154e-06, "loss": 0.4828, "step": 326670 }, { "epoch": 28.29, "learning_rate": 4.2717127327396784e-06, "loss": 0.6063, "step": 326680 }, { "epoch": 28.29, "learning_rate": 4.269544350641841e-06, "loss": 0.5015, "step": 326690 }, { "epoch": 28.29, "learning_rate": 4.267375968544003e-06, "loss": 0.6057, "step": 326700 }, { "epoch": 28.3, "learning_rate": 4.265207586446166e-06, "loss": 0.6054, "step": 326710 }, { "epoch": 28.3, "learning_rate": 4.263039204348329e-06, "loss": 0.5169, "step": 326720 }, { "epoch": 28.3, "learning_rate": 4.260870822250491e-06, "loss": 0.5415, "step": 326730 }, { "epoch": 28.3, "learning_rate": 4.258702440152654e-06, "loss": 0.4591, "step": 326740 }, { "epoch": 28.3, "learning_rate": 4.256534058054816e-06, "loss": 0.5425, "step": 326750 }, { "epoch": 28.3, "learning_rate": 4.254365675956979e-06, "loss": 0.6971, "step": 326760 }, { "epoch": 28.3, "learning_rate": 4.252197293859141e-06, "loss": 0.544, "step": 326770 }, { "epoch": 28.3, "learning_rate": 4.250028911761304e-06, "loss": 0.5908, "step": 326780 }, { "epoch": 28.3, "learning_rate": 4.247860529663467e-06, "loss": 0.6136, "step": 326790 }, { "epoch": 28.3, "learning_rate": 4.245692147565629e-06, "loss": 0.6307, "step": 326800 }, { "epoch": 28.3, "learning_rate": 4.243523765467792e-06, "loss": 0.5582, "step": 326810 }, { "epoch": 28.3, "learning_rate": 4.2413553833699546e-06, "loss": 0.6722, "step": 326820 }, { "epoch": 28.31, "learning_rate": 4.239187001272117e-06, "loss": 0.6136, "step": 326830 }, { "epoch": 28.31, "learning_rate": 4.23701861917428e-06, "loss": 0.5733, "step": 326840 }, { "epoch": 28.31, "learning_rate": 4.234850237076442e-06, "loss": 0.6368, "step": 326850 }, { "epoch": 28.31, "learning_rate": 4.232681854978605e-06, "loss": 0.4759, "step": 326860 }, { "epoch": 28.31, "learning_rate": 4.230513472880768e-06, "loss": 0.5314, "step": 326870 }, { "epoch": 28.31, "learning_rate": 4.22834509078293e-06, "loss": 0.7189, "step": 326880 }, { "epoch": 28.31, "learning_rate": 4.226176708685093e-06, "loss": 0.6387, "step": 326890 }, { "epoch": 28.31, "learning_rate": 4.224008326587255e-06, "loss": 0.5541, "step": 326900 }, { "epoch": 28.31, "learning_rate": 4.221839944489418e-06, "loss": 0.5673, "step": 326910 }, { "epoch": 28.31, "learning_rate": 4.21967156239158e-06, "loss": 0.5458, "step": 326920 }, { "epoch": 28.31, "learning_rate": 4.217503180293743e-06, "loss": 0.5646, "step": 326930 }, { "epoch": 28.32, "learning_rate": 4.215334798195906e-06, "loss": 0.525, "step": 326940 }, { "epoch": 28.32, "learning_rate": 4.213166416098068e-06, "loss": 0.5052, "step": 326950 }, { "epoch": 28.32, "learning_rate": 4.210998034000231e-06, "loss": 0.5859, "step": 326960 }, { "epoch": 28.32, "learning_rate": 4.208829651902394e-06, "loss": 0.6744, "step": 326970 }, { "epoch": 28.32, "learning_rate": 4.206661269804556e-06, "loss": 0.5868, "step": 326980 }, { "epoch": 28.32, "learning_rate": 4.204492887706719e-06, "loss": 0.534, "step": 326990 }, { "epoch": 28.32, "learning_rate": 4.202324505608881e-06, "loss": 0.5934, "step": 327000 }, { "epoch": 28.32, "learning_rate": 4.200156123511044e-06, "loss": 0.5995, "step": 327010 }, { "epoch": 28.32, "learning_rate": 4.197987741413206e-06, "loss": 0.6232, "step": 327020 }, { "epoch": 28.32, "learning_rate": 4.195819359315369e-06, "loss": 0.5466, "step": 327030 }, { "epoch": 28.32, "learning_rate": 4.193650977217532e-06, "loss": 0.4709, "step": 327040 }, { "epoch": 28.32, "learning_rate": 4.191482595119694e-06, "loss": 0.605, "step": 327050 }, { "epoch": 28.33, "learning_rate": 4.189314213021857e-06, "loss": 0.6553, "step": 327060 }, { "epoch": 28.33, "learning_rate": 4.1871458309240195e-06, "loss": 0.5939, "step": 327070 }, { "epoch": 28.33, "learning_rate": 4.184977448826182e-06, "loss": 0.6106, "step": 327080 }, { "epoch": 28.33, "learning_rate": 4.182809066728345e-06, "loss": 0.4786, "step": 327090 }, { "epoch": 28.33, "learning_rate": 4.180640684630507e-06, "loss": 0.4684, "step": 327100 }, { "epoch": 28.33, "learning_rate": 4.17847230253267e-06, "loss": 0.5827, "step": 327110 }, { "epoch": 28.33, "learning_rate": 4.176303920434833e-06, "loss": 0.6329, "step": 327120 }, { "epoch": 28.33, "learning_rate": 4.174135538336995e-06, "loss": 0.5624, "step": 327130 }, { "epoch": 28.33, "learning_rate": 4.171967156239158e-06, "loss": 0.6362, "step": 327140 }, { "epoch": 28.33, "learning_rate": 4.16979877414132e-06, "loss": 0.5297, "step": 327150 }, { "epoch": 28.33, "learning_rate": 4.167630392043483e-06, "loss": 0.5437, "step": 327160 }, { "epoch": 28.33, "learning_rate": 4.165462009945645e-06, "loss": 0.7136, "step": 327170 }, { "epoch": 28.34, "learning_rate": 4.163293627847808e-06, "loss": 0.5206, "step": 327180 }, { "epoch": 28.34, "learning_rate": 4.161125245749971e-06, "loss": 0.6199, "step": 327190 }, { "epoch": 28.34, "learning_rate": 4.1589568636521334e-06, "loss": 0.4693, "step": 327200 }, { "epoch": 28.34, "learning_rate": 4.156788481554296e-06, "loss": 0.6679, "step": 327210 }, { "epoch": 28.34, "learning_rate": 4.154620099456459e-06, "loss": 0.5254, "step": 327220 }, { "epoch": 28.34, "learning_rate": 4.152451717358621e-06, "loss": 0.4623, "step": 327230 }, { "epoch": 28.34, "learning_rate": 4.150283335260784e-06, "loss": 0.5184, "step": 327240 }, { "epoch": 28.34, "learning_rate": 4.148114953162946e-06, "loss": 0.6869, "step": 327250 }, { "epoch": 28.34, "learning_rate": 4.145946571065109e-06, "loss": 0.6527, "step": 327260 }, { "epoch": 28.34, "learning_rate": 4.143778188967272e-06, "loss": 0.5904, "step": 327270 }, { "epoch": 28.34, "learning_rate": 4.141609806869434e-06, "loss": 0.5701, "step": 327280 }, { "epoch": 28.35, "learning_rate": 4.139441424771597e-06, "loss": 0.5072, "step": 327290 }, { "epoch": 28.35, "learning_rate": 4.137273042673759e-06, "loss": 0.5885, "step": 327300 }, { "epoch": 28.35, "learning_rate": 4.135104660575922e-06, "loss": 0.5607, "step": 327310 }, { "epoch": 28.35, "learning_rate": 4.132936278478084e-06, "loss": 0.5847, "step": 327320 }, { "epoch": 28.35, "learning_rate": 4.1307678963802466e-06, "loss": 0.6011, "step": 327330 }, { "epoch": 28.35, "learning_rate": 4.1285995142824096e-06, "loss": 0.5559, "step": 327340 }, { "epoch": 28.35, "learning_rate": 4.1264311321845726e-06, "loss": 0.6106, "step": 327350 }, { "epoch": 28.35, "learning_rate": 4.124262750086735e-06, "loss": 0.6766, "step": 327360 }, { "epoch": 28.35, "learning_rate": 4.122094367988898e-06, "loss": 0.5783, "step": 327370 }, { "epoch": 28.35, "learning_rate": 4.11992598589106e-06, "loss": 0.4929, "step": 327380 }, { "epoch": 28.35, "learning_rate": 4.117757603793223e-06, "loss": 0.5978, "step": 327390 }, { "epoch": 28.35, "learning_rate": 4.115589221695385e-06, "loss": 0.6064, "step": 327400 }, { "epoch": 28.36, "learning_rate": 4.113420839597548e-06, "loss": 0.6423, "step": 327410 }, { "epoch": 28.36, "learning_rate": 4.111252457499711e-06, "loss": 0.485, "step": 327420 }, { "epoch": 28.36, "learning_rate": 4.109084075401873e-06, "loss": 0.4975, "step": 327430 }, { "epoch": 28.36, "learning_rate": 4.106915693304036e-06, "loss": 0.5148, "step": 327440 }, { "epoch": 28.36, "learning_rate": 4.104747311206198e-06, "loss": 0.5365, "step": 327450 }, { "epoch": 28.36, "learning_rate": 4.1025789291083605e-06, "loss": 0.4474, "step": 327460 }, { "epoch": 28.36, "learning_rate": 4.1004105470105235e-06, "loss": 0.492, "step": 327470 }, { "epoch": 28.36, "learning_rate": 4.098242164912686e-06, "loss": 0.5358, "step": 327480 }, { "epoch": 28.36, "learning_rate": 4.096073782814849e-06, "loss": 0.6278, "step": 327490 }, { "epoch": 28.36, "learning_rate": 4.093905400717011e-06, "loss": 0.773, "step": 327500 }, { "epoch": 28.36, "learning_rate": 4.091737018619174e-06, "loss": 0.5338, "step": 327510 }, { "epoch": 28.37, "learning_rate": 4.089568636521337e-06, "loss": 0.4886, "step": 327520 }, { "epoch": 28.37, "learning_rate": 4.087400254423499e-06, "loss": 0.6533, "step": 327530 }, { "epoch": 28.37, "learning_rate": 4.085231872325662e-06, "loss": 0.4806, "step": 327540 }, { "epoch": 28.37, "learning_rate": 4.083063490227824e-06, "loss": 0.6507, "step": 327550 }, { "epoch": 28.37, "learning_rate": 4.080895108129987e-06, "loss": 0.5972, "step": 327560 }, { "epoch": 28.37, "learning_rate": 4.078726726032149e-06, "loss": 0.5215, "step": 327570 }, { "epoch": 28.37, "learning_rate": 4.076558343934312e-06, "loss": 0.5668, "step": 327580 }, { "epoch": 28.37, "learning_rate": 4.0743899618364745e-06, "loss": 0.5992, "step": 327590 }, { "epoch": 28.37, "learning_rate": 4.0722215797386375e-06, "loss": 0.5338, "step": 327600 }, { "epoch": 28.37, "learning_rate": 4.0700531976408e-06, "loss": 0.5311, "step": 327610 }, { "epoch": 28.37, "learning_rate": 4.067884815542963e-06, "loss": 0.5832, "step": 327620 }, { "epoch": 28.37, "learning_rate": 4.065716433445125e-06, "loss": 0.5358, "step": 327630 }, { "epoch": 28.38, "learning_rate": 4.063548051347288e-06, "loss": 0.5793, "step": 327640 }, { "epoch": 28.38, "learning_rate": 4.06137966924945e-06, "loss": 0.6053, "step": 327650 }, { "epoch": 28.38, "learning_rate": 4.059211287151613e-06, "loss": 0.4756, "step": 327660 }, { "epoch": 28.38, "learning_rate": 4.057042905053776e-06, "loss": 0.5169, "step": 327670 }, { "epoch": 28.38, "learning_rate": 4.054874522955938e-06, "loss": 0.5494, "step": 327680 }, { "epoch": 28.38, "learning_rate": 4.052706140858101e-06, "loss": 0.5984, "step": 327690 }, { "epoch": 28.38, "learning_rate": 4.050537758760263e-06, "loss": 0.4949, "step": 327700 }, { "epoch": 28.38, "learning_rate": 4.048369376662426e-06, "loss": 0.5659, "step": 327710 }, { "epoch": 28.38, "learning_rate": 4.0462009945645884e-06, "loss": 0.394, "step": 327720 }, { "epoch": 28.38, "learning_rate": 4.044032612466751e-06, "loss": 0.501, "step": 327730 }, { "epoch": 28.38, "learning_rate": 4.041864230368914e-06, "loss": 0.5012, "step": 327740 }, { "epoch": 28.39, "learning_rate": 4.039695848271077e-06, "loss": 0.6773, "step": 327750 }, { "epoch": 28.39, "learning_rate": 4.037527466173239e-06, "loss": 0.5574, "step": 327760 }, { "epoch": 28.39, "learning_rate": 4.035359084075402e-06, "loss": 0.5834, "step": 327770 }, { "epoch": 28.39, "learning_rate": 4.033190701977564e-06, "loss": 0.6933, "step": 327780 }, { "epoch": 28.39, "learning_rate": 4.031022319879727e-06, "loss": 0.5729, "step": 327790 }, { "epoch": 28.39, "learning_rate": 4.028853937781889e-06, "loss": 0.6046, "step": 327800 }, { "epoch": 28.39, "learning_rate": 4.026685555684052e-06, "loss": 0.5642, "step": 327810 }, { "epoch": 28.39, "learning_rate": 4.024517173586215e-06, "loss": 0.4737, "step": 327820 }, { "epoch": 28.39, "learning_rate": 4.022348791488377e-06, "loss": 0.627, "step": 327830 }, { "epoch": 28.39, "learning_rate": 4.02018040939054e-06, "loss": 0.6831, "step": 327840 }, { "epoch": 28.39, "learning_rate": 4.018012027292702e-06, "loss": 0.702, "step": 327850 }, { "epoch": 28.39, "learning_rate": 4.0158436451948646e-06, "loss": 0.476, "step": 327860 }, { "epoch": 28.4, "learning_rate": 4.0136752630970276e-06, "loss": 0.4992, "step": 327870 }, { "epoch": 28.4, "learning_rate": 4.01150688099919e-06, "loss": 0.5808, "step": 327880 }, { "epoch": 28.4, "learning_rate": 4.009338498901353e-06, "loss": 0.5665, "step": 327890 }, { "epoch": 28.4, "learning_rate": 4.007170116803515e-06, "loss": 0.5591, "step": 327900 }, { "epoch": 28.4, "learning_rate": 4.005001734705678e-06, "loss": 0.4781, "step": 327910 }, { "epoch": 28.4, "learning_rate": 4.002833352607841e-06, "loss": 0.4853, "step": 327920 }, { "epoch": 28.4, "learning_rate": 4.000664970510003e-06, "loss": 0.4993, "step": 327930 }, { "epoch": 28.4, "learning_rate": 3.998496588412166e-06, "loss": 0.4922, "step": 327940 }, { "epoch": 28.4, "learning_rate": 3.996328206314328e-06, "loss": 0.4617, "step": 327950 }, { "epoch": 28.4, "learning_rate": 3.994159824216491e-06, "loss": 0.5482, "step": 327960 }, { "epoch": 28.4, "learning_rate": 3.991991442118653e-06, "loss": 0.5462, "step": 327970 }, { "epoch": 28.41, "learning_rate": 3.9898230600208155e-06, "loss": 0.6228, "step": 327980 }, { "epoch": 28.41, "learning_rate": 3.9876546779229785e-06, "loss": 0.7139, "step": 327990 }, { "epoch": 28.41, "learning_rate": 3.9854862958251415e-06, "loss": 0.5087, "step": 328000 }, { "epoch": 28.41, "learning_rate": 3.983317913727304e-06, "loss": 0.5901, "step": 328010 }, { "epoch": 28.41, "learning_rate": 3.981149531629467e-06, "loss": 0.5744, "step": 328020 }, { "epoch": 28.41, "learning_rate": 3.978981149531629e-06, "loss": 0.6099, "step": 328030 }, { "epoch": 28.41, "learning_rate": 3.976812767433792e-06, "loss": 0.6391, "step": 328040 }, { "epoch": 28.41, "learning_rate": 3.974644385335954e-06, "loss": 0.6189, "step": 328050 }, { "epoch": 28.41, "learning_rate": 3.972476003238117e-06, "loss": 0.7303, "step": 328060 }, { "epoch": 28.41, "learning_rate": 3.97030762114028e-06, "loss": 0.5888, "step": 328070 }, { "epoch": 28.41, "learning_rate": 3.968139239042442e-06, "loss": 0.5091, "step": 328080 }, { "epoch": 28.41, "learning_rate": 3.965970856944605e-06, "loss": 0.6465, "step": 328090 }, { "epoch": 28.42, "learning_rate": 3.963802474846767e-06, "loss": 0.6387, "step": 328100 }, { "epoch": 28.42, "learning_rate": 3.9616340927489295e-06, "loss": 0.5214, "step": 328110 }, { "epoch": 28.42, "learning_rate": 3.9594657106510925e-06, "loss": 0.6557, "step": 328120 }, { "epoch": 28.42, "learning_rate": 3.957297328553255e-06, "loss": 0.5995, "step": 328130 }, { "epoch": 28.42, "learning_rate": 3.955128946455418e-06, "loss": 0.5091, "step": 328140 }, { "epoch": 28.42, "learning_rate": 3.952960564357581e-06, "loss": 0.5571, "step": 328150 }, { "epoch": 28.42, "learning_rate": 3.950792182259743e-06, "loss": 0.4757, "step": 328160 }, { "epoch": 28.42, "learning_rate": 3.948623800161906e-06, "loss": 0.5686, "step": 328170 }, { "epoch": 28.42, "learning_rate": 3.946455418064068e-06, "loss": 0.5716, "step": 328180 }, { "epoch": 28.42, "learning_rate": 3.944287035966231e-06, "loss": 0.6156, "step": 328190 }, { "epoch": 28.42, "learning_rate": 3.942118653868393e-06, "loss": 0.4834, "step": 328200 }, { "epoch": 28.43, "learning_rate": 3.939950271770556e-06, "loss": 0.5274, "step": 328210 }, { "epoch": 28.43, "learning_rate": 3.937781889672719e-06, "loss": 0.5257, "step": 328220 }, { "epoch": 28.43, "learning_rate": 3.9356135075748805e-06, "loss": 0.5552, "step": 328230 }, { "epoch": 28.43, "learning_rate": 3.9334451254770435e-06, "loss": 0.495, "step": 328240 }, { "epoch": 28.43, "learning_rate": 3.9312767433792065e-06, "loss": 0.5913, "step": 328250 }, { "epoch": 28.43, "learning_rate": 3.929108361281369e-06, "loss": 0.501, "step": 328260 }, { "epoch": 28.43, "learning_rate": 3.926939979183532e-06, "loss": 0.6623, "step": 328270 }, { "epoch": 28.43, "learning_rate": 3.924771597085694e-06, "loss": 0.4933, "step": 328280 }, { "epoch": 28.43, "learning_rate": 3.922603214987857e-06, "loss": 0.5226, "step": 328290 }, { "epoch": 28.43, "learning_rate": 3.920434832890019e-06, "loss": 0.6144, "step": 328300 }, { "epoch": 28.43, "learning_rate": 3.918266450792182e-06, "loss": 0.6888, "step": 328310 }, { "epoch": 28.43, "learning_rate": 3.916098068694345e-06, "loss": 0.6387, "step": 328320 }, { "epoch": 28.44, "learning_rate": 3.913929686596507e-06, "loss": 0.5967, "step": 328330 }, { "epoch": 28.44, "learning_rate": 3.91176130449867e-06, "loss": 0.6556, "step": 328340 }, { "epoch": 28.44, "learning_rate": 3.909592922400832e-06, "loss": 0.5585, "step": 328350 }, { "epoch": 28.44, "learning_rate": 3.907424540302994e-06, "loss": 0.6479, "step": 328360 }, { "epoch": 28.44, "learning_rate": 3.905256158205157e-06, "loss": 0.6514, "step": 328370 }, { "epoch": 28.44, "learning_rate": 3.9030877761073196e-06, "loss": 0.6067, "step": 328380 }, { "epoch": 28.44, "learning_rate": 3.900919394009483e-06, "loss": 0.6033, "step": 328390 }, { "epoch": 28.44, "learning_rate": 3.898751011911646e-06, "loss": 0.6752, "step": 328400 }, { "epoch": 28.44, "learning_rate": 3.896582629813808e-06, "loss": 0.6391, "step": 328410 }, { "epoch": 28.44, "learning_rate": 3.894414247715971e-06, "loss": 0.4996, "step": 328420 }, { "epoch": 28.44, "learning_rate": 3.892245865618133e-06, "loss": 0.5949, "step": 328430 }, { "epoch": 28.44, "learning_rate": 3.890077483520296e-06, "loss": 0.5639, "step": 328440 }, { "epoch": 28.45, "learning_rate": 3.887909101422458e-06, "loss": 0.6146, "step": 328450 }, { "epoch": 28.45, "learning_rate": 3.885740719324621e-06, "loss": 0.581, "step": 328460 }, { "epoch": 28.45, "learning_rate": 3.883572337226784e-06, "loss": 0.5486, "step": 328470 }, { "epoch": 28.45, "learning_rate": 3.881403955128946e-06, "loss": 0.6201, "step": 328480 }, { "epoch": 28.45, "learning_rate": 3.879235573031108e-06, "loss": 0.6916, "step": 328490 }, { "epoch": 28.45, "learning_rate": 3.877067190933271e-06, "loss": 0.4809, "step": 328500 }, { "epoch": 28.45, "learning_rate": 3.8748988088354335e-06, "loss": 0.5734, "step": 328510 }, { "epoch": 28.45, "learning_rate": 3.8727304267375965e-06, "loss": 0.7019, "step": 328520 }, { "epoch": 28.45, "learning_rate": 3.870562044639759e-06, "loss": 0.5849, "step": 328530 }, { "epoch": 28.45, "learning_rate": 3.868393662541922e-06, "loss": 0.5984, "step": 328540 }, { "epoch": 28.45, "learning_rate": 3.866225280444085e-06, "loss": 0.6214, "step": 328550 }, { "epoch": 28.46, "learning_rate": 3.864056898346247e-06, "loss": 0.5164, "step": 328560 }, { "epoch": 28.46, "learning_rate": 3.86188851624841e-06, "loss": 0.5655, "step": 328570 }, { "epoch": 28.46, "learning_rate": 3.859720134150572e-06, "loss": 0.6765, "step": 328580 }, { "epoch": 28.46, "learning_rate": 3.857551752052735e-06, "loss": 0.5832, "step": 328590 }, { "epoch": 28.46, "learning_rate": 3.855383369954897e-06, "loss": 0.5534, "step": 328600 }, { "epoch": 28.46, "learning_rate": 3.85321498785706e-06, "loss": 0.5944, "step": 328610 }, { "epoch": 28.46, "learning_rate": 3.851046605759222e-06, "loss": 0.5226, "step": 328620 }, { "epoch": 28.46, "learning_rate": 3.848878223661385e-06, "loss": 0.544, "step": 328630 }, { "epoch": 28.46, "learning_rate": 3.8467098415635475e-06, "loss": 0.6598, "step": 328640 }, { "epoch": 28.46, "learning_rate": 3.8445414594657105e-06, "loss": 0.6279, "step": 328650 }, { "epoch": 28.46, "learning_rate": 3.842373077367873e-06, "loss": 0.6782, "step": 328660 }, { "epoch": 28.46, "learning_rate": 3.840204695270036e-06, "loss": 0.4851, "step": 328670 }, { "epoch": 28.47, "learning_rate": 3.838036313172198e-06, "loss": 0.6434, "step": 328680 }, { "epoch": 28.47, "learning_rate": 3.835867931074361e-06, "loss": 0.5777, "step": 328690 }, { "epoch": 28.47, "learning_rate": 3.833699548976524e-06, "loss": 0.5896, "step": 328700 }, { "epoch": 28.47, "learning_rate": 3.831531166878686e-06, "loss": 0.7076, "step": 328710 }, { "epoch": 28.47, "learning_rate": 3.829362784780849e-06, "loss": 0.5124, "step": 328720 }, { "epoch": 28.47, "learning_rate": 3.827194402683011e-06, "loss": 0.6015, "step": 328730 }, { "epoch": 28.47, "learning_rate": 3.825026020585173e-06, "loss": 0.5576, "step": 328740 }, { "epoch": 28.47, "learning_rate": 3.822857638487336e-06, "loss": 0.6267, "step": 328750 }, { "epoch": 28.47, "learning_rate": 3.8206892563894985e-06, "loss": 0.6487, "step": 328760 }, { "epoch": 28.47, "learning_rate": 3.8185208742916615e-06, "loss": 0.5782, "step": 328770 }, { "epoch": 28.47, "learning_rate": 3.816352492193824e-06, "loss": 0.5346, "step": 328780 }, { "epoch": 28.48, "learning_rate": 3.8141841100959866e-06, "loss": 0.6067, "step": 328790 }, { "epoch": 28.48, "learning_rate": 3.812015727998149e-06, "loss": 0.6243, "step": 328800 }, { "epoch": 28.48, "learning_rate": 3.8098473459003118e-06, "loss": 0.535, "step": 328810 }, { "epoch": 28.48, "learning_rate": 3.807678963802475e-06, "loss": 0.536, "step": 328820 }, { "epoch": 28.48, "learning_rate": 3.8055105817046374e-06, "loss": 0.7251, "step": 328830 }, { "epoch": 28.48, "learning_rate": 3.8033421996068e-06, "loss": 0.5262, "step": 328840 }, { "epoch": 28.48, "learning_rate": 3.8011738175089625e-06, "loss": 0.4858, "step": 328850 }, { "epoch": 28.48, "learning_rate": 3.799005435411125e-06, "loss": 0.5662, "step": 328860 }, { "epoch": 28.48, "learning_rate": 3.7968370533132873e-06, "loss": 0.5238, "step": 328870 }, { "epoch": 28.48, "learning_rate": 3.79466867121545e-06, "loss": 0.5999, "step": 328880 }, { "epoch": 28.48, "learning_rate": 3.7925002891176124e-06, "loss": 0.5132, "step": 328890 }, { "epoch": 28.48, "learning_rate": 3.790331907019775e-06, "loss": 0.5597, "step": 328900 }, { "epoch": 28.49, "learning_rate": 3.788163524921938e-06, "loss": 0.5496, "step": 328910 }, { "epoch": 28.49, "learning_rate": 3.7859951428241006e-06, "loss": 0.5258, "step": 328920 }, { "epoch": 28.49, "learning_rate": 3.783826760726263e-06, "loss": 0.5148, "step": 328930 }, { "epoch": 28.49, "learning_rate": 3.7816583786284257e-06, "loss": 0.5997, "step": 328940 }, { "epoch": 28.49, "learning_rate": 3.7794899965305883e-06, "loss": 0.4371, "step": 328950 }, { "epoch": 28.49, "learning_rate": 3.777321614432751e-06, "loss": 0.6509, "step": 328960 }, { "epoch": 28.49, "learning_rate": 3.7751532323349135e-06, "loss": 0.519, "step": 328970 }, { "epoch": 28.49, "learning_rate": 3.7729848502370765e-06, "loss": 0.6046, "step": 328980 }, { "epoch": 28.49, "learning_rate": 3.770816468139239e-06, "loss": 0.49, "step": 328990 }, { "epoch": 28.49, "learning_rate": 3.7686480860414012e-06, "loss": 0.6303, "step": 329000 }, { "epoch": 28.49, "learning_rate": 3.766479703943564e-06, "loss": 0.6588, "step": 329010 }, { "epoch": 28.5, "learning_rate": 3.7643113218457264e-06, "loss": 0.559, "step": 329020 }, { "epoch": 28.5, "learning_rate": 3.762142939747889e-06, "loss": 0.4498, "step": 329030 }, { "epoch": 28.5, "learning_rate": 3.7599745576500515e-06, "loss": 0.615, "step": 329040 }, { "epoch": 28.5, "learning_rate": 3.757806175552214e-06, "loss": 0.7061, "step": 329050 }, { "epoch": 28.5, "learning_rate": 3.7556377934543767e-06, "loss": 0.5589, "step": 329060 }, { "epoch": 28.5, "learning_rate": 3.7534694113565397e-06, "loss": 0.5981, "step": 329070 }, { "epoch": 28.5, "learning_rate": 3.7513010292587023e-06, "loss": 0.5173, "step": 329080 }, { "epoch": 28.5, "learning_rate": 3.749132647160865e-06, "loss": 0.607, "step": 329090 }, { "epoch": 28.5, "learning_rate": 3.7469642650630275e-06, "loss": 0.5085, "step": 329100 }, { "epoch": 28.5, "learning_rate": 3.74479588296519e-06, "loss": 0.5088, "step": 329110 }, { "epoch": 28.5, "learning_rate": 3.7426275008673526e-06, "loss": 0.5783, "step": 329120 }, { "epoch": 28.5, "learning_rate": 3.7404591187695148e-06, "loss": 0.6625, "step": 329130 }, { "epoch": 28.51, "learning_rate": 3.7382907366716773e-06, "loss": 0.6222, "step": 329140 }, { "epoch": 28.51, "learning_rate": 3.7361223545738404e-06, "loss": 