{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9838646202282566, "eval_steps": 500, "global_step": 80000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 8.19672131147541e-08, "loss": 42.728, "step": 10 }, { "epoch": 0.0, "learning_rate": 1.639344262295082e-07, "loss": 42.5633, "step": 20 }, { "epoch": 0.0, "learning_rate": 2.459016393442623e-07, "loss": 42.3511, "step": 30 }, { "epoch": 0.0, "learning_rate": 3.278688524590164e-07, "loss": 42.1928, "step": 40 }, { "epoch": 0.0, "learning_rate": 4.0983606557377047e-07, "loss": 42.3546, "step": 50 }, { "epoch": 0.0, "learning_rate": 4.918032786885246e-07, "loss": 42.5585, "step": 60 }, { "epoch": 0.0, "learning_rate": 5.737704918032787e-07, "loss": 42.6458, "step": 70 }, { "epoch": 0.0, "learning_rate": 6.557377049180328e-07, "loss": 43.0448, "step": 80 }, { "epoch": 0.0, "learning_rate": 7.377049180327869e-07, "loss": 42.9794, "step": 90 }, { "epoch": 0.0, "learning_rate": 8.196721311475409e-07, "loss": 42.4481, "step": 100 }, { "epoch": 0.0, "learning_rate": 9.016393442622952e-07, "loss": 41.6212, "step": 110 }, { "epoch": 0.0, "learning_rate": 9.836065573770493e-07, "loss": 41.3723, "step": 120 }, { "epoch": 0.0, "learning_rate": 1.0655737704918034e-06, "loss": 42.5345, "step": 130 }, { "epoch": 0.0, "learning_rate": 1.1475409836065575e-06, "loss": 41.8501, "step": 140 }, { "epoch": 0.0, "learning_rate": 1.2295081967213116e-06, "loss": 42.2224, "step": 150 }, { "epoch": 0.0, "learning_rate": 1.3114754098360657e-06, "loss": 41.6194, "step": 160 }, { "epoch": 0.0, "learning_rate": 1.3934426229508196e-06, "loss": 41.5141, "step": 170 }, { "epoch": 0.0, "learning_rate": 1.4754098360655739e-06, "loss": 41.2204, "step": 180 }, { "epoch": 0.0, "learning_rate": 1.557377049180328e-06, "loss": 41.0626, "step": 190 }, { "epoch": 0.0, "learning_rate": 1.6393442622950819e-06, "loss": 41.2386, "step": 200 }, { "epoch": 0.0, "learning_rate": 1.7213114754098362e-06, "loss": 40.6984, "step": 210 }, { "epoch": 0.0, "learning_rate": 1.8032786885245903e-06, "loss": 41.5098, "step": 220 }, { "epoch": 0.0, "learning_rate": 1.8852459016393442e-06, "loss": 40.804, "step": 230 }, { "epoch": 0.0, "learning_rate": 1.9672131147540985e-06, "loss": 41.693, "step": 240 }, { "epoch": 0.0, "learning_rate": 2.0491803278688526e-06, "loss": 40.4265, "step": 250 }, { "epoch": 0.0, "learning_rate": 2.1311475409836067e-06, "loss": 39.5724, "step": 260 }, { "epoch": 0.0, "learning_rate": 2.213114754098361e-06, "loss": 40.7129, "step": 270 }, { "epoch": 0.0, "learning_rate": 2.295081967213115e-06, "loss": 40.6633, "step": 280 }, { "epoch": 0.0, "learning_rate": 2.377049180327869e-06, "loss": 40.1102, "step": 290 }, { "epoch": 0.0, "learning_rate": 2.459016393442623e-06, "loss": 39.7031, "step": 300 }, { "epoch": 0.0, "learning_rate": 2.5409836065573773e-06, "loss": 39.5397, "step": 310 }, { "epoch": 0.0, "learning_rate": 2.6229508196721314e-06, "loss": 38.851, "step": 320 }, { "epoch": 0.0, "learning_rate": 2.704918032786886e-06, "loss": 39.2468, "step": 330 }, { "epoch": 0.0, "learning_rate": 2.786885245901639e-06, "loss": 39.2133, "step": 340 }, { "epoch": 0.0, "learning_rate": 2.8688524590163937e-06, "loss": 39.0269, "step": 350 }, { "epoch": 0.0, "learning_rate": 2.9508196721311478e-06, "loss": 38.0372, "step": 360 }, { "epoch": 0.0, "learning_rate": 3.032786885245902e-06, "loss": 38.2282, "step": 370 }, { "epoch": 0.0, "learning_rate": 3.114754098360656e-06, "loss": 37.5624, "step": 380 }, { "epoch": 0.0, "learning_rate": 3.1967213114754105e-06, "loss": 37.4344, "step": 390 }, { "epoch": 0.0, "learning_rate": 3.2786885245901638e-06, "loss": 36.9773, "step": 400 }, { "epoch": 0.01, "learning_rate": 3.3606557377049183e-06, "loss": 35.7986, "step": 410 }, { "epoch": 0.01, "learning_rate": 3.4426229508196724e-06, "loss": 35.2651, "step": 420 }, { "epoch": 0.01, "learning_rate": 3.5245901639344265e-06, "loss": 33.1556, "step": 430 }, { "epoch": 0.01, "learning_rate": 3.6065573770491806e-06, "loss": 31.9066, "step": 440 }, { "epoch": 0.01, "learning_rate": 3.6885245901639347e-06, "loss": 31.0781, "step": 450 }, { "epoch": 0.01, "learning_rate": 3.7704918032786884e-06, "loss": 30.5787, "step": 460 }, { "epoch": 0.01, "learning_rate": 3.852459016393443e-06, "loss": 30.7174, "step": 470 }, { "epoch": 0.01, "learning_rate": 3.934426229508197e-06, "loss": 30.2692, "step": 480 }, { "epoch": 0.01, "learning_rate": 4.016393442622951e-06, "loss": 29.993, "step": 490 }, { "epoch": 0.01, "learning_rate": 4.098360655737705e-06, "loss": 29.8367, "step": 500 }, { "epoch": 0.01, "learning_rate": 4.180327868852459e-06, "loss": 30.3889, "step": 510 }, { "epoch": 0.01, "learning_rate": 4.2622950819672135e-06, "loss": 30.0812, "step": 520 }, { "epoch": 0.01, "learning_rate": 4.3442622950819676e-06, "loss": 29.7899, "step": 530 }, { "epoch": 0.01, "learning_rate": 4.426229508196722e-06, "loss": 29.5817, "step": 540 }, { "epoch": 0.01, "learning_rate": 4.508196721311476e-06, "loss": 29.1851, "step": 550 }, { "epoch": 0.01, "learning_rate": 4.59016393442623e-06, "loss": 29.9939, "step": 560 }, { "epoch": 0.01, "learning_rate": 4.672131147540984e-06, "loss": 29.7766, "step": 570 }, { "epoch": 0.01, "learning_rate": 4.754098360655738e-06, "loss": 29.0335, "step": 580 }, { "epoch": 0.01, "learning_rate": 4.836065573770492e-06, "loss": 29.8774, "step": 590 }, { "epoch": 0.01, "learning_rate": 4.918032786885246e-06, "loss": 29.0639, "step": 600 }, { "epoch": 0.01, "learning_rate": 5e-06, "loss": 29.0456, "step": 610 }, { "epoch": 0.01, "learning_rate": 5.0819672131147545e-06, "loss": 29.2741, "step": 620 }, { "epoch": 0.01, "learning_rate": 5.163934426229509e-06, "loss": 29.4171, "step": 630 }, { "epoch": 0.01, "learning_rate": 5.245901639344263e-06, "loss": 29.0033, "step": 640 }, { "epoch": 0.01, "learning_rate": 5.327868852459017e-06, "loss": 28.7732, "step": 650 }, { "epoch": 0.01, "learning_rate": 5.409836065573772e-06, "loss": 28.931, "step": 660 }, { "epoch": 0.01, "learning_rate": 5.491803278688526e-06, "loss": 28.6373, "step": 670 }, { "epoch": 0.01, "learning_rate": 5.573770491803278e-06, "loss": 28.5153, "step": 680 }, { "epoch": 0.01, "learning_rate": 5.655737704918033e-06, "loss": 29.0248, "step": 690 }, { "epoch": 0.01, "learning_rate": 5.737704918032787e-06, "loss": 28.4666, "step": 700 }, { "epoch": 0.01, "learning_rate": 5.8196721311475415e-06, "loss": 28.6413, "step": 710 }, { "epoch": 0.01, "learning_rate": 5.9016393442622956e-06, "loss": 28.6983, "step": 720 }, { "epoch": 0.01, "learning_rate": 5.98360655737705e-06, "loss": 28.7854, "step": 730 }, { "epoch": 0.01, "learning_rate": 6.065573770491804e-06, "loss": 28.289, "step": 740 }, { "epoch": 0.01, "learning_rate": 6.147540983606558e-06, "loss": 28.7361, "step": 750 }, { "epoch": 0.01, "learning_rate": 6.229508196721312e-06, "loss": 28.7765, "step": 760 }, { "epoch": 0.01, "learning_rate": 6.311475409836066e-06, "loss": 28.5523, "step": 770 }, { "epoch": 0.01, "learning_rate": 6.393442622950821e-06, "loss": 28.3397, "step": 780 }, { "epoch": 0.01, "learning_rate": 6.475409836065575e-06, "loss": 28.098, "step": 790 }, { "epoch": 0.01, "learning_rate": 6.5573770491803276e-06, "loss": 28.1114, "step": 800 }, { "epoch": 0.01, "learning_rate": 6.6393442622950825e-06, "loss": 28.37, "step": 810 }, { "epoch": 0.01, "learning_rate": 6.721311475409837e-06, "loss": 28.302, "step": 820 }, { "epoch": 0.01, "learning_rate": 6.803278688524591e-06, "loss": 28.5721, "step": 830 }, { "epoch": 0.01, "learning_rate": 6.885245901639345e-06, "loss": 28.4993, "step": 840 }, { "epoch": 0.01, "learning_rate": 6.967213114754099e-06, "loss": 28.3435, "step": 850 }, { "epoch": 0.01, "learning_rate": 7.049180327868853e-06, "loss": 28.2901, "step": 860 }, { "epoch": 0.01, "learning_rate": 7.131147540983607e-06, "loss": 28.5831, "step": 870 }, { "epoch": 0.01, "learning_rate": 7.213114754098361e-06, "loss": 28.583, "step": 880 }, { "epoch": 0.01, "learning_rate": 7.295081967213115e-06, "loss": 28.0047, "step": 890 }, { "epoch": 0.01, "learning_rate": 7.3770491803278695e-06, "loss": 27.9143, "step": 900 }, { "epoch": 0.01, "learning_rate": 7.459016393442624e-06, "loss": 28.1182, "step": 910 }, { "epoch": 0.01, "learning_rate": 7.540983606557377e-06, "loss": 27.8796, "step": 920 }, { "epoch": 0.01, "learning_rate": 7.622950819672132e-06, "loss": 28.3695, "step": 930 }, { "epoch": 0.01, "learning_rate": 7.704918032786886e-06, "loss": 27.9613, "step": 940 }, { "epoch": 0.01, "learning_rate": 7.786885245901639e-06, "loss": 28.2338, "step": 950 }, { "epoch": 0.01, "learning_rate": 7.868852459016394e-06, "loss": 28.1708, "step": 960 }, { "epoch": 0.01, "learning_rate": 7.950819672131147e-06, "loss": 27.8797, "step": 970 }, { "epoch": 0.01, "learning_rate": 8.032786885245902e-06, "loss": 28.1845, "step": 980 }, { "epoch": 0.01, "learning_rate": 8.114754098360657e-06, "loss": 28.2, "step": 990 }, { "epoch": 0.01, "learning_rate": 8.19672131147541e-06, "loss": 27.9973, "step": 1000 }, { "epoch": 0.01, "learning_rate": 8.278688524590165e-06, "loss": 27.8465, "step": 1010 }, { "epoch": 0.01, "learning_rate": 8.360655737704919e-06, "loss": 27.5939, "step": 1020 }, { "epoch": 0.01, "learning_rate": 8.442622950819674e-06, "loss": 28.2279, "step": 1030 }, { "epoch": 0.01, "learning_rate": 8.524590163934427e-06, "loss": 27.9474, "step": 1040 }, { "epoch": 0.01, "learning_rate": 8.60655737704918e-06, "loss": 27.6415, "step": 1050 }, { "epoch": 0.01, "learning_rate": 8.688524590163935e-06, "loss": 27.682, "step": 1060 }, { "epoch": 0.01, "learning_rate": 8.770491803278688e-06, "loss": 28.0427, "step": 1070 }, { "epoch": 0.01, "learning_rate": 8.852459016393443e-06, "loss": 27.8246, "step": 1080 }, { "epoch": 0.01, "learning_rate": 8.934426229508197e-06, "loss": 28.0181, "step": 1090 }, { "epoch": 0.01, "learning_rate": 9.016393442622952e-06, "loss": 27.7555, "step": 1100 }, { "epoch": 0.01, "learning_rate": 9.098360655737707e-06, "loss": 27.6486, "step": 1110 }, { "epoch": 0.01, "learning_rate": 9.18032786885246e-06, "loss": 27.4092, "step": 1120 }, { "epoch": 0.01, "learning_rate": 9.262295081967215e-06, "loss": 27.8648, "step": 1130 }, { "epoch": 0.01, "learning_rate": 9.344262295081968e-06, "loss": 27.8314, "step": 1140 }, { "epoch": 0.01, "learning_rate": 9.426229508196723e-06, "loss": 27.5533, "step": 1150 }, { "epoch": 0.01, "learning_rate": 9.508196721311476e-06, "loss": 27.3973, "step": 1160 }, { "epoch": 0.01, "learning_rate": 9.59016393442623e-06, "loss": 27.5652, "step": 1170 }, { "epoch": 0.01, "learning_rate": 9.672131147540984e-06, "loss": 27.5649, "step": 1180 }, { "epoch": 0.01, "learning_rate": 9.754098360655738e-06, "loss": 27.124, "step": 1190 }, { "epoch": 0.01, "learning_rate": 9.836065573770493e-06, "loss": 27.5819, "step": 1200 }, { "epoch": 0.01, "learning_rate": 9.918032786885246e-06, "loss": 27.8305, "step": 1210 }, { "epoch": 0.02, "learning_rate": 1e-05, "loss": 27.4044, "step": 1220 }, { "epoch": 0.02, "learning_rate": 1.0081967213114754e-05, "loss": 27.0521, "step": 1230 }, { "epoch": 0.02, "learning_rate": 1.0163934426229509e-05, "loss": 27.2803, "step": 1240 }, { "epoch": 0.02, "learning_rate": 1.0245901639344262e-05, "loss": 27.7298, "step": 1250 }, { "epoch": 0.02, "learning_rate": 1.0327868852459017e-05, "loss": 27.7905, "step": 1260 }, { "epoch": 0.02, "learning_rate": 1.040983606557377e-05, "loss": 27.385, "step": 1270 }, { "epoch": 0.02, "learning_rate": 1.0491803278688525e-05, "loss": 26.9803, "step": 1280 }, { "epoch": 0.02, "learning_rate": 1.0573770491803279e-05, "loss": 27.5341, "step": 1290 }, { "epoch": 0.02, "learning_rate": 1.0655737704918034e-05, "loss": 27.0457, "step": 1300 }, { "epoch": 0.02, "learning_rate": 1.0737704918032787e-05, "loss": 26.7992, "step": 1310 }, { "epoch": 0.02, "learning_rate": 1.0819672131147544e-05, "loss": 26.3803, "step": 1320 }, { "epoch": 0.02, "learning_rate": 1.0901639344262295e-05, "loss": 26.9781, "step": 1330 }, { "epoch": 0.02, "learning_rate": 1.0983606557377052e-05, "loss": 26.7096, "step": 1340 }, { "epoch": 0.02, "learning_rate": 1.1065573770491805e-05, "loss": 26.8885, "step": 1350 }, { "epoch": 0.02, "learning_rate": 1.1147540983606557e-05, "loss": 26.9279, "step": 1360 }, { "epoch": 0.02, "learning_rate": 1.1229508196721313e-05, "loss": 26.6643, "step": 1370 }, { "epoch": 0.02, "learning_rate": 1.1311475409836066e-05, "loss": 26.6135, "step": 1380 }, { "epoch": 0.02, "learning_rate": 1.1393442622950821e-05, "loss": 27.1947, "step": 1390 }, { "epoch": 0.02, "learning_rate": 1.1475409836065575e-05, "loss": 26.6972, "step": 1400 }, { "epoch": 0.02, "learning_rate": 1.155737704918033e-05, "loss": 27.043, "step": 1410 }, { "epoch": 0.02, "learning_rate": 1.1639344262295083e-05, "loss": 26.7005, "step": 1420 }, { "epoch": 0.02, "learning_rate": 1.1721311475409838e-05, "loss": 26.4959, "step": 1430 }, { "epoch": 0.02, "learning_rate": 1.1803278688524591e-05, "loss": 27.1831, "step": 1440 }, { "epoch": 0.02, "learning_rate": 1.1885245901639346e-05, "loss": 26.7137, "step": 1450 }, { "epoch": 0.02, "learning_rate": 1.19672131147541e-05, "loss": 26.7359, "step": 1460 }, { "epoch": 0.02, "learning_rate": 1.2049180327868853e-05, "loss": 27.278, "step": 1470 }, { "epoch": 0.02, "learning_rate": 1.2131147540983608e-05, "loss": 27.4079, "step": 1480 }, { "epoch": 0.02, "learning_rate": 1.221311475409836e-05, "loss": 27.2985, "step": 1490 }, { "epoch": 0.02, "learning_rate": 1.2295081967213116e-05, "loss": 26.8436, "step": 1500 }, { "epoch": 0.02, "learning_rate": 1.2377049180327869e-05, "loss": 26.2591, "step": 1510 }, { "epoch": 0.02, "learning_rate": 1.2459016393442624e-05, "loss": 26.7118, "step": 1520 }, { "epoch": 0.02, "learning_rate": 1.2540983606557377e-05, "loss": 26.9512, "step": 1530 }, { "epoch": 0.02, "learning_rate": 1.2622950819672132e-05, "loss": 26.8143, "step": 1540 }, { "epoch": 0.02, "learning_rate": 1.2704918032786885e-05, "loss": 26.6907, "step": 1550 }, { "epoch": 0.02, "learning_rate": 1.2786885245901642e-05, "loss": 27.3007, "step": 1560 }, { "epoch": 0.02, "learning_rate": 1.2868852459016394e-05, "loss": 26.2503, "step": 1570 }, { "epoch": 0.02, "learning_rate": 1.295081967213115e-05, "loss": 26.993, "step": 1580 }, { "epoch": 0.02, "learning_rate": 1.3032786885245904e-05, "loss": 26.5596, "step": 1590 }, { "epoch": 0.02, "learning_rate": 1.3114754098360655e-05, "loss": 26.3575, "step": 1600 }, { "epoch": 0.02, "learning_rate": 1.3196721311475412e-05, "loss": 26.4031, "step": 1610 }, { "epoch": 0.02, "learning_rate": 1.3278688524590165e-05, "loss": 26.4806, "step": 1620 }, { "epoch": 0.02, "learning_rate": 1.336065573770492e-05, "loss": 26.5496, "step": 1630 }, { "epoch": 0.02, "learning_rate": 1.3442622950819673e-05, "loss": 26.387, "step": 1640 }, { "epoch": 0.02, "learning_rate": 1.3524590163934428e-05, "loss": 26.6426, "step": 1650 }, { "epoch": 0.02, "learning_rate": 1.3606557377049181e-05, "loss": 26.7206, "step": 1660 }, { "epoch": 0.02, "learning_rate": 1.3688524590163936e-05, "loss": 26.1517, "step": 1670 }, { "epoch": 0.02, "learning_rate": 1.377049180327869e-05, "loss": 26.1497, "step": 1680 }, { "epoch": 0.02, "learning_rate": 1.3852459016393445e-05, "loss": 26.8465, "step": 1690 }, { "epoch": 0.02, "learning_rate": 1.3934426229508198e-05, "loss": 26.5966, "step": 1700 }, { "epoch": 0.02, "learning_rate": 1.4016393442622951e-05, "loss": 26.3885, "step": 1710 }, { "epoch": 0.02, "learning_rate": 1.4098360655737706e-05, "loss": 26.5465, "step": 1720 }, { "epoch": 0.02, "learning_rate": 1.418032786885246e-05, "loss": 26.3944, "step": 1730 }, { "epoch": 0.02, "learning_rate": 1.4262295081967214e-05, "loss": 26.2267, "step": 1740 }, { "epoch": 0.02, "learning_rate": 1.4344262295081968e-05, "loss": 26.2037, "step": 1750 }, { "epoch": 0.02, "learning_rate": 1.4426229508196722e-05, "loss": 25.7021, "step": 1760 }, { "epoch": 0.02, "learning_rate": 1.4508196721311476e-05, "loss": 26.2906, "step": 1770 }, { "epoch": 0.02, "learning_rate": 1.459016393442623e-05, "loss": 25.7089, "step": 1780 }, { "epoch": 0.02, "learning_rate": 1.4672131147540984e-05, "loss": 25.8394, "step": 1790 }, { "epoch": 0.02, "learning_rate": 1.4754098360655739e-05, "loss": 25.9367, "step": 1800 }, { "epoch": 0.02, "learning_rate": 1.4836065573770492e-05, "loss": 25.9496, "step": 1810 }, { "epoch": 0.02, "learning_rate": 1.4918032786885249e-05, "loss": 26.1963, "step": 1820 }, { "epoch": 0.02, "learning_rate": 1.5000000000000002e-05, "loss": 25.9338, "step": 1830 }, { "epoch": 0.02, "learning_rate": 1.5081967213114754e-05, "loss": 25.6272, "step": 1840 }, { "epoch": 0.02, "learning_rate": 1.516393442622951e-05, "loss": 26.38, "step": 1850 }, { "epoch": 0.02, "learning_rate": 1.5245901639344264e-05, "loss": 26.0585, "step": 1860 }, { "epoch": 0.02, "learning_rate": 1.532786885245902e-05, "loss": 25.7464, "step": 1870 }, { "epoch": 0.02, "learning_rate": 1.5409836065573772e-05, "loss": 26.3311, "step": 1880 }, { "epoch": 0.02, "learning_rate": 1.5491803278688525e-05, "loss": 25.5722, "step": 1890 }, { "epoch": 0.02, "learning_rate": 1.5573770491803278e-05, "loss": 25.4848, "step": 1900 }, { "epoch": 0.02, "learning_rate": 1.5655737704918035e-05, "loss": 25.4869, "step": 1910 }, { "epoch": 0.02, "learning_rate": 1.5737704918032788e-05, "loss": 26.0235, "step": 1920 }, { "epoch": 0.02, "learning_rate": 1.5819672131147545e-05, "loss": 25.4484, "step": 1930 }, { "epoch": 0.02, "learning_rate": 1.5901639344262295e-05, "loss": 25.3224, "step": 1940 }, { "epoch": 0.02, "learning_rate": 1.598360655737705e-05, "loss": 25.8618, "step": 1950 }, { "epoch": 0.02, "learning_rate": 1.6065573770491805e-05, "loss": 25.9926, "step": 1960 }, { "epoch": 0.02, "learning_rate": 1.6147540983606558e-05, "loss": 25.8784, "step": 1970 }, { "epoch": 0.02, "learning_rate": 1.6229508196721314e-05, "loss": 25.5772, "step": 1980 }, { "epoch": 0.02, "learning_rate": 1.6311475409836068e-05, "loss": 25.3414, "step": 1990 }, { "epoch": 0.02, "learning_rate": 1.639344262295082e-05, "loss": 25.7102, "step": 2000 }, { "epoch": 0.02, "learning_rate": 1.6475409836065574e-05, "loss": 25.7019, "step": 2010 }, { "epoch": 0.02, "learning_rate": 1.655737704918033e-05, "loss": 25.6786, "step": 2020 }, { "epoch": 0.02, "learning_rate": 1.6639344262295084e-05, "loss": 25.8914, "step": 2030 }, { "epoch": 0.03, "learning_rate": 1.6721311475409837e-05, "loss": 26.1056, "step": 2040 }, { "epoch": 0.03, "learning_rate": 1.680327868852459e-05, "loss": 25.4805, "step": 2050 }, { "epoch": 0.03, "learning_rate": 1.6885245901639347e-05, "loss": 25.2237, "step": 2060 }, { "epoch": 0.03, "learning_rate": 1.69672131147541e-05, "loss": 25.8045, "step": 2070 }, { "epoch": 0.03, "learning_rate": 1.7049180327868854e-05, "loss": 25.4788, "step": 2080 }, { "epoch": 0.03, "learning_rate": 1.7131147540983607e-05, "loss": 25.4885, "step": 2090 }, { "epoch": 0.03, "learning_rate": 1.721311475409836e-05, "loss": 25.6791, "step": 2100 }, { "epoch": 0.03, "learning_rate": 1.7295081967213117e-05, "loss": 25.178, "step": 2110 }, { "epoch": 0.03, "learning_rate": 1.737704918032787e-05, "loss": 25.2685, "step": 2120 }, { "epoch": 0.03, "learning_rate": 1.7459016393442624e-05, "loss": 25.6115, "step": 2130 }, { "epoch": 0.03, "learning_rate": 1.7540983606557377e-05, "loss": 25.3796, "step": 2140 }, { "epoch": 0.03, "learning_rate": 1.7622950819672133e-05, "loss": 25.2294, "step": 2150 }, { "epoch": 0.03, "learning_rate": 1.7704918032786887e-05, "loss": 25.3355, "step": 2160 }, { "epoch": 0.03, "learning_rate": 1.7786885245901643e-05, "loss": 26.1178, "step": 2170 }, { "epoch": 0.03, "learning_rate": 1.7868852459016393e-05, "loss": 25.3262, "step": 2180 }, { "epoch": 0.03, "learning_rate": 1.795081967213115e-05, "loss": 25.5796, "step": 2190 }, { "epoch": 0.03, "learning_rate": 1.8032786885245903e-05, "loss": 25.4896, "step": 2200 }, { "epoch": 0.03, "learning_rate": 1.8114754098360656e-05, "loss": 25.4967, "step": 2210 }, { "epoch": 0.03, "learning_rate": 1.8196721311475413e-05, "loss": 25.2496, "step": 2220 }, { "epoch": 0.03, "learning_rate": 1.8278688524590166e-05, "loss": 25.3435, "step": 2230 }, { "epoch": 0.03, "learning_rate": 1.836065573770492e-05, "loss": 24.6709, "step": 2240 }, { "epoch": 0.03, "learning_rate": 1.8442622950819673e-05, "loss": 25.1664, "step": 2250 }, { "epoch": 0.03, "learning_rate": 1.852459016393443e-05, "loss": 25.315, "step": 2260 }, { "epoch": 0.03, "learning_rate": 1.8606557377049183e-05, "loss": 24.9705, "step": 2270 }, { "epoch": 0.03, "learning_rate": 1.8688524590163936e-05, "loss": 25.0718, "step": 2280 }, { "epoch": 0.03, "learning_rate": 1.877049180327869e-05, "loss": 25.4942, "step": 2290 }, { "epoch": 0.03, "learning_rate": 1.8852459016393446e-05, "loss": 25.1679, "step": 2300 }, { "epoch": 0.03, "learning_rate": 1.89344262295082e-05, "loss": 25.1532, "step": 2310 }, { "epoch": 0.03, "learning_rate": 1.9016393442622952e-05, "loss": 24.906, "step": 2320 }, { "epoch": 0.03, "learning_rate": 1.9098360655737706e-05, "loss": 25.0198, "step": 2330 }, { "epoch": 0.03, "learning_rate": 1.918032786885246e-05, "loss": 25.4099, "step": 2340 }, { "epoch": 0.03, "learning_rate": 1.9262295081967216e-05, "loss": 25.2965, "step": 2350 }, { "epoch": 0.03, "learning_rate": 1.934426229508197e-05, "loss": 25.2375, "step": 2360 }, { "epoch": 0.03, "learning_rate": 1.9426229508196722e-05, "loss": 25.2278, "step": 2370 }, { "epoch": 0.03, "learning_rate": 1.9508196721311475e-05, "loss": 24.8347, "step": 2380 }, { "epoch": 0.03, "learning_rate": 1.9590163934426232e-05, "loss": 25.2288, "step": 2390 }, { "epoch": 0.03, "learning_rate": 1.9672131147540985e-05, "loss": 24.872, "step": 2400 }, { "epoch": 0.03, "learning_rate": 1.975409836065574e-05, "loss": 24.6531, "step": 2410 }, { "epoch": 0.03, "learning_rate": 1.9836065573770492e-05, "loss": 25.0888, "step": 2420 }, { "epoch": 0.03, "learning_rate": 1.991803278688525e-05, "loss": 24.9576, "step": 2430 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 25.3078, "step": 2440 }, { "epoch": 0.03, "learning_rate": 1.9999999206724512e-05, "loss": 24.8604, "step": 2450 }, { "epoch": 0.03, "learning_rate": 1.999999682689816e-05, "loss": 25.2267, "step": 2460 }, { "epoch": 0.03, "learning_rate": 1.9999992860521337e-05, "loss": 25.4556, "step": 2470 }, { "epoch": 0.03, "learning_rate": 1.9999987307594658e-05, "loss": 25.0821, "step": 2480 }, { "epoch": 0.03, "learning_rate": 1.999998016811901e-05, "loss": 24.8077, "step": 2490 }, { "epoch": 0.03, "learning_rate": 1.9999971442095524e-05, "loss": 24.8103, "step": 2500 }, { "epoch": 0.03, "learning_rate": 1.999996112952559e-05, "loss": 24.7786, "step": 2510 }, { "epoch": 0.03, "learning_rate": 1.999994923041084e-05, "loss": 25.2657, "step": 2520 }, { "epoch": 0.03, "learning_rate": 1.9999935744753164e-05, "loss": 25.1923, "step": 2530 }, { "epoch": 0.03, "learning_rate": 1.9999920672554693e-05, "loss": 25.4469, "step": 2540 }, { "epoch": 0.03, "learning_rate": 1.999990401381783e-05, "loss": 25.7019, "step": 2550 }, { "epoch": 0.03, "learning_rate": 1.999988576854521e-05, "loss": 25.0008, "step": 2560 }, { "epoch": 0.03, "learning_rate": 1.9999865936739732e-05, "loss": 25.0297, "step": 2570 }, { "epoch": 0.03, "learning_rate": 1.999984451840454e-05, "loss": 24.8832, "step": 2580 }, { "epoch": 0.03, "learning_rate": 1.9999821513543037e-05, "loss": 25.4443, "step": 2590 }, { "epoch": 0.03, "learning_rate": 1.9999796922158868e-05, "loss": 25.0165, "step": 2600 }, { "epoch": 0.03, "learning_rate": 1.9999770744255934e-05, "loss": 25.0618, "step": 2610 }, { "epoch": 0.03, "learning_rate": 1.999974297983839e-05, "loss": 24.5812, "step": 2620 }, { "epoch": 0.03, "learning_rate": 1.9999713628910637e-05, "loss": 25.1198, "step": 2630 }, { "epoch": 0.03, "learning_rate": 1.999968269147734e-05, "loss": 24.7516, "step": 2640 }, { "epoch": 0.03, "learning_rate": 1.9999650167543402e-05, "loss": 24.8594, "step": 2650 }, { "epoch": 0.03, "learning_rate": 1.9999616057113983e-05, "loss": 25.3804, "step": 2660 }, { "epoch": 0.03, "learning_rate": 1.9999580360194496e-05, "loss": 25.0233, "step": 2670 }, { "epoch": 0.03, "learning_rate": 1.9999543076790602e-05, "loss": 25.2274, "step": 2680 }, { "epoch": 0.03, "learning_rate": 1.9999504206908226e-05, "loss": 24.4068, "step": 2690 }, { "epoch": 0.03, "learning_rate": 1.999946375055352e-05, "loss": 25.1119, "step": 2700 }, { "epoch": 0.03, "learning_rate": 1.9999421707732912e-05, "loss": 25.1095, "step": 2710 }, { "epoch": 0.03, "learning_rate": 1.999937807845307e-05, "loss": 25.3042, "step": 2720 }, { "epoch": 0.03, "learning_rate": 1.9999332862720917e-05, "loss": 24.3286, "step": 2730 }, { "epoch": 0.03, "learning_rate": 1.9999286060543627e-05, "loss": 25.2253, "step": 2740 }, { "epoch": 0.03, "learning_rate": 1.9999237671928622e-05, "loss": 25.2525, "step": 2750 }, { "epoch": 0.03, "learning_rate": 1.9999187696883582e-05, "loss": 24.7763, "step": 2760 }, { "epoch": 0.03, "learning_rate": 1.9999136135416436e-05, "loss": 24.7266, "step": 2770 }, { "epoch": 0.03, "learning_rate": 1.9999082987535365e-05, "loss": 25.2048, "step": 2780 }, { "epoch": 0.03, "learning_rate": 1.9999028253248797e-05, "loss": 24.9774, "step": 2790 }, { "epoch": 0.03, "learning_rate": 1.999897193256542e-05, "loss": 25.4292, "step": 2800 }, { "epoch": 0.03, "learning_rate": 1.9998914025494165e-05, "loss": 25.2419, "step": 2810 }, { "epoch": 0.03, "learning_rate": 1.9998854532044227e-05, "loss": 25.17, "step": 2820 }, { "epoch": 0.03, "learning_rate": 1.999879345222504e-05, "loss": 24.698, "step": 2830 }, { "epoch": 0.03, "learning_rate": 1.9998730786046295e-05, "loss": 25.2263, "step": 2840 }, { "epoch": 0.04, "learning_rate": 1.999866653351793e-05, "loss": 24.9513, "step": 2850 }, { "epoch": 0.04, "learning_rate": 1.9998600694650153e-05, "loss": 25.4221, "step": 2860 }, { "epoch": 0.04, "learning_rate": 1.9998533269453393e-05, "loss": 25.196, "step": 2870 }, { "epoch": 0.04, "learning_rate": 1.9998464257938358e-05, "loss": 24.9158, "step": 2880 }, { "epoch": 0.04, "learning_rate": 1.999839366011599e-05, "loss": 24.9507, "step": 2890 }, { "epoch": 0.04, "learning_rate": 1.9998321475997495e-05, "loss": 25.5247, "step": 2900 }, { "epoch": 0.04, "learning_rate": 1.999824770559432e-05, "loss": 25.4312, "step": 2910 }, { "epoch": 0.04, "learning_rate": 1.9998172348918178e-05, "loss": 24.4901, "step": 2920 }, { "epoch": 0.04, "learning_rate": 1.999809540598102e-05, "loss": 24.9661, "step": 2930 }, { "epoch": 0.04, "learning_rate": 1.9998016876795047e-05, "loss": 24.9178, "step": 2940 }, { "epoch": 0.04, "learning_rate": 1.9997936761372728e-05, "loss": 25.2357, "step": 2950 }, { "epoch": 0.04, "learning_rate": 1.999785505972677e-05, "loss": 25.1108, "step": 2960 }, { "epoch": 0.04, "learning_rate": 1.999777177187013e-05, "loss": 25.0108, "step": 2970 }, { "epoch": 0.04, "learning_rate": 1.999768689781603e-05, "loss": 24.8748, "step": 2980 }, { "epoch": 0.04, "learning_rate": 1.9997600437577935e-05, "loss": 25.0071, "step": 2990 }, { "epoch": 0.04, "learning_rate": 1.999751239116956e-05, "loss": 25.1276, "step": 3000 }, { "epoch": 0.04, "learning_rate": 1.999742275860487e-05, "loss": 25.0661, "step": 3010 }, { "epoch": 0.04, "learning_rate": 1.9997331539898093e-05, "loss": 24.3578, "step": 3020 }, { "epoch": 0.04, "learning_rate": 1.9997238735063695e-05, "loss": 24.7596, "step": 3030 }, { "epoch": 0.04, "learning_rate": 1.9997144344116407e-05, "loss": 24.9314, "step": 3040 }, { "epoch": 0.04, "learning_rate": 1.9997048367071197e-05, "loss": 25.4871, "step": 3050 }, { "epoch": 0.04, "learning_rate": 1.9996950803943295e-05, "loss": 24.4153, "step": 3060 }, { "epoch": 0.04, "learning_rate": 1.9996851654748185e-05, "loss": 24.9819, "step": 3070 }, { "epoch": 0.04, "learning_rate": 1.999675091950159e-05, "loss": 25.4436, "step": 3080 }, { "epoch": 0.04, "learning_rate": 1.9996648598219498e-05, "loss": 24.7625, "step": 3090 }, { "epoch": 0.04, "learning_rate": 1.999654469091814e-05, "loss": 25.1039, "step": 3100 }, { "epoch": 0.04, "learning_rate": 1.9996439197614e-05, "loss": 25.2747, "step": 3110 }, { "epoch": 0.04, "learning_rate": 1.9996332118323816e-05, "loss": 25.2124, "step": 3120 }, { "epoch": 0.04, "learning_rate": 1.9996223453064577e-05, "loss": 24.9606, "step": 3130 }, { "epoch": 0.04, "learning_rate": 1.9996113201853524e-05, "loss": 24.5669, "step": 3140 }, { "epoch": 0.04, "learning_rate": 1.999600136470815e-05, "loss": 25.0641, "step": 3150 }, { "epoch": 0.04, "learning_rate": 1.9995887941646196e-05, "loss": 24.5276, "step": 3160 }, { "epoch": 0.04, "learning_rate": 1.999577293268566e-05, "loss": 25.159, "step": 3170 }, { "epoch": 0.04, "learning_rate": 1.9995656337844785e-05, "loss": 24.18, "step": 3180 }, { "epoch": 0.04, "learning_rate": 1.9995538157142073e-05, "loss": 24.9421, "step": 3190 }, { "epoch": 0.04, "learning_rate": 1.999541839059627e-05, "loss": 24.3875, "step": 3200 }, { "epoch": 0.04, "learning_rate": 1.9995297038226383e-05, "loss": 24.7851, "step": 3210 }, { "epoch": 0.04, "learning_rate": 1.999517410005166e-05, "loss": 24.5867, "step": 3220 }, { "epoch": 0.04, "learning_rate": 1.999504957609161e-05, "loss": 25.0812, "step": 3230 }, { "epoch": 0.04, "learning_rate": 1.9994923466365983e-05, "loss": 24.8647, "step": 3240 }, { "epoch": 0.04, "learning_rate": 1.9994795770894796e-05, "loss": 25.468, "step": 3250 }, { "epoch": 0.04, "learning_rate": 1.99946664896983e-05, "loss": 24.2568, "step": 3260 }, { "epoch": 0.04, "learning_rate": 1.9994535622797013e-05, "loss": 24.7718, "step": 3270 }, { "epoch": 0.04, "learning_rate": 1.999440317021169e-05, "loss": 24.7892, "step": 3280 }, { "epoch": 0.04, "learning_rate": 1.9994269131963354e-05, "loss": 24.9733, "step": 3290 }, { "epoch": 0.04, "learning_rate": 1.9994133508073265e-05, "loss": 24.3024, "step": 3300 }, { "epoch": 0.04, "learning_rate": 1.999399629856294e-05, "loss": 25.6103, "step": 3310 }, { "epoch": 0.04, "learning_rate": 1.9993857503454152e-05, "loss": 24.7009, "step": 3320 }, { "epoch": 0.04, "learning_rate": 1.999371712276892e-05, "loss": 25.1231, "step": 3330 }, { "epoch": 0.04, "learning_rate": 1.9993575156529513e-05, "loss": 24.2778, "step": 3340 }, { "epoch": 0.04, "learning_rate": 1.9993431604758464e-05, "loss": 24.733, "step": 3350 }, { "epoch": 0.04, "learning_rate": 1.9993286467478533e-05, "loss": 24.8114, "step": 3360 }, { "epoch": 0.04, "learning_rate": 1.9993139744712763e-05, "loss": 24.3561, "step": 3370 }, { "epoch": 0.04, "learning_rate": 1.9992991436484418e-05, "loss": 25.3879, "step": 3380 }, { "epoch": 0.04, "learning_rate": 1.999284154281704e-05, "loss": 23.8962, "step": 3390 }, { "epoch": 0.04, "learning_rate": 1.9992690063734403e-05, "loss": 24.7544, "step": 3400 }, { "epoch": 0.04, "learning_rate": 1.999253699926054e-05, "loss": 25.0143, "step": 3410 }, { "epoch": 0.04, "learning_rate": 1.9992382349419742e-05, "loss": 24.4987, "step": 3420 }, { "epoch": 0.04, "learning_rate": 1.9992226114236538e-05, "loss": 24.3587, "step": 3430 }, { "epoch": 0.04, "learning_rate": 1.9992068293735716e-05, "loss": 25.012, "step": 3440 }, { "epoch": 0.04, "learning_rate": 1.9991908887942317e-05, "loss": 24.8483, "step": 3450 }, { "epoch": 0.04, "learning_rate": 1.9991747896881635e-05, "loss": 24.7096, "step": 3460 }, { "epoch": 0.04, "learning_rate": 1.9991585320579205e-05, "loss": 24.4604, "step": 3470 }, { "epoch": 0.04, "learning_rate": 1.9991421159060825e-05, "loss": 24.7097, "step": 3480 }, { "epoch": 0.04, "learning_rate": 1.999125541235254e-05, "loss": 24.367, "step": 3490 }, { "epoch": 0.04, "learning_rate": 1.9991088080480642e-05, "loss": 24.3909, "step": 3500 }, { "epoch": 0.04, "learning_rate": 1.9990919163471684e-05, "loss": 24.3412, "step": 3510 }, { "epoch": 0.04, "learning_rate": 1.9990748661352466e-05, "loss": 24.5649, "step": 3520 }, { "epoch": 0.04, "learning_rate": 1.9990576574150037e-05, "loss": 24.8146, "step": 3530 }, { "epoch": 0.04, "learning_rate": 1.99904029018917e-05, "loss": 25.0072, "step": 3540 }, { "epoch": 0.04, "learning_rate": 1.999022764460501e-05, "loss": 24.9415, "step": 3550 }, { "epoch": 0.04, "learning_rate": 1.999005080231777e-05, "loss": 24.3291, "step": 3560 }, { "epoch": 0.04, "learning_rate": 1.9989872375058034e-05, "loss": 24.8531, "step": 3570 }, { "epoch": 0.04, "learning_rate": 1.998969236285412e-05, "loss": 25.1282, "step": 3580 }, { "epoch": 0.04, "learning_rate": 1.998951076573458e-05, "loss": 24.3484, "step": 3590 }, { "epoch": 0.04, "learning_rate": 1.9989327583728228e-05, "loss": 25.1501, "step": 3600 }, { "epoch": 0.04, "learning_rate": 1.9989142816864127e-05, "loss": 24.456, "step": 3610 }, { "epoch": 0.04, "learning_rate": 1.998895646517159e-05, "loss": 25.6157, "step": 3620 }, { "epoch": 0.04, "learning_rate": 1.998876852868018e-05, "loss": 24.7954, "step": 3630 }, { "epoch": 0.04, "learning_rate": 1.9988579007419724e-05, "loss": 24.8893, "step": 3640 }, { "epoch": 0.04, "learning_rate": 1.998838790142028e-05, "loss": 24.6207, "step": 3650 }, { "epoch": 0.05, "learning_rate": 1.9988195210712174e-05, "loss": 24.8112, "step": 3660 }, { "epoch": 0.05, "learning_rate": 1.9988000935325973e-05, "loss": 24.8551, "step": 3670 }, { "epoch": 0.05, "learning_rate": 1.9987805075292504e-05, "loss": 24.8236, "step": 3680 }, { "epoch": 0.05, "learning_rate": 1.9987607630642837e-05, "loss": 24.5339, "step": 3690 }, { "epoch": 0.05, "learning_rate": 1.99874086014083e-05, "loss": 24.265, "step": 3700 }, { "epoch": 0.05, "learning_rate": 1.998720798762047e-05, "loss": 24.454, "step": 3710 }, { "epoch": 0.05, "learning_rate": 1.998700578931118e-05, "loss": 24.6033, "step": 3720 }, { "epoch": 0.05, "learning_rate": 1.99868020065125e-05, "loss": 24.7044, "step": 3730 }, { "epoch": 0.05, "learning_rate": 1.9986596639256763e-05, "loss": 24.6697, "step": 3740 }, { "epoch": 0.05, "learning_rate": 1.998638968757656e-05, "loss": 25.0206, "step": 3750 }, { "epoch": 0.05, "learning_rate": 1.998618115150472e-05, "loss": 24.9192, "step": 3760 }, { "epoch": 0.05, "learning_rate": 1.9985971031074324e-05, "loss": 24.9982, "step": 3770 }, { "epoch": 0.05, "learning_rate": 1.9985759326318717e-05, "loss": 23.9823, "step": 3780 }, { "epoch": 0.05, "learning_rate": 1.998554603727148e-05, "loss": 24.3654, "step": 3790 }, { "epoch": 0.05, "learning_rate": 1.9985331163966454e-05, "loss": 24.6956, "step": 3800 }, { "epoch": 0.05, "learning_rate": 1.9985114706437734e-05, "loss": 24.3876, "step": 3810 }, { "epoch": 0.05, "learning_rate": 1.9984896664719657e-05, "loss": 24.8851, "step": 3820 }, { "epoch": 0.05, "learning_rate": 1.998467703884682e-05, "loss": 24.7123, "step": 3830 }, { "epoch": 0.05, "learning_rate": 1.9984455828854067e-05, "loss": 24.9964, "step": 3840 }, { "epoch": 0.05, "learning_rate": 1.9984233034776487e-05, "loss": 24.2747, "step": 3850 }, { "epoch": 0.05, "learning_rate": 1.998400865664944e-05, "loss": 24.6305, "step": 3860 }, { "epoch": 0.05, "learning_rate": 1.9983782694508515e-05, "loss": 25.0437, "step": 3870 }, { "epoch": 0.05, "learning_rate": 1.9983555148389567e-05, "loss": 24.4758, "step": 3880 }, { "epoch": 0.05, "learning_rate": 1.9983326018328695e-05, "loss": 24.4291, "step": 3890 }, { "epoch": 0.05, "learning_rate": 1.9983095304362253e-05, "loss": 24.3748, "step": 3900 }, { "epoch": 0.05, "learning_rate": 1.9982863006526846e-05, "loss": 24.7051, "step": 3910 }, { "epoch": 0.05, "learning_rate": 1.9982629124859324e-05, "loss": 24.5649, "step": 3920 }, { "epoch": 0.05, "learning_rate": 1.99823936593968e-05, "loss": 24.9019, "step": 3930 }, { "epoch": 0.05, "learning_rate": 1.9982156610176626e-05, "loss": 25.1051, "step": 3940 }, { "epoch": 0.05, "learning_rate": 1.9981917977236415e-05, "loss": 24.459, "step": 3950 }, { "epoch": 0.05, "learning_rate": 1.9981677760614027e-05, "loss": 24.7265, "step": 3960 }, { "epoch": 0.05, "learning_rate": 1.9981435960347572e-05, "loss": 24.2788, "step": 3970 }, { "epoch": 0.05, "learning_rate": 1.9981192576475417e-05, "loss": 24.3874, "step": 3980 }, { "epoch": 0.05, "learning_rate": 1.9980947609036166e-05, "loss": 24.7302, "step": 3990 }, { "epoch": 0.05, "learning_rate": 1.99807010580687e-05, "loss": 24.6663, "step": 4000 }, { "epoch": 0.05, "learning_rate": 1.998045292361212e-05, "loss": 25.1816, "step": 4010 }, { "epoch": 0.05, "learning_rate": 1.9980203205705804e-05, "loss": 25.217, "step": 4020 }, { "epoch": 0.05, "learning_rate": 1.9979951904389364e-05, "loss": 24.338, "step": 4030 }, { "epoch": 0.05, "learning_rate": 1.997969901970268e-05, "loss": 25.3183, "step": 4040 }, { "epoch": 0.05, "learning_rate": 1.997944455168586e-05, "loss": 24.6901, "step": 4050 }, { "epoch": 0.05, "learning_rate": 1.997918850037929e-05, "loss": 24.6477, "step": 4060 }, { "epoch": 0.05, "learning_rate": 1.9978930865823582e-05, "loss": 24.9119, "step": 4070 }, { "epoch": 0.05, "learning_rate": 1.9978671648059622e-05, "loss": 24.7799, "step": 4080 }, { "epoch": 0.05, "learning_rate": 1.9978410847128525e-05, "loss": 24.9764, "step": 4090 }, { "epoch": 0.05, "learning_rate": 1.997814846307168e-05, "loss": 24.7056, "step": 4100 }, { "epoch": 0.05, "learning_rate": 1.997788449593071e-05, "loss": 24.261, "step": 4110 }, { "epoch": 0.05, "learning_rate": 1.997761894574749e-05, "loss": 24.6061, "step": 4120 }, { "epoch": 0.05, "learning_rate": 1.997735181256416e-05, "loss": 24.7915, "step": 4130 }, { "epoch": 0.05, "learning_rate": 1.9977083096423096e-05, "loss": 24.5038, "step": 4140 }, { "epoch": 0.05, "learning_rate": 1.997681279736693e-05, "loss": 24.4274, "step": 4150 }, { "epoch": 0.05, "learning_rate": 1.9976540915438553e-05, "loss": 24.6798, "step": 4160 }, { "epoch": 0.05, "learning_rate": 1.9976267450681095e-05, "loss": 24.5629, "step": 4170 }, { "epoch": 0.05, "learning_rate": 1.9975992403137946e-05, "loss": 24.5296, "step": 4180 }, { "epoch": 0.05, "learning_rate": 1.9975715772852743e-05, "loss": 24.8781, "step": 4190 }, { "epoch": 0.05, "learning_rate": 1.9975437559869367e-05, "loss": 24.5083, "step": 4200 }, { "epoch": 0.05, "learning_rate": 1.997515776423197e-05, "loss": 24.2887, "step": 4210 }, { "epoch": 0.05, "learning_rate": 1.9974876385984937e-05, "loss": 24.6184, "step": 4220 }, { "epoch": 0.05, "learning_rate": 1.997459342517291e-05, "loss": 24.538, "step": 4230 }, { "epoch": 0.05, "learning_rate": 1.997430888184078e-05, "loss": 24.5282, "step": 4240 }, { "epoch": 0.05, "learning_rate": 1.9974022756033696e-05, "loss": 24.6795, "step": 4250 }, { "epoch": 0.05, "learning_rate": 1.9973735047797053e-05, "loss": 24.9176, "step": 4260 }, { "epoch": 0.05, "learning_rate": 1.9973445757176498e-05, "loss": 23.9119, "step": 4270 }, { "epoch": 0.05, "learning_rate": 1.997315488421792e-05, "loss": 24.5437, "step": 4280 }, { "epoch": 0.05, "learning_rate": 1.9972862428967475e-05, "loss": 25.113, "step": 4290 }, { "epoch": 0.05, "learning_rate": 1.9972568391471565e-05, "loss": 24.5959, "step": 4300 }, { "epoch": 0.05, "learning_rate": 1.9972272771776837e-05, "loss": 24.6978, "step": 4310 }, { "epoch": 0.05, "learning_rate": 1.9971975569930187e-05, "loss": 24.5467, "step": 4320 }, { "epoch": 0.05, "learning_rate": 1.9971676785978777e-05, "loss": 24.7619, "step": 4330 }, { "epoch": 0.05, "learning_rate": 1.9971376419970005e-05, "loss": 24.5437, "step": 4340 }, { "epoch": 0.05, "learning_rate": 1.9971074471951527e-05, "loss": 24.7483, "step": 4350 }, { "epoch": 0.05, "learning_rate": 1.997077094197125e-05, "loss": 24.7618, "step": 4360 }, { "epoch": 0.05, "learning_rate": 1.997046583007733e-05, "loss": 24.4072, "step": 4370 }, { "epoch": 0.05, "learning_rate": 1.997015913631817e-05, "loss": 24.4685, "step": 4380 }, { "epoch": 0.05, "learning_rate": 1.9969850860742436e-05, "loss": 24.5359, "step": 4390 }, { "epoch": 0.05, "learning_rate": 1.9969541003399033e-05, "loss": 24.3548, "step": 4400 }, { "epoch": 0.05, "learning_rate": 1.9969229564337122e-05, "loss": 24.6333, "step": 4410 }, { "epoch": 0.05, "learning_rate": 1.9968916543606113e-05, "loss": 24.8061, "step": 4420 }, { "epoch": 0.05, "learning_rate": 1.9968601941255673e-05, "loss": 24.6622, "step": 4430 }, { "epoch": 0.05, "learning_rate": 1.996828575733571e-05, "loss": 24.9534, "step": 4440 }, { "epoch": 0.05, "learning_rate": 1.9967967991896393e-05, "loss": 24.4709, "step": 4450 }, { "epoch": 0.05, "learning_rate": 1.9967648644988135e-05, "loss": 24.3276, "step": 4460 }, { "epoch": 0.05, "learning_rate": 1.99673277166616e-05, "loss": 24.2676, "step": 4470 }, { "epoch": 0.06, "learning_rate": 1.9967005206967704e-05, "loss": 24.5079, "step": 4480 }, { "epoch": 0.06, "learning_rate": 1.996668111595762e-05, "loss": 24.5939, "step": 4490 }, { "epoch": 0.06, "learning_rate": 1.9966355443682762e-05, "loss": 24.5334, "step": 4500 }, { "epoch": 0.06, "learning_rate": 1.9966028190194802e-05, "loss": 24.3562, "step": 4510 }, { "epoch": 0.06, "learning_rate": 1.9965699355545662e-05, "loss": 24.7548, "step": 4520 }, { "epoch": 0.06, "learning_rate": 1.996536893978751e-05, "loss": 23.9627, "step": 4530 }, { "epoch": 0.06, "learning_rate": 1.9965036942972772e-05, "loss": 24.7824, "step": 4540 }, { "epoch": 0.06, "learning_rate": 1.996470336515412e-05, "loss": 25.1635, "step": 4550 }, { "epoch": 0.06, "learning_rate": 1.9964368206384467e-05, "loss": 24.8275, "step": 4560 }, { "epoch": 0.06, "learning_rate": 1.9964031466717003e-05, "loss": 24.1721, "step": 4570 }, { "epoch": 0.06, "learning_rate": 1.9963693146205148e-05, "loss": 24.1989, "step": 4580 }, { "epoch": 0.06, "learning_rate": 1.9963353244902576e-05, "loss": 24.4533, "step": 4590 }, { "epoch": 0.06, "learning_rate": 1.9963011762863217e-05, "loss": 24.2562, "step": 4600 }, { "epoch": 0.06, "learning_rate": 1.996266870014125e-05, "loss": 24.2692, "step": 4610 }, { "epoch": 0.06, "learning_rate": 1.9962324056791095e-05, "loss": 24.3142, "step": 4620 }, { "epoch": 0.06, "learning_rate": 1.996197783286744e-05, "loss": 24.8496, "step": 4630 }, { "epoch": 0.06, "learning_rate": 1.9961630028425214e-05, "loss": 24.7462, "step": 4640 }, { "epoch": 0.06, "learning_rate": 1.9961280643519597e-05, "loss": 25.0659, "step": 4650 }, { "epoch": 0.06, "learning_rate": 1.996092967820602e-05, "loss": 24.4044, "step": 4660 }, { "epoch": 0.06, "learning_rate": 1.9960577132540165e-05, "loss": 24.9168, "step": 4670 }, { "epoch": 0.06, "learning_rate": 1.996022300657797e-05, "loss": 24.8774, "step": 4680 }, { "epoch": 0.06, "learning_rate": 1.995986730037561e-05, "loss": 24.1602, "step": 4690 }, { "epoch": 0.06, "learning_rate": 1.9959510013989527e-05, "loss": 24.2732, "step": 4700 }, { "epoch": 0.06, "learning_rate": 1.9959151147476405e-05, "loss": 24.4093, "step": 4710 }, { "epoch": 0.06, "learning_rate": 1.9958790700893176e-05, "loss": 23.993, "step": 4720 }, { "epoch": 0.06, "learning_rate": 1.995842867429703e-05, "loss": 24.3103, "step": 4730 }, { "epoch": 0.06, "learning_rate": 1.9958065067745406e-05, "loss": 24.4675, "step": 4740 }, { "epoch": 0.06, "learning_rate": 1.995769988129599e-05, "loss": 24.2461, "step": 4750 }, { "epoch": 0.06, "learning_rate": 1.995733311500672e-05, "loss": 24.7322, "step": 4760 }, { "epoch": 0.06, "learning_rate": 1.9956964768935788e-05, "loss": 24.7001, "step": 4770 }, { "epoch": 0.06, "learning_rate": 1.995659484314163e-05, "loss": 24.4238, "step": 4780 }, { "epoch": 0.06, "learning_rate": 1.9956223337682935e-05, "loss": 24.5846, "step": 4790 }, { "epoch": 0.06, "learning_rate": 1.9955850252618655e-05, "loss": 24.9899, "step": 4800 }, { "epoch": 0.06, "learning_rate": 1.995547558800797e-05, "loss": 24.4184, "step": 4810 }, { "epoch": 0.06, "learning_rate": 1.9955099343910327e-05, "loss": 24.8782, "step": 4820 }, { "epoch": 0.06, "learning_rate": 1.995472152038542e-05, "loss": 24.4626, "step": 4830 }, { "epoch": 0.06, "learning_rate": 1.995434211749319e-05, "loss": 24.8099, "step": 4840 }, { "epoch": 0.06, "learning_rate": 1.9953961135293837e-05, "loss": 24.4939, "step": 4850 }, { "epoch": 0.06, "learning_rate": 1.99535785738478e-05, "loss": 23.9342, "step": 4860 }, { "epoch": 0.06, "learning_rate": 1.9953194433215778e-05, "loss": 23.9494, "step": 4870 }, { "epoch": 0.06, "learning_rate": 1.9952808713458713e-05, "loss": 24.6229, "step": 4880 }, { "epoch": 0.06, "learning_rate": 1.9952421414637803e-05, "loss": 24.3811, "step": 4890 }, { "epoch": 0.06, "learning_rate": 1.99520325368145e-05, "loss": 24.7037, "step": 4900 }, { "epoch": 0.06, "learning_rate": 1.9951642080050487e-05, "loss": 23.8055, "step": 4910 }, { "epoch": 0.06, "learning_rate": 1.995125004440773e-05, "loss": 24.2426, "step": 4920 }, { "epoch": 0.06, "learning_rate": 1.9950856429948417e-05, "loss": 24.0605, "step": 4930 }, { "epoch": 0.06, "learning_rate": 1.9950461236735e-05, "loss": 24.5315, "step": 4940 }, { "epoch": 0.06, "learning_rate": 1.9950064464830176e-05, "loss": 24.6932, "step": 4950 }, { "epoch": 0.06, "learning_rate": 1.9949666114296898e-05, "loss": 24.7879, "step": 4960 }, { "epoch": 0.06, "learning_rate": 1.9949266185198363e-05, "loss": 24.5258, "step": 4970 }, { "epoch": 0.06, "learning_rate": 1.9948864677598022e-05, "loss": 24.354, "step": 4980 }, { "epoch": 0.06, "learning_rate": 1.994846159155958e-05, "loss": 23.9269, "step": 4990 }, { "epoch": 0.06, "learning_rate": 1.994805692714699e-05, "loss": 24.1679, "step": 5000 }, { "epoch": 0.06, "learning_rate": 1.9947650684424443e-05, "loss": 24.2143, "step": 5010 }, { "epoch": 0.06, "learning_rate": 1.9947242863456407e-05, "loss": 24.052, "step": 5020 }, { "epoch": 0.06, "learning_rate": 1.9946833464307575e-05, "loss": 24.2856, "step": 5030 }, { "epoch": 0.06, "learning_rate": 1.9946422487042897e-05, "loss": 24.3727, "step": 5040 }, { "epoch": 0.06, "learning_rate": 1.9946009931727585e-05, "loss": 24.7897, "step": 5050 }, { "epoch": 0.06, "learning_rate": 1.9945595798427087e-05, "loss": 24.4636, "step": 5060 }, { "epoch": 0.06, "learning_rate": 1.9945180087207116e-05, "loss": 24.5299, "step": 5070 }, { "epoch": 0.06, "learning_rate": 1.9944762798133615e-05, "loss": 24.6935, "step": 5080 }, { "epoch": 0.06, "learning_rate": 1.99443439312728e-05, "loss": 24.2355, "step": 5090 }, { "epoch": 0.06, "learning_rate": 1.994392348669112e-05, "loss": 24.6241, "step": 5100 }, { "epoch": 0.06, "learning_rate": 1.9943501464455283e-05, "loss": 24.0297, "step": 5110 }, { "epoch": 0.06, "learning_rate": 1.9943077864632238e-05, "loss": 25.1312, "step": 5120 }, { "epoch": 0.06, "learning_rate": 1.99426526872892e-05, "loss": 24.2793, "step": 5130 }, { "epoch": 0.06, "learning_rate": 1.9942225932493627e-05, "loss": 24.1844, "step": 5140 }, { "epoch": 0.06, "learning_rate": 1.9941797600313217e-05, "loss": 24.0795, "step": 5150 }, { "epoch": 0.06, "learning_rate": 1.9941367690815933e-05, "loss": 24.3908, "step": 5160 }, { "epoch": 0.06, "learning_rate": 1.994093620406998e-05, "loss": 24.0364, "step": 5170 }, { "epoch": 0.06, "learning_rate": 1.994050314014382e-05, "loss": 23.8404, "step": 5180 }, { "epoch": 0.06, "learning_rate": 1.9940068499106153e-05, "loss": 24.3954, "step": 5190 }, { "epoch": 0.06, "learning_rate": 1.9939632281025944e-05, "loss": 24.7963, "step": 5200 }, { "epoch": 0.06, "learning_rate": 1.9939194485972398e-05, "loss": 24.4077, "step": 5210 }, { "epoch": 0.06, "learning_rate": 1.9938755114014975e-05, "loss": 24.1392, "step": 5220 }, { "epoch": 0.06, "learning_rate": 1.993831416522338e-05, "loss": 24.5027, "step": 5230 }, { "epoch": 0.06, "learning_rate": 1.9937871639667574e-05, "loss": 24.3514, "step": 5240 }, { "epoch": 0.06, "learning_rate": 1.9937427537417767e-05, "loss": 24.2398, "step": 5250 }, { "epoch": 0.06, "learning_rate": 1.993698185854442e-05, "loss": 24.3929, "step": 5260 }, { "epoch": 0.06, "learning_rate": 1.9936534603118237e-05, "loss": 24.5277, "step": 5270 }, { "epoch": 0.06, "learning_rate": 1.9936085771210183e-05, "loss": 24.7694, "step": 5280 }, { "epoch": 0.07, "learning_rate": 1.993563536289146e-05, "loss": 24.5949, "step": 5290 }, { "epoch": 0.07, "learning_rate": 1.993518337823354e-05, "loss": 24.0181, "step": 5300 }, { "epoch": 0.07, "learning_rate": 1.9934729817308115e-05, "loss": 24.7018, "step": 5310 }, { "epoch": 0.07, "learning_rate": 1.993427468018716e-05, "loss": 24.5961, "step": 5320 }, { "epoch": 0.07, "learning_rate": 1.993381796694288e-05, "loss": 24.2317, "step": 5330 }, { "epoch": 0.07, "learning_rate": 1.993335967764773e-05, "loss": 24.9055, "step": 5340 }, { "epoch": 0.07, "learning_rate": 1.9932899812374427e-05, "loss": 24.9409, "step": 5350 }, { "epoch": 0.07, "learning_rate": 1.9932438371195926e-05, "loss": 24.3472, "step": 5360 }, { "epoch": 0.07, "learning_rate": 1.993197535418544e-05, "loss": 24.6407, "step": 5370 }, { "epoch": 0.07, "learning_rate": 1.993151076141643e-05, "loss": 23.7955, "step": 5380 }, { "epoch": 0.07, "learning_rate": 1.9931044592962602e-05, "loss": 24.4674, "step": 5390 }, { "epoch": 0.07, "learning_rate": 1.993057684889792e-05, "loss": 23.9117, "step": 5400 }, { "epoch": 0.07, "learning_rate": 1.993010752929659e-05, "loss": 24.5586, "step": 5410 }, { "epoch": 0.07, "learning_rate": 1.9929636634233075e-05, "loss": 24.5149, "step": 5420 }, { "epoch": 0.07, "learning_rate": 1.992916416378208e-05, "loss": 24.434, "step": 5430 }, { "epoch": 0.07, "learning_rate": 1.9928690118018578e-05, "loss": 24.5712, "step": 5440 }, { "epoch": 0.07, "learning_rate": 1.9928214497017763e-05, "loss": 24.4062, "step": 5450 }, { "epoch": 0.07, "learning_rate": 1.9927737300855102e-05, "loss": 24.6374, "step": 5460 }, { "epoch": 0.07, "learning_rate": 1.9927258529606303e-05, "loss": 24.2215, "step": 5470 }, { "epoch": 0.07, "learning_rate": 1.9926778183347326e-05, "loss": 24.5144, "step": 5480 }, { "epoch": 0.07, "learning_rate": 1.9926296262154385e-05, "loss": 24.5789, "step": 5490 }, { "epoch": 0.07, "learning_rate": 1.992581276610393e-05, "loss": 24.0762, "step": 5500 }, { "epoch": 0.07, "learning_rate": 1.9925327695272678e-05, "loss": 24.5254, "step": 5510 }, { "epoch": 0.07, "learning_rate": 1.9924841049737583e-05, "loss": 25.1053, "step": 5520 }, { "epoch": 0.07, "learning_rate": 1.992435282957586e-05, "loss": 24.579, "step": 5530 }, { "epoch": 0.07, "learning_rate": 1.992386303486496e-05, "loss": 24.3352, "step": 5540 }, { "epoch": 0.07, "learning_rate": 1.9923371665682594e-05, "loss": 24.3193, "step": 5550 }, { "epoch": 0.07, "learning_rate": 1.9922878722106723e-05, "loss": 24.2245, "step": 5560 }, { "epoch": 0.07, "learning_rate": 1.992238420421555e-05, "loss": 23.9863, "step": 5570 }, { "epoch": 0.07, "learning_rate": 1.992188811208754e-05, "loss": 24.0468, "step": 5580 }, { "epoch": 0.07, "learning_rate": 1.9921390445801397e-05, "loss": 24.7273, "step": 5590 }, { "epoch": 0.07, "learning_rate": 1.9920891205436074e-05, "loss": 24.7287, "step": 5600 }, { "epoch": 0.07, "learning_rate": 1.9920390391070784e-05, "loss": 24.5309, "step": 5610 }, { "epoch": 0.07, "learning_rate": 1.991988800278498e-05, "loss": 24.609, "step": 5620 }, { "epoch": 0.07, "learning_rate": 1.991938404065837e-05, "loss": 24.6423, "step": 5630 }, { "epoch": 0.07, "learning_rate": 1.9918878504770914e-05, "loss": 24.9217, "step": 5640 }, { "epoch": 0.07, "learning_rate": 1.991837139520281e-05, "loss": 24.5743, "step": 5650 }, { "epoch": 0.07, "learning_rate": 1.991786271203452e-05, "loss": 24.2864, "step": 5660 }, { "epoch": 0.07, "learning_rate": 1.9917352455346744e-05, "loss": 24.9791, "step": 5670 }, { "epoch": 0.07, "learning_rate": 1.991684062522044e-05, "loss": 24.2466, "step": 5680 }, { "epoch": 0.07, "learning_rate": 1.9916327221736815e-05, "loss": 25.1627, "step": 5690 }, { "epoch": 0.07, "learning_rate": 1.9915812244977317e-05, "loss": 25.1379, "step": 5700 }, { "epoch": 0.07, "learning_rate": 1.9915295695023654e-05, "loss": 24.628, "step": 5710 }, { "epoch": 0.07, "learning_rate": 1.991477757195778e-05, "loss": 25.278, "step": 5720 }, { "epoch": 0.07, "learning_rate": 1.9914257875861892e-05, "loss": 24.4404, "step": 5730 }, { "epoch": 0.07, "learning_rate": 1.991373660681845e-05, "loss": 24.3918, "step": 5740 }, { "epoch": 0.07, "learning_rate": 1.991321376491015e-05, "loss": 24.6724, "step": 5750 }, { "epoch": 0.07, "learning_rate": 1.9912689350219944e-05, "loss": 24.2653, "step": 5760 }, { "epoch": 0.07, "learning_rate": 1.991216336283104e-05, "loss": 24.533, "step": 5770 }, { "epoch": 0.07, "learning_rate": 1.9911635802826877e-05, "loss": 24.23, "step": 5780 }, { "epoch": 0.07, "learning_rate": 1.9911106670291167e-05, "loss": 24.3108, "step": 5790 }, { "epoch": 0.07, "learning_rate": 1.991057596530785e-05, "loss": 24.9814, "step": 5800 }, { "epoch": 0.07, "learning_rate": 1.9910043687961134e-05, "loss": 24.5174, "step": 5810 }, { "epoch": 0.07, "learning_rate": 1.990950983833546e-05, "loss": 25.0494, "step": 5820 }, { "epoch": 0.07, "learning_rate": 1.990897441651553e-05, "loss": 24.7007, "step": 5830 }, { "epoch": 0.07, "learning_rate": 1.9908437422586288e-05, "loss": 24.3883, "step": 5840 }, { "epoch": 0.07, "learning_rate": 1.9907898856632935e-05, "loss": 24.8504, "step": 5850 }, { "epoch": 0.07, "learning_rate": 1.9907358718740914e-05, "loss": 24.6391, "step": 5860 }, { "epoch": 0.07, "learning_rate": 1.9906817008995924e-05, "loss": 24.2316, "step": 5870 }, { "epoch": 0.07, "learning_rate": 1.9906273727483905e-05, "loss": 24.4287, "step": 5880 }, { "epoch": 0.07, "learning_rate": 1.9905728874291052e-05, "loss": 24.0667, "step": 5890 }, { "epoch": 0.07, "learning_rate": 1.990518244950382e-05, "loss": 24.7445, "step": 5900 }, { "epoch": 0.07, "learning_rate": 1.9904634453208884e-05, "loss": 24.6955, "step": 5910 }, { "epoch": 0.07, "learning_rate": 1.9904084885493197e-05, "loss": 24.602, "step": 5920 }, { "epoch": 0.07, "learning_rate": 1.990353374644395e-05, "loss": 24.7539, "step": 5930 }, { "epoch": 0.07, "learning_rate": 1.990298103614858e-05, "loss": 24.4718, "step": 5940 }, { "epoch": 0.07, "learning_rate": 1.9902426754694788e-05, "loss": 24.3857, "step": 5950 }, { "epoch": 0.07, "learning_rate": 1.99018709021705e-05, "loss": 23.9171, "step": 5960 }, { "epoch": 0.07, "learning_rate": 1.9901313478663917e-05, "loss": 23.9727, "step": 5970 }, { "epoch": 0.07, "learning_rate": 1.9900754484263466e-05, "loss": 23.9484, "step": 5980 }, { "epoch": 0.07, "learning_rate": 1.990019391905784e-05, "loss": 23.9851, "step": 5990 }, { "epoch": 0.07, "learning_rate": 1.9899631783135973e-05, "loss": 24.068, "step": 6000 }, { "epoch": 0.07, "learning_rate": 1.9899068076587055e-05, "loss": 24.8214, "step": 6010 }, { "epoch": 0.07, "learning_rate": 1.989850279950052e-05, "loss": 24.4303, "step": 6020 }, { "epoch": 0.07, "learning_rate": 1.989793595196605e-05, "loss": 24.3905, "step": 6030 }, { "epoch": 0.07, "learning_rate": 1.9897367534073575e-05, "loss": 24.4801, "step": 6040 }, { "epoch": 0.07, "learning_rate": 1.9896797545913284e-05, "loss": 24.9059, "step": 6050 }, { "epoch": 0.07, "learning_rate": 1.9896225987575604e-05, "loss": 24.3359, "step": 6060 }, { "epoch": 0.07, "learning_rate": 1.989565285915122e-05, "loss": 24.6383, "step": 6070 }, { "epoch": 0.07, "learning_rate": 1.9895078160731056e-05, "loss": 24.3035, "step": 6080 }, { "epoch": 0.07, "learning_rate": 1.9894501892406293e-05, "loss": 25.0863, "step": 6090 }, { "epoch": 0.08, "learning_rate": 1.9893924054268363e-05, "loss": 24.5381, "step": 6100 }, { "epoch": 0.08, "learning_rate": 1.989334464640894e-05, "loss": 24.0101, "step": 6110 }, { "epoch": 0.08, "learning_rate": 1.989276366891995e-05, "loss": 24.0678, "step": 6120 }, { "epoch": 0.08, "learning_rate": 1.9892181121893568e-05, "loss": 24.5987, "step": 6130 }, { "epoch": 0.08, "learning_rate": 1.9891597005422217e-05, "loss": 24.618, "step": 6140 }, { "epoch": 0.08, "learning_rate": 1.989101131959857e-05, "loss": 24.0287, "step": 6150 }, { "epoch": 0.08, "learning_rate": 1.989042406451555e-05, "loss": 24.9081, "step": 6160 }, { "epoch": 0.08, "learning_rate": 1.9889835240266326e-05, "loss": 24.3679, "step": 6170 }, { "epoch": 0.08, "learning_rate": 1.9889244846944323e-05, "loss": 24.6125, "step": 6180 }, { "epoch": 0.08, "learning_rate": 1.9888652884643206e-05, "loss": 23.9614, "step": 6190 }, { "epoch": 0.08, "learning_rate": 1.9888059353456896e-05, "loss": 25.0453, "step": 6200 }, { "epoch": 0.08, "learning_rate": 1.9887464253479557e-05, "loss": 23.9821, "step": 6210 }, { "epoch": 0.08, "learning_rate": 1.98868675848056e-05, "loss": 24.4337, "step": 6220 }, { "epoch": 0.08, "learning_rate": 1.9886269347529702e-05, "loss": 23.9692, "step": 6230 }, { "epoch": 0.08, "learning_rate": 1.9885669541746762e-05, "loss": 24.6403, "step": 6240 }, { "epoch": 0.08, "learning_rate": 1.9885068167551956e-05, "loss": 24.6021, "step": 6250 }, { "epoch": 0.08, "learning_rate": 1.9884465225040686e-05, "loss": 24.263, "step": 6260 }, { "epoch": 0.08, "learning_rate": 1.9883860714308612e-05, "loss": 24.1291, "step": 6270 }, { "epoch": 0.08, "learning_rate": 1.9883254635451646e-05, "loss": 24.6281, "step": 6280 }, { "epoch": 0.08, "learning_rate": 1.9882646988565947e-05, "loss": 25.0111, "step": 6290 }, { "epoch": 0.08, "learning_rate": 1.9882037773747916e-05, "loss": 24.8293, "step": 6300 }, { "epoch": 0.08, "learning_rate": 1.988142699109421e-05, "loss": 23.6165, "step": 6310 }, { "epoch": 0.08, "learning_rate": 1.9880814640701738e-05, "loss": 24.567, "step": 6320 }, { "epoch": 0.08, "learning_rate": 1.9880200722667643e-05, "loss": 24.4052, "step": 6330 }, { "epoch": 0.08, "learning_rate": 1.9879585237089337e-05, "loss": 24.4013, "step": 6340 }, { "epoch": 0.08, "learning_rate": 1.9878968184064457e-05, "loss": 24.3499, "step": 6350 }, { "epoch": 0.08, "learning_rate": 1.9878349563690916e-05, "loss": 24.5157, "step": 6360 }, { "epoch": 0.08, "learning_rate": 1.987772937606685e-05, "loss": 24.8191, "step": 6370 }, { "epoch": 0.08, "learning_rate": 1.987710762129066e-05, "loss": 23.7717, "step": 6380 }, { "epoch": 0.08, "learning_rate": 1.987648429946099e-05, "loss": 24.3967, "step": 6390 }, { "epoch": 0.08, "learning_rate": 1.987585941067673e-05, "loss": 24.1426, "step": 6400 }, { "epoch": 0.08, "learning_rate": 1.9875232955037027e-05, "loss": 24.0838, "step": 6410 }, { "epoch": 0.08, "learning_rate": 1.987460493264127e-05, "loss": 25.1567, "step": 6420 }, { "epoch": 0.08, "learning_rate": 1.9873975343589094e-05, "loss": 24.4442, "step": 6430 }, { "epoch": 0.08, "learning_rate": 1.987334418798039e-05, "loss": 24.7934, "step": 6440 }, { "epoch": 0.08, "learning_rate": 1.9872711465915298e-05, "loss": 24.2411, "step": 6450 }, { "epoch": 0.08, "learning_rate": 1.9872077177494195e-05, "loss": 24.7736, "step": 6460 }, { "epoch": 0.08, "learning_rate": 1.9871441322817717e-05, "loss": 24.2621, "step": 6470 }, { "epoch": 0.08, "learning_rate": 1.9870803901986745e-05, "loss": 24.2924, "step": 6480 }, { "epoch": 0.08, "learning_rate": 1.9870164915102412e-05, "loss": 24.2827, "step": 6490 }, { "epoch": 0.08, "learning_rate": 1.9869524362266094e-05, "loss": 24.4326, "step": 6500 }, { "epoch": 0.08, "learning_rate": 1.9868882243579414e-05, "loss": 24.2781, "step": 6510 }, { "epoch": 0.08, "learning_rate": 1.9868238559144254e-05, "loss": 24.2697, "step": 6520 }, { "epoch": 0.08, "learning_rate": 1.986759330906274e-05, "loss": 23.9032, "step": 6530 }, { "epoch": 0.08, "learning_rate": 1.9866946493437234e-05, "loss": 25.1336, "step": 6540 }, { "epoch": 0.08, "learning_rate": 1.9866298112370367e-05, "loss": 24.3768, "step": 6550 }, { "epoch": 0.08, "learning_rate": 1.9865648165965e-05, "loss": 24.638, "step": 6560 }, { "epoch": 0.08, "learning_rate": 1.9864996654324257e-05, "loss": 24.1929, "step": 6570 }, { "epoch": 0.08, "learning_rate": 1.98643435775515e-05, "loss": 24.4508, "step": 6580 }, { "epoch": 0.08, "learning_rate": 1.986368893575034e-05, "loss": 24.5969, "step": 6590 }, { "epoch": 0.08, "learning_rate": 1.9863032729024643e-05, "loss": 24.4962, "step": 6600 }, { "epoch": 0.08, "learning_rate": 1.986237495747852e-05, "loss": 24.4151, "step": 6610 }, { "epoch": 0.08, "learning_rate": 1.9861715621216326e-05, "loss": 24.3736, "step": 6620 }, { "epoch": 0.08, "learning_rate": 1.9861054720342673e-05, "loss": 23.7175, "step": 6630 }, { "epoch": 0.08, "learning_rate": 1.9860392254962418e-05, "loss": 23.9924, "step": 6640 }, { "epoch": 0.08, "learning_rate": 1.9859728225180655e-05, "loss": 24.4977, "step": 6650 }, { "epoch": 0.08, "learning_rate": 1.9859062631102745e-05, "loss": 24.2849, "step": 6660 }, { "epoch": 0.08, "learning_rate": 1.9858395472834285e-05, "loss": 24.2635, "step": 6670 }, { "epoch": 0.08, "learning_rate": 1.9857726750481118e-05, "loss": 24.4326, "step": 6680 }, { "epoch": 0.08, "learning_rate": 1.985705646414935e-05, "loss": 24.2433, "step": 6690 }, { "epoch": 0.08, "learning_rate": 1.9856384613945314e-05, "loss": 24.3373, "step": 6700 }, { "epoch": 0.08, "learning_rate": 1.9855711199975613e-05, "loss": 24.3982, "step": 6710 }, { "epoch": 0.08, "learning_rate": 1.985503622234708e-05, "loss": 23.9251, "step": 6720 }, { "epoch": 0.08, "learning_rate": 1.9854359681166812e-05, "loss": 23.7173, "step": 6730 }, { "epoch": 0.08, "learning_rate": 1.985368157654213e-05, "loss": 23.8503, "step": 6740 }, { "epoch": 0.08, "learning_rate": 1.985300190858064e-05, "loss": 24.285, "step": 6750 }, { "epoch": 0.08, "learning_rate": 1.985232067739016e-05, "loss": 23.5844, "step": 6760 }, { "epoch": 0.08, "learning_rate": 1.9851637883078772e-05, "loss": 24.7652, "step": 6770 }, { "epoch": 0.08, "learning_rate": 1.985095352575481e-05, "loss": 23.9387, "step": 6780 }, { "epoch": 0.08, "learning_rate": 1.9850267605526847e-05, "loss": 24.1968, "step": 6790 }, { "epoch": 0.08, "learning_rate": 1.9849580122503708e-05, "loss": 23.4232, "step": 6800 }, { "epoch": 0.08, "learning_rate": 1.984889107679447e-05, "loss": 23.8916, "step": 6810 }, { "epoch": 0.08, "learning_rate": 1.984820046850845e-05, "loss": 24.5945, "step": 6820 }, { "epoch": 0.08, "learning_rate": 1.9847508297755213e-05, "loss": 24.7049, "step": 6830 }, { "epoch": 0.08, "learning_rate": 1.984681456464458e-05, "loss": 23.4752, "step": 6840 }, { "epoch": 0.08, "learning_rate": 1.9846119269286616e-05, "loss": 24.3581, "step": 6850 }, { "epoch": 0.08, "learning_rate": 1.9845422411791632e-05, "loss": 23.9373, "step": 6860 }, { "epoch": 0.08, "learning_rate": 1.984472399227019e-05, "loss": 24.2455, "step": 6870 }, { "epoch": 0.08, "learning_rate": 1.984402401083309e-05, "loss": 24.4249, "step": 6880 }, { "epoch": 0.08, "learning_rate": 1.9843322467591396e-05, "loss": 24.5095, "step": 6890 }, { "epoch": 0.08, "learning_rate": 1.984261936265641e-05, "loss": 24.9538, "step": 6900 }, { "epoch": 0.08, "learning_rate": 1.9841914696139678e-05, "loss": 23.6822, "step": 6910 }, { "epoch": 0.09, "learning_rate": 1.9841208468153007e-05, "loss": 23.9692, "step": 6920 }, { "epoch": 0.09, "learning_rate": 1.9840500678808437e-05, "loss": 24.4303, "step": 6930 }, { "epoch": 0.09, "learning_rate": 1.9839791328218262e-05, "loss": 24.3141, "step": 6940 }, { "epoch": 0.09, "learning_rate": 1.983908041649503e-05, "loss": 24.6573, "step": 6950 }, { "epoch": 0.09, "learning_rate": 1.983836794375153e-05, "loss": 24.2943, "step": 6960 }, { "epoch": 0.09, "learning_rate": 1.983765391010079e-05, "loss": 23.8802, "step": 6970 }, { "epoch": 0.09, "learning_rate": 1.9836938315656108e-05, "loss": 24.6169, "step": 6980 }, { "epoch": 0.09, "learning_rate": 1.983622116053101e-05, "loss": 24.7538, "step": 6990 }, { "epoch": 0.09, "learning_rate": 1.9835502444839277e-05, "loss": 23.9907, "step": 7000 }, { "epoch": 0.09, "learning_rate": 1.9834782168694935e-05, "loss": 24.517, "step": 7010 }, { "epoch": 0.09, "learning_rate": 1.9834060332212262e-05, "loss": 24.7551, "step": 7020 }, { "epoch": 0.09, "learning_rate": 1.983333693550578e-05, "loss": 23.9245, "step": 7030 }, { "epoch": 0.09, "learning_rate": 1.983261197869026e-05, "loss": 24.3435, "step": 7040 }, { "epoch": 0.09, "learning_rate": 1.983188546188072e-05, "loss": 24.0436, "step": 7050 }, { "epoch": 0.09, "learning_rate": 1.983115738519243e-05, "loss": 24.3236, "step": 7060 }, { "epoch": 0.09, "learning_rate": 1.9830427748740896e-05, "loss": 23.6828, "step": 7070 }, { "epoch": 0.09, "learning_rate": 1.9829696552641883e-05, "loss": 23.2498, "step": 7080 }, { "epoch": 0.09, "learning_rate": 1.9828963797011394e-05, "loss": 24.0547, "step": 7090 }, { "epoch": 0.09, "learning_rate": 1.9828229481965688e-05, "loss": 24.695, "step": 7100 }, { "epoch": 0.09, "learning_rate": 1.982749360762127e-05, "loss": 24.4125, "step": 7110 }, { "epoch": 0.09, "learning_rate": 1.9826756174094884e-05, "loss": 24.4779, "step": 7120 }, { "epoch": 0.09, "learning_rate": 1.9826017181503537e-05, "loss": 24.2115, "step": 7130 }, { "epoch": 0.09, "learning_rate": 1.9825276629964463e-05, "loss": 24.6113, "step": 7140 }, { "epoch": 0.09, "learning_rate": 1.9824534519595163e-05, "loss": 24.6144, "step": 7150 }, { "epoch": 0.09, "learning_rate": 1.9823790850513373e-05, "loss": 24.4142, "step": 7160 }, { "epoch": 0.09, "learning_rate": 1.9823045622837078e-05, "loss": 24.5967, "step": 7170 }, { "epoch": 0.09, "learning_rate": 1.9822298836684518e-05, "loss": 24.22, "step": 7180 }, { "epoch": 0.09, "learning_rate": 1.982155049217417e-05, "loss": 24.156, "step": 7190 }, { "epoch": 0.09, "learning_rate": 1.982080058942476e-05, "loss": 24.4925, "step": 7200 }, { "epoch": 0.09, "learning_rate": 1.982004912855527e-05, "loss": 24.0989, "step": 7210 }, { "epoch": 0.09, "learning_rate": 1.981929610968492e-05, "loss": 24.5381, "step": 7220 }, { "epoch": 0.09, "learning_rate": 1.981854153293318e-05, "loss": 24.0206, "step": 7230 }, { "epoch": 0.09, "learning_rate": 1.981778539841977e-05, "loss": 24.2044, "step": 7240 }, { "epoch": 0.09, "learning_rate": 1.9817027706264647e-05, "loss": 24.1859, "step": 7250 }, { "epoch": 0.09, "learning_rate": 1.9816268456588036e-05, "loss": 24.2025, "step": 7260 }, { "epoch": 0.09, "learning_rate": 1.9815507649510383e-05, "loss": 24.286, "step": 7270 }, { "epoch": 0.09, "learning_rate": 1.98147452851524e-05, "loss": 24.6235, "step": 7280 }, { "epoch": 0.09, "learning_rate": 1.981398136363504e-05, "loss": 24.839, "step": 7290 }, { "epoch": 0.09, "learning_rate": 1.9813215885079505e-05, "loss": 24.4451, "step": 7300 }, { "epoch": 0.09, "learning_rate": 1.9812448849607238e-05, "loss": 24.9785, "step": 7310 }, { "epoch": 0.09, "learning_rate": 1.981168025733993e-05, "loss": 24.0538, "step": 7320 }, { "epoch": 0.09, "learning_rate": 1.9810910108399534e-05, "loss": 24.0802, "step": 7330 }, { "epoch": 0.09, "learning_rate": 1.9810138402908222e-05, "loss": 24.9676, "step": 7340 }, { "epoch": 0.09, "learning_rate": 1.9809365140988442e-05, "loss": 24.5075, "step": 7350 }, { "epoch": 0.09, "learning_rate": 1.980859032276287e-05, "loss": 24.7303, "step": 7360 }, { "epoch": 0.09, "learning_rate": 1.980781394835444e-05, "loss": 24.2503, "step": 7370 }, { "epoch": 0.09, "learning_rate": 1.9807036017886324e-05, "loss": 24.3082, "step": 7380 }, { "epoch": 0.09, "learning_rate": 1.9806256531481943e-05, "loss": 24.7114, "step": 7390 }, { "epoch": 0.09, "learning_rate": 1.9805475489264967e-05, "loss": 24.3804, "step": 7400 }, { "epoch": 0.09, "learning_rate": 1.9804692891359315e-05, "loss": 24.7874, "step": 7410 }, { "epoch": 0.09, "learning_rate": 1.9803908737889147e-05, "loss": 24.2182, "step": 7420 }, { "epoch": 0.09, "learning_rate": 1.9803123028978877e-05, "loss": 24.216, "step": 7430 }, { "epoch": 0.09, "learning_rate": 1.9802335764753157e-05, "loss": 24.0453, "step": 7440 }, { "epoch": 0.09, "learning_rate": 1.9801546945336897e-05, "loss": 24.2269, "step": 7450 }, { "epoch": 0.09, "learning_rate": 1.9800756570855237e-05, "loss": 24.7872, "step": 7460 }, { "epoch": 0.09, "learning_rate": 1.9799964641433584e-05, "loss": 23.9532, "step": 7470 }, { "epoch": 0.09, "learning_rate": 1.9799171157197578e-05, "loss": 24.3048, "step": 7480 }, { "epoch": 0.09, "learning_rate": 1.979837611827311e-05, "loss": 24.1784, "step": 7490 }, { "epoch": 0.09, "learning_rate": 1.979757952478631e-05, "loss": 24.0368, "step": 7500 }, { "epoch": 0.09, "learning_rate": 1.9796781376863574e-05, "loss": 24.1054, "step": 7510 }, { "epoch": 0.09, "learning_rate": 1.979598167463152e-05, "loss": 24.3187, "step": 7520 }, { "epoch": 0.09, "learning_rate": 1.9795180418217038e-05, "loss": 25.0973, "step": 7530 }, { "epoch": 0.09, "learning_rate": 1.9794377607747236e-05, "loss": 23.6455, "step": 7540 }, { "epoch": 0.09, "learning_rate": 1.97935732433495e-05, "loss": 24.3119, "step": 7550 }, { "epoch": 0.09, "learning_rate": 1.9792767325151433e-05, "loss": 23.7743, "step": 7560 }, { "epoch": 0.09, "learning_rate": 1.9791959853280907e-05, "loss": 24.4069, "step": 7570 }, { "epoch": 0.09, "learning_rate": 1.9791150827866023e-05, "loss": 24.6152, "step": 7580 }, { "epoch": 0.09, "learning_rate": 1.9790340249035146e-05, "loss": 24.4936, "step": 7590 }, { "epoch": 0.09, "learning_rate": 1.9789528116916877e-05, "loss": 24.2301, "step": 7600 }, { "epoch": 0.09, "learning_rate": 1.978871443164006e-05, "loss": 24.3323, "step": 7610 }, { "epoch": 0.09, "learning_rate": 1.978789919333379e-05, "loss": 24.2113, "step": 7620 }, { "epoch": 0.09, "learning_rate": 1.9787082402127414e-05, "loss": 24.8747, "step": 7630 }, { "epoch": 0.09, "learning_rate": 1.9786264058150518e-05, "loss": 23.8486, "step": 7640 }, { "epoch": 0.09, "learning_rate": 1.9785444161532936e-05, "loss": 24.5549, "step": 7650 }, { "epoch": 0.09, "learning_rate": 1.978462271240475e-05, "loss": 24.2642, "step": 7660 }, { "epoch": 0.09, "learning_rate": 1.9783799710896283e-05, "loss": 24.3243, "step": 7670 }, { "epoch": 0.09, "learning_rate": 1.9782975157138112e-05, "loss": 24.5208, "step": 7680 }, { "epoch": 0.09, "learning_rate": 1.9782149051261057e-05, "loss": 24.4625, "step": 7690 }, { "epoch": 0.09, "learning_rate": 1.9781321393396183e-05, "loss": 23.7841, "step": 7700 }, { "epoch": 0.09, "learning_rate": 1.97804921836748e-05, "loss": 24.0014, "step": 7710 }, { "epoch": 0.09, "learning_rate": 1.977966142222847e-05, "loss": 24.1787, "step": 7720 }, { "epoch": 0.1, "learning_rate": 1.9778829109189e-05, "loss": 24.2934, "step": 7730 }, { "epoch": 0.1, "learning_rate": 1.977799524468843e-05, "loss": 24.1074, "step": 7740 }, { "epoch": 0.1, "learning_rate": 1.9777159828859068e-05, "loss": 23.7783, "step": 7750 }, { "epoch": 0.1, "learning_rate": 1.977632286183345e-05, "loss": 23.9655, "step": 7760 }, { "epoch": 0.1, "learning_rate": 1.9775484343744368e-05, "loss": 23.7111, "step": 7770 }, { "epoch": 0.1, "learning_rate": 1.9774644274724856e-05, "loss": 24.5512, "step": 7780 }, { "epoch": 0.1, "learning_rate": 1.9773802654908198e-05, "loss": 24.082, "step": 7790 }, { "epoch": 0.1, "learning_rate": 1.9772959484427917e-05, "loss": 24.1064, "step": 7800 }, { "epoch": 0.1, "learning_rate": 1.977211476341779e-05, "loss": 24.0584, "step": 7810 }, { "epoch": 0.1, "learning_rate": 1.9771268492011834e-05, "loss": 24.5307, "step": 7820 }, { "epoch": 0.1, "learning_rate": 1.9770420670344315e-05, "loss": 24.5208, "step": 7830 }, { "epoch": 0.1, "learning_rate": 1.9769571298549744e-05, "loss": 24.2229, "step": 7840 }, { "epoch": 0.1, "learning_rate": 1.976872037676288e-05, "loss": 24.4524, "step": 7850 }, { "epoch": 0.1, "learning_rate": 1.9767867905118724e-05, "loss": 24.4688, "step": 7860 }, { "epoch": 0.1, "learning_rate": 1.9767013883752525e-05, "loss": 24.0577, "step": 7870 }, { "epoch": 0.1, "learning_rate": 1.9766158312799776e-05, "loss": 24.2318, "step": 7880 }, { "epoch": 0.1, "learning_rate": 1.9765301192396227e-05, "loss": 24.386, "step": 7890 }, { "epoch": 0.1, "learning_rate": 1.976444252267785e-05, "loss": 24.6232, "step": 7900 }, { "epoch": 0.1, "learning_rate": 1.976358230378089e-05, "loss": 23.8479, "step": 7910 }, { "epoch": 0.1, "learning_rate": 1.976272053584182e-05, "loss": 23.8426, "step": 7920 }, { "epoch": 0.1, "learning_rate": 1.976185721899736e-05, "loss": 24.1812, "step": 7930 }, { "epoch": 0.1, "learning_rate": 1.9760992353384484e-05, "loss": 24.334, "step": 7940 }, { "epoch": 0.1, "learning_rate": 1.976012593914041e-05, "loss": 24.83, "step": 7950 }, { "epoch": 0.1, "learning_rate": 1.9759257976402595e-05, "loss": 24.3408, "step": 7960 }, { "epoch": 0.1, "learning_rate": 1.9758388465308745e-05, "loss": 24.3164, "step": 7970 }, { "epoch": 0.1, "learning_rate": 1.9757517405996815e-05, "loss": 24.2236, "step": 7980 }, { "epoch": 0.1, "learning_rate": 1.9756644798605004e-05, "loss": 23.7576, "step": 7990 }, { "epoch": 0.1, "learning_rate": 1.975577064327175e-05, "loss": 24.4664, "step": 8000 }, { "epoch": 0.1, "learning_rate": 1.9754894940135746e-05, "loss": 23.7182, "step": 8010 }, { "epoch": 0.1, "learning_rate": 1.9754017689335932e-05, "loss": 23.5912, "step": 8020 }, { "epoch": 0.1, "learning_rate": 1.9753138891011477e-05, "loss": 24.1121, "step": 8030 }, { "epoch": 0.1, "learning_rate": 1.9752258545301816e-05, "loss": 24.1152, "step": 8040 }, { "epoch": 0.1, "learning_rate": 1.9751376652346615e-05, "loss": 24.5118, "step": 8050 }, { "epoch": 0.1, "learning_rate": 1.9750493212285794e-05, "loss": 24.6393, "step": 8060 }, { "epoch": 0.1, "learning_rate": 1.974960822525951e-05, "loss": 23.8714, "step": 8070 }, { "epoch": 0.1, "learning_rate": 1.974872169140818e-05, "loss": 23.6483, "step": 8080 }, { "epoch": 0.1, "learning_rate": 1.974783361087245e-05, "loss": 24.5605, "step": 8090 }, { "epoch": 0.1, "learning_rate": 1.974694398379322e-05, "loss": 24.1174, "step": 8100 }, { "epoch": 0.1, "learning_rate": 1.974605281031164e-05, "loss": 24.1467, "step": 8110 }, { "epoch": 0.1, "learning_rate": 1.9745160090569085e-05, "loss": 24.5486, "step": 8120 }, { "epoch": 0.1, "learning_rate": 1.9744265824707203e-05, "loss": 24.3394, "step": 8130 }, { "epoch": 0.1, "learning_rate": 1.974337001286787e-05, "loss": 23.996, "step": 8140 }, { "epoch": 0.1, "learning_rate": 1.9742472655193205e-05, "loss": 24.4756, "step": 8150 }, { "epoch": 0.1, "learning_rate": 1.9741573751825587e-05, "loss": 24.8103, "step": 8160 }, { "epoch": 0.1, "learning_rate": 1.9740673302907627e-05, "loss": 24.1702, "step": 8170 }, { "epoch": 0.1, "learning_rate": 1.9739771308582188e-05, "loss": 24.092, "step": 8180 }, { "epoch": 0.1, "learning_rate": 1.9738867768992374e-05, "loss": 24.973, "step": 8190 }, { "epoch": 0.1, "learning_rate": 1.9737962684281535e-05, "loss": 23.8881, "step": 8200 }, { "epoch": 0.1, "learning_rate": 1.973705605459327e-05, "loss": 23.8866, "step": 8210 }, { "epoch": 0.1, "learning_rate": 1.9736147880071424e-05, "loss": 24.789, "step": 8220 }, { "epoch": 0.1, "learning_rate": 1.9735238160860075e-05, "loss": 23.8004, "step": 8230 }, { "epoch": 0.1, "learning_rate": 1.9734326897103558e-05, "loss": 24.342, "step": 8240 }, { "epoch": 0.1, "learning_rate": 1.9733414088946454e-05, "loss": 24.4173, "step": 8250 }, { "epoch": 0.1, "learning_rate": 1.9732499736533578e-05, "loss": 24.298, "step": 8260 }, { "epoch": 0.1, "learning_rate": 1.9731583840010003e-05, "loss": 24.4053, "step": 8270 }, { "epoch": 0.1, "learning_rate": 1.9730666399521034e-05, "loss": 24.2068, "step": 8280 }, { "epoch": 0.1, "learning_rate": 1.9729747415212234e-05, "loss": 24.2148, "step": 8290 }, { "epoch": 0.1, "learning_rate": 1.9728826887229398e-05, "loss": 24.1363, "step": 8300 }, { "epoch": 0.1, "learning_rate": 1.972790481571858e-05, "loss": 24.2691, "step": 8310 }, { "epoch": 0.1, "learning_rate": 1.9726981200826064e-05, "loss": 24.0531, "step": 8320 }, { "epoch": 0.1, "learning_rate": 1.9726056042698393e-05, "loss": 24.4429, "step": 8330 }, { "epoch": 0.1, "learning_rate": 1.9725129341482344e-05, "loss": 24.4376, "step": 8340 }, { "epoch": 0.1, "learning_rate": 1.9724201097324942e-05, "loss": 24.4666, "step": 8350 }, { "epoch": 0.1, "learning_rate": 1.972327131037346e-05, "loss": 24.5694, "step": 8360 }, { "epoch": 0.1, "learning_rate": 1.9722339980775414e-05, "loss": 24.7227, "step": 8370 }, { "epoch": 0.1, "learning_rate": 1.972140710867856e-05, "loss": 24.7042, "step": 8380 }, { "epoch": 0.1, "learning_rate": 1.972047269423091e-05, "loss": 24.7787, "step": 8390 }, { "epoch": 0.1, "learning_rate": 1.9719536737580703e-05, "loss": 24.3637, "step": 8400 }, { "epoch": 0.1, "learning_rate": 1.9718599238876442e-05, "loss": 24.0146, "step": 8410 }, { "epoch": 0.1, "learning_rate": 1.971766019826687e-05, "loss": 24.6968, "step": 8420 }, { "epoch": 0.1, "learning_rate": 1.9716719615900955e-05, "loss": 23.8031, "step": 8430 }, { "epoch": 0.1, "learning_rate": 1.971577749192794e-05, "loss": 24.3133, "step": 8440 }, { "epoch": 0.1, "learning_rate": 1.971483382649729e-05, "loss": 24.0775, "step": 8450 }, { "epoch": 0.1, "learning_rate": 1.9713888619758722e-05, "loss": 23.9682, "step": 8460 }, { "epoch": 0.1, "learning_rate": 1.9712941871862205e-05, "loss": 24.3013, "step": 8470 }, { "epoch": 0.1, "learning_rate": 1.971199358295794e-05, "loss": 24.3544, "step": 8480 }, { "epoch": 0.1, "learning_rate": 1.9711043753196377e-05, "loss": 24.0879, "step": 8490 }, { "epoch": 0.1, "learning_rate": 1.9710092382728213e-05, "loss": 24.2709, "step": 8500 }, { "epoch": 0.1, "learning_rate": 1.970913947170439e-05, "loss": 24.8152, "step": 8510 }, { "epoch": 0.1, "learning_rate": 1.970818502027609e-05, "loss": 24.6316, "step": 8520 }, { "epoch": 0.1, "learning_rate": 1.9707229028594737e-05, "loss": 24.6137, "step": 8530 }, { "epoch": 0.11, "learning_rate": 1.9706271496812012e-05, "loss": 24.0025, "step": 8540 }, { "epoch": 0.11, "learning_rate": 1.9705312425079827e-05, "loss": 24.2688, "step": 8550 }, { "epoch": 0.11, "learning_rate": 1.9704351813550348e-05, "loss": 24.5089, "step": 8560 }, { "epoch": 0.11, "learning_rate": 1.9703389662375975e-05, "loss": 24.1787, "step": 8570 }, { "epoch": 0.11, "learning_rate": 1.970242597170936e-05, "loss": 23.9267, "step": 8580 }, { "epoch": 0.11, "learning_rate": 1.9701460741703406e-05, "loss": 23.7045, "step": 8590 }, { "epoch": 0.11, "learning_rate": 1.9700493972511242e-05, "loss": 24.0935, "step": 8600 }, { "epoch": 0.11, "learning_rate": 1.969952566428625e-05, "loss": 24.7564, "step": 8610 }, { "epoch": 0.11, "learning_rate": 1.969855581718206e-05, "loss": 24.1325, "step": 8620 }, { "epoch": 0.11, "learning_rate": 1.969758443135255e-05, "loss": 24.6456, "step": 8630 }, { "epoch": 0.11, "learning_rate": 1.9696611506951825e-05, "loss": 24.6496, "step": 8640 }, { "epoch": 0.11, "learning_rate": 1.9695637044134245e-05, "loss": 24.3498, "step": 8650 }, { "epoch": 0.11, "learning_rate": 1.9694661043054423e-05, "loss": 24.3761, "step": 8660 }, { "epoch": 0.11, "learning_rate": 1.9693683503867196e-05, "loss": 24.2501, "step": 8670 }, { "epoch": 0.11, "learning_rate": 1.9692704426727663e-05, "loss": 23.9667, "step": 8680 }, { "epoch": 0.11, "learning_rate": 1.9691723811791155e-05, "loss": 23.617, "step": 8690 }, { "epoch": 0.11, "learning_rate": 1.9690741659213254e-05, "loss": 24.6176, "step": 8700 }, { "epoch": 0.11, "learning_rate": 1.968975796914978e-05, "loss": 25.1035, "step": 8710 }, { "epoch": 0.11, "learning_rate": 1.9688772741756805e-05, "loss": 25.0263, "step": 8720 }, { "epoch": 0.11, "learning_rate": 1.9687785977190643e-05, "loss": 23.9892, "step": 8730 }, { "epoch": 0.11, "learning_rate": 1.968679767560784e-05, "loss": 25.1092, "step": 8740 }, { "epoch": 0.11, "learning_rate": 1.96858078371652e-05, "loss": 23.9069, "step": 8750 }, { "epoch": 0.11, "learning_rate": 1.968481646201976e-05, "loss": 24.299, "step": 8760 }, { "epoch": 0.11, "learning_rate": 1.968382355032882e-05, "loss": 24.3922, "step": 8770 }, { "epoch": 0.11, "learning_rate": 1.9682829102249904e-05, "loss": 24.2753, "step": 8780 }, { "epoch": 0.11, "learning_rate": 1.968183311794078e-05, "loss": 24.9726, "step": 8790 }, { "epoch": 0.11, "learning_rate": 1.9680835597559476e-05, "loss": 24.7244, "step": 8800 }, { "epoch": 0.11, "learning_rate": 1.9679836541264244e-05, "loss": 24.1956, "step": 8810 }, { "epoch": 0.11, "learning_rate": 1.9678835949213596e-05, "loss": 24.6243, "step": 8820 }, { "epoch": 0.11, "learning_rate": 1.9677833821566282e-05, "loss": 23.9354, "step": 8830 }, { "epoch": 0.11, "learning_rate": 1.9676830158481287e-05, "loss": 24.8383, "step": 8840 }, { "epoch": 0.11, "learning_rate": 1.9675824960117857e-05, "loss": 24.0802, "step": 8850 }, { "epoch": 0.11, "learning_rate": 1.9674818226635465e-05, "loss": 24.4001, "step": 8860 }, { "epoch": 0.11, "learning_rate": 1.9673809958193834e-05, "loss": 24.3385, "step": 8870 }, { "epoch": 0.11, "learning_rate": 1.9672800154952937e-05, "loss": 24.3111, "step": 8880 }, { "epoch": 0.11, "learning_rate": 1.967178881707298e-05, "loss": 24.4698, "step": 8890 }, { "epoch": 0.11, "learning_rate": 1.9670775944714416e-05, "loss": 24.0014, "step": 8900 }, { "epoch": 0.11, "learning_rate": 1.9669761538037946e-05, "loss": 24.0328, "step": 8910 }, { "epoch": 0.11, "learning_rate": 1.966874559720451e-05, "loss": 24.1724, "step": 8920 }, { "epoch": 0.11, "learning_rate": 1.9667728122375286e-05, "loss": 24.1869, "step": 8930 }, { "epoch": 0.11, "learning_rate": 1.966670911371171e-05, "loss": 24.0289, "step": 8940 }, { "epoch": 0.11, "learning_rate": 1.9665688571375447e-05, "loss": 24.219, "step": 8950 }, { "epoch": 0.11, "learning_rate": 1.9664666495528418e-05, "loss": 24.5106, "step": 8960 }, { "epoch": 0.11, "learning_rate": 1.9663642886332773e-05, "loss": 24.0919, "step": 8970 }, { "epoch": 0.11, "learning_rate": 1.9662617743950918e-05, "loss": 23.8567, "step": 8980 }, { "epoch": 0.11, "learning_rate": 1.9661591068545494e-05, "loss": 24.0292, "step": 8990 }, { "epoch": 0.11, "learning_rate": 1.966056286027939e-05, "loss": 24.3611, "step": 9000 }, { "epoch": 0.11, "learning_rate": 1.9659533119315738e-05, "loss": 24.4284, "step": 9010 }, { "epoch": 0.11, "learning_rate": 1.9658501845817907e-05, "loss": 24.0667, "step": 9020 }, { "epoch": 0.11, "learning_rate": 1.9657469039949517e-05, "loss": 24.1714, "step": 9030 }, { "epoch": 0.11, "learning_rate": 1.9656434701874428e-05, "loss": 23.7838, "step": 9040 }, { "epoch": 0.11, "learning_rate": 1.965539883175674e-05, "loss": 24.3806, "step": 9050 }, { "epoch": 0.11, "learning_rate": 1.9654361429760803e-05, "loss": 24.234, "step": 9060 }, { "epoch": 0.11, "learning_rate": 1.9653322496051202e-05, "loss": 23.3207, "step": 9070 }, { "epoch": 0.11, "learning_rate": 1.9652282030792777e-05, "loss": 25.0555, "step": 9080 }, { "epoch": 0.11, "learning_rate": 1.9651240034150595e-05, "loss": 24.1155, "step": 9090 }, { "epoch": 0.11, "learning_rate": 1.965019650628998e-05, "loss": 24.1833, "step": 9100 }, { "epoch": 0.11, "learning_rate": 1.9649151447376485e-05, "loss": 24.6365, "step": 9110 }, { "epoch": 0.11, "learning_rate": 1.964810485757592e-05, "loss": 24.6101, "step": 9120 }, { "epoch": 0.11, "learning_rate": 1.964705673705433e-05, "loss": 24.189, "step": 9130 }, { "epoch": 0.11, "learning_rate": 1.964600708597801e-05, "loss": 24.4473, "step": 9140 }, { "epoch": 0.11, "learning_rate": 1.9644955904513483e-05, "loss": 24.9149, "step": 9150 }, { "epoch": 0.11, "learning_rate": 1.964390319282753e-05, "loss": 24.6982, "step": 9160 }, { "epoch": 0.11, "learning_rate": 1.964284895108717e-05, "loss": 24.0765, "step": 9170 }, { "epoch": 0.11, "learning_rate": 1.9641793179459657e-05, "loss": 24.5518, "step": 9180 }, { "epoch": 0.11, "learning_rate": 1.9640735878112506e-05, "loss": 23.9584, "step": 9190 }, { "epoch": 0.11, "learning_rate": 1.9639677047213452e-05, "loss": 24.4648, "step": 9200 }, { "epoch": 0.11, "learning_rate": 1.9638616686930492e-05, "loss": 24.1059, "step": 9210 }, { "epoch": 0.11, "learning_rate": 1.963755479743185e-05, "loss": 24.85, "step": 9220 }, { "epoch": 0.11, "learning_rate": 1.963649137888601e-05, "loss": 23.6147, "step": 9230 }, { "epoch": 0.11, "learning_rate": 1.963542643146168e-05, "loss": 24.3874, "step": 9240 }, { "epoch": 0.11, "learning_rate": 1.963435995532782e-05, "loss": 23.9253, "step": 9250 }, { "epoch": 0.11, "learning_rate": 1.9633291950653638e-05, "loss": 24.1175, "step": 9260 }, { "epoch": 0.11, "learning_rate": 1.9632222417608573e-05, "loss": 23.8474, "step": 9270 }, { "epoch": 0.11, "learning_rate": 1.9631151356362318e-05, "loss": 23.7071, "step": 9280 }, { "epoch": 0.11, "learning_rate": 1.9630078767084794e-05, "loss": 23.9534, "step": 9290 }, { "epoch": 0.11, "learning_rate": 1.9629004649946177e-05, "loss": 24.2413, "step": 9300 }, { "epoch": 0.11, "learning_rate": 1.962792900511688e-05, "loss": 24.2047, "step": 9310 }, { "epoch": 0.11, "learning_rate": 1.9626851832767563e-05, "loss": 24.0172, "step": 9320 }, { "epoch": 0.11, "learning_rate": 1.962577313306912e-05, "loss": 24.2259, "step": 9330 }, { "epoch": 0.11, "learning_rate": 1.9624692906192696e-05, "loss": 23.9907, "step": 9340 }, { "epoch": 0.11, "learning_rate": 1.9623611152309676e-05, "loss": 24.2731, "step": 9350 }, { "epoch": 0.12, "learning_rate": 1.9622527871591676e-05, "loss": 24.2703, "step": 9360 }, { "epoch": 0.12, "learning_rate": 1.962144306421058e-05, "loss": 24.7612, "step": 9370 }, { "epoch": 0.12, "learning_rate": 1.962035673033848e-05, "loss": 24.0949, "step": 9380 }, { "epoch": 0.12, "learning_rate": 1.9619268870147744e-05, "loss": 24.0113, "step": 9390 }, { "epoch": 0.12, "learning_rate": 1.9618179483810956e-05, "loss": 24.3159, "step": 9400 }, { "epoch": 0.12, "learning_rate": 1.961708857150096e-05, "loss": 23.9427, "step": 9410 }, { "epoch": 0.12, "learning_rate": 1.9615996133390825e-05, "loss": 24.3918, "step": 9420 }, { "epoch": 0.12, "learning_rate": 1.9614902169653885e-05, "loss": 24.2726, "step": 9430 }, { "epoch": 0.12, "learning_rate": 1.9613806680463693e-05, "loss": 23.9878, "step": 9440 }, { "epoch": 0.12, "learning_rate": 1.961270966599406e-05, "loss": 23.8853, "step": 9450 }, { "epoch": 0.12, "learning_rate": 1.9611611126419028e-05, "loss": 23.9866, "step": 9460 }, { "epoch": 0.12, "learning_rate": 1.9610511061912885e-05, "loss": 23.7879, "step": 9470 }, { "epoch": 0.12, "learning_rate": 1.9609409472650172e-05, "loss": 24.4017, "step": 9480 }, { "epoch": 0.12, "learning_rate": 1.960830635880565e-05, "loss": 24.053, "step": 9490 }, { "epoch": 0.12, "learning_rate": 1.960720172055434e-05, "loss": 24.2895, "step": 9500 }, { "epoch": 0.12, "learning_rate": 1.9606095558071498e-05, "loss": 24.622, "step": 9510 }, { "epoch": 0.12, "learning_rate": 1.960498787153262e-05, "loss": 24.3525, "step": 9520 }, { "epoch": 0.12, "learning_rate": 1.9603878661113444e-05, "loss": 23.7417, "step": 9530 }, { "epoch": 0.12, "learning_rate": 1.9602767926989962e-05, "loss": 23.8594, "step": 9540 }, { "epoch": 0.12, "learning_rate": 1.9601655669338387e-05, "loss": 24.157, "step": 9550 }, { "epoch": 0.12, "learning_rate": 1.9600541888335187e-05, "loss": 23.9096, "step": 9560 }, { "epoch": 0.12, "learning_rate": 1.959942658415707e-05, "loss": 23.6637, "step": 9570 }, { "epoch": 0.12, "learning_rate": 1.9598309756980988e-05, "loss": 24.5131, "step": 9580 }, { "epoch": 0.12, "learning_rate": 1.9597191406984123e-05, "loss": 24.5173, "step": 9590 }, { "epoch": 0.12, "learning_rate": 1.9596071534343915e-05, "loss": 24.0625, "step": 9600 }, { "epoch": 0.12, "learning_rate": 1.9594950139238035e-05, "loss": 24.0444, "step": 9610 }, { "epoch": 0.12, "learning_rate": 1.9593827221844403e-05, "loss": 24.1635, "step": 9620 }, { "epoch": 0.12, "learning_rate": 1.9592702782341163e-05, "loss": 24.0005, "step": 9630 }, { "epoch": 0.12, "learning_rate": 1.9591576820906722e-05, "loss": 24.0392, "step": 9640 }, { "epoch": 0.12, "learning_rate": 1.9590449337719716e-05, "loss": 24.005, "step": 9650 }, { "epoch": 0.12, "learning_rate": 1.958932033295903e-05, "loss": 24.6618, "step": 9660 }, { "epoch": 0.12, "learning_rate": 1.9588189806803785e-05, "loss": 24.3144, "step": 9670 }, { "epoch": 0.12, "learning_rate": 1.9587057759433342e-05, "loss": 24.1702, "step": 9680 }, { "epoch": 0.12, "learning_rate": 1.958592419102731e-05, "loss": 24.1163, "step": 9690 }, { "epoch": 0.12, "learning_rate": 1.9584789101765533e-05, "loss": 23.866, "step": 9700 }, { "epoch": 0.12, "learning_rate": 1.9583652491828097e-05, "loss": 24.425, "step": 9710 }, { "epoch": 0.12, "learning_rate": 1.9582514361395334e-05, "loss": 24.3649, "step": 9720 }, { "epoch": 0.12, "learning_rate": 1.9581374710647813e-05, "loss": 23.875, "step": 9730 }, { "epoch": 0.12, "learning_rate": 1.9580233539766348e-05, "loss": 24.4151, "step": 9740 }, { "epoch": 0.12, "learning_rate": 1.9579090848931988e-05, "loss": 23.9801, "step": 9750 }, { "epoch": 0.12, "learning_rate": 1.9577946638326025e-05, "loss": 24.2562, "step": 9760 }, { "epoch": 0.12, "learning_rate": 1.957680090813e-05, "loss": 24.6437, "step": 9770 }, { "epoch": 0.12, "learning_rate": 1.9575653658525685e-05, "loss": 23.9744, "step": 9780 }, { "epoch": 0.12, "learning_rate": 1.9574504889695098e-05, "loss": 24.168, "step": 9790 }, { "epoch": 0.12, "learning_rate": 1.9573354601820496e-05, "loss": 24.2061, "step": 9800 }, { "epoch": 0.12, "learning_rate": 1.9572202795084377e-05, "loss": 23.9875, "step": 9810 }, { "epoch": 0.12, "learning_rate": 1.9571049469669486e-05, "loss": 24.8741, "step": 9820 }, { "epoch": 0.12, "learning_rate": 1.95698946257588e-05, "loss": 23.6988, "step": 9830 }, { "epoch": 0.12, "learning_rate": 1.956873826353554e-05, "loss": 24.3898, "step": 9840 }, { "epoch": 0.12, "learning_rate": 1.9567580383183174e-05, "loss": 23.6308, "step": 9850 }, { "epoch": 0.12, "learning_rate": 1.95664209848854e-05, "loss": 23.6741, "step": 9860 }, { "epoch": 0.12, "learning_rate": 1.956526006882616e-05, "loss": 23.9864, "step": 9870 }, { "epoch": 0.12, "learning_rate": 1.956409763518965e-05, "loss": 24.0213, "step": 9880 }, { "epoch": 0.12, "learning_rate": 1.9562933684160288e-05, "loss": 24.2737, "step": 9890 }, { "epoch": 0.12, "learning_rate": 1.956176821592274e-05, "loss": 24.3736, "step": 9900 }, { "epoch": 0.12, "learning_rate": 1.9560601230661918e-05, "loss": 24.4834, "step": 9910 }, { "epoch": 0.12, "learning_rate": 1.955943272856297e-05, "loss": 24.7612, "step": 9920 }, { "epoch": 0.12, "learning_rate": 1.9558262709811284e-05, "loss": 23.9508, "step": 9930 }, { "epoch": 0.12, "learning_rate": 1.9557091174592482e-05, "loss": 24.147, "step": 9940 }, { "epoch": 0.12, "learning_rate": 1.9555918123092447e-05, "loss": 24.3494, "step": 9950 }, { "epoch": 0.12, "learning_rate": 1.9554743555497277e-05, "loss": 24.5488, "step": 9960 }, { "epoch": 0.12, "learning_rate": 1.9553567471993333e-05, "loss": 24.4045, "step": 9970 }, { "epoch": 0.12, "learning_rate": 1.95523898727672e-05, "loss": 24.9827, "step": 9980 }, { "epoch": 0.12, "learning_rate": 1.9551210758005716e-05, "loss": 23.6511, "step": 9990 }, { "epoch": 0.12, "learning_rate": 1.955003012789595e-05, "loss": 24.5095, "step": 10000 }, { "epoch": 0.12, "learning_rate": 1.9548847982625216e-05, "loss": 24.0016, "step": 10010 }, { "epoch": 0.12, "learning_rate": 1.9547664322381066e-05, "loss": 24.5985, "step": 10020 }, { "epoch": 0.12, "learning_rate": 1.954647914735129e-05, "loss": 23.9498, "step": 10030 }, { "epoch": 0.12, "learning_rate": 1.9545292457723935e-05, "loss": 24.357, "step": 10040 }, { "epoch": 0.12, "learning_rate": 1.954410425368726e-05, "loss": 24.8405, "step": 10050 }, { "epoch": 0.12, "learning_rate": 1.954291453542979e-05, "loss": 24.4025, "step": 10060 }, { "epoch": 0.12, "learning_rate": 1.9541723303140277e-05, "loss": 24.0332, "step": 10070 }, { "epoch": 0.12, "learning_rate": 1.9540530557007717e-05, "loss": 23.6611, "step": 10080 }, { "epoch": 0.12, "learning_rate": 1.953933629722134e-05, "loss": 24.4029, "step": 10090 }, { "epoch": 0.12, "learning_rate": 1.953814052397063e-05, "loss": 24.3448, "step": 10100 }, { "epoch": 0.12, "learning_rate": 1.953694323744529e-05, "loss": 23.886, "step": 10110 }, { "epoch": 0.12, "learning_rate": 1.9535744437835288e-05, "loss": 23.8999, "step": 10120 }, { "epoch": 0.12, "learning_rate": 1.9534544125330818e-05, "loss": 24.0883, "step": 10130 }, { "epoch": 0.12, "learning_rate": 1.9533342300122306e-05, "loss": 24.335, "step": 10140 }, { "epoch": 0.12, "learning_rate": 1.953213896240044e-05, "loss": 23.5998, "step": 10150 }, { "epoch": 0.12, "learning_rate": 1.9530934112356128e-05, "loss": 24.0151, "step": 10160 }, { "epoch": 0.13, "learning_rate": 1.9529727750180528e-05, "loss": 24.1579, "step": 10170 }, { "epoch": 0.13, "learning_rate": 1.9528519876065035e-05, "loss": 24.0508, "step": 10180 }, { "epoch": 0.13, "learning_rate": 1.9527310490201286e-05, "loss": 23.8514, "step": 10190 }, { "epoch": 0.13, "learning_rate": 1.952609959278115e-05, "loss": 24.3171, "step": 10200 }, { "epoch": 0.13, "learning_rate": 1.9524887183996753e-05, "loss": 24.2038, "step": 10210 }, { "epoch": 0.13, "learning_rate": 1.952367326404044e-05, "loss": 23.8353, "step": 10220 }, { "epoch": 0.13, "learning_rate": 1.9522457833104812e-05, "loss": 24.8362, "step": 10230 }, { "epoch": 0.13, "learning_rate": 1.95212408913827e-05, "loss": 24.2924, "step": 10240 }, { "epoch": 0.13, "learning_rate": 1.9520022439067178e-05, "loss": 24.363, "step": 10250 }, { "epoch": 0.13, "learning_rate": 1.951880247635156e-05, "loss": 23.8473, "step": 10260 }, { "epoch": 0.13, "learning_rate": 1.9517581003429398e-05, "loss": 24.187, "step": 10270 }, { "epoch": 0.13, "learning_rate": 1.9516358020494487e-05, "loss": 24.5298, "step": 10280 }, { "epoch": 0.13, "learning_rate": 1.9515133527740865e-05, "loss": 24.3881, "step": 10290 }, { "epoch": 0.13, "learning_rate": 1.9513907525362792e-05, "loss": 25.0746, "step": 10300 }, { "epoch": 0.13, "learning_rate": 1.9512680013554786e-05, "loss": 24.0546, "step": 10310 }, { "epoch": 0.13, "learning_rate": 1.95114509925116e-05, "loss": 24.1847, "step": 10320 }, { "epoch": 0.13, "learning_rate": 1.9510220462428218e-05, "loss": 24.1006, "step": 10330 }, { "epoch": 0.13, "learning_rate": 1.9508988423499878e-05, "loss": 24.0113, "step": 10340 }, { "epoch": 0.13, "learning_rate": 1.9507754875922043e-05, "loss": 24.1273, "step": 10350 }, { "epoch": 0.13, "learning_rate": 1.9506519819890423e-05, "loss": 23.9727, "step": 10360 }, { "epoch": 0.13, "learning_rate": 1.9505283255600965e-05, "loss": 23.7439, "step": 10370 }, { "epoch": 0.13, "learning_rate": 1.9504045183249864e-05, "loss": 24.1511, "step": 10380 }, { "epoch": 0.13, "learning_rate": 1.9502805603033536e-05, "loss": 23.969, "step": 10390 }, { "epoch": 0.13, "learning_rate": 1.950156451514865e-05, "loss": 24.1265, "step": 10400 }, { "epoch": 0.13, "learning_rate": 1.9500321919792116e-05, "loss": 23.9264, "step": 10410 }, { "epoch": 0.13, "learning_rate": 1.949907781716107e-05, "loss": 23.2233, "step": 10420 }, { "epoch": 0.13, "learning_rate": 1.9497832207452906e-05, "loss": 24.2816, "step": 10430 }, { "epoch": 0.13, "learning_rate": 1.9496585090865234e-05, "loss": 23.6315, "step": 10440 }, { "epoch": 0.13, "learning_rate": 1.9495336467595924e-05, "loss": 23.8562, "step": 10450 }, { "epoch": 0.13, "learning_rate": 1.949408633784307e-05, "loss": 23.8334, "step": 10460 }, { "epoch": 0.13, "learning_rate": 1.9492834701805015e-05, "loss": 24.1295, "step": 10470 }, { "epoch": 0.13, "learning_rate": 1.9491581559680343e-05, "loss": 23.7352, "step": 10480 }, { "epoch": 0.13, "learning_rate": 1.949032691166786e-05, "loss": 24.1385, "step": 10490 }, { "epoch": 0.13, "learning_rate": 1.948907075796663e-05, "loss": 24.0507, "step": 10500 }, { "epoch": 0.13, "learning_rate": 1.9487813098775946e-05, "loss": 24.2542, "step": 10510 }, { "epoch": 0.13, "learning_rate": 1.9486553934295344e-05, "loss": 24.0617, "step": 10520 }, { "epoch": 0.13, "learning_rate": 1.9485293264724595e-05, "loss": 23.9526, "step": 10530 }, { "epoch": 0.13, "learning_rate": 1.948403109026371e-05, "loss": 24.2067, "step": 10540 }, { "epoch": 0.13, "learning_rate": 1.9482767411112942e-05, "loss": 24.0355, "step": 10550 }, { "epoch": 0.13, "learning_rate": 1.9481502227472777e-05, "loss": 24.084, "step": 10560 }, { "epoch": 0.13, "learning_rate": 1.9480235539543944e-05, "loss": 23.6645, "step": 10570 }, { "epoch": 0.13, "learning_rate": 1.9478967347527407e-05, "loss": 24.5305, "step": 10580 }, { "epoch": 0.13, "learning_rate": 1.9477697651624378e-05, "loss": 24.389, "step": 10590 }, { "epoch": 0.13, "learning_rate": 1.9476426452036296e-05, "loss": 23.8083, "step": 10600 }, { "epoch": 0.13, "learning_rate": 1.9475153748964842e-05, "loss": 24.6046, "step": 10610 }, { "epoch": 0.13, "learning_rate": 1.947387954261194e-05, "loss": 24.1979, "step": 10620 }, { "epoch": 0.13, "learning_rate": 1.947260383317975e-05, "loss": 23.9171, "step": 10630 }, { "epoch": 0.13, "learning_rate": 1.9471326620870667e-05, "loss": 23.8918, "step": 10640 }, { "epoch": 0.13, "learning_rate": 1.9470047905887328e-05, "loss": 23.8302, "step": 10650 }, { "epoch": 0.13, "learning_rate": 1.9468767688432603e-05, "loss": 23.7906, "step": 10660 }, { "epoch": 0.13, "learning_rate": 1.9467485968709615e-05, "loss": 23.995, "step": 10670 }, { "epoch": 0.13, "learning_rate": 1.946620274692171e-05, "loss": 23.8411, "step": 10680 }, { "epoch": 0.13, "learning_rate": 1.9464918023272477e-05, "loss": 23.9304, "step": 10690 }, { "epoch": 0.13, "learning_rate": 1.9463631797965744e-05, "loss": 23.7916, "step": 10700 }, { "epoch": 0.13, "learning_rate": 1.946234407120558e-05, "loss": 24.0259, "step": 10710 }, { "epoch": 0.13, "learning_rate": 1.9461054843196284e-05, "loss": 24.2306, "step": 10720 }, { "epoch": 0.13, "learning_rate": 1.9459764114142407e-05, "loss": 24.479, "step": 10730 }, { "epoch": 0.13, "learning_rate": 1.945847188424872e-05, "loss": 24.2391, "step": 10740 }, { "epoch": 0.13, "learning_rate": 1.9457178153720248e-05, "loss": 24.5699, "step": 10750 }, { "epoch": 0.13, "learning_rate": 1.9455882922762245e-05, "loss": 24.294, "step": 10760 }, { "epoch": 0.13, "learning_rate": 1.945458619158021e-05, "loss": 24.1578, "step": 10770 }, { "epoch": 0.13, "learning_rate": 1.9453287960379877e-05, "loss": 24.5461, "step": 10780 }, { "epoch": 0.13, "learning_rate": 1.9451988229367208e-05, "loss": 24.3996, "step": 10790 }, { "epoch": 0.13, "learning_rate": 1.945068699874842e-05, "loss": 23.6728, "step": 10800 }, { "epoch": 0.13, "learning_rate": 1.9449384268729958e-05, "loss": 23.8057, "step": 10810 }, { "epoch": 0.13, "learning_rate": 1.94480800395185e-05, "loss": 24.1203, "step": 10820 }, { "epoch": 0.13, "learning_rate": 1.944677431132098e-05, "loss": 24.2954, "step": 10830 }, { "epoch": 0.13, "learning_rate": 1.9445467084344554e-05, "loss": 23.8943, "step": 10840 }, { "epoch": 0.13, "learning_rate": 1.944415835879662e-05, "loss": 24.3214, "step": 10850 }, { "epoch": 0.13, "learning_rate": 1.944284813488481e-05, "loss": 23.8901, "step": 10860 }, { "epoch": 0.13, "learning_rate": 1.9441536412817003e-05, "loss": 24.5518, "step": 10870 }, { "epoch": 0.13, "learning_rate": 1.9440223192801308e-05, "loss": 24.5289, "step": 10880 }, { "epoch": 0.13, "learning_rate": 1.9438908475046075e-05, "loss": 24.4826, "step": 10890 }, { "epoch": 0.13, "learning_rate": 1.9437592259759892e-05, "loss": 24.13, "step": 10900 }, { "epoch": 0.13, "learning_rate": 1.943627454715158e-05, "loss": 24.6666, "step": 10910 }, { "epoch": 0.13, "learning_rate": 1.94349553374302e-05, "loss": 23.6091, "step": 10920 }, { "epoch": 0.13, "learning_rate": 1.9433634630805056e-05, "loss": 23.7818, "step": 10930 }, { "epoch": 0.13, "learning_rate": 1.9432312427485683e-05, "loss": 24.6152, "step": 10940 }, { "epoch": 0.13, "learning_rate": 1.9430988727681856e-05, "loss": 24.015, "step": 10950 }, { "epoch": 0.13, "learning_rate": 1.9429663531603583e-05, "loss": 24.0866, "step": 10960 }, { "epoch": 0.13, "learning_rate": 1.9428336839461113e-05, "loss": 24.2567, "step": 10970 }, { "epoch": 0.14, "learning_rate": 1.942700865146494e-05, "loss": 23.9772, "step": 10980 }, { "epoch": 0.14, "learning_rate": 1.942567896782578e-05, "loss": 24.2896, "step": 10990 }, { "epoch": 0.14, "learning_rate": 1.94243477887546e-05, "loss": 24.3205, "step": 11000 }, { "epoch": 0.14, "learning_rate": 1.9423015114462592e-05, "loss": 23.8304, "step": 11010 }, { "epoch": 0.14, "learning_rate": 1.9421680945161195e-05, "loss": 24.4937, "step": 11020 }, { "epoch": 0.14, "learning_rate": 1.9420345281062084e-05, "loss": 24.1339, "step": 11030 }, { "epoch": 0.14, "learning_rate": 1.941900812237717e-05, "loss": 24.3125, "step": 11040 }, { "epoch": 0.14, "learning_rate": 1.9417669469318588e-05, "loss": 23.9973, "step": 11050 }, { "epoch": 0.14, "learning_rate": 1.9416329322098737e-05, "loss": 24.1937, "step": 11060 }, { "epoch": 0.14, "learning_rate": 1.9414987680930227e-05, "loss": 24.1077, "step": 11070 }, { "epoch": 0.14, "learning_rate": 1.9413644546025926e-05, "loss": 24.4623, "step": 11080 }, { "epoch": 0.14, "learning_rate": 1.9412299917598924e-05, "loss": 23.7403, "step": 11090 }, { "epoch": 0.14, "learning_rate": 1.9410953795862554e-05, "loss": 24.111, "step": 11100 }, { "epoch": 0.14, "learning_rate": 1.9409606181030383e-05, "loss": 24.1381, "step": 11110 }, { "epoch": 0.14, "learning_rate": 1.940825707331622e-05, "loss": 23.9113, "step": 11120 }, { "epoch": 0.14, "learning_rate": 1.9406906472934102e-05, "loss": 24.5744, "step": 11130 }, { "epoch": 0.14, "learning_rate": 1.9405554380098314e-05, "loss": 23.9869, "step": 11140 }, { "epoch": 0.14, "learning_rate": 1.9404200795023372e-05, "loss": 23.7833, "step": 11150 }, { "epoch": 0.14, "learning_rate": 1.940284571792403e-05, "loss": 24.0625, "step": 11160 }, { "epoch": 0.14, "learning_rate": 1.9401489149015278e-05, "loss": 24.5318, "step": 11170 }, { "epoch": 0.14, "learning_rate": 1.940013108851234e-05, "loss": 24.3526, "step": 11180 }, { "epoch": 0.14, "learning_rate": 1.9398771536630682e-05, "loss": 23.5795, "step": 11190 }, { "epoch": 0.14, "learning_rate": 1.9397410493585996e-05, "loss": 24.1185, "step": 11200 }, { "epoch": 0.14, "learning_rate": 1.939604795959423e-05, "loss": 24.0207, "step": 11210 }, { "epoch": 0.14, "learning_rate": 1.939468393487155e-05, "loss": 24.6091, "step": 11220 }, { "epoch": 0.14, "learning_rate": 1.939331841963437e-05, "loss": 24.554, "step": 11230 }, { "epoch": 0.14, "learning_rate": 1.939195141409933e-05, "loss": 23.9954, "step": 11240 }, { "epoch": 0.14, "learning_rate": 1.9390582918483315e-05, "loss": 23.665, "step": 11250 }, { "epoch": 0.14, "learning_rate": 1.9389212933003444e-05, "loss": 23.8878, "step": 11260 }, { "epoch": 0.14, "learning_rate": 1.9387841457877073e-05, "loss": 23.7388, "step": 11270 }, { "epoch": 0.14, "learning_rate": 1.9386468493321794e-05, "loss": 24.0573, "step": 11280 }, { "epoch": 0.14, "learning_rate": 1.938509403955543e-05, "loss": 24.3264, "step": 11290 }, { "epoch": 0.14, "learning_rate": 1.9383718096796053e-05, "loss": 24.6399, "step": 11300 }, { "epoch": 0.14, "learning_rate": 1.938234066526196e-05, "loss": 23.7796, "step": 11310 }, { "epoch": 0.14, "learning_rate": 1.9380961745171682e-05, "loss": 23.8447, "step": 11320 }, { "epoch": 0.14, "learning_rate": 1.9379581336743997e-05, "loss": 24.2082, "step": 11330 }, { "epoch": 0.14, "learning_rate": 1.9378199440197914e-05, "loss": 23.8484, "step": 11340 }, { "epoch": 0.14, "learning_rate": 1.937681605575268e-05, "loss": 24.027, "step": 11350 }, { "epoch": 0.14, "learning_rate": 1.937543118362777e-05, "loss": 23.7753, "step": 11360 }, { "epoch": 0.14, "learning_rate": 1.9374044824042906e-05, "loss": 24.243, "step": 11370 }, { "epoch": 0.14, "learning_rate": 1.937265697721804e-05, "loss": 24.1807, "step": 11380 }, { "epoch": 0.14, "learning_rate": 1.9371267643373353e-05, "loss": 23.8575, "step": 11390 }, { "epoch": 0.14, "learning_rate": 1.9369876822729285e-05, "loss": 23.9004, "step": 11400 }, { "epoch": 0.14, "learning_rate": 1.9368484515506487e-05, "loss": 23.5145, "step": 11410 }, { "epoch": 0.14, "learning_rate": 1.9367090721925858e-05, "loss": 24.1937, "step": 11420 }, { "epoch": 0.14, "learning_rate": 1.936569544220853e-05, "loss": 24.136, "step": 11430 }, { "epoch": 0.14, "learning_rate": 1.936429867657587e-05, "loss": 23.7991, "step": 11440 }, { "epoch": 0.14, "learning_rate": 1.936290042524948e-05, "loss": 23.8957, "step": 11450 }, { "epoch": 0.14, "learning_rate": 1.9361500688451206e-05, "loss": 23.8131, "step": 11460 }, { "epoch": 0.14, "learning_rate": 1.936009946640312e-05, "loss": 24.3102, "step": 11470 }, { "epoch": 0.14, "learning_rate": 1.9358696759327535e-05, "loss": 23.4187, "step": 11480 }, { "epoch": 0.14, "learning_rate": 1.9357292567446992e-05, "loss": 23.9914, "step": 11490 }, { "epoch": 0.14, "learning_rate": 1.9355886890984278e-05, "loss": 24.2055, "step": 11500 }, { "epoch": 0.14, "learning_rate": 1.9354479730162416e-05, "loss": 24.6328, "step": 11510 }, { "epoch": 0.14, "learning_rate": 1.9353071085204643e-05, "loss": 24.3475, "step": 11520 }, { "epoch": 0.14, "learning_rate": 1.9351660956334465e-05, "loss": 24.0788, "step": 11530 }, { "epoch": 0.14, "learning_rate": 1.9350249343775597e-05, "loss": 24.3084, "step": 11540 }, { "epoch": 0.14, "learning_rate": 1.9348836247752e-05, "loss": 24.3446, "step": 11550 }, { "epoch": 0.14, "learning_rate": 1.934742166848787e-05, "loss": 24.4797, "step": 11560 }, { "epoch": 0.14, "learning_rate": 1.9346005606207634e-05, "loss": 24.261, "step": 11570 }, { "epoch": 0.14, "learning_rate": 1.9344588061135963e-05, "loss": 24.4691, "step": 11580 }, { "epoch": 0.14, "learning_rate": 1.9343169033497754e-05, "loss": 24.3041, "step": 11590 }, { "epoch": 0.14, "learning_rate": 1.934174852351814e-05, "loss": 23.9145, "step": 11600 }, { "epoch": 0.14, "learning_rate": 1.9340326531422504e-05, "loss": 23.9422, "step": 11610 }, { "epoch": 0.14, "learning_rate": 1.9338903057436437e-05, "loss": 23.4472, "step": 11620 }, { "epoch": 0.14, "learning_rate": 1.9337478101785793e-05, "loss": 24.2922, "step": 11630 }, { "epoch": 0.14, "learning_rate": 1.933605166469664e-05, "loss": 23.3475, "step": 11640 }, { "epoch": 0.14, "learning_rate": 1.9334623746395295e-05, "loss": 23.4883, "step": 11650 }, { "epoch": 0.14, "learning_rate": 1.9333194347108298e-05, "loss": 24.2361, "step": 11660 }, { "epoch": 0.14, "learning_rate": 1.9331763467062436e-05, "loss": 23.5648, "step": 11670 }, { "epoch": 0.14, "learning_rate": 1.9330331106484725e-05, "loss": 24.3235, "step": 11680 }, { "epoch": 0.14, "learning_rate": 1.9328897265602416e-05, "loss": 24.4228, "step": 11690 }, { "epoch": 0.14, "learning_rate": 1.9327461944642992e-05, "loss": 24.0444, "step": 11700 }, { "epoch": 0.14, "learning_rate": 1.932602514383418e-05, "loss": 24.2238, "step": 11710 }, { "epoch": 0.14, "learning_rate": 1.932458686340393e-05, "loss": 24.3603, "step": 11720 }, { "epoch": 0.14, "learning_rate": 1.9323147103580434e-05, "loss": 23.8037, "step": 11730 }, { "epoch": 0.14, "learning_rate": 1.932170586459212e-05, "loss": 24.2969, "step": 11740 }, { "epoch": 0.14, "learning_rate": 1.9320263146667645e-05, "loss": 23.8096, "step": 11750 }, { "epoch": 0.14, "learning_rate": 1.9318818950035902e-05, "loss": 24.0511, "step": 11760 }, { "epoch": 0.14, "learning_rate": 1.9317373274926026e-05, "loss": 24.1948, "step": 11770 }, { "epoch": 0.14, "learning_rate": 1.9315926121567375e-05, "loss": 24.6529, "step": 11780 }, { "epoch": 0.14, "learning_rate": 1.9314477490189554e-05, "loss": 23.2538, "step": 11790 }, { "epoch": 0.15, "learning_rate": 1.9313027381022386e-05, "loss": 24.5688, "step": 11800 }, { "epoch": 0.15, "learning_rate": 1.9311575794295946e-05, "loss": 24.3508, "step": 11810 }, { "epoch": 0.15, "learning_rate": 1.931012273024053e-05, "loss": 24.0333, "step": 11820 }, { "epoch": 0.15, "learning_rate": 1.9308668189086685e-05, "loss": 23.8452, "step": 11830 }, { "epoch": 0.15, "learning_rate": 1.9307212171065168e-05, "loss": 23.7673, "step": 11840 }, { "epoch": 0.15, "learning_rate": 1.9305754676406988e-05, "loss": 23.7802, "step": 11850 }, { "epoch": 0.15, "learning_rate": 1.930429570534339e-05, "loss": 24.074, "step": 11860 }, { "epoch": 0.15, "learning_rate": 1.930283525810584e-05, "loss": 24.3065, "step": 11870 }, { "epoch": 0.15, "learning_rate": 1.9301373334926047e-05, "loss": 24.3024, "step": 11880 }, { "epoch": 0.15, "learning_rate": 1.9299909936035957e-05, "loss": 24.0101, "step": 11890 }, { "epoch": 0.15, "learning_rate": 1.9298445061667737e-05, "loss": 23.8801, "step": 11900 }, { "epoch": 0.15, "learning_rate": 1.9296978712053803e-05, "loss": 24.3349, "step": 11910 }, { "epoch": 0.15, "learning_rate": 1.92955108874268e-05, "loss": 24.0728, "step": 11920 }, { "epoch": 0.15, "learning_rate": 1.9294041588019604e-05, "loss": 23.7147, "step": 11930 }, { "epoch": 0.15, "learning_rate": 1.9292570814065324e-05, "loss": 23.7403, "step": 11940 }, { "epoch": 0.15, "learning_rate": 1.9291098565797307e-05, "loss": 24.3712, "step": 11950 }, { "epoch": 0.15, "learning_rate": 1.928962484344914e-05, "loss": 24.3014, "step": 11960 }, { "epoch": 0.15, "learning_rate": 1.9288149647254626e-05, "loss": 24.3053, "step": 11970 }, { "epoch": 0.15, "learning_rate": 1.9286672977447818e-05, "loss": 24.1158, "step": 11980 }, { "epoch": 0.15, "learning_rate": 1.9285194834262994e-05, "loss": 23.7079, "step": 11990 }, { "epoch": 0.15, "learning_rate": 1.9283715217934673e-05, "loss": 24.1164, "step": 12000 }, { "epoch": 0.15, "learning_rate": 1.9282234128697603e-05, "loss": 24.4227, "step": 12010 }, { "epoch": 0.15, "learning_rate": 1.9280751566786763e-05, "loss": 23.7225, "step": 12020 }, { "epoch": 0.15, "learning_rate": 1.9279267532437372e-05, "loss": 24.5004, "step": 12030 }, { "epoch": 0.15, "learning_rate": 1.9277782025884877e-05, "loss": 24.3343, "step": 12040 }, { "epoch": 0.15, "learning_rate": 1.9276295047364964e-05, "loss": 24.0298, "step": 12050 }, { "epoch": 0.15, "learning_rate": 1.9274806597113545e-05, "loss": 23.9137, "step": 12060 }, { "epoch": 0.15, "learning_rate": 1.9273316675366783e-05, "loss": 24.2872, "step": 12070 }, { "epoch": 0.15, "learning_rate": 1.9271825282361044e-05, "loss": 24.3313, "step": 12080 }, { "epoch": 0.15, "learning_rate": 1.927033241833296e-05, "loss": 23.962, "step": 12090 }, { "epoch": 0.15, "learning_rate": 1.9268838083519368e-05, "loss": 24.4391, "step": 12100 }, { "epoch": 0.15, "learning_rate": 1.9267342278157362e-05, "loss": 24.4613, "step": 12110 }, { "epoch": 0.15, "learning_rate": 1.9265845002484257e-05, "loss": 24.2988, "step": 12120 }, { "epoch": 0.15, "learning_rate": 1.9264346256737604e-05, "loss": 24.4877, "step": 12130 }, { "epoch": 0.15, "learning_rate": 1.926284604115518e-05, "loss": 24.0678, "step": 12140 }, { "epoch": 0.15, "learning_rate": 1.926134435597501e-05, "loss": 24.0294, "step": 12150 }, { "epoch": 0.15, "learning_rate": 1.925984120143534e-05, "loss": 23.9895, "step": 12160 }, { "epoch": 0.15, "learning_rate": 1.9258336577774658e-05, "loss": 23.9063, "step": 12170 }, { "epoch": 0.15, "learning_rate": 1.9256830485231674e-05, "loss": 24.8577, "step": 12180 }, { "epoch": 0.15, "learning_rate": 1.925532292404534e-05, "loss": 24.6828, "step": 12190 }, { "epoch": 0.15, "learning_rate": 1.925381389445484e-05, "loss": 24.4579, "step": 12200 }, { "epoch": 0.15, "learning_rate": 1.925230339669958e-05, "loss": 24.1294, "step": 12210 }, { "epoch": 0.15, "learning_rate": 1.925079143101922e-05, "loss": 24.0372, "step": 12220 }, { "epoch": 0.15, "learning_rate": 1.9249277997653636e-05, "loss": 23.5307, "step": 12230 }, { "epoch": 0.15, "learning_rate": 1.9247763096842945e-05, "loss": 23.9176, "step": 12240 }, { "epoch": 0.15, "learning_rate": 1.9246246728827485e-05, "loss": 24.4871, "step": 12250 }, { "epoch": 0.15, "learning_rate": 1.9244728893847847e-05, "loss": 24.2975, "step": 12260 }, { "epoch": 0.15, "learning_rate": 1.9243209592144833e-05, "loss": 24.629, "step": 12270 }, { "epoch": 0.15, "learning_rate": 1.92416888239595e-05, "loss": 24.2938, "step": 12280 }, { "epoch": 0.15, "learning_rate": 1.924016658953311e-05, "loss": 24.1026, "step": 12290 }, { "epoch": 0.15, "learning_rate": 1.9238642889107185e-05, "loss": 24.0235, "step": 12300 }, { "epoch": 0.15, "learning_rate": 1.9237117722923466e-05, "loss": 23.7903, "step": 12310 }, { "epoch": 0.15, "learning_rate": 1.9235591091223925e-05, "loss": 24.2052, "step": 12320 }, { "epoch": 0.15, "learning_rate": 1.9234062994250772e-05, "loss": 23.4747, "step": 12330 }, { "epoch": 0.15, "learning_rate": 1.923253343224645e-05, "loss": 24.5898, "step": 12340 }, { "epoch": 0.15, "learning_rate": 1.9231002405453623e-05, "loss": 24.0333, "step": 12350 }, { "epoch": 0.15, "learning_rate": 1.9229469914115207e-05, "loss": 23.9726, "step": 12360 }, { "epoch": 0.15, "learning_rate": 1.9227935958474333e-05, "loss": 23.5959, "step": 12370 }, { "epoch": 0.15, "learning_rate": 1.9226400538774374e-05, "loss": 24.2943, "step": 12380 }, { "epoch": 0.15, "learning_rate": 1.9224863655258926e-05, "loss": 25.0519, "step": 12390 }, { "epoch": 0.15, "learning_rate": 1.922332530817183e-05, "loss": 24.1909, "step": 12400 }, { "epoch": 0.15, "learning_rate": 1.9221785497757155e-05, "loss": 23.7182, "step": 12410 }, { "epoch": 0.15, "learning_rate": 1.9220244224259193e-05, "loss": 24.3, "step": 12420 }, { "epoch": 0.15, "learning_rate": 1.9218701487922476e-05, "loss": 24.1378, "step": 12430 }, { "epoch": 0.15, "learning_rate": 1.9217157288991774e-05, "loss": 24.5945, "step": 12440 }, { "epoch": 0.15, "learning_rate": 1.921561162771207e-05, "loss": 24.8097, "step": 12450 }, { "epoch": 0.15, "learning_rate": 1.92140645043286e-05, "loss": 23.6621, "step": 12460 }, { "epoch": 0.15, "learning_rate": 1.9212515919086822e-05, "loss": 24.5307, "step": 12470 }, { "epoch": 0.15, "learning_rate": 1.9210965872232422e-05, "loss": 24.4831, "step": 12480 }, { "epoch": 0.15, "learning_rate": 1.9209414364011326e-05, "loss": 24.6498, "step": 12490 }, { "epoch": 0.15, "learning_rate": 1.9207861394669692e-05, "loss": 23.8966, "step": 12500 }, { "epoch": 0.15, "learning_rate": 1.9206306964453905e-05, "loss": 24.6103, "step": 12510 }, { "epoch": 0.15, "learning_rate": 1.9204751073610577e-05, "loss": 23.5874, "step": 12520 }, { "epoch": 0.15, "learning_rate": 1.9203193722386567e-05, "loss": 23.9468, "step": 12530 }, { "epoch": 0.15, "learning_rate": 1.9201634911028952e-05, "loss": 23.4933, "step": 12540 }, { "epoch": 0.15, "learning_rate": 1.9200074639785044e-05, "loss": 23.9336, "step": 12550 }, { "epoch": 0.15, "learning_rate": 1.919851290890239e-05, "loss": 22.9658, "step": 12560 }, { "epoch": 0.15, "learning_rate": 1.9196949718628766e-05, "loss": 23.8889, "step": 12570 }, { "epoch": 0.15, "learning_rate": 1.9195385069212182e-05, "loss": 24.4255, "step": 12580 }, { "epoch": 0.15, "learning_rate": 1.9193818960900876e-05, "loss": 23.7015, "step": 12590 }, { "epoch": 0.15, "learning_rate": 1.9192251393943316e-05, "loss": 24.2959, "step": 12600 }, { "epoch": 0.16, "learning_rate": 1.919068236858821e-05, "loss": 23.5866, "step": 12610 }, { "epoch": 0.16, "learning_rate": 1.9189111885084492e-05, "loss": 23.9223, "step": 12620 }, { "epoch": 0.16, "learning_rate": 1.9187539943681323e-05, "loss": 24.244, "step": 12630 }, { "epoch": 0.16, "learning_rate": 1.91859665446281e-05, "loss": 24.0285, "step": 12640 }, { "epoch": 0.16, "learning_rate": 1.9184391688174453e-05, "loss": 24.1018, "step": 12650 }, { "epoch": 0.16, "learning_rate": 1.9182815374570237e-05, "loss": 23.9085, "step": 12660 }, { "epoch": 0.16, "learning_rate": 1.918123760406555e-05, "loss": 23.9537, "step": 12670 }, { "epoch": 0.16, "learning_rate": 1.9179658376910705e-05, "loss": 24.215, "step": 12680 }, { "epoch": 0.16, "learning_rate": 1.9178077693356263e-05, "loss": 24.1529, "step": 12690 }, { "epoch": 0.16, "learning_rate": 1.9176495553652998e-05, "loss": 24.1423, "step": 12700 }, { "epoch": 0.16, "learning_rate": 1.917491195805193e-05, "loss": 24.8051, "step": 12710 }, { "epoch": 0.16, "learning_rate": 1.9173326906804304e-05, "loss": 24.3083, "step": 12720 }, { "epoch": 0.16, "learning_rate": 1.9171740400161598e-05, "loss": 23.9483, "step": 12730 }, { "epoch": 0.16, "learning_rate": 1.9170152438375512e-05, "loss": 23.9918, "step": 12740 }, { "epoch": 0.16, "learning_rate": 1.9168563021697996e-05, "loss": 23.7363, "step": 12750 }, { "epoch": 0.16, "learning_rate": 1.9166972150381208e-05, "loss": 23.8512, "step": 12760 }, { "epoch": 0.16, "learning_rate": 1.9165379824677556e-05, "loss": 24.0625, "step": 12770 }, { "epoch": 0.16, "learning_rate": 1.9163786044839664e-05, "loss": 24.3125, "step": 12780 }, { "epoch": 0.16, "learning_rate": 1.91621908111204e-05, "loss": 24.4338, "step": 12790 }, { "epoch": 0.16, "learning_rate": 1.916059412377285e-05, "loss": 24.0471, "step": 12800 }, { "epoch": 0.16, "learning_rate": 1.915899598305034e-05, "loss": 23.983, "step": 12810 }, { "epoch": 0.16, "learning_rate": 1.915739638920642e-05, "loss": 24.2255, "step": 12820 }, { "epoch": 0.16, "learning_rate": 1.915579534249488e-05, "loss": 23.8776, "step": 12830 }, { "epoch": 0.16, "learning_rate": 1.9154192843169727e-05, "loss": 24.4224, "step": 12840 }, { "epoch": 0.16, "learning_rate": 1.915258889148521e-05, "loss": 24.0608, "step": 12850 }, { "epoch": 0.16, "learning_rate": 1.91509834876958e-05, "loss": 23.9537, "step": 12860 }, { "epoch": 0.16, "learning_rate": 1.914937663205621e-05, "loss": 24.859, "step": 12870 }, { "epoch": 0.16, "learning_rate": 1.9147768324821366e-05, "loss": 24.1576, "step": 12880 }, { "epoch": 0.16, "learning_rate": 1.9146158566246443e-05, "loss": 24.0946, "step": 12890 }, { "epoch": 0.16, "learning_rate": 1.9144547356586835e-05, "loss": 23.8571, "step": 12900 }, { "epoch": 0.16, "learning_rate": 1.9142934696098167e-05, "loss": 23.5641, "step": 12910 }, { "epoch": 0.16, "learning_rate": 1.9141320585036292e-05, "loss": 23.7186, "step": 12920 }, { "epoch": 0.16, "learning_rate": 1.9139705023657304e-05, "loss": 23.9974, "step": 12930 }, { "epoch": 0.16, "learning_rate": 1.9138088012217518e-05, "loss": 24.7126, "step": 12940 }, { "epoch": 0.16, "learning_rate": 1.913646955097348e-05, "loss": 24.1175, "step": 12950 }, { "epoch": 0.16, "learning_rate": 1.9134849640181965e-05, "loss": 24.4301, "step": 12960 }, { "epoch": 0.16, "learning_rate": 1.9133228280099984e-05, "loss": 24.6212, "step": 12970 }, { "epoch": 0.16, "learning_rate": 1.9131605470984775e-05, "loss": 24.0094, "step": 12980 }, { "epoch": 0.16, "learning_rate": 1.91299812130938e-05, "loss": 23.7187, "step": 12990 }, { "epoch": 0.16, "learning_rate": 1.912835550668476e-05, "loss": 23.9397, "step": 13000 }, { "epoch": 0.16, "learning_rate": 1.912672835201558e-05, "loss": 24.5713, "step": 13010 }, { "epoch": 0.16, "learning_rate": 1.9125099749344412e-05, "loss": 23.9529, "step": 13020 }, { "epoch": 0.16, "learning_rate": 1.9123469698929652e-05, "loss": 24.3717, "step": 13030 }, { "epoch": 0.16, "learning_rate": 1.912183820102991e-05, "loss": 24.6275, "step": 13040 }, { "epoch": 0.16, "learning_rate": 1.912020525590403e-05, "loss": 24.0146, "step": 13050 }, { "epoch": 0.16, "learning_rate": 1.9118570863811086e-05, "loss": 24.3874, "step": 13060 }, { "epoch": 0.16, "learning_rate": 1.9116935025010388e-05, "loss": 24.0428, "step": 13070 }, { "epoch": 0.16, "learning_rate": 1.9115297739761468e-05, "loss": 23.9197, "step": 13080 }, { "epoch": 0.16, "learning_rate": 1.911365900832409e-05, "loss": 23.6276, "step": 13090 }, { "epoch": 0.16, "learning_rate": 1.9112018830958242e-05, "loss": 23.6913, "step": 13100 }, { "epoch": 0.16, "learning_rate": 1.9110377207924155e-05, "loss": 24.2527, "step": 13110 }, { "epoch": 0.16, "learning_rate": 1.910873413948228e-05, "loss": 23.9806, "step": 13120 }, { "epoch": 0.16, "learning_rate": 1.910708962589329e-05, "loss": 23.7929, "step": 13130 }, { "epoch": 0.16, "learning_rate": 1.91054436674181e-05, "loss": 23.9123, "step": 13140 }, { "epoch": 0.16, "learning_rate": 1.910379626431785e-05, "loss": 24.4201, "step": 13150 }, { "epoch": 0.16, "learning_rate": 1.910214741685391e-05, "loss": 23.7169, "step": 13160 }, { "epoch": 0.16, "learning_rate": 1.910049712528788e-05, "loss": 23.8961, "step": 13170 }, { "epoch": 0.16, "learning_rate": 1.909884538988158e-05, "loss": 23.8494, "step": 13180 }, { "epoch": 0.16, "learning_rate": 1.909719221089707e-05, "loss": 23.7727, "step": 13190 }, { "epoch": 0.16, "learning_rate": 1.909553758859664e-05, "loss": 23.3636, "step": 13200 }, { "epoch": 0.16, "learning_rate": 1.9093881523242794e-05, "loss": 24.1079, "step": 13210 }, { "epoch": 0.16, "learning_rate": 1.9092224015098287e-05, "loss": 23.9549, "step": 13220 }, { "epoch": 0.16, "learning_rate": 1.9090565064426082e-05, "loss": 24.1303, "step": 13230 }, { "epoch": 0.16, "learning_rate": 1.9088904671489384e-05, "loss": 24.3418, "step": 13240 }, { "epoch": 0.16, "learning_rate": 1.9087242836551623e-05, "loss": 23.8346, "step": 13250 }, { "epoch": 0.16, "learning_rate": 1.9085579559876457e-05, "loss": 24.1, "step": 13260 }, { "epoch": 0.16, "learning_rate": 1.908391484172777e-05, "loss": 23.6609, "step": 13270 }, { "epoch": 0.16, "learning_rate": 1.9082248682369684e-05, "loss": 24.0198, "step": 13280 }, { "epoch": 0.16, "learning_rate": 1.9080581082066542e-05, "loss": 24.0608, "step": 13290 }, { "epoch": 0.16, "learning_rate": 1.907891204108291e-05, "loss": 23.7306, "step": 13300 }, { "epoch": 0.16, "learning_rate": 1.9077241559683603e-05, "loss": 24.344, "step": 13310 }, { "epoch": 0.16, "learning_rate": 1.907556963813364e-05, "loss": 24.448, "step": 13320 }, { "epoch": 0.16, "learning_rate": 1.9073896276698284e-05, "loss": 23.4931, "step": 13330 }, { "epoch": 0.16, "learning_rate": 1.907222147564302e-05, "loss": 24.0297, "step": 13340 }, { "epoch": 0.16, "learning_rate": 1.907054523523357e-05, "loss": 23.6877, "step": 13350 }, { "epoch": 0.16, "learning_rate": 1.9068867555735877e-05, "loss": 23.6038, "step": 13360 }, { "epoch": 0.16, "learning_rate": 1.9067188437416108e-05, "loss": 23.4188, "step": 13370 }, { "epoch": 0.16, "learning_rate": 1.9065507880540664e-05, "loss": 23.9109, "step": 13380 }, { "epoch": 0.16, "learning_rate": 1.9063825885376178e-05, "loss": 23.8756, "step": 13390 }, { "epoch": 0.16, "learning_rate": 1.9062142452189504e-05, "loss": 23.2376, "step": 13400 }, { "epoch": 0.16, "learning_rate": 1.906045758124773e-05, "loss": 24.1118, "step": 13410 }, { "epoch": 0.17, "learning_rate": 1.905877127281817e-05, "loss": 23.9485, "step": 13420 }, { "epoch": 0.17, "learning_rate": 1.905708352716836e-05, "loss": 23.7039, "step": 13430 }, { "epoch": 0.17, "learning_rate": 1.9055394344566075e-05, "loss": 23.431, "step": 13440 }, { "epoch": 0.17, "learning_rate": 1.9053703725279312e-05, "loss": 24.179, "step": 13450 }, { "epoch": 0.17, "learning_rate": 1.9052011669576292e-05, "loss": 23.8475, "step": 13460 }, { "epoch": 0.17, "learning_rate": 1.905031817772547e-05, "loss": 24.243, "step": 13470 }, { "epoch": 0.17, "learning_rate": 1.904862324999553e-05, "loss": 23.9042, "step": 13480 }, { "epoch": 0.17, "learning_rate": 1.904692688665538e-05, "loss": 24.0638, "step": 13490 }, { "epoch": 0.17, "learning_rate": 1.9045229087974154e-05, "loss": 24.6993, "step": 13500 }, { "epoch": 0.17, "learning_rate": 1.9043529854221215e-05, "loss": 23.8547, "step": 13510 }, { "epoch": 0.17, "learning_rate": 1.9041829185666163e-05, "loss": 24.3776, "step": 13520 }, { "epoch": 0.17, "learning_rate": 1.904012708257881e-05, "loss": 24.1637, "step": 13530 }, { "epoch": 0.17, "learning_rate": 1.9038423545229206e-05, "loss": 24.5479, "step": 13540 }, { "epoch": 0.17, "learning_rate": 1.903671857388763e-05, "loss": 23.6799, "step": 13550 }, { "epoch": 0.17, "learning_rate": 1.903501216882458e-05, "loss": 23.7544, "step": 13560 }, { "epoch": 0.17, "learning_rate": 1.903330433031078e-05, "loss": 24.1374, "step": 13570 }, { "epoch": 0.17, "learning_rate": 1.90315950586172e-05, "loss": 23.8687, "step": 13580 }, { "epoch": 0.17, "learning_rate": 1.9029884354015017e-05, "loss": 23.4758, "step": 13590 }, { "epoch": 0.17, "learning_rate": 1.9028172216775642e-05, "loss": 23.5783, "step": 13600 }, { "epoch": 0.17, "learning_rate": 1.902645864717072e-05, "loss": 24.0993, "step": 13610 }, { "epoch": 0.17, "learning_rate": 1.902474364547211e-05, "loss": 24.1591, "step": 13620 }, { "epoch": 0.17, "learning_rate": 1.9023027211951914e-05, "loss": 24.0931, "step": 13630 }, { "epoch": 0.17, "learning_rate": 1.9021309346882447e-05, "loss": 24.2667, "step": 13640 }, { "epoch": 0.17, "learning_rate": 1.9019590050536262e-05, "loss": 24.2926, "step": 13650 }, { "epoch": 0.17, "learning_rate": 1.9017869323186127e-05, "loss": 23.8994, "step": 13660 }, { "epoch": 0.17, "learning_rate": 1.9016147165105052e-05, "loss": 24.3082, "step": 13670 }, { "epoch": 0.17, "learning_rate": 1.901442357656626e-05, "loss": 23.9335, "step": 13680 }, { "epoch": 0.17, "learning_rate": 1.901269855784321e-05, "loss": 23.8972, "step": 13690 }, { "epoch": 0.17, "learning_rate": 1.9010972109209582e-05, "loss": 23.5807, "step": 13700 }, { "epoch": 0.17, "learning_rate": 1.9009244230939292e-05, "loss": 23.6262, "step": 13710 }, { "epoch": 0.17, "learning_rate": 1.900751492330647e-05, "loss": 24.5787, "step": 13720 }, { "epoch": 0.17, "learning_rate": 1.9005784186585488e-05, "loss": 24.1018, "step": 13730 }, { "epoch": 0.17, "learning_rate": 1.9004052021050926e-05, "loss": 24.2528, "step": 13740 }, { "epoch": 0.17, "learning_rate": 1.9002318426977608e-05, "loss": 24.7451, "step": 13750 }, { "epoch": 0.17, "learning_rate": 1.9000583404640574e-05, "loss": 24.6188, "step": 13760 }, { "epoch": 0.17, "learning_rate": 1.8998846954315095e-05, "loss": 24.1263, "step": 13770 }, { "epoch": 0.17, "learning_rate": 1.8997109076276668e-05, "loss": 24.6276, "step": 13780 }, { "epoch": 0.17, "learning_rate": 1.899536977080102e-05, "loss": 23.9662, "step": 13790 }, { "epoch": 0.17, "learning_rate": 1.8993629038164093e-05, "loss": 24.0036, "step": 13800 }, { "epoch": 0.17, "learning_rate": 1.8991886878642067e-05, "loss": 23.7958, "step": 13810 }, { "epoch": 0.17, "learning_rate": 1.8990143292511347e-05, "loss": 24.2468, "step": 13820 }, { "epoch": 0.17, "learning_rate": 1.8988398280048558e-05, "loss": 24.0755, "step": 13830 }, { "epoch": 0.17, "learning_rate": 1.8986651841530555e-05, "loss": 24.2843, "step": 13840 }, { "epoch": 0.17, "learning_rate": 1.898490397723442e-05, "loss": 24.5059, "step": 13850 }, { "epoch": 0.17, "learning_rate": 1.8983154687437465e-05, "loss": 24.3422, "step": 13860 }, { "epoch": 0.17, "learning_rate": 1.898140397241722e-05, "loss": 23.6847, "step": 13870 }, { "epoch": 0.17, "learning_rate": 1.897965183245144e-05, "loss": 24.0439, "step": 13880 }, { "epoch": 0.17, "learning_rate": 1.8977898267818124e-05, "loss": 24.1264, "step": 13890 }, { "epoch": 0.17, "learning_rate": 1.897614327879547e-05, "loss": 23.7698, "step": 13900 }, { "epoch": 0.17, "learning_rate": 1.8974386865661927e-05, "loss": 24.2084, "step": 13910 }, { "epoch": 0.17, "learning_rate": 1.897262902869615e-05, "loss": 24.0648, "step": 13920 }, { "epoch": 0.17, "learning_rate": 1.8970869768177033e-05, "loss": 23.9575, "step": 13930 }, { "epoch": 0.17, "learning_rate": 1.896910908438369e-05, "loss": 23.928, "step": 13940 }, { "epoch": 0.17, "learning_rate": 1.8967346977595466e-05, "loss": 24.4869, "step": 13950 }, { "epoch": 0.17, "learning_rate": 1.8965583448091925e-05, "loss": 23.9702, "step": 13960 }, { "epoch": 0.17, "learning_rate": 1.896381849615286e-05, "loss": 23.8268, "step": 13970 }, { "epoch": 0.17, "learning_rate": 1.8962052122058293e-05, "loss": 23.6403, "step": 13980 }, { "epoch": 0.17, "learning_rate": 1.8960284326088465e-05, "loss": 24.1462, "step": 13990 }, { "epoch": 0.17, "learning_rate": 1.8958515108523846e-05, "loss": 24.4824, "step": 14000 }, { "epoch": 0.17, "learning_rate": 1.895674446964513e-05, "loss": 23.6164, "step": 14010 }, { "epoch": 0.17, "learning_rate": 1.8954972409733246e-05, "loss": 24.3738, "step": 14020 }, { "epoch": 0.17, "learning_rate": 1.895319892906933e-05, "loss": 24.2826, "step": 14030 }, { "epoch": 0.17, "learning_rate": 1.895142402793476e-05, "loss": 24.0769, "step": 14040 }, { "epoch": 0.17, "learning_rate": 1.8949647706611128e-05, "loss": 24.0515, "step": 14050 }, { "epoch": 0.17, "learning_rate": 1.894786996538026e-05, "loss": 24.0444, "step": 14060 }, { "epoch": 0.17, "learning_rate": 1.8946090804524203e-05, "loss": 23.9571, "step": 14070 }, { "epoch": 0.17, "learning_rate": 1.8944310224325233e-05, "loss": 23.5221, "step": 14080 }, { "epoch": 0.17, "learning_rate": 1.8942528225065843e-05, "loss": 23.7205, "step": 14090 }, { "epoch": 0.17, "learning_rate": 1.894074480702876e-05, "loss": 24.1618, "step": 14100 }, { "epoch": 0.17, "learning_rate": 1.893895997049693e-05, "loss": 23.4872, "step": 14110 }, { "epoch": 0.17, "learning_rate": 1.893717371575353e-05, "loss": 24.1551, "step": 14120 }, { "epoch": 0.17, "learning_rate": 1.8935386043081952e-05, "loss": 23.8938, "step": 14130 }, { "epoch": 0.17, "learning_rate": 1.8933596952765824e-05, "loss": 23.9581, "step": 14140 }, { "epoch": 0.17, "learning_rate": 1.8931806445088994e-05, "loss": 23.6852, "step": 14150 }, { "epoch": 0.17, "learning_rate": 1.8930014520335536e-05, "loss": 24.5673, "step": 14160 }, { "epoch": 0.17, "learning_rate": 1.8928221178789746e-05, "loss": 23.8822, "step": 14170 }, { "epoch": 0.17, "learning_rate": 1.8926426420736142e-05, "loss": 23.8792, "step": 14180 }, { "epoch": 0.17, "learning_rate": 1.8924630246459486e-05, "loss": 23.8193, "step": 14190 }, { "epoch": 0.17, "learning_rate": 1.8922832656244738e-05, "loss": 24.2447, "step": 14200 }, { "epoch": 0.17, "learning_rate": 1.8921033650377095e-05, "loss": 24.5694, "step": 14210 }, { "epoch": 0.17, "learning_rate": 1.8919233229141986e-05, "loss": 23.8891, "step": 14220 }, { "epoch": 0.18, "learning_rate": 1.891743139282505e-05, "loss": 24.3299, "step": 14230 }, { "epoch": 0.18, "learning_rate": 1.891562814171216e-05, "loss": 23.7289, "step": 14240 }, { "epoch": 0.18, "learning_rate": 1.8913823476089404e-05, "loss": 23.6398, "step": 14250 }, { "epoch": 0.18, "learning_rate": 1.8912017396243118e-05, "loss": 24.0658, "step": 14260 }, { "epoch": 0.18, "learning_rate": 1.8910209902459834e-05, "loss": 23.8764, "step": 14270 }, { "epoch": 0.18, "learning_rate": 1.890840099502632e-05, "loss": 24.1189, "step": 14280 }, { "epoch": 0.18, "learning_rate": 1.8906590674229575e-05, "loss": 24.5511, "step": 14290 }, { "epoch": 0.18, "learning_rate": 1.8904778940356803e-05, "loss": 24.0108, "step": 14300 }, { "epoch": 0.18, "learning_rate": 1.890296579369546e-05, "loss": 24.388, "step": 14310 }, { "epoch": 0.18, "learning_rate": 1.8901151234533205e-05, "loss": 24.4691, "step": 14320 }, { "epoch": 0.18, "learning_rate": 1.8899335263157924e-05, "loss": 24.6387, "step": 14330 }, { "epoch": 0.18, "learning_rate": 1.889751787985773e-05, "loss": 23.9904, "step": 14340 }, { "epoch": 0.18, "learning_rate": 1.8895699084920965e-05, "loss": 23.3823, "step": 14350 }, { "epoch": 0.18, "learning_rate": 1.8893878878636186e-05, "loss": 24.1485, "step": 14360 }, { "epoch": 0.18, "learning_rate": 1.889205726129218e-05, "loss": 24.8737, "step": 14370 }, { "epoch": 0.18, "learning_rate": 1.8890234233177956e-05, "loss": 23.8244, "step": 14380 }, { "epoch": 0.18, "learning_rate": 1.8888409794582744e-05, "loss": 23.3918, "step": 14390 }, { "epoch": 0.18, "learning_rate": 1.8886583945796003e-05, "loss": 23.6288, "step": 14400 }, { "epoch": 0.18, "learning_rate": 1.8884756687107413e-05, "loss": 24.0943, "step": 14410 }, { "epoch": 0.18, "learning_rate": 1.8882928018806878e-05, "loss": 24.0495, "step": 14420 }, { "epoch": 0.18, "learning_rate": 1.8881097941184525e-05, "loss": 24.0392, "step": 14430 }, { "epoch": 0.18, "learning_rate": 1.8879266454530705e-05, "loss": 24.0153, "step": 14440 }, { "epoch": 0.18, "learning_rate": 1.8877433559135996e-05, "loss": 23.8803, "step": 14450 }, { "epoch": 0.18, "learning_rate": 1.8875599255291187e-05, "loss": 24.9088, "step": 14460 }, { "epoch": 0.18, "learning_rate": 1.8873763543287313e-05, "loss": 24.7807, "step": 14470 }, { "epoch": 0.18, "learning_rate": 1.8871926423415608e-05, "loss": 23.8637, "step": 14480 }, { "epoch": 0.18, "learning_rate": 1.8870087895967543e-05, "loss": 24.1423, "step": 14490 }, { "epoch": 0.18, "learning_rate": 1.8868247961234813e-05, "loss": 24.1969, "step": 14500 }, { "epoch": 0.18, "learning_rate": 1.8866406619509333e-05, "loss": 23.8287, "step": 14510 }, { "epoch": 0.18, "learning_rate": 1.8864563871083236e-05, "loss": 24.023, "step": 14520 }, { "epoch": 0.18, "learning_rate": 1.886271971624889e-05, "loss": 24.4365, "step": 14530 }, { "epoch": 0.18, "learning_rate": 1.8860874155298873e-05, "loss": 24.3895, "step": 14540 }, { "epoch": 0.18, "learning_rate": 1.8859027188525994e-05, "loss": 23.99, "step": 14550 }, { "epoch": 0.18, "learning_rate": 1.8857178816223292e-05, "loss": 24.0161, "step": 14560 }, { "epoch": 0.18, "learning_rate": 1.8855329038684013e-05, "loss": 24.7513, "step": 14570 }, { "epoch": 0.18, "learning_rate": 1.8853477856201634e-05, "loss": 23.4753, "step": 14580 }, { "epoch": 0.18, "learning_rate": 1.8851625269069854e-05, "loss": 23.7207, "step": 14590 }, { "epoch": 0.18, "learning_rate": 1.8849771277582598e-05, "loss": 23.8053, "step": 14600 }, { "epoch": 0.18, "learning_rate": 1.8847915882034013e-05, "loss": 23.4081, "step": 14610 }, { "epoch": 0.18, "learning_rate": 1.8846059082718463e-05, "loss": 23.9905, "step": 14620 }, { "epoch": 0.18, "learning_rate": 1.884420087993054e-05, "loss": 23.5943, "step": 14630 }, { "epoch": 0.18, "learning_rate": 1.884234127396505e-05, "loss": 24.5789, "step": 14640 }, { "epoch": 0.18, "learning_rate": 1.8840480265117046e-05, "loss": 23.7325, "step": 14650 }, { "epoch": 0.18, "learning_rate": 1.8838617853681772e-05, "loss": 23.9026, "step": 14660 }, { "epoch": 0.18, "learning_rate": 1.8836754039954713e-05, "loss": 23.9211, "step": 14670 }, { "epoch": 0.18, "learning_rate": 1.8834888824231578e-05, "loss": 24.0017, "step": 14680 }, { "epoch": 0.18, "learning_rate": 1.8833022206808286e-05, "loss": 24.1795, "step": 14690 }, { "epoch": 0.18, "learning_rate": 1.8831154187980983e-05, "loss": 23.8199, "step": 14700 }, { "epoch": 0.18, "learning_rate": 1.882928476804605e-05, "loss": 23.9704, "step": 14710 }, { "epoch": 0.18, "learning_rate": 1.8827413947300074e-05, "loss": 24.128, "step": 14720 }, { "epoch": 0.18, "learning_rate": 1.8825541726039868e-05, "loss": 23.4916, "step": 14730 }, { "epoch": 0.18, "learning_rate": 1.8823668104562473e-05, "loss": 23.8464, "step": 14740 }, { "epoch": 0.18, "learning_rate": 1.8821793083165148e-05, "loss": 24.3654, "step": 14750 }, { "epoch": 0.18, "learning_rate": 1.8819916662145373e-05, "loss": 24.0797, "step": 14760 }, { "epoch": 0.18, "learning_rate": 1.8818038841800858e-05, "loss": 23.4878, "step": 14770 }, { "epoch": 0.18, "learning_rate": 1.8816159622429516e-05, "loss": 24.0561, "step": 14780 }, { "epoch": 0.18, "learning_rate": 1.881427900432951e-05, "loss": 24.0713, "step": 14790 }, { "epoch": 0.18, "learning_rate": 1.88123969877992e-05, "loss": 24.068, "step": 14800 }, { "epoch": 0.18, "learning_rate": 1.8810513573137178e-05, "loss": 24.0102, "step": 14810 }, { "epoch": 0.18, "learning_rate": 1.8808628760642263e-05, "loss": 24.0336, "step": 14820 }, { "epoch": 0.18, "learning_rate": 1.880674255061348e-05, "loss": 24.3288, "step": 14830 }, { "epoch": 0.18, "learning_rate": 1.8804854943350098e-05, "loss": 24.0389, "step": 14840 }, { "epoch": 0.18, "learning_rate": 1.880296593915159e-05, "loss": 24.2182, "step": 14850 }, { "epoch": 0.18, "learning_rate": 1.880107553831765e-05, "loss": 24.5494, "step": 14860 }, { "epoch": 0.18, "learning_rate": 1.879918374114821e-05, "loss": 24.0277, "step": 14870 }, { "epoch": 0.18, "learning_rate": 1.8797290547943404e-05, "loss": 24.236, "step": 14880 }, { "epoch": 0.18, "learning_rate": 1.8795395959003607e-05, "loss": 23.9177, "step": 14890 }, { "epoch": 0.18, "learning_rate": 1.87934999746294e-05, "loss": 24.7762, "step": 14900 }, { "epoch": 0.18, "learning_rate": 1.8791602595121588e-05, "loss": 23.9773, "step": 14910 }, { "epoch": 0.18, "learning_rate": 1.8789703820781203e-05, "loss": 23.9609, "step": 14920 }, { "epoch": 0.18, "learning_rate": 1.878780365190949e-05, "loss": 24.1485, "step": 14930 }, { "epoch": 0.18, "learning_rate": 1.878590208880793e-05, "loss": 23.2855, "step": 14940 }, { "epoch": 0.18, "learning_rate": 1.878399913177821e-05, "loss": 24.031, "step": 14950 }, { "epoch": 0.18, "learning_rate": 1.8782094781122242e-05, "loss": 24.3228, "step": 14960 }, { "epoch": 0.18, "learning_rate": 1.8780189037142168e-05, "loss": 24.5841, "step": 14970 }, { "epoch": 0.18, "learning_rate": 1.8778281900140335e-05, "loss": 24.4002, "step": 14980 }, { "epoch": 0.18, "learning_rate": 1.8776373370419327e-05, "loss": 23.8255, "step": 14990 }, { "epoch": 0.18, "learning_rate": 1.8774463448281938e-05, "loss": 23.8118, "step": 15000 }, { "epoch": 0.18, "learning_rate": 1.877255213403119e-05, "loss": 23.9285, "step": 15010 }, { "epoch": 0.18, "learning_rate": 1.8770639427970318e-05, "loss": 23.8736, "step": 15020 }, { "epoch": 0.18, "learning_rate": 1.876872533040279e-05, "loss": 23.619, "step": 15030 }, { "epoch": 0.18, "learning_rate": 1.8766809841632278e-05, "loss": 23.7412, "step": 15040 }, { "epoch": 0.19, "learning_rate": 1.876489296196269e-05, "loss": 23.6915, "step": 15050 }, { "epoch": 0.19, "learning_rate": 1.876297469169815e-05, "loss": 24.0382, "step": 15060 }, { "epoch": 0.19, "learning_rate": 1.8761055031142995e-05, "loss": 24.0897, "step": 15070 }, { "epoch": 0.19, "learning_rate": 1.8759133980601798e-05, "loss": 23.769, "step": 15080 }, { "epoch": 0.19, "learning_rate": 1.8757211540379337e-05, "loss": 24.0791, "step": 15090 }, { "epoch": 0.19, "learning_rate": 1.8755287710780617e-05, "loss": 23.7392, "step": 15100 }, { "epoch": 0.19, "learning_rate": 1.8753362492110864e-05, "loss": 23.8448, "step": 15110 }, { "epoch": 0.19, "learning_rate": 1.875143588467553e-05, "loss": 24.1047, "step": 15120 }, { "epoch": 0.19, "learning_rate": 1.874950788878027e-05, "loss": 23.673, "step": 15130 }, { "epoch": 0.19, "learning_rate": 1.8747578504730978e-05, "loss": 24.0824, "step": 15140 }, { "epoch": 0.19, "learning_rate": 1.8745647732833757e-05, "loss": 24.4309, "step": 15150 }, { "epoch": 0.19, "learning_rate": 1.8743715573394938e-05, "loss": 24.2619, "step": 15160 }, { "epoch": 0.19, "learning_rate": 1.8741782026721065e-05, "loss": 24.0684, "step": 15170 }, { "epoch": 0.19, "learning_rate": 1.8739847093118904e-05, "loss": 24.534, "step": 15180 }, { "epoch": 0.19, "learning_rate": 1.8737910772895443e-05, "loss": 23.9706, "step": 15190 }, { "epoch": 0.19, "learning_rate": 1.873597306635789e-05, "loss": 23.9937, "step": 15200 }, { "epoch": 0.19, "learning_rate": 1.8734033973813672e-05, "loss": 24.2465, "step": 15210 }, { "epoch": 0.19, "learning_rate": 1.8732093495570436e-05, "loss": 23.8885, "step": 15220 }, { "epoch": 0.19, "learning_rate": 1.873015163193605e-05, "loss": 24.047, "step": 15230 }, { "epoch": 0.19, "learning_rate": 1.8728208383218596e-05, "loss": 23.8952, "step": 15240 }, { "epoch": 0.19, "learning_rate": 1.8726263749726384e-05, "loss": 23.3622, "step": 15250 }, { "epoch": 0.19, "learning_rate": 1.872431773176794e-05, "loss": 23.3455, "step": 15260 }, { "epoch": 0.19, "learning_rate": 1.8722370329652006e-05, "loss": 23.3597, "step": 15270 }, { "epoch": 0.19, "learning_rate": 1.8720421543687554e-05, "loss": 23.8842, "step": 15280 }, { "epoch": 0.19, "learning_rate": 1.871847137418376e-05, "loss": 23.9839, "step": 15290 }, { "epoch": 0.19, "learning_rate": 1.871651982145004e-05, "loss": 23.8694, "step": 15300 }, { "epoch": 0.19, "learning_rate": 1.8714566885796006e-05, "loss": 24.0701, "step": 15310 }, { "epoch": 0.19, "learning_rate": 1.871261256753151e-05, "loss": 24.663, "step": 15320 }, { "epoch": 0.19, "learning_rate": 1.8710656866966607e-05, "loss": 24.0502, "step": 15330 }, { "epoch": 0.19, "learning_rate": 1.8708699784411585e-05, "loss": 24.6269, "step": 15340 }, { "epoch": 0.19, "learning_rate": 1.870674132017694e-05, "loss": 25.1045, "step": 15350 }, { "epoch": 0.19, "learning_rate": 1.8704781474573396e-05, "loss": 23.5605, "step": 15360 }, { "epoch": 0.19, "learning_rate": 1.8702820247911895e-05, "loss": 23.9095, "step": 15370 }, { "epoch": 0.19, "learning_rate": 1.8700857640503592e-05, "loss": 23.9402, "step": 15380 }, { "epoch": 0.19, "learning_rate": 1.869889365265986e-05, "loss": 24.0752, "step": 15390 }, { "epoch": 0.19, "learning_rate": 1.8696928284692304e-05, "loss": 24.348, "step": 15400 }, { "epoch": 0.19, "learning_rate": 1.8694961536912738e-05, "loss": 23.8515, "step": 15410 }, { "epoch": 0.19, "learning_rate": 1.8692993409633193e-05, "loss": 23.8668, "step": 15420 }, { "epoch": 0.19, "learning_rate": 1.8691023903165923e-05, "loss": 23.8773, "step": 15430 }, { "epoch": 0.19, "learning_rate": 1.8689053017823405e-05, "loss": 24.0178, "step": 15440 }, { "epoch": 0.19, "learning_rate": 1.8687080753918322e-05, "loss": 24.0156, "step": 15450 }, { "epoch": 0.19, "learning_rate": 1.868510711176359e-05, "loss": 23.9698, "step": 15460 }, { "epoch": 0.19, "learning_rate": 1.8683132091672334e-05, "loss": 24.0937, "step": 15470 }, { "epoch": 0.19, "learning_rate": 1.8681155693957905e-05, "loss": 23.7859, "step": 15480 }, { "epoch": 0.19, "learning_rate": 1.8679177918933867e-05, "loss": 23.322, "step": 15490 }, { "epoch": 0.19, "learning_rate": 1.8677198766914e-05, "loss": 24.3949, "step": 15500 }, { "epoch": 0.19, "learning_rate": 1.8675218238212315e-05, "loss": 24.1541, "step": 15510 }, { "epoch": 0.19, "learning_rate": 1.8673236333143023e-05, "loss": 24.0611, "step": 15520 }, { "epoch": 0.19, "learning_rate": 1.867125305202057e-05, "loss": 24.3713, "step": 15530 }, { "epoch": 0.19, "learning_rate": 1.8669268395159613e-05, "loss": 23.1617, "step": 15540 }, { "epoch": 0.19, "learning_rate": 1.8667282362875024e-05, "loss": 23.7409, "step": 15550 }, { "epoch": 0.19, "learning_rate": 1.8665294955481902e-05, "loss": 23.4481, "step": 15560 }, { "epoch": 0.19, "learning_rate": 1.8663306173295556e-05, "loss": 24.188, "step": 15570 }, { "epoch": 0.19, "learning_rate": 1.8661316016631516e-05, "loss": 23.3732, "step": 15580 }, { "epoch": 0.19, "learning_rate": 1.8659324485805537e-05, "loss": 24.2227, "step": 15590 }, { "epoch": 0.19, "learning_rate": 1.8657331581133577e-05, "loss": 24.233, "step": 15600 }, { "epoch": 0.19, "learning_rate": 1.8655337302931825e-05, "loss": 24.0666, "step": 15610 }, { "epoch": 0.19, "learning_rate": 1.8653341651516684e-05, "loss": 23.9881, "step": 15620 }, { "epoch": 0.19, "learning_rate": 1.865134462720477e-05, "loss": 24.5686, "step": 15630 }, { "epoch": 0.19, "learning_rate": 1.8649346230312924e-05, "loss": 24.0365, "step": 15640 }, { "epoch": 0.19, "learning_rate": 1.8647346461158204e-05, "loss": 24.4222, "step": 15650 }, { "epoch": 0.19, "learning_rate": 1.864534532005788e-05, "loss": 24.4591, "step": 15660 }, { "epoch": 0.19, "learning_rate": 1.8643342807329447e-05, "loss": 23.895, "step": 15670 }, { "epoch": 0.19, "learning_rate": 1.864133892329061e-05, "loss": 24.2932, "step": 15680 }, { "epoch": 0.19, "learning_rate": 1.8639333668259295e-05, "loss": 24.0708, "step": 15690 }, { "epoch": 0.19, "learning_rate": 1.863732704255365e-05, "loss": 24.2273, "step": 15700 }, { "epoch": 0.19, "learning_rate": 1.8635319046492032e-05, "loss": 24.2074, "step": 15710 }, { "epoch": 0.19, "learning_rate": 1.8633309680393026e-05, "loss": 24.2459, "step": 15720 }, { "epoch": 0.19, "learning_rate": 1.863129894457542e-05, "loss": 23.3597, "step": 15730 }, { "epoch": 0.19, "learning_rate": 1.8629286839358232e-05, "loss": 24.1499, "step": 15740 }, { "epoch": 0.19, "learning_rate": 1.8627273365060693e-05, "loss": 23.7487, "step": 15750 }, { "epoch": 0.19, "learning_rate": 1.8625258522002253e-05, "loss": 23.8837, "step": 15760 }, { "epoch": 0.19, "learning_rate": 1.8623242310502573e-05, "loss": 24.2562, "step": 15770 }, { "epoch": 0.19, "learning_rate": 1.8621224730881535e-05, "loss": 23.6537, "step": 15780 }, { "epoch": 0.19, "learning_rate": 1.861920578345924e-05, "loss": 24.5953, "step": 15790 }, { "epoch": 0.19, "learning_rate": 1.8617185468556008e-05, "loss": 23.6728, "step": 15800 }, { "epoch": 0.19, "learning_rate": 1.8615163786492365e-05, "loss": 23.5996, "step": 15810 }, { "epoch": 0.19, "learning_rate": 1.861314073758907e-05, "loss": 23.7399, "step": 15820 }, { "epoch": 0.19, "learning_rate": 1.8611116322167077e-05, "loss": 23.7296, "step": 15830 }, { "epoch": 0.19, "learning_rate": 1.860909054054758e-05, "loss": 23.8029, "step": 15840 }, { "epoch": 0.19, "learning_rate": 1.860706339305198e-05, "loss": 24.0293, "step": 15850 }, { "epoch": 0.2, "learning_rate": 1.860503488000189e-05, "loss": 24.1921, "step": 15860 }, { "epoch": 0.2, "learning_rate": 1.8603005001719146e-05, "loss": 23.8959, "step": 15870 }, { "epoch": 0.2, "learning_rate": 1.8600973758525797e-05, "loss": 23.5305, "step": 15880 }, { "epoch": 0.2, "learning_rate": 1.859894115074411e-05, "loss": 23.693, "step": 15890 }, { "epoch": 0.2, "learning_rate": 1.8596907178696573e-05, "loss": 23.5212, "step": 15900 }, { "epoch": 0.2, "learning_rate": 1.859487184270588e-05, "loss": 24.1797, "step": 15910 }, { "epoch": 0.2, "learning_rate": 1.8592835143094953e-05, "loss": 24.2334, "step": 15920 }, { "epoch": 0.2, "learning_rate": 1.859079708018692e-05, "loss": 24.0607, "step": 15930 }, { "epoch": 0.2, "learning_rate": 1.8588757654305128e-05, "loss": 23.5306, "step": 15940 }, { "epoch": 0.2, "learning_rate": 1.8586716865773155e-05, "loss": 24.4817, "step": 15950 }, { "epoch": 0.2, "learning_rate": 1.8584674714914765e-05, "loss": 23.6565, "step": 15960 }, { "epoch": 0.2, "learning_rate": 1.858263120205397e-05, "loss": 23.794, "step": 15970 }, { "epoch": 0.2, "learning_rate": 1.8580586327514974e-05, "loss": 23.9813, "step": 15980 }, { "epoch": 0.2, "learning_rate": 1.8578540091622213e-05, "loss": 24.4042, "step": 15990 }, { "epoch": 0.2, "learning_rate": 1.8576492494700332e-05, "loss": 24.386, "step": 16000 }, { "epoch": 0.2, "learning_rate": 1.857444353707419e-05, "loss": 24.1712, "step": 16010 }, { "epoch": 0.2, "learning_rate": 1.8572393219068866e-05, "loss": 24.7082, "step": 16020 }, { "epoch": 0.2, "learning_rate": 1.8570341541009657e-05, "loss": 24.3568, "step": 16030 }, { "epoch": 0.2, "learning_rate": 1.8568288503222064e-05, "loss": 24.2135, "step": 16040 }, { "epoch": 0.2, "learning_rate": 1.8566234106031816e-05, "loss": 24.3724, "step": 16050 }, { "epoch": 0.2, "learning_rate": 1.8564178349764855e-05, "loss": 24.2598, "step": 16060 }, { "epoch": 0.2, "learning_rate": 1.8562121234747338e-05, "loss": 23.7695, "step": 16070 }, { "epoch": 0.2, "learning_rate": 1.8560062761305632e-05, "loss": 23.7325, "step": 16080 }, { "epoch": 0.2, "learning_rate": 1.8558002929766328e-05, "loss": 23.589, "step": 16090 }, { "epoch": 0.2, "learning_rate": 1.8555941740456228e-05, "loss": 24.3016, "step": 16100 }, { "epoch": 0.2, "learning_rate": 1.8553879193702352e-05, "loss": 23.9359, "step": 16110 }, { "epoch": 0.2, "learning_rate": 1.8551815289831935e-05, "loss": 23.2878, "step": 16120 }, { "epoch": 0.2, "learning_rate": 1.8549750029172418e-05, "loss": 24.2895, "step": 16130 }, { "epoch": 0.2, "learning_rate": 1.8547683412051468e-05, "loss": 23.7602, "step": 16140 }, { "epoch": 0.2, "learning_rate": 1.854561543879697e-05, "loss": 23.9311, "step": 16150 }, { "epoch": 0.2, "learning_rate": 1.8543546109737013e-05, "loss": 24.0611, "step": 16160 }, { "epoch": 0.2, "learning_rate": 1.854147542519991e-05, "loss": 24.3256, "step": 16170 }, { "epoch": 0.2, "learning_rate": 1.8539403385514184e-05, "loss": 23.4808, "step": 16180 }, { "epoch": 0.2, "learning_rate": 1.8537329991008574e-05, "loss": 23.437, "step": 16190 }, { "epoch": 0.2, "learning_rate": 1.8535255242012033e-05, "loss": 24.1418, "step": 16200 }, { "epoch": 0.2, "learning_rate": 1.8533179138853735e-05, "loss": 23.315, "step": 16210 }, { "epoch": 0.2, "learning_rate": 1.8531101681863063e-05, "loss": 24.0386, "step": 16220 }, { "epoch": 0.2, "learning_rate": 1.8529022871369612e-05, "loss": 23.7205, "step": 16230 }, { "epoch": 0.2, "learning_rate": 1.8526942707703202e-05, "loss": 23.9643, "step": 16240 }, { "epoch": 0.2, "learning_rate": 1.852486119119386e-05, "loss": 24.2543, "step": 16250 }, { "epoch": 0.2, "learning_rate": 1.8522778322171822e-05, "loss": 23.9938, "step": 16260 }, { "epoch": 0.2, "learning_rate": 1.8520694100967558e-05, "loss": 23.724, "step": 16270 }, { "epoch": 0.2, "learning_rate": 1.851860852791173e-05, "loss": 23.9814, "step": 16280 }, { "epoch": 0.2, "learning_rate": 1.851652160333523e-05, "loss": 24.321, "step": 16290 }, { "epoch": 0.2, "learning_rate": 1.851443332756916e-05, "loss": 24.0771, "step": 16300 }, { "epoch": 0.2, "learning_rate": 1.851234370094483e-05, "loss": 23.4897, "step": 16310 }, { "epoch": 0.2, "learning_rate": 1.8510252723793777e-05, "loss": 23.8505, "step": 16320 }, { "epoch": 0.2, "learning_rate": 1.8508160396447737e-05, "loss": 23.8935, "step": 16330 }, { "epoch": 0.2, "learning_rate": 1.8506066719238677e-05, "loss": 23.916, "step": 16340 }, { "epoch": 0.2, "learning_rate": 1.8503971692498764e-05, "loss": 24.0626, "step": 16350 }, { "epoch": 0.2, "learning_rate": 1.8501875316560388e-05, "loss": 24.2745, "step": 16360 }, { "epoch": 0.2, "learning_rate": 1.849977759175615e-05, "loss": 23.8125, "step": 16370 }, { "epoch": 0.2, "learning_rate": 1.849767851841886e-05, "loss": 24.3066, "step": 16380 }, { "epoch": 0.2, "learning_rate": 1.8495578096881548e-05, "loss": 24.4741, "step": 16390 }, { "epoch": 0.2, "learning_rate": 1.849347632747746e-05, "loss": 24.2617, "step": 16400 }, { "epoch": 0.2, "learning_rate": 1.849137321054005e-05, "loss": 23.3189, "step": 16410 }, { "epoch": 0.2, "learning_rate": 1.8489268746402992e-05, "loss": 23.8828, "step": 16420 }, { "epoch": 0.2, "learning_rate": 1.8487162935400162e-05, "loss": 23.4491, "step": 16430 }, { "epoch": 0.2, "learning_rate": 1.848505577786566e-05, "loss": 24.4642, "step": 16440 }, { "epoch": 0.2, "learning_rate": 1.8482947274133805e-05, "loss": 23.834, "step": 16450 }, { "epoch": 0.2, "learning_rate": 1.848083742453912e-05, "loss": 24.2819, "step": 16460 }, { "epoch": 0.2, "learning_rate": 1.8478726229416334e-05, "loss": 24.2955, "step": 16470 }, { "epoch": 0.2, "learning_rate": 1.8476613689100406e-05, "loss": 23.8953, "step": 16480 }, { "epoch": 0.2, "learning_rate": 1.8474499803926497e-05, "loss": 23.867, "step": 16490 }, { "epoch": 0.2, "learning_rate": 1.847238457422999e-05, "loss": 23.5175, "step": 16500 }, { "epoch": 0.2, "learning_rate": 1.847026800034648e-05, "loss": 23.782, "step": 16510 }, { "epoch": 0.2, "learning_rate": 1.8468150082611767e-05, "loss": 24.1402, "step": 16520 }, { "epoch": 0.2, "learning_rate": 1.846603082136187e-05, "loss": 24.4609, "step": 16530 }, { "epoch": 0.2, "learning_rate": 1.8463910216933017e-05, "loss": 23.6647, "step": 16540 }, { "epoch": 0.2, "learning_rate": 1.846178826966166e-05, "loss": 24.0723, "step": 16550 }, { "epoch": 0.2, "learning_rate": 1.8459664979884455e-05, "loss": 24.3833, "step": 16560 }, { "epoch": 0.2, "learning_rate": 1.8457540347938265e-05, "loss": 23.9932, "step": 16570 }, { "epoch": 0.2, "learning_rate": 1.8455414374160187e-05, "loss": 23.2575, "step": 16580 }, { "epoch": 0.2, "learning_rate": 1.8453287058887505e-05, "loss": 23.8635, "step": 16590 }, { "epoch": 0.2, "learning_rate": 1.845115840245774e-05, "loss": 23.8676, "step": 16600 }, { "epoch": 0.2, "learning_rate": 1.84490284052086e-05, "loss": 23.6483, "step": 16610 }, { "epoch": 0.2, "learning_rate": 1.8446897067478036e-05, "loss": 23.902, "step": 16620 }, { "epoch": 0.2, "learning_rate": 1.8444764389604183e-05, "loss": 24.0144, "step": 16630 }, { "epoch": 0.2, "learning_rate": 1.844263037192541e-05, "loss": 23.6505, "step": 16640 }, { "epoch": 0.2, "learning_rate": 1.8440495014780278e-05, "loss": 23.384, "step": 16650 }, { "epoch": 0.2, "learning_rate": 1.8438358318507585e-05, "loss": 24.4069, "step": 16660 }, { "epoch": 0.21, "learning_rate": 1.8436220283446324e-05, "loss": 24.0876, "step": 16670 }, { "epoch": 0.21, "learning_rate": 1.8434080909935703e-05, "loss": 23.986, "step": 16680 }, { "epoch": 0.21, "learning_rate": 1.8431940198315148e-05, "loss": 23.9222, "step": 16690 }, { "epoch": 0.21, "learning_rate": 1.842979814892429e-05, "loss": 23.4488, "step": 16700 }, { "epoch": 0.21, "learning_rate": 1.8427654762102984e-05, "loss": 23.6268, "step": 16710 }, { "epoch": 0.21, "learning_rate": 1.8425510038191278e-05, "loss": 24.2058, "step": 16720 }, { "epoch": 0.21, "learning_rate": 1.8423363977529448e-05, "loss": 23.5304, "step": 16730 }, { "epoch": 0.21, "learning_rate": 1.842121658045798e-05, "loss": 24.2827, "step": 16740 }, { "epoch": 0.21, "learning_rate": 1.8419067847317568e-05, "loss": 23.5699, "step": 16750 }, { "epoch": 0.21, "learning_rate": 1.8416917778449123e-05, "loss": 24.1352, "step": 16760 }, { "epoch": 0.21, "learning_rate": 1.8414766374193754e-05, "loss": 24.0961, "step": 16770 }, { "epoch": 0.21, "learning_rate": 1.8412613634892803e-05, "loss": 23.9519, "step": 16780 }, { "epoch": 0.21, "learning_rate": 1.8410459560887807e-05, "loss": 23.8531, "step": 16790 }, { "epoch": 0.21, "learning_rate": 1.8408304152520526e-05, "loss": 23.5547, "step": 16800 }, { "epoch": 0.21, "learning_rate": 1.8406147410132918e-05, "loss": 23.7533, "step": 16810 }, { "epoch": 0.21, "learning_rate": 1.840398933406717e-05, "loss": 24.5511, "step": 16820 }, { "epoch": 0.21, "learning_rate": 1.8401829924665668e-05, "loss": 24.5971, "step": 16830 }, { "epoch": 0.21, "learning_rate": 1.8399669182271012e-05, "loss": 23.7855, "step": 16840 }, { "epoch": 0.21, "learning_rate": 1.8397507107226016e-05, "loss": 23.3549, "step": 16850 }, { "epoch": 0.21, "learning_rate": 1.8395343699873706e-05, "loss": 24.1113, "step": 16860 }, { "epoch": 0.21, "learning_rate": 1.839317896055732e-05, "loss": 23.9562, "step": 16870 }, { "epoch": 0.21, "learning_rate": 1.8391012889620297e-05, "loss": 23.9809, "step": 16880 }, { "epoch": 0.21, "learning_rate": 1.83888454874063e-05, "loss": 24.1432, "step": 16890 }, { "epoch": 0.21, "learning_rate": 1.8386676754259196e-05, "loss": 23.8909, "step": 16900 }, { "epoch": 0.21, "learning_rate": 1.838450669052307e-05, "loss": 23.6748, "step": 16910 }, { "epoch": 0.21, "learning_rate": 1.8382335296542207e-05, "loss": 23.7545, "step": 16920 }, { "epoch": 0.21, "learning_rate": 1.8380162572661117e-05, "loss": 24.4153, "step": 16930 }, { "epoch": 0.21, "learning_rate": 1.837798851922451e-05, "loss": 24.502, "step": 16940 }, { "epoch": 0.21, "learning_rate": 1.8375813136577314e-05, "loss": 23.8558, "step": 16950 }, { "epoch": 0.21, "learning_rate": 1.8373636425064658e-05, "loss": 24.1399, "step": 16960 }, { "epoch": 0.21, "learning_rate": 1.8371458385031895e-05, "loss": 24.2652, "step": 16970 }, { "epoch": 0.21, "learning_rate": 1.8369279016824574e-05, "loss": 23.8935, "step": 16980 }, { "epoch": 0.21, "learning_rate": 1.8367098320788474e-05, "loss": 23.9683, "step": 16990 }, { "epoch": 0.21, "learning_rate": 1.8364916297269562e-05, "loss": 24.1132, "step": 17000 }, { "epoch": 0.21, "learning_rate": 1.8362732946614037e-05, "loss": 24.2998, "step": 17010 }, { "epoch": 0.21, "learning_rate": 1.8360548269168293e-05, "loss": 23.7547, "step": 17020 }, { "epoch": 0.21, "learning_rate": 1.8358362265278938e-05, "loss": 23.6924, "step": 17030 }, { "epoch": 0.21, "learning_rate": 1.8356174935292802e-05, "loss": 24.2883, "step": 17040 }, { "epoch": 0.21, "learning_rate": 1.835398627955691e-05, "loss": 23.7544, "step": 17050 }, { "epoch": 0.21, "learning_rate": 1.83517962984185e-05, "loss": 24.2243, "step": 17060 }, { "epoch": 0.21, "learning_rate": 1.8349604992225032e-05, "loss": 23.8735, "step": 17070 }, { "epoch": 0.21, "learning_rate": 1.8347412361324163e-05, "loss": 24.0548, "step": 17080 }, { "epoch": 0.21, "learning_rate": 1.834521840606376e-05, "loss": 23.3207, "step": 17090 }, { "epoch": 0.21, "learning_rate": 1.8343023126791915e-05, "loss": 23.7148, "step": 17100 }, { "epoch": 0.21, "learning_rate": 1.8340826523856915e-05, "loss": 24.0567, "step": 17110 }, { "epoch": 0.21, "learning_rate": 1.833862859760726e-05, "loss": 23.6406, "step": 17120 }, { "epoch": 0.21, "learning_rate": 1.833642934839167e-05, "loss": 24.26, "step": 17130 }, { "epoch": 0.21, "learning_rate": 1.8334228776559062e-05, "loss": 23.8457, "step": 17140 }, { "epoch": 0.21, "learning_rate": 1.8332026882458567e-05, "loss": 24.27, "step": 17150 }, { "epoch": 0.21, "learning_rate": 1.8329823666439527e-05, "loss": 24.117, "step": 17160 }, { "epoch": 0.21, "learning_rate": 1.8327619128851498e-05, "loss": 23.7001, "step": 17170 }, { "epoch": 0.21, "learning_rate": 1.8325413270044235e-05, "loss": 23.9541, "step": 17180 }, { "epoch": 0.21, "learning_rate": 1.832320609036771e-05, "loss": 23.6946, "step": 17190 }, { "epoch": 0.21, "learning_rate": 1.8320997590172108e-05, "loss": 23.5699, "step": 17200 }, { "epoch": 0.21, "learning_rate": 1.8318787769807814e-05, "loss": 24.3501, "step": 17210 }, { "epoch": 0.21, "learning_rate": 1.831657662962543e-05, "loss": 23.5042, "step": 17220 }, { "epoch": 0.21, "learning_rate": 1.8314364169975762e-05, "loss": 23.7022, "step": 17230 }, { "epoch": 0.21, "learning_rate": 1.8312150391209828e-05, "loss": 23.5235, "step": 17240 }, { "epoch": 0.21, "learning_rate": 1.830993529367886e-05, "loss": 24.3496, "step": 17250 }, { "epoch": 0.21, "learning_rate": 1.830771887773429e-05, "loss": 23.9044, "step": 17260 }, { "epoch": 0.21, "learning_rate": 1.8305501143727765e-05, "loss": 23.8575, "step": 17270 }, { "epoch": 0.21, "learning_rate": 1.830328209201114e-05, "loss": 23.8275, "step": 17280 }, { "epoch": 0.21, "learning_rate": 1.8301061722936478e-05, "loss": 24.0908, "step": 17290 }, { "epoch": 0.21, "learning_rate": 1.8298840036856055e-05, "loss": 23.5373, "step": 17300 }, { "epoch": 0.21, "learning_rate": 1.8296617034122346e-05, "loss": 23.8252, "step": 17310 }, { "epoch": 0.21, "learning_rate": 1.829439271508805e-05, "loss": 23.8611, "step": 17320 }, { "epoch": 0.21, "learning_rate": 1.829216708010606e-05, "loss": 23.4627, "step": 17330 }, { "epoch": 0.21, "learning_rate": 1.828994012952949e-05, "loss": 23.7022, "step": 17340 }, { "epoch": 0.21, "learning_rate": 1.828771186371165e-05, "loss": 23.7899, "step": 17350 }, { "epoch": 0.21, "learning_rate": 1.828548228300607e-05, "loss": 23.8543, "step": 17360 }, { "epoch": 0.21, "learning_rate": 1.8283251387766487e-05, "loss": 24.001, "step": 17370 }, { "epoch": 0.21, "learning_rate": 1.8281019178346838e-05, "loss": 23.7583, "step": 17380 }, { "epoch": 0.21, "learning_rate": 1.8278785655101283e-05, "loss": 23.804, "step": 17390 }, { "epoch": 0.21, "learning_rate": 1.8276550818384172e-05, "loss": 24.1401, "step": 17400 }, { "epoch": 0.21, "learning_rate": 1.827431466855008e-05, "loss": 23.0606, "step": 17410 }, { "epoch": 0.21, "learning_rate": 1.827207720595378e-05, "loss": 24.1444, "step": 17420 }, { "epoch": 0.21, "learning_rate": 1.826983843095025e-05, "loss": 22.5601, "step": 17430 }, { "epoch": 0.21, "learning_rate": 1.82675983438947e-05, "loss": 24.2274, "step": 17440 }, { "epoch": 0.21, "learning_rate": 1.826535694514252e-05, "loss": 23.5485, "step": 17450 }, { "epoch": 0.21, "learning_rate": 1.826311423504932e-05, "loss": 23.7139, "step": 17460 }, { "epoch": 0.21, "learning_rate": 1.8260870213970925e-05, "loss": 23.9576, "step": 17470 }, { "epoch": 0.21, "learning_rate": 1.825862488226335e-05, "loss": 23.8957, "step": 17480 }, { "epoch": 0.22, "learning_rate": 1.825637824028283e-05, "loss": 23.824, "step": 17490 }, { "epoch": 0.22, "learning_rate": 1.825413028838581e-05, "loss": 24.4632, "step": 17500 }, { "epoch": 0.22, "learning_rate": 1.825188102692894e-05, "loss": 24.6196, "step": 17510 }, { "epoch": 0.22, "learning_rate": 1.8249630456269077e-05, "loss": 23.7252, "step": 17520 }, { "epoch": 0.22, "learning_rate": 1.8247378576763278e-05, "loss": 23.8179, "step": 17530 }, { "epoch": 0.22, "learning_rate": 1.824512538876882e-05, "loss": 23.6786, "step": 17540 }, { "epoch": 0.22, "learning_rate": 1.8242870892643187e-05, "loss": 23.5444, "step": 17550 }, { "epoch": 0.22, "learning_rate": 1.8240615088744063e-05, "loss": 23.6507, "step": 17560 }, { "epoch": 0.22, "learning_rate": 1.823835797742934e-05, "loss": 24.0603, "step": 17570 }, { "epoch": 0.22, "learning_rate": 1.8236099559057124e-05, "loss": 23.7652, "step": 17580 }, { "epoch": 0.22, "learning_rate": 1.823383983398572e-05, "loss": 23.9412, "step": 17590 }, { "epoch": 0.22, "learning_rate": 1.823157880257365e-05, "loss": 23.5698, "step": 17600 }, { "epoch": 0.22, "learning_rate": 1.8229316465179635e-05, "loss": 24.0044, "step": 17610 }, { "epoch": 0.22, "learning_rate": 1.822705282216261e-05, "loss": 23.0581, "step": 17620 }, { "epoch": 0.22, "learning_rate": 1.822478787388171e-05, "loss": 23.6871, "step": 17630 }, { "epoch": 0.22, "learning_rate": 1.822252162069628e-05, "loss": 23.7919, "step": 17640 }, { "epoch": 0.22, "learning_rate": 1.822025406296587e-05, "loss": 24.5174, "step": 17650 }, { "epoch": 0.22, "learning_rate": 1.8217985201050253e-05, "loss": 24.0862, "step": 17660 }, { "epoch": 0.22, "learning_rate": 1.821571503530938e-05, "loss": 23.5376, "step": 17670 }, { "epoch": 0.22, "learning_rate": 1.8213443566103434e-05, "loss": 23.4822, "step": 17680 }, { "epoch": 0.22, "learning_rate": 1.821117079379279e-05, "loss": 24.3738, "step": 17690 }, { "epoch": 0.22, "learning_rate": 1.8208896718738037e-05, "loss": 24.2225, "step": 17700 }, { "epoch": 0.22, "learning_rate": 1.820662134129997e-05, "loss": 24.1555, "step": 17710 }, { "epoch": 0.22, "learning_rate": 1.8204344661839583e-05, "loss": 24.1412, "step": 17720 }, { "epoch": 0.22, "learning_rate": 1.8202066680718093e-05, "loss": 23.7333, "step": 17730 }, { "epoch": 0.22, "learning_rate": 1.8199787398296904e-05, "loss": 24.0452, "step": 17740 }, { "epoch": 0.22, "learning_rate": 1.819750681493764e-05, "loss": 24.6777, "step": 17750 }, { "epoch": 0.22, "learning_rate": 1.8195224931002126e-05, "loss": 23.6794, "step": 17760 }, { "epoch": 0.22, "learning_rate": 1.8192941746852398e-05, "loss": 23.7583, "step": 17770 }, { "epoch": 0.22, "learning_rate": 1.819065726285069e-05, "loss": 23.3713, "step": 17780 }, { "epoch": 0.22, "learning_rate": 1.818837147935945e-05, "loss": 23.4188, "step": 17790 }, { "epoch": 0.22, "learning_rate": 1.8186084396741327e-05, "loss": 24.2589, "step": 17800 }, { "epoch": 0.22, "learning_rate": 1.818379601535918e-05, "loss": 24.0021, "step": 17810 }, { "epoch": 0.22, "learning_rate": 1.818150633557607e-05, "loss": 24.1293, "step": 17820 }, { "epoch": 0.22, "learning_rate": 1.817921535775527e-05, "loss": 24.3031, "step": 17830 }, { "epoch": 0.22, "learning_rate": 1.8176923082260253e-05, "loss": 24.2465, "step": 17840 }, { "epoch": 0.22, "learning_rate": 1.81746295094547e-05, "loss": 24.618, "step": 17850 }, { "epoch": 0.22, "learning_rate": 1.8172334639702502e-05, "loss": 24.4909, "step": 17860 }, { "epoch": 0.22, "learning_rate": 1.817003847336775e-05, "loss": 24.2764, "step": 17870 }, { "epoch": 0.22, "learning_rate": 1.8167741010814735e-05, "loss": 24.3765, "step": 17880 }, { "epoch": 0.22, "learning_rate": 1.816544225240797e-05, "loss": 24.3581, "step": 17890 }, { "epoch": 0.22, "learning_rate": 1.816314219851216e-05, "loss": 24.2769, "step": 17900 }, { "epoch": 0.22, "learning_rate": 1.8160840849492226e-05, "loss": 23.9033, "step": 17910 }, { "epoch": 0.22, "learning_rate": 1.815853820571328e-05, "loss": 24.0478, "step": 17920 }, { "epoch": 0.22, "learning_rate": 1.815623426754066e-05, "loss": 24.2995, "step": 17930 }, { "epoch": 0.22, "learning_rate": 1.8153929035339887e-05, "loss": 23.6785, "step": 17940 }, { "epoch": 0.22, "learning_rate": 1.81516225094767e-05, "loss": 23.7395, "step": 17950 }, { "epoch": 0.22, "learning_rate": 1.8149314690317043e-05, "loss": 23.6728, "step": 17960 }, { "epoch": 0.22, "learning_rate": 1.8147005578227066e-05, "loss": 24.1397, "step": 17970 }, { "epoch": 0.22, "learning_rate": 1.8144695173573116e-05, "loss": 23.9106, "step": 17980 }, { "epoch": 0.22, "learning_rate": 1.814238347672175e-05, "loss": 23.607, "step": 17990 }, { "epoch": 0.22, "learning_rate": 1.8140070488039738e-05, "loss": 23.8368, "step": 18000 }, { "epoch": 0.22, "learning_rate": 1.8137756207894043e-05, "loss": 23.7693, "step": 18010 }, { "epoch": 0.22, "learning_rate": 1.8135440636651832e-05, "loss": 23.6231, "step": 18020 }, { "epoch": 0.22, "learning_rate": 1.813312377468049e-05, "loss": 23.5258, "step": 18030 }, { "epoch": 0.22, "learning_rate": 1.81308056223476e-05, "loss": 23.9659, "step": 18040 }, { "epoch": 0.22, "learning_rate": 1.8128486180020938e-05, "loss": 23.6504, "step": 18050 }, { "epoch": 0.22, "learning_rate": 1.8126165448068507e-05, "loss": 24.1806, "step": 18060 }, { "epoch": 0.22, "learning_rate": 1.8123843426858498e-05, "loss": 24.1402, "step": 18070 }, { "epoch": 0.22, "learning_rate": 1.812152011675931e-05, "loss": 23.731, "step": 18080 }, { "epoch": 0.22, "learning_rate": 1.8119195518139548e-05, "loss": 24.8028, "step": 18090 }, { "epoch": 0.22, "learning_rate": 1.811686963136803e-05, "loss": 24.1284, "step": 18100 }, { "epoch": 0.22, "learning_rate": 1.8114542456813758e-05, "loss": 24.2254, "step": 18110 }, { "epoch": 0.22, "learning_rate": 1.8112213994845954e-05, "loss": 24.104, "step": 18120 }, { "epoch": 0.22, "learning_rate": 1.810988424583404e-05, "loss": 24.3644, "step": 18130 }, { "epoch": 0.22, "learning_rate": 1.810755321014765e-05, "loss": 24.6181, "step": 18140 }, { "epoch": 0.22, "learning_rate": 1.8105220888156606e-05, "loss": 24.0764, "step": 18150 }, { "epoch": 0.22, "learning_rate": 1.8102887280230944e-05, "loss": 24.2542, "step": 18160 }, { "epoch": 0.22, "learning_rate": 1.8100552386740905e-05, "loss": 23.3886, "step": 18170 }, { "epoch": 0.22, "learning_rate": 1.809821620805693e-05, "loss": 24.2656, "step": 18180 }, { "epoch": 0.22, "learning_rate": 1.809587874454967e-05, "loss": 24.4582, "step": 18190 }, { "epoch": 0.22, "learning_rate": 1.8093539996589966e-05, "loss": 23.9831, "step": 18200 }, { "epoch": 0.22, "learning_rate": 1.8091199964548883e-05, "loss": 23.879, "step": 18210 }, { "epoch": 0.22, "learning_rate": 1.8088858648797673e-05, "loss": 23.7599, "step": 18220 }, { "epoch": 0.22, "learning_rate": 1.80865160497078e-05, "loss": 23.7514, "step": 18230 }, { "epoch": 0.22, "learning_rate": 1.8084172167650926e-05, "loss": 23.4744, "step": 18240 }, { "epoch": 0.22, "learning_rate": 1.8081827002998924e-05, "loss": 23.6152, "step": 18250 }, { "epoch": 0.22, "learning_rate": 1.8079480556123866e-05, "loss": 24.3701, "step": 18260 }, { "epoch": 0.22, "learning_rate": 1.8077132827398027e-05, "loss": 23.3391, "step": 18270 }, { "epoch": 0.22, "learning_rate": 1.8074783817193883e-05, "loss": 23.5094, "step": 18280 }, { "epoch": 0.22, "learning_rate": 1.807243352588412e-05, "loss": 23.8844, "step": 18290 }, { "epoch": 0.23, "learning_rate": 1.807008195384162e-05, "loss": 23.536, "step": 18300 }, { "epoch": 0.23, "learning_rate": 1.8067729101439476e-05, "loss": 23.4512, "step": 18310 }, { "epoch": 0.23, "learning_rate": 1.806537496905098e-05, "loss": 24.1148, "step": 18320 }, { "epoch": 0.23, "learning_rate": 1.8063019557049624e-05, "loss": 23.7717, "step": 18330 }, { "epoch": 0.23, "learning_rate": 1.806066286580911e-05, "loss": 24.2656, "step": 18340 }, { "epoch": 0.23, "learning_rate": 1.8058304895703336e-05, "loss": 24.2649, "step": 18350 }, { "epoch": 0.23, "learning_rate": 1.8055945647106403e-05, "loss": 23.4811, "step": 18360 }, { "epoch": 0.23, "learning_rate": 1.8053585120392627e-05, "loss": 24.7796, "step": 18370 }, { "epoch": 0.23, "learning_rate": 1.8051223315936507e-05, "loss": 23.7643, "step": 18380 }, { "epoch": 0.23, "learning_rate": 1.8048860234112768e-05, "loss": 24.5247, "step": 18390 }, { "epoch": 0.23, "learning_rate": 1.804649587529631e-05, "loss": 24.3004, "step": 18400 }, { "epoch": 0.23, "learning_rate": 1.8044130239862262e-05, "loss": 23.8342, "step": 18410 }, { "epoch": 0.23, "learning_rate": 1.8041763328185937e-05, "loss": 24.0461, "step": 18420 }, { "epoch": 0.23, "learning_rate": 1.8039395140642867e-05, "loss": 24.1921, "step": 18430 }, { "epoch": 0.23, "learning_rate": 1.8037025677608765e-05, "loss": 23.8664, "step": 18440 }, { "epoch": 0.23, "learning_rate": 1.8034654939459566e-05, "loss": 24.5391, "step": 18450 }, { "epoch": 0.23, "learning_rate": 1.80322829265714e-05, "loss": 23.9673, "step": 18460 }, { "epoch": 0.23, "learning_rate": 1.8029909639320593e-05, "loss": 24.2394, "step": 18470 }, { "epoch": 0.23, "learning_rate": 1.8027535078083684e-05, "loss": 24.127, "step": 18480 }, { "epoch": 0.23, "learning_rate": 1.802515924323741e-05, "loss": 24.0286, "step": 18490 }, { "epoch": 0.23, "learning_rate": 1.8022782135158707e-05, "loss": 23.754, "step": 18500 }, { "epoch": 0.23, "learning_rate": 1.8020403754224715e-05, "loss": 24.272, "step": 18510 }, { "epoch": 0.23, "learning_rate": 1.801802410081278e-05, "loss": 23.4162, "step": 18520 }, { "epoch": 0.23, "learning_rate": 1.801564317530044e-05, "loss": 23.476, "step": 18530 }, { "epoch": 0.23, "learning_rate": 1.8013260978065444e-05, "loss": 23.7751, "step": 18540 }, { "epoch": 0.23, "learning_rate": 1.801087750948574e-05, "loss": 24.0607, "step": 18550 }, { "epoch": 0.23, "learning_rate": 1.800849276993948e-05, "loss": 24.021, "step": 18560 }, { "epoch": 0.23, "learning_rate": 1.8006106759805013e-05, "loss": 24.6052, "step": 18570 }, { "epoch": 0.23, "learning_rate": 1.800371947946089e-05, "loss": 24.1545, "step": 18580 }, { "epoch": 0.23, "learning_rate": 1.8001330929285866e-05, "loss": 24.2362, "step": 18590 }, { "epoch": 0.23, "learning_rate": 1.7998941109658895e-05, "loss": 24.2185, "step": 18600 }, { "epoch": 0.23, "learning_rate": 1.799655002095914e-05, "loss": 24.1238, "step": 18610 }, { "epoch": 0.23, "learning_rate": 1.7994157663565953e-05, "loss": 23.963, "step": 18620 }, { "epoch": 0.23, "learning_rate": 1.7991764037858894e-05, "loss": 23.8807, "step": 18630 }, { "epoch": 0.23, "learning_rate": 1.798936914421773e-05, "loss": 24.2977, "step": 18640 }, { "epoch": 0.23, "learning_rate": 1.798697298302242e-05, "loss": 23.8535, "step": 18650 }, { "epoch": 0.23, "learning_rate": 1.7984575554653124e-05, "loss": 24.0203, "step": 18660 }, { "epoch": 0.23, "learning_rate": 1.7982176859490206e-05, "loss": 23.8376, "step": 18670 }, { "epoch": 0.23, "learning_rate": 1.797977689791424e-05, "loss": 23.8696, "step": 18680 }, { "epoch": 0.23, "learning_rate": 1.7977375670305984e-05, "loss": 24.0269, "step": 18690 }, { "epoch": 0.23, "learning_rate": 1.7974973177046406e-05, "loss": 23.6632, "step": 18700 }, { "epoch": 0.23, "learning_rate": 1.7972569418516675e-05, "loss": 23.8161, "step": 18710 }, { "epoch": 0.23, "learning_rate": 1.7970164395098157e-05, "loss": 23.6226, "step": 18720 }, { "epoch": 0.23, "learning_rate": 1.7967758107172425e-05, "loss": 23.5747, "step": 18730 }, { "epoch": 0.23, "learning_rate": 1.796535055512125e-05, "loss": 23.7762, "step": 18740 }, { "epoch": 0.23, "learning_rate": 1.79629417393266e-05, "loss": 24.4549, "step": 18750 }, { "epoch": 0.23, "learning_rate": 1.7960531660170642e-05, "loss": 24.2132, "step": 18760 }, { "epoch": 0.23, "learning_rate": 1.7958120318035754e-05, "loss": 24.0426, "step": 18770 }, { "epoch": 0.23, "learning_rate": 1.79557077133045e-05, "loss": 23.7172, "step": 18780 }, { "epoch": 0.23, "learning_rate": 1.7953293846359665e-05, "loss": 24.1263, "step": 18790 }, { "epoch": 0.23, "learning_rate": 1.795087871758421e-05, "loss": 23.8344, "step": 18800 }, { "epoch": 0.23, "learning_rate": 1.7948462327361306e-05, "loss": 23.3797, "step": 18810 }, { "epoch": 0.23, "learning_rate": 1.7946044676074334e-05, "loss": 24.4189, "step": 18820 }, { "epoch": 0.23, "learning_rate": 1.7943625764106864e-05, "loss": 24.0337, "step": 18830 }, { "epoch": 0.23, "learning_rate": 1.7941205591842666e-05, "loss": 24.0543, "step": 18840 }, { "epoch": 0.23, "learning_rate": 1.7938784159665715e-05, "loss": 23.8833, "step": 18850 }, { "epoch": 0.23, "learning_rate": 1.7936361467960184e-05, "loss": 24.3626, "step": 18860 }, { "epoch": 0.23, "learning_rate": 1.7933937517110442e-05, "loss": 24.4091, "step": 18870 }, { "epoch": 0.23, "learning_rate": 1.7931512307501063e-05, "loss": 23.4575, "step": 18880 }, { "epoch": 0.23, "learning_rate": 1.792908583951682e-05, "loss": 23.2221, "step": 18890 }, { "epoch": 0.23, "learning_rate": 1.7926658113542687e-05, "loss": 24.1331, "step": 18900 }, { "epoch": 0.23, "learning_rate": 1.792422912996383e-05, "loss": 23.8308, "step": 18910 }, { "epoch": 0.23, "learning_rate": 1.7921798889165623e-05, "loss": 23.8752, "step": 18920 }, { "epoch": 0.23, "learning_rate": 1.791936739153363e-05, "loss": 23.9494, "step": 18930 }, { "epoch": 0.23, "learning_rate": 1.791693463745363e-05, "loss": 24.16, "step": 18940 }, { "epoch": 0.23, "learning_rate": 1.7914500627311584e-05, "loss": 23.091, "step": 18950 }, { "epoch": 0.23, "learning_rate": 1.7912065361493664e-05, "loss": 24.2014, "step": 18960 }, { "epoch": 0.23, "learning_rate": 1.7909628840386238e-05, "loss": 23.7506, "step": 18970 }, { "epoch": 0.23, "learning_rate": 1.7907191064375872e-05, "loss": 24.1078, "step": 18980 }, { "epoch": 0.23, "learning_rate": 1.7904752033849324e-05, "loss": 23.9967, "step": 18990 }, { "epoch": 0.23, "learning_rate": 1.790231174919357e-05, "loss": 23.6798, "step": 19000 }, { "epoch": 0.23, "learning_rate": 1.789987021079577e-05, "loss": 23.9009, "step": 19010 }, { "epoch": 0.23, "learning_rate": 1.7897427419043282e-05, "loss": 23.8709, "step": 19020 }, { "epoch": 0.23, "learning_rate": 1.7894983374323674e-05, "loss": 23.8961, "step": 19030 }, { "epoch": 0.23, "learning_rate": 1.7892538077024703e-05, "loss": 23.8883, "step": 19040 }, { "epoch": 0.23, "learning_rate": 1.7890091527534324e-05, "loss": 24.1997, "step": 19050 }, { "epoch": 0.23, "learning_rate": 1.78876437262407e-05, "loss": 24.1671, "step": 19060 }, { "epoch": 0.23, "learning_rate": 1.7885194673532183e-05, "loss": 23.6051, "step": 19070 }, { "epoch": 0.23, "learning_rate": 1.7882744369797337e-05, "loss": 23.8763, "step": 19080 }, { "epoch": 0.23, "learning_rate": 1.7880292815424904e-05, "loss": 24.0601, "step": 19090 }, { "epoch": 0.23, "learning_rate": 1.7877840010803837e-05, "loss": 23.2894, "step": 19100 }, { "epoch": 0.24, "learning_rate": 1.787538595632329e-05, "loss": 23.8366, "step": 19110 }, { "epoch": 0.24, "learning_rate": 1.7872930652372614e-05, "loss": 24.2825, "step": 19120 }, { "epoch": 0.24, "learning_rate": 1.7870474099341352e-05, "loss": 23.8765, "step": 19130 }, { "epoch": 0.24, "learning_rate": 1.7868016297619245e-05, "loss": 23.8753, "step": 19140 }, { "epoch": 0.24, "learning_rate": 1.7865557247596238e-05, "loss": 24.2737, "step": 19150 }, { "epoch": 0.24, "learning_rate": 1.7863096949662475e-05, "loss": 23.6927, "step": 19160 }, { "epoch": 0.24, "learning_rate": 1.7860635404208292e-05, "loss": 24.314, "step": 19170 }, { "epoch": 0.24, "learning_rate": 1.7858172611624227e-05, "loss": 23.9489, "step": 19180 }, { "epoch": 0.24, "learning_rate": 1.7855708572301017e-05, "loss": 24.4008, "step": 19190 }, { "epoch": 0.24, "learning_rate": 1.7853243286629584e-05, "loss": 23.943, "step": 19200 }, { "epoch": 0.24, "learning_rate": 1.7850776755001072e-05, "loss": 23.8699, "step": 19210 }, { "epoch": 0.24, "learning_rate": 1.78483089778068e-05, "loss": 23.729, "step": 19220 }, { "epoch": 0.24, "learning_rate": 1.7845839955438292e-05, "loss": 23.7353, "step": 19230 }, { "epoch": 0.24, "learning_rate": 1.784336968828728e-05, "loss": 23.8969, "step": 19240 }, { "epoch": 0.24, "learning_rate": 1.784089817674568e-05, "loss": 23.9916, "step": 19250 }, { "epoch": 0.24, "learning_rate": 1.7838425421205605e-05, "loss": 24.1678, "step": 19260 }, { "epoch": 0.24, "learning_rate": 1.783595142205938e-05, "loss": 23.5409, "step": 19270 }, { "epoch": 0.24, "learning_rate": 1.7833476179699506e-05, "loss": 23.2806, "step": 19280 }, { "epoch": 0.24, "learning_rate": 1.7830999694518702e-05, "loss": 23.6718, "step": 19290 }, { "epoch": 0.24, "learning_rate": 1.782852196690987e-05, "loss": 23.4533, "step": 19300 }, { "epoch": 0.24, "learning_rate": 1.782604299726612e-05, "loss": 23.8773, "step": 19310 }, { "epoch": 0.24, "learning_rate": 1.7823562785980746e-05, "loss": 23.9108, "step": 19320 }, { "epoch": 0.24, "learning_rate": 1.7821081333447248e-05, "loss": 23.8051, "step": 19330 }, { "epoch": 0.24, "learning_rate": 1.7818598640059322e-05, "loss": 23.242, "step": 19340 }, { "epoch": 0.24, "learning_rate": 1.7816114706210862e-05, "loss": 23.2899, "step": 19350 }, { "epoch": 0.24, "learning_rate": 1.7813629532295956e-05, "loss": 24.1856, "step": 19360 }, { "epoch": 0.24, "learning_rate": 1.781114311870889e-05, "loss": 23.9065, "step": 19370 }, { "epoch": 0.24, "learning_rate": 1.780865546584414e-05, "loss": 23.6397, "step": 19380 }, { "epoch": 0.24, "learning_rate": 1.7806166574096397e-05, "loss": 23.3145, "step": 19390 }, { "epoch": 0.24, "learning_rate": 1.7803676443860522e-05, "loss": 24.1525, "step": 19400 }, { "epoch": 0.24, "learning_rate": 1.7801185075531595e-05, "loss": 24.7002, "step": 19410 }, { "epoch": 0.24, "learning_rate": 1.7798692469504885e-05, "loss": 23.0357, "step": 19420 }, { "epoch": 0.24, "learning_rate": 1.7796198626175854e-05, "loss": 24.0993, "step": 19430 }, { "epoch": 0.24, "learning_rate": 1.7793703545940165e-05, "loss": 23.7393, "step": 19440 }, { "epoch": 0.24, "learning_rate": 1.7791207229193674e-05, "loss": 23.6761, "step": 19450 }, { "epoch": 0.24, "learning_rate": 1.7788709676332433e-05, "loss": 23.7991, "step": 19460 }, { "epoch": 0.24, "learning_rate": 1.7786210887752694e-05, "loss": 23.7072, "step": 19470 }, { "epoch": 0.24, "learning_rate": 1.77837108638509e-05, "loss": 24.4742, "step": 19480 }, { "epoch": 0.24, "learning_rate": 1.7781209605023694e-05, "loss": 23.6793, "step": 19490 }, { "epoch": 0.24, "learning_rate": 1.7778707111667913e-05, "loss": 24.9482, "step": 19500 }, { "epoch": 0.24, "learning_rate": 1.777620338418059e-05, "loss": 24.1543, "step": 19510 }, { "epoch": 0.24, "learning_rate": 1.777369842295896e-05, "loss": 23.7215, "step": 19520 }, { "epoch": 0.24, "learning_rate": 1.7771192228400437e-05, "loss": 24.1692, "step": 19530 }, { "epoch": 0.24, "learning_rate": 1.776868480090265e-05, "loss": 24.4046, "step": 19540 }, { "epoch": 0.24, "learning_rate": 1.7766176140863413e-05, "loss": 23.5353, "step": 19550 }, { "epoch": 0.24, "learning_rate": 1.7763666248680736e-05, "loss": 24.0048, "step": 19560 }, { "epoch": 0.24, "learning_rate": 1.7761155124752827e-05, "loss": 23.8878, "step": 19570 }, { "epoch": 0.24, "learning_rate": 1.7758642769478087e-05, "loss": 24.5747, "step": 19580 }, { "epoch": 0.24, "learning_rate": 1.775612918325512e-05, "loss": 23.9416, "step": 19590 }, { "epoch": 0.24, "learning_rate": 1.7753614366482713e-05, "loss": 23.5013, "step": 19600 }, { "epoch": 0.24, "learning_rate": 1.7751098319559857e-05, "loss": 23.7243, "step": 19610 }, { "epoch": 0.24, "learning_rate": 1.7748581042885735e-05, "loss": 23.9033, "step": 19620 }, { "epoch": 0.24, "learning_rate": 1.7746062536859726e-05, "loss": 23.7137, "step": 19630 }, { "epoch": 0.24, "learning_rate": 1.7743542801881407e-05, "loss": 24.4859, "step": 19640 }, { "epoch": 0.24, "learning_rate": 1.7741021838350537e-05, "loss": 23.846, "step": 19650 }, { "epoch": 0.24, "learning_rate": 1.773849964666709e-05, "loss": 23.9226, "step": 19660 }, { "epoch": 0.24, "learning_rate": 1.7735976227231223e-05, "loss": 23.944, "step": 19670 }, { "epoch": 0.24, "learning_rate": 1.7733451580443288e-05, "loss": 23.8812, "step": 19680 }, { "epoch": 0.24, "learning_rate": 1.7730925706703828e-05, "loss": 23.76, "step": 19690 }, { "epoch": 0.24, "learning_rate": 1.7728398606413593e-05, "loss": 24.0276, "step": 19700 }, { "epoch": 0.24, "learning_rate": 1.7725870279973517e-05, "loss": 24.0128, "step": 19710 }, { "epoch": 0.24, "learning_rate": 1.772334072778473e-05, "loss": 24.4359, "step": 19720 }, { "epoch": 0.24, "learning_rate": 1.7720809950248567e-05, "loss": 23.7062, "step": 19730 }, { "epoch": 0.24, "learning_rate": 1.7718277947766537e-05, "loss": 24.1479, "step": 19740 }, { "epoch": 0.24, "learning_rate": 1.7715744720740363e-05, "loss": 24.1383, "step": 19750 }, { "epoch": 0.24, "learning_rate": 1.771321026957195e-05, "loss": 23.9552, "step": 19760 }, { "epoch": 0.24, "learning_rate": 1.7710674594663406e-05, "loss": 23.7156, "step": 19770 }, { "epoch": 0.24, "learning_rate": 1.7708137696417027e-05, "loss": 24.208, "step": 19780 }, { "epoch": 0.24, "learning_rate": 1.7705599575235303e-05, "loss": 23.5092, "step": 19790 }, { "epoch": 0.24, "learning_rate": 1.7703060231520922e-05, "loss": 24.1797, "step": 19800 }, { "epoch": 0.24, "learning_rate": 1.770051966567676e-05, "loss": 24.0138, "step": 19810 }, { "epoch": 0.24, "learning_rate": 1.7697977878105893e-05, "loss": 23.4246, "step": 19820 }, { "epoch": 0.24, "learning_rate": 1.7695434869211593e-05, "loss": 23.0667, "step": 19830 }, { "epoch": 0.24, "learning_rate": 1.769289063939732e-05, "loss": 24.684, "step": 19840 }, { "epoch": 0.24, "learning_rate": 1.769034518906672e-05, "loss": 24.3443, "step": 19850 }, { "epoch": 0.24, "learning_rate": 1.7687798518623652e-05, "loss": 23.506, "step": 19860 }, { "epoch": 0.24, "learning_rate": 1.7685250628472153e-05, "loss": 23.9263, "step": 19870 }, { "epoch": 0.24, "learning_rate": 1.768270151901646e-05, "loss": 23.6718, "step": 19880 }, { "epoch": 0.24, "learning_rate": 1.7680151190661004e-05, "loss": 23.9397, "step": 19890 }, { "epoch": 0.24, "learning_rate": 1.7677599643810403e-05, "loss": 23.9497, "step": 19900 }, { "epoch": 0.24, "learning_rate": 1.767504687886948e-05, "loss": 24.293, "step": 19910 }, { "epoch": 0.24, "learning_rate": 1.767249289624324e-05, "loss": 24.0676, "step": 19920 }, { "epoch": 0.25, "learning_rate": 1.766993769633688e-05, "loss": 23.8003, "step": 19930 }, { "epoch": 0.25, "learning_rate": 1.7667381279555805e-05, "loss": 23.2592, "step": 19940 }, { "epoch": 0.25, "learning_rate": 1.76648236463056e-05, "loss": 23.8168, "step": 19950 }, { "epoch": 0.25, "learning_rate": 1.7662264796992047e-05, "loss": 24.5823, "step": 19960 }, { "epoch": 0.25, "learning_rate": 1.7659704732021117e-05, "loss": 23.7224, "step": 19970 }, { "epoch": 0.25, "learning_rate": 1.765714345179898e-05, "loss": 24.5021, "step": 19980 }, { "epoch": 0.25, "learning_rate": 1.7654580956732e-05, "loss": 23.7399, "step": 19990 }, { "epoch": 0.25, "learning_rate": 1.765201724722672e-05, "loss": 24.4039, "step": 20000 }, { "epoch": 0.25, "learning_rate": 1.7649452323689897e-05, "loss": 23.5536, "step": 20010 }, { "epoch": 0.25, "learning_rate": 1.764688618652846e-05, "loss": 23.9562, "step": 20020 }, { "epoch": 0.25, "learning_rate": 1.7644318836149542e-05, "loss": 24.6637, "step": 20030 }, { "epoch": 0.25, "learning_rate": 1.764175027296047e-05, "loss": 23.9704, "step": 20040 }, { "epoch": 0.25, "learning_rate": 1.763918049736876e-05, "loss": 23.7303, "step": 20050 }, { "epoch": 0.25, "learning_rate": 1.7636609509782115e-05, "loss": 24.1104, "step": 