0.5518, "step": 329150 }, { "epoch": 28.51, "learning_rate": 3.733953972476003e-06, "loss": 0.5196, "step": 329160 }, { "epoch": 28.51, "learning_rate": 3.7317855903781655e-06, "loss": 0.7215, "step": 329170 }, { "epoch": 28.51, "learning_rate": 3.729617208280328e-06, "loss": 0.7205, "step": 329180 }, { "epoch": 28.51, "learning_rate": 3.7274488261824907e-06, "loss": 0.5072, "step": 329190 }, { "epoch": 28.51, "learning_rate": 3.7252804440846533e-06, "loss": 0.5476, "step": 329200 }, { "epoch": 28.51, "learning_rate": 3.723112061986816e-06, "loss": 0.4979, "step": 329210 }, { "epoch": 28.51, "learning_rate": 3.720943679888979e-06, "loss": 0.5398, "step": 329220 }, { "epoch": 28.51, "learning_rate": 3.7187752977911414e-06, "loss": 0.5919, "step": 329230 }, { "epoch": 28.51, "learning_rate": 3.716606915693304e-06, "loss": 0.5767, "step": 329240 }, { "epoch": 28.52, "learning_rate": 3.7144385335954666e-06, "loss": 0.5283, "step": 329250 }, { "epoch": 28.52, "learning_rate": 3.7122701514976287e-06, "loss": 0.5349, "step": 329260 }, { "epoch": 28.52, "learning_rate": 3.7101017693997913e-06, "loss": 0.5419, "step": 329270 }, { "epoch": 28.52, "learning_rate": 3.707933387301954e-06, "loss": 0.5252, "step": 329280 }, { "epoch": 28.52, "learning_rate": 3.7057650052041165e-06, "loss": 0.6231, "step": 329290 }, { "epoch": 28.52, "learning_rate": 3.703596623106279e-06, "loss": 0.4716, "step": 329300 }, { "epoch": 28.52, "learning_rate": 3.701428241008442e-06, "loss": 0.661, "step": 329310 }, { "epoch": 28.52, "learning_rate": 3.6992598589106046e-06, "loss": 0.5153, "step": 329320 }, { "epoch": 28.52, "learning_rate": 3.6970914768127672e-06, "loss": 0.6094, "step": 329330 }, { "epoch": 28.52, "learning_rate": 3.69492309471493e-06, "loss": 0.7732, "step": 329340 }, { "epoch": 28.52, "learning_rate": 3.6927547126170924e-06, "loss": 0.6527, "step": 329350 }, { "epoch": 28.52, "learning_rate": 3.690586330519255e-06, "loss": 0.6972, "step": 329360 }, { "epoch": 28.53, "learning_rate": 3.6884179484214175e-06, "loss": 0.6064, "step": 329370 }, { "epoch": 28.53, "learning_rate": 3.6862495663235805e-06, "loss": 0.6755, "step": 329380 }, { "epoch": 28.53, "learning_rate": 3.6840811842257427e-06, "loss": 0.6167, "step": 329390 }, { "epoch": 28.53, "learning_rate": 3.6819128021279053e-06, "loss": 0.6513, "step": 329400 }, { "epoch": 28.53, "learning_rate": 3.679744420030068e-06, "loss": 0.6866, "step": 329410 }, { "epoch": 28.53, "learning_rate": 3.6775760379322304e-06, "loss": 0.4612, "step": 329420 }, { "epoch": 28.53, "learning_rate": 3.675407655834393e-06, "loss": 0.4932, "step": 329430 }, { "epoch": 28.53, "learning_rate": 3.6732392737365556e-06, "loss": 0.5886, "step": 329440 }, { "epoch": 28.53, "learning_rate": 3.671070891638718e-06, "loss": 0.6115, "step": 329450 }, { "epoch": 28.53, "learning_rate": 3.668902509540881e-06, "loss": 0.6279, "step": 329460 }, { "epoch": 28.53, "learning_rate": 3.6667341274430438e-06, "loss": 0.7157, "step": 329470 }, { "epoch": 28.54, "learning_rate": 3.6645657453452063e-06, "loss": 0.5119, "step": 329480 }, { "epoch": 28.54, "learning_rate": 3.662397363247369e-06, "loss": 0.5516, "step": 329490 }, { "epoch": 28.54, "learning_rate": 3.6602289811495315e-06, "loss": 0.584, "step": 329500 }, { "epoch": 28.54, "learning_rate": 3.6580605990516937e-06, "loss": 0.5034, "step": 329510 }, { "epoch": 28.54, "learning_rate": 3.6558922169538562e-06, "loss": 0.6834, "step": 329520 }, { "epoch": 28.54, "learning_rate": 3.653723834856019e-06, "loss": 0.6444, "step": 329530 }, { "epoch": 28.54, "learning_rate": 3.6515554527581814e-06, "loss": 0.3924, "step": 329540 }, { "epoch": 28.54, "learning_rate": 3.6493870706603444e-06, "loss": 0.5654, "step": 329550 }, { "epoch": 28.54, "learning_rate": 3.647218688562507e-06, "loss": 0.6163, "step": 329560 }, { "epoch": 28.54, "learning_rate": 3.6450503064646696e-06, "loss": 0.666, "step": 329570 }, { "epoch": 28.54, "learning_rate": 3.642881924366832e-06, "loss": 0.597, "step": 329580 }, { "epoch": 28.54, "learning_rate": 3.6407135422689947e-06, "loss": 0.5471, "step": 329590 }, { "epoch": 28.55, "learning_rate": 3.6385451601711573e-06, "loss": 0.5613, "step": 329600 }, { "epoch": 28.55, "learning_rate": 3.63637677807332e-06, "loss": 0.818, "step": 329610 }, { "epoch": 28.55, "learning_rate": 3.634208395975483e-06, "loss": 0.5578, "step": 329620 }, { "epoch": 28.55, "learning_rate": 3.6320400138776455e-06, "loss": 0.5589, "step": 329630 }, { "epoch": 28.55, "learning_rate": 3.6298716317798076e-06, "loss": 0.654, "step": 329640 }, { "epoch": 28.55, "learning_rate": 3.62770324968197e-06, "loss": 0.5923, "step": 329650 }, { "epoch": 28.55, "learning_rate": 3.6255348675841328e-06, "loss": 0.61, "step": 329660 }, { "epoch": 28.55, "learning_rate": 3.6233664854862954e-06, "loss": 0.5587, "step": 329670 }, { "epoch": 28.55, "learning_rate": 3.621198103388458e-06, "loss": 0.5602, "step": 329680 }, { "epoch": 28.55, "learning_rate": 3.6190297212906205e-06, "loss": 0.644, "step": 329690 }, { "epoch": 28.55, "learning_rate": 3.616861339192783e-06, "loss": 0.6415, "step": 329700 }, { "epoch": 28.55, "learning_rate": 3.614692957094946e-06, "loss": 0.6355, "step": 329710 }, { "epoch": 28.56, "learning_rate": 3.6125245749971087e-06, "loss": 0.656, "step": 329720 }, { "epoch": 28.56, "learning_rate": 3.6103561928992713e-06, "loss": 0.4817, "step": 329730 }, { "epoch": 28.56, "learning_rate": 3.608187810801434e-06, "loss": 0.6374, "step": 329740 }, { "epoch": 28.56, "learning_rate": 3.6060194287035964e-06, "loss": 0.5481, "step": 329750 }, { "epoch": 28.56, "learning_rate": 3.603851046605759e-06, "loss": 0.6682, "step": 329760 }, { "epoch": 28.56, "learning_rate": 3.601682664507921e-06, "loss": 0.4399, "step": 329770 }, { "epoch": 28.56, "learning_rate": 3.5995142824100837e-06, "loss": 0.6121, "step": 329780 }, { "epoch": 28.56, "learning_rate": 3.5973459003122467e-06, "loss": 0.587, "step": 329790 }, { "epoch": 28.56, "learning_rate": 3.5951775182144093e-06, "loss": 0.5538, "step": 329800 }, { "epoch": 28.56, "learning_rate": 3.593009136116572e-06, "loss": 0.5793, "step": 329810 }, { "epoch": 28.56, "learning_rate": 3.5908407540187345e-06, "loss": 0.6808, "step": 329820 }, { "epoch": 28.57, "learning_rate": 3.588672371920897e-06, "loss": 0.5616, "step": 329830 }, { "epoch": 28.57, "learning_rate": 3.5865039898230596e-06, "loss": 0.6585, "step": 329840 }, { "epoch": 28.57, "learning_rate": 3.5843356077252222e-06, "loss": 0.6097, "step": 329850 }, { "epoch": 28.57, "learning_rate": 3.5821672256273852e-06, "loss": 0.5274, "step": 329860 }, { "epoch": 28.57, "learning_rate": 3.579998843529548e-06, "loss": 0.6289, "step": 329870 }, { "epoch": 28.57, "learning_rate": 3.5778304614317104e-06, "loss": 0.4964, "step": 329880 }, { "epoch": 28.57, "learning_rate": 3.575662079333873e-06, "loss": 0.5985, "step": 329890 }, { "epoch": 28.57, "learning_rate": 3.573493697236035e-06, "loss": 0.5305, "step": 329900 }, { "epoch": 28.57, "learning_rate": 3.5713253151381977e-06, "loss": 0.498, "step": 329910 }, { "epoch": 28.57, "learning_rate": 3.5691569330403603e-06, "loss": 0.4918, "step": 329920 }, { "epoch": 28.57, "learning_rate": 3.566988550942523e-06, "loss": 0.6111, "step": 329930 }, { "epoch": 28.57, "learning_rate": 3.5648201688446854e-06, "loss": 0.6018, "step": 329940 }, { "epoch": 28.58, "learning_rate": 3.5626517867468484e-06, "loss": 0.7151, "step": 329950 }, { "epoch": 28.58, "learning_rate": 3.560483404649011e-06, "loss": 0.5621, "step": 329960 }, { "epoch": 28.58, "learning_rate": 3.5583150225511736e-06, "loss": 0.5784, "step": 329970 }, { "epoch": 28.58, "learning_rate": 3.556146640453336e-06, "loss": 0.5353, "step": 329980 }, { "epoch": 28.58, "learning_rate": 3.5539782583554988e-06, "loss": 0.6324, "step": 329990 }, { "epoch": 28.58, "learning_rate": 3.5518098762576613e-06, "loss": 0.6609, "step": 330000 }, { "epoch": 28.58, "learning_rate": 3.549641494159824e-06, "loss": 0.6783, "step": 330010 }, { "epoch": 28.58, "learning_rate": 3.547473112061987e-06, "loss": 0.5521, "step": 330020 }, { "epoch": 28.58, "learning_rate": 3.545304729964149e-06, "loss": 0.5104, "step": 330030 }, { "epoch": 28.58, "learning_rate": 3.5431363478663117e-06, "loss": 0.4836, "step": 330040 }, { "epoch": 28.58, "learning_rate": 3.5409679657684742e-06, "loss": 0.5477, "step": 330050 }, { "epoch": 28.59, "learning_rate": 3.538799583670637e-06, "loss": 0.4735, "step": 330060 }, { "epoch": 28.59, "learning_rate": 3.5366312015727994e-06, "loss": 0.7646, "step": 330070 }, { "epoch": 28.59, "learning_rate": 3.534462819474962e-06, "loss": 0.622, "step": 330080 }, { "epoch": 28.59, "learning_rate": 3.5322944373771246e-06, "loss": 0.5244, "step": 330090 }, { "epoch": 28.59, "learning_rate": 3.5301260552792876e-06, "loss": 0.4815, "step": 330100 }, { "epoch": 28.59, "learning_rate": 3.52795767318145e-06, "loss": 0.5778, "step": 330110 }, { "epoch": 28.59, "learning_rate": 3.5257892910836127e-06, "loss": 0.5433, "step": 330120 }, { "epoch": 28.59, "learning_rate": 3.5236209089857753e-06, "loss": 0.5287, "step": 330130 }, { "epoch": 28.59, "learning_rate": 3.521452526887938e-06, "loss": 0.6009, "step": 330140 }, { "epoch": 28.59, "learning_rate": 3.5192841447901005e-06, "loss": 0.6047, "step": 330150 }, { "epoch": 28.59, "learning_rate": 3.5171157626922626e-06, "loss": 0.5063, "step": 330160 }, { "epoch": 28.59, "learning_rate": 3.514947380594425e-06, "loss": 0.5654, "step": 330170 }, { "epoch": 28.6, "learning_rate": 3.5127789984965878e-06, "loss": 0.4957, "step": 330180 }, { "epoch": 28.6, "learning_rate": 3.510610616398751e-06, "loss": 0.6735, "step": 330190 }, { "epoch": 28.6, "learning_rate": 3.5084422343009134e-06, "loss": 0.4553, "step": 330200 }, { "epoch": 28.6, "learning_rate": 3.506273852203076e-06, "loss": 0.6304, "step": 330210 }, { "epoch": 28.6, "learning_rate": 3.5041054701052385e-06, "loss": 0.4669, "step": 330220 }, { "epoch": 28.6, "learning_rate": 3.501937088007401e-06, "loss": 0.5988, "step": 330230 }, { "epoch": 28.6, "learning_rate": 3.4997687059095637e-06, "loss": 0.5713, "step": 330240 }, { "epoch": 28.6, "learning_rate": 3.4976003238117263e-06, "loss": 0.4969, "step": 330250 }, { "epoch": 28.6, "learning_rate": 3.4954319417138893e-06, "loss": 0.4917, "step": 330260 }, { "epoch": 28.6, "learning_rate": 3.493263559616052e-06, "loss": 0.5594, "step": 330270 }, { "epoch": 28.6, "learning_rate": 3.491095177518214e-06, "loss": 0.463, "step": 330280 }, { "epoch": 28.61, "learning_rate": 3.4889267954203766e-06, "loss": 0.5194, "step": 330290 }, { "epoch": 28.61, "learning_rate": 3.486758413322539e-06, "loss": 0.6101, "step": 330300 }, { "epoch": 28.61, "learning_rate": 3.4845900312247018e-06, "loss": 0.6306, "step": 330310 }, { "epoch": 28.61, "learning_rate": 3.4824216491268643e-06, "loss": 0.6946, "step": 330320 }, { "epoch": 28.61, "learning_rate": 3.480253267029027e-06, "loss": 0.494, "step": 330330 }, { "epoch": 28.61, "learning_rate": 3.4780848849311895e-06, "loss": 0.4823, "step": 330340 }, { "epoch": 28.61, "learning_rate": 3.4759165028333525e-06, "loss": 0.661, "step": 330350 }, { "epoch": 28.61, "learning_rate": 3.473748120735515e-06, "loss": 0.5504, "step": 330360 }, { "epoch": 28.61, "learning_rate": 3.4715797386376777e-06, "loss": 0.5242, "step": 330370 }, { "epoch": 28.61, "learning_rate": 3.4694113565398402e-06, "loss": 0.5503, "step": 330380 }, { "epoch": 28.61, "learning_rate": 3.467242974442003e-06, "loss": 0.5893, "step": 330390 }, { "epoch": 28.61, "learning_rate": 3.4650745923441654e-06, "loss": 0.4529, "step": 330400 }, { "epoch": 28.62, "learning_rate": 3.4629062102463276e-06, "loss": 0.5557, "step": 330410 }, { "epoch": 28.62, "learning_rate": 3.46073782814849e-06, "loss": 0.6279, "step": 330420 }, { "epoch": 28.62, "learning_rate": 3.458569446050653e-06, "loss": 0.5602, "step": 330430 }, { "epoch": 28.62, "learning_rate": 3.4564010639528157e-06, "loss": 0.4976, "step": 330440 }, { "epoch": 28.62, "learning_rate": 3.4542326818549783e-06, "loss": 0.616, "step": 330450 }, { "epoch": 28.62, "learning_rate": 3.452064299757141e-06, "loss": 0.5267, "step": 330460 }, { "epoch": 28.62, "learning_rate": 3.4498959176593035e-06, "loss": 0.6769, "step": 330470 }, { "epoch": 28.62, "learning_rate": 3.447727535561466e-06, "loss": 0.6089, "step": 330480 }, { "epoch": 28.62, "learning_rate": 3.4455591534636286e-06, "loss": 0.6049, "step": 330490 }, { "epoch": 28.62, "learning_rate": 3.4433907713657916e-06, "loss": 0.5369, "step": 330500 }, { "epoch": 28.62, "learning_rate": 3.441222389267954e-06, "loss": 0.6215, "step": 330510 }, { "epoch": 28.63, "learning_rate": 3.4390540071701168e-06, "loss": 0.5153, "step": 330520 }, { "epoch": 28.63, "learning_rate": 3.4368856250722794e-06, "loss": 0.5145, "step": 330530 }, { "epoch": 28.63, "learning_rate": 3.4347172429744415e-06, "loss": 0.5383, "step": 330540 }, { "epoch": 28.63, "learning_rate": 3.432548860876604e-06, "loss": 0.3774, "step": 330550 }, { "epoch": 28.63, "learning_rate": 3.4303804787787667e-06, "loss": 0.5023, "step": 330560 }, { "epoch": 28.63, "learning_rate": 3.4282120966809293e-06, "loss": 0.5726, "step": 330570 }, { "epoch": 28.63, "learning_rate": 3.426043714583092e-06, "loss": 0.4673, "step": 330580 }, { "epoch": 28.63, "learning_rate": 3.423875332485255e-06, "loss": 0.559, "step": 330590 }, { "epoch": 28.63, "learning_rate": 3.4217069503874174e-06, "loss": 0.6739, "step": 330600 }, { "epoch": 28.63, "learning_rate": 3.41953856828958e-06, "loss": 0.6406, "step": 330610 }, { "epoch": 28.63, "learning_rate": 3.4173701861917426e-06, "loss": 0.5687, "step": 330620 }, { "epoch": 28.63, "learning_rate": 3.415201804093905e-06, "loss": 0.6016, "step": 330630 }, { "epoch": 28.64, "learning_rate": 3.4130334219960677e-06, "loss": 0.4285, "step": 330640 }, { "epoch": 28.64, "learning_rate": 3.4108650398982303e-06, "loss": 0.535, "step": 330650 }, { "epoch": 28.64, "learning_rate": 3.4086966578003933e-06, "loss": 0.4076, "step": 330660 }, { "epoch": 28.64, "learning_rate": 3.406528275702555e-06, "loss": 0.5144, "step": 330670 }, { "epoch": 28.64, "learning_rate": 3.404359893604718e-06, "loss": 0.4443, "step": 330680 }, { "epoch": 28.64, "learning_rate": 3.4021915115068806e-06, "loss": 0.4934, "step": 330690 }, { "epoch": 28.64, "learning_rate": 3.4000231294090432e-06, "loss": 0.4533, "step": 330700 }, { "epoch": 28.64, "learning_rate": 3.397854747311206e-06, "loss": 0.7013, "step": 330710 }, { "epoch": 28.64, "learning_rate": 3.3956863652133684e-06, "loss": 0.6495, "step": 330720 }, { "epoch": 28.64, "learning_rate": 3.393517983115531e-06, "loss": 0.4858, "step": 330730 }, { "epoch": 28.64, "learning_rate": 3.3913496010176935e-06, "loss": 0.5079, "step": 330740 }, { "epoch": 28.65, "learning_rate": 3.3891812189198565e-06, "loss": 0.5486, "step": 330750 }, { "epoch": 28.65, "learning_rate": 3.387012836822019e-06, "loss": 0.5345, "step": 330760 }, { "epoch": 28.65, "learning_rate": 3.3848444547241817e-06, "loss": 0.5318, "step": 330770 }, { "epoch": 28.65, "learning_rate": 3.3826760726263443e-06, "loss": 0.5019, "step": 330780 }, { "epoch": 28.65, "learning_rate": 3.380507690528507e-06, "loss": 0.6511, "step": 330790 }, { "epoch": 28.65, "learning_rate": 3.378339308430669e-06, "loss": 0.687, "step": 330800 }, { "epoch": 28.65, "learning_rate": 3.3761709263328316e-06, "loss": 0.7018, "step": 330810 }, { "epoch": 28.65, "learning_rate": 3.374002544234994e-06, "loss": 0.6551, "step": 330820 }, { "epoch": 28.65, "learning_rate": 3.371834162137157e-06, "loss": 0.7416, "step": 330830 }, { "epoch": 28.65, "learning_rate": 3.3696657800393198e-06, "loss": 0.6129, "step": 330840 }, { "epoch": 28.65, "learning_rate": 3.3674973979414823e-06, "loss": 0.5236, "step": 330850 }, { "epoch": 28.65, "learning_rate": 3.365329015843645e-06, "loss": 0.6335, "step": 330860 }, { "epoch": 28.66, "learning_rate": 3.3631606337458075e-06, "loss": 0.4928, "step": 330870 }, { "epoch": 28.66, "learning_rate": 3.36099225164797e-06, "loss": 0.5269, "step": 330880 }, { "epoch": 28.66, "learning_rate": 3.3588238695501327e-06, "loss": 0.5209, "step": 330890 }, { "epoch": 28.66, "learning_rate": 3.3566554874522957e-06, "loss": 0.5828, "step": 330900 }, { "epoch": 28.66, "learning_rate": 3.3544871053544582e-06, "loss": 0.5733, "step": 330910 }, { "epoch": 28.66, "learning_rate": 3.352318723256621e-06, "loss": 0.7108, "step": 330920 }, { "epoch": 28.66, "learning_rate": 3.350150341158783e-06, "loss": 0.656, "step": 330930 }, { "epoch": 28.66, "learning_rate": 3.3479819590609456e-06, "loss": 0.6027, "step": 330940 }, { "epoch": 28.66, "learning_rate": 3.345813576963108e-06, "loss": 0.5835, "step": 330950 }, { "epoch": 28.66, "learning_rate": 3.3436451948652707e-06, "loss": 0.6131, "step": 330960 }, { "epoch": 28.66, "learning_rate": 3.3414768127674333e-06, "loss": 0.6054, "step": 330970 }, { "epoch": 28.66, "learning_rate": 3.339308430669596e-06, "loss": 0.7113, "step": 330980 }, { "epoch": 28.67, "learning_rate": 3.337140048571759e-06, "loss": 0.5525, "step": 330990 }, { "epoch": 28.67, "learning_rate": 3.3349716664739215e-06, "loss": 0.6303, "step": 331000 }, { "epoch": 28.67, "learning_rate": 3.332803284376084e-06, "loss": 0.5903, "step": 331010 }, { "epoch": 28.67, "learning_rate": 3.3306349022782466e-06, "loss": 0.468, "step": 331020 }, { "epoch": 28.67, "learning_rate": 3.328466520180409e-06, "loss": 0.5673, "step": 331030 }, { "epoch": 28.67, "learning_rate": 3.3262981380825718e-06, "loss": 0.5451, "step": 331040 }, { "epoch": 28.67, "learning_rate": 3.324129755984734e-06, "loss": 0.6555, "step": 331050 }, { "epoch": 28.67, "learning_rate": 3.3219613738868965e-06, "loss": 0.6267, "step": 331060 }, { "epoch": 28.67, "learning_rate": 3.3197929917890595e-06, "loss": 0.5401, "step": 331070 }, { "epoch": 28.67, "learning_rate": 3.317624609691222e-06, "loss": 0.5548, "step": 331080 }, { "epoch": 28.67, "learning_rate": 3.3154562275933847e-06, "loss": 0.6089, "step": 331090 }, { "epoch": 28.68, "learning_rate": 3.3132878454955473e-06, "loss": 0.4265, "step": 331100 }, { "epoch": 28.68, "learning_rate": 3.31111946339771e-06, "loss": 0.6857, "step": 331110 }, { "epoch": 28.68, "learning_rate": 3.3089510812998724e-06, "loss": 0.5644, "step": 331120 }, { "epoch": 28.68, "learning_rate": 3.306782699202035e-06, "loss": 0.7015, "step": 331130 }, { "epoch": 28.68, "learning_rate": 3.304614317104198e-06, "loss": 0.5784, "step": 331140 }, { "epoch": 28.68, "learning_rate": 3.3024459350063606e-06, "loss": 0.4665, "step": 331150 }, { "epoch": 28.68, "learning_rate": 3.300277552908523e-06, "loss": 0.5613, "step": 331160 }, { "epoch": 28.68, "learning_rate": 3.2981091708106857e-06, "loss": 0.5101, "step": 331170 }, { "epoch": 28.68, "learning_rate": 3.295940788712848e-06, "loss": 0.5241, "step": 331180 }, { "epoch": 28.68, "learning_rate": 3.2937724066150105e-06, "loss": 0.6096, "step": 331190 }, { "epoch": 28.68, "learning_rate": 3.291604024517173e-06, "loss": 0.5286, "step": 331200 }, { "epoch": 28.68, "learning_rate": 3.2894356424193356e-06, "loss": 0.5656, "step": 331210 }, { "epoch": 28.69, "learning_rate": 3.2872672603214982e-06, "loss": 0.6095, "step": 331220 }, { "epoch": 28.69, "learning_rate": 3.2850988782236612e-06, "loss": 0.6329, "step": 331230 }, { "epoch": 28.69, "learning_rate": 3.282930496125824e-06, "loss": 0.5048, "step": 331240 }, { "epoch": 28.69, "learning_rate": 3.2807621140279864e-06, "loss": 0.5601, "step": 331250 }, { "epoch": 28.69, "learning_rate": 3.278593731930149e-06, "loss": 0.6841, "step": 331260 }, { "epoch": 28.69, "learning_rate": 3.2764253498323115e-06, "loss": 0.5649, "step": 331270 }, { "epoch": 28.69, "learning_rate": 3.274256967734474e-06, "loss": 0.4627, "step": 331280 }, { "epoch": 28.69, "learning_rate": 3.2720885856366367e-06, "loss": 0.6081, "step": 331290 }, { "epoch": 28.69, "learning_rate": 3.2699202035387997e-06, "loss": 0.439, "step": 331300 }, { "epoch": 28.69, "learning_rate": 3.2677518214409614e-06, "loss": 0.582, "step": 331310 }, { "epoch": 28.69, "learning_rate": 3.2655834393431244e-06, "loss": 0.5684, "step": 331320 }, { "epoch": 28.7, "learning_rate": 3.263415057245287e-06, "loss": 0.5992, "step": 331330 }, { "epoch": 28.7, "learning_rate": 3.2612466751474496e-06, "loss": 0.5142, "step": 331340 }, { "epoch": 28.7, "learning_rate": 3.259078293049612e-06, "loss": 0.5777, "step": 331350 }, { "epoch": 28.7, "learning_rate": 3.2569099109517748e-06, "loss": 0.5669, "step": 331360 }, { "epoch": 28.7, "learning_rate": 3.2547415288539373e-06, "loss": 0.742, "step": 331370 }, { "epoch": 28.7, "learning_rate": 3.2525731467561e-06, "loss": 0.4817, "step": 331380 }, { "epoch": 28.7, "learning_rate": 3.250404764658263e-06, "loss": 0.5246, "step": 331390 }, { "epoch": 28.7, "learning_rate": 3.2482363825604255e-06, "loss": 0.5033, "step": 331400 }, { "epoch": 28.7, "learning_rate": 3.246068000462588e-06, "loss": 0.6423, "step": 331410 }, { "epoch": 28.7, "learning_rate": 3.2438996183647507e-06, "loss": 0.5516, "step": 331420 }, { "epoch": 28.7, "learning_rate": 3.2417312362669133e-06, "loss": 0.5152, "step": 331430 }, { "epoch": 28.7, "learning_rate": 3.2395628541690754e-06, "loss": 0.6183, "step": 331440 }, { "epoch": 28.71, "learning_rate": 3.237394472071238e-06, "loss": 0.487, "step": 331450 }, { "epoch": 28.71, "learning_rate": 3.2352260899734006e-06, "loss": 0.5776, "step": 331460 }, { "epoch": 28.71, "learning_rate": 3.2330577078755636e-06, "loss": 0.5292, "step": 331470 }, { "epoch": 28.71, "learning_rate": 3.230889325777726e-06, "loss": 0.6257, "step": 331480 }, { "epoch": 28.71, "learning_rate": 3.2287209436798887e-06, "loss": 0.6178, "step": 331490 }, { "epoch": 28.71, "learning_rate": 3.2265525615820513e-06, "loss": 0.5766, "step": 331500 }, { "epoch": 28.71, "learning_rate": 3.224384179484214e-06, "loss": 0.5833, "step": 331510 }, { "epoch": 28.71, "learning_rate": 3.2222157973863765e-06, "loss": 0.596, "step": 331520 }, { "epoch": 28.71, "learning_rate": 3.220047415288539e-06, "loss": 0.6962, "step": 331530 }, { "epoch": 28.71, "learning_rate": 3.217879033190702e-06, "loss": 0.5141, "step": 331540 }, { "epoch": 28.71, "learning_rate": 3.2157106510928646e-06, "loss": 0.547, "step": 331550 }, { "epoch": 28.72, "learning_rate": 3.2135422689950272e-06, "loss": 0.4566, "step": 331560 }, { "epoch": 28.72, "learning_rate": 3.2113738868971894e-06, "loss": 0.6296, "step": 331570 }, { "epoch": 28.72, "learning_rate": 3.209205504799352e-06, "loss": 0.6403, "step": 331580 }, { "epoch": 28.72, "learning_rate": 3.2070371227015145e-06, "loss": 0.6472, "step": 331590 }, { "epoch": 28.72, "learning_rate": 3.204868740603677e-06, "loss": 0.5291, "step": 331600 }, { "epoch": 28.72, "learning_rate": 3.2027003585058397e-06, "loss": 0.4164, "step": 331610 }, { "epoch": 28.72, "learning_rate": 3.2005319764080023e-06, "loss": 