20060 }, { "epoch": 0.25, "learning_rate": 1.7634037310608437e-05, "loss": 23.7327, "step": 20070 }, { "epoch": 0.25, "learning_rate": 1.7631463900255823e-05, "loss": 23.9749, "step": 20080 }, { "epoch": 0.25, "learning_rate": 1.7628889279132552e-05, "loss": 23.6198, "step": 20090 }, { "epoch": 0.25, "learning_rate": 1.76263134476471e-05, "loss": 24.2133, "step": 20100 }, { "epoch": 0.25, "learning_rate": 1.7623736406208142e-05, "loss": 23.9926, "step": 20110 }, { "epoch": 0.25, "learning_rate": 1.7621158155224532e-05, "loss": 23.8425, "step": 20120 }, { "epoch": 0.25, "learning_rate": 1.761857869510533e-05, "loss": 24.0824, "step": 20130 }, { "epoch": 0.25, "learning_rate": 1.7615998026259773e-05, "loss": 23.946, "step": 20140 }, { "epoch": 0.25, "learning_rate": 1.76134161490973e-05, "loss": 23.9785, "step": 20150 }, { "epoch": 0.25, "learning_rate": 1.761083306402754e-05, "loss": 24.0904, "step": 20160 }, { "epoch": 0.25, "learning_rate": 1.7608248771460315e-05, "loss": 23.7204, "step": 20170 }, { "epoch": 0.25, "learning_rate": 1.7605663271805634e-05, "loss": 23.6448, "step": 20180 }, { "epoch": 0.25, "learning_rate": 1.7603076565473695e-05, "loss": 24.0979, "step": 20190 }, { "epoch": 0.25, "learning_rate": 1.7600488652874902e-05, "loss": 24.0747, "step": 20200 }, { "epoch": 0.25, "learning_rate": 1.759789953441983e-05, "loss": 23.8468, "step": 20210 }, { "epoch": 0.25, "learning_rate": 1.759530921051926e-05, "loss": 23.9422, "step": 20220 }, { "epoch": 0.25, "learning_rate": 1.759271768158416e-05, "loss": 23.6583, "step": 20230 }, { "epoch": 0.25, "learning_rate": 1.7590124948025694e-05, "loss": 23.1515, "step": 20240 }, { "epoch": 0.25, "learning_rate": 1.7587531010255206e-05, "loss": 23.719, "step": 20250 }, { "epoch": 0.25, "learning_rate": 1.7584935868684242e-05, "loss": 23.4113, "step": 20260 }, { "epoch": 0.25, "learning_rate": 1.7582339523724533e-05, "loss": 23.7529, "step": 20270 }, { "epoch": 0.25, "learning_rate": 1.7579741975788e-05, "loss": 24.1172, "step": 20280 }, { "epoch": 0.25, "learning_rate": 1.7577143225286758e-05, "loss": 24.3849, "step": 20290 }, { "epoch": 0.25, "learning_rate": 1.7574543272633113e-05, "loss": 23.931, "step": 20300 }, { "epoch": 0.25, "learning_rate": 1.757194211823956e-05, "loss": 24.3759, "step": 20310 }, { "epoch": 0.25, "learning_rate": 1.756933976251879e-05, "loss": 24.2114, "step": 20320 }, { "epoch": 0.25, "learning_rate": 1.756673620588367e-05, "loss": 23.2024, "step": 20330 }, { "epoch": 0.25, "learning_rate": 1.756413144874728e-05, "loss": 23.949, "step": 20340 }, { "epoch": 0.25, "learning_rate": 1.7561525491522865e-05, "loss": 23.8169, "step": 20350 }, { "epoch": 0.25, "learning_rate": 1.7558918334623888e-05, "loss": 23.764, "step": 20360 }, { "epoch": 0.25, "learning_rate": 1.755630997846397e-05, "loss": 23.666, "step": 20370 }, { "epoch": 0.25, "learning_rate": 1.755370042345696e-05, "loss": 23.7273, "step": 20380 }, { "epoch": 0.25, "learning_rate": 1.755108967001686e-05, "loss": 23.3621, "step": 20390 }, { "epoch": 0.25, "learning_rate": 1.7548477718557887e-05, "loss": 23.9853, "step": 20400 }, { "epoch": 0.25, "learning_rate": 1.7545864569494442e-05, "loss": 24.1975, "step": 20410 }, { "epoch": 0.25, "learning_rate": 1.7543250223241112e-05, "loss": 23.609, "step": 20420 }, { "epoch": 0.25, "learning_rate": 1.7540634680212677e-05, "loss": 24.0299, "step": 20430 }, { "epoch": 0.25, "learning_rate": 1.7538017940824103e-05, "loss": 23.8854, "step": 20440 }, { "epoch": 0.25, "learning_rate": 1.7535400005490554e-05, "loss": 23.8678, "step": 20450 }, { "epoch": 0.25, "learning_rate": 1.7532780874627374e-05, "loss": 23.9058, "step": 20460 }, { "epoch": 0.25, "learning_rate": 1.7530160548650105e-05, "loss": 23.725, "step": 20470 }, { "epoch": 0.25, "learning_rate": 1.7527539027974477e-05, "loss": 24.1385, "step": 20480 }, { "epoch": 0.25, "learning_rate": 1.75249163130164e-05, "loss": 23.3558, "step": 20490 }, { "epoch": 0.25, "learning_rate": 1.752229240419199e-05, "loss": 23.9956, "step": 20500 }, { "epoch": 0.25, "learning_rate": 1.751966730191754e-05, "loss": 23.8766, "step": 20510 }, { "epoch": 0.25, "learning_rate": 1.751704100660953e-05, "loss": 23.7784, "step": 20520 }, { "epoch": 0.25, "learning_rate": 1.7514413518684642e-05, "loss": 23.629, "step": 20530 }, { "epoch": 0.25, "learning_rate": 1.751178483855974e-05, "loss": 23.84, "step": 20540 }, { "epoch": 0.25, "learning_rate": 1.7509154966651878e-05, "loss": 23.9682, "step": 20550 }, { "epoch": 0.25, "learning_rate": 1.7506523903378293e-05, "loss": 23.6836, "step": 20560 }, { "epoch": 0.25, "learning_rate": 1.7503891649156424e-05, "loss": 23.8296, "step": 20570 }, { "epoch": 0.25, "learning_rate": 1.7501258204403883e-05, "loss": 24.3339, "step": 20580 }, { "epoch": 0.25, "learning_rate": 1.7498623569538488e-05, "loss": 24.4088, "step": 20590 }, { "epoch": 0.25, "learning_rate": 1.7495987744978234e-05, "loss": 24.0378, "step": 20600 }, { "epoch": 0.25, "learning_rate": 1.7493350731141307e-05, "loss": 23.9294, "step": 20610 }, { "epoch": 0.25, "learning_rate": 1.7490712528446083e-05, "loss": 23.676, "step": 20620 }, { "epoch": 0.25, "learning_rate": 1.7488073137311124e-05, "loss": 23.5049, "step": 20630 }, { "epoch": 0.25, "learning_rate": 1.7485432558155188e-05, "loss": 23.8445, "step": 20640 }, { "epoch": 0.25, "learning_rate": 1.7482790791397215e-05, "loss": 23.5925, "step": 20650 }, { "epoch": 0.25, "learning_rate": 1.7480147837456333e-05, "loss": 23.6386, "step": 20660 }, { "epoch": 0.25, "learning_rate": 1.7477503696751862e-05, "loss": 24.2535, "step": 20670 }, { "epoch": 0.25, "learning_rate": 1.74748583697033e-05, "loss": 23.8571, "step": 20680 }, { "epoch": 0.25, "learning_rate": 1.7472211856730355e-05, "loss": 23.5149, "step": 20690 }, { "epoch": 0.25, "learning_rate": 1.7469564158252905e-05, "loss": 23.4195, "step": 20700 }, { "epoch": 0.25, "learning_rate": 1.7466915274691018e-05, "loss": 24.5604, "step": 20710 }, { "epoch": 0.25, "learning_rate": 1.746426520646495e-05, "loss": 23.8012, "step": 20720 }, { "epoch": 0.25, "learning_rate": 1.7461613953995158e-05, "loss": 24.2178, "step": 20730 }, { "epoch": 0.26, "learning_rate": 1.7458961517702265e-05, "loss": 23.5505, "step": 20740 }, { "epoch": 0.26, "learning_rate": 1.7456307898007102e-05, "loss": 23.9609, "step": 20750 }, { "epoch": 0.26, "learning_rate": 1.745365309533068e-05, "loss": 24.5078, "step": 20760 }, { "epoch": 0.26, "learning_rate": 1.745099711009419e-05, "loss": 23.7003, "step": 20770 }, { "epoch": 0.26, "learning_rate": 1.744833994271902e-05, "loss": 23.8098, "step": 20780 }, { "epoch": 0.26, "learning_rate": 1.744568159362675e-05, "loss": 24.3805, "step": 20790 }, { "epoch": 0.26, "learning_rate": 1.7443022063239127e-05, "loss": 23.8389, "step": 20800 }, { "epoch": 0.26, "learning_rate": 1.7440361351978114e-05, "loss": 23.6086, "step": 20810 }, { "epoch": 0.26, "learning_rate": 1.7437699460265836e-05, "loss": 24.069, "step": 20820 }, { "epoch": 0.26, "learning_rate": 1.743503638852462e-05, "loss": 23.8269, "step": 20830 }, { "epoch": 0.26, "learning_rate": 1.7432372137176975e-05, "loss": 24.0535, "step": 20840 }, { "epoch": 0.26, "learning_rate": 1.7429706706645598e-05, "loss": 23.9991, "step": 20850 }, { "epoch": 0.26, "learning_rate": 1.7427040097353373e-05, "loss": 23.2564, "step": 20860 }, { "epoch": 0.26, "learning_rate": 1.7424372309723373e-05, "loss": 24.5375, "step": 20870 }, { "epoch": 0.26, "learning_rate": 1.7421703344178853e-05, "loss": 23.9904, "step": 20880 }, { "epoch": 0.26, "learning_rate": 1.7419033201143258e-05, "loss": 24.7628, "step": 20890 }, { "epoch": 0.26, "learning_rate": 1.7416361881040224e-05, "loss": 24.1015, "step": 20900 }, { "epoch": 0.26, "learning_rate": 1.7413689384293565e-05, "loss": 24.1064, "step": 20910 }, { "epoch": 0.26, "learning_rate": 1.741101571132729e-05, "loss": 24.0048, "step": 20920 }, { "epoch": 0.26, "learning_rate": 1.740834086256559e-05, "loss": 24.0164, "step": 20930 }, { "epoch": 0.26, "learning_rate": 1.740566483843284e-05, "loss": 23.5512, "step": 20940 }, { "epoch": 0.26, "learning_rate": 1.7402987639353605e-05, "loss": 23.8252, "step": 20950 }, { "epoch": 0.26, "learning_rate": 1.740030926575264e-05, "loss": 22.8828, "step": 20960 }, { "epoch": 0.26, "learning_rate": 1.7397629718054885e-05, "loss": 23.7828, "step": 20970 }, { "epoch": 0.26, "learning_rate": 1.7394948996685458e-05, "loss": 23.7638, "step": 20980 }, { "epoch": 0.26, "learning_rate": 1.739226710206967e-05, "loss": 23.8572, "step": 20990 }, { "epoch": 0.26, "learning_rate": 1.738958403463302e-05, "loss": 23.8809, "step": 21000 }, { "epoch": 0.26, "learning_rate": 1.7386899794801188e-05, "loss": 23.7866, "step": 21010 }, { "epoch": 0.26, "learning_rate": 1.7384214383000042e-05, "loss": 23.7029, "step": 21020 }, { "epoch": 0.26, "learning_rate": 1.7381527799655637e-05, "loss": 23.501, "step": 21030 }, { "epoch": 0.26, "learning_rate": 1.7378840045194215e-05, "loss": 24.081, "step": 21040 }, { "epoch": 0.26, "learning_rate": 1.73761511200422e-05, "loss": 24.1896, "step": 21050 }, { "epoch": 0.26, "learning_rate": 1.7373461024626205e-05, "loss": 24.6137, "step": 21060 }, { "epoch": 0.26, "learning_rate": 1.7370769759373025e-05, "loss": 23.6948, "step": 21070 }, { "epoch": 0.26, "learning_rate": 1.7368077324709644e-05, "loss": 24.2268, "step": 21080 }, { "epoch": 0.26, "learning_rate": 1.7365383721063232e-05, "loss": 23.7453, "step": 21090 }, { "epoch": 0.26, "learning_rate": 1.736268894886114e-05, "loss": 24.1903, "step": 21100 }, { "epoch": 0.26, "learning_rate": 1.7359993008530917e-05, "loss": 23.6596, "step": 21110 }, { "epoch": 0.26, "learning_rate": 1.7357295900500273e-05, "loss": 24.1641, "step": 21120 }, { "epoch": 0.26, "learning_rate": 1.7354597625197126e-05, "loss": 23.8275, "step": 21130 }, { "epoch": 0.26, "learning_rate": 1.735189818304957e-05, "loss": 23.7874, "step": 21140 }, { "epoch": 0.26, "learning_rate": 1.734919757448588e-05, "loss": 24.3726, "step": 21150 }, { "epoch": 0.26, "learning_rate": 1.7346495799934534e-05, "loss": 23.7996, "step": 21160 }, { "epoch": 0.26, "learning_rate": 1.734379285982417e-05, "loss": 24.0488, "step": 21170 }, { "epoch": 0.26, "learning_rate": 1.7341088754583634e-05, "loss": 23.0665, "step": 21180 }, { "epoch": 0.26, "learning_rate": 1.7338383484641935e-05, "loss": 23.9671, "step": 21190 }, { "epoch": 0.26, "learning_rate": 1.7335677050428287e-05, "loss": 23.6235, "step": 21200 }, { "epoch": 0.26, "learning_rate": 1.733296945237207e-05, "loss": 24.3267, "step": 21210 }, { "epoch": 0.26, "learning_rate": 1.733026069090287e-05, "loss": 24.1926, "step": 21220 }, { "epoch": 0.26, "learning_rate": 1.7327550766450437e-05, "loss": 23.8633, "step": 21230 }, { "epoch": 0.26, "learning_rate": 1.7324839679444714e-05, "loss": 23.671, "step": 21240 }, { "epoch": 0.26, "learning_rate": 1.7322127430315836e-05, "loss": 23.8539, "step": 21250 }, { "epoch": 0.26, "learning_rate": 1.7319414019494108e-05, "loss": 24.5298, "step": 21260 }, { "epoch": 0.26, "learning_rate": 1.7316699447410033e-05, "loss": 24.2662, "step": 21270 }, { "epoch": 0.26, "learning_rate": 1.7313983714494288e-05, "loss": 24.0259, "step": 21280 }, { "epoch": 0.26, "learning_rate": 1.7311266821177736e-05, "loss": 23.4087, "step": 21290 }, { "epoch": 0.26, "learning_rate": 1.7308548767891427e-05, "loss": 24.1786, "step": 21300 }, { "epoch": 0.26, "learning_rate": 1.7305829555066595e-05, "loss": 23.8279, "step": 21310 }, { "epoch": 0.26, "learning_rate": 1.7303109183134657e-05, "loss": 23.6196, "step": 21320 }, { "epoch": 0.26, "learning_rate": 1.7300387652527215e-05, "loss": 24.4887, "step": 21330 }, { "epoch": 0.26, "learning_rate": 1.7297664963676052e-05, "loss": 23.6509, "step": 21340 }, { "epoch": 0.26, "learning_rate": 1.7294941117013138e-05, "loss": 23.8178, "step": 21350 }, { "epoch": 0.26, "learning_rate": 1.7292216112970625e-05, "loss": 23.4532, "step": 21360 }, { "epoch": 0.26, "learning_rate": 1.7289489951980845e-05, "loss": 23.6208, "step": 21370 }, { "epoch": 0.26, "learning_rate": 1.728676263447632e-05, "loss": 23.8794, "step": 21380 }, { "epoch": 0.26, "learning_rate": 1.7284034160889757e-05, "loss": 23.0627, "step": 21390 }, { "epoch": 0.26, "learning_rate": 1.7281304531654034e-05, "loss": 23.804, "step": 21400 }, { "epoch": 0.26, "learning_rate": 1.7278573747202227e-05, "loss": 24.1968, "step": 21410 }, { "epoch": 0.26, "learning_rate": 1.7275841807967585e-05, "loss": 24.2986, "step": 21420 }, { "epoch": 0.26, "learning_rate": 1.727310871438355e-05, "loss": 23.8552, "step": 21430 }, { "epoch": 0.26, "learning_rate": 1.7270374466883733e-05, "loss": 24.2517, "step": 21440 }, { "epoch": 0.26, "learning_rate": 1.7267639065901942e-05, "loss": 24.434, "step": 21450 }, { "epoch": 0.26, "learning_rate": 1.726490251187216e-05, "loss": 23.8986, "step": 21460 }, { "epoch": 0.26, "learning_rate": 1.7262164805228558e-05, "loss": 23.5965, "step": 21470 }, { "epoch": 0.26, "learning_rate": 1.7259425946405485e-05, "loss": 24.0335, "step": 21480 }, { "epoch": 0.26, "learning_rate": 1.7256685935837473e-05, "loss": 23.5977, "step": 21490 }, { "epoch": 0.26, "learning_rate": 1.725394477395924e-05, "loss": 23.3851, "step": 21500 }, { "epoch": 0.26, "learning_rate": 1.725120246120569e-05, "loss": 24.009, "step": 21510 }, { "epoch": 0.26, "learning_rate": 1.7248458998011896e-05, "loss": 23.9537, "step": 21520 }, { "epoch": 0.26, "learning_rate": 1.724571438481313e-05, "loss": 23.6904, "step": 21530 }, { "epoch": 0.26, "learning_rate": 1.7242968622044835e-05, "loss": 24.12, "step": 21540 }, { "epoch": 0.27, "learning_rate": 1.7240221710142644e-05, "loss": 24.0498, "step": 21550 }, { "epoch": 0.27, "learning_rate": 1.7237473649542362e-05, "loss": 24.2778, "step": 21560 }, { "epoch": 0.27, "learning_rate": 1.7234724440679992e-05, "loss": 24.2091, "step": 21570 }, { "epoch": 0.27, "learning_rate": 1.72319740839917e-05, "loss": 23.6086, "step": 21580 }, { "epoch": 0.27, "learning_rate": 1.722922257991385e-05, "loss": 23.947, "step": 21590 }, { "epoch": 0.27, "learning_rate": 1.722646992888298e-05, "loss": 24.2105, "step": 21600 }, { "epoch": 0.27, "learning_rate": 1.7223716131335816e-05, "loss": 23.9067, "step": 21610 }, { "epoch": 0.27, "learning_rate": 1.722096118770926e-05, "loss": 24.0804, "step": 21620 }, { "epoch": 0.27, "learning_rate": 1.7218205098440398e-05, "loss": 24.1458, "step": 21630 }, { "epoch": 0.27, "learning_rate": 1.7215447863966494e-05, "loss": 23.8023, "step": 21640 }, { "epoch": 0.27, "learning_rate": 1.7212689484725e-05, "loss": 23.8184, "step": 21650 }, { "epoch": 0.27, "learning_rate": 1.7209929961153546e-05, "loss": 23.8626, "step": 21660 }, { "epoch": 0.27, "learning_rate": 1.7207169293689947e-05, "loss": 23.8616, "step": 21670 }, { "epoch": 0.27, "learning_rate": 1.7204407482772195e-05, "loss": 24.0837, "step": 21680 }, { "epoch": 0.27, "learning_rate": 1.7201644528838466e-05, "loss": 24.2201, "step": 21690 }, { "epoch": 0.27, "learning_rate": 1.7198880432327115e-05, "loss": 23.6549, "step": 21700 }, { "epoch": 0.27, "learning_rate": 1.7196115193676686e-05, "loss": 24.4173, "step": 21710 }, { "epoch": 0.27, "learning_rate": 1.7193348813325886e-05, "loss": 23.3854, "step": 21720 }, { "epoch": 0.27, "learning_rate": 1.719058129171363e-05, "loss": 23.7961, "step": 21730 }, { "epoch": 0.27, "learning_rate": 1.718781262927899e-05, "loss": 23.6572, "step": 21740 }, { "epoch": 0.27, "learning_rate": 1.718504282646123e-05, "loss": 23.9115, "step": 21750 }, { "epoch": 0.27, "learning_rate": 1.7182271883699798e-05, "loss": 24.5567, "step": 21760 }, { "epoch": 0.27, "learning_rate": 1.717949980143431e-05, "loss": 23.6287, "step": 21770 }, { "epoch": 0.27, "learning_rate": 1.7176726580104577e-05, "loss": 24.2431, "step": 21780 }, { "epoch": 0.27, "learning_rate": 1.7173952220150582e-05, "loss": 23.7547, "step": 21790 }, { "epoch": 0.27, "learning_rate": 1.7171176722012494e-05, "loss": 24.3978, "step": 21800 }, { "epoch": 0.27, "learning_rate": 1.716840008613066e-05, "loss": 24.4114, "step": 21810 }, { "epoch": 0.27, "learning_rate": 1.7165622312945603e-05, "loss": 24.0919, "step": 21820 }, { "epoch": 0.27, "learning_rate": 1.7162843402898034e-05, "loss": 23.7267, "step": 21830 }, { "epoch": 0.27, "learning_rate": 1.716006335642884e-05, "loss": 23.9516, "step": 21840 }, { "epoch": 0.27, "learning_rate": 1.7157282173979093e-05, "loss": 23.4363, "step": 21850 }, { "epoch": 0.27, "learning_rate": 1.715449985599004e-05, "loss": 23.4491, "step": 21860 }, { "epoch": 0.27, "learning_rate": 1.7151716402903104e-05, "loss": 23.3798, "step": 21870 }, { "epoch": 0.27, "learning_rate": 1.7148931815159903e-05, "loss": 24.1385, "step": 21880 }, { "epoch": 0.27, "learning_rate": 1.7146146093202224e-05, "loss": 23.5982, "step": 21890 }, { "epoch": 0.27, "learning_rate": 1.714335923747203e-05, "loss": 24.4981, "step": 21900 }, { "epoch": 0.27, "learning_rate": 1.7140571248411477e-05, "loss": 22.905, "step": 21910 }, { "epoch": 0.27, "learning_rate": 1.713778212646289e-05, "loss": 24.0858, "step": 21920 }, { "epoch": 0.27, "learning_rate": 1.7134991872068778e-05, "loss": 23.8177, "step": 21930 }, { "epoch": 0.27, "learning_rate": 1.713220048567183e-05, "loss": 24.1577, "step": 21940 }, { "epoch": 0.27, "learning_rate": 1.7129407967714913e-05, "loss": 24.4745, "step": 21950 }, { "epoch": 0.27, "learning_rate": 1.7126614318641074e-05, "loss": 24.5528, "step": 21960 }, { "epoch": 0.27, "learning_rate": 1.712381953889354e-05, "loss": 23.9143, "step": 21970 }, { "epoch": 0.27, "learning_rate": 1.712102362891572e-05, "loss": 23.4798, "step": 21980 }, { "epoch": 0.27, "learning_rate": 1.711822658915119e-05, "loss": 23.9507, "step": 21990 }, { "epoch": 0.27, "learning_rate": 1.7115428420043722e-05, "loss": 24.0069, "step": 22000 }, { "epoch": 0.27, "learning_rate": 1.7112629122037263e-05, "loss": 23.2262, "step": 22010 }, { "epoch": 0.27, "learning_rate": 1.710982869557593e-05, "loss": 23.8889, "step": 22020 }, { "epoch": 0.27, "learning_rate": 1.7107027141104025e-05, "loss": 23.7878, "step": 22030 }, { "epoch": 0.27, "learning_rate": 1.7104224459066033e-05, "loss": 23.6856, "step": 22040 }, { "epoch": 0.27, "learning_rate": 1.7101420649906608e-05, "loss": 24.3917, "step": 22050 }, { "epoch": 0.27, "learning_rate": 1.709861571407059e-05, "loss": 23.2926, "step": 22060 }, { "epoch": 0.27, "learning_rate": 1.7095809652003006e-05, "loss": 24.3535, "step": 22070 }, { "epoch": 0.27, "learning_rate": 1.7093002464149036e-05, "loss": 23.6334, "step": 22080 }, { "epoch": 0.27, "learning_rate": 1.7090194150954073e-05, "loss": 23.5299, "step": 22090 }, { "epoch": 0.27, "learning_rate": 1.708738471286365e-05, "loss": 23.6082, "step": 22100 }, { "epoch": 0.27, "learning_rate": 1.7084574150323514e-05, "loss": 24.3062, "step": 22110 }, { "epoch": 0.27, "learning_rate": 1.7081762463779567e-05, "loss": 23.8731, "step": 22120 }, { "epoch": 0.27, "learning_rate": 1.7078949653677905e-05, "loss": 23.8624, "step": 22130 }, { "epoch": 0.27, "learning_rate": 1.7076135720464785e-05, "loss": 23.5789, "step": 22140 }, { "epoch": 0.27, "learning_rate": 1.7073320664586658e-05, "loss": 24.7013, "step": 22150 }, { "epoch": 0.27, "learning_rate": 1.7070504486490147e-05, "loss": 23.9408, "step": 22160 }, { "epoch": 0.27, "learning_rate": 1.7067687186622053e-05, "loss": 24.0325, "step": 22170 }, { "epoch": 0.27, "learning_rate": 1.706486876542935e-05, "loss": 23.8708, "step": 22180 }, { "epoch": 0.27, "learning_rate": 1.7062049223359204e-05, "loss": 24.198, "step": 22190 }, { "epoch": 0.27, "learning_rate": 1.705922856085894e-05, "loss": 23.7259, "step": 22200 }, { "epoch": 0.27, "learning_rate": 1.7056406778376076e-05, "loss": 23.8542, "step": 22210 }, { "epoch": 0.27, "learning_rate": 1.70535838763583e-05, "loss": 23.7234, "step": 22220 }, { "epoch": 0.27, "learning_rate": 1.7050759855253484e-05, "loss": 23.9183, "step": 22230 }, { "epoch": 0.27, "learning_rate": 1.704793471550967e-05, "loss": 23.7588, "step": 22240 }, { "epoch": 0.27, "learning_rate": 1.7045108457575078e-05, "loss": 23.8927, "step": 22250 }, { "epoch": 0.27, "learning_rate": 1.704228108189811e-05, "loss": 23.8758, "step": 22260 }, { "epoch": 0.27, "learning_rate": 1.7039452588927348e-05, "loss": 24.1231, "step": 22270 }, { "epoch": 0.27, "learning_rate": 1.7036622979111546e-05, "loss": 24.0858, "step": 22280 }, { "epoch": 0.27, "learning_rate": 1.703379225289963e-05, "loss": 24.7796, "step": 22290 }, { "epoch": 0.27, "learning_rate": 1.7030960410740714e-05, "loss": 23.8493, "step": 22300 }, { "epoch": 0.27, "learning_rate": 1.702812745308408e-05, "loss": 24.2032, "step": 22310 }, { "epoch": 0.27, "learning_rate": 1.70252933803792e-05, "loss": 23.8565, "step": 22320 }, { "epoch": 0.27, "learning_rate": 1.7022458193075702e-05, "loss": 23.5961, "step": 22330 }, { "epoch": 0.27, "learning_rate": 1.7019621891623414e-05, "loss": 23.8713, "step": 22340 }, { "epoch": 0.27, "learning_rate": 1.7016784476472325e-05, "loss": 24.3699, "step": 22350 }, { "epoch": 0.27, "learning_rate": 1.7013945948072607e-05, "loss": 24.5663, "step": 22360 }, { "epoch": 0.28, "learning_rate": 1.70111063068746e-05, "loss": 23.7063, "step": 22370 }, { "epoch": 0.28, "learning_rate": 1.7008265553328832e-05, "loss": 23.1776, "step": 22380 }, { "epoch": 0.28, "learning_rate": 1.7005423687886008e-05, "loss": 23.8436, "step": 22390 }, { "epoch": 0.28, "learning_rate": 1.7002580710997e-05, "loss": 23.8257, "step": 22400 }, { "epoch": 0.28, "learning_rate": 1.699973662311286e-05, "loss": 24.1672, "step": 22410 }, { "epoch": 0.28, "learning_rate": 1.699689142468482e-05, "loss": 23.9897, "step": 22420 }, { "epoch": 0.28, "learning_rate": 1.699404511616428e-05, "loss": 23.7312, "step": 22430 }, { "epoch": 0.28, "learning_rate": 1.699119769800283e-05, "loss": 23.531, "step": 22440 }, { "epoch": 0.28, "learning_rate": 1.6988349170652214e-05, "loss": 23.6726, "step": 22450 }, { "epoch": 0.28, "learning_rate": 1.698549953456438e-05, "loss": 24.3774, "step": 22460 }, { "epoch": 0.28, "learning_rate": 1.6982648790191426e-05, "loss": 24.3672, "step": 22470 }, { "epoch": 0.28, "learning_rate": 1.6979796937985646e-05, "loss": 23.7333, "step": 22480 }, { "epoch": 0.28, "learning_rate": 1.6976943978399495e-05, "loss": 24.1334, "step": 22490 }, { "epoch": 0.28, "learning_rate": 1.6974089911885612e-05, "loss": 24.5556, "step": 22500 }, { "epoch": 0.28, "learning_rate": 1.697123473889681e-05, "loss": 23.6292, "step": 22510 }, { "epoch": 0.28, "learning_rate": 1.696837845988607e-05, "loss": 24.143, "step": 22520 }, { "epoch": 0.28, "learning_rate": 1.6965521075306568e-05, "loss": 23.9191, "step": 22530 }, { "epoch": 0.28, "learning_rate": 1.6962662585611633e-05, "loss": 24.0035, "step": 22540 }, { "epoch": 0.28, "learning_rate": 1.695980299125478e-05, "loss": 24.192, "step": 22550 }, { "epoch": 0.28, "learning_rate": 1.69569422926897e-05, "loss": 23.6966, "step": 22560 }, { "epoch": 0.28, "learning_rate": 1.6954080490370253e-05, "loss": 23.826, "step": 22570 }, { "epoch": 0.28, "learning_rate": 1.6951217584750487e-05, "loss": 23.7338, "step": 22580 }, { "epoch": 0.28, "learning_rate": 1.6948353576284612e-05, "loss": 23.4338, "step": 22590 }, { "epoch": 0.28, "learning_rate": 1.6945488465427013e-05, "loss": 24.2884, "step": 22600 }, { "epoch": 0.28, "learning_rate": 1.6942622252632262e-05, "loss": 23.8569, "step": 22610 }, { "epoch": 0.28, "learning_rate": 1.6939754938355094e-05, "loss": 24.1071, "step": 22620 }, { "epoch": 0.28, "learning_rate": 1.6936886523050426e-05, "loss": 23.941, "step": 22630 }, { "epoch": 0.28, "learning_rate": 1.6934017007173347e-05, "loss": 24.2533, "step": 22640 }, { "epoch": 0.28, "learning_rate": 1.6931146391179118e-05, "loss": 23.7288, "step": 22650 }, { "epoch": 0.28, "learning_rate": 1.6928274675523172e-05, "loss": 24.0314, "step": 22660 }, { "epoch": 0.28, "learning_rate": 1.692540186066113e-05, "loss": 23.9815, "step": 22670 }, { "epoch": 0.28, "learning_rate": 1.692252794704877e-05, "loss": 23.854, "step": 22680 }, { "epoch": 0.28, "learning_rate": 1.6919652935142065e-05, "loss": 24.1373, "step": 22690 }, { "epoch": 0.28, "learning_rate": 1.6916776825397142e-05, "loss": 23.5428, "step": 22700 }, { "epoch": 0.28, "learning_rate": 1.6913899618270312e-05, "loss": 24.096, "step": 22710 }, { "epoch": 0.28, "learning_rate": 1.6911021314218055e-05, "loss": 24.6422, "step": 22720 }, { "epoch": 0.28, "learning_rate": 1.6908141913697033e-05, "loss": 23.5979, "step": 22730 }, { "epoch": 0.28, "learning_rate": 1.690526141716408e-05, "loss": 24.0118, "step": 22740 }, { "epoch": 0.28, "learning_rate": 1.69023798250762e-05, "loss": 23.8744, "step": 22750 }, { "epoch": 0.28, "learning_rate": 1.6899497137890565e-05, "loss": 23.9076, "step": 22760 }, { "epoch": 0.28, "learning_rate": 1.6896613356064535e-05, "loss": 23.8819, "step": 22770 }, { "epoch": 0.28, "learning_rate": 1.689372848005564e-05, "loss": 23.8925, "step": 22780 }, { "epoch": 0.28, "learning_rate": 1.6890842510321568e-05, "loss": 23.7662, "step": 22790 }, { "epoch": 0.28, "learning_rate": 1.6887955447320202e-05, "loss": 24.1332, "step": 22800 }, { "epoch": 0.28, "learning_rate": 1.6885067291509592e-05, "loss": 24.0358, "step": 22810 }, { "epoch": 0.28, "learning_rate": 1.688217804334795e-05, "loss": 23.9756, "step": 22820 }, { "epoch": 0.28, "learning_rate": 1.6879287703293673e-05, "loss": 24.2098, "step": 22830 }, { "epoch": 0.28, "learning_rate": 1.6876396271805333e-05, "loss": 24.1472, "step": 22840 }, { "epoch": 0.28, "learning_rate": 1.6873503749341668e-05, "loss": 24.0594, "step": 22850 }, { "epoch": 0.28, "learning_rate": 1.6870610136361586e-05, "loss": 23.5167, "step": 22860 }, { "epoch": 0.28, "learning_rate": 1.6867715433324182e-05, "loss": 23.4331, "step": 22870 }, { "epoch": 0.28, "learning_rate": 1.6864819640688708e-05, "loss": 24.0071, "step": 22880 }, { "epoch": 0.28, "learning_rate": 1.6861922758914597e-05, "loss": 24.2216, "step": 22890 }, { "epoch": 0.28, "learning_rate": 1.685902478846146e-05, "loss": 23.9785, "step": 22900 }, { "epoch": 0.28, "learning_rate": 1.685612572978907e-05, "loss": 23.2948, "step": 22910 }, { "epoch": 0.28, "learning_rate": 1.685322558335738e-05, "loss": 24.17, "step": 22920 }, { "epoch": 0.28, "learning_rate": 1.685032434962651e-05, "loss": 24.0191, "step": 22930 }, { "epoch": 0.28, "learning_rate": 1.6847422029056755e-05, "loss": 24.5177, "step": 22940 }, { "epoch": 0.28, "learning_rate": 1.6844518622108586e-05, "loss": 23.8092, "step": 22950 }, { "epoch": 0.28, "learning_rate": 1.684161412924264e-05, "loss": 23.9687, "step": 22960 }, { "epoch": 0.28, "learning_rate": 1.683870855091974e-05, "loss": 23.4647, "step": 22970 }, { "epoch": 0.28, "learning_rate": 1.683580188760086e-05, "loss": 23.4355, "step": 22980 }, { "epoch": 0.28, "learning_rate": 1.6832894139747158e-05, "loss": 24.2659, "step": 22990 }, { "epoch": 0.28, "learning_rate": 1.6829985307819964e-05, "loss": 23.9172, "step": 23000 }, { "epoch": 0.28, "learning_rate": 1.6827075392280777e-05, "loss": 23.8691, "step": 23010 }, { "epoch": 0.28, "learning_rate": 1.682416439359128e-05, "loss": 24.3101, "step": 23020 }, { "epoch": 0.28, "learning_rate": 1.68212523122133e-05, "loss": 23.6959, "step": 23030 }, { "epoch": 0.28, "learning_rate": 1.6818339148608872e-05, "loss": 23.6673, "step": 23040 }, { "epoch": 0.28, "learning_rate": 1.6815424903240177e-05, "loss": 23.9044, "step": 23050 }, { "epoch": 0.28, "learning_rate": 1.681250957656957e-05, "loss": 24.6443, "step": 23060 }, { "epoch": 0.28, "learning_rate": 1.680959316905959e-05, "loss": 23.5114, "step": 23070 }, { "epoch": 0.28, "learning_rate": 1.6806675681172933e-05, "loss": 24.2586, "step": 23080 }, { "epoch": 0.28, "learning_rate": 1.6803757113372482e-05, "loss": 24.0322, "step": 23090 }, { "epoch": 0.28, "learning_rate": 1.6800837466121273e-05, "loss": 24.0084, "step": 23100 }, { "epoch": 0.28, "learning_rate": 1.6797916739882528e-05, "loss": 24.0893, "step": 23110 }, { "epoch": 0.28, "learning_rate": 1.679499493511964e-05, "loss": 23.7399, "step": 23120 }, { "epoch": 0.28, "learning_rate": 1.6792072052296154e-05, "loss": 24.0321, "step": 23130 }, { "epoch": 0.28, "learning_rate": 1.6789148091875813e-05, "loss": 23.4226, "step": 23140 }, { "epoch": 0.28, "learning_rate": 1.6786223054322515e-05, "loss": 23.6697, "step": 23150 }, { "epoch": 0.28, "learning_rate": 1.6783296940100328e-05, "loss": 23.8937, "step": 23160 }, { "epoch": 0.28, "learning_rate": 1.67803697496735e-05, "loss": 24.0334, "step": 23170 }, { "epoch": 0.29, "learning_rate": 1.6777441483506446e-05, "loss": 23.7318, "step": 23180 }, { "epoch": 0.29, "learning_rate": 1.6774512142063745e-05, "loss": 24.0709, "step": 23190 }, { "epoch": 0.29, "learning_rate": 1.6771581725810154e-05, "loss": 24.3248, "step": 23200 }, { "epoch": 0.29, "learning_rate": 1.67686502352106e-05, "loss": 24.1127, "step": 23210 }, { "epoch": 0.29, "learning_rate": 1.6765717670730176e-05, "loss": 23.7361, "step": 23220 }, { "epoch": 0.29, "learning_rate": 1.6762784032834152e-05, "loss": 23.6266, "step": 23230 }, { "epoch": 0.29, "learning_rate": 1.6759849321987963e-05, "loss": 23.2712, "step": 23240 }, { "epoch": 0.29, "learning_rate": 1.6756913538657214e-05, "loss": 23.619, "step": 23250 }, { "epoch": 0.29, "learning_rate": 1.6753976683307686e-05, "loss": 23.9155, "step": 23260 }, { "epoch": 0.29, "learning_rate": 1.675103875640532e-05, "loss": 23.8357, "step": 23270 }, { "epoch": 0.29, "learning_rate": 1.674809975841624e-05, "loss": 23.8084, "step": 23280 }, { "epoch": 0.29, "learning_rate": 1.6745159689806723e-05, "loss": 23.8001, "step": 23290 }, { "epoch": 0.29, "learning_rate": 1.6742218551043236e-05, "loss": 23.7463, "step": 23300 }, { "epoch": 0.29, "learning_rate": 1.6739276342592405e-05, "loss": 23.6867, "step": 23310 }, { "epoch": 0.29, "learning_rate": 1.673633306492102e-05, "loss": 23.6038, "step": 23320 }, { "epoch": 0.29, "learning_rate": 1.6733388718496055e-05, "loss": 23.8593, "step": 23330 }, { "epoch": 0.29, "learning_rate": 1.673044330378464e-05, "loss": 23.5234, "step": 23340 }, { "epoch": 0.29, "learning_rate": 1.672749682125408e-05, "loss": 23.5456, "step": 23350 }, { "epoch": 0.29, "learning_rate": 1.672454927137185e-05, "loss": 23.5758, "step": 23360 }, { "epoch": 0.29, "learning_rate": 1.672160065460559e-05, "loss": 23.631, "step": 23370 }, { "epoch": 0.29, "learning_rate": 1.6718650971423123e-05, "loss": 24.0388, "step": 23380 }, { "epoch": 0.29, "learning_rate": 1.6715700222292426e-05, "loss": 23.6471, "step": 23390 }, { "epoch": 0.29, "learning_rate": 1.6712748407681648e-05, "loss": 23.5376, "step": 23400 }, { "epoch": 0.29, "learning_rate": 1.6709795528059113e-05, "loss": 24.1839, "step": 23410 }, { "epoch": 0.29, "learning_rate": 1.6706841583893312e-05, "loss": 23.5668, "step": 23420 }, { "epoch": 0.29, "learning_rate": 1.6703886575652894e-05, "loss": 23.9304, "step": 23430 }, { "epoch": 0.29, "learning_rate": 1.6700930503806696e-05, "loss": 23.2633, "step": 23440 }, { "epoch": 0.29, "learning_rate": 1.6697973368823713e-05, "loss": 23.792, "step": 23450 }, { "epoch": 0.29, "learning_rate": 1.6695015171173103e-05, "loss": 24.273, "step": 23460 }, { "epoch": 0.29, "learning_rate": 1.6692055911324204e-05, "loss": 23.981, "step": 23470 }, { "epoch": 0.29, "learning_rate": 1.668909558974652e-05, "loss": 23.7412, "step": 23480 }, { "epoch": 0.29, "learning_rate": 1.668613420690971e-05, "loss": 23.7701, "step": 23490 }, { "epoch": 0.29, "learning_rate": 1.6683171763283626e-05, "loss": 24.6431, "step": 23500 }, { "epoch": 0.29, "learning_rate": 1.6680208259338265e-05, "loss": 23.5913, "step": 23510 }, { "epoch": 0.29, "learning_rate": 1.6677243695543813e-05, "loss": 23.8706, "step": 23520 }, { "epoch": 0.29, "learning_rate": 1.66742780723706e-05, "loss": 23.6932, "step": 23530 }, { "epoch": 0.29, "learning_rate": 1.6671311390289147e-05, "loss": 23.8153, "step": 23540 }, { "epoch": 0.29, "learning_rate": 1.6668343649770126e-05, "loss": 23.5339, "step": 23550 }, { "epoch": 0.29, "learning_rate": 1.6665374851284392e-05, "loss": 24.222, "step": 23560 }, { "epoch": 0.29, "learning_rate": 1.6662404995302954e-05, "loss": 23.868, "step": 23570 }, { "epoch": 0.29, "learning_rate": 1.6659434082296998e-05, "loss": 23.9177, "step": 23580 }, { "epoch": 0.29, "learning_rate": 1.665646211273787e-05, "loss": 24.6962, "step": 23590 }, { "epoch": 0.29, "learning_rate": 1.6653489087097093e-05, "loss": 24.0982, "step": 23600 }, { "epoch": 0.29, "learning_rate": 1.6650515005846354e-05, "loss": 25.0599, "step": 23610 }, { "epoch": 0.29, "learning_rate": 1.66475398694575e-05, "loss": 23.8934, "step": 23620 }, { "epoch": 0.29, "learning_rate": 1.6644563678402554e-05, "loss": 24.0639, "step": 23630 }, { "epoch": 0.29, "learning_rate": 1.6641586433153708e-05, "loss": 24.1449, "step": 23640 }, { "epoch": 0.29, "learning_rate": 1.663860813418331e-05, "loss": 23.6849, "step": 23650 }, { "epoch": 0.29, "learning_rate": 1.6635628781963887e-05, "loss": 23.5969, "step": 23660 }, { "epoch": 0.29, "learning_rate": 1.663264837696813e-05, "loss": 23.818, "step": 23670 }, { "epoch": 0.29, "learning_rate": 1.6629666919668894e-05, "loss": 23.4711, "step": 23680 }, { "epoch": 0.29, "learning_rate": 1.6626684410539194e-05, "loss": 23.4485, "step": 23690 }, { "epoch": 0.29, "learning_rate": 1.662370085005223e-05, "loss": 23.9579, "step": 23700 }, { "epoch": 0.29, "learning_rate": 1.662071623868136e-05, "loss": 23.4172, "step": 23710 }, { "epoch": 0.29, "learning_rate": 1.6617730576900104e-05, "loss": 24.2334, "step": 23720 }, { "epoch": 0.29, "learning_rate": 1.661474386518215e-05, "loss": 23.5067, "step": 23730 }, { "epoch": 0.29, "learning_rate": 1.661175610400136e-05, "loss": 23.6438, "step": 23740 }, { "epoch": 0.29, "learning_rate": 1.6608767293831754e-05, "loss": 23.6513, "step": 23750 }, { "epoch": 0.29, "learning_rate": 1.6605777435147527e-05, "loss": 23.7894, "step": 23760 }, { "epoch": 0.29, "learning_rate": 1.660278652842303e-05, "loss": 24.6823, "step": 23770 }, { "epoch": 0.29, "learning_rate": 1.6599794574132786e-05, "loss": 23.4689, "step": 23780 }, { "epoch": 0.29, "learning_rate": 1.6596801572751488e-05, "loss": 24.3244, "step": 23790 }, { "epoch": 0.29, "learning_rate": 1.6593807524753986e-05, "loss": 24.4763, "step": 23800 }, { "epoch": 0.29, "learning_rate": 1.6590812430615304e-05, "loss": 24.2729, "step": 23810 }, { "epoch": 0.29, "learning_rate": 1.6587816290810627e-05, "loss": 23.801, "step": 23820 }, { "epoch": 0.29, "learning_rate": 1.658481910581531e-05, "loss": 23.2851, "step": 23830 }, { "epoch": 0.29, "learning_rate": 1.6581820876104872e-05, "loss": 23.3053, "step": 23840 }, { "epoch": 0.29, "learning_rate": 1.6578821602154992e-05, "loss": 23.9421, "step": 23850 }, { "epoch": 0.29, "learning_rate": 1.6575821284441523e-05, "loss": 24.2441, "step": 23860 }, { "epoch": 0.29, "learning_rate": 1.657281992344049e-05, "loss": 23.6173, "step": 23870 }, { "epoch": 0.29, "learning_rate": 1.656981751962806e-05, "loss": 23.5749, "step": 23880 }, { "epoch": 0.29, "learning_rate": 1.6566814073480586e-05, "loss": 23.8933, "step": 23890 }, { "epoch": 0.29, "learning_rate": 1.656380958547458e-05, "loss": 23.6139, "step": 23900 }, { "epoch": 0.29, "learning_rate": 1.6560804056086724e-05, "loss": 24.0581, "step": 23910 }, { "epoch": 0.29, "learning_rate": 1.655779748579385e-05, "loss": 23.5368, "step": 23920 }, { "epoch": 0.29, "learning_rate": 1.6554789875072973e-05, "loss": 23.5637, "step": 23930 }, { "epoch": 0.29, "learning_rate": 1.6551781224401265e-05, "loss": 24.2755, "step": 23940 }, { "epoch": 0.29, "learning_rate": 1.654877153425606e-05, "loss": 24.1571, "step": 23950 }, { "epoch": 0.29, "learning_rate": 1.6545760805114866e-05, "loss": 24.1233, "step": 23960 }, { "epoch": 0.29, "learning_rate": 1.6542749037455347e-05, "loss": 23.8344, "step": 23970 }, { "epoch": 0.29, "learning_rate": 1.6539736231755338e-05, "loss": 23.8973, "step": 23980 }, { "epoch": 0.3, "learning_rate": 1.6536722388492834e-05, "loss": 23.294, "step": 23990 }, { "epoch": 0.3, "learning_rate": 1.6533707508146e-05, "loss": 23.4463, "step": 24000 }, { "epoch": 0.3, "learning_rate": 1.653069159119315e-05, "loss": 23.9503, "step": 24010 }, { "epoch": 0.3, "learning_rate": 1.6527674638112792e-05, "loss": 24.4508, "step": 24020 }, { "epoch": 0.3, "learning_rate": 1.6524656649383573e-05, "loss": 24.3114, "step": 24030 }, { "epoch": 0.3, "learning_rate": 1.652163762548431e-05, "loss": 23.4861, "step": 24040 }, { "epoch": 0.3, "learning_rate": 1.6518617566893986e-05, "loss": 24.1693, "step": 24050 }, { "epoch": 0.3, "learning_rate": 1.6515596474091754e-05, "loss": 23.2861, "step": 24060 }, { "epoch": 0.3, "learning_rate": 1.651257434755692e-05, "loss": 23.8146, "step": 24070 }, { "epoch": 0.3, "learning_rate": 1.6509551187768967e-05, "loss": 23.7978, "step": 24080 }, { "epoch": 0.3, "learning_rate": 1.650652699520753e-05, "loss": 24.2626, "step": 24090 }, { "epoch": 0.3, "learning_rate": 1.650350177035241e-05, "loss": 23.9523, "step": 24100 }, { "epoch": 0.3, "learning_rate": 1.650047551368358e-05, "loss": 23.6399, "step": 24110 }, { "epoch": 0.3, "learning_rate": 1.649744822568117e-05, "loss": 24.5427, "step": 24120 }, { "epoch": 0.3, "learning_rate": 1.649441990682547e-05, "loss": 23.6425, "step": 24130 }, { "epoch": 0.3, "learning_rate": 1.6491390557596943e-05, "loss": 23.8714, "step": 24140 }, { "epoch": 0.3, "learning_rate": 1.648836017847621e-05, "loss": 23.3679, "step": 24150 }, { "epoch": 0.3, "learning_rate": 1.648532876994405e-05, "loss": 23.8981, "step": 24160 }, { "epoch": 0.3, "learning_rate": 1.6482296332481423e-05, "loss": 23.6482, "step": 24170 }, { "epoch": 0.3, "learning_rate": 1.6479262866569433e-05, "loss": 24.0643, "step": 24180 }, { "epoch": 0.3, "learning_rate": 1.6476228372689353e-05, "loss": 24.1254, "step": 24190 }, { "epoch": 0.3, "learning_rate": 1.6473192851322625e-05, "loss": 24.0154, "step": 24200 }, { "epoch": 0.3, "learning_rate": 1.647015630295085e-05, "loss": 24.012, "step": 24210 }, { "epoch": 0.3, "learning_rate": 1.6467118728055792e-05, "loss": 24.0592, "step": 24220 }, { "epoch": 0.3, "learning_rate": 1.6464080127119377e-05, "loss": 23.8151, "step": 24230 }, { "epoch": 0.3, "learning_rate": 1.646104050062369e-05, "loss": 23.718, "step": 24240 }, { "epoch": 0.3, "learning_rate": 1.645799984905099e-05, "loss": 23.7667, "step": 24250 }, { "epoch": 0.3, "learning_rate": 1.645495817288369e-05, "loss": 23.7022, "step": 24260 }, { "epoch": 0.3, "learning_rate": 1.6451915472604366e-05, "loss": 23.5753, "step": 24270 }, { "epoch": 0.3, "learning_rate": 1.6448871748695758e-05, "loss": 24.0228, "step": 24280 }, { "epoch": 0.3, "learning_rate": 1.6445827001640772e-05, "loss": 24.0671, "step": 24290 }, { "epoch": 0.3, "learning_rate": 1.6442781231922465e-05, "loss": 24.2944, "step": 24300 }, { "epoch": 0.3, "learning_rate": 1.643973444002407e-05, "loss": 24.0859, "step": 24310 }, { "epoch": 0.3, "learning_rate": 1.6436686626428976e-05, "loss": 23.3339, "step": 24320 }, { "epoch": 0.3, "learning_rate": 1.6433637791620732e-05, "loss": 24.2499, "step": 24330 }, { "epoch": 0.3, "learning_rate": 1.643058793608305e-05, "loss": 23.5344, "step": 24340 }, { "epoch": 0.3, "learning_rate": 1.6427537060299807e-05, "loss": 24.2171, "step": 24350 }, { "epoch": 0.3, "learning_rate": 1.642448516475504e-05, "loss": 23.6272, "step": 24360 }, { "epoch": 0.3, "learning_rate": 1.642143224993295e-05, "loss": 23.6746, "step": 24370 }, { "epoch": 0.3, "learning_rate": 1.6418378316317896e-05, "loss": 23.4404, "step": 24380 }, { "epoch": 0.3, "learning_rate": 1.64153233643944e-05, "loss": 24.4688, "step": 24390 }, { "epoch": 0.3, "learning_rate": 1.641226739464714e-05, "loss": 23.8362, "step": 24400 }, { "epoch": 0.3, "learning_rate": 1.6409210407560974e-05, "loss": 24.2723, "step": 24410 }, { "epoch": 0.3, "learning_rate": 1.6406152403620894e-05, "loss": 23.1456, "step": 24420 }, { "epoch": 0.3, "learning_rate": 1.6403093383312078e-05, "loss": 24.1454, "step": 24430 }, { "epoch": 0.3, "learning_rate": 1.640003334711985e-05, "loss": 23.5449, "step": 24440 }, { "epoch": 0.3, "learning_rate": 1.6396972295529706e-05, "loss": 23.6844, "step": 24450 }, { "epoch": 0.3, "learning_rate": 1.639391022902729e-05, "loss": 23.8959, "step": 24460 }, { "epoch": 0.3, "learning_rate": 1.639084714809842e-05, "loss": 24.1408, "step": 24470 }, { "epoch": 0.3, "learning_rate": 1.6387783053229064e-05, "loss": 24.3023, "step": 24480 }, { "epoch": 0.3, "learning_rate": 1.6384717944905367e-05, "loss": 23.9705, "step": 24490 }, { "epoch": 0.3, "learning_rate": 1.638165182361361e-05, "loss": 24.1421, "step": 24500 }, { "epoch": 0.3, "learning_rate": 1.637858468984026e-05, "loss": 23.9782, "step": 24510 }, { "epoch": 0.3, "learning_rate": 1.637551654407193e-05, "loss": 23.4608, "step": 24520 }, { "epoch": 0.3, "learning_rate": 1.6372447386795394e-05, "loss": 24.0648, "step": 24530 }, { "epoch": 0.3, "learning_rate": 1.636937721849759e-05, "loss": 23.9738, "step": 24540 }, { "epoch": 0.3, "learning_rate": 1.6366306039665617e-05, "loss": 24.0382, "step": 24550 }, { "epoch": 0.3, "learning_rate": 1.6363233850786736e-05, "loss": 23.7804, "step": 24560 }, { "epoch": 0.3, "learning_rate": 1.6360160652348364e-05, "loss": 24.0537, "step": 24570 }, { "epoch": 0.3, "learning_rate": 1.6357086444838076e-05, "loss": 23.7889, "step": 24580 }, { "epoch": 0.3, "learning_rate": 1.6354011228743612e-05, "loss": 23.9543, "step": 24590 }, { "epoch": 0.3, "learning_rate": 1.635093500455288e-05, "loss": 24.2738, "step": 24600 }, { "epoch": 0.3, "learning_rate": 1.6347857772753927e-05, "loss": 23.498, "step": 24610 }, { "epoch": 0.3, "learning_rate": 1.6344779533834972e-05, "loss": 24.0833, "step": 24620 }, { "epoch": 0.3, "learning_rate": 1.6341700288284397e-05, "loss": 24.2379, "step": 24630 }, { "epoch": 0.3, "learning_rate": 1.633862003659074e-05, "loss": 23.5369, "step": 24640 }, { "epoch": 0.3, "learning_rate": 1.63355387792427e-05, "loss": 23.8535, "step": 24650 }, { "epoch": 0.3, "learning_rate": 1.633245651672913e-05, "loss": 23.3915, "step": 24660 }, { "epoch": 0.3, "learning_rate": 1.632937324953905e-05, "loss": 24.0843, "step": 24670 }, { "epoch": 0.3, "learning_rate": 1.6326288978161634e-05, "loss": 23.781, "step": 24680 }, { "epoch": 0.3, "learning_rate": 1.6323203703086218e-05, "loss": 23.8035, "step": 24690 }, { "epoch": 0.3, "learning_rate": 1.6320117424802296e-05, "loss": 23.7275, "step": 24700 }, { "epoch": 0.3, "learning_rate": 1.631703014379953e-05, "loss": 24.1647, "step": 24710 }, { "epoch": 0.3, "learning_rate": 1.6313941860567723e-05, "loss": 24.0053, "step": 24720 }, { "epoch": 0.3, "learning_rate": 1.6310852575596844e-05, "loss": 24.3006, "step": 24730 }, { "epoch": 0.3, "learning_rate": 1.6307762289377035e-05, "loss": 23.4891, "step": 24740 }, { "epoch": 0.3, "learning_rate": 1.6304671002398578e-05, "loss": 23.846, "step": 24750 }, { "epoch": 0.3, "learning_rate": 1.6301578715151923e-05, "loss": 23.9006, "step": 24760 }, { "epoch": 0.3, "learning_rate": 1.6298485428127676e-05, "loss": 23.6495, "step": 24770 }, { "epoch": 0.3, "learning_rate": 1.6295391141816607e-05, "loss": 24.5392, "step": 24780 }, { "epoch": 0.3, "learning_rate": 1.6292295856709637e-05, "loss": 24.2192, "step": 24790 }, { "epoch": 0.3, "learning_rate": 1.6289199573297852e-05, "loss": 24.3607, "step": 24800 }, { "epoch": 0.31, "learning_rate": 1.6286102292072487e-05, "loss": 24.1199, "step": 24810 }, { "epoch": 0.31, "learning_rate": 1.6283004013524945e-05, "loss": 24.5419, "step": 24820 }, { "epoch": 0.31, "learning_rate": 1.6279904738146785e-05, "loss": 24.0651, "step": 24830 }, { "epoch": 0.31, "learning_rate": 1.627680446642972e-05, "loss": 24.0487, "step": 24840 }, { "epoch": 0.31, "learning_rate": 1.627370319886563e-05, "loss": 23.9191, "step": 24850 }, { "epoch": 0.31, "learning_rate": 1.627060093594654e-05, "loss": 23.3945, "step": 24860 }, { "epoch": 0.31, "learning_rate": 1.6267497678164642e-05, "loss": 24.0534, "step": 24870 }, { "epoch": 0.31, "learning_rate": 1.6264393426012284e-05, "loss": 24.0287, "step": 24880 }, { "epoch": 0.31, "learning_rate": 1.626128817998197e-05, "loss": 23.9866, "step": 24890 }, { "epoch": 0.31, "learning_rate": 1.6258181940566365e-05, "loss": 24.0151, "step": 24900 }, { "epoch": 0.31, "learning_rate": 1.625507470825829e-05, "loss": 23.872, "step": 24910 }, { "epoch": 0.31, "learning_rate": 1.6251966483550723e-05, "loss": 23.9357, "step": 24920 }, { "epoch": 0.31, "learning_rate": 1.6248857266936794e-05, "loss": 24.4874, "step": 24930 }, { "epoch": 0.31, "learning_rate": 1.6245747058909808e-05, "loss": 24.4526, "step": 24940 }, { "epoch": 0.31, "learning_rate": 1.624263585996321e-05, "loss": 23.9848, "step": 24950 }, { "epoch": 0.31, "learning_rate": 1.62395236705906e-05, "loss": 24.1054, "step": 24960 }, { "epoch": 0.31, "learning_rate": 1.6236410491285747e-05, "loss": 24.6667, "step": 24970 }, { "epoch": 0.31, "learning_rate": 1.623329632254258e-05, "loss": 24.2418, "step": 24980 }, { "epoch": 0.31, "learning_rate": 1.6230181164855174e-05, "loss": 23.8834, "step": 24990 }, { "epoch": 0.31, "learning_rate": 1.6227065018717758e-05, "loss": 23.6216, "step": 25000 }, { "epoch": 0.31, "learning_rate": 1.622394788462473e-05, "loss": 23.7723, "step": 25010 }, { "epoch": 0.31, "learning_rate": 1.622082976307064e-05, "loss": 23.9436, "step": 25020 }, { "epoch": 0.31, "learning_rate": 1.621771065455019e-05, "loss": 23.7476, "step": 25030 }, { "epoch": 0.31, "learning_rate": 1.621459055955825e-05, "loss": 23.8122, "step": 25040 }, { "epoch": 0.31, "learning_rate": 1.6211469478589828e-05, "loss": 23.8143, "step": 25050 }, { "epoch": 0.31, "learning_rate": 1.620834741214011e-05, "loss": 23.8278, "step": 25060 }, { "epoch": 0.31, "learning_rate": 1.6205224360704425e-05, "loss": 24.3667, "step": 25070 }, { "epoch": 0.31, "learning_rate": 1.6202100324778255e-05, "loss": 23.867, "step": 25080 }, { "epoch": 0.31, "learning_rate": 1.619897530485725e-05, "loss": 23.3159, "step": 25090 }, { "epoch": 0.31, "learning_rate": 1.619584930143721e-05, "loss": 23.6848, "step": 25100 }, { "epoch": 0.31, "learning_rate": 1.619272231501409e-05, "loss": 23.7062, "step": 25110 }, { "epoch": 0.31, "learning_rate": 1.6189594346084e-05, "loss": 23.886, "step": 25120 }, { "epoch": 0.31, "learning_rate": 1.618646539514321e-05, "loss": 24.1474, "step": 25130 }, { "epoch": 0.31, "learning_rate": 1.6183335462688146e-05, "loss": 23.9457, "step": 25140 }, { "epoch": 0.31, "learning_rate": 1.618020454921539e-05, "loss": 23.2917, "step": 25150 }, { "epoch": 0.31, "learning_rate": 1.6177072655221667e-05, "loss": 24.1469, "step": 25160 }, { "epoch": 0.31, "learning_rate": 1.617393978120388e-05, "loss": 23.6984, "step": 25170 }, { "epoch": 0.31, "learning_rate": 1.617080592765907e-05, "loss": 23.5455, "step": 25180 }, { "epoch": 0.31, "learning_rate": 1.6167671095084437e-05, "loss": 23.1803, "step": 25190 }, { "epoch": 0.31, "learning_rate": 1.6164535283977343e-05, "loss": 23.5975, "step": 25200 }, { "epoch": 0.31, "learning_rate": 1.6161398494835295e-05, "loss": 23.8957, "step": 25210 }, { "epoch": 0.31, "learning_rate": 1.6158260728155964e-05, "loss": 23.8431, "step": 25220 }, { "epoch": 0.31, "learning_rate": 1.6155121984437173e-05, "loss": 23.6428, "step": 25230 }, { "epoch": 0.31, "learning_rate": 1.61519822641769e-05, "loss": 23.8603, "step": 25240 }, { "epoch": 0.31, "learning_rate": 1.6148841567873272e-05, "loss": 23.8596, "step": 25250 }, { "epoch": 0.31, "learning_rate": 1.614569989602458e-05, "loss": 24.3436, "step": 25260 }, { "epoch": 0.31, "learning_rate": 1.6142557249129272e-05, "loss": 24.1391, "step": 25270 }, { "epoch": 0.31, "learning_rate": 1.6139413627685934e-05, "loss": 24.0862, "step": 25280 }, { "epoch": 0.31, "learning_rate": 1.613626903219333e-05, "loss": 23.5367, "step": 25290 }, { "epoch": 0.31, "learning_rate": 1.613312346315036e-05, "loss": 23.8866, "step": 25300 }, { "epoch": 0.31, "learning_rate": 1.6129976921056082e-05, "loss": 23.3283, "step": 25310 }, { "epoch": 0.31, "learning_rate": 1.612682940640971e-05, "loss": 24.2655, "step": 25320 }, { "epoch": 0.31, "learning_rate": 1.612368091971062e-05, "loss": 23.9261, "step": 25330 }, { "epoch": 0.31, "learning_rate": 1.612053146145833e-05, "loss": 23.4667, "step": 25340 }, { "epoch": 0.31, "learning_rate": 1.611738103215252e-05, "loss": 24.6647, "step": 25350 }, { "epoch": 0.31, "learning_rate": 1.611422963229302e-05, "loss": 23.5173, "step": 25360 }, { "epoch": 0.31, "learning_rate": 1.611107726237982e-05, "loss": 23.3872, "step": 25370 }, { "epoch": 0.31, "learning_rate": 1.610792392291305e-05, "loss": 23.5538, "step": 25380 }, { "epoch": 0.31, "learning_rate": 1.6104769614393014e-05, "loss": 23.6168, "step": 25390 }, { "epoch": 0.31, "learning_rate": 1.6101614337320155e-05, "loss": 24.3522, "step": 25400 }, { "epoch": 0.31, "learning_rate": 1.6098458092195074e-05, "loss": 23.73, "step": 25410 }, { "epoch": 0.31, "learning_rate": 1.6095300879518526e-05, "loss": 23.6345, "step": 25420 }, { "epoch": 0.31, "learning_rate": 1.6092142699791414e-05, "loss": 24.0214, "step": 25430 }, { "epoch": 0.31, "learning_rate": 1.608898355351481e-05, "loss": 23.9159, "step": 25440 }, { "epoch": 0.31, "learning_rate": 1.6085823441189915e-05, "loss": 23.9698, "step": 25450 }, { "epoch": 0.31, "learning_rate": 1.6082662363318107e-05, "loss": 24.3375, "step": 25460 }, { "epoch": 0.31, "learning_rate": 1.60795003204009e-05, "loss": 24.0111, "step": 25470 }, { "epoch": 0.31, "learning_rate": 1.6076337312939977e-05, "loss": 24.0566, "step": 25480 }, { "epoch": 0.31, "learning_rate": 1.607317334143716e-05, "loss": 24.3404, "step": 25490 }, { "epoch": 0.31, "learning_rate": 1.607000840639443e-05, "loss": 24.0548, "step": 25500 }, { "epoch": 0.31, "learning_rate": 1.6066842508313916e-05, "loss": 22.9385, "step": 25510 }, { "epoch": 0.31, "learning_rate": 1.606367564769791e-05, "loss": 23.4208, "step": 25520 }, { "epoch": 0.31, "learning_rate": 1.6060507825048846e-05, "loss": 23.4857, "step": 25530 }, { "epoch": 0.31, "learning_rate": 1.6057339040869322e-05, "loss": 24.0757, "step": 25540 }, { "epoch": 0.31, "learning_rate": 1.6054169295662073e-05, "loss": 24.3102, "step": 25550 }, { "epoch": 0.31, "learning_rate": 1.605099858993e-05, "loss": 23.9799, "step": 25560 }, { "epoch": 0.31, "learning_rate": 1.6047826924176155e-05, "loss": 23.7658, "step": 25570 }, { "epoch": 0.31, "learning_rate": 1.604465429890373e-05, "loss": 24.0371, "step": 25580 }, { "epoch": 0.31, "learning_rate": 1.6041480714616087e-05, "loss": 23.6048, "step": 25590 }, { "epoch": 0.31, "learning_rate": 1.6038306171816725e-05, "loss": 23.6457, "step": 25600 }, { "epoch": 0.31, "learning_rate": 1.6035130671009303e-05, "loss": 24.0677, "step": 25610 }, { "epoch": 0.32, "learning_rate": 1.6031954212697633e-05, "loss": 24.1592, "step": 25620 }, { "epoch": 0.32, "learning_rate": 1.6028776797385675e-05, "loss": 23.6979, "step": 25630 }, { "epoch": 0.32, "learning_rate": 1.602559842557754e-05, "loss": 24.0633, "step": 25640 }, { "epoch": 0.32, "learning_rate": 1.60224190977775e-05, "loss": 24.2821, "step": 25650 }, { "epoch": 0.32, "learning_rate": 1.601923881448996e-05, "loss": 23.9542, "step": 25660 }, { "epoch": 0.32, "learning_rate": 1.6016057576219495e-05, "loss": 23.9441, "step": 25670 }, { "epoch": 0.32, "learning_rate": 1.6012875383470825e-05, "loss": 23.7648, "step": 25680 }, { "epoch": 0.32, "learning_rate": 1.600969223674882e-05, "loss": 24.1907, "step": 25690 }, { "epoch": 0.32, "learning_rate": 1.6006508136558505e-05, "loss": 23.9958, "step": 25700 }, { "epoch": 0.32, "learning_rate": 1.6003323083405047e-05, "loss": 24.3011, "step": 25710 }, { "epoch": 0.32, "learning_rate": 1.6000137077793777e-05, "loss": 24.4302, "step": 25720 }, { "epoch": 0.32, "learning_rate": 1.5996950120230167e-05, "loss": 23.8833, "step": 25730 }, { "epoch": 0.32, "learning_rate": 1.599376221121985e-05, "loss": 23.9487, "step": 25740 }, { "epoch": 0.32, "learning_rate": 1.59905733512686e-05, "loss": 23.737, "step": 25750 }, { "epoch": 0.32, "learning_rate": 1.5987383540882344e-05, "loss": 24.1137, "step": 25760 }, { "epoch": 0.32, "learning_rate": 1.5984192780567166e-05, "loss": 23.6365, "step": 25770 }, { "epoch": 0.32, "learning_rate": 1.5981001070829293e-05, "loss": 23.8878, "step": 25780 }, { "epoch": 0.32, "learning_rate": 1.597780841217511e-05, "loss": 23.7856, "step": 25790 }, { "epoch": 0.32, "learning_rate": 1.597461480511114e-05, "loss": 24.0895, "step": 25800 }, { "epoch": 0.32, "learning_rate": 1.5971420250144075e-05, "loss": 23.7722, "step": 25810 }, { "epoch": 0.32, "learning_rate": 1.5968224747780743e-05, "loss": 24.0822, "step": 25820 }, { "epoch": 0.32, "learning_rate": 1.5965028298528126e-05, "loss": 24.1616, "step": 25830 }, { "epoch": 0.32, "learning_rate": 1.596183090289336e-05, "loss": 23.6045, "step": 25840 }, { "epoch": 0.32, "learning_rate": 1.5958632561383724e-05, "loss": 23.8641, "step": 25850 }, { "epoch": 0.32, "learning_rate": 1.5955433274506654e-05, "loss": 24.0198, "step": 25860 }, { "epoch": 0.32, "learning_rate": 1.5952233042769732e-05, "loss": 23.5012, "step": 25870 }, { "epoch": 0.32, "learning_rate": 1.5949031866680694e-05, "loss": 23.9841, "step": 25880 }, { "epoch": 0.32, "learning_rate": 1.5945829746747423e-05, "loss": 23.4809, "step": 25890 }, { "epoch": 0.32, "learning_rate": 1.5942626683477943e-05, "loss": 23.6624, "step": 25900 }, { "epoch": 0.32, "learning_rate": 1.5939422677380446e-05, "loss": 24.2558, "step": 25910 }, { "epoch": 0.32, "learning_rate": 1.5936217728963258e-05, "loss": 24.3139, "step": 25920 }, { "epoch": 0.32, "learning_rate": 1.5933011838734865e-05, "loss": 24.4704, "step": 25930 }, { "epoch": 0.32, "learning_rate": 1.5929805007203895e-05, "loss": 24.2824, "step": 25940 }, { "epoch": 0.32, "learning_rate": 1.5926597234879125e-05, "loss": 24.1909, "step": 25950 }, { "epoch": 0.32, "learning_rate": 1.5923388522269495e-05, "loss": 24.467, "step": 25960 }, { "epoch": 0.32, "learning_rate": 1.592017886988407e-05, "loss": 23.9017, "step": 25970 }, { "epoch": 0.32, "learning_rate": 1.591696827823209e-05, "loss": 23.9199, "step": 25980 }, { "epoch": 0.32, "learning_rate": 1.5913756747822925e-05, "loss": 24.4481, "step": 25990 }, { "epoch": 0.32, "learning_rate": 1.59105442791661e-05, "loss": 23.7022, "step": 26000 }, { "epoch": 0.32, "learning_rate": 1.5907330872771295e-05, "loss": 24.2219, "step": 26010 }, { "epoch": 0.32, "learning_rate": 1.5904116529148326e-05, "loss": 24.2151, "step": 26020 }, { "epoch": 0.32, "learning_rate": 1.590090124880717e-05, "loss": 23.7547, "step": 26030 }, { "epoch": 0.32, "learning_rate": 1.5897685032257946e-05, "loss": 24.0158, "step": 26040 }, { "epoch": 0.32, "learning_rate": 1.5894467880010926e-05, "loss": 23.7037, "step": 26050 }, { "epoch": 0.32, "learning_rate": 1.5891249792576523e-05, "loss": 24.9681, "step": 26060 }, { "epoch": 0.32, "learning_rate": 1.588803077046531e-05, "loss": 23.9304, "step": 26070 }, { "epoch": 0.32, "learning_rate": 1.588481081418799e-05, "loss": 23.9923, "step": 26080 }, { "epoch": 0.32, "learning_rate": 1.5881589924255434e-05, "loss": 24.004, "step": 26090 }, { "epoch": 0.32, "learning_rate": 1.587836810117865e-05, "loss": 23.499, "step": 26100 }, { "epoch": 0.32, "learning_rate": 1.5875145345468797e-05, "loss": 24.0943, "step": 26110 }, { "epoch": 0.32, "learning_rate": 1.5871921657637184e-05, "loss": 24.3287, "step": 26120 }, { "epoch": 0.32, "learning_rate": 1.5868697038195265e-05, "loss": 23.5576, "step": 26130 }, { "epoch": 0.32, "learning_rate": 1.5865471487654635e-05, "loss": 24.1383, "step": 26140 }, { "epoch": 0.32, "learning_rate": 1.586224500652705e-05, "loss": 24.4201, "step": 26150 }, { "epoch": 0.32, "learning_rate": 1.5859017595324413e-05, "loss": 24.0583, "step": 26160 }, { "epoch": 0.32, "learning_rate": 1.585578925455876e-05, "loss": 23.9366, "step": 26170 }, { "epoch": 0.32, "learning_rate": 1.585255998474229e-05, "loss": 23.7651, "step": 26180 }, { "epoch": 0.32, "learning_rate": 1.5849329786387337e-05, "loss": 24.0752, "step": 26190 }, { "epoch": 0.32, "learning_rate": 1.584609866000639e-05, "loss": 23.7076, "step": 26200 }, { "epoch": 0.32, "learning_rate": 1.5842866606112092e-05, "loss": 23.8832, "step": 26210 }, { "epoch": 0.32, "learning_rate": 1.5839633625217214e-05, "loss": 24.091, "step": 26220 }, { "epoch": 0.32, "learning_rate": 1.583639971783469e-05, "loss": 24.1749, "step": 26230 }, { "epoch": 0.32, "learning_rate": 1.5833164884477595e-05, "loss": 23.834, "step": 26240 }, { "epoch": 0.32, "learning_rate": 1.5829929125659152e-05, "loss": 23.7587, "step": 26250 }, { "epoch": 0.32, "learning_rate": 1.582669244189273e-05, "loss": 24.1579, "step": 26260 }, { "epoch": 0.32, "learning_rate": 1.5823454833691846e-05, "loss": 23.8553, "step": 26270 }, { "epoch": 0.32, "learning_rate": 1.5820216301570165e-05, "loss": 24.1461, "step": 26280 }, { "epoch": 0.32, "learning_rate": 1.5816976846041494e-05, "loss": 24.1862, "step": 26290 }, { "epoch": 0.32, "learning_rate": 1.5813736467619788e-05, "loss": 23.4259, "step": 26300 }, { "epoch": 0.32, "learning_rate": 1.5810495166819155e-05, "loss": 23.5434, "step": 26310 }, { "epoch": 0.32, "learning_rate": 1.5807252944153835e-05, "loss": 24.0224, "step": 26320 }, { "epoch": 0.32, "learning_rate": 1.5804009800138233e-05, "loss": 24.2075, "step": 26330 }, { "epoch": 0.32, "learning_rate": 1.5800765735286886e-05, "loss": 23.8479, "step": 26340 }, { "epoch": 0.32, "learning_rate": 1.579752075011448e-05, "loss": 24.2749, "step": 26350 }, { "epoch": 0.32, "learning_rate": 1.579427484513585e-05, "loss": 23.9587, "step": 26360 }, { "epoch": 0.32, "learning_rate": 1.579102802086597e-05, "loss": 23.832, "step": 26370 }, { "epoch": 0.32, "learning_rate": 1.5787780277819976e-05, "loss": 24.2598, "step": 26380 }, { "epoch": 0.32, "learning_rate": 1.578453161651313e-05, "loss": 23.8751, "step": 26390 }, { "epoch": 0.32, "learning_rate": 1.5781282037460854e-05, "loss": 23.7584, "step": 26400 }, { "epoch": 0.32, "learning_rate": 1.5778031541178707e-05, "loss": 23.7278, "step": 26410 }, { "epoch": 0.32, "learning_rate": 1.57747801281824e-05, "loss": 24.6193, "step": 26420 }, { "epoch": 0.33, "learning_rate": 1.5771527798987782e-05, "loss": 24.0783, "step": 26430 }, { "epoch": 0.33, "learning_rate": 1.5768274554110853e-05, "loss": 24.3887, "step": 26440 }, { "epoch": 0.33, "learning_rate": 1.5765020394067757e-05, "loss": 24.2698, "step": 26450 }, { "epoch": 0.33, "learning_rate": 1.576176531937479e-05, "loss": 24.2362, "step": 26460 }, { "epoch": 0.33, "learning_rate": 1.5758509330548376e-05, "loss": 24.0514, "step": 26470 }, { "epoch": 0.33, "learning_rate": 1.5755252428105095e-05, "loss": 23.7875, "step": 26480 }, { "epoch": 0.33, "learning_rate": 1.575199461256168e-05, "loss": 23.959, "step": 26490 }, { "epoch": 0.33, "learning_rate": 1.574873588443499e-05, "loss": 23.5397, "step": 26500 }, { "epoch": 0.33, "learning_rate": 1.5745476244242047e-05, "loss": 23.8874, "step": 26510 }, { "epoch": 0.33, "learning_rate": 1.5742215692500003e-05, "loss": 23.9672, "step": 26520 }, { "epoch": 0.33, "learning_rate": 1.5738954229726164e-05, "loss": 23.6193, "step": 26530 }, { "epoch": 0.33, "learning_rate": 1.573569185643798e-05, "loss": 24.4514, "step": 26540 }, { "epoch": 0.33, "learning_rate": 1.5732428573153038e-05, "loss": 23.8281, "step": 26550 }, { "epoch": 0.33, "learning_rate": 1.572916438038908e-05, "loss": 23.9211, "step": 26560 }, { "epoch": 0.33, "learning_rate": 1.5725899278663982e-05, "loss": 23.7026, "step": 26570 }, { "epoch": 0.33, "learning_rate": 1.5722633268495774e-05, "loss": 23.629, "step": 26580 }, { "epoch": 0.33, "learning_rate": 1.571936635040262e-05, "loss": 23.8867, "step": 26590 }, { "epoch": 0.33, "learning_rate": 1.5716098524902836e-05, "loss": 23.8914, "step": 26600 }, { "epoch": 0.33, "learning_rate": 1.571282979251488e-05, "loss": 24.2461, "step": 26610 }, { "epoch": 0.33, "learning_rate": 1.5709560153757348e-05, "loss": 23.6546, "step": 26620 }, { "epoch": 0.33, "learning_rate": 1.570628960914899e-05, "loss": 24.2063, "step": 26630 }, { "epoch": 0.33, "learning_rate": 1.5703018159208697e-05, "loss": 24.1464, "step": 26640 }, { "epoch": 0.33, "learning_rate": 1.5699745804455496e-05, "loss": 23.509, "step": 26650 }, { "epoch": 0.33, "learning_rate": 1.569647254540856e-05, "loss": 23.8654, "step": 26660 }, { "epoch": 0.33, "learning_rate": 1.5693198382587217e-05, "loss": 24.078, "step": 26670 }, { "epoch": 0.33, "learning_rate": 1.5689923316510923e-05, "loss": 24.5286, "step": 26680 }, { "epoch": 0.33, "learning_rate": 1.568664734769929e-05, "loss": 23.5339, "step": 26690 }, { "epoch": 0.33, "learning_rate": 1.568337047667206e-05, "loss": 23.7137, "step": 26700 }, { "epoch": 0.33, "learning_rate": 1.5680092703949127e-05, "loss": 23.58, "step": 26710 }, { "epoch": 0.33, "learning_rate": 1.567681403005053e-05, "loss": 23.8721, "step": 26720 }, { "epoch": 0.33, "learning_rate": 1.5673534455496448e-05, "loss": 24.0127, "step": 26730 }, { "epoch": 0.33, "learning_rate": 1.5670253980807195e-05, "loss": 23.6559, "step": 26740 }, { "epoch": 0.33, "learning_rate": 1.566697260650324e-05, "loss": 23.776, "step": 26750 }, { "epoch": 0.33, "learning_rate": 1.566369033310519e-05, "loss": 23.3536, "step": 26760 }, { "epoch": 0.33, "learning_rate": 1.566040716113379e-05, "loss": 23.8578, "step": 26770 }, { "epoch": 0.33, "learning_rate": 1.565712309110994e-05, "loss": 23.8421, "step": 26780 }, { "epoch": 0.33, "learning_rate": 1.5653838123554666e-05, "loss": 23.5497, "step": 26790 }, { "epoch": 0.33, "learning_rate": 1.5650552258989154e-05, "loss": 23.7938, "step": 26800 }, { "epoch": 0.33, "learning_rate": 1.5647265497934718e-05, "loss": 24.0382, "step": 26810 }, { "epoch": 0.33, "learning_rate": 1.5643977840912814e-05, "loss": 23.8806, "step": 26820 }, { "epoch": 0.33, "learning_rate": 1.5640689288445053e-05, "loss": 23.3545, "step": 26830 }, { "epoch": 0.33, "learning_rate": 1.563739984105318e-05, "loss": 23.9125, "step": 26840 }, { "epoch": 0.33, "learning_rate": 1.563410949925908e-05, "loss": 24.2597, "step": 26850 }, { "epoch": 0.33, "learning_rate": 1.5630818263584787e-05, "loss": 23.748, "step": 26860 }, { "epoch": 0.33, "learning_rate": 1.5627526134552465e-05, "loss": 23.811, "step": 26870 }, { "epoch": 0.33, "learning_rate": 1.562423311268443e-05, "loss": 23.7764, "step": 26880 }, { "epoch": 0.33, "learning_rate": 1.5620939198503142e-05, "loss": 22.9613, "step": 26890 }, { "epoch": 0.33, "learning_rate": 1.561764439253119e-05, "loss": 23.853, "step": 26900 }, { "epoch": 0.33, "learning_rate": 1.5614348695291313e-05, "loss": 23.9515, "step": 26910 }, { "epoch": 0.33, "learning_rate": 1.56110521073064e-05, "loss": 23.2089, "step": 26920 }, { "epoch": 0.33, "learning_rate": 1.5607754629099455e-05, "loss": 23.5927, "step": 26930 }, { "epoch": 0.33, "learning_rate": 1.5604456261193654e-05, "loss": 24.1339, "step": 26940 }, { "epoch": 0.33, "learning_rate": 1.5601157004112292e-05, "loss": 23.9706, "step": 26950 }, { "epoch": 0.33, "learning_rate": 1.5597856858378816e-05, "loss": 23.8346, "step": 26960 }, { "epoch": 0.33, "learning_rate": 1.5594555824516807e-05, "loss": 24.1769, "step": 26970 }, { "epoch": 0.33, "learning_rate": 1.5591253903049993e-05, "loss": 24.1906, "step": 26980 }, { "epoch": 0.33, "learning_rate": 1.5587951094502246e-05, "loss": 23.9829, "step": 26990 }, { "epoch": 0.33, "learning_rate": 1.5584647399397565e-05, "loss": 23.6515, "step": 27000 }, { "epoch": 0.33, "learning_rate": 1.5581342818260106e-05, "loss": 23.4487, "step": 27010 }, { "epoch": 0.33, "learning_rate": 1.557803735161415e-05, "loss": 23.626, "step": 27020 }, { "epoch": 0.33, "learning_rate": 1.557473099998413e-05, "loss": 24.4735, "step": 27030 }, { "epoch": 0.33, "learning_rate": 1.5571423763894618e-05, "loss": 24.7083, "step": 27040 }, { "epoch": 0.33, "learning_rate": 1.556811564387032e-05, "loss": 23.9088, "step": 27050 }, { "epoch": 0.33, "learning_rate": 1.556480664043609e-05, "loss": 23.9534, "step": 27060 }, { "epoch": 0.33, "learning_rate": 1.5561496754116908e-05, "loss": 23.9442, "step": 27070 }, { "epoch": 0.33, "learning_rate": 1.555818598543792e-05, "loss": 24.0116, "step": 27080 }, { "epoch": 0.33, "learning_rate": 1.5554874334924384e-05, "loss": 23.5007, "step": 27090 }, { "epoch": 0.33, "learning_rate": 1.5551561803101714e-05, "loss": 24.1498, "step": 27100 }, { "epoch": 0.33, "learning_rate": 1.5548248390495463e-05, "loss": 23.8849, "step": 27110 }, { "epoch": 0.33, "learning_rate": 1.5544934097631315e-05, "loss": 23.8675, "step": 27120 }, { "epoch": 0.33, "learning_rate": 1.5541618925035103e-05, "loss": 23.6366, "step": 27130 }, { "epoch": 0.33, "learning_rate": 1.55383028732328e-05, "loss": 23.6276, "step": 27140 }, { "epoch": 0.33, "learning_rate": 1.5534985942750508e-05, "loss": 23.7463, "step": 27150 }, { "epoch": 0.33, "learning_rate": 1.553166813411448e-05, "loss": 24.3261, "step": 27160 }, { "epoch": 0.33, "learning_rate": 1.5528349447851096e-05, "loss": 24.3931, "step": 27170 }, { "epoch": 0.33, "learning_rate": 1.5525029884486888e-05, "loss": 22.9999, "step": 27180 }, { "epoch": 0.33, "learning_rate": 1.552170944454852e-05, "loss": 23.7866, "step": 27190 }, { "epoch": 0.33, "learning_rate": 1.55183881285628e-05, "loss": 23.6765, "step": 27200 }, { "epoch": 0.33, "learning_rate": 1.5515065937056667e-05, "loss": 23.9044, "step": 27210 }, { "epoch": 0.33, "learning_rate": 1.5511742870557208e-05, "loss": 23.4037, "step": 27220 }, { "epoch": 0.33, "learning_rate": 1.550841892959164e-05, "loss": 23.9549, "step": 27230 }, { "epoch": 0.34, "learning_rate": 1.5505094114687328e-05, "loss": 23.786, "step": 27240 }, { "epoch": 0.34, "learning_rate": 1.5501768426371764e-05, "loss": 23.4423, "step": 27250 }, { "epoch": 0.34, "learning_rate": 1.5498441865172594e-05, "loss": 23.2725, "step": 27260 }, { "epoch": 0.34, "learning_rate": 1.549511443161759e-05, "loss": 23.7025, "step": 27270 }, { "epoch": 0.34, "learning_rate": 1.5491786126234663e-05, "loss": 23.3385, "step": 27280 }, { "epoch": 0.34, "learning_rate": 1.5488456949551868e-05, "loss": 24.0091, "step": 27290 }, { "epoch": 0.34, "learning_rate": 1.5485126902097397e-05, "loss": 23.2676, "step": 27300 }, { "epoch": 0.34, "learning_rate": 1.5481795984399578e-05, "loss": 24.1562, "step": 27310 }, { "epoch": 0.34, "learning_rate": 1.547846419698688e-05, "loss": 23.703, "step": 27320 }, { "epoch": 0.34, "learning_rate": 1.5475131540387903e-05, "loss": 23.9109, "step": 27330 }, { "epoch": 0.34, "learning_rate": 1.5471798015131394e-05, "loss": 24.4486, "step": 27340 }, { "epoch": 0.34, "learning_rate": 1.5468463621746237e-05, "loss": 23.9736, "step": 27350 }, { "epoch": 0.34, "learning_rate": 1.5465128360761444e-05, "loss": 24.0726, "step": 27360 }, { "epoch": 0.34, "learning_rate": 1.546179223270617e-05, "loss": 23.7691, "step": 27370 }, { "epoch": 0.34, "learning_rate": 1.545845523810972e-05, "loss": 23.6408, "step": 27380 }, { "epoch": 0.34, "learning_rate": 1.5455117377501514e-05, "loss": 23.4078, "step": 27390 }, { "epoch": 0.34, "learning_rate": 1.5451778651411125e-05, "loss": 23.4619, "step": 27400 }, { "epoch": 0.34, "learning_rate": 1.5448439060368258e-05, "loss": 23.8515, "step": 27410 }, { "epoch": 0.34, "learning_rate": 1.544509860490276e-05, "loss": 24.0204, "step": 27420 }, { "epoch": 0.34, "learning_rate": 1.54417572855446e-05, "loss": 24.0445, "step": 27430 }, { "epoch": 0.34, "learning_rate": 1.5438415102823907e-05, "loss": 24.2541, "step": 27440 }, { "epoch": 0.34, "learning_rate": 1.543507205727093e-05, "loss": 23.6017, "step": 27450 }, { "epoch": 0.34, "learning_rate": 1.5431728149416066e-05, "loss": 23.7841, "step": 27460 }, { "epoch": 0.34, "learning_rate": 1.5428383379789834e-05, "loss": 23.7225, "step": 27470 }, { "epoch": 0.34, "learning_rate": 1.5425037748922903e-05, "loss": 23.285, "step": 27480 }, { "epoch": 0.34, "learning_rate": 1.5421691257346082e-05, "loss": 23.9643, "step": 27490 }, { "epoch": 0.34, "learning_rate": 1.5418343905590295e-05, "loss": 23.8275, "step": 27500 }, { "epoch": 0.34, "learning_rate": 1.5414995694186625e-05, "loss": 24.9553, "step": 27510 }, { "epoch": 0.34, "learning_rate": 1.5411646623666282e-05, "loss": 24.3785, "step": 27520 }, { "epoch": 0.34, "learning_rate": 1.540829669456061e-05, "loss": 23.6721, "step": 27530 }, { "epoch": 0.34, "learning_rate": 1.5404945907401096e-05, "loss": 23.7216, "step": 27540 }, { "epoch": 0.34, "learning_rate": 1.5401594262719356e-05, "loss": 24.0172, "step": 27550 }, { "epoch": 0.34, "learning_rate": 1.539824176104715e-05, "loss": 24.3255, "step": 27560 }, { "epoch": 0.34, "learning_rate": 1.5394888402916362e-05, "loss": 23.6134, "step": 27570 }, { "epoch": 0.34, "learning_rate": 1.5391534188859026e-05, "loss": 23.6836, "step": 27580 }, { "epoch": 0.34, "learning_rate": 1.5388179119407304e-05, "loss": 24.1574, "step": 27590 }, { "epoch": 0.34, "learning_rate": 1.53848231950935e-05, "loss": 23.919, "step": 27600 }, { "epoch": 0.34, "learning_rate": 1.5381466416450036e-05, "loss": 23.8926, "step": 27610 }, { "epoch": 0.34, "learning_rate": 1.537810878400949e-05, "loss": 24.1323, "step": 27620 }, { "epoch": 0.34, "learning_rate": 1.5374750298304567e-05, "loss": 24.1367, "step": 27630 }, { "epoch": 0.34, "learning_rate": 1.537139095986811e-05, "loss": 24.1044, "step": 27640 }, { "epoch": 0.34, "learning_rate": 1.5368030769233088e-05, "loss": 24.0231, "step": 27650 }, { "epoch": 0.34, "learning_rate": 1.5364669726932617e-05, "loss": 24.3142, "step": 27660 }, { "epoch": 0.34, "learning_rate": 1.5361307833499947e-05, "loss": 23.8774, "step": 27670 }, { "epoch": 0.34, "learning_rate": 1.5357945089468452e-05, "loss": 24.3449, "step": 27680 }, { "epoch": 0.34, "learning_rate": 1.5354581495371652e-05, "loss": 24.3263, "step": 27690 }, { "epoch": 0.34, "learning_rate": 1.5351217051743203e-05, "loss": 23.5136, "step": 27700 }, { "epoch": 0.34, "learning_rate": 1.5347851759116882e-05, "loss": 23.6243, "step": 27710 }, { "epoch": 0.34, "learning_rate": 1.5344485618026617e-05, "loss": 23.5398, "step": 27720 }, { "epoch": 0.34, "learning_rate": 1.534111862900646e-05, "loss": 23.8775, "step": 27730 }, { "epoch": 0.34, "learning_rate": 1.5337750792590603e-05, "loss": 23.9475, "step": 27740 }, { "epoch": 0.34, "learning_rate": 1.5334382109313367e-05, "loss": 24.3237, "step": 27750 }, { "epoch": 0.34, "learning_rate": 1.5331012579709216e-05, "loss": 23.9805, "step": 27760 }, { "epoch": 0.34, "learning_rate": 1.532764220431274e-05, "loss": 24.2836, "step": 27770 }, { "epoch": 0.34, "learning_rate": 1.5324270983658667e-05, "loss": 23.963, "step": 27780 }, { "epoch": 0.34, "learning_rate": 1.5320898918281858e-05, "loss": 24.1432, "step": 27790 }, { "epoch": 0.34, "learning_rate": 1.531752600871731e-05, "loss": 24.0297, "step": 27800 }, { "epoch": 0.34, "learning_rate": 1.531415225550014e-05, "loss": 24.0753, "step": 27810 }, { "epoch": 0.34, "learning_rate": 1.531077765916563e-05, "loss": 23.1752, "step": 27820 }, { "epoch": 0.34, "learning_rate": 1.530740222024917e-05, "loss": 24.0225, "step": 27830 }, { "epoch": 0.34, "learning_rate": 1.5304025939286287e-05, "loss": 23.8036, "step": 27840 }, { "epoch": 0.34, "learning_rate": 1.5300648816812647e-05, "loss": 24.0298, "step": 27850 }, { "epoch": 0.34, "learning_rate": 1.5297270853364046e-05, "loss": 23.7775, "step": 27860 }, { "epoch": 0.34, "learning_rate": 1.5293892049476422e-05, "loss": 23.5664, "step": 27870 }, { "epoch": 0.34, "learning_rate": 1.529051240568583e-05, "loss": 23.5608, "step": 27880 }, { "epoch": 0.34, "learning_rate": 1.528713192252847e-05, "loss": 24.7328, "step": 27890 }, { "epoch": 0.34, "learning_rate": 1.5283750600540682e-05, "loss": 23.3877, "step": 27900 }, { "epoch": 0.34, "learning_rate": 1.5280368440258923e-05, "loss": 24.0781, "step": 27910 }, { "epoch": 0.34, "learning_rate": 1.5276985442219786e-05, "loss": 23.852, "step": 27920 }, { "epoch": 0.34, "learning_rate": 1.5273601606960007e-05, "loss": 24.1835, "step": 27930 }, { "epoch": 0.34, "learning_rate": 1.5270216935016447e-05, "loss": 24.1838, "step": 27940 }, { "epoch": 0.34, "learning_rate": 1.52668314269261e-05, "loss": 23.9898, "step": 27950 }, { "epoch": 0.34, "learning_rate": 1.5263445083226096e-05, "loss": 24.0574, "step": 27960 }, { "epoch": 0.34, "learning_rate": 1.52600579044537e-05, "loss": 23.7356, "step": 27970 }, { "epoch": 0.34, "learning_rate": 1.5256669891146292e-05, "loss": 23.2038, "step": 27980 }, { "epoch": 0.34, "learning_rate": 1.5253281043841405e-05, "loss": 23.821, "step": 27990 }, { "epoch": 0.34, "learning_rate": 1.5249891363076705e-05, "loss": 23.8749, "step": 28000 }, { "epoch": 0.34, "learning_rate": 1.5246500849389971e-05, "loss": 23.8604, "step": 28010 }, { "epoch": 0.34, "learning_rate": 1.5243109503319128e-05, "loss": 23.446, "step": 28020 }, { "epoch": 0.34, "learning_rate": 1.5239717325402237e-05, "loss": 24.3174, "step": 28030 }, { "epoch": 0.34, "learning_rate": 1.5236324316177471e-05, "loss": 23.8808, "step": 28040 }, { "epoch": 0.34, "learning_rate": 1.5232930476183161e-05, "loss": 23.4295, "step": 28050 }, { "epoch": 0.35, "learning_rate": 1.5229535805957751e-05, "loss": 23.8794, "step": 28060 }, { "epoch": 0.35, "learning_rate": 1.5226140306039825e-05, "loss": 24.448, "step": 28070 }, { "epoch": 0.35, "learning_rate": 1.5222743976968093e-05, "loss": 23.5676, "step": 28080 }, { "epoch": 0.35, "learning_rate": 1.5219346819281403e-05, "loss": 24.311, "step": 28090 }, { "epoch": 0.35, "learning_rate": 1.521594883351873e-05, "loss": 23.4992, "step": 28100 }, { "epoch": 0.35, "learning_rate": 1.5212550020219184e-05, "loss": 24.2672, "step": 28110 }, { "epoch": 0.35, "learning_rate": 1.5209150379921997e-05, "loss": 23.7986, "step": 28120 }, { "epoch": 0.35, "learning_rate": 1.520574991316655e-05, "loss": 23.893, "step": 28130 }, { "epoch": 0.35, "learning_rate": 1.5202348620492337e-05, "loss": 24.0511, "step": 28140 }, { "epoch": 0.35, "learning_rate": 1.519894650243899e-05, "loss": 24.4849, "step": 28150 }, { "epoch": 0.35, "learning_rate": 1.519554355954628e-05, "loss": 23.9759, "step": 28160 }, { "epoch": 0.35, "learning_rate": 1.5192139792354095e-05, "loss": 23.4332, "step": 28170 }, { "epoch": 0.35, "learning_rate": 1.5188735201402458e-05, "loss": 23.2182, "step": 28180 }, { "epoch": 0.35, "learning_rate": 1.5185329787231528e-05, "loss": 24.0544, "step": 28190 }, { "epoch": 0.35, "learning_rate": 1.5181923550381595e-05, "loss": 23.7919, "step": 28200 }, { "epoch": 0.35, "learning_rate": 1.5178516491393067e-05, "loss": 24.4377, "step": 28210 }, { "epoch": 0.35, "learning_rate": 1.5175108610806497e-05, "loss": 23.9316, "step": 28220 }, { "epoch": 0.35, "learning_rate": 1.5171699909162563e-05, "loss": 23.9327, "step": 28230 }, { "epoch": 0.35, "learning_rate": 1.5168290387002072e-05, "loss": 23.8105, "step": 28240 }, { "epoch": 0.35, "learning_rate": 1.516488004486596e-05, "loss": 24.1498, "step": 28250 }, { "epoch": 0.35, "learning_rate": 1.5161468883295296e-05, "loss": 24.1117, "step": 28260 }, { "epoch": 0.35, "learning_rate": 1.5158056902831281e-05, "loss": 24.4921, "step": 28270 }, { "epoch": 0.35, "learning_rate": 1.5154644104015242e-05, "loss": 23.8319, "step": 28280 }, { "epoch": 0.35, "learning_rate": 1.5151230487388633e-05, "loss": 23.8589, "step": 28290 }, { "epoch": 0.35, "learning_rate": 1.5147816053493047e-05, "loss": 23.8887, "step": 28300 }, { "epoch": 0.35, "learning_rate": 1.5144400802870203e-05, "loss": 23.9376, "step": 28310 }, { "epoch": 0.35, "learning_rate": 1.5140984736061937e-05, "loss": 23.8828, "step": 28320 }, { "epoch": 0.35, "learning_rate": 1.5137567853610236e-05, "loss": 24.0641, "step": 28330 }, { "epoch": 0.35, "learning_rate": 1.5134150156057201e-05, "loss": 23.5425, "step": 28340 }, { "epoch": 0.35, "learning_rate": 1.5130731643945069e-05, "loss": 23.4813, "step": 28350 }, { "epoch": 0.35, "learning_rate": 1.5127312317816202e-05, "loss": 23.6644, "step": 28360 }, { "epoch": 0.35, "learning_rate": 1.51238921782131e-05, "loss": 23.6344, "step": 28370 }, { "epoch": 0.35, "learning_rate": 1.5120471225678377e-05, "loss": 23.4094, "step": 28380 }, { "epoch": 0.35, "learning_rate": 1.511704946075479e-05, "loss": 24.2468, "step": 28390 }, { "epoch": 0.35, "learning_rate": 1.5113626883985217e-05, "loss": 24.2399, "step": 28400 }, { "epoch": 0.35, "learning_rate": 1.5110203495912667e-05, "loss": 23.0542, "step": 28410 }, { "epoch": 0.35, "learning_rate": 1.5106779297080282e-05, "loss": 23.9571, "step": 28420 }, { "epoch": 0.35, "learning_rate": 1.5103354288031324e-05, "loss": 23.6241, "step": 28430 }, { "epoch": 0.35, "learning_rate": 1.5099928469309193e-05, "loss": 23.9132, "step": 28440 }, { "epoch": 0.35, "learning_rate": 1.5096501841457405e-05, "loss": 23.3399, "step": 28450 }, { "epoch": 0.35, "learning_rate": 1.5093074405019619e-05, "loss": 23.738, "step": 28460 }, { "epoch": 0.35, "learning_rate": 1.5089646160539611e-05, "loss": 23.8351, "step": 28470 }, { "epoch": 0.35, "learning_rate": 1.5086217108561292e-05, "loss": 24.1023, "step": 28480 }, { "epoch": 0.35, "learning_rate": 1.5082787249628697e-05, "loss": 23.7405, "step": 28490 }, { "epoch": 0.35, "learning_rate": 1.5079356584285993e-05, "loss": 23.7594, "step": 28500 }, { "epoch": 0.35, "learning_rate": 1.507592511307747e-05, "loss": 23.7153, "step": 28510 }, { "epoch": 0.35, "learning_rate": 1.5072492836547548e-05, "loss": 23.911, "step": 28520 }, { "epoch": 0.35, "learning_rate": 1.506905975524078e-05, "loss": 23.5096, "step": 28530 }, { "epoch": 0.35, "learning_rate": 1.5065625869701838e-05, "loss": 23.6662, "step": 28540 }, { "epoch": 0.35, "learning_rate": 1.5062191180475523e-05, "loss": 23.471, "step": 28550 }, { "epoch": 0.35, "learning_rate": 1.505875568810677e-05, "loss": 23.6225, "step": 28560 }, { "epoch": 0.35, "learning_rate": 1.5055319393140635e-05, "loss": 23.8291, "step": 28570 }, { "epoch": 0.35, "learning_rate": 1.5051882296122306e-05, "loss": 23.3642, "step": 28580 }, { "epoch": 0.35, "learning_rate": 1.5048444397597094e-05, "loss": 23.9764, "step": 28590 }, { "epoch": 0.35, "learning_rate": 1.5045005698110441e-05, "loss": 24.0218, "step": 28600 }, { "epoch": 0.35, "learning_rate": 1.5041566198207912e-05, "loss": 23.8404, "step": 28610 }, { "epoch": 0.35, "learning_rate": 1.5038125898435204e-05, "loss": 23.6427, "step": 28620 }, { "epoch": 0.35, "learning_rate": 1.5034684799338135e-05, "loss": 23.951, "step": 28630 }, { "epoch": 0.35, "learning_rate": 1.5031242901462655e-05, "loss": 23.5411, "step": 28640 }, { "epoch": 0.35, "learning_rate": 1.5027800205354841e-05, "loss": 24.3194, "step": 28650 }, { "epoch": 0.35, "learning_rate": 1.5024356711560886e-05, "loss": 24.0958, "step": 28660 }, { "epoch": 0.35, "learning_rate": 1.5020912420627127e-05, "loss": 23.6645, "step": 28670 }, { "epoch": 0.35, "learning_rate": 1.501746733310001e-05, "loss": 23.8984, "step": 28680 }, { "epoch": 0.35, "learning_rate": 1.5014021449526124e-05, "loss": 24.0022, "step": 28690 }, { "epoch": 0.35, "learning_rate": 1.501057477045217e-05, "loss": 23.5394, "step": 28700 }, { "epoch": 0.35, "learning_rate": 1.5007127296424987e-05, "loss": 24.0867, "step": 28710 }, { "epoch": 0.35, "learning_rate": 1.5003679027991526e-05, "loss": 24.0338, "step": 28720 }, { "epoch": 0.35, "learning_rate": 1.500022996569888e-05, "loss": 23.1742, "step": 28730 }, { "epoch": 0.35, "learning_rate": 1.4996780110094256e-05, "loss": 24.0631, "step": 28740 }, { "epoch": 0.35, "learning_rate": 1.4993329461724992e-05, "loss": 23.4516, "step": 28750 }, { "epoch": 0.35, "learning_rate": 1.4989878021138556e-05, "loss": 23.4058, "step": 28760 }, { "epoch": 0.35, "learning_rate": 1.4986425788882529e-05, "loss": 23.5987, "step": 28770 }, { "epoch": 0.35, "learning_rate": 1.4982972765504627e-05, "loss": 23.6779, "step": 28780 }, { "epoch": 0.35, "learning_rate": 1.4979518951552692e-05, "loss": 23.5818, "step": 28790 }, { "epoch": 0.35, "learning_rate": 1.497606434757469e-05, "loss": 23.8864, "step": 28800 }, { "epoch": 0.35, "learning_rate": 1.4972608954118708e-05, "loss": 24.2817, "step": 28810 }, { "epoch": 0.35, "learning_rate": 1.4969152771732963e-05, "loss": 23.2941, "step": 28820 }, { "epoch": 0.35, "learning_rate": 1.4965695800965798e-05, "loss": 23.8727, "step": 28830 }, { "epoch": 0.35, "learning_rate": 1.4962238042365677e-05, "loss": 23.8866, "step": 28840 }, { "epoch": 0.35, "learning_rate": 1.4958779496481193e-05, "loss": 23.9752, "step": 28850 }, { "epoch": 0.35, "learning_rate": 1.4955320163861058e-05, "loss": 23.8527, "step": 28860 }, { "epoch": 0.36, "learning_rate": 1.4951860045054118e-05, "loss": 23.831, "step": 28870 }, { "epoch": 0.36, "learning_rate": 1.4948399140609336e-05, "loss": 24.3766, "step": 28880 }, { "epoch": 0.36, "learning_rate": 1.4944937451075801e-05, "loss": 23.5657, "step": 28890 }, { "epoch": 0.36, "learning_rate": 1.494147497700273e-05, "loss": 24.5577, "step": 28900 }, { "epoch": 0.36, "learning_rate": 1.493801171893946e-05, "loss": 23.7783, "step": 28910 }, { "epoch": 0.36, "learning_rate": 1.4934547677435456e-05, "loss": 23.2588, "step": 28920 }, { "epoch": 0.36, "learning_rate": 1.4931082853040302e-05, "loss": 23.4183, "step": 28930 }, { "epoch": 0.36, "learning_rate": 1.4927617246303716e-05, "loss": 24.02, "step": 28940 }, { "epoch": 0.36, "learning_rate": 1.4924150857775532e-05, "loss": 23.6498, "step": 28950 }, { "epoch": 0.36, "learning_rate": 1.4920683688005708e-05, "loss": 23.4097, "step": 28960 }, { "epoch": 0.36, "learning_rate": 1.4917215737544333e-05, "loss": 23.6693, "step": 28970 }, { "epoch": 0.36, "learning_rate": 1.4913747006941607e-05, "loss": 23.3648, "step": 28980 }, { "epoch": 0.36, "learning_rate": 1.4910277496747872e-05, "loss": 23.9252, "step": 28990 }, { "epoch": 0.36, "learning_rate": 1.4906807207513577e-05, "loss": 23.3533, "step": 29000 }, { "epoch": 0.36, "learning_rate": 1.49033361397893e-05, "loss": 23.7426, "step": 29010 }, { "epoch": 0.36, "learning_rate": 1.4899864294125746e-05, "loss": 23.4279, "step": 29020 }, { "epoch": 0.36, "learning_rate": 1.4896391671073742e-05, "loss": 23.9265, "step": 29030 }, { "epoch": 0.36, "learning_rate": 1.4892918271184236e-05, "loss": 23.53, "step": 29040 }, { "epoch": 0.36, "learning_rate": 1.4889444095008301e-05, "loss": 24.2866, "step": 29050 }, { "epoch": 0.36, "learning_rate": 1.4885969143097132e-05, "loss": 23.9067, "step": 29060 }, { "epoch": 0.36, "learning_rate": 1.4882493416002049e-05, "loss": 24.0833, "step": 29070 }, { "epoch": 0.36, "learning_rate": 1.4879016914274492e-05, "loss": 23.953, "step": 29080 }, { "epoch": 0.36, "learning_rate": 1.4875539638466027e-05, "loss": 24.6063, "step": 29090 }, { "epoch": 0.36, "learning_rate": 1.4872061589128343e-05, "loss": 23.7177, "step": 29100 }, { "epoch": 0.36, "learning_rate": 1.4868582766813248e-05, "loss": 23.4827, "step": 29110 }, { "epoch": 0.36, "learning_rate": 1.4865103172072677e-05, "loss": 23.5222, "step": 29120 }, { "epoch": 0.36, "learning_rate": 1.4861622805458682e-05, "loss": 24.208, "step": 29130 }, { "epoch": 0.36, "learning_rate": 1.4858141667523445e-05, "loss": 23.9059, "step": 29140 }, { "epoch": 0.36, "learning_rate": 1.4854659758819261e-05, "loss": 23.6968, "step": 29150 }, { "epoch": 0.36, "learning_rate": 1.4851177079898558e-05, "loss": 23.6495, "step": 29160 }, { "epoch": 0.36, "learning_rate": 1.484769363131388e-05, "loss": 24.079, "step": 29170 }, { "epoch": 0.36, "learning_rate": 1.4844209413617891e-05, "loss": 23.7933, "step": 29180 }, { "epoch": 0.36, "learning_rate": 1.484072442736338e-05, "loss": 24.2211, "step": 29190 }, { "epoch": 0.36, "learning_rate": 1.4837238673103263e-05, "loss": 23.7783, "step": 29200 }, { "epoch": 0.36, "learning_rate": 1.4833752151390565e-05, "loss": 23.7843, "step": 29210 }, { "epoch": 0.36, "learning_rate": 1.4830264862778445e-05, "loss": 24.3485, "step": 29220 }, { "epoch": 0.36, "learning_rate": 1.4826776807820183e-05, "loss": 24.0541, "step": 29230 }, { "epoch": 0.36, "learning_rate": 1.4823287987069168e-05, "loss": 23.7531, "step": 29240 }, { "epoch": 0.36, "learning_rate": 1.4819798401078923e-05, "loss": 23.8233, "step": 29250 }, { "epoch": 0.36, "learning_rate": 1.481630805040309e-05, "loss": 23.4819, "step": 29260 }, { "epoch": 0.36, "learning_rate": 1.4812816935595431e-05, "loss": 24.038, "step": 29270 }, { "epoch": 0.36, "learning_rate": 1.4809325057209827e-05, "loss": 23.7549, "step": 29280 }, { "epoch": 0.36, "learning_rate": 1.480583241580028e-05, "loss": 23.4676, "step": 29290 }, { "epoch": 0.36, "learning_rate": 1.4802339011920925e-05, "loss": 24.3114, "step": 29300 }, { "epoch": 0.36, "learning_rate": 1.4798844846125999e-05, "loss": 23.7988, "step": 29310 }, { "epoch": 0.36, "learning_rate": 1.4795349918969871e-05, "loss": 24.0681, "step": 29320 }, { "epoch": 0.36, "learning_rate": 1.4791854231007032e-05, "loss": 23.8625, "step": 29330 }, { "epoch": 0.36, "learning_rate": 1.478835778279209e-05, "loss": 23.6853, "step": 29340 }, { "epoch": 0.36, "learning_rate": 1.4784860574879774e-05, "loss": 23.5483, "step": 29350 }, { "epoch": 0.36, "learning_rate": 1.4781362607824929e-05, "loss": 23.6249, "step": 29360 }, { "epoch": 0.36, "learning_rate": 1.4777863882182534e-05, "loss": 23.908, "step": 29370 }, { "epoch": 0.36, "learning_rate": 1.4774364398507672e-05, "loss": 23.7436, "step": 29380 }, { "epoch": 0.36, "learning_rate": 1.4770864157355558e-05, "loss": 24.4533, "step": 29390 }, { "epoch": 0.36, "learning_rate": 1.4767363159281524e-05, "loss": 23.9931, "step": 29400 }, { "epoch": 0.36, "learning_rate": 1.4763861404841018e-05, "loss": 23.9022, "step": 29410 }, { "epoch": 0.36, "learning_rate": 1.4760358894589612e-05, "loss": 24.535, "step": 29420 }, { "epoch": 0.36, "learning_rate": 1.4756855629083e-05, "loss": 23.7148, "step": 29430 }, { "epoch": 0.36, "learning_rate": 1.4753351608876987e-05, "loss": 23.6242, "step": 29440 }, { "epoch": 0.36, "learning_rate": 1.474984683452751e-05, "loss": 24.3785, "step": 29450 }, { "epoch": 0.36, "learning_rate": 1.4746341306590618e-05, "loss": 23.4658, "step": 29460 }, { "epoch": 0.36, "learning_rate": 1.4742835025622477e-05, "loss": 24.1785, "step": 29470 }, { "epoch": 0.36, "learning_rate": 1.473932799217938e-05, "loss": 24.1021, "step": 29480 }, { "epoch": 0.36, "learning_rate": 1.4735820206817733e-05, "loss": 24.2066, "step": 29490 }, { "epoch": 0.36, "learning_rate": 1.4732311670094068e-05, "loss": 23.7975, "step": 29500 }, { "epoch": 0.36, "learning_rate": 1.4728802382565028e-05, "loss": 23.399, "step": 29510 }, { "epoch": 0.36, "learning_rate": 1.4725292344787381e-05, "loss": 23.6205, "step": 29520 }, { "epoch": 0.36, "learning_rate": 1.4721781557318016e-05, "loss": 23.5321, "step": 29530 }, { "epoch": 0.36, "learning_rate": 1.4718270020713929e-05, "loss": 23.7716, "step": 29540 }, { "epoch": 0.36, "learning_rate": 1.471475773553225e-05, "loss": 24.0543, "step": 29550 }, { "epoch": 0.36, "learning_rate": 1.471124470233022e-05, "loss": 24.2498, "step": 29560 }, { "epoch": 0.36, "learning_rate": 1.47077309216652e-05, "loss": 23.4664, "step": 29570 }, { "epoch": 0.36, "learning_rate": 1.4704216394094668e-05, "loss": 23.9148, "step": 29580 }, { "epoch": 0.36, "learning_rate": 1.4700701120176217e-05, "loss": 23.8361, "step": 29590 }, { "epoch": 0.36, "learning_rate": 1.4697185100467572e-05, "loss": 23.953, "step": 29600 }, { "epoch": 0.36, "learning_rate": 1.4693668335526563e-05, "loss": 23.8932, "step": 29610 }, { "epoch": 0.36, "learning_rate": 1.4690150825911143e-05, "loss": 24.0279, "step": 29620 }, { "epoch": 0.36, "learning_rate": 1.4686632572179382e-05, "loss": 23.6783, "step": 29630 }, { "epoch": 0.36, "learning_rate": 1.468311357488947e-05, "loss": 24.0849, "step": 29640 }, { "epoch": 0.36, "learning_rate": 1.467959383459971e-05, "loss": 23.1715, "step": 29650 }, { "epoch": 0.36, "learning_rate": 1.4676073351868534e-05, "loss": 23.8159, "step": 29660 }, { "epoch": 0.36, "learning_rate": 1.467255212725448e-05, "loss": 23.4399, "step": 29670 }, { "epoch": 0.37, "learning_rate": 1.4669030161316209e-05, "loss": 23.8834, "step": 29680 }, { "epoch": 0.37, "learning_rate": 1.46655074546125e-05, "loss": 23.822, "step": 29690 }, { "epoch": 0.37, "learning_rate": 1.4661984007702248e-05, "loss": 24.2306, "step": 29700 }, { "epoch": 0.37, "learning_rate": 1.4658459821144462e-05, "loss": 23.8303, "step": 29710 }, { "epoch": 0.37, "learning_rate": 1.4654934895498276e-05, "loss": 23.4825, "step": 29720 }, { "epoch": 0.37, "learning_rate": 1.4651409231322937e-05, "loss": 24.1798, "step": 29730 }, { "epoch": 0.37, "learning_rate": 1.4647882829177811e-05, "loss": 24.0836, "step": 29740 }, { "epoch": 0.37, "learning_rate": 1.4644355689622376e-05, "loss": 23.6959, "step": 29750 }, { "epoch": 0.37, "learning_rate": 1.4640827813216232e-05, "loss": 24.0861, "step": 29760 }, { "epoch": 0.37, "learning_rate": 1.4637299200519097e-05, "loss": 24.326, "step": 29770 }, { "epoch": 0.37, "learning_rate": 1.46337698520908e-05, "loss": 23.7329, "step": 29780 }, { "epoch": 0.37, "learning_rate": 1.4630239768491295e-05, "loss": 24.4056, "step": 29790 }, { "epoch": 0.37, "learning_rate": 1.4626708950280641e-05, "loss": 23.6502, "step": 29800 }, { "epoch": 0.37, "learning_rate": 1.4623177398019026e-05, "loss": 23.5648, "step": 29810 }, { "epoch": 0.37, "learning_rate": 1.4619645112266744e-05, "loss": 24.1469, "step": 29820 }, { "epoch": 0.37, "learning_rate": 1.4616112093584213e-05, "loss": 23.6679, "step": 29830 }, { "epoch": 0.37, "learning_rate": 1.4612578342531968e-05, "loss": 23.6435, "step": 29840 }, { "epoch": 0.37, "learning_rate": 1.4609043859670648e-05, "loss": 24.0713, "step": 29850 }, { "epoch": 0.37, "learning_rate": 1.4605508645561024e-05, "loss": 23.5647, "step": 29860 }, { "epoch": 0.37, "learning_rate": 1.4601972700763973e-05, "loss": 24.2552, "step": 29870 }, { "epoch": 0.37, "learning_rate": 1.4598436025840493e-05, "loss": 24.3914, "step": 29880 }, { "epoch": 0.37, "learning_rate": 1.4594898621351687e-05, "loss": 23.4825, "step": 29890 }, { "epoch": 0.37, "learning_rate": 1.4591360487858795e-05, "loss": 24.4432, "step": 29900 }, { "epoch": 0.37, "learning_rate": 1.458782162592315e-05, "loss": 24.1001, "step": 29910 }, { "epoch": 0.37, "learning_rate": 1.4584282036106214e-05, "loss": 23.5208, "step": 29920 }, { "epoch": 0.37, "learning_rate": 1.4580741718969565e-05, "loss": 24.1555, "step": 29930 }, { "epoch": 0.37, "learning_rate": 1.4577200675074888e-05, "loss": 24.0444, "step": 29940 }, { "epoch": 0.37, "learning_rate": 1.4573658904983985e-05, "loss": 23.541, "step": 29950 }, { "epoch": 0.37, "learning_rate": 1.4570116409258778e-05, "loss": 24.1197, "step": 29960 }, { "epoch": 0.37, "learning_rate": 1.4566573188461305e-05, "loss": 23.9652, "step": 29970 }, { "epoch": 0.37, "learning_rate": 1.4563029243153712e-05, "loss": 24.2821, "step": 29980 }, { "epoch": 0.37, "learning_rate": 1.4559484573898267e-05, "loss": 24.2102, "step": 29990 }, { "epoch": 0.37, "learning_rate": 1.455593918125735e-05, "loss": 24.0564, "step": 30000 }, { "epoch": 0.37, "learning_rate": 1.4552393065793453e-05, "loss": 24.0047, "step": 30010 }, { "epoch": 0.37, "learning_rate": 1.4548846228069189e-05, "loss": 24.0773, "step": 30020 }, { "epoch": 0.37, "learning_rate": 1.4545298668647279e-05, "loss": 23.508, "step": 30030 }, { "epoch": 0.37, "learning_rate": 1.4541750388090563e-05, "loss": 23.6923, "step": 30040 }, { "epoch": 0.37, "learning_rate": 1.453820138696199e-05, "loss": 23.531, "step": 30050 }, { "epoch": 0.37, "learning_rate": 1.453465166582463e-05, "loss": 24.0647, "step": 30060 }, { "epoch": 0.37, "learning_rate": 1.4531101225241666e-05, "loss": 23.6739, "step": 30070 }, { "epoch": 0.37, "learning_rate": 1.452755006577639e-05, "loss": 23.9023, "step": 30080 }, { "epoch": 0.37, "learning_rate": 1.4523998187992215e-05, "loss": 24.1414, "step": 30090 }, { "epoch": 0.37, "learning_rate": 1.452044559245266e-05, "loss": 24.1711, "step": 30100 }, { "epoch": 0.37, "learning_rate": 1.4516892279721368e-05, "loss": 23.7357, "step": 30110 }, { "epoch": 0.37, "learning_rate": 1.4513338250362088e-05, "loss": 23.9192, "step": 30120 }, { "epoch": 0.37, "learning_rate": 1.4509783504938684e-05, "loss": 23.9282, "step": 30130 }, { "epoch": 0.37, "learning_rate": 1.4506228044015134e-05, "loss": 23.6443, "step": 30140 }, { "epoch": 0.37, "learning_rate": 1.450267186815553e-05, "loss": 23.8316, "step": 30150 }, { "epoch": 0.37, "learning_rate": 1.4499114977924081e-05, "loss": 24.3394, "step": 30160 }, { "epoch": 0.37, "learning_rate": 1.4495557373885101e-05, "loss": 23.7193, "step": 30170 }, { "epoch": 0.37, "learning_rate": 1.4491999056603023e-05, "loss": 23.428, "step": 30180 }, { "epoch": 0.37, "learning_rate": 1.4488440026642394e-05, "loss": 23.9708, "step": 30190 }, { "epoch": 0.37, "learning_rate": 1.4484880284567872e-05, "loss": 24.1246, "step": 30200 }, { "epoch": 0.37, "learning_rate": 1.4481319830944227e-05, "loss": 24.4739, "step": 30210 }, { "epoch": 0.37, "learning_rate": 1.4477758666336343e-05, "loss": 23.6003, "step": 30220 }, { "epoch": 0.37, "learning_rate": 1.4474196791309221e-05, "loss": 24.0533, "step": 30230 }, { "epoch": 0.37, "learning_rate": 1.4470634206427965e-05, "loss": 23.7333, "step": 30240 }, { "epoch": 0.37, "learning_rate": 1.44670709122578e-05, "loss": 23.4547, "step": 30250 }, { "epoch": 0.37, "learning_rate": 1.4463506909364058e-05, "loss": 23.6887, "step": 30260 }, { "epoch": 0.37, "learning_rate": 1.4459942198312195e-05, "loss": 23.5349, "step": 30270 }, { "epoch": 0.37, "learning_rate": 1.445637677966776e-05, "loss": 24.1047, "step": 30280 }, { "epoch": 0.37, "learning_rate": 1.4452810653996428e-05, "loss": 24.0644, "step": 30290 }, { "epoch": 0.37, "learning_rate": 1.4449243821863985e-05, "loss": 23.885, "step": 30300 }, { "epoch": 0.37, "learning_rate": 1.4445676283836327e-05, "loss": 23.2498, "step": 30310 }, { "epoch": 0.37, "learning_rate": 1.4442108040479461e-05, "loss": 24.7107, "step": 30320 }, { "epoch": 0.37, "learning_rate": 1.4438539092359508e-05, "loss": 23.8735, "step": 30330 }, { "epoch": 0.37, "learning_rate": 1.4434969440042696e-05, "loss": 23.6968, "step": 30340 }, { "epoch": 0.37, "learning_rate": 1.4431399084095375e-05, "loss": 24.4826, "step": 30350 }, { "epoch": 0.37, "learning_rate": 1.4427828025083994e-05, "loss": 23.2943, "step": 30360 }, { "epoch": 0.37, "learning_rate": 1.4424256263575125e-05, "loss": 24.1219, "step": 30370 }, { "epoch": 0.37, "learning_rate": 1.4420683800135443e-05, "loss": 23.5897, "step": 30380 }, { "epoch": 0.37, "learning_rate": 1.4417110635331736e-05, "loss": 23.915, "step": 30390 }, { "epoch": 0.37, "learning_rate": 1.4413536769730912e-05, "loss": 23.8553, "step": 30400 }, { "epoch": 0.37, "learning_rate": 1.4409962203899974e-05, "loss": 23.3471, "step": 30410 }, { "epoch": 0.37, "learning_rate": 1.440638693840605e-05, "loss": 24.0996, "step": 30420 }, { "epoch": 0.37, "learning_rate": 1.4402810973816372e-05, "loss": 24.468, "step": 30430 }, { "epoch": 0.37, "learning_rate": 1.439923431069829e-05, "loss": 23.8797, "step": 30440 }, { "epoch": 0.37, "learning_rate": 1.4395656949619252e-05, "loss": 23.8786, "step": 30450 }, { "epoch": 0.37, "learning_rate": 1.439207889114683e-05, "loss": 23.7173, "step": 30460 }, { "epoch": 0.37, "learning_rate": 1.4388500135848702e-05, "loss": 24.1542, "step": 30470 }, { "epoch": 0.37, "learning_rate": 1.4384920684292651e-05, "loss": 23.3555, "step": 30480 }, { "epoch": 0.37, "learning_rate": 1.4381340537046578e-05, "loss": 23.9248, "step": 30490 }, { "epoch": 0.38, "learning_rate": 1.4377759694678492e-05, "loss": 23.7473, "step": 30500 }, { "epoch": 0.38, "learning_rate": 1.4374178157756514e-05, "loss": 24.2395, "step": 30510 }, { "epoch": 0.38, "learning_rate": 1.4370595926848868e-05, "loss": 23.8899, "step": 30520 }, { "epoch": 0.38, "learning_rate": 1.4367013002523896e-05, "loss": 24.3421, "step": 30530 }, { "epoch": 0.38, "learning_rate": 1.4363429385350048e-05, "loss": 23.8892, "step": 30540 }, { "epoch": 0.38, "learning_rate": 1.4359845075895877e-05, "loss": 23.3821, "step": 30550 }, { "epoch": 0.38, "learning_rate": 1.4356260074730062e-05, "loss": 23.7767, "step": 30560 }, { "epoch": 0.38, "learning_rate": 1.4352674382421377e-05, "loss": 24.4016, "step": 30570 }, { "epoch": 0.38, "learning_rate": 1.4349087999538706e-05, "loss": 24.5154, "step": 30580 }, { "epoch": 0.38, "learning_rate": 1.4345500926651053e-05, "loss": 23.9497, "step": 30590 }, { "epoch": 0.38, "learning_rate": 1.4341913164327525e-05, "loss": 23.8404, "step": 30600 }, { "epoch": 0.38, "learning_rate": 1.4338324713137337e-05, "loss": 23.8527, "step": 30610 }, { "epoch": 0.38, "learning_rate": 1.4334735573649817e-05, "loss": 24.2351, "step": 30620 }, { "epoch": 0.38, "learning_rate": 1.4331145746434394e-05, "loss": 23.2844, "step": 30630 }, { "epoch": 0.38, "learning_rate": 1.4327555232060619e-05, "loss": 23.4794, "step": 30640 }, { "epoch": 0.38, "learning_rate": 1.4323964031098142e-05, "loss": 23.1774, "step": 30650 }, { "epoch": 0.38, "learning_rate": 1.4320372144116728e-05, "loss": 23.2991, "step": 30660 }, { "epoch": 0.38, "learning_rate": 1.4316779571686249e-05, "loss": 24.1355, "step": 30670 }, { "epoch": 0.38, "learning_rate": 1.4313186314376679e-05, "loss": 23.928, "step": 30680 }, { "epoch": 0.38, "learning_rate": 1.4309592372758111e-05, "loss": 23.9163, "step": 30690 }, { "epoch": 0.38, "learning_rate": 1.4305997747400744e-05, "loss": 23.7971, "step": 30700 }, { "epoch": 0.38, "learning_rate": 1.4302402438874881e-05, "loss": 23.1898, "step": 30710 }, { "epoch": 0.38, "learning_rate": 1.4298806447750932e-05, "loss": 23.6379, "step": 30720 }, { "epoch": 0.38, "learning_rate": 1.4295209774599427e-05, "loss": 23.8561, "step": 30730 }, { "epoch": 0.38, "learning_rate": 1.4291612419990994e-05, "loss": 23.7024, "step": 30740 }, { "epoch": 0.38, "learning_rate": 1.4288014384496369e-05, "loss": 23.8107, "step": 30750 }, { "epoch": 0.38, "learning_rate": 1.42844156686864e-05, "loss": 24.0213, "step": 30760 }, { "epoch": 0.38, "learning_rate": 1.4280816273132046e-05, "loss": 23.4874, "step": 30770 }, { "epoch": 0.38, "learning_rate": 1.427721619840436e-05, "loss": 23.8349, "step": 30780 }, { "epoch": 0.38, "learning_rate": 1.427361544507452e-05, "loss": 23.6909, "step": 30790 }, { "epoch": 0.38, "learning_rate": 1.4270014013713803e-05, "loss": 23.1758, "step": 30800 }, { "epoch": 0.38, "learning_rate": 1.4266411904893591e-05, "loss": 23.6113, "step": 30810 }, { "epoch": 0.38, "learning_rate": 1.4262809119185377e-05, "loss": 23.8631, "step": 30820 }, { "epoch": 0.38, "learning_rate": 1.4259205657160769e-05, "loss": 23.9771, "step": 30830 }, { "epoch": 0.38, "learning_rate": 1.4255601519391463e-05, "loss": 23.068, "step": 30840 }, { "epoch": 0.38, "learning_rate": 1.4251996706449282e-05, "loss": 23.1079, "step": 30850 }, { "epoch": 0.38, "learning_rate": 1.4248391218906146e-05, "loss": 23.7625, "step": 30860 }, { "epoch": 0.38, "learning_rate": 1.4244785057334085e-05, "loss": 23.812, "step": 30870 }, { "epoch": 0.38, "learning_rate": 1.4241178222305232e-05, "loss": 24.0632, "step": 30880 }, { "epoch": 0.38, "learning_rate": 1.423757071439183e-05, "loss": 23.9563, "step": 30890 }, { "epoch": 0.38, "learning_rate": 1.4233962534166233e-05, "loss": 24.8522, "step": 30900 }, { "epoch": 0.38, "learning_rate": 1.4230353682200893e-05, "loss": 24.204, "step": 30910 }, { "epoch": 0.38, "learning_rate": 1.422674415906837e-05, "loss": 24.0838, "step": 30920 }, { "epoch": 0.38, "learning_rate": 1.4223133965341342e-05, "loss": 24.2259, "step": 30930 }, { "epoch": 0.38, "learning_rate": 1.4219523101592576e-05, "loss": 23.4749, "step": 30940 }, { "epoch": 0.38, "learning_rate": 1.421591156839496e-05, "loss": 24.4819, "step": 30950 }, { "epoch": 0.38, "learning_rate": 1.4212299366321479e-05, "loss": 24.3253, "step": 30960 }, { "epoch": 0.38, "learning_rate": 1.420868649594523e-05, "loss": 23.541, "step": 30970 }, { "epoch": 0.38, "learning_rate": 1.420507295783941e-05, "loss": 24.1132, "step": 30980 }, { "epoch": 0.38, "learning_rate": 1.4201458752577325e-05, "loss": 23.6533, "step": 30990 }, { "epoch": 0.38, "learning_rate": 1.4197843880732391e-05, "loss": 23.4491, "step": 31000 }, { "epoch": 0.38, "learning_rate": 1.4194228342878123e-05, "loss": 24.1759, "step": 31010 }, { "epoch": 0.38, "learning_rate": 1.4190612139588142e-05, "loss": 24.165, "step": 31020 }, { "epoch": 0.38, "learning_rate": 1.4186995271436183e-05, "loss": 24.0169, "step": 31030 }, { "epoch": 0.38, "learning_rate": 1.4183377738996078e-05, "loss": 23.7825, "step": 31040 }, { "epoch": 0.38, "learning_rate": 1.4179759542841765e-05, "loss": 23.5453, "step": 31050 }, { "epoch": 0.38, "learning_rate": 1.417614068354729e-05, "loss": 23.242, "step": 31060 }, { "epoch": 0.38, "learning_rate": 1.417252116168681e-05, "loss": 24.4437, "step": 31070 }, { "epoch": 0.38, "learning_rate": 1.4168900977834573e-05, "loss": 24.0534, "step": 31080 }, { "epoch": 0.38, "learning_rate": 1.4165280132564939e-05, "loss": 24.2889, "step": 31090 }, { "epoch": 0.38, "learning_rate": 1.4161658626452377e-05, "loss": 24.1744, "step": 31100 }, { "epoch": 0.38, "learning_rate": 1.4158036460071455e-05, "loss": 23.6503, "step": 31110 }, { "epoch": 0.38, "learning_rate": 1.4154413633996853e-05, "loss": 24.0831, "step": 31120 }, { "epoch": 0.38, "learning_rate": 1.4150790148803344e-05, "loss": 23.9102, "step": 31130 }, { "epoch": 0.38, "learning_rate": 1.414716600506582e-05, "loss": 23.7211, "step": 31140 }, { "epoch": 0.38, "learning_rate": 1.4143541203359263e-05, "loss": 24.1869, "step": 31150 }, { "epoch": 0.38, "learning_rate": 1.413991574425877e-05, "loss": 23.6988, "step": 31160 }, { "epoch": 0.38, "learning_rate": 1.4136289628339539e-05, "loss": 23.6081, "step": 31170 }, { "epoch": 0.38, "learning_rate": 1.4132662856176868e-05, "loss": 23.7382, "step": 31180 }, { "epoch": 0.38, "learning_rate": 1.4129035428346168e-05, "loss": 24.0129, "step": 31190 }, { "epoch": 0.38, "learning_rate": 1.4125407345422946e-05, "loss": 23.8698, "step": 31200 }, { "epoch": 0.38, "learning_rate": 1.4121778607982817e-05, "loss": 23.7254, "step": 31210 }, { "epoch": 0.38, "learning_rate": 1.4118149216601495e-05, "loss": 23.2607, "step": 31220 }, { "epoch": 0.38, "learning_rate": 1.4114519171854806e-05, "loss": 23.9116, "step": 31230 }, { "epoch": 0.38, "learning_rate": 1.4110888474318673e-05, "loss": 23.3074, "step": 31240 }, { "epoch": 0.38, "learning_rate": 1.4107257124569124e-05, "loss": 23.8839, "step": 31250 }, { "epoch": 0.38, "learning_rate": 1.4103625123182294e-05, "loss": 24.3718, "step": 31260 }, { "epoch": 0.38, "learning_rate": 1.4099992470734415e-05, "loss": 22.981, "step": 31270 }, { "epoch": 0.38, "learning_rate": 1.409635916780183e-05, "loss": 23.6086, "step": 31280 }, { "epoch": 0.38, "learning_rate": 1.4092725214960977e-05, "loss": 23.7248, "step": 31290 }, { "epoch": 0.38, "learning_rate": 1.4089090612788403e-05, "loss": 24.1787, "step": 31300 }, { "epoch": 0.39, "learning_rate": 1.4085455361860755e-05, "loss": 23.7664, "step": 31310 }, { "epoch": 0.39, "learning_rate": 1.4081819462754789e-05, "loss": 24.0835, "step": 31320 }, { "epoch": 0.39, "learning_rate": 1.407818291604735e-05, "loss": 23.7589, "step": 31330 }, { "epoch": 0.39, "learning_rate": 1.4074545722315403e-05, "loss": 23.6295, "step": 31340 }, { "epoch": 0.39, "learning_rate": 1.4070907882136002e-05, "loss": 24.2018, "step": 31350 }, { "epoch": 0.39, "learning_rate": 1.4067269396086311e-05, "loss": 23.5197, "step": 31360 }, { "epoch": 0.39, "learning_rate": 1.4063630264743596e-05, "loss": 24.1694, "step": 31370 }, { "epoch": 0.39, "learning_rate": 1.405999048868522e-05, "loss": 23.9195, "step": 31380 }, { "epoch": 0.39, "learning_rate": 1.4056350068488653e-05, "loss": 23.63, "step": 31390 }, { "epoch": 0.39, "learning_rate": 1.405270900473147e-05, "loss": 24.5599, "step": 31400 }, { "epoch": 0.39, "learning_rate": 1.4049067297991341e-05, "loss": 24.0647, "step": 31410 }, { "epoch": 0.39, "learning_rate": 1.4045424948846043e-05, "loss": 23.7385, "step": 31420 }, { "epoch": 0.39, "learning_rate": 1.404178195787345e-05, "loss": 23.9749, "step": 31430 }, { "epoch": 0.39, "learning_rate": 1.4038138325651545e-05, "loss": 23.6896, "step": 31440 }, { "epoch": 0.39, "learning_rate": 1.4034494052758405e-05, "loss": 23.8571, "step": 31450 }, { "epoch": 0.39, "learning_rate": 1.4030849139772215e-05, "loss": 24.0085, "step": 31460 }, { "epoch": 0.39, "learning_rate": 1.4027203587271261e-05, "loss": 24.4037, "step": 31470 }, { "epoch": 0.39, "learning_rate": 1.4023557395833924e-05, "loss": 24.3131, "step": 31480 }, { "epoch": 0.39, "learning_rate": 1.4019910566038693e-05, "loss": 23.4487, "step": 31490 }, { "epoch": 0.39, "learning_rate": 1.4016263098464155e-05, "loss": 24.3777, "step": 31500 }, { "epoch": 0.39, "learning_rate": 1.4012614993689003e-05, "loss": 23.9683, "step": 31510 }, { "epoch": 0.39, "learning_rate": 1.4008966252292022e-05, "loss": 23.9197, "step": 31520 }, { "epoch": 0.39, "learning_rate": 1.4005316874852108e-05, "loss": 23.8569, "step": 31530 }, { "epoch": 0.39, "learning_rate": 1.4001666861948251e-05, "loss": 23.9764, "step": 31540 }, { "epoch": 0.39, "learning_rate": 1.3998016214159547e-05, "loss": 23.8307, "step": 31550 }, { "epoch": 0.39, "learning_rate": 1.3994364932065185e-05, "loss": 23.9289, "step": 31560 }, { "epoch": 0.39, "learning_rate": 1.3990713016244464e-05, "loss": 24.1007, "step": 31570 }, { "epoch": 0.39, "learning_rate": 1.3987060467276773e-05, "loss": 24.0507, "step": 31580 }, { "epoch": 0.39, "learning_rate": 1.3983407285741614e-05, "loss": 23.5046, "step": 31590 }, { "epoch": 0.39, "learning_rate": 1.3979753472218582e-05, "loss": 23.5907, "step": 31600 }, { "epoch": 0.39, "learning_rate": 1.3976099027287374e-05, "loss": 23.2878, "step": 31610 }, { "epoch": 0.39, "learning_rate": 1.397244395152778e-05, "loss": 23.7275, "step": 31620 }, { "epoch": 0.39, "learning_rate": 1.39687882455197e-05, "loss": 24.0757, "step": 31630 }, { "epoch": 0.39, "learning_rate": 1.3965131909843133e-05, "loss": 23.8852, "step": 31640 }, { "epoch": 0.39, "learning_rate": 1.396147494507817e-05, "loss": 24.2196, "step": 31650 }, { "epoch": 0.39, "learning_rate": 1.3957817351805015e-05, "loss": 23.8662, "step": 31660 }, { "epoch": 0.39, "learning_rate": 1.395415913060396e-05, "loss": 24.0546, "step": 31670 }, { "epoch": 0.39, "learning_rate": 1.3950500282055395e-05, "loss": 23.6107, "step": 31680 }, { "epoch": 0.39, "learning_rate": 1.3946840806739818e-05, "loss": 23.7284, "step": 31690 }, { "epoch": 0.39, "learning_rate": 1.394318070523783e-05, "loss": 24.2571, "step": 31700 }, { "epoch": 0.39, "learning_rate": 1.3939519978130118e-05, "loss": 23.9024, "step": 31710 }, { "epoch": 0.39, "learning_rate": 1.3935858625997474e-05, "loss": 24.0026, "step": 31720 }, { "epoch": 0.39, "learning_rate": 1.3932196649420797e-05, "loss": 24.1285, "step": 31730 }, { "epoch": 0.39, "learning_rate": 1.3928534048981071e-05, "loss": 23.8828, "step": 31740 }, { "epoch": 0.39, "learning_rate": 1.3924870825259392e-05, "loss": 23.3353, "step": 31750 }, { "epoch": 0.39, "learning_rate": 1.3921206978836943e-05, "loss": 23.7892, "step": 31760 }, { "epoch": 0.39, "learning_rate": 1.391754251029502e-05, "loss": 23.3586, "step": 31770 }, { "epoch": 0.39, "learning_rate": 1.3913877420215001e-05, "loss": 23.883, "step": 31780 }, { "epoch": 0.39, "learning_rate": 1.3910211709178377e-05, "loss": 23.6622, "step": 31790 }, { "epoch": 0.39, "learning_rate": 1.3906545377766731e-05, "loss": 23.3811, "step": 31800 }, { "epoch": 0.39, "learning_rate": 1.3902878426561742e-05, "loss": 23.8636, "step": 31810 }, { "epoch": 0.39, "learning_rate": 1.3899210856145194e-05, "loss": 23.4877, "step": 31820 }, { "epoch": 0.39, "learning_rate": 1.389554266709896e-05, "loss": 23.7333, "step": 31830 }, { "epoch": 0.39, "learning_rate": 1.3891873860005028e-05, "loss": 23.8008, "step": 31840 }, { "epoch": 0.39, "learning_rate": 1.3888204435445462e-05, "loss": 23.688, "step": 31850 }, { "epoch": 0.39, "learning_rate": 1.3884534394002441e-05, "loss": 24.6374, "step": 31860 }, { "epoch": 0.39, "learning_rate": 1.3880863736258234e-05, "loss": 23.5553, "step": 31870 }, { "epoch": 0.39, "learning_rate": 1.3877192462795209e-05, "loss": 23.9452, "step": 31880 }, { "epoch": 0.39, "learning_rate": 1.3873520574195834e-05, "loss": 23.4602, "step": 31890 }, { "epoch": 0.39, "learning_rate": 1.3869848071042671e-05, "loss": 23.8408, "step": 31900 }, { "epoch": 0.39, "learning_rate": 1.3866174953918381e-05, "loss": 23.8639, "step": 31910 }, { "epoch": 0.39, "learning_rate": 1.3862501223405722e-05, "loss": 24.0715, "step": 31920 }, { "epoch": 0.39, "learning_rate": 1.3858826880087556e-05, "loss": 23.643, "step": 31930 }, { "epoch": 0.39, "learning_rate": 1.3855151924546828e-05, "loss": 23.5113, "step": 31940 }, { "epoch": 0.39, "learning_rate": 1.3851476357366596e-05, "loss": 23.408, "step": 31950 }, { "epoch": 0.39, "learning_rate": 1.3847800179130002e-05, "loss": 23.8569, "step": 31960 }, { "epoch": 0.39, "learning_rate": 1.3844123390420294e-05, "loss": 23.8326, "step": 31970 }, { "epoch": 0.39, "learning_rate": 1.3840445991820807e-05, "loss": 24.0869, "step": 31980 }, { "epoch": 0.39, "learning_rate": 1.3836767983914989e-05, "loss": 23.8733, "step": 31990 }, { "epoch": 0.39, "learning_rate": 1.3833089367286369e-05, "loss": 23.9834, "step": 32000 }, { "epoch": 0.39, "learning_rate": 1.3829410142518578e-05, "loss": 23.8429, "step": 32010 }, { "epoch": 0.39, "learning_rate": 1.3825730310195345e-05, "loss": 23.8167, "step": 32020 }, { "epoch": 0.39, "learning_rate": 1.382204987090049e-05, "loss": 23.879, "step": 32030 }, { "epoch": 0.39, "learning_rate": 1.3818368825217943e-05, "loss": 23.5651, "step": 32040 }, { "epoch": 0.39, "learning_rate": 1.3814687173731709e-05, "loss": 24.1577, "step": 32050 }, { "epoch": 0.39, "learning_rate": 1.3811004917025907e-05, "loss": 23.8266, "step": 32060 }, { "epoch": 0.39, "learning_rate": 1.3807322055684752e-05, "loss": 23.382, "step": 32070 }, { "epoch": 0.39, "learning_rate": 1.3803638590292536e-05, "loss": 23.9252, "step": 32080 }, { "epoch": 0.39, "learning_rate": 1.3799954521433664e-05, "loss": 24.0077, "step": 32090 }, { "epoch": 0.39, "learning_rate": 1.3796269849692641e-05, "loss": 23.5906, "step": 32100 }, { "epoch": 0.39, "learning_rate": 1.3792584575654047e-05, "loss": 24.0104, "step": 32110 }, { "epoch": 0.4, "learning_rate": 1.3788898699902573e-05, "loss": 24.3018, "step": 32120 }, { "epoch": 0.4, "learning_rate": 1.3785212223023009e-05, "loss": 23.9121, "step": 32130 }, { "epoch": 0.4, "learning_rate": 1.3781525145600225e-05, "loss": 23.6727, "step": 32140 }, { "epoch": 0.4, "learning_rate": 1.3777837468219195e-05, "loss": 24.0409, "step": 32150 }, { "epoch": 0.4, "learning_rate": 1.377414919146499e-05, "loss": 23.9508, "step": 32160 }, { "epoch": 0.4, "learning_rate": 1.377046031592278e-05, "loss": 23.7742, "step": 32170 }, { "epoch": 0.4, "learning_rate": 1.3766770842177812e-05, "loss": 23.8608, "step": 32180 }, { "epoch": 0.4, "learning_rate": 1.376308077081545e-05, "loss": 23.3589, "step": 32190 }, { "epoch": 0.4, "learning_rate": 1.3759390102421137e-05, "loss": 24.0268, "step": 32200 }, { "epoch": 0.4, "learning_rate": 1.3755698837580416e-05, "loss": 24.5734, "step": 32210 }, { "epoch": 0.4, "learning_rate": 1.3752006976878927e-05, "loss": 24.2808, "step": 32220 }, { "epoch": 0.4, "learning_rate": 1.3748314520902405e-05, "loss": 24.2034, "step": 32230 }, { "epoch": 0.4, "learning_rate": 1.3744621470236674e-05, "loss": 23.98, "step": 32240 }, { "epoch": 0.4, "learning_rate": 1.3740927825467654e-05, "loss": 24.0864, "step": 32250 }, { "epoch": 0.4, "learning_rate": 1.3737233587181362e-05, "loss": 23.5334, "step": 32260 }, { "epoch": 0.4, "learning_rate": 1.3733538755963909e-05, "loss": 23.8736, "step": 32270 }, { "epoch": 0.4, "learning_rate": 1.3729843332401498e-05, "loss": 23.8688, "step": 32280 }, { "epoch": 0.4, "learning_rate": 1.3726147317080425e-05, "loss": 24.3729, "step": 32290 }, { "epoch": 0.4, "learning_rate": 1.3722450710587086e-05, "loss": 23.9936, "step": 32300 }, { "epoch": 0.4, "learning_rate": 1.3718753513507959e-05, "loss": 23.8908, "step": 32310 }, { "epoch": 0.4, "learning_rate": 1.3715055726429629e-05, "loss": 23.7532, "step": 32320 }, { "epoch": 0.4, "learning_rate": 1.3711357349938772e-05, "loss": 23.4832, "step": 32330 }, { "epoch": 0.4, "learning_rate": 1.3707658384622146e-05, "loss": 24.0193, "step": 32340 }, { "epoch": 0.4, "learning_rate": 1.3703958831066615e-05, "loss": 23.7673, "step": 32350 }, { "epoch": 0.4, "learning_rate": 1.3700258689859133e-05, "loss": 23.8223, "step": 32360 }, { "epoch": 0.4, "learning_rate": 1.3696557961586745e-05, "loss": 24.2802, "step": 32370 }, { "epoch": 0.4, "learning_rate": 1.369285664683659e-05, "loss": 24.8658, "step": 32380 }, { "epoch": 0.4, "learning_rate": 1.3689154746195898e-05, "loss": 23.7799, "step": 32390 }, { "epoch": 0.4, "learning_rate": 1.3685452260252001e-05, "loss": 23.5825, "step": 32400 }, { "epoch": 0.4, "learning_rate": 1.3681749189592313e-05, "loss": 24.1974, "step": 32410 }, { "epoch": 0.4, "learning_rate": 1.3678045534804343e-05, "loss": 24.1192, "step": 32420 }, { "epoch": 0.4, "learning_rate": 1.36743412964757e-05, "loss": 23.3248, "step": 32430 }, { "epoch": 0.4, "learning_rate": 1.3670636475194075e-05, "loss": 23.5161, "step": 32440 }, { "epoch": 0.4, "learning_rate": 1.366693107154726e-05, "loss": 23.7259, "step": 32450 }, { "epoch": 0.4, "learning_rate": 1.3663225086123135e-05, "loss": 24.2401, "step": 32460 }, { "epoch": 0.4, "learning_rate": 1.3659518519509679e-05, "loss": 23.9048, "step": 32470 }, { "epoch": 0.4, "learning_rate": 1.3655811372294947e-05, "loss": 24.0693, "step": 32480 }, { "epoch": 0.4, "learning_rate": 1.3652103645067103e-05, "loss": 23.7237, "step": 32490 }, { "epoch": 0.4, "learning_rate": 1.3648395338414397e-05, "loss": 23.1716, "step": 32500 }, { "epoch": 0.4, "learning_rate": 1.364468645292517e-05, "loss": 24.0308, "step": 32510 }, { "epoch": 0.4, "learning_rate": 1.3640976989187856e-05, "loss": 23.9426, "step": 32520 }, { "epoch": 0.4, "learning_rate": 1.3637266947790977e-05, "loss": 23.888, "step": 32530 }, { "epoch": 0.4, "learning_rate": 1.3633556329323158e-05, "loss": 24.1412, "step": 32540 }, { "epoch": 0.4, "learning_rate": 1.36298451343731e-05, "loss": 23.5922, "step": 32550 }, { "epoch": 0.4, "learning_rate": 1.3626133363529604e-05, "loss": 23.8469, "step": 32560 }, { "epoch": 0.4, "learning_rate": 1.3622421017381566e-05, "loss": 23.9033, "step": 32570 }, { "epoch": 0.4, "learning_rate": 1.3618708096517963e-05, "loss": 23.6406, "step": 32580 }, { "epoch": 0.4, "learning_rate": 1.3614994601527874e-05, "loss": 23.9938, "step": 32590 }, { "epoch": 0.4, "learning_rate": 1.3611280533000459e-05, "loss": 23.7455, "step": 32600 }, { "epoch": 0.4, "learning_rate": 1.3607565891524977e-05, "loss": 24.417, "step": 32610 }, { "epoch": 0.4, "learning_rate": 1.3603850677690771e-05, "loss": 23.405, "step": 32620 }, { "epoch": 0.4, "learning_rate": 1.3600134892087284e-05, "loss": 23.6104, "step": 32630 }, { "epoch": 0.4, "learning_rate": 1.3596418535304043e-05, "loss": 23.9949, "step": 32640 }, { "epoch": 0.4, "learning_rate": 1.3592701607930665e-05, "loss": 23.3253, "step": 32650 }, { "epoch": 0.4, "learning_rate": 1.3588984110556857e-05, "loss": 23.5233, "step": 32660 }, { "epoch": 0.4, "learning_rate": 1.3585266043772427e-05, "loss": 23.9219, "step": 32670 }, { "epoch": 0.4, "learning_rate": 1.358154740816726e-05, "loss": 24.1313, "step": 32680 }, { "epoch": 0.4, "learning_rate": 1.3577828204331334e-05, "loss": 23.559, "step": 32690 }, { "epoch": 0.4, "learning_rate": 1.3574108432854725e-05, "loss": 23.4138, "step": 32700 }, { "epoch": 0.4, "learning_rate": 1.3570388094327592e-05, "loss": 23.6267, "step": 32710 }, { "epoch": 0.4, "learning_rate": 1.3566667189340183e-05, "loss": 23.9767, "step": 32720 }, { "epoch": 0.4, "learning_rate": 1.3562945718482841e-05, "loss": 24.4115, "step": 32730 }, { "epoch": 0.4, "learning_rate": 1.3559223682345997e-05, "loss": 23.9332, "step": 32740 }, { "epoch": 0.4, "learning_rate": 1.3555501081520169e-05, "loss": 23.7079, "step": 32750 }, { "epoch": 0.4, "learning_rate": 1.3551777916595965e-05, "loss": 24.1991, "step": 32760 }, { "epoch": 0.4, "learning_rate": 1.354805418816409e-05, "loss": 24.2673, "step": 32770 }, { "epoch": 0.4, "learning_rate": 1.3544329896815328e-05, "loss": 23.8151, "step": 32780 }, { "epoch": 0.4, "learning_rate": 1.3540605043140556e-05, "loss": 24.0374, "step": 32790 }, { "epoch": 0.4, "learning_rate": 1.3536879627730745e-05, "loss": 23.9509, "step": 32800 }, { "epoch": 0.4, "learning_rate": 1.3533153651176949e-05, "loss": 24.1726, "step": 32810 }, { "epoch": 0.4, "learning_rate": 1.3529427114070315e-05, "loss": 23.8209, "step": 32820 }, { "epoch": 0.4, "learning_rate": 1.3525700017002072e-05, "loss": 24.0849, "step": 32830 }, { "epoch": 0.4, "learning_rate": 1.3521972360563549e-05, "loss": 24.2881, "step": 32840 }, { "epoch": 0.4, "learning_rate": 1.3518244145346154e-05, "loss": 23.4747, "step": 32850 }, { "epoch": 0.4, "learning_rate": 1.3514515371941386e-05, "loss": 24.001, "step": 32860 }, { "epoch": 0.4, "learning_rate": 1.3510786040940839e-05, "loss": 23.9378, "step": 32870 }, { "epoch": 0.4, "learning_rate": 1.3507056152936189e-05, "loss": 23.9809, "step": 32880 }, { "epoch": 0.4, "learning_rate": 1.3503325708519194e-05, "loss": 24.0672, "step": 32890 }, { "epoch": 0.4, "learning_rate": 1.3499594708281722e-05, "loss": 23.6519, "step": 32900 }, { "epoch": 0.4, "learning_rate": 1.3495863152815702e-05, "loss": 23.7637, "step": 32910 }, { "epoch": 0.4, "learning_rate": 1.3492131042713172e-05, "loss": 23.6199, "step": 32920 }, { "epoch": 0.4, "learning_rate": 1.348839837856625e-05, "loss": 24.128, "step": 32930 }, { "epoch": 0.41, "learning_rate": 1.348466516096714e-05, "loss": 23.469, "step": 32940 }, { "epoch": 0.41, "learning_rate": 1.3480931390508136e-05, "loss": 23.5818, "step": 32950 }, { "epoch": 0.41, "learning_rate": 1.3477197067781618e-05, "loss": 23.6623, "step": 32960 }, { "epoch": 0.41, "learning_rate": 1.347346219338006e-05, "loss": 23.6947, "step": 32970 }, { "epoch": 0.41, "learning_rate": 1.3469726767896015e-05, "loss": 23.2983, "step": 32980 }, { "epoch": 0.41, "learning_rate": 1.3465990791922128e-05, "loss": 23.7277, "step": 32990 }, { "epoch": 0.41, "learning_rate": 1.3462254266051133e-05, "loss": 24.1583, "step": 33000 }, { "epoch": 0.41, "learning_rate": 1.3458517190875845e-05, "loss": 24.0372, "step": 33010 }, { "epoch": 0.41, "learning_rate": 1.3454779566989174e-05, "loss": 23.9435, "step": 33020 }, { "epoch": 0.41, "learning_rate": 1.3451041394984112e-05, "loss": 23.2785, "step": 33030 }, { "epoch": 0.41, "learning_rate": 1.3447302675453737e-05, "loss": 23.9835, "step": 33040 }, { "epoch": 0.41, "learning_rate": 1.3443563408991219e-05, "loss": 24.5264, "step": 33050 }, { "epoch": 0.41, "learning_rate": 1.3439823596189807e-05, "loss": 23.63, "step": 33060 }, { "epoch": 0.41, "learning_rate": 1.3436083237642846e-05, "loss": 23.9556, "step": 33070 }, { "epoch": 0.41, "learning_rate": 1.343234233394376e-05, "loss": 24.2557, "step": 33080 }, { "epoch": 0.41, "learning_rate": 1.3428600885686063e-05, "loss": 23.0483, "step": 33090 }, { "epoch": 0.41, "learning_rate": 1.3424858893463359e-05, "loss": 24.0738, "step": 33100 }, { "epoch": 0.41, "learning_rate": 1.3421116357869328e-05, "loss": 24.2885, "step": 33110 }, { "epoch": 0.41, "learning_rate": 1.3417373279497744e-05, "loss": 23.5104, "step": 33120 }, { "epoch": 0.41, "learning_rate": 1.341362965894247e-05, "loss": 23.8729, "step": 33130 }, { "epoch": 0.41, "learning_rate": 1.3409885496797443e-05, "loss": 24.3019, "step": 33140 }, { "epoch": 0.41, "learning_rate": 1.3406140793656702e-05, "loss": 23.3954, "step": 33150 }, { "epoch": 0.41, "learning_rate": 1.3402395550114354e-05, "loss": 24.0195, "step": 33160 }, { "epoch": 0.41, "learning_rate": 1.339864976676461e-05, "loss": 23.9973, "step": 33170 }, { "epoch": 0.41, "learning_rate": 1.339490344420175e-05, "loss": 24.0237, "step": 33180 }, { "epoch": 0.41, "learning_rate": 1.3391156583020151e-05, "loss": 24.3178, "step": 33190 }, { "epoch": 0.41, "learning_rate": 1.3387409183814272e-05, "loss": 24.3394, "step": 33200 }, { "epoch": 0.41, "learning_rate": 1.3383661247178656e-05, "loss": 24.1386, "step": 33210 }, { "epoch": 0.41, "learning_rate": 1.337991277370793e-05, "loss": 23.8826, "step": 33220 }, { "epoch": 0.41, "learning_rate": 1.3376163763996813e-05, "loss": 23.668, "step": 33230 }, { "epoch": 0.41, "learning_rate": 1.3372414218640103e-05, "loss": 23.727, "step": 33240 }, { "epoch": 0.41, "learning_rate": 1.3368664138232681e-05, "loss": 24.0055, "step": 33250 }, { "epoch": 0.41, "learning_rate": 1.3364913523369519e-05, "loss": 23.7719, "step": 33260 }, { "epoch": 0.41, "learning_rate": 1.3361162374645674e-05, "loss": 24.333, "step": 33270 }, { "epoch": 0.41, "learning_rate": 1.3357410692656282e-05, "loss": 23.9431, "step": 33280 }, { "epoch": 0.41, "learning_rate": 1.3353658477996565e-05, "loss": 23.9368, "step": 33290 }, { "epoch": 0.41, "learning_rate": 1.3349905731261834e-05, "loss": 24.5842, "step": 33300 }, { "epoch": 0.41, "learning_rate": 1.334615245304748e-05, "loss": 23.5357, "step": 33310 }, { "epoch": 0.41, "learning_rate": 1.3342398643948977e-05, "loss": 23.9788, "step": 33320 }, { "epoch": 0.41, "learning_rate": 1.3338644304561887e-05, "loss": 24.1169, "step": 33330 }, { "epoch": 0.41, "learning_rate": 1.3334889435481864e-05, "loss": 24.2878, "step": 33340 }, { "epoch": 0.41, "learning_rate": 1.3331134037304625e-05, "loss": 24.3204, "step": 33350 }, { "epoch": 0.41, "learning_rate": 1.3327378110625988e-05, "loss": 23.6996, "step": 33360 }, { "epoch": 0.41, "learning_rate": 1.3323621656041852e-05, "loss": 23.6137, "step": 33370 }, { "epoch": 0.41, "learning_rate": 1.3319864674148192e-05, "loss": 23.7472, "step": 33380 }, { "epoch": 0.41, "learning_rate": 1.3316107165541077e-05, "loss": 24.1022, "step": 33390 }, { "epoch": 0.41, "learning_rate": 1.3312349130816658e-05, "loss": 23.7155, "step": 33400 }, { "epoch": 0.41, "learning_rate": 1.3308590570571159e-05, "loss": 24.0237, "step": 33410 }, { "epoch": 0.41, "learning_rate": 1.3304831485400897e-05, "loss": 23.6783, "step": 33420 }, { "epoch": 0.41, "learning_rate": 1.3301071875902272e-05, "loss": 23.7481, "step": 33430 }, { "epoch": 0.41, "learning_rate": 1.3297311742671763e-05, "loss": 22.7065, "step": 33440 }, { "epoch": 0.41, "learning_rate": 1.3293551086305935e-05, "loss": 23.5886, "step": 33450 }, { "epoch": 0.41, "learning_rate": 1.3289789907401439e-05, "loss": 23.8557, "step": 33460 }, { "epoch": 0.41, "learning_rate": 1.3286028206555e-05, "loss": 23.6414, "step": 33470 }, { "epoch": 0.41, "learning_rate": 1.3282265984363431e-05, "loss": 24.2118, "step": 33480 }, { "epoch": 0.41, "learning_rate": 1.3278503241423631e-05, "loss": 23.2615, "step": 33490 }, { "epoch": 0.41, "learning_rate": 1.3274739978332578e-05, "loss": 23.3962, "step": 33500 }, { "epoch": 0.41, "learning_rate": 1.3270976195687333e-05, "loss": 24.2668, "step": 33510 }, { "epoch": 0.41, "learning_rate": 1.3267211894085037e-05, "loss": 23.9298, "step": 33520 }, { "epoch": 0.41, "learning_rate": 1.3263447074122916e-05, "loss": 23.6306, "step": 33530 }, { "epoch": 0.41, "learning_rate": 1.325968173639828e-05, "loss": 23.9664, "step": 33540 }, { "epoch": 0.41, "learning_rate": 1.3255915881508514e-05, "loss": 23.4275, "step": 33550 }, { "epoch": 0.41, "learning_rate": 1.3252149510051098e-05, "loss": 23.8893, "step": 33560 }, { "epoch": 0.41, "learning_rate": 1.3248382622623581e-05, "loss": 23.7736, "step": 33570 }, { "epoch": 0.41, "learning_rate": 1.3244615219823598e-05, "loss": 23.5349, "step": 33580 }, { "epoch": 0.41, "learning_rate": 1.324084730224887e-05, "loss": 23.8436, "step": 33590 }, { "epoch": 0.41, "learning_rate": 1.3237078870497195e-05, "loss": 23.2913, "step": 33600 }, { "epoch": 0.41, "learning_rate": 1.3233309925166453e-05, "loss": 24.0307, "step": 33610 }, { "epoch": 0.41, "learning_rate": 1.3229540466854606e-05, "loss": 23.6892, "step": 33620 }, { "epoch": 0.41, "learning_rate": 1.32257704961597e-05, "loss": 24.4017, "step": 33630 }, { "epoch": 0.41, "learning_rate": 1.3222000013679861e-05, "loss": 23.5018, "step": 33640 }, { "epoch": 0.41, "learning_rate": 1.3218229020013289e-05, "loss": 23.2711, "step": 33650 }, { "epoch": 0.41, "learning_rate": 1.321445751575828e-05, "loss": 24.0223, "step": 33660 }, { "epoch": 0.41, "learning_rate": 1.3210685501513196e-05, "loss": 24.0611, "step": 33670 }, { "epoch": 0.41, "learning_rate": 1.3206912977876485e-05, "loss": 23.4684, "step": 33680 }, { "epoch": 0.41, "learning_rate": 1.3203139945446684e-05, "loss": 24.0731, "step": 33690 }, { "epoch": 0.41, "learning_rate": 1.31993664048224e-05, "loss": 24.601, "step": 33700 }, { "epoch": 0.41, "learning_rate": 1.3195592356602322e-05, "loss": 23.585, "step": 33710 }, { "epoch": 0.41, "learning_rate": 1.3191817801385226e-05, "loss": 24.0893, "step": 33720 }, { "epoch": 0.41, "learning_rate": 1.3188042739769965e-05, "loss": 24.0837, "step": 33730 }, { "epoch": 0.41, "learning_rate": 1.3184267172355467e-05, "loss": 23.6181, "step": 33740 }, { "epoch": 0.42, "learning_rate": 1.3180491099740751e-05, "loss": 23.8555, "step": 33750 }, { "epoch": 0.42, "learning_rate": 1.3176714522524905e-05, "loss": 24.2301, "step": 33760 }, { "epoch": 0.42, "learning_rate": 1.3172937441307105e-05, "loss": 24.2382, "step": 33770 }, { "epoch": 0.42, "learning_rate": 1.3169159856686602e-05, "loss": 23.5765, "step": 33780 }, { "epoch": 0.42, "learning_rate": 1.3165381769262731e-05, "loss": 24.2046, "step": 33790 }, { "epoch": 0.42, "learning_rate": 1.3161603179634905e-05, "loss": 24.1042, "step": 33800 }, { "epoch": 0.42, "learning_rate": 1.3157824088402615e-05, "loss": 24.0694, "step": 33810 }, { "epoch": 0.42, "learning_rate": 1.3154044496165433e-05, "loss": 24.0276, "step": 33820 }, { "epoch": 0.42, "learning_rate": 1.3150264403523015e-05, "loss": 24.2982, "step": 33830 }, { "epoch": 0.42, "learning_rate": 1.3146483811075084e-05, "loss": 23.1862, "step": 33840 }, { "epoch": 0.42, "learning_rate": 1.3142702719421456e-05, "loss": 24.2011, "step": 33850 }, { "epoch": 0.42, "learning_rate": 1.3138921129162023e-05, "loss": 24.0288, "step": 33860 }, { "epoch": 0.42, "learning_rate": 1.3135139040896747e-05, "loss": 23.8666, "step": 33870 }, { "epoch": 0.42, "learning_rate": 1.313135645522568e-05, "loss": 23.998, "step": 33880 }, { "epoch": 0.42, "learning_rate": 1.3127573372748943e-05, "loss": 24.0697, "step": 33890 }, { "epoch": 0.42, "learning_rate": 1.3123789794066748e-05, "loss": 23.325, "step": 33900 }, { "epoch": 0.42, "learning_rate": 1.3120005719779374e-05, "loss": 23.5162, "step": 33910 }, { "epoch": 0.42, "learning_rate": 1.3116221150487188e-05, "loss": 23.7068, "step": 33920 }, { "epoch": 0.42, "learning_rate": 1.311243608679063e-05, "loss": 23.6113, "step": 33930 }, { "epoch": 0.42, "learning_rate": 1.3108650529290215e-05, "loss": 23.8935, "step": 33940 }, { "epoch": 0.42, "learning_rate": 1.3104864478586545e-05, "loss": 24.7242, "step": 33950 }, { "epoch": 0.42, "learning_rate": 1.3101077935280297e-05, "loss": 24.4217, "step": 33960 }, { "epoch": 0.42, "learning_rate": 1.3097290899972229e-05, "loss": 24.0041, "step": 33970 }, { "epoch": 0.42, "learning_rate": 1.3093503373263165e-05, "loss": 24.1087, "step": 33980 }, { "epoch": 0.42, "learning_rate": 1.308971535575402e-05, "loss": 23.6086, "step": 33990 }, { "epoch": 0.42, "learning_rate": 1.308592684804578e-05, "loss": 24.593, "step": 34000 }, { "epoch": 0.42, "learning_rate": 1.3082137850739513e-05, "loss": 23.7541, "step": 34010 }, { "epoch": 0.42, "learning_rate": 1.3078348364436364e-05, "loss": 23.5261, "step": 34020 }, { "epoch": 0.42, "learning_rate": 1.307455838973755e-05, "loss": 23.8387, "step": 34030 }, { "epoch": 0.42, "learning_rate": 1.3070767927244377e-05, "loss": 23.6492, "step": 34040 }, { "epoch": 0.42, "learning_rate": 1.3066976977558212e-05, "loss": 23.727, "step": 34050 }, { "epoch": 0.42, "learning_rate": 1.3063185541280514e-05, "loss": 23.6357, "step": 34060 }, { "epoch": 0.42, "learning_rate": 1.3059393619012816e-05, "loss": 23.7549, "step": 34070 }, { "epoch": 0.42, "learning_rate": 1.3055601211356721e-05, "loss": 23.9949, "step": 34080 }, { "epoch": 0.42, "learning_rate": 1.3051808318913916e-05, "loss": 24.0283, "step": 34090 }, { "epoch": 0.42, "learning_rate": 1.304801494228616e-05, "loss": 24.1247, "step": 34100 }, { "epoch": 0.42, "learning_rate": 1.3044221082075297e-05, "loss": 23.9899, "step": 34110 }, { "epoch": 0.42, "learning_rate": 1.3040426738883236e-05, "loss": 23.5947, "step": 34120 }, { "epoch": 0.42, "learning_rate": 1.3036631913311974e-05, "loss": 24.0177, "step": 34130 }, { "epoch": 0.42, "learning_rate": 1.3032836605963578e-05, "loss": 24.1734, "step": 34140 }, { "epoch": 0.42, "learning_rate": 1.302904081744019e-05, "loss": 23.8344, "step": 34150 }, { "epoch": 0.42, "learning_rate": 1.3025244548344033e-05, "loss": 23.7325, "step": 34160 }, { "epoch": 0.42, "learning_rate": 1.3021447799277406e-05, "loss": 23.4729, "step": 34170 }, { "epoch": 0.42, "learning_rate": 1.3017650570842682e-05, "loss": 23.6946, "step": 34180 }, { "epoch": 0.42, "learning_rate": 1.3013852863642309e-05, "loss": 23.633, "step": 34190 }, { "epoch": 0.42, "learning_rate": 1.3010054678278816e-05, "loss": 24.1513, "step": 34200 }, { "epoch": 0.42, "learning_rate": 1.30062560153548e-05, "loss": 23.596, "step": 34210 }, { "epoch": 0.42, "learning_rate": 1.3002456875472942e-05, "loss": 23.943, "step": 34220 }, { "epoch": 0.42, "learning_rate": 1.299865725923599e-05, "loss": 23.5308, "step": 34230 }, { "epoch": 0.42, "learning_rate": 1.2994857167246781e-05, "loss": 23.9756, "step": 34240 }, { "epoch": 0.42, "learning_rate": 1.2991056600108211e-05, "loss": 23.4828, "step": 34250 }, { "epoch": 0.42, "learning_rate": 1.2987255558423262e-05, "loss": 24.1184, "step": 34260 }, { "epoch": 0.42, "learning_rate": 1.2983454042794989e-05, "loss": 23.9045, "step": 34270 }, { "epoch": 0.42, "learning_rate": 1.2979652053826522e-05, "loss": 23.7625, "step": 34280 }, { "epoch": 0.42, "learning_rate": 1.2975849592121063e-05, "loss": 23.6501, "step": 34290 }, { "epoch": 0.42, "learning_rate": 1.2972046658281899e-05, "loss": 23.8379, "step": 34300 }, { "epoch": 0.42, "learning_rate": 1.2968243252912376e-05, "loss": 23.7237, "step": 34310 }, { "epoch": 0.42, "learning_rate": 1.2964439376615931e-05, "loss": 23.4829, "step": 34320 }, { "epoch": 0.42, "learning_rate": 1.2960635029996064e-05, "loss": 23.3181, "step": 34330 }, { "epoch": 0.42, "learning_rate": 1.2956830213656358e-05, "loss": 23.9769, "step": 34340 }, { "epoch": 0.42, "learning_rate": 1.2953024928200462e-05, "loss": 23.4696, "step": 34350 }, { "epoch": 0.42, "learning_rate": 1.2949219174232104e-05, "loss": 23.4204, "step": 34360 }, { "epoch": 0.42, "learning_rate": 1.294541295235509e-05, "loss": 23.514, "step": 34370 }, { "epoch": 0.42, "learning_rate": 1.2941606263173294e-05, "loss": 23.7974, "step": 34380 }, { "epoch": 0.42, "learning_rate": 1.2937799107290666e-05, "loss": 23.9856, "step": 34390 }, { "epoch": 0.42, "learning_rate": 1.2933991485311234e-05, "loss": 23.5341, "step": 34400 }, { "epoch": 0.42, "learning_rate": 1.2930183397839094e-05, "loss": 23.7443, "step": 34410 }, { "epoch": 0.42, "learning_rate": 1.2926374845478416e-05, "loss": 23.5753, "step": 34420 }, { "epoch": 0.42, "learning_rate": 1.2922565828833453e-05, "loss": 24.0746, "step": 34430 }, { "epoch": 0.42, "learning_rate": 1.2918756348508522e-05, "loss": 23.3624, "step": 34440 }, { "epoch": 0.42, "learning_rate": 1.2914946405108013e-05, "loss": 23.974, "step": 34450 }, { "epoch": 0.42, "learning_rate": 1.2911135999236396e-05, "loss": 23.8832, "step": 34460 }, { "epoch": 0.42, "learning_rate": 1.2907325131498212e-05, "loss": 23.2147, "step": 34470 }, { "epoch": 0.42, "learning_rate": 1.2903513802498074e-05, "loss": 23.4279, "step": 34480 }, { "epoch": 0.42, "learning_rate": 1.2899702012840665e-05, "loss": 23.6755, "step": 34490 }, { "epoch": 0.42, "learning_rate": 1.289588976313075e-05, "loss": 24.1121, "step": 34500 }, { "epoch": 0.42, "learning_rate": 1.2892077053973162e-05, "loss": 24.018, "step": 34510 }, { "epoch": 0.42, "learning_rate": 1.2888263885972802e-05, "loss": 24.1404, "step": 34520 }, { "epoch": 0.42, "learning_rate": 1.2884450259734654e-05, "loss": 24.0035, "step": 34530 }, { "epoch": 0.42, "learning_rate": 1.2880636175863766e-05, "loss": 23.5619, "step": 34540 }, { "epoch": 0.42, "learning_rate": 1.287682163496526e-05, "loss": 23.9178, "step": 34550 }, { "epoch": 0.43, "learning_rate": 1.2873006637644336e-05, "loss": 23.7565, "step": 34560 }, { "epoch": 0.43, "learning_rate": 1.2869191184506264e-05, "loss": 23.9939, "step": 34570 }, { "epoch": 0.43, "learning_rate": 1.286537527615638e-05, "loss": 24.3416, "step": 34580 }, { "epoch": 0.43, "learning_rate": 1.2861558913200099e-05, "loss": 24.2975, "step": 34590 }, { "epoch": 0.43, "learning_rate": 1.285774209624291e-05, "loss": 24.1843, "step": 34600 }, { "epoch": 0.43, "learning_rate": 1.2853924825890368e-05, "loss": 23.9972, "step": 34610 }, { "epoch": 0.43, "learning_rate": 1.28501071027481e-05, "loss": 23.9928, "step": 34620 }, { "epoch": 0.43, "learning_rate": 1.284628892742181e-05, "loss": 23.3436, "step": 34630 }, { "epoch": 0.43, "learning_rate": 1.2842470300517272e-05, "loss": 23.9804, "step": 34640 }, { "epoch": 0.43, "learning_rate": 1.2838651222640325e-05, "loss": 23.717, "step": 34650 }, { "epoch": 0.43, "learning_rate": 1.2834831694396893e-05, "loss": 24.0217, "step": 34660 }, { "epoch": 0.43, "learning_rate": 1.283101171639296e-05, "loss": 23.3965, "step": 34670 }, { "epoch": 0.43, "learning_rate": 1.2827191289234584e-05, "loss": 24.0214, "step": 34680 }, { "epoch": 0.43, "learning_rate": 1.2823370413527896e-05, "loss": 23.944, "step": 34690 }, { "epoch": 0.43, "learning_rate": 1.28195490898791e-05, "loss": 24.1415, "step": 34700 }, { "epoch": 0.43, "learning_rate": 1.2815727318894463e-05, "loss": 24.0268, "step": 34710 }, { "epoch": 0.43, "learning_rate": 1.281190510118033e-05, "loss": 23.6587, "step": 34720 }, { "epoch": 0.43, "learning_rate": 1.2808082437343119e-05, "loss": 23.9035, "step": 34730 }, { "epoch": 0.43, "learning_rate": 1.2804259327989314e-05, "loss": 23.9634, "step": 34740 }, { "epoch": 0.43, "learning_rate": 1.2800435773725469e-05, "loss": 24.1429, "step": 34750 }, { "epoch": 0.43, "learning_rate": 1.279661177515821e-05, "loss": 23.8674, "step": 34760 }, { "epoch": 0.43, "learning_rate": 1.2792787332894236e-05, "loss": 24.1792, "step": 34770 }, { "epoch": 0.43, "learning_rate": 1.2788962447540314e-05, "loss": 23.8747, "step": 34780 }, { "epoch": 0.43, "learning_rate": 1.278513711970328e-05, "loss": 23.7642, "step": 34790 }, { "epoch": 0.43, "learning_rate": 1.2781311349990042e-05, "loss": 23.9499, "step": 34800 }, { "epoch": 0.43, "learning_rate": 1.2777485139007581e-05, "loss": 24.2077, "step": 34810 }, { "epoch": 0.43, "learning_rate": 1.2773658487362938e-05, "loss": 24.1049, "step": 34820 }, { "epoch": 0.43, "learning_rate": 1.276983139566324e-05, "loss": 23.9135, "step": 34830 }, { "epoch": 0.43, "learning_rate": 1.2766003864515666e-05, "loss": 24.0982, "step": 34840 }, { "epoch": 0.43, "learning_rate": 1.2762175894527478e-05, "loss": 24.1915, "step": 34850 }, { "epoch": 0.43, "learning_rate": 1.2758347486305999e-05, "loss": 23.8011, "step": 34860 }, { "epoch": 0.43, "learning_rate": 1.2754518640458636e-05, "loss": 24.0383, "step": 34870 }, { "epoch": 0.43, "learning_rate": 1.2750689357592841e-05, "loss": 23.1697, "step": 34880 }, { "epoch": 0.43, "learning_rate": 1.2746859638316156e-05, "loss": 24.4788, "step": 34890 }, { "epoch": 0.43, "learning_rate": 1.2743029483236188e-05, "loss": 23.9881, "step": 34900 }, { "epoch": 0.43, "learning_rate": 1.2739198892960606e-05, "loss": 23.7031, "step": 34910 }, { "epoch": 0.43, "learning_rate": 1.2735367868097157e-05, "loss": 24.2168, "step": 34920 }, { "epoch": 0.43, "learning_rate": 1.2731536409253647e-05, "loss": 24.2601, "step": 34930 }, { "epoch": 0.43, "learning_rate": 1.2727704517037962e-05, "loss": 24.3282, "step": 34940 }, { "epoch": 0.43, "learning_rate": 1.2723872192058045e-05, "loss": 23.9442, "step": 34950 }, { "epoch": 0.43, "learning_rate": 1.272003943492192e-05, "loss": 24.1127, "step": 34960 }, { "epoch": 0.43, "learning_rate": 1.2716206246237672e-05, "loss": 24.2955, "step": 34970 }, { "epoch": 0.43, "learning_rate": 1.2712372626613454e-05, "loss": 24.008, "step": 34980 }, { "epoch": 0.43, "learning_rate": 1.270853857665749e-05, "loss": 23.6555, "step": 34990 }, { "epoch": 0.43, "learning_rate": 1.2704704096978072e-05, "loss": 23.5731, "step": 35000 }, { "epoch": 0.43, "learning_rate": 1.270086918818356e-05, "loss": 24.1461, "step": 35010 }, { "epoch": 0.43, "learning_rate": 1.269703385088238e-05, "loss": 23.8053, "step": 35020 }, { "epoch": 0.43, "learning_rate": 1.2693198085683031e-05, "loss": 23.6261, "step": 35030 }, { "epoch": 0.43, "learning_rate": 1.2689361893194079e-05, "loss": 23.7027, "step": 35040 }, { "epoch": 0.43, "learning_rate": 1.2685525274024145e-05, "loss": 24.1892, "step": 35050 }, { "epoch": 0.43, "learning_rate": 1.2681688228781937e-05, "loss": 23.5768, "step": 35060 }, { "epoch": 0.43, "learning_rate": 1.267785075807622e-05, "loss": 24.0822, "step": 35070 }, { "epoch": 0.43, "learning_rate": 1.2674012862515827e-05, "loss": 23.8173, "step": 35080 }, { "epoch": 0.43, "learning_rate": 1.2670174542709659e-05, "loss": 23.3472, "step": 35090 }, { "epoch": 0.43, "learning_rate": 1.2666335799266691e-05, "loss": 23.6577, "step": 35100 }, { "epoch": 0.43, "learning_rate": 1.266249663279595e-05, "loss": 23.4051, "step": 35110 }, { "epoch": 0.43, "learning_rate": 1.2658657043906546e-05, "loss": 23.4313, "step": 35120 }, { "epoch": 0.43, "learning_rate": 1.265481703320765e-05, "loss": 23.3761, "step": 35130 }, { "epoch": 0.43, "learning_rate": 1.2650976601308495e-05, "loss": 24.1644, "step": 35140 }, { "epoch": 0.43, "learning_rate": 1.2647135748818386e-05, "loss": 23.7114, "step": 35150 }, { "epoch": 0.43, "learning_rate": 1.2643294476346693e-05, "loss": 23.7201, "step": 35160 }, { "epoch": 0.43, "learning_rate": 1.2639452784502858e-05, "loss": 23.0878, "step": 35170 }, { "epoch": 0.43, "learning_rate": 1.2635610673896381e-05, "loss": 23.5948, "step": 35180 }, { "epoch": 0.43, "learning_rate": 1.2631768145136831e-05, "loss": 24.0506, "step": 35190 }, { "epoch": 0.43, "learning_rate": 1.2627925198833852e-05, "loss": 24.0549, "step": 35200 }, { "epoch": 0.43, "learning_rate": 1.262408183559714e-05, "loss": 24.0946, "step": 35210 }, { "epoch": 0.43, "learning_rate": 1.2620238056036465e-05, "loss": 23.7414, "step": 35220 }, { "epoch": 0.43, "learning_rate": 1.2616393860761667e-05, "loss": 23.8327, "step": 35230 }, { "epoch": 0.43, "learning_rate": 1.261254925038264e-05, "loss": 23.9585, "step": 35240 }, { "epoch": 0.43, "learning_rate": 1.2608704225509359e-05, "loss": 23.8736, "step": 35250 }, { "epoch": 0.43, "learning_rate": 1.2604858786751847e-05, "loss": 23.8568, "step": 35260 }, { "epoch": 0.43, "learning_rate": 1.2601012934720217e-05, "loss": 23.4676, "step": 35270 }, { "epoch": 0.43, "learning_rate": 1.2597166670024617e-05, "loss": 24.0448, "step": 35280 }, { "epoch": 0.43, "learning_rate": 1.2593319993275283e-05, "loss": 23.5583, "step": 35290 }, { "epoch": 0.43, "learning_rate": 1.2589472905082516e-05, "loss": 23.0704, "step": 35300 }, { "epoch": 0.43, "learning_rate": 1.2585625406056667e-05, "loss": 23.7983, "step": 35310 }, { "epoch": 0.43, "learning_rate": 1.2581777496808165e-05, "loss": 23.9004, "step": 35320 }, { "epoch": 0.43, "learning_rate": 1.2577929177947504e-05, "loss": 23.8559, "step": 35330 }, { "epoch": 0.43, "learning_rate": 1.2574080450085233e-05, "loss": 23.744, "step": 35340 }, { "epoch": 0.43, "learning_rate": 1.2570231313831975e-05, "loss": 23.614, "step": 35350 }, { "epoch": 0.43, "learning_rate": 1.2566381769798415e-05, "loss": 23.1839, "step": 35360 }, { "epoch": 0.43, "learning_rate": 1.2562531818595307e-05, "loss": 23.5219, "step": 35370 }, { "epoch": 0.44, "learning_rate": 1.2558681460833459e-05, "loss": 24.3006, "step": 35380 }, { "epoch": 0.44, "learning_rate": 1.255483069712375e-05, "loss": 23.9845, "step": 35390 }, { "epoch": 0.44, "learning_rate": 1.255097952807713e-05, "loss": 23.6497, "step": 35400 }, { "epoch": 0.44, "learning_rate": 1.2547127954304599e-05, "loss": 24.1541, "step": 35410 }, { "epoch": 0.44, "learning_rate": 1.2543275976417232e-05, "loss": 23.5978, "step": 35420 }, { "epoch": 0.44, "learning_rate": 1.2539423595026165e-05, "loss": 24.2459, "step": 35430 }, { "epoch": 0.44, "learning_rate": 1.25355708107426e-05, "loss": 24.088, "step": 35440 }, { "epoch": 0.44, "learning_rate": 1.2531717624177797e-05, "loss": 23.9907, "step": 35450 }, { "epoch": 0.44, "learning_rate": 1.2527864035943087e-05, "loss": 24.2774, "step": 35460 }, { "epoch": 0.44, "learning_rate": 1.252401004664986e-05, "loss": 22.9237, "step": 35470 }, { "epoch": 0.44, "learning_rate": 1.2520155656909571e-05, "loss": 23.6784, "step": 35480 }, { "epoch": 0.44, "learning_rate": 1.2516300867333738e-05, "loss": 24.362, "step": 35490 }, { "epoch": 0.44, "learning_rate": 1.2512445678533944e-05, "loss": 23.8621, "step": 35500 }, { "epoch": 0.44, "learning_rate": 1.2508590091121833e-05, "loss": 24.0838, "step": 35510 }, { "epoch": 0.44, "learning_rate": 1.2504734105709116e-05, "loss": 24.0591, "step": 35520 }, { "epoch": 0.44, "learning_rate": 1.2500877722907561e-05, "loss": 24.1405, "step": 35530 }, { "epoch": 0.44, "learning_rate": 1.2497020943329006e-05, "loss": 24.2045, "step": 35540 }, { "epoch": 0.44, "learning_rate": 1.2493163767585348e-05, "loss": 23.3965, "step": 35550 }, { "epoch": 0.44, "learning_rate": 1.2489306196288546e-05, "loss": 24.4582, "step": 35560 }, { "epoch": 0.44, "learning_rate": 1.2485448230050627e-05, "loss": 23.4629, "step": 35570 }, { "epoch": 0.44, "learning_rate": 1.2481589869483675e-05, "loss": 23.6011, "step": 35580 }, { "epoch": 0.44, "learning_rate": 1.2477731115199836e-05, "loss": 24.5476, "step": 35590 }, { "epoch": 0.44, "learning_rate": 1.2473871967811326e-05, "loss": 24.1909, "step": 35600 }, { "epoch": 0.44, "learning_rate": 1.2470012427930415e-05, "loss": 23.5643, "step": 35610 }, { "epoch": 0.44, "learning_rate": 1.2466152496169439e-05, "loss": 24.4704, "step": 35620 }, { "epoch": 0.44, "learning_rate": 1.2462292173140798e-05, "loss": 23.7575, "step": 35630 }, { "epoch": 0.44, "learning_rate": 1.245843145945695e-05, "loss": 24.3587, "step": 35640 }, { "epoch": 0.44, "learning_rate": 1.2454570355730415e-05, "loss": 23.5955, "step": 35650 }, { "epoch": 0.44, "learning_rate": 1.2450708862573779e-05, "loss": 24.1799, "step": 35660 }, { "epoch": 0.44, "learning_rate": 1.2446846980599689e-05, "loss": 24.5774, "step": 35670 }, { "epoch": 0.44, "learning_rate": 1.2442984710420852e-05, "loss": 23.4188, "step": 35680 }, { "epoch": 0.44, "learning_rate": 1.2439122052650033e-05, "loss": 23.8878, "step": 35690 }, { "epoch": 0.44, "learning_rate": 1.2435259007900067e-05, "loss": 23.649, "step": 35700 }, { "epoch": 0.44, "learning_rate": 1.2431395576783842e-05, "loss": 23.8448, "step": 35710 }, { "epoch": 0.44, "learning_rate": 1.2427531759914313e-05, "loss": 23.7174, "step": 35720 }, { "epoch": 0.44, "learning_rate": 1.2423667557904496e-05, "loss": 23.4169, "step": 35730 }, { "epoch": 0.44, "learning_rate": 1.2419802971367463e-05, "loss": 24.4081, "step": 35740 }, { "epoch": 0.44, "learning_rate": 1.2415938000916352e-05, "loss": 24.5603, "step": 35750 }, { "epoch": 0.44, "learning_rate": 1.2412072647164359e-05, "loss": 24.2177, "step": 35760 }, { "epoch": 0.44, "learning_rate": 1.2408206910724745e-05, "loss": 23.6413, "step": 35770 }, { "epoch": 0.44, "learning_rate": 1.2404340792210825e-05, "loss": 23.616, "step": 35780 }, { "epoch": 0.44, "learning_rate": 1.240047429223598e-05, "loss": 24.1405, "step": 35790 }, { "epoch": 0.44, "learning_rate": 1.2396607411413652e-05, "loss": 23.19, "step": 35800 }, { "epoch": 0.44, "learning_rate": 1.239274015035734e-05, "loss": 24.2478, "step": 35810 }, { "epoch": 0.44, "learning_rate": 1.2388872509680603e-05, "loss": 23.626, "step": 35820 }, { "epoch": 0.44, "learning_rate": 1.2385004489997064e-05, "loss": 23.3168, "step": 35830 }, { "epoch": 0.44, "learning_rate": 1.2381136091920404e-05, "loss": 23.4434, "step": 35840 }, { "epoch": 0.44, "learning_rate": 1.2377267316064363e-05, "loss": 23.8878, "step": 35850 }, { "epoch": 0.44, "learning_rate": 1.237339816304274e-05, "loss": 23.6687, "step": 35860 }, { "epoch": 0.44, "learning_rate": 1.2369528633469401e-05, "loss": 23.6979, "step": 35870 }, { "epoch": 0.44, "learning_rate": 1.2365658727958261e-05, "loss": 23.77, "step": 35880 }, { "epoch": 0.44, "learning_rate": 1.2361788447123304e-05, "loss": 23.8847, "step": 35890 }, { "epoch": 0.44, "learning_rate": 1.2357917791578572e-05, "loss": 23.7703, "step": 35900 }, { "epoch": 0.44, "learning_rate": 1.2354046761938157e-05, "loss": 23.6448, "step": 35910 }, { "epoch": 0.44, "learning_rate": 1.2350175358816222e-05, "loss": 24.1383, "step": 35920 }, { "epoch": 0.44, "learning_rate": 1.2346303582826987e-05, "loss": 23.5686, "step": 35930 }, { "epoch": 0.44, "learning_rate": 1.2342431434584725e-05, "loss": 23.6557, "step": 35940 }, { "epoch": 0.44, "learning_rate": 1.2338558914703774e-05, "loss": 23.7483, "step": 35950 }, { "epoch": 0.44, "learning_rate": 1.2334686023798525e-05, "loss": 22.8375, "step": 35960 }, { "epoch": 0.44, "learning_rate": 1.2330812762483442e-05, "loss": 23.7431, "step": 35970 }, { "epoch": 0.44, "learning_rate": 1.2326939131373025e-05, "loss": 24.0206, "step": 35980 }, { "epoch": 0.44, "learning_rate": 1.2323065131081854e-05, "loss": 23.8243, "step": 35990 }, { "epoch": 0.44, "learning_rate": 1.2319190762224558e-05, "loss": 23.8762, "step": 36000 }, { "epoch": 0.44, "learning_rate": 1.2315316025415819e-05, "loss": 23.574, "step": 36010 }, { "epoch": 0.44, "learning_rate": 1.2311440921270392e-05, "loss": 23.8303, "step": 36020 }, { "epoch": 0.44, "learning_rate": 1.2307565450403078e-05, "loss": 23.3034, "step": 36030 }, { "epoch": 0.44, "learning_rate": 1.2303689613428742e-05, "loss": 23.822, "step": 36040 }, { "epoch": 0.44, "learning_rate": 1.22998134109623e-05, "loss": 23.0302, "step": 36050 }, { "epoch": 0.44, "learning_rate": 1.229593684361874e-05, "loss": 24.159, "step": 36060 }, { "epoch": 0.44, "learning_rate": 1.2292059912013092e-05, "loss": 24.1198, "step": 36070 }, { "epoch": 0.44, "learning_rate": 1.2288182616760453e-05, "loss": 23.7238, "step": 36080 }, { "epoch": 0.44, "learning_rate": 1.2284304958475977e-05, "loss": 23.236, "step": 36090 }, { "epoch": 0.44, "learning_rate": 1.2280426937774873e-05, "loss": 23.5616, "step": 36100 }, { "epoch": 0.44, "learning_rate": 1.227654855527241e-05, "loss": 23.3113, "step": 36110 }, { "epoch": 0.44, "learning_rate": 1.227266981158391e-05, "loss": 23.6148, "step": 36120 }, { "epoch": 0.44, "learning_rate": 1.2268790707324759e-05, "loss": 24.1345, "step": 36130 }, { "epoch": 0.44, "learning_rate": 1.2264911243110396e-05, "loss": 23.7521, "step": 36140 }, { "epoch": 0.44, "learning_rate": 1.2261031419556315e-05, "loss": 24.1493, "step": 36150 }, { "epoch": 0.44, "learning_rate": 1.2257151237278073e-05, "loss": 24.0738, "step": 36160 }, { "epoch": 0.44, "learning_rate": 1.2253270696891282e-05, "loss": 23.4554, "step": 36170 }, { "epoch": 0.44, "learning_rate": 1.2249389799011607e-05, "loss": 23.6792, "step": 36180 }, { "epoch": 0.45, "learning_rate": 1.224550854425477e-05, "loss": 23.9828, "step": 36190 }, { "epoch": 0.45, "learning_rate": 1.2241626933236555e-05, "loss": 23.9641, "step": 36200 }, { "epoch": 0.45, "learning_rate": 1.2237744966572798e-05, "loss": 24.0738, "step": 36210 }, { "epoch": 0.45, "learning_rate": 1.2233862644879392e-05, "loss": 24.003, "step": 36220 }, { "epoch": 0.45, "learning_rate": 1.2229979968772289e-05, "loss": 24.9026, "step": 36230 }, { "epoch": 0.45, "learning_rate": 1.2226096938867498e-05, "loss": 23.4652, "step": 36240 }, { "epoch": 0.45, "learning_rate": 1.2222213555781075e-05, "loss": 23.3972, "step": 36250 }, { "epoch": 0.45, "learning_rate": 1.2218329820129142e-05, "loss": 24.0285, "step": 36260 }, { "epoch": 0.45, "learning_rate": 1.2214445732527878e-05, "loss": 23.696, "step": 36270 }, { "epoch": 0.45, "learning_rate": 1.2210561293593503e-05, "loss": 23.5207, "step": 36280 }, { "epoch": 0.45, "learning_rate": 1.2206676503942314e-05, "loss": 23.7942, "step": 36290 }, { "epoch": 0.45, "learning_rate": 1.2202791364190643e-05, "loss": 23.0954, "step": 36300 }, { "epoch": 0.45, "learning_rate": 1.2198905874954893e-05, "loss": 23.9616, "step": 36310 }, { "epoch": 0.45, "learning_rate": 1.2195020036851518e-05, "loss": 24.472, "step": 36320 }, { "epoch": 0.45, "learning_rate": 1.2191133850497018e-05, "loss": 23.6463, "step": 36330 }, { "epoch": 0.45, "learning_rate": 1.2187247316507967e-05, "loss": 23.9954, "step": 36340 }, { "epoch": 0.45, "learning_rate": 1.2183360435500973e-05, "loss": 23.7431, "step": 36350 }, { "epoch": 0.45, "learning_rate": 1.2179473208092717e-05, "loss": 24.1434, "step": 36360 }, { "epoch": 0.45, "learning_rate": 1.2175585634899927e-05, "loss": 24.079, "step": 36370 }, { "epoch": 0.45, "learning_rate": 1.2171697716539381e-05, "loss": 24.0331, "step": 36380 }, { "epoch": 0.45, "learning_rate": 1.2167809453627921e-05, "loss": 23.4672, "step": 36390 }, { "epoch": 0.45, "learning_rate": 1.216392084678244e-05, "loss": 23.8536, "step": 36400 }, { "epoch": 0.45, "learning_rate": 1.2160031896619886e-05, "loss": 23.769, "step": 36410 }, { "epoch": 0.45, "learning_rate": 1.2156142603757257e-05, "loss": 23.7982, "step": 36420 }, { "epoch": 0.45, "learning_rate": 1.215225296881161e-05, "loss": 24.6162, "step": 36430 }, { "epoch": 0.45, "learning_rate": 1.214836299240006e-05, "loss": 23.4315, "step": 36440 }, { "epoch": 0.45, "learning_rate": 1.2144472675139766e-05, "loss": 23.5375, "step": 36450 }, { "epoch": 0.45, "learning_rate": 1.2140582017647948e-05, "loss": 23.6764, "step": 36460 }, { "epoch": 0.45, "learning_rate": 1.2136691020541883e-05, "loss": 23.6613, "step": 36470 }, { "epoch": 0.45, "learning_rate": 1.2132799684438892e-05, "loss": 23.252, "step": 36480 }, { "epoch": 0.45, "learning_rate": 1.2128908009956354e-05, "loss": 23.7243, "step": 36490 }, { "epoch": 0.45, "learning_rate": 1.2125015997711714e-05, "loss": 24.0925, "step": 36500 }, { "epoch": 0.45, "learning_rate": 1.2121123648322448e-05, "loss": 23.6067, "step": 36510 }, { "epoch": 0.45, "learning_rate": 1.21172309624061e-05, "loss": 23.8811, "step": 36520 }, { "epoch": 0.45, "learning_rate": 1.2113337940580268e-05, "loss": 23.6926, "step": 36530 }, { "epoch": 0.45, "learning_rate": 1.2109444583462598e-05, "loss": 23.939, "step": 36540 }, { "epoch": 0.45, "learning_rate": 1.2105550891670787e-05, "loss": 24.4052, "step": 36550 }, { "epoch": 0.45, "learning_rate": 1.2101656865822594e-05, "loss": 24.2121, "step": 36560 }, { "epoch": 0.45, "learning_rate": 1.2097762506535823e-05, "loss": 23.5814, "step": 36570 }, { "epoch": 0.45, "learning_rate": 1.2093867814428338e-05, "loss": 23.4952, "step": 36580 }, { "epoch": 0.45, "learning_rate": 1.2089972790118045e-05, "loss": 23.7729, "step": 36590 }, { "epoch": 0.45, "learning_rate": 1.208607743422292e-05, "loss": 23.1332, "step": 36600 }, { "epoch": 0.45, "learning_rate": 1.2082181747360968e-05, "loss": 23.8882, "step": 36610 }, { "epoch": 0.45, "learning_rate": 1.207828573015027e-05, "loss": 24.0394, "step": 36620 }, { "epoch": 0.45, "learning_rate": 1.2074389383208946e-05, "loss": 23.8094, "step": 36630 }, { "epoch": 0.45, "learning_rate": 1.207049270715517e-05, "loss": 24.7677, "step": 36640 }, { "epoch": 0.45, "learning_rate": 1.2066595702607169e-05, "loss": 23.5302, "step": 36650 }, { "epoch": 0.45, "learning_rate": 1.206269837018322e-05, "loss": 24.2851, "step": 36660 }, { "epoch": 0.45, "learning_rate": 1.2058800710501665e-05, "loss": 23.9334, "step": 36670 }, { "epoch": 0.45, "learning_rate": 1.2054902724180877e-05, "loss": 23.5619, "step": 36680 }, { "epoch": 0.45, "learning_rate": 1.2051004411839296e-05, "loss": 23.5049, "step": 36690 }, { "epoch": 0.45, "learning_rate": 1.2047105774095412e-05, "loss": 23.7539, "step": 36700 }, { "epoch": 0.45, "learning_rate": 1.2043206811567756e-05, "loss": 24.1121, "step": 36710 }, { "epoch": 0.45, "learning_rate": 1.2039307524874922e-05, "loss": 23.7938, "step": 36720 }, { "epoch": 0.45, "learning_rate": 1.2035407914635556e-05, "loss": 23.5676, "step": 36730 }, { "epoch": 0.45, "learning_rate": 1.2031507981468342e-05, "loss": 23.6749, "step": 36740 }, { "epoch": 0.45, "learning_rate": 1.2027607725992035e-05, "loss": 24.1855, "step": 36750 }, { "epoch": 0.45, "learning_rate": 1.202370714882542e-05, "loss": 23.4765, "step": 36760 }, { "epoch": 0.45, "learning_rate": 1.2019806250587353e-05, "loss": 23.8847, "step": 36770 }, { "epoch": 0.45, "learning_rate": 1.2015905031896726e-05, "loss": 23.7594, "step": 36780 }, { "epoch": 0.45, "learning_rate": 1.2012003493372486e-05, "loss": 24.6287, "step": 36790 }, { "epoch": 0.45, "learning_rate": 1.2008101635633636e-05, "loss": 23.7894, "step": 36800 }, { "epoch": 0.45, "learning_rate": 1.2004199459299221e-05, "loss": 24.2494, "step": 36810 }, { "epoch": 0.45, "learning_rate": 1.2000296964988345e-05, "loss": 23.8566, "step": 36820 }, { "epoch": 0.45, "learning_rate": 1.1996394153320156e-05, "loss": 23.9839, "step": 36830 }, { "epoch": 0.45, "learning_rate": 1.199249102491386e-05, "loss": 24.3287, "step": 36840 }, { "epoch": 0.45, "learning_rate": 1.1988587580388702e-05, "loss": 23.6911, "step": 36850 }, { "epoch": 0.45, "learning_rate": 1.1984683820363987e-05, "loss": 24.0313, "step": 36860 }, { "epoch": 0.45, "learning_rate": 1.1980779745459068e-05, "loss": 23.7763, "step": 36870 }, { "epoch": 0.45, "learning_rate": 1.197687535629334e-05, "loss": 23.9151, "step": 36880 }, { "epoch": 0.45, "learning_rate": 1.1972970653486258e-05, "loss": 23.778, "step": 36890 }, { "epoch": 0.45, "learning_rate": 1.1969065637657327e-05, "loss": 23.9538, "step": 36900 }, { "epoch": 0.45, "learning_rate": 1.196516030942609e-05, "loss": 23.9912, "step": 36910 }, { "epoch": 0.45, "learning_rate": 1.196125466941215e-05, "loss": 23.9354, "step": 36920 }, { "epoch": 0.45, "learning_rate": 1.1957348718235158e-05, "loss": 23.6357, "step": 36930 }, { "epoch": 0.45, "learning_rate": 1.1953442456514817e-05, "loss": 23.8599, "step": 36940 }, { "epoch": 0.45, "learning_rate": 1.1949535884870867e-05, "loss": 23.8359, "step": 36950 }, { "epoch": 0.45, "learning_rate": 1.194562900392311e-05, "loss": 23.7341, "step": 36960 }, { "epoch": 0.45, "learning_rate": 1.1941721814291394e-05, "loss": 23.93, "step": 36970 }, { "epoch": 0.45, "learning_rate": 1.193781431659561e-05, "loss": 24.0449, "step": 36980 }, { "epoch": 0.45, "learning_rate": 1.1933906511455708e-05, "loss": 23.6868, "step": 36990 }, { "epoch": 0.46, "learning_rate": 1.1929998399491677e-05, "loss": 23.8562, "step": 37000 }, { "epoch": 0.46, "learning_rate": 1.1926089981323562e-05, "loss": 23.9488, "step": 37010 }, { "epoch": 0.46, "learning_rate": 1.192218125757145e-05, "loss": 24.2376, "step": 37020 }, { "epoch": 0.46, "learning_rate": 1.191827222885548e-05, "loss": 23.4668, "step": 37030 }, { "epoch": 0.46, "learning_rate": 1.1914362895795846e-05, "loss": 23.8869, "step": 37040 }, { "epoch": 0.46, "learning_rate": 1.1910453259012775e-05, "loss": 23.3794, "step": 37050 }, { "epoch": 0.46, "learning_rate": 1.1906543319126554e-05, "loss": 23.6817, "step": 37060 }, { "epoch": 0.46, "learning_rate": 1.1902633076757516e-05, "loss": 24.2844, "step": 37070 }, { "epoch": 0.46, "learning_rate": 1.1898722532526041e-05, "loss": 24.0926, "step": 37080 }, { "epoch": 0.46, "learning_rate": 1.1894811687052554e-05, "loss": 23.7629, "step": 37090 }, { "epoch": 0.46, "learning_rate": 1.1890900540957537e-05, "loss": 23.7866, "step": 37100 }, { "epoch": 0.46, "learning_rate": 1.1886989094861508e-05, "loss": 23.6009, "step": 37110 }, { "epoch": 0.46, "learning_rate": 1.1883077349385035e-05, "loss": 23.8907, "step": 37120 }, { "epoch": 0.46, "learning_rate": 1.1879165305148741e-05, "loss": 24.143, "step": 37130 }, { "epoch": 0.46, "learning_rate": 1.1875252962773294e-05, "loss": 24.0197, "step": 37140 }, { "epoch": 0.46, "learning_rate": 1.1871340322879399e-05, "loss": 24.1407, "step": 37150 }, { "epoch": 0.46, "learning_rate": 1.1867427386087824e-05, "loss": 24.0081, "step": 37160 }, { "epoch": 0.46, "learning_rate": 1.1863514153019372e-05, "loss": 23.4805, "step": 37170 }, { "epoch": 0.46, "learning_rate": 1.1859600624294901e-05, "loss": 24.3308, "step": 37180 }, { "epoch": 0.46, "learning_rate": 1.1855686800535308e-05, "loss": 23.2398, "step": 37190 }, { "epoch": 0.46, "learning_rate": 1.1851772682361547e-05, "loss": 24.2173, "step": 37200 }, { "epoch": 0.46, "learning_rate": 1.1847858270394605e-05, "loss": 23.4941, "step": 37210 }, { "epoch": 0.46, "learning_rate": 1.1843943565255531e-05, "loss": 23.7758, "step": 37220 }, { "epoch": 0.46, "learning_rate": 1.1840028567565407e-05, "loss": 23.7581, "step": 37230 }, { "epoch": 0.46, "learning_rate": 1.1836113277945372e-05, "loss": 23.9343, "step": 37240 }, { "epoch": 0.46, "learning_rate": 1.18321976970166e-05, "loss": 23.2679, "step": 37250 }, { "epoch": 0.46, "learning_rate": 1.1828281825400325e-05, "loss": 23.5951, "step": 37260 }, { "epoch": 0.46, "learning_rate": 1.1824365663717818e-05, "loss": 23.2779, "step": 37270 }, { "epoch": 0.46, "learning_rate": 1.1820449212590396e-05, "loss": 24.3192, "step": 37280 }, { "epoch": 0.46, "learning_rate": 1.1816532472639426e-05, "loss": 24.2445, "step": 37290 }, { "epoch": 0.46, "learning_rate": 1.181261544448632e-05, "loss": 23.8466, "step": 37300 }, { "epoch": 0.46, "learning_rate": 1.1808698128752529e-05, "loss": 23.5264, "step": 37310 }, { "epoch": 0.46, "learning_rate": 1.180478052605956e-05, "loss": 24.4149, "step": 37320 }, { "epoch": 0.46, "learning_rate": 1.180086263702896e-05, "loss": 24.0136, "step": 37330 }, { "epoch": 0.46, "learning_rate": 1.1796944462282323e-05, "loss": 24.0671, "step": 37340 }, { "epoch": 0.46, "learning_rate": 1.1793026002441281e-05, "loss": 23.733, "step": 37350 }, { "epoch": 0.46, "learning_rate": 1.1789107258127525e-05, "loss": 23.4026, "step": 37360 }, { "epoch": 0.46, "learning_rate": 1.1785188229962783e-05, "loss": 23.9275, "step": 37370 }, { "epoch": 0.46, "learning_rate": 1.1781268918568824e-05, "loss": 23.7718, "step": 37380 }, { "epoch": 0.46, "learning_rate": 1.177734932456747e-05, "loss": 24.1236, "step": 37390 }, { "epoch": 0.46, "learning_rate": 1.1773429448580584e-05, "loss": 23.542, "step": 37400 }, { "epoch": 0.46, "learning_rate": 1.1769509291230078e-05, "loss": 23.7077, "step": 37410 }, { "epoch": 0.46, "learning_rate": 1.1765588853137895e-05, "loss": 23.7472, "step": 37420 }, { "epoch": 0.46, "learning_rate": 1.1761668134926045e-05, "loss": 23.8992, "step": 37430 }, { "epoch": 0.46, "learning_rate": 1.1757747137216562e-05, "loss": 23.8859, "step": 37440 }, { "epoch": 0.46, "learning_rate": 1.1753825860631534e-05, "loss": 24.1823, "step": 37450 }, { "epoch": 0.46, "learning_rate": 1.1749904305793092e-05, "loss": 23.1084, "step": 37460 }, { "epoch": 0.46, "learning_rate": 1.174598247332341e-05, "loss": 24.0395, "step": 37470 }, { "epoch": 0.46, "learning_rate": 1.1742060363844708e-05, "loss": 23.9121, "step": 37480 }, { "epoch": 0.46, "learning_rate": 1.1738137977979245e-05, "loss": 23.7247, "step": 37490 }, { "epoch": 0.46, "learning_rate": 1.1734215316349333e-05, "loss": 23.9758, "step": 37500 }, { "epoch": 0.46, "learning_rate": 1.1730292379577319e-05, "loss": 24.1464, "step": 37510 }, { "epoch": 0.46, "learning_rate": 1.1726369168285596e-05, "loss": 23.9289, "step": 37520 }, { "epoch": 0.46, "learning_rate": 1.1722445683096606e-05, "loss": 24.2454, "step": 37530 }, { "epoch": 0.46, "learning_rate": 1.1718521924632824e-05, "loss": 23.7965, "step": 37540 }, { "epoch": 0.46, "learning_rate": 1.1714597893516777e-05, "loss": 23.8932, "step": 37550 }, { "epoch": 0.46, "learning_rate": 1.1710673590371035e-05, "loss": 24.0603, "step": 37560 }, { "epoch": 0.46, "learning_rate": 1.1706749015818206e-05, "loss": 23.6996, "step": 37570 }, { "epoch": 0.46, "learning_rate": 1.1702824170480942e-05, "loss": 24.1675, "step": 37580 }, { "epoch": 0.46, "learning_rate": 1.1698899054981943e-05, "loss": 24.3352, "step": 37590 }, { "epoch": 0.46, "learning_rate": 1.169497366994395e-05, "loss": 24.004, "step": 37600 }, { "epoch": 0.46, "learning_rate": 1.1691048015989739e-05, "loss": 23.5581, "step": 37610 }, { "epoch": 0.46, "learning_rate": 1.168712209374214e-05, "loss": 23.2049, "step": 37620 }, { "epoch": 0.46, "learning_rate": 1.1683195903824018e-05, "loss": 24.2446, "step": 37630 }, { "epoch": 0.46, "learning_rate": 1.1679269446858288e-05, "loss": 24.3922, "step": 37640 }, { "epoch": 0.46, "learning_rate": 1.1675342723467898e-05, "loss": 22.9526, "step": 37650 }, { "epoch": 0.46, "learning_rate": 1.167141573427584e-05, "loss": 24.6745, "step": 37660 }, { "epoch": 0.46, "learning_rate": 1.1667488479905158e-05, "loss": 23.5724, "step": 37670 }, { "epoch": 0.46, "learning_rate": 1.1663560960978928e-05, "loss": 24.0798, "step": 37680 }, { "epoch": 0.46, "learning_rate": 1.165963317812027e-05, "loss": 23.7052, "step": 37690 }, { "epoch": 0.46, "learning_rate": 1.1655705131952344e-05, "loss": 23.2121, "step": 37700 }, { "epoch": 0.46, "learning_rate": 1.165177682309836e-05, "loss": 24.2583, "step": 37710 }, { "epoch": 0.46, "learning_rate": 1.164784825218156e-05, "loss": 23.5141, "step": 37720 }, { "epoch": 0.46, "learning_rate": 1.1643919419825234e-05, "loss": 23.9092, "step": 37730 }, { "epoch": 0.46, "learning_rate": 1.1639990326652713e-05, "loss": 24.325, "step": 37740 }, { "epoch": 0.46, "learning_rate": 1.1636060973287364e-05, "loss": 23.6856, "step": 37750 }, { "epoch": 0.46, "learning_rate": 1.1632131360352598e-05, "loss": 23.8724, "step": 37760 }, { "epoch": 0.46, "learning_rate": 1.1628201488471875e-05, "loss": 23.9703, "step": 37770 }, { "epoch": 0.46, "learning_rate": 1.1624271358268681e-05, "loss": 23.6428, "step": 37780 }, { "epoch": 0.46, "learning_rate": 1.1620340970366559e-05, "loss": 23.6179, "step": 37790 }, { "epoch": 0.46, "learning_rate": 1.161641032538908e-05, "loss": 23.7398, "step": 37800 }, { "epoch": 0.46, "learning_rate": 1.1612479423959862e-05, "loss": 23.8192, "step": 37810 }, { "epoch": 0.47, "learning_rate": 1.1608548266702562e-05, "loss": 23.3878, "step": 37820 }, { "epoch": 0.47, "learning_rate": 1.1604616854240875e-05, "loss": 23.2304, "step": 37830 }, { "epoch": 0.47, "learning_rate": 1.160068518719855e-05, "loss": 23.088, "step": 37840 }, { "epoch": 0.47, "learning_rate": 1.1596753266199354e-05, "loss": 23.4121, "step": 37850 }, { "epoch": 0.47, "learning_rate": 1.1592821091867112e-05, "loss": 23.5613, "step": 37860 }, { "epoch": 0.47, "learning_rate": 1.1588888664825687e-05, "loss": 23.9144, "step": 37870 }, { "epoch": 0.47, "learning_rate": 1.1584955985698972e-05, "loss": 23.465, "step": 37880 }, { "epoch": 0.47, "learning_rate": 1.1581023055110909e-05, "loss": 23.5467, "step": 37890 }, { "epoch": 0.47, "learning_rate": 1.1577089873685479e-05, "loss": 23.7827, "step": 37900 }, { "epoch": 0.47, "learning_rate": 1.1573156442046701e-05, "loss": 23.7104, "step": 37910 }, { "epoch": 0.47, "learning_rate": 1.156922276081863e-05, "loss": 23.6757, "step": 37920 }, { "epoch": 0.47, "learning_rate": 1.156528883062537e-05, "loss": 24.2075, "step": 37930 }, { "epoch": 0.47, "learning_rate": 1.1561354652091057e-05, "loss": 23.708, "step": 37940 }, { "epoch": 0.47, "learning_rate": 1.1557420225839867e-05, "loss": 24.5865, "step": 37950 }, { "epoch": 0.47, "learning_rate": 1.1553485552496018e-05, "loss": 23.8544, "step": 37960 }, { "epoch": 0.47, "learning_rate": 1.1549550632683766e-05, "loss": 23.4193, "step": 37970 }, { "epoch": 0.47, "learning_rate": 1.1545615467027407e-05, "loss": 23.9409, "step": 37980 }, { "epoch": 0.47, "learning_rate": 1.1541680056151271e-05, "loss": 24.3498, "step": 37990 }, { "epoch": 0.47, "learning_rate": 1.1537744400679739e-05, "loss": 23.7592, "step": 38000 }, { "epoch": 0.47, "learning_rate": 1.1533808501237216e-05, "loss": 23.9651, "step": 38010 }, { "epoch": 0.47, "learning_rate": 1.1529872358448153e-05, "loss": 23.7791, "step": 38020 }, { "epoch": 0.47, "learning_rate": 1.1525935972937044e-05, "loss": 24.1185, "step": 38030 }, { "epoch": 0.47, "learning_rate": 1.152199934532841e-05, "loss": 23.801, "step": 38040 }, { "epoch": 0.47, "learning_rate": 1.1518062476246821e-05, "loss": 23.5512, "step": 38050 }, { "epoch": 0.47, "learning_rate": 1.151412536631688e-05, "loss": 24.136, "step": 38060 }, { "epoch": 0.47, "learning_rate": 1.1510188016163233e-05, "loss": 23.3588, "step": 38070 }, { "epoch": 0.47, "learning_rate": 1.1506250426410555e-05, "loss": 23.1937, "step": 38080 }, { "epoch": 0.47, "learning_rate": 1.1502312597683567e-05, "loss": 24.2212, "step": 38090 }, { "epoch": 0.47, "learning_rate": 1.1498374530607031e-05, "loss": 24.6782, "step": 38100 }, { "epoch": 0.47, "learning_rate": 1.1494436225805732e-05, "loss": 23.8361, "step": 38110 }, { "epoch": 0.47, "learning_rate": 1.1490497683904506e-05, "loss": 23.7976, "step": 38120 }, { "epoch": 0.47, "learning_rate": 1.1486558905528229e-05, "loss": 24.1594, "step": 38130 }, { "epoch": 0.47, "learning_rate": 1.1482619891301799e-05, "loss": 23.8253, "step": 38140 }, { "epoch": 0.47, "learning_rate": 1.1478680641850164e-05, "loss": 23.7605, "step": 38150 }, { "epoch": 0.47, "learning_rate": 1.1474741157798305e-05, "loss": 23.644, "step": 38160 }, { "epoch": 0.47, "learning_rate": 1.1470801439771249e-05, "loss": 24.2138, "step": 38170 }, { "epoch": 0.47, "learning_rate": 1.1466861488394041e-05, "loss": 23.807, "step": 38180 }, { "epoch": 0.47, "learning_rate": 1.146292130429178e-05, "loss": 23.4276, "step": 38190 }, { "epoch": 0.47, "learning_rate": 1.1458980888089597e-05, "loss": 23.5689, "step": 38200 }, { "epoch": 0.47, "learning_rate": 1.1455040240412654e-05, "loss": 24.0875, "step": 38210 }, { "epoch": 0.47, "learning_rate": 1.1451099361886162e-05, "loss": 24.0135, "step": 38220 }, { "epoch": 0.47, "learning_rate": 1.1447158253135357e-05, "loss": 23.7306, "step": 38230 }, { "epoch": 0.47, "learning_rate": 1.1443216914785518e-05, "loss": 23.1903, "step": 38240 }, { "epoch": 0.47, "learning_rate": 1.1439275347461955e-05, "loss": 23.6416, "step": 38250 }, { "epoch": 0.47, "learning_rate": 1.1435333551790026e-05, "loss": 23.4223, "step": 38260 }, { "epoch": 0.47, "learning_rate": 1.1431391528395108e-05, "loss": 24.316, "step": 38270 }, { "epoch": 0.47, "learning_rate": 1.1427449277902627e-05, "loss": 24.0428, "step": 38280 }, { "epoch": 0.47, "learning_rate": 1.1423506800938037e-05, "loss": 23.354, "step": 38290 }, { "epoch": 0.47, "learning_rate": 1.141956409812684e-05, "loss": 23.8394, "step": 38300 }, { "epoch": 0.47, "learning_rate": 1.141562117009456e-05, "loss": 23.9764, "step": 38310 }, { "epoch": 0.47, "learning_rate": 1.1411678017466762e-05, "loss": 23.8002, "step": 38320 }, { "epoch": 0.47, "learning_rate": 1.1407734640869052e-05, "loss": 24.0866, "step": 38330 }, { "epoch": 0.47, "learning_rate": 1.1403791040927063e-05, "loss": 23.8304, "step": 38340 }, { "epoch": 0.47, "learning_rate": 1.139984721826647e-05, "loss": 23.9169, "step": 38350 }, { "epoch": 0.47, "learning_rate": 1.1395903173512977e-05, "loss": 24.7961, "step": 38360 }, { "epoch": 0.47, "learning_rate": 1.1391958907292334e-05, "loss": 24.1092, "step": 38370 }, { "epoch": 0.47, "learning_rate": 1.1388014420230306e-05, "loss": 23.8883, "step": 38380 }, { "epoch": 0.47, "learning_rate": 1.138406971295272e-05, "loss": 24.4609, "step": 38390 }, { "epoch": 0.47, "learning_rate": 1.1380124786085412e-05, "loss": 23.3344, "step": 38400 }, { "epoch": 0.47, "learning_rate": 1.1376179640254275e-05, "loss": 24.1788, "step": 38410 }, { "epoch": 0.47, "learning_rate": 1.137223427608522e-05, "loss": 24.3482, "step": 38420 }, { "epoch": 0.47, "learning_rate": 1.1368288694204202e-05, "loss": 23.456, "step": 38430 }, { "epoch": 0.47, "learning_rate": 1.1364342895237207e-05, "loss": 23.9495, "step": 38440 }, { "epoch": 0.47, "learning_rate": 1.1360396879810254e-05, "loss": 23.5788, "step": 38450 }, { "epoch": 0.47, "learning_rate": 1.13564506485494e-05, "loss": 24.0193, "step": 38460 }, { "epoch": 0.47, "learning_rate": 1.1352504202080739e-05, "loss": 23.7449, "step": 38470 }, { "epoch": 0.47, "learning_rate": 1.1348557541030385e-05, "loss": 24.0805, "step": 38480 }, { "epoch": 0.47, "learning_rate": 1.1344610666024507e-05, "loss": 24.3437, "step": 38490 }, { "epoch": 0.47, "learning_rate": 1.1340663577689291e-05, "loss": 23.9619, "step": 38500 }, { "epoch": 0.47, "learning_rate": 1.133671627665096e-05, "loss": 23.9532, "step": 38510 }, { "epoch": 0.47, "learning_rate": 1.133276876353578e-05, "loss": 24.1841, "step": 38520 }, { "epoch": 0.47, "learning_rate": 1.132882103897004e-05, "loss": 23.7386, "step": 38530 }, { "epoch": 0.47, "learning_rate": 1.1324873103580068e-05, "loss": 23.4014, "step": 38540 }, { "epoch": 0.47, "learning_rate": 1.1320924957992221e-05, "loss": 24.7298, "step": 38550 }, { "epoch": 0.47, "learning_rate": 1.1316976602832896e-05, "loss": 23.8465, "step": 38560 }, { "epoch": 0.47, "learning_rate": 1.131302803872852e-05, "loss": 23.2258, "step": 38570 }, { "epoch": 0.47, "learning_rate": 1.130907926630555e-05, "loss": 23.8069, "step": 38580 }, { "epoch": 0.47, "learning_rate": 1.130513028619048e-05, "loss": 24.2743, "step": 38590 }, { "epoch": 0.47, "learning_rate": 1.130118109900984e-05, "loss": 24.2001, "step": 38600 }, { "epoch": 0.47, "learning_rate": 1.1297231705390184e-05, "loss": 23.8241, "step": 38610 }, { "epoch": 0.47, "learning_rate": 1.1293282105958099e-05, "loss": 23.5746, "step": 38620 }, { "epoch": 0.48, "learning_rate": 1.1289332301340217e-05, "loss": 24.0988, "step": 38630 }, { "epoch": 0.48, "learning_rate": 1.1285382292163196e-05, "loss": 23.5555, "step": 38640 }, { "epoch": 0.48, "learning_rate": 1.1281432079053716e-05, "loss": 23.4488, "step": 38650 }, { "epoch": 0.48, "learning_rate": 1.1277481662638506e-05, "loss": 23.3934, "step": 38660 }, { "epoch": 0.48, "learning_rate": 1.1273531043544317e-05, "loss": 23.4946, "step": 38670 }, { "epoch": 0.48, "learning_rate": 1.1269580222397933e-05, "loss": 23.3409, "step": 38680 }, { "epoch": 0.48, "learning_rate": 1.1265629199826176e-05, "loss": 23.4529, "step": 38690 }, { "epoch": 0.48, "learning_rate": 1.1261677976455897e-05, "loss": 23.4797, "step": 38700 }, { "epoch": 0.48, "learning_rate": 1.1257726552913968e-05, "loss": 23.8816, "step": 38710 }, { "epoch": 0.48, "learning_rate": 1.1253774929827315e-05, "loss": 23.8872, "step": 38720 }, { "epoch": 0.48, "learning_rate": 1.1249823107822874e-05, "loss": 24.4635, "step": 38730 }, { "epoch": 0.48, "learning_rate": 1.1245871087527627e-05, "loss": 23.0103, "step": 38740 }, { "epoch": 0.48, "learning_rate": 1.1241918869568578e-05, "loss": 24.0235, "step": 38750 }, { "epoch": 0.48, "learning_rate": 1.1237966454572769e-05, "loss": 23.958, "step": 38760 }, { "epoch": 0.48, "learning_rate": 1.1234013843167274e-05, "loss": 24.0413, "step": 38770 }, { "epoch": 0.48, "learning_rate": 1.1230061035979188e-05, "loss": 23.694, "step": 38780 }, { "epoch": 0.48, "learning_rate": 1.1226108033635645e-05, "loss": 23.8008, "step": 38790 }, { "epoch": 0.48, "learning_rate": 1.1222154836763816e-05, "loss": 24.1541, "step": 38800 }, { "epoch": 0.48, "learning_rate": 1.121820144599089e-05, "loss": 23.84, "step": 38810 }, { "epoch": 0.48, "learning_rate": 1.1214247861944092e-05, "loss": 23.7979, "step": 38820 }, { "epoch": 0.48, "learning_rate": 1.1210294085250683e-05, "loss": 24.1576, "step": 38830 }, { "epoch": 0.48, "learning_rate": 1.1206340116537948e-05, "loss": 24.2088, "step": 38840 }, { "epoch": 0.48, "learning_rate": 1.1202385956433199e-05, "loss": 23.9596, "step": 38850 }, { "epoch": 0.48, "learning_rate": 1.119843160556379e-05, "loss": 24.2899, "step": 38860 }, { "epoch": 0.48, "learning_rate": 1.1194477064557099e-05, "loss": 24.0436, "step": 38870 }, { "epoch": 0.48, "learning_rate": 1.1190522334040529e-05, "loss": 24.1477, "step": 38880 }, { "epoch": 0.48, "learning_rate": 1.118656741464152e-05, "loss": 23.7995, "step": 38890 }, { "epoch": 0.48, "learning_rate": 1.1182612306987544e-05, "loss": 23.7365, "step": 38900 }, { "epoch": 0.48, "learning_rate": 1.1178657011706096e-05, "loss": 24.0298, "step": 38910 }, { "epoch": 0.48, "learning_rate": 1.11747015294247e-05, "loss": 23.1068, "step": 38920 }, { "epoch": 0.48, "learning_rate": 1.1170745860770925e-05, "loss": 23.958, "step": 38930 }, { "epoch": 0.48, "learning_rate": 1.1166790006372343e-05, "loss": 23.9841, "step": 38940 }, { "epoch": 0.48, "learning_rate": 1.1162833966856584e-05, "loss": 23.9439, "step": 38950 }, { "epoch": 0.48, "learning_rate": 1.1158877742851282e-05, "loss": 23.0448, "step": 38960 }, { "epoch": 0.48, "learning_rate": 1.115492133498412e-05, "loss": 23.9433, "step": 38970 }, { "epoch": 0.48, "learning_rate": 1.1150964743882802e-05, "loss": 23.8886, "step": 38980 }, { "epoch": 0.48, "learning_rate": 1.1147007970175057e-05, "loss": 23.5041, "step": 38990 }, { "epoch": 0.48, "learning_rate": 1.114305101448865e-05, "loss": 24.0413, "step": 39000 }, { "epoch": 0.48, "learning_rate": 1.1139093877451373e-05, "loss": 23.6883, "step": 39010 }, { "epoch": 0.48, "learning_rate": 1.1135136559691045e-05, "loss": 23.9561, "step": 39020 }, { "epoch": 0.48, "learning_rate": 1.1131179061835513e-05, "loss": 23.5807, "step": 39030 }, { "epoch": 0.48, "learning_rate": 1.1127221384512656e-05, "loss": 23.771, "step": 39040 }, { "epoch": 0.48, "learning_rate": 1.1123263528350382e-05, "loss": 24.1328, "step": 39050 }, { "epoch": 0.48, "learning_rate": 1.1119305493976617e-05, "loss": 24.3236, "step": 39060 }, { "epoch": 0.48, "learning_rate": 1.1115347282019337e-05, "loss": 23.6357, "step": 39070 }, { "epoch": 0.48, "learning_rate": 1.111138889310652e-05, "loss": 24.2464, "step": 39080 }, { "epoch": 0.48, "learning_rate": 1.110743032786619e-05, "loss": 24.1558, "step": 39090 }, { "epoch": 0.48, "learning_rate": 1.1103471586926389e-05, "loss": 24.1245, "step": 39100 }, { "epoch": 0.48, "learning_rate": 1.1099512670915198e-05, "loss": 23.3239, "step": 39110 }, { "epoch": 0.48, "learning_rate": 1.1095553580460715e-05, "loss": 24.493, "step": 39120 }, { "epoch": 0.48, "learning_rate": 1.1091594316191073e-05, "loss": 23.8083, "step": 39130 }, { "epoch": 0.48, "learning_rate": 1.1087634878734426e-05, "loss": 23.5198, "step": 39140 }, { "epoch": 0.48, "learning_rate": 1.1083675268718958e-05, "loss": 24.4294, "step": 39150 }, { "epoch": 0.48, "learning_rate": 1.1079715486772887e-05, "loss": 23.776, "step": 39160 }, { "epoch": 0.48, "learning_rate": 1.1075755533524451e-05, "loss": 23.4233, "step": 39170 }, { "epoch": 0.48, "learning_rate": 1.1071795409601914e-05, "loss": 23.2495, "step": 39180 }, { "epoch": 0.48, "learning_rate": 1.1067835115633567e-05, "loss": 23.9904, "step": 39190 }, { "epoch": 0.48, "learning_rate": 1.1063874652247738e-05, "loss": 24.2784, "step": 39200 }, { "epoch": 0.48, "learning_rate": 1.105991402007277e-05, "loss": 23.7697, "step": 39210 }, { "epoch": 0.48, "learning_rate": 1.1055953219737038e-05, "loss": 24.4775, "step": 39220 }, { "epoch": 0.48, "learning_rate": 1.1051992251868948e-05, "loss": 23.3916, "step": 39230 }, { "epoch": 0.48, "learning_rate": 1.1048031117096923e-05, "loss": 23.9542, "step": 39240 }, { "epoch": 0.48, "learning_rate": 1.1044069816049417e-05, "loss": 24.0063, "step": 39250 }, { "epoch": 0.48, "learning_rate": 1.104010834935491e-05, "loss": 24.0107, "step": 39260 }, { "epoch": 0.48, "learning_rate": 1.1036146717641914e-05, "loss": 23.5062, "step": 39270 }, { "epoch": 0.48, "learning_rate": 1.1032184921538955e-05, "loss": 24.2059, "step": 39280 }, { "epoch": 0.48, "learning_rate": 1.1028222961674597e-05, "loss": 23.7967, "step": 39290 }, { "epoch": 0.48, "learning_rate": 1.1024260838677427e-05, "loss": 23.5552, "step": 39300 }, { "epoch": 0.48, "learning_rate": 1.1020298553176052e-05, "loss": 24.0071, "step": 39310 }, { "epoch": 0.48, "learning_rate": 1.1016336105799108e-05, "loss": 23.3609, "step": 39320 }, { "epoch": 0.48, "learning_rate": 1.101237349717526e-05, "loss": 23.5764, "step": 39330 }, { "epoch": 0.48, "learning_rate": 1.1008410727933195e-05, "loss": 24.4251, "step": 39340 }, { "epoch": 0.48, "learning_rate": 1.1004447798701624e-05, "loss": 22.9785, "step": 39350 }, { "epoch": 0.48, "learning_rate": 1.1000484710109294e-05, "loss": 24.3565, "step": 39360 }, { "epoch": 0.48, "learning_rate": 1.0996521462784961e-05, "loss": 23.9154, "step": 39370 }, { "epoch": 0.48, "learning_rate": 1.099255805735742e-05, "loss": 23.4217, "step": 39380 }, { "epoch": 0.48, "learning_rate": 1.0988594494455478e-05, "loss": 23.6398, "step": 39390 }, { "epoch": 0.48, "learning_rate": 1.0984630774707986e-05, "loss": 24.058, "step": 39400 }, { "epoch": 0.48, "learning_rate": 1.0980666898743795e-05, "loss": 24.4347, "step": 39410 }, { "epoch": 0.48, "learning_rate": 1.0976702867191804e-05, "loss": 24.2015, "step": 39420 }, { "epoch": 0.48, "learning_rate": 1.0972738680680923e-05, "loss": 24.1327, "step": 39430 }, { "epoch": 0.49, "learning_rate": 1.0968774339840094e-05, "loss": 23.5232, "step": 39440 }, { "epoch": 0.49, "learning_rate": 1.0964809845298271e-05, "loss": 23.7601, "step": 39450 }, { "epoch": 0.49, "learning_rate": 1.096084519768445e-05, "loss": 23.2892, "step": 39460 }, { "epoch": 0.49, "learning_rate": 1.0956880397627642e-05, "loss": 24.0543, "step": 39470 }, { "epoch": 0.49, "learning_rate": 1.095291544575688e-05, "loss": 24.6009, "step": 39480 }, { "epoch": 0.49, "learning_rate": 1.094895034270122e-05, "loss": 23.9254, "step": 39490 }, { "epoch": 0.49, "learning_rate": 1.0944985089089755e-05, "loss": 24.1814, "step": 39500 }, { "epoch": 0.49, "learning_rate": 1.0941019685551585e-05, "loss": 23.6929, "step": 39510 }, { "epoch": 0.49, "learning_rate": 1.0937054132715846e-05, "loss": 23.4269, "step": 39520 }, { "epoch": 0.49, "learning_rate": 1.0933088431211691e-05, "loss": 23.6907, "step": 39530 }, { "epoch": 0.49, "learning_rate": 1.0929122581668301e-05, "loss": 23.6041, "step": 39540 }, { "epoch": 0.49, "learning_rate": 1.0925156584714873e-05, "loss": 24.6126, "step": 39550 }, { "epoch": 0.49, "learning_rate": 1.0921190440980639e-05, "loss": 24.3848, "step": 39560 }, { "epoch": 0.49, "learning_rate": 1.0917224151094843e-05, "loss": 23.865, "step": 39570 }, { "epoch": 0.49, "learning_rate": 1.0913257715686758e-05, "loss": 24.5884, "step": 39580 }, { "epoch": 0.49, "learning_rate": 1.0909291135385682e-05, "loss": 23.3132, "step": 39590 }, { "epoch": 0.49, "learning_rate": 1.0905324410820931e-05, "loss": 24.3404, "step": 39600 }, { "epoch": 0.49, "learning_rate": 1.0901357542621848e-05, "loss": 23.9827, "step": 39610 }, { "epoch": 0.49, "learning_rate": 1.0897390531417794e-05, "loss": 24.0069, "step": 39620 }, { "epoch": 0.49, "learning_rate": 1.089342337783816e-05, "loss": 23.6634, "step": 39630 }, { "epoch": 0.49, "learning_rate": 1.0889456082512348e-05, "loss": 23.473, "step": 39640 }, { "epoch": 0.49, "learning_rate": 1.0885488646069798e-05, "loss": 23.6513, "step": 39650 }, { "epoch": 0.49, "learning_rate": 1.0881521069139957e-05, "loss": 24.3062, "step": 39660 }, { "epoch": 0.49, "learning_rate": 1.0877553352352306e-05, "loss": 22.9707, "step": 39670 }, { "epoch": 0.49, "learning_rate": 1.0873585496336341e-05, "loss": 24.2835, "step": 39680 }, { "epoch": 0.49, "learning_rate": 1.086961750172158e-05, "loss": 23.4644, "step": 39690 }, { "epoch": 0.49, "learning_rate": 1.0865649369137574e-05, "loss": 24.2929, "step": 39700 }, { "epoch": 0.49, "learning_rate": 1.0861681099213878e-05, "loss": 23.6222, "step": 39710 }, { "epoch": 0.49, "learning_rate": 1.0857712692580085e-05, "loss": 24.0589, "step": 39720 }, { "epoch": 0.49, "learning_rate": 1.0853744149865796e-05, "loss": 23.2564, "step": 39730 }, { "epoch": 0.49, "learning_rate": 1.0849775471700652e-05, "loss": 24.2662, "step": 39740 }, { "epoch": 0.49, "learning_rate": 1.0845806658714293e-05, "loss": 23.4475, "step": 39750 }, { "epoch": 0.49, "learning_rate": 1.0841837711536398e-05, "loss": 23.3844, "step": 39760 }, { "epoch": 0.49, "learning_rate": 1.0837868630796658e-05, "loss": 24.0411, "step": 39770 }, { "epoch": 0.49, "learning_rate": 1.0833899417124785e-05, "loss": 24.2842, "step": 39780 }, { "epoch": 0.49, "learning_rate": 1.0829930071150518e-05, "loss": 24.2848, "step": 39790 }, { "epoch": 0.49, "learning_rate": 1.0825960593503618e-05, "loss": 23.8851, "step": 39800 }, { "epoch": 0.49, "learning_rate": 1.0821990984813856e-05, "loss": 23.3159, "step": 39810 }, { "epoch": 0.49, "learning_rate": 1.081802124571103e-05, "loss": 23.7797, "step": 39820 }, { "epoch": 0.49, "learning_rate": 1.0814051376824968e-05, "loss": 24.1005, "step": 39830 }, { "epoch": 0.49, "learning_rate": 1.0810081378785504e-05, "loss": 23.8019, "step": 39840 }, { "epoch": 0.49, "learning_rate": 1.0806111252222498e-05, "loss": 23.4501, "step": 39850 }, { "epoch": 0.49, "learning_rate": 1.080214099776583e-05, "loss": 23.9617, "step": 39860 }, { "epoch": 0.49, "learning_rate": 1.079817061604541e-05, "loss": 23.7484, "step": 39870 }, { "epoch": 0.49, "learning_rate": 1.0794200107691148e-05, "loss": 24.1736, "step": 39880 }, { "epoch": 0.49, "learning_rate": 1.079022947333299e-05, "loss": 24.463, "step": 39890 }, { "epoch": 0.49, "learning_rate": 1.0786258713600896e-05, "loss": 24.2219, "step": 39900 }, { "epoch": 0.49, "learning_rate": 1.078228782912485e-05, "loss": 23.5593, "step": 39910 }, { "epoch": 0.49, "learning_rate": 1.0778316820534853e-05, "loss": 23.9327, "step": 39920 }, { "epoch": 0.49, "learning_rate": 1.0774345688460921e-05, "loss": 24.1189, "step": 39930 }, { "epoch": 0.49, "learning_rate": 1.0770374433533102e-05, "loss": 23.9515, "step": 39940 }, { "epoch": 0.49, "learning_rate": 1.0766403056381448e-05, "loss": 24.0566, "step": 39950 }, { "epoch": 0.49, "learning_rate": 1.0762431557636042e-05, "loss": 23.4111, "step": 39960 }, { "epoch": 0.49, "learning_rate": 1.0758459937926984e-05, "loss": 23.5273, "step": 39970 }, { "epoch": 0.49, "learning_rate": 