0.5803, "step": 331620 }, { "epoch": 28.72, "learning_rate": 3.1983635943101653e-06, "loss": 0.6009, "step": 331630 }, { "epoch": 28.72, "learning_rate": 3.196195212212328e-06, "loss": 0.534, "step": 331640 }, { "epoch": 28.72, "learning_rate": 3.1940268301144904e-06, "loss": 0.5494, "step": 331650 }, { "epoch": 28.72, "learning_rate": 3.191858448016653e-06, "loss": 0.7263, "step": 331660 }, { "epoch": 28.72, "learning_rate": 3.1896900659188156e-06, "loss": 0.5743, "step": 331670 }, { "epoch": 28.73, "learning_rate": 3.187521683820978e-06, "loss": 0.6003, "step": 331680 }, { "epoch": 28.73, "learning_rate": 3.1853533017231408e-06, "loss": 0.6752, "step": 331690 }, { "epoch": 28.73, "learning_rate": 3.183184919625303e-06, "loss": 0.4664, "step": 331700 }, { "epoch": 28.73, "learning_rate": 3.181016537527466e-06, "loss": 0.7028, "step": 331710 }, { "epoch": 28.73, "learning_rate": 3.1788481554296285e-06, "loss": 0.5684, "step": 331720 }, { "epoch": 28.73, "learning_rate": 3.176679773331791e-06, "loss": 0.5409, "step": 331730 }, { "epoch": 28.73, "learning_rate": 3.1745113912339537e-06, "loss": 0.5282, "step": 331740 }, { "epoch": 28.73, "learning_rate": 3.1723430091361162e-06, "loss": 0.7714, "step": 331750 }, { "epoch": 28.73, "learning_rate": 3.170174627038279e-06, "loss": 0.5911, "step": 331760 }, { "epoch": 28.73, "learning_rate": 3.1680062449404414e-06, "loss": 0.4694, "step": 331770 }, { "epoch": 28.73, "learning_rate": 3.1658378628426044e-06, "loss": 0.4486, "step": 331780 }, { "epoch": 28.74, "learning_rate": 3.163669480744767e-06, "loss": 0.5574, "step": 331790 }, { "epoch": 28.74, "learning_rate": 3.1615010986469296e-06, "loss": 0.578, "step": 331800 }, { "epoch": 28.74, "learning_rate": 3.159332716549092e-06, "loss": 0.6231, "step": 331810 }, { "epoch": 28.74, "learning_rate": 3.1571643344512543e-06, "loss": 0.3784, "step": 331820 }, { "epoch": 28.74, "learning_rate": 3.154995952353417e-06, "loss": 0.6038, "step": 331830 }, { "epoch": 28.74, "learning_rate": 3.1528275702555795e-06, "loss": 0.4947, "step": 331840 }, { "epoch": 28.74, "learning_rate": 3.150659188157742e-06, "loss": 0.5785, "step": 331850 }, { "epoch": 28.74, "learning_rate": 3.1484908060599046e-06, "loss": 0.6075, "step": 331860 }, { "epoch": 28.74, "learning_rate": 3.1463224239620676e-06, "loss": 0.6565, "step": 331870 }, { "epoch": 28.74, "learning_rate": 3.14415404186423e-06, "loss": 0.5914, "step": 331880 }, { "epoch": 28.74, "learning_rate": 3.1419856597663928e-06, "loss": 0.5158, "step": 331890 }, { "epoch": 28.74, "learning_rate": 3.1398172776685554e-06, "loss": 0.4935, "step": 331900 }, { "epoch": 28.75, "learning_rate": 3.137648895570718e-06, "loss": 0.5405, "step": 331910 }, { "epoch": 28.75, "learning_rate": 3.1354805134728805e-06, "loss": 0.5813, "step": 331920 }, { "epoch": 28.75, "learning_rate": 3.133312131375043e-06, "loss": 0.4233, "step": 331930 }, { "epoch": 28.75, "learning_rate": 3.131143749277206e-06, "loss": 0.515, "step": 331940 }, { "epoch": 28.75, "learning_rate": 3.128975367179368e-06, "loss": 0.5818, "step": 331950 }, { "epoch": 28.75, "learning_rate": 3.126806985081531e-06, "loss": 0.6297, "step": 331960 }, { "epoch": 28.75, "learning_rate": 3.1246386029836934e-06, "loss": 0.574, "step": 331970 }, { "epoch": 28.75, "learning_rate": 3.122470220885856e-06, "loss": 0.5905, "step": 331980 }, { "epoch": 28.75, "learning_rate": 3.1203018387880186e-06, "loss": 0.5872, "step": 331990 }, { "epoch": 28.75, "learning_rate": 3.118133456690181e-06, "loss": 0.5798, "step": 332000 }, { "epoch": 28.75, "learning_rate": 3.1159650745923437e-06, "loss": 0.6107, "step": 332010 }, { "epoch": 28.76, "learning_rate": 3.1137966924945063e-06, "loss": 0.5558, "step": 332020 }, { "epoch": 28.76, "learning_rate": 3.1116283103966693e-06, "loss": 0.4705, "step": 332030 }, { "epoch": 28.76, "learning_rate": 3.109459928298832e-06, "loss": 0.5348, "step": 332040 }, { "epoch": 28.76, "learning_rate": 3.1072915462009945e-06, "loss": 0.6183, "step": 332050 }, { "epoch": 28.76, "learning_rate": 3.105123164103157e-06, "loss": 0.7333, "step": 332060 }, { "epoch": 28.76, "learning_rate": 3.1029547820053196e-06, "loss": 0.5046, "step": 332070 }, { "epoch": 28.76, "learning_rate": 3.100786399907482e-06, "loss": 0.4907, "step": 332080 }, { "epoch": 28.76, "learning_rate": 3.0986180178096444e-06, "loss": 0.6784, "step": 332090 }, { "epoch": 28.76, "learning_rate": 3.096449635711807e-06, "loss": 0.5615, "step": 332100 }, { "epoch": 28.76, "learning_rate": 3.09428125361397e-06, "loss": 0.482, "step": 332110 }, { "epoch": 28.76, "learning_rate": 3.0921128715161325e-06, "loss": 0.6065, "step": 332120 }, { "epoch": 28.76, "learning_rate": 3.089944489418295e-06, "loss": 0.6172, "step": 332130 }, { "epoch": 28.77, "learning_rate": 3.0877761073204577e-06, "loss": 0.6346, "step": 332140 }, { "epoch": 28.77, "learning_rate": 3.0856077252226203e-06, "loss": 0.6322, "step": 332150 }, { "epoch": 28.77, "learning_rate": 3.083439343124783e-06, "loss": 0.5664, "step": 332160 }, { "epoch": 28.77, "learning_rate": 3.0812709610269454e-06, "loss": 0.6137, "step": 332170 }, { "epoch": 28.77, "learning_rate": 3.0791025789291084e-06, "loss": 0.6203, "step": 332180 }, { "epoch": 28.77, "learning_rate": 3.076934196831271e-06, "loss": 0.6085, "step": 332190 }, { "epoch": 28.77, "learning_rate": 3.0747658147334336e-06, "loss": 0.6224, "step": 332200 }, { "epoch": 28.77, "learning_rate": 3.0725974326355958e-06, "loss": 0.6284, "step": 332210 }, { "epoch": 28.77, "learning_rate": 3.0704290505377583e-06, "loss": 0.634, "step": 332220 }, { "epoch": 28.77, "learning_rate": 3.068260668439921e-06, "loss": 0.4941, "step": 332230 }, { "epoch": 28.77, "learning_rate": 3.0660922863420835e-06, "loss": 0.5787, "step": 332240 }, { "epoch": 28.77, "learning_rate": 3.063923904244246e-06, "loss": 0.5973, "step": 332250 }, { "epoch": 28.78, "learning_rate": 3.0617555221464087e-06, "loss": 0.4837, "step": 332260 }, { "epoch": 28.78, "learning_rate": 3.0595871400485717e-06, "loss": 0.6706, "step": 332270 }, { "epoch": 28.78, "learning_rate": 3.0574187579507342e-06, "loss": 0.6588, "step": 332280 }, { "epoch": 28.78, "learning_rate": 3.055250375852897e-06, "loss": 0.6234, "step": 332290 }, { "epoch": 28.78, "learning_rate": 3.0530819937550594e-06, "loss": 0.49, "step": 332300 }, { "epoch": 28.78, "learning_rate": 3.050913611657222e-06, "loss": 0.6681, "step": 332310 }, { "epoch": 28.78, "learning_rate": 3.0487452295593846e-06, "loss": 0.4566, "step": 332320 }, { "epoch": 28.78, "learning_rate": 3.046576847461547e-06, "loss": 0.5968, "step": 332330 }, { "epoch": 28.78, "learning_rate": 3.0444084653637093e-06, "loss": 0.6805, "step": 332340 }, { "epoch": 28.78, "learning_rate": 3.0422400832658723e-06, "loss": 0.6134, "step": 332350 }, { "epoch": 28.78, "learning_rate": 3.040071701168035e-06, "loss": 0.7145, "step": 332360 }, { "epoch": 28.79, "learning_rate": 3.0379033190701975e-06, "loss": 0.5325, "step": 332370 }, { "epoch": 28.79, "learning_rate": 3.03573493697236e-06, "loss": 0.613, "step": 332380 }, { "epoch": 28.79, "learning_rate": 3.0335665548745226e-06, "loss": 0.5882, "step": 332390 }, { "epoch": 28.79, "learning_rate": 3.031398172776685e-06, "loss": 0.5579, "step": 332400 }, { "epoch": 28.79, "learning_rate": 3.0292297906788478e-06, "loss": 0.5874, "step": 332410 }, { "epoch": 28.79, "learning_rate": 3.027061408581011e-06, "loss": 0.5246, "step": 332420 }, { "epoch": 28.79, "learning_rate": 3.0248930264831734e-06, "loss": 0.5902, "step": 332430 }, { "epoch": 28.79, "learning_rate": 3.022724644385336e-06, "loss": 0.6642, "step": 332440 }, { "epoch": 28.79, "learning_rate": 3.0205562622874985e-06, "loss": 0.6906, "step": 332450 }, { "epoch": 28.79, "learning_rate": 3.018387880189661e-06, "loss": 0.4854, "step": 332460 }, { "epoch": 28.79, "learning_rate": 3.0162194980918233e-06, "loss": 0.4649, "step": 332470 }, { "epoch": 28.79, "learning_rate": 3.014051115993986e-06, "loss": 0.6042, "step": 332480 }, { "epoch": 28.8, "learning_rate": 3.0118827338961484e-06, "loss": 0.5748, "step": 332490 }, { "epoch": 28.8, "learning_rate": 3.009714351798311e-06, "loss": 0.6071, "step": 332500 }, { "epoch": 28.8, "learning_rate": 3.007545969700474e-06, "loss": 0.6007, "step": 332510 }, { "epoch": 28.8, "learning_rate": 3.0053775876026366e-06, "loss": 0.5442, "step": 332520 }, { "epoch": 28.8, "learning_rate": 3.003209205504799e-06, "loss": 0.5492, "step": 332530 }, { "epoch": 28.8, "learning_rate": 3.0010408234069617e-06, "loss": 0.5994, "step": 332540 }, { "epoch": 28.8, "learning_rate": 2.9988724413091243e-06, "loss": 0.6107, "step": 332550 }, { "epoch": 28.8, "learning_rate": 2.996704059211287e-06, "loss": 0.6734, "step": 332560 }, { "epoch": 28.8, "learning_rate": 2.9945356771134495e-06, "loss": 0.5091, "step": 332570 }, { "epoch": 28.8, "learning_rate": 2.9923672950156125e-06, "loss": 0.5673, "step": 332580 }, { "epoch": 28.8, "learning_rate": 2.9901989129177742e-06, "loss": 0.5274, "step": 332590 }, { "epoch": 28.81, "learning_rate": 2.9880305308199372e-06, "loss": 0.533, "step": 332600 }, { "epoch": 28.81, "learning_rate": 2.9858621487221e-06, "loss": 0.6088, "step": 332610 }, { "epoch": 28.81, "learning_rate": 2.9836937666242624e-06, "loss": 0.5062, "step": 332620 }, { "epoch": 28.81, "learning_rate": 2.981525384526425e-06, "loss": 0.5503, "step": 332630 }, { "epoch": 28.81, "learning_rate": 2.9793570024285875e-06, "loss": 0.5217, "step": 332640 }, { "epoch": 28.81, "learning_rate": 2.97718862033075e-06, "loss": 0.5797, "step": 332650 }, { "epoch": 28.81, "learning_rate": 2.9750202382329127e-06, "loss": 0.7215, "step": 332660 }, { "epoch": 28.81, "learning_rate": 2.9728518561350757e-06, "loss": 0.5017, "step": 332670 }, { "epoch": 28.81, "learning_rate": 2.9706834740372383e-06, "loss": 0.6464, "step": 332680 }, { "epoch": 28.81, "learning_rate": 2.968515091939401e-06, "loss": 0.5553, "step": 332690 }, { "epoch": 28.81, "learning_rate": 2.9663467098415635e-06, "loss": 0.5128, "step": 332700 }, { "epoch": 28.81, "learning_rate": 2.964178327743726e-06, "loss": 0.5009, "step": 332710 }, { "epoch": 28.82, "learning_rate": 2.962009945645888e-06, "loss": 0.5179, "step": 332720 }, { "epoch": 28.82, "learning_rate": 2.9598415635480508e-06, "loss": 0.6496, "step": 332730 }, { "epoch": 28.82, "learning_rate": 2.9576731814502133e-06, "loss": 0.4924, "step": 332740 }, { "epoch": 28.82, "learning_rate": 2.9555047993523764e-06, "loss": 0.5518, "step": 332750 }, { "epoch": 28.82, "learning_rate": 2.953336417254539e-06, "loss": 0.5645, "step": 332760 }, { "epoch": 28.82, "learning_rate": 2.9511680351567015e-06, "loss": 0.5232, "step": 332770 }, { "epoch": 28.82, "learning_rate": 2.948999653058864e-06, "loss": 0.4582, "step": 332780 }, { "epoch": 28.82, "learning_rate": 2.9468312709610267e-06, "loss": 0.4867, "step": 332790 }, { "epoch": 28.82, "learning_rate": 2.9446628888631893e-06, "loss": 0.6011, "step": 332800 }, { "epoch": 28.82, "learning_rate": 2.942494506765352e-06, "loss": 0.4291, "step": 332810 }, { "epoch": 28.82, "learning_rate": 2.940326124667515e-06, "loss": 0.6004, "step": 332820 }, { "epoch": 28.83, "learning_rate": 2.9381577425696774e-06, "loss": 0.5763, "step": 332830 }, { "epoch": 28.83, "learning_rate": 2.93598936047184e-06, "loss": 0.6177, "step": 332840 }, { "epoch": 28.83, "learning_rate": 2.933820978374002e-06, "loss": 0.6389, "step": 332850 }, { "epoch": 28.83, "learning_rate": 2.9316525962761647e-06, "loss": 0.6011, "step": 332860 }, { "epoch": 28.83, "learning_rate": 2.9294842141783273e-06, "loss": 0.4652, "step": 332870 }, { "epoch": 28.83, "learning_rate": 2.92731583208049e-06, "loss": 0.4817, "step": 332880 }, { "epoch": 28.83, "learning_rate": 2.9251474499826525e-06, "loss": 0.4714, "step": 332890 }, { "epoch": 28.83, "learning_rate": 2.922979067884815e-06, "loss": 0.6745, "step": 332900 }, { "epoch": 28.83, "learning_rate": 2.920810685786978e-06, "loss": 0.6491, "step": 332910 }, { "epoch": 28.83, "learning_rate": 2.9186423036891406e-06, "loss": 0.5432, "step": 332920 }, { "epoch": 28.83, "learning_rate": 2.9164739215913032e-06, "loss": 0.5046, "step": 332930 }, { "epoch": 28.83, "learning_rate": 2.914305539493466e-06, "loss": 0.6222, "step": 332940 }, { "epoch": 28.84, "learning_rate": 2.9121371573956284e-06, "loss": 0.547, "step": 332950 }, { "epoch": 28.84, "learning_rate": 2.909968775297791e-06, "loss": 0.6047, "step": 332960 }, { "epoch": 28.84, "learning_rate": 2.9078003931999535e-06, "loss": 0.4944, "step": 332970 }, { "epoch": 28.84, "learning_rate": 2.9056320111021157e-06, "loss": 0.5209, "step": 332980 }, { "epoch": 28.84, "learning_rate": 2.9034636290042787e-06, "loss": 0.6976, "step": 332990 }, { "epoch": 28.84, "learning_rate": 2.9012952469064413e-06, "loss": 0.5493, "step": 333000 }, { "epoch": 28.84, "learning_rate": 2.899126864808604e-06, "loss": 0.5547, "step": 333010 }, { "epoch": 28.84, "learning_rate": 2.8969584827107664e-06, "loss": 0.6997, "step": 333020 }, { "epoch": 28.84, "learning_rate": 2.894790100612929e-06, "loss": 0.5696, "step": 333030 }, { "epoch": 28.84, "learning_rate": 2.8926217185150916e-06, "loss": 0.5489, "step": 333040 }, { "epoch": 28.84, "learning_rate": 2.890453336417254e-06, "loss": 0.6005, "step": 333050 }, { "epoch": 28.85, "learning_rate": 2.888284954319417e-06, "loss": 0.5657, "step": 333060 }, { "epoch": 28.85, "learning_rate": 2.8861165722215798e-06, "loss": 0.4847, "step": 333070 }, { "epoch": 28.85, "learning_rate": 2.8839481901237423e-06, "loss": 0.4888, "step": 333080 }, { "epoch": 28.85, "learning_rate": 2.881779808025905e-06, "loss": 0.5637, "step": 333090 }, { "epoch": 28.85, "learning_rate": 2.8796114259280675e-06, "loss": 0.6111, "step": 333100 }, { "epoch": 28.85, "learning_rate": 2.8774430438302297e-06, "loss": 0.5589, "step": 333110 }, { "epoch": 28.85, "learning_rate": 2.8752746617323922e-06, "loss": 0.5219, "step": 333120 }, { "epoch": 28.85, "learning_rate": 2.873106279634555e-06, "loss": 0.4486, "step": 333130 }, { "epoch": 28.85, "learning_rate": 2.8709378975367174e-06, "loss": 0.5091, "step": 333140 }, { "epoch": 28.85, "learning_rate": 2.8687695154388804e-06, "loss": 0.6322, "step": 333150 }, { "epoch": 28.85, "learning_rate": 2.866601133341043e-06, "loss": 0.4675, "step": 333160 }, { "epoch": 28.85, "learning_rate": 2.8644327512432056e-06, "loss": 0.4503, "step": 333170 }, { "epoch": 28.86, "learning_rate": 2.862264369145368e-06, "loss": 0.6758, "step": 333180 }, { "epoch": 28.86, "learning_rate": 2.8600959870475307e-06, "loss": 0.5041, "step": 333190 }, { "epoch": 28.86, "learning_rate": 2.8579276049496933e-06, "loss": 0.5174, "step": 333200 }, { "epoch": 28.86, "learning_rate": 2.855759222851856e-06, "loss": 0.6509, "step": 333210 }, { "epoch": 28.86, "learning_rate": 2.853590840754019e-06, "loss": 0.6004, "step": 333220 }, { "epoch": 28.86, "learning_rate": 2.8514224586561806e-06, "loss": 0.3952, "step": 333230 }, { "epoch": 28.86, "learning_rate": 2.8492540765583436e-06, "loss": 0.5112, "step": 333240 }, { "epoch": 28.86, "learning_rate": 2.847085694460506e-06, "loss": 0.5642, "step": 333250 }, { "epoch": 28.86, "learning_rate": 2.8449173123626688e-06, "loss": 0.6971, "step": 333260 }, { "epoch": 28.86, "learning_rate": 2.8427489302648314e-06, "loss": 0.4913, "step": 333270 }, { "epoch": 28.86, "learning_rate": 2.840580548166994e-06, "loss": 0.5382, "step": 333280 }, { "epoch": 28.87, "learning_rate": 2.8384121660691565e-06, "loss": 0.5232, "step": 333290 }, { "epoch": 28.87, "learning_rate": 2.836243783971319e-06, "loss": 0.646, "step": 333300 }, { "epoch": 28.87, "learning_rate": 2.834075401873482e-06, "loss": 0.6831, "step": 333310 }, { "epoch": 28.87, "learning_rate": 2.8319070197756447e-06, "loss": 0.546, "step": 333320 }, { "epoch": 28.87, "learning_rate": 2.8297386376778073e-06, "loss": 0.7086, "step": 333330 }, { "epoch": 28.87, "learning_rate": 2.82757025557997e-06, "loss": 0.5627, "step": 333340 }, { "epoch": 28.87, "learning_rate": 2.8254018734821324e-06, "loss": 0.528, "step": 333350 }, { "epoch": 28.87, "learning_rate": 2.8232334913842946e-06, "loss": 0.6333, "step": 333360 }, { "epoch": 28.87, "learning_rate": 2.821065109286457e-06, "loss": 0.5242, "step": 333370 }, { "epoch": 28.87, "learning_rate": 2.8188967271886197e-06, "loss": 0.6266, "step": 333380 }, { "epoch": 28.87, "learning_rate": 2.8167283450907827e-06, "loss": 0.4935, "step": 333390 }, { "epoch": 28.87, "learning_rate": 2.8145599629929453e-06, "loss": 0.5347, "step": 333400 }, { "epoch": 28.88, "learning_rate": 2.812391580895108e-06, "loss": 0.5691, "step": 333410 }, { "epoch": 28.88, "learning_rate": 2.8102231987972705e-06, "loss": 0.6909, "step": 333420 }, { "epoch": 28.88, "learning_rate": 2.808054816699433e-06, "loss": 0.4705, "step": 333430 }, { "epoch": 28.88, "learning_rate": 2.8058864346015956e-06, "loss": 0.633, "step": 333440 }, { "epoch": 28.88, "learning_rate": 2.8037180525037582e-06, "loss": 0.6486, "step": 333450 }, { "epoch": 28.88, "learning_rate": 2.8015496704059212e-06, "loss": 0.5653, "step": 333460 }, { "epoch": 28.88, "learning_rate": 2.799381288308084e-06, "loss": 0.5863, "step": 333470 }, { "epoch": 28.88, "learning_rate": 2.7972129062102464e-06, "loss": 0.6721, "step": 333480 }, { "epoch": 28.88, "learning_rate": 2.7950445241124085e-06, "loss": 0.5713, "step": 333490 }, { "epoch": 28.88, "learning_rate": 2.792876142014571e-06, "loss": 0.6971, "step": 333500 }, { "epoch": 28.88, "learning_rate": 2.7907077599167337e-06, "loss": 0.65, "step": 333510 }, { "epoch": 28.88, "learning_rate": 2.7885393778188963e-06, "loss": 0.5157, "step": 333520 }, { "epoch": 28.89, "learning_rate": 2.786370995721059e-06, "loss": 0.5771, "step": 333530 }, { "epoch": 28.89, "learning_rate": 2.7842026136232214e-06, "loss": 0.4988, "step": 333540 }, { "epoch": 28.89, "learning_rate": 2.7820342315253844e-06, "loss": 0.4724, "step": 333550 }, { "epoch": 28.89, "learning_rate": 2.779865849427547e-06, "loss": 0.5563, "step": 333560 }, { "epoch": 28.89, "learning_rate": 2.7776974673297096e-06, "loss": 0.4739, "step": 333570 }, { "epoch": 28.89, "learning_rate": 2.775529085231872e-06, "loss": 0.624, "step": 333580 }, { "epoch": 28.89, "learning_rate": 2.7733607031340348e-06, "loss": 0.5567, "step": 333590 }, { "epoch": 28.89, "learning_rate": 2.7711923210361973e-06, "loss": 0.5658, "step": 333600 }, { "epoch": 28.89, "learning_rate": 2.76902393893836e-06, "loss": 0.6094, "step": 333610 }, { "epoch": 28.89, "learning_rate": 2.766855556840522e-06, "loss": 0.5711, "step": 333620 }, { "epoch": 28.89, "learning_rate": 2.764687174742685e-06, "loss": 0.6577, "step": 333630 }, { "epoch": 28.9, "learning_rate": 2.7625187926448477e-06, "loss": 0.5433, "step": 333640 }, { "epoch": 28.9, "learning_rate": 2.7603504105470102e-06, "loss": 0.4204, "step": 333650 }, { "epoch": 28.9, "learning_rate": 2.758182028449173e-06, "loss": 0.5014, "step": 333660 }, { "epoch": 28.9, "learning_rate": 2.7560136463513354e-06, "loss": 0.5114, "step": 333670 }, { "epoch": 28.9, "learning_rate": 2.753845264253498e-06, "loss": 0.5282, "step": 333680 }, { "epoch": 28.9, "learning_rate": 2.7516768821556606e-06, "loss": 0.431, "step": 333690 }, { "epoch": 28.9, "learning_rate": 2.7495085000578236e-06, "loss": 0.5451, "step": 333700 }, { "epoch": 28.9, "learning_rate": 2.747340117959986e-06, "loss": 0.5913, "step": 333710 }, { "epoch": 28.9, "learning_rate": 2.7451717358621487e-06, "loss": 0.6582, "step": 333720 }, { "epoch": 28.9, "learning_rate": 2.7430033537643113e-06, "loss": 0.5899, "step": 333730 }, { "epoch": 28.9, "learning_rate": 2.740834971666474e-06, "loss": 0.4963, "step": 333740 }, { "epoch": 28.9, "learning_rate": 2.738666589568636e-06, "loss": 0.7063, "step": 333750 }, { "epoch": 28.91, "learning_rate": 2.7364982074707986e-06, "loss": 0.682, "step": 333760 }, { "epoch": 28.91, "learning_rate": 2.734329825372961e-06, "loss": 0.5155, "step": 333770 }, { "epoch": 28.91, "learning_rate": 2.7321614432751238e-06, "loss": 0.5095, "step": 333780 }, { "epoch": 28.91, "learning_rate": 2.729993061177287e-06, "loss": 0.5015, "step": 333790 }, { "epoch": 28.91, "learning_rate": 2.7278246790794494e-06, "loss": 0.6614, "step": 333800 }, { "epoch": 28.91, "learning_rate": 2.725656296981612e-06, "loss": 0.5034, "step": 333810 }, { "epoch": 28.91, "learning_rate": 2.7234879148837745e-06, "loss": 0.5307, "step": 333820 }, { "epoch": 28.91, "learning_rate": 2.721319532785937e-06, "loss": 0.5712, "step": 333830 }, { "epoch": 28.91, "learning_rate": 2.7191511506880997e-06, "loss": 0.5093, "step": 333840 }, { "epoch": 28.91, "learning_rate": 2.7169827685902623e-06, "loss": 0.4897, "step": 333850 }, { "epoch": 28.91, "learning_rate": 2.7148143864924253e-06, "loss": 0.6888, "step": 333860 }, { "epoch": 28.92, "learning_rate": 2.712646004394588e-06, "loss": 0.491, "step": 333870 }, { "epoch": 28.92, "learning_rate": 2.71047762229675e-06, "loss": 0.5127, "step": 333880 }, { "epoch": 28.92, "learning_rate": 2.7083092401989126e-06, "loss": 0.5948, "step": 333890 }, { "epoch": 28.92, "learning_rate": 2.706140858101075e-06, "loss": 0.5733, "step": 333900 }, { "epoch": 28.92, "learning_rate": 2.7039724760032378e-06, "loss": 0.6816, "step": 333910 }, { "epoch": 28.92, "learning_rate": 2.7018040939054003e-06, "loss": 0.6699, "step": 333920 }, { "epoch": 28.92, "learning_rate": 2.699635711807563e-06, "loss": 0.6346, "step": 333930 }, { "epoch": 28.92, "learning_rate": 2.6974673297097255e-06, "loss": 0.5572, "step": 333940 }, { "epoch": 28.92, "learning_rate": 2.6952989476118885e-06, "loss": 0.5718, "step": 333950 }, { "epoch": 28.92, "learning_rate": 2.693130565514051e-06, "loss": 0.4622, "step": 333960 }, { "epoch": 28.92, "learning_rate": 2.6909621834162137e-06, "loss": 0.5553, "step": 333970 }, { "epoch": 28.92, "learning_rate": 2.6887938013183762e-06, "loss": 0.6579, "step": 333980 }, { "epoch": 28.93, "learning_rate": 2.686625419220539e-06, "loss": 0.6379, "step": 333990 }, { "epoch": 28.93, "learning_rate": 2.684457037122701e-06, "loss": 0.6442, "step": 334000 }, { "epoch": 28.93, "learning_rate": 2.6822886550248636e-06, "loss": 0.6476, "step": 334010 }, { "epoch": 28.93, "learning_rate": 2.680120272927026e-06, "loss": 0.6195, "step": 334020 }, { "epoch": 28.93, "learning_rate": 2.677951890829189e-06, "loss": 0.5433, "step": 334030 }, { "epoch": 28.93, "learning_rate": 2.6757835087313517e-06, "loss": 0.5586, "step": 334040 }, { "epoch": 28.93, "learning_rate": 2.6736151266335143e-06, "loss": 0.6023, "step": 334050 }, { "epoch": 28.93, "learning_rate": 2.671446744535677e-06, "loss": 0.5202, "step": 334060 }, { "epoch": 28.93, "learning_rate": 2.6692783624378395e-06, "loss": 0.6508, "step": 334070 }, { "epoch": 28.93, "learning_rate": 2.667109980340002e-06, "loss": 0.6016, "step": 334080 }, { "epoch": 28.93, "learning_rate": 2.6649415982421646e-06, "loss": 0.6209, "step": 334090 }, { "epoch": 28.94, "learning_rate": 2.6627732161443276e-06, "loss": 0.4894, "step": 334100 }, { "epoch": 28.94, "learning_rate": 2.66060483404649e-06, "loss": 0.5889, "step": 334110 }, { "epoch": 28.94, "learning_rate": 2.6584364519486528e-06, "loss": 0.6079, "step": 334120 }, { "epoch": 28.94, "learning_rate": 2.656268069850815e-06, "loss": 0.4965, "step": 334130 }, { "epoch": 28.94, "learning_rate": 2.6540996877529775e-06, "loss": 0.6057, "step": 334140 }, { "epoch": 28.94, "learning_rate": 2.65193130565514e-06, "loss": 0.5769, "step": 334150 }, { "epoch": 28.94, "learning_rate": 2.6497629235573027e-06, "loss": 0.585, "step": 334160 }, { "epoch": 28.94, "learning_rate": 2.6475945414594653e-06, "loss": 0.7149, "step": 334170 }, { "epoch": 28.94, "learning_rate": 2.645426159361628e-06, "loss": 0.4894, "step": 334180 }, { "epoch": 28.94, "learning_rate": 2.643257777263791e-06, "loss": 0.4782, "step": 334190 }, { "epoch": 28.94, "learning_rate": 2.6410893951659534e-06, "loss": 0.658, "step": 334200 }, { "epoch": 28.94, "learning_rate": 2.638921013068116e-06, "loss": 0.6612, "step": 334210 }, { "epoch": 28.95, "learning_rate": 2.6367526309702786e-06, "loss": 0.5349, "step": 334220 }, { "epoch": 28.95, "learning_rate": 2.634584248872441e-06, "loss": 0.5356, "step": 334230 }, { "epoch": 28.95, "learning_rate": 2.6324158667746037e-06, "loss": 0.7014, "step": 334240 }, { "epoch": 28.95, "learning_rate": 2.6302474846767663e-06, "loss": 0.4749, "step": 334250 }, { "epoch": 28.95, "learning_rate": 2.6280791025789285e-06, "loss": 0.6852, "step": 334260 }, { "epoch": 28.95, "learning_rate": 2.6259107204810915e-06, "loss": 0.5111, "step": 334270 }, { "epoch": 28.95, "learning_rate": 2.623742338383254e-06, "loss": 0.5247, "step": 334280 }, { "epoch": 28.95, "learning_rate": 2.6215739562854166e-06, "loss": 0.6642, "step": 334290 }, { "epoch": 28.95, "learning_rate": 2.6194055741875792e-06, "loss": 0.6714, "step": 334300 }, { "epoch": 28.95, "learning_rate": 2.617237192089742e-06, "loss": 0.7013, "step": 334310 }, { "epoch": 28.95, "learning_rate": 2.6150688099919044e-06, "loss": 0.5856, "step": 334320 }, { "epoch": 28.96, "learning_rate": 2.612900427894067e-06, "loss": 0.5315, "step": 334330 }, { "epoch": 28.96, "learning_rate": 2.61073204579623e-06, "loss": 0.4281, "step": 334340 }, { "epoch": 28.96, "learning_rate": 2.6085636636983925e-06, "loss": 0.6759, "step": 334350 }, { "epoch": 28.96, "learning_rate": 2.606395281600555e-06, "loss": 0.4902, "step": 334360 }, { "epoch": 28.96, "learning_rate": 2.6042268995027177e-06, "loss": 0.5375, "step": 334370 }, { "epoch": 28.96, "learning_rate": 2.6020585174048803e-06, "loss": 0.5545, "step": 334380 }, { "epoch": 28.96, "learning_rate": 2.5998901353070424e-06, "loss": 0.5539, "step": 334390 }, { "epoch": 28.96, "learning_rate": 2.597721753209205e-06, "loss": 0.3975, "step": 334400 }, { "epoch": 28.96, "learning_rate": 2.5955533711113676e-06, "loss": 0.5056, "step": 334410 }, { "epoch": 28.96, "learning_rate": 2.59338498901353e-06, "loss": 0.6269, "step": 334420 }, { "epoch": 28.96, "learning_rate": 2.591216606915693e-06, "loss": 0.5104, "step": 334430 }, { "epoch": 28.96, "learning_rate": 2.5890482248178558e-06, "loss": 0.6633, "step": 334440 }, { "epoch": 28.97, "learning_rate": 2.5868798427200183e-06, "loss": 0.7363, "step": 334450 }, { "epoch": 28.97, "learning_rate": 2.584711460622181e-06, "loss": 0.4617, "step": 334460 }, { "epoch": 28.97, "learning_rate": 2.5825430785243435e-06, "loss": 0.4963, "step": 334470 }, { "epoch": 28.97, "learning_rate": 2.580374696426506e-06, "loss": 0.5464, "step": 334480 }, { "epoch": 28.97, "learning_rate": 2.5782063143286687e-06, "loss": 0.5287, "step": 334490 }, { "epoch": 28.97, "learning_rate": 2.5760379322308317e-06, "loss": 0.7082, "step": 334500 }, { "epoch": 28.97, "learning_rate": 2.5738695501329942e-06, "loss": 0.489, "step": 334510 }, { "epoch": 28.97, "learning_rate": 2.5717011680351564e-06, "loss": 0.5027, "step": 334520 }, { "epoch": 28.97, "learning_rate": 2.569532785937319e-06, "loss": 0.4869, "step": 334530 }, { "epoch": 28.97, "learning_rate": 2.5673644038394816e-06, "loss": 0.4778, "step": 334540 }, { "epoch": 28.97, "learning_rate": 2.565196021741644e-06, "loss": 0.5784, "step": 334550 }, { "epoch": 28.98, "learning_rate": 2.5630276396438067e-06, "loss": 0.534, "step": 334560 }, { "epoch": 28.98, "learning_rate": 2.5608592575459693e-06, "loss": 0.6706, "step": 334570 }, { "epoch": 28.98, "learning_rate": 2.558690875448132e-06, "loss": 0.6872, "step": 334580 }, { "epoch": 28.98, "learning_rate": 2.556522493350295e-06, "loss": 0.5781, "step": 334590 }, { "epoch": 28.98, "learning_rate": 2.5543541112524575e-06, "loss": 0.5067, "step": 334600 }, { "epoch": 28.98, "learning_rate": 2.55218572915462e-06, "loss": 0.5292, "step": 334610 }, { "epoch": 28.98, "learning_rate": 2.5500173470567826e-06, "loss": 0.4886, "step": 334620 }, { "epoch": 28.98, "learning_rate": 2.547848964958945e-06, "loss": 0.6027, "step": 334630 }, { "epoch": 28.98, "learning_rate": 2.5456805828611078e-06, "loss": 0.6034, "step": 334640 }, { "epoch": 28.98, "learning_rate": 2.54351220076327e-06, "loss": 0.6645, "step": 334650 }, { "epoch": 28.98, "learning_rate": 2.5413438186654325e-06, "loss": 0.5527, "step": 334660 }, { "epoch": 28.98, "learning_rate": 2.5391754365675955e-06, "loss": 0.4857, "step": 334670 }, { "epoch": 28.99, "learning_rate": 2.537007054469758e-06, "loss": 0.5062, "step": 334680 }, { "epoch": 28.99, "learning_rate": 2.5348386723719207e-06, "loss": 0.5346, "step": 334690 }, { "epoch": 28.99, "learning_rate": 2.5326702902740833e-06, "loss": 0.6654, "step": 334700 }, { "epoch": 28.99, "learning_rate": 2.530501908176246e-06, "loss": 0.5256, "step": 334710 }, { "epoch": 28.99, "learning_rate": 2.5283335260784084e-06, "loss": 0.5172, "step": 334720 }, { "epoch": 28.99, "learning_rate": 2.526165143980571e-06, "loss": 0.6106, "step": 334730 }, { "epoch": 28.99, "learning_rate": 2.523996761882734e-06, "loss": 0.6499, "step": 334740 }, { "epoch": 28.99, "learning_rate": 2.5218283797848966e-06, "loss": 0.4518, "step": 334750 }, { "epoch": 28.99, "learning_rate": 2.519659997687059e-06, "loss": 0.5618, "step": 334760 }, { "epoch": 28.99, "learning_rate": 2.5174916155892213e-06, "loss": 0.5979, "step": 334770 }, { "epoch": 28.99, "learning_rate": 2.515323233491384e-06, "loss": 0.6462, "step": 334780 }, { "epoch": 28.99, "learning_rate": 2.5131548513935465e-06, "loss": 0.5047, "step": 334790 }, { "epoch": 29.0, "learning_rate": 2.510986469295709e-06, "loss": 0.6349, "step": 334800 }, { "epoch": 29.0, "learning_rate": 2.5088180871978716e-06, "loss": 0.6672, "step": 334810 }, { "epoch": 29.0, "learning_rate": 2.5066497051000342e-06, "loss": 0.5476, "step": 334820 }, { "epoch": 29.0, "learning_rate": 2.5044813230021972e-06, "loss": 0.6195, "step": 334830 }, { "epoch": 29.0, "learning_rate": 2.50231294090436e-06, "loss": 0.4997, "step": 334840 }, { "epoch": 29.0, "eval_cer": 0.5328037374805152, "eval_em": 0.06601340897369778, "eval_f1": 0.06601340897369778, "eval_loss": 1.588804006576538, "eval_runtime": 257.6801, "eval_samples_per_second": 7.525, "eval_steps_per_second": 0.943, "eval_wer": 0.9339865910263022, "step": 334848 }, { "epoch": 29.0, "learning_rate": 2.5001445588065224e-06, "loss": 0.5547, "step": 334850 }, { "epoch": 29.0, "learning_rate": 2.497976176708685e-06, "loss": 0.5793, "step": 334860 }, { "epoch": 29.0, "learning_rate": 2.4958077946108475e-06, "loss": 0.5699, "step": 334870 }, { "epoch": 29.0, "learning_rate": 2.49363941251301e-06, "loss": 0.5708, "step": 334880 }, { "epoch": 29.0, "learning_rate": 2.4914710304151727e-06, "loss": 0.5458, "step": 334890 }, { "epoch": 29.0, "learning_rate": 2.489302648317335e-06, "loss": 0.4991, "step": 334900 }, { "epoch": 29.01, "learning_rate": 2.487134266219498e-06, "loss": 0.6146, "step": 334910 }, { "epoch": 29.01, "learning_rate": 2.4849658841216604e-06, "loss": 0.632, "step": 334920 }, { "epoch": 29.01, "learning_rate": 2.482797502023823e-06, "loss": 0.561, "step": 334930 }, { "epoch": 29.01, "learning_rate": 2.4806291199259856e-06, "loss": 0.5655, "step": 334940 }, { "epoch": 29.01, "learning_rate": 2.478460737828148e-06, "loss": 0.5804, "step": 334950 }, { "epoch": 29.01, "learning_rate": 2.4762923557303108e-06, "loss": 0.632, "step": 334960 }, { "epoch": 29.01, "learning_rate": 2.4741239736324733e-06, "loss": 0.541, "step": 334970 }, { "epoch": 29.01, "learning_rate": 2.4719555915346364e-06, "loss": 0.685, "step": 334980 }, { "epoch": 29.01, "learning_rate": 2.469787209436799e-06, "loss": 0.6402, "step": 334990 }, { "epoch": 29.01, "learning_rate": 2.4676188273389615e-06, "loss": 0.4698, "step": 335000 }, { "epoch": 29.01, "learning_rate": 2.465450445241124e-06, "loss": 0.5533, "step": 335010 }, { "epoch": 29.01, "learning_rate": 2.4632820631432867e-06, "loss": 0.6169, "step": 335020 }, { "epoch": 29.02, "learning_rate": 2.461113681045449e-06, "loss": 0.637, "step": 335030 }, { "epoch": 29.02, "learning_rate": 2.4589452989476114e-06, "loss": 0.5425, "step": 335040 }, { "epoch": 29.02, "learning_rate": 2.456776916849774e-06, "loss": 0.7232, "step": 335050 }, { "epoch": 29.02, "learning_rate": 2.4546085347519366e-06, "loss": 0.4684, "step": 335060 }, { "epoch": 29.02, "learning_rate": 2.4524401526540996e-06, "loss": 0.4451, "step": 335070 }, { "epoch": 29.02, "learning_rate": 2.450271770556262e-06, "loss": 0.4424, "step": 335080 }, { "epoch": 29.02, "learning_rate": 2.4481033884584247e-06, "loss": 0.4923, "step": 335090 }, { "epoch": 29.02, "learning_rate": 2.4459350063605873e-06, "loss": 0.5334, "step": 335100 }, { "epoch": 29.02, "learning_rate": 2.44376662426275e-06, "loss": 0.6385, "step": 335110 }, { "epoch": 29.02, "learning_rate": 2.4415982421649125e-06, "loss": 0.649, "step": 335120 }, { "epoch": 29.02, "learning_rate": 2.439429860067075e-06, "loss": 0.4983, "step": 335130 }, { "epoch": 29.03, "learning_rate": 2.437261477969238e-06, "loss": 0.589, "step": 335140 }, { "epoch": 29.03, "learning_rate": 2.4350930958714006e-06, "loss": 0.6944, "step": 335150 }, { "epoch": 29.03, "learning_rate": 2.432924713773563e-06, "loss": 0.5701, "step": 335160 }, { "epoch": 29.03, "learning_rate": 2.4307563316757254e-06, "loss": 0.6279, "step": 335170 }, { "epoch": 29.03, "learning_rate": 2.428587949577888e-06, "loss": 0.5363, "step": 335180 }, { "epoch": 29.03, "learning_rate": 2.4264195674800505e-06, "loss": 0.5632, "step": 335190 }, { "epoch": 29.03, "learning_rate": 2.424251185382213e-06, "loss": 0.5654, "step": 335200 }, { "epoch": 29.03, "learning_rate": 2.4220828032843757e-06, "loss": 0.5629, "step": 335210 }, { "epoch": 29.03, "learning_rate": 2.4199144211865383e-06, "loss": 0.4817, "step": 335220 }, { "epoch": 29.03, "learning_rate": 2.4177460390887013e-06, "loss": 0.5531, "step": 335230 }, { "epoch": 29.03, "learning_rate": 2.415577656990864e-06, "loss": 0.4563, "step": 335240 }, { "epoch": 29.03, "learning_rate": 2.4134092748930264e-06, "loss": 0.5978, "step": 335250 }, { "epoch": 29.04, "learning_rate": 2.411240892795189e-06, "loss": 0.5916, "step": 335260 }, { "epoch": 29.04, "learning_rate": 2.4090725106973516e-06, "loss": 0.5995, "step": 335270 }, { "epoch": 29.04, "learning_rate": 2.406904128599514e-06, "loss": 0.557, "step": 335280 }, { "epoch": 29.04, "learning_rate": 2.4047357465016763e-06, "loss": 0.5088, "step": 335290 }, { "epoch": 29.04, "learning_rate": 2.402567364403839e-06, "loss": 0.7064, "step": 335300 }, { "epoch": 29.04, "learning_rate": 2.400398982306002e-06, "loss": 0.6362, "step": 335310 }, { "epoch": 29.04, "learning_rate": 2.3982306002081645e-06, "loss": 0.5139, "step": 335320 }, { "epoch": 29.04, "learning_rate": 2.396062218110327e-06, "loss": 0.5667, "step": 335330 }, { "epoch": 29.04, "learning_rate": 2.3938938360124897e-06, "loss": 0.4506, "step": 335340 }, { "epoch": 29.04, "learning_rate": 2.3917254539146522e-06, "loss": 0.4848, "step": 335350 }, { "epoch": 29.04, "learning_rate": 2.389557071816815e-06, "loss": 0.4733, "step": 335360 }, { "epoch": 29.05, "learning_rate": 2.3873886897189774e-06, "loss": 0.6137, "step": 335370 }, { "epoch": 29.05, "learning_rate": 2.3852203076211404e-06, "loss": 0.6084, "step": 335380 }, { "epoch": 29.05, "learning_rate": 2.383051925523303e-06, "loss": 0.5857, "step": 335390 }, { "epoch": 29.05, "learning_rate": 2.3808835434254656e-06, "loss": 0.5244, "step": 335400 }, { "epoch": 29.05, "learning_rate": 2.378715161327628e-06, "loss": 0.6859, "step": 335410 }, { "epoch": 29.05, "learning_rate": 2.3765467792297903e-06, "loss": 0.4964, "step": 335420 }, { "epoch": 29.05, "learning_rate": 2.374378397131953e-06, "loss": 0.6137, "step": 335430 }, { "epoch": 29.05, "learning_rate": 2.3722100150341155e-06, "loss": 0.5246, "step": 335440 }, { "epoch": 29.05, "learning_rate": 2.370041632936278e-06, "loss": 0.581, "step": 335450 }, { "epoch": 29.05, "learning_rate": 2.3678732508384406e-06, "loss": 0.601, "step": 335460 }, { "epoch": 29.05, "learning_rate": 2.3657048687406036e-06, "loss": 0.5213, "step": 335470 }, { "epoch": 29.05, "learning_rate": 2.363536486642766e-06, "loss": 0.6478, "step": 335480 }, { "epoch": 29.06, "learning_rate": 2.3613681045449288e-06, "loss": 0.5814, "step": 335490 }, { "epoch": 29.06, "learning_rate": 2.3591997224470914e-06, "loss": 0.5577, "step": 335500 }, { "epoch": 29.06, "learning_rate": 2.357031340349254e-06, "loss": 0.6289, "step": 335510 }, { "epoch": 29.06, "learning_rate": 2.3548629582514165e-06, "loss": 0.5847, "step": 335520 }, { "epoch": 29.06, "learning_rate": 2.352694576153579e-06, "loss": 0.6053, "step": 335530 }, { "epoch": 29.06, "learning_rate": 2.3505261940557413e-06, "loss": 0.6417, "step": 335540 }, { "epoch": 29.06, "learning_rate": 2.3483578119579043e-06, "loss": 0.5401, "step": 335550 }, { "epoch": 29.06, "learning_rate": 2.346189429860067e-06, "loss": 0.4183, "step": 335560 }, { "epoch": 29.06, "learning_rate": 2.3440210477622294e-06, "loss": 0.5669, "step": 335570 }, { "epoch": 29.06, "learning_rate": 2.341852665664392e-06, "loss": 0.5818, "step": 335580 }, { "epoch": 29.06, "learning_rate": 2.3396842835665546e-06, "loss": 0.5182, "step": 335590 }, { "epoch": 29.07, "learning_rate": 2.337515901468717e-06, "loss": 0.5221, "step": 335600 }, { "epoch": 29.07, "learning_rate": 2.3353475193708797e-06, "loss": 0.6302, "step": 335610 }, { "epoch": 29.07, "learning_rate": 2.3331791372730427e-06, "loss": 0.5002, "step": 335620 }, { "epoch": 29.07, "learning_rate": 2.3310107551752053e-06, "loss": 0.5605, "step": 335630 }, { "epoch": 29.07, "learning_rate": 2.3288423730773675e-06, "loss": 0.5559, "step": 335640 }, { "epoch": 29.07, "learning_rate": 2.32667399097953e-06, "loss": 0.5832, "step": 335650 }, { "epoch": 29.07, "learning_rate": 2.3245056088816926e-06, "loss": 0.6787, "step": 335660 }, { "epoch": 29.07, "learning_rate": 2.3223372267838556e-06, "loss": 0.5577, "step": 335670 }, { "epoch": 29.07, "learning_rate": 2.3201688446860182e-06, "loss": 0.5615, "step": 335680 }, { "epoch": 29.07, "learning_rate": 2.318000462588181e-06, "loss": 0.6236, "step": 335690 }, { "epoch": 29.07, "learning_rate": 2.315832080490343e-06, "loss": 0.5935, "step": 335700 }, { "epoch": 29.07, "learning_rate": 2.313663698392506e-06, "loss": 0.5952, "step": 335710 }, { "epoch": 29.08, "learning_rate": 2.3114953162946685e-06, "loss": 0.5408, "step": 335720 }, { "epoch": 29.08, "learning_rate": 2.309326934196831e-06, "loss": 0.4827, "step": 335730 }, { "epoch": 29.08, "learning_rate": 2.3071585520989937e-06, "loss": 0.5543, "step": 335740 }, { "epoch": 29.08, "learning_rate": 2.3049901700011563e-06, "loss": 0.5715, "step": 335750 }, { "epoch": 29.08, "learning_rate": 2.302821787903319e-06, "loss": 0.5112, "step": 335760 }, { "epoch": 29.08, "learning_rate": 2.3006534058054814e-06, "loss": 0.5565, "step": 335770 }, { "epoch": 29.08, "learning_rate": 2.298485023707644e-06, "loss": 0.5329, "step": 335780 }, { "epoch": 29.08, "learning_rate": 2.2963166416098066e-06, "loss": 0.5331, "step": 335790 }, { "epoch": 29.08, "learning_rate": 2.294148259511969e-06, "loss": 0.5638, "step": 335800 }, { "epoch": 29.08, "learning_rate": 2.2919798774141318e-06, "loss": 0.5677, "step": 335810 }, { "epoch": 29.08, "learning_rate": 2.2898114953162948e-06, "loss": 0.6439, "step": 335820 }, { "epoch": 29.09, "learning_rate": 2.287643113218457e-06, "loss": 0.5865, "step": 335830 }, { "epoch": 29.09, "learning_rate": 2.2854747311206195e-06, "loss": 0.5949, "step": 335840 }, { "epoch": 29.09, "learning_rate": 2.283306349022782e-06, "loss": 0.5693, "step": 335850 }, { "epoch": 29.09, "learning_rate": 2.2811379669249447e-06, "loss": 0.5264, "step": 335860 }, { "epoch": 29.09, "learning_rate": 2.2789695848271077e-06, "loss": 0.6194, "step": 335870 }, { "epoch": 29.09, "learning_rate": 2.2768012027292702e-06, "loss": 0.6736, "step": 335880 }, { "epoch": 29.09, "learning_rate": 2.274632820631433e-06, "loss": 0.5104, "step": 335890 }, { "epoch": 29.09, "learning_rate": 2.272464438533595e-06, "loss": 0.5274, "step": 335900 }, { "epoch": 29.09, "learning_rate": 2.270296056435758e-06, "loss": 0.6113, "step": 335910 }, { "epoch": 29.09, "learning_rate": 2.2681276743379206e-06, "loss": 0.6694, "step": 335920 }, { "epoch": 29.09, "learning_rate": 2.265959292240083e-06, "loss": 0.4225, "step": 335930 }, { "epoch": 29.09, "learning_rate": 2.2637909101422457e-06, "loss": 0.5182, "step": 335940 }, { "epoch": 29.1, "learning_rate": 2.2616225280444083e-06, "loss": 0.3292, "step": 335950 }, { "epoch": 29.1, "learning_rate": 2.259454145946571e-06, "loss": 0.5337, "step": 335960 }, { "epoch": 29.1, "learning_rate": 2.2572857638487335e-06, "loss": 0.6119, "step": 335970 }, { "epoch": 29.1, "learning_rate": 2.255117381750896e-06, "loss": 0.6469, "step": 335980 }, { "epoch": 29.1, "learning_rate": 2.2529489996530586e-06, "loss": 0.5023, "step": 335990 }, { "epoch": 29.1, "learning_rate": 2.250780617555221e-06, "loss": 0.5331, "step": 336000 }, { "epoch": 29.1, "learning_rate": 2.2486122354573838e-06, "loss": 0.5297, "step": 336010 }, { "epoch": 29.1, "learning_rate": 2.246443853359547e-06, "loss": 0.6092, "step": 336020 }, { "epoch": 29.1, "learning_rate": 2.244275471261709e-06, "loss": 0.5487, "step": 336030 }, { "epoch": 29.1, "learning_rate": 2.2421070891638715e-06, "loss": 0.5573, "step": 336040 }, { "epoch": 29.1, "learning_rate": 2.239938707066034e-06, "loss": 0.602, "step": 336050 }, { "epoch": 29.1, "learning_rate": 2.2377703249681967e-06, "loss": 0.5677, "step": 336060 }, { "epoch": 29.11, "learning_rate": 2.2356019428703597e-06, "loss": 0.7438, "step": 336070 }, { "epoch": 29.11, "learning_rate": 2.2334335607725223e-06, "loss": 0.518, "step": 336080 }, { "epoch": 29.11, "learning_rate": 2.2312651786746844e-06, "loss": 0.4663, "step": 336090 }, { "epoch": 29.11, "learning_rate": 2.229096796576847e-06, "loss": 0.4713, "step": 336100 }, { "epoch": 29.11, "learning_rate": 2.22692841447901e-06, "loss": 0.4861, "step": 336110 }, { "epoch": 29.11, "learning_rate": 2.2247600323811726e-06, "loss": 0.5343, "step": 336120 }, { "epoch": 29.11, "learning_rate": 2.222591650283335e-06, "loss": 0.5168, "step": 336130 }, { "epoch": 29.11, "learning_rate": 2.2204232681854977e-06, "loss": 0.5444, "step": 336140 }, { "epoch": 29.11, "learning_rate": 2.2182548860876603e-06, "loss": 0.6437, "step": 336150 }, { "epoch": 29.11, "learning_rate": 2.216086503989823e-06, "loss": 0.4546, "step": 336160 }, { "epoch": 29.11, "learning_rate": 2.2139181218919855e-06, "loss": 0.5439, "step": 336170 }, { "epoch": 29.12, "learning_rate": 2.211749739794148e-06, "loss": 0.687, "step": 336180 }, { "epoch": 29.12, "learning_rate": 2.2095813576963106e-06, "loss": 0.5781, "step": 336190 }, { "epoch": 29.12, "learning_rate": 2.2074129755984732e-06, "loss": 0.5723, "step": 336200 }, { "epoch": 29.12, "learning_rate": 2.205244593500636e-06, "loss": 0.4338, "step": 336210 }, { "epoch": 29.12, "learning_rate": 2.2030762114027984e-06, "loss": 0.6504, "step": 336220 }, { "epoch": 29.12, "learning_rate": 2.200907829304961e-06, "loss": 0.5895, "step": 336230 }, { "epoch": 29.12, "learning_rate": 2.1987394472071235e-06, "loss": 0.6092, "step": 336240 }, { "epoch": 29.12, "learning_rate": 2.196571065109286e-06, "loss": 0.4972, "step": 336250 }, { "epoch": 29.12, "learning_rate": 2.194402683011449e-06, "loss": 0.6395, "step": 336260 }, { "epoch": 29.12, "learning_rate": 2.1922343009136117e-06, "loss": 0.5238, "step": 336270 }, { "epoch": 29.12, "learning_rate": 2.190065918815774e-06, "loss": 0.5821, "step": 336280 }, { "epoch": 29.12, "learning_rate": 2.1878975367179364e-06, "loss": 0.6704, "step": 336290 }, { "epoch": 29.13, "learning_rate": 2.185729154620099e-06, "loss": 0.53, "step": 336300 }, { "epoch": 29.13, "learning_rate": 2.183560772522262e-06, "loss": 0.5905, "step": 336310 }, { "epoch": 29.13, "learning_rate": 2.1813923904244246e-06, "loss": 0.6009, "step": 336320 }, { "epoch": 29.13, "learning_rate": 2.179224008326587e-06, "loss": 0.5783, "step": 336330 }, { "epoch": 29.13, "learning_rate": 2.1770556262287498e-06, "loss": 0.5922, "step": 336340 }, { "epoch": 29.13, "learning_rate": 2.1748872441309124e-06, "loss": 0.523, "step": 336350 }, { "epoch": 29.13, "learning_rate": 2.172718862033075e-06, "loss": 0.555, "step": 336360 }, { "epoch": 29.13, "learning_rate": 2.1705504799352375e-06, "loss": 0.543, "step": 336370 }, { "epoch": 29.13, "learning_rate": 2.1683820978374e-06, "loss": 0.5776, "step": 336380 }, { "epoch": 29.13, "learning_rate": 2.1662137157395627e-06, "loss": 0.6688, "step": 336390 }, { "epoch": 29.13, "learning_rate": 2.1640453336417253e-06, "loss": 0.5153, "step": 336400 }, { "epoch": 29.14, "learning_rate": 2.161876951543888e-06, "loss": 0.7087, "step": 336410 }, { "epoch": 29.14, "learning_rate": 2.1597085694460504e-06, "loss": 0.5338, "step": 336420 }, { "epoch": 29.14, "learning_rate": 2.157540187348213e-06, "loss": 0.5144, "step": 336430 }, { "epoch": 29.14, "learning_rate": 2.1553718052503756e-06, "loss": 0.5582, "step": 336440 }, { "epoch": 29.14, "learning_rate": 2.153203423152538e-06, "loss": 0.5502, "step": 336450 }, { "epoch": 29.14, "learning_rate": 2.151035041054701e-06, "loss": 0.6916, "step": 336460 }, { "epoch": 29.14, "learning_rate": 2.1488666589568633e-06, "loss": 0.6819, "step": 336470 }, { "epoch": 29.14, "learning_rate": 2.146698276859026e-06, "loss": 0.4817, "step": 336480 }, { "epoch": 29.14, "learning_rate": 2.1445298947611885e-06, "loss": 0.5471, "step": 336490 }, { "epoch": 29.14, "learning_rate": 2.142361512663351e-06, "loss": 0.5462, "step": 336500 }, { "epoch": 29.14, "learning_rate": 2.140193130565514e-06, "loss": 0.6736, "step": 336510 }, { "epoch": 29.14, "learning_rate": 2.1380247484676766e-06, "loss": 0.6155, "step": 336520 }, { "epoch": 29.15, "learning_rate": 2.1358563663698392e-06, "loss": 