1.075448819788439e-05, "loss": 23.3692, "step": 39980 }, { "epoch": 0.49, "learning_rate": 1.0750516338138397e-05, "loss": 23.5833, "step": 39990 }, { "epoch": 0.49, "learning_rate": 1.0746544359319157e-05, "loss": 24.4243, "step": 40000 }, { "epoch": 0.49, "learning_rate": 1.0742572262056854e-05, "loss": 23.8471, "step": 40010 }, { "epoch": 0.49, "learning_rate": 1.0738600046981671e-05, "loss": 23.6798, "step": 40020 }, { "epoch": 0.49, "learning_rate": 1.0734627714723827e-05, "loss": 23.5098, "step": 40030 }, { "epoch": 0.49, "learning_rate": 1.0730655265913552e-05, "loss": 23.5659, "step": 40040 }, { "epoch": 0.49, "learning_rate": 1.072668270118109e-05, "loss": 23.515, "step": 40050 }, { "epoch": 0.49, "learning_rate": 1.0722710021156715e-05, "loss": 23.9481, "step": 40060 }, { "epoch": 0.49, "learning_rate": 1.0718737226470712e-05, "loss": 23.7668, "step": 40070 }, { "epoch": 0.49, "learning_rate": 1.0714764317753381e-05, "loss": 23.0312, "step": 40080 }, { "epoch": 0.49, "learning_rate": 1.0710791295635047e-05, "loss": 23.7712, "step": 40090 }, { "epoch": 0.49, "learning_rate": 1.0706818160746053e-05, "loss": 23.5425, "step": 40100 }, { "epoch": 0.49, "learning_rate": 1.0702844913716755e-05, "loss": 23.4535, "step": 40110 }, { "epoch": 0.49, "learning_rate": 1.0698871555177524e-05, "loss": 23.1279, "step": 40120 }, { "epoch": 0.49, "learning_rate": 1.0694898085758758e-05, "loss": 23.4742, "step": 40130 }, { "epoch": 0.49, "learning_rate": 1.0690924506090872e-05, "loss": 23.7466, "step": 40140 }, { "epoch": 0.49, "learning_rate": 1.0686950816804287e-05, "loss": 23.4834, "step": 40150 }, { "epoch": 0.49, "learning_rate": 1.0682977018529452e-05, "loss": 23.9055, "step": 40160 }, { "epoch": 0.49, "learning_rate": 1.0679003111896832e-05, "loss": 24.1252, "step": 40170 }, { "epoch": 0.49, "learning_rate": 1.0675029097536909e-05, "loss": 23.772, "step": 40180 }, { "epoch": 0.49, "learning_rate": 1.0671054976080175e-05, "loss": 24.0372, "step": 40190 }, { "epoch": 0.49, "learning_rate": 1.066708074815715e-05, "loss": 23.696, "step": 40200 }, { "epoch": 0.49, "learning_rate": 1.0663106414398362e-05, "loss": 23.8317, "step": 40210 }, { "epoch": 0.49, "learning_rate": 1.065913197543436e-05, "loss": 23.9829, "step": 40220 }, { "epoch": 0.49, "learning_rate": 1.0655157431895713e-05, "loss": 23.8872, "step": 40230 }, { "epoch": 0.49, "learning_rate": 1.0651182784412997e-05, "loss": 23.7317, "step": 40240 }, { "epoch": 0.5, "learning_rate": 1.0647208033616812e-05, "loss": 23.4892, "step": 40250 }, { "epoch": 0.5, "learning_rate": 1.0643233180137774e-05, "loss": 23.8621, "step": 40260 }, { "epoch": 0.5, "learning_rate": 1.0639258224606512e-05, "loss": 23.9712, "step": 40270 }, { "epoch": 0.5, "learning_rate": 1.0635283167653672e-05, "loss": 23.9856, "step": 40280 }, { "epoch": 0.5, "learning_rate": 1.0631308009909919e-05, "loss": 23.6536, "step": 40290 }, { "epoch": 0.5, "learning_rate": 1.0627332752005934e-05, "loss": 23.2737, "step": 40300 }, { "epoch": 0.5, "learning_rate": 1.0623357394572407e-05, "loss": 23.7609, "step": 40310 }, { "epoch": 0.5, "learning_rate": 1.0619381938240052e-05, "loss": 24.3744, "step": 40320 }, { "epoch": 0.5, "learning_rate": 1.0615406383639597e-05, "loss": 23.5601, "step": 40330 }, { "epoch": 0.5, "learning_rate": 1.0611430731401781e-05, "loss": 23.7857, "step": 40340 }, { "epoch": 0.5, "learning_rate": 1.0607454982157361e-05, "loss": 23.8758, "step": 40350 }, { "epoch": 0.5, "learning_rate": 1.0603479136537109e-05, "loss": 23.9869, "step": 40360 }, { "epoch": 0.5, "learning_rate": 1.0599503195171818e-05, "loss": 23.7821, "step": 40370 }, { "epoch": 0.5, "learning_rate": 1.059552715869229e-05, "loss": 24.0078, "step": 40380 }, { "epoch": 0.5, "learning_rate": 1.059155102772934e-05, "loss": 24.995, "step": 40390 }, { "epoch": 0.5, "learning_rate": 1.0587574802913802e-05, "loss": 24.1677, "step": 40400 }, { "epoch": 0.5, "learning_rate": 1.0583598484876529e-05, "loss": 24.1616, "step": 40410 }, { "epoch": 0.5, "learning_rate": 1.0579622074248379e-05, "loss": 23.5454, "step": 40420 }, { "epoch": 0.5, "learning_rate": 1.0575645571660231e-05, "loss": 23.7807, "step": 40430 }, { "epoch": 0.5, "learning_rate": 1.0571668977742982e-05, "loss": 24.2551, "step": 40440 }, { "epoch": 0.5, "learning_rate": 1.0567692293127532e-05, "loss": 24.0898, "step": 40450 }, { "epoch": 0.5, "learning_rate": 1.0563715518444807e-05, "loss": 23.9845, "step": 40460 }, { "epoch": 0.5, "learning_rate": 1.0559738654325742e-05, "loss": 24.1845, "step": 40470 }, { "epoch": 0.5, "learning_rate": 1.0555761701401283e-05, "loss": 23.9393, "step": 40480 }, { "epoch": 0.5, "learning_rate": 1.0551784660302397e-05, "loss": 23.7892, "step": 40490 }, { "epoch": 0.5, "learning_rate": 1.0547807531660063e-05, "loss": 23.8557, "step": 40500 }, { "epoch": 0.5, "learning_rate": 1.054383031610527e-05, "loss": 24.0282, "step": 40510 }, { "epoch": 0.5, "learning_rate": 1.0539853014269026e-05, "loss": 23.8697, "step": 40520 }, { "epoch": 0.5, "learning_rate": 1.0535875626782346e-05, "loss": 23.9087, "step": 40530 }, { "epoch": 0.5, "learning_rate": 1.053189815427627e-05, "loss": 23.9995, "step": 40540 }, { "epoch": 0.5, "learning_rate": 1.0527920597381837e-05, "loss": 23.9346, "step": 40550 }, { "epoch": 0.5, "learning_rate": 1.0523942956730112e-05, "loss": 24.2111, "step": 40560 }, { "epoch": 0.5, "learning_rate": 1.0519965232952167e-05, "loss": 23.9061, "step": 40570 }, { "epoch": 0.5, "learning_rate": 1.0515987426679085e-05, "loss": 23.7935, "step": 40580 }, { "epoch": 0.5, "learning_rate": 1.0512009538541966e-05, "loss": 23.8677, "step": 40590 }, { "epoch": 0.5, "learning_rate": 1.0508031569171922e-05, "loss": 23.8205, "step": 40600 }, { "epoch": 0.5, "learning_rate": 1.0504053519200081e-05, "loss": 23.468, "step": 40610 }, { "epoch": 0.5, "learning_rate": 1.050007538925758e-05, "loss": 23.4372, "step": 40620 }, { "epoch": 0.5, "learning_rate": 1.0496097179975568e-05, "loss": 23.5012, "step": 40630 }, { "epoch": 0.5, "learning_rate": 1.049211889198521e-05, "loss": 23.5162, "step": 40640 }, { "epoch": 0.5, "learning_rate": 1.048814052591768e-05, "loss": 24.0362, "step": 40650 }, { "epoch": 0.5, "learning_rate": 1.0484162082404168e-05, "loss": 24.1989, "step": 40660 }, { "epoch": 0.5, "learning_rate": 1.0480183562075873e-05, "loss": 24.6432, "step": 40670 }, { "epoch": 0.5, "learning_rate": 1.0476204965564008e-05, "loss": 24.2827, "step": 40680 }, { "epoch": 0.5, "learning_rate": 1.0472226293499798e-05, "loss": 23.3304, "step": 40690 }, { "epoch": 0.5, "learning_rate": 1.0468247546514476e-05, "loss": 24.2096, "step": 40700 }, { "epoch": 0.5, "learning_rate": 1.0464268725239298e-05, "loss": 23.3175, "step": 40710 }, { "epoch": 0.5, "learning_rate": 1.0460289830305517e-05, "loss": 24.3141, "step": 40720 }, { "epoch": 0.5, "learning_rate": 1.0456310862344405e-05, "loss": 23.1555, "step": 40730 }, { "epoch": 0.5, "learning_rate": 1.0452331821987253e-05, "loss": 23.526, "step": 40740 }, { "epoch": 0.5, "learning_rate": 1.044835270986535e-05, "loss": 24.3801, "step": 40750 }, { "epoch": 0.5, "learning_rate": 1.0444373526610003e-05, "loss": 23.4672, "step": 40760 }, { "epoch": 0.5, "learning_rate": 1.0440394272852532e-05, "loss": 24.192, "step": 40770 }, { "epoch": 0.5, "learning_rate": 1.0436414949224263e-05, "loss": 23.7939, "step": 40780 }, { "epoch": 0.5, "learning_rate": 1.0432435556356535e-05, "loss": 23.5136, "step": 40790 }, { "epoch": 0.5, "learning_rate": 1.0428456094880705e-05, "loss": 23.9934, "step": 40800 }, { "epoch": 0.5, "learning_rate": 1.0424476565428132e-05, "loss": 24.3099, "step": 40810 }, { "epoch": 0.5, "learning_rate": 1.0420496968630185e-05, "loss": 24.0135, "step": 40820 }, { "epoch": 0.5, "learning_rate": 1.041651730511825e-05, "loss": 24.0393, "step": 40830 }, { "epoch": 0.5, "learning_rate": 1.0412537575523722e-05, "loss": 23.4108, "step": 40840 }, { "epoch": 0.5, "learning_rate": 1.0408557780478004e-05, "loss": 23.5111, "step": 40850 }, { "epoch": 0.5, "learning_rate": 1.0404577920612511e-05, "loss": 24.1862, "step": 40860 }, { "epoch": 0.5, "learning_rate": 1.040059799655867e-05, "loss": 23.8732, "step": 40870 }, { "epoch": 0.5, "learning_rate": 1.0396618008947913e-05, "loss": 24.3903, "step": 40880 }, { "epoch": 0.5, "learning_rate": 1.0392637958411687e-05, "loss": 24.0018, "step": 40890 }, { "epoch": 0.5, "learning_rate": 1.0388657845581446e-05, "loss": 24.1748, "step": 40900 }, { "epoch": 0.5, "learning_rate": 1.0384677671088659e-05, "loss": 23.7682, "step": 40910 }, { "epoch": 0.5, "learning_rate": 1.0380697435564798e-05, "loss": 24.5061, "step": 40920 }, { "epoch": 0.5, "learning_rate": 1.0376717139641345e-05, "loss": 24.0853, "step": 40930 }, { "epoch": 0.5, "learning_rate": 1.03727367839498e-05, "loss": 24.3778, "step": 40940 }, { "epoch": 0.5, "learning_rate": 1.0368756369121662e-05, "loss": 23.6643, "step": 40950 }, { "epoch": 0.5, "learning_rate": 1.0364775895788446e-05, "loss": 23.5836, "step": 40960 }, { "epoch": 0.5, "learning_rate": 1.0360795364581677e-05, "loss": 23.2408, "step": 40970 }, { "epoch": 0.5, "learning_rate": 1.0356814776132882e-05, "loss": 23.8426, "step": 40980 }, { "epoch": 0.5, "learning_rate": 1.0352834131073602e-05, "loss": 23.3843, "step": 40990 }, { "epoch": 0.5, "learning_rate": 1.034885343003539e-05, "loss": 23.2972, "step": 41000 }, { "epoch": 0.5, "learning_rate": 1.0344872673649804e-05, "loss": 24.7323, "step": 41010 }, { "epoch": 0.5, "learning_rate": 1.0340891862548407e-05, "loss": 24.0788, "step": 41020 }, { "epoch": 0.5, "learning_rate": 1.0336910997362781e-05, "loss": 24.3944, "step": 41030 }, { "epoch": 0.5, "learning_rate": 1.0332930078724509e-05, "loss": 23.8251, "step": 41040 }, { "epoch": 0.5, "learning_rate": 1.0328949107265179e-05, "loss": 23.7393, "step": 41050 }, { "epoch": 0.5, "learning_rate": 1.0324968083616395e-05, "loss": 24.2599, "step": 41060 }, { "epoch": 0.51, "learning_rate": 1.0320987008409773e-05, "loss": 23.5593, "step": 41070 }, { "epoch": 0.51, "learning_rate": 1.0317005882276923e-05, "loss": 23.2997, "step": 41080 }, { "epoch": 0.51, "learning_rate": 1.031302470584947e-05, "loss": 23.8649, "step": 41090 }, { "epoch": 0.51, "learning_rate": 1.0309043479759058e-05, "loss": 24.2754, "step": 41100 }, { "epoch": 0.51, "learning_rate": 1.0305062204637317e-05, "loss": 24.0072, "step": 41110 }, { "epoch": 0.51, "learning_rate": 1.0301080881115904e-05, "loss": 23.4927, "step": 41120 }, { "epoch": 0.51, "learning_rate": 1.0297099509826475e-05, "loss": 24.1133, "step": 41130 }, { "epoch": 0.51, "learning_rate": 1.0293118091400693e-05, "loss": 23.3718, "step": 41140 }, { "epoch": 0.51, "learning_rate": 1.0289136626470233e-05, "loss": 24.256, "step": 41150 }, { "epoch": 0.51, "learning_rate": 1.0285155115666772e-05, "loss": 23.323, "step": 41160 }, { "epoch": 0.51, "learning_rate": 1.0281173559621997e-05, "loss": 23.5737, "step": 41170 }, { "epoch": 0.51, "learning_rate": 1.0277191958967605e-05, "loss": 23.86, "step": 41180 }, { "epoch": 0.51, "learning_rate": 1.0273210314335293e-05, "loss": 23.4969, "step": 41190 }, { "epoch": 0.51, "learning_rate": 1.0269228626356775e-05, "loss": 24.1172, "step": 41200 }, { "epoch": 0.51, "learning_rate": 1.026524689566376e-05, "loss": 23.8661, "step": 41210 }, { "epoch": 0.51, "learning_rate": 1.0261265122887974e-05, "loss": 24.1416, "step": 41220 }, { "epoch": 0.51, "learning_rate": 1.0257283308661143e-05, "loss": 24.3981, "step": 41230 }, { "epoch": 0.51, "learning_rate": 1.0253301453615005e-05, "loss": 23.7152, "step": 41240 }, { "epoch": 0.51, "learning_rate": 1.0249319558381299e-05, "loss": 24.3285, "step": 41250 }, { "epoch": 0.51, "learning_rate": 1.0245337623591774e-05, "loss": 22.9003, "step": 41260 }, { "epoch": 0.51, "learning_rate": 1.0241355649878189e-05, "loss": 23.4047, "step": 41270 }, { "epoch": 0.51, "learning_rate": 1.0237373637872294e-05, "loss": 24.1388, "step": 41280 }, { "epoch": 0.51, "learning_rate": 1.0233391588205862e-05, "loss": 23.8218, "step": 41290 }, { "epoch": 0.51, "learning_rate": 1.0229409501510665e-05, "loss": 24.1252, "step": 41300 }, { "epoch": 0.51, "learning_rate": 1.0225427378418482e-05, "loss": 24.2071, "step": 41310 }, { "epoch": 0.51, "learning_rate": 1.0221445219561094e-05, "loss": 24.3197, "step": 41320 }, { "epoch": 0.51, "learning_rate": 1.0217463025570294e-05, "loss": 23.649, "step": 41330 }, { "epoch": 0.51, "learning_rate": 1.0213480797077876e-05, "loss": 24.4831, "step": 41340 }, { "epoch": 0.51, "learning_rate": 1.0209498534715642e-05, "loss": 23.8642, "step": 41350 }, { "epoch": 0.51, "learning_rate": 1.0205516239115396e-05, "loss": 23.8495, "step": 41360 }, { "epoch": 0.51, "learning_rate": 1.0201533910908954e-05, "loss": 24.0674, "step": 41370 }, { "epoch": 0.51, "learning_rate": 1.0197551550728128e-05, "loss": 24.0706, "step": 41380 }, { "epoch": 0.51, "learning_rate": 1.0193569159204742e-05, "loss": 24.3472, "step": 41390 }, { "epoch": 0.51, "learning_rate": 1.0189586736970618e-05, "loss": 24.2866, "step": 41400 }, { "epoch": 0.51, "learning_rate": 1.0185604284657598e-05, "loss": 23.9918, "step": 41410 }, { "epoch": 0.51, "learning_rate": 1.0181621802897506e-05, "loss": 23.9148, "step": 41420 }, { "epoch": 0.51, "learning_rate": 1.0177639292322192e-05, "loss": 23.9249, "step": 41430 }, { "epoch": 0.51, "learning_rate": 1.0173656753563499e-05, "loss": 23.9455, "step": 41440 }, { "epoch": 0.51, "learning_rate": 1.0169674187253276e-05, "loss": 24.0297, "step": 41450 }, { "epoch": 0.51, "learning_rate": 1.0165691594023374e-05, "loss": 24.0002, "step": 41460 }, { "epoch": 0.51, "learning_rate": 1.0161708974505661e-05, "loss": 24.3063, "step": 41470 }, { "epoch": 0.51, "learning_rate": 1.0157726329331994e-05, "loss": 23.3576, "step": 41480 }, { "epoch": 0.51, "learning_rate": 1.0153743659134241e-05, "loss": 23.9637, "step": 41490 }, { "epoch": 0.51, "learning_rate": 1.0149760964544274e-05, "loss": 23.8398, "step": 41500 }, { "epoch": 0.51, "learning_rate": 1.0145778246193965e-05, "loss": 24.2768, "step": 41510 }, { "epoch": 0.51, "learning_rate": 1.0141795504715191e-05, "loss": 24.074, "step": 41520 }, { "epoch": 0.51, "learning_rate": 1.0137812740739839e-05, "loss": 24.0394, "step": 41530 }, { "epoch": 0.51, "learning_rate": 1.0133829954899797e-05, "loss": 24.3959, "step": 41540 }, { "epoch": 0.51, "learning_rate": 1.0129847147826946e-05, "loss": 23.8461, "step": 41550 }, { "epoch": 0.51, "learning_rate": 1.0125864320153183e-05, "loss": 23.3837, "step": 41560 }, { "epoch": 0.51, "learning_rate": 1.0121881472510404e-05, "loss": 24.0701, "step": 41570 }, { "epoch": 0.51, "learning_rate": 1.011789860553051e-05, "loss": 23.6596, "step": 41580 }, { "epoch": 0.51, "learning_rate": 1.01139157198454e-05, "loss": 23.7149, "step": 41590 }, { "epoch": 0.51, "learning_rate": 1.010993281608698e-05, "loss": 23.8201, "step": 41600 }, { "epoch": 0.51, "learning_rate": 1.0105949894887161e-05, "loss": 23.2169, "step": 41610 }, { "epoch": 0.51, "learning_rate": 1.0101966956877846e-05, "loss": 24.4608, "step": 41620 }, { "epoch": 0.51, "learning_rate": 1.0097984002690958e-05, "loss": 24.194, "step": 41630 }, { "epoch": 0.51, "learning_rate": 1.0094001032958407e-05, "loss": 23.9296, "step": 41640 }, { "epoch": 0.51, "learning_rate": 1.009001804831211e-05, "loss": 24.0168, "step": 41650 }, { "epoch": 0.51, "learning_rate": 1.0086035049383993e-05, "loss": 23.585, "step": 41660 }, { "epoch": 0.51, "learning_rate": 1.0082052036805979e-05, "loss": 23.8833, "step": 41670 }, { "epoch": 0.51, "learning_rate": 1.0078069011209988e-05, "loss": 23.4555, "step": 41680 }, { "epoch": 0.51, "learning_rate": 1.0074085973227949e-05, "loss": 24.1133, "step": 41690 }, { "epoch": 0.51, "learning_rate": 1.0070102923491798e-05, "loss": 23.4622, "step": 41700 }, { "epoch": 0.51, "learning_rate": 1.0066119862633456e-05, "loss": 24.1157, "step": 41710 }, { "epoch": 0.51, "learning_rate": 1.0062136791284862e-05, "loss": 23.436, "step": 41720 }, { "epoch": 0.51, "learning_rate": 1.0058153710077951e-05, "loss": 23.8338, "step": 41730 }, { "epoch": 0.51, "learning_rate": 1.005417061964466e-05, "loss": 23.8509, "step": 41740 }, { "epoch": 0.51, "learning_rate": 1.0050187520616916e-05, "loss": 24.4892, "step": 41750 }, { "epoch": 0.51, "learning_rate": 1.004620441362667e-05, "loss": 23.8312, "step": 41760 }, { "epoch": 0.51, "learning_rate": 1.004222129930586e-05, "loss": 23.7029, "step": 41770 }, { "epoch": 0.51, "learning_rate": 1.0038238178286424e-05, "loss": 24.5051, "step": 41780 }, { "epoch": 0.51, "learning_rate": 1.0034255051200304e-05, "loss": 24.3811, "step": 41790 }, { "epoch": 0.51, "learning_rate": 1.0030271918679448e-05, "loss": 23.6605, "step": 41800 }, { "epoch": 0.51, "learning_rate": 1.0026288781355798e-05, "loss": 23.2368, "step": 41810 }, { "epoch": 0.51, "learning_rate": 1.0022305639861297e-05, "loss": 23.5358, "step": 41820 }, { "epoch": 0.51, "learning_rate": 1.0018322494827895e-05, "loss": 23.7282, "step": 41830 }, { "epoch": 0.51, "learning_rate": 1.0014339346887534e-05, "loss": 23.7592, "step": 41840 }, { "epoch": 0.51, "learning_rate": 1.0010356196672165e-05, "loss": 23.5843, "step": 41850 }, { "epoch": 0.51, "learning_rate": 1.0006373044813729e-05, "loss": 23.4403, "step": 41860 }, { "epoch": 0.51, "learning_rate": 1.000238989194418e-05, "loss": 23.828, "step": 41870 }, { "epoch": 0.52, "learning_rate": 9.998406738695457e-06, "loss": 23.5088, "step": 41880 }, { "epoch": 0.52, "learning_rate": 9.994423585699515e-06, "loss": 22.9541, "step": 41890 }, { "epoch": 0.52, "learning_rate": 9.990440433588305e-06, "loss": 23.976, "step": 41900 }, { "epoch": 0.52, "learning_rate": 9.986457282993768e-06, "loss": 23.0902, "step": 41910 }, { "epoch": 0.52, "learning_rate": 9.982474134547848e-06, "loss": 23.8166, "step": 41920 }, { "epoch": 0.52, "learning_rate": 9.978490988882496e-06, "loss": 23.3791, "step": 41930 }, { "epoch": 0.52, "learning_rate": 9.974507846629661e-06, "loss": 23.9491, "step": 41940 }, { "epoch": 0.52, "learning_rate": 9.970524708421284e-06, "loss": 24.1406, "step": 41950 }, { "epoch": 0.52, "learning_rate": 9.966541574889312e-06, "loss": 23.5167, "step": 41960 }, { "epoch": 0.52, "learning_rate": 9.962558446665693e-06, "loss": 22.7822, "step": 41970 }, { "epoch": 0.52, "learning_rate": 9.958575324382365e-06, "loss": 23.7476, "step": 41980 }, { "epoch": 0.52, "learning_rate": 9.954592208671272e-06, "loss": 24.1964, "step": 41990 }, { "epoch": 0.52, "learning_rate": 9.950609100164357e-06, "loss": 24.0469, "step": 42000 }, { "epoch": 0.52, "learning_rate": 9.94662599949356e-06, "loss": 23.7064, "step": 42010 }, { "epoch": 0.52, "learning_rate": 9.942642907290822e-06, "loss": 24.2241, "step": 42020 }, { "epoch": 0.52, "learning_rate": 9.938659824188077e-06, "loss": 23.604, "step": 42030 }, { "epoch": 0.52, "learning_rate": 9.934676750817266e-06, "loss": 24.0037, "step": 42040 }, { "epoch": 0.52, "learning_rate": 9.93069368781032e-06, "loss": 23.8547, "step": 42050 }, { "epoch": 0.52, "learning_rate": 9.926710635799173e-06, "loss": 23.5544, "step": 42060 }, { "epoch": 0.52, "learning_rate": 9.92272759541576e-06, "loss": 24.4825, "step": 42070 }, { "epoch": 0.52, "learning_rate": 9.918744567292006e-06, "loss": 23.1997, "step": 42080 }, { "epoch": 0.52, "learning_rate": 9.91476155205984e-06, "loss": 23.9405, "step": 42090 }, { "epoch": 0.52, "learning_rate": 9.91077855035119e-06, "loss": 23.3101, "step": 42100 }, { "epoch": 0.52, "learning_rate": 9.906795562797977e-06, "loss": 24.2591, "step": 42110 }, { "epoch": 0.52, "learning_rate": 9.902812590032123e-06, "loss": 23.9156, "step": 42120 }, { "epoch": 0.52, "learning_rate": 9.898829632685548e-06, "loss": 24.445, "step": 42130 }, { "epoch": 0.52, "learning_rate": 9.89484669139017e-06, "loss": 23.8703, "step": 42140 }, { "epoch": 0.52, "learning_rate": 9.890863766777898e-06, "loss": 24.0651, "step": 42150 }, { "epoch": 0.52, "learning_rate": 9.886880859480644e-06, "loss": 23.9745, "step": 42160 }, { "epoch": 0.52, "learning_rate": 9.88289797013032e-06, "loss": 23.9875, "step": 42170 }, { "epoch": 0.52, "learning_rate": 9.87891509935883e-06, "loss": 23.5228, "step": 42180 }, { "epoch": 0.52, "learning_rate": 9.874932247798076e-06, "loss": 23.6674, "step": 42190 }, { "epoch": 0.52, "learning_rate": 9.870949416079958e-06, "loss": 23.5387, "step": 42200 }, { "epoch": 0.52, "learning_rate": 9.866966604836376e-06, "loss": 23.1521, "step": 42210 }, { "epoch": 0.52, "learning_rate": 9.862983814699217e-06, "loss": 23.6959, "step": 42220 }, { "epoch": 0.52, "learning_rate": 9.859001046300375e-06, "loss": 23.4741, "step": 42230 }, { "epoch": 0.52, "learning_rate": 9.855018300271739e-06, "loss": 23.879, "step": 42240 }, { "epoch": 0.52, "learning_rate": 9.851035577245185e-06, "loss": 23.239, "step": 42250 }, { "epoch": 0.52, "learning_rate": 9.847052877852596e-06, "loss": 23.9368, "step": 42260 }, { "epoch": 0.52, "learning_rate": 9.843070202725852e-06, "loss": 23.7707, "step": 42270 }, { "epoch": 0.52, "learning_rate": 9.839087552496816e-06, "loss": 23.9652, "step": 42280 }, { "epoch": 0.52, "learning_rate": 9.835104927797362e-06, "loss": 24.2465, "step": 42290 }, { "epoch": 0.52, "learning_rate": 9.831122329259352e-06, "loss": 24.0361, "step": 42300 }, { "epoch": 0.52, "learning_rate": 9.827139757514646e-06, "loss": 23.8719, "step": 42310 }, { "epoch": 0.52, "learning_rate": 9.823157213195096e-06, "loss": 23.2834, "step": 42320 }, { "epoch": 0.52, "learning_rate": 9.81917469693256e-06, "loss": 24.1134, "step": 42330 }, { "epoch": 0.52, "learning_rate": 9.815192209358877e-06, "loss": 23.352, "step": 42340 }, { "epoch": 0.52, "learning_rate": 9.811209751105893e-06, "loss": 24.2063, "step": 42350 }, { "epoch": 0.52, "learning_rate": 9.807227322805446e-06, "loss": 24.0235, "step": 42360 }, { "epoch": 0.52, "learning_rate": 9.803244925089366e-06, "loss": 23.54, "step": 42370 }, { "epoch": 0.52, "learning_rate": 9.799262558589484e-06, "loss": 23.8471, "step": 42380 }, { "epoch": 0.52, "learning_rate": 9.795280223937616e-06, "loss": 23.5004, "step": 42390 }, { "epoch": 0.52, "learning_rate": 9.791297921765586e-06, "loss": 24.5132, "step": 42400 }, { "epoch": 0.52, "learning_rate": 9.787315652705206e-06, "loss": 23.3674, "step": 42410 }, { "epoch": 0.52, "learning_rate": 9.78333341738828e-06, "loss": 24.0987, "step": 42420 }, { "epoch": 0.52, "learning_rate": 9.779351216446613e-06, "loss": 24.1481, "step": 42430 }, { "epoch": 0.52, "learning_rate": 9.775369050512002e-06, "loss": 23.8256, "step": 42440 }, { "epoch": 0.52, "learning_rate": 9.771386920216232e-06, "loss": 24.1927, "step": 42450 }, { "epoch": 0.52, "learning_rate": 9.767404826191095e-06, "loss": 24.2928, "step": 42460 }, { "epoch": 0.52, "learning_rate": 9.763422769068371e-06, "loss": 23.7552, "step": 42470 }, { "epoch": 0.52, "learning_rate": 9.759440749479827e-06, "loss": 23.9045, "step": 42480 }, { "epoch": 0.52, "learning_rate": 9.755458768057234e-06, "loss": 23.9743, "step": 42490 }, { "epoch": 0.52, "learning_rate": 9.751476825432358e-06, "loss": 23.5023, "step": 42500 }, { "epoch": 0.52, "learning_rate": 9.747494922236948e-06, "loss": 23.3758, "step": 42510 }, { "epoch": 0.52, "learning_rate": 9.743513059102755e-06, "loss": 24.1577, "step": 42520 }, { "epoch": 0.52, "learning_rate": 9.739531236661528e-06, "loss": 23.6363, "step": 42530 }, { "epoch": 0.52, "learning_rate": 9.735549455544993e-06, "loss": 23.5378, "step": 42540 }, { "epoch": 0.52, "learning_rate": 9.731567716384886e-06, "loss": 24.3685, "step": 42550 }, { "epoch": 0.52, "learning_rate": 9.727586019812929e-06, "loss": 23.826, "step": 42560 }, { "epoch": 0.52, "learning_rate": 9.723604366460841e-06, "loss": 24.0945, "step": 42570 }, { "epoch": 0.52, "learning_rate": 9.719622756960327e-06, "loss": 23.497, "step": 42580 }, { "epoch": 0.52, "learning_rate": 9.715641191943093e-06, "loss": 23.9185, "step": 42590 }, { "epoch": 0.52, "learning_rate": 9.711659672040838e-06, "loss": 23.7411, "step": 42600 }, { "epoch": 0.52, "learning_rate": 9.707678197885245e-06, "loss": 23.8848, "step": 42610 }, { "epoch": 0.52, "learning_rate": 9.70369677010799e-06, "loss": 24.0497, "step": 42620 }, { "epoch": 0.52, "learning_rate": 9.699715389340757e-06, "loss": 23.8858, "step": 42630 }, { "epoch": 0.52, "learning_rate": 9.695734056215209e-06, "loss": 24.0741, "step": 42640 }, { "epoch": 0.52, "learning_rate": 9.691752771363003e-06, "loss": 23.8937, "step": 42650 }, { "epoch": 0.52, "learning_rate": 9.68777153541579e-06, "loss": 23.5374, "step": 42660 }, { "epoch": 0.52, "learning_rate": 9.683790349005218e-06, "loss": 23.8224, "step": 42670 }, { "epoch": 0.52, "learning_rate": 9.679809212762916e-06, "loss": 22.8363, "step": 42680 }, { "epoch": 0.53, "learning_rate": 9.675828127320515e-06, "loss": 24.1999, "step": 42690 }, { "epoch": 0.53, "learning_rate": 9.671847093309635e-06, "loss": 23.8257, "step": 42700 }, { "epoch": 0.53, "learning_rate": 9.66786611136189e-06, "loss": 23.3576, "step": 42710 }, { "epoch": 0.53, "learning_rate": 9.663885182108875e-06, "loss": 24.1602, "step": 42720 }, { "epoch": 0.53, "learning_rate": 9.659904306182191e-06, "loss": 23.6769, "step": 42730 }, { "epoch": 0.53, "learning_rate": 9.655923484213425e-06, "loss": 23.8776, "step": 42740 }, { "epoch": 0.53, "learning_rate": 9.651942716834149e-06, "loss": 23.834, "step": 42750 }, { "epoch": 0.53, "learning_rate": 9.647962004675936e-06, "loss": 23.2411, "step": 42760 }, { "epoch": 0.53, "learning_rate": 9.643981348370347e-06, "loss": 23.173, "step": 42770 }, { "epoch": 0.53, "learning_rate": 9.640000748548932e-06, "loss": 23.7064, "step": 42780 }, { "epoch": 0.53, "learning_rate": 9.636020205843233e-06, "loss": 23.652, "step": 42790 }, { "epoch": 0.53, "learning_rate": 9.632039720884786e-06, "loss": 23.8827, "step": 42800 }, { "epoch": 0.53, "learning_rate": 9.62805929430511e-06, "loss": 23.5324, "step": 42810 }, { "epoch": 0.53, "learning_rate": 9.624078926735725e-06, "loss": 23.374, "step": 42820 }, { "epoch": 0.53, "learning_rate": 9.620098618808138e-06, "loss": 24.3448, "step": 42830 }, { "epoch": 0.53, "learning_rate": 9.61611837115384e-06, "loss": 24.5248, "step": 42840 }, { "epoch": 0.53, "learning_rate": 9.612138184404319e-06, "loss": 23.4777, "step": 42850 }, { "epoch": 0.53, "learning_rate": 9.608158059191051e-06, "loss": 24.4362, "step": 42860 }, { "epoch": 0.53, "learning_rate": 9.604177996145504e-06, "loss": 23.8053, "step": 42870 }, { "epoch": 0.53, "learning_rate": 9.600197995899139e-06, "loss": 23.836, "step": 42880 }, { "epoch": 0.53, "learning_rate": 9.596218059083398e-06, "loss": 23.789, "step": 42890 }, { "epoch": 0.53, "learning_rate": 9.59223818632972e-06, "loss": 24.1638, "step": 42900 }, { "epoch": 0.53, "learning_rate": 9.588258378269535e-06, "loss": 23.7286, "step": 42910 }, { "epoch": 0.53, "learning_rate": 9.584278635534255e-06, "loss": 24.1234, "step": 42920 }, { "epoch": 0.53, "learning_rate": 9.58029895875529e-06, "loss": 23.827, "step": 42930 }, { "epoch": 0.53, "learning_rate": 9.576319348564036e-06, "loss": 24.3883, "step": 42940 }, { "epoch": 0.53, "learning_rate": 9.572339805591875e-06, "loss": 23.9603, "step": 42950 }, { "epoch": 0.53, "learning_rate": 9.568360330470186e-06, "loss": 24.1058, "step": 42960 }, { "epoch": 0.53, "learning_rate": 9.564380923830333e-06, "loss": 23.8996, "step": 42970 }, { "epoch": 0.53, "learning_rate": 9.560401586303663e-06, "loss": 23.8707, "step": 42980 }, { "epoch": 0.53, "learning_rate": 9.556422318521526e-06, "loss": 23.538, "step": 42990 }, { "epoch": 0.53, "learning_rate": 9.552443121115248e-06, "loss": 23.7916, "step": 43000 }, { "epoch": 0.53, "learning_rate": 9.548463994716151e-06, "loss": 23.541, "step": 43010 }, { "epoch": 0.53, "learning_rate": 9.544484939955542e-06, "loss": 23.7848, "step": 43020 }, { "epoch": 0.53, "learning_rate": 9.540505957464724e-06, "loss": 23.6532, "step": 43030 }, { "epoch": 0.53, "learning_rate": 9.536527047874974e-06, "loss": 23.4577, "step": 43040 }, { "epoch": 0.53, "learning_rate": 9.53254821181757e-06, "loss": 23.7327, "step": 43050 }, { "epoch": 0.53, "learning_rate": 9.528569449923782e-06, "loss": 23.3754, "step": 43060 }, { "epoch": 0.53, "learning_rate": 9.52459076282485e-06, "loss": 24.1505, "step": 43070 }, { "epoch": 0.53, "learning_rate": 9.520612151152018e-06, "loss": 24.1891, "step": 43080 }, { "epoch": 0.53, "learning_rate": 9.51663361553651e-06, "loss": 23.5677, "step": 43090 }, { "epoch": 0.53, "learning_rate": 9.512655156609544e-06, "loss": 23.7072, "step": 43100 }, { "epoch": 0.53, "learning_rate": 9.508676775002323e-06, "loss": 24.0275, "step": 43110 }, { "epoch": 0.53, "learning_rate": 9.504698471346037e-06, "loss": 24.4826, "step": 43120 }, { "epoch": 0.53, "learning_rate": 9.50072024627186e-06, "loss": 23.2869, "step": 43130 }, { "epoch": 0.53, "learning_rate": 9.496742100410967e-06, "loss": 23.9833, "step": 43140 }, { "epoch": 0.53, "learning_rate": 9.492764034394501e-06, "loss": 23.6753, "step": 43150 }, { "epoch": 0.53, "learning_rate": 9.48878604885361e-06, "loss": 23.5367, "step": 43160 }, { "epoch": 0.53, "learning_rate": 9.484808144419417e-06, "loss": 23.7603, "step": 43170 }, { "epoch": 0.53, "learning_rate": 9.480830321723038e-06, "loss": 24.3152, "step": 43180 }, { "epoch": 0.53, "learning_rate": 9.476852581395576e-06, "loss": 24.0106, "step": 43190 }, { "epoch": 0.53, "learning_rate": 9.472874924068121e-06, "loss": 24.4903, "step": 43200 }, { "epoch": 0.53, "learning_rate": 9.468897350371743e-06, "loss": 24.0754, "step": 43210 }, { "epoch": 0.53, "learning_rate": 9.46491986093751e-06, "loss": 23.8871, "step": 43220 }, { "epoch": 0.53, "learning_rate": 9.460942456396471e-06, "loss": 23.5291, "step": 43230 }, { "epoch": 0.53, "learning_rate": 9.456965137379657e-06, "loss": 23.8094, "step": 43240 }, { "epoch": 0.53, "learning_rate": 9.452987904518092e-06, "loss": 23.6065, "step": 43250 }, { "epoch": 0.53, "learning_rate": 9.449010758442786e-06, "loss": 23.9487, "step": 43260 }, { "epoch": 0.53, "learning_rate": 9.445033699784733e-06, "loss": 23.9458, "step": 43270 }, { "epoch": 0.53, "learning_rate": 9.44105672917491e-06, "loss": 23.4865, "step": 43280 }, { "epoch": 0.53, "learning_rate": 9.437079847244295e-06, "loss": 23.6443, "step": 43290 }, { "epoch": 0.53, "learning_rate": 9.433103054623824e-06, "loss": 23.8907, "step": 43300 }, { "epoch": 0.53, "learning_rate": 9.429126351944446e-06, "loss": 23.8648, "step": 43310 }, { "epoch": 0.53, "learning_rate": 9.425149739837084e-06, "loss": 23.631, "step": 43320 }, { "epoch": 0.53, "learning_rate": 9.421173218932641e-06, "loss": 23.3824, "step": 43330 }, { "epoch": 0.53, "learning_rate": 9.417196789862024e-06, "loss": 24.0151, "step": 43340 }, { "epoch": 0.53, "learning_rate": 9.413220453256105e-06, "loss": 23.305, "step": 43350 }, { "epoch": 0.53, "learning_rate": 9.409244209745751e-06, "loss": 23.9789, "step": 43360 }, { "epoch": 0.53, "learning_rate": 9.405268059961818e-06, "loss": 23.2142, "step": 43370 }, { "epoch": 0.53, "learning_rate": 9.401292004535139e-06, "loss": 23.5284, "step": 43380 }, { "epoch": 0.53, "learning_rate": 9.397316044096535e-06, "loss": 23.7634, "step": 43390 }, { "epoch": 0.53, "learning_rate": 9.393340179276813e-06, "loss": 23.8981, "step": 43400 }, { "epoch": 0.53, "learning_rate": 9.389364410706764e-06, "loss": 23.6628, "step": 43410 }, { "epoch": 0.53, "learning_rate": 9.385388739017165e-06, "loss": 23.8804, "step": 43420 }, { "epoch": 0.53, "learning_rate": 9.381413164838775e-06, "loss": 23.8335, "step": 43430 }, { "epoch": 0.53, "learning_rate": 9.377437688802341e-06, "loss": 23.833, "step": 43440 }, { "epoch": 0.53, "learning_rate": 9.373462311538593e-06, "loss": 23.8173, "step": 43450 }, { "epoch": 0.53, "learning_rate": 9.369487033678242e-06, "loss": 23.6081, "step": 43460 }, { "epoch": 0.53, "learning_rate": 9.36551185585199e-06, "loss": 23.4549, "step": 43470 }, { "epoch": 0.53, "learning_rate": 9.361536778690513e-06, "loss": 24.3559, "step": 43480 }, { "epoch": 0.53, "learning_rate": 9.357561802824483e-06, "loss": 23.8429, "step": 43490 }, { "epoch": 0.53, "learning_rate": 9.35358692888455e-06, "loss": 24.5827, "step": 43500 }, { "epoch": 0.54, "learning_rate": 9.349612157501344e-06, "loss": 23.2781, "step": 43510 }, { "epoch": 0.54, "learning_rate": 9.345637489305489e-06, "loss": 24.8447, "step": 43520 }, { "epoch": 0.54, "learning_rate": 9.34166292492758e-06, "loss": 23.5513, "step": 43530 }, { "epoch": 0.54, "learning_rate": 9.337688464998204e-06, "loss": 24.0376, "step": 43540 }, { "epoch": 0.54, "learning_rate": 9.333714110147929e-06, "loss": 23.8023, "step": 43550 }, { "epoch": 0.54, "learning_rate": 9.329739861007307e-06, "loss": 24.1747, "step": 43560 }, { "epoch": 0.54, "learning_rate": 9.325765718206873e-06, "loss": 23.6787, "step": 43570 }, { "epoch": 0.54, "learning_rate": 9.321791682377147e-06, "loss": 23.5368, "step": 43580 }, { "epoch": 0.54, "learning_rate": 9.317817754148628e-06, "loss": 24.3214, "step": 43590 }, { "epoch": 0.54, "learning_rate": 9.313843934151797e-06, "loss": 23.4779, "step": 43600 }, { "epoch": 0.54, "learning_rate": 9.30987022301713e-06, "loss": 23.6283, "step": 43610 }, { "epoch": 0.54, "learning_rate": 9.305896621375065e-06, "loss": 23.4215, "step": 43620 }, { "epoch": 0.54, "learning_rate": 9.301923129856042e-06, "loss": 23.8472, "step": 43630 }, { "epoch": 0.54, "learning_rate": 9.297949749090475e-06, "loss": 23.9967, "step": 43640 }, { "epoch": 0.54, "learning_rate": 9.293976479708757e-06, "loss": 24.169, "step": 43650 }, { "epoch": 0.54, "learning_rate": 9.29000332234127e-06, "loss": 23.2742, "step": 43660 }, { "epoch": 0.54, "learning_rate": 9.286030277618378e-06, "loss": 23.6312, "step": 43670 }, { "epoch": 0.54, "learning_rate": 9.282057346170422e-06, "loss": 22.9978, "step": 43680 }, { "epoch": 0.54, "learning_rate": 9.278084528627726e-06, "loss": 23.7805, "step": 43690 }, { "epoch": 0.54, "learning_rate": 9.274111825620603e-06, "loss": 23.8281, "step": 43700 }, { "epoch": 0.54, "learning_rate": 9.270139237779338e-06, "loss": 23.9037, "step": 43710 }, { "epoch": 0.54, "learning_rate": 9.266166765734203e-06, "loss": 24.086, "step": 43720 }, { "epoch": 0.54, "learning_rate": 9.262194410115453e-06, "loss": 23.7117, "step": 43730 }, { "epoch": 0.54, "learning_rate": 9.258222171553322e-06, "loss": 23.8838, "step": 43740 }, { "epoch": 0.54, "learning_rate": 9.254250050678024e-06, "loss": 24.7059, "step": 43750 }, { "epoch": 0.54, "learning_rate": 9.250278048119762e-06, "loss": 23.4152, "step": 43760 }, { "epoch": 0.54, "learning_rate": 9.246306164508704e-06, "loss": 23.792, "step": 43770 }, { "epoch": 0.54, "learning_rate": 9.24233440047502e-06, "loss": 24.0849, "step": 43780 }, { "epoch": 0.54, "learning_rate": 9.238362756648844e-06, "loss": 23.9691, "step": 43790 }, { "epoch": 0.54, "learning_rate": 9.234391233660298e-06, "loss": 23.4989, "step": 43800 }, { "epoch": 0.54, "learning_rate": 9.23041983213949e-06, "loss": 23.4707, "step": 43810 }, { "epoch": 0.54, "learning_rate": 9.226448552716496e-06, "loss": 23.8849, "step": 43820 }, { "epoch": 0.54, "learning_rate": 9.222477396021382e-06, "loss": 23.7463, "step": 43830 }, { "epoch": 0.54, "learning_rate": 9.218506362684197e-06, "loss": 23.3557, "step": 43840 }, { "epoch": 0.54, "learning_rate": 9.214535453334958e-06, "loss": 23.8885, "step": 43850 }, { "epoch": 0.54, "learning_rate": 9.210564668603676e-06, "loss": 23.613, "step": 43860 }, { "epoch": 0.54, "learning_rate": 9.206594009120334e-06, "loss": 23.7356, "step": 43870 }, { "epoch": 0.54, "learning_rate": 9.202623475514896e-06, "loss": 24.4411, "step": 43880 }, { "epoch": 0.54, "learning_rate": 9.19865306841731e-06, "loss": 23.7414, "step": 43890 }, { "epoch": 0.54, "learning_rate": 9.194682788457501e-06, "loss": 24.1478, "step": 43900 }, { "epoch": 0.54, "learning_rate": 9.190712636265372e-06, "loss": 23.5848, "step": 43910 }, { "epoch": 0.54, "learning_rate": 9.18674261247081e-06, "loss": 22.9506, "step": 43920 }, { "epoch": 0.54, "learning_rate": 9.182772717703679e-06, "loss": 23.3827, "step": 43930 }, { "epoch": 0.54, "learning_rate": 9.178802952593822e-06, "loss": 23.4945, "step": 43940 }, { "epoch": 0.54, "learning_rate": 9.174833317771062e-06, "loss": 24.1435, "step": 43950 }, { "epoch": 0.54, "learning_rate": 9.170863813865203e-06, "loss": 24.0933, "step": 43960 }, { "epoch": 0.54, "learning_rate": 9.166894441506031e-06, "loss": 23.6028, "step": 43970 }, { "epoch": 0.54, "learning_rate": 9.1629252013233e-06, "loss": 23.1563, "step": 43980 }, { "epoch": 0.54, "learning_rate": 9.158956093946758e-06, "loss": 23.9574, "step": 43990 }, { "epoch": 0.54, "learning_rate": 9.154987120006116e-06, "loss": 23.7179, "step": 44000 }, { "epoch": 0.54, "learning_rate": 9.151018280131077e-06, "loss": 24.3688, "step": 44010 }, { "epoch": 0.54, "learning_rate": 9.147049574951314e-06, "loss": 24.3678, "step": 44020 }, { "epoch": 0.54, "learning_rate": 9.143081005096484e-06, "loss": 23.7268, "step": 44030 }, { "epoch": 0.54, "learning_rate": 9.139112571196224e-06, "loss": 23.6075, "step": 44040 }, { "epoch": 0.54, "learning_rate": 9.135144273880142e-06, "loss": 23.2733, "step": 44050 }, { "epoch": 0.54, "learning_rate": 9.131176113777829e-06, "loss": 23.467, "step": 44060 }, { "epoch": 0.54, "learning_rate": 9.127208091518857e-06, "loss": 23.995, "step": 44070 }, { "epoch": 0.54, "learning_rate": 9.12324020773277e-06, "loss": 23.3081, "step": 44080 }, { "epoch": 0.54, "learning_rate": 9.119272463049095e-06, "loss": 23.9257, "step": 44090 }, { "epoch": 0.54, "learning_rate": 9.115304858097335e-06, "loss": 23.6577, "step": 44100 }, { "epoch": 0.54, "learning_rate": 9.111337393506967e-06, "loss": 24.3517, "step": 44110 }, { "epoch": 0.54, "learning_rate": 9.107370069907452e-06, "loss": 23.9366, "step": 44120 }, { "epoch": 0.54, "learning_rate": 9.103402887928227e-06, "loss": 23.1909, "step": 44130 }, { "epoch": 0.54, "learning_rate": 9.099435848198707e-06, "loss": 23.347, "step": 44140 }, { "epoch": 0.54, "learning_rate": 9.095468951348279e-06, "loss": 24.0437, "step": 44150 }, { "epoch": 0.54, "learning_rate": 9.091502198006313e-06, "loss": 23.5948, "step": 44160 }, { "epoch": 0.54, "learning_rate": 9.087535588802158e-06, "loss": 24.1727, "step": 44170 }, { "epoch": 0.54, "learning_rate": 9.08356912436513e-06, "loss": 24.5013, "step": 44180 }, { "epoch": 0.54, "learning_rate": 9.079602805324534e-06, "loss": 23.8836, "step": 44190 }, { "epoch": 0.54, "learning_rate": 9.075636632309648e-06, "loss": 23.4707, "step": 44200 }, { "epoch": 0.54, "learning_rate": 9.07167060594972e-06, "loss": 24.2876, "step": 44210 }, { "epoch": 0.54, "learning_rate": 9.067704726873987e-06, "loss": 23.5762, "step": 44220 }, { "epoch": 0.54, "learning_rate": 9.063738995711648e-06, "loss": 23.9966, "step": 44230 }, { "epoch": 0.54, "learning_rate": 9.059773413091894e-06, "loss": 23.4166, "step": 44240 }, { "epoch": 0.54, "learning_rate": 9.055807979643878e-06, "loss": 23.9522, "step": 44250 }, { "epoch": 0.54, "learning_rate": 9.05184269599674e-06, "loss": 23.4101, "step": 44260 }, { "epoch": 0.54, "learning_rate": 9.047877562779594e-06, "loss": 23.6702, "step": 44270 }, { "epoch": 0.54, "learning_rate": 9.043912580621528e-06, "loss": 23.646, "step": 44280 }, { "epoch": 0.54, "learning_rate": 9.039947750151604e-06, "loss": 23.5809, "step": 44290 }, { "epoch": 0.54, "learning_rate": 9.035983071998862e-06, "loss": 23.2967, "step": 44300 }, { "epoch": 0.54, "learning_rate": 9.032018546792324e-06, "loss": 23.6323, "step": 44310 }, { "epoch": 0.55, "learning_rate": 9.028054175160976e-06, "loss": 23.6965, "step": 44320 }, { "epoch": 0.55, "learning_rate": 9.024089957733788e-06, "loss": 24.0334, "step": 44330 }, { "epoch": 0.55, "learning_rate": 9.020125895139706e-06, "loss": 23.6384, "step": 44340 }, { "epoch": 0.55, "learning_rate": 9.016161988007644e-06, "loss": 23.673, "step": 44350 }, { "epoch": 0.55, "learning_rate": 9.012198236966499e-06, "loss": 23.6982, "step": 44360 }, { "epoch": 0.55, "learning_rate": 9.008234642645143e-06, "loss": 23.8248, "step": 44370 }, { "epoch": 0.55, "learning_rate": 9.004271205672415e-06, "loss": 23.8677, "step": 44380 }, { "epoch": 0.55, "learning_rate": 9.000307926677136e-06, "loss": 23.8102, "step": 44390 }, { "epoch": 0.55, "learning_rate": 8.996344806288102e-06, "loss": 24.0596, "step": 44400 }, { "epoch": 0.55, "learning_rate": 8.99238184513408e-06, "loss": 23.9175, "step": 44410 }, { "epoch": 0.55, "learning_rate": 8.988419043843817e-06, "loss": 23.3625, "step": 44420 }, { "epoch": 0.55, "learning_rate": 8.98445640304603e-06, "loss": 23.5267, "step": 44430 }, { "epoch": 0.55, "learning_rate": 8.98049392336941e-06, "loss": 23.9263, "step": 44440 }, { "epoch": 0.55, "learning_rate": 8.976531605442633e-06, "loss": 24.0199, "step": 44450 }, { "epoch": 0.55, "learning_rate": 8.97256944989433e-06, "loss": 23.798, "step": 44460 }, { "epoch": 0.55, "learning_rate": 8.96860745735312e-06, "loss": 23.8882, "step": 44470 }, { "epoch": 0.55, "learning_rate": 8.9646456284476e-06, "loss": 23.6425, "step": 44480 }, { "epoch": 0.55, "learning_rate": 8.960683963806326e-06, "loss": 24.166, "step": 44490 }, { "epoch": 0.55, "learning_rate": 8.956722464057839e-06, "loss": 24.1137, "step": 44500 }, { "epoch": 0.55, "learning_rate": 8.952761129830656e-06, "loss": 23.9699, "step": 44510 }, { "epoch": 0.55, "learning_rate": 8.948799961753255e-06, "loss": 23.9228, "step": 44520 }, { "epoch": 0.55, "learning_rate": 8.944838960454101e-06, "loss": 23.6041, "step": 44530 }, { "epoch": 0.55, "learning_rate": 8.940878126561628e-06, "loss": 24.6055, "step": 44540 }, { "epoch": 0.55, "learning_rate": 8.936917460704236e-06, "loss": 23.4367, "step": 44550 }, { "epoch": 0.55, "learning_rate": 8.932956963510308e-06, "loss": 24.5431, "step": 44560 }, { "epoch": 0.55, "learning_rate": 8.928996635608202e-06, "loss": 23.8859, "step": 44570 }, { "epoch": 0.55, "learning_rate": 8.925036477626238e-06, "loss": 23.6128, "step": 44580 }, { "epoch": 0.55, "learning_rate": 8.921076490192716e-06, "loss": 24.0876, "step": 44590 }, { "epoch": 0.55, "learning_rate": 8.917116673935913e-06, "loss": 24.438, "step": 44600 }, { "epoch": 0.55, "learning_rate": 8.913157029484067e-06, "loss": 23.4669, "step": 44610 }, { "epoch": 0.55, "learning_rate": 8.9091975574654e-06, "loss": 24.0896, "step": 44620 }, { "epoch": 0.55, "learning_rate": 8.905238258508102e-06, "loss": 23.7558, "step": 44630 }, { "epoch": 0.55, "learning_rate": 8.901279133240337e-06, "loss": 23.9992, "step": 44640 }, { "epoch": 0.55, "learning_rate": 8.897320182290237e-06, "loss": 24.0986, "step": 44650 }, { "epoch": 0.55, "learning_rate": 8.893361406285911e-06, "loss": 23.7679, "step": 44660 }, { "epoch": 0.55, "learning_rate": 8.889402805855446e-06, "loss": 24.2416, "step": 44670 }, { "epoch": 0.55, "learning_rate": 8.885444381626886e-06, "loss": 23.8826, "step": 44680 }, { "epoch": 0.55, "learning_rate": 8.881486134228254e-06, "loss": 23.8024, "step": 44690 }, { "epoch": 0.55, "learning_rate": 8.87752806428755e-06, "loss": 23.8915, "step": 44700 }, { "epoch": 0.55, "learning_rate": 8.873570172432742e-06, "loss": 24.0547, "step": 44710 }, { "epoch": 0.55, "learning_rate": 8.869612459291769e-06, "loss": 24.1754, "step": 44720 }, { "epoch": 0.55, "learning_rate": 8.86565492549254e-06, "loss": 24.0091, "step": 44730 }, { "epoch": 0.55, "learning_rate": 8.861697571662943e-06, "loss": 23.6096, "step": 44740 }, { "epoch": 0.55, "learning_rate": 8.857740398430828e-06, "loss": 23.9382, "step": 44750 }, { "epoch": 0.55, "learning_rate": 8.853783406424023e-06, "loss": 23.4718, "step": 44760 }, { "epoch": 0.55, "learning_rate": 8.849826596270326e-06, "loss": 23.9224, "step": 44770 }, { "epoch": 0.55, "learning_rate": 8.845869968597501e-06, "loss": 23.6965, "step": 44780 }, { "epoch": 0.55, "learning_rate": 8.84191352403329e-06, "loss": 24.0898, "step": 44790 }, { "epoch": 0.55, "learning_rate": 8.837957263205405e-06, "loss": 23.5103, "step": 44800 }, { "epoch": 0.55, "learning_rate": 8.834001186741522e-06, "loss": 23.2954, "step": 44810 }, { "epoch": 0.55, "learning_rate": 8.830045295269295e-06, "loss": 23.786, "step": 44820 }, { "epoch": 0.55, "learning_rate": 8.826089589416347e-06, "loss": 23.9063, "step": 44830 }, { "epoch": 0.55, "learning_rate": 8.822134069810271e-06, "loss": 23.4822, "step": 44840 }, { "epoch": 0.55, "learning_rate": 8.818178737078631e-06, "loss": 23.475, "step": 44850 }, { "epoch": 0.55, "learning_rate": 8.814223591848957e-06, "loss": 23.738, "step": 44860 }, { "epoch": 0.55, "learning_rate": 8.810268634748759e-06, "loss": 23.4509, "step": 44870 }, { "epoch": 0.55, "learning_rate": 8.806313866405505e-06, "loss": 23.4844, "step": 44880 }, { "epoch": 0.55, "learning_rate": 8.802359287446643e-06, "loss": 23.7411, "step": 44890 }, { "epoch": 0.55, "learning_rate": 8.798404898499587e-06, "loss": 23.5827, "step": 44900 }, { "epoch": 0.55, "learning_rate": 8.794450700191722e-06, "loss": 23.3996, "step": 44910 }, { "epoch": 0.55, "learning_rate": 8.790496693150397e-06, "loss": 23.8562, "step": 44920 }, { "epoch": 0.55, "learning_rate": 8.786542878002936e-06, "loss": 23.5896, "step": 44930 }, { "epoch": 0.55, "learning_rate": 8.782589255376637e-06, "loss": 23.5602, "step": 44940 }, { "epoch": 0.55, "learning_rate": 8.778635825898755e-06, "loss": 24.0461, "step": 44950 }, { "epoch": 0.55, "learning_rate": 8.77468259019653e-06, "loss": 24.0209, "step": 44960 }, { "epoch": 0.55, "learning_rate": 8.770729548897157e-06, "loss": 24.1411, "step": 44970 }, { "epoch": 0.55, "learning_rate": 8.766776702627809e-06, "loss": 24.0859, "step": 44980 }, { "epoch": 0.55, "learning_rate": 8.762824052015625e-06, "loss": 23.8338, "step": 44990 }, { "epoch": 0.55, "learning_rate": 8.75887159768771e-06, "loss": 24.1571, "step": 45000 }, { "epoch": 0.55, "learning_rate": 8.754919340271148e-06, "loss": 23.9645, "step": 45010 }, { "epoch": 0.55, "learning_rate": 8.750967280392977e-06, "loss": 23.5369, "step": 45020 }, { "epoch": 0.55, "learning_rate": 8.747015418680216e-06, "loss": 23.9175, "step": 45030 }, { "epoch": 0.55, "learning_rate": 8.743063755759848e-06, "loss": 24.254, "step": 45040 }, { "epoch": 0.55, "learning_rate": 8.739112292258822e-06, "loss": 23.7068, "step": 45050 }, { "epoch": 0.55, "learning_rate": 8.735161028804059e-06, "loss": 23.867, "step": 45060 }, { "epoch": 0.55, "learning_rate": 8.731209966022448e-06, "loss": 24.0154, "step": 45070 }, { "epoch": 0.55, "learning_rate": 8.727259104540843e-06, "loss": 24.1675, "step": 45080 }, { "epoch": 0.55, "learning_rate": 8.723308444986072e-06, "loss": 23.4602, "step": 45090 }, { "epoch": 0.55, "learning_rate": 8.719357987984926e-06, "loss": 23.6042, "step": 45100 }, { "epoch": 0.55, "learning_rate": 8.71540773416416e-06, "loss": 24.0584, "step": 45110 }, { "epoch": 0.55, "learning_rate": 8.711457684150509e-06, "loss": 24.2514, "step": 45120 }, { "epoch": 0.56, "learning_rate": 8.707507838570668e-06, "loss": 23.7663, "step": 45130 }, { "epoch": 0.56, "learning_rate": 8.703558198051296e-06, "loss": 23.3079, "step": 45140 }, { "epoch": 0.56, "learning_rate": 8.699608763219028e-06, "loss": 24.4457, "step": 45150 }, { "epoch": 0.56, "learning_rate": 8.695659534700454e-06, "loss": 23.7757, "step": 45160 }, { "epoch": 0.56, "learning_rate": 8.691710513122147e-06, "loss": 23.5833, "step": 45170 }, { "epoch": 0.56, "learning_rate": 8.687761699110639e-06, "loss": 23.6289, "step": 45180 }, { "epoch": 0.56, "learning_rate": 8.683813093292424e-06, "loss": 23.8455, "step": 45190 }, { "epoch": 0.56, "learning_rate": 8.679864696293974e-06, "loss": 23.8968, "step": 45200 }, { "epoch": 0.56, "learning_rate": 8.675916508741722e-06, "loss": 23.6015, "step": 45210 }, { "epoch": 0.56, "learning_rate": 8.671968531262063e-06, "loss": 23.9256, "step": 45220 }, { "epoch": 0.56, "learning_rate": 8.668020764481369e-06, "loss": 24.4379, "step": 45230 }, { "epoch": 0.56, "learning_rate": 8.664073209025971e-06, "loss": 24.5508, "step": 45240 }, { "epoch": 0.56, "learning_rate": 8.660125865522169e-06, "loss": 23.721, "step": 45250 }, { "epoch": 0.56, "learning_rate": 8.65617873459623e-06, "loss": 24.0892, "step": 45260 }, { "epoch": 0.56, "learning_rate": 8.652231816874388e-06, "loss": 23.6975, "step": 45270 }, { "epoch": 0.56, "learning_rate": 8.648285112982837e-06, "loss": 23.9061, "step": 45280 }, { "epoch": 0.56, "learning_rate": 8.644338623547744e-06, "loss": 24.1063, "step": 45290 }, { "epoch": 0.56, "learning_rate": 8.640392349195242e-06, "loss": 24.1293, "step": 45300 }, { "epoch": 0.56, "learning_rate": 8.636446290551424e-06, "loss": 24.2736, "step": 45310 }, { "epoch": 0.56, "learning_rate": 8.632500448242354e-06, "loss": 23.6047, "step": 45320 }, { "epoch": 0.56, "learning_rate": 8.628554822894062e-06, "loss": 23.9078, "step": 45330 }, { "epoch": 0.56, "learning_rate": 8.624609415132535e-06, "loss": 23.4274, "step": 45340 }, { "epoch": 0.56, "learning_rate": 8.620664225583737e-06, "loss": 23.8396, "step": 45350 }, { "epoch": 0.56, "learning_rate": 8.616719254873593e-06, "loss": 23.6664, "step": 45360 }, { "epoch": 0.56, "learning_rate": 8.612774503627995e-06, "loss": 23.9738, "step": 45370 }, { "epoch": 0.56, "learning_rate": 8.608829972472795e-06, "loss": 23.614, "step": 45380 }, { "epoch": 0.56, "learning_rate": 8.604885662033808e-06, "loss": 23.887, "step": 45390 }, { "epoch": 0.56, "learning_rate": 8.600941572936822e-06, "loss": 23.8672, "step": 45400 }, { "epoch": 0.56, "learning_rate": 8.596997705807592e-06, "loss": 24.3412, "step": 45410 }, { "epoch": 0.56, "learning_rate": 8.593054061271829e-06, "loss": 23.9714, "step": 45420 }, { "epoch": 0.56, "learning_rate": 8.589110639955209e-06, "loss": 23.7877, "step": 45430 }, { "epoch": 0.56, "learning_rate": 8.585167442483381e-06, "loss": 23.9846, "step": 45440 }, { "epoch": 0.56, "learning_rate": 8.58122446948195e-06, "loss": 23.8416, "step": 45450 }, { "epoch": 0.56, "learning_rate": 8.577281721576492e-06, "loss": 23.8361, "step": 45460 }, { "epoch": 0.56, "learning_rate": 8.573339199392543e-06, "loss": 24.0142, "step": 45470 }, { "epoch": 0.56, "learning_rate": 8.569396903555602e-06, "loss": 23.933, "step": 45480 }, { "epoch": 0.56, "learning_rate": 8.565454834691136e-06, "loss": 24.0476, "step": 45490 }, { "epoch": 0.56, "learning_rate": 8.561512993424575e-06, "loss": 23.8603, "step": 45500 }, { "epoch": 0.56, "learning_rate": 8.557571380381309e-06, "loss": 22.9959, "step": 45510 }, { "epoch": 0.56, "learning_rate": 8.5536299961867e-06, "loss": 23.8902, "step": 45520 }, { "epoch": 0.56, "learning_rate": 8.549688841466063e-06, "loss": 23.5909, "step": 45530 }, { "epoch": 0.56, "learning_rate": 8.545747916844688e-06, "loss": 23.8913, "step": 45540 }, { "epoch": 0.56, "learning_rate": 8.541807222947819e-06, "loss": 23.4294, "step": 45550 }, { "epoch": 0.56, "learning_rate": 8.53786676040067e-06, "loss": 24.3589, "step": 45560 }, { "epoch": 0.56, "learning_rate": 8.533926529828413e-06, "loss": 23.7553, "step": 45570 }, { "epoch": 0.56, "learning_rate": 8.529986531856187e-06, "loss": 23.7048, "step": 45580 }, { "epoch": 0.56, "learning_rate": 8.52604676710909e-06, "loss": 23.7909, "step": 45590 }, { "epoch": 0.56, "learning_rate": 8.522107236212192e-06, "loss": 23.3881, "step": 45600 }, { "epoch": 0.56, "learning_rate": 8.518167939790516e-06, "loss": 24.1461, "step": 45610 }, { "epoch": 0.56, "learning_rate": 8.51422887846905e-06, "loss": 24.0839, "step": 45620 }, { "epoch": 0.56, "learning_rate": 8.510290052872743e-06, "loss": 24.1066, "step": 45630 }, { "epoch": 0.56, "learning_rate": 8.506351463626519e-06, "loss": 23.4736, "step": 45640 }, { "epoch": 0.56, "learning_rate": 8.502413111355248e-06, "loss": 24.3566, "step": 45650 }, { "epoch": 0.56, "learning_rate": 8.498474996683771e-06, "loss": 23.5322, "step": 45660 }, { "epoch": 0.56, "learning_rate": 8.494537120236894e-06, "loss": 23.9427, "step": 45670 }, { "epoch": 0.56, "learning_rate": 8.490599482639375e-06, "loss": 23.2208, "step": 45680 }, { "epoch": 0.56, "learning_rate": 8.486662084515944e-06, "loss": 23.8178, "step": 45690 }, { "epoch": 0.56, "learning_rate": 8.482724926491289e-06, "loss": 23.7011, "step": 45700 }, { "epoch": 0.56, "learning_rate": 8.478788009190062e-06, "loss": 24.5863, "step": 45710 }, { "epoch": 0.56, "learning_rate": 8.474851333236871e-06, "loss": 23.9427, "step": 45720 }, { "epoch": 0.56, "learning_rate": 8.470914899256291e-06, "loss": 23.7051, "step": 45730 }, { "epoch": 0.56, "learning_rate": 8.466978707872861e-06, "loss": 23.7789, "step": 45740 }, { "epoch": 0.56, "learning_rate": 8.46304275971107e-06, "loss": 24.6065, "step": 45750 }, { "epoch": 0.56, "learning_rate": 8.459107055395383e-06, "loss": 23.3234, "step": 45760 }, { "epoch": 0.56, "learning_rate": 8.45517159555022e-06, "loss": 23.8059, "step": 45770 }, { "epoch": 0.56, "learning_rate": 8.451236380799958e-06, "loss": 24.178, "step": 45780 }, { "epoch": 0.56, "learning_rate": 8.447301411768939e-06, "loss": 23.7063, "step": 45790 }, { "epoch": 0.56, "learning_rate": 8.443366689081469e-06, "loss": 24.1495, "step": 45800 }, { "epoch": 0.56, "learning_rate": 8.439432213361807e-06, "loss": 23.4024, "step": 45810 }, { "epoch": 0.56, "learning_rate": 8.435497985234181e-06, "loss": 23.8663, "step": 45820 }, { "epoch": 0.56, "learning_rate": 8.43156400532278e-06, "loss": 24.1556, "step": 45830 }, { "epoch": 0.56, "learning_rate": 8.427630274251744e-06, "loss": 23.3397, "step": 45840 }, { "epoch": 0.56, "learning_rate": 8.423696792645183e-06, "loss": 23.2456, "step": 45850 }, { "epoch": 0.56, "learning_rate": 8.41976356112716e-06, "loss": 23.747, "step": 45860 }, { "epoch": 0.56, "learning_rate": 8.415830580321702e-06, "loss": 24.3582, "step": 45870 }, { "epoch": 0.56, "learning_rate": 8.411897850852803e-06, "loss": 23.5843, "step": 45880 }, { "epoch": 0.56, "learning_rate": 8.407965373344404e-06, "loss": 23.3145, "step": 45890 }, { "epoch": 0.56, "learning_rate": 8.404033148420416e-06, "loss": 24.1746, "step": 45900 }, { "epoch": 0.56, "learning_rate": 8.400101176704707e-06, "loss": 23.361, "step": 45910 }, { "epoch": 0.56, "learning_rate": 8.396169458821101e-06, "loss": 23.4778, "step": 45920 }, { "epoch": 0.56, "learning_rate": 8.392237995393386e-06, "loss": 23.7075, "step": 45930 }, { "epoch": 0.56, "learning_rate": 8.388306787045314e-06, "loss": 24.0502, "step": 45940 }, { "epoch": 0.57, "learning_rate": 8.384375834400585e-06, "loss": 23.5535, "step": 45950 }, { "epoch": 0.57, "learning_rate": 8.380445138082864e-06, "loss": 23.8817, "step": 45960 }, { "epoch": 0.57, "learning_rate": 8.376514698715784e-06, "loss": 24.1961, "step": 45970 }, { "epoch": 0.57, "learning_rate": 8.37258451692292e-06, "loss": 24.0104, "step": 45980 }, { "epoch": 0.57, "learning_rate": 8.36865459332782e-06, "loss": 23.7518, "step": 45990 }, { "epoch": 0.57, "learning_rate": 8.364724928553989e-06, "loss": 23.5923, "step": 46000 }, { "epoch": 0.57, "learning_rate": 8.360795523224882e-06, "loss": 24.058, "step": 46010 }, { "epoch": 0.57, "learning_rate": 8.356866377963923e-06, "loss": 24.0656, "step": 46020 }, { "epoch": 0.57, "learning_rate": 8.35293749339449e-06, "loss": 24.0389, "step": 46030 }, { "epoch": 0.57, "learning_rate": 8.349008870139922e-06, "loss": 24.2478, "step": 46040 }, { "epoch": 0.57, "learning_rate": 8.345080508823512e-06, "loss": 24.0597, "step": 46050 }, { "epoch": 0.57, "learning_rate": 8.341152410068517e-06, "loss": 23.9144, "step": 46060 }, { "epoch": 0.57, "learning_rate": 8.337224574498154e-06, "loss": 23.6907, "step": 46070 }, { "epoch": 0.57, "learning_rate": 8.333297002735588e-06, "loss": 23.7227, "step": 46080 }, { "epoch": 0.57, "learning_rate": 8.329369695403947e-06, "loss": 24.2799, "step": 46090 }, { "epoch": 0.57, "learning_rate": 8.32544265312632e-06, "loss": 23.7012, "step": 46100 }, { "epoch": 0.57, "learning_rate": 8.321515876525756e-06, "loss": 24.1032, "step": 46110 }, { "epoch": 0.57, "learning_rate": 8.317589366225255e-06, "loss": 24.1094, "step": 46120 }, { "epoch": 0.57, "learning_rate": 8.313663122847777e-06, "loss": 24.0499, "step": 46130 }, { "epoch": 0.57, "learning_rate": 8.309737147016247e-06, "loss": 23.4824, "step": 46140 }, { "epoch": 0.57, "learning_rate": 8.30581143935353e-06, "loss": 23.8269, "step": 46150 }, { "epoch": 0.57, "learning_rate": 8.301886000482466e-06, "loss": 23.2881, "step": 46160 }, { "epoch": 0.57, "learning_rate": 8.29796083102585e-06, "loss": 24.0299, "step": 46170 }, { "epoch": 0.57, "learning_rate": 8.294035931606422e-06, "loss": 24.7698, "step": 46180 }, { "epoch": 0.57, "learning_rate": 8.29011130284689e-06, "loss": 24.1031, "step": 46190 }, { "epoch": 0.57, "learning_rate": 8.286186945369922e-06, "loss": 24.0066, "step": 46200 }, { "epoch": 0.57, "learning_rate": 8.28226285979813e-06, "loss": 23.4669, "step": 46210 }, { "epoch": 0.57, "learning_rate": 8.278339046754091e-06, "loss": 23.5263, "step": 46220 }, { "epoch": 0.57, "learning_rate": 8.27441550686034e-06, "loss": 24.151, "step": 46230 }, { "epoch": 0.57, "learning_rate": 8.270492240739369e-06, "loss": 23.2917, "step": 46240 }, { "epoch": 0.57, "learning_rate": 8.266569249013619e-06, "loss": 24.2825, "step": 46250 }, { "epoch": 0.57, "learning_rate": 8.262646532305496e-06, "loss": 24.0771, "step": 46260 }, { "epoch": 0.57, "learning_rate": 8.25872409123736e-06, "loss": 22.7913, "step": 46270 }, { "epoch": 0.57, "learning_rate": 8.254801926431522e-06, "loss": 23.8408, "step": 46280 }, { "epoch": 0.57, "learning_rate": 8.250880038510258e-06, "loss": 24.0668, "step": 46290 }, { "epoch": 0.57, "learning_rate": 8.246958428095796e-06, "loss": 23.7816, "step": 46300 }, { "epoch": 0.57, "learning_rate": 8.243037095810314e-06, "loss": 23.7795, "step": 46310 }, { "epoch": 0.57, "learning_rate": 8.239116042275952e-06, "loss": 23.8581, "step": 46320 }, { "epoch": 0.57, "learning_rate": 8.235195268114808e-06, "loss": 23.7847, "step": 46330 }, { "epoch": 0.57, "learning_rate": 8.231274773948935e-06, "loss": 23.6757, "step": 46340 }, { "epoch": 0.57, "learning_rate": 8.227354560400333e-06, "loss": 23.9863, "step": 46350 }, { "epoch": 0.57, "learning_rate": 8.223434628090968e-06, "loss": 23.4866, "step": 46360 }, { "epoch": 0.57, "learning_rate": 8.21951497764276e-06, "loss": 23.6367, "step": 46370 }, { "epoch": 0.57, "learning_rate": 8.215595609677574e-06, "loss": 23.2904, "step": 46380 }, { "epoch": 0.57, "learning_rate": 8.211676524817244e-06, "loss": 24.1571, "step": 46390 }, { "epoch": 0.57, "learning_rate": 8.20775772368355e-06, "loss": 24.632, "step": 46400 }, { "epoch": 0.57, "learning_rate": 8.203839206898232e-06, "loss": 23.3853, "step": 46410 }, { "epoch": 0.57, "learning_rate": 8.19992097508298e-06, "loss": 24.1218, "step": 46420 }, { "epoch": 0.57, "learning_rate": 8.19600302885944e-06, "loss": 23.4375, "step": 46430 }, { "epoch": 0.57, "learning_rate": 8.192085368849223e-06, "loss": 23.3768, "step": 46440 }, { "epoch": 0.57, "learning_rate": 8.188167995673874e-06, "loss": 24.1949, "step": 46450 }, { "epoch": 0.57, "learning_rate": 8.18425090995491e-06, "loss": 23.3399, "step": 46460 }, { "epoch": 0.57, "learning_rate": 8.1803341123138e-06, "loss": 23.6455, "step": 46470 }, { "epoch": 0.57, "learning_rate": 8.176417603371958e-06, "loss": 24.5552, "step": 46480 }, { "epoch": 0.57, "learning_rate": 8.17250138375076e-06, "loss": 23.6951, "step": 46490 }, { "epoch": 0.57, "learning_rate": 8.168585454071536e-06, "loss": 23.8402, "step": 46500 }, { "epoch": 0.57, "learning_rate": 8.164669814955566e-06, "loss": 23.6932, "step": 46510 }, { "epoch": 0.57, "learning_rate": 8.160754467024086e-06, "loss": 24.2154, "step": 46520 }, { "epoch": 0.57, "learning_rate": 8.156839410898291e-06, "loss": 23.7413, "step": 46530 }, { "epoch": 0.57, "learning_rate": 8.152924647199317e-06, "loss": 23.5966, "step": 46540 }, { "epoch": 0.57, "learning_rate": 8.149010176548265e-06, "loss": 24.1225, "step": 46550 }, { "epoch": 0.57, "learning_rate": 8.145095999566184e-06, "loss": 23.5576, "step": 46560 }, { "epoch": 0.57, "learning_rate": 8.14118211687408e-06, "loss": 23.9906, "step": 46570 }, { "epoch": 0.57, "learning_rate": 8.137268529092912e-06, "loss": 24.1722, "step": 46580 }, { "epoch": 0.57, "learning_rate": 8.133355236843586e-06, "loss": 24.0567, "step": 46590 }, { "epoch": 0.57, "learning_rate": 8.129442240746969e-06, "loss": 23.8633, "step": 46600 }, { "epoch": 0.57, "learning_rate": 8.125529541423878e-06, "loss": 23.5919, "step": 46610 }, { "epoch": 0.57, "learning_rate": 8.121617139495081e-06, "loss": 24.202, "step": 46620 }, { "epoch": 0.57, "learning_rate": 8.117705035581302e-06, "loss": 23.8127, "step": 46630 }, { "epoch": 0.57, "learning_rate": 8.113793230303216e-06, "loss": 23.6135, "step": 46640 }, { "epoch": 0.57, "learning_rate": 8.109881724281449e-06, "loss": 24.5958, "step": 46650 }, { "epoch": 0.57, "learning_rate": 8.105970518136584e-06, "loss": 22.7759, "step": 46660 }, { "epoch": 0.57, "learning_rate": 8.102059612489153e-06, "loss": 23.2307, "step": 46670 }, { "epoch": 0.57, "learning_rate": 8.098149007959638e-06, "loss": 23.4842, "step": 46680 }, { "epoch": 0.57, "learning_rate": 8.09423870516848e-06, "loss": 24.0914, "step": 46690 }, { "epoch": 0.57, "learning_rate": 8.09032870473607e-06, "loss": 23.7656, "step": 46700 }, { "epoch": 0.57, "learning_rate": 8.086419007282745e-06, "loss": 23.4599, "step": 46710 }, { "epoch": 0.57, "learning_rate": 8.0825096134288e-06, "loss": 23.7992, "step": 46720 }, { "epoch": 0.57, "learning_rate": 8.078600523794483e-06, "loss": 24.1683, "step": 46730 }, { "epoch": 0.57, "learning_rate": 8.074691738999986e-06, "loss": 23.6858, "step": 46740 }, { "epoch": 0.57, "learning_rate": 8.07078325966546e-06, "loss": 24.2906, "step": 46750 }, { "epoch": 0.58, "learning_rate": 8.066875086411012e-06, "loss": 23.5811, "step": 46760 }, { "epoch": 0.58, "learning_rate": 8.062967219856682e-06, "loss": 23.6671, "step": 46770 }, { "epoch": 0.58, "learning_rate": 8.05905966062248e-06, "loss": 24.1025, "step": 46780 }, { "epoch": 0.58, "learning_rate": 8.055152409328356e-06, "loss": 23.9648, "step": 46790 }, { "epoch": 0.58, "learning_rate": 8.051245466594218e-06, "loss": 23.9326, "step": 46800 }, { "epoch": 0.58, "learning_rate": 8.047338833039923e-06, "loss": 24.347, "step": 46810 }, { "epoch": 0.58, "learning_rate": 8.043432509285276e-06, "loss": 23.5191, "step": 46820 }, { "epoch": 0.58, "learning_rate": 8.039526495950036e-06, "loss": 23.5423, "step": 46830 }, { "epoch": 0.58, "learning_rate": 8.035620793653915e-06, "loss": 23.4046, "step": 46840 }, { "epoch": 0.58, "learning_rate": 8.03171540301657e-06, "loss": 24.35, "step": 46850 }, { "epoch": 0.58, "learning_rate": 8.027810324657607e-06, "loss": 24.4932, "step": 46860 }, { "epoch": 0.58, "learning_rate": 8.023905559196596e-06, "loss": 23.4794, "step": 46870 }, { "epoch": 0.58, "learning_rate": 8.020001107253039e-06, "loss": 24.0382, "step": 46880 }, { "epoch": 0.58, "learning_rate": 8.016096969446403e-06, "loss": 23.6894, "step": 46890 }, { "epoch": 0.58, "learning_rate": 8.012193146396098e-06, "loss": 23.7063, "step": 46900 }, { "epoch": 0.58, "learning_rate": 8.008289638721483e-06, "loss": 23.6984, "step": 46910 }, { "epoch": 0.58, "learning_rate": 8.00438644704187e-06, "loss": 23.3236, "step": 46920 }, { "epoch": 0.58, "learning_rate": 8.000483571976522e-06, "loss": 23.8698, "step": 46930 }, { "epoch": 0.58, "learning_rate": 7.996581014144653e-06, "loss": 24.0167, "step": 46940 }, { "epoch": 0.58, "learning_rate": 7.992678774165415e-06, "loss": 23.6839, "step": 46950 }, { "epoch": 0.58, "learning_rate": 7.988776852657925e-06, "loss": 24.5401, "step": 46960 }, { "epoch": 0.58, "learning_rate": 7.984875250241245e-06, "loss": 23.4441, "step": 46970 }, { "epoch": 0.58, "learning_rate": 7.980973967534375e-06, "loss": 23.3416, "step": 46980 }, { "epoch": 0.58, "learning_rate": 7.977073005156279e-06, "loss": 23.8657, "step": 46990 }, { "epoch": 0.58, "learning_rate": 7.97317236372587e-06, "loss": 24.0167, "step": 47000 }, { "epoch": 0.58, "learning_rate": 7.969272043861997e-06, "loss": 24.2259, "step": 47010 }, { "epoch": 0.58, "learning_rate": 7.965372046183464e-06, "loss": 23.6857, "step": 47020 }, { "epoch": 0.58, "learning_rate": 7.961472371309029e-06, "loss": 24.1227, "step": 47030 }, { "epoch": 0.58, "learning_rate": 7.957573019857398e-06, "loss": 23.9444, "step": 47040 }, { "epoch": 0.58, "learning_rate": 7.953673992447218e-06, "loss": 24.116, "step": 47050 }, { "epoch": 0.58, "learning_rate": 7.949775289697092e-06, "loss": 23.0318, "step": 47060 }, { "epoch": 0.58, "learning_rate": 7.94587691222557e-06, "loss": 24.0276, "step": 47070 }, { "epoch": 0.58, "learning_rate": 7.941978860651148e-06, "loss": 23.3023, "step": 47080 }, { "epoch": 0.58, "learning_rate": 7.938081135592271e-06, "loss": 24.0336, "step": 47090 }, { "epoch": 0.58, "learning_rate": 7.934183737667334e-06, "loss": 23.806, "step": 47100 }, { "epoch": 0.58, "learning_rate": 7.930286667494681e-06, "loss": 23.8526, "step": 47110 }, { "epoch": 0.58, "learning_rate": 7.926389925692598e-06, "loss": 23.5807, "step": 47120 }, { "epoch": 0.58, "learning_rate": 7.922493512879327e-06, "loss": 24.1392, "step": 47130 }, { "epoch": 0.58, "learning_rate": 7.918597429673052e-06, "loss": 24.346, "step": 47140 }, { "epoch": 0.58, "learning_rate": 7.914701676691904e-06, "loss": 23.5003, "step": 47150 }, { "epoch": 0.58, "learning_rate": 7.910806254553969e-06, "loss": 23.6665, "step": 47160 }, { "epoch": 0.58, "learning_rate": 7.906911163877274e-06, "loss": 23.5884, "step": 47170 }, { "epoch": 0.58, "learning_rate": 7.90301640527979e-06, "loss": 23.0732, "step": 47180 }, { "epoch": 0.58, "learning_rate": 7.899121979379446e-06, "loss": 24.1375, "step": 47190 }, { "epoch": 0.58, "learning_rate": 7.895227886794113e-06, "loss": 23.6374, "step": 47200 }, { "epoch": 0.58, "learning_rate": 7.891334128141604e-06, "loss": 24.4974, "step": 47210 }, { "epoch": 0.58, "learning_rate": 7.887440704039686e-06, "loss": 23.8255, "step": 47220 }, { "epoch": 0.58, "learning_rate": 7.883547615106075e-06, "loss": 24.5768, "step": 47230 }, { "epoch": 0.58, "learning_rate": 7.879654861958423e-06, "loss": 24.1021, "step": 47240 }, { "epoch": 0.58, "learning_rate": 7.875762445214335e-06, "loss": 24.1, "step": 47250 }, { "epoch": 0.58, "learning_rate": 7.871870365491365e-06, "loss": 24.0795, "step": 47260 }, { "epoch": 0.58, "learning_rate": 7.867978623407013e-06, "loss": 24.0441, "step": 47270 }, { "epoch": 0.58, "learning_rate": 7.864087219578722e-06, "loss": 23.7241, "step": 47280 }, { "epoch": 0.58, "learning_rate": 7.860196154623881e-06, "loss": 23.5406, "step": 47290 }, { "epoch": 0.58, "learning_rate": 7.85630542915983e-06, "loss": 23.3568, "step": 47300 }, { "epoch": 0.58, "learning_rate": 7.852415043803855e-06, "loss": 23.3045, "step": 47310 }, { "epoch": 0.58, "learning_rate": 7.84852499917318e-06, "loss": 23.6398, "step": 47320 }, { "epoch": 0.58, "learning_rate": 7.84463529588498e-06, "loss": 23.4066, "step": 47330 }, { "epoch": 0.58, "learning_rate": 7.840745934556385e-06, "loss": 23.8551, "step": 47340 }, { "epoch": 0.58, "learning_rate": 7.836856915804451e-06, "loss": 24.2572, "step": 47350 }, { "epoch": 0.58, "learning_rate": 7.832968240246196e-06, "loss": 23.4035, "step": 47360 }, { "epoch": 0.58, "learning_rate": 7.829079908498581e-06, "loss": 24.2633, "step": 47370 }, { "epoch": 0.58, "learning_rate": 7.825191921178503e-06, "loss": 23.5053, "step": 47380 }, { "epoch": 0.58, "learning_rate": 7.821304278902815e-06, "loss": 23.9371, "step": 47390 }, { "epoch": 0.58, "learning_rate": 7.817416982288313e-06, "loss": 23.7981, "step": 47400 }, { "epoch": 0.58, "learning_rate": 7.813530031951732e-06, "loss": 24.1956, "step": 47410 }, { "epoch": 0.58, "learning_rate": 7.809643428509758e-06, "loss": 23.5919, "step": 47420 }, { "epoch": 0.58, "learning_rate": 7.805757172579024e-06, "loss": 23.6895, "step": 47430 }, { "epoch": 0.58, "learning_rate": 7.8018712647761e-06, "loss": 23.7894, "step": 47440 }, { "epoch": 0.58, "learning_rate": 7.797985705717507e-06, "loss": 23.5626, "step": 47450 }, { "epoch": 0.58, "learning_rate": 7.794100496019711e-06, "loss": 23.2743, "step": 47460 }, { "epoch": 0.58, "learning_rate": 7.790215636299114e-06, "loss": 23.903, "step": 47470 }, { "epoch": 0.58, "learning_rate": 7.786331127172075e-06, "loss": 23.2118, "step": 47480 }, { "epoch": 0.58, "learning_rate": 7.782446969254885e-06, "loss": 23.6035, "step": 47490 }, { "epoch": 0.58, "learning_rate": 7.77856316316379e-06, "loss": 24.2714, "step": 47500 }, { "epoch": 0.58, "learning_rate": 7.774679709514976e-06, "loss": 24.0865, "step": 47510 }, { "epoch": 0.58, "learning_rate": 7.77079660892457e-06, "loss": 23.6916, "step": 47520 }, { "epoch": 0.58, "learning_rate": 7.766913862008646e-06, "loss": 23.9825, "step": 47530 }, { "epoch": 0.58, "learning_rate": 7.763031469383226e-06, "loss": 23.8004, "step": 47540 }, { "epoch": 0.58, "learning_rate": 7.759149431664265e-06, "loss": 23.6677, "step": 47550 }, { "epoch": 0.58, "learning_rate": 7.755267749467674e-06, "loss": 23.571, "step": 47560 }, { "epoch": 0.59, "learning_rate": 7.751386423409297e-06, "loss": 23.8866, "step": 47570 }, { "epoch": 0.59, "learning_rate": 7.74750545410493e-06, "loss": 23.3885, "step": 47580 }, { "epoch": 0.59, "learning_rate": 7.743624842170303e-06, "loss": 23.5704, "step": 47590 }, { "epoch": 0.59, "learning_rate": 7.7397445882211e-06, "loss": 23.8402, "step": 47600 }, { "epoch": 0.59, "learning_rate": 7.735864692872942e-06, "loss": 22.9426, "step": 47610 }, { "epoch": 0.59, "learning_rate": 7.731985156741392e-06, "loss": 23.9378, "step": 47620 }, { "epoch": 0.59, "learning_rate": 7.728105980441963e-06, "loss": 23.7174, "step": 47630 }, { "epoch": 0.59, "learning_rate": 7.724227164590101e-06, "loss": 23.9509, "step": 47640 }, { "epoch": 0.59, "learning_rate": 7.720348709801202e-06, "loss": 23.4814, "step": 47650 }, { "epoch": 0.59, "learning_rate": 7.716470616690602e-06, "loss": 24.3656, "step": 47660 }, { "epoch": 0.59, "learning_rate": 7.712592885873581e-06, "loss": 23.8629, "step": 47670 }, { "epoch": 0.59, "learning_rate": 7.708715517965362e-06, "loss": 23.6471, "step": 47680 }, { "epoch": 0.59, "learning_rate": 7.70483851358111e-06, "loss": 23.8057, "step": 47690 }, { "epoch": 0.59, "learning_rate": 7.700961873335925e-06, "loss": 23.1918, "step": 47700 }, { "epoch": 0.59, "learning_rate": 7.697085597844862e-06, "loss": 23.6966, "step": 47710 }, { "epoch": 0.59, "learning_rate": 7.693209687722908e-06, "loss": 23.2198, "step": 47720 }, { "epoch": 0.59, "learning_rate": 7.689334143584997e-06, "loss": 23.9141, "step": 47730 }, { "epoch": 0.59, "learning_rate": 7.685458966046007e-06, "loss": 24.0462, "step": 47740 }, { "epoch": 0.59, "learning_rate": 7.68158415572075e-06, "loss": 24.5833, "step": 47750 }, { "epoch": 0.59, "learning_rate": 7.677709713223987e-06, "loss": 23.7887, "step": 47760 }, { "epoch": 0.59, "learning_rate": 7.67383563917042e-06, "loss": 23.5987, "step": 47770 }, { "epoch": 0.59, "learning_rate": 7.669961934174685e-06, "loss": 24.4164, "step": 47780 }, { "epoch": 0.59, "learning_rate": 7.66608859885137e-06, "loss": 23.9991, "step": 47790 }, { "epoch": 0.59, "learning_rate": 7.662215633814999e-06, "loss": 23.6073, "step": 47800 }, { "epoch": 0.59, "learning_rate": 7.658343039680034e-06, "loss": 23.3657, "step": 47810 }, { "epoch": 0.59, "learning_rate": 7.654470817060885e-06, "loss": 23.9809, "step": 47820 }, { "epoch": 0.59, "learning_rate": 7.650598966571899e-06, "loss": 24.5931, "step": 47830 }, { "epoch": 0.59, "learning_rate": 7.646727488827365e-06, "loss": 24.3702, "step": 47840 }, { "epoch": 0.59, "learning_rate": 7.642856384441513e-06, "loss": 23.3034, "step": 47850 }, { "epoch": 0.59, "learning_rate": 7.638985654028511e-06, "loss": 24.5318, "step": 47860 }, { "epoch": 0.59, "learning_rate": 7.635115298202476e-06, "loss": 23.9507, "step": 47870 }, { "epoch": 0.59, "learning_rate": 7.631245317577452e-06, "loss": 23.9499, "step": 47880 }, { "epoch": 0.59, "learning_rate": 7.627375712767436e-06, "loss": 24.49, "step": 47890 }, { "epoch": 0.59, "learning_rate": 7.6235064843863615e-06, "loss": 23.7711, "step": 47900 }, { "epoch": 0.59, "learning_rate": 7.619637633048097e-06, "loss": 23.682, "step": 47910 }, { "epoch": 0.59, "learning_rate": 7.615769159366462e-06, "loss": 24.4052, "step": 47920 }, { "epoch": 0.59, "learning_rate": 7.611901063955201e-06, "loss": 24.0502, "step": 47930 }, { "epoch": 0.59, "learning_rate": 7.608033347428013e-06, "loss": 23.9773, "step": 47940 }, { "epoch": 0.59, "learning_rate": 7.604166010398528e-06, "loss": 24.1286, "step": 47950 }, { "epoch": 0.59, "learning_rate": 7.60029905348032e-06, "loss": 24.074, "step": 47960 }, { "epoch": 0.59, "learning_rate": 7.596432477286902e-06, "loss": 23.9759, "step": 47970 }, { "epoch": 0.59, "learning_rate": 7.592566282431726e-06, "loss": 23.5174, "step": 47980 }, { "epoch": 0.59, "learning_rate": 7.588700469528183e-06, "loss": 24.077, "step": 47990 }, { "epoch": 0.59, "learning_rate": 7.584835039189602e-06, "loss": 23.7558, "step": 48000 }, { "epoch": 0.59, "learning_rate": 7.580969992029259e-06, "loss": 23.7497, "step": 48010 }, { "epoch": 0.59, "learning_rate": 7.577105328660357e-06, "loss": 24.0257, "step": 48020 }, { "epoch": 0.59, "learning_rate": 7.573241049696046e-06, "loss": 23.8583, "step": 48030 }, { "epoch": 0.59, "learning_rate": 7.569377155749418e-06, "loss": 23.8937, "step": 48040 }, { "epoch": 0.59, "learning_rate": 7.565513647433494e-06, "loss": 23.5046, "step": 48050 }, { "epoch": 0.59, "learning_rate": 7.561650525361242e-06, "loss": 23.9674, "step": 48060 }, { "epoch": 0.59, "learning_rate": 7.5577877901455675e-06, "loss": 24.0947, "step": 48070 }, { "epoch": 0.59, "learning_rate": 7.553925442399308e-06, "loss": 24.1136, "step": 48080 }, { "epoch": 0.59, "learning_rate": 7.550063482735249e-06, "loss": 23.8162, "step": 48090 }, { "epoch": 0.59, "learning_rate": 7.546201911766111e-06, "loss": 23.54, "step": 48100 }, { "epoch": 0.59, "learning_rate": 7.542340730104547e-06, "loss": 23.4793, "step": 48110 }, { "epoch": 0.59, "learning_rate": 7.538479938363157e-06, "loss": 23.4905, "step": 48120 }, { "epoch": 0.59, "learning_rate": 7.534619537154475e-06, "loss": 23.6536, "step": 48130 }, { "epoch": 0.59, "learning_rate": 7.5307595270909716e-06, "loss": 23.8981, "step": 48140 }, { "epoch": 0.59, "learning_rate": 7.526899908785062e-06, "loss": 23.2717, "step": 48150 }, { "epoch": 0.59, "learning_rate": 7.523040682849087e-06, "loss": 23.5146, "step": 48160 }, { "epoch": 0.59, "learning_rate": 7.519181849895334e-06, "loss": 23.9535, "step": 48170 }, { "epoch": 0.59, "learning_rate": 7.515323410536033e-06, "loss": 24.1933, "step": 48180 }, { "epoch": 0.59, "learning_rate": 7.511465365383337e-06, "loss": 24.4022, "step": 48190 }, { "epoch": 0.59, "learning_rate": 7.507607715049347e-06, "loss": 23.7318, "step": 48200 }, { "epoch": 0.59, "learning_rate": 7.503750460146104e-06, "loss": 24.1788, "step": 48210 }, { "epoch": 0.59, "learning_rate": 7.499893601285573e-06, "loss": 24.439, "step": 48220 }, { "epoch": 0.59, "learning_rate": 7.49603713907967e-06, "loss": 24.4391, "step": 48230 }, { "epoch": 0.59, "learning_rate": 7.492181074140242e-06, "loss": 23.6923, "step": 48240 }, { "epoch": 0.59, "learning_rate": 7.488325407079071e-06, "loss": 23.922, "step": 48250 }, { "epoch": 0.59, "learning_rate": 7.484470138507879e-06, "loss": 23.9877, "step": 48260 }, { "epoch": 0.59, "learning_rate": 7.480615269038326e-06, "loss": 23.5259, "step": 48270 }, { "epoch": 0.59, "learning_rate": 7.4767607992820034e-06, "loss": 23.5216, "step": 48280 }, { "epoch": 0.59, "learning_rate": 7.472906729850444e-06, "loss": 24.4242, "step": 48290 }, { "epoch": 0.59, "learning_rate": 7.469053061355118e-06, "loss": 24.0989, "step": 48300 }, { "epoch": 0.59, "learning_rate": 7.465199794407425e-06, "loss": 23.5601, "step": 48310 }, { "epoch": 0.59, "learning_rate": 7.461346929618708e-06, "loss": 23.5901, "step": 48320 }, { "epoch": 0.59, "learning_rate": 7.457494467600244e-06, "loss": 23.9816, "step": 48330 }, { "epoch": 0.59, "learning_rate": 7.453642408963246e-06, "loss": 23.8751, "step": 48340 }, { "epoch": 0.59, "learning_rate": 7.44979075431886e-06, "loss": 24.4394, "step": 48350 }, { "epoch": 0.59, "learning_rate": 7.445939504278173e-06, "loss": 24.0186, "step": 48360 }, { "epoch": 0.59, "learning_rate": 7.442088659452207e-06, "loss": 23.4548, "step": 48370 }, { "epoch": 0.59, "learning_rate": 7.438238220451916e-06, "loss": 23.3336, "step": 48380 }, { "epoch": 0.6, "learning_rate": 7.434388187888191e-06, "loss": 23.705, "step": 48390 }, { "epoch": 0.6, "learning_rate": 7.430538562371859e-06, "loss": 24.0009, "step": 48400 }, { "epoch": 0.6, "learning_rate": 7.4266893445136855e-06, "loss": 24.205, "step": 48410 }, { "epoch": 0.6, "learning_rate": 7.422840534924365e-06, "loss": 24.2305, "step": 48420 }, { "epoch": 0.6, "learning_rate": 7.418992134214533e-06, "loss": 23.7183, "step": 48430 }, { "epoch": 0.6, "learning_rate": 7.41514414299476e-06, "loss": 23.6747, "step": 48440 }, { "epoch": 0.6, "learning_rate": 7.411296561875545e-06, "loss": 24.3998, "step": 48450 }, { "epoch": 0.6, "learning_rate": 7.407449391467327e-06, "loss": 24.1241, "step": 48460 }, { "epoch": 0.6, "learning_rate": 7.403602632380485e-06, "loss": 24.3642, "step": 48470 }, { "epoch": 0.6, "learning_rate": 7.399756285225319e-06, "loss": 23.6463, "step": 48480 }, { "epoch": 0.6, "learning_rate": 7.395910350612075e-06, "loss": 24.4252, "step": 48490 }, { "epoch": 0.6, "learning_rate": 7.39206482915093e-06, "loss": 23.2586, "step": 48500 }, { "epoch": 0.6, "learning_rate": 7.388219721451999e-06, "loss": 23.7048, "step": 48510 }, { "epoch": 0.6, "learning_rate": 7.3843750281253215e-06, "loss": 24.5571, "step": 48520 }, { "epoch": 0.6, "learning_rate": 7.380530749780881e-06, "loss": 24.3611, "step": 48530 }, { "epoch": 0.6, "learning_rate": 7.3766868870285946e-06, "loss": 24.2604, "step": 48540 }, { "epoch": 0.6, "learning_rate": 7.372843440478307e-06, "loss": 23.9574, "step": 48550 }, { "epoch": 0.6, "learning_rate": 7.369000410739799e-06, "loss": 24.1557, "step": 48560 }, { "epoch": 0.6, "learning_rate": 7.365157798422793e-06, "loss": 23.9059, "step": 48570 }, { "epoch": 0.6, "learning_rate": 7.3613156041369335e-06, "loss": 23.7667, "step": 48580 }, { "epoch": 0.6, "learning_rate": 7.357473828491806e-06, "loss": 23.9335, "step": 48590 }, { "epoch": 0.6, "learning_rate": 7.35363247209693e-06, "loss": 23.5443, "step": 48600 }, { "epoch": 0.6, "learning_rate": 7.349791535561752e-06, "loss": 24.3294, "step": 48610 }, { "epoch": 0.6, "learning_rate": 7.345951019495663e-06, "loss": 24.4387, "step": 48620 }, { "epoch": 0.6, "learning_rate": 7.3421109245079705e-06, "loss": 24.1339, "step": 48630 }, { "epoch": 0.6, "learning_rate": 7.3382712512079336e-06, "loss": 24.1163, "step": 48640 }, { "epoch": 0.6, "learning_rate": 7.33443200020473e-06, "loss": 23.6665, "step": 48650 }, { "epoch": 0.6, "learning_rate": 7.330593172107481e-06, "loss": 24.0127, "step": 48660 }, { "epoch": 0.6, "learning_rate": 7.326754767525232e-06, "loss": 23.9718, "step": 48670 }, { "epoch": 0.6, "learning_rate": 7.322916787066973e-06, "loss": 23.8499, "step": 48680 }, { "epoch": 0.6, "learning_rate": 7.319079231341611e-06, "loss": 24.0461, "step": 48690 }, { "epoch": 0.6, "learning_rate": 7.3152421009579965e-06, "loss": 23.8518, "step": 48700 }, { "epoch": 0.6, "learning_rate": 7.311405396524914e-06, "loss": 24.0017, "step": 48710 }, { "epoch": 0.6, "learning_rate": 7.307569118651068e-06, "loss": 23.7776, "step": 48720 }, { "epoch": 0.6, "learning_rate": 7.30373326794511e-06, "loss": 23.6486, "step": 48730 }, { "epoch": 0.6, "learning_rate": 7.299897845015617e-06, "loss": 23.8259, "step": 48740 }, { "epoch": 0.6, "learning_rate": 7.296062850471094e-06, "loss": 23.8296, "step": 48750 }, { "epoch": 0.6, "learning_rate": 7.292228284919985e-06, "loss": 23.6789, "step": 48760 }, { "epoch": 0.6, "learning_rate": 7.2883941489706675e-06, "loss": 23.5513, "step": 48770 }, { "epoch": 0.6, "learning_rate": 7.2845604432314385e-06, "loss": 23.8732, "step": 48780 }, { "epoch": 0.6, "learning_rate": 7.280727168310541e-06, "loss": 23.8618, "step": 48790 }, { "epoch": 0.6, "learning_rate": 7.276894324816144e-06, "loss": 23.3508, "step": 48800 }, { "epoch": 0.6, "learning_rate": 7.273061913356342e-06, "loss": 23.8462, "step": 48810 }, { "epoch": 0.6, "learning_rate": 7.2692299345391705e-06, "loss": 24.5055, "step": 48820 }, { "epoch": 0.6, "learning_rate": 7.265398388972595e-06, "loss": 24.2575, "step": 48830 }, { "epoch": 0.6, "learning_rate": 7.261567277264504e-06, "loss": 23.9508, "step": 48840 }, { "epoch": 0.6, "learning_rate": 7.257736600022732e-06, "loss": 23.3868, "step": 48850 }, { "epoch": 0.6, "learning_rate": 7.2539063578550215e-06, "loss": 24.0956, "step": 48860 }, { "epoch": 0.6, "learning_rate": 7.25007655136907e-06, "loss": 24.0366, "step": 48870 }, { "epoch": 0.6, "learning_rate": 7.246247181172494e-06, "loss": 24.4303, "step": 48880 }, { "epoch": 0.6, "learning_rate": 7.242418247872842e-06, "loss": 23.7067, "step": 48890 }, { "epoch": 0.6, "learning_rate": 7.238589752077593e-06, "loss": 23.4571, "step": 48900 }, { "epoch": 0.6, "learning_rate": 7.23476169439416e-06, "loss": 23.6293, "step": 48910 }, { "epoch": 0.6, "learning_rate": 7.230934075429881e-06, "loss": 24.3306, "step": 48920 }, { "epoch": 0.6, "learning_rate": 7.227106895792027e-06, "loss": 23.4309, "step": 48930 }, { "epoch": 0.6, "learning_rate": 7.223280156087804e-06, "loss": 23.6554, "step": 48940 }, { "epoch": 0.6, "learning_rate": 7.219453856924339e-06, "loss": 23.1158, "step": 48950 }, { "epoch": 0.6, "learning_rate": 7.2156279989086945e-06, "loss": 24.1138, "step": 48960 }, { "epoch": 0.6, "learning_rate": 7.211802582647866e-06, "loss": 24.0617, "step": 48970 }, { "epoch": 0.6, "learning_rate": 7.207977608748771e-06, "loss": 23.7699, "step": 48980 }, { "epoch": 0.6, "learning_rate": 7.204153077818263e-06, "loss": 23.7826, "step": 48990 }, { "epoch": 0.6, "learning_rate": 7.200328990463126e-06, "loss": 23.3744, "step": 49000 }, { "epoch": 0.6, "learning_rate": 7.1965053472900635e-06, "loss": 23.7101, "step": 49010 }, { "epoch": 0.6, "learning_rate": 7.192682148905723e-06, "loss": 23.678, "step": 49020 }, { "epoch": 0.6, "learning_rate": 7.188859395916674e-06, "loss": 23.4936, "step": 49030 }, { "epoch": 0.6, "learning_rate": 7.185037088929411e-06, "loss": 23.2994, "step": 49040 }, { "epoch": 0.6, "learning_rate": 7.1812152285503645e-06, "loss": 23.7243, "step": 49050 }, { "epoch": 0.6, "learning_rate": 7.177393815385895e-06, "loss": 23.2397, "step": 49060 }, { "epoch": 0.6, "learning_rate": 7.173572850042287e-06, "loss": 24.0385, "step": 49070 }, { "epoch": 0.6, "learning_rate": 7.16975233312576e-06, "loss": 23.319, "step": 49080 }, { "epoch": 0.6, "learning_rate": 7.16593226524245e-06, "loss": 23.2536, "step": 49090 }, { "epoch": 0.6, "learning_rate": 7.162112646998436e-06, "loss": 24.3724, "step": 49100 }, { "epoch": 0.6, "learning_rate": 7.158293478999719e-06, "loss": 24.0861, "step": 49110 }, { "epoch": 0.6, "learning_rate": 7.154474761852228e-06, "loss": 23.804, "step": 49120 }, { "epoch": 0.6, "learning_rate": 7.1506564961618245e-06, "loss": 23.7617, "step": 49130 }, { "epoch": 0.6, "learning_rate": 7.146838682534296e-06, "loss": 23.5313, "step": 49140 }, { "epoch": 0.6, "learning_rate": 7.143021321575354e-06, "loss": 23.7486, "step": 49150 }, { "epoch": 0.6, "learning_rate": 7.139204413890646e-06, "loss": 23.9681, "step": 49160 }, { "epoch": 0.6, "learning_rate": 7.135387960085744e-06, "loss": 24.0729, "step": 49170 }, { "epoch": 0.6, "learning_rate": 7.131571960766145e-06, "loss": 23.891, "step": 49180 }, { "epoch": 0.6, "learning_rate": 7.127756416537279e-06, "loss": 23.8263, "step": 49190 }, { "epoch": 0.61, "learning_rate": 7.1239413280045e-06, "loss": 23.7391, "step": 49200 }, { "epoch": 0.61, "learning_rate": 7.120126695773095e-06, "loss": 24.0107, "step": 49210 }, { "epoch": 0.61, "learning_rate": 7.11631252044827e-06, "loss": 23.9902, "step": 49220 }, { "epoch": 0.61, "learning_rate": 7.112498802635166e-06, "loss": 23.9192, "step": 49230 }, { "epoch": 0.61, "learning_rate": 7.108685542938849e-06, "loss": 23.7658, "step": 49240 }, { "epoch": 0.61, "learning_rate": 7.10487274196431e-06, "loss": 23.7877, "step": 49250 }, { "epoch": 0.61, "learning_rate": 7.101060400316471e-06, "loss": 23.587, "step": 49260 }, { "epoch": 0.61, "learning_rate": 7.097248518600182e-06, "loss": 24.7038, "step": 49270 }, { "epoch": 0.61, "learning_rate": 7.0934370974202104e-06, "loss": 24.4974, "step": 49280 }, { "epoch": 0.61, "learning_rate": 7.089626137381262e-06, "loss": 23.8304, "step": 49290 }, { "epoch": 0.61, "learning_rate": 7.085815639087967e-06, "loss": 23.9634, "step": 49300 }, { "epoch": 0.61, "learning_rate": 7.082005603144881e-06, "loss": 24.1525, "step": 49310 }, { "epoch": 0.61, "learning_rate": 7.078196030156479e-06, "loss": 23.615, "step": 49320 }, { "epoch": 0.61, "learning_rate": 7.074386920727173e-06, "loss": 23.5268, "step": 49330 }, { "epoch": 0.61, "learning_rate": 7.0705782754613e-06, "loss": 23.84, "step": 49340 }, { "epoch": 0.61, "learning_rate": 7.0667700949631146e-06, "loss": 24.0665, "step": 49350 }, { "epoch": 0.61, "learning_rate": 7.062962379836809e-06, "loss": 23.171, "step": 49360 }, { "epoch": 0.61, "learning_rate": 7.059155130686497e-06, "loss": 24.0748, "step": 49370 }, { "epoch": 0.61, "learning_rate": 7.055348348116218e-06, "loss": 23.8599, "step": 49380 }, { "epoch": 0.61, "learning_rate": 7.0515420327299325e-06, "loss": 23.7431, "step": 49390 }, { "epoch": 0.61, "learning_rate": 7.047736185131537e-06, "loss": 23.729, "step": 49400 }, { "epoch": 0.61, "learning_rate": 7.04393080592485e-06, "loss": 23.473, "step": 49410 }, { "epoch": 0.61, "learning_rate": 7.040125895713608e-06, "loss": 24.1059, "step": 49420 }, { "epoch": 0.61, "learning_rate": 7.036321455101483e-06, "loss": 24.3496, "step": 49430 }, { "epoch": 0.61, "learning_rate": 7.032517484692072e-06, "loss": 24.084, "step": 49440 }, { "epoch": 0.61, "learning_rate": 7.028713985088889e-06, "loss": 24.0983, "step": 49450 }, { "epoch": 0.61, "learning_rate": 7.024910956895379e-06, "loss": 24.1254, "step": 49460 }, { "epoch": 0.61, "learning_rate": 7.021108400714917e-06, "loss": 24.5819, "step": 49470 }, { "epoch": 0.61, "learning_rate": 7.017306317150793e-06, "loss": 23.6486, "step": 49480 }, { "epoch": 0.61, "learning_rate": 7.013504706806227e-06, "loss": 23.5499, "step": 49490 }, { "epoch": 0.61, "learning_rate": 7.009703570284368e-06, "loss": 23.9547, "step": 49500 }, { "epoch": 0.61, "learning_rate": 7.00590290818828e-06, "loss": 23.8127, "step": 49510 }, { "epoch": 0.61, "learning_rate": 7.0021027211209625e-06, "loss": 23.9575, "step": 49520 }, { "epoch": 0.61, "learning_rate": 6.9983030096853324e-06, "loss": 23.6415, "step": 49530 }, { "epoch": 0.61, "learning_rate": 6.9945037744842335e-06, "loss": 24.134, "step": 49540 }, { "epoch": 0.61, "learning_rate": 6.990705016120435e-06, "loss": 23.1519, "step": 49550 }, { "epoch": 0.61, "learning_rate": 6.986906735196625e-06, "loss": 24.1356, "step": 49560 }, { "epoch": 0.61, "learning_rate": 6.983108932315423e-06, "loss": 23.5233, "step": 49570 }, { "epoch": 0.61, "learning_rate": 6.979311608079371e-06, "loss": 23.5023, "step": 49580 }, { "epoch": 0.61, "learning_rate": 6.975514763090931e-06, "loss": 24.0529, "step": 49590 }, { "epoch": 0.61, "learning_rate": 6.971718397952495e-06, "loss": 24.1653, "step": 49600 }, { "epoch": 0.61, "learning_rate": 6.967922513266375e-06, "loss": 23.4404, "step": 49610 }, { "epoch": 0.61, "learning_rate": 6.964127109634805e-06, "loss": 23.8148, "step": 49620 }, { "epoch": 0.61, "learning_rate": 6.960332187659946e-06, "loss": 23.861, "step": 49630 }, { "epoch": 0.61, "learning_rate": 6.956537747943885e-06, "loss": 23.8538, "step": 49640 }, { "epoch": 0.61, "learning_rate": 6.9527437910886256e-06, "loss": 24.2514, "step": 49650 }, { "epoch": 0.61, "learning_rate": 6.948950317696099e-06, "loss": 23.1813, "step": 49660 }, { "epoch": 0.61, "learning_rate": 6.945157328368161e-06, "loss": 23.6613, "step": 49670 }, { "epoch": 0.61, "learning_rate": 6.9413648237065865e-06, "loss": 23.4907, "step": 49680 }, { "epoch": 0.61, "learning_rate": 6.9375728043130756e-06, "loss": 23.5401, "step": 49690 }, { "epoch": 0.61, "learning_rate": 6.933781270789255e-06, "loss": 23.7178, "step": 49700 }, { "epoch": 0.61, "learning_rate": 6.929990223736665e-06, "loss": 23.8986, "step": 49710 }, { "epoch": 0.61, "learning_rate": 6.926199663756778e-06, "loss": 23.3893, "step": 49720 }, { "epoch": 0.61, "learning_rate": 6.922409591450989e-06, "loss": 24.097, "step": 49730 }, { "epoch": 0.61, "learning_rate": 6.918620007420606e-06, "loss": 23.8697, "step": 49740 }, { "epoch": 0.61, "learning_rate": 6.914830912266868e-06, "loss": 24.2396, "step": 49750 }, { "epoch": 0.61, "learning_rate": 6.911042306590935e-06, "loss": 23.9165, "step": 49760 }, { "epoch": 0.61, "learning_rate": 6.907254190993893e-06, "loss": 23.7517, "step": 49770 }, { "epoch": 0.61, "learning_rate": 6.903466566076739e-06, "loss": 23.7408, "step": 49780 }, { "epoch": 0.61, "learning_rate": 6.899679432440398e-06, "loss": 24.1276, "step": 49790 }, { "epoch": 0.61, "learning_rate": 6.895892790685723e-06, "loss": 23.6727, "step": 49800 }, { "epoch": 0.61, "learning_rate": 6.892106641413483e-06, "loss": 24.0617, "step": 49810 }, { "epoch": 0.61, "learning_rate": 6.888320985224368e-06, "loss": 24.0997, "step": 49820 }, { "epoch": 0.61, "learning_rate": 6.884535822718994e-06, "loss": 24.0263, "step": 49830 }, { "epoch": 0.61, "learning_rate": 6.880751154497895e-06, "loss": 23.4699, "step": 49840 }, { "epoch": 0.61, "learning_rate": 6.876966981161528e-06, "loss": 23.7821, "step": 49850 }, { "epoch": 0.61, "learning_rate": 6.873183303310271e-06, "loss": 24.1016, "step": 49860 }, { "epoch": 0.61, "learning_rate": 6.869400121544426e-06, "loss": 23.6, "step": 49870 }, { "epoch": 0.61, "learning_rate": 6.865617436464211e-06, "loss": 24.1309, "step": 49880 }, { "epoch": 0.61, "learning_rate": 6.8618352486697685e-06, "loss": 24.2985, "step": 49890 }, { "epoch": 0.61, "learning_rate": 6.858053558761164e-06, "loss": 23.6766, "step": 49900 }, { "epoch": 0.61, "learning_rate": 6.854272367338383e-06, "loss": 23.9013, "step": 49910 }, { "epoch": 0.61, "learning_rate": 6.850491675001326e-06, "loss": 23.6394, "step": 49920 }, { "epoch": 0.61, "learning_rate": 6.846711482349821e-06, "loss": 23.9434, "step": 49930 }, { "epoch": 0.61, "learning_rate": 6.84293178998362e-06, "loss": 24.6524, "step": 49940 }, { "epoch": 0.61, "learning_rate": 6.839152598502384e-06, "loss": 23.875, "step": 49950 }, { "epoch": 0.61, "learning_rate": 6.835373908505701e-06, "loss": 23.7332, "step": 49960 }, { "epoch": 0.61, "learning_rate": 6.831595720593085e-06, "loss": 24.151, "step": 49970 }, { "epoch": 0.61, "learning_rate": 6.827818035363959e-06, "loss": 23.6277, "step": 49980 }, { "epoch": 0.61, "learning_rate": 6.8240408534176746e-06, "loss": 23.8528, "step": 49990 }, { "epoch": 0.61, "learning_rate": 6.820264175353505e-06, "loss": 23.8535, "step": 50000 }, { "epoch": 0.62, "learning_rate": 6.816488001770632e-06, "loss": 23.6544, "step": 50010 }, { "epoch": 0.62, "learning_rate": 