0.585, "step": 336530 }, { "epoch": 29.15, "learning_rate": 2.1336879842720014e-06, "loss": 0.5343, "step": 336540 }, { "epoch": 29.15, "learning_rate": 2.1315196021741644e-06, "loss": 0.4645, "step": 336550 }, { "epoch": 29.15, "learning_rate": 2.129351220076327e-06, "loss": 0.746, "step": 336560 }, { "epoch": 29.15, "learning_rate": 2.1271828379784895e-06, "loss": 0.6363, "step": 336570 }, { "epoch": 29.15, "learning_rate": 2.125014455880652e-06, "loss": 0.5438, "step": 336580 }, { "epoch": 29.15, "learning_rate": 2.1228460737828147e-06, "loss": 0.5115, "step": 336590 }, { "epoch": 29.15, "learning_rate": 2.1206776916849773e-06, "loss": 0.6056, "step": 336600 }, { "epoch": 29.15, "learning_rate": 2.11850930958714e-06, "loss": 0.4837, "step": 336610 }, { "epoch": 29.15, "learning_rate": 2.1163409274893024e-06, "loss": 0.5943, "step": 336620 }, { "epoch": 29.15, "learning_rate": 2.114172545391465e-06, "loss": 0.648, "step": 336630 }, { "epoch": 29.16, "learning_rate": 2.1120041632936276e-06, "loss": 0.5806, "step": 336640 }, { "epoch": 29.16, "learning_rate": 2.10983578119579e-06, "loss": 0.5477, "step": 336650 }, { "epoch": 29.16, "learning_rate": 2.107667399097953e-06, "loss": 0.5782, "step": 336660 }, { "epoch": 29.16, "learning_rate": 2.1054990170001153e-06, "loss": 0.5486, "step": 336670 }, { "epoch": 29.16, "learning_rate": 2.103330634902278e-06, "loss": 0.7172, "step": 336680 }, { "epoch": 29.16, "learning_rate": 2.1011622528044405e-06, "loss": 0.5026, "step": 336690 }, { "epoch": 29.16, "learning_rate": 2.098993870706603e-06, "loss": 0.5291, "step": 336700 }, { "epoch": 29.16, "learning_rate": 2.096825488608766e-06, "loss": 0.502, "step": 336710 }, { "epoch": 29.16, "learning_rate": 2.0946571065109287e-06, "loss": 0.55, "step": 336720 }, { "epoch": 29.16, "learning_rate": 2.092488724413091e-06, "loss": 0.6333, "step": 336730 }, { "epoch": 29.16, "learning_rate": 2.0903203423152534e-06, "loss": 0.4332, "step": 336740 }, { "epoch": 29.16, "learning_rate": 2.0881519602174164e-06, "loss": 0.5713, "step": 336750 }, { "epoch": 29.17, "learning_rate": 2.085983578119579e-06, "loss": 0.5733, "step": 336760 }, { "epoch": 29.17, "learning_rate": 2.0838151960217416e-06, "loss": 0.4828, "step": 336770 }, { "epoch": 29.17, "learning_rate": 2.081646813923904e-06, "loss": 0.4439, "step": 336780 }, { "epoch": 29.17, "learning_rate": 2.0794784318260667e-06, "loss": 0.749, "step": 336790 }, { "epoch": 29.17, "learning_rate": 2.0773100497282293e-06, "loss": 0.5395, "step": 336800 }, { "epoch": 29.17, "learning_rate": 2.075141667630392e-06, "loss": 0.6043, "step": 336810 }, { "epoch": 29.17, "learning_rate": 2.0729732855325545e-06, "loss": 0.5345, "step": 336820 }, { "epoch": 29.17, "learning_rate": 2.070804903434717e-06, "loss": 0.559, "step": 336830 }, { "epoch": 29.17, "learning_rate": 2.0686365213368796e-06, "loss": 0.6436, "step": 336840 }, { "epoch": 29.17, "learning_rate": 2.066468139239042e-06, "loss": 0.7137, "step": 336850 }, { "epoch": 29.17, "learning_rate": 2.0642997571412048e-06, "loss": 0.5141, "step": 336860 }, { "epoch": 29.18, "learning_rate": 2.0621313750433674e-06, "loss": 0.4577, "step": 336870 }, { "epoch": 29.18, "learning_rate": 2.05996299294553e-06, "loss": 0.3935, "step": 336880 }, { "epoch": 29.18, "learning_rate": 2.0577946108476925e-06, "loss": 0.6873, "step": 336890 }, { "epoch": 29.18, "learning_rate": 2.0556262287498555e-06, "loss": 0.5573, "step": 336900 }, { "epoch": 29.18, "learning_rate": 2.053457846652018e-06, "loss": 0.6782, "step": 336910 }, { "epoch": 29.18, "learning_rate": 2.0512894645541803e-06, "loss": 0.6802, "step": 336920 }, { "epoch": 29.18, "learning_rate": 2.049121082456343e-06, "loss": 0.634, "step": 336930 }, { "epoch": 29.18, "learning_rate": 2.0469527003585054e-06, "loss": 0.5437, "step": 336940 }, { "epoch": 29.18, "learning_rate": 2.0447843182606684e-06, "loss": 0.5057, "step": 336950 }, { "epoch": 29.18, "learning_rate": 2.042615936162831e-06, "loss": 0.534, "step": 336960 }, { "epoch": 29.18, "learning_rate": 2.0404475540649936e-06, "loss": 0.6123, "step": 336970 }, { "epoch": 29.18, "learning_rate": 2.038279171967156e-06, "loss": 0.5648, "step": 336980 }, { "epoch": 29.19, "learning_rate": 2.0361107898693187e-06, "loss": 0.6646, "step": 336990 }, { "epoch": 29.19, "learning_rate": 2.0339424077714813e-06, "loss": 0.5872, "step": 337000 }, { "epoch": 29.19, "learning_rate": 2.031774025673644e-06, "loss": 0.4019, "step": 337010 }, { "epoch": 29.19, "learning_rate": 2.0296056435758065e-06, "loss": 0.4577, "step": 337020 }, { "epoch": 29.19, "learning_rate": 2.027437261477969e-06, "loss": 0.5626, "step": 337030 }, { "epoch": 29.19, "learning_rate": 2.0252688793801316e-06, "loss": 0.6882, "step": 337040 }, { "epoch": 29.19, "learning_rate": 2.0231004972822942e-06, "loss": 0.4466, "step": 337050 }, { "epoch": 29.19, "learning_rate": 2.020932115184457e-06, "loss": 0.6114, "step": 337060 }, { "epoch": 29.19, "learning_rate": 2.0187637330866194e-06, "loss": 0.5953, "step": 337070 }, { "epoch": 29.19, "learning_rate": 2.016595350988782e-06, "loss": 0.6012, "step": 337080 }, { "epoch": 29.19, "learning_rate": 2.0144269688909445e-06, "loss": 0.5899, "step": 337090 }, { "epoch": 29.19, "learning_rate": 2.0122585867931075e-06, "loss": 0.4782, "step": 337100 }, { "epoch": 29.2, "learning_rate": 2.01009020469527e-06, "loss": 0.5595, "step": 337110 }, { "epoch": 29.2, "learning_rate": 2.0079218225974323e-06, "loss": 0.5098, "step": 337120 }, { "epoch": 29.2, "learning_rate": 2.005753440499595e-06, "loss": 0.651, "step": 337130 }, { "epoch": 29.2, "learning_rate": 2.0035850584017574e-06, "loss": 0.6257, "step": 337140 }, { "epoch": 29.2, "learning_rate": 2.0014166763039204e-06, "loss": 0.5459, "step": 337150 }, { "epoch": 29.2, "learning_rate": 1.999248294206083e-06, "loss": 0.4648, "step": 337160 }, { "epoch": 29.2, "learning_rate": 1.9970799121082456e-06, "loss": 0.5523, "step": 337170 }, { "epoch": 29.2, "learning_rate": 1.9949115300104078e-06, "loss": 0.5509, "step": 337180 }, { "epoch": 29.2, "learning_rate": 1.9927431479125708e-06, "loss": 0.5649, "step": 337190 }, { "epoch": 29.2, "learning_rate": 1.9905747658147333e-06, "loss": 0.4125, "step": 337200 }, { "epoch": 29.2, "learning_rate": 1.988406383716896e-06, "loss": 0.5729, "step": 337210 }, { "epoch": 29.21, "learning_rate": 1.9862380016190585e-06, "loss": 0.5891, "step": 337220 }, { "epoch": 29.21, "learning_rate": 1.984069619521221e-06, "loss": 0.4871, "step": 337230 }, { "epoch": 29.21, "learning_rate": 1.9819012374233837e-06, "loss": 0.4408, "step": 337240 }, { "epoch": 29.21, "learning_rate": 1.9797328553255462e-06, "loss": 0.4725, "step": 337250 }, { "epoch": 29.21, "learning_rate": 1.977564473227709e-06, "loss": 0.5217, "step": 337260 }, { "epoch": 29.21, "learning_rate": 1.9753960911298714e-06, "loss": 0.6211, "step": 337270 }, { "epoch": 29.21, "learning_rate": 1.973227709032034e-06, "loss": 0.5826, "step": 337280 }, { "epoch": 29.21, "learning_rate": 1.9710593269341966e-06, "loss": 0.6299, "step": 337290 }, { "epoch": 29.21, "learning_rate": 1.9688909448363596e-06, "loss": 0.5682, "step": 337300 }, { "epoch": 29.21, "learning_rate": 1.9667225627385217e-06, "loss": 0.5928, "step": 337310 }, { "epoch": 29.21, "learning_rate": 1.9645541806406843e-06, "loss": 0.6799, "step": 337320 }, { "epoch": 29.21, "learning_rate": 1.962385798542847e-06, "loss": 0.508, "step": 337330 }, { "epoch": 29.22, "learning_rate": 1.9602174164450095e-06, "loss": 0.505, "step": 337340 }, { "epoch": 29.22, "learning_rate": 1.9580490343471725e-06, "loss": 0.5796, "step": 337350 }, { "epoch": 29.22, "learning_rate": 1.955880652249335e-06, "loss": 0.5552, "step": 337360 }, { "epoch": 29.22, "learning_rate": 1.953712270151497e-06, "loss": 0.5119, "step": 337370 }, { "epoch": 29.22, "learning_rate": 1.9515438880536598e-06, "loss": 0.5318, "step": 337380 }, { "epoch": 29.22, "learning_rate": 1.949375505955823e-06, "loss": 0.4813, "step": 337390 }, { "epoch": 29.22, "learning_rate": 1.9472071238579854e-06, "loss": 0.5097, "step": 337400 }, { "epoch": 29.22, "learning_rate": 1.945038741760148e-06, "loss": 0.5553, "step": 337410 }, { "epoch": 29.22, "learning_rate": 1.9428703596623105e-06, "loss": 0.6227, "step": 337420 }, { "epoch": 29.22, "learning_rate": 1.940701977564473e-06, "loss": 0.5034, "step": 337430 }, { "epoch": 29.22, "learning_rate": 1.9385335954666357e-06, "loss": 0.5156, "step": 337440 }, { "epoch": 29.23, "learning_rate": 1.9363652133687983e-06, "loss": 0.6302, "step": 337450 }, { "epoch": 29.23, "learning_rate": 1.934196831270961e-06, "loss": 0.4708, "step": 337460 }, { "epoch": 29.23, "learning_rate": 1.9320284491731234e-06, "loss": 0.5658, "step": 337470 }, { "epoch": 29.23, "learning_rate": 1.929860067075286e-06, "loss": 0.5772, "step": 337480 }, { "epoch": 29.23, "learning_rate": 1.9276916849774486e-06, "loss": 0.544, "step": 337490 }, { "epoch": 29.23, "learning_rate": 1.925523302879611e-06, "loss": 0.5396, "step": 337500 }, { "epoch": 29.23, "learning_rate": 1.9233549207817738e-06, "loss": 0.6171, "step": 337510 }, { "epoch": 29.23, "learning_rate": 1.9211865386839363e-06, "loss": 0.6129, "step": 337520 }, { "epoch": 29.23, "learning_rate": 1.919018156586099e-06, "loss": 0.6271, "step": 337530 }, { "epoch": 29.23, "learning_rate": 1.916849774488262e-06, "loss": 0.546, "step": 337540 }, { "epoch": 29.23, "learning_rate": 1.9146813923904245e-06, "loss": 0.6168, "step": 337550 }, { "epoch": 29.23, "learning_rate": 1.9125130102925867e-06, "loss": 0.5515, "step": 337560 }, { "epoch": 29.24, "learning_rate": 1.9103446281947492e-06, "loss": 0.6, "step": 337570 }, { "epoch": 29.24, "learning_rate": 1.908176246096912e-06, "loss": 0.5577, "step": 337580 }, { "epoch": 29.24, "learning_rate": 1.9060078639990746e-06, "loss": 0.4979, "step": 337590 }, { "epoch": 29.24, "learning_rate": 1.9038394819012374e-06, "loss": 0.6838, "step": 337600 }, { "epoch": 29.24, "learning_rate": 1.9016710998034e-06, "loss": 0.5909, "step": 337610 }, { "epoch": 29.24, "learning_rate": 1.8995027177055626e-06, "loss": 0.5209, "step": 337620 }, { "epoch": 29.24, "learning_rate": 1.897334335607725e-06, "loss": 0.5402, "step": 337630 }, { "epoch": 29.24, "learning_rate": 1.8951659535098875e-06, "loss": 0.5792, "step": 337640 }, { "epoch": 29.24, "learning_rate": 1.8929975714120503e-06, "loss": 0.5257, "step": 337650 }, { "epoch": 29.24, "learning_rate": 1.8908291893142129e-06, "loss": 0.5751, "step": 337660 }, { "epoch": 29.24, "learning_rate": 1.8886608072163755e-06, "loss": 0.5657, "step": 337670 }, { "epoch": 29.25, "learning_rate": 1.8864924251185382e-06, "loss": 0.5479, "step": 337680 }, { "epoch": 29.25, "learning_rate": 1.8843240430207006e-06, "loss": 0.5161, "step": 337690 }, { "epoch": 29.25, "learning_rate": 1.8821556609228632e-06, "loss": 0.4843, "step": 337700 }, { "epoch": 29.25, "learning_rate": 1.8799872788250258e-06, "loss": 0.522, "step": 337710 }, { "epoch": 29.25, "learning_rate": 1.8778188967271884e-06, "loss": 0.5362, "step": 337720 }, { "epoch": 29.25, "learning_rate": 1.8756505146293511e-06, "loss": 0.5126, "step": 337730 }, { "epoch": 29.25, "learning_rate": 1.8734821325315137e-06, "loss": 0.4707, "step": 337740 }, { "epoch": 29.25, "learning_rate": 1.8713137504336763e-06, "loss": 0.4448, "step": 337750 }, { "epoch": 29.25, "learning_rate": 1.8691453683358387e-06, "loss": 0.5488, "step": 337760 }, { "epoch": 29.25, "learning_rate": 1.8669769862380015e-06, "loss": 0.6278, "step": 337770 }, { "epoch": 29.25, "learning_rate": 1.864808604140164e-06, "loss": 0.5562, "step": 337780 }, { "epoch": 29.25, "learning_rate": 1.8626402220423266e-06, "loss": 0.6538, "step": 337790 }, { "epoch": 29.26, "learning_rate": 1.8604718399444894e-06, "loss": 0.5316, "step": 337800 }, { "epoch": 29.26, "learning_rate": 1.858303457846652e-06, "loss": 0.5748, "step": 337810 }, { "epoch": 29.26, "learning_rate": 1.8561350757488144e-06, "loss": 0.5891, "step": 337820 }, { "epoch": 29.26, "learning_rate": 1.853966693650977e-06, "loss": 0.4759, "step": 337830 }, { "epoch": 29.26, "learning_rate": 1.8517983115531395e-06, "loss": 0.5148, "step": 337840 }, { "epoch": 29.26, "learning_rate": 1.8496299294553023e-06, "loss": 0.6097, "step": 337850 }, { "epoch": 29.26, "learning_rate": 1.847461547357465e-06, "loss": 0.5426, "step": 337860 }, { "epoch": 29.26, "learning_rate": 1.8452931652596275e-06, "loss": 0.4695, "step": 337870 }, { "epoch": 29.26, "learning_rate": 1.8431247831617903e-06, "loss": 0.5703, "step": 337880 }, { "epoch": 29.26, "learning_rate": 1.8409564010639526e-06, "loss": 0.4525, "step": 337890 }, { "epoch": 29.26, "learning_rate": 1.8387880189661152e-06, "loss": 0.6788, "step": 337900 }, { "epoch": 29.27, "learning_rate": 1.8366196368682778e-06, "loss": 0.5667, "step": 337910 }, { "epoch": 29.27, "learning_rate": 1.8344512547704406e-06, "loss": 0.5192, "step": 337920 }, { "epoch": 29.27, "learning_rate": 1.8322828726726032e-06, "loss": 0.4405, "step": 337930 }, { "epoch": 29.27, "learning_rate": 1.8301144905747657e-06, "loss": 0.5189, "step": 337940 }, { "epoch": 29.27, "learning_rate": 1.8279461084769281e-06, "loss": 0.5688, "step": 337950 }, { "epoch": 29.27, "learning_rate": 1.8257777263790907e-06, "loss": 0.5431, "step": 337960 }, { "epoch": 29.27, "learning_rate": 1.8236093442812535e-06, "loss": 0.5912, "step": 337970 }, { "epoch": 29.27, "learning_rate": 1.821440962183416e-06, "loss": 0.5342, "step": 337980 }, { "epoch": 29.27, "learning_rate": 1.8192725800855786e-06, "loss": 0.4851, "step": 337990 }, { "epoch": 29.27, "learning_rate": 1.8171041979877414e-06, "loss": 0.5824, "step": 338000 }, { "epoch": 29.27, "learning_rate": 1.8149358158899038e-06, "loss": 0.5767, "step": 338010 }, { "epoch": 29.27, "learning_rate": 1.8127674337920664e-06, "loss": 0.4284, "step": 338020 }, { "epoch": 29.28, "learning_rate": 1.810599051694229e-06, "loss": 0.5454, "step": 338030 }, { "epoch": 29.28, "learning_rate": 1.8084306695963916e-06, "loss": 0.745, "step": 338040 }, { "epoch": 29.28, "learning_rate": 1.8062622874985543e-06, "loss": 0.5145, "step": 338050 }, { "epoch": 29.28, "learning_rate": 1.804093905400717e-06, "loss": 0.7355, "step": 338060 }, { "epoch": 29.28, "learning_rate": 1.8019255233028795e-06, "loss": 0.4938, "step": 338070 }, { "epoch": 29.28, "learning_rate": 1.7997571412050419e-06, "loss": 0.6446, "step": 338080 }, { "epoch": 29.28, "learning_rate": 1.7975887591072047e-06, "loss": 0.6325, "step": 338090 }, { "epoch": 29.28, "learning_rate": 1.7954203770093672e-06, "loss": 0.5088, "step": 338100 }, { "epoch": 29.28, "learning_rate": 1.7932519949115298e-06, "loss": 0.5126, "step": 338110 }, { "epoch": 29.28, "learning_rate": 1.7910836128136926e-06, "loss": 0.3952, "step": 338120 }, { "epoch": 29.28, "learning_rate": 1.7889152307158552e-06, "loss": 0.7477, "step": 338130 }, { "epoch": 29.29, "learning_rate": 1.7867468486180176e-06, "loss": 0.4836, "step": 338140 }, { "epoch": 29.29, "learning_rate": 1.7845784665201801e-06, "loss": 0.45, "step": 338150 }, { "epoch": 29.29, "learning_rate": 1.7824100844223427e-06, "loss": 0.5741, "step": 338160 }, { "epoch": 29.29, "learning_rate": 1.7802417023245055e-06, "loss": 0.6945, "step": 338170 }, { "epoch": 29.29, "learning_rate": 1.778073320226668e-06, "loss": 0.5788, "step": 338180 }, { "epoch": 29.29, "learning_rate": 1.7759049381288307e-06, "loss": 0.6092, "step": 338190 }, { "epoch": 29.29, "learning_rate": 1.7737365560309935e-06, "loss": 0.5611, "step": 338200 }, { "epoch": 29.29, "learning_rate": 1.7715681739331558e-06, "loss": 0.5592, "step": 338210 }, { "epoch": 29.29, "learning_rate": 1.7693997918353184e-06, "loss": 0.561, "step": 338220 }, { "epoch": 29.29, "learning_rate": 1.767231409737481e-06, "loss": 0.5458, "step": 338230 }, { "epoch": 29.29, "learning_rate": 1.7650630276396438e-06, "loss": 0.6892, "step": 338240 }, { "epoch": 29.29, "learning_rate": 1.7628946455418064e-06, "loss": 0.4809, "step": 338250 }, { "epoch": 29.3, "learning_rate": 1.760726263443969e-06, "loss": 0.5137, "step": 338260 }, { "epoch": 29.3, "learning_rate": 1.7585578813461313e-06, "loss": 0.5507, "step": 338270 }, { "epoch": 29.3, "learning_rate": 1.7563894992482939e-06, "loss": 0.5797, "step": 338280 }, { "epoch": 29.3, "learning_rate": 1.7542211171504567e-06, "loss": 0.5859, "step": 338290 }, { "epoch": 29.3, "learning_rate": 1.7520527350526193e-06, "loss": 0.6151, "step": 338300 }, { "epoch": 29.3, "learning_rate": 1.7498843529547818e-06, "loss": 0.5341, "step": 338310 }, { "epoch": 29.3, "learning_rate": 1.7477159708569446e-06, "loss": 0.5465, "step": 338320 }, { "epoch": 29.3, "learning_rate": 1.745547588759107e-06, "loss": 0.6444, "step": 338330 }, { "epoch": 29.3, "learning_rate": 1.7433792066612696e-06, "loss": 0.6036, "step": 338340 }, { "epoch": 29.3, "learning_rate": 1.7412108245634322e-06, "loss": 0.5546, "step": 338350 }, { "epoch": 29.3, "learning_rate": 1.7390424424655947e-06, "loss": 0.5152, "step": 338360 }, { "epoch": 29.3, "learning_rate": 1.7368740603677575e-06, "loss": 0.4701, "step": 338370 }, { "epoch": 29.31, "learning_rate": 1.7347056782699201e-06, "loss": 0.6283, "step": 338380 }, { "epoch": 29.31, "learning_rate": 1.7325372961720827e-06, "loss": 0.6781, "step": 338390 }, { "epoch": 29.31, "learning_rate": 1.730368914074245e-06, "loss": 0.5197, "step": 338400 }, { "epoch": 29.31, "learning_rate": 1.7282005319764079e-06, "loss": 0.5633, "step": 338410 }, { "epoch": 29.31, "learning_rate": 1.7260321498785704e-06, "loss": 0.5453, "step": 338420 }, { "epoch": 29.31, "learning_rate": 1.723863767780733e-06, "loss": 0.644, "step": 338430 }, { "epoch": 29.31, "learning_rate": 1.7216953856828958e-06, "loss": 0.5574, "step": 338440 }, { "epoch": 29.31, "learning_rate": 1.7195270035850584e-06, "loss": 0.5162, "step": 338450 }, { "epoch": 29.31, "learning_rate": 1.7173586214872208e-06, "loss": 0.5443, "step": 338460 }, { "epoch": 29.31, "learning_rate": 1.7151902393893833e-06, "loss": 0.6301, "step": 338470 }, { "epoch": 29.31, "learning_rate": 1.713021857291546e-06, "loss": 0.5385, "step": 338480 }, { "epoch": 29.32, "learning_rate": 1.7108534751937087e-06, "loss": 0.5396, "step": 338490 }, { "epoch": 29.32, "learning_rate": 1.7086850930958713e-06, "loss": 0.4675, "step": 338500 }, { "epoch": 29.32, "learning_rate": 1.7065167109980339e-06, "loss": 0.6858, "step": 338510 }, { "epoch": 29.32, "learning_rate": 1.7043483289001967e-06, "loss": 0.559, "step": 338520 }, { "epoch": 29.32, "learning_rate": 1.702179946802359e-06, "loss": 0.5355, "step": 338530 }, { "epoch": 29.32, "learning_rate": 1.7000115647045216e-06, "loss": 0.5384, "step": 338540 }, { "epoch": 29.32, "learning_rate": 1.6978431826066842e-06, "loss": 0.6363, "step": 338550 }, { "epoch": 29.32, "learning_rate": 1.6956748005088468e-06, "loss": 0.5247, "step": 338560 }, { "epoch": 29.32, "learning_rate": 1.6935064184110096e-06, "loss": 0.4812, "step": 338570 }, { "epoch": 29.32, "learning_rate": 1.6913380363131721e-06, "loss": 0.566, "step": 338580 }, { "epoch": 29.32, "learning_rate": 1.6891696542153345e-06, "loss": 0.5689, "step": 338590 }, { "epoch": 29.32, "learning_rate": 1.687001272117497e-06, "loss": 0.4576, "step": 338600 }, { "epoch": 29.33, "learning_rate": 1.6848328900196599e-06, "loss": 0.5038, "step": 338610 }, { "epoch": 29.33, "learning_rate": 1.6826645079218225e-06, "loss": 0.687, "step": 338620 }, { "epoch": 29.33, "learning_rate": 1.680496125823985e-06, "loss": 0.4255, "step": 338630 }, { "epoch": 29.33, "learning_rate": 1.6783277437261478e-06, "loss": 0.6167, "step": 338640 }, { "epoch": 29.33, "learning_rate": 1.6761593616283104e-06, "loss": 0.511, "step": 338650 }, { "epoch": 29.33, "learning_rate": 1.6739909795304728e-06, "loss": 0.6006, "step": 338660 }, { "epoch": 29.33, "learning_rate": 1.6718225974326354e-06, "loss": 0.4738, "step": 338670 }, { "epoch": 29.33, "learning_rate": 1.669654215334798e-06, "loss": 0.5388, "step": 338680 }, { "epoch": 29.33, "learning_rate": 1.6674858332369607e-06, "loss": 0.6154, "step": 338690 }, { "epoch": 29.33, "learning_rate": 1.6653174511391233e-06, "loss": 0.552, "step": 338700 }, { "epoch": 29.33, "learning_rate": 1.6631490690412859e-06, "loss": 0.5458, "step": 338710 }, { "epoch": 29.34, "learning_rate": 1.6609806869434483e-06, "loss": 0.6188, "step": 338720 }, { "epoch": 29.34, "learning_rate": 1.658812304845611e-06, "loss": 0.6465, "step": 338730 }, { "epoch": 29.34, "learning_rate": 1.6566439227477736e-06, "loss": 0.4747, "step": 338740 }, { "epoch": 29.34, "learning_rate": 1.6544755406499362e-06, "loss": 0.573, "step": 338750 }, { "epoch": 29.34, "learning_rate": 1.652307158552099e-06, "loss": 0.5854, "step": 338760 }, { "epoch": 29.34, "learning_rate": 1.6501387764542616e-06, "loss": 0.6045, "step": 338770 }, { "epoch": 29.34, "learning_rate": 1.647970394356424e-06, "loss": 0.5486, "step": 338780 }, { "epoch": 29.34, "learning_rate": 1.6458020122585865e-06, "loss": 0.4381, "step": 338790 }, { "epoch": 29.34, "learning_rate": 1.6436336301607491e-06, "loss": 0.5137, "step": 338800 }, { "epoch": 29.34, "learning_rate": 1.641465248062912e-06, "loss": 0.5843, "step": 338810 }, { "epoch": 29.34, "learning_rate": 1.6392968659650745e-06, "loss": 0.5395, "step": 338820 }, { "epoch": 29.34, "learning_rate": 1.637128483867237e-06, "loss": 0.5666, "step": 338830 }, { "epoch": 29.35, "learning_rate": 1.6349601017693999e-06, "loss": 0.4848, "step": 338840 }, { "epoch": 29.35, "learning_rate": 1.6327917196715622e-06, "loss": 0.5942, "step": 338850 }, { "epoch": 29.35, "learning_rate": 1.6306233375737248e-06, "loss": 0.5409, "step": 338860 }, { "epoch": 29.35, "learning_rate": 1.6284549554758874e-06, "loss": 0.551, "step": 338870 }, { "epoch": 29.35, "learning_rate": 1.62628657337805e-06, "loss": 0.5599, "step": 338880 }, { "epoch": 29.35, "learning_rate": 1.6241181912802128e-06, "loss": 0.56, "step": 338890 }, { "epoch": 29.35, "learning_rate": 1.6219498091823753e-06, "loss": 0.5702, "step": 338900 }, { "epoch": 29.35, "learning_rate": 1.6197814270845377e-06, "loss": 0.5358, "step": 338910 }, { "epoch": 29.35, "learning_rate": 1.6176130449867003e-06, "loss": 0.5713, "step": 338920 }, { "epoch": 29.35, "learning_rate": 1.615444662888863e-06, "loss": 0.5488, "step": 338930 }, { "epoch": 29.35, "learning_rate": 1.6132762807910257e-06, "loss": 0.657, "step": 338940 }, { "epoch": 29.36, "learning_rate": 1.6111078986931882e-06, "loss": 0.6007, "step": 338950 }, { "epoch": 29.36, "learning_rate": 1.608939516595351e-06, "loss": 0.5085, "step": 338960 }, { "epoch": 29.36, "learning_rate": 1.6067711344975136e-06, "loss": 0.4954, "step": 338970 }, { "epoch": 29.36, "learning_rate": 1.604602752399676e-06, "loss": 0.5139, "step": 338980 }, { "epoch": 29.36, "learning_rate": 1.6024343703018386e-06, "loss": 0.5884, "step": 338990 }, { "epoch": 29.36, "learning_rate": 1.6002659882040011e-06, "loss": 0.6364, "step": 339000 }, { "epoch": 29.36, "learning_rate": 1.598097606106164e-06, "loss": 0.737, "step": 339010 }, { "epoch": 29.36, "learning_rate": 1.5959292240083265e-06, "loss": 0.6226, "step": 339020 }, { "epoch": 29.36, "learning_rate": 1.593760841910489e-06, "loss": 0.5324, "step": 339030 }, { "epoch": 29.36, "learning_rate": 1.5915924598126515e-06, "loss": 0.5448, "step": 339040 }, { "epoch": 29.36, "learning_rate": 1.5894240777148142e-06, "loss": 0.4686, "step": 339050 }, { "epoch": 29.36, "learning_rate": 1.5872556956169768e-06, "loss": 0.6136, "step": 339060 }, { "epoch": 29.37, "learning_rate": 1.5850873135191394e-06, "loss": 0.5191, "step": 339070 }, { "epoch": 29.37, "learning_rate": 1.5829189314213022e-06, "loss": 0.618, "step": 339080 }, { "epoch": 29.37, "learning_rate": 1.5807505493234648e-06, "loss": 0.7031, "step": 339090 }, { "epoch": 29.37, "learning_rate": 1.5785821672256271e-06, "loss": 0.5601, "step": 339100 }, { "epoch": 29.37, "learning_rate": 1.5764137851277897e-06, "loss": 0.5225, "step": 339110 }, { "epoch": 29.37, "learning_rate": 1.5742454030299523e-06, "loss": 0.5084, "step": 339120 }, { "epoch": 29.37, "learning_rate": 1.572077020932115e-06, "loss": 0.5952, "step": 339130 }, { "epoch": 29.37, "learning_rate": 1.5699086388342777e-06, "loss": 0.5912, "step": 339140 }, { "epoch": 29.37, "learning_rate": 1.5677402567364403e-06, "loss": 0.5795, "step": 339150 }, { "epoch": 29.37, "learning_rate": 1.565571874638603e-06, "loss": 0.4716, "step": 339160 }, { "epoch": 29.37, "learning_rate": 1.5634034925407654e-06, "loss": 0.5651, "step": 339170 }, { "epoch": 29.38, "learning_rate": 1.561235110442928e-06, "loss": 0.6163, "step": 339180 }, { "epoch": 29.38, "learning_rate": 1.5590667283450906e-06, "loss": 0.5157, "step": 339190 }, { "epoch": 29.38, "learning_rate": 1.5568983462472532e-06, "loss": 0.5423, "step": 339200 }, { "epoch": 29.38, "learning_rate": 1.554729964149416e-06, "loss": 0.5413, "step": 339210 }, { "epoch": 29.38, "learning_rate": 1.5525615820515785e-06, "loss": 0.512, "step": 339220 }, { "epoch": 29.38, "learning_rate": 1.550393199953741e-06, "loss": 0.5754, "step": 339230 }, { "epoch": 29.38, "learning_rate": 1.5482248178559035e-06, "loss": 0.6107, "step": 339240 }, { "epoch": 29.38, "learning_rate": 1.5460564357580663e-06, "loss": 0.57, "step": 339250 }, { "epoch": 29.38, "learning_rate": 1.5438880536602289e-06, "loss": 0.6927, "step": 339260 }, { "epoch": 29.38, "learning_rate": 1.5417196715623914e-06, "loss": 0.6531, "step": 339270 }, { "epoch": 29.38, "learning_rate": 1.5395512894645542e-06, "loss": 0.6199, "step": 339280 }, { "epoch": 29.38, "learning_rate": 1.5373829073667168e-06, "loss": 0.5837, "step": 339290 }, { "epoch": 29.39, "learning_rate": 1.5352145252688792e-06, "loss": 0.5864, "step": 339300 }, { "epoch": 29.39, "learning_rate": 1.5330461431710418e-06, "loss": 0.5949, "step": 339310 }, { "epoch": 29.39, "learning_rate": 1.5308777610732043e-06, "loss": 0.5285, "step": 339320 }, { "epoch": 29.39, "learning_rate": 1.5287093789753671e-06, "loss": 0.6016, "step": 339330 }, { "epoch": 29.39, "learning_rate": 1.5265409968775297e-06, "loss": 0.5226, "step": 339340 }, { "epoch": 29.39, "learning_rate": 1.5243726147796923e-06, "loss": 0.6118, "step": 339350 }, { "epoch": 29.39, "learning_rate": 1.5222042326818547e-06, "loss": 0.5935, "step": 339360 }, { "epoch": 29.39, "learning_rate": 1.5200358505840174e-06, "loss": 0.7406, "step": 339370 }, { "epoch": 29.39, "learning_rate": 1.51786746848618e-06, "loss": 0.6344, "step": 339380 }, { "epoch": 29.39, "learning_rate": 1.5156990863883426e-06, "loss": 0.5649, "step": 339390 }, { "epoch": 29.39, "learning_rate": 1.5135307042905054e-06, "loss": 0.6667, "step": 339400 }, { "epoch": 29.4, "learning_rate": 1.511362322192668e-06, "loss": 0.5208, "step": 339410 }, { "epoch": 29.4, "learning_rate": 1.5091939400948306e-06, "loss": 0.4918, "step": 339420 }, { "epoch": 29.4, "learning_rate": 1.507025557996993e-06, "loss": 0.4562, "step": 339430 }, { "epoch": 29.4, "learning_rate": 1.5048571758991555e-06, "loss": 0.5855, "step": 339440 }, { "epoch": 29.4, "learning_rate": 1.5026887938013183e-06, "loss": 0.5316, "step": 339450 }, { "epoch": 29.4, "learning_rate": 1.5005204117034809e-06, "loss": 0.6005, "step": 339460 }, { "epoch": 29.4, "learning_rate": 1.4983520296056435e-06, "loss": 0.5971, "step": 339470 }, { "epoch": 29.4, "learning_rate": 1.4961836475078062e-06, "loss": 0.6207, "step": 339480 }, { "epoch": 29.4, "learning_rate": 1.4940152654099686e-06, "loss": 0.4647, "step": 339490 }, { "epoch": 29.4, "learning_rate": 1.4918468833121312e-06, "loss": 0.6641, "step": 339500 }, { "epoch": 29.4, "learning_rate": 1.4896785012142938e-06, "loss": 0.5783, "step": 339510 }, { "epoch": 29.4, "learning_rate": 1.4875101191164564e-06, "loss": 0.5718, "step": 339520 }, { "epoch": 29.41, "learning_rate": 1.4853417370186191e-06, "loss": 0.5348, "step": 339530 }, { "epoch": 29.41, "learning_rate": 1.4831733549207817e-06, "loss": 0.642, "step": 339540 }, { "epoch": 29.41, "learning_rate": 1.481004972822944e-06, "loss": 0.5224, "step": 339550 }, { "epoch": 29.41, "learning_rate": 1.4788365907251067e-06, "loss": 0.6028, "step": 339560 }, { "epoch": 29.41, "learning_rate": 1.4766682086272695e-06, "loss": 0.5571, "step": 339570 }, { "epoch": 29.41, "learning_rate": 1.474499826529432e-06, "loss": 0.5677, "step": 339580 }, { "epoch": 29.41, "learning_rate": 1.4723314444315946e-06, "loss": 0.5695, "step": 339590 }, { "epoch": 29.41, "learning_rate": 1.4701630623337574e-06, "loss": 0.5535, "step": 339600 }, { "epoch": 29.41, "learning_rate": 1.46799468023592e-06, "loss": 0.6887, "step": 339610 }, { "epoch": 29.41, "learning_rate": 1.4658262981380824e-06, "loss": 0.6563, "step": 339620 }, { "epoch": 29.41, "learning_rate": 1.463657916040245e-06, "loss": 0.6318, "step": 339630 }, { "epoch": 29.41, "learning_rate": 1.4614895339424075e-06, "loss": 0.7272, "step": 339640 }, { "epoch": 29.42, "learning_rate": 1.4593211518445703e-06, "loss": 0.5408, "step": 339650 }, { "epoch": 29.42, "learning_rate": 1.457152769746733e-06, "loss": 0.5023, "step": 339660 }, { "epoch": 29.42, "learning_rate": 1.4549843876488955e-06, "loss": 0.5303, "step": 339670 }, { "epoch": 29.42, "learning_rate": 1.4528160055510578e-06, "loss": 0.5435, "step": 339680 }, { "epoch": 29.42, "learning_rate": 1.4506476234532206e-06, "loss": 0.4837, "step": 339690 }, { "epoch": 29.42, "learning_rate": 1.4484792413553832e-06, "loss": 0.5099, "step": 339700 }, { "epoch": 29.42, "learning_rate": 1.4463108592575458e-06, "loss": 0.5404, "step": 339710 }, { "epoch": 29.42, "learning_rate": 1.4441424771597086e-06, "loss": 0.5646, "step": 339720 }, { "epoch": 29.42, "learning_rate": 1.4419740950618712e-06, "loss": 0.5872, "step": 339730 }, { "epoch": 29.42, "learning_rate": 1.4398057129640338e-06, "loss": 0.5718, "step": 339740 }, { "epoch": 29.42, "learning_rate": 1.4376373308661961e-06, "loss": 0.3067, "step": 339750 }, { "epoch": 29.43, "learning_rate": 1.4354689487683587e-06, "loss": 0.6666, "step": 339760 }, { "epoch": 29.43, "learning_rate": 1.4333005666705215e-06, "loss": 0.5565, "step": 339770 }, { "epoch": 29.43, "learning_rate": 1.431132184572684e-06, "loss": 0.6402, "step": 339780 }, { "epoch": 29.43, "learning_rate": 1.4289638024748467e-06, "loss": 0.5995, "step": 339790 }, { "epoch": 29.43, "learning_rate": 1.4267954203770094e-06, "loss": 0.5865, "step": 339800 }, { "epoch": 29.43, "learning_rate": 1.4246270382791718e-06, "loss": 0.617, "step": 339810 }, { "epoch": 29.43, "learning_rate": 1.4224586561813344e-06, "loss": 0.5463, "step": 339820 }, { "epoch": 29.43, "learning_rate": 1.420290274083497e-06, "loss": 0.4652, "step": 339830 }, { "epoch": 29.43, "learning_rate": 1.4181218919856596e-06, "loss": 0.5583, "step": 339840 }, { "epoch": 29.43, "learning_rate": 1.4159535098878223e-06, "loss": 0.4855, "step": 339850 }, { "epoch": 29.43, "learning_rate": 1.413785127789985e-06, "loss": 0.5007, "step": 339860 }, { "epoch": 29.43, "learning_rate": 1.4116167456921473e-06, "loss": 0.5868, "step": 339870 }, { "epoch": 29.44, "learning_rate": 1.4094483635943099e-06, "loss": 0.5083, "step": 339880 }, { "epoch": 29.44, "learning_rate": 1.4072799814964727e-06, "loss": 0.5323, "step": 339890 }, { "epoch": 29.44, "learning_rate": 1.4051115993986352e-06, "loss": 0.559, "step": 339900 }, { "epoch": 29.44, "learning_rate": 1.4029432173007978e-06, "loss": 0.5412, "step": 339910 }, { "epoch": 29.44, "learning_rate": 1.4007748352029606e-06, "loss": 0.5669, "step": 339920 }, { "epoch": 29.44, "learning_rate": 1.3986064531051232e-06, "loss": 0.5783, "step": 339930 }, { "epoch": 29.44, "learning_rate": 1.3964380710072856e-06, "loss": 0.5959, "step": 339940 }, { "epoch": 29.44, "learning_rate": 1.3942696889094481e-06, "loss": 0.5999, "step": 339950 }, { "epoch": 29.44, "learning_rate": 1.3921013068116107e-06, "loss": 0.5072, "step": 339960 }, { "epoch": 29.44, "learning_rate": 1.3899329247137735e-06, "loss": 0.4896, "step": 339970 }, { "epoch": 29.44, "learning_rate": 1.387764542615936e-06, "loss": 0.522, "step": 339980 }, { "epoch": 29.45, "learning_rate": 1.3855961605180987e-06, "loss": 0.5896, "step": 339990 }, { "epoch": 29.45, "learning_rate": 1.383427778420261e-06, "loss": 0.5969, "step": 340000 }, { "epoch": 29.45, "learning_rate": 1.3812593963224238e-06, "loss": 0.592, "step": 340010 }, { "epoch": 29.45, "learning_rate": 1.3790910142245864e-06, "loss": 0.5752, "step": 340020 }, { "epoch": 29.45, "learning_rate": 1.376922632126749e-06, "loss": 0.6822, "step": 340030 }, { "epoch": 29.45, "learning_rate": 1.3747542500289118e-06, "loss": 0.5996, "step": 340040 }, { "epoch": 29.45, "learning_rate": 1.3725858679310744e-06, "loss": 0.5598, "step": 340050 }, { "epoch": 29.45, "learning_rate": 1.370417485833237e-06, "loss": 0.5136, "step": 340060 }, { "epoch": 29.45, "learning_rate": 1.3682491037353993e-06, "loss": 0.5488, "step": 340070 }, { "epoch": 29.45, "learning_rate": 1.3660807216375619e-06, "loss": 0.5362, "step": 340080 }, { "epoch": 29.45, "learning_rate": 1.3639123395397247e-06, "loss": 0.6032, "step": 340090 }, { "epoch": 29.45, "learning_rate": 1.3617439574418873e-06, "loss": 0.5574, "step": 340100 }, { "epoch": 29.46, "learning_rate": 1.3595755753440498e-06, "loss": 0.552, "step": 340110 }, { "epoch": 29.46, "learning_rate": 1.3574071932462126e-06, "loss": 0.7012, "step": 340120 }, { "epoch": 29.46, "learning_rate": 1.355238811148375e-06, "loss": 0.5894, "step": 340130 }, { "epoch": 29.46, "learning_rate": 1.3530704290505376e-06, "loss": 0.683, "step": 340140 }, { "epoch": 29.46, "learning_rate": 1.3509020469527002e-06, "loss": 0.6266, "step": 340150 }, { "epoch": 29.46, "learning_rate": 1.3487336648548627e-06, "loss": 0.5773, "step": 340160 }, { "epoch": 29.46, "learning_rate": 1.3465652827570255e-06, "loss": 0.5156, "step": 340170 }, { "epoch": 29.46, "learning_rate": 1.3443969006591881e-06, "loss": 0.644, "step": 340180 }, { "epoch": 29.46, "learning_rate": 1.3422285185613505e-06, "loss": 0.618, "step": 340190 }, { "epoch": 29.46, "learning_rate": 1.340060136463513e-06, "loss": 0.6893, "step": 340200 }, { "epoch": 29.46, "learning_rate": 1.3378917543656759e-06, "loss": 0.705, "step": 340210 }, { "epoch": 29.47, "learning_rate": 1.3357233722678384e-06, "loss": 0.5363, "step": 340220 }, { "epoch": 29.47, "learning_rate": 1.333554990170001e-06, "loss": 0.5318, "step": 340230 }, { "epoch": 29.47, "learning_rate": 1.3313866080721638e-06, "loss": 0.5076, "step": 340240 }, { "epoch": 29.47, "learning_rate": 1.3292182259743264e-06, "loss": 0.5699, "step": 340250 }, { "epoch": 29.47, "learning_rate": 1.3270498438764888e-06, "loss": 0.5326, "step": 340260 }, { "epoch": 29.47, "learning_rate": 1.3248814617786513e-06, "loss": 0.562, "step": 340270 }, { "epoch": 29.47, "learning_rate": 1.322713079680814e-06, "loss": 0.6852, "step": 340280 }, { "epoch": 29.47, "learning_rate": 1.3205446975829767e-06, "loss": 0.5863, "step": 340290 }, { "epoch": 29.47, "learning_rate": 1.3183763154851393e-06, "loss": 0.5437, "step": 340300 }, { "epoch": 29.47, "learning_rate": 1.3162079333873019e-06, "loss": 0.5659, "step": 340310 }, { "epoch": 29.47, "learning_rate": 1.3140395512894642e-06, "loss": 0.5786, "step": 340320 }, { "epoch": 29.47, "learning_rate": 1.311871169191627e-06, "loss": 0.7004, "step": 340330 }, { "epoch": 29.48, "learning_rate": 1.3097027870937896e-06, "loss": 0.6024, "step": 340340 }, { "epoch": 29.48, "learning_rate": 1.3075344049959522e-06, "loss": 0.6916, "step": 340350 }, { "epoch": 29.48, "learning_rate": 1.305366022898115e-06, "loss": 0.4419, "step": 340360 }, { "epoch": 29.48, "learning_rate": 1.3031976408002776e-06, "loss": 0.6614, "step": 340370 }, { "epoch": 29.48, "learning_rate": 1.3010292587024401e-06, "loss": 0.6002, "step": 340380 }, { "epoch": 29.48, "learning_rate": 1.2988608766046025e-06, "loss": 0.4714, "step": 340390 }, { "epoch": 29.48, "learning_rate": 1.296692494506765e-06, "loss": 0.5416, "step": 340400 }, { "epoch": 29.48, "learning_rate": 1.2945241124089279e-06, "loss": 0.5502, "step": 340410 }, { "epoch": 29.48, "learning_rate": 1.2923557303110905e-06, "loss": 0.4652, "step": 340420 }, { "epoch": 29.48, "learning_rate": 1.290187348213253e-06, "loss": 0.5788, "step": 340430 }, { "epoch": 29.48, "learning_rate": 1.2880189661154158e-06, "loss": 0.4937, "step": 340440 }, { "epoch": 29.49, "learning_rate": 1.2858505840175782e-06, "loss": 0.4951, "step": 340450 }, { "epoch": 29.49, "learning_rate": 1.2836822019197408e-06, "loss": 0.5287, "step": 340460 }, { "epoch": 29.49, "learning_rate": 1.2815138198219034e-06, "loss": 0.4958, "step": 340470 }, { "epoch": 29.49, "learning_rate": 1.279345437724066e-06, "loss": 0.4178, "step": 340480 }, { "epoch": 29.49, "learning_rate": 1.2771770556262287e-06, "loss": 0.4698, "step": 340490 }, { "epoch": 29.49, "learning_rate": 1.2750086735283913e-06, "loss": 0.5945, "step": 340500 }, { "epoch": 29.49, "learning_rate": 1.2728402914305539e-06, "loss": 0.6567, "step": 340510 }, { "epoch": 29.49, "learning_rate": 1.2706719093327163e-06, "loss": 0.5908, "step": 340520 }, { "epoch": 29.49, "learning_rate": 1.268503527234879e-06, "loss": 0.6299, "step": 340530 }, { "epoch": 29.49, "learning_rate": 1.2663351451370416e-06, "loss": 0.6567, "step": 340540 }, { "epoch": 29.49, "learning_rate": 1.2641667630392042e-06, "loss": 0.6099, "step": 340550 }, { "epoch": 29.49, "learning_rate": 1.261998380941367e-06, "loss": 0.542, "step": 340560 }, { "epoch": 29.5, "learning_rate": 1.2598299988435296e-06, "loss": 0.4735, "step": 340570 }, { "epoch": 29.5, "learning_rate": 1.257661616745692e-06, "loss": 0.5586, "step": 340580 }, { "epoch": 29.5, "learning_rate": 1.2554932346478545e-06, "loss": 0.5881, "step": 340590 }, { "epoch": 29.5, "learning_rate": 1.2533248525500171e-06, "loss": 0.5624, "step": 340600 }, { "epoch": 29.5, "learning_rate": 1.25115647045218e-06, "loss": 0.4252, "step": 340610 }, { "epoch": 29.5, "learning_rate": 1.2489880883543425e-06, "loss": 0.4457, "step": 340620 }, { "epoch": 29.5, "learning_rate": 1.246819706256505e-06, "loss": 0.6154, "step": 340630 }, { "epoch": 29.5, "learning_rate": 1.2446513241586674e-06, "loss": 0.5489, "step": 340640 }, { "epoch": 29.5, "learning_rate": 1.2424829420608302e-06, "loss": 0.6023, "step": 340650 }, { "epoch": 29.5, "learning_rate": 1.2403145599629928e-06, "loss": 0.6389, "step": 340660 }, { "epoch": 29.5, "learning_rate": 1.2381461778651554e-06, "loss": 0.5893, "step": 340670 }, { "epoch": 29.51, "learning_rate": 1.2359777957673182e-06, "loss": 0.7287, "step": 340680 }, { "epoch": 29.51, "learning_rate": 1.2338094136694808e-06, "loss": 0.5175, "step": 340690 }, { "epoch": 29.51, "learning_rate": 1.2316410315716433e-06, "loss": 0.5135, "step": 340700 }, { "epoch": 29.51, "learning_rate": 1.2294726494738057e-06, "loss": 0.6395, "step": 340710 }, { "epoch": 29.51, "learning_rate": 1.2273042673759683e-06, "loss": 0.5451, "step": 340720 }, { "epoch": 29.51, "learning_rate": 1.225135885278131e-06, "loss": 0.458, "step": 340730 }, { "epoch": 29.51, "learning_rate": 1.2229675031802937e-06, "loss": 0.5651, "step": 340740 }, { "epoch": 29.51, "learning_rate": 1.2207991210824562e-06, "loss": 0.584, "step": 340750 }, { "epoch": 29.51, "learning_rate": 1.218630738984619e-06, "loss": 0.7417, "step": 340760 }, { "epoch": 29.51, "learning_rate": 1.2164623568867814e-06, "loss": 0.477, "step": 340770 }, { "epoch": 29.51, "learning_rate": 1.214293974788944e-06, "loss": 0.6018, "step": 340780 }, { "epoch": 29.51, "learning_rate": 1.2121255926911066e-06, "loss": 0.5417, "step": 340790 }, { "epoch": 29.52, "learning_rate": 1.2099572105932691e-06, "loss": 0.6894, "step": 340800 }, { "epoch": 29.52, "learning_rate": 1.207788828495432e-06, "loss": 0.5783, "step": 340810 }, { "epoch": 29.52, "learning_rate": 1.2056204463975945e-06, "loss": 0.456, "step": 340820 }, { "epoch": 29.52, "learning_rate": 1.203452064299757e-06, "loss": 0.5719, "step": 340830 }, { "epoch": 29.52, "learning_rate": 1.2012836822019195e-06, "loss": 0.5296, "step": 340840 }, { "epoch": 29.52, "learning_rate": 1.1991153001040822e-06, "loss": 0.6128, "step": 340850 }, { "epoch": 29.52, "learning_rate": 1.1969469180062448e-06, "loss": 0.5866, "step": 340860 }, { "epoch": 29.52, "learning_rate": 1.1947785359084074e-06, "loss": 0.4764, "step": 340870 }, { "epoch": 29.52, "learning_rate": 1.1926101538105702e-06, "loss": 0.5187, "step": 340880 }, { "epoch": 29.52, "learning_rate": 1.1904417717127328e-06, "loss": 0.5273, "step": 340890 }, { "epoch": 29.52, "learning_rate": 1.1882733896148951e-06, "loss": 0.4832, "step": 340900 }, { "epoch": 29.52, "learning_rate": 1.1861050075170577e-06, "loss": 0.5753, "step": 340910 }, { "epoch": 29.53, "learning_rate": 1.1839366254192203e-06, "loss": 0.6076, "step": 340920 }, { "epoch": 29.53, "learning_rate": 1.181768243321383e-06, "loss": 0.5793, "step": 340930 }, { "epoch": 29.53, "learning_rate": 1.1795998612235457e-06, "loss": 0.6323, "step": 340940 }, { "epoch": 29.53, "learning_rate": 1.1774314791257083e-06, "loss": 0.6154, "step": 340950 }, { "epoch": 29.53, "learning_rate": 1.1752630970278706e-06, "loss": 0.5715, "step": 340960 }, { "epoch": 29.53, "learning_rate": 1.1730947149300334e-06, "loss": 0.5803, "step": 340970 }, { "epoch": 29.53, "learning_rate": 1.170926332832196e-06, "loss": 0.5048, "step": 340980 }, { "epoch": 29.53, "learning_rate": 1.1687579507343586e-06, "loss": 0.6512, "step": 340990 }, { "epoch": 29.53, "learning_rate": 1.1665895686365214e-06, "loss": 0.6098, "step": 341000 }, { "epoch": 29.53, "learning_rate": 1.1644211865386837e-06, "loss": 0.5893, "step": 341010 }, { "epoch": 29.53, "learning_rate": 1.1622528044408463e-06, "loss": 0.5149, "step": 341020 }, { "epoch": 29.54, "learning_rate": 1.1600844223430091e-06, "loss": 0.4359, "step": 341030 }, { "epoch": 29.54, "learning_rate": 1.1579160402451715e-06, "loss": 0.6117, "step": 341040 }, { "epoch": 29.54, "learning_rate": 1.1557476581473343e-06, "loss": 0.5205, "step": 341050 }, { "epoch": 29.54, "learning_rate": 1.1535792760494969e-06, "loss": 0.531, "step": 341060 }, { "epoch": 29.54, "learning_rate": 1.1514108939516594e-06, "loss": 0.5616, "step": 341070 }, { "epoch": 29.54, "learning_rate": 1.149242511853822e-06, "loss": 0.5534, "step": 341080 }, { "epoch": 29.54, "learning_rate": 1.1470741297559846e-06, "loss": 0.54, "step": 341090 }, { "epoch": 29.54, "learning_rate": 1.1449057476581474e-06, "loss": 0.5709, "step": 341100 }, { "epoch": 29.54, "learning_rate": 1.1427373655603098e-06, "loss": 0.5237, "step": 341110 }, { "epoch": 29.54, "learning_rate": 1.1405689834624723e-06, "loss": 0.4903, "step": 341120 }, { "epoch": 29.54, "learning_rate": 1.1384006013646351e-06, "loss": 0.5274, "step": 341130 }, { "epoch": 29.54, "learning_rate": 1.1362322192667975e-06, "loss": 0.5472, "step": 341140 }, { "epoch": 29.55, "learning_rate": 1.1340638371689603e-06, "loss": 0.4757, "step": 341150 }, { "epoch": 29.55, "learning_rate": 1.1318954550711229e-06, "loss": 0.5463, "step": 341160 }, { "epoch": 29.55, "learning_rate": 1.1297270729732854e-06, "loss": 0.5202, "step": 341170 }, { "epoch": 29.55, "learning_rate": 1.127558690875448e-06, "loss": 0.5018, "step": 341180 }, { "epoch": 29.55, "learning_rate": 1.1253903087776106e-06, "loss": 0.5011, "step": 341190 }, { "epoch": 29.55, "learning_rate": 1.1232219266797734e-06, "loss": 0.5272, "step": 341200 }, { "epoch": 29.55, "learning_rate": 1.1210535445819358e-06, "loss": 0.6618, "step": 341210 }, { "epoch": 29.55, "learning_rate": 1.1188851624840983e-06, "loss": 0.6392, "step": 341220 }, { "epoch": 29.55, "learning_rate": 1.1167167803862611e-06, "loss": 0.4086, "step": 341230 }, { "epoch": 29.55, "learning_rate": 1.1145483982884235e-06, "loss": 0.5896, "step": 341240 }, { "epoch": 29.55, "learning_rate": 1.1123800161905863e-06, "loss": 0.5965, "step": 341250 }, { "epoch": 29.56, "learning_rate": 1.1102116340927489e-06, "loss": 0.4423, "step": 341260 }, { "epoch": 29.56, "learning_rate": 1.1080432519949115e-06, "loss": 0.5247, "step": 341270 }, { "epoch": 29.56, "learning_rate": 1.105874869897074e-06, "loss": 0.5775, "step": 341280 }, { "epoch": 29.56, "learning_rate": 1.1037064877992366e-06, "loss": 0.6577, "step": 341290 }, { "epoch": 29.56, "learning_rate": 1.1015381057013992e-06, "loss": 0.486, "step": 341300 }, { "epoch": 29.56, "learning_rate": 1.0993697236035618e-06, "loss": 0.6189, "step": 341310 }, { "epoch": 29.56, "learning_rate": 1.0972013415057246e-06, "loss": 0.5657, "step": 341320 }, { "epoch": 29.56, "learning_rate": 1.095032959407887e-06, "loss": 0.7371, "step": 341330 }, { "epoch": 29.56, "learning_rate": 1.0928645773100495e-06, "loss": 0.4502, "step": 341340 }, { "epoch": 29.56, "learning_rate": 1.0906961952122123e-06, "loss": 0.6232, "step": 341350 }, { "epoch": 29.56, "learning_rate": 1.0885278131143749e-06, "loss": 0.6273, "step": 341360 }, { "epoch": 29.56, "learning_rate": 1.0863594310165375e-06, "loss": 0.6963, "step": 341370 }, { "epoch": 29.57, "learning_rate": 1.0841910489187e-06, "loss": 0.5607, "step": 341380 }, { "epoch": 29.57, "learning_rate": 1.0820226668208626e-06, "loss": 0.5519, "step": 341390 }, { "epoch": 29.57, "learning_rate": 1.0798542847230252e-06, "loss": 0.4608, "step": 341400 }, { "epoch": 29.57, "learning_rate": 1.0776859026251878e-06, "loss": 0.5652, "step": 341410 }, { "epoch": 29.57, "learning_rate": 1.0755175205273506e-06, "loss": 0.6735, "step": 341420 }, { "epoch": 29.57, "learning_rate": 1.073349138429513e-06, "loss": 0.4758, "step": 341430 }, { "epoch": 29.57, "learning_rate": 1.0711807563316755e-06, "loss": 0.4944, "step": 341440 }, { "epoch": 29.57, "learning_rate": 1.0690123742338383e-06, "loss": 0.611, "step": 341450 }, { "epoch": 29.57, "learning_rate": 1.0668439921360007e-06, "loss": 