6.812712333268168e-06, "loss": 24.1414, "step": 50020 }, { "epoch": 0.62, "learning_rate": 6.8089371704451405e-06, "loss": 22.8352, "step": 50030 }, { "epoch": 0.62, "learning_rate": 6.8051625139005006e-06, "loss": 24.2216, "step": 50040 }, { "epoch": 0.62, "learning_rate": 6.8013883642331165e-06, "loss": 23.4625, "step": 50050 }, { "epoch": 0.62, "learning_rate": 6.797614722041775e-06, "loss": 24.3429, "step": 50060 }, { "epoch": 0.62, "learning_rate": 6.793841587925184e-06, "loss": 23.6082, "step": 50070 }, { "epoch": 0.62, "learning_rate": 6.790068962481973e-06, "loss": 24.1225, "step": 50080 }, { "epoch": 0.62, "learning_rate": 6.786296846310685e-06, "loss": 23.7849, "step": 50090 }, { "epoch": 0.62, "learning_rate": 6.782525240009785e-06, "loss": 23.5676, "step": 50100 }, { "epoch": 0.62, "learning_rate": 6.778754144177664e-06, "loss": 23.8028, "step": 50110 }, { "epoch": 0.62, "learning_rate": 6.7749835594126165e-06, "loss": 23.7214, "step": 50120 }, { "epoch": 0.62, "learning_rate": 6.771213486312871e-06, "loss": 23.4653, "step": 50130 }, { "epoch": 0.62, "learning_rate": 6.767443925476567e-06, "loss": 23.5677, "step": 50140 }, { "epoch": 0.62, "learning_rate": 6.763674877501765e-06, "loss": 24.2749, "step": 50150 }, { "epoch": 0.62, "learning_rate": 6.759906342986442e-06, "loss": 23.5652, "step": 50160 }, { "epoch": 0.62, "learning_rate": 6.756138322528498e-06, "loss": 23.4047, "step": 50170 }, { "epoch": 0.62, "learning_rate": 6.752370816725746e-06, "loss": 24.0845, "step": 50180 }, { "epoch": 0.62, "learning_rate": 6.748603826175922e-06, "loss": 23.2037, "step": 50190 }, { "epoch": 0.62, "learning_rate": 6.744837351476679e-06, "loss": 23.9098, "step": 50200 }, { "epoch": 0.62, "learning_rate": 6.741071393225585e-06, "loss": 23.6596, "step": 50210 }, { "epoch": 0.62, "learning_rate": 6.7373059520201275e-06, "loss": 23.8027, "step": 50220 }, { "epoch": 0.62, "learning_rate": 6.733541028457717e-06, "loss": 23.8126, "step": 50230 }, { "epoch": 0.62, "learning_rate": 6.729776623135677e-06, "loss": 23.8002, "step": 50240 }, { "epoch": 0.62, "learning_rate": 6.726012736651247e-06, "loss": 23.6194, "step": 50250 }, { "epoch": 0.62, "learning_rate": 6.722249369601587e-06, "loss": 24.0211, "step": 50260 }, { "epoch": 0.62, "learning_rate": 6.718486522583774e-06, "loss": 24.4067, "step": 50270 }, { "epoch": 0.62, "learning_rate": 6.714724196194809e-06, "loss": 23.8948, "step": 50280 }, { "epoch": 0.62, "learning_rate": 6.710962391031595e-06, "loss": 23.8492, "step": 50290 }, { "epoch": 0.62, "learning_rate": 6.707201107690966e-06, "loss": 23.4587, "step": 50300 }, { "epoch": 0.62, "learning_rate": 6.703440346769671e-06, "loss": 23.7618, "step": 50310 }, { "epoch": 0.62, "learning_rate": 6.6996801088643715e-06, "loss": 24.0507, "step": 50320 }, { "epoch": 0.62, "learning_rate": 6.695920394571648e-06, "loss": 23.0471, "step": 50330 }, { "epoch": 0.62, "learning_rate": 6.692161204488001e-06, "loss": 23.6352, "step": 50340 }, { "epoch": 0.62, "learning_rate": 6.68840253920984e-06, "loss": 24.0202, "step": 50350 }, { "epoch": 0.62, "learning_rate": 6.684644399333501e-06, "loss": 23.7437, "step": 50360 }, { "epoch": 0.62, "learning_rate": 6.680886785455232e-06, "loss": 24.3457, "step": 50370 }, { "epoch": 0.62, "learning_rate": 6.677129698171194e-06, "loss": 23.9035, "step": 50380 }, { "epoch": 0.62, "learning_rate": 6.673373138077471e-06, "loss": 23.7569, "step": 50390 }, { "epoch": 0.62, "learning_rate": 6.669617105770061e-06, "loss": 23.6239, "step": 50400 }, { "epoch": 0.62, "learning_rate": 6.6658616018448755e-06, "loss": 23.6935, "step": 50410 }, { "epoch": 0.62, "learning_rate": 6.662106626897744e-06, "loss": 23.8882, "step": 50420 }, { "epoch": 0.62, "learning_rate": 6.6583521815244164e-06, "loss": 23.8754, "step": 50430 }, { "epoch": 0.62, "learning_rate": 6.654598266320549e-06, "loss": 23.6702, "step": 50440 }, { "epoch": 0.62, "learning_rate": 6.650844881881724e-06, "loss": 24.0529, "step": 50450 }, { "epoch": 0.62, "learning_rate": 6.6470920288034325e-06, "loss": 23.7728, "step": 50460 }, { "epoch": 0.62, "learning_rate": 6.643339707681089e-06, "loss": 24.0597, "step": 50470 }, { "epoch": 0.62, "learning_rate": 6.639587919110011e-06, "loss": 23.8232, "step": 50480 }, { "epoch": 0.62, "learning_rate": 6.635836663685441e-06, "loss": 23.7194, "step": 50490 }, { "epoch": 0.62, "learning_rate": 6.632085942002535e-06, "loss": 23.8998, "step": 50500 }, { "epoch": 0.62, "learning_rate": 6.628335754656368e-06, "loss": 23.4347, "step": 50510 }, { "epoch": 0.62, "learning_rate": 6.62458610224192e-06, "loss": 23.8156, "step": 50520 }, { "epoch": 0.62, "learning_rate": 6.620836985354096e-06, "loss": 23.7829, "step": 50530 }, { "epoch": 0.62, "learning_rate": 6.617088404587715e-06, "loss": 23.627, "step": 50540 }, { "epoch": 0.62, "learning_rate": 6.613340360537503e-06, "loss": 23.6405, "step": 50550 }, { "epoch": 0.62, "learning_rate": 6.609592853798108e-06, "loss": 24.3561, "step": 50560 }, { "epoch": 0.62, "learning_rate": 6.605845884964096e-06, "loss": 23.8457, "step": 50570 }, { "epoch": 0.62, "learning_rate": 6.602099454629936e-06, "loss": 23.6115, "step": 50580 }, { "epoch": 0.62, "learning_rate": 6.598353563390021e-06, "loss": 24.271, "step": 50590 }, { "epoch": 0.62, "learning_rate": 6.594608211838655e-06, "loss": 23.671, "step": 50600 }, { "epoch": 0.62, "learning_rate": 6.59086340057006e-06, "loss": 23.6671, "step": 50610 }, { "epoch": 0.62, "learning_rate": 6.587119130178365e-06, "loss": 23.9911, "step": 50620 }, { "epoch": 0.62, "learning_rate": 6.58337540125762e-06, "loss": 23.6366, "step": 50630 }, { "epoch": 0.62, "learning_rate": 6.57963221440179e-06, "loss": 24.0989, "step": 50640 }, { "epoch": 0.62, "learning_rate": 6.575889570204743e-06, "loss": 24.2246, "step": 50650 }, { "epoch": 0.62, "learning_rate": 6.572147469260274e-06, "loss": 24.2583, "step": 50660 }, { "epoch": 0.62, "learning_rate": 6.568405912162086e-06, "loss": 24.1557, "step": 50670 }, { "epoch": 0.62, "learning_rate": 6.564664899503794e-06, "loss": 24.2857, "step": 50680 }, { "epoch": 0.62, "learning_rate": 6.56092443187893e-06, "loss": 23.7719, "step": 50690 }, { "epoch": 0.62, "learning_rate": 6.557184509880943e-06, "loss": 23.3269, "step": 50700 }, { "epoch": 0.62, "learning_rate": 6.553445134103183e-06, "loss": 23.5168, "step": 50710 }, { "epoch": 0.62, "learning_rate": 6.549706305138923e-06, "loss": 24.0244, "step": 50720 }, { "epoch": 0.62, "learning_rate": 6.545968023581347e-06, "loss": 23.9785, "step": 50730 }, { "epoch": 0.62, "learning_rate": 6.542230290023557e-06, "loss": 24.2661, "step": 50740 }, { "epoch": 0.62, "learning_rate": 6.538493105058556e-06, "loss": 23.8159, "step": 50750 }, { "epoch": 0.62, "learning_rate": 6.534756469279273e-06, "loss": 23.9525, "step": 50760 }, { "epoch": 0.62, "learning_rate": 6.531020383278543e-06, "loss": 23.9743, "step": 50770 }, { "epoch": 0.62, "learning_rate": 6.527284847649118e-06, "loss": 23.1931, "step": 50780 }, { "epoch": 0.62, "learning_rate": 6.5235498629836536e-06, "loss": 23.9937, "step": 50790 }, { "epoch": 0.62, "learning_rate": 6.519815429874727e-06, "loss": 23.8448, "step": 50800 }, { "epoch": 0.62, "learning_rate": 6.5160815489148275e-06, "loss": 23.4019, "step": 50810 }, { "epoch": 0.62, "learning_rate": 6.51234822069635e-06, "loss": 24.3456, "step": 50820 }, { "epoch": 0.63, "learning_rate": 6.50861544581161e-06, "loss": 24.0457, "step": 50830 }, { "epoch": 0.63, "learning_rate": 6.50488322485283e-06, "loss": 23.9542, "step": 50840 }, { "epoch": 0.63, "learning_rate": 6.5011515584121446e-06, "loss": 23.4174, "step": 50850 }, { "epoch": 0.63, "learning_rate": 6.4974204470816015e-06, "loss": 23.5866, "step": 50860 }, { "epoch": 0.63, "learning_rate": 6.493689891453164e-06, "loss": 23.9573, "step": 50870 }, { "epoch": 0.63, "learning_rate": 6.4899598921187e-06, "loss": 23.2906, "step": 50880 }, { "epoch": 0.63, "learning_rate": 6.4862304496699925e-06, "loss": 23.821, "step": 50890 }, { "epoch": 0.63, "learning_rate": 6.482501564698742e-06, "loss": 23.6421, "step": 50900 }, { "epoch": 0.63, "learning_rate": 6.47877323779655e-06, "loss": 24.2268, "step": 50910 }, { "epoch": 0.63, "learning_rate": 6.475045469554935e-06, "loss": 23.8321, "step": 50920 }, { "epoch": 0.63, "learning_rate": 6.471318260565331e-06, "loss": 23.6718, "step": 50930 }, { "epoch": 0.63, "learning_rate": 6.46759161141907e-06, "loss": 24.2997, "step": 50940 }, { "epoch": 0.63, "learning_rate": 6.463865522707413e-06, "loss": 24.2879, "step": 50950 }, { "epoch": 0.63, "learning_rate": 6.460139995021517e-06, "loss": 23.7771, "step": 50960 }, { "epoch": 0.63, "learning_rate": 6.4564150289524585e-06, "loss": 23.9479, "step": 50970 }, { "epoch": 0.63, "learning_rate": 6.452690625091222e-06, "loss": 23.8609, "step": 50980 }, { "epoch": 0.63, "learning_rate": 6.448966784028703e-06, "loss": 23.5584, "step": 50990 }, { "epoch": 0.63, "learning_rate": 6.445243506355707e-06, "loss": 23.5911, "step": 51000 }, { "epoch": 0.63, "learning_rate": 6.441520792662954e-06, "loss": 23.889, "step": 51010 }, { "epoch": 0.63, "learning_rate": 6.437798643541067e-06, "loss": 24.2608, "step": 51020 }, { "epoch": 0.63, "learning_rate": 6.434077059580586e-06, "loss": 23.9073, "step": 51030 }, { "epoch": 0.63, "learning_rate": 6.430356041371961e-06, "loss": 23.5713, "step": 51040 }, { "epoch": 0.63, "learning_rate": 6.42663558950555e-06, "loss": 24.5752, "step": 51050 }, { "epoch": 0.63, "learning_rate": 6.422915704571617e-06, "loss": 24.1975, "step": 51060 }, { "epoch": 0.63, "learning_rate": 6.419196387160349e-06, "loss": 23.4838, "step": 51070 }, { "epoch": 0.63, "learning_rate": 6.415477637861828e-06, "loss": 24.109, "step": 51080 }, { "epoch": 0.63, "learning_rate": 6.411759457266054e-06, "loss": 23.8935, "step": 51090 }, { "epoch": 0.63, "learning_rate": 6.408041845962938e-06, "loss": 23.5892, "step": 51100 }, { "epoch": 0.63, "learning_rate": 6.404324804542294e-06, "loss": 24.1063, "step": 51110 }, { "epoch": 0.63, "learning_rate": 6.40060833359385e-06, "loss": 24.0188, "step": 51120 }, { "epoch": 0.63, "learning_rate": 6.396892433707249e-06, "loss": 23.9574, "step": 51130 }, { "epoch": 0.63, "learning_rate": 6.393177105472031e-06, "loss": 23.7615, "step": 51140 }, { "epoch": 0.63, "learning_rate": 6.389462349477652e-06, "loss": 23.3095, "step": 51150 }, { "epoch": 0.63, "learning_rate": 6.385748166313484e-06, "loss": 23.4928, "step": 51160 }, { "epoch": 0.63, "learning_rate": 6.38203455656879e-06, "loss": 23.3621, "step": 51170 }, { "epoch": 0.63, "learning_rate": 6.378321520832764e-06, "loss": 23.673, "step": 51180 }, { "epoch": 0.63, "learning_rate": 6.37460905969449e-06, "loss": 24.137, "step": 51190 }, { "epoch": 0.63, "learning_rate": 6.370897173742972e-06, "loss": 24.0839, "step": 51200 }, { "epoch": 0.63, "learning_rate": 6.367185863567121e-06, "loss": 24.4664, "step": 51210 }, { "epoch": 0.63, "learning_rate": 6.363475129755751e-06, "loss": 24.016, "step": 51220 }, { "epoch": 0.63, "learning_rate": 6.359764972897593e-06, "loss": 24.3497, "step": 51230 }, { "epoch": 0.63, "learning_rate": 6.356055393581282e-06, "loss": 23.7279, "step": 51240 }, { "epoch": 0.63, "learning_rate": 6.3523463923953584e-06, "loss": 23.893, "step": 51250 }, { "epoch": 0.63, "learning_rate": 6.3486379699282775e-06, "loss": 23.5113, "step": 51260 }, { "epoch": 0.63, "learning_rate": 6.344930126768401e-06, "loss": 23.6145, "step": 51270 }, { "epoch": 0.63, "learning_rate": 6.3412228635039905e-06, "loss": 24.0544, "step": 51280 }, { "epoch": 0.63, "learning_rate": 6.337516180723228e-06, "loss": 23.6462, "step": 51290 }, { "epoch": 0.63, "learning_rate": 6.333810079014198e-06, "loss": 24.2308, "step": 51300 }, { "epoch": 0.63, "learning_rate": 6.330104558964887e-06, "loss": 24.3491, "step": 51310 }, { "epoch": 0.63, "learning_rate": 6.326399621163197e-06, "loss": 23.1714, "step": 51320 }, { "epoch": 0.63, "learning_rate": 6.322695266196938e-06, "loss": 24.0022, "step": 51330 }, { "epoch": 0.63, "learning_rate": 6.318991494653824e-06, "loss": 23.575, "step": 51340 }, { "epoch": 0.63, "learning_rate": 6.315288307121474e-06, "loss": 24.0544, "step": 51350 }, { "epoch": 0.63, "learning_rate": 6.311585704187419e-06, "loss": 23.8217, "step": 51360 }, { "epoch": 0.63, "learning_rate": 6.3078836864391e-06, "loss": 23.8129, "step": 51370 }, { "epoch": 0.63, "learning_rate": 6.304182254463856e-06, "loss": 23.6092, "step": 51380 }, { "epoch": 0.63, "learning_rate": 6.300481408848944e-06, "loss": 23.6998, "step": 51390 }, { "epoch": 0.63, "learning_rate": 6.296781150181512e-06, "loss": 23.7742, "step": 51400 }, { "epoch": 0.63, "learning_rate": 6.293081479048633e-06, "loss": 23.7747, "step": 51410 }, { "epoch": 0.63, "learning_rate": 6.289382396037275e-06, "loss": 23.9325, "step": 51420 }, { "epoch": 0.63, "learning_rate": 6.285683901734316e-06, "loss": 24.0107, "step": 51430 }, { "epoch": 0.63, "learning_rate": 6.281985996726547e-06, "loss": 23.8904, "step": 51440 }, { "epoch": 0.63, "learning_rate": 6.278288681600652e-06, "loss": 24.4519, "step": 51450 }, { "epoch": 0.63, "learning_rate": 6.274591956943231e-06, "loss": 23.4732, "step": 51460 }, { "epoch": 0.63, "learning_rate": 6.27089582334079e-06, "loss": 24.2353, "step": 51470 }, { "epoch": 0.63, "learning_rate": 6.267200281379741e-06, "loss": 24.0516, "step": 51480 }, { "epoch": 0.63, "learning_rate": 6.263505331646396e-06, "loss": 23.8114, "step": 51490 }, { "epoch": 0.63, "learning_rate": 6.259810974726979e-06, "loss": 23.7558, "step": 51500 }, { "epoch": 0.63, "learning_rate": 6.256117211207622e-06, "loss": 24.1913, "step": 51510 }, { "epoch": 0.63, "learning_rate": 6.252424041674355e-06, "loss": 24.6382, "step": 51520 }, { "epoch": 0.63, "learning_rate": 6.248731466713121e-06, "loss": 24.2895, "step": 51530 }, { "epoch": 0.63, "learning_rate": 6.245039486909765e-06, "loss": 23.6203, "step": 51540 }, { "epoch": 0.63, "learning_rate": 6.241348102850037e-06, "loss": 23.6345, "step": 51550 }, { "epoch": 0.63, "learning_rate": 6.237657315119594e-06, "loss": 23.846, "step": 51560 }, { "epoch": 0.63, "learning_rate": 6.233967124304003e-06, "loss": 24.3636, "step": 51570 }, { "epoch": 0.63, "learning_rate": 6.230277530988726e-06, "loss": 23.6257, "step": 51580 }, { "epoch": 0.63, "learning_rate": 6.2265885357591374e-06, "loss": 24.2089, "step": 51590 }, { "epoch": 0.63, "learning_rate": 6.222900139200517e-06, "loss": 23.5704, "step": 51600 }, { "epoch": 0.63, "learning_rate": 6.219212341898044e-06, "loss": 23.7574, "step": 51610 }, { "epoch": 0.63, "learning_rate": 6.215525144436814e-06, "loss": 23.7782, "step": 51620 }, { "epoch": 0.63, "learning_rate": 6.211838547401809e-06, "loss": 23.8458, "step": 51630 }, { "epoch": 0.64, "learning_rate": 6.208152551377933e-06, "loss": 24.1396, "step": 51640 }, { "epoch": 0.64, "learning_rate": 6.204467156949987e-06, "loss": 23.6077, "step": 51650 }, { "epoch": 0.64, "learning_rate": 6.200782364702676e-06, "loss": 23.5549, "step": 51660 }, { "epoch": 0.64, "learning_rate": 6.197098175220611e-06, "loss": 24.0854, "step": 51670 }, { "epoch": 0.64, "learning_rate": 6.193414589088313e-06, "loss": 23.941, "step": 51680 }, { "epoch": 0.64, "learning_rate": 6.189731606890193e-06, "loss": 23.1245, "step": 51690 }, { "epoch": 0.64, "learning_rate": 6.186049229210581e-06, "loss": 23.7796, "step": 51700 }, { "epoch": 0.64, "learning_rate": 6.182367456633704e-06, "loss": 23.6556, "step": 51710 }, { "epoch": 0.64, "learning_rate": 6.178686289743694e-06, "loss": 23.8681, "step": 51720 }, { "epoch": 0.64, "learning_rate": 6.175005729124583e-06, "loss": 23.8048, "step": 51730 }, { "epoch": 0.64, "learning_rate": 6.171325775360318e-06, "loss": 23.9847, "step": 51740 }, { "epoch": 0.64, "learning_rate": 6.1676464290347375e-06, "loss": 23.5638, "step": 51750 }, { "epoch": 0.64, "learning_rate": 6.163967690731587e-06, "loss": 24.2441, "step": 51760 }, { "epoch": 0.64, "learning_rate": 6.160289561034522e-06, "loss": 24.2057, "step": 51770 }, { "epoch": 0.64, "learning_rate": 6.156612040527093e-06, "loss": 23.6129, "step": 51780 }, { "epoch": 0.64, "learning_rate": 6.1529351297927575e-06, "loss": 23.4202, "step": 51790 }, { "epoch": 0.64, "learning_rate": 6.1492588294148795e-06, "loss": 23.1437, "step": 51800 }, { "epoch": 0.64, "learning_rate": 6.145583139976719e-06, "loss": 23.4375, "step": 51810 }, { "epoch": 0.64, "learning_rate": 6.141908062061443e-06, "loss": 23.7896, "step": 51820 }, { "epoch": 0.64, "learning_rate": 6.1382335962521256e-06, "loss": 23.7788, "step": 51830 }, { "epoch": 0.64, "learning_rate": 6.134559743131734e-06, "loss": 23.8258, "step": 51840 }, { "epoch": 0.64, "learning_rate": 6.130886503283147e-06, "loss": 23.656, "step": 51850 }, { "epoch": 0.64, "learning_rate": 6.127213877289145e-06, "loss": 23.6239, "step": 51860 }, { "epoch": 0.64, "learning_rate": 6.123541865732402e-06, "loss": 23.9172, "step": 51870 }, { "epoch": 0.64, "learning_rate": 6.119870469195508e-06, "loss": 23.7762, "step": 51880 }, { "epoch": 0.64, "learning_rate": 6.116199688260944e-06, "loss": 23.971, "step": 51890 }, { "epoch": 0.64, "learning_rate": 6.1125295235111e-06, "loss": 23.9487, "step": 51900 }, { "epoch": 0.64, "learning_rate": 6.108859975528268e-06, "loss": 23.8549, "step": 51910 }, { "epoch": 0.64, "learning_rate": 6.105191044894638e-06, "loss": 23.8521, "step": 51920 }, { "epoch": 0.64, "learning_rate": 6.101522732192304e-06, "loss": 23.5917, "step": 51930 }, { "epoch": 0.64, "learning_rate": 6.097855038003266e-06, "loss": 23.6902, "step": 51940 }, { "epoch": 0.64, "learning_rate": 6.0941879629094205e-06, "loss": 23.6878, "step": 51950 }, { "epoch": 0.64, "learning_rate": 6.090521507492566e-06, "loss": 23.2916, "step": 51960 }, { "epoch": 0.64, "learning_rate": 6.086855672334409e-06, "loss": 23.6517, "step": 51970 }, { "epoch": 0.64, "learning_rate": 6.083190458016547e-06, "loss": 23.6198, "step": 51980 }, { "epoch": 0.64, "learning_rate": 6.079525865120488e-06, "loss": 23.7581, "step": 51990 }, { "epoch": 0.64, "learning_rate": 6.075861894227641e-06, "loss": 23.7677, "step": 52000 }, { "epoch": 0.64, "learning_rate": 6.072198545919308e-06, "loss": 23.7755, "step": 52010 }, { "epoch": 0.64, "learning_rate": 6.068535820776701e-06, "loss": 24.1663, "step": 52020 }, { "epoch": 0.64, "learning_rate": 6.0648737193809296e-06, "loss": 23.8016, "step": 52030 }, { "epoch": 0.64, "learning_rate": 6.061212242313006e-06, "loss": 24.39, "step": 52040 }, { "epoch": 0.64, "learning_rate": 6.057551390153841e-06, "loss": 23.7621, "step": 52050 }, { "epoch": 0.64, "learning_rate": 6.053891163484247e-06, "loss": 23.399, "step": 52060 }, { "epoch": 0.64, "learning_rate": 6.05023156288494e-06, "loss": 24.0238, "step": 52070 }, { "epoch": 0.64, "learning_rate": 6.04657258893653e-06, "loss": 23.8445, "step": 52080 }, { "epoch": 0.64, "learning_rate": 6.04291424221954e-06, "loss": 23.873, "step": 52090 }, { "epoch": 0.64, "learning_rate": 6.0392565233143745e-06, "loss": 23.7842, "step": 52100 }, { "epoch": 0.64, "learning_rate": 6.0355994328013575e-06, "loss": 23.3519, "step": 52110 }, { "epoch": 0.64, "learning_rate": 6.0319429712607006e-06, "loss": 23.9506, "step": 52120 }, { "epoch": 0.64, "learning_rate": 6.02828713927252e-06, "loss": 23.6674, "step": 52130 }, { "epoch": 0.64, "learning_rate": 6.024631937416836e-06, "loss": 23.8197, "step": 52140 }, { "epoch": 0.64, "learning_rate": 6.020977366273562e-06, "loss": 24.1589, "step": 52150 }, { "epoch": 0.64, "learning_rate": 6.017323426422512e-06, "loss": 23.6626, "step": 52160 }, { "epoch": 0.64, "learning_rate": 6.013670118443407e-06, "loss": 24.2274, "step": 52170 }, { "epoch": 0.64, "learning_rate": 6.010017442915864e-06, "loss": 23.7226, "step": 52180 }, { "epoch": 0.64, "learning_rate": 6.006365400419394e-06, "loss": 24.4687, "step": 52190 }, { "epoch": 0.64, "learning_rate": 6.002713991533412e-06, "loss": 23.8139, "step": 52200 }, { "epoch": 0.64, "learning_rate": 5.999063216837237e-06, "loss": 23.7457, "step": 52210 }, { "epoch": 0.64, "learning_rate": 5.995413076910079e-06, "loss": 23.8603, "step": 52220 }, { "epoch": 0.64, "learning_rate": 5.991763572331053e-06, "loss": 24.0737, "step": 52230 }, { "epoch": 0.64, "learning_rate": 5.988114703679173e-06, "loss": 24.0346, "step": 52240 }, { "epoch": 0.64, "learning_rate": 5.984466471533348e-06, "loss": 24.0015, "step": 52250 }, { "epoch": 0.64, "learning_rate": 5.9808188764723884e-06, "loss": 23.8991, "step": 52260 }, { "epoch": 0.64, "learning_rate": 5.977171919075007e-06, "loss": 24.0302, "step": 52270 }, { "epoch": 0.64, "learning_rate": 5.973525599919809e-06, "loss": 24.1491, "step": 52280 }, { "epoch": 0.64, "learning_rate": 5.969879919585302e-06, "loss": 24.3179, "step": 52290 }, { "epoch": 0.64, "learning_rate": 5.966234878649895e-06, "loss": 23.3582, "step": 52300 }, { "epoch": 0.64, "learning_rate": 5.9625904776918865e-06, "loss": 23.2978, "step": 52310 }, { "epoch": 0.64, "learning_rate": 5.958946717289488e-06, "loss": 24.1266, "step": 52320 }, { "epoch": 0.64, "learning_rate": 5.95530359802079e-06, "loss": 24.1723, "step": 52330 }, { "epoch": 0.64, "learning_rate": 5.9516611204637985e-06, "loss": 23.2623, "step": 52340 }, { "epoch": 0.64, "learning_rate": 5.94801928519641e-06, "loss": 23.5085, "step": 52350 }, { "epoch": 0.64, "learning_rate": 5.944378092796419e-06, "loss": 23.7253, "step": 52360 }, { "epoch": 0.64, "learning_rate": 5.940737543841521e-06, "loss": 23.8253, "step": 52370 }, { "epoch": 0.64, "learning_rate": 5.937097638909308e-06, "loss": 23.991, "step": 52380 }, { "epoch": 0.64, "learning_rate": 5.933458378577267e-06, "loss": 23.8659, "step": 52390 }, { "epoch": 0.64, "learning_rate": 5.9298197634227865e-06, "loss": 24.119, "step": 52400 }, { "epoch": 0.64, "learning_rate": 5.9261817940231536e-06, "loss": 24.7702, "step": 52410 }, { "epoch": 0.64, "learning_rate": 5.922544470955546e-06, "loss": 24.0147, "step": 52420 }, { "epoch": 0.64, "learning_rate": 5.9189077947970465e-06, "loss": 23.5896, "step": 52430 }, { "epoch": 0.64, "learning_rate": 5.915271766124633e-06, "loss": 23.5655, "step": 52440 }, { "epoch": 0.65, "learning_rate": 5.911636385515178e-06, "loss": 23.8203, "step": 52450 }, { "epoch": 0.65, "learning_rate": 5.908001653545453e-06, "loss": 23.7949, "step": 52460 }, { "epoch": 0.65, "learning_rate": 5.904367570792131e-06, "loss": 23.8707, "step": 52470 }, { "epoch": 0.65, "learning_rate": 5.900734137831772e-06, "loss": 23.0151, "step": 52480 }, { "epoch": 0.65, "learning_rate": 5.897101355240841e-06, "loss": 23.637, "step": 52490 }, { "epoch": 0.65, "learning_rate": 5.8934692235957e-06, "loss": 24.0385, "step": 52500 }, { "epoch": 0.65, "learning_rate": 5.8898377434726005e-06, "loss": 23.6228, "step": 52510 }, { "epoch": 0.65, "learning_rate": 5.886206915447699e-06, "loss": 23.8659, "step": 52520 }, { "epoch": 0.65, "learning_rate": 5.882576740097044e-06, "loss": 24.3581, "step": 52530 }, { "epoch": 0.65, "learning_rate": 5.87894721799658e-06, "loss": 23.9865, "step": 52540 }, { "epoch": 0.65, "learning_rate": 5.875318349722155e-06, "loss": 23.4167, "step": 52550 }, { "epoch": 0.65, "learning_rate": 5.871690135849496e-06, "loss": 24.1473, "step": 52560 }, { "epoch": 0.65, "learning_rate": 5.868062576954246e-06, "loss": 24.0372, "step": 52570 }, { "epoch": 0.65, "learning_rate": 5.864435673611936e-06, "loss": 23.8825, "step": 52580 }, { "epoch": 0.65, "learning_rate": 5.8608094263979886e-06, "loss": 23.0496, "step": 52590 }, { "epoch": 0.65, "learning_rate": 5.857183835887727e-06, "loss": 24.0053, "step": 52600 }, { "epoch": 0.65, "learning_rate": 5.853558902656375e-06, "loss": 23.3762, "step": 52610 }, { "epoch": 0.65, "learning_rate": 5.849934627279039e-06, "loss": 24.322, "step": 52620 }, { "epoch": 0.65, "learning_rate": 5.846311010330734e-06, "loss": 23.7451, "step": 52630 }, { "epoch": 0.65, "learning_rate": 5.842688052386365e-06, "loss": 23.8751, "step": 52640 }, { "epoch": 0.65, "learning_rate": 5.839065754020731e-06, "loss": 23.9703, "step": 52650 }, { "epoch": 0.65, "learning_rate": 5.835444115808527e-06, "loss": 23.942, "step": 52660 }, { "epoch": 0.65, "learning_rate": 5.83182313832435e-06, "loss": 23.3446, "step": 52670 }, { "epoch": 0.65, "learning_rate": 5.828202822142677e-06, "loss": 23.6073, "step": 52680 }, { "epoch": 0.65, "learning_rate": 5.824583167837901e-06, "loss": 23.4663, "step": 52690 }, { "epoch": 0.65, "learning_rate": 5.82096417598429e-06, "loss": 24.3217, "step": 52700 }, { "epoch": 0.65, "learning_rate": 5.817345847156018e-06, "loss": 23.9454, "step": 52710 }, { "epoch": 0.65, "learning_rate": 5.813728181927152e-06, "loss": 23.9475, "step": 52720 }, { "epoch": 0.65, "learning_rate": 5.8101111808716545e-06, "loss": 24.1061, "step": 52730 }, { "epoch": 0.65, "learning_rate": 5.806494844563375e-06, "loss": 24.4244, "step": 52740 }, { "epoch": 0.65, "learning_rate": 5.802879173576073e-06, "loss": 23.8537, "step": 52750 }, { "epoch": 0.65, "learning_rate": 5.7992641684833875e-06, "loss": 23.5111, "step": 52760 }, { "epoch": 0.65, "learning_rate": 5.7956498298588536e-06, "loss": 23.7003, "step": 52770 }, { "epoch": 0.65, "learning_rate": 5.792036158275919e-06, "loss": 23.3682, "step": 52780 }, { "epoch": 0.65, "learning_rate": 5.788423154307892e-06, "loss": 23.4764, "step": 52790 }, { "epoch": 0.65, "learning_rate": 5.7848108185280074e-06, "loss": 23.679, "step": 52800 }, { "epoch": 0.65, "learning_rate": 5.781199151509377e-06, "loss": 23.3516, "step": 52810 }, { "epoch": 0.65, "learning_rate": 5.777588153825004e-06, "loss": 23.7331, "step": 52820 }, { "epoch": 0.65, "learning_rate": 5.773977826047804e-06, "loss": 23.519, "step": 52830 }, { "epoch": 0.65, "learning_rate": 5.7703681687505665e-06, "loss": 23.7534, "step": 52840 }, { "epoch": 0.65, "learning_rate": 5.7667591825059785e-06, "loss": 23.8025, "step": 52850 }, { "epoch": 0.65, "learning_rate": 5.763150867886633e-06, "loss": 23.5979, "step": 52860 }, { "epoch": 0.65, "learning_rate": 5.759543225465003e-06, "loss": 24.9479, "step": 52870 }, { "epoch": 0.65, "learning_rate": 5.755936255813455e-06, "loss": 24.1389, "step": 52880 }, { "epoch": 0.65, "learning_rate": 5.7523299595042615e-06, "loss": 23.5525, "step": 52890 }, { "epoch": 0.65, "learning_rate": 5.748724337109575e-06, "loss": 24.0404, "step": 52900 }, { "epoch": 0.65, "learning_rate": 5.745119389201443e-06, "loss": 23.5184, "step": 52910 }, { "epoch": 0.65, "learning_rate": 5.741515116351817e-06, "loss": 23.5648, "step": 52920 }, { "epoch": 0.65, "learning_rate": 5.737911519132528e-06, "loss": 23.4492, "step": 52930 }, { "epoch": 0.65, "learning_rate": 5.734308598115303e-06, "loss": 24.2929, "step": 52940 }, { "epoch": 0.65, "learning_rate": 5.7307063538717685e-06, "loss": 23.7736, "step": 52950 }, { "epoch": 0.65, "learning_rate": 5.727104786973439e-06, "loss": 23.4478, "step": 52960 }, { "epoch": 0.65, "learning_rate": 5.723503897991716e-06, "loss": 23.7873, "step": 52970 }, { "epoch": 0.65, "learning_rate": 5.719903687497904e-06, "loss": 23.5752, "step": 52980 }, { "epoch": 0.65, "learning_rate": 5.716304156063195e-06, "loss": 24.0792, "step": 52990 }, { "epoch": 0.65, "learning_rate": 5.712705304258665e-06, "loss": 22.9737, "step": 53000 }, { "epoch": 0.65, "learning_rate": 5.709107132655307e-06, "loss": 24.7486, "step": 53010 }, { "epoch": 0.65, "learning_rate": 5.705509641823969e-06, "loss": 23.7175, "step": 53020 }, { "epoch": 0.65, "learning_rate": 5.701912832335425e-06, "loss": 23.1521, "step": 53030 }, { "epoch": 0.65, "learning_rate": 5.698316704760323e-06, "loss": 24.1789, "step": 53040 }, { "epoch": 0.65, "learning_rate": 5.694721259669203e-06, "loss": 24.3165, "step": 53050 }, { "epoch": 0.65, "learning_rate": 5.691126497632509e-06, "loss": 23.9046, "step": 53060 }, { "epoch": 0.65, "learning_rate": 5.687532419220564e-06, "loss": 23.6301, "step": 53070 }, { "epoch": 0.65, "learning_rate": 5.683939025003582e-06, "loss": 24.0026, "step": 53080 }, { "epoch": 0.65, "learning_rate": 5.680346315551684e-06, "loss": 23.5939, "step": 53090 }, { "epoch": 0.65, "learning_rate": 5.676754291434862e-06, "loss": 23.9955, "step": 53100 }, { "epoch": 0.65, "learning_rate": 5.673162953223013e-06, "loss": 23.8486, "step": 53110 }, { "epoch": 0.65, "learning_rate": 5.669572301485924e-06, "loss": 23.8186, "step": 53120 }, { "epoch": 0.65, "learning_rate": 5.665982336793266e-06, "loss": 23.7077, "step": 53130 }, { "epoch": 0.65, "learning_rate": 5.662393059714602e-06, "loss": 23.7746, "step": 53140 }, { "epoch": 0.65, "learning_rate": 5.658804470819399e-06, "loss": 23.1723, "step": 53150 }, { "epoch": 0.65, "learning_rate": 5.655216570676998e-06, "loss": 25.126, "step": 53160 }, { "epoch": 0.65, "learning_rate": 5.6516293598566355e-06, "loss": 23.8094, "step": 53170 }, { "epoch": 0.65, "learning_rate": 5.6480428389274474e-06, "loss": 24.0808, "step": 53180 }, { "epoch": 0.65, "learning_rate": 5.644457008458451e-06, "loss": 23.8273, "step": 53190 }, { "epoch": 0.65, "learning_rate": 5.6408718690185525e-06, "loss": 23.8851, "step": 53200 }, { "epoch": 0.65, "learning_rate": 5.637287421176561e-06, "loss": 23.8461, "step": 53210 }, { "epoch": 0.65, "learning_rate": 5.63370366550116e-06, "loss": 23.8895, "step": 53220 }, { "epoch": 0.65, "learning_rate": 5.63012060256093e-06, "loss": 23.9727, "step": 53230 }, { "epoch": 0.65, "learning_rate": 5.626538232924354e-06, "loss": 23.577, "step": 53240 }, { "epoch": 0.65, "learning_rate": 5.622956557159776e-06, "loss": 24.2186, "step": 53250 }, { "epoch": 0.66, "learning_rate": 5.61937557583546e-06, "loss": 24.4259, "step": 53260 }, { "epoch": 0.66, "learning_rate": 5.615795289519541e-06, "loss": 23.9723, "step": 53270 }, { "epoch": 0.66, "learning_rate": 5.612215698780048e-06, "loss": 23.7505, "step": 53280 }, { "epoch": 0.66, "learning_rate": 5.608636804184909e-06, "loss": 24.0852, "step": 53290 }, { "epoch": 0.66, "learning_rate": 5.60505860630193e-06, "loss": 23.7499, "step": 53300 }, { "epoch": 0.66, "learning_rate": 5.601481105698805e-06, "loss": 23.8421, "step": 53310 }, { "epoch": 0.66, "learning_rate": 5.597904302943131e-06, "loss": 23.3864, "step": 53320 }, { "epoch": 0.66, "learning_rate": 5.5943281986023835e-06, "loss": 23.708, "step": 53330 }, { "epoch": 0.66, "learning_rate": 5.5907527932439235e-06, "loss": 23.2938, "step": 53340 }, { "epoch": 0.66, "learning_rate": 5.587178087435017e-06, "loss": 23.5182, "step": 53350 }, { "epoch": 0.66, "learning_rate": 5.583604081742805e-06, "loss": 23.4511, "step": 53360 }, { "epoch": 0.66, "learning_rate": 5.580030776734317e-06, "loss": 23.8609, "step": 53370 }, { "epoch": 0.66, "learning_rate": 5.576458172976484e-06, "loss": 23.6712, "step": 53380 }, { "epoch": 0.66, "learning_rate": 5.572886271036115e-06, "loss": 23.8174, "step": 53390 }, { "epoch": 0.66, "learning_rate": 5.569315071479906e-06, "loss": 24.1137, "step": 53400 }, { "epoch": 0.66, "learning_rate": 5.565744574874454e-06, "loss": 23.811, "step": 53410 }, { "epoch": 0.66, "learning_rate": 5.562174781786232e-06, "loss": 24.2706, "step": 53420 }, { "epoch": 0.66, "learning_rate": 5.5586056927816025e-06, "loss": 23.7632, "step": 53430 }, { "epoch": 0.66, "learning_rate": 5.555037308426826e-06, "loss": 23.893, "step": 53440 }, { "epoch": 0.66, "learning_rate": 5.551469629288043e-06, "loss": 23.9496, "step": 53450 }, { "epoch": 0.66, "learning_rate": 5.547902655931279e-06, "loss": 23.0875, "step": 53460 }, { "epoch": 0.66, "learning_rate": 5.544336388922461e-06, "loss": 23.7038, "step": 53470 }, { "epoch": 0.66, "learning_rate": 5.540770828827391e-06, "loss": 24.4544, "step": 53480 }, { "epoch": 0.66, "learning_rate": 5.537205976211764e-06, "loss": 24.3088, "step": 53490 }, { "epoch": 0.66, "learning_rate": 5.53364183164116e-06, "loss": 23.3829, "step": 53500 }, { "epoch": 0.66, "learning_rate": 5.530078395681047e-06, "loss": 23.7523, "step": 53510 }, { "epoch": 0.66, "learning_rate": 5.526515668896789e-06, "loss": 24.0721, "step": 53520 }, { "epoch": 0.66, "learning_rate": 5.522953651853628e-06, "loss": 23.4802, "step": 53530 }, { "epoch": 0.66, "learning_rate": 5.519392345116691e-06, "loss": 24.5102, "step": 53540 }, { "epoch": 0.66, "learning_rate": 5.515831749251006e-06, "loss": 23.7109, "step": 53550 }, { "epoch": 0.66, "learning_rate": 5.512271864821475e-06, "loss": 23.6272, "step": 53560 }, { "epoch": 0.66, "learning_rate": 5.508712692392888e-06, "loss": 23.6004, "step": 53570 }, { "epoch": 0.66, "learning_rate": 5.5051542325299345e-06, "loss": 23.8009, "step": 53580 }, { "epoch": 0.66, "learning_rate": 5.501596485797178e-06, "loss": 23.4726, "step": 53590 }, { "epoch": 0.66, "learning_rate": 5.4980394527590695e-06, "loss": 23.209, "step": 53600 }, { "epoch": 0.66, "learning_rate": 5.494483133979959e-06, "loss": 23.8132, "step": 53610 }, { "epoch": 0.66, "learning_rate": 5.490927530024068e-06, "loss": 23.4109, "step": 53620 }, { "epoch": 0.66, "learning_rate": 5.487372641455511e-06, "loss": 23.7038, "step": 53630 }, { "epoch": 0.66, "learning_rate": 5.483818468838294e-06, "loss": 23.6878, "step": 53640 }, { "epoch": 0.66, "learning_rate": 5.4802650127363e-06, "loss": 23.9594, "step": 53650 }, { "epoch": 0.66, "learning_rate": 5.476712273713301e-06, "loss": 23.9088, "step": 53660 }, { "epoch": 0.66, "learning_rate": 5.473160252332967e-06, "loss": 23.5018, "step": 53670 }, { "epoch": 0.66, "learning_rate": 5.469608949158836e-06, "loss": 23.3155, "step": 53680 }, { "epoch": 0.66, "learning_rate": 5.466058364754339e-06, "loss": 24.0501, "step": 53690 }, { "epoch": 0.66, "learning_rate": 5.462508499682801e-06, "loss": 23.9532, "step": 53700 }, { "epoch": 0.66, "learning_rate": 5.458959354507425e-06, "loss": 23.0195, "step": 53710 }, { "epoch": 0.66, "learning_rate": 5.455410929791299e-06, "loss": 24.3099, "step": 53720 }, { "epoch": 0.66, "learning_rate": 5.451863226097396e-06, "loss": 24.1105, "step": 53730 }, { "epoch": 0.66, "learning_rate": 5.448316243988577e-06, "loss": 23.8059, "step": 53740 }, { "epoch": 0.66, "learning_rate": 5.444769984027596e-06, "loss": 23.6147, "step": 53750 }, { "epoch": 0.66, "learning_rate": 5.44122444677708e-06, "loss": 23.6076, "step": 53760 }, { "epoch": 0.66, "learning_rate": 5.437679632799544e-06, "loss": 23.1406, "step": 53770 }, { "epoch": 0.66, "learning_rate": 5.4341355426574e-06, "loss": 24.7967, "step": 53780 }, { "epoch": 0.66, "learning_rate": 5.430592176912928e-06, "loss": 23.5934, "step": 53790 }, { "epoch": 0.66, "learning_rate": 5.427049536128299e-06, "loss": 23.5816, "step": 53800 }, { "epoch": 0.66, "learning_rate": 5.42350762086558e-06, "loss": 24.0599, "step": 53810 }, { "epoch": 0.66, "learning_rate": 5.41996643168671e-06, "loss": 23.4442, "step": 53820 }, { "epoch": 0.66, "learning_rate": 5.41642596915351e-06, "loss": 23.8456, "step": 53830 }, { "epoch": 0.66, "learning_rate": 5.4128862338277035e-06, "loss": 23.2857, "step": 53840 }, { "epoch": 0.66, "learning_rate": 5.409347226270879e-06, "loss": 24.0756, "step": 53850 }, { "epoch": 0.66, "learning_rate": 5.405808947044524e-06, "loss": 23.9096, "step": 53860 }, { "epoch": 0.66, "learning_rate": 5.402271396710004e-06, "loss": 23.4283, "step": 53870 }, { "epoch": 0.66, "learning_rate": 5.398734575828561e-06, "loss": 24.1836, "step": 53880 }, { "epoch": 0.66, "learning_rate": 5.395198484961342e-06, "loss": 23.9341, "step": 53890 }, { "epoch": 0.66, "learning_rate": 5.391663124669359e-06, "loss": 24.0417, "step": 53900 }, { "epoch": 0.66, "learning_rate": 5.388128495513513e-06, "loss": 23.5856, "step": 53910 }, { "epoch": 0.66, "learning_rate": 5.3845945980545965e-06, "loss": 23.735, "step": 53920 }, { "epoch": 0.66, "learning_rate": 5.381061432853278e-06, "loss": 24.118, "step": 53930 }, { "epoch": 0.66, "learning_rate": 5.377529000470112e-06, "loss": 24.2859, "step": 53940 }, { "epoch": 0.66, "learning_rate": 5.373997301465537e-06, "loss": 23.8471, "step": 53950 }, { "epoch": 0.66, "learning_rate": 5.3704663363998745e-06, "loss": 23.826, "step": 53960 }, { "epoch": 0.66, "learning_rate": 5.3669361058333265e-06, "loss": 23.934, "step": 53970 }, { "epoch": 0.66, "learning_rate": 5.363406610325989e-06, "loss": 23.7758, "step": 53980 }, { "epoch": 0.66, "learning_rate": 5.359877850437832e-06, "loss": 23.7265, "step": 53990 }, { "epoch": 0.66, "learning_rate": 5.3563498267287075e-06, "loss": 23.7209, "step": 54000 }, { "epoch": 0.66, "learning_rate": 5.352822539758361e-06, "loss": 24.0152, "step": 54010 }, { "epoch": 0.66, "learning_rate": 5.349295990086407e-06, "loss": 24.1736, "step": 54020 }, { "epoch": 0.66, "learning_rate": 5.345770178272359e-06, "loss": 23.8425, "step": 54030 }, { "epoch": 0.66, "learning_rate": 5.3422451048756e-06, "loss": 23.2209, "step": 54040 }, { "epoch": 0.66, "learning_rate": 5.338720770455399e-06, "loss": 23.7388, "step": 54050 }, { "epoch": 0.66, "learning_rate": 5.335197175570916e-06, "loss": 23.7685, "step": 54060 }, { "epoch": 0.66, "learning_rate": 5.331674320781183e-06, "loss": 23.4228, "step": 54070 }, { "epoch": 0.67, "learning_rate": 5.328152206645115e-06, "loss": 24.3912, "step": 54080 }, { "epoch": 0.67, "learning_rate": 5.324630833721522e-06, "loss": 23.0938, "step": 54090 }, { "epoch": 0.67, "learning_rate": 5.321110202569084e-06, "loss": 23.596, "step": 54100 }, { "epoch": 0.67, "learning_rate": 5.3175903137463635e-06, "loss": 23.7513, "step": 54110 }, { "epoch": 0.67, "learning_rate": 5.314071167811815e-06, "loss": 24.3162, "step": 54120 }, { "epoch": 0.67, "learning_rate": 5.310552765323767e-06, "loss": 23.7207, "step": 54130 }, { "epoch": 0.67, "learning_rate": 5.3070351068404255e-06, "loss": 23.416, "step": 54140 }, { "epoch": 0.67, "learning_rate": 5.303518192919894e-06, "loss": 23.8878, "step": 54150 }, { "epoch": 0.67, "learning_rate": 5.3000020241201455e-06, "loss": 24.0019, "step": 54160 }, { "epoch": 0.67, "learning_rate": 5.296486600999039e-06, "loss": 23.6581, "step": 54170 }, { "epoch": 0.67, "learning_rate": 5.2929719241143115e-06, "loss": 23.7045, "step": 54180 }, { "epoch": 0.67, "learning_rate": 5.289457994023584e-06, "loss": 25.0986, "step": 54190 }, { "epoch": 0.67, "learning_rate": 5.285944811284364e-06, "loss": 23.7729, "step": 54200 }, { "epoch": 0.67, "learning_rate": 5.282432376454033e-06, "loss": 23.6119, "step": 54210 }, { "epoch": 0.67, "learning_rate": 5.278920690089852e-06, "loss": 23.7074, "step": 54220 }, { "epoch": 0.67, "learning_rate": 5.275409752748979e-06, "loss": 24.1692, "step": 54230 }, { "epoch": 0.67, "learning_rate": 5.271899564988437e-06, "loss": 23.7518, "step": 54240 }, { "epoch": 0.67, "learning_rate": 5.268390127365128e-06, "loss": 23.5326, "step": 54250 }, { "epoch": 0.67, "learning_rate": 5.264881440435852e-06, "loss": 24.1312, "step": 54260 }, { "epoch": 0.67, "learning_rate": 5.261373504757277e-06, "loss": 23.5258, "step": 54270 }, { "epoch": 0.67, "learning_rate": 5.2578663208859515e-06, "loss": 23.6216, "step": 54280 }, { "epoch": 0.67, "learning_rate": 5.254359889378313e-06, "loss": 23.9999, "step": 54290 }, { "epoch": 0.67, "learning_rate": 5.2508542107906745e-06, "loss": 23.9989, "step": 54300 }, { "epoch": 0.67, "learning_rate": 5.247349285679223e-06, "loss": 24.0344, "step": 54310 }, { "epoch": 0.67, "learning_rate": 5.243845114600041e-06, "loss": 23.9039, "step": 54320 }, { "epoch": 0.67, "learning_rate": 5.240341698109082e-06, "loss": 24.1461, "step": 54330 }, { "epoch": 0.67, "learning_rate": 5.236839036762172e-06, "loss": 23.8756, "step": 54340 }, { "epoch": 0.67, "learning_rate": 5.2333371311150386e-06, "loss": 24.2179, "step": 54350 }, { "epoch": 0.67, "learning_rate": 5.22983598172327e-06, "loss": 23.19, "step": 54360 }, { "epoch": 0.67, "learning_rate": 5.226335589142339e-06, "loss": 23.7816, "step": 54370 }, { "epoch": 0.67, "learning_rate": 5.2228359539276075e-06, "loss": 23.6596, "step": 54380 }, { "epoch": 0.67, "learning_rate": 5.219337076634309e-06, "loss": 22.8563, "step": 54390 }, { "epoch": 0.67, "learning_rate": 5.215838957817555e-06, "loss": 24.0482, "step": 54400 }, { "epoch": 0.67, "learning_rate": 5.212341598032342e-06, "loss": 23.6012, "step": 54410 }, { "epoch": 0.67, "learning_rate": 5.2088449978335385e-06, "loss": 24.2431, "step": 54420 }, { "epoch": 0.67, "learning_rate": 5.205349157775907e-06, "loss": 23.8022, "step": 54430 }, { "epoch": 0.67, "learning_rate": 5.201854078414077e-06, "loss": 23.6863, "step": 54440 }, { "epoch": 0.67, "learning_rate": 5.198359760302556e-06, "loss": 24.3517, "step": 54450 }, { "epoch": 0.67, "learning_rate": 5.194866203995741e-06, "loss": 23.6276, "step": 54460 }, { "epoch": 0.67, "learning_rate": 5.191373410047902e-06, "loss": 23.856, "step": 54470 }, { "epoch": 0.67, "learning_rate": 5.1878813790131835e-06, "loss": 23.7452, "step": 54480 }, { "epoch": 0.67, "learning_rate": 5.184390111445621e-06, "loss": 24.0634, "step": 54490 }, { "epoch": 0.67, "learning_rate": 5.180899607899119e-06, "loss": 23.849, "step": 54500 }, { "epoch": 0.67, "learning_rate": 5.17740986892746e-06, "loss": 23.5917, "step": 54510 }, { "epoch": 0.67, "learning_rate": 5.173920895084314e-06, "loss": 23.6479, "step": 54520 }, { "epoch": 0.67, "learning_rate": 5.170432686923226e-06, "loss": 24.0476, "step": 54530 }, { "epoch": 0.67, "learning_rate": 5.166945244997607e-06, "loss": 23.6501, "step": 54540 }, { "epoch": 0.67, "learning_rate": 5.163458569860773e-06, "loss": 23.7482, "step": 54550 }, { "epoch": 0.67, "learning_rate": 5.159972662065893e-06, "loss": 23.5115, "step": 54560 }, { "epoch": 0.67, "learning_rate": 5.156487522166022e-06, "loss": 23.9587, "step": 54570 }, { "epoch": 0.67, "learning_rate": 5.153003150714102e-06, "loss": 23.7865, "step": 54580 }, { "epoch": 0.67, "learning_rate": 5.149519548262946e-06, "loss": 23.599, "step": 54590 }, { "epoch": 0.67, "learning_rate": 5.146036715365237e-06, "loss": 23.6312, "step": 54600 }, { "epoch": 0.67, "learning_rate": 5.142554652573555e-06, "loss": 23.5681, "step": 54610 }, { "epoch": 0.67, "learning_rate": 5.139073360440341e-06, "loss": 24.2024, "step": 54620 }, { "epoch": 0.67, "learning_rate": 5.13559283951792e-06, "loss": 24.0019, "step": 54630 }, { "epoch": 0.67, "learning_rate": 5.132113090358497e-06, "loss": 23.8295, "step": 54640 }, { "epoch": 0.67, "learning_rate": 5.128634113514146e-06, "loss": 24.2897, "step": 54650 }, { "epoch": 0.67, "learning_rate": 5.1251559095368305e-06, "loss": 24.1993, "step": 54660 }, { "epoch": 0.67, "learning_rate": 5.1216784789783846e-06, "loss": 24.0442, "step": 54670 }, { "epoch": 0.67, "learning_rate": 5.1182018223905165e-06, "loss": 24.0891, "step": 54680 }, { "epoch": 0.67, "learning_rate": 5.114725940324819e-06, "loss": 23.8749, "step": 54690 }, { "epoch": 0.67, "learning_rate": 5.111250833332759e-06, "loss": 24.0557, "step": 54700 }, { "epoch": 0.67, "learning_rate": 5.107776501965676e-06, "loss": 23.6473, "step": 54710 }, { "epoch": 0.67, "learning_rate": 5.104302946774795e-06, "loss": 24.3249, "step": 54720 }, { "epoch": 0.67, "learning_rate": 5.1008301683112114e-06, "loss": 24.0294, "step": 54730 }, { "epoch": 0.67, "learning_rate": 5.0973581671258954e-06, "loss": 23.7909, "step": 54740 }, { "epoch": 0.67, "learning_rate": 5.093886943769705e-06, "loss": 23.0741, "step": 54750 }, { "epoch": 0.67, "learning_rate": 5.090416498793363e-06, "loss": 23.6347, "step": 54760 }, { "epoch": 0.67, "learning_rate": 5.086946832747471e-06, "loss": 24.1685, "step": 54770 }, { "epoch": 0.67, "learning_rate": 5.083477946182516e-06, "loss": 23.6201, "step": 54780 }, { "epoch": 0.67, "learning_rate": 5.0800098396488495e-06, "loss": 23.9329, "step": 54790 }, { "epoch": 0.67, "learning_rate": 5.076542513696703e-06, "loss": 24.0014, "step": 54800 }, { "epoch": 0.67, "learning_rate": 5.073075968876191e-06, "loss": 24.2994, "step": 54810 }, { "epoch": 0.67, "learning_rate": 5.069610205737295e-06, "loss": 24.5624, "step": 54820 }, { "epoch": 0.67, "learning_rate": 5.066145224829874e-06, "loss": 23.9671, "step": 54830 }, { "epoch": 0.67, "learning_rate": 5.0626810267036695e-06, "loss": 23.9477, "step": 54840 }, { "epoch": 0.67, "learning_rate": 5.059217611908292e-06, "loss": 23.9184, "step": 54850 }, { "epoch": 0.67, "learning_rate": 5.0557549809932305e-06, "loss": 23.2489, "step": 54860 }, { "epoch": 0.67, "learning_rate": 5.0522931345078485e-06, "loss": 23.9941, "step": 54870 }, { "epoch": 0.67, "learning_rate": 5.048832073001382e-06, "loss": 24.2434, "step": 54880 }, { "epoch": 0.68, "learning_rate": 5.045371797022953e-06, "loss": 23.8547, "step": 54890 }, { "epoch": 0.68, "learning_rate": 5.041912307121548e-06, "loss": 24.0306, "step": 54900 }, { "epoch": 0.68, "learning_rate": 5.03845360384603e-06, "loss": 23.5057, "step": 54910 }, { "epoch": 0.68, "learning_rate": 5.034995687745145e-06, "loss": 23.8335, "step": 54920 }, { "epoch": 0.68, "learning_rate": 5.0315385593675105e-06, "loss": 24.5653, "step": 54930 }, { "epoch": 0.68, "learning_rate": 5.028082219261607e-06, "loss": 23.6527, "step": 54940 }, { "epoch": 0.68, "learning_rate": 5.024626667975814e-06, "loss": 23.7753, "step": 54950 }, { "epoch": 0.68, "learning_rate": 5.0211719060583655e-06, "loss": 23.732, "step": 54960 }, { "epoch": 0.68, "learning_rate": 5.017717934057373e-06, "loss": 23.9887, "step": 54970 }, { "epoch": 0.68, "learning_rate": 5.014264752520835e-06, "loss": 24.0962, "step": 54980 }, { "epoch": 0.68, "learning_rate": 5.010812361996613e-06, "loss": 24.1013, "step": 54990 }, { "epoch": 0.68, "learning_rate": 5.007360763032443e-06, "loss": 23.7017, "step": 55000 }, { "epoch": 0.68, "learning_rate": 5.0039099561759454e-06, "loss": 24.0338, "step": 55010 }, { "epoch": 0.68, "learning_rate": 5.000459941974604e-06, "loss": 23.2103, "step": 55020 }, { "epoch": 0.68, "learning_rate": 4.997010720975778e-06, "loss": 24.1289, "step": 55030 }, { "epoch": 0.68, "learning_rate": 4.993562293726711e-06, "loss": 24.1329, "step": 55040 }, { "epoch": 0.68, "learning_rate": 4.990114660774512e-06, "loss": 24.2826, "step": 55050 }, { "epoch": 0.68, "learning_rate": 4.986667822666159e-06, "loss": 24.2556, "step": 55060 }, { "epoch": 0.68, "learning_rate": 4.98322177994852e-06, "loss": 23.8384, "step": 55070 }, { "epoch": 0.68, "learning_rate": 4.979776533168321e-06, "loss": 24.5055, "step": 55080 }, { "epoch": 0.68, "learning_rate": 4.976332082872167e-06, "loss": 23.5933, "step": 55090 }, { "epoch": 0.68, "learning_rate": 4.972888429606547e-06, "loss": 23.6527, "step": 55100 }, { "epoch": 0.68, "learning_rate": 4.9694455739178e-06, "loss": 23.9531, "step": 55110 }, { "epoch": 0.68, "learning_rate": 4.966003516352163e-06, "loss": 23.5488, "step": 55120 }, { "epoch": 0.68, "learning_rate": 4.9625622574557346e-06, "loss": 23.7901, "step": 55130 }, { "epoch": 0.68, "learning_rate": 4.959121797774481e-06, "loss": 24.1585, "step": 55140 }, { "epoch": 0.68, "learning_rate": 4.955682137854259e-06, "loss": 23.8762, "step": 55150 }, { "epoch": 0.68, "learning_rate": 4.952243278240781e-06, "loss": 23.7205, "step": 55160 }, { "epoch": 0.68, "learning_rate": 4.9488052194796395e-06, "loss": 23.4705, "step": 55170 }, { "epoch": 0.68, "learning_rate": 4.945367962116307e-06, "loss": 23.3796, "step": 55180 }, { "epoch": 0.68, "learning_rate": 4.941931506696116e-06, "loss": 24.0187, "step": 55190 }, { "epoch": 0.68, "learning_rate": 4.938495853764276e-06, "loss": 24.1109, "step": 55200 }, { "epoch": 0.68, "learning_rate": 4.9350610038658765e-06, "loss": 23.4636, "step": 55210 }, { "epoch": 0.68, "learning_rate": 4.931626957545871e-06, "loss": 23.5993, "step": 55220 }, { "epoch": 0.68, "learning_rate": 4.928193715349086e-06, "loss": 24.259, "step": 55230 }, { "epoch": 0.68, "learning_rate": 4.924761277820229e-06, "loss": 24.2871, "step": 55240 }, { "epoch": 0.68, "learning_rate": 4.92132964550387e-06, "loss": 23.9721, "step": 55250 }, { "epoch": 0.68, "learning_rate": 4.917898818944451e-06, "loss": 23.8655, "step": 55260 }, { "epoch": 0.68, "learning_rate": 4.914468798686299e-06, "loss": 24.091, "step": 55270 }, { "epoch": 0.68, "learning_rate": 4.911039585273596e-06, "loss": 23.2707, "step": 55280 }, { "epoch": 0.68, "learning_rate": 4.90761117925041e-06, "loss": 23.6698, "step": 55290 }, { "epoch": 0.68, "learning_rate": 4.9041835811606746e-06, "loss": 23.3341, "step": 55300 }, { "epoch": 0.68, "learning_rate": 4.900756791548189e-06, "loss": 23.4688, "step": 55310 }, { "epoch": 0.68, "learning_rate": 4.89733081095664e-06, "loss": 24.2693, "step": 55320 }, { "epoch": 0.68, "learning_rate": 4.893905639929576e-06, "loss": 23.5253, "step": 55330 }, { "epoch": 0.68, "learning_rate": 4.890481279010407e-06, "loss": 23.8945, "step": 55340 }, { "epoch": 0.68, "learning_rate": 4.887057728742437e-06, "loss": 24.1992, "step": 55350 }, { "epoch": 0.68, "learning_rate": 4.8836349896688264e-06, "loss": 23.7422, "step": 55360 }, { "epoch": 0.68, "learning_rate": 4.880213062332604e-06, "loss": 23.8064, "step": 55370 }, { "epoch": 0.68, "learning_rate": 4.876791947276685e-06, "loss": 23.5213, "step": 55380 }, { "epoch": 0.68, "learning_rate": 4.873371645043845e-06, "loss": 23.7407, "step": 55390 }, { "epoch": 0.68, "learning_rate": 4.869952156176726e-06, "loss": 24.1985, "step": 55400 }, { "epoch": 0.68, "learning_rate": 4.866533481217854e-06, "loss": 23.8644, "step": 55410 }, { "epoch": 0.68, "learning_rate": 4.863115620709615e-06, "loss": 23.7422, "step": 55420 }, { "epoch": 0.68, "learning_rate": 4.859698575194276e-06, "loss": 23.8657, "step": 55430 }, { "epoch": 0.68, "learning_rate": 4.856282345213964e-06, "loss": 23.6697, "step": 55440 }, { "epoch": 0.68, "learning_rate": 4.852866931310678e-06, "loss": 23.5995, "step": 55450 }, { "epoch": 0.68, "learning_rate": 4.849452334026299e-06, "loss": 24.1033, "step": 55460 }, { "epoch": 0.68, "learning_rate": 4.846038553902566e-06, "loss": 23.3497, "step": 55470 }, { "epoch": 0.68, "learning_rate": 4.8426255914810895e-06, "loss": 24.173, "step": 55480 }, { "epoch": 0.68, "learning_rate": 4.83921344730336e-06, "loss": 23.8005, "step": 55490 }, { "epoch": 0.68, "learning_rate": 4.835802121910729e-06, "loss": 23.858, "step": 55500 }, { "epoch": 0.68, "learning_rate": 4.832391615844415e-06, "loss": 23.8114, "step": 55510 }, { "epoch": 0.68, "learning_rate": 4.828981929645521e-06, "loss": 23.7249, "step": 55520 }, { "epoch": 0.68, "learning_rate": 4.825573063855007e-06, "loss": 23.6812, "step": 55530 }, { "epoch": 0.68, "learning_rate": 4.822165019013704e-06, "loss": 23.5804, "step": 55540 }, { "epoch": 0.68, "learning_rate": 4.818757795662322e-06, "loss": 23.3071, "step": 55550 }, { "epoch": 0.68, "learning_rate": 4.815351394341436e-06, "loss": 23.984, "step": 55560 }, { "epoch": 0.68, "learning_rate": 4.811945815591476e-06, "loss": 23.4519, "step": 55570 }, { "epoch": 0.68, "learning_rate": 4.808541059952765e-06, "loss": 23.9247, "step": 55580 }, { "epoch": 0.68, "learning_rate": 4.80513712796548e-06, "loss": 24.0337, "step": 55590 }, { "epoch": 0.68, "learning_rate": 4.801734020169677e-06, "loss": 23.6189, "step": 55600 }, { "epoch": 0.68, "learning_rate": 4.798331737105275e-06, "loss": 23.5415, "step": 55610 }, { "epoch": 0.68, "learning_rate": 4.794930279312059e-06, "loss": 23.7767, "step": 55620 }, { "epoch": 0.68, "learning_rate": 4.791529647329694e-06, "loss": 23.9529, "step": 55630 }, { "epoch": 0.68, "learning_rate": 4.788129841697705e-06, "loss": 23.3557, "step": 55640 }, { "epoch": 0.68, "learning_rate": 4.784730862955484e-06, "loss": 24.3557, "step": 55650 }, { "epoch": 0.68, "learning_rate": 4.781332711642305e-06, "loss": 24.2582, "step": 55660 }, { "epoch": 0.68, "learning_rate": 4.777935388297298e-06, "loss": 23.8521, "step": 55670 }, { "epoch": 0.68, "learning_rate": 4.77453889345946e-06, "loss": 23.3476, "step": 55680 }, { "epoch": 0.68, "learning_rate": 4.771143227667671e-06, "loss": 24.2681, "step": 55690 }, { "epoch": 0.69, "learning_rate": 4.767748391460668e-06, "loss": 23.606, "step": 55700 }, { "epoch": 0.69, "learning_rate": 4.764354385377055e-06, "loss": 23.8504, "step": 55710 }, { "epoch": 0.69, "learning_rate": 4.760961209955315e-06, "loss": 23.7703, "step": 55720 }, { "epoch": 0.69, "learning_rate": 4.757568865733788e-06, "loss": 23.7709, "step": 55730 }, { "epoch": 0.69, "learning_rate": 4.754177353250685e-06, "loss": 23.8943, "step": 55740 }, { "epoch": 0.69, "learning_rate": 4.750786673044093e-06, "loss": 23.9111, "step": 55750 }, { "epoch": 0.69, "learning_rate": 4.747396825651957e-06, "loss": 23.5766, "step": 55760 }, { "epoch": 0.69, "learning_rate": 4.744007811612092e-06, "loss": 23.6741, "step": 55770 }, { "epoch": 0.69, "learning_rate": 4.740619631462186e-06, "loss": 23.9947, "step": 55780 }, { "epoch": 0.69, "learning_rate": 4.737232285739789e-06, "loss": 23.9593, "step": 55790 }, { "epoch": 0.69, "learning_rate": 4.733845774982322e-06, "loss": 24.0117, "step": 55800 }, { "epoch": 0.69, "learning_rate": 4.730460099727071e-06, "loss": 23.4542, "step": 55810 }, { "epoch": 0.69, "learning_rate": 4.7270752605111866e-06, "loss": 23.8768, "step": 55820 }, { "epoch": 0.69, "learning_rate": 4.723691257871699e-06, "loss": 24.8955, "step": 55830 }, { "epoch": 0.69, "learning_rate": 4.720308092345493e-06, "loss": 23.9189, "step": 55840 }, { "epoch": 0.69, "learning_rate": 4.716925764469322e-06, "loss": 24.1495, "step": 55850 }, { "epoch": 0.69, "learning_rate": 4.713544274779818e-06, "loss": 23.7719, "step": 55860 }, { "epoch": 0.69, "learning_rate": 4.710163623813465e-06, "loss": 23.9592, "step": 55870 }, { "epoch": 0.69, "learning_rate": 4.706783812106619e-06, "loss": 24.6681, "step": 55880 }, { "epoch": 0.69, "learning_rate": 4.703404840195512e-06, "loss": 23.4696, "step": 55890 }, { "epoch": 0.69, "learning_rate": 4.700026708616229e-06, "loss": 23.5022, "step": 55900 }, { "epoch": 0.69, "learning_rate": 4.696649417904726e-06, "loss": 23.6462, "step": 55910 }, { "epoch": 0.69, "learning_rate": 4.693272968596836e-06, "loss": 24.2088, "step": 55920 }, { "epoch": 0.69, "learning_rate": 4.689897361228243e-06, "loss": 23.9675, "step": 55930 }, { "epoch": 0.69, "learning_rate": 4.686522596334503e-06, "loss": 23.609, "step": 55940 }, { "epoch": 0.69, "learning_rate": 4.683148674451047e-06, "loss": 23.9169, "step": 55950 }, { "epoch": 0.69, "learning_rate": 4.67977559611316e-06, "loss": 23.5707, "step": 55960 }, { "epoch": 0.69, "learning_rate": 4.676403361855994e-06, "loss": 24.1814, "step": 55970 }, { "epoch": 0.69, "learning_rate": 4.673031972214581e-06, "loss": 24.0822, "step": 55980 }, { "epoch": 0.69, "learning_rate": 4.6696614277238035e-06, "loss": 23.7357, "step": 55990 }, { "epoch": 0.69, "learning_rate": 4.666291728918413e-06, "loss": 23.48, "step": 56000 }, { "epoch": 0.69, "learning_rate": 4.662922876333036e-06, "loss": 23.7725, "step": 56010 }, { "epoch": 0.69, "learning_rate": 4.6595548705021535e-06, "loss": 24.5675, "step": 56020 }, { "epoch": 0.69, "learning_rate": 4.656187711960118e-06, "loss": 23.9609, "step": 56030 }, { "epoch": 0.69, "learning_rate": 4.6528214012411475e-06, "loss": 23.687, "step": 56040 }, { "epoch": 0.69, "learning_rate": 4.649455938879318e-06, "loss": 24.1377, "step": 56050 }, { "epoch": 0.69, "learning_rate": 4.646091325408587e-06, "loss": 24.0173, "step": 56060 }, { "epoch": 0.69, "learning_rate": 4.642727561362763e-06, "loss": 23.8521, "step": 56070 }, { "epoch": 0.69, "learning_rate": 4.63936464727552e-06, "loss": 23.6963, "step": 56080 }, { "epoch": 0.69, "learning_rate": 4.636002583680409e-06, "loss": 23.3834, "step": 56090 }, { "epoch": 0.69, "learning_rate": 4.6326413711108365e-06, "loss": 23.4816, "step": 56100 }, { "epoch": 0.69, "learning_rate": 4.62928101010007e-06, "loss": 23.8366, "step": 56110 }, { "epoch": 0.69, "learning_rate": 4.6259215011812574e-06, "loss": 24.1071, "step": 56120 }, { "epoch": 0.69, "learning_rate": 4.622562844887396e-06, "loss": 23.0418, "step": 56130 }, { "epoch": 0.69, "learning_rate": 4.619205041751353e-06, "loss": 23.693, "step": 56140 }, { "epoch": 0.69, "learning_rate": 4.615848092305865e-06, "loss": 23.9669, "step": 56150 }, { "epoch": 0.69, "learning_rate": 4.612491997083527e-06, "loss": 24.0457, "step": 56160 }, { "epoch": 0.69, "learning_rate": 4.609136756616798e-06, "loss": 24.1755, "step": 56170 }, { "epoch": 0.69, "learning_rate": 4.60578237143801e-06, "loss": 23.6526, "step": 56180 }, { "epoch": 0.69, "learning_rate": 4.60242884207935e-06, "loss": 23.5581, "step": 56190 }, { "epoch": 0.69, "learning_rate": 4.599076169072868e-06, "loss": 23.5192, "step": 56200 }, { "epoch": 0.69, "learning_rate": 4.595724352950492e-06, "loss": 23.9644, "step": 56210 }, { "epoch": 0.69, "learning_rate": 4.592373394244e-06, "loss": 24.4119, "step": 56220 }, { "epoch": 0.69, "learning_rate": 4.589023293485034e-06, "loss": 23.5467, "step": 56230 }, { "epoch": 0.69, "learning_rate": 4.585674051205113e-06, "loss": 23.3715, "step": 56240 }, { "epoch": 0.69, "learning_rate": 4.582325667935609e-06, "loss": 23.7467, "step": 56250 }, { "epoch": 0.69, "learning_rate": 4.578978144207756e-06, "loss": 23.974, "step": 56260 }, { "epoch": 0.69, "learning_rate": 4.57563148055266e-06, "loss": 24.1655, "step": 56270 }, { "epoch": 0.69, "learning_rate": 4.572285677501282e-06, "loss": 23.3444, "step": 56280 }, { "epoch": 0.69, "learning_rate": 4.568940735584455e-06, "loss": 24.1003, "step": 56290 }, { "epoch": 0.69, "learning_rate": 4.56559665533287e-06, "loss": 23.382, "step": 56300 }, { "epoch": 0.69, "learning_rate": 4.562253437277078e-06, "loss": 23.9354, "step": 56310 }, { "epoch": 0.69, "learning_rate": 4.558911081947505e-06, "loss": 24.1581, "step": 56320 }, { "epoch": 0.69, "learning_rate": 4.555569589874429e-06, "loss": 23.5043, "step": 56330 }, { "epoch": 0.69, "learning_rate": 4.552228961587991e-06, "loss": 24.026, "step": 56340 }, { "epoch": 0.69, "learning_rate": 4.548889197618206e-06, "loss": 23.7047, "step": 56350 }, { "epoch": 0.69, "learning_rate": 4.545550298494942e-06, "loss": 23.7008, "step": 56360 }, { "epoch": 0.69, "learning_rate": 4.5422122647479275e-06, "loss": 24.2371, "step": 56370 }, { "epoch": 0.69, "learning_rate": 4.538875096906765e-06, "loss": 23.6782, "step": 56380 }, { "epoch": 0.69, "learning_rate": 4.535538795500913e-06, "loss": 23.8499, "step": 56390 }, { "epoch": 0.69, "learning_rate": 4.532203361059684e-06, "loss": 23.4721, "step": 56400 }, { "epoch": 0.69, "learning_rate": 4.528868794112273e-06, "loss": 23.602, "step": 56410 }, { "epoch": 0.69, "learning_rate": 4.525535095187722e-06, "loss": 23.2605, "step": 56420 }, { "epoch": 0.69, "learning_rate": 4.522202264814934e-06, "loss": 23.4592, "step": 56430 }, { "epoch": 0.69, "learning_rate": 4.5188703035226875e-06, "loss": 23.5163, "step": 56440 }, { "epoch": 0.69, "learning_rate": 4.515539211839612e-06, "loss": 24.1691, "step": 56450 }, { "epoch": 0.69, "learning_rate": 4.512208990294199e-06, "loss": 23.2768, "step": 56460 }, { "epoch": 0.69, "learning_rate": 4.50887963941481e-06, "loss": 24.0738, "step": 56470 }, { "epoch": 0.69, "learning_rate": 4.505551159729663e-06, "loss": 23.0071, "step": 56480 }, { "epoch": 0.69, "learning_rate": 4.502223551766835e-06, "loss": 23.7175, "step": 56490 }, { "epoch": 0.69, "learning_rate": 4.4988968160542725e-06, "loss": 24.067, "step": 56500 }, { "epoch": 0.69, "learning_rate": 4.49557095311977e-06, "loss": 23.7521, "step": 56510 }, { "epoch": 0.7, "learning_rate": 4.492245963491005e-06, "loss": 23.7665, "step": 56520 }, { "epoch": 0.7, "learning_rate": 4.488921847695495e-06, "loss": 23.9865, "step": 56530 }, { "epoch": 0.7, "learning_rate": 4.48559860626063e-06, "loss": 24.2627, "step": 56540 }, { "epoch": 0.7, "learning_rate": 4.482276239713661e-06, "loss": 23.3587, "step": 56550 }, { "epoch": 0.7, "learning_rate": 4.4789547485816994e-06, "loss": 24.5184, "step": 56560 }, { "epoch": 0.7, "learning_rate": 4.47563413339171e-06, "loss": 23.8597, "step": 56570 }, { "epoch": 0.7, "learning_rate": 4.472314394670535e-06, "loss": 23.7046, "step": 56580 }, { "epoch": 0.7, "learning_rate": 4.468995532944862e-06, "loss": 23.5085, "step": 56590 }, { "epoch": 0.7, "learning_rate": 4.465677548741243e-06, "loss": 23.617, "step": 56600 }, { "epoch": 0.7, "learning_rate": 4.4623604425861e-06, "loss": 23.5991, "step": 56610 }, { "epoch": 0.7, "learning_rate": 4.459044215005706e-06, "loss": 24.0058, "step": 56620 }, { "epoch": 0.7, "learning_rate": 4.455728866526192e-06, "loss": 23.9401, "step": 56630 }, { "epoch": 0.7, "learning_rate": 4.4524143976735635e-06, "loss": 24.5479, "step": 56640 }, { "epoch": 0.7, "learning_rate": 4.449100808973674e-06, "loss": 24.0403, "step": 56650 }, { "epoch": 0.7, "learning_rate": 4.445788100952237e-06, "loss": 23.4927, "step": 56660 }, { "epoch": 0.7, "learning_rate": 4.4424762741348395e-06, "loss": 24.3186, "step": 56670 }, { "epoch": 0.7, "learning_rate": 4.439165329046911e-06, "loss": 23.9638, "step": 56680 }, { "epoch": 0.7, "learning_rate": 4.435855266213757e-06, "loss": 24.4183, "step": 56690 }, { "epoch": 0.7, "learning_rate": 4.432546086160532e-06, "loss": 23.619, "step": 56700 }, { "epoch": 0.7, "learning_rate": 4.429237789412251e-06, "loss": 23.8939, "step": 56710 }, { "epoch": 0.7, "learning_rate": 4.4259303764938045e-06, "loss": 23.6077, "step": 56720 }, { "epoch": 0.7, "learning_rate": 4.422623847929918e-06, "loss": 24.0098, "step": 56730 }, { "epoch": 0.7, "learning_rate": 4.4193182042451895e-06, "loss": 23.7895, "step": 56740 }, { "epoch": 0.7, "learning_rate": 4.416013445964084e-06, "loss": 23.5008, "step": 56750 }, { "epoch": 0.7, "learning_rate": 4.412709573610912e-06, "loss": 23.5077, "step": 56760 }, { "epoch": 0.7, "learning_rate": 4.40940658770985e-06, "loss": 23.5279, "step": 56770 }, { "epoch": 0.7, "learning_rate": 4.406104488784938e-06, "loss": 24.0391, "step": 56780 }, { "epoch": 0.7, "learning_rate": 4.402803277360068e-06, "loss": 24.214, "step": 56790 }, { "epoch": 0.7, "learning_rate": 4.39950295395899e-06, "loss": 23.9605, "step": 56800 }, { "epoch": 0.7, "learning_rate": 4.396203519105326e-06, "loss": 24.4011, "step": 56810 }, { "epoch": 0.7, "learning_rate": 4.392904973322543e-06, "loss": 23.0494, "step": 56820 }, { "epoch": 0.7, "learning_rate": 4.389607317133971e-06, "loss": 23.5107, "step": 56830 }, { "epoch": 0.7, "learning_rate": 4.386310551062802e-06, "loss": 23.7015, "step": 56840 }, { "epoch": 0.7, "learning_rate": 4.383014675632082e-06, "loss": 23.9062, "step": 56850 }, { "epoch": 0.7, "learning_rate": 4.379719691364723e-06, "loss": 24.0834, "step": 56860 }, { "epoch": 0.7, "learning_rate": 4.37642559878349e-06, "loss": 23.9179, "step": 56870 }, { "epoch": 0.7, "learning_rate": 4.373132398411003e-06, "loss": 23.437, "step": 56880 }, { "epoch": 0.7, "learning_rate": 4.369840090769751e-06, "loss": 23.2715, "step": 56890 }, { "epoch": 0.7, "learning_rate": 4.366548676382073e-06, "loss": 23.7567, "step": 56900 }, { "epoch": 0.7, "learning_rate": 4.363258155770165e-06, "loss": 24.1458, "step": 56910 }, { "epoch": 0.7, "learning_rate": 4.359968529456092e-06, "loss": 24.1053, "step": 56920 }, { "epoch": 0.7, "learning_rate": 4.356679797961766e-06, "loss": 23.4264, "step": 56930 }, { "epoch": 0.7, "learning_rate": 4.353391961808957e-06, "loss": 23.8307, "step": 56940 }, { "epoch": 0.7, "learning_rate": 4.3501050215193095e-06, "loss": 23.8925, "step": 56950 }, { "epoch": 0.7, "learning_rate": 4.346818977614301e-06, "loss": 23.6496, "step": 56960 }, { "epoch": 0.7, "learning_rate": 4.34353383061528e-06, "loss": 23.6674, "step": 56970 }, { "epoch": 0.7, "learning_rate": 4.3402495810434584e-06, "loss": 23.3868, "step": 56980 }, { "epoch": 0.7, "learning_rate": 4.336966229419892e-06, "loss": 24.309, "step": 56990 }, { "epoch": 0.7, "learning_rate": 4.333683776265509e-06, "loss": 23.9744, "step": 57000 }, { "epoch": 0.7, "learning_rate": 4.330402222101083e-06, "loss": 24.3545, "step": 57010 }, { "epoch": 0.7, "learning_rate": 4.3271215674472466e-06, "loss": 23.947, "step": 57020 }, { "epoch": 0.7, "learning_rate": 4.323841812824499e-06, "loss": 23.4145, "step": 57030 }, { "epoch": 0.7, "learning_rate": 4.320562958753185e-06, "loss": 23.6306, "step": 57040 }, { "epoch": 0.7, "learning_rate": 4.317285005753511e-06, "loss": 23.8357, "step": 57050 }, { "epoch": 0.7, "learning_rate": 4.314007954345546e-06, "loss": 23.9008, "step": 57060 }, { "epoch": 0.7, "learning_rate": 4.310731805049208e-06, "loss": 23.7725, "step": 57070 }, { "epoch": 0.7, "learning_rate": 4.30745655838427e-06, "loss": 23.6888, "step": 57080 }, { "epoch": 0.7, "learning_rate": 4.304182214870376e-06, "loss": 23.5003, "step": 57090 }, { "epoch": 0.7, "learning_rate": 4.300908775027012e-06, "loss": 24.1489, "step": 57100 }, { "epoch": 0.7, "learning_rate": 4.297636239373523e-06, "loss": 23.794, "step": 57110 }, { "epoch": 0.7, "learning_rate": 4.294364608429119e-06, "loss": 23.4406, "step": 57120 }, { "epoch": 0.7, "learning_rate": 4.29109388271286e-06, "loss": 23.6555, "step": 57130 }, { "epoch": 0.7, "learning_rate": 4.2878240627436594e-06, "loss": 23.2363, "step": 57140 }, { "epoch": 0.7, "learning_rate": 4.284555149040296e-06, "loss": 23.5174, "step": 57150 }, { "epoch": 0.7, "learning_rate": 4.281287142121398e-06, "loss": 23.6217, "step": 57160 }, { "epoch": 0.7, "learning_rate": 4.278020042505445e-06, "loss": 23.9239, "step": 57170 }, { "epoch": 0.7, "learning_rate": 4.274753850710795e-06, "loss": 23.7873, "step": 57180 }, { "epoch": 0.7, "learning_rate": 4.271488567255625e-06, "loss": 23.2535, "step": 57190 }, { "epoch": 0.7, "learning_rate": 4.2682241926580046e-06, "loss": 23.5429, "step": 57200 }, { "epoch": 0.7, "learning_rate": 4.264960727435838e-06, "loss": 24.0447, "step": 57210 }, { "epoch": 0.7, "learning_rate": 4.261698172106887e-06, "loss": 24.2182, "step": 57220 }, { "epoch": 0.7, "learning_rate": 4.25843652718878e-06, "loss": 23.8413, "step": 57230 }, { "epoch": 0.7, "learning_rate": 4.25517579319899e-06, "loss": 23.8394, "step": 57240 }, { "epoch": 0.7, "learning_rate": 4.251915970654846e-06, "loss": 23.0971, "step": 57250 }, { "epoch": 0.7, "learning_rate": 4.248657060073542e-06, "loss": 23.9565, "step": 57260 }, { "epoch": 0.7, "learning_rate": 4.245399061972118e-06, "loss": 24.2466, "step": 57270 }, { "epoch": 0.7, "learning_rate": 4.242141976867466e-06, "loss": 23.0584, "step": 57280 }, { "epoch": 0.7, "learning_rate": 4.238885805276349e-06, "loss": 23.9546, "step": 57290 }, { "epoch": 0.7, "learning_rate": 4.2356305477153715e-06, "loss": 24.0029, "step": 57300 }, { "epoch": 0.7, "learning_rate": 4.232376204700992e-06, "loss": 23.7104, "step": 57310 }, { "epoch": 0.7, "learning_rate": 4.229122776749537e-06, "loss": 23.9878, "step": 57320 }, { "epoch": 0.71, "learning_rate": 4.225870264377173e-06, "loss": 23.6208, "step": 57330 }, { "epoch": 0.71, "learning_rate": 4.222618668099928e-06, "loss": 24.0227, "step": 57340 }, { "epoch": 0.71, "learning_rate": 4.219367988433688e-06, "loss": 23.5362, "step": 57350 }, { "epoch": 0.71, "learning_rate": 4.216118225894189e-06, "loss": 22.7468, "step": 57360 }, { "epoch": 0.71, "learning_rate": 4.212869380997018e-06, "loss": 24.0444, "step": 57370 }, { "epoch": 0.71, "learning_rate": 4.209621454257626e-06, "loss": 23.8541, "step": 57380 }, { "epoch": 0.71, "learning_rate": 4.206374446191313e-06, "loss": 24.1871, "step": 57390 }, { "epoch": 0.71, "learning_rate": 4.203128357313228e-06, "loss": 24.0578, "step": 57400 }, { "epoch": 0.71, "learning_rate": 4.19988318813839e-06, "loss": 23.8202, "step": 57410 }, { "epoch": 0.71, "learning_rate": 4.196638939181647e-06, "loss": 24.0007, "step": 57420 }, { "epoch": 0.71, "learning_rate": 4.193395610957727e-06, "loss": 23.846, "step": 57430 }, { "epoch": 0.71, "learning_rate": 4.190153203981198e-06, "loss": 23.5214, "step": 57440 }, { "epoch": 0.71, "learning_rate": 4.1869117187664785e-06, "loss": 23.8019, "step": 57450 }, { "epoch": 0.71, "learning_rate": 4.183671155827856e-06, "loss": 23.1808, "step": 57460 }, { "epoch": 0.71, "learning_rate": 4.180431515679458e-06, "loss": 23.6484, "step": 57470 }, { "epoch": 0.71, "learning_rate": 4.177192798835267e-06, "loss": 23.8411, "step": 57480 }, { "epoch": 0.71, "learning_rate": 4.173955005809127e-06, "loss": 23.794, "step": 57490 }, { "epoch": 0.71, "learning_rate": 4.170718137114729e-06, "loss": 23.6203, "step": 57500 }, { "epoch": 0.71, "learning_rate": 4.167482193265613e-06, "loss": 24.1874, "step": 57510 }, { "epoch": 0.71, "learning_rate": 4.164247174775188e-06, "loss": 23.8815, "step": 57520 }, { "epoch": 0.71, "learning_rate": 4.1610130821567e-06, "loss": 24.4609, "step": 57530 }, { "epoch": 0.71, "learning_rate": 4.157779915923253e-06, "loss": 23.6986, "step": 57540 }, { "epoch": 0.71, "learning_rate": 4.154547676587809e-06, "loss": 23.7295, "step": 57550 }, { "epoch": 0.71, "learning_rate": 4.151316364663179e-06, "loss": 23.3545, "step": 57560 }, { "epoch": 0.71, "learning_rate": 4.148085980662024e-06, "loss": 23.6225, "step": 57570 }, { "epoch": 0.71, "learning_rate": 4.144856525096864e-06, "loss": 23.455, "step": 57580 }, { "epoch": 0.71, "learning_rate": 4.141627998480068e-06, "loss": 24.1676, "step": 57590 }, { "epoch": 0.71, "learning_rate": 4.138400401323855e-06, "loss": 23.5842, "step": 57600 }, { "epoch": 0.71, "learning_rate": 4.1351737341403055e-06, "loss": 23.5997, "step": 57610 }, { "epoch": 0.71, "learning_rate": 4.131947997441344e-06, "loss": 23.6549, "step": 57620 }, { "epoch": 0.71, "learning_rate": 4.128723191738745e-06, "loss": 24.0501, "step": 57630 }, { "epoch": 0.71, "learning_rate": 4.125499317544152e-06, "loss": 23.8144, "step": 57640 }, { "epoch": 0.71, "learning_rate": 4.122276375369037e-06, "loss": 23.5812, "step": 57650 }, { "epoch": 0.71, "learning_rate": 4.119054365724743e-06, "loss": 24.043, "step": 57660 }, { "epoch": 0.71, "learning_rate": 4.115833289122458e-06, "loss": 24.0141, "step": 57670 }, { "epoch": 0.71, "learning_rate": 4.112613146073215e-06, "loss": 23.887, "step": 57680 }, { "epoch": 0.71, "learning_rate": 4.109393937087917e-06, "loss": 24.117, "step": 57690 }, { "epoch": 0.71, "learning_rate": 4.106175662677303e-06, "loss": 23.7906, "step": 57700 }, { "epoch": 0.71, "learning_rate": 4.102958323351965e-06, "loss": 23.2364, "step": 57710 }, { "epoch": 0.71, "learning_rate": 4.0997419196223575e-06, "loss": 23.4585, "step": 57720 }, { "epoch": 0.71, "learning_rate": 4.096526451998775e-06, "loss": 24.1341, "step": 57730 }, { "epoch": 0.71, "learning_rate": 4.093311920991365e-06, "loss": 23.8313, "step": 57740 }, { "epoch": 0.71, "learning_rate": 4.090098327110137e-06, "loss": 23.3155, "step": 57750 }, { "epoch": 0.71, "learning_rate": 4.086885670864938e-06, "loss": 23.2397, "step": 57760 }, { "epoch": 0.71, "learning_rate": 4.083673952765472e-06, "loss": 24.2419, "step": 57770 }, { "epoch": 0.71, "learning_rate": 4.080463173321299e-06, "loss": 23.8332, "step": 57780 }, { "epoch": 0.71, "learning_rate": 4.077253333041824e-06, "loss": 23.9402, "step": 57790 }, { "epoch": 0.71, "learning_rate": 4.0740444324363e-06, "loss": 23.7321, "step": 57800 }, { "epoch": 0.71, "learning_rate": 4.070836472013842e-06, "loss": 23.9168, "step": 57810 }, { "epoch": 0.71, "learning_rate": 4.067629452283406e-06, "loss": 23.4318, "step": 57820 }, { "epoch": 0.71, "learning_rate": 4.064423373753799e-06, "loss": 23.956, "step": 57830 }, { "epoch": 0.71, "learning_rate": 4.0612182369336885e-06, "loss": 23.7021, "step": 57840 }, { "epoch": 0.71, "learning_rate": 4.058014042331583e-06, "loss": 23.9106, "step": 57850 }, { "epoch": 0.71, "learning_rate": 4.0548107904558385e-06, "loss": 24.0025, "step": 57860 }, { "epoch": 0.71, "learning_rate": 4.0516084818146806e-06, "loss": 23.2375, "step": 57870 }, { "epoch": 0.71, "learning_rate": 4.048407116916157e-06, "loss": 24.3798, "step": 57880 }, { "epoch": 0.71, "learning_rate": 4.045206696268189e-06, "loss": 23.4142, "step": 57890 }, { "epoch": 0.71, "learning_rate": 4.042007220378539e-06, "loss": 24.0862, "step": 57900 }, { "epoch": 0.71, "learning_rate": 4.0388086897548155e-06, "loss": 24.5239, "step": 57910 }, { "epoch": 0.71, "learning_rate": 4.0356111049044885e-06, "loss": 23.6051, "step": 57920 }, { "epoch": 0.71, "learning_rate": 4.032414466334867e-06, "loss": 23.7614, "step": 57930 }, { "epoch": 0.71, "learning_rate": 4.029218774553113e-06, "loss": 24.3802, "step": 57940 }, { "epoch": 0.71, "learning_rate": 4.026024030066243e-06, "loss": 23.6699, "step": 57950 }, { "epoch": 0.71, "learning_rate": 4.022830233381119e-06, "loss": 23.9699, "step": 57960 }, { "epoch": 0.71, "learning_rate": 4.019637385004448e-06, "loss": 23.4408, "step": 57970 }, { "epoch": 0.71, "learning_rate": 4.016445485442798e-06, "loss": 23.6998, "step": 57980 }, { "epoch": 0.71, "learning_rate": 4.013254535202579e-06, "loss": 24.1284, "step": 57990 }, { "epoch": 0.71, "learning_rate": 4.010064534790046e-06, "loss": 23.5572, "step": 58000 }, { "epoch": 0.71, "learning_rate": 4.0068754847113155e-06, "loss": 23.7288, "step": 58010 }, { "epoch": 0.71, "learning_rate": 4.003687385472344e-06, "loss": 24.0895, "step": 58020 }, { "epoch": 0.71, "learning_rate": 4.000500237578937e-06, "loss": 23.8401, "step": 58030 }, { "epoch": 0.71, "learning_rate": 3.997314041536758e-06, "loss": 23.4069, "step": 58040 }, { "epoch": 0.71, "learning_rate": 3.99412879785131e-06, "loss": 23.7782, "step": 58050 }, { "epoch": 0.71, "learning_rate": 3.990944507027944e-06, "loss": 23.6181, "step": 58060 }, { "epoch": 0.71, "learning_rate": 3.987761169571871e-06, "loss": 23.7841, "step": 58070 }, { "epoch": 0.71, "learning_rate": 3.984578785988137e-06, "loss": 23.8428, "step": 58080 }, { "epoch": 0.71, "learning_rate": 3.98139735678165e-06, "loss": 23.657, "step": 58090 }, { "epoch": 0.71, "learning_rate": 3.978216882457161e-06, "loss": 23.4591, "step": 58100 }, { "epoch": 0.71, "learning_rate": 3.975037363519258e-06, "loss": 23.3268, "step": 58110 }, { "epoch": 0.71, "learning_rate": 3.971858800472396e-06, "loss": 24.4198, "step": 58120 }, { "epoch": 0.71, "learning_rate": 3.9686811938208694e-06, "loss": 24.2911, "step": 58130 }, { "epoch": 0.72, "learning_rate": 3.965504544068818e-06, "loss": 24.2114, "step": 58140 }, { "epoch": 0.72, "learning_rate": 3.962328851720239e-06, "loss": 23.8633, "step": 58150 }, { "epoch": 0.72, "learning_rate": 3.959154117278968e-06, "loss": 24.2303, "step": 58160 }, { "epoch": 0.72, "learning_rate": 3.955980341248693e-06, "loss": 23.7427, "step": 58170 }, { "epoch": 0.72, "learning_rate": 3.952807524132953e-06, "loss": 23.6111, "step": 58180 }, { "epoch": 0.72, "learning_rate": 3.949635666435131e-06, "loss": 23.632, "step": 58190 }, { "epoch": 0.72, "learning_rate": 3.946464768658452e-06, "loss": 24.4185, "step": 58200 }, { "epoch": 0.72, "learning_rate": 3.943294831306004e-06, "loss": 23.5403, "step": 58210 }, { "epoch": 0.72, "learning_rate": 3.94012585488071e-06, "loss": 23.3222, "step": 58220 }, { "epoch": 0.72, "learning_rate": 3.936957839885341e-06, "loss": 23.7177, "step": 58230 }, { "epoch": 0.72, "learning_rate": 3.933790786822525e-06, "loss": 24.4009, "step": 58240 }, { "epoch": 0.72, "learning_rate": 3.9306246961947235e-06, "loss": 24.2456, "step": 58250 }, { "epoch": 0.72, "learning_rate": 3.9274595685042615e-06, "loss": 24.0231, "step": 58260 }, { "epoch": 0.72, "learning_rate": 3.924295404253299e-06, "loss": 23.4592, "step": 58270 }, { "epoch": 0.72, "learning_rate": 3.921132203943842e-06, "loss": 23.4421, "step": 58280 }, { "epoch": 0.72, "learning_rate": 3.917969968077756e-06, "loss": 24.1974, "step": 58290 }, { "epoch": 0.72, "learning_rate": 3.914808697156743e-06, "loss": 24.1069, "step": 58300 }, { "epoch": 0.72, "learning_rate": 3.91164839168235e-06, "loss": 23.5652, "step": 58310 }, { "epoch": 0.72, "learning_rate": 3.908489052155984e-06, "loss": 23.8244, "step": 58320 }, { "epoch": 0.72, "learning_rate": 3.9053306790788856e-06, "loss": 23.4624, "step": 58330 }, { "epoch": 0.72, "learning_rate": 3.902173272952148e-06, "loss": 23.61, "step": 58340 }, { "epoch": 0.72, "learning_rate": 3.899016834276709e-06, "loss": 23.2141, "step": 58350 }, { "epoch": 0.72, "learning_rate": 3.895861363553353e-06, "loss": 23.6634, "step": 58360 }, { "epoch": 0.72, "learning_rate": 3.892706861282709e-06, "loss": 23.323, "step": 58370 }, { "epoch": 0.72, "learning_rate": 3.889553327965262e-06, "loss": 23.7024, "step": 58380 }, { "epoch": 0.72, "learning_rate": 3.886400764101329e-06, "loss": 24.0065, "step": 58390 }, { "epoch": 0.72, "learning_rate": 3.883249170191087e-06, "loss": 23.7992, "step": 58400 }, { "epoch": 0.72, "learning_rate": 3.880098546734549e-06, "loss": 23.6264, "step": 58410 }, { "epoch": 0.72, "learning_rate": 3.876948894231573e-06, "loss": 23.4171, "step": 58420 }, { "epoch": 0.72, "learning_rate": 3.8738002131818765e-06, "loss": 23.8916, "step": 58430 }, { "epoch": 0.72, "learning_rate": 3.870652504085009e-06, "loss": 24.2209, "step": 58440 }, { "epoch": 0.72, "learning_rate": 3.8675057674403674e-06, "loss": 24.095, "step": 58450 }, { "epoch": 0.72, "learning_rate": 3.864360003747204e-06, "loss": 23.8479, "step": 58460 }, { "epoch": 0.72, "learning_rate": 3.861215213504608e-06, "loss": 24.2019, "step": 58470 }, { "epoch": 0.72, "learning_rate": 3.858071397211512e-06, "loss": 24.1843, "step": 58480 }, { "epoch": 0.72, "learning_rate": 3.854928555366705e-06, "loss": 23.9023, "step": 58490 }, { "epoch": 0.72, "learning_rate": 3.8517866884688125e-06, "loss": 23.9624, "step": 58500 }, { "epoch": 0.72, "learning_rate": 3.8486457970163036e-06, "loss": 24.6943, "step": 58510 }, { "epoch": 0.72, "learning_rate": 3.845505881507503e-06, "loss": 23.775, "step": 58520 }, { "epoch": 0.72, "learning_rate": 3.842366942440573e-06, "loss": 24.0092, "step": 58530 }, { "epoch": 0.72, "learning_rate": 3.839228980313517e-06, "loss": 23.8664, "step": 58540 }, { "epoch": 0.72, "learning_rate": 3.836091995624196e-06, "loss": 23.7933, "step": 58550 }, { "epoch": 0.72, "learning_rate": 3.832955988870305e-06, "loss": 24.0963, "step": 58560 }, { "epoch": 0.72, "learning_rate": 3.829820960549387e-06, "loss": 23.8005, "step": 58570 }, { "epoch": 0.72, "learning_rate": 3.826686911158832e-06, "loss": 23.9772, "step": 58580 }, { "epoch": 0.72, "learning_rate": 3.823553841195869e-06, "loss": 23.986, "step": 58590 }, { "epoch": 0.72, "learning_rate": 3.82042175115758e-06, "loss": 23.7779, "step": 58600 }, { "epoch": 0.72, "learning_rate": 3.8172906415408865e-06, "loss": 24.0824, "step": 58610 }, { "epoch": 0.72, "learning_rate": 3.8141605128425498e-06, "loss": 23.9664, "step": 58620 }, { "epoch": 0.72, "learning_rate": 3.811031365559189e-06, "loss": 23.7818, "step": 58630 }, { "epoch": 0.72, "learning_rate": 3.807903200187255e-06, "loss": 23.5752, "step": 58640 }, { "epoch": 0.72, "learning_rate": 3.8047760172230442e-06, "loss": 23.2208, "step": 58650 }, { "epoch": 0.72, "learning_rate": 3.801649817162706e-06, "loss": 23.9828, "step": 58660 }, { "epoch": 0.72, "learning_rate": 3.7985246005022258e-06, "loss": 23.804, "step": 58670 }, { "epoch": 0.72, "learning_rate": 3.7954003677374306e-06, "loss": 23.7478, "step": 58680 }, { "epoch": 0.72, "learning_rate": 3.792277119364003e-06, "loss": 24.0534, "step": 58690 }, { "epoch": 0.72, "learning_rate": 3.7891548558774595e-06, "loss": 23.7765, "step": 58700 }, { "epoch": 0.72, "learning_rate": 3.7860335777731594e-06, "loss": 23.8388, "step": 58710 }, { "epoch": 0.72, "learning_rate": 3.782913285546317e-06, "loss": 23.8768, "step": 58720 }, { "epoch": 0.72, "learning_rate": 3.779793979691977e-06, "loss": 24.1628, "step": 58730 }, { "epoch": 0.72, "learning_rate": 3.7766756607050313e-06, "loss": 23.8717, "step": 58740 }, { "epoch": 0.72, "learning_rate": 3.7735583290802247e-06, "loss": 23.8375, "step": 58750 }, { "epoch": 0.72, "learning_rate": 3.770441985312132e-06, "loss": 24.201, "step": 58760 }, { "epoch": 0.72, "learning_rate": 3.767326629895176e-06, "loss": 23.8534, "step": 58770 }, { "epoch": 0.72, "learning_rate": 3.764212263323629e-06, "loss": 23.9607, "step": 58780 }, { "epoch": 0.72, "learning_rate": 3.7610988860915976e-06, "loss": 24.0448, "step": 58790 }, { "epoch": 0.72, "learning_rate": 3.757986498693037e-06, "loss": 23.9066, "step": 58800 }, { "epoch": 0.72, "learning_rate": 3.7548751016217413e-06, "loss": 23.7242, "step": 58810 }, { "epoch": 0.72, "learning_rate": 3.751764695371346e-06, "loss": 23.814, "step": 58820 }, { "epoch": 0.72, "learning_rate": 3.7486552804353404e-06, "loss": 23.8536, "step": 58830 }, { "epoch": 0.72, "learning_rate": 3.7455468573070465e-06, "loss": 23.8587, "step": 58840 }, { "epoch": 0.72, "learning_rate": 3.742439426479627e-06, "loss": 23.9154, "step": 58850 }, { "epoch": 0.72, "learning_rate": 3.739332988446098e-06, "loss": 23.422, "step": 58860 }, { "epoch": 0.72, "learning_rate": 3.73622754369931e-06, "loss": 24.2438, "step": 58870 }, { "epoch": 0.72, "learning_rate": 3.7331230927319527e-06, "loss": 23.3004, "step": 58880 }, { "epoch": 0.72, "learning_rate": 3.7300196360365714e-06, "loss": 23.1421, "step": 58890 }, { "epoch": 0.72, "learning_rate": 3.72691717410554e-06, "loss": 23.2536, "step": 58900 }, { "epoch": 0.72, "learning_rate": 3.7238157074310777e-06, "loss": 23.3883, "step": 58910 }, { "epoch": 0.72, "learning_rate": 3.7207152365052547e-06, "loss": 23.8782, "step": 58920 }, { "epoch": 0.72, "learning_rate": 3.7176157618199725e-06, "loss": 22.8133, "step": 58930 }, { "epoch": 0.72, "learning_rate": 3.7145172838669762e-06, "loss": 24.1872, "step": 58940 }, { "epoch": 0.72, "learning_rate": 3.711419803137861e-06, "loss": 23.9825, "step": 58950 }, { "epoch": 0.73, "learning_rate": 3.708323320124055e-06, "loss": 23.7346, "step": 58960 }, { "epoch": 0.73, "learning_rate": 3.705227835316828e-06, "loss": 23.6877, "step": 58970 }, { "epoch": 0.73, "learning_rate": 3.7021333492072997e-06, "loss": 23.7994, "step": 58980 }, { "epoch": 0.73, "learning_rate": 3.6990398622864255e-06, "loss": 23.6147, "step": 58990 }, { "epoch": 0.73, "learning_rate": 3.6959473750449957e-06, "loss": 23.3444, "step": 59000 }, { "epoch": 0.73, "learning_rate": 3.6928558879736597e-06, "loss": 24.2513, "step": 59010 }, { "epoch": 0.73, "learning_rate": 3.6897654015628923e-06, "loss": 23.7865, "step": 59020 }, { "epoch": 0.73, "learning_rate": 3.686675916303015e-06, "loss": 23.8887, "step": 59030 }, { "epoch": 0.73, "learning_rate": 3.683587432684191e-06, "loss": 23.7308, "step": 59040 }, { "epoch": 0.73, "learning_rate": 3.6804999511964203e-06, "loss": 23.8967, "step": 59050 }, { "epoch": 0.73, "learning_rate": 3.677413472329554e-06, "loss": 24.0506, "step": 59060 }, { "epoch": 0.73, "learning_rate": 3.674327996573275e-06, "loss": 24.0301, "step": 59070 }, { "epoch": 0.73, "learning_rate": 3.671243524417106e-06, "loss": 23.5758, "step": 59080 }, { "epoch": 0.73, "learning_rate": 3.6681600563504203e-06, "loss": 23.835, "step": 59090 }, { "epoch": 0.73, "learning_rate": 3.665077592862424e-06, "loss": 23.3002, "step": 59100 }, { "epoch": 0.73, "learning_rate": 3.6619961344421605e-06, "loss": 23.5848, "step": 59110 }, { "epoch": 0.73, "learning_rate": 3.6589156815785276e-06, "loss": 23.6983, "step": 59120 }, { "epoch": 0.73, "learning_rate": 3.6558362347602503e-06, "loss": 24.1352, "step": 59130 }, { "epoch": 0.73, "learning_rate": 3.6527577944758953e-06, "loss": 23.9383, "step": 59140 }, { "epoch": 0.73, "learning_rate": 3.6496803612138786e-06, "loss": 23.9459, "step": 59150 }, { "epoch": 0.73, "learning_rate": 3.6466039354624494e-06, "loss": 23.4698, "step": 59160 }, { "epoch": 0.73, "learning_rate": 3.6435285177096937e-06, "loss": 23.6456, "step": 59170 }, { "epoch": 0.73, "learning_rate": 3.6404541084435485e-06, "loss": 24.0689, "step": 59180 }, { "epoch": 0.73, "learning_rate": 3.637380708151782e-06, "loss": 24.1365, "step": 59190 }, { "epoch": 0.73, "learning_rate": 3.6343083173220005e-06, "loss": 23.8505, "step": 59200 }, { "epoch": 0.73, "learning_rate": 3.631236936441663e-06, "loss": 23.0877, "step": 59210 }, { "epoch": 0.73, "learning_rate": 3.6281665659980547e-06, "loss": 23.8905, "step": 59220 }, { "epoch": 0.73, "learning_rate": 3.625097206478303e-06, "loss": 24.0448, "step": 59230 }, { "epoch": 0.73, "learning_rate": 3.622028858369384e-06, "loss": 23.7946, "step": 59240 }, { "epoch": 0.73, "learning_rate": 3.6189615221581033e-06, "loss": 23.8469, "step": 59250 }, { "epoch": 0.73, "learning_rate": 3.6158951983311095e-06, "loss": 23.7636, "step": 59260 }, { "epoch": 0.73, "learning_rate": 3.61282988737489e-06, "loss": 23.636, "step": 59270 }, { "epoch": 0.73, "learning_rate": 3.609765589775769e-06, "loss": 24.0973, "step": 59280 }, { "epoch": 0.73, "learning_rate": 3.6067023060199203e-06, "loss": 24.0497, "step": 59290 }, { "epoch": 0.73, "learning_rate": 3.6036400365933454e-06, "loss": 24.1498, "step": 59300 }, { "epoch": 0.73, "learning_rate": 3.600578781981886e-06, "loss": 24.1967, "step": 59310 }, { "epoch": 0.73, "learning_rate": 3.5975185426712313e-06, "loss": 24.0468, "step": 59320 }, { "epoch": 0.73, "learning_rate": 3.594459319146902e-06, "loss": 23.7257, "step": 59330 }, { "epoch": 0.73, "learning_rate": 3.591401111894256e-06, "loss": 24.1888, "step": 59340 }, { "epoch": 0.73, "learning_rate": 3.5883439213984993e-06, "loss": 23.5769, "step": 59350 }, { "epoch": 0.73, "learning_rate": 3.585287748144668e-06, "loss": 23.5706, "step": 59360 }, { "epoch": 0.73, "learning_rate": 3.582232592617636e-06, "loss": 24.2834, "step": 59370 }, { "epoch": 0.73, "learning_rate": 3.5791784553021268e-06, "loss": 24.0479, "step": 59380 }, { "epoch": 0.73, "learning_rate": 3.57612533668269e-06, "loss": 23.2864, "step": 59390 }, { "epoch": 0.73, "learning_rate": 3.573073237243716e-06, "loss": 23.729, "step": 59400 }, { "epoch": 0.73, "learning_rate": 3.5700221574694425e-06, "loss": 23.9236, "step": 59410 }, { "epoch": 0.73, "learning_rate": 3.566972097843936e-06, "loss": 23.6004, "step": 59420 }, { "epoch": 0.73, "learning_rate": 3.5639230588511e-06, "loss": 23.3258, "step": 59430 }, { "epoch": 0.73, "learning_rate": 3.560875040974686e-06, "loss": 23.7623, "step": 59440 }, { "epoch": 0.73, "learning_rate": 3.557828044698277e-06, "loss": 23.9583, "step": 59450 }, { "epoch": 0.73, "learning_rate": 3.5547820705052884e-06, "loss": 23.1156, "step": 59460 }, { "epoch": 0.73, "learning_rate": 3.5517371188789873e-06, "loss": 23.8044, "step": 59470 }, { "epoch": 0.73, "learning_rate": 3.5486931903024637e-06, "loss": 24.0947, "step": 59480 }, { "epoch": 0.73, "learning_rate": 3.545650285258663e-06, "loss": 23.7597, "step": 59490 }, { "epoch": 0.73, "learning_rate": 3.5426084042303466e-06, "loss": 23.8537, "step": 59500 }, { "epoch": 0.73, "learning_rate": 3.5395675477001257e-06, "loss": 23.7015, "step": 59510 }, { "epoch": 0.73, "learning_rate": 3.536527716150454e-06, "loss": 24.0634, "step": 59520 }, { "epoch": 0.73, "learning_rate": 3.5334889100636125e-06, "loss": 23.4786, "step": 59530 }, { "epoch": 0.73, "learning_rate": 3.5304511299217205e-06, "loss": 23.8234, "step": 59540 }, { "epoch": 0.73, "learning_rate": 3.5274143762067437e-06, "loss": 23.9857, "step": 59550 }, { "epoch": 0.73, "learning_rate": 3.524378649400474e-06, "loss": 23.8932, "step": 59560 }, { "epoch": 0.73, "learning_rate": 3.521343949984544e-06, "loss": 24.1586, "step": 59570 }, { "epoch": 0.73, "learning_rate": 3.518310278440429e-06, "loss": 23.8352, "step": 59580 }, { "epoch": 0.73, "learning_rate": 3.5152776352494346e-06, "loss": 23.4159, "step": 59590 }, { "epoch": 0.73, "learning_rate": 3.5122460208927e-06, "loss": 23.1496, "step": 59600 }, { "epoch": 0.73, "learning_rate": 3.5092154358512143e-06, "loss": 23.9584, "step": 59610 }, { "epoch": 0.73, "learning_rate": 3.506185880605791e-06, "loss": 23.9565, "step": 59620 }, { "epoch": 0.73, "learning_rate": 3.5031573556370825e-06, "loss": 24.0926, "step": 59630 }, { "epoch": 0.73, "learning_rate": 3.5001298614255853e-06, "loss": 23.605, "step": 59640 }, { "epoch": 0.73, "learning_rate": 3.49710339845162e-06, "loss": 23.4209, "step": 59650 }, { "epoch": 0.73, "learning_rate": 3.4940779671953584e-06, "loss": 23.8985, "step": 59660 }, { "epoch": 0.73, "learning_rate": 3.4910535681367963e-06, "loss": 22.8712, "step": 59670 }, { "epoch": 0.73, "learning_rate": 3.4880302017557656e-06, "loss": 22.9791, "step": 59680 }, { "epoch": 0.73, "learning_rate": 3.4850078685319475e-06, "loss": 23.9236, "step": 59690 }, { "epoch": 0.73, "learning_rate": 3.481986568944845e-06, "loss": 23.7236, "step": 59700 }, { "epoch": 0.73, "learning_rate": 3.4789663034738018e-06, "loss": 23.7934, "step": 59710 }, { "epoch": 0.73, "learning_rate": 3.475947072598007e-06, "loss": 24.5144, "step": 59720 }, { "epoch": 0.73, "learning_rate": 3.4729288767964676e-06, "loss": 24.0213, "step": 59730 }, { "epoch": 0.73, "learning_rate": 3.4699117165480356e-06, "loss": 24.1166, "step": 59740 }, { "epoch": 0.73, "learning_rate": 3.4668955923314053e-06, "loss": 24.1213, "step": 59750 }, { "epoch": 0.73, "learning_rate": 3.4638805046250955e-06, "loss": 23.2822, "step": 59760 }, { "epoch": 0.74, "learning_rate": 3.4608664539074645e-06, "loss": 24.4876, "step": 59770 }, { "epoch": 0.74, "learning_rate": 3.4578534406567123e-06, "loss": 23.5592, "step": 59780 }, { "epoch": 0.74, "learning_rate": 3.4548414653508642e-06, "loss": 23.9173, "step": 59790 }, { "epoch": 0.74, "learning_rate": 3.4518305284677844e-06, "loss": 23.8784, "step": 59800 }, { "epoch": 0.74, "learning_rate": 3.448820630485179e-06, "loss": 23.2792, "step": 59810 }, { "epoch": 0.74, "learning_rate": 3.4458117718805773e-06, "loss": 23.1756, "step": 59820 }, { "epoch": 0.74, "learning_rate": 3.442803953131355e-06, "loss": 23.6073, "step": 59830 }, { "epoch": 0.74, "learning_rate": 3.4397971747147173e-06, "loss": 24.0679, "step": 59840 }, { "epoch": 0.74, "learning_rate": 3.4367914371076994e-06, "loss": 23.8111, "step": 59850 }, { "epoch": 0.74, "learning_rate": 3.433786740787185e-06, "loss": 24.0518, "step": 59860 }, { "epoch": 0.74, "learning_rate": 3.4307830862298806e-06, "loss": 23.7123, "step": 59870 }, { "epoch": 0.74, "learning_rate": 3.4277804739123276e-06, "loss": 23.7478, "step": 59880 }, { "epoch": 0.74, "learning_rate": 3.4247789043109127e-06, "loss": 24.3408, "step": 59890 }, { "epoch": 0.74, "learning_rate": 3.4217783779018476e-06, "loss": 24.359, "step": 59900 }, { "epoch": 0.74, "learning_rate": 3.418778895161177e-06, "loss": 24.4198, "step": 59910 }, { "epoch": 0.74, "learning_rate": 3.41578045656479e-06, "loss": 24.0757, "step": 59920 }, { "epoch": 0.74, "learning_rate": 3.412783062588402e-06, "loss": 23.7154, "step": 59930 }, { "epoch": 0.74, "learning_rate": 3.4097867137075613e-06, "loss": 23.2116, "step": 59940 }, { "epoch": 0.74, "learning_rate": 3.406791410397661e-06, "loss": 23.8659, "step": 59950 }, { "epoch": 0.74, "learning_rate": 3.4037971531339177e-06, "loss": 23.6602, "step": 59960 }, { "epoch": 0.74, "learning_rate": 3.4008039423913854e-06, "loss": 24.1182, "step": 59970 }, { "epoch": 0.74, "learning_rate": 3.3978117786449517e-06, "loss": 24.0257, "step": 59980 }, { "epoch": 0.74, "learning_rate": 3.3948206623693346e-06, "loss": 23.6729, "step": 59990 }, { "epoch": 0.74, "learning_rate": 3.3918305940390996e-06, "loss": 23.8709, "step": 60000 }, { "epoch": 0.74, "learning_rate": 3.3888415741286307e-06, "loss": 23.8717, "step": 60010 }, { "epoch": 0.74, "learning_rate": 3.385853603112148e-06, "loss": 23.4954, "step": 60020 }, { "epoch": 0.74, "learning_rate": 3.382866681463716e-06, "loss": 24.7481, "step": 60030 }, { "epoch": 0.74, "learning_rate": 3.37988080965722e-06, "loss": 23.4865, "step": 60040 }, { "epoch": 0.74, "learning_rate": 3.376895988166381e-06, "loss": 23.6608, "step": 60050 }, { "epoch": 0.74, "learning_rate": 3.373912217464764e-06, "loss": 24.2054, "step": 60060 }, { "epoch": 0.74, "learning_rate": 3.3709294980257556e-06, "loss": 23.5324, "step": 60070 }, { "epoch": 0.74, "learning_rate": 3.3679478303225756e-06, "loss": 24.0557, "step": 60080 }, { "epoch": 0.74, "learning_rate": 3.364967214828289e-06, "loss": 23.6276, "step": 60090 }, { "epoch": 0.74, "learning_rate": 3.3619876520157792e-06, "loss": 23.9016, "step": 60100 }, { "epoch": 0.74, "learning_rate": 3.3590091423577675e-06, "loss": 23.8376, "step": 60110 }, { "epoch": 0.74, "learning_rate": 3.356031686326817e-06, "loss": 23.7784, "step": 60120 }, { "epoch": 0.74, "learning_rate": 3.3530552843953125e-06, "loss": 23.7143, "step": 60130 }, { "epoch": 0.74, "learning_rate": 3.3500799370354706e-06, "loss": 23.9476, "step": 60140 }, { "epoch": 0.74, "learning_rate": 3.3471056447193553e-06, "loss": 23.0172, "step": 60150 }, { "epoch": 0.74, "learning_rate": 3.3441324079188463e-06, "loss": 24.1166, "step": 60160 }, { "epoch": 0.74, "learning_rate": 3.3411602271056607e-06, "loss": 24.0659, "step": 60170 }, { "epoch": 0.74, "learning_rate": 3.338189102751358e-06, "loss": 23.616, "step": 60180 }, { "epoch": 0.74, "learning_rate": 3.3352190353273172e-06, "loss": 23.5193, "step": 60190 }, { "epoch": 0.74, "learning_rate": 3.332250025304755e-06, "loss": 23.7186, "step": 60200 }, { "epoch": 0.74, "learning_rate": 3.3292820731547217e-06, "loss": 23.8953, "step": 60210 }, { "epoch": 0.74, "learning_rate": 3.326315179348092e-06, "loss": 24.1326, "step": 60220 }, { "epoch": 0.74, "learning_rate": 3.3233493443555877e-06, "loss": 23.5299, "step": 60230 }, { "epoch": 0.74, "learning_rate": 3.320384568647749e-06, "loss": 23.8069, "step": 60240 }, { "epoch": 0.74, "learning_rate": 3.317420852694949e-06, "loss": 23.2224, "step": 60250 }, { "epoch": 0.74, "learning_rate": 3.314458196967405e-06, "loss": 24.2042, "step": 60260 }, { "epoch": 0.74, "learning_rate": 3.3114966019351513e-06, "loss": 23.4096, "step": 60270 }, { "epoch": 0.74, "learning_rate": 3.30853606806806e-06, "loss": 23.617, "step": 60280 }, { "epoch": 0.74, "learning_rate": 3.3055765958358386e-06, "loss": 23.6332, "step": 60290 }, { "epoch": 0.74, "learning_rate": 3.302618185708021e-06, "loss": 23.7758, "step": 60300 }, { "epoch": 0.74, "learning_rate": 3.2996608381539685e-06, "loss": 23.6947, "step": 60310 }, { "epoch": 0.74, "learning_rate": 3.2967045536428887e-06, "loss": 23.5888, "step": 60320 }, { "epoch": 0.74, "learning_rate": 3.2937493326438074e-06, "loss": 23.5274, "step": 60330 }, { "epoch": 0.74, "learning_rate": 3.2907951756255807e-06, "loss": 23.951, "step": 60340 }, { "epoch": 0.74, "learning_rate": 3.287842083056908e-06, "loss": 23.6309, "step": 60350 }, { "epoch": 0.74, "learning_rate": 3.284890055406309e-06, "loss": 23.8905, "step": 60360 }, { "epoch": 0.74, "learning_rate": 3.2819390931421356e-06, "loss": 23.9993, "step": 60370 }, { "epoch": 0.74, "learning_rate": 3.2789891967325794e-06, "loss": 23.6778, "step": 60380 }, { "epoch": 0.74, "learning_rate": 3.276040366645652e-06, "loss": 23.941, "step": 60390 }, { "epoch": 0.74, "learning_rate": 3.2730926033491974e-06, "loss": 23.6775, "step": 60400 }, { "epoch": 0.74, "learning_rate": 3.270145907310901e-06, "loss": 23.8975, "step": 60410 }, { "epoch": 0.74, "learning_rate": 3.267200278998267e-06, "loss": 23.4016, "step": 60420 }, { "epoch": 0.74, "learning_rate": 3.264255718878634e-06, "loss": 23.8172, "step": 60430 }, { "epoch": 0.74, "learning_rate": 3.261312227419172e-06, "loss": 23.674, "step": 60440 }, { "epoch": 0.74, "learning_rate": 3.2583698050868773e-06, "loss": 23.9752, "step": 60450 }, { "epoch": 0.74, "learning_rate": 3.255428452348588e-06, "loss": 24.0363, "step": 60460 }, { "epoch": 0.74, "learning_rate": 3.252488169670959e-06, "loss": 24.7198, "step": 60470 }, { "epoch": 0.74, "learning_rate": 3.249548957520481e-06, "loss": 23.4806, "step": 60480 }, { "epoch": 0.74, "learning_rate": 3.246610816363478e-06, "loss": 23.6826, "step": 60490 }, { "epoch": 0.74, "learning_rate": 3.2436737466661006e-06, "loss": 24.1673, "step": 60500 }, { "epoch": 0.74, "learning_rate": 3.240737748894326e-06, "loss": 23.879, "step": 60510 }, { "epoch": 0.74, "learning_rate": 3.2378028235139715e-06, "loss": 23.6812, "step": 60520 }, { "epoch": 0.74, "learning_rate": 3.2348689709906732e-06, "loss": 23.5563, "step": 60530 }, { "epoch": 0.74, "learning_rate": 3.231936191789902e-06, "loss": 24.1602, "step": 60540 }, { "epoch": 0.74, "learning_rate": 3.229004486376961e-06, "loss": 23.9744, "step": 60550 }, { "epoch": 0.74, "learning_rate": 3.226073855216979e-06, "loss": 23.6256, "step": 60560 }, { "epoch": 0.74, "learning_rate": 3.223144298774913e-06, "loss": 24.2241, "step": 60570 }, { "epoch": 0.75, "learning_rate": 3.220215817515556e-06, "loss": 23.7677, "step": 60580 }, { "epoch": 0.75, "learning_rate": 3.2172884119035263e-06, "loss": 24.3762, "step": 60590 }, { "epoch": 0.75, "learning_rate": 3.214362082403266e-06, "loss": 23.8542, "step": 60600 }, { "epoch": 0.75, "learning_rate": 3.2114368294790598e-06, "loss": 23.9587, "step": 60610 }, { "epoch": 0.75, "learning_rate": 3.2085126535950105e-06, "loss": 23.5771, "step": 60620 }, { "epoch": 0.75, "learning_rate": 3.205589555215051e-06, "loss": 23.5446, "step": 60630 }, { "epoch": 0.75, "learning_rate": 3.2026675348029513e-06, "loss": 23.6065, "step": 60640 }, { "epoch": 0.75, "learning_rate": 3.1997465928223014e-06, "loss": 23.7166, "step": 60650 }, { "epoch": 0.75, "learning_rate": 3.196826729736525e-06, "loss": 22.6474, "step": 60660 }, { "epoch": 0.75, "learning_rate": 3.1939079460088717e-06, "loss": 24.0496, "step": 60670 }, { "epoch": 0.75, "learning_rate": 3.1909902421024186e-06, "loss": 23.874, "step": 60680 }, { "epoch": 0.75, "learning_rate": 3.188073618480082e-06, "loss": 24.2239, "step": 60690 }, { "epoch": 0.75, "learning_rate": 3.185158075604594e-06, "loss": 24.1188, "step": 60700 }, { "epoch": 0.75, "learning_rate": 3.182243613938519e-06, "loss": 23.3062, "step": 60710 }, { "epoch": 0.75, "learning_rate": 3.1793302339442567e-06, "loss": 23.4363, "step": 60720 }, { "epoch": 0.75, "learning_rate": 3.1764179360840253e-06, "loss": 23.6898, "step": 60730 }, { "epoch": 0.75, "learning_rate": 3.1735067208198745e-06, "loss": 24.1344, "step": 60740 }, { "epoch": 0.75, "learning_rate": 3.170596588613689e-06, "loss": 23.9412, "step": 60750 }, { "epoch": 0.75, "learning_rate": 3.1676875399271733e-06, "loss": 23.9112, "step": 60760 }, { "epoch": 0.75, "learning_rate": 3.164779575221859e-06, "loss": 23.8077, "step": 60770 }, { "epoch": 0.75, "learning_rate": 3.161872694959116e-06, "loss": 23.3455, "step": 60780 }, { "epoch": 0.75, "learning_rate": 3.158966899600132e-06, "loss": 24.4103, "step": 60790 }, { "epoch": 0.75, "learning_rate": 3.156062189605924e-06, "loss": 24.4062, "step": 60800 }, { "epoch": 0.75, "learning_rate": 3.1531585654373452e-06, "loss": 23.3785, "step": 60810 }, { "epoch": 0.75, "learning_rate": 3.1502560275550674e-06, "loss": 23.8711, "step": 60820 }, { "epoch": 0.75, "learning_rate": 3.14735457641959e-06, "loss": 23.8438, "step": 60830 }, { "epoch": 0.75, "learning_rate": 3.1444542124912482e-06, "loss": 23.7806, "step": 60840 }, { "epoch": 0.75, "learning_rate": 3.1415549362301966e-06, "loss": 23.6935, "step": 60850 }, { "epoch": 0.75, "learning_rate": 3.138656748096418e-06, "loss": 23.4455, "step": 60860 }, { "epoch": 0.75, "learning_rate": 3.135759648549731e-06, "loss": 24.3152, "step": 60870 }, { "epoch": 0.75, "learning_rate": 3.1328636380497714e-06, "loss": 23.6114, "step": 60880 }, { "epoch": 0.75, "learning_rate": 3.1299687170560054e-06, "loss": 23.7892, "step": 60890 }, { "epoch": 0.75, "learning_rate": 3.1270748860277287e-06, "loss": 23.5041, "step": 60900 }, { "epoch": 0.75, "learning_rate": 3.1241821454240586e-06, "loss": 23.991, "step": 60910 }, { "epoch": 0.75, "learning_rate": 3.121290495703948e-06, "loss": 24.0701, "step": 60920 }, { "epoch": 0.75, "learning_rate": 3.118399937326171e-06, "loss": 23.3035, "step": 60930 }, { "epoch": 0.75, "learning_rate": 3.115510470749323e-06, "loss": 23.8076, "step": 60940 }, { "epoch": 0.75, "learning_rate": 3.1126220964318434e-06, "loss": 23.5337, "step": 60950 }, { "epoch": 0.75, "learning_rate": 3.1097348148319807e-06, "loss": 23.2412, "step": 60960 }, { "epoch": 0.75, "learning_rate": 3.1068486264078155e-06, "loss": 23.6504, "step": 60970 }, { "epoch": 0.75, "learning_rate": 3.1039635316172623e-06, "loss": 23.9739, "step": 60980 }, { "epoch": 0.75, "learning_rate": 3.101079530918052e-06, "loss": 24.2222, "step": 60990 }, { "epoch": 0.75, "learning_rate": 3.098196624767744e-06, "loss": 23.2311, "step": 61000 }, { "epoch": 0.75, "learning_rate": 3.0953148136237323e-06, "loss": 24.3987, "step": 61010 }, { "epoch": 0.75, "learning_rate": 3.0924340979432264e-06, "loss": 24.1625, "step": 61020 }, { "epoch": 0.75, "learning_rate": 3.089554478183265e-06, "loss": 24.0373, "step": 61030 }, { "epoch": 0.75, "learning_rate": 3.0866759548007196e-06, "loss": 23.7477, "step": 61040 }, { "epoch": 0.75, "learning_rate": 3.0837985282522775e-06, "loss": 23.2507, "step": 61050 }, { "epoch": 0.75, "learning_rate": 3.0809221989944627e-06, "loss": 23.096, "step": 61060 }, { "epoch": 0.75, "learning_rate": 3.0780469674836167e-06, "loss": 24.3493, "step": 61070 }, { "epoch": 0.75, "learning_rate": 3.0751728341759056e-06, "loss": 24.2069, "step": 61080 }, { "epoch": 0.75, "learning_rate": 3.072299799527333e-06, "loss": 23.5445, "step": 61090 }, { "epoch": 0.75, "learning_rate": 3.069427863993716e-06, "loss": 23.8211, "step": 61100 }, { "epoch": 0.75, "learning_rate": 3.0665570280307032e-06, "loss": 23.9786, "step": 61110 }, { "epoch": 0.75, "learning_rate": 3.0636872920937653e-06, "loss": 23.9178, "step": 61120 }, { "epoch": 0.75, "learning_rate": 3.0608186566382024e-06, "loss": 23.2358, "step": 61130 }, { "epoch": 0.75, "learning_rate": 3.0579511221191336e-06, "loss": 24.0028, "step": 61140 }, { "epoch": 0.75, "learning_rate": 3.0550846889915166e-06, "loss": 23.7496, "step": 61150 }, { "epoch": 0.75, "learning_rate": 3.05221935771012e-06, "loss": 23.7755, "step": 61160 }, { "epoch": 0.75, "learning_rate": 3.0493551287295396e-06, "loss": 23.5327, "step": 61170 }, { "epoch": 0.75, "learning_rate": 3.046492002504209e-06, "loss": 23.7053, "step": 61180 }, { "epoch": 0.75, "learning_rate": 3.043629979488373e-06, "loss": 24.2858, "step": 61190 }, { "epoch": 0.75, "learning_rate": 3.0407690601361027e-06, "loss": 23.6847, "step": 61200 }, { "epoch": 0.75, "learning_rate": 3.0379092449013047e-06, "loss": 23.3221, "step": 61210 }, { "epoch": 0.75, "learning_rate": 3.0350505342376955e-06, "loss": 24.0813, "step": 61220 }, { "epoch": 0.75, "learning_rate": 3.032192928598833e-06, "loss": 23.2357, "step": 61230 }, { "epoch": 0.75, "learning_rate": 3.0293364284380854e-06, "loss": 24.443, "step": 61240 }, { "epoch": 0.75, "learning_rate": 3.026481034208648e-06, "loss": 23.7123, "step": 61250 }, { "epoch": 0.75, "learning_rate": 3.0236267463635516e-06, "loss": 23.8533, "step": 61260 }, { "epoch": 0.75, "learning_rate": 3.020773565355639e-06, "loss": 24.1329, "step": 61270 }, { "epoch": 0.75, "learning_rate": 3.0179214916375787e-06, "loss": 23.568, "step": 61280 }, { "epoch": 0.75, "learning_rate": 3.015070525661875e-06, "loss": 23.8514, "step": 61290 }, { "epoch": 0.75, "learning_rate": 3.0122206678808417e-06, "loss": 24.2568, "step": 61300 }, { "epoch": 0.75, "learning_rate": 3.009371918746623e-06, "loss": 24.007, "step": 61310 }, { "epoch": 0.75, "learning_rate": 3.0065242787111927e-06, "loss": 23.7828, "step": 61320 }, { "epoch": 0.75, "learning_rate": 3.0036777482263402e-06, "loss": 23.9461, "step": 61330 }, { "epoch": 0.75, "learning_rate": 3.0008323277436824e-06, "loss": 24.1669, "step": 61340 }, { "epoch": 0.75, "learning_rate": 2.997988017714659e-06, "loss": 23.931, "step": 61350 }, { "epoch": 0.75, "learning_rate": 2.9951448185905354e-06, "loss": 23.7462, "step": 61360 }, { "epoch": 0.75, "learning_rate": 2.992302730822395e-06, "loss": 23.7113, "step": 61370 }, { "epoch": 0.75, "learning_rate": 2.989461754861157e-06, "loss": 23.9765, "step": 61380 }, { "epoch": 0.75, "learning_rate": 2.9866218911575505e-06, "loss": 23.7244, "step": 61390 }, { "epoch": 0.76, "learning_rate": 2.98378314016214e-06, "loss": 23.0659, "step": 61400 }, { "epoch": 0.76, "learning_rate": 2.980945502325304e-06, "loss": 24.1928, "step": 61410 }, { "epoch": 0.76, "learning_rate": 2.978108978097247e-06, "loss": 24.1541, "step": 61420 }, { "epoch": 0.76, "learning_rate": 2.9752735679280043e-06, "loss": 23.7053, "step": 61430 }, { "epoch": 0.76, "learning_rate": 2.9724392722674234e-06, "loss": 23.6924, "step": 61440 }, { "epoch": 0.76, "learning_rate": 2.969606091565176e-06, "loss": 23.2292, "step": 61450 }, { "epoch": 0.76, "learning_rate": 2.96677402627077e-06, "loss": 23.2494, "step": 61460 }, { "epoch": 0.76, "learning_rate": 2.9639430768335208e-06, "loss": 23.619, "step": 61470 }, { "epoch": 0.76, "learning_rate": 2.9611132437025712e-06, "loss": 23.5389, "step": 61480 }, { "epoch": 0.76, "learning_rate": 2.958284527326896e-06, "loss": 24.2278, "step": 61490 }, { "epoch": 0.76, "learning_rate": 2.95545692815528e-06, "loss": 23.9658, "step": 61500 }, { "epoch": 0.76, "learning_rate": 2.9526304466363352e-06, "loss": 23.4892, "step": 61510 }, { "epoch": 0.76, "learning_rate": 2.949805083218502e-06, "loss": 24.0904, "step": 61520 }, { "epoch": 0.76, "learning_rate": 2.9469808383500364e-06, "loss": 24.1708, "step": 61530 }, { "epoch": 0.76, "learning_rate": 2.9441577124790154e-06, "loss": 23.7037, "step": 61540 }, { "epoch": 0.76, "learning_rate": 2.9413357060533498e-06, "loss": 23.5578, "step": 61550 }, { "epoch": 0.76, "learning_rate": 2.9385148195207613e-06, "loss": 23.6961, "step": 61560 }, { "epoch": 0.76, "learning_rate": 2.935695053328795e-06, "loss": 22.9846, "step": 61570 }, { "epoch": 0.76, "learning_rate": 2.9328764079248308e-06, "loss": 23.1805, "step": 61580 }, { "epoch": 0.76, "learning_rate": 2.9300588837560474e-06, "loss": 23.7337, "step": 61590 }, { "epoch": 0.76, "learning_rate": 2.927242481269471e-06, "loss": 24.309, "step": 61600 }, { "epoch": 0.76, "learning_rate": 2.9244272009119325e-06, "loss": 23.5059, "step": 61610 }, { "epoch": 0.76, "learning_rate": 2.921613043130089e-06, "loss": 24.1967, "step": 61620 }, { "epoch": 0.76, "learning_rate": 2.918800008370428e-06, "loss": 23.6591, "step": 61630 }, { "epoch": 0.76, "learning_rate": 2.9159880970792476e-06, "loss": 24.2174, "step": 61640 }, { "epoch": 0.76, "learning_rate": 2.9131773097026673e-06, "loss": 24.3878, "step": 61650 }, { "epoch": 0.76, "learning_rate": 2.9103676466866416e-06, "loss": 23.8844, "step": 61660 }, { "epoch": 0.76, "learning_rate": 2.907559108476934e-06, "loss": 23.8659, "step": 61670 }, { "epoch": 0.76, "learning_rate": 2.904751695519129e-06, "loss": 24.2929, "step": 61680 }, { "epoch": 0.76, "learning_rate": 2.901945408258645e-06, "loss": 23.7583, "step": 61690 }, { "epoch": 0.76, "learning_rate": 2.8991402471407094e-06, "loss": 23.2741, "step": 61700 }, { "epoch": 0.76, "learning_rate": 2.896336212610372e-06, "loss": 23.7593, "step": 61710 }, { "epoch": 0.76, "learning_rate": 2.8935333051125147e-06, "loss": 24.1478, "step": 61720 }, { "epoch": 0.76, "learning_rate": 2.89073152509183e-06, "loss": 24.3323, "step": 61730 }, { "epoch": 0.76, "learning_rate": 2.8879308729928303e-06, "loss": 23.258, "step": 61740 }, { "epoch": 0.76, "learning_rate": 2.8851313492598598e-06, "loss": 24.6774, "step": 61750 }, { "epoch": 0.76, "learning_rate": 2.8823329543370736e-06, "loss": 24.3248, "step": 61760 }, { "epoch": 0.76, "learning_rate": 2.87953568866845e-06, "loss": 24.2166, "step": 61770 }, { "epoch": 0.76, "learning_rate": 2.876739552697795e-06, "loss": 23.8089, "step": 61780 }, { "epoch": 0.76, "learning_rate": 2.8739445468687256e-06, "loss": 23.8149, "step": 61790 }, { "epoch": 0.76, "learning_rate": 2.8711506716246805e-06, "loss": 24.1132, "step": 61800 }, { "epoch": 0.76, "learning_rate": 2.8683579274089334e-06, "loss": 23.0865, "step": 61810 }, { "epoch": 0.76, "learning_rate": 2.8655663146645542e-06, "loss": 23.8535, "step": 61820 }, { "epoch": 0.76, "learning_rate": 2.8627758338344556e-06, "loss": 23.6789, "step": 61830 }, { "epoch": 0.76, "learning_rate": 2.859986485361358e-06, "loss": 23.3885, "step": 61840 }, { "epoch": 0.76, "learning_rate": 2.857198269687804e-06, "loss": 23.7659, "step": 61850 }, { "epoch": 0.76, "learning_rate": 2.8544111872561633e-06, "loss": 23.9632, "step": 61860 }, { "epoch": 0.76, "learning_rate": 2.8516252385086172e-06, "loss": 23.8567, "step": 61870 }, { "epoch": 0.76, "learning_rate": 2.848840423887169e-06, "loss": 24.1822, "step": 61880 }, { "epoch": 0.76, "learning_rate": 2.8460567438336495e-06, "loss": 23.5165, "step": 61890 }, { "epoch": 0.76, "learning_rate": 2.8432741987896996e-06, "loss": 23.8441, "step": 61900 }, { "epoch": 0.76, "learning_rate": 2.8404927891967824e-06, "loss": 23.6221, "step": 61910 }, { "epoch": 0.76, "learning_rate": 2.8377125154961882e-06, "loss": 23.5513, "step": 61920 }, { "epoch": 0.76, "learning_rate": 2.83493337812902e-06, "loss": 23.8789, "step": 61930 }, { "epoch": 0.76, "learning_rate": 2.8321553775361964e-06, "loss": 23.8108, "step": 61940 }, { "epoch": 0.76, "learning_rate": 2.8293785141584694e-06, "loss": 24.1048, "step": 61950 }, { "epoch": 0.76, "learning_rate": 2.8266027884363987e-06, "loss": 23.9636, "step": 61960 }, { "epoch": 0.76, "learning_rate": 2.8238282008103647e-06, "loss": 24.1817, "step": 61970 }, { "epoch": 0.76, "learning_rate": 2.8210547517205754e-06, "loss": 23.5484, "step": 61980 }, { "epoch": 0.76, "learning_rate": 2.818282441607051e-06, "loss": 24.5486, "step": 61990 }, { "epoch": 0.76, "learning_rate": 2.815511270909628e-06, "loss": 24.1613, "step": 62000 }, { "epoch": 0.76, "learning_rate": 2.8127412400679734e-06, "loss": 23.5001, "step": 62010 }, { "epoch": 0.76, "learning_rate": 2.8099723495215637e-06, "loss": 23.7494, "step": 62020 }, { "epoch": 0.76, "learning_rate": 2.807204599709694e-06, "loss": 23.4105, "step": 62030 }, { "epoch": 0.76, "learning_rate": 2.8044379910714926e-06, "loss": 23.3678, "step": 62040 }, { "epoch": 0.76, "learning_rate": 2.801672524045882e-06, "loss": 23.7791, "step": 62050 }, { "epoch": 0.76, "learning_rate": 2.7989081990716283e-06, "loss": 23.4681, "step": 62060 }, { "epoch": 0.76, "learning_rate": 2.7961450165873018e-06, "loss": 24.1641, "step": 62070 }, { "epoch": 0.76, "learning_rate": 2.7933829770312926e-06, "loss": 24.1211, "step": 62080 }, { "epoch": 0.76, "learning_rate": 2.7906220808418183e-06, "loss": 23.8533, "step": 62090 }, { "epoch": 0.76, "learning_rate": 2.7878623284569063e-06, "loss": 23.5842, "step": 62100 }, { "epoch": 0.76, "learning_rate": 2.7851037203144014e-06, "loss": 24.3756, "step": 62110 }, { "epoch": 0.76, "learning_rate": 2.7823462568519797e-06, "loss": 24.2295, "step": 62120 }, { "epoch": 0.76, "learning_rate": 2.7795899385071213e-06, "loss": 23.3175, "step": 62130 }, { "epoch": 0.76, "learning_rate": 2.776834765717127e-06, "loss": 24.5865, "step": 62140 }, { "epoch": 0.76, "learning_rate": 2.774080738919127e-06, "loss": 24.3967, "step": 62150 }, { "epoch": 0.76, "learning_rate": 2.771327858550057e-06, "loss": 24.3238, "step": 62160 }, { "epoch": 0.76, "learning_rate": 2.7685761250466725e-06, "loss": 23.4925, "step": 62170 }, { "epoch": 0.76, "learning_rate": 2.765825538845557e-06, "loss": 23.7675, "step": 62180 }, { "epoch": 0.76, "learning_rate": 2.763076100383102e-06, "loss": 23.5242, "step": 62190 }, { "epoch": 0.76, "learning_rate": 2.7603278100955165e-06, "loss": 23.5836, "step": 62200 }, { "epoch": 0.77, "learning_rate": 2.7575806684188364e-06, "loss": 23.6203, "step": 62210 }, { "epoch": 0.77, "learning_rate": 2.7548346757889066e-06, "loss": 24.14, "step": 62220 }, { "epoch": 0.77, "learning_rate": 2.7520898326413914e-06, "loss": 23.7894, "step": 62230 }, { "epoch": 0.77, "learning_rate": 2.7493461394117794e-06, "loss": 23.5502, "step": 62240 }, { "epoch": 0.77, "learning_rate": 2.7466035965353668e-06, "loss": 23.5409, "step": 62250 }, { "epoch": 0.77, "learning_rate": 2.7438622044472717e-06, "loss": 23.8487, "step": 62260 }, { "epoch": 0.77, "learning_rate": 2.7411219635824373e-06, "loss": 24.3154, "step": 62270 }, { "epoch": 0.77, "learning_rate": 2.7383828743756047e-06, "loss": 24.6242, "step": 62280 }, { "epoch": 0.77, "learning_rate": 2.7356449372613545e-06, "loss": 24.0484, "step": 62290 }, { "epoch": 0.77, "learning_rate": 2.732908152674071e-06, "loss": 23.6412, "step": 62300 }, { "epoch": 0.77, "learning_rate": 2.7301725210479546e-06, "loss": 23.7453, "step": 62310 }, { "epoch": 0.77, "learning_rate": 2.7274380428170345e-06, "loss": 23.5962, "step": 62320 }, { "epoch": 0.77, "learning_rate": 2.7247047184151477e-06, "loss": 23.608, "step": 62330 }, { "epoch": 0.77, "learning_rate": 2.721972548275944e-06, "loss": 23.6075, "step": 62340 }, { "epoch": 0.77, "learning_rate": 2.7192415328329057e-06, "loss": 24.5155, "step": 62350 }, { "epoch": 0.77, "learning_rate": 2.7165116725193165e-06, "loss": 23.9784, "step": 62360 }, { "epoch": 0.77, "learning_rate": 2.7137829677682812e-06, "loss": 24.3491, "step": 62370 }, { "epoch": 0.77, "learning_rate": 2.711055419012728e-06, "loss": 23.184, "step": 62380 }, { "epoch": 0.77, "learning_rate": 2.7083290266853932e-06, "loss": 24.3862, "step": 62390 }, { "epoch": 0.77, "learning_rate": 2.7056037912188317e-06, "loss": 24.0085, "step": 62400 }, { "epoch": 0.77, "learning_rate": 2.70287971304542e-06, "loss": 24.4681, "step": 62410 }, { "epoch": 0.77, "learning_rate": 2.7001567925973437e-06, "loss": 23.7112, "step": 62420 }, { "epoch": 0.77, "learning_rate": 2.6974350303066065e-06, "loss": 24.2898, "step": 62430 }, { "epoch": 0.77, "learning_rate": 2.694714426605035e-06, "loss": 23.4153, "step": 62440 }, { "epoch": 0.77, "learning_rate": 2.6919949819242643e-06, "loss": 23.9222, "step": 62450 }, { "epoch": 0.77, "learning_rate": 2.6892766966957452e-06, "loss": 24.0079, "step": 62460 }, { "epoch": 0.77, "learning_rate": 2.686559571350753e-06, "loss": 24.4253, "step": 62470 }, { "epoch": 0.77, "learning_rate": 2.683843606320368e-06, "loss": 23.5768, "step": 62480 }, { "epoch": 0.77, "learning_rate": 2.6811288020354976e-06, "loss": 23.2771, "step": 62490 }, { "epoch": 0.77, "learning_rate": 2.6784151589268592e-06, "loss": 24.0266, "step": 62500 }, { "epoch": 0.77, "learning_rate": 2.675702677424977e-06, "loss": 23.6642, "step": 62510 }, { "epoch": 0.77, "learning_rate": 2.672991357960211e-06, "loss": 23.6819, "step": 62520 }, { "epoch": 0.77, "learning_rate": 2.6702812009627187e-06, "loss": 23.8736, "step": 62530 }, { "epoch": 0.77, "learning_rate": 2.6675722068624823e-06, "loss": 24.046, "step": 62540 }, { "epoch": 0.77, "learning_rate": 2.664864376089299e-06, "loss": 23.8595, "step": 62550 }, { "epoch": 0.77, "learning_rate": 2.66215770907278e-06, "loss": 23.9212, "step": 62560 }, { "epoch": 0.77, "learning_rate": 2.6594522062423477e-06, "loss": 23.9677, "step": 62570 }, { "epoch": 0.77, "learning_rate": 2.65674786802725e-06, "loss": 23.8406, "step": 62580 }, { "epoch": 0.77, "learning_rate": 2.6540446948565414e-06, "loss": 23.6902, "step": 62590 }, { "epoch": 0.77, "learning_rate": 2.6513426871590896e-06, "loss": 23.5804, "step": 62600 }, { "epoch": 0.77, "learning_rate": 2.6486418453635897e-06, "loss": 23.6443, "step": 62610 }, { "epoch": 0.77, "learning_rate": 2.6459421698985375e-06, "loss": 23.9378, "step": 62620 }, { "epoch": 0.77, "learning_rate": 2.6432436611922564e-06, "loss": 23.1078, "step": 62630 }, { "epoch": 0.77, "learning_rate": 2.640546319672874e-06, "loss": 23.9395, "step": 62640 }, { "epoch": 0.77, "learning_rate": 2.637850145768337e-06, "loss": 23.3263, "step": 62650 }, { "epoch": 0.77, "learning_rate": 2.6351551399064114e-06, "loss": 24.0188, "step": 62660 }, { "epoch": 0.77, "learning_rate": 2.6324613025146693e-06, "loss": 23.7402, "step": 62670 }, { "epoch": 0.77, "learning_rate": 2.629768634020502e-06, "loss": 24.3811, "step": 62680 }, { "epoch": 0.77, "learning_rate": 2.627077134851119e-06, "loss": 24.6472, "step": 62690 }, { "epoch": 0.77, "learning_rate": 2.6243868054335374e-06, "loss": 24.331, "step": 62700 }, { "epoch": 0.77, "learning_rate": 2.6216976461945897e-06, "loss": 23.3984, "step": 62710 }, { "epoch": 0.77, "learning_rate": 2.6190096575609302e-06, "loss": 24.4538, "step": 62720 }, { "epoch": 0.77, "learning_rate": 2.616322839959021e-06, "loss": 24.7091, "step": 62730 }, { "epoch": 0.77, "learning_rate": 2.6136371938151315e-06, "loss": 23.8525, "step": 62740 }, { "epoch": 0.77, "learning_rate": 2.610952719555362e-06, "loss": 24.0641, "step": 62750 }, { "epoch": 0.77, "learning_rate": 2.6082694176056154e-06, "loss": 23.9194, "step": 62760 }, { "epoch": 0.77, "learning_rate": 2.6055872883916076e-06, "loss": 23.8055, "step": 62770 }, { "epoch": 0.77, "learning_rate": 2.602906332338878e-06, "loss": 23.99, "step": 62780 }, { "epoch": 0.77, "learning_rate": 2.6002265498727685e-06, "loss": 23.9502, "step": 62790 }, { "epoch": 0.77, "learning_rate": 2.5975479414184456e-06, "loss": 23.8731, "step": 62800 }, { "epoch": 0.77, "learning_rate": 2.594870507400882e-06, "loss": 23.6574, "step": 62810 }, { "epoch": 0.77, "learning_rate": 2.5921942482448626e-06, "loss": 23.61, "step": 62820 }, { "epoch": 0.77, "learning_rate": 2.5895191643749963e-06, "loss": 23.9681, "step": 62830 }, { "epoch": 0.77, "learning_rate": 2.586845256215694e-06, "loss": 24.4873, "step": 62840 }, { "epoch": 0.77, "learning_rate": 2.5841725241911843e-06, "loss": 23.8715, "step": 62850 }, { "epoch": 0.77, "learning_rate": 2.5815009687255145e-06, "loss": 23.2039, "step": 62860 }, { "epoch": 0.77, "learning_rate": 2.578830590242537e-06, "loss": 23.9384, "step": 62870 }, { "epoch": 0.77, "learning_rate": 2.57616138916592e-06, "loss": 23.8166, "step": 62880 }, { "epoch": 0.77, "learning_rate": 2.5734933659191485e-06, "loss": 23.8408, "step": 62890 }, { "epoch": 0.77, "learning_rate": 2.570826520925518e-06, "loss": 22.9203, "step": 62900 }, { "epoch": 0.77, "learning_rate": 2.5681608546081327e-06, "loss": 23.9544, "step": 62910 }, { "epoch": 0.77, "learning_rate": 2.56549636738992e-06, "loss": 23.5944, "step": 62920 }, { "epoch": 0.77, "learning_rate": 2.5628330596936124e-06, "loss": 23.9785, "step": 62930 }, { "epoch": 0.77, "learning_rate": 2.5601709319417534e-06, "loss": 23.6978, "step": 62940 }, { "epoch": 0.77, "learning_rate": 2.5575099845567097e-06, "loss": 23.6502, "step": 62950 }, { "epoch": 0.77, "learning_rate": 2.5548502179606506e-06, "loss": 23.6949, "step": 62960 }, { "epoch": 0.77, "learning_rate": 2.5521916325755615e-06, "loss": 24.3474, "step": 62970 }, { "epoch": 0.77, "learning_rate": 2.5495342288232415e-06, "loss": 23.6333, "step": 62980 }, { "epoch": 0.77, "learning_rate": 2.5468780071252974e-06, "loss": 23.7187, "step": 62990 }, { "epoch": 0.77, "learning_rate": 2.5442229679031583e-06, "loss": 24.2127, "step": 63000 }, { "epoch": 0.77, "learning_rate": 2.541569111578057e-06, "loss": 23.525, "step": 63010 }, { "epoch": 0.78, "learning_rate": 2.5389164385710383e-06, "loss": 23.8611, "step": 63020 }, { "epoch": 0.78, "learning_rate": 2.536264949302969e-06, "loss": 24.2769, "step": 63030 }, { "epoch": 0.78, "learning_rate": 2.5336146441945177e-06, "loss": 23.3603, "step": 63040 }, { "epoch": 0.78, "learning_rate": 2.5309655236661647e-06, "loss": 24.3577, "step": 63050 }, { "epoch": 0.78, "learning_rate": 2.528317588138215e-06, "loss": 23.3702, "step": 63060 }, { "epoch": 0.78, "learning_rate": 2.5256708380307715e-06, "loss": 23.7319, "step": 63070 }, { "epoch": 0.78, "learning_rate": 2.523025273763753e-06, "loss": 24.0145, "step": 63080 }, { "epoch": 0.78, "learning_rate": 2.5203808957568975e-06, "loss": 23.1995, "step": 63090 }, { "epoch": 0.78, "learning_rate": 2.517737704429747e-06, "loss": 24.0497, "step": 63100 }, { "epoch": 0.78, "learning_rate": 2.515095700201653e-06, "loss": 24.0232, "step": 63110 }, { "epoch": 0.78, "learning_rate": 2.512454883491788e-06, "loss": 24.0609, "step": 63120 }, { "epoch": 0.78, "learning_rate": 2.5098152547191303e-06, "loss": 23.8185, "step": 63130 }, { "epoch": 0.78, "learning_rate": 2.5071768143024664e-06, "loss": 23.7429, "step": 63140 }, { "epoch": 0.78, "learning_rate": 2.5045395626604054e-06, "loss": 24.6378, "step": 63150 }, { "epoch": 0.78, "learning_rate": 2.5019035002113556e-06, "loss": 24.295, "step": 63160 }, { "epoch": 0.78, "learning_rate": 2.49926862737354e-06, "loss": 24.1158, "step": 63170 }, { "epoch": 0.78, "learning_rate": 2.4966349445650007e-06, "loss": 23.4537, "step": 63180 }, { "epoch": 0.78, "learning_rate": 2.494002452203581e-06, "loss": 23.7006, "step": 63190 }, { "epoch": 0.78, "learning_rate": 2.4913711507069406e-06, "loss": 23.8446, "step": 63200 }, { "epoch": 0.78, "learning_rate": 2.488741040492548e-06, "loss": 23.4453, "step": 63210 }, { "epoch": 0.78, "learning_rate": 2.486112121977682e-06, "loss": 23.6677, "step": 63220 }, { "epoch": 0.78, "learning_rate": 2.4834843955794373e-06, "loss": 23.3534, "step": 63230 }, { "epoch": 0.78, "learning_rate": 2.480857861714715e-06, "loss": 23.8196, "step": 63240 }, { "epoch": 0.78, "learning_rate": 2.4782325208002243e-06, "loss": 24.1292, "step": 63250 }, { "epoch": 0.78, "learning_rate": 2.475608373252496e-06, "loss": 23.4542, "step": 63260 }, { "epoch": 0.78, "learning_rate": 2.47298541948786e-06, "loss": 23.6561, "step": 63270 }, { "epoch": 0.78, "learning_rate": 2.4703636599224588e-06, "loss": 24.0885, "step": 63280 }, { "epoch": 0.78, "learning_rate": 2.467743094972255e-06, "loss": 24.6024, "step": 63290 }, { "epoch": 0.78, "learning_rate": 2.46512372505301e-06, "loss": 23.7678, "step": 63300 }, { "epoch": 0.78, "learning_rate": 2.4625055505802986e-06, "loss": 24.2239, "step": 63310 }, { "epoch": 0.78, "learning_rate": 2.459888571969513e-06, "loss": 23.8174, "step": 63320 }, { "epoch": 0.78, "learning_rate": 2.457272789635846e-06, "loss": 23.6555, "step": 63330 }, { "epoch": 0.78, "learning_rate": 2.4546582039943045e-06, "loss": 24.0947, "step": 63340 }, { "epoch": 0.78, "learning_rate": 2.452044815459709e-06, "loss": 23.6923, "step": 63350 }, { "epoch": 0.78, "learning_rate": 2.449432624446685e-06, "loss": 24.2447, "step": 63360 }, { "epoch": 0.78, "learning_rate": 2.446821631369667e-06, "loss": 23.8533, "step": 63370 }, { "epoch": 0.78, "learning_rate": 2.444211836642909e-06, "loss": 24.1624, "step": 63380 }, { "epoch": 0.78, "learning_rate": 2.4416032406804647e-06, "loss": 23.5413, "step": 63390 }, { "epoch": 0.78, "learning_rate": 2.438995843896198e-06, "loss": 23.8458, "step": 63400 }, { "epoch": 0.78, "learning_rate": 2.436389646703791e-06, "loss": 23.3203, "step": 63410 }, { "epoch": 0.78, "learning_rate": 2.4337846495167282e-06, "loss": 24.0346, "step": 63420 }, { "epoch": 0.78, "learning_rate": 2.431180852748306e-06, "loss": 24.1578, "step": 63430 }, { "epoch": 0.78, "learning_rate": 2.428578256811629e-06, "loss": 23.7121, "step": 63440 }, { "epoch": 0.78, "learning_rate": 2.4259768621196102e-06, "loss": 23.7797, "step": 63450 }, { "epoch": 0.78, "learning_rate": 2.4233766690849793e-06, "loss": 24.4706, "step": 63460 }, { "epoch": 0.78, "learning_rate": 2.4207776781202676e-06, "loss": 23.8007, "step": 63470 }, { "epoch": 0.78, "learning_rate": 2.418179889637816e-06, "loss": 23.1063, "step": 63480 }, { "epoch": 0.78, "learning_rate": 2.415583304049781e-06, "loss": 24.1014, "step": 63490 }, { "epoch": 0.78, "learning_rate": 2.412987921768123e-06, "loss": 23.351, "step": 63500 }, { "epoch": 0.78, "learning_rate": 2.4103937432046085e-06, "loss": 24.1878, "step": 63510 }, { "epoch": 0.78, "learning_rate": 2.407800768770824e-06, "loss": 24.4937, "step": 63520 }, { "epoch": 0.78, "learning_rate": 2.4052089988781547e-06, "loss": 24.4021, "step": 63530 }, { "epoch": 0.78, "learning_rate": 2.4026184339377954e-06, "loss": 23.876, "step": 63540 }, { "epoch": 0.78, "learning_rate": 2.400029074360757e-06, "loss": 23.2677, "step": 63550 }, { "epoch": 0.78, "learning_rate": 2.397440920557854e-06, "loss": 23.4246, "step": 63560 }, { "epoch": 0.78, "learning_rate": 2.3948539729397047e-06, "loss": 24.8218, "step": 63570 }, { "epoch": 0.78, "learning_rate": 2.39226823191675e-06, "loss": 24.4159, "step": 63580 }, { "epoch": 0.78, "learning_rate": 2.3896836978992256e-06, "loss": 23.935, "step": 63590 }, { "epoch": 0.78, "learning_rate": 2.3871003712971796e-06, "loss": 24.0448, "step": 63600 }, { "epoch": 0.78, "learning_rate": 2.384518252520476e-06, "loss": 23.4274, "step": 63610 }, { "epoch": 0.78, "learning_rate": 2.381937341978776e-06, "loss": 23.8428, "step": 63620 }, { "epoch": 0.78, "learning_rate": 2.3793576400815533e-06, "loss": 23.7867, "step": 63630 }, { "epoch": 0.78, "learning_rate": 2.3767791472380964e-06, "loss": 23.1562, "step": 63640 }, { "epoch": 0.78, "learning_rate": 2.3742018638574927e-06, "loss": 23.8324, "step": 63650 }, { "epoch": 0.78, "learning_rate": 2.3716257903486426e-06, "loss": 23.627, "step": 63660 }, { "epoch": 0.78, "learning_rate": 2.3690509271202523e-06, "loss": 23.6473, "step": 63670 }, { "epoch": 0.78, "learning_rate": 2.366477274580833e-06, "loss": 23.9891, "step": 63680 }, { "epoch": 0.78, "learning_rate": 2.363904833138716e-06, "loss": 24.4503, "step": 63690 }, { "epoch": 0.78, "learning_rate": 2.3613336032020273e-06, "loss": 23.4068, "step": 63700 }, { "epoch": 0.78, "learning_rate": 2.358763585178705e-06, "loss": 23.5895, "step": 63710 }, { "epoch": 0.78, "learning_rate": 2.356194779476498e-06, "loss": 23.9095, "step": 63720 }, { "epoch": 0.78, "learning_rate": 2.3536271865029613e-06, "loss": 23.4815, "step": 63730 }, { "epoch": 0.78, "learning_rate": 2.3510608066654504e-06, "loss": 23.7684, "step": 63740 }, { "epoch": 0.78, "learning_rate": 2.3484956403711435e-06, "loss": 23.0608, "step": 63750 }, { "epoch": 0.78, "learning_rate": 2.3459316880270123e-06, "loss": 24.1943, "step": 63760 }, { "epoch": 0.78, "learning_rate": 2.343368950039837e-06, "loss": 24.2399, "step": 63770 }, { "epoch": 0.78, "learning_rate": 2.3408074268162185e-06, "loss": 23.5114, "step": 63780 }, { "epoch": 0.78, "learning_rate": 2.338247118762549e-06, "loss": 23.7683, "step": 63790 }, { "epoch": 0.78, "learning_rate": 2.3356880262850333e-06, "loss": 23.6163, "step": 63800 }, { "epoch": 0.78, "learning_rate": 2.33313014978969e-06, "loss": 24.0111, "step": 63810 }, { "epoch": 0.78, "learning_rate": 2.3305734896823363e-06, "loss": 23.9755, "step": 63820 }, { "epoch": 0.79, "learning_rate": 2.3280180463685966e-06, "loss": 24.2734, "step": 63830 }, { "epoch": 0.79, "learning_rate": 2.325463820253909e-06, "loss": 23.977, "step": 63840 }, { "epoch": 0.79, "learning_rate": 2.322910811743515e-06, "loss": 24.0411, "step": 63850 }, { "epoch": 0.79, "learning_rate": 2.3203590212424574e-06, "loss": 24.0856, "step": 63860 }, { "epoch": 0.79, "learning_rate": 2.3178084491555975e-06, "loss": 24.1961, "step": 63870 }, { "epoch": 0.79, "learning_rate": 2.3152590958875898e-06, "loss": 23.4093, "step": 63880 }, { "epoch": 0.79, "learning_rate": 2.312710961842911e-06, "loss": 24.078, "step": 63890 }, { "epoch": 0.79, "learning_rate": 2.310164047425827e-06, "loss": 23.912, "step": 63900 }, { "epoch": 0.79, "learning_rate": 2.3076183530404193e-06, "loss": 23.8028, "step": 63910 }, { "epoch": 0.79, "learning_rate": 2.3050738790905804e-06, "loss": 24.1374, "step": 63920 }, { "epoch": 0.79, "learning_rate": 2.302530625980002e-06, "loss": 23.4769, "step": 63930 }, { "epoch": 0.79, "learning_rate": 2.29998859411218e-06, "loss": 23.6685, "step": 63940 }, { "epoch": 0.79, "learning_rate": 2.2974477838904286e-06, "loss": 23.6972, "step": 63950 }, { "epoch": 0.79, "learning_rate": 2.294908195717854e-06, "loss": 23.654, "step": 63960 }, { "epoch": 0.79, "learning_rate": 2.292369829997376e-06, "loss": 23.8815, "step": 63970 }, { "epoch": 0.79, "learning_rate": 2.2898326871317223e-06, "loss": 24.2549, "step": 63980 }, { "epoch": 0.79, "learning_rate": 2.2872967675234205e-06, "loss": 24.0718, "step": 63990 }, { "epoch": 0.79, "learning_rate": 2.2847620715748066e-06, "loss": 24.2427, "step": 64000 }, { "epoch": 0.79, "learning_rate": 2.2822285996880254e-06, "loss": 24.2112, "step": 64010 }, { "epoch": 0.79, "learning_rate": 2.279696352265023e-06, "loss": 23.9901, "step": 64020 }, { "epoch": 0.79, "learning_rate": 2.277165329707557e-06, "loss": 23.8621, "step": 64030 }, { "epoch": 0.79, "learning_rate": 2.2746355324171853e-06, "loss": 23.9324, "step": 64040 }, { "epoch": 0.79, "learning_rate": 2.2721069607952683e-06, "loss": 22.8691, "step": 64050 }, { "epoch": 0.79, "learning_rate": 2.269579615242985e-06, "loss": 23.3984, "step": 64060 }, { "epoch": 0.79, "learning_rate": 2.2670534961613066e-06, "loss": 24.0907, "step": 64070 }, { "epoch": 0.79, "learning_rate": 2.2645286039510128e-06, "loss": 24.0521, "step": 64080 }, { "epoch": 0.79, "learning_rate": 2.2620049390126974e-06, "loss": 24.0425, "step": 64090 }, { "epoch": 0.79, "learning_rate": 2.259482501746748e-06, "loss": 23.8767, "step": 64100 }, { "epoch": 0.79, "learning_rate": 2.2569612925533615e-06, "loss": 23.6223, "step": 64110 }, { "epoch": 0.79, "learning_rate": 2.2544413118325457e-06, "loss": 23.8936, "step": 64120 }, { "epoch": 0.79, "learning_rate": 2.2519225599841034e-06, "loss": 24.5876, "step": 64130 }, { "epoch": 0.79, "learning_rate": 2.249405037407646e-06, "loss": 24.018, "step": 64140 }, { "epoch": 0.79, "learning_rate": 2.2468887445025966e-06, "loss": 24.1075, "step": 64150 }, { "epoch": 0.79, "learning_rate": 2.244373681668176e-06, "loss": 23.0719, "step": 64160 }, { "epoch": 0.79, "learning_rate": 2.2418598493034072e-06, "loss": 23.2017, "step": 64170 }, { "epoch": 0.79, "learning_rate": 2.23934724780713e-06, "loss": 24.5973, "step": 64180 }, { "epoch": 0.79, "learning_rate": 2.2368358775779754e-06, "loss": 23.5969, "step": 64190 }, { "epoch": 0.79, "learning_rate": 2.23432573901439e-06, "loss": 24.1647, "step": 64200 }, { "epoch": 0.79, "learning_rate": 2.231816832514617e-06, "loss": 23.5885, "step": 64210 }, { "epoch": 0.79, "learning_rate": 2.229309158476706e-06, "loss": 23.8338, "step": 64220 }, { "epoch": 0.79, "learning_rate": 2.2268027172985173e-06, "loss": 23.7829, "step": 64230 }, { "epoch": 0.79, "learning_rate": 2.2242975093777066e-06, "loss": 23.874, "step": 64240 }, { "epoch": 0.79, "learning_rate": 2.221793535111736e-06, "loss": 23.6297, "step": 64250 }, { "epoch": 0.79, "learning_rate": 2.2192907948978794e-06, "loss": 24.2883, "step": 64260 }, { "epoch": 0.79, "learning_rate": 2.2167892891332066e-06, "loss": 23.4704, "step": 64270 }, { "epoch": 0.79, "learning_rate": 2.2142890182145915e-06, "loss": 24.015, "step": 64280 }, { "epoch": 0.79, "learning_rate": 2.21178998253872e-06, "loss": 24.0066, "step": 64290 }, { "epoch": 0.79, "learning_rate": 2.209292182502074e-06, "loss": 23.5925, "step": 64300 }, { "epoch": 0.79, "learning_rate": 2.2067956185009397e-06, "loss": 24.4252, "step": 64310 }, { "epoch": 0.79, "learning_rate": 2.204300290931416e-06, "loss": 23.6228, "step": 64320 }, { "epoch": 0.79, "learning_rate": 2.2018062001893947e-06, "loss": 23.6486, "step": 64330 }, { "epoch": 0.79, "learning_rate": 2.199313346670575e-06, "loss": 23.7158, "step": 64340 }, { "epoch": 0.79, "learning_rate": 2.1968217307704685e-06, "loss": 23.0061, "step": 64350 }, { "epoch": 0.79, "learning_rate": 2.194331352884371e-06, "loss": 24.2309, "step": 64360 }, { "epoch": 0.79, "learning_rate": 2.1918422134074034e-06, "loss": 23.8466, "step": 64370 }, { "epoch": 0.79, "learning_rate": 2.1893543127344763e-06, "loss": 24.0892, "step": 64380 }, { "epoch": 0.79, "learning_rate": 2.1868676512603058e-06, "loss": 24.189, "step": 64390 }, { "epoch": 0.79, "learning_rate": 2.1843822293794183e-06, "loss": 24.3819, "step": 64400 }, { "epoch": 0.79, "learning_rate": 2.1818980474861372e-06, "loss": 24.0171, "step": 64410 }, { "epoch": 0.79, "learning_rate": 2.1794151059745863e-06, "loss": 24.0735, "step": 64420 }, { "epoch": 0.79, "learning_rate": 2.176933405238704e-06, "loss": 23.7516, "step": 64430 }, { "epoch": 0.79, "learning_rate": 2.1744529456722197e-06, "loss": 24.1156, "step": 64440 }, { "epoch": 0.79, "learning_rate": 2.1719737276686705e-06, "loss": 23.5414, "step": 64450 }, { "epoch": 0.79, "learning_rate": 2.169495751621401e-06, "loss": 23.969, "step": 64460 }, { "epoch": 0.79, "learning_rate": 2.1670190179235527e-06, "loss": 24.0278, "step": 64470 }, { "epoch": 0.79, "learning_rate": 2.1645435269680705e-06, "loss": 23.8275, "step": 64480 }, { "epoch": 0.79, "learning_rate": 2.162069279147707e-06, "loss": 23.9507, "step": 64490 }, { "epoch": 0.79, "learning_rate": 2.1595962748550114e-06, "loss": 24.071, "step": 64500 }, { "epoch": 0.79, "learning_rate": 2.157124514482338e-06, "loss": 24.096, "step": 64510 }, { "epoch": 0.79, "learning_rate": 2.1546539984218475e-06, "loss": 24.5021, "step": 64520 }, { "epoch": 0.79, "learning_rate": 2.152184727065498e-06, "loss": 23.527, "step": 64530 }, { "epoch": 0.79, "learning_rate": 2.14971670080505e-06, "loss": 23.958, "step": 64540 }, { "epoch": 0.79, "learning_rate": 2.147249920032072e-06, "loss": 23.5281, "step": 64550 }, { "epoch": 0.79, "learning_rate": 2.14478438513793e-06, "loss": 23.7847, "step": 64560 }, { "epoch": 0.79, "learning_rate": 2.1423200965137914e-06, "loss": 24.6499, "step": 64570 }, { "epoch": 0.79, "learning_rate": 2.139857054550635e-06, "loss": 23.8681, "step": 64580 }, { "epoch": 0.79, "learning_rate": 2.1373952596392243e-06, "loss": 23.8284, "step": 64590 }, { "epoch": 0.79, "learning_rate": 2.1349347121701446e-06, "loss": 23.753, "step": 64600 }, { "epoch": 0.79, "learning_rate": 2.132475412533771e-06, "loss": 23.4293, "step": 64610 }, { "epoch": 0.79, "learning_rate": 2.1300173611202813e-06, "loss": 23.8611, "step": 64620 }, { "epoch": 0.79, "learning_rate": 2.1275605583196635e-06, "loss": 23.6168, "step": 64630 }, { "epoch": 0.79, "learning_rate": 2.125105004521698e-06, "loss": 23.9148, "step": 64640 }, { "epoch": 0.8, "learning_rate": 2.12265070011597e-06, "loss": 24.0286, "step": 64650 }, { "epoch": 0.8, "learning_rate": 2.1201976454918727e-06, "loss": 24.1839, "step": 64660 }, { "epoch": 0.8, "learning_rate": 2.117745841038592e-06, "loss": 23.7413, "step": 64670 }, { "epoch": 0.8, "learning_rate": 2.115295287145117e-06, "loss": 23.9937, "step": 64680 }, { "epoch": 0.8, "learning_rate": 2.112845984200245e-06, "loss": 23.2984, "step": 64690 }, { "epoch": 0.8, "learning_rate": 2.1103979325925693e-06, "loss": 24.3101, "step": 64700 }, { "epoch": 0.8, "learning_rate": 2.1079511327104817e-06, "loss": 23.8525, "step": 64710 }, { "epoch": 0.8, "learning_rate": 2.1055055849421856e-06, "loss": 23.9135, "step": 64720 }, { "epoch": 0.8, "learning_rate": 2.1030612896756777e-06, "loss": 24.0974, "step": 64730 }, { "epoch": 0.8, "learning_rate": 2.1006182472987537e-06, "loss": 23.7317, "step": 64740 }, { "epoch": 0.8, "learning_rate": 2.0981764581990206e-06, "loss": 24.2534, "step": 64750 }, { "epoch": 0.8, "learning_rate": 2.0957359227638774e-06, "loss": 23.8071, "step": 64760 }, { "epoch": 0.8, "learning_rate": 2.0932966413805265e-06, "loss": 23.7219, "step": 64770 }, { "epoch": 0.8, "learning_rate": 2.0908586144359765e-06, "loss": 23.6482, "step": 64780 }, { "epoch": 0.8, "learning_rate": 2.08842184231703e-06, "loss": 23.8281, "step": 64790 }, { "epoch": 0.8, "learning_rate": 2.085986325410292e-06, "loss": 23.9614, "step": 64800 }, { "epoch": 0.8, "learning_rate": 2.083552064102172e-06, "loss": 23.5201, "step": 64810 }, { "epoch": 0.8, "learning_rate": 2.0811190587788787e-06, "loss": 23.2618, "step": 64820 }, { "epoch": 0.8, "learning_rate": 2.0786873098264193e-06, "loss": 23.499, "step": 64830 }, { "epoch": 0.8, "learning_rate": 2.076256817630604e-06, "loss": 23.7232, "step": 64840 }, { "epoch": 0.8, "learning_rate": 2.073827582577038e-06, "loss": 24.1099, "step": 64850 }, { "epoch": 0.8, "learning_rate": 2.07139960505114e-06, "loss": 23.9278, "step": 64860 }, { "epoch": 0.8, "learning_rate": 2.0689728854381153e-06, "loss": 24.4678, "step": 64870 }, { "epoch": 0.8, "learning_rate": 2.0665474241229753e-06, "loss": 23.9939, "step": 64880 }, { "epoch": 0.8, "learning_rate": 2.064123221490536e-06, "loss": 24.1675, "step": 64890 }, { "epoch": 0.8, "learning_rate": 2.0617002779254057e-06, "loss": 23.6646, "step": 64900 }, { "epoch": 0.8, "learning_rate": 2.0592785938119964e-06, "loss": 23.7794, "step": 64910 }, { "epoch": 0.8, "learning_rate": 2.0568581695345247e-06, "loss": 24.0555, "step": 64920 }, { "epoch": 0.8, "learning_rate": 2.0544390054770012e-06, "loss": 24.0962, "step": 64930 }, { "epoch": 0.8, "learning_rate": 2.052021102023235e-06, "loss": 23.8469, "step": 64940 }, { "epoch": 0.8, "learning_rate": 2.049604459556844e-06, "loss": 24.4543, "step": 64950 }, { "epoch": 0.8, "learning_rate": 2.0471890784612404e-06, "loss": 23.8133, "step": 64960 }, { "epoch": 0.8, "learning_rate": 2.044774959119632e-06, "loss": 24.3086, "step": 64970 }, { "epoch": 0.8, "learning_rate": 2.0423621019150363e-06, "loss": 23.8246, "step": 64980 }, { "epoch": 0.8, "learning_rate": 2.039950507230264e-06, "loss": 23.6242, "step": 64990 }, { "epoch": 0.8, "learning_rate": 2.037540175447923e-06, "loss": 23.5325, "step": 65000 }, { "epoch": 0.8, "learning_rate": 2.035131106950431e-06, "loss": 23.7899, "step": 65010 }, { "epoch": 0.8, "learning_rate": 2.0327233021199967e-06, "loss": 24.2172, "step": 65020 }, { "epoch": 0.8, "learning_rate": 2.0303167613386264e-06, "loss": 23.1956, "step": 65030 }, { "epoch": 0.8, "learning_rate": 2.0279114849881366e-06, "loss": 23.9215, "step": 65040 }, { "epoch": 0.8, "learning_rate": 2.0255074734501335e-06, "loss": 23.9109, "step": 65050 }, { "epoch": 0.8, "learning_rate": 2.0231047271060257e-06, "loss": 23.5967, "step": 65060 }, { "epoch": 0.8, "learning_rate": 2.0207032463370214e-06, "loss": 23.6067, "step": 65070 }, { "epoch": 0.8, "learning_rate": 2.018303031524125e-06, "loss": 24.3036, "step": 65080 }, { "epoch": 0.8, "learning_rate": 2.0159040830481493e-06, "loss": 24.2883, "step": 65090 }, { "epoch": 0.8, "learning_rate": 2.0135064012896943e-06, "loss": 23.9591, "step": 65100 }, { "epoch": 0.8, "learning_rate": 2.011109986629165e-06, "loss": 23.7505, "step": 65110 }, { "epoch": 0.8, "learning_rate": 2.008714839446768e-06, "loss": 23.924, "step": 65120 }, { "epoch": 0.8, "learning_rate": 2.006320960122503e-06, "loss": 23.7405, "step": 65130 }, { "epoch": 0.8, "learning_rate": 2.0039283490361694e-06, "loss": 23.959, "step": 65140 }, { "epoch": 0.8, "learning_rate": 2.0015370065673723e-06, "loss": 23.7017, "step": 65150 }, { "epoch": 0.8, "learning_rate": 1.999146933095507e-06, "loss": 23.9436, "step": 65160 }, { "epoch": 0.8, "learning_rate": 1.9967581289997693e-06, "loss": 23.8455, "step": 65170 }, { "epoch": 0.8, "learning_rate": 1.9943705946591595e-06, "loss": 24.2416, "step": 65180 }, { "epoch": 0.8, "learning_rate": 1.99198433045247e-06, "loss": 23.9012, "step": 65190 }, { "epoch": 0.8, "learning_rate": 1.989599336758291e-06, "loss": 23.3068, "step": 65200 }, { "epoch": 0.8, "learning_rate": 1.987215613955019e-06, "loss": 23.9758, "step": 65210 }, { "epoch": 0.8, "learning_rate": 1.984833162420842e-06, "loss": 24.3031, "step": 65220 }, { "epoch": 0.8, "learning_rate": 1.9824519825337427e-06, "loss": 23.9039, "step": 65230 }, { "epoch": 0.8, "learning_rate": 1.9800720746715163e-06, "loss": 23.3627, "step": 65240 }, { "epoch": 0.8, "learning_rate": 1.977693439211743e-06, "loss": 23.6677, "step": 65250 }, { "epoch": 0.8, "learning_rate": 1.9753160765318023e-06, "loss": 23.5157, "step": 65260 }, { "epoch": 0.8, "learning_rate": 1.9729399870088817e-06, "loss": 23.7113, "step": 65270 }, { "epoch": 0.8, "learning_rate": 1.970565171019956e-06, "loss": 24.0175, "step": 65280 }, { "epoch": 0.8, "learning_rate": 1.9681916289418013e-06, "loss": 23.6515, "step": 65290 }, { "epoch": 0.8, "learning_rate": 1.9658193611509948e-06, "loss": 23.4361, "step": 65300 }, { "epoch": 0.8, "learning_rate": 1.9634483680239027e-06, "loss": 23.5688, "step": 65310 }, { "epoch": 0.8, "learning_rate": 1.961078649936703e-06, "loss": 23.7515, "step": 65320 }, { "epoch": 0.8, "learning_rate": 1.958710207265361e-06, "loss": 24.0765, "step": 65330 }, { "epoch": 0.8, "learning_rate": 1.9563430403856377e-06, "loss": 24.0033, "step": 65340 }, { "epoch": 0.8, "learning_rate": 1.9539771496731018e-06, "loss": 23.9059, "step": 65350 }, { "epoch": 0.8, "learning_rate": 1.9516125355031134e-06, "loss": 23.6331, "step": 65360 }, { "epoch": 0.8, "learning_rate": 1.9492491982508254e-06, "loss": 23.367, "step": 65370 }, { "epoch": 0.8, "learning_rate": 1.9468871382912015e-06, "loss": 23.7552, "step": 65380 }, { "epoch": 0.8, "learning_rate": 1.944526355998988e-06, "loss": 23.4687, "step": 65390 }, { "epoch": 0.8, "learning_rate": 1.9421668517487356e-06, "loss": 23.6096, "step": 65400 }, { "epoch": 0.8, "learning_rate": 1.9398086259147963e-06, "loss": 23.644, "step": 65410 }, { "epoch": 0.8, "learning_rate": 1.9374516788713105e-06, "loss": 23.8636, "step": 65420 }, { "epoch": 0.8, "learning_rate": 1.935096010992219e-06, "loss": 23.9549, "step": 65430 }, { "epoch": 0.8, "learning_rate": 1.932741622651265e-06, "loss": 23.2876, "step": 65440 }, { "epoch": 0.8, "learning_rate": 1.930388514221979e-06, "loss": 23.8387, "step": 65450 }, { "epoch": 0.81, "learning_rate": 1.9280366860777e-06, "loss": 23.7179, "step": 65460 }, { "epoch": 0.81, "learning_rate": 1.9256861385915525e-06, "loss": 23.4986, "step": 65470 }, { "epoch": 0.81, "learning_rate": 1.923336872136462e-06, "loss": 23.799, "step": 65480 }, { "epoch": 0.81, "learning_rate": 1.920988887085157e-06, "loss": 23.735, "step": 65490 }, { "epoch": 0.81, "learning_rate": 1.918642183810153e-06, "loss": 23.7386, "step": 65500 }, { "epoch": 0.81, "learning_rate": 1.9162967626837683e-06, "loss": 23.8043, "step": 65510 }, { "epoch": 0.81, "learning_rate": 1.913952624078115e-06, "loss": 23.5261, "step": 65520 }, { "epoch": 0.81, "learning_rate": 1.9116097683651037e-06, "loss": 23.7806, "step": 65530 }, { "epoch": 0.81, "learning_rate": 1.909268195916436e-06, "loss": 23.7599, "step": 65540 }, { "epoch": 0.81, "learning_rate": 1.9069279071036207e-06, "loss": 24.9182, "step": 65550 }, { "epoch": 0.81, "learning_rate": 1.9045889022979524e-06, "loss": 24.0242, "step": 65560 }, { "epoch": 0.81, "learning_rate": 1.902251181870527e-06, "loss": 24.075, "step": 65570 }, { "epoch": 0.81, "learning_rate": 1.8999147461922374e-06, "loss": 24.4836, "step": 65580 }, { "epoch": 0.81, "learning_rate": 1.8975795956337672e-06, "loss": 23.6976, "step": 65590 }, { "epoch": 0.81, "learning_rate": 1.8952457305656058e-06, "loss": 23.8807, "step": 65600 }, { "epoch": 0.81, "learning_rate": 1.892913151358029e-06, "loss": 24.3157, "step": 65610 }, { "epoch": 0.81, "learning_rate": 1.8905818583811108e-06, "loss": 23.2133, "step": 65620 }, { "epoch": 0.81, "learning_rate": 1.8882518520047277e-06, "loss": 24.4779, "step": 65630 }, { "epoch": 0.81, "learning_rate": 1.8859231325985438e-06, "loss": 23.7411, "step": 65640 }, { "epoch": 0.81, "learning_rate": 1.8835957005320194e-06, "loss": 23.1125, "step": 65650 }, { "epoch": 0.81, "learning_rate": 1.88126955617442e-06, "loss": 23.7761, "step": 65660 }, { "epoch": 0.81, "learning_rate": 1.878944699894798e-06, "loss": 23.7526, "step": 65670 }, { "epoch": 0.81, "learning_rate": 1.8766211320619987e-06, "loss": 23.8691, "step": 65680 }, { "epoch": 0.81, "learning_rate": 1.8742988530446759e-06, "loss": 24.3718, "step": 65690 }, { "epoch": 0.81, "learning_rate": 1.8719778632112662e-06, "loss": 23.6752, "step": 65700 }, { "epoch": 0.81, "learning_rate": 1.8696581629300048e-06, "loss": 23.7148, "step": 65710 }, { "epoch": 0.81, "learning_rate": 1.8673397525689297e-06, "loss": 24.1635, "step": 65720 }, { "epoch": 0.81, "learning_rate": 1.8650226324958654e-06, "loss": 23.7848, "step": 65730 }, { "epoch": 0.81, "learning_rate": 1.8627068030784357e-06, "loss": 23.1641, "step": 65740 }, { "epoch": 0.81, "learning_rate": 1.8603922646840567e-06, "loss": 24.2633, "step": 65750 }, { "epoch": 0.81, "learning_rate": 1.85807901767994e-06, "loss": 23.6227, "step": 65760 }, { "epoch": 0.81, "learning_rate": 1.8557670624331003e-06, "loss": 23.6612, "step": 65770 }, { "epoch": 0.81, "learning_rate": 1.8534563993103383e-06, "loss": 23.3395, "step": 65780 }, { "epoch": 0.81, "learning_rate": 1.8511470286782484e-06, "loss": 23.656, "step": 65790 }, { "epoch": 0.81, "learning_rate": 1.84883895090323e-06, "loss": 23.1252, "step": 65800 }, { "epoch": 0.81, "learning_rate": 1.846532166351469e-06, "loss": 24.5634, "step": 65810 }, { "epoch": 0.81, "learning_rate": 1.844226675388946e-06, "loss": 23.922, "step": 65820 }, { "epoch": 0.81, "learning_rate": 1.8419224783814438e-06, "loss": 23.7203, "step": 65830 }, { "epoch": 0.81, "learning_rate": 1.8396195756945335e-06, "loss": 23.744, "step": 65840 }, { "epoch": 0.81, "learning_rate": 1.8373179676935781e-06, "loss": 23.3703, "step": 65850 }, { "epoch": 0.81, "learning_rate": 1.8350176547437448e-06, "loss": 23.4897, "step": 65860 }, { "epoch": 0.81, "learning_rate": 1.8327186372099892e-06, "loss": 23.7179, "step": 65870 }, { "epoch": 0.81, "learning_rate": 1.8304209154570573e-06, "loss": 24.3291, "step": 65880 }, { "epoch": 0.81, "learning_rate": 1.828124489849501e-06, "loss": 23.4924, "step": 65890 }, { "epoch": 0.81, "learning_rate": 1.8258293607516563e-06, "loss": 23.8912, "step": 65900 }, { "epoch": 0.81, "learning_rate": 1.8235355285276557e-06, "loss": 23.8072, "step": 65910 }, { "epoch": 0.81, "learning_rate": 1.8212429935414322e-06, "loss": 23.7646, "step": 65920 }, { "epoch": 0.81, "learning_rate": 1.8189517561567049e-06, "loss": 23.9379, "step": 65930 }, { "epoch": 0.81, "learning_rate": 1.8166618167369886e-06, "loss": 23.9795, "step": 65940 }, { "epoch": 0.81, "learning_rate": 1.8143731756455985e-06, "loss": 23.819, "step": 65950 }, { "epoch": 0.81, "learning_rate": 1.812085833245636e-06, "loss": 23.5317, "step": 65960 }, { "epoch": 0.81, "learning_rate": 1.8097997899000009e-06, "loss": 23.8743, "step": 65970 }, { "epoch": 0.81, "learning_rate": 1.8075150459713853e-06, "loss": 24.2678, "step": 65980 }, { "epoch": 0.81, "learning_rate": 1.8052316018222725e-06, "loss": 24.1146, "step": 65990 }, { "epoch": 0.81, "learning_rate": 1.8029494578149476e-06, "loss": 24.1095, "step": 66000 }, { "epoch": 0.81, "learning_rate": 1.8006686143114826e-06, "loss": 23.5975, "step": 66010 }, { "epoch": 0.81, "learning_rate": 1.7983890716737407e-06, "loss": 24.1385, "step": 66020 }, { "epoch": 0.81, "learning_rate": 1.7961108302633901e-06, "loss": 24.1151, "step": 66030 }, { "epoch": 0.81, "learning_rate": 1.7938338904418816e-06, "loss": 23.8259, "step": 66040 }, { "epoch": 0.81, "learning_rate": 1.7915582525704622e-06, "loss": 23.6927, "step": 66050 }, { "epoch": 0.81, "learning_rate": 1.7892839170101762e-06, "loss": 23.8827, "step": 66060 }, { "epoch": 0.81, "learning_rate": 1.787010884121858e-06, "loss": 23.5155, "step": 66070 }, { "epoch": 0.81, "learning_rate": 1.784739154266133e-06, "loss": 23.9979, "step": 66080 }, { "epoch": 0.81, "learning_rate": 1.7824687278034269e-06, "loss": 24.2668, "step": 66090 }, { "epoch": 0.81, "learning_rate": 1.780199605093952e-06, "loss": 23.691, "step": 66100 }, { "epoch": 0.81, "learning_rate": 1.777931786497714e-06, "loss": 23.6883, "step": 66110 }, { "epoch": 0.81, "learning_rate": 1.7756652723745205e-06, "loss": 23.9123, "step": 66120 }, { "epoch": 0.81, "learning_rate": 1.7734000630839598e-06, "loss": 23.9075, "step": 66130 }, { "epoch": 0.81, "learning_rate": 1.7711361589854192e-06, "loss": 23.8548, "step": 66140 }, { "epoch": 0.81, "learning_rate": 1.7688735604380813e-06, "loss": 24.1044, "step": 66150 }, { "epoch": 0.81, "learning_rate": 1.766612267800919e-06, "loss": 23.7435, "step": 66160 }, { "epoch": 0.81, "learning_rate": 1.764352281432693e-06, "loss": 24.0876, "step": 66170 }, { "epoch": 0.81, "learning_rate": 1.7620936016919688e-06, "loss": 23.2557, "step": 66180 }, { "epoch": 0.81, "learning_rate": 1.7598362289370929e-06, "loss": 23.6262, "step": 66190 }, { "epoch": 0.81, "learning_rate": 1.7575801635262102e-06, "loss": 23.6047, "step": 66200 }, { "epoch": 0.81, "learning_rate": 1.7553254058172564e-06, "loss": 23.772, "step": 66210 }, { "epoch": 0.81, "learning_rate": 1.7530719561679587e-06, "loss": 23.1688, "step": 66220 }, { "epoch": 0.81, "learning_rate": 1.7508198149358424e-06, "loss": 24.3341, "step": 66230 }, { "epoch": 0.81, "learning_rate": 1.7485689824782194e-06, "loss": 24.1145, "step": 66240 }, { "epoch": 0.81, "learning_rate": 1.7463194591521926e-06, "loss": 23.666, "step": 66250 }, { "epoch": 0.81, "learning_rate": 1.7440712453146657e-06, "loss": 23.8133, "step": 66260 }, { "epoch": 0.82, "learning_rate": 1.7418243413223268e-06, "loss": 23.7743, "step": 66270 }, { "epoch": 0.82, "learning_rate": 1.739578747531656e-06, "loss": 23.8604, "step": 66280 }, { "epoch": 0.82, "learning_rate": 1.7373344642989342e-06, "loss": 23.6072, "step": 66290 }, { "epoch": 0.82, "learning_rate": 1.7350914919802243e-06, "loss": 23.5482, "step": 66300 }, { "epoch": 0.82, "learning_rate": 1.7328498309313835e-06, "loss": 23.8184, "step": 66310 }, { "epoch": 0.82, "learning_rate": 1.7306094815080677e-06, "loss": 24.232, "step": 66320 }, { "epoch": 0.82, "learning_rate": 1.7283704440657179e-06, "loss": 23.6268, "step": 66330 }, { "epoch": 0.82, "learning_rate": 1.7261327189595644e-06, "loss": 23.6056, "step": 66340 }, { "epoch": 0.82, "learning_rate": 1.7238963065446412e-06, "loss": 23.8517, "step": 66350 }, { "epoch": 0.82, "learning_rate": 1.7216612071757633e-06, "loss": 24.2022, "step": 66360 }, { "epoch": 0.82, "learning_rate": 1.7194274212075369e-06, "loss": 23.8127, "step": 66370 }, { "epoch": 0.82, "learning_rate": 1.7171949489943685e-06, "loss": 23.2036, "step": 66380 }, { "epoch": 0.82, "learning_rate": 1.7149637908904504e-06, "loss": 24.2009, "step": 66390 }, { "epoch": 0.82, "learning_rate": 1.712733947249764e-06, "loss": 23.3739, "step": 66400 }, { "epoch": 0.82, "learning_rate": 1.7105054184260905e-06, "loss": 24.3555, "step": 66410 }, { "epoch": 0.82, "learning_rate": 1.7082782047729929e-06, "loss": 23.7489, "step": 66420 }, { "epoch": 0.82, "learning_rate": 1.7060523066438307e-06, "loss": 23.6948, "step": 66430 }, { "epoch": 0.82, "learning_rate": 1.703827724391759e-06, "loss": 23.5719, "step": 66440 }, { "epoch": 0.82, "learning_rate": 1.701604458369711e-06, "loss": 23.9215, "step": 66450 }, { "epoch": 0.82, "learning_rate": 1.699382508930425e-06, "loss": 23.9745, "step": 66460 }, { "epoch": 0.82, "learning_rate": 1.697161876426423e-06, "loss": 23.7044, "step": 66470 }, { "epoch": 0.82, "learning_rate": 1.6949425612100167e-06, "loss": 24.2301, "step": 66480 }, { "epoch": 0.82, "learning_rate": 1.6927245636333178e-06, "loss": 23.6779, "step": 66490 }, { "epoch": 0.82, "learning_rate": 1.6905078840482192e-06, "loss": 24.4587, "step": 66500 }, { "epoch": 0.82, "learning_rate": 1.6882925228064063e-06, "loss": 23.8953, "step": 66510 }, { "epoch": 0.82, "learning_rate": 1.6860784802593622e-06, "loss": 23.3887, "step": 66520 }, { "epoch": 0.82, "learning_rate": 1.683865756758355e-06, "loss": 23.6772, "step": 66530 }, { "epoch": 0.82, "learning_rate": 1.6816543526544405e-06, "loss": 23.4991, "step": 66540 }, { "epoch": 0.82, "learning_rate": 1.6794442682984745e-06, "loss": 24.4444, "step": 66550 }, { "epoch": 0.82, "learning_rate": 1.677235504041096e-06, "loss": 23.8045, "step": 66560 }, { "epoch": 0.82, "learning_rate": 1.675028060232734e-06, "loss": 23.9609, "step": 66570 }, { "epoch": 0.82, "learning_rate": 1.6728219372236154e-06, "loss": 23.971, "step": 66580 }, { "epoch": 0.82, "learning_rate": 1.6706171353637513e-06, "loss": 24.4252, "step": 66590 }, { "epoch": 0.82, "learning_rate": 1.6684136550029418e-06, "loss": 23.8299, "step": 66600 }, { "epoch": 0.82, "learning_rate": 1.6662114964907838e-06, "loss": 24.0118, "step": 66610 }, { "epoch": 0.82, "learning_rate": 1.6640106601766614e-06, "loss": 24.3826, "step": 66620 }, { "epoch": 0.82, "learning_rate": 1.6618111464097443e-06, "loss": 23.8988, "step": 66630 }, { "epoch": 0.82, "learning_rate": 1.6596129555390016e-06, "loss": 24.2144, "step": 66640 }, { "epoch": 0.82, "learning_rate": 1.6574160879131852e-06, "loss": 24.1645, "step": 66650 }, { "epoch": 0.82, "learning_rate": 1.6552205438808356e-06, "loss": 23.9005, "step": 66660 }, { "epoch": 0.82, "learning_rate": 1.6530263237902965e-06, "loss": 23.8754, "step": 66670 }, { "epoch": 0.82, "learning_rate": 1.6508334279896821e-06, "loss": 24.113, "step": 66680 }, { "epoch": 0.82, "learning_rate": 1.6486418568269125e-06, "loss": 24.3674, "step": 66690 }, { "epoch": 0.82, "learning_rate": 1.6464516106496898e-06, "loss": 23.7737, "step": 66700 }, { "epoch": 0.82, "learning_rate": 1.6442626898055058e-06, "loss": 23.1789, "step": 66710 }, { "epoch": 0.82, "learning_rate": 1.6420750946416486e-06, "loss": 23.4078, "step": 66720 }, { "epoch": 0.82, "learning_rate": 1.6398888255051882e-06, "loss": 23.7202, "step": 66730 }, { "epoch": 0.82, "learning_rate": 1.6377038827429848e-06, "loss": 23.5352, "step": 66740 }, { "epoch": 0.82, "learning_rate": 1.6355202667016968e-06, "loss": 23.7726, "step": 66750 }, { "epoch": 0.82, "learning_rate": 1.6333379777277636e-06, "loss": 23.888, "step": 66760 }, { "epoch": 0.82, "learning_rate": 1.6311570161674128e-06, "loss": 24.189, "step": 66770 }, { "epoch": 0.82, "learning_rate": 1.6289773823666699e-06, "loss": 24.0706, "step": 66780 }, { "epoch": 0.82, "learning_rate": 1.6267990766713438e-06, "loss": 23.615, "step": 66790 }, { "epoch": 0.82, "learning_rate": 1.62462209942703e-06, "loss": 23.3568, "step": 66800 }, { "epoch": 0.82, "learning_rate": 1.6224464509791238e-06, "loss": 24.7189, "step": 66810 }, { "epoch": 0.82, "learning_rate": 1.6202721316727987e-06, "loss": 24.0707, "step": 66820 }, { "epoch": 0.82, "learning_rate": 1.618099141853019e-06, "loss": 24.2899, "step": 66830 }, { "epoch": 0.82, "learning_rate": 1.6159274818645465e-06, "loss": 23.3898, "step": 66840 }, { "epoch": 0.82, "learning_rate": 1.6137571520519213e-06, "loss": 24.3881, "step": 66850 }, { "epoch": 0.82, "learning_rate": 1.6115881527594813e-06, "loss": 24.2098, "step": 66860 }, { "epoch": 0.82, "learning_rate": 1.6094204843313476e-06, "loss": 23.5385, "step": 66870 }, { "epoch": 0.82, "learning_rate": 1.607254147111429e-06, "loss": 23.8568, "step": 66880 }, { "epoch": 0.82, "learning_rate": 1.605089141443431e-06, "loss": 24.288, "step": 66890 }, { "epoch": 0.82, "learning_rate": 1.6029254676708428e-06, "loss": 24.1463, "step": 66900 }, { "epoch": 0.82, "learning_rate": 1.600763126136935e-06, "loss": 23.5332, "step": 66910 }, { "epoch": 0.82, "learning_rate": 1.5986021171847821e-06, "loss": 23.9869, "step": 66920 }, { "epoch": 0.82, "learning_rate": 1.5964424411572343e-06, "loss": 23.5477, "step": 66930 }, { "epoch": 0.82, "learning_rate": 1.5942840983969354e-06, "loss": 23.6155, "step": 66940 }, { "epoch": 0.82, "learning_rate": 1.5921270892463215e-06, "loss": 23.4327, "step": 66950 }, { "epoch": 0.82, "learning_rate": 1.5899714140476108e-06, "loss": 23.2997, "step": 66960 }, { "epoch": 0.82, "learning_rate": 1.5878170731428088e-06, "loss": 24.1787, "step": 66970 }, { "epoch": 0.82, "learning_rate": 1.585664066873719e-06, "loss": 24.1802, "step": 66980 }, { "epoch": 0.82, "learning_rate": 1.5835123955819232e-06, "loss": 23.2943, "step": 66990 }, { "epoch": 0.82, "learning_rate": 1.581362059608793e-06, "loss": 24.0259, "step": 67000 }, { "epoch": 0.82, "learning_rate": 1.5792130592954945e-06, "loss": 23.7982, "step": 67010 }, { "epoch": 0.82, "learning_rate": 1.577065394982974e-06, "loss": 23.988, "step": 67020 }, { "epoch": 0.82, "learning_rate": 1.574919067011973e-06, "loss": 23.7669, "step": 67030 }, { "epoch": 0.82, "learning_rate": 1.572774075723016e-06, "loss": 23.352, "step": 67040 }, { "epoch": 0.82, "learning_rate": 1.570630421456414e-06, "loss": 23.8214, "step": 67050 }, { "epoch": 0.82, "learning_rate": 1.5684881045522726e-06, "loss": 24.2027, "step": 67060 }, { "epoch": 0.82, "learning_rate": 1.5663471253504793e-06, "loss": 23.8761, "step": 67070 }, { "epoch": 0.82, "learning_rate": 1.5642074841907108e-06, "loss": 24.1376, "step": 67080 }, { "epoch": 0.83, "learning_rate": 1.5620691814124334e-06, "loss": 24.1244, "step": 67090 }, { "epoch": 0.83, "learning_rate": 1.5599322173549002e-06, "loss": 24.1663, "step": 67100 }, { "epoch": 0.83, "learning_rate": 1.5577965923571492e-06, "loss": 23.9427, "step": 67110 }, { "epoch": 0.83, "learning_rate": 1.55566230675801e-06, "loss": 23.7527, "step": 67120 }, { "epoch": 0.83, "learning_rate": 1.5535293608961012e-06, "loss": 23.5552, "step": 67130 }, { "epoch": 0.83, "learning_rate": 1.5513977551098169e-06, "loss": 23.62, "step": 67140 }, { "epoch": 0.83, "learning_rate": 1.5492674897373539e-06, "loss": 23.703, "step": 67150 }, { "epoch": 0.83, "learning_rate": 1.547138565116686e-06, "loss": 23.7524, "step": 67160 }, { "epoch": 0.83, "learning_rate": 1.5450109815855818e-06, "loss": 24.2607, "step": 67170 }, { "epoch": 0.83, "learning_rate": 1.5428847394815915e-06, "loss": 24.4054, "step": 67180 }, { "epoch": 0.83, "learning_rate": 1.540759839142052e-06, "loss": 23.1429, "step": 67190 }, { "epoch": 0.83, "learning_rate": 1.5386362809040934e-06, "loss": 24.2835, "step": 67200 }, { "epoch": 0.83, "learning_rate": 1.5365140651046273e-06, "loss": 23.4886, "step": 67210 }, { "epoch": 0.83, "learning_rate": 1.5343931920803512e-06, "loss": 23.835, "step": 67220 }, { "epoch": 0.83, "learning_rate": 1.5322736621677581e-06, "loss": 23.784, "step": 67230 }, { "epoch": 0.83, "learning_rate": 1.530155475703119e-06, "loss": 24.1585, "step": 67240 }, { "epoch": 0.83, "learning_rate": 1.5280386330224939e-06, "loss": 23.733, "step": 67250 }, { "epoch": 0.83, "learning_rate": 1.525923134461733e-06, "loss": 24.0285, "step": 67260 }, { "epoch": 0.83, "learning_rate": 1.5238089803564716e-06, "loss": 24.0923, "step": 67270 }, { "epoch": 0.83, "learning_rate": 1.5216961710421252e-06, "loss": 24.2002, "step": 67280 }, { "epoch": 0.83, "learning_rate": 1.5195847068539105e-06, "loss": 23.616, "step": 67290 }, { "epoch": 0.83, "learning_rate": 1.517474588126816e-06, "loss": 23.3973, "step": 67300 }, { "epoch": 0.83, "learning_rate": 1.515365815195623e-06, "loss": 24.3225, "step": 67310 }, { "epoch": 0.83, "learning_rate": 1.5132583883949014e-06, "loss": 23.5781, "step": 67320 }, { "epoch": 0.83, "learning_rate": 1.5111523080590063e-06, "loss": 24.3876, "step": 67330 }, { "epoch": 0.83, "learning_rate": 1.5090475745220722e-06, "loss": 24.5137, "step": 67340 }, { "epoch": 0.83, "learning_rate": 1.5069441881180325e-06, "loss": 23.7844, "step": 67350 }, { "epoch": 0.83, "learning_rate": 1.5048421491805975e-06, "loss": 24.1302, "step": 67360 }, { "epoch": 0.83, "learning_rate": 1.5027414580432652e-06, "loss": 23.4587, "step": 67370 }, { "epoch": 0.83, "learning_rate": 1.5006421150393236e-06, "loss": 23.6152, "step": 67380 }, { "epoch": 0.83, "learning_rate": 1.4985441205018392e-06, "loss": 23.5449, "step": 67390 }, { "epoch": 0.83, "learning_rate": 1.4964474747636748e-06, "loss": 24.1164, "step": 67400 }, { "epoch": 0.83, "learning_rate": 1.4943521781574732e-06, "loss": 24.0613, "step": 67410 }, { "epoch": 0.83, "learning_rate": 1.4922582310156596e-06, "loss": 24.3945, "step": 67420 }, { "epoch": 0.83, "learning_rate": 1.490165633670455e-06, "loss": 23.9314, "step": 67430 }, { "epoch": 0.83, "learning_rate": 1.4880743864538583e-06, "loss": 24.0987, "step": 67440 }, { "epoch": 0.83, "learning_rate": 1.4859844896976538e-06, "loss": 23.7745, "step": 67450 }, { "epoch": 0.83, "learning_rate": 1.4838959437334188e-06, "loss": 23.376, "step": 67460 }, { "epoch": 0.83, "learning_rate": 1.4818087488925103e-06, "loss": 23.0028, "step": 67470 }, { "epoch": 0.83, "learning_rate": 1.4797229055060702e-06, "loss": 23.3482, "step": 67480 }, { "epoch": 0.83, "learning_rate": 1.4776384139050316e-06, "loss": 23.6836, "step": 67490 }, { "epoch": 0.83, "learning_rate": 1.4755552744201075e-06, "loss": 23.5352, "step": 67500 }, { "epoch": 0.83, "learning_rate": 1.473473487381798e-06, "loss": 24.1348, "step": 67510 }, { "epoch": 0.83, "learning_rate": 1.4713930531203912e-06, "loss": 24.2046, "step": 67520 }, { "epoch": 0.83, "learning_rate": 1.4693139719659587e-06, "loss": 24.03, "step": 67530 }, { "epoch": 0.83, "learning_rate": 1.4672362442483533e-06, "loss": 24.2433, "step": 67540 }, { "epoch": 0.83, "learning_rate": 1.4651598702972235e-06, "loss": 24.0431, "step": 67550 }, { "epoch": 0.83, "learning_rate": 1.4630848504419925e-06, "loss": 24.26, "step": 67560 }, { "epoch": 0.83, "learning_rate": 1.4610111850118725e-06, "loss": 23.4724, "step": 67570 }, { "epoch": 0.83, "learning_rate": 1.4589388743358635e-06, "loss": 24.0985, "step": 67580 }, { "epoch": 0.83, "learning_rate": 1.4568679187427459e-06, "loss": 23.9449, "step": 67590 }, { "epoch": 0.83, "learning_rate": 1.4547983185610893e-06, "loss": 24.0143, "step": 67600 }, { "epoch": 0.83, "learning_rate": 1.4527300741192452e-06, "loss": 23.8156, "step": 67610 }, { "epoch": 0.83, "learning_rate": 1.450663185745349e-06, "loss": 23.9326, "step": 67620 }, { "epoch": 0.83, "learning_rate": 1.4485976537673274e-06, "loss": 23.5078, "step": 67630 }, { "epoch": 0.83, "learning_rate": 1.446533478512886e-06, "loss": 23.1778, "step": 67640 }, { "epoch": 0.83, "learning_rate": 1.4444706603095126e-06, "loss": 24.2013, "step": 67650 }, { "epoch": 0.83, "learning_rate": 1.4424091994844914e-06, "loss": 23.6989, "step": 67660 }, { "epoch": 0.83, "learning_rate": 1.440349096364878e-06, "loss": 23.7966, "step": 67670 }, { "epoch": 0.83, "learning_rate": 1.4382903512775192e-06, "loss": 23.8493, "step": 67680 }, { "epoch": 0.83, "learning_rate": 1.436232964549047e-06, "loss": 23.3459, "step": 67690 }, { "epoch": 0.83, "learning_rate": 1.4341769365058766e-06, "loss": 22.8966, "step": 67700 }, { "epoch": 0.83, "learning_rate": 1.432122267474203e-06, "loss": 23.5954, "step": 67710 }, { "epoch": 0.83, "learning_rate": 1.4300689577800164e-06, "loss": 24.221, "step": 67720 }, { "epoch": 0.83, "learning_rate": 1.4280170077490807e-06, "loss": 23.6162, "step": 67730 }, { "epoch": 0.83, "learning_rate": 1.4259664177069477e-06, "loss": 23.407, "step": 67740 }, { "epoch": 0.83, "learning_rate": 1.423917187978957e-06, "loss": 23.3224, "step": 67750 }, { "epoch": 0.83, "learning_rate": 1.4218693188902278e-06, "loss": 23.3686, "step": 67760 }, { "epoch": 0.83, "learning_rate": 1.419822810765662e-06, "loss": 23.4774, "step": 67770 }, { "epoch": 0.83, "learning_rate": 1.4177776639299545e-06, "loss": 23.6625, "step": 67780 }, { "epoch": 0.83, "learning_rate": 1.4157338787075748e-06, "loss": 23.8622, "step": 67790 }, { "epoch": 0.83, "learning_rate": 1.4136914554227789e-06, "loss": 23.6096, "step": 67800 }, { "epoch": 0.83, "learning_rate": 1.4116503943996096e-06, "loss": 23.9611, "step": 67810 }, { "epoch": 0.83, "learning_rate": 1.4096106959618927e-06, "loss": 23.6624, "step": 67820 }, { "epoch": 0.83, "learning_rate": 1.4075723604332336e-06, "loss": 23.7766, "step": 67830 }, { "epoch": 0.83, "learning_rate": 1.405535388137027e-06, "loss": 23.3531, "step": 67840 }, { "epoch": 0.83, "learning_rate": 1.4034997793964457e-06, "loss": 23.8207, "step": 67850 }, { "epoch": 0.83, "learning_rate": 1.401465534534454e-06, "loss": 23.4591, "step": 67860 }, { "epoch": 0.83, "learning_rate": 1.3994326538737923e-06, "loss": 23.5303, "step": 67870 }, { "epoch": 0.83, "learning_rate": 1.3974011377369868e-06, "loss": 23.8702, "step": 67880 }, { "epoch": 0.83, "learning_rate": 1.3953709864463516e-06, "loss": 23.2887, "step": 67890 }, { "epoch": 0.84, "learning_rate": 1.3933422003239772e-06, "loss": 23.6713, "step": 67900 }, { "epoch": 0.84, "learning_rate": 1.3913147796917403e-06, "loss": 23.9347, "step": 67910 }, { "epoch": 0.84, "learning_rate": 1.389288724871305e-06, "loss": 23.7145, "step": 67920 }, { "epoch": 0.84, "learning_rate": 1.3872640361841139e-06, "loss": 24.5399, "step": 67930 }, { "epoch": 0.84, "learning_rate": 1.3852407139513913e-06, "loss": 24.1947, "step": 67940 }, { "epoch": 0.84, "learning_rate": 1.3832187584941514e-06, "loss": 23.8054, "step": 67950 }, { "epoch": 0.84, "learning_rate": 1.3811981701331868e-06, "loss": 23.6567, "step": 67960 }, { "epoch": 0.84, "learning_rate": 1.3791789491890718e-06, "loss": 23.6433, "step": 67970 }, { "epoch": 0.84, "learning_rate": 1.3771610959821691e-06, "loss": 23.2643, "step": 67980 }, { "epoch": 0.84, "learning_rate": 1.3751446108326205e-06, "loss": 23.7971, "step": 67990 }, { "epoch": 0.84, "learning_rate": 1.373129494060349e-06, "loss": 23.9818, "step": 68000 }, { "epoch": 0.84, "learning_rate": 1.3711157459850677e-06, "loss": 23.9081, "step": 68010 }, { "epoch": 0.84, "learning_rate": 1.3691033669262666e-06, "loss": 25.0153, "step": 68020 }, { "epoch": 0.84, "learning_rate": 1.3670923572032158e-06, "loss": 24.2815, "step": 68030 }, { "epoch": 0.84, "learning_rate": 1.3650827171349778e-06, "loss": 23.8574, "step": 68040 }, { "epoch": 0.84, "learning_rate": 1.36307444704039e-06, "loss": 24.0057, "step": 68050 }, { "epoch": 0.84, "learning_rate": 1.3610675472380752e-06, "loss": 23.1765, "step": 68060 }, { "epoch": 0.84, "learning_rate": 1.3590620180464376e-06, "loss": 24.1023, "step": 68070 }, { "epoch": 0.84, "learning_rate": 1.3570578597836626e-06, "loss": 23.6886, "step": 68080 }, { "epoch": 0.84, "learning_rate": 1.3550550727677236e-06, "loss": 23.251, "step": 68090 }, { "epoch": 0.84, "learning_rate": 1.3530536573163722e-06, "loss": 23.2948, "step": 68100 }, { "epoch": 0.84, "learning_rate": 1.351053613747141e-06, "loss": 23.5455, "step": 68110 }, { "epoch": 0.84, "learning_rate": 1.3490549423773502e-06, "loss": 24.0155, "step": 68120 }, { "epoch": 0.84, "learning_rate": 1.3470576435240978e-06, "loss": 23.3761, "step": 68130 }, { "epoch": 0.84, "learning_rate": 1.3450617175042635e-06, "loss": 23.7727, "step": 68140 }, { "epoch": 0.84, "learning_rate": 1.343067164634515e-06, "loss": 23.4874, "step": 68150 }, { "epoch": 0.84, "learning_rate": 1.341073985231296e-06, "loss": 23.5858, "step": 68160 }, { "epoch": 0.84, "learning_rate": 1.3390821796108332e-06, "loss": 23.4373, "step": 68170 }, { "epoch": 0.84, "learning_rate": 1.3370917480891398e-06, "loss": 24.31, "step": 68180 }, { "epoch": 0.84, "learning_rate": 1.3351026909820076e-06, "loss": 23.6092, "step": 68190 }, { "epoch": 0.84, "learning_rate": 1.333115008605007e-06, "loss": 23.8453, "step": 68200 }, { "epoch": 0.84, "learning_rate": 1.3311287012734986e-06, "loss": 23.9967, "step": 68210 }, { "epoch": 0.84, "learning_rate": 1.3291437693026177e-06, "loss": 23.8437, "step": 68220 }, { "epoch": 0.84, "learning_rate": 1.3271602130072825e-06, "loss": 23.7753, "step": 68230 }, { "epoch": 0.84, "learning_rate": 1.325178032702198e-06, "loss": 24.096, "step": 68240 }, { "epoch": 0.84, "learning_rate": 1.3231972287018424e-06, "loss": 24.1824, "step": 68250 }, { "epoch": 0.84, "learning_rate": 1.3212178013204868e-06, "loss": 23.9426, "step": 68260 }, { "epoch": 0.84, "learning_rate": 1.3192397508721733e-06, "loss": 24.174, "step": 68270 }, { "epoch": 0.84, "learning_rate": 1.3172630776707273e-06, "loss": 24.2006, "step": 68280 }, { "epoch": 0.84, "learning_rate": 1.3152877820297671e-06, "loss": 22.9209, "step": 68290 }, { "epoch": 0.84, "learning_rate": 1.3133138642626741e-06, "loss": 23.7596, "step": 68300 }, { "epoch": 0.84, "learning_rate": 1.3113413246826211e-06, "loss": 23.9861, "step": 68310 }, { "epoch": 0.84, "learning_rate": 1.309370163602567e-06, "loss": 24.1577, "step": 68320 }, { "epoch": 0.84, "learning_rate": 1.3074003813352442e-06, "loss": 23.0721, "step": 68330 }, { "epoch": 0.84, "learning_rate": 1.3054319781931646e-06, "loss": 23.5848, "step": 68340 }, { "epoch": 0.84, "learning_rate": 1.3034649544886324e-06, "loss": 24.3321, "step": 68350 }, { "epoch": 0.84, "learning_rate": 1.3014993105337226e-06, "loss": 23.8062, "step": 68360 }, { "epoch": 0.84, "learning_rate": 1.2995350466402911e-06, "loss": 23.1166, "step": 68370 }, { "epoch": 0.84, "learning_rate": 1.2975721631199856e-06, "loss": 24.3624, "step": 68380 }, { "epoch": 0.84, "learning_rate": 1.2956106602842222e-06, "loss": 24.1172, "step": 68390 }, { "epoch": 0.84, "learning_rate": 1.2936505384442043e-06, "loss": 23.8657, "step": 68400 }, { "epoch": 0.84, "learning_rate": 1.2916917979109177e-06, "loss": 23.5427, "step": 68410 }, { "epoch": 0.84, "learning_rate": 1.2897344389951217e-06, "loss": 23.8225, "step": 68420 }, { "epoch": 0.84, "learning_rate": 1.2877784620073675e-06, "loss": 24.2556, "step": 68430 }, { "epoch": 0.84, "learning_rate": 1.2858238672579771e-06, "loss": 24.0628, "step": 68440 }, { "epoch": 0.84, "learning_rate": 1.2838706550570556e-06, "loss": 23.4693, "step": 68450 }, { "epoch": 0.84, "learning_rate": 1.2819188257144932e-06, "loss": 23.2248, "step": 68460 }, { "epoch": 0.84, "learning_rate": 1.2799683795399577e-06, "loss": 23.748, "step": 68470 }, { "epoch": 0.84, "learning_rate": 1.2780193168428934e-06, "loss": 23.6035, "step": 68480 }, { "epoch": 0.84, "learning_rate": 1.2760716379325334e-06, "loss": 23.5628, "step": 68490 }, { "epoch": 0.84, "learning_rate": 1.2741253431178847e-06, "loss": 24.0916, "step": 68500 }, { "epoch": 0.84, "learning_rate": 1.2721804327077359e-06, "loss": 24.055, "step": 68510 }, { "epoch": 0.84, "learning_rate": 1.2702369070106623e-06, "loss": 23.8356, "step": 68520 }, { "epoch": 0.84, "learning_rate": 1.2682947663350075e-06, "loss": 23.4652, "step": 68530 }, { "epoch": 0.84, "learning_rate": 1.2663540109889028e-06, "loss": 23.2577, "step": 68540 }, { "epoch": 0.84, "learning_rate": 1.264414641280264e-06, "loss": 23.4824, "step": 68550 }, { "epoch": 0.84, "learning_rate": 1.2624766575167779e-06, "loss": 23.5789, "step": 68560 }, { "epoch": 0.84, "learning_rate": 1.2605400600059137e-06, "loss": 24.1596, "step": 68570 }, { "epoch": 0.84, "learning_rate": 1.2586048490549286e-06, "loss": 24.0054, "step": 68580 }, { "epoch": 0.84, "learning_rate": 1.2566710249708468e-06, "loss": 23.7954, "step": 68590 }, { "epoch": 0.84, "learning_rate": 1.2547385880604867e-06, "loss": 23.9799, "step": 68600 }, { "epoch": 0.84, "learning_rate": 1.2528075386304338e-06, "loss": 23.5825, "step": 68610 }, { "epoch": 0.84, "learning_rate": 1.2508778769870589e-06, "loss": 23.2325, "step": 68620 }, { "epoch": 0.84, "learning_rate": 1.2489496034365168e-06, "loss": 23.721, "step": 68630 }, { "epoch": 0.84, "learning_rate": 1.247022718284735e-06, "loss": 23.7357, "step": 68640 }, { "epoch": 0.84, "learning_rate": 1.2450972218374224e-06, "loss": 23.8042, "step": 68650 }, { "epoch": 0.84, "learning_rate": 1.2431731144000724e-06, "loss": 23.4912, "step": 68660 }, { "epoch": 0.84, "learning_rate": 1.241250396277952e-06, "loss": 24.4489, "step": 68670 }, { "epoch": 0.84, "learning_rate": 1.2393290677761094e-06, "loss": 23.8495, "step": 68680 }, { "epoch": 0.84, "learning_rate": 1.237409129199376e-06, "loss": 24.676, "step": 68690 }, { "epoch": 0.84, "learning_rate": 1.2354905808523587e-06, "loss": 23.4869, "step": 68700 }, { "epoch": 0.85, "learning_rate": 1.2335734230394435e-06, "loss": 24.0408, "step": 68710 }, { "epoch": 0.85, "learning_rate": 1.2316576560647996e-06, "loss": 23.9495, "step": 68720 }, { "epoch": 0.85, "learning_rate": 1.2297432802323716e-06, "loss": 23.8152, "step": 68730 }, { "epoch": 0.85, "learning_rate": 1.2278302958458844e-06, "loss": 24.1374, "step": 68740 }, { "epoch": 0.85, "learning_rate": 1.2259187032088472e-06, "loss": 23.4418, "step": 68750 }, { "epoch": 0.85, "learning_rate": 1.2240085026245362e-06, "loss": 23.7614, "step": 68760 }, { "epoch": 0.85, "learning_rate": 1.2220996943960206e-06, "loss": 23.5564, "step": 68770 }, { "epoch": 0.85, "learning_rate": 1.220192278826141e-06, "loss": 24.2136, "step": 68780 }, { "epoch": 0.85, "learning_rate": 1.2182862562175167e-06, "loss": 23.6586, "step": 68790 }, { "epoch": 0.85, "learning_rate": 1.2163816268725515e-06, "loss": 24.1432, "step": 68800 }, { "epoch": 0.85, "learning_rate": 1.214478391093422e-06, "loss": 24.1025, "step": 68810 }, { "epoch": 0.85, "learning_rate": 1.2125765491820852e-06, "loss": 24.078, "step": 68820 }, { "epoch": 0.85, "learning_rate": 1.2106761014402813e-06, "loss": 23.7985, "step": 68830 }, { "epoch": 0.85, "learning_rate": 1.2087770481695238e-06, "loss": 24.1777, "step": 68840 }, { "epoch": 0.85, "learning_rate": 1.2068793896711062e-06, "loss": 23.2994, "step": 68850 }, { "epoch": 0.85, "learning_rate": 1.2049831262461053e-06, "loss": 23.3905, "step": 68860 }, { "epoch": 0.85, "learning_rate": 1.2030882581953717e-06, "loss": 23.4525, "step": 68870 }, { "epoch": 0.85, "learning_rate": 1.2011947858195316e-06, "loss": 23.8913, "step": 68880 }, { "epoch": 0.85, "learning_rate": 1.1993027094190002e-06, "loss": 23.3721, "step": 68890 }, { "epoch": 0.85, "learning_rate": 1.1974120292939629e-06, "loss": 23.8196, "step": 68900 }, { "epoch": 0.85, "learning_rate": 1.195522745744382e-06, "loss": 23.713, "step": 68910 }, { "epoch": 0.85, "learning_rate": 1.1936348590700086e-06, "loss": 23.907, "step": 68920 }, { "epoch": 0.85, "learning_rate": 1.1917483695703613e-06, "loss": 24.289, "step": 68930 }, { "epoch": 0.85, "learning_rate": 1.1898632775447395e-06, "loss": 23.9022, "step": 68940 }, { "epoch": 0.85, "learning_rate": 1.1879795832922281e-06, "loss": 23.5184, "step": 68950 }, { "epoch": 0.85, "learning_rate": 1.1860972871116817e-06, "loss": 23.8105, "step": 68960 }, { "epoch": 0.85, "learning_rate": 1.1842163893017344e-06, "loss": 23.0644, "step": 68970 }, { "epoch": 0.85, "learning_rate": 1.1823368901608067e-06, "loss": 23.4795, "step": 68980 }, { "epoch": 0.85, "learning_rate": 1.1804587899870823e-06, "loss": 24.0233, "step": 68990 }, { "epoch": 0.85, "learning_rate": 1.178582089078536e-06, "loss": 24.166, "step": 69000 }, { "epoch": 0.85, "learning_rate": 1.1767067877329163e-06, "loss": 23.9237, "step": 69010 }, { "epoch": 0.85, "learning_rate": 1.1748328862477466e-06, "loss": 24.1925, "step": 69020 }, { "epoch": 0.85, "learning_rate": 1.1729603849203342e-06, "loss": 23.771, "step": 69030 }, { "epoch": 0.85, "learning_rate": 1.1710892840477606e-06, "loss": 24.2904, "step": 69040 }, { "epoch": 0.85, "learning_rate": 1.1692195839268806e-06, "loss": 23.734, "step": 69050 }, { "epoch": 0.85, "learning_rate": 1.167351284854339e-06, "loss": 23.7157, "step": 69060 }, { "epoch": 0.85, "learning_rate": 1.1654843871265464e-06, "loss": 24.3247, "step": 69070 }, { "epoch": 0.85, "learning_rate": 1.1636188910396939e-06, "loss": 23.6212, "step": 69080 }, { "epoch": 0.85, "learning_rate": 1.161754796889758e-06, "loss": 23.5801, "step": 69090 }, { "epoch": 0.85, "learning_rate": 1.1598921049724832e-06, "loss": 23.5296, "step": 69100 }, { "epoch": 0.85, "learning_rate": 1.158030815583393e-06, "loss": 23.9414, "step": 69110 }, { "epoch": 0.85, "learning_rate": 1.1561709290177946e-06, "loss": 23.9732, "step": 69120 }, { "epoch": 0.85, "learning_rate": 1.1543124455707666e-06, "loss": 23.5914, "step": 69130 }, { "epoch": 0.85, "learning_rate": 1.1524553655371652e-06, "loss": 23.5675, "step": 69140 }, { "epoch": 0.85, "learning_rate": 1.15059968921163e-06, "loss": 24.1438, "step": 69150 }, { "epoch": 0.85, "learning_rate": 1.1487454168885692e-06, "loss": 23.9089, "step": 69160 }, { "epoch": 0.85, "learning_rate": 1.1468925488621741e-06, "loss": 23.936, "step": 69170 }, { "epoch": 0.85, "learning_rate": 1.1450410854264126e-06, "loss": 23.2422, "step": 69180 }, { "epoch": 0.85, "learning_rate": 1.143191026875028e-06, "loss": 23.9597, "step": 69190 }, { "epoch": 0.85, "learning_rate": 1.1413423735015395e-06, "loss": 23.8035, "step": 69200 }, { "epoch": 0.85, "learning_rate": 1.1394951255992515e-06, "loss": 24.4951, "step": 69210 }, { "epoch": 0.85, "learning_rate": 1.1376492834612307e-06, "loss": 23.2479, "step": 69220 }, { "epoch": 0.85, "learning_rate": 1.1358048473803363e-06, "loss": 24.3858, "step": 69230 }, { "epoch": 0.85, "learning_rate": 1.133961817649194e-06, "loss": 24.0756, "step": 69240 }, { "epoch": 0.85, "learning_rate": 1.1321201945602101e-06, "loss": 23.6983, "step": 69250 }, { "epoch": 0.85, "learning_rate": 1.1302799784055695e-06, "loss": 24.187, "step": 69260 }, { "epoch": 0.85, "learning_rate": 1.1284411694772313e-06, "loss": 23.5992, "step": 69270 }, { "epoch": 0.85, "learning_rate": 1.1266037680669294e-06, "loss": 24.1221, "step": 69280 }, { "epoch": 0.85, "learning_rate": 1.12476777446618e-06, "loss": 23.4305, "step": 69290 }, { "epoch": 0.85, "learning_rate": 1.122933188966272e-06, "loss": 23.8695, "step": 69300 }, { "epoch": 0.85, "learning_rate": 1.1211000118582694e-06, "loss": 23.4163, "step": 69310 }, { "epoch": 0.85, "learning_rate": 1.1192682434330193e-06, "loss": 23.6633, "step": 69320 }, { "epoch": 0.85, "learning_rate": 1.1174378839811384e-06, "loss": 23.8047, "step": 69330 }, { "epoch": 0.85, "learning_rate": 1.115608933793021e-06, "loss": 23.4548, "step": 69340 }, { "epoch": 0.85, "learning_rate": 1.1137813931588426e-06, "loss": 23.1774, "step": 69350 }, { "epoch": 0.85, "learning_rate": 1.1119552623685515e-06, "loss": 24.1544, "step": 69360 }, { "epoch": 0.85, "learning_rate": 1.11013054171187e-06, "loss": 23.7893, "step": 69370 }, { "epoch": 0.85, "learning_rate": 1.1083072314783027e-06, "loss": 23.3329, "step": 69380 }, { "epoch": 0.85, "learning_rate": 1.1064853319571256e-06, "loss": 24.0111, "step": 69390 }, { "epoch": 0.85, "learning_rate": 1.1046648434373896e-06, "loss": 24.022, "step": 69400 }, { "epoch": 0.85, "learning_rate": 1.1028457662079295e-06, "loss": 23.9388, "step": 69410 }, { "epoch": 0.85, "learning_rate": 1.1010281005573498e-06, "loss": 23.7614, "step": 69420 }, { "epoch": 0.85, "learning_rate": 1.0992118467740276e-06, "loss": 24.075, "step": 69430 }, { "epoch": 0.85, "learning_rate": 1.0973970051461302e-06, "loss": 23.5911, "step": 69440 }, { "epoch": 0.85, "learning_rate": 1.0955835759615808e-06, "loss": 24.0002, "step": 69450 }, { "epoch": 0.85, "learning_rate": 1.0937715595080966e-06, "loss": 23.9433, "step": 69460 }, { "epoch": 0.85, "learning_rate": 1.0919609560731615e-06, "loss": 24.008, "step": 69470 }, { "epoch": 0.85, "learning_rate": 1.090151765944033e-06, "loss": 23.6986, "step": 69480 }, { "epoch": 0.85, "learning_rate": 1.0883439894077551e-06, "loss": 24.0126, "step": 69490 }, { "epoch": 0.85, "learning_rate": 1.0865376267511363e-06, "loss": 23.9605, "step": 69500 }, { "epoch": 0.85, "learning_rate": 1.0847326782607658e-06, "loss": 23.5868, "step": 69510 }, { "epoch": 0.85, "learning_rate": 1.0829291442230095e-06, "loss": 23.8084, "step": 69520 }, { "epoch": 0.86, "learning_rate": 1.0811270249240057e-06, "loss": 24.6432, "step": 69530 }, { "epoch": 0.86, "learning_rate": 1.0793263206496696e-06, "loss": 23.8207, "step": 69540 }, { "epoch": 0.86, "learning_rate": 1.0775270316856935e-06, "loss": 23.6003, "step": 69550 }, { "epoch": 0.86, "learning_rate": 1.0757291583175434e-06, "loss": 24.0237, "step": 69560 }, { "epoch": 0.86, "learning_rate": 1.0739327008304589e-06, "loss": 23.5125, "step": 69570 }, { "epoch": 0.86, "learning_rate": 1.0721376595094613e-06, "loss": 23.503, "step": 69580 }, { "epoch": 0.86, "learning_rate": 1.0703440346393413e-06, "loss": 23.9902, "step": 69590 }, { "epoch": 0.86, "learning_rate": 1.0685518265046624e-06, "loss": 24.041, "step": 69600 }, { "epoch": 0.86, "learning_rate": 1.066761035389775e-06, "loss": 24.1583, "step": 69610 }, { "epoch": 0.86, "learning_rate": 1.064971661578792e-06, "loss": 23.4177, "step": 69620 }, { "epoch": 0.86, "learning_rate": 1.0631837053556072e-06, "loss": 23.4965, "step": 69630 }, { "epoch": 0.86, "learning_rate": 1.0613971670038914e-06, "loss": 23.4871, "step": 69640 }, { "epoch": 0.86, "learning_rate": 1.0596120468070848e-06, "loss": 24.0028, "step": 69650 }, { "epoch": 0.86, "learning_rate": 1.0578283450484095e-06, "loss": 23.2121, "step": 69660 }, { "epoch": 0.86, "learning_rate": 1.056046062010857e-06, "loss": 24.1079, "step": 69670 }, { "epoch": 0.86, "learning_rate": 1.0542651979771955e-06, "loss": 24.1012, "step": 69680 }, { "epoch": 0.86, "learning_rate": 1.0524857532299681e-06, "loss": 23.6787, "step": 69690 }, { "epoch": 0.86, "learning_rate": 1.0507077280514933e-06, "loss": 23.7683, "step": 69700 }, { "epoch": 0.86, "learning_rate": 1.0489311227238609e-06, "loss": 23.7481, "step": 69710 }, { "epoch": 0.86, "learning_rate": 1.0471559375289431e-06, "loss": 23.7547, "step": 69720 }, { "epoch": 0.86, "learning_rate": 1.0453821727483793e-06, "loss": 23.5648, "step": 69730 }, { "epoch": 0.86, "learning_rate": 1.0436098286635865e-06, "loss": 24.2042, "step": 69740 }, { "epoch": 0.86, "learning_rate": 1.0418389055557577e-06, "loss": 23.7106, "step": 69750 }, { "epoch": 0.86, "learning_rate": 1.0400694037058568e-06, "loss": 23.77, "step": 69760 }, { "epoch": 0.86, "learning_rate": 1.038301323394625e-06, "loss": 23.9524, "step": 69770 }, { "epoch": 0.86, "learning_rate": 1.0365346649025787e-06, "loss": 24.0457, "step": 69780 }, { "epoch": 0.86, "learning_rate": 1.0347694285100052e-06, "loss": 23.7285, "step": 69790 }, { "epoch": 0.86, "learning_rate": 1.033005614496968e-06, "loss": 23.6712, "step": 69800 }, { "epoch": 0.86, "learning_rate": 1.0312432231433078e-06, "loss": 23.8698, "step": 69810 }, { "epoch": 0.86, "learning_rate": 1.0294822547286343e-06, "loss": 23.9354, "step": 69820 }, { "epoch": 0.86, "learning_rate": 1.0277227095323372e-06, "loss": 23.897, "step": 69830 }, { "epoch": 0.86, "learning_rate": 1.0259645878335744e-06, "loss": 23.7033, "step": 69840 }, { "epoch": 0.86, "learning_rate": 1.0242078899112806e-06, "loss": 23.8214, "step": 69850 }, { "epoch": 0.86, "learning_rate": 1.0224526160441683e-06, "loss": 24.1214, "step": 69860 }, { "epoch": 0.86, "learning_rate": 1.0206987665107183e-06, "loss": 23.8079, "step": 69870 }, { "epoch": 0.86, "learning_rate": 1.0189463415891866e-06, "loss": 24.0003, "step": 69880 }, { "epoch": 0.86, "learning_rate": 1.0171953415576064e-06, "loss": 24.1915, "step": 69890 }, { "epoch": 0.86, "learning_rate": 1.0154457666937833e-06, "loss": 24.0599, "step": 69900 }, { "epoch": 0.86, "learning_rate": 1.0136976172752945e-06, "loss": 23.9584, "step": 69910 }, { "epoch": 0.86, "learning_rate": 1.011950893579493e-06, "loss": 23.7758, "step": 69920 }, { "epoch": 0.86, "learning_rate": 1.0102055958835066e-06, "loss": 23.6364, "step": 69930 }, { "epoch": 0.86, "learning_rate": 1.0084617244642325e-06, "loss": 24.1428, "step": 69940 }, { "epoch": 0.86, "learning_rate": 1.0067192795983494e-06, "loss": 24.0873, "step": 69950 }, { "epoch": 0.86, "learning_rate": 1.0049782615623028e-06, "loss": 23.5655, "step": 69960 }, { "epoch": 0.86, "learning_rate": 1.0032386706323127e-06, "loss": 23.8454, "step": 69970 }, { "epoch": 0.86, "learning_rate": 1.0015005070843764e-06, "loss": 23.5435, "step": 69980 }, { "epoch": 0.86, "learning_rate": 9.99763771194261e-07, "loss": 23.7218, "step": 69990 }, { "epoch": 0.86, "learning_rate": 9.980284632375093e-07, "loss": 23.364, "step": 70000 }, { "epoch": 0.86, "learning_rate": 9.962945834894377e-07, "loss": 23.5295, "step": 70010 }, { "epoch": 0.86, "learning_rate": 9.945621322251308e-07, "loss": 23.5903, "step": 70020 }, { "epoch": 0.86, "learning_rate": 9.928311097194566e-07, "loss": 24.5456, "step": 70030 }, { "epoch": 0.86, "learning_rate": 9.911015162470483e-07, "loss": 23.4847, "step": 70040 }, { "epoch": 0.86, "learning_rate": 9.893733520823112e-07, "loss": 23.7195, "step": 70050 }, { "epoch": 0.86, "learning_rate": 9.876466174994336e-07, "loss": 23.6106, "step": 70060 }, { "epoch": 0.86, "learning_rate": 9.859213127723665e-07, "loss": 23.5321, "step": 70070 }, { "epoch": 0.86, "learning_rate": 9.841974381748387e-07, "loss": 24.0813, "step": 70080 }, { "epoch": 0.86, "learning_rate": 9.824749939803535e-07, "loss": 23.9759, "step": 70090 }, { "epoch": 0.86, "learning_rate": 9.807539804621847e-07, "loss": 24.4115, "step": 70100 }, { "epoch": 0.86, "learning_rate": 9.790343978933769e-07, "loss": 23.5311, "step": 70110 }, { "epoch": 0.86, "learning_rate": 9.773162465467568e-07, "loss": 24.0952, "step": 70120 }, { "epoch": 0.86, "learning_rate": 9.755995266949126e-07, "loss": 23.8249, "step": 70130 }, { "epoch": 0.86, "learning_rate": 9.738842386102132e-07, "loss": 23.8278, "step": 70140 }, { "epoch": 0.86, "learning_rate": 9.721703825647966e-07, "loss": 23.6664, "step": 70150 }, { "epoch": 0.86, "learning_rate": 9.70457958830574e-07, "loss": 24.1363, "step": 70160 }, { "epoch": 0.86, "learning_rate": 9.687469676792327e-07, "loss": 24.338, "step": 70170 }, { "epoch": 0.86, "learning_rate": 9.670374093822298e-07, "loss": 23.975, "step": 70180 }, { "epoch": 0.86, "learning_rate": 9.65329284210792e-07, "loss": 23.7156, "step": 70190 }, { "epoch": 0.86, "learning_rate": 9.636225924359255e-07, "loss": 23.56, "step": 70200 }, { "epoch": 0.86, "learning_rate": 9.619173343284062e-07, "loss": 24.1552, "step": 70210 }, { "epoch": 0.86, "learning_rate": 9.602135101587772e-07, "loss": 24.0457, "step": 70220 }, { "epoch": 0.86, "learning_rate": 9.585111201973651e-07, "loss": 23.3555, "step": 70230 }, { "epoch": 0.86, "learning_rate": 9.568101647142592e-07, "loss": 23.7972, "step": 70240 }, { "epoch": 0.86, "learning_rate": 9.551106439793233e-07, "loss": 23.1561, "step": 70250 }, { "epoch": 0.86, "learning_rate": 9.534125582622e-07, "loss": 23.985, "step": 70260 }, { "epoch": 0.86, "learning_rate": 9.517159078322958e-07, "loss": 23.3882, "step": 70270 }, { "epoch": 0.86, "learning_rate": 9.500206929587907e-07, "loss": 23.3233, "step": 70280 }, { "epoch": 0.86, "learning_rate": 9.483269139106454e-07, "loss": 23.9822, "step": 70290 }, { "epoch": 0.86, "learning_rate": 9.466345709565827e-07, "loss": 23.826, "step": 70300 }, { "epoch": 0.86, "learning_rate": 9.449436643650999e-07, "loss": 24.4612, "step": 70310 }, { "epoch": 0.86, "learning_rate": 9.432541944044716e-07, "loss": 23.5568, "step": 70320 }, { "epoch": 0.86, "learning_rate": 9.415661613427396e-07, "loss": 23.9246, "step": 70330 }, { "epoch": 0.87, "learning_rate": 9.398795654477155e-07, "loss": 23.7526, "step": 70340 }, { "epoch": 0.87, "learning_rate": 9.381944069869931e-07, "loss": 24.0228, "step": 70350 }, { "epoch": 0.87, "learning_rate": 9.36510686227926e-07, "loss": 23.4316, "step": 70360 }, { "epoch": 0.87, "learning_rate": 9.348284034376476e-07, "loss": 23.3663, "step": 70370 }, { "epoch": 0.87, "learning_rate": 9.331475588830596e-07, "loss": 23.8104, "step": 70380 }, { "epoch": 0.87, "learning_rate": 9.314681528308345e-07, "loss": 23.9623, "step": 70390 }, { "epoch": 0.87, "learning_rate": 9.297901855474234e-07, "loss": 24.1744, "step": 70400 }, { "epoch": 0.87, "learning_rate": 9.281136572990412e-07, "loss": 23.7806, "step": 70410 }, { "epoch": 0.87, "learning_rate": 9.264385683516775e-07, "loss": 23.7174, "step": 70420 }, { "epoch": 0.87, "learning_rate": 9.24764918971095e-07, "loss": 23.2937, "step": 70430 }, { "epoch": 0.87, "learning_rate": 9.23092709422827e-07, "loss": 23.9435, "step": 70440 }, { "epoch": 0.87, "learning_rate": 9.214219399721758e-07, "loss": 23.7909, "step": 70450 }, { "epoch": 0.87, "learning_rate": 9.197526108842192e-07, "loss": 24.2372, "step": 70460 }, { "epoch": 0.87, "learning_rate": 9.180847224238065e-07, "loss": 24.3432, "step": 70470 }, { "epoch": 0.87, "learning_rate": 9.164182748555517e-07, "loss": 23.7926, "step": 70480 }, { "epoch": 0.87, "learning_rate": 9.147532684438498e-07, "loss": 23.9308, "step": 70490 }, { "epoch": 0.87, "learning_rate": 9.130897034528618e-07, "loss": 24.2784, "step": 70500 }, { "epoch": 0.87, "learning_rate": 9.114275801465178e-07, "loss": 24.5176, "step": 70510 }, { "epoch": 0.87, "learning_rate": 9.097668987885256e-07, "loss": 23.5846, "step": 70520 }, { "epoch": 0.87, "learning_rate": 9.081076596423599e-07, "loss": 23.6819, "step": 70530 }, { "epoch": 0.87, "learning_rate": 9.064498629712659e-07, "loss": 24.2249, "step": 70540 }, { "epoch": 0.87, "learning_rate": 9.047935090382631e-07, "loss": 23.3734, "step": 70550 }, { "epoch": 0.87, "learning_rate": 9.031385981061414e-07, "loss": 23.7021, "step": 70560 }, { "epoch": 0.87, "learning_rate": 9.014851304374561e-07, "loss": 23.7247, "step": 70570 }, { "epoch": 0.87, "learning_rate": 8.998331062945442e-07, "loss": 24.2642, "step": 70580 }, { "epoch": 0.87, "learning_rate": 8.98182525939506e-07, "loss": 22.8952, "step": 70590 }, { "epoch": 0.87, "learning_rate": 8.965333896342132e-07, "loss": 24.0276, "step": 70600 }, { "epoch": 0.87, "learning_rate": 8.948856976403098e-07, "loss": 23.8761, "step": 70610 }, { "epoch": 0.87, "learning_rate": 8.932394502192088e-07, "loss": 24.5274, "step": 70620 }, { "epoch": 0.87, "learning_rate": 8.915946476321003e-07, "loss": 23.8942, "step": 70630 }, { "epoch": 0.87, "learning_rate": 8.899512901399376e-07, "loss": 23.9283, "step": 70640 }, { "epoch": 0.87, "learning_rate": 8.883093780034469e-07, "loss": 23.8657, "step": 70650 }, { "epoch": 0.87, "learning_rate": 8.866689114831284e-07, "loss": 24.1244, "step": 70660 }, { "epoch": 0.87, "learning_rate": 8.850298908392496e-07, "loss": 23.6498, "step": 70670 }, { "epoch": 0.87, "learning_rate": 8.833923163318481e-07, "loss": 23.5549, "step": 70680 }, { "epoch": 0.87, "learning_rate": 8.817561882207359e-07, "loss": 24.2292, "step": 70690 }, { "epoch": 0.87, "learning_rate": 8.801215067654922e-07, "loss": 23.5689, "step": 70700 }, { "epoch": 0.87, "learning_rate": 8.784882722254651e-07, "loss": 24.6806, "step": 70710 }, { "epoch": 0.87, "learning_rate": 8.768564848597793e-07, "loss": 23.8123, "step": 70720 }, { "epoch": 0.87, "learning_rate": 8.752261449273258e-07, "loss": 23.7495, "step": 70730 }, { "epoch": 0.87, "learning_rate": 8.73597252686762e-07, "loss": 23.5431, "step": 70740 }, { "epoch": 0.87, "learning_rate": 8.719698083965244e-07, "loss": 23.9913, "step": 70750 }, { "epoch": 0.87, "learning_rate": 8.703438123148145e-07, "loss": 23.7427, "step": 70760 }, { "epoch": 0.87, "learning_rate": 8.687192646996024e-07, "loss": 23.4873, "step": 70770 }, { "epoch": 0.87, "learning_rate": 8.670961658086352e-07, "loss": 23.6757, "step": 70780 }, { "epoch": 0.87, "learning_rate": 8.654745158994227e-07, "loss": 23.8998, "step": 70790 }, { "epoch": 0.87, "learning_rate": 8.638543152292478e-07, "loss": 23.9613, "step": 70800 }, { "epoch": 0.87, "learning_rate": 8.62235564055165e-07, "loss": 24.439, "step": 70810 }, { "epoch": 0.87, "learning_rate": 8.606182626339976e-07, "loss": 23.8535, "step": 70820 }, { "epoch": 0.87, "learning_rate": 8.590024112223372e-07, "loss": 23.8314, "step": 70830 }, { "epoch": 0.87, "learning_rate": 8.573880100765485e-07, "loss": 23.9723, "step": 70840 }, { "epoch": 0.87, "learning_rate": 8.557750594527614e-07, "loss": 23.4889, "step": 70850 }, { "epoch": 0.87, "learning_rate": 8.54163559606882e-07, "loss": 24.1801, "step": 70860 }, { "epoch": 0.87, "learning_rate": 8.525535107945826e-07, "loss": 24.1387, "step": 70870 }, { "epoch": 0.87, "learning_rate": 8.509449132713032e-07, "loss": 24.0987, "step": 70880 }, { "epoch": 0.87, "learning_rate": 8.493377672922598e-07, "loss": 23.9595, "step": 70890 }, { "epoch": 0.87, "learning_rate": 8.477320731124317e-07, "loss": 24.2374, "step": 70900 }, { "epoch": 0.87, "learning_rate": 8.461278309865694e-07, "loss": 24.4852, "step": 70910 }, { "epoch": 0.87, "learning_rate": 8.445250411691974e-07, "loss": 24.2822, "step": 70920 }, { "epoch": 0.87, "learning_rate": 8.429237039146054e-07, "loss": 24.3277, "step": 70930 }, { "epoch": 0.87, "learning_rate": 8.413238194768514e-07, "loss": 23.3094, "step": 70940 }, { "epoch": 0.87, "learning_rate": 8.397253881097678e-07, "loss": 24.213, "step": 70950 }, { "epoch": 0.87, "learning_rate": 8.381284100669529e-07, "loss": 23.1896, "step": 70960 }, { "epoch": 0.87, "learning_rate": 8.365328856017751e-07, "loss": 24.3724, "step": 70970 }, { "epoch": 0.87, "learning_rate": 8.349388149673732e-07, "loss": 23.5761, "step": 70980 }, { "epoch": 0.87, "learning_rate": 8.333461984166557e-07, "loss": 24.0925, "step": 70990 }, { "epoch": 0.87, "learning_rate": 8.317550362022953e-07, "loss": 23.771, "step": 71000 }, { "epoch": 0.87, "learning_rate": 8.301653285767431e-07, "loss": 24.4255, "step": 71010 }, { "epoch": 0.87, "learning_rate": 8.285770757922107e-07, "loss": 24.1028, "step": 71020 }, { "epoch": 0.87, "learning_rate": 8.269902781006833e-07, "loss": 23.8164, "step": 71030 }, { "epoch": 0.87, "learning_rate": 8.254049357539151e-07, "loss": 24.3531, "step": 71040 }, { "epoch": 0.87, "learning_rate": 8.238210490034305e-07, "loss": 23.8716, "step": 71050 }, { "epoch": 0.87, "learning_rate": 8.222386181005171e-07, "loss": 23.9372, "step": 71060 }, { "epoch": 0.87, "learning_rate": 8.206576432962388e-07, "loss": 24.1948, "step": 71070 }, { "epoch": 0.87, "learning_rate": 8.190781248414226e-07, "loss": 23.8554, "step": 71080 }, { "epoch": 0.87, "learning_rate": 8.175000629866703e-07, "loss": 23.494, "step": 71090 }, { "epoch": 0.87, "learning_rate": 8.159234579823483e-07, "loss": 23.5115, "step": 71100 }, { "epoch": 0.87, "learning_rate": 8.14348310078591e-07, "loss": 24.0785, "step": 71110 }, { "epoch": 0.87, "learning_rate": 8.127746195253061e-07, "loss": 23.8407, "step": 71120 }, { "epoch": 0.87, "learning_rate": 8.112023865721686e-07, "loss": 24.1286, "step": 71130 }, { "epoch": 0.87, "learning_rate": 8.096316114686165e-07, "loss": 24.2476, "step": 71140 }, { "epoch": 0.88, "learning_rate": 8.080622944638661e-07, "loss": 24.4565, "step": 71150 }, { "epoch": 0.88, "learning_rate": 8.06494435806896e-07, "loss": 23.7715, "step": 71160 }, { "epoch": 0.88, "learning_rate": 8.049280357464528e-07, "loss": 23.6662, "step": 71170 }, { "epoch": 0.88, "learning_rate": 8.033630945310578e-07, "loss": 24.4739, "step": 71180 }, { "epoch": 0.88, "learning_rate": 8.017996124089944e-07, "loss": 23.3432, "step": 71190 }, { "epoch": 0.88, "learning_rate": 8.002375896283154e-07, "loss": 23.9775, "step": 71200 }, { "epoch": 0.88, "learning_rate": 7.986770264368482e-07, "loss": 23.7239, "step": 71210 }, { "epoch": 0.88, "learning_rate": 7.971179230821802e-07, "loss": 23.8797, "step": 71220 }, { "epoch": 0.88, "learning_rate": 7.955602798116735e-07, "loss": 23.8492, "step": 71230 }, { "epoch": 0.88, "learning_rate": 7.940040968724561e-07, "loss": 23.6145, "step": 71240 }, { "epoch": 0.88, "learning_rate": 7.924493745114214e-07, "loss": 23.5536, "step": 71250 }, { "epoch": 0.88, "learning_rate": 7.90896112975238e-07, "loss": 24.1676, "step": 71260 }, { "epoch": 0.88, "learning_rate": 7.893443125103362e-07, "loss": 24.4101, "step": 71270 }, { "epoch": 0.88, "learning_rate": 7.877939733629169e-07, "loss": 24.0364, "step": 71280 }, { "epoch": 0.88, "learning_rate": 7.862450957789514e-07, "loss": 23.8143, "step": 71290 }, { "epoch": 0.88, "learning_rate": 7.846976800041773e-07, "loss": 23.5594, "step": 71300 }, { "epoch": 0.88, "learning_rate": 7.831517262840948e-07, "loss": 24.0794, "step": 71310 }, { "epoch": 0.88, "learning_rate": 7.816072348639836e-07, "loss": 23.9746, "step": 71320 }, { "epoch": 0.88, "learning_rate": 7.800642059888819e-07, "loss": 23.1279, "step": 71330 }, { "epoch": 0.88, "learning_rate": 7.785226399035972e-07, "loss": 24.0188, "step": 71340 }, { "epoch": 0.88, "learning_rate": 7.769825368527128e-07, "loss": 24.0579, "step": 71350 }, { "epoch": 0.88, "learning_rate": 7.754438970805688e-07, "loss": 24.0049, "step": 71360 }, { "epoch": 0.88, "learning_rate": 7.739067208312801e-07, "loss": 24.1154, "step": 71370 }, { "epoch": 0.88, "learning_rate": 7.723710083487279e-07, "loss": 23.5618, "step": 71380 }, { "epoch": 0.88, "learning_rate": 7.708367598765576e-07, "loss": 23.7814, "step": 71390 }, { "epoch": 0.88, "learning_rate": 7.693039756581911e-07, "loss": 23.4051, "step": 71400 }, { "epoch": 0.88, "learning_rate": 7.677726559368092e-07, "loss": 24.1361, "step": 71410 }, { "epoch": 0.88, "learning_rate": 7.662428009553624e-07, "loss": 23.2671, "step": 71420 }, { "epoch": 0.88, "learning_rate": 7.64714410956574e-07, "loss": 23.9506, "step": 71430 }, { "epoch": 0.88, "learning_rate": 7.631874861829269e-07, "loss": 23.9537, "step": 71440 }, { "epoch": 0.88, "learning_rate": 7.616620268766761e-07, "loss": 23.7956, "step": 71450 }, { "epoch": 0.88, "learning_rate": 7.601380332798459e-07, "loss": 23.8521, "step": 71460 }, { "epoch": 0.88, "learning_rate": 7.586155056342237e-07, "loss": 24.1011, "step": 71470 }, { "epoch": 0.88, "learning_rate": 7.570944441813654e-07, "loss": 23.8003, "step": 71480 }, { "epoch": 0.88, "learning_rate": 7.555748491625981e-07, "loss": 24.3399, "step": 71490 }, { "epoch": 0.88, "learning_rate": 7.54056720819012e-07, "loss": 23.6823, "step": 71500 }, { "epoch": 0.88, "learning_rate": 7.525400593914633e-07, "loss": 23.756, "step": 71510 }, { "epoch": 0.88, "learning_rate": 7.51024865120582e-07, "loss": 23.6319, "step": 71520 }, { "epoch": 0.88, "learning_rate": 7.495111382467623e-07, "loss": 23.6804, "step": 71530 }, { "epoch": 0.88, "learning_rate": 7.479988790101578e-07, "loss": 23.2713, "step": 71540 }, { "epoch": 0.88, "learning_rate": 7.464880876507019e-07, "loss": 24.097, "step": 71550 }, { "epoch": 0.88, "learning_rate": 7.449787644080875e-07, "loss": 24.0205, "step": 71560 }, { "epoch": 0.88, "learning_rate": 7.434709095217774e-07, "loss": 24.039, "step": 71570 }, { "epoch": 0.88, "learning_rate": 7.419645232310002e-07, "loss": 24.1074, "step": 71580 }, { "epoch": 0.88, "learning_rate": 7.404596057747493e-07, "loss": 23.9022, "step": 71590 }, { "epoch": 0.88, "learning_rate": 7.389561573917914e-07, "loss": 24.3424, "step": 71600 }, { "epoch": 0.88, "learning_rate": 7.374541783206557e-07, "loss": 24.0704, "step": 71610 }, { "epoch": 0.88, "learning_rate": 7.359536687996349e-07, "loss": 23.9724, "step": 71620 }, { "epoch": 0.88, "learning_rate": 7.344546290667965e-07, "loss": 23.9877, "step": 71630 }, { "epoch": 0.88, "learning_rate": 7.329570593599711e-07, "loss": 24.2029, "step": 71640 }, { "epoch": 0.88, "learning_rate": 7.31460959916751e-07, "loss": 23.4266, "step": 71650 }, { "epoch": 0.88, "learning_rate": 7.299663309745053e-07, "loss": 24.0957, "step": 71660 }, { "epoch": 0.88, "learning_rate": 7.284731727703631e-07, "loss": 23.7709, "step": 71670 }, { "epoch": 0.88, "learning_rate": 7.269814855412183e-07, "loss": 23.7143, "step": 71680 }, { "epoch": 0.88, "learning_rate": 7.254912695237393e-07, "loss": 23.719, "step": 71690 }, { "epoch": 0.88, "learning_rate": 7.240025249543548e-07, "loss": 23.9673, "step": 71700 }, { "epoch": 0.88, "learning_rate": 7.22515252069259e-07, "loss": 23.7324, "step": 71710 }, { "epoch": 0.88, "learning_rate": 7.210294511044203e-07, "loss": 23.3074, "step": 71720 }, { "epoch": 0.88, "learning_rate": 7.195451222955652e-07, "loss": 23.8443, "step": 71730 }, { "epoch": 0.88, "learning_rate": 7.180622658781888e-07, "loss": 24.1441, "step": 71740 }, { "epoch": 0.88, "learning_rate": 7.165808820875586e-07, "loss": 24.3866, "step": 71750 }, { "epoch": 0.88, "learning_rate": 7.151009711586998e-07, "loss": 24.1107, "step": 71760 }, { "epoch": 0.88, "learning_rate": 7.136225333264102e-07, "loss": 23.9101, "step": 71770 }, { "epoch": 0.88, "learning_rate": 7.121455688252488e-07, "loss": 24.0611, "step": 71780 }, { "epoch": 0.88, "learning_rate": 7.106700778895448e-07, "loss": 24.0949, "step": 71790 }, { "epoch": 0.88, "learning_rate": 7.091960607533932e-07, "loss": 24.4003, "step": 71800 }, { "epoch": 0.88, "learning_rate": 7.077235176506536e-07, "loss": 24.16, "step": 71810 }, { "epoch": 0.88, "learning_rate": 7.062524488149514e-07, "loss": 24.0125, "step": 71820 }, { "epoch": 0.88, "learning_rate": 7.04782854479682e-07, "loss": 23.5007, "step": 71830 }, { "epoch": 0.88, "learning_rate": 7.033147348780012e-07, "loss": 23.6431, "step": 71840 }, { "epoch": 0.88, "learning_rate": 7.018480902428349e-07, "loss": 23.8566, "step": 71850 }, { "epoch": 0.88, "learning_rate": 7.003829208068747e-07, "loss": 23.6168, "step": 71860 }, { "epoch": 0.88, "learning_rate": 6.989192268025757e-07, "loss": 23.2288, "step": 71870 }, { "epoch": 0.88, "learning_rate": 6.974570084621591e-07, "loss": 23.7623, "step": 71880 }, { "epoch": 0.88, "learning_rate": 6.959962660176167e-07, "loss": 23.956, "step": 71890 }, { "epoch": 0.88, "learning_rate": 6.945369997007023e-07, "loss": 23.9682, "step": 71900 }, { "epoch": 0.88, "learning_rate": 6.930792097429317e-07, "loss": 23.6139, "step": 71910 }, { "epoch": 0.88, "learning_rate": 6.916228963755967e-07, "loss": 23.8625, "step": 71920 }, { "epoch": 0.88, "learning_rate": 6.901680598297444e-07, "loss": 23.5291, "step": 71930 }, { "epoch": 0.88, "learning_rate": 6.887147003361927e-07, "loss": 23.8222, "step": 71940 }, { "epoch": 0.88, "learning_rate": 6.87262818125527e-07, "loss": 23.5358, "step": 71950 }, { "epoch": 0.88, "learning_rate": 6.858124134280941e-07, "loss": 23.751, "step": 71960 }, { "epoch": 0.89, "learning_rate": 6.843634864740067e-07, "loss": 23.7615, "step": 71970 }, { "epoch": 0.89, "learning_rate": 6.829160374931476e-07, "loss": 23.6613, "step": 71980 }, { "epoch": 0.89, "learning_rate": 6.814700667151597e-07, "loss": 24.168, "step": 71990 }, { "epoch": 0.89, "learning_rate": 6.800255743694551e-07, "loss": 23.7035, "step": 72000 }, { "epoch": 0.89, "learning_rate": 6.785825606852082e-07, "loss": 23.569, "step": 72010 }, { "epoch": 0.89, "learning_rate": 6.771410258913591e-07, "loss": 22.9489, "step": 72020 }, { "epoch": 0.89, "learning_rate": 6.757009702166195e-07, "loss": 24.0478, "step": 72030 }, { "epoch": 0.89, "learning_rate": 6.742623938894588e-07, "loss": 24.3781, "step": 72040 }, { "epoch": 0.89, "learning_rate": 6.728252971381122e-07, "loss": 23.6149, "step": 72050 }, { "epoch": 0.89, "learning_rate": 6.713896801905861e-07, "loss": 23.2565, "step": 72060 }, { "epoch": 0.89, "learning_rate": 6.699555432746486e-07, "loss": 23.7846, "step": 72070 }, { "epoch": 0.89, "learning_rate": 6.685228866178284e-07, "loss": 23.8432, "step": 72080 }, { "epoch": 0.89, "learning_rate": 6.670917104474284e-07, "loss": 23.3485, "step": 72090 }, { "epoch": 0.89, "learning_rate": 6.656620149905113e-07, "loss": 23.3449, "step": 72100 }, { "epoch": 0.89, "learning_rate": 6.642338004739024e-07, "loss": 24.2827, "step": 72110 }, { "epoch": 0.89, "learning_rate": 6.628070671241982e-07, "loss": 24.0312, "step": 72120 }, { "epoch": 0.89, "learning_rate": 6.613818151677576e-07, "loss": 24.1299, "step": 72130 }, { "epoch": 0.89, "learning_rate": 6.599580448307008e-07, "loss": 24.2842, "step": 72140 }, { "epoch": 0.89, "learning_rate": 6.585357563389194e-07, "loss": 23.5661, "step": 72150 }, { "epoch": 0.89, "learning_rate": 6.571149499180662e-07, "loss": 23.9517, "step": 72160 }, { "epoch": 0.89, "learning_rate": 6.556956257935576e-07, "loss": 23.569, "step": 72170 }, { "epoch": 0.89, "learning_rate": 6.542777841905789e-07, "loss": 23.8793, "step": 72180 }, { "epoch": 0.89, "learning_rate": 6.528614253340771e-07, "loss": 23.9164, "step": 72190 }, { "epoch": 0.89, "learning_rate": 6.514465494487631e-07, "loss": 23.7368, "step": 72200 }, { "epoch": 0.89, "learning_rate": 6.500331567591167e-07, "loss": 24.3119, "step": 72210 }, { "epoch": 0.89, "learning_rate": 6.486212474893793e-07, "loss": 23.4717, "step": 72220 }, { "epoch": 0.89, "learning_rate": 6.472108218635564e-07, "loss": 24.1465, "step": 72230 }, { "epoch": 0.89, "learning_rate": 6.4580188010542e-07, "loss": 23.9763, "step": 72240 }, { "epoch": 0.89, "learning_rate": 6.443944224385046e-07, "loss": 24.754, "step": 72250 }, { "epoch": 0.89, "learning_rate": 6.429884490861115e-07, "loss": 23.9321, "step": 72260 }, { "epoch": 0.89, "learning_rate": 6.41583960271307e-07, "loss": 23.7488, "step": 72270 }, { "epoch": 0.89, "learning_rate": 6.40180956216917e-07, "loss": 23.6985, "step": 72280 }, { "epoch": 0.89, "learning_rate": 6.387794371455392e-07, "loss": 23.9112, "step": 72290 }, { "epoch": 0.89, "learning_rate": 6.373794032795299e-07, "loss": 24.0007, "step": 72300 }, { "epoch": 0.89, "learning_rate": 6.359808548410096e-07, "loss": 23.8666, "step": 72310 }, { "epoch": 0.89, "learning_rate": 6.345837920518683e-07, "loss": 23.7804, "step": 72320 }, { "epoch": 0.89, "learning_rate": 6.331882151337565e-07, "loss": 23.4703, "step": 72330 }, { "epoch": 0.89, "learning_rate": 6.317941243080872e-07, "loss": 23.327, "step": 72340 }, { "epoch": 0.89, "learning_rate": 6.304015197960434e-07, "loss": 23.5313, "step": 72350 }, { "epoch": 0.89, "learning_rate": 6.290104018185672e-07, "loss": 23.7555, "step": 72360 }, { "epoch": 0.89, "learning_rate": 6.276207705963655e-07, "loss": 24.5247, "step": 72370 }, { "epoch": 0.89, "learning_rate": 6.262326263499119e-07, "loss": 24.1871, "step": 72380 }, { "epoch": 0.89, "learning_rate": 6.248459692994435e-07, "loss": 23.428, "step": 72390 }, { "epoch": 0.89, "learning_rate": 6.234607996649577e-07, "loss": 24.2333, "step": 72400 }, { "epoch": 0.89, "learning_rate": 6.22077117666221e-07, "loss": 23.3992, "step": 72410 }, { "epoch": 0.89, "learning_rate": 6.20694923522761e-07, "loss": 23.1569, "step": 72420 }, { "epoch": 0.89, "learning_rate": 6.193142174538691e-07, "loss": 23.8535, "step": 72430 }, { "epoch": 0.89, "learning_rate": 6.179349996786022e-07, "loss": 23.8645, "step": 72440 }, { "epoch": 0.89, "learning_rate": 6.165572704157807e-07, "loss": 24.386, "step": 72450 }, { "epoch": 0.89, "learning_rate": 6.151810298839878e-07, "loss": 24.0319, "step": 72460 }, { "epoch": 0.89, "learning_rate": 6.138062783015708e-07, "loss": 23.2216, "step": 72470 }, { "epoch": 0.89, "learning_rate": 6.124330158866399e-07, "loss": 24.0324, "step": 72480 }, { "epoch": 0.89, "learning_rate": 6.110612428570717e-07, "loss": 24.0345, "step": 72490 }, { "epoch": 0.89, "learning_rate": 6.096909594305056e-07, "loss": 23.715, "step": 72500 }, { "epoch": 0.89, "learning_rate": 6.083221658243421e-07, "loss": 23.7634, "step": 72510 }, { "epoch": 0.89, "learning_rate": 6.069548622557498e-07, "loss": 24.0712, "step": 72520 }, { "epoch": 0.89, "learning_rate": 6.055890489416572e-07, "loss": 23.9731, "step": 72530 }, { "epoch": 0.89, "learning_rate": 6.042247260987544e-07, "loss": 23.723, "step": 72540 }, { "epoch": 0.89, "learning_rate": 6.028618939435038e-07, "loss": 23.436, "step": 72550 }, { "epoch": 0.89, "learning_rate": 6.015005526921224e-07, "loss": 23.5232, "step": 72560 }, { "epoch": 0.89, "learning_rate": 6.001407025605943e-07, "loss": 23.8935, "step": 72570 }, { "epoch": 0.89, "learning_rate": 5.987823437646667e-07, "loss": 24.2654, "step": 72580 }, { "epoch": 0.89, "learning_rate": 5.974254765198517e-07, "loss": 24.1485, "step": 72590 }, { "epoch": 0.89, "learning_rate": 5.960701010414193e-07, "loss": 23.8062, "step": 72600 }, { "epoch": 0.89, "learning_rate": 5.94716217544411e-07, "loss": 24.3039, "step": 72610 }, { "epoch": 0.89, "learning_rate": 5.933638262436259e-07, "loss": 23.55, "step": 72620 }, { "epoch": 0.89, "learning_rate": 5.920129273536257e-07, "loss": 23.7169, "step": 72630 }, { "epoch": 0.89, "learning_rate": 5.906635210887402e-07, "loss": 24.2893, "step": 72640 }, { "epoch": 0.89, "learning_rate": 5.893156076630568e-07, "loss": 24.6173, "step": 72650 }, { "epoch": 0.89, "learning_rate": 5.879691872904325e-07, "loss": 24.4307, "step": 72660 }, { "epoch": 0.89, "learning_rate": 5.866242601844818e-07, "loss": 24.2454, "step": 72670 }, { "epoch": 0.89, "learning_rate": 5.852808265585852e-07, "loss": 23.909, "step": 72680 }, { "epoch": 0.89, "learning_rate": 5.839388866258833e-07, "loss": 23.8768, "step": 72690 }, { "epoch": 0.89, "learning_rate": 5.825984405992824e-07, "loss": 23.5698, "step": 72700 }, { "epoch": 0.89, "learning_rate": 5.812594886914502e-07, "loss": 23.9932, "step": 72710 }, { "epoch": 0.89, "learning_rate": 5.799220311148213e-07, "loss": 24.1187, "step": 72720 }, { "epoch": 0.89, "learning_rate": 5.785860680815891e-07, "loss": 23.2926, "step": 72730 }, { "epoch": 0.89, "learning_rate": 5.772515998037076e-07, "loss": 23.5919, "step": 72740 }, { "epoch": 0.89, "learning_rate": 5.759186264929018e-07, "loss": 23.446, "step": 72750 }, { "epoch": 0.89, "learning_rate": 5.745871483606513e-07, "loss": 23.7411, "step": 72760 }, { "epoch": 0.89, "learning_rate": 5.732571656182029e-07, "loss": 23.5466, "step": 72770 }, { "epoch": 0.9, "learning_rate": 5.719286784765665e-07, "loss": 23.7101, "step": 72780 }, { "epoch": 0.9, "learning_rate": 5.706016871465103e-07, "loss": 23.8765, "step": 72790 }, { "epoch": 0.9, "learning_rate": 5.692761918385714e-07, "loss": 23.7706, "step": 72800 }, { "epoch": 0.9, "learning_rate": 5.67952192763046e-07, "loss": 23.7292, "step": 72810 }, { "epoch": 0.9, "learning_rate": 5.666296901299906e-07, "loss": 23.5305, "step": 72820 }, { "epoch": 0.9, "learning_rate": 5.653086841492305e-07, "loss": 23.7825, "step": 72830 }, { "epoch": 0.9, "learning_rate": 5.639891750303484e-07, "loss": 24.0043, "step": 72840 }, { "epoch": 0.9, "learning_rate": 5.626711629826887e-07, "loss": 23.412, "step": 72850 }, { "epoch": 0.9, "learning_rate": 5.613546482153654e-07, "loss": 24.1241, "step": 72860 }, { "epoch": 0.9, "learning_rate": 5.60039630937248e-07, "loss": 24.2666, "step": 72870 }, { "epoch": 0.9, "learning_rate": 5.587261113569686e-07, "loss": 24.2748, "step": 72880 }, { "epoch": 0.9, "learning_rate": 5.574140896829283e-07, "loss": 23.9299, "step": 72890 }, { "epoch": 0.9, "learning_rate": 5.561035661232828e-07, "loss": 24.0523, "step": 72900 }, { "epoch": 0.9, "learning_rate": 5.547945408859534e-07, "loss": 24.07, "step": 72910 }, { "epoch": 0.9, "learning_rate": 5.534870141786264e-07, "loss": 24.1196, "step": 72920 }, { "epoch": 0.9, "learning_rate": 5.521809862087446e-07, "loss": 24.0832, "step": 72930 }, { "epoch": 0.9, "learning_rate": 5.508764571835156e-07, "loss": 23.6548, "step": 72940 }, { "epoch": 0.9, "learning_rate": 5.495734273099129e-07, "loss": 23.2041, "step": 72950 }, { "epoch": 0.9, "learning_rate": 5.482718967946643e-07, "loss": 23.9314, "step": 72960 }, { "epoch": 0.9, "learning_rate": 5.469718658442691e-07, "loss": 23.5215, "step": 72970 }, { "epoch": 0.9, "learning_rate": 5.456733346649812e-07, "loss": 23.5694, "step": 72980 }, { "epoch": 0.9, "learning_rate": 5.443763034628192e-07, "loss": 23.7572, "step": 72990 }, { "epoch": 0.9, "learning_rate": 5.430807724435638e-07, "loss": 23.5264, "step": 73000 }, { "epoch": 0.9, "learning_rate": 5.417867418127587e-07, "loss": 23.878, "step": 73010 }, { "epoch": 0.9, "learning_rate": 5.404942117757061e-07, "loss": 24.099, "step": 73020 }, { "epoch": 0.9, "learning_rate": 5.392031825374755e-07, "loss": 23.9893, "step": 73030 }, { "epoch": 0.9, "learning_rate": 5.37913654302894e-07, "loss": 23.8925, "step": 73040 }, { "epoch": 0.9, "learning_rate": 5.366256272765502e-07, "loss": 23.4166, "step": 73050 }, { "epoch": 0.9, "learning_rate": 5.353391016627985e-07, "loss": 23.7844, "step": 73060 }, { "epoch": 0.9, "learning_rate": 5.340540776657521e-07, "loss": 23.8375, "step": 73070 }, { "epoch": 0.9, "learning_rate": 5.327705554892848e-07, "loss": 24.4981, "step": 73080 }, { "epoch": 0.9, "learning_rate": 5.314885353370369e-07, "loss": 23.9342, "step": 73090 }, { "epoch": 0.9, "learning_rate": 5.302080174124058e-07, "loss": 23.9098, "step": 73100 }, { "epoch": 0.9, "learning_rate": 5.289290019185511e-07, "loss": 23.8587, "step": 73110 }, { "epoch": 0.9, "learning_rate": 5.276514890583972e-07, "loss": 23.5079, "step": 73120 }, { "epoch": 0.9, "learning_rate": 5.263754790346288e-07, "loss": 23.7009, "step": 73130 }, { "epoch": 0.9, "learning_rate": 5.251009720496869e-07, "loss": 23.338, "step": 73140 }, { "epoch": 0.9, "learning_rate": 5.238279683057857e-07, "loss": 23.869, "step": 73150 }, { "epoch": 0.9, "learning_rate": 5.225564680048856e-07, "loss": 24.0066, "step": 73160 }, { "epoch": 0.9, "learning_rate": 5.21286471348722e-07, "loss": 24.0128, "step": 73170 }, { "epoch": 0.9, "learning_rate": 5.200179785387848e-07, "loss": 23.9646, "step": 73180 }, { "epoch": 0.9, "learning_rate": 5.187509897763266e-07, "loss": 23.7068, "step": 73190 }, { "epoch": 0.9, "learning_rate": 5.174855052623617e-07, "loss": 24.0209, "step": 73200 }, { "epoch": 0.9, "learning_rate": 5.162215251976665e-07, "loss": 23.5722, "step": 73210 }, { "epoch": 0.9, "learning_rate": 5.149590497827761e-07, "loss": 23.2564, "step": 73220 }, { "epoch": 0.9, "learning_rate": 5.136980792179913e-07, "loss": 23.6033, "step": 73230 }, { "epoch": 0.9, "learning_rate": 5.124386137033699e-07, "loss": 23.3061, "step": 73240 }, { "epoch": 0.9, "learning_rate": 5.11180653438732e-07, "loss": 23.0578, "step": 73250 }, { "epoch": 0.9, "learning_rate": 5.099241986236603e-07, "loss": 23.4119, "step": 73260 }, { "epoch": 0.9, "learning_rate": 5.086692494574985e-07, "loss": 24.1229, "step": 73270 }, { "epoch": 0.9, "learning_rate": 5.074158061393475e-07, "loss": 23.7236, "step": 73280 }, { "epoch": 0.9, "learning_rate": 5.06163868868077e-07, "loss": 23.2999, "step": 73290 }, { "epoch": 0.9, "learning_rate": 5.049134378423104e-07, "loss": 24.0364, "step": 73300 }, { "epoch": 0.9, "learning_rate": 5.036645132604345e-07, "loss": 24.3202, "step": 73310 }, { "epoch": 0.9, "learning_rate": 5.024170953205987e-07, "loss": 23.9682, "step": 73320 }, { "epoch": 0.9, "learning_rate": 5.011711842207134e-07, "loss": 23.1085, "step": 73330 }, { "epoch": 0.9, "learning_rate": 4.999267801584451e-07, "loss": 23.9309, "step": 73340 }, { "epoch": 0.9, "learning_rate": 4.986838833312291e-07, "loss": 23.9867, "step": 73350 }, { "epoch": 0.9, "learning_rate": 4.974424939362555e-07, "loss": 24.5886, "step": 73360 }, { "epoch": 0.9, "learning_rate": 4.962026121704755e-07, "loss": 23.4652, "step": 73370 }, { "epoch": 0.9, "learning_rate": 4.949642382306063e-07, "loss": 24.3451, "step": 73380 }, { "epoch": 0.9, "learning_rate": 4.937273723131175e-07, "loss": 24.129, "step": 73390 }, { "epoch": 0.9, "learning_rate": 4.924920146142487e-07, "loss": 24.012, "step": 73400 }, { "epoch": 0.9, "learning_rate": 4.912581653299931e-07, "loss": 23.3314, "step": 73410 }, { "epoch": 0.9, "learning_rate": 4.900258246561063e-07, "loss": 23.9532, "step": 73420 }, { "epoch": 0.9, "learning_rate": 4.887949927881086e-07, "loss": 24.3193, "step": 73430 }, { "epoch": 0.9, "learning_rate": 4.875656699212761e-07, "loss": 23.1516, "step": 73440 }, { "epoch": 0.9, "learning_rate": 4.86337856250646e-07, "loss": 23.8563, "step": 73450 }, { "epoch": 0.9, "learning_rate": 4.851115519710192e-07, "loss": 24.4562, "step": 73460 }, { "epoch": 0.9, "learning_rate": 4.838867572769545e-07, "loss": 23.7859, "step": 73470 }, { "epoch": 0.9, "learning_rate": 4.826634723627699e-07, "loss": 23.8419, "step": 73480 }, { "epoch": 0.9, "learning_rate": 4.8144169742255e-07, "loss": 23.7771, "step": 73490 }, { "epoch": 0.9, "learning_rate": 4.80221432650132e-07, "loss": 23.5188, "step": 73500 }, { "epoch": 0.9, "learning_rate": 4.790026782391166e-07, "loss": 23.2938, "step": 73510 }, { "epoch": 0.9, "learning_rate": 4.777854343828692e-07, "loss": 23.7262, "step": 73520 }, { "epoch": 0.9, "learning_rate": 4.765697012745085e-07, "loss": 24.445, "step": 73530 }, { "epoch": 0.9, "learning_rate": 4.7535547910691705e-07, "loss": 23.8635, "step": 73540 }, { "epoch": 0.9, "learning_rate": 4.741427680727395e-07, "loss": 24.0265, "step": 73550 }, { "epoch": 0.9, "learning_rate": 4.729315683643776e-07, "loss": 24.1468, "step": 73560 }, { "epoch": 0.9, "learning_rate": 4.71721880173992e-07, "loss": 23.5353, "step": 73570 }, { "epoch": 0.9, "learning_rate": 4.705137036935092e-07, "loss": 23.9984, "step": 73580 }, { "epoch": 0.91, "learning_rate": 4.6930703911461263e-07, "loss": 24.3957, "step": 73590 }, { "epoch": 0.91, "learning_rate": 4.6810188662874126e-07, "loss": 23.6847, "step": 73600 }, { "epoch": 0.91, "learning_rate": 4.668982464271055e-07, "loss": 24.4117, "step": 73610 }, { "epoch": 0.91, "learning_rate": 4.656961187006626e-07, "loss": 23.9405, "step": 73620 }, { "epoch": 0.91, "learning_rate": 4.644955036401411e-07, "loss": 24.236, "step": 73630 }, { "epoch": 0.91, "learning_rate": 4.6329640143602196e-07, "loss": 23.8176, "step": 73640 }, { "epoch": 0.91, "learning_rate": 4.620988122785475e-07, "loss": 23.5465, "step": 73650 }, { "epoch": 0.91, "learning_rate": 4.6090273635772565e-07, "loss": 23.1348, "step": 73660 }, { "epoch": 0.91, "learning_rate": 4.597081738633169e-07, "loss": 23.4177, "step": 73670 }, { "epoch": 0.91, "learning_rate": 4.585151249848452e-07, "loss": 23.7278, "step": 73680 }, { "epoch": 0.91, "learning_rate": 4.5732358991159466e-07, "loss": 24.0812, "step": 73690 }, { "epoch": 0.91, "learning_rate": 4.561335688326074e-07, "loss": 23.2706, "step": 73700 }, { "epoch": 0.91, "learning_rate": 4.5494506193668573e-07, "loss": 24.4533, "step": 73710 }, { "epoch": 0.91, "learning_rate": 4.537580694123955e-07, "loss": 24.2299, "step": 73720 }, { "epoch": 0.91, "learning_rate": 4.525725914480561e-07, "loss": 24.0283, "step": 73730 }, { "epoch": 0.91, "learning_rate": 4.5138862823174924e-07, "loss": 23.9753, "step": 73740 }, { "epoch": 0.91, "learning_rate": 4.502061799513202e-07, "loss": 23.6383, "step": 73750 }, { "epoch": 0.91, "learning_rate": 4.4902524679436677e-07, "loss": 23.9163, "step": 73760 }, { "epoch": 0.91, "learning_rate": 4.4784582894825125e-07, "loss": 23.0106, "step": 73770 }, { "epoch": 0.91, "learning_rate": 4.4666792660009506e-07, "loss": 23.1778, "step": 73780 }, { "epoch": 0.91, "learning_rate": 4.4549153993677876e-07, "loss": 23.6373, "step": 73790 }, { "epoch": 0.91, "learning_rate": 4.4431666914493966e-07, "loss": 23.9135, "step": 73800 }, { "epoch": 0.91, "learning_rate": 4.431433144109798e-07, "loss": 23.9464, "step": 73810 }, { "epoch": 0.91, "learning_rate": 4.4197147592105583e-07, "loss": 23.9308, "step": 73820 }, { "epoch": 0.91, "learning_rate": 4.408011538610857e-07, "loss": 23.3231, "step": 73830 }, { "epoch": 0.91, "learning_rate": 4.3963234841675193e-07, "loss": 23.8249, "step": 73840 }, { "epoch": 0.91, "learning_rate": 4.3846505977348296e-07, "loss": 23.782, "step": 73850 }, { "epoch": 0.91, "learning_rate": 4.372992881164828e-07, "loss": 23.5781, "step": 73860 }, { "epoch": 0.91, "learning_rate": 4.361350336307024e-07, "loss": 24.3822, "step": 73870 }, { "epoch": 0.91, "learning_rate": 4.3497229650085736e-07, "loss": 23.808, "step": 73880 }, { "epoch": 0.91, "learning_rate": 4.3381107691142453e-07, "loss": 24.3529, "step": 73890 }, { "epoch": 0.91, "learning_rate": 4.3265137504663436e-07, "loss": 24.1346, "step": 73900 }, { "epoch": 0.91, "learning_rate": 4.3149319109048074e-07, "loss": 23.5373, "step": 73910 }, { "epoch": 0.91, "learning_rate": 4.3033652522671554e-07, "loss": 24.272, "step": 73920 }, { "epoch": 0.91, "learning_rate": 4.291813776388498e-07, "loss": 23.9551, "step": 73930 }, { "epoch": 0.91, "learning_rate": 4.280277485101514e-07, "loss": 23.8584, "step": 73940 }, { "epoch": 0.91, "learning_rate": 4.2687563802365273e-07, "loss": 23.5108, "step": 73950 }, { "epoch": 0.91, "learning_rate": 4.2572504636214096e-07, "loss": 23.4854, "step": 73960 }, { "epoch": 0.91, "learning_rate": 4.2457597370816117e-07, "loss": 24.0383, "step": 73970 }, { "epoch": 0.91, "learning_rate": 4.234284202440231e-07, "loss": 23.9823, "step": 73980 }, { "epoch": 0.91, "learning_rate": 4.2228238615178996e-07, "loss": 23.8465, "step": 73990 }, { "epoch": 0.91, "learning_rate": 4.211378716132841e-07, "loss": 23.9477, "step": 74000 }, { "epoch": 0.91, "learning_rate": 4.1999487681009257e-07, "loss": 23.2763, "step": 74010 }, { "epoch": 0.91, "learning_rate": 4.1885340192355573e-07, "loss": 23.9995, "step": 74020 }, { "epoch": 0.91, "learning_rate": 4.177134471347721e-07, "loss": 23.2973, "step": 74030 }, { "epoch": 0.91, "learning_rate": 4.1657501262460596e-07, "loss": 24.1071, "step": 74040 }, { "epoch": 0.91, "learning_rate": 4.154380985736706e-07, "loss": 23.6868, "step": 74050 }, { "epoch": 0.91, "learning_rate": 4.1430270516234717e-07, "loss": 23.5723, "step": 74060 }, { "epoch": 0.91, "learning_rate": 4.1316883257077054e-07, "loss": 23.7998, "step": 74070 }, { "epoch": 0.91, "learning_rate": 4.120364809788335e-07, "loss": 23.4991, "step": 74080 }, { "epoch": 0.91, "learning_rate": 4.1090565056619236e-07, "loss": 23.5427, "step": 74090 }, { "epoch": 0.91, "learning_rate": 4.0977634151225686e-07, "loss": 23.9719, "step": 74100 }, { "epoch": 0.91, "learning_rate": 4.086485539961971e-07, "loss": 23.5356, "step": 74110 }, { "epoch": 0.91, "learning_rate": 4.075222881969454e-07, "loss": 23.2929, "step": 74120 }, { "epoch": 0.91, "learning_rate": 4.0639754429318777e-07, "loss": 24.5081, "step": 74130 }, { "epoch": 0.91, "learning_rate": 4.052743224633693e-07, "loss": 24.2271, "step": 74140 }, { "epoch": 0.91, "learning_rate": 4.0415262288569737e-07, "loss": 23.2346, "step": 74150 }, { "epoch": 0.91, "learning_rate": 4.0303244573813406e-07, "loss": 23.6064, "step": 74160 }, { "epoch": 0.91, "learning_rate": 4.019137911983995e-07, "loss": 23.6448, "step": 74170 }, { "epoch": 0.91, "learning_rate": 4.007966594439772e-07, "loss": 24.1454, "step": 74180 }, { "epoch": 0.91, "learning_rate": 3.996810506521054e-07, "loss": 23.9935, "step": 74190 }, { "epoch": 0.91, "learning_rate": 3.9856696499977697e-07, "loss": 23.6794, "step": 74200 }, { "epoch": 0.91, "learning_rate": 3.974544026637528e-07, "loss": 24.3452, "step": 74210 }, { "epoch": 0.91, "learning_rate": 3.963433638205427e-07, "loss": 23.955, "step": 74220 }, { "epoch": 0.91, "learning_rate": 3.9523384864642135e-07, "loss": 23.7201, "step": 74230 }, { "epoch": 0.91, "learning_rate": 3.941258573174178e-07, "loss": 23.9077, "step": 74240 }, { "epoch": 0.91, "learning_rate": 3.9301939000931934e-07, "loss": 23.9286, "step": 74250 }, { "epoch": 0.91, "learning_rate": 3.9191444689767435e-07, "loss": 23.6456, "step": 74260 }, { "epoch": 0.91, "learning_rate": 3.9081102815778706e-07, "loss": 23.3477, "step": 74270 }, { "epoch": 0.91, "learning_rate": 3.8970913396471964e-07, "loss": 24.1736, "step": 74280 }, { "epoch": 0.91, "learning_rate": 3.886087644932945e-07, "loss": 23.6032, "step": 74290 }, { "epoch": 0.91, "learning_rate": 3.8750991991809317e-07, "loss": 24.1293, "step": 74300 }, { "epoch": 0.91, "learning_rate": 3.864126004134461e-07, "loss": 23.9, "step": 74310 }, { "epoch": 0.91, "learning_rate": 3.853168061534551e-07, "loss": 23.725, "step": 74320 }, { "epoch": 0.91, "learning_rate": 3.842225373119712e-07, "loss": 24.1794, "step": 74330 }, { "epoch": 0.91, "learning_rate": 3.831297940626033e-07, "loss": 24.3295, "step": 74340 }, { "epoch": 0.91, "learning_rate": 3.820385765787249e-07, "loss": 23.5891, "step": 74350 }, { "epoch": 0.91, "learning_rate": 3.8094888503346083e-07, "loss": 23.5462, "step": 74360 }, { "epoch": 0.91, "learning_rate": 3.798607195996973e-07, "loss": 24.4412, "step": 74370 }, { "epoch": 0.91, "learning_rate": 3.7877408045007723e-07, "loss": 23.5717, "step": 74380 }, { "epoch": 0.91, "learning_rate": 3.776889677569995e-07, "loss": 23.8114, "step": 74390 }, { "epoch": 0.91, "learning_rate": 3.7660538169262515e-07, "loss": 24.138, "step": 74400 }, { "epoch": 0.92, "learning_rate": 3.7552332242887014e-07, "loss": 23.9059, "step": 74410 }, { "epoch": 0.92, "learning_rate": 3.7444279013740703e-07, "loss": 23.7537, "step": 74420 }, { "epoch": 0.92, "learning_rate": 3.7336378498966987e-07, "loss": 23.908, "step": 74430 }, { "epoch": 0.92, "learning_rate": 3.722863071568483e-07, "loss": 23.7538, "step": 74440 }, { "epoch": 0.92, "learning_rate": 3.712103568098868e-07, "loss": 23.5424, "step": 74450 }, { "epoch": 0.92, "learning_rate": 3.7013593411949435e-07, "loss": 23.9234, "step": 74460 }, { "epoch": 0.92, "learning_rate": 3.690630392561312e-07, "loss": 24.122, "step": 74470 }, { "epoch": 0.92, "learning_rate": 3.679916723900168e-07, "loss": 23.743, "step": 74480 }, { "epoch": 0.92, "learning_rate": 3.6692183369113175e-07, "loss": 24.1874, "step": 74490 }, { "epoch": 0.92, "learning_rate": 3.6585352332921045e-07, "loss": 23.5773, "step": 74500 }, { "epoch": 0.92, "learning_rate": 3.6478674147374273e-07, "loss": 23.9777, "step": 74510 }, { "epoch": 0.92, "learning_rate": 3.6372148829398325e-07, "loss": 23.5462, "step": 74520 }, { "epoch": 0.92, "learning_rate": 3.6265776395893905e-07, "loss": 23.679, "step": 74530 }, { "epoch": 0.92, "learning_rate": 3.615955686373729e-07, "loss": 23.5153, "step": 74540 }, { "epoch": 0.92, "learning_rate": 3.6053490249781e-07, "loss": 23.4519, "step": 74550 }, { "epoch": 0.92, "learning_rate": 3.5947576570852906e-07, "loss": 24.0555, "step": 74560 }, { "epoch": 0.92, "learning_rate": 3.584181584375679e-07, "loss": 23.7136, "step": 74570 }, { "epoch": 0.92, "learning_rate": 3.573620808527212e-07, "loss": 24.5501, "step": 74580 }, { "epoch": 0.92, "learning_rate": 3.5630753312154154e-07, "loss": 23.6315, "step": 74590 }, { "epoch": 0.92, "learning_rate": 3.552545154113374e-07, "loss": 24.3578, "step": 74600 }, { "epoch": 0.92, "learning_rate": 3.542030278891773e-07, "loss": 23.8807, "step": 74610 }, { "epoch": 0.92, "learning_rate": 3.531530707218822e-07, "loss": 24.5827, "step": 74620 }, { "epoch": 0.92, "learning_rate": 3.521046440760356e-07, "loss": 23.4304, "step": 74630 }, { "epoch": 0.92, "learning_rate": 3.5105774811797555e-07, "loss": 23.7996, "step": 74640 }, { "epoch": 0.92, "learning_rate": 3.500123830137947e-07, "loss": 23.3045, "step": 74650 }, { "epoch": 0.92, "learning_rate": 3.4896854892934926e-07, "loss": 23.6731, "step": 74660 }, { "epoch": 0.92, "learning_rate": 3.479262460302479e-07, "loss": 22.9679, "step": 74670 }, { "epoch": 0.92, "learning_rate": 3.4688547448185396e-07, "loss": 24.0653, "step": 74680 }, { "epoch": 0.92, "learning_rate": 3.4584623444929633e-07, "loss": 23.1485, "step": 74690 }, { "epoch": 0.92, "learning_rate": 3.4480852609745207e-07, "loss": 23.9349, "step": 74700 }, { "epoch": 0.92, "learning_rate": 3.437723495909595e-07, "loss": 24.1546, "step": 74710 }, { "epoch": 0.92, "learning_rate": 3.427377050942149e-07, "loss": 23.6682, "step": 74720 }, { "epoch": 0.92, "learning_rate": 3.417045927713691e-07, "loss": 23.5322, "step": 74730 }, { "epoch": 0.92, "learning_rate": 3.406730127863289e-07, "loss": 24.1809, "step": 74740 }, { "epoch": 0.92, "learning_rate": 3.396429653027622e-07, "loss": 23.5564, "step": 74750 }, { "epoch": 0.92, "learning_rate": 3.386144504840905e-07, "loss": 24.167, "step": 74760 }, { "epoch": 0.92, "learning_rate": 3.375874684934921e-07, "loss": 23.535, "step": 74770 }, { "epoch": 0.92, "learning_rate": 3.3656201949390344e-07, "loss": 23.3521, "step": 74780 }, { "epoch": 0.92, "learning_rate": 3.355381036480165e-07, "loss": 24.2348, "step": 74790 }, { "epoch": 0.92, "learning_rate": 3.345157211182837e-07, "loss": 24.1364, "step": 74800 }, { "epoch": 0.92, "learning_rate": 3.3349487206690734e-07, "loss": 23.602, "step": 74810 }, { "epoch": 0.92, "learning_rate": 3.324755566558524e-07, "loss": 23.6881, "step": 74820 }, { "epoch": 0.92, "learning_rate": 3.314577750468384e-07, "loss": 23.6461, "step": 74830 }, { "epoch": 0.92, "learning_rate": 3.304415274013417e-07, "loss": 23.9872, "step": 74840 }, { "epoch": 0.92, "learning_rate": 3.294268138805934e-07, "loss": 23.8615, "step": 74850 }, { "epoch": 0.92, "learning_rate": 3.284136346455857e-07, "loss": 23.2922, "step": 74860 }, { "epoch": 0.92, "learning_rate": 3.274019898570635e-07, "loss": 24.079, "step": 74870 }, { "epoch": 0.92, "learning_rate": 3.2639187967552724e-07, "loss": 24.2677, "step": 74880 }, { "epoch": 0.92, "learning_rate": 3.2538330426123977e-07, "loss": 23.4599, "step": 74890 }, { "epoch": 0.92, "learning_rate": 3.2437626377421537e-07, "loss": 24.4137, "step": 74900 }, { "epoch": 0.92, "learning_rate": 3.2337075837422404e-07, "loss": 23.5393, "step": 74910 }, { "epoch": 0.92, "learning_rate": 3.22366788220797e-07, "loss": 23.6674, "step": 74920 }, { "epoch": 0.92, "learning_rate": 3.213643534732191e-07, "loss": 23.5437, "step": 74930 }, { "epoch": 0.92, "learning_rate": 3.2036345429052873e-07, "loss": 23.3258, "step": 74940 }, { "epoch": 0.92, "learning_rate": 3.193640908315276e-07, "loss": 24.131, "step": 74950 }, { "epoch": 0.92, "learning_rate": 3.1836626325476793e-07, "loss": 24.1318, "step": 74960 }, { "epoch": 0.92, "learning_rate": 3.1736997171855964e-07, "loss": 24.2004, "step": 74970 }, { "epoch": 0.92, "learning_rate": 3.163752163809708e-07, "loss": 23.9103, "step": 74980 }, { "epoch": 0.92, "learning_rate": 3.153819973998229e-07, "loss": 23.626, "step": 74990 }, { "epoch": 0.92, "learning_rate": 3.1439031493269765e-07, "loss": 24.2388, "step": 75000 }, { "epoch": 0.92, "learning_rate": 3.134001691369271e-07, "loss": 23.8188, "step": 75010 }, { "epoch": 0.92, "learning_rate": 3.124115601696054e-07, "loss": 24.0907, "step": 75020 }, { "epoch": 0.92, "learning_rate": 3.114244881875794e-07, "loss": 24.2008, "step": 75030 }, { "epoch": 0.92, "learning_rate": 3.104389533474539e-07, "loss": 23.6604, "step": 75040 }, { "epoch": 0.92, "learning_rate": 3.0945495580558814e-07, "loss": 23.6798, "step": 75050 }, { "epoch": 0.92, "learning_rate": 3.0847249571809845e-07, "loss": 23.8605, "step": 75060 }, { "epoch": 0.92, "learning_rate": 3.074915732408579e-07, "loss": 24.0932, "step": 75070 }, { "epoch": 0.92, "learning_rate": 3.0651218852949417e-07, "loss": 23.1953, "step": 75080 }, { "epoch": 0.92, "learning_rate": 3.055343417393919e-07, "loss": 23.7534, "step": 75090 }, { "epoch": 0.92, "learning_rate": 3.0455803302569255e-07, "loss": 23.511, "step": 75100 }, { "epoch": 0.92, "learning_rate": 3.0358326254328884e-07, "loss": 23.8762, "step": 75110 }, { "epoch": 0.92, "learning_rate": 3.0261003044683713e-07, "loss": 24.184, "step": 75120 }, { "epoch": 0.92, "learning_rate": 3.01638336890745e-07, "loss": 24.0411, "step": 75130 }, { "epoch": 0.92, "learning_rate": 3.006681820291735e-07, "loss": 23.4018, "step": 75140 }, { "epoch": 0.92, "learning_rate": 2.996995660160462e-07, "loss": 24.3062, "step": 75150 }, { "epoch": 0.92, "learning_rate": 2.9873248900503805e-07, "loss": 24.0716, "step": 75160 }, { "epoch": 0.92, "learning_rate": 2.9776695114957845e-07, "loss": 23.9594, "step": 75170 }, { "epoch": 0.92, "learning_rate": 2.968029526028582e-07, "loss": 23.6124, "step": 75180 }, { "epoch": 0.92, "learning_rate": 2.9584049351781943e-07, "loss": 23.829, "step": 75190 }, { "epoch": 0.92, "learning_rate": 2.9487957404715993e-07, "loss": 23.5219, "step": 75200 }, { "epoch": 0.92, "learning_rate": 2.9392019434333564e-07, "loss": 23.8496, "step": 75210 }, { "epoch": 0.93, "learning_rate": 2.9296235455855805e-07, "loss": 24.6198, "step": 75220 }, { "epoch": 0.93, "learning_rate": 2.920060548447912e-07, "loss": 23.3598, "step": 75230 }, { "epoch": 0.93, "learning_rate": 2.910512953537581e-07, "loss": 23.791, "step": 75240 }, { "epoch": 0.93, "learning_rate": 2.900980762369343e-07, "loss": 23.7227, "step": 75250 }, { "epoch": 0.93, "learning_rate": 2.8914639764555554e-07, "loss": 24.1321, "step": 75260 }, { "epoch": 0.93, "learning_rate": 2.8819625973060984e-07, "loss": 23.683, "step": 75270 }, { "epoch": 0.93, "learning_rate": 2.8724766264284e-07, "loss": 23.8476, "step": 75280 }, { "epoch": 0.93, "learning_rate": 2.863006065327478e-07, "loss": 23.7796, "step": 75290 }, { "epoch": 0.93, "learning_rate": 2.853550915505876e-07, "loss": 23.9302, "step": 75300 }, { "epoch": 0.93, "learning_rate": 2.844111178463693e-07, "loss": 22.9261, "step": 75310 }, { "epoch": 0.93, "learning_rate": 2.83468685569861e-07, "loss": 23.8378, "step": 75320 }, { "epoch": 0.93, "learning_rate": 2.825277948705829e-07, "loss": 23.906, "step": 75330 }, { "epoch": 0.93, "learning_rate": 2.815884458978113e-07, "loss": 24.1348, "step": 75340 }, { "epoch": 0.93, "learning_rate": 2.806506388005814e-07, "loss": 23.9679, "step": 75350 }, { "epoch": 0.93, "learning_rate": 2.7971437372767975e-07, "loss": 23.0683, "step": 75360 }, { "epoch": 0.93, "learning_rate": 2.7877965082764856e-07, "loss": 23.4682, "step": 75370 }, { "epoch": 0.93, "learning_rate": 2.778464702487882e-07, "loss": 23.8662, "step": 75380 }, { "epoch": 0.93, "learning_rate": 2.769148321391513e-07, "loss": 23.9105, "step": 75390 }, { "epoch": 0.93, "learning_rate": 2.759847366465462e-07, "loss": 23.6947, "step": 75400 }, { "epoch": 0.93, "learning_rate": 2.750561839185395e-07, "loss": 23.9255, "step": 75410 }, { "epoch": 0.93, "learning_rate": 2.7412917410244985e-07, "loss": 23.6321, "step": 75420 }, { "epoch": 0.93, "learning_rate": 2.7320370734535083e-07, "loss": 24.0293, "step": 75430 }, { "epoch": 0.93, "learning_rate": 2.722797837940738e-07, "loss": 23.1516, "step": 75440 }, { "epoch": 0.93, "learning_rate": 2.7135740359520377e-07, "loss": 23.7138, "step": 75450 }, { "epoch": 0.93, "learning_rate": 2.704365668950826e-07, "loss": 23.9178, "step": 75460 }, { "epoch": 0.93, "learning_rate": 2.695172738398033e-07, "loss": 23.6552, "step": 75470 }, { "epoch": 0.93, "learning_rate": 2.68599524575216e-07, "loss": 23.7184, "step": 75480 }, { "epoch": 0.93, "learning_rate": 2.676833192469286e-07, "loss": 23.4259, "step": 75490 }, { "epoch": 0.93, "learning_rate": 2.667686580003004e-07, "loss": 24.3106, "step": 75500 }, { "epoch": 0.93, "learning_rate": 2.658555409804475e-07, "loss": 24.1868, "step": 75510 }, { "epoch": 0.93, "learning_rate": 2.6494396833223966e-07, "loss": 23.8361, "step": 75520 }, { "epoch": 0.93, "learning_rate": 2.6403394020030446e-07, "loss": 24.229, "step": 75530 }, { "epoch": 0.93, "learning_rate": 2.631254567290198e-07, "loss": 23.555, "step": 75540 }, { "epoch": 0.93, "learning_rate": 2.6221851806252474e-07, "loss": 24.4062, "step": 75550 }, { "epoch": 0.93, "learning_rate": 2.6131312434470647e-07, "loss": 23.6679, "step": 75560 }, { "epoch": 0.93, "learning_rate": 2.6040927571921113e-07, "loss": 23.6063, "step": 75570 }, { "epoch": 0.93, "learning_rate": 2.5950697232944076e-07, "loss": 23.9086, "step": 75580 }, { "epoch": 0.93, "learning_rate": 2.586062143185486e-07, "loss": 23.9747, "step": 75590 }, { "epoch": 0.93, "learning_rate": 2.5770700182944364e-07, "loss": 23.2549, "step": 75600 }, { "epoch": 0.93, "learning_rate": 2.568093350047929e-07, "loss": 23.257, "step": 75610 }, { "epoch": 0.93, "learning_rate": 2.5591321398701353e-07, "loss": 23.9952, "step": 75620 }, { "epoch": 0.93, "learning_rate": 2.550186389182818e-07, "loss": 23.3024, "step": 75630 }, { "epoch": 0.93, "learning_rate": 2.541256099405265e-07, "loss": 23.6823, "step": 75640 }, { "epoch": 0.93, "learning_rate": 2.532341271954286e-07, "loss": 23.9931, "step": 75650 }, { "epoch": 0.93, "learning_rate": 2.523441908244295e-07, "loss": 23.8433, "step": 75660 }, { "epoch": 0.93, "learning_rate": 2.5145580096871957e-07, "loss": 23.8416, "step": 75670 }, { "epoch": 0.93, "learning_rate": 2.505689577692483e-07, "loss": 23.7756, "step": 75680 }, { "epoch": 0.93, "learning_rate": 2.496836613667186e-07, "loss": 24.08, "step": 75690 }, { "epoch": 0.93, "learning_rate": 2.4879991190158494e-07, "loss": 24.1293, "step": 75700 }, { "epoch": 0.93, "learning_rate": 2.479177095140584e-07, "loss": 24.0657, "step": 75710 }, { "epoch": 0.93, "learning_rate": 2.4703705434410695e-07, "loss": 23.267, "step": 75720 }, { "epoch": 0.93, "learning_rate": 2.461579465314501e-07, "loss": 23.6064, "step": 75730 }, { "epoch": 0.93, "learning_rate": 2.4528038621556285e-07, "loss": 23.6751, "step": 75740 }, { "epoch": 0.93, "learning_rate": 2.444043735356738e-07, "loss": 23.7621, "step": 75750 }, { "epoch": 0.93, "learning_rate": 2.4352990863076854e-07, "loss": 23.6985, "step": 75760 }, { "epoch": 0.93, "learning_rate": 2.426569916395849e-07, "loss": 23.8304, "step": 75770 }, { "epoch": 0.93, "learning_rate": 2.4178562270061435e-07, "loss": 24.351, "step": 75780 }, { "epoch": 0.93, "learning_rate": 2.409158019521041e-07, "loss": 23.8007, "step": 75790 }, { "epoch": 0.93, "learning_rate": 2.400475295320581e-07, "loss": 23.6026, "step": 75800 }, { "epoch": 0.93, "learning_rate": 2.3918080557823076e-07, "loss": 23.2785, "step": 75810 }, { "epoch": 0.93, "learning_rate": 2.3831563022812975e-07, "loss": 23.6494, "step": 75820 }, { "epoch": 0.93, "learning_rate": 2.374520036190242e-07, "loss": 23.6736, "step": 75830 }, { "epoch": 0.93, "learning_rate": 2.3658992588792894e-07, "loss": 23.5041, "step": 75840 }, { "epoch": 0.93, "learning_rate": 2.3572939717161903e-07, "loss": 23.9937, "step": 75850 }, { "epoch": 0.93, "learning_rate": 2.3487041760662077e-07, "loss": 23.6286, "step": 75860 }, { "epoch": 0.93, "learning_rate": 2.340129873292174e-07, "loss": 23.7904, "step": 75870 }, { "epoch": 0.93, "learning_rate": 2.331571064754423e-07, "loss": 23.9041, "step": 75880 }, { "epoch": 0.93, "learning_rate": 2.323027751810869e-07, "loss": 23.8681, "step": 75890 }, { "epoch": 0.93, "learning_rate": 2.3144999358169384e-07, "loss": 23.5957, "step": 75900 }, { "epoch": 0.93, "learning_rate": 2.305987618125627e-07, "loss": 23.5247, "step": 75910 }, { "epoch": 0.93, "learning_rate": 2.297490800087454e-07, "loss": 23.5866, "step": 75920 }, { "epoch": 0.93, "learning_rate": 2.289009483050475e-07, "loss": 23.7427, "step": 75930 }, { "epoch": 0.93, "learning_rate": 2.2805436683603022e-07, "loss": 23.6983, "step": 75940 }, { "epoch": 0.93, "learning_rate": 2.2720933573600835e-07, "loss": 24.2008, "step": 75950 }, { "epoch": 0.93, "learning_rate": 2.2636585513904796e-07, "loss": 23.7406, "step": 75960 }, { "epoch": 0.93, "learning_rate": 2.2552392517897536e-07, "loss": 23.9395, "step": 75970 }, { "epoch": 0.93, "learning_rate": 2.246835459893637e-07, "loss": 24.2442, "step": 75980 }, { "epoch": 0.93, "learning_rate": 2.238447177035452e-07, "loss": 23.3629, "step": 75990 }, { "epoch": 0.93, "learning_rate": 2.2300744045460342e-07, "loss": 23.6058, "step": 76000 }, { "epoch": 0.93, "learning_rate": 2.2217171437537878e-07, "loss": 23.5304, "step": 76010 }, { "epoch": 0.93, "learning_rate": 2.213375395984596e-07, "loss": 23.9279, "step": 76020 }, { "epoch": 0.94, "learning_rate": 2.2050491625619453e-07, "loss": 23.596, "step": 76030 }, { "epoch": 0.94, "learning_rate": 2.1967384448068453e-07, "loss": 24.0843, "step": 76040 }, { "epoch": 0.94, "learning_rate": 2.1884432440377966e-07, "loss": 23.7314, "step": 76050 }, { "epoch": 0.94, "learning_rate": 2.1801635615709137e-07, "loss": 23.8936, "step": 76060 }, { "epoch": 0.94, "learning_rate": 2.1718993987198012e-07, "loss": 23.9802, "step": 76070 }, { "epoch": 0.94, "learning_rate": 2.1636507567955878e-07, "loss": 23.7787, "step": 76080 }, { "epoch": 0.94, "learning_rate": 2.1554176371069823e-07, "loss": 23.8186, "step": 76090 }, { "epoch": 0.94, "learning_rate": 2.147200040960218e-07, "loss": 23.9747, "step": 76100 }, { "epoch": 0.94, "learning_rate": 2.138997969659029e-07, "loss": 23.9335, "step": 76110 }, { "epoch": 0.94, "learning_rate": 2.130811424504753e-07, "loss": 23.3386, "step": 76120 }, { "epoch": 0.94, "learning_rate": 2.1226404067962058e-07, "loss": 23.9268, "step": 76130 }, { "epoch": 0.94, "learning_rate": 2.1144849178297622e-07, "loss": 23.8479, "step": 76140 }, { "epoch": 0.94, "learning_rate": 2.1063449588993313e-07, "loss": 23.7022, "step": 76150 }, { "epoch": 0.94, "learning_rate": 2.0982205312963688e-07, "loss": 23.4895, "step": 76160 }, { "epoch": 0.94, "learning_rate": 2.0901116363098438e-07, "loss": 23.4021, "step": 76170 }, { "epoch": 0.94, "learning_rate": 2.082018275226283e-07, "loss": 23.8982, "step": 76180 }, { "epoch": 0.94, "learning_rate": 2.0739404493297255e-07, "loss": 24.1892, "step": 76190 }, { "epoch": 0.94, "learning_rate": 2.0658781599017796e-07, "loss": 23.4638, "step": 76200 }, { "epoch": 0.94, "learning_rate": 2.0578314082215666e-07, "loss": 23.4453, "step": 76210 }, { "epoch": 0.94, "learning_rate": 2.049800195565721e-07, "loss": 23.6217, "step": 76220 }, { "epoch": 0.94, "learning_rate": 2.0417845232084675e-07, "loss": 23.6667, "step": 76230 }, { "epoch": 0.94, "learning_rate": 2.0337843924215117e-07, "loss": 23.6705, "step": 76240 }, { "epoch": 0.94, "learning_rate": 2.0257998044741046e-07, "loss": 24.1357, "step": 76250 }, { "epoch": 0.94, "learning_rate": 2.017830760633077e-07, "loss": 24.1617, "step": 76260 }, { "epoch": 0.94, "learning_rate": 2.0098772621627405e-07, "loss": 23.3861, "step": 76270 }, { "epoch": 0.94, "learning_rate": 2.001939310324952e-07, "loss": 23.7038, "step": 76280 }, { "epoch": 0.94, "learning_rate": 1.9940169063791149e-07, "loss": 23.683, "step": 76290 }, { "epoch": 0.94, "learning_rate": 1.9861100515821574e-07, "loss": 23.7661, "step": 76300 }, { "epoch": 0.94, "learning_rate": 1.9782187471885317e-07, "loss": 23.7004, "step": 76310 }, { "epoch": 0.94, "learning_rate": 1.9703429944502583e-07, "loss": 23.5873, "step": 76320 }, { "epoch": 0.94, "learning_rate": 1.962482794616849e-07, "loss": 23.542, "step": 76330 }, { "epoch": 0.94, "learning_rate": 1.9546381489353616e-07, "loss": 24.2525, "step": 76340 }, { "epoch": 0.94, "learning_rate": 1.9468090586504007e-07, "loss": 24.1521, "step": 76350 }, { "epoch": 0.94, "learning_rate": 1.9389955250040838e-07, "loss": 24.4291, "step": 76360 }, { "epoch": 0.94, "learning_rate": 1.9311975492360635e-07, "loss": 23.7823, "step": 76370 }, { "epoch": 0.94, "learning_rate": 1.9234151325835394e-07, "loss": 23.6512, "step": 76380 }, { "epoch": 0.94, "learning_rate": 1.9156482762812344e-07, "loss": 23.7589, "step": 76390 }, { "epoch": 0.94, "learning_rate": 1.9078969815613858e-07, "loss": 24.7016, "step": 76400 }, { "epoch": 0.94, "learning_rate": 1.9001612496537758e-07, "loss": 23.8977, "step": 76410 }, { "epoch": 0.94, "learning_rate": 1.892441081785723e-07, "loss": 23.6846, "step": 76420 }, { "epoch": 0.94, "learning_rate": 1.884736479182081e-07, "loss": 23.8324, "step": 76430 }, { "epoch": 0.94, "learning_rate": 1.8770474430652163e-07, "loss": 23.7648, "step": 76440 }, { "epoch": 0.94, "learning_rate": 1.86937397465502e-07, "loss": 23.9344, "step": 76450 }, { "epoch": 0.94, "learning_rate": 1.8617160751689512e-07, "loss": 23.8563, "step": 76460 }, { "epoch": 0.94, "learning_rate": 1.8540737458219605e-07, "loss": 23.328, "step": 76470 }, { "epoch": 0.94, "learning_rate": 1.8464469878265446e-07, "loss": 24.2673, "step": 76480 }, { "epoch": 0.94, "learning_rate": 1.8388358023927245e-07, "loss": 23.8658, "step": 76490 }, { "epoch": 0.94, "learning_rate": 1.8312401907280675e-07, "loss": 24.1432, "step": 76500 }, { "epoch": 0.94, "learning_rate": 1.823660154037632e-07, "loss": 24.1012, "step": 76510 }, { "epoch": 0.94, "learning_rate": 1.816095693524056e-07, "loss": 23.8508, "step": 76520 }, { "epoch": 0.94, "learning_rate": 1.808546810387468e-07, "loss": 23.6893, "step": 76530 }, { "epoch": 0.94, "learning_rate": 1.8010135058255218e-07, "loss": 24.5351, "step": 76540 }, { "epoch": 0.94, "learning_rate": 1.7934957810334496e-07, "loss": 23.604, "step": 76550 }, { "epoch": 0.94, "learning_rate": 1.785993637203942e-07, "loss": 23.7244, "step": 76560 }, { "epoch": 0.94, "learning_rate": 1.7785070755272692e-07, "loss": 23.1496, "step": 76570 }, { "epoch": 0.94, "learning_rate": 1.7710360971912034e-07, "loss": 22.9869, "step": 76580 }, { "epoch": 0.94, "learning_rate": 1.7635807033810737e-07, "loss": 23.3388, "step": 76590 }, { "epoch": 0.94, "learning_rate": 1.7561408952796898e-07, "loss": 23.4875, "step": 76600 }, { "epoch": 0.94, "learning_rate": 1.7487166740674299e-07, "loss": 23.8534, "step": 76610 }, { "epoch": 0.94, "learning_rate": 1.741308040922185e-07, "loss": 23.4624, "step": 76620 }, { "epoch": 0.94, "learning_rate": 1.7339149970193703e-07, "loss": 23.5298, "step": 76630 }, { "epoch": 0.94, "learning_rate": 1.7265375435319364e-07, "loss": 23.5419, "step": 76640 }, { "epoch": 0.94, "learning_rate": 1.7191756816303252e-07, "loss": 23.0734, "step": 76650 }, { "epoch": 0.94, "learning_rate": 1.7118294124825686e-07, "loss": 23.8478, "step": 76660 }, { "epoch": 0.94, "learning_rate": 1.7044987372541787e-07, "loss": 23.934, "step": 76670 }, { "epoch": 0.94, "learning_rate": 1.6971836571081924e-07, "loss": 24.2382, "step": 76680 }, { "epoch": 0.94, "learning_rate": 1.689884173205203e-07, "loss": 24.0437, "step": 76690 }, { "epoch": 0.94, "learning_rate": 1.6826002867033065e-07, "loss": 24.2791, "step": 76700 }, { "epoch": 0.94, "learning_rate": 1.6753319987581119e-07, "loss": 23.8985, "step": 76710 }, { "epoch": 0.94, "learning_rate": 1.6680793105227854e-07, "loss": 24.1047, "step": 76720 }, { "epoch": 0.94, "learning_rate": 1.660842223148007e-07, "loss": 23.7731, "step": 76730 }, { "epoch": 0.94, "learning_rate": 1.653620737781958e-07, "loss": 23.9552, "step": 76740 }, { "epoch": 0.94, "learning_rate": 1.6464148555703884e-07, "loss": 23.7612, "step": 76750 }, { "epoch": 0.94, "learning_rate": 1.6392245776565396e-07, "loss": 23.7551, "step": 76760 }, { "epoch": 0.94, "learning_rate": 1.6320499051811656e-07, "loss": 24.027, "step": 76770 }, { "epoch": 0.94, "learning_rate": 1.624890839282589e-07, "loss": 23.9776, "step": 76780 }, { "epoch": 0.94, "learning_rate": 1.6177473810966348e-07, "loss": 23.9253, "step": 76790 }, { "epoch": 0.94, "learning_rate": 1.6106195317566188e-07, "loss": 23.7393, "step": 76800 }, { "epoch": 0.94, "learning_rate": 1.603507292393447e-07, "loss": 23.9335, "step": 76810 }, { "epoch": 0.94, "learning_rate": 1.5964106641354948e-07, "loss": 22.8645, "step": 76820 }, { "epoch": 0.94, "learning_rate": 1.5893296481086616e-07, "loss": 24.1055, "step": 76830 }, { "epoch": 0.95, "learning_rate": 1.5822642454364267e-07, "loss": 23.9724, "step": 76840 }, { "epoch": 0.95, "learning_rate": 1.5752144572397266e-07, "loss": 23.9705, "step": 76850 }, { "epoch": 0.95, "learning_rate": 1.5681802846370442e-07, "loss": 24.0942, "step": 76860 }, { "epoch": 0.95, "learning_rate": 1.5611617287443982e-07, "loss": 23.3055, "step": 76870 }, { "epoch": 0.95, "learning_rate": 1.5541587906752976e-07, "loss": 24.0161, "step": 76880 }, { "epoch": 0.95, "learning_rate": 1.5471714715408315e-07, "loss": 23.5212, "step": 76890 }, { "epoch": 0.95, "learning_rate": 1.5401997724495353e-07, "loss": 23.9713, "step": 76900 }, { "epoch": 0.95, "learning_rate": 1.5332436945075247e-07, "loss": 24.1777, "step": 76910 }, { "epoch": 0.95, "learning_rate": 1.5263032388184163e-07, "loss": 23.9321, "step": 76920 }, { "epoch": 0.95, "learning_rate": 1.5193784064833405e-07, "loss": 24.328, "step": 76930 }, { "epoch": 0.95, "learning_rate": 1.5124691986009631e-07, "loss": 23.9846, "step": 76940 }, { "epoch": 0.95, "learning_rate": 1.505575616267474e-07, "loss": 24.1235, "step": 76950 }, { "epoch": 0.95, "learning_rate": 1.4986976605765537e-07, "loss": 24.1526, "step": 76960 }, { "epoch": 0.95, "learning_rate": 1.49183533261944e-07, "loss": 23.6068, "step": 76970 }, { "epoch": 0.95, "learning_rate": 1.484988633484874e-07, "loss": 24.2135, "step": 76980 }, { "epoch": 0.95, "learning_rate": 1.4781575642591196e-07, "loss": 23.3968, "step": 76990 }, { "epoch": 0.95, "learning_rate": 1.471342126025954e-07, "loss": 23.7765, "step": 77000 }, { "epoch": 0.95, "learning_rate": 1.464542319866691e-07, "loss": 24.2578, "step": 77010 }, { "epoch": 0.95, "learning_rate": 1.4577581468601443e-07, "loss": 23.6204, "step": 77020 }, { "epoch": 0.95, "learning_rate": 1.4509896080826757e-07, "loss": 24.104, "step": 77030 }, { "epoch": 0.95, "learning_rate": 1.4442367046081262e-07, "loss": 23.937, "step": 77040 }, { "epoch": 0.95, "learning_rate": 1.4374994375078833e-07, "loss": 24.4726, "step": 77050 }, { "epoch": 0.95, "learning_rate": 1.4307778078508694e-07, "loss": 23.6187, "step": 77060 }, { "epoch": 0.95, "learning_rate": 1.4240718167034874e-07, "loss": 24.1552, "step": 77070 }, { "epoch": 0.95, "learning_rate": 1.4173814651296746e-07, "loss": 24.1338, "step": 77080 }, { "epoch": 0.95, "learning_rate": 1.4107067541908937e-07, "loss": 24.1253, "step": 77090 }, { "epoch": 0.95, "learning_rate": 1.4040476849461303e-07, "loss": 24.5918, "step": 77100 }, { "epoch": 0.95, "learning_rate": 1.397404258451851e-07, "loss": 24.1452, "step": 77110 }, { "epoch": 0.95, "learning_rate": 1.3907764757621122e-07, "loss": 23.7934, "step": 77120 }, { "epoch": 0.95, "learning_rate": 1.3841643379284065e-07, "loss": 23.9674, "step": 77130 }, { "epoch": 0.95, "learning_rate": 1.3775678459998053e-07, "loss": 23.9558, "step": 77140 }, { "epoch": 0.95, "learning_rate": 1.370987001022872e-07, "loss": 23.8724, "step": 77150 }, { "epoch": 0.95, "learning_rate": 1.3644218040416936e-07, "loss": 24.0507, "step": 77160 }, { "epoch": 0.95, "learning_rate": 1.3578722560978587e-07, "loss": 23.6548, "step": 77170 }, { "epoch": 0.95, "learning_rate": 1.3513383582305028e-07, "loss": 24.0045, "step": 77180 }, { "epoch": 0.95, "learning_rate": 1.3448201114762417e-07, "loss": 24.4648, "step": 77190 }, { "epoch": 0.95, "learning_rate": 1.3383175168692474e-07, "loss": 24.3157, "step": 77200 }, { "epoch": 0.95, "learning_rate": 1.3318305754411954e-07, "loss": 23.4197, "step": 77210 }, { "epoch": 0.95, "learning_rate": 1.3253592882212506e-07, "loss": 24.2072, "step": 77220 }, { "epoch": 0.95, "learning_rate": 1.3189036562361257e-07, "loss": 23.4446, "step": 77230 }, { "epoch": 0.95, "learning_rate": 1.3124636805100456e-07, "loss": 24.2796, "step": 77240 }, { "epoch": 0.95, "learning_rate": 1.306039362064726e-07, "loss": 24.0057, "step": 77250 }, { "epoch": 0.95, "learning_rate": 1.299630701919441e-07, "loss": 24.2054, "step": 77260 }, { "epoch": 0.95, "learning_rate": 1.2932377010909547e-07, "loss": 23.524, "step": 77270 }, { "epoch": 0.95, "learning_rate": 1.2868603605935225e-07, "loss": 23.5668, "step": 77280 }, { "epoch": 0.95, "learning_rate": 1.2804986814389796e-07, "loss": 24.2727, "step": 77290 }, { "epoch": 0.95, "learning_rate": 1.2741526646366077e-07, "loss": 23.9849, "step": 77300 }, { "epoch": 0.95, "learning_rate": 1.2678223111932563e-07, "loss": 23.1039, "step": 77310 }, { "epoch": 0.95, "learning_rate": 1.261507622113256e-07, "loss": 23.4287, "step": 77320 }, { "epoch": 0.95, "learning_rate": 1.255208598398472e-07, "loss": 23.2115, "step": 77330 }, { "epoch": 0.95, "learning_rate": 1.248925241048271e-07, "loss": 23.6524, "step": 77340 }, { "epoch": 0.95, "learning_rate": 1.242657551059534e-07, "loss": 23.4656, "step": 77350 }, { "epoch": 0.95, "learning_rate": 1.2364055294266652e-07, "loss": 23.6802, "step": 77360 }, { "epoch": 0.95, "learning_rate": 1.2301691771416046e-07, "loss": 23.8431, "step": 77370 }, { "epoch": 0.95, "learning_rate": 1.2239484951937497e-07, "loss": 23.858, "step": 77380 }, { "epoch": 0.95, "learning_rate": 1.2177434845700443e-07, "loss": 24.0938, "step": 77390 }, { "epoch": 0.95, "learning_rate": 1.2115541462549563e-07, "loss": 24.15, "step": 77400 }, { "epoch": 0.95, "learning_rate": 1.205380481230467e-07, "loss": 23.7791, "step": 77410 }, { "epoch": 0.95, "learning_rate": 1.1992224904760374e-07, "loss": 24.561, "step": 77420 }, { "epoch": 0.95, "learning_rate": 1.193080174968675e-07, "loss": 23.5651, "step": 77430 }, { "epoch": 0.95, "learning_rate": 1.1869535356828887e-07, "loss": 23.6167, "step": 77440 }, { "epoch": 0.95, "learning_rate": 1.1808425735907014e-07, "loss": 23.834, "step": 77450 }, { "epoch": 0.95, "learning_rate": 1.1747472896616485e-07, "loss": 23.1235, "step": 77460 }, { "epoch": 0.95, "learning_rate": 1.1686676848627787e-07, "loss": 23.6526, "step": 77470 }, { "epoch": 0.95, "learning_rate": 1.1626037601586426e-07, "loss": 23.5913, "step": 77480 }, { "epoch": 0.95, "learning_rate": 1.1565555165113262e-07, "loss": 24.1834, "step": 77490 }, { "epoch": 0.95, "learning_rate": 1.1505229548804064e-07, "loss": 23.693, "step": 77500 }, { "epoch": 0.95, "learning_rate": 1.1445060762229843e-07, "loss": 24.019, "step": 77510 }, { "epoch": 0.95, "learning_rate": 1.1385048814936628e-07, "loss": 23.093, "step": 77520 }, { "epoch": 0.95, "learning_rate": 1.1325193716445692e-07, "loss": 23.1658, "step": 77530 }, { "epoch": 0.95, "learning_rate": 1.126549547625333e-07, "loss": 24.0326, "step": 77540 }, { "epoch": 0.95, "learning_rate": 1.1205954103830962e-07, "loss": 23.4698, "step": 77550 }, { "epoch": 0.95, "learning_rate": 1.1146569608625036e-07, "loss": 23.9995, "step": 77560 }, { "epoch": 0.95, "learning_rate": 1.1087342000057345e-07, "loss": 23.9776, "step": 77570 }, { "epoch": 0.95, "learning_rate": 1.1028271287524594e-07, "loss": 23.5902, "step": 77580 }, { "epoch": 0.95, "learning_rate": 1.0969357480398624e-07, "loss": 23.48, "step": 77590 }, { "epoch": 0.95, "learning_rate": 1.091060058802651e-07, "loss": 24.2368, "step": 77600 }, { "epoch": 0.95, "learning_rate": 1.085200061973024e-07, "loss": 23.5654, "step": 77610 }, { "epoch": 0.95, "learning_rate": 1.0793557584807046e-07, "loss": 23.6176, "step": 77620 }, { "epoch": 0.95, "learning_rate": 1.0735271492529175e-07, "loss": 24.0452, "step": 77630 }, { "epoch": 0.95, "learning_rate": 1.067714235214401e-07, "loss": 24.1696, "step": 77640 }, { "epoch": 0.95, "learning_rate": 1.061917017287406e-07, "loss": 23.9215, "step": 77650 }, { "epoch": 0.96, "learning_rate": 1.0561354963916859e-07, "loss": 23.6259, "step": 77660 }, { "epoch": 0.96, "learning_rate": 1.050369673444529e-07, "loss": 23.5997, "step": 77670 }, { "epoch": 0.96, "learning_rate": 1.0446195493606703e-07, "loss": 24.0038, "step": 77680 }, { "epoch": 0.96, "learning_rate": 1.0388851250524467e-07, "loss": 23.9991, "step": 77690 }, { "epoch": 0.96, "learning_rate": 1.0331664014296195e-07, "loss": 23.8384, "step": 77700 }, { "epoch": 0.96, "learning_rate": 1.027463379399496e-07, "loss": 23.4137, "step": 77710 }, { "epoch": 0.96, "learning_rate": 1.021776059866908e-07, "loss": 24.3387, "step": 77720 }, { "epoch": 0.96, "learning_rate": 1.0161044437341672e-07, "loss": 24.0668, "step": 77730 }, { "epoch": 0.96, "learning_rate": 1.0104485319010982e-07, "loss": 22.9153, "step": 77740 }, { "epoch": 0.96, "learning_rate": 1.0048083252650498e-07, "loss": 23.5505, "step": 77750 }, { "epoch": 0.96, "learning_rate": 9.991838247208618e-08, "loss": 23.7673, "step": 77760 }, { "epoch": 0.96, "learning_rate": 9.935750311608983e-08, "loss": 23.4501, "step": 77770 }, { "epoch": 0.96, "learning_rate": 9.879819454750251e-08, "loss": 24.3419, "step": 77780 }, { "epoch": 0.96, "learning_rate": 9.824045685505878e-08, "loss": 23.9484, "step": 77790 }, { "epoch": 0.96, "learning_rate": 9.768429012725011e-08, "loss": 23.9123, "step": 77800 }, { "epoch": 0.96, "learning_rate": 9.712969445231257e-08, "loss": 24.0503, "step": 77810 }, { "epoch": 0.96, "learning_rate": 9.657666991823689e-08, "loss": 23.5896, "step": 77820 }, { "epoch": 0.96, "learning_rate": 9.602521661276398e-08, "loss": 23.4454, "step": 77830 }, { "epoch": 0.96, "learning_rate": 9.547533462338276e-08, "loss": 24.0935, "step": 77840 }, { "epoch": 0.96, "learning_rate": 9.492702403733567e-08, "loss": 23.6393, "step": 77850 }, { "epoch": 0.96, "learning_rate": 9.438028494161533e-08, "loss": 24.4254, "step": 77860 }, { "epoch": 0.96, "learning_rate": 9.383511742296459e-08, "loss": 24.1869, "step": 77870 }, { "epoch": 0.96, "learning_rate": 9.329152156787535e-08, "loss": 23.6197, "step": 77880 }, { "epoch": 0.96, "learning_rate": 9.274949746259531e-08, "loss": 23.6064, "step": 77890 }, { "epoch": 0.96, "learning_rate": 9.22090451931168e-08, "loss": 24.2253, "step": 77900 }, { "epoch": 0.96, "learning_rate": 9.167016484518565e-08, "loss": 23.5505, "step": 77910 }, { "epoch": 0.96, "learning_rate": 9.113285650429793e-08, "loss": 23.6641, "step": 77920 }, { "epoch": 0.96, "learning_rate": 9.059712025570211e-08, "loss": 24.0199, "step": 77930 }, { "epoch": 0.96, "learning_rate": 9.006295618439242e-08, "loss": 23.2968, "step": 77940 }, { "epoch": 0.96, "learning_rate": 8.953036437511997e-08, "loss": 24.3913, "step": 77950 }, { "epoch": 0.96, "learning_rate": 8.899934491238049e-08, "loss": 23.9661, "step": 77960 }, { "epoch": 0.96, "learning_rate": 8.846989788042326e-08, "loss": 23.7476, "step": 77970 }, { "epoch": 0.96, "learning_rate": 8.794202336324997e-08, "loss": 24.1633, "step": 77980 }, { "epoch": 0.96, "learning_rate": 8.74157214446092e-08, "loss": 23.4054, "step": 77990 }, { "epoch": 0.96, "learning_rate": 8.689099220799968e-08, "loss": 24.0243, "step": 78000 }, { "epoch": 0.96, "learning_rate": 8.636783573667596e-08, "loss": 24.1206, "step": 78010 }, { "epoch": 0.96, "learning_rate": 8.584625211363717e-08, "loss": 23.9901, "step": 78020 }, { "epoch": 0.96, "learning_rate": 8.532624142163493e-08, "loss": 24.024, "step": 78030 }, { "epoch": 0.96, "learning_rate": 8.480780374317321e-08, "loss": 23.6075, "step": 78040 }, { "epoch": 0.96, "learning_rate": 8.42909391605018e-08, "loss": 23.7735, "step": 78050 }, { "epoch": 0.96, "learning_rate": 8.37756477556284e-08, "loss": 23.3991, "step": 78060 }, { "epoch": 0.96, "learning_rate": 8.326192961030321e-08, "loss": 23.9278, "step": 78070 }, { "epoch": 0.96, "learning_rate": 8.274978480602992e-08, "loss": 23.6359, "step": 78080 }, { "epoch": 0.96, "learning_rate": 8.223921342406571e-08, "loss": 23.6428, "step": 78090 }, { "epoch": 0.96, "learning_rate": 8.173021554541361e-08, "loss": 24.2642, "step": 78100 }, { "epoch": 0.96, "learning_rate": 8.122279125082788e-08, "loss": 23.8994, "step": 78110 }, { "epoch": 0.96, "learning_rate": 8.071694062081636e-08, "loss": 23.8241, "step": 78120 }, { "epoch": 0.96, "learning_rate": 8.021266373563374e-08, "loss": 23.628, "step": 78130 }, { "epoch": 0.96, "learning_rate": 7.97099606752838e-08, "loss": 23.8591, "step": 78140 }, { "epoch": 0.96, "learning_rate": 7.920883151952718e-08, "loss": 23.8181, "step": 78150 }, { "epoch": 0.96, "learning_rate": 7.870927634786807e-08, "loss": 23.7144, "step": 78160 }, { "epoch": 0.96, "learning_rate": 7.821129523956194e-08, "loss": 24.1361, "step": 78170 }, { "epoch": 0.96, "learning_rate": 7.771488827362007e-08, "loss": 24.1306, "step": 78180 }, { "epoch": 0.96, "learning_rate": 7.722005552879719e-08, "loss": 23.9921, "step": 78190 }, { "epoch": 0.96, "learning_rate": 7.672679708360054e-08, "loss": 24.2577, "step": 78200 }, { "epoch": 0.96, "learning_rate": 7.623511301628972e-08, "loss": 23.826, "step": 78210 }, { "epoch": 0.96, "learning_rate": 7.574500340487346e-08, "loss": 23.8936, "step": 78220 }, { "epoch": 0.96, "learning_rate": 7.525646832710732e-08, "loss": 24.106, "step": 78230 }, { "epoch": 0.96, "learning_rate": 7.476950786050264e-08, "loss": 23.6948, "step": 78240 }, { "epoch": 0.96, "learning_rate": 7.42841220823165e-08, "loss": 24.3564, "step": 78250 }, { "epoch": 0.96, "learning_rate": 7.380031106955954e-08, "loss": 23.7874, "step": 78260 }, { "epoch": 0.96, "learning_rate": 7.331807489898923e-08, "loss": 23.3795, "step": 78270 }, { "epoch": 0.96, "learning_rate": 7.283741364711439e-08, "loss": 23.6091, "step": 78280 }, { "epoch": 0.96, "learning_rate": 7.235832739019732e-08, "loss": 23.6787, "step": 78290 }, { "epoch": 0.96, "learning_rate": 7.188081620424392e-08, "loss": 24.1443, "step": 78300 }, { "epoch": 0.96, "learning_rate": 7.14048801650158e-08, "loss": 23.8992, "step": 78310 }, { "epoch": 0.96, "learning_rate": 7.093051934802142e-08, "loss": 23.5726, "step": 78320 }, { "epoch": 0.96, "learning_rate": 7.045773382852283e-08, "loss": 23.7287, "step": 78330 }, { "epoch": 0.96, "learning_rate": 6.99865236815267e-08, "loss": 23.8531, "step": 78340 }, { "epoch": 0.96, "learning_rate": 6.951688898179543e-08, "loss": 23.3171, "step": 78350 }, { "epoch": 0.96, "learning_rate": 6.904882980383832e-08, "loss": 23.8386, "step": 78360 }, { "epoch": 0.96, "learning_rate": 6.858234622191485e-08, "loss": 23.5135, "step": 78370 }, { "epoch": 0.96, "learning_rate": 6.81174383100358e-08, "loss": 23.4799, "step": 78380 }, { "epoch": 0.96, "learning_rate": 6.765410614196e-08, "loss": 23.9319, "step": 78390 }, { "epoch": 0.96, "learning_rate": 6.719234979119859e-08, "loss": 23.4589, "step": 78400 }, { "epoch": 0.96, "learning_rate": 6.673216933101078e-08, "loss": 24.8889, "step": 78410 }, { "epoch": 0.96, "learning_rate": 6.627356483440706e-08, "loss": 23.7488, "step": 78420 }, { "epoch": 0.96, "learning_rate": 6.581653637414698e-08, "loss": 23.7729, "step": 78430 }, { "epoch": 0.96, "learning_rate": 6.536108402274033e-08, "loss": 23.855, "step": 78440 }, { "epoch": 0.96, "learning_rate": 6.49072078524482e-08, "loss": 23.5749, "step": 78450 }, { "epoch": 0.96, "learning_rate": 6.445490793527853e-08, "loss": 23.693, "step": 78460 }, { "epoch": 0.97, "learning_rate": 6.400418434299282e-08, "loss": 23.7314, "step": 78470 }, { "epoch": 0.97, "learning_rate": 6.355503714709943e-08, "loss": 23.504, "step": 78480 }, { "epoch": 0.97, "learning_rate": 6.310746641885801e-08, "loss": 24.4968, "step": 78490 }, { "epoch": 0.97, "learning_rate": 6.266147222927842e-08, "loss": 24.2917, "step": 78500 }, { "epoch": 0.97, "learning_rate": 6.221705464911854e-08, "loss": 24.6928, "step": 78510 }, { "epoch": 0.97, "learning_rate": 6.17742137488897e-08, "loss": 23.9618, "step": 78520 }, { "epoch": 0.97, "learning_rate": 6.133294959884906e-08, "loss": 24.2361, "step": 78530 }, { "epoch": 0.97, "learning_rate": 6.089326226900616e-08, "loss": 23.516, "step": 78540 }, { "epoch": 0.97, "learning_rate": 6.045515182911966e-08, "loss": 23.7335, "step": 78550 }, { "epoch": 0.97, "learning_rate": 6.001861834869838e-08, "loss": 23.8025, "step": 78560 }, { "epoch": 0.97, "learning_rate": 5.9583661896999157e-08, "loss": 23.5471, "step": 78570 }, { "epoch": 0.97, "learning_rate": 5.9150282543031235e-08, "loss": 23.5841, "step": 78580 }, { "epoch": 0.97, "learning_rate": 5.871848035555294e-08, "loss": 24.4589, "step": 78590 }, { "epoch": 0.97, "learning_rate": 5.828825540307059e-08, "loss": 24.319, "step": 78600 }, { "epoch": 0.97, "learning_rate": 5.785960775384181e-08, "loss": 23.8671, "step": 78610 }, { "epoch": 0.97, "learning_rate": 5.74325374758744e-08, "loss": 24.0395, "step": 78620 }, { "epoch": 0.97, "learning_rate": 5.70070446369253e-08, "loss": 23.6336, "step": 78630 }, { "epoch": 0.97, "learning_rate": 5.658312930450049e-08, "loss": 24.0905, "step": 78640 }, { "epoch": 0.97, "learning_rate": 5.616079154585619e-08, "loss": 24.6124, "step": 78650 }, { "epoch": 0.97, "learning_rate": 5.5740031427998775e-08, "loss": 24.3876, "step": 78660 }, { "epoch": 0.97, "learning_rate": 5.532084901768486e-08, "loss": 23.992, "step": 78670 }, { "epoch": 0.97, "learning_rate": 5.490324438141792e-08, "loss": 23.6116, "step": 78680 }, { "epoch": 0.97, "learning_rate": 5.448721758545494e-08, "loss": 23.7366, "step": 78690 }, { "epoch": 0.97, "learning_rate": 5.40727686957998e-08, "loss": 23.7749, "step": 78700 }, { "epoch": 0.97, "learning_rate": 5.365989777820657e-08, "loss": 24.0387, "step": 78710 }, { "epoch": 0.97, "learning_rate": 5.3248604898180624e-08, "loss": 23.705, "step": 78720 }, { "epoch": 0.97, "learning_rate": 5.2838890120974204e-08, "loss": 23.8956, "step": 78730 }, { "epoch": 0.97, "learning_rate": 5.243075351159088e-08, "loss": 24.1669, "step": 78740 }, { "epoch": 0.97, "learning_rate": 5.2024195134784405e-08, "loss": 23.5067, "step": 78750 }, { "epoch": 0.97, "learning_rate": 5.161921505505651e-08, "loss": 23.6652, "step": 78760 }, { "epoch": 0.97, "learning_rate": 5.1215813336660256e-08, "loss": 23.2063, "step": 78770 }, { "epoch": 0.97, "learning_rate": 5.081399004359777e-08, "loss": 23.6797, "step": 78780 }, { "epoch": 0.97, "learning_rate": 5.041374523961806e-08, "loss": 22.9744, "step": 78790 }, { "epoch": 0.97, "learning_rate": 5.0015078988223665e-08, "loss": 24.2431, "step": 78800 }, { "epoch": 0.97, "learning_rate": 4.9617991352665094e-08, "loss": 24.0878, "step": 78810 }, { "epoch": 0.97, "learning_rate": 4.922248239594196e-08, "loss": 23.9809, "step": 78820 }, { "epoch": 0.97, "learning_rate": 4.882855218080407e-08, "loss": 23.7029, "step": 78830 }, { "epoch": 0.97, "learning_rate": 4.8436200769750305e-08, "loss": 24.131, "step": 78840 }, { "epoch": 0.97, "learning_rate": 4.8045428225028664e-08, "loss": 23.6511, "step": 78850 }, { "epoch": 0.97, "learning_rate": 4.765623460863733e-08, "loss": 23.4237, "step": 78860 }, { "epoch": 0.97, "learning_rate": 4.726861998232579e-08, "loss": 23.4745, "step": 78870 }, { "epoch": 0.97, "learning_rate": 4.6882584407588194e-08, "loss": 23.4795, "step": 78880 }, { "epoch": 0.97, "learning_rate": 4.649812794567221e-08, "loss": 23.5658, "step": 78890 }, { "epoch": 0.97, "learning_rate": 4.6115250657574606e-08, "loss": 23.3882, "step": 78900 }, { "epoch": 0.97, "learning_rate": 4.573395260404012e-08, "loss": 23.9232, "step": 78910 }, { "epoch": 0.97, "learning_rate": 4.53542338455637e-08, "loss": 23.8254, "step": 78920 }, { "epoch": 0.97, "learning_rate": 4.4976094442389376e-08, "loss": 23.5463, "step": 78930 }, { "epoch": 0.97, "learning_rate": 4.4599534454511374e-08, "loss": 23.1659, "step": 78940 }, { "epoch": 0.97, "learning_rate": 4.422455394167302e-08, "loss": 24.3378, "step": 78950 }, { "epoch": 0.97, "learning_rate": 4.385115296336562e-08, "loss": 23.9148, "step": 78960 }, { "epoch": 0.97, "learning_rate": 4.3479331578831775e-08, "loss": 23.6381, "step": 78970 }, { "epoch": 0.97, "learning_rate": 4.31090898470643e-08, "loss": 23.5016, "step": 78980 }, { "epoch": 0.97, "learning_rate": 4.274042782680177e-08, "loss": 23.5501, "step": 78990 }, { "epoch": 0.97, "learning_rate": 4.237334557653627e-08, "loss": 23.7266, "step": 79000 }, { "epoch": 0.97, "learning_rate": 4.200784315450568e-08, "loss": 24.5764, "step": 79010 }, { "epoch": 0.97, "learning_rate": 4.164392061869915e-08, "loss": 23.9248, "step": 79020 }, { "epoch": 0.97, "learning_rate": 4.128157802685495e-08, "loss": 24.2592, "step": 79030 }, { "epoch": 0.97, "learning_rate": 4.0920815436461536e-08, "loss": 23.7548, "step": 79040 }, { "epoch": 0.97, "learning_rate": 4.056163290475423e-08, "loss": 24.0163, "step": 79050 }, { "epoch": 0.97, "learning_rate": 4.020403048872079e-08, "loss": 23.9363, "step": 79060 }, { "epoch": 0.97, "learning_rate": 3.984800824509472e-08, "loss": 23.8791, "step": 79070 }, { "epoch": 0.97, "learning_rate": 3.9493566230361936e-08, "loss": 24.0055, "step": 79080 }, { "epoch": 0.97, "learning_rate": 3.9140704500756354e-08, "loss": 23.8249, "step": 79090 }, { "epoch": 0.97, "learning_rate": 3.8789423112262084e-08, "loss": 23.9033, "step": 79100 }, { "epoch": 0.97, "learning_rate": 3.843972212061009e-08, "loss": 23.2766, "step": 79110 }, { "epoch": 0.97, "learning_rate": 3.809160158128267e-08, "loss": 23.7752, "step": 79120 }, { "epoch": 0.97, "learning_rate": 3.7745061549512295e-08, "loss": 23.8038, "step": 79130 }, { "epoch": 0.97, "learning_rate": 3.740010208027722e-08, "loss": 23.0058, "step": 79140 }, { "epoch": 0.97, "learning_rate": 3.705672322830811e-08, "loss": 23.8151, "step": 79150 }, { "epoch": 0.97, "learning_rate": 3.671492504808472e-08, "loss": 23.1617, "step": 79160 }, { "epoch": 0.97, "learning_rate": 3.637470759383366e-08, "loss": 24.5344, "step": 79170 }, { "epoch": 0.97, "learning_rate": 3.6036070919532875e-08, "loss": 23.702, "step": 79180 }, { "epoch": 0.97, "learning_rate": 3.569901507890716e-08, "loss": 24.5766, "step": 79190 }, { "epoch": 0.97, "learning_rate": 3.5363540125434856e-08, "loss": 24.3411, "step": 79200 }, { "epoch": 0.97, "learning_rate": 3.5029646112338945e-08, "loss": 24.1188, "step": 79210 }, { "epoch": 0.97, "learning_rate": 3.469733309259371e-08, "loss": 23.7696, "step": 79220 }, { "epoch": 0.97, "learning_rate": 3.436660111892254e-08, "loss": 23.9176, "step": 79230 }, { "epoch": 0.97, "learning_rate": 3.40374502437979e-08, "loss": 23.7469, "step": 79240 }, { "epoch": 0.97, "learning_rate": 3.370988051944135e-08, "loss": 23.7335, "step": 79250 }, { "epoch": 0.97, "learning_rate": 3.338389199782244e-08, "loss": 23.9957, "step": 79260 }, { "epoch": 0.97, "learning_rate": 3.3059484730662e-08, "loss": 22.9269, "step": 79270 }, { "epoch": 0.98, "learning_rate": 3.273665876942889e-08, "loss": 23.8827, "step": 79280 }, { "epoch": 0.98, "learning_rate": 3.241541416533989e-08, "loss": 23.5234, "step": 79290 }, { "epoch": 0.98, "learning_rate": 3.2095750969364236e-08, "loss": 24.0811, "step": 79300 }, { "epoch": 0.98, "learning_rate": 3.1777669232215815e-08, "loss": 23.3104, "step": 79310 }, { "epoch": 0.98, "learning_rate": 3.146116900436091e-08, "loss": 23.126, "step": 79320 }, { "epoch": 0.98, "learning_rate": 3.1146250336014926e-08, "loss": 23.99, "step": 79330 }, { "epoch": 0.98, "learning_rate": 3.0832913277139e-08, "loss": 23.7022, "step": 79340 }, { "epoch": 0.98, "learning_rate": 3.05211578774478e-08, "loss": 23.4691, "step": 79350 }, { "epoch": 0.98, "learning_rate": 3.021098418640178e-08, "loss": 23.7493, "step": 79360 }, { "epoch": 0.98, "learning_rate": 2.9902392253212676e-08, "loss": 24.1757, "step": 79370 }, { "epoch": 0.98, "learning_rate": 2.95953821268391e-08, "loss": 24.1075, "step": 79380 }, { "epoch": 0.98, "learning_rate": 2.9289953855989872e-08, "loss": 23.5501, "step": 79390 }, { "epoch": 0.98, "learning_rate": 2.8986107489122894e-08, "loss": 24.1369, "step": 79400 }, { "epoch": 0.98, "learning_rate": 2.8683843074444052e-08, "loss": 23.5559, "step": 79410 }, { "epoch": 0.98, "learning_rate": 2.8383160659911645e-08, "loss": 23.9156, "step": 79420 }, { "epoch": 0.98, "learning_rate": 2.808406029322752e-08, "loss": 24.0891, "step": 79430 }, { "epoch": 0.98, "learning_rate": 2.778654202184816e-08, "loss": 24.2354, "step": 79440 }, { "epoch": 0.98, "learning_rate": 2.7490605892973587e-08, "loss": 23.989, "step": 79450 }, { "epoch": 0.98, "learning_rate": 2.7196251953558462e-08, "loss": 23.5618, "step": 79460 }, { "epoch": 0.98, "learning_rate": 2.6903480250300982e-08, "loss": 23.5508, "step": 79470 }, { "epoch": 0.98, "learning_rate": 2.6612290829652887e-08, "loss": 24.0324, "step": 79480 }, { "epoch": 0.98, "learning_rate": 2.632268373781166e-08, "loss": 24.2663, "step": 79490 }, { "epoch": 0.98, "learning_rate": 2.6034659020724995e-08, "loss": 23.841, "step": 79500 }, { "epoch": 0.98, "learning_rate": 2.574821672408967e-08, "loss": 23.5646, "step": 79510 }, { "epoch": 0.98, "learning_rate": 2.5463356893351553e-08, "loss": 24.2139, "step": 79520 }, { "epoch": 0.98, "learning_rate": 2.5180079573704497e-08, "loss": 23.9244, "step": 79530 }, { "epoch": 0.98, "learning_rate": 2.4898384810092546e-08, "loss": 23.8649, "step": 79540 }, { "epoch": 0.98, "learning_rate": 2.461827264720773e-08, "loss": 24.293, "step": 79550 }, { "epoch": 0.98, "learning_rate": 2.4339743129491167e-08, "loss": 23.2673, "step": 79560 }, { "epoch": 0.98, "learning_rate": 2.4062796301133063e-08, "loss": 23.5893, "step": 79570 }, { "epoch": 0.98, "learning_rate": 2.3787432206072712e-08, "loss": 24.0108, "step": 79580 }, { "epoch": 0.98, "learning_rate": 2.35136508879974e-08, "loss": 23.8978, "step": 79590 }, { "epoch": 0.98, "learning_rate": 2.3241452390344587e-08, "loss": 24.0018, "step": 79600 }, { "epoch": 0.98, "learning_rate": 2.2970836756299742e-08, "loss": 23.5261, "step": 79610 }, { "epoch": 0.98, "learning_rate": 2.2701804028796292e-08, "loss": 24.2959, "step": 79620 }, { "epoch": 0.98, "learning_rate": 2.2434354250520097e-08, "loss": 23.2703, "step": 79630 }, { "epoch": 0.98, "learning_rate": 2.2168487463901655e-08, "loss": 23.5791, "step": 79640 }, { "epoch": 0.98, "learning_rate": 2.1904203711121674e-08, "loss": 24.0114, "step": 79650 }, { "epoch": 0.98, "learning_rate": 2.1641503034111056e-08, "loss": 23.7477, "step": 79660 }, { "epoch": 0.98, "learning_rate": 2.1380385474549792e-08, "loss": 24.4798, "step": 79670 }, { "epoch": 0.98, "learning_rate": 2.1120851073862524e-08, "loss": 23.6319, "step": 79680 }, { "epoch": 0.98, "learning_rate": 2.0862899873228537e-08, "loss": 23.5648, "step": 79690 }, { "epoch": 0.98, "learning_rate": 2.0606531913570648e-08, "loss": 24.4312, "step": 79700 }, { "epoch": 0.98, "learning_rate": 2.035174723556521e-08, "loss": 24.333, "step": 79710 }, { "epoch": 0.98, "learning_rate": 2.0098545879633224e-08, "loss": 24.579, "step": 79720 }, { "epoch": 0.98, "learning_rate": 1.9846927885948108e-08, "loss": 23.9197, "step": 79730 }, { "epoch": 0.98, "learning_rate": 1.9596893294429042e-08, "loss": 23.9745, "step": 79740 }, { "epoch": 0.98, "learning_rate": 1.9348442144746516e-08, "loss": 23.7268, "step": 79750 }, { "epoch": 0.98, "learning_rate": 1.9101574476316776e-08, "loss": 23.691, "step": 79760 }, { "epoch": 0.98, "learning_rate": 1.885629032830849e-08, "loss": 23.3914, "step": 79770 }, { "epoch": 0.98, "learning_rate": 1.861258973963609e-08, "loss": 24.2582, "step": 79780 }, { "epoch": 0.98, "learning_rate": 1.8370472748964197e-08, "loss": 23.9658, "step": 79790 }, { "epoch": 0.98, "learning_rate": 1.812993939470542e-08, "loss": 24.133, "step": 79800 }, { "epoch": 0.98, "learning_rate": 1.7890989715022566e-08, "loss": 24.0574, "step": 79810 }, { "epoch": 0.98, "learning_rate": 1.765362374782642e-08, "loss": 23.5374, "step": 79820 }, { "epoch": 0.98, "learning_rate": 1.7417841530774638e-08, "loss": 23.6882, "step": 79830 }, { "epoch": 0.98, "learning_rate": 1.7183643101277293e-08, "loss": 23.7561, "step": 79840 }, { "epoch": 0.98, "learning_rate": 1.695102849648911e-08, "loss": 24.132, "step": 79850 }, { "epoch": 0.98, "learning_rate": 1.671999775331723e-08, "loss": 23.6946, "step": 79860 }, { "epoch": 0.98, "learning_rate": 1.649055090841567e-08, "loss": 23.7915, "step": 79870 }, { "epoch": 0.98, "learning_rate": 1.6262687998185313e-08, "loss": 23.5585, "step": 79880 }, { "epoch": 0.98, "learning_rate": 1.6036409058780566e-08, "loss": 23.8152, "step": 79890 }, { "epoch": 0.98, "learning_rate": 1.5811714126100496e-08, "loss": 23.5021, "step": 79900 }, { "epoch": 0.98, "learning_rate": 1.5588603235794362e-08, "loss": 23.9131, "step": 79910 }, { "epoch": 0.98, "learning_rate": 1.5367076423259408e-08, "loss": 23.3981, "step": 79920 }, { "epoch": 0.98, "learning_rate": 1.514713372364196e-08, "loss": 23.5376, "step": 79930 }, { "epoch": 0.98, "learning_rate": 1.492877517183744e-08, "loss": 24.4572, "step": 79940 }, { "epoch": 0.98, "learning_rate": 1.4712000802490356e-08, "loss": 23.7117, "step": 79950 }, { "epoch": 0.98, "learning_rate": 1.4496810649990978e-08, "loss": 24.4231, "step": 79960 }, { "epoch": 0.98, "learning_rate": 1.4283204748481994e-08, "loss": 23.3788, "step": 79970 }, { "epoch": 0.98, "learning_rate": 1.4071183131851851e-08, "loss": 23.5974, "step": 79980 }, { "epoch": 0.98, "learning_rate": 1.3860745833740308e-08, "loss": 23.8628, "step": 79990 }, { "epoch": 0.98, "learning_rate": 1.365189288753399e-08, "loss": 24.2237, "step": 80000 } ], "logging_steps": 10, "max_steps": 81312, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 4000, "total_flos": 0.0, "train_batch_size": 1, "trial_name": null, "trial_params": null }