0.6755, "step": 341460 }, { "epoch": 29.57, "learning_rate": 1.0646756100381635e-06, "loss": 0.5005, "step": 341470 }, { "epoch": 29.57, "learning_rate": 1.062507227940326e-06, "loss": 0.4681, "step": 341480 }, { "epoch": 29.58, "learning_rate": 1.0603388458424886e-06, "loss": 0.5342, "step": 341490 }, { "epoch": 29.58, "learning_rate": 1.0581704637446512e-06, "loss": 0.463, "step": 341500 }, { "epoch": 29.58, "learning_rate": 1.0560020816468138e-06, "loss": 0.5245, "step": 341510 }, { "epoch": 29.58, "learning_rate": 1.0538336995489766e-06, "loss": 0.6335, "step": 341520 }, { "epoch": 29.58, "learning_rate": 1.051665317451139e-06, "loss": 0.5942, "step": 341530 }, { "epoch": 29.58, "learning_rate": 1.0494969353533015e-06, "loss": 0.6327, "step": 341540 }, { "epoch": 29.58, "learning_rate": 1.0473285532554643e-06, "loss": 0.5203, "step": 341550 }, { "epoch": 29.58, "learning_rate": 1.0451601711576267e-06, "loss": 0.6377, "step": 341560 }, { "epoch": 29.58, "learning_rate": 1.0429917890597895e-06, "loss": 0.641, "step": 341570 }, { "epoch": 29.58, "learning_rate": 1.040823406961952e-06, "loss": 0.5636, "step": 341580 }, { "epoch": 29.58, "learning_rate": 1.0386550248641147e-06, "loss": 0.5771, "step": 341590 }, { "epoch": 29.58, "learning_rate": 1.0364866427662772e-06, "loss": 0.5068, "step": 341600 }, { "epoch": 29.59, "learning_rate": 1.0343182606684398e-06, "loss": 0.6581, "step": 341610 }, { "epoch": 29.59, "learning_rate": 1.0321498785706024e-06, "loss": 0.435, "step": 341620 }, { "epoch": 29.59, "learning_rate": 1.029981496472765e-06, "loss": 0.5523, "step": 341630 }, { "epoch": 29.59, "learning_rate": 1.0278131143749278e-06, "loss": 0.5916, "step": 341640 }, { "epoch": 29.59, "learning_rate": 1.0256447322770901e-06, "loss": 0.6453, "step": 341650 }, { "epoch": 29.59, "learning_rate": 1.0234763501792527e-06, "loss": 0.3872, "step": 341660 }, { "epoch": 29.59, "learning_rate": 1.0213079680814155e-06, "loss": 0.5932, "step": 341670 }, { "epoch": 29.59, "learning_rate": 1.019139585983578e-06, "loss": 0.5968, "step": 341680 }, { "epoch": 29.59, "learning_rate": 1.0169712038857407e-06, "loss": 0.7417, "step": 341690 }, { "epoch": 29.59, "learning_rate": 1.0148028217879032e-06, "loss": 0.6343, "step": 341700 }, { "epoch": 29.59, "learning_rate": 1.0126344396900658e-06, "loss": 0.5255, "step": 341710 }, { "epoch": 29.6, "learning_rate": 1.0104660575922284e-06, "loss": 0.4266, "step": 341720 }, { "epoch": 29.6, "learning_rate": 1.008297675494391e-06, "loss": 0.5875, "step": 341730 }, { "epoch": 29.6, "learning_rate": 1.0061292933965538e-06, "loss": 0.6178, "step": 341740 }, { "epoch": 29.6, "learning_rate": 1.0039609112987161e-06, "loss": 0.4854, "step": 341750 }, { "epoch": 29.6, "learning_rate": 1.0017925292008787e-06, "loss": 0.5634, "step": 341760 }, { "epoch": 29.6, "learning_rate": 9.996241471030415e-07, "loss": 0.706, "step": 341770 }, { "epoch": 29.6, "learning_rate": 9.974557650052039e-07, "loss": 0.6105, "step": 341780 }, { "epoch": 29.6, "learning_rate": 9.952873829073667e-07, "loss": 0.562, "step": 341790 }, { "epoch": 29.6, "learning_rate": 9.931190008095293e-07, "loss": 0.684, "step": 341800 }, { "epoch": 29.6, "learning_rate": 9.909506187116918e-07, "loss": 0.6623, "step": 341810 }, { "epoch": 29.6, "learning_rate": 9.887822366138544e-07, "loss": 0.5949, "step": 341820 }, { "epoch": 29.6, "learning_rate": 9.86613854516017e-07, "loss": 0.4112, "step": 341830 }, { "epoch": 29.61, "learning_rate": 9.844454724181798e-07, "loss": 0.6833, "step": 341840 }, { "epoch": 29.61, "learning_rate": 9.822770903203422e-07, "loss": 0.5285, "step": 341850 }, { "epoch": 29.61, "learning_rate": 9.801087082225047e-07, "loss": 0.4272, "step": 341860 }, { "epoch": 29.61, "learning_rate": 9.779403261246675e-07, "loss": 0.6479, "step": 341870 }, { "epoch": 29.61, "learning_rate": 9.757719440268299e-07, "loss": 0.5934, "step": 341880 }, { "epoch": 29.61, "learning_rate": 9.736035619289927e-07, "loss": 0.5818, "step": 341890 }, { "epoch": 29.61, "learning_rate": 9.714351798311553e-07, "loss": 0.5852, "step": 341900 }, { "epoch": 29.61, "learning_rate": 9.692667977333178e-07, "loss": 0.6914, "step": 341910 }, { "epoch": 29.61, "learning_rate": 9.670984156354804e-07, "loss": 0.4291, "step": 341920 }, { "epoch": 29.61, "learning_rate": 9.64930033537643e-07, "loss": 0.4728, "step": 341930 }, { "epoch": 29.61, "learning_rate": 9.627616514398056e-07, "loss": 0.5938, "step": 341940 }, { "epoch": 29.62, "learning_rate": 9.605932693419682e-07, "loss": 0.4558, "step": 341950 }, { "epoch": 29.62, "learning_rate": 9.58424887244131e-07, "loss": 0.6427, "step": 341960 }, { "epoch": 29.62, "learning_rate": 9.562565051462933e-07, "loss": 0.7256, "step": 341970 }, { "epoch": 29.62, "learning_rate": 9.54088123048456e-07, "loss": 0.4876, "step": 341980 }, { "epoch": 29.62, "learning_rate": 9.519197409506187e-07, "loss": 0.5142, "step": 341990 }, { "epoch": 29.62, "learning_rate": 9.497513588527813e-07, "loss": 0.7307, "step": 342000 }, { "epoch": 29.62, "learning_rate": 9.475829767549438e-07, "loss": 0.6119, "step": 342010 }, { "epoch": 29.62, "learning_rate": 9.454145946571064e-07, "loss": 0.5019, "step": 342020 }, { "epoch": 29.62, "learning_rate": 9.432462125592691e-07, "loss": 0.5876, "step": 342030 }, { "epoch": 29.62, "learning_rate": 9.410778304614316e-07, "loss": 0.6862, "step": 342040 }, { "epoch": 29.62, "learning_rate": 9.389094483635942e-07, "loss": 0.6649, "step": 342050 }, { "epoch": 29.62, "learning_rate": 9.367410662657569e-07, "loss": 0.6197, "step": 342060 }, { "epoch": 29.63, "learning_rate": 9.345726841679193e-07, "loss": 0.5778, "step": 342070 }, { "epoch": 29.63, "learning_rate": 9.32404302070082e-07, "loss": 0.6111, "step": 342080 }, { "epoch": 29.63, "learning_rate": 9.302359199722447e-07, "loss": 0.6385, "step": 342090 }, { "epoch": 29.63, "learning_rate": 9.280675378744072e-07, "loss": 0.5537, "step": 342100 }, { "epoch": 29.63, "learning_rate": 9.258991557765698e-07, "loss": 0.5587, "step": 342110 }, { "epoch": 29.63, "learning_rate": 9.237307736787324e-07, "loss": 0.5225, "step": 342120 }, { "epoch": 29.63, "learning_rate": 9.215623915808951e-07, "loss": 0.6231, "step": 342130 }, { "epoch": 29.63, "learning_rate": 9.193940094830576e-07, "loss": 0.5365, "step": 342140 }, { "epoch": 29.63, "learning_rate": 9.172256273852203e-07, "loss": 0.6592, "step": 342150 }, { "epoch": 29.63, "learning_rate": 9.150572452873829e-07, "loss": 0.5877, "step": 342160 }, { "epoch": 29.63, "learning_rate": 9.128888631895453e-07, "loss": 0.4744, "step": 342170 }, { "epoch": 29.63, "learning_rate": 9.10720481091708e-07, "loss": 0.5886, "step": 342180 }, { "epoch": 29.64, "learning_rate": 9.085520989938707e-07, "loss": 0.6279, "step": 342190 }, { "epoch": 29.64, "learning_rate": 9.063837168960332e-07, "loss": 0.5983, "step": 342200 }, { "epoch": 29.64, "learning_rate": 9.042153347981958e-07, "loss": 0.5481, "step": 342210 }, { "epoch": 29.64, "learning_rate": 9.020469527003585e-07, "loss": 0.6278, "step": 342220 }, { "epoch": 29.64, "learning_rate": 8.998785706025209e-07, "loss": 0.5004, "step": 342230 }, { "epoch": 29.64, "learning_rate": 8.977101885046836e-07, "loss": 0.6157, "step": 342240 }, { "epoch": 29.64, "learning_rate": 8.955418064068463e-07, "loss": 0.6026, "step": 342250 }, { "epoch": 29.64, "learning_rate": 8.933734243090088e-07, "loss": 0.6014, "step": 342260 }, { "epoch": 29.64, "learning_rate": 8.912050422111714e-07, "loss": 0.5289, "step": 342270 }, { "epoch": 29.64, "learning_rate": 8.89036660113334e-07, "loss": 0.569, "step": 342280 }, { "epoch": 29.64, "learning_rate": 8.868682780154967e-07, "loss": 0.573, "step": 342290 }, { "epoch": 29.65, "learning_rate": 8.846998959176592e-07, "loss": 0.5442, "step": 342300 }, { "epoch": 29.65, "learning_rate": 8.825315138198219e-07, "loss": 0.6337, "step": 342310 }, { "epoch": 29.65, "learning_rate": 8.803631317219845e-07, "loss": 0.5205, "step": 342320 }, { "epoch": 29.65, "learning_rate": 8.781947496241469e-07, "loss": 0.6327, "step": 342330 }, { "epoch": 29.65, "learning_rate": 8.760263675263096e-07, "loss": 0.5009, "step": 342340 }, { "epoch": 29.65, "learning_rate": 8.738579854284723e-07, "loss": 0.4721, "step": 342350 }, { "epoch": 29.65, "learning_rate": 8.716896033306348e-07, "loss": 0.5828, "step": 342360 }, { "epoch": 29.65, "learning_rate": 8.695212212327974e-07, "loss": 0.5063, "step": 342370 }, { "epoch": 29.65, "learning_rate": 8.673528391349601e-07, "loss": 0.696, "step": 342380 }, { "epoch": 29.65, "learning_rate": 8.651844570371225e-07, "loss": 0.467, "step": 342390 }, { "epoch": 29.65, "learning_rate": 8.630160749392852e-07, "loss": 0.4969, "step": 342400 }, { "epoch": 29.65, "learning_rate": 8.608476928414479e-07, "loss": 0.5949, "step": 342410 }, { "epoch": 29.66, "learning_rate": 8.586793107436104e-07, "loss": 0.5069, "step": 342420 }, { "epoch": 29.66, "learning_rate": 8.56510928645773e-07, "loss": 0.5714, "step": 342430 }, { "epoch": 29.66, "learning_rate": 8.543425465479356e-07, "loss": 0.6445, "step": 342440 }, { "epoch": 29.66, "learning_rate": 8.521741644500983e-07, "loss": 0.6767, "step": 342450 }, { "epoch": 29.66, "learning_rate": 8.500057823522608e-07, "loss": 0.6557, "step": 342460 }, { "epoch": 29.66, "learning_rate": 8.478374002544234e-07, "loss": 0.6122, "step": 342470 }, { "epoch": 29.66, "learning_rate": 8.456690181565861e-07, "loss": 0.4657, "step": 342480 }, { "epoch": 29.66, "learning_rate": 8.435006360587485e-07, "loss": 0.5728, "step": 342490 }, { "epoch": 29.66, "learning_rate": 8.413322539609112e-07, "loss": 0.5912, "step": 342500 }, { "epoch": 29.66, "learning_rate": 8.391638718630739e-07, "loss": 0.6546, "step": 342510 }, { "epoch": 29.66, "learning_rate": 8.369954897652364e-07, "loss": 0.6256, "step": 342520 }, { "epoch": 29.67, "learning_rate": 8.34827107667399e-07, "loss": 0.6066, "step": 342530 }, { "epoch": 29.67, "learning_rate": 8.326587255695617e-07, "loss": 0.5644, "step": 342540 }, { "epoch": 29.67, "learning_rate": 8.304903434717241e-07, "loss": 0.6245, "step": 342550 }, { "epoch": 29.67, "learning_rate": 8.283219613738868e-07, "loss": 0.5628, "step": 342560 }, { "epoch": 29.67, "learning_rate": 8.261535792760495e-07, "loss": 0.5707, "step": 342570 }, { "epoch": 29.67, "learning_rate": 8.23985197178212e-07, "loss": 0.4546, "step": 342580 }, { "epoch": 29.67, "learning_rate": 8.218168150803746e-07, "loss": 0.585, "step": 342590 }, { "epoch": 29.67, "learning_rate": 8.196484329825372e-07, "loss": 0.6676, "step": 342600 }, { "epoch": 29.67, "learning_rate": 8.174800508846999e-07, "loss": 0.5956, "step": 342610 }, { "epoch": 29.67, "learning_rate": 8.153116687868624e-07, "loss": 0.4561, "step": 342620 }, { "epoch": 29.67, "learning_rate": 8.13143286689025e-07, "loss": 0.5736, "step": 342630 }, { "epoch": 29.67, "learning_rate": 8.109749045911877e-07, "loss": 0.5481, "step": 342640 }, { "epoch": 29.68, "learning_rate": 8.088065224933501e-07, "loss": 0.5463, "step": 342650 }, { "epoch": 29.68, "learning_rate": 8.066381403955128e-07, "loss": 0.5709, "step": 342660 }, { "epoch": 29.68, "learning_rate": 8.044697582976755e-07, "loss": 0.4181, "step": 342670 }, { "epoch": 29.68, "learning_rate": 8.02301376199838e-07, "loss": 0.6096, "step": 342680 }, { "epoch": 29.68, "learning_rate": 8.001329941020006e-07, "loss": 0.5938, "step": 342690 }, { "epoch": 29.68, "learning_rate": 7.979646120041633e-07, "loss": 0.4326, "step": 342700 }, { "epoch": 29.68, "learning_rate": 7.957962299063257e-07, "loss": 0.5416, "step": 342710 }, { "epoch": 29.68, "learning_rate": 7.936278478084884e-07, "loss": 0.6457, "step": 342720 }, { "epoch": 29.68, "learning_rate": 7.914594657106511e-07, "loss": 0.4775, "step": 342730 }, { "epoch": 29.68, "learning_rate": 7.892910836128136e-07, "loss": 0.5722, "step": 342740 }, { "epoch": 29.68, "learning_rate": 7.871227015149762e-07, "loss": 0.6018, "step": 342750 }, { "epoch": 29.69, "learning_rate": 7.849543194171388e-07, "loss": 0.5106, "step": 342760 }, { "epoch": 29.69, "learning_rate": 7.827859373193015e-07, "loss": 0.5758, "step": 342770 }, { "epoch": 29.69, "learning_rate": 7.80617555221464e-07, "loss": 0.7489, "step": 342780 }, { "epoch": 29.69, "learning_rate": 7.784491731236266e-07, "loss": 0.5089, "step": 342790 }, { "epoch": 29.69, "learning_rate": 7.762807910257893e-07, "loss": 0.5426, "step": 342800 }, { "epoch": 29.69, "learning_rate": 7.741124089279517e-07, "loss": 0.5341, "step": 342810 }, { "epoch": 29.69, "learning_rate": 7.719440268301144e-07, "loss": 0.5234, "step": 342820 }, { "epoch": 29.69, "learning_rate": 7.697756447322771e-07, "loss": 0.6466, "step": 342830 }, { "epoch": 29.69, "learning_rate": 7.676072626344396e-07, "loss": 0.4906, "step": 342840 }, { "epoch": 29.69, "learning_rate": 7.654388805366022e-07, "loss": 0.5464, "step": 342850 }, { "epoch": 29.69, "learning_rate": 7.632704984387649e-07, "loss": 0.4494, "step": 342860 }, { "epoch": 29.69, "learning_rate": 7.611021163409273e-07, "loss": 0.5726, "step": 342870 }, { "epoch": 29.7, "learning_rate": 7.5893373424309e-07, "loss": 0.5157, "step": 342880 }, { "epoch": 29.7, "learning_rate": 7.567653521452527e-07, "loss": 0.6129, "step": 342890 }, { "epoch": 29.7, "learning_rate": 7.545969700474153e-07, "loss": 0.5739, "step": 342900 }, { "epoch": 29.7, "learning_rate": 7.524285879495778e-07, "loss": 0.5559, "step": 342910 }, { "epoch": 29.7, "learning_rate": 7.502602058517404e-07, "loss": 0.5972, "step": 342920 }, { "epoch": 29.7, "learning_rate": 7.480918237539031e-07, "loss": 0.4942, "step": 342930 }, { "epoch": 29.7, "learning_rate": 7.459234416560656e-07, "loss": 0.5248, "step": 342940 }, { "epoch": 29.7, "learning_rate": 7.437550595582282e-07, "loss": 0.5638, "step": 342950 }, { "epoch": 29.7, "learning_rate": 7.415866774603909e-07, "loss": 0.4963, "step": 342960 }, { "epoch": 29.7, "learning_rate": 7.394182953625533e-07, "loss": 0.5546, "step": 342970 }, { "epoch": 29.7, "learning_rate": 7.37249913264716e-07, "loss": 0.4759, "step": 342980 }, { "epoch": 29.71, "learning_rate": 7.350815311668787e-07, "loss": 0.4949, "step": 342990 }, { "epoch": 29.71, "learning_rate": 7.329131490690412e-07, "loss": 0.5407, "step": 343000 }, { "epoch": 29.71, "learning_rate": 7.307447669712038e-07, "loss": 0.5492, "step": 343010 }, { "epoch": 29.71, "learning_rate": 7.285763848733664e-07, "loss": 0.6187, "step": 343020 }, { "epoch": 29.71, "learning_rate": 7.264080027755289e-07, "loss": 0.5864, "step": 343030 }, { "epoch": 29.71, "learning_rate": 7.242396206776916e-07, "loss": 0.536, "step": 343040 }, { "epoch": 29.71, "learning_rate": 7.220712385798543e-07, "loss": 0.5459, "step": 343050 }, { "epoch": 29.71, "learning_rate": 7.199028564820169e-07, "loss": 0.7206, "step": 343060 }, { "epoch": 29.71, "learning_rate": 7.177344743841793e-07, "loss": 0.5507, "step": 343070 }, { "epoch": 29.71, "learning_rate": 7.15566092286342e-07, "loss": 0.5615, "step": 343080 }, { "epoch": 29.71, "learning_rate": 7.133977101885047e-07, "loss": 0.4786, "step": 343090 }, { "epoch": 29.71, "learning_rate": 7.112293280906672e-07, "loss": 0.5324, "step": 343100 }, { "epoch": 29.72, "learning_rate": 7.090609459928298e-07, "loss": 0.5682, "step": 343110 }, { "epoch": 29.72, "learning_rate": 7.068925638949925e-07, "loss": 0.6957, "step": 343120 }, { "epoch": 29.72, "learning_rate": 7.047241817971549e-07, "loss": 0.5818, "step": 343130 }, { "epoch": 29.72, "learning_rate": 7.025557996993176e-07, "loss": 0.6804, "step": 343140 }, { "epoch": 29.72, "learning_rate": 7.003874176014803e-07, "loss": 0.6614, "step": 343150 }, { "epoch": 29.72, "learning_rate": 6.982190355036428e-07, "loss": 0.6587, "step": 343160 }, { "epoch": 29.72, "learning_rate": 6.960506534058054e-07, "loss": 0.5368, "step": 343170 }, { "epoch": 29.72, "learning_rate": 6.93882271307968e-07, "loss": 0.6377, "step": 343180 }, { "epoch": 29.72, "learning_rate": 6.917138892101305e-07, "loss": 0.6443, "step": 343190 }, { "epoch": 29.72, "learning_rate": 6.895455071122932e-07, "loss": 0.6084, "step": 343200 }, { "epoch": 29.72, "learning_rate": 6.873771250144559e-07, "loss": 0.5238, "step": 343210 }, { "epoch": 29.73, "learning_rate": 6.852087429166185e-07, "loss": 0.5609, "step": 343220 }, { "epoch": 29.73, "learning_rate": 6.830403608187809e-07, "loss": 0.5347, "step": 343230 }, { "epoch": 29.73, "learning_rate": 6.808719787209436e-07, "loss": 0.6439, "step": 343240 }, { "epoch": 29.73, "learning_rate": 6.787035966231063e-07, "loss": 0.7162, "step": 343250 }, { "epoch": 29.73, "learning_rate": 6.765352145252688e-07, "loss": 0.5789, "step": 343260 }, { "epoch": 29.73, "learning_rate": 6.743668324274314e-07, "loss": 0.6378, "step": 343270 }, { "epoch": 29.73, "learning_rate": 6.721984503295941e-07, "loss": 0.5495, "step": 343280 }, { "epoch": 29.73, "learning_rate": 6.700300682317565e-07, "loss": 0.4896, "step": 343290 }, { "epoch": 29.73, "learning_rate": 6.678616861339192e-07, "loss": 0.6054, "step": 343300 }, { "epoch": 29.73, "learning_rate": 6.656933040360819e-07, "loss": 0.5235, "step": 343310 }, { "epoch": 29.73, "learning_rate": 6.635249219382444e-07, "loss": 0.5582, "step": 343320 }, { "epoch": 29.73, "learning_rate": 6.61356539840407e-07, "loss": 0.522, "step": 343330 }, { "epoch": 29.74, "learning_rate": 6.591881577425696e-07, "loss": 0.4824, "step": 343340 }, { "epoch": 29.74, "learning_rate": 6.570197756447321e-07, "loss": 0.5779, "step": 343350 }, { "epoch": 29.74, "learning_rate": 6.548513935468948e-07, "loss": 0.6353, "step": 343360 }, { "epoch": 29.74, "learning_rate": 6.526830114490575e-07, "loss": 0.4728, "step": 343370 }, { "epoch": 29.74, "learning_rate": 6.505146293512201e-07, "loss": 0.6125, "step": 343380 }, { "epoch": 29.74, "learning_rate": 6.483462472533825e-07, "loss": 0.6072, "step": 343390 }, { "epoch": 29.74, "learning_rate": 6.461778651555452e-07, "loss": 0.5701, "step": 343400 }, { "epoch": 29.74, "learning_rate": 6.440094830577079e-07, "loss": 0.6144, "step": 343410 }, { "epoch": 29.74, "learning_rate": 6.418411009598704e-07, "loss": 0.5861, "step": 343420 }, { "epoch": 29.74, "learning_rate": 6.39672718862033e-07, "loss": 0.5508, "step": 343430 }, { "epoch": 29.74, "learning_rate": 6.375043367641957e-07, "loss": 0.5717, "step": 343440 }, { "epoch": 29.74, "learning_rate": 6.353359546663581e-07, "loss": 0.5612, "step": 343450 }, { "epoch": 29.75, "learning_rate": 6.331675725685208e-07, "loss": 0.6704, "step": 343460 }, { "epoch": 29.75, "learning_rate": 6.309991904706835e-07, "loss": 0.5416, "step": 343470 }, { "epoch": 29.75, "learning_rate": 6.28830808372846e-07, "loss": 0.5993, "step": 343480 }, { "epoch": 29.75, "learning_rate": 6.266624262750086e-07, "loss": 0.613, "step": 343490 }, { "epoch": 29.75, "learning_rate": 6.244940441771712e-07, "loss": 0.6102, "step": 343500 }, { "epoch": 29.75, "learning_rate": 6.223256620793337e-07, "loss": 0.6142, "step": 343510 }, { "epoch": 29.75, "learning_rate": 6.201572799814964e-07, "loss": 0.6915, "step": 343520 }, { "epoch": 29.75, "learning_rate": 6.179888978836591e-07, "loss": 0.591, "step": 343530 }, { "epoch": 29.75, "learning_rate": 6.158205157858217e-07, "loss": 0.5779, "step": 343540 }, { "epoch": 29.75, "learning_rate": 6.136521336879841e-07, "loss": 0.485, "step": 343550 }, { "epoch": 29.75, "learning_rate": 6.114837515901468e-07, "loss": 0.6301, "step": 343560 }, { "epoch": 29.76, "learning_rate": 6.093153694923095e-07, "loss": 0.4852, "step": 343570 }, { "epoch": 29.76, "learning_rate": 6.07146987394472e-07, "loss": 0.6911, "step": 343580 }, { "epoch": 29.76, "learning_rate": 6.049786052966346e-07, "loss": 0.5251, "step": 343590 }, { "epoch": 29.76, "learning_rate": 6.028102231987973e-07, "loss": 0.6611, "step": 343600 }, { "epoch": 29.76, "learning_rate": 6.006418411009597e-07, "loss": 0.5245, "step": 343610 }, { "epoch": 29.76, "learning_rate": 5.984734590031224e-07, "loss": 0.4207, "step": 343620 }, { "epoch": 29.76, "learning_rate": 5.963050769052851e-07, "loss": 0.5601, "step": 343630 }, { "epoch": 29.76, "learning_rate": 5.941366948074476e-07, "loss": 0.554, "step": 343640 }, { "epoch": 29.76, "learning_rate": 5.919683127096102e-07, "loss": 0.5991, "step": 343650 }, { "epoch": 29.76, "learning_rate": 5.897999306117728e-07, "loss": 0.5147, "step": 343660 }, { "epoch": 29.76, "learning_rate": 5.876315485139353e-07, "loss": 0.5614, "step": 343670 }, { "epoch": 29.76, "learning_rate": 5.85463166416098e-07, "loss": 0.5911, "step": 343680 }, { "epoch": 29.77, "learning_rate": 5.832947843182607e-07, "loss": 0.5013, "step": 343690 }, { "epoch": 29.77, "learning_rate": 5.811264022204232e-07, "loss": 0.5806, "step": 343700 }, { "epoch": 29.77, "learning_rate": 5.789580201225857e-07, "loss": 0.5298, "step": 343710 }, { "epoch": 29.77, "learning_rate": 5.767896380247484e-07, "loss": 0.5563, "step": 343720 }, { "epoch": 29.77, "learning_rate": 5.74621255926911e-07, "loss": 0.504, "step": 343730 }, { "epoch": 29.77, "learning_rate": 5.724528738290737e-07, "loss": 0.6627, "step": 343740 }, { "epoch": 29.77, "learning_rate": 5.702844917312362e-07, "loss": 0.475, "step": 343750 }, { "epoch": 29.77, "learning_rate": 5.681161096333987e-07, "loss": 0.6657, "step": 343760 }, { "epoch": 29.77, "learning_rate": 5.659477275355614e-07, "loss": 0.5525, "step": 343770 }, { "epoch": 29.77, "learning_rate": 5.63779345437724e-07, "loss": 0.5639, "step": 343780 }, { "epoch": 29.77, "learning_rate": 5.616109633398867e-07, "loss": 0.4721, "step": 343790 }, { "epoch": 29.78, "learning_rate": 5.594425812420492e-07, "loss": 0.6643, "step": 343800 }, { "epoch": 29.78, "learning_rate": 5.572741991442118e-07, "loss": 0.5575, "step": 343810 }, { "epoch": 29.78, "learning_rate": 5.551058170463744e-07, "loss": 0.5982, "step": 343820 }, { "epoch": 29.78, "learning_rate": 5.52937434948537e-07, "loss": 0.5574, "step": 343830 }, { "epoch": 29.78, "learning_rate": 5.507690528506996e-07, "loss": 0.4625, "step": 343840 }, { "epoch": 29.78, "learning_rate": 5.486006707528623e-07, "loss": 0.6563, "step": 343850 }, { "epoch": 29.78, "learning_rate": 5.464322886550248e-07, "loss": 0.5554, "step": 343860 }, { "epoch": 29.78, "learning_rate": 5.442639065571874e-07, "loss": 0.5308, "step": 343870 }, { "epoch": 29.78, "learning_rate": 5.4209552445935e-07, "loss": 0.5226, "step": 343880 }, { "epoch": 29.78, "learning_rate": 5.399271423615126e-07, "loss": 0.7669, "step": 343890 }, { "epoch": 29.78, "learning_rate": 5.377587602636753e-07, "loss": 0.5546, "step": 343900 }, { "epoch": 29.78, "learning_rate": 5.355903781658378e-07, "loss": 0.544, "step": 343910 }, { "epoch": 29.79, "learning_rate": 5.334219960680003e-07, "loss": 0.5071, "step": 343920 }, { "epoch": 29.79, "learning_rate": 5.31253613970163e-07, "loss": 0.5249, "step": 343930 }, { "epoch": 29.79, "learning_rate": 5.290852318723256e-07, "loss": 0.539, "step": 343940 }, { "epoch": 29.79, "learning_rate": 5.269168497744883e-07, "loss": 0.5777, "step": 343950 }, { "epoch": 29.79, "learning_rate": 5.247484676766508e-07, "loss": 0.627, "step": 343960 }, { "epoch": 29.79, "learning_rate": 5.225800855788133e-07, "loss": 0.5678, "step": 343970 }, { "epoch": 29.79, "learning_rate": 5.20411703480976e-07, "loss": 0.5779, "step": 343980 }, { "epoch": 29.79, "learning_rate": 5.182433213831386e-07, "loss": 0.4651, "step": 343990 }, { "epoch": 29.79, "learning_rate": 5.160749392853012e-07, "loss": 0.6139, "step": 344000 }, { "epoch": 29.79, "learning_rate": 5.139065571874639e-07, "loss": 0.5365, "step": 344010 }, { "epoch": 29.79, "learning_rate": 5.117381750896264e-07, "loss": 0.5361, "step": 344020 }, { "epoch": 29.8, "learning_rate": 5.09569792991789e-07, "loss": 0.6065, "step": 344030 }, { "epoch": 29.8, "learning_rate": 5.074014108939516e-07, "loss": 0.5191, "step": 344040 }, { "epoch": 29.8, "learning_rate": 5.052330287961142e-07, "loss": 0.5617, "step": 344050 }, { "epoch": 29.8, "learning_rate": 5.030646466982769e-07, "loss": 0.5662, "step": 344060 }, { "epoch": 29.8, "learning_rate": 5.008962646004394e-07, "loss": 0.5129, "step": 344070 }, { "epoch": 29.8, "learning_rate": 4.987278825026019e-07, "loss": 0.5178, "step": 344080 }, { "epoch": 29.8, "learning_rate": 4.965595004047646e-07, "loss": 0.669, "step": 344090 }, { "epoch": 29.8, "learning_rate": 4.943911183069272e-07, "loss": 0.7118, "step": 344100 }, { "epoch": 29.8, "learning_rate": 4.922227362090899e-07, "loss": 0.6549, "step": 344110 }, { "epoch": 29.8, "learning_rate": 4.900543541112524e-07, "loss": 0.5304, "step": 344120 }, { "epoch": 29.8, "learning_rate": 4.878859720134149e-07, "loss": 0.5975, "step": 344130 }, { "epoch": 29.8, "learning_rate": 4.857175899155776e-07, "loss": 0.4923, "step": 344140 }, { "epoch": 29.81, "learning_rate": 4.835492078177402e-07, "loss": 0.4945, "step": 344150 }, { "epoch": 29.81, "learning_rate": 4.813808257199028e-07, "loss": 0.5644, "step": 344160 }, { "epoch": 29.81, "learning_rate": 4.792124436220655e-07, "loss": 0.5638, "step": 344170 }, { "epoch": 29.81, "learning_rate": 4.77044061524228e-07, "loss": 0.5478, "step": 344180 }, { "epoch": 29.81, "learning_rate": 4.7487567942639064e-07, "loss": 0.5319, "step": 344190 }, { "epoch": 29.81, "learning_rate": 4.727072973285532e-07, "loss": 0.6903, "step": 344200 }, { "epoch": 29.81, "learning_rate": 4.705389152307158e-07, "loss": 0.5561, "step": 344210 }, { "epoch": 29.81, "learning_rate": 4.6837053313287843e-07, "loss": 0.5379, "step": 344220 }, { "epoch": 29.81, "learning_rate": 4.66202151035041e-07, "loss": 0.6015, "step": 344230 }, { "epoch": 29.81, "learning_rate": 4.640337689372036e-07, "loss": 0.4796, "step": 344240 }, { "epoch": 29.81, "learning_rate": 4.618653868393662e-07, "loss": 0.5262, "step": 344250 }, { "epoch": 29.82, "learning_rate": 4.596970047415288e-07, "loss": 0.4842, "step": 344260 }, { "epoch": 29.82, "learning_rate": 4.5752862264369144e-07, "loss": 0.4884, "step": 344270 }, { "epoch": 29.82, "learning_rate": 4.55360240545854e-07, "loss": 0.6057, "step": 344280 }, { "epoch": 29.82, "learning_rate": 4.531918584480166e-07, "loss": 0.7018, "step": 344290 }, { "epoch": 29.82, "learning_rate": 4.5102347635017923e-07, "loss": 0.6176, "step": 344300 }, { "epoch": 29.82, "learning_rate": 4.488550942523418e-07, "loss": 0.4672, "step": 344310 }, { "epoch": 29.82, "learning_rate": 4.466867121545044e-07, "loss": 0.5046, "step": 344320 }, { "epoch": 29.82, "learning_rate": 4.44518330056667e-07, "loss": 0.5887, "step": 344330 }, { "epoch": 29.82, "learning_rate": 4.423499479588296e-07, "loss": 0.5983, "step": 344340 }, { "epoch": 29.82, "learning_rate": 4.4018156586099224e-07, "loss": 0.6167, "step": 344350 }, { "epoch": 29.82, "learning_rate": 4.380131837631548e-07, "loss": 0.5673, "step": 344360 }, { "epoch": 29.82, "learning_rate": 4.358448016653174e-07, "loss": 0.4991, "step": 344370 }, { "epoch": 29.83, "learning_rate": 4.3367641956748003e-07, "loss": 0.563, "step": 344380 }, { "epoch": 29.83, "learning_rate": 4.315080374696426e-07, "loss": 0.5399, "step": 344390 }, { "epoch": 29.83, "learning_rate": 4.293396553718052e-07, "loss": 0.5572, "step": 344400 }, { "epoch": 29.83, "learning_rate": 4.271712732739678e-07, "loss": 0.5628, "step": 344410 }, { "epoch": 29.83, "learning_rate": 4.250028911761304e-07, "loss": 0.6585, "step": 344420 }, { "epoch": 29.83, "learning_rate": 4.2283450907829304e-07, "loss": 0.5886, "step": 344430 }, { "epoch": 29.83, "learning_rate": 4.206661269804556e-07, "loss": 0.6701, "step": 344440 }, { "epoch": 29.83, "learning_rate": 4.184977448826182e-07, "loss": 0.6499, "step": 344450 }, { "epoch": 29.83, "learning_rate": 4.1632936278478083e-07, "loss": 0.5342, "step": 344460 }, { "epoch": 29.83, "learning_rate": 4.141609806869434e-07, "loss": 0.6742, "step": 344470 }, { "epoch": 29.83, "learning_rate": 4.11992598589106e-07, "loss": 0.5792, "step": 344480 }, { "epoch": 29.84, "learning_rate": 4.098242164912686e-07, "loss": 0.6177, "step": 344490 }, { "epoch": 29.84, "learning_rate": 4.076558343934312e-07, "loss": 0.6081, "step": 344500 }, { "epoch": 29.84, "learning_rate": 4.0548745229559383e-07, "loss": 0.6165, "step": 344510 }, { "epoch": 29.84, "learning_rate": 4.033190701977564e-07, "loss": 0.4568, "step": 344520 }, { "epoch": 29.84, "learning_rate": 4.01150688099919e-07, "loss": 0.5105, "step": 344530 }, { "epoch": 29.84, "learning_rate": 3.9898230600208163e-07, "loss": 0.5266, "step": 344540 }, { "epoch": 29.84, "learning_rate": 3.968139239042442e-07, "loss": 0.6058, "step": 344550 }, { "epoch": 29.84, "learning_rate": 3.946455418064068e-07, "loss": 0.5977, "step": 344560 }, { "epoch": 29.84, "learning_rate": 3.924771597085694e-07, "loss": 0.5597, "step": 344570 }, { "epoch": 29.84, "learning_rate": 3.90308777610732e-07, "loss": 0.6227, "step": 344580 }, { "epoch": 29.84, "learning_rate": 3.8814039551289463e-07, "loss": 0.6392, "step": 344590 }, { "epoch": 29.84, "learning_rate": 3.859720134150572e-07, "loss": 0.5902, "step": 344600 }, { "epoch": 29.85, "learning_rate": 3.838036313172198e-07, "loss": 0.6154, "step": 344610 }, { "epoch": 29.85, "learning_rate": 3.816352492193824e-07, "loss": 0.6276, "step": 344620 }, { "epoch": 29.85, "learning_rate": 3.79466867121545e-07, "loss": 0.5339, "step": 344630 }, { "epoch": 29.85, "learning_rate": 3.7729848502370764e-07, "loss": 0.5279, "step": 344640 }, { "epoch": 29.85, "learning_rate": 3.751301029258702e-07, "loss": 0.58, "step": 344650 }, { "epoch": 29.85, "learning_rate": 3.729617208280328e-07, "loss": 0.5937, "step": 344660 }, { "epoch": 29.85, "learning_rate": 3.7079333873019543e-07, "loss": 0.6337, "step": 344670 }, { "epoch": 29.85, "learning_rate": 3.68624956632358e-07, "loss": 0.5616, "step": 344680 }, { "epoch": 29.85, "learning_rate": 3.664565745345206e-07, "loss": 0.5585, "step": 344690 }, { "epoch": 29.85, "learning_rate": 3.642881924366832e-07, "loss": 0.6039, "step": 344700 }, { "epoch": 29.85, "learning_rate": 3.621198103388458e-07, "loss": 0.7151, "step": 344710 }, { "epoch": 29.85, "learning_rate": 3.5995142824100844e-07, "loss": 0.4749, "step": 344720 }, { "epoch": 29.86, "learning_rate": 3.57783046143171e-07, "loss": 0.4961, "step": 344730 }, { "epoch": 29.86, "learning_rate": 3.556146640453336e-07, "loss": 0.5157, "step": 344740 }, { "epoch": 29.86, "learning_rate": 3.5344628194749623e-07, "loss": 0.5493, "step": 344750 }, { "epoch": 29.86, "learning_rate": 3.512778998496588e-07, "loss": 0.6143, "step": 344760 }, { "epoch": 29.86, "learning_rate": 3.491095177518214e-07, "loss": 0.5657, "step": 344770 }, { "epoch": 29.86, "learning_rate": 3.46941135653984e-07, "loss": 0.5365, "step": 344780 }, { "epoch": 29.86, "learning_rate": 3.447727535561466e-07, "loss": 0.6504, "step": 344790 }, { "epoch": 29.86, "learning_rate": 3.4260437145830924e-07, "loss": 0.5668, "step": 344800 }, { "epoch": 29.86, "learning_rate": 3.404359893604718e-07, "loss": 0.7435, "step": 344810 }, { "epoch": 29.86, "learning_rate": 3.382676072626344e-07, "loss": 0.509, "step": 344820 }, { "epoch": 29.86, "learning_rate": 3.3609922516479703e-07, "loss": 0.5292, "step": 344830 }, { "epoch": 29.87, "learning_rate": 3.339308430669596e-07, "loss": 0.4301, "step": 344840 }, { "epoch": 29.87, "learning_rate": 3.317624609691222e-07, "loss": 0.5751, "step": 344850 }, { "epoch": 29.87, "learning_rate": 3.295940788712848e-07, "loss": 0.5427, "step": 344860 }, { "epoch": 29.87, "learning_rate": 3.274256967734474e-07, "loss": 0.4509, "step": 344870 }, { "epoch": 29.87, "learning_rate": 3.2525731467561004e-07, "loss": 0.6543, "step": 344880 }, { "epoch": 29.87, "learning_rate": 3.230889325777726e-07, "loss": 0.6065, "step": 344890 }, { "epoch": 29.87, "learning_rate": 3.209205504799352e-07, "loss": 0.6033, "step": 344900 }, { "epoch": 29.87, "learning_rate": 3.1875216838209783e-07, "loss": 0.4923, "step": 344910 }, { "epoch": 29.87, "learning_rate": 3.165837862842604e-07, "loss": 0.5993, "step": 344920 }, { "epoch": 29.87, "learning_rate": 3.14415404186423e-07, "loss": 0.5476, "step": 344930 }, { "epoch": 29.87, "learning_rate": 3.122470220885856e-07, "loss": 0.6291, "step": 344940 }, { "epoch": 29.87, "learning_rate": 3.100786399907482e-07, "loss": 0.5459, "step": 344950 }, { "epoch": 29.88, "learning_rate": 3.0791025789291083e-07, "loss": 0.5383, "step": 344960 }, { "epoch": 29.88, "learning_rate": 3.057418757950734e-07, "loss": 0.542, "step": 344970 }, { "epoch": 29.88, "learning_rate": 3.03573493697236e-07, "loss": 0.6358, "step": 344980 }, { "epoch": 29.88, "learning_rate": 3.0140511159939863e-07, "loss": 0.4991, "step": 344990 }, { "epoch": 29.88, "learning_rate": 2.992367295015612e-07, "loss": 0.5109, "step": 345000 }, { "epoch": 29.88, "learning_rate": 2.970683474037238e-07, "loss": 0.4941, "step": 345010 }, { "epoch": 29.88, "learning_rate": 2.948999653058864e-07, "loss": 0.5727, "step": 345020 }, { "epoch": 29.88, "learning_rate": 2.92731583208049e-07, "loss": 0.5528, "step": 345030 }, { "epoch": 29.88, "learning_rate": 2.905632011102116e-07, "loss": 0.6347, "step": 345040 }, { "epoch": 29.88, "learning_rate": 2.883948190123742e-07, "loss": 0.5543, "step": 345050 }, { "epoch": 29.88, "learning_rate": 2.8622643691453685e-07, "loss": 0.5974, "step": 345060 }, { "epoch": 29.89, "learning_rate": 2.8405805481669937e-07, "loss": 0.6036, "step": 345070 }, { "epoch": 29.89, "learning_rate": 2.81889672718862e-07, "loss": 0.5712, "step": 345080 }, { "epoch": 29.89, "learning_rate": 2.797212906210246e-07, "loss": 0.6624, "step": 345090 }, { "epoch": 29.89, "learning_rate": 2.775529085231872e-07, "loss": 0.6647, "step": 345100 }, { "epoch": 29.89, "learning_rate": 2.753845264253498e-07, "loss": 0.6722, "step": 345110 }, { "epoch": 29.89, "learning_rate": 2.732161443275124e-07, "loss": 0.5603, "step": 345120 }, { "epoch": 29.89, "learning_rate": 2.71047762229675e-07, "loss": 0.5402, "step": 345130 }, { "epoch": 29.89, "learning_rate": 2.6887938013183764e-07, "loss": 0.6, "step": 345140 }, { "epoch": 29.89, "learning_rate": 2.6671099803400017e-07, "loss": 0.6136, "step": 345150 }, { "epoch": 29.89, "learning_rate": 2.645426159361628e-07, "loss": 0.4928, "step": 345160 }, { "epoch": 29.89, "learning_rate": 2.623742338383254e-07, "loss": 0.5653, "step": 345170 }, { "epoch": 29.89, "learning_rate": 2.60205851740488e-07, "loss": 0.6497, "step": 345180 }, { "epoch": 29.9, "learning_rate": 2.580374696426506e-07, "loss": 0.6052, "step": 345190 }, { "epoch": 29.9, "learning_rate": 2.558690875448132e-07, "loss": 0.5499, "step": 345200 }, { "epoch": 29.9, "learning_rate": 2.537007054469758e-07, "loss": 0.5493, "step": 345210 }, { "epoch": 29.9, "learning_rate": 2.5153232334913844e-07, "loss": 0.5216, "step": 345220 }, { "epoch": 29.9, "learning_rate": 2.4936394125130097e-07, "loss": 0.5635, "step": 345230 }, { "epoch": 29.9, "learning_rate": 2.471955591534636e-07, "loss": 0.5632, "step": 345240 }, { "epoch": 29.9, "learning_rate": 2.450271770556262e-07, "loss": 0.752, "step": 345250 }, { "epoch": 29.9, "learning_rate": 2.428587949577888e-07, "loss": 0.5451, "step": 345260 }, { "epoch": 29.9, "learning_rate": 2.406904128599514e-07, "loss": 0.6027, "step": 345270 }, { "epoch": 29.9, "learning_rate": 2.38522030762114e-07, "loss": 0.5277, "step": 345280 }, { "epoch": 29.9, "learning_rate": 2.363536486642766e-07, "loss": 0.5933, "step": 345290 }, { "epoch": 29.91, "learning_rate": 2.3418526656643922e-07, "loss": 0.5731, "step": 345300 }, { "epoch": 29.91, "learning_rate": 2.320168844686018e-07, "loss": 0.6901, "step": 345310 }, { "epoch": 29.91, "learning_rate": 2.298485023707644e-07, "loss": 0.5022, "step": 345320 }, { "epoch": 29.91, "learning_rate": 2.27680120272927e-07, "loss": 0.5735, "step": 345330 }, { "epoch": 29.91, "learning_rate": 2.2551173817508962e-07, "loss": 0.4089, "step": 345340 }, { "epoch": 29.91, "learning_rate": 2.233433560772522e-07, "loss": 0.546, "step": 345350 }, { "epoch": 29.91, "learning_rate": 2.211749739794148e-07, "loss": 0.5295, "step": 345360 }, { "epoch": 29.91, "learning_rate": 2.190065918815774e-07, "loss": 0.5487, "step": 345370 }, { "epoch": 29.91, "learning_rate": 2.1683820978374001e-07, "loss": 0.5407, "step": 345380 }, { "epoch": 29.91, "learning_rate": 2.146698276859026e-07, "loss": 0.7313, "step": 345390 }, { "epoch": 29.91, "learning_rate": 2.125014455880652e-07, "loss": 0.5781, "step": 345400 }, { "epoch": 29.91, "learning_rate": 2.103330634902278e-07, "loss": 0.6104, "step": 345410 }, { "epoch": 29.92, "learning_rate": 2.0816468139239041e-07, "loss": 0.671, "step": 345420 }, { "epoch": 29.92, "learning_rate": 2.05996299294553e-07, "loss": 0.5651, "step": 345430 }, { "epoch": 29.92, "learning_rate": 2.038279171967156e-07, "loss": 0.5858, "step": 345440 }, { "epoch": 29.92, "learning_rate": 2.016595350988782e-07, "loss": 0.6026, "step": 345450 }, { "epoch": 29.92, "learning_rate": 1.9949115300104081e-07, "loss": 0.6094, "step": 345460 }, { "epoch": 29.92, "learning_rate": 1.973227709032034e-07, "loss": 0.5444, "step": 345470 }, { "epoch": 29.92, "learning_rate": 1.95154388805366e-07, "loss": 0.5957, "step": 345480 }, { "epoch": 29.92, "learning_rate": 1.929860067075286e-07, "loss": 0.458, "step": 345490 }, { "epoch": 29.92, "learning_rate": 1.908176246096912e-07, "loss": 0.5107, "step": 345500 }, { "epoch": 29.92, "learning_rate": 1.8864924251185382e-07, "loss": 0.5968, "step": 345510 }, { "epoch": 29.92, "learning_rate": 1.864808604140164e-07, "loss": 0.5237, "step": 345520 }, { "epoch": 29.93, "learning_rate": 1.84312478316179e-07, "loss": 0.5213, "step": 345530 }, { "epoch": 29.93, "learning_rate": 1.821440962183416e-07, "loss": 0.442, "step": 345540 }, { "epoch": 29.93, "learning_rate": 1.7997571412050422e-07, "loss": 0.6596, "step": 345550 }, { "epoch": 29.93, "learning_rate": 1.778073320226668e-07, "loss": 0.5527, "step": 345560 }, { "epoch": 29.93, "learning_rate": 1.756389499248294e-07, "loss": 0.5063, "step": 345570 }, { "epoch": 29.93, "learning_rate": 1.73470567826992e-07, "loss": 0.573, "step": 345580 }, { "epoch": 29.93, "learning_rate": 1.7130218572915462e-07, "loss": 0.6856, "step": 345590 }, { "epoch": 29.93, "learning_rate": 1.691338036313172e-07, "loss": 0.5283, "step": 345600 }, { "epoch": 29.93, "learning_rate": 1.669654215334798e-07, "loss": 0.5366, "step": 345610 }, { "epoch": 29.93, "learning_rate": 1.647970394356424e-07, "loss": 0.6208, "step": 345620 }, { "epoch": 29.93, "learning_rate": 1.6262865733780502e-07, "loss": 0.5175, "step": 345630 }, { "epoch": 29.93, "learning_rate": 1.604602752399676e-07, "loss": 0.6103, "step": 345640 }, { "epoch": 29.94, "learning_rate": 1.582918931421302e-07, "loss": 0.5526, "step": 345650 }, { "epoch": 29.94, "learning_rate": 1.561235110442928e-07, "loss": 0.4603, "step": 345660 }, { "epoch": 29.94, "learning_rate": 1.5395512894645542e-07, "loss": 0.6333, "step": 345670 }, { "epoch": 29.94, "learning_rate": 1.51786746848618e-07, "loss": 0.5257, "step": 345680 }, { "epoch": 29.94, "learning_rate": 1.496183647507806e-07, "loss": 0.6122, "step": 345690 }, { "epoch": 29.94, "learning_rate": 1.474499826529432e-07, "loss": 0.5655, "step": 345700 }, { "epoch": 29.94, "learning_rate": 1.452816005551058e-07, "loss": 0.6394, "step": 345710 }, { "epoch": 29.94, "learning_rate": 1.4311321845726842e-07, "loss": 0.6034, "step": 345720 }, { "epoch": 29.94, "learning_rate": 1.40944836359431e-07, "loss": 0.5499, "step": 345730 }, { "epoch": 29.94, "learning_rate": 1.387764542615936e-07, "loss": 0.5258, "step": 345740 }, { "epoch": 29.94, "learning_rate": 1.366080721637562e-07, "loss": 0.583, "step": 345750 }, { "epoch": 29.95, "learning_rate": 1.3443969006591882e-07, "loss": 0.5937, "step": 345760 }, { "epoch": 29.95, "learning_rate": 1.322713079680814e-07, "loss": 0.4447, "step": 345770 }, { "epoch": 29.95, "learning_rate": 1.30102925870244e-07, "loss": 0.4448, "step": 345780 }, { "epoch": 29.95, "learning_rate": 1.279345437724066e-07, "loss": 0.5236, "step": 345790 }, { "epoch": 29.95, "learning_rate": 1.2576616167456922e-07, "loss": 0.5229, "step": 345800 }, { "epoch": 29.95, "learning_rate": 1.235977795767318e-07, "loss": 0.5324, "step": 345810 }, { "epoch": 29.95, "learning_rate": 1.214293974788944e-07, "loss": 0.5426, "step": 345820 }, { "epoch": 29.95, "learning_rate": 1.19261015381057e-07, "loss": 0.6234, "step": 345830 }, { "epoch": 29.95, "learning_rate": 1.1709263328321961e-07, "loss": 0.5371, "step": 345840 }, { "epoch": 29.95, "learning_rate": 1.149242511853822e-07, "loss": 0.5522, "step": 345850 }, { "epoch": 29.95, "learning_rate": 1.1275586908754481e-07, "loss": 0.5815, "step": 345860 }, { "epoch": 29.95, "learning_rate": 1.105874869897074e-07, "loss": 0.5293, "step": 345870 }, { "epoch": 29.96, "learning_rate": 1.0841910489187001e-07, "loss": 0.5889, "step": 345880 }, { "epoch": 29.96, "learning_rate": 1.062507227940326e-07, "loss": 0.6476, "step": 345890 }, { "epoch": 29.96, "learning_rate": 1.0408234069619521e-07, "loss": 0.6008, "step": 345900 }, { "epoch": 29.96, "learning_rate": 1.019139585983578e-07, "loss": 0.467, "step": 345910 }, { "epoch": 29.96, "learning_rate": 9.974557650052041e-08, "loss": 0.5442, "step": 345920 }, { "epoch": 29.96, "learning_rate": 9.7577194402683e-08, "loss": 0.7603, "step": 345930 }, { "epoch": 29.96, "learning_rate": 9.54088123048456e-08, "loss": 0.606, "step": 345940 }, { "epoch": 29.96, "learning_rate": 9.32404302070082e-08, "loss": 0.532, "step": 345950 }, { "epoch": 29.96, "learning_rate": 9.10720481091708e-08, "loss": 0.5232, "step": 345960 }, { "epoch": 29.96, "learning_rate": 8.89036660113334e-08, "loss": 0.4644, "step": 345970 }, { "epoch": 29.96, "learning_rate": 8.6735283913496e-08, "loss": 0.5389, "step": 345980 }, { "epoch": 29.96, "learning_rate": 8.45669018156586e-08, "loss": 0.5923, "step": 345990 }, { "epoch": 29.97, "learning_rate": 8.23985197178212e-08, "loss": 0.5037, "step": 346000 }, { "epoch": 29.97, "learning_rate": 8.02301376199838e-08, "loss": 0.528, "step": 346010 }, { "epoch": 29.97, "learning_rate": 7.80617555221464e-08, "loss": 0.637, "step": 346020 }, { "epoch": 29.97, "learning_rate": 7.5893373424309e-08, "loss": 0.458, "step": 346030 }, { "epoch": 29.97, "learning_rate": 7.37249913264716e-08, "loss": 0.5238, "step": 346040 }, { "epoch": 29.97, "learning_rate": 7.155660922863421e-08, "loss": 0.5447, "step": 346050 }, { "epoch": 29.97, "learning_rate": 6.93882271307968e-08, "loss": 0.5974, "step": 346060 }, { "epoch": 29.97, "learning_rate": 6.721984503295941e-08, "loss": 0.5982, "step": 346070 }, { "epoch": 29.97, "learning_rate": 6.5051462935122e-08, "loss": 0.6233, "step": 346080 }, { "epoch": 29.97, "learning_rate": 6.288308083728461e-08, "loss": 0.6123, "step": 346090 }, { "epoch": 29.97, "learning_rate": 6.07146987394472e-08, "loss": 0.4664, "step": 346100 }, { "epoch": 29.98, "learning_rate": 5.8546316641609804e-08, "loss": 0.5995, "step": 346110 }, { "epoch": 29.98, "learning_rate": 5.6377934543772404e-08, "loss": 0.4962, "step": 346120 }, { "epoch": 29.98, "learning_rate": 5.4209552445935004e-08, "loss": 0.5537, "step": 346130 }, { "epoch": 29.98, "learning_rate": 5.2041170348097604e-08, "loss": 0.5023, "step": 346140 }, { "epoch": 29.98, "learning_rate": 4.9872788250260203e-08, "loss": 0.6275, "step": 346150 }, { "epoch": 29.98, "learning_rate": 4.77044061524228e-08, "loss": 0.4874, "step": 346160 }, { "epoch": 29.98, "learning_rate": 4.55360240545854e-08, "loss": 0.6577, "step": 346170 }, { "epoch": 29.98, "learning_rate": 4.3367641956748e-08, "loss": 0.5318, "step": 346180 }, { "epoch": 29.98, "learning_rate": 4.11992598589106e-08, "loss": 0.5262, "step": 346190 }, { "epoch": 29.98, "learning_rate": 3.90308777610732e-08, "loss": 0.5261, "step": 346200 }, { "epoch": 29.98, "learning_rate": 3.68624956632358e-08, "loss": 0.6025, "step": 346210 }, { "epoch": 29.98, "learning_rate": 3.46941135653984e-08, "loss": 0.5722, "step": 346220 }, { "epoch": 29.99, "learning_rate": 3.2525731467561e-08, "loss": 0.6602, "step": 346230 }, { "epoch": 29.99, "learning_rate": 3.03573493697236e-08, "loss": 0.4814, "step": 346240 }, { "epoch": 29.99, "learning_rate": 2.8188967271886202e-08, "loss": 0.696, "step": 346250 }, { "epoch": 29.99, "learning_rate": 2.6020585174048802e-08, "loss": 0.6566, "step": 346260 }, { "epoch": 29.99, "learning_rate": 2.38522030762114e-08, "loss": 0.6201, "step": 346270 }, { "epoch": 29.99, "learning_rate": 2.1683820978374e-08, "loss": 0.5071, "step": 346280 }, { "epoch": 29.99, "learning_rate": 1.95154388805366e-08, "loss": 0.5423, "step": 346290 }, { "epoch": 29.99, "learning_rate": 1.73470567826992e-08, "loss": 0.4506, "step": 346300 }, { "epoch": 29.99, "learning_rate": 1.51786746848618e-08, "loss": 0.4899, "step": 346310 }, { "epoch": 29.99, "learning_rate": 1.3010292587024401e-08, "loss": 0.516, "step": 346320 }, { "epoch": 29.99, "learning_rate": 1.0841910489187e-08, "loss": 0.5763, "step": 346330 }, { "epoch": 30.0, "learning_rate": 8.6735283913496e-09, "loss": 0.5081, "step": 346340 }, { "epoch": 30.0, "learning_rate": 6.5051462935122004e-09, "loss": 0.4469, "step": 346350 }, { "epoch": 30.0, "learning_rate": 4.3367641956748e-09, "loss": 0.44, "step": 346360 }, { "epoch": 30.0, "learning_rate": 2.1683820978374e-09, "loss": 0.637, "step": 346370 }, { "epoch": 30.0, "learning_rate": 0.0, "loss": 0.6531, "step": 346380 }, { "epoch": 30.0, "eval_cer": 0.5330904090444878, "eval_em": 0.06498194945848375, "eval_f1": 0.06498194945848375, "eval_loss": 1.590509057044983, "eval_runtime": 252.9879, "eval_samples_per_second": 7.664, "eval_steps_per_second": 0.961, "eval_wer": 0.9350180505415162, "step": 346380 } ], "logging_steps": 10, "max_steps": 346380, "num_train_epochs": 30, "save_steps": 500, "total_flos": 2.712419467941768e+18, "trial_name": null, "trial_params": null }